Intermediate Algebra

Richard Rusczyk
Mathew Crawford
Art of Problem Solving
© 2008, AoPS Incorporated. All Rights Reserved.

Reproduction of any portion of this book without the written permission of AoPS Incorporated is strictly
prohibited, except for "fair use" or other noncommercial uses as defined in Sections 107 and 108 of the
U.S. Copyright Act.

Published by: AoPS Incorporated

P.O. Box 2185
Alpine, CA 91903-2185
(619) 659-1612
[email protected]

ISBN-13: 978-1-934124-04-8

Visit the Art of Problem Solving website at http: ∕∕www. artofproblemsolving. com

Cover image designed by Vanessa Rusczyk using KaleidoTile software.

Printed in the United States of America.

First Printing 2008.
 HOW TO USE THIS BOOK

Organization

Y°u can think of this book as divided into three sections:

• Chapters 1 through 4. If you completed Art of Problem Solving's Introduction to Algebra, then much
of the first four chapters will be review. Don't just skip these four chapters; there is some new
material, and some of the review material is extended to more challenging problems. If you
have not completed Introduction to Algebra, then you should have experience factoring quadratic
expressions before starting this book. If you have completed a typical Algebra 1 course in the
United States, you may be ready for this text, but the first four chapters might not be review. Study
them closely, and if they give you great difficulty, you may wish to study those topics in greater
depth in Introduction to Algebra.

• Chapters 5 through 16. These chapters form the core of the book. Chapter 5 is the least important of
the first 16 chapters of the book; much of the material can be read for a qualitative understanding,
rather than for detailed study. Section 5.3 is the most important section of this chapter; the
remaining sections can be read lightly. All students who use this book should study chapters 6
through 16 closely.

• Chapters 17 through 20. These chapters are primarily intended for those students preparing for
mathematics competitions, or for those who want to pursue more advanced study in the subjects
of these chapters.

Learn by Solving Problems

We believe that the best way to learn mathematics is by solving problems. Lots and lots of problems.
In fact, we believe that the best way to learn mathematics is to try to solve problems that you don't
know how to do. When you discover something on your own, you'll understand it much better than if
someone had just told you.
Most of the sections of this book begin with several problems. The solutions to these problems will
be covered in the text, but try to solve the problems before reading the section. If you can't solve some
of the problems, that's OK, because they will all be fully solved as you read the section. Even if you can
solve all of the problems, it's still important to read the section, both to make sure that your solutions
are correct, and also because you may find that the book's solutions are simpler or easier to understand
than your own.

iii
HOW TO USE THIS BOOK

Explanation of Icons

Throughout the book, you will see various shaded boxes and icons.

Concept: This will be a general problem-solving technique or strategy. These are the
Q—l3⅛ "keys" to becoming a better problem solver!

Important: This will be something important that you should learn. It might be a

ψ formula, a solution technique, or a caution.

WARNING!! Beware if you see this box! This will point out a common mistake or
V pitfall.

Sidenote: This box will contain material that, although interesting, is not part of the
J> main material of the text. It's OK to skip over these boxes, but if you read
them, you might learn something interesting!

Bogus Solution: Just like the impossible cube shown to the left, there's something
wrong with any "solution" that appears in this box.

Proof: Some proofs in the text will be presented in boxes like this one. These are intended to be
0 examples that the students can use as guides for how to write their own proofs.

Exercises, Review Problems, and Challenge Problems

Most sections end with several Exercises. These will test your understanding of the material that was
covered in the section. You should try to solve all of the exercises. Exercises marked with a ★ are more
difficult.
Each chapter concludes with many Review Problems and Challenge Problems. The Review Prob­
lems test your basic understanding of the material covered in the chapter. Your goal should be to solve
most or all of the Review Problems for every chapter—if you're unable to do this, it means that you
haven't yet mastered the material, and you should probably go back and read the chapter again.
The Challenge Problems are generally more difficult than the other problems in the book, and will
really test your mastery of the material. Some of them are very hard—the hardest ones are marked with
a ★. Don't expect to be able to solve all of the Challenge Problems on your first try—these are difficult
problems even for experienced problem solvers. If you are able to solve a large number of Challenge
Problems, then congratulations, you are on your way to becoming an expert problem solver!

iv
 HOW TO USE THIS BOOK

Hints
Many problems come with one or more hints. You can look up the hints in the Hints section in the
back of the book. 1 he hints are numbered in random order, so that when you're looking up a hint to a
problem you don’t accidentally glance at the hint to the next problem at the same time.
It is very important that you first try to solve each problem without resorting to the hints. Only after
you ve seriously thought about a problem and are stuck should you seek a hint. Also, for problems
which have multiple hints, use the hints one at a time; don't go to the second hint until you've thought
about the first one.

Solutions
The solutions to all of the Exercises, Review Problems, and Challenge Problems are in the separate
solutions book. If you are using this textbook in a regular school class, then your teacher may decide not
to make this solutions book available to you, and instead present the solutions him/herself. However, if
you are using this book to learn on your own, then you probably have a copy of the solutions book, in
which case there are some very important things to keep in mind:

1. Make a serious attempt to solve each problem before looking at the solution. Don't use the
solutions book as a crutch to avoid really thinking about the problem first. You should think hard
about a problem before deciding to look at the solution. On the other hard, after you've thought
hard about a problem, don't feel bad about looking at the solution if you're really stuck.

2. After you solve a problem, it's usually a good idea to read the solution, even if you think you
know how to solve the problem. The solutions book might show you a quicker or more concise
way to solve the problem, or it might have a completely different solution method than yours.

3. If you have to look at the solution in order to solve a problem, make a note of that problem. Come
back to it in a week or two to make sure that you are able to solve it on your own, without resorting
to the solution.

Resources
Here are some other good resources for you to further pursue your study of mathematics:

• The Art of Problem Solving Intermediate Series of texts. This series of texts will form a full
curriculum for high-performing students in grades 8-12. In addition to this Intermediate Algebra
text, the text Intermediate Counting & Probability is also available. More titles will be added to this
series in 2009-2011.

• The Art of Problem Solving books, by Sandor Lehoczky and Richard Rusczyk. Whereas the book
that you're reading right now will go into great detail of one specific subject area—algebra—the
Art of Problem Solving books cover a wide range of problem solving topics across many different
areas of mathematics.

v
HOW TO USE THIS BOOK

• The www. artofproblemsolving. com website. The publishers of this book are also the webmasters
of the Art of Problem Solving website, which contains many resources for students:
- a discussion forum
- online classes
- resource lists of books, contests, and other websites
- a IΔTgX tutorial
- a math and problem solving Wiki
- and much more!
• You can hone your problem solving skills (and perhaps win prizes!) by participating in various
math contests. For middle school students in the United States, the major contests are MOEMS,
MATHCOUNTS, and the AMC 8. For U.S. high school students, some of the best-known contests
are the AMC/AIME/USAMO series of contests (which are used to choose the U.S. team for the
International Mathematical Olympiad), the American Regions Math League (ARML), the Mandel­
brot Competition, the Harvard-MIT Mathematics Tournament, and the USA Mathematical Talent
Search. More details about these contests are on page vii, and links to these and many other
contests are available on the Art of Problem Solving website.

A Note to Teachers
We believe students learn best when they are challenged with hard problems that at first they may not
know how to do. This is the motivating philosophy behind this book.
Rather than first introducing new material and then giving students exercises, we present problems
at the start of each section that students should try to solve before new material is presented. The goal
is to get students to discover the new material on their own. Often, complicated problems are broken
into smaller parts, so that students can discover new techniques one piece at a time. After the problems,
new material is formally presented in the text, and full solutions to each problem are explained, along
with problem-solving strategies.
We hope that teachers will find that their stronger students will discover most of the material in this
book on their own by working through the problems. Other students may learn better from a more
traditional approach of first seeing the new material, then working the problems. Teachers have the
flexibility to use either approach when teaching from this book.
Sections marked with a ★ contain supplementary material that may be safely skipped. In general,
chapters are not equal in length, so different chapters may take different amounts of classroom time.

Links
The Art of Problem Solving website has a page containing links to websites with content relating to
material in this book, as well as an errata list for the book. This page can be found at:
http:∕∕www.artofproblemsolving.com/BookLinks/IntermAlgebra/links.php

Extra! Occasionally, you'll see a box like this at the bottom of a page. This is an "Extra!" and
ni» 11» hi» u» might be a quote, some biographical or historical background, or perhaps an interesting
idea to think about.

vi
 ACKNOWLEDGEMENTS

I
Acknowledgements

Contests
We would like to thank the following contests for allowing us to use a wide selection of their problems
in this book:

• The American Mathematics Competitions, a series of contests for U.S. middle and high school
students. The AMC 8, AMC 10, and AMC 12 contests are multiple-choice tests that are taken by
over 400,000 students every year. Top scorers on the AMC 10 and AMC 12 are invited to take the
American Invitational Mathematics Examination (AIME), which is a more difficult, short-answer
contest. Approximately 10,000 students every year participate in the AIME. Then, based on the
results of the AMC and AIME contests, about 500 students are invited to participate in the USA
Mathematical Olympiad (USAMO), a 2-day, 9-hour examination in which each student must
show all of his or her work. Results from the USAMO are used to invite a number of students to
the Math Olympiad Summer Program, at which the U.S. team for the International Mathematical
Olympiad (IMO) is chosen. More information about the AMC contests can be found on the AMC
website at www.unl.edu/amc. (Note: Problems marked AHSME are from the American High
School Math Exam, the predecessor of the AMC 12.)

• The Mandelbrot Competition, which was founded in 1990 by Sandor Lehoczky, Richard Rusczyk,
and Sam Vandervelde. The aim of the Mandelbrot Competition is to provide a challenging,
engaging mathematical experience that is both competitive and educational. Students compete
both as individuals and in teams. The Mandelbrot Competition is offered at the national level
for more advanced students and at the regional level for less experienced problem solvers. More
information can be found at www. mandelbrot. org.

• The USA Mathematical Talent Search (USAMTS), which was founded in 1989 by Professor
George Berzsenyi. The USAMTS is a free mathematics competition open to all United States
middle and high school students. As opposed to most mathematics competitions, the USAMTS
allows students a full month to work out their solutions. Carefully written justifications are
required for each problem. More information is available at www. usamts. org.

• The American Regions Math League (ARML), which was founded in 1976. The annual ARML
competition brings together nearly 2,000 of the nation's finest students. They meet, compete
against, and socialize with one another, forming friendships and sharpening their mathematical
skills. The contest is written for high school students, although some exceptional junior high
students attend each year. The competition consists of several events, which include a team
round, a power question (in which a team solves proof-oriented questions), an individual round,
and two relay rounds. More information is available at www. arml. com. Some questions from the
New York State Math League (NYSML) also appear in this text. NYSML is a New York statewide
competition that inspired the start of ARML.

I
vii
ACKNOWLEDGEMENTS

• The Harvard-MIT Mathematics Tournament, which is an annual math tournament for high school
students, held at MIT and at Harvard in alternating years. It is run exclusively by MIT and Harvard
students, most of whom themselves participated in math contests in high school. More information
is available at web .mit. edu∕hmmt∕.

Some other sources for problems in the book are listed below.

MATHCOUNTS, the premier national math contest for U.S. middle school students.

The William Lowell Putnam Competition, an annual math competition for undergraduate students in
North America.

Mathematical journals Mathematics & Informatics Quarterly (M&IQ), Mathematical Mayhem, and Crux
Mathematicorum.

Various Canadian national competitions, such as the Canadian Invitational Mathematics Contest
(CIMC), the Canadian Open Mathematics Challenge (COMC), the Canadian Mathematics Olympiad
(CMO), and contests run by the Centre for Education in Mathematics and Computing (CEMC).

The Duke Math Meet and the Michigan Mathematics Prize Competition (MMPC).

Problems from various countries' national math contests. These problems are cited by country name
in the text.

How We Wrote This Book

This book was written using the L¾X document processing system. We thank the authors of the various
LTpX packages that we used while preparing this book, and also the authors of The K⅛X Companion for
writing a reference book that is not only thorough but also very readable. The diagrams were prepared
using METAPOST and Asymptote.

About Us
Even more than our other texts, this book is a collaborative effort of the staff of the Art of Problem
Solving. Richard Rusczyk was the lead author of the text and primary editor of the solutions manual.
Mathew Crawford prepared a draft of some portions of the text, and thorough notes on other portions.
He also selected and organized many of the problems in the text, and wrote a draft of many of the
solutions. Naoki Sato was the lead author of the solutions manual, and selected and organized many
of the problems in the text. David Patrick read several drafts of the text and solutions and made many,
many helpful suggestions. The following people also made contributions through proofreading, creating
diagrams, suggesting extra material for the textbook, or writing problems for the text: Tom Belulovich,
Yakov Berchenko-Kogan, Tim Black, Greg Brockman, Chris Chang, Lisa Davis, Kenan Diab, Larry Evans,
Mike Kling, Joseph Laurendi, Daniel Li, Linda Liu, Sean Markan, Maria Monks, Jeff Nanney, Dimitar
Popov, Tianren Qi, Adrian Sanborn, Nathan Savir, Beth Schaffer, Arnav Tripathy, Philip Tynan, Valentin
Vomicu, Samson Zhou, and Olga Zverovich. Vanessa Rusczyk designed the cover.

viii
 CONTENTS

How to Use This Book in

•∙
Acknowledgements Vll

Basic Techniques for Solving Equations 1

1.1 Isolation.......................................................................................................................................... 1
1.2 Substitution................................................................................................................................... 6
1.3 Elimination................................................................................................................................... 9
1.4 Larger Systems of Linear Equations......................................................................................... 14
1.5 Summary...................................................................................................................................... 19

Functions Review 24
2.1 Function Basics............................................................................................................................. 24
2.2 Graphing Functions.................................................................................................................... 32
2.3 Composition................................................................................................................................. 39
2.4 Inverse Functions....................................................................................................................... 43
2.5 Summary....................................................................................................................................... 49

Complex Numbers 54
3.1 Arithmetic of Complex Numbers............................................................................................ 55
3.2 The Complex Plane.................................................................................................................... 61
3.3 Real and Imaginary Parts........................................................................................................... 65
3.4 Graphing in the Complex Plane............................................................................................... 70
3.5 Summary....................................................................................................................................... 74

ix
CONTENTS

Quadratics 79
4.1 Factoring Quadratics ................................................................................................................. 79
4.2 Relating Roots and Coefficients .............................................................................................. 84
4.3 Completing the Square............................................................................................................. 88
4.4 The Discriminant.......................................................................................................................... 96
4.5 Quadratic Inequalities................................................................................................................ 99
4.6 Summary......................................................................................................................................... 102

Conics 106
5.1 Parabolas.........................................................................................................................................107
5.2 Problem Solving With Parabolas................................................................................................. 118
5.3 Maxima and Minimaof Quadratics............................................................................................ 121
5.4 Circles............................................................................................................................................... 126
5.5 Ellipses............................................................................................................................................129
5.6 Hyperbolas......................................................................................................................................142
5.7 Summary......................................................................................................................................... 156

Polynomial Division 162

6.1 Polynomial Review.......................................................................................................................163
6.2 Introduction to Polynomial Division .........................................................................................166
6.3 Synthetic Division.......................................................................................................................... 177
6.4 The Remainder Theorem.............................................................................................................. 185
6.5 Summary......................................................................................................................................... 189

Polynomial Roots Part I 193

7.1 The Factor Theorem....................................................................................................................... 193
7.2 Integer Roots................................................................................................................................... 198
7.3 Rational Roots................................................................................................................................ 206
7.4 Bounds............................................................................................................................................ 212
7.5 Graphing and the Fundamental Theorem of Algebra.............................................................215
7.6 Algebraic Applications of the Fundamental Theorem.............................................................219
7.7 Summary......................................................................................................................................... 225

x
 CONTENTS

Polynomial Roots Part II 230

8.1 Irrational Roots.................................................................................................................................. 230
8.2 Nonreal Roots .................................................................................................................................. 239
8.3 Vieta's Formulas............................................................................................................................... 244
8.4 Using Roots to Make Equations................................................................................................... 253
8.5 Summary........................................................................................................................................... 256

Factoring Multivariable Polynomials 261

9.1 Grouping........................................................................................................................................... 261
9.2 Sums and Differences of Powers................................................................................................... 268
9.3* The Factor Theoremfor Multivariable Polynomials................................................................... 277
9.4 Summary............................................................................................................................................ 282

Sequences and Series 286

Arithmetic Sequences..................................................................................................................... 286
Arithmetic Series.............................................................................................................................. 292
Geometric Sequences..................................................................................................................... 298
Geometric Series.............................................................................................................................. 302
Sequence, Summation, and Product Notation.......................................................................... 313
Nested Sums and Products............................................................................................ 325
Summary........................................................................................................................................... 331

I Identities, Manipulations, and Induction 338

11.1 Brute Force..................................................................................................................................... ....
11.2 Ratios...................................................................................................................................................
11.3 Induction........................................................................................................................................ ....
11.4 Binomial Theorem............................................................................................................................ 358
11.5 Summary..................................................... 365

12 Inequalities 371
12.1 Manipulating Inequalities............................................................................................................ 371
12.2 The Trivial Inequality.................................................................................................................. 378
12.3 AM-GM Inequality with Two Variables................................................................................... 384

xi
CONTENTS

12.4 AM-GM with More Variables........................................................................................................389

12.5 The Cauchy-Schwarz Inequality..................................................................................................392
12.6 Maxima and Minima .................................................................................................................... 400
12.7 Summary......................................................................................................................................... 406

13 Exponents and Logarithms 413

13.1 Exponential Function Basics..................................................................................................... 413
13.2 Introduction to Logarithms........................................................................................................ 416
13.3 Logarithmic Identities................................................................................................................. 424
13.4 Using Logarithm Identities........................................................................................................... 433
13.5 Switching Between Logs andExponents.................................................................................... 441
13.6 Natural Logarithms and Exponential Decay............................................................................ 444
13.7 Summary......................................................................................................................................... 451

14 Radicals 456
14.1 Raising Radicals to Powers.......................................................................................................... 456
14.2 Evaluating Expressions With Radicals........................................................................................ 465
14.3 Radical Conjugates .......................................................................................................................470
14.4 Summary......................................................................................................................................... 473

15 Special Classes of Functions

Rational Functions and Their Graphs
476
15.1 476
15.2 Rational Function Equations and Inequalities.......................................................................... 486
15.3 Even and Odd Functions.............................................................................................................. 492
15.4 Monotonic Functions 497
15.5 Summary 503

Piecewise Defined Functions 508

16.1 Introduction to Piecewise Defined Functions............................................................................ 508
16.2 Absolute Value.................................................................................................................................517
16.3 Graphing Absolute Value......................................................................................... 523
16.4 Floor and Ceiling...................................................................................... 529
16.5 Problem Solving with the FloorFunction....................................................................................535
16.6 Summary.......................................................................................................................................... 539

xii
 CONTENTS

17 More Sequences and Series 544

17.1 Algebra of Recursive Sequences................................................................................................ 544
17.2 Telescoping....................................................................................................................................... 552
17.3 Sums of Polynomial Series............................................................................................................ 561
17.4 Arithmetico-Geometric Series 567
17.5 Finite Differences 571
17.6 Summary 576

More Inequalities 581

18.1 Mean Inequality Chain............................................................................................................... 581
18.2 The Rearrangement Inequality................................................................................................... 593
18.3 When Formulas Fail..................................................................................................................... 600
18.4 Summary 606

19 Functional Equations 611

19.1 Finding Values.............................................................................................................................. 611
19.2 Finding Functions with Substitution ...................................................................................... 615
19.3 Separation.......................................................................................................................................... 619
19.4 Cyclic Functions.............................................................................................................................. 621
19.5 Summary 627

20 Some Advanced Strategies 631

20.1 Symmetry.......................................................................................................................................... 631
20.2 Substitution for Simplification......................................................................................................638
20.3 Method of Undetermined Coefficients ...................................................................................... 644
20.4 Constructing Polynomials From Roots...................................................................................... 649
20.5 Common Divisors of Polynomials................................................................................................655
20.6 Symmetric Sums Revisited............................................................................................................ 658
20.7 Summary.......................................................................................................................................... 665

xiii
CONTENTS

References 673

Hints to Selected Problems 674

Index 693

xiv
For Vanessa's casita. -RR

xv
xvi
 Extra! In this text, we use as the chapter headings images generated from a specific cellular
'∣ι⅜ ∣u+ '∣ι⅜> ιn+
automaton that is often referred to as Rule 110. A cellular automaton is a grid of cells
in which some cells are initially assigned a value. The cellular automaton also has a set
of rules that are used to assign values to subsequent cells. As an example, we'll explain
Rule Γ10 in more detail.
In Rule 110, each generation of cells consists of a row of cells, and the value of each
cell is either 1 or 0. We color a cell white if its value is 0 and black if its value is 1. Suppose
our initial generation consists of a single black cell in an otherwise infinite row of white
cells. Below, we show a small snippet of this generation that contains the black cell:

The next generation of the cellular automaton appears in the next row. The value of
each cell in that row is determined by the values of the three cells directly above it. There
are 23 = 8 possible combinations of values for three consecutive cells, which we label A
through H as follows:

We define a rule for the cellular automaton by determining what each combination of
three cells produces. Rule 110 is defined as follows:

1p ∣ sb
A B C D E F G H

■tv ■
So, for example, combination A tells us that if there are three black cells in a row in one
generation, then the cell directly below the center cell of these three black cells is white
(that is, has value 0). Recalling that black cells have value 1 and white cells have value
0, we can read the results of the 8 possible combinations above as 01101110. Evaluating
this as a binary number, we find OllOlllO2 = 110, which is where Rule 110 gets its name.
Now, let's see what happens when we apply Rule 110 to our first generation:

■
(1) (2) (3)

Most of the cells in the second generation are below three white cells in the first
generation. Rule H tells us that such cells must be white. There are only three cells
in the second generation that are not below three white cells; these are marked (1), (2)
and (3) above. Reading left to right, cell (1) is below cells that read white-white-black.
So, we apply Rule G to see that cell (1) is black. Similarly, the cells above cell (2) are
white-black-white, so Rule F tells us to color it black. Finally, the cells above cell (3) are
black-white-white, so Rule D tells us cell (3) is white.
xvii
 Extra! . . . continued from the previous page
I∣l∣+∣lll⅜∣ll∣+∣lll
* All the rest of the cells in the second generation are white, and therefore our first two
generations are:

J
(Again, the generations continue infinitely to the left and right, but all those cells are

As chapter headings in this book, we use various snippets of the subsequent gener­
ations of Rule 110. Rule 110 has been closely studied by mathematicians and computer
scientists. Stephen Wolfram conjectured that Rule 110 is Turing complete, which loosely
means that it can be adapted to replicate the logic of any possible computer. This con­
jecture was proved by Matthew Cook while working with Wolfram, and the proof was
published in 2004 in the journal Complex Systems.
On the website cited on page vi, you'll find links to more webpages with information
about cellular automata in general, and Rule 110 in specific, including many images that
display the intricate structure of Rule 110.

xviii
It nil began in the remote past, with a simple turn of thought from the declarative to the interrogative, from "this
plus this equals this to "this plus what equals this?" - John Derbyshire

CHAPTER

Basic Techniques for Solving Equations

We'll start by reviewing some basic techniques for solving linear equations. As we review these tactics,
we'll show how these same strategies can also be used to solve inequalities and more complex equations.

1.1 Isolation

One of the most straightforward ways to solve a one-variable equation or inequality is simply to
manipulate it until the variable is alone on one side and a constant is on the other side. In other words,
we isolate the variable.
II I I_________________________________________
Problems

Problem 1.2: Solve each of the following inequalities:

⅛<
(a) 2-4x≥Tθ (b) 7-3f<5-2t (c) 3-3x<4x + 7≤8-x

Problem 1.3: Find all x such that 2x2 + 5 = x2 + 18.

1
CHAPTER 1. BASIC TECHNIQUES FOR SOLVING EQUATIONS

Problem 1.5: In this problem, we find all values of c such that the equation

has no solutions for x.

(a) Solve the equation for x in terms of c.
(b) Find all values of c for which there is no solution for x to the equation you found in part (a).
(c) What values of x cause division by 0 in the original equation? To what values of c do these
values of x correspond in the equation you found in part (a)?
(d) List all values of c for which there is no solution for x to the original equation._______________

Problem 1.1: Solve the following equations:

(a) 3r-2 + 2r = 7- r + 3 (b) + 7 = 3(f - 9)

Solution for Problem 1.1:

(a) Simplifying both sides gives 5r — 2 = 10 - r. We add r to both sides to give 6r — 2 — 10, then add 2
to both sides to give 6r = 12. Finally, we divide by 6 to find r = 2.
We can test our answer by substituting r = 2 back into the equation. On the left we have
3r-2 + 2r = 3∙2-2 + 2∙2 = 8, and on the right we have 7-r + 3 = 7- 2 + 3 = 8. So, the equation
is satisfied when r = 2.
Important: Whenever you solve an equation, you can check an answer by substitut­

ψ ing it back into the original equation.

We won't do this on every problem in the text to save space, but it's a very useful habit to develop
to prevent errors.
Expanding the right side gives ⅛ + 7 = 3t - 27. Subtracting 3f and subtracting 7 from both sides
gives -⅞ = -34. Multiplying both sides by gives us t = ψ.

Just as we isolate the variable to solve a linear equation, we isolate the variable to solve a linear
inequality.

Solution for Problem 1.2:

(a) What's wrong with the following solution:

2
 1.1. ISOLATION

Bogus Solution: Subtracting 2 from both sides gives -4x > 8. Dividing both sides by
∣¾] “4 gives x > -2 as our solution.

We see that we've made a mistake when we test our answer. If x ≥ -2 is the solution, then
x = 0 should satisfy the original inequality (because 0 is greater than —2). However, if x = 0, then
it is not true that 2 — 4x ≥ 10. So, we've made a mistake.
Our mistake is that we forgot to reverse the direction of the inequality sign when we divided
by -4. When we divide -4x ≥ 8 by -4, we have x ≤ -2.
WARNING!! If we multiply or divide an inequality by a negative number, we must
* reverse the direction of the inequality sign.
(b) Adding 3t to both sides and subtracting 5 from both sides gives us 2 < t. It's common to write the
variable first when writing the solution to an equation or inequality, so we might also write our
answer to this problem as t > 2.
WARNING!! The inequality 2 < t is not the same as t <2. If we rewrite an inequality
* by swapping the sides of the inequality, we must reverse the direction
of the inequality sign. So, 2 < t is the same as t > 2.
(c) We are given three expressions linked together in what we call an inequality chain. Any value of
x that satisfies the inequality chain must satisfy both 3 - 3x < 4x + 7 and 4x + 7 ≤ 8 - x. Isolating
x in 3 - 3x < 4x + 7 gives us -∣ < x, while isolating xin4x + 7≤8-x gives x ≤ ∣. Therefore, we
must have both — ∣ < x and x ≤ ∣, which we can write as —∣ < x ≤
□
We say that an inequality is strict if the two sides cannot be equal, and that it is nonstrict if they can
be equal. So, t > 2 is a strict inequality, while x ≥ -2 is a nonstrict inequality.

Solutions to inequality problems are sometimes written using interval notation. For example, we
can denote "all real numbers greater than 3 and less than 5" by the interval (3,5). We use "(" and ")" to
indicate that the values 3 and 5 are not included. Notice that we don't write (5,3); the lesser end of the
interval always comes first. To use this notation to indicate the solutions to the inequality 3 < r < 5, we
write r ∈ (3,5), where "r ∈" means "r is in," and (3,5) indicates the real numbers greater than 3 and less
than 5, as we just described.

To include a boundary value in the interval, we use "[" for the lower bound and "]" for the upper
bound. For example, the statement x ∈ (-3,5] means -3 < x ≤ 5 and y ∈ [-12.2,0] means -12.2 ≤ y ≤ 0.
Finally, to indicate an interval that has either no upper bound or no lower bound (or neither), we use
the ∞ symbol. For example, we write t > -3 as t ∈ (-3, +∞). The "(-3" part indicates that no numbers
equal to -3 or lower than -3 are in the interval. The "+∞)" part indicates that the interval continues
forever in the positive direction. That is, there is no upper bound. So, the interval (-3, +∞) is all real
numbers greater than -3. Similarly, the statement w ∈ (-∞, -2] is the same as w ≤ -2. Notice that we
always use ( with —∞ and ) with +∞, instead of "[" and "]". In both cases, there is not a boundary
value to include in the interval, so we must use "(" or ")" instead of "[" or "]".
To write all real numbers in interval notation, we can write (-oo, +∞). We often use the symbol 1R to
denote "all real numbers." Therefore, we typically write the interval (-∞, ∞) as "1R", so x ∈ 1R means "x
is a real number."

3
CHAPTER 1. BASIC TECHNIQUES FOR SOLVING EQUATIONS

Problem 1.3: Find all x such that 2xz + 5 = x2 + 18.

Solution for Problem 1.3: Just as we isolate the variable to solve a linear equation, we isolate x2 in this
equation by subtracting x2 from both sides, then subtracting 5 from both sides. This gives us x = 13.
We then take the square root of both sides to find x = ± V13. □

WARNING!! When taking the square root of both sides of an equation, we must
a<* not forget negative values. Specifically, if we have x2 = a2, then we
have x = ±ar where the ± sign indicates that x can equal either a or -a.

Here's another example:

Solution for Problem 1.4: First, we notice that the denominators of the fractions are the same. So, we can
combine the fractions by subtracting 1∕( Vz - 3 - 1) from both sides, which gives us

^ Vz-^3-l
We get rid of the fraction by multiplying both sides by Vz - 3 -1, which gives us 2 Vz - 3 -2 = 2. Adding
2 to both sides, then dividing by 2 to isolate √z-3 gives us √z-3 = 2. We get rid of the square root
sign by squaring both sides, which yields z — 3 = 4, so z = 7.

WARNING!! Whenever we square an equation as a step in solving it, we have

*jka to check that our solutions are valid. If the solution to the squared
equation does not satisfy our original equation, then the solution is
called extraneous. An extraneous solution is not a valid solution to
the original equation.

Here, if z = 7, we have Vz - 3 -1=2-1 = 1, so our equation reads 1+2 = 3, which is true. So, z = 7 is
the only solution to the equation. □
To see how extraneous solutions might occur, consider the equation √x + 5 = 2. Subtracting 5 gives
√x = -3, which clearly has no solution. However, if we square y/x = -3, we produce x = 9, which
doesn't satisfy √x + 5 = 2. So, x = 9 is extraneous and there is no solution to y/x + 5 = 2. The problem
arises when we take the square root of x = 9 to evaluate y/x for our original equation—this gives us
y/x = 3, not y/x = -3, because we define y/x for positive numbers x to be the positive number whose
square is x.
One key step in each of the last two problems was isolating an expression. In Problem 1.3, we isolated
x2, and in Problem 1.4, we isolated Vz — 3.

Concept: Just as we can solve a linear equation by isolating the variable, we can
O∙~~dιJ often solve more complex equations by isolating an expression within the
equation.

4
 1.1. ISOLATION

Problem 1.5: Find all values of c such that the equation

has no solutions for x.

Solution for Problem 1.5: We want to know when there are no solutions for x, so we solve the equation
for x in terms of c. Then, we can use the resulting expression to determine what values of c fail to give
us a value for x.

Multiplying both sides of the equation by 2 - i gives us 3 = 2c - We isolate the term with x by
adding ∣ - 3 to both sides, which gives = 2c - 3. Taking the reciprocal of both sides gives = 2⅛3,
and multiplying both sides by c gives
c

We cannot have c = ∣, because we cannot divide by 0. It appears that for any other value of c, we can
use x ~ 2⅛3 t° find the solution for x that results for that value of c. But we have to be careful—we
cannot have a value of x that results in dividing by 0 in our original equation. So, we cannot have x = 0,
and we cannot have 2—^=0. The latter only occurs if x = ∣. Therefore, we must check for values of c
in x = 2⅛5 that give x = 0 or x = ∣.

If x = 0 in x = 2⅛, then we must have c = 0, so it appears that we cannot have c = 0 in our original
equation. Indeed, letting c = 0 gives us

which clearly has no solution, because the numerator is not 0 for any x. If x = 1 in x = 7¾, then we have
2 = 2F=3∙ Cross-multiplying gives 2c - 3 = 2c, which has no solutions. Therefore, there are no values of
c for which x = ∣.

We conclude that the only values of c for which there is no solution for x in the original equation are
c = 0 and c = ∣. For all other values of c, we can use x = 2⅛ to find the x that satisfies the original
equation for that value of c. □

1.1.1 Solve the following equations:

(a) 9-4x + 4 =-5x + 3-x (b) 5(y + 2)∕2 = 7∖

j -4

1.1.2 Solve the following inequalities. Write your solutions using interval notation.
(a) 8-2f≥ll + f (b) 7-2x≤6x-l <23-2x

1.1.3 Find all x such that 5x2 + 2x - 1 = 4x2 + 2x + 7.

1.1.4 Find all values of a such that there is no value of b that satisfies the equation —- = 5a.
3 b

5
CHAPTER 1. BASIC TECHNIQUES FOR SOLVING EQUATIONS

1.1.5
* Find all values of t such that t-l,t + l, and 4 could be the lengths of the sides of a right triangle.

1.1.6
* Find all a such that

Hints: 115

1.2 Substitution

A group of equations for which we seek values that satisfy all of the equations at the same time is
called a system of equations. In the next two sections, we explore the two most common strategies for
solving systems of equations: substitution and elimination. We'll start each section by reviewing these
methods by solving a system of two-variable linear equations. Then, we'll extend the strategies to more
challenging systems.
Both strategies involve trying to create an equation that has a single variable. Then, we solve that
equation for that variable. The key step in our first strategy, substitution, is to solve one of the equations
for one of the variables, and then substitute the result into other equations in the system.

Problems

Problem 1.6: Consider the system of equations

x - y = 4,
2x + y = 29.

(a) Find x in terms of y using the first equation.

(b) Substitute your answer from (a) for x into the second equation to create a new equation. Solve
this equation for y.
(c) Use your answers from (a) and (b) to find the values of x and y that satisfy the system.

Problem 1.7: Consider the following system of equations:

χ - y = 3, jP2

1 1 _ 1
x y ^ 2’

(a) Solve the first equation for x in terms of y.

(b) Create a new equation by replacing x in the second equation with your expression for x from
part (a).
(c) Find all solutions to the original system of equations.

Problem 1.8: Jackie and Rachel both worked during last summer and made $960 each. Rachel worked
16 hours more than Jackie, but Rachel earned $2 less per hour. How many hours did Jackie work?

6
 1.2. SUBSTITUTION

Problem 1.9: Suppose that a, b, c, and d are constants such that ax + b = ex + d for all values of x. In
this problem, we show that we must have a = c and b = d.
(a) The equation must hold for all values of x. Choose an appropriate value for x that allows you
to deduce that b = d.
♦

(b) Choose another value of x that, along with the fact that b — d, allows you to deduce that a - c.

We begin our study of substitution by solving a simple system of linear equations.

Problem 1.6: Solve the system of equations x — y = 4,2x + y = 29.

Solution for Problem 1.6: The first equation allows us to easily express x in terms of y, though we could
isolate either variable in either linear equation. Adding y to both sides of the first equation gives us
x = y + 4. This rearrangement of the first equation allows us to replace x in the second equation with
the expression y + 4 to give
2(y + 4) + y = 29.
Solving for y gives y = 7, so x = y + 4 = 7 + 4 = ll. The solution to the system of equations is x = 11 and
y = 7. We often write solutions to two-variable systems of equations as ordered pairs. For example, we
write (x, y) = (11,7) to indicate x = 11 and y = 7. □
Let's try a slightly more complicated system of equations.

Problem 1.7: Find x and y if x - y = 3 and 1 + 1 = ∣.

Solution for Problem 1.7: From the first equation, we have x = y + 3. Substituting for x in the second
equation, we have
1 1 _ 1
y+3 y 2
Multiplying this equation by 2y(y + 3) to get rid of the denominators gives us

2y + 2(y + 3) = y(y + 3).

Rearranging this equation gives us y2 - y - 6 = 0. Factoring the left side gives us (y - 3)(y + 2) = 0, so
we have y = 3 or y = -2. Since x = y + 3, the two solutions to our system of equations are (x, y) = (6,3)
and (x, y) = (1, -2). □

As we saw in the previous problem, and will see in the next one as well, when one or more equations
look complicated, substitution can quickly provide us with an equation that's not as intimidating.

Problem 1.8: Jackie and Rachel both worked during last summer and made $960 each. Rachel j
worked 16 hours more than Jackie, but Rachel earned $2 less per hour. How many hours did Jackie J

Solution for Problem 1.8: First, we translate this word problem into algebra. We let hj and hχ be the
numbers of hours Jackie and Rachel worked, respectively, and we let wj and ior be their corresponding

7
CHAPTER 1. BASIC TECHNIQUES FOR SOLVING EQUATIONS

hourly wages. Notice that each variable is clearly related to what it stands for. If we had used a, b, c,
and d instead, we would have had a harder time remembering what each variable represents.

Concept: If you have to assign multiple variables in a problem, use variables that
Q uη-) are clearly related to what they represent.

Here, we use h for hours and w for wages, and attach the subscripts J and R to stand for Jackie and
Rachel. So, for example, it's easy to remember that Hr means the number of hours Rachel worked.

We can now turn the information in the problem into equations:

hjivj = 960, (1∙1)

HrWr = 960, (1∙2)
Hr = hj + 16, (1∙3)
wjr = wj-2. (1-4)

The last two equations are already set up for substitution into Equation (1.2). We replace 1ir and wr in
Equation (1.2) with the right sides of Equations (1.3) and (1.4), and we have

(hj + 16)(wf - 2) = 960.

Expanding the left side gives hjwj — 2hj + 16wj — 32 = 960. Equation (1.1) tells us that hjWj = 960, so our
equation now is 960 - 2hj + 16ipj - 32 = 960. Simplifying this equation gives us hj = 8wj - 16. Now we
can substitute hj = 8wj - 16 into hjwj = 960 to create an equation with one variable:

(8wj - 16)w∕ = 960.

Simplifying this equation gives us the quadratic equation wj - 2wj -120 = 0, and factoring the quadratic
gives (wj - 12)(w∕ + 10) = 0. Since wj > 0, we know wj = 12.
Finally, hj = 8wj - 16 = 80 is the number of hours Jackie worked.

WARNING!! Everyone makes mistakes! After performing a great deal of problem

*∕ solving and computation to find a solution, it's a good idea to check
and see if your answer really works.

Jackie worked 80 hours at a wage of $12 per hour, for a total of

$12
80 hours ∙ - ---- = $960.
hour
Rachel worked 96 hours at a wage of $10 per hour, for a total of

$10
96 hours ∙ :------= $960.
hour
We see that our solution satisfies the conditions of the problem. □

8
 1.3. ELIMINATION

Problem 1.9: Suppose that a, b, c, and d are constants such that ax + b = ex + d for all values of x
Show that a = c and b = d.

Solution for Problem 1.9: Because the equation must be true for all values of x, we can assign any value to
x in the equation. We start by letting x = 0, since this will eliminate the ax and ex terms in the equation,
leaving us with b = d. Since b = d, our equation now is ax + b = ex + b, so ax = ex. Again, this must hold
for all values of x, so we let x = 1 to find a = c. □
Problem 1.9 gives us another use of substitution:

Concept: If you have an equation that must be true for all values of a variable, then
o≡≡g substituting values for that variable can give you information about the
equation. Specifically, setting the variable equal to 0 can eliminate some
terms from the equation. Setting the variable equal to 1 is also often a
useful simplifying substitution.

Ixeκi¼e¼

1.2.1 Find all ordered pairs (x, y) such that -x + 2y = 8 and x- y = -5.
1.2.2 Find b if a is a constant such that Sax - Sb = 4x + 6a for all values of x.
1.2.3 Find all ordered pairs (x, y) such that x + y = 6 and x2 + xy + 2y2 = 32.
*
1.2.4 Solve the system of equations for x:

x + 2y - z = 5,
v 3x + 2y + z = 11,
(x + 2y)2 - z2 = 15.

(Source: COMC) Hints: 154

1.3 Elimination

Elimination is the process of solving or simplifying a system of equations by combining the equations
to produce new equations with fewer variables.

Here's an example. Suppose that we know both x + 2y = 7 and x - y = 4. When we add twice the
second equation to the first, we get

(x + 2y) + 2(x - y) = 7 + 2 ∙ 4.

This equation simplifies to 3x = 15, which has only one variable instead of two. We say that we have
"eliminated y."

9
CHAPTER 1. BASIC TECHNIQUES FOR SOLVING EQUATIONS

Problems

Problem 1.10: In this problem we solve the following system of linear equations:

2x + 3y = 35,
13x - 6y = -2.

(a) Multiply the first equation by a constant to produce an equation whose y coefficient is the
opposite of the y coefficient in the second equation.
(b) Add your result from (a) to the second equation to produce a new equation.
(c) Solve the original system of equations.

Problem 1.11: The sum of three numbers is 12. The third number is 2 less than the sum of the second
number and three times the first number. Five times the first number is equal to the sum of the second
and third number. Find the numbers.

Problem 1.12: Find all ordered pairs (x, y) that satisfy the following system of equations:

x2 + xy = 126,
x2 - xy = 36.

Problem 1.13: Consider the following system of equations:

xy2 = 108, — = 1010.

y

(a) Find positive integers A and B such that the simplification of (xy2)x (y) has only one variable.
(b) Solve the given system of equations. (Source: COMC)

Problem 1.14: Find all ordered pairs (x, y) such that 5x2 - 3xy = 280 and 3y2 - 5xy = 56.

Problem 1.10: Solve the following system of equations using elimination:

2x + 3y = 35,
13x - 6y = -2.

Solution for Problem 1.10: If the coefficient of y in one equation were the opposite of the coefficient of y
in the other equation, then we could add the equations to eliminate y. So, we multiply the first equation
by 2 to give us the system

4x + 6y = 70,
13x - 6y = -2.

10
 1.3. ELIMINATION

Adding these equations eliminates y and leaves 17x = 68. This equation is true whenever both the
original equations are true. Solving 17x = 68 gives us x = 4.

Now that we know x = 4, we can replace x with 4 in either of the original equations to solve for the
value of y. For instance, from the first equation in the original system, we have 2(4) + 3y = 35, so y = 9.
Therefore, the solution to the system of equations is (x, y) = (4,9). □

Notice that it is much simpler to eliminate y than x in the system in Problem 1.10. Had we wanted to
eliminate x, we would have had to multiply the first equation by 13 and the second by 2, which is more
complicated than simply multiplying just the first equation by 2.

Concept: When using elimination to solve a system of equations, look for the easiest
O—°⊂> variables to eliminate first.

Problem 1.11: The sum of three numbers is 12. The third number is 2 less than the sum of the >r
second number and three times the first number. Five times the first number is equal to the sum of ∣
the second and third number. Find the numbers.

Solution for Problem 1.11: We begin by rewriting the given information as equations. Let the three
numbers be a, b, and c, in that order. We have

a + b + c — 12,
c = 3a + b-2,
5a = b + c.

We could go ahead and begin eliminating variables from the system, but first we get more organized by
rewriting the equations with all the variables on the left side of the equations:

a + b + c = 12,
-3a - b + c - -2,
5α - b - c = 0.

Organizing our equations like this makes it easier to see ways to eliminate variables. There are a few
quick ways to eliminate a variable by adding two equations, but our organization also helps us see that
adding the first and third equation eliminates two variables and leaves us with just 6a = 12, so a = 2.
Letting a = 2 in our equations gives us

b + c = 10,
-b + c = 4,
-b-c = -10.

Note that the last equation is just -1 times the first equation, so the last of these three equations is
redundant. Adding the first and second equations eliminates b and leaves 2c = 14, soc = 7. Finally, we
have b = 10 - c = 3, so our solution is (a, b, c) = (2,3,7). □

Concept: Don't just blindly eliminate variables. Keep your eyes open for quick ways
θ=⅛zd t° eliminate variables.

11
CHAPTER 1. BASIC TECHNIQUES FOR SOLVING EQUATIONS

Problem 1.12: Find all ordered pairs (x, y) that satisfy the following system of equations:

x2 + xy = 126,
x2 - xy = 36.

Solution for Problem 1.12: Elimination is not just a method for removing linear terms from a system of
equations. Elimination can also help us to remove other expressions that complicate a system. Here,
we add the given equations to get rid of the xy term. This gives us a single equation in one variable,
2x2 = 162, that we can easily solve to find x = ±9. When x = 9 in either of the original equations, we
find y = 5. When x = -9, we have y = -5. So, our solutions are (x, y) = (9,5) and (x, y) = (—9, —5). □

Concept: Elimination can be an effective tool when working with systems of non-
θ≡~-lznj linear equations.

r3
Problem 1.13: Find all ordered pairs (x, y) that satisfy both xy2 = 108 and — = 1010. (Source: COMC)

Solution for Problem 1.13: We can't eliminate x or y by simply adding these two equations, nor can
we eliminate a variable by multiplying the two equations by nonzero constants and adding the results.
However, we can use multiplication and division to eliminate whichever variable we like.
Solution 1: Squaring the second equation gives us x6∕y2 = 1020. Multiplying this by the first equation
gives us
*y 2) (^ς) = (1°8) (1°20)'
<

so x7 = 1028, which means x = 104 = 10000. Substituting this value of x into x3∕y = 1010, we find that
y = 102 = 100, so our only solution is (x,y) = (10000,100).
Solution 2: Dividing the cube of the first equation by the second equation eliminates x:

y 2)3 _ io24
(*

~ 1010'
y

so y7 = 1014. Therefore, we have y = 102 = 100, and we find x = 104 by substituting y = 100 into either
of the original equations. □

Concept: We can use multiplication and division to eliminate variables in some

O—unJ systems of equations.

But, sometimes we have to do some work to be able to do so.

Problem 1.14: Find all ordered pairs (x, y) such that 5x2 - 3xy = 280 and 3y2 - 5xy = 56.

12
 1.3. ELIMINATION

Solution for Problem 1.14: We might try eliminating the xy terms, but we'd still have x2 and y2 terms to
deal with. Instead, we note that we can easily factor the left sides of both equations:

x(5x - 3y) = 280,

y(3y - 5x) = 56.

Concept: Factoring is your friend. When stuck on an algebra problem, it's very often
Qr^~t∣iJ useful to factor expressions that appear in the problem.

The equations share a similar factor. In fact, we have 5x - 3y = -(3y - 5x), so (5x - 3y)∕(3y - 5x) = -1.
This means we can eliminate these factors by dividing the first equation by the second, which leaves
us with -x∕y — 280/56 = 5. We have to be careful when we do this division, though. What if 3y = 5x?
Then we will be dividing by 0. Fortunately, we can rule out the possibility that 3y = 5x by noting that
this makes the left sides of our equations both equal to 0, so we can't have 3y = 5x.
After dividing the first equation by the second, we have -x∣y = 5. Solving for x in terms of y gives
x = -5y. Substituting x = —5y into the first equation in the original system gives 5(-5y)2-3(-5y)y = 280.
Simplifying this equation gives 140y2 = 280, which has solutions y = ±√2. Using x = -5y again, we see
that if y = √2, we have x = -5 √2, and if y = - √2, then x = 5 √2. So, the solutions to the system of
equations are (x, y) = (5 √2, - √2) and (x, y) = (-5 √2, √2). □

Notice that a key step in our solution to Problem 1.14 was eliminating a whole expression, not just a
variable.

Concept: While we usually substitute for a variable or eliminate a variable, we can

Q -uj also sometimes substitute for or eliminate complicated expressions.

Ixer<i¼e⅜

1.3.1 Find all ordered pairs («, b) such that 3a - 2b = -8 and 5a + 4b = 5.

1.3.2 Let x and y be real numbers satisfying 2 = ∣ = 1. Determine x3. (Source: Mandelbrot)

1.3.3 Find all ordered pairs (x, y) such that x3y2 = 234 and x4y = 232.

1.3.4 Find the solution to the system of equations,

x + 2z = 1,
-xy + 2z = 13,
xy + z2 = -14.

1.3.5 Find all r and s such that r2 - 2rs = 27 and rs - 2s2 = 9.

Extra! Basic research is what I am doing when 1 don't know what I am doing.
*-l"*,"*ll"* _ Wernher von Braun

13
CHAPTER 1. BASIC TECHNIQUES FOR SOLVING EQUATIONS

1.4 Larger Systems of Linear Equations

In this section, we use our strategies from the last two sections to discuss larger systems of linear
equations. We also introduce some terminology used to describe systems of equations.
ιιι∣ .
Problems

Problem 1.15: Consider the system of equations

x + y + z = 1,
2x + 2y + 3z = -2,
3x + y - z = 13.

(a) Use the first equation to eliminate the variable z in each of the other two equations.
(b) Solve the resulting system of two equations, and solve the original system above.

Problem 1.16: Find all solutions to the system of equations

7x + 5y -z = 28,
x - 4y - z = 5,
4x - 5y - 2z = 19.

Problem 1.17: Find all solutions to the system of equations

a + b + c = 6,
a + 2b + 3c = 14,
2rz + 5b + 8c = 36.

Problem 1.18: Find a - ft + c if 3α - + 5c = 44 and a + 2c = 12.

Elimination is also effective with systems of equations involving many variables, as we see in the
next problem:

Problem 1.15: Use elimination to solve the system

x + y + z = l,
2x + 2y + 3z = -2,
3x + y - z = 13.

Solution for Problem 1.15: Why is this not the best start we can make:

Extra! There are two kinds of people, those who finish what they start and so on.
- Robert Byrne

14
 1.4. LARGER SYSTEMS OF LINEAR EQUATIONS

Bogus Solution: Aciding the first equation to the third eliminates z and gives us
4x + 2y = 14, and dividing this by 2 gives 2x + y = 7. Subtracting
the first equation from the third eliminates y and gives 2x — 2z = 12,
so x -z = 6. So, now we just have 2 two-variable equations.

Indeed, we have 2 two-variable equations, but there are different variables in the two equations. (We
could use these two equations to find y and z in terms of x, and then use substitution to finish. See if
you can figure out how.)

We need to eliminate the same variable in our first two eliminations so that we,ll then have a system
of 2 two-variable equations with the same variables. So, after eliminating z by adding the first and third
equations, we then eliminate z from the first and second equations. We do so by multiplying the first
equation by 3 to get 3x + 3y + 3z = 3, then subtracting this from the second equation in the given system
to find -x - y = -5.

We have reduced the system of three linear equations in three variables to a system of two linear
equations in two variables:

2x + y = 7,
-x - y = -5.

Adding these two equations gives x = 2. Substituting x = 2 into 2x + y — 7 gives us y = 3. Finally, we

let x = 2 and y = 3 in the first equation of the original system, x + y + z = l, to find z = -4. The single
solution to the system is (x, y, z) = (2,3, -4). □

In our solution to Problem 1.15, we first eliminate a variable from a system of three linear equations
to get a system of two linear equations with two variables. We then eliminate a variable from that
simpler system to solve the system.

Concept: If a system of linear equations has solutions, they can be found by repeat-
ιz∏j edly eliminating variables from the system.

Following this process, we can find any solution(s) to any system of linear equations that has
solutions, no matter how many variables and equations are in the system.

Note that in Problem 1.15, it doesn't matter which variable we choose to eliminate first. If we wanted
to, we could have eliminated x or y first (instead of z).

Problem 1.16: Find all solutions to the system of equations

7x + 5y - z = 28,
x - 4y - z = 5,
4x - 5y - 2z = 19.

Solution for Problem 1.16: Subtracting the second equation from the first gives us 6x+9y = 23. Subtracting
twice the second equation from the third gives us 2x + 3y = 9.

15
CHAPTER 1. BASIC TECHNIQUES FOR SOLVING EQUATIONS

Now we have a simpler system of equations. However, a problem occurs when we try to eliminate
one of the remaining variables. If we subtract three times 2x + 3y = 9 from 6x + 9y = 23, we get 0 = -4.
There is no triple (x, y, z) that makes this statement true. Since the statement 0 = -4 is the result of
combining the equations in our system, any values of (x, y, z) that satisfy all three equations must also
satisfy this equation. However, the equation 0 = -4 is never true, so there are no solutions to the system
of equations. □

Definitions: When there is at least one solution to a system of equations, we say that the system is
consistent. Otherwise, we say that the system is inconsistent.

The system in Problem 1.15 is a consistent system, and the system in Problem 1.16 is inconsistent.
So far, all the consistent systems we've solved had a finite number of solutions. This isn t always the
case.
Problem 1.17: Find all solutions (a, b, c) to the system of equations

cι ÷ b ÷ c — 6,
a ÷ 2b + 3c — 14,
2a + 5b + 8c = 36.

Solution for Problem 1.17: We eliminate a by subtracting the first equation from the second, then
subtracting twice the second equation from the third. This gives us a simpler system in two variables:
b + 2c — 8,
b + 2c = 8.
These two new equations are the same! There aren't any more equations we can use to eliminate
variables or simplify the system—we've used all three of the original equations to find these two. So,
every solution to the equation b + 2c = 8 can be used to give a solution to the system of equations. The
equation b + 2c = 8 has infinitely many solutions, so there are infinitely many solutions to the system of
equations.
To express all possible solutions to our system of equations, we note that b + 2c - 8 means that
b = 8 - 2c. From a + b + c = 6, we have a = 6- b- c = 6-(8-2c)-c = -2 + c. For any value of c, we
can use a = -2 + c and b = 8 - 2c to generate a solution to our system of equations. So, for all values
of c, the triple (a, b, c) = (-2 + c, 8 - 2c, c) gives us a solution to the system. We can check this by letting
(a, b, c) = (-2 + c, 8 - 2c, c) in each of our equations:
(-2 + c) + (8 - 2c) + c = 6,
(-2 + c) + 2(8 - 2c) + 3c = 14,
2(—2 + c) + 5(8 — 2c) + 8c = 36.
Indeed, all three equations are satisfied if (a,b,c) = (-2 + c, 8 — 2c, c) for any value of c, so these are the
solutions to our system.
We could instead write our solutions in terms of a different variable. For instance, since λ = -2+c, we
have c = a + 2, which we can substitute into (αz b, c) = (-2 + c, 8 - 2c, c) to see that (a, b, c) = (a, -2a + 4, a + 2)
is a solution for all values of a.

16
 1.4. LARGER SYSTEMS OF LINEAR EQUATIONS

Don't worry about the many different ways we can represent the solutions to a system of equations
that has infinitely many solutions. The most important point to understand now is how to determine
that a given linear system has infinitely many solutions. □

As mentioned before, we can solve systems of linear equations by continually simplifying them,
removing one variable at a time from the system. In general, we can find any solution(s) to a system
of linear equations by simplifying the system one step at a time until we have a single linear equation
with as few variables as possible (such as x = 2 in Problem 1.15, or b + 2c = 8 in Problem 1.17). That last
remaining equation can then be used to describe all solutions to the system. If this remaining equation
has no solutions, then the system has no solutions.
When we multiply two equations by different constants and add the results, the resulting equation
is called a linear combination of the first two equations. For example, consider the two equations

2x + y = 3,
x + 7y = -5.

If we multiply the first equation by 2 and the second by -3 and add the results, we get x - 19y = 21. So,
the equation x — 19y = 21 is a linear combination of our original two equations.

In Problem 1.17, each equation in the system is a linear combination of the other two. For example,
when we add 3 times a + 2b + 3c = 14 to —1 times 2a + 5b + 8c = 36, we get a + b + c = 6.
We call such a system of equations a dependent system.

Definitions:

• When some equation in a system of linear equations is a linear combination of the others, we
call that system dependent. Dependent systems are also sometimes called redundant since at
least one of the equations can be removed without changing the solutions to the system.

• When no equation in a system of linear equations is a linear combination of the others, the
system of equations is independent.

While the system in Problem 1.17 is an example of a dependent system, the systems in Problems 1.15
and 1.16 are examples of independent systems of linear equations.

Don't worry too much about all these terms right now. The main point to take from this discussion
of larger systems of linear equations is that systems of linear equations can have no solution, a single
unique solution, or infinitely many solutions. Using the tools we have discussed so far, you should now
be able to determine if a linear system with a relatively small number of variables has a solution or not,
and, if the solution is unique, you should be able to find it. When you study a field of mathematics
called linear algebra, you will learn much more effective general strategies for working with systems of
linear equations.

Problem 1.18: Find a - b + c if 3a — b + 5c = 44 and a + 2c = 12. -

Solution for Problem 1.18: We might try to solve the system with substitution. We solve the second
equation for a to get a = 12 — 2c. Substituting this into our first equation gives us 3(12 — 2c) — b + 5c — 44.

17
chapter 1. bas∣c techniques for solving equations

Simplifying this equation gives us b + c = -8. Now what? We don't have any more equations left to
work with, so it appears that we can't find a, b, and c.
However, we're not asked to find a, b, and c.

Concept: Keep your eye on the ball. For some problems, you don't have to find the
Q c∏g values of all the variables in the problem to solve the problem.

We're asked to find a — b + c. We know that α = 12 - 2c and that b + c — -8. Solving the second equation
for b gives b = -8 - c. Substituting our expressions for a and b into a-b + c gives us

a - b + c = (12 - 2c) - (-8 - c) + c = 20.

Without even finding α, b, and c, we have found a - b + cl

We could also have tackled this problem by trying to combine 3a-b + 5c and a + 2c to give a — b + c.
We notice that subtracting twice the second expression from the first gives us the correct a and b terms.
Let's see what happens with the c terms:

(3a - b + 5c) - 2(α + 2c) = a - b + c.

So, subtracting twice a + 2c = 12 from 3a-b + 5c = 44 gives us a - b + c = 20. □

You might find Problem 1.18 somewhat gimmicky (and you're more than a little right!), but it's an
example of how we can sometimes evaluate an expression without knowing the values of the variables
in the expression.

Exercises

1.4.1 Solve each of the following systems of equations:

(a) 2fl + b + c — 2, (b) 2a ÷ b ÷ c — 3, (c) x + y + z = 13,

5a - b + 3c = 5, 5a — b + 3c = 5, x + y-z = -1,
7a + 4c = 8. 7a + 4c = 8. x - y + z = 9.

1.4.2 Consider the system of equations in Problem 1.17. Find constants k1 and k2 such that k↑ times the
first equation in the system plus k2 times the second equation in the system gives us the third equation
in the system.
1.4.3 Find the value of a⅛7c14 given that a'ib2c = 108 and α2b3c5 = 240. Hints: 122

1.4.4 Solve each of the following systems of equations:

(a) a + 3b-2c = 18, (b) χi + χz + *3 + *4 = 1,
2a-4b + 2c-d = -16, Xi + x2 - X3 = 2,
-a + 2b-5c + d = 23, x2 + X3 = 0.
3b-7c + d = 35.

18
 1.5. SUMMARY

*
1.4.5 Suppose we have a system of three linear equations in three variables such that a times the
first equation plus b times the second equation equals the third equation, where a and b are nonzero
constants.

(a) Must there be constants c and d such that c times the second equation plus d times the third
equation equals the first equation?
(b) If the answer to part (a) is "yes," is it possible to find c and d in terms of a and b?

1.5 Summary

We often solve linear equations and inequalities with one variable by isolating the variable. We can also
sometimes solve more complicated equations by isolating non-linear expressions such as x2 or √z - 3.

Two general strategies for solving systems of equations are:

• Substitution. Solve one of the equations for one variable in terms of the others, and then substitute
the result into the other equations.

• Elimination. Add, subtract, multiply, or divide two equations so that one variable is eliminated,
and the resulting equation has one fewer variable than the original equations.

As we saw with larger systems of equations, sometimes we have to use substitution or elimination
several times to solve a problem.
Once we have determined how many solutions a system of equations has, there is some special
terminology we can use to describe the system:

Definitions:

• When there is at least one solution to a system of equations, we say that the system is consistent.
Otherwise, we say that the system is inconsistent.

• When some equation in a system of linear equations is a linear combination of the others, we
call that system dependent. Dependent systems are also sometimes called redundant since at
least one of the equations can be removed without changing the solutions to the system.

• When no equation in a system of linear equations is a linear combination of the others, the
system of equations is independent.

Things To Watch Out For!___________________________________________

• If we multiply or divide an inequality by a negative number, we must reverse the direction of the
inequality sign.

19
CHAPTER 1. BASIC TECHNIQUES FOR SOLVING EQUATIONS

• When taking the square root of both sides of an equation, we must not forget negative values.
Specifically, if we have λ'2 = fl2, then we have x = ±a, where the ± sign indicates that x can equal
either a or -a.
• Whenever we square an equation as a step in solving it, we have to check that our solutions are
valid. If a solution to the squared equation does not satisfy our original equation, then that solution
is called extraneous. An extraneous solution is not a valid solution to the original equation.

• Everyone makes mistakes! After performing a great deal of problem solving and computation to
find a solution, it's a good idea to check and see if your answer really works.

Problem Solving Strategies__________________________-—

Concepts: ∙ Just as we can solve a linear equation by isolating the variable, we

can often solve more complex equations by isolating an expression
within the equation.

• If you have to assign multiple variables in a problem, use variables

that are clearly related to what they represent.

• If you have an equation that must be true for all values of a variable,
then substituting values for that variable can give you information
about the equation. Specifically, setting the variable equal to 0 can
eliminate terms from the equation. Setting the variable equal to 1 is
also often a useful simplifying substitution.

• When using elimination to solve a system of equations, look for the

easiest variables to eliminate first.

• Elimination can be an effective tool when working with systems of

nonlinear equations.

• We can use multiplication and division to eliminate variables in some

systems of equations.

• Factoring is your friend. When stuck on an algebra problem, it's very

often useful to try to factor expressions that appear in the problem.

• While we usually substitute for a variable or eliminate a variable,

we can also sometimes substitute for or eliminate more complex
expressions.
• Keep your eye on the ball. For some problems, you don't have to find
the values of all the variables in the problem to solve the problem.
 REVIEW PROBLEMS

Review Problems

1.19 Solve the following systems:

(a) 4x + 5y = 43, (b) 2x + 4y = 18, (c) 3x - y = 1,

9x - 2y = 57. 3x - 3y = -12. 12x - 4y = 4.

1.20 Jeff is 4 times older than his daughter. Five years ago he was 9 times older than his daughter.
How old is his daughter?

1.21 Find all solutions to each of the following systems of equations:

(a) 8x + y - z = 46, (b) 2a — 3b + 5c = 17, (c) x + y - z = 11,
3x + 4y-2z = 27, 3a + b-6c = -4, 3x-2ιj + 2z = 3,
4x - y + z = 14. a + 4b - 11c =-15. 5x + 3y - 3z = 43.

1.22 Solve the following system of equations for a, b, c, d, and e:

Ω! 4^ b — 2,
b + c = 13,
c + 3d — 37,
3d + 4e = —23,
8e + 9α = -43.

1.23 Solve each of the following inequalities. Write your answers in interval notation.
(a) ∣(x-l)≤^ + ∣ (b) x~1 ≤3x + 2≤2x + 6 (c) 14 - (x - 1) > -(5 - 3x)

1.24 Find all ordered pairs (x, y) that satisfy both ∙√x + yjy -7 and 3√x - 4√y = -14.
1.25 Solve the following system of equations:

Xl + X2 - X3 - X4 = 1,
xι + 2x2 + 3x3 - X4 = 2,
3x] + 5x2 + 5x3 - 3x4 = 6.

1.26 Find all values of r and s such that----- = — and-------- = -.

s - r 3r r s 6
1.27 Find the value ofa + b + c given that

2a-b + 5c = 13,
2a + 3b + c = 75.

21
CHAPTER 1. BASIC TECHNIQUES FOR SOLVING EQUATIONS

1.28 Find all ordered pairs (π, b) such that a*19b = 4 and a∕3h = 18.
1.29 Find all ordered pairs (x, y) that satisfy both x2 + xy = 28 and y2*+ xy = -12.
1.30 Find all solutions to the system of equations a - 2b = -4, a2 - 2b2 = -14.

Challenge Problems

1.31 A tennis player computes her "win ratio" by dividing the number of matches she has won by the
total number of matches she has played. At the start of a weekend, her win ratio is exactly .500. During
the weekend she plays four matches, winning three and losing one. At the end of the weekend her win
ratio is greater than .503. What is the largest number of matches that she could have won before the
weekend began? (Source: AIME)
1
1.32 Find all x such that -4 < - < 3. Hints: 278
x
1.33 Suppose that a, b, c, d, e, and f are constants such that ax2 + bx + c = dx2 + ex + f for all values of x.
Prove that a = d, b = e, and c - f.
1.34 A right triangle has both a perimeter and an area of 30. Find the side lengths of the triangle.
1.35 Find all pairs of real numbers (a, b) such that (x - a)2 + (2x - b)2 = (x - 3)2 + (2x)2 for all x.

1.36 If — = 24 and — = 30, find the value of y⅛.

z χ (yz)5

1.37 Ten people form a circle. Each picks a number and tells it to the two neigh­
bors adjacent to him in the circle. Then each person computes and announces
the average of the numbers of his two neighbors. The figure shows the average
announced by each person (not the original number the person picked).
"8"
What number was picked by the person who announced the average 6?
(Source: AHSME) Hints: 198
"6"
1.38 Find all values (a, b, c, d) that satisfy ac - 6, ad = 4, be = 9, and bd = 6.
1.39 Find all ordered pairs (x, y) such that 9x + 3y - 2 - 0 and 9x2 + 3y2 - 7x = 0. (Source: CEMC)

1.40 Find all ordered pairs (x, y) of solutions to the system of equations,
2y-3x _ 7
xy “ 12'
5x + 3y _ 25
xy 4 ’
Hints: 294
1.41 When we place four numbers in a 2 × 2 grid, we form a matrix. We find the determinant of such

22
 CHALLENGE PROBLEMS

a 2 × 2 matrix as follows:
a b
= ad - be.
c d
For example,
3 -2
5 7

Suppose a, b, c, d, e, and f are constants in the system of linear equations

ax + by = e,
cx + dy = f.

a b
Cramer's Rule states that if is nonzero, then the solution to this system is
c d

e b a e
c f
and
a b
c d c d

a b
Prove Cramer's Rule. What happens if the determinant equals 0?
c d
1.42
* For what values of k does the linear system of equations

kx + y + z = k,
x + ky + z = k,
x + y + kz = k,

(a) have no solution?

(b) have an infinite number of solutions?
(c) have precisely one solution?

(Source: CEMC) Hints: 104

1.43
* Let a, b, c be nonzero constants. Solve the system

ay + bx = c,
az + ex = b,
bz + cy = a,

for (x, y, z) in terms of a, b, and c.

1.44
* The binomial coefficients can be arranged in rows to form Pascal's Triangle (where row n is
(o)< (")∕ ∙∙∙> (nn))∙ ha which row of Pascal's Triangle do three consecutive entries occur that are in the ratio
3:4:5? (Source: AIME) Hints: 325

23
CHAPTER 2. FUNCTIONS REVIEW

Any impatient student of mathematics or science or engineering who is irked by having algebraic symbolism thrust
upon him should try to get along without it for a week. - Eric Temple Bell

CHAPTER
I
I Functions Review

This chapter is a review of essential facts about functions that are covered in greater detail in Introduction
to Algebra. If any of this is entirely new to you, we recommend working through the appropriate chapters
in that text.

2.1 Function Basics

Suppose we have a machine that accepts any number, multiplies it by two, adds three to this product,
and then outputs the result. Mathematically speaking, this machine is a function because there is only
one possible output from the machine for each input to the machine. We can give this function a label,
f, and write the function as
∕(x) = 2x + 3.
This simple equation describes the machine. The "(x)" after f on the left side indicates that we are
putting x into the function f. When speaking, ∕(x) is read "f of x." The x in the equation ∕(x) = 2x + 3 is
a dummy variable, which means that it is essentially a placeholder. When we put a specific number in
our machine, we replace x with that number in the equation ∕(x) = 2x + 3 to determine what the machine
outputs. For example,
∕(5) = 2 ∙ 5 + 3 = 13,
so the machine outputs 13 when we put 5 into it.
Functions are really that simple. We define the function, then whenever we input a number to the
function, we follow the definition to get an output. Usually, we use an equation, such as ∕(x) = 2x + 3,
to define a function. For the obvious reason, f is the most commonly used label for /unctions.
In this chapter, we will only discuss functions that take real number inputs and give real number
outputs. The domain of a function consists of all the values we are able to input to the function and

24
 2.1. FUNCTION BASICS

get an output, and the range of the function consists of all the values that can possibly come out of the
function.
For example, consider the function

>>√3∙
The value x = 3 is not part of the domain of this function, because is not defined when x = 3. We
can safely put any other value of x into this function, so the domain of f is "all real numbers except 3."
Similarly, there is no value of x for which it is possible to make the function output 0. However, we can
make the function output any other real number, so the range of f is "all real numbers except 0."
We call a function that can only output real numbers a real-valued function, or, even more simply, a
real function. Unless a problem states otherwise, you can assume that all functions in this chapter are
real functions. Moreover, you can assume that they are only defined for real number inputs. Later in
this book we will explore a variety of functions that can take nonreal inputs and give nonreal outputs.
I i I L_____
Problems

Problem 2.1: Let ∕(x) = 2x2 - 3.

(a) Find∕(2).
(b) Find all values of t such that f(t) = 47.
(c) Find ∕(3x + 1).

Problem 2.2: A function can be defined to accept multiple inputs. For example, let

g(α, b, c) = 3a -2b + 7c2.

(a) Evaluate g(3, -5, -1). (b) Find b if g(b, 2, -1) = 21.

Problem 2.3: Suppose f(x) = αx4 + bx2 + x + 5, where a and b are constants. In this problem, we find
t(4) if f(-4) = 3.
(a) Use t(-4) = 3 to write an equation in terms of a and b.
(b) Express t(4) in terms of a and b.
(c) Find f(4).

Problem 2.4: Find the domain and range of each of the following functions:
, ' 2f
(a) ∕(x) = 2x - 3 (c) gω=∕-ι

(b) ∕(x) = √-2x + 7 (d) ∕z(x) = 9x2 + 4

Problem 2.5: Suppose p(x) = 2 √x + 3, but that p(x) is only defined for 4 ≤ x ≤ 9.
(a) Is there a value of x for which p(x) = 5?
(b) What is the range of p?

25
CHAPTER 2. FUNCTIONS REVIEW

Problem 2.6: Find the domain of each of the following functions:

√2a - 5 2t - 4
w <ω (b) g(D= 1 l
t 3∕-4

χ2 — χ _ (x—h.υ _ * }----- 1 ancj g(χ) _ χ qle same function7

Problem 2.7: Notice that ----- — ≡
x-1

Problem 2.1: Let f(x) = 2x2 - 3.

(a) Find ∕(2).
(b) Find all values of t such that ∕(f) = 47.
(c) Find∕(3x + 1).

Solut ion for Problem 2.1:

(a) We simply replace x with 2 in the function definition, which gives us ∕(2) = 2(2)^^ - 3 = 8- 3 = 5.
(b) If ∕^(∕) : ■ 47, we must have 2f2 - 3 = 47. Adding 3 to both sides gives 2t2 = 50, and dividing by 2
gives /■' = 25. The two values of t that satisfy this equation are f = 5 and t = —5. Checking our
answer, we find ∕(5) = 2(5)2 - 3 = 47 and ∕(-5) = 2(-5)2 - 3 = 47.
Concept: We can often check our answers in algebra problems by plugging the
solutions we find back into the original question.

(c) Just as we replace the dummy variable in the function definition with a number to evaluate
a function for a specific numerical input, we can replace the dummy variable with an entire
expression. Moreover, we can replace the x in the function definition ∕(x) = 2x2 - 3 with an
expression that contains x. For example, we have

∕(3x + 1) = 2(3x + 1)2 - 3 = 2(9x2 + 6x + 1) - 3 = 18x2 + 12x - 1.

□
Problem 2.2: A function can be defined to accept multiple inputs. For example, let

g(a, b, c) = 3a -2b + 7c2.

(a) Evaluate^,-5,-1). (b) Find b if g(b,2,-1) ≡ 21.

Solution for Problem 2.2:

(a) We replace the dummy variables a, b, and c in the function definition with the values 3, -5, and
-1, in that order. This gives usι⅛(3z -5,-1) = 3(3) - 2(-5) + 7(-l)2 ≡ 26.
(b) What's wrong with this solution:

26
 2.1. FUNCTION BASICS

Bogus Solution: We have

g(b, 2, -1) = 3(2) - 2b + 7(-l)2 = -2b + 13,

so we seek the value of b such that -2b + 13 = 21, which is b = -4.

We see our mistake immediately if we check our answer. We have
g(-4,2,-l) = 3(-4) - 2(2) + 7(-l)2 = -9,
but we want g(b,2, — 1) = 21. Looking at our Bogus Solution above, we see that in finding an
expression for g(b, 2, -1), we set a = 2 and c = -1, and left b as b. But this isn't what we must do
to evaluate g(b, 2, -1).
WARNING!! When evaluating a function that accepts multiple inputs, make sure
• you assign the values to the dummy variables in the correct order.
To evaluate g(b,2, — 1), we replace a in the function definition with b, we replace b with 2, and
replace c with -1. This gives us
g(b,2,-l) = 3b- 2(2) + 7(-l)2 = 3b + 3.
So, we seek the value of b such that 3b + 3 = 21, which gives us b = 6. Checking our answer, we
find that g(6,2, -1) = 21, as desired.

Problem 2.3: Suppose t(x) = ax4 + bx2 + x + 5, where a and b are constants. Find t(4) if t(—4) = 3.

Solution for Problem 2.3: We have

t(4) = 256« + 16b + 4 + 5 = 256« + 16b + 9.
If we can find « and b, then we can evaluate t(4). We turn to the only other piece of information we have,
which is t(—4) = 3. We have
f(-4) = 256« + 16b - 4 + 5 = 256« + 16b + 1,
so f(-4) = 3 gives us 256« + 16b + 1 = 3, which means 256« + 16b = 2. Unfortunately, this doesn't tell us
a or b. However, looking back at our expression for f(4), we see that we don't need a and b. We need
256« + 16b, which we know equals 2. So, we have f(4) = 256« + 16b + 9 = 2 + 9 = 11. □

Concept: Keep your eye on the ball. Sometimes we don't need to evaluate every
ζ~) j□ variable in a problem in order to solve the problem.

Now that we have a little practice evaluating functions, let's try finding the domain and range of
specific functions.

Problem 2.4: Find the domain and range of each of the following functions:

(a) ∕(x) = 2x - 3 (c)

(b) ∕(x) = √-2x + 7 (d) b(x) = 9x2 + 4

27
CHAPTER 2. FUNCTIONS REVIEW

Solution for Problem 2.4:

(a) We can input any real number to f, so the domain of f is all real numbers. It appears that any
real number can be output from f. To show that the range of f is all real numbers, we show that
for any real number y, we can find an x such that ∕(x) = ι∕. Since ∕(x) = 2x - 3, we seek the value
of a such that 2x - 3 = y. Solving for x gives x = (y + 3)∕2. So, for any real number y, if we let
a = (y + 3)∕2, then we have ∕(x) = y. This tells us that every real number is in the range of ∕.
Therefore, both the domain and range are all real numbers. Remember, we can use the symbol R
to indicate the real numbers, so we can write, "The domain and range of ∕ are both iR."
(b) The square root of a negative number is not a real number, so we must have -2x + 7 ≥ 0. This
gives us x ≤ 7/2, so the domain is all real numbers less than or equal to 7/2. In interval notation,
the domain is (-∞, 7∕2].
Intuitively, we might guess that because -2x + 7 can equal any nonnegative number, the range
of f is all nonnegative real numbers. We can explicitly show that this is the range of f as we did
in part (a). We let
y = ∕(χ) = √-2x + 7.
We then solve this equation for x in terms of y. Squaring both sides of y = V~2x + 7 gives
y2 = -2x + 7. We solve for x by subtracting 7 from both sides, then dividing by -2, to find
x = (7 - y2)∕2. So, for any nonnegative value of y, if we let x = (7 - y2)∕2, then we have

∕w=z(⅛^)∙ ∕^2(¼r)+7“√-7+y2+7 = ∕7 = y-
Γhis final step, ∖[if - y, is only valid because y is nonnegative. We therefore see that all nonnegative
real numbers are in the range of/, because for every nonnegative value of y, we can find an x such
that f(x) y. Since √-2x + 7 cannot be negative, no negative numbers are in the range of/. So,
the range of f is all nonnegative real numbers, which we can write as [0, +∞).
(c) We cannot have t = 1 in g(t) = 2t∕(t -1), since this would make the denominator equal to 0. There
are no other restrictions on the input to g, so the domain is all real numbers except 1.
Finding the range of g is a little trickier. We let y = g(f), so we have
2t

Then we solve for t in terms of y. Multiplying both sides by t - 1 gives yt - y = 2t. Solving this
equation for t in terms of y gives

For any desired output y except y = 2, we can use this equation to find the input, t. to g(f) that will
produce the desired output. Therefore, the range is all real numbers except 2. We can denote "all
real numbers except 2" with interval notation as (-∞, 2) U (2, ÷⅛χ>).
Γhe "U" in ( oo, 2) U (2, +oo) means "or," so, "y ∈ (-oo, 2) U (2, +<■»)" means

ι∕ is in the interval ( ∞, 2) or the interval (2, +∞).

28
 2.1. FUNCTION BASICS

Sidenote: The union of two intervals consists of all numbers that are in either one
or both of the intervals. As we just saw, we use the symbol U to denote
a union of two intervals, so (-∞, 2) U (2,+∞) means "all numbers in the
interval (-∞, 2) or the interval (2, +∞)."
The intersection of two intervals consists of all numbers that are in both
of the intervals. We use the symbol ∩ to refer to the intersection of two
intervals. For example, the numbers that are in both the interval [3,7] and
[5,11] form the interval [5,7], so we can write

[3,7] ∩ [5,11] = [5,7].

On the other hand, the numbers that are either in the interval [3,7] or
[5,11] or both form the interval [3,11]. Therefore, we have

[3,7] U [5,11] = [3,11].

We also have a special notation for "all real numbers except a few specific values." We can
write "all real numbers except 2" as R∖{2}. If we wish to exclude several specific values, we just
list them inside the curly braces. So, we write "all real numbers except 5, 6, and 7" as 1R∖{5,6,7}.
(d) We can input any real number to the function h(x) = 9x2 + 4, so the domain of h is IR. Because
the square of a real number is always nonnegative, the expression 9x2 is always nonnegative.
Therefore, the expression 9x2 + 4 must be greater than or equal to 4. Since 9x2 + 4 can equal any
number that is greater than or equal to 4, the range of h is [4, +∞).
□
Note that in parts (b) and (c), we find the domain by considering operations that we are not able to
perform. Specifically, we cannot take the square root of a negative number in a real function, and we
cannot divide by zero. This is typically how we find the domain when it is not obvious, as it is in parts
(a) and (d), that the domain is all real numbers. Finding the range is also often a matter of considering
the possible outputs of special expressions like square roots, perfect squares, or absolute values. But, as
shown in part (c), the restrictions on the range can also be more subtle. There, we found a strategy that
is often useful for determining the range of complicated functions:

Concept: We can often find the range of a function by setting the function definition
Q) -τ3c) equal to a new variable, then solving for the dummy variable in terms of
the new variable.

Sometimes functions are defined with explicit constraints on the domain of the function. The range
of such a function consists only of those outputs that can be obtained from the permitted inputs. Here's
an example:

Problem 2.5: Suppose p(x) = 2λ∕x + 3, but that p(x) is only defined for 4 ≤ x ≤ 9. Find the range of p. ∣

Solution for Problem 2.5: What's wrong with this solution:

29
CHAPTER 2. FUNCTIONS REVIEW

Bogus Solution: The output of √x can be any nonnegative real number, so the result
of 2 √x+3 can be any nonnegative real number greater than or equal
to 3.

Our Bogus Solution completely ignores the restriction on the domain of p.

WARNING!! We must take into account any restrictions on the domain of a function
A when finding the range of the function.

Since p is only defined for inputs x such that 4 ≤ x ≤ 9, the range consists of all values of p that
can result from inputting a value of x such that 4 ≤ x ≤ 9. Because 4 ≤ x ≤ 9, we have 2 ≤ √x < 3.
Multiplying all parts of this inequality chain by 2 gives 4 ≤ 2 χ∕x ≤ 6 and adding 3 to all parts of this
chain gives 7≤2√x + 3≤9. Therefore, for all x such that 4 ≤ x ≤ 9, we have 7 < p(x) < 9.
We're not finished yet! We've only shown that all values in the range of p are between 7 and 9,
inclusive (which means 7 and 9 are included). We haven't shown that all real numbers from 7 to 9 are
in the range. This is somewhat "obvious," since p(x) goes "smoothly" from 7 to 9 as x goes from 4 to 9.
I lowever, to explicitly prove that all possible values from 7 to 9 are in the range, we let y = p(x) = 2 χ∕x + 3
and solve for x in terms of y. This gives us

For any value of y from 7 to 9, we can use this equation to find the value of x for which p(x) equals this
y. (Make sure you see why all the resulting values of x are in the domain of p!) □
Finding the domain and range of more complicated functions can be pretty tricky. We'll try finding
the domain of a couple of such functions.

Problem 2.6: Find the domain of each of the following functions:

(b) 2f- 4
g<0 = ΓΓ÷
7 3Z-4

Solution for Problem 2.6:

(a) The input to a square root must be nonnegative, so we must have 2x - 5 ≥ 0. This gives us x ≥ 5/2.
However, we cannot have x = 3, since this will make the denominator equal to 0.
WARNING!! Sometimes there is more than one constraint on the domain of a func-
V tιon.
Combining these two constraints tells us that the domain is all real numbers greater than or equal
to 5/2 except 3. We can write this in interval notation as [5/2,3) U (3, +<∞).
(b) First, we notice that we cannot have t = 0 or 3t - 4 = 0, because we cannot divide by t)∙ This tells
us that we must exclude ∕ 0 and t ■ 4/3 from the domain I lowever, we're not finished, because
we cannot have the entire denominator of χ(t) equal to 0. Γheretorv, we must exclude from our
domain those values of t for w hich

30
 2.1. FUNCTION BASICS

Multiplying both sides by ∕(3∕ - 4) to get rid of the fractions gives us 3t 4 f = 0, from which
we find ∕ = 2. So, the domain of y is all real numbers except 0, 4/3, and 2. We could write this as
intervals with the unwieldy (-oo,0) U (0, U (∣,2) U (2z +∞). We can also write it as R∖∣0, ^,2),
which reads "all real numbers except 0, ∣, and 2."

x2-x _ (χ - l)(χ) v2 - X
Problem 2.7: Notice that = x. Are f(x) = ------- and σ(x) = x the same function?
x-1 x-1 x -1

Solution for Problem 2.7: No! The function f is not defined for x = 1, even though we can simplify our
expression for ∕(x) to just x. The simplification

x2 - x _ x(x - 1)
x-1 x-1

is only valid if x ≠ 1. If x = 1, then y5f is g, which is not defined. In other words, 1 is not in the domain
of f, but it is in the domain of g. So, f and g are not the same function. □
We finish this section with the relatively obvious observation that the sum, product, and quotient of
two functions are all themselves functions. For example, if f and g are functions, we can define a new
function h ash = f + g, where this means that ∕t(x) = ∕(x) + g(x). Note that ∕∕(x) is only defined if both
∕(x) and g(x) are defined. Therefore, a number must be in the domains of both f and y in order to be in
the domain of f + g.
In much the same way, we can define a new function as a linear combination of any two functions f
and g. Specifically, if α and b are constants, then we can define a new function t as t = α ■ f + b ■ gr where
this means that f(x) = α∙ ∕(x) + b ■ g(x).
We can also define the function p = f ∙ g such that p(x) = ∕(x) ∙ g(x) and the function q = f ∕g such
that q(x) = f(x)∕g(x). Note that if q = f∕g, then we must exclude values of x for which g(x) - 0 from the
domain of q.

*e
txer<i

2.1.1 Let ∕(x) = x2 - x - 6.

(a) Compute ∕(2). (d) Is there a real value of x such that ∕(x) = 6?
(b) Find all x such that ∕(x) = 0. (e) ★ Find the range of f. Hints: 134
(c) Find f(x - 1)

2.1.2 Find the domain and range of each of the following real-valued functions:

(a) ∕(x) = ∣1 - x∣ (b) ∕(t) = √2→ (c) ∕ι(x) = 1 - x2 (d) g(u) = -Jτ

2.1.3 Let ∕(x) = V4 - x and g(x) = x∣2x - 6. Find the domain of / ∙ g and the domain of f∣g. Explain
why these domains are different.

31
CHAPTER 2. FUNCTIONS REVIEW

∕2χ - 5 √2λ - 5
2.1.4 Do the functions ∕(x) x∕ — and y(x) = —- ---- r∙ have the same domain? Why or why not?
V x —8 , √χ- 8

2.1.5 Let ∕(x) = ~"~γ∙ Suppose ∣x∣ ≠ 1. Simplify f(x) ∙ f(-x).

2.1.6 A function g(x) is defined for -3 < x < 4 such that g(x) = (2 + x)2. Find the range of g.
2.1.7 Let T(a, b, c) = 3ab - c.

(a) Find T(2,3,-5).

(b) Find all values of x such that T(x,2,6) = 21.

2.2 Graphing Functions

When we graph an equation in which the only variables are x and/or y, we 4

I
plot all points (x,y) on the Cartesian plane that satisfy the equation. When we
graph a function f, we graph the equation y = ∕(x). For example, the graph
of the function f(x) = x2 is shown at right. ∕
∕
The x*
intercepts of a graph are the points where the graph crosses the x-
axis, while the y-intercepts are the points where the graph crosses the y-axis.
I or example, the point (0,0) is both an x-interceptand a y-intercept of the graph
of f (x) ∖ (Note: Some sources use "x-intercept" to refer to the x-coordinate
I
of a point where a graph crosses the x-axis, and likewise for "y-intercept." So, these sources would
describe both the x-intercept and the y-intercept of the graph of ∕(x) = x2 as 0 instead of (0,0).)
In this section, we review the basics of graphing functions, getting information from the graph of a
function, and transforming the graph of a function.

Problems

Problem 2.8: For each of the functions below, graph the function and find the x-intercepts and the
y-intercepts of the graph.

(a) f(χ) - (x - 6)2 - 4 (c) ∕(x) = √x + 5 - 3

(b) ∕(x) ≡ 4 (d) ∕(x) = ∣x + 2∣ + 1

Problem 2.9: I he graph of y = ∕(x) is pictured at right. y

(a) According to the graph, what is /(2)?
(b) According io the graph, for what values of a is ∕(rt) = 2?

32
 2.2. GRAPHING FUNCTIONS

Problem 2.10:
(a) Is it possible for a vertical line to intersect the graph of a function at two different points? Why
or why not?
(b) If there is no vertical line that passes through more than one point of a curve on the Cartesian
plane, then is the curve the graph of a function?

Problem 2.11: At right is the graph of y = ∕^(x). y

(a) Name three points on the graph of y = ∕(x) + 2.

(b) Explain why the graph of y = f(x) + 2 is a 2-unit vertical upward ∖
∖
shift of the graph of y = ∕(χ).
∕
(c) Name three points on the graph of y = ∕(x + 2).
A 7 * I

(d) Explain why the graph of y = f(χ + 2) is a 2-unit leftward shift of

the graph of y = f(x).
(e) Find the domain and range of h if ∕7(x) = ∕(x + 3) - 4. 1

Problem 2.12: Let ∕(x) be the function graphed in the previous problem.
(a) Draw the graph of y = -∕(x).
(b) Draw the graph of y = ∕(-x).
(c) Draw the graph of y = 2∕(x).
(d) Draw the graph of y = ∕(2x).

Problem 2.8: For each of the functions below, graph the function and find the x-intercepts and the
y-intercepts of the graph.
(a) ∕(x) = (x-6)2-4 (c) ∕(x) = Vx^+^5 - 3
(b) ∕(x) = 4 (d) ∕(x) = ∣x + 2∣ + l

Solution for Problem 2.8:

(a) The graph of y = (x - 6)2 - 4 is an upward-opening parabola with vertex (6, -4). (We will discuss
parabolas in much more detail in Chapter 5.) We find several points on the graph by finding y for
various values of x. A table with these points is at left below, and the graph is at right below.

X y = ∕(χ)
3 5
4 0
5 -3
6 -4
7 -3
8 0
9 5

33
CHAPTER 2. FUNCTIONS REVIEW

The y-intercept of the graph is the point where the graph intersects the y-axis. All points on the
ι∕-a∖is have an x-coordinate of 0, so evaluating ∕^(0) gives us the y-coordinate of the y-intercept of
the graph of y = ∕(x). Since ∕(0) = (0 - 6)2 - 4 = 32, the y-intercept of the graph is (0,32).
The a-intercepts of the graph are those points where the graph intersects the x-axis. The y-
coordinate of such a point is 0, so the x-coordinates of the x-intercepts are the solutions to the
equation ∕(x) = 0. Therefore, we seek the values of x such that (x - 6)2 - 4 = 0. Expanding and
rearranging gives us x2 - 12x + 32 = 0. Factoring gives (x - 4)(x - 8) = 0, so our solutions are x = 4
and x = 8. Therefore, the x-intercepts of our graph are (4,0) and (8,0). (Notice that we could also
read these off our graph.)
(b) Since ∕(x) - 4, the graph of y = ∕(x) is the graph of y = 4, which is a y
♦ I I iL -
horizontal line as shown at right. This line meets the y-axis at (0,4), so Γ
...

this is the y-intercept. The graph never hits the x-axis, so it does not have _ IJ
i
an x-intercept.

1 1

(c) As with the first part, we choose values of x, and then find y such that y = ∕(x). Notice that we
start with x = -5 and then substitute greater values of x, because the domain of f is [-5, +oo).
'∕

X y ≡ f(×)
-5 -3
-4 -2
-1 -1
4 0
11 1

When x - 0, we have y = ∕(0) = -3 + λ∕5, so the y-intercept is (0, -3 + ∖∕5). When y = 0, we have
0 √a + 5 - 3. Adding 3 to both sides gives 3 = √x + 5, and squaring both sides of this gives
9 - x + 5. Solving this gives x = 4, so (4,0) is the x-intercept of the graph.
(d) If x > -2, we have ∣x+2∣ = x+2, so if x > -2, we have∕(x) - x+2+1 = x+3.
So, when graphing y ≡ ∕(x), we graph y = x + 3 for x > -2. Similarly, if
x < -2, we have ∣x + 2∣ = -x - 2, so graphing y = ∕(x) for x < -2 is the
same as graphing y = -x - 2 + 1 - -x - 1. The result is shown at right.
When x = 0, we have y = ∕(0) = 3, so (0,3) is the y-intercept of
the graph. When y = 0, we have 0 = ∣x + 2∣ -t- 1. Subtracting 1 gives
∣λ + 21 - -1, which has no solutions because absolute value is always nonnegative. So, the graph
has no a intercept. (We can also see this from our graph.)

Now that we've used the definition of a function to get information about the graph of the function,
let's try going the other direction—using a graph to get information about the function.

34
 2.2. GRAPHING FUNCTIONS

Solution for Problem 2.9:

(a) The graph of y = ∕(x) passes through (2,3), so we have ∕(2) = 3.

Important: If the point (a, b) is on the graph of y = ∕(x), then f(a) — b. Conversely,
f(a) = b, then (a, b) is on the graph of y = ∕(x).

(b) To determine the values of a such that ∕(fl) = 2, we locate the points on the graph of y = ∕(x) such
that y = 2. The x-coordinates of these points then are our desired values of fl. Since (0,2) is on the
graph of y = ∕(x), we have ∕(0) = 2.
There is another point on the graph with y-coordinate equal to 2, but it's not immediately ob­
vious what the x-coordinate of this point is. However, this point is on the line segment connecting
(2,3) to (3,0), so we can use the equation of the line through these two points to find the desired
x-coordinate. The slope of this line is (3 - 0)∕(2 - 3) = -3, so a point-slope equation of the line is
y - 0 = -3(x - 3), or y = -3x + 9. When y = 2 in this equation, we have x = 7/3, so the line passes
through (7/3,2). Therefore, we have ∕(7∕3) = 2, so the two values of a for which f(a) = 2 are a - 0
and a = r7/3.
□
Problem 2.10:
(a) Is it possible for a vertical line to intersect the graph of a function at two different points? Why
or why not?
(b) If there is no vertical line that passes through more than one point of a curve on the Cartesian
plane, then is the curve the graph of a function?

Solution for Problem 2.10:

(a) For each value of x in the domain of f, there is exactly one value of ∕(x). So, for each such value of
x, there is only one point (x, y) on the graph of y = ∕(x). Therefore, it is impossible for the graph
of a function to pass through two points with the same x-coordinate. Because the equation of anv
vertical line has the form x = fl for some constant a, this means that it is impossible for a vertical
line to pass through more than one point on the graph of a function.
(b) If there is not a vertical line that passes through more than one point of a graph, then for every
value of x, there is at most one value of y such that (x, y) is on the graph. In other words, for every
x for which there is a point (x,y) on the graph, there is exactly one corresponding value of ι∕, so
the graph does indeed represent a function.
□
35
CHAPTER 2. FUNCTIONS REVIEW

Important: A curve on the Cartesian plane is the graph of some function if and only
ψ it every vertical line passes through no more than one point on the curve.
We call this the vertical line test.

Problem 2.11: At right is the graph of y = ∕(x).

(a) Explain why the graph of y = ∕(x) + 2 is a 2-unit vertical upward
shift of the graph of y = f(χ).
(b) Explain why the graph of y = f(x + 2) is a 2-unit leftward shift of
the graph of y = ∕(x).
(c) hind the domain and range of h if ∕∕(x) = f(x + 3) - 4.

Solution for Problem 2.11:

(a) Suppose x = a, where a is in the domain of f. Then, the point (a,f(a))

is on the graph of y = ∕(x) and the point (a,f(a) + 2) is on the graph
of y = ∕(x) + 2. Therefore, every point of the graph of y = ∕(x) + 2 is 2
units above the corresponding point on the graph of y = ∕(x). So, the
graph of y = ∕(x) + 2 is a 2-unit upward shift of the graph of y = ∕(x).
The graph of y = ∕(x) + 2 is solid in the graph at right, and the graph
of y = ∕(x) is dashed.
Similarly, if k is positive, then the graph of y = ∕(x) + k is a fc-unit
upward shift of the graph of y = ∕(x), and the graph of y = ∕(x) - k is
a A-unil downward shift of the graph of y = ∕(x).

(b) Suppose a ∣ι, where a is in the domain of f. Then, the point (fl,∕(α))
is on the graph of y = ∕(x) and the point (λ, f(a + 2)) is on the graph of
y f(a + 2) Unfortunately, it isn't so clear how f(a) and f(a + 2) are
related. Γo get a point on the graph of y = ∕(x + 2) with y-coordinate
equal to f(a)t we have to let x = a- 2. When x = α- 2iny = ∕(x + 2),
we have y = ∕(λ), so (a - 2,f(a)) is on the graph of y = ∕(χ + 2).
We've shown that for any a in the domain of f, the point (a,f(a))
is on the graph of y = ∕(x) and the point (a - 2, f(a)) is on the graph
of y = ∕(x + 2). Therefore, every point on the graph of y = ∕(χ + 2) is
2 units to the left of the corresponding point on the graph of y = ∕(χ).
So, the graph of y = f(x + 2) is a 2-unit leftward shift of the graph of y = f(x).
y = f(χ + 2) is solid in the graph at right above, and the graph of y = ∕(x) is dashed.
Similarly, the graph of y = ∕(x - k), where k > 0, is a A-unit rightward shift of the graph of
y ≡ ∕(r).
WARNING!! Λ common mistake is to think that the graph of y = f(x + k) is a
A rightward shift of the graph ot ι∕ ∕(x) when k ∣~∙ positive. This is
incorrect; make sure you see why.

36
 2.2. GRAPHING FUNCTIONS

(c) The graph of y = ∕(x ÷ 3) 4 is the result of shifting the graph of y = ∕(x) downward 4 units and
leftward 3 units. Γhe leftward shift means the domain of lι is 3 units "to the left" of the domain of
f. The domain of f is ∣ -5,3], so the domain of lι is [-8,0]. The downward shift means the range
of h is 4 units "below" the range of f. The range of f is [-3,4), so the range of h is [~7,0∣.

In Problem 2.11, we learned the effect of adding a constant to the input or to the output of a function:

Important: When k > 0, the graph of y = ∕(x) + k results from shifting the graph
kl ? y - f(χ) vertically upward by k units, while the graph of y = ∕(x + k)
results from shifting the graph of y - ∕(x) horizontally to the left by k
units. Similarly, the graph of y = ∕(x) - k is a A-unit downward shift of
the graph of y = ∕(x) and the graph of y = ∕(x - k) is a A-unit rightward
shift.

Problem 2.12: At right is the graph of y = ∕(x).

(a) Find the graph of y = -∕(x).
(b) Find the graph of y = f(-x).
(c) Find the graph of y = 2∕(x).
(d) Find the graph of y = ∕(2x).

Solution for Problem 2.12:

(a) As before, we let x = a, where a is in the domain of f. Then, the point y
(a, f(a∖) is on the graph of y = ∕(x) and the point (a, -/(a)) is on the
graph of y = - ∕^(x). So, the y-coordinate of each point on the graph
of y = -∕(x) is the opposite of the y-coordinate of the corresponding
point on the graph of y = ∕(x). Therefore, we form the graph of
y = -∕(x) by reflecting the graph of y = ∕(x) over the x-axis. At right,
the graph of y = -∕(x) is solid and the graph of y - f(x) is dashed.

Extra! The word algebra is derived from the works of the ninth century Arabic mathematician
*'" ι* n⅜ -∣ ⅜ Abu ∫a'far Muhammad ibn Musa al-Khwarizmi. His book al-Kitab al-mιιkhtasar fi hisab
al-jabr wa'l-muqabala classified quadratic equations and discussed approaches to solve
them. However, like many early mathematicians, all of al-Khwarizmi's "algebra" was
rhetorical and geometric, rather than symbolic. For example, he described the equation
x2 + 10x = 39 as "One square and ten roots are equal to 39 units." Imagine doing all the
mathematics in this book with words rather than symbols!

37
CHAPTER 2. FUNCTIONS REVIEW

y
(b) We ∣el v a, where a is in the domain of f. Then, the point (a, f(a))
is on the graph of y = ∕(x) and the point (a,f(-a)) is on the graph of
y ≡ f(- v). We can find a point on the graph of y = /(—x) with f(a) as
the y-coordinate by letting x = -a, which tells us that (-arf(a')) is on
the graph ot y = ∕(-x). Therefore, the graph of y = ∕(-x) is the result
of reflecting the graph of y = ∕(x) over the y-axis. At right, the graph
of y = f(-x) is solid and the graph of y = ∕(x) is dashed.

Important: The graph of y = -∕(x) is the result of reflecting the graph of y = ∕(x)
over the x-axis.
The graph of y = ∕(-x) is the result of re∩ecting the graph of y = ∕(x)
over the y-axis.

(c) For each point (a, f (a)) on the graph of y = ∕(x), the point (a, 2f(a)) is on
the graph of y = 2∕(x). The latter point is twice as far from the x-axis
as the former. Similarly, each point on the graph of y = 2∕(x) can be
found by doubling the distance from the x-axis of the corresponding
point on the graph of y = ∕(x). Therefore, the graph of y = 2∕(x) can
be found by scaling the graph of y = ∕(x) vertically by a factor of 2.
Λl right, the graph of y = 2∕(x) is solid and the graph of y = ∕(x) is
dashed.
’ I

(d) l or each point (2a,f(2a)) on the graph of y = ∕(x), the point (a,∕(2π)) is y
on the graph of y = ∕(2x). The latter point is half as far from the y-axis
as the former. Similarly, each point on the graph of y = ∕(2x) can be
found by halving the distance from the y-axis of the corresponding
point on the graph of = f(x). Therefore, the graph of y = f(2x) can be
found by scaling the graph of y = ∕(x) horizontally by a factor of I/2.
At right, the graph of y = ∕(2x) is solid and the graph of y = ∕(x) is
dashed.

Important: When A > 0, the graph of y = kf(x) results from scaling the graph of
y = f(x) vertically by a factor of k, while the graph of y = f(kx) results
from scaling the graph of y = ∕(x) horizontally by a factor of 1∕A.

WARNING!! I he graph of y f(kx) is a horizontal scaling of the graph of y = ∕(x)

V by a factor of l∕A, not by k. So, the graph ot y ∕(3.γ) is the result of
compressing the graph of y ∕( t> towaids the y-axis, not stretching
it away from the y-axis.

38
 2.3. COMPOSITION

2.2.1 Graph each of the following, and find the x-intercepts and (/-intercepts of each graph.
(a) ∕(x) = 3x-7 (b) ∕(x) = 2-∣x∣ (c) ∕(x) = x2 - 5x ÷ 6

2.2.2 Determine if each of the following graphs represents a function:

2.2.3 Suppose ∕(9) = 2. For each of the parts below, find a point that must be on the graph of the given
equation.

(a) y = ∕(x-3) + 5 (b) y = 2∕(x∕4) (c) y = 2∕(3x-l) + 7

*
2.2.4 Suppose ∕(4) = 8. Note that the point (2,8) is on the graph of y = ∕(2x) and that the point (5,8)
is on the graph of y = ∕(2x - 6). So, the point we have found on the graph of y = ∕(2x - 6) is 3 units to
the right of the point we found on the graph of y = ∕(2x). Which of the following is true: (a) we made a
mistake; (b) it's a coincidence—not every point on y = ∕(2x - 6) is 3 to the right of a point on y = /(2x);
(c) the graph of y = ∕(2x - 6) is a 3-unit rightward shift of the graph of y = ∕(2x). Explain your answer.
Hints: 285
Aλ ■?-

2.3 Composition

Sometimes we want to hook two (or more) machines together, taking the output from one machine and
putting it into another. When we connect functions together like this, we are performing a composition
of the functions. For example, the expression ∕(g(x)) means we put our input x into function g, and
then take the output, g(x), and put that into function f. A composition of functions is also sometimes
indicated with the symbol o. For example, when we write h = f o g, we define the function h such that
h(x) = f(g(χ))-

Problems

Problem 2.13: Suppose ∕(x) = 3x + 7 and g(x) = 2 Vx - 3.

(a) Find f(g(3)) and g(∕(3)).
(b) Find fl if ∕(g(rt)) = 25.
(c) Is g(∕(-7)) defined? Why or why not?
(d) Suppose that lι = g o f. What is the domain of //?

39
CHAPTER 2. FUNCTIONS REVIEW

Problem 2.14:
(a) Let ∕j(.v) f(χ(>)), where f and y are functions, and suppose ∕∕(x) is defined for all values of x in
the domain of £. Explain why the range of g must be part of the domain of f.
(b) Again, let ∕ι(x) = ∕(g(x)), where f and g are functions. Suppose every value in the range of g is
in the domain of f. Must the domain of h be the same as the domain of g?

Problem 2.15: Let f(a) = a-2 and F(afb) = b2 + a.

(a) Compute ∕(4).
(b) Find F(3,∕(4)). (Source: AHSME)

Problem 2.16:
(a) Let f(x) ≈ 3x -1. Find ∕(∕(x)) and ∕(∕(∕(x))).
(b) Suppose g(x) - ax + b, where a and b are real constants. Find all possible pairs (a,b) such that
g(g(x)) = 9x + 28.

Problem 2.17: Let ∕(x) = √x and g(x~) = x2. Is it true that ∕(g(x)) and g(∕(x)) are the same function?
Why or why not?

Problem 2.13: Suppose ∕(x) = 3x + 7 and g(x) = 2 √x - 3.

(a) Find f(g(3)) and χ(∕(3)). I
(b) Find a if f(g(a)) = 25. I
(c) Is g(f (-7)) defined? Why or why not? I

(d) Suppose that h = g o f. What is the domain of h?

Solution for Problem 2.13:

(a) We have y(3) = 2 y∕3~ 3 = 0, so ∕(g(3)) = ∕(0) = 3(0) + 7 = 7. Similarly, we have ∕(3) = 3(3) + 7 = 16,
,,. ,,√ι. = >i gOQ. ¾Vlt-3g2√i3- Notice that∕⅛(3)) ≠ g(∕(3)).

WARNING!! If ∕(x) and g(x) are functions, then the functions f(g(x)) and g(∕(x))
*λ are not necessarily the same the same function.
(b) We have
∕(X(λ)) = ∕(2 √λt3) = 3(2 √a^3) + 7 = 6 √T^3 + 7,

so f(g(a)) ≡ 25 means
6 √Γr3 + 7 = 25.

Subtracting 7 from both sides and then dividing both sides by 6 gives √u - 3 = 3. Squaring both
bides gives us a - 3 9, son ^ 12. Checking our answer, we find f(y(12)) - ∕(6) = 25, as expected.
(c) We have ∕(-7) 3( 7) + 7 14, so χ(∕( 7)) χ(-14). However, the value -14 is not in the
domain of g. Since g( 14) is not defined, χ(∕(-7)) is not defined.

40
 2.3. COMPOSITION

(d) Since ∕z(x) = g(∕(x)), in order for x to be in the domain of lι, both ∕^(x) and g(∕(.r)) must be defined.
Tlιe domain of f is all real numbers; however, as we saw in the previous part, g(∕(.v)) is not defined
for all real values of x.
In order for g(∕(.v)) to be defined, the value of ∕^(x) must be in the domain of g. The domain of g
is [3, +∞). Therefore, in order for g(f(x)) to be defined, we must have ∕(.v) > 3. So, we must have
3x + 7 ≥ 3, which gives us x > -4/3. For all values of x such that x > -4/3, the expression g(∕(x))
is defined. For any value of x less than -4/3, the expression g(f(x)) is not defined. Therefore, the
domain of h is [—4/3, +∞).

Problem 2.14:
(a) Let ∕ι(x) = ∕(g(x)), where f and g are functions, and suppose ∕z(x) is defined for all values of x
in the domain of g. Explain why the range of g must be part of the domain of f.
(b) Again, let ∕t(x) = ∕(g(x)), where f and g are functions. Suppose every value in the range of g is
in the domain of f. Must the domain of h be the same as the domain of g?

Solution for Problem 2.14:

(a) Since ∕∕(x) = ∕(g(x)) and ∕ι(x) is defined for all values of x in the domain of g, the expression ∕(g(x))
must be defined for all values of x in the domain of g. Because ∕(g(x)) is defined, the value of g(x)
is in the domain of f. So, every number in the range of g must be in the domain of f.
(b) The expression ∕(g(x)) is not defined if g(x) is not defined. Therefore, all values in the domain
of h must be in the domain of g. Conversely, if x is in the domain of g, then g(x) is defined.
Furthermore, we are given that every value in the range of g is in the domain of f, so f(g(x)) is
defined for all x in the domain of g. Therefore, every value in the domain of g is in the domain of
h. Since every value in the domain of g is in the domain of h, and vice versa, we conclude that the
domain of lι is the same as the domain of g.

□
Problem 2.15: If ∕(fl) =a-2 and F(a, b) = b2 + a, find the value of F(3, ∕(4)). (Source: AHSME)

Solution for Problem 2.15: We note that ∕(4) = 4 - 2 = 2, so

F(3,∕(4)) = F(3,2) = 22 + 3 = 7.

□
Although the expression F(3,∕(4)) looks a little intimidating, the process of evaluating it is just simple
plug-and-chug arithmetic.

WARNING!! Don't be intimidated by notation. Often the notation is a fancy way

V of representing very simple ideas.

We'll see the value of notation as we move to more complicated problems throughout this book.
However, we'll usually introduce notation with simple problems like Problem 2.15 so you can get used
to the notation before using it on harder problems.

41
CHAPTER 2. FUNCTIONS REVIEW

Problem 2,16:
(.ι) Lot ∕(Λ∙) - 3x - 1. Find ∕(∕(x)) and ∕(∕(∕(x)))-
(b) Suppose ι∖,(.v) - ax + bf where a and b are real constants. Find all possible pairs (a,b) such that
χ(χ(.v)) = 9x + 28.

Solution for Problem 2.16:

(a) We have
f2(x) = ∕(∕(x)) = f(3x - 1) = 3(3x - 1) - 1 = 9x - 4.
We can now use this result to find ∕3(x).∙

f3(x) = /(/(/(
)))
* = ∕(∕2(x)) = ∕(9x - 4) = 3(9x - 4) -1 = 27x - 13.

(b) In order to find a and b, we first need an expression for g2(x). We have

g2(x) = g(g(x)) = g(ax + b) = a(ax + b) + b = a2x + ab + b.

So, our equation g2(x) = 9x + 28 is now a2x + ab + b = 9x + 28. In order for this equation to be true
for all values of x, the coefficient of x on both sides must be the same and the constant terms on
both sides must be the same. Therefore, we have the system of equations

a2 = 9,
ab + b = 28.

From ιr 9, we have a = ±3. Letting fl = 3 in the second equation gives b = 7 and letting
a 3 in the second equation gives b = -14. So, our two possible pairs are (a,b) = (3,7) and
(α,t>) ≡ (-3,-14).

□
We have a special notation for repeatedly feeding the output of a function back into the function
itself. In this text, we write ∕2(x) to mean ∕(∕(x)). Likewise, fn(x) refers to applying the function ∕(x)
exactly n times; for example, ∕5(x) = ∕(∕(∕(∕(∕(
*
)))))∙

WARNING!! Some sources use f2(x) to mean∕(x)∙∕(x) instead of ∕(∕(x)). Likewise,

*ia f (x) is sometimes used to denote [∕(x)∣π. You'll often have to deter­
mine by context whether ∕n(x) refers to a function applied n times, or
to a function raised to the nth power.

Problem 2.17: Let ∕(x) = √x and g(.v) = x2 Is it true that f(g(x)) and g(∕(x)) are the same function?
Why or why not?

Solution for Problem 2.17: At first glance, we might be tempted to reason as follows:

Bogus Solution: We have f∖χ(x)) V< ι and y(∕(v)) (√x)2 = x, so f(χ(x)) and
λ,(∕( v)) are the same.
, . - - ... ................................................................ ............. ................... ..... _____________ _ __ _________ _____________ ______________________ 1

42
 2.4. INVERSE FUNCTIONS

This Bogus Solution contains two significant errors. First, the domain of/is all nonnegative numbers,
so the domain of g(∕(x)) is all nonnegative numbers. However, because the domain of g is all reals while
the range of g is all nonnegative numbers, the expression ∕(g(x)) is defined for all real x. Therefore, the
domain of ∕(g(x)) is all real numbers. So, the two functions g(∕(x)) and ∕(g(x)) have different domains,
which means that the functions are not the same. For example, g(∕(-l)) is undefined, because ∕(-l) is
undefined, but ∕(g(-l)) = ∕(1) = 1.

The other error occurs when we write Vx2 = χ. This statement is only true if x ≥ 0. If x < 0, then we
have Vx2 = -x. So, it is not true that ∕(g(x)) = x for all x. If x ≥ 0, then ∕(g(x)) = x, and if x < 0, then
∕(g(λ')) = -x∙ We can capture both of these by writing ∕(g(x)) = ∣x∣. □

2.3.1 Check the answer to part (b) of Problem 2.16 by evaluating g(g(x)) for g(x) = 3x + 7 and for
g(x) = -3x - 14.

2.3.2 Let ∕(x) = x2 - 2x and g(x) = √1 - χ.

(a) Find ∕(g(x)).
(b) Find the domain of ∕(g(x)).

2.3.3 Are both ∕(∕2(x)) and ∕2(∕(x)) the same as /3(x)?

2.3.4 Let a be a constant, and let ∕(x) = ax. Find an expression for fn(x) in terms of a and π, where n is
a positive integer.
2.3.5 Suppose ∕z, ∕, and g are functions such that ∕ι(x) = ∕(g(x)). If (3,0) is the only x-intercept of the
graph of g, then can we determine the x-intercepts or y-intercepts of h?

2.4 Inverse Functions

You're probably very familiar with using Control-Z when typing on a computer: that's the "undo"
command for many applications. Being able to "undo" the work of a function can be extremely useful.
The machine that "undoes" the work of a function ∕ is called the inverse function of ∕.

Definition: Function g is the inverse of function ∕ if and only if

g(∕(x)) = x for all values of x in the domain of ∕, and

∕(g(x)) = x for all values of x in the domain of g.

From our definition, we see that if g is the inverse function of ∕, then ∕ is the inverse function of g.
We often write the inverse of f as ∕^1.

WARNING!! When working with functions, f 1 does not mean ∣! The expression
* ∕^1 is a special notation that denotes the inverse of the function ∕.

43
CHAPTER 2. FUNCTIONS REVIEW

Problems

Problem 2.18: In this problem we find the inverse of the function ∕(x) = 3x - 5.
(a) Let χ be the inverse of f, so that we have f(g(x)) = x and g(f(x)) = x∙ Which of these equations
is easier to use to find χ(x)7
(b) Use the definition of f to solve the equation ∕(y) = x for y.
(c) Whatisχ(x)7
(d) Check your answer by confirming that χ(∕(x)) = x.

Problem 2.19: Let∕(x) = ——-.

3 - 2x
(a) Find ∕-1 (7), without finding ∕^1(x).
(b) Find ∕~,(x).

Problem 2.20:
(a) Docs the function χ(x) = 2x2 - 9 have an inverse? If it does, find it. If it doesn't, explain why it
doesn't.
(b) Dot s the function h(x) = 3 √2x + 1 have an inverse? If it does, find it. If it doesn't, explain why
it doesn't. (Be careful on this part!)

Problem 2.21:
(a) 1 xplain how to determine from the graph of a function whether or not the function has an
inverse.
(b) I low are the graph of a function and the graph of its inverse related?

Problem 2.22: Suppose functions∕ and g both have an inverse. Show that if h = f ogr then the inverse
of h is χ^, o ∕~,.

Problem 2.18: Find the inverse of the function ∕(x) = 3x - 5.

Solution for Problem 2.18: Suppose g is the inverse of f. Then, we must have

g(f(×)) ≈ f(g(χ)) = χ.

The equation f(g{x)) = x is more helpful, because we know ∕(χ). So, we can put χ(x) into ∕(x), which
gives us ∕(χ(x)) = 3χ(x) - 5. Therefore, we must have 3g(x) - 5 = χ.

Important: Just as we can sometimes solve an equation for a variable by isolating the
CΓ7 variable, we can sometimes solve an equation for a function by isolating
*' that function.

We isolate χ(x) by adding 5 to both sides, then dividing by 3. This gives us χ(x) - ly. We can check our

44
 2.4. INVERSE FUNCTIONS

work by confirming that ∕(g(x)) = g(∕(x)) = X∙ Therefore, we have f 1(x) = a⅛p.□

Note that we found the inverse of f by solving the equation ∕(g(x)) = x for g(x). Make sure you see
why this produces a function g that is the inverse of f.

Important: If f has an inverse g, then we can often find that inverse by solving the

ψ equation ∕(g(x)) = x for g(x).

If you find the idea of "solving for a function" confusing, you can instead let y = g(x) and solve for
y. For example, when ∕(x) = 3x - 5, we can solve for y in ∕(y) = x to find the inverse of f. Since ∕(y) = x,
we have 3y - 5 = x. Solving for y gives y = so the inverse of f is ∕~1(x) =
-------- ---- — - - - - ------- -------- -. —
Problem 2.19: Let ∕(x) = ——-.
jv ’ 3-2x
(a) Find ∕^1 (7), without finding f~1 (χ).
(b) Find∕-1(x).

Solution for Problem 2.19:

(a) Suppose y = ∕-1(7). Because f~1 is the inverse of f, we have ∕(∕-1(x)) = x for all values of x.
Specifically, we have ∕(∕~1(7)) = 7, so because y = ∕-1(7), we have ∕(y) = 7. We also could have
deduced ∕(y) = 7 by noting that y = ∕-1(7) means that the y is the input to f that produces 7 as an
output.

Important: If function f has an inverse, then f(a) = b means that f 1(b) = a, and vice
versa.

Since ∕(y) = 7, we have

-5-y
= 7.
3-2y
Multiplying both sides by 3 - 2y gives -5 - y = 7(3 - 2y). Solving this equation gives us y = 2.
Therefore, ∕-1(7) = 2.
(b) As we just learned, we can find the inverse of f by solving the equation ∕(y) = x for y in terms of
x. Using our definition of f, the equation ∕(y) = x becomes

3-2y

Multiplying both sides by 3 - 2y gives -5 - y = x(3 - 2y). Isolating the terms with y on the left
side gives 2xy - y = 3x + 5, so y(2x - 1) = 3x + 5. Dividing by 2x - 1 gives

3x + 5
r1ω = y =
2x-Γ

With this, we can check our answer to the first part by directly evaluating f 1 (7) = 2.
□

45
CHAPTER 2. FUNCTIONS REVIEW

We say that a function is invertible if it has an inverse. As you might have guessed, not all functions
are invertible.
Problem 2.20:
(a) Does the function g(x) = 2x2 - 9 have an inverse? If it does, find it. If it doesn't, explain why it
doesn't.
(b) Does the function h(x) = 3 √2x + 1 have an inverse? If it does, find it. If it doesn't, explain why
it doesn't.

Solution for Problem 2.20:

(a) Both g(1) and g(-l) equal -7. If g has an inverse, should this inverse return 1 or -1 when -7 is
input? We can't tell! Therefore, g is not reversible; it does not have an inverse.
Important: If a function gives the same output for two different inputs, then the
φ function does not have an inverse.

On the other hand, if a function never gives the same output for two different inputs, then the
function is reversible—we can tell from the output of the function what input produced that
output.
Important: If there are no two different inputs to a function that produce the same
output, then the function has an inverse.

(b) We start by looking for two different inputs to h that give the same output. After inputting a few
values to h, we notice that as we input higher and higher values of x, we get higher and higher
∖ alues of h(x). This makes us think that lι might indeed have an inverse, since it seems impossible
to find two different inputs that give the same output. So, we try to find the inverse of h by solving
the equation ∕t(y) = x for y. This gives us the equation

3 √2y + 1 = χ.

Squaring both sides gives us 9(2y + 1) = x2. Solving for y gives

= ≤.1
y 18 2'

This isn't the end of the story, however. Recall our definition of inverse functions. Functions f
and h are inverse functions of each other if and only if

⅛(∕(x)) = X for all values of x in the domain of fl and

∕(∕ι(x)) = x for all values of x in the domain of lι.

We let ∕(χ) - Our work above suggests that this is the inverse of h. However, it f is the
inverse of It, then lι is the inverse of ∕ So, we should have ∕ι(∕(v)) ∙- v for all x in the domain of
f. Unfortunately, we have ∕( 3) - 0, but ∕∕(∕( 3)) ∕ι(0) 3, so we don't have ∕ι(∕(x)) = x when
x = -3. The problem here is that the function ∕(ι) ∣rt ∙, does not have an inverse. We can see this
quickly by noting that ∕(3) - ∕(- 3) = 0. So, where did we go wrong?

46
 2.4. INVERSE FUNCTIONS

Let's evaluate ∕t(∕(x)) ar>d see if it really does equal x. Maybe this will shed some light on where
we went wrong. We have

⅛(∕(x)) = 3 √2∕(x) + 1 = 3 √≤ - 1 + 1

Here, we have to be careful.

WARNING!! We only have = χ if χ is nonnegative.

Now, we've found our problem. We only have Jz(∕(x)) = x if y∕x^ = x. This only occurs if x is
nonnegative. So, if we restrict the domain of f to nonnegative numbers, then we have h(f(x)) = x
for all values of x in the domain of f. We can quickly confirm that ∕(∕z(x)) = x for all x in the
domain of h as well. So, the inverse of h is ∕Γ1(x) = f∣ - ∖, where the function h~} is only defined
for nonnegative values of x.

Problem 2.21:
(a) Explain how to determine from the graph of a function whether or not the function has an
inverse.
(b) How are the graph of a function and the graph of its inverse related?

Solution for Problem 2.21:

(a) A function has an inverse if and only if it is reversible. That is, a function has an inverse if we
can tell from the output of the function what the input was. If f has an inverse, then when we
graph y = ∕(x), there can be no two points on the graph with the same y-coordinate. If there were,
then we would have two different inputs (the x-coordinates of points on the graph) that provide
the same output (the common y-coordinate) for the function, so it couldn't have an inverse. "No
two points on the graph have the same y-coordinate" means the same thing as "No horizontal
line intersects the graph in more than one point." Therefore, we have a horizontal line test for
whether or not a function has an inverse:
Important: A function has an inverse if and only if there does not exist a horizontal

ψ line that passes through more than one point on the graph of the function.

(b) Suppose the point (a, b) is on the graph of y = ∕(x). Then, we have ∕(α) = b, so f~1(b) = a. Because
f~1(b) = a, the point (b,a) is on the graph of y = ∕-1(x). Therefore, if (a,b) is on the graph of a
function, then (b, a) is on the graph of its inverse.

Extra! I've missed more than 9000 shots in my career. I've lost almost 300 games. 26 times, I’ve been
' ∣ *∙ i∣∣*∙' ∣*∙ trusted to take the game winning shot and missed. I've failed over and over and over again in my
"∣*
life. And that is why I succeed.
- Michael Jordan

47
CHAPTER 2. FUNCTIONS REVIEW

The fact that the points on the graph of y = f '1(x) can be found
by reversing the coordinates of the points on the graph of ι∕ = ∕(x)
has a geometric interpretation. We just learned that if (a,b) is on
the graph of y = f(x), then (b,a) is on the graph of y = ∕-1(x)∙
In the graph at right, we have plotted the points (5,6), (6,-3),
and (-7,1). We have also plotted the points that result when the
coordinates of these three points are reversed. We see that the new
points are the result of reflecting the old points over the line ι∕ = x.
The graph of y = χ is the dashed line in the diagram at right.
In general, reversing the coordinates of a point has the same
effect as reflecting the point over the graph of y = x. To prove
this, we let point P be (a,b) and Q be (b,a). The slope of PQ is -1, so PQ is perpendicular to the
graph of y - x. Furthermore, the midpoint of PQ is (2yz, ηr), which is on the graph of y = x- So,
the graph of y = x is the perpendicular bisector of PQ, which means that P and Q are images of
each other upon reflection over the graph of y = x. So, each point on the graph of y = f 1 (x) is
the mirror image of a point on the graph of y = ∕(x) when reflected over the line y = x (and vice
versa). Therefore, the entire graph of y = ∕^1(x) is the mirror image of the graph of y = ∕(x) when
reflected over the line x = y.
_—— — -
Important: If the function f has an inverse f 1, then the graph of y = f 1(x) is the
τ reflection of the graph of y = ∕(x) over the line x = ι∕.

∣ I
Problem 2.22: Suppose functions f and g both have an inverse. Show that if lι = f o g, then the I
inverse of h is g 1 o f ,.

Solution for Problem 2.22: Let t ≈ g l o f l. To show that t is the inverse of It, we must show that
h(t(x)) x and f(∕∕(x)) = x. We'll start with ∕z(f(x)). Using our definitions of h and I, we have

∕ι(f(x)) ≡ ∕(g(g^1(∕^1(x)))).

That looks pretty scary! But seeing both g and g~l right next to each other reminds us that g(g~1(x)) = x
for all ∖ alues of a. In other words, the output of g o g 1 is always the same as the input. So, when we
put ∕-1(x) into g o g-,, we get ∕^l(x) out, which means

- Γ1(χ)∙

So, using this in our expression for h(t(x)) above, we have

Λ(f(x)) = ∕⅛⅛^,(∕-l⅛)))) = ∕(∕-,(x)).

Because ∕ 1 is the inverse of f, we have ∕(∕"l(x)) = x, so h(f(x)) = χ.

We can do essentially the same thing tor ∕(h(x))∙ We start with

t(∕∕(x)) - g ,(∕ l(∕(g(A))))

48
 2.5. SUMMARY

Because f is the inverse of ∕^1, we have f~l (∕(g(x))) = g(x), so t(∕ι(x)) = g^1 (∕~1 (∕(g(x)))) = g~1 (g(x)) = x,
as desired. Because ∕∕(f(x)) = f(7ι(x)) = x when h = f o g and t = g~1 o ∕^1, the function g^1 o f~i is the
inverse of f o g. □

I
Ivercises

2.4.1 If ∕(x) = what is the value of f 1(3)7

2.4.2 Determine whether or not each of the following functions has an inverse. If the function has an
inverse, find it.
(a) ∕(x) = 2x-7 (c) ∕(x) = √2^x (e) ∕(x) = √4≡x + √xz2

(b) = 2x + 3 (d) = ∣5 - x∣ *
(0 = χ2 ~6χ + 3, where x ≥ 4

2.4.3 Describe all ordered pairs (α,b) such that the inverse of ∕(x) = ax + b satisfies ∕(x) = ∕^1(x) for
all x.

2.4.4 Let g(x) = where a, b, c, and d are positive and ad ≠ be. Which of ∣, and j cannot be in the
domain of g-1? Hints: 336

2.5 Summary

A function gives a single output for each input. The domain of a function consists of all the values we
are able to input to the function and get an output. Meanwhile, the range of the function consists of all
the values that can possibly come out of the function.
When we graph the function f, we graph the equation y = ∕(x) on the Cartesian plane. Therefore, if
the point («, b) is on the graph of y = ∕(x), then f(a) = b. Conversely, if f(a) = b, then («, b) is on the graph
of y = ∕(x).

Important: A curve on the Cartesian plane is the graph of some function if and only
∖∣∕ if every vertical line passes throughl no more than one point on the curve.
We call this the vertical line test.

Important: When k > 0, the graph of y - f(x) + k results from shifting the graph

ψ of y = ∕(x) vertically upward by k units, while the graph of y = ∕(x + k)

results from shifting the graph of y = ∕(x) horizontally to the left by k
units. Similarly, the graph of y - f{x) - k is a k-unit downward shift of
the graph of y = ∕(x) and the graph of y = ∕(x - k) is a ∕c-unit rightward
shift.

Important: The graph of y = -∕(x) is the result of reflecting the graph of y = ∕(x)

ψ over the x-axis. The graph of y = ∕( -x) is the result of reflecting the
graph of y = ∕(x) over the y-axis.

I
49
CHAPTER 2. FUNCTIONS REVIEW

I
Important: When k > 0, the graph of y = kf(x) results from scaling the graph of

ψ y = ∕(x) vertically by a factor of k, while the graph of y = f(kx) results

from scaling the graph of y = ∕(x) horizontally by a factor of ↑∕k.

When we perform a composition of two functions, we place the output from the first function into
the second. For example, both ∕(g(χ)) and (∕ o g)(x) indicate that we take the output g(x) and input this
result to f. We have a special notation for when we feed the output of a function back into the function
itself. In this text, we write ∕2(x) to mean ∕(∕(x)). Likewise, ∕7'(x) refers to applying the function ∕(x)
exactly n times; for example, ∕5(x) = ∕(∕(∕(∕(∕(x)))))∙
Finally, we have a special name for a pair of functions such that each function undoes the action
of the other:

Definition: Functions f and g are inverse functions of each other if and only if

g(∕(x)) = x for all values of x in the domain of f, and

∕(g(x)) = x f°r a∏ values of x in the domain of g.

If f is the inverse function of g, then g is the inverse function of f.

Not every function has an inverse. If a function gives the same output for two different inputs, then
(he function does not have an inverse. If there are no two different inputs to a function that produce the
same output, then the function has an inverse. If f has an inverse g, then we can often find that inverse
by solving the equation ∕(g(x)) = x for g(x).
We usually denote the inverse of the function f as ∕-1. If f has an inverse, then the graph of ∕~1 is
the reflection of the graph of f over the line y = x.

Important: A function has an inverse if and only if there does not exist a horizontal
line that passes through more than one point on the graph of the function.

Things To Watch Out For! ---------------------- ------------------------------------------

• When evaluating a function that accepts multiple inputs, make sure you assign the values to the
dummy variables in the correct order.
• We must take into account any restrictions on the domain of a function when finding the range of
the function.
• Sometimes there is more than one constraint on the domain of a function.

• A common mistake is to think that the graph of y ≡ ∕(x + k) is a rightward shift of the graph of
y ≡ f(x) when k is positive. This is incorrect; make sure you see why.
• The graph of y - f(kx) is a horizontal scaling ot the graph ot ι∕ ∕(v) by a factor of l∕fc, not by k.
So, the graph of y ∕(3x) is the result of compressing the graph of y = ∕(.v) towards the y-axis, not
stretching it away from the y-axis.

50
 REVIEW PROBLEMS

• Some sources use ∕"2(x) to mean ∕(x) ∙ ∕(x) instead of ∕(∕(x)). Likewise, fn(x) is sometimes used to
denote [∕(x)]". You'll often have to determine by context whether fn(x) refers to a function applied
n times, or to a function raised to the πth power.

• Don't be intimidated by notation. Often the notation is a fancy way of representing very simple
ideas.

• When working with functions, ∕~1 does not mean ∣! The expression ∕^1 is a special notation that
denotes the inverse of the function f.

Problem Solving Strategies____________________________ _

Concepts: • We can often check our answers in algebra problems by plugging the
O≡^ solutions we find back into the original question.

• Keep your eye on the ball. Sometimes we don't need to evaluate

every variable in a problem in order to solve the problem.

• We can often find the range of a function by setting the function def­
inition equal to a new variable, then solving for the dummy variable
in terms of the new variable.

L
Review Problems

2.23 Find the domain and range of each of the following functions:

(a) A(x) = 4x2 + 1 (c) P(x) =------ .

3 + √x+ 1
(b) 0(x) = 3 + √16-(x-3)2 (d) S(,τ) = ⅛≡∣

2.24 Find the domain of each of the following functions:

(a) ∕(x) = —= + √9≡3x

√2x - 5
(b) ∕(x) = ∣√x-2∣ + ∣ Vx^-2∣
(c) g(x) = √∣x∣ - 2 + V∣x- 2∣
4y 2x
2.25 Let ∕(x) = —and g(x) = Find ∕(g(x)).

2.26 Let ∕(x) = ax2 + bx + c and g(x) = ax2 - bx + c. If ∕(1) = g(l) + 2 and ∕(2) = 2, find g(2).

51
CHAPTER 2. FUNCTIONS REVIEW

2.27 At right is the graph of y - f(x). (.Jraph each of the following:

(a) y = ∕(x-3)
(b) y = 2∕(x) + l
(c) y = ∕(-x)
(d) y = ∕(2-x)
(e) * y = ∣∕(2x - 1) + 3 Hints: 24

2.28 Suppose that f (∕(x)) = g(f(x)) for all real x. Must f and g be the same
function?
2.29 For each of the following three graphs, determine if the graph could represent a function. If the
graph can represent a function, then determine if the function has an inverse. If it does have an inverse,
then graph the inverse.

2.30 If f(f(x)) ; ∕(x2), then must there be some constant c such that ∕(x) = c for all values of x in the
domain of /?
x-z
2.31 Let ∕(x,y,z) = If ∕(rt, b, c) = 1, what is the value of f(a, c, b)2
y^2
2.32 Suppose that ∕(x) = -x2 + bx + c and g(x) = dx + e, where b, c, d, and e are real constants. Is it
possible that ∕(χ(x)) = x2? Why or why not?
* + bx + c and g(x) = (5a + 2b + c)x. If ∕(1) = 0 and ∕(3) = 2, compute g(2).
2.33 Let ∕(x) = tfx
2.34 Must the in∖ erse of a linear function ∕(x) = ax + b, where a ≠ 0, be a linear function?
2.35 Suppose that a function ∕(x) is defined for all real x. How can you use the graph of ∕(x) to produce
the graph of y = |/(x)|?

2.36 Let /(I) = ------, where t ≠ 1. If y = f(x), then x can be expressed as which of the following:

f(My). -f(yl -f(-yl f(-y),

(Source: AHSME)
2.37 The function ∕ defined by ∕(x) —J* - satisfies ∕(∕(.ι)) v for all real numbers x except -3/2.
Find c. (Source: AHSMΓ)

52
 CHALLENGE PROBLEMS

2.38 Suppose that ∕(x) has an inverse, f 1(x). Must f l(2x) be the inverse of /(2x)? If so, prove it. If
not, find an expression that is the inverse of ∕(2x).
2.39 The function ∕(x) = x2 is not invertible. However, if we let g(x) = y[x, then ∕(g(x)) = (λ∕x)2 = x. So,
why isn't f invertible?

Challenge Problems

2.40 Describe all constants a and b such that ∕(χ) = ------- and f(x) = ∕'1(x) for all x in the domain of f.
bx- 2
2.41 Suppose that a function ∕(x) has domain (-1,1). Find the domains of the following functions:

(a) ∕(x + 1) (b) ∕(l∕x) (C) ∕(√J)

2.42 Find the domain and range of the function ∕(x) = V2 - x - x2. Hints: 165
2.43 Let ∕(x2 + 1) = χ4 + 5x2 + 3. What is ∕(x2 - 1)? (Source: AMC12) Hints: 277

2.44 Suppose that a function ∕(x) is defined for all real x. How can we obtain the graph of ∕(∣x∣) from
the graph of /(x)? Hints: 35
2.45 If ∕(∕(x)) = x for all x for which ∕(x) is defined and ∕(0) = 2003, find all roots of the equation
∕(x) = 0.
3x + l
2.46 Find g 1(3) given that g(x) =
2x + g(x)-

2.47 For which constants a, b, c, and d does the function ∕(x) = have an inverse?

2.48 Suppose that g(x) = ∕(x)∕(∣x∣), where ∕(x) is a function with JR as its domain. Must all x-intercepts
of the graph of f be x-intercepts of the graph of g? Must all x-intercepts of the graph of g be x-intercepts
of the graph of /?

53
CHAPTER 3. COMPLEX NUMBERS

Don't let us make imaginary evils, when you know we have so many real ones to encounter. - Oliver Goldsmith

r
CHAPTER
I
I Complex Numbers

Many great discoveries, inventions, and mathematical advances have been the result of people assum­
ing the impossible is not so impossible, after all. One such step in mathematics came when some
mathematicians confronted equations like x2 = -1.
Your first instinct might be, "No solutions. A positive number squared is positive. A negative
number squared is positive. 0 squared is 0. So, there's no way the square of a number can be negative!"
And for centuries, your first instinct was common knowledge among mathematicians, until eventually
some mathematicians started wondering, "What if there were a number that, when squared, gave -1 as
a result? What would such a number be like?"
Some mathematicians were not impressed by such musings. The great mathematician and philoso­
pher Rene Descartes even scoffed at such an idea. No real number has a negative square; such a number
is just imaginary, he believed. The name stuck. The great mathematician Leonhard Euler would give
later give us i, the symbol we use to represent a number whose square is -1:

i2 = -1.

An imaginary number is a number whose square is a real number that is not positive. The "real
numbers" Descartes was used to are those numbers we have worked with so far in this book. This name
stuck, too. The numbers we can plot on the number line are called real numbers.
A complex number is a number of the form a + bi, where a and b are real numbers and i2 - -1. We
call a the real part of a + bi and we call b the imaginary part. If b ≠ 0, then a + bi is nonreal. If a = 0, then
the complex number a + bi is called an imaginary number. (Some sources refer to imaginary numbers
as "pure imaginary.") All real numbers and all imaginary numbers are complex numbers. Just as we
use the symbol R to refer to "all real numbers," we use the symbol C to refer to "all complex numbers."

54
 3.1. ARITHMETIC OF COMPLEX NUMBERS

3.1 Arithmetic of Complex Numbers

Problems

Problem 3.1: Find all values of z such that z2 + 81 = 0.

Problem 3.2: Let w = 5 - 3i, x = -1 + 4/, y = 2 - z, and z = 1 + 2i. Compute each of the following:
(a) x+y+z (c) wz -2w-z (e) 2w2z + 3wz,
(b) xyz (d) w2 + 2wz + z2

Problem 3.3: Evaluate z2007.

Problem 3.4: Let f(x) — (—2 + 3z)x2 — (7 + 2z)x + 37 - 13z, where the domain of f is all complex numbers.
Determine each of the following:
<a) ∕<1) (b) ∕(z) (c) ∕(2 + z)

Problem 3.5:
(a) Find all real values of k such that the product (3 + 2z)(3 + ∕cz') is a real number.
(b) Let z = a + bi, where a and b are real numbers, and let z = a - bi. Prove that z ∙ z is real.

1
Problem 3.6: Write as a complex number. In other words, find real numbers a and b such that
3 + 2z'
1
= a + bi.
3 + 2z

Problem 3.7: Let w = 2-5i and z = -3 + z. Express each of the following in the form a + bi, where a
and b are real numbers.
1 z , x z Z
(a) (C) (e) 1 .+ 1
w 10 1+z 1-z
1 z
(b) (d)
w w

Problem 3.8: Find all complex numbers z such that------ — = -5 + z'.

r z+4

[Problem 3.1: Find all values of z such that z2 + 81 - 0.

Solution for Problem 3.1: Isolating z gives z2 = -81. Since the square of z is negative, z must be imaginary.
We might immediately see that our solutions are z = ±9z, because z2 = -1 and both 92 and (-9)2 equal
81. We could also "take the square root" of both sides of z2 = -81 and write z = ± √-81, where V-81 is

55
CHAPTER 3. COMPLEX NUMBERS

a shorthand for z VδT. So, we have z = ±z' VsT = ±9z. □

WARNING!! We have to be careful about using notation like √ξ81∙ The real-valued
function ∕(x) = fx is only defined for x ≥ 0. When we write V—81 as
an intermediate step to solving an equation like z2 = -81, we should
keep in mind that we mean i VδT.

Problem 3.2: Let w = 5 - 3z, x = -1 + 4z, y = 2 - i, and z = 1 + 2i. Compute each of the following:
(a) x+y+z (c) wz -2w -z (e) 2w2z + 3wz
(b) xyz (d) w2 + 2wz + z2

Solution for Problem 3.2:

(a) We add complex numbers by adding the real parts and the imaginary parts separately:

x + y = (-1 + 4z) + (2 - z) = (-1 + 2) + (4 - l)z = 1 + 3z,

so x + y + z = (x + y) + z = (1 + 3z) + (1 + 2z) - (1 + 1) + (3 + 2)z = 2 + 5i.

(b) We use the distributive property to multiply complex numbers:

xy = (-1 + 4z)(2 - z) = -1(2 - z) + 4z(2 - z)

= -2 + z + 8z - 4z'2
= -2 + 9z - 4(-l)
= 2 + 9z.

WARNING!! Be careful with signs when working with complex numbers. One
V common error is to incorrectly simplify an expression like 1 - 4z2 as
1-4, rather than 1+4. Make sure you see why 1 - 4z2 = 1 + 4.
Similarly, for any two complex numbers a + bi and c + di, we have

(a + bi)(c + di) = a(c + di) + bi(c + di)

= ac + adi + bci + bdi2
= ac + adi + bci — bd
= (ac - bd) + (ad + bc)i.

You shouldn't memorize this as a formula; this is just an example of the expansion of the product
of two binomials using the distributive property.
Since xy = 2 + 9z, we have xyz = (xy)z = (2 + 9z)(l + 2z) = 2(1 + 2z') + 9z(l + 2z) = -16 + 13z.
(c) Since wz = (5 - 3z)(l + 2z) = 5 + 10/ - 3z - 6z2 = 11 + 7z, we have

wz - 2τv - z = (5 - 3z)(l + 2z) - 2(5 - 3z) - (1 + 2z) = 11 + 7z - 10 + 6z - 1 - 2z = llz.

56
 3.1. ARITHMETIC OF COMPLEX NUMBERS

(d) We note that <υ2 + 2wz + z2 = (w + z)2, so

w2 + 2wz + z2 = (w + z)2 = [(5 - 3z) + (1 + 2z)]2 = (6 - i)2 = 62 + 2(6)(-i) + (-z)2.

In simplifying the last expression on the right, we note that (-z)2 = (-l)2(z)2 = -1, so we have
w2 + 2wz + z2 = 36 - 12z - 1 = 35 - 12z.
Concept: Factoring and other manipulations can make some calculations easier.
O≡≡^

(e) The given expression factors as wz{2w + 3z), so

2w2z + 3wz2 = wz(2w + 3z) = (5 - 3z)(l + 2z)[2(5 - 3z) + 3(1 + 2z)]

= (5 - 3z)(l + 2z)(13) = (11 + 7z)(13) = 143 + 91z.

Notice how factoring saved a lot of work. Without factoring we would have had to perform
several multiplications of binomials, but factoring allowed us to get the answer with just one.
□
Problem 3.3: Evaluate z2007. 1

Solution for Problem 3.3: At first, it's not obvious how to compute z2007, so we experiment with smaller
powers of z, hoping to find a pattern. We have z1 = z, z2 = -1, z3 = z2 ∙ z = -z, and z4 = z2 ■ z2 = (-l)2 = 1.
Then, the powers repeat: z5 = z4 ∙ z = z, z6 = z4 ∙ z2 = z2, z7 = z4 ∙ z3 = z3, and z8 = z4 ∙ z4. Therefore, we see that:

Important: The powers of z repeat in blocks of 4:

ψ z1
z2
= z5 =
= z6 =
i9 =
z1° =
∙∙∙
...
=
=
z,

z3 = z7 = z11 = ■∙∙ = -z,

z∙4 = z∙8 = z∙12 = ... = 1

Because 2007 leaves a remainder of 3 when divided by 4, we have z2007 = z3 = -z. □

In Chapter 2, we reviewed functions with real domains and real ranges. WeTl continue practicing
complex number arithmetic by considering a function that accepts nonreal inputs and has nonreal
numbers in its range.

Problem 3.4: Let ∕(x) = (-2 + 3z)x2 - (7 + 2z)x + 37 - 13z, where the domain of f is all complex numbers
Determine each of the following:
(a) ∕(1) (b) ∕(z) (c) ∕(2 + z)

Solution for Problem 3.4: Just as with the real functions we have studied, we substitute each input for x
to determine the value of the function f for that input.

(a) ∕(1) = (-2 + 3z) ∙ 12 - (7 + 2z) ∙ 1 + 37 - 13z = 28 - 12z.

(b) f(i) = (-2 + 3z) ∙ z2 — (7 + 2z) ∙ z + 37 - 13z = (-2 + 3z)(-l) - (7z - 2) + 37 - 13z = 41 - 23z.

57
CHAPTER 3. COMPLEX NUMBERS

⅜κ
(c) ∕(2 + z) = (-2 + 3z) ∙ (2 + z)2 - (7 + 2z) ■ (2 + z) + 37 - 13z = (-2 + 3z)(3 + 4z) - (12 + llz) + 37 - 13z =
(-18 + ∕) - (12 + llz) + 37 - 13z = 7 - 23z.

□
As we've seen in our examples so far in this chapter, the product of two complex numbers is itself a
complex number. Let's take a look at when the product of two nonreal complex numbers turns out to
be a real number.
Problem 3.5:
(a) Find all values of k such that the product (3 + 2z)(3 + fcz) is a real number.
(b) Let z = fl + bi, where a and b are real numbers, and let z = fl —J

Solution for Problem 3.5:

(a) Expanding the given product, we have (3 + 2i)(3 + ki) = 9 — 2k + (6 + 3k)i. The result is a real number
when 6 + 3k = 0, so k = —2.
(b) In part (a) we saw that the product of the nonreal numbers 3 + 2z and 3 — 2z is real. Notice that the
only difference between these two numbers is the sign of the imaginary part. Let's see if we can
generalize our observation by checking whether or not the product (fl + bi)(a — bi) is a real number.
We find
(a + fez)(fl - bi) = fl(a - bi) + fez(fl - bi) = a2 - abi + abi - b2i2 = a2 + b2,
which is real because a and b are real.

Definition: If z = fl + bi, then a - bi is called the conjugate of z. We denote the conjugate of z as z.

So, we have fl + fez = fl - bi. Together, a number and its conjugate are referred to as a conjugate pair.

Important: The product of a complex number and its conjugate is a real number.

Note that in order to prove this fact in Problem 3.5, we wrote a general complex number in terms of
its real and imaginary parts by writing z = fl + bi.

Concept: In general, if you are given a problem involving a general complex number
O≡^ z, it may help to write z in the form a + fez, and then express the problem in
terms of a and fe.

We've covered addition, subtraction, and multiplication of complex numbers. Let's put conjugates
to work by turning to division.
f - - _
Problem 3.6: Write - ----- ■ as a complex number. In other words, find real numbers fl and fe such that I

58
 3.1. ARITHMETIC OF COMPLEX NUMBERS

Solution for Problem 3.6: Knowing that (3 + 2z')(3 - 2ι) is real, we multiply the given fraction by L-∣⅞z
which equals 1, to produce an equivalent fraction with a real denominator:

1 _ 1 3 -2z _ 3- 2i _ 3-2i 3-2i

3 + 2i 3 + 2i' 3- 2i ^ (3 + 2z)(3 - 2z) ^ 32 + 22 ^ 13 ’

1 3 2
Therefore, we have -—— = —---- -z. □
3 + 2z 13 13
The fact that zz is real for all complex numbers z gives us a way to divide by a complex number.

Important: If w and z are complex numbers and z ≠ 0, then we express the quo­

ψ tient w∣z as a complex number by multiplying both the numerator and

denominator by the conjugate of the denominator, z:

w _ w z _wz
z z z zz'
Since zz is always real, we can easily express wz∕{zz) as a complex num­
ber.

Let's give this technique some practice.

Problem 3.7: Let w = 2 — 5z and z - — 3 + i. Express each of the following in the form a + bi, where a I
and b are real numbers.

(a)
1
(C)
z
(e)
∕∖1
z
∙ ÷
z
w zυ l+z l—z
1 z
(b) (d)
w w
mrmimmmhm

Solution for Problem 3.7:

2 5i
^ 29 + 29’
2 _ 5i_
~ 29 29'
13z
29'
13z
29'
(e) We could simplify each fraction, and then add. Instead, we recognize that the denominators in
the problem are conjugates, so their product is real. Therefore, the common denominator of the
two fractions is a real number:
-3 + i + -3 - z _ (-3 + z')(l - z) + (-3 - z)(l + z) _ -2 + 4z + -2 - 4z _ -4 _
ι + f + ι-z - (i + z)(i-z) ^ -T~"2∙
□

59
CHAPTER 3. COMPLEX NUMBERS

Notice that the answer to part (b) is the conjugate of the answer to part (a), and the answer to part
(d) is the conjugate of the answer to part (c). This isn't a coincidence. We'll explore why in Section 3.3.
See if you can figure out why on your own before getting there.

2z - 3/
Problem 3.8: Find z such that = -5 + i.
z+4

Solution for Problem 3.8: We begin by multiplying both sides by the denominator of the left side, which
gives us 2z — 3z = (-5 + i)(z + 4). Expanding the right side then gives 2z - 3z = -5z — 20 + zz + 4z. Putting
all terms with z on one side and terms without z on the other gives us

7z - zz = -20 + 7z.

Factoring z out of the left side gives z(7 - z) = —20 + 7z. Dividing both sides by 7 — z then gives us

-20 + 7z -20 + 7z 7 + z -147 + 29z _ 147 29 .

7-z 7-z ’ 7 + z 50 50 + 50Z

Sidenote: Mathematicians frequently use w and z as variables to represent complex

numbers, and use x and y as variables to represent real numbers. Of course,
this is not a hard-and-fast rule! As you'll see in this book, the variable x is
so universally common that we will have many equations in terms of x to
which some solutions turn out to be nonreal.

Exercises
♦

3.1.1 Find all z such that 4z2 + 12 = 0.

3.1.2 Let a = 3 + 4z and b = 12 - 5z. Compute each of the following:

(a) a-b (c) fl2 + 3z7 + 2

a a
(b) ab (d) ~f + ~b

3.1.3 Write each of the following as a complex number:

1
(a) 2z + 7 - 2z (b)
2 + 3z + 1 + 2z

3.1.4 f(x) = . Find each of the following:

(a) ∕(z) (b) ∕(-z) (c) ∕(z-l)

3.1.5 What is (z - Γ1)"1 ? (Source: AHSME)

60
 3.2. THE COMPLEX PLANE

3.1.6 Solve each of the following equations for z:

(a) ≡±⅞=2 (b) li⅛ = 4+5i (c) 3z +-L. = 10 _ 4i'
z-3 3z 1 +1
3.1.7 Let S = in + i~n, where n is an integer. Find the total number of possible distinct values of S.
(Source: AHSME)

3.2 The Complex Plane

A complex number can be represented as a point in the

complex plane. Like the coordinate plane, the complex
plane has two axes, a horizontal real axis for the real part,
and a vertical imaginary axis for the imaginary part. So, 1 +2z
plotting the complex number x + yi on the complex plane --t
1
is the same as plotting the point (x, y) in the Cartesian coor­ I I l I R
1 1 1 1 1 1 1 *
dinate plane. Tire figure at right shows the numbers 1 + 2z, ”I
1
I I

1 1
—5 — 3z, and 2 — 4z plotted in the complex plane. 1 I
1 I
1
Just as the point (0,0) is called the origin on the Cartesian 1
o---------------------
1
∣
plane, the point that represents 0 on the complex plane is -3z 1
also called the origin of the complex plane. ’’ ^-4z
*
2

The real axis of the complex plane is labeled Re and the imaginary axis is labeled Im. Similarly, we
can use Re(z) to refer to the real part of the complex number z, and use Im(z) to refer to the imaginary
part of z. So, for example Re(2 - 4z) = 2 and Im(2 - 4z) = -4. Some sources use 'X(z) and 3 (z) in place of
Re(z) and Im(z).

Problems

Problem 3.9:
(a) Let w = 2 + 3z. Plot w, w, and -w on the complex plane.
(b) For any complex number z, how is z related to z on the complex plane? How is z related to -z
on the complex plane?

Problem 3.10:
(a) Let w = 2 + 4z and z = 3 - 2z. Plot w, z, and w + z on the complex plane. Connect the origin to
w, then connect w to ιυ + z, then connect w + z to z, then connect z to the origin. Notice anything
interesting?
(b) Suppose w and z are nonzero complex numbers such that w/z is not a real constant. (In other
words, there is no real number a such that w = az.) Connect the origin to w, then connect w to
w + z, then connect w + z to z, then connect z to the origin. What type of quadrilateral must
result? Why?

61
CHAPTER 3. COMPLEX NUMBERS

Problem 3.11:
(a) What is the distance between 3 + 4z and the origin in the complex plane?
(b) What is the distance between a + bi and the origin in the complex plane?

Problem 3.12: Let ∣fl + bi∖ = + b2.

(a) Evaluate ∣5 - 12z∣ and ∣3 - 3z∣.
(b) Suppose w = 3 - 5z and z = -2 + 7z. Find ∣w - z∖. Find the distance between w and z on the
complex plane. Notice anything interesting?
(c) For any complex numbers w and z, how is ∣w - z∣ related to the distance between w and z on the
complex plane?

Problem 3.9: For any complex number z, how is z related to z on the complex plane? How is z
related to —z on the complex plane?

Solution for Problem 3.9: We start by experimenting with a specific value of z.

Concept: Trying specific examples can be a good way to discover general relation-
Q≡s≡ ships-

Suppose z = 2 + 3z. Then, we have z = 2 - 3z' and -z = -2 - 3z.

1
We see that z is the reflection of z over the real axis and -z is the
z — z. I ■
reflection of z over the imaginary axis. Is this just a coincidence?
To see if this is true in general, we let z = fl + bi. Then, we have
z = a-bi and -z = -fl - bi. So, z and z have the same real part .
and opposite imaginary parts. Therefore, z and z are reflections of
each other over the real axis. Similarly, z and -z have the same i
r, .* " ,
imaginary part, but opposite real parts, so z and —z are reflections
of each other over the imaginary axis. fc- -
1
But how are z and —z related? Because z = a + bi and —z = —« — bi, -z -2--3ι z » _ 3z~
z and —z are in exactly opposite directions from the origin. We also 1 1 ∣ r
I
1
see that z and -z are the same distance from the origin. Therefore, -z is a 180o rotation of z around the
origin on the complex plane. (Note: We sometimes refer to a 180o rotation of a point A around a point
P as a "reflection of point A through point P.") □

Important: For every complex number z, we have

ψ • The point -z is a 180o rotation of point z about the origin.

• The points z and z are reflections of each other over the real axis.

• The points z and -z are reflections of each other over the imaginary
axis.

62
 3.2. THE COMPLEX PLANE

Don't memorize these relationships! If you understand why they are true, they should become
obvious to you.

Problem 3.10: Suppose rz, and z are nonzero complex numbers such that w/Z is not a real constant.
(In other words, there is no real number a such that w = az.) Connect the origin to w, then connect w
to w + z, then connect w + z to z, then connect z to the origin. What type of quadrilateral must result?
Why?

Solution for Problem 3.10: Again, we start with an example. Im

1k 1 I 1 f 1
Suppose w = 2 + 4z and z = 3 - 2i, so w + z = 5 + 2i. In — W ■
2÷4z I-
___ f

the diagram at right, we connect the points as suggested. We j—j I

s 7/1 -L- r7 -
appear to have a parallelogram. (2 + 3) + (4 - 2)i
>
∕
To see if this is always the case, we let w = a + bi and
z = c + di. So, we have w + z = (a + c) + (b + d)i. Ifα ≠ 0, then the
slope between w and the origin is b∕a, as is the slope between
w + z and z. If a = 0, then both of these lines are vertical lines. Z == 3 2z, — —
In either case, these two sides of our quadrilateral are parallel. ∣
Similarly, the slopes of the other two sides are either both equal
to d∣c or they are both undefined (when both sides are vertical). So, these two sides are parallel as well.
Therefore, both pairs of opposite sides of the quadrilateral are parallel, which means the quadrilateral
is a parallelogram. □

Problem 3.11:
(a) What is the distance between 3 + 4z and the origin in the complex plane?
(b) What is the distance between a + bi and the origin in the complex plane?

Solution for Problem 3.11:

(a) Finding the distance between 3 + 4z and the origin on the complex plane is
the same as finding the distance between (3,4) and (0,0) on the Cartesian
plane. We can use the distance formula, or build a right triangle as shown
at right, to find that the desired distance is √(3 - 0)2 + (4 - 0)2 = 5.
(b) The distance between a + bi and 0 on the complex plane is the same as the
distance between (zz, b) and (0,0) on the Cartesian plane. So, the desired
distance is vzz2 + b2.

Just as we use ∣x∣ to denote the distance between a real number x and 0 on the number line, we use ∣z∣
to represent the distance between z and the origin on the complex plane. We call ∣z∣ the magnitude of z.

Important: The magnitude of z, denoted by ∣z∣, equals the distance from z to the

ψ origin on the complex plane. If z = a + bi, we have

∣z∣ = √zz2 + b2.

63
CHAPTER 3. COMPLEX NUMBERS

Problem 3.12:
(a) Evaluate ∣5 - 12z∣ and ∣3 - 3z∣.
(b) Suppose w = 3-5i and z = -2 + 7z. Find ∖w - z∣. Find the distance between w and z on the
complex plane. Notice anything interesting?
(c) For any complex numbers w and z, how is ∣w — z∣ related to the distance between w and z on
the complex plane?

Solution for Problem 3.12:

(a) We have ∣5 - 12z∣ = y∕¾- + (-12)i2 = 13 and ∣3 - 3z∣ = √32 + (-3)2 = 3 √2.
(b) We have w - z = 5 - 12/, and in the previous part we found that
∣5 - 12z∣ = 13. We plot the two points in the diagram at right.
We see that the horizontal distance between them is 5 and the
vertical distance between them is 12, so the segment connecting
the two points is the hypotenuse of a right triangle with legs
of lengths 5 and 12. Therefore, the Pythagorean Theorem tells
us that the distance between w and z on the complex plane is
√52 + 122 = 13.
Notice that the distance between w and z equals ∖w - z∣. Is
this a coincidence?

(c) No, it's not a coincidence. Suppose w = a + bi and z = c + di. Finding the distance between w and
z in the complex plane is the same as finding the distance between (a, b) and (c, d) on the Cartesian
plane. So, the distance between w and z is √(α - c)2 + (& - d)2.
Because w - z = (a + bi) - (c + di) = (a-c) + (b- d)i, we have
∣w - z∣ = >∕(fl - c)2 + (& - d)2.

This tells us that:

Important: The distance between complex numbers w and z on the complex plane

ψ is ∖w - z∖.

i 11
Exercises

3.2.1 Plot each of the following in the complex plane:

(a) 4 + 7z (b) -6 - 2z (c) (3 + z)(-2 + 5z)

3.2.2 Find the magnitude of each of the following complex numbers:

(a) 24 -7i (b) 2 + 2√3z (c) (l+2z)(2 + z)

3.2.3 Let w = 3 + 5z and z = 12 + 2i. Find the area of the convex quadrilateral in the complex plane that
has vertices w, z, w, and z.

64
 3.3. REAL AND IMAGINARY PARTS

3.2.4 Find the distance between the points 4 + 7z and -3 - 17/ in the complex plane.
3.2.5 Show that the midpoint of the segment connecting zι and z^ on the complex plane is (zj + Z2)∕2.
3.2.6

(a) Find the magnitude of ~γ^~ ∙

C 1 XXz
(b) Find the magnitude of (You can use a calculator for this part.)

(c) Notice anything interesting? Can you generalize your observations from the first two parts?
3.2.7★ Four complex numbers lie at the vertices of a square in the complex plane. Three of the numbers
are 1 + 2i, -2 + z and -1 - 2z. What is the fourth number? (Source: AMC12)

3.3 Real and Imaginary Parts

II
I

Problems

Problem 3.13: Let z and w be complex numbers.

(a) Let z = a + bi and w = c + di, where a, b, c, and d are real. Show that z + w = z + w.
(b) Show that zw = z∙w.

Problem 3.14:
(a) Show that z = z for all complex numbers z.
(b) Show that z = z if and only if z is real.
(c) Show that z = —z if and only if z is imaginary.

Problem 3.15:
(a) Prove that zz = ∣z∣2 for all complex numbers z.
(b) Prove that ∖zw∖ = ∣z∣∣zσ∣ for all complex numbers z and w.

Problem 3.16: Solve the equation z + 2z = 6 - 4/ for z.

Problem 3.17: In this problem, we find all complex numbers z such that z2 = 21 - 20/.
(a) Let z = a + bi in the given equation. Find a system of equations involving a and b.
(b) Solve for b in terms of a in one of the equations, and substitute the expression you found for b
into the other equation.
(c) Solve the equation you formed in part (b) for all possible values of a. Hints: 90
(d) Find all complex numbers z such that z2 = 21 - 20z.

65
CHAPTER 3. COMPLEX NUMBERS

Problem 3.18: In this problem, we find ∣z∣ if the complex number z satisfies z + ∣z∣ = 2 + 8i.
(a) Let r = ∣z∣. Express z in terms of r.
(b) What is the real part of z? What is the imaginary part of z? Find ∣z∣.
(c) Solve the problem again by letting z = a + bi.
(Source: AHSME)

Problem 3.13: Let z and w be complex numbers.

(a) Show that z + w = z + w.
(b) Show thatzw = z∙w. _________________________ ∣

Solution for Problem 3.13: Let z = a + bi and w — c + di, where a, b, c, and d are real numbers.
(a) Since z + w = a + bi + c + di = (a + c) + (b + d)i, the conjugate of z + w is

z + w = (α + c) + (b + d)i = (a + c)- (b + d)i.

Similarly, the conjugate of z is z = a - bi, and the conjugate of w is w= c— di, and their sum is
a - bi + c - di = (a + c) - (b + d)i. So, we have z + w = z + w.
(b) We have
zw - (α + bz)(c + di) = ac + adi + bci + bdi1 = (ac - bd) +(ad +bc)i,
so the conjugate of zw is zw = (ac - bd) - (ad + bc)i. Also, we have

z ∙ w = (a - bi)(c - di) = ac- adi - bci + bdi2 = (ac - bd) — (ad + bc)i,

so zw = z ∙w.

Important: For any complex numbers w and z, we have w + z = w + z and wz = w∙z.

ψ

We'll be seeing these two very important relationships again when we study nonreal roots of poly­
nomials in Section 8.2.
Problem 3.14:
(a) Show that z = z for all complex numbers z.
(b) Show that z = z if and only if z is real.
(c) Show that z = -z if and only if z is imaginary.

Solution for Problem 3.14: Let z = a + bi, where a and b are real numbers.

(a) Since z = a - bi, we have z = a - bi = a + bi = z.

66
 3.3. REAL AND IMAGINARY PARTS

(b) Since z = a - bi, the equation z = z becomes a - bi = a + bi. Subtracting a from both sides gives
-bi = bi. Therefore, we have 2bi = 0, so b = 0. This gives us z = a + bi = aι + 0z = a, so z is real. This
shows that z = z only if z is real.
We're not finished with our proof! We have only shown that when z = z, then z is real. (In
other words, we have z-z only if z is real.) We have not shown that the equation z = z is true for
all real numbers. (That is, that z = z if z is real.)
To show that z = z whenever z is real, we let z = a, where a is a real number. We then have
z = a = a, so z = z if z is real.
Important: Suppose A and B are two mathematical statements. The statement "A if
and only if B" means both of the following:

• If A is true, then B is also true.

__ ∙ If B is true, then A is also true.

Therefore, as we just saw, proving an "if and only if" statement requires proving two statements,
not just one.
(c) Since z = a - bi, the equation z = -z becomes a - bi = -a - bi. Adding bi to both sides gives a = -a,
so α = 0. Therefore, we have z = a + bi = bi, so z is imaginary. This shows that z = —z only if z is
imaginary.
As with part (b), this is an "if and only if" statement. We must also show that z = -z if z is
imaginary. Letting z = bi, we have z = bi = -bi = -z, as desired.

Problem 3.15: ∣
(a) Prove that zz = ∣z∣2 for all complex numbers z.
(b) Prove that ∖zw∖ = ∖z∖∖w∖ for all complex numbers z and w.

Solution for Problem 3.15: Let z = a + bi and w = c + di for real numbers a, b, c, and d.

(a) Since z = a- bi, we have zz = (α + bι)(a - bi) = a2 + b2 = + b2) = ∖a + bz∣2 = ∣z∣2.

Extra! In 1806, bookkeeper Jean-Robert Argand published his concept of a geometric interpre-
l∣mt∣ ∣uk∣∣4>∣∣∣∣* tation of complex numbers. Since he wasn't a professional mathematician, he used his
knowledge of the bookselling industry to get his work out. A copy of his work reached
the mathematician Legendre, who sent a copy to Franςois Franςais, yet neither could
figure out who the author was—apparently Argand had left out his name!
Franςois Franςais passed away, and his brother Jacques Franςais discovered Argand's
book among his deceased brother's papers. Jacques Franςais then published a work in
1813 featuring a geometric representation of complex numbers, but instead of claiming
that concept as his own, he mentioned in his paper that the idea was from some anony­
mous mathematician's work, and so Argand came forward and his work was recognized.
For this reason, some now call the complex plane the Argand plane, and call graphs of
complex number relationships on this plane Argand diagrams.

67
CHAPTER 3. COMPLEX NUMBERS

(b) First, we note that zw = (« + bi)(c + di) = ac-bd + (ad + bc)i. So, we have
∣zw∣ = ∣ac -bd + (ad + bc)i∖.
Next, we use the fact that ∣r + sz∣ = >[r2 + s2 for any real numbers r and s. Here, we have r = ac-bd
and s = ad + be. So, we have:
∣zzυ∣ = ∣ac - bd + (ad + bc)i∖ = y∕(ac - bzi)2 + (ad + be)2
= Vβ2c2 _ fabed + b2rf2 + λ2^2 + 2abcd + b2c2
= √α2c2 + β2c∕2 + ⅛2c2 + ⅛2^2

= λ∕λ2(c2 + d2) + b2(e2 + ^2)

= √(fl2 + b2)(c2 + rf2)
= Vo2 + b2Vc2 + d2
= ∖a + bi∖∖c + rfz∣ = ∖z∖∖w∖.

Important: For any complex numbers z and w, we have

• ∖zw∖ = ∣z∣∣w∣.

Problem 3.16: Solve z + 2z = 6 - 4z for z. 1

Solution for Problem 3.16: Letting z = a + bi, for some real numbers a and b, the given equation becomes
(zz + bi) + 2(zz - bi) = 6 - 4z. Simplifying the left side gives 3a - bi = 6 - 4z. Since two complex numbers
are equal if and only if their real parts are equal and their imaginary parts are equal, we have 3a = 6 and
-b = -4. Solving these two equations gives a = 2 and b - 4, so z = a + bi = 2 + 4z. □

Problem 3.17: Find all complex numbers z such that z2 = 21 - 20z.

Solution for Problem 3.17: We might start by taking the square root of both sides, so that z = ± √21 - 20z.
But what does V21 - 20z really mean? The real-valued function f(χ) = -√χ is only defined for x ≥ 0;
if x ≥ 0, then fx is the nonnegative number whose square is x. Unfortunately, this doesn't help us
define the square root of a nonreal number. Instead of trying to define √21 - 20z, let's look for complex
numbers whose square is 21 — 20z.
Let z = a + bi, so we have (a + bz)2 = 21 - 20z. Expanding the left side gives
a2 - b2 + 2abi = 21 - 20z.
Since a and b are real, we can equate the real parts of both sides to get a2 — b2 = 21, and equate the
imaginary parts to find 2ab = —20. Solving 2ab = -20 for b gives b = -10/a. Substituting this into
a2 - b2 - 21 gives

68
 3.3. REAL AND IMAGINARY PARTS

Multiplying this by a2 gives us a4 - 21α2 - 100 = 0. Letting c = a2 in this equation yields the quadratic
equation c2 - 21c - 100 = 0, which has roots 25 and -4. However, a is real and a2 = c, so c cannot be -4.

Therefore, we have c = a2 = 25, so a = ±5. Since b = -10∕a we have b = -2 when a = 5, and b = 2

when a = -5. So, the two possible values of z are ±(5 - 2z). □
Having found the two complex numbers whose squares equal 21 - 20z, we can see some of the
difficulty that arises from defining √21 - 20/. Which of the values 5 - 2i or -5 + 2i should we take to be
the square root? We can't simply call one "positive" and one "negative."

Due to these difficulties, we try to avoid using expressions like V21 — 20z, and instead think of the
"numbers whose square is 21 - 20z." We may sometimes refer to these two numbers as "the square
roots" of 21 - 20z, and may even refer to one or the other as "a square root" of 21 - 20z. But we generally
won't choose one or the other to equal V21 — 20z the way we choose 5 to equal y∣25.

We have these same difficulties when trying to define V~81. For example, which of 9z or -9z is "the"
square root of -81? However, as we saw in Problem 3.1, we may occasionally stumble on expressions
like V—81 when solving equations like z2 = —81. When this happens, the square root should always be
preceded by ±, as in z = + V-81. In other words, we are referring to both "square roots" of -81; that is,
we are considering both numbers whose squares are —81.

Problem 3.18: The complex number z satisfies z + ∣z∣ = 2 + 8z. What is |z|? (Source: AHSME)

Solution for Problem 3.18: Solution 1: Keep your eye on the ball. We want ∣z∣, and our equation involves ∣z∣,
so we let ∣z∣ = r. Then, we have z = 2 + 8z - ∣z∣ = (2 - r) + 8z. Written this way, we can see that the complex
number z has real part 2 - r and imaginary part 8. Therefore, we have

∣z∣2 = (2 - f)2 + 82 = r2 - 4r + 68.

Concept: Often, dealing with ∣z∣2 is nicer than dealing with ∣z∣, since squaring ∣z∣ gets
Q oj rid of the square root sign in the definition of ∣z∣.

We also have ∣z∣2 = r2, and combining this with our equation above gives f2 = r2 - 4r + 68. Solving this
equation gives us r = 17, so ∣z∣ = 17.
Solution 2: Let z = a + bi. Then, the given equation is

a + bi + λJcP- + t>2 = 2 + 8i.

The only imaginary term on the left is bi, so we have b = 8. Equating the real terms of both sides
then gives a + Va2 + 64 = 2. Isolating the radical gives √α2 + 64 = 2-a and squaring both sides gives
fl2 + 64 = 4 - 4zz + a2. Solving this equation gives us a = -15. Because we squared the equation
Vzz2 + 64 = 2 - a, we must check if the solution is extraneous. It isn't, so ∣z∣ = |—15 + 8z∣ = 17. □

.1 ~' “ , ,
Γxeκi*e⅜

3.3.1 Find ∣(10 + 24z)(8 - 6z)∣ without finding the product (10 + 24z)(8 - 6z).

69
CHAPTER 3. COMPLEX NUMBERS

3.3.2 Show that ∣z∣ = ∣z∣.

3.3.3 Solve 3z + 4z = 12 - 5z' for z.
3.3.4 Solve each of the following equations:
(a) z2 = 2z (b) z2 = -5 + 12z (c) z2 = 24 - 10z

3.3.5 Let z = 3 + 4z and w = 5 — 12z. Evaluate the following expressions:

1 z M
w
(a)
z <b, ⅛ (c) (d)
∣w∣
Compare your answers in parts (a) and (b), and compare your answers in parts (c) and (d). Notice any­
thing interesting? (You should!) Prove that your observations will hold for any nonzero complex
numbers.
3.3.6 Find all complex numbers z such that z/z is
(a) a real number. (b) an imaginary number.

*
3.3.7 Two solutions of x4 - 3x3 + 5x2 - 27x - 36 = 0 are pure imaginary numbers. Find these two
solutions. (Source: ARML) Hints: 213

3.4 Graphing in the Complex Plane

We graph an equation in terms of x and y in the Cartesian plane by plotting all the points (x, y) that
satisfy the equation. Similarly we graph an equation in terms of a complex number z in the complex
plane by plotting all the values of z that satisfy the equation.

Problems

Z+Z Z—Z
Problem 3.19: Show that Re(z) = -y- and Im(z) =

Problem 3.20: Graph in the complex plane all complex numbers z that satisfy each of the following:

(a) =6 (c) (3 + 2z)z + (3 - 2z)z = 36

(b) z-z = -2√7∕

Problem 3.21: Let w be a constant complex number and c be a nonnegative real constant. Describe the
graphs of each of the following:

(a) ∣z∣ = 3 (c) ∣z + 5 - 4z∣ = 2√2

(b) ∣z-4∣ = 3 (d) ∖z-w∖ = c

Problem 3.22: Find and graph all z such that ∣z - 3∣ = ∣z + 2z∣.

70
 3.4. GRAPHING IN THE COMPLEX PLANE

Problem 3.19: Show that Re(z) = - and lm(z) = Z-.2.

Solution for Problem 3.19: Let z = a + bi so that z = a- bi. Then

z + z _ (a + bi) + (a- bi) _ 2a

z - z _ (a + bi) - (α - bi) _ 2bi

~2Γ ~ 2i ~ 2?

z+z
So, Re(z) = a =
^τ~ and Im(z) = b =

Important: For any complex number z, we have

ψ Re(z) =
z+z
and
~2~

Problem 3.20: Graph in the complex plane all complex numbers z that satisfy each of the following:
(a) r42 = 6 (c) (3 + 2z)z + (3 - 2i)z = 36

(b) z-z = -2y∕7i

Solution for Problem 3.20:

(a) In Problem 3.19, we found that Re(z) = —y-. So, we seek all complex numbers z such that
Re(z) = 6. These form a vertical line in the complex plane. This line is shown in the graph on the
left below.
Im

(b) The left side of the.given equation looks a lot like our expression for Im(z) from Problem 3.19. We
only have to divide both sides by 2i to match this form on the left side. Doing so gives us

71
CHAPTER 3. COMPLEX NUMBERS

so we have Im(z) = -y∣7. All z with imaginary part -λ∕7 satisfy this equation, so the graph of the
given equation is the horizontal line V7 units below the real axis, which is shown at right above.
(c) We can't easily relate the given equation to either of the forms we
found in Problem 3.19. So, we start by writing z as fl + bi, where a
and b are real numbers. The given equation becomes

(3 + 2z)(fl + fez) + (3 - 2z)(zz - bi) = 36.

Expanding and rearranging this equation gives us 3α — 2b = 18. This

is simply a linear equation. To graph it on the complex plane, we
note that a = Re(z) and b = Im(z), so we have

3Re(z) - 2Im(z) = 18.

The graph of this line is shown at right. (If you don't see why this is the graph of the equation, try
graphing the line 3x — 2y = 18 on the Cartesian plane. In the complex plane, Re(z) takes the place
of x and Im(z) takes the place of y. )

In the next problem, we explore how distance can help us graph in the complex plane.

Problem 3.21: Let w be a constant complex number and c be a nonnegative real constant. Describe
the graphs of each of the following:

(a) ∣z∣ = 3 (c) ∣z + 5 - 4z∣ = 2√2

(b) ∣z-4∣ = 3 (d) ∣z - w∣ = c

Solution for Problem 3.21:

(a) If ∣z∣ = 3, then the distance from z to the origin of the complex plane is 3. Therefore, z must be on
the circle centered at the origin with radius 3. Moreover, every point z on this circle satisfies the
equation ∣z∣ = 3, because every point on this circle is 3 units from the origin. Therefore, the graph
∣z∣ = 3 describes a circle of radius 3 centered the origin.
(b) The equation ∣z - 4∣ = 3 tells us that the distance between z and 4 in the complex plane is 3, so the
graph is a circle of radius 3 centered at the point that represents 4. Since 4 = 4 + 0z, the center of
the circle is on the real axis, 4 emits to the right of the origin.
(c) Writing z + 5 - 4z as z - (-5 + 4z), we see that the graph of the equation Im
∖z - (-5 + 4z)∣ = 2^2 consists of all points that are 2 V2 from the point
—5+4z. Therefore, the graph is a circle of radius 2λ∕2 centered at —5+4z.
(d) Generalizing our work from parts (a)-(c), we note that the graph of
∣z — w∖ = c consists of all points that are a distance c from the complex
number w. In other words, the graph is a circle of radius c centered at
w (shown at right for w = 2 + z and c - 6).

72
 3.4. GRAPHING IN THE COMPLEX PLANE

Solution for Problem 3.22: Solution 1: Algebra. Letting z = a + bi, for

real numbers a and b, the equation becomes ∖a + bi - 3∣ = ∣α + bi + 2i∖, or
I (fl - 3) + bi∖ = ∣α + (b + 2)z∣.

Applying the definition of the magnitude of a complex number to

both sides of this equation, we have

V(fl - 3)2 + b2 = √fl2 + (b + 2)2.

Squaring both sides of this equation and expanding the squares of

binomials gives
fl2 — 6a + 9 + b2 = a2 + b2 + 4b + 4.

Simplifying this equation gives 6a + 4b = 5, so our equation is 6Re(z) + 4Im(z) = 5. The graph of this
equation is a line, shown at right above. We can write the equation 6Re(z) + 4Im(z) = 5 in terms of z and
z by noting that Re(z) = (z + z)∕2 and Im(z) = (z - z)∕(2z). Substituting these into 6Re(z) + 4Im(z) = 5
gives us (3 - 2z)z + (3 + 2z)z = 5.

Solution 2: Geometric intuition. We rewrite the equation so that both magnitudes are of differences
between complex numbers, ∣z - 3∣ = ∖z- (-2z)∣. The quantity on the left side of the equation is the distance
from z to 3 in the complex plane. Similarly, the right side is the distance from z to -2z in the complex
plane. Therefore, our equation tells us that z is equidistant from 3 and —2z' in the complex plane. So, z
must be on the perpendicular bisector of the segment that connects 3 and -2z in the complex plane.
The slope of the line connecting 3 and -2z' is 2/3, so the slope of its perpendicular bisector is -∣. The
perpendicular bisector passes through the midpoint of 3 and -2z, which is = ∣ - z. Using this point
and the slope of the line, we can write a point-slope form of the line:
Im(z) - (-1) = -∣ ^Re(z) - ∣).

Rearranging this equation gives us 6Re(z) + 4Im(z) = 5, as before. □

The key to our second solution was using the geometric interpretations of ∣z - 3∣ and ∣z + 2z∣.

Concept: Many complex number concepts have a strong geometric significance as

Q uj well as an algebraic one, so if you have trouble understanding an algebraic
relationship regarding complex numbers, try considering it in the complex
plane.

Exerci⅜es

3.4.1 Graph each of the following equations on the complex plane.

(a) z+z =4 (c) ∣z - 2 + z∣ = 6
(b) (1 - 2z)z + (-1 - 2z)z = -6z (d) ∣z - i∖ = ∣z + z∣

73
CHAPTER 3. COMPLEX NUMBERS

3.4.2 In each of the following parts, find an equation whose graph in the complex plane is the following:

(a) A circle centered at -2 + 5i with radius 3 √2.

(b) The perpendicular bisector of the segment connecting 4 + i and 7 - 2i.
(c) The line that passes through 3 + i and has slope 2.

3.4.3 Let S be the set of points z in the complex plane such that (3 + 4z)z is a real number. Describe the
graph of S. (Source: AMC 12)
3.4.4* Let O be the origin of the complex plane, and let w and z be two nonzero complex numbers.
Draw the line joining O to w, and the line joining O to z. Show that these two lines are perpendicular if
and only if w/z is imaginary.
*
3.4.5 Let u be a complex number, and let v be the reflection of u over the line Re(z) = Im(z). Express
υ in terms of u and u. Hints: 301

3.5 Summary

A complex number is a number of the form a + bi, where a and b are real numbers and z2 = -1.

Definitions: For the complex number a+bi, where a and b are real, we have the following definitions:

• We call a the real part and b the imaginary part of a + bi. We use the notation Re(z) and Im(z)
to denote the real and imaginary parts of a complex number z. So, if z = a + bi, then we have
Re(z) = a and Im(z) = b.

• The conjugate of a + bi is a - bi, and is denoted by a + bi.

• The magnitude of a + bi is defined as Vα2 + b2, and is denoted by ∖a + bz∣.

We add two complex numbers by adding their real parts and adding their imaginary parts:

(a + bi) + (c + di) =a + c + bi + di = (a + c) + (b + d)i.

We multiply two complex numbers by using the distributive property and the fact that z2 = -1:

(a + bi)(c + di) = (ac - bd) + (ad + bc)i.

For every complex number z = a + bi, the number zz = a2 + b2 is real. We can use this to write the
ratio of two complex numbers as a single complex number:

w _ wz _ wz
z zz ∣z∣2

If a + bi = c + di where a, b, c, and d are real numbers, then a = c and b = d. In other words, complex
numbers are equal if and only if both their real and imaginary parts are equal.

74
 3.5. SUMMARY

Concept: In general, if you are given a problem involving a complex number z, it

O~ τ,J may help to write z in the form a + bi, and then express the problem in
terms of ∩ and b.

For any complex numbers w and z, we have

(i) z + w = z + w.

(ii) zw = zw.

(iii) z = z.

(iv) zz = ∣z∣2.

(v) ∖zw∖ = ∣z∣∣w∣.

(vi) Re(z) = zl^-z∙ and Im(z) =

A complex number can be represented as a point in the

complex plane. Like the coordinate plane, the complex
plane has two axes, the real axis for the real part, and the
imaginary axis for the imaginary part. So, the complex
number x + yi corresponds to the point (x, y) in the Cartesian
coordinate plane. The figure at right shows the numbers
1 + 2z', -5 — 3z, and 2 - 4z plotted in the complex plane. Just
as we can graph equations in terms of two real variables on
the Cartesian plane, we can graph equations in terms of a
single complex variable on the complex plane.
The magnitude of a complex number has a geometric
significance, as well. For any complex numbers w and z, the expression ∣z∣ equals the distance from z to
the origin in the complex plane, and the expression ∖w - z∖ equals the distance between w and z in the
complex plane.

Problem Solving Strategies________ __ ι__________________ _______

Concepts: ∙ Factoring and other manipulations can make some calculations eas-
O≡≡^ ier∙
• Trying specific examples can be a good way to discover general
relationships.

• Often, dealing with ∣z∣2 is nicer than dealing with ∣z∣, since squaring
∣z∣ gets rid of the square root sign in the definition of ∣z∣.

• Many complex number concepts have a strong geometric significance

as well as an algebraic one.

75
CHAPTER 3. COMPLEX NUMBERS

Review Problems

3.23 Let w = 2 + 3z and z = 4 - 5i. Express each of the following as a complex number:

(c) =Λ- -y + xι
(a) 2w - 3z (e)
w+z x+ψ
1 w3 + 2w2z + wz2 (1 - >)4
(b) (d) 2 2 (f)
w w2z + wz2 (1 + 03

3.24 Solve the equation -3 — x2 = 9 + x2.

1
3.25 Simplify---------- ------ .
1÷
1-l÷i

3.26 Find the complex number z such that ------ = -1 + z.

1 +Z

3.27 Four complex numbers are plotted in the plane, as shown. One of them is
z. The other three are -z, z, and -z. Which is which?
3.28 Let w and z be complex numbers.
(a) Show that wz + wz is real.
(b) Show that wz - wz is imaginary.
3.29 Find two complex numbers whose squares equal 5 - 12z.
3.30 Find all real numbers c such that 7 + z is 5 units from 10 + ci on the complex plane.

3.31 Find the area of the region enclosed by the graph of ∣z - 4 + 5z∣ = 2√3.

3.32 The diagram to the right shows several numbers in the complex plane.
The circle has radius 1 and is centered at the origin. One of these numbers is
the reciprocal of F. Which one? (Source: AHSME)
3.33 Find all complex numbers z such that ∣z + 1 - z∣ = ∣z - 2∣.
3.34 The product of the complex numbers a + bi and c + di is real, where a,
b, c, and d are real numbers. Prove that the product of the complex numbers
b + ai and d + ci is also real.
3.35 Describe the graph of each of the following:
(a) z-z = -8z (b) (4 - z)z - (4 + z)z = 16z (c) ∣7 + z-2z∣ = 4

3.36 Graph ∣z - 5 + 2z∣ ≤ 4 on the complex plane.

3.37 Simplify (∕ + 1)3200 - (∕ - I)3200.

76
 CHALLENGE PROBLEMS

3.38 Show that ∣z - 1∣2 + ∣z + 1∣2 = 4 for all z such that ∣z∣ = 1.

Challenge Problems

3.39 Evaluate i + 2z2 + 3z3 + 4z4 + ∙ ∙ ∙ + 64z64.

3.40 Find all complex numbers z such that ∣z - 1∣ = ∣z + 3∣ = ∣z - i∣. (Source: ARML) Hints: 163
3.41 Find the area of the region of points z in the complex plane such that

∣z + 4 - 4z∣ ≤ 4√2 and ∣z - 4 - 4z∣ ≥ 4√2.

ZU zv
3.42 Let w and z be complex numbers and z ≠ 0. Prove that — 4— is a real number.
z z
3.43 Show that the points w, x, y, and z are the vertices of a parallelogram in the complex plane if and
only if the sum of some two of them is equal to the sum of the other two. Hints: 56
3.44
* Define a sequence of complex numbers by Zι = 0, zπ+ι = z2 + z' for n≥l. How far away from the
origin is z1117 (Source: AHSME) Hints: 60
3.45
* A function f is defined on the complex numbers by ∕(z) = (a + bi)z, where a and b are positive
numbers. This function has the property that the image of each point in the complex plane is equidistant
from that point and the origin. Given that ∖a + bz∣ = 8, find the value of b2. (Source: AIME) Hints: 330,146
3.46 For a nonzero complex number z, let ∕(z) = l∕z.
(a) Show that f(f(z)) = z for all z ≠ 0.
(b) * Let w = ∕(z). As z varies along the line (1 + 2z)z - (1 - 2z)z = z, what curve does w trace?
Hints: 59
3.47
* Find all ordered pairs (a, b) such that a + -^+ = 2 and b + ⅛—X = 0. Hints: 143
a 2 + b 2- a2 + b2
3.48
* Let u and v be complex numbers such that ∣u∣ = ∖v∖ = 1, u ≠ v, and u ≠ -v. Show that

4wυ
(u + υ)2

is real, and that

u+v
u-v
is imaginary. Hints: 334

77
CHAPTER 3. COMPLEX NUMBERS

Extra! Suppose we defined a "special number" as an ordered pair of real numbers, so that
"* ι*∙""' (3z _2) and (-2,7) are "special numbers." Next, let's define addition and multiplication
of these "special numbers" as follows:

(a, b) + (c, d) = (a + c,b + d),

(a, b) ■ (c, d) = (ac - bd, ad + be).

It turns out that this definition of "special numbers" can be used as an alternative
definition of complex numbers. To get a feel for why, let's compare a few sums and
products of "special numbers" with their /-notation counterparts:

(3,4) ∙ (-2,1) = (-10,-5)

(3 + 4∕) ∙ (-2 + ∕) = -10 - 5i

(0,1) ∙ (0,1) = (-1,0) (4,-l)∙ (4,1) = (17,0)

i∙i = -1 (4 - ∕)(4 + z) = 17

So, we see that the first number in the ordered pair (α,b) of a "special number"
corresponds to what we have been calling the "real part" of a complex number and
the second number in the ordered pair corresponds to what we have been calling the
"imaginary part." (If you don't see why, note that (1,0) ∙ (1,0) = (1,0), but (0,1) ■ (0,1) =
(-l,0).)
We can also define the magnitude of (a, b) the same way we have defined ∖a + bi∖ by
letting (a, b) = Vfl2 + f2. Moreover, we can define the conjugate of (a, b) as (a, b) = (a, ~b).
Just to make sure, let's check the product of (a, b) and its "conjugate:"

(a, b) ■ (a, -b) = (a2 - b2, -ab + ab) = (a2 - b2,0).

As expected, the second component of the product is 0.

This ordered pair notation very obviously lends itself to the geometric depiction of
complex numbers on the complex plane. Moreover, it was this formal definition of
complex numbers that helped mathematicians accept the concept of nonreal numbers.
The first use of imaginary numbers is often attributed to Girolamo Cardano in the
16th century, who referred to square roots of negative numbers. For around two cen­
turies thereafter, mathematicians used the notation V—T until the great mathematician
Leonhard Euler introduced i in 1777.
Despite all the interesting results mathematicians found regarding imaginary num­
bers, many were very uncomfortable with them, since they appeared to have no "physi­
cal" meaning. Mathematicians would add graphical representations of complex numbers
in the 17th and 18th centuries, but it wasn't until 1833 that Lord William Rowan Hamilton
gave complex numbers the definition described above. With this description of complex
numbers, mathematicians could use complex numbers without having to ever encounter
such an intuitively "unreal" concept as V-T.
(Source: MacTulor History of Mathematics archive)

78
 Λ
* mMM j<
Γm very well acquainted, too, with matters mathematical,
I understand equations, both the simple and quadratical. - Gilbert & Sullivan

CHAPTER 4 ι
Quadratics

An expression of the form

ax2 + bx + c,
where a, b, and c are constants and a ≠ 0, is called a quadratic. The solutions to the equation ax2+bx+c = 0
are called the roots of the quadratic. The roots are also sometimes called zeros of the quadratic. The ax2
term of ax2 + bx + c is called the quadratic term, the bx term is the linear term, and c is the constant term.
In this chapter, we review techniques for manipulating and solving quadratic equations.

4.1 Factoring Quadratics

When we expand the product of two linear expressions, such as

(x - 2)(x - 3) = x2 - 5x + 6,

the result is a quadratic. When we reverse this process, writing a quadratic as the product of two linear
terms, we say we factor the quadratic.

I Problems

Problem 4.1: Find the solutions to each of the following equations.

(a) x2 - 5x + 6 = 0
(b) x2 + x - 20 = 0 .
(c) r4-13r2 + 36 = 0

79
CHAPTER 4. QUADRATICS

Problem 4.2: The quadratic x2 - 12x - 540 has two integer roots. In this problem, we find those roots.
(a) How can we tell from the coefficients of the quadratic that both roots are divisible by 6?
(b) Find the roots of the quadratic.

Problem 4.3: Find the solutions to each of the following equations:

(a) 3f2 +Ilf+ 10 = 0 (c) 36/-25 = 0
(b) 2x2 -7x = 15 (d) 9f2 - 2f + 5 = 25 - 90f

Problem 4.4:
(a) Factor the expression 6x2 + x - 15.
(b) Find all possible values of the ratio of x to y if x and y are nonzero numbers that satisfy the
equation 6x2 + xy = 15y2.

Problem 4.5: How many integers x satisfy the equation (x2 — x — l)x+2 = 1? (Source: AHSME)

We start by factoring quadratics in which the coefficient of the quadratic term is 1.

Problem 4.1: Find the solutions to each of the following equations.

(a) x2-5x +6=0 (c) r4-13r2 + 36 = 0
(b) x2 + x - 20 = 0

Solution for Problem 4.1: We know that the product (x + p)(x + q) will give us an x2 term when expanded,
so we guess that the quadratics in parts (a) and (b) can be factored in this form. Expanding this product
gives
(x + p)(x + q) = x2 + px + qx + pq = x2 + (p + q)χ + pq.

To factor each quadratic, we must find the appropriate p and q.

(a) From the coefficient of x, we have p + q = -5. The constant term tells us that pq = 6. Because pq is
positive, the numbers p and q have the same sign. Because p + q is negative, we know that they
are both negative. Now we just try different pairs of p and q that multiply to 6, looking for a pair
that add to -5. We find that if p and q are -2 and -3, then p + q = -5 and pq = 6, so

x2 - 5x + 6 = (x - 2)(x - 3).

Therefore, our equation is (x - 2)(x - 3) = 0, so our solutions are x = 2 and x = 3.

(b) Here, we have p + q = 1 and pq = -20. Because pq is negative, p and q have opposite signs. Trying
different pairs of numbers that multiply to -20, we find that 5 and -4 satisfy both equations. (We
might have tried this first, since p + q - 1 tells us that p and q are not far apart in magnitude.) So,
we have (x + 5)(x - 4) = 0, which gives us x = -5 and x = 4 as solutions.
(c) This equation isn't a quadratic, but we can still factor it as one. If we don't see how at first, we can
let z = r2, which makes our equation z2 - 13z + 36 = 0.

80
 4 1, FACTORING QUADRATICS

Concept: Some equationscan be turned into quadratic equations with a substitution

O≡^

Factoring the left side gives (z - 4)(z - 9) = 0, from which we have 4 or 9. Since r, we
have r - 4 or r = 9. From r2 = 4, we have r ≡≈ 2 and r = -2, and r = 9 gives us r = 3 and r - -3
as solutions

□
let’s try a slightly harder quadratic.

Problem 4.2: Find the roots of f(χ) = χ2 - 12x - 540.

Solution for Problem 4.2: Suppose ∕(x) = (x + p)(χ + q). As before, we have pq = -540 and p + *∕ ~ —12. We
could start searching through all the pairs of numbers that multiply to -540. but that might take quite a
while.

Instead, we first notice that 12and 540 are both even. Ifxisodd, then v2 is odd while both I2xand
—540 are even, so the expression x2 - 12x - 540 is odd. This tells us that the quadratic cannot equal 0 it ι
is odd. So, if the quadratic has integer roots, they must be even.
Next, we notice that 12 and 540 are both divisible by 3. Therefore, the last two terms of t 12 v 140
are always divisible by 3 if x is an integer. So, ∕(x) can only equal 0 for an integer value of ι if ι is
divisible by 3, since otherwise x2 — 12x — 540 will not be divisible by 3. This tells us that both root» are
divisible by 3.
Now, because an integer that is divisible by 2 and by 3 must also be divisible by 6, we know tfι.ιt both
roots arc divisible by 6 if they are integers. If ∕(x) = (x + p)(x + q), then the roots are -p and q. So, if the
roots are integers, then both p and q must be divisible by 6. Therefore, we let ρ 6p∣ and q 6qi, and
we have (6p∣)(6<∕j) = -540 and 6pι + 6q↑ = -12. Simplifying these gives us piq↑ =■ -15 and p∣ f ∕∕∣ 2.
These are much easier to handle than our original equations! We quickly find that p1 and ι∕∣ are 5 and
3, so p and q are 6 ■ (-5) = -30 and 6 ∙ 3 = 18. This gives us

x2 - 12x - 540 = (x - 30)(x + 18),

so the roots of ∕(x) are x = 30 and x = -18. □

Concept: Using a little number theory can help factor quadratics.

o≡≡≡
Let's try some quadratics in which the coefficient of the quadratic term is not 1.

81
CHAPTER 4. QUADRATICS

Solution for Problem 4.3:

(a) Because the quadratic term is 3t2, we know that (t + p)(f + t∕) won't do, since this will produce
a t2 term instead of a 3t2 term. Instead, we try (3f + p)(f + g), since this will produce a 3f2 term.
Expanding this product gives

(3t + p)(t + t∕) = 3t2 + (p + 3q)t + pq.

So, we seek p and q such that p + 3q = 11 and pq = 10. Therefore, p and q are both positive. Trying
different pairs of numbers that multiply to 10, we find p = 5 and q = 2. (Notice that it does matter
which is p and which is q.) We therefore have (3t + 5)(f + 2) = 0, so we must have 3t + 5 = 0 or
t + 2 = 0. Solving these two equations gives the solutions t = —5/3 and t = —2.
(b) First, we move all terms to the left, which gives 2x2 — 7x — 15 = 0. We guess that the quadratic
factors as (2x + p)(x + q). Expanding this product gives 2x2 + (p + 2q)x + pq, so we seek p and q
such that p + 2q = -7 and pq = -15. We know that p and q have opposite signs, and we find that
p = 3 and q = -5 satisfy both equations. So, we have 2x2 - 7x - 15 = (2x + 3)(x - 5). This gives us
x = -3/2 and x = 5 as solutions to the original equation.
(c) We can solve this equation without factoring. Adding 25 to both sides gives 36y2 = 25, then taking
the square root of both sides gives 6y = ±5. Therefore, our solutions are y = ±5/6.
We also can use the difference of squares factorization, a2 — b2 = (α — b)(a + &), to note that
36y2 -25 = (6y - 5)(6y + 5).
Important:

ψ a2 - b2 = (a - b)(a + b)

We'll use the difference of squares factorization often throughout this book.
(d) Putting all terms on the same side gives us 9t2 + 88f - 20 = 0. We would like to factor this as
(At + B)(Ct + D). Expanding this product and comparing the result to our quadratic gives us

AC = 9,
AD + BC = 88,
BD = -20.

From AC = 9, we see that A and C are factors of 9, and from BD = -20, we know that B and D are
factors of 20. Next, we consider the middle equation. Because 88 is much larger than 9 and 20, we
know that we must choose large factors of 9 to pair with large factors of 20 when assigning values
of A, B, C, and D. In other words, AD + BC is pretty large, so we need both A and D (or both B
and C) to be pretty large.
For example, if we let A = 3, C = 3, B = -4 and D = 5, then AD + BC = 3, which is way too
small. If we instead let A = 9 and D = 10, we have C = 1 and B = -2, so AD + BC = 88. This gives

9t2 + 88t - 20 = (9t - 2)(t + 10),

which means the solutions are t = 2/9 and t = -10.

82
 4.1. FACTORING QUADRATICS

Concept: Factoring quadratics is just a number game in which we use the signs, the
divisors, and the magnitudes of the coefficients of the quadratic to find the
quadratic ,s factors.

Problem 4.4: Find all possible values of the ratio of x to y if x and v are nonzero number ∙ such that
* + xy - 15y2.
6x

Solution for Problem 4.4: What's wrong with this solution:

Bogus Solution: Subtracting 10xy from both sides of the equation gives us t>v2-lh t∕
-10xy + 15y2. We can then factor both sides as

3x(2x - 3y) = -5y(2x - 3y).

Dividing both sides by 2x - 3y, we get 3x = -5y. so x/y V3

While this solution begins with a nice, clever step to allow us to factor, dividing that factor out might
involve division by 0, so we lose a possible solution. Here are two valid solutions
Solution 1: Keep your eye on the ball. We seek the ratio of x to ι∕, so we let r -. x/y. which means ∖ ry.
The given equation then is
6(ry)2 + (ry)y = 15y2.
Since y is nonzero, we can divide by y2, which gives 6r2 + r = 15. Rearranging and factoring this equation
gives us (3r + 5)(2r - 3) = 0, so r = -5/3 or r = 3/2.
Solution 2: Il's a quadratic;factor it. Subtracting 15y2 from both sides gives 6x2 + xy 15y2 0. FSven
though there are two variables, the expression 6x2 + xy - 15y2 is still a quadratic expression. We can
view this as a quadratic with x as the variable and y as a constant. However, if we don't see at first how
to factor it, we can look at a simpler, related problem.

Concept: When facing a problem you don't know how to do, try relating it to a
∙p3 problem you know how to do.

Letting y = 1 gives 6x2 + x - 15, which we can factor as (2x - 3)(3x + 5).
Returning to 6.r2 + xy - 15y2, we see that 6x2 + x - 15 is the same as 6x2 + xy - 15y2, but with the y's
removed. So maybe the factorization of 6x2 + xy - 15y2 is the factorization of 6x2 ÷ x - 15, but with the
y's "put back in." Let's try it:

(2x - 3y)(3x + 5y) = 2x(3x + 5y) - 3y(3x + 5y) = 6x2 + xy - 15y2.

Success! We have 6x2 + xy - 15ι∕2 = (2x - 3y)(3x + 5y). Solving (2x - 3y)(3x + 5y) = 0 gives us x/y = 3/2
and x/y = -5/3, as before. □
Here is an example of how factoring can help solve more complicated equations.
CHAPTER 4. QUADRATICS

Problem 4.5: How many integers x satisfy the equation (x2 - x- l)v+2 = 1? (Source: AHSME)

Solution for Problem 4.5: If a and b are integers such that ab = 1, then at least one of the following must
be true:

• b = 0 and a ≠ 0. Any nonzero number raised to the 0th power equals 1. So, if x + 2 — 0 and x2 — x — 1
is nonzero, then x is a solution to the equation. Solving x + 2 = 0 gives us x = -2, for which
x2 - x — 1 = 5. So, x = -2 is a solution.

• a = 1. We note that 1 raised to any power is 1. If x2 - x - 1 = 1, then rearranging and factoring

gives (x - 2)(x + 1) = 0, so x = 2 and x = -1 are solutions since any power of 1 is equal to 1.

• a = -1 and b is even. If x2 - x - 1 = -1, rearranging and factoring gives x(x - 1) = 0, so x = 0 and

x = 1 need to be considered. When x = 0, we have (x2 — x — l)x+2 = (—1)2 = 1, so x = 0 is a solution.
When x = 1, we have (x2 - x - l)x+2 = (-l)3 = -1, so x = 1 is not a solution.

In total, there are 4 integer solutions: -2, -1,0, and 2. □

I xeι<i⅜es

4.1.1 Factor and find the roots of each of the following quadratics:
(a) x2 + 19x + 90 (c) 6a2 + 13α - 8
(b) 2y2-15y + 18 (d) 7x2 + 296x-215

4.1.2 Find the solutions to each of the following equations:

679 ________
(a) 3x4 + 26 = 19x2 (b) x+~ = lθ4 (c) √x2 + x - 7 = √x≡73

4.1.3 Find all possible values of a/b if α2 + 4b2 = 4ab.

4.1.4 Fix the Bogus Solution to Problem 4.4 by rearranging the equation 3x(2x - 3y) = -5y(2x - 3y) into
an equation that's easy to factor.
4.1.5 Find the solutions to (x4 - llx3 + 24x2) - (4x2 - 44x + 96) = 0.

4.2 Relating Roots and Coefficients

While factoring quadratics in which the coefficient of the quadratic term is 1, you might have noticed
an interesting relationship between the roots of a quadratic and its coefficients. If not, find the roots of
x2 - 7x + 10 and see if you can find the relationship.
In this section, we explore this relationship in more detail.

84
 42 REL ATING R∞TS AND COEFFICIENTS

Problem
*

Problem 4.6: Suppose the πhHs of the quadratic <∕v f bx + c are r and
(a) Why must we have a.v ÷ bx + c - a(x - r)(x - s) for all values of t7
(b) Express r + .- in terms of only a and b.
(c) Express rs in terms of only a and c.

Problem 4.7: Find the sum and the product of the roots of each of the following quadratic equations
j (a) x2-13x + 36≈0 (b) -2006√ = 2007r ÷ 1977

∣--- ~--- ------ --- “-- - ' ' ‘--- - ' ---- --- ---- -- —-------------- ----- ------ . . l -................... . ... ...... ... —-------------------
Γroblem 4.8: Let r and s be the roots of x2 - pχ + q - 0. In this problem, we find r + $ in terms of p
and q.
(a) Find equations relating r and s to p and q.
(b) Express r + s2 in terms of r + s and rs.
(c) Find r2 + s2 in terms of p and q. (Source: AHSME)

Problem 4.9: li∣nl real numbersα and biffl+ 2/and ft-5/are the roots o| v t ( 1» ,∙m S(t3 + i) 0

Problem 4.10: The sum of the zeros, the product of the zeros, and the sum of the iwlitcients of the
function ∕^(x) = ax2 + bx + c are equal. Show that all three of these quantities must equal ι∕. (Source
AMC 12)

Problem 4.6: Express the sum and product of the roots of the quadratic ax2 + bx + c in terms of a, b,
and c.

Solution for Problem 4.6: In order to look for a relationship between the roots and the coefficients of a
general quadratic, we write the quadratic using its roots. Suppose the roots are r and s. Then, the
product (x - r)(x - s) will give us a quadratic with roots r and s. However, to produce the quadratic
ax2 + bx ÷ c, we must have a as the coefficient of the quadratic term. So, we multiply (x - r)(x - s) by a,
and we have
λx2 + bx + c = a(x - r)(x - s). (4.1)

Expanding the right side of Equation (4.1), we have

ax1 + bx + c = (ix2 - a(r ÷ s)x + ars. (4.2)

These two quadratics are the same for all values of x, so the corresponding coefficients of the two
quadratics must be the same. Equating the coefficients of each power of x, we get

a = a,
b = -rt(r + s),
c ≈ ars,

85
CHAPTER 4. QUADRATICS

While the first of these equations is true by design (we chose to multiply (x - r)(x - s) by a in order to
match the coefficients of the x2 term in Equation (4.2)), the other two equations tell an interesting story.
Isolating the sum and product of r and s in these equations, we have
b , c
r+s = — and rs = -.
a a
In other words, the sum of the roots of ax2 + bx + c is -b∕a, and the product of those roots is c∣a. □
These relationships between the roots and coefficients of a quadratic are called Vieta's Formulas.

Important: If r and s are the roots of the quadratic ax2 + bx + c, then

ψ i λ c
r + s- — and rs = -.
a a

When using these formulas, we will often simply write, "By Vieta...." For example, by Vieta, we
know that the sum of the roots of x2 - 4x + 3 is-(-4)∕l = 4.

Problem 4.7: Find the sum and the product of the roots of each of the following quadratic equations:
(a) x2-13x + 36 = 0 (b) -2006x2 = 2007x + 1977

Solution for Problem 4.7:

(a) We can factor this quadratic as (x - 4)(x - 9) = 0, so the roots are 4 and 9. Or, we could have used
Vieta to find that the sum of the roots is -(-13)∕l = 13 and the product of the roots is 36/1 = 36.
(b) Factoring doesn't look like much fun on this problem, so let's try Vieta. What's wrong with this
solution:
Bogus Solution: We have a = —2006, b = 2007, and c = 1977, so the sum of the roots is
-b∕a - 2007/2006 and the product of the roots is c/a = -1977/2006.

The problem here is that Vieta's Formulas are for quadratics in the form ax2 + bx + c. So, we have
to move all terms to one side of the equation before proceeding with Vieta. Reorganizing the
terms on one side of the equation gives us 2006x2 + 2007x + 1977 = 0. So, the sum of the roots is
-2007/2006 and their product is 1977/2006.

□
We can also use Vieta to find other relationships between the roots and the coefficients of a quadratic.

Problem 4.8: If r and s are the roots of x2 - px + q = 0, then find r2 + s2 (in terms of p and t∕). (Source: ∣
AHSME)

Solution for Problem 4.8: While it's not immediately obvious how to express r2 + s2 in terms of p and t∕,
we can use Vieta to relate r and s to p and q:
r + S = p,
rs = q.

86
 4.2 RELATING R∞TS AND COEFFICIENTS

Our goal is to find an expression for »' + But Vieta formulas don't directly provide us with
any expressions involving squares ol the roots However, we can get an equation involving r' + s by
squaring both sides of ? 4 ι- p. This gives us r 4 2r< + s2 p'. Now, subtracting 2r≤ from both sides
gives us what we want
r2 + ff = p2 -2rs = p2 - 2q.

∖∖e call a two-variable expression symmetric it it remains the same w hen the variables an swapped
I or example, r ÷ s2 is symmetric because swapping the variables gives s2÷r2. which is the same as r ÷
But r + 2s is not symmetric because swapping the variables gives s + 2r, which is different from r + 2∙
As shown in Problem 4.8, we can use r + s and rs to evaluate more complicated symmetric expressions
in r and s.

Concept: Symmetric expressions involving the roots of a quadratic are a clue to try
O≡¾⅞¾ using Vieta's Formulas.

Problem 4.9: Find real numbers a and b if a + 2i and b - 5i are the roots of r t (- 4 , 30λ * (∣3 4 ∕) 0 ∣

Solution for Problem 4.9: Nothing in our proof of Vieta's Formulas prevents the roots or the coefficients
of the quadratic from being non real. So, we can use Vieta for quadratics in which the roots and/or the
coefficients are not real. By Vieta, the sum of the roots of the given quadratic is (-4 + 3i)∕ 1 I W, .o

4 - 3i = a + 2i + b - 5i = (a + b) - 3i,

and the product of the roots is (13 + ∕)∕1 = 13 + ∕, so

13 + i = (it + 2i)(b - 5i) = ab- 5ai + 2bi - 10r2 = (ab + 10) + (-5α + 2b)i.

Equating real parts of (fl+b)-3i = 4-3ιgivesfl+b = 4. Equating the real parts of (Λb+10)+(-5a+2b)∕ 13+i
gives ab + 10 = 13, and equating the imaginary parts gives -5a + 2b = 1.
Solving the system of linear equations a + b = 4 and -5α + 2b = 1 gives us a = 1 and b = 3, which is
consistent with the equation ab + 10 = 13. Therefore, we haveα = 1 and b = 3. □

Problem 4.10: The sum of the zeros, the product of the zeros, and the sum of the coefficients of the ∣
function f(x) = ax2 + bx + c are equal. Show' that all three of these quantities must equal a. (Source: I
AMC 12)

Solution for Problem 4.10: By Vieta, the sum of the zeros is -b∕a and their product is c∕a. So, we can write
the given information as
b c
— = - = a + b + c.
a a
From the first two parts of this equation chain, w,e see that -b = c. Therefore, we have <7+b+ι urb b a,
so the sum of the coefficients, and therefore the sum and product of the zeros, equals a.

87
CHAPTER 4. QUADRATICS

I vrι < iMk⅜

4.2.1 Find the sum and product of the roots of each of the following quadratic equations:
(a) x2-5x + 7 = 0 (c) 12x2 + 13x + 14 = 15x2 + 16x + 17
(b) 2x2 = 7x + 120

4.2.2 Find the product of the roots of the equation (x - l)(x - 2) + (x - 3)(x - 4) + (x - 5)(x - 6) = 0.
4.2.3 One root of the equation a(b - c)x2 + b(c - a)x + c(a - b) = 0 is x = 1. Find the other root in terms
of a, b, and c. (Source: AHSME)
4.2.4 Let p(x) = 2x2 + 7x + c, where c ≠ 0. If r1 and r2 are the roots of p(x), then find the value of ⅛ + ⅛
in terms of c.
4.2.5 Let m and n be the roots of αx2 + bx + c = 0. Prove that if m2 + n2 = 1, then 2ac = b2 — α2.

4.2.6 Back on page 85, we made the claim that if two quadratics are the same for all values of the
variable, then the corresponding coefficients of the two quadratics are the same. In other words, if
ax2 + bx + c = dx2 + ex + f for all x, then we must have a = d, b = e, and c = f. However, we didn't prove
this fact! Remedy that oversight now by proving it yourself.

4.3 Completing the Square

Not all quadratics can be factored easily. For example, try factoring x2 + 6x - 1. However, we can still
find the roots of such quadratics using a process called completing the square. When we complete the
square of a quadratic ax2 + bx + c, we write it in the form

a(x - h)2 + k,

for some constants a, h, and k. In this section, we review how to complete the square, and how to use
completing the square to solve quadratic equations.

Problems

Problem 4.11: Expand each of the following squares of binomials:

(a) (x + 4)2 (b) (4f - 5)2

Problem 4.12: Suppose A: is a constant. In terms of k, what constant must be added to x2 + kx to produce
a quadratic that is the square of a binomial?

Problem 4.13: Complete the square in each of the following quadratics by writing each in the form
n(x + p)2 +1], where a, p, and q are constants.
(a) x2 + 8x + 2 (b) -6x2 - 18x - 7

88
 4 3. COMPLETING THE SQUARE

Problem 4,14: l or each of the eqιι.Uυ∙n∙- below, add a <<>nfant to both -ides su< h that the left side
becomes the square of a binomial I hen, solve the equation.

(a) t2 -4f - 18
(b) 4y2 + 24v = -37

Problem 4.15: Find a formula for the roots of the quadratic equation ∣ιv2 + bx + / where √ b. and art­
constants and a ≠ 0. Test your formula on the quadratic r2 - 5x + 6 = 0. Does your formula give the
correct roots?

Γroblcm 4.16: Find the solutions of each of the following equations:

(a) x2 4 4x + 7 = 0
(b) 3x2 = 10x + 24
(c) * x2 — (11 ih (i

Problem 4.17: I w< > numbers sum to 6 and have product -19. f ind t >

Problem 4.18: Suppose∕(x) = x+√xand^(x) = x+l∕4. In this problem, we evaluate χ( fr(χ(∕(Λ,(∕(7))))))

(a) Express g(∕(x)) as the square of a binomial.
(b) Use your answer to part (a) to find ^(∕(g(∕(g(∕(7)))))) quickly.

Problem 4.11: Expand each of the following squares of binomials:

(a) (x + 4)2 (b) (4f - 5)2

Solution for Problem 4.11:

(a) We have (x + 4)(x + 4) = x2 + 4x + 4x + 4^ = x2 + 8x + 16.
(b) We have (4f - 5)(4∕ - 5) = (4∕)2 - 5(4f) - 5(4f) + (-5)2 = 16t2 - 40f + 25.

□
Problem 4,11 gives us two examples of the expansion of the square of a binomial. In general, we
have
(a + b)2 = a~ + ab + ab + b2 = a2 + 2ab + b2.

------- ------ ------------- -- --------------------- --------- —w------------------------------------------- , -

Important:

ψ (a + b)2 = ar + 2ab + b2

Problem 4.12: Suppose k is a constant. In terms of k, what constant must be added to .v + ⅛,v to
produce a quadratic that is the square of a binomial?

89
CHAPTER 4. QUADRATICS

Solution for Problem 4.12: We have x2 + 2xb + b2 = (x + b')2. We can turn the 2xb into kx by letting b = k∕2,
which gives us
x2 + fcx + θθ =(x+|

So, if we add (k∕Z)2 to x2 + kx, we produce a quadratic that is the square of a binomial. □

Important: Adding the square of half the coefficient of the linear term to a quadratic
of the form x2 + kx produces a quadratic that is the square of a binomial.

Let's see how we use this to complete the square.

Problem 4.13: Complete the square in each of the following quadratics by writing each in the form
a(x + pj2 + q, where a, p, and q are constants.

(a) x2 + 8x + 2

Solution for Problem 4.13:

(a) We must add (8∕2)2 = 16 to x2 + 8x in order to produce a quadratic that is the square of a binomial.
However, when we add 16 to an expression, we must also subtract 16, so that our new expression
is still equivalent to the original x2 + 8x + 2. This gives us

x2 + 8x + 2 = x2 + 8x + 16 - 16 + 2 = (x + 4)2 - 14.

(b) Let ∕(x) = -6x2 - 18x - 7. We know what to do when the coefficient of x2 is 1, so we factor -6 out
of the first two terms, to give

f(x) - -6x2 - 18x - 7 = -6(x2 + 3x) - 7.

To complete the square in x2 + 3x, we must add (3∕2)2 = 9/4 to the quadratic. However, if we just
add 9/4 to f(x), we have
∕(x)+∣ = -6(√ + 3x)-7+∣.

This doesn't give us a square of a binomial inside the parentheses. In order to get the 9/4 inside
the parentheses, we instead add -6(9/4) to f(x):

∕(x) + (-6) (?) = -6(x2 + 3x) - 7 + (-6) (?) = -6 (x2 + 3x + ?) - 7.

Subtracting (—6)(9∕4) from both sides of ∕(x) + (-6) Q) = -6 (x2 + 3x + ∣) - 7 gives

f(x) = -6 (x2 + 3x + ?) - 7 - (-6) (?) = -6 (x + ∣) ÷≡.

90
 4.3. COMPLETING THE SQUARE

Note that this is the same as starting with f(x) 6f.τ∙f ∙ W) 7, and then both adding and
subtracting (-6)(9∕4) to the right side

∕(<) -- -ħ(√ + 3x)-7= -ħ(√ + 3ι)÷(-6)Q-( 6)( j)-7

,∕ 3? 13
= -6(x + -) +y.

We can check our answer by expanding our final result. This gives us

-β(x + + y - -fi(x2 + 3x-ty)+ y = -fix2 - 18x -7,

which is indeed our original quadratic.

O

Now that we know how to complete the square, let's use our new tool to solve a couple equations

Problem 4.14: For each of the equations below, add a constant to both sides such that the left side
becomes the square of a binomial. Then, solve the equation.
(a) t2 - 41 = 18 (b) 4y2 + 24y = -37

Solution for Problem 4.14:

(a) We add (-4∕2)2 = 4 to both sides of our original equation to produce the square of a binomial on
the left side:
t2 - 4f + 4 = 18 + 4.
As planned, the left side is now the square of a binomial, so we have (t - 2)2 22. Taking the
square root of both sides, we have t - 2 = ± yf22, so our solutions are t = 2 ± >∣22.
(b) We factor 4 out of the left side, which gives us 4(y2 + 6y) = -37. Now, we want the quadratic on
the left side to be the square of a binomial. We recognize that (y + 3)2 = y2 + 6y + 9, so we need to
add 9 inside the parentheses on the left. To do so, we add 4(9) to both sides, which gives us

4(y2 + 6y) + 4(9) = 4(y2 + 6y + 9)

on the left and -37 + 4(9) = -1 on the right. Factoring the quadratic then gives us

4(y + 3)2 = -l.

Taking the square root of both sides of this equation gives us 2(y + 3) = ±i. Dividing by 2 gives us
y + 3 = ±∣, so our solutions are y = -3 ± ∣.
We could also have divided by 4 in our first step, to get y2 + 6y - -l See it you can finish the
problem from here.

91
CHAPTER 4. QUADRATICS

Since we can complete the square for any quadratic, we can complete the square of a general quadratic
to produce a formula that will give us the roots of any quadratic.

Problem 4.15: Find a formula for the roots of the quadratic equation ax2 + bx + c, where a, b, and c
are constants and a ≠ 0.

Solution for Problem 4.15: Since a ≠ 0, we can divide the whole equation by a to make the coefficient of x2

fl fl
equal to 1:
2 b c n
x + -x + - = 0.
2 2
jj. (b∕a∖
This lets us use the result of Problem 4.12. We complete the square by adding I — I I to both

sides:
2 2
2 b
X + -X +
fl (⅛) ÷-fl =
Because x2 + -x +
a
∣| 2 , we now have II x + —b ( b ∖2
I — I . We then isolate the square of
the binomial:

Writing the terms on the right with a common denominator gives us

∕ ∖2 _ c _ b2 4flc _ b2 - 4flc
∖ + 2fl ∕ 4a2 fl 4fl2 4a2 4fl2 '
so we have
2 _ b2 - 4flc
(x+⅛ = 4fl2
Then we take the square root of both sides, and remember the ±
b ∕b2 - 4flc _ ι Vb2 - 4flc _ Vb2 - 4flc
X+ Ta = i V ~~4a2~ ^ i √^2 = 2fl
Subtracting b∕2a from both sides gives us our desired formula:
-b ± Vb2 - 4flc
x=
2fl

WARNING!! In one of our steps above, we replace λ∕4fl2 with 2a. However, it is not
*∙
* true that √4fl2 = 2a if a < 0, so we can't write Vz4fl2 = 2a for all a. Why
is it OK for us to do so above?
While we can't write λ∕4a2 = 2a for all a, we can write ± V4fl2 = ±2fl,
because ± Vlfl2 means "take both numbers that, when squared, give
us 4a2." These two numbers can be depicted as ±2fl whether or not a is
positive or negative.

92
 4.3. COMPLETING THE SQUARE

Important: The quadratic formula Male

* that (hr ∙√d∣ιtion,∙ to the quadratic equation

ψ axi + hx ÷ c = 0

arc
-b ± - 4tf<^
x=
2a

Before we use the quadratic formula to solve a bunch of problems, we should test it on a quadratic
we know how to solve.

C oncept: Using sample cases to test a formula vou deτιvε is a good wax to check
o≡gg your formula.

We'll try x2 - 5x + 6 - 0. We know we can factor the quadratic to get (x- 2)( x -3) 0. so the solutions
are x = 2 and x = 3. Let's see what the quadratic formula gives us. In this case, o 1, b -5, and ∣r ≡ 6,
so
v _ 4-5) ± √(-5)2 - 4(1)(6) 5 ± √25 - 24 5 ± 1
2(1) “ 2 “ 2 '
Therefore, our solutions are x = (5 + 1 )∕2 = 3 and x = (5 - 1)∕2 = 2, as expected.
Let's try it on quadratics we can't factor easily.

Problem 4.16: Find the roots of each of the following:

(a) x2 + 4x + 7 = 0 (b) 3x2 = 10x + 24 (c)
* x2 - (11 - 2∕)x + 23 11/ 0

Solution for Problem 4.16:

(a) Applying the quadratic formula, the roots of x2 + 4x + 7 are

-4 ± √42 -4(1)(7) _ -4 ± √-12 _ -4 ± √12 √≡1 -4 ± (2 √3)(i) r

----------- 2----------- ~-------- ∩------ ------------- 2-------- =----------2-------- ~ ~2±i v3.

(b) We first move all the terms to one side so we can apply the quadratic formula correctly. This gives
us 3x2 - 10x - 24 = 0. So, applying the quadratic formula gives

-(-10) ± √(-10)2 - 4(3)(-24) 10±√388 10±2√97 5±√97

X~ 2(3) 6 " 6 ~ 3~'

We can check this quickly by noticing that Vieta tells us that the sum of the roots of 3x2 - 10x 24
is 10/3 and the product of the roots is -24/3 = -8. Adding and multiplying the roots we found,
we have
5 + √97 5 - √97 10
3 + 3 ^ ~3'
∕5÷ √97∖f5-√97∖ 25-97
I 3 A 3 )= 9

93
CHAPTER 4. QUADRATICS

(c) Nothing in our proof of the quadratic formula assumes that the coefficients are real, so we can use
the quadratic formula when the coefficients are nonreal. Applying the quadratic formula gives us

(11 - 2∕) ± √(11 - 2∕)2 - 4(1)(23 - 11∕) 11-2i± ∖∣YV7 - 44/ - 92 + 44/ _ 11-2/±5
2 - 2 2
Therefore, the roots are
11-2/+ 5 11-2/-5
and
2 2
We can check our work by confirming that the sum of the roots is 11 — 2/ and their product is
23 - 11/, as required by Vieta.

Problem 4.17: Two numbers sum to 6 and have product —19. Find the two numbers.
J
Solution for Problem 4.17: Solution 1: Make a system of equations. Let r and s be the numbers. We then have
r + s = 6 and rs = -19. Solving the first equation for r gives r = 6 - s. Substituting this into the second
equation gives (6 — s)s = —19. Rearranging this equation gives s2 - 6s — 19 = 0. The quadratic formula
then gives us

-(-6) ± √(-6)2 - 4(1)(-19) _ 6 + √36 + 76 6±4√7

= 3 ± 2 √7.
2(1) ^ 2 2
If s = 3 + 2 √7, then r = 6- s = 3- 2√7. Ifs = 3- 2√7, then r = 6- s = 3 + 2√7. In both cases, the two
numbers are 3 -2^^7 and 3 + 2 V7.
Solution 2: Use Vieta. We know the sum and the product of two numbers. This reminds us of Vieta's
Formulas. Suppose the numbers are the roots of the quadratic x2 + bx + c = 0. Then, -b is the sum of the
numbers and c is their product. Therefore, we must have b = -6 and c = -19, so the numbers are the
roots of the quadratic x2 - 6x - 19 = 0. Applying the quadratic formula gives us the same answers as in
our first solution. □
Our second solution shows us a useful way to apply Vieta's Formulas:

Concept: When we have information about the sum and the product of two numbers,
O≡^ we can let the numbers be the roots of a quadratic and use Vieta to construct
a quadratic whose roots are the numbers.

Problem 4.18: Suppose ∕(x) = x + √x and g(x) = x + 1/4. Evaluate g(∕(g(∕(g(∕(7)))))). ∣

Solution for Problem 4.18: We might try just plugging 7 into ∕(x) and grinding out the answer. Our first
step gives us ∕(7) = 7 + √7. Next, we have g(∕(7)) = 7+ V7+∣ = ψ + √7. Sticking that into ∕(x) doesn't
look like fun. Let's take a step back and look for another approach.
We're repeatedly sticking the output of f into g, so let's look at g(f(x)f

g(f(×)) = x + √v +

94
 4.3. COMPLETING THE SQUARE

If we don't see how to simplify this right away, we <an try a few values of r to ^<τ if anything interesting
happens. We choose perf<x t squares to avoid radicals in our result
.
*

jrt∕oυ = ∣,

g(fW) = 7.
49
g(∕(9)) = -.

∖∖hv do we keep getting perfect squares divided by 4 as our results? This makes us wonder if y( f(χ)) ∣⅛
itself the square of an expression.

Important: Not only must you know how to create a perfect square of a binomial,
you must know how to recognize one when you see one.

2
We have (√x) ≡ x, (1∕2)2 = 1/4, and 2(√x)(l∕2) = √r, so

g(∕(r)) = x+ √x+^ = (√x+^) .

Aha! Suppose we let ∕∕(x) = g(∕(x)). Our problem is to evaluate

g(∕(gσW(7)))))) = g(∕W(Λ(7)))) = g(fWh(7))) = ∕z(∕1(h(7))).

We have ∕∕(7) = (√7 + ∙ We don't expand the square, though, because our next step is to evaluate

W(7))≡
⅛(h(7)) = (√fc<7j + ∙

Using our expression for ∕∕(7) from above, we have y∕h(7) = √7 + ⅛, so

ft(Λ(7)) = (√S(η + l) =(√7 + l)2.

Therefore, we have h(h(Λ(7))) = (∖∕h(lι{7)) + - j = (√7 + ~ j = ^-+3√7.□

I Mt∏iMk⅜

4.3.1 Complete the square for each of the following quadratics:

(a) x2 +8x (c) 2z2 + 20z + 3

(b) y2 - 6y + 1 (d) 9χ2 + ∣χ +

95
CHAPTER 4. QUADRATICS

4.3.2 Find all real solutions to each of the following quadratic equations:
(a) x2 + 8x = 12 (c) 2z2 + 20z + 3 =-30
(b) y2 - 6y + 1 = -4 (d) 9x2 + ⅛x + f = 2

4.3.3 What positive number is 8 more than its reciprocal?

4.3.4 Let g(x) = -7ax1 + 3α2x + 13 for some constant a. If g(3) = -59, find all possible values of a.

4.3.5 Find the roots of each of the following quadratic equations:

(a) 2x2 + 17x = -21 (c) 22 + 4z + 5 = 0
(b) y2 - 16y + 51 = 0 *
(d) √5x2 - (8√5 + 10)x + (20√5 + 40) = 0

4.3.6 Let g(f) = 2f2 - 8λ∕5f + 25. Find all values of f such that g(t) = -13.

4.3.7 Find all z such that z2 - (3 - 8z)z - (14 + 12z) = 0.

4.4 The Discriminant

Problems

Problem 4.19: Withoutfinding the roots, determine whether the roots of each of the following quadratics
are nonreal, irrational, or rational. Also, determine the number of distinct roots each quadratic has.
(a) x2-4x + 20 = 0 (c) r2 - r - 462 = 0
(b) 2x2 - 3x - 7 = 0 (d) —^2 + 4t — 8 = 0

The quadratic formula gives us a quick way to determine the nature of the roots of a quadratic with
real coefficients without even determining the roots of the quadratic. To see how, we note that the roots of
ax2 + bx + c are -----------
-b W2 - 4zzc
x = — ±---- z------ ■
2zz 2zz
So, if a, b, and c are real, these roots are nonreal if and only if b2 - 4ac < 0. Moreover, the ± tells us
that these roots form a conjugate pair, since -b∕2a is the real part of each root and the ± tells us that the
imaginary parts of the roots are opposites.
If the coefficients are real and b2 - 4zzc = 0, then both roots equal -~. We say that such a quadratic
has a double root at x = and it has only one distinct root. Finally, if b2 - 4ac > 0, then the quadratic
has two different real roots.
If the coefficients are rational, we can go a step further. If b2 - 4ac is the square of a rational number,
then both roots are rational, since each equals the sum or difference of two rational numbers divided by
another rational number. Otherwise, the roots are both irrational.

96
 4 4 THE DISCRIMINANT

Important: The discriminant of lhr quadratic tn ∙ ∕π ∙cisfr2 ¼- When .t hr arid <'

∖ Tz aιv re al we can use Hie disc πminanl to determine the nature of the roots
ol the quadratic

• If b2 - 4λc > 0, the roots of the quadratic are real.

• If b2 - 4ac < 0, the roots of the quadratic are a nonreal conjugate

pair.

I urthermore, if b2 - 4ac = 0, then the quadratic has a double root.

If the coefficients of the quadratic are rational, then the roots of the
quadratic arc rational if and only if b2 - 4ac is the square of a rational
number.

Problem 4.19: Without finding the roots, determine whether the roots (if each of the following
quadratics arc nonreal, irrational, or rational. Also, determine the number of distinct roots the ;
quadratic has.
(a) x2 - 4x + 20 = 0 (c) τ2 - r - 462 = 0
(b) 2x2-3x-7 = 0 (d) -£+41-8 = 0 i

Solution for Problem 4.19:

(a) The discriminant is (-4)2 - 4(l)(20) = -64, so the two distinct roots are not real
(b) The discriminant is (-3)2 -4(2)(-7) = 65, which is positive but not the square of a rational number.
So, the roots are two distinct irrational numbers.
(c) The discriminant is (-l)2 - 4(1 )(-462) = 1849 = 432, which is the square of a rational numbe r I he
coefficients of the quadratic are rational, so the roots are two different rational numbers
(d) The discriminant is (4)2 - 4 (-⅛) (-8) = 0, so this quadratic has one distinct root. Γhe coefficients
of the quadratic are rational, so the double root is rational.
□
Problems

Problem 4.20:
(a) Find the quadratic with real coefficients and quadratic term x2 that has 3 + 2i as a root.
(b) Find the quadratic with rational coefficients and quadratic term x2 that has -3 + 2 V7 as a root

Problem 4.21: In this problem, we find all pairs of real numbers (x,y) such that there is at least one
real number t for which ι∕ = 2∕.v - t2.
(a) Treat x and y as constants, and write the equation as a quadratic with t as the variable. Find the
discriminant of the quadratic in terms of x and y.
(b) Use the discriminant from part (a) to describe all pairs of real numbers (x, y) such that the
quadratic has real roots. For all of these pairs, is it true that the equation y 2t∖ t: has at least
one real solution /?

97
CHAPTER 4. QUADRATICS

Problem 4.20:
(a) Find the quadratic with real coefficients and quadratic term x2 that has 3 + 2z as a root.
(b) Find the quadratic with rational coefficients and quadratic term x2 that has —3 + 2 f7 as a root.

Solution for Problem 4.20:

(a) Since the roots are nonreal but the coefficients are real, the roots form a conjugate pair. Therefore,
the other root is 3 + 2z = 3 - 2i. To find the quadratic, we note that the sum of the roots is
3 + 2z + 3 - 2z = 6 and the product of the roots is (3 + 2z)(3 - 2z) = 13. Using Vieta's Formulas, we
now see that the quadratic is x2 - 6x + 13.
(b) The roots of ax2 + bx + c are
-b Vb2 - 4ac
x — — ±------------ ∙
2a 2a
If a, b, and c are rational, the only way a square root can be part of a root is from the second term.
The ± before that term tells us that if —3 + 2 V7 is one root of a quadratic with rational coefficients,
then —3 — 2 V7 is the other root. To build the quadratic, we again use Vieta. The sum of the roots
is -6 and their product is (-3 + 2 V7)(-3 - 2 V7) = -19, so the quadratic is x2 + 6x — 19.

Problem 4.21: Find all pairs of real numbers (x, y) such that there is at least one real number t for
which y = 2fx - t2.

Solution for Problem 4.21: We wish to know when we can find real values of t that satisfy the equation,
so we view the given equation as a quadratic equation in t,

t2 - 2xt + y = 0.

The coefficient of t is -2x and the "constant term" of the quadratic (i.e., the term without t) is y. If t is
real, the discriminant of the quadratic must be nonnegative, so we must have

(-2x)2 - 4(l)(y) ≥ 0.

Simplifying this inequality gives us x2 > y. Therefore, if the ordered pair (x, y) satisfies x2 ≥ y, then there
is at least one real solution t to the equation y = 2tx -12. □

The hardest part of Problem 4.21 was figuring out where to start. The subsequent algebraic manip­
ulation in the problem was pretty straightforward. What helped us get started was focusing on just one
of the three variables in the given equation. Our goal was to find the circumstances under which real
values of t exist, so we rewrote the equation as a quadratic in the variable t. This allowed us to use the
discriminant to answer the question.

If an equation with multiple variables is quadratic in one of the variables,

O≡≡s then viewing the equation as a quadratic in that variable allows us to use
all our quadratic equation tools on the problem.

98
 4.5 QUADRATIC INEQUALITIES

4.4.1 One of the roots of a quadratic with real coefficients is 6 ÷ VSf Find the other r<κ∣t
4.4.2 Find all a such that ΛΛx - 5x + 9 ≈ 0 has only one distinct root.
4.4.3 Is it possible for a quadratic with real coefficients to have a double root that is non real’ Why or
why not?

4.4.4 In part (c) of Problem 4.16, we found the roots of the quadratic to be 3 - i and 8 - i. Why is it
possible for the quadratic to have roots that are not a conjugate pair’
4.4.5 II is given that one root of 2r2 + rx + s = 0, with r and s real numbers, is 3 + 2i. Find s. (Source:
AHSME)

4.4.6 Find the real values of K for which the equation x = K2(x - l)(x - 2) has real roots.

4.5 Quadratic Inequalities

Problems

Problem 4.22:
(a) Find all x such that (x - 2)(x - 3) > 0.
(b) Find all ∕ such that 2∕2 + 9/ + 4 < 0.
(c) Find all r such that r2 - 4r + 2 > 0.

Problem 4.23: Let ∣>(x) = 2x2 + 8x + 27.

(a) Complete the square to write p(x) in the form a{x - h)2 + k, where a, h, and A are constants
(b) Use part (a) to determine all values of x for which p(x) > 0.

Problem 4.22:
(a) Find all x such that (x - 2)(x - 3) ≥ 0.
(b) Find all t such that 2t ' + 9f ⅛ 4 < 0.
(c) Find all r such that r2 - 4r ÷ 2 > t).
CHAPTER 4. QUADRATICS

Solution for Problem 4.22:

(a) Both x — 2 and x - 3 are nonnegative for x ≥ 3, so (x - 2)(x — 3) ≥ 0 when x ≥ 3. Similarly, both
x — 2 and x — 3 are nonpositive for x ≤ 2, so (x - 2)(x — 3) ≥ 0 when x ≤ 2. Therefore, all x such
that x ≤ 2 or x ≥ 3 satisfy the inequality. In interval notation, our solution is x ∈ (—∞, 2] U [3, +∞).
Recall from page 28 that the "U" in this statement means "or," so, "x ∈ (-∞, 2] U [3, +∞)" means

x is in the interval (-∞,2] or the interval [3, +oo).

Note that we use "]" after the 2 and "[" before the 3 to indicate that 2 and 3 are valid solutions.
(b) Factoring the quadratic gives us (2t + l)(f+4) < 0. f + 4 2t + l (t + 4)(2t + 1)
The product (2t + l)(f + 4) is only negative if 2t +1 + + +
t>
and t + 4 have opposite signs. We have 2f +1 > 0
for t > -1/2 and 2t +1 < 0 for t < -1/2. Similarly, -4 < f < -∣ ÷ — —
t + 4 > Ofort > -4andf+4 < Oforf < -4. We can f <-4 - — ÷
capture this information in a table, as shown on
the right. Our table quickly shows us that (2f + l)(f + 4) < 0 for -4 < t < -1/2. Notice that neither
t = —1/2 nor t = —4 satisfy the inequality because the inequality is strict. In interval notation, our
solution is t ∈ (-4, -∣).

(c) We can't easily factor this quadratic. However, we can use the quadratic formula to find that the
roots of r2 — 4r + 2 are 2 ± V2. So, we can write the inequality as

(r - (2 + √2)) (r - (2 - √2)) > 0.

Both of the factors on the left are positive if r > 2 + y/2 and they are both negative if r < 2 - V2, so
the inequality is satisfied if r ∈ (-∞,2 - √2) U (2 + -√2, +∞). Notice that we use ")" and "(" in the
interval notation because 2 - V∑ and 2 + √2 do not satisfy the strict inequality.
We also could have solved the problem by completing the square. We must add (-4∕2)2 = 4
to r2 - 4r to complete the square. Adding 4 to both sides of our original inequality gives us
r2 _ 4r + 4 + 2 > 4, so we have (r - 2)2 > 2. From this inequality, we see that we must have either
r-2>√2orr-2< - √2, which leads to the same solution as before.
Concept: Solving the same problem in two different ways is a great way to check
o≡=g your answer.

□
Problem 4.23: Let p(x) = 2x2 + 8x + 27. Find all real values of x such that p(x) > 0.

Solution for Problem 4.23: We start by trying to find the roots of p(x). Factoring doesn't work, and when
we find the roots of p(x), we see that the discriminant is 82 - 4(2)(27) = 8(8) - 8(27) = 8(-19), which is
negative. So, the roots of p(x) are not real. Now what?
Completing the square worked when the roots were irrational. Let's try it on this quadratic, too.

Concept: Completing the square isn't just for finding roots of quadratics.

100
 4 5 QUADRATIC INEQUALITIES

We have pix) ≡ 2(x, 4 4ι) 4 27 Completing the square gives us

p(x) = 2(x2 4 4x) 4 2(4) - 2(4) 4 27 ≡ 2(x 4 2)2 4 19

If X is real, then 2(x 4 2)2 is nonnegative. Therefore, p(x) equals 19 plus some nonnegative number. In
other words, p(ι) is positive for all real x. □

As an Exercise, you'll extend Problem 4.23 to show that:

Important: Suppose fix) is a quadratic with nonreal roots If the coefficient of v2 1-

positive, then fix) > 0 for all x. If the coefficient of v2 is negative, then
∕(x)<0forall x.

Problem 4.24: Let ∕(x) =----- - —----- z where the domain of f is all real numbers Find the range
of f. * + 1

Solution for Problem 4.24: If k is in the range of f, then there must be some real value of v for which
∕(x) = k. So, the equation
5x2 - 4x 4 8 _ j
x2 4 1 ~k'
must have a real solution for x. Multiplying this equation by x2 4 1, and then rearranging, gives us

(k - 5)x2 4 4x 4 (k - 8) = 0.

There is at least one real solution x when the discriminant is nonnegative. Therefore, we mu a have
42 - 4(A' - 5)(A' - 8) > 0. Dividing both sides by 4 gives 4-(k- 5)(A - 8) ≥ 0. Expanding and rearranging
gives -∕r 4 12A - 36 ≥ 0, so multiplying by -1 (and reversing the inequality sign) gives k2 - 12k 4 36 ≤ 0
Factoring the quadratic gives (A-4)(A-9) ≤ 0, and solving this inequality gives us k € [4,9∣. So, only for
values of k in this interval is there a value of x such that ∕(x) = k. This means [4,9] is the range of f. □

I ven ive*

4.5.1 Find all values of t such that t2 - 13f ≤ -40.

4.5.2 Find all real values of x such that -x2 4 7x - 13 > 0.

4.5.3 Let ∕(x) = where the domain of f is all real numbers. Find the range of f.
4.5.4

(a) Let f(x) be a quadratic with nonreal roots. Show that if the coefficient of x2 in ∕(x) is positi∖ e, then
fix) > 0 for all x.
(b) Let f(x) be a quadratic with nonreal roots. Show that if the coefficient of λ in ∕(x) is negative,
then fix) < 0 for all x.
(c) What if a quadratic fix) has a real double root? What then can we say about the possible values
of /(x)?
CHAPTER 4. QUADRATICS

4.5.5 If x is real and 4y2 + 4xy + x + 6 = 0, then find the complete set of values of x for which y is real.
(Source: AHSME) Hints: 269
4.5.6 Find all real values of x such that x3 - 6x2 + 9x > 0.

4.6 Summary

An expression of the form ax2 + bx + c, where a, b, and c are constants and a ≠ 0, is called a quadratic.
The solutions to the equation ax2 + bx + c = 0 are called the roots or zeros of the quadratic.
There are two common methods for finding the roots of quadratics. First, we can sometimes easily
factor the quadratic by writing it as the product of two linear expressions. Setting each of the linear
expressions equal to zero gives us the roots. Second, the quadratic formula tells us that the roots of
ax2 + bx + c = 0 are
-b ± Vb2 - 4αc

We prove the quadratic formula by completing the square, which means writing a quadratic ex­
pression ax2 + bx + c in the form a(x + p')2 + q for some constants a, p, and q.

Important: The discriminant of the quadratic ax2 + bx + c is b2 - 4ac. When a, b, and c

ψ are real, we can use the discriminant to determine the nature of the roots
of the quadratic:

• If b2 - 4ac > 0, the roots of the quadratic are real.

• If b2 - 4ac < 0, the roots of the quadratic are a nonreal conjugate

pair.

Furthermore, if b2 - 4flc = 0, then the quadratic has a double root.

If the coefficients of the quadratic are rational, then the roots of the
quadratic are rational if and only if b2 - 4ac is the square of a rational
number.

Vieta's Formulas for a quadratic tell us that if r and s are the roots of the

ψ quadratic flx2 + bx + c, then

, . c
r+s = — and rs = -.
fl a

We can solve many quadratic inequalities by factoring the quadratic, then determining for what
values of the variable each factor is positive or negative. If ∕(x) is a quadratic that does not have real
roots, then either ∕(x) > 0 for all x or ∕(x) < 0 for all x.

102
 REVIEW PROBLEMS

Problem Solving Strategies

Concept*: • Sonic equations can be turned into quadratic equations with a uh
stitution.

• Using a little number theory can greatly help factor quadratics

• When facing a problem you don't know how to do, tn relating it to

a problem you know how to do.

• We call a two-variable expression symmetric if if remains the same

when the variables are swapped. Symmetric expressions involving
the roots of a quadratic are a clue to try using Vieta s Formulas.

• Using sample cases to test a formula you derive is a good wav to

check your formula.

• When we have information about the sum and the product of two
numbers, we can let the numbers be the roots of a quadratic and use
Vieta to construct a quadratic whose roots are the numbers

• If an equation with multiple variables is quadratic in one of the

variables, then viewing the equation as a quadratic in that variable
allows us to use all our quadratic equation tools on the problem

• Solving the same problem in two different ways is a gre.∣t way to

check your answer.

Review Problems

4.25 Find all the solutions of the following equations:

(a) x2 + 12x + 27 = 0 (b) x2-6x + l=0 (c) 2y2-5y + 2 = 0 (d) 3x2 + 15x = 7

4.26 Let ∕(x) = <ι2x2 + sflx + 3 and ∕(2) = 2. Find all possible values of the constant a.

4.27 Find the value of x if x is positive and x - 1 is the reciprocal of x + ∣.

4.28 The sum of the squares of the roots of the equation x2 + 2∕rτ = 3 is 10. Find ∣∕ι∣. (Source: AHSME)
4.29 Solve each of the following equations:
(a) √6β2 + 5a + 21 =14 (b) √4x2 + 20x + 25 = 2v + 5

4.30 Let f be a function for which ∕(x∕3) .v + v4 I f ind the sum ot all values of : for which/(3.2) = 7.
(Source: AMC 12)

103
CHAPTER 4. QUADRATICS

x+7
4.31 What is the domain of the function ∕(x) =

4.32 Let a and b be the roots of the equation x2 — mx + 2 = 0. Suppose that a + ∣ and b + J are the roots
of the equation x2 - px + q = 0. What is q? (Source: AMC10)
4.33 One of the roots of x2 - ax + 2a + 3 = 0 is 3 times the other. Find all possible values of a.

4.34 Find all x such that 6x2 + 5x < 4. (Source: AHSME)

4.35 Find all values of k such that x2 + kx + 27 — 0 has two distinct real solutions for x.
4.36 Find all possible values of x/y if x and y are nonzero real numbers such that x2 = xy + 12y2.

4.37 Find x if y = and x + y = 2. (Source: AHSME)

4.38 The Bolts charge $50 for each ticket, and they sell 20,000 tickets per game. For each dollar they
raise their prices, they will sell 100 fewer tickets. For what values of k can the Bolts increase ticket prices
by $/c and have more than $1,000,000 in revenue per game?
4.39 Find all x such that x2 + 50x - 2079 ≥ 0.
4.40 Let P(x) be a quadratic polynomial such that P(0) = -1, P(l) = 9, and P(2) = 25. Find P(-1).
4.41 Find all real solutions to (x2 - 5x + 5)∙r2-9x+20 = 1.

4.42 Find all solutions to the system of equations,

x2 + yz = 39,
x-yz = -33,
y + z = 13.

4.43 Find the roots of x2 + = 0 in terms of a.

[^Challenge Problems

4.44 jf P∑+ι = ≡zl pιas solutions for x such that each solution is the negative of the other, then find m
in terms of a and b. (Source: AHSME)
4.45 Find constants a and b such that b - a is as small as possible, and the entire graph of the equation
y = (1 - x2)∕(l + x2) lies within a < y ≤ b.
4.46 Let t be a real number such that ∣f∣ ≤ 2. Show that both roots of the quadratic z2 + tz + 1 = 0 have
magnitude 1.

4.47 Prove that if = —then a and b can't both be real numbers. Hints: 191
a a+b
1 p2 -1
4.48 Find the positive difference between the roots of x2 - px H----- - = 0. Hints: 133

104
 CHALLENGE PROBLEMS

4.49 Find all ordered pan

*- (.r, v) that satisfy both ι vv « 6 and »y * ι 4
4.50 A quadratic expression is monk if the < oeff∣c∣ent of its quadratic term is 1 Let ∕(.r). ^(r). and ∕ι(x)
be monic quadratics such that the sum of the roots of f(r) is a, the sum of the roots of <(.r) is b, and the
sum of the roots of ∕tfι) is ι Find the sum of the solutions to ∕(x) + χ(x) + h(x) a 0 in terms of a, b, and c

4.51 Let a and b be nonzero real numbers such that rt2 + b2 + 2λ + 1 = 2ab + 2b. Find the value of
Hints: 34
4.52 Let a be a constant such that the roots of ax2 - 6x + 12 = 0 are r and s. Find the value of -ζ ÷ 4 in
terms of a.

4.53 Find the sum of the real roots of (2 + (2 + (2 + (2 ÷ x)2)2)2)2 = 15129.

4.54 Prove that if a quadratic has real coefficients such that the sum of the coefficients is 0. then the
quadratic has real roots. Hints: 208

4.55 Find all ordered pairs (rfs) that satisfy both s2 - 1 = 4r2 + 4r and 4r + s2 I + 3rs.
4.56 Prove that if (r + l)(s + 1) = 13, then there is some constant c such that r and s are roots of the
quadratic equation x2 - ex + (12 - c) = 0. Hints: 303

4.57 Find all real solutions (x, y) of the system x2 + y = 12 = y2 + x. (Source: IiMMT) Hints: 160

4.58 Let a and b be real numbers such that 0 < a < b and a2 + b2 ≈ 6ab. Find * . Hints: 2M
a-b
9
4.59 Find all ordered pairs (x,y) that satisfy x + y = - and 2x2y2 - 13xy + 18 0. (Source: C l'M( )
Hints: 292

4.60 Find all the values of m for which the zeros of 2x2 - mx - 8 differ by m-↑. (Source I IMMT)
„ j ,ι I, x ~ a x -b b a
4.61
* Find all x such that —-— +------ =------- +------ -, where a and b are constants.
b a x-a x-b
4.62
* Solve the following system of equations in real numbers:

2xι = x| - 23,
4λ^7 = x^ + 7,
6x3 = x^ + 14,
8x4 = x^ + 23,
lθx5 = x2 + 34.

(Source: Austria) Hints: 177

CHAPTER 5. CONICS

Along a parabola life like a rocket flies,

Mainly in darkness, now and then on a rainbow. - Andrei Voznesensky

CHAPTER
I
I Conics

A locus is a collection of points that satisfy specific criteria. For example, the locus of points on the
Cartesian plane that satisfy the equation x — y is a line, and the locus of points in a plane that are 2 units
from a given point P is a circle with radius 2. In this chapter, we explore some special loci ("loci" is the
plural of "locus"), and we discuss equations whose graphs on the Cartesian plane produce these loci.
The loci we study in this chapter are collectively called conic sections, or conics for short. The
diagrams below illustrate why. In each diagram, a plane intersects a pair of congruent cones that share
a vertex and axis, but open in opposite directions.

If the plane is parallel to a line on the surface of the cones such that the line is not in the plane, then
the intersection of the plane and the cones forms a shape called a parabola. If the plane is perpendicular
to the axis, then we get a circle. If it intersects one cone in a closed curve, we have an ellipse. Finally,
we produce a hyperbola when the plane intersects both cones as in the final diagram.

Sidenole: Wh<∣t happens when the plane passes through the vertex of the cones?

106
 5.1. PARABOLAS

In this chapter, we'll learn definitions of the conic sections, and the important points, lines, and
segments associated with each type of conic section. We'll use these definitions to create general
formulas that can be used to solve specific problems, and we'll also learn how to use our definitions to
tackle some problems without resorting to formulas.
The many formulas we will prove are not very important, so you shouldn't feel the need to memorize
them all. You can usually just look them up when you need them.
As stated in the introduction to this text on page iii, the details of this chapter are not as important as
those presented in most of the other chapters of this book. The two main items you should learn from
this chapter are the importance of using completing the square to work with quadratic expressions, and
a qualitative understanding of equations whose graphs are conic sections.

5.1 Parabolas

Geometrically speaking, a parabola is the locus of points in a plane that are equidistant from a given line
and a point not on the line. The line is called the directrix of the parabola and the point is the parabola's
focus.

The solid curve at right is a parabola. The dashed horizontal line £ below
the parabola is the parabola's directrix and point F "inside" the parabola
is its focus. The line through the focus perpendicular to the directrix is the
parabola's axis of symmetry. As the name indicates (and we'll prove in the
first problem below), a parabola is symmetric about its axis of symmetry.
The point where the axis intersects the parabola is the vertex of the parabola.
I
As shown at right, the vertex of a parabola with a horizontal directrix is the ▼

bottommost point of the parabola.

You may already know that the graph of a quadratic produces a parabola.
In this section, we learn how our geometric definition of a parabola produces a quadratic equation.

Problems

Problem 5.1: In the diagram, F is the focus of the parabola, P is on the

parabola, £ is the directrix, and Q is the foot of the altitude from P to £.
Line k passes through the focus and is perpendicular to the directrix. In
this problem, we prove that the parabola is symmetric about line k.
Let P' be the image of P upon reflection over the line k. Our goal is to
show that P' is also on the parabola.
(a) Explain why PQ = PF.
(b) Explain why P'F - PF.
(c) Explain why the distance from P' to £ equals PQ.
(d) Explain why P' is on the parabola. Why does this mean that the parabola is symmetric about
line k?

107
CHAPTER 5. CONICS

Problem 5.2: Show that the vertex of a parabola is the midpoint of the segment that connects the focus
of the parabola to the point where the parabola's axis of symmetry intersects the parabola's directrix.

Problem 5.3: Suppose a parabola on the Cartesian plane has focus (0, f) and vertex (0,0), where t > 0.
(a) First, we show that the equation whose graph is this parabola can be written in the form y = ax2,
where a = l∕(4f).

(i) Find the equation whose graph is the directrix of the parabola.
(ii) Suppose that (x, y) is a point on the parabola. In terms of x, y, and t, how far is the point
(x, y) from the focus? How far is the point from the directrix?
(iii) Because (x, y) is on the parabola, what must be true about the two distances you found in
part (ii)?
(iv) Find an equation that must be satisfied by x and y. Show that your equation can be written
in the form y = ax2, where a = l∕(4f). Does the same equation result if t < 0?
(b) Let a be a nonzero constant. Find the focus and directrix of the graph of y — ax2 in terms of a.
(c) Graph the equation y — ax2 for several values of a. How does the value of a affect the graph?

Problem 5.4:
(a) Find an equation whose graph is a parabola with vertex (0,0) and focus (0,2).
(b) Suppose we wish to find an equation whose graph is a parabola with vertex (-6,3) and focus
(-6,5). How can we use our answer to part (a) to quickly find the desired equation?
(c) Use the strategy from part (b) and your final equation from Problem 5.3 to find an equation
whose graph is a parabola with vertex (h, k) and focus (h,k+ t).
(d) Find the vertex, focus, axis of symmetry, and directrix of the graph of y = α(x - h)2 + k, where a,
h, and k are constants, and a ≠ 0.

Problem 5.5: Graph the equations y = (x + 2)2 + 3 and x = (y + 2)2 + 3.

Problem 5.6:
(a) A parabola has vertex (-2,5) and directrix y = 5. Find an equation whose graph is this parabola,
and graph the parabola.
(b) Find an equation whose graph is a parabola with vertex (4, -1) and focus (0, -1), and graph the
parabola.

Problem 5.7:
(a) Find the vertex and axis of symmetry of the graph of y2 - 8y + x = 2.
(b) Find the vertex and axis of symmetry of the graph of 2x2 + 14x — y = 1.

Problem 5.1: Prove that a parabola is symmetric about the line through its focus that is perpendicular
to its directrix. In other words, show why this line earns the name "axis of symmetry."

108
 5.1. PARABOLAS

Solution for Problem 5.1: We let C be the directrix, F be the focus, P be a point
on the parabola, and k be the line through F perpendicular to C. To show
that k is a line of symmetry of the parabola, we consider the reflection of P
over k, which we call point P'. We wish to show that P' is on the parabola.
To show that P' is on the parabola, we must show that it is just as
far from f as it is from F. Since P is on the parabola, we already know
that PF = PQ, where Q is the foot of the altitude from P to F (so PQ is the
distance from P to F). Because P' is the image of P upon reflection over k, and F is its own image upon
this reflection, we have P'F = PF. Meanwhile, because PQ is parallel to k, the image of PQ, which is
P'Q', is parallel to both k and PQ. Since PQ ± F and PQ ∣∣ P'Q', we know that P'Q' ± F as well. This
means that P'Q' is the distance from P' to F. Since P'Q' is the image of PQ upon reflection over k, we
have P'Q' = PQ. Finally, because PF = PQ, PF = P'F, and PQ = P'Q', we have P'F = P'Q'. So, P' is
equidistant from F and from F, which means that P' is on the parabola. Therefore, the image of each
point on the parabola upon reflection over k is also on the parabola, which means that k is a line of
symmetry of the parabola.

This is why the line through the focus of a parabola that is perpendicular to the directrix is called
the axis of symmetry. □

Problem 5.2: Show that the vertex of a parabola is the midpoint of the segment that connects the ∣
focus of the parabola to the point where the parabola's axis of symmetry intersects the parabola's ⅛
directrix.

Solution for Problem 5.2: Let V be the point where the axis intersects the
parabola, and let Z be the point where the axis meet the directrix. Because
the axis is perpendicular to the directrix, VZ is perpendicular to the directrix.
Therefore, VZ is the distance between the vertex and the directrix. Since V
is on the parabola, it is equidistant from the focus and the directrix. Letting
the focus be F, we therefore have VF = VZ. Because F, V, and Z are different
points on the same line such that VF = VZ, point V is the midpoint of FZ,
as desired. □
Problem 5.3: Suppose a parabola on the Cartesian plane has focus (0, t) and vertex (0,0), where t ≠ 0.
(a) Let (x, y) be a point on the parabola. Find an equation that must be satisfied by x and y. Show
that your equation can be written in the form y = ax1, where a = l∕(4f).
(b) Let a be a nonzero constant. Find the vertex, focus, and directrix of the graph of ι∕ = axl in
terms of a.
(c) Graph the equation y - ax2 for several values of a. How does the value of a affect the graph?

Solution for Problem 5.3:

(a) We'll first tackle the case in which t > 0. The axis of symmetry of a parabola passes through the
vertex and the focus. Since the vertex and focus are both on the y-axis, the axis of symmetry of
this parabola is vertical. The axis of symmetry and directrix of a parabola are perpendicular, so
the directrix, which we'll call F, is horizontal. Moreover, the vertex of a parabola is equidistant

109
CHAPTER 5. CONICS

from the focus and the directrix. So, because the vertex is t below the focus, the directrix is t below
the vertex. This means that the directrix is the graph of y -
Let point P be on the parabola, where P = (x, y)∙ Let the
focus be point F and the directrix be line C. By the definition
of a parabola, the length of PF equals the distance from P
to C. We let Q be the foot of the altitude from P to Λ so PQ
is the distance from P to F.
The distance formula gives us PF — √(x - 0)2 + (y - 02 ∙
P and F cannot be on opposite sides of Λ since this would
force PF to be greater than the distance from P to Λ Therefore, P is above <f, as shown, so y > t
and the distance PQ is y - (-f) = y + f. This distance must equal PF, so we have

λ∕(x-0)2 + (y-f)2 = y + f∙

We square both sides to get rid of the radical, and we have x2 + (y - t)2 = (y + 02∙ Subtracting
(y - t)2 from both sides gives

x2 = (y +1)2 - (y - 02 = 4fy∙

Solving this equation for y gives y = ^x2.

If f < 0, then the focus is below the directrix, as shown

at right. The distance from P = (x, y) to £ is -t - y, but PF
equals the same expression as before, so we have

λ∕(x-o)2 + (y-t)2 = -f-y∙

Because (-f - y)2 = (y + 02z squaring this equation leads

to the same equation as the one we found for t > 0.
We could have knocked off both cases at once by noting that in either case, the distance from P
to F is ∣y - (-f)l∙ This expression equals ∣y + t∣, and it must equal PF, so we have

√(x - 0)2 + (y - t)2 = ly + f∣∙

Squaring both sides gives (x - 0)2 + (y - f)2 = (ly +11)2∙ We have (ly +1∣)2 = (y + f )2, because (∣z∣)2 = z2
for all z, so (x - 0)2 + (y - t)2 = (y + 02z as before.

We can reverse our steps in Problem 5.3 to show that the graph of any equation of the form
v = l√z where t is a nonzero constant, is a parabola with vertex (0,0), focus (0, f), and directrix
<7 4f
y = -l
We just found that a parabola with vertex (0,0), focus (0,f), and directrix y =-t is the graph of
(b)
the equation y = ⅛x2. If we let a = ⅛, then this equation is y = αx2. Solving a = ⅛ for t in terms
of a gives t = ⅛. The vertex is still just (0,0), but using this expression for f allows us to write the
focus as (0, ⅛) and the directrix as y = -⅛.

110
 5.1. PARABOLAS

(c) Below are graphs of y = ax2 for a = 5, a = l∕5z and a = -1/5.

These three graphs give us a good idea how the coefficient of x2 affects the graph of the quadratic.
The most striking difference is that the graph of ι∕ = —x2∕5 opens downward, while our other
two graphs open upward. This suggests that whenever the coefficient of x2 is negative, the graph
opens downward, and when this coefficient is positive, the graph opens upward.
We can see why this is true by considering very large values of x. When x gets very large and
y = x2∕5, then y becomes a large positive number. When x gets very large and y = -x2∕5, then y
becomes much less than 0.
The coefficient of x2 also affects the width of the parabola.
The graph of y = 5x2 is much narrower than that of y = x2,
which is narrower than that of y = x2∕5. To see this effect
more clearly, consider the graphs of y = 5x2, y = x2, and
y = x2∕5 all on the same Cartesian plane at right.
Another feature we note from these graphs is that we have
an algebraic explanation for why a parabola is symmetric
about its axis. Focusing on the graph of y = x2, we see that
for every point (x, x2) on the graph with x ≠ 0, there is a point
(-x,x2) also on the graph. The point (-x,x2) is the reflection
of the point (x,x2) over the y-axis, so the parabola is symmetric about the y-axis, which is its axis
of symmetry.

□
Problem 5.4:
(a) Find an equation whose graph is a parabola with vertex (0,0) and focus (0,2).
(b) Suppose we wish to find an equation whose graph is a parabola with vertex (-6,3) and focus
(-6,5). How can we use our answer to part (a) to quickly find the desired equation?
(c) Use the strategy from part (b) and your final equation from Problem 5.3 to find an equation
whose graph is a parabola with vertex (h,k) and focus (h,k + ∕).
(d) Find the vertex, focus, axis of symmetry, and directrix of the graph of y = a(x - ∕ι)2 + k, where
a, hr and k are constants.

111
CHAPTER 5. CONICS

Solution for Problem 5.4:

(a) A parabola with vertex (0,0) and focus (0, is the graph of the equation y = ax2. Therefore, to
find an equation whose graph is a parabola with vertex (0,0) and focus (0, 2), we must find the
value of a such that = 2. Solving this equation gives us fl = ∣, so the parabola is the graph of

(b) Let fP be the parabola with vertex (-6,3) and focus (-6,5). The vertex of eP is 2 units below the
focus, just as in part (a). But the vertex of *P is not the origin, so it appears at first that we cannot
use what we know about the graphs of equations of the form y = flx2∙
However, if we shift the parabola with vertex (0,0) and focus (0,2) left 6 units and up 3 units,
we get a parabola with vertex (-6,3) and focus (-6,5). Mathematically speaking, this "shift" is
a translation. But is the result of this translation the only parabola with vertex (-6,3) and focus
(-6,5)?
To answer this question, we note that once we know the vertex and the focus of a parabola, we
can find the directrix of the parabola. Moreover, there is only one possible parabola with a given
focus and directrix. Combining these observations tells us that there is only one possible parabola
with a given focus and vertex. Therefore, there is only one parabola with vertex (-6,3) and focus
(-6,5), so the result of our translation is the parabola eP we seek.
Now, we,re ready to find the equation that has P as its graph. eP is the result of translating the
parabola in part (a) left 6 units and up 3 units. Therefore, for every point (x, y) on the graph of
y = gx2, there is a point (x', y') on <P where (x', y') = (x - 6, y + 3). For example, the vertex of P is
(0-6,0 + 3) = (-6,3).
Parabola P is graphed as a solid curve at right, and the graph
of y - ∣x2 is dashed. Notice that every point on P is 6 units to
the left and 3 units above a corresponding point on the dashed
parabola.
We can now use the relationship (x', y') = (x - 6, y + 3) to find —
an equation whose graph is P. For any point (x', y') on P, we
must have x' = x - 6 and y' = y + 3, where (x, y) is a point on the graph of y = 1√2 ∣x2. We solve
x' = χ - 6 and y' = y + 3 for x and y, respectively, to find x = x' + 6 and y = y' - 3. Since (x, y) must
satisfy y = ∣x2, we substitute our expressions for x and y to get an equation relating x' and y':

y'-3 = ∣(x' + 6)2.

Therefore, P consists of every point (x', y') that satisfies y' - 3 = ∣(x' + 6)2. In other words, P is
the graph of the equation y - 3 = ∣(x + 6)2.
We can test our equation by confirming that the vertex satisfies the equation. Indeed, when
x = -6 and y = 3, both sides of the equation equal 0. This check doesn't mean our equation is
definitely correct; there are infinitely many equations whose graphs pass through (—6,3). However,
quick checks like this help us catch errors.

You may have noticed that our using a translation to produce the equation y — 3 = ∣(x + 6)2
from the equation y = ∣x2 is a lot like our discussion of transforming functions in Section 2.2. It's

112
 5.1. PARABOLAS

not only "a lot like" transforming functions; it's exactly the same! If we let ∕(x) = ∣x2, then the
parabola in part (a) is the graph of f. In part (b), we must shift this graph to the left 6 units and
up 3. Our work in Section 2.2 tells us that the result is the graph of y = f(x + 6) + 3, which is
y = ∣(x + 6)2 + 3, or y - 3 = j(x + 6)2.

(c) Let lP be the parabola with vertex (∕z, k) and focus (∕z, k + f)∙ We follow the same steps as in part (b)
to find an equation whose graph is tP.
The graph of the equation y = x2 is a parabola with vertex (0z 0) and focus (0z t). So, the
vertex and focus of fP are h to right and k above the vertex and focusz respectively, of the graph of
V = ⅛χ2∙ Therefore, fP is the result of shifting the graph of y = ⅛x2 right h units and up k units.
In other words, for every point (xz y) on the graph of y = j,x2z there is a point (x', y') on tP where
(x'z y') = (x + ft, y + k). This means *P consists of all points (x', y') such that x' = x + h and y' = y + k.
Solving these equations for x and yz respectively, gives x = x' — h and y = y' - k. Because (xz y)
must satisfy the equation y = ^x2, we can substitute our expressions for x and y to find

y' -⅛ ⅛(∕ - ⅛)2∙

As before, we can get rid of the prime symbols and write that fP is the graph of the equation
y - = ⅛(x - ^z)2∙ We can also write this equation as y = j-t (x - ∕z)2 + k.
(d) We just learned that the graph of the equation y = ^(x - ∕r)2 + k is a parabola with vertex (h, ∕c) and
focus (∕z, k +1). If we let« = then the equation becomes y = α(x - ∕z)2 + k. The vertex is still (∕z, k),
and we can solve « = ⅛ for t to find that the focus is (∕z, k+ -⅛f).
Because the focus has the same x-coordinate as the vertex, the axis of symmetry is the vertical
line through the vertex, which is x = h. Since the axis of symmetry is vertical, the directrix is
horizontal. The y-coordinate of the vertex is k, the y-coordinate of the focus is fc+ and the vertex
is equidistant from the focus and directrix, so the directrix must be the line y = k - ⅛.

□
Important: The standard form of an equation whose graph is a parabola that opens
ψ vertically (upward or downward) is

y = α(x - ∕z)2 + k,

where a, h, and k are constants with a ≠ 0. If the α > 0, then the parabola
opens upward, and if a < 0, then the parabola opens downward. The
vertex of the parabola is (∕z, k) and the axis of symmetry is x = h.

The power of this standard form is that every equation whose graph is a parabola that opens upward or
downward can be written in this form. Once we have the equation in this form, we can easily determine
the vertex and axis of symmetry of the parabola as described above. We can also note that the focus of
such a parabola is (h, + ⅛) and the directrix is y = k -

Now, let's graph a couple particular parabolas.

Problem 5.5: Graph the equations y = (x + 2)2 + 3 and x = (y + 2)2 + 3.

113
CHAPTER 5. CONICS

Solution for Problem 5.5: From our work in Problem 5.4, we recognize y = (x + 2)2 + 3 y
I⅜ i 1 J
T
i
as an equation that is in the standard form of an equation whose graph is an upward- -j
il
__ ∖ ~T~
— J
∖
opening parabola. Since (x + 2)2 is nonnegative for all real x, the smallest possible fc 1 1∕ I
∖
value of (x + 2)2 + 3 occurs when (x + 2)2 = 0, which is when x = -2. When x = -2, ∖z
I
we have y = 3, so the bottommost point of the parabola, which is the vertex of the
J
parabola, is (-2,3). 1
I 1 1"

We find more points on the graph by choosing various values of x and find the corresponding values
of y. We thereby produce the graph at right above.

y At first glance, the equation x = (y + 2)2 + 3 resembles the standard form we

discussed in Problem 5.4. However, in this equation, the y term is squared, rather
than the x term. We can get a sense for the graph the same way we found the graph
of y = (x + 2)2 + 3. First, we note that (y + 2)2 is nonnegative for all real y, and is
zero for y = —2. Therefore, (y + 2)2 + 3 is minimized when y = —2, which makes
x = 3. For all other values of y, the value of x is greater than 3, which means that
(x, y) = (3, -2) is the leftmost point of the graph.
Choosing more values of y and computing the corresponding values of x, we produce the graph at
left above. As expected, the graph is a parabola with vertex (3, —2), but its axis of symmetry is horizontal
rather than vertical. □
Using similar methods to those we used in Problem 5.4, we can show that the graph of any equation
of the form x = α(y - k)2 + h is a horizontally-opening parabola:

Important: The standard form of an equation whose graph is a parabola that opens

φ horizontally (rightward or leftward) is

x = a(y - k)2 + h,

where a, h, and k are constants with a ≠ 0. If a > 0, then the parabola

opens rightward and if a < 0, it opens leftward. The vertex of the
parabola is (h, k) and the axis of symmetry is y = k.

We can also show that the graph of x = α(y -k)2 + h has focus (ft + k} and directrix y = ∕z -

Combining our observations about the graphs of equations of the form y = α(x - ∕z)2 + k and
x = ^(y - k)2 + h, where a, h, and k are constants with a ≠ 0, we have the following:

Important: Suppose we have an equation of the form Ax2 + By2 + Cx + Dy + E = 0,

ψ where A, B, C, D, E are constants such that one of A or B is zero and the

other is nonzero. The graph of the equation is a parabola. If A ≠ 0, then
the parabola opens upward or downward. If B ≠ 0, then the parabola
opens leftward or rightward.

An equation does not have to be in the form described above in order to have a graph that is a
parabola. Specifically, if the axis of symmetry of a parabola on the Cartesian plane is not horizontal or

114
 5.1. PARABOLAS

vertical, then the equation whose graph is this parabola does not fit either of the standard forms we
discuss in this section. We need more advanced mathematical tools to study these parabolas thoroughly,
so we'll focus on vertically-opening and horizontally-opening parabolas in this text.

Problem 5.6:
(a) A parabola has vertex (—2,5) and directrix y = ⅛. Find an equation whose graph is this parabola,
and graph the parabola.
(b) Find an equation whose graph is a parabola with vertex (4, -1) and focus (0, -1), and graph the
parabola.

Solution for Problem 5.6:

(a) Because the vertex is ∣ above the directrix, the focus is ∣ above the vertex, at (-2, y). Therefore,
any point (x, y) on the parabola is equidistant from (-2, y) (the focus) and y = ∣ (the directrix), so
we have
^τ + 2)2 + (y-y) = y -1∙

Squaring both sides and rearranging gives

u + 2)2 = (y-∣) -(y-⅞) =(y-∣ + y-y)(y-∣-(y-y)) = 6y-30.

Solving for y in terms of x gives y = ∣(x + 2)2 + 5.

We could also have used the standard form of the equation of a parabola to find the desired
equation. Specifically, we know that since the directrix is a horizontal line, the equation must be
of the form
y = a(x - h)2 + fc.
Furthermore, since the vertex is (-2,5) and the vertex must be (∕z, k), we have h = -2 and k = 5.
Now all we have to do is find a. For that, we use the directrix. The directrix must have the form
y = k - ⅛. Since we are given that the directrix is y = and we found above that k = 5, we must
have 5 - ⅛ = j∙ Therefore, a = ∣, so our equation is

y = j(x + 2)2 + 5.

To graph the parabola, we start with the vertex (-2,5). Since x is squared
and a is positive, the parabola opens upward. We then choose values of x
greater than -2 and find the corresponding values of y to graph the right side
of the parabola. The left side is the reflection of the right side over the axis of
symmetry x = —2, and we have the graph shown at right.

(b) The focus and vertex are on the same horizontal line, so the axis of symmetry is horizontal.
Therefore, the directrix is vertical. Since the focus is 4 to the left of the vertex, the directrix must
be 4 to the right of the vertex, which means the directrix is the graph of x = 8.

115
CHAPTER 5. CONICS

Because the parabola has focus (0, -1) and directrix x = 8, every point (x, y) on the parabola
must satisfy
√(x-0)2 + (y + l)2 = 8 - x.

We have 8 - x on the right rather than x - 8 because the fact that the focus is to the left of the
directrix tells us that every point on the parabola is to the left of the directrix. Therefore, if (x, y) is
on the parabola, we have x < 8, and the distance from this point to the directrix is 8 - x.
Squaring both sides of this equation gives us x2 + (y + 1)2 = x2 — 16x + 64, and solving for x
gives us x = -⅛(y + 1)2 + 4.

As in part (a), we could also have used our understanding of the standard form of
a parabola. The vertex (4, -1) is 4 units to the right of the focus (0, -1), so the parabola
opens leftward, and its equation has the form x = α(y - k)2 + h. From the vertex, we
have h = 4 and k = -1. From the x-coordinate of the focus we have h + ⅛ =0, from
which we find ⅛ = = -4, so a = - ⅛. Therefore, our equation is

x=-⅛(y+1)2+4∙

To graph this equation, we again start with the vertex (4, -1). We then choose values
of y (because we have x in terms of y in our equation above) and find corresponding
values of x to produce the graph shown at right.

In Problem 5.6, we showed two methods to get the equation for each parabola to highlight what's
important in this chapter. In each case, using our standard form looks easier. This shows the power of
understanding how to use standard forms of equations, and why we bother developing them. However,
our other approach, starting from the geometric definition of a parabola, shows that even when we don't
know or remember our formulas, all is not lost. We can start from the fundamentals and follow the
footsteps of the work we did when deriving the standard form.

Problem 5.7:
(a) Find the vertex and axis of symmetry of the graph of y2 - 8y + x = 2.
(b) Find the vertex and axis of symmetry of the graph of 2x2 + 14x - y = 1.

Solution for Problem 5.7:

(a) First, we note that y is squared, not x. So, the parabola opens left or right. If we write the equation
in the form x = α(y - k)2 + h, we can easily identify the vertex and axis.
We start by solving for x to find x = -y2 + 8y + 2. We then complete the square on the right side.
To do so, we first factor —1 out of the first two terms, to get x = —l(y2 — 8y) + 2. We then complete
the square by adding (—1)(16) - (-l)(16) (which equals 0) to the right side to give

x = —(y2 _ 8y) + (-l)(16) - (-l)(16) + 2 = -l(y2 - 8y + 16) + 18 = -l(y - 4)2 + 18.

What's wrong with this next step:

116
 5.1. PARABOLAS

Bogus Solution: We have a = -1, ∣ι = 4, and k = 18, so the vertex is (4,18).

We see our mistake immediately if we check to see if the point (4,18) satisfies the original equation.
The vertex is on the parabola, so this point should satisfy the equation. However, when we
substitute (x, y) = (4,18) in the left side of the equation y2 - 8y + x = 2, we have y2 - 8y + x =
182 — 8(18) + 4 = 184, so clearly something is wrong.
We overlooked which variable is squared. Our equation in standard form is x = -l(y - 4)2 +18.
Since y is squared, the parabola opens horizontally. We can read off the correct answer by
comparing x = -l(y - 4)2 + 18 to the standard form we found earlier, x = -l(y - k)2 + h. We then
have a = -1, h = 18, and k = 4, so the vertex is (18,4) and the axis is y = 4.
We can also simply think about the equation to determine the axis and vertex. We can see that
the parabola opens leftward by noting that as y gets much larger than 0 or much smaller than 0,
the corresponding value of x is much less than 0. Since x = -l(y - 4)2 + 18 and -l(y — 4)2 ≤ 0 for
all real y, the rightmost point on the parabola occurs when y = 4. This makes our squared term 0,
and gives x = 18, so our vertex is (18,4). The axis of symmetry is the horizontal line through the
vertex (because the parabola opens horizontally), so the axis is y = 4.
We again start by isolating the non-squared variable, to get y = 2x2 + 14x — 1, then we complete
the square to get

This fits our standard form, y = α(x — ∕ι)2 + k, of a parabola that opens vertically. What's wrong
with this next step:

Bogus Solution: We therefore have α = 2, ∕z = y and k = -y, so the vertex is (y -y).

We see our error by putting the supposed vertex into our equation in standard form. We quickly
see it doesn't work, because the expression 2(x + j)2 isn't 0 when x = (so y ≠ -y when x = j).
This time, we made a sign error. Notice that in our standard form, we have (x - ∕ι)2, not (x + h)2.
So, for our equation y = 2 (x + g) - y, we have h = -y not h = y We also have k = -y, so the
vertex is (-y -y) and the axis is x = -y

WARNING!! Be very wary of sign errors when working with parabolas (or other
V conic sections). One quick way to check for mistakes when working
with parabolas is to plug the point that you think is the vertex into
your equation. If the resulting equation isn't true, then you've made
a mistake somewhere.
We could have avoided the sign error by thinking about the equation rather than reaching
for our formulas right away. Since y = l(x + ∣) - y and 2(x + is always nonnegative, the
bottommost point on the graph will occur when 2 (x + ∣f = 0, which gives usx=-y,soy = -y.
Therefore, the vertex is (~y~y)∙ The axis of symmetry is the vertical line through the vertex,
which is x = -y□

117
CHAPTER 5. CONICS

*⅛
lixeιciM

5.1.1 Find an equation whose graph is a parabola with directrix y = -1 and vertex (1,3).

5.1.2
(a) What is the directrix of a parabola with focus (3,5) and vertex (2,5)?
(b) What is the directrix of a parabola with focus (3,7) and vertex (2,5)? Hints: 250
5.1.3 Find the vertex and axis of symmetry of the parabolas that result from graphing the following
equations. Draw the graph of each.
(a) y=l(x-3)2+2 (b) x = 5y2 - 20y + 23

5.1.4 For each of the following parts, find an equation whose graph is a parabola with the given
properties.

(a) The focus is (2,2) and the directrix is x = 4.

(b) The vertex is (4,6), the axis of symmetry is horizontal, and the parabola passes through (—2,2).

5.1.5
* The latus rectum of a parabola is a line segment passing through
the focus of the parabola such that the segment is parallel to the directrix
and has both endpoints on the parabola, as shown in the diagram. Show
that the latus rectum of the graph of the equation y = a(x- h)2 + k has length
1 ∕α. Hints: 327

5.2 Problem Solving With Parabolas

In Section 5.1, we learned the standard form for parabolas, but this is not the only useful tool for working
with parabolas. In fact, you've already learned some other tools for dealing with parabolas.

II Problems

Problem 5.8: A parabola has vertex (4,3) and its axis of symmetry parallel to the y-axis. If one
x-intercept is (1,0), find the other x-intercept. (Source: CMO)

Problem 5.9: A parabolic arch has a height of 20 feet and a span of 30 feet. 1I
How high is the arch 6 feet from its center?

20

1 6 I
30

118
 5.2. PROBLEM SOLVING WITH PARABOLAS

Problem 5.10: Parabola ePhas a vertical axis of symmetry and passes through the points (4,5), (-2,11),
and (-4,21). Find an equation whose graph is fP.

Problem 5.8: A parabola has vertex (4,3) and its axis of symmetry parallel to the y-axis. If one
x-intercept is (1,0), find the other x-intercept. (Source: CMO)

Solution for Problem 5.8: Solution 1: Find the equation. The axis of symmetry is vertical, so the parabola is
the graph of an equation of the form y = a(x - h)2 + k. From the vertex, we have h = 4 and k = 3, so our
equation is y = a(x - 4)2 + 3. The parabola passes through the point (1,0), so we can find a by plugging
this point into our equation. This gives 0 = a(l - 4)2 + 3, so a = -∣ and our equation is y = -∣(x - 4)2 + 3.

To find the x-intercepts of the graph, we let y = 0, which gives 0 = — ∣(x — 4)2 + 3. Subtracting 3, and
then multiplying by -3, gives (x - 4)2 = 9, so x - 4 = ±3. This gives us x = 7 and x = 1 as solutions, so
(7,0) is the other x-intercept.

Hmmm.... The x-intercepts are (1,0) and (7,0), which are 3 to the left and 3 to the right, respectively,
of the axis of symmetry x = 4. Aha! There's a much faster solution!
Solution 2: Forget the formulas, think about the problem. The axis of symmetry of the parabola is the
vertical line through the vertex. The vertex is (4,3), so the axis is x = 4. The point (1,0) is on the parabola,
so its reflection over x = 4, which is (7,0), is also on the parabola. □

Concept: Don't depend too heavily on your formulas. A little thinking about a
Q=~mlJ problem can go a long way.

Concept: Symmetry can be a very powerful problem-solving tool.

O≡^

Problem 5.9: A parabolic arch has a height of 20 feet and a span (width at its base) of 30 feet. How
high is the arch 6 feet from its center?

Solution for Problem 5.9: We know how to describe parabolas on

the Cartesian plane with equations, so we create a model of the
arch on the Cartesian plane. We have our choice of what point
to choose as the origin when placing the arch on the Cartesian
plane. We want to do so in a way that will make our algebra as
simple as possible. Specifically, we want many of the points we
know something about to have 0 as one or both of its coordinates.
In order to make calculations simplest, we let the parabola be
a downward opening parabola whose line of symmetry is the
y-axis. We place its vertex at (0,20) so that its span is the distance
between its x-intercepts, which are symmetric about the y-axis.

119
CHAPTER 5. CONICS

Since the span is 30, these intercepts are (15,0) and (-15,0). We therefore seek the value of c such that
(6, c) is on the parabola.
Here are a couple ways we can find the equation for our parabola:
Method 1: Use standard form. The parabola opens downward and has vertex (0,20), so its equation is
y = «(x - 0)2 + 20 for some negative value of a. Since the parabola passes through (15,0), we let x = 15
and y = 0 in this equation to find a = =
Method 2: Use the roots. Since the x-intercepts of our graph are (15,0) and (—15,0), the roots of our
desired quadratic are 15 and -15. Therefore, our quadratic is y = a(x + 15)(x — 15) for some constant a.
Since (x, y) = (0,20) must also satisfy this equation, we have 20 = a(0 — 15)(0 + 15), which gives a = —gg,
as before.
Our equation for the parabola is therefore = _ 4 χ2 + 20. If (6,c) is on the graph of this equation,
we have
84
5 '
So, the height of the arch 6 feet from its center is y = 16.8 feet. □

Problem 5.10: Parabola lP has a vertical axis of symmetry and passes through the points (4,5),
(-2,11), and (-4,21). Find an equation whose graph is tP.

Solution for Problem 5.10: We might start by noting that because tP has a vertical axis of symmetry, the
desired equation has the form y = a(x - h)2 + k. We can then use the three given points to produce the
system of equations

5 = «(4 - ∕ι)2 + k,
11 = α(-2 - h)2 + k,
21 = «(-4 - h)2 + k.

At first glance this system of equations looks pretty scary. (But it's not impossible to solve; see if you
can do so.) Fortunately, there's an easier way to approach the problem. Standard form is not the only
form we know for a quadratic. We can also write the equation in the form y = ax2 + bx + c. Plugging the
three given points into this form gives a much simpler system:

16« + 4b + c = 5,
4a-2b + c = 11,
16« — 4⅛ + c = 21.

Subtracting the last equation from the first gives b = -2. Subtracting the second from the first gives
12« + 6b = -6. Letting b = -2 in this equation gives a = 5. Finally, c = 5 - 16« - 4b = 5, so our desired
equation is y = ∣x2 - 2x + 5. □
A common thread in the three problems in this section is the importance of being flexible in your
approach to problems. We used symmetry, roots of quadratics, standard form, and general form.

120
 5.3. MAXIMA AND MINIMA OF QUADRATICS

Concept: Don't get locked into one way of thinking about quadratics and parabolas.
O≡≡S

l∖<'MiM¼

5.2.1 Find an equation whose graph is a horizontally-opening parabola that passes through (20, -3),
(12,1), and (33,2).

5.2.2 A parabolic arch is 30 meters wide at its base. A fly sits on the arch 25 meters above a point on
the ground that is 5 meters from one point where the arch touches the ground. How high is the center
of the arch?

5.2.3 The graph of y = x2 + 2x — 2 is reflected over the line y = x. How many points of intersections
does the original graph have with its reflection?
5.2.4 If the graph of the equation y = ax2 + 6 is tangent to the graph of y = x, then what is a? (Source:
Mandelbrot) (A line is tangent to a parabola if it touches the parabola at one point, but is otherwise
always “outside" the parabola.)

5.3 Maxima and Minima of Quadratics

Problems

Problem 5.11: Let f(x) = x2 + 6x + 5.

(a) Graph f.
(b) Find the smallest possible value of ∕(x) if x is a real number.
(c) How is the value of x for which ∕(x) is minimized related to the roots of f(x)2

Problem 5.12: Let ∕(x) = ax2 + bx + c, where a, b, and c are real numbers and a > 0. In terms of a, b,
and/or c, find the real value of x that makes ∕(x) as small as possible. What happens if a is negative?

Problem 5.13: Let k be a real number. Find the minimum possible distance between k(l + 3z) and 2z in
the complex plane.

Problem 5.14: Greg can bicycle at 20 miles per hour on pavement and 10 miles per hour on gravel.
Greg is building a rectangular track with one side paved and three sides gravel. Greg would like to
build the track so that it takes him 3 hours to ride all the way around the track. In this problem, we
find how long the paved side should be if he wishes his track to enclose as large an area as possible.
(We assume the track itself has negligible width.)
(a) Let the dimensions of the track be x and y, where the paved side has length x. Use Greg's speed
on pavement and gravel to find an equation for x and y, and find an expression in terms of x
that we wish to maximize.
(b) Determine how long the paved side should be if he wishes his track to enclose as large an area
as possible.

121
CHAPTER 5. CONICS

Problem 5.15: Let ∕(x) = Vαx2 + bx. For how many real values of a is there at least one positive value
of b for which the domain of f and the range f are the same? (Source: AMC12)

Problem 5.11: Let ∕(x) = x2 ÷ 6x + 5.

(a) Graph f.
(b) Find the smallest possible value of ∕(x) if x is a real number.
(c) How is the value of x for which ∕(x) is minimized related to the roots of /(x)?

Solution for Problem 5.11:

(a) Since ∕(x) is a quadratic, its graph is a parabola. We complete the square to
give ∕(x) = (x + 3)2 - 4, and we see that the graph of f is an upward-opening
parabola with vertex (-3, -4), as shown at right.
(b) Our graph gives the answer right away. The smallest value of ∕(x) corresponds
to the y-coordinate of the lowest point on the graph of f. This lowest point is
the vertex, (-3, -4), so the smallest possible value of ∕(x) is —4, which occurs
when x - -3.
Algebraically speaking, we can see that -4 is the smallest possible value of ∕(x) after we
have completed the square to find ∕(x) = (x + 3)2 - 4. Since (x + 3)2 ≥ 0 for all real x, we have
(x + 3)2 - 4 ≥ -4, which means ∕(x) > -4. We also have ∕(-3) = -4, so -4 is indeed the minimum
of f(x)-

(c) Factoring gives us ∕(x) = (x + l)(x + 5), so the roots of ∕(x) are —1 and -5. The value of x that
minimizes ∕(x) is the average of these roots. Our graphical approach to finding the minimum
explains why. The vertex of a parabola is on the axis of symmetry. The parabola's x-intercepts,
whose x-coordinates are the roots of ∕(x), are symmetric about the axis of symmetry. So, the
x-coordinate of the vertex is the average of the roots of ∕(x).

□
By thinking about the graph of a parabola, we have now learned how to find the minimum or
maximum of a quadratic expression:

Concept: To find the minimum or maximum value of a quadratic, complete the

θ .ol-j square in the quadratic.

We can use this tactic to generalize our solution to Problem 5.11.

Problem 5.12: Let ∕(x) = ax2 + bx + c, where a, b, and c are real numbers and a > 0. hi terms of a, b,
and/or c, find the real value of x that makes ∕(x) as small as possible. What happens if a is negative?

Solution for Problem 5.12: As we often do with generalizations, we use the solution to a specific case
as our guide. In Problem 5.11, we completed the square, so we do the same here. Rather than walk

122
 5.3. MAXIMA AND MINIMA OF QUADRATICS

through all the steps again, we recall that we completed the square for ax2 + bx + c when proving the
quadratic formula back on page 92, and we found

∕ω = o(^ + ⅛) +c-^∙

The graph of f is an upward-opening parabola with vertex (~⅛,c~ Therefore, its lowest point
has y-coordinate c - which occurs when x = -½-, which means that the minimum value of ∕(x) is
∕(-⅛H-g.

If a is negative, the parabola opens downwards, so (-⅛, c - is the highest point on the parabola.
Therefore, f (-⅛) = c - is the maximum value of ∕(x). □

Important: Let ∕(x) = ax2 + bx + c, where a, b, and c are constants.

ψ • If a > 0, then the minimum value of ∕(x) is c - ■£, which occurs

when x =

• If a < 0, then the maximum value of ∕(x) is c - which occurs

when x =

You don't have to memorize these formulas. As we have seen, we simply complete the square in a
quadratic to find its minimum or maximum.
Notice that this result also verifies our observation in Problem 5.11 that the value of x that minimizes
∕(x) = ax2 + bx + c is the average of the roots of ∕(x). Vieta's Formulas tell us that the sum of the roots of
∕(x) is -⅞, so the average of the roots is -γa. Furthermore, this relationship holds even when the roots
are nonreal.
Problem 5.13: Let k be a real number. Find the minimum possible distance between k(l + 3z) and 2z
in the complex plane.

Solution for Problem 5.13: As discussed in Section 3.2, the distance between 2i and the point fc(l + 3z) in
the complex plane is
∣fc(l + 3z) - 2z∣ = ∖k + (3⅛ - 2)z∣ = √k2 + (3k - 2)2.

To minimize √k2 + (3fc - 2)2, we must minimize k2 + (3k - 2)2. This is just a quadratic, which we know
how to minimize. Expanding the square and combining like terms gives

k2 + (3k - 2)2 = k2 + 9k2 - 12k + 4 = 10k2 - 12fc + 4.

We could complete the square, but instead we use the result of Problem 5.12 to note that this quadratic
is minimized when k = —(—12)∕(2 ∙ 10) = 3/5. Therefore, the minimum value of the quadratic is
10 (∣) - 12 Q) + 4 = ∣, which means the minimum distance from k(l + 3z) to 2i is √2∕5 = V10∕5. □

123
CHAPTER 5. CONICS

Problem 5.14: Greg can bicycle at 20 miles per hour on pavement and 10 miles per hour on gravel, j
Greg is building a rectangular track with one side paved and three sides gravel. Greg would like to
build the track so that it takes him 3 hours to ride all the way around the track. How long should
the paved side be if he wishes his track to enclose as large an area as possible? (You can assume the
track itself has negligible width.)

Solution for Problem 5.14: We start as we usually do with word problems—we turn the words into math.
Let the dimensions of the rectangular track be x and y miles, where one of the x-mile sides is paved.
Therefore, we wish to maximize xy. Of course, we need more information about x and y to maximize
xy.
We turn to the only other information we have, the fact that it should take Greg 3 hours to ride all
the way around the track. Because Greg rides 20 miles per hour on paved road, he can complete the
paved side in hours. Similarly, he can cover the other three sides in hours. So, we must have
⅛+ ~ 3∙ Multiplying both sides by 20 to get rid of fractions, then rearranging, gives 3x + 4y = 60.
We now wish to maximize xy given that 3x + 4y = 60. Specifically, we want to find the value of
x that gives us this maximum, so we solve 3x + 4y = 60 for y in terms of x, which gives y = 15 - ∣x.
Substituting this into our area expression gives

xy = x (15 - ∣xj = -∣x2 + 15x.

We complete the square to find

-∣x2 + 15x = -∣ (x2 - 20x) = -∣ (x2 - 20x + lθθ) + ∣(100) = -∣(x - 10)2 + 75.

The expression -∣(x - 10)2 + 75 is maximized when x = 10. (Alternatively, we could have simply noted
that the maximum of ~∣x2 + 15x occurs when χ = = —15∕(2(-3∕4)) = 10.)
Therefore, Greg should make the paved side 10 miles long. □

Problem 5.15: Let ∕(x) = Vαx2 + bx. For how many real values of a is there at least one positive value I
of b for which the domain of f and the range f are the same? (Source: AMC 12)

Solution for Problem 5.15: First, we take a little time to understand exactly what the question is asking.

Concept: If a problem is difficult to understand, take extra time reading and re-
o=≡) reading the problem until you are sure you understand what the problem
is. You don't want to spend a lot of time working on the wrong problem!

Our problem requires us to count the values of a for which we can find a b that makes the domain
and range of f the same. We start our search for such a and b by finding the domain and range of f in
terms of a and b.
There are no negative numbers in the range of f because Vflx2 + bx cannot be negative. For the
domain of f, we must have ax2 + bx > 0 in order for f to be defined. We solved some quadratic

124
 5.3. MAXIMA AND MINIMA OF QUADRATICS

inequalities in Section 4.5 by factoring and considering the signs of the factors. Trying the same here,
we get x(ax + b) ≥ 0. Solutions to this inequality are in the domain of f. In the problem, we are told
that b must be positive. However, we don't know if a is positive, negative, or 0. So, we have 3 cases to
consider:

• Case 1: a > 0. We know that there cannot be negative numbers in the range of f, so we check if
there are negative numbers in the domain of f when a > 0. If a and b are positive, then the solution
to the inequality x(ax + ⅛) ≥ 0 is x ∈ (-oo, - £] U [0, +∞). If you don't see this right away, note that if
a > 0, then the graph of g(x) = x(ax + b) is an upward-opening parabola. Since g(x) has roots x = 0
and x = -∣, the graph of g is below the y-axis for 0 < x < -∣, and above or on the y-axis otherwise.
Therefore, there are negative numbers in the domain of f.
Since there are no negative numbers in the range of f, but there are negative numbers in the
domain if a > 0, there are no positive values of a that satisfy the problem.

• Case 2: a = 0. In this case, our inequality becomes bx ≥ 0, which is true for all x ≥ 0. Therefore, if
« = 0, the domain of f is all nonnegative numbers. Because ∕(x) = Vbx when a — 0, the range of
f is also all nonnegative numbers when b > 0. We see this by noting that for any nonnegative y,
setting x = y2∣b gives Vbx = y, so y is in the range of f. Therefore, a = 0 satisfies the problem.

• Case 3: a < 0. When a < 0 and b > 0, then the solution to the inequality x(ax + b) ≥ 0 is x ∈ [0, —1].
(Note that a < 0 and b > 0, so -~ is positive.) We can visualize the solution to this inequality with
graphing. If a < 0, then the graph of g(x) = x(ax + b) is a downward-opening parabola. Since g(x)
has roots x = 0 and χ = -⅛, the graph of g is above or on the y-axis for 0 ≤ x ≤ —1, and below the
y-axis otherwise.
The domain of f when a < 0 is therefore [0,-∣]. Next, we focus on the range of f. We first
note that the smallest possible value of ∕(x) is 0, which we attain at both x = 0 and x = -b∣a. But
what is the largest possible value of /? To find the maximum value of ∕, we must first maximize
ax2 + bx. Now, we're in familiar territory. The maximum of this expression occurs when x =
So, if a < 0, the maximum of ∕(x) is

This must match the maximum value of our range, -∣, so we have = -∣. Rearranging this
equation gives a = -2 y∣^a. Squaring both sides gives α2 = -4λ, from which we find a - 0 or a = -4.
We have therefore found another value of a, namely a = -4, for which the domain and range of f
are the same for some positive value of b.

Combining all our cases, we have 2 values of a for which there exists a positive value of b such that
the domain and range of ∕(x) = Vαx2 + bx are the same. □

5.3.1 Find the largest possible value of ∕(x) = -2x2 + 6x + 9.

125
CHAPTER 5. CONICS

5.3.2 A farmer is going to fence his field by first forming a rectangle with his fence, then cutting this
rectangle into three congruent rectangles with two additional fences, where these additional fences must
be parallel to each other. He has 80 yards of fence. What is the maximum area, in square yards, of field
that he can enclose in this manner?
5.3.3 Suppose that p and q are constants such that the smallest possible value of x2 + px + q is 0. Express
q in terms of p.
5.3.4 Suppose the temperature of each point (x, y) in the plane is given by the function x2 + y2 — 4x + 6y.
Find the coldest point in the plane and determine its temperature.
5.3.5⅛ A table is to be constructed by gluing together 68 1 × 1 × 1 wooden cubes. All four legs and the
rectangular top will be formed by the cubes. The four legs must be the same length and need to be one
cube thick, and the top is one cube thick, as well. What is the maximum possible volume of the space
between the table's top and the floor, excluding the space taken up by the table's legs? Hints: 3

5.4 Circles

The locus of points in a plane that are a fixed distance from a given point in that plane is called a circle.
The fixed distance is the radius of the circle, and the given point is the center of the circle.
You're probably already familiar with representing circles in the Cartesian plane with equations. In
this chapter, we review this material and apply it to a couple more challenging problems.

Problems
____________
Problem 5.16: Show that the circle with center (ft, k) and radius r, where r > 0, is the graph of the
equation (x - ∕z)2 + (y - k)2 - r2.

Problem 5.17: Find the distance from (5,7) to the nearest point on the graph of 2x2 + 4x + 2y2 - 8y = 6.

Problem 5.18: Let x and y be real numbers satisfying the equation x2 - 4x + y2 + 3 = 0. Find the
maximum and minimum values of x2 + y2. (Source: NYSML)

Problem 5.19: A car travels due east at ∣ mile per minute on a long, straight road. At the same time,
a circular storm, whose radius is 51 miles, moves southeast at -y- mile per minute. At time t = 0, the
center of the storm is 110 miles due north of the car. At time t = t-↑ minutes, the car enters the storm
circle, and at time t - tz minutes, the car leaves the storm circle. Find (L + ⅛)∕2. (Source: AIME)
Hints: 346

Problem 5.16: Show that the circle with center (h, k) and radius r, where r > 0, is the graph of the
equation (x - h)2 + (y - k)2 = r2.

Solution for Problem 5.16: Suppose that the point (x,y) is on the circle with center (∕z,fc) and radius r.

126
 5.4. CIRCLES

Then, the distance between (x,y) and (lι,k) is r, so the distance formula gives us

y∣(x - ⅛)2 + (y - k)2 = r.

Conversely, any point (x, y) that satisfies this equation is r away from (/?, k), which means that (x, y) is on
the circle with center (h, k) and radius r. Because (x - ∕ι)2 + (y - k)2 is nonnegative, the equation above is
equivalent to
(x - h)2 + (y - k)2 = r2.
The graph of this equation is a circle with center (∕z, k) and radius r. □

Important: The standard form of an equation whose graph is a circle with center
(∕ι, ∕c) and radius r is
(x - h)2 + (y - k)2 = r2.

Problem 5.17: Find the distance from (5,7) to the nearest point on the graph of 2x2 + 4x + 2y2 - 8y = 6

Solution for Problem 5.17: We begin by finding the graph of the given equation.
We first divide by 2 to make the coefficients of x2 and y2 equal to 1. This gives
x2 + 2x + y2 - 4y = 3. We complete the square in x and in y by adding both 1
and 4 to both sides. This gives us x2 + 2x + 1 + y2 - 4y + 4 = 3 + 1 + 4, so

(x + 1)2 + (y - 2)2 = 8.

The graph of this equation is a circle with center (-1,2) and radius √8 = 2 λ∕2. The circle is shown in the
graph at right. The center of the circle is labeled C, and the point (5,7) is labeled X.

The nearest point on the circle to X is the point where CX intersects the circle. To see why, we let
CX intersect the circle at Y, and let Z be some other point on the circle. The Triangle Inequality tells us
that CZ + XZ > CX. We also have CX = XY + CY and CY = CZ (because both are radii of the circle), so
CZ + XZ > CX becomes CY + XZ > XY + CY. This means that XZ > XY, so Y is closer to X than any
other point on the circle is.
Because CY is a radius of the circle, we have CY = 2 y∕2. The distance formula gives us

CX = √(5 + l)2 + (7-2)2 = √61,

so XY = CX - CY = √61 - 2 √2. □

Problem 5.18: Let x and y be real numbers satisfying the equation x2 - 4x + y2 + 3 = 0. Find the I
maximum and minimum values of x2 + y2. (Source: NYSML)

Solution for Problem 5.18: Let k = x2 + y2. We now seek an expression in terms of k that we can maximize
and minimize. We might at this point substitute k for x2 + y2 in the given equation, but before we do so,
we recognize the form of the two equations x2 + y2 =k and x2 - 4x + y2 + 3 = 0. The graphs of these two
equations are circles; maybe we can solve this problem with geometry instead of algebra.

127
CHAPTER 5. CONICS

The graph of x2 + y2 = k is a circle with center (0,0) and radius Vfc. Completing I
thesquareinxandrearrangingtheequationx 2-4x+y2+3 = Ogives (x-2)2+y2 = 1.
The graph of this equation is a circle with center (2,0) and radius 1. We'll call this z Λ,.. -
∕ ×
circle C, and its graph is the solid circle shown at right. Every point (x, y) that I

∖ J
satisfies x2 - 4x + y2 + 3 = 0 must be on this circle. If we also have x2 + y2 = k, then (V
∖ ∕
the point (x, y) must be on the circle with center (0,0) and radius Vk- . j 1

So, to find the largest and smallest possible values of k, we must simply find i
I

the largest and smallest circles centered at the origin that intersect circle C. The largest is shown as a
dashed circle, and has radius 3. The smallest is shown as a dotted circle, and has radius 1. Since the
radius in each case equals √fc, the largest possible value of k is 9 and the smallest possible value is 1. □

Concept: Be alert for equations you know how to graph easily. Sometimes geometry
o≡≡ can be used to solve algebra problems.

Problem 5.19: A car travels due east at ∣ mile per minute on a long, straight road. At the same time,
a circular storm, whose radius is 51 miles, moves southeast at mile per minute. At time t = 0, the
center of the storm is 110 miles due north of the car. At time f = f1 minutes, the car enters the storm
circle, and at time t = ⅛ minutes, the car leaves the storm circle. Find (fι + f2)∕2. (Source: AIME)

Solution for Problem 5.19: We have a long, wordy problem, so we first try to turn all these words into
equations and expressions we can manipulate. We set the problem up on the Cartesian plane, since this
will give us a way to model the location of the car and the storm at any time t. We let the car start at the
origin. Because it travels due east at ∣ mile per minute, its location is (∣f, 0) after t minutes.

The storm is a little harder to work with. At time t = 0, the center of the storm is 110 miles north of
the car, at (0,110). Because the storm moves southeast at ⅛ √2 mile per minute, its center moves south ∣
mile and east ∣ mile each minute. Therefore, the center of the storm at time t is (½, 110 -1). The radius of
the storm is always 51, so the equation whose graph represents the storm's boundary on the Cartesian
plane is
(x _ t)2+(y _ (110 -1)) =5f.
The values of t at which the car enters and leaves the storm correspond to the values of t at which the
point representing the car is on the circle representing the storm. So, we seek the values of t such that
(∣f,0) satisfies the equation for our storm. Setting x = ∣t and y = 0 in our storm equation gives us

ff-0,∙(∙-l""-3,->',
Simplifying the left side gives us

(ι4)2÷(-ιιθ÷^24÷^≡÷-⅛-≡÷-2∙
so our equation is - 110t + 1102 = 512. Solving that doesn't look like much fun. Before we dive into
the quadratic formula, we reread the problem and notice that all we want is the average of the solutions
to this equation, not the solutions themselves.

128
 5.5. ELLIPSES

Concept: Keep your eye on the ball! Don't forget what you're looking for in a
O≡≡^≡ problem—you don't always have to solve for every variable in a problem
to solve the problem itself.

The average of the solutions to this quadratic is half the sum of the quadratic's roots. Vieta's Formulas
give us this sum quickly: -(-110)∕(5∕18) = 396. Half this sum is our desired average, 198. □

Exercises

5.4.1 Find the center and the radius of the graph of 4x2 + 4y2 + 12x + 16y + 9 = 0.
5.4.2 Find an equation whose graph is a circle that has a diameter with endpoints (3,2) and (7, -4).
5.4.3 A chord of a circle is a line segment whose endpoints are on the circle. Find the length of the
common chord of the two circles whose equations are x2 + ι∕2 =4 and x2 + ι∕2 — 6x + 2 = 0. (Source:
CEMC)

5.4.4 Square ABCD has sides of length 4, and M is the midpoint of CD. A circle with radius 2 and
center M intersects a circle with radius 4 and center A at points P and D. What is the distance from P to
AD? (Source: AMC)

5.5 Ellipses
Suppose we have points Fι and F2 in a plane. For any positive Bi
constant k greater than the distance FγF2, the locus of all points P ~"~∣
in the plane such that PF-[ + PF2 equals k is called an ellipse, an y< ∣
example of which is shown at right. The points F↑ and F2 are the ∕ [ ∖
foci of the ellipse. ("Foci" is the plural of focus.) The midpoint of ^↑Q -----→- -4Aj
the segment connecting the foci is the center of the ellipse. ∖ ! J
A line segment with its endpoints on the ellipse is called a chord
of the elhpse. The chord that passes through both foci is the major B2
axis of the ellipse, while the chord through the center that is perpendicular to the major axis is called
the minor axis. So, in the given diagram, A1A2 is the major axis and B1B2 is the minor axis of the ellipse
above. These axes meet at the center, point C.

Sidenote: You can make your own ellipse with two pins, a string, and a pencil. Push
your pins into a piece of paper on a hard surface. These are the foci of
your ellipse. Make sure that your string is at least twice as long as the
distance between the pins. Tie the ends of the string together so that you
now have a loop. Place your loop around both pins. Finally, make your
ellipse: hook your pencil inside the loop, pull the loop taut against the pins
and pencil, then, keeping the string taut and pencil perpendicular to the
paper throughout, trace the entire boundary you can reach with the pencil.
Why does this process produce an ellipse?

129
CHAPTER 5. CONICS

Problems

Problem 5.20: Suppose the ellipse 8 has foci Fι = (c, 0) and F2 = (-G θ) for some positive c. Moreover,
suppose that for every point P on 8, we have PF] + PF2 = 2π for some nonzero constant a such that
« > c.
(a) Why must we have a > c?
(b) In terms of a and/or c, find the coordinates of the points on the x-axis that are on 8, and prove
that PFχ + PF2 equals the length of the major axis of 8 for any point P on 8.
(c) Suppose 8 passes through the point (0, b), where b > 0. Find b in terms of a and c.
(d) Suppose P = (x, y) is on 8. Use the equation PF1 + PP2 = 2a to show that 8 is the graph of the
equation

= 1.

(This part requires a considerable amount of algebraic manipulation to get rid of square root
symbols; don't give up too quickly!)

y2
Problem 5.21: Consider the graph of the equation — + 7- = 1.
a2- bz
(a) What happens to the graph if a = b?
(b) In which of the three graphs below is a/b largest? The smallest?

(c) What happens if a < b?

Problem 5.22: Suppose that ellipse & has (h,k) as its center, (h + c,k) as a focus, and a major axis of
length 2a. Show that 8 is the graph of the equation

(x - ∕z)2 (y - ^)2 _ 1
fl2 - i'

where b2 = a2 - c2.

Problem 5.23: Let C be the center of an ellipse, Aχ and A2 be the endpoints of the ellipse's major axis,
Bi and B2 be the endpoints of its minor axis, and F↑ and F2 be its foci.
(a) Why must B↑F↑ = B1F2? Why must B1F1 + B1F2 = A1A2?
(b) Show that B]Fj equals half the length of the major axis of the ellipse.
(c) Find an equation that relates the distance between the foci to the lengths of the axes of the ellipse.

130
 5.5. ELLIPSES

Problem 5.24: Graph each of the ellipses described below, and find an equation that produces the
given graph. Then, find the center, foci, and the lengths of both axes of each ellipse.
(a) The center is (4, -2), the major axis is horizontal with length 6, and (4, -3) is one endpoint of the
minor axis.
(b) The foci are (-3, -1) and (-3,3), and the ellipse passes through (-5,1).

Problem 5.25: Find the center, foci, and the lengths of both axes of the graph of each of the equations
below, and then graph each equation.
(λ-1)2 (y + 7)2
(a) = 1 (b) x2 + 4y2 + 6x + 16y + 24 = 0
16 25

Problem 5.26: What is the graph of each of the following equations:

(b) 9x2 + 4y2 - 16y + 52 = 0

Problem 5.27: The graphics program we used to write this book doesn't have y
a command to graph an ellipse. It does, however, have a command to graph
a circle given its center and radius. It also allows the user to scale a graph
horizontally or vertically. (When we "scale a graph horizontally," we multiply
the x-coordinate of every point on the graph by some fixed constant.) In this
problem, we figure out how to use this program to produce the ellipse at right,
and explain why this process does indeed produce the correct ellipse.

(a) Draw a circle on the Cartesian plane that you can scale horizontally by a factor of 2 to produce
the shown ellipse. What is the equation whose graph is your circle?
(b) Suppose that (x, y) is on the circle you produced in part (a). What point must be on the graph
produced by scaling your part (a) graph horizontally by a factor of 2?
(c) Find the equation whose graph results from scaling the circle in part (a) horizontally by a factor
of 2.
(d) Does this process of scaling a circle horizontally produce an ellipse?
(e) Take a guess at a formula for the area of an ellipse with axes of lengths 2a and 2b.

Problem 5.20: Suppose the ellipse 8 has foci F↑ = (c, 0) and F2 = (-c, 0) for some positive c. Moreover,
suppose that for every point P on 8, we have PF↑ + PF2 = 2a for some nonzero constant a such that
a > c. Find the equation whose graph is this ellipse.

Solution for Problem 5.20: We start by wondering why we must have a > c. The Triangle Inequality tells
us that PF] + PF2 ≥ F-[F2, where equality holds if and only if P is on F1F2∙ Since PF↑ + PF2 = 2a and
F1F2 = 2c, we must have a >c, with equality if and only if P is on F1F2. Therefore, if a < c, then there are
no points P that satisfy PF↑ + PF2 = 2a; and ifa = c, then only points on F↑F2 satisfy PF↑ + PF2 = 2a (and

131
CHAPTER 5. CONICS

our locus of points such that PF↑ + PF2 = 2a is a line segment). So, we must have a > c to produce an
ellipse as our locus.

To find the equation whose graph is this ellipse, we first

find a few points on the ellipse. We aim first at the easy points:
the points where the ellipse intersects the axes. Let A↑ = (%ι,0)
be a point where 8 intersects the x-axis. Because Λ] is on the
ellipse, we have A↑Fγ + A1F2 = 2a. We know that A↑ is not on
F1F2, since the sum of the distances from any point on F1F2 to
the endpoints of this segment is 2c, not 2a. So, suppose Xj > c,
so Aι is to the right of Fχ. Then, we have A1F1 = χ1 - c and
A1F2 = x↑ + c, so AiFj + A1F2 = 2xι. Since this must equal 2a,
we have xi = a, which means Ai = (a, 0). Similarly, the other
point where 8 intersects the x-axis is A2 = {-a, 0). Therefore, the
length of the major axis is 2a, and this length equals the constant
sum PFι + PF2 for points P on the ellipse.

Important: If P is on an ellipse with foci F↑ and F2, then PF↑ + PF2 equals the length
of the major axis of the ellipse.

Next, suppose that B↑ = (0,&), with b > 0, is a point where 8 intersects the y-axis. Then, since
Fι = (c, 0) and F2 = (-c, 0), we have

B1F1 + B1F2 = √⅛2 + c2 + √b2 + c2 = 2 √b2 + c2.

Because Bi is on the ellipse, this sum must equal 2a, which gives us Vb2 + c2 = a. Squaring both sides of
this and rearranging the result gives us b2 - a2 - c2. Similarly, the point B2 = (0, -b), where b2 = a2 - c2
and b > 0, is on 8. So, the minor axis has length 2b.
We now have the endpoints of both axes. Let's try to find an equation whose graph is the whole
ellipse. Let P = (x, y) be a point on the ellipse. From PF↑ + PF2 = 2a, we have

^(x - c)2 + y2 + y∣(x + c)2 + y2 = 2α.

Unfortunately, we have no choice but to square this equation to get rid of the radicals. Instead of
squaring right away, we first move one of the radicals to the right to get

^(x-c)2 + y2 = 2a - ^(x + c)2 + y2.

This allows us to square the equation without having to multiply the two radical expressions. Squaring
this equation gives (x - c)2 + y2 = 4α2 - 4α √(x + c)2 + y2 + (x + c)2 + y2. The y2 terms cancel, and we can
isolate the remaining square root to find

4α ^(x + c)2 + y2 = 4a2 + (x + c)2 - (x - c)2 = 4λ2 + 4xc.

132
 5.5. ELLIPSES

Dividing by 4 gives us α ^∕(.v + c)2 + y2 = a2 + xc. Squaring again gives a2(x + c)2 + a2y2 = a4 + 2a2xc + x2c2.
Expanding the square on the left gives

a2x2 + 2a2xc + a2c2 + a2y2 = a4 + 2a2xc + x2c2.

The 2a2xc terms cancel. We then move the x2 and y2 terms to the left and the other terms to the right to
give
(α2 - c2)x2 + α2y2 - a4 - a2c2.
Aha! We recognize a2 - c2 = b2 on the left. We find b2 again on the right when we factor a2 out:
a4 - a2c2 = a2(a2 - c2) = a2b2. Now, our equation is b2x2 + α2y2 = a2b2. Dividing by a2b2 gives us the
surprisingly simple
χ2 y2 1
— + — = ι.
a2 b2
□

We can reverse the steps above to show that:

Important: The graph of the equation + f∑ = 1, where a > b > 0, is an ellipse with

ψ major axis of length 2a, minor axis of length 2b, and foci (c, 0) and (—c, 0)z
where c = Va2 - b2.

__ 2 ∣A
The endpoints of the horizontal axis of the graph of ⅛r + = 1 are (a, 0) and (-a, 0). To get a sense for
why, we note that in the equation + p- = 1, both and p∙ are squares, so they are both nonnegative.
Since their sum is 1 and both are nonnegative, each must be no greater than 1. We can only have = 1
if jL = 0, so we only have x = a or x = -a if y = 0. Moreover, we cannot have x2 > a2, so x can be no
farther from 0 than a is. Therefore, («, 0) and (-a, 0) are the most extreme points horizontally from the
center. Similarly, (0, b) and (0, -b) are the most extreme points vertically from the center.
With this in mind, we can now get a feel for this equation by thinking about what happens as we
vary a and b.

133
CHAPTER 5. CONICS

Solution for Problem 5.21:

(a) If a -b, then we can rewrite the equation as x2 + y2 = a2. The graph of this equation is a circle with
center (0,0) and radius a. But how can this be? We're studying ellipses, not circles.
Important: Circles are ellipses.
Ψ ___ __________ !
But if circles are ellipses, what are the foci of a circle? We know that the foci of an ellipse that is
the graph of the equation + ⅛ = 1 are the points (c, 0) and (-c, 0), where c = *4a 2- - b2. But when
a = b, we have c = 0, which puts both of our foci at the center of the circle, (0,0).
Now, we see why a circle can be considered an ellipse. If we let the foci f ι and F2 of an ellipse
be the same point, which we'll call O, then our condition that "PF∖ + PF2 is constant" becomes
"2PO is constant." In other words, every point on the ellipse is the same distance from O, which
means the ellipse is a circle.
(b) We know we have a circle when a = b. To analyze the shape of the ellipses that result as we choose
values of a that are larger than b, we consider the endpoints of the axes of the ellipses.

The endpoints of the major axis of the ellipse are (a, 0) and (-αz 0), while the endpoints of the
minor axis of the ellipse are (0, b) and (0, -b). As we saw in part (a), when a = b, we have a circle.
If a/b is slightly greater than 1, then the ratio of the length of the major axis to the length of the
minor axis is slightly greater than 1. This means the ellipse is almost a circle, as in the ellipse at
left above. Increasing the ratio a∣b increases the ratio of the major axis to the minor axis. The
larger this ratio is, the more "stretched" the resulting ellipse is. So, the ellipse at right above has
the largest ratio a/b and the one at left above has the smallest ratio a∣b.
Sidenote: Eccentricity is one measure of how "stretched" an ellipse is. The eccentric-
Jl> ity of an ellipse is defined as the ratio of the distance between the ellipse's
foci to the length of its major axis. So, the eccentricity of a circle is 0, and
the greater the eccentricity of an ellipse, the more "stretched" the ellipse is.
(c) Now that we've tackled what happens if a = b and as a gets greater and
greater than b, we wonder what happens if a < b. We start by finding a few
points on the resulting graph when a < b. The easiest, as usual, are those
on the coordinate axes. As before, (a, 0) and (-α, 0) are on the graph, as are
(0, b) and (0, -b). However, this time, we have a < b. In other words, the
major axis is vertical, not horizontal! So, the graph is still an ellipse, but the
foci and major axis are on the y-axis, not the x-axis. At right, we have the
graph of an ellipse in which the major axis is vertical.

134
 5.5. ELLIPSES

We have just discovered that the major axis of the graph of an equation of the form + p∙ = 1
2 ∖f^
is horizontal when a > b and vertical when b > a. We don't need to memorize this. As described in
Problem 5.21, we can simply think about where the endpoints of the axes are to see when the major axis
is vertical and when it is horizontal.
We now have a pretty good handle on ellipses centered at the origin. Of course, not all ellipses are
centered at the origin.

Problem 5.22: Suppose that ellipse 8 has (h, k) as its center, (h + c, ∕c) as a focus, and a major axis of
length 2a. Show that 8 is the graph of the equation

(¾-⅞)2 ι (y-fc)2 rrl

a2 b2
where b2 = a2 - c2.

Solution for Problem 5.22: Back in Section 5.1, we started with a parabola with its vertex at the origin, and
then translated the parabola so that its vertex was at (h,k). We use a similar strategy here.
In Problem 5.20, we found that an ellipse with foci (c, 0) and (-c, 0) and major axis of length 2(7 is the
graph of the equation
χ2 y2
-------∣- — = 1
a2 b2 '
where a > b > 0 and b2 = a2 — c2. Let this ellipse be !D. The origin is the center of this ellipse, since it is
the midpoint of the major and minor axes.
Because (h, fc) is the center of 8 and (∕ι + c, k) is one focus of 8, the other focus is c to the left of the
center, at (h -c,k). So, the center and the foci of 8 result from translating the center and foci of D right h
units and up k units. Furthermore, the major axes of the two ellipses are equal in length, so we suspect
that 8 is the result of translating D right h units and up k units.

To show this is the case, we let F↑ and F2 be the y

foci of Σ) and F' and F'2 be the foci of 8. We also let
P' = (x + h, y + k)
P = (x, y) be on D and let P' - (x + hry + k), so P,
is the result of translating P right h units and up k
units. We wish to show that P' is on 8.
___ Since P is on !D, we have PF↑ +PF2 = 2,a. Because
P'F' is a translation of PF↑, we have P'F^ = PF-[.
Similarly, we have P'F'2 = PF2, so P'F' + P'F'2 =
PF↑ + PF2 = 2a, which means that P' is on ellipse
8. Conversely, every point on 8 is the result of
translating some point on £) right h units and up k
units.
2 l√
So, for every point (x, y) on the graph of + = 1, there is a point (x', y') on the graph of 8, where
(x', y') = (x + /7, y + k). We therefore have x = x' -h and y = y' -k, where (x, y) is on D and (x', y') is on

135
CHAPTER 5. CONICS

8. Substituting these into our equation for £) tells us that 8 consists of all points (x', y') such that

(x' - h)2 (y' “ fc)2 = 1

a2 b2
Therefore, we have:

Important: The standard form of an equation whose graph is an ellipse with a

ψ horizontal major axis is

(x-∕ι)2 ι (y-k)2 1
«2 + b2

where a > b > 0. The center of the ellipse is (h,k), its major axis has
length 2ar its minor axis has length 2b, and its foci are (∕z ± c, k), where
c - V«2 - b2.

Note that we don't have to memorize the expressions for the lengths of the axes of the graph of

(x-∕z)2 (y-fc)2
a2 b2

Since the major axis is horizontal, its endpoints have y = k, just like the center. This leaves us - 2-■ = 1,
so (x - h)2 = a2, which means x -h = ±a. This tells us that the endpoints of the major axis are a to the
left and right of the center, so the major axis has length 2a. In other words, thinking about the values of
x and y that make each fraction equal 1 tells us the lengths of the axes. □
In much the same way, we can show that for an ellipse with a vertical major axis, we have:

Important: The standard form of an equation whose graph is an ellipse with a

ψ vertical major axis is

(x-h)2 (y-k)2
a2 b2 ~ '

where b > a > 0. The center of the ellipse is (∕ι,k), its major axis has
length 2b, its minor axis has length 2a, and its foci are (∕t, k ± c), where
c = yjb2 - a2.

Make sure you see that there's nothing really new here. This is basically the same information
that we learned in Problem 5.22. You shouldn't memorize all this information. Even the relationship
between c (the distance from each focus and the center) and the lengths of the axes has a nice a geometric
relationship that prevents us from having to memorize a formula for a, b, and c.

Problem 5.23: Show that in an ellipse, the distance between an endpoint of the minor axis to either
focus equals half the length of the major axis. Use this to find an equation that relates the distance
between the foci to the lengths of the axes.

136
 5.5. ELLIPSES

Solution for Problem 5.23: We start with a diagram at right. Let C be B1

the center of the ellipse, A↑ and A2 be the endpoints of the major
axis, B] and B2 be the endpoints of the minor axis, and F↑ and F2
be the foci. We wish to relate B1 F↑ to A]A2, the length of the major
axis, so we note that B↑F↑ + B↑F2 = A↑A2 because B-∣ is on the ellipse. 7^2
All we have to do now is take care of BxF2. Our diagram suggests
that B↑Fι = BxF2. To see that this is the case, we note that CFx = CF2
because the center of the ellipse is the midpoint of the segment
β2
connecting the foci. Since we also have B1C ± F1F2, we can apply
the Pythagorean Theorem to triangles ∆B1CF1 and ∆BxCF2 to see that

We also could have determined this by noting that the two right triangles are congruent.
Combining B1F1 = BxF2 and BxFι + B1F2 = AxA2 gives us 2BxFι = AxA2, so B↑F↑ = A∖A2∕2. The
Pythagorean Theorem applied to ∆B]CFι gives us (BxFι)2 = (CBf)2 + (CFf)2. Substituting B↑F↑ = A↑A2∕2
into this gives (AxA2∕2)2 = (CBx)2 + (CFx)2. Multiplying by 4 gives us

(A1A2)2 = 4(CB1)2 + 4(CF1)2 = (2CBi)2 + (2CFx)2.

Since C is the midpoint of BxB2, we have 2CBx = BxB2. Similarly, we have 2CFx = FxF2, so the equation
(AxA2)2 = (2CBx)2 + (2CFx)2 gives us (AxA2)2 = (BxB2)2 + (FxF2)2. This means we have

(major axis length)2 = (minor axis length)2 + (distance between foci)2.

This problem revealed a couple useful facts about ellipses:

Important: Let B be an endpoint of the minor axis of an ellipse, F be a focus of the

ellipse, and C be the center of an ellipse. Then, ∆BFC is a right triangle,
BC equals half the length of the minor axis, and BF equals half the length
of the major axis. Therefore, the length of the major axis, the length of
the minor axis, and the distance between the foci can be related by the
Pythagorean Theorem:

(major axis length)2 = (minor axis length)2 + (distance between foci)2.

As with our other formulas, this one doesn't need to be memorized. Once you understand it, you
can produce it whenever you need it.
To recap, we're now able to relate many notable parts of the graph of an ellipse with horizontal and
vertical axes to the equation that produced the graph. We first put the equation in the form

(x - ∕7)2 (y - fc)2
a2 b2 ~
The center is (h, k), since this point makes both fractions 0. The endpoints of the horizontal axis are those
points that make the first fraction 1 and the second 0, and the endpoints of the vertical axis make the

137
CHAPTER 5. CONICS

first fraction 0 and the second 1. Finally, we can find the distance from the foci to the center with the
relationship we just discovered with the Pythagorean Theorem.
If the axes of an ellipse are not horizontal and vertical, then the equation that produces the ellipse
cannot be written in the standard form + , 'z^√- = 1. We will learn how to handle such ellipses in a
later text in the Art of Problem Solving series.
We'll now get a little practice working with specific ellipses.

Problem 5.24: Graph each of the ellipses described below, and find an equation that produces the
given graph. Then, find the center, foci, and the lengths of both axes of each ellipse.
(a) The center is (4, —2), the major axis is horizontal with length 6, and (4, —3) is one endpoint of
the minor axis.
(b) The foci are (-3, -1) and (-3,3), and the ellipse passes through (-5,1).

Solution for Problem 5.24:

(a) We start with a sketch of the graph. Because the major axis is horizontal y
with length 6, its endpoints are 3 to the left and to the right of the center, at ∣ .l ...i— 1
(1, -2) and (7, -2), as shown. Since the given endpoint of the minor axis is 1 JJ
“Hl—
emit below the center, the other endpoint is 1 unit above the center, at (4, —1) I ∣ !

as shown. '
-≡ 1

To find the equation whose graph is this ellipse, we start with the standard 1 1 1
form
(x-∕z)2 (y-fc)2
a2 b2
From the center (4, -2), we know that our numerators on the left are (x - 4)2 and (y + 2)2. From
the lengths of the axes, the denominator of the x term is (6∕2)2 = 9 and the denominator of the y
term is (2∕2)2 = 1. (The x term has the larger denominator because the major axis is horizontal.)

Finally, we can find the distance between the foci with the relationship
(major axis length)2 = (minor axis length)2 + (distance between foci)2,

from which we find that the foci are 2 V2 apart. The major axis is horizontal, so the foci are V2 to
the left and to the right of the center, at (4 + ∖∕2, -2) and (4 - √2, -2).

(b) We start with a sketch of the ellipse. At first, we only know one point on
the ellipse, so weTl have to see what information about the ellipse we can
get from the foci. The foci of this ellipse are on the same vertical line, so the
major axis is vertical. The center is the midpoint of the segment connecting
the foci, which is (-3,1). Aha! This point is 2 units to the right of the given
point on the ellipse (-5,1). From this, we know that the minor axis has
length 2 ∙ 2 = 4. The minor axis is horizontal and the center is (—3,1), so the
equation whose graph is the ellipse is
(x + 3)2 i (y-D2 _
22 b2

138
 5.5. ELLIPSES

for some value of b. (We know the denominator of the first fraction is (4∕2)2 because the minor
axis is horizontal with length 4. We can even check our work at this point by noting that (-5,1)
satisfies this equation.) To find the denominator of the second fraction, we have to find the length
of the major axis. Since
(major axis length)2 = (minor axis length)2 + (distance between foci)2,
the length of the major axis is 4 V2. Therefore, the denominator of the (y -1)2 term in the equation
is (4 √2∕2)2 = 8, and our equation is
(x + 3)2 (y-l)2
4 8
Now that we have the length of the major axis, we see that the endpoints of the major axis are
(-3,1 + 2 √2) and (-3,1 - 2 √2). Using 2 √2 ≈ 2.8, we can locate these endpoints on the Cartesian
plane and produce the graph shown.

Problem 5.25: Graph each of the equations below, then find the center, foci, and the lengths of both
axes of the graph.
(x —1)2 (y + 7)2
(a) 16 + 25 ^1 (b) x2 + 4y2 + 6x + 16y + 24 = 0

Solution for Problem 5.25:

(a) First, we note that the denominator of the y term is larger than that of y
the x term, so the major axis is vertical. The numerators are both zero
for (x, y) = (1, -7), so this point is the center of the ellipse. The endpoints
of the major axis are directly above and below the center, so they occur
when x = 1, which leaves us = 1, so y + 7 = ± V∑5. This tells us that
the endpoints of the major axis are λ∕25 above and below the center, so
the major axis has length 2 V∑5 = 10. Similarly, the minor axis has length
2√16 = 8.
Using the relationship among the distance between the foci and the
lengths of the axes, we find that the foci are V100 - 64 = 6 apart, so they
are 3 from the center. Because the major axis is vertical, the foci are 3 above and 3 below the center,
at (1, -10) and (1, -4). Using the center and the lengths of the axes, we can produce the graph at
right.

(b) The equation has both an x2 and a y2 term, but the coefficients of these two terms are different. So,
we know the graph of the equation isn't a circle, since the coefficients of x2 and y2 are the same
in the general form of an equation whose graph is a circle. Therefore, we suspect the graph is an
ellipse, so we try to write the equation in the standard form of an equation whose graph is an
ellipse.
We start by completing the square. We first group the x terms and the y terms, and move the
constant to the other side. This gives us
x2 + 6x + 4y2 + 16y = -24.

139
CHAPTER 5. CONICS

Adding 9 to both sides completes the square in x. Factoring 4 out of each of the y terms gives
4(y2 + 4y), so we must add 4 ∙ 4 to both sides to complete the square in y:

x2 + 6x + 9 + 4(y2 + 4y + 4) = -24 + 9 + 4(4).

Therefore, our equation is (x + 3)2 + 4(y + 2)2 = 1. Uh-oh. Where are the denominators on the left
side? We can view the denominator of the (x + 3)2 term as 1, but what about the 4(y + 2)2 term; do
we let that denominator be 1 also? But then what do we do with the 4?

We move the 4 to the denominator of the 4(y + 2)2 term by writing y

this term with a denominator of ∣. This makes our equation

(x + 3)2 (y + 2)2
1 1/4
We know how to deal with this equation. The center is (-3,-2).
The major axis is horizontal because the denominator of the x term is
greater than that of the y term. The length of the major axis is 2 Vl = 2
and the length of the minor axis is 2 = 1. The foci are -^1 - ∖-~γ

units horizontally from the center, so the foci are (-3 + —, —2) and
(-3 - -2). Finally, we can produce the graph shown at right.

□
i Problem 5.26: What is the graph of each of the following equations:
½0
(a) j+ (b) 9x2 + 4y2 - 16y + 52 = 0
9

Solution for Problem 5.26:

(a) A quick glance at the left side might make us suspect that the graph of this equation is an ellipse.
However, the right side is 0, not 1. Because the numerators of the terms on the left are squares,
both terms are nonnegative for all real x and y. Therefore, we can only have 0 on the left side if
both terms equal 0, which happens only when (x, y) = (0,2). So, the graph of this equation consists
of this single point.
(b) The 4x2 and 9y2 terms suggest that the graph is an ellipse, so we try to write the equation in the
standard form for an ellipse. Completing the square in y and moving the constant to the right
gives 9x2 + 4(y - 2)2 = -36. Then, we divide by 36 to get

≤ + ^2)2 = _i
4 9
As in part (a), the two terms on the left must be nonnegative, so their sum cannot be negative.
Therefore, no ordered pair (x, y) of real numbers satisfies this equation. Hence, the graph of
this equation is empty—there are no points on the Cartesian plane whose coordinates satisfy the
equation. □

140
 5.5. ELLIPSES

In part (b) of Problem 5.25, we rearranged an equation of the form Ax2 + By2 + Cx + Dy + E = 0,
where A, Br C, D, and E are constants with AB > 0, into the standard form of an ellipse. However, in
Problem 5.26, we see that not all such equations can be put in the standard form of an ellipse. Therefore,
we see that:

WARNING!! Not every equation of the form Ax2 + By2 + Cx + Dy + E = 0, where

>Λ
Λ A, B, C, D, E are constants with AB > 0, has a graph that is an ellipse.
The graph of the equation may be a point, or may be empty.

In order to tell if the graph of an equation of this form is an ellipse, we try to write it in the standard
form of an ellipse by making one side —+ and the other side constant. If the constant is
positive, we can divide by it to produce our standard form, and the graph is an ellipse. If the constant
is nonpositive, then the graph is not an ellipse.

Sidenote: If the graph of Ax2 + By2 + Cx + Dy + E = 0, where A, B, C, D, E are constants

with AB > 0, is a point, the graph is sometimes referred to as a degenerate
ellipse.

Problem 5.27: The graphics program we used to write this book doesn't have
a command to graph an ellipse. It does, however, have a command to graph
a circle given its center and radius. It also allows the user to scale a graph
horizontally or vertically. (When we "scale a graph horizontally," we multiply
the x-coordinate of every point on the graph by a fixed constant.) How can we
use this program to produce the ellipse at right?

Solution for Problem 5.27: We've already seen that a circle is a specific type of
ellipse. This wasn't a surprise, because an ellipse looks like a circle that has been
"stretched." Our graphics program allows us to stretch a graph horizontally or
vertically, so maybe we can simply stretch a circle into an ellipse.
Our desired ellipse has a horizontal major axis with length 8 and a vertical
minor axis with length 4. So, we start with a circle with diameter 4 (shown at
right), and we plan to scale it horizontally by a factor of 2 to make our ellipse.

We have to show that this scaling does produce an ellipse, and not just some curve that looks a lot
like an ellipse but isn't quite an ellipse. To do this, we start with the equation whose graph is the circle,
and consider the effect of the scaling on this equation. The equation of the circle is simply x2 + y2 = 4.
But what does the scaling do?
The scaling doubles the x-coordinate of every point on the graph. So, for every point (x, y) on the
graph of x2 + y2 = 4, there is a point (x', y') on the scaled graph such that (x', y') = (2xz y). From x' = 2x,
we have x = x'∕2. Letting x = x'∕2 and y = y' in x2 + y2 = 4 gives us (x'∕2)2 + (y')2 = 4, which looks a lot

141
CHAPTER 5. CONICS

like an equation of an ellipse. Dividing by 4 then gives us

∞i + ⅛2 = ι
16 4
Graphing this equation does indeed give us the desired ellipse with horizontal major axis of length 8
and vertical minor axis of length 4. □
There's nothing special about scaling horizontally. If we similarly scale a circle in any direction, we
produce an ellipse. With some more advanced geometric tools, we can show that if we scale a closed
figure in some direction by a factor k, then the area of the scaled figure is k times the area of the original
figure. As an Exercise, you'll use this fact to explain why we have the following:

Important: The area of an ellipse with major axis of length 2a and minor axis of

ψ length 2b is abπ.

Exercises

5.5.1 Find an equation whose graph is the ellipse shown at right. y

5.5.2 Find the foci, center, and lengths of the axes of the graph of the equation
x2 + 4y2 + 8y = -3. Graph the equation.
5.5.3 Find the foci, the lengths of the major and minor axes, and the center
of the graph of the equation x2 + 2y2 - 12y + 4x + 2 = 0.
5.5.4 An ellipse has foci at (0,0) and (14,0) and passes through the vertex of
the parabola with equation y = x2 - 10x + 37. Find the length of the major axis
of this ellipse.
5.5.5 Scaling a closed figure in some direction by a factor k produces a figure with area k times the area
of the original figure. Use this fact to explain why the area of an ellipse with major axis of length 2a and
minor axis of length Zb is abπ.

5.6 Hyperbolas

Suppose we have points F↑ and F2 in a plane. For any positive constant k, the
locus of all points P in the plane such that ∣PFj - PF2∣ = A: is called a hyperbola, an
example of which is shown at right. As you can see, unlike the other conic sections
we have studied, a hyperbola consists of two pieces, which are called branches of
the hyperbola.
As with an ellipse, the points Fι and F2 are the foci of the hyperbola. The line
connecting the foci meets the hyperbola at the vertices of the hyperbola (just as with a parabola, the
vertex of each branch of a hyperbola is an "extreme" point of a branch of the hyperbola). The midpoint
of the segment connecting a hyperbola's vertices is the hyperbola's center.

142
 5.6. HYPERBOLAS

I
Problems

Problem 5.28:
(a) Let 7/ be a hyperbola with foci F1 = {c, 0) and F2 = (-G θ) and vertices V∖ = {a, 0) and Vz2 = (-fl, 0),
where a and c are positive constants with c > a. Show that if P is on the hyperbola, then
∣PFι - PF2∣ = 2a. Suppose that P is closer to F2 than to F↑. Then, how can we write the equation
∣PF] - PF2∣ = 2a without the absolute value signs?
(b) Let P be (x, y). Show that if P is on 7√, then 4 - = 1, where b2 = c2- a2. Let 77' be a hyperbola
with foci (h ± c, k) and vertices (∕ι ± a, k), where a, c, h, and k are positive constants and c > a. Find
an equation whose graph is 77'.

Problem 5.29:
(a) Graph the equation ⅜∙ - = 1 by finding several points on the graph.
(b) Graph the equations ∣ = ∣ and ∣ = -∣ on the same Cartesian plane as your graph in part (a).
(c) Is it possible for your graphs in parts (a) and (b) to meet?
(d) What happens to these graphs as x gets very large? Why?
(e) How can you use the graphs of ~ = and 2⅛~ = -⅛~ to help you graph the hyperbola
(x-⅛)2 (y-∕c)2 _ o
a2 b2 ~ 1∙

Problem 5.30:
9 2
(a) Graph the hyperbola = 1 for various values of a and b. How do the values of a and b
affect the graph?
(b) What happens if the y term is positive and the x term is negative? For example, what is the
graph of ⅜ - ⅛ = 1?

Problem 5.31: Find the center and vertices of the graph of each equation given below, and then graph
the equation.
(x - 2)2
16
(b) 4y2 - x2 + 16y + 6x - —23

Problem 5.32: Back in Problem 5.26, we looked at equations that produce "degenerate" ellipses. These
equations almost exactly matched the standard form of an equation whose graph is an ellipse, except
that the isolated constant term was 0 or negative rather than being 1. In this problem, we investigate
corresponding equations that look like they have hyperbolas as their graphs. Find the graphs of each
of the following:
(a) ⅞⅛=0

143
CHAPTER 5. CONICS

Problem 5.28:
(a) Let Wbea hyperbola with foci F↑ = (c,0) and F2 = (-c,0) and vertices V1 = (α,0) and
V2 = (-a, 0), where a and c are positive constants with c> a. Find an equation whose graph is
7√.
(b) Let fH' be a hyperbola with foci (∕ι ± c, ∕c) and vertices (⅛ ± a, k), where a, c, h, and k are positive
constants. Find an equation whose graph is 777.
~hmmmhmmm

Solution for Problem 5.28:

(a) We start with the definition of a hyperbola. Suppose the point P — (x, y) is on the hyperbola. Then,
the difference ∣PFι — PF2∣ is constant for all possible P. But what is this constant?
We find this constant difference the same way we found the constant sum for an ellipse.
Concept: When faced with a new problem that's very similar to a problem you know
Q —-∣J how to do, use your solution to the known problem as a guide for the new
problem.
We consider a specific point on the hyperbola, namely, vertex V↑. Since VjFι = c—a and V-[F2 = c+a,
we have ∣V1F1 - V1F2∣ = ∣(c - a) - (c + α)∣ = ∣-2a∣ = 2a. Because V↑ is on the hyperbola and
∣Vr1F1 - V1F2∣ = 2α, we know that the constant difference ∣PFι - PF2∖ is 2a.
To get rid of the absolute value, we'll first investigate points on the hyperbola that are closer to
F2 than to F↑. If P is closer to F2 than to Fj, then we have PFχ > PF2. This means PFχ - PF2 > 0, so
∣PF1 - PF2∖ = PF↑ - PF2. Therefore, the equation ∣PFι - PF2∣ = 2a is PF2 - PF2 = 2a. We can use the
distance formula to find PF-∣l and PF2, so we now have an equation for our hyperbola:

This is very close to the initial equation we had on page 132 for an ellipse, which was

^(x - c)2 + y2 + ^(x + c)2 + y2 = 2n.

The only difference between these two equations is the sign before the second radical. We have a
situation very much like our ellipse equation, so we use the same strategy we used to simplify the
ellipse equation. We first isolate one radical,
^(x-c)2 + y2 = 2a + ^(x + c)2 + y2,

then square both sides, to get

(x - c)2 + y2 = (2α)2 + 2(2«) (-^(x + c)2 + y2j + (x + c)2 + y2.

Rearranging, we have

4« y∣(x + c)2 + y2 = (x - c)2 + y2 - (x + c)2 - y2 - (2α)2

= x2 - 2xc + c2 + y2 - x2 - 2xc -c2 - y1 - 4«2

= -4xc - 4a2.

144
 5.6. HYPERBOLAS

Dividing both sides by 4 and squaring again will get rid of the radical and give us

α2((x + c)2 + y2) = x2c2 + 2xcα2 + α4.

Expanding the left side gives us α2x2 + 2xca2 + a2c2 + a2y2 = x2c2 + 2xca2 + a4. Bringing the x's and
y's to one side and moving the constants to the other gives

x2(c2 - α2) - fl2y2 = α2(c2 - a2).

Dividing both sides by α2(c2 - a2) gives us

x2 w2
Finally, we let b2 = c2 - a2, and we have — - ⅛ = 1.
«2 b2
We've now shown that if P = (χ, y) is on the hyperbola such that PFι > PF2∕ then (x, y) satisfies
the equation = 1. If P is closer to F↑ than to p2, we have PPι < Pp2, so PF↑ - PF½ < 0. In this
case, we have ∖PF↑ - PF2∣ = -(PPι - PF2), so the equation ∣PFι - PF2∣ = 2a becomes PF↑ - PF2 = —2a.
We can then work through essentially the same steps as above to find that (x, y) satisfies —12 = 1
in this case, as well.
(b) The center of the hyperbola in part (a) is (0,0) and the center of the hyperbola in this part is (Ji, k).
Otherwise, the two hyperbolas are the same. That is, hyperbola eH' in this part is the result of
translating hyperbola 7F rightward h units and upward k units. So, for each point (x, y) on 7Y,
there is a point (x', y') on fFF' such that (x', y') = (x + h,y + k). Therefore, we have x = x' -h and
y = y' -k. Substituting these into our equation in part (a), we have

(χ' - ∕z)2 (y' - k)2

a2 b2
where b2 = c2 - a2. We can reverse all these steps to show that:
Important: The graph of the equation

(x - h)2 ⅛ ~ fc)2
b2
is a hyperbola with center (∕z, k~), vertices (⅛ ± a, k), and foci (h ± c, k), where
c2 = a2 + b2. This is the standard form of an equation whose graph is a
hyperbola that opens horizontally.

You shouldn't need to memorize that the vertices of the graph of

(x - ∕z)2 (y - k)2
a2 b2
are (h ± α,fc). The vertices are the points closest to the center of the hyperbola. These points occur when
the negative term of the left side of the equation is 0. When this happens, we have = 1, so x-h = ±a,

145
CHAPTER 5. CONICS

which tells us that the two resulting points on the graph are a to the left and right from the center. If
the negative term of our equation is nonzero, then y ≠ k, and the fraction fl2 is greater than 1, which
means that x is more than a from h. Therefore, (x, y) is both horizontally and vertically farther from the
center than the points we found by letting y = k, so the points that result from letting the negative term
equal 0 are the vertices of the hyperbola.
-------------------------------------- --------- ---------------------------------------------------------------------------------------------------------------------------- —— -------------------——-

Problem 5.29:
(a) Graph the equation y - = 1 by finding several points on the graph.
(b) Graph the equations 5 = 5 and 5 = -5∙
(c) Is it possible for your graphs in parts (a) and (b) to meet?
(d) What happens to these graphs as x gets very large? Why?
(e) How can you use the graphs of 2yi = and 2yi = -⅞^ to help you graph the hyperbola
(χ-h)2 (y-k)2 _ 1 o
a2 b2 ~ i∙

Solution for Problem 5.29:

(a) Below is our table and our graph. Note that because the equation only has x2 and y2 and no other
variable expressions, if the point (x, y) is on the graph, then so are (x, -y), (-x, y), and (-x, -y). So,
aside from first row (the vertices of the hyperbola), each row in the table gives us four points on
the hyperbola. (Several of the y-coordinates are approximations.)

y
0
+1.76
±2.67
+3.46
±4.22
±4.94
±5.66

(b) We add the graphs of the lines to the graph of our hyperbola, and
the result is shown at right.
The lines appear to "frame" the hyperbola. It looks like the
hyperbola gets closer and closer to the lines, but doesn't intersect
them.
(c) If the hyperbola intersects the graph of ∣ = ∣, then it must be at a
point (x, y) that satisfies the system of equations

χ2 _ y2 _
9 4 ~ '
x = If
3 2’

146
 5.6. HYPERBOLAS

But squaring the second equation gives us y = ¾-, and rearranging this gives y - y = 0. This
equation and the first equation in our system above, ⅞ ~ ⅜ = 1/ cannot both be true, so the
hyperbola cannot intersect the graph of ∣ = ∣. Similarly, it cannot intersect the graph of ∣ = -∣.

(d) We investigate what happens as x gets larger by finding points X Hyperbola's y Line's y
on the upper portion of the right branch and on the graph of 6 3.46 4
3 = 2 for larger and larger values of x. At right is a table 10 6.36 6.67
showing the y-coordinates (rounded to the nearest hundredth) 20 13.18 13.33
of points on the hyperbola and the line for various values of 50 33.27 33.33
x. We see that the y-coordinates get closer and closer to each 100 66.64 66.67
other.
Let's take a look at why the y-coordinates get closer and closer to each other. We'll focus on
why the line ∣ = ∣ approaches the upper half of the right branch of the hyperbola when x is large.
Solving for y in ∣ = ∣ gives y = Solving for y2 in y - y = 1 gives y2 = 4(y - 1). We take the
positive square root because we are interested in the upper half of the right branch, and we have
y=2√⅛÷L

Intuitively, we can see that as x gets very large, the value of y - 1 is (relatively speaking) very
close to y, which makes y = 2 — 1 very close to y = 2 λJ⅜ = y. We can be a little more precise
with our explanation by noting that if x > 0, we have

2x
y=2 =2
~3

As x gets large, the value of J gets closer and closer to 0, so 1 - J gets closer and closer to 1. This
means that y gets closer and closer to 2x∕3, as we saw in our table.
We call a line an asymptote of a graph if the graph gets closer and closer to, butdoes not
intersect, the line as x and/or y gets
2
far
y2
from 0. As we have just seen, the lines I5 =2 5 and3 ½= -%■2
are asymptotes of the hyperbola y - = 1.
Similarly, we can show that:
χ-h y-k
Important: The lines —— = ±—— are asymptotes of the graph of

Ψ (x - h)2 (y ^ fc)∖ 1
fl2 b2

(e) Suppose we wish to graph the equation = 1. We know that the graph is a hyperbola
with center (h, k) and vertices (∕z ± a, k), but this just gives us two points on the hyperbola. To get
a quick sketch of the hyperbola, we can draw the asymptotes of the hyperbola first, then use the
asymptotes as a guide to fill out the hyperbola.
The asymptotes of the hyperbola are yp = . Multiplying both sides by b gives the equation
y~k = ±⅞(x-∕z). Therefore, the asymptotes are the lines through (∕ι,k) with slopes b/a and -b∕a.

147
CHAPTER 5. CONICS

One quick way to graph these asymptotes is to make a rectangle centered at (h,k) with vertical
sides of length 2b and horizontal sides of length 2a. The midpoints of the vertical sides then
are the vertices of the hyperbola, and the asymptotes of the hyperbola contain the diagonals of
the rectangle. The result is shown below, with the sides of the rectangle dashed and asymptotes
drawn with thin lines.

□
If we add the graph of ⅛r + ¾r- = 1 to the graph we just made of = 1 (including the
rectangle we made to plot the asymptotes), we have a visual sense for how these equations are related:

Problem 5.30: 2
(a) Graph the hyperbola - ∣ς = 1 for various values of a and b. How do the values of a and b
affect the graph?
(b) What happens if the y term is positive and the x term negative? For example, what is the graph
of⅛-⅞ = 17

148
 5.6. HYPERBOLAS

Solution for Problem 5.30:

(a) In our first three hyperbolas below, we keep b fixed and vary a. In the second three hyperbolas,
we keep ∩ fixed and vary b. We use the asymptotes of each hyperbola as described in the last part
of the previous problem to quickly generate the graphs.

First, we notice that increasing a moves the branches farther apart. This makes sense because
the vertices of a hyperbola are 2a apart. We also notice that decreasing a or increasing b makes
the hyperbola "open more widely." Examining the asymptotes explains why. Decreasing a or
increasing b makes the slopes of the asymptotes, and farther from 0. In other words, the
asymptotes are "steeper," thereby making the hyperbola open wider. So, as £ increases, the
hyperbola opens wider, and as £ decreases, the hyperbola opens more narrowly.

Sidenote: Just as with ellipses, we define the eccentricity of a hyperbola as the ratio of
the distance between the foci to the distance between the vertices. Because
c2 = α2 + b2, we have

c y[a2 + b2
a a

So, the "wider" a hyperbola is (meaning J is large), the more eccentric it is.

149
CHAPTER 5. CONICS

(b) We've seen that having y squared instead of x squared in a parabola y

means the parabola opens horizontally rather than vertically. Simi­
larly, when the y term has the larger denominator in the standard form
equation of an ellipse, then the foci are on a vertical line, rather than
on a horizontal line. So, we're not surprised to find that the graph of
y - iy = 1 is a hyperbola that opens up and down rather than left and
right. The graph is shown at right.
We can also find the graph of y - j∙ = 1 by noting that for each point
(a, b) on this graph, there is a point (b, α) on the graph of y - 4- = 1, and
vice versa. We've seen something like this before! Back on page 48, we noted that thepoints (a, b)
and (b, a) are reflections of each other about the line y = x. Therefore, the graph of ⅜- - ¾∙ = 1 is
the image of the graph of y — = 1 when reflected over the line y — X∙

Important: The standard form of an equation whose graph is a hyperbola that opens
vertically is
(y - fc)2 (* - ^)2
α2 b2
The center of the hyperbola is (h, k~), its vertices are (h, k ± a), and its foci
are (h,k ± c), where c2 = a2 + b2. The asymptotes of the hyperbola are
y-fc = ±f(χ-^)∙

Important: Don't worry too much about memorizing all these various forms and
formulas. It's more important to understand how we derive them, and
how we apply them once we have found them. You can usually look
them up when you need them.

Problem 5.31: Find the center and vertices of the graph of each equation given below, then graph
the equation.
(* - 2)2
= 1 (b) 4y2 - x2 + 16y + 6x = -23
16

Solution for Problem 5.31:

(a) What's wrong with this start:

Bogus Solution: We have h = 2,k = -1, a = ½, and b = 3. So, the center is (2, -1). The
y term is positive, so the branches of the hyperbola open vertically.
So, the vertices are (2, -1 ± 4), which are (2,3) and (2, -5).
The problem with this start is that the standard form of a hyperbola that opens up and down is
(y^)— (x√τr _ 1 gθ, we have _ 3, not a _ 4 yye couu have caught our mistake by checking
our work:

150
 5.6. HYPERBOLAS

Important: You can easily check if you have the vertices of a hyperbola correct. They

φ must satisfy the equation of the hyperbola.

This shows yet another reason you shouldn't simply memorize y

formulas without thinking about them. If you don't have a formula
quite right, and don't understand it, you won't notice you've made a
mistake.
Returning to the problem, we know the hyperbola opens upward
and downward, and the center is (2, -1). Rather than relying on our
formulas, we note that the negative term is 0 for x = 2, which gives
us (y + 1)2∕9 = 1. From this, we find that our vertices are 3 above and
below the center, at (2,2) and (2z -4).
We draw our asymptotes, y + 1 = ±∣(x - 2), and then we use these and our vertices to graph
the hyperbola, as shown at right above.

(b) We first complete the square in x and y. Factoring 4 out of the y

terms and -1 out of the x terms, we have -l(x2 - 6x) + 4(y2 + 4y) =
—23. We complete the square by adding (—1)(9) for x and 4(4) for
y, and we have

-l(x2 - 6x + 9) + 4(y2 + 4y + 4) = -23 + (-l)(9) + (4)(4) = -16.

Finally, we divide by —16 (make sure you see why we must divide
by -16, not just 16) to give
(x - 3)2 (y + 2)2
16 4
The center of the hyperbola is (3, -2). The negative term is 0 for y = -2, from which we find that
the vertices are (7, -2) and (-1, -2).
As before, we graph the asymptotes, y + 2 = ±∣(x - 3), and use these along with the vertices to
graph the hyperbola, shown above.

Solution for Problem 5.32:

(a) From the left side, it looks like the graph of the equation is a hyperbola. However, y
the right side is 0, not 1. We can add to both sides to get γ = —~2^ . Taking
the square root of both sides then gives us ∣ = ±^. These two possibilities give
us the equations 3x = 2(y - 2) and 3x = -2(y - 2). The graphs of these are the two
lines shown at right.

151
CHAPTER 5. CONICS

The original equation is very close to an equation that we know has a hyperbola
as its graph. So, we look for how this equation is related to that hyperbola. In our
graph at right, we have graphed with dashed lines the hyperbola that is the graph
of — — 92^ = 1. From the graph, we quickly see the relationship—the lines that
form the graph of — ^y 92^ = 0 are the asymptotes of the graph of ¾∙ — ^y 9 > = 1∙

The graph of an equation such as - — 92- = 0 is considered a degenerate

hyperbola.
(b) Our first instinct might be to guess that the graph of this equation is not y
a hyperbola, just like the graph of the equation ¾∙ + l^y 9~ = —1 is not an
ellipse. However, multiplying the equation given in this part by —1 produces
an equation that is in the standard form of a hyperbola:

(y ~ 2)2 _ √ =
9 4
Graphing this equation produces the hyperbola at right. Algebraically speak­
ing, while there are no real solutions to the equation j∙ + (y 9-) = -1, there are
solutions to the equation ψ - <y 92∙ - = -1.

We now summarize what we've learned about the possible graphs that can be produced from an
equation which is quadratic in x and/or y:

Important: Graphs of equations of the form Ax2 + By2 + Cx + Dy + E = 0, where A,

ψ B, C, D, and E are constants such that not both A and B are zero, can be
grouped into three categories:
• One of A or B is 0. The graph is a parabola.

• AB > 0. In other words, A and B have the same sign. The graph is
an ellipse (a circle if A = B), a point, or empty.

• AB < 0. The graph is either a hyperbola or a pair of lines.

These are not the only equations that produce parabolas, ellipses, or hyperbolas as their graphs.

Problems

Problem 5.33: In this problem, we examine the graph of the equation xy = 1.

(a) Graph the equation. Does it look like a hyperbola?
(b) What lines appear to be the asymptotes of the graph in part (a)?
(c) Assuming the graph is a hyperbola, what are its vertices?

Problem 5.34: Find the graph of the equation (x - 3)(y + 1) = -4.

152
 5.6. HYPERBOLAS

Problem 5.33: In this problem, we examine the graph of the equation xy = 1.

(a) Graph the equation. Does it look like a hyperbola?
(b) What lines appear to be the asymptotes of the graph in part (a)?
(c) Assuming the graph is a hyperbola, what are its vertices?

Solution for Problem 5.33:

(a) We solve the equation for y to find y = ∣. Choosing various values of

x, we generate the graph at right.
The graph sure looks like a hyperbola, but this hyperbola doesn't
open horizontally or vertically. Instead, it appears to be the result of
rotating a horizontally-opening hyperbola 45 degrees counterclock­
wise. In a later text in the Art of Problem Solving series we'll study
trigonometry, which we can use to prove that this graph is the result
of rotating the graph of y — y = 1 forty-five degrees counterclockwise
about the origin.
Important: If fc is a nonzero constant, then the graph of the equation xy = k is a

ψ hyperbola.

(b) The graph appears to approach, but not intersect, either coordinate axis. Clearly, we can't have
x = 0ory = 0in the equation xy = 1, since this will give us 0 = 1, which can never be true. So, the
graph of xy = 1 does not intersect either axis. However, because we have y = 1, as x gets large, y
gets very close to 0. The same happens if x is much less than 0. Therefore, the x-axis, which is the
graph of y = 0, is an asymptote of the graph.
Similarly, when y is very large, x is close to 0. The graph of xy - 1 never intersects the y-axis,
but approaches the y-axis as y gets large. So, the y-axis is also an asymptote of the graph.
Note that the asymptotes of the graph are perpendicular. We call a hyperbola with perpendic­
ular asymptotes a rectangular hyperbola.
(c) For each point (a, ∣) on the graph of xy - 1, we see that the point (∣z α) is also on the graph. In other
words, the graph xy = 1 is symmetric about the line x = y. The line that divides both branches
into two symmetric pieces must pass through the vertices of the hyperbola, so the vertices of the
hyperbola occur where x = y. Setting x = y in xy = 1 gives us y2 = 1, so y = ±1. Therefore, the
vertices are (1,1) and (-1, -1).

Problem 5.34: Find the graph of the equation (x - 3)(y + 1) = -4.

Extra! For any set of five points in a plane such that no three of the points lie on the same line,
ιι* ιι*-" ιιι∣*∙ there is a single conic section that passes through all five points.

153
CHAPTER 5. CONICS

Solution for Problem 5.34: We suspect that the graph is a hyperbola, both because
the equation is similar to the equation in the previous problem, and because the
problem is in the "Hyperbola" section. To graph the equation, we solve for y in
terms of x. We know that we can't have x = 3, since this would give us 0 = -4.
So, we can divide both sides by x - 3, and then subtract 1, to find y = - 1.
Choosing several values of x and computing the corresponding y gives us the
graph at right, which sure looks like a hyperbola.
Let's take a look at how we can reason our way to the graph from the equation
(x — 3)(y + 1) = —4 once we know the graph is a hyperbola. First, we could have seen that the lines x = 3
and y = —1 are asymptotes. If x = 3 or y = -1 in (x — 3)(y + 1) - -4, we will have 0 = -4, which is
impossible. Therefore, the graph doesn't intersect either x = 3 or y = -1. However, from y = -j75 - L
we see that if x is just above 3, but very close to 3, then y is the negative of a large number. So, x = 3 is
an asymptote of the graph. Since is very close to 0 when x is far from 0, we see that y = - j∑5 - 1
approaches, but never equals, -1 when x is far from 0. So, the line y = -1 is an asymptote as well. The
asymptotes meet at (3, -1), so this point is the center of the hyperbola.

The asymptotes divide the Cartesian plane into four quarters. The branches
of the hyperbola are in the upper left and lower right quarters, which we also
could have reasoned from the equation (x - 3)(y + 1) = -4. If x > 3 and y > —1,
then (x - 3)(y +1) is positive, so no points with x > 3 and y > -1 are on the graph
of (x - 3)(y +1) = -4. Similarly, we cannot have x < 3 and y < -1. Therefore, the
hyperbola is in the upper left and lower right quarters formed by the asymptotes.
The vertices of the hyperbola are the intersections of the hyperbola and
the line through (3, -1) with slope -1. This line has equation y + 1 = -(x - 3).
Substituting this into (x-3)(y+l) = -4givesus-(x-3)2 = -4, from which we have x-3 = ±2. Therefore,
x equals 5 or 1, and our vertices are (5, -3) and (1,1). We have our vertices and our asymptotes, and
now we can use these as a guide to graph the hyperbola.
The resemblance of the equation (x - 3)(y +1) = -4 and its graph to the equation xy = 1 and its graph
suggests we might have used our function transformation tools from Section 2.2 to find the graph of
(x - 3)(y + 1) = -4 starting with the graph of xy = 1. To see how, we solve both equations for y, so that
we have y = ∣ and y = -⅛ - 1. Letting ∕(x) = J and g(x) = -1⅛ - 1, we seek to relate ∕(x) and g(x).
We do so in a series of transformations, as shown on the following page.

Extra! When a ray reflects off a curve, it behaves as if it is being

l∣∣∣*- ∣ ∣ Φ∣ u* ∣ ∣ φ. reflected by the line tangent to the curve at the point where ∖ ∕∕
the ray meets the curve. In the diagram, the incoming ray ∖, Focus
is parallel to the axis of symmetry of the parabola, and the χχ∙
reflection of the ray passes through the focus. This is not xs* - --^7
a coincidence! As shown in the various dotted rays, the z∕
reflection of every ray parallel to the axis of symmetry passes through the focus. This
is why satellite dishes are parabolic. When we point our satellite dish so that its axis
is pointed directly at the satellite, the rays coming from the satellite are all essentially
parallel to the axis of the dish, so if we place the receiver for the satellite signals at the
focus of the dish, all the rays will be focused on the receiver.

154
 5.6. HYPERBOLAS

Transformation Effect on Graph

-∕ω = 4 Reflect graph of ∕(x) over the ι∕-axis
-4∕(-V) = -∣ Scale the graph of -∕(x) vertically by 4
-4∕(x-3) = -⅛ Shift the graph of -4∕(x) right 3 units
-4∕(x-3)-l = -⅛-l Shift the graph of -4∕(x - 3) down 1 unit
Our final expression for -4∕(x - 3) - 1 matches g(x), so our final graph is the graph of g. The series of
transformations is shown below:

Similarly, we can show that if t is a nonzero constant, then the graph of (x — h)(y — k) — f is a hyperbola
with center (∕t, ∕c).

We have not yet seen all the forms of equations whose graphs are parabolas, ellipses, or hyperbolas.
In a later text in the Art of Problem Solving series, we'll consider what happens when we include an xy
term along with x2 and y2 terms.

HI Exercises

5.6.1 A hyperbola centered at the origin has one focus at (-13,0) and one vertex at (5,0). Find an
equation whose graph is this hyperbola.

5.6.2 For each of the graphs below, find an equation whose graph is shown. Each graph is a hyperbola.

Note: The graph in (a) passes through (6,7) and the graph in (b) passes through (0,6).

155
CHAPTER 5. CONICS
τ

5.6.3 Find the center, asymptotes, and vertices of the graph of each equation below, and then graph
each equation.
(a) ^-⅛-y2 = l (b) x2 + 12x-3y2 + 18y+ 21 = 0
9

5.6.4 If a and b are positive constants such that the graph of - -∙jj2- = 1 is a rectangular hyperbola,
then what is a/b?
5.6.5 Hyperbola 7^∕ is the graph of the equation = 1. One of the points of intersection of 7Y
and the line y = ax has x-coordinate 0. Find the x-coordinates of the points of intersection of the line
y = bx and 7/.
5.6.6
* Find the foci of the graph of xy = 1. Hints: 315

5.7 Summary

In this chapter, we saw how to find equations whose graphs fit geometric definitions of certain curves.
The process we used to find these equations and understand them is far more important than the
equations themselves, or the many related facts we discovered. Many areas of study require building
mathematical models. The process for doing so is often very similar to what we did in this chapter:
build equations that depend on certain constants (like our h, k, and a for parabolas, for example), then
try different values of those constants to see how they affect your model.
A locus is a set of points that satisfy certain given conditions. We covered four special loci in this
chapter:

• Parabola: The locus of points in a plane equidistant from a given line (the directrix) and a point
(the focus) not on the line. The line through the focus that is perpendicular to the directrix is the
axis of symmetry, which divides the parabola into two pieces that are mirror images of each other.
The axis intersects the parabola at the vertex of the parabola.
The standard form of a vertically-opening parabola is y = α(x - h)2 + k, where a ≠ 0. The graph
of this equation has vertex (h,k) and axis x = h.
The standard form of a horizontally-opening parabola is x = α(y - k)2 + h, where a ≠ 0. The
graph of this equation has vertex (h,k) and axis y = k.

• Circle: The locus of points in a plane a specified distance (the radius) from a given point (the
center).
The standard form of a circle is (x - h)2 + (y - k)2 = r2f where r > 0. The graph of this equation
has center (∕ι, k) and radius r.
• Ellipse: The locus of points P in a plane such that PFχ + PF2 is a given constant, where Pi and F2
are two points in the plane. The points F↑ and F2 are the foci (plural of focus) of the ellipse, and
the midpoint of F↑F2 is the center of the ellipse. The chord (a line segment with both endpoints
on the ellipse) that passes through both foci is the major axis of the ellipse, and the chord through
the center that is perpendicular to the major axis is the minor axis.

156
 5.7. SUMMARY

The standard form for an ellipse is c*~⅛- ■ ÷ = 1. The center is (h,k). Tine axes have lengths
2a and 2b. If a > b, then the major axis is horizontal; if a < b, then the major axis is vertical. (If
a = br then the graph is a circle.) The distance between the foci can be found with the relationship:
(major axis length)2 = (minor axis length)2 + (distance between foci)2.

• Hyperbola: The locus of points P in a plane such that ∣PjFi - PF2∣ is a given constant, where Fj and
F2 are two points in the plane called the foci of the hyperbola. The line connecting the foci meets
the hyperbola at two points called the vertices of the hyperbola.

The standard form for a horizontally-opening hyperbola is = 1. The center is (h, k),
the vertices are (h ± a,k), and the foci are (∕z ± c,k), where c2 = a2 + b2. The asymptotes (lines the
graph approaches, but never intersects) of the hyperbola are = ±⅛-∙

The standard form for a vertically-opening hyperbola is = 1. The center is (h, k),
the vertices are (h,k ± «), and the foci are (h,k ± c), where c2 = a2 + b2. The asymptotes of the
hyperbola are = ±2y1.

Collectively, these curves are known as conic sections.

Important: Don't waste time memorizing all the formulas we discussed in this chap­
ter. The formulas themselves are not terribly important—you can usu­
ally look them up when you need them. It's far more important to
understand how we derived the formulas and how to use them.

Important: Graphs of equations of the form Ax2 + By2 + Cx + Dy + E = 0, where A,

ψ B, C, D, and E are constants such that not both A and B are zero, can be
grouped into three categories:

• One of A or B is 0. The graph is a parabola.

• AB > 0. In other words, A and B have the same sign. The graph is
an ellipse (a circle if A = B), a point, or empty.

• AB < 0. The graph is either a hyperbola or a pair of lines.

We used our understanding of parabolas and completing the square to learn how to find the maxi­
mum or minimum of a quadratic expression.

Important: Let ∕(x) = ax2 + bx + c, where a, b, and c are constants.

ψ • If a > 0, then the minimum value of ∕(x) is c - ξ, which occurs

when x = -½.

• If a < 0, then the maximum value of ∕(x) is c - ⅛, which occurs

when x = —

157
 I
CHAPTER 5. CONICS

Problem Solving Strategies___________ _________________________

Concepts: ∙ Don't depend too heavily on your formulas. A little thinking about
Q~'lj∣J a problem can go a long way.

• Symmetry can be a very powerful problem-solving tool.

• Don't get locked into one way of thinking about quadratics and
parabolas.
• To find the minimum or maximum value of a quadratic, complete
the square in the quadratic.

• If a problem is difficult to understand, take extra time reading and

re-reading the problem until you are sure you understand what the
problem is. You don't want to spend a lot of time working on the
wrong problem!

• Be alert for equations you know how to graph easily. Sometimes

geometry can be used to solve algebra problems.

• Keep your eye on the ball! Don't forget what you're looking for in
a problem—you don't always have to solve for every variable in a
problem to solve the problem itself.

• When faced with a new problem that's very similar to a problem you
know how to do, use your solution to the known problem as a guide
for the new problem.

Review Problems

5.35 For each of the parts below, describe the graph of the equation with one of the following terms:
parabola, circle, ellipse, hyperbola, point, line, two lines, empty.

(a) x2 — —y + 7x — 3 (d) x2 + 4y2 + 2x = 0 (g) x2 + 2x = y2 + 4y + 3

(χ-3)2 + y2-9 x= y1
(b) 16 + 4 ^9 (e) 6x2 + 12x + 6y2 = 18y (h)
x+1
4
(c) x2 + y2 + 18y = -81 (f) x= 0 (i) x2 - 4y2 + 8y = 0
v-2

5.36 Find the vertex and axis of symmetry of the graph of y = 4x2 - 12x + 8.

5.37 Find the vertex and axis of symmetry of the graph of x = -(y - 4)2 - 1. Graph the equation.

158
 REVIEW PROBLEMS

5.38 The vertex of the parabola x = 4y2 + 6y + c lies on the y-axis. Find the value of c.
5.39 Find an equation whose graph is a parabola with vertex (1,5) and focus (1,3).
5.40 For each of the following, find an equation whose graph is shown in the diagram.

The hyperbola in part (d) passes through (-2,1), (-2, -3), and (4,3).
5.41 For how many values of a is it true that the line y = x + a passes through the vertex of the parabola
y = x2 + a2? (Source: AMC)
5.42
(a) Find the minimum value of p(y) = 2y2 - 4y + 19.
(b) Find the maximum value of 37 - 16r - r2.
(c) Find the minimum value of the function f(x) = x4 — 3x2 — 13.
(d) Find the minimum value of x2 + 4y2 - 6x + 4y + 5.

5.43 GetThere Airlines currently charges $200 per ticket, and sells 40,000 tickets. For every $10 they
increase the ticket price, they sell 1000 fewer tickets. How much should they charge to maximize their
revenue?

5.44 Find the smallest possible value of the length of a diagonal of a rectangle with perimeter 36. (Note:
You cannot simply assume the rectangle is a square; you must show that the shortest possible diagonal
occurs when the rectangle is a square.)
5.45 Find the area enclosed in the graph of x2 + y2 = 16x + 32y.

5.46 Find r if r is positive and the line whose equation is x + y = r is tangent to the circle whose equation
is x2 + y2 = r. (Source: AHSME)

159
CHAPTERS. CONICS

5.47 The circumcircle of a triangle is the circle that passes through all three vertices of the triangle.
Find an equation whose graph is the circumcircle of a triangle with vertices (—2,5), (-4z -3), and (0, -3).
5.48 Find an equation whose graph is an ellipse with axes parallel to the coordinate axes such that the
endpoints of the axes of the ellipse are (0,2), (3,6), (3, -2), and (6,2).
5.49 Graph each of the equations below, then find the foci, center, and lengths of the axes of each graph.

(a) (X~3)2 + y*2 = 1 (b) 4x2 + y2 + 16x-6y = 11

9
5.50 Ellipse & has a horizontal major axis, one focus at (4,1), and one axis with an endpoint at (6,5).
Find an equation whose graph is this ellipse.
5.51 Find the center, asymptotes, and vertices of the graphs of each equation below, then graph each
equation.
(a) x2 - 4(y + 1)2 = 1 (b) 2x2 - y2 + 8x + 6y = 9

5.52 Al begins on a vertex of the hyperbola - = 1. He walks in a straight line to the closest
focus of this hyperbola, and then he moves to the farthest vertex. Find an expression, in terms of a, b, k,
and h, for the total distance Al walked.
5.53 One of the graphs of the following equations comes within 1 emit of the point (265,346). Which
one?
(a) y2 + x + 4y + 7 - 0 (c) x2 + y2 + 4y - 17 = 0
(b) 16x2 - 96x - 9y2 - 54y = 81 (d) 25x2 + 9y2 - 36y = 9

Challenge Problems

5.54 Points A and B are on the parabola y = 4x2 + 7x - 1, and the origin of the Cartesian plane is the
midpoint of AB. What is the length of AB? (Source: AMC)
5.55 A latus rectum of an ellipse is a line segment with both endpoints on the ellipse that is parallel to
the minor axis and passes through a focus of the ellipse. Find the length of a latus rectum of the graph
2 ι∕2
of ⅛ + ⅛ = 1 in terms of a and b.

5.56 A parabola with equation y = ax2 + bx + c is reflected about the x-axis. The parabola and its
reflection are translated horizontally five units in opposite directions to become the graphs of y = ∕(x)
and y = g(x), respectively. Which of the following describes the graph of y = (f + g)(x)?
(A) a parabola tangent to the x-axis (D) a non-horizontal line
(B) a parabola not tangent to the x-axis (E) the graph of a cubic function
(C) a horizontal line

(Source: AMC) Hints: 119

160
 CHALLENGE PROBLEMS

5.57 If x is real, then compute the maximum value of —----- -------- . (Source: ARML)

5.58 Let ∕(x) = x2 + 6x + 1, and let R denote the set of points (x,y) in the coordinate plane such that
∕(x) + ∕(ι∕j ≤ 0 and ∕(x) - ∕(y) ≤ 0. Find the area of R. (Source: AMC) Hints: 75

5.59 A chord of a parabola is a segment with both endpoints on the parabola. Two parabolas have the
same focus, namely the point (3, -28). Their directrixes are the x-axis and y-axis, respectively. Compute
the slope of their common chord. (Source: NYSML) Hints: 209

5.60 Find the largest value of x for which x2 + y2 = x + y has a solution, if x and y are real. (Source:
ARML)

5.61 P is a fixed point on the diameter AB of a circle. Prove that for any chord CD of the circle that is
parallel to AB, we have PC2 + PD2 = PA2 + PB2. Hints: 58

5.62 Let (2,3) be the vertex of a parabola whose equation is of the form y = ax2 + bx + c. If the graph
of the parabola has two x-intercepts with positive x-coordinates, compute the greatest possible integral
value of k such that « = ⅛. (Source: ARML) Hints: 202

5.63 An equilateral triangle is inscribed in the ellipse whose equation is x2 + 4y2 = 4. One vertex of the
triangle is (0,1), and one altitude is contained in the y-axis. Find the length of each side of the triangle.
(Source: AIME)
5.64 Given that x2 + y2 = 14x + 6y + 6, what is the largest possible value that 3x + 4y can have? (Source:
AHSME)

5.65 Three circles with radius 2 are drawn in a plane such that each circle is tangent to the other two. Let
the centers of the circles be A, B, and C. Point X is on the circle with center C such that AX+XB = AC+CB.
Find the area of triangle AXB. (Source: USAMTS) Hints: 66
5.66★ Let ∕(x) = (x + 3)2 + ∣ for x ≥ -3. Compute the shortest possible distance between a point on the
graph of f and a point on the graph of f~1. (Source: ARML) Hints: 39, 36

5.67 Determine the unique pair of real numbers (x, y) that satisfy the equation
(4x2 + 6x + 4)(4y2 - 12y + 25) = 28.

(Source: USAMTS)
5.68
* Find an equation whose graph is a parabola with vertex (1,1) and focus (2,2).
*
5.69 Consider the two functions
f(x) = x2 + 2bx + 1 and g(x) = 2a(x + b^),

where the constants a and b may be considered as a point (a, b) in an αfe-ρlane. Let S be the set of such
points (a, b) for which the graphs of y = ∕(x) and y = g(x) do NOT intersect (in the xy-plane). What is
the area of S? (Source: AHSME) Hints: 357
5.70
* The graph of 2x2 + xy + 3y2 - llx — 20y + 40 = 0 is an ellipse in the first quadrant of the xy-plane.
(This means that all points on the ellipse have nonnegative coordinates.) Let a and b be the maximum
and minimum values of over all points (x, y) on the ellipse. What is the value of a + b? (Source: AMC)
Hints: 270

161
CHAPTER 6. POLYNOMIAL DIVISION

Life is too short for long division - Unknown

CHAPTER
I
I Polynomial Division

A polynomial in one variable is an expression of the form

anxn + an~]Xn~1 + an-2xtl~2 + ■■■ + ∏ιx + «0,
where x is a variable and each αi is a constant. Each individual summand αt∙x' is called a term, and each
number a∣ is the coefficient of the corresponding xl. The degree of a polynomial is the greatest integer
n such that the coefficient an is not equal to 0. So for example, the polynomial 2x4 - x3 + 3x2 -x-7 has
degree 4, and the coefficient of x3 in this polynomial is -1. The degree of a nonzero constant polynomial
is zero, and the degree of the polynomial 0 is undefined. We call the polynomial 0 the zero polynomial,
and any other polynomial is a nonzero polynomial.
We say the function f is a polynomial function if ∕(x) is a polynomial, and we denote the degree of
∕(x) as deg f. We will often refer to a polynomial function simply as a polynomial.
There are some special names for polynomials of certain degrees. For example, a polynomial of
degree 1 is called linear, and a polynomial of degree 2 is called quadratic. The following table lists some
names used to describe polynomials of specific degrees.

Degree Name
0 constant
1 linear
2 quadratic
3 cubic
4 quartic

Polynomials are also sometimes described according to the number of terms they contain. A poly­
nomial with one term is a monomial, with two terms is a binomial, and with three terms is a trinomial.
For instance, 3x4 is a quartic monomial, 2x2 + 1 is a quadratic binomial, and 8x5 -3x4 + 2x3 is a trinomial
of degree 5.

162
 6.1. POLYNOMIAL REVIEW

6.1 Polynomial Review

In this section, we review polynomial addition, multiplication, and evaluation.

Problems

Problem 6.1: Let ∕(x) = 2x2 + 5x + 13 and g(χ) = x3 - 3x - 4.

(a) Find∕(g(3)).
(b) Find ∕(x) + g(x) and determine its degree.
(c) For what value of the constant c does the degree of x ∙ ∕(x) + c ■ g(x) equal 2?
(d) Find ∕(x)g(x) and determine its degree.

Problem 6.2: Prove that if f and g are nonzero polynomials, then deg(∕ ∙ g) = (deg/) + (degg).

Problem 6.3:
(a) Let ∕(x) = 3x3 - 4x2 + ax - 11 be a polynomial such that ∕(1) = 2. Find the value of a.
(b) Suppose (3x - 1)7 = aγx7 + α6χ6 + ∙ ∙ ∙ + α0 for constants ⅜,flι,...,a7. Find 07 + + ■ ∙ ■ + a↑ + Uq.
(Source: AHSME)

Problem 6.1: Let ∕(x) = 2x2 + 5x + 13 and g(x) = x3 - 3x - 4.

(a) Find∕(g(3)).
(b) Find ∕(x) + g(χy) and determine its degree.
(c) For what value of the constant c does the degree of x ∙ ∕(x) + c ■ g(x) equal 2?
(d) Find ∕(x)g(x) and determine its degree.

Solution for Problem 6.1:

(a) We have g(3) = 33 - 3(3) - 4 = 14, so ∕(g(3)) = ∕(14) = 2(14)2 + 5(14) + 13 = 475.
(b) We add the two polynomials the way we add any two expressions—we combine like terms:

∕(x) + g(x) = 2x2 + 5x + 13 + x3 - 3x - 4

= x3 + 2x2 + (5x - 3x) + (13 - 4)
= x3 + 2x2 + 2x + 9.

Since x3 is the highest power of x in ∕(x) + g(χ), the degree of ∕(x) + g(x) is 3.
(c) We have

x ∙ ∕(x) + c ■ g(x) = 2x3 + 5x2 + 13x + cx3 - 3cx - 4c

= (2 + c)x3 + 5x2 + (13 - 3c)x - 4c.

If the degree of this polynomial is 2, then there must be no x3 term. So, we must have 2 + c = 0,

163
CHAPTER 6. POLYNOMIAL DIVISION

which gives us c = -2. When c = -2, we have x ■ f(x) + c ■ g(x) = 5x2 + 19x + 8, which does indeed
have degree 2.
(d) We can use the distributive property to multiply any two polynomials:

(2x2 + 5x + 13)(x3 - 3x - 4) = 2x2(x3 - 3x - 4) + 5x(x3 - 3x - 4) + 13(x3 - 3x - 4)

= (2x5 - 6x3 - 8x2) + (5x4 - 15x2 - 20x) + (13x3 - 39x - 52) (0)
= 2x5 + 5x4 + 7x3 - 23x2 - 59x - 52.

So, the degree of the product is 5.

□
We can also multiply two polynomials by "stacking" them, just as we do for long multiplication of
numbers. Here's an example:

x3 -3x -4
× 2x2 +5x +13
+13x3 -39x -52 * )(
+5x4 -15x2 -20x (*
)
+ 2x5 -6x3 -8x2 (*
)
2x5 +5x4 +7x3 -23x2 -59x -52

The lines marked (* ) in our "stacked" multiplication result from multiplying x3 - 3x - 4 by each
term, in turn, of 2x2 + 5x + 13. In other words, we're just doing the distributive property. We can see
this clearly by noticing that the polynomials in these lines are the same as the polynomials that appear
in the line marked (4>) in our distributive expansion above.
Notice in our final part above that deg∕ = 2, degg = 3, and deg(∕ ∙ g) = 2 + 3 = 5. Is this just a
coincidence?
Problem 6.2: Prove that if f and g are nonzero polynomials, then deg(∕ ∙ g) = (deg ∕) + (deg g).

Solution for Problem 6.2: Let deg f = n and degg = m. To relate the degrees of ∕(x) and g(x) to ∕(x) ∙ g(x),
we start by writing ∕(x) and g(x) in general form:

∕(x) = anxn + fln-ιxπ~1 + an-2xn~2 + ∙ ■ ■ + a↑x + ao,

g(x) = bmxm + bm-ιxm^1 + bm-2Xnι~2 + ∙ ∙ ∙ + b1x + b0.

Because the degree of ∕(x) is n, we know that an is nonzero. (Otherwise, ∕(x) would have no xn term.)
Similarly, we know that bm is nonzero because the degree of g(x) is m. When we take the product
f(x) ■ g(x), the highest power of x will occur when we multiply the highest powers of x in ∕(x) and g(x)
to get (anxn)(bmxm) = anbmxn+m. Because an and bιn are nonzero, we know that a,lblll ≠ 0, so the product
∕(x) ∙ g(x) has a term with xn+m. All other terms in ∕(x) ∙ g(x) have a lower power of x. Since x"+"' is the
highest power of x in ∕ι(x), the degree of h is n + m. Since the degree of f is n and the degree of g is nι,
we see that
(deg/) + (degg) = deg//.

164
 6.1. POLYNOMIAL REVIEW

This relationship holds even if ∕(x) or g(x) is just a nonzero constant, because the degree of a nonzero
constant is 0. However, our rule above breaks if we try to define the degree of the polynomial 0. To see
why, suppose f(x) = 0, g(x) = x2, and ∕z(x) = ∕(x) ∙ g(x). Then, we have h(x) = 0. Our relationship above
gives us
deg ∕ + 2 = deg h,
so we have (degθ) + 2 = (degθ). There's no value we can assign to (degθ) to make this equation true,
so we can't define the degree of 0.

WARNING!! The degree of the polynomial 0 is undefined.

Problem 6.3:
(a) Let ∕(x) = 3x3 - 4x2 + ax - 11 be a polynomial such that ∕(1) = 2. Find the value of a.
(b) Suppose (3x-1)7 = α7x7 + α6χ6 ■+------ι-α1 + α0 for constants ⅜,αι, ...,a7. Find aγ + a^ -t------ ι-flι +flo
(Source: AHSME)

Solution for Problem 6.3:

(a) Substituting x = 1 into the polynomial, we have ∕(1) = 3- 4 + α-ll = a - 12. Since ∕(1) = 2, we
have a - 12 = 2, so a = 14.
(b) Multiplying out (3x - 1)7 looks a little scary. However, a quick glance at part (a) gives us an idea.
In part (a), we see that ∕(1) = 3 — 4 + a —11. These numbers look familiar—they're the coefficients
of ∕(x)! This is no coincidence; when we let x = 1, each term is equal to the coefficient of that
term. Returning to the problem in this part, we see that letting x = 1 gives us the expression
«7 + r⅛ + ∙ ∙ ∙ + «1 + «0 on the right, and our equation then is
(3 ∙ 1 - 1)7 = #7 + #6 + ∙ ∙ ∙ + fl} + πθ.
Evaluating the left side gives usaγ + a^ +---- F + ⅜ = 27 = 128.

Concept: Substituting a value for the variable in a polynomial can give us useful
Q lj1J information about the coefficients of the polynomial. In particular, letting
the variable equal 1 gives the sum of the coefficients of the polynomial.

Speaking of coefficients, we have a little more vocabulary before we continue. The term of a
polynomial with the highest power of the variable is called the leading term and the coefficient of this
term is the leading coefficient. If the leading coefficient of the polynomial is 1, then the polynomial is
called a monic polynomial.

I 1
lxer<J⅜e¼

6.1.1 Let g(x) = 2x2 + 5x + 7. If h(x) is a polynomial such that ∕z(x) = [g(x) + 3][g(x) + 9], then what is the
degree of ∕z(x)7

165
CHAPTER 6. POLYNOMIAL DIVISION

6.1.2 Let ∕(x) = x4 + x2 - 2 and g(x) = 2x3 + 3x2 + 4x + 5.

(a) What is the degree of /(x)?

(b) What is the degree of g(x)?
(c) Find the degree of ∕(g(x)).
(d) Find the degree of ∕4(x).
(e) Find the sum of the coefficients in the expansion of g(∕(x)).

6.1.3 If ax3 + bx2 + ex + d = (x2 + 2x - 8)(x - 3) - (x - 2)(x2 + 5x + 4) for all values of x, then what is the
value of a + b + c + d?
6.1.4 Find the degree of ∕(g(x)) given that f and g are polynomials such that the degree of ∕ is m and
the degree of g is n, where m and n are nonnegative integers.

6.2 Introduction to Polynomial Division

R
I
Problems

Problem 6.4: Suppose that a, br and c are constants such that

(2x2 + 3x + 7)(ax2 + bx + c) = 2x4 + llx3 + 9x2 + 13x - 35.

(a) Use the coefficient of x4 to find a.

(b) Use the coefficient of x3 to find b.
(c) Find c. Check your answer by expanding the product on the left side using your values of a, b,
and c.
-γ3 _j_ rj γ∙2_ "V -1- 11
Problem 6.5: In this problem we divide x3 + 7x2 - x + 11 by x - 5. Let gi(x) = -------- - —- ------- , so
x-5
(x - 5)gι(x) = x3 + 7x2 - x + 11.
(a) Why must the highest degree term of gι(x) be x2?
(b) Let g2(x) = x2 + gι(x), so that we have
x3 + 7x2 - x + ∏ 2 ∕ ∖
= x + g2(x)∙
x-5
Find (x - 5)g2(x) by multiplying this equation by x - 5. Find the highest degree term of g2(x)∙
(c) Let g2 (x) = ax + g3(x)z where ax is the highest degree term from part (b), so we have

x3 + 7x2 - x + 11 2 ∕ x
= xz + ax + g3(x).
x-5
Find g3(x).
÷ 7x^ — χ 4-11 r
(d) Combine the previous parts to write ------ —— -------- = <∕(x) + _ l-, where ι∕(x) is a polynomial
X
x-5□ X c)
and r is a constant.

166
 6.2. INTRODUCTION TO POLYNOMIAL DIVISION

Problem 6.4: Find the constants a, bl c such that

(2x2 + 3x + 7)(«x2 + bx + c) = 2x4 +llx3 + 9x2 + 13x - 35.

Solution for Problem 6.4: The only way to produce an x4 term when expanding the product on the left
side is when multiplying (2x2)(αx2) = 2αx4. This must equal the 2x4 on the right, so we must have 2a - 2,
which gives us a = 1. Now we seek the constants b and c such that
(2x2 + 3x + 7)(x2 + bx + c) = 2x4 + llx3 + 9x2 + 13x - 35.

Looking at the x4 term helped us find a, so let's take a look at the x3 term. On the right, we have llx3,
and the only x3 terms that will appear in the expansion on the left side are (2x2)(bx) and (3x)(x2). So, we
must have
2bx3 + 3x3 = 11x3.
Therefore, we must have 2b + 3 = 11, so b = 4. We could move on to x2 to find c, or we can use a bit of
a shortcut by noticing that the only constant term in the expansion of (2x2 + 3x + 7)(tzx2 + bx + c) is 7c,
so we must have 7c = -35 to match the constant term of the given quartic polynomial. This gives us
c = -5. Checking our answer, we find that
(2x2 + 3x + 7)(x2 + 4x - 5) = 2x4 + llx3 + 9x2 + 13x - 35.

□
This section is titled "Introduction to Polynomial Division," but the last problem is a multiplication
problem. What does it have to do with division?
One way to think of division is as the "reverse" of multiplication. In other words, when trying to
evaluate the division 35/7, we can ask ourselves, "By what must we multiply 7 to get 35?" Because
7 × 5 = 35, we know that 35/7 = 5. In much the same way, because our work in Problem 6.4 tells us that
(2x2 + 3x + 7)(x2 + 4x - 5) = 2x4 + llx3 + 9x2 + 13x - 35,

we know that
2x4 + 11x3 + 9x2 + 13x-35 7
---------- —z— ------ ----------= x + 4x - 5.

Let's try using this insight to perform a polynomial division.

Problem 6.5: Let ∕(x) = x3 + 7x2 - x + 11 and rf(x) = x - 5. Find a polynomial q(x) and a constant r
such that
f(×) - ∕ A r
d(x) + rf(x)'

Solution for Problem 6.5: Let

x3 + 7x2 - x + 11
gι(
*
) =
x-5
We hope to write g√x) in the form t∕(x) + r∕(χ - 5), where q(x) is a polynomial and r is a constant. We
multiply the equation by x - 5 to get rid of the denominator, and we have
(x - 5)gι(x) = x3 + 7x2 - x + 11.

167
CHAPTER 6. POLYNOMIAL DIVISION

This means we seek the expression that we must multiply x - 5 by to get x3 + 7x2-x+H. The leading term
on the right is x3, so we know the leading term of (x - 5)gι(x) must also be x3. So, we let gι(x) = x2 + g2(x)
for some expression g2(x), since this makes the leading term of (x — 5)(x2 + g2(x)) be χ3■ So, now we have

(x - 5)(x2 + g2(x)) = x3 + 7x2 - x + 11.

Expanding the product on the left gives

(x - 5)x2 + (x - 5)g2(x) = x3 + 7x2 - x + 11.

Subtracting (x - 5)x2 from both sides gives us

(x - 5)g2(x) = x3 + 7x2 - x + 11 - (x - 5)x2 = 12x2 - x + 11.

We need a 12x2 term in the expansion of (x - 5)g2(x) to match the 12x2 term in 12x2 - x + 11, so we let
g2(x) = 12x + g3(x), which gives us

(x - 5)(12x + g3(x)) = 12x2 - x + 11.

Expanding the product on the left gives

(x - 5)(12x) + (x - 5)g3(x) = 12x2 - x + 11,

so we have
(x - 5)g3(x) = 12x2 - x + 11 - (x - 5)(12x) = 59x + 11.
We need a 59x term in the expansion of (x-5)g3(x) to match the 59x on the right, so we let g3(x) = 59+g4(x),
which gives us
(x - 5)(59 + g4(x)) = 59x + ll.
Expanding the left side gives us

59(x - 5) + (x - 5)g4(x) = 59x + 11.

Simplifying this equation, we have

(x - 5)g4(x) = 306.
Uh-oh. Clearly g4(x) isn't a polynomial. Did we make a mistake?

No! Dividing this equation by x - 5 gives us

If we look back through our steps, we have

gι(x) = x2 + g2(x)
= x2 + 12x + g3(x)
= x2 + 12x + 59 + g4(x)
9 γ-λ 3θ6
= x2 + 12x + 59 H------- -.
x-5

168
 6.2. INTRODUCTION TO POLYNOMIAL DIVISION

So, we have
x3 + 7x2-x + 11 2 1o rn 306
------------- ------- = x2 + I2x + 59 + -,
x-5 x-5
which means our desired q(x) and r are q(x) = x2 + 12x + 59 and r = 306. □
When we divide one integer into another, we have a quotient and a remainder. For example, 233
divided by 5 gives a quotient of 46 and a remainder of 3:

Multiplying this equation by the divisor, we see that we can also express our division in terms of
multiplication and addition:
233 = 5 × 46 + 3.
The remainder of a division is always less than the divisor (the number we are dividing by).
In general, we can express the fact that n divided by d gives a quotient q with remainder r (where
0 ≤ r < d) as
n r

Multiplying both sides by d gives us

n = qd + r.
So, when dividing an integer n by another integer d, we seek the integers q and r such that n = qd + r
and 0 <r < d.

Dividing polynomials is essentially the same game. When we divide the polynomial ∕(x) by the
polynomial d(x), we seek polynomials q(x) and r(x) such that

∕(*)
rf(x)
Multiplying both sides by rf(x) gives us

∕(x) = q(x) ■ d(x) + r(x).

When dividing one integer into another, the remainder must be less than the divisor. For polynomials,
our measure of the "size" of a polynomial is its degree. When dividing the polynomial d(x) into ∕(x), the
remainder must either be 0, or the degree of the remainder must be less than the degree of the divisor.

Important: When the polynomial ∕(x) is divided by the polynomial rf(x), the quotient
ψ q(x) and remainder r(x) are the polynomials such that

∕(x) = q(x) ■ d(x^) + r(x)

and either r(x) = 0 or deg r < deg d.

If the remainder is 0 when the polynomial ∕(x) is divided by the polynomial rf(x), then we say that
∕(x) is divisible by rf(x). Again, this is just like division of integers: if the remainder is 0 when the integer
n is divided by an integer d, then n is divisible by d.

169
CHAPTER 6. POLYNOMIAL DIVISION

We've already seen a couple examples of polynomial division. In Problem 6.4, we found that

2x4 + llx3 + 9x2 + 13x - 35 2 , λ k

= xz + 4x - 5,
2x2 ÷ 3x + 7
so the quotient when 2x4 + llx3 + 9x2 + 13x - 35 is divided by 2x2 + 3x + 7 is x2 + 4x - 5, and the remainder
is 0. Therefore, 2x4 + llx3 + 9x2 + 13x - 35 is divisible by 2x2 + 3x + 7.

In Problem 6.5, we found that

x3 + 7x2 - x + 11 2 -10 l c∩ , 306

---------------------- = xz + 12x + 59 +------ -. (*)
x-5 *
~5

The quotient is x2 + 12x + 59 and the remainder is 306.

WARNING!! The remainder is 306, not 306∕(x - 5), just as the remainder when 33
V is divided by 5 is 3, not 3/5.

Multiplying Equation (⅛) by x - 5 gives us

x3 + 7x2 - x + 11 = (x - 5)(x2 + 12x + 59) + 306.

We can check our work by expanding the product on the right side:

(x - 5)(x2 + 12x + 59) + 306 = x(x2 + 12x + 59) - 5(x2 + 12x + 59) + 306
= x3 + 12x2 + 59x - 5x2 - 60x - 295 + 306
= x3 + 7x2 - x + 11.

This matches our original ∕(x), so our quotient and remainder are correct.

Concept: Just as you can check integer division with multiplication and addition,
you can check polynomial division. If ∕(x) divided by d(x) gives a quotient
of (∕(x) and remainder of r(x), we must have

rf(x) ∙ q(x) + r(x) = ∕(χ).

You may have found a somewhat faster solution to Problem 6.5 on your own. However, we can
use our solution above to build a method of long division for polynomials that is very much like the
long division we perform to divide one integer into another. We'll now walk through our solution to
Problem 6.5 again, and develop our long division method as we go.

Our initial set-up is just like integer long division; this is shown (6.1)
at right. When we are finished, line (6.1) will contain the quotient. v _ 5 ∣χ3 + 7χ2 _ χ + 11 (6 2)
On line (6.2), we place the polynomial we are dividing inside the
division bracket and we place the divisor on the left.

170
 6.2. INTRODUCTION TO POLYNOMIAL DIVISION

Our first step in our solution to Problem 6.5 was to find the highest degree term of g↑ (x) such that

(x - 5)gι (x) = x3 + 7x2 - x + 11.

Since x times the highest degree term of g] (x) must equal x3, the highest degree term of g↑ (x) is x3∕x = x2.
This is the leading term of our quotient.

In our long division, we find the leading term of our quotient by x2 (6.1)
dividing tire leading term of the divisor (x - 5) into the leading term χ _ 5 ∣χ3 + 7χ2 _ γ + ∩ 2)
of x3 + 7x2 - x +11. We place the resulting x2 in line (6.1) as the leading
term of our quotient.

)∕
Next in our solution to Problem 6.5, we write gι(x) = x2 + g2(
* so (x-5)(x2 + g2(x))= x3 + 7x-x + ll.
Expanding the left side gives us (x - 5)x2 + (x - 5)g2(x) = x3 + 7x-x + ll. Rearrangingthis gives us

(x - 5)g2(x) = x3 + 7x - x + 11 - (x - 5)x2 - 12x2 - x + 11. (*

)

So, (x - 5)g2(x) gives us what's left when (x - 5)x2 is subtracted from x3 + 7x-x + ll.

In our long division, we multiply the quotient term we just found x2 (6.1)
(x2) by the divisor (x—5), and subtract this product from x3+7x-x+ll x - 5∣x3 + 7x2 - x + 11 (6.2)
on line (6.3). The subtraction is exactly the same as the subtraction
x3 - 5x2__________ (6.3)
on line (♦). Notice that the resulting difference, 12x2 - x + 11, on line
(6.4) is the same as the final right side on line (4) above. 12x2 - x + 11 (6.4)

Now, we simply repeat the process, finding the highest degree term of g2(x) using our equation
(x - 5)g2(x) = 12x2 - x + 11.

Notice that we are essentially just dividing x - 5 into 12x2 - x + 11 in this step. The highest degree term
of g2(x) times x gives us 12x2, so the highest degree term of g2(x) is 12x2∕x = 12x.

In our long division, we are also now dividing x — 5 into the x2 + 12x (6.1)
polynomial 12x2 — x + 11 on line (6.4). We find the next term of the x - 5 ∣x3 + 7x2 - χ + 11 (6.2)
quotient by dividing the leading term of the divisor x - 5 into the
x3 - 5x2__________ (6.3)
leading term of our difference 12x2 - x + 11 on line (6.4). This gives
us 12x2∕x = 12x, which we place as our next term of the quotient in 12x2 - x + 11 (6.4)
line (6.1).

Extra! The Universe is a grand book which cannot be read until one first learns to comprehend the
" l* language and become familiar with the characters in which it is composed. It is written in the
l"* l**
language of mathematics.
- Galilei Galileo

171
CHAPTER 6. POLYNOMIAL DIVISION

Next up in our solution to Problem 6.5, we write g2(x) - 12x + g3(x), sθ we have
* 2 - x + 11.
(x - 5)(12x + g3(x)) = 12

As before, we expand the left side to get (x - 5)(12x) + (x- 5)g3(x) = 12x2 - x +11. Rearranging then gives

(x - 5)g3(x) = 12x2 - x + 11 - (x - 5)(12x) = 59x + 11. (♦)

So, (x - 5)g3(x) gives us what's left when (x - 5)(12x) is subtracted from 12x2 - x + 11.

Now back to our long division. We multiply the quotient term x2 + 12x (6.1)
we just found (12x) by the divisor (x - 5), and subtract this product -5 x3 + 7x2 - X + 11 (6.2)
from 12x2 - x +11 on line (6.4). The subtraction is exactly the same x3 - 5x2 (6∙3)
as the subtraction on line (♦). Notice that the resulting difference
12x2 - X + 11 (6∙4)
on line (6.6), 59x + 11, is the same as the final right side on line
(A) above. This step should look very familiar. It's essentially the 12x2 - 60x (6∙5)
same process that produced line (6.4). 59x + 11 (6∙6)

You know what's coming next. Our equation (x - 5)g3(x) = 59x + 11 tells us that the highest degree
term of g3(x) is 59x∕x = 59. So, we have g3(x) = 59 + g4(x), and (x - 5)(59 + g4(x)) = 59x + 11. Expanding
the left side and rearranging gives
(x - 5)g4 = 59x + 11 - (x - 5)(59) = 306.
We can't divide anymore, so 306 is our remainder and our quotient is x2 + 12x + 59.

We continue with our long division by dividing x - 5 into x2 + 12x + 59 (6.1)

59x + 11 on line (6.6). This tells us the next term of our quotient ∣x3 + 7x2 - X ÷ 11 (6.2)
is 59x∕x = 59, which we place on line (6.1). We then subtract the
x3 - 5x2 (6.3)
product of this term and the divisor (59(x - 5)) on line (6.7), which
gives us a difference of 306 on line (6.8). We can't divide any 12x2 - X ÷ 11 (6.4)
further, because the leading term of the divisor doesn't divide 12x2 -- 60x (6.5)
into the leading term of line (6.8). So, line (6.1) gives us our 59x + 11 (6.6)
quotient and line (6.8) gives us our remainder, and we have
59x 295 (6.7)
x3 + 7x - x + 11 306 306 (6.8)
= x2 + 12x + 59 +
x-5 x - 5’

Now, it's your turn to try long division. Don't be intimidated by more complicated divisors such as
2x - 1 or x2 - 3x + 5. The long division process for these works exactly the same way it did above for
x - 5. At each step, we are dividing the divisor into a polynomial. We divide the leading term of the
polynomial by the leading term of our divisor to get the next term in the quotient. Then, we subtract
the product of this next term and the divisor from the polynomial. This gives us a new polynomial to
divide by the divisor in our next step.

Problems

Problem 6.6: Find the quotient and remainder when x + 6 is divided into 3x2 + 3x - 7.

172
 6.2. INTRODUCTION TO POLYNOMIAL DIVISION

Problem 6.7: Steve divided 2x - I into 6x', - 7x3 + 22x2 - 24x - 13 and came up with a quotient of
6x3 - x2 + 21x - 3 and a remainder of -16. Looking at his answer, he immediately knew he had
made a mistake. How could he quickly tell he made a mistake, and what are the correct quotient and
remainder?

Problem 6.8: Steve's at it again. He divides x2 - 3x + 5 into x4 - 3x3 + 7x - 1, and gets a quotient of
x with a remainder of 2x - 1. Again, he immediately realizes he has made a mistake. How could he
quickly tell he made a mistake, and what are the correct quotient and remainder?

Problem 6.9: In this problem we explore why the quotient and remainder are unique when one
polynomial is divided into another. Suppose we are dividing the polynomial d(x) into ∕(x). Let q↑(x)
and Γι(x) be polynomials such that either ∏(x) = 0 or deg r↑ < degd, and

∕(x) = d(x) ■ q1(x) + r1(x).

Let q2(x) and i'2(x) also be polynomials such that either r2(x) = 0 or deg r2 < deg d, and

∕(x) = d(x) ∙ q2(x) + r2(x).

We will show that q2(x) and t∕ι(x) must be the same, and that r2(x) and fι(x) must be the same.
(a) Eliminate ∕(x) from our two equations by subtracting the second from the first.
(b) Rearrange the equation in part (a) by putting the remainders on the right side and the other
terms on the left side.
(c) Compare the degrees of both sides of the equation from part (b) to explain why ⅛ι(x) and t∕2(x)
must be the same, and rι(x) and r2(x) must be the same.

Problem 6.6: Find the quotient and remainder when x + 6 is divided into 3x2 + 3x - 7.

Solution for Problem 6.6: We divide the leading term of x + 6 into the 3x - 15 (6.6.1)
leading term of 3x2+3x-7 to get the first term of our quotient, 3x2∕x = 3x. x + 6∣3x2 + 3x -- 7
(6.6.2)
We place this term on line (6.6.1), and then place the product of this term
3x2 + 18x (6.6.3)
and the divisor on line (6.6.3). We then subtract this product on line
(6.6.3) from 3x2 + 3x - 7 to get line (6.6.4). Then, we divide the leading — 15x -- 7 (6.6.4)
term of x + 6 into the leading term on line (6.6.4), which gives us the - 15x -- 90 (6.6.5)
next term of our quotient, -15. We place this term on line (6.6.1), then
83 (6.6.6)
subtract the product of this term and x + 6 from line (6.6.4). The degree
of the difference (line (6.6.6)) is less than the degree of x + 6, so we can't divide anymore.
The quotient is 3x - 15 and the remainder is 83. Checking our answer, we have

(3x - 15)(x + 6) + 83 = 3x2 + 3x - 90 + 83 = 3x2 + 3x - 7,

as expected. □

173
CHAPTER 6. POLYNOMIAL DIVISION

Problem 6.7: Steve divided 2x - 1 into 6x4 - 7x3 + 22x2 - 24x - 13 and came up with a quotient
of 6x3 - X2 + 21 x - 3 and a remainder of -16. Looking at his answer, he immediately knew he had
made a mistake. How could he quickly tell he made a mistake, and what are the correct quotient and
remainder?

Solution for Problem 6.7: If Steve's answer is correct, then we must have

6x4 - 7x3 + 22x2 - 24x - 13 = (6x3 - x2 + 21x - 3)(2x - 1) - 16.

We don't have to multiply out the entire right side to see that the two sides are not the same. The leading
term of the right side is clearly 12x4, which doesn't match the leading term on the left. So, Steve made
a mistake. Let's figure out what the answer should have been. (As an extra challenge, see if you can
figure out how Steve made his mistake.)

As before, we form our quotient by dividing the 3x3 - 2x2 + 10x - 7 (6.7.1)
leading term of the divisor into the leading terms of 2x-1∣6x4 - 7x3 + 22x2 - 24x - 13 (6.7.2)
each of the polynomials on lines (6.7.2), (6.7.4), (6.7.6), 6x4 - 3x3___________________ (6.7.3)
and (6.7.8). On lines (6.7.3), (6.7.5), (6.7.7), and (6.7.9),
we subtract the product of the divisor and each suc­ - 4x3 + 22x2 - 24x - 13 (6.7.4)
cessive term of the quotient, thereby creating the next - 4x3 + 2x2____________ (6.7.5)
polynomial into which we divide 2x — 1. The degree 20x2 - 24x - 13 (6.7.6)
of the polynomial we produce on line (6.7.10) is less
____________ 20x2 - 10x (6.7.7)
than the degree of the divisor, so we can't divide any
further (and line (6.7.10) is our remainder). - 14x - 13 (6.7.8)
_________________ - 14x + 7 (6.7.9)
So, the quotient is 3x3 - 2x2 + 10x — 7 and the re­
mainder is -20. We can check this in the usual way: - 20 (6.7.10)

(2x - 1)(3x3 - 2x2 + 10x - 7) - 20 = 2x(3x3 - 2x2 + 10x - 7) - l(3x3 - 2x2 + 10x - 7) - 20
= 6x4 - 4x3 + 20x2 - 14x - 3x3 + 2x2 - 10x + 7 - 20
= 6x4 - 7x3 + 22x2 - 24x - 13.

This matches the original polynomial we are dividing into on line (6.7.2), so our quotient and remainder
are correct. □
Problem 6.8: Steve's at it again. He divides x2 - 3x + 5 into x4 - 3x3 + 7x - 1, and gets a quotient of I
x with a remainder of 2x - 1. Again, he immediately realizes he has made a mistake. How could he ∣
quickly tell he made a mistake, and what are the correct quotient and remainder?

Solution for Problem 6.8: If Steve is correct, then we must have x4 - 3x3 + 7x - 1 = (x2 - 3x + 5)(x) + 2x - 1,
but the degree of the right side is not the same as the degree of the left side. Therefore, the two sides
cannot be the same.
We'll have to fix Steve's mistake. Before we get started with our division, we notice that the
polynomial we are dividing has no x2 term. (Hmmm.... Maybe that's what Steve overlooked.)

174
 6.2. INTRODUCTION TO POLYNOMIAL DIVISION

WARNING!! When dividing one polynomial by another, be careful not to overlook

V "missing" terms. These terms aren't really missing—their coefficients
are 0. Include such terms in your long division by writing them with
a coefficient of 0.

Even though our divisor is not linear, we can use x2 - 5 (6.8.1)

long division just as we have done before. To find the χ2 _ ^χ + 5 ∣χ4 _ 3χ3 + θχ2 + yχ _ (5 g.2)
first term of our quotient, we divide the leading term
x4 - 3x3 + 5x2____________ (6.8.3)
of tlιe divisor into the leading term of the polynomial
on line (6.8.2). This gives us x4∕x2 = x2 as the first - 5x2 + 7x - 1 (6.8.4)
term of the quotient, and we place this term on line _________ - 5x2 + 15x - 25 (6.8.5)
(6.8.1). We then subtract the product x2(χ2 - 3x + 5) - 8x + 24 (6.8.6)
on line (6.8.3). Notice that both the x4 and x3 terms
are eliminated. So, when we divide the leading term of the divisor into the leading term of the difference
on line (6.8.4), the next term of our quotient is a constant term, -5x2∕x2 = -5, not a linear term. We then
subtract (-5)(x2 — 3x + 5) on line (6.8.5). The resulting difference on line (6.8.6) has a smaller degree than
the divisor, so we can't divide anymore.
The quotient is x2 — 5 and the remainder is —8x + 24. As before, we check our work:
(x2 - 3x + 5)(x2 - 5) - 8x + 24 = (x2 - 3x + 5)(x2) + (x2 - 3x + 5)(-5) - 8x + 24
= x4 - 3x3 + 5x2 - 5x2 + 15x - 25 - 8x + 24
= x4 - 3x3 + 7x - 1.

Throughout this section, we have assumed that it is always possible to find a quotient, r∕(x), and
remainder, r(x), such that ∕(x) = q(x) ■ d(x) + r(x) and either deg r < deg d or r(x) = 0. Our long division
process makes it intuitively obvious why this is true, but we won't go through all the details proving
this fact. We will, however, prove that there's not more than one possible quotient and remainder for
each polynomial division.

Problem 6.9: Show that for any two polynomials ∕(x) and d(x), where d(x) ≠ 0, there is not more
than one pair of polynomials q(x) and r(x) such that ∕(x) = q(x) ■ d(x) + r(x), and either r(x) — 0 or
deg r < deg d.

Solution for Problem 6.9: Suppose we have

∕(x) = ^](x) ∙ d(x) + r1(x)

and
∕(x) = q2(x) ■ d(x) + r2(x),
where deg∏ < degd (or r1(x) = 0) and degr2 < degd (or r2(x) = 0).
Our two equations have a lot of different polynomials in them. We can at least eliminate ∕(x) by
subtracting the second equation from the first. This gives us

0 = <∕1(x) ∙ d(x) - q2(x) ■ d(x) + r1(x) - r2(x).

175
CHAPTER 6. POLYNOMIAL DIVISION

We bring the two terms with d(x) to the left and factor d(x) out of both terms to give

d(x)(t∕2(x) - <∕ι(x)) = fιW - f2W∙

The left side of this equation either is 0 or has degree at least degd. However, because the degree of
a remainder cannot equal or exceed that of the divisor, the degrees of ∏(x) and r2(x) cannot equal or
exceed degd, which means that deg(r1 - r2) ≤ degd or rι(x) - r2(x) = 0. Therefore, the right side of the
equation either is 0 or has degree less than deg d. Combining this with our observation about the degree
of the left side, we see that the only way the two sides are equal for all x is if both sides are 0.
Since ri(x) - r2(x) = 0, we have r1(x) = r2(x). Since d(x)(q2(x) ~ <7i(χ)) = θ and rf(x) is a nonzero
polynomial, we must have tj2(x) - <j1(x) = 0, so zy2(x) = <∕ι(x) for all x. (This glosses over the possibility
that t}2(x) = t∕ι(x) for all x except where d{x) = 0. That this is impossible is a consequence of the Identity
Theorem, which we study in Section 7.6.)
We have therefore shown that for any two polynomials ∕(x) and c∕(x), there cannot be more than one
pair of polynomials t∕(x) and r(x) such that either r(x) = 0 or deg r < deg d, and

∕(x) = q(x) ■ d(x) + r(x).

Note that we've only proved that there is no more than one pair of polynomials q(x) and r(x) that satisfy
these conditions. We haven't proved that these polynomials tj(x) and r(x) must always exist, though our
process for polynomial division described earlier in this section gives us some intuition for why they
must exist. □

Important: The quotient and remainder in polynomial division are unique. In other
words, for any two polynomials ∕(x) and d(x), there's exactly one pair of
polynomials q(x) and r(x) such that either r(x) = 0 or deg r < deg d, and

∕(x) = t∕(x) ∙ d(x) + r(x).

Exercises

6.2.1 Find the quotient and remainder for each of the following polynomial divisions:

(a) x2 - 19x + 17 divided by x + 7

(b) 3x4 + 13x2 - 9x + 122 divided by x2 + x + 4
(c) 4x3 + 2x - 1 divided by 2x + 3
(d) 19x5 divided by x - 1
6.2.2 Find a constant c such that there is no remainder when x3 + cx2 + 4x - 21 is divided by x - 3.
6.2.3 Find a if the remainder is a constant when x3 + 3x2 + ax + 13 is divided by x2 + 3x - 2.

6.2.4 Teresa divides 3x4 + 2x3 - 7x2 + 4x -1 by x + 2 and gets a quotient of 3x3 -4x2 + x + 2 and a remainder
of 5. How can Teresa quickly realize that she made a mistake without performing the division again,
and without multiplying x + 2 by her quotient?

176
 6.3. SYNTHETIC DIVISION

6.3 Synthetic Division

Dividing polynomials by binomials of the form x - a will prove to be very useful. But long division is,
well, pretty long. Fortunately, there's a more compact way to divide a linear binomial into a polynomial.
We call this method synthetic division.
We'll walk through an example of synthetic division, then let you try.

____________________ (6.1) 1 7 -1 11 (?)

x - 51 x3 + 7x2- x÷∏ (6.2) (»)
’ (*
)

Above, we have the initial set-up for long division on the left and for synthetic division on the right.
When we divide a polynomial by x - a using synthetic division, we start by placing a to the left of our
division bracket and the coefficients of the polynomial we are dividing (including coefficients of 0 for
any missing terms) inside the bracket.

WARNING!! When we divide by x — a using synthetic division, we put a to the left

* of the division bracket, not -a. So, for example, when we divide by
x - 5, we put 5 to the left of the division bracket.

We have one column inside the bracket for each power of the variable in the polynomial, just as we have
one column in our long division for each power of the variable. The row (⅛) is used for our quotient and
remainder.

lx2 (6.1) 1 7 -1 11 (?)

x - 51 x3 + 7x2 - x + 11 (6.2) _________________ (*)
1 (♦)
Because we are dividing by x — 5, the leading term of the quotient has the same coefficient as the
leading term of the polynomial we are dividing. So, we copy the leading coefficient of the polynomial
we are dividing to the quotient row (♦).

WARNING!! Synthetic division only works when the coefficient of the linear term
in the divisor is 1.

Later in this section, we'll learn how to handle divisors in which the coefficient of the linear term is not 1.

Extra! Synthetic division is also occasionally referred to as Ruffini's rule, after Italian math-
' ∣∙∙i"
∣"* ∣* * ematician Paolo Ruffini, who was supposedly one of the first to use this method of
polynomial division. Though Ruffini was appointed professor of the foundations of
analysis (a branch of mathematics) at the University of Modena in 1788 when he was
only 23, he also held degrees in philosophy, medicine, and surgery.

177
CHAPTER 6. POLYNOMIAL DIVISION

lx2 (6.1) 1 7 -1 11 (?)

x - 51 x3 + 7x2 - x + 11 (6.2) 5 (0)
1 (*)
x3 - 5x2 (6.3)

In the long division above, after finding the first term of the quotient, we will subtract x2(x - 5) from
the polynomial on line (6.3) above. We know that the x3 terms will cancel in this subtraction, so we
focus on the x2 terms. Notice how the —5x2 term is formed: it is the product of the term we just put in
the quotient (x2) and the constant term of the divisor (-5). Therefore, the coefficient equals the product
of the coefficient of the term we just put in the quotient and the constant term of the divisor.
Turning to our synthetic division, the number to the left of the line is the negative of the constant
term of the divisor, and the number we just placed in the quotient row (♦) is the coefficient of the term
we just found in the quotient. So, we place the product of these two numbers, 5 ∙ 1 = 5, in the next
column on row (0). The 5 we place in this row corresponds to the -5x2 in line (6.3) of our long division.
The 5 in row (0) is the opposite of the coefficient of x2 in line (6.3) because the sign of the number to the
left of the line in the synthetic division is the opposite of the constant term of the divisor.

lx2 + 12x (6∙1) 5 1 7 -1 11 (?)

5∣ x3 + 7x2 - x + 11 (6.2) 5 (0)
1 12 (♦)
x3 - 5x2 (6∙3)
12x2 - x + 11 (6.4)

In the long division, we subtract -5x2 on line (6.3) from the 7x2 on line (6.2) to give us 12x2 on line
(6.4). This is the leading term of the next polynomial into which we divide x - 5. So, the next term of
our quotient is 12x2∕x = 12x.
Turning to our synthetic division, the 5 in row (<0>) is the opposite of the corresponding coefficient of
x2 in line (6.3). Therefore, instead of subtracting the 5 in row (0) from the 7 in row (?), we add the 5 in
row (0) and the 7 in row (?) to produce the 12 in row (*). This represents adding 7x2 + 5x2 (which is the
same as the subtraction 7x2 - (-5x2)) to get 12x2, and it indicates that the coefficient of this sum is the
next term in our quotient.

Sidenote: In addition to preferring synthetic division because it's more compact and
K faster than long division, we also prefer it because it involves addition at
each step rather than subtraction. We're less likely to make sign errors or
other arithmetic errors with addition than subtraction.

That the coefficient of the leading term of our polynomial on line (6.4) is also the coefficient of the
next term in the quotient is a direct result of the fact that the coefficient of the linear term of the divisor
x - 5 is 1. This is why we must have this coefficient equal to 1 in order to use synthetic division.

178
 6.3. SYNTHETIC DIVISION

lx2 + 12x (6.1) 5 1 7 -1 11 (<7)

x - 51 x3 + 7x2 - x + 11 (6.2) 5 60 (9)

1 12 (♦)
x3 - 5x2 (6.3)
12x2 - x + 11 (6.4)
12x2 - 60x (6.5)

We repeat our steps to produce the next term of the quotient. In the long division, we subtract
12x(x - 5) = 12x2 - (5)(12)x on line (6.5), which is the same as adding -12x2 + 60x. In our synthetic
division, we multiply the 5 to the left of the line by the 12 in line (♦) and place the product in the next
column of line (❖).

lx2 + 12x + 59 (6.1) 1 7 -1 11 (9)

x - 51 x3 + 7x2 - x + 11 (6.2) 5 60 (❖)
1 12 59 (♦)
x3 - 5x2____________ (6.3)
12x2 - x + 11 (6.4)
12x2 - 60x (6.5)
59x + 11 (6.6)

Our next step is subtracting line (6.5) from line (6.4). The x2 terms cancel, and subtracting -lx- (-60x)
gives us 59x as the leading term of our polynomial on line (6.6). Turning to our synthetic division, we
add the -1 in row (c7) to the 60 below it in row (0) to produce the 59 in row (♦). This represents our
adding —lx + 60x (which is the same as -lx - (-60x)) to get 59x, and it indicates that the coefficient of
the next term in our quotient is 59.

lx2 + 12x + 59 (6.1) 5 1 7 -1 11 (ς>)

∣ x3 + 7x2 - x + 11 (6.2) 5 60 295 (9)

1 12 59 306 *
)
(
x3 - 5x2 (6.3)
12x2 - x + 11 (6.4)
12x2 - 60x (6∙5)
59x + 11 (6.6)
59x - 295 (6.7)
306 (6.8)
We repeat the process one more time to find the remainder. Subtracting 59(x - 5) on line (6.7) is the
same as adding -59x + 295. The x terms cancel, and we have 11 + 295 = 306 as our remainder. Over in
the synthetic division, multiplying 5 by the 59 in row (♦) gives us the 295 in row (❖) that corresponds to
the -295 we subtract in row (6.7) in our long division. Adding the 11 in row (t7) to the 295 in row (<0>) to
get 306 in row (⅜) corresponds to our subtraction 11 — (—295) to find the remainder in our long division.
We typically separate the remainder from the quotient with a vertical bar in row (⅜) of the synthetic

179
CHAPTER 6. POLYNOMIAL DIVISION

division. This helps us avoid making a mistake when reading the quotient and remainder from row (♦).
The number after the bar is our remainder, and the other numbers are the coefficients of the quotient.
The number just before the bar is the constant term. The next number to the left is the coefficient of the
linear term, the next number to the left is the quadratic term's coefficient, and so on. So, our synthetic
division gives us a quotient of x2 + 12x + 59 with a remainder of 306. □
Below is one more example, in which we divide x - 2 into 3x4 — 9x3 + 2x2 + 17x — 24.

3x3 - 3x2 - 4x + 9 3 -9 2 17 -24 (Y)

x-2∣3x4 - 9x3 + 2x2 + 17x - 24 6 -6 -8 18 (0)
3 -3 -4 9 -6 (♦)
3x4 - 6x3
- 3x3 + 2x2 + 17x - 24
- 3x3 + 6x2
- 4x2 + 17x - 24
- 4x2 + 8x
+ 9x - 24
+ 9x - 18
- 6
Notice that all the numbers in our synthetic division, or their opposites, appear as coefficients (or
constants) in the long division.

Concept: Synthetic division is nothing new. It's just a more compact way to represent
Q l√⅛ long division of a polynomial ∕(x) by an expression of the form x - a, where
α is a constant.

Problems

Problem 6.10: Divide x + 5 into x3 - 3x2 + 5x - 2 using long division, then using synthetic division.

Problem 6.11: In this problem we use synthetic division to divide 2x - 7 into ∕(x) = 4x2 - 6x + 9.
(a) Write 2x - 7 in the form c(x - fl), where c and a are constants.
(b) Use synthetic division to divide your x - fl from part (a) into ∕(x).
(c) Use your result from (b) to find the quotient and remainder when ∕(x) is divided by 2x - 7.

Problem 6.12: Perform the following divisions with synthetic division:

(a) (x3 - x + 7)∕(x + 3)
(b) (3y4 -4y3 + 5y2 - lly + 2)∕(2 + 3y)

Problem 6.10: Divide x + 5 into x3 - 3x2 + 5x - 2 using long division, then using synthetic division.

180
 6.3. SYNTHETIC DIVISION

Solution for Problem 6.10: What's wrong with the synthetic division below:

Bogus Solution:
5 1 -3 5 -2
5 10 75
1 2 15 73

When we divide a polynomial by x - a with synthetic division, we put a on the left of the division
bracket. Here, we are dividing by x + 5, which is x — (-5). So, we should put —5, not 5, to the left of the
division bracket.

WARNING!! When we divide a polynomial by x - a with synthetic division, we put

a a on the left of the division bracket. So, when we divide by x + c, we
put —c to the left of the division bracket, not c.

Below are the correct long and synthetic divisions.

x2 - 8x + 45 -5 1 -3 5 -2
x + 51 x3 - 3x2 + 5x - 2 -5 40 -225
1 -8 45 -227
x3 + 5x2
- 8x2 + 5x - 2
- 8x2 - 40x
45x - 2
45x + 225
- 227

The quotient is x2 - 8x + 45 and the remainder is -227. □

Problem6.il

Solution for Problem 6.11: Why is the following solution wrong:

Bogus Solution:
7 4-6 9
28 154
4 22 ∣ 163

So, the quotient is 4x + 22 and the remainder is 163.

We see that we've made a mistake when we check our answer, because we have

(2x - 7)(4x + 22) + 163 = 8x2 + 16x + 9.

Uh-oh. If we had divided correctly, we should have gotten 4x2 - 6x + 9 when we multiplied 2x - 7 by
the quotient and added the remainder. What went wrong?

181
CHAPTER 6. POLYNOMIAL DIVISION

We forgot that synthetic division only works if the leading coefficient of the divisor is 1. So, are we
stuck using long division on this problem?
No! We are performing the division
4x2 - 6x + 9
2x - 7
We would like to convert the division into an equivalent problem such that the leading coefficient of the
divisor is 1. We can make the coefficient of x equal to 1 in the divisor by factoring 2 out of 2x - 7 to give
us 2(x - j):
4x2 - 6x + 9 _ 4x2 - 6x ÷ 9 _ 1 4x2 - 6x + 9
2x - 7 2(x -1) 2 x- ∖

Aha! We can divide x - 5 into 4x2 - 6x + 9 using synthetic division.

7 4 -6 9
Our synthetic division is shown at right. We appear to have a quotient of 2
4x + 8 and a remainder of 37. Let's check our answer: 14 28
4 8 37
(2x - 7)(4x + 8) + 37 = 8x2 - 12x - 19.

Uh-oh. Wrong again. But why?

Our synthetic division tells us that

4x2 - 6x + 9
= 4x + 8 +

However, we want to perform the division (4x2 - 6x + 9)∕(2x - 7). Looking back at how we came up
with (4x2 - 6x + 9)∕(x - 5) tells us how to use the result of this division to find the desired quotient and
remainder:
4x2 — 6x + 9 1 4x2 - 6x + 9
2x-7 2 χ— -
Λ 2

1
4x + 8 +
2 <
37
= 2x + 4 +
2(x-Z)
37
= 2x + 4 + ------ -
2x-7
Now, we see that the quotient is 2x + 4 and the remainder is 37.
We've been wrong twice, so we better check our answer:
(2x - 7)(2x + 4) + 37 = 4x2 - 6x + 9.

Phew. It works. □
Problem 6.∏ shows us that we can use synthetic division to divide by a linear expression with a
leading coefficient besides 1. However, we must first turn the division into an equivalent problem in

182
 6.3. SYNTHETIC DIVISION

which the divisor has leading coefficient 1. As we saw in our solution to Problem 6.11, we must be
careful about the remainder. Here are two ways to make sure you have your division correct when
dividing ∕(χ) by d(x):

1. Write your answer in the form

∕ω = w + -⅛>
rf(x) ’ d(x)'
where ti(x) is the original divisor. In Problem 6.11, doing this enabled us to get the correct remainder.

2. Check your answer! Once you have found q(x) and r(x), expand d(x) ■ c∣(x) + r(x). If the result is not
∕(x), then you made a mistake.

For one more example, we divide 9x3 — 24x2 + 12x —14 by — 3x — 1. First, we convert the problem into
a division by an expression with leading coefficient 1, by factoring -3 out of the divisor:

9x3 - 24x2 + 12x - 14 _ 9x3 - 24x2 + 12x - 14

-3x-l ^ -3(x÷i)
1 9x3 - 24x2 + 12x - 14

We divide x + ∣ into 9x3 - 24x2 + 12x - 14 at right. Note that we put 9 -24 12 -14
—∣ to the left of the division bracket, not ∣. Our synthetic division gives
-3 9 -7
us 9 -27 21 ∣ -21
9x3 - 24x2 + 12x - 14
= 9x2 - 27x + 21 +----- -.
x+ ∣ x+ 1
So, we have

9x3 - 24x2 + 12x - 14

-3x - 1

= -3x2 + 9x - 7 + -1- ∙ —⅛
-3 χ÷l
= -3x2 + 9x - 7 + ■ ~2-- .
-3x- 1
Notice that the denominator at the end is our original divisor. So, when we divide 9x3 - 24x2 + 12x - 14
by -3x - 1, the quotient is -3x2 + 9x - 7 and the remainder is -21. We check our answer and find

(-3x - 1)(-3x2 + 9x - 7) - 21 = 9x3 - 24x2 + 12x - 14,

so our quotient and remainder are correct.

We'll do a couple more for practice.
CHAPTER 6. POLYNOMIAL DIVISION

Problem 6.12: Perform the following divisions with synthetic division:

(a) (x3 — χ + 7)∕ (x + 3)
(b) (3y4-4y3+ 5y2 - lly+ 2)∕(2 + 3y)

Solution for Problem 6.12:

(a) What's wrong with this solution:

Bogus Solution: -3 1 -1 7
-3 12
1 -4 19
Our first clue that something is wrong is that our quotient is x — 4. When we divide a cubic by a
linear expression, we should get a quadratic as a quotient. So, we made a mistake, but where?
Our mistake was leaving out the 0x2 term in x3 - x + 7 in our division.
WARNING!! Be on the lookout for "missing terms" when doing synthetic division.
λ* They aren't really missing—the coefficients of these terms are 0. So,
we include Os for these coefficients in synthetic division.
The correct synthetic division is shown at right. We find ■3 1 0 -1 7
-3 9 -24
x3 - x + 7 = (x + 3)(x2 - 3x + 8) - 17.
So, the quotient is x2 - 3x + 8 and the remainder is -17.
1 -3 81 -17

(b) First, we have to adjust for the fact that the coefficient of y in the divisor is not 1. We have
3y4 - 4y3 + 5y2 - lly + 2 _ 1 3y4 - 4y3 + 5y2 - lly + 2
2 + 3y '^ 3 ’ y+ l

From our synthetic division at right, we have _2

3 3 -4 5 -11 2
3y4 - 4y3 + 5y2 - lly + 2 _ 1 3y4 - 4y3 + 5y2 - lly + 2 -2 34
4 -6 3
2 + 3y “ 3 ’
y+5 3 -6 9 -17 40
3
if, , 40/3)
= -1 3y3 - 6y2 + 9y -17 +-∕y
3jι y + 3/
3 n 2 n 17 1 40/3
= y3-2y2 + 3y-- + -∙-y
3 3 y+s
3 n 2 n 17 40/3
= y≈-2∕÷3y-- + -.

So, the quotient is y3 - 2y2 + 3y - y and the remainder is y.

6.3.1 Find the quotient and remainder when x3 + 4x2 - 5x + 19 is divided by x - 3.

184
 6.4. THE REMAINDER THEOREM

6.3.2 Check the two divisions we performed in Problem 6.12 by multiplying the quotient by the divisor,
then adding the remainder.
6.3.3 Find the quotient and remainder when x5 - 23x3 + llx2 - 14x + 20 is divided by x + 5.
6.3.4 Find the quotient and remainder when 4x4 - 10x3 + 14x2 + 7x - 19 is divided by 2x - 1.
6.3.5 Find the quotient and remainder when 3z3 - 4z2 - 14z + 3 is divided by 3z + 5.
6.3.6 Find the quotient and remainder when x4 + 3x3 - x2 + 7x - 1 is divided by 2 - x.
6.3.7 Stan divided x3 + 3x2 - 7x + 5 by 2x - 3 using synthetic division. He found a quotient of x2 + 6x +11
and a remainder of 38. By looking at his answer, he immediately realized that he had made a mistake.
How could he tell? What mistake did he probably make?

6.4 The Remainder Theorem

111 i______________
Problems

Problem 6.13: Let ∕(x) = x4 - 3x3 + 7x2 - x + 5.

(a) What is the remainder when ∕(x) is divided by x - 3?
(b) What is/(3)?

Problem 6.14: In this problem we prove the Remainder Theorem for polynomials. Suppose t∕(x) and
r(x) are the quotient and remainder, respectively, when the polynomial ∕(x) is divided by x - a, where
a is a constant.
(a) What is the degree of r(x)?
(b) Write an equation expressing ∕(x) in terms of q(x), r(x), and x- a.
(c) Use part (b) to show that ∕(α) = r(a).
(d) Why do (a) and (c) together tell us that the remainder when the polynomial ∕(x) is divided by
x - a is f(a)?

Problem 6.15: Let ∕(x) = x4 - 6x3 - 9x2 + 20x + 7. Use synthetic division to find ∕(7).

Problem 6.16: Find the remainder when 2x10 - 3zx8 + (1 + ∕)x2 - (3+ 2i)x + 1 is divided by x - i.

Problem 6.17: Let ∕(x) = x1° - 2x5 + 3.

(a) Find the remainder when ∕(x) is divided by x - 2.
(b) Find the remainder when ∕(x) is divided by 2x - 4.

Extra! Imagination is the beginning of creation. You imagine what you desire, you will what you imagine
l∣ Φ∣∣⅜l∣^lκ and at last you create what you will.
- George Bernard Shaw

185
CHAPTER 6. POLYNOMIAL DIVISION

Problem 6.18: Let P(x) be a polynomial such that when P(x) is divided by x - 19, the remainder is 99,
and when P(x) is divided by x - 99, the remainder is 19. In this problem we find the remainder when
P(x) is divided by (x - 19)(x - 99). (Source: AHSME)
(a) Find P(19) and P(99).
(b) Let r(x) be the desired remainder. Write an equation representing the division of P(x) by
(x - 19)(x - 99).
(c) What is the degree of r(x)?
(d) Choose two convenient values of x and substitute these into your equation from (b). Use the
results to find r(x).

Problem 6.19: Find the remainder when the polynomial x81 + x49 + x25 + x9 + x is divided by x3 - x.
(Source: Great Britain)

Problem 6.13: Let ∕(x) = x4 - 3x3 + 7x2 - x + 5.

(a) What is the remainder when ∕(x) is divided by x - 3?
(b) What is/(3)? I

Solution for Problem 6.13:

(a) We find the remainder using the synthetic division shown at 1 -3 7 -1 5
right. The remainder is 65. 3 0 21 60
1 0 7 20 65
(b) We have ∕(3) = 34 — 3 ∙ 33 + 7 ∙ 32 - 3 + 5 = 65.

Hmm.... We see that ∕(3) equals the remainder when ∕(x) is divided by x - 3. Is this a coincidence?

Problem 6.14: Suppose t∕(x) and r(x) are the quotient and remainder, respectively, when the poly- I
nomial ∕(x) is divided by x - α, where a is a constant. Is it always true that the remainder equals I
/w? I

Solution for Problem 6.14: Because q(x) and r(x) are the quotient and remainder, respectively, when the
polynomial ∕(x) is divided by x - fl, we have
∕(x) = (x - fl)ty(x) + r(x).
Because either the remainder is 0 or its degree is less than the degree of the divisor x - fl, we see that r(x)
must be a constant. So, we let r(x) = c for some constant c, and we have
∕(x) = (x - a)q(x) + c.
This equation must hold for all x, and we're interested in /(fl), so we let x = a in the equation. This gives
∕(fl) = (fl - fl)<∕(fl) + c = 0 ∙ <∕(fl) + c = c.
Therefore, when a polynomial ∕(x) is divided by x - fl, the remainder is /(fl). □

186
 6.4. THE REMAINDER THEOREM

Important: The Remainder Theorem states that when a polynomial ∕(x) is divided

ψ by x - a, the remainder is ∕(zz).

Solution for Problem 6.15: We can put x = 7 into ∕(x) to evaluate 7 1 -6 -9 20 7

∕(7), but the Remainder Theorem allows us to evaluate f(7) with less 7 7 -14 42
complicated arithmetic. The synthetic division at right tells us that 1 -2 ^6^] 49
1
the remainder when ∕(x) is divided by x - 7 is 49. Therefore, we have ∕(7) = 49. □
The Remainder Theorem can also be used to find remainders without division.

Problem 6.16: Find the remainder when 2x10 - 3ix8 + (1 + i)x2 - (3 + 2i)x + 1 is divided by x - i.
J
Solution for Problem 6.16: We could use synthetic division, even though some of our coefficients are
nonreal. However, because we're dividing a 10th degree polynomial, this synthetic division would be
pretty lengthy. Instead, we use the Remainder Theorem. We let ∕(x) = 2x10 - 3zx8 + (I + z)x2 - (3 + 2z)x +1.
By the Remainder Theorem, the remainder when we divide ∕(x) by x — z is

∕(z) = 2 ∙ z1° - 3z ■ z8 + (1 + z) ∙ z2 - (3 + 2z') ∙ i + 1

= 2(-l) - 3z(l) + (1 + z)(-l) - 3z - 2z2 + 1
= —7z.

□
Problem 6.17: Let ∕(x) = x1° - 2x5 + 3. Find the remainder when ∕(x) is divided by 2x - 4.

Solution for Problem 6.17: What's wrong with this solution:

Bogus Solution: The remainder is ∕(4) = 410 - 2 ∙ 45 + 3 = 1046531.

The Remainder Theorem tells us that when ∕(x) is divided by x - ar the remainder is f(a). The value
of ∕(4) equals the remainder when ∕(x) is divided by x - 4, not 2x - 4.

WARNING!! The Remainder Theorem applies to division by linear expressions

*A
* with leading coefficient 1.

While we can't directly apply the Remainder Theorem, we can use our proof of the Remainder
Theorem as guidance. Because we are dividing by a linear expression, the remainder must be a constant.
So, we seek the constant r such that
∕(x) = (2x - 4)zj(x) + r

187
CHAPTER 6. POLYNOMIAL DIVISION

for all values of x. When proving the Remainder Theorem, we eliminated q(x) by setting x equal to a
value that made the divisor 0. Here, we let x = 2 to find ∕(2) = (2 ∙ 2 - 4)g(2) + r = r. So, our remainder
is ∕(2) = 210 - 2 ■ 25 + 3 = 963. □
Notice that the key step in this solution is not to apply the Remainder Theorem directly, but rather
to use the same general technique we used to prove the Remainder Theorem.

Concept: Don't just memorize important theorems. Learn how to prove them. The
Q—pe-) methods used to prove important theorems can also be used to solve other
problems to which it is not easy to directly apply these theorems.________

Here, the technique we use is:

Concept: Suppose tj(x) and r(x) are the quotient and remainder, respectively, when
Q—∙lλj the polynomial ∕(x) is divided by the polynomial d(x). One powerful way
to learn more about r(x) is to express the division as

∕(x) = q(x) ∙ d(x) + r(x),

and then eliminate the q(x) ■ d(x) term by choosing values of x such that
d(x) = 0. Combining this with the fact that degr is no greater than degd
will often allow us to determine r(x).

This strategy works even when d(x) is not linear.

Problem 6.18: Let P(x) be a polynomial such that when P(x) is divided by x - 19, the remainder is 99,
and when P(x) is divided by x - 99, the remainder is 19. What is the remainder when P(x) is divided
by (x - 19)(x - 99)? (Source: AHSME)

Solution for Problem 6.18: Because we are dividing by a quadratic, the degree of the remainder is no
greater than 1. So, the remainder is ax + b, for some constants a and b. Therefore, we have

P(x) = (x - 19)(x - 99)Q(x) +ax + b,

where Q(x) is the quotient when P(x) is divided by (x - 19)(x - 99). We can eliminate the Q(x) term by
letting x = 19 or by letting x = 99. Doing each in turn gives us the system of equations

P(19) = 19« + b = 99,

P(99) = 99a + b = 19.

Solving this system of equations gives us a = — 1 and b = 118. So, the remainder is —x + 118. □

Problem 6.19: Find the remainder when the polynomial x81 + x49 + x25 + X9 + x is divided by x3 - x. I
: (Source: Great Britain) ∖

Solution for Problem 6.19: Let P(x) = x81 + x49 + x25 + x9 + x and let Q(x) be the quotient when P(x)
is divided by x3 - x. We are dividing by a cubic, so the degree of the remainder is no more than 2.

188
 6.5. SUMMARY

Therefore, the remainder is ax2 + bx + c for some constants a, b, and c. So, we have
P(x) = (x3 - x)Q(x) + ax2 + bx + c.

Since x3 - x = x(x + l)(x -1), we can let x = -1,0, or 1 to make x3 — x equal to 0. Doing each in turn gives
us the system of equations

P(-l) = a-b + c,
P(0) = c,
P(l) =a + b + c.
Since P(0) = 0, we have c = 0. Adding the first and third equations then gives 2a = P(-l) + P(l) =
-5 + 5 = 0, so a = 0. Finally, from P(l) = a + b + c, we have b = P(l) = 5, so the desired remainder is 5x. □

Exercises

6.4.1 Let g(x) = 2x6 — x4 + 4x2 - 8. Find the remainder when g(x) is divided by each of the following:
(a) x+1 (c) x+3 (e) x2 + 4x + 3
(b) x-1 (d) x-3 (f) x2 - 4x + 3

6.4.2 When y2 + my+2 is divided by y -1, the quotient is ∕(y) and the remainder is R↑. When y2 + my + Z
is divided by y+1, the quotient is g(y) and the remainder is ¾∙ If Pi = R2, then find m. (Source: AHSME)
6.4.3 Find the remainder when x10° is divided by (x — l)(x - 2). (Source: AHSME) (You can leave
constants in the form ab in your answer.)
6.4.4 Suppose q(x) and r(x) are the quotient and remainder, respectively, when the polynomial ∕(x) is
divided by the polynomial d(x). Show that if x = a is a root of d(x), then r(α) = /(a).
6.4.5
* Find the remainder when x10° - 4x98 + 5x + 6 is divided by x3 - 2x2 - x + 2. Hints: 91

6.5 Summary

A polynomial in one variable is an expression or function of the form

anχn + Λn-'[Xn 1 + flw-2Xπ 2 + ∙ ∙ ∙ ÷ ∕7']X + Λq,

where x is a variable and each α,∙ is a constant. Each individual summand αl∙χ, is called a term, and each
number cy is the coefficient of the corresponding x,. The degree of a polynomial is the greatest integer n
such that the⅛oefficient an is not equal to 0. (The degree of the polynomial 0 is undefined.)

Important: When the polynomial ∕(x) is divided by the polynomial d(x), the quotient
t∕(x) and remainder r(x) are the polynomials such that

∕(x) = q(x) ■ d(x) + r(x)

and either r(x) = 0 or deg r < deg d.

189
CHAPTER 6. POLYNOMIAL DIVISION

We covered a method for performing polynomial division that is a lot like "long division" of integers.
We then also covered a shorthand, called synthetic division, for dividing a polynomial by a linear
polynomial in which the linear term has coefficient 1. We can also use synthetic division even if the
coefficient of the linear term of the divisor is not 1, by converting the division into a division in which
this coefficient is 1.

Important: The Remainder Theorem states that when a polynomial ∕(x) is divided

ψ by x - a, the remainder is f(a).

Things To Watch Out For!________

• When dividing one polynomial by another, be careful not to overlook "missing" terms. These
terms aren't really missing—their coefficients are 0. Include such terms in your long division by
writing them with a coefficient of 0.

• When we divide by x - fl using synthetic division, we put a to the left of the division bracket, not
-fl. So, when we divide by x + c, we put —c to the left of the division bracket, not c.

• Synthetic division only works when the coefficient of the linear term in the divisor is 1.

Problem Solving Strategies---------------------------------------------------------------

Concepts: ∙ Substituting a value for the variable in a polynomial can give us

θ o~j useful information about the coefficients of the polynomial. In par­
ticular, letting the variable equal 1 gives the sum of the coefficients
of the polynomial.

• Just as you can check integer division with multiplication and addi­
tion, you can check polynomial division. If ∕(x) divided by d(x) gives
a quotient of ty(x) and remainder of r(x), we must have

d(x) ∙ q(x) + r(x) = ∕(χ).

• Don't just memorize important theorems. Learn how to prove them.

The methods used to prove important theorems can also be used to
solve other problems to which it is not easy to directly apply these
theorems.

Continued on the next page. . .

190
 REVIEW PROBLEMS

Concepts: . . . continued from the previous page

θ uju^, ∙ Suppose q(x) and r(x) are the quotient and remainder, respectively,

when the polynomial f(x) is divided by the polynomial d(x). One

powerful way to learn more about r(x) is to express the division as

∕(x) = <7(x) ∙ d(x) + r(x),

and then eliminate the q(x) ■ d(x) term by choosing values of x such
that d(x) = 0. Combining this with the fact that deg r is no greater
than degd will often allow us to determine r(x).

Review Problems

6.20 Let f (x) - x2 + 3x + 3 and g(x) = 2x2 -x + 2. Find all constants c such that h(x) = f(x) + c ■ g(x) is a
binomial.

6.21 Find the quotient and remainder in each of the following divisions:
(a) 9x4 + x3 — 12x + 21 divided by x + 4.
(b) x3 - 3x2 — 9x + 27 divided by x - 3.
(c) 2y6 + 3y4 + 4y2 + 5 divided by y2 — 1.
(d) 8x3 + 16x2 — 7x + 4 divided by 5 + 2x.
(e) 3r4 + 16r3 — 5r + 19 divided by 3r - 2.
(f) t4 - t2 +3t — 7 divided by t2 -3t + 8.
6.22 If P(x) is a polynomial in x, and x23 + 23x17 - 18x16 - 24x15 + 108x14 = (x4 - 3x2 - 2x + 9) ∙ P(χ) for
all values of x, compute the sum of the coefficients of P(x). (Source: ARML)
6.23 In Exercise 6.4.1, you should have found that g(x) = 2x6 - x4 + 4x2 - 8 leaves the same remainder
when divided by x + 1 and by x - 1, and that g(x) leaves the same remainder when divided by x + 3 and
by x - 3. Is it true that for every polynomial ∕(x), the remainder when ∕(x) is divided by x - a is the same
as when it is divided by x + a, for any constant a?
6.24 Beth divided 2x3 - x + 5 by x + 3 and came up with a quotient of 2x + 7 and a remainder of 26. She
immediately knew from her answer that she had made a mistake. How did she know so quickly that
she had erred? Can you figure out what mistake she made?
6.25 Find the remainder when x13 + 1 is divided by x - 1. (Source: AHSME)

6.26 Suppose the polynomial f(x) is divided by the polynomial d(x) to give a quotient q(x) and remain­
der r(x). If deg f = 12 and deg r = 2, then what are the possible values of deg q?
6.27 If x4 + 4x3 + βpx2 + 4t?x + r is exactly divisible by x3 + 3x2 + 9x + 3, then find (p + q)r. (Source:
AHSME)

191
CHAPTER 6. POLYNOMIAL DIVISION

6.28 Find the remainder when 13x6 + 3x4 + 9x3 + 2x2 + 17 is divided by x2 - 1.

6.29 ∕(x) is a polynomial of degree greater than 3. If ∕(1) = 2, ∕(2) = 3 and ∕(3) = 5, find the remainder
when ∕(x) is divided by (x - l)(x - 2)(x - 3). (Source: ARML)
6.30 Show that if the remainder when ∕(x) is divided by x - « is r, then the remainder when ∕(x) is
divided by c(x — a) is also r for any nonzero constant c.

ι
Challenge Problems

6.31 Let ∕(x) = x2 - 3x + 3 and g(x) = 3x3 - 5x2 - 3x +12. The solutions of the equation ∕(x) = 0 are also
solutions of the equation g(x) - 0. The equation g(x) = 0 has a third solution besides the two solutions
of ∕(x) = 0; what is this third solution?
6.32 Show that if a polynomial ∕(x) leaves a remainder of the form px + q when it is divided by
(x - rt)(x - b)(x - c), where a, b, and c are all distinct, then (b-c)f(a) + (c-a)f(b) + (a- b)f(c) = 0. Hints: 20
6.33 How can we use synthetic division to divide x4 - 3x3 + 4x2 + llx - 9 by x2 — 3x + 2? Hints: 162

6.34
* When P(x) = x81 + Lx57 + Gx41 + Hx19 + 2x + 1 is divided by x - lz the remainder is 5, and when
P(x) is divided by x — 2, the remainder is —4. However, x81 + Lx57 + Gx41 + Hx19 + Kx + R is exactly
divisible by (x - l)(x - 2). If L, G, H, K, and R are real, compute the ordered pair (K, R). (Source: NYSML)
6.35
* Let P(x) = (x - l)(x - 2)(x - 3). For how many polynomials Q(x) does there exist a polynomial
R(x) of degree 3 such that P(Q(x)) = P(x) ∙ R(x)? (Source: AMC12) Hints: 231,356, 224
6.36
* Find the remainder when the polynomial x81 + x49 + x25 + x9 + x is divided by x3 + x. Note: This
is not the exact same problem as Problem 6.19! Hints: 317
6.37
* Find a polynomial ∕(x) of degree 5 such that ∕(x) - 1 is divisible by (x - 1)3 and ∕(x) is itself
divisible by x3. Hints: 286, 267

Extra! The ancient Greeks made many geometric discoveries, but their algebraic accomplish-
∣ 0-∣∣∣∣>∣ < l∣ ∙> ments were much more limited. One of the main reasons they didn't make much progress
with algebra is that they had not developed sophisticated notation like the notation we
now use. Indeed, much of their algebraic reasoning was done in geometric terms, such
as the following translation of Euclid's Elements:

To a given straight line to apply a parallelogram equal to a given rectilineal figure

and deficient by a parallelogrammic figure similar to the given one: thus the given
rectilineal figure must not be greater than the parallelogram described on the half of
the straight line and similar to the defect.

In his book Unknown Quantity about the history of algebra, John Derbyshire referred to
this passage when he wrote, "This, say Bashmakova and Smirnova, is tantamount to
solving the quadratic equation x(a - x) = S. I am willing to take their word for it." Me
too.

192
When solving problems, dig at the roots instead of just hacking at the leaves. - Anthony J. D'Angelo

CHAPTER 7
Polynomial Roots Part I

The number r is a root of the polynomial ∕(x) if ∕(r) = 0. Roots of polynomials are also sometimes called
zeros of the polynomial. In this chapter, we investigate tactics for finding rational roots of polynomials.

7.1 The Factor Theorem

Recall that a polynomial d(x) evenly divides a polynomial f(x) if the remainder is 0 when we divide ∕(x)
by d(x). We then call d(x) a factor of ∕(x). In this section, we use the Remainder Theorem from Section 6.4
to find linear factors of polynomials.

Problems

Problem 7.1: Let ∕(x) = 6x3 - tx2 + 7x- 12, where t is a constant such that ∕(x) is divisible by x - 2.
(a) Find ∕(2) in terms of t. (b) Find t.

Problem 7.2: Let p(x) be a polynomial.

(a) Show that if x - a is a factor of p(x), then p(a) = 0.
(b) Show that if p(α) = 0, then x - a is a factor of p(x).

Problem 7.3: The polynomial ∕(x) = x4 + ax3 + bx2 + ex + d has roots 1, 3, 5, and 7. Determine all the
coefficients of ∕(x).

Problem 7.4: Let g(x) and ∕(x) be nonzero polynomials.

(a) If g(x) is a factor of ∕(x), then must every root of g(x) also be a root of /(x)?
(b) If every root of g(x) is a root of ∕(x), then must g(x) divide ∕(x) evenly?

193
CHAPTER 7. POLYNOMIAL ROOTS PART I

Problem 7.5: Let p(x) = 3x4 + 5x3 - 49x2 + llx + 30 and note that p(-2∕3) = p(3) = 0.
(a) Find two linear factors of p(x). Divide p(x) by one of these factors. Why are the roots of the
quotient also roots of p(x)?
(b) Find all four roots of p(x).

Problem 7.1: Find the constant t such that the polynomial 6x3 - tx2 + 7x - 12 is divisible
■■■■■■■■■■■■■■■■■■■■■■■■■■■I
byx^2∙ I
Solution for Problem 7.1: We let ∕(x) = 6x3 - tx2 + 7x- 12. If ∕(x) is divisible by x - 2, then the remainder
is 0 when ∕(x) is divided by x - 2. By the Remainder Theorem, if the remainder when f (x) is divided by
x - 2 equals 0, then we must have ∕(2) = 0.
Since ∕(2) = 0, we let x = 2 in our definition of ∕(x) and we have 6 ∙ 23 — f ∙ 22 + 7 ∙ 2 — 12 — 0. Solving
for t gives us t = 12.5.

At right, we confirm that x - 2 divides ∕(x) evenly when we have 2 6 -12.5 7 -12
t = 12.5. We find that 12 --1 12
6 -0.5 6 ∣ 0
6x3 - 12.5x2 + 7x - 12 = (x - 2)(6x2 - 0.5x + 6),

so x — 2 is a factor of 6x3 - 12.5x2 + 7x - 12.

Not only can we use synthetic division to test our solution, 2 6 -t 7 -12
but we could also have used it to find the answer. At right, we 12 24-2f 62-4f
synthetically divide x - 2 into ∕(x) = 6x3 - tx2 + 7x-12. In order 6 12 -t 31-2f ∣ 50 - 4t
for x - 2 to divide ∕(x) evenly, the remainder must be 0. Solving
50 - 4f = 0 gives us t = 12.5, as before. □

' Problem 7.2: Let p(x) be a polynomial. Prove that x - a is a factor of p(x) if and only if p(a) = 0.

Solution for Problem 7.2: We begin by proving that if x - α is a factor of p(x), then p(a) = 0.
If x - a is a factor of p(x), then p(x) = (x - tz)t∕(x) for some polynomial q(x). Letting x = a gives us
p(a) = (a- d)q(a) = 0 ■ q(a) = 0, so a is a root of p(x).
Next, we prove that if a is a root of p(x), then x - a is a factor of p(x).
If p(μ) is a root of p(x), then p(a) = 0. By the Remainder Theorem, the remainder when p(x) is divided
by x - a is p(a). Since p(a) = 0, this remainder is 0. This means that x - a is a factor of p(x). □
Our solution to Problem 7.2 gives us a proof of the Factor Theorem:

Important: The Factor Theorem states that the expression x-a, where a is a constant,
is a factor of the polynomial p(x) if and only if p(q) = 0.

194
 7.1. THE FACTOR THEOREM

This theorem tells us that factors and roots of polynomials go hand in hand. Since the polynomial
p(x) has the factor x - a if and only if p(a) = 0, we know each of the following:

• If x - a is a factor of p(x)r then p(fl) = 0.

• If p(a) = 0, then x - fl is a factor of p(x).

In other words, if a is a constant and p(x) is a polynomial, then the statements "x - fl is a factor of
p(x),' and "p(fl) = 0" mean the same thing.

As you might have suspected from its name, the Factor Theorem can be used to help us factor
polynomials if we know their roots.

Problem 7.3: The polynomial ∕(x) = x4 + ax3 + bx2 + ex + d has roots 1,3,5, and 7. Determine all the ∣
coefficients of ∕(x).

Solution for Problem 7.3: From the roots of ∕(x), we note that x - 1, x - 3, x - 5, and x - 7 are all factors
°f ∕(x)∙ But are these the only ones? We see that they are the only factors by working through the roots
one at a time.

Because 1 is a root of ∕(x), we have ∕(x) = (x - l)gι(x), for some polynomial gι(x). Since the leading
term of ∕(x) is x4, the leading term of gι(x) must be x3. Next, because 3 is a root of ∕(x), we have
∕(3) = 0. This means that (3 — l)gι(3) = 0, so gι(3) = 0, which means 3 is a root of gι(x). This means that
gi(x) = (x — 3)g2(x), where g2(x) is a monic quadratic polynomial.
We now have ∕(x) = (x-l)(x-3)g2(x). Since 5 is a root of ∕(x), wehave(5-l)(5-3)g2(5) = 0,sog2(5) = 0
and g2(x) = (x — 5)g3(x) for some linear polynomial g3(x). Finally, because ∕(x) = (x - l)(x - 3)(x - 5)g3(x)
and 7 is a root of ∕(x), we find that g3(7) = 0, which means that g3(x) is the linear polynomial x - 7. So,
we can now deduce that

∕(x) = (x - l)(x - 3)(x - 5)(x - 7)

= (x2 - 4x + 3)(x2 - 12x + 35)
= x2(x2 - 12x + 35) - 4x(x2 - 12x + 35) + 3(x2 - 12x + 35)
= x4 - 16x3 + 86x2 - 176x + 105.

Notice that the coefficient of x3 is -(1 +3 + 5 + 7) and the constant term is 1 ∙ 3 ∙ 5 ∙ 7. We'll be exploring
this "coincidence" in Section 8.3. Maybe you can figure out before then why this happens. □
In general, if we know all the roots of a polynomial, we don't have to go through the long-winded
process of our solution to Problem 7.3 to write the polynomial as a product of the corresponding factors.
We can follow essentially the same steps to show the following:

Important: Let the polynomial ∕(x) = anxn + flπ-ιxn 1 + flπ-2xπ 2 + ∙ ∙ ∙ + axx + a0 have

ψ roots rι, r2, r3, ...frn. Then we can write ∕(x) as

∕(x) = a„(x - r1)(x - r2) ∙ ∙ ∙ (x - rπ).

195
CHAPTER 7. POLYNOMIAL ROOTS PART I

We can use this to show that a polynomial with degree n cannot have more than n roots. If a
polynomial ∕(x) has n + 1 roots, then it can be written as

∕(x) = c(x - ∏)(x - r2) ∙ ∙ ∙ (x - rn+1),

where c is a constant and rj, r2, r^,..., rtl+↑ are the roots. There are n + 1 roots, so there are n + 1 linear
factors. When this product is expanded, the highest degree term is formed by taking x from each of the
n + 1 linear factors, which gives us the term cx"+1. Therefore, the degree of f is n + 1, not n. Similarly,
any polynomial with more than n roots has degree greater than n, so no polynomial with degree n can
have more than n roots.
With tools more advanced than we will cover in this text, we can show that every one-variable
polynomial with degree n has exactly n roots. This fact is so fundamental to the study of algebra that it
is called the Fundamental Theorem of Algebra.

Important: The Fundamental Theorem of Algebra states that every one-variable

ψ polynomial of degree n has exactly n complex roots.

You can get a sense for why this theorem is true by noting that multiplying n one-variable linear
terms will produce a polynomial with degree n. The Fundamental Theorem of Algebra tells us that this
process runs in reverse as well—if ∕(x) is a polynomial with degree n, then we can write ∕(x) as

∕(x) = α,1(x - r1)(x - r2) ∙ ∙ ∙ (x - rn),

where an is a constant and the constants r1, r2, r3, ...,rn are the roots of the polynomial. These roots need
not be distinct. For example, the polynomial (x - 1)3 has a roots 1,1, and 1. That is, 1 is a triple root of
(x - 1)3. We therefore say that the root 1 of (x - 1)3 has multiplicity 3.

We have already seen that the Factor Theorem tells us that if g(x) is a linear polynomial, then g(x) is a
factor of the polynomial ∕(x) if and only if the root of g(x) is a root of ∕(x). But what if g(x) is not linear?

Problem 7.4: Let g(x) and ∕(x) be nonzero polynomials.

(a) If g(x) is a factor of ∕(x), then must every root of g(x) also be a root of /(x)?
(b) If every root of g(x) is a root of ∕(x), then must g(x) divide ∕(x) evenly?

Solution for Problem 7.4:

(a) If g(x) is a factor of ∕(x), then we have ∕(x) = g(x)<f(x) for some polynomial z∕(x). Therefore, if r is a
root of g(x), then we have ∕(r) = g(f)q(r) = 0 ∙ q(f) = 0,sor is also a root of ∕(x). This means that if
g(x) is a factor of ∕(x), then every root of g(x) is a root of ∕(x).
(b) Suppose g(x) = (x - 1)2 and ∕(x) = x(x - 1). Then, every root of g(x) is also a root of ∕(x), but g(x)
does not evenly divide ∕(x). Therefore, the converse of part (a) is not true in general.

Important: If g(x) is a factor of ∕(x), then every root of g(x) is a root of ∕(x).

ψ
196
 7.1. THE FACTOR THEOREM

Let's see an example of why this fact is useful.

Problem 7.5: Let p(x) = 3x4 + 5x3 - 49x2 + llχ + 30 and note that p(-2∕3) = p(3) = 0. Find all four
roots of p(x).

Solution for Problem 7.5: Since p(-2∕3) = p(3) = 0, the Factor Theo­ 3 5 -49 11 30
rem tells us that x + 2/3 and x — 3 are factors of p(x). Now, we can 9 42 -21 -30
use synthetic division to help us factor p(x). We first divide ρ(x) by 3 14 -7 -10 I 0
x - 3 (at right) to get 3x3 + 14x2 - 7x - 10. Therefore, we have

p(x) = (x - 3)(3x3 + 14x2 - 7x - 10).

Since 3x3 + 14x2 — 7x -10 is a factor of p(x), the roots of 3x3 + 14x2 - 7x —10 are also roots of p(x). Therefore,
we can continue our search for roots of p(x) by finding roots of 3x3 + 14x2 — 7x — 10.

We don't have to search far because we know that x + 2/3 is a

K
-2/3 3 14 -10

1
factor of p(x). So, we divide 3x3 + 14x2 - 7x - 10 by x + 2/3 (at right)

oo
-2 10

1
to get 3x2 + 12x - 15. We now have 3 12 -15 ∣ 0

p(x) = (x - 3)(x + 2∕3)(3x2 + 12x - 15).

All three coefficients of the quadratic are divisible by 3, so we factor out a 3 from the quadratic to find

p(x) = 3(x - 3)(x + 2∕3)(x2 + 4x - 5).

Concept: If all the coefficients of a polynomial you must factor are divisible by the
Q- oe) same integer (other than 1), then factor that integer out so you can work
with a simpler polynomial.

We finish factoring p(x) by factoring the quadratic x2 + 4x - 5 as (x + 5)(x - 1), so we have

p(x) = 3(x - 3)(x + 2∕3)(x + 5)(x - 1).

Therefore, the roots ofp(x) are -5, -2/3, 1, and 3. Note that because 3(x + 2/3) = 3x + 2, we can also write
the factorization as p(x) = (x — 3)(3x + 2)(x + 5)(x - 1). □

Important: If we find that a is a root of the polynomial p(x), we can simplify our

ψ search for more roots of p(x) by dividing p(x) by x - a. Since a is a root of

p(x), this division will leave no remainder, and we will have

p(x) = (x - α)t∕(x).

We can then continue our search for roots of p(x) by searching for roots
of t∕(x).

197
CHAPTER 7. POLYNOMIAL ROOTS PART I

Ixer<i
*es

7.1.1 For what values of a is x + 2 a factor of x4 - 2x3 + ax2 -ax+ 7?

7.1.2 How can we quickly tell that x - 1 is a factor of x5 + 6x4 - 7x3 + 2x2 - 2 without performing the long
division?
7.1.3 Let ∕(x) = 6x4 + 19x3 - 36x2 - 49x + 60. One root of ∕(x) is -5/3. Find all four roots of ∕(x).
7.1.4 Suppose ∕(x) and <j(x) are polynomials such that ∕(x) = (x-a)q(x). Is it true that t∕(x) = 0 whenever
∕(x) = 0?
7.1.5 The polynomial p(x) = 3x3 - 202 + kx + 12 is divisible by x - 3 for some constant k. Factor p(x)
completely.

7.2 Integer Roots

Finding roots of polynomials can be extremely difficult. However, if we are looking for integer roots
of a polynomial with integer coefficients, we can use the coefficients of the polynomial to considerably
narrow our search.
In the next three sections, we'll explore methods for finding rational roots of polynomials. In
Chapter 8, we'll study polynomials that have irrational or nonreal roots.

Problems

Problem 7.6: In this problem we find the roots of ∕(x) = x3 - 5x2 - 77x + 441.
(a) Consider /(2):
∕(2) = 23 - 5(22) - 77(2) + 441.

How can we determine that ∕(2) ≠ 0 without evaluating /(2)?

(b) Consider /(5):
∕(5) = 53 - 5(52)-77(5)+ 441.

Which of the four terms on the right are divisible by 5? Which are not? Is 5 a root of /(x)?
(c) Suppose that n is an integer that is a solution to the equation ∕(n) = 0. Show that n is a divisor
(positive or negative) of 441.
(d) Find the roots of ∕(x).

Problem 7.7: Suppose k is a root of ∕(x) = anxn + an-ιXn~i +an-2Xn~2Λ------t-αιx + ⅜z where k is a nonzero
integer, and all the coefficients of ∕(x) are integers. Why must k divide ⅜7

Extra! The polynomial f(ri) = n2 + n + 41 cannot be factored as the product of two polynomials
"*- ι*"" with integer coefficients. Evaluate ∕(l)z ∕(2), ∕(3), and ∕(4). Do you notice anything
interesting about the integers that result? l∣ so, explore further. (Continued on page 209.)

198
 7.2. INTEGER ROOTS

Problem 7.8: Let ∕(x) = x3 - 3x2 - 10x + 24.

(a) Graph ∕(x).
(b) Notice that ∕(0) = 24 and ∕(3) = -6. Explain why there must be a root between 0 and 3.
(c) Notice that ∕(1) = 12 and ∕(5) = 24. Does this mean that there is not a root between 1 and 5?
(d) Find the roots of ∕(x).

Problem 7.10: Let g(χ) = χ6 + 15x5 + 81x4 + 193x3 + 198x2 + 72x.

(a) How can we quickly see that 0 is a root of g(x)?
(b) How can we quickly see that g(x) has no positive roots?
(c) Find the roots of g(x).

Problem 7.11: In this problem, we solve the inequality x3 + 5x2 — 16x — 80 ≤ 0.

(a) Factor the left side.
(b) Find all x that satisfy the inequality.

Problem 7.6: Find all solutions to the equation x3 - 5x2 - 77x + 441 = 0.

Solution for Problem 7.6: Let ∕(x) = x3 - 5x2 - 77x + 441. We wish to find the roots of ∕(x). We could just
go hunting for roots by evaluating f(a) for various values of a. We'll start with integer values of a, since
they're easier to work with than fractions:
∕(0) = 02 - 5(02) - 77(0) + 441 = 441,
∕(1) = 13 - 5(12) - 77(1) + 441 = 360,
∕(2) = 23 - 5(22) - 77(2) + 441 = 275,
∕(3) = 33 - 5(32) - 77(3) + 441 = 192,
∕(4) = 43 - 5(42) - 77(4) + 441 = 117,
∕(5) = 53 - 5(52) - 77(5) + 441 = 56.
Hmmm.... This search might take a very long time. Let's see if we can find more clever method. First,
we note that ∕(0), ∕(2), and ∕(4) are all odd. Is this just a coincidence, or will we always get an odd
output when we put an even number in for x?

Suppose x is an even integer. Then, the first three terms of ∕(x) are even, but the last term, 441, is
odd. Therefore, the sum of these four terms is odd. So, because 0 is even and ∕(x) is odd for even x, we
cannot have ∕(x) = 0 if x is an even integer. This tells us that ∕(x) cannot have a root that is divisible by 2.
Considering values of x that are divisible by 2 narrowed our search for roots. Let's look at other
possible divisors of x.

199
CHAPTER 7. POLYNOMIAL ROOTS PART I

Concept: If a strategy gives you some information about a problem, but doesn't
completely solve it, then try that strategy again in a slightly different way.

If x is a multiple of 3, then the first three terms of ∕(x) are divisible by 3. The constant term, 441, is
also divisible by 3. Therefore, if x is divisible by 3, then all four terms of ∕(x) are divisible by 3. So, it is
possible to have ∕(x) = 0 if x is divisible by 3.
That didn't help too much. Let's try multiples of 5. (We skip multiples of 4 because we already know
that the roots can't be even.) If x is divisible by 5, then all terms of x3 - 5x2 - 77x + 441 except the constant
term are divisible by 5. Therefore, if x is divisible by 5, then ∕(x) equals a multiple of 5 plus 441. This
sum cannot be divisible by 5, so ∕(x) cannot equal 0 if x is a multiple of 5.
We don't have to try multiples of 6 since there are no even roots. If x is divisible by 7, then all four
terms of ∕(x) are divisible by 7. So, it is possible to have ∕(x) = 0 if x is divisible by 7.
We have a pattern emerging. In each case, we let x be divisible by some positive integer d. If d does
not divide 441 (such as when d = 2 or d = 5), then it seems that we cannot have ∕(x) = 0. If d does divide
441 (such as when d — 3 or d = 7), then it seems that we cannot eliminate the possibility that f (x) = 0.
But does the pattern hold in general?
We suspect our hunt for integer solutions of x3 - 5x2 - 77x + 441 = 0 depends on whether or not 441/x
is an integer. So, we isolate 441 in this equation, then divide by x (clearly, x = 0 is not a solution, so we
can divide by it). This gives us
441 2 r- r7r7
---- = -x2 + 5x + 77.
x
All three terms on the right are integers when x is an integer, so 441/x must be an integer if the integer x
is a solution to the equation. Therefore, if ∕(x) = 0 for an integer x, then x must evenly divide 441.

This narrows our search for integer roots a great deal. Since 6 doesn't 7 1 -5 -77 441
divide 441 but 7 does, we try dividing ∕(x) by x - 7 next to see if 7 is 7 14 -441
a root of ∕(x). We find that x - 7 divides ∕(x) evenly, and we have 1 2 -63 ∣ 0
f(x) = (x - 7)(x2 + 2x - 63). Factoring the quadratic gives us

∕(x) = (x - 7)(x - 7)(x +9).

So, there are two solutions to the equation ∕(x) = 0, namely x = 7 and x = -9. Because ∕(x) has two
factors equal to x - 7, we say that 7 is a double root of ∕(x). So, we say that ∕(x) has 3 roots, which are 7,
7, and -9, but only 2 distinct (different) roots, which are 7 and -9. □
The number of times a number is a root of a polynomial is called the multiplicity of the root. So, the
root 7 of ∕(x) = (x - 7)(x - 7)(x + 9) has multiplicity 2, while the root -9 has multiplicity 1.
Our work in Problem 7.6 suggests that any nonzero integer root of a polynomial with integer
coefficients must divide the constant term of the polynomial. Let's see if this is the case.

Problem 7.7: Suppose x = k is a root of ∕(x) = allxn + ⅜-ιx"^1 + a,l-2Xn~2 + ∙ ∙ ∙ + *hx + ⅜∕ where k is a '
nonzero integer, and all the coefficients of ∕(x) are integers. Why must k divide «0?

200
 7.2. INTEGER ROOTS

Solution for Problem 7.7: We take essentially the same approach we took in Problem 7.6. If x = k is a root
f(χ)∕ then we must have ank', + all~↑kll~i + an-2kn~2 + ∙ ∙ ∙ + a↑k + ao = 0. Since k is a nonzero integer, we
can divide by k to find
⅜F^1 + an-ikn~2 + an-2kn~3 + ∙∙∙+a↑ + ^- =0.
k
Isolating the fraction then gives us

— — —n P7~l ∕ι l∏-2 n t-n~3 n

RrιK Cln-'[K. ⅞rι-2^∙ * ’ ’ ^1∙

Since k and all the coefficients are integers, the right side is an integer. Therefore, ao/k must be an integer,
which tells us that k divides ao. So, if k is a root of ∕(x), then it must divide the constant term of f(x).
We can also show that ao is a multiple of k by isolating «0 hi

a∏k + tzπ-jJcπ 1 + an-2kn 2 + ∙ ■ ∙ + a↑k + λq = 0.

This gives us a0 = -(ankn + an^kn~1 + an-2kn~2 + ■■■+ a1k) = -k(anktl~1 + an.1kn~2 + an-2kn~3 + ∙ ∙ ∙ + a1).
Because all the coefficients and k are integers, the expression (ankn~^ + an--∣kn~2 + an-2kn~3 4------ 1- a↑) is an
integer. Therefore, ao equals k times an integer, which means ao is a multiple of k. □
We indicate that a number d divides a number n by writing d ∣ n.

Important: Let ∕(x) - anxn + an-∖Xn 1 + an-2Xn 2 + ∙ ∙ ∙ + «0, where the coefficients of

ψ ∕(x) are integers. If r is a nonzero integer root of f(x), then r ∣ ao.

This fact considerably reduces the number of integers we must try when searching for roots of a
polynomial with integer coefficients. If f(x) is a polynomial with integer coefficients and k is an integer
that does not divide the constant term of ∕(x), then we know that x - k does not divide ∕(x).
Now, we'll get some practice finding integer roots of polynomials, and find a few more strategies for
finding roots quickly.

Problem 7.8: Let ∕(x) = x3 - 3x2 - 10x + 24. Graph ∕(x) and find its roots.

Solution for Problem 7.8: If ∕(x) has any integer roots, each of these roots must evenly divide 24, the
constant term of ∕(x). We start with 1 and -1 because it's so easy to evaluate ∕(1) and ∕(-l).

Concept: We start our search for nonzero integer roots of any polynomial ∕(x) with
Q- l,j 1 and -1, since it's so easy to evaluate ∕(1) and ∕(-l).

Unfortunately, neither of these work, as we find ∕(1) = 12 and ∕(-l) = 30. So, we must continue our
search. Since 3 divides 24 evenly, we check if 3 is a root. (We see that x3 - 3x2 = 0 for x = 3, so evaluating
∕(3) is particularly easy.) We find that ∕(3) = -6, so 3 is not a root. However, we notice that ∕(3) < 0
while ∕(1) > 0. Because ∕(3) is negative and ∕(1) is positive, we guess that there is a value between 1
and 3 that is a root. Are we right?

201
CHAPTER 7. POLYNOMIAL ROOTS PART I

The only integer between 1 and 3 is 2, and it divides 24 evenly, so it 2 1 -3 -10 24

could be a root. The synthetic division at right shows that 2 is indeed a 2 -2 -24
root, and we have ∕(x) = (x - 2)(x2 -x -12). Factoring the quadratic gives 1 -1 -12] 0
us y^(χ) = (χ - 2)(x + 3)(x - 4), so the roots are 2, -3, and 4.
We can use the graph of f(x) to see why our guess that there is a root between 1 and 3 was correct.
The roots of ∕(x) tell us where the graph of ∕(x) meets the x-axis, and we choose a few more values of x
to fill out the graph. Also, ∕(x) is much greater than 0 for large values of x, and is much less than 0 when
x is much less than 0. This information gives us a good idea what the graph looks like.
y
X ∕(*)
-4 -48
-3 0
-2 24
-1 30
0 24
1 12
2 0
3 -6
4 0
5 24

The graph passes through (1,12) and (3, -6). So, as the graph goes from x = 1 to x = 3, it goes from
above the x-axis to below the x-axis. Somewhere in between, it must cross the x-axis. Therefore, there
must be a root between 1 and 3. This argument assumes that ∕(x) is continuous, which means that its
graph has no "jumps" in it. In other words, we can draw the portion of the graph between any two
points without lifting our pencil. Proving that ∕(x) is continuous requires too advanced tools to cover
in this book. □

WARNING!! Once we determined both that there is a root between 1 and 3, and
*jk
* that 2 is the only integer between 1 and 3 that divides 24, we still had
to check if 2 is in fact a root.
To see why we can't immediately conclude that 2 is a root, consider
the polynomial g(x) = x3 — 14x2 - 16x + 15. We have g(0) = 15 and
g(-3) = -90, so there is a root between 0 and -3. Moreover, -1 is the
only integer between 0 and -3 that divides 15 evenly. However, we
have g(-l) = 16, so -1 is not a root of g(x). The root of g(x) that is
between 0 and -3 is not an integer. (See if you can find it!)

Our method of determining that there is a root of ∕(x) between 1 and 3 is a very handy tool for
simplifying the search for roots of a polynomial.

Important: If ∕(x) is a polynomial and a and b are real numbers such that ∕(fl) and

ψ f(b) have different signs, then there is a root of ∕(x) between a and b.

Notice that we cannot determine whether or not there is a root between a and b if f(a) and f(b) are

202
 7.2. INTEGER ROOTS

the same sign. For example, we have ∕(l) = 12 and ∕(5) = 24 in Problem 7.8. Even though ∕(1) and ∕(5)
are the same sign, we cannot deduce that there are no roots between 1 and 5. (Indeed, as we saw, there
are two roots of ∕(x) between 1 and 5.)

Problem 7.9: Let ∕(x) = x4 - 5x3 - 4x2 + 44x - 48. Factor f(x), find its roots, and graph it.

Solution for Problem 7.9: We start our hunt for roots by trying divisors of 48. We start with 1 and -1
because it's so easy to evaluate ∕(1) and f(-l).

Concept: We start our search for nonzero integer roots of any polynomial f (x) with
o=≡a 1 and —1, since it's so easy to evaluate ∕(1) and /(—1).

Unfortunately, we have ∕(1) = -12and∕(-l) = -90, so neither

T}<
1 -5 44 -48

1
1 nor —1 is a root. Since ∕(1) is closer to 0, we try 2 next instead of

x©
2 -20 48

1
—2. The synthetic division at right shows us that 2 is a root, and 1 -3 -10 24 0
we have
∕(x) = (x - 2)(x3 - 3x2 - 10x + 24).
We continue our search for roots by finding roots of x3 — 3x2 — 10x + 24. We find that x = 2 is a root of
this polynomial as well.

WARNING!! Don't forget about multiplicity. If we find that a is a root of ∕(x), we

< can't forget that a may be a double root, or triple root, etc. So, we
check if x - a also divides the quotient ∕(x)∕(x - a).

Dividing x - 2 into x3 - 3x2 - 10x + 24 gives us a quotient of x2 - x - 12. So, we have

∕(x) = (x - 2)(x3 - 3x2 - 10x + 24) = (x - 2)2(x2 - x - 12).

Factoring the quadratic then gives us ∕(x) = (x - 2)2(x - 4)(x + 3), so the roots are -3, 2,2, and 4.

When graphing f, the roots tell us where graph intersects the x-axis. We choose a few more values
of x in the table at left below to find a few more points on the graph. The graph of f extends far off the
bottom and top of the diagram shown below. To indicate this, we place arrows where the graph extends
off the diagram. The graph is so steep near x = -3 that it almost appears to be a vertical line.

X ∕(χ) y
ik u — -" k
- —
-4 288
-3 0
-2 -96 I
-1 -90
0 -48
1 -12
2 0 -
■ -— ——
3 -6
4 0 . V
I
5 72
1 ’1

203
CHAPTER 7. POLYNOMIAL ROOTS PART I

From our table, we see that ∕(x) is so much less than 0 for most values of x between -3 and 2 that our
graph of y = ∕(x) extends well beyond the bottom of the shown graph. However, we do see interesting
behavior of the graph near x = 2. It looks like the graph just touches the x-axis, but doesn't cross it. We
know that the graph passes through (2,0). We can look at the factored form of f (x) to see why the graph
behaves the way it does near x = 2:

∕(x) = (x-2)2(x-4)(x + 3).

If 2 < x < 4, then x - 4 is negative, x + 3 is positive, x - 4 (x - 2)2 x + 3

and (x - 2)2 is positive, so ∕(x) is negative for 2 < x < 4. x>4 ÷ + + +
Similarly, if -3 < x < 2, then (x - 2)2 is positive, x - 4 2<x<4 - ÷ + —
is negative, and x + 3 is positive, so ∕(x) is negative for -3 < x < 2 + + —
-3 < x < 2 as well. We capture all of this information in x < -3 + +
the table at right. Therefore, the graph of f(x) is below
the x-axis both just to the left and just to the right of (2,0). (Notice that we could have omitted the (x 2)
column, since (x - 2)2 is always nonnegative.)
When x is very far from 0, the value of ∕(x) is primarily determined by the x4 term. So, in both cases,
∕(x) is a large positive number. This is depicted in our graph by the fact that the graph of ∕(x) goes far
above the x-axis both when it is far to the right of the y-axis and when it is far to the left of the y-axis. □

Problem 7.10: Find the roots of g(x) = x6 + 15x5 + 81x4 + 193x3 + 198x2 + 72x.

Solution for Problem 7.10: Our first step here is pretty easy. There's no constant term, so there is an x in
every term. This means we can factor out an x, and that 0 is a root of g(x):
g(x) = x(x5 + 15x4 + 81x3 + 193x2 + 198x + 72).

Important: If a polynomial has no constant term, then 0 is a root of the polynomial.

We continue our search for roots with x5 + 15x4 + 81x3 + 193x2 + 198x + 72. Let this polynomial be
t∕(x). Before we start trying factors of 72, we notice that the coefficients of q(x) are all positive. Therefore,
whenever x is positive, each term of q(x) is positive, which means q(x) is positive. So, there are no
positive roots. This cuts our search in half.
We start with -1 because q(-T) is easy to evaluate. We find that q(-l) = 0, so x + 1 is a factor of q(x).
Synthetic division gives us
g(x) = xq(x) = x(x + 1)(
* 4 + 14x3 + 67x2 + 126x + 72).

That worked, so we try it again. We find that x + 1 divides x4 + 14x3 + 67x2 + 126x + 72 because
1 - 14 + 67 - 126 + 72 = 0, and we have
g(χ) = x(x + l)2(x3 + 13x2 + 54x + 72).

We quickly see that x + 1 does not divide x3 + 13x2 + 54x + 72 (because —1 + 13 — 54 + 72 ≠ 0). So we try
x + 2, which fails, then x + 3, which does divide x3 + 13x2 + 54x + 72. Then, all we have left is a quadratic,

204
 7.2. INTEGER ROOTS

and we find
g(x) = x(x + l)2(χ + 3)(x2 + 10x + 24) = x(x + l)2(x + 3)(x + 4)(x + 6).
So, the roots of g(x) are -6, -4, -3, -1, -1, and 0. □
Finding roots is not the only application of factoring.

Problem 7.11: Solve the inequality x3 + 5x2 - 16x - 80 ≤ 0.

Solution for Problem 7.11: We tackled quadratic inequalities by factoring them, so we try factoring the
polynomial on the lesser side of the inequality. We find that

x3 + 5x2 - 16x - 80 = (x - 4)(x2 + 9x + 20) = (x - 4)(x + 4)(x + 5).

So, our inequality is (x - 4)(x + 4)(x + 5) ≤ 0.

Lθt ∕(x) = (x - 4)(x + 4)(x + 5). We can find the solutions x-4 x + 4 x ÷ 5 ∕ω
to f (x) < 0 by considering the signs of the factors of ∕(x) for x>4 + + + +
different values of x. All three factors of ∕(x) are positive for -4 < x < 4 - + + —
x > 4, so ∕(x) > 0 if x > 4. The factor x — 4 is negative and the -5 < x < -4 - - + +
other two factors are positive if —4 < x < 4, so ∕(x) < 0 when x < -5 — — — —
-4 < x < 4. Similarly, ∕(x) is positive if -5 < x < -4 and ∕(x)
is negative if x < —5. We can organize all this information in a table, as show n.

Because the inequality is nonstrict, the roots of ∕(x) also satisfy it. So, the values of x that satisfy the
inequality are x ≤ —5 and —4 ≤ x ≤ 4. In interval notation, our solution is x ∈ (—∞, —5] U [—4,4].

We can also reason our way to the answer by thinking about the graph of y
y = (χ - 4)(x + 4)(x + 5). When x is much less than 0, then ∕(x) is also much less
than 0. As x gets larger, the graph of y = ∕(x) approaches the x-axis, crossing the
x-axis at (-5,0) because -5 is a root of ∕(x). There are no roots smaller than -5,
so the graph of y = ∕(x) is below the x-axis for all x less than -5. So, ∕(x) < 0 for
x < -5. After passing through (-5,0), the graph goes above the x-axis, because as
x goes from just below -5 to just above it, the sign of x + 5 changes from negative
to positive. The signs of x - 4 and x + 4 stay negative, so ∕(x) > 0 just after the
graph passes through (-5,0).
The graph again crosses the x-axis at (-4,0), because -4 is a root of ∕(x). There
are no roots of ∕(x) between -5 and -4, so the graph of y = ∕(x) is above the x-axis
for all x such that -5 < x < -4. Therefore, ∕(x) > 0 for all x such that -5 < x < -4.

After passing through (-4,0), the graph goes below the x-axis, because the sign of x + 4 changes
from negative to positive as x goes from just below -4 to just above it. The signs of x + 5 and x - 4
don't change, so ∕(x) < 0 just after the graph passes through (-4,0). The graph next crosses the x-axis at
(4,0), so for x between -4 and 4, the graph of y = ∕(x) is below the x-axis. (In fact, it is so far below the
x-axis for most x between -4 and 4 that the graph extends well off the bottom of the diagram shown.)
Therefore, we have ∕(x) < 0 for all x such that -4 < x < 4. As x goes from a little less than 4 to a little
more than 4, the value of ∕(x) increases so fast from being much less than 0 to being much greater than
0 that the graph is almost vertical. Finally, for all x > 4, we have ∕(x) > 0. Combining the intervals of x

205
CHAPTER 7. POLYNOMIAL ROOTS PART I

for which our graph is below the x-axis with the points where the graph intersects the x-axis (because
the inequality is nonstrict), we have the same answer as before, x ∈ (-∞, -5] U [-4,4].
Our consideration of the graph of y = f(x) is essentially the same approach to the problem as the
table. In both, we determine the sign of ∕(x) for different intervals of x. We use the roots of ∕(x) to
determine the intervals we must consider. □

Exercises

7.2.1 Find all roots of each of the following polynomials.

(a) ∕(x) = x3 - 4x2 - llx + 30 (c) ∕(x) = -X5 - 12x4 + 6x3 + 64x2 - 93x + 36
(b) g(f) = f4 + 5f3 _ 19f2 _ 65f + 150 (d) ∕z(y) = 6y3 - 5y2 - 22y + 24

7.2.2 Find all solutions to the inequalities below.

(a) f3 + 10t2 + 17f > 28 (c) r4 + r3 + 5r2 - 13r —18 < 0
(b) r2(6r + 12 - r2) ≤ 5(14r + 15)

7.2.3 There are four roots of ∕(x) = x4 — 8x3 + 24x2 — 32x + 16. We can easily test to find that f(2) = 0.
We can then check all the other divisors of 16, both positive and negative, and find that no divisors of
16 besides 2 are roots of the polynomial. Is it correct to deduce that the other three roots of ∕(x) are not
integers?
*
7.2.4 Suppose that ∕(x) is a polynomial with integer coefficients such that ∕(2) = 3 and ∕(7) = -5.
Show that ∕(x) has no integer roots. Hints: 1,319

7.3 Rational Roots

A rational number is a number that can be expressed in the form p∕q, where p and q are integers.
A number that is not rational is called an irrational number. Just as we can use the coefficients of
a polynomial with integer coefficients to narrow the search for integer roots, we can also use these
coefficients to narrow the search for non-integer rational roots.

Problems

Problem 7.12: In this problem we find the roots of the polynomial g(x) = 12x3 + 16x2 - 31x + 10.
(a) Can you find any integers n such that g(ιι) = 0?
(b) Suppose g(p∕q) = 0, where p and q are integers and p/q is in reduced form. Rewrite g(p∕q) = 0
using the definition of g.
(c) Get rid of the fractions in your equation from (b) by multiplying by the appropriate power of q.
(d) What terms in your equation from (c) have p? Why must p divide 10?
(e) What terms in your equation from (c) have q? Why must q divide 12?
(f) Find all the roots of g(x).

206
 7.3. RATIONAL ROOTS

Problem 7.13: Let ∕'(x) be the polynomial ∕(x) = fl,lx'7 + a∣j-∣x"~1 + ιιu~2Xn~2 + ∙ ∙ ∙ + ∏↑x + ao, where all
the coefficients are integers and both an and ∏o are nonzero. Let p and q be integers such that p/q is a
fraction in simplest terms, and f(p∕q) = 0. In this problem, we show that p divides ⅜ and q divides an.
(a) Suppose f(p∕q) = 0- Why must we have

anpn + an-λpn~γq + απ-2pn"⅛2 + an-3p"~3q3 + ∙ ∙ ∙ + aλpq"~λ + a0qn = 0?

(b) Which terms in the equation in (a) must be divisible by p? Use this to show that ao must be
divisible by p.
(c) Which terms in the equation in (a) must be divisible by q? Use this to show that an must be
divisible by q.

Problem 7.14: Find all the roots of each of the following polynomials:
(a) ∕(x) = 12x3 - 107x2 - 15x + 54.
(b) g(x) = 30x4 - 133x3 - 121x2 + 189x - 45.

Problem 7.15: Find all r such that 12r4 - 16r3 > 41r2 - 69r + 18.

We've found a way to narrow our search for integer roots. How about rational roots?

: Problem 7.12: Find all x such that 12x3 + 16x2 - 31x + 10 = 0.

■■■■■■■■■■■■■

Solution for Problem 7.12: We let ∕(x) = 12x3 + 16x2 - 31x + 10 and begin our search for roots of ∕(x) by
seeing if we can find any integer roots. An integer root of ∕(x) must be a divisor of its constant term,
which is 10 = 21 ∙ 51. We try each divisor (positive and negative):
∕(1) = 12 ■ 13 +16∙12 -31-1 +10 = 7,
∕(-l) = 12 ∙ (-l)3 +16∙(-1)2 -31 ■ (-1) +10 = 45,
∕(2) = 12 ∙ 23 +16 ■ 22 -31 ∙ 2 +10 = 108,
∕(-2) =12-(-2)3 +16 ∙ (-2)2 -31 ∙ (-2) +10 = 40,
∕(5) = 12 ∙ 53 +16 ∙ 52 -31 ∙ 5 +10 = 1755,
∕(-5) = 12 ∙ (-5)3 +16 ∙ (-l)2 -31 ∙ (-5) +10 = -935,
∕(10) = 12 ∙ 103 +16 ∙ 102 -31∙10 +10 = 13300,
∕(-10) = 12 ∙ (-10)3 +16 ∙ (-10)2 -31 ∙ (-10) +10 = -10080.

(We might also have checked some of these a little faster with synthetic division.) We know we don't
have to check any other integers, since any integer root must divide the constant term of ∕(x). Since all
of these fail, we know there are no integer roots of ∕(x). Back to the drawing board.

A little number theory helped us find integer roots, so we hope that a similar method might help us
find rational roots. Letting p and q be relatively prime integers (meaning they have no common positive
divisor besides 1), we examine the equation f(p∕q) = 0:

+ 10 = 0.

207
CHAPTER 7. POLYNOMIAL ROOTS PART I

Multiplying by t∕3 to get rid of all the denominators, we have the equation

12p3 + lβp2q - 31pq2 + 10t∕3 = 0. (7.1)

Previously, we isolated the last term of such an expansion in order to find information about integer
roots. Doing so here gives us
10ty3 = p(-12p2 - 16pq + 31ty2).

Dividing this equation by p gives us

= -12p2 - 16pq + 31t∕2.

P
The right side of this equation must be an integer, so 10q3∕p must also be an integer. Therefore, p divides
10tj3. But, since p and q are relatively prime, p must divide 10.
If we isolate the first term of the left side of Equation (7.1) instead of the last term, we have

12p3 = q(-16p2 + 31pq - 10t∕2).

Dividing this equation by q gives us

= -16p2 + 31pq - 10i∕2.

As before, the right side is an integer, so the left side must be as well. Thus q divides 12p3. Because p
and q are relatively prime, we know that q must divide 12.
Combining what we have learned about p and q, we see that if p/q is a root of ∕(x), then p is a divisor
of 10 and q is a divisor of 12.
Now, we make a list of possible rational roots p/q:

2 5 10 1 5 1 5 1 5
i3'i3'i 3 ,i4'i4'i6,i6'

While this is not a terribly small list, it is a complete list of possible rational roots of a polynomial with leading
coefficient 12 and constant term 10. We already know that ∕(x) has no integer roots, so we continue our
search for roots with the fractions in the list above. We start with the fractions with small denominators,
since these will be easiest to check.

After a little experimentation, we find that ∕(1∕2) = 0 via the synthetic 12 16 -31 10
division at right. (You may have found one of the other roots first.) We 6 11 -10
therefore have 12 22 -20 ∣ 0
∕(x) = (x - I) (12x2 + 22x - 20).

We continue our search by finding roots of 12x2 + 22x - 20. We find 12x2 + 22x - 20 = 2(6x2 + llx - 10) =
2(3x - 2)(2x + 5), so we have
∕(x)=2(x-l)(3x-2)(2x + 5).

208
 7.3. RATIONAL ROOTS

When factoring polynomials, we often factor constants out of each linear factor so that the coefficient of
x in each factor is 1. This makes identifying the roots particularly easy:

∕(.r) = 2(x - ∣) (3) (x - ?)(2) (x + I) = 12 (x - ∣) (x - ∣) (x + ∣).

Now we can easily see that the roots of ∕(x) are j, and -∣. □

In our solution, we found a way to narrow our search for rational roots of f(x) in much the same way
we can narrow our search for integer roots. Let's see if we can apply this method to any polynomial
with integer coefficients.

Problem 7.13: Let∕(x) be the polynomial∕(x) = anxtl + ⅜-ιxπ~1 + an-2xn~2 + + a↑x + ao, where all
the coefficients are integers and both all and «o are nonzero. Let p and q be integers such that p/q is a
fraction in simplest terms, and f(p∕q) = 0. Show that p divides «0 and q divides an.

Solution for Problem 7.13: We use our solution to Problem 7.12 as a guide. Since f(p∕q) = 0, we have

"n(lj) +*,^1(¾) + + '→'≈ι(^+'⅛=0.

We get rid of the fractions by multiplying both sides by qn, which gives us

anpn + an-γpn~1q + an-2pn~2q2 + ■■■ + a1pqn~1 + a0qn = 0. (7.2)

To see why q must divide an, we divide Equation (7.2) by q to produce an∕q in the first term:

+ an-ιpn~1 + an-2pn~2q + ■ ∙ ∙ + a1pqn~2 + a0qn~1 = 0.

Isolating this first term gives us

= -an-1pn~1 - an-2pn~2q-------- a1pqn~2 - a0qtl~1.

cl
Each term on the right side of this equation is an integer, so the entire right side is an integer. Therefore,
anpn ∕ q must be an integer. Since p/qis in lowest terms and anpn ∣q is an integer, we know that q divides an.
We take essentially the same approach to show that p ∣ ⅜. We divide Equation (7.2) by p and isolate
the term with an in it, which gives us

u°tf _ _n n17-l _ n .nn-2 _ ^t,n-3n2 _ _n nn-l

— ^nP ⅞n-lP 7

Each term on the right is an integer, so the entire right side equals an integer. So, the expression aoq', ∣p
must equal an integer, which means that p must divide a$. □

Extra! If f(ri) = n2 + n + 41, then ∕(1), ∕(2), ∕(3), and ∕(4) are all prime. Maybe /(;/) is prime for
"* llll* , *,"l*∙ all integers 11. Is it? (Continued on page 220)

209
CHAPTER 7. POLYNOMIAL ROOTS PART I

In Problem 7.13, we have proved the Rational Root Theorem:

Important: Let ∕(x) be the polynomial

V ∕(x) = anxn + an-ιxn~γ + an-2xn~2 + ∙∙∙ + a1x + ao,

where all the a∣ are integers, and both an and «0 are nonzero. If p and q
are relatively prime integers and f(p∕q) = 0, then p ∣ Λq and q ∣ an.

Problem 7.14: Find all the roots of each of the following polynomials:
(a) ∕(x) = 12x3 - 107x2 - 15x + 54
(b) g(x) = 30x4 - 133x3 - 121x2 + 189x - 45

Solution for Problem 7.14:

(a) We first look for easy roots by evaluating ∕(1) and ∕(-l). Wefindthat ∕(1) = 12—107—15+54 = —56
and ∕(-l) = -12 - 107 + 15 + 54 = -50. Since ∕(0) = 54, we see that there is a root between -1 and
0 (because ∕(-l) < 0 and ∕(0) > 0) and a root between 0 and 1 (because ∕(0) > 0 and ∕(1) < 0).
Concept: We often start our hunt for rational roots by evaluating /(—1), ∕(0), and
Oβ ∕(1), because these are easy to compute and the results may tell us where
to continue our search.
We continue our hunt for roots with ∣, which may be a root because 2 divides the leading coefficient
of ∕(x). Using synthetic division, we find ∕(∣) = 211, so there is a root between ∣ and 1. We try ∣
and it works, giving us

∕(x) = (x - ∣j(12x2 - 99x - 81) = 3(x-∣l(4x2-33x-27).

Factoring the quadratic then gives us ∕(x) = 3 (x - ∣) (4x + 3)(x - 9). So, the roots of ∕(x) are -∣, ∣,
and 9.
(b) We start our hunt for roots by finding g(0) = -45, g(l) = -80, and g(-l) = -192. Unfortunately,
that doesn't help too much; we can't immediately tell if there are any roots between -1 and 0,
or between 0 and -1. So, we test 3 and -3, knowing that 2 and -2 cannot be roots because the
constant term of g(x) is odd. By synthetic division, we find that g(3) = -1728. This has the same
sign as g(l), so we continue with g(-3). We find that g(-3) = 4320, so because g(-3) > 0 and
g(-l) < 0, we know there is a root between -3 and -1. (Note that we didn't need to find the actual
value of g(-3). Once it is clear that g(-3) is a large positive number, we know that there is a root
between -1 and -3.) With synthetic division, we find that g(-2) is also a large positive number,
so there is a root between -2 and -1.
3 189 -45
VJe try -3/2. We know that -3/2 might be a root 2 30 -133 -121
of g(x) because -3 divides the constant term and 2 -45 267 -219 45
divides the leading coefficient. If -3/2 doesn't work, 30 -178 146 -30 ∣ 0
we will at least narrow our search to between -2 and

210
 7.3. RATIONAL ROOTS

-3/2 or between -3/2 and -1. Fortunately, the synthetic division shows us that

g(x) = (* + ∣) (30x3 - 178x2 + 146x - 30) = 2 (x + ∣) (15x3 - 89x2 + 73x - 15).

We now note that 15x3 - 89x2 + 73x -15 has no negative roots because if x is negative, each term of
this polynomial is negative. So, we only have to search for positive roots. We already know that
1 and 3 don't work, and the only two other positive integers that divide 15 are 5 and 15. We find
that 5 works, and we have

g(x) = 2 (x + ∣) (x - 5)(15x2 - 14x + 3).

Now, we can factor the quadratic or use the quadratic formula to find

g(x) = 2 (x + ∣) (χ - 5)(3x - l)(5x - 3) = 30 (x + ∣) (x - 5) (x - ∣)(x - ∣).

The roots of g(x) are -3/2, 1/3, 3/5, and 5. Notice that there are roots between 0 and 1. Why
doesn't testing g(0) and g(l) reveal the fact that these roots exist? (You'll have a chance to answer
this question as an Exercise.)

Problem 7.15: Find all r such that 12r4 - 16r3 > 41r2 - 69r + 18.

Solution for Problem 7.15: First, we move all the terms to the left, which gives us

12r4 - 16r3 - 41r2 + 69r - 18 > 0.

Now, we must factor the polynomial on the left. Let this polynomial be ∕(r). Because ∕(0) = -18 and
∕(1) = 6, there is a root between 0 and 1. We find that ∕(1∕2) = 5, so the root is between 0 and 1/2. Trying
1/3 gives us

Continuing with 4r3 — 4r2 — 15r + 18, we find that -1 and 2 are not roots, but —2 is, and we have

∕(2) = 3 (r - ∣) (r + 2)(4r2 - 12r + 9) = 3 (r - ∣) (r + 2)(2r - 3)2.

So, our inequality is 3 (r - ∣) (r + 2)(2r - 3)2 > 0.

The expression on the left side of the inequality equals

r+2 r-1 (2r - 3)2 ∕(r)
0 when r = -2, 1/3, or 3/2. We consider the intervals
between (and beyond) these values to build the table at r < -2 - - + +
right. We see that f(f) > 0forr ∈ (-∞, -2)U(j, ∣)U(∣, +∞). -2<r<⅜ + ÷ —
The inequality is strict, so the roots of ∕(r) do not satisfy 13 < r < 52 + ÷ + ÷
the inequality. □ q
r> - + + + +

211
CHAPTER 7. POLYNOMIAL ROOTS PART I

Exercises

7.3.1 Find all roots of the following polynomials:

(a) g(y) = 12y3 - 28y2 - 9y + 10 (b) ∕(x) = 45x3 + 48x2 + 17x + 2

7.3.2 In part (b) of Problem 7.14, we saw that g(0) and g(l) are both negative, and yet we still found
roots between 0 and 1. How can that be? Shouldn't g(0) and g(l) have different signs if g(x) = 0 for
some x between 0 and 1?
7.3.3 Suppose f(x) = x3 + ∣x2 - 4x - 3. Notice that ∕(2) = 0, so 2 is a root of ∕(x). But 2 doesn't evenly
divide the constant term of ∕(x). This seems to violate the Rational Root Theorem! Does it really?
7.3.4 In Problem 7.15, we built a table to find the values of r for which 3(r - ∣)(r + 2)(2r - 3)2 > 0. How
could we have solved this inequality by thinking about the graph of y = 3(x - j)(x + 2)(2x - 3)2?

7.3.5 Solve the two inequalities below:

(a) 24x3 + 26x2 ≥ 21x + 9 *
(b) 6s4 + 13s3 - 2s2 + 35s - 12 < 0

7.4 Bounds

In this section, we learn how to use synthetic division to determine bounds on the roots of a polynomial.

Problems

Problem 7.16:
(a) Let f (x) = x3 + 13x2 + 39x + 27. Without even finding the roots, how can we tell that there are no
positive values of x for which ∕(x) = 0?
(b) Let ∕(x) = 3x3 - 28x2 + 51x - 14. Without finding the roots, how can we tell that there are no
negative values of x for which ∕(x) = 0?

Problem 7.17: Let ∕(x) = anxn + an-ιxn~1 + ∙ ∙ ■ + a1x + a0.

(a) Show that if all the coefficients of ∕(x) have the same sign (positive or negative), then ∕(x) has
no positive roots.
(b) Show that if a∣ > 0 for all odd i and a∣ < 0 for all even i, then ∕(x) has no negative roots.
(c) Is it possible for ∕(x) to have negative roots if a, is negative for all odd i and positive for all
even z?

Problem 7.18: Let g(x) = x4 - 3x3 - 12x2 + 52x - 48.

(a) Use synthetic division to divide g(x) by x - 6.
(b) What do you notice about the coefficients in the quotient and the remainder from part (a)?
(c) How can you use your answer to part (b) to deduce that there are no roots of g(x) greater than
6? (In other words, how do you know you don't have to test 8,12,16, etc.?)

212
 7A. BOUNDS

Problem 7.19: Let ∕(x) be a polynomial and a be a positive constant.

(a) Let ∕(x) = (.v - a)q↑ (x) + r∣, so that z∕ι(x) and r↑ are the quotient and remainder, respectively, when
∕(x) is divided by x - a. If ∩ is positive and the coefficients of z∕ι(x) are all nonnegative, explain
why ∕(x) has no roots that are greater than a.
(b) Why must a be positive to draw the conclusions we found in part (a)?

Problem 7.16:
(a) Let ∕(x) = x3 + 13x2 + 39x + 27. Without even finding the roots, how can we tell that there are
no positive values of x for which ∕(x) = 0?
(b) Let ∕(x) = 3x3 - 28x2 + 51x - 14. Without finding the roots, how can we tell that there are no
negative values of x for which ∕(x) = 0?

Solution for Problem 7.16:

(a) If x > 0, then each term of ∕(x) is positive, so ∕(x) is positive. Thus, there are no positive roots of
∕(x)∙
(b) If x < 0, then all four terms of ∕(x) are negative, so ∕(x) is negative, which means there are no
negative roots of ∕(x).
□
Problem 7.16 suggests a general method we can sometimes use to quickly determine if a polynomial
has no positive roots or has no negative roots.

Problem 7.17: Let ∕(x) = anxn + an-∖Xn 1 + ∙ ∙ ∙ + «jx + clq.

(a) Show that if all the coefficients of ∕(x) have the same sign (positive or negative), then ∕(x) has
no positive roots.
(b) Show that if zz;- > 0 for all odd i and αi∙ < 0 for all even z, then ∕(x) has no negative roots.
(c) Is it possible for ∕(x) to have negative roots if αz∙ is negative for all odd i and positive for all
even z?

Solution for Problem 7.17:

(a) Suppose that all the a, are positive. Then, if x > 0, every term of ∕(x) is positive, so f (x) > 0.
Therefore, ∕(x) has no positive roots. Similarly, if all the zz,∙ are negative and x > 0, then every term
of ∕(x) is negative. So, ∕(x) < 0, which means ∕(x) has no positive roots.
(b) If z is odd and x < 0, then a∣ > 0 and x' < 0, so α1∙x, < 0. Similarly, if i is even and x < 0, then a, < 0
and x' > 0, so zz,x' < 0 in this case, as well. Therefore, every term zz,∙x' of ∕(x) is negative if x < 0, so
∕(x) < 0 for all negative values of x. This tells us that ∕(x) has no negative roots.
(c) We follow essentially the same steps as in the last part. If z is odd and x < 0, then α1∙ < 0 and x' < 0,
so zz,x' > 0. If z is even and x < 0, then a, > 0 and xl > 0, so zz,x' > 0. Therefore, zz,x' is positive for all
z, which means ∕(x) > 0. So, ∕(x) cannot have negative roots.
□

213
CHAPTER 7. POLYNOMIAL ROOTS PART I

Important: If the coefficients of a polynomial all have the same sign, then the polyno­
mial has no positive roots. If the coefficients of a polynomial are nonzero
and alternate in sign, then the polynomial has no negative roots.

As an Exercise, you'll explore what information we can deduce if some of the coefficients equal 0.
But now, we'll apply the same reasoning we used on these first two problems to determine upper and
lower bounds for roots of polynomials.
Problem 7.18: Find the quotient and remainder when g(x) = x4 - 3x3 - 12x2 + 52% - 48 is divided by ∣
x - 6. Explain how the quotient and remainder tell us that there are no roots of g(x) greater than 6.

Solution for Problem 7.18: Our synthetic division at right shows us 1 -3 -12 52 -48
that 6 18 36 528
g(x) = (x - 6)(x3 + 3x2 + 6x + 88) + 480. 1 3 6 88 ∣ 480

To explore whether or not g(x) has any roots that are greater than 6, suppose x > 6 and consider each
expression on the right side of our equation. The expression x - 6 is positive, and because the coefficients
of x3 + 3x2 + 6x + 88 are all positive, this polynomial is positive, too. Finally, the remainder is positive
for all values of x. So, if x > 6, the equation

g(x) = (x - 6)(x3 + 3x2 + 6x + 88) + 480

tells us that g(x) is the product of two positive numbers, plus another positive number. Therefore, we
know that g(x) cannot equal 0. We've found that g(x) has no roots greater than 6. So, we say that 6 is an
upper bound for the roots of g(x). □
Let's generalize Problem 7.18.

Problem 7.19: Let ∕(x) be a polynomial and a be a positive constant.

(a) Let ∕(x) = (x - a)q∖(x) + rι, so that (∕ι(x) and r1 are the quotient and remainder, respectively,
when ∕(x) is divided by x - fl. If rι is positive and the coefficients of t∕ι(x) are all nonnegative,
explain why ∕(x) has no roots that are greater than a.
(b) Why must a be positive to draw the conclusion we found in part (a)?

Solution for Problem 7.19:

(a) We follow essentially the same argument we used in Problem 7.18.
Concept: To prove a general mathematical statement, it's often helpful to work out
θ —j a specific example of that statement, then build your general proof by
following your example.
If x > fl, then x-a is positive. If the coefficients of q↑ (x) are all nonnegative, then q↑ (x) is nonnegative
for all positive values of x. The value of ∏ does not depend on x and we are given that r1 is positive.
Therefore, if x > fl in
∕(x) = (x - fltyι(x) + n,
then ∕(x) is positive. This means that ∕(x) cannot have any roots greater than fl.

214
 7.5. GRAPHING AND THE FUNDAMENTAL THEOREM OF ALGEBRA

(b) In showing that ∕(x) > 0 if x > a in part (a), we used the fact that x is positive because a is positive.
However, if a is negative, then x is not positive for all x > a. If x is not necessarily positive, then
we cannot deduce that ∣∕ι(x) is nonnegative merely from the fact that the coefficients of ^∣(x) are
nonnegative. So, we don't know the sign of <∕ι(x), which means we cannot determine the sign of
(x - a)q↑ (x) + Γ]. Therefore, ∕(x) may have roots greater than a.

Important: Let ∕(x) = (χ - fl)tj1(x) + r↑, so that q↑(x) and r↑ are the quotient and
remainder, respectively, when ∕(x) is divided by x-a, where a is a positive
constant. If r↑ is positive and the coefficients of q↑ (x) are all nonnegative,
explain why ∕(x) has no roots that are greater than a.

WARNING!! The rule above does not apply if a is positive.

iiI
*κi⅜e>
lκ<

7.4.1 Find the solutions to each of the following:

(a) 6r3 + 31r2 + 34r = 15 (b) 4r4 - 28r3 + 61r2 - 42r + 9 = 0

7.4.2 Larry is finding the roots of a polynomial ∕(x). He notes that ∕(1) = 34 and ∕(0) = -3, so he
knows that there is a root between 0 and 1. He finds that x = 1/3 is a root. Can he then deduce that all
the other roots are either greater than 1 or less than 0?
7.4.3 In the text, we noted that if the coefficients of a polynomial all have the same sign, then the
polynomial has no positive roots. What if some of the coefficients equal 0, but all the others still have
the same sign? Then is it possible for the polynomial to have a positive root?
*
7.4.4 Consider the set of all equations x3 + aιx2 + a↑x + α0 = 0, where a^, α1, a0 are real constants and
∣αt-∣ ≤ 2 for i = 0,1,2. Let r be the largest positive real number that satisfies at least one of these equations.
Which of the inequalities below does r satisfy?
(A)l≤r<∣ (B)∣<r<2 (C) 2 ≤ r < ∣ (D) ∣ ≤ r < 3 (E)3≤r<∣.

(Source: AHSME) Hints: 197

7.5 Graphing and the Fundamental Theorem of Algebra

In this section, we get a further feel for the Fundamental Theorem of Algebra by tackling a few questions
about graphs of polynomials.

Problems

Problem 7.20: Let ∕(x) be a polynomial of degree n, where n > 0. Show that the graphs of y = c and
y = f (x) cannot intersect at more than n points.

215
CHAPTER 7. POLYNOMIAL ROOTS PART I

Problem 7.21: The graph of a polynomial intersects the x-axis at two distinct points. Does this mean
that the polynomial is a quadratic?

Problem 7.22: At left below is the graph of y = ∕(x), where ∕(x) is a polynomial.
(a) Explain why the degree of the polynomial is at least 5.
(b) Explain why the polynomial must have an odd degree.

Figure 7.1: Diagram for Problem 7.22 Figure 7.2: Diagram for Problem 7.23

Problem 7.23: At right above is the graph of y = ∕(x), where ∕(x) is a polynomial. The only intercepts
of the graph are (-1,0), (1,0), (3, 0)z and (0z3)z and we have deg f < 5. Find ∕(x).

Problem 7.20: Let ∕(x) be a polynomial of degree n, where n > 0. Show that the graphs of y = c and
y - f{x) cannot intersect at more than n points.
-mB≡M⅛B≡≡β≡≡≡≡≡≡≡≡≡M≡≡≡"≡≡≡≡"≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡"≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡≡B≡S

Solution for Problem 7.20: For each point (xz y) at which the graphs of y = c and y = f(x) intersect, we
must have ∕(x) = c. Subtracting c from both sides, we have ∕(x) - c = 0. The values of x for which
∕(x) - c = 0 are the roots of ∕(x) - c. Because c is a constant, the degree of ∕(x) - c equals the degree of
∕(x), which is n. So, by the Fundamental Theorem of Algebra, there are n roots of ∕(x) - c, and therefore
n values of x such that ∕(x) - c = 0. This tells us that there are no more than n real values of x such that
∕(x) - c = 0, so there are no more than n points (x, y) at which the graphs of y = c and y = ∕(x) intersect.
□
Problem 7.21: The graph of a polynomial intersects the x-axis at two distinct points. Does this mean
that the polynomial is a quadratic?

Solution for Problem 7.21: Let the polynomial be ∕(x). Because the graph of y = ∕(x) intersects the x-axis
at two different points, there are two real values of x such that ∕(x) = 0. So, ∕(x) has two distinct real
roots. However, this doesn't say anything about nonreal roots of ∕(x), or about double roots (or about

216
 7.5. GRAPHING AND THE FUNDAMENTAL THEOREM OF ALGEBRA

triple roots, etc.). For example, at left below is the graph of y = (x - 2)2(x + 2)2 and at right is the graph
of y = x4 - 16.

yn j 1

Figure 7.3: Graph of y = (χ - 2)2(x + 2)2 Figure 7.4: Graph of y = x4 - 16

In both cases, the graph of the polynomial has only two x-intercepts, but the polynomial has degree
4. In the first case, the polynomial has two double roots, while in the second, the polynomial has two
nonreal roots.

While we cannot say for sure that the polynomial is a quadratic, the fact that the polynomial has at
least 2 roots tells us that its degree is at least 2, because the degree of a polynomial equals the number
of roots it has. □
Problem 7.22: At right below is the graph of y = ∕(x), where ∕(x) is a polynomial.
(a) Explain why the degree of the polynomial is at least 5.
(b) Explain why the polynomial must have an odd degree.

Solution for Problem 7.22:

(a) The graph of y = -4 (dashed in the diagram) intersects the graph at 5
points. Therefore, there are 5 values of x for which ∕(x) = -4, so there
are 5 values of x such that ∕(x) + 4 = 0. Since ∕(x) + 4 is a polynomial
with 5 distinct roots, the degree of ∕(x) + 4 is at least 5. Since deg f > 0,
the degree of ∕(x) + 4 equals deg/, so the degree of ∕ is at least 5.
(b) When x is much more than 0 or much less than 0, the value of ∕(x) is
largely determined by its leading term. If a polynomial ∕(x) has even
degree, then its leading term is ax2n for some constant a and some integer
n. When x is much more than 0 or much less than 0, the term x2" is a
large positive number. So, the sign of ax2n is same as the sign of a, and it
is the same both when x is much more than 0 and τυhen x is much less than 0.
In our graph, for large values of x, the value of ∕(x) is positive, but when
x is much less than 0, the value of ∕(x) is negative. Therefore, the degree of ∕(x) cannot be even, so
the degree of ∕(x) must be odd.

□
Notice that we cannot be sure that the degree of the polynomial graphed in Problem 7.22 is exactly

217
CHAPTER 7. POLYNOMIAL ROOTS PART I

5. We only know that the degree is at least 5. For example, in Problem 7.21, we graphed ∕(x) = x4 - 16.
No horizontal line hits more than 2 points on the graph of f(x), but the degree of f(x) is 4, not 2. So,
the number of intersection points of a horizontal line with the graph of a polynomial only gives a lower
bound for the degree of the polynomial.

Problem 7.23: At right below is the graph of y = f(x), where f{x) is a polynomial. The only intercepts
of the graph are (-1,0), (1,0), (3,0), and (0,3), and we have degf < 5. Find fix).

intercepts of the graph, we see that at one of them, (-1,0), the graph does not cross the x-axis. Specifically,
we see that ∕(x) is positive both for x slightly larger than -1 and for x slightly smaller than -1. This is
exactly the behavior we expect if (x + 1)I2 divides ∕(x) evenly, since (x + 1)2 is positive both for x > -1
and x < -1. We've found our double root! (If x + 1 were a single root, then ∕(x) would change sign as x
goes from slightly less than -1 to slightly greater than -1.) So, now that we know all four roots of ∕(x),
we know that
∕(x) = c(x + l)2(x - l)(x - 3)
for some constant c. The graph passes through (0,3), so ∕(0) = 3. Combining ∕(0) = 3 with our expression
for ∕(x) above gives us c = 1, so ∕(x) = (x + l)2(x - l)(x - 3). □

I *x<
ι<,ise

7.5.1 Sketch the graphs of each of the following functions. Your graphs don't have to be precisely
correct, but they should have the correct intercepts and be above or below the x-axis at the right places.

(a) ∕(x) = 0.2(x - 3)(x + 2)(x + 3) (c) ∕(x) = 0.1(x + l)3(x - 2)(x - 4)
(b) ∕(x) =-(x - 2)2(x - l)2(x + 1)2

7.5.2
(a) What are the real roots of the polynomial x4 - 8x2 + 16 = 0?
(b) Solve the inequality x4 - 8x2 + 16 < 0 for real values of x.
(c) Graph the function ∕(x) = x4 - 8x2 + 16.

218
 7.6. ALGEBRAIC APPLICATIONS OF THE FUNDAMENTAL THEOREM

7.5.3 Let ∕(x) be the polynomial that is graphed at right. The graph passes through
(-4,0), (-3,0), (2,0), and the origin.

(a) Explain how we know that -4 is not a double root of f(x).

(b) What is the minimum possible degree of /?
(c) Suppose f has the degree found in part (b) and passes through (-2,4). Find ∕(x).

7.5.4 Suppose that ∕(x) is a polynomial with degree 6. Is it possible for the graph of ι∕ = x to intersect
the graph of y = ∕(x) at 7 different points? Why or why not?

7.6 Algebraic Applications of the Fundamental Theorem

Suppose that we are given two polynomials, pι(x) and pz{x)∙ How can we tell if they are the same
polynomial? At first, this appears to be an obvious question, because we can tell if two polynomials are
the same simply by comparing their coefficients. However, what if we don't know the coefficients, but
instead just know the values of pι(x) and pι(x) for a few values of x?
To investigate this question, let's play a game called "Guess the Polynomial." In this game, I think
of a polynomial ∕(x), and I give you ∕(x) for certain values of x. Then, based on these values, you guess
what polynomial I am thinking of. Ready?

OK, here are the values: ∕(1) = 1, ∕(2) = 8, ∕(3) = 27, and ∕(4) = 64. What do you think ∕(x) is?
You're probably guessing that ∕(x) = x3, right? Sorry, but I was thinking of the polynomial

∕(x) = x3 + (x - l)(x - 2)(x - 3)(x - 4) (16,000,000x427 - 473.15x101 - -^x23 + x5 + 99^.

But you were close. Thanks for playing "Guess the Polynomial."
Well, obviously that wasn't a very fair game. But Γm not trying to be whimsical (not completely,
anyway); I want to show that the values of a polynomial are not enough to deduce what the polynomial
is. We need to know something else about the polynomial, something more fundamental. The key piece
of additional information is the degree of the polynomial.

Problems

Problem 7.24: Suppose ∕(x) is a polynomial such that

∕(x) = anxn + σn-ιx"^1 + an-2Xn~2 + ∙ ∙ ■ + «ix + ⅜z

and there are at least n + 1 different values of x for which ∕(x) = 0.

(a) If an ≠ 0, then what does the Fundamental Theorem of Algebra tell us about the number of
different values of x for which ∕(x) = 0?
(b) Why must we have ∕(x) = 0 for all values of x?

219
CHAPTER 7. POLYNOMIAL ROOTS PART I

Problem 7.25: Let ∕(x) be a quadratic polynomial such that ∕(2) = 4, ∕(3) = 9, and ∕(4) = 16.
(a) Let g(x) = f(x) - x2. What's the greatest possible degree of g?
(b) Find at least three solutions to g(x) = 0.
(c) Prove that g(x) = 0 for all values of x.
(d) What does this tell us about /(x)?

Problem 7.26: Let pι(x) and p2(x) be polynomials of degree at most n. In this problem, we prove that
if pι(x) = p2(x) for n + 1 distinct values of x, then p1(x) = p2(x) for all values of x.
(a) Consider the polynomial p(x) = pι(x) - p2(x)- What is the greatest possible degree of p(x)?
(b) Based on your answer from (a), what is the greatest number of roots the polynomial p(x) can
have if p(x) is not 0 for all values of x?
(c) If pι(x) = p2(x) for n + 1 distinct values of x, then how at least how many roots does pι(x) - p2(x)
have?
(d) Combine your answers to parts (b) and (c) to prove that pι(x) = p2{x) for all values of x.

Problem 7.27:
Consider the system of equations

«1 + 8α2 ÷ 27α3 + 64(74 = L

8«i + 27a2 + 64α3 + 125a4 = 27,
27a1 + 64α2 ÷ 125α3 + 216a 4 = 125,
64flι + 125λ2 ÷ 216fl3 + 343a4 = 343.

These four equations determine a4, a2, a2,, and «4. Show that

aγm3 + a2(m + 1)3 + a2,{m + 2)3 + α4(m + 3)3 = (2m - 1)3

for all values of m.

(b) Prove that a↑ + a2 + <⅛ + ⅜ = 8 and 64«i + Ua2 + 8α3 + fl4 = 729.
(Source: Mandelbrot)

Problem 7.28: There is a unique polynomial P(x) of the form P(x) = 7x7 + c1x6 + c2x5 + ••■ + Cf,x + Cγ
such that P(l) = 1, P(2) = 2,..., and P(7) = 7.
(a) Find the roots of P(x) - x.
(b) Find the constants a, r↑, r2, r3, ...lr2 such that P(x) - x = «(x - r1)(x - r2) ∙ ∙ ∙ (x - r7) for all values
of x.
(c) Determine P(0). (Source: Mandelbrot)

Extra! The polynomial ∕(n) = n2 + π + 41 is prime for all integers n such that 0 < 11 < 40, but
-ι* < ih»κ ∕(40) = 402 + 40 + 41 = 40(41) + 41 = 412, which is not prime. This might make us wonder
if there is any polynomial that is prime for all integers. Try to find one (but don't try too
hard). (Continued on page 226.)

220
 7.6. ALGEBRAIC APPLICATIONS OF THE FUNDAMENTAL THEOREM

Problem 7.29: Let g(x) and ∕(x) be polynomials. Back in Problem 7.4 we showed that it is not
necessarily true that g(x) is a factor of f(x) if every root of g(x) is a root of ∕(x). We showed this by
considering g(x) = (x - 1)2 and ∕(x) = x(x - 1). Notice that our choice of g(x) has a double root.
What if g(x) does not have any multiple roots? Specifically, suppose that the roots of the polynomial
g(x) are distinct, and that each of these roots is a root of the polynomial ∕(x). Must g(x) be a factor of
/(x)?

Problem 7.24: Suppose ∕(x) is a polynomial such that

∕(x) = anxn + fln-1χ"^1 + an-2xn~2 + ■■■ + a-ιx + ao

and there are at least n + 1 different values of x for which f(x) = 0. Show that ∕(x) = 0 for all x.

Solution for Problem 7.24: If any of the a∣ are nonzero, then deg∕ ≤ n. Therefore, by the Fundamental
Theorem of Algebra, if any of the a∣ are nonzero, then ∕(x) cannot have more than n roots. However, we
are told that there are at least n + 1 values of x for which ∕(x) = 0. So, it is impossible for any of the to
be nonzero (since then the Fundamental Theorem of Algebra would force ∕(x) to have no more than n
roots). Therefore, all the ai must be 0, which means ∕(x) = 0 for all values of x. □

Important: If ∕(x) is a polynomial such that

∕(x) = anxn + ⅜-1xπ^1 + an-2Xn~2 + ∙ ∙ ■ + a↑x + ao,

and there are at least n + 1 different values of x for which ∕(x) = 0, then
∕(x) = 0 for all values of x. That is, an = an-^ = an~2 = ∙ ∙ ∙ = t⅛ = 0.

Problem 7.25: Suppose that ∕(x) is a quadratic polynomial such that ∕(2) = 4, ∕(3) = 9, and ∕(4) = 16. ∣
Prove that ∕(x) = x2.

Solution for Problem 7.25: By inspection, we note that ∕(x) = x2 satisfies ∕(2) = 4, ∕(3) = 9, and ∕(4) = 16.
However, maybe there are other quadratics ∕(x) such that ∕(2) = 4, ∕(3) = 9, and ∕(4) = 16. How can we
tell if ∕(x) = x2 is the only possible one?
We let g(x) = ∕(x) - x2 and focus on g(x), because we know that 2, 3, and 4 are all roots of g(x).
Therefore, the Factor Theorem tells us that x — 2, x - 3, and x - 4 are all factors of g(x). So, we have

g(x) = (x - 2)(x - 3)(x - 4)^(x),

for some polynomial q(x). This tells us that either the degree of g(x) is at least 3, or t∕(x) = g(x) = 0 for all
values of x. However, because g(x) = ∕(x) - x2 and ∕(x) is a quadratic, we know that the degree of g(x)
cannot be larger than 2. Therefore, we must have q(x) = g(x) = 0 for all x. Finally, since ∕(x) = g(x) + x2,
we know that ∕(x) = x2 for all x.
Notice that we have used the Fundamental Theorem of Algebra in this problem. The degree of
∕(x) — x2 is at most 2. If f(x) - x2 is not 0 for all x, then by the Fundamental Theorem of Algebra, it must

221
CHAPTER 7. POLYNOMIAL ROOTS PART I

have no more than 2 roots. However, we know that 2,3, and 4 are all roots of ∕(x) - x2. So, we violate the
Fundamental Theorem of Algebra if f(x) is anything except x2. (If ∕(x) = x2, then f(x) - x2 = 0 for all x,
so we cannot apply the Fundamental Theorem of Algebra to ∕(x) - x2 because its degree is undefined.)
□
Let's see if we can generalize our work in Problem 7.25.

Problem 7.26: Let pi(x) and p2(x) be polynomials of degree at most n. Prove that if pι(x) = p2(x) for
77 + 1 distinct values of x, then p1(x) = p2(x) for all x.

Solution for Problem 7.26: In Problem 7.25, we saw that ∕(x) = x2 for x = 2, 3, and 4, so we examined the
polynomial ∕(x) - χ2 to prove that ∕(x) = x2. Let's try the same general strategy here.
Let X], x2,..., xn+1 be the 71 + 1 distinct values of x for which p1(x) = p2(x). We let p(x) = pι(x) - P2CO∙
Since p1(x,∙) = p2(xi) for 1 < z ≤ 77 + 1, we have ρ(xι) = p(x2) = ∙ ∙ ∙ = p(xπ+ι) = 0, which means that either
degp ≥ n + 1 or p(x) = 0 for all x. However, because pι(x) and p2(χ) have degree at most n, there are
no terms of p(x) with degree greater than n. Therefore, we cannot have degp > n + 1, so we must have
p(x) = 0 for all x. □

Important: Let pι(x) and p2(x) be polynomials of degree at most 77. If pι(x) = pz(xy)

ψ for 77 + l distinct values of x, then pι(x) = p2(x) for all x. We call this the
Identity Theorem for polynomials.

Problem 7.27: Consider the system of equations

U∖ + 8λ2+ 27^3 + 64z?4 =1,

8«i + 27λ2+ 64tz3 + 125/74 = 27,
27a∖ + 64∕72+ 125fl3 + 216/74 — 125,
64/z 1 + 125∕72 + 216«3 ^t^ 343«4 — 343.

These four equations determine z?i, ∕z2, «3/ and ⅜∙ Find zz1 + zz2 + zz3 + «4 and 64« 1 + 27«2 + 8/73 + «4.
(Source: Mandelbrot)

Solution for Problem 7.27: In each equation in the system, the coefficients on the left side form a sequence
of consecutive cubes. The constants on the right are the cubes of consecutive odd integers. We therefore
see that all four equations have the form
?rz3«i + (m + 1)3«2 + (w + 2)3«3 + (777 + 3)3«4 = (2zτz - 1)3,

where m is 1, 2, 3, or 4. The right side of this equation is a cubic polynomial in m. We can also view the
left side as a cubic polynomial, where 777 is the variable and «i, ∕72, ∕z3, and «4 are constants. Since the two
cubic polynomials are equal for the four values m = 1,2,3, and 4, the Identity Theorem tells us that they
must be the same cubic, so we have

777 3Λ] + (∕77 + 1)3Λ2 + + 2)3«3 + (T71 + 3)3∕74 = (2/7/ - 1)3

for all values of m.

222
 7.6. ALGEBRAIC APPLICATIONS OF THE FUNDAMENTAL THEOREM

We wish to evaluate a↑ + a2 + «3 + ⅜. In the expansion of

∕7z3∕71 + (m + l)3α2 + (m + 2)⅛ + (m + 3)3⅜,

the coefficient of m3 is a↑ + a2 + 0.3 + fl4∙ (We don't have to multiply the whole thing out; we can see that
the leading term of each of (m + 1)3, (m + 2)3 and (m + 3)3 is m3 without expanding the cubes.) This must
match the coefficient of m3 in (2m - 1)3, so we have fl1 + a2 + fl3 + fl4 = 8.

We can't easily find 64«i + 27z⅛ + 8<⅞ + «4 just by matching up coefficients. However, since
m3a↑ + (m + l)3α2 + (m + 2)3α3 + (m + 3)3α4 = (2m - 1)3

for all m, we might be able to choose a value of m that enables us to evaluate 64«i + 27zz2 + 8α3 + α4. We
first try m = 4, since this makes the coefficient of a↑ equal to 64. Unfortunately, this makes the coefficient
of zz2 equal to 125. So, we instead try m = -4, which makes the coefficient of a↑ equal to -64 and gives us

-64∕q — 27z?2 - 8fl3 — Λ4 = —729.

Aha! Multiplying by -1 gives us 64αi + 27a2 + 8α3 + α4 = 729. □
Our key step in Problem 7.27 was noticing that all four equations in the given system had the same
form. This enabled us to express all four equations using a single equation.

Problem 7.28: There is a unique polynomial P(x) of the form

P(x) = 7x7 + cιx6 + C2x5 + ∙ ∙ ∙ + C(,x + C7

such that P(l) = 1, P(2) = 2,..., and P(7) = 7. Find P(0). (Source: Mandelbrot)

Solution for Problem 7.28: The statements P(l) = 1, P(2) = 2,..., and P(7) = 7 all have the same form, so
we write them all at once by noting that P(x) = x for x = 1, 2, 3,..., 7. Now, we're in familiar territory.
We rearrange this equation to give P(x) - x = 0 for x = 1, 2, 3,..., 7, so these seven values are all roots
of the polynomial P(x) - x. Because P(x) is a 7th degree polynomial, we know from the Fundamental
Theorem of Algebra that P(x) - x has exactly 7 roots. We've already found the seven roots, so we know
that
P(x) - x = a(x - l)(x - 2)(x - 3)(x - 4)(x - 5)(x - 6)(x - 7)
for some constant a. Because the leading term of P(x) - x is 7x7, the leading term of the right side must
also be 7x7. Therefore, we have a = 7r and

P(x) - χ = 7(x - l)(x - 2)(x - 3)(x - 4)(x - 5)(x - 6)(x - 7).

So, we find P(0) = 7(0 - l)(0 - 2)(0 - 3)(0 - 4)(0 - 5)(0 - 6)(0 - 7) + 0 = -35280. □

Our solution to Problem 7.28 revisits an important use of the Fundamental Theorem of Algebra that
we'll be seeing many more times throughout this book:

Important: Let the polynomial ∕(x) = anx" + ⅜-ιxn 1 + απ-2x" 2 + ∙ ∙ ∙ + a1x + β0 have

ψ roots r↑, T2, r3, ...,rn. Then we can write ∕(x) as

∕(x) = an(x - ri)(x - r2) ∙ ∙ ∙ (x - rn).

223
CHAPTER 7. POLYNOMIAL ROOTS PART I

We'll often find it useful to express a polynomial in terms of its roots instead of its coefficients.
We started this chapter with Factor Theorem. WeTl finish with one more look at relationships
between factors and roots of polynomials.
Let g(x) and ∕(x) be polynomials. Back in Problem 7.4 we showed that it is not necessarily true that
g(x) is a factor of ∕(x) if every root of g(x) is a root of ∕(x). We showed this by considering g(x) = (x —1)I2
and ∕(x) = x{x - 1). Notice that our choice of g(x) has a double root.
What if g(x) does not have any multiple roots?

Problem 7.29: Suppose that the roots of the polynomial g(x) are distinct, and that each of these roots I
is a root of the polynomial ∕(x). Must g(x) be a factor of /(x)?

Solution for Problem 7.29: Let t∕(x) and r(x) be the quotient and remainder, respectively, when ∕(x) is
divided by g(x). We then have ∕(x) = q(x)g(x) + r(x), where r(x) = 0 or deg r < deg g. Let the degree of
g(x) be n, so that it has n roots. Let these roots be Si, S2, ■ ■ ∙, sn. Since g(s1∙) = 0 for each i with 1 ≤ i < n,
we have
∕(s,) = tj(sz∙)g(s1∙) + r(s1∙) = t∕(sz∙) ∙ 0 + r(sz∙) = r(sz∙)
for 1 < i < n. But these s1∙ are roots of ∕(x), so ∕(s,∙) = 0. Therefore, we have r(sl) = 0 for 1 ≤ i ≤ n. But this
means that r(x) has at least n roots, because the sl∙ are distinct. However, we must have deg r < deg g if r
is nonzero. We therefore conclude that r(x) must be 0 for all x, which means that g(x) is a factor of ∕(x).
□

I xeι
*< ises

7.6.1
(a) Show that if f is a polynomial of degree 4, such that ∕(0) = ∕(1) = ∕(2) = ∕(3) = 1 and ∕(4) = 0,
then ∕(5) = -4.
(b) Show that if f is a polynomial of degree n, such that ∕(0) = ∕(1) = ∙ ∙ ∙ = f(n - 1) = 1 and f(ri) = 0,
then /(« + 1) = ~n.
7.6.2 Find all possible polynomials p(x) of degree 3 such that ∕(1) = 0, ∕(2) = 7, ∕(3) = 26, and ∕(4) = 63.
7.6.3 A certain polynomial p(x) leaves a remainder of a when divided by x - fl, a remainder of b when
divided by x - b, and a remainder of c when divided by x - c, where fl, b, and c are different. What is the
remainder when p(x) is divided by (x — fl)(x — b)(x — c)?

*
7.6.4 Solve the system

fl+ b÷ c + d= 1,
8fl + ½b+ 2c + d = 16,
27fl + 9b+ 3c + d = 81,
64a + 16b + 4b + d =256.

Hints: 205

224
 7.7. SUMMARY

*
7.6.5 Let P(x) be a polynomial whose degree is 1996. If P(π) = for n = 1,2,3,... ,1997, compute the
value of P(1998). (Source: ARML) Hints: 324

7.7 Summary

The value r is a root of the polynomial ∕(x) if ∕(r) = 0. Roots of polynomials are also sometimes called
zeros of the polynomial. We say that a polynomial d(x) evenly divides a polynomial ∕(x) if the remainder
is 0 when we divide ∕(x) by d(x). We then call rf(x) a factor of ∕(x).

Important: The Factor Theorem states that the expression x-a, where a is a constant,

ψ is a factor of the polynomial p(x) if and only if p(a) = 0.

Important: The Fundamental Theorem of Algebra states that every one-variable

ψ polynomial of degree n has exactly n complex roots.

Important: Let the polynomial ∕(x) = anxn + ⅜-ιx" 1 + απ-2Xw 2 + ∙ ■ ∙ + apx + «o have
roots rι, r2, r3, ...,rn. Then we can write ∕(x) as

∕(x) = an(x - r1)(x -r2)-∙-(x- rn).

Important: If we find that a is a root of the polynomial p(x), we can simplify our

ψ search for more roots of p(x) by dividing p(x) by x - a. Since a is a root of

p(x), this division will leave no remainder, and we will have

p(x) = (x - a)q(x').

We can then continue our search for roots of p(x) by searching for roots
of q(x).

We discussed several methods for narrowing our search for rational roots of polynomials.

Important: Let ∕(x) be the polynomial

ψ ∕(x) = anxn + αn-ιxπ 1 + flπ-2xπ 2 + ∙ ∙ ∙ + aix + a0,

where all the λi∙ are integers and both an and ao are nonzero. If p and q
are relatively prime integers and f(p∕q) = 0, then p ∣ o0 and q ∣ an. We
call this the Rational Root Theorem.

Important: If ∕(x) is a polynomial and a and b are real numbers such that f(a) and

ψ f(b) have different signs, then there is a root of ∕(x) between a and b.

225
CHAPTER 7. POLYNOMIAL ROOTS PART I

Important: If a polynomial has no constant term, then 0 is a root of the polynomial.

ψ
Important: If the coefficients of a polynomial all have the same sign, then the polyno­

ψ mial has no positive roots. If the coefficients of a polynomial are nonzero

and alternate in sign, then the polynomial has no negative roots.

We applied the Fundamental Theorem of Algebra to various questions about graphs of polynomials,
and also used it to prove the Identity Theorem for polynomials:

Important: Let pι(x) and p2W be polynomials of degree at most n. If pι(x) - p2(x)

ψ for n + 1 distinct values of x, then pι(x) = p2(

*) f°r x∙

One common application of the Identity Theorem is:

Important: If ∕(x) is a polynomial such that

ψ
∕(x) = anxtl + απ-ιx"^1 + an-2xn~2 + ∙ ∙ ∙ + flι% + «0/

and there are at least n + 1 different values of x for which ∕(x) = 0, then
∕(x) = 0 for all values of x. That is, an = an-χ = an-2 = ■ ■ ∙ = ao = 0.

Problem Solving Strategies----------------------------------------------------------------

Concepts: ∙ If all the coefficients of a polynomial you must factor are divisible by
Q l,7¾ thθ same integer (other than 1), then factor that integer out so you
can work with a simpler polynomial.

• If a strategy gives you some information about a problem, but doesn't

completely solve it, then try that strategy again in a slightly different
way.

• We often start our hunt for rational roots by evaluating ∕(-l), ∕(0),
and ∕(1), because these are easy to compute and the results may tell
us where to continue our search.

• To prove a general mathematical statement, it's often helpful to work

out a specific example of that statement, then build your general
proof by following your example.

Extra! French mathematician Adrien-Marie Legendre proved that no polynomial (and no ratio
ll∣*l∣ *∣l∣*l∣* of polynomials) gives a prime number as output for every integer input.

226
 REVIEW PROBLEMS

7.30 For what value(s) of k is x - 2 a factor of x3 + 2kx2 + k2x + k - 4?

7.31 Let ∕(x) = x2 + 4x. For what values of x is ∕(∕(x)) = /(x)? (Source: NYSML)
7.32 Let p(x) = 2x3 - x2 - 5x + 3. Note that p(0) = 3 and p(l) = -1, so there is a root between 0 and 1.
Applying the Rational Root Theorem, we next compute p(l∕2) = 1/2. But there aren't any other fractions
between 0 and 1 such that the numerator divides the constant term of p(x) and the denominator divides
the leading coefficient of p(x). So, how can there possibly be a root between 0 and 1?
7.33 Compute the positive integer x such that 4x3 - 41x2 + 10x = 1989. (Source: NYSML)
7.34 Let ∕(x) be a quartic polynomial with integer coefficients and four integer roots.

(a) If the constant term of f(x) is 6, is it possible for 3 to be a root of /(x)?

(b) If the constant term of ∕(x) is 6, is it possible for 3 to be a double root of /(x)?

r7.35 Why must the coefficients of a polynomial be integers in order to apply the Rational Root Theorem?
It's pretty clear that the leading coefficient and the constant term of a polynomial have to be integers in
order for the Rational Root Theorem to be applicable, but why can we not have non-integer fractions
among the other coefficients?
7.36 Solve the following equations:
(a) 2x3 + 35 = -19x2 - 52x (b) 16y2(y + 3) = 61y - 12

7.37 Find all sets of four consecutive positive integers such that the sum of the cubes of three of them
equals the cube of the fourth.
7.38 Find all the roots of each of the following:
(a) f(r) = βr3 - 31r2 - 31r + 6 (c) ∕z(f) = 12t3 - 40f2 + 13f + 30
(b) g(f) = 6t4 + 4113 + 88t2 + 671 + 14

7.39 Kai has programmed his calculator to accept as inputs two positive integers, a and b, that have no
common divisors. The calculator then compares 30 times the cube of a∣b plus 11 times the square of a/b
to the number that is 12 more than 59a∕b. When these two quantities are equal, the calculator lets him
play a game. Unfortunately, Kai has forgotten what two integers to input to be able to play his game.
What integers should Kai input?
7.40 Solve the following inequalities:
(a) 6t5 + 4f4 > 2f4 (b) 2x4 + 13x3 + 9x2 - 40x + 16 ≥ 0

7.41 In Section 7.4, we noted that "If the coefficients of a polynomial are nonzero and alternate in sign,
then the polynomial has no negative roots."

(a) Suppose some of the coefficients of the polynomial ∕(x) equal 0. Then, if the nonzero coefficients

227
CHAPTER 7. POLYNOMIAL ROOTS PART I

alternate in sign, can we still deduce that ∕(x) has no negative roots?
(b) How can we amend our statement above to address a polynomial that has some coefficients equal
to zero?

7.42 What is the maximum number of points of intersection of the graphs of two different fourth degree
polynomial functions y = p(χ) and y = c∣(x), each with leading coefficient 1? (Source: AHSME)
7.43 Sketch the graphs of each of the following functions. Your graphs don't have to be precisely
correct, but should have the correct intercepts and be above or below the x-axis in the right places.
(a) ∕(x) =-(x2 - l)(x + 4) (b) ∕(x) = (x - l)2(x + 2)(x + 3)

7.44 The graph at right has y-intercepts (0,3), (0, -2), and (0, -4), and it has x-intercept y
! I I j
÷÷+--l
(3,0). Find an equation whose graph matches this graph.
7.45 Let ∕(x) be a polynomial with degree n, where n > 1. Prove that no line intersects
the graph of y - f{x) in more than n points. (Note: Your proof must address all possible
lines, not just horizontal lines.)
7.46 If ∕(x) is a monic quartic polynomial such that ∕(-l) = -1,∕(2) = -4,∕(-3) = -9,
and ∕(4) = -16, find ∕(1). (Source: HMMT)

Challenge Problems

7.47 How many real solutions are there to the equation 2(x - l)2(2x + 3)(x + 5)3 = .001?
7.48 Let ∕(x) = x3 - x2 - 13x + 24. Find three pairs (x, y) such that if y = ∕(x), then x = ∕(y).
7.49 The polynomial g(x) = x3 - x2 - (m2 + m)x + lm2 + 4m + 2 is divisible by x - 4 and all of the roots
of g(x) are integers. Find m.
7.50 Show that if a polynomial ∕(x) leaves the same remainder when divided by x - a and by x + a for
all constants a, then the polynomial does not have any terms with odd degree. Hints: 49
7.51 Given an integer n > 2, for how many different rational numbers x does there exist a polynomial
P(x) of degree n -1 with P(0) = 0, and with all integer coefficients, such that 54x" + P(x) = 315? Hints: 253
7.52 Let P(x) = x4 + ax3 + bx3 + ex + d, where a, b, c, and d are constants. If P(l) = 10, P(2) = 20, and
P(3) = 30, compute
P(12) + P(-8)
10
(Source: ARML) Hints: 42
7.53
* Let ∕(x) = x2 + (∏ - a)x + a, where a and n are integers such that 1 ≤ a < n ≤ 49. For how many
values of n does ∕(x) have irrational roots for all possible values of a?
7.54 Prove that if all coefficients of a quadratic equation are odd integers, then the roots of the equation
cannot be rational.

228
 CHALLENGE PROBLEMS

7.55
* Shown is the complete graph of y = ∕(x), a polynomial function of 1/
degree 10 whose domain is restricted to [1,5]. Function f is symmetric about
x = 3. Compute the number of solutions to the equation ∕(x) = ∕(∕(x)). (Source:
ARML) Hints: 30
7.56
* In order to get to Mars, you must win a video game. The video game
chooses 10 points (a, b), where a and b are single-digit integers, and places a disk
with radius 1/3 on each of the points. You must find a polynomial f such that
the graph of f hits all 10 discs. However, you must choose your polynomial
before seeing where the disks are. Find a polynomial that guarantees you a trip to
Mars. Hints: 151

7.57⅛ Let ∕(x) - xn + a↑xn 1 + α2x" 2 +---- t∙ an-]X + an, where all the coefficients are integers. Suppose
there are four distinct integers p, q, r, and s such that f(p) = f(q) = f(r) = f(s) = 5. Show that there is no
integer k such that f(k) = 8. (Source: Canada) Hints: 217

*
7.58 Find all pairs of nonzero integers (α,b) such that (a2 + b)(a + b2) = (a - b)3. (Source: USAMO)
Hints: 46

Extra! Having learned the quadratic formula, you might wonder if there is a similar formula
∣∣n⅜∣∣∣∣^∣∣∣Φ∣∣∣∣⅛ for polynomials with higher degrees. Finding such formulas was a notable quest for
centuries. In 1545, Girolamo Cardano published his method to solve any cubic and the
method his pupil, Ludovico Ferrari, derived for solving quartic equations.
Both methods were much more complicated than the quadratic formula, and both
were controversial, since Cardono's and Ferrari's work was based on a method showed
to them by Nicolo Fontana Tartaglia for solving equations of the for x3 + bx = c, where
b and c are constants. Tartaglia had not published his results, and had allegedly sworn
Cardano and Ferrari to secrecy.
Following Cardano's and Ferrari's successes, mathematicians spent over two cen­
turies searching for a method to solve quintic equations. They searched in vain. Amateur
mathematician Paolo Ruffini (about whom we read on page 177 with respect to synthetic
division) was one of the first to suggest in print that it was impossible to find a general
solution like those of Cardano and Ferrari for cubics and quartics. Being an amateur,
Ruffini found his work largely overlooked, perhaps because mathematicians of his day
(he published in 1799—over 250 years after Cardano and Ferrari) mostly still believed a
solution for quintics was possible.
Later examination of Ruffini's work would expose some flaws in his arguments, but
he was one of the first to be barking up the right tree. The great mathematician Niels
Abel would finally put the problem to rest in 1824, proving that Ruffini was correct in his
claim that there is no general solution for quintics as there is for quadratics, cubics, and
quartics. See if you can figure out how to use the fact that there is no general solution for
quintics to show that there is no general solution for higher degree polynomials, either.

229
CHAPTER 8. POLYNOMIAL ROOTS PART II

The roots of education are bitter, but the fruit is sweet. - Aristotle

CHAPTER

I
Polynomial Roots Part II

We continue our investigation of roots of polynomials by considering polynomials with roots that are
not rational numbers. We'll consider both irrational roots and nonreal roots. We then will explore a
useful relationship between the roots and the coefficients of a polynomial.

8.1 Irrational Roots

The Pythagoreans, an ancient Greek society of mathematicians and philosophers headed by the great
mathematician Pythagoras, believed that "all is number." However, as we shall see, they were missing
a few numbers.
The Pythagoreans started with the numbers we now know as the positive integers:

1,2,3,4,5,...

They didn't accept zero as a number, nor did they use what we now call negative numbers. They did,
however, use ratios of positive integers, such as

4 19 203 1
3, 4 '-Γ,17-

Nowadays, we use the term rational numbers to describe ratios of integers, including 0 and negative
ratios. Just as we use R to refer to the real numbers and C for the set of complex numbers, we use Z to
refer to the set of all integers and Q for the set of rational numbers.
At first, the Pythagoreans believed that every number could be expressed as the ratio of two positive
integers. Then, one of them made a startling discovery. The Greek Hippasus proved that the square
root of 2 could not be written as a ratio of two positive integers.

230
 8.1. IRRATIONAL ROOTS

We use the term irrational number to describe a real number that is not rational; that is. an irrational
number cannot be written as the ratioof two integer s According to legend, I fippasus made his discovery
while at sea and was rewarded by being thrown overboard. But his fate didn't change the fact that there
arc, indeed, irrational numbers. And the discovery of new numbers didn't stop there. Centuries later,
mathematicians of various cultures would introduce 0, negative numbers, and nonreal numbers

We ll start this section with a proof that V2 is irrational, but don't worry, we won't throw you
overboard if you're able to prove it yourself!

*
Problem

Problem 8.1: In this problem, we prove that yJl cannot be written as a ratio of two positive integers.
(a) Suppose that λ∕2 = p∕q, where p and q are positive integers that have no common factors. Square
this equation and use the resulting equation to prove that p must be even.
(b) Part (a) tells us that p = 2r for some positive integer r. Substitute this into your equation in (a)
and use the result to prove that q is even.
(c) We started by assuming that √2 = p∕q, where p and q are positive integers that have no common
factors. Why do your conclusions in parts (a) and (b) show that this is impossible7

Problem 8.2:
(a) Prove that the ratio of two rational numbers must be rational.
(b) Prove that if V∣∕ is irrational, and b and c are rational numbers such that b + cVd 0, then we
must have b = c = 0.

Problem 8.3: If b, c, and d are rational and V√ is irrational, then let b + c Vd = b-c ∖fd. So, lor e x arnpl∣ ,
we have -2 + 3 V2 = -2 - 3 λ∕2 and 4 - 5 V3 = 4 + 5 V3.
For the following parts, let ∕,∙ = b∣ + cl∙ Vd, where bi, c∣, and d are rational and √d is irrational
(a) Show' that f] + b = + ⅛∕
(b) Show that f√2 = t↑ ■ tι.
(c) Show that if ∕(x) is a polynomial with rational coefficients such that f1 is a root of ∕(x), then t1 is
also a root of ∕(x).

Problem 8.1: Show that V2 is irrational.

Solution for Problem 8.1: We show' that V2 cannot be w'ritten as the ratio of two positive numbers by
using proof by contradiction. To do so, we start by assuming that V2 is rational, and then show' that this
assumption leads to an impossible conclusion. Assuming V2 is rational, it can be written as a fraction
p∕q, where p and q are integers and p/q is in low'est terms. So, we have

231
CHAPTER 8. POLYNOMIAL ROOTS PART II

We square both sides to get 2 = p2∕q2, so 2q2 = p2. Since p and q are integers, this means that p is even.
So, we have p = 2r, for some integer r. Substituting this into 2q2 - p2 gives 2r?2 = (2r)2 = 4r2. Dividing
both sides of 2<j2 = 4r2 by 2 gives us t∕2 = 2r1. So, now we see that q is even, too! However, we assumed
that p/q is in lowest terms, so p and q can't both be divisible by 2.

Because our assumption (that V2 can be written as a ratio of positive integers in lowest terms) leads
to the impossible conclusion that the ratio is also not in lowest terms, we must conclude that our original
assumption is wrong. Specifically, we conclude that V2 cannot be written as a ratio of integers. □

Important: We can sometimes prove a statement by showing that if we assume the

ψ statement is false, then an impossible conclusion results. We call this

strategy proof by contradiction or indirect proof.

Essentially the same argument that we used in Problem 8.1 can be used to show that if d is a positive
integer that is not a perfect square, then >[d is irrational.

Problem 8.2:
(a) Prove that the ratio of two rational numbers must be rational.
(b) Prove that if Vd is irrational, and b, c, and d are rational numbers such that b + c ^∕d = 0, then
we must have b = c = 0. _______________

Solution for Problem 8.2:

(a) Let p and q be two nonzero rational numbers. Because p and q are rational, we have p = m/n and
q = r/s for some integers m, n, r, and s. Therefore, we have

p _ m/n _ ms
q r/s rn'

Since m, n, r, and s are integers, both ms and rn are integers. So, we can write the ratio p/q as the
ratio of two integers, which means p/q is rational.
(b) Ifb + cVd = 0, then we have c^[d~ -b. From here, there are two possibilities. First, we could have
c = 0, which makes b = 0, too. Second, if c ≠ 0, then b ≠ 0 also, and dividing our equation by c
gives √i∕ = -b∕c. This means that we can express Vd as the ratio of two nonzero rational numbers,
-b and c. From part (a), we know that the ratio of these two rational numbers is also rational. But
this means Vd is equal to a rational number, which we are given is not true. So, we conclude that
we cannot have c ≠ 0, which leaves b = c = 0 as the only possibility.

Important: If √d is irrational, and b and c are rational numbers such that b+c Vd = 0,

ψ then b = c = 0.

Back in Section 3.1, we introduced a special notation for complex conjugates. Specifically, we let
a + bi = a - bi. We'll now introduce a similar notation for numbers like 1+2 √3. If b, c, and d are rational

232
 8 1 IRRATIONAL ROOTS

and x/t? is irrational, then we let b + £ V⅛ b -1 Vrf. So, for example, we have - 2 + 3 √2 -2 - 3 √2 and
4 - 5 V3 = 4 + 5 V3. We call b c √∣∕ the radical conjugate of b + < x∕j.

The notation b ÷ < ^V√ is not standard. There isn't a standard notation for radical conjugates the way
there is for complex conjugates. However, this new terminology and notation will help us with our next
proof.

Problem 8.3: Let ll = b∣ + c∣ Vd, where b∣, c∣, and d are rational and ∖∣d is irrational.
(a) Show that f1 + f2 = f1 + t2,
(b) Show that f1f2 = f] ∙ h∙
(c) Show' that if ∕(x) is a polynomial with rational coefficients such that is a root of ∕(.v). then
is also a root of ∕(x).

Solution for Problem 8.3:

(a) We have t] + f2 = (bj + c↑y∕d) + (b2 + c2y<∕d) = (b1 +½) + (cι + C2) y∕d, so f> + /2 = (^,ι + ⅛)~(cι +Q) ^∙
Because L = lη - C]Vd and f2 = b2 - C2yfd, we have

tι + f2 = b1 - C] Vrf + b2 ~ cιy∣d = (b↑ + b2) ~ (ci + ¾) Vd = t↑ + /2-

(b) We have

f√2 = Gh +C1V⅞)(b2 + C2V⅜ = (bιb2 + C1C2^) + ((b1C2 + ⅛cι) Vrf) = (b↑b2 + c↑c2d) ~ ((b1C2 + b2c∣) √7∕)

and h , ⅛ = 0’1 ~ ∩ ~ c2'∕d) = (b↑b2 + c↑c2d) - ((b-lc2 + b2c↑) √d), so b↑t2 = h ∙ ⅛.

(c) Let ∕(x) = anxu + ∩n-↑xn~l + ∙ ∙ ∙ + a↑x + ⅜. Applying parts (a) and (b) gives us

∕(h) = ⅜(h)" + ⅜-ι(h)" 1 ■<------ +∙ flι(h) + ⅜

= a∏(t'f) + ^n-l(f” 1) ^+------ <^ al(h) + ⅜

= ⅜f] + On-lf” 1 + ∙ ∙ ∙ + ∏]t] + flp

= ant" + an-ιt∏ 1 4— + «ifi + «o

= ∕(h)-

If fι is a root of ∕(x), we have ∕(fι) = 0, so f(t↑) = f(t↑) = 0 = 0, which means t↑ is also a root of
f(χ).
□

Important: If b + cVd, w'here b, c, and d are rational and V<? is irrational, is the root
of a polynomial with rational coefficients, then b - c √i∕ is also a root of
the polynomial.

233
CHAPTER 8. POLYNOMIAL ROOTS PART II

WARNING!! The notation ft + c Vd = b - c y[d is not standard! We only introduced

V it here to be able to write our solution to Problem 8.3 more simply. We
won't use this notation again in this book.

Problems

Problem 8.4: Let ∕(x) = x3 - χ - 2.

(a) Show that ∕(x) has no rational roots.
(b) Show that f (x) has at least one irrational root.

Problem 8.5: Find a polynomial with rational coefficients that has 2 - √3 and 3 + 2 λ∕5 as roots.

Problem 8.6: Find a polynomial with rational coefficients that has -ψ3 + V7 as a root.

Problem 8.7: Let g(x) = 4x4 - 16x2 + 15.

(a) Find all roots of g(x).
(b) Factor g(x) completely into factors with rational coefficients.
(c) Factor g(x) completely into factors with real coefficients.

Problem 8.8: Let r = √2 + √3.

(a) Show that r is a root of x4 - 10x2 + 1 = 0.
(b) Prove that r is irrational.

Problem 8.9: Prove that vz5 + V∑5 is irrational.

Problem 8.4: Show that ∕(x) = x3 - x - 2 has at least one irrational root.

Solution for Problem 8.4: We have ∕(0) = -2 and ∕(2) = 4, so there is a root between 0 and 2. By the
Rational Root Theorem, the only possible rational root of ∕(x) between 0 and 2 is 1. However, we have
∕(1) = -2, so 1 is not a root. Therefore, there is an irrational root between 0 and 2. □

Problem 8.5: Find a polynomial with rational coefficients that has 2 - λ∕3 and 3 + 2 √5 as roots.

Solution for Problem 8.5: Let f(x) be our desired polynomial. Because 2 - λ∕3 is a root and f(x) has rational
coefficients, we know that 2 + √3 is also a root. Similarly, because 3 + 2 √5 is a root of ∕(x), so is 3 - 2 √5.
We can now guess that

∕(x) = (x - (2 - √3)) (x - (2 + √3)) (x - (3 - 2 √5)) (x - (3 + 2 √5))

will work. We'll look at a couple ways to multiply this out without getting lost in a forest of square

234
 8.1 IRRATIONAL ROOTS

roots. Both involve taking advantage ol the radical conjugates

First, we might note that because 2 V5 and 2 + V3 are radical conjugates, their sum and their
product arc integers, namelv (2 √3) + (2 ÷ √3) = 4 and (2 √3)(2 + √3) = I. So, by Vieta's Formulas, the
monic quadratic with these numbers as roots is x2 - 4x + 1, which is nice because it has no radicals in its
coefficients Similarly, we have (3 + 2 √5) + (3 - 2 √5) = 6 and (3 + 2 √5)(3 - 2 √5) = 9 - 20 I 1, so the
product of the second two factors in our expression for fix) is x2 — 6x - 11. Therefore, the polynomial

f(x) = (x2 - 4x + l)(x2 - 6x- 11) = x4- 10x3 + 14? + 38x - 11

has rational coefficients and has 2 - √3 and 3 + 2 √5 as roots.

Another way we can use the radical conjugates is to rewrite the product (x - (2 - V5))(λ - (2 + √5))
as a factored difference of squares, and then expand the product:

(x - (2 - √3)) (x - (2 + √3)) = ((χ - 2) + √3) ((χ - 2) - √3) = (x - 2)2 - (√3)2 = ? - 4 v + 1.

Similarly, we can multiply the other two factors of ∕(x) to find

(x - (3 - 2 √5)) (x - (3 + 2 √5)) = ((x - 3) + 2 √5) ((x - 3) - 2 √5) = (x - 3)2 - (2 √5)2 = ? - 6v 11,

so f(x) = (x2 - 4x + l)(x2 - 6x - 11) = x4 - 10x3 + 14x2 + 38x - 11, as before. □

Problem 8.6: Find a polynomial with rational coefficients that has

Solution for Problem 8.6: Clearly, this can't be the only root of the polynomial, since the polynomial must
have rational coefficients. However, it's not immediately clear what other numbers must be roots in
order to create a polynomial with rational coefficients. So, we take an educated guess.

Concept: There's nothing wrong with guessing! Guessing has a long and glorious
Q)≡[j¾ tradition in math and science. Guessing, and then confirming that your
guess is correct, is a perfectly valid way to solve problems.

We know that if 0 + b χ[c, where b and c are rational but √c is not, is a root of a polynomial with
rational coefficients, then 0 - b y/c is also a root. So, we guess that another root of our polynomial must
be - -^3 + √7. The equation ? = 3 + V7 clearly has ± ^3 + √7 as solutions, so a quadratic with these

two ugly numbers as roots is

? - (3 + √7).

That's not so bad. But we still have one irrational coefficient. Writing x2 - (3 + a∕7) as (x2 - 3) - √7, we
see that can get rid of the square root by multiplying by (? - 3) + √7, which gives us the polynomial

(? - 3 - V7)(? - 3 + √7) = (? - 3)2 - (√7)2 = x4 - 6.x2 + 2.

5∕T- √7. □
Success! Notice that the other two roots of this polynomial are t

235
CHAPTER 8. POLYNOMIAL ROOTS PART II

Problem 8.7: Let g(x) = 4x4 - 16x2 + 15.

(a) Find all roots of g(x).
(b) Factor g(x) completely into factors with rational coefficients.
(c) Factor g(x) completely into factors with real coefficients.

Solution for Problem 8.7:

(a) We can factor g(x) as a quadratic by letting y = x2. This gives us

4x4 - 16x2 + 15 = 4y2 - 16y + 15 = (2y - 3)(2y - 5).

Since y = x2, we have

g(x) = (2x2 - 3)(2x2 - 5).
So, the roots of 2x2 — 3 and 2x2 — 5 are the roots of g(x). Therefore, the roots of g(x) are ±Vβ∕2 and
±√Tθ∕2.
(b) If g{x} has a factor of the form ax + b, where a and b are rational, then the rational number
x = -b∣a is a root of g(x). However, we have already found that g(x) has no rational roots.
Therefore, g(x) does not have any linear factors with rational coefficients. So, our factorization
above, g(x) = (2x2 - 3)(2x2 - 5), is the most we can factor g(x) such that all the factors have rational
coefficients.
(c) The roots of g(x) are and ±2φ∕ and the leading coefficient of g(x) is 4, so

g(x) = 4

We also could have factored 2x2 - 3 and 2x2 - 5 in g(x) = (2x2 - 3)(2x2 - 5) as differences of squares
to find
g(x) = (√2x - √3) (√2x + √3) (√2x - √5) (√2x + √5).
The advantage of our first factorization is that it is much easier to read the roots of g(x) than it is
in the second factorization.

□
The final two parts of Problem 8.7 show that we can sometimes factor a polynomial into more factors
if we allow irrational coefficients among the factors than we can if we only allow rational coefficients.
When we factor a polynomial as completely as possible such that each factor has rational coefficients,
we say we "factor the polynomial over the rational numbers." Similarly, if we factor the polynomial
as completely as possible such that each factor has real coefficients, we "factor the polynomial over
the reals." Likewise, we can "factor a polynomial over the integers" (meaning the factors have integer
coefficients) or "factor a polynomial over the complex numbers" (meaning the factors have complex
coefficients).
The Rational Root Theorem also gives us another method to prove that a number is irrational.

Problem 8.8: Let r = ∖∕2 + ∖∕3. Prove that r is irrational.

236
 8.1. IRRATIONAL ROOTS

Solution for Problem 8.8: In Problem 8.4, we deduced that a polynomial had an irrational root by first
showing that it had a real root, then showing that it didn't have any rational roots. This gives us an idea
for the present problem. If we can find a polynomial ∕(x) with rational coefficients but no rational roots
such that ∕(r) = 0, then we can deduce that r is irrational. Problem 8.6 gives us a guide for building a
polynomial with √2 + √3 as a root.

Let ∕(x) be a polynomial with rational coefficients and r as a root. We start by guessing that √2 - √3
is also a root of ∕(x). Because (√2 + √3) + (√2 - √3) = 2 √2 and (√2 + √3)(√2 - √3) = -1, a quadratic
with √2 + √3 and √2 - √3 as roots is
x2 - 2 √2x - 1.

We can use the same tactic we used in our solution to Problem 8.6 to get rid of the square root. We
write x2 - 2 λ∕2x - 1 as x2 - 1 - 2 √2x. Multiplying by x2 - 1 + 2 V2x will get rid of the radicals:

(x2 - 1 - 2 √2x) (x2 - 1 + 2 √2x) = (x2 - 1)2 - (2 √2x)2 = x4 - 10x2 + 1.

So, ∕(x) - x4 - 10x2 + 1 has V2 + √3 as a root. By the Rational Root Theorem, the only possible rational
roots of ∕(x) are 1 and -1. Since ∕(1) = -8 and ∕(-l) = -8, we know that ∕(x) has no rational roots.
Therefore, all the real roots of ∕(x) are irrational. Since V2 + √3 is one of these roots, we know that it is
irrational. □

Solution for Problem 8.9: We'd like to take the same approach as in the previous problem. We let
r = V5 + V25, and we try to find a polynomial ∕(x) with integer coefficients and r as a root.
We start off by guessing that f must have degree 3, because we expect we'll have to cube r to get rid
of the cube roots. We can simplify our steps when cubing r by letting a = 1/5 and b = j∖∕25, so r = a + b.
This makes the algebraic manipulation of cubing r a little easier:

r3 = (a + b)3 = (α + b)(a + b)2 = (a + b)(a2 + 2ab + b2) = a3 + 3a2b + 3ab2 + b3.

Because a3 = 5 and b3 = 25, we have

r3 = 30 + 3a2b + 3ab2.
We can factor 3a2b + 3ab2 on the right to give r3 = 30 + 3ab(a + b).

Concept: When in doubt, factor. Factoring often simplifies expressions.

But what can we do with 3ab(a + b)?

WARNING!! When you make a substitution to help with algebra, don't forget what
*∕ your variables stand for.

237
CHAPTER 8. POLYNOMIAL ROOTS PART II

We have r = a + b and ab = 7125 = 5, so 3ab(a +b) = 15r, and we have r3 = 30 + 15r. Rearranging this gives
r3 - 15r - 30 = 0z and we have our polynomial! The number r = 75 + 725 is a root of f(x) = x3 - 15x - 30
because r3 - 15r - 30 = 0. Clearly, r is positive, so we only have to test f(x) for positive rational roots.
We can quickly work through the possibilities given by the Rational Root Theorem (namely, the divisors

irrational. □

Sidenote: A real number that is not a root of any nonzero polynomial with rational
coefficients is called a transcendental number. For example, the number
72 is not transcendental, because it is a root of x2 — 2. All transcendental
numbers are irrational, because any rational number r is a root of the
polynomial x - r.
Naturally, we also have a name for a number that is a root of some
nonzero polynomial with rational coefficients. Such a number is called an
algebraic number. All rational numbers are algebraic, as are all square
roots. (Of course, as we have seen in several problems in this section,
rational numbers and square roots are not the only algebraic numbers!)
One interesting fact about transcendental numbers is that, given a seg­
ment of length 1, it is impossible to construct a segment of length k with
straightedge and compass if k is transcendental. In 1882, mathematician
Ferdinand von Lindemann used this fact to put to rest a problem that had
baffled mathematicians for millennia. The problem, known as "squaring
the circle," was to construct a circle with area equal to the area of a given
square. Performing this construction essentially requires being able to con­
struct a segment of length √π given a segment of length 1. Ferdinand von
Lindemann proved that π is transcendental, thus putting this problem to
rest at last.

8.1.1 Find the roots of the polynomial x3 - 7x2 + 12x - 4.

8.1.2 Find a monic quartic polynomial with rational coefficients such that 2 - 3 72 and 1-73 are roots
of the polynomial.
8.1.3 Prove that 74 is irrational.

8.1.4 Let h(x) = x4 - 3x2 + 2.

(a) Factor ∕z(x) completely over the rational numbers.

(b) Factor h(x) completely over the reals.

8.1.5 Let p(x) be a polynomial with rational coefficients. Is it true that if p(4+77) = 0, thenp(4-77) = 0?

*
8.1.6 Find a polynomial with integer coefficients that has + 73 as a root. Hints: 153

238
 8.2. NONREAL ROOTS

8.2 Nonreal Roots

Problems

Problem 8.10: Find the roots of the following two polynomials:

(a) ∕(x) = x3 - 2x2 + 3x - 18
(b) ιy(x) = 2x4 - 8x3 - 3x2 + 22x - 40

Problem 8.11: Let p(x) = anxn + an-↑xn 1 + ∙ ∙ ∙ + a↑x + α0 and t∕(x) = amxm be polynomials with real
coefficients, and recall that if z = c + di, where c and d are real, then z = c — di.
(a) Show that zy(z) = z∕(z) for all complex numbers z.
(b) Show that p(z) = p(z) for all complex numbers z.
(c) Show that if p(z) = 0, then p(z) = 0.

Problem 8.12:
(a) Find the monic quadratic with real coefficients that has 1 + i as a root.
(b) Find the monic quadratic with real coefficients that has 1 - z as a root.
(c) Find the monic polynomial ∕(x) of minimal degree such that ∕(1 + z) = ∕(-l + z) = 0 and all the
coefficients of ∕(x) are real.

Problem 8.13: Let ∕(x) = anxn + an--∣xn 1 + ∙ ∙ ■ + uq, where all the α,∙ are real. In this problem, we show
that if r is a nonreal double root of the polynomial ∕(x), then r is also a double root of ∕(x).
(a) Let r = c + di. Find a quadratic g(x) with real coefficients (in terms of c and rf) such that g(x)
evenly divides ∕(x).
(b) Show that when ∕(x) is divided by g(x), the quotient has real coefficients.
(c) Show that r is a double root of ∕(x).

Problem 8.14: Prove that a nonconstant polynomial with real coefficients can be factored into a product
of linear and/or quadratic polynomials with real coefficients.

Problem 8.15: Consider the equation C4Z4 + z'c3z3 + C2Z2 + zc1z + Co = 0, where co, Cj, C2, C3, and C4 are
real constants.
(a) Find a function f such that substituting z = ∕(x) into the polynomial above produces a fourth
degree polynomial in x with real coefficients.
(b) If z = α + bi is a solution to the original equation, where a and b are real, then what value of x
solves your equation from part (a)?
(c) What other value of x must be a solution to your equation from (a)?
(d) Prove that z = -a + bi is a solution to the original equation.
(Source: AHSME)

239
CHAPTER 8. POLYNOMIAL ROOTS PART II

Problem 8.10: Find the roots of the following two polynomials:

(a) ∕(x) = x3 - 2x2 + 3x - 18
(b) g(x) = 2x4 - 8x3 - 3x2 + 22x - 40

Solution for Problem 8.10:

(a) We first note that because the signs of the coefficients of ∕(x) alternate, there are no negative roots
of ∕(x). Next, we use the Rational Root Theorem to hunt for roots by testing divisors of 18 to see
if they are roots of ∕(x). We find that ∕(3) = 0 and
∕(x) = (x - 3)(x2 + x + 6).
We can't factor x2 + x + 6 easily, so we turn to the quadratic formula to find that the roots of x2 + x + 6
are
-l±z√23
X~ 2

So, the three roots of ∕(x) are 3 and ∙

(b) We start by searching for rational roots. We find that x + 2 divides g(x) evenly and gives us
g(x) = (x + 2)(2x3 - 12x2 + 21x - 20).
We continue by searching for roots of 2x3 — 12x2 + 21x — 20. First, we note that there are no negative
roots because the signs of the coefficients alternate (and none of them is 0). So, we test for positive
roots and eventually find that x - 4 is a factor. Dividing 2x3 - 12x2 + 21x - 20 by x - 4 gives us
g(x) = (x + 2)(x - 4)(2x2 - 4x + 5).
As in part (a), we tackle the quadratic with the quadratic formula and find that its roots are
4 ± 2z λ∕6 . r—
x ----------- = 2 ± ι √6.

So, the four roots of g(x) are —2,4,2 - i >∣6r and 2 + i ^fβ.

□
Hmmm.... For both polynomials in Problem 8.10, the nonreal roots come in conjugate pairs (in
other words, if a + bi is a root, then so is a - bi). We've already seen that nonreal roots of a quadratic
with real coefficients must form a conjugate pair. Let's see if we can prove a similar statement for higher
degree polynomials.
As a reminder, we use z to denote the conjugate of the complex number z.

Problem 8.11: Let p(x) = anxn + ⅜-ιXn^1 + ∙ ∙ ∙ + a1x + ⅜ and z∕(x) = amxm be polynomials with real
coefficients.
(a) Show that q(z) = q(z) for all complex numbers z.
(b) Show that p(z) = p(z) for all complex numbers z.
(c) Show that if p(z) = 0, then p(z) = 0.

240
 8.2. NONREAL ROOTS

Solution for Problem 8.11:

(a) In Chapter 3, we showed that for all complex numbers τυ and z, we have wz = w ∙ z. Using this
relationship, along with the fact that am is real, we see that

amzm = a^∙zm = amzm.

Now, we note that

2^ — 2 ∙ 1 — 2 ∙2 _ ∙ ∙ ∙ _ Z , Z * * * Z —-

m z's

which gives us q(z) = a^z™ = am^ = amzm = q{z).

(b) In Chapter 3, we showed that for all complex numbers w and z, we have w + z — w + z. Combining
this with part (a), we have

P(z) = ΛnZn + flM_jZn—1 + ∙ ∙ ∙ + Λ∖Z + Λq

= anzn + αn-ιzn^1 + ∙ ∙ ∙ + afz + af
— anz + flπ~ιz + ∙ ∙ ∙ + Λjz + Λθ
= P(2)λ

as desired.
(c) From part (b), we have p(z) = p(z). So, if p(z) = 0, then we have p(z) = p(z) = 0 = 0.

Important: Let p{x) be a polynomial with real coefficients. If z is a root of p(x), then

ψ z is also a root of p(x). In other words, the nonreal roots of p(x) come in
conjugate pairs.

We can sometimes use this fact to build a polynomial given some, but not all, of its roots.

Problem 8.12: Find the monic polynomial f(x) of minimal degree such that ∕(1 + z) = ∕(-l + z)
and all the coefficients of f(x) are real.

Solution for Problem 8.12: Nonreal roots of a polynomial with real coefficients come in conjugate pairs,
so we know that 1 ± z and —1 ± z are all roots of f(xy). So, the degree of f is at least four.

Now, we could find the product of four linear factors whose roots are the four known roots of f(x).
However, finding quadratics with the given roots is easier. The sum of 1 + z and 1 — z is 2 and the product
(1 + z)(l - z) equals 2. So, Vieta's Formulas tell us that a quadratic with roots 1 ± z is

x2 - 2x + 2.

Similarly, we have (-1 + z) + (-1 - z) = -2 and (-1 + z)(-l - z) = 2, so a quadratic with roots -1 ± z is

x2 + 2x + 2.

241
CHAPTER 8. POLYNOMIAL ROOTS PART II

The product of these quadratics contains all the known roots of ∕(x). Multiplying these quadratics gives
us
(x2 -2x + 2)(x2 + 2x + 2) = x4 + 4,

which is of degree four, so ∕(x) = x4 + 4 is the desired polynomial. □

For all the polynomials with real coefficients that we,ve factored so far, we've been able to write
the polynomial as the product of linear and quadratic factors with real coefficients. In our next two
problems, we prove that we can do so with all polynomials that have real coefficients.

Problem 8.13: Let ∕(x) = anxn + ⅜-ιx"^1 + ∙ ∙ ∙ + «o, where all the αj are real. Show that if r is a nonreal 1
double root of the polynomial ∕(x), then r is also a double root of ∕(x)∙ J

Solution for Problem 8.13: Let r = c + di be a root of ∕(x), where d ≠ 0 because r is nonreal. Because
the coefficients of ∕(x) are real, r = c - di must also be a root of ∕(x), and we have r ≠ r because d ≠ 0.
Therefore, both x - r and x - r are factors of ∕(x), which means there is some polynomial t∕(x) such that

∕(x) = (x - r)(x - r)<∕(x).

Writing (x - r)(x - r) in terms of c and d, and we have

(x - r)(x — r) = x2 — (r + r)x + r∙r = x2- 2cx + c2 - d2,

so we have
∕(x) = (x2 - 2cx + c2 - d2)q{x).
In other words, q(x) is the quotient when ∕(x) is divided by a particular quadratic with real coefficients.
This means all of the coefficients of t∕(x) are real. (Consider our process in Section 6.2 for polynomial
division to see why.)
We know that r is a double root of ∕(x), so x - r must appear twice as a factor in the factorization of
∕(x). We only have one factor of x - r so far in ∕(x) = (x - r)(x - r)^(x), so t∕(x) must contribute another
factor of x - r. Therefore, r is a root of <∕(x). Since all the coefficients of q(x) are real but r is nonreal, we
know that r is also a root of q{x). Therefore, x - r is a factor of t∕(x), which means that r is a double root
of ∕(x). □
Problem 8.14: Prove that a nonconstant polynomial with real coefficients can be factored into a ∣∣
product of linear and/or quadratic polynomials with real coefficients.________________________ ∣

Solution for Problem 8.14: Let p(x) be a polynomial of degree n, for some positive integer n, such that all
the coefficients of p(x) are real. By the Fundamental Theorem of Algebra, p has exactly n complex roots.
Let these roots be r↑, r2,.. ∙, rn, so we have

p(x) = c(x - rι)(x - r2) ■ ∙ ∙ (x - rll), (8.1)

for some nonzero real constant c.

When r; is real, the expression x — rz∙ is a linear factor with real coefficients. Otherwise, r∣ is nonreal,
so η is also a root of p(x), because the coefficients of p(x) are real. This means that (x - Yl') is also one of

242
 8.2. NONREAL ROOTS

the linear factors of p in Equation (8.1). Letting r, = a + bi, we have r∣∙ = a - bi, and

(x - r,)(x -r^i) = x2 - (ri + η)x + rir^i

= x2 - 2ax + a2 + b2.

Since la and a2 + b2 are real, this quadratic has only real coefficients. So, it appears we can pair the linear
factors of p(x) that don't have real coefficients to create quadratic factors that do have real coefficients.
We can't yet simply deduce that all the linear factors with nonreal roots can be paired off this way!
For example, suppose ∕(x) has (x + z)2 and (x - z) among its factors. Then, from our work above, we can
pair off either x + i with the x — i to produce a quadratic with real coefficients. But we can't pair off both
x + i factors with the same x - z factor; once we pair off x + i and x — z, we'll have one x + z "left over." So,
it s not enough to simply note that for every factor x - r, with r1∙ nonreal, there is a factor x - r∕. We have
to be careful about multiplicity.

WARNING!! Watch out for special cases.

This is where our work in Problem 8.13 comes into play. We showed there that if a polynomial with
real coefficients has a double root that is nonreal, then the conjugate of the root is also a double root.
Similarly, we can show that if a nonreal root of a polynomial with real coefficients has multiplicity k
(that is, it appears k times among the roots), then the conjugate of the nonreal root also has multiplicity
k. So, we can pair off linear factors with nonreal roots and not have any extra factors with nonreal roots
"left over."

We're still not quite finished, because at this point, we've only explained that p(x) can be written as
the product of some nonzero constant, linear expressions with real coefficients, and quadratics with real
coefficients. We still have to deal with the constant. Fortunately, that's easy. We can either include it in
one of our linear expressions by writing c(x - r1∙) = ex- cri for any one of the real roots, or we can include
it similarly in one of the quadratics.
So, every polynomial with real coefficients can be written as the product of linear and quadratic
expressions with real coefficients. □

Problem 8.15: Suppose z = a + bi is a solution of the polynomial equation

c4z4 + ic3z3 + C2z2 + icpz + co = 0,

where co, c1, C2, 03, C4, a, and b are real constants. Prove that -a + bi is also a solution. (Source:
AHSME)

Solution for Problem 8.15: Let g(z) = c4z4 + zc3z3 + c2z2 + ic↑z + c0. The coefficients of g(z) are not all
real, so we can't deduce that the nonreal roots come in conjugate pairs. We don't know much about
roots of polynomials with nonreal coefficients, so we'd like to relate g(z) to a polynomial in which all the
coefficients are real. If we had such a polynomial, we'd at least then know that its nonreal roots come in
conjugate pairs.

We notice that the imaginary coefficients are of the odd powers of z, so if we let z = xz, the resulting

243
CHAPTER 8. POLYNOMIAL ROOTS PART II

coefficients of x are real:

g(xi) = c4x4 + c3x3 - c2x2 -c1x + c0.

Aha! Now, we've related g(z) to a polynomial with real coefficients. We let

∕(x) = g(xi) = C4x4 + C3x3 - c2x2 - c1x + c0

and investigate the roots of ∕(x). We are given that g(a + bi) = 0. Because f(x) = g(xi), we therefore
know that f(x) is 0 when xi = a + bi. Solving for x gives us x = (fl + bi)∕i = b - ai, and we have
f(b - ai) = g((b - ai)i) = g(a + bi) = 0. Since f(b - ai) = 0 and ∕(x) has real coefficients, we know that
b - ai = b + ai is also a root of ∕(x). Now, because g(xi) = ∕(x) = 0 when x = b + ai, we see that g(z) = 0
when z = (b + ai)i = -a + bi, as desired. □

Exer<i⅜e⅜

8.2.1 Find all roots of g(x) = 2x3 + 5x2 + 15x + 18.

8.2.2 Find all roots of f(t) = 2t4 - 23t2 + 271 - 36.
8.2.3 Let g(x) be a polynomial with real coefficients such that g(2 + i) = g(3 + i) = 0, and deg g ≤ 4.
Other than 2 + i and 3 + i, find all possible roots of g.
8.2.4 Prove that a fourth degree polynomial with a real root and real coefficients must have another
(not necessarily distinct) real root.
8.2.5 Use the roots of x4 + 4 we found in Problem 8.12 to find the four fourth roots of -1. In other
words, find the four numbers whose fourth power equals -1.
*
8.2.6 A polynomial of degree four with leading coefficient 1 and integer coefficients has two real
zeros, both of which are integers. Which of the following can also be a zero of the polynomial?

1 + i √∏ 1 + i 1 i 1 + i V13
2 ' 2 ' 2 + l'1 + 2' 2 -

(Source: AMC 12) Hints: 101

*
8.2.7 Find all values of x such that x4 + 5x2 + 4x + 5 = 0. Hints: 112

8.3 Vieta's Formulas

We began our discussion of Vieta's Formulas in Section 4.2 by examining the relationship between the
coefficients of a quadratic and the sum and product of the roots of the quadratic. In this section, we
extend Vieta's Formulas to all polynomials.

Extra! Franςois Vieta was a French mathematician in the late 1500's who once noted, "There
< ∣ ⅜ii∣ φ ∣∣⅜ is no problem that cannot be solved." He was likely inspired in this belief by his many
mathematical successes, which were in part due to his being one of the first Europeans
to use letters to represent unknown quantities in his published mathematics.

244
 8.3. VIETA’S FORMULAS

∣fiτ
Problems

Problem 8.16: Let ∕(x) = .v3 - 4x2 + 15.r - 7, and let p, q, and r be the roots of ∕(x).
(a) Why must ∕(x) = (χ - p)(χ - <j)(x - r)?
(b) Expand (x - p)(χ - q)(χ - r).
(c) Compare the result in (b) to ∕(x) = x3 - 4x2 + 15x - 7 to find p + q + r.
(d) Find pq + qr + rp.
(e) Find pqr.

Problem 8.17: Let ∕(x) = c(x - p)(χ - <j)(χ - r), where c, p, q, and r are constants.
(a) What are the roots of /(x)?
(b) Expand the given expression for f (x).
(c) Suppose we write ∕(x) in the form ∕(x) = a3x3 + a2x2 + a↑x + uq. Use your expansion in part (b)
to write each of a3, a2,aγr and ⅜ in terms of c, p, q, and r.
(d) Why must p + q + r = -fl2∕β37
(e) What is the product of the roots of ∕(x) in terms of the coefficients of /(x)?
(f) Express pq + qr + rp in terms of the coefficients of ∕(x).

Problem 8.18: Let p, q, r, and s be the roots of ∕(x) = 3x4 + 2x3 - 7x2 + 9x -10. Find pqr + pqs + prs + qrs.

Problem 8.19: Let ∕(x) = anxn + on_jx” 1 + λπ-2xπ~2 + ■ ∙ ∙ + a-∖x + ciq, and let the roots of the polynomial
∕(x) be r1, r2, r3,..., rn.
(a) Why can we also write ∕(x) as ∕(x) = an{x - r1)(x - r2)(x - r3) ∙ ∙ ∙ (x - rπ)7
(b) When we expand the product in (a), what terms have xm^^1 in them? What is the coefficient of
x"^^1 in the full expansion?
(c) When we expand the product in (a), what is the coefficient of x”-2?
(d) When we expand the product in (a), what is the coefficient of x"-3?
(e) When we expand the product in (a), what is the constant term?
(f) How are the coefficients of a polynomial related to its roots?

Problem 8.20: Let r, s, and t be the solutions to the equation 3x3 - 4x2 + 5x + 7 = 0.
(a) Find r + s + t.
(b) Find r2 + s2 + f2.
(c) Find 7 + 7 + 7.

Extra! Vieta was well-positioned in the French government in the late 1500s. He was a favorite
n» ∣ *n ∣ *∣∣∣* of King Henry IV, and the King relied on Vieta for a variety of mathematical matters.
Vieta was so successful in decoding the messages of Henry IV's enemy, King Philip II of
Spain, that Philip complained to the Pope that black magic was being used.

245
CHAPTER 8. POLYNOMIAL ROOTS PART II

Problem 8.21: One of the roots of x3 + ax + b = 0 is 1 + 2z, where a and b are real numbers.
(a) Find the other nonreal root of the polynomial above.
(b) Find the sum of the roots of the polynomial above.
(c) Find a and b.

Problem 8.22: Evaluate (2 - r)(2 - s)(2 - t) if r, s, and t are the roots of ∕(x) = 2x3 - 4x2 + 3x - 9.

Problem 8.23: The polynomial ∕(x) = x4 + 6x3 + ax2 - 54x + c has four real roots r1, r2, r3, and r4 such
that r-∣ + r2 = 0 and r4 - r3 = 4. Find a and c.

Problem 8.24: The roots Γχ, r2, and r3 of x3 - 2x2 - llx + a - 0 satisfy r↑ + 2r2 + 3r3 = 0. Find all possible
values of a.

Problem 8.16: Let p, z∕, and r be the roots of ∕(x) = x3 — 4x2 + 15x — 7. Determine the values of p + q
pq + qr + rp, and pqr.

Solution for Problem 8.16: By the Factor Theorem, since p, q, and r are the roots of ∕(x), we have

∕(x) = x3 - 4x2 + 15x-7 = k(x - p)(x - q)(x - r)

for some constant k. The coefficient of the x3 term of the expansion on the right must be 1, so k = 1.
Expanding the right side then gives
χ3 - 4x2 + 15x - 7 = x3 - (p + q + r)x2 + (pq + qr + rp)x - pqr.

Since these polynomials are the same, each coefficient on the right must match the corresponding
coefficient on the left, so we have
p + q + r = 4,
pq + qr + rp = 15,
pqr = 7.

For a set of variables x↑, x3, ■ ■ ∙, χn> lθt s⅛ denote the sum of all products of the x1∙ taken A: at a time. So
for example, for two variables Xχ and x2, we have s1 = x1 + x2 and s2 = XχX2. For three variables xχ, x2,
and x3, we have
Sχ = Xχ + X2 + X3,

52 = X1X2 + X2X3 + Λ⅛X1∕

53 = X1X2
*
3∙

The expressions sχz s2,..., sn are known as the elementary symmetric polynomials in the variables
X1, X2,..., x„. We will call them symmetric sums for short. Problem 8.16 suggests a general relationship
between the coefficients and the symmetric sums of the roots of a cubic polynomial.

246
 8.3. VIETA’S FORMULAS

Problem 8.17: Let ∕(x) = α3x3 + π2x2 + ιρx + ⅜, and let p, q, and r be the roots of ∕(x). Find the
symmetric sums of p, q, and r in terms of only ao, a↑, a2, and fl3.

Solution for Problem 8.17: Because p, q, and r are the roots of ∕(x), we have

ι⅛x3 + a2x2 + a↑x + a0 = «3(x - p)(x - q)(x - r)

for all x. The a3 on the right side ensures that the coefficient of x3 on the right side matches the coefficient
of x3 on the left.

Expanding the right side, we have

α3x3 + a2x2 + a1χ + α0 = α3χ3 - (fl3p + ∩3q + α3r)χ2 + (a3pq + a3qr + a3rp)x - a3pqr.

Because these two polynomials are identical, their coefficients must be the same. Therefore, we have

fl3 = «3/
«2 = -«3(p + q + r),
aι = a3(pq + qr + rp),
⅜ = ~a3pqr.

Isolating a symmetric sum of p, q, and r in each of the last three equations gives us
«2
p + q + r =---- ,
«3
«1
pq + qr + rp =
«3
«0
pqr = --.
«3
□

These look a lot like the relationships we found between the roots and coefficients of a quadratic.
Let's see what happens when we consider a quartic.

Problem 8.18: Let p, q, r, and s be the roots of ∕(x) = 3x4 + 2x3 - 7x2 + 9x -10. Find pqr + pqs + prs + qrs. I

Solution for Problem 8.18: We use our work on cubics as a guide. Because p, q, r, and s are the roots of
∕(x), and the leading coefficient of ∕(x) is 3, we can write

∕(x) = 3(x - p)(x - t∕)(χ - r)(x - s).

Each term in the expansion of the right side consists of the product of 3 and one term from each of the
four linear factors. We seek expressions like pqr in this expansion. Such terms will occur when we take
x from one linear factor, and take the constant from each of the other three factors. There are 4 factors
from which to choose the x, and these generate all four linear terms in the expansion of ∕(x). Therefore,
the coefficient of x in the expansion is

7>(-pqr - pqs — prs - qrsf

247
CHAPTER 8. POLYNOMIAL ROOTS PART II

The coefficient of x in ∕(x) is 9, so we have 3(—pqr - pqs - prs — qrs) — 9. Dividing by 3 gives us
pqr + pqs + prs + qrs = -3. □
We've found some relationships between coefficients of a polynomial and symmetric sums of the
roots of the polynomials for quadratic, cubic, and quartic polynomials. This suggests we can find similar
relationships between the roots and coefficients of any polynomial. Let's try.
Problem 8.19: Let ∕(x) = anxn + a„_ixH_1 + an-2xtl~2 + ∙ ∙ ∙ + a↑x + ao, and let the roots of ∕(x) be r-[, r2,1
r3, ...,rn. Find the symmetric sums of the roots of ∕(x) in terms of the coefficients of ∕(x).

Solution for Problem 8.19: Since the leading coefficient of ∕(x) is an, and the roots are r↑,r2,∙∙∙, rn, we can
write
⅜x" + απ-ιxπ~1 + an-2xn~2 + ∙ ∙ ∙ + a-∖x + «0 = ‰(χ - rι)(x - r2)(
* - ^3) ∙ ∙ ∙ (x - rn).
When tackling quadratics and cubics, we expanded the product on the right side and equated the
resulting coefficients in the expansion to the corresponding coefficients on the left side. However, our
product with n binomials is pretty intimidating to expand. To form each term in the expansion, we
choose one of the terms in each binomial, and then take the product of all the chosen terms and the
leading constant, an. The full expansion consists of the sum of all the possible products we can thus
form.
Let's see what terms appear in the expansion of
an(x - rι)(x - r2)(x ~r3)--∙(x- rn).
When we choose x from each of the binomials, we form the term anxn. What about the rest of the
terms in the expansion? Suppose we choose -∏ from the first binomial and x from the remaining n -1
binomials. Then, we form the term -anrkxn~1. Similarly, we can form terms by choosing each -r,∙ in turn,
and choosing x from the remaining n-1 binomials. All these terms together give us
-anr↑xn~1 ~ anr2xn~1 - anr3xn~1-------- anrnxn~1 = -an(∏ + r2 + ∙∙∙ + rn)xn~1 = -ans1xn~1,

where Si is the symmetric sum of the roots taken one at a time.

Perhaps you see where this is going. Let's see what happens when we choose -rz∙ from two of the
binomials. If we choose -Γχ and -r2 from the first two binomials and x from the other n - 2 binomials,
we form the term anrkr2Xn~2∙ Continuing in this way, we choose every possible combination of two r,∙
along with x from the remaining n - 2 binomials to give
⅜rιr2x"^2 + + ⅜r1r4x"~2 + ∙ ∙ ■ + ⅜rr,-ιr∏xtt^2 = ⅜(nr2 + rιr3 + ∙ ∙ ■ + rn-λrn)xn~2 = ans2xn~2,

where s2 is the symmetric sum of the roots taken two at a time.

Similarly, if we choose all possible combinations of k of the -ri and choose x from the other n - k
binomials, we form the term (-l)kanskxn~k, where sk is the sum of the roots taken k at a time. We have
(-V)k because in each term we form by choosing k of the -r1∙, we multiply k terms of the form -ri. So, if k
is even, this product equals the product of the k chosen roots, but if k is odd, the product is the opposite
of the product of the chosen roots.
So, when we expand ∕(x) = atfx - r↑')(x - r2)(x - r3) ∙ ∙ ∙ (x - r∏), we have
y^(x) = anxt' -ans∣x"~1 + ans2x"~2 + ∙ ∙ ∙ + (-l)kallskxn~k + ∙ ■ ∙ + (-l)'' 1α∏sn-ιx + (-l)nansn.

248
 8.3. VIETA’S FORMULAS

Equating the coefficient of each term to the corresponding coefficient in

∕(x) = anxn + an-↑xn~1 + an-2xn~2 + ∙ ∙ ∙ + a↑x + a0

gives us our desired relationship between the roots and coefficients of a polynomial:

Important: Let ∕(x) = anxn + an-↑xn 1 + an-2xtl 2 + ∙ ∙ ∙ + a↑x + ao, and let the roots of

ψ ∕(x) be r↑, r2,.. .f rn. Then, we have

si - r-l+r2 + r3 +■■■+ rn
dn
52 ~ r1r2 + r1r3 + r1r4 + ■ ■ ■ + rn-irn-2 =
ιan
53 = r1r2r3 + r1r2r4 + ■ ∙ ■ + rn-2rn-1rn =
cιn

sn= r1r2r3∙--rn = (-1)"—,

where each s⅛ is the symmetric sum of the roots of ∕(,x) taken k at a time.
We call these equations Vieta's Formulas.

Let's put Vieta's Formulas to work.

Problem 8.20: Let r, s, and t be the solutions to the equation 3x3 - 4x2 + 5x + 7 = 0.
(a) Find r + s + t.
(b) Find r2 + s2 + t2.
(c) Find I+ 1 + 1.

Solution for Problem 8.20:

(a) By Vieta, we have r + s + t = —(-4)∕3 = 4/3.
(b) We already have r + s + t, and we know that squaring r + s + t will give us r2, s2, and t2, so we
expand (r + s + t)2 = (4∕3)2 to find

16
r2 + s2 + t2 +2rs + 2st + 2tr = —.
9
We can isolate the sum of the squares by subtracting 2(rs + st + fr) from both sides:

r2 + s2 + t2 = ^- - 2(rs + st + tr).
9
We recognize rs + st + tr as the symmetric sum of the roots taken two at a time. By Vieta, we have
rs + st + tr = 5/3, so r2 + s2 + f2 = y - 2 ∙ ∣ = ~y.

249
CHAPTER 8. POLYNOMIAL ROOTS PART II

(c) Writing the fractions with a common denominator, we get a ratio of symmetric sums we can
evaluate with Vieta's Formulas:
1 1 1 _ st + tr + rs _ I _ 5
r + s+l rst -Z 7*

□
A symmetric expression in multiple variables is an expression that is unchanged if any two of the
variables are swapped. For example, p2r2 + r2q2 + q2p2 is symmetric because swapping any two of the
variables results in the same expression. The last two parts of Problem 8.20 are examples of writing
symmetric expressions in terms of elementary symmetric polynomials (that is, symmetric sums).

Important: We can think of elementary symmetric polynomials as building blocks

ψ because any symmetric polynomial can be built from them.

You'll have more practice building symmetric polynomials from symmetric sums as Exercises. Now,
we use Vieta's Formulas as a problem solving tool to tackle a few more challenging polynomial problems.

Problem 8.21: One of the roots of x3 + ax + b = 0 is 1 + 2z, where a and b are real numbers. Find a
and b. ?’
liβMMMMMI

Solution for Problem 8.21: Since the coefficients of the cubic are real, one of the other roots must be the
conjugate of 1 +2z, which is 1 -2z. Letting the third root be r, Vieta's Formulas give us r+(l+2z)+(l-2z) = 0,
because the coefficient of x2 in x3+ax+b is 0. Therefore, the third root must be r = -[(l+2z) + (l-2z)] = -2.
Now, by Vieta, we know each of the following:
a = (-2)(l + 2z) + (-2)(l - 2z) + (1 + 2z)(l - 2z) = 1,
b = -(-2)(l + 2z)(l - 2z) = 10.

So, the polynomial is x3 + x + 10. (We could also have found the polynomial by finding the quadratic
with 1 + 2z and 1 - 2z as roots, and then multiplying this quadratic by x + 2.) □

Problem 8.22: Evaluate (2 - r)(2 - s)(2 - t) if r, s, and t are the roots of f(x) = 2x3 - 4x2 + 3x - 9.

Solution for Problem 8.22: Solution 1: Vieta. When we expand the desired expression, symmetric sums
pop out:
(2 - r)(2 - s)(2 - t) = (4 - 2(r + s) + rs)(2 - t)
= 4(2 -t) - 2(r + s)(2 -1) + rs(2 - t)
= 8 - 4t - 4r + 2rt - 4s + 2st + 2rs - rst
= 8 - 4(r + s + t) + 2(rf + st + rs) - rst.
From Vieta's Formulas, we have r + s + t = -(-4)∕2 = 2, rt + st + rs - 3/2, and rst = -(-9)∕2 = 9/2.
Therefore, we have
9 3
(2 - r)(2 - s)(2 - t) = 8 - 4(r + s + t) + 2(rf + st + rs) - rst = 8 - 8 + 3 - - = --.

250
 8.3. VIETA’S FORMULAS

Solution 2: Recognize the desired expression. The desired expression looks a lot like the factored form
of the polynomial. Because the coefficient of the leading term of ∕(x) is 2, we have
∕(x) = 2x3 - 4x2 + 3x - 9 = 2(x - r)(x - s)(x - t),
where r,s, and t are the roots of ∕(x). Letting x = 2 gives us the desired expression on the far right, and
we have
2(2 - r)(2 - s)(2 - f) = ∕(2) = 2 ∙ 23 - 4 ∙ 22 + 3 ∙ 2 - 9 = -3.
Therefore, we have (2 - r)(2 - s)(2 - f) = -3/2. □

In each of the previous two problems, we have suggested solutions or significant steps that don't
require Vieta s Formulas, in addition to providing approaches that use Vieta's Formulas.

WARNING!! While Vieta's Formulas are often very useful, don't forget about the
* many other tools we have for working with polynomials.

Problem 8.23: The polynomial ∕(x) = χ4 + 6x3 + ax2 - 54x + c has four real roots rj, r2, r3, and r4 such
that r1 + r2 = 0 and r4-r3 = 4. Find a and c.

Solution for Problem 8.23: We have a problem involving the roots and coefficients of a polynomial, so we
try using Vieta. We have information about the sum of two roots, so we start by focusing on the sum of
all four roots. Vieta's Formulas tell us
∏ + Γ∑ + 7⅛ + r4 - ~6.

Since r1 + r2 = 0, we have r3 + r4 - -6. Adding this to r4 - r3 = 4 gives us 2r4 = -2, so r4 = -1 and

r3 = -5. However, we don't know r1 and r2, so how do we find a and c?

Concept: When stuck on a problem, look for information you haven't used yet.

We haven't used the fact that the coefficient of x is -54, so we focus on that. Vieta's Formulas give us
rιr2r3 + r↑r2r4 + r1r3r4 + r2r3r4 = 54.
Since r2 = -r4, r4 = -1, and r3 = -5, we have
5r2 + r2 + 5r4 - 5r1 = 54,
so r2 = 9, which means r1 and r2 are 3 and -3. (It doesn't matter which is which.) Letting r1 be 3 and r2
be -3, we can now use Vieta's Formulas to compute
a = r1r2 + r1r3 + r1r4 + r2r3 + r2r4 + r3r4 = -4
and c = r↑r2r3r4 = -45. Note that we could have used a little clever factorization, together with the fact
that r↑ + r2 = 0, to compute a more quickly:
« = rιr2 + rιr3 + r-ir4 + r2r3 + r2r4 + r3r4
= r-ir2 + rι(r3 + r4) + r2(r3 + r4) + r3 + r4
= r↑r2 + (r1 + r2)(r3 + r4) + r3r4 = -9 + (0)(-6) + 5 = -4.
□

251
CHAPTER 8. POLYNOMIAL ROOTS PART II

Concept: Many problems can be solved by using Vieta's Formulas in conjunction

O≡^ with given information about the roots or coefficients of polynomials.

Problem 8.24: The roots r↑, r2, and τ⅛ of x3 - 2x2 - llx + a = 0 satisfy rι + 2r2 + 3^3 - Im­
possible values of a.

Solution for Problem 8.24: Vieta's Formulas give us several equations involving the roots of the polyno­
mial. Combining these with r↑ + 2f⅛ + 3?'3 = 0 gives us the system

r^ι ÷ 2r2 ÷ 3r⅛ — 0,

ri + r2 + r3 = 2,
rιr2 + r1r3 + r2r3 = -11,
r↑r2r3 = -a.

Yikes! Four equations, four variables, and only two of the equations are linear. Let's try to solve the first
three equations for ∏, r2, and r3, and then use the result to find a.
Subtracting the second equation from the first eliminates r↑ and gives r2 + 2r3 - -2. So, we have
r2 = —2 — 2r3. We can also write r-[ in terms of r3 by subtracting twice the second equation from the first.
This gives us -r↑ + r3 - -4, so η = 4 + r3. Substituting these expressions for ∏ and r2 into the third
equation gives us an equation with only r3,

(4 + r3)(-2 — 2r3) + (4 + t3)(r3) + (—2 — 2r3)(r3) = —11.

Simplifying this equation gives us 3r2 + 8r3 - 3 = 0. Factoring the quadratic gives (r3 + 3)(3r3 - 1), so
r3 = 1/3 or r3 - -3. When r3 = 1/3, we have ∏ = 13/3, r2 = -8/3 and a = -r∖r2r3 = 104/27. When
r3 = —3, we have ∏ = 1, r2 = 4, and a = -r↑r2r3 = 12. □
The key step in solving the unusual system of equations we faced in Problem 8.24 was solving for
Γ] and r2 in terms of r3. This allowed us to create an equation with one variable, which we could then
solve for that variable.

Concept: One useful approach when solving a system of equations is to solve for
all but one of the variables in terms of the remaining variable. Then, use
O≡≡≡ substitution to create an equation involving only this final variable.

Ixeπi⅜e⅛

8.3.1 Find the product of the roots of 3x3 - 7x2 + 2x - 6 = 0.

Ill
8.3.2 Let v(x) = x3 - 5x2 + 12x - 19 have roots a, b, and c. Find the value of — + — + —.
rv ' ab be ca
8.3.3 If two factors of x3 - t↑x + t2 are x + 2 and x - 1, find the roots of x2 - t∖X + t2.
8.3.4 Let a and b be real numbers, such that one of the roots of x3 + ax2 - 4x + b = 0 is 1 + i. Find the
other two roots.

252
 8.4. USING ROOTS TO MAKE EQUATIONS

8.3.5 Find (2 + r)(2 + s)(2 + t)(2 + ιι) if r, s, t, and u are the roots of ∕(x) = 3x4 - x3 + 2x2 + 7x + 2.
8.3.6 Find the two values of k for which 2x3 - 9x2 + 12x - k has a double root. (Source: ARML)
*
8.3.7 For certain real values of a, b, c, and d, the equation x4 + ax? + bx2 + ex + d = 0 has four nonreal
roots. The product of two of these roots is 13 + i and the sum of the other two roots is 3 + 4z. Find b.
(Source: AIME)

8.4 Using Roots to Make Equations

Vieta's Formulas are not the only way to relate roots of a polynomial to the coefficients of the polynomial.
An even more straightforward way is to note that if r is a root of f, then ∕(r) = 0.
1111____________
II
Problems

Problem 8.25: Let r be a root of x2 - x + 7.

(a) Show that r3 = r2 - 7r.
(b) Find the value of r3 + 6r + τι.

Problem 8.26: In this problem, we find the sum of the 20th powers of the roots of z20 - 19z + 2.
(a) Suppose r-[ is a root of the polynomial. Write an equation in terms of rj.
(b) Solve the equation in part (a) for r20.
(c) Use part (b) as inspiration to find an expression for the sum of the 20th powers of the roots.
(d) Use your expression in (c) and the original polynomial to evaluate the sum.

Problem 8.27: Let a, b, c be the roots of x3 + 3x2 — 24x + 1=0. All three roots are real. Prove that
tya + ^∖∕b + -tyc = 0.
(a) Find a polynomial f(a) such that = f(a). Hints: 41, 69
(b) Use part (a) to show that i∣a + Vb + i/c = 0.

Problem 8.25: Let r be a root of x2 - x + 7. Find the value of r3 + 6r + π.

Solution for Problem 8.25: We could find the possible values of r, and then evaluate the given expression
for each of them. But that would take a lot of time. Instead, we note that r being a root of x2 - x + 7
means that r2 — r + 7 = 0, and then relate this equation to the expression r3 + 6r + π.
Solution 1: From r2 - r + 7 = 0, we have r2 = r - 7, so we can substitute r - 7 for f2 until we reduce
r, + 6r + π to an equivalent linear or constant expression:

r3 + 6r + π = r (r2) + 6r + π = r(r-7) + 6r + π = r2-r + π = (r-7)-r + π = π-7.

So, we have r3 + 6r + π = π-7. Notice that all our steps hold for any r that satisfies r2 - r + 7 = 0. In

253
CHAPTER 8. POLYNOMIAL ROOTS PART II

other words, we have r3 + 6r + π = π- 7no matter which root of x2 - x + 7 we take to be r.

Solution 2: Looking for a way to relate r2 — r + 7 = 0 to the cubic expression we're trying to evaluate,
we multiply our equation by r to produce an r3, which gives us r3 - r2 + 7r = 0. Therefore, we have
r3 = r2 - 7r, so r3 + 6r + π = r2 - r + π. We know that r2-r = -7, so we have r3 + 6r + π = r2-r + π = π-7.
□
Important: Vieta's Formulas are not the only way to use a polynomial to create

ψ equations in terms of the polynomial's roots. We can substitute the roots

into the polynomial itself! Specifically, if x = k is a root of f(x), then
∕(fc) = 0.

Problem 8.26: Find the sum of the 20th powers of the roots of z20 - 19z + 2.

Solution for Problem 8.26: Let the roots be n, r2,..., r20∙ We seek
r20 + r20 + r20 + ... + r20.

We can use Vieta to evaluate the symmetric sums of the roots, but expressing the sum of the 20th powers
of the roots in terms of these symmetric sums is daunting. So, we look for another approach.
Because ∏ is a root of z20 - 19z + 2, we have
r20 - 19r1 + 2 = 0.
We're interested in adding the 20th powers of the roots so we solve this equation for r2° to find that
r2° = 19r1 - 2. Aha! We can do the same for every root. Adding the resulting equations to get the sum
of the 20th powers gives us
r20 + r20 + ... + r20 = 19(ri+r2 + ... + r2o)_2(2O).

By Vieta, the sum of the roots of z20 - 19z+2 is 0, so the sum of the 20th powers of the roots is -2(20) = -40.
□
Problem 8.27: Let a, b, c be the roots of x3 + 3x2 - 24x + 1 = 0, and all three roots are real. Prove that

Solution for Problem 8.27: What's wrong with this solution:

Bogus Solution: We start by working backwards, hoping to produce an equation we

know how to prove is true. We isolate yf∩ to get tya = -y∕b - ∖∕c.
Cubing both sides and rearranging gives

a + b + c = -3 ^fb2c - 3 Vbc3 = 3 yfbc(- fb - y∕c).

Aha! We recognize that -y∕b- ⅛[c = -fa from above, and we have
a + b+c = 3 y∕iibc. Turning to our polynomial, we use Vieta to see that
a + b + c = -3 and abc = -1, so we do indeed have a + b + c = 3 ∖Γabc.
Then, reversing our steps above, we find that ^∕a + fb + ^fc = 0, as
desired.

254
 8.4. USING ROOTS TO MAKE EQUATIONS

The error in this argument is a subtle one, and had we written our proof forwards, we would have
found it.

WARNING!! When you solve a problem by working backwards, always write your
a solution forwards. This will help you catch flaws in your reasoning.

Here, when we write our solution forwards, we see that we cannot go from a + b + c = 3 ^∖[abc to

a + b + c = 3 ^!bc{- ^!b - -∖[c),

because we haven't shown that 1/a = — y/b — ψc. So, our steps are not reversible! WeTl have to find
another way.

The main complication in this problem is the cube roots. It would be nice if could get rid of them.

Concept: Focus on removing the main complication in difficult problems.

O≡≡g

We get rid of cube roots by cubing, so we look for cubes in the problem. We first might rewrite the
polynomial equation as x3 = —3x2 + 24x - 1, but this doesn't seem to help. Then, we notice that the
polynomial itself is almost a cube. Specifically, the polynomial equals (x + 1)3 - 27x, and 27 itself is a
cube! So, now our equation is (x +1)3 - 27x = 0, and we have a path to our solution. Adding 27x to both
sides, and then dividing by 27, gives
(x + 1)3
X~ 27 ‘

We know that a, b, and c are solutions to this equation, so we have a = Taking the cube root of
both sides gives
{μ + 1)3 _ a + 1
27 ^ ^^3~'
and likewise for b and c. No more cube roots! We now have
a+b+c
3
Using Vieta, we have a + b + c = —3, so we have the desired + ^∖∕b + i/c = 0. □

111L
Exercises

8.4.1 Let t be a root of ∕(x) = x3 - x + 2. Evaluate f6 - f2 + 4f.

8.4.2 Let r be a root ofx2 + 7x + 17 = 0. Compute each of the following:

(a) (r + 3)(r + 4).

(b) (r-2)(r + 9).
(c) * (r - l)(r + 2)(r + 8)(r + 5). Hints: 84

255
CHAPTER 8. POLYNOMIAL ROOTS PART II

8.4.3
* Find the sum of the 20th powers of the roots of z20 - 19z2 + 2 = 0. Hints: 52
8.4.4
* Let a, b, c, and d be four distinct real numbers such that a and b are the roots of the equation
χ2 - 3cx - 8d = 0, and c and d are the roots of the equation x2 - 3ax - 8b = 0. Find a + b + c + d. (Source:
Costa Rica)

8.5 Summary

Important: Let p(x) be a polynomial with real coefficients. If z is a root of p(x), then

ψ z is also a root of p(x). In other words, the nonreal roots of p(x) come in
conjugate pairs.

Important: If b + c ^∖[d, where b, c, and d are rational and ^∖fd is irrational, is the root
of a polynomial with rational coefficients, then b — c y[d is also a root of
the polynomial.

Important: We can sometimes prove a statement by showing that if we assume the

ψ statement is false, then an impossible conclusion results. We call this

strategy proof by contradiction or indirect proof.

Important: Let ∕(x) = αnxn + an-∖Xn 1 + an-2xn 2 4------ 1- a↑x + a$, and let the roots of
∕(x) be rι, r2, ■.., rn. Then, we have

si = r1 + r2 + r3 + ∙ ∙ ∙ +

S2 = rιr2 + rι^3 + + ■ ∙ ∙ + r∏-ιr∏-2 = ----- ,

∩n

S3 = nr2r3 + r1r2r4 + ■ ∙ ■ + rn-2rn-ιrn = -⅛-2,

cιn

sn= r↑r2r3∙∙∙rn = (-l)"-,

where each s⅛ is the symmetric sum of the roots of ∕(x) taken k at a time.
We call these equations Vieta's Formulas.

Important: Vieta's Formulas are not the only way to use a polynomial to create
equations in terms of the polynomial's roots. We can substitute the roots
into the polynomial itself! Specifically, if x = k is a root of ∕(x), then
∕(∕c) = 0.

256
 REVIEW PROBLEMS

Things To Watch Out For!___________________________________________

• Watch out for special cases.

• When you make a substitution to help with algebra, don't forget what your variables stand for.

• When you solve a problem by working backwards, always write your solution forwards. This will
help you catch ∩aws in your reasoning.

Problem Solving Strategies__________________________ ________________

Concepts: • When facing a new type of problem, comparing it to similar problems

you have already solved can guide you to a solution.

• There's nothing wrong with guessing! Guessing has a long and glo­
rious tradition in math and science. Guessing, and then confirming
that your guess is correct, is a perfectly valid way to solve problems.

• When in doubt, factor. Factoring often simplifies expressions.

• When stuck on a problem, look for information you haven't used yet.

• Many problems can be solved by using Vieta's Formulas in con­

junction with given information about the roots or coefficients of
polynomials.

• One useful approach when solving a system of equations is to solve

for all but one of the variables in terms of the remaining variable.
Then, use substitution to create an equation involving only this final
variable.

• Focus on removing the main complication in difficult problems.

ι
Review Problems

8.28 Find the roots of each of the following polynomials:

(a) ∕(r) = 2r3 + 7r2 - 4r - 21 (b) g(s) = s3 + 8s + 24

8.29 One of the roots of x4 + 9x3 + 48x2 + 78x — 136 = 0 is —3 + 5i. Find the other three roots.

257
CHAPTER 8. POLYNOMIAL ROOTS PART II

8.30 Let ∕(x) be a polynomial with degree 6 and rational coefficients. The polynomial has five distinct
roots: 1,2 - i, 2 + i, 1 — i, and 1 + z.

(a) Explain why the polynomial must have a double root.

(b) Which root is the double root?

8.31 Let rι = α + bi and r½ = a - bi, where a and b are nonzero real numbers. Find the minimum degree
of a polynomial with real coefficients and roots r∖i and r2
*.
8.32 One root of x3 - llx2 + 29x - 7 = 0 is 2 + √3. Find the other two roots.
8.33 Find the monic polynomial with minimal degree and integer coefficients such that 3 — 2z'V5 and
√3- 1 are roots of the polynomial.
8.34 Prove that if c is a positive integer that is not a perfect square, then y[c is irrational.

8.35 Prove that -^2 - √2 is not rational.

8.36 Factor 8 - 2y + 4y3 - y4 completely

(a) over the integers.

(b) over the reals.
(c) * over the complex numbers.

8.37 Let p(x) be a cubic polynomial with real coefficients.

(a) Use the graph of p(x) to give an intuitive explanation for why p(x) must have at least one real root.
(b) Find an algebraic proof that p(x) has at least one real root using the tools we developed in this
chapter. Hints: 170

8.38 Let f(x) be a polynomial with rational coefficients.

(a) If √2 is a root of ∕(x), must - √2 also be a root of /(x)?

(b) If ^∕2 is a root of ∕(x)z must - V? also be a root of /(x)?

8.39 Let g(x) = x4 - 5x3 + 2x2 + 7x - 11, and let the roots of g(x) be p, q, r, and s. Find the following:

(a) pqr + pqs + prs + qrs (c) p2 + z∕2 + r2 + s2

1111
(b) -+-+-+- (d) p2qrs + pq2rs + pqr2s + pqrs
p z∕ r s

8.40 What is the sum of the reciprocals of the roots of the equation + 1 + " = θ∙ (Source: AMC
10)
8.41 If zz ± bi are complex roots of the equation x3 +qx + r = 0, where a, b, q, and r are real numbers and
b ≠ 0, then find q in terms of a and b. (Source: AHSME)
8.42 A cubic polynomial p(x) has leading coefficient!, all real coefficients, and p(3-2z) = 0. Ifp(O) = -52,
find p(x).

258
 CHALLENGE PROBLEMS

8.43 The equation x3 - 4x2 + 5x -1.9 = 0 has real roots r, s, and t. Find the length of an interior diagonal
of a box with sides r, s, and t. (Source: Mandelbrot)
8.44 Let x be a real number such that x3 + 4x = 8. Determine the value of x7 + 64x2. (Source: HMMT)

Challenge Problems

8.45 Let r be a root of x2 — 2x + 3. Find the value of r4 - 4r3 + 2r2 + 4r + 3.

8.46 Let f(x) = xn + α↑x,'~γ + a2xn~2 + 4x + 7, where a∖ and «2 are constant. Prove that the sum of the
squares of the roots of f(x) is the same for all integers n≥4.
8.47 Find the polynomial P of smallest degree with rational coefficients and leading coefficient 1 such
that P (√49 + = 4. (Source: NYSML) Hints: 62

8.48 Find all roots of the polynomial zy3 - 8y2 — 22zy + 21. Hints: 43
8.49 Let %ι and X2 be the roots of the equation x2 - (a + d)x + (ad - be) = 0. Show that x3 and x3 are the
roots of the equation
y2 - (a3 + d3 + 3abc + 3bcd)y + (ad - be)3 = 0.
(Source: Hungary)
8.50 Let rι, r2, r3 be the 3 zeroes of the cubic polynomial x3 - x - 1 = 0. Then, the expression

rι(r2 ~ r3)2 + r2(r5 - r1)2 + r3(r1 - r2)2

equals a rational number. Find this number. (Source: ARML)

8.51 Let f(x) be a polynomial with rational coefficients. If >∕2 is a root of f(x), must - λ∕2 also be a root
of∕(x)7
8.52 Determine (r + s)(s + t)(t + r) if r, s, and t are the three real roots of the polynomial x3 + 9x2 - 9x - 8.
(Source: Mandelbrot) Hints: 147
8.53 Show that if the roots of x3+αx+b = 0 are rational numbers m, n, and p, then the roots of mx2+nx+p
are also rational. Hints: 172
8.54 Let p, q, and r be the distinct roots of x3 - x2 + x - 2 = 0. Find p3 + q3 + r3. (Source: AHSME)
Hints: 28
8.55 Let n be a positive integer. Prove that

— -j- — * ∙ ∙ + — — ---------- — ----------

12 n 1∙2 1∙3 (n - l)n

where each term consists of (-l∕+W divided by a product of k of positive integers from 1 to n, and all
possible such terms are included in the sum. (Source: Mandelbrot) Hints: 167,148

259
CHAPTER 8. POLYNOMIAL ROOTS PART II

8.56 Consider all lines that meet the graph of y = 2x4 + 7x3 + 3x-5 in four distinct points, (x1, y1),
(x2,y2)∕ U3,y3)∕ and (x¼y4)∙ Show that
X1 + X2 + X3 + M
4
is independent of the line and find its value. (Source: Putnam)
8.57 Let f(x) = αnxn + an-↑xn~1 + ■ ■ ■ + a1x + a0, and let tk be the stun of the fcth powers of the roots of
∕(x)∙ Prove that if k ≥ n, then antk + απ-ι⅛-ι + ‰-2⅛-2 + ∙ ∙ ∙ + flo⅞-n = θ∙ Hints: 149
1+a 1 +β 1+γ
8.58 If a, β, γ are the roots of x3 - x -1 = 0, compute
1-a w +w (Source: Canada) Hints: 179

8.59** Find the remainder when x81 + x48 + 2x27 + x6 + 3 is divided by x3 + 1.

8.60
* Let α, b, c be the roots of x3 — 9x2 + llx — 1 = 0, and let s = y/a + Vfe + y[c. Find s4 - 18s2 — 8s.
(Source: HMMT)
8.61 ★ Let and ct2 be the roots of the quadratic x2 - 5x - 2 = 0, and let βι, β2, and ∕⅛ be the roots of
the cubic x3 — 3x - 1 = 0. Compute (aq + βι)(a↑ + β2)(tfι + β3)(^2 + ∕h)(n2 + β2)(a2 + β3)∙ (Source: WOOT)
Hints: 164,105
8.62
* Let ∕(x) be a polynomial in x of degree greater than 1. Define g1∙(x) by gι(x) = ∕(x), and
g∕f+ι(x) = ∕(gk(^))∙ Lθt hi be the average of the roots of gk. Determine r⅛9 if D9 = 89. (Source:
Mathematical Mayhem) Hints: 273
8.63
* Let α be a root of the cubic x3 - 21x + 35 = 0. Prove that <x2 + 2α — 14 is a root of the cubic.
Hints: 74, 7

Extra! Descartes Rule of Signs is one more tool that helps us search for polynomial roots.
∣∣⅜i∣ ∣⅛ ∣∣∣⅜ ∣∣∣φ To use it, we order the nonzero coefficients of ∕(x) from the coefficient of the highest
degree term to the coefficient of the lowest degree term. We similarly order the nonzero
coefficients of ∕(-x). Let the number of sign changes in the list for ∕(x) be z and the
number of sign changes in the list for ∕(—x) be ∕. Descartes Rule of Signs then tells us the
following:

• The number of positive roots of ∕(x) is z - 2k for some nonnegative

*
integer k.

• The number of negative roots of ∕(x) is j - 2k for some nonnegative integer k.

So, for example, let ∕(x) = 3x5 - 2x2 - x +1. There are two sign changes when reading
the coefficients from the highest degree term to the lowest, so there are either 2 positive
roots or 0 positive roots. (That is, we cannot have exactly 1,3,4, or 5 positive roots.) Since
∕(-x) = -3χ5 - 2x2 + x +1, there is only one sign change when we read the coefficients of
∕(-x) from the highest degree term to the lowest. Therefore, there is exactly 1 negative
root. Combining our possibilities for the numbers of positive and negative roots with
the fact that 0 is not a root of ∕(x), we see that we can have at most 3 real roots. Since
deg∕ = 5, the polynomial has 5 roots, so there must be at least 2 nonreal roots.

260
Ideas are the factors that lift civilization. They create revolutions. There is more dynamite in an idea than in many
bombs. - Bishop Vincent

ι
CHAPTER

L_ Factoring Multivariable Polynomials

An expression or function with more than one variable is a multivariable polynomial if it is a sum of
terms such that each term is a product of a constant and a nonnegative integer power of each variable.
For example, the expressions

r2t5
x3y2 -3y + 2t, -y- +r + rst -7, and a2b + b2c + c2a

are multivariable polynomials, while

- + -, λ∕x2 + y3 - 3, and ry/s - 3

V x v ∙7

are not polynomials.

We extend the idea of degree to multivariable polynomials by defining the degree of a term to be the
sum of the exponents of the variables in the term. For example, the degree of 2x3y7 is 3 + 7 = 10. The
degree of a multivariable polynomial is defined as the maximum degree of the terms in the polynomial.
For example, the degree of a5b + a4b2 + a3b3 is 6 and the degree of x9 + 2xy3 + xy is 9.

In this chapter, we use some of our strategies for working with single-variable polynomials to develop
strategies for dealing with multivariable polynomials. We also turn around and use some of these new
strategies to tackle problems with one variable.

9.1 Grouping

In this section, we factor expressions by splitting them into groups of terms, factoring each group, and
then discovering that the factored groups have factors in common.

261
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

Problems

Problem 9.1:
(a) Find a common factor of x3 - x2 and 4x - 4.
(b) Factor x3 - x2 + 4x - 4 completely into factors with real coefficients.

Problem 9.2: In this problem, we find all ordered pairs of positive integers (x, y) such that we have
6xy + 4x + 9y + 6 = 253.
(a) Factor 6xy + 4x.
(b) Factor 9y + 6.
(c) Factor 6xy + 4x + 9y + 6.
(d) Use your factorization from part (c) together with possible factorizations of 253 to solve the
problem.

Problem 9.3: Write each of the following expressions as the product of two binomials, plus (or minus)
a constant. For example, we have xy + x + y = (x + l)(y + 1) - 1.
(a) ∩b + 3a - 5b (c) 6x2y2 - 3x2 — 2y2
(b) 2mn + 14m - 3n

Problem 9.4: Find all positive integers m and n, where n is odd, that satisfy ⅛ + ⅛ = ⅛∙ (Source: Great
Britain)

Problem 9.5: Factor 2x5 + x4y - 21x3y2 + 16x2 + 8xy - 168y2 as the product of linear and quadratic
polynomials with real coefficients.

Problem 9.6: Factor afr(c2 + d2) + crf(a2 + b2).

Problem 9.7: In this problem, we find all values of x that satisfy

6 1 7 12 λ
---------- 1—. — ∣—- _________ 1—— —------- = 0.
√χ - 8 - 9 √x - 8 - 4 √χ - 8 + 4 √x - 8 + 9

(a) Let y = √x - 8. Solve the resulting equation for y.

(b) Find all values of x that satisfy the original equation. (Source: ARML)

Problem 9.1: Factor x3 - x2 + 4x - 4 completely into factors with real coefficients.

Solution for Problem 9.1: We note that x3 - x2 = x2(x - 1) and 4x - 4 = 4(x - 1) share x - 1 as a common
factor. So, we have
χ3 - x2 + 4x - 4 = x2(x - 1) + 4(x - 1) = (x2 ÷ 4)(x - 1).

262
 9.1. GROUPING

The roots of x2+4 are imaginary, so we cannot factor it as the product of linear terms with real coefficients.
□
In Problem 9.1, we applied the technique of grouping in order to factor x3 - x2 + 4x - 4. We looked
for groups of terms that shared a common factor, factored those groups, and then factored the common
factor, x — 1, out of the two groups.

Problem 9.2: Find all ordered pairs of positive integers (x, y) such that 6xy + 4x + 9y + 6 = 253. I

Solution for Problem 9.2: We could attempt trial and error, but that would take quite a while, and it would
be hard to be sure that we found all the solutions. We've solved many equations with factorization, so
we try factorization here. We start by factoring 2x out of our first two terms, which gives

χx 2x(3y + 2) + 9y + 6 = 253.
We notice that factoring 3 out of 9y + 6 also produces a factor of 3y + 2:

2x(3y + 2) + 3(3y + 2) = 253.

So, we have
(2x + 3)(3y + 2) = 253.
We seek integer solutions (x, y), so 2x + 3 and 3y + 2 must be integers. Since x and y must be positive,
neither 2x+3 nor 3y+2 can equal 1. So, the two factors can't be 1 and 253. We find the prime factorization
of 253 to find other pairs of integers that multiply to 253. Since 253 = 11 ∙ 23, the values of x and y must
satisfy either the system of equations 2x + 3 = 11, 3y + 2 = 23, or the system 2x + 3 = 23, 3y + 2 = 11.
Solving these gives us the solutions (x, y) = (4,7) and (x, y) = (10,3). □
We call equations for which we seek integer solutions Diophantine equations. Problem 9.2 highlights
a powerful problem solving strategy for tackling Diophantine equations.

Concept: Factorization is a powerful tool for solving Diophantine equations. If we

o=§ can write a Diophantine equation as a product of expressions that equal
an integer, such as
(2x + 3)(3y + 2) = 253,
we can then use the divisors of the integer to find solutions to the equation.

Our factorization of 6xy + 4x + 9y + 6 is an example of this factorization:

Important:

ψ ap + aq + bp + bq = (« + t>)(p + q).

This really isn't something new; it's just the reverse of expanding the product of binomials (a+b)(p+q).
This factorization is so useful that we sometimes "force" it into problems. For example, if Problem 9.2
had asked us to find all positive integer pairs such that 6xy + 4x + 9y = 247, we would add 6 to both sides
to give us 6xy + 4x + 9y + 6 = 253, because we could then factor the left side to give (2x + 3)(3y + 2) = 253.
We call this process Simon's Favorite Factoring Trick.

263
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

Sidenote: Simon's Favorite Factoring Trick is named after Art of Problem Solving
K Community member Simon Rubinstein-Salzedo, who described this clever
manipulation as his favorite factoring trick when discussing the problem
of finding all pairs of positive integers {x, y) such that

1 _ 1 _ _1_
x y 97'

See if you can solve the problem!

Problem 9.3: Write each of the following expressions as the product of two binomials, plus (or
minus) a constant. For example, we have xy + x + y — (x + l)(y + 1) — 1.
(a) ab + 3a - 5b (c) 6x2y2 - 3x2 - 2y2
(b) 2mn + 14m - 3n

Solution for Problem 9.3:

(a) We guess that our product of binomials will be of the form (fl + m)(b + ri) for some constants m and
n, because this product will include an ab term in its expansion. Since we need a 3a term and a
-5b term in our expansion, we let m = -5 and n = 3, which gives us
(a - 5)(b + 3) = ab + 3a - 5b - 15.
Adding 15 to both sides, we have ab + 3a - 5b = (fl - 5)(b + 3) + 15.
(b) We know that a product of binomials of the form (m + c)(n + d), where c and d are constants, won't
do, since this will produce an mn term instead of a 2mn term. Instead, we try (2m + c)(h + d), where
we give m a coefficient of 2 rather than n because the coefficient of m in 2mn + 14m - 3n is even. In
order to produce 14m and -3n terms, we let c = -3 and d = 7, and we have
(2m - 3)(n + 7) = 2mn + 14m -3n- 21.
Adding 21 to both sides, we have 2mn + 14m -3n = (2m - 3)(κ + 7) + 21.
We could also have found this result by using grouping. We start by factoring 2m out of the
first two terms to give
2m(n + 7) - 3n.
We need a factor of n + 7 in the second term of this expression in order to factor further. Adding
-3(7) gives us such a factor, but we also need to subtract -3(7) to keep the value of expression
unchanged. We have
2τn(n + 7) — 3n = 2m(n + 7) — 3n — 3(7) — (—3(7)) = 2m(n + 7) — 3(n + 7) + 21 = (2m — 3)(∕z + 7) + 21.

(c) We factor 3x2 out of the first two terms to give 3x2(2y2 - 1) - 2y2. Adding +1 - 1 will allow us to
factor further:
3x2(2y2 - 1) - 2y1 = 3x2(2y2 - 1) - 2y2 + 1 - 1
= 3x2(2y2 - 1) - l(2y2 - 1) - 1 = (3x2 - l)(2y2 - 1) - 1.

□
264
 9.1. GROUPING

Problem 9.4: Find all positive integers nι and «, where n is odd, that satisfy ⅛ + ~ = (Source:
Great Britain)

Solution for Problem 9.4: We start by getting rid of the fractions. Multiplying both sides by 12mn leaves
us 12« +48w = mn. Rearranging then gives us mn -12« -48?« = 0. We apply Simon's Favorite Factoring
Trick by adding (12)(48) to both sides, which gives us mn - 12« - 48m + (12)(48) = (12)(48). We can now
factor the left side, to give (?« - 12)(« - 48) = 576.

Before we go through listing all the divisors of 576 (there are a lot of them!), we note that « must be
odd. Therefore, the quantity n — 48 must be an odd divisor of 576. Moreover, since m and « are positive,
we must have < ⅛∕ which means that « > 48. This greatly limits our search. The prime factorization
of 576 is 26 ∙ 32, so all the odd divisors of 576 are divisors of 32 = 9. This means « — 48 must divide 9.
We have three possibilities. First, « - 48 = 9 and m — 12 = 576/9 = 64, which gives us (m,«) = (76,57).
Second, « - 48 = 3 and m - 12 = 576/3 = 192, which gives us (m,«) = (204,51). Third, « - 48 = 1 and
m -12 = 576, which gives us (m,π) = (588,49). □

Problem 9.5: Let ∕(x, y) = 2x5 + χ4y - 21x3y2 + 16x2 + 8xy - 168y2. Factor ∕(x, y) completely into
linear and quadratic factors with real coefficients.

Solution for Problem 9.5: It's not obvious how we might find an expression that evenly divides f(x, y). We
try to group terms with common factors to get us started. Looking for clues, we note that gcd(16,168) = 8,
so we can factor 8 out of the last three terms of f(x, y):

16x2 + 8xy - 168y2 = 8(2x2 + xy - 21y2).

The factor 2x2 + xy — 21 y2 looks a lot like the first few terms of ∕(x, y). Indeed, we note that

2x5 + x4y - 21x3y2 = x3(2x2 + xy - 21y2).

So, we see that 16x2 + 8xy — 168y2 and 2x5 + x4y - 21x3y2 have a common factor:

2x5 + x4y - 21x3y2 + 16x2 + 8xy - 168y2 = x3(2x2 + xy - 21y2) + 8(2x2 + xy - 21y2)
= (x3 + 8)(2x2 + xy - 21y2).

Next, we check if we can factor either x3 + 8 or 2x2 - xy - 21y2. We factor 2x2 - xy - 21y2 as a quadratic
to find 2x2 + xy - 21y2 = (2x + 7y)(x - 3y). Using our tactics from Chapter 7, we find that x + 2 is a factor
of x3 + 8, and we have x3 + 8 = (x + 2)(x2 - 2x + 4). The discriminant of x2 - 2x + 4 is negative, so we can't
factor x2 - 2x + 4 over the real numbers, and our factorization of ∕(x, y) is:

∕(x, y) = 2x5 + x4y - 21x3y2 + 16x2 + 8xy - 168y2 = (x + 2)(x2 - 2x + 4)(2x + 7y)(x - 3y).

□
Problem 9.6: Factor αb(c2 + d2) + cd(a2 + b2).

Solution for Problem 9.6: We begin by expanding the given expression to get

ab(c2 + d2) + cd(a2 + fo2) = abc2 + abd2 + a2cd + b2cd.

265
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

Looking for similarities between terms to guide us, we note a couple of partial factorizations not present
in the original expression we expanded:

abc2 + a2cd = ac(bc + ad),

abd2 + b2cd = bd(ad + be).

Both right sides include the factor ad + be. Putting these equations together, we have factors of the
original expression:

ab(c2 + d2) + cd(a2 + b2) = abc2 + abd2 + a2cd + b2cd

= abc2 + a2cd + abd2 + b2cd = ac(bc + ad) + bd(ad + be) = (ac + bd)(ad + be).

□
Problem 9.7: Find all values of x that satisfy

a/x -8-9 √x - 8 - 4 a/x —8+4 a/x -8+9

(Source: ARML) ∣

Solution for Problem 9.7: Let y = Vx - 8. Then the equation becomes

+ _J_ + _Z_ +
y-9 + y-4 + y + 4 + y + 9 -
-0
which is considerably less intimidating. We multiply both sides by all four denominators to find

6(y - 4)(y + 4)(y + 9) + l(y - 9)(y + 4)(y + 9) + 7(y - 9)(y - 4)(y + 9) + 12(y - 9)(y - 4)(y + 4) = 0.

Multiplying out all those products doesn't look easy. Instead of diving into expanding all those products,
we notice that many of the products have factors in common. Maybe we can use clever factorization to
simplify the equation.
-- —------------------?--------- - ^ ~~ " ---------------------------------
Concept: Look for ways to use factorization to avoid having to expand complicated
Q)≡≡^ products.

We group the two terms with (y - 4)(y + 4) as factors, since expanding this product gives us a binomial,
y2 - 16, instead of a 3-term quadratic. We find

6(y - 4)(y + 4)(y + 9) + 12(y - 9)(y - 4)(y + 4) = (6(y + 9) + 12(y - 9))(y - 4)(y + 4)
= 18(y - 3)(y2 - 16).

Grouping the two terms with (y - 9)(y + 9) gives

l(y - 9)(y + 4)(y + 9) + 7(y - 9)(y - 4)(y + 9) = (y + 4 + 7(y - 4))(y - 9)(y + 9)

= 8(y - 3)(y2 - 81).

266
 9.1. GROUPING

So, the left side of the equation we had after multiplying by the denominators can be simplified as:
18(y - 3)(y2 - 16) + 8(y - 3)(y2 - 81) = (y- 3)(18(y2 - 16) + 8(y2 - 81))
= (y - 3)(2)(9(y2 - 16) + 4(j∕2 - 81))
= 2(y - 3)(13y2 - 468)
= 2(y - 3)(13)(y2 - 36)
= 26(y-3)(y-6)(y + 6).

So, our equation now is 26(y - 3)(y - 6)(y + 6) = 0. Our careful factorization spared us a lot of work.
Now, we have y = 3, y = 6, and y = —6 as solutions. However, we must find x. Since y = λ∕x — 8,
the solution y = —6 is extraneous, because a square root cannot be negative. The solution y = 3 gives us
3 = √χ-8, sox = 17, and y = 6 gives us 6 = √x-8, sox = 44. Therefore, 17 and 44 are the only values
of x that satisfy the original equation. □

In general, grouping can be very helpful on problems that allow us to rearrange an equation as a
product of factors equal to 0, and on Diophantine equations. However, it isn't always successful. For
example, suppose we must find all the solutions to the equation x3 - 4x2 + 8x - 32 = -27. We might be
think we have an opportunity to use grouping on the left side by noting that

x3 - 4x2 + 8x - 32 = x2(x - 4) + 8(x - 4) = (x2 + 8)(x - 4),

so we have (x2 + 8)(x — 4) = -27. But now what? This isn't a Diophantine equation, and we don't have a
product of factors equal to 0. We might proceed with the factors of 27, but this will only succeed if all the
possible solutions are integers. Instead, we're better off rearranging the equation as x3 - 4x2 + 8x - 5 = 0
and using our usual techniques for finding the roots of polynomials.

WARNING!! There's an old saying that "When all you have is a hammer, every
V problem looks like a nail." Keep your problem solving toolbox well-
stocked, so you won't find yourself swinging a hammer when you
really need a wrench. In other words, don't get so attached to one
strategy that you can't let it go and try something else when it isn't
working.

*
Exercise

9.1.1 Factor each of the following polynomials:

(a) xy + 3x - 4y - 12 (b) 2ab - 14« + 5b - 35

9.1.2 Find all ordered pairs of integers (x, y) such that 1 + 1 = ∣.

9.1.3 Factor x3 + 2xy2 + 5x2 + 10y2 completely over the reals.

9.1.4 Find all ordered pairs of real numbers (x, y) such that 4x2y2 + 36x2 = y2 + 9.
9.1.5 If x and y are positive integers and x + y + xy = 54, then find x + y.
9.1.6 Find all real roots of 8 - 2y + 4y3 - y4.

267
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

9.1.7 Factor x2y + xy2 + x2 + 2xy + y2 + x + y.

*
9.1.8 Find all solutions to the system of equations

xy + 2x + 3y = 4,
yz - y + 2z = 5,
xz- x + 3z = 33.

9.2 Sums and Differences of Powers

11 IL.________
Problems

Problem 9.8: Use the Factor Theorem to find a factor of each of the following. Rewrite each as the
product of a linear polynomial and another polynomial.
(a) x3 - 1 (c) x5 -1
(b) x4 — 1 (d) xn — 1, where n is an integer with n> 1

Problem 9.9: Factor x5 - 25.

Problem 9.10: Let f(x) = xn - yn for some constant y and some integer n, where n> 1.
(a) What is/(y)?
(b) Factor xn - yn into a product of two polynomial factors such that one factor is linear in x and y
and the coefficient of x in this factor is 1.

Problem 9.11:
(a) Can you factor x2 + y2 as the product of two linear expressions with real coefficients?
(b) Factor x3 + y3 as the product of two polynomials with nonzero degree.
(c) Let n be a positive integer. Factor x2n+1 + y2'1+1 as the product of two polynomials with nonzero
degrees.

Problem 9.12: In this problem, we show that 350 + 850 is divisible by 73.
(a) Use the factorization of x5 + y5 to explain why 35 + 85 is divisible by 11.
(b) Explain why 350 + 850 is divisible by 31° + 810.
(c) Show that 350 + 850 is divisible by 73.

Extra! French mathematician Marie-Sophie Germain found a factorization during her study
<∣ * ∣n* ∣ ∣* of number theory that still bears her name. Sophie Germain's Identity is a factorization
of x4 + 4y4 over the integers. See if you can find the factorization!
Germain carried on a notable mathematical correspondence with the great mathe­
matician Carl Gauss. She used a pseudonym when writing Gauss to hide her gender.
Gauss would learn her identity when she asked a French commander to ensure Gauss's
safety during a French occupation of Gauss's hometown in Germany.

268
 9.2. SUMS AND DIFFERENCES OF POWERS

Problem 9.13: In this problem, we compute the number of ordered triples of integers (x, ∖
j,z) with
1729 < x, y,z < 1999 such that

x2 + xy + y2 = y3 - x3 and yz + 1 = y2 + z.

(Source: ARML)
(a) Factor the right side of the first equation. How can you simplify the resulting equation?
(b) Rearrange the second equation until you find an expression in the equation that you can factor.
(This expression need not involve every term in the equation.)
(c) There are several ways to accomplish part (b). Keep experimenting until you find a way to
rewrite the second equation with a product of two factors on one side and 0 on the other.
(d) Combine parts (a) and (c) to find restrictions on x, y, and z that you can use to solve the problem.

Problem 9.14:
(a) Find a positive integer n such that x4 + x2 + 1 is a factor of xn - 1.
(b) Factor xn — 1 as a difference of squares for your choice of n in part (a).
(c) Factor the two factors produced in part (b).
(d) Factor x4 + x2 + 1 completely over the real numbers.

Problem 9.15: In this problem, we factor

∕(x) = x1° + 2x9 + 3x8 + 4x7 + 5x6 + 6x5 + 5x4 + 4x3 + 3x2 + 2x + 1.

(a) Expand each of the following: (x +1)2, (x2 + x +1)2, (x3 + x2 + x +1)2. Notice anything interesting?
(b) Express ∕(x) as the product of linear and quadratic expressions with real coefficients.

Problem 9.16: Let N = 48 + 68 + 98. In this problem, we find two 4-digit numbers whose product is N.
(a) What expression must be added to x2 + xy + y2 to give a square of a binomial?
(b) Why is N + 68 a perfect square? (Note that 6 = 2 ∙ 3.)
(c) Write N + 68 - 68 as the product of two 4-digit numbers.

Problem 9.17: In this problem, we evaluate

(104 + 324)(224 + 324)(344 + 324)(464 + 324)(584 + 324)

(44 + 324)(164 + 324)(284 + 324)(404 + 324)(524 + 324)

ιoithout a calculator. (Source: AIME)

(a) Let ∕(x) = x4 + 324. Find a function g(x) such that ∕(x) + g(x) is the square of a binomial.
(b) Use part (a) to factor x4 + 324 as the product of polynomials with integer coefficients.
(c) Solve the problem.

269
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

Problem 9.8: Express each of the following as the product of a linear polynomial and another
polynomial.
(a) x3 - 1 (c) x5 -1
(b) x4 -1 (d) x” - 1, where n is an integer with n > 1

Solution for Problem 9.8: We see that x = 1 is a root of each of the given polynomials. So, by the Factor
Theorem, x - 1 is a factor of each. We divide x - 1 out of each to get another factor.

(a) x3 - 1 = (χ - l)(x2 + χ + 1).

(b) x4 - 1 = (χ - l)(χ3 + χ2 + χ + 1).
(c) x5 - 1 = (χ - l)(χ4 + x3 + χ2 + χ + 1).
(d) Our first three parts suggest that x" - 1 = (x - l)(xπ^1 + x"^2 + xn~3 + ∙ ∙ ∙ + x + 1). We confirm this
by expanding the product on the right side:

(x - l)(x"^1 + xn~2 + ∙ ∙ ∙ + x + 1) = χ(χπ^1 + xn^^2 + ∙ ∙ ∙ + x + 1) - l(xπ^1 + xn~2 + ∙ ∙ ∙ + x + 1)

= x” + x"^1 + ∙ ∙ ■ + x2 + x - xm"1 - x"~2----- -- x - 1
= xn-l.

Maybe the pattern we see in Problem 9.8 is just a quirk that happens when we subtract 1 from xn.
What if we try another constant? We might try factoring x3 - 2 over the rational numbers, but quickly
see that we get nowhere, since x3 - 2 has no rational roots. But what if our constant is raised to the same
power as x is?

Problem 9.9: Factor x5 - 25._ _ '

Solution for Problem 9.9: Let ∕(x) = x5 - 25. Clearly, we have ∕(2) = 0, so x - 2 is a factor. We divide x - 2
into x5 - 25 and find that

x5 - 25 = (x - 2)(x4 + 2x3 + 4x2 + 8x + 16) = (x - 2)(x4 + 21x3 + 22x2 + 23x + 24).

Now, we,re ready to generalize our results from Problems 9.8 and 9.9.

Problem 9.10: Let n be a positive integer. Factor xn - yn into a product of two polynomial factors
such that one factor is linear in x and y and the coefficient of x in this factor is 1.

Solution for Problem 9.10: We use the previous two problems as a guide. Suppose ∕(x) = xn - yn, where
y is a constant. We have f(y) = yn - yn = 0, so by the Factor Theorem, x - y is a factor of xn - y". Our
work in Problem 9.9 strongly suggests that we then have

Xπ - y" = (χ - y) (χ" 1 + χ" 2y + χn 3y2 + ∙ ∙ ∙ + χyn 2 + yn 1) .

270
 9.2. SUMS AND DIFFERENCES OF POWERS

As before, expanding the right side confirms this factorization:

(x - y)(xw^1 + x"-2y + ∙ ∙ ∙ + xy"~2 + y"^1)
= x (χ"~1 + χ"-2y + ∙ ∙ ∙ + χy"^2 + y''^1) - y (x”_1 + χn~2y + ∙ ∙ ∙ + χyn~2 + y”_1)
= x'' + x"-1y + ∙ ∙ ∙ + x2y"~2 + xy"^1 - x"-1y - xn~2y2-------- χytl~*x ~ yn

Important:

ψ χn - yn = (χ - y) (xn 1 + x” 2y + χn 3y2 + ... + χyn 2 + y" 1J

The most common instances of this factorization are the difference of squares (n — 2) and difference
of cubes (π = 3) factorizations:

X2 - y2 = (χ - y)(χ + y),
χ3 - y3 = (χ _ y)(χ2 + χy + y2).

What about sums of nth powers? If we try factoring x2 + 1, we immediately get stuck. The roots of
x2 + 1 are nonreal, so we can't factor x2 +1 as a product of linear factors with real coefficients. However,
if we look at x3 + 1, we see that —1 is a root, so x + 1 is a factor of x3 + 1. Similarly, -1 is a root of x5 + 1,
x7 + 1, x9 + 1, and so on. This gives us an idea:

Problem 9.11: Let n be a positive integer. Factor x2"+1 + y2π+1 as the product of a linear factor and
another polynomial.

Solution for Problem 9.11: Before jumping into factoring x2π+1 + y2"+1, we try a few simple examples.
First, we lety = 1,and try small values of n. Since x = -1 is a root of x2π+1 + 1 for every nonnegative
integer n, we quickly find
x3 + 1 = (x + l)(x2 - x + 1),
x5 + 1 = (x + l)(x4 - x3 + x2 - x + 1),
x7 + 1 = (x + l)(x6 - x5 + x4 - x3 + x2 - x + 1).

Aha! We see a pattern, so we guess that

x2"+1 + 1 = (χ + l)(x2" - x2"^^1 + x2n~2----- + x2 - x + 1).

We can multiply the right side out to confirm it is true. Let's see what happens if y is something more
interesting than 1. If x = -y, then no matter what y is, we have x2',+1 + y2"+1 = 0. Using this, we can
quickly find a linear factor for each of our examples. Here's what we find:
x3 + 33 = (x + 3)(x2 - 31x + 32),
x5 + 25 = (x + 2)(x4 - 21x3 + 22x2 - 23x + 24),
x7 + 27 = (x + 2)(x6 - 21x5 + 22x4 - 23x3 + 24x2 - 25x + 26).

271
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

Our examples strongly suggest that

x2"+1 + y2'1+1 = (x + y)(x2'1 - x2'7^^1y + x2w-2y2----- + x2y2"~2 - xy2'7-1 + y2”).

As before, we expand the right side to confirm this factorization:

(x + y)(x2" - x2"~1y + x2"^2y2----- + x2y2,7"2 - xy2"^1 + y2z7)
= x(x2,' - x2"^1y + x2"^2y2----- + x2y2"~2 - xy2n~λ + y2”)
+ y(x2n - x2"~1y + x2"~2y2----- + x2y2n~2 - xy2"~l + y2")
= x2"÷1 - Λ-2"y + x2''~1y2 - ∙ ■ ■ + x,y2n~2 - x2y2"-1 + *
y 2"
+ x2"y - χ2"-1y2 + ■ ■ ■ - χ3y2"-2 + X2y2n~' - xy2" + y2"+, ■

All terms but the first and last cancel, and we are left with

Important:
x2',+ι + y2f7+1 = (x + y)(x2'7 - x2"^1y + x2"^2y2----- + x2y2" 2 - xy2n 1 + y2π)

□
The most common instance of this factorization is the sum of cubes factorization:
x3 + y3 = (x + y)(x2 - xy + y2).

You might wonder why we focus on sums of odd powers. To see why we cannot factor x2 + y2 over
the reals, notice that if x2 + y2 = 0, then x2 = -y2. Taking the square root of both sides gives x = ±zy. In
other words, if we view x2 + y2 as a polynomial in x, its roots are zy and -zy, which means we can factor
it as x2 + y2 = (x + zy)(x - zy). That is, we can factor it over complex numbers, but not over real numbers.
However, as we'll see in our next problem, we can sometimes still use our sum of powers factorization
when the powers are even.

Problem 9.12: Show that 350 + 850 is divisible by 73.

Solution for Problem 9.12: We certainly aren't going to compute 350 + 850. Instead, we'll look for a way
to factor it, hoping we can eventually write it as the product of 73 and an integer. We don't have a
factorization for the sum of 50th powers, but any 50th power is also a 5th power and a 25th power. So,
we factor 350 + 850 as a sum of 5th powers and as a sum of 25th powers:
3≡0 + 850 = (310)5 + (810)5 = (310 + 810)((310)4 _ (310)3(gl0) + (3W)2(glO)2 _ (310)(gl0)3 + (glθ)4y
350 +g50 = (32)25 + (δ2)25 = (32 + 82)((32)24 - (32)23(82) +----- (32)(82)23 + (82)24).

The first factorization requires more work to be useful (try factoring 310 + 8l0), but the second is
immediately a winner, because 32 + 82 = 73. Our second factorization tells us that 350 + 850 equals
73 times an integer, which means 350 + 850 is divisible by 73. □

Concept: Factorization can be useful in number theory problems involving divisi-

O≡≡s bility∙

272
 9.2. SUMS AND DIFFERENCES OF POWERS

Notice that we were able to apply our factorization for x2"^, 1 + y2''+1 to a sum of even powers in
Problem 9.12 because the even power, 50, is divisible by odd numbers besides 1 (namely, 5 and 25). So,
we can view the 50th powers as 5th or 25th powers and use the factorization of x2',+1 + y2"+1.

Problem 9.13: Compute the number of ordered triples of integers (x, y,z) with 1729 < x, y,z < 1999
such that
x2 + xy + y2 = yi - x3 and yz + 1 = y2 + z.
(Source: ARML)

Solution for Problem 9.13: We hope to simplify the system of equations, so we look for helpful factoriza­
tions or manipulations. We note that the right side of the first equation is a difference of cubes. Factoring
the right side of x2 + xy + y2 = yi - χ3 gives us

x2 + xy + y2 = (y - χ)(y2 + xy + x2).

Since x and y are positive, the factor x2 + xy + y2 is nonzero. So, we can divide both sides of our equation
by x2 + xy + y2 to find 1 = y - x. Therefore, y = x + 1.

Unfortunately, the relationship we discovered between x and y doesn't help us with the second
equation. Neither side factors and nothing jumps out of the expressions on either side that seems
helpful, so we move all the terms to one side of the equation to give ourselves another perspective:

y2 - yz + z - 1 = 0.

Experiment with equations that you're unsure how to handle. Rearrange

Concept:
Q -them and look for familiar expressions or forms. Don't just stare at them!

We could factor y out of the first two terms, but this doesn't help much. However, we also notice that
we have a difference of squares, y2 - 1. Factoring this gives us

(y + i)(y -1) - yz + z = o.

Factoring z out of the last two terms gives (y + l)(y -1) - z(y -1) = 0. Aha! We can factor out y -1 to get

(y + l-z)(y-l) = 0.

Since y > 1729, we know that y - 1 is nonzero. Dividing by y - 1 gives us y + 1 - z = 0, so y = z - 1.

From y = x + 1, we have x = y - 1, and y = z - 1 gives z = y + 1. So, for each possible value of y,
the triple (x, y,z) = (y - 1, y, y + 1) is the only possible solution to the system with that value of y. Since
1729 < x,y,z < 1999, we must have 1730 < y < 1998 in order to keep χ = y - 1 and z = y + 1 in the
permissible range of values. Since there are 1998 - 1730 -1 = 267 integers between 1730 and 1998, there
are 267 triples of integers (x, y, z) that are solutions to the given equations. □

Problem 9.14: Factor x4 + x2 + 1 over the real numbers.

Solution for Problem 9.14: The expression we must factor has the form y2 + y +1 for y = x2. We recognize
y2 + y + 1 as a factor of y3 — 1. Since y3 - 1 = (y - l)(y2 + y + 1), we have y2 + y + 1 = (y3 - l)∕(y - 1).

273
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

Letting ∖
j = x2, we have

Now we factor the numerator, first using the difference of squares factorization, then both the sum and
difference of cubes factorizations:
v4 , v2 , π χ6 - 1 (χ3 + l)(χ3 - 1) (*x + 1)(x - l)(χ2 + x + 1)(χ2 - x + 1)
x2 -1 x2 -1 X2 -1
Since x2 - 1 = (x + l)(χ - 1), we have

x4 + x2 +1 = (x2 + x + l)(x2 - x + 1).

Neither of the quadratics on the right side has real roots, so we cannot factor any further over the reals.
□
In Chapter 10, we'll take another look at expressions like 1 + x2 + x4 when we discuss geometric
series.
Problem 9.15: Factor

∕(x) = x1° + 2x9 + 3x8 + 4x7 + 5x6 + 6x5 + 5x4 + 4x3 + 3x2 + 2x + 1

as the product of linear and quadratic expressions with real coefficients.

Solution for Problem 9.15: Before tackling this tenth degree polynomial, we try to think of similar
polynomials that we know how to deal with. We notice that the coefficients of the polynomial go up
by 1 until they reach 6, then back down by 1 until they reach 1. The expansion (x + 1)2 = x2 + 2x + 1 is
similar in this regard: the coefficients go up to 2, then back down. So, does this give us any insight into,
say, x4 + 2x3 + 3x2 + 2x + 1?
Because the leading term of x4 + 2x3 + 3x2 + 2x +1 is x4, the constant term is 1, and all the coefficients
are positive, we guess that
x4 + 2x3 + 3x2 + 2x + 1 = (x2 + αx + l)(x2 + bx+ 1),

for some constants a and b. We need the coefficients of x3 and of x in the expansion of the right side both
to be 2, so we try a = b = 1. Sure enough, we find that
(x2 + x + 1)2 = x4 + 2x3 + 3x2 + 2x + 1.

Hmmm.... Comparing this to our expansion of (x + 1)2, we start to see a pattern. We guess that ∕(x)
equals (x5 + x4 + x3 + x2 + x + 1)2. We expand

(x5 + x4 + x3 + x2 + x + l)(x5 + x4 + x3 + x2 + x + 1)

to check if it works. To expand this product, we must add every possible product of a term in tire first
factor and a term in the second. The expansion clearly has x1° as its first term. There are two x9 terms,
(x5)(x4) and (x4)(x5). There are three x8 terms, (x5)(x3), (x4)(x4), and (x3)(x5). Continuing in this manner,
we find that
x1° + 2x9 + 3x8 + 4x7 + 5x6 + 6x5 + 5x4 + 4x3 + 3x2 + 2x + 1 = (x5 + x4 + x3 + x2 + x + 1)2.

274
 9.2. SUMS AND DIFFERENCES OF POWERS

We're not finished. We have to try to factor x5 + x4 + x3 + x2 + x + 1. We turn to our strategy from
Problem 9.14. We recognize this polynomial as a factor of x6 - 1, and we have

5 4 3 2 1 χβ - I (X3 — l)(x3 + l) (x - l)(x2 + X + l)(x + l)(x2 - X + l)

x-l x-l x-l
= (x + l)(x2 + x + l)(x2 - x + l).

The quadratics in this final expression have nonreal roots, so we cannot factor any further. So, we finally
have
∕(x) = (x5 + x4 + χ3 + χ2 + l)2 = (χ + l)2(χ2 + χ + l)2(x2 -χ + l)2.
□
Notice that we did a lot of guessing on the way to finding a solution. These weren't blind, lucky
guesses; each was based on a few observations or patterns. Moreover, we confirmed each guess by
working out our hunches to see if they were correct. This is not only a perfectly valid approach to
problems, it s also an important problem solving strategy.

Concept: Don't be afraid to guess. A great many problems can be solved by looking
O=^τ>∣j for patterns or clues, guessing an answer, then showing that your guess is
indeed the correct answer.

Problem 9.16: Find two 4-digit numbers whose product is 48 + 68 + 98.

Solution for Problem 9.16: None of our factorizations is immediately useful. We notice that the bases in
the expression are 22, 2 ∙ 3, and 32, so we write the expression in terms of 2s and 3s:

48 + 68 + 98 = (22)8 + (2 ∙ 3)8 + (32)8 = 216 + (2 ∙ 3)8 + 316.

This is very close to the expansion of (28 + 38)2. If the middle term were 2(2 ∙ 3)8, then the expression
would be the expansion of (28 + 38)2. This gives us an idea. Let's "complete the square":

216 + (2 ∙ 3)8 + 316 = 216 + (2 ∙ 3)8 + 316 + (2 ∙ 3)8 - (2 ∙ 3)8

= 216 + 2(2 ∙ 3)8 + 316 - (2 ∙ 3)8
= (28 + 38)2 - (2 ∙ 3)8.

We have a difference of squares! Factoring it gives

48 + 68 + 98 = (28 + 38)2 - (2 ∙ 3)8 = (28 + 38 + (2 ∙ 3)4)(28 + 38 - (2 ∙ 3)4) = 8113 ■ 5521.

Problem 9.16 gives us both another method of completing the square, and another use for doing so.

Important: We can sometimes factor the sum of even powers of expressions by

ψ creating a difference of squares. We do so by adding and subtracting an

appropriate perfect square so that the result is the square of a binomial
minus the perfect square.

275
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

Here's another problem that illustrates this strategy:

Problem 9.17: Evaluate

(104 + 324)(224 + 324)(344 + 324)(464 + 324)(584 + 324)

(44 + 324)(164 + 324)(284 + 324)(404 + 324)(524 + 324)

without a calculator. (Source: AIME)

Solution for Problem 9.17: We can't cancel anything, but we do notice that each factor in the numerator
and denominatoi" has the form n4 + 324. So, if we find a way to factor n4 + 324, we may be able to use this
factorization to do some canceling and simplify the fraction. Since 324 = 182, we have w4 + 182. This is a
sum of squares, which gets us thinking about using our new "building a difference of squares" strategy.
We seek a term we can add to n4 + 182 to make the result the square of a binomial. This binomial
must be n2 + 18, since squaring this will give us both an n4 term and an 182 term. We find

(n2 + 18)2 = n4 + 36π2 + 182,

so h4 + 182 = (h2 + 18)2 — 36π2. Success! We can factor the right side as a difference of squares, to find

n4 + 182 = (π2 + 18)2 - 36π2 = (n2 - 6n + 18)(n2 + 6n + 18).

So, now our fraction is

(102 - 6 ■ 10 + 18)(102 + 6 ∙ 10 + 18)(222 - 6 ∙ 22 + 18)(222 + 6 ∙ 22 + 18) ∙ ∙ ∙ (582 - 6 ∙ 58 + 18)(582 + 6 - 58 + 18)

(42 _ 6.4 + 18)(42 + 6 ∙ 4 + 18)(162 - 6 ∙ 16 + 18)(162 + 6 ∙ 16 + 18) ∙ ∙ ∙ (522 - 6 ∙ 52 + 18)(522 + 6 - 52 + 18)
We might think we're stuck again, but we notice that

102 - 6 ∙ 10 = (10 - 6)(10) = 4(10) = 4(4 + 6) = 42 + 6 ∙ 4,

so the factors (102 — 6 ∙ 10 +18) and (42 + 6 ∙ 4 +18) are equal, and therefore cancel. Maybe it's a coincidence.
Let's see if there's a factor in the denominator equal to (102 + 6 ∙ 10 + 18). There is—the very next factor,
(162 -616 + 18), because

102 + 6 ∙ 10 = (10 + 6)(10) = (16)(10) = 16(16 - 6) = 162 - 6 ∙ 16.

Doing a little more factorization reveals why these factors in the numerator match factors in the denom­
inator. We write each factor of the form n2 — 6n +18 as n(n — 6) +18 and each factor of the form n2 + 6rι +18
as n(n + 6) + 18:
(10 ∙ 4 + 18)(10 ∙ 16 + 18)(22 ∙ 16 + 18)(22 ■ 28 + 18) ∙ ∙ ∙ (58 ∙ 52 + 18)(58 ∙ 64 + 18)
(4 ∙ (-2) + 18)(4 ∙ 10 + 18)(16 ∙ 10 + 18)(16 ∙ 22 + 18) ∙ ∙ ∙ (52 ■ 46 + 18)(52 ∙ 58 + 18) ’
Now we see that everything cancels except the last term in the numerator and the first term in the
denominator, and we have
58-64 + 18 _ 3730 _ 373
4-(-2) + 18^ 10
□
276
 9.3*. THE FACTOR THEOREM FOR MULTIVARIABLE POLYNOMIALS

Concept: Experimentation and searching for patterns are critical factoring skills.
O≡^

∣l∣L
ι xer< ises
9.2.1 Find all 6 roots of ∕(x) = x6 - 1.

9.2.2 Compute all ordered pairs (x, y) such that χy + 9 = y2 and xy + 7 = x2. (Source: ARML)
9.2.3 Show that 227 + 327 is divisible by 35.
9.2.4 Let x3 = 8x + y and y3 = x + 8y. If ∣x∣ ≠ ∣y∣, then find x2 + y2.
1 11
9.2.5 Find (a∕c)3 if = — + -. (Source: Mandelbrot)

9.2.6 Given that 7,999,999,999 has at most two prime factors, find its largest prime factor. (Source:
HMMT) Hints: 347

9.2.7 Compute the real value of x such that (x2 + x + l)(χ6 + x3 + 1) = ------ . (Source: ARML)
x-1
*
9.2.8 Show that n4 - 20rz2 + 4 is composite for all integers n, where n > 4. (Source: Crux Mathemati-
corum) Hints: 283

9.3⅛ The Factor Theorem for Multivariable Polynomials

II Problems

Problem 9.18: In this problem, we factor the polynomial f(α, b) = b3 - ab2 - ab - b + a2 + a.

(a) Start by viewing f(a, b) as a polynomial in a, with b being just a constant. In other words, let
g(a) = a2 - ab2 - ab + a + b3 - b. Find a function h(b) such that g(h(b)) = 0 for all values of b.
(b) Find the polynomial q(a) such that g(a) = (a-h(b))q(a). In other words, use your result from part
(a) to factor g(a).

Problem 9.18: Factor the polynomial /(«, b) = b3 - ab2 - ab - b + a2 + a.

Solution for Problem 9.18: We could try grouping, but it's not at all clear which terms to group. For
example, we could factor b from the first 2 terms and a from the last two, and we have
f(a,b) = b(b2 - ab-a-l) + a(a + 1).
That's not too helpful.

We have 3 positive terms and 3 negative terms in our original f(a, b), and each coefficient is 1 or -1.
If we just let a = b, all the terms will cancel:
f(a, a) = a3 - a(a2) - a(a) - a + a2 + a = a3 - a3 - a2 - a + a2 + a - 0.

277
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

Hmmm.... We have f(a, b) = 0 whenever a = b. Does this mean a - b is a factor, just as x - r is a factor
of a polynomial p(x) if p(r) = 0?
To see why a - b is a factor of f(a, b), let's take a different view of f{a, b). We know a lot about factoring
one-variable polynomials, so we view f(a, b) as a polynomial in a, and treat b as a constant. We let

g(fl) = b3 — ab2 — ab - b + a2 + a = a2 + (—b2 - b + l)a + (b3 — b),

where b is a constant. Notice that we organize the terms of g(fl) by the degree of a in each term, in order
to focus on the fact that it is a polynomial in a.
As shown above, we have g(b) = 0, so b is a root of g(a). Therefore, a — b is a factor of the one-variable
polynomial g(fl), and we can write
g(«) = (fl - b)q(a)
for some polynomial q(a). Note that we only know that q(a) is a polynomial in a at this point. We can
find q(d) with long division, or with a little logic. We must have

(α - b)q(a) = g(a) = a2 + (-b2 - b + l)fl + (b3 - b),

Since the far right side is quadratic in a, the polynomial z∕(fl) must be linear in a. The leading term of g(fl)
is fl2, so the leading term of q(a') is simply a, which means that q(a) = a + k for some constant k (where the
constant k may be in terms of b, which we are treating as a constant). We therefore have

(a - b)(fl + k) = fl2 - ab2 - ab + a + b3 — b.

Expanding the left gives a2 -ab + ak-bk = a2 - ab2 - ab + a + b3 -b. Removing common terms leaves us
with ak-bk = -ab2 + a + b3 - b. Since k must be a constant, we equate the coefficients of a on both sides
to find k = -b2 + 1. (Remember, we are treating b as a constant.) So, we have

g(fl) = (a - fe)(fl -b2 + 1).

But this holds for any constant b, so we have our factorization:

∕(fl, b) = (a - b)(fl - b2 + 1).

As a check, we can multiply it out:

(a - b)(a - b2 + 1) = fl(fl - b2 + 1) - b(a - b2 + 1) = a2 - ab2 + a - ab + b3 - b,

as desired. □
Taking a closer look at our solution to Problem 9.18, our key step was finding a polynomial h(b) such
that f(h(b),b) = 0 for all b. Specifically, we found that if h(b) = b, then f(h(b), b) = f(b,b) = 0 for all b.
This observation guided us to guess that a - h(b) (that is, fl - b) is a factor of f(a, b). Viewing f(a, b) as a
one-variable polynomial in a and applying the Factor Theorem made it particularly clear that we could
write f(a,b) as the product of a - b and another factor. We then found that this other factor is also a
polynomial.
In much the same way, we can show that:

278
 9.3*. THE FACTOR THEOREM FOR MULTIVARIABLE POLYNOMIALS

Important: If f(a,b) is a polynomial and there is a polynomial h(b) such that

f(lι(b'), b) = 0 for all b, then we can write f(a,b) = (a - h(b))g(a,b), where
g(a,b) is a polynomial.

We can also extend this to a polynomial in any number of variables. For the rest of this section, we'll
focus on applying the Factor Theorem for multivariable polynomials to various problems. You'll also
get a lot of practice in dealing with complicated expressions and expressions involving several variables.

Problems

Problem 9.19: Let f(a, b, c) = (a + b + c)3 - a3 - b3 - c3.

(a) Show that f(-b, b, c) = 0. What expression must therefore be a factor of f{a, b, c)2
(b) Show that a + c and b + c are factors of f(a, b, c).
(c) Why are the factors found in parts (a) and (b) the only nonconstant factors of ∕(fl, b, c)?
(d) Factor f (a, b, c). Check your factorization by expanding f(a, b, c) and the factorization you found.

Problem 9.20: Factor α2(b - c) + b2(c -a) + c2(a - b).

Problem 9.21: Factor (α + b + c)3 - (b + c - a)3 - (c + a - b)3 - (α + b - c)3.

Problem 9.19: Factor f(a, b,c) = (a + b + c)3 - a3 -b3 - c3.

Solution for Problem 9.19: We notice that if a = -b, then we have

f(a, b, c) = f(-b, b, c) = (-b + b + c)3 - (-b)3 - b3 - c3 = c3 - c3 = 0.

Since f(-b, b, c) = 0, the Factor Theorem tells us that a + b is a factor of f(a, b, c). Similarly, if α = -c, we
have
f(-c, b, c) = (-c + b + c)3 - (-c)3 - b3 - c3 = b3 - b3 = 0,
so a + c is a factor of f(a, b, c). Finally, if b = -c, we have ∕(α, -c, c) = 0, so b + c is a factor of f(a, b,c).
Since deg∕ = 3, and a + b, a + c, and b + c are factors of f(a, b, c), we have
/(a, b, c) = k(a + b)(a + c)(b + c)

for some constant k. There can be no more factors, since the product of (a + b)(μ + c)(b + c) and any
nonconstant polynomial has a degree greater than 3.
Since this factorization must equal /(«, b, c) for all a, b, and c, we can choose specific values of a, b,
and c to find k. So, we let a = b = c = 1, and we have
(1 + 1 + 1)3 - 13 - 13 - 13 = k(l + 1)(1 + 1)(1 + 1).

Solving this equation gives us k = 3, so we have

(a + b + c)3-a3-b3-c3 = 3(« + b)(a + c)(b + c).

279
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

We can spot-check our factorization by trying a few triples (a, b, c), and making sure the two sides
are equal. For example, if (π, b, c) = (1,1,0), then both sides equal 6.

Sidenote: We can also recognize that both (α+b+c)3-a3-b3-c3 and 3(a+b)(b+c)(c+fl) are
J∖ symmetric. By symmetric, we mean that if we exchange any 2 variables in
the expression, then the expression is unchanged. For example, exchanging
any 2 of the variables in the expression abc leaves the expression unchanged,
so it is symmetric. On the other hand, the a2bc is not symmetric, because
exchanging a and b gives ab2c, which is not the same as a2bc. A symmetric
expression cannot be equivalent to a non-symmetric expression, so noting
that our two expressions are symmetric gives us a quick check that we
haven't made a pretty obvious mistake (but it doesn't show the two are the
same).

The best check is to multiply out both sides. The (α + b + c)3 on the left side presents a bit of a
challenge. We start by writing it as

(a + b + c)(α + b + c)(α + b + c).

We could repeatedly use the distributive property, but we can also carefully reason our way to the
expansion. Each term in the expansion results from taking the product of one term from each of the
factors in the product. Because all the terms in each factor are linear, and we multiply three of them
to form each term in the product, each term in the expansion has degree 3. So, we have three types of
terms in the expansion:

• Terms of the form a3. If we take a from each expression, we get a3. Similarly, we get b3 by taking b
from each expression and c3 by taking c from each expression. Each of these can only be done in
one way, so a3 + b3 + c3 is part of the expansion.

• Terms of the form a2b. If we take a from two expressions and b from the other, we form a2b. We can
do so in 3 ways, since there are 3 ways to choose the expression from which we take the b. So,
there is a 3a2b term in the expansion. Similarly, we see that the expansion must include

3a2b + 3ab2 + 3a2c + 3ac2 + 3b2c + 3bc2.

• Terms of the form abc. If we take a from one expression, b from another, and c from the remaining
expression, we form the product abc. There are 3 ∙ 2 ∙ 1 = 6 ways to form abc in the expansion, so
we must have 6abc in our expansion.

So, we have

(a + b + c)3 = a3 + b3 + c3 + 3a2b + 3ab2 + 3a2c + 3ac2 + 3b2c + 3bc2 + 6abc.

Notice that both sides are symmetric, and they both equal 27 when a = b = c = 1. (This is just a check; it
doesn't prove our expansion is correct.)

280
 9.3*. THE FACTOR THEOREM FOR MULTIVARIABLE POLYNOMIALS

Important: Don't memorize the expansion of (a + b + c)3. Just understand how we

V∣ ? cameup withit, and you'll be able to re-derive it if you need it. You'll also
's' be able to use this expansion method on other complicated expansions.

Using our expansion of (fl + b + c)3 gives us the expansion of the left side of

(a + b + c)3 -a3 -b3 -c3 = 3(fl + b)(a + c)(b + c)

as
3<72t> + 3ab2 + 3a2c + 3ac2 + 3b2c + 3bc2 + 6abc.

We could use the same approach to expand 3(fl + b)(a + c)(b + c), or we could just use the distributive
property:

3(α + b)(a + c)(b + c) = 3(a2 + ac + ab + bc)(b + c)

= 3(fl⅛ + a2c + abc + ac2 + ab2 + abc + b2c + be2)
= 3a2b + 3ab2 + 3a2c + 3ac2 + 3b2c + 3bc2 + 6abc.

So, our factorization is correct. □

Problem 9.20: Factor a2(b - c) + b2(c - a) + c2(fl - b).

Solution for Problem 9.20: Let the polynomial be f(a, b, c). We have f(b, b, c) = b2(b-c)+b2(c-b)+c2(b-b) =
0. Since f(a, b,c) = 0 when a = b, the expression a - b is a factor of ∕(fl, b, c). Similarly, we have f(a, b,c) = 0
for c = α and b = c, so c - a and b - c are factors of ∕(fl, b, c). Because deg f = 3 and we have found 3 linear
factors of f, we have

∕(fl, b, c) = a2(b - c) + b2(c - a) + c2{a -b) = k(a - b)(b - c)(c - fl)

for some constant k. This must hold for all a, b, and c, so we let a = 3, b = 2, and c = 1, and we have
32(2 - 1) + 22(1 - 3) + 12(3 - 2) = k(3 - 2)(2 - 1)(1 - 3),

which gives us k = -1. So, we have

a2(b - c) + b2(c - fl) + c2(fl -b) = -(a - b)(b - c)(c - a).

Note that we can write the expression -(a - b)(b - c)(c - a) in many different ways. For example, because
-(fl - b) = b - a, we can write our final factorization as (b - a)(b - c)(c - fl). □

Problem 9.21: Factor (fl + b + c)3 - (b + c - a)3 - (c + a - b)3 - (fl + b - c)3.

Solution for Problem 9.21: Let the polynomial be ∕(fl, b, c). We try f(b, b, c) to see if a - b is a factor, but
this gives us
f(b, b, c) = (2b + c)3 -c3-c3- (2b - c)3.
However, this expression doesn't equal 0 for all b and c. (We can see this by multiplying out the cubes,
or simply by noting that if b = c = 1, the expression equals 24, not 0.) Similarly, we test f(-b, b, c) to see
if fl + & is a factor, but this fails, too.

281
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

We could start trying more complicated expressions, but first, maybe we should try something
simpler. Instead of letting a equal some expression involving b and/or c, we check what happens when
a = 0. Then, we have
∕(0, b, c) = (b + c)3 - (b + c)3 - (c - b)3 -(b- c)3,
which does indeed equal 0 for all b and c because (b + c)3 - (b + c)3 = 0 and -(c - b)3 - (b - c)3 -
(b - c)3 -(b- c)3 = 0. Since ∕(α, br c) = 0 when a = 0, the expression a - 0 is a factor of f(a, b, c). That is, a
itself is a factor of f(a, b, c). Similarly, b and c are factors of f(a, b, c). Since the degree of f{a, b, c) is clearly
no greater than 3, and we have 3 factors of ∕(fl, b, c), we have

f(a,b,c) = (a + b + c)3-(b + c-a)3-(c + a-bj3-(a + b-c)3 = kabc

for some constant k. We find k by letting a = b = c = 1, which gives us k = 24, so

(a + b + c)3 - (b + c - a)3 - (c + a - b)3 - (a + b — c)3 = 24afec.

llxercises

9.3.1 Show that (« - b)3 + (b- c)3 + (c - a)3 = 3(« - b)(b - c)(c — a).
9.3.2 Prove that
(a + b + c + rf)4 + (a + b - c - d)4 + (a - b + c - d)4 + (α - b - c + d)4
- (a + b + c-d)4 - (a + b-c + df -(a-b + c + d)4 - (-a + b + c + d)4 = 192abcd.

9.3.3 Factor abc - ab -be- ca + a + b + c-l.

b - c c-a a-b (c- b)(a - c)(b -d) ., ,
9.3.4 Show that-------- 1--- r—∣---------=------------- ;------------ if abc ≠ 0.
abc abc
9.3.5 Suppose that x 1 + y 1 + z 1 = (x + y + z) 1 for nonzero numbers x, y, and z such that x + y + z≠0.
Prove that x" + yn + zn = (x + y + z)n for all odd integers n. Hints: 99
*
9.3.6 Factor (x + y)7 - (x7 + y7) over the reals.

9.4 Summary

We can sometimes factor expressions by splitting them into groups of terms, factoring the groups, and
then discovering that the factored groups have factors in common. One common example is

ap + aq + bp + bq = a(p + q) + b(p + q) = (a + b)(p + <7).

This factorization is so useful that we sometimes "force" it into problems. For example, if we must
find all positive integer pairs such that 6xy + 4x + 9y = 247, we would add 6 to both sides to give us
6xy + 4x + 9y + 6 = 253, because we could then factor the left side to give (2x + 3)(3y + 2) = 253. We call
this process Simon's Favorite Factoring Trick.

282
 REVIEW PROBLEMS

Two useful factorizations of sums and differences of powers are

We can sometimes factor the sum of even powers of expressions by creating a difference of squares.
We do so by adding and subtracting an appropriate perfect square so that the result is the square of a
binomial minus the perfect square.

The Factor Theorem can be extended to multivariable polynomials. For example:

Important: If f(a,b) is a polynomial and there is a polynomial h(tf) in b such that

ψ ∕(∕7(fe)z b) = 0 for all b, then we can write f{a, b) = (a- h(b))g(a, b), where
g(a, b) is a polynomial.

We can also apply this to polynomials with more than 2 variables.

Problem Solving Strategies__________________________________________

Concepts: • A Diophantine equation is an equation for which we must find

o≡⅛ integer solutions. Factorization is a powerful tool for solving Dio­
phantine equations. If we can write a Diophantine equation as a
product of expressions set equal to a constant integer, we can then
use the divisors of the integer to find solutions to the equation.

• Look for ways to use factorization to avoid having to expand com­

plicated products.

• Factorization can be useful in number theory problems about divisi­

bility.

• Experiment with equations that you're unsure how to handle. Re­

arrange them and look for familiar expressions or forms. Don't just
stare at them!

• Don't be afraid to guess. A great many problems can be solved by

looking for patterns or clues, guessing an answer, then showing that
your guess is indeed the correct answer.

Review Problems

9.22 Let x = 2OO11002 - 2OO1^1°02 and y = 2OO11002 + 2001^ioo2. Find x2 - y1. (Source: HMMT)

283
CHAPTER 9. FACTORING MULTIVARIABLE POLYNOMIALS

9.23 Factor both 16y4 - 1 and 243x5 + 32z10 over the integers.
9.24 Find all tire roots of x7 - x4 - x3 + 1.
9.25 Find all 8 roots of x8 - 1.
9.26 If a ≠ b, a3 - b3 = 19x3, and a-b -x, then find a in terms of x. (Source: AHSME)
9.27 Find all ordered pairs such that x - y = 2 and x3 — y3 = 98.
9.28 Call a pair of numbers whose product is equal to their sum Simonish. For instance, 1.25 and 5 are
a Simonish pair: 5 + 1.25 = 5 ∙ 1.25 = 6.25. The number 2 is a Simonish pair with itself: 2 + 2 = 2∙2 = 4.
Find all Simonish pairs.
9.29 Graph all solutions to the equation 6xy + 1 = 3x + 2y.
9.30 Factor each of the following as much as possible:
(a) p2 - 2p + 1- q2 - 2qr - r2 (c) (a2 - b2)(b2 - c2) + (a- b)(b - c)
(b) 4x2y2 - (x2 -z2 + y2)2 (d)
* x2 + 5xy + 6y2 + x + 3y

9.31 Find all ordered pairs of integers (c, d) such that 6cd + 4c = 79 + 3d.
9.32 Factor 9α5 - 4α3 - 81α2 + 36 completely
(a) over the integers. (c) over the reals.
(b) over the rationals. (d) over the complex numbers.

9.33 Suppose a, b, and c are distinct numbers such that d2 - be = b2 - ca. Show that a = -(b + c).
9.34 Factor abc + abe + acd + ade + bef + bef + cdf + def completely.
9.35 Factor x12 - y12 completely:
(a) over the rationals. (b)
* over the reals.

b+c a+c a+b

9.36 Show that _ c) + _ fl) + (c _ a}(c _ ty = 0 for all distinct a, b, and c.

9.37 Factor x4 + 4y4 over real coefficients.

9.38 Factor (x - a)3(b - c)3 + (x - b')3(c - a)3 + (x - c)3(α - b)3 into a product of linear terms.

9.39 Let a, b, and c be three real numbers such that a + b,b + c, and c + a are nonzero. Show that
2a + 2b + 2c + (b - c)(c - a)(a - b) _
a+b b + c c + a (b + c)(c + a)(a + b)

Challenge Problems

9.40 The distinct real numbers x and y satisfy x2 = 33y + 907 and y2 = 33x + 907. Find xy. (Source:
WOOT)

284
 CHALLENGE PROBLEMS

9.41 If x3 = 1 and x ≠ 1, find the value of (1 - x + x2)( 1 + x - x2). Hints: 188

9.42 Find all ordered pairs (x, ι∕) such that x - xy3 = 7 and xy2 - xy = 3.

9.43 If x ≠ y and - + x = - + y, what is the value of - + -? Hints: 246

y x y x y
9.44 Let g(x) = x5 + x4+x3+χ2 + χ+ l. What is the remainder when the polynomial g(x12) is divided
by the polynomial g(x)? (Source: AHSME) Hints: 109, 284

9.45 Let Q(x) be the quotient when 37x73 - 73x37 + 36 is divided by x - 1. Compute the sum of the
coefficients of Q(x). (Source: ARML) Hints: 73
1 1
9.46 Given x + - = 1, calculate x7 + —
x x7
9.47 Factor each of the following as completely as possible over the reals.

(a) a3(c -b) + b3(a - c) + c3(b - a).

(b) * a4(c -b) + b4(a - c) + c4(b - a).

9.48 Let x and y be real numbers that differ in absolute value and satisfy the equations

x3 = 13x + 3yz
y3 = 3x + 13y.

Find (x2 - y2)2. (Source: AIME) Hints: 266

9.49 Let n be a positive integer. Show that (x - 1)2 is a factor of x" - n(x - 1) - 1. Hints: 287
111
9.50 Let z be such that z7 = 1 and z ≠ 1. Evaluate z10° + — + z300 + -ς-- + z500 d------ . Hints: 70
^KJO £,500

9.51 Factor x5 + x4y + x3y2 + x2y3 + xy4 + y5 completely over the reals.

9.52 Find two positive integers greater than 1 whose product is 66 + 84 + 274. Hints: 117

9.54 ★ Factor p3 + q3 + r3 - 3pqr. Hints: 85

9.55
* Write in the form fl(l - bc), where a, b, and c are constants.
Hints: 94
*
9.56 Factor x2(y + z) + y2(z + x) + z2(x + y) - x3 - y3 - z3 - 2xyz. Hints: 127
e. 1., 1 . a3(b + c) b3(c + a) c3(a + b)
*
9.57 S1mphfy the expression (a _ fe)(a _ c) + (fc _ c)((, _ a) ÷ (c _ a)(c _ 6) ∙

*
9.58 Determine all the solutions to x4 - 4x = 1. Hints: 354
*
9.59 Find the two real roots of 2000x6 + 100x5 + 10x3 + x - 2 = 0. (Source: AIME) Hints: 276

285
CHAPTER 10. SEQUENCES AND SERIES

We are continually faced with a series of great opportunities brilliantly disguised as insoluble problems.
- John W. Gardner

CHAPTER 1

Sequences and Series

A sequence is an ordered list of numbers, while a series is an ordered sum of numbers. For instance,
10,11,12,..., 99 is a sequence, while 10 + 11 + 12 + ∙ ∙ ∙ + 99 is a series. The order of the numbers in a
sequence or series matters. For example, the sequence 1, 2, 3, 4 is not the same as the sequence 3, 1, 2,
4. Sequences and series can also have infinitely many terms. For example, both the sequence 1,2,3,...
and the series l + ∣ + ∣ + ∣ + ⅛ + ∙∙∙ have infinitely many terms.

10.1 Arithmetic Sequences

Definition: An arithmetic sequence is a sequence in which the difference between any two
consecutive terms is the same. This fixed difference is known as the common difference of the
sequence.

For example,
28,33,38,43,48,53,58,63,68
is an arithmetic sequence, because each term is 5 greater than the previous term.
Arithmetic sequences are sometimes referred to as arithmetic progressions, and terms that form
an arithmetic sequence are said to be "in arithmetic progression." For example, 7, 9, 11, and 13 are in
arithmetic progression.

Problems

Problem 10.1: The first term of an arithmetic sequence is 5 and the sixth term is 40.
(a) Find the common difference of the sequence.
(b) Find a formula for the nth term of the sequence.
(c) Find the 717th term of the sequence.

286
 10.1, ARITHMETIC SEQUENCES

Problem 10.2: Let a be the first term of an arithmetic sequence with common difference d. Find a
formula for the πth term of the sequence in terms of a, d, and n.

Problem 10.3: Prove that if a, bf and c are in arithmetic progression, then for any constant k, the
numbers ak, bk, and ck are also in arithmetic progression.

Problem 10.4: An arithmetic sequence has first and last terms of 42 and -138 respectively. Its common
difference is -10.
(a) Find a formula for the fcth term of the sequence in terms of k.
(b) How many terms are there in the sequence?

Problem 10.5: The first four terms of an arithmetic sequence, in order, are x + y, x - y, xy, and x∕y.
(a) Using the first two terms, express the common difference of the sequence in terms of y.
(b) Using the common difference, find new expressions for the third and fourth terms in the se­
quence.
(c) Using your expressions from part (b), build and solve a system of equations to find x and y.
(d) Find the fifth term in the sequence. (Source: AMC10)

Problem 10.6: The roots of 64x3 - 144x2 + 92x - 15 = 0 are in arithmetic progression. Find them.
(Source: AHSME)

Problem 10.7: Let a, b, and c be in arithmetic progression.

(a) Prove that 2b = a + c.
(b) Prove that a2 - be, b2 - ac, and c2 - ab are in arithmetic progression.

Problem 10.1: The first term of an arithmetic sequence is 5 and the sixth term is 40. Find the 717th
term of the sequence.

Solution for Problem 10.1: Finding the common difference will allow us to calculate any term of the
sequence. In going from the first term to the sixth, we add the common difference 5 times. So, if the
common difference is d, then 5 + 5d = 40. Therefore, we have d = ⅛2 = 7.

For any positive integer n, the πth term in the arithmetic sequence is produced by adding the common
difference to the first term of the sequence n-1 times. Therefore, the nth term of the sequence is

5 + (n- 1)7 = 7n-2.

We can test this formula with the terms we already know. Letting n = 1 and n = 6 in our formula gives
7(1) -2 = 5 and 7(6) - 2 = 40. These are indeed the first and sixth terms.
We've done all the heavy lifting. Now we simply substitute n = 717 into our formula and find that
7(717) — 2 = 5017 is the 717th term of the sequence. □

287
CHAPTER 10. SEQUENCES AND SERIES

Problem 10.2: Let a be the first term of an arithmetic sequence with common difference d. Find a
formula for the nth term of the arithmetic sequence in terms of a, d, and n.

Solution for Problem 10.2: The nih term of an arithmetic progression is the result of adding the common
difference to the first term exactly n-1 times. The resulting nth term then is a + (n - l)d. □

Important: The Hth term of an arithmetic sequence that has first term a and common

ψ difference d is
cι + (n — l)d.

Problem 10.3: Prove that if a, b, and c are in arithmetic progression, then for any constant k, the
numbers ak, bk, and ck are also in arithmetic progression.

Solution for Problem 10.3: Because a, b, and c are in arithmetic progression, the difference between b and
a equals the difference between c and b. Therefore, we have

b - a — c - b.

To show that ok, bk, and ck are in arithmetic progression, we must show that the difference between
bk and ak equals the difference between ck and bk. In other words, we must show that bk- ak = ck - bk.
Factoring k out of both sides of this equation gives k(b - a) = k(c - b). This is just k times our equation
b - a = c - b from above. We can now write our proof as follows:

Proof: Because a, b, and c are in arithmetic progression, we have

0
b - a = c - b.

Multiplying this equation by k gives k(b - d) = k(c - b). Expanding both sides gives us
bk- ak = ck - bk. Therefore, ak, bk, and ck are in arithmetic progression.

In Problem 10.3, we multiplied an arithmetic sequence by a constant to generate another arithmetic

sequence. As an Exercise, you'll show that adding a constant to each term in an arithmetic sequence
produces another arithmetic sequence.

Problem 10.4: An arithmetic sequence has first and last terms of 42 and -138 respectively. How
many terms are there in the sequence if it has a common difference of -10?

Solution for Problem 10.4: Let n be the number of terms in the sequence. To get from the first term to the
last, we add the common difference n-1 times. Since the common difference is -10, we must therefore
have 42 + (n - l)(-10) = -138. Solving this equation, we find that n = 19 is the number of terms in the
sequence. □

288
 10.1. ARITHMETIC SEQUENCES

Concept: Each term in an arithmetic sequence can be expressed in terms of the first
θ-⅛jij term of the sequence (usually denoted a), the common difference of the
sequence (usually denoted t∕), and the position of the term in the sequence
(often denoted /?). When a problem does not provide some of these values,
we can introduce variables for them to create equations.

For example, in Problem 10.4, we are given a = 42,d = -10, and the nth term, so we can quickly write
an equation for n.

Problem 10.5: The first four terms of an arithmetic sequence, in order, are x + y, x - ι∕, xy, and x∕y, in I
that order. What is the fifth term? (Source: AMC10) I

Solution for Problem 10.5: The first two terms tell us that the common difference of the sequence is
(λ' - y) ~ (χ + J∕) = ~2y. Starting from our second term, x - y, and adding this common difference gives
us third and fourth terms of x — 3y and x — 5y respectively. These must match the given expressions for
the third and fourth terms of the sequence, so we have the system of equations

x - 3y = xy,
c x

x - 5y = -.
V

We solve the first equation for x in terms of y to find

3y
x= r-^-.
ι-y
We could go ahead and substitute for x in the second equation. However, we can first manipulate the
second equation a bit to get rid of the fraction, making substitution easier. Multiplying both sides by y
gives xy - 5y2 = x, and rearranging this gives 5y2 + x - xy = 0, from which we have

5y2 + (1 - y)χ = o.

Substituting for x, we get 5y2 + 3y = 0, which gives us the possible solutions y = 0 and y = -3/5.
However, because y is in the denominator of one of the terms of the sequence, we must have y ≠ 0.
Therefore, the only possible value of y is -3/5, which gives us

3y -I 9
ι-y^ I ^^s'

Now, using the common difference -2y again, the fifth term is

∕ r x + (-2y)
(χ-5y) r7 9 ∕ 3∖ = --
∕ α x = x-7y=---7(--) 21 = -.
9 +- 123

□
Problem 10.6: The roots of 64x3 - 144x2 + 92x - 15 = 0 are in arithmetic progression. Find them.
(Source: AHSME)

289
CHAPTER 10. SEQUENCES AND SERIES

Solution for Problem 10.6: We could assign variables to the values of the roots in many ways. We might
write them as a, a + d, and a + 2d, and then use Vieta's Formulas to note that
-144 9
a + (a + d) + (fl ÷ 2d) —
64 4'
However, we can also write the roots as r - d, r, and r + d. In other words, we write them in terms of
the middle term of the sequence instead of the first term. The advantage of this approach becomes clear
when we add the roots, which gives

(r - d) + r + (r + d) = 3r.

By Vieta, this sum must equal 9/4, so we have 3r = 9/4, which gives us r = 3/4.
Vieta's Formulas helped us find r, so we try it again to find d. By Vieta, the product of the roots is
15
(r-d)r(r + d) = —.

Since r = 3/4, we have

4∖16 1 64
Solving this equation gives d = ±1/2. The sign of d does not matter since the roots are r and r±d. From
our values of r and d, the roots of the cubic are 1/4,3/4, and 5/4. □
Note that in Problem 10.6, we set the middle term in the arithmetic sequence equal to a variable.

Concept: Setting the middle term of an arithmetic sequence with an odd number of
θ--ctj terms equal to a variable often introduces a symmetry that allows us to
solve problems more easily.

If the sequence has an even number of terms, we can often accomplish the same feat by assigning a
variable to the average of the two middle terms.

Problem 10.7: Prove that if the numbers a, b, and c are in arithmetic progression, then so are a2 - be, fc
b2 - ac, and c2 - ab.

Solution for Problem 10.7: The fact that a, b, and c are in arithmetic progression gives us b - a = c - b.
Solving this equation for b in terms of a and c gives
a+c

In other words, b is the average of a and c.

To show that the sequence a2 — be, b2 - ac, c2 - ab is an arithmetic sequence, we must show that

(c2 - ab) - (b2 - ac) = (b2 - ac) - (fl2 - be).

We rearrange this to put the terms with b on one side, since we know b in terms of a and c:

a2 + 2ac + c2 = 2b2 +ab + be.

290
 10.1. ARITHMETIC SEQUENCES

Aha! We have the square of a binomial on the left. Factoring b out of the right side leaves us the equation

(α + c)2 = b(2b + a + c)

to prove. Substituting b = (fl + c)∕2 makes the right side

(^T^)(2(⅛^) + fl + c)= ^(α + c)(" + c + fl + c) = (fl + c)2,

as desired.

We started from an equation we wish to show true,

(c2 - ab) - (b2 - ac) = (b2 - ac) - (fl2 - be),

and manipulated it to produce an equation that we proved is true. Make sure you see why this isn't yet
a complete proof! We must make sure that all our steps are reversible. That is, we must show that we
can start from true equations and work forwards to the desired equation. Using our work above as a
guide, we write our proof below:

Proof: Because a, b, and c are in arithmetic progression, we have c — b = b — a, so 2b = a + c. Multiplying

0 both sides of 2b = a + c by a + c gives 2fo(fl + c) = (a + c)2, so

(a + c)2 = b(2a + 2c) = b(a + c + a + c) = b(2b + a + c) = 2b2 + ab + be.

Expanding the left side of (fl + c)2 = b2 + ab + be gives

fl2 + 2ac + c2 = 2b2 + ab + be,

and rearranging this equation gives us

(c2 - ab) - (b2 - ac) = (b2 - ac) - (a2 - be),

so fl2 - be, b2 — ac, and c2 — ab are in arithmetic progression.

WARNING’! Working backwards is a great way to find a solution. However, if

4k you work backwards to find a solution, you must make sure that all
your steps are reversible. The best way to do so is to write your proof
forwards, starting from what is known and ending with what you
wanted to prove.

At the beginning of our solution, we showed that if fl, b, and c are in arithmetic progression, then b
is the average of a and c. As you'll show in the Exercises, the converse is also true, so that we have:

Important: The quantities fl, b, and c are in arithmetic progression if and only if b is

ψ the average of a and c.

291
CHAPTER 10. SEQUENCES AND SERIES

III
Exercises

10.1.1 The fourth term of a particular infinite arithmetic sequence is 203 and the thirteenth term is 167.
What is the smallest value of n such that the nth term of the sequence is negative?
10.1.2 The fifth term of an arithmetic sequence is 10 and the tenth term is 4.

(a) Find the twentieth term in the sequence.

(b) Find a formula for the nih term of the sequence.

10.1.3 Show that if b is the average of a and c, then a, b, and c are in arithmetic progression.
10.1.4 Prove that when a constant k is added to each term in an arithmetic progression, the resulting
sequence is also arithmetic.
10.1.5 In an attempt to copy down a sequence of six positive integers in arithmetic progression, a student
wrote down the five numbers 113,137,149,155,173, accidentally omitting one. He later discovered that
he also miscopied one of them. Can you help him recover the original sequence? (Source: USAMTS)
10.1.6
* Find all values of k such that x4 - (3k + 4)x2 + k2 = 0 has 4 real roots in arithmetic progression.
Hints: 265

10.2 Arithmetic Series

Now that we've solved a few problems involving arithmetic sequences, we move on to problems that
involve summing the terms of arithmetic sequences.

Definition: The sum of the terms of an arithmetic sequence is called an arithmetic series.

Problems

Problem 10.8: Consider the arithmetic sequence with first term a, last term b, and common difference
d, and also the sequence produced by reversing the order of the terms:

a, a + d, a + 2d,... ,b - 2d,b - d,b,

b, b - d, b - 2d,..., a + 2d, a + d, a.

(a) First, we find a formula for the sum of the terms in the original sequence.

(i) Add the first term of the first sequence to the first term of the second sequence. Do the
same with the second term of each sequence, and the third term of each sequence. Notice
anything interesting?
(ii) Suppose there are n terms in each sequence. Find the sum of all 2/z terms in both sequences,
then use your result to find a formula for the sum of the terms in the original sequence.

(b) Show that the average of all the terms in an arithmetic sequence is equal to the average of the
first and last terms.

r
292
 10.2. ARITHMETIC SERIES

Problem 10.9: Compute the sum of the arithmetic series 5 + 11 +17+∙∙∙ + 107.

Problem 10.10: Find a formula for the sum of the terms in an n-term arithmetic series with first term
a and common difference d.

Problem 10.11: The pages of a book are numbered 1 through n. When the page numbers of the book
are added, one of the page numbers was mistakenly added twice, resulting in the incorrect sum of
1986. What was the number of the page that was added twice? (Source: AIME)

Problem 10.12: Find all integers k such that 46 + 44 + 42 + ∙ ∙ ∙ + k = 510.

Problem 10.13: Let Λ1,α2∕Λ3,... be an arithmetic progression with common difference 1 and let

S] = flj + Λ2 ÷ ÷ ', - + ^98z

S2 = Ω2 ÷ fl4 + ⅜ + ' ’- + fl98 ■

(a) Find an equation relating Si to S2.

(b) Find the value of S2 given that Si = 137. (Source: AIME)

Sidenote: German mathematician Carl Friedrich Gauss (1777-1855) is widely con­

sidered one of the greatest mathematicians ever. Due to his significant
contributions to so many fields, he is sometimes called the "prince of math­
ematicians."
Gauss was a child prodigy, amazing adults with his mathematical
prowess at a young age in the way Mozart as a child wowed audiences
with his mastery of music. One account of Gauss has him discovering
an error in his father's payroll calculations at the age of three. Another
story recounts how one of Gauss's elementary school teachers attempted
to occupy students in class by having them compute the sum of the inte­
gers from 1 to 100. Gauss immediately wrote down the sum, 5050, having
recognized a convenient pairing of the terms:

1 + 2 + 3 + 4 + ∙ ■ ∙ +97 + 98 + 99 + 100
= (1 + 100) + (2 + 99) + (3 + 98) + (4 + 97) + ∙ ∙ ■ + (50 + 51)
= 101 + 101 + 101 + 101 + ∙ ∙ ∙ + 101
= 50(101) = 5050.

Expanding on Gauss's clever work, we now derive a formula for an arithmetic series given its first
term, last term, and total number of terms.

Extra! I mean the word proof not in the sense of the lawyers, who set two half proofs equal to a whole
∣ ∣⅛∣∣<l∣<∣l∣⅜. onC/ tout in the sense of a mathematician, where half proof equals nothing, and it is demanded for
proof that every doubt becomes impossible.
- Carl Gauss

293
CHAPTER 10. SEQUENCES AND SERIES

Problem 10.8: Consider an arithmetic series with n terms whose first term is a and whose last term
is b:
a + (a + d) + (a + 2d) + ∙ ■ ■ + (b - 2d) + (b - d) + b.
(a) Find a formula for the series (that is, a formula for the sum of all the terms) in terms of a, b,
and n.
(b) Compute the average of the terms in the series.

Solution for Problem 10.8:

(a) Inspired by the creativity of Gauss, we write the series both forward and backward. If S is the
sum of the series, then
S = a + (a + d) + (a + 2d) + --∙ +(b-2d) + (b-d) + b,
S = b + (b - d) + (b - 2d) + ■ ■ ∙ + (α + 2d) + (a + d) + a.

Instead of trying to sum across in each row, we add the equations by adding the terms in each
column to get
2S = (α + b) + (a + b) + (« + b) + ∙ ∙ ∙ + (c + b) + (« + b) + (α + b).
There are n terms in each of the series we just added, so the sum consists of n instances of a + b.
Therefore, we have 2S = n(μ + b). Dividing this equation by 2, we have S = n(a + b)∣2.
n κ r,. . r 11 . n(a + b) 1 . . n(a + b) a+b
(b) Since the sum of all n terms is —-—, their average is —-— ÷n = -.

□
In Problem 10.8 we proved the following highly useful fact:

Important: The average of the terms in an arithmetic sequence is equal to the average

ψ of the first and last terms of the sequence.

Therefore, we see that:

The sum of the terms in an arithmetic sequence equals the product of

the number of terms and the average of the first and last terms.

Problem 10.9: Compute the sum of the arithmetic series 5 + ll + 17 + 23 + 29 + ∙∙∙ + 107.

Sol ιιl ion for Problem 10.9: We have the first and last terms, so all we need is the number of teπns to sum
the series. We first note that the common difference of the series is 11 -5 = 6. So, we take(107-5)∕6 = 17
steps of 6 going from 5 to 107. Therefore, there are 18 terms in the series—the first term plus one for
each step.

WARNING!! The term that is A terms after the Ist term of a sequence is the (A' + l)th
V term of the sequence, not the Alh term.

294
 10.2. ARITHMETIC SERIES

The average of the first and the last terms is (5 + 107)∕2 = 112/2 = 56. This is also the average of all
the terms, so the sum of the arithmetic series is 18 ∙ 56 = 1008. □

Concept: While we often use the sum of a group of numbers to compute their
Q σcJ average, we can also use the average of a group of numbers to compute
their sum.

Problem 10.10: Find a formula for the sum of the terms in an n-term arithmetic series with first term
a and common difference d.
KMHi

Solution for Problem 10.10: The first term of the arithmetic series is a and the last term is a + (n — l)rf. The
average of these terms, which is also the average of all the terms in the series, is

a + [a + (n - 1)<2] 2a + (n- l)d

2 = 2 '
Therefore, we find

Important: The sum of an n-term arithmetic series with first term a and common

ψ difference d is
n(2fl + (n- l)d)
2

Don't bother memorizing this formula. It's just another way of expressing the fact that the sum of an
arithmetic series is the product of the number of terms in the series and the average of the first and last
terms. □
n(2a + (n- l)d)
The simple formula 2— is much easier to work with than the equivalent series

u + (fl + d) + (fl + 2d) + ∙ ∙ ∙ + (a + (π — l)d).

We call this simple formula a closed form for the series. A closed form is basically an expression that
doesn't involve a summation with "∙ ∙ ∙" or some other repeated operation.

Problem 10.11: The pages of a book are numbered 1 through n. When the page numbers of the book
are added, one of the page numbers was mistakenly added twice, resulting in the incorrect sum of
1986. What was the number of the page that was added twice? (Source: AIME)

Solution for Problem 10.11: The sum of the page numbers in the book is 1 + 2 + ∙ ∙ ∙ + n. This is an arithmetic
series with first term 1, last term n, and n terms, so its sum is

Important:
n(n + 1)
ψ l+2 + 3 + ∙∙∙ + n =
2

But one page was added twice to get our sum of 1986. Suppose page k was added twice, so we
have '112±∖1 + jς = 1986. Solving for k gives us k = 1986 - . We don't know n, but we do know that

295
CHAPTER 10. SEQUENCES AND SERIES

1 ≤ A' ≤ π, because k is a page number of the book. So, now we can simply guess values of n, evaluate
1986 - "---,+1 -, and see if this expression gives us a positive integer from 1 to n.

Concept: Don't overlook trial and error as a problem-solving strategy. Some prob-
Q" l⅛ lems are most easily solved by guessing solutions and checking if they
work.

Letting n = 60 gives 1986 - π-"2+- = 156, so we need a larger value of n. Letting n = 61 gives k = 95,
n = 62 gives k = 33, and n = 63 gives k = -30. Larger values of n give negative values of k, and values
of n smaller than 60 give values of k larger than 60. So, the only value of n that gives a value of k from 1
to n is n = 62, which gives k = 33, the number of the page that was added twice. □
Notice that in our solution, we used trial and error to find a value of n that solves the problem, and
we also explained why no other values of n work. The latter step is very important, as our next problem
exhibits.
Problem 10.12: Find all integers k such that 46 + 44 + 42 +---- ι-fc = 510.

Solution for Problem 10.12: A complete understanding of arithmetic series helps us avoid mistakes like
the following:

Bogus Solution: We start with 46 + 44, then keep adding successive terms until we
have a sum of 510. We eventually find that 46+44+42 4---- t-16+14 =
510, so we know that k = 14.

While k = 14 is a solution to the problem, it's not the only one!

WARNING!! If you must find all possible solutions to a given problem, simply find­
λ.⅛
A ing one solution isn't enough. You must not only find the solutions to
the problem, but you must also show that these are the only solutions.
Using "trial and error" or "guess and check" isn't enough to do the
latter—you might overlook a solution.

We'll use an algebraic approach to the problem to make sure we find all possible solutions. We
take (k - 46)∕(-2) steps of size -2 in going from 46 down to k, so the number of terms in the series
46 + 44 + 42 + ∙∙∙+kis
k - 46 , 48 - k
—+1 = -
. 46 + k . .
The average of these terms is —-—, so their sum is

48 - k 46 + k _ 2208 + 2k - k2
~~2 2^^~ ^^ 4
This sum must be 510, so we have
2208 + 2A' - k2
= 510.
4

296
 10.2. ARITHMETIC SERIES

Rearranging this equation gives k2 - 2k - 168 = 0, and solving this quadratic gives us k = 14 and k = -12
as solutions. Notice how these two solutions are related. The series for k = —12 equals the series for
A: = 14 plus the sum 12 + 10 + ∙ ∙ ∙ + (-10) + (-12). This latter sum obviously equals 0, since each nonzero
term is canceled by its opposite. □

Problem 10.13: Find the value of «2 + «4 + «6 + ∙ ∙ ∙ + «98 if «1/ «2/ «3/ ∙ ∙ ∙ is an arithmetic progression I
with common difference 1, and α1 + a2 + «3 + ∙ ∙ ∙ + a9s = 137. (Source: AIME)

Solution for Problem 10.13: Solution 1: Write everything in terms of a↑. Because we know the common
difference of the arithmetic sequence, we can write the left side of the given equation entirely in terms
of «1. Specifically, we have a2 = «1 + 1, a2, = a↑ + 2, «4 = a↑ + 3, and so on, so our equation is

«1 + («1 + 1) + (β] + 2) + ∙ ∙ ∙ + (a-[ + 97) = 137.

Concept: We can often tackle problems involving arithmetic series by writing all the
given information in terms of the first term, the common difference, and
the number of terms in the series.

There are 98 terms, and the sum of the first 97 positive integers is 97(98)∕2 = 4753, so we find that
98«i + 4753 = 137. This gives us a-y = -2308/49. Therefore, the series a2 + «4 + «6 ■+-------- ∏ «98 has first term
«2 = «1 + 1 = —2259/49, common difference 2, and 49 terms, so its sum is
49(2flg + (48)(2)) _ 93

There's a lot of computation in this solution, and a lot of ways to make careless arithmetic mistakes.
As soon as we saw that «1 is a pretty ugly fraction, we might instead take a few minutes to look for a
simpler solution.

Concept: Before wading through excessive computation or algebraic manipulation

Q lj∣j in a problem, look for ways to solve the problem more creatively.

Solution 2: Clever manipulation. We have the sum of one arithmetic series and we want to find the
sum of a related series. Instead of finding «1, we look for a way to use the relationship between these
two series to manipulate the series we know into the one we want. To form the series we want, we must
eliminate all the odd-numbered terms from «1 + «2 + «3 +, ∙ ∙ + «98- Each odd-numbered term in the series
is 1 less than the next even-numbered term, so we have

137 = «i + «2 + «3 ÷ ■'' «98

= («2 — 1) ÷ «2 ÷ («4 - 1) + «4 ÷ ‘, - ÷ («98 - 1) + «98

= (2a2 — 1) + (2«4 — 1) + ∙ ∙ ∙ + (2«98 — 1)

= 2(a2 + «4 + ∙ ■ ∙ + «9e) - 49.

We now have a nice equation, 2(a2 + «4 + «6 ^,--------«9s) — 49 = 137. Adding 49 to both sides and dividing
by 2, we see that «2 + «4 + «6 + ∙ ∙ ∙ + «98 = 93. □
Notice that we didn't solve for any variable in our second solution.

297
CHAPTER 10. SEQUENCES AND SERIES

Concept: Algebraic relationships between sequences and series can help us compute
Q. v,l9 some series without evaluating any of the terms in the series.

I κeι ci¼es

10.2.1 Find the sum of the arithmetic series 639 + 631 + 623 + ■ ∙ ∙ + (—97).
10.2.2 Mrs. White has 13 grandchildren. If the youngest is 5 years old and they are spaced 2 years
apart, what is the sum of the ages of all of Mrs. White's grandchildren?
10.2.3 Let n be a positive integer.
(a) Find the sum of the n smallest positive odd integers.
(b) Find the sum of the n smallest positive odd integers greater than 100.
10.2.4 Let a and b be positive integers with a < b. Find the sum of the positive integers from a to b,
including a and b, in terms of a and b.
10.2.5 Let a↑, ..., and b↑, b2, ... be arithmetic progressions such that a↑ = 25, ½ = 75, and
flιoo + b↑oo = 100. Find the sum of the first 100 terms of the progression a↑ + bi, a2 + ½, ∙ ∙ ∙ ∙ (Source:
AHSME)
10.2.6 For every n, the sum of the first n terms of a certain arithmetic progression is 2h + 3n2. Find a
formula for the ∕ιth term in the progression. (Source: AHSME)
10.2.7 If the sum of the first 3n positive integers is 150 more than the sum of the first n positive integers,
then find the sum of the first 4n positive integers. (Source: AHSME)
10.2.8
* In a certain arithmetic progression, the ratio of the sum of the first r terms to the sum of the
first s terms is r2∕s2 for any r and s. Find the ratio of the 8th term to the 23rd term. (Source: ARML)

10.3 Geometric Sequences

Definition: A geometric sequence is a sequence for which there exists a constant r such that each
term (except for the first term) is r times the previous term. This constant r is called the common
ratio of the sequence.

For example,
1,3,9,27,81,243
is a geometric sequence, because each term is 3 times the previous term. Geometric sequences are also
known as geometric progressions.

Problems

Problem 10.14: The first term of a geometric sequence is a, and the common ratio of the sequence is r.
Find the πth term of the sequence in terms of rt, r, and u.

298
 10.3. GEOMETRIC SEQUENCES

Problem 10.15: The first term of a geometric sequence of positive numbers is 12, and the fourth term
is 24.
(a) Find the common ratio of the geometric sequence.
(b) Find a formula for the πlh term of the geometric sequence.
(c) Find the 10lh term of the geometric sequence.

Problem 10.16: The geometric mean of n positive numbers is the ∕Th root of the product of the
numbers. Let a be the first term of a geometric sequence with n terms, and let r be the common ratio
of the sequence. Show that if a and r are positive, the geometric mean of the terms in the sequence is
equal to the geometric mean of the first term and the last term.

Problem 10.17: Suppose x,y,z is a geometric sequence with common ratio r and x ≠ t∕. If x, 2ι∕, 3z is
an arithmetic sequence, find the value of r. (Source: AHSME)

Problem 10.18: Prove that if a, b, c, and d are four consecutive terms in a geometric sequence, then

(& - c)2 + (c - a)2 + (d - ⅛)2 = (fl - d)2.

(Source: CEMC)

Problem 10.14: The first term of a geometric sequence is a, and the common ratio of the sequence
is r. Find the nth term of the sequence in terms of a, r, and n.

Solution for Problem 10.14: The first term of the geometric sequence is a. The second is ar. The third is
ar2. In general, we get the κth term of the sequence by multiplying the first term of the sequence by r a
total of n - 1 times, so the nth term is flr',-1. □

Important: If a geometric sequence has first term a and common ratio r, then the nth

ψ term of the sequence is equal to flr"^1.

Problem 10.15: The first term of a geometric sequence of positive numbers is 12, and the fourth term
is 24. Find the 10th term of the geometric sequence.

Solution for Problem 10.15: Solution 1: Find the common ratio. Let r be the common ratio. We multiply the
first term by r three times to get the fourth term, so 12r3 = 24. Dividing by 12 and taking the cube root
of both sides, we find that r = ^2 is the common ratio of the sequence. The tenth term of the sequence
is therefore 12 ∙ (y∕2)10^^1 = 12 ■ (21/3)9 = 12 ∙ 23 = 96.

Solution 2: Think about it! Rather than finding the common ratio, we notice that in addition to
multiplying by the common ratio three times to get from the first term to the fourth, we also multiply by
the common ratio three times to get from the fourth term to the seventh, and to get from the seventh term
to the tenth. Since multiplying 12 by the common ratio three times gives us 24, the result of multiplying

299
CHAPTER 10. SEQUENCES AND SERIES

a number by the common ratio three times is doubling the number. So, the seventh term of the sequence
is 24(2) = 48 and the tenth term is 48(2) = 96. □
In Section 10.2, we saw that the average of all the terms in an arithmetic sequence equals the average
of the first and last term. Is there a similar relationship for the terms in a geometric sequence?

Problem 10.16: The geometric mean of n positive numbers is the nth root of the product of the
numbers. Prove that for any finite geometric sequence of positive numbers, the geometric mean of
the terms in the sequence is equal to the geometric mean of the first term and the last term.________ I

Solution for Problem 10.16: Let n be the number of terms of a geometric sequence with first term a and
common ratio r. The last term of this sequence is arn~∖ so the geometric mean of the first and last terms
is
Vα ∙ αrπ^1 = Vα2rπ-1 = α∕w-1^2.
The geometric mean of all n terms is

^Ja(ar)(a72') ∙ ∙ ∙ (αrn^1) = Vα" ∙ rli+2+∙→(n-i)l

= . r(π-l)n∕2
= fl/"’1)/2,

which matches our expression for the geometric mean of the first and last terms of the sequence. □

Problem 10.17: Suppose x,y,z is a geometric sequence with common ratio r and x ≠ y. If x,2y,3z is
an arithmetic sequence, find the value of r. (Source: AHSME)

Solution for Problem 10.17: We can start by reducing the number of variables we have to deal with. From
our geometric sequence, we know that y = rx and z = r2x. Now that we can write y and z in terms of r
and xr we can write our arithmetic sequence in terms of just x and r. So, we know that x, 2rx, and 3r2x
are in arithmetic progression, which gives us

2rx - x = 3r2x - 2rx.

Rearranging this equation gives us 3r2x - 4rx + x = 0. Factoring this equation gives

x(r - l)(3r -1) = 0.

We cannot have x = 0 or r = 1, since both of these give us x = y, which is forbidden in the problem
statement. This leaves us with r = 1/3 as the only possible value of r. □

Problem 10.18: Prove that if a, b, c, and d are four consecutive terms in a geometric sequence, then

(b - c)2 + (c - λ)2 + (d - bj2 = (a- d)2.

(Source: CEMC)

300
 10.3. GEOMETRIC SEQUENCES

Solution for Problem 10.18: Let r be the common ratio of the geometric sequence, so that b = ar, c = ar2,
and d = arf Then, on the left side of the relationship we must prove, we have
(b - c)2 + (c - λ)2 + (d - b)2 = (αr - ar2)2 + (ar2 - a)2 + (flr3 - ar)2
= a2r2(l - r)2 + a2(r2 - 1)2 + a2r2(r2 - 1)2
= α2((r4 - 2r3 + r2) + (r4 - 2r2 + 1) + (r6 - 2r4 + r2))
= α2(r6 - 2r3 + 1)
= α2(r3 -1)2.
Over on the right side of the desired equation, we have (α - d)2 = (α - αr3)2 = α2(l - r3)2 = α2(r3 - 1)2, so
we see that (b - c)2 + (c - «)2 + (d - b)2 = (α - d)2. □

Concept: Many problems involving geometric sequences can be tackled by writing

Oa-L⅛J the sequence in terms of the sequence's first term and its common ratio.

π
Exercises

10.3.1 The second term in a geometric sequence of positive numbers is 6 and the sixth term is ∣.
(a) Find the common ratio of the sequence.
(b) Find a formula for the mth term of the sequence.
(c) Find the value of a if the 1977th term of the sequence is ^∙.
10.3.2 The Steering Committee of the EmailTag Club sends a message to all the members of the club
as follows: On Day 1, each member of the Steering Committee emails three other people in the club. On
each subsequent day, each person who received an email the day before emails three other people in the
club. If no one ever gets two copies of the email on the same day, and there are 972 more emails on Day
5 than on Day 4, then how many people are on the Steering Committee?

10.3.3 A sequence of three real numbers forms an arithmetic progression with a first term of 9. If 2 is
added to the second term and 20 is added to the third term, the three resulting numbers form a geometric
progression. What is the smallest possible value for the third term in the geometric progression? (Source:
AMC12)
10.3.4
(a) Show that nonzero numbers x, y, and z are in geometric progression if and only if y2 = xz.
(b) Is it true that nonzero numbers x, y and z are in geometric progression if and only if y = -√χz7
10.3.5 The roots of 2x3 - 19x2 + kx - 54 - 0 are in geometric progression for some constant k. Find k.

*
10.3.6 The sum of three consecutive terms in a geometric sequence is 39, and the sum of their squares
is 741. Find the three terms. Hints: 51

Extra! Great things are not done by impulse, but by a series of small things brought together.
*""
* l"* lll* - Vincent van Gogh

301
CHAPTER 10. SEQUENCES AND SERIES

10.4 Geometric Series

Definition: The sum of the numbers in a geometric sequence is called a geometric series.

For example, the series 4 + 8 + 16 + 32 + 64 is a geometric series. In this section, we will also work
with infinite geometric series, which are geometric series with infinitely many terms. For example, the
infinite geometric series that starts with 1 and has common ratio 5 is
111
1 + 2 + 4 + 8+'"'
where the "∙ ■ ∙" indicates that the series continues infinitely.

Problems

Problem 10.19: In this problem, we evaluate the sum S = a + ar + ar2 ■+----- F artl 1, where r ≠ 1, in terms
of a, r, and n.
(a) By what expression can we multiply our expression for S to produce another series that has
many terms in common with S?
(b) Compute S in terms of a, r, and n.
(c) Explain the relationship between your result from (b) and a factoring method we've discussed
in a previous chapter.

Problem 10.20: Note that 38 = 6561. Compute the geometric series 1 + 3 + 9 +27+ ∙∙∙ + 2187.

Problem 10.21: Write the following expression as a single polynomial:

(f2 + f + l)(f12 + t9 + f6 + t3 + 1)
f1° +15 + 1

(Source: Mandelbrot)

Problem 10.22: In this problem we learn how to evaluate the infinite geometric series

1111
2 + 4 + 8 + 16+ -

We start by letting s=∣ + ^ + j + ⅛ + ''* + ^∙

(a) Explain why S cannot be greater than 1. z

(b) By what number can we multiply S to create another series that has nearly all of its terms in
common with S?
1
(c) Use part (b) to explain why S = 1 - —.
(d) Evaluate S for n — 5, n = 10, and n ≈ 20. (Yes you can use a calculator!)
Ill I
(e) Use parts (c) and (d) to explain why 3 4^ 4 + ~ ,

302
 10.4. GEOMETRIC SERIES

Problem 10.23: Find a formula for the sum of an infinite geometric series with first term a and common
ratio r, where ∣r∣ < 1. Why must we have the ∣r∣ < 1 restriction?

8 16
Problem 10.24: Evaluate the sum 4 -

Problem 10.25: Find all infinite geometric series that have 4 as the second term and a sum of 18.

Problem 10.26: The sum of the geometric series a + ar+ar2+ar3-∖— is 15, and the sum of the geometric
series a + ar2 + ar4 + ar6 + ∙ ∙ ∙ is 9. Find the sum of the geometric series a + ari + ar6 + αr9 + ∙ ∙ ∙.

Problem 10.27: Let P be the product of n numbers in geometric progression, S their sum, and S' the
sum of their reciprocals. In this problem, we find P in terms of S, S', and n. (Source: AHSME)
(a) Let the first term of the progression be a and the common ratio be r. Find P, S, and S' in terms
of a, r, and n.
(b) Combine your expressions from part (a) to write P in terms of S and S'.

Problem 10.19: Find the sum of the geometric series a + ar + ar2 + ∙ ∙ ∙ + arn 1, where r ≠ 1, in terms
of a, r, and n.

Solution for Problem 10.19: We set the series equal to S, hoping to perform a manipulation similar to the
one we used to find a formula for arithmetic series:

S = a + ar + ar2 + -∙- + arn~1.

We can't simply write this series backwards and add it to the original series to get our answer. However,
if we multiply this series by r, we create another series that has many terms in common with S:

rS = ar + ar2 + ∙ ∙ • + arn 1 + ar'

S - a + ar + af2 + ∙∙ • + arn~1.

Subtracting the second equation from the first cancels most of the terms and leaves

rS - S = artl - a,

so S(r -1) = a(rn - 1). Dividing by r - 1 gives us our formula, S - a(rn - lf∣(r - 1). To avoid dividing
by 0, we must note that we cannot have r = 1. If we have an n-term geometric series in which r = 1, the
series simply consists of n copies of the first term. So, if we call the first term a, the sum is just an.

Important: A geometric series with n terms, first term a, and common ratio r (with

ψ r ≠ 1) has sum
a(rn -1)
r-1 ‘

303
CHAPTER 10. SEQUENCES AND SERIES

We might also have found this formula by factoring a out of the whole series:
arn~l + arn~2 + ∙ ∙ ∙ + arλ + ar + a = α(r"^1 + rn~2 + ∙ ∙ ∙ + r2 + r + 1).
We recognize r',^1 + rn~2 + ∙∙∙ + r2 + r+ las one of the factors of rn - 1:
rn - 1 = (r - l)(r"^1 + r”-2 + ∙ ∙ ■ + r2 + r + 1).
Dividing both sides by r - 1 gives rn~1 + rn~2 + ∙ ∙ ∙ + r2 + r + 1 = (r" — l)∕(r — 1), so we have

arn~1 + arn~2 + ■ ■ ■ + ar2 + ar + a = a(rn~1 + rn~2 + ∙ ∙ ■ + r2 + r + 1) =

□
Problem 10.20: Note that 38 = 6561. Compute the geometric series 1 + 3 + 9 +27+ ∙∙∙ + 2187. ∣

1 2 _y 38 - 1
Solution for Problem 10.20: We simply apply the formula we just found: 1+3+3 H------ 1-3 - _∣ -
6561 - 1 nnnn
---- - ---- = 3280. □

Problem 10.21: Write the following expression as a single polynomial:

(t2 +1 + l)(t12 + ⅜9 + ⅜6 + f3 + 1)
t1° + f5 + 1
(Source: Mandelbrot)
Solution for Problem 10.21: We could multiply out the numerator and wade through long division
of the product by the denominator. But each factor is a geometric series, so maybe we can use our
understanding of geometric series to avoid long division. Writing t2 + t + 1 as 1 + t + t2, we recognize it
as a geometric series with first term 1 and common ratio t, so we have
l(f3 - 1) t3-l
1 + t + t2 =
t-1 t-1 ’
You might also have recognized this relationship by noting that t2 + t + 1 is a factor of t3 -1.
Similarly, the series 1 + t3 + tβ + t9 + t12 is geometric with first term 1 and common ratio t3, and the
series 1 + t5 + t1° is geometric with first term 1 and common ratio t5. So, we have

1+f3 + f6 + f9 + f12zz l((t3)5 - 1) f15 - 1

t3 - 1 ^ t3 - Γ
l((t5)3 - 1) _ ti5 - 1
1 + t5 + t1° =
t5 - 1 t5 -1 ‘

We see a lot of common factors among our new expressions. So, we expect we can do a lot of
cancellation:
(t2 + t + l)(t12 + t9 + t6 + t3 + 1) _ ⅛ ∙ ⅞⅛ _ (t3 - l)(t15 - l)(t5 - 1)
t10 + t5 + l - ~ (t15-l)(t-l)(t3-l)
= √1(t - ±f12±υ = t< + t3 + t2 + t + l.

t-1 t-1

304
 10.4. GEOMETRIC SERIES

Note that if t = 1, we have division by 0 in many steps of this solution. So, our solution only shows that
the original expression equals t4 + t3 + f2 + f + 1 if t ≠ 1. We take care of the case t = 1 by noting that
when t = 1, the original expression equals (3)(5)∕3 = 5, and f4 + f3 +12 + f +1 = 5. Therefore, the original
expression does indeed equal t4 + t3 + t2 + t + 1 when t = 1, as well. □

Concept: When x ≠ 0 and x ≠ y, the factorization

Ore
x" - y” = (χ- y)(x"~1 + xn~2y + xn~3y2 + ■■■ + xyn~2 + y"~1)

is the same as the relationship found by summing the geometric series

with first term x,,~1 and common ratio y/x:

χn-ι ((2)n) _ 1
X» 1 + χn 2y + χn-2>y2 + ... + χy∏-2 ψ yι-l _ ------ -------------
x 1
y” γn-l
X λ
7-1
y” - x” x" - y"
y-x - x-y -

We most frequently use this relationship when confronted with a polynomial that is a geometric
series, such as t2 + t + 1, t1° + t5 + 1, or t12 + t9 + t6 + t3 + 1.

Some geometric series never end. These are called infinite geometric series. Despite the fact that
these series are infinite, we can still find the sum of some of them.

Problem 10.22: Evaluate the infinite geometric series - + - + - + ■ ∙ ∙.

Solution for Problem 10.22: We can view our series as starting from 0 on the number line, and taking
steps to the right of length ∖, then ∣, then §, and so on. Our first step, ∣, takes us half the distance from
0 to 1. We are therefore ∣ away from 1. The next step, ∣, takes us half the distance from ∣ to 1. Wezre
then at ∣, which is still ∣ from 1. The next step, ∣, covers half the distance from ∣ to 1. Continuing in
this way, each step halves our distance from 1.

Since each step just halves our distance from 1, it appears that we never quite get to 1 and we certainly
can't get beyond it.
It appears that the series
1 1 1
-+-+-+

305
CHAPTER 10. SEQUENCES AND SERIES

never quite gets to 1. But what number does it equal?

To answer that question, we start with something that we know how to deal with: a finite geometric
series.

Concept: When stuck on a problem, compare it to a similar problem you know how
o≡eg to solve.

Suppose we instead try to find an expression for the sum

This is a finite geometric series with first term ∣, common ratio ∣, and n terms, so we can use our earlier
formula to find a closed form for it:

S(n) =

Now, we can easily evaluate S(n) for different values of n. A few such values are shown in the table
below. (The last two are approximations.)

n ∣S(H)
1 0.5
5 0.96875
10 0.9990234
20 0.9999990

We see that the larger n gets, the closer S(n) gets to 1. Let's see why. Since

S(n) = 1 - ⅛,

the value of S(n) cannot ever be larger than 1. However, the value of ⅛ gets closer to 0 as n gets larger,
so S(ιι) will get closer to 1 as n gets larger.
If "π is infinite," meaning our series never stops, we have our original infinite series,

1 1 1
2+4+8+ ’

Moreover, when "n is infinite," the value of is 0, so S(π) = 1 - equals 1 when n is infinite. Therefore,
we have

This explanation omits several details about exactly what we mean by "n is infinite," but it should
give you an intuitive understanding why our infinite series has a finite sum.

306
 10.4. GEOMETRIC SERIES

Sidenote: The details we omitted are some of the beginning steps of calculus. You'll
learn much more about these details later when you study the mathematical
concept of a limit.

Once we know our infinite series has a finite sum, we can find that sum even more quickly using the
same manipulation we used in Problem 10.19.

Let our infinite geometric series equal T. We multiply T by 1 /2, producing a series that has all of its
terms in common with T:

Subtracting the second equation from the first gives ⅜ = ⅛∙ All the other terms cancel! Therefore, T = 1.
□

A key step in our investigation of this infinite series was looking at sums of a finite number of terms
of the series. These sums of just part of a series are referred to as partial sums. For example, j +
Σ + 4 + and 2 + I + I + ⅛ are a^ partial sums of the infinite geometric series j + j + § + "••

Concept: We can often learn more about an infinite series by investigating partial
O≡ sums of the series.

Let's find a formula for computing the sum of an infinite geometric series.

Problem 10.23: Find a formula for the sum of an infinite geometric series with first term a and
common ratio r, where ∣r∣ < 1. Why must we have the ∣r∣ < 1 restriction?

Solution for Problem 10.23: The series we wish to evaluate is

a + ar + ar2 + ar3 + ■ ∙ ■ .

We use our previous problem as a guide, and let T(n) be the finite series

T(n) = a + ar + ar2 + ar3 + ∙ ∙ ∙ + arn~1.

This finite geometric series has first term a, common ratio r, and n terms, so we can use our formula for
the sum of a finite geometric series to give

arn - a arn a
T(H) =
r-1 rτT r-l'

As we did before, we evaluate the infinite series by thinking about what happens to T(n) when n is
infinite. If ∣r∣ < 1, then rn becomes 0 when n is infinite. This is true no matter how close to 1 that r is.

307
CHAPTER 10. SEQUENCES AND SERIES

(Grab your calculator and calculate O.999,oooooo.) So, when ∣r∣ < 1 and n is infinite, the term arn∣(r - 1)
equals 0, and we have
n □ ~u U
a + ar + ar + ar h— = —- = ------ .
r-1 1—r

If r = 1, then our infinite series is just a + a + a + ■ ■ ■, which we clearly cannot evaluate. If r > 1, then
rn just gets bigger and bigger as n gets larger, so T(n) either keeps getting bigger (if a > 0) or it keeps
getting smaller (if a < 0). In either case, we say that the series diverges, and it does not have a finite
sum.

Sidenote: If r = —1, we have a different problem. Then, our infinite series is

a-a +a-a+a-a+a-a+ ∙∙∙ .

After 2 terms, the sum is 0, but after 3 terms, the sum is a. After 4 terms,
the sum is 0 again, and after 5 terms, the sum is a. So, if the series has an
even number of terms, then the sum is 0. If the sum has an odd number of
terms, then the sum is a. But our infinite series doesn't have an even or an
odd number of terms.
It turns out that there is no way to define what the sum of this series is,
and we call such a series an indeterminate series.
If r < -1, then we have similar problems. As we add more and more
terms of the series, the sum goes back and forth from being much less than
0 to being much greater than 0. Therefore, we can't even say whether or
not the sum is negative or positive, so we certainly can't evaluate it.

Once we know that an infinite series does have a sum, we could also use the shortcut we used in our
last solution to Problem 10.19. We let our infinite series equal S:

S = a + ar + arλ + arl + ∙ ∙ ∙ .

We multiply this equation by r to produce a series that has all of its terms in common with S:

S = a + ar + ar2 + ,
rS = ar + aι2 + ∙∙∙ .

We then subtract, forcing nearly all the terms on the right to cancel and leaving

S - rS = α.

Factoring out S and dividing by 1 - r gives us:

Important: The sum of an infinite geometric series with first term a and common

ψ ratio r, where ∣r∣ < 1, is

1 -r

308
 10.4. GEOMETRIC SERIES

Concept: While the formula for an infinite geometric series is very handy, pay more
θ≡Uj attention to the method we used to derive it. Wezll use this method again,
and you can always use it to rederive the formula if you forget it.

As we saw earlier, we must include the restriction ∣r∣ < 1 because otherwise the geometric series
cannot be evaluated. An infinite series that we can evaluate is called convergent, and we say that such a
series converges. On the other hand, an infinite series that doesn't converge, such asl + 2+4+8 + 16 + ∙∙∙,
is called divergent, or is said to diverge. □
Let's try out our new formula.

Problem 10.24: Evaluate the sum 4-- + - - — + ∙∙∙.

............... 3 9 27

Solution for Problem 10.24: We have an infinite geometric series with first term 4. The common ratio
between the terms is the ratio of the second term to the first, or

_8 7
3 _ 2
4 3’
We have an infinite geometric series with first term 4 and a common ratio with absolute value less than
1, so we can use our formula to find that

8 16 32 _ 4 _ 4 _ 12
3+ 9 ^27+ ∙=1,(.∣)= ∣ = S'

Just as we drew a picture on page 305 for an infinite geometric series with a positive ratio, we illustrate
below the addition of the first few terms of the geometric series with negative ratio from Problem 10.24.

z xX
*-
1'∖"'
I C∖ I "1 r
0 3
κ
^-^^27
12
5
ZIQ Λ
81. '' 9
8 .-≈4

The first long dashed arrow on top represents the initial 4. Each subsequent dashed arrow represents
adding another term in the series. For example, the term after the 4, which is -∣, is the longest arrow on
the bottom. As we add more and more terms, we'll continue to get closer and closer to y. Each positive
term we add makes our sum a little greater than y, and each negative term we add makes our sum a
little less than y. In both cases, we get closer to y with each term we add.
∣ —-—————— ---------------------------------------—---------------------------------------- ■.---------------------------------------------------- ----- I

j Problem 10.25: Find all infinite geometric series that have 4 as the second term and a sum of 18.

309
CHAPTER 10. SEQUENCES AND SERIES

Solution for Problem 10.25: Let a be the first term of the series, and let r be the common ratio. Since the
second term is 4 and the sum is 18, we have the system of equations

ar = 4,

Multiplying the second equation by 1 - r gives us a = 18 - 18r. From the first equation, we have a - ⅛fr.
We equate these two expressions for a to give 18-18r = 4. Rearranging this equation gives 9r2-9r+2 = 0,
from which we find the solutions r = 1/3 and r = 2/3. If r = 1/3, we have a = 12, so the series is

4 4
12÷4÷3÷9+"

If r — 2/3, then a = ⅛∣r = 6, and our series is

z, λ 8 16
6 + 4+3 + V + '"
We can use our formula for the sum of an infinite geometric series to quickly confirm that each of these
sums equals 18, as required. □

Problem 10.26: If a + ar + ar2 + ar3 + ∙ ∙ ∙ = 15 and a + ar2 + ar4 + ar6 + ∙ ∙ ∙ = 9, then what is the sum of
the geometric series a + ari + ar6 + ar9 + ■ ■ ■?

Solution for Problem 10.26: Both of the given series are infinite geometric series, one with common ratio
r and the other with common ratio r2. Applying our formula for infinite geometric series to both series
gives us

There are a variety of ways to solve this system of equations. Perhaps the most straightforward is to
notice that we can factor the denominator of the left side in the second equation to give us
a
= 9.
(1 + r)(l - r)

We already know that a∕(l -r) = 15, so we have

a a
(1 + r)(l -r) 1-r 1 + r

Solving 9 = 15∕(1 + r) gives r = 2/3. From a∕(l — r) = 15, we find that a = 15(1 - r) = 5, so we have

310
 10.4. GEOMETRIC SERIES

Problem 10.27: If P is the product of π numbers in geometric progression, S their sum, and S' the
sum of their reciprocals, then find P in terms of S, S', and r∕. (Source: AHSME)

Solution for Problem 10.27: At first glance, it seems unlikely that there's a simple relationship between P,
S, S', and π. To search for one, we take our usual approach for working with geometric sequences: we
let the first term be a and the common ratio be r. Then, we write the quantities in the problem in terms
of a, r, and n. First, we take care of the product of the n terms:

P = a ■ ar ■ ar2 ■■ ■ arn~2 ∙ arn~1.

Since there are n terms, our product equals an times all the r terms:

P = a ∙ ar ■ ar2 ■ ∙ ∙ arn~2 ■ artl~^ = flπp+2+3+∙∙∙+(h-i) — fl"pj(π-i)∕2ι

We also know how to write S in closed form:

a(rn -1)
S = a + ar+ ar2 + —f arn 1
r-1

Writing S' in closed form is a little trickier. We start with factoring:

The series on the right is a geometric series with first term 1 and common ratio ∣, so we have

U√1≡-ι))
∕ a 1 -1
r 7

That's not too pretty, but after a little algebra, we can get the expression in a much nicer form:
∕ . ∕∕-1 ∖n
√1((1)"-1)) 1 fJ--l rtl∖ - 1f rL∑l
rn -l 1
) rn (r-l)'
αrπ-1 -1
S' = -
α v l-1 ) a 1 - 1
J r
r” ~ a∖rn - r"^1
i

Next, we write our expressions for P, S, and S' all together, hoping we see some way to relate them:
p _ anr∏(n-l)∕2
_ a(rf -1)
r-1 '
rn -1
S' =__ -___ -__
arn~fr -1)

Our expression for P doesn't have r'1 - 1 or r - 1, so we have to combine S and S' in a way that cancels
these. Fortunately, we see that dividing S by S' (or vice versa) eliminates both r” - 1 and r-1:

S ⅛r a(rn - 1) arn~1(r - 1) 2 1 r" -1 r-1 2 π.1

— = ---------- = ------------ - ---------------- = err - -------- - -------- = ar
s' -⅞∙1∙-n r~1 r"~1 r-1 rn-l

311
CHAPTER 10. SEQUENCES AND SERIES

Aha! We have S∕S' = a2rn 1 and P = ar,rtl^n 1^2. To get from a2rn 1 to anrn^n 1^2, we only have to raise
the former to the h/2 power. So, we have
n/2

= (fl2r"-1)"z2 = fl"r"(π^1)∕2 _ p
We now have the desired (and surprising) relationship: P = (S∕S')n^2.

Seeing such a simple relationship, we check to see if we could have derived it by manipulating the
series for S and S'.

Concept: If you derive a simple algebraic relationship after complex algebraic ma-
Q^ mJ nipulation, use the simple relationship as a guide to look for a simpler
proof.

Now that we know what to look for, we start with S∕S':

s a + ar + ar2 + ■■ ■ +arn 1
S' a + 1ar++1ar+
^
X + .. - + ⅛
a(l +r + r2 + ∙∙ • + rπ-1)
j(ι÷l÷⅛÷∙ ∙→ ⅛)

- Λ7z9∙ . 1 + r + r2 + ∙ • ■ + rn~1
rn-l+rn-2+,.,+r2+r+∣
rn-l

= a2rn~∖

Aha! Comparing this to P = attrn^n~1^2, we have P = (a2rn~1)n^2 = (S∣S')n∣2r as before. (Notice also that
S∕S' = a ■ arn~1, which is the product of the first and last terms of the sequence. This also is a clue that
there is a simple relationship between S∕S' and P.) □

Γxercises

10.4.1 Find the sum of each of the following series:

(a) 2 + 8 + 32 + ∙∙∙ + 2048 *
(c) 1 + 2 + 3 + 6 + 9 + 18 + ■ ∙ ∙ + 729 + 1458
(b) 54-18 + 6-2 + ∙→2∣3

10.4.2 Sum each of the following infinite geometric series:

<a> ⅛ + ⅛ + ⅛+∙∙ <c) 2-√2 + l-^ + l-...

(b) 192+ 144+ 108 +∙∙∙

12 12
10.4.3 Find the sum -
7
+ rz∙
7z
+ -7
73
+ -τ
74
+ -∙∙. (Source: AHSME)

10.4.4 Find the value of x that satisfies the equation 1 + x + x2 + ∙ ∙ ∙ = 4. (Source: Mandelbrot)

312
 10.5. SEQUENCE, SUMMATION, AND PRODUCT NOTATION

(fl8 + a4 + 1 )(fl3 + a2 + a + 1)
10.4.5 Find all real values of π such that = 21.
λ9 + β6 + λ3 ÷ 1
10.4.6 The first term of a finite geometric series of real numbers is positive, and the sum of the series is
negative. Show that the series has an even number of terms.
10.4.7 Note the following relationship between sums of powers of 5 and other powers of 5:

51 + 50 = 6 6 ■ 100 + 25 = 625 = 54
52 + 51 +50 = 31 31 100 + 25= 3125 = 55
53 + 52 + 51 + 50 = 156 156 ∙ 100 + 25 = 15625 = 56
Explain the pattern and why it works.

10.5 Sequence, Summation, and Product Notation

We have special notations that allow us to write sequences, series, and products more compactly. We
introduce these notations with a few examples. We start with sequences:

∣3n + ⅛1 = 8,11,14,17,20
{2fc + l}∞0 = 2,3,5,9,17,...

In the first row, we write the sequence 8,11,14,17,20 by enclosing the formula 3n + 5 in curly braces.
The variable in our formula is a dummy variable, also called the index, that takes on the range of values
indicated by the indices outside the braces to the right. The lower index represents the lower bound
(starting point) for the dummy variable. The upper index represents the upper bound (endpoint) for the
dummy variable. Note that these indices are not necessarily the same as the least and greatest numbers
in the sequence.
The second sequence is an infinite sequence, so we denote the upper bound of the dummy variable
with the symbol for infinity: ∞.
Sometimes a sequence will be defined without indices. We should only write a sequence with braces
without indices when it's clear what the dummy variable is and at what value of the dummy variable
we should start the sequence. For example, if we say that Fo = 0, Fι = 1, and Fn = Fπ-ι + F∏-2 for all
n ≥ 2, we can then comfortably refer to the sequence {Fπ}, where it is implied that this refers to the
infinite sequence Fo, F1,F2,....
To express some series, we use the uppercase Greek letter Σ, which is called sigma. For example, we
have
10
y 2n = 2 + 4 + 6 + 8 + 10 + 12 + 14 + 16 + 18 + 20
n=l

We call this summation notation. Each Σ symbol tells us to sum the sequence whose general formula
immediately follows the sigma. Below each sigma is the starting point for the dummy variable in the

313
CHAPTER 10. SEQUENCES AND SERIES

sum, and above each sigma is the dummy variable's upper bound. Note that the first sum is a finite
arithmetic series and the second is an infinite geometric series.
Finally, we use the uppercase Greek letter ∏ for products in much the same way we use Σ for sums:
7
^J(k + 1) = 3 ∙ 4 ∙ 5 ∙ 6 ∙ 7 ∙ 8
k=2
∞
1 2 3
2’3’4

You might notice the resemblance of ∏ to n. This isn't a coincidence; ∏ is the uppercase version of the
Greek letter pi, while π is the lowercase version. Just as with summation notation, the index below the pi
represents the starting point of the dummy variable in the product while the number on top represents
the endpoint.
Series are extremely important in mathematics and most sciences (including, if not especially, com­
puter science). You'll encounter Σ frequently, so it's important to understand what it means and how to
manipulate expressions involving it.

Problems

Problem 10.28: Rewrite each of the following sequences, series, and products by expanding them
term by term. You do not need to calculate the sums and products, but feel free to do so if you like.
6 ∞ 10
(a) {7fc + 1⅛ (c) ∑ n3 (e) y, xn (g) ∏
n=l n=0 ι=l
8 ∞ λ i
(b) {2"-⅛2 (d) ∑(2t-l)
k=l
ω Σ- ι=3
(h) ∏
n=l

Problem 10.29: Write each of the following using sequence, summation, or product notation. Assume
that each sequence or series in parts (a) through (f) is arithmetic or geometric in nature.

(a) 4 → 11z —12' 1

i*z _2 4' —18 (e) 2-6 + 18-54 + ∙∙∙ + 118098
(b) ...,-10,-7,-4,-l,2,5,... (f) 3 + 0.9 + 0.27 + 0.081 + ∙ ∙ ∙
12 3 8
(c) 1 + 3 + 5 + ∙∙∙ + 101 (g) 3 ’ 4 5 ’‘‘ 10
22-l 32-l 42-l 182-1
(d) 83 + 79 + 75 + ∙ ∙ ∙ + (-105) + (-109) (h) 22+l 32+l 42+l 182+1

Problem 10.30: In each of the following parts, state whether or not the two sums are the same, and
explain your answer.
21 16 15 17
(a) and y (3k + 15) (c) T(-3)' and y(-3)7+2
∕=-2
∕=2 k=-3 y=θ
10 1 11 1 00 ∞
Σ
(b) and y -—- (d) £(0.5)' and £>)'-’
k=2 K
∕=0 ∕=1

314 ■
 10.5. SEQUENCE, SUMMATION, AND PRODUCT NOTATION

Problem 10.31:
12
2i
(a) Express the sum
Σ i-3 as
i-Λ
a summation in which the dummy variable ranges from 7 to 15.

00

(b) Express the sum θ∙l7 as a summation in which the dummy variable starts at -5.
∕=o
100
Problem 10.32: Evaluate the summation J~t(5∕ + 7).
∕=ι

10 10
Problem 10.33: Suppose fl], a2,..., «10 and bχ, b2,..., b^ are numbers such that a∣i - 6 and J~
* ι ⅛ = 13.

10
k=ι ⅛=ι
Find J2∕3⅜ - 2b⅛ + 1).
k=l

Problem 10.34:
k
(a) Find a closed form expression for 2n. In other words, find a formula in terms of k that equals
n=l
k
J-1,2n for all integers k.
n=l

OQ
l + 4^
∑
k=0
5k
D

(a) Write the first five terms of the sum without adding the terms in each numerator.
(b) Separate
. the sum into two sums that you can evaluate, and then compute these sums.
00
1+f
(c) Compute
5k

n
Problem 10.36: Find the polynomial f(x) such that f(k) = n2 -3n for all positive integers n.
k=l

315
CHAPTER 10. SEQUENCES AND SERIES

We begin by getting a little practice interpreting these new notations.

Problem 10.28: Rewrite each of the following sequences, series, and products by expanding them
term by term. (You do not need to calculate the sums and products.)
6 ∞ 10
(a) {7k + 12⅛,1 (c) y n3 (e) f+" (g) ∏.
n=l n=0 i=l
8 ∞ ~ i

(b) (2"-∏∣L2 (d) ∑(2fc-l) ∑H) <h> ∏J⅛

k=l z=3 n=l

Solution for Problem 10.28:

(a) To produce the terms in the sequence {7k + 12}®=1, we evaluate 7k + 12 for each value of k from
k = 1 to k = 8. This gives us 19,26,33,40,47,54,61,68. Notice that this is an arithmetic sequence.
(b) To produce the terms in the sequence {2n - n}6n=2, we evaluate 2n -n for each value of n from n = 2
to n = 6. This gives us 2,5,12,27,58.

(c) ∑n3 = l3 + 23 + 33 + 43 + 53 + 63.

n=l
8
(d) J~,(2⅛ -1) = 1 + 3 + 5 + 7 + 9 + 11+ 13+ 15. This is an arithmetic series.
fc=ι
∞
(e) J~lj xn = 1 + x + x2 + x3 + ∙ ∙ ∙. We might be tempted to simplify this by noting that it is an infinite
n=0
geometric series with first term 1 and common ratio x. However, we don't know whether or not
∣x∣ < 1, so we don't know if the sum converges to 1∕(1 - x) or not. So, we must leave it as the sum
shown.
(f) We have

This is an infinite geometric series with first term (-2∕3)3 = -8/27 and common ratio -2/3, so its
sumis
27 _ 27 _ Q
ι-H)" I =’45’

10
(g) p = l∙2∙3∙4∙5∙6∙7∙8∙9∙10 = 10!.
i=l

(h) π
n=l
12345678 9 _ 1
2'3'4 ’ 5 ’ 6 ’ 7 ’ 8 ’ 9 ’ 10 " 10'

316
 10.5. SEQUENCE, SUMMATION, AND PRODUCT NOTATION

Now, let's get a little practice writing sequences, series, and products using our new notation.

Problem 10.29: Write each of the following using sequence, summation, or product notation.
Assume that each sequence or series in parts (a) through (f) is arithmetic or geometric in nature.
(a) 4-21-11-1
→ 1∕ 2'4' 8 (e) 2-6+ 18-54 +∙∙∙ +118098
(b) ...,-10,-7,-4,-l,2,5,... (f) 3 + 0.9 + 0.27 + 0.081 + ∙∙∙
(c) 1 + 3 + 5 + ∙ ∙ ■ + 101 12 3 8
(g) 3 ' 4 ' 5 10
(d) 83 + 79 + 75 + ∙ ∙ ∙ + (-105) + (-109) 22-l 32-l 42-l 182-1
(h) 22+l 32+l 42+l 182+1

Solution for Problem 10.29: If your answers don't match the answers we come up with below, don't
panic! Your answers may still be correct, as we'll discuss shortly.

(a) The first term is 4 and the common ratio is -1 /2, so the general formula for each term is 4(—1 ∕2)"-1.
There are six terms, so we write the sequence as ∣4 (—⅜ j ∣

(b) There is no beginning or end to this arithmetic sequence. We do know that the common difference
is 3. From this we could create any number of general formulas. We choose 3k + 2 here, but we
could just as easily choose 3k -l,3k + 11, or any other that produces all the numbers that are two
more than a multiple of 3. Using 3k + 2 as our formula, we can write the sequence as {3k + 2}∞ oo.
(c) What's wrong with this solution:
Bogus Solution: This series consists of the first 51 positive odd numbers. We can
write odd numbers as 2n + 1, and we want the first 51 positive odd
51
numbers, so we can write the series as ∑J2n + 1).
n=l
51
If we write out the sum represented by J~*
*j(2n + 1), the first term is 2(1) +1 = 3, and the last term is
n=l
2(51) + 1 = 103. So, the resulting series is3 + 5 + 7 + ∙∙∙ + 103, not 1 + 3 + 5 + ∙ ∙ ∙ + 101. Our starting
and ending values of the dummy variable are 1 too large! We can fix this by subtracting 1 from
50
each, so the sum can be written as y>+1).
n=0

WARNING!! Pay close attention to the starting and ending values of the dummy
*
*a variable when writing a series in summation notation. The easiest
way to make sure you have the correct values is to simply substitute
them into the formula in your summation. The starting value should
produce the first term of the sum, and the ending value should pro­
duce the last term.
We also could have noted that the common difference of this arithmetic series is 2, and a general
formula for each term is 1 + 2(rz -1) = 2n — 1. Since 2n -1 = 101 when n = 51, our series consists of
51
the sum of 2n -1 from n = 1 to n = 51, which we write as Σ<2" -1). While this summation does not
H=l

317
CHAPTER 10. SEQUENCES AND SERIES

50
look the same as ^(2π +1), it is indeed the same series. Both represent the sum 1 + 3 + 5 4---- ÷∙101.
n=0
(d) The common difference of this arithmetic series is -4, and a general formula for each term is
83 - 4(n - 1) = 87 - 4n. Looking at the last term, we see that 87 - 4n = -109 when n = 49, so our
49
series consists of the sum of 87 - 4n from κ = 1 to n = 49. We can write this sum as ∑(87 - 4n).
n=l

(e) The common ratio of this geometric series is -3, and a general formula for each term is 2 ∙ (-3)f,-1.
Since 2 ∙ (-3)"^1 = 118098 = 2 ∙ 310 when n = 11, the series is the sum of 2 ∙ (-3)"^1 as n goes from 1
11
to 11. We can write this sum as ∑l(2 ∙ (-3)π"1).
n=l
(f) The common ratio of this infinite geometric series is 0.3, and a general formula for each term is
∞
* ι 3(0.3)"-1.
3(0.3)π-1. We write the sum as J~
∏=ι
(g) As we did with the other parts, we look for a formula that generates each term. The denominator
is two more than the numerator in each fraction, so each term can be written as n∕(n + 2). As we
let n go from n = lton = 8in n∕(n + 2), we produce all the terms in the product, so the product
8
can be written as I ----- -.
11 ∏ + 2
n=l

(h) One formula that generates the terms in our product is pretty clearly (n2 — l)∕(n2 + 1). However,
we must be careful to note that our product does not start with n = 1. It starts instead with n = 2

and goes through n - 18, so we write it as J

18
n2z -1ι
n=2
n2 + 1'

□
You might have different answers from those above, but your answers might still be correct. Here
are a couple of reasons why you might have different answers:

49
• You used a different variable as your dummy variable. For example, the sum ^(87 - 4κ) is the
n=l
49
same as the sum £(87 - 4∕).
;=i

• You used a different formula that produces the same sequence, sum, or product. For example,
instead of using 2h - 1 to express each odd number in the sum 1+3 + 5 + 74------ F 101, you might
use 2n + 1. However, if you do so, you must be careful about the bounds of the dummy variable.
If we use In + 1 to represent each term in the sum 1 + 3 + 5 + ∙ ∙ ∙ + 101, then n goes from n = 0 to
50 51 49
n = 50, so our sum is Σ<2" +1). Notice that this is the same sum as ^(2π -1), and as (2n + 3),
n=0 n=l n=-l

318
 10.5. SEQUENCE, SUMMATION, AND PRODUCT NOTATION

53
so on.

Let's take a closer look at this latter point.

Problem 10.30: In each of the following parts, state whether or not the two sums are the same, and
explain your answer.
16 15 17
T, (3⅛
k=-3
+ 15) (C) Σ (-3∑ and y,(-3),+2
j=-2 j=0
∞ ∞
1
Σ
1—d
k-2
(d) ∑(0.5)'
i=0
and 2j0,5Γ1
∕=1

Solution for Problem 10.30:

(a) One way to see that two summations represent the same series is to write both summations as
series without summation notation. We have
21
y, 3z
i=2
= 6 + 9 + 12 + 15 + ∙ ∙ ∙ + 63,

16

k=-3
y (3k + 15) = 6 + 9 + 12 + 15 + ∙ ∙ ∙ + 63.

Yep, these are the same series. Now, let's think about how we might have seen this without writing
out the series. Let
21 16
A = y 3z and B = y (3k + 15).
i=2 k=-3

We first note that both series A and series B have 20 terms. We could also see that the two series
have the same number of terms by noting that each limit of A is 5 more than the corresponding
limit of B.
Since the limits of A are 5 more than the limits of B, the dummy variable in each term of A is 5
more than the dummy variable in the corresponding term in B. This means that for each value of
z in A and the corresponding value of k in B, we have i = k + 5. In A, we add 3z for each value of
z. Letting i = k + 5, adding 3z for each value of z from 2 to 21 is the same as adding 3(fc + 5), which
equals 3k + 15, for each value of k from -3 to 16. (We get the range -3 to 16 for k by noting that
k = i-5, and z ranges from 2 to 21.) Therefore, series A and series B are the same.
(b) The first series has 10-2 + 1 = 9 terms and the second has 11-4 + 1 = 8 terms, so these two series
are not the same. (Their sums may or may not be equal, but the series are not the same.)
(c) Both series have 18 terms. To make the two series easier to discuss, we assign a variable to each,
and change the dummy variable in the second series to k. (Changing the dummy variable doesn't

319
CHAPTER 10. SEQUENCES AND SERIES

change the sum.) So, we write

15 17
C=∑(-3)' and D = ∑(-3)k+2.
j=-2 k=0

What's wrong here:

Bogus Solution: Each k is 2 more than the corresponding ∕, so we let j = k + 2 in
the term (-3)^ of C to find the corresponding term (-3)fc+2 of D.
Therefore, the two series are the same.
Be careful! Because eachk is 2 more than the corresponding j, we let j = k-2, not j = k + 2. Letting
; = k - 2 in the expression we sum in C gives us (—3)fc-2, which is not the expression we sum in D.
Specifically, we have
15 17
£(-3)' = £(-3)‘-2.
∕=-2 k=0

We also could have caught our mistake by writing out a few terms of each summation:

15
∑(-3)' = (-3)^2 + (-3)-, + (-3)° + ■■■ + (-3)15,
∕=-2

∑(-3)fc+2 = (-3)2 + (-3)3 + (-3)4 + ∙ ∙ ∙ + (-3)19.

k=0

These are clearly not the same.

Concept: When in doubt with summation notation, write out the series without
Q -⊂√j summation notation.

(d) At first, we might think that the two series are different because the second series appears to have
one fewer term than the first. But, if we write both series without summation notation, we have:
∞
∑(0,5)z = (O.5)o + (0.5)1 + (0.5)2 + (0.5)3 + ∙ ∙ ∙ ,
z=0
∞
∑(0∙5),'^1 = (O.5)o + (0.5)1 + (0.5)2 + (0.5)3 + ∙ ∙ ∙ .
i=ι
These are the same series. Because the series are infinite, we can't count the number of terms. In
other words, even though the dummy variable starts at a higher value in the second summation,
we can't say the second series has "more" terms than the first.

Extra! Here's a surprising summation:

320
 10.5. SEQUENCE, SUMMATION, AND PRODUCT NOTATION

Problem 10.31:
12

(a) Express the sum y - —- as a summation in which the dummy variable ranges from 7 to 15.
∕=4 '
∞
(b) Express the sum y 0.V as a summation in which the dummy variable starts at -5.
∕=o

Solution for Problem 10.31:

(a) Let the dummy variable of our new summation be k. We wish to have k range from 7 to 15, so we
already have the limits on our summation. We need only to find the expression that we sum in
the new series. Each z in the old series is 3 less than the corresponding k in the new series, so to
find our new expression, we let z = k - 3 in ⅛j. This gives us = So our new summation
k-6
15
2k-6
i5∑= ^k∑Γζ^ As a check, we can write out a few terms in each summation:
k=7

2z _ 8 10 12 24
∑z-3~l + 2 + 3 + ’ ’- + 9'
ι=4

v 2⅛-6 8 10 12 24
1+ τ+ T + '" + ^9^∙
k=7
Indeed, these two series are the same.
(b) Let the dummy variable of our new summation again be k. Our new summation starts with
the dummy variable 5 less than the dummy variable in our old summation. So, for each k and
corresponding 7, we have k = 7 - 5, or 7 = k + 5. This means that the expression we sum in the new
summation is 0.1
* r+5. But what are the limits of the new summation? We know the new summation
starts at k = —5, but where does it end?
The new summation must still be an infinite series, so the upper limit is still to:
∞ 00

0.1/ = y 0.1fc+5.
7=0 k=-5

When we write a few terms of each series, we see that each equals (0.1)0 + (0.1)2 + (0.1)3 + ∙ ∙ ∙.

12 15

When we write y -—- as the equivalent summation Σ- — we say we shift the index of
j=4 k=7
the original summation. Shifting the index of a summation does not produce a new series; it merely
produces a different way to represent the series. As we saw in our second example, when we shift the
index of an infinite series, only the starting limit changes.
Now that we're comfortable with summation notation, let's try evaluating a sum expressed in
summation notation.

321
CHAPTER 10. SEQUENCES AND SERIES

100
Problem 10.32: Evaluate the summation ∑(5∕ + 7).
7=1
L

Solution for Problem 10.32: Solution 1: We write out the first few terms of the series. We see that increasing
j by 1 at each step increases 5; + 7 by 5. In other words, our sum is an arithmetic series with common
difference 5. We get the first term by letting ; = 1 in 5/ ÷ 7 and we get the last term by letting j = 100 in
5j + 7. This gives us the arithmetic series
ιoo
∑<5∕ + 7) = 12 + 17 + 22 + ∙ ∙ ∙ + 507.
7=ι

ιθθ i2 ∣ 5Q7 519

We know that there are 100 terms, so our sum is J~lj(5∕ + 7) = 100 ------- ------= 100 ∙ —= 25950.
7=ι

Solution 2: Writing each term in the form of the general term, we see that we can group and factor in
a way that makes each group easy to evaluate:
100
∑(5∕ + 7) = (5 ∙ 1 + 7) + (5 ∙ 2 + 7) + (5 ∙ 3 + 7) + ∙ ∙ ∙ + (5 ∙ 99 + 7) + (5 ∙ 100 + 7)
7=ι
= (5∙l+5∙2 + 5∙3 + ∙∙∙ + 5∙99 + 5∙ 100) + (7 + 7 + 7 + ∙∙∙ + 7 + 7)

100 7's
= 5(1 + 2 + 3 + ∙ ∙∙ + 99 + 100) + (7 + 7 + 7 + ∙∙∙ + 7 + 7)

100 7's
100 100
= 5∑7÷∑
*
7=1 7=1

We could have evaluated the expression

5(1 + 2 + 3 + ∙ ∙ ∙ + 99 + 100) + (7 + 7 + 7 + ∙∙∙ + 7 + 7)

100 7's
as 5(100 ∙ 101/2) + 7 ∙ 100 = 25950, but instead we rewrote it in terms of two other summations above to
make a point. We've shown how to break the original sum into simpler sums, and in the future we can
jump straight to this sum of simpler sums in order to calculate similar sums more easily. Specifically,
we break the initial summation into twosums, then factor the 5 out of the first sum tofind:
100 100 100 100 100 ∩∩ , 1 ∖
∑(5∕ + 7) = ∑(5∕) + ∑ι7 = 5∑7 + ∑l 7 = 5 ∙ - 2 +1 + 100 ∙ 7 = 5 ∙ 5050 + 700 = 25950.
7=1 7=1 7=1
∕=ι 7=1 7=1

While the second solution may not look simpler at first, these sorts of summation manipulations will
become more natural as you gain experience working with summation notation, and they become more
important as summation problems get harder. □
Our second solution is an example of the following two manipulations of summation notation:

322
 10.5. SEQUENCE, SUMMATION, AND PRODUCT NOTATION

Important: For any real numbers a↑, a2,..., an, b↑,b2,..., bn, and constant c,
ψ ιι n n
Jj⅜ + bk) = y, ak + yii bk,

k=l fc=l ⅛=1

n
∑c⅜ =
c∑⅛∙
fe=l fc=l

Here's a little more practice using these facts.

Solution for Problem 10.33: We separate the sum into parts we can evaluate, using the given summation
values where possible:
10 10 10 10
J"*, (3ak ~ 2bfc + 1) = y 3ak + y (-2bk) + y 1
k=l k=l k=l k=l
10 10
=y ak y bk+10
k=l fc=l
= 3 ∙ 6 - 2 ∙ 13 + 10 = 18 - 26 + 10 = 2.
□

Solution for Problem 10.34: We separate the summation into two sums, then sum the results to get a
closed form:

yι(2n + 3) = yι(2n) + y3 = 2yn + 3k = 2∙^^+3fc = k2 + 4k.

n=l n=l n=l n=l

Notice that our final answer gives us a nice, simple closed form in terms of k for the summation. Closed
forms are usually much easier to deal with than summations. For example, which would you rather
have to compute for some value of k,
k
Σ<2" + 3) or k2 + 4k?
n=l
□
∞ 1 +4k
Problem 10.35: Compute J —-—.
k=0 5

323
CHAPTER 10. SEQUENCES AND SERIES

Solution for Problem 10.35: Let's start by taking a look at the first few terms of the sum:
∞
1 +4fc 1+1 1+4 1 + 16 1 + 64 1 + 256
Σ
∕c=0
5k 1 + 5 + 25 + 125 + 625

1 1 1 4 1 16 1 64
1 + 1 + 5 + 5 + 25 + 25 + 125 + 125

Starting with the first term and taking every other term, we see that we have the infinite geometric series
l + 5 + ⅛ + i⅛ + ∙∙∙∙ The remaining fractions also form an infinite geometric series, with first term 1
and common ratio 4/5. We could have seen this directly by separating the initial summation into a pair
of summations:

Problem 10.36: Find the polynomial ∕(x) such that ∕(^) = ∏2 -3n for all positive integers n.
k=l

Solution for Problem 10.36: Letting n = x in our summation gives us

∕(1) + ∕(2) + ∕(3) h---- + ∕(x - 1) + ∕(x) = x2 - 3x.

If only we could get rid of the first x — 1 terms on the left.

We can! We use our summation again, this time letting n = x-l. This gives us

∕(1) + ∕(2) + ∕(3) + ∙ ∙ ∙ + ∕(x - 1) = (x - 1)2 - 3(x - 1).

Subtracting this from our n = x equation eliminates every term on the left except ∕(x), and we have

∕(x) = x2 - 3x - (x - 1)2 + 3(x - 1) = 2x - 4.

Concept: If we have a summation with a variable among its bounds, we can of­
O≡≡^ ten learn more about the summation by choosing different values for the
variable.

Exerci⅜e⅜

10.5.1 Rewrite each of the following arithmetic and geometric sequences using sequence notation:
(a) -91,-81,-71,..., 1999 (c) 15,11,7,3,-1,...
1 1 32
(b) -,-,2,8,32,.. .,8192 (d) 54,36,24,16, y,...

324
 10.6. NESTED SUMS AND PRODUCTS

10.5.2 Write the following using summation notation:

1 2 3 n
(a) 3+3+3+3+3+3+3 (d) 2 + 3 + 4 n+1
1,3 5 x x 2 x3
(b) -÷l÷-+2÷-+3 (e) 1! + 2! + 3!+-

(c) n + (n - 1) + (n - 2) + ∙ ∙ ∙ + 1 (f) l-x + x2-x3 + ∙∙∙

10.5.3 Express each of the following as a summation in which the dummy variable starts at 5.

v÷ 2' ∞ ι
(a) ∑TT5 «» ∑⅛
10.5.4 Compute or express in closed form each of the following sums:
15 6
(a) J2<3' “ 4>
2=1
(≈) Σ4-⅛
n=l
(e) X<3" - 3«)
H=l

n n 100
(b) ^(2 - 0
z=l
(d) Σ∕ (f)
* ∑(-υ⅛2
fc=0 k=l

99 99 99
10.5.5 If y~
* ι Xk = 7 and J~
* ι y⅛ = 17, find the value of y^(107xn - 74yn).
k=l k=l n=l
oo
_ 1 ∖r 3k + 6k
ι
10.5.6 Evaluate the sum y ———
k=0
10.5.7
* Express, as concisely as possible, the value of the product

(03 - 350)(l3 - 349)(23 - 348)(33 - 347) ∙ ∙ ∙ (3493 - l)(3503 - 0).

(Source: HMMT) Hints: 18

10.6 Nested Sums and Products

10
We know how to tackle a summation of a simple expression, like J~l(2z). We simply let i equal each of 1
i=l
through 10 in the expression 2z, then sum the results. But what if we must find the summation not of a
simple expression, but of another summation?

Extra! We should be taught not to wait for inspiration to start a thing. Action always generates
∣< <∣∣Φ. ∣ ⅜ ∣ ∣> inspiration. Inspiration seldom generates action.
- Frank Tibolt

325
CHAPTER 10. SEQUENCES AND SERIES

Problem 10.37: Evaluate each of the following:

4 6 λ 4 ( 6 ∖

<”) ΣΣ'
i=ι U=ι 7
(c) ∑ ∑(' + ∕)
r=l ∖∕=1 7

Problem 10.38:
( 5 ∖
f 5 '
(a) Evaluate l∑√ll∑4
v=ι 7∖y=ι 7
5(5 λ
(b) Evaluate Σ Σ i' ∙
>=l b⅛l J

(c) Explain the relationship between your answers.

∞ ' ∞ ''
(d) Evaluate 7
Z—< Z_j7 — J .
∕=ι G=ι
2,+∕

Problem 10.39:
n ∞ j
Find a closed form for ∑(1). (c) Find a closed form for 1√ ’
(3)
(a)

10 ( i y
;=i f ∞ . } CO
j=i

(b) Evaluate ∑∑ω

≈=ι b=ι 7
(d) Evaluate ∑∑∣
i∙=o v j=i 7

Problem 10.40: Bill thinks of a number that equals 5 raised to a positive integer power. Betty thinks
of a number that equals 5, raised to a positive integer power. Bob then multiplies Bill's number and
Betty's number. Find the sum of all the possible products Bob can form.

We call the summation of a summation, such as

a nested sum. We will often leave out the parentheses around the inner sum and simply write such a
nested summation as
10 5
∑ ∑(2>7)∙
∕=1 ∕=1

326
 10.6. NESTED SUMS AND PRODUCTS

Solution for Problem 10.37: We begin by breaking the expression being summed (and summed again)
into pieces:
4 6 4 6

ΣΣ<'∙÷7)=ΣΣ'∙+Σ∕ =ΣΣ-∙÷ΣΣ∕∙
1=1 7=1 ι=l ∖∕=1 7=1 J i=l j=l f=l y=l
Now we evaluate the inner sum for each of our pieces:

ΣΣ>∙=ΣΣ'∙ =Σ⅛
i=l 7∙=1 j∙=ι (j=i J ι=l

4 6 4 ( λ 4

ΣΣ7 = Σ Σ7 =∑21∙
1=1 7∙=1 i=ι (j=l J z=ι
Next, we evaluate each of these summations:

i=l 7=1 z=l i=l

4 6 4
∑∑∕ = ∑21=4(21) = 84.
ι=l 7=1 i=ι

Finally, we add the pieces to evaluate the whole sum:

4 6 4 6 4 6

i=l 7=1 z=l 7=1 z=l 7=1

Problem 10.38:
5 5
(a) Evaluate Σ Σ ,,
i=l 7=1

Solution for Problem 10.38:

5
(a) We focus first on the inner sum, ij. In evaluating this sum, we treat i as constant, and let j
7=1
5
range from 1 to 5. Therefore, we can factor i out of the sum
Σ>
7=1
5 5

∑,7"=i∑7∙
∕=1 7=1

327
CHAPTER 10. SEQUENCES AND SERIES

This is true for any value of z, so we can write our nested sum as
5 5 5
∑∑'7 = ∑['∑∕
ι=l y=ι 7=1
■ i=ι k 7
Now, the inner sum doesn't depend on the variable of the outer sum, so we evaluate the inner
sum:
5f 5 1
∑-∑M∑ z(l + 2 + 3 + 4 + 5).
5

'=ι I 7=ι J
i=l
We can now factor the constant (1+ 2 + 3 + 4 +5) out of the remaining sum, and then evaluate
what remains:
5 5
* i z(l + 2 + 3 + 4 + 5) = (1 + 2 + 3 + 4 + 5) Γ i = (1 + 2 + 3 + 4 + 5)(1 + 2 + 3 + 4 + 5),
y
ι=l z=l

which is 152 = 225.

(b) With our first part as inspiration, we try to separate the nested sum into two summations. First,
we must separate the variables in the inner expression:
(∞
∞ (∞ ∖ ∞ ∖

Σ Σ⅛ =Σ*-f ∑i∙1
∕=1 b=ι 7 l=l ∖7=1
^2l 2>
27
7

Now, we can factor 1/2’ out of the inner sum:

∞ f

”
OQ

ι'
ΣΣ∑ Z—∣ 2—∣ 1~ 7 = ∑1 -Σ~ 7 ■
= Σ i∑
∖ ~ ∕
∞
1 °o 1
r oo

l=l G=l J
2' 2) 2, * i 2J
'=1 K 7=1 "=1 ∖J=1 7
∞
Since the sum of the infinite geometric series ~^^ *s i∑1(i∕2) = 1/ wθ have
7=1 21
M≡

∞ f ∞ 1 1 ( ∞ 1 A
÷-i

o° 1

Σ- ∑i
∣

Σ Z-ι = ∑j = l∙
I
I

z-∣ 27
m

2'+7
i=l 7=1 J z=l 7=1 J t=l

□
We found in Problem 10.38 that each nested sum was the product of two simpler sums. For example,
in the first part, we found
5 ( 5 λ
∑∑v=∑∣>∑4

i=l ∕=1 1∙=ι 7=ι ;

So, our outer summation now consists of summing the product of i and another summation that does not
depend on i. So, we can factor this other summation out of the outer summation, and write the expression
as the product of two summations, instead of as a nested sum:
5 5 5 ( 5 λ ( 5 √ 5 }

Σ∑-7 = ∑-∑∕
ι=l ∕=1 V 7 = v'=ιΣ-7∖Σ∕.
j=ι∕=1 7=1 7

328
 10.6. NESTED SUMS AND PRODUCTS

Similarly, in the second part, we saw that

Concept: When evaluating the inner sum of a nested sum, the outer variable can be
O u∣J treated as a constant. It (or any function of it) can sometimes be factored
out of the inner sum to separate a nested sum into the product of two
simple summations.

Now that we know how to deal with the nested summation of an expression involving the dummy
variables of both sums, let's take a look at what happens if the dummy variable of the outer summation
is part of the limits of the inner summation.

Solution for Problem 10.39:

(a) As usual, we work from the inside out, focusing first on } (1). We can forget for the moment that
∕=ι
i is itself the dummy variable of a summation and just treat it as a variable, so we can write this
summation in closed form in terms of i. The sum is the sum of i copies of 1, so we have

∑(1) = '∙
∕=ι

10 i 1
Therefore, we have ∑∑ω=∑ i = 1 + 2 + 3 + ∙ ∙∙ + 10= = 55.
!=1 2
i=ι 0=1 ∕
∞ ∣
(b) Once again, we work from the inside out, seeking a closed form for Σ -τ. This is an infinite
3
)=t
geometric series with first term 1/3' and ratio 1/3. Therefore, the series equals
1
3'
3'"

This means that

329
CHAPTER 10. SEQUENCES AND SERIES

This is just another geometric series! The first term is 1 and the common ratio is 1/3, so we have

□
Concept: Evaluating nested summations sometimes requires writing the inner sum-
Q " ljJ mation in closed form in terms of the dummy variable of the outer sum-
mation.

Problem 10.40: Bill thinks of a number that equals ∣ raised to a positive integer power. Betty thinks
of a number that equals ∣ raised to a positive integer power. Bob then multiplies Bill's number and
Betty's number. Find the sum of all the possible products Bob can form.
~OSBMNMBMMMH --------------------------------------------------------------------------

Solution for Problem 10.40: * for some positive integer i. Betty's is in

Bill's number is in the form Q)
the form Q)y for some positive integer j. Therefore, the desired sum is the sum of Q) ( j)7 for all pairs
of positive integers i and j. It's not immediately clear how to find this sum, but we can write it in
summation notation. We write "Add Q) Q)7 for all integers i, j ≥ 1" as

Make sure you see why this is the desired sum; it produces a term Q) Q)7 for every pair of positive
integers i and j. Moreover, it tells us how to evaluate the sum. We can factor out Q)' from the inner
summation, and we have

In other words, our desired sum is the product of two infinite geometric series! We evaluate each infinite
geometric series to find that the sum is

Now that we know what to look for, we can see why this is true without using summation notation.
Consider the product of these two infinite geometric series:

111 ∖∕1 1 1 ∖
2 + 22 + 23 + ∕ ∖3 + 32 + 33 + /?

When we expand this product, each term is the product of a power of ∣ from the first sum and a product
of ∣ from the second, and every such product is formed. □

330
 10.7. SUMMARY

Our key step in this solution was writing the sum in summation notation. Then, it was clear how to
evaluate it.

Concept: Sometimes expressing a sum in summation notation will tell us how to

Q—oj evaluate the sum.

liL

_I ∖m ises
_____
10.6.1 Evaluate each of the following as completely as possible:

(a) ΣΣ0 + ^
≈=1 7=l
(b)
ΣΣ3'+'
i=l 7=1
(c) Σ Σ Σ ''k
i=1 ;=1 k=l

10.6.2 Evaluate each of the following:

5 i
(≈>) ∑∑>

2=1 ∕=1
(b)
ΣΣ
*
z=0 ∕=0

5 12
10.6.3 Evaluate ∑ ^(2/ - 7).
f=l j=i

n i n
∑∑x.' = ∑<"
10.6.4 Show that
i=l 7=1 i=!

10.7 Summary

A sequence is a list of numbers, while a series is a sum of numbers.

In an arithmetic sequence, the difference between any two consecutive terms is the same. This fixed
difference is known as the common difference of the sequence. When we add the terms of an arithmetic
sequence, we form an arithmetic series.
Here are some useful facts about arithmetic sequences and series:

• The nth term of an arithmetic sequence that has first term a and common difference d is a + (n - l)d.

• The sequence a, b, c is an arithmetic sequence if and only if b is the average of a and c.

• The average of the terms in an arithmetic sequence is equal to the average of the first and last
terms of the sequence.

• The sum of the terms in an arithmetic sequence equals the product of the number of terms and the
average of the first and last terms.

331
CHAPTER 10. SEQUENCES AND SERIES

rz(2α + (π - l)d)
The sum of an n-term arithmetic series with first term a and common difference d is
2
n(n + 1)
The sum of the first n positive integers is
2

A geometric sequence is a sequence for which there exists a constant r such that each term (except
for the first term) is r times the previous term. This constant r is called the common ratio of the sequence.
When we add the terms in a geometric sequence, we form a geometric series.
Here are some important facts about geometric sequences and series:

• If a geometric sequence has first term a and common ratio r, then the κth term of the sequence is
equal to αrπ^1.
fl(r" - 1)
• A geometric series with n terms, first term a, and common ratio r (with r ≠ 1) has sum —r _ —.

∣ Π
• The sum of an infinite geometric series with first term a and common ratio r, where ∣r∣ < 1, is -.

We have special notations for sequences, series, and products. We use curly braces for sequences,
the capital Greek letter sigma, Σ, for sums, and the capital Greek letter pi, ∏, for products.

Problem Solving Strategies__________________________________________

Concepts: ∙ Setting the middle term of an arithmetic sequence with an odd num-
Q uj ber of terms equal to a variable often introduces a symmetry that
allows us to solve problems more easily.

• We can often tackle problems involving arithmetic series by writing

all the given information in terms of the first term, the common
difference, and the number of terms in the series.

• Before wading through excessive computation or algebraic manip­

ulation in a problem, look for ways to solve the problem more cre­
atively.

• Don't overlook trial and error as a problem-solving strategy. Some

equations are most easily solved by guessing solutions and checking
if they work.

• Many problems involving geometric sequences can be tackled by

writing the terms of the sequence in terms of the first term of the
sequence and the common ratio of the sequence.

Continued on the next page. . .

332
 REVIEW PROBLEMS

Concepts: . . . continued from the previous page

O≡^
• When stuck on a problem, compare it to a similar problem you know
how to solve.

• We can often learn more about an infinite series by investigating the

sums of first n terms of the series for small values of n.

• If you derive a simple algebraic relationship after complex algebraic

manipulation, use the simple relationship as a guide to look for a
simpler proof.

• When in doubt with summation notation, write out the series without
summation notation.
∕
• If we have a summation with a variable among its bounds, we can
often learn more about the summation by choosing different values
for the variable.

Review Problems

10.41 The first term of an arithmetic sequence is 4 and the tenth term is 85. Find the sum of the first
fifty terms in the sequence.
10.42 Evaluate each of the following series:
∞ o
(a) 12 + 15 + 18 + ∙ ∙ ∙ + 1002 (c) ∑31
k=0 °
40
(b) ∑(-3n + 4)
n=l

10.43 Evaluate the following geometric series:

1 1 -√6 + √2-^ +
(a) l÷j÷ig÷- (c)
(b) -3 + 6-12 + 24-------- 768

10.44 Write the following using summation or product notation:

(a) Vl + √2 + √3 + ∙∙∙ + √κ (c) 5 ∙ 6 ∙ 7 ∙ 8 ∙ 9 ∙ 10 ∙ 11

(b) 1 + 4 + 7 + ∙ ∙ ■ + 100 (d) (X-X1)(X-X2)∙∙∙(X

10.45 Find all ordered pairs (x, y) such that 3, x, y is a geometric sequence and x, y, 9 is an arithmetic
sequence.
10.46 Find the value of⅜ + l + l + ^ + ^ + ⅛-7. (Source: AHSME)

333
CHAPTER 10. SEQUENCES AND SERIES

10.47 Find all real numbers a such that the roots of x3 - 6x2 + 21x + a = 0 are not all real and form an
arithmetic progression.
10.48 The sum of all the terms in an infinite geometric progression is 6. The sum of the first two terms
is 9/2. Find the possible first terms of the progression. (Source: AHSME)
10.49 In an arithmetic sequence t1, t2, t3,..., ⅛7, the sum of the odd numbered terms is 1272. What is
the sum of all 47 terms in the sequence? (Source: COMC)
10.50 Prove that for any linear function ∕(x), the sequence ∕(0),∕(l),∕(2),∕(3),... is an arithmetic
sequence.
10.51 In an arithmetic sequence, the pth term is q and the qth term is p, where p ≠ q. Find the (p + <∕)th
term.
10.52
(a) Find the arithmetic sequence with first term 1 and common difference not equal to 0 such that the
second, tenth, and thirty-fourth terms are the first three terms of a geometric sequence.
(b) The fourth term in the geometric sequence in part (a) appears as the nth term in the arithmetic
sequence in part (a). Find the value of n.
(Source: CEMC)
10.53 Find the sum of the series l + 5 + ⅜ + ⅜ + ⅛ + ∙∙∙∕ where we alternately multiply by 1/2 and 1/5
to get successive terms.
, . 1 . l+3 + 5 + ∙∙∙ + (2n-l) 115
10.54 Find a positive integer solution to the equation —2 + 4 + 6 +---- ^+2n~ ~ 116 ’ (Source: AHSME)

10.55 Consider the ten numbers ar, ar2, ar3,..., ar10. If their sum is 18 and the sum of their reciprocals
is 6, determine their product. (Source: COMC)
10.56 Show that if ∣r∣ < 1, then the product of a + ar + ar2 + ■■ ■ and a - ar + ar2 - ■■■ is equal to
a2 + a2r2 + a2r4 + ∙ ∙ ∙.
10.57 The sum of an infinite geometric series with common ratio r such that ∣r∣ < 1 is 15, and the sum
of the squares of the terms of this series is 45. Find the first term of this series. (Source: AHSME)
n
ιt∑gW = n
10.58 find a formula for g(E). (Source: ARML)
3~n - 2'
k=l
10.59 In each of the following parts, state whether or not the two summations are equal, and explain
why or why not the sums are equal.
19 21 20 i 20 71 - lc
(a) ∑√2^ι2> and X,(-2+4fc-fc2) (c) ∑22^7 and ∑τ+τ^
i=-4 k=-2 j=l , fc=l

00 1 00 1
(b) y —----- and y —-----
⅛2*÷4 ⅛
* +l

10.60 Show that if c is a positive constant, then ⅛ = 4 + j⅜ + ^∙ + ∙∙∙ for all x such that ∣x∕c∣ < 1.

334
 CHALLENGE PROBLEMS

10.61 Evaluate each of the following:

(a) ΣΣ°'^^2

i=l 7∙=1

10.62 Evaluate each of the following:

(a) ∑∑3
7=0 k=0

k k
10.63 Find a closed form for ΣΣ2∕
7=0 i=0

10.64 A harmonic sequence (or harmonic progression) is a sequence in which the reciprocals of the
terms form an arithmetic sequence. A harmonic series is a sum of the numbers in a harmonic sequence.

(a) Let Sn denote the sum of the first n terms of a harmonic sequence whose first two terms are 3 and
4. Compute S4.
(b) The first term of a harmonic sequence is 1/3, and the fifth term is 1. Find the second, third, and
fourth terms.
(c) Find all real numbers x such that x + 4, x + 1, x is a harmonic progression.
(d) Let a, b, and c be positive real numbers. Prove that if a, b, c is a harmonic sequence, then so is
«/(b + c), b∕(α + c), c∕(α + b).

Challenge Problems

10.65 It is possible to place positive integers into the vacant twenty-one

squares of the 5 × 5 square shown on the right so that the numbers in each ★
row and column form arithmetic sequences. Find the number that must
74
occupy the vacant square marked by the ★. (Source: AIME) Hints: 125

10.66 Write xι(xι + 2x2)(x1 + 2%2 + 3xβ) ∙ ∙ ∙ (xι + 2x2 + 3xβ + ∙ ∙ ∙ + nxn) using 186
product and summation notation.
103
10.67 A set of consecutive positive integers beginning with 1 is written
on a blackboard. One number is erased. The average (arithmetic mean) 0
of the remaining numbers is 35 What number was erased? (Source:
AHSME) Hints: 181
10.68 If a and b are the roots of llx2 — 4x — 2 = 0, then compute the product

(1 + a + α2 + a3 -ι---- )(1 + b + b2 + b3 + ∙ ∙ ∙).

(Source: ARML) Hints: 235

335
CHAPTER 10. SEQUENCES AND SERIES

5 i j
10.69 Evaluate ∑∑∑<1>∙
i=l j=l fc=l

10.70 Let a, b, and c be positive real numbers. Prove that if a2, b2, c2 is an arithmetic sequence, then so
is l∕(fo + c), l∕(β + c), 1∕(∏ + b).
τrι ιι
10.71 Let p(x1, x2, *3, ∙ ∙ ∙, xπι) = ∏ V Xk' Find the sum of the coefficients of the θxPanded form of the
n=l k=l
m-variable polynomial p.
10.72 The roots of the equation x5 - 80x4 + Px3 + Qx2 + Rx + S = 0 are in geometric progression. The
sum of their reciprocals is 5. Compute ∣S∣. Hints: 243
10.73 Consider the triangular array of numbers formed when the numbers 0, 1, 2, 3, ... are placed
along the sides and interior numbers are obtained by adding the two adjacent numbers in the previous
row. Rows 1 through 6 are shown below.

Let f(ri) denote the sum of the numbers in row n. What is the remainder when ∕(100) is divided by 100?
(Source: AHSME) Hints: 130,218
10.74 If the integer k is added to each of the numbers 36,300, and 596, one obtains the squares of three
consecutive terms of an arithmetic sequence. Find k. (Source: AIME) Hints: 237
10.75 Si is the sum of an infinite geometric series with first term a and common ratio r. S2 is the sum of
another geometric series with first term a2 and common ratio r2. Similarly, for all n≥2, let Sn be the sum
of the infinite geometric series with first term an and common ratio rn. If ∣r∣ < 1 and ∣α∣ > 1, determine
the sum of the infinite series

in terms of a and r. (Source: CIMC) Hints: 110

∞ 1 00 1 00 00 |
10.76 Define p = £ - and <∕ = ∑,μ∙ Write — in terms of p and q. (Source: Mandelbrot)
k=l k=l ∕=1 k=l J
Hints: 262,13
10.77 Set A consists of m consecutive integers whose sum is 2m, and set B consists of 2m consecutive
integers whose sum is m. The absolute value of the difference between the greatest element of A and the
greatest element of B is 99. Find m. (Source: AIME)
10.78 An infinite geometric series has sum 2005. A new series, obtained by squaring each term of the
original series, has sum 10 times the sum of the original series. Find the common ratio of the original
series. (Source: A1ME)

336
 CHALLENGE PROBLEMS

*
10.79 Find all pairs of integers n and k, with 2 < k < n, such that the binomial coefficients

form an increasing arithmetic sequence. (Source: USAMTS) Hints: 23,150

*
10.80 If a, b, and c form an arithmetic progression, and

α = x2 + xy + y2,
b = x2 +χz + z2,
c = y2 + yz + z2,

where x + y + z ≠ 0, prove that x, y, and z also form an arithmetic progression. Hints: 194
10.81
* A sequence of positive integers with a↑ = 1 and a<) + αιo = 646 is formed so that the first three
terms are in geometric progression, the second, third, and fourth terms are in arithmetic progression,
and, in general, for all j > 1, the terms fl2∕-i∕ a2j, a2j+ι are in geometric progression, and the terms a^j,
a2j+ι, and α2∕+2 are in arithmetic progression. Find the greatest term in this sequence that is less than
1000. (Source: AIME) Hints: 159,333

10.82
* In an increasing sequence of four positive integers, the first three terms form an arithmetic
progression, the last three terms form a geometric progression, and the first and fourth terms differ by
30. Find the sum of the four terms. (Source: AIME) Hints: 295
10.83 Let∣x∣ < 1. Express (l+x)(l+x2)(l+x4)(l+x8) ∙ ∙ ∙ as ∕(x)∕g(x), where ∕(x) and g(x) are polynomials
with finite degree. Hints: 26

10.84
* For 0 < x < 1, let ∕(x) = (1 + x)(l + x4)(l + x16)(l + x64)(l + x256) ∙ ∙ ∙. Compute: f~1

(Source: ARML) Hints: 228

∞ 2"
Xz
10.85
* For 0 < x < 1, express J~lj as ∕(x)∕g(x), where ∕(x) and g(x) are polynomials with finite
1 - x2''+1
n=0

degree. (Source: Putnam) Hints: 53

Extra! Mathematicians have closely studied a great many different sequences, such as
HlφHI∣+∣∣llΦ∣ll∏t
1,1,2,5,14,42,132,... .

If you come across a sequence of numbers that you'd like to learn more about, you can
visit The On-Line Encyclopedia of Integer Sequences, which is a collaborative project
to collect information about sequences. A link to The On-Line Encyclopedia of Integer
Sequences can be found on the Links page cited on page vi. In case you're wondering,
the sequence 1, 1, 2, 5, 14, 42, 132, ... consists of the Catalan numbers, which you can
learn more about in Art of Problem Solving's Intermediate Counting & Probability.

337
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

Science is nothing but the finding of analogy, identity, in the most remote parts. - Ralph Waldo Emerson

CHAPTER 11 ------- I
I Identities, Manipulations, and Induction

You've seen many identities already. The difference of squares factorization

χ2 - y2 = (χ + y)(χ - y)

is an identity because it's true for any ordered pair (x, y). On the other hand, the equation

x2 - y2 = 0

is not an identity, because it is not true for all values of x and y.

Definition: An identity is an equation that is true for all values of all variables for which the
equation is defined.

We've already seen how useful many algebraic identities can be in our discussion of factorization.
In this chapter, we develop more methods for finding algebraic identities. Most of these identities
themselves are not terribly important (we'll highlight the ones that are), but the procedures we use to
develop them are.

11.1 Brute Force

We can prove the difference of squares identity, χ2 - y2 = (χ - y)(x + y), by simply multiplying out the
right side:
(x - y)(x + y) = x(x + y) - y(x + y) = x2 + xy - xy - y2 = x2 - y2.

Many identities can be proved similarly, though often the necessary algebra is more complicated.
We call this approach brute force.

338
 11.1. BRUTE FORCE

Problems

Problem 11.1: Show that (a + b + c)(a2 + b2 + c2 - ab - be - ra) = a3 + b3 + c3 - 3abc.

Problem 11.3:
(a) Find an identity that relates the expressions x2+y2+z2, ⅛+Λ + ⅛,and(^ + ^)2÷(f + yf+(f + i↑-

(b) Given that x2 + y2 + z2 = 6 and + j) + (7 + ∑)2 + (f + f )2 = 16.5, what is + ⅛ + ⅛7 (Source:

Duke Math Meet)

Problem 11.4: In this problem, we evaluate λ∕50 ∙ 51 ∙ 52 ∙ 53 + 1 without a calculator. (Source: MATH­
COUNTS)
(a) Experiment. Evaluate √1 ∙2 ∙ 3 ∙4 + 1, √2∙3∙4∙5 + l, √3 ∙4 ■ 5 ∙ 6 + 1,and √4∙5∙6∙7 + l. Can
you guess the answer to the problem?
(b) Let x = 50 and write the expression we wish to evaluate in terms of x. Let the expression inside
the radical be f(χ). What is deg/?
(c) If there is a polynomial g(x) such that (g(x))2 = ∕(x), then what is deg g?
(d) Find V50 ∙ 51 ∙ 52 ∙ 53 + 1 without a calculator.

Problem 11.1: Show that (a + b + c)(a2 + b2 + c2 - ab-be - co) = a3 + b3 + c3 - 3abc.

Solution for Problem 11.1: We have

(a + b + c)(a2 + b2 + c2 - ab - be - c«) = a(a2 + b2 + c2 - ab - be - ca)

+ b(a2 + b2 + c2 - ab - be - ca)
+ c(a2 + b2 + c2 - ab-be- ca)
= a3 + ab2 + ac2 - a2b - abc - a2c
+ a2b + b3 + bc2 - ab2 - b2c - abc
+ a2c + b2c + c3 - abc - bc2 - ac2
= a3 + b3 + c3 - 3abc.

Make sure you see how all the other terms cancel out. □
Our work in Problem 11.1 doesn't require any particular insight; it's simply careful algebraic manip­
ulation.

339
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

Concept: When performing complicated algebra, stay organized and write neatly.
O≡^

Problem 11.2:
(a) Prove that (a? + a^)(b^ + b?) - (a↑b1 + fl2⅛)2 = (β1b2 ~ ^2b1)2.
(b) Prove that (α2 + a^)(b^ + b?) > (a↑b1 + fl2⅛)2 for all real numbers αι,fl2∕ bi, and b2.

Solution for Problem 11.2:

(a) We have

(λJ + a⅛)(b2 + b⅛) - (fl1bι + α2½)2 = a2b2 + a2b2 + a2b2 + fl^2 - fl2⅛ - 2λiΛ2^i⅛ - ⅛⅛
= a2b2 - 2aγa2b∖b2 + a⅛b∖
= (β1½ -a2bf)2.

This is yet another example of how useful it is to be able to recognize the square of a binomial.
(b) Our identity from part (a) gives us

(a2 + a2)(b2 + b2) - («ih + a2b2)2 = (a1b2 - a2b1)2.

Because a↑b2 - a2bγ is a real number, we have (a1b2 - α2b1)2 ≥ 0∙ Combining this with the equation
above gives us
(a2 + a2)(b2 + b2) - + a2b2)2 ≥ 0.
Adding (a↑b↑ + α2½)2 to both sides of this inequality gives us the desired

(a2 + α2)(b2 + b⅛) ≥ (aιbχ + α2½)2∙

The identity we proved in part (a) of Problem 11.2 is a special case of Lagrange's Identity. The
inequality we proved in part (b) is a similar special case of the Cauchy-Schwarz Inequality, which we
discuss in much greater detail in Chapter 12.
We continue with a couple problems in which we must discover an identity to evaluate expressions.

1 1 1 π
Problem 11.3: Given that x2+y2+z2 = 6and 16.5, what is
x~ y- z2
ource: Duke Math Meet)

Solution for Problem 11.3: We start by expanding the left side of the second equation, since doing so will
produce squares and reciprocals of squares. This gives us

= 16.5,

340
 11.1. BRUTE FORCE

so

Our hope is to relate this equation to the first given equation and to the expression we must evaluate.
The left side looks like some of what we expect from expanding the product of x1 + y2 + z2 (the left side
of the first given equation) and ⅜ + ~⅛ + (the expression we're trying to evaluate). So, we take a look
at that product:

Aha we recognize many of the expressions in this identity! We know that x2 + y2 + z2 = 6, and the
expression on the right equals 3 + 10.5 = 13.5. Substituting these into our new identity, we have

Recall from our discussion of Vieta's Formulas that expressions like x2 + y2 + z2 and ~5 + ]⅛ + ⅛ are
called symmetric, because exchanging any two of the variables in the expression does not change the
expression. Our solution to Problem 11.3 is an example that:

Concept: We can often relate symmetric expressions with identities.

O≡^

As we saw in Problem 11.3, we'll sometimes have to develop identities as we need them. Often these
identities won't be immediately obvious, so we'∏ have to experiment a little to find them, as we did
when we multiplied x2 + y2 + z2 and X⅜ + ⅛
y
+ Z⅛ ∙

Concept: If it's not immediately obvious how to solve a problem, don't just sit and
θ- ljlj stare at it. Experiment!

Problem 11.4: Evaluate √50 ∙ 51 ∙ 52 ∙ 53 + 1 without a calculator. (Source: MATHCOUNTS)

Solution for Problem 11.4: Multiplying that out doesn't look like much fun. Then, finding the square root
of the result looks like even less fun. Let's find a more clever approach. We start by making the problem
simpler by evaluating similar expressions for smaller numbers.

Concept: When asked to evaluate an expression involving large numbers, you can
O,'~ ∣->∣j gθt a fθθl for the problem by evaluating similar expressions with small
numbers.

341
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

We find:

√0 ■ 1 ∙ 2 ∙ 3 + 1 = 1,
√1 ■ 2 ∙ 3 ∙ 4 + 1 = 5,
√2 ■ 3 ∙ 4 ∙ 5 + 1 = 11,
√3∙4∙5∙6 + 1 = 19.

There's a subtle pattern here. The second result is 4 more than the first, the third is 6 more than the
second, and the fourth is 8 more than the third. However, it's not clear how to prove this pattern
continues. To see if it does, we replace 50 with x in √50 ∙ 51 ∙ 52 ∙ 53 + 1, and see if we can find the square
root of the resulting expression:

y∕x(x + l)(x + 2)(x + 3) + 1 = V(χ2 + x)(χ2 + 5x + 6) + 1 = Vx4 + 6x3 + llx2 + 6x + 1.

Our examples make us suspect that this quartic is the perfect square of some polynomial. If it is, then
that polynomial must be a quadratic. Because the first and last terms of the quartic are x4 and 1, and
all the coefficients of the quartic are positive, we guess that the square root is of the form x2 + ax + 1 for
some positive value of a. Rather than square x2 + ax + 1, we note that the x3 term of the expansion of
(χ2 + ax + l)(x2 + ax + 1) is 2αx3, so we guess that a - 3 to make this term equal to 6x3. Testing, we find
that
(χ2 + 3x + 1)2 = x4 + 6x3 + llx2 + 6x + 1,
so we have
λ∕x(x + l)(x + 2)(x + 3) + 1 = Vx4 + 6x3 + llx2 + 6x + 1 = x2 + 3x + 1.
Letting x = 50, we find V50 ∙ 51 ∙ 52 ∙ 53 + 1 = 502 + 3 ∙ 50 + 1 = 2651. □
You might be wondering if it's possible to perform expansions like

∖x2 + y2

and
(x2 + 3x + l)(x2 + 3x + 1) = x4 + 6x3 + llx2 + 6x + 1
in your head without filling up a lot of paper with intermediate steps. It is possible, but you have to be
careful.
For example, consider

(x2 + y2 + z2)

When expanding the left side, we are multiplying two expressions with 3 terms each. So, the expansion
must have 3 × 3 = 9 terms. Because each term in the expansion is the product of one term of x2 + y2 + z2
and one term of -⅛ + ⅛ + ⅛, we see that each of the possible ratios of one of x2, j∕2, or z2 to one of x2, ι∕2,
or z2 is formed in the expansion. Three of these ratios equal 1, which gives us the 3 on the right side.
The remaining possible ratios give us the other terms. We can quickly test our expansion by checking if
the identity holds for x = y = z = 1. We find (1 + 1 + 1)(1 + 1 + 1) = 3 + 1 + 1+1 + 1 + 1 + 1, which is true.

342
 11.2. RATIOS

Concept: Because an identity must be true for all permissible values of the vari-
Q—mJ ables in the identity, we can test an identity we develop by substituting
____ convenient values for the variables.____________________

To expand (x2 + 3x + l)(x2 + 3x + 1), we focus on each power of x in the expansion in turn.

• x4. There's only one way to form x4: (x2)(x2) = x4.

• x3. There are two ways to form x3: (x2)(3x) + (3x)(x2) = 6x3.

• x2. There are three ways to form x2: (χ2)(l) + (3x)(3x) + (l)(x2) = llx2.

• x. There are two ways to form x: (3x)(l) + (l)(3x) = 6x.

• Constant. There's only one way to form a constant: (1)(1) = 1.

Notice that across all five of these cases, we form a total of 9 different products of terms in the two
expressions we are multiplying. Combining all five cases gives us

(x2 + 3x + l)(χ2 + 3x + 1) = x4 + 6x3 + llx2 + 6x + 1.

We can quickly test this by letting x = 1, which gives us (5)(5) on the left and 1+6 + 11 + 6 + 1 on the
right. These both equal 25, so the expansion passes our quick test.

i I 11----------------------------------------------------- ----- — ------------------------------ ---- ---

Exercises

11.1.1 Show that a2(b -c) + b2(c -a) + c2(a -b) = bc(b - c) + ca(c - a) + ab{a - b) for all a, br and c.
11.1.2 Show that 2(x - y)2 - 3x2 + 3xy = (y - x)(x + 2y) for all values of x and y.
11.1.3 Consider the following equations:

12 + 22 + 22 = 32,
22 + 32 + 62 = 72,
32 + 42 + 122 = 132,
42 + 52 + 202 = 212.

State and prove a one-variable identity that is suggested by these examples.

11.1.4 If xyz = x + y + z = ∣ + → ∣ = 8, then find (x - i)(y - ∣ξz - lj.

11.2 Ratios

In this section, we explore some very handy manipulations involving ratios.

343
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

Problems

Problem 11.5: Let γ = ~.

b d
(a) Let 7 = k. Use this to show that 7 = 3 = 7—3 = 7—3 for those values of b and d for which the
b bdb + db-d
denominators are nonzero.
n . r,, , a + b c + d .r t , ,
(b) Show that-----7 =------ if a ≠ b and c ≠ d.
a-b c-d

x> ½ anc^ z such

Problem 11.6: Find the numerical value of k for which------ =------- — ------ fθr a^ *
x+y x+z z-y
that the denominators of the fractions are nonzero. (Source: ARML) _____________________

r a c
Problem 11.5: Let 7 = 3∙
b d
(a) Show that = = ⅛~j = —∣ fθr those values of b and d for which the denominators are
nonzero.
, a+b c + d .. . , ,,
(b) Show that----- 7 =---- 3 if a ≠ b and c ≠ d.
v ’ a-b c-d

Solution for Problem 11.5:

(a) Solution 1: Set the given equal ratios equal to a variable. We'd like to get rid of the fractions. One way
to do so is to let a c
k=b = d'
This gives us a = bk and c = dk, so we have
a+c bk + dk k(b + d) , . a-c bk-dk k(b-d) 1
------ =---------- = —--------- = k and ;----- ∑ = —-----— = —----- — = k.
b+d b+d b+d b-d b-d b-d
a + c_a-c
Therefore, we have 7 = - = bTd ~ b^d'
b d
Concept: When given equal ratios, setting the ratios equal to a variable helps elimi-
0≡≈g nate the fractions.

344
 11.2. RATIOS

Solution2∙. Bruteforce. We also could have used brute force to show that (a + c)∕(b + d) = a/b. It'snot
so clear how to start with a/b = c/d and prove (fl + c)∕(b + d) = a∕b, so we instead work backwards.
We start with (fl + c)∕{b + d) = a/b and try to work from here to an equation we know to be true.
We start by multiplying both sides of this equation by b(b + rf) to get rid of the fractions. This
gives us b(a + c) = a(b + d), so ab + bc = ab + ad, which means be = ad. Dividing both sides of this by
bd gives us c/d = a/b, which we know to be true. This gives us a path to prove (fl + c)∕{b + d) - a/b
starting from a/b = c/d. We simply reverse our "working backwards" steps:
Proof: We multiply a/b = c/d by bd to give ad = be. Adding ab to both sides gives the equation
0 ad + ab = be + ab. Factoring gives a(b + d) = b(a + c). Because b and b + d are nonzero,
we can divide both sides by b(b + d) to give the desired

fl _ fl + c
b ~ bTd'
We can tackle a/b = (a — c)∕(b - d) in much the same way.
Our key step in finding this brute force solution was working backwards.
Concept: Working backwards from a relationship you wish to prove is a powerful
O—°d problem-solving method.

WARNING!! If you do work backwards to find a solution, make sure you write
* your solution "forwards" as we've done in the Proof box above. This
will make your work easier to read, and help you catch any mistakes
you might have made.

Also, some steps are not reversible. For example, it's true that if x = y, then x2 = y2. But we can't
reverse this and say that "If x2 = y2, then x = y." The step going from x = y to x2 = y2 is not
reversible; if we know that x2 = y2, we cannot deduce that x = y because it's possible that x = - y
instead. (Look back to Problem 8.27 on page 254 to see an example in which writing our solution
forwards saves us from a subtle error that might occur when we work backwards.)
(b) Solution 1: Set the ratios equal to k. Letting a/b = c∕d = k as before, we have a = bk and c = dk, so

a + b_bk + b_k+l c + d dk + d k + 1
7→ ~ bk^b ~ k-1 an ^^d~ dk^d = k^V

so (a + b)∕(a -b) = (c + d')∕{c - d).

Solution 2: Clever manipulation. We wish to prove an equation involving a + b and c + d. Adding 1
to a/b = c/d gives us
a+b_c+d
b d
Similarly, we need a - b and c - d, so we subtract 1 from a/b = c/d, which gives

a-b c- d
b d '
c+d u ~b c—d , b a+b c÷d
Dividing ~η~ = “ , by = gives us the desired----- 7 - ------- □
d b d σ a-b c-d

345
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

This problem gives us examples of the following useful ratio manipulations:

Important: Suppose a/b = c∣d. Then, for any constant k for which b + kd ≠ 0, we
have
Ψ a _ c _ a + ^c
b d b + kd'
For any constant k for which a — kb and c — kd are nonzero, we have

a + kb _ c + kd
a - kb c — kd ____ ___ .

You'll have a chance to prove these as Exercises. We can also extend the fact that £ — cd implies b+d
to any number of equal ratios:

Important: M ⅛ = ⅛ = ∙∙∙ = ⅛ and y1+y2 + ∙∙∙ + yw≠0, then

ψ 2
Xl = * Xn _ Xi + X2 + ∙ ∙ ∙ + Xn
yn ~ yι + y2 + ∙ ∙ ∙ + y∕
yι ι∕2

You'll have a chance to prove this as an Exercise.

Using these relationships allows us to combine equal ratios to produce other ratio equations. Let's
take a look at how.
7 k 11
Problem 11.6: Find the numerical value of k for which------ =------- =-------- for all x, y, and z such
∣
x+y x+z z-y
that the denominators of the fractions are nonzero. (Source: ARML)
Solution for Problem 11.6: We notice that the sum of the denominators of the left and right fractions is
the denominator of the middle fraction. So, we can use the result of the previous problem to determine
that 7 _ 11 _ 7 + 11 _ 18
x + y -z-y^x + y + z-y-x + z'
Therefore, we have k = 18. □
x+y y + z _ x + z and abc ≠ 0, then -⅜ = -⅛ = -⅞.
Problem 11.7: Show that if
a2 + b2 b2 + c2 a2 + c2 az b2 co­

Solution for Problem 11.7: Brute-forcing this doesn't look like much fun. Let's try applying the identities
we just learned. Because
χ + y = y+z
a2 + b2 b2+c2'

we can apply the relationship we just proved to find

x+y _ y+z _ (x + y) + (y + z) _ x + 2y + z
a2 + b2 b2 + c2 (fl2 + b2) + (h2 + c2) fl2 + 2b2 + c2

346
 11.2. RATIOS

Hmmm.... These ratios also equal (x + z)∕(π2 + c2), so we have

x + y _ y + z _ x + z _ x + 2y + z _ (x + z) + (x + 2y + z) _ 2x + 2y + 2z
fl2 + b2 b2 + c2 a2 + c2 a2 + 2b2 + c2 (a2 + c2) + (α2 + 2b2 + c2) 2a2 + 2b2 + 2c2

So, we have
x+y _ y+z χ+z x+y+z
a2 + b2 b2 + c2 a2 + c2 a2 + b2 + c2

In part (a) of Problem 11.5, we showed that if = then we have = = Applying this to the
first and last equal ratios in our equality chain above, we have

x + V + z = * + y = x + y + z-(x + y) = £
a2 + b2 + c2 a2 + b2 a2 + b2 + c2 - (α2 + b2) c2

Similarly, we can show that ——------ equals both -⅜ and f⅞∙, so -r—----- - x_ = V_ _ £
az + b~ + c2 a2 b2 a2 + b2 + c2 a2 b2 c2 ’
We can also use our ratio manipulation tactics on inequalities.

Solution for Problem 11.8: We try letting k = a/b = c∕d, as before. This gives us a = bk and c = dk. Now,
we work backwards from the inequality we wish to prove. Letting a = bk and c = dk in a2 + d2 > b2 + c2,
we must now show that b2k2 + d2 > b2 + d2k2. We subtract b2 and d2 to give

b2k2 -b2> d2k2 - d2.

Factoring then gives us b2(k2 - 1) > d2(k2 - 1) to prove. We're told that b > d > 0, so b2 > d2. (Make sure
you see why it's important that b and d are positive for this step!) We also are given that a/b > l,sok > 1.
This means that k2 -1 is positive, so multiplying b > d by k2 -1 gives us the desired b2(k2 -1) > d2(k2 -1).

Our work above is not a great way to write a solution, but it is a great way to find a solution.
Specifically, we worked backwards from a2 +d2 > b2 + c2 to produce an inequality, b2(k2 - 1) > d2(k2 - 1),
that we know how to prove from the given information. So, we figured out how to solve the problem
using a combination of working forwards and working backwards. As our last step, we write our
solution entirely forwards, starting from the given information and ending at the desired inequality.

Proof: Let k = a/b = c∕d. Because a/b > 1, we have k > 1, so k2 - 1 > 0. Because b > d > 0, we have
0 b2 > d2. Multiplying this inequality by k2 - 1 gives b2(k2 - 1) > d2(k2 - 1). Since k = a/b = c∕d,
we have
fc⅜2 -1) = ⅛2 _ ιj = fl2 _ ∖fi i

Similarly, we have d2(∕k2 -1) = c2 - d2, so our inequality is now a2 -b2 > c2 - d2. Adding b2 + d2
to both sides gives the desired a2 + d2 > b2 + c2.

347
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

„ λ cι c
11.2.1 Suppose that 7 = -.
b d
cι ÷ kc
(a) Show that 7 = 3 =---------- for any constant k such that b + kd ≠ 0.
b d
b + kd--------- j
cι ^4- Jeb c ∣ led
(b) Show that-----— =------- for any constant k such that a- kb and c -kd are nonzero.
a-kb c-kd 7

11.2.2 Show that if — = — • — and yi + y2 + ∙ ∙ ∙ + y„ ≠ 0, then

Vi Vi y∏
X∖ _ X2 _ _ Xn _ X1 + x2 + ∙,, + χn
yi y2 yπ yi + y2 + ∙ ∙, + y«

α b
11.2.3 If ------------- and x + y + z ≠ 0, must all three fractions equal -—
x + y-z x-y + z -x + y + z------------------------------------------------------------------ x + y + z

x + y _ c2 + d2
11.2.4 Show that if and yd ≠ 0, then -
x - y c2 -d2 y

11.2.5 Let - = -.
y z
χ2 + y2 y2 xz + xy + ιr χ
(a) Show that -3----- 7 = -y *
(b) Show that -3------------- - -.
y2 + z2 z2- z2 + yz + y2 z

11.2.6 ★ If a, b, c are nonzero real numbers such that

a+b-c a -b + c _ -a + b + c
c b a

(α + b)(b + c)(c + α)
X~ rt^c '
and x < 0, then find x. (Source: AHSME) Hints: 89,199
*
11.2.7 Suppose that x1 > 0 for 1 ≤ i ≤ 6, and that — < — Show that — < — + *3 + λ5 < —.
rr X2 X4 Xβ X2 X2 + *
4 + Xβ xf>
Hints: 351

11.3 Induction

The principle of mathematical induction (or often just induction) is one of the key proof methods that
we use to prove statements about positive integers. The classic metaphor that is usually used to describe
mathematical induction is a row of dominos all standing on end:

348
 11.3. INDUCTION

If we tip the first domino over:

and if each domino, as it falls, knocks over the next domino in line:

then ultimately, all of the dominos will fall over:

This is the basis of mathematical induction. Each "domino" is a mathematical statement that depends
on n. The first domino is the statement where n = 1, the second domino is the statement where n = 2,
and so on. If we can prove that the first statement (where n = 1) is true, and that each statement implies
the next one, then we can knock all the dominos down, and prove that the statement is true for all n.
Obviously, this is all rather vague at the moment. We'll start by walking through a basic proof by
induction to see how it works. Mathematical induction is one of the most common proof techniques,
and you should master it.

Problems

Problem 11.9: In this problem, we prove the identity 1 + 2 + 22 + 23 + ∙ ∙ ∙ + 2π = 2'i+1 - 1 for all positive
integers n.
(a) Show that the identity holds for n = 1.
(b) Show that if the equation l+2 + 4 + ∙∙∙ + 2^ = 2k+1 - 1 is true, then we also have

l÷2÷4 + *∙ + 2^÷ 2^c+1 ~ 2^+2 — 1

(c) Why can we use part (b) together with part (a) to deduce that 1 + 21 + 22 = 23 - 1?
(d) Explain why we can use part (b) together with part (a) to conclude that 1+2+4+∙ ∙ ∙+2" = 2"+1-l,
where n is any positive integer.

We'll start with an identity that you may know how to prove with other methods.

349
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

Problem 11.9: In this problem, we prove the identity 1 + 2 + 22 + 23 H------ 1-2” = 2π+1 -1 for all positive
integers n.
(a) Show that the identity holds for n = 1.
(b) Show that if the equation l+2 + 4 + ∙∙∙ + 2λ = 2fc+1 - 1 is true, then we also have

l + 2 + 4 + ∙∙∙ + 2fc + 2fc+1 = 2k+2 - 1.

(c) Why can we use part (b) together with part (a) to deduce that 1 + 2*
1 + 22 = 23 — 1?
(d) Explain why we can use 1 + 21 = 22 -1 together with part (b) to conclude that l + 2+4-ι---- 1-2” =
2"+1 -1, where n is any positive integer.

Solution for Problem 11.9:

(a) The identity holds for n = 1 because 1 + 21 = 21+1 - 1. We call this the base case of our induction.
(b) Suppose that the identity holds if n = k, so that

1 + 21 + 22 + ∙ ∙ ∙ + 2k = 2* +1 - 1. (11-D

We wish to show that

l + 2 + 4 + ∙∙∙ + 2fc + 2fc+1 = 2k+2 -1. (11.2)
By simply adding 2fc+1 to both sides of Equation (11.1), we make the left side equal the left side of
Equation (11.2), and we have

l+2 + 4 + ∙∙∙ + 2fc + 2fc+1 = 2fc+1 - 1 + 2k+1.

Simplifying the right side of this equation gives 2(2fc+1) — 1 = 2k+2 - 1, so we have

l+2 + 4 + ∙∙∙ + 2fc + 2fc+1 = 2k+2 - 1,

as desired.
(c) We've shown that the identity holds for n = 1. We've also shown that if the identity holds for
n = k, then it holds for n = k + 1. So, because the identity holds for n = 1, it holds for n =2, which
means 1 + 2 + 22 = 23 — 1.
(d) Because the identity holds for n = 2, it holds for n = 2 + 1 = 3. Because it holds for n = 3, it holds
for n = 4. Because it holds for.... You get the picture. We can continue this reasoning indefinitely,
thereby showing that the identity

1 + 2 + 22 + 23 + ■ ∙ ∙ + 2” = 2”+1 — 1

holds for every positive integer n.

Our solution above is a basic example of proof by induction. The proof involves two main steps:

1. The base case. We showed that the identi ty is true for n = 1. This is usually pretty easy.

350
 11.3. INDUCTION

2. The inductive step. We show that if the identity is true for n = k, then the identity holds for n = k +1.

Combining the base case with the inductive step tells us that because the identity is true for n = 1,
it is true for n = 1 + 1 = 2. We can then use n = 2 together with the inductive step to deduce that the
identity is true for n = 2 + 1 = 3. And so on.

The inductive step is usually the hardest part of proving a statement by induction. It begins with
the inductive assumption, which is the assumption that the statement is true for some integer k. The
inductive assumption is also sometimes called the inductive hypothesis. In Problem 11.9, our inductive
assumption is our identity for n = k, which is

1 + 21 + 22 + ∙ ∙ ∙ + 2k = 2fc+1 - 1.

In the inductive step we used this assumption to prove the identity for n = k + 1, which is

I 2^1 ∣ 2^ ∣ ., . ∣ 2^ ∣ _ '2'k+'2._2
This, combined with our base case n = 1, completed the proof by induction.
Solve each of the following problems with induction. In each case, clearly identify the base case, the
inductive assumption, and the inductive step.
I I 11 ________________________________

Problems

Problem 11.12: Use induction, together with the identity established in Problem 11.10, to show that

13 + 23 + 33 + ∙ ∙ ∙ + n3 = (1 + 2 + 3 + ∙ ∙ ∙ + n)2

for all positive integers n.

Problem 11.14: Show that if n is a positive integer greater than 2, then

111
----- — 4- ------ 4- ----- — 4- ... 4- _13
— > —
n +1 n + 2 n + 3 2n 5

351
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

Solution for Problem 11.10: We wish to show that

n(n + 1)
l + 2 + 3 + ∙∙∙ + n =
2

For our base case, we have 1 = which is true.

For the inductive step, we assume the identity holds for n — k; that is, we assume

k(k + 1)
l + 2 + 3 + ∙∙∙+fc = (11.3)
2
We must now show that the identity holds for n = k + l. When n = k + l, the identity is

(⅛ + l)(fc + 2)
l+2 + 3 + ∙∙∙ + k + (k + l) — (H-4)
2
We wish to use our inductive assumption, Equation (11.3), to show that Equation (11.4) is true. Adding
k + 1 to both sides of Equation (11.3) gives us

l + 2 + 3 + ∙∙∙ + fc + (∕c+l) = —-———- + k + 1

k2 + k 2k + 2
~ 2 + 2
_ k2 + 3k + 2
2
(⅛ + l)(fe + 2)
2
which shows that Equation (11.4) is true. So, we have shown that if the identity holds for n = k, then
it also holds for n = k + 1. Since we have also shown that the identity holds for n = 1, our induction is
complete and we have proved the identity for all positive integers. □

Important: The sum of the first n positive integers is .l(w+1).

When writing up an induction proof, we don't have to be so long-winded. For example, we can
write our solution to Problem 11.10 as:

352
 11.3. INDUCTION

Proof: We proceed by induction. For the base case, we have 1 = ---ly1 ), which is true. Now, we assume
0 that the identity holds for n = k, which gives

k(k + 1)
1 +2+3+∙∙∙+k=
2
Adding k + 1 to both sides of this equation gives us

(k + l)(Ar + 2)
2
so the identity holds for n — k + 1. This completes the inductive step, so the identity holds for
all positive integers n.

Problem 11.11: Let n be a positive integer. Prove that 1 ■ 1! + 2 ∙ 2! + 3 ∙ 3! + ∙ ∙ ∙ + n ■ n! = (n + 1)! - 1.

■■■■■■■a ■MB

Solution for Problem 11.11: What's wrong with this solution:

Bogus Solution: We assume that

1 ∙ 1! + 2 ∙ 2! + 3 ∙ 3! + ∙ ∙ ∙ + k ∙ k'. = (k + 1)! - 1

for some positive integer k. Using this inductive assumption, we

have

1 ∙ 1! + ■ ∙ ∙ + k ■ kl + (k + 1) ∙ (k + 1)! = (k + 1)! — 1 + (k + 1) ∙ (k + 1)!

= (k + 2)∙(k + l)!-l
= (k + 2)!-l.

This completes the inductive step, so our induction is complete.

The error here is that we forgot the base case! If we don't have a starting point for our induction,
then showing that the identity holds for n = k +1 if it holds for n = k does us no good. Trying to perform
an induction without establishing the base case is like trying to knock over a row of dominos without
knocking over the first domino to start the dominos falling!

Fortunately, the base case is easy. For rz = 1, we have 1 ∙ 1! = (1 + 1)! - 1, which is true. Combining
this with the work from our Bogus Solution gives us a complete proof by induction. □

WARNING!! Don't forget the base case when using induction.

Notice that each of the identities we have proved with induction so far is an identity that holds "for
all positive integers." This is usually a strong sign that induction might succeed.

353
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

Concept: Mathematicians often use the letter n to stand for a nonnegative or positive
Q... ⅝r⅞=> integer. Because of this, youTl sometimes hear people say about a proof
problem, "If it has an n, you can use induction." While this isn't always the
case, it is generally true that if you must prove a mathematical statement
for all positive integers, then you should consider trying induction.

Problem 11.12: Show that for any positive integer n, we have

l3 + 23 +33 + ∙∙∙ + π3 = (1+2 + 3+ ∙∙∙ + n)2.

Solution for Problem 11.12: That square of a huge expression is pretty intimidating. Fortunately, we have
already proved that
n(n + 1)
l + 2 + 3 + ∙∙∙ + n —
so the identity we wish to prove can be written as
n2(n + 1)2
13 + 23 + 33 + ∙ ∙ ∙ + n3 = (11.5)
4
First, we take care of the base case: 13 = (1)2(1 +1)2∕4, which is clearly true. Now, we assume the identity
holds for n = k, so we have
q q q q k2(k + 1)2
13 + 23 + 33 + ∙ ∙ ∙ + k3 = v - -.
4
Adding (k + 1)3 to both sides of this equation gives us

k2(k + 1)2 + (4k + 4)(fc + 1)2 __ (k2 + 4k + 4)(k + 1)2

4 ~ 4
(k + l)2(k + 2)2 _ (k + l)2((k + 1) + 1)2
4 - 4 -
This is our identity in Equation (11.5) for n = k + 1, so we have completed our induction. □

Important: For all positive integers n, we have

13 + 23 + 33 + ∙ ∙ ∙ + n3 = (1 + 2 + 3 + ∙ ∙ ∙ + n)2 = w2<n + 1)∖

4

So far, we have used induction on problems in which there is only one variable. We can also use
induction to prove identities involving multiple variables.

Problem 11.13: Prove that

m! (m + l)! (m + 2)! (m + n)! _ (m + n + 1)!
0! + 1! + 2! + + n! n↑(m + l)

for all nonnegative integers nι and n. (Note that 0! = 1.)

354
 11.3. INDUCTION

Solution for Problem 11.13: In this problem, we have our choice of which variable to use for our induction.
We'll focus on n, since there are fewer terms that involve n in the identity we wish to prove. First, we
tackle our base case, n = 0. When n = 0, we have ⅛ = m! on the left and = ml on the right, so
the identity holds when n = 0. (Notice that our base case works for any m, as well.)

WARNING!! Note that we must start this induction with a base case of 0, not 1,
* because we must prove the identity for all nonnegative integers, not
just all positive integers.

Next, we assume the identity holds for n = k, so we have

m! + (w + 1)! (m + 2)! (m + λr)! (m + k + l)!

0! 1! + 2! +"'+ kl ~ k∖(m + l')

(m + k + T) !
for all nonnegative integers m. To show that the identity holds for n = k + l, we add - to both
(* + 1)!
sides:

ml ι (m + l)! ι (m÷2)! (m + fc)! (m + k + 1)! _ (m + k + 1)! (m + k + T)l

0! 1! 21 +"'+ kl + (fc + l)! ~ fc!(m + l) + (k + l)!
_ (k + l)(m + k + 1)! (m + l)(zπ + k + 1)!
(k + l)k!(m + 1) + (m + l)(∕c + l)!
(k + l)(m + k + 1)! + (m + l)(m + k + 1)!
(k + l)!(m + 1)
(m + k + 2)(m + k + 1)!
(k + l)!(m + 1)
(m + k + 2)!
(k + l)!(m + 1)'

This is our identity for n = k + 1, so our induction is complete. Notice that every part of the induction
holds for all nonnegative values of m. So, even though we must prove this statement for all nonnegative
integers m and n, we can treat m as a constant and use induction on n. Because all of our steps are valid
for any value of m, our induction shows that for all nonnegative values of m, the identity holds for all
nonnegative values of n. □
We can also use induction on inequalities that we wish to prove for positive integers.

Problem 11.14: Show that if n is a positive integer greater than 2, then

1 1 1 1 3
n +1 n+2 n+3 2n 5’

Solution for Problem 11.14: What's wrong with this solution:

Bogus Solution: For our base case, we have n = 1, which gives us ⅛ on the left. This
is not greater than ∣, so the inequality is not true.

355
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

Read the question carefully: we have to prove this inequality for all integers n greater than 2. So, our
base case is n = 3z for which we have a left side of
1 1 1 _ 15 + 12 + 10 _ 37
4 5 6 60 60,
which is indeed greater than 3/5, because 3/5 = 36/60.
Our next step is to assume that the inequality holds for n = k,so we have
1 1 1 1^3
k~+l + k + 2 + k + 3 +"' + 2k> 5' (11.6)

We wish to use this to show that the inequality holds for n = k + 1; that is, we wish to show
1 1 1 1 . 3 (11.7)
Γf2 + k + 3 + fc + 4+ +2fc + 2>5’
To produce the left side of (11.7) from the left side of (11.6), we must add 2⅛γ + 5⅛2 - £+1 *o both sides
of (11.6). This gives us
111 1 3 1 1 1
k+2+k+3+ k + l + "'+ 2k +2 > 5 + 2fr+l + 2k + 2 k+1
3 1 1 2
~5 + 2k + l + 2k + 2 2k + 2
3 1 1
^ 5 + 2fc + l ^ 2k + 2
3 (2k + 2) - (2k + 1)
- 5 + (2k + l)(2fc + 2)
_ 3 1
^ 5 + (2k + l)(2k + 2) ’

1
Because is positive, we have
(2k + 2)(2fc + 1)
111 1 3 1
Jt + 2 + fc + 3 + Ar + 4+ '" + 2k + 2 > 5 + (2k + 2)(2k + 1)

so the inequality holds for n = k + 1. This completes our induction. □

00 ∕ 2n > j∣
Problem 11.15: Evaluate J 3-Σ2-t
n=l k=0 )

Solution for Problem 11.15: We begin by evaluating the inner sum, which is a geometric series with
common ratio 2_1:
1 _ 2-(2n+i)
1 -2"1
so
∞ 00

∏(3-(2-2^)] = ∏(l + 2-η.

H≡l «=1
Now what? We might be tempted to try this step:

356
 11.3. INDUCTION

Bogus Solution: We separate the product into two products:

∞ 00 oo
∏(1÷2-2>∏1 ∏2-.
n=l n=l n=l

The first product is just 1, so we focus on the second product.

But before we focus on that second product, we hopefully notice that we can't separate products of sums
like this. For example, it is not generally true that (1 + χ)(l + y)(l + z) = 1 ∙ 1 ∙ 1 + xyz. There are a bunch
more terms in the expansion of (1 + x)(l + y)(l + z).

So, back to the drawing board. To get a guess at what the infinite product is, let's try looking at the
product of the first few terms.

Concept: When faced with a complicated summation or product, work out the result
—ljlJ fθr small numbers of terms and look for a pattern.

The one-term product is easy: 1 + 2-2.

If we have two terms in the product, we get (1 + 2-2)(l + 2^4) = 1 + 2~2 + 2^4 + 2^6.
The product of the first three terms in the product is
(1 + 2"2)(1 + 2^4)(1 + 2^8) = (1 + 2^2 + 2~4 + 2^6)(1 + 2^8)
= 1(1 + 2~2 + 2^4 + 2^6) + 2-8(l + 2~2 + 2~4 + 2~6)
= 1 + 2^2 + 2^4 + 2^6 + 2^8 + 2^10 + 2^12 + 2^14.

Aha! Our experimentation suggests that our product equals the infinite geometric series with first
term 1 and ratio ∣. Moreover, our experimentation suggests a way to prove it. We used the product of
the first two terms to get the product of the first three terms. Similarly, we can use the product of the
first k terms to get the product of the first k + 1 terms. Yep—it's induction time! We wish to show that

∏(1 + 2^2'') = 1 + 2“2 + 2^4 + ∙ ∙ ∙ + 2-27+1÷2.

n=l

We've already taken care of the base case. For the inductive step, we assume that
m
J(l + 2^2") = 1 + 2^2 + 2^4 + ∙ ∙ ∙ + 2-2",+1+2.
n=l

Multiplying both sides of our inductive assumption by 1 + 2^2 "+1 gives us

m+1
](1 + 2-2") = (1 + 2-2"* ,)(l + 2-2 + 2-4 + ■ ■ ■ + 2-2",÷2)
n=l
= 1 + 2^2 + 2"4 + ∙ ∙ ∙ + 2-2",+1÷2 + 2-2'π+1(1 + 2-2 + 2-4 + ∙ ∙ ∙ + 2-2"'+1÷2)
= l + 2^2 + 2^4 + ∙ ∙ ■ + 2-2"'+1÷2 + 2"2'π+1 + 2-2'"+1-2 + 2-2"'+1-4 + ∙ ∙ ∙ + 2-2m+2÷2.

357
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

This completes our induction, thus showing that for all positive integers ∕, we have

∩(1 + 2^2") = 1 + 2"2 + 2^4 + ∙ ∙ ∙ + 2~2'+'+2.

n=l
oo

So, the infinite product j^J(l + 2 2") equals the infinite geometric series with first term 1 and common
κ=l
ratio ∣. The sum of this geometric series isl∕(l-∣) = ∣. □

Concept: Many problems, particularly those involving sequences or series, can be

solved with the strategy:

Experiment → See Pattern → Prove Pattern with Induction

txeιcises

11.3.1 Use induction to prove each of the following for all positive integers n:

(a) 1 + 4 + 7 + 10 + ■ ■ ■ + (3n - 2) = **

(b) 12 + 42 + 72 + ... + pn-2)2^(⅛"2-3"-D

(c) 22 +52 + 82 + ... + (3n-1)2^(6"2÷3"^)

11.3.2 The ∕cth triangular number is given by the formula Prove that the sum of the first n
triangular numbers is π<π+1×,L!⅜

11.3.3 Use induction to prove that for any complex number z, we have ∖zn∖ = ∣z∣π for all positive
integers n.
Ill 1 ∕∕
11.3.4 Use induction to show that —— + —— + —— + ∙ ∙ ∙ + —---- ——----- — = ------ - for all positive
1 ∙ 3 3-5 5 ∙ 7 (2n-1)(2∏ + 1) 2n + 1 r
integers n.
11 1
*
11.3.5 Show that if n is a positive integer, then 1 + — + — + ••• + —<2 √n.
√2 √3 √n

*
11.3.6 The sequence %ι, xz, X31... is defined by = 2 and ‰ι = x2 - x∣ς + 1 for all k ≥ 1. Find

Hints: 144

11.4 Binomial Theorem

If you are not familiar with combinations or the notation ('r'), you may wish to review them in our
Introduction to Counting & Probability text before studying this section.

358
 11.4. BINOMIAL THEOREM

The numbers (”) are sometimes called binomial coefficients for their relationship to the expansions
of powers of binomials. That relationship is:

Important: The Binomial Theorem states that

(* +'i')" = ∑(Λ∙W∙
i=0 ' '

Problem 11.17: Use the Binomial Theorem to expand each of the following:
(a) (x + y)3 (c) (2z - l)i
(b) (* + y)4 (d) (x2 - x),

Problem 11.18: Find the real solution to the equation x3 + 3x2 + 3x = 3. (Source: Mandelbrot)

Problem 11.19: In this problem, we find the remainder when 6520 is divided by 512.
(a) Consider the expansion of (64 +1)20, and notice that 512 = 29. Which terms in the expansion are
not divisible by 512?
(b) Find the desired remainder.

Problem 11.20: In the expansion of (αx + b)2000, where a and b are relatively prime positive integers
(meaning their greatest common divisor is l)z the coefficients of x2 and x3 are equal. Find a and b.
(Source: AIME)

359
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

Problem 11.21: Find the stun of all the roots, real and nonreal, of the equation

given that there are no multiple roots. (Source: AIME)

Problem 11.22:
(a) Show that 0 < λ∕3 - √2 < 1.
(b) Compute (√3 + √2)6 + (√3 - √2)6.
(c) What is the smallest integer larger than (V5 + V2)6? (Source: AHSME)

Solution for Problem 11.16:

(a) A little brute force goes a long way:

n∖ [ n ∖ _ nl n∖
r∕ + ∖r + 1/ r!(n - r)! + (r + l)!(n - r - 1)!
(r + l)rz! (n — f)nl
(r + l)r!(n - r)! (r + l)!(n - r)(n - r - 1)!
(r + l)n! + (n - r)κ!
(r + 1)!(m - r)!
(π + 1)∏! _ (n + 1)!
(r + l)!(n-r)! (r+l)!(n-r)!

This useful identity is known as Pascal's Identity.

(b) We must prove a statement for all positive integers n, so we try using induction. For our base
case, we must show (x + y)1 = (})x1 + (θ)ι∕1. Both sides of this equation equal x + yz so we have
established the base case. Next, we let n = k, and we assume that

360
 11.4. BINOMIAL THEOREM

Multiplying both sides by x + y gives us

*
(Δ),",1' -
(x + y)λ'+1 = (x + y) ∣xλ' +

j
= X l∕ + Jt-l
∖xk 1y + ∙ ∙ ∙ +

x∕ + ∣√ry + ∖xk 1y2 + ∙ ∙ ■ +

+ ∣∕+1
Since (fc) - (fc+∣) and (θ) = (fcθ1) because all four of these expressions equal 1, we have

(x+y)λ'+1

For all the other terms, we apply Pascal's Identity from part (a). For example, we have the
equations (jfc*1) + (J) = (fc÷1), and (kk2) + (kk1) = (J÷J), and so on down to (J) + (J) = (ψ), which gives
us the desired

u+= (‘ ÷ J)√÷1+f ÷ 1)√y + (* ÷ })√-y +...+f ÷ 1)xv-> + f ÷ 1)y + f ÷ 1)y-,

This completes our induction.
□
As shown at the beginning of this section, we can write the Binomial Theorem more succinctly with
summation notation:

(x + y)” = |x” + lx” 2y2 + ∙ ∙ ∙ + ∣x"-⅛≈.

There are other ways we can express the Binomial Theorem. For example, because

π! π!
(π - i)!(n -(n- i))! (n - z)!z!

we can write it as

Let's put the Binomial Theorem to work.

Problem 11.17: Expand each of the following:

(a) (* + y)3 (c) (2z - 1)5
(b) (x + y)4 (d) (x2 - x)4

361
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

Solution for Problem 11.17:

(a) (χ + y)3 = (i>3 + (2)^2y + (J)χy2 + (0)y3 = χ3 + 3χ2y + 3χy2 + y3∙
Φ) (χ+y)4 = φχ4 + (t)χ3y + (4)χ2y2 + (1)χy3 + (o)y4 =χ4 + 4χ3y + 6χ2y2+4χy3 + y4∙
(c) We have
(2z - 1)5 = Q(2z)5 + ∩(2z)4(-l) + ^(2z)3(-l)2 + ^(2z)2(-l)3 + ^(2z)(-l)4 + (θj(-l)5

= 32z5 - 80z4 + 80z3 - 40z2 + 10z - 1.

(d) (x2 - .γ∣4 = (4)(z2)4 + (4)(x2)3(-x) + (4)(z2)2(-x)2 + φ(z2)(-z)3 + φ(-z)4 = z8 - 4z7 + 6x6 - 4x5 + x4.

Notice that we're careful to keep our signs correct in parts (c) and (d). □
For the rest of this section, we apply the Binomial Theorem to more challenging algebra problems.

Problem 11.18: Find the real solution to the equation x3 + 3x2 + 3x = 3. (Source: Mandelbrot)
J
Solution for Problem 11.18: The left side looks a lot like the expansion of (x +1)3, which is x3 + 3x2 + 3x +1.
So, we add 1 to both sides of the equation, which gives us x3 + 3x2 + 3x + 1 = 4, so we have
(x + 1)3 = 4.
Taking the cube root of both sides of the equation, we get x + 1 = "∖∕4, so x = ^4-l is the real root.
(There are two nonreal numbers whose cube is 4; see if you can find them, and thereby find the other
two solutions to the equation x3 + 3x2 + 3x = 3.) □
The key to our solution to Problem 11.18 was recognizing the resemblance between x3 + 3x2 + 3x and
the expansion of (x + 1)3.

Concept: Keep an eye out for expressions that are expansions of powers of binomials,
Q 'lSjg or are very similar to expansions of powers of binomials.

Problem 11.19: Find the remainder when 6520 is divided by 512.

Solution for Problem 11.19: We surely don't want to evaluate 6520. We notice that 512 is an integer power
of 2, namely 29 = 512. Next, we see that 65 is very close to 64, which equals 26. So, we consider the
binomial expansion of (64 + l)20:
(64 + 1)20 = ^°)e420 + (^6419(1) + ... + ^(642)(1m) + ^j(641)(l1,) + Ql“

Because 642 = (26)2 = 212, we see that all but the last two terms of the expansion are divisible by 2q.
Therefore, the remainder when 6520 is divided by 512 equals the remainder when the sum of these last
two terms is divided by 512. The sum of these last two terms is
(20)(64) + 1 = 1280 + 1 = 1281.
Since 1281 divided by 512 is 2 with a remainder of 257, our answer is 257. □

362
 11.4. BINOMIAL THEOREM

Concept: Whenever faced with a problem that involves large powers of expressions
Q—pO or numbers, consider looking for a way to apply the Binomial Theorem.

Problem 11.20: In the expansion of (ax + b)2000, where a and b are relatively prime positive integers,
the coefficients of x*12 and x3*are equal. Find a and b. (Source: AIME)

Solution for Problem 11.20: Using the Binomial Theorem, the first few terms of (ax + I?)2000 are

⅛l''n ■ ax + *'"8 ∙ a2x2 + ∣f,1w7.03x3 + ...

Since (∞°) = (20200) = ≡≡S and (≡) = (2∞0) = coomms ™ setttag the coefficients of χ2 and χ3 equal
gives us
2000 ∙ 1999 71998 2 2000 ∙ 1999 ∙ 1998,1997 3
------ ------- bιwβαz =----------- ----------- bιwa,
2 6
which simplifies to 3b1998a2 = 1998b1"7α3. Dividing both sides by 3fl2b1"7 gives b = 666a, so a divides b
evenly. Since the greatest common divisor of a and b is 1 and a divides br we must have a = 1, so b = 666.
□
Problem 11.21: Find the sum of all the roots, real and nonreal, of the equation

given that there are no multiple roots. (Source: AIME)

Solution for Problem 11.21: Solution 1: We can use Vieta's Formulas to find the sum of the roots of a
∕1 ∖2001
polynomial, so we expand 1j - x] with the Binomial Theorem:

■ + ,2000 + (_x)2001

Since ® = (2°°1) = 2001 and (≡) = ≡≡, we have

1 _ rf°1 1 2001 τ + .. - 2001'2≡ ι⅜99 + 2001 2∞ _ γ200ι

2 ) 220°1 22oo° 2 ∙ 22 2

Adding x2001 gives us

1 2001 _ 2001 ∙ 2000 1999 2001 2000

x2001 + 22001 22000 2∙22 2
2001-2000
By Vieta's Formulas, the sum of the roots is —— = 500.
~

363
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

Solution 2: We focus on the expressions that are raised to the same power, x and (∣ - x). Noting
that the average of x and ∣ - x is ∣ allows us to make a clever substitution that simplifies the problem.
Letting y = x - ∣, we have the equation

The expansions of (⅜ + y) and Q - y) are the same except that the signs of the terms with
odd exponents are opposites. When the expansions are added, these terms with odd exponents cancel,
leaving a degree 2000 polynomial that has only terms with even powers of y. Now when we apply
Vieta's Formulas, the coefficient of y19" is 0, which means that the sum of the 2000 solutions for y is 0.
Since each value of y is 1/4 less than a solution for x in the original equation, the sum of the values of x
that satisfy the original equation is 2000(∣) = 500. □
Our second solution is a clever use of symmetry that makes applications of the Binomial Theorem
and Vieta's Formulas much easier. In this solution, we make use of the fact that the terms with odd
powers of b in the expansions of (α + b)n and (« — b)n cancel each other out. This can be a very useful
problem solving strategy.

, Problem 11.22: What is the smallest integer larger than (V3 + V2)6? (Source: AHSME)
Solution for Problem 11.22: When we expand (V3 + V2) , we have terms with square roots still in them.
We could work hard to estimate them, but, inspired by our second solution to the previous problem,
we can also use a related binomial expansion to help us. Note that 0 < V3 - V2 < 1, so (√3 - √2) < 1.
Now, we note that

(vs w=0 «+(') (<«+(3««+(3 i(⅛ «

÷(3(<(<÷(3^),^)5÷(o)(<∙

(√3 - √2)6 = (J (√3)6 - g(√3)5 (√2)1 + Q(√3)4 (√2)2 - Q« ∣ <

Adding these expansions rids us of those pesky square roots, leaving an integer to work with:

(√3 + √2)6 + (√3 - √2)6 = 2 (33) + 30 (32) (21) + 30 (31) (22) + 2 (23)

= 54 + 540 + 360 + 16 = 970.

So, wehave(√3 + √2)6 = 970-(V3 - V2)6. Since 0 < (√3 - >/2^ < 1, we see that 969 < (√3 + V2)6 < 970,

which means the smallest integer greater than (√3 + V2j is 970. □

364
 11.5. SUMMARY

*
Exercise

11.4.1 Expand each of the following:

(a) (l-x)5 (b) (2x2 + 3y)4
∖9
(χ + .

11.4.3

(a)Without using a calculator, show that 3119 + 33" is divisible by 64.

(b)
* Without using a calculator, show that 3119 + 33" is not divisible by 128.

(a) Calculate (100) + (1ι0j÷(120) + ... + (l°),

V, ∕99∖
*
(b) Find y Hints: 326

11.4.5 Let n be a positive integer and x ≥ 0. Show that (1 + x)n ≥ 1 + nx.

11.4.6
* Let P be the product of the nonreal roots of x4 - 4x3 + βx2 — 4x = 2005. Find [PJ, where [xj
stands for the greatest integer that is less than or equal to x. (Source: AIME)

11.5 Summary

An identity is an equation that is true for all values of all variables for which the equation is defined.
In this chapter, we discussed a few useful identities and covered several methods for proving specific
identities.
First, we discussed brute force, which is essentially expanding both sides of a proposed identity to
see if the two sides match.
Next, we worked with ratios. We proved a particularly useful pair of ratio manipulations:

Important: Suppose a/b = c∕d. Then, for any constant k for which b + kd ≠ 0, we
I 2 have
∖J∕ a _ c _ a + kc
b d b + kd'
For any constant k for which a- kb and c -kd are nonzero, we have

a + kb _ c + kd
a-kb c-kd'

We then moved on to an extremely powerful tool, mathematical induction. While we primarily

used induction to prove identities in this chapter, you will see it in many more contexts throughout your

365
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

study of mathematics. To use induction to prove an identity holds for all positive integers n, we must
perform two steps:

1. The base case. Show that the identity is true for n = 1. This is usually pretty easy.

2. The inductive step. Show that if the identity is true for n = k, then the identity holds for n - k + 1.

We used induction to prove several identities, including:

Important: The sum of the first n positive integers is n(n + 1)∕2.

ψ
Important: For all positive integers n, we have

ψ „ , , , ~
13 + 23 + 33 + ∙ ∙ ∙ + n3 = (1 + 2 + 3 + ∙ ∙ ∙ + n)2 =
n2(n +1)2
— ■

Finally, we used induction to prove the Binomial Theorem:

Important: The Binomial Theorem states that

Problem Solving Strategies________________

Concepts: ∙ When performing complicated algebra, stay organized and write

Q≡W neatly.

• We can often relate symmetric expressions with identities.

• If it's not immediately obvious how to solve a problem, don't just sit
and stare at it. Experiment!

• When asked to evaluate an expression involving large numbers, you

can get a feel for the problem by evaluating similar expressions with
small numbers.

• Because an identity must be true for all permissible values of the

variables in the identity, we can test an identity we develop by sub­
stituting convenient values for the variables.

Continued on the next page. . .

366
 REVIEW PROBLEMS

Concepts: . . . continued from the previous page

O _ ______ :--------------------1
ta∏

• When given equal ratios, setting the ratios equal to a variable helps
eliminate the fractions.

• Working backwards from a relationship you wish to prove is a pow­

erful problem-solving method.

• If you must prove a mathematical statement for all positive integers,

then you should probably consider trying induction.

• When faced with a complicated summation or product, work out the

result for small numbers of terms and look for a pattern.

• Many problems, particularly those involving sequences or series, can

be solved with the strategy:

Experiment → See Pattern → Prove Pattern with Induction

• Keep an eye out for expressions that are expansions of powers of

binomials, or are very similar to expansions of powers of binomials.

• Whenever faced with a problem that involves large powers of expres­

sions or numbers, consider looking for a way to apply the Binomial
Theorem.

11.23 In each of the following parts, state whether or not the equation is an identity. If it is an identity,
prove that it is an identity. If it is not an identity, find values of the variables for which the equation is
not true.

(a) a(b — c) + b(c - a) + c(a - b) = 0 (c) , where b ≠ 0 and b ≠ 1

b b- y∕b'
(b) (x - y)4 + (x + y)4 = 2(x2 + y2)2 + 8x2y2 (d) *2(* - y) + y2(y - χ) = χ3 + y3

11.24 Find x + y + z if— + = 12 and xyz=-.

j xy yz zx 18

11.25

c
11.26 a
Suppose - = -c , where a, b,c,d,e, f > 0. Show that - =

p + q-r u+v
11.27 Let----- ----- . Prove that pv = u(q - r).
-q+
P^<7 + rr u—v

367
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

(IX cι — X C) —1∕
11.28 Suppose - = —, where a,brx,y > 0. Show that = .
V ' ya2 + x2 Λjb2 + y2
11.29 Use induction to prove each of the following for all positive integers n:
r, ~ n(n + l)(2n + 1)
(a) 1 + 3 + 5 + ∙ ∙ ∙ + (2π - 1) = n2 (b) 1 + 4 + 9 + ∙ ∙ ∙ + n2 = —------- 7-------- -
o

11.30 Show that 77 + 77 + 77 + ∣7 + ∙∙∙ + -—- = 1 - -7.

1! 2! 3! 4! n'. n'.
11.31 Let «2/ «3,.. ∙, an and bγ, b2, b3,..., bn be two sequences of numbers. Prove the identity

a↑bγ + a2b2 + a3b3 + ∙ ∙ ∙ + anbn = (a↑ - a2)bγ + («2 - fl3)(^1 + ⅛) + («3 - fl4)(b1+b2 + b3) + ■■■
+ (an_i -απ)(bι +b2 + b3 + ■ ■■ + bn-↑) + an(bι + b2 + b3 + ∙∙∙ + bn),

where n > 1. This identity is sometimes referred to as the Abel Summation.

11.32 Find the sum of the digits of the number 20034 — 19974. (Source: Mandelbrot)
11.33 State and prove a one-variable identity that is suggested by the examples below:

1+2 = 3,
4 + 5 + 6 = 7 + 8,
9 + 10 + 11 + 12 = 13 + 14 + 15,
16 + 17 + 18 + 19 + 20 = 21 + 22 + 23 + 24.

11.34 Show that for all x, y, and z, we have

x2(2x2 -y2 -z2) + yλ(2y1 - x2 - z2) + z2(2z2 - x2 - y2) = (x- y)2(x + y)2 + (y - z)2(y + z)2 + (z - x)2(z + x)2.

11.35 Find the coefficient of x14 in the expansion of (2x3 - x)8.

23 - 1 33 - 1 43 - 1 n3 - 1 2(π2 + n + 1) r
11.36 Show that 77l ■ 77γ ■ ....... = 77777 for any positive integer n > 2.

11.37
(a) Verify that for all x ≠ 0, we have x"+1 + = (x + |) (x" + - (x"~1 + 3Γι)∙

1 1
(b) ⅛ Let x + 7 = a, for some integer a. Prove that x" + — is an integer for all n ≥ 0. Hints: 241
χ X
11.38 Without using a calculator, prove that 2005 ∙ 20073 - 2006 ∙ 20043 is the cube of an integer .

11.39 In Problem 11.4.5, we showed that (1 + x)" ≥ 1 + nx if x > 0 and n is a positive integer. In this
problem, we show that this inequality still holds for all positive integers n even if -1 ≤ x < 0.

(a) Explain why we cannot use the exact same Binomial Theorem approach we used in Problem 11.4.5.
(b) What's another tool we can use to prove a statement “for all positive integers n"? Prove that
(1 + x)" ≥ 1 + hx for all positive integers ιι even if -1 < x < 0.

368
 CHALLENGE PROBLEMS

11.40 Find x if (x2 - 3x + 4)∕(x2 - 5x + 5) = (x2 - 3x + 9)∕(x2 - 5x + 10).

11.41 Let a, b, c, d, and e be real numbers such that 7 = - = 5 = - = -. Find the value of ⅛-.
b c d e a c2d

Challenge Problems

11.42 Find all real values x that satisfy x4 - 4x3 + 5x2 - 4x + 1 = 0.

11.43 Prove that nn > (κ + 1)" 1 for all positive integers n. Hints: 175,195
χ y a x3 + y3 _ x3 - xyλ
11.44 Let ------ = ——, where x ≠ y and b ≠ 0. Show that χ + y = b and a2 + 3fc2 = 2b(fl + ⅛)'
a+b a-b j x-y
Hints: 305

11.45 Find the units digit of the decimal expansion of (15 + √22θ)19 + (15 + V22θ)82. (Source: AHSME)
Hints: 64

11.46 State and prove a one-variable identity that is suggested by the examples below:

22 + 32 + 42 + 142 = 152,
42 + 52 + 62 + 382 = 392,
62 + 72 + 82 + 742 = 752,
82 + 92 + 102 + 1222 = 1232.

Hints: 214

(π∖ = —-----
h!
—, we can write the Binomial Theorem as
z∕ ιl(n - z)!
(χ + y)" = ∑ ⅛χiyy,
0≤i,j≤n ''
i+j=n

where the 0 ≤ i,j ≤n and i + j = n under the summation means, "Sum the expression for all possible i
and j such that i and j sum to n, and both i and j are nonnegative integers no greater than n." Express
the expansion of (x + y + z)9 using a similar summation notation.
11.48 Prove that n4 ≤ 4n for all positive integers n greater than 3.
11.49 Let (1 + x + x2)" = «0 + «1X + β2X2 ----- + «2nX2" for all X∙ Find flo + ∏2 + «4 + ∙ ∙ ∙ + am in terms of n.
(Source: AHSME) Hints: 129
11.50 Show that if a}X + b↑y + c↑z = a½x + b2y + C2Z = 0, then

x _ y _ z
b∖C2 - b2Cι C∖a2-C2a∖ a∖b2-a2b∖
for all values of the variables for which the three fractions are defined. Hints: 120

369
CHAPTER 11. IDENTITIES, MANIPULATIONS, AND INDUCTION

11.51 Prove that, for each positive integer n, we have

1 1_ 1 1 _ 1 ∣ 1 ∣ ∣ 1
2 + 3 + 2n -1 2n n+1 n+2 2n'
Hints: 118
11.52 Find all ordered pairs of real numbers (x, y) such that x3 - y3 - 19 and x2y - xy2 = 6.

11.53 The polynomial 1 - χ + χλ - χ3 + ∙ ∙ ∙ + χ16 - x17 may be written in the form

«0 + aγy + ¾y2 ÷ z⅛y3 + ∙ ∙ ∙ + fli6y16 ÷ fl17y^,

where y = x + 1 and the fl/s are constants. Find the value of «2- (Source: AIME) Hints: 192
11.54 Factor (x + y)5 - (χ5 + y5) completely into a product of polynomials with real coefficients.
11.55 Find the smallest positive integer n for which the expansion of (xy — 3x + 7y — 21)”, after like
terms have been collected, has at least 1996 terms. (Source: AIME) Hints: 182
bc(b + c) ca(c + α) ι ab(a + b~)
11.56 Prove that ----- ——— + (& - c)(b - a) + (c - fl)(c - &) — fl + b + c.
(a - b)(a - c)
11.57 The Binomial Theorem is valid for exponents that are not integers. That is, for all real numbers
x, y, and r with ∣x∣ > ∣y∣, we have
(x ÷ vγ=3f ÷ r√->y ÷ ⅞⅛-y + ÷....

What are the first three digits to the right of the decimal point in the decimal representation of the
number (1O2002 + l)10/7? (Source: AIME) Hints: 310

11.58
* Let n be a positive integer. Prove that----- - +----- - + ∙ ∙ ∙ + — ≥ - ----- -.
r o n+1 n+2 2n 3n + l
11.59
* Prove that if
b2 + c2 — fl2 fl2 + c2 — b2 a2 + fe2 — c2
A = 2bc ' = 2∑c ' C= 2αb '
and A + B + C = 1, then ABC = -1. Hints: 275
11.60
* Write (x2 - yz)3 + (y2 - zx)3 + (z2 - xy)3 - 3(x2 - yz)(y2 - zx)(z2 - xy) as the square of a polynomial.
Hints: 100
« ∣
11.61
* Prove that for all positive integers n, we have V — < 2. Hints: 10

*
11.62
(a) Find a polynomial f(a, b, c, d) in four variables such that
∕(χ1,χ2,χ3,χ4)∕(y1,y2,y3,y4) = (*
ιyι + ^2y2 ÷ χ3y3 + χ4y4)2 + (χ1y2-χ2y1 +χ3y4 -χ4y3)2
+ (Xly3 - X3yl + X4y2 - X2y4)2 + (Xly4 - X4yl + X2y3 - X3y2)2

for all values of x↑, X2, X3, X4, yι, y2, yι> 1∕4∙ Hints: 102
(b) Every prime number can be written as the sum of four squares. Combine this fact with part (a) to
prove that every positive integer can be written as the sum of four perfect squares.

370
Nature is unfair? So much the better. Inequality is the only bearable thing; the monotony of equality can only
lead us to boredom. - Francis Picabia

CHAPTER 12 ι
I Inequalities

So far in this book, much of our work has involved solving equations or inequalities by finding those
specific values of variables that make these equations or inequalities true. We have also proved some
equations are always true; we called these identities. In this chapter, we turn to proving that certain
inequalities are always true.

12.1 Manipulating Inequalities

We'll sometimes impose certain conditions on the variables in an inequality we wish to prove, and we'll
occasionally use a shorthand for writing some of these conditions. For example, rather than writing,
"a > 0, & > 0, and c > 0," we can write "a, b,c > 0." Similarly, we can write 0 < w, x, y, z < 1 to indicate
that w, x, y, and z are all between 0 and 1.
∣∣∣∣ ----
Problems
111 ■■■■■■»■■■■■

Problem 12.1: Suppose that > ∣ a∏d we wish to prove that —- > -.
Cl c
(a) Suppose we start with, "We multiply - > - by bd to get ad > be." Wiry is this invalid?

n . „ , a+b c+d
(b) Prove that —— > ——.
b d

Problem 12.2: hi this problem, we show that if a > b and c≥ d, then ac + bd ≥ ad + be.
(a) Move all the terms of ac + bd ≥ ad + be to one side and factor the result.
(b) Explain why the inequality you produced in part (a) must be true if a ≥ b and c≥ d.
(c) In parts (a) and (b), we worked backwards to figure out why ac + bd≥ ad + be. Write a complete
solution forwards, starting with a ≥ b and c≥ d and ending with ac + bd ≥ ad + be.

371
CHAPTER 12. INEQUALITIES

Problem 12.3:
(a) Find some positive numbers a, b, c, and d such that j < § < 1.
(b) Arrange the quantities and 1 from greatest to least according to your values from part (a).
(c) Show that the ordering found in part (b) is the same for all positive numbers a, b, c, and d.

445721 445725
Problem 12.4: In this problem, we will determine which is greater, g9Q2923 or 9902927'

(a) Prove that if b > a > 0 and x > 0, then 7 <

b b+x
nx 445721 445725
(b) Without using a calculator or computer, determine which is greater, 99Q2923 °r 9902927’

Problem 12.5: Let 0 < x, y < c. Prove that

ab cd
Problem 12.6: Show that if 0 < a < c and 0 <b < d, then <
cι + b c+d

a+b
Problem 12.1: Suppose that - > -. Show that >
~~b~

Solution for Problem 12.1: What's wrong with this solution:

Bogus Solution: Multiplying both sides of f> 2 by bd gives us ad > be. Adding bd
to both sides gives us ad + bd > be + bd. Factoring both sides gives
d(a + b) > b(c + d). Dividing both sides by b gives ⅛(a + b) > c + d,
and dividing both sides by d gives the desired > .

Our first step, multiplying ∣ > j by bd to get ad > be, is only valid if bd is positive. But we aren't
guaranteed that this is the case; if bd is negative, then multiplying f > ⅛ by bd produces ad < be. Each of
our divisions have similar difficulties, as well.

WARNING!! Pay special attention to whether or not the variables can be negative
*a4 in inequalities you wish to prove. Some common inequality manipu­
lations are only valid if the variables are positive.

We could fix our Bogus Solution with some casework, but instead, let's look for manipulations that
don't depend on the signs of the variables.
Comparing the left side of the inequality we are given, a∕b, to the left side of the inequality we want
to prove, (fl + b)∕b, we see a very simple relationship:

372
________________________________________________ 12.1. MANIPULATING INEQUALITIES

So, we add 1 to both sides of 5 > ⅛, which gives us £ + 1 > j + 1. Writing each side of the inequality
with a common denominator gives the desired > ej-p. □

Concept: Look for ways to manipulate one side of a given inequality to match the
θ—3Q corresponding side of the desired inequality. Sometimes simple manipu
lations will be enough to prove the desired inequality.

Problem 12.2: Show that if a ≥ b and c ≥d, then ac + bd ≥ad + be.

Solution for Problem 12.2: What's wrong with starting as follows:

Bogus Solution: We need an ac and a bd, so we start by multiplying a ≥ b by c ≥ d to

get ac ≥ bd.

This is bad for two reasons. First, it doesn't get us any closer to what we want to prove, since ac > bd
has bd on the lesser side, but the desired inequality has bd on the greater side. Even subtracting bd from
both sides doesn't help, since that gives ac-bd≥ 0, which is quite different from the ac + bd we want on
the larger side.

Concept: Keep your eye on the ball! Don't just blindly manipulate inequalities—try
Q lj∣J tθ make sure you're moving in the right direction.

Even worse than the fact that we're moving in the wrong direction is the fact that the step "multiplying
a ≥ bbyc ≥ d gives ac ≥ bd" is just plain invalid. What if a — 2, b = —1, c = —3, and d — —4? Then, we
have ac < bd, not ac ≥ bd.

WARNING!! If a, b, c, and d can be any real numbers, then we cannot multiply

V a ≥ b by c ≥ d to get ac ≥ bd. If a, b, c, and d are positive, then this
manipulation is valid.

So, back to the drawing board. We might note that we need ac on the larger side and be on the
smaller, so we could try multiplying a > bby c to give ac > be. But that's invalid too! Since c might be
negative, we can't say that ac ≥ be. Now, we seem stuck. There are two ways we might proceed from
here:
Solution 1: Focus on why we are stuck. We are stuck because every time we try multiplying inequalities,
we run into the problem of possibly multiplying by a negative number. So, we try to find a way to
multiply inequalities such that we are guaranteed to multiply by a nonnegative number.

Concept: When stuck on a problem, focus on why you are stuck, and try to find
Q—pD ways to remove that obstacle.

373
CHAPTER 12. INEQUALITIES

We can turn our given inequalities into ones involving nonnegative expressions by subtracting b from
a ≥ b and subtracting d from c ≥ d. This gives us a - b ≥ 0 and c - d ≥ 0. Aha! Nonnegative numbers.
The product of nonnegative expressions a-b and c - d must be nonnegative, and it will produce all the
products in our desired inequality. We therefore have (a - b){c - d) ≥ 0. Expanding the left side gives
ac — ad - be + bd ≥ 0, and rearranging gives the desired ac + bd ≥ ad + be.
Solution 2: Work backwards. We're stuck going forwards, so we try manipulating the desired inequality
to produce an inequality we know how to prove.

Concept: Working backwards is a very powerful method for proving inequalities.

O≡≡sg

We start by moving everything to one side, which gives us ac + bd — ad — be ≥ 0. We see pairs of terms
we can factor, so we try factoring, which gives us c(a -b) + d(b -a)≥0. Aha! We can factor some more.
We have
c(a -b) + d(b -a) = c(a -b) - d(a -b) = (c - d)(a - b),
so our desired inequality is now (c - d)(a - b) ≥ 0. Now, we see the path to a solution. The given
inequalities a ≥ b and c ≥ d give us a - b ≥ 0 and c - d > 0, and multiplying these gives us (c - d~)(a -b)≥0.
This second solution is a combination of working backwards and forwards. As usual when we find
a solution wholly or partially by working backwards, we finish by writing the whole solution forwards:

Proof: Since a ≥ b and c ≥ d, we have a-b ≥ 0 and c - d ≥ 0. Therefore, we have (a — b)(c - d) ≥ 0.

0 Expanding the product on the left gives ac - ad — be + bd, and rearranging gives the desired
ac + bd ≥ ad + be.

Problem 12.3: Let a, b, c, and d be positive numbers such that f < ⅜ < 1∙ Arrange the four quantities
o' c' λc' and *1 from greatest to least, and show that the ordering is the same for all positive numbers
a, b, c, and d.

Solution for Problem 12.3: Before diving into manipulating the given inequalities, we might simply
choose some positive numbers that fit the description of the problem. Since f and ⅞ must be less than
1, we must choose a < b and c < d, and we must also choose the numbers so that f <⅛∙ Letting a = 1,
b = 2,c = 3, and d = 4 works just fine.

Concept: When possible, testing numbers that fit an inequality problem can help
o≈≡ you get started with a solution.

For our choice of values for the variables, we have = 2, = ∣, and = f∕so the order from greatest
to least is ~τ, ~, 7,1.
ac' a' c'

We now have a target. We'd like to prove that

bd b d
— > -> - >1.
ac a c

374
 12.1. MANIPULATING INEQUALITIES

We see that the last three terms in this inequality chain are reciprocals of the terms in the given inequality,
so we wonder what happens when we take the reciprocal of both sides of an inequality.
If x and ι∕ are positive and x ≥ y, then dividing both sides by xy gives j ≥ ½∙ So, when we take the
reciprocal of both sides of an inequality (and both sides are positive), we reverse the direction of the
inequality sign.

Applying this fact to our given inequality, ∣ < ⅞ < 1, gives us

b
- >
a
So, all we have left to prove is that ^^ > -^∙ We note that the left side of this inequality is times the right
side. But we already know that > 1. Multiplying this by gives us > ∣. Combining this with our
earlier inequality chain gives us
bd b d ^
— > - > - > 1,
ac a c
as desired. □

Not all problems that involve inequality proofs say "Prove this inequality."

445721
Problem 12.4: Without using a calculator or computer, determine which is greater, 9902923 or 1
445725
9902927’

Solution for Problem 12.4: We're certainly not interested in performing the long division. So, we look
for clues in the fractions that will help us find a slicker solution. We notice that the numerator and
denominator in the second fraction are each 4 larger than in the first fraction. So, we can recast our
problem as comparing
a α+4
to 1 λ'
b b+4
where a = 445721 and b = 9902923. Working with a and b will probably be a lot easier than working
with those huge numbers.

Concept: If a problem is given in terms of gigantic numbers, it's often useful to

O≡≡^ replace those numbers with variables.

In fact, we could even go further and replace 4 with x, so that we are comparing
a a+x
τ t0 I—∙
b b+x
We can experiment to get a sense for which is larger. For example, if we let a = 1, b = 2, and x = 3, we
have a∣b = 1/2 and (α + x)∕(b + x) = 4/5. So, we think we have
a+x
b+x

375
CHAPTER 12. INEQUALITIES

We're dealing with positive numbers in the original problem, so we let a,b,x > 0. It's not clear how to
get from a, b, x > 0 to a/b < (a + x)∕(b + x), so we work backwards. Multiplying both sides of the desired
inequality by b(b + x) to get rid of the fractions gives

ab + ax < ab + bx.

Rearranging this inequality gives us (α - b)x < 0. This inequality is only true if b > a. However, for the
two numbers we wish to compare, we do indeed have b > a.
So, now we're ready to prove an inequality that we can use to solve our problem. If b > a > 0 and
x > 0, then we have (α - V)x < 0, so ax < bx. Adding ab to both sides and factoring gives a(b + x) < b(a + x).
Dividing both sides by b(b + x) gives
a a+x
b < b + x'
445721 445725
Letting a = 445721, b = 9902923, and x = 4 gives us
9902923 9902927'

Problem 12.5: Let 0 < x, y ≤ c. Prove that

Solution for Problem 12.5: We don't have any clear starting point for working forwards from 0 ≤ x,y ≤ c
to the desired inequality, so we'll have to try working backwards.
We'd like to start by getting rid of all the ugly fractions in the desired inequality. Multiplying both
sides by the denominator of the left side would be easiest, though we must first check that this expression
is positive. Because x, y, and c are positive, the expression 1 + is positive, too. Therefore, multiplying
both sides of the desired inequality by 1 + does not change the direction of the inequality. This gives
us χy
x + y ≤c+—.
j c
Similarly, since c > 0, we can multiply both sides of this inequality by c to get

(x + y)c < c2 + xy.

That's much nicer than the original inequality. We move all terms to the left side, which gives

-c2 + (x + y)c - xy ≤ 0.

We can view the left side as a quadratic in c. Multiplying the inequality by -1 (and reversing the
direction of the inequality sign) gives

c2 - (x + y)c + xy ≥ 0,

which makes the quadratic, and its factorization, much clearer. We now factor the left side, and we have

(c - x)(c - y) ≥ 0

as the inequality we wish to prove. Since c — x > 0 and c - y ≥ 0, this inequality is true.
Now, we can walk backwards through our steps to write our proof:

376
 12.1. MANIPULATING INEQUALITIES

Proof: Since x,y <c, we have c x > 0 and c - y ≥ 0. Multiplying these two gives (c - x)(c - y) > 0.
0 Expanding the left side and rearranging gives c2 + xy ≥ c(x + y). Since c > 0, dividing both
sides by c gives
xy
c + — ≥ x + y.
c v
Factoring c out of the left side gives

c ≥ x + y.

Since x, y, and c are positive, the expression 1 + →r is positive. So, we can divide both sides by
this expression to give the desired
x+y
c >-----
~l + x-½
cz

Problem 12.6: Show that if 0 < a < c and 0 < b < d, then <
a+b c+d

Solution for Problem 12.6: We're definitely going to try working backwards.
Solution 1: Because a + b and c + d are positive, we can multiply both sides of the desired inequality
by (α + b)(c + d) to produce the equivalent inequality
abc + abd < acd + bed.

We can produce some of these terms from our given inequalities. Multiplying b < d by ac gives
abc < acd and multiplying a < cb>ybd gives abd < bed. Adding abc < acd and abd < bed gives
abc + abd < acd + bed.
We know how to finish from here; we retrace our initial "working backwards" steps. Factoring gives
ab(c + d) < cd(a + b) and dividing both sides by (a + b)(c + dy) gives < ^⅛∙

Solution 2: Multiplying by denominators is not the only way to deal with fractions. Sometimes,
reciprocals of fractions are much nicer than the original fractions. (This is particularly true when the
denominator of a fraction is a sum and the numerator is a product.)

Concept: Dealing with reciprocals of complicated fractions is often nicer than dealing
Q~~uP with the original fractions.

Taking the reciprocal of both sides of the desired inequality (and therefore reversing the direction of the
inequality) gives us
a+b c+d
ab cd
Splitting both sides into the sum of two fractions, we have
1111
7 + ~ > 7 + -∙
b a d c

377
CHAPTER 12. INEQUALITIES

Because we have 0 < a < c, we have ∣ > ∣. Similarly, 0 < b < d gives ⅛ > ⅛∙ Adding these reciprocal
inequalities puts us on track to walk backwards through our steps above:

Proof: Because we have 0 < a < c and 0 < b < d, we have ∣ > ⅜ and M∙ Adding these gives
0
1111
- + - > - + -.
b a d c
Writing each side with a common denominator, we have

a + b > c+_d
ab cd
Because a, b,c,d> 0, we can take the reciprocal of both sides of this inequality to give

ab cd
a+b c + d'

WARNING!! Many of the manipulations we performed in our solution to Prob-

λλ lem 12.6 are valid only because the variables are all positive.

Exerciser

12.1.1 Let x > y > 0 and z ≠ 0. In each of the parts below, determine whether or not the inequality
always holds. If it does, prove that it does. If it doesn't, provide an example in which it fails.
(a) x+z > z+y (c) x - z > y - z
(b) (x + y)z > 0 (d) xz > yz

12.1.2 Let -1 < x < 0. Arrange x, x2, and x3 inorder from least to greatest.
12.1.3 Show that if 0 < x ≤ y≤ c, then y - x ≤ c (1 -

12.1.4 Suppose that a, b, c, and d are positive real numbers such that γ ≤-1. Show that —≤ —.
rr b d a+b c+d

12.1.5
∙y 7 b+y
f + *.
Show that if a, b, x, y are constants such that a > b > 0, and x > y > 0, then γ + - >
b+x
*
12.1.6 Let 0 ≤ a <b < c and a + b + c = 1. Show that a2 + 3b2 + 5c2 ≥ 1. Hints: 244

12.2 The Trivial Inequality

Already throughout this book, we have used the fact that the square of any real number is nonnegative.
This fact is so simple, you might even consider it trivial.

378
 12.2. THE TRIVIAL INEQUALITY

Important: The Trivial Inequality states that if x is a real number, then x2 ≥ 0.

ψ Equality holds if and only if x = 0.

The second sentence above seems rather simple, but, in fact, it's very important. We call the condition
under which equality holds in a nonstrict inequality the equality condition of the inequality. Don't
overlook the equality conditions of inequalities—we'll see that they can be very helpful in solving
problems.

We've already seen in our discussion of finding maxima or minima of quadratic functions that this
Trivial Inequality is not so trivial. But it's useful for much more than just quadratics; it's the foundation
of many of the inequalities we will study throughout the rest of this chapter and in Chapter 18.
il,' _ _
Problems

x2 + 2
Problem 12.7: Show that —- ≥ 2 for all real numbers x.
√x2 + l

Problem 12.8: In this problem, we prove that if x and y are positive, then - 4— ≥ - -----.
X ∖J X I l∕
(a) Manipulate the desired inequality to produce an inequality that is equivalent to the Trivial
Inequality.
(b) Reverse your steps in part (a) to prove the desired inequality.

Problem 12.9:
(a) Prove that a2 + b2 ≥ 2ab for all real numbers a and b. When does equality occur?
(b) Prove that a2 + b2 + c2 ≥ ab + ac + be for all real numbers a, b, and c. When does equality occur?
(c) Prove that a4 + b4 + c4 ≥ abc(a + b + c) for all real numbers a, b, and c.

Problem 12.10: Let a and b be positive real numbers. Show that -z + Λ ≥ ~ + τ∙

az b2 a b

Problem 12.11: In this problem, we find all ordered pairs of real numbers (x, y) such that we have
(4x2 + 4x. + 3)(y2 - 6y + 13) = 8.
(a) Complete the square in both quadratics.
(b) Solve the problem.

χ2 + 2
Problem 12.7: Show that - = ≥ 2 for all real numbers x.
y[χ2- + 1

Solution for Problem 12.7: As usual, we start by working backwards. We multiply by Vx2 + 1 to get rid
of the fractions, and we have x2 + 2 ≥ 2 √x2 + 1. Because x2 ≥ 0 for all real x, we know that both sides of

379
CHAPTER 12. INEQUALITIES

x2 + 2 ≥ 2 √x2 + 1 are positive. So, we can square both sides to give

x4 + 4x2 + 4 ≥ 4(x2 + 1).

Rearranging this inequality gives us simply x4 ≥ 0, which by the Trivial Inequality is always true.
Now we have our path to a solution:

Proof: Because x2 is a real number, the Trivial Inequality gives us (x2)2 ≥ 0, so x4 ≥ 0. Adding 4x2 + 4
0 to both sides gives
x4 + 4x2 + 4 ≥ 4(x2 + 1),
so we have (x2 + 2)2 ≥ 4(x2 + 1). Because x2 ≥ 0, we know that both x2 + 2 and x2 +1 are positive.
So, we can take the square root of both sides of (x2 + 2)2 ≥ 4(x2 + 1) to give x2 + 2 ≥ 2 Vx2 + 1.
Dividing both sides by Vx2 + 1 gives the desired

Vx2 + 1
for all real numbers x.

□
While the Trivial Inequality is stated in terms of one variable, it's a powerful tool for multi-variable
inequalities, too.

Problem 12.8: Prove that if x and y are positive, then - + - ≥------ .

x y x+y

Solution for Problem 12.8: Again, we work backwards. Because x and y are positive, multiplying both
sides of the desired inequality by xy(x + y) to get rid of the fractions gives us

y(x + y) + x(x + y) ≥ 4xy.

Rearranging this gives us x2 - 2xy + y2 ≥ 0. Factoring the left side gives us (x - y)2 ≥ 0, which is true by
the Trivial Inequality. So, now we can write our solution forwards:

Proof: The Trivial Inequality gives us (x - y)2 ≥ 0. Expanding the left side gives x2 - 2xy + y2 ≥ 0.
0 Adding 4xy to both sides gives x2 + xy + y2 + xy > 4xy. Factoring on the left side gives
x(x + y) + y(x + y) ≥ 4xy. Since x and y are positive, we can divide both sides by xy(x + y) to
give the desired
1 1 4
÷ ≥
x y x+y

For many of the problems in the rest of this chapter, we will not be writing "forwards" solutions
after discovering solutions by working backwards. After each such problem, you should try closing the
book and re-creating the forwards solution on your own.

380
 12.2. THE TRIVIAL INEQUALITY

Problem 12.9:
(a) Prove that a2 ÷ b2 ≥ 2ab for all real numbers a and b. When does equality occur?
(b) Prove that a2 + b2 + c2 >ab + ac + be for all real numbers a, b, and c. When does equality occur?
(c) Prove tha t a4 + b4 + c4 ≥ abc(a + b + c) for all real numbers a, b, and c.

Solution for Problem 12.9:

(a) Rearranging the desired inequality gives a2 — 2ab + b2 ≥ 0. We recognize that a2 — 2ab + b2 = (a- b)2,
so the given inequality is just an instance of the Trivial Inequality. Equality occurs if and only if
(a - b)2 = 0, which is equivalent to a = b.
(b) Rearranging the desired inequality gives

a2 + b2 + c2 - ab - ac - be ≥ 0.

The expression on the left side is not a square, nor can we easily factor it. However, individual
parts of the expression look familiar. The terms a2, b2, and ab should remind us of the inequality
we just derived, a2 + b2 ≥ 2ab.
Concept: Don't treat every problem as if it is brand-new. Try to relate problems or
O^" ulJ parts of problems to other problems that you already know how to do.
Notice that a2 + b2 + c2 — ab - ac — be is symmetric, so it may help to write inequalities of the form
α2 + i>2 ≥ 2ab for all three pairs of variables:

a2 + b2 ≥ 2ab,
a2 + c2 ≥ 2ac,
b2 + c2 ≥ 2bc.

Adding these inequalities gives us twice our desired sum of squares on the greater side:

2a2 + 2b2 + 2c2 ≥ 2ab + 2ac + 2bc.

Aha! Dividing by 2, we get the desired inequality, a2 + b2 + c2 ≥ ab + ac + be.

For equality to occur, we must have equality in each of the three inequalities we added, and
this occurs if and only if a = b = c.
Concept: We can sometimes prove a complicated inequality by proving simpler
o ≡sg inequalities that can be combined to give us the desired complicated one.

(c) We just proved an inequality that involved a sum of squares. The sum α4 + b4 + c4 is a sum of
squares, too. Letting x = a2,y = b2, and z = c2 in

x2 + y2 + z2 ≥ xy + yz + zx (12.1)

gives us
a4 + b4 + c4 ≥ a2b2 + b2c2 + c2a2. (12.2)

381
CHAPTER 12. INEQUALITIES

We're not quite there, but if we can show that the right side of (12.2) is at least abc(μ + b + c), we'll
have our desired inequality. The right side of (12.2) is also a sum of three squares. We let x = ab,
y - be, and z = ca in (12.1), and we have
a2b2 + b2c2 + c2a2 ≥ ab2c + abc2 + a2bc.
Success! Factoring the right side of this inequality gives us
a2b2 + b2c2 + c2a2 ≥ abc(a + b + c). (12.3)
Combining inequalities (12.2) and (12.3) gives us z?4 + b4 + c4 ≥ a2b2 + b2c2 + c2a2 ≥ abc(a + b + c).

□
Our solutions to parts (b) and (c) show how we can build on inequalities we have already proved in
order to prove more complicated inequalities. First, in part (b), we recognized that the greater side of
the desired inequality was similar to that of the inequality in part (a). Moreover, the lesser side of the
desired inequality was also similar to the lesser side in the inequality of part (a). So, we added three
applications of part (a) to prove the desired inequality in part (b).

Concept: Once we have proved an inequality, it becomes a tool to prove other

Q..^-τ-,j inequalities, just like we can use proven identities to find new identities.

In going from part (b) to part (c), we first recognized that the desired greater side in (c) had the
same form as the greater side in (b)—they were both sums of three squares. So, we applied part (b) to
a4 + b4 + c4. Notice that we used a little wishful thinking here; we hadn't even thought about the lesser
side yet.

Concept: In more complicated inequality problems, we often just focus on one side
θ J3 of the desired inequality, hoping to be able to manipulate it into or compare
it to something that we can show is greater than the lesser side.

In other words, if we want to show that P ≥ R, we sometimes first find an expression Q such that
we can prove P ≥ Q. Then, we try to show that Q ≥ R, so we can deduce that P ≥ Q ≥ R. Above,
we did this by showing that a4 + b4 + c4 ≥ a2b2 + b2c2 + c2a2 and a2b2 + b2c2 + c2a2 ≥ abc(a + b + c), so
a4 + b4 + c4 ≥ abc(a + b + c).

Sidenote: An old legend has it that a man once willed his herd of camels to his three
children, giving the oldest 1/2 of the herd to the oldest, 1/3 to the middle
child, and 1/9 to the youngest child. Unfortunately, the herd only had
17 camels, so the children did not know how to divide the camels among
them. They asked their wise uncle what to do. He responded, zT will give
you a magic camel to help you divide the herd. But after you divide the
herd, you must return the camel."
When they added the magic camel to the herd, the herd then had 18
camels. The oldest then took 1/2 the herd, 9 camels. The middle child took
1 /3 of the herd, 6 camels. And the youngest took 1 /9 of the herd, 2 camels.
Together, then, they took 9 + 6 + 2 = 17 camels, and returned the 18th, the
magic camel, to their wise uncle.

382
 12.2. THE TRIVIAL INEQUALITY

Our intermediary expression a2b2 + b2c2 + c2a2 is a magic camel—it doesn't exist in the problem when
we start, and it's not there in the final result. But it's a crucial step from the beginning to the end.

Concept: Be on the lookout for magic camels.

O≡^

Problem 12.10: Let a and b be positive real numbers. Show that — 4—- ≥ —F -.
a2 b2 a b

Solution for Problem 12.10: First, we get rid of the fractions by multiplying by a2b2, which gives

a3 + &3 ≥ a2b + ab2.

Now, we move all the terms to one side and try to factor. This gives us a3 — a2b — ab2 + b3 > 0. Factoring
fl2 out of the first two terms on the left and b2 out of the last two terms gives

a3 - a2b - ab2 + b3 = a∖a -b) + b2(-a + b) = a2(a -b)- b2(μ -b) = (a2 - b2)(a -b) = (a- b)2(a + b).

So, our inequality now is (α — b)2(a + b) ≥ 0. We now have a path to a solution.

Since a > 0 and b > 0, we have a + b > 0. The Trivial Inequality gives us (a — b)2 > 0. Multiplying
these two gives (a — fe)2(α + b) ≥ 0, and working backwards through our steps above gives us the desired
inequality. □

Concept: Many inequality problems are essentially manipulation and factorization

Q uj problems.

Problem 12.11: Find all ordered pairs of real numbers (x, y) such that (4x2 + 4x + 3)(y2 - 6ι∕ +13) = 8. >

Solution for Problem 12.11: We certainly aren't interested in expanding the product of those quadratics,
so we look for other ways to work with the quadratics.

Concept: When stuck on a problem involving quadratics, try completing the square.
O≡^

We complete the square in the two quadratics, because we know plenty about perfect squares. The
resulting equation is
((2x + 1)2 + 2)((y - 3)2 + 4) = 8.
The Trivial Inequality tells us that (2x + 1)2 ≥ 0 and (y - 3)2 ≥ 0, so we have

(2x + 1)2 + 2 ≥ 2,
(y-3)2 + 4≥4.

383
CHAPTER 12. INEQUALITIES

The product of these inequalities looks a lot like our equation:

((2x + 1)2 + 2)((y - 3)2 + 4) ≥ 8. (12.4)
This inequality holds for all real x and y. In order for equality to hold, we must have equality in both
of the inequalities we multiplied together to produce line (12.4). In other words, equality only holds if
(2x + 1)2 = (y — 3)2 = 0. Therefore, we have ((2x +1)2 + 2)((y — 3)2 + 4) = 8 if and only if (x, y) = (-j, 3). □

Concept: Some equations can be solved by viewing them as the equality case of a
Q Ln_) nonstrict inequality. We can then use the equality condition of the inequal­
ity to find the solutions to the equation.

Exeιci⅝e⅜

12.2.1 Show that x2 + x + 1 ≥ 3x for all real x.

12.2.2 Show that (α2 + l)(b2 + 1) ≥ (ab + 1)2 for all real a and b.
12.2.3 We wish to show that if a > b, then a3 + a2b ≥b3 + ab2.

(a) What is wrong with this solution: "Because a> b, we have a2 > b2. Multiplying both sides by a + b
gives a3 + a2b > b3 + ab2, so a3 + a2b ≥b3 + ab2, as desired."
(b) Prove that if a > b, then a3 + a2b ≥b3 + ab2. When does equality hold?

12.2.4 Show that for all nonzero x, y, and z, we have x2 + y2 + z2 + + + >2 + Hints: 186

*
12.2.5 Prove that if a and b are nonnegative real numbers, then (μ7 + fe7)(α2 + b2) ≥ (α5 + b5)(a4 + b4).
Hints: 212
12.2.6
* Prove that + y2 + z2 ≥ xy - xz + 2yz for all real numbers x, y, and z. Hints: 189

12.3 AM-GM Inequality with Two Variables

While the Trivial Inequality is the most fundamental of the inequalities we will study, there are many
others that are powerful problem solving tools. As we'll see, the Trivial Inequality is a key tool in
proving many of these other inequalities.
The first new inequality we will explore is the Arithmetic Mean-Geometric Mean Inequality, which
is usually more succinctly called the AM-GM Inequality.

Definition: The arithmetic mean (AM) of the nonnegative numbers a↑, a2,..., an is

n
and the geometric mean (GM) of these numbers is

ψπ1α2∙∙∙⅜∙

384
 12.3. AM-GM INEQUALITY WITH TWO VARIABLES

You probably already know the arithmetic mean of a group of numbers as the average of the numbers.
We'll start by comparing the arithmetic mean and geometric mean of pairs of nonnegative numbers.

Problem 12.12: Show that ≥ Vflb for all a,b≥0. For what values of a and b are the two sides
equal?

Problem 12.13: Let x and y be positive real numbers.

(a) Find the geometric mean of - and -.
y

(b) Show that y + y ≥ 2. When does equality occur?

Problem 12.14: Show that ifx,y> 0, then -⅛ + — ≥ —.

x2 yλ xy

Problem 12.15: Show that if x, y, and z are nonnegative, then xy + yz + zx > x fyz + y y[zx + z y[xy.

Problem 12.16: Show that (x + y)(y + z)(z + x) ≥ 8xyz for all nonnegative numbers x, y, and z.

Problem 12.12: Show that

Cl + b Γ~Γ
----- > yab
2
for all a,b ≥ 0. For what values of a and b are the two sides equal?

So∕ufioπ for Problem 12.12: Working backwards, we start by multiplying both sides by 2 to give
a + b ≥ 2 y[ab. Because a, b ≥ 0, we can square both sides to give (a + b)2 ≥ 4ab. Expanding the left side
and rearranging gives a2 - 2ab + b2 ≥ 0. Factoring then gives fa - b)2 ≥ 0, which is true by the Trivial
Inequality. We can now walk backwards through our steps to write our proof:

Proof: Let a, b > 0. The Trivial Inequality gives us (α - b)2 ≥ 0, and expanding the left side gives
ltZ a2 - 2ab + b2 ≥ 0. Adding 4ab to both sides gives a2 + 2ab + b2 ≥ 4ab, so we have (a + b)2 ≥ 4ab.
Because both sides are positive, we can take the square root of both sides to give a + b ≥ 2 yfab.
Dividing both sides by 2 gives us fa + V)∣2 > *lfab.

We only have equality in our first step, (λ - b)2 ≥ 0, if a = b. Otherwise, the left side is strictly larger than
the right. If we have a = b, we have fa + b)∕2 - ^[ab, so equality holds in fa + b~)∕2 ≥ y/ab if and only if
a = b. □
We have now proved the AM-GM Inequality for two variables:

385
CHAPTER 12. INEQUALITIES

Important: The arithmetic mean of two nonnegative numbers is greater than or

ψ equal to the geometric mean of the numbers. The two means are equal
if and only if the two numbers are equal. In other words, if a and b are
nonnegative, then
Ω÷b ∕~τ
— ≥ yabr

and equality holds if and only if fl = b.

Don't overlook the equality condition of the AM-GM Inequality. The sentence, "The two means are
equal if and only if the two numbers are equal," not only tells us that the two means are the same if the
two numbers are equal, but it also tells us that the two numbers must be equal if the two means are the
same.

Let's try AM-GM on some problems.

Problem 12.13: Let x and y be positive real numbers. Show that —t- — ≥ 2. When does equality
1/
occur?

Solution for Problem 12.13: Because x and y are positive, so are x∣y and y∕x. So, we can apply in the
AM-GM Inequality to find
χ∕y + y∕χ > ∕χ y ,
2 ~y∣y'χ ∙

Multiplying the far left side and the far right side by 2, we get

- + - ≥ 2.
y χ

Equality occurs if and only if J = Cross-multiplying gives x2 = y2. So, because x and y are positive,
equality occurs if and only if x = y. □

One of our biggest clues to try AM-GM in Problem 12.13 is the fact that and have a constant
product. This means that if we apply AM-GM to these quantities, we will have a constant on the lesser
side of the inequality, just as our desired inequality has.

Similarly, if we have nonnegative expressions with a constant sum, we can apply AM-GM to produce
an inequality in which the larger side is constant.

Concept: When trying to prove an inequality involving nonnegative expressions

Q"- oj that have a constant sum or product, consider using AM-GM.

Problem 12.14: Show that if x, y > 0, then — d—- ≥ —.

x2 y2 xy

Solution for Problem 12.14: We have a sum on the greater side and a product on the lesser side. This

386
 12.3. AM-GM INEQUALITY WITH TWO VARIABLES

looks like a job for AM-GM. Applying AM-GM to 4 and Λ gives

112
Multiplying both sides by 2 and simplifying the right side gives the desired — + — ≥ —. ∏
V J

ProbleπGL2Λ5^^ho^th^Hf^^^nd^arenonnegative3her^^^^+zx≥xy^+^Λ^xj^^^J
-

Solution for Problem 12.15: We're not sure where to start, but we see sums of expressions, so we try
AM-GM. What's wrong with the following start:

Bogus Solution: We try applying AM-GM to the sum x fyz + y >fzx. This gives us

≥ √(
√yz)(y√≡)∙
*
2

Not only is the expression on the right a mess, but the result is an inequality in which x y/yz is on the
greater side, not the lesser side.

WARNING!! If you want to show that A ≥ B, then inequalities like C ≥ A or B ≥ D

aA4 are not going to help you. You need to work with inequalities in which
A is on the greater side or B is on the lesser side.

The y[yz term is a big clue how to use AM-GM, particularly because the square root is on the lesser side
of the inequality. We produce the square roots by applying AM-GM to each pair of x, y, and z:

χ+y
2
y+z
2
Z + X
~τ~

Multiplying these by z, x, and y, respectively, gives us all the terms we want on the lesser side:

≥ Zy]xy,
> xyfyz,
z+x
≥ y^fzx.

Alaa! This gives us the terms we want on the greater side, too!

387
CHAPTER 12. INEQUALITIES

Concept: Focus on one side of an inequality and the other will often fall into place.
O≡^

Adding all three of these inequalities gives the desired xy + yz + zx ≥ x yfyz + y √zx + z √xy. □
Addition is not the only way we can combine inequalities.
Problem 12.16: Show that (x + y)(y + z)(z + x) ≥ 8xyz for all normegative numbers x, y, and z.

Solution for Problem 12.16: We have a product of sums on the greater side of the inequality. Sums on
the greater side—that sounds like AM-GM. Let's try it! We use AM-GM to produce each sum on the
desired left side, which gives us

We want the product of these sums, so we multiply all three inequalities to get

(χ + y)(y + z)(z + x)
8
We have √x2y2z2 = xyz because x,y,z ≥ 0, so multiplying both sides by 8 gives the desired inequality
(x + y)(y + z)(z + x) ≥ 8xyz. □

12.3.1 Show that the sum of any positive number and its reciprocal is no less than 2.

12.3.2 Show that a6 + b4 ≥ 2a3b2 for all positive real numbers a and b.

12.3.3 Let a, b, c, and d be nonnegative real numbers. Show that (ab + cd)(ac + bd) ≥ 4abcd.

12.3.4 Show that if x, y > 0, then yJx + ∣ + y + j ≥ 2 V2.

12.3.5 Let x and y be positive real numbers such that x + y = 1. Show that (1 + ∣) (1 + ∣) > 9. (Source:
CMO)
x2 + 2
12.3.6 Use AM-GM to show that ι > 2 for all real x.
Vx2 + 1

Extra! The beginning of knowledge is the discovery of something we do not understand.

ll<∣∣n+l∣<ll∣* _ Frank Herbert

388
 12.4. AM-GM WITH MORE VARIABLES

12.4 AM-GM with More Variables

We've seen that AM-GM with two variables is basically just a routine application of the Trivial Inequality.
If that were all there were to AM-GM, it probably wouldn't even have its own name. But having seen
that AM-GM holds for two variables, we should wonder if it holds for more variables.
The AM-GM Inequality does, indeed, hold for more variables.

Important: The AM-GM Inequality tells us that the arithmetic mean of any n non­
negative numbers is greater than or equal to the geometric mean of the
numbers. That is, if α1, a2,..., an are nonnegative, then

Equality holds if and only if all α,∙ are equal.

The proof is very intricate (and beautiful!). We'll give you a chance to prove the general AM-GM
Inequality as a series of Challenge Problems. Through this series of problems, you'll walk in the footsteps
of the great mathematician Augustin Louis Cauchy.

The French mathematician and engineer Augustin Louis Cauchy was one
of the most prolific mathematicians of the 1800s. He made numerous
contributions relating to inequalities, one of which bears his name. He also
made important contributions to number theory, geometry, calculus, and
even physics. It's hard to go very far in math and science without reading
about his many accomplishments. However, the name Cauchy would
probably be more well-known were it not for Cauchy's abrasiveness and
the many contentious rifts he had with other mathematicians of his day.

Fortunately, using AM-GM for any number of variables is a lot easier than proving it works! Try it
and see.

Problems

Problem 12.17: Let x, y, and z be positive real numbers. Show that

t*+y+⅜+H)a9∙

∕n + l∖n
Problem 12.18: In this problem, we show that that I—-—I > nl for all positive integers n≥2.
(a) Take the nth root of both sides. Why does the result make us think of the AM-GM Inequality?
(b) Solve the problem.

389
CHAPTER 12. INEQUALITIES

Problem 12.19: Prove that for all nonnegative real numbers x, y, z, we have

x2(l + y2) + y2(l + z2) + z2(l + x2) ≥ 6xyz.

When does equality occur?

1 1
Problem 12.17: Let x, y, and z be positive real numbers. Show that (x + y + z) - + -
x y

Solution for Problem 12.17: We have two sums on the greater side, so we apply AM-GM to both. We find

x+y+z _
------ ---- ≥ √xyz,
1+1 1
χ___ y_
3 ^∕xyz'

Multiplying these inequalities, we have

(χ + y + z)(l + l + l)
>
v≡ = 1
9

Multiplying both sides by 9 gives the desired result. □

1 (n + l∖n
Problem 12.18: Prove that I —— I∣ > π! for all positive integers n ≥ 2.
■■■■■■■■■■■■■■■■■■■■■
Solution for Problem 12.18: First, we take the nth root of both sides to get

n+1
2

We now have an πth root of a product of n numbers on the lesser side, which is a strong indication to
use the AM-GM Inequality. So, we apply AM-GM to the n numbers 1, 2,..., n, which gives

(Note that this is a strict inequality, since the numbers 1, 2,..., π are not all equal, and therefore, they
do not satisfy the equality condition of AM-GM.) Writing the left side in closed form gives us
n(n+l)
1 +2 + +n 2 n+1
n n 2 '

so we have ,jγ- > y∕ip.. Raising both sides to the rzth power gives the desired result. □

390
 12.4. AM-GM WITH MORE VARIABLES

Problem 12.19: Prove that for all nonnegative real numbers x, y, z, we have

2(l + y2) + y2(l + z2) + z2(l + x2) ≥ 6xyz.

x*

When does equality occur?

Solution for Problem 12.19: We have a sum on the greater side and a product on the lesser, which suggests
using the AM-GM Inequality. However, when we apply AM-GM to the three added terms on the left,
their geometric mean is a very messy expression. Going the other direction, we could view xyz as the
geometric mean of x3, y3, and z3, but that doesn't seem to help much, either.

Concept: It's often necessary to manipulate a desired inequality into a form for
θy~ mJ which tools like AM-GM apply. Factor! Expand! Regroup terms!

Here, the 6 is a big clue. Expanding all the terms on the left gives us 6 terms:
x2 + x2y2 + y2 + y2z2 + z2 + z2x2 ≥ 6xyz.
The larger side is now the sum of six terms, and the smaller is a product, so the inequality is a natural
candidate for AM-GM. All six terms are nonnegative, so by the AM-GM Inequality, we have

≥ ψ2 ■ χ2y2 ■ y2 ∙ y⅛2 ■ z2 ■ z2χ2

= ^x6y6z6

= xyz.
Multiplying both sides by 6 gives the desired x2 + x2y2 + y2 + y2z2 + z2 + z2x2 ≥ 6xyz.
We must also determine when equality occurs. Equality occurs if and only if x2 = x2y2 = y2 = y2z2 =
z2 = z2x2. Since x, y, and z are nonnegative and x2 = y2 = z2, we get x = y = z. Then x2 = x4, so x = 0 or
x = 1. Therefore, equality occurs only for (x, y,z) = (0,0,0) and (x, y, z) = (1,1,1). □

Exercises

12.4.1 Let a,b,c≥ 0. Show that a2b + ab2 + a2c + ac2 + b2c + bc2 ≥ 6abc.
, . . 1 a b c
12.4.2 Show that if a, b, and c are positive, then - + - + - ≥ 3.
r b c a
2 4
12.4.3 Find the minimum value of xy + - + - for positive real numbers x and y.

12.4.4 Show that, for all positive real numbers p, q, r, s, we have

(p2 + p + l)(t∕2 + q + l)(r2 + r + l)(s2 + s + 1) ≥ Slpqrs.

12.4.5
* Show that — * — + — + — + CL∏ ÷ #1 a↑ + ^2
•∙ + + ≥ 2n for all positive reals aχ, a2, ■ ■ ult.
«1 «2
Hints: 210

*
12.4.6 If fl > b > 0, find the minimum value of fl + ----- ——. Hints: 345
(a - b)b
391
CHAPTER 12. INEQUALITIES

12.5 The Cauchy-Schwarz Inequality

The AM-GM Inequality isn't the only inequality that Cauchy proved. He also developed a powerful
inequality that still bears his name.

Problems

Problem 12.20: Let a, b, c, and d be real numbers. Prove that (α2 + b2)(c2 + d2) ≥ (ac + bd)2. When does
equality occur?

Problem 12.21: Let a∖, a2,..., an, bχ, ½,..., bn be real numbers. In this problem, we find two proofs
that
(αξ + «2 + ∙ ∙ ∙ + a2)(b2 + b2 + ■ ■ ■ + b2) ≥ (a-lb1 + a2b2 + ∙ ∙ ∙ + anbn)2.

We call this the Cauchy-Schwarz Inequality.

(a) Proof 1: Consider the expansion of both sides. What terms have no variables besides a↑, a2, b↑,
and b2?
(b) Apply the inequality from the previous problem to the terms in part (a).
(c) Use your observations in parts (a) and (b) to prove the Cauchy-Schwarz Inequality.
(d) Proof 2: Consider the quadratic ∕(x) = (ayx — bf)2 + (a2x — b2)2 + ■ ■■ + (anx — bn)2. What is the
discriminant of this quadratic?
(e) Why must the discriminant of ∕(x) be nonpositive?
(f) Find the equality condition of the Cauchy-Schwarz Inequality.

7 .0 O (fl + fo + C)2
Problem 12.22: Show that a2 + b2 + c2 >------ - ------ .

Problem 12.23: Let x, y, and z be positive real numbers. Use the Cauchy-Schwarz Inequality to prove
that

<ι+y+z,(H+l)≥9∙
Cl u
Problem 12.24: Let a and b be positive real numbers with a + b = 1. Prove that------ + -------- ≥ -.
fl+1 &+1 3
(Source: Hungary)

Problem 12.25: If a, br c, x, y, and z are nonzero real numbers and a2 + b2 + c2 = 25, x2 + y2 + z2 = 36,
and ax + by + cz = 30, compute
a+b+c
x + y + z'
(Source: ARML)

Just as we did with AM-GM, we start with the 2-variable version of the Cauchy-Schwarz Inequality.

392
 12.5. THE CAUCHY-SCHWARZ INEQUALITY

Problem 12.20: Let a, b, c, and d be real numbers. Prove that (α2 + b2')(c2 + d2) ≥ (ac + bd~)2. When
does equality occur?

Solution for Problem 12.20: Solution 1: Bash it out. We start by working backwards. We expand both
sides, which gives
a2c2 + a2d2 + b2c2 + b2d2 ≥ a2c2 + 2abcd + b2d2.
Simplifying this gives a2d2 + b2c2 ≥ 2abcd. What's wrong with this finish:

Bogus Solution: Because a2d2 and b2c2 are positive, we can apply AM-GM to find

-4 h^∙r^∙ _________
— γ-÷ ≥ y[(^dW2c⅛ = abed.

Multiplying both sides by 2 gives a2d2 + b2c2 ≥ 2abcd. Adding

a2c2 + b2d2 to both sides, and then factoring the result, gives the
desired
(α2 + fc2)(c2 + d2) ≥ (ac + bd')2.
We have equality when we have equality in our application of AM-
GM, which is when a2d2 = b2c2.

We've overlooked one detail, and that is the fact that a, b, c, or d might be negative, so abed might be
negative above. Therefore, we must write λ∕(λ2z72)(⅛2c2) = ∣abcd∣. We can still patch our proof with a little
careful casework, or we can remember that 2-variable AM-GM has its roots in the Trivial Inequality.
Subtracting 2abcd from both sides of a2d2 + b2c2 ≥ 2abcd leads us to the following solution:

Proof: The Trivial Inequality gives us (ad - be)2 ≥ 0. Expanding the left side and rearranging gives
0 a2d2 + b2c2 ≥ 2abcd. Adding a2c2 + b2d2 to both sides, and then factoring the result, gives the
desired
(a2 + b2)(c2 + d2) > (ac + bd)2.

We have equality when we have equality in our initial application of the Trivial Inequality,
which was (ad - be)2 ≥ 0. Therefore, our equality condition is ad = be.

Solution 2: Recall a similar identity. Back on page 340, we proved the identity

(a2 + b2)(c2 + d2) = (ac + bd)2 + (ad - be)2.

By the Trivial Inequality, we have (ad - be)2 ≥ 0, so we have

(a2 + b2)(c2 + d2) - (ac + bd)2 = (ad - be)2 ≥ 0,

which means that

(a2 + b2)(c2 + d2) ≥ (ac + bd)2.
Equality holds when (ad - be)2 = 0, which occurs if and only if ad = be. □

393
CHAPTER 12. INEQUALITIES

Don't worry if you didn't remember the identity we used in our second solution. Our first solution
shows that, when all else fails, some algebra and the Trivial Inequality will often do the trick.
We were able to extend AM-GM to any number of variables. Can we do the same with the inequality
we proved in Problem 12.20?

Problem 12.21: Let flι, ∏2, ■.., an, b↑,b2, ■ ■ ■, brι be real numbers. Prove that
(α^ + fl2 + ... + fl2^2 + &2 + ... + > + a2b2 + ... + anbn)1.

Solution for Problem 12.21: Solution 1: Extend our previous solution. If we expand the products on both
sides, we have terms of the form a2b2 on both sides. On the left, we also have all the possible terms a2b2
for which i ≠ j and 1 ≤ z, j < n, which we'll write as a^' ancl on t^e right/ we also have all
i≠ j,l≤i,j≤n
possible terms 2aibiajbj for which 1 ≤ i < j ≤ nl which we'll write as 2albiajbj. We can therefore
l<i<j≤n
write the expansions as

∑⅛⅛+. ∑ ⅛^≥∑,4⅛ + ∑ 2aibiajbj.

∕=1 ι≠j,l≤f,∕≤n ι=l l≤i<j≤n

The first sum on the left matches the first sum on the right, so we don't have to worry about those. Next,
rather than work with the remaining complicated summations, we look at one 2aibjajbj on the right and
the corresponding a2b2 and a2b2 on the left. This is exactly the same problem we faced in Problem 12.20!
The Trivial Inequality gives us
(aibj - ajbi)2 ≥ 0.
Expanding the left side and rearranging gives a2b2 + a2b2 ≥ 2zz,⅛7fy. If we do this for all pairs i and j
such that 1 ≤ i < j ≤ n and i ≠ ∕, and then add the results, we have

∑ ⅛6∕≥ ∑ 2aibiajbj.
i≠j,l≤i,j≤n l≤i<j≤n

n
Adding ajtf to both sides gives
1=1

Σ⅛⅛+ Σ ⅛fc∕≥∑⅛⅛+ ∑ 2zz,⅛⅛

i=l i≠j,l≤ifj≤n i=l l≤i<j≤n

which is our expansion above of the desired inequality. Factoring both sides gives

(«2 + «2 + ∙,, + cfy(b} + b2 + ■ ∙ ■ + b2) ≥ (a↑b1 + a2b2 + ∙ ∙ ∙ + anbn)2.

We have equality if and only if we have equality in all of our applications of the Trivial Inequality, which
were the inequalities (zz∣b∣ - a∣bz)2 ≥ 0, where z ≠ j. In other words, for any z and j such that i ≠ j, we must
have aibj = ajb∣. If bi and b∣ are nonzero, we can rearrange this as z7∣√fe∣ = cij∕bj. Since this relationship
must hold for any z and j, the ratio a∣Jbk must be constant for all k. There is however a subtlety to this

394
 12.5. THE CAUCHY-SCHWARZ INEQUALITY

equality condition. Some of the variables can equal 0. For example, equality occurs if (a↑, a2ra3) = (0,2,3)
and (b↑, b2, b3) = (0,4,6). Also, equality occurs if either all the a∣ or all the b∣ are zero. So, to be completely
accurate, equality occurs if and only if either all a∣ are 0, or there is a constant t such that b∣ = ta∣ for all i.
(The case that all the b∣ are 0 is addressed by letting t = 0.)
Solution 2: Use the discriminant. Discriminant? What discriminant? To see how we might think to
use the discriminant, we let

P = α∣ + «2 + ∙ ∙ ∙ + α2,
Q = a-lb↑ + a2b2 + ∙ ∙ ∙ + anbn,
P — b∣ + b2 + ∙ ∙ ∙ + b^.

This allows us to write our desired inequality as simply PR ≥ Q2, or Q2 - PR ≤ 0. A square minus a
product less than or equal to zero; this might make us think of the discriminant of a quadratic that has
no more than one real root. (Yes, this is a leap, but great insights come from such creative jumps.)
But what quadratic might have such a discriminant? The coefficients of its quadratic term and
constant term must be a2 + α2 -ι ∖-a2 and b2 + b∣ -+------ +- b2l, so we can try to produce a suitable quadratic
f(x) by writing
∕(x) = (fllχ - b1)2 + (a2x - b2)2 + ∙ ∙ ∙ + (anx - bn)2.
This gives us the appropriate quadratic and constant terms. But what about the linear term? When we
expand each of the squares, we have

f(x) = (a2 + ^2 + ■- ■ + β2)x2 - 2(∏ιbj + fl2½ + ∙ ∙ ∙ + anbn)x + (b2 + b2 + ∙ ∙ ∙ + b2).

Now, we're in business. We have ∕(x) = Px2 - 2Qx + R, and we wish to prove that PR ≥ Q2.
Before we dive into using our quadratic, we have to address the possibility that P = 0. If P =
a2 + a2 + ∙∙∙+a2 = 0, then each a, must be equal to 0. Then we have Q = 0, so the statement PR > Q2 is
obviously true. Otherwise, we have P > 0, so f{x) = Px2 — 2Qx + R is a quadratic.
We thought of building this quadratic by thinking about discriminants. To create an inequality from
the discriminant, we must learn more about the roots of ∕(x). Because ∕(x) is a sum of squares, we
have ∕(x) ≥ 0 for any real number x. Since ∕(x) is never negative, it cannot have two distinct real roots.
Therefore, its discriminant must be nonpositive. This gives us (-2Q)2 - 4PP ≤ 0, from which we have
the desired Q2 ≤ PR, or PR ≥ Q2.

Substituting back for P, Q, and R, we have the Cauchy-Schwarz Inequality:

(β∣ + #2 T ‘- - ^* ^∏)(b2 -f- b2 -f- ∙ ∙ ∙ + b2) ≥ (a^b] + a2b2 + ∙ ∙ ∙ ÷ anbn')~.

We have equality if and only if the discriminant of our quadratic is 0. There are a couple ways this
could happen. First, we might have all the a, or all the b∣ equal to 0. (If all the a∣ equal 0, then ∕(x) isn't
even a quadratic, but we clearly still have equality in the Cauchy-Schwarz Inequality.) Second, if some
of the αl are nonzero, the discriminant of ∕(x) is still 0 if ∕(x) has a double root. Let's take another look
at/(x):
∕(x) = (α1χ - b1)2 + (fl2x - b2)2 + ∙ ∙ ∙ + (flnx - bn)2.

395
CHAPTER 12. INEQUALITIES

Because ∕(x) is a sum of squares, the Trivial Inequality tells us that ∕(x) can never be negative. It can
equal 0, but only if all of the squares equal 0 for the same value of x. In other words, we must have
(a∣x - frl∙)2 = 0 for the same value of x for all values of i. This means that there is some x such that fe, = α1∙x
for all i. In this case, this value of x is the only value of x such that f{x) = 0. Since f(x) is a quadratic, this
means that this value of x is a double root of the quadratic, so the discriminant of the quadratic equals
0, as desired.
Therefore, the equality condition for the Cauchy-Schwarz Inequality is that either all the α,∙ equal 0,
or there exists a constant t such that bi = ta, for all i. □

Important: The Cauchy-Schwarz Inequality states that for any two sequences of n

ψ real numbers a↑, a∙χ,..., an and b∖, ⅛∕ ∙ ∙ ∙∕ bn, we have

(α2 + «2 + ∙ ∙ ∙ + ⅛)(b2 + b2 -ι— + b2) ≥ («i&i + a2b2 + ∙ ∙ ∙ + anbn)2.

Equality occurs if and only if either αz∙ = 0 for all z, or there exists a
constant t such that b∣ = tαz∙ for all i.

Sidenote: The Cauchy-Schwarz Inequality is also sometimes called the Cauchy-

Schwarz-Buniakowsky Inequality.

Rather than writing Cauchy-Schwarz Inequality (or Cauchy-Schwarz-Buniakowsky Inequality) over

and over, we will often simply refer to the inequality as "Cauchy." From example, we might write, "By
Cauchy, we have (a2 + l)(fe2 + 1) ≥ (α ■ b + 1 ∙ 1)2, so (tz2 + l)(fo2 + 1) ≥ (ab + 1)2."

Sidenote: If you ever have trouble remembering which side of the Cauchy-Schwarz
Inequality is the larger side and you don't have time to re-derive it, just
compute a simple case. For example, let α1 = 1, α2 = 2, bi = 3, and b2 = 4.
Then, we have (a^+a^^ + bf) = (5)(25) = 125, and (αχbι +π2b2)2 = (3 + 8)2 =
121, so (fl2 + zz2)(fr2 + b∣) is on the larger side.
If you like to remember things with words, you can think of Cauchy
as, "The product of the sum of the squares is greater than or equal to the
square of the sum of the products."
Or perhaps you like symbols: ∏ ∑ □ ≥ □ ∑ ∏. (Remember, ∏ is the
symbol we use for products and ∑ is the symbol for sums.)

It should not be a surprise that the Cauchy-Schwarz Inequality is particularly useful for inequalities
that involve sums of squares.

Solution for Problem 12.22: We have a sum of squares on the greater side, and the square of a sum on the
lesser side. This may be a job for Cauchy. We isolate the square of the sum by multiplying both sides

396
 12.5. THE CAUCHY-SCHWARZ INEQUALITY

by 3:
3(∕72 + b2 + c2) ≥ (a + b + c)2.
To use Cauchy here, one of our sequences clearly must be a, b, c, since we have a2 + b2 + c2 on the greater
side. What is the other sequence? If we don't see it right away, we can let the other sequence be x, y, z,
so that Cauchy gives us
(x2 + y2 + z2)(α2 + b2 + c2) ≥ (xα + yb + zc)2.
Comparing this to our desired inequality, we see that if we let x = y = z = 1, then we have the desired

3(a2 + b2 + c2) > (α + b + c)2.

□

Concept: If an inequality involves a sum of squares or a product of two sums with

O≡^≡)
the same number of terms, try using the Cauchy-Schwarz Inequality to
prove it.

We also saw another useful inequality manipulation in Problem 12.22 when we let one of the
sequences be 1,1,1 in the Cauchy-Schwarz Inequality:

Concept: Some inequalities result from setting variables or even entire sequences of
CD~ l-,lJ variables equal to carefully chosen constants, and then applying known
inequalities to those variables or sequences.

Just as sums of squares are a tip-off to consider Cauchy, so, too, are products of sums:

Problem 12.23: Let x, y, and z be positive real numbers. Show that (x + y + z) ∣μ÷ι÷i]∣ >9. ∣
lx y χ)
Solution for Problem 12.23: We tackled this once already in Problem 12.17, but Cauchy now offers us an
even faster solution. The product of sums on the greater side is our clue to try Cauchy. We first write
the Cauchy-Schwarz Inequality for three variables:

(α2 +a2+ a2) (b2 + b2 + b2) ≥ (α1b1 + α2b2 + a3bf)2.

In the case of the desired inequality, we let a↑ = y∕x, α2 = √y, fl 3 = √z, and b1
get

which is the desired inequality. □

Problem 12.23 is an example of how Cauchy can help deal with variables in denominators. While
the product of sums was our first clue to use Cauchy, a more experienced problem-solver might see the
denominators themselves as a clue to try Cauchy.

397
CHAPTER 12. INEQUALITIES

Problem 12.24: Let a and b be positive real numbers with a + b = 1. Prove that

fl2 + — > -
fl + 1 b+1 3
(Source: Hungary)

Solution for Problem 12.24: The inconvenient denominators make the expression on the left side difficult
to deal with. We just saw an example of the Cauchy-Schwarz Inequality eliminating denominators,
so we try that here. However, in this problem, we aren't given a second sequence that eliminates the
denominators for us. So, we introduce it. Specifically, we pair the denominator fl + 1 with another
appearance of fl + 1, and likewise for b + 1:

[(β + l) + (t, + l)](-≤-

∖a + 1 —b
+ b + l)~ √fl + 1 ∙ + √fe + 1 ∙

(a + b)2,

Concept: Cauchy can help us prove inequalities that have a sum of fractions on the
θ∙∙, "jjj greater side. To try using Cauchy on such an inequality, let the sum of the
fractions be one sum of squares, and the sum of the denominators be the
other sum of squares:

(y? + Vl + , ∙ ∙ ÷ xJ2n) ≥ (*1 + *2 + ∙ ∙ ∙ + Xn)1-

This strategy should particularly come to mind when the numerators of

the fractions are perfect squares.

Problem 12.25: If fl, b, c, x, y, and z are nonzero real numbers and a2 + b2 + c2 = 25, x2 + y2 + z2 = 36,
and ax + by + cz = 30, compute
a+b+c
x + y + z'
(Source: ARML)

Solution for Problem 12.25: The given equations include two sums of squares, and a sum of products of
the variables that are squared in the other two equations. These sums remind us of the Cauchy-Schwarz
Inequality. Applying Cauchy, we get
(fl2 + b2 + c2)(x2 + ι∕2 + z2) ≥ (ax + by + cz)2.

But from the given equations, we have

(fl2 + if + c2)(x2 + y2 + z2) = 900 = (ax + by + cz)2.

398
 12.5. THE CAUCHY-SCHWARZ INEQUALITY

Therefore, the information in the problem fits the case in which equality holds for the Cauchy-Schwarz
Inequality. This tells us that the equality condition for the Cauchy-Schwarz Inequality must be met.
Specifically, it means that for some real number f, we must have a = tx,b = ty, and c = tz. Substituting
into ax + by + cz = 30, the left side becomes

ax + by + cz = tx2 + ty2 + tz2 = t(x2 + y2 + z2),

so f(.r^ + ι∕2 + z2) = 30. Since x2 + y2 + z2 = 36, we find that f = 5/6. Finally, we note that

fl + b + c _ tx + ty + tz 5
x+y+z ~ x+y+z ^ 6'
□

Concept: Some equations are really inequality problems in disguise. If an equation

O^ mJ tells us that equality holds in some inequality, then we can use the equality
condition of the inequality to solve the equation.

Sidenote: The Cauchy-Schwarz Inequality is a special case of the Holder Inequality,

j∖ which states that if p and q are positive numbers such that + ∣ = 1 then

^Jι+X2y2 + --- + χnyn≤(xp1+χp2 + --→xpn)"(yq1 + yq2 + --- + y^

for all nonnegative numbers Xi, x2, ...,xn and yi, y2,..., yn. What happens
when p = q = 2?

∣∣∣l
Exercise
*

12.5.1 Show that (flι⅛ι + fl2b2 + ∙ ∙ ∙ + flπbπ) + ^- + ∙∙∙ + ^∙j≥(flι+fl2 + ∙∙∙ + flπ)2, where fl∕, b1 > 0 for

all i.
12.5.2 Use the Cauchy-Schwarz Inequality to prove that a2 + b2 + c2 ≥ ab + ac + be for all positive a, b,
and c.

12.5.3 Jane has just drawn 6 rectangles, and each rectangle has a different length and width. The sum
of the squares of the lengths is 40 and the sum of the squares of the widths is 20. What is largest possible
value of the sum of the areas of the rectangles?

12.5.4 Suppose that a, b, c, and d are positive. Prove that 2 Vfl + b + c +1? ≥ -√fl + Vb + √c + √d. Hints: 8

64
12.5.5 Let a, b, c, and d be positive real numbers. Show that - + - 4----- 1----- ≥ (Source:
abed a + b + c + d'
Leningrad) Hints: 126

(3x + 4y)2
12.5.6 Let x, y > 0. Find the maximum value of
x2 + y2

399
CHAPTER 12. INEQUALITIES

12.6 Maxima and Minima

In this section, we use our inequality tools to find the minimum or maximum value of various expres­
sions.

Important: Maximization and minimization problems are two-part problems. We

must show that the maximum or minimum value is attainable, and that
no better value (no higher for maximization, no lower for minimization)
can be attained.

We sometimes refer to maximization or minimization problems as optimization problems.

Problems

Problem 12.26: Suppose that a + b < c + d + 3 and c - b ≤ 3. Find the maximum possible value of a - ci.

Problem 12.27: Let x2 + 3xy + y2 = 60, where x and y are real. In this problem, we determine the
maximum possible value of xy. (Source: CEMC)
(a) Isolate xy in the equation, and then maximize the expression that xy equals.
(b) Confirm that your maximum value in part (a) can be attained. If it cannot, return to part (a),
and find a different way to isolate xy.

Problem 12.28: Let US be a rectangle with perimeter 36. In this problem, we find the largest possible
area of 2?.
(a) Use one of the inequalities in this chapter to relate the perimeter and the area of the rectangle.
(b) Use your inequality from part (a) to maximize the area of the rectangle. What must be true of
the rectangle if it has this maximum area?

Problem 12.29: BigBox Company is making big rectangular cardboard boxes. They don't want to use
any more cardboard than they have to. A customer requests boxes that have a volume of 1000 cubic
inches.
(a) Suppose the customer wants each box to have a top, a bottom, and all four sides. Show that
the BigBox Company should make boxes that are cubes in order to minimize the amount of
cardboard they have to use.
(b) Suppose the customer wants each box to have a bottom and all four sides, but no top. Wrhat
dimensions should BigBox Company use to minimize the amount of cardboard they use on each
box?

x2 + 16
Problem 12.30: For x > 0, let ∕(x) = Find the minimum possible value of ∕(x).
3x

Extra! Which facts arc likely to reappear? The simple facts. How to recognize them? Choose those that
ι* κ l>* κ seetn simple.
- Robert M. Pirsig

400
 12.6. MAXIMA AND MINIMA

Problem 12.31: Let Xj, x2,..., x,l be a sequence of integers such that

(i) -1 ≤ Xj ≤ 2, for z = 1, 2,3,..., n,

(ii) x1 + x2 + ∙ ∙ ∙ + x,∣ = 19, and

(iii) x2 + γ2 + ... + χ2 _ 99

Find the minimal and maximal possible values of x∣ + x^ + ∙ ∙ ∙ + x„. (Source: AHSME)

Solution for Problem 12.26: What's wrong with the following:

Bogus Solution: If equality holds in our inequalities, we have c - b = 3 and a + b =

∣X∣ c + d + 3. Solving for c in the first equation gives us c = b + 3, and
substituting this in the second equation gives usa + b = b + 3 + d + 3.
Simplifying this equation gives us a - d = 6, so the maximum value
of a - d is 6.

As we'll see, the numerical answer is correct, but we haven't proved that a-d cannot possibly equal
anything higher.

WARNING!! We can't be sure we have the maximum or minimum value of an

expression until we show that no better value can be achieved.

We want to maximize a -d,so we want to find an inequality of the form

a - d ≤ some number.

If we also show that a - d can equal this number, then we have found the maximum of a - d.
Since we want a - d on the lesser side of an inequality, we subtract b and d from both sides of
a + b<c + d + 3to get a-d<c-b + 3. We also know that c - b < 3, so we have

ci — d<c-b + 3<3 + 3 = 6.

Therefore, we have a - d < 6. We now want to show that it is possible to have a - d = 6. Our Bogus
Solution above shows us how. We choose values such that equality holds in all the inequalities. For
example, if (a, b, c, d) = (6,2,5,0), then both inequalities are satisfied and a - d = 6. □

Problem 12.27: If x2 + 3xy + y2 = 60, where x and y are real, then determine the maximum possible
value of xy. (Source: CEMC)

Solution for Problem 12.27: Here's another common error in maximization and minimization problems:

401
CHAPTER 12. INEQUALITIES

Bogus Solution: We recognize the left side as xy+(x+y)2, so we have xy+(x+y)2 = 60.
Isolating xy gives xy = 60 - (x + y)2. By the Trivial Inequality, we
have (x + y)2 ≥ 0, so 60 - (x + y)2 ≤ 60. Therefore, we have xy < 60,
which means the maximum possible value of xy is 60.

What we haven't done in this Bogus Solution is shown that the value 60 can be achieved. We only
have xy = 60 if 60 - (x + y)2 = 60. That is, we achieve our maximum when (x + y)2 = 0. We have
(x + y)2 = 0 if and only if x = -y. But if x = -y, then we have -x2 = 60, so x2 = -60. There are no real
solutions to this equation! So, we see that we cannot have xy = 60.

WARNING!! To show that an expression has a specific maximum value, there are
V two steps:

1. Show that the expression cannot be greater than this value.

2. Show that the expression can be equal to this value.

Both steps are necessary! Similarly, if we wish to show that an expres­

sion has a minimum value, we show both that the expression cannot
be less than this value, and that the expression can equal this value.

Back to the drawing board. We might next try to isolate xy directly, and we find that

x2
xy = 20 ^ 3^ “

Here, we have x2 ≥ 0 and y2 ≥ 0, so xy ≤ 20. But once again, showing that this value can be attained
stymies us. We only have xy = 20 in our equation if x2∕3 = y2∕3 = 0. But in this case, we have x = y = 0,
so xy = 0, not 20.
Each of our false leads has failed us when we looked at the equality condition of the Trivial Inequality.
In the first failed solution, the equality condition had x = -y, which failed. The second failed solution
offered x = 0 and y = 0 as the equality condition for the Trivial Inequality. That's a clear loser. What
might a successful equality condition be? Due to symmetry, a reasonable guess is x = y. So, we think
about what Trivial Inequality application has x = y as the equality condition. This leads us to (x - y)2.
The Trivial Inequality gives us (x - y)2 ≥ 0. Expanding the left side gives x2 - 2xy + y2 ≥ 0. Adding
5xy to both sides of this inequality to match the left side of the given equation yields x2 + 3xy + y2 > 5xy.
Because x2 + 3xy + y2 = 60, we have 60 > 5xy, so 12 > xy. We've been fooled before, so we check our
equality condition again. Here, the equality condition is x = y. Substituting this in our given equation
gives us 5y2 = 60, so y2 = 12. If y2 = 12 and x - y, we have xy = y2 = 12, so 12 is the maximum. □

Problem 12.28: Let fB be a rectangle with perimeter 36. Find the largest possible area of ,B.

Solution for Problem 12.28: Here's a common mistake in geometric optimization problems:

Bogus Solution: The maximum area occurs when the rectangle is a square. Since the
perimeter is 36, each side has length 9, so the area is 9- = 81.

402
 12.6. MAXIMA AND MINIMA

The answer is correct, but we've skipped the most important step: we haven't proved the statement
"The maximum area occurs when the rectangle is a square."
To do so, we let the side lengths of the rectangle be x and y. From the information about the perimeter,
we have 2x + 2y = 36, so x + y = 18. We wish to maximize the area, which is xy. We have an inequality
that relates x + y and xy. The AM-GM Inequality gives us

Since x + y = 18, we have yTy ≤ 9. Squaring both sides gives xy ≤ 81. This shows that the area of the
rectangle cannot be greater than 81. To show that it can possibly equal 81, we note that the equality
condition of AM-GM gives us x = y, which describes the case in which the rectangle is a square with
perimeter 36 and area 81. □

With Problem 12.28 as a guide, we're ready to tackle a tougher geometric problem.

use any more cardboard than they have to. A customer requests boxes that have a volume of 10001
cubic inches.
(a) Suppose the customer wants each box to have a top, a bottom, and all four sides. Show that
the BigBox Company should make boxes that are cubes in order to minimize the amount of ∣
cardboard they have to use.
(b) Suppose the customer wants each box to have a bottom and all four sides, but no top. What
dimensions should BigBox Company use to minimize the amount of cardboard they use on
each box?

Solution for Problem 12.29: We let the dimensions of the box be x, y, and z. The volume of the box is
xyz = 1000.

(a) We wish to minimize the surface area of the box, which equals 2(xy + yz + zx). This means we want
xy + yz + zx on the greater side of an inequality. A sum on the greater side of an inequality—this
is a job for AM-GM. Applying AM-GM to the nonnegative quantities xy, yz, and zx gives

≥ ^(χy)(yz)(zx) = ^2y⅛2.

We have xyz = 1000, so ^2J∕2z2 = 100. Multiplying our inequality above by 6 therefore gives
2(xy + yz + zx) ≥ 6 -^x2y2z2 = 600. The equality condition of AM-GM tells us that this surface area
occurs when xy - yz = zx. From xy = yz, we have x = z, and from yz = zx, we have y = x. So, our
minimum surface area occurs when x = y = z, which means that BigBox Company should make
cubes. If we have x = y = z = 10, then we do indeed have a box with volume 1000 in3 and surface
area 600 in2.
(b) The results of Problem 12.28 and part (a) of this problem are intuitively obvious. However, if
BigBox Company must make boxes without tops, then the answer is not so intuitively obvious.
They could go ahead and still make cubes of volume 1000 in3. Without the top, such a cubical box
would have surface area 500 in2. But can they do better?

403
CHAPTER 12. INEQUALITIES

Let x be the "height" of the box, so that y and z are the dimensions of the top and bottom. So,
the surface area of the box is 2xy + 2xz + yz. Again, we have a sum that we must minimize, so we
try AM-GM:
2xy + 2xz + yz > ^2xy)(2xz)(yz) = ^4x2y2z2 = 100λ∕4.

Multiplying by 3 gives us 2xy + 2xz + yz ≥ 300λ∕4. Since 300 ∖z4 ≈ 476.2, this does indeed suggest
we can make a box with less than 500 square inches of cardboard. To find out how to build the
boxes, we use the equality condition of AM-GM, which gives us 2xy = 2xz = yz. From 2xy = 2xz,
we have y = z. From 2xz = yz, we have y = 2x. Therefore, we must have 2x = y = z. So, the
dimensions of the bottom must each be twice the height of the box.
To find out what these dimensions are exactly, we let y = 2x and z = 2x in xyz = 1000 to find
4x3 = 1000, so x3 = 250. This gives us x = 5λ^, so y = z = 10√2. A topless box whose bottom is
a square with side length 10 λ∕2 inches and whose sides have height 5 λ^2 inches does indeed have
volume 1000 in3 and surface area 300λ∕4, so our minimum surface area is 300λ∕4 in2.

So far, all of our optimization problems have involved expressions with more than one variable.
We can also sometimes optimize one-variable functions using the inequalities we have studied in this
chapter. We've already seen examples of this when we studied quadratic optimization in Section 5.3.

_ n . r. . x2 +16
Problem 12.30: For x > 0, let ∕(x) = ——— Find the minimum possible value of ∕(x).

Solution for Problem 12.30: As written, it's not clear how we can minimize ∕(x). So, we look for other
ways in which we can write ∕(x). We can rewrite the function as a sum of two expressions, one with the
variable in the numerator, and one with the variable in the denominator:

We do so because this produces two expressions whose product is a constant. This suggests we can use
AM-GM to compare ∕(x) to a constant. (Note that we can only use AM-GM because x > 0, so both ∣ and
1∣ are positive.) AM-GM gives us
x , 16 /16 _ 4
3 ^l^ 3x
2 V 9 ^ 3'

Multiplying by 2 gives us ∕(x) ≥ 8/3. We have equality when x/3 = 16∕(3x). Solving this equation (and
remembering that we must have x > 0) gives us x = 4. We confirm that

42 + 16 8
∕(4) =
3∙4 3'
so the minimum value 8/3 can be achieved. □
We finish our study of maximization and minimization with a problem that doesn't quite fit the
mold of any of the inequalities we have studied in this chapter.

404
 12.6. MAXIMA AND MINIMA

Problem 12.31: Let x1, χ1,..., xn be a sequence of integers such that

(i) -1 ≤ Xj ≤ 2, for i = 1,2,3,..., n,

(ii) X] + X2 + ∙ ∙ ∙ + x,l = 19, and

2 + ∙ ∙ ∙ + x2 = 99
(iii) x^ + x*

Find the minimal and maximal possible values of x3 + x3 + ∙ ∙ ∙ + x3. (Source: AHSME)

Solution for Problem 12.31: The sum of squares suggests Cauchy, but before we go too far with Cauchy,
we note the odd restrictions on the x1∙. First, all the x1∙ must be integers. Second, each must be —1, 0,1,
or 2. This doesn't fit the restrictions of any of our inequalities from this chapter. Furthermore, it looks
pretty difficult to express the sum of the cubes of the xz∙ in terms of the sum of the x,∙ and the sum of the
squares of the x,-. So, we need to try something different than just applying our toolbox of inequalities.
We focus on the most unusual restriction, which is that each x,∙ is either —1, 0,1, or 2.

Concept: When stuck on an unusual problem, try focusing on the what makes the
o problem unusual.

We can use this idea to recast the problem by counting the number of x1∙ that equal each of —1, 0,1, or
2. We let a, b, c, and d denote the number of appearances among the x,∙ of -1, 0,1, and 2, respectively.
Now, we have 4 variables instead of n variables. Writing our two given equations in terms of a, b, c, and
d gives

—a + c + 2d — 19,
a + c + 4d = 99.

We want the minimum and maximum values of (-l)3α + 03b + l3c + 23d = -a + c + 8d. From the first of
our two linear equations above, we have a = c + 2d - 19, so the expression we want to maximize and
minimize is -a + c + 8d = -c - 2d + 19 + c + 8d - 19 + 6d. Now, our task is simply to find the minimum
and maximum possible values of d.
The only restrictions we have on d are the two linear equations in our system above. We must choose
d such that both a and c are nonnegative integers. So, we solve the linear equations for a and c in terms of
d. Subtracting the first equation from the second gives 2a + 2d = 80, so a = 40 - d. Adding the equations
gives 2c + 6d = 118, so c = 59 - 3d. For a to be nonnegative, we therefore must have d < 40. For c to be
nonnegative, we must have rf≤f = 191. Since d is an integer, this means d < 19.

Now, we're ready to maximize and minimize 19 + 6d. We know that d must be a nonnegative integer,
so the smallest we can allow d to be is 0, which gives us 19 + 6d = 19. To show that this case is possible,
we note that it gives us a = 40 - d = 40 and c = 59-3d = 59, so the xi∙ must have 40 -Is and 59 Is. If we let
these be all the xj∙, then the sum of the xl is indeed 19 and the sum of their squares is 40(-l)2 + 59(1)2 = 99.
To maximize 19 + 6d, we must have d as large as possible. Above, we found that d ≤ 19 and
d ≤ 40. Clearly, the former is more restrictive. So, we let d = 19, which gives us a = 40 - d = 21 and
c = 59-3d = 2. If the xl consist of 21 -Is, 2 Is, and 19 2s, then the sum of the x,is21(-l) + 2(l) + 19(2) = 19

405
CHAPTER 12. INEQUALITIES

and the sum of their squares is 21(-1)2 + 2(1)2 + 19(2)2 = 99, as required. So, the maximum value of
19 + 6d = 19 + 6(19) = 133 can be achieved.
To summarize, the minimum is 19 and the maximum is 133. □
Because the Trivial Inequality, AM-GM, and Cauchy can be used to solve so many problems, some
people fall into the trap of forgetting the rest of their problem solving toolbox when faced with an
inequality problem.

WARNING!! Not all optimization and inequality problems are straightforward ap­
plications of the inequalities we have studied in this chapter. Don't
forget all your basic problem solving skills when faced with an op­
timization or inequality problem. The Trivial Inequality, AM-GM,
and Cauchy are nice tools, but they won't solve every problem. Be
ready to use all your problem solving strategies: experiment, recast
the problem in different forms, focus on what makes the problem un­
usual, compare the problem to others you know how to do, and so
on.

UH
L∖(,∣( ise⅜

12.6.1 A box contains chips, each of which is red, white, or blue. The number of blue chips is at least
half the number of white chips, and at most one third the number of red chips. The number of chips that
are white or blue is at least 55. What is the minimum possible number of red chips? (Source: AHSME)
12.6.2 ]fa-7 = b + 3 = c- 9 = d + 22, which of a, b, c, and d is largest?
12.6.3 What is the largest possible volume of a rectangular box whose diagonal length is 12? Prove
that your answer is correct.
x2 — 2x + 2
12.6.4 Find the minimum value of — — for x > 1. (Source: AHSME)

12.6.5
* Find the minimum value of λ∕x2 + y2 if 5x + 12y = 60. (Source: AHSME)

12.7 Summary

In this chapter, we covered many strategies for proving that certain inequalities are always true, or that
these inequalities are true under specific conditions.
Many inequalities can be proved by using basic inequality manipulations.

WARNING!! Pay special attention to whether or not the variables can be negative
*∙4 in inequalities you wish to prove. Some common inequality manipu­
lations are only valid if the variables are always positive.

Three powerful tools for proving inequalities are:

406
 12.7. SUMMARY

• The Trivial Inequality. If x is a real number, then x2 ≥ 0. Equality holds if and only if x = 0.

• The AM-GM Inequality. The arithmetic mean of any n nonnegative numbers is greater than or
equal to the geometric mean of the numbers. That is, if a↑, a2, ...,an are nonnegative, then

«1 + ÷ * ÷ a∏ ≥ ∖[∩}02 ∙ ∙ ∙fl∏
n
Equality holds if and only if all α∣ are equal.

• The Cauchy-Schwarz Inequality. For any two sequences of n real numbers a↑, a2, ..., an and b↑,
½, ∙ ■ ∙, bn, we have

(ai + a2 + ‘, ‘ + fln)(^ι + ⅛2 + ’ ‘' + ⅛) ≥ + β2½ + ,' ’ + dnbn)2.

Equality occurs if and only if either αi = 0 for all i, or there exists a constant t such that b∣ = t∏i for
all i.

We then used these strategies to solve maximization and minimization problems:

Important: Maximization and minimization problems are two-part problems. We

must show that the maximum or minimum value is attainable, and that
no better value (no higher for maximization, no lower for minimization)
can be attained.

Problem Solving Strategies__________________________________________

We encountered so many useful strategies in this chapter, that we split the concepts summary into
inequality-specific strategies and general strategies.

Concepts: Useful general strategies discussed in this chapter include:

Q "lj ∙ When stuck on a problem, focus on why you are stuck, and try to

find ways to remove that obstacle.

• If a problem is given in terms of gigantic numbers, it's often useful

to replace those numbers with variables.

• Dealing with reciprocals of complicated fractions is often nicer than

dealing with the original fractions.

• Don't treat every problem as if it is brand-new. Try to relate problems

or parts of problems to other problems you already know how to do.

• When stuck on a problem involving quadratics, try completing the

square.

407
CHAPTER 12. INEQUALITIES

Concepts: Useful inequality strategies discussed in this chapter include:

(^--∣τ3
• Keep your eye on the ball! Don't just blindly manipulate
inequalities—try to make sure you're moving in the right direction.

• When possible, testing numbers that fit an inequality problem can

help you get started with a solution.

• Working backwards is a very powerful method for proving inequal­

ities.

• We can sometimes prove a complicated inequality by proving sim­

pler inequalities that can be combined to give us the desired compli­
cated one.

• Some equations can be solved by viewing them as the equality case

of a nonstrict inequality. We can then use the equality condition of
the inequality to find the solutions to the equation.

• When trying to prove an inequality involving nonnegative expres­

sions that have a constant sum or product, consider using AM-GM.

• Focus on one side of an inequality and the other will often fall into
place.

• It's often necessary to manipulate a desired inequality into a form

for which tools like AM-GM apply. Factor! Expand! Regroup terms!

• If an inequality involves a sum of squares or a product of two sums

with the same number of terms, try using the Cauchy-Schwarz In­
equality to prove it.

• Some inequalities result from setting variables or even entire se­

quences of variables equal to carefully chosen constants, and then
applying known inequalities to those variables or sequences.

• Cauchy can help us prove inequalities that have a sum of fractions

on the greater side. To try using Cauchy on such an inequality, let
the sum of the fractions be one sum of squares, and the sum of the
denominators be the other sum of squares.

Review Problems

12.32 Let x > y > 0 and z ≠ 0. Determine whether each of the following inequalities always holds. If it
does, then prove that it always holds. If it doesn't, then give an example for which it doesn't hold.
- , x ι∕
(a) xz2 > yzl (b) -7 > ⅛
z z

408
 REVIEW PROBLEMS

12.33 Show that if 0 < b < a < 1, then we must have------— > 1.
a—b
12.34 Show that if x < a < 0, then x2 > ax > a2. (Source: AHSME)
12.35 Show that if x - y > x and x + y < ∖
j, then both x and y must be negative.(Source: AHSME)
12.36 Show that if a + b > 0 then a3 + b3 ≥ a2b + αb2.
12.37 If 2⅛j < < 2⅛Γ' comPute the number of possible integer values of ,j.(Source: ARML)

12.38 Let s = x + y and p = xy for real numbers x and y. Show that s2 > 4p.
12.39 Let a, b, and c be nonnegative real numbers. Prove that a2b + b2c + c2a > Babe.
∩11 + 1 ∣ 17∕7^÷I
12.40 Let n be a positive integer, and let a,b≥0. Show that -----——— ≥ abt,.
n+1
12.41 Let a, b, c and d be integers with a < 2b,b < 3c, and c < 4d. If d < 100, find the largest possible
value for a. (Source: AHSME)

12.42 Let a, b, c, and d be positive numbers such that f < 3. Show that ∣ < < j.
12.43 A student attempted to compute the average, A, of x, y and z by computing the average of x and
y, and then computing the average of that result with z. The student's final result was B. If x < y < z, is
either A or B necessarily larger? (Source: AHSME)

12.44 Determine the minimum value of the sum rτ- + — + —, where a, b, and c are positive real
2b 4c 8a r
numbers. (Source: Mandelbrot)
12.45 A three-dimensional box has a surface area of 96 cm2. What is the maximum volume of the box?
4x2 + 8τ + 13
12.46 For x ≥ 0, find the minimum value of----- ------- -—. (Source: AHSME)
6(1 + x)
12.47 In counting n colored balls, some red and some black, it was found that 49 of the first 50 counted
were red. Thereafter, the balls were counted in groups of 8, and 7 balls in each of these groups were red.
Find the maximum possible value of n if 90% or more of the balls counted were red. (Source: AHSME)
12.48 A dealer bought n radios for d dollars total (where d is a positive integer). He sold two radios to
a community bazaar at half their cost. The rest he sold at a profit of $8 on each radio sold. If the overall
profit was $72, find the least possible value of n. (Source: AHSME)
12.49 Let a + b + c + d = l. Prove that a2 + b2 + c2 + d2 ≥

12.50 A gambler played the following game with a friend. The gambler bet half the money in his
pocket on the toss of a coin; he won on heads and lost on tails. The coin was tossed and the money
handed over. The game was repeated, each time for half the money held by the gambler. At the end,
the number of times the gambler lost was equal to the number of times he won. Did he gain, lose, or
break even?

12.51 What's the smallest possible value of k + m + n if k, m, and n are positive integers such that
ik + 1
m
+ in < 1?

12.52 Show that for all real numbers a, b, c, d we have (π2 - b2)(c2 - d2) < (ac - bd)2.

409
CHAPTER 12. INEQUALITIES

. . . fl2 + b2 + c2 a3 + b3 + c3
12.53 Show that for all positive numbers a, b, c, we have ———---- τ ≥ —∏---- τ.
r a3 + b3 + c3 a4 + b4 + c4
12.54 Let P, A, G, and Q be points on a line, in that order, with PG = a and GQ = b. Let A be be the
midpoint of PQ. Construct a semicircle with diameter PQ, and let M be a point on the semicircle such
that MG is perpendicular to PQ.

(a) Compute AM and GM in terms of a and b.

(b) Find a geometric proof of the 2-variable AM-GM Inequality.

12.55 Letx, y, and z be real numbers such that x2 + y2+z2 = 1. Find the maximum value of 3x + 4y + 12z.

Challenge Problems

12.56 Prove that the inequality 3α4 — 4a3b + b4 ≥ 0 holds for all real numbers a and b. Hints: 338
12.57 Let x, y, and z be positive real numbers such that xy2z3 = 108. What is the minimum value of
x + y + z? Hints: 355, 296
12.58 Let a↑,a2,... ,an> 0, and «1^2 ∙∙∙an = 1. Show that (1 + flι)(l + ai) ∙ ∙ ∙ (1 + an) ≥ 2n.
12.59 Find all ordered triples (x, y,z) such that 3x2 + y2 + z2 = 2x(y + z).
12.60 Find all ordered triples of real numbers (a, b, c) that satisfy the system of equations

a2 + 10b = -34,
b2 + 4c = 17,
c2 + 8a = -28.

12.61 Show that a2 ÷ ab + b2 ≥ 0 for all real numbers a and b. When does equality occur? Hints: 14

12.62 Find the ordered pair of real numbers (x, y) that satisfies the equation below, and demonstrate
that it is unique:
^ + A=42-9√≡-√y.
vχ √y
(Source: USAMTS) Hints: 215
x2 + y2 y2 + z2 z2 + x2
12.63 Show that for all positive x, y, and z, we have--------- +---------- +---------- > x + 1/ + z.
r ,7 x+y y+z z+x j
12.64 The sum of the following seven numbers is exactly 19: ai = 2.56, a2 = 2.61, «3 = 2.65, α⅛ = 2.71,
α5 = 2.79, ⅝ = 2.82, (i7 = 2.86. It is desired to replace each a1 by an integer approximation Al, 1 ≤ ∕ ≤ 7,
so that the sum of the A∣'s is also 19, and so that M, the maximum of the “errors" ∣√4∣ - rz∣∣, is as small as
possible. For this minimum M, what is 100A4? (Source: AIME)

410
 CHALLENGE PROBLEMS

12.65 Show that for all positive real numbers x ≠ 1 and nonnegative integers n, we have
X2"+1 _ 2
≥ 2n + 1.
xn+^l- xn
Hints: 338

12.66 Let 0 ≤ x < 4. Find the maximum value of x3(4 - x). Hints: 257
12.67 Show that if a + b + c = 3, then αb + k + cα ≤ 3. Hints: 282
12.68 Arrange the following products in increasing order from left to right:
1000! (400!)(400!)(200!) (500!)(500!) (600!)(300!)(100!) (700!)(300!).
(Source: ARML)

12.69 Show that for all positive integers n with n >2, we have
1 1 1
4- -------- + * ∙ ∙ + ---------- > n(21'n -1).
n n+1 2n-l
Hints: 145,4, 279
1 _ γπ+l 1 _ γ∏
12.70 Let 0 ≤ x < 1. Prove that------- — ≤---- —. Hints: 338,289
n+1 n
12.71 Let S be a list of positive integers—not necessarily distinct—in which the number 68 appears.
The average (arithmetic mean) of the numbers in S is 56. However, if 68 is removed, the average of
the remaining numbers drops to 55. What is the largest number that can appear in S? (Source: AIME)
Hints: 152
12.72 Let a, b, c, be positive real numbers. Prove that (l + ∣)(l + ^θ(l + j)≥2(l + Hints: 11

12.73
* At a wedding reception, n guests have assembled into m groups to converse. (The groups are
not necessarily equal sized.) The host is preparing m square cakes, each with an ornate ribbon adorning
its perimeter, to serve the m groups. No guest is allowed to have more than 25 cm2 of cake. Prove that no
more than 20 y[mn cm of ribbon is needed to embellish the m cakes. (Source: Mandelbrot) Hints: 233,32
12.74
* Find all values of k such that 36x2 + 25y2 ≥ kxy for all real values of x and y. Hints: 27

12.75
* Prove that for all real numbers a, b, x, and y, we have Vα2 + b2 + λ∕x2 + y2 ≥ √(tz - x)2 + (b - y)2.
Hints: 341
12.76
* Show that for positive numbers a, b, c, we have abc(a + b + c) ≤ ac3 + cb3 + ba3. Hints: 281
12.77
* Given that a, b, c, d, e are real numbers such that
u + b + c + d + e = 8,
a2 + b2 + c2 + d2 + e2 = 16,
determine the maximum value of e. (Source: USAMO) Hints: 81,107, 216
12.78
* Prove that if xz > 0 for all i then

(Source: Mandelbrot) Hints: 230, 309

411
CHAPTER 12. INEQUALITIES

12.79* Suppose that a, b, and c are positive numbers and ab + be + ca = 1/3. Show that
a
-------------- b
------------- c
- -∣- -------------- 1 t
a2 - be + 1 b2 - ca + 1 c2 - ab + 1 a+b+c
Hints: 311,176
12.80
* Let a, b, and c be positive real numbers such that abc — 1. Prove that

1 1 1 > 3
α3(b + c) b∖a + c) <?(a + b) 2
(Source: IMO) Hints: 226, 221

In the following problems, we prove the general AM-GM Inequality. You can use AM-GM for two
variables in the following problems, but you cannot use any other AM-GM inequalities until you prove
them.
12.81 Prove that zLtx+y+z > for all nonnegative numbers w, x, y, and z. Under what conditions
are the two sides equal?
12.82 Prove that the arithmetic mean of any 2” nonnegative numbers, where n > 1, is greater than
or equal to the geometric mean of those numbers. Under what conditions are these two means equal?
Hints: 340
12.83
* Show that if a, b, and c are nonnegative, then + + C ≥ y[abc. When does equality hold?
Hints: 16, 5
12.84
* Show that the arithmetic mean of any n nonnegative numbers, where n > 1, is greater than or
equal to the geometric mean of those numbers. Under what conditions does equality hold? Hints: 19,72

Extra! Here are a few more special inequalities:

(■+ iN+ MtlM+

• Chebyshev's Inequality. If a↑ ≥ a2≥ ≥ an and b↑ > b2 ≥ ∙ ∙ ∙ > bn, then

H A

∖ i=ι ∕

• Schur's Inequality. If x, ∖
j, and z are nonnegative real numbers and t is positive,
then
xt(x - y)(x — z) + yf(y - z)(y - x) + zf(z - x)(z - y) ≥ 0.

• Minkowski's Inequality. If p > 1 and the sequences {α∣∙} and [bl∣ consist of positive
numbers, then

412
Exponential growth looks like nothing is happening, and then suddenly you get this explosion at the end.
- Ray Kurzweil

CHAPTER 13 ι
I Exponents and Logarithms
Most of the expressions and functions we have worked with so far either are polynomials or are closely
related to polynomials. In the next four chapters, we will explore several other common types of
expressions and functions. These present some complexities that are quite different from those we faced
when working with polynomials. While it may at first feel like we are presenting a bunch of unrelated
tricks, where each trick is associated with a particular type of expression, we are really presenting many
different examples of a very important general strategy:

Concept: When working with an unusual expression in a problem, try to find ways
tθ convert the problem into one involving expressions you know how to
handle.

Think of most of the tactics you learn in the next four chapters as putting this strategy into practice.
Then, you won't have to memorize a bunch of different tricks—you'll have a strategy that you can use to
re-create whatever tool you need whenever you need it. Even better, you'll develop the ability to create
new tools when you encounter new unusual expressions in problems.

13.1 Exponential Function Basics

In the expression 23, the 3 is called the exponent and the 2 is called the base. We call a function ∕(x) = ax,
where fl is a positive constant, an exponential function with base fl.

Suppose g(x) = 2x. At right are values of g(x) for several values of x. Notice that as x X g(χ)
gets larger and larger, so does g(x). -2 0.25
0 1
In general, if ∕(x) = ax, where fl > 1, then f(x) grows as x grows. More formally we
say that a function f is a monotonically increasing function if f(x) > f(y) whenever x > y.
2 4
5 32
Often we will just say the function is "increasing" rather than "monotonically increasing."
All exponential functions ∕(x) = ax such that fl > 1 are monotonically increasing. Showing 8 256

413
CHAPTER 13. EXPONENTS AND LOGARITHMS

that exponential functions with base greater than 1 are increasing functions requires more advanced
tools than we study in this book.

Sidenote: Rigorously defining the output of exponential functions requires more ad-
vanced tools, too. For example, consider g(x) = 2x. It's very clear that
g(2) = 4 and g(5) = 32. We can also easily define the result when we input
rational numbers to g(x), such as g(3∕2) = 23^2 = (23)1^2 = 81/2 = 2 V2. But
what about g{τi)2 We have g(n) = 2π, but how do we raise 2 to the power
of n? We could estimate it with 2314, which we can evaluate because 3.14 is
rational. But this doesn't tell us how to define 2ττ. We have to rely on more
advanced tools to define g(x) = 2x for all real numbers x. You'll learn these
tools when you study calculus.

blems

Problem 13.1: Let f(x) = 2x and g(x) = Q∫∖ Graph f and g on the same Cartesian plane. Notice
anything interesting? If so, explain the relationship you see between the graphs.

Problem 13.2: Suppose that a is a constant greater than 1 and that ax = axJ. Must it be true that x = y?
Why or why not?

Problem 13.3:
(a) Find x if 5χ2-3 = 252x.
(b) Find t if327' = 273'.

13.1: Let f(x) = 2x and g(x) = (J) . Graph f and g on the same Cartesian plane. Notice
interesting? If so, explain the relationship you see between the graphs.

Solution for Problem 13.1: By choosing various values of x, we produce

the graphs of y = 2x and y = β) at right. The graph of y = 2x is bold.
We see that the graph of y = 2x is the graph of y = Q) reflected over
the x-axis. In other words, for any point (fl, b) on the graph of y = (∣V,
there is a point (-a, b) on the graph of y = 2x, and vice-versa. To see why
this is the case, we note that g(x) = Q) = (2-1)x = 2~x = f(-x). So, for
each point (a,g(a)) on the graph of g, there is a point (-«,/(-«)) on the
graph of f (and vice versa) with the same y-coordinate, but opposite
x-coordinate. □
The graph of y = 2x above exhibits what we call exponential growth.
When x is negative, the value of 2x increases very slowly as x increases. But when x is positive, we reach
a point where the value of 2x skyrockets as x increases. Every exponential function ax with fl > 1 exhibits

414
 13.1. EXPONENTIAL FUNCTION BASICS

this behavior. Even if a is very close to 1, the value of ax will appear to "explode" at some point as x
increases.
On the other hand, if 0 < a < 1 in ax, we have the behavior depicted by the graph of y = Q) .
* In this
case, as x increases, the value of ax decreases. As x gets very large, the value of ax gets very close to 0,
but never equals 0. This behavior is called exponential decay, which we'll discuss more in Section 13.6.
Problem 13.2: Suppose that a is a constant greater than 1 and that ax = ay. Must it be true that x = y? ⅛
Why or why not?

Solution for Problem 13.2: Let ∕(x) = ax. Because f is increasing, we know that ∕(x) > ∕(y) for any x and
y such that x > y. So, to test if we must have x = y when ∕(x) = ∕(y), we consider what happens in each
of the three possible cases, x > y, x < y, and x = y.
If x > y, then we must have ∕(x) > ∕(y) because f is increasing. Therefore, if x > y, then ax > a'-P
If x < y, then we must have ∕(x) < ∕(y) because f is increasing. Therefore, if x < y, then ax < a∙,.
If x = y, then ∕(x) = ∕(y), so ax = a*.

We must have either x > y, x < y, or x = y, and in the first two cases, we cannot also have ax = aV. So,
if ax = ay, the only possible relationship between x and y is x = y. (Notice that this is not true if a = 1.
However, if a = 1, then the function ∕(x) = ax is just ∕(x) = 1.) □
We can use similar reasoning to our solution in Problem 13.2 to show that if 0 < a < 1 and ax = ay,
then x = y. (You'll have a chance to do this on your own as an Exercise.)

Important: If a is a positive constant such that a ≠ 1 and ax = ay, then x = y.

φ
Let's see how this simple fact can be used to solve equations in which the variables are in exponents.

Problem 13.3:
(a) Find all x such that 5λ~^3 = 25x.
(b) Find all t such that 327' = 273'.

Solution for Problem 13.3:

(a) We don't know much about dealing with variables in exponents, so we'd like to convert this
equation into one in which the variables are not in exponents. We know that ax = ay implies x = y.
So, if we can rewrite the equation such that the base is the same on both sides, then we can produce
a simpler equation. Since 25 = 52, we have 25x = (52)x = 52x, so our equation now is

5x2-3 _ g2x.

Since the base is the same on both sides, the exponents must be equal. So, we have x2 - 3 = 2x.
Rearranging gives x2 - 2x - 3 = 0 and factoring gives (x - 3)(x + 1) = 0, so the two values of x that
satisfy the equation are x = 3 and x = -1.

415
CHAPTER 13. EXPONENTS AND LOGARITHMS

Concept: If you have an equation with constants raised to powers, it's often helpful
O≡≡∈> to write all those constants with the same base if possible.

(b) Our first step is to note that 27 = 33, so we can write each side as a power of 3. Specifically, on the
left we have 327' = = 333' and on the right, we have 273' = (33)3' = 33'3^ = 33,"', so our equation
is
33* = 33'+1.
Because the base of both sides is 3, the exponents of both sides must be equal, which gives us
33f = 3f+1. Again, the base of both sides is 3, so the exponents must be equal. This gives us
3f = t + 1, from which we find t = 1/2.

Lxer<i¼es

13.1.1 Find all t such that 16f = 82-f.

13.1.2 The equation 2x" = 323x+8 has two real solutions. Find their product.
13.1.3 A function f is monotonically decreasing if ∕(x) < ∕(y) whenever x > xj. Let g(x) = ax, where
0 < a < 1.

(a) Show that g is monotonically decreasing.

(b) Show that if g(x) = g(y), then x = y.

13.1.4 Find the maximum value of 3~χ2+2x+3.

13.1.5 Find all values of y such that 4y - 12 ∙ 2y = -32. Hints: 93

13.2 Introduction to Logarithms

In addition to helping us solve equations, the fact that exponential functions with base greater than 1
are monotonically increasing tells us something very important about them.

*rob1ems

Problem 13.4: Suppose that ∕(x) = ax, where a is constant such that a > 1. Show that ∕(x) has an
inverse. (If you don't remember what an inverse function is, read Section 2.4, then try this problem
again.)

Problem 13.5: Letg(x) = 2x. Find g 1(32).

Problem 13.4: Suppose that ∕(x) = ax, where a is constant such that ∕j > 1. Show that ∕(x) has an
inverse.

416
 13.2. INTRODUCTION TO LOGARITHMS

Solution for Problem 13.4: A function has an inverse if it does not ever give the same output for two
different inputs. Because f is increasing, we know that if x↑ > X2, then f(x↑) > f(x2f In other words, if
two inputs to f are not equal, then their outputs from f cannot be equal. So, f must have an inverse. □

Problem 13.5: Let g(x) = 2x. Find g-1(32). I

Solution for Problem 13.5: The previous problem tells us that g has an inverse. The value of g 1(32) is the
input to g(x) such that g(x) = 32. Therefore, we seek the value of x such that 2x = 32. In other words,
g-1(32) is the power to which we must raise 2 in order to get 32. Since 32 = 25, we find that ^^1(32) = 5.
□

The inverses of exponential functions are so common that they have their own name: logarithms.
To get a feel for logarithms, we'll look at a specific exponential function and its inverse. We write

x = log2 32
to indicate that x is the power to which we must raise 2 in order to get 32. In other words, the equations

x = log2 32 and 2x = 32
are equivalent. Since 25 = 32, we know that

log2 32 = 5.
Using this notation, we can now write the inverse of ∕(x) = 2x as ∕~1(x) = log2 x.
Just as 2 is the base of the expression 25, we say that 2 is the base of the logarithm log2 32. The base
of a logarithm must be a positive number, and it cannot equal 1. In the expression log232, the 32 is
sometimes referred to as the argument of the logarithm. When speaking, we say log2 32 = 5 as "The
logarithm base 2 of 32 is 5," and this means "The exponent to which we must raise 2 to get 32 is 5."
As we will see, much of understanding logarithms requires being able to convert between logarithmic
form, logrt b - c, and exponential form, ac = b.

Important: If a > 0, b > 0, and a ≠1, then

ψ logfl b = c and ac = b

are equivalent.

In this book, we only define logarithms for positive bases, and we exclude 1. (To see why we exclude
1, try to figure out what log11 is. We have ly = 1 for all real y, so log1 x would not be a function.) We
also will restrict each logarithm to having only real numbers in its domain and range.

Problems

Problem 13.6: Evaluate each of the following:

(a) 1°g24 (c) log216
(b) log2δ (d) logι1 af where a > 0 and a ≠ 1

417
CHAPTER 13. EXPONENTS AND LOGARITHMS

Problem 13.7:
(a) Find the domain and range of ∕(x) = 2x.
(b) Find the domain and range of g(x) = log2 X∙
(c) Graph f and g on the same Cartesian plane.

Problem 13.8: If log5 a = log5 b, then must we have a = b?

Problem 13.9: Solve the following inequalities:

(a) log4(2x - 3) > 2
(b) 1°gl∕5x>3

Problem 13.10:
(a) Find the domain of h(x) = log1/2 x.
(b) Find the domain of g(x) = log1/2 (log3 x).
(c) Find the domain of ∕(x) = log]/2 (log5 (log1∕3 x)).

Problem 13.11: Let x = log4128.

(a) Evaluate 4x.
(b) Evaluate 2x.
(c) Find x.

Problem 13.12: Evaluate each of the following:

(a) 1°gι6 8 (b) log3/24/9 (c) lθg3 √3 243

Problem 13.13: Suppose that a and b are constants such that ∕(x) = axb, ∕(3) = 14, and ∕(27) = 686.
Find a and b.

Problem 13.14: Find all ordered pairs of positive integers (x, y) that satisfy both y < 5 -x and y ≥ log2 x.

We start with a little practice computing a few simple logarithms.

Problem 13.6: Evaluate each of the following:

(a) 1θg24 (c) log216

(b) log28 (d) logfl aP, where a > 0 and a ≠ 1

Solution for Problem 13.6:

(a) Because 22 = 4, we have log2 4 = log2 22 = 2.
(b) Because 23 = 8, we have log2 8 = log2 23 = 3.

418
 13.2. INTRODUCTION TO LOGARITHMS

(c) Because 24 = 16, we have log2 16 = log2 24 = 4.

(d) We raise a to the power p to get ap, so logι7 ap = p.

Important: If a > 0 and a ≠ 1, then

logn cf = p.

One way to learn more about a new function is to find the domain, range, and graph of the function.
Problem 13.7:
(a) Find the domain and range of ∕(χ) = 2x.
(b) Find the domain and range of g(χ) = log2 x.
(c) Graph f and g on the same Cartesian plane.

Solution for Problem 13.7:

(a) We can raise 2 to any real power, so the domain of ∕(x) = 2x is all real numbers. All powers of
2 are positive, so there are no nonpositive numbers in the range of 2x. We can make 2x as small
a positive number as we like by choosing a suitably small negative number for x. For example,
when x = —20, we have 2x = 2-20 = 1∕220 ≈ 0.000000954. That's pretty small. Imagine what 2x is
if x = —200. So, the range of f is all positive real numbers.
(b) Because g(x) = log2 x is the inverse of ∕(x), the domain of g is the range of ∕, and vice versa.
Therefore, the domain of g is all positive real numbers and the range of g is all real numbers. We
can also see this by noting that log2 x equals the power to which we must raise 2 in order to get x.
So, the range is all real numbers because we can raise 2 to any real power. Similarly, the domain of
log2 x consists of all numbers that equal 2 raised to some power. As we just discussed, all positive
real numbers can be expressed as 2 raised to some power. So, the domain of g is all positive real
numbers.

Problem 13.8: If log5 a = log5 b, then must we have a = b?

Solution for Problem 13.8: Let both logarithms equal x. The equation x = log5fl means that 5x = a.
Similarly, x = log5 b means that 5x = b. Therefore, we have a = 5x = b. □

419
CHAPTER 13. EXPONENTS AND LOGARITHMS

Problem 13.9: Solve the following inequalities:

(a) log4(2x-3)>2
(b) logv5x>3

Solution for Problem 13.9:

(a) If log4(2x - 3) is greater than 2, then the expression 2x - 3 equals 4 raised to a power greater than
2. Because the function 4f is increasing, when we raise 4 to any power greater than 2, the result is
greater than 42. Therefore, we have 2% - 3 > 42. Solving this inequality gives x > 19/2.
We also could have tackled this problem in much the same way we solved Problem 13.8. We
start by assigning a variable to the logarithm, letting y = log4(2x - 3). Writing this in exponential
form gives 4'z = 2x - 3. Because y > 2, we have 4jz > 16. Therefore, we have 2x - 3 > 16, so
x > 19/2, as before.
(b) In part (a), we rewrote log4(2x - 3) > 2 as 2x - 3 > 42. So, we might be inspired to do this:

Bogus Solution: Because log1,5 x > 3, we have x > (1∕5)3, so x > 1/125.
__ X__________________________
At first glance, this looks fine. But then we test it. If the solution really is x > 1/125, then x = 1
should satisfy the inequality. However, we have log1∕51=0, which clearly is not greater than 3.
Where did we go wrong?
Our mistake was rewriting the logarithm inequality in exponential form without thinking about
what the inequality means.
WARNING!! Don't just perform algebraic manipulations blindly. Think about each
a>a step and make sure you know why it is valid.
If log1/5 x > 3, then we must raise 1/5 to a power greater than 3 to get x. Because 1/5 is less than
1, when we raise 1/5 to a power greater than 3, we get a number that is smaller than (1∕5)3, not
greater than (1∕5)3. Therefore, we must have x < (1∕5)3, or x < 1/125.
We're not quite finished, and testing our answer will reveal why. When we consider trying
numbers less than 1/125 in the original inequality, log1∕5x > 3, we see that the logarithm is not
defined if x is nonpositive, so we must have x > 0. Therefore, the full solution to the inequality is
0 < x < 1/125.
As with the previous part, we can also solve this problem by assigning a variable to the
logarithm. Letting y = log1/5x, we have (l∕5)∙⅛z = x, and the given inequality is y > 3. Since 1/5
is between 0 and lz we have (l∕5)zz < (1∕5)3, so x < (1∕5)3 = 1/125. Combining this with the
restriction x > 0 from the domain of the logarithm, we have 0 < x < 1/125, as before.

Problem 13.10: Find the domain of ∕(x) = log1/2 (log5 (l°gι∕3 x))∙

Solution for Problem 13.10: We can either work from the inside out, by starting with the base 1/3
logarithm, or we can work from the outside in, by starting with the base 1/2 logarithm. We'll work from
the outside in; see if you can find a solution working from the inside out.

420
 13.2. INTRODUCTION TO LOGARITHMS

We let ∕z(x) = logl^2x, which has all positive numbers as its domain. Next, we look at g(x) =
Iog1∕2^og5 χ)∙ We know that the domain of log1/2 x is all positive numbers, so the output of log5 x must
be positive in order for log1∕2(log5x) to be defined. The output of log5x is positive if and only if χ is a
positive power of 5. We have 50 = 1, and 5l is an increasing function, so all real numbers greater than 1
are positive powers of 5. Therefore, the domain of g(x) = log1/2 (log5 x) is (1, +∞).
Finally, we consider ∕(x) = log1/2 (log5 (log1/3x)), so that we have f(x) = g(log1/3x). We know that
the domain of g is (1, +∞), so, in order to be able to evalute g(log1∕3 x), we must restrict x to those values
for which log1∕3 x > 1. The statement log1∕3 x > 1 means that the power to which we raise 1/3 to get x is
greater than 1. Raising 1/3 to a power greater than 1 gives a number less than (1∕3)1, so x < j. Therefore,
the domain of f is (0,1 /3).

We can check our answer by noting that if 0 < x < j, then log1∕3 x > 1, so log5(log1∕3 x) > 0. Because
^0g5^0g1∕3 x) *s positive, we can take its logarithm with any positive base. □
Now that we have a feel for how logarithmic functions behave, let's compute a few more specific
logarithms.

Problem 13.11: Compute log4128.

Solution for Problem 13.11: Let x = log4128. We don't know much yet about dealing with logarithms in
equations, so we convert this equation to exponential form:

4x = 128.

We know how to tackle this equation; we write 4 and 128 with a common base, 2, and we have (22)x = 27,
so 22x = 27. The bases on both sides are the same, so the exponents must be the same. Therefore, we
have 2x = 7, so x = 7/2. □

Concept: We can often evaluate logarithms by setting the logarithm equal to x, then
Q ljlj writing the resulting equation in exponential form.

After we do so, we use the same tactic we used to solve exponential equations; we write constants
in the problem as powers of a common base. We could also have done this from the beginning with our
logarithm problem, so we have
log4 128 = log22 27.
Now, it's easy to see that 22 must be raised to the 7/2 power to get 27, so we have log4128 = log22 27 = 7/2.

Concept: Finding a common exponential base between two numbers a and b often
∣
Q - l> J helps compute logi7 b more easily.

Here's a little more practice.

Problem 13.12: Evaluate each of the following:

(a) ⅛68 (b) log3/24/9 (c) 1°g3√3243

421
CHAPTER 13. EXPONENTS AND LOGARITHMS

Solution for Problem 13.12:

(a) We let x = log16 8, so we have 16x = 8. Writing the constants with a common base gives (24)x = 23,
so 24λ' = 23. Therefore, we have 4x = 3, so x = 3/4.
We also could have noted that log16 8 = log24 23, and we must raise 24 to the 3/4 power to get
23, so log16 8 = 3/4.
(b) We let x = log3/2 4/9, so (3∕2)x = 4/9 = (2∕3)2 = (3∕2)-2. Therefore, x = -2.
(c) We let x = l0g3 √3 243, so (3 √3)λ' = 243. We have 3 √3 = 3 ∙ 31/2 = 33/2 and 243 = 35, so our equation
is 33xy,2 = 35. Therefore, we have 3x∕2 - 5, so x = 10/3.

□
Problem 13.13: Suppose that a and b are constants such that f(x) = axb, f(3) = 14, and ∕(27) = 686. ∣
Find c? and/’. '

Solution for Problem 13.13: Because ∕(3) = 14 and ∕(27) = 686, we have the system of equations
a ∙ 3b = 14,
a ∙ 27b = 686.

Dividing the second equation by the first eliminates a and leaves 27b∣3b = 49. The bases on the left
are both powers of 3, so we can simplify the left side by noting 27b∕3b = (33)6∕3b = 33b∕3b = 32b, which
makes our equation 32fo = 49. Taking the square root of both sides gives 3b = ±7. We can't have 3b = —7,
so our equation is now just 3b = 7. Now what? All this tells us is that b is the power to which we raise
3 in order to get 7. We can't find this power without a calculator or computer, but we can write it as a
logarithm: b = log3 7.
Since 3b = 7, we can use the equation a ∙ 3b = 14 from above to find a = 14∕3b = 2. □

Problem 13.14: Find all ordered pairs of positive integers (x, y) that satisfy both y ≤ 5 - x and ∣
y>log2x.

Solution for Problem 13.14: Solution 1: Casework. There aren't many pairs of positive integers that satisfy
the inequality y <5-x, because if x ≥ 5, there are no positive integers y such that y <5-x. So, we only
have to try the cases x = 1,2, 3, and 4.
When x = 1, we have y < 4 and y ≥ log21 = 0, so there are 4 solutions for this case with y as a
positive integer, (x, y) = (1,1), (1,2), (1,3), and (1,4).
When x = 2, we have y ≤ 3 and y ≥ 1 as our inequalities, so there are 3 solutions for this case,
(x,y) = (2,l), (2,2), and (2,3).
When x = 3, we have y < 2 and y ≥ log7 3 as our inequalities. Because 21 = 2 and 22 = 4, we know
that we must raise 2 to a power that is between 1 and 2 to get 3. So, we have 2 > log2 3, which means
(x, y) = (3,2) satisfies both inequalities. That gives us 1 solution for this case.
When x = 4, we have y ≤ 1 and y ≥ 2 as our inequalities. There are no solutions for this case.
This gives us a total of4 + 3 + l = 8 solutions.

422
 13.2. INTRODUCTION TO LOGARITHMS

Concept: Not all problems have nice neat solutions. Sometimes a little casework is
Q≡oJ required to find a solution.

Solution 2: Graphing. We graph the two inequalities as shown at right. The y

graph of ∖ j < 5 - x consists of the diagonal line shown and the region below
this line. The graph of y > log2 x consists of the graph of y = log2 x and the
shaded region above this graph. The points that satisfy both these inequalities
are the dark shaded region and the solid portions of the graphs of y - 5 - x
and y = log2 x.

We see that there are 8 lattice points (points with integer coordinates) with
positive coordinates that are either in the darkest shaded region, or on the solid
lines in the diagram. These points are in bold in the diagram. So, there are 8 ordered pairs of positive
integers that satisfy both inequalities. □

Our first solution is more rigorous than our second solution. While graphs can be excellent guides
to finding an answer, they are not proofs. For example, our graph above of y = log2 x appears to pass
through (2,1). But we can't just trust what the graph appears to be in order to be completely confident
in our answer. We should confirm that it passes through (2,1) by testing that (2,1) satisfies the equation
!/ = 1°§2 x∙ Similarly, our graph of y = log2 x appears to pass between the points (3,1) and (3,2). While
this looks obvious, our casework solution is more thorough in explaining why this is the case.

Concept: Graphing can be a useful tool when working with unusual equations or
o≡i5j≡) inequalities.

WARNING!! Graphs are excellent for providing intuition about a problem, but they
V shouldn't be used as a key part of a proof or a rigorous solution to a
problem.

Exercises
___________

13.2.1 Evaluate each of the following:

(a) log4256 (b) log31 (c) log1255

13.2.2 Evaluate each of the following:

(a) log12525 (b) log2y216 (c) log1/39
* «

13.2.3 Solve the equation 3 ∙ 2x — 9 = 14.

13.2.4 Solve the equation log2(x2 - 2x - 7) = 3.
13.2.5 Find the domain and range of each of the following:
(a) ∕(x) = logv5x (b) g(x) = log6(logv5x)

13.2.6* Let f(x) = log3 (x - 1) and g(x) = Find the domain of g(∕(x)). Hints: 68

423
CHAPTER 13. EXPONENTS AND LOGARITHMS

13.3 Logarithmic Identities

At right are exponential relationships sometimes re­ (i) <v) ⅛)w = 7

ferred to as exponent laws. In the next two sections, we
use these laws to develop an analogous set of identities (ii) x^=Xm~n (Vi) X~n = ⅛
for logarithms, then put those identities to use. (iii) (x"1)" = xmn Mi) (r=αr

I
(iv) (%y)", = xmym
Problems

3 Xx
Problem 13.15: In this problem, we find all positive values of x such that (x)(xlz'r) = —. (Source:
NYSML)
(a) Use the exponent laws to simplify both sides, writing the equation in the form = xs^x~
,.
(b) Use your equation from part (a) to find all values of x that satisfy the original equation. Don't
forget to check special cases!

Problem 13.16:
(a) Evaluate log2 8, log2 82, log2 83, and log2 84.
(b) How do you think logfl b and logfl bc are related?
(c) Let logfl b = x and logπ bc = y. Write these equations in exponential form.
(d) Use the exponential equations in part (c) to show that y = ex. Does this prove your intuition
from part (b)?

Problem 13.17: Prove that if a and b are positive and n is nonzero, then logβπ bn = logi7 b.

Problem 13.18:
(a) Evaluate log3 9 + log3 3 and log3 27.
(b) Evaluate log4 2 + log4 64 and log4128.
(c) Take a guess as to how we might simplify logfl b + logfl c.
(d) Let logι7 b = x and logrt c = y. Write these equations in exponential form.
(e) Prove that logfl b + logfl c = logβ bc.
(f) Evaluate log6 2 + log6 27 + log6 4.

Problem 13.19: Prove that logfl b - logfl c - logfl £.

Problem 13.20:
(a) Evaluate (log2 8)(log8 64).
(b) Evaluate (log3 9) (log9 ∣).
(c) Use the first two parts as a guide to guess how we might simplify (logιl b)(log⅛ c)∙
(d) Prove your guess.

424
 13.3. LOGARITHMIC IDENTITIES

log,, b
Problem 13.21: Prove that;------ = log,, b.
lθgrtc

Problem 13.22: Evaluate ;------ - - ∑----- -.

logw5 log45

Problem 13.23: Evaluate 3lo‰4.

We start off by practicing with some of the exponent laws.

Problem 13.15: Find all positive values of x such that (x) (x1/x)3 = —. (Source: NYSML)

Solution for Problem 13.15: We start by simplifying both sides using the exponent laws. On the left, we
have
x = x^ ∙ x⅜ — χ*
+ 4

On the right, we have — = xx~1, so our equation now is x

* +1 = xx~1. What's wrong with the following
solution:

Bogus Solution: Since the base on both sides is x, the exponents must be equal, so
we must have ∣ + 1 = x - 1. Multiplying both sides by x, and then
rearranging, gives x2 - 2x - 3 = 0, which has solutions x = 3 and
x = -1. Since x must be positive, we discard the solution x = -1, so
x = 3 is the only solution to the original equation.

The Bogus Solution does correctly identify those values of x for which the exponents are equal, and
x = 3 is a solution to the original equation. However, the solution is incomplete. There is a special case
we must consider. Specifically, if x is 1, then the equation xa = xb does not require a and b to be equal. If
x = 1 in the original equation, then both sides equal 1, so x = 1 is a solution. Therefore, the two solutions
to our equation are x = 1 and x = 3. □
In general, the case x = 1 is not the only special case to worry about when dealing with an equation
of the form xa = xb. If x = 0, then xa = xb for all nonzero numbers a and b. If we restrict a and b to being
integers, then if x = -1, we have xa = xb when a and b are either both even or both odd.

WARNING!! When working with equations involving exponential expressions,

*>4 watch out for special cases in which -1, 0, or 1 is the base.

Problem 13.16: Let logfl b = x and logfl bc = y. How are x and y related?

Solution for Problem 13.16: We start by exploring various logarithms to look for a pattern:

425
CHAPTER 13. EXPONENTS AND LOGARITHMS

log2 8 = log2 23 = 3,
log2 82 = log2 26 = 6,
log2 83 = log2 29 = 9,
log284 = log2212 = 12

Concept: When trying to find a relationship between two expressions involving vari-
θ--ljO ables, experiment by choosing simple values for the variables. Sometimes,
a clear pattern will emerge.

We see that in each case, log2 8c is c times log2 8. So, we guess that

logflfoc = clogflb.

But how do we prove it? We don't have a lot of tools for working with logarithms, but we know a lot
about exponents. So, we let

x= ⅛& and y = logfl bc,

and then write these equations in exponential form:

ax = b and αy = bc.

We believe that y = ex. We can relate x to y by substituting our expression for b in the first equation,
b = ax, into the second equation. This gives us

ay = {μx)c = acx.

Since the bases are the same in the equation αy = acx, we have y = ex, as desired. (Note that we do not
have to worry about the possibilities of a being -1, 0, or 1 in the equation aλj = acx, because a is the base
of a logarithm in the problem.) □
We have our first logarithmic identity:

Important:

ψ logafcc = clogflk

This identity is a direct result of the exponent laws.

Concept: Most logarithmic identities are the result of the exponent laws. There-
Q— 1∑n√) fore, we can usually prove logarithmic identities by converting them to
exponential form and using our exponent laws.

Problem 13.17: Prove that if a and b are positive and n is nonzero, then logrt,, b" = log(; b.
■MNMMMMMMMMHMHMMMI Ml MMMMMMMMMMMMM ∣

Solution for Problem 13.17: This problem looks a lot like the last one, so we try the same strategy.

426
 13.3. LOGARITHMIC IDENTITIES

Concept: When facing a problem that is very similar to a problem that you have
Oξl mJ already solved, try using the tactics you used on the solved problem to
tackle the new problem.

We start with
x = logπ b and y = logπ,, bn.
Converting these to exponential notation gives

x=b
a* and (flπ)V = bn.

As we did in the previous problem, we substitute our expression for b from the first equation into the
second. We then have (α")
* = (αx)". Because (αψ = a* n and (ax)n = axn, we have axn = a'Jn. Therefore, we
have xn = yn, which means x = y.

We also could have written this proof by raising both sides of ax = b to the πth power to give (flx)λ, = bn,
so a = bn, which tells us that (α',)x = bn. The equation (απ)x = bn means that raising an to the power of x
gives us bn, which means logfl,, bn = x, so we have logfl,, bn = logfl b. □

Concept: Your first solution to a problem often will not be the simplest solution.
O mJ Reviewing your first solution will sometimes reveal new, quicker methods.

Problem 13.18:
(a) Prove that logfl b + logi3 c = logfl be.
(b) Evaluate log6 2 + log6 27 + log6 4.

Solution for Problem 13.18:

(a) Exponential form helped us so much in the last two problems that we try it again here. We let

x = logfl b and y = logfl c.

We wish to show that logfl be = x + y. We write our equations for x and y in exponential form:

ax = b and ay - c.

Since we are interested in the product be, we multiply these two equations:

ax+* = be.

Writing this equation in logarithmic form gives the desired

logfl bc = x + y = logfl b + logπ c.

Notice that this identity is just another way of writing the exponent law ax+V = axay.

427
CHAPTER 13. EXPONENTS AND LOGARITHMS

(b)As an example of how this identity helps us, consider the expression

lθg62 + 1°g627 + log64∙

We can't easily evaluate any one of these logarithms. But when we use our new identity to
combine them, we have

log6 2 + log6 27 + log6 4 = log6(2)(27) + log6 4 = log6(2)(27)(4) = log6 63 = 3.

□
You won't be surprised that there is an identity for the difference of two logarithms.

Problem 13.19: Prove that logfl b - logfl c = logfl -.

I
Solution for Problem 13.19: Solution 1: Use exponential form. This looks so much like our previous identity
that we figure we can use the same approach. We let

x = logβ b and y = ⅛ c.

We wish to show that logfl - = x - y. We write our equations for x and y in exponential form:

ax = b and ay = c.

We want b∣c, so we divide:

ax _ b
ay c'
oχ _ b
Since — = ax~y, we have ax~v = -. Writing this in logarithmic form, we have

Solution 2: Use our other logarithm identities.

Concept: Once we have proved an identity, we can use it to prove new identities.

We've already proved an identity that involves adding logarithms, and we wish to prove that

logafc-logflc = logfl⅛.

Therefore, if we can turn the subtraction sign in front of logn c into an addition sign, we can use our
addition identity. Fortunately, we can use our very first identity to write

- ⅛ c = (-1) ⅛ c = ⅛ c"1 = ⅛ -∙

428
 13.3. LOGARITHMIC IDENTITIES

So, we have
logfl b - logi7 c = logfl b + logrt ∣ = logn(fr) (∣) = logfl

We also could have used the addition identity to deduce that logrt + logfl c = logπ β ∙ c) = logfl b, so
logflb - logac = logfl £. □

We can now combine sums and differences of logarithms that have the same base:

WARNING!! The logarithms in our sum and difference identities must have the same
*■ base. For example, we cannot use the addition identity to combine the
logarithms in the sum
log2 5 + log3 6.

We have identities for adding and subtracting logs; what about multiplying them?

Problem 13.20:
(a) Evaluate (log2 8)(logg 64).
(b) Evaluate (log3 9) (log9 ∣).
(c) Use the first two parts as a guide to guess how we might simplify (logπ b)(logfc c).
(d) Prove your guess.

Solution for Problem 13.20:

(a) Since log2 8 = 3 and log8 64 = 2, we have (log2 8)(logg 64) = (2)(3) = 6.
(b) Since log3 9 = 2 and log9 J = -⅜, we have (log3 9) (log9 ∣ j = (2) (-∣) = -1.

(c) Our final result in the first part equals log2 64, and the answer in the second part equals log3

(log28)(log8 64) = log264,

(log39) (log9 ∣) = log3 ∣.

Interesting! These both follow the pattern (logfl fc)(logi, c) = logβ c. Let's see if we can prove this
identity.
(d) As usual, we shift to exponential notation. We let

x = ⅛fe and z = logbc.

Therefore,
ax = b and bz = c.

429
CHAPTER 13. EXPONENTS AND LOGARITHMS

We wish to show that logrt c = (logfl b)(logb c) = xz. Writing logf7 c = xz in exponential form gives
us axz = c. Since we want axz, we raise both sides of the equation ax = b to the z power:

axz = bz.

However, we already know that bz = c, so we have axz = bz = c. Because axz - c, we have

logfl c = xz = (logfl b)(logb c),

as desired.

One step that helped guide our proof was working backwards, writing the equation we wanted to
prove, logπ c - xz, in exponential form, axz = c. This gave us an equation to aim for when working with
our exponential equations.

Concept: Working backwards can be very useful in guiding your exploration when
o≡≡≡g trying to find a proof.

We can use our multiplication identity to deal with dividing logarithms with the same base.

log,, b
Problem 13.21: Prove that------- = log. b.
logac

Solution for Problem 13.21: We know how to deal with the product of logarithms, so we multiply both
sides of the identity we'd like to prove by logfl c, which gives

logfl b = (logc fc)(logfl c).

But this is just the identity we proved in the previous problem! Because we have already proved that

(logc b)(logfl c) = logfl b,

we can divide this equation by logrt c to get the desired = logcb. □

⅛c
Once again, we worked backwards to find our solution. Then, we wrote our solution forwards,
starting from the multiplication identity and ending with the division identity we wanted to prove.

Problem 13.22: Evaluate ;------ - - ;---- -.

1°gιo5 1o845

Solution for Problem 13.22: We might start by writing the expression with a common denominator, which
gives us
4_______ 2 = 41og45-21og105
1°gιo5 l0g45 (lθg105)(log45)
Unfortunately, we don't have a good way to deal with the numerator or the denominator. None of our
identities gets us anywhere. One reason we can't simplify the difference in the numerator is that the

430
 13.3. LOGARITHMIC IDENTITIES

bases of the logarithms are different. However, in both cases, the argument (the number we are taking
the logarithm of) is 5.

We know how to simplify sums and differences of logarithms if their bases are the same, so we put
our wishful thinking hat on and try to find a way to turn the arguments of our logarithms into bases.

Concept: Wishful thinking is an excellent problem solving tool. When stuck on a

O ^, i-,lJ problem, ask yourself what, if it were true, would make the problem easier.

We do have an identity that turns the argument of a logarithm into a base, , — = log. b. This turns
102; c oc
the argument of the denominator on the left side into the base of the right side. So, if we let a = 10 and
c = 5, we have

This is true for any value of b. Choosing b = 10 gives us —-— = log5 10 because log1010 = 1. Similarly,
1 , 1°gιo5
log b
letting a = b in —- = logc b gives us

Applying this to our problem, we have —— = log510 and — = log54, so our expression is

i⅛-⅛=4⅛1°-2⅛4∙

Now, we can use some of our other identities to evaluate the expression:

⅛5 ^ ⅛ = 41°g510 ^ 21°ffe 4 = 1°g51°4 “lθffe 42 = foffe (¥) * '°fe625 = 4'

□
Note that we discovered a new identity on the way to our solution:

Important:

ψ
Actually, this isn't a new identity at all! As we saw, it's just a special case of an identity we already
logfl b
knew, ------- = logc b, which itself is just a rearrangement of another identity we already proved,
(logfl c)(l°gc ty = l°gn b- So, there's really no need to memorize all three identities. If you know this last
identity, you can quickly produce the others when you need them.

431
CHAPTER 13. EXPONENTS AND LOGARITHMS

We finish our discovery of identities with an identity that involves both exponents and logarithms.

Problem 13.23: Evaluate 3⅛4.

Solution for Problem 13.23: What makes this expression so complicated is the very unusual exponent.
Rather than diving into our exponent and logarithm tools right away, we stop and think about what this
expression means.

Concept: Don't be intimidated by unusual notation. Often, it represents a very

Q-'- -mJ simple idea.

Here, we focus on what the exponent means. The expression log3 4 equals the power to which we must
raise 3 in order to get 4. Therefore, when we raise 3 to this power, by definition, the result is 4. □
Similarly, for any positive numbers a and x with a ≠ 1, the expression logfl x is the power to which
we must raise a to get x, so fllo‰ x = x.
Putting all the identities we proved in this section in one place, we have:

Important: logabc = clogflb

ψ loga,i bn = loga b
logfl b + loga c = loga be

⅛fe-⅛c = logfl
(loga b)(logb c) = logac

1 = l°gc b
logα c
α1°g
* x=x

Hopefully you understand these identities so well that you do not need to memorize them. After
working with them for a while, and proving them on your own, they should become as clear to you as
the exponent laws we listed at the start of the section. In fact, most of these are simply restatements of
exponent laws in different forms.

WARNING!! Be careful when you apply the laws of logarithms. Here are some
V common mistakes that students make when working with logarithms:

⅛(x + J∕) = ⅛ × + logb y,

⅛
* _1 χ
ιθg∕, y °8b yz
(logfc x)(logb y) = logbxy.

None of these are true in general!

432
 13.4. USING LOGARITHM IDENTITIES

I *e ⅜
ver<i

13.3.1 Evaluate each of the following without a calculator:

(a) log3 (272007) (d) (log25)(log512) + (log27)(log7∣)

log2125
(b) 61o86418 (e)
log2 25
21og4108
(c) log2 j+log26 (f)
*
21o84 3

13.3.2 Prove that logb 1 = - ⅛fl.

13.3.3 If c = logv b, c ≠ 0, and d = 2 logy3 (b3), then find the value of

13.3.4 Prove that if ∣a∣ ≠ 1, ∣c∣ ≠ l,ac ≠ 0, and we have both ax = ccι and c'z = az, then xy = qz.
13.3.5 To the nearest hundredth, log10 2 is .301 and log10 3 is 0.477. Which of the following is the best
approximation of log5 10?
8 9 10 11 12
7, 7' 7' 7' 7 '
(Source: AHSME)
*
13.3.6 Prove that (logfl b)(logc d) - (logc b)(logfl d). Hints: 249

13.4 Using Logarithm Identities

In this section, we use our logarithm identities to solve more challenging problems. In most of these
problems, our main strategy is to use our identities to simplify expressions and equations. In fact, that's
the main strategy for solving a great deal of algebra problems:

Concept: Use basic tools to simplify expressions and to convert complicated equa-
Q ~^□ tions into simpler equations you know how to solve.

Try using this strategy on the following problems.

Problems

Problem 13.24: In this problem, we find all values of x such that 2 log3(x + 4) - log3(4x - 11) = 2.
(a) Use logarithm identities to write the left side as a single logarithm.
(b) Convert your equation from part (a) to exponential form, and solve the resulting equation.
(c) Confirm that your solutions satisfy the original equation.

433
CHAPTER 13. EXPONENTS AND LOGARITHMS

Problem 13.25: Find all ordered pairs of real numbers (xt y) that satisfy the system of equations

1°gl0g) = 5,

1°gιo (χ2y5) = u∙

Problem 13.26: Suppose that 3x = 12. Find 9x 1.

Problem 13.27: Suppose I have a calculator that only computes base 10 logarithms. How can I use my
calculator to compute log2 3?

Problem 13.28: In this problem, we find the largest real value of b such that the solutions to the
following equation are integers:
(log2ιo X2b) =lθg210x4.

(a) Rewrite the equation so that the logarithm terms are all log2 x instead of log2ι0 x2b and log2w x4-
(b) Solve the problem. (Source: ARML)

Problem 13.29: The solutions to the system of equations

1θg225x + 1<‰y = 4'

logx 225-logy 64 = 1,

are (x↑, yι) and (x2, y∑)∙ Find log30(x1y1X2y2)∙ (Source: AIME) Hints: 47,169

∣ Problem 13.24: Find all values of x such that 2 log3(x + 4) - log3(4x - 11) = 2.

Solution for Problem 13.24: We use our logarithm identities to simplify the left side. First, we have
21og3(x + 4) = log3(x + 4)2, so our equation is log3(x + 4)2 - log3(4x - 11) = 2. We now have a difference
of logarithms with a common base, which we can write as

log3(x + 4)2 - log3(4x - 11) = log3

(x + 4)2
This makes our original equation log3 = 2. Writing this in exponential form, we have
4x-ll
(x + 4)2 2
4x -11
We know how to solve this equation. We multiply both sides by 4x -11 and expand (x + 4)2 to give

x2 + 8x+16 = 9(4x-11).

Rearranging this equation gives x2 - 28x + 115 = 0. Factoring the quadratic yields (x - 23)(x — 5) = 0, so
the solutions are x = 23 and x = 5.

434
 13.4. USING LOGARITHM IDENTITIES

We test these solutions by substituting them in the original equation. For x = 23, we have

2 log3 27 - log3 81 = 2(3) - 4 = 2,

so this solution works. For x = 5, we have

21og39 - log39 = 2(2) - 2 = 2,

so this solution works, too. □

Ouι first example exhibited two powerful strategies for solving equations involving logarithms:

Concept: Equations involving several logarithms are often solved by using loga­
O≡≡≡ rithm identities to write the equation with a single logarithm.

Concept: Many logarithmic equations can be solved by converting them into expo­
O---- cr⊂>
nential equations, then solving the exponential equations.

However, we must be careful:

WARNING!! When you solve a logarithmic equation, be on the lookout for extrane­
>Λ
A ous solutions. The best way to avoid extraneous solutions is to check
that your solutions satisfy the original equation.

For example, consider the equation log2x + log2(x + 2) = 3. Combining the logarithms on the left
gives log2(x2 + 2x) = 3, and converting to exponential form gives x2 + 2x = 23 = 8. Solving this equation
gives x = 2 and x = -4 as solutions. While x = 2 does satisfy the original equation, the value x = -4
does not, because log2(-4) and log2(-2) are not defined. This extraneous solution is produced when we
add log2x and log2(x + 2) to get log2(x2 + 2x). The expression log2(x2 + 2x) is defined for x = -4, even
though log2 x and log2(x + 2) are not.

Another way we can be on the lookout for extraneous solutions when working with an equation
like log2 x + log2(x + 2) = 3 is to start by determining those values of x for which the expressions in the
equation are defined. Since log2 x is defined only for x > 0 and log2(x + 2) is defined only for x > -2, we
know that any valid solution must satisfy x > 0.

Problem 13.25: Find all ordered pairs of real numbers (x, y) that satisfy the system of equations

1°gw (* 2y5) = n∙
Solution for Problem 13.25: Solution 1: Logarithm identities. The mix of variables inside the logarithms is
the chief obstacle, so we start by rewriting the system with the variables separated:
log10x3-log10y4 = 5,
log10 x2 + log10 y5 = 11.

435
CHAPTER 13. EXPONENTS AND LOGARITHMS

Concept: Just as our sum and difference identities can be used to combine logarithms,
Q ■ ι^Ξ) they can also be used to separate logarithms of products and quotients into
easier to handle logarithms.

Unfortunately, we can't yet combine these equations to eliminate x or y. However, we can use the
identity logfl bc = c logfl b to write both equations in terms of log10 x and log10 y∙ This gives us

3 log10 x - 4 log10 y = 5,
21og10x + 51ogι0y = 11∙

Adding 5 times the first equation to 4 times the second eliminates log10 y and gives us 23 log10 x = 69,
so log10 x = 3. Therefore, we have x = 10* 3 = 1000. Substituting this into either of our original equations
gives log10 y = 1, so y = 10. Therefore, the solution to our system of equations is (x, y) = (1000,10).
Solution 2: Convert to exponential form. Exponential equations are often easier to work with than
logarithmic equations, so we rewrite the system as

x2y5 = 1011.

Now, we can eliminate y by taking the first equation to the fifth power and multiplying it by the second
equation raised to the fourth power:

(χ2y5)4 = (105)5 ∙ (1011)4.

Simplifying both sides of this equation gives us x23 = 1069, so x = 103. Substituting this value into the
equation x2y5 = 1011 and solving for y, we get (x, y) = (1000,10). □

Problem 13.26: Suppose that 3x = 12. Find 9x 1.

Solution for Problem 13.26: Solution 1: Exponential manipulation. We are working with two different
powers of 3, so we start by writing everything as a power of 3. We already have 3x = 12. We write
9x^1 as 9x^1 = (32)x^1 = 32(x_1) = 32x~2. If we can write this in terms of 3x, we can substitute 3x = 12 to
evaluate it. Fortunately, it's not hard to do so:

32x _ (3x)2 _ (12)2

= 16.
32^ - 9 “ 9

Solution 2: x is a logarithm. From 3x = 12, we have x = log3 12. Substituting this into 9x 1 gives us

9iog312-1

We only have one tool in our bag of tricks for dealing with logarithms as exponents:

436
 13.4. USING LOGARITHM IDENTITIES

In order to use this identity, we must rewrite 9lo8312-1 so that the complicated exponential expression is in
the form αlo‰ b. We must therefore have only a logarithm in the exponent, and the base of the exponential
must be the same as the base of the logarithm. We could isolate the logarithm in the exponent first:

9iog,12-1 _ 9l0g312 _ (32)1°δ312 _ 32l°g312 3iog3144 l44

91 - 9 ~ 9 - 9 - ~9~ ~ 16"

We could also note that log312 = log32122 = log9 144, so

9iog√2-1 _ 9⅛12 _ 9l0‰2122 giogoin 144

91 ~ 9 ^ 9 - V - 16∙

Or, we could use identities to write the exponent as a single logarithm:

We also could have evaluated 9lo‰4 by noting that log3 4 = log32 42 = log9 16, so 9lo&4 = 9loS<>16 = 16. □

Problem 13.27: Suppose I have a calculator that only computes base 10 logarithms. How can I use
my calculator to compute log2 3?

Solution for Problem 13.27: Because my calculator only understands base 10 logarithms, I can only
compute numbers of the form log10 x. However, I need to compute something that doesn't have any
10's at all! So, we look through all our logarithm tools in search of an identity that has a base on one
side of the equation but doesn't have that base on the other side. Our division identity stands out:

We want log2 3, and our identity gives us the way to get it:

So, to compute log2 3,1 compute log10 3 and log10 2 with my calculator, then divide the former by the
latter. □
Problem 13.28: Compute the largest real value of b such that the solutions to the following equation
are integers:
2
(lθg210 X2b) = lθg2]0 x4.

(Source: ARML)
Solution for Problem 13.28: The two logarithmic expressions are pretty complicated, so we start by
looking for ways to simplify them. First, we can remove the exponents from the arguments, x2b and x4,
and we have
2
(2blog2wx) =41og2ι0x.

437
CHAPTER 13. EXPONENTS AND LOGARITHMS

Now, the two logarithm expressions are the same. We can simplify them both even further by noting
that
log2ι0 x = log2 x1/10 = — log2 x,

so our equation becomes

(∣log2x) =∣log2x.

We multiply both sides by 25 to get rid of the fractions, which gives us

b2(log2 x)2 = 10 log2 x.

Since the logarithm expressions are the same, we bring all terms to the left and factor the resulting
expression to give
(b2 log2 x - 10) log2 x = 0.

WARNING!! Notice that we did not divide both sides by log2 x to cancel out a factor
λλ of log2x from both sides of b2(log2x)2 = 101og2x. This is because
log2 x might be 0, in which case we cannot divide by it. Instead, we
rearrange and factor, so that we don't lose track of the fact that the
original equation is true when log2 x = 0.

The equation (b2 log2 x - 10) log2 x = 0 tells us that either log2 x = 0, which gives x = 1 no matter what b
is, or b2 log-, x = 10. Isolating the logarithm gives log2 x = 10∕b2, and converting this to exponential form
gives
x = 2≡2.

In order for 21°∕fc2 to be an integer, the expression 10∕b2 must be a positive integer. The largest value
of b for which 10∕b2 is an integer is V10. For any value of b with b > V10, we have 10∕t>2 < 1, so we
have 1 < 210^2 < 2. Therefore, the largest value of b for which all solutions to the original equation are
integers is b = VW∙ □
The key to our solution is using logarithm identities to simplify the logarithm expressions in the
equation. In fact, this has been the key to most of the problems in this section:

Concept: When working with a complicated equation involving logarithms, try

Q) ■■ cτj using logarithm identities to simplify the expressions involved. Common
ways to do so are:

• Simplify the bases or arguments of logarithms using the identities

logfl bc = c logfl b and logrt,, b" = log,7 b.

• Combine logarithms of the same base with the sum, difference, or

quotient identities.

• Break logarithms of products or quotients into simpler pieces with

the sum or difference identities.

438
 13.4. USING LOGARITHM IDENTITIES

Our final two points seem to contradict each other. Here's where a little problem-solving savvy
is required. We combine logarithms with our identities when the result is a simpler expression. For
example, if we have log-, y + log2 ξ, then using our addition identity cancels the 7s and gets rid of the
fractions, leaving
32 7 /32 7 ∖
log2 7 + * 1°g2 4 = 1*°g2 (7 ■ 4) = lθg2 8 = 3.

On the other hand, when confronted with log10 y in Problem 13.25, breaking the logarithm up gave
us 31og10x - 41og10 y, which has much simpler logarithms. Notice that we can also use our "breaking
logarithms up for simplification" strategy to evaluate log2 y + log2 ?:
l°g2 7 + 1°g2 | = 1°g2 32 - log2 7 + log2 7 - log2 4 = 5 - 2 = 3.

Problem 13.29: The solutions to the system of equations

log225 x + log64 y = 4,
logx225-logy64 = 1,

are (xι,yι) and (x2, y2). Find log30(x1y1X2y2). (Source: AIME)

Solution for Problem 13.29: First, we note that the logarithms in the second equation are the reciprocals
of those in the first equation, so our system of equations is
* + 1°g64y = 4'
1°g225
_!________ 2_ = 1.
log225x log64y
Now, at least the logarithms are the same. Rather than continuing to write log225 x and logb4 ∖
j over and
over, we let a = log225 x and b = log64 y to make the algebraic manipulation simpler. Our system is now
α + b = 4,
14=ι∙
a b
This is a lot less difficult to deal with.

Concept: Substituting variables for complex expressions in equations can make the
Q—∙crj equations easier to solve.

We solve the first equation for a to find a = 4- b. Substituting this in the second equation gives

4-b b
Multiplying both sides by (4 - b)b gives b-(4-b) = (4 - b)b, and rearranging gives b2 - 2b - 4 = 0. If we
find b, then we can find y, since the equation b = log64 y tells us that y = 64fc. Unfortunately, the roots of
b2-2b-4 = 0 are irrational. However, looking at the expression we seek,

lθg30χiyiχ2y2,
we see that we don't actually have to find y1 and y2. We only need their product.

439
CHAPTER 13. EXPONENTS AND LOGARITHMS

Concept: Keep your eye on the ball. Sometimes we don't need to find the values of
Q ul) all the variables in a problem to solve the problem.

Fortunately, we can find this product without ever finding the values of b that satisfy b2 - 2b - 4 - 0.
Suppose b↑ and b2 are the solutions to this equation, so we have yι = 64bl and y2 = 64z’2. Then, we have

y1y2 = 64bl ■ 64fa2 = 64bl+∖

Since b↑ and b2 are the solutions to the equation b2 - 2b — 4 = 0, Vieta tells us that bι +b2 — —(—2)∕1 — 2.
So, we have yιy2 = 642.
Similarly, from a = log225 x, we have x = 225a. Since a = 4 - b, we have x = 2254^i,. So, from our two
values of b, namely b↑ and b2, we get two values of x, which we call Xi and x2, and we have

x1x2 = 2254~bl ∙ 2254~b2 = 2258~(bl+b2) = 2256.

So, we have log30 xιyιx2y2 = log30(642)(2256) = log30(212 ∙ 312 ∙ 512) = log30 3012 = 12. □

13.4.1 If r and s are the roots of 3x2 - 16x + 12 = 0, then find log2 r + log2 s.
log. i≡
13.4.2 Given that log1017 = r and log10 2 = s, find the value of ^θ^1° in terms of r and s.

13.4.3 Find all t such that 2 log3(l - 5t) = log3(2f + 5) + 2.

13.4.4 Let X], x2,..., xn be a geometric sequence of positive real numbers. Prove that for any positive
number b, with b ≠ 1, the sequence logfc X∖, logb x2,..., logfc xn is an arithmetic sequence.

13.4.5 Suppose the ordered pair (x, y) satisfies------- —— - -. If y is increased by 50%, by what fraction
1°gιo(p) 2
must x be multiplied to keep this equation true? (Source: ARML)
13.4.6 A line x = k intersects the graph of y = log5 x and the graph of y = log5 (x + 4). The distance
between the points of intersection is 0.5. Find k. (Source: AHSME)

13.4.7 For all integers n greater than 1, define

1
' logπ2002∙

Let b = «2 + «3 + «4 + «5 and c = O10 + flu + flι2 + «13 + «14- What is b - c? (Source: AMC12) Hints: 219
13.4.8 Compute the integer k, where k > 2, for which

log10 (fc - 2)! + log10 (k - 1)! + 2 = 2 log10 k∖.

(Source: ARML) Hints: 86

440
 13.5. SWITCHING BETWEEN LOGS AND EXPONENTS

13.5 Switching Between Logs and Exponents

We've already switched from logarithmic notation to exponential notation as a step in several problems.
In this section, we highlight this strategy with some more challenging problems. We also introduce the
reverse strategy: using logarithms to simplify complicated exponential expressions.

Problems

Problem 13.30: If log10 5 ≈ 0.699, how many digits are in 52006?

lo‰c = clogl,fl.
Problem 13.31: In this problem, we prove that α*
(a) Simplify logfl αlogι> c. Simplify logσ clog⅛a.
(b) Use part (a) to prove that βlo‰c = c1°8i,β.

Problem 13.32: Suppose that p and q are positive numbers for which log9p = log12i∕ = log16(p + q)-
What is the value of q/p? (Source: AHSME)

Problem 13.30: If log10 5 ≈ 0.699, how many digits are in 52006?

Solution for Problem 13.30: There's no way we're going to compute 52006, so we need to find some way
to count the digits of a number without knowing the number exactly. There are some large numbers
whose digits are easy to count: powers of 10. For example, 1054 is a 1 followed by 54 zeroes, so it has 55
digits. We can use this observation to count the digits in other numbers, as well. For example, 3.5 × 1063
starts with "35," and then has 62 zeroes (one power of 10 turns 3.5 into 35: 3.5(1063) = 35(1062)).
So, we'd like to write 52006 in terms of powers of 10. This gives us the idea of writing the equation
log10 5 ≈ 0.699 in exponential form, since this will give us 5 as a power of 10:

100 6" ~ 5

Therefore, we have
52006 ≈ (1q0.699)2006 _ 1q(0.699)(2006) _ 1θl402.19

This tells us that 52006 is between 1014°2 and 101403. Since 1014°2 has 1403 digits and 1014°3 is the smallest
positive base 10 number with 1404 digits, the number 52006 must have 1403 digits.

We could also have solved this problem by taking the base 10 logarithm of 52006:

log1052006 = 20061og105 ≈ (2006)(0.699) = 1402.19.

Putting this in exponential form gives us 52006 ≈ io140219, and the rest of our solution proceeds as before.
□

Problem 13.30 is an example of how we can take the base 10 logarithm of a number to count the
digits in the decimal expansion of the number.

441
CHAPTER 13. EXPONENTS AND LOGARITHMS

Problem 13.31: Prove that α1°81,c = cl°81,a

Solution for Problem 13.31: If the base of the logarithm in αlog<,c were a instead of b, we'd know how
to simplify it. But it isn't, so we'll have to find some other way to deal with such a complicated
exponent. We do at least have an identity that gives us a way to turn exponents into factors of a product:
logv yz = z logγ y. So, we can get logb c out of the exponent of Λlog<,c by taking the logarithm of fllogf,c. But
what base should we use?
We try using a, since we have

logβ αlo⅛c = (logfc c)(logfl a) = logf, c,

which is pretty simple. Let's see what happens when we take the base a logarithm of clog⅛fl.∙

logflc1°8^ = (⅛fl)(⅛c)∙

Back in Problem 13.20, we showed that (logi, α)(logfl c) = logfc c, so we have

logβ α1°8fc c = logb c = logrt c1°8⅛a.

We have logrt x = logfl y if and only if x = y, so the equation above gives us the desired α1°8*> c = clo‰a.

Our proof does overlook the fact that a could be 1, in which case we can't take the logarithm base a.
Jf a = 1, we have alo⅛ι>c = 1 and c1θ8<>fl = clogU = c0 = 1, so we still have alo⅛>c = clogfca. We also could have
gotten around this difficulty by seeing that we could use any base when we took the logarithms of tzlog⅛c
and clog⅛a. Using the multiplication identity in Exercise 13.3.6, we have

logrffl1°8l,c = (logb c)(logrf fl) = (logd c)(logt, a) = ⅛c⅛β.

In our last step, we used the logarithm identity logχyn = nlogxy in reverse, writing the product
(logz, fl)(logt/ c) as logdc1°8i,n. Because the bases are the same in k>gdfllogfrc = logdclo‰fl, so are the ar­
guments: fllogf,c = clog',β. □
Problem 13.31 gives us a strategy for dealing with complicated exponents:

Concept: If you have an expression containing a complicated exponent, take the

Q u□ logarithm of it to turn the exponent into a factor.

WARNING!! This "take the logarithm to remove complicated exponents" strategy

V does not work on sums or differences of expressions. For example,

logπ(bx + cy) does not equal x log(? b + y logιj c

Problem 13.32: Suppose that p and q are positive numbers for which log9p = log12 q ~ log16(p + q).
What is the value of q/p? (Source.

442
 13.5. SWITCHING BETWEEN LOGS AND EXPONENTS

Solution for Problem 13.32: All the bases are different, as are the arguments p, q, and p + q. Moreover, the
bases aren't all powers of the same integer. So, there's no obvious way to use our identities.
In other problems, when we've been stuck working with logarithms, we've tried using exponential
notation instead. Let's try this strategy here. We start with log9p = log12^. Writing this in exponential
form gives
9k‰^=p θr 121°S^ = φ
Yuck. It might be possible to find a solution from here, but those are pretty forbidding equations.
We ve introduced new variables to simplify complicated equations before. Let's try that here:

x - log9 p = log12 q = log16(p + q).

This gives us much simpler exponential equations:

9λ - p, 12x = q, and * = p + q.
16

We might then note that this makes our original equation 9x + 12* = 16x, but this doesn't seem to help
us with q∕p. Instead, we divide the second equation by the first to find a simple expression for q/p:

(i∖x
p~ 9x ~∖3J ‘

But how can we find (4∕3)x7

* — p + q, and we expect we'll have to. Comparing this to our other

We haven't used the equation 16
equations, we note that 16x∕12x = (4∕3)x, which is the quantity we seek. So, we divide 16x = p + q by
12x = q to give
P + <7 = P +1
<7 q
So, we have
1 = E+1.
p 9
Letting f = this means we have t = ∣ + 1, so we have t2 - t - 1 = 0. Applying the quadratic formula
gives
1±√5
2 -
We must have t > 0, since = (∣)λ > 0. Since is negative, the only possible value of j is f =

We could also have divided our equation 9x + 12x = 16x by 9x to get 1 + (∣)λ = (∣)2∖ because
(γ) = (3) ∙ Letting t = (∣) leads to the same quadratic as before.□

Recall our story about the magic camel from page 382. In Problem 13.32, the variable x is a magic
camel—we don't ever find x, but we use it to recast the problem in a more convenient form that allows
us to solve the problem.

443
CHAPTER 13. EXPONENTS AND LOGARITHMS

Exercises
II l ≡H≡B≡≡ffi≡≡≡BillB

13.5.1 Suppose that log10 xy3 = 1 and log10 x2y = 1. What is log10 xy? (Source: AMC 12)
13.5.2 If 60λ = 3 and 60b = 5, then find i2∣(1-fl-W2(i-M. (Source: AHSME)

13.5.3 If log2p = log3 r = log3617, then determine pr. (Source: Mandelbrot) Hints: 67
13.5.4 Suppose that α > b > 0. Solve the equation (α4 - 2a2b2 + b4)y~1 = (a — h)2y(fl + b)-2 for y in terms
of a and b.
13.5.5 Find all x such that χlo8ιox = (Source: AHSME)

13.6 Natural Logarithms and Exponential Decay

■

Problems

∕ 1 ∖π 111 1
Problem 13.33: Let f(ri) - 11 + - I andg(ri) = + 3[7 h~ 2j + 3? + ’'' + ’
(a) Evaluate ∕(1), ∕(2), ∕(5), and ∕(10). (Yes, you can use a calculator on this problem.)
(b) Evaluate g(l), g(2), g(5), and g(10).
(c) Use a calculator or computer to evaluate f(ri) and g(n) for large values of n. Notice anything
interesting?

We start this section by introducing a special constant that, much like π, has a great many uses in
mathematics and science.
∕ 1∖" 111 1 I
Problem 13.33: Let f(n) = 11 + - I and g(n) = 1+ p + 2!+5y + "' + ^∙ Use a calculator or computer I
to evaluate f(n) and g(n) for large values of n. Notice anything interesting?

Solution for Problem 13.33: The table below shows f(ri) and g(n) to the nearest 10 8 for several values

n ∕(w) g(»)
1 2.00000000 2.00000000
2 2.25000000 2.50000000
5 2.48832000 2.71666668
10 2.59374246 2.71828180
100 2.70481383 2.71828183
1000 2.71692393 2.71828183
100000 2.71826824 2.71828183
10000000 2.71828169 2.71828183
It looks like both f{ri) and g(n) approach constant values as n gets very large. Moreover, it looks like

444
 13.6. NATURAL LOGARITHMS AND EXPONENTIAL DECAY

they might approach the same constant value. □

In fact, ∕(π) and g(n) in Problem 13.33 do indeed approach the same constant value as n gets very
large. We call this constant e, named after the great mathematician Leonhard Euler. (Euler is pronounced
"oiler.") Not until we have the tools of calculus can we really appreciate the many fascinating properties
of e, but e does appear in a great many scientific contexts. The base e logarithm is also frequently used in
mathematics and science. We call the base e logarithm the natural logarithm. Rather than writing loge,
we usually write In, so In x = loge x.

WARNING!! Many mathematicians will also write log without a base to indicate the
> natural logarithm. However, it is also common to write log without a
base to indicate the base 10 logarithm. In fact, it's so common that such
a logarithm is sometimes called the common logarithm. Because of
the ambiguity surrounding the use of log without an indicated base,
we will avoid doing so in this book. When you see it used in other
sources, you'll have to use context to decide which of logt, and log10
is intended.

The letter e and exponential functions are frequently used to describe natural processes. One example
is radioactive decay. A radioactive substance decays into non-radioactive substances over time. It does
so following a process called exponential decay. Under exponential decay, the same proportion of a
substance decays in any fixed time t. In other words, the fraction of a 100 gram sample of a radioactive
substance that decays in 3 seconds is the same as the fraction of a 30 kilogram sample of the same
substance that decays in 3 seconds.

The half-life of a radioactive substance is the amount of time it takes for exactly half the radioactive
substance to decay. Because the same proportion of a substance decays in any fixed time t, the half-life
of a substance is the same no matter how much of the radioactive substance is present. So, if the half-life
of a substance is 40 days and initially there are 400 grams of the substance, then after 40 days there will
be 200 grams of radioactive substance remaining, and after another 40 days, there will be 100 grams,
and so on.
I

Problems

Problem 13.34: Initially there are 8 grams of a radioactive material in a container. The half-life of the
material is 2 days.
(a) After how many days will there be 0.5 grams of radioactive material left in the container?
(b) Suppose there are x grams of the radioactive material remaining after 5 days. Use the amounts
of radioactive material remaining after 4 days and after 6 days to determine x. (Warning: The
answer is not 1.5.)
(c) Let ∕(x) be the number of grams of the substance that has not decayed after x days. Graph
y = ∕(x).
(d) Let fι, f2, and f3 be three amounts of time after the material is placed in the container such that
⅛ - h = ⅛ - ⅛∙ (For example, we could have h = 2 days, t2 = 3.5 days, and f3 = 5 days.) Let
∕(x) be the function you found in part (c). Show that ∕(f2)∕∕(f1) = ∕(⅛)∕∕(f2)∙

445
CHAPTER 13. EXPONENTS AND LOGARITHMS

Problem 13.35: Suppose that a radioactive element has a half-life of t↑∕2, and that we run an experiment
in which we initially place No atoms of this element in a container. After a time t, suppose there are
Nu atoms of the radioactive element remaining in the container. In this problem, we will show that

t χ ln(Nw∕N0)
1/2 ’ 0.693 '

(This formula frequently appears in science textbooks.)

n √∕fι∕2 Nn
(a) Explain why I- I
No­
(b) Take the natural logarithm of both sides of part (a) to show that (t∕t1∕2) 1∏(1∕2) - ln(Nπ∕N0).
ln(N√N0)
(c) Show that t = -⅛∕2 ∙
In 2

∕1∖
1√∕fv2 m .. ln2
Problem 13.36: Explain why the formula I- I = — can be written as Nn = N0e , where Λ = -—
No t1∕2

Problem 13.37: Dr. Curie took two measurements of a radioactive sample exactly 100 seconds apart.
There were 403 grams of the sample in the first measurement and 83 grams remained undecayed in
the second measurement.
(a) Find the half-life of the sample.
(b) If she measures the sample a third time, 38 seconds after the second measurement, how many
grams of undecayed material will there be in the sample?

Problem 13.34: Initially there are 8 grams of a radioactive material in a container. The half-life of
the material is 2 days.
(a) After how many days will there be 0.5 grams of radioactive material left in the container?
(b) How much of the radioactive substance will remain after 5 days?
(c) Let ∕(x) be the number of grams of the substance that has not decayed after x days. Graph
)∙
y = ∕(
* ⅛J
(d) Let tι, t2, and t3 be three amounts of time after the material is placed in the container such that
t2-t↑ = t3-t2. Let ∕(x)be the function you found in part (c). Show that ∕(⅛)∕∕(h) = ∕(⅛)∕∕(⅛)∙ ⅛1

Solution for Problem 13.34:

(a) Because the material has a half-life of 2 days, exactly half the material decays in 2 days. So, after
two days, there are 8/2 = 4 grams of the radioactive material remaining. Over the next two days,
half the remaining radioactive material decays, leaving 4/2 = 2 grams of the radioactive material.
Two days later (6 days from the beginning), there is 2/2 = 1 gram of radioactive material. Finally,
two days later, which is 8 days from the beginning, there are 1/2 = 0.5 grams of radioactive
material.
(b) Above, we saw that after 4 days, there are 2 grams of radioactive material, and after 6 days, there

446
 13.6. NATURAL LOGARITHMS AND EXPONENTIAL DECAY

is 1 gram of radioactive material. Let there be x grams of radioactive material after 5 days. The
same fraction of the remaining radioactive substance decays each day. So, the fraction of the
radioactive substance that decays from day 4 to day 5, which is (2 - x)∕2, must equal the fraction
of the radioactive substance that decays from day 5 to day 6, which is (x - l)∕x. Therefore, we
have
2-x _ x-1
~T~ ^ ~x~'
Cross-multiplying gives 2x - x2 = 2x - 2, from which we find x = √2.

(c) Every 2 days, the amount of radioactive substance remaining is

divided by 2. So, after x days, the amount of radioactive substance
remaining has been divided by 2 exactly x/2 times, and we have

This is an exponential function! Its graph is shown at right.

(d) We can check that the function ∕(x) satisfies our definition of ex­
ponential decay. Specifically, suppose that times fι, ⅛∕ arid ⅛ are
equally spaced with fι < t2 < ⅛∙ Then, we have f2 - fι = ⅛ - ⅛∕ arid

f(ti) _ 8(5)2 _ ∕i∖a*il

∕(⅛) = 8(⅛)2 = ∕i∖s⅛a

Because t2 - fι = ⅛ - ⅛∕ we have ∕(t2)∕∕(f1) = ∕(⅛)∕∕(f2)∙ We can subtract both sides of this

equation from 1 to give
_ ∕(⅛) = _ ∕(⅛)
∕(fi) ∕(f2)'

which means that the fraction of the sample that decays from time f1 to time t2 is the same as the
fraction of the sample that decays from time f2 to time f3.

□
We still haven't seen e. But if you open a science text, you'll usually find e in any discussion of
exponential decay.

Problem 13.35: Suppose that a radioactive element has a half-life of t1∕2, and that we run an
experiment in which we initially place No atoms of this element in a container. After a time t, ∣
suppose there are N„ atoms of the element remaining in the container. Show that

i ι ln(Nπ ∕N0)
'≈^fiz2' ' 0.693 '

447
CHAPTER 13. EXPONENTS AND LOGARITHMS

Solution for Problem 13.35: We start with the definition of half-life. In any time period of length fι∕2,
the number of atoms of the radioactive element is halved. In a time period of length t, there are t∕t1∕2
time periods of length fι∕2- (For example, if fι∕2 = 3 days and t = 12 days, then 12/3 = 4 half-lives have
transpired.) Therefore, the number of atoms of the radioactive element is halved t∕t1∕2 times in a time
period of t. So, we must have
Nn = ∕1√λv2
No ∖2∕
Taking the natural logarithm of both sides gets the f ∕f 1 /2 out of the exponent, and introduces the ln(Nπ∕N0)
term that's in the formula we want to prove:

Solving for t gives us

± i ↑n(Nn∕N0)

This looks close, but our desired formula has both a negative sign and that 0.693. Where did those come
from? We pull out our calculator and find that In ∣ ≈ -0.693. Aha! Now, we have our desired

+ ln(Nn∕N0)
1/2 ’ 0.693 ’

You might be wondering why science books report a formula that relies on taking the natural
logarithm of both sides of
Nn = ∕l√zfv2
No ∖2∕ ∙

We might instead write the equation as = 2 tftl'2, then take the base 2 logarithm of both sides to find

This gives us a formula for t without any constant to remember like the 0.693 in the previous formula
we found.

WARNING!! Formulas, particularly those involving logarithms, can be written in

*
*> many different forms.

The reason we use e and In in so many formulas in math and science is because of many nice
properties of the function ∕(x) = ex that we need calculus to explain.
You certainly don't want to memorize all the different forms in which a formula can be written, but
it's very important to be able to manipulate a formula from one form to another.

448
 13.6. NATURAL LOGARITHMS AND EXPONENTIAL DECAY

∕1 ∖
√,∕,U2 M
Problem 13.36: Explain why the formula I -I∣ = τ-p∙ can be written as Ntl = N0e~λt, where
∖ - —
h/2

Solution for Problem 13.36: We can isolate Nll by multiplying both sides by No to give

Now, our task is to write the j term with e as the base. We want e as a base, so we write 1/2 as a
power of e. Specifically, we have 1/2 = t so we have

M, = n0QPNnd(^≡)''',,2.
We need a negative sign and In 2, so we note that ln(l∕2) = In 1 - In 2 = 0 - In 2 = - In 2, and we have

= No = Noe~t(ln2)/t1'2.

Letting Λ = (ln2)∕t1∕2, we have Nn - N0e~λt. □

Notice that taking the natural logarithm of this final formula gives us

lnNπ = ln(N0e~λf) = lnN0 + lne^λf = lnN0 - λt,

so
= lnNπ - lnN0 = t1,2]n(Nn∕N0)
-Λ ln2
which is the formula we found back in Problem 13.35. Hopefully, this convinces you of the futility of
memorizing all the different forms of these types of formulas. All of them can be derived from:

Important: If the half-life of a substance is tι∕2, and there is initially No of the sub­

ψ stance, then the amount remaining after a time t is given by the function

/1 ∖f∕h∕2
∕ω = N0(-) .

Problem 13.37: Dr. Curie took two measurements of a radioactive sample exactly 100 seconds apart.
There were 403 grams of the radioactive sample in the first measurement and 83 grams remained
undecayed in the second measurement. If she measures the sample a third time, 38 seconds after the
second measurement, how many grams will remain undecayed in the sample?

Solution for Problem 13.37: If we can find the half-life of the sample, we can determine what fraction
of the sample remains radioactive 138 seconds after the first measurement (which is 38 seconds after

449
CHAPTER 13. EXPONENTS AND LOGARITHMS

the second measurement). We know that in 100 seconds, the initial sample of 403 grams of radioactive
material decays to 83 grams. So, if t1∕2 is the half-life of the sample, then
100∕tl∕2
-=∩
403 ∖2∕
To get the variable out of the exponent, we take the natural logarithm of both sides (we could use any
base for the logarithm). This gives us

1 83 1 ∕l∖i00∕fv2 _ 100 1
lnSδ3=ln
∖2∕ f1∕2 2

Solving for fj∕2, we find

100 Ini
f1∕2 =----- ≈ 43.9 seconds.
1∩ _22_
ul 403
Now that we have the half-life, we can find the amount of the sample left in the third measurement.
This measurement occurs when t = 100 + 38 = 138, so the amount remaining is
, -1 χ138/43.9
403 M ≈ 45.6 grams.

We can quickly sanity-check this answer. The half-life is approximately 43.9 seconds, so in the 38 seconds
after Dr. Curie measured 83 radioactive grams remaining, we should expect a little less than half of this
amount to decay. Indeed, the 45.6 grams we expect to remain is a bit more than half of 83 grams.
Notice that we didn't have to rely on memorizing any formulas to solve the problem. We simply
used the definition of half-life. □

Exercises

13.6.1 There are 20 grams of a radioactive substance at the start of an experiment. In the first 10 minutes
of the experiment, 15 grams of the substance decays.

(a) What is the half-life of the substance?

(b) How much of the substance remains radioactive 24 minutes after the experiment starts?

A
13.6.2 Solve the equation-----—— = D for t in terms of A, B, C, D, and k.
B + Ce~κt
13.6.3 400 kilograms of a radioactive element is buried. The half-life of the element is 120.3 years. At
the end of the year 2314, the buried material is unearthed, and 96.2 kilograms of the radioactive element
remains. In what year is the material buried?
13.6.4 Find a real-valued function ∕(f) that satisfies f(t + 2) = ⅛∕(f) for all real values of t.

13.6.5 Three grams of a radioactive substance is placed in a vial. Find the half-life of the substance if
the number of grams of the substance remaining after t seconds is 3e-5f.

13.6.6 Solve y = In (2 - e'j) - In 3 for y.

450
 13.7. SUMMARY

13.7 Summary

We call a function ∕(x) = ax, where a is a positive constant, an exponential function with base a. We say
that a function f is a monotonically increasing function if ∕(x) > ∕(y) whenever x > y. If « > 1, then the
exponential function ∕(x) = ax is monotonically increasing.

Important: If a is a positive constant such that a ≠ 1 and ax = a',, then x = y.

ψ
We write x - logfl b, where a and b are positive and a ≠ 1, to indicate that x is the power to which we
must raise a to get b. In other words, the equations
x = logfl b and ax = b
are equivalent. We call the expression logfl b a logarithm, where a is the base of the logarithm and b
is the argument of the logarithm. The logarithmic function ∕(x) = logrt x and the exponential function
g(x) = ax are inverses of each other.

Logarithms satisfy the following identities:

Important: logflαP = p
ψ logabc = clogπb
loga,l bn = logfl b
∖oga b + logn c = logfl be
logβ b - logfl c = logfl ⅛

(logfl b)(logb c) = logfl c

⅛b 1 ,
1 = 1°gc b
logπ c
aio^x = x

We use the letter e to denote a special constant that equals the value that the expression (1 + ⅜)
approaches when n is very large. This constant appears in many mathematical and scientific contexts,
but we require calculus to fully understand many of the interesting properties of e. Since e is so common
in exponential and logarithmic expressions, the special notation In is often used to denote the base e
logarithm.

One place that e often appears in science is in the study of radioactive decay. The half-life of a
radioactive substance is the amount of time it takes for exactly half the radioactive substance to decay.

Important: If the half-life of a substance is h∕2, and there is initially No of the sub­
stance, then the amount remaining after a time t is given by the function

451
CHAPTER 13. EXPONENTS AND LOGARITHMS

Things To Watch Out For!_____________________ .—_------------------------------

• Don't just perform algebraic manipulations blindly. Think about each step and make sure you
know why it is valid.

• Graphs are excellent for providing intuition about a problem, but they shouldn't be used as a key
part of a proof or a rigorous solution to a problem.
• When working with equations involving exponential expressions, watch out for special cases in
which -1, 0, or 1 is the base.
• The logarithms in our sum and difference identities must have the same base. For example, we
cannot use the addition identity to combine the logarithms in the sum log2 5 + log3 6.
• Be careful when you apply the laws of logarithms. Here are some common mistakes that students
make when working with logarithms:

Bogus Solution: logb(x + y) = logfc x + logi7 y,

⅛y
⅛xlogby = logf,xy.

None of these are true in general!

• When you solve a logarithmic equation, be on the lookout for extraneous solutions. The best way
to avoid extraneous solutions is to check that your solutions satisfy the original equation.

• Formulas, particularly those involving logarithms, can be written in many different forms.

Problem Solving Strategies —- -------------------------------------------------------------

Concepts: ∙ When working with an unusual expression in a problem, try to find
θ - caΞ) ways to convert the problem into one involving expressions you
know how to handle.

• If you have an equation with constants raised to powers, it's often

helpful to write all those constants with the same base if possible.

• We can often evaluate logarithms by setting the logarithm equal to

x, then writing the resulting equation in exponential form.

• Finding a common exponential base between two numbers a and b

often helps compute logrt b more easily.

Continued on the next page. . .

452
 REVIEW PROBLEMS

Concepts: . . continued from the previous page

• Wishful thinking is an excellent problem solving tool. When stuck

on a problem, ask yourself what, if it were true, would make the
problem easier.

• Equations involving several logarithms are often solved by using

logarithm identities to write the equation with a single logarithm.

• Many logarithmic equations can be solved by converting them into

exponential equations, then solving the exponential equations.

• Just as our sum and difference identities can be used to combine

logarithms, they can also be used to separate logarithms of products
and quotients into easier to handle logarithms.

• If you have an expression containing a complicated exponent, take

the logarithm of it to turn the exponent into a factor.

Review Problems

13.38 Find all values of n such that 62 ■ (6")" = 6w ∙ 6” ∙ 6".

13.39 Find the real value of t such that 2f+1 + 2f = 23 + 25 + 27~t.
13.40 Find the value of a given that 4 ∙ 9fl - 21 = 25 ∙ 3fl.
13.41 Evaluate each of the following:
(a) log3243 (b) log4⅜ (c) log10W)T

13.42 Evaluate the following:

(a) log4 12 + 21og43 - 31og46 (b) log27∙log72

13.43 Find x such that log2 (log2 (log2 x)) = 2. (Source: AMC 12)
13.44 Find the value of b if 4rt = 81’ and 3f, = 2 ∙ 3α.

13.45 Solve the following inequalities:

(a) log5(3 -x)≥7 (b) log1/2(2x) > 3

13.46 Let n be a positive integer. If the number of integers in the domain of f(x) = log ((1 - λ)(.v - rι))
equals 2n - 6, compute n. (Source: ARML)
13.47 Compute 3λ, where A = ^∙°-⅛-∙9..~ ⅛ (Source: Mandelbrot)
(log3 1 - log3 9)(log3 8 - log34)

453
CHAPTER 13. EXPONENTS AND LOGARITHMS

13.48 Prove that logfl,, b = - logfl b∙

13.49 Find all ordered pairs (a, b) such that logb(9tz) = 3 logrt b = -3.
13.50 Compute the numerical value of 5logιo2 ∙ 2logιθ3 ∙ 2logιo6 ■ 5logιo9. (Source: NYSML)

13.51 If log10 5 = x, find the value of log2 5 in terms of x.

13.52 At what point do the graphs of y = log2(x + 3) and y = 3 - log2(x - 1) intersect?
13.53 What is the value of the expression
1 1 1 1
log2100! + log3100! + log4100! + ’ ‘ ’ + log100100!'

where 100! = 100 ∙ 99 ∙ 98 ■ 97....... 2-1? (Source: AHSME)

13.54 The value of log3 2 is 0.631, rounded to the nearest thousandth. Compute the smallest positive
integer n for which 3" > 2102. (Source: Mandelbrot)
39
13.55 Solve 241 = 2 + ∑ logιo ^or rea^ value(s) °f x∙ (Source: NYSML)
k=0
13.56 Solve the equation a + bine = d for c in terms of a, b, and d.
13.57 A certain radioactive isotope has a half-life of 4.3 seconds. If 4.2 ounces of the isotope remain 12
seconds after an experiment starts, how much was present when the experiment began?
13.58 My clock reads 1:01 PM when I take a magic pill. Every time the minutes number on my clock
changes from 01 to 02, my height increases 50%. Every time the minutes number changes from 31 to 32,
my height decreases 40%. I was 6 feet tall when I took the pill. What time is it when I am first less than
1 foot tall?
13.59 Jamie forgot to take a measurement of how much of a radioactive sample she had at the start of
an experiment. She also doesn't know what the half-life of the sample is. She did note that there were
12.2 grams exactly 63 seconds after she started the experiment, and 4.1 grams remained radioactive
exactly 107 seconds after the experiment started. How many grams in the sample were radioactive at
the start of the experiment?

∣
Challenge Problems

13.60 Solve the equation (x + 2)2y - x2 = 3 for y in terms of x. For what values of x is there no value of
y that satisfies the equation?
13.61 A triangle with sides a < b ≤ c is log-right if log10 (fl2) + logl0 (b2) = log10 (c2). Compute the largest
possible value of a in a right triangle that is also log-right. (Source: ARML) Hints: 322
13.62 The graph, C, of y = log10x is rotated 90o counterclockwise about the origin to obtain a new
graph, G'. Find an equation whose graph is C'. (Source: AMC 12) Hints: 121, 318

454
 CHALLENGE PROBLEMS

13.63 If x, ∖
j > 0, logv x + logv ι∕ = 10/3, and xy = 144, then find (x + y)∕2. (Source: AHSME) Hints: 222
13.64 For integers x and y with 1 < x,y < 100, compute the number of ordered pairs (x, y) such that
logγ y + loglz x2 = 3. (Source: ARML)

13.65 Find the product of the positive roots of √1995 xlog∣w∙r = χ2. (Source: AIME) Hints: 335
13.66 lf ⅛ = ⅛'then c = ck- Compute k. (Source: ARML) Hints: 302

13.67 I
183,344

13.68 The points A — (xι,ι∕j) and B = (x2,y2) are two points on the graph of y = Iog2x. Through the
midpoint of the line segment AB, a horizontal line is drawn to intersect the graph at C = (x3,1∕3)∙ Prove
thatx∣ = x1χ2. (Source: CEMC)

13.69 Find x in terms of a and y given that y = Hints: 229

13.70 Let x = 2lo⅛ 3 and y = 3lo8∣> 2. Find x - y. Hints: 335

13.71 If log2 a + log2 b ≥ 6, then find the smallest possible value ofa+b. (Source: AHSME) Hints: 344,40
13.72 The equation 2333x^2 + 2lllx+2 = 2222x+1 + 1 has three real roots. Find their sum. (Source: AIME)
Hints: 245

13.73 How many digits does it take to write the base 10 integer 418 in base 3, given that log3 2 ≈ .631.
13.74 Let x, y, and z all exceed 1 and let w be a positive number such that logχ. w = 24, logy w = 40, and
logri/, w = 12. Find log, w. (Source: AIME)
13.75 Let S be the set of ordered triples (x, y, z) of real numbers for which

1°gιo(χ + y)≈2 and log10(x2 + y2) = z + 1.

There are real numbers a and b such that for all ordered triples (x, y, z) in S, we have x3+y3 = Λ∙103z+b∙102z.
What is the value of a + b? (Source: AMC 12) Hints: 174
13.76 Find (log2 x)2 if log2(log8 x) = log8(log2 x). (Source: AIME) Hints: 33
13.77
* Let x and y be positive numbers such that x1°8*x - 2 and ylo‰ l/ = 16. Find x. (Source: Bay Area
Math Meet)
*
13.78 Find both ordered triplets (x, y,z) that satisfy the system of equations

log10 (2000xy) - (log10 x)(log10 y) = 4,

logι0 (2yz) - (log10 y)(Iog10 z) = 1,
!og10 (zx) - (log10 z)(log10 x) = 0.

(Source: AIME) Hints: 353

455
CHAPTER 14. RADICALS

1 am a radical in thought and a conservative in method. - Rutherford B. Hayes

CHAPTER 14 ι
L_ Radicals

At the start of Chapter 13, we noted that the key to working with unusual expressions or inconvenient
notation in a problem is to convert the problem into one involving expressions that we understand
better.
In this chapter, we'll explore some common methods for dealing with expressions that have radicals.
The goal is almost always to remove the radicals from the problem entirely, so that we're left with less
complicated notation.

14.1 Raising Radicals to Powers

Problems

Problem 14.1: Consider the equation x = λ∕2x + 24.

(a) Square the given equation to produce a new equation. Solve your new equation.
(b) Do all your solutions from (a) satisfy x = √2x + 24?

Problem 14.2: Find all x such that √6 - x - V10 +~3x ~ 2.

Problem 14.3: In this problem, we solve the equation Vx2 - 5x + 4 - Vx2 - 10x + 9 = x - 1.
(a) Factor the quadratics.
(b) Solve the original equation.
(c) Check your solutions. You should have 2 solutions. If you don't, try the problem again.

456
 14.1. RAISING RADICALS TO POWERS

Problem 14.4: Find all real numbers x such that 3 ∖∕x ≥ x + 2.

Problem 14.5:
(a) Find all r such that Vr + 2 ≥ -r.
(b) Test your solution by substitu ting a variety of values of r into the inequality. If you find values
of r that satisfy the inequality but are not among the solutions you found in part (a), do part (a)
over again.

Problem 14.6: Prove that -■..... > - ----- - for all nonnegative integers π.
√3τT+4 2n + 2 6 6

Problem 14.7: Find all x such that ∖,2x2 + 3x - 26 = x - 2.

Problem 14.8: In this problem, we solve the equation y∕13x + 37 - ¼3x - 37 = √2.
(a) Let a = V13x + 37 and b = ¾∕13x — 37. Find two equations relating a and b.
(b) Find ab.
(c) Find x.

Problem 14.1: Find all solutions to the equation x = √2x + 24.

Solution for Problem 14.1: The radical on the right side of our equation is the major complication in this
problem. We get rid of it by squaring the equation.

Concept: One way to get rid of a radical in an equation is to raise both sides to a
O≡≡^ power that gets rid of the radical.

Squaring both sides of the equation gives us x2 = 2x+24. Rearranging and factoring gives (x-6)(x+4) = 0,
which has solutions x = 6 and x = -4.
Because we squared the equation as a step in our solution, we must check that our solutions are
not extraneous. When x = 6, our equation is 6 = √2(6) + 24, which is true. When x = -4, we have
√2x + 24 = 4, so we do not have x = √2x + 24. Our original equation is not satisfied by x = -4, but our
squared equation is; therefore, the solution x = -4 is extraneous.

WARNING!! Whenever we square an equation as a step in solving it, we have

V to check that our solutions are valid. If a solution to the squared
equation does not satisfy our original equation, then the solution is
called extraneous. An extraneous solution is not a valid solution to
the original equation.

In conclusion, the only solution to x = √2x + 24 is x = 6. □

457
CHAPTER 14. RADICALS

Let's try our radical-removal strategy on some more complicated problems.

Solution for Problem 14.2: Solution 1: Square first. Squaring both sides of the equation to get rid of the
radicals gives us 6 - x - 2λ∕(6 - x)(10 + 3x) + 10 + 3x = 4, so

2x + 12 = 2√(6 - x)(10 + 3x).

Dividing both sides by 2 and then squaring again, we have

x2 + 12x + 36 = -3x2 + 8x + 60,

so 4x2 + 4x — 24 = 0. Dividing by 4 gives us x2 + x - 6 = 0, and factoring gives (x + 3)(x — 2) = 0. Our

solutions appear to be x = -3 and x = 2. Because we squared the equation (twice) on the way to our
solution, we have to check if our solutions are extraneous.
For x = -3, we have √6 - (-3) - √10 - 9 = 3-1 = 2, so this solution is valid. For x = 2, we have
V6 — 2 — √10 + 6 = 2 - 4 = —2 on the left side, which does not equal the right side, 2. Therefore, x = 2 is
an extraneous solution, and x = -3 is the only solution to the original equation.
Solution 2: We first move one of the radicals to the right side of the equation to give the equation
√6 - x = √10 + 3x + 2, and then we square both sides to find:

6 - x = 10 + 3x + 4V10 + 3x + 4.

Simplifying this equation gives -8 - 4x = 4√10 + 3x. Dividing by 4 gives us -2 - x = √10 + 3x. Squaring
both sides of this equation gives x2 + 4x + 4 = 10 + 3x, so x2 + x - 6 = 0. From here, the rest of the solution
is the same as before. □
Either solution to Problem 14.2 is fine, but the second solution has the advantage of avoiding
quadratic expressions until the final equation.

Concept: Rearranging an equation with radicals before raising it to a power may

o≡≡≡ simplify the algebra you have to perform to solve the equation.

Problem 14.3: Solve the equation Vx2 - 5x + 4 - Vx2 - 10x + 9 = x - 1.

Solution for Problem 14.3: We first factor the quadratics inside the radicals to see if there are any useful
relationships to exploit:
λ∕(x - l)(x - 4) - V(x - l)(x - 9) = x - 1.

We immediately see that if x = 1, then both sides equal 0, so x = 1 is one solution. What's wrong with
the following:

Extra! Inspiration is the impact of a fact on a well-prepared mind.

I∣ll⅜-∣llfl⅜∣lllφ∣∣ll^ - Louis Pasteur

458
 14.1. RAISING RADICALS TO POWERS

Bogus Solution: If x ≠ 1, we can divide both sides by Vx - L to give

Vx-4 - Vx - 9 = Vx - 1.

We could start squaring and manipulating, and so on, but instead

we can do something clever. We note that x - 4 is smaller than x -1
for all x, so Vx-4 is less than Vx -1 Moreover, Vx-4 - √x-9
is less than Vx-4, so Vx-4 — Vx — 9 has to be less than Vx- 1.
Therefore, the equation Vx-4 - Vx - 9 = Vx - 1 has no solutions.
So, the only solution to the original equation is x = 1.

Pretty convincing. But maybe you noticed that the constants in the radicals, 4 and 9, are perfect
squares. If we let x = 0 in the original equation, we have V4 — V9 = -1, which is true! So, x = 0 is also a
solution to the original equation. However, x = 0 is not a solution to the equation Vx - 4- Vx - 9 = Vx - 1
that we produced by dividing the original equation by Vx - 1. This suggests that it's the "divide the
original equation by Vx — 1" step that's the problem. Looking more closely at this step, we realize that
we can only divide the equation by Vx - 1 if x > 1. If x < 1, then Vx - 1 is undefined, and if x = 1, then
dividing by Vx - 1 is dividing by 0.

Blindly manipulating the equation misled us into discarding the valid solution x = 0. But did we
miss any other solutions? We'll have to find a valid solution to the problem to figure that out. Rather
than dividing the equation
V(x - l)(x - 4) - V(x - l)(x - 9) = x - 1
by Vx - 1, let's try squaring it to get rid of the radicals. This gives us

2(x-40(x-9) + (x - l)(x - 9) = (x - 1)2.

(x - l)(x - 4) - 2 V(x-1)*

Once again, we don't blindly multiply everything out. But we still have to be careful. For example, we
can't do this:

Bogus Solution: Since x - 1 divides every term, we divide both sides by x -1 to give

x - 4 - 2 VU - 4)(x - 9) + x- 9 = x-l.

This Bogus Solution assumes that

2 V(^ - 1)2(* - 4)(x - 9) ,-----------------

—------ . ∖l ------- - = 2 √(x - 4)(x - 9),

but this isn't true! If x < 1, then V(x - l)2∕(x - 1) = -1∕ n°t 1∙

We could resort to casework, and handle the cases x < 1, x = 1, and x > 1 separately, or we could
avoid the difficulty altogether by rearranging the equation and squaring again. With a little organization,
the latter approach is easier than casework. We start with

(x - l)(x - 4) - 2 V(x - l)2(x - 4)(x - 9) + (x - l)(x - 9) = (x - 1)2.

459
CHAPTER 14. RADICALS

We isolate the radical to give

2 √(x - l)2(χ - 4)(x - 9) = (x - l)(x - 4) + (x - l)(x - 9) - (x - 1)2.

We then factor (x — 1) out of each term on the right to give us

2 √(x - l)2(x - 4)(x - 9) = (x - l)((x - 4) + (x - 9) - (x - 1)) = (x - l)(x - 12).

Now, squaring both sides of the equation 2 √(χ - l)2(x - 4)(x - 9) = (x - l)(x - 12) gives us the equation
4(x - l)2(x - 4)(x - 9) = (x - l)2(x - 12)2. At this point, we might make yet another mistake:

Bogus Solution: Dividing both sides of 4(x - l)2(x - 4)(x - 9) = (x - l)2(x - 12)2 by
(x —1)2 gives us 4(x — 4)(x — 9) = (x —12)2. Expanding both sides and
rearranging gives us 3x2 - 28x = 0. The solutions to this equation
are x = 0 and x = 28/3. Because we squared the equation while
solving it, we must check if our solutions are extraneous. When
x = 28/3, we have

√⅛-lXx-4) - √(x-l)(x-9) = 7(f)(τ) ^ V(y)G) = 5'

but x - 1 = 25/3. So, x = 28/3 is extraneous. As we've already seen,

x = 0 is a valid solution. Therefore, the only solution to the original
equation is x = 0.

This Bogus Solution overlooks the solution x = 1. The step where this Bogus Solution "loses" the
solution x = 1 occurs when we divide both sides of 4(x - l)2(x - 4)(x - 9) = (x — l)2(x - 12)2 by (x - 1)2.
If x = 1, we cannot divide by (x - 1)2, since this would be division by 0. Indeed, if x = 1, then both sides
of the equation equal 0, so x = 1 is a solution to the equation. If x ≠ 1, then we can divide both sides
by (x - 1)2. From there, we can follow our Bogus Solution above to find the solution x = 0. So, the two
solutions to the equation are x = 0 and x = 1.

WARNING!! If you divide a common factor from two sides of an equation, you
V cannot forget about the solutions that might arise if the factor equals
0.

We also could have avoided forgetting the solution x = 1 by continuing to rearrange and factor.
Subtracting (x — l)2(x —12)2 from both sides of 4(x - l)2(x - 4)(x — 9) = (x - l)2(x —12)2 gives a left side of
4(x - l)2(x - 4)(x - 9) - (x - l)2(x - 12)2 = (x - l)2(4(x - 4)(x - 9) - (x - 12)2) = (x - l)2(3x2 - 28x).

This expression is 0 when x = 1, x = 0, or x = 28/3. As before, x = 28/3 is extraneous, but the other two
are valid. Again, the only solutions to the equation are x = 0 and x = 1.
This solution shows yet again the benefit of not immediately bashing through equations. Here, we
factor relentlessly rather than multiply everything out right away. As a result, the necessary algebraic
manipulations are much simpler. Factoring is your friend—imagine trying to solve the problem by
squaring the original equation without even factoring the quadratics in the radicals!

460
 14.1. RAISING RADICALS TO POWERS

Concept: Don't be hasty to jump into lengthy computations or algebraic manipula-

Qr~3J tions. First check and see if there is a more clever approach.

Take some extra time to think about all the Bogus Solutions we pursued while finding the solution to
Problem 14.3. These Bogus Solutions all stem from blindly pursuing algebraic manipulations without
thinking about what these manipulations mean. Each seems valid, but each exhibits an error that stems
from overlooking subtle, but critical, concepts.

We now turn from equations to inequalities. Can we square to remove square roots from inequalities?

Problem 14.4: Find all real numbers x such that 3 √x > x + 2.

Solution for Problem 14.4: First, for λ∕x to be defined, x must be nonnegative. So, we must have x ≥ 0,
which means that x + 2 ≥ 0. Since we also have y/x ≥ 0, both sides of the inequality are nonnegative,
which means we can square both sides to get 9x ≥ x2 + 4x + 4. Rearranging gives x2 — 5x + 4 ≤ 0, and
factoring gives (x — l)(x - 4) ≤ 0. This inequality holds for all x ∈ [1,4]. □
Just as with equations, we must be careful when we square inequalities. Above, we made sure that
both sides are nonnegative before squaring. Here's an example of what can go wrong if you don't:

Problem 14.5: Find all r such that Vr + 2 ≥ -r.

Solution for Problem 14.5: What's wrong here:

Bogus Solution: Squaring both sides gives r + 2≥r2. Rearranging gives r2-r-2 ≤0,
and factoring gives us (r - 2)(r + 1) ≤ 0. This gives us the solutions
r∈[-l,2].

If we test r = -1, r - 0, and r - 2, we see that all of these values satisfy the original inequality, so
we might think our answer is correct. However, then we might try r = 7, and see that this value of r
satisfies the inequality, too!

Our mistake is that we aren't careful about our signs. Our solution only tells us when ( Vr + 2)2 is
greater than or equal to (-r)2, not when Vr + 2 is greater than or equal to (-r). Considering signs first,
we see that for all positive values ofr, the left side of Vr + 2 ≥ -r is positive and the right side is negative.
So, for all positive values of r, the left side is larger than the right.

To finish, we also analyze what happens when r ≤ 0. Since Vr + 2 is only defined for r ≥ 2, we only
need to consider r ∈ [-2,0]. When r ∈ [-2,0], both sides of the inequality are nonnegative. Therefore,
we can follow the steps in the Bogus Solution to see that the inequality is satisfied when r ∈ [-1,0] and
not satisfied when r ∈ [-2, -1). So, our full solution to the inequality is r ∈ [-1, +∞). □
This is yet another example of the importance of thinking about your algebraic manipulations before
performing them.

461
CHAPTER 14. RADICALS

WARNING!! Don't just blindly manipulate equations and inequalities. Doing so

*<a can introduce subtle errors and significant oversights.

Sidenote: See if you can find the flaw in this proof that 1 = 2. Suppose x and y
J∖ are positive numbers, and that x = y. Multiplying by y gives xy = y2.
Subtracting x2 from both sides gives xy - x2 = y2 - x2∙ Factoring then gives
x(y - x) = (y - x)(x + y). Dividing by y - x gives x = x + y. Since x = y, we
now have x = 2x, and dividing by x gives us 1 = 2.

√3π +1 2n +1 . „ . .
Problem 14.6: Prove that — ■ ≥ - ----- r for all nonnegative integers n.
√‰T4 2n + 2 δ 6

Solution for Problem 14.6: We'll deal with the radicals before dealing with the fractions. Since both sides
of the inequality are positive, we can simply square both sides to get an equivalent inequality:

3n + 1 4h2 + 4n + 1
------- ≥---------------- .
3n + 4 4rz2 + 8n + 4
Because both denominators are positive, we can multiply both sides by both denominators without
changing the direction of the inequality. This gives us

12π3 + 28π2 + 20π + 4 ≥ 12π3 + 28/r + 19π + 4.

Simplifying this inequality gives us n ≥ 0, which is true for all nonnegative integers n, so the original
inequality holds.
We found our solution by working backwards. It's important to note that all of our steps are
reversible, so we can start from n ≥ 0 and derive the desired inequality. We can check that all steps are
indeed reversible by writing a "forwards" solution:

Proof: Since n is nonnegative, we have n ≥ 0. Adding 12n3 + 28n2 + 19n + 4 to both sides gives
0
12n3 + 28n2 + 20n + 4 ≥ 12h3 + 28h2 + 19n + 4.

Factoring gives (3n + 1)(4∏2 + 8n + 4) ≥ (3n + 4)(4π2 + 4n + 1), and dividing both sides by
(3h + 4)(4h2 + 8n + 4) gives
3n + 1 > 4π2 + 4h + 1
3n + 4 - 4h2 + 8n + 4
Because both sides are positive, we can take the square root of both sides to give the desired

V3n + 1 > 2n + 1
λ∕3h + 4 - 2h + 2

462
 14.1. RAISING RADICALS TO POWERS

Our forwards proof shows that all our steps working backwards are reversible. However, it's very
hard to tell from reading the forwards proof how we found the solution—why did we think to add
12h3 + 28zz2 + 19zz + 4 to both sides in our first step?

In general, writing forwards is best when trying to convey that your solution is correct. It is very
clear in our forwards proof that every step is valid. However, it's not always clear from such solutions
how you would think of such a solution. Sometimes, you'll read a solution that will appear to be almost
magic. It'll be clear that the solution is correct, but it won't be clear how the writer ever thought of it.
In these cases, working through the solution backwards can sometimes reveal how the writer found the
solution.

Now that you've mastered square roots, let's try cube roots.

Problem 14.7: Find allx such that ‰2 + 3x - 26 = x-2.

■■■■■■■■■■I

Solution for Problem 14.7: We squared to get rid of square roots, so here, we'll cube to get rid of cube roots.
Cubing gives us 2x2 + 3x - 26 = (x - 2)3. Expanding the right side gives 2x2 + 3x - 26 = x3 — 6x2 + 12x - 8,
and moving all terms to the same side gives us x3 - 8x2 + 9x + 18 = 0. Factoring this polynomial gives
us (x + l)(x - 3)(x - 6) = 0, so we appear to have the solutions x = -1, x = 3, and x = 6. Checking each
solution, we find that all three work, so all three are valid solutions.
It's a good idea to check each solution, just to make sure we didn't make an algebra error. However,
we don't have to worry about cubing the equation producing extraneous roots the way that squaring
an equation might lead to extraneous roots.
Extraneous roots can occur when we square an equation because it is possible for two different real
numbers to have the same square; for example, (-l)2 = 12. So, it's possible to have a ≠ b, but a2 = b2.
While two different real numbers can have the same square, no two different real numbers have the
same cube. So, if a3 = b3, then we must have a = b. In other words, the function ∕(x) = x3 is an invertible
function; its inverse is ∕~1(x) = λ∕x∙ This means that if we cube an equation, we cannot introduce any
extraneous roots, since any value that satisfies the cubed equation must satisfy the original equation. □

Important: If we raise an equation to an even power as a step in solving it, we must

ψ check for extraneous solutions. If we raise an equation to an odd power

while solving it, we don't have to check for extraneous solutions. (But it
wouldn't hurt to check your solution for algebra or arithmetic mistakes.)

Problem 14.8: Solve λ^l3x + 37 - y∕13x - 37 = y∕2.

Solution for Problem 14.8: We might try isolating one of the roots and cubing first, but cubing both sides
of
λ∕13x + 37 = √2 + ^13x - 37

is pretty daunting. While it will give us 13x + 37 on the left, the right side will have 4 terms, and two
of them will still have complicated cube root expressions that we can't easily combine. So, cubing this
equation won't help us much.

463
CHAPTER 14. RADICALS

Cubing the original equation looks pretty scary, too. So, we look for ways to simplify our task. We
start by substituting variables for the ugly cube roots. We let a = y∕13x + 37 and b = ∖,13x - 37, so our
equation becomes a - b = ^∕2. That's a pretty simple equation. If we can find another simple equation
for a and b, perhaps we can use the resulting system to solve the problem, because it will be easy to find
x if we know a or b.
We can get rid of the cube roots in a = y∕13x + 37 and b = y∕13x - 37 by cubing both equations, which
gives

a3 = 13x + 37,
b3 = 13x - 37.
Aha! Subtracting the second equation from the first eliminates x and gives us another equation for a
and b, namely a3 - b3 = 74. We now have the system of equations

a-b = ¾∕2,
a3-b3 = 74.

We have two attractive first steps for trying to solve this system. Let's explore both a little bit, and
then decide which is worth pursuing further.

Concept: When you have multiple options for how to proceed with a problem, but
θ-7 u-ij aren't sure which will work, try taking a few steps down each path before
committing yourself to one.

• Option 1: Factor a3 - b3. Since we know a-b, and a - b is a factor of a3 - b3, we factor the left side
of a3 -b3 = 74 to find (a - b)(a2 + ab + b2) = 74. We therefore have

74
a2 + ab + b2 =
a-b √2 ⅛2 √4
Unfortunately, it's not so clear how to proceed from here. Rather than continuing with this option,
we take a look at our other possible approach. If our second option fails, we can always come back
to Option 1.
• Option 2: Cube a-b. Cubing a - b will give us both an a3 term and a -b3 term, and we know that
a3 - b3 = 74. So, we cube both sides of a- b = ∖∣2, and we have

a3 - 3a2b + 3ab2 -b3 =2.

Since a3 - b3 = 74, we have -3a2b + 3ab2 = -72. Dividing by 3 and factoring the left side gives
-ab(a - b) = -24. We have a - b = λ∕2, and substituting this in our equation gives -ab(V2) = -24.
Therefore, we have ab = 12^^4.

The equation that results from Option 2 is simpler than the one from Option 1, so we proceed with
that one. Substituting our expressions for a and b into ab = 12 >∕4 gives

√13x + 37 ■ ^13x-37 = 12√4, so λ∕169x2 - 1369 = 12∖z4.

464
 14.2. EVALUATING EXPRESSIONS WITH RADICALS

Cubing both sides of this last equation gives us 169x2 - 1369 = 6912, from which we find x2 = 49.
Therefore, we have x = ±7.
Checking our solutions to see if they work, we note that if x = 7, then we have

λ∕13 ∙ 7 + 37 - ‰ ∙ 7 - 37 = √128 - ^54 = 4λz2 - 3λ∕2 = √2,

and if x = -7, then we have

√13 ∙ (-7) + 37 - -ξ∕13 ∙ (-7) - 37 = √≡54 - √-128 = √128 - √54 =

So, both solutions satisfy the given equation. □

Notice that we didn't lock in to one solution tactic; we explored both factoring the left side of
a3 -b3 - 72 and cubing both sides of aι- b = V∑ Only after taking a couple steps down each path did we
see which offered a more promising route to a solution. (Both paths will get us to a solution eventually,
though. See if you can figure out how to finish from where we stopped on the first path.)

Exercises

14.1.1
(a) Is it true that √(x — 1)2 = x — 1 for all real values of x? Why or why not?
(b) Is it true that -∖[(x- 1)3 = x — 1 for all real values of x? Why or why not?
(c) Is it true that λ∕(x — 1)4 = (x - 1)2 for all real values of x? Why or why not?

14.1.2 Find all positive numbers x that satisfy the inequality yfx < 2x. (Source: AHSME)
14.1.3 Find all solutions to the equation √x -1 -Vx + 1 + 1 = 0.

14.1.4 Find all solutions to the equation √x + 8----- , θ = 5.

√xTδ

14.1.5 Find all values of x such that 9V2 — 3x + x2 = Vx6 - 3x5 + 2x4.
14.1.6 Find all r such that λ^5r2 + 24r + 8-2 <r.
14.1.7
* Solve: ‰-5 + ‰-4 = ‰ - 9. (Source: ARML) Hints: 252,157

14.2 Evaluating Expressions With Radicals

We've seen that raising equations with radicals to powers is an effective way to simplify them. In this
section, we use this strategy to evaluate expressions with radicals.

Extra! Mathematics takes us still further from what is human, into the region of absolute necessity, to
"*∙ ,l"* ,l*ιll* which not only the actual world, but every possible world, must conform.
- Bertrand Russell

465
CHAPTER 14. RADICALS

Problems

Problem 14.9:
(a) Simplify -^3 + 2 √2.

(b) Suppose √74- 12√30 = √Ξ - √b, where a and b are integers. Find a and b, and simplify

√74 -12 √30.

(c) Simplify ∙^45 - 29 √2.

Problem 14.10: Between which two consecutive integers is V23 + V39? (No calculators!) (Source:
Mandelbrot)

- 6 and let the expression we seek to simplify be x. Write a

relationship air
(b) Find two equations without radical signs relating a and b.
(c) Combine your equations from (b) with your equation from (a) in a way that produces an equation
you can solve for x. Find x.

Problem 14.9:
(a) Simplify -^3 + 2 √2.

(b) Suppose y74 - 12 √30 = √fl - 4b, where a and b are integers. Find a and b, then simplify

√74 -12 √30.

(c) Simplify ∙^45 - 29 V2.

Solution for Problem 14.9:

(a) We know that squaring gets rid of radicals, but we don't have an equation to square. If we start
with x = y∣3 + 2 V2 and square, we get x2 = 3 + 2 V2. This doesn't directly tell us x, but it does
suggest what form x has. The square of a number of the form a + b V2 is a2 + 2b2 + 2ab √2, so we
guess that x = a + b √2 for some integers a and b. This gives us

a2 + 2b2 + 2ab ∖∣2 = 3 + 2 √2.

If there are integers a and b that satisfy this equation, then a2 + 2b2 = 3 and 2ab = 2. We could do

466
 14.2. EVALUATING EXPRESSIONS WITH RADICALS

more algebra to find a and b, or notice that a = b = } and a = b = -1 satisfy the equations. Our
answer must be positive, so we have √3 + 2λ∕2 = l + ∖z2.

(b) Our basic strategy in the previous part was to guess that the answer had the form a + b √2.
Unfortunately, there are several different forms of numbers that we can square to get a number of
the form p - q V30∙ For example, squaring a - b V30 or a √2 - b V15 will give a number of the form
p-q V30. And there are several other cases we'd have to investigate. So, rather than guessing one
of these forms specifically, we try the more generic expression √α - Vb, where a and b are integers.
If y∣74 - 12 √30 = y∕α - ^∖∕b, then squaring both sides gives

74 - 12 √30 = α + b - 2 √αb.

So, we must have a + b = 74 and 2 Vαb = 12 V30. Dividing the second equation by 2, then squaring,
gives ab = 1080. Solving a + b = 74 for a gives α = 74 - b. Substituting this into ab = 1080 and
rearranging gives b2 — 74b + 1080 = 0. Notice that factoring this equation is exactly the same
guessing game as solving the system of equations a + b = 74 and ab = 1080. We find that a and
b are 54 and 20. Since ∙^74 — 12 V30 must be positive, we must choose a and b so that ∖[a — Vb is
positive. Therefore, we have

-^74-12√30 = √54-√20 = 3√6-2√5.

(c) As with the first two parts, we take a guess at the form of the answer, and let

- 29 √2 = a - b √2,

where a and b are integers. We cube both sides to find 45 - 29 V2 = α3 - 3a2b V2 + βab2 - 2b3 V∑
Equating the integer parts of both sides, and equating the parts with √2, gives us the system of
equations

a3 + 6ab2 = 45,
-3a2b - 2b3 = -29.

Factoring these two equations gives us

a(a2 + 6b2) = 45,

b(3a2 + 2b2) = 29.

If a and b are integers, then the second equation tells us that b times an integer is 29. Therefore,
b must be 1 or 29. If b = 29, then our second equation gives us 3α2 + 2(292) = 1, which clearly
doesn't have an integer solution. If b = 1, we have 3a2 + 2 = 29, which gives a = ±3. Trying
both values in the first equation reveals that (a, b) = (3,1) is a solution to the system, so we have
^45 - 29 √2 = 3 - √2,

467
CHAPTER 14. RADICALS

Problem 14.10: Between which two consecutive integers is λ∕23 + V39? (Source: Mandelbrot)

Solution for Problem 14.10: Again, we have square roots. Again, we'd like to square to get rid of the
square roots. And again, we don't have an equation to square. This is just like our last problem, so we
try a similar strategy.

Concept: When faced with the same complications as in problems you know how
Q—otJ to solve, try the same strategies that succeeded on the problems you've
already solved.

We introduce an equation to work with by letting x = √23 + V39. Squaring x will result in a quantity
with only one radical, so we give that a try to see what we can learn:

x2 = 62 + 2√897.

Since ∖Γsf)7 is just a little less than √900 = 30, we see that x2 is a little less than 62 + 2 ∙ 30 = 122. Because
x2 < 122, we know that x < 12. But is x greater than or less than 11? We might have x2 = 120.9, in which
case x < 11. Or we might have x2 - 121.1, which makes x > 11. So, we must determine if 62 + 2 V897 is
greater than or less than 121. This is the same as comparing 2 √897 to 121 - 62 = 59, which is the same
as comparing Vδ97 to 59/2 = 29.5. We have (29.5)2 = 870.25 < 897, so V897 > 29.5. Therefore, we have

x2 = 62 + 2√897 > 62 + 2(29.5) = 121,

so x > 11, and √Z3 + "√39 is between 11 and 12. □

Solution for Problem 14.11: We start off with the same tactic that we used on the previous problem. We
let ________ _________

We might cube both sides now, but those are pretty scary looking radicals. Back in Problem 14.8, we
made a similarly scary looking problem simpler using substitution. We try the same here. We let

and we seek the value of α

We try getting rid of the outer radicals by cubing our expressions for a and b, which gives

Subtracting the second equation from the first gets rid of the square roots and leaves a3 - b3 = 12.

468
 14.2. EVALUATING EXPRESSIONS WITH RADICALS

This is a much simpler equation, but we still need to find a — b. Unfortunately, factoring a - b out of
the left side of a3 - b3 = 12 won't help much, since this just gives (a - b)(a2 +ab + b2) = 12. Instead, we go
back to our original strategy of setting the expression we wish to evaluate equal to x. However, now we
can express x in terms of a and b; we have x = a-b. We hope to perform some algebra that will allow us
to find x. Cubing a — b will give us an expression including a3 — b3r which we know is 12. So we try that:

x3 = (α - bj3 = a3- 3a2b + 3ab2 -b3 = a3- b3 - 3ab(a -b) = 12- 3ab(a - b).

This is promising, because we can replace a-b with x. Turning to ab, we have

Now our equation is x3 = 12 — 2x. Rearranging this equation gives x3 + 2x — 12 = 0. Searching for
rational roots of x3+2x-12, we find thatx = 2isaroot, and factoring gives x2+2x-12 = (x—2)(x2+2x+6),
so we have
(x - 2)(x2 + 2x + 6) = 0.
But x2 + 2x + 6 has no real roots, so x = 2 is the only real solution to x3 + 2x - 12 = 0. Therefore, we have

In our solution to Problem 14.11, our general strategy is the same as in all the other problems we've
faced in this chapter: get rid of the radicals. In addition to cubing equations, we added a new tactic
that allows us to remove radicals from a problem: substitution. WeTl see more examples of using
substitution to simplify expressions in Section 20.2.

Exercises

14.2.1 Simplify y[β + V∏ + y6 - V∏∙

14.2.2 Simplify each of the following:

(a) √49 + 28λ^ (b) ^49 + 20√6

14.2.3 Find the ordered pair of positive integers (a, b), with a <b, for which

(Source: ARML)
14.2.4 Which is greater, V2 + V3 or V10? (No calculators!)

14.2.5 Find the value of (52 + 6√43)3z2 - (52 - 6√43)3z2. (Source: AIME)

469
CHAPTER 14. RADICALS

14.3 Radical Conjugates

hi our solution to Problem 14.11, we encountered the product

This might inspire us to notice that raising radicals to powers is not the only way to get rid of them.
In this section, we explore ways to use the difference of squares and difference of cubes factorizations to
get rid of radicals in problems.
Some problems in this section involve rationalizing the denominators of fractions. When we
rationalize the denominator of a fraction, we write the fraction as an equivalent fraction with a rational
denominator. For example, to rationalize the denominator of -^=, we multiply the numerator and
denominator by V2:
J_ _ J_ √2 _ √2
√2 √2 √2 2 '

Problems

Problem 14.12: Rationalize the denominator of

x + Vx2 - 1 x - Vx2 - 1
Problem 14.13: Find all real values of x such that = 98.
x- -1 x + √x2 - 1

Problem 14.14: What is the smallest positive integer n such that √n-√π - 1 < 0.01? (Source: AHSME)

Problem 14.15: Rationalize the denominator of —---- .

y/3-1

Problem 14.12: Rationalize the denominator of —----- -.

√7+ √2

Solution for Problem 14.12: We need to get rid of the square roots in the denominator. The difference of
squares factorization, (x + y)(x - y) = x2 - y2, gives us a way to square away both square roots in the
denominator. Since we have V7 + λ∕2 in the denominator, we multiply the numerator and denominator
by √7 - √Σ
1 _ 1 √7-√2 _ √7- √2 _ √7- √2
√7 + √2 ^ √7 + √2 √7 - √2 ~ ( √7)2 _ ( √2)2 5

□
470
 14.3. RADICAL CONJUGATES

γ a. a∕y2 _ 1 γ_ 'f'f'2∙_ I
Problem 14.13: Find all real values of x such that ------- —-----d--------------- = 98.

Solution for Problem 14.13: We can eliminate the square root from the denominator of the first fraction
by multiplying the numerator and denominator by x + Vx2 - 1:

x + √x2 - 1 _ χ + √χ2 - 1 χ + √χ2 _ j _ (χ + √χ2 - 1)2 _

x - √x2 - 1 X - √X2-1 X + √χ2-l x2 - (x2 - 1) λ+

Similarly, we can rationalize the denominator of the second fraction on the left of our original equation
to give
(x + √x2 - 1)2 + (χ - √χ2 - 1)2 = 98.
Expanding the squares gives x2 + 2x√x2 - 1 + χ2 - 1 + χ2 - 2x√x2 - 1 + x2 - 1 = 98, so 4x2 = 100. This
gives us x = ±5. □

Notice that in each of the first two problems, a key step in our solution is multiplying an expression
of the form -fa + *fb by an expression of the form fa — fb. The resulting product has no radicals:

(fa + Vt>) (fa - V&) = a-b.

The expressions fa + Vb and fa- Vb are sometimes referred to as radical conjugates.

Concept: One way to eliminate radicals in a problem involving an expression of the

o form fa + fb or fa - Vb is to multiply by the radical conjugate, fa —fb or
fa + fb, respectively.

Of course, this tactic also works when the radicals have rational coefficients besides 1. For example,
we can rationalize the denominator of 3∕(3 f5 - 2 √2) by multiplying the numerator and denominator
by 3 √5 + 2 √2r

3 _ 3 3√5 + 2√2 _ 9√5 + 6√2 _ 9√5 + 6√2

3√5-2√2 ^ 3√5-2√2 3√5 + 2√2 ^ (3√5)2 _ (2√2)2 ~ 37

Multiplying by radical conjugates isn't just for problems with square roots in a denominator:

Problem 14.14: What is the smallest positive integer n such that √n - √n - 1 < 0.01? (Source:
AHSME)

Solution for Problem 14.14: We don't yet have a very good method for approximating fn - fn - 1, even
if n is a perfect square. However, estimating fn + V∏ - 1 isn't too hard, particularly if n is a perfect
square. So, we multiply the inequality by fn + V∏ - 1, because we know this will give us 1 on the lesser
side. We are left with
1 < 0.01 (fn + Vn - 1).

471
CHAPTER 14. RADICALS

Multiplying both sides by 100 gives us

100 < √n + Vπ - 1.

If n ≤ 502, then √n ≤ 50 and √π - 1 < 50, so y/n + Vn - 1 < 100 when n ≤ 502. However, when
n = 502 + 1, we have y/n > 50 and V∏ -1 = 50, so y/n + V∏ - 1 > 100 when n = 502 + 1. Therefore, the
smallest value of n that satisfies the inequality is n = 502 + 1 = 2501. □

Let's see if our "multiply by the conjugate" tactic works on cube roots, too:

1
Problem 14.15: Rationalize the denominator of - ------- .
.
I
Solution for Problem 14.15: We might start by multiplying our numerator and denominator, — 1 + λ∕3z by
-1 — V3, hoping that our "multiply by the conjugate" strategy will still work with cube roots. We find:

1 1 -l--^3 -l-√3
+ + l-√9 '

Hmmm.... That didn't quite work. We could keep at it, multiplying next by 1 + y∣9, but that will make
quite a mess in the numerator, and still won't make our denominator rational.
Instead, we take a step back and think about why multiplying by the radical conjugate works with
square roots. As we saw earlier, it works because the product {y∕a + Vb) (y/a - y[b^ is the difference of
squares factorization of a - b.
But here, we have a cube root. The difference of squares helped us out with square roots; maybe the
difference of cubes factorization will help us with cube roots. So, we try to use the difference of cubes
factorization χ3 - y3 = (x - y)(x2 + xy + y2) just as we used the difference of squares factorization in
Problem 14.12. Letting x = y∕3 and y = 1 in this equation, we see that ^^3 - 1 is a factor of 3 - 1:

3-l = (^3-l)(^9 + ^3 + l).

We can use this to rationalize the denominator of our fraction:

1 1 √9 + ^3 + l ∖z9 + √3 + l _-^9 + y∕3 + l

√3-l ^3-l ^9 + ^3 + Γ 3-1 “ 2 ’

Extra! Here's a clever proof that it is possible for an irrational number raised to an irrational
⅜>∣ l<>∣liφllllφ ∕ r—∖ V2
power to result in a rational number. Consider the number x = ( V2J . If x is rational,
then we have our sought-after example. If x is irrational, then consider the number x λ^2.
This number equals ( V2) = ( √2^ = 2. So, either x is rational, or xχf2 is rational. In
either case, we have found an irrational number raised to an irrational power that results
in a rational number.

472
 14.4. SUMMARY

*e¼
I xeιτi

14.3.1 Rationalize the denominator in each of the following:

4
(a)
√lT + 3
(c) ------ ——- (Source: AHSME)
√2 + √3 + √5
√4 + 2√3 + √4 - 2√3
(b) (d)
√4 + 2√3 - √4 - 2√3

14.3.2 Evaluate the product (√5 + √6 + √7) (√5 + √6 - √7) (√5 - √6 + √7) (-√5 + √6 + ^√7). (Source:
AIME)
1
14.3.3 Write the number ——---- — as the sum of terms of the form 2cι, where q is rational. (For example,
√2- ^2
21 + 2_1/3 + 28/5 is a sum of this form.) (Source: USAMTS)
ι ___
14.3.4 Show that — < √101 - V99 without using a calculator.

14.4 Summary

The key idea in this chapter is to deal with problems involving radicals by getting rid of the radicals.

Concept: One way to get rid of a radical in an equation is to raise both sides to a
O≡^ power that gets rid of the radical.

Concept: One way to eliminate the radicals in a problem involving an expression of

Q ud the form y[a + y/b or y[a - y/b is to multiply by the radical conjugate, y[a-y[b
or -√α + y/b, respectively.

We often use this idea when rationalizing the denominator of a fraction, which means writing the
fraction as an equivalent fraction with a rational (usually integer) denominator.

Things To Watch Out For!___________________________________________

• Whenever we square an equation as a step in solving it, we have to check that our solutions
are valid. If the solution to the squared equation does not satisfy our original equation, then
the solution is called extraneous. An extraneous solution is not a valid solution to the original
equation.

• If you divide a common factor from two sides of an equation, you cannot forget about the solutions
that might arise if the factor equals 0.

473
CHAPTER 14. RADICALS

Problem Solving Strategies__________________________________________ -

Concepts: ∙ Rearranging an equation with radicals before raising it to a power

Q∙-'c⅛) may simplify the algebra you have to perform to solve the equation.

• Don't just manipulate equations blindly. If you think about what an

equation means, you might find a fast route to solve the equation or
see that it has no solutions.

• Don't be hasty to jump into lengthy computations or algebraic ma­

nipulations. First check and see if there is a more clever approach.

• When you have multiple options for how to proceed with a problem,
but aren't sure which will work, try taking a few steps down each
path before committing yourself to one.

• When faced with the same complications as in problems you know

how to solve, try the same strategies that succeeded on the problems
you've already solved.

Review Problems

14.16 Rationalize the denominator in each of the following:

1 1
(a) (c)
√3-√2 ^7-l
4 2
(b) (d)
3 + √7 yz25 + ^5 + l

14.17 Simplify each of the following:

(a) λ∕6 + 2√5 ⅛89 + 28√Tθ

(b)

Simplify -^4 + V7 - y∣4 - λ∕7.

14.18

14.19 Solve the equation V5x- 1 + yx-1 = 2. (Source: AHSME)

14.20 Find all solutions to the equation 4λ∕x3 - 2x2 + x = Vx3 - x2.

Find all pairs of rational numbers (x,y) such that -∖^2λ∕3 - 3 = ^.v∖∕3 - ^ι∕V3.
14.21

474
 CHALLENGE PROBLEMS

14.22 Simplify (Source: AHSME)

14.23
(a) Find all values of x such that Vx2 + 7x + 10 = x + 2 + y[x + 2.
(b) * Find all values of x such that Vx2 + 7x + 10 > x + 2 + Vx + 2.

14.24 Simplify the sum √18 + 5√13 + λ∕18-5√13.

14.25 For x > 1, the equation y∣x + √2x - 1 - yj~χ — y∕2x - 1 = Vfc is an identity. Find k. (Source: IMO)

14.26 The number

7104√6 + 468√10 + 144√15
+ 2006
can be written as ax/2 + b V3 + c>∕5, where a, b, and c are positive integers. Find abc. (Source: AIME)

14.27 If x + √χ2 - 1 4-------- = 20, then find x2 + Vx4 - 1 +----------------- . (Source: AHSME)
x - Vx2 - 1 χ2 + √χ4 - 1
14.28 If yjri^+ Vn2 + 8 + y[tι- Vπ2 + 8 = 8, where n is an integer, then find n.

Challenge Problems

14.29 Solve the equation χ∕60 - x + x/x - 11 = x∕i. (Source: Sweden) Hints: 79,106

14.30 Find the largest integer N such that N < + V128 + V2 - δj . (Source: ARML)

14.31 Find the ordered triple of positive integers (a, b, c) for which
(√5 + √2 - √3) ^3√fl + √fe - 2√c) = 12.

(Source: NYSML)

14.32 Let x = ——---- ——---- 77—------ x z ----- 7 Find (x + 1)48. (Source: AIME)
(√5 + l)(√5 + l)(^5 + l)(1V5 + l)

14.33 Show that y[x - 4λ∕x - 4 + 2 = -Jx + 4Vx - 4 - 2 for all x ≥ 8.

14.34
* Solve the equation (V2 + l)x + (V2 - l)x = 6. (Source: Sweden) Hints: 236

14.35
* Find a polynomial ∕(x, y, z) in three variables, with integer coefficients, such that for all integers
a, b, c, the sign of f(a, b, c) (that is, positive, negative, or zero) is the same as the sign of a + bx/2 + c λ^4.
(Source: USAMTS) Hints: 190, 95

475
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

CHAPTER
l_
15 ----------- ι
Special Classes of Functions

15.1 Rational Functions and Their Graphs

Any function that can be expressed as one polynomial divided by another is called a rational function.
So, ∕(x) is a rational function if
rtx) =
t(x)'
where s(x) and t(x) are polynomials, and t(x) is not the polynomial 0. For example, each of the following
is a rational function:

^)∙=⅛

,, 1 5y3-7y + 13
<y) = 4/ '

z10°
,(2)_ (z-l)(z-2)'

Problems

1
Problem 15.1: Let∕(x) = -.
X
(a) Find the domain of f. (b) Find the range of f. (c) Graph f.

476
 15.1. RATIONAL FUNCTIONS AND THEIR GRAPHS

Problem 15.1: Let ∕(x) = -.

(a) Find the domain of f. (b) Find the range of f. (c) Graph f.

Solution for Problem 15.1:

(a) The only value of x for which f(x) is not defined is x = 0, so the domain is all nonzero real numbers.
(b) We let y = j. Solving for x in terms of y gives x = so we cannot have y = 0. For any other value
of y, letting x = ∣ in ∕(x) = 1 gives us ∕(x) = y, so the range of f is all nonzero real numbers.

(c) The graph of f is the graph of the equation y = zg Multiplying this by x

gives us xy = 1. We found in Section 5.6 that the graph of this equation
is a hyperbola, as shown at right.
As we noted back in Section 5.6, an asymptote of a graph is a line
that the graph approaches but does not intersect as x or y becomes very
far from 0. For example, as x gets very large, the value of y approaches,
but never equals, 0. So, the graph approaches, but doesn't intersect, the
x-axis (where y = 0) as x gets large. Similarly, as x gets much less than 0,
the graph again approaches but does not intersect the x-axis. So, we call the x-axis a horizontal
asymptote of the graph. Similarly, the y-axis is a vertical asymptote of the graph.

I HL
Problems
1 i

τ, _ _ - τ rz . x+4, , . x2 + 3x-4
Problem 15.2: Let ∕(x)= -—- and g(x) = -z------------- .
(a) Factor the numerator and denominator of g. For what values of x are ∕(x) and g(x) equal?
(b) Find the domain and range of f, and graphy = ∕(x). Find all asymptotes of the graph.
(c) Find the domain and range of g, and graphy = g(x). Find all asymptotes of the graph.
(d) Are f and g the same function? Why or why not?

Problem 15.3: Consider your graph of ∕(x) = from the previous problem. Notice that the graph
looks like the hyperbolas of the form (x - h)(y -k) = c from Section 5.6. In this problem, we learn why.
(a) Let y = Get rid of the fraction, and move all the variables to one side. Can you factor the
side with the variables?
(b) Show that the graph of f is a hyperbola of the form (x - Zz)(y - fc) = c.

p(x)
Problem 15.4: Let ∕(x) = where p(x) and rf(x) are polynomials.
d(x)'
(a) Suppose the line x = c, where c is a constant, is an asymptote of the graph of f. Must d(c) = 0?
(b) If c is a constant such that d(c) = 0, then must the line x = c be a vertical asymptote of the graph
of /? Why or why not?

477
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

Problem 15.5: Which of the graphs below could be the graph of g(x) =

For each graph that is not the graph of g, explain why the graph cannot possibly be the graph of g.

Problem 15.6: Let g(x) =------- —---- .

(a) Graph y = g(x).
(b) Is there a non-vertical line that appears to be an asymptote of the graph?
(c) Find the quotient and remainder when 2x2 - 5x - 1 is divided by x - 3. Use the result to explain
your observations from part (b).
(d) What is the range of g?

Problem 15.7: Let ∕(x) = ———j—----- -. In this problem, we find a horizontal asymptote of the
A + ⅛3λ H- z_A
* O
graph of f.
(a) Divide the numerator and denominator by ∕(x) by x4, thereby finding an expression for ∕(x) for
all x ≠ 0 such that the highest degree term in the denominator is a constant.
(b) Use the form you found in part (a) to investigate what happens to ∕(x) when x is very far from
0. What line is a horizontal asymptote of the graph of /?
(c) Generalize this problem. In other words, find a general statement about a certain class of rational
functions that this problem represents.

478
 15.1. RATIONAL FUNCTIONS AND THEIR GRAPHS

Problem 15.2: Let f(x) = - - and g(x) = * ∙+ —4

x-3 ‘ x2 - 4x + 3
(a) Factor the numerator and denominator of g. For what values of x are ∕(x) and g{x) equal?
(b) Find the domain and range of f, and graph y = ∕(χ).
(c) Find the domain and range of g, and graph y = g(χ).
(d) Are f and g the same function? Why or why not?

Solution for Problem 15.2:

(a) Factoring the numerator of g(x) gives us g(x) = y + 4)<* 1∖ SO/ it appears that we have
l(x-3)(x-l)
= (X + 4)(x - 1) _ x + 4 _
g (x - 3)(x - 1) “ χ - 3 ^ -'^t

for all x. However, while ∕(1) = (1 + 4)∕(1 - 3) = -5/2, the function g(x) is not defined for x = 1,
because the denominator of g(x) is 0 if x = 1. For all other values of x, the x - 1 factors in the
numerator and denominator of g(x) do cancel. This leaves us with ∕(x) = g(x) for all x besides
x = 1 and x = 3 (both functions are undefined for x = 3).
(b) The domain of f is all real numbers except 3. To find the range,
we let y = ∕(x) = (x + 4)∕(x - 3). Solving for x in terms of y
gives us x = (3y + 4)∕(y - 1). For any value of y except y = 1,
we can use this equation to find a value of x such that ∕(x) = y.
Therefore, the range of f is all real numbers except 1.
Choosing several values of x reveals the graph of y = f (x)
shown at right. The line x = 3 is a vertical asymptote, because
choosing values of x closer and closer to 3 makes y much greater
than 0 (when x > 3) or much less than 0 (when x < 3). Similarly,
y = 1 is a horizontal asymptote. To see why, consider the
equation x = (3y+4)∕(y-l), which we found by solving y = ∕(x)
for x in terms of y. As y gets very close to 1, the value of x is
much greater than 0 (when y > 1) or much less than 0 (when y < 1). The asymptotes are dashed
in the graph.

(c) Because the denominator of g(x) is (x - 3)(x - 1), the domain of g(x) is all real numbers except 1
and 3. We've seen that g(x) = ∕(x) for all values of x except x = 1 and x = 3. So, the range of g
appears to be the same as the range of f, except that it may not include ∕(1), because x = 1 is not
in the domain of g. There are a couple ways to see that ∕(1), which equals -5/2, is not in the range
of g.
Approach 1: Solve g(x) = -5/2. Cross-multiplying

x2 + 3x - 4 5
x2 - 4x + 3 = ~2

gives 2x2 + 6x — 8 = —5x2 + 20x — 15, so we have 7x2 — 14x + 7 = 0. This gives us 7(x — 1)2 = 0, so
we only have g(x) = -5/2 if x = 1. This is as expected, since ∕(1) = -5/2. However, because 1 is

479
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

not in the domain of g, there are no values of x for which g(x) = -5/2. Therefore, the range of g is
all real numbers except -5/2 and 1 (remember, except for -5/2, the range of g matches the range
of/).
Approach 2: Graph g. Because g(x) = f(x) for all x except for
x = 1, the graph of y = g(x) is the same as the graph of y = f(x)
everywhere except where x = 1. So, to graph g(x), we draw
the graph of f(x), but place an open circle at the point (1, -5/2),
since this point is on f(x), but not on g(x). This graph is shown
at right. Again, the asymptotes are shown as dashed lines.
Note that we do not consider the line x = 1 an asymptote. The
graph of g does approach this line, and does not intersect it,
but the graph of g does not approach the line x = 1 as y gets
far from 0.
We say that the graph of g has a hole at (1, -5/2).
(d)Because ∕(x) and g(x) are not the same for x = 1, they are not
the same function.

□
Problem 15.3: Consider your graph of
x+4
x-3
from the previous problem. Notice that it looks like the hyperbolas of the form (x - ∕z)(y - k) = c from
Section 5.6. Is this a coincidence? ⅛⅛. . ∣ -∣ ∣∣∣∣⅜w∣∣∣∣w∣∣∣m-∣m∣∣ww∣w∣w∣m∣∣-wm——■■

Solution for Problem 15.3: The graph of ∕ is the graph of the equation y = (x + 4)∕ (x - 3). Multiplying both
sides by x - 3 gives xy - 3y = x + 4. Hey, this looks like an equation for Simon's Favorite Factoring Trick!
Moving the variables to the left gives xy - 3y - x = 4. Adding 3 to both sides gives xy - 3y - x + 3 = 7,
and factoring the left side of this gives us

(x - 3)(y - 1) = 7.

So, rearranging the equation y = ∕(x) gives us an equation of the form (x - ∕z)(y -k) = c. Therefore, the
graph of ∕ is a hyperbola with horizontal and vertical asymptotes. □
Now that we've graphed a few rational functions, perhaps you have some intuition for how to find
the vertical asymptotes of the graph of a rational function.

p(χ)
Problem 15.4: Let ∕(x) = where ρ(x) and rf(x) are polynomials.
d(x)' r r 7
(a) Suppose the line x = c, where c is a constant, is a vertical asymptote of the graph of f. Must
rf(c) = 0?
(b) If c is a constant such that d(c) = 0, then must the line x = c be a vertical asymptote of the graph
of /? Why or why not?

480
 15.1. RATIONAL FUNCTIONS AND THEIR GRAPHS

Solution for Problem 15.4:

(a) If d(c) ≠ 0, then ∕(c) = p(c)∕d(c) is defined. In other words, the graph of f passes through the point
(c, a(<5>' If χ = c were an asymptote of the graph of f, then the graph would approach the line, but
not intersect it. Instead, the graph approaches the line, then passes through a point on the line.
So, x = c is not an asymptote of the graph of f if d(c) ≠ 0. This means that we must have rf(c) = 0
if x = c is a vertical asymptote of the graph of f.
(b) Consider our graph of y
x2 + 3x - 4
g(χ) =
x2 - 4x + 3
in Problem 15.2. This graph is shown at right. The denominator
equals 0 when x = 1, but x = 1 is not a vertical asymptote of
the graph. Why isn't it?
The numerator of g(x) is also 0 when x = 1. As we saw in
our discussion of Problem 15.2, if x ≠ 1, then we can cancel the
factors of x — 1 in the numerator and denominator of g(x) to
find g(x) = ∣⅛∣. So, the graph of g(x) approaches ∣3∣ - —∣ as
x gets close to 1. In other words, g(x) is not far from 0 when x
gets close to 1. This tells us that x = 1 is not a vertical asymptote of the graph of g.
What if the numerator is not also 0 when the denominator equals 0? We'll answer this by
considering the rational function ∕(x) = p(x)∕d(x), where d(c) = 0 and p(c) ≠ 0 for some constant c.
Because d(c) — 0, one of the factors of d(c) is x - c. Therefore, as x gets very close to 0, the value of
d(x) gets very close to 0. Since p(c) ≠ 0, when we compute ∕(x) for values of x very close to x = c,
we are dividing a nonzero quantity by a number that is very close to 0. The result of dividing a
nonzero constant by a number that is very close to 0 is a quotient that is very far from 0. This tells
us that if d(c) = 0 and p(c) ≠ 0, the line x = c is a vertical asymptote of the graph of f.
Note that if p(x) and d(x) are polynomials such that p(c) = d(c) = 0 for some constant c, then we cannot
conclude that x = c is definitely not a vertical asymptote of the graph of y = p(x)∕d(x). For example,
consider the function ∕(x) = x∕x2. Here, both the numerator and denominator are 0 for x = 0. However,
we have ∕(x) = 1 ∕x for all nonzero x, so the graph of ∕(x) = x∕x2 is the same as the graph of ι∕ = 1 ∕x. This
latter graph clearly has a vertical asymptote at x = 0.

Together, our two parts of Problem 15.4 give us some guidance for finding vertical asymptotes of the
graph of a rational function:

Important: Let p(x) and d(x) be polynomials such that √(x) is not 0 for all x. If c is a

ψ constant such that d(c) = 0 and p(c) ≠ 0, then the graph of ∕(x) = p(x)∕d(x)
has a vertical asymptote atx = c.

We can use our understanding of asymptotes and polynomial roots to quickly get qualitative infor­
mation about the graph of a rational function.

481
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

Solution for Problem 15.5: We proceed by process of elimination. First, we note that g(x) = 0 if and only
if x = —4 or x = 1. So, the graph of g intersects the x-axis once to the right of the y-axis and once to the
left of the y-axis. That eliminates graph (a) and graph (f). It looks like it eliminates graph (e) as well,
but it's not completely clear where that graph intersects the x-axis, so we should look for another way
to eliminate (e).
Next, we note that the two lines x = -3 and x = 2 are vertical asymptotes of g. These two asymptotes
are on opposite sides of the y-axis. This does eliminate option (e).
Comparing graphs (c) and (d), we see that the only difference is the left branch of the graph. Consider
what happens when x < -4. This makes (x - 1), (x + 3), and (x - 2) all negative. However, (x + 4)2 is
always positive. So, if x < -4, then our expression for g(x) has a negative numerator and a positive
denominator. Therefore g(x) < 0 for x < -4. This eliminates graph (d).
All we have left are graphs (b) and (c). The only difference between these two is the middle branch.
There are several ways to eliminate (c). First, we could note that g(0) = 4/3, so g(0) > 0, which is not the
case in graph (c). We could also note that if -3 < x < 0, then the numerator of g(x) is negative, as is the
denominator, so g(x) > 0 if -3 < x < 0. This is the case in graph (b), but not in graph (c). (Similarly, we
could have examined the behaviors of the graphs near x = 2.)

∣ ∣⅛a∣⅜ l∣ Φ ∣ *
Extra! One factor that has remained constant through all the twists and turns of the history of physical
science is the decisive importance of the mathematical imagination.
- Freeman Dyson

482
 15.1. RATIONAL FUNCTIONS AND THEIR GRAPHS

We have eliminated all but graph (b). Graph (x + 4)2 x + 3 x - 1 x-2 gW

(b) nas roots and vertical asymptotes in the cor­ x>2 + + + ÷ +
rect locations. To check if the graph is below 1 <x<2 ÷ + ÷ _ _
the y-axis where g is negative and above the -3 < x < 1 + + — +
y-axis where g is positive, we make the table -4 < x < -3 + — — — _
at right. We see that g(x) is positive for x > 2 x < -4 + - — - -
and -3 < x < 1, and negative for 1 < χ < 2,
-4 < x < -3, and x < -4. Though we don't know for sure that these exactly match the corresponding
ranges in graph (b) for which the graph is above or below the x-axis, there are no points (x, y) on graph
(b) at which the sign of y obviously cannot match the sign of g(x). □
Our graph of g(x) in Problem 15.5 exhibits some interesting behavior as x becomes far from 0.
Specifically, the graph doesn't appear to have a horizontal asymptote, but it does appear to "straighten
out and look more and more like a straight line. Let's take a look at what might cause this behavior.

Problem 15.6: Letg(x) = —------------- -.

x 3
(a) Graph y = g(χ).
(b) Is there a non-vertical line that appears to be an asymptote of the graph?
(c) Find the quotient and remainder when 2x2 — 5x — 1 is divided by x — 3. Use the result to explain ∣
your observations from part (b).
(d) What is the range of g?

Solution for Problem 15.6:

(a) We start with the table at left below, and use it to generate the graph of y = g(x) at right below.
X g(
*) X g(
*)
-7 -13.2 3.1 27.2
-5 -9.25 3.5 12
-3 -5.33 4 11
-1 -1.5 5 12
1 2 6 13.67
2 3 7 15.5
2.5 2 8 17.4
2.9 -13.2 9 19.33

(b) The graph has two pieces, and a vertical asymptote at x = 3 (where the denominator of g(x) is 0, as
expected). The graph does not seem to have a horizontal asymptote. However, as x gets far from
0, the graph appears to become more and more like a line. Both the graph and our table suggest
that the slope of the line is close to 2.
(c) We explained the hyperbola form of the graph y = (x+4)∕(x-3) by writing the equation in another
form. We try the same here. We hope to explain the line-like behavior of the graph of g by finding
another form in which to write g(x). One natural way to do so is to perform polynomial division,

483
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

which gives us
2x2 - 5x - 1 2
gW = — 2x + 1 H-------- .
x-3 x—3
Aha! This tells the story. When x is very large, the term 2∕(x — 3) is very close to zero, so g(x) is
very close to 2x + 1. The same is true if x is much less than 0. So, as x becomes far from 0, the
graph of g gets closer and closer to the line y = 2x + 1. We call this line a slant asymptote or an
oblique asymptote of the graph.
In fact, a horizontal asymptote can be thought of as a slant asymptote with slope 0, and we can
discover horizontal asymptotes the same way we found the vertical asymptote of the graph of g.
For example, we earlier found that the graph of ∕(x) = (x + 4)∕(x — 3) has a horizontal asymptote
at y = 1. Performing the polynomial division (x + 4)∕(x - 3) gives us

x+4 1 7
------ 3 - 1 ^l-------- 3∙
x-3 x-3
As x gets very far from 0, the term 7∕(x - 3) gets very close to 0, so ∕(x) gets very close to 1.
While horizontal asymptotes are essentially slant asymptotes with slope 0, when we use the
term "slant asymptote" in this book, we refer specifically to asymptotes that are neither horizontal
nor vertical.
(d) Our graph and table suggest that the range of g is (-∞, 3] U [11, +∞). To show that this is the case,
we again rewrite
2x2 - 5x - 1
y= x-3 '

Multiplying both sides by x - 3 gives

xy - 3y = 2x2 - 5x - 1,

and bringing all terms to one side gives us a quadratic in x:

2x2 + (-5 - y)x + (3y - 1) = 0.

We wish to know for what values of y there is a real solution x to this equation. This is a job for the
discriminant. This quadratic has real solutions for x if and only if its discriminant is nonnegative,
so we must have
(-5 - y)2 - 4(2)(3y - 1) > 0.
Simplifying the left side of this inequality gives us y2 - 14y + 33 ≥ 0. Factoring the left side gives
us (y - ll)(y - 3) ≥ 0, which has solutions y ∈ (-∞,3] U [11, +∞), as expected.

In Problem 15.6, we learned a lot about g and its graph by writing the equation y = y(x) in different
forms.

Concept: Don't limit yourself to an equation or expression you're given. Different

algebraic forms of the same equation or expression can offer different
information, so experiment with manipulating equations and expressions
into different forms.

484
 15.1. RATIONAL FUNCTIONS AND THEIR GRAPHS

2x2 - 3
Problem 15.7: Let ∕(x) = — * ——-. Find the horizontal asymptote of the graph of f.
■
Solution for Problem 15.7: Intuitively, it seems like y = 0 is a horizontal asymptote of the graph because
when x is very large, the value of x4 is much greater than that of 2x2. To more clearly show that this is
the case, we look back to Problem 15.6 for inspiration. In that problem, we rewrote the given rational
function in a more convenient form by performing polynomial division. Here, polynomial division isn' t
so obviously useful, since the quotient is 0 (a clue that y = 0 is a horizontal asymptote), which means
∕(x) can't be rewritten with polynomial division.

We wish to analyze what happens when x is very large. We don't have good tools for comparing
what happens to different powers of x as x gets large, but we do know that dividing by large values
produces numbers close to 0. So, we rewrite our rational function such that in both the numerator and
denominator, all the instances of x are in denominators. Specifically, for all x ≠ 0, we can divide the
numerator and denominator of ∕(x) by the highest power of x in either, x4, to eliminate the positive
powers of x and leave

f(χ
j
∖ — __ fy 2 o
__________
x4 + 3x3 + 2x-5
X
χ4 _
X ^ l+-+4-⅛
X-J.
χ2 x4

xi X x3 x4
for x ≠ 0. Notice that we have a constant term in the denominator, but not the numerator. This is
because the denominator of f (x) has a higher degree than the numerator.

Now, the behavior of ∕(x) when x is far from 0 is clear. Except for the constant term, 1, every term in
the numerator and the denominator is very close to 0 when x is far from 0. Therefore, when x is large,
the numerator is very close to 0, while the denominator is close to 1. This means ∕(x) is very close to 0.
To see that ∕(x) is not equal to 0, we can note that the roots of 2x2 - 3 are ±√3∕2, so ∕(x) ≠ 0 when x is
very far from 0. Since ∕(x) gets very, very close to 0, but never equals 0, as x gets far from 0, we know
that y = 0 is a horizontal asymptote of the graph of f. □
We can follow essentially the same steps to generalize Problem 15.7. Suppose that

( . _ ⅜x" + ⅜-ιxn~1 + ∙ ∙ ∙ + fl0

SX bmxm + bm-1xm~1 + -∙- + b0'

where an and bm are nonzero. The key feature of our specific ∕(x) above was that the degree of the
denominator was greater than that of the numerator. So, we let m > n. Then, for x ≠ 0, we can divide
the numerator and denominator of g(x) by xm to get
a∏ | an-∖ _i_ . . . ∣ £0_
jH-∕π ‘ ' * * * ' %m

for x ≠ 0. When x is far from zero, all the terms except the bm in the denominator are very close zero.
So, g(x) approaches 0 when x is far from 0.

Important: If ∕(x) = g(x)∕∕z(x), where g(x) and ∕z(x) are nonzero polynomials such

ψ that deg g < deg h, then y = 0 is a horizontal asymptote.

485
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

txeιτlses

15.1.1 Find the domain and range of each of the following rational functions. Also, sketch the graph
of y = ∕(x) for each function, and identify any asymptotes.
r∕ ∖ x+
(a) ∕ω = x.ι (C) Λτ> = (x - 2)(x + 3)
x2 — 1 x3
(b) ∙flx> = 2x + 3 <d) ∕ω=χ2.1

15.1.2 Find a rational function that has vertical asymptotes x = 1 and x = —2. Sketch the graph of you:
function.

15.1.3 For what values of the constant c does the graph of ∕(x) = ——“—Tξ have exactly one vertical
X oX ÷ J. 3
asymptote?
15.1.4 Let f and g be rational functions. Suppose the graph of f has a vertical asymptote at x = —1 and
the graph of g has a vertical asymptote at x = 3, and that these graphs have no other vertical asymptotes.
What can we say about the vertical asymptotes of f(x) ∙ g(x)? Of f(x) + g(x)? Bθ careful—there are some
special cases that must be considered.
15.1.5
2χ -1
(a) Does the graph of f(x) = 2 have a horizontal asymptote? If so, what is it?
(x - 2)(3x -1) 1 , . , r, τ. , . . _
(b) Does the graph of f(x) = -------- —------ have a horizontal asymptote? If so, what is it?

(c) Does the graph of ∕(x) = γ-ηττ- —∑γ have a horizontal asymptote? If so, what is it?
yλ I J- ∕ ∖X" v5y

x-2
(d) Does the graph of ∕(x) = — have a horizontal asymptote? If so, what is it?

(e) Suppose ∕(x) = g(x)∕h(x), where g(x) and ∕ι(x) are non-constant polynomials. What must be true
about g(x) and ∕ι(x) if the graph of f has a horizontal asymptote?

15.2 Rational Function Equations and Inequalities

In this section, we continue our work with rational functions by solving equations and inequalities
involving rational functions.

Problems

3x-2 2-5x
Problem 15.8: Find all values of x such that----- -
x-1 x-3 4

486
 15.2. RATIONAL FUNCTION EQUATIONS AND INEQUALITIES

Problem 15.9: In this problem, we find all solutions to the equation

o d π 5x3 + 13x2 + 6x n x3 - 3x + 2
8x + 12 +------- ------------ = 2x +----- =—-—.
x2 + x x2 -1
(a) Can you simplify either side of the equation before getting rid of the fractions?
(b) Solve the equation and check that your solutions satisfy the original equation.

Q t∑
Problem 15.10: In this problem, we solve the inequality-------- ∣- 2 ≥------ .
π j r+4 r+1
(a) What s wrong with simply multiplying the inequality by (r + 4)(r + 1) and solving the resulting
inequality?
(b) Move all the terms to the left side and write that side with a common denominator.
(c) Factor the numerator and denominator of the fraction you formed in part (b). Use the result to
solve the inequality.

Problem 15.11: There is a unique real value A such that for all x with 1 < x < 3 and x ≠ 2,

A 1
~7-------- 7 + ~<7-----7? < 1999.
x2 - x - 2 x2 - 6x + 8
Compute A. (Source: Duke Math Meet)

x 3x — 2 2 — 5x
Problem 15.8: Find all values of x such that
x-1 x-3 4
■■■■■■■■■■■■■■I

Solution for Problem 15.8: We first form a common denominator on the left, which gives us

x 3x-2 _ x(x - 3) + (3x - 2)(x - 1) _ 4x2 - 8x + 2

x-l+x-3 (x- l)(x - 3) “ (x-l)(x-3)'

1 . . 4x2 - 8x + 2
so the equation is ------ —----- — —. Multiplying both sides by 4(x - l)(x - 3) gets rid of the fractions
4 (x-l)(x-3)
and gives us
4(4x2 - 8x + 2) = (2 - 5x)(x - l)(χ - 3).
Next, we expand the product on left to give 16x2 - 32x + 8, and we expand the product on the right to
give

(2 - 5x)(x - l)(x - 3) = (2 - 5x)(x2 - 4x + 3) = 2(x2 - 4x + 3) - 5x(x2 - 4x + 3) = -5x3 + 22x2 - 23x + 6.

So, our equation now is

16x2 - 32x + 8 = -5x3 + 22x2 - 23x + 6.
Rearranging this equation gives 5x3 - 6x2 - 9x + 2 = 0. We first try x = 1 and x = -1, hoping to find a
root quickly. We see that x = — 1 solves the equation, so x + 1 is a factor of the left side. Factoring this

487
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

out gives (x + l)(5x2 - llx + 2) = 0. Factoring this quadratic gives us (x + l)(5x - l)(x - 2) = 0, so our
solutions are x = -1, x = 1/5, and x = 2. Testing each solution, we find that all three satisfy the original
equation. □
There are many other ways we could have solved Problem 15.8 that essentially amount to the same
solution as above. For example, we could have multiplied by 4(x - l)(x - 3) as our very first step to get
rid of the fractions, rather than first finding the common denominator on the left side. We also could
have moved all the terms to one side initially, and then found a common denominator on that side.

All these possible solutions have one common general strategy:

Concept: When faced with a new type of equation, focus on what makes the equation
o difficult to solve. Try to remove this difficulty and turn the equation into
one you already know how to handle.

For equations involving rational functions, the difficulty is typically the variables in the denominator.
We usually take care of this the way we did in Problem 15.8, by multiplying the equation by the
denominators to get rid of the fractions. However, we must be careful to check our answers at the end.
Our next problem shows why.

5x3 + 13x2 + 6x x3 - 3x + 2
Problem 15.9: Find all solutions to the equation 8x + 12 + = 2x +
x2 + x x2 -1

Solution for Problem 15.9: We start by simplifying the fractions. We can factor x out of the numerator
and denominator of (5x3 + 13x2 + 6x)∕(x2 + x), then cancel these common factors to leave

5x2 + 13x + 6 x3 - 3x + 2
8x + 12 ÷ = 2x +
x+1 x2 -1

Concept: Don't make problems harder than they are! When solving equations in­
volving rational functions, eliminate any common factors in the numerator
and denominator of each rational function.

We can't simplify the fraction on the left any further, but we quickly see that x -1 divides the numerator
and denominator of the fraction on the right (since both the numerator and denominator equal 0 when
x = 1). We have
x3 -3x + 2 (x- l)(x2 + x-2) _ x2 + x-2
x2-l ~^ (x-l)(x + l) x+1 ,
and our equation is now
, „ 5x2 + 13x + 6 _ x2 + x - 2
8x + 12 +---------- ------ = 2x +--------- —.
x+1 x+1
Before multiplying by x +1 to get rid of the fractions, we rearrange the equation to combine the fractions.
We then have
4x2 + 12x ÷ 8 _
----------------- = -6x - 12.
x+1

488
 15.2. RATIONAL FUNCTION EQUATIONS AND INEQUALITIES

Cross-multiplying gives us 4x2 + 12x + 8 = -6x2 - 18x - 12, so we have 10x2 + 30x + 20 = 0. Solving this
equation gives us x = -1 and .v = -2 as solutions. However, x = -1 makes denominators in our original
equation equal to 0, so we must discard it as extraneous.

WARNING!! When you solve an equation involving rational functions, you must
V check that your solutions do not make any denominator equal to 0.

The only solution to our equation is x = -2. □

When we turn from equations to inequalities, we have to be more careful how we handle the
complications caused by rational functions.

Problem 15.10: Find all values of r that satisfy -—- + 2 > -—-
........ _ . _ _____ y r+4 “ r +1

Solution for Problem 15.10: What's wrong with this approach:

Bogus Solution: We multiply both sides by (r + l)(r + 4) to get rid of the fractions.
On the left, this gives us

(r - 3)(r +1) + 2(r + l)(r + 4) = [(r - 3) + 2(r + 4)](r +1) = 3r2 + 8r + 5,

and on the right we have

(5 - r)(r + 4) = -r2 + r + 20,

so our inequality is now 3r2 + 8r + 5 ≥ -r2 + r + 20. Rearranging

this inequality gives 4r2 + 7r -15 > 0, so we have (r + 3)(4r - 5) > 0,
which gives us r ∈ (-∞, -3] U [∣, +∞).

All the steps look OK, but if we look at our solution closely, we see one problem right away. Our
final answer suggests that r = -4 satisfies the inequality, but it clearly doesn't, since it makes the left
side undefined. A little more experimenting will also reveal that r = -2 satisfies the inequality, but isn't
included in our answer. Uh-oh. Something went wrong, but what?

The problem comes when we multiply the inequality by (r + l)(r + 4). If this expression is negative,
then we have to reverse the direction of the inequality.

WARNING!! Be careful when multiplying an inequality by a variable expression.

V If the expression is sometimes negative and sometimes positive, then
you can't just multiply the inequality by the expression and proceed
as usual.

We can avoid multiplying the inequality by a variable expression by moving all the terms to one side
and finding a common denominator. Moving all the terms of our inequality to the left gives us a left

489
 CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

side of
r-3 5 - r _ (r - 3)(r + 1) 2(r + l)(r + 4) _ (5 - r)(r + 4)
r + 4+ r + l-(r + l)(r + 4) + (r + l)(r + 4) (r + l)(r + 4)
(r2 - 2r - 3) + (2r2 + 10r + 8) - (-r2 + r + 20)
(r + l)(r + 4)
_ 4r2 + 7r -15
(r + l)(r + 4)
_ (r + 3)(4r - 5)
(r + l)(r + 4)

Our inequality is now r+4 r+3 r+ 1 4r - 5 ∕(r)

(r + 3)(4r -5) r>l + ÷ + + ÷

(r + l)(r + 4) ^ ‘ -1 < r < ∣ + + + - —

To determine when the expression on the left is -3 < r < -1 + + — — +

positive, we have to analyze the values of r for — 4 < r < —3 + _ _ _ —
which each factor is positive. This is just like our
previous work with quadratic inequalities in Sec­ r < -4 +
-
tion 4.5 and polynomial inequalities in Sections 7.2
and 7.3. We make a table as shown, letting f(r) be the expression on the greater side of our inequality.
We see that f(r) is positive for r ∈ (-∞, -4) U (-3, -1) U (∣, +∞). However, this isn't the whole story!

WARNING!! When solving an inequality, we have to pay special attention to

V whether or not the inequality is strict. If the inequality is nonstrict,
we must include those values of the variable for which the two sides
of the inequality are equal.

Here, the inequality is indeed nonstrict, so we must include those values of r such that ∕(r) = 0. This
adds r = -3 and r = ∣ to the solutions we already found, and makes our final answer

r ∈ (-∞, -4) U [-3, -1) U [∣, +∞).

□
Problem 15.11: There is a unique real value A such that for all x with 1 < x < 3 and x ≠ 2,

A 1
< 1999.
x2 - x - 2 x2 - 6x + 8
■
Compute/!. (Source: Duke Math Meet)
Ml

Solution for Problem 15.11: We start by letting ∕(x) = a,2-λ-5 + v⅞-⅛r+g and writing ∕(x) with a common
denominator: κ
A
A 1 _ A(x - 4) + (x + 1)
~ (x - 2)(x + 1) + (x - 2)(x - 4) (x - 2)(x + l)(x - 4)'

490
 15.2. RATIONAL FUNCTION EQUATIONS AND INEQUALITIES

So, our inequality now is

A(x - 4) + (x + 1)
< 1999.
(x-2)(x + l)(χ-4)
Trying to solve for A in terms of x, or for x in terms of A, looks pretty difficult. It's also not clear how
doing either of these would help. Instead, we stop and think about the expression on the left side.

o=≡∈>
Concept: Don't just blindly bash equations and inequalities with algebra when
you're stuck on a problem. Take a little time to think about properties
of the expressions in the problem; sometimes a little qualitative under­
standing will go a long way. Thinking about graphs is a particularly good
way to look for useful properties of complicated expressions.

Considering the graph of ∕(x) for values of x such that 1 < x < 3, it appears that ∕(x) has a vertical
asymptote at x = 2 as a result of the factor x - 2 in the denominator. If this is the case, then ∖f(x)∣ grows
without bound as x gets close to 2. Specifically, if x = 2 is a vertical asymptote of the graph of f, then
there are values of x such that 1 < x < 3 and ∣∕(x)∣ > 1999.
So, we must choose A such that ∕(x) does not have a vertical asymptote at x = 2. The only way that
∕(x) does not have a vertical asymptote at x = 2 is if the numerator of ∕(x) is also 0 for x = 2. Then, ∕(x)
would have a hole in its graph at x = 2 rather than a vertical asymptote. Therefore, we must have
A(2 - 4) + (2 + 1) = 0,
which gives us A = 3/2. When A = 3/2 and x ≠ 2, the left side of our inequality simplifies to
∣(x - 4) + (x + 1) ∣x -5 ∣(x-2) 5
(x - 2)(x + l)(x - 4) (x - 2)(x + l)(x - 4) (x - 2)(x + l)(x - 4) 2(x + l)(χ - 4)
To confirm that this final expression is less than 1999 for all x between 1 and 3, we note that the
expression is maximized when 2(x + l)(x - 4) is as close as possible to 0. Completing the square, we have
2(x+l)(x-4) = 2(x-∣)2-^. The graph of g(x) = 2(x-∣)2-^ is an upward-opening parabola with vertex
(∣, -y). For 1 ≤ x ≤ 3, the graph reaches its highest point to the left of the vertex at (1, g(l)) = (1, -12),
and it reaches its highest point to the right of the vertex at (3,g(3)) = (3, -8). Therefore, the closest g(x)
gets to 0 is -8, which means that ∕(x) is no greater than 5/8 for 1 ≤ x ≤ 3 when A = ∣. □

Exercises
1 t _ 7
15.2.1 Find all values of t such that - ---- - + - ---- -
2t - 3 3f -1 “ 5’
3r2 - 12r
15.2.2 Find all r such that —-----------
r2 - 3r - 4 r+7
x -12 2 _ 5
15.2.3 Find all solutions to the equation —------
x ^⅛~ 3x x x + 3'
χ^ _ q
15.2.4 Find all real values of x such that —------ > 0.
x2-25
6
15.2.5 Find all real values of x that satisfy —-— + —≥ 1. (Source: ARML)
7 x+1

491
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

15.3 Even and Odd Functions

A function f is called an even function if ∕(x) = ∕(-x) for all x in the domain of f, and it is called an odd
function if ∕(x) — -f(-x) for all x in the domain of f. As well see, these names are much less arbitrary
than they seem!

Problems

Problem 15.12: For each of the following functions, state if the function is even, odd, or neither.

^=χ41-2 ∕(x) =
(a) ∕W = χ (C) (e)

(b) ∕ω = 3χ (d) ∕(X) = 5∙'2 (f) ∕(x) - -Vχ2

Problem 15.13:
(a) For each of the following polynomials, determine if the polynomial is even, odd, or neither:

p(x) = x3 - x,
q(x) = x4 - 3x3 + 2x,
r(x) = 2x5 -3x-3,
s(x) = 6x42 +x6 - 2x2 + 1.

(b) Suppose f(x) is a polynomial such that f is an even function. Is it possible for the coefficient of
x to be nonzero?
(c) Suppose ∕(x) is a polynomial such that f is an odd function. Is it possible for the coefficient of
x4 to be nonzero?
(d) Explain why the name "even" is appropriate for functions f that satisfy ∕(x) = ∕(-x) for all x in
_____ their domains. What about the name "odd"?_________________

Problem 15.14: Is it possible to tell just by looking at the graph of a function whether the function is
even, odd, or neither? If so, how can you tell? If not, why is it not possible?

Problem 15.15: Let f and g be even functions that are defined for all real numbers.
(a) Must ∕ + g be even?
(b) Must f ∙ g be even?

Problem 15.16: Let f be a function with all real numbers as its domain.
(a) Show that the function g(x) = ∕(x) + ∕(-x) is even.
(b) Show that the function ∕z(x) = ∕(x) - f(~x) is odd.
(c) Show that every function with domain 1R can be written as the sum of an even function and an
odd function. For example, if ∕(x) = x2+x + 3^∕x-9, we can write ∕(x) = (x2 - 9) + (x + 3 ^∕x)∙

492
 15.3. EVEN AND ODD FUNCTIONS

Problem 15.12: For each of the following functions, state if the function is even, odd, or neither.

(a) ∕(x) = x (c) ∕(x) = - 4l (e) f(x) = tfx

x* -2
(b) ∕(x) = 3v (d) ∕(x) = 5χ2 (f) ∕(x) = λ⅞2

Solution for Problem 15.12:

(a) We have ∕(x) - x and ∕(-x) = -x, so ∕(x) = -∕(-x), which means ∕(x) = x is an odd function.
(b) We have ∕(x) = 3λ and ∕(-x) = 3~x = l∕3x. For instance, we have ∕(1) = 3 and ∕(-l) = 1/3, so it is
not true that ∕(x) = ∕(-χ) or ∕(x) = -∕(-x) for all x.
1 11
(c) We have ∕(x) = and ∕(-x) = ——------ = ------- , so ∕(x) = ∕(-x) for all x. Therefore, f is an
even function. ' * , ^2
f^ * ^2

(d) We have f (-x) - = 5x = ∕(x) for all x, so f is an even function.

*
(e) We have ∕(-x) = V^^ = -i∕x = -f(x) for all x, so ∕(x) = ^fx is an odd function.
(f) We have /(—x) = -ζ∕(-x)2 = Vx2 = ∕(x) for all x, so ∕(x) = Vx2 is an even function.

Problem 15.13:
(a) For each of the following polynomials, determine if the polynomial is even, odd, or neither:

p(x) = x3 - X,
<j(x) = x4 - 3x3 + 2x,
r(x) = 2x5 - 3x - 3,
s(x) = 6x42 + x6 - 2x2 + 1.

(b) Suppose ∕(x) is a polynomial such that f is an even function. Is it possible for the coefficient of
x to be nonzero?
(c) Suppose ∕(x) is a polynomial such that f is an odd function. Is it possible for the coefficient of
x4 to be nonzero?
(d) Explain why the name "even" is appropriate for functions f that satisfy ∕(x) = ∕(-x) for all x
in their domains. What about the name "odd"?

Solution for Problem 15.13:

(a) We have p(-x) = (-x)3 - (-x) = -x3 + x = -p(x), so p(x) is odd.
We have q(-x) = (-x)4 - 3(-x)3 + 2(-x) = x4 + 3x3 - 2x, which is not the same as t∕(x) or -q(x).
Therefore, q(x) is neither even nor odd.
We have r(-x) = 2(-x)5 - 3(-x) - 3 = —2x5 + 3x — 3, which is not the same as r(x) or -r(x).
Therefore, r(x) is neither even nor odd.
We have s(-x) = 6(-x)42 + (-x)6 - 2(-x)2 + 1 = 6x42 + x6 - 2x2 + 1 = s(x), so s(x) is even.

493
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

For each of the following parts, let ∕(x) = anxtl + an-↑xn 1 + ∙ ∙ ∙ + a↑x + z⅞∙
(b) Two polynomials are equal for all values of x if and only if all their corresponding coefficients are
the same. If f is an even function, then we must have ∕(x) = ∕(-x) for all x, which means that the
coefficients of ∕(x) must be the same as the coefficients of ∕(-x). However, the coefficient of x in
∕(-x) is -«i, while the coefficient of x in ∕(x) is a-[. Therefore, we must have -a↑ = a↑, which means
«1 = 0. We conclude that if polynomial ∕(x) is an even function, then the coefficient of x must be 0.
(c) If f is odd, then ∕(x) = -∕(-x), which means the coefficients of -∕(-x) match those of ∕(x). The
coefficient of x4 in ∕(-x) is a⅛, so the coefficient of x4 in -∕(-x) is -⅜. This must equal the coefficient
of x4 in ∕(x), which is a½. So, we have c⅛ = -a⅛, which means «4 = 0. This tells us that if polynomial
∕(x) is an odd function, then the coefficient of x4 must be 0.
(d) The first two parts guide us. We let
∕(x) = anxn + an-ιxn~1 + + a3x3 + a2x2 + ∩ιx + ^o∙

Then, we have
∕(-x) = ⅜(-l)πx" + ⅜-ι(-l)π-1xn^1 +----- a3x3 + ∩2X2 ~ aιχ + ao-

If f is even, then these two polynomials must be the same, which means that their corresponding
coefficients must match. However, the coefficients of corresponding terms with odd powers of x
are opposites. Since we can only have α1∙ = -α,∙ if αz = 0, we have ∕(x) = ∕(-x) only if α1- = 0 for all
odd i.
The coefficients of the even powers of x always match in ∕(x) and ∕(-x), so a polynomial ∕(x)
that only has even powers of x satisfies ∕(x) = ∕(-x) for all x. Therefore, a polynomial ∕(x) is an
even function if and only if it only has even powers of x. Hence the name "even function."
We now have a pretty good idea now where the name "odd function" comes from. We have

-∕(-x) = an(-l)n+1xn + an-↑(-l)nxn~1 + ∙ ∙ ∙ + a3x3 - a2x2 + «ix - r?o-

Therefore, we have ∕(x) = ~f(~x) only if αl∙ = -ai when i is even. This tells us that ai = 0 for all
even i. The coefficients of the odd powers of x always match in ∕(x) and -∕(-x), so a polynomial
∕(x) that only has odd powers of x satisfies ∕(x) = -∕(-x) for all x. Therefore, a polynomial ∕(x) is
an odd function if and only if it only has odd powers of x. Hence the name "odd function."

Important: A polynomial function ∕(x) is an even function if and only if its only
nonzero coefficients are coefficients of even powers of x, and ∕(x) is an
odd function if and only if its only nonzero coefficients are coefficients
of odd powers of x.

So, we can tell at a glance if a polynomial is even or odd. Can we tell by a glance at a function's
graph if it is even or odd?

Problem 15.14: Is it possible to tell just by looking at the graph of a function whether the function is
even, odd, or neither? If so, how can you tell? If not, why is it not possible?

494
 15.3. EVEN AND ODD FUNCTIONS

Solution for Problem 15.14: If a function f is even, then ∕(x) = ∕(-x). Suppose (a,b) is on the graph of
y = ∕(x), so we have b = f(a). Because f is even, we have ∕(-λ) = f(a) = b, which means the point (-a,b)
must also be on the graph of y = ∕(x). The point (-fl, b) is the reflection of the point (a, b) over the y-axis.
Therefore, for each point on the graph of an even function, the reflection of that point over the y-axis
must also be on the graph of the function. This means that the graph of an even function must have the
y-axis as a line of symmetry, as shown in the graph at left below.
Conversely, suppose the graph of a function f is symmetric about the y-axis. This means that if (a, b)
is on the graph, then (-a,b) must be as well, which means f(a) = f(-a). Therefore, the graph of any
function that is symmetric about the y-axis must be the graph of an even function.

Figure 15.1: Graph of an Even Function Figure 15.2: Graph of an Odd Function

If a function f is odd, then ∕(x) = — ∕(-χ). Suppose the point (a,b) is on the graph of y = ∕(x), so
we have b = f(μ), as before. Since f(x) = -f{-x'), we have f(-a) = -∕(-(-zz)) = ~f(a) - ~b, so the point
(—a, —b) is on the graph of f. The point (—α, -b) is the image of (a, b) upon a 180o rotation about the origin.
We say that two such points are "symmetric about the origin." So, for each point on the graph of an odd
function, the rotation of that point 180o about the origin is also on the graph of the function. This means
that the graph of an odd function must be symmetric about the origin, as shown at right above.
Conversely, suppose the graph of a function f is symmetric about the origin. This means that if (a, b)
is on the graph, then (-a, -b) must be as well, which means that f(a) - -f(-a). So, the graph of any
function that is symmetric about the origin must be the graph of an odd function. □

Important: A function is even if and only if its graph is symmetric about the y-axis,

ψ and a function is odd if and only if its graph is symmetric about the
origin.

Solution for Problem 15.15:

(a) Because f and g are even, we have ∕(x) = ∕(-x) and g(x) = g(-x), so we have ∕(-x) + g(-x) =
∕(x) + g(x). So, if h = f + g, we have h(x) = h(-x), which means h is even.
(b) We have ∕(-x) ∙ g(-x) = ∕(x) ∙ g(x) for all x, so f ∙ g is even.
□
In the Exercises, you'll have a chance to see what happens to f + g and f ∙ g if f and/or g is odd
instead of even.

495
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

We'll finish our exploration with a surprising fact about functions.

Problem 15.16: Show that every function with domain JR can be written as the sum of an even
function and an odd function.

Solution for Problem 15.16: We can obviously write any polynomial as the sum of an even and an odd
function, since we can break any polynomial f(x) into two polynomials such that one has all the even
powers of x in ∕(x) and the other has all the odd powers. For example, if ∕(x) = x4 - 3x3 + 7x - 13, we
have
∕(x) = (x4 - 13) + (-3x3 + 7x).
But what about functions like g(x) = (x4 - x + 9)∕(x2 + 5) or ∕z(x) = ∣x2 - x|?
One way to show that any function g can be written as the sum of an even function and an odd
function is to find those two functions. But g can be just about anything! How can we possibly find the
even and odd functions into which we can split g? The only thing we have to work with is g, so we'll
have to find an even function and an odd function in terms of g.
We look back to our polynomial ∕(x) = x4 — 3x3 + 7x - 13 for inspiration.

Concept: When trying to prove a general fact, consider specific examples. You may
Q. lj J be able to use them as inspiration for a proof.

We know we can split ∕(x) into x4 - 13 and -3x3 + 7x. We hope to find a way to write these both in
terms of f. Let's focus on the even piece first. We get the even powers of x when we write ∕(x), but we
also get the odd powers of x. We'd like to cancel those odd powers out. We know how to do that! The
coefficient of each odd power of x in ∕(-x) is the opposite of the corresponding odd power of x in ∕(x).
So, ∕(x) + ∕(-x) will not have any odd powers of x:
∕(x) + ∕(-x) = (x4 - 3x3 + 7x - 13) + (x4 + 3x3 -7x- 13) = 2(x4 - 13).

Therefore, ∕(x) + ∕(~x) equals twice the "even part" of ∕(x). So, we can write the even part of ∕(x) as
(∕(χ) + ∕(-x))∕2. '

Now that we've found the even part, the path to the "odd part" is clear. If we want to eliminate the
even powers of x in ∕(x), we subtract ∕(-x) instead of adding it:
∕(x) - ∕(-x) = (x4 - 3x3 + 7x - 13) - (x4 + 3x3 -7x- 13) = 2(-3x3 + 7x).

So, the odd part of ∕(x) is just (∕(x) - ∕(-x))∕2.

Well, that worked for our specific ∕(x). Will it work for any function g(x)? Specifically, can we write
g(x) as the sum of an even function and an odd function by writing it as the sum of (g(x) + g(-x))∕2 and
(g(x) - g(—x))/2? First, we confirm that these pieces do in fact add to g(x):
)
+ g(-
*)
g(
* gW-g(→)
2 2
= gW∙
So far, so good. Now, we must confirm that one piece is even and the other is odd. If we let p(x) =
*)
(&( + g(-^))∕X we have

496
 15.4. MONOTONIC FUNCTIONS

so p is even. Similarly, letting q{x) = (g(χ) - g(-x))∕2, we have

Wγ(-υ
,(-λ-) . ×lΞ±Ξxω . .* =
so q is odd. Therefore, we can write any function g as the sum of an even function and an odd function
by writing g(x) = p(χ) + q(χ), where p(x) = (g(χ) + g(-χ))∕2 and q(x) = (g(x) - g(-x))∕2. □

Fveιτi⅜es

15.3.1 State whether each of the following functions is odd, even, or neither:
(a) f(x) = x4 + x6 + χ8 + 2008 - ∣x∣ (c) ∕ι(χ) = ∣3x3 - 7x∣
(b) g(χ) = 3λ λ (d) y(x) = √χ2 _ 4

15.3.2 Suppose that f and g are both odd functions.

(a) Must f + g be an odd function, an even function, or neither?

(b) Must f ■ g be odd function, an even function, or neither?

15.3.3 Suppose that g(x) is a polynomial that is an odd function. Find all possible values of the product
of the roots of g(x).

15.3.4 Suppose that ∕(x) = g(x)∕∕z(x) is a rational function, where g(x) and ∕z(x) are distinct polynomials.
If f is an even function, must g and h be even functions?

15.4 Monotonic Functions

A function f is a monotonically increasing function if f(a) > f(b) whenever a> b. Often we will just say
the function is "increasing" rather than "monotonically increasing." For example, back in Section 13.1,
we noted that exponential functions ∕(x) = cx, where c is a constant, are monotonically increasing if
c > 1. Similarly, a function is monotonically decreasing if f(a) < f(b) whenever a > b, and we will
often say that such a function is "decreasing" rather than "monotonically decreasing." The graph of
an increasing function goes strictly upward as it goes from left to right, and the graph of a decreasing
function goes strictly downward as it goes from left to right.

y y

Figure 15.3: An Increasing Function Figure 15.4: A Decreasing Function

497
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

Problems

Problem 15.17: In this problem, we show that f(x) = x3 is a monotonically increasing function.
(a) Let c be a positive constant. Find ∕(x + c) - ∕(x).
(b) Your answer in part (a) should be a quadratic in x. Show that this quadratic does not have real
roots.
(c) Show that ∕(x + c) - ∕(x) is positive for all x and all positive c.
(d) Why does part (c) tell us that ∕(x) is a monotonically increasing function?

Problem 15.18:
(a) Is the sum of two monotonically increasing functions necessarily monotonically increasing? If
so, prove it. If not, give an example of two monotonically increasing functions whose sum is
not monotonically increasing.
(b) Is the product of two monotonically increasing functions necessarily monotonically increasing?
If so, prove it. If not, give an example of two monotonically increasing functions whose product
is not monotonically increasing.

Problem 15.19: If a function is monotonically increasing, must the function be invertible? Why or
why not? What if a function is monotonically decreasing?

Problem 15.20: Show that if ∕(x) is a monotonically increasing function, and ∕(x) is positive for all
values of x in the domain of f, then l∕∕(x) is a monotonically decreasing function.

Problem 15.21:
(a) Show that ∕(x) = √x is monotonically increasing.
(b) Find the largest possible value of √x + 4 - √x.

Problem 15.22:
(a) Show that p(x) = x(15 - x) decreases monotonically when x ∈ (7.5, +∞).
(b) Which is greater, π2 or 7.5?
(c) Without using a calculator, determine if π2(15 - π2) is greater than, less than, or equal to 50.

Problem 15.23:
(a) Use the fact that 8λ is an increasing function to show that log8 x is monotonically increasing.
(b) Find the smallest possible value of log8(x3 - 4x2 + x + 26) - log8(x + 2).

Problem 15.17:
(a) Suppose that ∕(x + c) - ∕(x) > 0 for all x and c such that c > 0 and both x and x + c are in the
domain of/. Explain why ∕ must be monotonically increasing.
(b) Use part (a) to show that ∕(x) = x3 increases monotonically.

498
 15.4. MONOTONIC FUNCTIONS

Solution for Problem 15.17:

(a) Adding ∕(x) to both sides of f(x + cy) - f{x) > 0 gives f{x + c) > f(x) for c > 0. Suppose that a and
b are both in the domain of f, and that b > a. Because f(x + c) > f(x) for all c > 0, letting x = a
and c = b - a gives us f(b) > f(μ). Therefore, we have f(b) > f(a) for any numbers a and b in the
domain of f such that b > a. This means that f is monotonically increasing.
Important: If ∕(x+c) - f (x) is positive for all values of x and c such that c > 0 and both
M? x + c and x are in the domain of f, then f is monotonically increasing.
Similarly, if ∕(x + c) - ∕(x) is negative for all values of x and c such that
c > 0 and both x + c and x are in the domain of f, then f is monotonically

(b) To use part (a), we must show that ∕(x + c) - ∕(x) > 0 if c > 0 and ∕(x) = x3. We have

∕(x + c) - ∕(x) = (χ + c)3 - x3 = 3cx2 + 3c2x + c3 = c(3x2 + 3cx + c2).

Because c > 0, the expression c(3x2 + 3cx + c2) is positive whenever 3x2 + 3cx + c2 is positive.
The graph of g(x) = 3x2 + 3cx + c2 is an upward-opening parabola. If it never crosses the x-axis,
then g(x) must be positive for all x. So, we examine its discriminant to determine if g(x) = 0 for
any real values of x. The discriminant is

(3c)2 - 4(3)(c2) = 9c2 - 12c2 = -3c2,

which is negative for all real values of c, so it is negative for all positive values of c. Therefore,
3x2 + 3cx + c2 has no real roots. So, we know that 3x2 + 3cx + c2 is positive for all values of x and
all positive c, which tells us that ∕(x + c) - ∕(x) = c(3x2 + 3cx + c2) is positive for all x and all c > 0.
This means that ∕(x) = x3 increases monotonically.

Problem 15.18:
(a) Is the sum of two monotonically increasing functions necessarily monotonically increasing? If I
so, prove it. If not, give an example of two monotonically increasing functions whose sum is ∣
not monotonically increasing.
(b) Is the product of two monotonically increasing functions necessarily monotonically increasing? I
If so, prove it. If not, give an example of two monotonically increasing functions whose product
is not monotonically increasing.

Solution for Problem 15.18:

(a) Yes. Suppose f and g are monotonically increasing, and h = f + g. We have

h(x + c)- ∕z(x) = ∕(x + c) + g(x + c) - (∕(x) + g(x)) = ∕(x + c) - f(x) + g(x + c) - g(x).

Because f and g are increasing, when c > 0 we have ∕(x + c) - ∕(x) > 0 and g(x + c) - g(x) > 0. So,
we have
h(x + c) - ∕ι(x) = ∕(x + c) - ∕(x) + g(x + c) - g(x) > 0,
which means that h is monotonically increasing.

499
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

(b) No. The function ∕(x) = x is obviously monotonically increasing—as x increases, so does ∕(x). If
∕(x) = x, g(x) = xr and ∕z(x) = ∕(x) ∙ g(x) = x2, then f and g are monotonically increasing, but h is
not. For example, ∕z(-2) > ∕z(0). (There are infinitely many other choices of f and g that we could
pick such that f and g are monotonically increasing, but f ∙ g is not.)

Back in Section 13.2, we used the fact that an exponential function ∕(x) = ax with a > 1 is mono­
tonically increasing to prove that the function is invertible. This might make us wonder if every
monotonically increasing function is invertible.

Problem 15.19: If a function is monotonically increasing, must the function be invertible? Why
why not? What if a function is monotonically decreasing?

Solution for Problem 15.19: Let ∕(x) be a monotonically increasing function. Because ∕(x) is increasing,
we know that if a > b, then f(a) > f(b). In other words, if two inputs to f are not equal, then their
outputs from f cannot be equal. So, f must have an inverse.
If ∕(x) is monotonically decreasing, we have f(b) < f(a) for all a > b. Again, we see that if two inputs
to f are not equal, then their outputs from f are not equal, so f has an inverse. □

Important: If a function is either monotonically increasing or monotonically decreas­

ing, then it has an inverse.

Problem 15.20: Show that if ∕(x) is a monotonically increasing function, and ∕(x) is positive for all
values of x in the domain of f, then l∕∕(x) is a monotonically decreasing function.

Solution for Problem 15.20: Let g(x) = jfa. We wish to show that if a > b, then g(b) > g(α).

Because ∕(x) is monotonically increasing, we have ∕(α) > f(b) whenever a> b. Since ∕(x) is positive
for all x, we can divide our inequality by ∕(zz) and ∕(b) to give j⅛y > y⅛)∙ Because g(x) = we have
g(b) > g(a). Therefore, g(x) is monotonically decreasing, as desired. □
Let's put this principle to work on a specific function.

Problem 15.21:
(a) Show that ∕(x) = √x is monotonically increasing.
(b) Find the largest possible value of √x + 4 - √x.

Solution for Problem 15.21:

(a) We wish to show that if a > b ≥ 0, then y∕a> y∕b. (We must have a and b nonnegative so that both
are in the domain of ∕.) What's wrong with this proof:

Bogus Solution: Since a > b, taking the square root of both sides gives us yfa > >∕b.
Therefore, ∕(x) = √x is monotonically increasing.

500
 15.4. MONOTONIC FUNCTIONS

This "proof" assumes that if a > b, then √<i > Vb. But this is the very fact that we are asked to
prove! So, we have to find another way to show that f is increasing.
What's wrong with this proof:

Bogus Solution: If y/a > Vb, then, because both sides are positive, we can square
∣J‰I both sides to give a > b. Therefore, if a > b, then y/a > ∖[b.

This argument is backwards; we show that if ∖∕a > >∕b, then a > b. But we can't conclude just
from this fact that "If a > b, then √a > √b."
WARNING!! Let X and Y be mathematical statements. If we show that "If X, then
a Y," then we cannot conclude that "If Y, then X." Doing so would be
like reasoning that the statement "If I live in America, then I live on
the Earth," implies that "If I live on the Earth, then I live in America."

We've shown that a > b if y/a > y∕b, but we haven't addressed the possibility that a > b if
y/a < y/b or y/a = Vb. So, we investigate these other two possibilities.
If y/a < y∕b, then squaring both sides gives a < b. So, we can't have a > b if y/a < VR
If y/a = y∕b, then squaring both sides gives a = b, so we can't have a> & if y/a = y∕b, either.
So, of the three possibilities y/a > y[b, y/a < Vb, and y∕α = Vb, only y/a > yJb allows a > b.
Therefore, if a > b, then we must also have y∕α > ^∕b. Make sure you see why this solution is valid,
but the second Bogus Solution above is not.
Let g(x) = Vx + 4- y/x. We start by experimenting. We evaluate (or approximate) g(x) X gW
for various values of x at right. It looks like g(x) is monotonically decreasing. If we 0 2
can show that it is, then we can conclude that g(x) is maximized when x is minimized, 1 1.24
which is when x = 0. 2 1.04
If we start with g(x + c) - g(x), we have 5 0.76
10 0.58
g(x + c) - g(x) = (Vx + 4 + c - √x + c) - (√x + 4 - √χ),

which is quite a mess. Before we dive into that scary expression, we look for a faster way.
Our observation in Problem 15.20 offers us a considerably simpler approach. We note that

1 _ 1 _ 1 Vx + 4 + √x _ √x + 4 + √χ
)
#(
* √x + 4 - √x √x + 4 - √x Vx + 4 + y/x 4

Because both √x + 4 and y/x are monotonically increasing, our result from Problem 15.18 tells
us that their sum is monotonically increasing. Therefore, the function l∕(g(x)) is monotonically
increasing. Since g(x) > 0 for all x, this means that g(x) is monotonically decreasing. So, the
function g(x) is maximized when x is minimized, which means the maximum value of g(x) is
g(0) = 2.

Many functions are neither monotonically increasing nor monotonically decreasing, but they are
increasing or decreasing over parts of their domains.

501
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

Problem 15.22: Without using a calculator, determine if π2(15 - π2) is greater than, less than, or
equal to 50.

Solution for Problem 15.22: Multiplying out π2(15 - π2) doesn't look like much fun. So, we look for
different ways to think about this expression. We can view it as ∕(π2), where ∕(x) = x(15 - x). This seems
much simpler. Moreover, we have ∕(5) = ∕(10) = 50, so our problem now is to compare ∕(π2) to ∕(5) or
∕(10). The graph of ∕(x) is a downward-opening parabola. We can see exactly what parabola this is by
completing the square:
„ x 9 ∕ 15∖2 225
∕(x) = 15x-x2 = -(x- —j + —∙

So, the axis of symmetry of the parabola is x = 7^. To the right of this line, f is decreasing. Both x = π2
and x = 10 are to the right of this line, so now we must compare n2 to 10. Since π2 ≈ 9.87, we have
π2 < 10. Because ∕(x) is strictly decreasing for all x > 7.5, we have ∕(π2) > ∕(lθ) = 50. Therefore,
π2(15 - π2) is greater than 50. □
Note that our invoking the graph of a parabola is not a proof that ∕(x) is decreasing for x > 7.5; it's
merely an intuitive explanation. You'll have a chance to prove it formally as an Exercise.

Problem 15.23: Find the smallest possible value of log8(x3 - 4x2 + x + 26) - log8(x + 2).

Solution for Problem 15.23: We first simplify the expression by using a logarithm identity:

log8(x3 - 4x2 + x + 26) - log8(x + 2) = log8 ( x3 - 4x2 + x + 26 ∖

x+2 f

Next, we note that x + 2 divides the polynomial evenly and leaves us with

------- -------------- j = 1°g8<χ2 ~ 6λ' + 13)∙

x+2

Now what?
We might think that log8(x2 — 6x + 13) is minimized when x2 - 6x + 13 is minimized, but this is only
true if log8 x is a monotonically increasing function. It seems intuitively obvious that it is increasing, but
let's check to be sure. If ∕(x) = log8 x, we have

∕(x + c) - ∕(x) = log8(x + c) - log8 x = log8 = log8 (1 + ∣) ∙

Is the power to which we must raise 8 to get 1 + ξ positive or negative? We first note that c and x must
be positive, so 1 + ξ > 1. Since 80 = 1 and 8x is monotonically increasing, we must raise 8 to a positive
power to get any number greater than 1. So, log8 (1 + is positive, which means that ∕(x) = log8x is
monotonically increasing.
We now know that log8(x2 - 6x +13) is minimized when x2 - 6x +13 is minimized. Since x2 - 6x +13 =
(x-3)2+4, the expression x2-6x+13 is minimized when x = 3. So, the minimum value of log8(x2-6x+13)

2
log8(32-6∙3 + 13) = loga4= -.

502
 15.5. SUMMARY

As a final check, we let x = 3 in our original expression to ensure that both logarithms in the original
expression are defined for x = 3. They are, so we have indeed found the correct minimum. □

I ver< i¼v¼

15.4.1 Suppose that f (x) and g(x) are positive for all x. If ∕(x) and g(x) are both monotonically increasing,
then must ∕(x) ∙ g(x) be monotonically increasing?
15.4.2 Prove that x(15 — x) is a decreasing function for x > 7.5.
15.4.3 Find all x such that 8λ(3x + 1) = 4, and prove that you have found all values of x that satisfy this
equation.

15.4.4 Show that ∕(x) = x” increases monotonically for any positive odd integer n.
15.4.5 For each of the following functions, find the maximum of the function, and prove that the
function cannot equal any greater value.

(a) ∕(x) = √x + 3 - √x-3, for x ≥ 3.

(b) * g(χ) = √x + l - ψχf for x > 0. Hints: 316

15.5 Summary

Any function that can be expressed as one polynomial divided by another is called a rational function.
An asymptote of a graph is a line that the graph approaches but does not intersect as x or y becomes
very far from 0. Determining the asymptotes of the graph of a rational function can be very helpful in
graphing the function and understanding its behavior.

Important: Let p(x) and d(x) be polynomials, where rf(x) is not 0 for all x. If c is a
constant such that d(c) = 0andp(c) ≠ 0, then the graph of ∕(x) = p(x)∕d(χ)
has a vertical asymptote at x = c.

To find horizontal and slant asymptotes, we divide the denominator of a rational function into the
numerator of the function. If the resulting quotient is a constant, c, then y = c is a horizontal asymptote.
If the quotient is a linear function with a nonzero linear coefficient, then the graph of the quotient is a
slant asymptote of the graph of the rational function.

Solving equations involving rational functions usually involves getting rid of all the denominators
by multiplying through by them. But we have to be careful:

WARNING!! When you solve an equation involving rational functions, you must
V check that your solutions do not make any denominator equal to 0.

When solving inequalities involving rational functions, we usually move all the terms to one side of
the inequality, form a single rational function on that side, then compare that rational function to 0. This

503
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

typically involves factoring the numerator and denominator of the rational function, and analyzing the
signs of the factors for different values of the variable.

WARNING!! When solving an inequality, we have to pay special attention to

Y whether or not the inequality is strict. If the inequality is nonstrict,
we must include those values of the variable for which the two sides
of the inequality are equal.

A function f is called an even function if ∕(x) = f(~x) for all x in the domain of f, and it is called an
odd function if ∕(x) = -∕(-x) for all x in the domain of f.

Important: A function is even if and only if its graph is symmetric about the y-axis,

ψ and a function is odd if and only if its graph is symmetric about the
origin.

A function f is a monotonically increasing function if f(d) > f(b) whenever a > b. Similarly,
a function is monotonically decreasing if f(a) < f(b) whenever a > b. We often exclude the word
"monotonically," and simply refer to a function as "increasing" or "decreasing." The graph of an
increasing function goes strictly upward as it goes from left to right, and the graph of a decreasing
function goes strictly downward as it goes from left to right.

Important: If ∕(x+c) - f(x) is positive for all values of x and c such that c > 0 and both

ψ χ + c and x are in the domain of f, then f is monotonically increasing.

Similarly, if ∕(x + c) - ∕(x) is negative for all values of x and c such that
c > 0 and both x + c and x are in the domain of f, then f is monotonically
decreasing.

Problem Solving Strategies---------------------------- ------------------------------------

ι-Concepts:
------ • Don't limit yourself to an equation or expression you're given. Dif­
ferent algebraic forms of the same equation or expression can offer
O≡≡≡ different information, so experiment with manipulating equations
and expressions into different forms.

• When faced with a new type of equation, focus on what makes the
equation difficult to solve. Try to remove this difficulty and turn the
equation into one you already know how to handle.

• When you're stuck on an algebra problem, take a little time to think

about properties of the expressions in the problem; sometimes a
little qualitative understanding will go a long way. Thinking about
graphs is a particularly good way to look for useful properties of
complicated expressions.

• When trying to prove a general fact, consider specific examples. You

may be able to use them as inspiration for a proof.

504
 REVIEW PROBLEMS

Review Problems

15.24 Sketch graphs of each of the following. Identify the domain, range, and any asymptotes of each
function.
2x + 1 x2 - 4x + 3
(a) *
)
∕( = (b) ∕(x) =
3-x x-2
15.25 Suppose ∕(x) = p(x)∕q(x), where p(x) = anx'1 + an-↑xn~1 + —fa0 and q(x) = bnxn + bn~↑xn~l + ∙∙∙ + bo
are nonzero polynomials. Suppose also that anbn ≠ 0, and that the graph of y = f(x) has a horizontal
asymptote. Find the equation of the horizontal asymptote in terms of the coefficients of p(x) and q(x).
x3 - 2x2 - 15x0
15.26 Which of the four graphs below could be the graph of f(x) = For each that
3x2 + 12x + 12'
cannot be the graph of f, explain why not.

15.28 For each of the following functions, state whether or not the graph of the function has a slant
asymptote. If it does have a slant asymptote, then find it.
2x-l = (x-2)(3x-l) 2x3 + 5x2 - 4x - 3
(c) , (x + l)(x-3) (e) ∕(x) =
x+2 x2 + 2x
(x - 2)(3x - 1) x-2 2x3 + 5x2 - 4x - 3
(b) (d) (f)
x+1 x(x + 1) x+2
Suppose ∕(x) = g(x)∕∕ι(x), where g(x) and ∕z(x) are polynomials. What must be true about g(x) and ∕z(x) if
the graph of f has a slant asymptote?
15.29 Is it possible for the graph of a rational function to have:

(a) Two horizontal asymptotes?

(b) Two vertical asymptotes?
(c) A horizontal and a vertical asymptote?
(d) A horizontal and a slant asymptote?
(e) A slant and a vertical asymptote?
(f) Two slant asymptotes?

505
CHAPTER 15. SPECIAL CLASSES OF FUNCTIONS

x2 — 2x + 3 9
15.30 What is the range of the function ∕(x) =
x2 + 2x - 3 ‘
1
15.31 Find all positive real values of x that satisfy------ — + ≤ 1.
x+ √x χ - √X

15.32 Find the solutions of the equation —-------------- - = 1. (Source: AHSME)

x2 — 4 x - 2
15.33 Find all x such that -—∣ - — = 3x + 1.
x+5 1 -x
1
15.34 Find the maximum value of the function ∕(x) = 2 ∙

15.35 Find all real numbers x such that —----------- < 1.

x2 + 3x + 2
15.36 Suppose that ∕(x) is a polynomial that is an even function. Find all possible values of the sum of
the roots of ∕(x).
15.37 Let f and g be functions such that the domain of g includes the range of f.

(a) If f is even, then must go f be even, odd, or neither?

(b) If f is odd, then must g o f be even, odd, or neither?

15.38 If ∕(x) is even and g(x) is odd, identify each of the following as definitely even, definitely odd, or
neither:
(a) ∕(-x) (c) ∕(
)
* ∙ g(x) (e) ∕(g(2x))
(b) g(x2) (d) ∕(χ) + g(χ)

15.39 Letg(x) = ∣∕(x)∣.

(a) If f is even, must g be an even function?

(b) If f is odd, must g be an even function, an odd function, or not necessarily either?
(c) If g is even, then must it be true that f is either even or odd?

15.40 Suppose that ∕(x) is nonzero and monotonically increasing. Must l∕∕(x) be monotonically
decreasing? (Make sure you see why this is a different question from Problem 15.20.)
15.41 Show that ⅛[x is a monotonically increasing function.
15.42 Suppose a and b are constants greater than 1. How many solutions x are there to the equation
(logfl x)(logfo x) = 2007? Prove your answer is correct.
X
15.43 Show that the function ∕(x) = —is increasing for all positive x.

15.44 Without using a calculator, order the following from greatest to least:

10 - 3√∏, 7 - 4√3, 5√41 - 32, 9 - 4√5.

506
 CHALLENGE PROBLEMS

Challenge Problems

15.45 Let a, b, and c be positive constants. Show that x = 0 is the only value of x such that

ya + bx + Vb + ex + y/c + ax = yfb - ax + Vc - bx + y/a - ex.

x + √x + 4
15.46 Solve the inequality < 2. Hints: 306,15
x -1
15.47 Solve the equation √x + 2 + √x + 3 + √χ + 4 = 0. (Source: Finland) Hints: 87

15.48 Find all values of the constant A such that the equation ——1- 2 — -------- has exactly one
solution. x+2 Ax +1 j

15.49
* Find the minimum possible value of ∕(x) = log10(x3 - 5x2 + llx -15) - log10(x - 3), or show that
the function has no minimum. Hints: 258,25

15.50
* Find the largest positive value attained by the function ∕(x) = Vδx - x2 - V14x - x2 - 48, where
x is a real number. (Source: AHSME) Hints: 288,45

15.51
* Find the number of real solutions to the equation 567x + 678r = 789x. Hints: 298

Extra! The Nobel Prize is awarded in chemistry, economics, medicine, and physics, but not
"* l"* mathematics. Shortly after the Nobel Prize was established in the late 1800s, Norwegian
mathematician Sophus Lie began advocating that the Nobel Prize's lack of a mathematics
prize be remedied by establishing an Abel Prize, named after the great mathematician
Niels Abel. Over a century later, Lie's vision was realized, as the Norwegian government
announced in August, 2001, the formation of an Abel Fund worth 200 million Norwegian
krone (which is worth around 40 million dollars now) to fund an international Abel Prize.
The first Abel Prize was awarded in 2003 to Jean-Pierre Serre. He was awarded
6,000,000 Norwegian krone (worth around $850,000 at the time). In 2008, the prize
was awarded to John Griggs Thompson and Jacques Tits, who shared the 6,000,000
Norwegian krone (worth around $1,200,000 now).
Another major prize given to mathematicians is the Fields Medal. The Fields Medal
is awarded every four years to two, three, or four mathematicians not more than 40
years old. Whereas the Abel Prize has been used mainly to reward a lifetime of achieve­
ment in mathematics, the Fields Medal is designed to recognize and encourage young
mathematicians.
There are even special prizes for specific problems. In 2000, the Clay Mathematics
Institute announced it would award $1,000,000 for solutions to each of seven Millen­
nium Problems. So far, only one of the seven problems has been solved, so there are still
prizes to be claimed!
For more information about these prizes, and about the Millennium Problems, visit
the pages linked from our links page cited on page vi.

507
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Always watch where you are going. Otherwise, you may step on a piece of the forest that ιυas left out by mistake.
- Winnie the Pooh

CHAPTER

I I
Piecewise Defined Functions

Most of the functions we have studied so far have a single rule that is applied to every valid input.
For example, the function ∕(x) = 2x doubles any input. In this chapter, we look at functions that apply
different rules to different inputs, as well as some functions that require special notation to denote the
rules they represent.
While we'll see several new types of problems in this chapter, our key strategy for solving them is
not new:

Concept: When working with an unusual expression in a problem, try to find ways
Q ue) to convert the problem into one involving expressions you know how to
handle.

16.1 Introduction to Piecewise Defined Functions

We,ve described functions as machines that take an input and produce an output according to some
rule. We could also imagine a machine that applies different rules to different inputs in order to produce
an output. We call such a function a piecewise defined function, and to describe such a function, we
must both give the rules used to produce outputs, and describe when to apply each rule. Here is an
example:
2x if x < 0,
k
3x if x > 0.
This notation describes a function that doubles negative inputs and triples positive inputs. Notice that
∕(x) is not defined for x = 0, so 0 is not in the domain of f.

508
 16.1. INTRODUCTION TO PIECEWISE DEFINED FUNCTIONS

Problems

Problem 16.1: Let g be the function defined for all real numbers such that

if x < 2,
if x ≥ 2.

(a) Compute g(0), g(l), g(2), g(3), and g(4).

(b) Graph g(x).
(c) Is g an even function, an odd function, or neither?

Problem 16.2: Let

x + 5 if -9 ≤ x ≤ -3,
2x2-1 if -3 < x ≤ 2,
-x∕2 if x > 2.
(a) What are the domain and range of /?
(b) For what values of x is ∕(x) = x?

Problem 16.3: Loosely speaking, we say that a function is continuous if we can draw the graph of
the function between any two points without lifting our pencil from the paper. If a function is not
continuous, we say it is discontinuous. Determine which of the following functions are continuous
and which are discontinuous.
x+5 if -9 ≤ x ≤ -3,
(a) 2x2-1 if -3 < x ≤ 2,
-x∕2 ifx>2.
if x < 2,
(b) ifx≥2.
if x < 2,
(c) if x > 2.

Problem 16.4: Each week, Alex and Barbara get paid based on how many hours each works. Alex
receives $30 per hour for each of his first 40 hours. He receives $50 per hour for each hour after his
first 40 hours. (He receives the respective fraction of these amounts if he works some fraction of an
hour.) Barbara receives $22 per hour for each of her first 20 hours, then $40 per hour for each of her
next 10 hours, then $60 for each hour after that.
Let the function A(x) represent how much Alex gets paid if he works x hours, and let B(x) represent
how much Barbara gets paid if she works x hours. Write A(x) and B(x) as piecewise defined functions
in order to answer the following questions.
(a) How many hours must Alex work to earn $800?
(b) How many hours must Barbara work to earn $800?
(c) For what positive number k will they both get the same pay for working k hours each?

509
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Problem 16.5: Determine whether each of the following functi<3ns has an inverse. If a function has an
inverse, find the inverse.
2x if x < 2,
<∙> ,w={54⅛ <b> -rt-rt=, 2x if x ≥ 2.

Problem 16.6: The function f is defined for all integers and satisfies

( n-3 if n ≥ 1000,
∕(∕(n + 5)) ifn< 1000.

Find ∕(84). (Source: AIME)

Problem 16.1: Let g be the function defined for all real numbers such that

if x < 2,
if x ≥ 2.

(a) Compute g(0), g(l), g(2), g(3), and g(4).

(b) Graph g(x).
(c) Is g(x) even or odd?

Solution for Problem 16.1:

(a) Since g(x) = 2x for x < 2, we have g(0) = 2 ∙ 0 = 0 and g(l) = 2 ∙ 1 = 2. Since g(x) = 2x for all x ≥ 2,
we have g(2) = 22 = 4, g(3) = 23 = 8, and g(4) = 24 = 16.
(b) For x < 2, we have g(x) = 2x. So, for all x < 2, the graph of g is the same as
the graph of y = 2x. Similarly, for x ≥ 2, we have g(x) = 2x, so for all x ≥ 2, the
graph of g is the same as the graph of y = 2x. So, we graph y = 2x for x < 2
and y = 2x for x ≥ 2, and we produce the graph at right.
(c) Our graph at right is certainly not that of an even function, because it is not
symmetric about the y-axis. We can also show that g is not even by noting
that g(-l) = -2 and g(l) = 2, so g(-l) ≠ g(l).
To see that g(x) is not odd, we note that g(-3) = -6 and g(3) = 8, so
g(-3) ≠ -g(3).
□
Problem 16.2: Let
x+5 if -9 ≤ x ≤ -3,
2x2-1 if —3 < x < 2,
-X/2 if x > 2.
(a) What are the domain and range of /?
(b) For what values of x is ∕(x) = x?

510
 16.1. INTRODUCTION TO PIECEWISE DEFINED FUNCTIONS

Solution for Problem 16.2:

(a) First, we note that f is defined for all x greater than or equal to —9,
so its domain is [-9, +∞). To determine the range, we can find the
range of each piece. As an aid in doing so, we graph y = ∕(x) at
right.
In our graph, the open circles mark points that our graph ap­
proaches but never quite reaches, and the solid circles mark points
that are on the graph. We can clearly see the three pieces of our
function.
From the right piece of the graph, we see that f appears to
continue forever downward. Since ∕(x) = -χ∕2 for x > 2, this
piece has a range of (-∞, -1). The lowest point on the middle piece is (0, -1), so -1 is in the range
of f. The middle piece extends up to, but does not include, the boundary point on its left end. The
y-coordinate of this point is 2(-3)2 - 1 = 17. So, the range of the middle piece is [-1,17). Make
sure you see why 17 itself is not in the range. The range of the leftmost piece of f is [—4,2]. We
combine the ranges of all three pieces to find the range of f, so the range of f is (-∞, 17).

(b) We wish to find the solutions to the equation ∕(x) = x. First, we check for solutions where x ≤ —3.
These must satisfy the equation x + 5 - x because ∕(x) = x + 5 if x ≤ -3. There are no solutions to
the equation x + 5 = x, so there are no values of x such that x ≤ -3 and ∕(x) = x.
Next, we check for solutions between —3 and 2. These must satisfy 2x2 — 1 = x because
∕(x) = 2x2 — 1 when —3 < x ≤ 2. Rearranging 2x2 — 1 = x gives 2x2 — x — 1 = 0 and factoring gives
(2x + l)(x-1) = 0. The solutions to this equation are x = 1 and x = -1/2, both of which are between
-3 and 2.
Finally, we look for solutions that are greater than 2. These must satisfy -x∕2 = x. The only
solution to this equation is x = 0, but 0 is not greater than 2, so there are no values of x such that
x > 2 and ∕(x) = x.
Therefore, the only values of x for which ∕(x) = x are x = 1 and x = -1/2.

□
Loosely speaking, we say that a function is continuous if we can draw the graph of the function
between any two points without lifting our pencil from the paper. If a function is not continuous, we
say it discontinuous.

511
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Solution for Problem 16.3:

(a) This is the function from Problem 16.2. We can see from our graph that f is not continuous because
there are two "breaks" in the graph. We therefore see that f is discontinuous.
(b) This is the function from Problem 16.1. We reproduce its graph at left below. There are no "jumps"
in either of the pieces of the graph of g. In other words, both 2x and 2x are continuous functions.
So, we know that g(x) is continuous for x <2, and for x > 2. In order to determine whether or not
g(x) is continuous at x = 2, we must check if the pieces meet when x = 2. Since 2x and 2x both
equal 4 when x = 2, the two pieces do indeed meet when x -2. So, to graph g, we graph y -2x
up to the point (2,4), and then graph y = 2x from (2,4) on. There is no jump in this graph, so the
function is continuous.
y y
r i I i J k

~ψ--~t~ τ
-4—

— -÷ η-√

∕
∕
Figure 16.1: Graph of y = g(x) Figure 16.2: Graph of y = h(x)

The only difference between this function and the prior one is that this function is not defined at
x - 2. So, to graph ⅛, we graph y = 2x to the left of x = 2 and y = 2x to the right of x = 2. However,
at x = 2 there is a hole in the graph because h(x) is not defined for x = 2. We indicate this hole in
the graph with an open circle, as shown in the graph at right above. Since we would have to lift
our pencil at x = 2 to graph y = h(x), the function h is discontinuous.

You may be a little uncomfortable with our definition of continuity. When you study calculus, you'll
learn a rigorous definition of continuity. However, our intuitive description of continuity based on
graphing is usually enough to determine if a function is continuous or not.

Problem 16.4: Each week, Alex and Barbara get paid based on how many hours each works. Alex
receives $30 per hour for each of his first 40 hours. He receives $50 per hour for each hour after his
first 40 hours. (He receives the respective fraction of these amounts if he works some fraction of an
hour.) Barbara receives $22 per hour for each of her first 20 hours, then $40 per hour for each of her
next 10 hours, then $60 for each hour after that.
(a)How many hours must Alex work to earn $800?
(b) How many hours must Barbara work to earn $800?
(c) For what positive number k will they both get the same pay for working k hours each?

Solution for Problem 16.4: Alex's and Barbara's salaries are defined in pieces—different definitions for

512
 16.1. INTRODUCTION TO PIECEWISE DEFINED FUNCTIONS

different amounts of hours. Piecewise defined functions help us organize piecewise defined data. We
let A(x) be Alex's salary in dollars if he works x hours. If Alex works no more them 40 hours, he earns
$30/hr, so A(x) = 30x if 0 ≤ x ≤ 40. If he works more than 40 hours, he gets ($30)(40) = $1200 for the first
40 hours, plus $50 per hour for the remaining x - 40 hours. So, his salary is 1200 + 50(x - 40) = 50x - 800
dollars if x > 40. Putting these pieces together gives

30x if 0 ≤ x < 40,

50x - 800 if x > 40.

Similarly, we let B(x) be Barbara's salary in dollars if she works x hours. We consider three cases to
determine B(x):

• 0 ≤ x ≤ 20. Since Barbara earns $22/hr for her first 20 hours, we have B(x) = 22x dollars if
0 ≤ x ≤ 20.

• 20 < x ≤ 30. She earns 22(20) = 440 dollars for the first 20 hours, then she earns $40/hr for the
rest of the time. So, she earns 440 + 40(x — 20) dollars. Simplifying this expression, we find that
B(x) = 40x - 360 dollars if 20 < x < 30.

• x > 30. She earns 440 dollars for the first 20 hours, then 400 dollars for the next 10 hours, then
60(x - 30) dollars for the time past 30 hours. This gives her a total pay of 440 + 400 + 60x - 1800 =
60x — 960 dollars.

Putting the information from all three cases together, we have

22x if 0 ≤ x ≤ 20,
40x - 360 if 20 < x ≤ 30,
60x - 960 if x > 30.

We can test both A(x) and B(x) by trying a specific value of x. If Alex works 60 hours, he gets
($30)(40) = $1200 for the first 40 hours and ($50)(20) = $1000 for the last 20 hours, for a total of
$2200. Indeed, we find that A(60) = 2200. Similarly, if Barbara works 60 hours, she should make
($22)(20) + ($40)(10) + ($60)(30) = $2640, and we find B(60) = 2640.

(a) Alex makes $1200 if he works 40 hours, so the amount he must work to make $800 is less than 40
hours. So, we must have 30x = 800, which gives us x = 800/30 = 26∣ hours that Alex must work
to make $800.
(b) Barbara makes $440 if she works 20 hours, so she must work more than 20 hours to make $800.
She makes 40(30) - 360 = 840 dollars if she works 30 hours, so she must work between 20 and 30
hours to make $800. Therefore, we must have 40x - 360 = 800, which gives us x = 29 hours.
(c) We seek the positive value of k such that A(k) = B(k). Since these two functions are defined in
pieces, we must check for equality by pieces. We won't have equality if 0 < k < 20, since then we
must have 22k = 30k, which only gives k = 0 as a solution.
If 20 < k ≤ 30, then A(k) = 30k and B(k) = 40k - 360, which gives us 30k = 40k - 360. Solving
this gives us k = 36, which does not fit the 20 < k ≤ 30 restriction of this case.

513
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Next, we consider 30 < k ≤ 40. Then, we have A(Jc) = 30k and B(k) = 60k — 960. Setting these
equal gives us k = 32, so they make the same amount if they both work 32 hours.
Finally, we check k > 40, to see if there's a second possible solution. If k > 40, we have
A(k) = 50k - 800 and B(k) = 60k - 960. Setting these equal gives us k = 16, which isn't greater than
40. So, the only way they can earn the same pay for the same nonzero time is if they both work 32
hours.

Problem 16.5: Determine whether each of the following functions has an inverse. If a function has I
an inverse, find the inverse.
if x < 5, if x < 2,
(a) (b) if x ≥ 2.
if x > 5.

Solution for Problem 16.5:

(a) What's wrong with the following reasoning:

Bogus Solution: We consider each piece separately. Let g(x) = 4x. Then, we have
g-1(x) = χ∕4 because then we ha veg-1 (g((x)) = (4x)∕4 = x. Similarly,
we let k(x) = 5∕x, and we find that h-1(x) = 5∕x, because then we
have k^1(k(x)) = 5∕(5∕x) = x. Therefore, we have

if x < 5,
if x ≥ 5.

We see that our answer is wrong by checking a few specific values. We have t(2) = 8, but
Γ1(8) = 5/8, so we don't have t^1(t(2)) = 2. Something went wrong.
By experimenting with a few more values, or by graphing f, we find that t does not have an
inverse. For example, because t(l∕4) = f(5) = 1, we know that t does not have an inverse. So, even
though each piece of t has an inverse, t itself does not have an inverse.
(b) We've already graphed f, so we start by looking at the graph of f to see if y
f has an inverse. No horizontal line intersects f at more than one point, so
there are no two inputs to f that give the same output. This means that f
does have an inverse.
We might be tempted to think that we simply take the inverse of each piece
of f to find that its inverse is

if x < 2,
if x ≥ 2.

Inverting each piece failed in part (a), so we take a closer look here. Again,
we try a few values. We find that k(3) = log2 3, but because log2 3 < 2, we have ∕(log2 3) = 2 log2 3.
So, we don't have ∕(∕z(3)) = 3. Uh-oh. Back to the drawing board.
We correctly determined the inverses of both pieces; the inverse of 2x is x∕2, and the inverse
of 2x is log2 x. We erred in making the domains of the two pieces in our inverse the same as the

514
 16.1. INTRODUCTION TO PIECEWISE DEFINED FUNCTIONS

domains of the pieces of f. To see why this is a mistake, let's think about ∕-1(3). The value of
f 1(3) is the value of x for which ∕(x) = 3. Examining both pieces of f, we see that the output of
the piece with x ≥ 2 is always at least 4, and that the output of the piece with x < 2 is 3 for x = 1.5.
Since we have ∕(1.5) = 3, we must have ∕-1(3) = 1.5. Aha! In determining the domains of the
pieces of ∕^1(x), we must consider the ranges of each piece of ∕(x).
If x < 2, then we have ∕(x) = 2x, so ∕(x) < 4. In other words, the left piece of ∕(x) takes numbers
in the interval (—∞, 2) as input and outputs numbers in the interval (—∞,4). The inverse of this
piece must "undo" this process, taking numbers in the interval (-∞,4) as input and returning
numbers in the interval (-∞,2) as output. Since the inverse of 2x is x∕2, we therefore have
∕^1(x) = x/2 for x < 4.
Similarly, if x ≥ 2, then we have ∕(x) = 2x, which means ∕(x) ≥ 4. This piece of ∕(x) takes the
numbers in the interval [2, +∞) as input and outputs numbers in the interval [4, +oo). The inverse
of 2x is log2 x, so we have ∕-1(x) = log2 x for x ≥ 4.
We therefore have

□
Problem 16.6: The function f is defined for all integers and satisfies ⅛

n-3 if n ≥ 1000,
/(«) = < f(f(n + 5)) if n < 1000.

Find ∕(84). (Source: AIME)

Solution for Problem 16.6: For n ≥ 1000, the behavior of the function is easy to understand. To compute
∕(84), we might first note that ∕(84) = ∕(∕(89)) from our function definition. But what's f (89)? We can
use our function definition to find that ∕(89) = ∕(∕(94)), so ∕(84) = ∕(∕(89)) = ∕(∕(∕(94))). Uh-oh. It's
going to take a long time to reach a number that's greater than 1000. Let's try a different strategy.
It's easy to find f(n) if n ≥ 1000. To see what happens when n < 1000, let's take a look at values of n
that are just below 1000:
/(999) = ∕(∕(1004)) = ∕(1001) = 998,
/(998) = ∕(∕(1003)) = ∕(1000) = 997,
/(997) = ∕(∕(1002)) = ∕(999) = 998,
∕(996) = ∕(∕(1001)) = ∕(998) = 997,
/(995) = ∕(∕(1000)) = ∕(997) = 998,
and so on. We appear to have a pattern. It looks like f(ri) = 997 if n is an even number less than 1000,
and ∕(n) = 998 if n is an odd number less than 1000.

We've already seen that this pattern holds for all integers n such that 995 ≤ n < 999. Now, we want
to show that it holds for all n < 1000. Intuitively, it seems pretty clear that this pattern will hold. It looks
like we'll always have

/(odd number) = /(/(odd number + 5)) = /(/(even number)) = ∕(997) = 998,

/(even number) = /(/(even number + 5)) = /(/(odd number)) = f(998) = 997.

515
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

To prove this formally, we use induction. We assume that ∕(∕ς) = 997 if k is even and ∕(∕c) = 998 if k is
odd, for some specific k less than 1000. To "induct downward" to prove that ∕(84) = 997, we must use
this assumption to show that f(k - 1) = 997 if k — 1 is even and f(k — 1) = 998 if k — 1 is odd.
So, what is f(k - 1)? Since k-1 < 1000, we have f(k - 1) = f(f(k + 4)). But we don't know anything
about f(k + 4). What will we do?
We have to go back and adjust our induction slightly. We cannot relate f(k - 1) to f(k); we can only
relate it to f(k + 4). In our exploration above, we found several values of k for which f(k) = 997 if k is
even and f(k) = 998 if k is odd. So, we adjust our inductive assumption so that it includes f(k + 4) in
addition to f(k).
Specifically, we assume that f(n) = 997 if n is even and f(ri) = 998 if n is odd, for all n such that
k ≤ n < 1000. So, if k < 996, our assumption tells us that f(ri) = 997 if n is even and f(ri) = 998 if n is
odd for both n = k and n = k + 4, instead of only telling us this for n = k. We've already shown that the
assumption is true for all n from 995 to 999, so we can "start" our induction with k = 995.
-We're now ready to return to f(k - 1), for k < 995. Above, we found that f(k — T) = f(f(k + 4)).
Moreover, since k < 995, we have k + 4 ≤ 999, so k < k + 4 < 1000. Because k + 4 is between k and 1000,
our inductive assumption tells us that f(k + 4) = 997 if k + 4 is even and f(k + 4) = 998 if k + 4 is odd. If
k — 1 is even, then k + 4 is odd, so f(k - 1) = f(f(k + 4)) = ∕(998) = 997. If k — 1 is odd, then k + 4 is even,
and we have f(k — 1) = f(f(k + 4)) = ∕(997) = 998. Combining these, we have f(k — 1) = 997 if k - 1 is
even and f(k — 1) = 998 if k — 1 is odd.
This completes our inductive step, so we have shown that f(ri) = 998 if n is an odd integer less than
1000, and f(n) = 997 if n is an even integer less than 1000. So, we have ∕(84) = 997.
Make sure you see why it is important that we showed that the assumption is true for all n from 995
to 999 as the base case of the induction. This is because our inductive step uses the inductive assumption
for n = k + 4 to prove that the statement also holds for n = k -1. In other words, we are taking steps of 5
in our induction, so we need 5 base cases. If we only showed that the inductive assumption is true for
n = 999 as the base case, then our inductive step can only be used to show that the statement is true for
n = 994, 989, 984,979, etc. □
Our solution to Problem 16.6 is a great example of a very common problem solving process: experi­
ment, see a pattern, and prove the pattern with induction.

Concept: Don't just stare at a problem when you're stuck. Experiment and look for
patterns.

Exercise
*

16.1.1 Let f be the function defined for all real numbers such that

λ ∕ if x < -2,
2x, if -2 ≤ x < 0,
3x, if 0 ≤ x < 2,
x2 + 3, ifx ≥ 2.

(a) Find ∕(3∕2). (b) Graph/. Is∕continuous? (c) Find the range of ∕(x).

516
 16.2. ABSOLUTE VALUE

16.1.2 What real value of c will make the following function continuous?

»={ x2 + 2c
2cx + 1
if x ≤ 3,
if x > 3.

Graph f for this choice of c.

16.1.3 √2≡x if x < 0,

Find the inverse of f if ∕(χ) = <
l-x2 if x ≥ 0.
16.1.4 Mientka Publishing Company prices its bestseller Where's Walter? as follows:

if 1 ≤ n < 24,
if 25 ≤ n < 48,
if 49 ≤ n,
where n is the number of books ordered and C(π) is the cost in dollars of n books. Notice that 25 books
cost less than 24 books. For how many values of n is it cheaper to buy more than n books than to buy
exactly n books? (Source: AHSME)

16.2 Absolute Value

We can think of the absolute value of a number as the distance between that number and 0 on the
number line. However, we can also view absolute value as a piecewise defined function.

Definition: The absolute value of the real number x, denoted as ∣x∣, is defined as

if x ≥ 0,
if x < 0.

For example, we have ∣4∣ = 4, ∣-2∣ = -(-2) = 2, and ∣0∣ = 0. Furthermore, because ∣x∣ is positive if x is
positive or negative, and ∣0∣ = 0, we have ∣x∣ ≥ 0 for all x, where ∣x∣ = 0 if and only if x = 0.

Problems

Problem 16.7: Write each of the following as a piecewise defined function:

(a) ∕(x) = ∣3 - x∣
(b) g(x) = ∣x + 2∣ + ∣x + 5∣

Problem 16.8:
(a) Solve the equation ∣3 - x∣ = 5.
(b) Solve the equation ∣x + 2∣ + ∣x + 5∣ = 8.
(c) Find all solutions to ∣x - 1∣ = ∣x - 2∣ + ∣x - 3∣. (Source: HMMT)

Problem 16.9: Find all values of x such that ∣x + 3∣ + ∣x + 4∣ < 7.

517
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Problem 16.10:
(a) If ∣fl∣ + ∣b∣ = 0, then what can be said about a and b?
(b) Find all constants c such that ∣cx - 1∣ + ∣χ2 - x - 2∣ = 0 has a solution in x.

Problem 16.11: How many ordered triples of integers (a,b,c) satisfy ∖a + b∖ + c - 19 and ab + ∣c∣ = 97?
(Source: AHSME)

Many absolute value problems are solved by thinking of the absolute value expression as a piecewise
defined function, so we'll start by getting a little practice writing absolute value expressions as piecewise
defined functions.
Problem 16.7: Write each of the following as a piecewise defined function:
(a) ∕(x) = ∣3 - x∣
(b) g(x) = ∣x + 2∣ + ∣x + 5∣
∣.................................... ..................

Solution for Problem 16.7:

(a) If 3 — x ≥ 0, then ∣3 - x∣ = 3 - x. Rearranging 3 - x ≥ 0 gives us x ≤ 3, so we have ∣3 — x∣ = 3 — x if

x ≤ 3. Similarly, if 3 - x < 0, then ∣3 - x∣ = -(3 - x) = x - 3. So, we have ∣3 - x∣ = x - 3 if x > 3.
Therefore, we can write f as
if x > 3,
if x ≤ 3.

(b) In order to write ∣x + 2∣ and ∣x + 5∣ without the absolute value notation, we consider three cases.

• Both x + 2 and x + 5 are nonnegative. If x ≥ -2, then ∣x + 2∣ = x + 2 and ∣x + 5∣ = x + 5, so

∣x + 2∣ + ∣x + 5∣ = 2x + 7.
• x + 2is negative and x + 5 is nonnegative. If -5 ≤ x < -2, then ∣x + 2∣ = -(x + 2) = -x - 2 and
∣x + 5∣ = x + 5, so ∣x + 2∣ + ∣x + 5∣ = 3. (Note that we cannot have x + 5 negative and x + 2
nonnegative.)
• Both x + 2andx + 5are negative. If x < -5, then ∣x + 2∣ = -(x + 2) = -x -2 and ∣x + 5∣ = -x - 5,
so ∣x + 2∣ + ∣x + 5∣ = -2x - 7.

Putting these three cases together gives

2x + 7 if x ≥ -2,
∕ω = 3 if -5 ≤ x < —2,
-2x-7 if x < -5.

Let's see how we can use this technique to solve equations:

Extra! You cannot depend on your eyes when your imagination is out offocus.
i∣∣⅜l∣∣∣φ l∣∣
* ∣∣⅜ - Mark Twain

518
 16.2. ABSOLUTE VALUE

Problem 16.8:
(a) Solve the equation ∣3 — x∣ = 5.
(b) Solve the equation ∣x + 2∣ + ∣x + 5∣ = 8.
(c) Find all solutions to ∣x - 1∣ = ∣x - 2∣ + ∣x - 3∣. (Source: HMMT)

Solution for Problem 16.8:

(a) For simple equations like this one, we have several straightforward approaches:
Solution 1: Use the piecewise defined function. If we let ∕(x) = ∣3 - x∣, then our problem is to solve
the equation ∕(x) = 5. In part (a) of Problem 16.7, we saw that ∕(x) = ∣3 - x∣ can be written as

x - 3 if x > 3,
3 - x if x ≤ 3.
So, we test each piece to see what values of x give ∕(x) = 5. If x - 3 = 5, we have x = 8, which
satisfies x > 3 (the restriction for the first piece). If 3 — x = 5, then x — —2, which satisfies x ≤ 3 (the
restriction for the second piece). Checking our answers, we see that x = -2 and x = 8 both satisfy
∣3 - x∣ = 5.
Solution 2: Use the number line. The expression ∣3 - x∣ equals the distance between 3 and x on
the number line. So, if ∣3 — x∣ = 5, then x is 5 away from 3 on the number line, which means either
x = -2(5 to the left of 3) or x = 8 (5 to the right of 3).
Solution 3: What numbers have absolute value 5? The only two numbers with absolute value 5 are
-5 and 5. So, either we must have 3 - x = -5, which gives x = 8, or we must have 3 - x = 5, which
gives x = -2.
Solution 4: Square the absolute value signs away. Because the square of a real number is always
nonnegative, we have (∣x∣)2 = ∣x2∣ = x2 for all real x. So, we can square both sides of our equation
to get rid of the absolute value signs. This gives (∣3 - x∣)2 = 52, so (3 - x)2 = 25. Expanding the
square and rearranging gives us x2 - 6x - 16 = 0, which has solutions x = 8 and x = -2. Because
we squared the equation as a step in our solution, we must check for extraneous solutions. We
find that both solutions are acceptable.
All four solutions are valid, and all four have the same key step: finding a way to get rid of the
absolute value signs and turn the problem into one we know how to solve.
(b) Of our four solutions to part (a), you might find the first one the most complicated. But as absolute
value problems get harder, some of our simpler methods start to fail us. For example, in solving
∣x + 2∣ + ∣x + 5∣ = 8, we won't get far with our third or fourth approaches to part (a). Let's give our
piecewise defined function approach a shot. We saw in part (b) of Problem 16.7 that

2x + 7 if x ≥ -2,
∣x + 2∣ + ∣x + 5∣ — < 3 if -5 < x < -2,
-2x-7 if x < -5.

Clearly, the middle piece cannot ever equal 8. But each of the other two pieces can. For the first
piece, we have 2x + 7 = 8, so x = 1/2, which satisfies x > —2, as required. For the third piece, we
have -2x - 7 = 8, so x = -15/2, which satisfies x < -5. Therefore, our two solutions are x = -15/2
and x = 1/2. (As an Exercise, you'll have a chance to use the number line to solve this problem.)

519
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

(c) We aren't too sure how to deal with an equation that has absolute values on both sides, so we move
all the absolute value signs to the left, giving ∣x — 1∣ - ∣x - 2∣ — ∣x - 3∣ = 0. Now, our previous two
parts give us a guide. We let the left side be ∕(x) and write ∕(x) as a piecewise defined function.
We consider the cases x<l, l≤x<2, 2≤x<3, and x ≥ 3, and we have

x-4 if x < 1,
3x - 6 if 1 ≤ x < 2,
∕(x) = ∣x-l∣-∣x-2∣-∣x-3∣ = if 2 ≤ x < 3,
x-2
∖
-x + 4 if x ≥ 3.

As before, we check each piece for solutions. Setting our first piece equal to 0 gives us x = 4,
which doesn't satisfy x < 1, so we have no solutions for the first piece. The second piece equals 0
for x = 2, which doesn't satisfy the restriction 1 ≤ x < 2. So, there are no solutions for the second
piece. The third piece gives us x = 2, which satisfies 2 ≤ x < 3, and the fourth piece gives us X = 4,
which satisfies x ≥ 3. Therefore, the only solutions to our equation are x = 2 and x = 4.

Concept: An effective technique for solving some equations with absolute values is to
o≡≡^ break the problem into cases by viewing the absolute values as piecewise
defined functions. Remember to check each solution. In each case, the
solution must satisfy the restrictions of that case in order to be a valid
solution.

Casework helped us with absolute value equations. Let's try an absolute value inequality.

Problem 16.9: Find all values of x such that ∣x + 3∣ + ∣x + 4∣ < 7.

Solution for Problem 16.9: Solution 1: Casework. When x ≥ -3, we have ∣x + 3∣ + ∣x + 4∣ = 2x + 7, so our
inequality is 2x + 7 < 7 if x ≥ -3. Solving 2x + 7 < 7 gives x < 0. Combining x < 0 with the restriction
x ≥ -3 of this case gives us -3 ≤ x < 0.
When-4 ≤ x < -3, we have ∣x + 3∣ + ∣x + 4∣ = -(x + 3) + x + 4 = 1. Sincel < 7, wehave∣x + 3∣ + ∣x + 4∣ < 7
for all x such that -4 ≤ x < -3.
When x < -4, we have ∣x + 3∣ + ∣x + 4∣ = -(x + 3) - (x + 4) = -2x - 7, so our inequality is -2x - 7 < 7
if x < -4. Solving -2x - 7 < 7 gives x > -7. Combining x > -7 with the restriction x < -4 gives us
-7 < x < -4.
Putting these three cases together, the values of x that satisfy the original inequality are -3 < x < 0,
-4 ≤ x < -3, and -7 < x < -4. We can combine all three of these ranges for x; our original inequality
holds when -7 < x < 0.
Solution 2: Use the number line. The quantity ∣x + 3∣ = ∣x - (-3)∣ is the distance between x and -3 on the
number line. Likewise, the quantity ∣x+4∣ is the distance between x and—4. Therefore, if ∣x+3∣ + ∣x+4∣ < 7,
then the sum of the distances from x to -3 and to -4 must be no more than 7. Clearly this is the case if
-4 ≤ x ≤ —3. For x < —4, we consider the following number line:

520
 16.2. ABSOLUTE VALUE

c+1

Since —3 and -4 are 1 apart, if x is c to the left of —4 as shown above, then the sum of distances from x to
—3 and to —4 is 2c + 1. We must then have 2c + 1 < 7, so c < 3, which means x > —7. Similarly, if x is to
the right of -3, it must be within 3 of -3, so x < 0. This gives us -7 < x < 0 as our solution. □
Note that in the second solution to Problem 16.9, we avoided the need for any algebra by interpreting
the absolute value function as a measure of distance.

Concept: Viewing the absolute value of the difference between two real numbers as
O mJ the distance between the numbers on the number line can help solve some
absolute value problems.

Problem 16.10: Find all constants c such that the following equation has a solution in x:

∣cx - 1∣ + ∣x2 - x - 2∣ = 0.

Solution for Problem 16.10: We have ∣cx - 1∣ ≥ 0 and ∣x2 - x - 2∣ ≥ 0 for all x, so ∖cx - 1∣ + ∣x2 - x - 2∣ = 0 if
and only if ∣cx - 1∣ = ∣x2 — x - 2∣ = 0. So, we must have ex - 1 = x2 - x - 2 = 0.
The quadratic x2 — x — 2 factors as (x — 2)(x +1), so x = 2 or x = —1. If x = 2 is a solution of the original
equation, then ∣cx — 1∣ = 0 gives us ∖2c —1∣ = 0, so c — 1/2. Similarly, if x = —1 is a solution, then we must
have |— c — 1∣ = 0, so c = -1. Therefore, the only possible values for c are 1/2 and — 1. □
Don't forget that the absolute value of any expression must be nonnegative. This fact alone will help
simplify many problems.

Problem 16.11: How many ordered triples of integers (fl, b, c) satisfy ∣a + b∖ + c = 19 and ab + ∣c∣ = 97?
'WBaMHaBaBBHBMHMaa∣Haaι∣MaaaaaBMH∣aaMBHaMBHBMaMMB ∣MM∣aaHMal

Solution for Problem 16.11: At first glance, it looks like we can easily eliminate c by subtracting the first
equation from the second. However, the pesky absolute value signs get in the way. But we know how
to take care of them! WeTl consider two cases:

Case 1: c≥0. If c ≥ 0, then ∣c∣ = c, so our system of equations is ∖a + b∖ + c = 19, ab + c = 97. Subtracting
the first equation from the second gives us ab - ∖a + b∣ = 78. Once again, if it weren't for the absolute
value signs, we'd know what to do with this equation: we'd use Simon's Favorite Factoring Trick. So,
we get rid of the absolute value signs by considering two subcases:

• Subcase la: a + b ≥ 0, c ≥ 0. If a + b ≥ 0, then ∖a + b∣ = a + b, so we have ab - a - b = 78. Adding

1 to both sides and then factoring the left side gives (a - l)(b - 1) = 79. Since 79 is prime and
a + b ≥ 0, we must have a - 1 = 79, b - 1 = 1 or a - 1 = 1, b - 1 = 79. So, we have (a, b) = (80,2) or
(a, b) = (2,80). In either case, we have a + b = 82, so we have c = 19 - ∖a + b∖ = -63. But we must
have c ≥ 0, so there are no solutions for this subcase.

521
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

• Subcase lb: a + b < 0, c ≥ 0. If a + b < 0, then ∖a + b∖ = -(a + b), so we have ab + a + b = 78. Adding
1 to both sides and then factoring the left side gives (a + l)(b + 1) = 79. Since a + b < 0, we have
(αz b) = (—80, -2) or (a, b) = (-2, -80). In either case, we again have c = 19-∖a + b∖ = -63, and since
we must have c ≥ 0, we have no solutions for this subcase, either.

Case 2: c < 0. If c < 0, we have ∣c∣ = —c, so our system of equations is ∖a + b∖ + c - 19, ab - c - 97.
Adding these gives ab + ∖a + b∖ = 116. Once again, we consider two subcases:

• Subcase 2a: a + b ≥ 0, c < 0. If a + b ≥ 0, then ∖a + b∖ = a + b, so we have ab + a + b = 116, so

(a + l)(b + 1) = 117. Since 117 = 1 ∙ 117 = 3 ∙ 39 = 9 ∙ 13, we have three pairs of numbers whose
product is 117. Each gives us two solutions (a,b'). For example, from 117 = 1 ∙ 117, we have
(λ, b) = (0,116) and (a,b) = (116,0). For both these solutions, we have c = 19 — ∖a + b∖ = —97,
which satisfies the restrictions of this subcase. Similarly, 117 = 3 ∙ 39 gives us the solutions
(a,b,c) = (2,38,-21) and (a,b,c) = (38,2,-21), and 117 = 9 ∙ 13 gives us (a,b,c) = (8,12,-1) and
(a, b, c) = (12,8, -1). So, there are six solutions in this subcase.

• Subcase2b: a+b < 0,c < 0. In this case, we have ab-a-b = 116, so (α-l)(b-l) = 117. Proceeding as in
the previous subcase, we get the six solutions (a, b, c) = (-116,0, —97), (0, -116, —97), (-2, —38, —21),
(-38, -2, -21), (-12, -8, -1), and (-8, -12, -1).

Phew! Putting all our cases together, we have a total of 12 solutions to the original system of
equations. □

Exercises

16.2.1 Let x be a real number no greater than -4.

(a) Simplify ∣x + 2∣.

(b) Simplify ∣2 - ∣x + 2∣∣.
(c) Is ∣π2 - 10∣ equal to π2 - 10 or 10 - π27
(d) Simplify ∣x + π2 - 6∣.

16.2.2 Use the number line to solve the equation ∣x + 2∣ + ∣x + 5∣ = 8.

16.2.3 Find all x such that 1 ≤∖x-2∖<7. (Source: AHSME)

16.2.4 Find all solutions to the equation ∣x - ∣2x + 1∣∣ = 3.
16.2.5 If ∣x∣ + x + y = 10 and x + ∖y∖-y = 12, find x + y. (Source: AHSME)
ly - lyll
16.2.6 Find all real values of y for which--------- > 0.
y
*
16.2.7 If ∕(x) = (x - 3)2 - 1, compute the set of real numbers such that ∕(∣x∣) = ∣∕(x)∣. (Source: ARML)
Hints: 251

522
 16.3. GRAPHING ABSOLUTE VALUE

16.3 Graphing Absolute Value

In the previous section, we explored several algebraic approaches for handling problems involving
absolute value. In this section, we take a more geometric approach to absolute value.

Problems

Problem 16.12: Graph the function ∕(x) = χ + ∣χ∣.

Problem 16.13: Let ∕(x) = ∣x - 2∣ + ∣x - 4∣ + ∣2x - 6∣, for 2 ≤ x ≤ 8.

(a) Rewrite ∕(x) as a piecewise defined function.
(b) Graph f.
(c) Sketch the largest and smallest possible values of ∕(x).

Problem 16.14:
(a) Graph y = ∣x + 1∣ - 2 and y = -∣x - 4∣ + 7.
(b) The graphs of y = -∣x — «| + b and y = ∣x - c∣ + d intersect at points (2,5) and (8,3). Find a + c.
(Source: AHSME)

Problem 16.15: Find the area of the region of points that satisfy the inequality ∣x + y∣ + ∣x - y∣ ≤ 8.

Problem 16.16:
(a) Solve the equation ∣x2 + 6x - 3∣ = x + 3.
(b) Graph the equation y = ∣x2 + 6x - 3∣ and use your graph to help you solve the inequality
∣x2 + 6x - 3∣ ≥ x + 3.
(c) Find the value of c if the line y = c intersects the graph of y = ∣x2 + 6x - 3∣ in exactly three points.

Problem 16.12: Graph the function ∕(x) = x + ∣x∣.

Solution for Problem 16.12: When x > 0, we have ∕(x) = x + ∣x∣ = x + χ = 2x.
When x < 0, we have ∕(x) = x + ∣x∣ = x- x = 0. So, for x ≥ 0, the graph of ∕(x)
is the same as the graph of y = 2x, and for x < 0, the graph of f is the same as
the graph of y = 0. Putting these together, the graph of f is shown at right. □

Concept: The absolute value function is a piecewise defined function, so it's often
Qz-∣-nJ easiest to graph functions involving absolute values piece-by-piece.

523
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Solution for Problem 16.13: We begin by rewriting ∕(x) as a piecewise defined function: y

8-2x if 2 ≤ x < 3,
∕ω = 2x-4 if 3 ≤ x < 4,
4x-12 if 4 ≤ x ≤ 8.

The graph of f consists of three line segments, as shown at right. Each segment
is non-horizontal, so we know that the highest and lowest points of the graph of each
segment are the endpoints of the segment. With this in mind, we could have skipped
actually graphing f, and evaluated ∕(x) at the x values of each of these endpoints. These
x values are the boundary values of our cases in the piecewise definition of f, so to
calculate the minimum and maximum of f, we find
∕(2) = 4, ∕(3) = 2, ∕(4) = 4, ∕(8) = 20.

So, the minimum value of f(x) is 2 and the maximum value is 20. □

Problem 16.14: The graphs of y = -∣x - a∖ + b and y = ∣x - c∣ + d intersect at points (2,5) and (8,3).
Find a + c. (Source: AMC)

Solution for Problem 16.14: We start by figuring out what the graphs of equations of the formsy = -∣x-fl∣+fe
and y = ∖x - c∖ + d look like. We do so by first graphing some example equations, by choosing specific
values for a, b, c, and d. On the left below, we graph y = ∣x+l∣-2. Forx ≥ -1, wegraphy = x+1-2 = x— 1,
and for x < -1, we graph y = -(x + 1) - 2 = -x - 3.

Figure 16.4: Graph of y = -∣x - 4∣ + 7

On the right above, we graph y = -∣x - 4∣ + 7. For x ≥ 4, the resulting graph is the graph of
xj = -(χ - 4) + 7 = -χ + 11, and for x < 4, the graph matches the graph of y = -(4 - x) + 7 = x + 3.

We see that graphs of the form y = -∣x - fl∣ + b have an inverted V shape. The top of the inverted
V occurs when x = a, since this makes -∣x - a∖ + b as large as possible. For any other value of x, the
expression -∣x - a∖ is negative, so -∣x - a∖ + b is less than b. When x ≥ a, the graph of y = -∣x - fl∣ + b
matches that of a line with slope -1, and if x < fl, then the graph matches a line with slope 1. Similarly,
graphs of the form y = ∣x + c∣ + d form an upward V shape, where the branches have slopes ±1 and the
bottommost point is (c,d).

Extra! It has long been an axiom of mine that the little things are infinitely the most important.
∣u* ∣* ∣ ⅜ ii#. - Sir Arthur Conan Doyle

524
 16.3. GRAPHING ABSOLUTE VALUE

Armed with these observations, we're ready to tackle the problem. We know y
what the graphs look like, so we start from the intersection points (2,5) and (8,3)
and construct a V and an inverted V, using rays with slopes 1 and -1 for both
branches of each V. The bottommost point of the V is (6,1), so c = 6. The highest
point of the inverted V is (4,7), so a = 4, and a + c = 10.

Notice that a + c equals the sum of the x-coordinates of the given intersection
points. Is this a coincidence? □

Problem 16.15: Find the area of the region of points that satisfy the inequality ∖x + y∣ + ∣x - y∣ ≤ 8.

Solution for Problem 16.15: We've tackled other absolute value graphs by considering cases, so we try the
same here.

Case 1: x + y ≥ 0. This makes our inequality x + y + ∣x - y∣ ≤ 8. Hmmm.... We still have an absolute

value to get rid of. Let's try subcases.

• Subcase la: x + y ≥ 0, x - y ≥ 0. Now, we have ∣x - y∣ = x - y, so our inequality is 2x ≤ 8, or x ≤ 4.

Therefore, if both x + y ≥ 0 and x — y ≥ 0, then we must have x ≤ 4.

• Subcase lb: x + y ≥ 0, x - y < 0. This gives us ∣x - y∣ = -(χ - y) = y - χ, so our inequality is 2y ≤ 8,

or y ≤ 4. Therefore, if x + y ≥ 0 and x - y < 0, then we have y ≤ 4.

But how do we graph these solutions? We start by splitting the Cartesian plane into regions that
correspond to our subcases. At left below, we have graphed x + y = 0 and x - y = 0 with dashed lines.
Region [T∣ is where x + y > 0 and x - y > 0. So, to graph our solutions in Subcase la, we shade the region
of points inside region ∣T∣ such that x ≤ 4, as shown in the middle diagram below. Notice that we also
make solid the portions of the lines x + y = 0 and x - y = 0 that border region [T~j for x ≤ 4, since we have
x≤4forx + y≥0 and x - y > 0. This includes the appropriate points on these lines in the region we're
graphing.

Figure 16.5: Our Regions Figure 16.6: Subcase la Figure 16.7: Case 1

Region [Σ is where x + y > 0 and x - y < 0, so to graph Subcase lb, we shade the region of points
inside region 1Γ∣ for which y ≤ 4 (including the appropriate portion of the line x + y = 0). Putting this
together with our graph from Subcase la gives us the graph at right above, which is the graph for the
entire Case 1.

We have a pretty good guess what we'll find for Case 2. Let's see:

525
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Case 2: x + y < 0. This gives us -x - y + ∣x - y∣ ≤ 8, and we turn again to subcases.

• Subcase 2a: x + y < 0, x - y ≥ 0. This gives us -2y ≤ 8, so y ≥ -4. This subcase corresponds to
region ∣~3~∣ in our graph, so we shade y ≥ -4 inside that region.

• Subcase 2b: x + y<0, x-y<0. This gives us -2x ≤ 8, so x > -4. This subcase corresponds to
region ∣^4~∣ in our graph, so we shade x ≥ -4 inside that region.

x and y. Therefore, ∣x + y∣ + ∣x - y∣ ≤ 8 tells us that the sum of these two distances is no more than 8. If
∣y∣ > 4, then no matter what x is, the sum of these distances is greater than 8. To see why, consider the
diagram below.

*3 ~y l∣ 1
* l y xτ2 r

-4 0 4

Here, we have y > 4, and three different possibilities for x. For x↑, which is between y and —y, the
sum of the distances from Xi to y and to -y equals y - (-y) = 2y, which is greater than 8. For X2, we
have X2 > y, so the distance between X2 and -y is greater than 8. Similarly, the distance between x3 and
y is greater than 8.
By similar arguments, if ∣y∣ ≤ 4 and ∣x∣ > 4, then the sum of the distances x - y and ∣x - (-y)∣ is 2∣x∣,
which is greater than 8. But if ∣y∣ ≤ 4 and ∣x∣ ≤ 4, then the sum of these distances is at most 8, so the
square described by -4 ≤ x ≤ 4, -4 ≤ y ≤ 4 is the solution to ∣x + y∣ + ∣x - y∣ ≤ 8. □
Don't feel bad if you didn't see the second solution right away. Often, we will work through
complicated casework or computations, only to arrive at a very simple answer. When you do so, you
shouldn't just immediately move on.

Concept: When you get a simple answer after doing a lot of casework, look for a
o==g simple explanation for the simple answer.

Finding a simple explanation will not only provide more confidence that your answer is correct, but
it may also give you greater insight into future problems.

Extra! German mathematician Karl Weierstrass is credited with introducing the notation ∣x∣
ι*∙"'" for absolute value in 1841. While mathematics is often regarded as a young person's
pursuit, Weierstrass published a result at age 70 that still bears his name (the Weierstrass
Polynomial Approximation Theorem)!

526
 16.3. GRAPHING ABSOLUTE VALUE

Problem 16.16:
(a) Solve the equation ∣x2 + 6x - 3∣ = x + 3.
(b) Solve the inequality ∣x2 + 6x - 3∣ > x + 3.
j = c intersects the graph of y = ∣x2 + 6x — 3∣ in exactly three points. I
(c) Find the value of c if the line ∖

Solution for Problem 16.16:

(a) Using casework based on where x2 + 6x - 3 is positive or negative isn't too attractive (but it can be
done), because the roots of this quadratic are not rational. Squaring to get rid of the absolute value
signs doesn't look like much fun either. However, we can apply casework in a slightly different
way. If ∣x2 + 6x - 3∣ = x + 3, then either x2 + 6x - 3 and x + 3 are equal, or they are opposites. If
x2 + 6x - 3 = x + 3, then x2 + 5x - 6 = 0, so x = -6 or x = 1. If χ2 + 6x - 3 = -(x + 3), then x2 + 7x = 0,
so x = 0 or x = -7.
Therefore, we appear to have four solutions: x = -7, -6, 0, and 1. However, if we test the
solutions, something interesting happens. When x = -7, we have ∣x2 + 6x - 3∣ = ∣49 - 42 - 3∣ = 4,
but x + 3 = -4. In other words, we don't have ∣x2 + 6x - 3∣ = x + 3 for x = -7. What went wrong?
Our approach to the solution overlooked an important fact: absolute values must be non­
negative. When we rewrote ∣x2 + 6x — 3∣ = x + 3 as the two equations x2 + 6x- 3 = x + 3 and
x2+6x-3 = —(x+3), we lost sight of this fact. Whilex = -7 is indeed a solution to x2+6x-3 = —(x+3),
it is not a solution to ∣x2 + 6x — 3∣ = x + 3. Similarly, x = -6 does not satisfy the original equation,
but both x = 0 and x = 1 do.
WARNING!! When solving an equation of the form ∣∕(x)∣ = g(x) by considering the !
**4 cases ∕(x) = g(x) and ∕(x) = -g(x), you must check that the answers
you find satisfy the original equation. Any solution for which g(x) is
negative is extraneous, since ∣∕(x)∣ cannot be negative.
So, the only solutions to ∣x2 + 6x - 3∣ = x + 3 are x = 0 and x = 1.
(b) Unfortunately, we can't simply use the same reasoning as in part (a). It is true that if x2+6x-3 ≥ x+3,
then ∣x2 + 6x - 3∣ ≥ x + 3, because ∣x2 + 6x - 3∣ ≥ x2 + 6x - 3. However, if we compare x2 + 6x - 3
to —(x + 3) as our next step, we have a problem. It is not true that if x2 + 6x - 3 ≥ -(x + 3), then
∣x2 + 6x - 3∣ ≥ x + 3. For example, if x = ∖, we have x2 + 6x - 3 = ⅛ and -(x + 3) = -3jz so we have
x2 + 6x - 3 ≥ -(x + 3). But it is not true that ∣x2 + 6x - 3∣ ≥ x + 3 when x = ⅛. Dealing with the
inequality is much trickier than dealing with the equation in (a)!
We've tried using graphs to understand complicated equations before;
let's try that here. We'll graph both sides of the equation and see how the
graphs are related. The right side is easy. The graph of y = x + 3 is just a line
through (0,3) with slope 1. Graphing y - ∣x2 + 6x - 3∣ is tougher. We start by
graphing y = x2 + 6x - 3 = (x + 3)2 -12. This is an upward-opening parabola
as shown at right.

527
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

For those values of x for which x2 + 6x - 3 is nonnegative, the graphs of

y = ∣x2 + 6x - 3∣ and y = x2 + 6x - 3 are the same. But if x2 - 6x + 3 <0, then the
y values of the corresponding points on the graphs are opposites. Therefore,
we get the graph of y = ∣x2 + 6x - 3∣ by copying the graph of y = x2 + 6x - 3
when y is nonnegative, and by reflecting the graph of y = x2 + 6x - 3 over
the x-axis when y is negative. The result is the graph at right.

Graphing y = x + 3 and y = ∣x2 + 6x - 3∣ on the same Cartesian plane helps y

us set up our solution to the inequality. We could just read the answer off
the graph, trusting that the two graphs don't intersect again somewhere off
to the right of the portion of the graph shown. However, we'd like to be sure
that there are no more intersection points, so we'll use our graph as a guide
to an algebraic solution.
First, we note that x + 3 < 0 for x < -3. Since ∣x2 + 6x - 3∣ is nonnegative
for all x, we have ∣x2 + 6x - 3∣ > x + 3 for all x < -3. Next, we must check for solutions for which
x+3 ≥ 0. If ∣x2 + 6x-3∣ > x+3 ≥ 0, then, in addition to x ≥ -3, we must have either x2 + 6x-3 ≥ x+3
or —(x2 + 6x — 3) ≥ x + 3. The former gives us (x + 6)(x — 1) ≥ 0, so x > 1 or x ≤ -6. Since we must
also have x ≥ —3, we have x ≥ 1 for this case. If -(x2 + 6x - 3) ≥ x + 3, then we have x(x + 7) < 0,
so 0 ≥ x ≥ —7. Since we must also have x > -3, we have -3 ≤ x ≤ 0.
Combining x < -3, -3 ≤ x ≤ 0, and x > 1 gives x ∈ (-∞, 0] U [1, +∞) as our solution.

(c) Solution 1: Graphing. With part (b) as inspiration, we seek a horizontal line that passes through
exactly 3 points on the graph of y = ∣x2 + 6x - 3∣. We use our graph from part (b). We see that any
horizontal line below the x-axis misses the graph of y = ∣x2 + 6x - 3∣ entirely. Any horizontal line
below the "vertex" of the middle "hump" of the graph intersects the graph in 4 points, and any
horizontal line above this vertex of the middle hump intersects the graph at 2 points.
The only horizontal line that intersects the graph in 3 points is the line through the vertex. The
vertex of the parabola formed by graphing y = x2 + 6x - 3 = (x + 3)2 - 12 is (-3, -12). When we
graph y = ∣x2 + 6x - 3∣, this point is reflected over the x-axis to give (-3,12). Therefore, the desired
horizontal line is y = 12.
Solution 2: Algebra. If the graph of y = c intersects the graph of y = ∣x2 + 6x - 3∣ at three points,
then the equation ∣x2 + 6x - 3∣ = c has three real solutions. If ∣x2 + 6x - 3∣ = c, then x2 + 6x - 3 = c
or x2 + 6x - 3 = -c. Each of these is a quadratic equation. Together, these quadratics must have
3 distinct real solutions. The only way the two quadratics can share a solution is if c = 0, in
which case the two quadratics share both solutions, and we only have 2 distinct solutions. So,
the quadratics can't share any solutions. This means that the only way they can together have 3
distinct solutions is if one of them has only 1 distinct solution.
Since the quadratic and linear terms of each quadratic equation are x2 and 6x, one of these
equations must be equivalent to (x + 3)2 = 0, which means we must have c = ±12. Clearly, c must
be positive, since ∣x2 + 6x - 3∣ = c has no solutions if c is negative. So, the desired horizontal line
is y = 12.
We can quickly check our answer by solving the equation ∣x2 + 6x - 3∣ = 12. We must have
either x2 + 6x - 3 = 12 or x2 + 6x - 3 = -12, so either x2 + 6x - 15 = 0 or x2 + 6x + 9 = 0. The former

528
 16.4. FLOOR AND CEILING

has solutions x = -3 ± 2 Vβ and the latter only has the solution x = -3. Therefore, the equation
∣x2 + 6x - 3∣ = 12 has exactly 3 solutions.

□
Concept: Sometimes we can use the graph of a function to give us intuition about
C⅛ an algebraic problem.

l I “ -------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -

Exer<i%es

16.3.1 Graph the function ∕(x) = ∣2 - 4x∣ - 4, and find the x-intercepts and the y-intercepts of the graph.
16.3.2 Find all intersection points of the graphs of ∕(x) = ∣χ - 1∣ and g(x) = ∣2 - x∣.
16.3.3 Find the total area of the region of points that satisfy both ∣χ + y∣ ≤ 8 and xy ≥ 0.
16.3.4 Find the area of the smallest region bounded by the graphs of ι√ = ∣x∣ and x2 + v2 = 4. (Source:
AHSME) j o r y

16.3.5
(a) Graph ∕(x) = ∣3 - x∣.
(b) Graph ∕(x) = ∣x + 2∣ + ∣x + 5∣.
(c) Find another solution to part (c) of Problem 16.8 using graphing.
16.3.6 Find the minimum possible value of the function ∕z(x) = ∣x + 1∣ + ∣x - 4∣ + ∣x - 6∣.
16.3.7 Find the area of the region described by ∣5x∣ + ∣6y∣ ≤ 30.

16.4 Floor and Ceiling

Definitions:

• The floor of x is defined as the greatest integer that is less than or equal to x, and is denoted
by ∣xJ∙

• Similarly, the ceiling of x is defined as the least integer that is greater than or equal to xz and
is denoted by ∣^x^∣.

• The fractional part of x is defined as x - LxJ, and is denoted by (x}.

In other words, floor "rounds down" and ceiling "rounds up." The floor function is sometimes
referred to as the greatest integer function and the ceiling function is sometimes referred to as the least
integer function. Moreover, the notation [x] is sometimes used instead of Lx] for the floor function.

Extra! Let x and y be positive irrational numbers such that z + p = 1. Use a computer to
l"* ,'* *l,*l l* compute the first 50 terms in the sequences Lx], ∣2xj, [3xj, ... and |_yj, ∣2yJ, ∣3yJ, ....
Notice anything interesting?

529
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Here are a few examples:

L3J = 3, Γ3^∣ = 3, {3} = 0,

L6.25J = 6, Γ6.251 = 7, {6.25} = 0.25z
L-6.25J = -7, Γ-6.25∏ = -6, {-6.25} = 0.75z
LπJ = 3, KI =4, {π} = π - 3.

In a sense, ∣x], Γx], and {x) can all be thought of as piecewise defined functions. For example, Lx]
equals -2 if -2 ≤ x < -lz equals -1 if -1 ≤ x < 0, equals 0 if 0 ≤ x < 1, equals 1 if 1 ≤ x < 2, and so on.
Most of our problems will focus on the floor function, because it is the most common of these three
functions.

Problems

Problem 16.17: Sketch the graphs of y = LxJ∕ y = ΓχJ∕ and xJ - W∙

Problem 16.18:
(a) Find the smallest integer n such that 2" > 1000.
(b) Compute Llog2 lθθθJ∙

Problem 16.20:
(a) Find all x such that ∣xj = 7.
(b) Find all x such that [3xJ = 7.
(c) Find all x such that [3x + ⅜J = 7.

Problem 16.21: Find all y such that 3 < ∣4y - 5J ≤ 8.

Problem 16.23:
(a) Prove that ∖x + nJ = ∣xJ + n f°r aU real numbers x and integers n.
(b) Prove that {x + n] = {x} for all real numbers x and integers n.
(c) Prove that ∣x + yj > ∣xJ + LyJ fθr ali real x> xJ∙

20073 20053
Problem 16.24: Evaluate 2005 ∙ 2006 ~ 2006 ∙ 2007 without a calculator.

530
 16.4. FLOOR AND CEILING

Problem 16.17: Sketch the graphs of y = ∣χj, ι∕ = [^χj, and ι∕ = (x}.

Solution for Problem 16.17: We start with y = [χj. We have y = 0 for all x such that 0 ≤ x < 1. Similarly,
y = 1 for 1 ≤ x < 2, and y = 2 for 2 ≤ x < 3, and so on. Looking at negative values of x, we have y = -1
for -1 ≤ x < 0, and y = -2 for -2 ≤ x < -1, and so on. Putting all this together gives us the step-like
graph at left below.

y = LχJ y = Γ*l y = {x)

We find the graph of y — ∣^x^∣ similarly, and discover the step-like graph in the middle above.
In graphing y - {x}, we note that as x goes from one integer to just less than the next larger integer,
the value of {x} increases linearly from 0 to just less than 1. This produces the repeating saw-tooth
pattern at right above.

These three graphs clearly depict the piecewise nature of these three functions. □
Just to make sure we have a handle on the floor function, we'll start by computing the floor of a
specific number.

Problem 16.18: Compute Llog2 1000J.

Solution for Problem 16.18: If [log21000J = n, then we must have n < log21000 < n + 1, where n is an
integer.

Concept: The equation |_xj = n, where n is an integer, is equivalent to the inequality

O~^ Ji3 chain n < x < n + 1. Similarly, fx] = n is equivalent to π - 1 < x ≤ ∏. This
provides a method for getting rid of floor or ceiling notation in a problem.

Converting our inequality chain from logarithmic notation to exponential notation gives

2'' ≤ 1000 < 2"+1.

Since 512 = 29 < 1000 < 210 = 1024, we have n = Uog21000J = 9. □

Problem 16.19: Graph each of the following:

(a) ∕(x) = Lχ] + L-χJ (b) g(x) = {x} + {-x} (C) f+g
■■■■■■■■■■■■■■■■■■■■■■■■■Mi

531
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Solution for Problem 16.19:

(a) When x is an integer, we have ∕(x) = LxJ+L-xJ = x+(~x) = θ∙ But what if

x is not an integer? Then, we have n < x < n+1 for some integer n, so we
have -n > -x > -n-1. This gives ∕(x) = [xj + [-xJ = n + (-n -1) =
We can therefore write ∕(x) as

ft . ∫ ° if x is an integer,
∕(x) = i
—1 if x is not an integer.

The graph of f is shown at right.

(b) We start with the definition of {x}, and we have J
1 I

g(x) = {x} + {-x} = x - ∣xj + (-χ) - L-χJ = - (LχJ + L-

J)
* ∙ |
1
Aha! This is just the negative of the function in part (a)! So, we graph
.A—4 ,,
g by reflecting the graph of f in part (a) over the x-axis, as shown at
i
right. [ ^"

(c) Looking at the graphs of f and g in parts (a) and (b), we expect the
graph of f+g to simply bethex-axis. We can also see that ∕(x)+g(x) = 0 1 r I
by using our definition of {x}. Since {x} = x - LxJ∕ we have {x} + LxJ = X∕ so

∕(x) + g(x) = LxJ + L-

*J + W + l→}
= LχJ + W + L-χJ + {-χ} = χ + (-χ) = 0.

So, the graph of f + g is the x-axis.

Now that we have a little experience computing and graphing the floor function, let's try solving
some equations and inequalities involving the floor function.

Problem 16.20: Find all x such that L3x + jj = 7.

Solution for Problem 16.20: Because L3x + jJ = 7, we know that 3x + 5 is at least 7 and less than 8:

7 ≤ 3x + ⅛ < 8.
2

We subtract ∣ from all three parts of this inequality chain to get y ≤ 3x < y, and then divide this chain
by 3 to find y ≤ x < ∣. □
The process we used to turn the equation in Problem 16.20 into an inequality is yet another example
of removing complicated notation from a problem to turn it into a problem we know how to handle.

Problem 16.21: Find all y such that 3 < [4y - 5J ≤ 8.

Solution for Problem 16.21: Because L4y - 5 J must be an integer, we can rewrite the inequality as
4 ≤ L4y - 5J ≤ 8. This tells us that 4y - 5 is at least 4 and less than 9, so we can get rid of the floor function

532
 16.4. FLOOR AND CEILING

notation, leaving 4 ≤ 4y - 5 < 9. Adding 5 to all three parts gives 9 ≤ 4y < 14, and then dividing by 4
gives ∣ < y < 2. □ ,

Problem 16.22: Find all real values of a such that Jj-fl2 + 10« - 16∣J = 1.

Solution for Problem 16.22: In order to solve this problem, we get rid of all the complicated notation, one
step at a time. First, removing the floor function notation gives

1 ≤ ∣-fl2 + 10« - 16∣ < 2.

Now, removing the absolute value notation, we're looking for values of a such that either

1 ≤-«2 + 10« - 16 < 2 or -2 <-«2 + 10« - 16 ≤-1.

Ratheι than solve 4 different quadratic inequalities, we complete the square in —fl2 + 10« — 16, which
gives us
1 ≤-(o-5)2+ 9 < 2 or -2 < -(« -5)2 + 9 ≤ -1.
Isolating (« — 5)2 in these inequalities gives

8 > (« - 5)2 > 7 or 11 > (« - 5)2 ≥ 10.

All of the quantities in each inequality chain are positive, so we can take the square root of both
inequalities to obtain
√8≥a-5>√7 or . 1√∏ > a - 5 ≥ √10.
Noting that y/8 = 2 √2 and isolating a in each inequality chain gives us

5 + 2V2 ≥fl>5 + V7 or 5 + √T1 > a ≥ 5 + √10.

□
Problem 16.23:
(a) Prove that \_x + n] = LxJ + n for all real numbers x and integers n.
(b) Prove that {x + n} = {x} for all real numbers x and integers n.
(c) Prove that Lx + yj ≥ LxJ + LyJ for all real x, y.

Solution for Problem 16.23:

(a) To show that Lx + n] = LxJ + n, we must show that LxJ + n < x + n < LxJ + n + 1. Subtracting n from
all three parts gives Lx J ≤ x < LxJ + 1, which we know is true for all x. So,we have our proof:
Proof: For all x, we have LxJ ≤ x < LxJ + 1∙ Adding n to all three parts of this inequality gives
0 LxJ + n ≤ x + n < LxJ + n + l, so we have Lx + n∖ = LxJ + n.
(b) First, we write both {x + n} and {x} in terms of the floor function. We have ∣x + n} = x + n — [x + nJ
and {x) = x - LxJ. We can use part (a) to write Lx + nJ as LxJ + n, and we have

{x + n} = x + n-[x + n]=x + n- (LxJ + n) = x - LxJ = {x}.

533
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

(c) We need to have both LxJ and LyJ on the lesser side of an inequality. We know two inequalities
with these on the lesser sides: Lx J ≤ x < LxJ + 1 anc^ LyJ ≤ y < LyJ + 1∙ Adding these, we get

LχJ + LyJ ≤ χ + y < LχJ + LyJ + 2.

So, the number x + y is greater than or equal to the integer LxJ + LyJ and less than the integer
LxJ + LyJ + 2. This means that Lx + yj is equal to LxJ + LyJ θr LxJ + LyJ + 1∙ either case, we have
Lχ + yJ ≥ LχJ + LyJ-

We'll finish this section where we started: evaluating the floor of an expression. This one's a little
more complicated than Llog21000J.
ι------------------------------------ 1
20073 20053
Problem 16.24: Evaluate without a calculator.
2005 ∙ 2006 2006 ∙ 2007
I.,

Solution for Problem 16.24: We certainly aren't going to multiply and divide everything out. We hope
we can simplify the expression with some algebra, so we let n = 2005, write the expression inside the
floor function in terms of n, and hope something good happens. We have

20073 20053 _ (n + 2)3 n3 _ {n + 2)4 - n4

2005 ∙ 2006 ~ 2006 ∙ 2007 ^^ rfn + 1) ~ (n + l)(n + 2) ^ n(n + l)(n + 2)
8n3 + 24m2 + 32m + 16 _ 8(n + 1)(∏2 + 2n + 2)
n(n + l)(n + 2) n(n + 1)(∏ + 2)
_ 8(h2 + 2n + 2) _ 8n(n + 2) + 16
n(n + 2) n(n + 2)

n(n + 2)

16
Something good happened! Because 0 < < 1 for n = 2005, we have
n(n + 2)

20073 20053
2005 ∙ 2006 ^^ 2006 ∙ 2007

Concept: When you have to compute an expression involving very large numbers,
θ uj assign a variable to one of the numbers in the problem. Write the expression
in terms of that variable, and then simplify the expression with algebra
before evaluating it.

I xcr< J⅜e%

16.4.1 Find all values of x that satisfy LxJ + ∣^x] = 9.

534
 16.5. PROBLEM SOLVING WITH THE FLOOR FUNCTION

16.4.2 Find all integers n such that [rz∕6J = 5.

16.4.3 Evaluate [ -y∕log3 1000J in your head. Do not use a calculator and do not use paper and pencil.

16.4.4 Graph both ∕(x) = Γx] + ∣xj and g(x) = Γχ^∣ - ∣χ].
16.4.5 Find all x such that -2 ≤ [3x + 2j ≤ 2.
π. ∣√3}2-2{√2}2 ,.1
16.4.6 FUld {√3} 2{√2} with°ut a calculator∙ (Source: NYSML)

331 +231
16.4.7* Compute without a calculator. (Source: ARML) Hints: 260, 22
329 + 229

16.5 Problem Solving with the Floor Function

In this section, we explore a couple more approaches to solving problems involving the floor function.
You shouldn t view these as brand-new techniques; they have the same motivation as nearly all the
problem solving approaches wθ,ve employed in the last few chapters:

Concept: When faced with a problem involving a new notation or new function, try
to find a way to turn the problem into one you know how to do.

Problems

Problem 16.26: Determine the value of 10(x + y), given that [xj + LyJ + y = 43.8 and x + y - [χj = 18.4.
(Source: AIME)

Problem 16.28: Let f(x) = [2xJ + ∣,4xJ + L6xJ + L8x].

(a) At what values of x does the graph of y = ∣2x] "step up" from one value of y to another? How
about the graph of y = [4xj? The graph of y = [6xJ? The graph of y = [_8xJ?
(b) How much does the graph of f increase at each of the values of x you found in part (a)?
(c) How many of the first 1000 positive integers are in the range of/(x)? (Source: AIME)

535
 CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Problem 16.25:
(a) Find all values of x such that x + LxJ = 6.7.
∖
(b) Find all real numbers k satisfying [A:J = 5k- 14.

Solution for Problem 16.25:

(a) We know that [xj is an integer, so if x + L*J equals 6.7, then the ".7'' must be the fractional part of
x. Therefore, we can write x as (xj + 0.7, so our equation becomes 2LxJ + 0.7 = 6.7, from which we
find Lx] = 3. This gives us x = 3.7.
The key step in this solution was viewing x as the sum of its floor and its fractional part:
χ = L*J + W∙
Concept: Writing a variable as the sum of its fractional part and its floor can help
o≡≡g unravel problems involving the floor function.

(b) We use our previous part as inspiration, writing k = L^J + W∙ This gives us

Lkj = 5k - 14 = 5L&J + 5{fc} - 14.

Solving for {k}, we get {k} = 4∣ΛJ g∣nce o ≤ {∕c} < 1, we have 0 ≤ ----- < 1- Solving this
inequality chain for L^J gives ∣ < LAJ ≤ j∙ The only integer that lies in this interval is 3, so L^J = 3.
Therefore, we have [k} = (14 — 4 ∙ 3)∕5 = 2/5, so k - |_AJ + [k} = 17/5.
We also could have started with the "build an inequality" approach from the previous section.
We first note that L^J ≤ k < L^J + 1. Solving the given equation for k gives us

, _ L/cJ + 14
k~ 5 '

litl + 14
so [A:] ≤ —------ < [AJ + 1. From here, we can obtain the same bounds on [AJ as before.
5
□
In both parts, we use the fact that:

Important: The fractional part of a number is greater than or equal to 0 and less
than 1.

The inequality chain 0 ≤ {x} < 1 is really just another way of writing [xj ≤ x < *JL + 1. (If you don't see
why these inequality chains are equivalent, add |_xj to all three parts of 0 ≤ {x} < 1.)
Here's another problem in which viewing a number as the sum of its integer and fractional parts
helps immensely.
Problem 16.26: Determine the value of 10(x + y), given that [xj + [yj + ι∕ = 43.8 and x + ι∕ - [xj = 18.4. ∣∣
(Source: AIME)

536
 16.5. PROBLEM SOLVING WITH THE FLOOR FUNCTION

Solution for Problem 16.26: From our first equation, we know that the fractional part of y is 0.8. This
makes us think to write x = ∣xj + {x) and y = LyJ + {y}∙ Then the given equations become

W+2LyJ + {y)=43.8,
W + LyJ + {y} = 18.4.

The only non-integer on the left side of the first equation is {y}, so we must have {y} = 0.8 to
account foι the fractional part of 43.8 on the right side. Therefore, our second equation gives us
+ LyJ — 18.4 — 0.8 = 17.6. This tells us that {x} must be the fractional part of 17.6, so {x} = 0.6. Now
our system of equations is

LxJ + 2 LyJ = 43,

LyJ = 17.
Subtracting twice the second equation from the first gives LxJ = 43 - 2 ∙ 17 = 9, so x = LxJ + {x} = 9.6 and
y = LyJ + {y} = 17.8. Finally, we have 10(x + y) = 274. □
As we saw in Problem 16.17, the graph of y = LxJ looks somewhat like a series of steps. Some
problems can be solved by considering where these "steps" take place.
99
Problem 16.27: Find ∑ [√ξ∣.
k=l

Solution for Problem 16.27: First, we get rid of the summation notation to get a look at the terms in the
sum:
99
∑ [√⅛] = [√i] + [√2] + [√3] + [√4] + [√5] + ∙ ■ ■ + [√W].
k=l

Let's start with the first few terms. The first term is easy: [√lj = fl J = 1. What is [V2]? Because
1 < 2 < 4, we have 1 < V2 < 2, so ∣^V2j = 1. Similarly, we have [V3j = 1.

Continuing in this manner, each term L>∕fcJ in the sum is equal to the previous term until k becomes
a perfect square, at which point the term "steps up" by 1. We see that 1 appears 3 times in the sum, and
2 appears 5 times. We might guess that 3 appears 7 times, 4 appears 9 times, and so on. Let's see if we
can prove that this is the case.
We do so by finding all integers k such that ∣~Vkj = m, where m is a positive integer. If |_ = m, then
we must have m ≤ x/k < m + 1, so m2 < k < (m + 1)2. In other words, k must be one of the integers m2,
m2 + 1, m2 + 2,..., (m + 1)2 - 1. There are (m + 1)2 - 1 - m2 + 1 - 2m + 1 integers in this list, so there are
2m + 1 integers k such that = m.

So, each integer m with 1 ≤ m ≤ 9 appears 2m + 1 times in the sum. We have to be careful to check
the boundary cases. Since our sum starts at Vl and ends one term shy of ^V102j, we do indeed have all
2 ∙ 1 + 1 = 3 terms that equal 1 and all 2 ∙ 9 + 1 = 19 terms that equal 9.

537
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Now, we can evaluate our sum:

99 9
ΣM = Σ
k=l m=l
(2m + l)m = 3 + 10 + 21 + 36 + 55 + 78 + 105 + 136 + 171 = 615.

□
In this problem, we removed the floor function by counting how many times each integer appears
as a term in the sum. This approach was inspired by checking the first few terms of the sum, which
revealed a pattern in the number of times that each integer appears in the sum. Then we proved that
our pattern always holds.

Concept: Finding patterns is an important problem solving tool. This is particularly

Q--∙ l,j true of problems involving sequences and series—looking at early terms
in a sum can reveal clues about how to evaluate the sum.

Problem 16.28: How many of the first 1000 positive integers can be expressed in the form

L2xJ + ∣4xJ + L6xJ + L8xJ,

where x is a real number? (Source: AIME)

Solution for Problem 16.28: Let ∕(x) = ∣2x] + L4xJ + L6xJ + ∣8*J∙ We experiment with this function by
examining what happens to ∕(x) as x goes from 0 to 1. For all x ∈ [0, §), we have ∕(x) = 0, since 2x, 4x,
6χ, and 8x are positive numbers less than 1. But when we hit x = ∣, the value of L8xJ steps up from 0 to
1, and we have /(§) = 1. For all x ∈ [∣, ∣), the first three terms of ∕(x) equal zero, while the last equals 1.
But when we hit x = j, ∕(x) steps up again, and we have ∕(∣) = 2.

Continuing in this vein, we see that ∕(x) = 2 for all x ∈ [g, ⅜)∙ But when x = ∣, something interesting
happens: ∕(x) steps up by 2, because both ∣4xJ and ∣8xJ step up by 1 when we hit x = 5. So, we find
y(l) = 0 + l + l+ 2 = 4. Now, we're starting to get a grasp on ∕(x). We see that ∕(x) steps up whenever
at least one of its terms steps up, so we focus on those values of x at which each term steps up.

We find that
n
[2x] steps up when x = -, for any integer n,
∏
[4xJ steps up when x = ^7> for any integer n,
∏
[6xj steps up when x = -, for any integer n,
n
L8xJ steps up when x = —, for any integer n.
0
This means that steps occur at ⅛, g, 5, 3, ∣, ∣, ∣, 3, ∣, ∣, ∣, and 1. So, from ∕(0) to ∕(1), ∕(x) steps up a total
of 12 different times. However, ∕(x) sometimes steps up more than 1 unit at a time, so some integers
are not in its range. Because ∕(0) = 0 and ∕(1) = 20, as f takes 12 steps from 0 to 20, it hits 12 positive
integers (including 20).

538
 16.6. SUMMARY

Now, we have a handle on how ∕(x) behaves for 0 ≤ x ≤ 1. What about other values of x? Do we
need to start from scratch for 1 ≤ x < 2?

No! From our analysis above, we see that ∕(x) steps up exactly 12 times as x goes from n to n + 1 for
any integer/?, not just for n = 0. Moreover, for any integer π, we have f(n + l)-f(n) = 20n + 20-20n = 20.
So, during these 12 steps, ∕(x) hits 12 of the integers from 20h to 20n+20, including 20n+20 but excluding
20n.

Putting all this together, we note that as x goes from 0 to 50, the value of f(x) starts at 0, steps up
12 ∙ 50 = 600 times, and ends at 1000. Each step gives us another integer than can be expressed in the
form ∣2xJ + ∣4xJ + L6xJ + [8x], so we have 600 such integers. □

Concept: Much of the behavior of many functions that involve the floor function is
Or~ ulJ determined by when and by how much the floor function part(s) step up.

Exercises

16.5.1 Find all values of x such that x + |_xj = 5.7.

16.5.2 Compute the number of ordered pairs (x, y) with x > 0 and y > 0 that satisfy x + LyJ = 5.3 and
y + ∣xj = 5.7. (Source: ARML)
16.5.3 Solve the equation 7t + [2t] = 52.

16.5.4 For how many positive integers n is it true that n < 1000 and that ∣,log9 n] is a positive even
integer? (Source: AIME)

16.5.5
* Let Sn = LVlJ + L√2J + L√3J + ∙ ∙ ∙ + L√κJ. Compute the largest value of k < 1997 such that
Si997 - Sk is a perfect square. (Source: ARML)
2002
16.5.6
* Find the smallest real number x such that;—- . (Source: ARML) Hints: 261, 290
LxJ 2003

16.6 Summary

A piecewise defined function applies different rules to produce output for different inputs. To define
such a function, we must describe both the rules used to produce output, and the inputs each rule is for.
In this chapter, we studied a few special classes of functions that can be thought of as piecewise
defined functions.

Definition: The absolute value of the real number x, denoted as ∣x∣, is defined as

if x > 0,
if x < 0.

539
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

Definitions:

• The floor of x is defined as the greatest integer that is less than or equal to x, and is denoted
by LxJ.
• Similarly, the ceiling of x is defined as the least integer that is greater than or equal to x, and
is denoted by ∣^x].

• The fractional part of x is defined as x - L*J, and is denoted by {x}.

Problem Solving Strategies_____________________

Concepts: ∙ When working with an unusual expression in a problem, try to find

Q∙~ljJ ways to convert the problem into one involving expressions you
know how to handle.

• Don't just stare at a problem when you're stuck. Experiment and

look for patterns.

• An effective technique for solving some equations with absolute val­

ues is to break the problem into cases by viewing the absolute values
as piecewise defined functions. Remember to check each solution.
In each case, the solution must satisfy the restrictions of that case in
order to be a valid solution.

• Viewing the absolute value of the difference between two real num­
bers as the distance between the numbers on the number line can
help solve some absolute value problems.

• When you get a simple answer after doing a lot of casework, look for
a simple explanation for the simple answer.

• Sometimes we can use the graph of a function to give us intuition

about an algebraic problem.

• The equation L*J = n, where n is an integer, is equivalent to the

inequality chain n < x < n + 1. Similarly, [^x^∣ = n is equivalent to
n - 1 < x ≤ n. This provides a method for getting rid of floor or
ceiling notation in a problem.

• When you have to compute an expression involving very large num­

bers, assign a variable to one of the numbers in the problem. Write
the expression in terms of that variable, then simplify the expression
with algebra before evaluating it.

• Writing a variable as the sum of its fractional part and its floor can
help unravel problems involving the floor function.

540
 REVIEW PROBLEMS

Review Problems

if x ≤ 3,
16.29 Let∕(x) = < 2 √x if 3 < x < 8,
2∣17-x∣ if x > 8.

(a) Find ∕(-2),∕(3), and ∕(8).

(b) Find all c such that ∕(c) = 6.

16.30 Find all intersection points of the graphs of f and g if

if x ≤ -3,
if x ≤ 4,
if -3 < x < 2, and
if x > 4.
if 2 ≤ x ≤ 9,

16.31 Simplify ∣χ - λ∕(x - 9)2∣ for x < 0.

16.32 Find all x that satisfy the equation x∣x + 1∣ — 1 = x — ∣x + l∣.

16.33 Given that a, b, and c are all real numbers, find all possible values of the expression

a b c abc
W + i⅛i+ id + ∣Hbd'
(Source: AHSME)
16.34 Find the minimum possible value of ∣x∣ + ∣x + 3∣, where x is a real number.
16.35 Let a and b be distinct real numbers. Find all x in terms of a and b such that ∣x - α∣ = ∣χ - b∖.
16.36 Let a, b, and c be real numbers such that ∖a∖ - ∣b - 2∣z ∣b∣ = ∣c - 2∣, and ∣c∣ = ∣a - 2∣. Prove that
Λ + + c = 3.
16.37 Find all possible values of a such that there are real numbers x for which ∣x - 4∣ + ∣x + 5∣ < a.
16.38 Find all ordered pairs of real numbers (x, y) such that ∣x∣ + ∣y∣ = 10 and xy = 24.
16.39 Find ∕-1(x) if ∕(x) = x∣x∣ + 2.
16.40 If x and y are nonzero real numbers such that ∣x∣ + y = 3 and ∣x∣y + x3 = 0, then find the integer
nearest to x - y. (Source: AHSME)
16.41 Solve each of the following:
(a) ∣x2 + 8x - 3∣ = 19 (c) ∣2 - 4r - r2∣ + 3 > 0
(b) ∣z2 - z - 18∣ + 3 = z (d) ∖f2 - 4f + 5∣ ≤ 6

16.42 A line intersects the graph of y = ∣x∣ at the points A and B. The midpoint of AB is (3,4). Find the
length of AB.

541
CHAPTER 16. PIECEWISE DEFINED FUNCTIONS

16.43 Evaluate Llog2 V999J.

16.44 What values of r satisfy ∣~V2r - 7J = 9?

16.45 Find all real x such that 1990Lx] + 1989L~xJ = 1.

16.46 Find all x such that x3 - [xj = 47.

16.47 Show that ∣^x^∣ = —1_—xj for all real numbers x.

16.48 For each real number x, let ∕(x) be the minimum of the numbers 4x + 1, x + 2, and -2x + 4. Find
the maximum value of ∕(x). (Source: AHSME)

Challenge Problems

16.49 Suppose r is a real number for which

19 20 21
Γ 4- ---------
100 J
+ L
T 4^ ---------
100 J
÷ Γ+ . ~
100J
+ = 546.

Find LTOOrJ. (Source: AIME) Hints: 207

16.50 Compute Lλ∕π2 - 10κ + 29J when n = 19941994. (Source: ARML) Hints: 332
16.51 Find the largest positive number δ such that ∣√x - 2∣ < 0.1 whenever ∣x - 4∣ < δ.
16.52 Find the area of the region enclosed by thegraphof ∣x-60∣+∣y∣ = ∣x∕4∣. (Source: AIME) Hints:349
2-3y
16.53 Find all values of y such that 3+ι∕
Hints: 248

16.54 Let S(x) = 1 if x ≥ 0 and S(x) = 0 if x < 0. Express the function

= { 0 if -3 ≤ x ≤ 3,
otherwise,

in terms of S(x) without using piecewise notation. Hints: 111

√64 - x2
16.55 Find the domain of ∕(x) =
^16-∣2x + 5∣'
16.56 Find 100(x + y + z) given that

χ + LyJ + {z} = i∙08,

Lx J + {y} + z = 3.31,
{x) + y + LzJ = 2.99.

Hints: 291
16.57 What positive real number xhas the property that x, L
*J, and x-LxJ form a geometric progression?
Hints: 242,193

542
 CHALLENGE PROBLEMS

16.58 Compute the smallest positive integer x greater than 9 such that
x x
∣xj-19∙ = 9 = [xj - 89 ∙
.19. 89.
(Source: ARML) Hints: 98

16.59 Find polynomials f(x), g(x), and h(x), such that, for all x,
r
-1 if x < -1,
∣∕(x)l - lg(x)l + h(x) = < 3x + 2 if -1 ≤ x ≤ 0,
-2x + 2 if x > 0.
(Source: Putnam) Hints: 88,142

16.60 Let x be a real number selected uniformly at random between 100 and 200. If L√xj = 12, find the
probability that L√100xJ = 120. (Source: AHSME)
16.61 Prove that
1 2
[nxj = kJ + x + - + x + — ÷ ∙ ∙ ∙ ÷ x+
n. n.
This relationship is called Hermite's Identity. Hints: 57, 204
16.62 Prove that [2xj + [2yJ > [xj + Lyj + [x + yj for all real numbers x and y.
16.63
* Let S be the set of points in the Cartesian plane that satisfy ∣∣∣x∣ - 2∣ - 1∣ + ∣∣∣y∣ - 2∣ - 1∣ = 1. If
you build a model of S from wire of negligible thickness, then exactly how much wire would you need?
(Source: AIME)

*
16.64 Let ∕(x) = x ∙ |_x ∙ [x ∙ [xjjj for all positive real numbers x.

(a) Find x such that ∕(x) = 2001.

(b) Prove there are no solutions to ∕(x) = 2002.

(Source: USAMTS)
16.65
* Find the number of points of intersection of the graphs of (x - [xj)2 + y2 = χ - [χj and y = -.
(Source: ARML) Hints: 211, 331
16.66
* Suppose that α is positive, {a~1} = {a2}, and 2 < a2 < 3. Find the value of a12 - 144∏~1. (Source:
AIME) Hints: 158, 232
1995 -
16.67
* Let∕(n) be the integer closest to -tyn. Find V ——. (Source: AIME) Hints: 161,97
2~4 J∖k)
k=l

16.68
* Define f(n) by
f(n) = | n∕2, if n is even,
(n + 1023)∕2, if n is odd.
Find the least positive integer n such that ∕(∕(∕(∕(∕(∏))))) = n. (Source: Duke Math Meet) Hints: 78,
271,131
1q20000

*
16.69 What is the units digit of ? (Source: Putnam) Hints: 156,128
101°0 + 3

543
CHAPTER 17. MORE SEQUENCES AND SERIES

In order to understand recursion, one must first understand recursion. - Anonymous

CHAPTER
l_
17 ι
More Sequences and Series

In this chapter, we explore more types of sequences and series, and we develop more techniques for
working with them. Many of these techniques have a common thread:

Concept: We can often learn a lot about a sequence or series by comparing each term
Q -pg to neighboring terms.

17.1 Algebra of Recursive Sequences

A recursive sequence is a sequence in which, after one or more initial terms, each term is defined as
a function of previous terms in the sequence. Such a sequence definition is called a recursion. The
recursion is sometimes also called the recursive formula or the recurrence relation.
The Fibonacci sequence is one of the most famous recursive sequences. The Fibonacci sequence
begins 0,1, and then each subsequent term is the sum of the two preceding terms. The first ten terms of
the Fibonacci sequence are
0,1,1,2,3,5,8,13,21,34.
Algebraically, we label the first two terms of this sequence Fq and F↑, so Fo = 0 and Fj = 1. Each
subsequent term is given by the recursion Fn = F„_i + Fn-2∙
In this section, we explore how to work with recursions algebraically. Art of Problem Solving's
Intermediate Counting & Probability text includes much more information about recursions, such as how
to solve counting problems with recursions.
One common way to work with recursive sequences is to find a closed form for the sequence. A
closed form for a sequence {«„} is a formula that can be used to compute a∣l for any n, without knowing
any other terms of the sequence, and without summations or notation involving "∙ ■ ∙". So, for example,

544
 17.1. ALGEBRA OF RECURSIVE SEQUENCES

an = 2n - 3 is a closed form for a sequence {απ}, but b,1 = 2bπ-ι and cn = 1 + 2 + 3 + ∙ ∙ ■ + (∏ - 1)2 + ∏2 are
not closed forms.

Problem 17.1:
(a) Let ⅜ = 3 and an = an~↑ + 7 for n > 0. What type of sequence is {fl∏}7 Find a closed form for an.
(b) Let bo = 4 and bn = 3bll^ for n > 0. What type of sequence is {bn}2 Find a closed form for bn.
(c) Let co = 7 and cu = 2cn-↑ + 1 for n > 0. Compute several terms of {cπ). Find a constant k such
that {c,1 + k} is a sequence you recognize. Find a closed form for cn.

Problem 17.2: Let {<⅛} be a sequence of integers such that a↑ = 1 and am+n = a∣n + an + mn for all positive
integers m and n. In this problem, we find α12. (Source: AMC10)
(a) Choose a convenient value for m to reduce the number of variables in the recursion.
(b) Use the recursion that results from part (a) to find flι2.

Problem 17.3: For n ≥ 1, the terms in a sequence satisfy the general relation a2 — an-ιan+3 = (-1)”.
(a) Find a formula for αtl+j in terms of απ and αn,1.
(b) Given a0 = 1 and a1 = 3, find the value of a3. (Source: AHSME)

Problem 17.4: Let a be a positive constant. Consider the sequence defined recursively by ιq = a and
⅝+l - -1∕(m7i + 1), for n = 1,2,3,.... Find the smallest positive integer d such that un+a = un for all
positive integers n and all possible values of a.

Problem 17.5: A sequence is defined such that α1 = α2 = 1 and ⅜+2 = g'i+1 + for n ≥ 1.

(a) Compute fl3,Λ4,z75,⅜,fl7, and as-

(b) Find the smallest positive integer d such that an+cj = an for all n ≥ 1.
(c) Compute α∣ where t = 19985. (Source: ARML)

Problem 17.6: A sequence of positive real numbers a∖, a2,a3,... has the property that for i ≥ 2, each al
is equal to the sum of all the previous terms.
(a) If α19 = 99, then what is «2o? (Source: Mandelbrot)
(b) Let a↑ = k. Find a closed form for an in terms of k for n ≥ 2.

Problem 17.7: Define two sequences of rational numbers as follows. Let ∏o = 2 and bo = 3, then
recursively define

a∏+ι ~ > and bn+↑ —

for n ≥ 0. Determine b3, leaving your answer in the form 3m∕2'l. (Source: Mandelbrot)

545
CHAPTER 17. MORE SEQUENCES AND SERIES

Problem 17.1:
(a) Find a closed form for an given that flo = 3 and an = an-∖ + 7 for n≥l.
(b) Find a closed form for bn given that bo = 4 and bn = 3bn-↑ for n≥l.
(c) Find a closed form for cn given that cq = 7 and cn = 2cn_i + 1 for n≥l.

Solution for Problem 17.1:

(a) We start by computing the first few terms of the sequence {an}, and we have ao = 3, «i = ⅜ + 7 = 10,
α2 = «i + 7 = 17, and α3 = α2 + 7 = 24. Our sequence starts 3,10,17,24; this looks like an arithmetic
sequence. Subtracting an-↑ from both sides of our recursive formula gives us an — an-∖ - 7.
This shows that the difference between consecutive terms in.the sequence {an} is constant, so {an}
is indeed an arithmetic sequence with common difference 7. To find an, we add the common
difference n times to ao, which gives us an = 7n + ao = 7n + 3. (We add the common difference n
times, not n-1 times, because we start with ao, not a-[.)
(b) Computing the first few terms gives us 4,12, 36,108. This looks like a geometric sequence. This
shouldn't be surprising; the recursion bn = 3b„_i tells us that after the first term, each term in the
sequence is 3 times the previous term, so the sequence {feπ} is geometric. To find bn, we multiply
bo by the common ratio n times: bn = 3" ∙ bo = 4 ■ 3”.
(c) This one's tougher than the last two. Again, we write out a few terms in the sequence, hoping to
spot something helpful. We note that the sequence is almost, but not quite, geometric:

7, 15, 31, 63, 127, 255, 511, 1023, 2047, 4095, ...

If we let dn = cn + 1, then {dn} appears to be a geometric sequence. To prove that it is, we let
cn = dn-l in the recursion cn - 2cn-ι + 1, and we have dn - 1 = 2(d„_i - 1) + 1. Rearranging this
gives dn = 2dn-∖. Therefore, after do, each term in {dn} is double the term before it, so the sequence
{dn} is geometric. Because do = co +1 = 8 = 23, we have dn = 2π+3. Therefore, cn = dn-l = 2n+3 -1.

□
Problem 17.1 showed us a very common way of investigating recursions:

Concept: To learn more about a recursion, compute the first few terms and look for
θ 'jo a pattern. If you find a pattern, use the recursion to figure out why the
pattern exists.

Some recursive formulas use more than one variable to define the recursion. Here's an example:

Problem 17.2: Let {⅜) be a sequence of integers such that ai = 1 and am+tl = am + a,l + mn for all ∣
positive integers m and n. Find Λ]2. (Source: AMC10) _______ I
Solution for Problem 17.2: We can assign any positive integer values to m and n. Because we know a↑ = 1,
we can simplify the recursion by letting m = 1, which gives us

an+∖ = λ1 + all + n = an + n + 1.

546
 17.1. ALGEBRA OF RECURSIVE SEQUENCES

Using this simpler recursion, we relate successive terms in the sequence:

(∣2 - fl] + 2,
«3 = «2 + 3,
= t?3 + 4,

∏12 — flu + 12.

Now we can calculate terms rather easily. In particular,

12 ∙ 13
fli2 - 12 + flu = 12 + 11 + α10 = ∙ ∙ ∙ = 12 + ∏ + ∙ ∙ ∙ + 2 + 1 = = 78.
2

Problem 17.3: Given ao - 1, a1 = 3, and the general relation of - an-τβn+ι = (-1)" for n ≥ 1, find «3.
(Source: AHSME)

Solution for Problem 17.3: We could jump right into substituting values for n and evaluating terms of the
sequence. However, if we first solve the equation for απ+ι, we'll have a simple computation at each step,
rather than a linear equation to solve. Isolating απ-1απ+1 gives us a„_i«n+i = α2 — (—1)”, and dividing by
⅜-ι gives
„ _ ⅛ - (-1)”
^n+l - z
∩n-l
as long as απ-ι ≠ 0.

Concept: Just as we can do with any other equation, we can manipulate recursive
Q—ljj formulas into more convenient forms.

Now that we have an+ι written in terms of earlier terms in the sequence, we quickly compute terms
up to «3:
⅛ ~ (-l)1
a2 =-------------
ao

4 - (~1)2 102 - 1 oo
«3 = — = —z— = 33.
a↑ 3
We could continue, but we have the term we needed. □

Sidenote: Compute a few more terms of the sequence in Problem 17.3. Are all the
J∖ terms of the sequence integers?

A periodic sequence is a sequence that repeats regularly. For example, we say the sequence

547
CHAPTER 17. MORE SEQUENCES AND SERIES

has period 3 because it repeats every three terms. More formally, we say that the sequence {an J is periodic
with period k, where k > 0, if an = arι+∣ς for all n for which an and an+jc are defined.

Problem 17.4: Let a be a positive number. Consider the sequence defined recursively by ιq = a
and u,1+1 = -l∕(un + 1), for n = 1,2,3,.... Find the smallest number d such that {un} is periodic with
period d for all values of a.

Solution for Problem 17.4: We start by computing the first few terms of {un} in terms of a. We are trying
to determine where the sequence begins to repeat.
We note that u2 = -1∕(∏ + 1) and then compute the next couple of terms:

1 1 a +1

1 1

a
We've determined that U4 = a = u^. Because the recursion relates each term only to the term immediately
prior, once we have found that ιq = ιq, we know that un+3 = un for all n ≥ 1. For example, because
«4 = ιq, we know that 1/5 = —l∕(z√4 + 1) = -l∕(ιq + 1) = ∏2∙ Similarly, because ιq = w5, we know that
«6 = «3/ and so on. Therefore, regardless of the value of a, we have un+3 = un for all positive integers n,
so the sequence {un} has a period of 3. We now also confirm that the sequence does not have a smaller
period. If a = 1, then the first three terms are 1, -∣, -2, so the period cannot be smaller than 3. Therefore,
the desired value of d is 3. □
Here's an example of periodic sequences in action:

+1 , „ c-
Problem 17.5: If ai = a2 = 1 and an+2 = ---------- for n ≥ 1, compute at where t = 1998< (Source:
a∏
ARML)

Solution for Problem 17.5: We begin by computing a few terms of the sequence. We sure aren't going to
write out 19985 terms, but hopefully we'll find a pattern.

Concept: When you're unsure how to begin with a sequence problem, consider
O≡^≡ writing out a few terms to see if anything interesting happens.

Extra! Consider the recursion

* ∣<∣n<∙∣ιι⅜∙ Λ,∣ι-1 ÷ x 1

χn = q

Let Xo = 4, and compute several terms in the sequence {x,,∣. Do the same with different
positive values of x0. Notice anything interesting? Try replacing the 9 in the recursion
with different positive constants and computing the first few terms in the sequence {.rπ}
for various positive values of %o∙

548
 17.1. ALGEBRA OF RECURSIVE SEQUENCES

We find:

«1 = 1,
fl2 = lz
ci2 ÷ 1 1+1
03 = = 2,
«1 1
«3 + 1 2+1
α4 = = 3,
∏2 1
#4 ÷ 1 3+1
«5 = = 2,
«3 2
a5 + 1 2+1
a6 = ≡ ------ = 1,
#4 3
⅜+1 1+1
ay = = ------ = 1,
«5 2
ay + 1 1+1
a8 = “ ------ = 2,
«6 1
and so on.

Each term depends only on the previous two terms, so once a pair of consecutive terms repeats, the
whole sequence repeats. From our work above, we see that the consecutive terms a∖, az are the same as
the consecutive terms a^, ay. Since «3 depends on aγ and «2 i∏ the same way as as depends on a⅛ and α7,
we have α3 = as. Similarly, az = ay and α3 = as give a4 = a9, and so on. In other words, the sequence has
period 5.
To figure out αf when t = 19985, we must find the remainder when 19985 is divided by 5. To do so,
we can write

19985 = (1995 + 3)5 = 19955 + 5(19954)(3) + 10(19953)(32) + 10(19952)(33) + 5(1995)(34) + 35.

All terms but the last are divisible by 5. Since 35 = 243, we therefore know that 19985 is 3 more than
a multiple of 5. We could also have used modular arithmetic: t = 19985 ≡ 35 ≡ 243 ≡ 3 (mod 5).
Therefore, the 19985th term in the sequence is the same as the 3rd term, and we have αf = α3 = 2. □

Concept: When looking for useful relationships to help solve sequence problems
Q—l3j (recursion problems in particular), keep an eye out for periodicity.

Problem 17.6: A sequence of positive real numbers α1,α2,α3,... has the property that for i ≥ 2, each
α1∙ is equal to the sum of all the previous terms.
(a) If α19 = 99, then what is azo7. (Source: Mandelbrot)
(b) Let α4 = k. Find a closed form for an in terms of k for n ≥ 2.

Solution for Problem 17.6:

(a) Our sequence definition tells us that for n≥2, we have

fln — Λn-1 + ⅜-2 + ,', + ∏2 + a↑.

549
CHAPTER 17. MORE SEQUENCES AND SERIES

Since #19 = 99, we have

#20 = #19 + #18 + ''' + a2 ÷ βl = 99 + flj8 + ∙ ∙, + #2 + fll∙

So, if we can find #is H------ ∏ #2 + ^i∕ we can find #20- From our sequence definition, this sum simply
equals #19! Therefore, we have

#20 = 99 + #18 + ∙ ∙ ∙ + #2 ^t ~ 99 + #19 = 198.

(b) We use part (a) as a guide. If we can simplify fln-21------- ÷^ fl2 + fli > then we can simplify our recursion
for #„. From our sequence definition, this sum is simply an-γ.

CLγι-∖ — Clrl~2 ÷ ∙ ∙ ∙ ÷ $2 ^1∕

so we have
Cln ” ^n-l + (‰-2 ÷ * * ’ ÷ #2 ÷ ^1) = ⅞n-l + (βn-l) = ^n-l∙

Aha! Each term an is double the preceding term. We might conclude that the sequence is a
geometric sequence with first term k and common ratio 2, and deduce that an = k2n 1. But testing
this formula reveals our error.
WARNING!! After deriving a closed form for a sequence or sum, test it with a few
V simple cases.
Our formula gives #i = k, as expected, but it also gives #2 = 2k. However, our recursion tells us
that #2 = #i- Where did we go wrong?
Our recursive formula an = an-∖ + #„-2 + ∙ ∙ ∙ + #2 + #1 only holds for n ≥ 2. Therefore, we can
only have #„_i = fln-2 + ∙ ∙ ∙ + #2 + «1 when n - 1 ≥ 2, which means we must have n ≥ 3. In other
words, we only have an = 2an-∖ for n ≥ 3. So, from the second term on, our sequence is geometric
with common ratio 2 and first term k, which means an = k2n~2 for n≥2.
WARNING!! When developing a general formula for a sequence or series, be careful
V about the initial terms. Sometimes a general relationship that holds
for most terms will not hold for the first few terms.

Our key step in Problem 17.6 is rewriting the recursion for an,

Un ~ ÷ ^n-2 ÷ ”• ÷¾ ^-^l∕

by replacing n with n — 1 to get

Ct∏-∖ = a∏-2 4- ∙ ∙ ■ 4- #2 + ^1∙

We call this "shifting the index" of the recursion.

Concept: Shifting the index of a recursion can be an effective way to learn more
Q-^□τ=) about the recursion.

Extra! There is # difference between not knowing and not knowing yet.
∣ >*d ∣⅜ ∣ ⅜>∣ a⅜ - Shelia Tobias

550
 17.1. ALGEBRA OF RECURSIVE SEQUENCES

Problem 17.7: Define two sequences of rational numbers as follows. Let ⅜ = 2 and bo = 3, then
recursively define

απ+ι = — and bn+1 = —

υ∏ (In
for n ≥ 0. Determine bg, leaving your answer in the form 3m ∕2fl. (Source: Mandelbrot)

Solution for Problem 17.7: We first notice that simply multiplying the two given recursions allows us to
cancel some terms and thereby obtain a simpler relationship among the α's and b's.

Concept: Seek simplicity.

O≡≡a

This gives us
, _an bn _ u
^n+l^n+l ~~ ι * “ a∏un∙
CLn
Since flo⅛ = 6 and an+γbn+ι = anbn, we have anbn = 6 for all n. Therefore, we have an = 6∕bn, so we can
write our recursion for bπ+1 solely in terms of bn:

From here, we can just crank out ⅛

⅛ _ (⅛∕6)3 _ ⅛2 _ (⅛∕6)32 _ ⅛ _ _ ⅛8
6^ ~ 6 " 63~6 “ 63∙6 ^ 632 ∙ 63 ∙ 61 ^ ^ ^ 637 ∙ 636 ∙ ∙ ∙ 631 ∙ 61'

The exponents of 6 in the denominator form a geometric sequence with sum 1 + 31 + 32 + ∙ ∙ ∙ + 37 =

(38 - 1)∕(3 - 1) = 6560/2 = 3280, so we have

⅛∙ β6561 g3281

637+36+-+31+30 63280 23280 ’

A key step in our solution was finding a recursive definition for bn that did not include terms in the
sequence {an}.

Concept: If a sequence {xπ} is defined in terms of numbers in another sequence, try

θ~ lj1ς) to find a way to define {xfz} without the other sequence.

Exercises

17.1.1 Find a closed form for an if ao = 4 and an = an-ι + ∣.

17.1.2 The sequence a↑, a2, (⅛,... satisfies a↑ = 19, «9 = 99, and, for all n ≥ 3, a,, is the arithmetic mean
of the first n - 1 terms. Find «2- (Source: AHSME)

551
CHAPTER 17. MORE SEQUENCES AND SERIES

17.1.3 The sequence of numbers t-∣, ⅛, ⅛, ∙ ∙ ∙ is defined by fι = 2 and

_ in - 1
,"l ^ ⅛ +1

for every positive integer n. Determine the value of t999. (Source: COMC)
17.1.4 Let {tn) be a sequence of positive real numbers, and let Sn = fι + ⅛ + '' ' + f∏∙

(a) Find tn if Sn = 1 - (l∕2)n.

(b) Find tn and Sn if tn + 2Sn = 3 for all positive integers n.

17.1.5 The increasing sequence of positive integers «i,«2/«3/ ∙ ∙ ∙ has the property that ««+2 = ⅜ + <⅞+ι
for all π ≥ 1. If α7 = 120, then find as- (Source: AHSME)
17.1.6
* A survey is given to a number of "sheeple" to determine how often sheeple go out to dinner.
Sheeple will always lie to make themselves seem as socially acceptable as possible. Sheeple all feel like
it would be best to go out to dinner slightly more than the average. Therefore, when asked how often
they go out to dinner on average per month, each sheeple determines his answer by averaging all of the
previous answers and then adding 0.6 (answers need not be integers). If the 7th sheeple claims to go out
to dinner 5.47 times per month, how often does the first sheeple claim to go out to dinner per month?
Hints: 141
17.1.7 The sequence ao, a∖,... satisfies an = kan-γ - an-2 for all n ≥ 2. Show that {an} is periodic for each
of the following values of k:

(a) -1
(b) √2
1 + √5
(c) * 2- Hints: 259

17.2 Telescoping

In certain sums and products, the terms can be manipulated so that most of them cancel, vastly simpli­
fying calculations. This technique is called telescoping after the collapsible telescopes that sea captains
once used.
Problems

Problem 17.8: Evaluate

Problem 17.9: Simplify the product ∣

H) H) H)H ∣. (Source: AHSME)

552
 17.2. TELESCOPING

Problem 17.10: Let T„ = l+ 2 + 3 + ∙∙∙ + n and

p Iz Iz Ii Ill
" T2 - 1 T3 - 1 ’ T4 - 1 ’ ’ ’ Tn - 1

for 11 = 2, 3, 4,.... Find P1991. (Source: AHSME)

Problem 17.11: In this problem we evaluate the sum

1 1 1 1 , , 1
T√2 + Γ3 + 3~4 + 4÷5+"' + 99 100’

(a) Find the sums of the first 2 terms, the first 3 terms, the first 4 terms, and the first 5 terms.
Conjecture a formula for the sum of the first n terms.
(b) Use your formula to compute a guess for the given sum. Is your answer the same as you found
in Problem 17.8? If not, do both that problem and this one again. If so, explain why.

Problem 17.8: Evaluate V (—---- -—

^-f∖n n + 1
n=l

Solution for Problem 17.8: Writing out the summation reveals a pattern that greatly simplifies the sum:

∑(∣-⅛)=(ι-l)÷(M)÷(l4)÷∙→(⅛-⅛)÷(⅛-ι⅛)∙
H=l

Each negative term except the last cancels with the next positive term, leaving only

y∣ /1 1 ∖_ 1 1 _ 99
n + l)~ 1 100 ^ 100’

The type of cancellation that we used to solve Problem 17.8 is the key step in tackling many series
problems. We call any sum that can be evaluated with this technique a telescoping series.
As well as telescoping sums, there are telescoping products.

Solution for Problem 17.9: When we rewrite each term in the product as a proper fraction, it becomes
clear that most of the terms cancel:
2 3 4 n-1 _ 2∙(3∙4∙5∙∙∙(n~l)) _ 2
3 4 5 n (3 ∙ 4 ∙ 5 ∙ ∙ ∙ (n — 1)) ∙ κ n

553
CHAPTER 17. MORE SEQUENCES AND SERIES

It sometimes takes a bit of algebraic manipulation to see how terms will cancel in a telescoping sum
or product:

Problem 17.10: Let Tn =1+2+34------ ∣- n and

p T2 T3 τ4 Tn
n T2 - 1 ' T3 - 1 T4 - 1 Tn-1

for n = 2,3,4,.... Find P1991. (Source: AHSME)

Solution for Problem 17.10: First, we note that Tk = k(k + 1)∕2. Our hope is that the product telescopes,
so we look for a way to factor the denominators, which are of the form Tk — 1. We find that

rr 1 k2 + k 1 k2 + k-2 (k-l)(k + 2)
= = = -2 ■

Now we write out the product to see which factors cancel, and which remain:

P1991 =
τ2 T3 T4 T1990 Ti99i
T2-l T3 - 1 ’ T4 - 1 Ti990 - 1 Ti99i - 1
2-3/2 3-4/2 4-5/2 1990 ∙ 1991/2 1991 ∙ 1992/2
1 ∙4∕2 2 ■ 5/2 ’ 3 ∙ 6/2 1989 ∙ 1992/2'1990 ∙ 1993/2
2∙3 3∙4 4-5 1990-1991 1991-1992
1∙42∙53∙6 1989 ∙ 1992 ’ 1990 ∙ 1993
2 ∙ 32 ∙ 42 ∙ ∙ ∙ 19912 ∙ 1992
1 ∙ 2 ∙ 3 ■ 42 ∙ 52 ■ ∙ ∙ 19902 ∙ 1991 ■ 1992 ∙ 1993
3∙1991 __ 5973
1993 - 1993’
□
Problem 17.11: Evaluate the sum — + — + — + — + ∙ ∙ ∙ + 99 iqq∙

Solution for Problem 17.11: Adding term-by-term would take a long time, so we'd like to find a nicer
approach. We begin by adding the first few terms of the series to see if partial sums of the series shed
any light on the problem:

1 1 _ 1 1 _ 2
l-2 + 2-3~ 2 + 2~3~ 3'

1 1 1 _ 2 1 _ 3
Γ^2 + 2∙3 + 3∙4-3 + 3∙4-4'

1 1 1 1 _ 3 1 _ 4
T^√2 + 2^^3 + 3^∙^4 + 4 ∙ 5 - 4 + 4 ∙ 5 - 5'

1 1 1 1 14 15
— + τ~r + r—τ + ~r~z + z~τ = ≈ + Z~7 = 7∙

554
 17.2. TELESCOPING

Concept: Simplifying a problem and looking for patterns is a great way to learn
θ=∙jj∣J more about the problem.

It looks like we have a pattern, and if we had to guess, we'd guess that
1 , 1 l 1 ι 1 1 _ 99
1∙2 2∙3 3-4 4 ∙ 5 + + 99 ∙ 100 - 100'
Hey, that's the same answer we found when adding the 99 terms in the summation of Problem 17.8.
Maybe this is the same problem as that one! Let's take a closer look.
First, we write down the general statement we think is true:

1 + ι + 1 1 1 _ κ
1∙2 2∙3 3∙4 4∙5 + '" + (n)∙(n + l) ^ ^+l'

Concept: When trying to prove that a pattern exists, write a general form that rep-
resents the pattern.

We want to compare this equation to what we found in Problem 17.8. So, we write a general result
for Problem 17.8:

(1 _ 1 V√1 --V
2) -M+∙.→f∖n -1 1 n∕
1b 1 11-1 1 n
k+l∕ ∖1 ∖2 3/ ∖rz n + l) H+1 n+1

Now, we should be very suspicious!

Concept: There aren't many coincidences in mathematics. If two types of problems

Q —j consistently give the same answer, look for a reason why.

Specifically, we suspect that the sum of the first n terms of the form equals the sum of the first
n terms of the form Q - j⅛J. Now that we know what to look for, the answer is staring us right in the
face. We find a common denominator in the latter expression, and we have
1 1 _ k+1 k _ 1
k k + l~k(k+T) k(k+l)~ k(k+l)'
Aha! Each term in the sum
1111 1 _ y∣ 1
1 ∙ 2 + 2 ∙ 3 + 3 ∙ 4 + 4 ∙ 5 + ” ’ + (n) ∙ (π + 1) ^ k(k + 1)
K— 1

π /1 1 ∖
equals the corresponding term in the sum J~,, (- - - j. So, we have
fc=ι
fi—3-=⅛(r--)∙
2-jfc(fc + l) Z-∣U k+1

555
CHAPTER 17. MORE SEQUENCES AND SERIES

We know that the latter sum telescopes to give n∕(n + 1), so the answer to our problem is indeed 99/100.
□
A key step in our solution to Problem 17.11 was writing the rational function in terms of rational
functions with simpler denominators:

1 _ 1 1_
k(k^+V)~ k k+Γ

We call this the partial fraction decomposition of

When we find the partial fraction decomposition of a rational function, we express it as the sum of
rational functions with linear or quadratic denominators. As we saw in Problem 17.11, partial fraction
decomposition sometimes helps us evaluate complicated series by turning them into telescoping series.
WeTl find the partial fraction decomposition of one more rational function, and then put this tech­
nique to work on a few more series.

Problems

12x + ll A B
Problem 17.12: Find constants A and B such that x-2 + 2x + 3'
(x - 2)(2x + 3)

(7n + 32) ∙ 3”
Problem 17.14: Find the value of the sum
n(n + 2) ∙ 4”
. (Source: Thomas Mildorf)

Problem 17.15: Let α↑, a2, ■ ■ ., an be an arithmetic sequence with nonzero common difference d. Find a
closed form in terms of «i, n, and art for
n-1
1
Σ
∕=1
β∕β∕+l

12x + 11 A B
Problem 17.12: Find constants A and B such that
(x - 2)(2x + 3) x-2 2x + 3 ‘

Solution for Problem 17.12: Multiplying both sides by (x - 2)(2x + 3) to get rid of the fractions gives

12x + 11 = A(2x + 3) + B(x - 2).

We could expand the products on the right side and equate coefficients of like terms to produce a system

556
 17.2. TELESCOPING

of equations, but we have a much faster way to find A and B. The equation must hold for all x; we can
choose values of x that enable us to find A and B quickly. Specifically, letting x = 2 eliminates the B term
and leaves 35 = 7A, so A = 5. Similarly, letting x = -3/2 eliminates A and leaves -7 = -7B∕2, so B = 2.
This gives us the partial fraction decomposition
12x +11 _ 5 2
2x2 -x- 6 x-2 + 2x + 3'

We can check our work by writing the right side with a common denominator:

5 2 _ 5(2x + 3) 2(x - 2) _ 12x +11

x - 2 + 2x + 3 (x — 2)(2x + 3) + (2x + 3)(x - 2) ~ 2x2 - x - 6 ’

□
We,ve already seen partial fraction decomposition help us telescope one series. Let's try it on a few
more:

100 co 1
Problem 17.13: Evaluate V ——— and V ———.
⅛4 k(k + 2> ⅛ι k<~k + 2)

Solution for Problem 17.13: We first let = $ + d⅛∙ Multiplying by k(k + 2) gives 1 = A(k + 2) + Bk.
Letting k = 0 gives A = ⅛ and letting k = -2 gives B = —1, so we have
ιoo 1
Σ k(k + 2) 1
2
k=l v 7

Now, we see that the last 98 terms of the first sum cancel with the first 98 terms of the second sum,
leaving:
1 ∕1 1 1 1 ∖ _ 7625
2∖ + 2 101 102/ 10302'

We also could have seen the cancellation by rewriting the second summation as a summation of p to
match the expression we sum in the first summation. We do so by letting k range from 3 to 102 instead
of 1 to 100. Then, we clearly see which terms cancel and which are left over:
1 100 q 1 100 1
1 y 1 1 y 1 7625
2 2-∣ 2 2-∣ k + 2 10302’
k=↑ k=l

The shift of index made it easy to see exactly which terms would be left.
To examine what happens when the sum has infinitely many terms, we first let there be n terms, and
find the sum in terms of n. From our work above, we see that
n
1 1
Σ
fc=l
k(k + 2) n+2

As n becomes infinite, the terms and equal 0, so the infinite series has sum ∣(1 + ∣) = ∣. □

557
CHAPTER 17. MORE SEQUENCES AND SERIES

We also might have tried a shift of index to see that the infinite series telescopes infinitely:

We have to be careful about this type of approach, for it can be misleading. For example, consider the
following:

Clearly something is wrong there. We see the mistake when we consider the corresponding finite sum:
n n n n n+1 n
y ι = y((∕c÷i)-∕c) = y(k+i)-y k=y k-y fc=(n+i)-ι=n.
fc=l k=l k=l k=l k=2 k=l

Now, we see that as n becomes infinite, so does the sum.

WARNING!! Be careful with infinity!

Solution for Problem 17.14: We're not so sure what to do about the 3n and the 4", but we have a pretty
good idea what to do about the rest of the expression we are summing. We let
7n + 32 _ A B
n(n + 2) n n + 2'

Multiplying both sides by n(n + 2) gives 7n + 32 = A(n + 2) + Bn. Letting n = 0 gives A = 16 and letting
n = -2 gives B = 9. Aha! We have A = 42 and B = 32. More 3's and 4's. We must be on the right track.
Our partial fraction decomposition is

7n + 32 42 32
n(n + 2) n n + 2'

We now have ∞ ∞
(7n + 32) ∙ 3» 3"
Σ
n=l
n(n + 2) ∙ 4n
n=l

Extra! The infinite! No other question has ever moved so profoundly the spirit of man.
∣∣∣φ∣∣∣⅜∣∣∣φ∣k - David Hilbert

558
 17.2. TELESCOPING

To get a feel for the sum, we write out a few terms:

y ∕42 32 ∖ 3”^ ∕42 3 32 3∖ ∕42 3^_3^

n=l x
n + 2'∣ 4n x11 4 3 4/ ∖2 42 4 42∕
3^~3^ 3^∖
∖ 3 ' 43 5 43 ∕ + ∖ 4 44 ~ 6 ' 44) +

1 3*∖
4 42∕
1 3^_1 ⅛ /1 3^_1 3^∖
3 ' 4 5 43∕ + ∖4 42 6 44∕ +

We expect to see terms cancel, and we're not disappointed. Because the denominator of the positive
term in the summation is n and the denominator of the negative term is n + 2, we expect to see each
negative term cancel with a positive term two positive expressions later in the series. That is, we expect
the negative term for n = k to cancel with the positive term for n = k + 2. Indeed, when n = k, we have

_32 3fc _ 3fc+2 _ 423fc+2 _ 42 3fc+2

k + 2 ’ 4fc ^ ~4fc(k + 2) ^ ~4t+2(k + 2) = ~k + 2 4⅛+2'

which is the opposite of the positive term when n = k + 2, as expected. This makes it even more clear
how our sum telescopes, leaving us only with the first two terms in the positive summation above,
which give us

We can confirm that the infinite series does indeed converge by finding the first k terms, then
considering what happens when n is infinite. We first break our summation into two summations:

k
32 32 3"+2
n+2 n + 2 4n
n=l n=l

We then shift the index by 2 in the summation we are subtracting to get

k
(∖n ∙ 34”n~2J]-Yl(((n +3π+2 ∖ _ y 3n y2( 3" 1 1
'H> 2) ∙ 4" ∕ - 2-> ι∖n∙4"-2∕ ^∖n ■ 4"~2J
κ=l M=1 ' , rι=1 n=4

All the terms in the first summation except the first two are subtracted in the second summation, and
all terms in the second summation are eliminated except the last two terms. We're left with

3n 3n \_ 3 32 3fc+1 3k+2
)~ 4^r + 2 ■ 40 (∕c+l)∙4fc^1 (k + 2)-4r

The last two terms become 0 as k becomes infinite (note that 3k+^^ ∕4k 1 = 9 ∙ (3∕4)fc 1), so the sum does
indeed converge to 33/2, as claimed. □

559
CHAPTER 17. MORE SEQUENCES AND SERIES

Concept: Many sums involving fractions can be rearranged so that they telescope.
O' 1^0

Here's one more:

Problem 17.15: Let ci∖, a^, ■ ■ ∙, an be an arithmetic sequence with nonzero common difference d. Find
a closed form in terms of «1, n, and an for
n-l
1
Σ
Solution for Problem 17.15: A summation of a fraction: let's take a closer look with our telescope! We'd
like to break the fraction into a difference of two fractions that will allow us to telescope. Inspired by
partial fraction decomposition, we look for constants B and C such that

1 _ B ι C
β∕βf+l dj di+1

Multiplying by α1∙α∕+ι gives 1 = Bα⅛ι + Cfl1∙. We don't get to choose the α1∙, but we do know that α⅛ι = d,
so we have (d + afB + alC - 1. This must hold for all the α1∙ in the sequence, so we choose B and C to
eliminate a,. This means we want B = —C, which leaves us with dB = 1, so B = 1/d and C = —l/d.
Checking that these work, we have

1/d -1/d _ 1 ∕⅜+ι -fli∖ _ 1 d _ 1

∏i Ui+1 d ∖ Uj(ii+] ∕ d UiUi+'[

as desired. Now, it's telescope time:

All the terms except l∕a↑ and -l∕an cancel, leaving a sum of

Sidenote: Partial fraction decomposition isn't only useful for evaluating telescoping
series. You'll see it again when you study calculus.

I xer<i¼e*
13
∏(∏ + 2)
17.2.1 Calculate J ------ —, leaving your answer in simplest form. (Source: Mandelbrot)
n=l

560
 17.3. SUMS OF POLYNOMIAL SERIES

17.2.2 Find------ - + —------- + —----- -- + ∙ ∙ ∙ +------- i------- ,

1 + √2 √2+√3 √3 + √4 √99 + √iθθ
17.2.3 Consider the sequence 1, —2,3, -4z 5, -6,..., whose nth term is (-l)"+1 ∙ n. What is the average of
the first 200 terms of the sequence? (Source: AHSME)

17.2.4 12 3 n -1
Let n be a positive integer. Find the sum — + — + — +
2! 3! 4! n∖
1 1 J___ 1
17.2.5 Find the sum------- 1------- + ’ '+ —
(2n- l)(2n + 1) + ” •+ . (Source: AHSME)
1∙3 3-5 255 ∙ 257
20
17.2.6 Find the product (1 + 2n + 1∖
n=l
7
____________1___________
*
17.2.7 Evaluate Hints: 312
k=l
^2 + V^ + l)+ ⅜ + l)2'

1 1 1 1
17.2.8
* Let Hn — 1 + _ + + ... + _z ancj ∣et γn = (κ + i)¾⅛√ provethatτ,+τ2 + r3 + '"°1∙
z d n

17.3 Sums of Polynomial Series

τι
We have seen a variety of ways to find a closed form for k. We used induction to prove that
k=l
n
=
*
Σ n(n + 1)(2∏ + 1) ∑fc3 = + 1)2
and
6 k=l 4
k=l

but these inductions didn't give any insight into how we could have found the expressions + +
0 .∙7 6
n2(∏ + 1)~
and------------- in the first place. In this section, we'll find a way to derive them.

Problems

n
Problem 17.16: In this problem, we use telescoping series to find a closed form for k.
k=l
(a) Simplify the expression (k + 1)2 - k2.
n n
(b) How are ^((A, + 1)2 - A2) and
fc=l
Σ k related?
k=l
n
(c) Find a closed form for Σ*∙L=1
561
CHAPTER 17. MORE SEQUENCES AND SERIES

n
Problem 17.17: Let n be a positive integer. Express J~
* j k2 as a polynomial. Hints: 12
Jt=l

20
Problem 17.18: Let ∕(x) = 3x2 -7x + 2. Evaluate Σ ∕(H)∙
n≈l

Problem 17.19: Find a closed form for the series l∙4 + 3∙6 + ∙∙∙ + (2n — l)(2n + 2).

Problem 17.20: Notice the pattern

32 + 42 = 52,
102 + 112 + 122 = 132 + 142,
212 + 222 + 232 + 242 = 252 + 262 + 272.

Prove that for any positive integer n, there exist n + 1 consecutive squares whose sum equals the sum
of the next n larger consecutive squares. Hints: 220

n
Problem 17.16: Use telescoping series to find a closed form for > k.
L k=l
I
Solution for Problem 17.16: We need to find a telescoping series that we can relate to our desired sum.
We try summing (k + 1)2 - k2 both because such a sum obviously telescopes, and because simplifying
this expression gives us a linear expression in k:

(k+V)2-k2 = k2+ 2k + l-k1 = 2k + l.

So we have
n H
J2<2fc+1) = + 1)2 - ^2) = (22 - 12) + (32 - 22) + ∙ ∙ ∙ + [(n + 1)2 - n2] = (n + 1)2 - 12 = n2 + 2n.
fc=l ⅛=ι

Now, we have our target in sight, because

i>÷β⅛Ht
⅛=1 k=l k=l
n
1 = 2 Σ*
λ

r ∏ > n
n2 + 2n- n _ n(n + 1)
So, we have 2 k + n = n2 + 2n, which means y ι k = 9 2 'α
<k=l > k=↑

562
 17.3. SUMS OF POLYNOMIAL SERIES

Let's see if we can use a similar approach to find the sum of the first n squares.

Solution for Problem 17.17: Inspired by our previous success, we start by summing (k + 1)3 - ki, both
because we know this sum will telescope, and because the k3 terms will cancel when we expand (fc + 1)3
in this expression, leaving (k + 1)3 - fc3 = 3k2 + 3k + l for all positive integers k. Therefore, we have

y,j(3k2 + 3k + 1) = Jπ[(∕c + 1)3 -k3] = (n + 1)3 - 13 = n3 + 3n2 + 3π.

fc=ι k=l

We also have
11 11 ( n ∖
∑(3fc2 + 3∕c + l) = ∑3k2 + Σ3* + ∑l=3∣∑⅛2∣ + 3 ÷ n.
k=l fc=l k=l k=l ∖fc=ι 7
( 11 ∖ ∕ ιι >
* ι fc2 +3
Therefore, we have 3 J~ k ÷ n = n3 ÷ 3π2 + 3/7, so
∖k=l 7 ∖fc=ι 7

2π3 + 3h2 + n _ n(n + l)(2n + 1)

6 ^ Γ^

We can test our formula by trying a few small values of n. If n - 2, our formula gives 2(3)(5)∕6 = 5,
which is indeed the sum of the first 2 positive squares. If n = 3, our formula gives 3(4)(7)∕6 = 14, which
matches the sum of the first 3 perfect squares.

Concept: Whenever you derive a general formula for a sequence or series, you can
Q -ljU usually test it with a few simple cases.

20
Problem 17.18: Let ∕(x) = 3x2 -7x + 2. Evaluate
n=l

Solution for Problem 17.18: We simply break the sum into pieces and use the formulas we derived in the
first two problems:
20 20 20 20 20
∑∕(") = ∑(3n2 - 7n + 2) = 3^ι∏2-7^2 « + 2^1
h=1 n=l n=l n=l n=l
-3(20(21)(41)) 7(20(21))
+ 2(20) = 8610 - 1470 + 40 = 7180.

563
CHAPTER 17. MORE SEQUENCES AND SERIES

Problem 17.19: Find a closed form for the series l∙4 + 3∙6 + ∙∙ + (2n - l)(2n + 2).________________ I

Solution for Problem 17.19: We can use the general term to help us rewrite the series using sigma notation:
n
l∙4 + 3∙6 + ∙∙∙ + (2n - l)(2n + 2) = ∑(2fc - l)(2Ar + 2).
k=↑

Now we expand the summed expression, break it into parts, and evaluate:
n rι
∑(2i - l)(2k + 2) = ∑(4⅛2 + 2k - 2)
k=l k=l

We went through some fairly complicated algebra to get this formula, so we test it on a few simple cases.
When n = 1, the series is simply 1 ∙ 4 = 4. Indeed, our formula gives (4 + 9 - 1)∕3 = 4 for n = 1. When
n = 2, the series is 1 ∙ 4 + 3 ∙ 6 = 22, and our formula gives (4 ∙ 8 + 9 ∙ 4 - 2)∕3 = 22. □
While we could have solved Problem 17.19 in several ways, rewriting the series with sigma notation
made it very clear how to relate the series to ones we already knew how to evaluate.

Concept: Notation itself can be a powerful problem solving tool. Rewriting some
series with sigma notation can help evaluate them more easily.
O≡^
Problem 17.20: Notice the pattern

32 + 42 = 52,
102 + 112 + 122 = 132 + 142,
212 + 222 + 232 + 242 = 252 + 262 + 272.

Prove that for any positive integer n, there exist n + 1 consecutive squares whose sum equals the sum
. of the next n higher consecutive squares.

Solution for Problem 17.20: One way to prove that there are n + 1 consecutive squares whose sum equals
the sum of the next n consecutive squares is to find the squares in terms of n. We could write 2n + 1
consecutive squares as m2, (m + 1)2, (∕π + 2)2, ...,(m + 2ri)2r but if we instead let m be the middle square,
then the 2n + 1 consecutive squares are (m - n)2, (m - n + 1 )2,..., (in — 1)2, m2, (m + 1)2, ...,(m + ri)2. Our
second list has a nice symmetry.

564
 17.3. SUMS OF POLYNOMIAL SERIES

Our problem now is to show that for every positive integer n, we can find an integer in with m > n
such that
(m - m)2 + (in — n + 1)2 + (m - n + 2)2 + ∙ ∙ ∙ + ∣n2 = (m + 1)2 + (m + 2)2 + (m + 3)3 + ∙ ∙ ∙ + (m + n)2.

Concept: When working with an odd number of consecutive integers, try assigning
O—°d a variable to the middle integer and letting symmetry do some of your
work for you.

To show that such an integer m always exists, we try to solve this equation for m in terms of n. We
first write the equation more compactly with summations:
11 ∏
y,(tκ - ^)2=
k=0
y (m+^)2∙
k=l
n n
Expanding both squares gives ∑(m2 - 2ιnk + k2) = ^j(m2 + 2mk + k2), and breaking both sides of this
k=0 k=l
into several sums gives

Σπ,2-Σ 2mk + Yjk2 = ^jm2 + Yj2mk + Yιk2.

k=0 k=0 k=0 k=l k=l k=l
n n
Subtracting Σ m2 + Σ k2 from both sides, we get
k=l k=l
n n
m2 - y 2mk = y, 2∕7 ∕ . zc
k=0 k=l

(Make sure you see why we still have an m2 on the left side; the summation of m2 on the left side had
n + 1 terms, while the one on the right side had n terms, so one m2 is "left over" on the left. There's a
"leftover" k2 term, too, but that leftover term is for k = 0, which makes k2 = 0.)
n n _n n
Since 2mk is 0 when k = 0, we have y 2mk = y 2mk, so our equation now is m2 — y 2mk =
k=0 k=l
y 2mk,
k=l Jt=l
n n
which gives us m2 = 2 y 2mk. Factoring 2m out of the sum, we have m2 = 4m y, k, so
k=l k=l
n
ffl=4^⅛ = 4.⅛^=2π2 + 2n.
fc=l
n
Thus, we have shown that if ∑<m - k)2 = ^(m + k)2, then m = 2n2 + 2n. All of our steps are reversible,
k=0 k=l
so we have
n n
y (2n2 + 2n- k^)2 = ∑(2n2 + 2n + k~)2
k=0 k=l

565
CHAPTER 17. MORE SEQUENCES AND SERIES

for all positive integers n. Therefore, for any positive integer n, there exist n + 1 consecutive squares
whose sum equals the sum of the next n higher consecutive squares. □

Solution for Problem 17.21: Problems 17.16 and 17.17 have already paved the way for us. We note that
(k + 1)4 - fc4 = 4k3 + βk2 + 4k + 1 for all positive integers k, and we have
n ∏
∑(4k3 + 6fc2 + 4k + 1) = V [(k + 1)4 - k4] = (n + 1)4 - 1 = n4 + 4n3 + 6n2 + 4n.
fc=l k=l
We also have
n n n n n
yι (4p+βk2+4fc+1)=4 yι a3+6 y k2+4 yι k+yι
k=l k=1 fc=l k=l k=l
( n ∖
=4 Σ
* 3 + n(n + l)(2zz + 1) + 2n(n + 1) + n
∖fc=l
∖
7
fn
=4 + 2zz3 + 5n2 + 4n.
∖fc=l 7
n ∖
We therefore have 4 ∕c3 + 2zz3 + 5n2 + 4π = n4 + 4κ3 + βn2 + 4n, so
Uk=l 7
n
Σ^= rz4 + 2n3 + n2 n(n + 1)∖2
4
fc=l

xeι< ises
l.
50
17.3.1 Evaluate Jjz'3 - 2z2).
ι=l

17.3.2 Find the sum of the series 1 ∙ 2 + 3 ∙ 4 + ∙ ∙ ∙ + (2n - 1) ∙ 2h.

n
17.3.3 Let g(x) = x2 - lh - 28. Find the positive integer n such that Σ gW = 0.
fc=l
n
17.3.4 Find a closed form for k4.
k=l

*
17.3.5 Compute the value of

Mo2÷^)÷⅛o2÷^+3η+l(^+^÷3¼4η+.→⅛5(l≈+2≡+...÷60≡).
(Source: ARML) Hints: 339

566
 17.4. ARITHMETICO-GEOMETRIC SERIES

17.4 Arithmetico-Geometric Series

We know how to handle arithmetic and geometric series. In this section, we see what happens when we
combine an arithmetic series with a geometric series of the same length by multiplying each term in the
arithmetic series by the corresponding term in the geometric series, then summing all these products to
form what we call an arithmetico-geometric series.
∣ ill- ___ . _ ____ - -
Problems

Problem 17.23: Determine the value of the infinite product

21∕3.41∕9 . gl/27.161∕81 ...

(Source: Mandelbrot)
∞
Problem 17.24: Let r be a constant such that ∣r∣ < 1. Find a closed form for krk.
k=0

11 14
Problem 17.25: Evaluate the sum - + — + 43 + 44
4 4* 2

Problem 17.26: Find a closed form for 1 + 4x + 9x2 + ∙ ∙ ■ + (π + l)2x" + ∙ ∙ ∙, where ∣x∣ < 1.

Problem 17.22: Evaluate the series 1 ∙ 20 + 2 ∙ 21 + 3 ∙22 4---- + n∙2" 1 4------ F 11 ∙ 210.

Solution for Problem 17.22: This looks a lot like a geometric series, so we try to evaluate it the same way
we tackled geometric series.

Concept: You know what we're going to say here. Compare new problems to similar
Q-1^) problems you already know how to do. The tactics that worked on the
similar problems might solve the new problem, too.

We let
S = 1 ∙ 20 + 2 ∙ 21 + 3 ∙ 22 + ∙ ∙ ∙ + n ∙ 2π~1 + ∙ ∙ ∙ + 11 ∙ 210,
then multiply this equation by 2 to get the series

2S = 1 ∙ 21 + 2 ∙ 22 + 3 ■ 23 + ∙ ∙ ∙ + 11 ∙ 211.

567
CHAPTER 17. MORE SEQUENCES AND SERIES

Subtracting the first equation from the second one, we get

10
210 + 11 ∙2n = 11 ∙211
k=0

The sum subtracted at the end is a geometric series, so we can now finish computing S:

io
= 11 ∙ 211 - (211 - 1) = 10 ∙ 211 + 1 = 20481.
k=0

□
When we multiply each term in an arithmetic series by the corresponding term in a geometric
series, we form an arithmetico-geometric series. For example, the series in Problem 17.22 results from
multiplying each term in the arithmetic series l + 2 + 3 + ∙∙∙ + llby the corresponding term in the
geometric series 20 + 21 + 22 + ∙ ∙ ∙ + 210. The result is the arithmetico-geometric series

1 ∙ 20 + 2 ∙ 21 + 3 ∙ 22 + ∙ ∙ ∙ + 11 ∙ 210.

The common ratio of the geometric series is also called the common ratio of the resulting arithmetico-
geometric series. For example, the common ratio of the series in Problem 17.22 is 2.

Concept: Arithmetico-geometric series can be evaluated using the methods we

o≡≡=) developed for solving geometric series. We multiply or divide the
arithmetico-geometric series by its common ratio to create a new series.
We then evaluate the difference between the two series.

As we've mentioned many times, our methods are more important than our formulas. Indeed, way
back in Section 10.4, we noted that the method for evaluating geometric series is more important than
the formulas we derived for geometric series. Here's yet another problem on which we can use this
method:
Problem 17.23: Determine the value of the infinite product 21/3 ∙ 41/9 ∙ 81/27 ∙ 161/81 ∙ ∙ ∙ . (Source: ∣
Mandelbrot) ___ ________

Solution for Problem 17.23: We first write each term as a power of 2:

The product then equals 2s, where

12 3 4
S= — 4^ — 4“ — 4- — 4"
3 32 33 34
That's an arithmetico-geometric series! We know how to tackle this. We multiply this series by 1/3 to

S 12 3
— zχ -""" 4- — 4- — 4-
3 32 33 34

568
 17.4. ARITHMETICO-GEOMETRIC SERIES

Subtracting this from our expression for S above gives

2S = 1 2 J_ 3 _ 1
3 3 + 32 + 33 + ‘'' ^ Γ∑l - 2'
3

so S = 3/4. Therefore, our product equals 23/4 = y∕8.

Even if you didn't remember our method for deriving the formulas for geometric series, you still
have another option for finding S above. We see that it resembles a geometric series with ratio 1/3. So,
we start by writing such a series:
111
3 + 32 + 33
To get S, we need one more 1∕32 term, 2 more 1∕33 terms, 3 more 1 ∕34 terms, and so on. This gives us
an idea; we can write all these powers of 1 /3 in a grid:

+ ⅛ + +
1 + ± 1
3 + 32
34
1
32
+ ⅜ + 1
34
+
1
33
+ 1
34
+
1
34
+

Adding the columns of our grid gives us S, as there are k copies of l∕3fc for each positive integer k. The
rows of the grid form an infinite collection of infinite geometric series!
We have:
1
1 X + .. _ 3 _ 1
+ ⅜ ÷
33
+ 34 -t-
1-3 - 2
1
1 1 1 32 _ 1 1
+ + ÷ ∙ ∙
32 33 34 “ 1-1 2,3
i 3
J_
1 1 — 33 1 1
+ ÷ ∙ ∙
33 34 14 2,32
_1_
1 _ 34 1 1
+ ∙ ∙
34 14 “ 2 * 33

So, the sums of the rows form another geometric sequence. When we add these sums to form S, we get
1 11 1 _1_ 1 J_ = 1 1 _ 3
^2 + 2 - 3 + 2'32 + 2 - 33 +"'^2'i-l - 4‘
1 3

Now, we're ready to tackle a general infinite arithmetico-geometric series.

∞
Problem 17.24: Let r be a constant such that ∣r∣ < 1. Find a closed form for *J~ ι M.
k=0

569
CHAPTER 17. MORE SEQUENCES AND SERIES

Solution for Problem 17.24: We again use our tactic for summing a geometric series. Letting the sum be
∞ 00

S, we have rS = r ^rk ~ T ■ Subtracting this from our expression for S gives

k=o k=0

00 ∞
S-rS = ∑kιjc -∑kιjc+1.
k=0 k=0

We want to group the like powers of r, so we shift the index of the second summation by 1 to give
00 oo

S-rS = ∑krk-∑(k-V)rk.
k=0 k=l

We also note that the k = 0 term of the first sum equals 0, so we can start that sum at k = 1 as well, and
we have ∞
∞ 00 00 00

S-rS = ∑kS- ∑(k - l)√c = ∑kt

*
- ∑(fc - 1)∕ = Jjfc/ -(k- 1)∕).
k=0 fc=l fc=l fc=l ⅛=1

Since krk - (k — l)rk - rk, our final summation is just an infinite geometric series with first term r and
common ratio r: OO ∞
S-rS = ∑fkrk -(k- l)rfc) = ∑τk =
k=l k=l
r
Therefore, we have S = —----- -τ. □
(1 - r)2
What if the arithmetic series doesn't have common difference 1?
5 8 11 14
Problem 17.25: Evaluate the sum - + ^ + ^3+^ + ∙∙∙∙
9M

Solution for Problem 17.25: We'll be pretty surprised if our "multiply by r" tactic doesn't work. We let S
be the sum, and we multiply by 1/4 to get

S _ _5 8 11 14 + ...
4 ^ 42 + 43 44 45

Subtracting this from the original series gives

3.
zπ S 5 3 3 3 5 42 _ 5 1 _ 3
S — 7 = 7 + ∙τ7 + ττ + 44 + ’ ’ ‘ ~ 4 + 1—1~4 + 4~2'
4 4 42 43 1 4

So, we have 3S∕4 = 3/2, which means S = 2. □

OO oo

We know how to sum P, rk and ^or ∣r∣ < 1∙ Y°u can guess what's next.
k=0 k=0

Problem 17.26: Find a closed form for 1 + 4x + 9x2 + ∙ ∙ ∙ + (π + 1)2λ" + ∙ ∙ ∙, where ∣x∣ < 1. I

570
 17.5. FINITE DIFFERENCES

Solution for Problem 17.26: And you can guess how we,ll attack it. We begin by setting a variable equal
to the whole series:
S = 1 + 4x + 9x2 + ∙ ∙ ■ + (∏ + l)2xn + ■ ∙ ∙ .
Multiplying this series by x, we have

xS = x + 4x2 + 9x3 + ∙ ∙ ■ + ∏2xn + ∙ ∙ ∙ .

Subtracting the second equation from the first, we get

S(l - x) = 1 + 3x + 5x2 + ∙ ∙ ∙ + (2π + l)χ" + ∙ ∙ ∙,

which is an arithmetico-geometric series!

Continuing with more of the same, we multiply our latest equation by x:

Sx(l - x) = χ + 3x2 + 5x3 + ∙ ∙ ∙ + (2n - l)χ" + ∙ ∙ ∙ .

Subtracting this equation from the previous one, we get

S(l - x)2 = 1 + 2x + 2x2 + 2x3 + ∙ ∙ ∙ + 2xn + ∙ ∙ ∙ ,

which is 1 more than a geometric series. Summing the geometric series then gives us

x+1
S(l - x)2 = 1 + —
1- 1 - x'
x+1
so S =
(1 - X)3
Another triumph of methods over formulas!

Exercises

17.4.1 Find the sum of each of the following arithmetico-geometric series:

(a) - + -÷-÷- + ... (b) ⅛(⅛-l)3^ (c)

k=1 n=l

17.4.2 Evaluate: log2 V2 + log2 λ^4 + log2 λ∕8 + ∙ ∙ ∙ + log2 2√2" + ∙ ∙ ∙ .

17.4.3 Find a closed form for the series 1 + 2x + 3x2 + ∙ ∙ ∙ + nxn~1 for x ≠ 1.

17.5 Finite Differences

Given a sequence ao, a^, a2, (⅛, ..., we can construct a table as follows. First, write the terms of the
sequence:
«0 fll O2 ∏3 ■ ■ ∙

571
CHAPTER 17. MORE SEQUENCES AND SERIES

Next, write the differences between consecutive terms under the original sequence:

Uq Ctγ Cl2 a3
$1 — $2 “ βl β3 “ a2

To make things easier, we introduce some notation. We let ∆απ = an+↑ - an for all n ≥ 0. (The symbol
Δ is the uppercase Greek letter delta.) We can then rewrite the table as follows:

«0 «1 #2 fl3
∆αo ∆αi ∆∏2

The next step, naturally, is to write the differences between consecutive terms of the sequence {∆απ}:

#0 Clγ ∕⅛ ^3
∆flo ∆flj Δfl2
∆Λ1 - ∆flo Δz72 -

Again, to make things easier, we denote the difference ∆απ+ι - ∆α,j by ∆2αn:

«o flι #2 fl3
Δ⅜ ∆flι Δfl2
Δ2α0 Δ26Zi

In this way, we can produce as many rows as we please:

«0 fll «2 «3
∆flo ∆flι Δfl2
∆2αo Δ¼
Δ3α0

The sequence ∆αoz ∆flι, Δfl2∕ ∙ ∙ ∙ is the "first difference sequence," the sequence Δ2(⅛, ∆⅛ι, Δ⅞2∕.. ■ is
the "second difference sequence," and so on. The table is known as the finite difference table for the
sequence flo,flι,tf2∕ ∙ ∙ ∙ ∙ By convention, the rows are labeled by the power of Δ, so the row containing the
sequence {an∖ is known as the 0th row, the row containing the sequence {∆απ} is known as the first row,
and so on.
In this section, we use finite difference tables to analyze polynomials. If p(x) is a polynomial, we
can construct a finite difference table for p(x) by starting with the sequence {p(n)}∞≈θ. The first difference
sequence is {∆p(π)}~ 0, where ∆p(π) = p(n +1) -p(n). The second difference sequence then is {∆2p(n)}~ 0,
where ∆2p(n) = ∆p(n + 1) - ∆p(π), and so on.

,roblems

Problem 17.27: Let an = n3. Construct the finite difference table for the sequence a0, n↑,U2,.... Do you
notice anything interesting?

572
 17.5. FINITE DIFFERENCES

Problem 17.28:
(a) Let ∕(x) = ∕7.v2 + bx + c, where a, b, and c are constants and a ≠ 0. Consider the sequence {∕(κ)}∞ θ.
Prove that Δ2∕(h) is constant and nonzero, but ∆f(n) is not constant.
(b) Let p(Λ-) be a polynomial with degree m, where m > 0. Show that ∆mp(n) is constant and nonzero,
but Δ"'-1p(π) is not constant.

Problem 17.29: Find all polynomials p(x) such that p(x + 1) - p(x) = 2x + 1 for all x.

Problem 17.30: In order to draw a graph of ∕(x) = ax2 + bx + c, a table of values was constructed. These
values of the function for a set of equally spaced increasing values of x were 3844, 3969, 4096, 4227,
4356, 4489, 4624, and 4761. Unfortunately, one of these values was determined incorrectly. Find the
incorrect value. (Source: AHSME)

Problem 17.27: Let an = n3. Construct the finite difference table for the sequence Λq, a↑r a2,....

Solution for Problem 17.27: We write the first several terms of the sequence, and then take the finite
differences:
an : 0 1 8 27 64 125 216
∆an : 1 7 19 37 61 91
∆2an : 6 12 18 24 30
∆3απ : 6 6 6 6 ...
∆4ar, : 0 0 0

We notice that the third row seems to consist of all 6s, and that all further rows seem to consist of
Os. Since we have only written down a portion of the table, can we be certain that this pattern continues
indefinitely?
We can, by explicitly computing the formula for ∆kan∙.

an = n3,
∆an = an+ι -an = (n + 1)3 - n3 = 3n2 + 3n + 1,
∆2an = Δan+↑ — ∆an = 3(n + 1)2 + 3(n + 1) + 1 — (3n2 + 3n + 1) = 6n + 6,
∆3an = ∆2an+1 - ∆2ar, = 6(n + 1) + 6 - (6n + 6) = 6,
∆4an = Δ3an+↑ - ∆3an = 6-6 = 0.

573
CHAPTER 17. MORE SEQUENCES AND SERIES

We see that the third row does indeed consist entirely of 6s, and that the fourth row consists entirely
of Os. Furthermore, once a difference table hits a row of all Os, then clearly all further rows are also all
Os. □
In Problem 17.27, we saw that if an = n3, then the sequence {∆2απ} is not constant, but the sequence
{Δ3⅜} is constant and nonzero. Let's generalize this observation by considering the difference table of a
general polynomial.

Problem 17.28: Let p(x) be a polynomial with degree m, where m > 0. Show that ∆'"p(n) is constant I
and nonzero, but ∆m-1p(π) is not constant.

Solution for Problem 17.28: We let

p(x) = cmxm + cm-1xm 1 + ∙ ∙ ∙ + cιx + co­
Building the whole difference table looks like a nightmare, but we can at least explore a little by finding
∆p(π). We find
∆p(π) = p(n +1) -p(ri) = cm(n + l)m + cm-ι(n + l)m^1 + ■ ∙ ∙ + c⅛(n +1) + <¾ - (cmnm + cm-ιnm 1 + ■ ■ ■ + c↑n + co).

We're certainly not interested in multiplying that all out, so we look back to Problem 17.27 for guidance.
There, we saw that for the cubic an = n3, we can express ∆απ as a quadratic in n. So, we wonder if ∆p(π)
similarly has degree 1 less than the degree of p(ri). Going back to our expression above for ∆p(n), we see
that the only two terms of degree m come from cm(n + 1)", and -cmnm. The degree m term of cm(n + l)m
is cmnm, so the degree m terms in our expression for ∆p(κ) cancel.
But what about the degree m - 1 terms? There are three:

The last two cancel, and we're left with cm(ff)nm 1. We know that cm is nonzero because p(n) has degree
m. So, ∆p(n) is a degree m - 1 polynomial.
By the exact same argument, we can show that deg∆2p(n) is 1 less than deg∆p(n). Continuing in
this vein, we can show with induction that ifk<m, then deg ∆fcp(n) = m - k. So, when k = m, we have
deg∆'"p(rz) = 0, which means ∆mp(ri) is a nonzero constant. Because deg∆m~1p(n) = 1, we know that
∆m-1p(κ) is nonconstant. □

Important: If p(n) is a degree m polynomial, then ∆"'p(π) is a nonzero constant, while

ψ Δ'"-1p(n) is linear.

Problem 17.29: Find all polynomials p(x) such that p(x + 1) - p(x) = 2x + l for all x.

Solution for Problem 17.29: From the given equation, we can deduce that ∆p(n) = In + 1. Therefore,
∆2p(π) = ∆p{n +1) - ∆p(n) = 2(n +1) +1 - (2π ÷ 1) = 2. Because ∆2p(n) is constant and nonzero, we know
that p(x) is a quadratic. So, we let p(x) = ax2 + bx + c, and we have
a(x + 1)2 + b(x + 1) + c - ax2 - bx - c = 2x + 1.

574
 17.5. FINITE DIFFERENCES

Simplifying the left side gives 2ax + a + b = 2x + 1. Equating the coefficients of x gives a ≈ 1, and equating
the constants gives b = 0. So, the polynomials p(χ) such that p(x + 1)- p(x) = 2x + 1 are p(x) = x* 2 + c, for
any constant c. □

Problem 17.30: In order to draw a graph of ∕(x) = ax2 + bx + c, a table of values was constructed.
These values of the function for a set of equally spaced increasing values of x were 3844, 3969, 4096,
422/, 4356,4489,4624, and 4761. Unfortunately, one of these values was determined incorrectly. Find
the incorrect value. (Source: AHSME)

Solution for Problem 17.30: Because ∕(x) is a quadratic, we know that {∆2∕(π)} is a nonzero constant
sequence. So, we may be able to detect the incorrect entry by constructing the difference table for /(x):

3844 3969 4096 4227 4356 4489 4624 4761

125 127 131 129 133 135 137
2 4 -2 4 2 2

Looking at the table, we suspect that the incorrect value is 4227. All the entries in the second row are
2, except the 4, —2, and 4, and the entry 4227 affects all three of these. Indeed, if we were to set all the
entries in the second row to 2 and rebuild the table, then the 4227 would change to 4225:

3844 3969 4096 4225 4356 4489 4624 4761

125 127 129 131 133 135 137
2 2 2 2 2 2

Therefore, the incorrect value is 4227.

But wait a minute! The problem only said that the given outputs were the result of equally spaced
inputs x. We weren't told that the outputs were ∕(0), ∕(1), ∕(2), and so on. Can we still use the difference
table as we did above?
Yes! We can transform the quadratic ∕(x) into another quadratic g(y) such that our outputs correspond
to g(0)∕ g(l)∕ g(2), etc.
Suppose our first input x is x = k, and that the x's are spaced d apart, where k and d are constants.
So, our erroneous original table has f(k) = 3844, f(k + d) = 3969, f(k + 2d) = 4096, and so on. Now we
see how to make g(y). We let g(y) = f(k + dy). Then, we have g(0) = f(k) = 3844, g(l) = f(k + d) = 3969,
g(2) = f(k + 2d) = 4096, and so on. Because f(x) is quadratic, f(k + dy) is quadratic in y (remember, k and
d are constants). Our difference tables above are therefore the difference tables for g(n), which we can
use as described to determine that 4227 was the incorrect value. □
Problem 17.31: Assume that x↑, x%, ■. , X7 are real numbers such that

x^[ + 4x2 ^t^ 9xβ + I6x4 + 25xς + 36x6 ^t 49xχ = 1,

4x] + 9x2 + I6x3 + 25x4 + 36x5 + 49x6 + 64x7 = 12,
9x] ÷ I6x2 ÷ 25x3 ^F 36x4 + 49x5 + 64x6 ^t^ 8lx7 = 123.

Find the value of 16xj + 25x2 + 36x3 + 49x4 + 64xs + 81x6 + WOx7. (Source: AIME)

575
CHAPTER 17. MORE SEQUENCES AND SERIES

Solution for Problem 17.31: We're not too thrilled with the idea of trying to solve the system of equations.
Moreover, with 3 linear equations in 7 variables, we won't even have a unique solution. Rather than
trying to solve the system and substitute into the desired equation, we look for a way to manipulate the
given equations to give us the expression we seek.
We first notice that the coefficients of the left sides of the given equations are perfect squares.
Moreover, we can model all three left sides at once with the expression
an = (n + l)2x1 + (n + 2)⅛ + (∏ + 3)2X3 + ∙ ∙ ∙ + (∏ + 7)2xχ∙

Then ao, a^, and «2 are the left sides of the three given equations. And «3 is the desired expression!
Furthermore, an is quadratic in n. We're given that ao = 1, a↑ = 12, and «2 = 123. So, now we can build
our difference table for the sequence of an.

We know that the ∆2tzπ row is constant, so after using 1, ∙ 1 12 123 334
12, and 123 in the top row to find 100 in the bottom row, we Δλ∏ : 11 111 211
can add another 100 in the ∆2αjl row and work backwards ∆2flπ : 100 100
to the top, finding a$ = 334. Therefore, we have

16x∣ + 25x2 ^t^ 36x3 + ∙ ■ ∙ + lθθx7 — fl3 — 334.

Exercises

17.5.1 A quadratic polynomial p(rz) satisfies p(0) = 1, p(l) = 2, and p(2) = 7.

(a) Use a difference table to find p(3).
(b) Use a difference table to find p(n).
17.5.2 Part of the difference table of a sequence an is as shown:

an ■ 1 * * *
∆απ : 2 * *
∆2απ : 4 *
Δ3 ⅜ : 8

Find Λι, «2/ and «3.

17.5.3 If ∆3p(κ) = 3n + 5, then must degp = 4?
17.5.4 For any sequence of real numbers A = a∖la2,a3,..., define ΔA as the sequence of numbers
r?2 - «1, «3 - ^2∕ «4 - ^3∕ ∙ ∙ ∙ ∕ whose nth term is αn+ι - ⅜. Suppose that all of the terms of the sequence Δ(ΔA)
are 1, and that «19 = «92 = θ∙ Find <h∙ (Source: AIME)

17.6 Summary

A recursive sequence is a sequence in which, after one or more initial terms, each term is defined as a
function of previous terms in the sequence. Such a sequence definition is called a recursion.

576
 17.6. SUMMARY

A series in which parts of each term cancel with parts of other terms in the scries is called a telescoping
series. Some products also telescope. One tool that is helpful in discovering and evaluating telescoping
series is partial fraction decomposition, through which we write a rational function as a sum of rational
functions with simpler denominators.
We can use telescoping series to find formulas for the sums of powers of consecutive integers, such
as:
n
n(n + 1)
Σ**=
k=l
2
n
n(n + 1)(2∏ + 1)
Σ* 2=
k=l
6
n
n2(n + 1)2
4
k=l

We form an arithιnetico-geometric series when we multiply each term of an arithmetic series by

the corresponding term of a geometric series of the same length, then sum all these products. We can
evaluate such series using similar tactics to those we used to derive formulas for geometric series.
Finite differences can be used to analyze the polynomial p(x) by first constructing the sequence
lp(π)l,7=o- We can ^^ιen construct the first difference sequence, {∆p(rz)}, where ∆p(n) = p(n + 1) - p(n).
From this, we can construct the second difference sequence, {∆2p(π)), where ∆2p(π) = ∆p(n + 1) - ∆p(>∕).
And so on.

Important: If p(n) is a degree m polynomial, then ∆'"p(n) is a nonzero constant, while

ψ ∆','~1p(n) is linear.

Problem Solving Strategies--------------------------------------------------- ------------

Concepts: ∙ We can often learn a lot about a sequence or series by comparing

θ each term to neighboring terms.

• To learn more about a recursion, compute the first few terms and look
for a pattern. Then, once you've found the pattern, use the recursion
to figure out why the pattern exists.

• When looking for useful relationships to help solve sequence prob­

lems (recursion problems in particular), keep an eye out for period­
icity.

• Shifting the index of a recursion can be an effective way to learn more

about the recursion.

Continued on the next page. . .

577
CHAPTER 17. MORE SEQUENCES AND SERIES

Concepts: . . . continued from the previous page

Q=g
• When trying to prove that a pattern exists, write a general form that
represents the pattern.

• There aren't many coincidences in mathematics. If two types of

problems consistently give the same answer, look for a reason why.

• Many sums involving fractions can be rearranged so that they tele­

scope.

• Whenever you derive a general formula for a sequence or series, you

can usually test it with a few simple cases.

• Notation itself can be a powerful problem solving tool. Rewriting

some series with sigma notation can help evaluate them more easily.

• When a strategy gives you information about a problem, but doesn't

completely solve it, try using the strategy again. Maybe it has more
to reveal.

Review Problems

17.32 Let a-∣,a2,... be a sequence for which a↑ = 2,a2 = 3, and an — fr± for each positive integer n ≥ 3.
What is «2006? (Source: AMC 10)
17.33 Let tn be the κth term of an geometric sequence with first term a and common ratio r. Write a
recursion for the sequence {t,7}.
17.34 Let a and b be nonzero constants such that f(x) = ax+ b and a ≠ 1, and let n be a positive integer.
Express fn(x) in terms of a, b, and n.
17.35 On Halloween, Mrs. Pi tells each of the neighborhood children that they can take a handful of
candy, hi order to decide how much candy to take, each kid takes the average of the amount of candy
that the 3 kids immediately before her took. If this average is not an integer, she always rounds up. If
the first kid took 3 pieces of candy, the second kid took 2, and the third kid took 5, how many pieces will
the 2007th kid take?
17.36 For the sequence of numbers n↑, n2, n3,..., the relation ni = 2ni~↑ + a, where a is a constant, holds
for all i > 1. If n2 = 5 and ns = 257, what is «5? (Source: HMMT)
17.37 Find log10 ∣ + log10 j + log10 ∣ + ∙ ∙ ∙ + log10 ⅛.

17.38 Determine the value ofl∙2-2∙3 + 3∙4-4∙5d------ +- 2001 ∙ 2002. (Source: HMMT)
17.39 Simplify the product

(Source: AHSME)
(l-⅛)H)∙∙∙H)(t-⅛)∙
578
 CHALLENGE PROBLEMS

17.40 Compute each of the following:

2 3 π
v , 7 49 7n*^ 1
o>)∑⅜3
n=l

17.41 Compute the unique positive integer n such that 2 ■ 22 + 3 ∙ 23 + 4 ∙ 24 + 5 ∙ 25 + ∙ ∙ ∙ + n ∙ 2n = 2"+10.

(Source: ARML)
r , , (19982 - 19962)(19982 - 19952) ∙ ∙ ∙ (19982 - 02) .z, . ., n
17.42 Compute the product ------- ----------- --------- ----------------- --------------- -. (Source: Mandelbrot)
f (19972 - 19962)(19972 - 19952) ∙ ∙ ∙ (19972 - 02) '
20
17.43 Evaluate ^(2fc2 - k).
fc=l
17.44 Define a sequence of numbers by an = 3n2 + 3n + 1, so that π1 = 7,a2 = 19, a2, = 37, and so on.
Calculate a↑ + a2 + ∙ ∙ ∙ + «ioo- (Source: Mandelbrot)
17.45 If 1 ∙ 1987 + 2 ∙ 1986 + 3 ∙ 1985 + ∙ ∙ ∙ + 1986 ∙ 2 + 1987 ∙ 1 = 1987 ∙ 994 ∙ x, compute the integer x.
(Source: ARML)

17.46 From the table shown, find the polynomial of least degree that expresses ι∕ in terms of x:

X 1 2 3 4 5
y 3 7 13 21 31

(Source: AHSME)
17.47 A cubic polynomial f(ri) satisfies ∕(0) = 5, ∕(1) = 4, ∕(2) = 17, and ∕(3) = 56. Find ∕(4).

Challenge Problems

17.48 For a finite sequence aι,a2,..., an of numbers, the Cesaro sum of the sequence is defined to be

Si + S2 + ∙ ∙ ∙ + Sn
n
where
S]ζ = Uγ + ∏2 + ∙ ∙ ∙ + U]i

for 1 ≤ k < n. If the Cesaro sum of the 99-term sequence fli,«2z ...,fl99 is 1000, what is the Cesaro sum of
the 100-term sequence l,a↑,a2,. ..,ayp. (Source: AHSME)
17.49 Let Po(x) = x3 + 313x2 - 77x - 8. For integers n ≥ 1, define P∏(x) = Pl∣-ι(x - n). What is the
coefficient of x in P20W? (Source: AIME) Hints: 136

579
CHAPTER 17. MORE SEQUENCES AND SERIES

17.50 A sequence of integers a∖, az, a$,... is chosen so that an = an-∖ - an-2 for each n ≥ 3. What is the
sum of the first 2001 terms of the sequence if the sum of the first 1492 terms is 1985, and the sum of the
first 1985 terms is 1492? (Source: AIME) Hints: 82
17.51 Consider all sets of two distinct positive integers less than or equal to 21. Find the sum of the
products of the elements in all such sets. (Source: Mandelbrot) Hints: 313,155

17.53 If f(n + 1) = (-1)"+1h - 2∕(n) for integers n ≥ 1, and f(l) = ∕(1986), find the value of the sum
∕(1) + ∕(2) + ∕(3) + ∙ ∙ ∙ + ∕(1985). (Source: ARML) Hints: 280
n
17.54 Find a closed form for *V —------1 - ----- -.
⅛Jk(k+l)(k + 2)
k=l
17.55 Determine the value of

(Source: USAMTS) Hints: 268

17.56 Find the value of -τ------- F -∑------- 1- —-------t- ∙ ∙ ∙. (Source: HMMT)

32 + 1 4 2 + 2 5 2 + 3
n
17.57
* Find a closed form for V ————-. Hints: 300
A-J k4 + k2 + 1

17.58 A sequence of numbers a-i, az, a3, ... satisfies «1 = 1/2 and ai + az + ∙ ∙ ∙ + an = n2an for all n ≥ 1.
Determine the value of an for n ≥ 1. (Source: CMO) Hints: 17
17.59 Let m be a positive integer, and let α0, a↑,..., am be a sequence of real numbers such that a0 = 37,
α1 = 72, am = 0, and ak+1 = ak~↑ - ½ for k = 1,2,..., m - 1. Find m. (Source: AIME)

*
17.60 Evaluate the infinite product ■q—j-. (Source: Putnam) Hints: 139

17.61
* Let x0 = xι = 1 and xn = 4xw-ι - xn-2 for n ≥ 2. Prove that xπ-1xπ+1 - x2 = 2 for all n ≥ 1.
Hints: 135

*
17.62 Find the sum ——γ + —- + —- + —- + ∙ ∙ ∙ . (Source: HMMT) Hints:
Hi 61

*
17.63 Let Fq = 0,Fχ = 1, and Fn = Fn~ι + Fn-2. Find the value of the infinite sum

112 Fn
---- I------ 1-------- ∣-∙∙∙H------ +

(Source: Mandelbrot) Hints: 140

580
There is always inequality in life. - John F. Kennedy

ι
CHAPTER

L_ More Inequalities

18.1 Mean Inequality Chain

We've already studied the Arithmetic Mean-Geometric Mean Inequality, which states that the arithmetic
mean of a sequence of nonnegative numbers is greater than or equal to the geometric mean of those
numbers. In this section, we introduce two more means, and find relationships between them and both
the arithmetic mean and the geometric mean.
First, we'll define the two new means:

Definition: The quadratic mean of a set of real numbers is the square root of the arithmetic mean
of the squares of the numbers. In other words, the quadratic mean of the real numbers «1, a2,..., an
is __________________
+ β2 + β2 + ... + fl2

n
The quadratic mean is also sometimes referred to as the root mean square.

Definition: The harmonic mean of a set of nonzero real numbers is the reciprocal of the average of
the reciprocals of the numbers. In other words, the harmonic mean of the nonzero numbers aχ, a2,
a$,..., an is
n
i +JL + ± + ... +±∙
a-[ (I2 an

The quadratic mean, root mean square, and harmonic mean are sometimes abbreviated QM, RMS,
and HM, respectively.

581
CHAPTER 18. MORE INEQUALITIES

Problems

Problem 18.1: In this problem, we show that the quadratic mean of the nonnegative numbers a↑, (⅛,
a3,..., a„ is greater than or equal to the arithmetic mean of the numbers.
(a) Show that the desired inequality is equivalent to κ(α∣ + a2-i---- + a2) ≥ (aι + a2 + ■ ■ ∙ + an)2.
(b) Use an inequality from Chapter 12 to prove that n(a2 + α∣ + ∙ ∙ ∙ + a21) ≥ (α1 + «2 3------ ∏ «n)2-
(c) Prove that

f?l + ∏2 ^* ‘ ’ + > llι + Ω2 + ''' + ∩n

n ~ n
Must this inequality still hold if some of the fl1∙ are negative? ∙

Problem 18.2: Let Xι, X2, ..., xn > 0. Prove the Geometric Mean-Harmonic Mean Inequality, which
states that
ι2
V
* ■ ∙ ∙ x∏ > η-------- ;------------------ r∙
— + — + —1- —
n χ2 χn
What is the equality condition?

Problem 18.3: Let «1, «2/ ∙ ∙ ∙ ∕ ⅝ be positive real numbers such that a↑ + a2 + ∙ ■ ∙ + an = 1. Prove that

11 1 7
— + — ÷ * ∙ ∙ + — ≥ n .

Problem 18.4: Let a, b, c, and d be positive real numbers. Prove that

a2 + b2 + c2 b2 + c2 + d2
+ ^, j— ≥fl + fe + c + t∕.
a+b+c b+c+d

a+b
Problem 18.5: Prove that ≥ y∣ab- ----- - for any positive numbers a and b. Hints: 342
~ a + b

Problem 18.6: Find the minimum possible value of

where x, y, and z are positive. Hints: 328,320

Problem 18.1: Prove that the quadratic mean of the nonnegative numbers α1, a2, (13,..., an is greater
than or equal to the arithmetic mean of the numbers. Must this inequality still hold if some of the a∣
are negative?

582
 18.1. MEAN INEQUALITY CHAIN

Solution for Problem 18.1: First, we write the inequality we wish to prove,

We avoid the square root by considering the squares of both sides. We wish to show that

fll + fl2 + " ’ ÷ fln > /«1 + fl2 + ∙∙' + ⅜

n
Both sides of this inequality are nonnegative. Therefore, if we prove this inequality, we can take the
square root of both sides to establish the desired inequality.
Our new inequality has a sum of squares on the greater side and the square of a sum on the lesser
side. That sounds like the Cauchy-Schwarz Inequality. We multiply both sides by n2, both to get rid of
the fractions and to isolate the square of a sum on the lesser side:

n(∏^ ÷ ' ’ ^t fl∏) ≥ (βl + fl⅛ ^F ‘'' ^t an)2∙

Now we can see the similarity between this and the Cauchy-Schwarz Inequality. But the Cauchy-
Schwarz Inequality involves two sequences. It states that

(a2 + «2 + ,∙∙ + ^t)(b2 + b2 + ∙∙∙ + b2)≥ (a↑b↑ + a2b2 + ∙∙∙ + anbnj2.

Aha! We need an n on the greater side, and we want the bj's to disappear on the lesser side. Letting
½ = ½ = ∙ ∙ ∙ = b∏ = 1 in the Cauchy-Schwarz Inequality gives us

(a2+a2 + -∙- + a2)(n) ≥(a1+a2 + -∙∙ + an)2,

as desired. Once again, we've found our proof working backwards, but we write the proof forwards:

Proof: The Cauchy-Schwarz Inequality gives us(a2 + a2 + ∙ ∙ ∙ + a2)(l2 +12 + ■ ∙ ∙ +12) ≥ (α1 + α2 + ∙ ∙ ∙ + ««)2,
$ so we have n(a2 + a2 +---- 1- «2) ≥ («1 + (⅛ + ∙ ∙ ■ + an)2. Dividing both sides by n2 gives

fl2 ÷ S∣ + ‘ ‘- ÷ > (α1 + ∏2 ÷ , ∙ ∙ + flπ)2

Because both sides are nonnegative, we can take the square root of both sides to give

n n

From the equality condition of the Cauchy-Schwarz Inequality, we see that equality holds if
and only if a↑/1 = tz2∕l = «3/1 = ■, - = an∕^-', that is, if all the a∣ are equal.

If some of the αi∙ are negative, then the sum flχ + α2 + ∙ ∙ ∙ + an might be negative. If a↑ + α2 H F at∣ ≥ 0,
then our proof above is valid, and shows that the arithmetic mean of the αi∙ is no greater than the quadratic
mean of the α,∙. If a↑ + α2 + ∙ ∙ ∙ + an <0, then the square root of («i + α2 + ∙ ■ ∙ + an)2 is -(«1 + «2 + ∙,, + ⅜)∙ This

583
CHAPTER 18. MORE INEQUALITIES

means our proof above doesn't work if a↑ + a2 + ∙ ∙ ∙ + an <0. However, if the sum of the Λl∙ is negative,
then their quadratic mean is positive, but the arithmetic mean is negative, which means the quadratic
mean is greater than the arithmetic mean. So, the inequality still holds if some of the α1∙ are negative. □
The inequality we proved in Problem 18.1 is sometimes referred to as the Quadratic Mean-Arithmetic
Mean Inequality, or QM-AM.
One key step in our solution was recognizing that the sum of squares and the square of a sum suggest
using the Cauchy-Schwarz Inequality.

Concept: Much of success in proving inequalities is recognizing expressions that are

C‰'ljiJ part of well-known inequalities such as the Cauchy-Schwarz Inequality or
the AM-GM Inequality.

This will be a common theme throughout this section, and you'll have plenty of practice with it when
you tackle the problems at the end of the chapter. When stuck on an inequality, ask yourself, "Where
have I seen an expression like this before?"
Recognizing the sum of squares as a clue to try Cauchy wasn't enough to solve Problem 18.1. We also
had to figure out how to use the Cauchy-Schwarz Inequality, which is an inequality for two sequences,
to prove an inequality involving only one sequence. The led us to a useful Cauchy-specific strategy:

Concept: To derive some inequalities involving a single sequence of real numbers,

θ--. ije) we can set one of the sequences in the Cauchy-Schwarz Inequality equal
to constants.

In fact, as we saw above, the QM-AM Inequality really isn't something new. It's just a specific
application of the Cauchy-Schwarz Inequality in which we let one sequence be a sequence of l's.

Problem 18.2: Let Xi, X2, ∙ ∙ ∙, x∏ > 0. Prove the Geometric Mean-Harmonic Mean Inequality, which
states that n
Vλ'l*2 ∙ ∙ -¾ ≥

What is the equality condition?

Solution for Problem 18.2: We don't know much about the harmonic mean, but we do know how to relate
the geometric mean to the arithmetic mean. The right side is the reciprocal of the arithmetic mean of the
numbers ..., y-. Applying AM-GM to these positive numbers gives
±*1 + ± + ... + X
χn
X2______×n > - 1
n Vxlχ2 ■ ∙ ∙ xn ‘

— + — + ••• + — 1
Both sides of the inequality ———---------- — ≥----------------- are positive, so taking the reciprocal of both
n Vχ1χ2∙∙∙⅞
sides reverses the inequality sign and gives us the desired

584
 18.1. MEAN INEQUALITY CHAIN

We have equality when equality holds in our application of AM-GM, which is when all the l∕xi∙ terms
are equal. Therefore, equality holds if and only if all the x, are equal. □
Putting this together with the result of the previous problem, we have the Mean Inequality Chain,
which is also known as the QM-AM-GM-HM Inequality:

Important: Let xι, X2,..., x∏ > 0, and let QM, AM, GM, and HM denote the quadratic

φ mean, arithmetic mean, geometric mean, and harmonic mean of these n

numbers, respectively. Then, we have

QM ≥ AM ≥ GM > HM.

For each inequality sign, equality occurs if and only if all of the xl∙ are
equal.

The AM-GM Inequality is the most commonly applied part of this inequality chain, but inequalities
involving sums of reciprocals or sums of squares sometimes call for the harmonic mean or quadratic
mean, respectively.

Problem 18.3: Let «1, az,..., an be positive real numbers, such that a↑ + a2 + ■ ∙ ∙ + att = 1. Prove that

11 1 2
—+—+ +— ≥n .
Uγ CI2

Solution for Problem 18.3: The sum of the reciprocals suggests comparing the harmonic mean of the
numbers to some other mean. So, we manipulate the given inequality to isolate the HM:

Concept: If you suspect a possible application of the QM-AM-GM-HM inequality

O≡^ chain, try isolating one of those means in your desired inequality.

The right side is the harmonic mean of the α1. But what about the left side? We know that the HM is
no greater than any of the other means. Which other mean should we compare it to?
Both the on the greater side and the fact that a↑ + a2 + ■ ■ ■ + an = 1 suggest using the AM-HM
Inequality, which gives us
÷ U2 ÷ * * ’ ÷ &n >
n

Because a-[ + az + ■ ■ ■ + an = 1, we have

-1 ≥
n

which we can rearrange to the desired inequality. □

585
 CHAPTER 18. MORE INEQUALITIES

Problem 18.4: Let a, b, c, and d be positive real numbers. Prove that

I
1 a2 + b2 + c2 a2 + b2 + d2 a2 + c2 + d2 b2 + c2 + d2
------- -------- +-------- -,---- — +------------ — + —--------- — ≥a + b + c + d.
a+b+c a+b+d a+c+d b+c+d

Solution for Problem 18.4: The sums of squares in the numerators on the left side of the inequality suggest
comparing quadratic means to some other mean(s). In particular, the sums of variables on the right side
and in the denominators of the left side make the QM-AM Inequality a prime candidate. So, we take a
look at the QM-AM Inequality for three variables:

Both sides are positive, so we can square both sides to get an equivalent inequality without radicals:

a2 +b2 + c2 (a + b + c)2
3 ≥ 9 ‘

Rearranging this inequality to fit the fractions on the left side of the inequality we're trying to prove
gives us
fl2 + b2 + c2 a+b+c
-------------- ≥------------ .
a+b+c 3
Likewise, we have similar inequalities for all the terms on the left side of the desired inequality:

fl2 + b2 + d2 fl + b + d
>
a+b+d 3 '
fl2 + c2 + d2 fl + c + d
>
a+c+d 3 '
b2 + c2 + d2 b+c+d
>
b+c+d 3 ‘
Summing these four inequalities, we have the desired result. □
Notice that in our solutions to the last two problems, we aren't just blindly trying to apply different
inequalities in a search for something that will work. We use the forms of the expressions in the
inequalities we wish to prove as a guide. When we see a sum of reciprocals, we focus on inequalities
involving the harmonic mean. When we see a sum of variables, we focus on the arithmetic mean.
Products suggest the geometric mean, and sums of squares suggest the quadratic mean or Cauchy-
Schwarz.
And if we're at a loss for anything to try, we can rearrange the desired inequality, hoping to find a
familiar form:

a+b r~r 2 . . . ,
Problem 18.5: Prove that ■ ■ ≥ vflb - τ---- - for any positive numbers a and b.
ι 2 -a + τb

Solution for Problem 18.5: Although all four of the means in the Mean Inequality Chain are present,
it's not at all clear how we can use the Mean Inequality Chain to prove this inequality. So, we start by

586
 18.1. MEAN INEQUALITY CHAIN

rearranging the inequality. We put the radicals on one side and the other means on the other side. This
gives us
∣a2 + b2 i— a + b 2
∖ 2 ~yβ,i 2 ->+l∙
a b
Aha! We see the QM on the greater side and the AM on the lesser side. Maybe if we start with QM ≥ AM,
we can get to this desired inequality. Let's see what we need to show to accomplish this. The QM-AM
Inequality gives us
la2 + b2 a+b
V 2 “ 2 '
To get to the desired inequality, we must show that

since adding this to the QM-AM Inequality above gives us the desired inequality. But this inequality
that we want to prove says that -(GM) ≥ -(HM), and multiplying this by -1 gives GM ≤ HM. Uh-oh.
This last inequality isn't even true! So, if we start with QM ≥ AM, we'll be stuck, since it's impossible
to show that —(GM) ≥ —(HM), which is what we need to combine with QM-AM to get the desired
inequality.

Back to the drawing board. We wish to show that

Because QM ≥ GM and AM ≥ HM, both sides are nonnegative. So, we can square both sides of the
inequality. Before doing so, we simplify the lesser side:

a+b 2 _ a+b 2ab _ (a + b)2 - 4ab _ a2 — 2ab + b2 _ (a - b)2

^^2 1 + J = ^^2 ^a + b ~ 2(α + b) - 2(α + b) = 2(α + b)
*

(α - b~)2
Now, our desired inequality is Squaring both sides gives us
2{a + b)

0* - ft)4
4(α + b)2'
This is a little scary at first, but when we combine the terms without radicals on the left, we have

a2 + b2 , a2 + 2ab + b2 (a + b)2
—77— +ab =------- - ------- = —-—.
2 2 2
That's a lot simpler. We can also simplify the radical term and write our desired inequality as

(a + b)2
2

587
CHAPTER 18. MORE INEQUALITIES

The radical term is now the geometric mean of 2ab and a2 + b2. That's interesting, so we isolate it, and
we have
⅛J⅛-
2
<l,-.⅜> √v wv 2 + ⅛2),7
4(fl + b)2
Now, we combine the terms on the greater side, hoping to be able to simplify it. We find

(fl + b? _ (a - b)4 _ 2(fl + fr)4 _ (fl - ⅜)4 = 2(fl + fr)4 - (fl - b)4
2 4(fl + b)2 4(fl + b)2 4(fl + b)2 4(fl + b)2

Rather than multiplying out the numerator, we note that we can apply the difference of squares factor­
ization to yield

2(fl + b)4 - (a - b)4 _ (fl + ⅛)4 + (fl + fr)4 _ (fl - fr)4

4(fl + b)2 4(fl + b)2
(fl + b)4 + ((fl + ⅛)2 - (fl - ⅛)2)((fl + fe)2 + (fl ~ fr)2)
4(fl + b)2
_ (μ + bf + (4fl⅛)(2fl2 + 2fr2)
4(fl + b)2-

Aha! This makes our desired inequality

(a + b)4 + 8flb(fl2 + b2) ∕ 9—τy-

-------- £ √2*÷ il2)∙

On the right, we have a geometric mean involving ab(a2 + b2). On the left, we have the sum of two
terms, one of which involves ab(a2 + b2). Maybe we can connect the two sides using AM-GM. Applying
AM-GM to (fl + b)4 and 8ab(a2 + b2) gives

(fl + &)4 + 8flfe(fl2 + b2) ≥ ^ + i,)4(8flh)(fl2 + ⅛2) = 2(fl + fr)2 √2flb(fl2 + b2).

Dividing both sides by 2(fl + b)2 gives us

(a + b)4 + 8flb(fl2 + b2) f- —-τ-

---- ---------- -≥√W 2÷⅛
which is the last desired inequality from above! All of our steps are reversible, so we can reverse our
steps above to complete our proof. □

Problem 18.6: Find the minimum possible value of

positive.
(ι+⅜) (ι+⅜) (ι+⅞)'where x' and z are ∣
Solution for Problem 18.6: In order to find the minimum of the expression, we try to prove an inequality
of the form

(1 + ⅞)(1+½)(l+⅛)≥someconstant'
588
 18.1. MEAN INEQUALITY CHAIN

and then show that the expression can, in fact, equal that "some constant" for some values of x, y, and z.
A natural tool to try is AM-GM, since we have three sums on the larger side of the inequality. Applying
AM-GM to each gives

Therefore, we have

WARNING!! We're not finished. In a minimization or maximization problem, we

X must show that the claimed minimum or maximum value can be
achieved.

It's not immediately obvious what choices of x, y, and z give us this minimal value. (This alone might
make us suspicious about our answer.) In order to have equality in our final nonstrict inequality, we
must have equality in each of the three applications of AM-GM. From the equality condition of AM-GM,
we must therefore have 1 = 1 = ⅛, and 1 = ⅛. Rearranging these gives x - 2y, y = 2z, and z = 2x.
Substituting for z in the second equation gives y = 4x, and substituting this into x = 2y gives x = 8x.
Uh-oh. That gives us x = 0, which is a big problem, since we can't divide by 0.
Back to the drawing board.

Solution 1: Focus on the equality condition. With a little experimentation, we might guess that the
expression is minimized when x = y = z. Of course, a guess isn't good enough, but maybe we can find
a solution starting from this guess.

If the expression is minimized when x = y = z, then we will have to build our inequality

(1+⅞)(1+⅛)(1+δ)asomeconstant

from inequalities that give us an equality condition of x = y = z.

Concept: We can sometimes use the equality conditions of inequalities as a guide to

θ≡7'u~J sol ving inequality problems.

589
CHAPTER 18. MORE INEQUALITIES

We start with the first factor, 1 + We already know that directly applying AM-GM to 1 and is
a loser, so we look for ways to apply inequalities such that x = y is the equality condition. Writing the
sum with a common denominator gives us ¾y-∙ We don't want to apply AM-GM to 2y and x, since that
will still give us an equality condition of x = 2y. But if we write 2y as y + y, our numerator then is a sum
of 3 terms that we are happy to have equal: - 2^+*.

Applying AM-GM to y, y, and x gives ≥ y∣y⅛, soy + y + x≥3 ^y⅛. Therefore, we have
1 + ⅜ = ~+2y+λ ≥ ~ 2y ~> where equality holds if y - y = x. Aha! We do the same to each of the other two
factors in our original expression, and we have

>
2y 2y 2y
y _ z+z+y > 3√∑⅛
2z 2z 2z ’
z x+x+z 3λ⅞⅛
>

Multiplying these three, we have

27 y∣y^x ∙ z2y ∙ x2z 27

8xyz 8 ’

Success! Equality is achieved in all three applications of AM-GM when x = y = z. A quick check reveals
that the expression (1 + 2y)(l + ⅛)(l + ⅜) equals y whenever x = y = z (and all three are nonzero).

Solution 2: Substitution. The expression we must minimize is complicated. Worse yet, it's not
even symmetric. If we exchange x and y, for example, the resulting expression is not the same as the
original expression. Simple, symmetric expressions are a lot easier to work with than complicated,
non-symmetric expressions. So, we make a substitution to simplify the expression we must minimize.
We let a - ^,b= and c = ⅛∙ This makes the expression we must minimize become

(1 +c)(l +fe)(l +c).

That's much easier to deal with. We also have abc = (∙g) (⅛) = so when we expand the product
(1 + π)(l + b)(l + c), we have

So, to minimize (1 + α)(l + b)(l + c), we must minimize a + b + c + ab + be + ca. Minimizing sums—that's
a job for AM-GM. What's wrong with this approach:

Bogus Solution: AM-GM gives us

a + b + c + ab + bc + ca £ = = √⅛ ’
6 √g

590
 18.1. MEAN INEQUALITY CHAIN

The trouble here becomes apparent when we investigate the equality condition. For equality to hold,
we must have a = b = c = ab = be = c∩. But this gives us a — b = c = 1, so x = 2y, y = 2z, and z = 2x,
which we've already seen gives us the impossible x = y = z = 0.
Instead, we address a + b + c and ab + bc + ca separately. AM-GM gives us

a+b+c
3
≥ -^(flb)(bc)(ca) = -7(obc)2 =

Multiplying these inequalities by 3 and adding them gives us

3 3 9
a + b + c + ab + bc + ca> - + - = -.
~ 2 4 4
Adding 1 + abc = ∣ to this inequality then gives us

27
∖+a + b + c + ab + bc + ca + abc > —.
- 8

We must still show that this minimum can be achieved. Both of our applications of AM-GM have the
same equality condition, a = b = c, which gives u≡ = ⅛ = ⅛∙ Applying one of the ratio manipulations
from Section 11.2, this chain of equal ratios tells us that

χ = V_ = _z_ = χ + y + z = ι
2y 2z 2x 2x + 2y + 2z 2’

Therefore, we have x = y = z as our equality condition. Indeed, when x = y = z, we have a = b = c =

and (1 + ∏)(1 + fr)(l + c) = y, as expected. □
Our second solution shows us a very powerful approach to difficult inequality problems:

Concept: When faced with a problem involving unwieldy expressions, try making
O≡≡^≡ substitutions for those expressions to simplify the problem.

You might be wondering why we end a section about the Mean Inequality Chain with two com­
plicated problems that we solve with plain old AM-GM. We do so because AM-GM is by far the most
commonly used of the inequalities regarding means. In fact, a vast majority of the inequalities you will
encounter can be proved with some combination of the Trivial Inequality, AM-GM, and the Cauchy-
Schwarz Inequality. Other inequalities will have their moments, but if you master these three, you will
be able to tackle most inequalities.

Extra! Some of the inequalities in the Mean Inequality Chain are specific cases of the Power
,l*l"* l"* ιll* Mean Inequality, which tells us that if x↑, X2,..., xn are positive and r and s are nonzero
numbers with r > s, then
1r
>
n

591
CHAPTER 18. MORE INEQUALITIES

Sidenote: Below is trapezoid WXYZ, with WX ∣∣ YZ.

Inside the trapezoid, we have constructed four segments that are parallel
to the bases:

• Q1Q2 is the segment that divides the trapezoid into two regions of
equal area.

• Aι is the median of the trapezoid, connecting the midpoints of legs

XY and WZ.

• G1G2 is the segment that divides the trapezoid into two similar trape­
zoids.

• H1H2 passes through the intersection of the diagonals of the trape­

zoid.

Note that Q1Q2 ≥ A1A2 > G1G2 ≥ H1H2∙ Can you find a geometric expla­
nation for the order of these lengths? Furthermore, suppose WX = a and
YZ = b. Find the four lengths above in terms of a and b. (As an additional
challenge, find a geometric construction of each of these segments using a
straightedge and compass.)

*
t'xer<J⅜e

18.1.1 Prove that if a, b, c are positive, then----- - + ------ +------- ≤ - - + - ÷ -

r a+b b+c c+a 2 ∖a b c
1 ι 1 ι 1 ∖
18.1.2 Prove that if a, b, c are positive, then----------- ≤ 2
r a+b+c a + b b + c c + a)
18.1.3 Which parts of the Mean Inequality Chain still hold even if some of the variables take on negative
values?

18.1.4 Find all real numbers 0 ≤ X3 ≤ λ⅛ ≤ x↑ ≤ 1 such that (1 - %ι)2 + (%ι - Λ'2)2 + (×2 ~ *3) 2 + x3 =

a+b a+b
18.1.5 Let a and b be positive. Show that . (Source: Mandelbrot)
~2~ ~2~

592
 18.2. THE REARRANGEMENT INEQUALITY

18.2 The Rearrangement Inequality

Problem 18.7: Suppose we have a game called Who Has Common Sense? The way the game works is
that there are 4 giant bags full of money. The first bag contains $100 bills. The second bag has $10
bills. The third has $5 bills, and the fourth has $1 bills. You are allowed to take 50 bills from any one
bag, and then 8 bills from another bag, and then 2 bills from a third bag, and then 1 bill from the last
bag. How can you make the most money?

Problem 18.8: Let x1 = -2,x2 = 2, and x$ = 3. Let y1 = -l,y2 = 3, and y3 = 5. Which of the following
expressions has the greatest value? Which has the least value?

χιyι + χ2y2 + χ3j∕3 xιyι + x2y3 + x3y2 xiy2 + x2yι + x3y3

χ1y2 + χ2j∕3 + χ3y1 X1y3 + λ'2y1 + X3y2 xιy3 + x2y2 + x3yι

Notice anything interesting? If so, create a few more examples of your own and test your ideas before
continuing.

Problem 18.7: Suppose we have a game called Who Has Common Sense? The way the game works
is that there are 4 giant bags full of money. The first bag contains $100 bills. The second bag has $10
bills. The third has $5 bills, and the fourth has $1 bills. You are allowed to take 50 bills from any one ⅛
bag, and then 8 bills from another bag, and then 2 bills from a third bag, and then 1 bill from the last
bag. How can you make the most money?

Solution for Problem 18.7: Clearly, you want as many $100 bills as possible, so you start by taking 50 of
them. Next, the best you can do is get as many $10 bills as possible, so you grab 8 of them. Then, you
take 2 of the $5 bills and 1 of the $1 bills, giving you a total of $5091. □

This intuitively obvious way to maximize your winnings is an application of the Rearrangement
Inequality. Before we get into the details of the Rearrangement Inequality, let's see what happens when
we throw in some negative numbers, too.

Problem 18.8: Let x1 = -2, x2 = 2, and x3 = 3. Let yι = -lz y2 = 3, and y3 = 5. Which of the following
expressions has the greatest value? Which has the least value?

xιyι + x2y2 + x3y3 xiyι + x2y3 + x3y2 xiy2 + X2y1 + X3J∕3

*ιj∕2 + X2y3 + X3y1 X1y3 + X2y1 + X3y2 xiy3 + x2y2 + x3y1

593
CHAPTER 18. MORE INEQUALITIES

Solution for Problem 18.8: We find:

xιyι + X2y2 + X3y3 = (-2) ■ (-1) + 2 ∙ 3 + 3 ∙ 5 = 23,
xiyι + x2y3 + X3y2 = (-2) ∙ (-1) + 2 ∙ 5 + 3 ∙ 3 = 21,
*1y2 + χ2y1 + X3y3 = (-2) ∙ 3 + 2 ∙ (-1) + 3 ∙ 5 = 7,
^1y2 + X2y3 + X3y1 = (-2)∙3 + 2∙5 + 3∙(-l) = 1,
χ1y3 + ^2y1 + χ3y2 = (—2) ■ 5 + 2 ∙ (—1) + 3 ∙ 3 = —3,
xιy3 + x2y2 + x3yι = (-2)∙5 + 2∙3 + 3∙(-l) = -7.
We have the largest result when we pair the two sequences off in the same order, multiplying the least
of the X{ by the least of the y∣, the middle of the x1∙ by the middle of the y,∙, and the greatest of the xi by
the greatest of the y1∙. We have the smallest result when we pair the sequences off in opposite orders,
multiplying the least of the x1∙ by the greatest of the y1∙, the middle of the Xj by the middle of the y,∙, and
the greatest of the x∣∙ by the least of the y1∙. □
After experimenting with a few more examples, we might think that this pattern holds with any two
sequences of numbers. Let's see.
Before diving into the next problem, we'll introduce a little vocabulary to make the problem easier to
discuss. We'll say that we take a “sequence-product" of two sequences of length n when we group the
terms of the sequences into n pairs, one term from each sequence in each pair, take the product of each
pair, and then sum all n products. For example, in Problem 18.8, we found all six sequence-products of
the sequences -2, 2,3 and -1, 3,5.
The term “sequence-product" is not standard; we introduce it to make discussing the Rearrangement
Inequality easier.
Problems

Problem 18.9: Consider all possible sequence-products of the two sequences xι, x2, x3,..., xn and yι,
ι∕2∕ y3, ∙ ∙ ∙∕ y∏, where x↑ ≤ x2 ≤ x3 < ■ ■ ■ < xn and y∖ ≤ y2 ≤ y3 ≤ ■ ■ ■ ≤ yn. In this problem, we show
that the sum x↑y↑ + x2y2 + χ3y3 + ∙ ∙ ∙ + ⅞y∏ is the maximum possible sequence-product of the two
sequences.
(a) Show that xγyι + x2y2 ≥ *1y2 + *2y1∙
(b) Show that Xγy2 + X2y1 + *3y3 ≥ χ1y2 + χ2y3 + χ3y1∙
(c) Show that if n = 3, then the maximum possible sequence-product is x∙1y1 + x2y2 + χ3y3.
(d) Use the results of parts (a) and (c) to show that if n = 4, then the maximum possible sequence­
product is X1y1 + X2y2 + X3y3 + Uy4∙
(e) Show that xιyι + X2y2 + χ3y3 + ∙ ∙ ∙ + χ∏yn is the maximum possible sequence-product of the two
sequences for all n.

n cι n 1
Problem 18.10: Let a↑, a2, a3,..., an be n different positive integers. Show that yι > yι -.
λ-=ι κ k=↑ κ

cr, b^, c^,

Problem 18.11: Prove that — + — ÷ — ≥ « + + c for all positive a, b, and c. (Source: Canada)
be αc αb

594
 18.2. THE REARRANGEMENT INEQUALITY

Problem 18.9: Consider all possible sequence-products of the two sequences x↑, X2, X3r ..., xll and I
yι, J∕2, }∕3∕ ∙ ∙ ∙, y∣i∕ where X] ≤ X2 ≤ ×3 ≤ ∙ ∙, ≤ x» and y1 ≤ y2 ≤ y3 ≤ ∙, ∙ ≤ JΛι∙ Show that the sum

*1y1 + *2y2 + X3J∕3 + ∙ ∙ ∙ + xnyn

is the maximum possible sequence-product of the two sequences.

Solution for Problem 18.9: We start by investigating the n = 2 case, hoping that this simpler problem will
give us some insight into the general problem.

Concept: Tackling simple versions of a problem can often provide a guide to solving
Q,—ctj ⅛e problem.

When n = 2, we have x↑, x2 and y↑, y2 as our sequences, where x↑ ≤ x2 and y∖ ≤ yι∙ Then, we wish to
prove that Xj 1/1+x2y2 ≥ xιy2 + x2y1. Moving all the terms to the left gives us Xjyι-X1y2+ X2y2-*2J∕1 ≥ θ∙
Factoring the left side gives

xiyι - xiy2 + x2y2 - x2y1 = (x2 - x1)(y2 - y1).

Because x2 ≥ Xj and y2 ≥ y↑, we have x2 — xχ ≥ 0 and y2 - yι ≥ 0, which means (x2 — X1)(y2 — yι) ≥ 0. We
can then reverse our steps above to show that X1y1 + X2y2 ≥ X∖yι + X2yi∕ as desired.
Therefore, the maximum possible sequence-product of the sequences is xιyι +X2y2 and the minimum
possible sequence-product is X∖y2 + X2y1∙
Let's take a look at the n - 3 case, so our sequences are Xi, X2, X3 and y1, y2, y3, where x↑ ≤ x2 ≤ X3 and
yι ≤ y2 ≤ y3∙ We wish to show that the maximum possible sequence-product of these two sequences is
χιyι + *2y2 + *3y3-
We first consider what happens when we swap two of the y1∙ by comparing X1y1 + x2y2 + χ3y3 to
X1y2 + X2y1 + X3y3. But this is the same as the 2-term sequences case that we already solved! We showed
that if Xi ≤ x2 and yι ≤ y2, then x↑y↑ + X2y2 ≥ X1y2 + *2y1∙ Adding x3y3 to both sides gives

xιyι + X2y2 + X3y3 > X1y2 + X2y1 + ^3y3∙

Similarly, we know that exchanging any two of the y,∙ in X1y1 +x2y2 +X3y3 will produce a sequence-product
that is no greater than x∖y∖ + X2y2 + X3y3∙
But what about sequence-products like X∖y2 + X2y3 + X3y{l We can't just swap a pair of y1∙ to
compare this to xiyι + X2y2 + X3y3∙ However, swapping y1 and y3 in xiy2 + X2y3 + X3y1 will give us
X1y2 + X2y1 + X3y3, and we know that this latter sequence-product is no greater than x1y1 + x2y2 + x3y3.
But which sequence-product is larger,

xιy2 + X2y3 + x3y∖ or x∖y2 + X2y∖ + X3y3rl

The first term in each is the same, so we must compare X2y3 + X3y1 to X2y1 + *
3y 3. But this is just our
2-term sequences case again! We have x2 ≤ X3 and yι ≤ y3, so

X2y∖ + X3y3 ≥ X2y3 + X3y1∙

595
CHAPTER 18. MORE INEQUALITIES

Adding Xi 1/2 to both sides gives

x1y2 + *2y1 + X3y3 > X1y2 + X2y3 + ^3y1∙
We already know that X1y1+X2y2+
*3y3 ≥ χ1y2+^2y1+Λ'3y3,so wehavex1y1+X2y2+T3y3 ≥ X1y2+
*2y3+X3y1∙
Similarly, we can show thatxιyι +X2j∕2+
*3y3 ≥ X1y3+X2y1 +X3y2∕ so we have shown that X∖y∖ +X2y2+Λ¾y3
is indeed the largest sequence-product we can form.
We used the 2-term case to tackle the 3-term case. Maybe we can similarly use the 3-term case to
prove the 4-term case, and then use the 4-term case to prove the 5-term case, and so on. In other words,
maybe we can use induction.
We take the 2-term case as our base case. We then assume that if Xi ≤ xz ≤ ∙ ∙ ∙ ≤ Xk and yι ≤
y2 ≤ ∙ ∙ ∙ ≤ y<t, then the largest sequence-product we can form of the Ar-term sequences {xz} and {yl∙} is
%ιyι + X2y2 + ∙ ∙ ∙ + Xk^k∙
We then look at the (k + l)-term case. Suppose Xi ≤ X2 ≤ ∙∙∙ ≤ ⅞+ι ar*d yι ≤ y2 ≤ ∙∙∙ ≤ yfc+ι∙
We wish to show that the largest sequence-product we can form of the sequences {x1} and {y1∙} is
yι + %2ψ2 + ∙ ∙ ∙ + Xjt+ιyk+ι∙ We do so by showing that every other possible sequence-product is no
greater than this one. In our proof of the 3-term case, we tackled separately the sequence-products with
X3y3 and those without X3y3. We do the same here.
Case 1: Sequence-products that contain Xk+ιPk+ι∙ Each such sequence-product has the form S + x⅛+ιy⅛+ι,
where S is a fc-term sequence-product of the sequences Xi, X2, ∙∙∙,Xk and yι, y2, ∙ ∙ ∙, Pk∙ By our inductive
assumption, each such sequence-product is no greater than X∖y∖ + X2y2 + ∙ ∙ ∙ + XkVk∙ Therefore, we have
X1yι + x2y2 + ∙ ∙ ∙ + Xk⅛k + Xk+1yk+1 ≥ S + Xk+iJ/k+1
for all possible S.
Case 2: Sequence-products that do not contain xfc+1y⅛+1. Each such sequence-product has an xfc+1y,∙ term
and an x7y⅛+ι term, where x7∙ ≤ x⅛+ι and yz∙ ≤ y⅛+ι∙ We can then write the sequence as T + X)⅛1y1∙ + XjPk+ι>
where T consists of the other k -1 terms in the sequence-product. Applying our 2-term case (that is, our
base case), we have x7∙yz + Xk+iVk+i ≥ Xjyk+ι + ‰ιy∕∙ Adding T to both sides, we have
T + x7yz∙ + Xk+1yk+1 ≥T + x∣yk+ι + ⅞+ιy1∙∙
The sequence-product on the greater side has an x⅛1yfc+1 term, so Case 1 tells us that it is no greater than
xiyi + X2y2 + ∙ ∙ ∙ + Xkyk + ‰ιy⅛+i∕ and we have
x↑yι + X2y2 + —h χkyk + ‰ιy⅛+ι ≥ T + X7∙y,∙ + Xjt+1y⅛+ι ≥ T + x7∙y⅛+1 + x⅛+ιyz∙.

Combining our two cases, we have shown that Xιy7 + X2y2 + ∙ ∙ ∙ + ⅞+1yk+1 is the largest possible
sequence-product of the (k + l)-term sequences {x, ) and {y1∙}, so our proof by induction is complete.
Therefore, the maximum possible sequence-product of the two sequences x↑, x2, X3, ∙ ■ ∙, x∣ι a∏d 1/1,1/2,
y3,..., yn is
x1y1 +X2y2 + ∙∙∙ +χny∏∙
As an Exercise, you'll have the chance to prove that
xiy„ + x2y,,-ι + λ'3½i-2 + ■ ∙ ∙ + x∏ι∕ι
is the minimum possible sequence-product of the two sequences. □

596
 18.2. THE REARRANGEMENT INEQUALITY

Important: Suppose we have sequencesι x↑, x2,

* * j. ------
x∏ and ∙∕ι, J∕2z
m
y∣ι, where
' t∕ ∙» r
v∕
* ---------
r⅜∕∙-'

Ψ *1 ≤ x2 ≤ X3 ≤ , ∙, ≤ xl∣ and ι∕j ≤ 1/2 ≤ J/3 ≤ ,, - ≤ V∏∙ If we group the

terms of the sequences into n pairs with one term from each pair, take the
product of each pair, then sum all n products, then the largest possible
sum is
xιyι + x2y2 + ■■■ + xnyn,
and the smallest possible sum is

λ∙iV∏ + x2yn-ι + ∙ ∙ ∙ + xny↑.

This is called the Rearrangement Inequality.

Another way to look at the Rearrangement Inequality is that the maximum sum is formed when we
pair off the terms of the sequences in order, and the smallest possible sum is formed when we pair off
the terms of the sequences in opposite order.
»1 r>

Solution for Problem 18.10: First, we note that we need only consider the sequences in which the a,
consist of the first n integers. If any a∣ς — s, where s > n, then there is some integer m with 1 ≤ m < n such
that none of the fl∣∙ equals m. We therefore have s∕fc2 > m∕∕c2, so the sum on the left side of the desired
inequality can be reduced by letting α⅛ = m rather than a∣~ = s. Therefore, any sequence that contains
numbers greater than n can be converted into a sequence with no terms greater than n such that the
n

that consist of the numbers 1 through n in some order, then we know it holds for all sequences.
If the c⅛ terms consist of the first n positive integers, we can view the left side of the inequality
as a sequence-product of the sequence 1, 2, 3, ..., n and the sequence ..., T∙. We might be

inspired to do this by noting that the right side of the inequality can be written as is also a

sequence-product of these two sequences.

We're comparing two sequence-products of the same two sequences—this is a job for the Rearrange­
ment Inequality! Because l<2≤3<∙∙∙≤n and ⅛ ≤ ≤ < ∙ ∙ ∙ ≤ ⅜, the Rearrangement
Inequality tells us that the minimum possible sequence-product isl∙^+2∙^ + ∙∙∙ + ∏∙-L. Therefore,
we have

as desired. □

u b c
Problem 18.11: Prove that — + — + — ≥ a + b + c for all positive a, b, and c. (Source: Canada)
be αc αb
■■■■■■■■■■■■■■■■■I

597
CHAPTER 18. MORE INEQUALITIES

Solution for Problem 18.11: We might first try the Mean Inequality Chain on this problem, but after trying
AM-GM for the sum on the larger side, or QM-AM for the sum on the smaller, we feel a little stuck.
The Rearrangement Inequality seems like a possible candidate because if we could change the
denominators on the left to a2, b2, c2 instead of be, ac, ab, then we would have the right side. In other
words, if we could take a3, b3, c3 as one of our sequences, then maybe we can choose another sequence
as our second sequence and get somewhere with the Rearrangement Inequality.
However, we have a problem with using the sequence α3, b3, c3. We don't know the order of fl3, b3,
and c3. We could break the problem into cases. First, we'd assume a > b ≥ c (so α3 ≥ b3 ≥ c3). Then,
we'd assume b≥a≥c. And so on.
But all these cases are essentially the same! Both sides of the inequality are symmetric, so if we prove
the inequality for the case a > b > c, then we know it's also true for b > a > c, because we know we can
just exchange a and b.
We have a special term for when we work with one case that is equivalent to all possible cases. We
say, "Without loss of generality, we let a ≥ b ≥ c." This means that we will consider the case a ≥ b ≥ c,
and that this is the only case we must consider because the other cases are equivalent. The phrase
"without loss of generality" is sometimes shortened to WLOG.
However, while "without loss of generality" can make writing proofs much easier, we do have to be
careful that we don't go astray, as we do below:

Bogus Solution: Without loss of generality, let a > b ≥ c. Because a ≥ b ≥ c, we have

a2 ≥ be, so ≥ -⅜. Multiplying both sides of this inequality by a3
3 _3 3 l3 Λ
gives us ‰ > ^2, so ‰ > a. Similarly, we have ≥ b and ‰ > c, and
adding these three inequalities gives the desired result.

Our "Without loss of generality" start is OK. Exchanging any two of a, b, and c doesn't make the
expression on either side of the desired inequality change, so if we prove the desired inequality for one
order of the variables, the same proof will hold for any order of the variables.
The problem comes with the "Similarly" step.

WARNING!! Be careful when you use "similarly" in a proof. You must be sure
*A4 that the conclusions that you state are "similarly" true can, in fact,
be proved by essentially the same steps. We particularly have to be
careful about this when starting with "without loss of generality."

/3 3
If we go through the same steps with as we did for ‰ in the Bogus Solution, we have a problem
right away. Starting from a > b ≥ c (we can't change the order of the variables for each part of the proof,
of course), we can't assume that b2 ≥ ac. So, we can't show that ⅛ ≥ ⅛∕ or that ‰ ≥b.

However, the ordering of a, b, and c does allow us to break out the Rearrangement Inequality!
Solution 1: Use the Rearrangement Inequality. Without loss of generality, we let a > b > c. So, we have
a3 > b3 ≥ c3. We need to multiply a3, b3, and c3 by ⅛, ±, and respectively. In other words, we need to

598
 18.2. THE REARRANGEMENT INEQUALITY

form a sequence-product, which points us to the Rearrangement Inequality. But first, we have to order
the terms ⅛, ⅜, and ⅛. Since ⅛ = α ∙ ⅛, ⅛ = & ∙ ⅛, and ⅛ ≡ c ∙ ⅛, and we have a ≥ b ≥ c > 0, we
conclude that ⅛ ≥ ⅛ ≥ ⅛.

We can now apply the Rearrangement Inequality to a3 ≥ b3 ≥ c3 and ⅛ ≥ ⅛ ≥ ⅛. The greatest

possible sequence-product is the greater side of our desired inequality. We compare this sequence­
product to one in which we pair each of the cubes with a denominator that allows us to do some
cancellation:
a3 b3 c3 a3 b3 c3 a**& b2
23 c2
—1—abT. + Γ^
ibe 1 ac , abii — ac c 1 a ,^ v∙
be = — b
Well, that didn't finish the problem off, but it left us with something much simpler to deal with. We now
must show that
a2 b2 c2
—1^--- 1—— ≥ Λ + b + c.
cab

Concept: If a tactic simplifies a problem but doesn't solve it, look for ways to apply
the tactic again.

Thisinequality looks a lot like theoriginal one, so we try the Rearrangement Inequality again. Since
a > b ≥ c > 0, wehave both a2 ≥ b2 ≥ c2 > 0 and ∣≥ ∣ ≥ > 0. So, the Rearrangement Inequality gives
us
a2 b2 c2 a2 b2 c2
+ — ^∣—— ≥ — + — + — — a + b + c,
c a b a b c
as desired.

If we don't find the Rearrangement Inequality solution to this problem, we can still fall back on
AM-GM.
3
Solution 2: Use AM-GM. In order to produce a from we need to multiply by be. AM-GM gives us
a way to build an inequality that will leave ∣∣ on the greater side, and to multiply by b and c to leave
a alone on the lesser side. Let's give it a try:

⅛ + fe + c

so we have
a3
-— + b + c ≥ 3λ.
be
Well, we have some of what we wanted. We have a3∕be on the greater side, and we have a on the lesser
side. But we have b and c on the greater side, which we don't want, and we have 3«, not just a, on the
lesser side. Let's see what we happens if we repeat the process for ∣∣ and Maybe we'll get lucky and
something good will happen. Applying AM-GM as above gives

&3
— + a + c ≥ 3b,
ac
c3
~ + a + b ≥ 3c.
ab

599
CHAPTER 18. MORE INEQUALITIES

Indeed, something good happened. Adding all three of our inequalities gives

+ — + ⅛ + 2(« + + c) ≥ 3(« + b + c),

be ac ab
which then gives us the desired inequality. □
Not only does our second solution show yet again how versatile AM-GM is, but it also shows a
general problem-solving approach we use with many inequalities. We've already seen how important
it is to look for familiar forms when proving an inequality, but our second solution takes a slightly
different approach:

Concept: Don't expect to be able to prove an inequality in one step. Use your
o⅛g inequality tools to produce pieces of the inequality you want to prove.
Look for ways to produce inequalities that have a term or two on the same
side of the inequality sign as in the inequality you must prove. As you
find these inequalities, you may then see ways to combine them to finish
the problem.

Exercises

„ 1 a+b+c 1 1 1
18.2.1 Let a,b,c > 0. Show that----------- ≤ — + -∑∙ + -z.
abc a2 b2 c2
18.2.2 Suppose that x1 ≤ X2 ≤ X3 ≤ ≤ xn and 1/1 ≤ 1/2 ≤ 1/3 ≤ ≤ ‰ Show that the sum
x1yπ + X2½1-1 + X3j∕n-2 + ∙ ∙ ∙ + x∏yι is the minimum possible sequence-product of the two sequences.
18.2.3 Let x and y be nonnegative real numbers. Show that x5 + y5 ≥ x4y + xy4 ≥ x3y2 + x2y3.
b3 c,
18.2.4
* Prove that — + — + — ≥ ab + be + ca for all positive a, b, c.
b c a

18.3 When Formulas Fail

After learning the Trivial Inequality, Cauchy-Schwarz, the Mean Inequality Chain, and the Rearrange­
ment Inequality, it can become tempting to try to fit every inequality problem into one of these molds.
This temptation becomes even stronger as you learn more and more of these "classical" inequalities—you
attack every inequality by just trying to find which one of these tools solves the problem immediately.
If you take this approach, you'll find that there are a great many inequality problems you simply can't
solve.
In this section, we look at a variety of techniques we use to solve inequality problems when our
formulas fail us.

WARNING!! Don't rely too heavily on formulas and famous theorems. Many of
*∕ the other problem solving strategies we've used to solve equations in
this text can also be used to tackle inequality problems.

600
 18.3. WHEN FORMULAS FAIL

Problems

1111
---- -j- ---- -∣- ----
Problem 18.12: The mathematical constant e can be defined as e = ---
0! 1! 2! 3!

(a) Show that — ≤ -—- for all positive integers n.

(b) Show that 2 < e < 3.

Problem 18.13: Prove that 2" ≥ n2 for all positive integers n > 4.

V
Problem 18.14: Show that if p and q are positive numbers less than 1, then ——
q+1

Problem 18.15: Consider a square array of numbers consisting of m rows and m columns. For each i,
let r,∙ denote the sum of the numbers in the /-th row, and c∕ denote the sum of the numbers in the /’-th
column. Show that there are distinct numbers ∕ and ∕' for which (rz∙ - cl∙)(r∕ - cf) ≤ 0.

1111 ∖* 1
Problem 18.12: The mathematical constant e can be defined ase= - + —∣—-4—- + ■••=/ —
0! 1! 2! 3! n'. n-Q
Show that 2 < c < 3.

Solution for Problem 18.12: We begin by investigating the series for e:

The sum of the first two terms in the series is 1 + 1 =2. Since the other terms are all positive, we see that
e > 1 + 1 = 2.
Proving that e < 3 is trickier. The series isn't arithmetic or geometric. It doesn't telescope. It seems
that we are stuck—we can't evaluate the series at all.
But we don't have to evaluate it! All we have to do is show that it is less than 3. So, if we can show
that the series is less than a series we know how to handle, and that this latter series has a sum less than
or equal to 3, then we have a solution!
A natural candidate to consider is a geometric series, since we get each term of our series for e by
multiplying the previous term by some factor. Suppose, instead of making that factor smaller at each
step as we do for e, we keep that factor the same. Then, we get a geometric series, which we know how
to handle. Let's try it.
The series for e starts 1 + 1 + + ∣. Clearly, we don't want to use 1 as our constant factor, since this
will give us 1 + 1 + 1 + 14---- . We might instead use which gives usS = l + j + j + ∙∙∙. This doesn't
quite work, because some of the terms of S are less than the corresponding terms of our series for e, so

601
CHAPTER 18. MORE INEQUALITIES

we cannot conclude that e < S. However, if we add 1 to S and call the result T, we have

Let's compare this to e:

Aha! It looks like each term of T is greater than or equal to the corresponding term of e. This is clearly
true for the first 5 terms shown. Each subsequent term of T is half the preceding term, while each term
of e after the fifth is found by multiplying the preceding term by a number that is smaller than 1/2. So,
each subsequent term of e is less than the corresponding term of T. We therefore have

The key step in our solution to Problem 18.12 was comparing the unwieldy series for e to a simple
geometric series we know how to handle.

Concept: To find an upper or lower bound of a complicated expression that we

Q—ljJ cannot evaluate directly, we can sometimes compare the complicated ex­
pression to an expression we do know how to evaluate.

Problem 18.13: Prove that 2n ≥ n2 for all positive integers n ≥ 4.

Solution for Problem 18.13: It seems that the function 2” "grows faster" than the function n2 for n > 4,
but this observation is not a proof.
Since we're proving a statement for all integers starting at a certain point, induction seems like a
promising tool. For the base case, we note that when n = 4, we have 24 = 16 = 42. For the inductive
step, we assume that 2k ≥ k2 for some k > 4. We must use this to prove our inequality holds for n - k +1;
that is, we must show that 2fc+1 ≥(k + 1)2.
Multiplying 2k ≥ k2 by 2 gives 2k+1 ≥ 2k2, so we must now show that 2k2 ≥ (k + 1)2. Expanding the
right side, we must show that 2k2 > k2 + 2k +1. Rearranging this leaves us k2 ≥ 2k +1 to prove. We know
that k ≥ 4, so k2 ≥ 4⅛ = 2k + 2k ≥ 2k + 1.
We can now reverse our steps to complete our induction. Adding k2 to both sides of k2 ≥ 2k +1 gives
2k2 ≥k2 + 2k + 1, so 2k2 ≥ (k + 1)2. From our inductive assumption, we have 2k ≥ k2. Multiplying this by
2 gives 2^+1 ≥ 2k2, so 2fc+1 ≥ (k + 1)2, and our induction is complete. □

Concept: If you must prove an inequality for all integers greater than some value,
o≡^≡ then consider induction.

p <∕ 1
Problem 18.14: Show that if p and
∣∣t' q are positive numbers less than 1, thenq+1 p+1 < 1.
------ + ------

602
 18.3. WHEN FORMULAS FAIL

Solution for Problem 18.14: We could try various pieces of the Mean Inequality Chain, but none seems
to deal well with those fractions. We might also try our Cauchy-Schwarz trick for getting rid of
denominators. Cauchy-Schwarz gives us

[⅛+i) + ⅛+ + =(√r+√i)2.

We got rid of the fractions on the right side, but we still have square roots. Worse yet, the expression
^+T + p+T *s on *h e lesser s'de i∏ the desired inequality, but it is on the greater side in the inequality above.

WARNING’! Pay attention to which side of the inequality your target expressions
* are on. If an expression is on the lesser side in your desired inequality,
an inequality that has that expression on the greater side probably
won't be helpful (and vice versa).

Another clue that Cauchy and the Mean Inequality Chain won't work is that our desired inequality
is strict, whereas Cauchy and the Mean Inequality Chain are not. This doesn't completely eliminate the
possibility that these inequalities can help, but it does suggest we'll need something else.
Back to the drawing board. We can get rid of the fractions by multiplying by (<? + l)(p + 1). We don't
have to worry about changing the direction of the inequality because p and q are positive. This gives us
p(p + 1) + <fo∣ + 1) < (p + 1)(^ + 1) to prove. Rearranging this gives us p2 + q2 < pq + 1 to prove. Now, we
appear stuck again. The squares might make us think of the quadratic mean, but they are on the lesser
side, so the QM portion of the Mean Inequality Chain probably won't help.
We appear stuck working backwards, so let's try working forwards. We only know that 0 < p < 1
and 0 < q < 1. We wish to build p2 + q2 < pq + 1 from these inequalities. We might try producing a
pq term by multiplying the given inequalities, which gives 0 < pq < 1. Unfortunately, pq > 0 doesn't
appear helpful, and pq < 1 has pq on the lesser side, not the greater side. So, we try manipulating the
given information in other ways to produce pq on the greater side.
We write p < 1 and <∕<lasl-p>0 and 1 - q > 0. Aha! Multiplying these gives us (1 - p)(l - q) > 0,
which rearranges to 1 + pq > p + q. This gives us both 1 and pq on the greater side. We're really close
now! If only we could turn p and q on the lesser side into p2 and q2. Fortunately, we can do so easily.
Because p is positive, we can multiply 1 > p by p to produce p > p2. Similarly, we have q > q2, so
1 + pq > p + q > p2 + q2, as desired. Now, we're ready to write our proof:

Proof: Because 1 > p and 1 > q, we have 1 - p > 0 and 1 - q > 0. Multiplying these gives us
0 (1 - p)(l - q) > 0, and rearranging gives 1 + pq > p ÷ q. Because p and q are positive, we can
multiply 1 > p and 1 > q by p and q, respectively, to produce p > p2 and q > q2. Therefore,
we have 1 + pq > p + q > p2 + q2. Adding p + q to both sides of 1 + pq > p2 + q2 gives
l+p + q + pq>p2+p + q2 + q. Factoring then gives us (1 + p)(l + q) > p(p + 1) + q(q + 1). Since
p + 1 and q + 1 are both positive, we can divide both sides of the inequality by (p + l)(47 + 1)
without reversing the inequality to produce the desired

□
603
CHAPTER 18. MORE INEQUALITIES

Our solution to Problem 18.14 involves a lot of experimentation and reasoning both forwards and
backwards. This is one key to not getting stuck on problems: don't get locked into one way of looking
at a problem. When you feel stuck on an algebra problem, manipulate the expressions in the problem
in a variety of ways, and look for familiar forms. If you've been working backwards for a while, stop
and try going forwards, working from the given information to produce expressions like those you
need to finish the problem. Conversely, if you're stuck working forwards, turn around and try working
backwards.
Our previous two problems also exhibited a very common technique for tackling inequalities. In
Problem 18.13, we showed that 2fc+1 ≥(k +1)2 by proving 2fc+1 ≥ 2k2 and 2k2 ≥(k + 1)2, and then putting
these together to get 2fc+1 ≥ 2k2 ≥ (k+l)2. In Problem 18.14, we combined 1+pq > p + qwithp + q > p2 + q2
to get 1 + pq > p2 + q2. In both cases, we found a relationship between two expressions by finding a third
expression that we could "sandwich" between them.

Concept: To prove A > B for some expressions A and B, we can sometimes find a
Cl-mJ third expression, C, such that A ≥ C and C ≥ B.

We typically find expression C by focusing on either A or B alone, not both. For example, in
Problem 18.14, we found 1 + pq > p + q by focusing on 1 + pq and ignoring p2 + q2. We then look for a
way to relate this intermediary expression to the other expression in our desired inequality, as we did
when we hunted for a way to show p + q > p2 + q2.

Problem 18.15: Consider a square array of numbers consisting of m rows and m columns. For each i,
let ri denote the sum of the numbers in the z-th row, and c7 denote the sum of the numbers in the j-th
column. Show that there are distinct numbers i and j for which (r1∙ - c1)(r7∙ - c7) ≤ 0.

Solution for Problem 18.15: We must show that (r1∙ - c1)(r7∙ - c7) ≤ 0 for some distinct i and j. 13 2
This is a different problem from showing that we always have (η∙ - cl∙)(r7∙ - c7∙) ≤ 0. Indeed, 12 4
it's not true that we must always have (ri - Ci)(rj - c7∙) ≤ 0. For example, in the array shown 3 12
at right, we have (ri - C1)(r2 - c2) = (1)(1) = 1, which is greater than 0.
So, this is a fundamentally different problem from most of the inequalities we have looked at so far.
We must prove the existence of a pair of distinct numbers i and j such that (r, - c1∙)(r7 - c,∙) ≤ 0, but this
statement doesn't need to be true for any specific i and j. Therefore, trying to prove the statement for a
specific i and j will probably be fruitless. Proving the statement directly seems impossible, so what do
we do?
We fry a proof by contradiction. We think about what happens if there are no distinct i and j such
that (r1 - C∕)(r7 - c7∙) ≤ 0. Then, we must have (r∣∙ - cl∙)(r7∙ - c7∙) > 0 for all i and j. This has the advantage of
giving us inequalities we can work with. If (η∙ - c1)(r7∙ - c7∙) > 0 for all i and j, then the differences z> - c∣i
have the same sign for all k. Is this possible? Below, we compute all r∣ς - ⅛ values for three example
arrays:

1 3 2 r j - c1 = 4 -1 -3 0 r1 - c1 = -4 0 1 3
1 2 4 r2-c2 = l 1 2 3 r2-c2 = 6 3 2 -6
0 1 4 r3-C3 = -5 0 1 4 r3-C3 = -2 0 3 5

604
 18.3. WHEN FORMULAS FAIL

Concept: Specific examples can be useful guides.

O≡≡^ ∣
Above, one fact that stands out in all three examples is that, in each case, the sum of the rk - ck equals 0.
Is this always the case?
Yes! Adding all the rk - ck is the same as adding all the rk, then subtracting the sum of all the ck.
The sum of all the rk and the sum of all the ck are the same, because they both equal the sum of all the
numbers in the array. So, the sum of all the rk - ck must be 0. This gives us our contradiction. If the sum
of all the rk - ck is 0, then it is impossible for all the rk - ck to be positive or for all of them to be negative.
Either they are all 0 (in which case we have (r,∙ - c,∙)(ry - c;) = 0 for all i and ∕), or there are two numbers i
and j such that r,∙ - c∣ < 0 and r∣ - c,∙ > 0.
Now, we're ready to write our proof:

in m in nι
Proof: Because both * ι ck give the sum of all the entries in the array, we have J~
rk and J~ * j ck.
* j rk = J~
0 fc=l fc=l k=l fc=l
m
Rearranging this equation gives ∑Jrk - ck) = 0. If any of the rk - ck are zero, then we choose ∕
fc=l

such that r1∙ - cz∙ = 0, and let j be any other number from 1 to m, and we have (rz∙ - c1∙)(ry∙ - Cy) ≤ 0,
as desired. Otherwise, we proceed by contradiction.
We assume that the rk - ck are either all positive or all negative for all k such that 1 <k <m.
Then, we must have
m
- ck) ≠ θ∕
fc=l

which is a contradiction of the fact proved above. Therefore, the rk — ck cannot all have the
same sign, so there are two numbers i and ∕ such that r; - c1∙ and r; - cy have opposite signs.
This means we have (η - c1∙)(ry - cy) ≤ 0 for some distinct numbers i and j, as desired.

Concept: Existence problems, in which we must prove that something does or

o≡≡^ doesn't exist, can often be solved with contradiction.

Problem 18.15 is an example of using contradiction to prove existence. We showed that there exist
numbers i and j such that (ri∙ - c,∙)(ry - cf) ≤ 0 by assuming that no such i and j exist, and then showing
that this leads to an impossible conclusion. Back on page 231, we used contradiction to show that √2 is
irrational. This is an example of using contradiction to prove that something doesn't exist. Specifically,
we assumed that there exists a rational number, p∕q, that equals y∕2, and then showed that this leads to
an impossible conclusion.
In both cases, the fact that the problem is about existence is a clue to try contradiction. Furthermore,
in both cases, using contradiction gave us equations or inequalities to work with. In other words, it

605
CHAPTER 18. MORE INEQUALITIES

gave us a way to turn the words in the problem into equations or inequalities that we could manipulate
to reach the desired conclusion.

Exercises

18.3.1 In the n-term series Stl = 5.5 + 5.55 + 5.555 + ∙ ∙ ∙, the zzth term has one five before the decimal
point and n fives after it (e.g., the sixth term is 5.555555). Show that 550 < Sιoo < 556. (Source: CIMC)
18.3.2 Show that 2 < 1.01100 < e.
18.3.3 Prove that
2(x3 + y3 + z3) ≥ x2y + xy2 + y2z + yz2 + x2z + xz2
for all nonnegative x, y, and z.
18.3.4 Prove that n∖ ≥ 3” for all integers n ≥ 7.
„___ __ , , _ 1 1 ι 1
18.3.5 Let xyz — 1. Find the minimum and maximum values of S = ---------------∏ τr~-----■------ ,^ π l ~~-
∙7 1 + x + xyl + y + yzl+z + zx
18.3.6 Let a, b, and c be positive real numbers such that a + b + c ≤ 4 and ab + be + ca ≥ 4. Prove that at
least two of the inequalities ∣α - b∖ < 2, ∖b - c∣ ≤ 2, and ∖c-a∖<2 are true. Hints: 171
*
18.3.7 Let x, y, and z be real numbers such that x + y + z = 2 and xy + xz + yz = 1. Find the minimum
and maximum value of x. Hints: 173
18.3.8
* Prove that if a, b, and c are positive real numbers, then (aa)(bb)(cc) ≥ (abc∕a+b+c^3. (Source:
USAMO) Hints: 247

18.4 Summary

Definition: The quadratic mean of a set of real numbers is the square root of the arithmetic mean
of the squares of the numbers. In other words, the quadratic mean of the real numbers tq, az, ...,an

e2 + e2 + u2 + ∙ ∙ ∙ + u2
n
The quadratic mean is also sometimes referred to as the root mean square.

Definition: The harmonic mean of a set of nonzero real numbers is the reciprocal of the average of
the reciprocals of the numbers. In other words, the harmonic mean of the nonzero numbers n↑, a2,
^3∕ ∙ * ∙∕ ⅛
n
l + ± + ± + ... + ±∙
#1 a3 fl∏

The quadratic mean, root mean square, and harmonic mean are sometimes abbreviated QM, RMS,
and HM, respectively.

606
 18.4. SUMMARY

Important: Letxι,*2∕ ∙ ∙ ∙ ∕⅞ > 0, and let QM, AM, GM, and HM denote the quadratic

ψ mean, arithmetic mean, geometric mean, and harmonic mean of these n

numbers, respectively. Then, we have

QM > AM ≥ GM ≥ HM.

For each inequality sign, equality occurs if and only if all of the xi are
equal.

Important: Suppose we have sequences x1, x2, ..., xn and yι, y2, ∙∙∙∕ y∏, where

ψ x↑ < ×2 < ×3 ≤ ∙∙∙ ≤ xn and yι < ι∕2 ≤ t/3 ≤ ∙∙∙ ≤ y∏∙ If we group the
terms of the sequences into n pairs with one term from each pair, take the
product of each pair, then sum all n products, then the largest possible
sum is
xιyι + x2y2 + ■■■ + xnyn,
and the smallest possible sum is

Xiy„ + x2yn-ι + ∙ ∙ ∙ + Xnyι.

This is called the Rearrangement Inequality.

Problem Solving Strategies __________________________________________

Concepts: ∙ Much of success in proving inequalities is recognizing expressions

Q- c7iJ that are part of well-known inequalities such as the Cauchy-Schwarz
Inequality or the AM-GM Inequality.

• To derive some inequalities involving a single sequence of real num­

bers, we can set one of the sequences in the Cauchy-Schwarz Inequal­
ity equal to constants.

• If you suspect a possible application of the QM-AM-GM-HM in­

equality chain, try isolating one of those means in your desired in­
equality.

• We can sometimes use the equality conditions of inequalities as a

guide to solving inequality problems.

• When faced with a problem involving unwieldy expressions, try

making substitutions for those expressions to simplify the problem.

Continued on the next page. . .

607
CHAPTER 18. MORE INEQUALITIES

Concepts: . . . continued from the previous page

θ 1^,lj ∙ If a tactic simplifies a problem but doesn't solve it, look for ways to

apply the tactic again.

• To find an upper or lower bound of a complicated expression that we

cannot evaluate directly, we can sometimes compare the complicated
expression to an expression we do know how to evaluate.

• If you must prove an inequality for all integers greater than some
value, consider induction.

• Existence problems, in which we must prove that something does or

doesn't exist, can often be solved with contradiction.

Review Problems

18.16 Show that 3α2 + 3b2 + 3c2 ≥(a + b + c)2 for all real a, b, and c.
18.17
(a) What is the maximum perimeter of a rectangle inscribed in a circle of radius 1?
(b) * What is the maximum perimeter of a rectangle inscribed in a semicircle of radius 1?

18.18 Let a, b, and c be real numbers such that α2 + b2 + c2 = 1. Prove that -- ≤ ab + ac + be ≤ 1.

18.19 Prove that x2y2 + x2z2 + y2z2 ≥ xyz(x + y + z) for all xty,z≥ 0.
18.20 Show that if a, b, and c are positive numbers such that a + b + c = 6, then

(Source: ARML)
18.21 Find the maximum of x2y if x + y = 6 and x ≥ 0.

18.22 If a and b are positive integers such that a < b and ab < a + 3b, how many possible values are
there for a?
18.23 Let α1, α2, ∙ ∙ ∙ ∕an,b↑,b2,...,bn be positive real numbers. Show that either

«1 02 1 . <2 n bi 1 ½ , bn
≥n or ≥ n.
b1 b2 ”■ bn flι a2 an

18.24 Let a, b, c, d, and e be positive integers such that

abede = a + b + c + d + e.

Find the maximum possible value of e.

608
 CHALLENGE PROBLEMS

18.25 Let a, b, c be positive real numbers. Prove that (a + b + c) (—+ —— + —— ≥ -

v ∖a + b a + c b + c) 2
2 2
18.26 Let a and b be positive real numbers such that a + b = l. Prove that (« + J) + (b + ∣) ≥

18.27 Let 0 ≤ x ≤ 1. Find the maximum value of x*12(l - χ).

18.28 Prove that x7 + 1 ≥ x4S+ x3 for all nonnegative x.

18.29 Show that -r + — ≥ α6 + b6 for any nonzero a and b

b 2 a 2-
18.30 Let a, b, and c be nonnegative real numbers. Prove that abc ≥(a + b - c)(a - b + c)(-fl + b + c).
18.31 Show that if a,b,c> 0, then

1 1 1
fl(l -b)> -, b(l - c) > -, and c(l -a)> -

cannot all simultaneously be valid.

Xl∕ l/Z ZX
18.32 Show that if x, y, and z are positive numbers whose sum is 1, then — + -—ι----- ≥ 1.
z x y
18.33 Prove that nn ≥ 1 ∙ 3 ∙ 5 ∙ ■ ∙ (2rz - 1) for all positive integers n.

ι , , '

Challenge Problems

18.34 Show that x6 + 2≥x3 + x2+x for all real values of x.

18.35 A walk consists of a sequence of steps of length 1 taken in directions north, south, east, or west.
A walk is self-avoiding if it never passes through the same point twice. Let f(ri) denote the number of
rz-step self-avoiding walks that begin at the origin. Show that 2n < f(ri) < 4(3n-1). (Source: Canada)
1 111 1 1 1
18.36 Show that if a, b, and c are positive, then — + — + — ≥ —7= . + -—
——7= + ——. Hints: 180
α2 b2 c2 α ybc b gac
. ■ cyab
18.37 Let S = Xj + X2 + ∙ ∙ ∙ + x∏, where x1∙ > 0 for z = 1,2,..., n. Prove that

S S S n2
—------------ ∏ 77------------ F ■ ∙ ∙ + —--------- ≥ ----------z
S x↑ S λ^2 S xtj ιι 1

with equality if and only if Xi = X2 = ■ ∙ ■ = x∏∙

xy xz yz
18.38 Find the minimum value of — + — + —, where x2 + 1∕2 + z2 = 1 and x,v,z> 0. Hints: 29
z y x j v
18.39 Let x, y, and z be nonnegative real numbers such that x + y + z = 1. Prove that

and find when equality occurs. Hints: 321

609
CHAPTER 18. MORE INEQUALITIES

18.40 Expanding (1 + O.2)1000 by the Binomial Theorem and doing no further manipulation gives

(100‰2)10°°
∖ιooo∕k 7 = Ao + A] + A2 + ∙ ∙ ∙ + A1000∕
where A⅛ = (1θoo)(O.2)fc for k = 0, 1, 2, ..., 1000. For which k is A⅛ the largest? (Source: AIME)
Hints: 77,178
18.41 Determine the minimum value of

for all real numbers r, s, t with l<r≤s≤f≤4. (Source: Putnam) '

18.42 Let α and b be positive real numbers. Prove that

x + a x — b < (a + b)2
x x ~ ⅛ab
for all nonzero x. Hints: 240
I c 3
18.43 Let a, b, c be positive real numbers. Prove that -------- 1---------- 1------- 7 ≥ -∙ (This is Nesbitt's
r b+c a+c a+b 2
Inequality.) Hints: 166, 65
18.44
* Let a, b, and c be real numbers. Prove that the smallest of (a — b)2, (a — c)2, and (b — c)2 is no
greater than (α2 + b2 + c2)∕2. Hints: 37

18.45 If a,b,c > 0, what is the smallest possible value of [~~] + + (Source: HMMT)
Hints: 307
18.46
* Let P(x) be a polynomial with positive coefficients. Prove that if Pθ) ≥ pLy holds for x = 1,
then it holds for all x > 0. (Source: Titu Andreescu)
*
18.47 Let a, b, and c be positive. Show that
a b c
------- - 4--------- . . 4--------- . - ≤ 1.
a + √(rt + b)(a + c) b+ yj(b + c)(b + a) c + √(c + a)(c + b)

Hints: 124
18.48
* Determine the largest real number z such that there exist real numbers x, y, and z with x+y+z = 5
and xy + xz + yz = 3. (Source: CMO)
18.49
* Determine the value of LxJ, where
1111 1
X = 1 H----- — H----- --- ^∣----- -- H----- — + ∙ ∙ ∙ 4----- . ~ ", ~∙
√2 √3 √4 √5 √1,000,000
(Source: USAMTS) Hints: 272, 256
18.50
* Prove that (a5 - a2 + 3)(b5 - b2 + 3)(c5 - c2 + 3) ≥ (a + b + c)3 for all nonnegative numbers a, b,
and c. (Source: USAMO) Hints: 50,352

610
Equations are more important to me, because politics is for the present, but an equation is something for eternity.
- Albert Einstein

CHAPTER

I
Functional Equations

Definition: A functional equation is an equation that includes a function.

For example, f(2x + 1) — 2f(x) + 3 is a functional equation. So are each of the following:

∕(χ'+y) = χ + ∕(y), /(^)=/(X'2/(y)- ∕(χ + y) = ∕M∕(y)-

19.1 Finding Values

Before we dive into solving functional equations, we'll first explore tactics for finding specific values of
a function that satisfies a given functional equation

Problems

Problem 19.1: Let ∕(x) be a function such that ∕(x + y) = x + ∕(y) for any two real numbers x and y,
and ∕(0) = 2. What is the value of /(1998)? (Source: AHSME)

Problem 19.2: The function ∕ has the interval [0,1] as its domain and its range. It also satisfies f(Q) = 0,
∕(1) = 1, and
√χ÷y∖ ∕(χ)+∕(y)

for all x, y ∈ [0,1].

(a) Find all possible values of ∕(1∕2). (b) Find all possible values of ∕(1 /3).

611
CHAPTER 19. FUNCTIONAL EQUATIONS

Problem 19.3: Let f be a real-valued function such that f(m + ri) = f(m)f(ri) for all real numbers m and
n. If ∕(4) = 625, then find a value of k such that ∕(∕c) = 0.04.

Problem 19.4: The sequence ⅜, al, a2,... satisfies

fyn+n ÷ fyn-n = xffl2-m + ^2n)

for all nonnegative integers m and n with m ≥ n. If = 1, determine α1995∙ (Source: Russia)

Problem 19.1: Let f(x) be a function such that f(x + y) = x + ∕(y) fθr any two real numbers x and y,
and ∕(0) = 2. What is the value of /(1998)? (Source: AHSME) __________

Solution for Problem 19.1: We are given ∕(0) = 2, and we want ∕(1998). By letting y — 0 in the functional
equation, we get ∕(x) = x + ∕(0) = x + 2. Now, we just let x = 1998 to find ∕(1998) = 1998 + 2 = 2000.
Note that our work above tells us that the function ∕(x) = x + 2 is the only function that satisfies both
properties of the problem. So, not only can we find ∕(1998), but we can now evaluate ∕(x) for any value
ofx. □
The key step in Problem 19.1 is simply choosing appropriate values for the variables in the functional
equation.

Concept: Choosing values for the variables in a functional equation is a great way
to learn more about the function involved in the equation.

Typically we choose values for which we know the value of the function (as when we chose y = 0
above), or at which we want the value of the function (as when we chose x = 1998).

Problem 19.2: The function ∕ has the interval [0,1] as its domain and its range. It also satisfies 1
∕(0) = 0,∕(l) = l,and
^χ + y)^ ∕(χ) + ∕(y) I

for all x, y ∈ [0,1].

(a) Find all possible values of ∕(1∕2). (b) Find all possible values of ∕(1∕3).
i-WMaH|MaaBaBaiiaaiHaaaiBHaBHiaMHaaaHaaiHaHaBaaHBaHHHaaaiaHHMaaaaaHaHaHaHBaaHBnaBHaBBBaaaHaiai^BBJ|
Solution for Problem 19.2:

(a) What's wrong with this solution:

Bogus Solution: The function ∕(x) = x satisfies the constraints of the problem, since
then we have ∕(0) = 0, ∕(1) = 1, and ∕((x + y)∕2) = (x + y)∕2 =
(∕(x) + ∕(y))∕2. Therefore, we have ∕(1∕2) = 1/2.
All the Bogus Solution tells us is that it is possible to have ∕(1∕2) = 1/2. It does not tell us that this
is the only possible value of ∕(1 /2).

612
 19.1. FINDING VALUES

We know ∕(0) and ∕(1), so letting x = 0 and y = 1 gives us ∕(1∕2) directly:

*
∕β)-⅛±≡.
Notice that this solution does not assume that ∕(x) equals some expression. We only use the given
information in the problem, and show that any function ∕(x) that satisfies the problem constraints
must have ∕( 1/2) = 1/2.
(b) To use the given functional equation to find ∕(1∕3), we seek x and y such that (x + y)∕2 = 1/3. In
other words, we want 1/3 to be the average of x and y. Since we know ∕(0) = 0, we can try taking
x = 0 and y = 2/3. Then we get

∕1∖ _ ∕(0)+∕(2∕3) = ∕(2∕3)

7∖3∕ 2 2 '

so ∕(2∕3) = 2∕(1∕3).
We need another equation relating ∕(1∕3) and ∕(2∕3). We note that the average of 1/3 and 2/3
is 1/2, and we know ∕(1∕2). We have

^(1) ∕(1∕3) + ∕(2∕3)

and we also have ∕(1∕2) = 1/2, so ∕(1∕3) + ∕(2∕3) = 1. Substituting ∕(2∕3) = 2∕(1∕3) into this
equation gives us ∕(1∕3) = 1/3.

Problem 19.3: Let f be a real-valued function such that f(m + ri) = f(m)f(ri) for all real numbers m
and n. If ∕(4) = 625, then find a value of k such that f(k) = 0.04.

Solution for Problem 19.3: We start by experimenting. We find f(m) for various values of m. Maybe wezll
get lucky and find an m such that ∕(m) = 0.04. We know ∕(4), so we start by letting m = n = 4, which
gives us ∕(4 + 4) = ∕(4)∕(4), so ∕(8) = 6252. Looks like we're moving in the wrong direction!

Another way we can use our value of ∕(4) is to let m = 4 and n = 0. This gives us ∕(4 + 0) = ∕(4)∕(0),
which means 625 = 625∕(0). So, we have ∕(0) = 1. That's progress, but we still don't have ∕(k) = 0.04.
Next, let's try m = n = 2, which gives us ∕(2 + 2) = f(2)f(2), so (∕(2))2 = 625. Therefore, we have
∕(2) = ±25. But is ∕(2) equal to 25 or -25? We can rule out ∕(2) = -25 by finding ∕(1) the same way we
just found ∕(2). We let m = n = 1, and we have ∕(2) = (/(I))2. Since ∕(x) is real for all x and ∕(2) is the
square of ∕(1), we know that ∕(2) is nonnegative.

We have ∕(2) = 25, and we can similarly show that ∕(1) = √∕(2) = 5. Hmmm.... We now have
∕(0) = 1, ∕(1) = 5, ∕(2) = 25, and ∕(4) = 625. We can easily fill in the gap to find ∕(3) = ∕(2)∕(1) = 53. We
could use induction to show that f(ri) = 5n for all positive integers n, but that won't give us the value of
k such that f(k) - 0.04. Writing 0.04 as a fraction, we have ∕(k) = 1 /25, and now we suspect that k = -2.
So, we let n = -2, and we have f{m - 2) = ∕(m)∕(-2). A natural choice for m now is m = 2, since we
know both ∕(0) and ∕(2). We then find ∕(0) = ∕(2)∕(-2), so 1 = 25∕(-2), which gives us the expected
∕(-2) = 1/25. Therefore, we have k = -2. □

613
CHAPTER 19. FUNCTIONAL EQUATIONS

Sidenote: That the function in Problem 19.3 turns out to be an exponential function
is no coincidence. In fact, the exponential function ∕(x) = ax, where a is
a positive constant, can be defined as the continuous function such that
∕(1) = a and ∕(x + y) = f(x)f(y) for all real numbers x and y.

Problem 19.3 exhibits one of our most useful strategies in dealing with new functional equations:

Concept: If you can't find the answer to a functional equation question right away,
Q∙ l, j explore the problem by trying various values for the variables. Evaluate
the function at many different values, and hopefully a pattern will emerge.

Here's a harder problem on which to try this strategy.

Problem 19.4: The sequence #0, «1/ #2/ ∙ ∙ ∙ satisfies

∩tn+n ÷ O-m-n = ~(Λ2m ÷ ^2n)

for all nonnegative integers m and n with m ≥ n. If «i = 1, determine fl1995∙ (Source: Russia)

Solution for Problem 19.4: At first glance, this doesn't look like a functional equation problem. But the
given recursion is essentially the same thing as a functional equation. Instead of being written in terms
of ∕(x), it is written in terms of α,∙. For each nonnegative integer input i, there is an output fl1∙, just as we
input x to a function f to get an output ∕(x). So, we try the same substitution tactics we've used on the
functional equations so far.
We can reduce our equation by one variable by letting n = 0, which gives us 2am = (μ2m + ∏o)∕2. We'd
like to evaluate am for some values of m to get a feel for our sequence. A natural substitution now is
m = 0, since this gives us the equation 2#o = («0 + ⅜)∕2, which we can solve to find #o = 0. Substituting
this back into our expression for am, we have 2am = a2m∣2, or 4tzm = fl2m.
Now, we can let m = 1 to find «2 = 4#i = 4. Letting m = 2 gives a$ = 4α2 = 42, and so on. Continuing
in this vein, we can show that λ2* = 4k for all k. Unfortunately, we still can't find am for any other specific
values, so we turn back to our given equation. We note that α2m and #2„ both appear, so we apply
α2m = 4am and tz2π = 4an to find

Qm+n ÷ a∣n-n —

Now, letting n = 1 gives am+↑ + flm-ι = 2am + 2a↑. We solve this for α,π+1 to get the recursion

Qrn+1 = 2∏m ~ fyn-l ÷ 2#i = 2(lm — Um-↑ ÷ 2.

We know #2 = 4#j = 4, and we can compute successive terms as follows:

03 = 2α2 — #1+2 = 8 — 1 + 2 = 9,
#4 = 2#3 - #2 + 2 = 18 - 4 + 2 = 16,
#5 = 2#4 — #3 + 2 = 32 — 9 + 2 = 25,

614
 19.2. FINDING FUNCTIONS WITH SUBSTITUTION

and so on. We've found our pattern! It looks like am = m2. We'd like to show that this formula works
for all positive integers—that sounds like a job for induction. We've already showed that am = m2 for
m = 0 and m = 1. Suppose that k>l and a,n = m2 for all m < k, so that ak = k2 and ak-↑ = (k- 1)2. Then,
we have
⅜+l = 2ak - ak.λ + 2 = 2k2 - (k - 1)2 + 2 = k2 +2k + 1 = (k + 1)2.

So, by induction, we have am = m2 for all positive integers m. In particular, we have α1995 = 19952. □

. H¾ .. ----- ----- ------------- -------------------------------------------------- —-------------------------------------- -----------------------------

Exercises

19.1.1 The function f is such that ∕(0) = 0 and ∕(2x + 1) = 2f(x) + 3 for all real x. Find ∕(1), ∕(3), and
∕(7)∙
19.1.2 F(n) is a function such that F(l) = F(2) = F(3) = 1 and
= F(n)F(n -1) ÷ 1
, F(n- 2)
for n ≥ 3. Find F(6). (Source: AHSME)

19.1.3 For all ordered pairs of positive integers (x, y), we define f(x, y) as follows:
(a) f(xr 1) = χ,
(b) ∕(χ∕ y) = 0 if y > x,
(c) f(x + 1, y) = y[∕(x, y) + ∕(x, y - 1)] if x + 1 ≥ y > 1.
Compute ∕(5,5). (Source: ARML)

19.1.4 If f(x - y) = f(x) ■ f(y) for all x and y, and ∕(x) never equals zero, find all possible values of
∕(1977).

19.1.5
* Assume that ∕(1) = 0, and that for all integers m and n, we have f(m+n) = f(m)+f(n')+3(4mn-V).
Determine f( 19). (Source: USAMTS)

19.2 Finding Functions with Substitution

In the previous section, we found specific values of functions that satisfy certain functional equations.
In the next three sections, we explore strategies for finding the functions themselves.
WeTl start by introducing some new notation. If we write, "Consider a function f : JR → [0,1],"
we mean that function f only accepts real numbers as inputs and only outputs numbers in the interval
[0,1]. In other words, the domain of f is JR, and the range of f only has numbers in [0,1]. Note that
f : JR → [0,1] does not indicate that the range is [0,1]. It merely states that all values in the range must be
in the interval [0,1]. We say that [0,1] is the codomain of the function f, which means that ∕(x) ∈ [0,1]
for all x.

To describe a function fully, we typically give its domain, its codomain, and a formula for the
function. For example, the function ∕ : Z → [-l,3],∕(x) = ∣ is a function with domain Z (the set of all
integers) and codomain [-1,3], but whose range is just the value ∣.

615
CHAPTER 19. FUNCTIONAL EQUATIONS

Finally, we have a special notation for excluding specific values from a set. For example, if a function
is defined for all real numbers except 0, we can write its domain as R∖{0}, where the "∖{0}" indicates that
0 is excluded from the domain. Similarly, Z∖{2,5} means "all integers except 2 and 5."
Don't worry too much about all this new notation. It will seem natural after you've seen it a little
more.
Our goal in solving a functional equation is to find a definition for the function. Usually this means
manipulating the equation until we have ∕(x) equal to some expression in terms of x. In this section, we
use substitutions to find these expressions.

Problems

Problem 19.5: Let f : 1R → R be a function that satisfies ∕(xy) = x2∕(y) fθr x an*^ V∙
(a) Find a substitution for y that reduces the functional equation to a simpler one-variable equation
that you can solve for ∕(x).
(b) Find all functions f that satisfy the functional equation.
(c) Confirm that the functions you found in part (b) satisfy f(xy) = x2f(y) for χ and y∙

Problem 19.6: Find all functions f : R → ]R that satisfy the functional equation ∕(x) + ∕(x + y) — y +10
for all real x and y.

Problem 19.7: Find ∕(x) if

f(a + b)- f(a -b) = 4αb
for all a, b ∈ R.

Problem 19.8: Find all functions f : R → R such that f(x)f(y) - ∕(xy) = x + y for all x, y.

Problem 19.9: Determine all polynomials P(x) such that P(x2 + 1) = [P(x)]2 + 1 and P(0) = 0. (Source:
Putnam)

Problem 19.5: Find all functions f : R → R that satisfy f(xy) = x2f(y) for all x and y.

Solution for Problem 19.5: We want to produce an equation of the form

∕(x) = some expression in terms of x.

We might think to isolate ∕(y) and write ∕(y) = ∕(xy)∕x2. But then we have that annoying ∕(xy) on the
right, and the x2 as well. We can get rid of x by letting x = 1, but that just tells us ∕(y) = f(y)∕l, which
isn't very helpful.
Instead, we focus on the f(xy) on the left side, and try substituting for y rather than for x. Letting
y = 1 gives us ∕(x) = x2∕(l). We don't know what ∕(1) is, but we do know that it is a constant. But what
constant is it?

616
 19.2. FINDING FUNCTIONS WITH SUBSTITUTION

WARNING!! Don't get thrown off by ∕(0) or ∕(1), or /(some number) showing up
k in a functional equation after a substitution. Such an expression is just
a constant. So, call that constant c, and then see if you can learn more
about it.

We let ∕(1) = c, so we have ∕(x) = x2∕(l) = cχ2∙ We now must check which values of c (if any) give us
a valid solution to the functional equation. Letting ∕(χ) = cx2 in the original functional equation gives
us ∕(xy) = c(xy)2 = cx2y2 on the left and x2∕(y) = x2(cy2) = cx2ι∕2 on the right. The two sides of the
equation are the same for all c, so ∕(x) = cx2 is a solution for any real constant c. □

Concept: When dealing with a functional equation that involves several variables,
O~∙ uJ setting one or more of the variables equal to simple constants like 0, 1, or
-1 often results in a simpler functional equation.

Our solution to Problem 19.5 essentially consisted of two parts:

1. We showed that every function that satisfies the functional equation must have the form ∕(x) = cx2
for some constant c.

2. We showed that every function of the form ∕(x) = cx2 satisfies the functional equation.

We must not forget the second step! Our next problem illustrates why this second step is necessary.

Problem 19.6: Find all functions f : JR. → JR. that satisfy the functional equation ∕(x) + ∕(x + y) = y +10
for all real x and y.

Solution for Problem 19.6: The complication here is the ∕(x + y) term. We can reduce this term to ∕(x) by
letting y = 0. This makes our equation 2∕(x) = 10, so ∕(x) = 5 for all x.
We check our solution by substituting ∕(x) = 5 back into the functional equation. This gives us
5 + 5 = y + 10. In order for our solution to be valid, this equation must be true for all possible values
of x and y. This is clearly not the case; if y is anything but 0, then the equation is false. Did we make a
mistake?
No! With our initial substitution, we showed that any function that satisfies the functional equation
must also satisfy ∕(x) = 5. However, we showed that this function does not satisfy the functional
equation. Therefore, there are no solutions to the functional equation. □

Important: After you solve a functional equation, check that your solutions actually

ψ satisfy the given equation. This is not just to check your work; as seen
in Problem 19.6, a solution that you derive from a functional equation
does not automatically satisfy the functional equation.

Problem 19.7: Find ∕(x) if f(μ + b)- f{a -b) = ½ab for all a, b ∈ 1R.

617
CHAPTER 19. FUNCTIONAL EQUATIONS

Solution for Problem 19.7: With f(a + b) and f(a — b) as our main complications, we start off by letting
b = 0, thus reducing both of these to f(a). This gives us f(a + b) - f(a -b) = f(a) - ∕(α) = 0 on the left.
Uh-oh. Over on the right side of the functional equation, we have 4ab = 4α(0) = 0. So, we have 0 = 0.
Not very helpful.
Another way we can reduce the functional equation to a single variable is to let b = a, which makes
the equation
∕(2α) - ∕(0) = 4,√.
Aha! This substitution also turned f(a - b) into the constant ∕(0). We can now let 2a = x, so we have
∕(x)-∕(0) = x2, which gives us∕(x) = x2+∕(0). Since ∕(0) is a constant, we have the solution ∕(x) = x2+c,
for some constant c.
We now let ∕(x) = x2 + c in the original equation. We find that

f(a + b) - f(a - b) = (a + bjz + c - (a - b)2 - c = 4ab.

Therefore, any constant c works, so the solutions of the equation are all functions of the form ∕(x) = x2 + c,
where c is a constant. □
Problem 19.8: Find all functions ∕ : R → R such that ∕(x)∕(y) - f(xy) ~_ffy *z V∙ ∣

Solution for Problem 19.8: Letting y = 0 turns two of our functional expressions into ∕(0), which will
leave us an equation we can solve for ∕(x). This gives us ∕(x)∕(0) - ∕(0) = x. Therefore, we have
∕(x)∕(0) = x + ∕(0). Notice that we don't divide by ∕^(0) right away; what if ∕(0) = 0? If we have ∕(0) = 0,
then our equation ∕(x)∕(0) = x + ∕(0) becomes 0 = x + 0. But this equation must be true for all x, which
clearly isn't the case. So, we can't have ∕(0) = 0.
This means we can let ∕(0) = c, where c ≠ 0, and divide ∕(x)∕(0) = x + ∕(0) by c to give ∕(x) = * + 1.
Next, we substitute back into our functional equation. The left side becomes
x y χy xy
- + - + -γ -
c c c2 c
Clearly, this doesn't equal x + y for all values of c. It does equal x + y for c = 1. To see that c = 1 is
the only possible value of c, we let x = 0 in ∕(x) = f + 1, which gives us ∕(0) = 1. Since c = ∕(0), we
have c = 1, so the only solution to our functional equation is ∕(x) = x + 1, which does indeed satisfy the
original equation. □

WARNING!! Once you let ∕(0) or ∕(1) (or whatever) in your expression for ∕(x)
a equal some constant c, you must remember to plug your expression
for ∕(x) back into the functional equation to check if you need any
restrictions on c.

Problem 19.9: Determine all polynomials P(x) such that P(x2 + 1) = [P(x)]2 + 1 and P(0) = 0. (Source: ∣
Putnam)
■■■■■■■■■■■■■■■■■■■■■■I

Solution for Problem 19.9: Let's start by substituting some values for x. Since we are given P(0) = 0, we
should first try x = 0. This gives P(l) = [P(0)]2 + 1 = 1.

618
 19.3. SEPARATION

We have P(l) now, so we let x = 1. This gives us P(2) = [P(1)]2 + 1=2.

We have P(2) now, so we let x = 2. This gives us P(5) = [P(2)]2 + 1=5.
Hmmm.... We think we know where this is headed. We let x - 5 and find P(26) = 26. Then we let
x = 26 and find P(677) = 677.

We see that P(x) = x for a lot of values of x. In fact, there are infinitely many values of x that
satisfy this equation. We can prove this using induction. We have P(l) = 1, and if P(n) = n, then
P(∕22 + 1) = [P(∏)]2 + 1 = n2 + 1. Therefore, if we have P(k) = k for any integer k, then there is a larger
integer, namely m = fc2 + 1, for which P(m) = m. So, there are infinitely many solutions to P(x) = x.

But P(x) is a polynomial! If there are infinitely many solutions to P(x) = x, then the polynomial
P(x) - x has infinitely many roots. The number of roots of a polynomial with finite degree must equal
the degree of the polynomial. So, P(x) - x must be the polynomial 0, which means P(x) = x for all x. □

ιι Ixeκi¼e¼

19.2.1 Suppose that for all x > 0 we have ∕(2x) = 5⅛. What is 2∕(x)7
19.2.2 Find all functions ∕(x) such that ∕(x + 2y) + ∕(x - 2y) = 2x2 + 8y2 for all real x and y.
19.2.3 The function ∕(x) is defined for every positive integer x and satisfies the equation

∕(x + y) = ∕(x)∕(y) - ∕(xy) + 1

for all positive integers x and y. If ∕(1) = 2, find ∕(x).

19.2.4 Find all functions f : R → R such that 1 + ∕(x + y) = 2∕(x)∕(y) for all real x, y.
19.2.5 Find all functions ∕ : Z → Z such that f(a + b) = f{a) + ∕(α)∕(b) + ∕(b) for all a, b, and ∕(1) = k-l,
where k is some integer greater than 1.

19.3 Separation

Problems

Problem 19.10: Suppose that x∕(x) = yf(y) for all nonzero real numbers x and y.
(a) Suppose that 3∕(3) = 7. What is 5/(5)? What is 8/(8)? What is nf(π)2
(b) Find all functions that satisfy the equation.

Problem 19.11: Find all functions ∕ : R → R such that y∕(x) = xf(y) for all real numbers x and y.

Problem 19.12: Find all functions ∕ : R∖{0} → R such that y2∕(x) - x2∕(y) = y2 - x2 for all nonzero real
numbers x, y.

619
CHAPTER 19. FUNCTIONAL EQUATIONS

Problem 19.10: Find all functions f : R∖{0} → R such thatx∕(χ) = y∕(y) for all nonzero real numbers
x and y.

Solution for Problem 19.10: If we let y = 1, then we have x∕(x) = ∕(1), so ∕(x) = f(l)∕χ∙ Letting c = ∕(1),
we have ∕(x) = c∕x. Substituting this into the functional equation, we have x∕(x) = c and yf(y) = c, so
the function ∕(x) = c/x satisfies the functional equation for any value of the constant c.
We can also solve this functional equation by noticing that the two sides of the equation x∕(x) = yf(y}
are essentially the same. This equation tells us that no matter what value we choose for x, the value of
x∕(x) is always the same. For example, we have x∕(x) = 1∕(1) = 2∕(2) = 3∕(3).
In other words, if we let g(f) = tf(f), then we have g(x) = g(y) for all x and y, which tells us that g(f)
is the same for all t. Therefore, g(f) must be a constant. So, the expression x∕(x) is constant, and we
have x∕(x) = c for some constant c. This gives us ∕(x) = c/x, which we can then substitute into the given
functional equation to see that c can take on any value. □
I Problem 19.11: Find all functions f : JR → R such that yf(x) = χf(y) for all real numbers x and y. ∣

Solution for Problem 19.11: Once again, we could use substitution. Letting y = 1 gives us ∕(x) — x∕(l) — ex
for some constant c.
But let's see if we can also apply our second strategy from Problem 19.10. In that problem, we had
an equation in which the two sides were essentially the same, but with different variables. We try to
manipulate the present equation into a similar form by separating the variables, putting one variable on
each side of the equation. Dividing by xy quickly separates both variables and gives us

∕W = Ay)
x y

This equation must hold for all (nonzero) x and y, so the expression f(x)∣x must be constant. This
means we have f(x~)∣x = c for some constant c, which gives us ∕(x) = ex for all x ≠ 0. To confirm that
∕(x) = ex for x = 0 as well, we let x = 0 in yf(x) = χf(y), and we find y∕(0) = 0 for all y. Therefore,
∕(0) = 0, so ∕(x) must satisfy ∕(x) = ex for all x.
Substituting this back into the original functional equation, we have y∕(x) = xf(y) = cxy, so the
function ∕(x) = ex satisfies the equation for all constants c. □

Concept: In some functional equations with two variables, we can separate the
Q u,□ variables so that each variable appears only on one side of the equation.
If the two sides are the same except for the different variables (in other
words, swapping the variables results in the same equation), then the two
expressions must equal a constant for all values of the variables.

Problem 19.12: Find all functions f : R∖{0) → R such that y2∕(x) - x2∕(y) = y2 ~ χ2 for all nonzero
real numbers x, y.

620
 19.4. CYCLIC FUNCTIONS

Solution for Problem 19.12: Inspired by our success with separation in the previous two problems, we try
the same here. We'll have to separate the y2 from y2∕(χ) and the x2 from x2f(y), so we start by dividing
both sides of the equation by x2y2, which leaves

∕(χ) ∕(y) _ 1 _ J_
X2 y2 x2 y2 ∙

Now, separating the variables is easy. We add - ± to both sides to get

∕ω 1 ∕(V) 1

Aha! If we let g(f) _ - 1, θur equation above tells us that g(x) = g(y) for all x and y. In other words,
g(x) is constant, so we have = c for some constant c. Solving for ∕(x) gives ∕(x) = cx2 + 1.
Substituting this back into the original functional equation gives us

y2∕(x) - x2f(y) = cx2y2 + y2- cx2y2 -x2 = y2-x2,

so all functions of the form ∕(x) = cx2 + 1 satisfy the equation. □

1111 -— -----------------------------------------------------------------
Exercises

19.3.1 For every pair of numbers a and b, the function f satisfies b2f(a) = a2f(b). If f (2) ≠ 0, find the
value of ,^'~ (Source: Canada)

19.3.2 Determine all functions with domain (0, +∞) that satisfy the functional equation y2∙^x> = χ2^.

19.3.3 Find all functions f : R∖{-l} → R such that = whenever x, y, and z do not
y+ ι x+1
equal -1.

19.4 Cyclic Functions

Back in Section 2.3, we introduced a special notation for composing a function with itself. In this text,
we write ∕2(x) to mean ∕(∕(x)). Likewise, ∕'1(x) refers to applying the function ∕(x) exactly n times; for
example, ∕5(x) = ∕(∕(∕(∕(∕(x)))))∙

WARNING!! Some sources use ∕2(x) to mean ∕(x) ∙∕(χ) instead of ∕(∕(x)). Likewise,
V fn(x) is sometimes used to denote [∕(x)]". You'll often have to deter­
mine by context whether ∕"(x) refers to a function applied n times, or
to a function raised to the πth power.

621
CHAPTER 19. FUNCTIONAL EQUATIONS

Problems

Problem 19.14: Given ffx) = ⅜½, let ∕n+ι(x) = ∕ι(∕∏W) for n = 1, 2, 3, .... It can be shown that
/35 = ∕s∙ Find a rational function that equals fιs(x) for all values of x for which ∕28(x) is defined.
(Source: AHSME)____________________________________________ - ________________________

Problem 19.15: The function g : R∖{0} → R satisfies the equation g(x) + 2gQ) = ~x+ χ ^or x ≠ θ∙

(a) Suppose ∕(x) = 1 ∕x. What is ∕2(x)7 What is f3(x)?

(b) Find all functions g(x) that satisfy the given functional equation.

Problem 19.16: Suppose that f satisfies the functional equation 2∕(x) + 3∕(2**∣9) = 100x + 80 for all
x ≠ 2. Find ∕(3). (Source: USAMTS)

Problem 19.17: Let E(x) be a real-valued function defined for all real x except for x = 0 and x = 1 such
that F(x) satisfies the functional equation

F(x) + f(^-) = 1+x

for all x besides 0 and 1. Find all functions F(x) satisfying these properties. (Source: Putnam)

Problem 19.13: Let u(x) — ------ Find u23(x) and u 1(x).

1 — X

Solution for Problem 19.13: There's no obvious way to jump straight to u23(x), so we start by calculating
un(x) for smaller n. We find:

w(x) = 7~-∕
1-x
1-x
1-x 1 -x 1-x
u2 (x) = u(u(x)) = u = 1 .~T^ ^1≤-J- 1-x-l x
i 1-x 1-x 1-x
m= 1 X
u3(x) = ∏(w2(x)) = u 1 =------- ;-------- = λ'∕
1+1→ x+l-x
X
u4(x) = w(u3(x)) = u(x).
It repeats! Because m3(x) = x, we have ι∕5(x) = w2(u3(x)) = w2(x), and ι∕6(x) = u3(w3(x)) = u3(x), and so on,
so that u"+3(χ) = w'1(x) for all positive integers n. Because 23 is one less than a multiple of 3, we have

622
 19.4. CYCLIC FUNCTIONS

u23(x) = u2(x) = -(1 - x)∕x.

To find ιι 1 (x), we could go through the usual steps of solving ∕z(ι∕) = x for y in terms of x, but instead,
we note that we saw π(u2(x)) = x above. In other words, the function ∕(x) that satisfies u(∕(x)) = x is
∕(x) = u2(x). This means that the inverse of u(x) is u2(x), and we have ιΓ1(x) = w2(x) = -(1 - x)∕x. □

We say a function f is cyclic if ∕tt(x) = x for some positive integer n. The smallest positive integer
π for which ∕',(x) = x is the order of f, and fk(x) = fk+n(χ) for all k. For example, above we saw that
u(x) = 1/(1 — x) has order 3, and we have ιz(x) = u4(x) = u7(χ) = ... l w2(χ) = ∏5(χ) = w8(χ) = ∙∙∙, and
m3(x) = ιz6(x) = u9(x) = ∙ ∙ ■.

Problem 19.14: Given ∕1(x) = γ, let ∕π+ι(x) = ∕ι(∕,j(x)) for n = 1, 2, 3, .... It can be shown
that /35 = fζ. Find a rational function that equals ∕28(x) for all values of x for which ∕⅛β(x) is defined.
(Source: AHSME)

Solution for Problem 19.14: Because f35 = f5, we have ∕30(∕5(x)) = ∕35(x) ÷ ∕s(x). Letting y = f5(x), we
have ∕30(y) = y. We might deduce from this that f has order 30, but it is possible that the order is some
smaller n that evenly divides 30. So, we compute the first few fn(x) and hope we get lucky. We have

Λ(X) = ∕1(∕1(X)) =
x-1
∕
X

∕3(X)=∕1(∕2(X)) =
x-2
2x-Γ
We could keep on plugging away, but what if the order is 10, or 15, or 30? Before chugging through all
that algebra, let's stop and think a bit.

Concept: Whenever it looks like you have a great deal of algebra ahead of you, take
Q —t√cj a ^ew minutes to look for another approach to the problem.

We have ∕30(x) = x, and we want to find ∕28(x)∙ Since 28 isn't that far from 30, we look for a way
to start with ∕30(x) and go back to /28 W- I∏ other words, we want to undo f twice. But that's what an
inverse function does! Since f30(x) = f29(f1(x)) = x, we see that ∕29(x) is the inverse of ∕1(x). Solving
∕i(y) = x for y in terms of x gives us ffλ(x') = y = (χ + 1)∕(2 - x). Since f29(χ') is the inverse of ∕1(x), we
have f29(x) = (x + 1)∕(2 - x).
Now, we're just a step away from /28W- We have ∕29(x) = ∕1(∕28(x))∙ Letting y = f28(x), we have
Λ(y) = f29(x) = (x + 1)∕(2 - x). So, to find f28(x), we solve (2y - l)∕(y + 1) = (χ + 1)∕(2 - x) for y in
terms of x. Multiplying both sides by (2 - x)(y + 1) gives (2y - 1)(2 - x) = (χ + l)(y + 1). Expanding and
rearranging gives 3y - 3xy = 3, so f2s(x) = y = 1∕(1 - x)∙ Note that we could have found f2s(x) a little
faster by noticing that because ∕30(x) = x, we have ∕29(∕29(λ')) = ∫5δU) = f30(f28(x)) = f2s(x)- Or, we could
see that ∕f 1(∕1^1(∕30W)) = ∕2δ(x), so f28(x) = ∕1^1(∕f1 W)∙ □

Concept: Just as we can go from fn(x) to ∕n+1(x) by performing the composition

O≡≡b ∕σ"ω), we can also step backwards from ∕n(x) to fn 1(x) when f is in­
vertible by performing the composition ∕"~1(x) = ∕~1(∕''(x)).

623
CHAPTER 19. FUNCTIONAL EQUATIONS

As we saw in Problem 19.14, this tactic can be particularly useful with cyclic functions, since these
functions have values of n such that ∕"(x) = x. This allows us to work both backwards and forwards to
find fk(x) for high values of k.

Problem 19.15: The function g : 1R∖{O} → 1R satisfies the equation g(x) + 2#Q) = -x + χ ^or x ≠ θ∙
Find g(x).

Solution for Problem 19.15: We start with a few substitutions to get a feel for g. The most obvious is x = 1,
which gives us 3g(l) = 0, so g(l) = 0. Unfortunately, after that, there aren't any substitutions that give
us immediate results for g.
We feel a bit stuck, so we examine the equation for any notable features that we might use as a
starting point. We focus on 2g(l ∕x), since this unusual expression makes the equation difficult to handle.
We let t = 1/x to get the variable out of the denominator. This gives us x = 1/t and makes our equation

g(∣) + 2g(f) = -→*-

That doesn't seem to have worked. We took the variable out of the denominator of 2g(l∕x), but ended
up with a variable in the denominator in the g(l∕t) term. But wait a minute—while our equations are
similar, they aren't exactly the same. If we let x = t in the original equation and write it next to our
equation for x = 1/t, we get a system of equations:

gω + 2g(l) = -f÷l,

g(∣)+⅛ω = -→t∙

We can now get rid of the annoying g(l ∕f) term. We subtract twice the second equation from the first
to get -3g(it) = -3t + 7, from which we find g(f) = t - 7. We substitute this into the original equation to
check, and we have

so our solution is g(x) = x - -.□

In hindsight, our clue that our approach in Problem 19.15 might work was the fact that 1/x is cyclic.
So, by letting x = t and x = 1/t in our original equation, we produce two equations with g(f) and g(l∕t)
on the left side, which allows us to eliminate g(l∕t).
In general, any time we have a functional equation in which an argument of a function is itself a
cyclic function, we should consider using appropriate substitutions to create a system of equations as
we did in Problem 19.15.
Note that we used a second variable (namely f) in Problem 19.15 to generate our equations, because
this makes the substitutions easier to follow and less prone to error. However, this substitution is not
necessary. If you feel comfortable going straight from the equation

S(x) + 2gQj = -x÷i

624
 19.4. CYCLIC FUNCTIONS

to the equation
⅝W
*-⅛(j) ∣+Λ
then feel free to do so. However, for more complicated problems, the extra variable will help avoid
some confusion and errors.

Sometimes it's not so obvious that a function is cyclic.

Problem 19.16: Suppose that f satisfies the functional equation

*2 9) = W0x + 80
2f(x) + 3f (~ I

for all x ≠ 2. Find ∕(3). (Source: USAMTS)

Solution for Problem 19.16: Since we want to find ∕(3), we start with substituting x = 3:

2∕(3) + 3∕(35) = 380.

The term ∕(35) appears, so next we try substituting x = 35:

2∕(35) + 3∕(3) = 3580.

Luckily, we obtain another equation involving ∕(3) and ∕(35). Solving this system of equations gives
∕(3) = 1996.
On second thought, were we really so lucky to get that second equation?

Concept: There are few coincidences in mathematics. When you think you have
Q lj1J found one, look more closely.

2x + 29
Let r(x) = Then we have
x-2
2 ∙ ⅛∣9 + 29 _ 2(2x + 29) + 29(x - 2) 33x
I- 2x+29 2 2x + 29-2(x-2) 33 ~X'

And now the mysterious "coincidence" is solved—the function r(x) is cyclic with order 2. In particular,
r(r(3)) = r(35) = 3, which explains why when we substituted x = 35, the term ∕(3) appeared again. □

Problem 19.17: Let F(x) be a real-valued function defined for all real x except for x = 0 and x = 1
such that F(x) satisfies the functional equation

w+f(⅛)= 1 + x
for all x besides 0 and 1. Find all functions F(x) satisfying these properties. (Source: Putnam)

625
CHAPTER 19. FUNCTIONAL EQUATIONS

Solution for Problem 19.17: We have an unusual argument of a function in the functional equation, so we
check if it's cyclic. We let r(x) = (x - l)∕x. Then we have

x-1-x _ 1
r(r(x)) = x-1 x-1 x - 1,
X
and
r(r(r(x))) = —-—- =----- 7-— =------- 7-----= X∙
r(x) -1 ηj-~l x-l-x
Therefore, the function r has order 3.
We've used functions with order 2 to solve functional equations. Let's see if we can do the same
with functions with order 3. First, we let x = t, so

F(0 + F(L_l) = 1 + f. (191)

Next, to get the term F(^), we let x = t-γ, which gives us

(19.2)

Now we have 2 equations, but 3 functional terms: F(f), F(^p), and F(—⅛)∙ Here's where the cyclic
function comes in. To produce another equation with the term F(—777), we let x = — j⅛

(19.3)

Success! We now have three equations and three functional terms. We can eliminate F(^y∙) and
and solve for F(t) by adding equations (19.1) and (19.3), subtracting (19.2) from the result, and
then dividing by 2 to get
f3 - t2 - 1
F(O = 5(l÷<÷l-⅛ 2f(t-l) '

Our last step is to check that this function satisfies the given equation:

x3 - x2 - 1 (⅛)3 - (⅛)2 -1
2x(x -1) 2∙⅛∙(⅛-l)
x3 - x2 - 1 (x - 1)3 - x(x - 1)2 - x3
2x(x -1) 2x(x - l)(x - 1 - x)
x3 - x2 - 1 -x3 - x2 + 2x - 1
2x(x -1) -2x(x-l)
2x3 - 2x
2x(x -1)
= x + 1,

626
 19.5. SUMMARY

for all real x not equal to 0 or 1, so the equation is satisfied. □

<∣ 11______ ■ ^ -- . 1 ~ -- ------- ~------------------------------------------------------

Ixeκi¼<k¼

19.4.1 Let ∕(t) = . Prove that f is cyclic, and find its order.

19.4.2 The function f has the property that for each real number x in its domain, 1/x is also in its
domain, and f(x) + ∕(∣) = x. What is the largest set of real numbers that can be in the domain of f?
(Source: AMC 12)

19.4.3 If 2∕(x) + 3∕(1 — x) = 5x2 — 3x — 11 for all real x, then find f(x).

19.4.4 Find all functions f : R∖{0,1} → R such that xf(χ) + f (⅛) = 2x.

19.4.5 If ∕(x) is a real-valued function such that ∕(x) = [∕(l∕x)]2, find the value of [∕(1977)]2 - ∕(1977).

19.5 Summary

Definition: A functional equation is an equation that includes a function.

Perhaps the most common tactic for working with functional equations is substitution. We choose
values for the variables in the functional equation to learn more about the function.
If a functional equation has two variables, sometimes we can manipulate the equation so that the
variables are separated on opposite sides of the equation. If the two sides then have the same form, we
can often use this information to solve for the function.

We say that a function is cyclic with order n if n is the smallest positive integer such that fn(x) = χ
for all x. If one of the arguments of a function in a functional equation is cyclic, we can sometimes use
the cyclic nature of this argument to create a system of equations that we can then solve for the function.

Important: After you solve a functional equation, check that your solutions actually
satisfy the given equation.

Extra! The Cauchy Functional Equation is

hi∣+iiihHmM∣Φ-

∕(x + y) = ∕(χ) + ∕(y).

This simple-looking equation is one of the most researched functional equations. In

1821, Cauchy proved that the only continuous functions f : R → R that satisfy this
equation are functions of the form ∕(x) = ex for some real constant c. But what if we
allow non-continuous functions, or restrict the domain to Q (rational numbers) or Z
(integers)? Then, are there other functions that satisfy the Cauchy Functional Equation?
These, and many other related questions, have been researched for nearly two centuries.

627
CHAPTER 19. FUNCTIONAL EQUATIONS

Problem Solving Strategies_______________________ ___________________

Concepts: ∙ If you can't find the answer to a functional equation question right
Q' uJ away, explore the problem by trying various values for the variables.

• Whenever it looks like you have a great deal of algebra ahead of you,
take a few minutes to look for another approach to the problem.

• There are few coincidences in mathematics. When you think you

have found one, look more closely.

L
Review Problems

19.18 Suppose f is a function with domain IR, and suppose the equation

∕(X1 + * 3 +*
2 +* 4 +*
δ) = f(×l) + ∕U2) + f(X3) + f(x½) + f(
*
5) ~ θ

holds for all real numbers x↑, X2, X3, X4, λ⅛∙ What is /(0)? (Source: HMMT)
19.19 Let f be a function satisfying ∕(xy) = f(x)∕y for all positive real numbers x and y. If ∕(500) = 3,
what is the value of /(600)? (Source: AMC12)

19.20 Find all functions ∕ : R → R such that ∕(x) + ∕(x + y) + ∕(x + 2y) = 6x + 6y for all real numbers
x, y.
19.21 The function ∕ has the property that, for all nonzero real numbers x, we have
∕ω+2∕(∣) = 3χ.

How many nonzero solutions are there to the equation f(x) = f(-x)2 (Source: AHSME)

19.22 A sequence a↑, a2, a3, ∙ ■ ■ satisfies aχ - 2 and αffl+∏ = 4mnaman for all m and n. Determine an.

19.23 Let u(x) = ■ Show that u(x) is cyclic, and find its order.
1 -x
19.24 Determine all functions that satisfy the identity ∕(xy) = (y2 - y + l)∕(x).

19.25 If f(x) is defined for all integers x ≥ 0, f(a + b) = f(a) + f(b) - 2f(ab), and ∕(1) = 1, compute
∕(2007).
19.26 Find all functions ∕ : R → JR such that ∕(z) + z∕(l - z) = 1 + z for all real z.
γ ∕]
19.27 Let ∕(x) = ------ . If f(f(x)) = x for all x when both sides are defined, then find c.

19.28 The function ∕ satisfies the functional equation ∕(x) + ∕(y) = ∕(x + y) - xy -1 for every pair x, y of
real numbers. If ∕(1) = 1, then find the number of integers n ≠ 1 for which f(n) = n. (Source: AHSME)

628
 CHALLENGE PROBLEMS

19.29 Let f be a function defined on ∣he set of all integers, and assume that it satisfies the following
properties:
(i) ∕(0) ≠ 0,
(ii) ∕(1) = 3,

(iii) ∕(x)∕(y) = ∕(x + y) + ∕(x - y) for all integers x and y.

Determine ∕(7). (Source: USAMTS)

19.30 The sequence of numbers..., α.3, α.2, fl.1, a0r aι,a2f a3^.. is defined by an - (n + l)α2-n = (« + 3)2
for all integers n. Calculate ⅜. (Source: COMC)

19.31 The function f(x) is defined for all real numbers x ≠ 0, and satisfies x∕(x) + 2x∕(-x) = -1 for all
x ≠ 0. Find ∕(x).

19.32 Find all functions f : R → R such that for all real x and y, we have yf(2x) - x∕(2y) = 8xy(x2 - y2).
(Source: Saudi Arabia)

Challenge Problems

19.33 Let f(x) - 4x - x2. Consider the sequence x↑, x2, X3, ... where xl- = f(xj-ι) for i > 1. Compute
the number of values x1 such that x1, x2, and x3 are distinct, but xi = x3 for all i > 3. (Source: ARML)
Hints: 304

19.34 The function f has the property that, for each real number x, we have f(x) + ∕(χ - 1) = χ2. If
∕(19) = 94, find ∕(94). (Source: AIME) Hints: 200
19.35 Let g(x) = ? ∙ Find g2000(2). Hints: 6

19.36 Let F(z) = *⅛ for all complex numbers z ≠ z, and let zn = F(zn.γ) for all positive integers n. Given
that Zo = 1^7 + z, find z2002∙ (Source: AIME) Hints: 239

19.37 The function f satisfies ∕(0) = 0, ∕(1) = 1, and f = — * — for all x, y ∈ R. Show that
∕(x) = x for all rational numbers x. Hints: 44

19.38 Let ∕(x) = —. Find ∕12(x).

j 1 +x j

19.39 The real function f has the property that, whenever α, b, n are positive integers such that a + b = 2",
then f(a) + f(b) = n2. What is /(2002)? (Source: HMMT) Hints: 323
19.40 Solve the functional equation ∕(x + t) - ∕(x - f) = 24xf.

19.41 Find all polynomials p(x) such that (x + 3)p(x) = χp(χ + 1) for all real x. Hints: 263, 31

19.42 Find all functions ∕ : R∖{0) → R such that f(a + b) - f(b) = ----- — for all real values of a and
b for which both sides of the equation are defined.

629
CHAPTER 19. FUNCTIONAL EQUATIONS

19.43 Prove that ∕(n) = 1 — n is the only integer-valued function defined on the integers that satisfies
the following conditions:

(i) ∕(∕(∏)) = n for all integers n,

(ii) f(f(n + 2) + 2) = n for all integers n,

(iii) ∕(0) = 1.

(Source: Putnam) Hints: 108

19.44
* The function f(x, y) satisfies

(1) ∕(θ,y) = y + ι,

(2) ∕(x+l,0) = ∕(x,l),

(3) ∕(x + l,y + l)=∕(x√(x + l,y))

for all integers x and y. Find ∕(4,1981). (Source: IMO)

*
19.45 Find all polynomials p such that p(x) + p(y) + p(z) + p(x + y + z) = p(x + y) + p(x + z) + p(y + z) + p(0)
for all real numbers x, y, and z. Hints: 138
*
19.46 Find all solutions to the functional equation ∕(1 - x) = f(x) + l-2x. (Source: M&IQ) Hints: 48,
274,314

630
Kh∕ soιt of staι t thinking anything s possible if you've got enough nerve. -J. K. Rowling

CHAPTER

I Some Advanced Strategies

In this chapter, we discuss a variety of advanced problem solving strategies. Many of these are extensions
of strategies that we have used several other times in the text.

20.1 Symmetry

We've worked a lot with symmetric sums already. In this section, we address symmetric systems and
symmetric polynomials of one variable.
i

Problems

Problem 20.1: Determine 3x4 + 2x5, if x1, x2, X3, x4, and χ5 satisfy the system of equations given below:

2x1 + X2 + X3 + *4 + X5 = 6,
x1 + 2x2 + X3 + x4 + X5 = 12,
x1 + X2 + 2x3 + x4 + X5 = 24,
x1 + X2 + X3 + 2x4 + X5 = 48,
Xj + x2 ÷ X3 + x4 ÷ 2x5 = 96.

(Source: AIME)

Problem 20.2: Solve the following system of equations:

xy + xz = 8 - x2,
xy + yz = 12 - y2,
yz + zx = -4 - z2.

631
CHAPTER 20. SOME ADVANCED STRATEGIES

Problem 20.3: Let x, y, and z be numbers satisfying x + ∣ = 4, ι∕ + ∣ = 1, and z + £ = ∖. Find xyz.

(Source: AMC12)

Problem 20.4: Find all real solutions to x4 + (2 - x)4 = 34. (Source: HMMT)

Problem 20.5: Suppose x + ∣ = 7. Evaluate each of the following:

(a) x2+⅛.
(b) x3 + ⅛.

Problem 20.6: Let ∕(x) = x6 + χ4 - ∏5x3 + x2 + 1. (Source: Mandelbrot)

(a) Show that if r is a root of f(x), then ∣ is also a root.
(b) Consider the equation f(x)∣xi = 0. What substitution would help you solve this equation?
(c) Find the sum of the real roots of the polynomial x6 + x4 — 115x3 + x2 + 1. (Source: Mandelbrot)

Problem 20.7: The polynomial function f satisfies ∕(6 + x) = ∕(6 - x) for all real numbers x. Moreover,
f (x) = 0 has exactly four distinct real roots. What is the sum of these roots?

Problem 20.1: Determine 3x⅛ + 2xs, if Xi, X2, X3, X4, and X5 satisfy the system of equations given
below:

2xι + X2 + X3 + X4 + X5 — 6,
Xi + 2x2 + X3 + X4 + X5 = 12,
Xi + X2 + 2x3 + X4 + X5 = 24,
Xi + X2 + X3 + 2x4 + X5 = 48,
Xi + X2 + X3 + X4 + 2x5 = 96.

(Source: AIME)

Solution for Problem 20.1: We can solve this system by eliminating or substituting for variables one-by-
one. However, we notice that there is symmetry among the left sides of the equation, and exploiting
this symmetry makes the solution much easier. If we don't see how to take advantage of the symmetry
right away, we can let s = Xi + X2 + X3 + *4 + X5, and our system becomes

Xi + s = 6,
x2+s = 12,
x3 + s = 24,
X4 + s = 48,
X5 + s = 96.

Now, our path is clear. We add all the equations to give 6s = 186, so we have s = 31. We therefore have
x4 = 48 - s = 17 and x5 = 96 - s = 65, so 3x4 + 2x5 = 3 ∙ 17 + 2 ∙ 65 = 181. □

632
 20.1. SYMMETRY

Our solution to Problem 20.1 depended on our recognition of the pattern of the coefficients in the
system of equations.

Concept: Look everywhere for patterns.

Problem 20.2: Solve the following system of equations:

xy + xz = 8 - x2,
xy + yz = 12-y2,
yz + zx = -4 - z2.

Solution for Problem 20.2: It's not immediately clear how we might make use of symmetry, but we can
rewrite each equation with all the variables on the left sides in order to factor a variable out of each:

x2 + xy + xz = x(x + y + z) = 8,
χy + y2 + l∕z = y(x + y + z) = 12,
xz + yz + z2 = z(x + y + z) = -4.

We let S = x + y + z, since we see this factor in all three equations, so we have

xS = 8,
yS = 12,
zS = -4.

Adding these equations gives us (x + y + z)S = 16. Aha—we have S again! Since x + y + z = S, our
equation is now S2 = 16, which means S = +4.
If S = x + y + z = 4, then x = 8/4 = 2, y = 12/4 = 3, and z = -4/4 = -1. If S = x + y + z = -4, then
x = 8∕(-4) = -2, y = 12∕(-4) = -3, and z = -4∕(-4) = 1. So, our two solutions are (x,y,z) = (2,3,-l)
and (-2, -3,1).
If you are very comfortable working with expressions with three variables, you might have added
all the equations prior to any manipulation, giving

2(xy + yz + zx) = 16 - x2 - y2 - z2.

Adding x2 + y2 + z2 to both sides gives

x2 + y2 + z2 + 2(xy + yz + zx) = 16.

Here's where familiarity with three-variable expressions comes in handy. The left side is just (x+y+z)2.□
Note that we found a nice symmetry in the system of equations in Problem 20.2 after rearranging
and factoring each equation.

633
CHAPTER 20. SOME ADVANCED STRATEGIES

Concept: If you don't see how to work with an expression, try rewriting it. Many
θ≡-l⅛) problems can be simplified by a different look at the same information.

Problem 20.3: Let x, y, and z be numbers satisfying

x+-=4, y+- = l, and z+- = -.

y z x 3

Find xyz. (Source: AMC 12)

Solution for Problem 20.3: We could try isolating variables and substituting for them one-by-one, but
that approach looks pretty messy. The similarity among the equations suggests we might be able to find
a nicer solution.

We're looking for the value of xyz, which is an expression that will appear when we multiply the
given equations together:
111 1 28
xyz +x+y+z+-+-+-+---- = —.
xyz xyz 3
In addition to xyz, we have terms we don't want. However, the unwanted terms all appear in the sum
of the three given equations:
1 1 1 ∕ n ∕ ∕1∖ 1∖ 7 22
x+y+z+ +-+ Hz + - =4÷1÷5
+
+

÷
>

— — “
1
1

Xy z ∖ y∕’ ∖ z∕ ∖ x∕ 3 ‘
Subtracting the sum of the given equations from their product, we have an equation involving only the
product of the variables:
1 28 22 π
xyz +---- = —------- — = 2.
y xyz 3 3
Letting P = xyz, we have P+p =2. Rearranging this equation gives us (P - 1)2 = 0, which means P = 1,
so xyz = 1. □
Not all problems have obvious symmetry to exploit. But in some problems, we can introduce a
useful symmetry.

Problem^O^^in^ll^al^olutions^o^H^^^M^ource^MMT^^^^^^^^^J

Solution for Problem 20.4: The obvious place to start with this equation is by expanding the left side. Γm
guessing that's what you tried first, too. And you probably ran into a nasty fourth degree polynomial
with no rational roots. Maybe you then recalled the Fundamental Theorem of Algebra, which tells us
that any quartic with rational coefficients can be factored into the product of two quadratics with real
coefficients. And maybe you even found those quadratics, and their roots.
Or maybe you saw how daunting all that would be, and looked for a slicker solution. We know that
expanding the left side will be a chore, and the resulting quartic won't be much fun to work with. But
maybe we can make a clever substitution to at least help us with the grindwork.
We see two terms raised to fourth powers. The sum of these terms is constant: x + (2 - x) = 2. This
means x and 2 - x are symmetric around the number 1. Letting y - 1 - x allows us to take advantage

634
 20.1. SYMMETRY

of this symmetry. We have x = 1 - ι∕ and 2 - χ = 1 + yz s0 x4 + (2 - x)4 = (1 - y)4 + (1 + y)<. Dojng this

simplifies expanding the fourth powers considerably because the odd degree terms cancel out:

(1 - y)4 + (1 + J∕)4 = 1 - 4y + 6y2 - 4y3 + y4 + 1 + 4y + 6y2 + 4y3 + y4

= 2 + 12y2 + 2y4.

So, our equation now is 2 + 12y2 + 2y4 = 34, so y4 + 6y2 - 16 = 0.

Our new equation has only terms with even exponents. This allows us to factor it as a quadratic
equation in y2, which gives us (y2 + 8)(y2 - 2) = 0. Since x is real, so is y. Therefore, we have y = ± V2,
which means x = 1 + √2 are the real solutions to the original equation. □

Problem 20.5: Suppose x + 1 = 7. Evaluate each of the following:

(a) x2 + ⅛.
(b) x3 + ⅜.

Solution for Problem 20.5: We could start off by solving for x, but the possible values of x are irrational,
which makes our arithmetic a bit messy. Instead, we try to manipulate the given equation to construct
the expressions we intend to evaluate.

(a) We need squared terms, so we square the given equation, which gives us x2 + 2 + =49.
Subtracting 2 from both sides of the resulting equation, we get x2 + 4? = 47.
(b) Solution 1: Proceeding as in (a), we cube the given equation, which gives x3 + 3x + 2 + ± = 343.
We have cubed terms, but we also have other terms. Fortunately, we can evaluate 3x + 2 using the
given equation:

3 o 3 1 3 1 ∕ 1∖ , 1 o 1
x +3xh----- F — — x + — + 3 x + -1 = x3 + —- + 3(7) = x3 h—- + 21.
X x3 x3 ∖ x∕ x3 x3

So, we have x3 + ⅛ + 21 = 343, which means x3 + 4, = 322.

Solution 2: Building off our work from part (a), we note that (x + (x2 + 4) = 7 ∙ 47, so we have
x3 + x+ 2 + ^ = 329. Subtracting x+⅛ = ? from this equation, we get x3 + = 322.

□
We can also use manipulations similar to those in Problem 20.5 to simplify some problems involving
one-variable polynomials.

Problem 20.6: Find the sum of the real roots of the polynomial ∕(x) = x6 + χ4 - 115x3 + x2 +1. (Source:
Mandelbrot)

Solution for Problem 20.6: First, we try finding rational roots. By the Rational Root Theorem, the only
possibilities are 1 and —1, and both fail. Guessing irrational roots seems impossible, so we'll have to
look for some nice way to relate this sixth degree polynomial to something simpler.

635
CHAPTER 20. SOME ADVANCED STRATEGIES

Concept: When stuck trying to deal with an algebraic expression, focus on any
Q— h⅞ξ) unusual aspects of the expression.

The most interesting aspect of the given polynomial is the symmetry of its coefficients; the coefficients
read the same forwards as backwards. To investigate whether or not this is important, let's look at some
simpler examples. We start by looking at quadratics whose coefficients read the same forwards as
backwards. (Linear polynomials with this property are not very interesting; they all are of the form
rx + r, which means -1 is the root of each such polynomial.)
So, we look for what's interesting about the roots of a quadratic of the form ax2 + bx + a. We
focus on what's important about the constant and quadratic coefficients being the same. This suggests
considering Vieta's Formulas, which tell us that the product of the roots is a∕a = 1. In other words, the
roots are reciprocals of each other!
Hmmm.... Let's see if there's something similar true about our given sixth degree polynomial.
Suppose r is a root of ∕(x). We then have ∕(r) = 0, so
r6 + r4 - 115r3 + r2 + 1 = 0.

We want to see if 1 ∕r is also a root. We have

1 1 115 1 , 1 + r2 - 115r3 + r4 + rβ f(r)
+ 1 = ^÷^-7Γ÷^ + 1 =------------------------------
r6 = = 0∙

Success! If r is a root of ∕(x), then so is l∕r.

But how can we use this information? We know that for each root, the reciprocal of that root is also
a root of ∕(x). So, we try rearranging the equation so that it contains both x and J. One way to do so is
to divide ∕(x) = 0 by x3, which gives us
, 1 1 λ
x3 + x - 115 + - + -r = 0.
x x3
This introduces reciprocals, and highlights the fact that the coefficients read the same both forwards and
backwards. We see that our new expression includes both x + 1 and x3 + We know how to relate
these two expressions. We let y = x + j, and we cube y:

y3 = χ3 + 3x+- + -≈=x3 + ~7÷3(x+-^ = x3 + ^+ 3y.

y X x3 x∙i ∖ x∕ xj
gO/ χ3 + = y3 _ 3y NOW/ we can rewrite x3 + x - 115 + ⅛ + ⅛ in terms of y:

13+rll5 + l + ^ = p + L) + (al + l)-115

= (y3-3y) + y-115 = y3-2y-115.

We therefore have y3 - 2y - 115 = 0. Fortunately, there is a rational root, and we can factor the left side
3g (y _ ζ)(↑∕2 ^y + 23) = 0. We need the sum of the real roots of x, and whenever x is real and nonzero,
we know that y is also real. But the roots of y2 + 5y + 23 are nonreal, so we are only concerned with the
root y = 5, from which we have x + J = 5. Rearranging this equation gives x2 - 5x + 1 = 0. Both roots of
this quadratic equation are real, and by Vieta's Formulas, their sum is 5. □

636
 20.1. SYMMETRY

Concept: The substitution y = x+∣ can help us simplify problems involving polyno-
θ-q⅛⅞ mials in x in which the coefficients read the same forwards and backwards.

Problem 20.7: The polynomial function f satisfies ∕(6+x) = f(6-x) for a 11 real numbers x. Moreover,
∕(x) = 0 has exactly four distinct real roots. What is the sum of these roots? ∣

Solution for Problem 20.7: Let r be a root of ∕(χ) = 0. We can find another root in terms of r using
∕(6 + x) = ∕(6 - x). Letting r = 6 + x, we have χ = r - 6 and

∕(r) = ∕(6 + x) = ∕(6 - x) = ∕(6 - (r - 6)) = ∕(12 - r).

Since ∕(r) = 0, we have ∕(12 - r) = 0 as well. Therefore, both r and 12 - r are roots, which means we can
group the roots of ∕(x) into two pairs such that sum of each pair is 12. So, the sum of the roots of f is
2 ∙ 12 = 24. □

I IL
Exercises

20.1.1 Four positive integers are given. Select any three of these integers, find their average, and add
this result to the fourth integer. In this fashion, the numbers 29, 23, 21, and 17 are obtained. Find the
four original integers. (Source: AHSME)
1 1
20.1.2 Find the value of r3 + — given that r+ - = 3. (Source: AHSME)

20.1.3 If y = x + 1, then write the equation x4 + x3 - 4x2 + x + 1 = 0 in terms of y. (Source: AHSME)

20.1.4 Solve the following system of equations:

ab = 1, be = 2, cd = 3, de = 4, ea = 6.

(Source: Sweden)
20.1.5
(a) Show that the only solution to the system

X1 + x2 + x3 = 0,
X2 + x3 + X4 = o,

X98 + X99 + X100 = 0,

X99 + *100 + Xl = 0,
X100 + Xl + x2 = 0.

is x1 = x2 = x3 = ∙ ∙ ■ = x100 = 0.
(b)
* In part (a), if there were 99 variables in 99 equations instead of 100 variables in 100 equations,
would all of the variables still have to be equal to 0?

637
CHAPTER 20. SOME ADVANCED STRATEGIES

20.2 Substitution for Simplification

Way back in Chapter 1, we reviewed a most basic use of substitution: turning a system of two linear
equations in two variables into a single one-variable linear equation. The substitution simplified the
system of equations. In this section, we expand on this strategy, simplifying various equations and
systems of equations with strategically chosen substitutions.
∣ f ; ∣_________________________________________________

Problems

Problem 20.8: Consider the following system of equations, where x and y are real numbers:

x + y + √x + y = 30,
x - y + √x-y = 12.

(a) Find two substitutions that simplify the system of equations.

(b) Solve each equation individually for your two new variables from (a).
(c) Solve the original system of equations.

Problem 20.9: In this problem, we find all solutions to the system of equations

2x1 + 3y2 - 4xy = 3,

2χ2 -if = 7.

(a) Solve the second equation for 2x2 and substitute the result into the first equation.
(b) Can you solve the resulting equation in part (a) for x or y?
(c) Solve the problem.
(d) In this part, we find a second, more subtle, solution. Instead of eliminating a variable from the
system, combine the equations in a way that eliminates the constants. In other words, combine
the two equations into a new equation such that each nonzero term in the new equation has at
least one variable in it. Use the new equation to solve the system.

Problem 20.10: In this problem, we find all ordered pairs of real numbers (x, y) that satisfy the system

x3 - 4y3 = 4,
3y3 -x2y+ xy2 = 1.

(a) Combine the equations to produce an equation that does not have a nonzero constant term.
(b) Suppose (x, y) = («, b) is a solution to your equation in part (a). Explain why (x, y) = (ra, rb) is
also a solution for any constant r.
(c) Suppose a and b are nonzero constants such that (x, y) = (m, rb) is a solution for any constant r.
Notice that all of these solutions have the same ratio x∕y. Use this fact to find a substitution that
you can use to reduce the equation to a one-variable equation.
(d) Find all ordered pairs of real numbers (x, y) that satisfy the original system.

638
 20.2. SUBSTITUTION FOR SIMPLIFICATION

Problem 20.11: Consider the equation V3x2 - 4χ + 34 + V3x2 - 4x - 11 = 9.

(a) Set a and b equal to the two radical parts. Rewrite the original equation using a and b.
(b) Find a second equation relating a2 and b2.
(c) Solve the system of equations you came up with in parts (a) and (b).
(d) Find all real values of x that satisfy the original equation.

Problem 20.13: Find all values of x such that 2x2 + 3x + 3 = 7x √x + 1. (Source: ARML)

Problem 20.8: Solve the following system of equations, where x and y are real numbers:

x + y + ∙√x + y = 30,
≡-m≡k X~y+ = 12'

Solution for Problem 20.8: There's no easy way to isolate x or y, and no easy way to eliminate the square
roots. However, since x + y is repeated in the first equation and x - y repeated in the second, substitution
might make the equations a lot easier to work with.

We could substitute for x + y and x - y. However, this would still leave us with square roots in our
equation. Instead, we let u = fx + y and v = y∣x=y, which turns our equations into simple quadratics:
u2 + u = 30,
v2 + v = 12.

Concept: Use substitutions to turn complicated equations into simple ones. Re-
O≡≡B peated expressions within an equation are a clue that substitution might
be helpful.

The solutions to the first equation are u = 5 and u = -6. The solutions to the second equation are v = 3
and v = -4. However, u and v are equal to square roots of expressions, so they can't be negative. This
means u = 5 and v = 3. Now, we have yfx + y = u = 5 and -fx — y = v = 3. Squaring both gives x + y = 25
and x - y = 9, from which we find that (x, y) = (17,8). □

To solve Problem 20.8, we used two related substitutions. Though we introduced multiple variables,
the result was still a significant simplification of the given problem.

Concept: Multiple substitutions can make a complicated problem easier to solve.

O≡≡s

639
CHAPTER 20. SOME ADVANCED STRATEGIES

Sometimes these substitutions aren't so obvious, though.

Problem 20.9: Find all solutions to the system of equations

12*+ 3y2 - 4xy = 3,

2x*
2x2 - y2 = 7.

Solution for Problem 20.9: Solution 1: Substitute for x2. We can't solve the first equation for x or y, but we
can easily solve the second equation for either x2 or y2 in terms of the other. Since 2x2 appears in both
equations, it's easiest to simply solve the second equation for 2x2, and substitute the result in the first
equation.

Concept: Don't make problems harder than they have to be. When you have multiple
Q—u∣J options for substitution, choose the one that appears to offer the easiest
path.

From the second equation, we have 2x2 = 7 + y2, so substituting for 2x2 in the first equation gives us

4y2 - 4xy - -4.

Dividing this equation by 4 gives y2 — xy = -1. We still have two variables. We could solve 2x2 = 7 + y2
for x and substitute for x, but this would both introduce radicals and fractions, not to mention a ± sign.
So, back to the drawing board.
But not all the way back to the beginning. The equation y2 - xy = -1 is linear in x, which means it is
easy to solve this equation for x in terms of y, which gives

1
x = y + -.
y

Important: If you have an equation in a system that is linear in any variable, you

ψ can solve for that variable and substitute the result in the rest of the
equations.

Substituting this expression for x into 2x2 = 7 + y2 gives us

2y2 + 4*+ ∣z = 7 + y2∙

Rearranging this equation gives us y4 - 3y2 + 2 = 0, which gives us (y2 - l)(y2 - 2) = 0, so y = ±1 or

y = ±√2,
From these values of y, we find the four solutions (x, y) = (2,1), (-2, -1), Q √2, V2), and (-∣ √2, -√2).

Solution 2: Turn the problem into one you know how to do. We have tackled equations like 6x2+xy- 15y2 =
0 by factoring the left side, or by letting r = x∕y. In the present problem, both of our equations have

640
 20.2. SUBSTITUTION FOR SIMPLIFICATION

the complication that there is a nonzero constant on the right side of the equation. So, we eliminate the
constants on the right side in our system of equations by subtracting 3 times 2x2 - y2 -r7 from 7 times
2x2 + 3i∕2 - 4xy = 3 to give
8x2 + 24y2 - 28xy = 0.
Dividing by 4 gives 2x2 + 6y2 - 7xy = 0. Now, we're in familiar territory. We factor this equation to
find (2x - 3y)(x - 2y) = 0, from which we have x = ∣y or x = 2y. Substituting x = 2y into either of our
original equations gives (x, y) = (2,1) and (x,y) = (-2,-1) as solutions. Substituting x = ∣y into either
of our original equations gives (x, y) = (∣ √2, √2), and (x, y) = (-∣ V2, as solutions. □

Our second solution is an example of a powerful problem solving strategy called homogenization.
A homogeneous equation is one in which all nonzero terms have the same degree. So, 2x2 + 6y2 - 7xy = 0
is homogeneous, but 2x2 - y2 = 7 is not, because the 7 term has degree 0 while the other terms have
degree 2. hi our next problem, we explore why this strategy is so useful.

Problem 20.10: Find all ordered pairs of real numbers (x, y) that satisfy the system
: . .∙√ ■ V V.
x3 - 4y3 = 4,
3y3 - x2y + χy2 = 1.

Solution for Problem 20.10: It's not too appealing to solve the first equation for x or y and substitute
into the second equation, so we look for another approach. In the previous problem, eliminating the
constants worked well, so we try the same here. Subtracting 4 times the second equation from the first
equation gives
x3 + 4x2y - 4xy2 - 16y3 - 0.
Immediately, we see that (0,0) is a solution, and if x or y is nonzero, then the other must also be nonzero.
We also see that all four nonzero terms have the same degree, so the equation is homogeneous. But how
does that help?
Suppose we find one solution (x, y) = («, b) to this equation, so

a3 + 4a⅛ - 4ab2 - 16b3 = 0.

Because every term has degree 3, if we multiply x and y both by r, we multiply each term by r3. So, if
we let (x, y) = (ra, rb), then

x3 + 4x2y - 4xy2 - 16y3 = (ra)3 + 4(ra)2(rb) - 4(ra)(rb)2 - 16(rb)3 = r3(α3 + 4λ⅛ - ⅛ab2 - 16b3) = 0.

Therefore, if (x, y) = (a, b) is a solution of the equation, then so is (x, y) = (ra, rb) for any constant r.

Concept: If we find a solution to a homogeneous equation, then multiplying every

Q—⅛∣j variable in this solution by the same constant produces another solution
to the equation.

Of course, we still haven't found any solution besides (0,0) to our equation. But we now know that
if (x,y) = (a,b) is a solution, then so are all pairs of the form (ra,rb). Investigating what's special about
all these pairs, we find that, as long as a, b, and r are nonzero, they all have the same ratio of x to y,

641
CHAPTER 20. SOME ADVANCED STRATEGIES

namely (ra)∕(rb) = a∕b. Aha! Now, our problem is reduced to finding all possible values of this ratio.
Letting x/y = k gives us x = yk, and substituting this into our equation gives

k3y3 + 4k2y3 - 4ky3 - 16y3 = 0.

For y ≠ 0, we can divide by y3 to give k3 + 4k2 - 4k - 16 = 0. Finally, an equation we know how to solve!

Concept: When working with a two-variable homogeneous equation, it's often help­
er ud ful to set the ratio of the variables equal to a new variable and write the
equation in terms of the new variable.

Factoring this polynomial gives (k - 2)(k + 2)(k + 4) = 0, so k = 2, k = -2, or k = -4.

We handle these three cases separately. When k = 2, we have x = 2y, and substituting this into the
first equation gives 4y3 = 4. The only real solution to this equation is y = 1, which gives (2,1) as a
solution to the system. Similarly, x = -2y gives us -12y3 = 4, so y = —⅛, which gives us
as a solution. Finally, letting x = -4y gives us -68y3 = 4, so y = -^=, which means J=j is a

solution. Those last two solutions would have been very difficult to guess! □

∣Problem20.11: Solve √3x2 - 4x + 34 + √3x2 - 4x - 11 = 9. ∣

Solution for Problem 20.11: We have a nasty equation with a repeated expression, so we look for a
substitution to simplify the equation.
Solution 1: Let y = 3x2 - 4x - 11, so our equation is

√y + 45 + √y = 9.

We know how to tackle this equation! We isolate y,y + 45 to give -√y + 45 = 9 — √y, and squaring this
gives
y + 45 = y-18√y + 81.
Simplifying this equation gives 18>∕y = 36, from which we find y = 4. Since y = 3x2 — 4x — 11, we have
4 = 3x2 - 4x - 11, which means 3x2 - 4x - 15 = 0. The roots of this quadratic are 3 and -5/3. Because
we squared an equation to find these solutions, we must check if they are extraneous. Both solutions
satisfy the original equation, so our solutions are x = 3 and x = -5/3.
Solution 2: Our first solution takes advantage of the fact that the expressions under the radicals differ
only by a constant. Perhaps we can assign a variable to each radical expression and find another way
to take advantage of this fact. We let a = V3x2 - 4x + 34 and b = V3x2 - 4x - 11, so that we are given the
equation a + b = 9. We haven't yet used the similarity between the expressions under the radicals, so we
focus on that.

Concept: When stuck on a problem, focus on information you haven't used yet.
O≡≡≡

642
 20.2. SUBSTITUTION FOR SIMPLIFICATION

Because the expressions under the radicals differ by a constant, the difference of the squares a2 -b2 is a
constant:
a2-b2 = (3x2 - 4x + 34) - (3x2 - 4x - 11) = 45.
This makes us happy—now we can factor! We have (a - b)(fl + b) = 45, and we know a + b = 9, so we
have a-b = 5. Since a + b = 9 and a-b = 5,we have (α, b) = (7,2). Now, we have V3x2 - 4x + 34 = a = 7.
Squaring this gives 3x2 — 4x + 34 = 49, which results in the same values of x as in our first solution. □

Sometimes, more is better. While it's not always a good idea to bring lots of extra variables into a
problem, our first solution to Problem 20.11 shows that a little substitution can go a long way. Our second
solution shows that a little more substitution can sometimes go even further. The primary advantage
of the second solution over the first is that the necessary algebraic manipulations after substitution are
much simpler.

Here's another problem in which we have multiple options for substitution, and choosing the right
one helps simplify the post-substitution algebra.

Solution for Problem 20.12: This problem seems intimidating at first glance. Not only does the given
equation involve several quadratics, but those quadratics are in the denominators of fractions.
Fortunately, the quadratics are related in that they differ only by constants. This allows us to substitute
for all three quadratics at once. We could choose many possible substitutions such as a = x2 - 10x or
b = x2 - 10x - 29. We choose to let y = x2 - 10x - 69 with the goal of keeping the algebra as simple
as possible. This makes one of the denominators equal y, and leaves no negative constants in the
denominators after substitution:
1__ 1 2 _
y + 40 y + 24^y^

Now we just get rid of the fractions by multiplying both sides by the product of the denominators,
which gives y(y + 24) + y(y + 40) - 2(y + 24)(y + 40) = 0. Simplifying the left side gives 64y + 1920 = 0,
so y = -30.
From our original y = x2 - 10x - 69, we have x2 - 10x - 69 = -30. Rearranging and factoring gives
(x + 3)(x - 13) = 0, so x = 13 is the positive solution. □

Problem 20.13: Find all values of x such that 2x2 + 3x + 3 = 7x Vx + 1. (Source: ARML)

Solution for Problem 20.13: We might start by squaring the equation, but that produces a huge mess
on the left side. At a loss for anything else to do, we might try it anyway. What results is a quartic
polynomial that has no rational roots.
Before we give up, we look for a helpful substitution. What makes this equation nasty is the square
root expression, so we substitute y = Vx + 1, which gives us 2x2 + 3x + 3 = 7xy. This, at least, is a

643
CHAPTER 20. SOME ADVANCED STRATEGIES

quadratic. We might then try to use the quadratic formula to find x in terms of y. But that looks messy,
and besides, we already have a simple relationship between x and y, namely y = Vx + 1. So, what
should we do?
We look for other ways to use our substitution, and see that x + 1 appears on the left side, too. We
have 2x2 + 3x + 3 = 2x2 + 3(x + 1). We can use our substitution again! Since y = Vx + 1, we have y2 = x +1,
so our equation is
2x2 + 3y2 = 7xy.
Now we're in familiar territory. Subtracting 7xy from both sides and factoring gives (2x - y)(x - 3y) = 0,
so 2x = y or x = 3y.
If 2x = y, then we have 2x = Vx + 1- Squaring both sides gives 4x2 = x + 1, which gives us the
solutions x = (1 ± V17)∕8. Because Vx + 1 = 2x, we can't have x < 0, so we discard the extraneous root
(1 - √17)∕8.
If x = 3y, then we have x = 3 Vx + 1. Squaring both sides gives x2 = 9x + 9, which gives us the
solutions x = (9 ± 3 V13)∕2. Once again, x must be positive, so we discard (9-3 V13)∕2 as extraneous.

Finally, our two solutions are x = (1 + V17)∕8 and x = (9 + 3 V13)∕2. □

11
Exercises

20.2.1 Find all solutions to the equation 72a = 45 ∙ r7a + 4 ∙ 72.

20.2.2 Find all real numbers y such that y8 + 108 = 21y4.

20.2.3
(a) Find all values of r such that (r2 + 5r)(r2 + 5r + 3) = 4.
(b) If x is a positive integer and x(x + l)(x + 2)(x + 3) + 1 = 3792, compute x. (Source: ARML)

20.2.4 Find all solutions to the equation (6√4 - x - 17)x = 36V4 - x - 68.
1
20.2.5 Find all values of x such that 4x2 - 6x - 41 + = 0.
2x2 - 3x - 19
20.2.6 Find all values of x such that √x2 + 17x + 59 - 2√x2 + 17x - 85 = 3.

20.2.7 Find all ordered pairs (x, y) such that

2x4 + 5x3y + 45xy3 = 34x2y2 + 18y4 and 2x - 3y = 7.

20.3 Method of Undetermined Coefficients

We have used the Method of Undetermined Coefficients several times already in this book. Simply
put, the Method of Undetermined Coefficients is the use of variables to represent coefficients whose
values we don't yet know.

644
 20.3. METHOD OF UNDETERMINED COEFFICIENTS

Problems

Problem 20.14: The polynomial x2 + 2x + 5 is a factor of x4 + px2 + q. Find p and q. (Source: AHSME)

Problem 20.15: Let ∕(x) and g(x) be two quadratic polynomials such that ∕(x) = 2x2 - 3x + 5 and
∕(g(x)) = 18x4 ~ 21x2 + 10. Find g(x).

Problem 20.16: Suppose p is a polynomial such that p(l) = 3 and p(x) = p(2x) - 6x2 + 3x.
(a) Show that the degree of p(x) is greater than 1.
(b) Show that the degree of p(x) is no greater than 2.
(c) Let p(x) = a2x2 + a↑x + ao. Find p(x).

Problem 20.17: Real roots x1 and X2 of the equation x5 - 55x + 21 = 0 satisfy X1X2 = L In this problem,
we find X] and x^. (Source: Bosnia and Herzegovnia)
(a) What must the constant term of a monic quadratic be if the product of its roots is 1 ?
(b) Suppose x5 - 55x + 21 is the product of a quadratic and a cubic. Suppose x↑ and X2 are the roots
of the quadratic. Why can we let the quadratic and cubic be monic? What are the constant terms
of the quadratic and the cubic?
(c) Find the rest of the coefficients of the quadratic, and find x1 and x2.

Problem 20.14: The polynomial x2 + 2x + 5 is a factor of x4 + px2 + q. Find p and q. (Source: AHSME) '

Solution for Problem 20.14: Solution 1: Regardless of the values of p and q, when we divide x4 + px2 + q
by x2 + 2x + 5, the leading term of the quotient is x2. So, we let the quotient be x2 + αx + b, so that
x4 + px2 + q = (x2 + 2x + 5)(x2 + ax + b) - x4 + (a + 2)x3 + (2a + b + 5)x2 + (5a + 2b)x + 5b.
Equating the coefficients on the far left side above to the corresponding coefficients on the far right, we
have a system of equations;
a + 2 = 0,
2a + b + 5 = p,
5a + 2b = 0,
5b = q.
From the first of these equations, we have a = -2. The third equation then gives us -10 + 2b - 0, so
b - 5. The second equation therefore gives p = 2a + b + 5 =-4+ 5 + 5 = 6, and the last equation yields
q = 5b = 25. Now, we have the coefficients of the second quadratic factor, and the quartic itself;
(x2 + 2x + 5)(x2 - 2x + 5) = x4 + 6x2 + 25.

In practice, we can often use a number of shortcuts on a problem like this one. We start as before,
with
x4 + px2 + q = (x2 + 2x + 5)(x2 + ax + b).

645
CHAPTER 20. SOME ADVANCED STRATEGIES

Rather than expanding the product on the right, we focus on important pieces of information that will
help us find a, b, p, and q quickly. For example, the coefficient of x3 on the left is 0. The only x3 terms
in the expansion of the right side come from (2x)(x2) and (x2)(ax), so we must have 2x3 + ax3 = 0, which
means a = -2. This means we have

x4 + px2 +q = (χ2 + 2x + 5)(x2 - 2x + b).

Similarly, we can focus on the x term of the expansion on the right to deduce that (2x)(b) + (5)(-2x) = 0,
which gives us b = 5, so x4 + px2 + q = (x2 + 2x + 5)(x2 - 2x + 5). Now, looking at the constant terms gives
q = 25, and looking at the x2 terms (or simply letting x = 1) gives us p = 6.

Solution 2: While the Method of Undetermined Coefficients provides a nice solution to this problem,
some students might cleverly sidestep its use by noting that ∕(x) = x4 +px2 + q is an even polynomial
function. So, ∕(x) = ∕(-x), which means that if r and s are roots of ∕(x), then —r and —s are also roots.
We are given that ∕(x) = (x2 + 2x + 5)g(x) for some quadratic polynomial g(x). Since the roots r and
s of x2 + 2x + 5 don't sum to 0, they are not opposites. Therefore, the roots of g(x) must be —r and —s.
Because r and s are roots of x2 + 2x + 5, we have r + s = -2 and rs = 5. Therefore, we have (-r) + (—s) = 2
and (—r)(-s) = 5. So, the sum of the roots of g(x) is 2 and the product of the roots is 5, which means
g(x) = x2 — 2x + 5. We can now compute ∕(x) = (x2 - 2x + 5)(x2 + 2x + 5) = x4 + 6x2 + 25. □

While our second solution to Problem 20.14 is slicker than the first, the first solution is much
more broadly applicable to problems. Specifically, it is an example of the Method of Undetermined
Coefficients:

Concept: When working with an unknown polynomial, it is often useful to assign

O≡^ variables to the coefficients of the polynomial, and then construct equations
in terms of those variables.

Problem 20.15: Let ∕(x) and g(x) be two quadratic polynomials such that ∕(x) = 2x2 - 3x + 5 and
∕(g(x)) = 18x4 - 21x2 + 10. Find g(x). ______

Solution for Problem 20.15: Let g(x) = ax2 + bx + cf where a, bl and c are constants. Then

∕(g(x)) = f(aχ2 + bx + c)
= 2(αx2 + bx + c)2 - 3(ax2 + bx + c) + 5
= 2λ2x4 + ⅛abx3 + (4λc + 2b2 - 3λ)x2 + (4bc - 3b)x + 2c2 - 3c + 5.

Equating the coefficients of this expression to the corresponding coefficients of the given 18x4 - 21x2 + 10
gives us a system of equations:

2a2 = 18,
4flfe = 0,
4αc + 2b2 -3a = -21,
4bc -3b = 0,
2c2 - 3c + 5 = 10.

646
 20.3. METHOD OF UNDETERMINED COEFFICIENTS

From the first equation, we have a = ±3. From the second equation, we have ab = 0, but a ≠ 0, so b = 0.
The last equation gives us 2c2-3c-5 = 0, so (c + l)(2c-5) = 0, which means c = -1 orc = 5/2. Finally, we
match the possible values of fl and c using the third equation, 4ac - 3fl = -21, to find that (a c) = (3 -1)
or (-3,5/2).

Each of the two possible ordered triples (fl, b, c) = (3,0, -1) and (fl, b, c) = (-3,0,5/2) gives us a solution
to the problem. The two possible polynomials g are g(x) = 3x2 - 1 or g(x) = -3x2 + f. □

Now, we take a look at a functional equation that can be solved using the Method of Undetermined
Coefficients.

Problem 20.16: Suppose p is a polynomial such that p(l) = 3 and p(x) = p(2x) - 6x2 + 3x. Find p(x).

Solution for Problem 20.16: By considering the leading terms of both sides of p(x) = p(2x) - 6x2 + 3x, we
suspect that the degree of p is 2. The degree of p cannot be less than 2, since this would make the degree
of the left side less than that of the right. To show that degp cannot be greater than 2, we let the leading
term of p(x) be flπx", where n > 2. The leading term of p(x) then is anxn, while that of p(2x) — 6x2 + 3x is
2nanxtl. We must therefore have an = 2nan, which means an = 0. However, this is a contradiction, since
we can't have an = 0 if anxn is the leading term of p(x). So, the degree of p cannot be greater than 2.
Since degp cannot be greater than 2 or less than 2, it must equal 2, so we have p(x) = fl2x2 + fljx + flo∙
Substituting this into p(x) = p(2x) - 6x2 + 3x gives

fl2x2 + flιx + flo = (4fl2 - 6)x2 + (2flι + 3)x + aq.

Equating the coefficients of x2, we see that 4fl2 — 6 = «2, so «2 = 2. From the coefficients of x, we have
flι = 2flι + 3, so flι = -3. We now have p(x) = 2x2 - 3x + fl0. How will we get a07

Concept: When stuck on a problem, try focusing on information in the problem that
θ ∏-j you haven't used yet.

We haven't used the fact that p(l) = 3, so we try applying that. We have p(l) = 2 - 3 + flo = flo - 1, so we
must have aq - 1 = 3. Therefore, we have flo = 4, so p(x) = 2x2 - 3x + 4. □

Problem 20.17: Real roots x↑ and X2 of the equation x5 - 55x + 21 = 0 satisfy X1X2 = 1. Find Xi and X2.
(Source: Bosnia and Herzegovnia)

Solution for Problem 20.17: Let p(x) = x5 - 55x + 21. A quick check reveals no rational roots, so it's
not clear how we can factor p(x). However, the fact that X1X2 = 1 gives us some information about a
quadratic factor of p(x). Specifically, because X1X2 = 1, we know that x↑ and X2 are the roots of a quadratic
of the form x2 + ax + 1 for some constant α. This quadratic must be a factor of p(x) because its roots are
also roots of p(x). So, we must have

x5 - 55x + 21 = (x2 + ax + l)(fex3 + cx2 + dx + f).

From the leading term of both sides, we have b = 1, and from the constant term, we have f = 21. So, we
have
x5 - 55x + 21 = (x2 + ax + l)(x3 + cx2 + dx + 21).

647
CHAPTER 20. SOME ADVANCED STRATEGIES

There's no x4 term on the left, so the two x4 terms in the expansion of the right side, which are ax4 and
cχ4, must cancel each other. This gives us a = -c, and we have

x5 - 55x + 21 = (χ2 - ex + l)(x3 + cx2 + dx + 21).

We could continue this reasoning coefficient by coefficient, but it's not so obvious which one will give
us immediately useful information next. So, we multiply out the right side, and we have

x5 - 55x + 21 = x5 + (d - c2 + l)x3 + (21 - cd + c)x2 + (-21c + d)x + 21.

Equating coefficients of like terms gives us the system

d — c2 + 1 = 0,
21 - cd + c = 0,
-21c + d = -55.

The first equation gives us d = c2 -1. Substituting this into the third equation gives us c2 — 21c -1 = -55.
Solving this quadratic gives us c = 3 or c = 18. If c = 3, then a = —3 and d = 8. The triple (a, c, d) = (-3,3,8)
does indeed satisfy all three equations. If c = 18, then a = -18 and d = 323. However, this solution does
not satisfy the second equation in our system.
So, we have one solution for (a, c, d), and we have our factorization:

x5 - 55x + 21 = (x2 - 3x + l)(x3 + 3x2 + 8x + 21).

The roots of x2 -3x+1 have product 1. From the quadratic formula, we find that these roots are (3± λ∕5)∕2.
As an Exercise, you'll show that the other factor, x3 + 3x2 + 8x + 21, does not have a pair of real roots
with product 1. □

Sidenote: This probably isn't the last you'll see of the Method of Undetermined Co-
K efficients. You'll see it again if you study differential equations in calculus.

Exercises

20.3.1 If a and b are integers such that x2 - x -1 is a factor of ax3 + bx2 + l, then find b. (Source: AHSME)

20.3.2 If 3x2 + kxy - 2y2 -7x + 7y-6 is the product of two linear factors with integral coefficients, find
the value of k. (Source: CMO)
20.3.3 Let p(x) be a fourth degree polynomial with leading coefficient 2 such that p(-2) = 34, p(-l) = 10,
p(l) = 10, and p(2) = 34. Find p(0).
20.3.4 Show that we have found the only possible solution to Problem 20.17 by showing that the
polynomial x3 + 3x2 + 8x + 21 does not have a pair of real roots with product 1.
20.3.5
* Solve for all complex numbers z such that z4 + 4z2 + 6 = z. (Source: HMMT) Hints: 2

648
 20.4. CONSTRUCTING POLYNOMIALS FROM ROOTS

20.4 Constructing Polynomials From Roots

We've discussed several techniques and theorems that help us determine relationships between poly­
nomials and their roots. However, we've focused mostly on finding those roots. In this section, we start
with information about the roots of polynomials, and construct the polynomials themselves, or solve
other related problems.

Problems

Problem 20.18: The roots of the polynomial P(x) are 2, 3, 7, and -5. The constant term of P(x) is 105.
(a) What is the product of the roots of P(x)?
(b) What is the coefficient of x4 of P(x)?
(c) WhatisP(x)?
(d) What is the sum of the coefficients of P(x)?

Problem 20.19: Let a, bf and c be the roots of ∕(x) = x3 + 8x + 5.

(a) Construct the monic polynomial with roots 1 + a, 1 + b, and 1 + c.
(b) Find the value of (1 + α)(l + b)(l + c).
111
(c) Find the value of------ +------- +-------
1+a l + ⅜ 1+c

Problem 20.20: Let the roots of the αx2 + bx + c - 0 be r and s. Find the monic quadratic whose roots
are ar + b and as + b. (Source: AHSME)

Problem 20.21: Consider the following system of equations:

x + y + z = 2,
x∖
j + xz + yz = -19,
xyz = -20.

(a) Find a monic cubic polynomial that has x, y, and z as its roots.
(b) Solve the system of equations.

Problem 20.22: If P(x) denotes a polynomial of degree n such that P(k) = for k = 0, 1, 2, ..., n,
determine P(n + 1). (Source: USAMO)

Extra! Suppose we have a degree n polynomial whose graph passes through the n + 1 points
"* l"* l"* ,* (xq, yo)∕ Ui∕ yι)∕ ∙ ∙ ∙ 0⅛∕ y∏)∙ We can find the polynomial as follows. First, we note that
the graph of ∕(x) = (x - xo)gι(x) + yo≠ where gι(x) is any polynomial, will pass through
Uθ∕yo)∙ Now, we need gι(x) to be a degree n - 1 polynomial whose graph passes
through (x1, jpz⅞), (*2, ∙ ∙ ∙ ∕ (*n, ~ξ)∙ So, we just repeat the process, by letting
*)
g↑(x} = (x - X1)g2( + Continuing in this manner, we can find the polynomial ∕(χ)
whose graph passes through all n + 1 given points.

649
CHAPTER 20. SOME ADVANCED STRATEGIES

Problem 20.23: Suppose P(x) is a polynomial with finite degree such that P(l) = 1 and

P(2x) _ 56
P(x + 1) x+7

for all real x for which both sides are defined.

(a) Show that P(x) = xQ(x) for some polynomial Q(x).
(b) Express Q(x) as the product of a Linear factor and a polynomial R(x).
(c) Express R(x) as the product of a linear factor and a polynomial S(x).
(d) Find P(-l). (Source: HMMT)

Problem 20.18: The roots of the polynomial P(x) are 2,3,7, and -5. The constant term of P(x) is 105.
What is the sum of the coefficients of P(x)?

Solution for Problem 20.18: LetP(x) = ax4+ bx3 + cx2+ dx + 105. By Vieta, the product of the roots is 105∕α.
The product of the given roots is 2 ∙ 3 ∙ 7 ∙ (-5) = -210, so we have 105/« = -210, which means a = -1/2.
We could plow ahead with Vieta's Formulas to find the rest of the coefficients, but we just want the sum
of the coefficients. We know that the sum of the coefficients equals P(l), so maybe we can find a faster
way than Vieta's Formulas to construct an expression for P(x) that we can use to find P(l).
Looking for another way to construct P(x), we note that we have all the roots, and thus know all
the monic linear factors of P(x). We also have the leading coefficient of P(x), so we can use the Factor
Theorem to write
P(x) = a(x - 2)(x - 3)(x - 7)(x + 5) = -∣(x - 2)(x - 3)(x - 7)(x + 5).

We could expand this product to find the coefficients, but instead we note that P(l) equals the sum of
the coefficients of P(x). So, the sum of the coefficients of P(x) is P(l) = -1(1 - 2)(1 - 3)(1 - 7)(1 + 5) = 36.
□

Important: Vieta's Formulas and the Factor Theorem can be quite helpful in con­

ψ structing polynomials from information about their roots.

Solution for Problem 20.19:

(a) We could expand the product (1 + «)(1 + b)(T + c) an<l Lise Vieta's Formulas to find the values of the
symmetric sums. However, if we could find a polynomial with roots 1 +a, 1 + b, and 1 + cr we could

650
 20.4. CONSTRUCTING POLYNOMIALS FROM ROOTS

use Vieta on that polynomial to quickly find the desired product. So, we seek the polynomial g(y)
with roots 1 + a, 1 + bf and 1 + c. We relate the roots of g(y) to those of ∕(x) by letting y = 1 + x.
This makes y = 1 + α when x = a, and y = l + b when x = b, and y = 1 + c when x = c.
From y = 1 + x, we have x = y - 1. Because the roots of ∕(x) are a, bf and c, we have
∕(-γ) = ∕(y ~ 1) = 0 when y equals a +1, b +1, or c +1. So, f(y -1) produces our desired polynomial
g(y):
g(y) = ∕(y - 1) = (y - 1)3 + 8(y - 1) + 5 = y3 - 3y2 + lly - 4.
Since (l+α)(l+fe)(l+c) is the product of the roots of^(y)z Vieta gives us (l+Λ)(l+fc)(l+c) = =4.
Note that our answer equals -∕(-l). See if you can figure out why this is not a coincidence.
(b) Solution 1: Find a polynomial with roots ⅛, 5lτ, and ⅛. The polynomial g(y) = y3 - 3y2 + lly - 4
from part (a) has roots a + l,b + l, and c +1. We seek the polynomial whose roots are the reciprocals
of these roots, since we can then use Vieta on the new polynomial to find our desired sum. So,
just as we let y = x + 1 in part (a), we let z = 1 here. Since g(y) = 0ify-a + l, we have g(4) = 0 if
z = a+r Similarly, and are also solutions to the equation g(∣) = 0. So, we find g(|):

∕1∖ 1 3 11 λ 1-3z + 11z2-4z3

# ∖7z∕ = zj - ~
zl
+ Z 4 =-------------z35----------- ∙
-----

Notice that the coefficients of the numerator are the same as those of g(y), but with their order
reversed. (You'll prove in the Exercises that this is not a coincidence.)
We have g(4) = 0 if and only if 1 - 3z + llz2 - 4z3 = 0, and g(∣) = 0 for z = ;dL, z = and
2 = ⅛> so 1 - 3z + llz2 - 4z3 has roots Ap and Finally, Vieta's Formulas tell us that the
sum of these roots is -ll∕(-4) = 11/4.
Solution 2: Use Vieta's Formulas with g(y)from part (a). The polynomial

g(y) = y3 - 3y2 + lly - 4

from part (a) has roots a +1, b +1, and c +1. If we let a +1 = r,b + l = s, and c+1 = t, we can write
our desired expression in terms of symmetric sums of r, s, and t:

1 4- 1 — -∣ 1— _ — st + rt + rs β
1 ∣ 1 1—1 __-------------
a + l- b + 1------- c+1 r s t rst
By Vieta, we have st + rt + rs = 11 and rst = -(-4) = 4, so our desired sum is 11/4.

Our key step in Problem 20.19 was transforming one polynomial into another that allows us to
compute the desired expression directly with Vieta's Formulas. See if you can use the polynomial
transformation tactics we developed in Problem 20.19 on the next problem.

Problem 20.20: Let the roots of the ax2 + bx + c = 0 be r and s. Find the monic quadratic whose roots
are ar + b and as + b. (Source: AHSME)

Solution for Problem 20.20: We use our work in Problem 20.19 as guide. We seek a monic quadratic with
roots ar + b and as + br so we let y = ax + b, which means that y = ar + b when x = r and y = as + b when

651
CHAPTER 20. SOME ADVANCED STRATEGIES

x = s. Solving y = ax + b for x gives x = (y — b)∕a. Substituting this expression for x into ax2 + bx + c gives

+ bx + c = a
(?)
1.2b b2 b
÷c

b2
- -↑f------- y +----- F -y------ + C
aj a j a aj a

We have ax2 + bx + c = 0 for x = r and x = sr and we have ax2 + bx + c = ^y2 — ⅛y ÷ c when y = ax + b.

So, when x = r, we have y = ar + b and ∖y2 -fy + c = ax2 + bx + c = 0, which means that ar + b is a root
of ∣y2 - ⅛y + c. Similarly, as + b is also a root of ∣y2 - ^y + c. Therefore, multiplying this quadratic by a
gives us our desired monic quadratic, y2 - by + ca. □

Solution for Problem 20.21: This system of equations initially looks quite frightening. While we could
grind out a series of substitutions to find the answer, we'd prefer to find a nicer solution.
The left sides of the system are the elementary symmetric sums of three variables, which makes us
think of Vieta's Formulas. So, we let x, y, and z be roots of a cubic polynomial. We let the leading
coefficient of the polynomial be 1, so that -(x + y + z) gives us the quadratic coefficient, xy + yz + zx gives
us the linear coefficient, and -xyz is the constant. Letting the variable be t, we therefore have

(f - x)(f - y)(f -z) = t3- 2t2 - 19t + 20.

Factoring the polynomial on the right, we find that t = -4, t = 1, and t = 5 are its roots. We can
assign these three to x, y, and z in any order. So, the six solutions to the system of equations are
(x, y, z) = (-4,1,5), (1, -4,5), (-4,5,1), (1,5, -4), (5,1, -4), and (5, -4,1). More succinctly, we can say that
these six solutions are the six permutations of (-4,1,5). □
Our fairly simple solution to Problem 20.21 comes from recognizing that the relationships between
symmetric sums and roots of a polynomial work both ways. We can calculate symmetric sums of roots
of a given polynomial, or we can construct a polynomial whose roots are the variables in a system of
symmetric sum equations.

Concept: Just as we can find the values of symmetric sums given a polynomial, we
O≡^ can construct polynomials given symmetric sums.

Problem 20.22: If P(x) denotes a polynomial of degree n such that P(k) = for k = 0, 1, 2,..., n,
determine P(n+ 1). (Source: USAMO)
652
 20.4. CONSTRUCTING POLYNOMIALS FROM ROOTS

Solution for Problem 20.22: It is tempting to set

β(χ) = p<x) -

and then analyze Q(x), since we are given n + 1 values of x for which Q(x) = 0. But Q(χ) is not a
polynomial, so we cannot use all our polynomial tools on Q(x). However, if we write the right side of
the above equation as a rational function, we have

P(x} — - — (x + ~ ∙y
' ' x ψ1 ~ x÷l
The numerator is a polynomial, so we instead let R(x) = (x + l)P(χ) - x, and analyze R(x).
Because P(x) = for x = 0,1, 2,..., n, but x + 1 ≠ 0 for all of these values, we have

R(x) = (x + l)P(x) - x = (χ + 1) ∙ —----- χ = 0

x+1
for x — 0, 1, 2, ..., n. Furthermore, since P(x) has degree n, the polynomial R(x) has degree n + 1. We
have already found all n + 1 roots of R(x), so we have

K(x) = cx(x - l)(χ - 2) ∙ ∙ ∙ (χ - n) (20.1)

for some constant c. Now, we're in familiar territory. All we have to do is find c. We don't know P(x)
for any more values of x, but we can eliminate P(x) from R(x) = (x + l)P(x) — x by letting x = —1, which
gives us R(-l) = 1. So, letting x = -1 in Equation (20.1), we get

R(-l) = c ■ (-1) ∙ (-2) ∙ (-3) ∙ ∙ ∙ (-1 - n) = c(-l)w+1(n + 1)!,

*(-υ (-D"+1 j
soc = (-ιr-(,, + i)! = (^ij!-and
(-l)n+1
*)
K( = fr + i)! ■ χ(χ ^ 1Xx - 2) ∙ ∙ ∙ (x - n).

Then, we have
(-l)π+1
R(n + 1) = ■ (n + 1) ∙ n ∙ (n - 1) ∙ ∙ ∙ 1 = (-1)m+1.
(κ + 1)!
R(x) + x
Solving R(x) = (x + l)P(x) - x for P(x) gives P(x) = for x ≠ -1, so
x+1
(-l)n+1 + n + 1 if n is even,
P(n + 1) =
n+2 if n is odd.
□
Problem 20.23: Suppose P(x) is a polynomial with finite degree such that P(l) = 1 and

P(2x) = 56
P(x + 1) x+7

for all real x for which both sides are defined. Find P(-l). (Source: HMMT)

653
CHAPTER 20. SOME ADVANCED STRATEGIES

Solution for Problem 20.23: The left side is written as a ratio of polynomials. We begin by rewriting the
right side that way so that we can compare apples to apples:
P(2x) _ 8x
P(x + 1) x + 7'
We'd like to use the information that P(l) = 1. We could do so by letting x = 1/2 or by letting x = 0. We
choose the latter since it makes the right side of our equation 0, and thus identifies a root of P(x). Letting
x = 0 gives us P(0)∕P(l) = 0, so P(0) = 0.
Since x = 0 is a root of P(x), we know that x is a factor of P(x). So, we let P(x) = xQ(x)∕ ar>d we have
8x P(2x) 2xQ(2x) Q(2x) 4(x + l)
------------- XΞ --------------------- — --------------------------------------- SO ---------------------- —-------------------- ∙
x+7 P(x + 1) (x + l)Q(x + l)' Q(x + 1) *+7

This looks familiar. Similar to what we did in our first step, we let x = -1 to find a root of Q(x)∙

Concept: If a tactic gives some information in a problem, but doesn't completely

Q uj solve it, try using the tactic again.

We find that Q(-2)∕Q(0) = 0, so Q(-2) = 0. This means Q(x) = (x + 2)R(x) for some polynomial R(x).
Substituting for Q(x), we get
4(x + 1) _ Q(2x) __ (2x + 2)P(2x) ^ K(2x) _ 2(x + 3)
x + 7 - Q(x+1) - (x + 3)R(x + l)' S° R(x + 1) x + 7'

Next, we let x = -3 to get R(-6)∕R(-2) = 0, so R(-6) = 0. This means R(x) = (x + 6)S(x) for some
polynomial S(x). Substituting for R(x), we get
2(x + 3) R(2x) _ (2x + 6)S(2x) c Sf2x)
x + 7 ^ R(x + 1) ^^ (x + 7)S(x + l)' s° S(x + 1)

which means that S(2x) = S(x + 1) for all x.

We've reduced the problem to a much simpler equation. Here are two ways to use S(2x) = S(x + 1)
to show that S(x) is constant for all x:

• Method 1: Experiment. Letting x = 2 in S(2x) = S(x + 1) gives S(4) = S(3). Letting x = 3 gives
S(6) = S(4), and letting x = 5 gives S(10) = S(6). Continuing in this way, we can find infinitely
many values of x such that S(x) = S(3). This means that the polynomial S(x) - S(3) has infinitely
many roots. We know that S(x) is not the zero polynomial, since S(2x)∕S(x + 1) = 1, so S(x) has
finite degree. Therefore, the fact that S(x) - S(3) has infinitely many roots tells us that S(x) - S(3) = 0
for all x. This means that S(x) = S(3) for all x, so S(x) is constant.

• Method 2: Undetermined coefficients. We know that S(x) cannot be the 0 polynomial, since we
would then have division by 0 at various points in the problem. Therefore, we can let S(x) =
anxn + απ-ιxn^1 + ∙ ∙ ∙ + «ix + ⅜, where an ≠ 0, so deg S = n. The leading coefficient of S(2x) then is
2nan and the leading coefficient of S(x + 1) is an. If S(2x) = S(x + 1) for all x, we must then have
2nan = an. Since an ≠ 0, we must have 2" = 1, so n = 0, which means S(x) is a constant.

654
 20.5. COMMON DIVISORS OF POLYNOMIALS

Now, we can find P(χ). We have

p(x) = xQ(x) = x(x + 2)R(x) = x(x + 2)(x + 6)S(x) = cx(x + 2)(x + 6).
Because P(l) = 1, we have c = 1/21, so

P(-l) = (-l>(-l + 2)(-l + 6) =

21 21
See if you can find another solution to this problem by using the Method of Undetermined Coefficients
from the beginning, in the spirit of our solution to Problem 20.16. □
Problem 20.23 is a great example of the importance of perseverance in problem solving.

Concept: Don't give up!

z
ι
Exercises

20.4.1 Compute + + given that r, s, and t are the roots of x3 — 2x2 + 3x — 4 = 0.

20.4.2 Let p, q, r, s, t satisfy the three equations

p3 + sp + t = 1,
q3 + sq + t = 1,
r3 + sr + t = ^[.

If p, q, and r are distinct, evaluate p2 + q2 + r2 in terms of s and t.

20.4.3 Let ∕(x) = anxn + ⅜-ιx"^1 + ∙ ∙ ■ + a1x + a0 and g(x) = a0xn + a1xn~1 + ■■■ + an~1x + an, where a0an ≠ 0.
Show that each root of f is the reciprocal of a root of g.
20.4.4 In part (a) of Problem 20.19, we let a, b, and c be the roots of ∕(x) = x3 + 8x + 5. We used a
transformation and Vieta's Formulas to evaluate (1 +fl)(l+b)(l+c). We then noticed that (l+α)(l+b)(l+c) =
-∕(-l). Why is this not a coincidence?
20.4.5
* Let p(x) be a degree 2 polynomial such that p(l) = 1, p(2) = 3, and p(3) = 2. Then p(p(x)) = x has
four real solutions. Find the only such solution that is not an integer. (Source: Mandelbrot) Hints: 114
*
20.4.6 Let P(x) be the polynomial of minimal degree that satisfies
1
P(∕c) =
fc(fc + 1)
for k = 1, 2, 3,..., 10. Find P(∏). (Source: Thomas Mildorf) Hints: 76

20.5 Common Divisors of Polynomials

A common divisor of two integers is an integer that divides both integers. We have a similar definition
for a common divisor of two polynomials.

655
CHAPTER 20. SOME ADVANCED STRATEGIES

Definition: Let ∕(x) and g(x) be polynomials. A common divisor of ∕(x) and g(x) is a polynomial
p(x) such that p(x) is a factor of both f(x) and g(x).

Polynomials with common polynomial divisors of degree greater than 0 share all the roots of those
common divisors. For example, the quadratics x2 - 3x + 2 and x2 — 4x + 3 are both divisible by x - 1, and
they share x = 1 as a root. This means that when we are looking for common roots of polynomials, we
can find them by finding their common divisors. Similarly, we can use common roots to find common
divisors.

Concept: Finding common roots of polynomials is equivalent to finding common

o^≡τi≡) divisors of the polynomials.

Problems

Problem 20.24: Let ∕(x), g(x), and ∕z(x) be polynomials such that ∕z(x) is a factor of both ∕(x) and g(x)∙
Prove that ∕ι(x) is a factor of α(x)∕(x) + b(x)g(x) for any polynomials «(x) and b(x).

Problem 20.25: Let p(x) = x2 + bx + cr where b and c are integers. If p(x) is a factor of both x4 + 6x2 + 25
and 3x4 + 4x2 + 28x + 5, then find b and c. (Source: AHSME)

Problem 20.26: For what values of b do the equations 1988x2 + bx + 8891 = 0 and 8891x2 + bx +1988 = 0
have a common root? (Source: CMO)

Problem 20.27: The equations x3 + ax2 + 7x - 6 = 0 and x3 - x2 + bx + 3 = 0 have two common roots
for some integers a and b. Find a and b.

Problem 20.24: Let ∕(xj, g(x), and ∕z(x) be polynomials such that ∕ι(x) is a factor of both ∕(x) and g(x). I
Prove that ∕ι(x) is a factor of α(x)∕(x) + ⅛(x)g(x) for any polynomials fl(x) and fr(x).__________ ∣

Solution for Problem 20.24: Since ∕z(x) is a factor of both ∕(x) and g(x), we know that there are some
polynomials <∕ι(x) and c∣2(χ) such that

∕(x) = q↑(x)h(x) and g(x) = q1(x)h(x).

Therefore, we have

fl(x)∕(x) + b(x)g(x) = fl(x)z∕ι(x)fi(x) + b(x)q2(x)h(x) = [a(x)q1(x) + b(x)q2(x)]h(x),

so ∕z(x) is a factor of a(x)f(x) + b(x)g(x). □

Important: If ∕(x), g(x), and h(x) are polynomials such that ∕z(x) is a factor of both ∕(x)

ψ and g(x), then ∕z(x) is a factor of fl(x)∕(x) + b(x)g(x) for any polynomials
fl(x) and b(xf

656
 20.5. COMMON DIVISORS OF POLYNOMIALS

Let's put this fact to work on some problems.

Problem 20.25: Let p(x) = x2 + bx + c, where b and c are integers. If p(x) is a factor of both χ4 + 6x2 + 25
and 3x4 + 4x2 + 28x + 5, then find b and c. (Source: AHSME)

Solution for Problem 20.25: At first, we might approach this problem by trying to find rational roots of
both polynomials, and use these to find a common quadratic factor. Unfortunately, this strategy fails,
because neither x4 + 6x2 + 25 nor 3x4 + 4x2 + 28x + 5 has any rational roots. We could try to factor each
polynomial as a product of quadratics using the Method of Undetermined Coefficients, but instead, we
note that the result of Problem 20.24 offers a more attractive strategy. Perhaps we can combine the two
polynomials to produce a third polynomial that will allow us to find the desired quadratic quickly.
Specifically, we know that if r and s are constants, then any polynomial that divides both x4 + 6x2 + 25
and 3x4 + 4x2 + 28x + 5 also divides

r(x4 + 6x2 + 25) + s(3x4 + 4x2 + 28x + 5).

Choosing r = 3 and s = — 1 eliminates the x4 term and leaves us

3(x4 + 6x2 + 25) - (3x4 + 4x2 + 28x + 5) = 14x2 - 28x + 70 = 14(x2 - 2x + 5).

Therefore, the polynomial p(x) = x2 - 2x + 5 divides both x4 + 6x2 + 25 and 3x4 + 4x2 + 28x + 5, so b = -2
and c = 5. We can check our answer by performing polynomial division to find

x4 + 6x2 + 25 = (x2 - 2x + 5)(x2 + 2x + 5),

3x4 + 4x2 + 28x + 5 = (x2 - 2x + 5)(3x2 + 6x + 1).

Problem 20.26: For what values of b do the equations 1988x2 + bx+8891 = 0 and 8891x2 + bx +1988 = 0
have a common root? (Source: CMO)

Solution for Problem 20.26: We let ∕(x) = 1988x2 + bx + 8891 and g(x) = 8891x2 + bx + 1988. Finding a
common root of ∕(x) and g(x) is equivalent to finding a common linear factor of the two polynomials.
Any common factor of ∕(x) and g(x) is also a factor of the difference g(x) - ∕(χ), which conveniently
eliminates b and leaves us

g(x) - ∕(x) = 6903x2 - 6903 = 6903(x2 - 1) = 6903(x - l)(χ + 1).

Therefore, if ∕(x) and g(x) have a common linear factor, it is either x -1 or x+1. So, the common root of ∕(x)
and g(x) is either x = 1 or x = -1. When x = 1 is the common root, we have b = -1988 - 8891 = -10879.
When x = -1 is the common root, we have b = 1988 + 8891 = 10879. These are the two values of b for
which the given equations have a common root. □

Problem 20.27: The polynomials ∕(x) = x3 + ax2 + 7x-6 and g(x) = x3 - x2 + bx + 3 have two common
roots for some integers a and b. Find a and b.

657
CHAPTER 20. SOME ADVANCED STRATEGIES

Solution for Problem 20.27: The given cubic polynomials have the same leading coefficient, so their
difference is a quadratic whose roots are the two common roots of the cubics:

(x3 + ax2 +7x-6)- (x3 - x2 + bx + 3) = (a + l)χ2 + (7 - b)x - 9.

Unfortunately, we still can't find a and b, so we need more information.

Constructing a second polynomial with the common roots of the given cubics might help. Just as
we eliminated the cubic term in our first step, we could also eliminate the constant term. We have

∕(x) + 2g(x) = x3 + ax2 + 7x - 6 + 2(x3 - x2 + bx + 3) = x[3x2 + (a- 2)x + (2b + 7)].

The common roots of ∕(x) and g(x) must also be roots of ∕(x) + 2g(x). Since x is clearly not a factor of
either of the cubics, the roots of 3x2 + (a- 2)x + (2b + 7) must be the common roots of the given cubics.
We now have two quadratics, (a + l)x2 + (7 - b)x - 9 and 3x2 + (fl - 2)x + (2b + 7), with the same
roots. This means that one of them is a constant times the other. In other words, their corresponding
coefficients are proportional:
a + l_7-b_ -9
3 ~a-2~2b + 7'
From 2yi = we find that
6 = 7-∣(<j + 1)(<j-2) = —a2 + fl + 23
3
From 2yi = wehave (fl + l)(2b+7) = -27. Substituting for b and simplifying gives 2a3 - 69a -148 = 0.
Factoring gives (fl + 4)(2a2 - 8fl - 37) = 0. The discriminant of the quadratic is (—8)2 + 4 ∙ 2 ∙ 37 = 360,
which is not a perfect square. Therefore, the only integer solution for fl is fl = -4. Substituting this into
our expression above for b gives b = 1. We can check our work by factoring the given polynomials for
these values of a and b:

x3 — 4x2 + 7x — 6 = (x — 2)(x2 - 2x + 3),

x3 - x2 + x + 3 = (x + l)(x2 - 2x + 3).

Exer<ises

20.5.1 Determine all real numbers a such that the two polynomials x2 + ax + 1 and x2 + x + fl have at
least one real root in common. (Source: CMO)
20.5.2 Find a and b so that the equations x3 + ax2 + llx + 6 = 0 and x3 + bx2 + 14x + 8 = 0 have two roots
in common.

20.6 Symmetric Sums Revisited

We finish our study of polynomials by taking a look at more complex problems involving symmetric
sums. We've already seen with Vieta's Formulas that elementary symmetric sums of the roots of a

658
 20.6. SYMMETRIC SUMS REVISITED

polynomial are related to the coefficients of the polynomial. In this section, we,ll frequently combine
Vieta with the following concept:

Important: We can think of elementary symmetric sums as building blocks because

ψ any symmetric polynomial can be built from them.

Problems

Problem 20.28: Let a and b be the roots of the equation 2x2 - 10x + 23 = 0.
(a) Find the value of a2 + b2.
(b) Find the value of a4 + b4.

Problem 20.29: If a + b = 1 and α2 + b2 = 2, then find a3 + b3. (Source: AHSMEJ____________________

Problem 20.30: In this problem, we find all ordered pairs of solutions (x, y) that satisfy

2x + 3y + 6xy = 9,

2x2y + 3xy2 = —.

(a) Factor the expression 2x2y + 3xy2.

(b) Make substitutions that turn the left sides of the two equations into symmetric polynomials.
(c) Solve the equations you formed in part (b), then solve the original system.

Problem 20.31: Suppose that the sum of the squares of two complex numbers x and y is 7 and the sum
of their cubes is 10. What is the largest real value that x + y can have? (Source: AIME)

Problem 20.32: In this problem, we solve the system of equations

x + y + z = 8,
x2 + y2 + z2 = 62,
111 1
-x ^l-----
y z i^ - - “ To-
42

(a) Find the values of xy + yz + zx and xyz.

(b) Find all solutions (x, y, z) of the system of equations above.

Problem 20.33: Let a, b, c, d be complex numbers satisfying

cι + b + c + d = crl + b3 + c3 + dj = 0.

Prove that at least one pair of a, b, c, and d must add up to 0. (Source: Tournament of Towns)

659
CHAPTER 20. SOME ADVANCED STRATEGIES

Problem 20.28: Let a and b be the roots of the equation 2x*12 - 10x + 23 = 0. Find the value of α4 + b4. fj

Solution for Problem 20.28: The quadratic formula gives us x = ∣ + as the roots. We could raise these
to the fourth power using the Binomial Theorem and add the results. However, all that arithmetic (not
to mention the radicals and imaginary numbers) leaves a lot of room for error. Let's see if we can find
an approach that requires less complicated arithmetic.
By Vieta, the sum and product of the roots of 2x2 - 10x + 23 = 0 are a + b = 5 and ab =
Solution 1: We first find a2 + b2 using the given equations, and hope to use that as a stepping stone
to fl4 + b4. We need squares, so we square a + b and find that a2 + b2 = (a + b)2 - 2ab = 52 - 2 ∙ y =2.
Squaring a2 + b2 now gives us something very close to what we want:

22 = (α2 + b2)2 = at + 2a2b2 + b4 = a4 + b4 + 2(ab)2.

Substituting for ab, we have 22 = α4 + b4 + 2 (y)2 = a4 + bi + ψ, so α4 + b4 = 22 - φ = -5y∙

Our first solution builds up to α4 + b4 by first finding a2 + b2. However, a straightforward attempt to
expand (a + bf to get a4 + b4 leads to a different set of manipulations.
Solution 2: We start by expanding (α + i>)4 i∏ order to get an expression with a4 and b4 terms:
625 = (a + b)4 = a4 + 4a⅛ + 0a2b2 + ⅛ab3 + b4.

This expansion includes terms we don't need. In particular, we have

625 = α4 + b4 + 2ab(2az + 3ab + 2&2) = a4 + b4 + 23(2λ2 + 3ab + 2b2).

We need to determine the value of 2a2 + 3ab + 2b2. The squares lead us to squaring a + b. We have
(α + b)2 = a2 + 2ab + b2, so 2a2 + 3ab + 2b2 is ab less than 2(a + b)2:
23 77
2a2 + 3ab + 2b2 = 2(a + b)2 -ab = 2∙52 - — =

We therefore have 625 = a4 + b4 + 23 ∙ y, so α4 + fe4 = 625 — 23 ∙ y = -^y. □

Problem 20.29: If a + b = 1 and a2 + b2 = 2, then find a3 + b3. (Source: AHSME)

Solution for Problem 20.29: We could find a and b, and then compute a3 + b3 directly. However, if we
instead try to build a3 + b3 from a + b and a2 + b2, we will reduce the amount of computation we must
perform, and reduce the possibility of making an arithmetic error.
We need a sum of cubes, so the natural step is to multiply a + b and a2 + b2 to produce the cubes. This
gives us
1 ∙ 2 = (a + b)(a2 + fo2) = a3 + a2b + ab2 + b3 = a3 + b3 + ab(a + t>). (20.2)
The value of ab would finish off the problem. We know we get an ab from squaring a + b. This also gives
us a2 + b2, when we know equals 2. So, we have (a + b)2 = a2 + b2 + 2ab = 2 + 2ab. Because a + b = 1, we
have 1 = 2 + 2ab, so ab = -1/2. Equation (20.2) now gives us

2 = a3 + b3 + ab(a + b) =a3 + b3 + = fl3 + ~

660
 20.6. SYMMETRIC SUMS REVISITED

so a3 + b3 = ∣. □

Notice that in each of our first two problems, we squared expressions or multiplied expressions
without knowing for sure that the result would be useful. We saw that these manipulations would
produce some terms we needed, and hoped we'd be able to find a way to deal with the other terms that
were produced.

Concept: You don't have to see a path all the way to the end of a solution before
o≡≡≡i≡) taking your first step. Experiment! Use what you need as a guide to work
with what you have.

Solution for Problem 20.30: What's wrong with this start:

Bogus Solution: We apply Simon's Favorite Factoring Trick to the first equation to
∣X∣ gθt (2x + l)(3y + 1) = 10. So, we can use the factors of 10 to find
possible values of x and y.

The problem here is that x and y are not necessarily integers, so the prime factorization of 10 doesn't
help us find x and y.

Instead, we turn our factoring skills to the second equation, and find 2xλy + 3xy2 = xy(2x + 3y). The
expressions xy, 2x, and 3y are all present in the left side of the first equation in the problem. With this
in mind, we let t = 2x and u = 3y, which allows us to write the left sides of both equations as symmetric
polynomials:
f + u + fu = 9,
tu(t + u) _ 10
6 = ~3'
It's still not yet clear how to solve this system, but at least it involves symmetric expressions t + u and
tu, which we are a little more comfortable with than 2x + 3y + 6xy and 2xy2 + 3xy2.

Concept: A well-chosen substitution can turn an equation into a more familiar form.
O≡≡^

Substitution worked once, so seeing t + u and tu in both of our equations might inspire us to do it
again. We let s = t + u and p = tu, so we can now write our equations as
s + p = 9,
sp _ 10
6 ^ ^3^'

661
CHAPTER 20. SOME ADVANCED STRATEGIES

Therefore, we have s + p = 9 and sp = 20, which means that s and p are the roots of the quadratic
x2 - 9x + 20 = 0. So, we have (s,p) = (4,5) or (s,p) = (5,4).
If (s,p) = (4,5), we have t + u = 4 and tu = 5, which means that t and u are the roots of z2 - 4z + 5.
Applying the quadratic formula gives 2 ± z' as the roots, so (f, zz) = (2 + z, 2 - z) or (t, z√) = (2 - z, 2 + z). These
values of t and u give us (x, y) = (1 + ∣z, ∣ - jz) or (x, y) = (1 - ∣z, ∣ + jz).

Similarly, if (s, p) = (5,4), then t and u are the roots of z2 - 5z + 4, which are 1 and 4. These give us the
solutions (x, y) = (1, ∣) and (x, y) = (2, j). □

Concept: Any symmetric expression with two variables can be written in terms of
O~ uιJ the sum and product of those two variables. Therefore, problems involving
symmetric expressions in two variables can often'be solved by substituting
s for the sum of the variables and p for their product.

Let's see this strategy in action again:

Problem 20.31: Suppose that the sum of the squares of two complex numbers x and y is 7 and the
sum of the cubes is 10. What is the largest real value that x + y can have? (Source: AIME)

Solution for Problem 20.31: We're given two symmetric polynomial equations:
x2 + y2 = 7,
X3 + y3 = 10.
We seek the largest value of x + y, so we let s = x + y and try to write the equations in terms of s. We
have s2 = x2 + y2 + 2xy = 7 + 2xy, which reminds us that we can write symmetric polynomials with two
variables in terms of the sum and product of the variables. So, we let p = xy and we have s2 = 7 + 2p.
Turning to the other equation, we have x3 + y3 = (x + y)(x2 - xy + y2) = s(s2 - 3p), so s(s2 - 3p) = 10.

Our equations in s and p aren't all that nice, but solving s2 = 7 + 2p for p gives us p = We can
use this equation to substitute for p in the second equation, giving us a cubic equation in s:
sP-3∙⅛Z) = 10,

Simplifying this equation gives s3 - 21s + 20 = 0. Using the Rational Root Theorem, we factor the left
side to find (s - l)(s - 4)(s + 5) = 0. The largest possible value of x + y = s is therefore 4. □
Now that we have some experience manipulating two-variable symmetric polynomials, let's try
more variables.
Problem 20.32: Solve the following system of equations:

x + y + z = 8,
x2 + y2 + z2 = 62,
111 1
---- 1------ 1---- —
x y z 42

662
 20.6. SYMMETRIC SUMS REVISITED

Solution for Problem 20.32: The three symmetric polynomials on the left remind us of Problem 20.21,
in which we let the variables be the roots of a polynomial, then found that polynomial using Vieta's
Formulas. But in this problem, only one of our symmetric polynomials can be directly related to a
polynomial via Vieta. We are given x + y + z = 8, but we don't know xy + yz + zx or xyz. Can we find
them?

Concept: Start new problems by asking yourself if you have solved a similar problem
Q ljiJ before. If you have, then try to turn the new problem into the same problem
as the old problem. _

Squaring x + y + z = 8 will produce xy, yz, and zx terms. This gives us

82 = x2 + y2 + z2 + 2xy + 2yz + 2zx = 62 + 2(xy + yz + zx),

so xy + yz + zx = 1. Success!
We'll probably have to use the third equation for something, so we start by writing j + ∣ + ∣ whh a
common denominator:
1 1 1 1 xy + yz + zx
42 x y z xyz
We know that xy + yz + zx = 1, so we have xyz = -42.
Now, by Vieta's Formulas, x, y, and z are the roots of the cubic
t3 - (x + y + z)f2 + (xy + xz + yz)t - xyz = t3 - 8f2 + t + 42.

This factors as (f - 3)(f - 7)(f + 2) = 0. Therefore, our six solutions are (x, y,z) = (3,7, -2), and each of the
other five permutations (orders) of (3,7, -2). □

Problem 20.33: Let a, b, c, d be complex numbers satisfying

a + b + c + d — a3 + b3 + c3 + d3 = 0.

Prove that a pair of the fl, b, c, d must add up to 0. (Source: Tournament of Towns)

Solution for Problem 20.33: It's not at all clear even where to start, so we'll have to play around a little.
We'll start by brainstorming a little bit, and see what we come up with.

Concept: On a tough problem, it often pays to start off by listing a bunch of ideas
Q about how to approach the problem. You may not start off with the right
one, but this brainstorming process will help prevent you from getting
stuck in one track of thought. It also might feed you ideas once you start
working with the problem more.

Keeping in mind that we're looking for information about sums of pairs of variables, let's brainstorm:
Try 1: Rearrange and factor. We can rearrange a + b + c + d = a3 + b3 + c3 + d3, then factor, to get

fl(fl - l)(fl + 1) + b(b - T)(b + 1) + c(c - l)(c + 1) + d(d - l)(d + 1) = 0.

663
CHAPTER 20. SOME ADVANCEDSTRATEGIES

No sums of variables. Probably not useful.

Try 2: Factor sums of cubes. We have a3 + b3 + c3 + d3 = (a + b)(a2 - ab + b2) + (c + d)(c2 - cd + d2). We
subtract a3 + b3 + c3 + d3 = a + b + c + d from this equation to give

(fl + b)(a2 - ab + b2 - 1) + (c + d)(c2 - cd + d2 - 1) = 0.

It's not clear where to go from here. We have sums of variables, which is good, but we have nasty terms
we don't know what to do with, which is bad.
Try 3: Rearrange a+b+c+d=Q and cube. We have a + b = -(c + d), which is good because we have
sums of pairs of variables. Then, we cube, so we can use our information about the cubes. This gives us

fl3 + 3a2b + 3ab2 + b3 = -c3 - 3c2d - 3cd2 - d3.

Since we are given a3 + b3 = -c3 - d3, we have 3a2b + 3ab2 = -3c2d - 3cd2. A little factoring gives

flb(fl + b) = -cd(c + d).

Combining this with a + b = -(c + d) gives us either a + b = c + d = 0, in which case we are done, or ab = cd.
It's not too clear what we'll do with this. It tells us about products of the variables instead of sums, but
it seems like it might be important. We might spend a few more minutes here exploring where ab = cd
will lead us, but if we don't get anywhere, it's back to brainstorming.

WARNING!! If you find a promising method of attack on a problem, explore it, but
*Z don't get stuck to it. If it doesn't work out after a little while, go back
to exploring the problem from different angles.

Try 4: Rearrange differently before cubing. Although a + b = -(c + d) didn't give the answer right away,
it did seem to give us progress. So, we feel close. We try rearranging a little differently, and we write
a = -(b + c + d). Cubing this is a bit tricky, but we find
a3 = -b3 - c3 -d3 - 3b2c - 3c2b - 3b2d - 3d2b - 3c2d - 3d2c - 6bcd.

We know that a3 = -b3 -c3 -d3, so we can get rid of those terms, then divide the equation by -3 to have

b2c + c2b + b2d + d2b + c2d + d2c + 2bcd = 0.

Before we get scared and quit, we remember what we're trying to prove. We're trying to prove that one
of the sums of a pair of fl, b, c, and d equals 0. Obviously, the messy equation we just hit has nothing
to do with a, but it does have b, c, and d. We'd be happy if we could show that b + c = 0, which makes
us wonder, "Is b + c a factor of the messy expression?" We have a way to check! We let b = -c in the
expression, and we have
b2c + c2b + b2d + d2b + c2d + d2c + 2bcd = c3 - c3 + c2d - d2c + c2d + d2c - 2c2d = 0.

Aha! We see that b + c is a factor. The polynomial is symmetric, so we aren't surprised to find that b + d
and c + d are factors, as well, and we have

b2c + c2b + b2d + d2b + c2d + d2c + 2bcd = (b + c)(c + d)(d + b).

664
 20.7. SUMMARY

So, if a + b + c + d = a3 + b3 + c3 + d3 = 0, then our manipulations above show that we must have

(b + c)(c + d)(d + b) = 0, which means that one of the sums of a pair of b, c, and d equals 0. □

Exercises

20.6.1 Ifx + y + z = l and - + 1 + i = 0, then find x2 + y2 + z2.

20.6.2 If xy = 5 and x2 + y2 = 21, then compute x4 + y4. (Source: HMMT)

20.6.3 Find all pairs of real numbers (x, y) such that x2 + y2 = 34 and x3 + y3 = 98.

20.6.4 Show that two of the sums of pairs of a, b, c, and d must equal 0 in Problem 20.33.
20.6.5 Find all solutions of the equations
x + y + z = 2,
x2 + y2 + z2 = 14,
xyz = -6.

20.6.6
* Find x, y, and z:
x ÷ y ÷ z = 6,
x2 + y2 + z2 = 26,
x3 + y3 + z3 = 90.

20.7 Summary

Concepts: ∙ Look everywhere for patterns.

• If you don't immediately see how to work with an expression, try

rewriting it. Many problems can be simplified by a different look at
the same information.

• The substitution y = x + j can help us simplify problems involving

polynomials in x with symmetric coefficients.

• When stuck trying to deal with an algebraic expression, focus on any

unusual aspects of the expression.

• Use substitutions to turn complicated equations into simple ones.

Repeated expressions within an equation are a clue that substitution
might be helpful.

• Don't make problems harder than they have to be. When you have
multiple options for substitution, choose the one that appears to offer
the easiest path.

Continued on the next page. . .

665
CHAPTER 20. SOME ADVANCED STRATEGIES

Concepts: . . . continued from the previous page

θ ∙ A homogeneous equation is one in which all nonzero terms have

the same degree. When working with a two-variable homogeneous

equation, it's often helpful to set the ratio of the variables equal to a
new variable and write the equation in terms of the new variable.

• When stuck on a problem, focus on information you haven't used

yet.

• Try to turn new problems you don't know how to do into old prob­
lems you know how to do.

• Don't get intimidated by complicated notation. Take your time,

read the problem closely, and play with simple examples. Often,
complicated notation is just shorthand for a very simple idea.

• When working with an unknown polynomial, it is often useful to

assign variables to the coefficients of the polynomial, and then con­
struct equations in terms of those variables.

• Just as we can find the values of symmetric sums given a polynomial,

we can construct polynomials given symmetric sums.

• If a tactic gives some information in a problem, but doesn't com­

pletely solve it, try using the tactic again.

• You don't have to see a path all the way to the end of a solution
before taking your first step. Experiment! Use what you need as a
guide to work with what you have.

• On a tough problem, it often pays to start off by listing a bunch of

ideas about how to approach the problem. You may not start off with
the right one, but this brainstorming process will help prevent you
from getting stuck in one track of thought. It also might feed you
ideas once you start working with the problem more.

review Problems

20.34 Solve the system of equations

(x + y)(x + y + z) = 15,
(y + z)(x + y + z) = 6,
(z + x)(x + y + z) = -3.

20.35 Find three constants p, q, and r such that x4 + 4x3 - 2x2 - 12x + 9 = (px2 + qx + r)2 for all x.

666
 REVIEW PROBLEMS

20.36 Find all m for which x2 + 3xy + x + my - m has two linear factors with integer coefficients (Source
AHSME)

20.37 A polynomial P(x) has four roots, ξ, 5, 2,4. Find P(0) if P(l) = 1. (Source: HMMT)
20.38 Find all ordered triples (a, b, c) that satisfy the system

a + b + c = 4,
ab + bc + ca = -4,
abc = -21.

20.39 Solve for x where √4x - 3 +------ ---

√4x-3
20.40 Solve for β17 if

$1 ÷ #2 4" ^3 — 1/

¾ ÷ ¾ ÷ ¾ = 2z

fl16 ÷ fl17 + — 16,

#17 ÷^1 + ⅛ = 17.

(Source: Mandelbrot)

20.41 Find all ordered triples (x, y, z) of real numbers that have the property that each number is the
product of the other two.

20.42 What is the product of the real roots of the equation x2 + 18x + 30 = 2 Vx2 + 18x + 45? (Source:
AIME)
20.43 The equation x4 - 16x3 + 94x2 + px + q = 0 has two double roots. Find p + q.
20.44 Let ∕(x) = 2x3 - 3x2 + 2x - 1 have roots r, s, and t.

1
Evaluate---- - H------ - +
r-2 s-2 f-2'
1 1
(b) Evaluate - ---- - + 2s + 1 + 2t + 1'
*
2r + 1
20.45 Let P(x) = x3 + x2 - 1 and Q(x) = x3 - x - 1. Given that r and s are two distinct solutions of
P(x) = 0, prove that rs is a solution of Q(x) = 0. (Source: MMPC)
20.46 Real numbers a, b, c satisfy the equations a + b + c = 26 and ∣ + I + ∣ = 28. Find the value of

a b c a c b
7 4------ 1------ 1------ F
b c a c
(Source: HMMT)

667
CHAPTER 20. SOME ADVANCED STRATEGIES

20.47 Find all ordered triples (x, y, z) that satisfy the system

√x + √y + √z = 10,
x + y + z = 38,
√xy + √xz + Vyz = 30.

20.48 Find the four values of x that satisfy (x - 3)4 + (x - 5)4 = -8. (Source: ARML)

20.49 Given that λ∕2x2 - 3xy + 64 + y,2x2 - 3xy = 16, find the value of √,2x2 - 3xy.

20.50 Find x2 + y2 given that x and y are natural numbers such that xy + x + y = 71 and x2y + xy2 = 880.
(Source: AIME)
20.51 If a, b, c, d are the solutions of the equation x4 — bx - 3 = 0, then find a polynomial whose roots
are the numbers
a+b+c cι + b + d cι + c + d b+c÷d

(Source: AHSME)
20.52 The equations x3 - 7x2 + px + c∣ = 0 and x3 - 9x2 + px + r = 0 have two roots in common. If the
third root of each equation is represented by xj and Xi, respectively, compute the ordered pair (X1,X2)∙

20.53 What is the sum of the solutions to the equation -tyx =----- — ? (Source.

20.54 Find all ordered pairs (x, y) such that 2x2 + y2 + 3xy = 12 and (x + y)2 - y = 7.
20.55 If ∕(x) is a ninth degree polynomial such that f(n) = ⅛l for n = 1,2,3,..., 10, then find the value
of∕(ll).

Since this is the final group of Challenge Problems, we've mixed in problems that require strategies
from some other chapters, as well.
20.56 Show that there exists a polynomial ∕(x) with rational coefficients such that ∕(√2 + √3) = √2.
Hints: 38
20.57 Let x, y, and z be nonzero real numbers such that

111,
X+V + Z= -H-------- F- = l.
x y z

Prove that at least one of x, y, and z is equal to 1.

668
 CHALLENGE PROBLEMS

20.58 If x = 2007(α -b),y = 2007(b -c),z = 2007(c - a), and xy + yz + zx≠0, then compute the numerical
value of
x2 + y2 + z2
xy + yz + zx
Hints: 123

20.59 Find all real solutions of (2x - 4)3 + (4λ' - 2)3 = (4x + 2x - 6)3. (Source: AHSME) Hints: 116
20.60
* Let a, b, c, and d be distinct real numbers such that a>b>c>d>0, abed = 1, and

ι, j 1 1 1 1
a+b+c+d= - + - 4----- 1- -.
abed
Show that ad = bc = l. Hints: 80, 337

20.61 Find the real values of x that satisfy the equation (x + l)(x2 + l)(x3 + 1) = 30x3.
20.62 For each positive number x, let

∕(x)=H)6-(
* 6÷⅜)-2.

(i'+93+L3 + ⅛)
Find the minimum value of ∕(x). (Source: AHSME) Hints: 225
20.63 Suppose that x, y, and z are three positive numbers that satisfy the equations

xyz = 1,
1 c
x + - = 5z
,,L≈,
j X

Find the value of z + l∕y. (Source: AIME) Hints: 343

20.64 Find all ordered pairs (x, y) such that 3xy - x2 - y2 = 5 and 7x2y2 - χ4 — y4 = 155.
20.65 Find the value of
a b c ------ _[_ ------ -∣- ------
a 4-1 b 4-1 c + 1
given that a + b + c = -1, ab + be + ca = -1, and none of a, b, or c is equal to -1. Hints: 54, 348
20.66 Suppose that the roots of x3+3x2+4x-11 = 0areα,b, and c, and that the roots of x3 + rx2+sx+t = 0
are a + brb + c, and c + a. Find t. (Source: AIME) Hints: 196
20.67

(a) Let x, y, and z be real numbers such that x + y + z, xy + xz + yz, and xyz are all positive. Prove that
x, y, and z are also positive.
(b) Is it possible for x + y + z, xy + xz + yz, and xyz to be positive real numbers if x, y, and z can be
nonreal?

669
CHAPTER 20. SOME ADVANCED STRATEGIES

20.68 Let P(x, y) be a polynomial in x and y such that

(i) P(x, y) is symmetric, i.e., P(x, y) = P(y, x) for all x and y,

(ii) x - y is a factor of P(x, y), i.e., P(x, y) = (x- y)Q(x, y) for some polynomial Q(x, y).

Prove that (x - y)2 is a factor of P(x, y). (Source: Canada)

20.69 If p, q, and r are the roots of x3 - 2x + 5 = 0, find an equation with roots + ∣, ∣ + ∣, and ∣ + p∙
Hints: 185
20.70
* A polynomial P(x) of degree n satisfies P(k) = ∖∣kAcrk = 1, 2, 4, 8,..., 2”. Find P(0). (Source:
Poland)
20.71 A sequence of numbers Xi, X2, X3, ■ ■ ∙, X100 has the property that, for every integer k between 1
and 100, inclusive, the number x⅛ is k less than the sum of the other 99 numbers. Find X50∙ (Source:
AIME) Hints: 168
20.72 Solve for x and y:

(x + y)2/3 + 2(x - y)2/3 = 3(x2 - y2)ιz3,

3x-2y = 13.

Hints: 255
20.73 Let p(x, y) be a polynomial function whose domain and range are the real numbers. Prove or find
a counterexample to the statement, "If [p(x, y)]2 is a symmetric function, then p(x, y) is also a symmetric
function."
20.74 Show that every symmetric two-variable polynomial in x and y can be written as a polynomial
in s and p, where s = x + y and p = xy. Hints: 227,203,350
20.75 Find all solutions to the system

Λ + ⅛ + C = 1,
α2 + b2 + c2 = 2,
β4 + b4 + c4 = 3.

20.76 Let x, y, z be real numbers such that x + y + z≠0. Prove that if

*(y - z) + y(z - *) + & - y) = 0

y+z z+x x+y

then two of x, y, and z must be equal. Hints: 83

20.77 The product of two of the four roots of the quartic equation

x4 - 18x3 +kx2 + 200x - 1984 = 0

is -32. Determine the value of k. (Source: USAMO) Hints: 329, 21, 71

670
 challenge problems

20.78
* Find all polynomials p such that p(χ + 2) = p(x) + 6x2 + 12x + 8.

20.79
* Let f(x) — a2nX ÷ ^2∏-iΛ' , + ∙ ∙ ∙ + mx + Λq, where fl2∏ θ and ci2∏-i = ∩i for all i such that
0 ≤ z ≤ 2n. Show that there is a polynomial g(χ) such that the solutions of g(x + ⅛) = 0 coincide with the
solutions of ∕(x) = 0. Hints: 206, 9
20.80
* Find αx5 + by5 if the real numbers a, b, x, and y satisfy the equations

ax + by = 3,
ax2 + by2 = 7,
ax3 + by3 = 16,
ax4 + by4 = 42.

(Source: AIME) Hints: 201

20.81
* Let Xχ, x2, and x3 be the roots of x3 - 2x2 - x + 1 =0. Let y1∙ = x2 + x1∙ + 1. Find the monic cubic
whose roots are y1, y2, and y3. Hints: 238,103

20.82
* Find all solutions to the equation (z2 - 3z + 1)2 - 3(z2 - 3z + 1) + 1 = z. Hints: 96
*
20.83 Let p(x) = χ5 + x2 + 1 have roots r1, r2, r3, r4, r5. Let q(x) = x2 - 2. Determine the product
q(r1)q(r2)q(r3)q(r4)q(r5). (Source: USAMTS) Hints: 297,223
20.84
* Curves A, B, C, and D are defined in the plane as follows:

A is the graph of x2 - ι∕2 = ~z~—^,

6 r y x2 + y2

C is the graph of x3 - 3xy2 + 3y = 1,

D is the graph of 3x2y - 3x - y3 = 0.

Prove that the intersection of curves A and B is the same as the intersection of curves C and D. (Source:
Putnam) Hints: 187
20.85
* Let a, b, c, d be distinct real numbers such that

a = x[4^+ Vδ + «,
a
b = a/T- √5 +^∕>,
b=
c = χ[4 + y/5
c,

Find abed. (Source: Brazil) Hints: 63

671
CHAPTER 20. SOME ADVANCED STRATEGIES

20.86* Determine x2 + t∕2 + z2 + w2 if

x2 y2 z2 w2
22 - 12 + 22 - 32 + 22 - 52 + 22-72
x2 y2 z2 w2
42 - 12 + 42 - 32 + 42 - 52 + 42 - 72
x2 y2 z2 w2
6r≡l2 + 62≡32 + 62 - 52 + 62^72
x2 y2 z2 w2
82≡712 + 82 -32 + 82 -52 + 82-^72

(Source: AIME) Hints: 132

Extra! The Art of Problem Solving's Intermediate Series of textbooks continues with Intermediate
∣∣∣⅜l∣n*.i lιΦ∣∣∣^ Counting & Probability by David Patrick.
Intermediate Counting & Probability is the sequel to the acclaimed Introduction to Count­
ing & Probability textbook that has been used by thousands of students. Some of the
topics covered in Intermediate Counting & Probability are:

• The Principle of Inclusion and Exclusion

• 1-1 correspondences

• The Pigeonhole Principle

• Distributions

• Fibonacci and Catalan numbers

• Recursion

• Conditional probability

• Generating functions

• Graph theory

And many more! The book contains over 650 problems and exercises.

672
 references

In addition to the books cited below, there are links to many useful websites on the book's links page at

http:∕∕www.artofproblemsolving.com/BookLinks/IntermAlgebra/links.php

1. J. Derbyshire, Unknown Quantity, Penguin Group, 2006.

2. A. Engel, Problem-Solving Strategies, Springer-Verlag, 1998.

3. D. Fomin, S. Genkin, and I. Itenberg, Mathematical Circles (Russian Experience), American Mathe­
matical Society, 1996.

4. L. Larson, Problem-Solving Through Problems, Springer-Verlag, 1983.

5. S. Lehoczky and R. Rusczyk, the Art of Problem Solving, Volume 1: the Basics, 7th edition, AoPS
Incorporated, 2006.

6. S. Lehoczky and R. Rusczyk, the Art of Problem Solving, Volume 2: and Beyond, 7th edition, AoPS
Incorporated, 2006.

7. D. Patrick, Intermediate Counting & Probability, AoPS Incorporated, 2007.

8. D. Patrick, Introduction to Counting & Probability, 2nd edition, AoPS Incorporated, 2007.

9. R. Rusczyk, Introduction to Algebra, AoPS Incorporated, 2007.

10. P. Zeitz, The Art and Craft of Problem Solving, 2nd edition, John Wiley & Sons, Inc., 2007.

673
hints to selected problems

--------- 1
Hints to Selected Problems

1. Note that 3 and -5 are both odd.

2. You've probably already moved all the terms to one side, and determined that there aren't any
linear factors with rational coefficients. So, what kind of factors with rational coefficients must
there be?

3. Let ∕ and w be the length and width of the top of the table, and let h be the height of the table. Find
lι in terms of ∕ and w. Express the volume in terms of h.

4. After rearranging due to the first hint, you probably have n + 1 terms on the greater side. One of
them should stand out from the others. How can you split that term up, and join one piece with
each of the other n terms?

5. Let d = (a + b + c)∕3.

6. It is wise to generalize. Find gn(x).

7. Let r be the root you found in the first hint. Now, go backwards. Find the polynomial with roots a,
a2 + 2a - 14, and r if a is a root of the given quadratic. (You can't assume anything about the latter
two roots!)
8. What happens if you let one sequence in Cauchy-Schwarz be √fl, ∖[b, ∖fc, y/d?

9. Manipulate the equation ∕(x) = 0 such that its highest power of x is the reciprocal of its lowest
power of x. Pair off the terms, each power of x with its reciprocal. Can you express the sum of each
pair in terms of x + |?

10. Compare the sum tol + ⅛ + ⅛ + ∙∙∙ + Can you find a simple expression that is equal to this
sum with experimentation and induction?

11. Consider the product (fl + b + c)(± + ⅛ + ±).

12. What is (x + 1)3 - x3?

674
 HINTS TO SELECTED PROBLEMS

13. How many times does p∙ appear? How about ^?

14. Complete the square.

15. Such as y = √x.

16. Start with the 4-variable AM-GM applied to a, b, c, and d. Choose an expression for d in terms of a, b,
and c that will give us the AM of a, b, and c on the greater side of the 4-variable AM-GM Inequality.

17. Find the first few terms. See a pattern? Prove it!

18. Write the product with ∏ notation. You should have a polynomial. What do we usually do to learn
more about a polynomial?

19. You used AM-GM for 4 variables to prove it for 3 variables. You should already have proved
AM-GM for 23 variables, and for 24 variables, etc.

20. Can you find ∕(α) in terms of a, p, and q?

21. Can you factor the polynomial now using what you know about products of pairs of roots?

22. 331 is way bigger than 231, and 329 is way bigger than 229. So, the fraction is probably very close to
331∕329 = 9. But is it a little larger, or a little smaller? (If a little larger, you must show it is less than
10; if a little smaller, you must show it is more than 8.)

23. Once you have an equation, simplify it as much as possible. Do you see any terms that can be
grouped and factored?

24. Focus on the four key points: the two ends of the graph and the two points where the graph
changes direction. Where are these points on the new graph?

25. For what values of x is f defined? Is f monotonically increasing or monotonically decreasing for
these values of x? Is there a smallest value in the domain of /?

26. Multiply by 1 - x.

TJ. What expression must 36%2 + 25y2 be greater than if x and y have the same sign? What if they have
opposite signs?

28. Can you express p3 in terms of p2 and p?

29. What can you do to the expression you must minimize in order to produce squared terms?

30. If f(a) = f(b), where 1 ≤ a, b ≤ 5, then what two possibilities are there for how a and b are related?
How can we find solutions to ∕(∕(x)) for each of these cases?

31. When a strategy works a little, try it again. Let p(x) = d(x)q(x), where d(x) is the factor you already
found. Find a factor of q(x).

32. Your setup should have plenty of sums, as well as square root signs to get rid of. Squaring gets rid
of square root signs. Sums and squaring—what inequality should help with that?

33. It would probably be easier if all the logarithms had the same base. Notice that 23 = 8.

675
HINTS TO SELECTED PROBLEMS

34. Let c = b - 1.

35. Try some specific examples, such as ∕(x) = x3 or ∕(x) = x + 3.

36. Suppose P and Q are on f and ∕-1, respectively, such that PQ is the desired shortest distance. So,
we have P = (a, /(a)) for some a. Can you express the coordinates of Q in terms of a?

37. Focus on just a pair of variables. Can you create an inequality that relates (« - b)2 to a2 + b2?

38. It's a puzzle! Compute a few powers of V2 + V3. Try to fit them together to get V2.

39. How are the graphs of f and ∕^^1 related?

40. What is the smallest possible value of ab?

41. Because a is a root of the polynomial, we have «3 + 3«2 - 24« + 1 = 0. We'd like to find a simple
expression for -ζ∕α. Reorganize «3 + 3«2 — 24« + 1 = 0 such that« is on one side, and the cube of a
polynomial is on the other.

42. Think about the polynomial P(x) - 10x.

43. We know a lot more about finding the roots of polynomials whose coefficients are all real.

44. What is f(x + y)?

45. Express f(x) as the product of two quadratics that are either monotonically increasing or monoton­
ically decreasing.

46. Rearrange the equation, and remember that« and b are nonzero. You should get a quadratic. What
must be true about a quadratic if it has integer roots?

47. Let a = log225 x and b = log64 y. Rewrite the given equations in terms of a and b.

48. You probably already tried using the fact that 1 - x is cyclic. And that failed. Can you try another
tactic from the chapter?

49. Compare ∕(x) to ∕(-x)∙ For what values of x must these polynomials be equal?

50. Find a polynomial /(«) such that a5 - a2 + 3 > /(«), and f(a) is easier to work with than a5 - a2 + 3.
Note that a5 - a2 = «2(«3 - 1)∙

51. Let a be the first term and r be the common ratio, and write the given information in terms of a and
r. How can you factor r4 + r2 +1? (If you don't remember, then try thinking of a notable polynomial
that has r4 + r2 + 1 as a factor.)

52. Suppose you wrote an equation for each root. What can you do with these 20 equations to produce
an equation that contains the sum of the 20th powers of the roots?

53. Each term in the series looks a lot like our formula for the sum of an infinite geometric series. Find
infinite geometric series that correspond to each.

54. Can you find a substitution that simplifies the expression you must find, but doesn't greatly
complicate the given equations?

676
 HINTS TO SELECTED PROBLEMS

55. Suppose you can find the length of wire needed to create the portion of the graph with x ≥ 0 and
xJ ≥ 0∙ How can you then quickly find the answer?
56. What is true about the diagonals of a parallelogram?

57. How many different integers can appear in the sum on the right side?

58. Suppose the circle is the graph of x2 + t∕2 = r2, and let P be (p, 0).

59. Write the equation for the line in terms of w, then let w = a + bi.

60. Experiment! Compute a few of the z1∙. Notice anything interesting?

61. Find pattern, prove pattern. Experiment!

62. Let x = λ∕49 + ^∖∣7. How do you get rid of cube roots?

63. If we could find a polynomial with roots a, b, c, and d, we'd be done. But it's not so obvious how to
do that. Can we at least find a polynomial with a as a root?

64. What expression can you add to (15 + √220) that, when both are expanded, will eliminate all the
terms?

65. Forget about the numerators for a minute. Deal first with j⅛ + ⅛⅛ + ∏⅛∙

66. Place the problem on the Cartesian plane. Let A be (-2,0) and B be (2,0). Where is C? What does
AX + XB = AC + CB tell us about point X?

67. How can you convert the problem to exponential form?

68. First, consider what inputs are allowed to f. Then, consider what inputs are allowed to g, then
determine what inputs to f give outputs that are valid inputs to g.

69. Expand (x + 1)3.

70. Write the expression as a sum of positive powers of z.

71. A product of a pair of roots won't tell you much about a linear factor of the polynomial. What type
of factor will it tell you something about?

72. Combine the previous two problems in a way-cool induction. Start from AM-GM for 2k variables,
and use the same tactic you used to prove AM-GM for 3 variables to prove it for 2k - 1 variables.
Can you then set up an induction to show that it holds for 2k — 2 variables, 2k — 3 variables, and so
on?

73. An easy way to divide 23149 by 7 is to break into pieces that are divisible by 7. We have 23149 =
21000 + 2100 + 49, so dividing by 7 gives 3000 + 300 + 7 = 3307. Can you do something similar here?

74. If a and a2 + 2a - 14 are roots of the cubic, what must the third root be?

75. For both inequalities, try completing the square in x and y. You should recognize one of the results.
You may have to factor the other.

677
HINTS TO SELECTED PROBLEMS

76. We have 10 solutions to P(x) - = 0. But that's not a polynomial on the left side.

77. What goes up must come down. But when?

78. Experiment. Try various small values of n.

79. A substitution makes the algebra easier. Let a = λ∕60 - x and b = ^∕x - U∙

80. There are a whole lot of elementary symmetric polynomials (or parts thereof) in this problem.

81. Sum of squares; try Cauchy.

82. Experiment. Find several terms of the sequence in terms of a↑ and «2- Can you find the two given
sums in terms of a↑ and az!

83. What equation is equivalent to the statement, "Two of x, ∖ j, and z must be equal?" Use the answer
to this, along with the restriction x + y + z ≠ 0, as inspiration to find a factored form of the left side.

84. Can you pair the factors so that you can use the quadratic to evaluate the results of expanding the
product of each pair?

85. Experiment. Choose simple values of p, q, and r, and look for a pattern in the factors of the resulting
numbers.

86. Express both sides as a single base 10 logarithm.

87. It's not as hard as it looks. Think about it before cubing.

88. All the pieces are linear, so what's a good guess for the form of each of the polynomials?

89. Add 1 to each of the three ratios in the first given equation.

90. Let c = a2.

91. Can you factor the cubic? Try factoring x2 out of the first two terms. Can you then factor further?

92. What is (x - l)(x - 2)(x - 3)?

93. Let x = 2y. What is in terms of x?

94. What happens if you multiply the product by 1 - 5?

95. We'd like to get rid of the cube roots. The polynomial we seek probably has zz3, b3, and c3 terms. Do
some wishful thinking. Find g(x, y, z) and ∕z(x, y, z) such that x3+y3 +z3+g(x, y, z) = (x+y+z)∕z(x, y, z),
and g(a, b^2,c y∕4) is a polynomial with integer coefficients.

96. Make the obvious substitution for simplification. You will then have two equations; the result of the
substitution and the variable definition that you used to make the substitution. What's interesting
about those equations? Can you combine them in any interesting ways?

97. In terms of n, how many integers satisfy the inequality chain from the first hint?

98. Experiment. Evaluate ∣xj - 19 ∙ [⅛] for different positive integers x. Notice anything interesting?

678
 hints to SELECTED PROBLEMS

99. Rearrange and factor the given equation.

100. Yikes, that's ugly. Let a = x2 - yz, b = y2 - zx, and c = z2 — xy. (But don't forget what a, b, and c
stand for as you work!)
101. All five choices are nonreal, so one of the factors of the polynomial is a quadratic with nonreal
roots. Which of the given choices can be a root of a monic quadratic with integer coefficients?

102. Let (yι, y2, y3, y½) = (1,0,0,0). What does this suggest as a possible /(«, b, c, d)2 Confirm it works.

103. Find x∣∙ in terms of y1∙. Don't forget what the xz∙ are!

104. What happens if you add all three equations?

105. Express a3 in the form ca↑ + d, where c and d are constants.

106. What is a3 + b3? Use the given equation to find a3b3.

107. You care most about e. Write Cauchy in a way that allows you to substitute expressions in terms of
e, to produce an inequality that has e as the only variable.
108. Use the three rules to find f(ri) for various values of n. Can you use your experimentation to prove
a pattern?
109. The polynomial g(x) should look familiar. Find a linear expression x — c such that (x — c)g(x) is nice
to work with.
110. Find each sum in terms of a and r. What happens to the desired sum? Do you find any more
geometric series?
111. Graph S and g, then use transformations of the graph of S to build the graph of g.

112. Find two quadratics whose product is the given quartic.

113. Consider the portion of the graph with x ≥ 0 and y ≥ 0. Why does this simplify the equation? Can
you repeat your first step? In other words, can you reduce the problem to considering a smaller
portion of the graph that, if you could find the length of wire needed for that portion of the graph,
then you could find the answer quickly?

114. Suppose r is a solution to p(p(r)) = r. What might p(r) be?

115. Multiply both sides by the product of the denominators.

116. What is interesting about the three quantities in parentheses; how are they related?

117. Write the bases in terms of 2 and 3. Let x = 2 and y = 3, and factor.

118. Write the left side as the difference of two sums.

119. Express the original equation in standard form. Can you express the equations corresponding to
the parabola's reflection and both translated parabolas in standard form?

120. Solve a↑x + b↑y + c↑z = 0 for z.

679
hints to selected problems

121. What is the result of rotating the point (a, b) 90 degrees counterclockwise about the origin?

122. For what x and y does (a3b2c)x(a⅛3c5)y = α⅛7c147

123. The expression we must evaluate is symmetric. What simple symmetric expression might be useful
to form from the three given equations?

124. What icky denominators! How can you get the square roots out of them?

125. Let b be the number bordering 0 to the right, a be the number one box above 0, and let c be the
number in the other box adjacent to both a and b. Find other boxes in terms of these three.

126. If we multiply both sides by a + b + c + d, we have a product of two sums. How will we then apply
Cauchy?

127. Group the x3 and x2 terms, and group the terms that are linear in x and those without x. Factor.
Then factor again. Then again.

128. Include an extra term in the numerator such that the denominator is a factor of the numerator.

129. How can you get the sum of all the coefficients? How can you eliminate those with odd subscripts?

130. Sum the first six rows and look for a pattern. Then, try to explain why the pattern exists.

131. Let n = 2m + bo, where ⅛ is the last digit of the binary representation of n. What happens to f(ri) if
bo is 0? If it is 1?
132. Can we introduce a single variable, then express all four equations as a single equation in which
letting the new variable take on 4 different values produces the four given equations?

133. Vieta doesn't appear to help with the difference of the roots. But what about the square of the
difference of the roots?
134. Complete the square. (That means write the quadratic in the form a(x -h)2 +k for some constants
a, h, and k. If you're not familiar with completing the square, don't worry about this part. We'll
review completing the square in a Chapter 4.)

135. A proof of a statement for all positive integers n....

136. Experiment. What is P↑(x) in terms of Po? What is P2(x) in terms of Po? How about Pι(x)2 Notice
anything interesting?
137. How many copies of xn' do you need to cancel with how many dr in order to have a constant GM?

138. For any polynomial ρ(x), how is the degree of p(x + 1) - p(x) related to degp? What substitution
does this suggest?

139. Factor. Write out a few terms. See anything interesting?

140. When stuck, go back to the beginning. Get back to your basic series-evaluation tools.

141. Shift that index.

680
 HINTS TO SELECTED PROBLEMS

142. Graph the piecewise function in the problem. The result should resemble other absolute value
graphs you have seen. In those other graphs, how have the "corners" of the graphs been related to
the functions the graphs represent?

143. The a2 + b2 suggests complex numbers. Let z = a + bi. Can you write an equation in terms of z that
is equivalent to the given system?

144. Experiment. Find the first few xi. Then find the first term of the series, the sum of the first two
terms, the sum of the first three terms. Notice a pattern? Can you take a guess as to what the sum
of the first k terms is in terms of one of the x∣7

145. Rearrange the inequality so that all terms on both sides are positive. Notice that you have a sum
on the greater side. Looks like a job for AM-GM. How many terms does the 2ιz" suggest you need
in your AM?

146. Use the described property of the function to write an equation in terms of a, b, and z. Remember,
this equation must hold for all z.

147. Express r + s in terms of t, and s + t in terms of r, and t + r in terms of s.

148. Expand (x - y)(x - 2)'‘' (x - ∏)∙

149. Let r be a root of ∕(x). Produce an equation with rk.

150. There are infinitely many solutions. Let a - n - 2k; your original equation should have an (π - 2k')2
in it somewhere.

151. In many of the graphs of polynomials that appear in the text, there are portions of the graph that
are nearly vertical lines. Where are these nearly vertical lines located on these graphs, and what
causes them to be nearly vertical?

152. Suppose there are n numbers in the list, and m is their sum. Find n and m.

153. In the text, we found a polynomial that has V2 + V3 as a root. How might this polynomial be
helpful?

154. Do you see an expression that appears in more than one equation? We can substitute for whole
expressions, not just variables!

155. If you have three variables, you are trying to evaluate a sum of the form ab + bc + ca. What algebraic
expression will produce a sum like that?

156. Find an integer that is close to the given fraction.

157. Let a = 3x - 5, b = 2x - 4, and c = 5x - 9. Then cube.

158. We'd like to get rid of those fractional part symbols. Find expressions for {fl-1} and {a2} that don't
have any special functions in them.

159. Experiment. Forget about «9 and «io- Pick different values of «2 and find the next few terms. Do you
see a pattern?

681
HINTS TO SELECTED PROBLEMS

160. Can you combine the equations in a way that allows you to factor the result?

161. We'd like to find an expression for the number of positive integers such that ∕(k) = n. If the closest
integer to Vk is n, then what inequality chain can we write with tfk in the middle?

162. Note that dividing a number by 6 is the same as dividing it by 2, then dividing the result by 3. Can
you factor the quadratic?

163. Let z = a + bi, and use the definition ∣z∣ = Vα2 + b2. For another approach, what is the graph of the
equation ∣z - 1∣ = ∣z + 3∣ on the complex plane?
164. Write (a↑ + βι)(aq + β2)(tfι + β3) as a polynomial in a↑. You should be able to evaluate the coefficients
of this polynomial.

165. Two useful tactics with quadratics are factoring and completing the square. Try them both!

166. Start by finding an inequality that will get a + b,a + c, and b + c out of the denominators on one side
of the inequality.
167. How can the first n terms of the sum be combined to produce the terms with n2 as the numerator?
To produce the terms with n3 as the numerator? With nn as the numerator?

168. If you were to write an equation for each λ⅛, you'd have a long list of very similar-looking equations.
What strategy can you use with a list of very similar-looking equations?

169. Suppose b↑ = log64 yι and b2 = log64 y2. Write yιy2 i∏ terms of b1 and b2.

170. Suppose one root is a + bi. What is another root? What is the sum of the roots?

171. What's a tactic that will both get rid of absolute value symbols, and introduce terms like ab into the
three inequalities that have absolute value symbols?

172. The cubic has no x2 term. What does that tell us about m, n, and p? Does this help us find a root of
the quadratic?
173. Back to basics. Get rid of a variable. When does the remaining equation have real solutions?

174. Factor x3 + y3. Can you use the given equations to write the factored expression in terms of z?

175. The right side is an integer power of a binomial. We have a tool for that.

176. After applying Cauchy, write the inequality you must prove to finish, then work backwards. Don't
forget that ab + be + ca = 1/3. And don't forget to factor, factor, factor.

177. It would be a lot easier if we could combine both instances of each variable in a single equation.
What happens if we do so for all the variables?

178. Look at the ratio of consecutive terms. How do we know we've reached the maximum A(?

179. Write the expression with a common denominator. Symmetric expressions suggest Vieta's Formu­
las.
180. When you arrange things twice, you rearrange them.

682
 HINTS TO SELECTED PROBLEMS

181. Suppose there were originally n numbers. In terms of n, what is the largest possible average of the
numbers that remain after erasing one? What is the smallest possible such average?
182. Can you factor xy - 3x + 7y - 21?

183. Write the expression in terms of logfl b and logfc a.

184. Back to basics. It looks like an arithmetico-geometric series; what general tactic did we use to tackle
those?

185. Can you first find a polynomial with roots 1, ∣, and |? We're interested in sums of pairs of these
roots; what is the sum of all three?
186. Squaring what expression will give us x2, -⅛, and 2-?
y y
187. The expression x2 + y2 suggests complex numbers. What can we do to the complex number x + yz
to produce x2 - y2 and 2xy?

188. Use x3 = 1 and x ≠ 1 to express x2 as a linear function of x.

189. Look for squares of binomials. (And of other expressions!)

190. Can you find a polynomial in a, b, and c with integer coefficients such that the polynomial has
a + b y/2 + c ^∖[⅛ as a factor?

191. Write the equation as a quadratic in a (or b). What must be true about the discriminant if a and b
are real?

192. What kind of series is the polynomial in x?

193. Write the three quantities in terms of [xj and {x}. Make an equation that includes both these
quantities. Can you solve for one in terms of the other? (And don't forget what they mean!)

194. Start with a - b = b - c, rearrange, and factor relentlessly. You should be able to express the equation
as the product of two factors (in terms of x, y, and z) that must equal 0.

195. Compare each term in the expansion of the right side to n.

196. Vieta asks, "What is a + b in terms of c?"

197. First, try to figure out which polynomial of the type described in the problem has the largest root.
Then, try to find which of the intervals the root is in.

198. Let Xi be the number picked by the person who announced the average i. What is xg in terms of X6?
Then, what is Xιo in terms of Xf>2

199. Let r equal the ratios that follow from the first hint. Find a useful way to combine the three equations
you then have for r.

200. Experiment. Find ∕(94) in terms of ∕(92). In terms of ∕(91). In terms of ∕(90).

201. What can you do to one of the four equations to produce zzx5 + fey5? (You'll get some extra terms,
but you might be able to evaluate some of those, too.)

683
hints to selected problems

202. If we write the equation in standard form, and use the vertex, what is c in terms of a?

203. We wish to show that we can express cxnym + cxmyn, where m ≠ n, in terms of p = xy and s = x + y.
Factoring leaves what?

204. Suppose k of the terms on the right equal a. Find both sides of the equality in terms of a, n, and k.

205. Can we write a single equation in the variable x such that letting x equal 1, then 2, then 3, then 4,
produces the four equations?
206. Start simple. Let ∕(x) = x2 + 1. How about ∕(x) = x4 + 1? f(x) = x4 + 6x2 + 1?

207. The sum on the left consists of 73 integers. How many distinct integers are in this sum?

208. Can you find a root of the quadratic?

209. Before writing any equations, think about the problem. Let P be one of the endpoints of the common
chord. What do you know about the distance between P and the x-axis?

210. We have a In on the right, and, if you look closely, there are 2rz terms on the left.

211. What is x - |_xj? What does the first equation look like?

212. Move everything to one side and factor, factor, factor. Be on the lookout for squares!

213. Suppose ki is one of the imaginary roots. Let x — kiin the equation. What must both the resulting
real and imaginary parts equal?
214. The numbers being squared on the right side are 3 more than multiples of 12. Express each in the
form 12« + 3. Do you recognize the values of a that give you the numbers?

215. Move all the variables to one side. Does that side suggest any of the inequalities we have studied?

216. a+b+c+d=8- e.
217. Let g(x) = ∕(x) - 5. Write g(x) in factored form. What is g(k)?

218. When going from row k to row k + 1, how many times is each number in row k included in a sum?

219. The arguments of the logarithms are the same. It sure would be nice if the bases were the same
instead. How can we make that happen?

220. Let m be the middle square.

221. Let« = J.

222. Let a = logx y. Find all possible values of a.

223. The first hint should inspire you to seek a polynomial whose roots are the squares of rl. So, let
y = x2, and try to build the necessary polynomial in y from p(x) = 0.

224. Based on the previous hint, what are the possible values of Q(l)?

225. How are the terms in the numerator related to those in the denominator?

684
 HINTS TO SELECTED PROBLEMS

226. Ugh. Fractions with nasty denominators. Cauchy should help. What clever substitution could you
use as a first step to move the cubes from the denominator to the numerator right away?

227. If the term cx1lym, where m ≠ n, is in a symmetric two-variable polynomial, what other term must
appear?

228. This looks a lot like the previous problem. In terms of f, what can you multiply ∕(χ) by to get the
product in the previous problem?

229. Let b = ax. Solve the equation for b in terms of y.

230. Expand, rearrange, factor. Summation notation will help a lot here.
231. What is deg Q?

232. Use your equation from the first hint to express a3 as a linear expression in a. Then, use this equation
to express α6 as a linear expression in a. And so on.

233. Let the number of people in the m groups be x1, X2,..., xnt, and let the side lengths of the cakes be
Si, S2,..., sm. Write inequalities and equations in terms of the x, and s1∙, as well as an expression for
the total length of ribbon needed.

234. Take the base 10 logarithm of both sides to get the logarithm out of the exponent. Then, let
1/ = l°gιo x∙ ^-an You write your equation in terms of y?

235. Can you write the product in a more convenient form? What expressions in terms of a and b are
easy to evaluate?

236. Let a = (V2 + l)λ and b = (V2 — l)x. Find two equations in terms of a and b.

237. Let a be the middle term of the arithmetic sequence.

238. The expression for y1∙ should look familiar.

239. Forget the nasty Zq. Experiment and learn more about F(z). Notice that Z2 = F(F(z0)).

240. This problem looks like a nail. Break out your hammer and bash away. Then, factor the pieces.

241. There is an n; induct! (Be careful—we need more than one base case.)

242. What is x — [xj?

243. Vieta says, "Let the roots be a, ar, ar2, ar3, and ar4."

244. The desired inequality has squares. Square the given equation. Don't forget that we are given the
order of a, b, and c.

245. Make a substitution that turns the equation into something you know how to handle.

246. Get rid of the fractions and move everything to one side. Factor, factor, factor.

247. Simplify the problem. Tackle a similar problem for 2 variables.

685
HINTS TO SELECTED PROBLEMS

248. Think first. Are there values of y (or a whole range of values) that obviously cannot satisfy the
inequality? Use the answer to this question to simplify the absolute value expression before you
even start with the algebra.

249. Express every logarithm in terms of base a logarithms.

250. Can you find a point on the directrix? What is the slope of the axis? Of the directrix?

251. Think both about what happens when x ≥ 0 and when x < 0, as well as about what happens when
∕(x) ≥ 0 and when f(x) < 0.

252. Notice that (3x - 5) + (2x - 4) = 5x - 9.

253. What does P(0) = 0 tell us about P(x)? And what does this tell us about all possible rational solutions
to the equation?

254. 2 + 2 = 4. Try applying AM-GM to a and b, then to c and d.

255. That first equation is quite ugly. A substitution for simplification might help. What substitution
would give you a two-variable equation that isn't so ugly?

256. Can you create a series with radicals that looks a lot like the given series, but telescopes?

257. We want the maximum of a product of three x's and a 4 - x, so AM-GM is a natural candidate. But,
we'll want our AM to have a constant sum, so that we can say our GM is at most some constant.
Find four nonnegative expressions that have a constant sum and a GM that involves x3(4 — x).

258. Be careful! After simplifying the logarithm and finding what value of x minimizes the argument,
check if ∕(x) is defined for that value of x.

259. Play a bit. Find a3 in terms of αi and a0.

260. Start with your intuition. What number is the answer probably near?

261 "Smallest real number" means we probably have to build an inequality of some sort. How can we
get an inequality from the equation?

262. Write out some terms.

263. Find a factor of p(x).

264. Let x = a⅛. We're given information about a2 + b2. What should we do with x?

265. Let the roots bea,a + d,a + 2d, and a + 3d. Use Vieta's Formulas to relate a and d. Vieta was helpful
once; try again!
266. We know how to handle x3 + y3 and x3 - y3.

267. Find an equation in terms of a, b, and c that must hold if x - 1 divides ∕(x) - 1. What equation must
be true if x - 1 divides the quotient of ∕(x) - 1 and x - 1?

268. What is the πth term in terms of n? Simplify it as much as possible.

686
 HINTS TO SELECTED PROBLEMS

269. The equation is a quadratic in y. If y is real, what must be true about the discriminant?

270. For what values of k does the graph of y = kx intersect the ellipse?

271. In your experimenting, you should have done a lot of dividing by 2. Also, 1023 is very close to a
power of 2. All those 2s might suggest thinking about the binary representation of n.

272. Compare the series to a very similar series that you can evaluate.

273. Let p(x) = pnχn + pn.1χn~1 + ∙∙∙ + p-lχ + p0 and q(x) = qnxn + q∏-ιxn~l + ■■■ + q↑x + qo- What is the
average of the roots of p(q(x)) in terms of the coefficients of these polynomials?

274. Can you separate x and 1 — x, as in the separation section?

275. After writing A + B + C = lin terms of a, b, and c, you should get a cubic polynomial equal to
0. To find factors of this polynomial, experiment by choosing values for b and c and factoring the
resulting polynomial in a. After trying a few, you should have a guess as to what the factors are.
Find the product of these factors to confirm your guess.

276. Regroup the terms in a way that lets you factor piece by piece.
277. What is f(t)2

278. Don't forget that x can be negative. Handle x < 0 and x ≥ 0 separately.
279. 1 + 1 = 2±1.

280. Imagine you wrote out the given recursion for every value of n from 2 to 1986 (remember, ∕(1986) =
∕(1)). What could you do with these equations to get an equation involving the desired sum?

281. We have a sum on each side. That's almost Cauchy! We need another sum on one side and a square
of a sum on the other. What can we multiply the inequality by to achieve this?

282. How can you get ab + be + ca terms from the equation a + b + c = 3?

283. Write the expression as a difference of squares.

284. If a polynomial is divisible by (x - c)g(x), is it divisible by g(x)?

285. What graph is produced when the graph of y = ∕(x) is shifted 6 to the right, and then the result
scaled horizontally by a factor of 1/2?

286. Let ∕(x) = z⅛x5 + α<x4 + α3X3 + α2X2 + a-∖X + ⅜. What does the fact that ∕(x) is divisible by x3 tell us
about some of the coefficients of /(x)?

287. It's obviously divisible by x - 1. What is the quotient?

288. Factor those quadratics.

289. After factoring, you should have a bunch of powers of x. Don't forget that 0 ≤ x ≤ 1.

290. Isolate the fractional part of x. What does the result tell us about |xj?

291. How can we get an equation with x + y + z in it from the three given equations?

687
HINTS TO SELECTED PROBLEMS

292. The second equation is quadratic in xy.

293. What do Vieta's Formulas tell us about Xi + Xz and x1x27 Use these to find expressions for x3 + x?
and x3x3.

294. Write the left sides as the sum or difference of two fractions. Does the result resemble a system of
two-variable linear equations?

295. Let a be the first term and d be the common difference of the arithmetic sequence. Once you have
aιι equation, solve for a. Remember that the sequence consists of positive integers.

296. How can you turn the y in x + y + z into two y's?

297. We seek the product (r2 - 2)(r2 - 2)(r2 - 2)(r2 - 2)(r∣ - 2). What does this look like if we replace the
2 with a variable?

298. Rewrite the equation in a form that allows you to use what you know about the monotonic nature
of exponential functions to solve the problem.
299. Use Vieta's Formulas and the fact that γ is a root of x2 + Px + Q to express (α - γ)(β - y) in the form
cγ + d, where c and d are constants.

300. Can you factor the denominator? (Yes.)

301. Let u = a + bi. Find v and u in terms of a and b.

302. It sure would be nice if the logarithms had the same base instead of the same argument.

303. What does the given equation tell us about how the sum and the product of the would-be roots are
related?

304. For what values of X3 is X4 = X3?

305. For the second ratio, rearrange the desired ratio equation to have the x and y terms all in one ratio,
and a and b in the other. Let k equal each of the given ratios.

306. We know how to handle rational function inequalities. Turn this inequality into one with a substi­
tution.
307. Get rid of the floor function, ha terms of x, what must Lx J be greater than?

308. We have ∕(∕(x)) = x for all x. Choose a convenient x.

309. After some algebra, you should have terms of the form x∣9x93 + xj9x93 on the greater side and terms
of the form x20x92 + x20x92 on the lesser side. Focus on a single i, j pair. Rearrange, factor.

310. Which terms in the expansion affect the first three digits after the decimal point?

311. Sums and nasty denominators. Try Cauchy. You'll get another nasty denominator, but one nasty
denominator on the greater side might be better than three.

312. How can you get rid of that nasty denominator with those cube roots?

688
 hints to selected problems

313. Start with fewer numbers; use 3 or 4 rather than 21. Write out the products (don't compute them),
and look for a clever way to sum them.

314. Suppose we have a function such that h(x) = ∕ι(l - χ). About what value of x is such a function
symmetric? What other class of functions is symmetric about some specific value of x? How might
h be related to one of these functions?

315. The graph of xy = 1 is the result of rotating the graph of what equation 45 degrees counterclockwise?
What are the foci of the latter equation?

316. We took care of a difference of square roots by multiplying by a sum of square roots. What will we
have to multiply a difference of cube roots by in order to get an expression without cube roots?

317. For what values of x does x3 + x = 0?

318. Suppose {a, b) is on the graph of y = log10 x. Express both coordinates of the corresponding point
on G' in terms of a. Use these coordinates to express y in terms of x to produce an equation whose
graph is G'.

319. Is it possible for ∕(0) to be even?

320. Try letting a = x∕(2y), b = y∕(2z), and c = z∕(2x).

321. Multiply out the left side. Can you minimize groups of terms that result? Start with

322. After using a logarithm identity and the Pythagorean Theorem to write the logarithmic equation
as an equation without logarithms in terms of just a and b, Simon would know how to finish the
problem. Don't forget that a <b.

323. Can you express ∕(2002) in terms of f(n), where n < 2002?

324. We can't speak of the roots of "polynomial P(x) - because this is not a polynomial. But xP(x) -1
is.
325. The expression (”) equals r∣("lrγ∣, where n∖ = n(n - l)(n - 2) ∙ ∙ ∙ (1). Let the binomial coefficients be
(A), (£)/ and (fc"1). Write some equations with the ratio information.

326. The sum consists of the coefficients of the even degree terms in the expansion of (1 + x)99. How can
we get rid of the odd degree terms?

327. What are the y-coordinates of the endpoints of the latus rectum?

328. What is likely to be the condition that will provide the desired minimum? Try using this to figure
out what tools you need to build an inequality that you can use to prove the expression cannot be
smaller than this minimum.

329. You have the product of two roots, so what can you immediately find by looking at the coefficients
of the polynomial?

330. Let z be a complex number (point in the complex plane), and let ∕(z) be its image. How can we
write the distance between these two points as an algebraic expression?

689
HINTS TO SELECTED PROBLEMS

331. That first equation resembles an equation whose graph is a circle. Experiment with values of x in
different ranges. For example, what happens if ∣xj = 0?

332. Can you find any squares near n2 - 10n + 29?

333. Confirm your pattern by letting «2 = Use ag + flιo to find n.

334. How is a real number related to its conjugate?

335. What can you sometimes do to get complicated expressions out of exponents?

336. Find g^1(x) in terms of a, b, c, and d.

337. Build a polynomial with roots a, b, c, and d. What is interesting about the coefficients of this
polynomial?

338. Factor!

339. Write the whole series in summation notation. (And don't use ■ ∙ ∙ to express each term.)

340. In the previous problem, we used AM-GM for 2 variables to prove it for 4. How do you think we 11
get to 8? To 16? To 2k2

341. Work backwards; square and rearrange. Does a familiar inequality appear?

342. Rearrange the inequality into various different forms. Try squaring some of the radicals away. Don't
give up! Keep manipulating it until you see some expressions you can work with.

343. Let z + 1 = a. Can you combine this with the final two given equations to produce xyz terms?
y
344. What inequality helps with finding the minimum value of a sum?

345. The a - b in the denominator is annoying. How can you rewrite the expression so that there's an
a - b in a numerator, that you can then cancel with the denominator when you apply AM-GM?

346. Find an equation whose graph is the storm at time f. (Your equation should have t in it.) Where is
the car at time t?
347. Find a perfect square or perfect cube near this number.

348. Those denominators are annoying.

349. Isolate ∣y∣. How can you most easily examine the behavior of the other side of the equation after
doing so?
350. Show that x2 + y2 can be expressed in terms of p and s. What about x3 + y3? x4 + y4? What, essentially,
are you doing as you go from one power to the next?

351. Tackle it in two parts. Handle the left end by letting m = x√X2∙

352. Explain why (fl3 — l)(o2 - 1) ≥ 0.

353. Simon would first substitute variables for logl0 x, logl0 y, and logl0 z.

690
 hints to selected problems

o54. Squared polynomials are often easier to work with. See if you can find a way to express the given
polynomial as a linear combination of squared polynomials.

355. How many x's, y's, and z,s do you need to produce a GM of xy2z3?

356. Find three roots of P(Q(x)).

357. How can you tell whether or not a quadratic equation has any real solutions?

691
HINTS to selected problems

692
 INDEX

Index

[x],529 Arithmetic Mean-Geometric Mean Inequality, 384-

A, 572 391
∖, 29 arithmetic sequence, 286-292
∩, 29 arithmetic series, 292-298
U, 28 arithmetico-geometric series, 567-571
∞, 3 ARML, vii
Γx], 529-539 Art of Problem Solving, v, 672, 702
Lx], 529-539 asymptote, 147,477
1,201 horizontal, 477
∏,313-331 oblique, 484
∑,313-331 slant, 484
{a1∙}, sequence notation, 313 vertical, 477
{x}, fractional part, 529-539 average, 385
axis of symmetry
Abel Prize, 507 of a parabola, 107,109
Abel Summation, 368
Abel, Niels, 229, 507 base, 413,417
absolute value, 517-529 base case, 350
Abu Ja'far Muhammad ibn Musa al-Khwarizmi, 37 binomial, 162
algebraic number, 238 binomial coefficient, 359
AM, 384 Binomial Theorem, 358-365
AM-GM Inequality, 384-391 brute force, 338-343
AMC, vii Buniakowsky, 396
American Mathematics Competitions, see AMC
American Regions Math League, see ARML camel, magic, 382,443
Argand plane, 67 Cardano, Girolamo, 78, 229
Argand, Jean-Robert, 67 Catalan numbers, 337
argument, 417 Cauchy Functional Equation, 627
arithmetic mean, 384, 581-593 Cauchy, Augustin Louis, 389
Cauchy-Schwarz Inequality, 340,392-399

693
INDEX

proof of, 394 degenerate hyperbola, 152

ceiling function, 529-539 degree, 162,189, 261
cellular automaton, xvii delta, 572
center dependent, 17,19
of a circle, 126 Descartes Rule of Signs, 260
of a hyperbola, 142 Descartes, Rene, 54
of an ellipse, 129 determinant, 22
Chebyshev's Inequality, 412 difference of cubes, 271
chord, 129 difference of squares, 82,271
circle, 126-129 Diophantine equation, 263
center, 126 directrix, 107
standard form, 127 discontinuous, 511
Clay Mathematics Institute, 507 discriminant, 96-99,102
closed form, 295, 544 diverge, 308
codomain, 615 divergent series, 309
coefficient, 162,189 division
common divisor, polynomial, 655 of polynomials, 166-189
common factor, polynomial, 658 synthetic, 177-185
common logarithm, 445 divisor, 169
completing the square, 88-96 domain, 24
complex number, 54-77 double root, 96, 200
complex plane, 61-65, 70-74 double sum, 325-331
graphing, 61-65, 70-74 dummy variable, 24,313
history, 78
imaginary part, 54
e, 444-450
eccentricity, 134,149
magnitude, 63
elimination, 9-13
real part, 54
ellipse, 129-142
complex plane, 61-65, 70-74
area of, 142
composition (of functions), 39-43
center, 129
conic section, 106-161
chord, 129
conjugate, 58
degenerate, 141
conjugate pair, 58
eccentricity, 134
consistent system, 16
focus, 129
constant term, 79
major axis, 129
continuous, 202, 511
minor axis, 129
contradiction, 231
standard form, 136
converge, 309
equality condition, 379
convergent series, 309
equation
Cook, Matthew, xviii
linear, 1-23
Cramer's Rule, 23
quadratic, 79-105
cubic, 162
Euler, Leonhard, 54, 78, 445
general solution of, 229
even function, 492-497
cyclic function, 621-627
exponent, 413
order, 623
exponent law, 424
degenerate ellipse, 141 exponential decay, 415, 444-450

694
 INDEX

exponential function, 413 Fundamental Theorem of Algebra. 196 215-225

exponential growth, 414
extraneous, 4, 457 Gauss, Carl, 268, 293
generation, xvii
factor, 193, 225 geometric mean, 384,581-593
Factor Theorem, 193-198, 277-282 geometric progression, 298
factoring geometric sequence, 298-301
xn + y'l, 271 geometric series, 302-312
xw - y”, 270 GM, 384
grouping, 261-268 greatest integer function, 529
quadratics, 79-84 grouping, 261-268
factorization
over the complex numbers, 236 Holder Inequality, 399
over the integers, 236 half-life, 445
over the rationals, 236 Hamilton, Lord William Rowan, 78
over the reals, 236 harmonic mean, 581-593
Ferrari, Ludovico, 229 harmonic progression, 335
Fibonacci sequence, 544 harmonic sequence, 335
Fields Medal, 507 harmonic series, 335
finite difference, 571-576 Harvard-MIT Math Tournament, sec HMMT
floor function, 529-539 Hermite's Identity, 543
focus Hippasus, 230
of a hyperbola, 142 HM, 581-593
of a parabola, 107 HMMT, vii
of an ellipse, 129 hole, 480, 512
fractional part, 529-539 homogeneous equation, 641
function, 24-53 homogenization, 641
composition, 39-43 horizontal asymptote, 477
cyclic, 621-627 horizontal line test, 47
decreasing, 497-503 hyperbola, 142-156
even,492-497 center, 142
graphing, 32-39 eccentricity, 149
increasing, 497-503 focus, 142
inverse, 43-49 rectangular, 153
invertible, 46 standard form, 145,150
monotonic, 497-503 vertex, 142
odd, 492^97 i, 54
piecewise defined, 508-516 identity, 338-370
rational, 476-491
Identity Theorem, 219-225
real, 25
if and only if, 67
real-valued, 25 Im, 61
transformation, 36-39
imaginary axis, 61
functional equation, 611-630
imaginary number, 54
cyclic function, 621-627
imaginary part, 54
separation, 619-621
inconsistent system, 16
substitution, 615-619 independent, 17, 19

695
INDEX

indeterminate series, 308 natural, 444—450

index, 313
indirect proof, 232, 256 magic camel, 382,443
induction, 348-358 magnitude, 63
inductive assumption, 351 major axis, 129
inductive hypothesis, 351 Mandelbrot Competition, vii
inductive step, 351 matrix, 22
inequality, 371-412 maximization, 400-406
AM-GM, 384-391 quadratic, 121-126
Arithmetic Mean-Geometric Mean, 384-391 mean
Cauchy-Schwarz, 392-399 arithmetic, 384
Holder Inequality, 399 geometric, 384
Mean Inequality Chain, 581-593 harmonic, 581
nonstrict, 3 quadratic, 581
polynomial, 205, 211 Mean Inequality Chain, 581-593
Method of Undetermined Coefficients, 644-648
quadratic, 99-102
Rearrangement Inequality, 593-600 Millennium Problems, 507
strict, 3 minimization, 400-406
Trivial Inequality, 378-384 quadratic, 121-126
Minkowski's Inequality, 412
inequality chain, 3
infinite geometric series, 302,305 minor axis, 129
monic polynomial, 165
intersection, 29
monomial, 162
interval, 3
monotonic, 497-503
inverse function, 43-49
monotonically increasing, 413
irrational number, 206, 231
multiple root, 200
isolation, 1-6
multiplicity, 196, 200
Lagrange's Identity, 340 multivariable polynomial, 261-285
latus rectum, 118 natural logarithm, 444-450
leading term, 165 Nesbitt's Inequality, 610
least integer function, 529 nested sum, 325-331
Legendre, Adrien-Marie, 226
New York State Math League, vii
Lie, Sophus, 507 Nobel Prize, 507
limit, 307 nonreal, 54
linear, 162 nonstrict inequality, 3
linear algebra, 17 nonzero polynomial, 162
linear combination, 17 number
linear term, 79 algebraic, 238
links, vi complex, 54
In, 445 irrational, 231
locus, 106 nonreal, 54
logarithm, 413-455
rational, 230
argument, 417
transcendental, 238
base, 417
NYSML, vii
common, 445
identity, 424-440 oblique asymptote, 484

696
 INDEX

odd function, 492—497 proof by induction, 348-358

optimization, 400—406 Pythagoras, 230
order (of a cyclic function), 623
ordered pair, 7 Q, 230
origin, 61 QM, 581-593
QM-AM, 584
parabola, 107-126 QM-AM-GM-HM Inequality, 585
axis of symmetry, 107 quadratic, 79-105,162
directrix, 107 completing the square, 88-96
focus, 107 discriminant, 96-99
latus rectum, 118 factoring, 79-84
standard form, 113 inequality, 99-102
vertex, 107 maximization, 121-126
partial fraction decomposition, 556 minimization, 121-126
partial sum, 307 sum and product of roots, 84-88
Pascal's Identity, 360 quadratic formula, 93
period, 548 quadratic mean, 581-593
periodic sequence, 547 Quadratic Mean-Arithmetic Mean Inequality, 584
permutation, 652 quadratic term, 79
∏, 313-331 quartic, 162
piecewise defined function, 508-516 general solution of, 229
polynomial, 162-191, 631-665 quotient, 169
common divisor, 655
common factors, 658 1R,3
division, 166-189 radical conjugate, 233, 470-473
finite difference, 572 radioactive decay, 445
graphing, 215-219 radius, 126
Identity Theorem, 219-225 range, 25
inequality, 205, 211 ratio, 343-348
integer root, 198-206 rational function, 476-491
monic, 165 rational number, 206,230
multivariable, 261-285 Rational Root Theorem, 206-212
product of roots, 244-253 rationalize the denominator, 470
rational root, 206-212 Re, 61
Remainder Theorem, 185-189 real axis, 61
roots and coefficients, 244-253 real function, 25
roots of, 193-260 real number, 54
sum of roots, 244-253 real part, 54
synthetic division, 177-185 Rearrangement Inequality, 593-600, 607
Vieta, 244-253 rectangular hyperbola, 153
polynomial function, 162 recurrence relation, 544
Power Mean Inequality, 591 recursion, 544-551
problem solving, iii recursive formula, 544
product redundant, 17,19
nested, 325-331 remainder, 169
proof by contradiction, 231 Remainder Theorem, 185-189

697
INDEX

resources, v summation
RMS, 581 double, 325-331
root, 79,193 nested, 325-331
root mean square, 581 ∑,313-331
Rubinstein-Salzedo, Simon, 264 summation notation, 313
Ruffini's rule, 177 symmetric, 87, 280
Ruffini, Paolo, 177, 229 symmetric expression, 250
Rule 110, xvii symmetric sum, 246, 658-665
symmetry, 631-637
Schur's Inequality, 412 synthetic division, 177-185
Schwarz, 392 system of equations, 6-19
sequence, 286 consistent, 16
arithmetic, 286-292 dependent, 17
geometric, 298-301 inconsistent, 16
harmonic, 335 independent, 17
notation, 313 linear, 10
periodic, 547 nonlinear, 12
series, 286
arithmetic, 292-298 tangent, 121
arithmetico-geometric, 567-571 Tartaglia, Nicolo Fontana, 229
convergent, 309 telescoping series, 552-560
divergent, 309 term, 162,189
geometric, 302-312 transcendental number, 238
harmonic, 335 translation, 112
polynomial, 561-566 trinomial, 162
telescoping, 552-560 Trivial Inequality, 378-384
Serre, Jean-Pierre, 507 Turing complete, xviii
shift the index
undetermined coefficient, 644-648
recursion, 550
union, 29
summation, 321
upper bound (of roots), 214
∑,313-331
USA Mathematical Talent Search, see USAMTS
sigma, 313
USAMTS, vii
Simon's Favorite Factoring Trick, 263
slant asymptote, 484 variable
Sophie Germain's Identity, 268 dummy, 24
square root, 548 vertex
squaring the circle, 238 of a hyperbola, 142
standard form of a parabola, 107
of a circle, 127 vertical asymptote, 477
of a hyperbola, 145,150 vertical line test, 36
of a parabola, 113,114 Vieta's Formulas, 84-88, 244-253
of an ellipse, 136 von Lindemann, Ferdinand, 238
strict inequality, 3
substitution, 6-9, 638-644 Weierstrass, Karl, 526
functional equation, 615-619 without loss of generality, 598
sum of cubes, 272 WLOG, 598

698
 INDEX

Wolfram, Stephen, xviii

x-intercept, 32

y-intercept, 32

Z, 230
z, 58
zero (of a polynomial), 193
zero (of a quadratic), 79
zero polynomial, 162

699
 www.artofproblemsolving.com
The Art of Problem Solving (AoPS) is:

• Books
For over 14 years, the classic Art of Problem Solving books have been used by students as a
resource for the American Mathematics Competitions and other national and local math events.

Every school should have this in their math library.

- Paul Zeitz, past coach of the U.S. International Mathematical Olympiad team

Our Introduction series of textbooks—covering Algebra, Counting & Probability, Geometry, and
Number Theory—constitutes a complete curriculum for outstanding math students in grades
6-10. Our Intermediate series of textbooks will form a complete curriculum for outstanding math
students in grades 8-12. Intermediate books covering Counting & Probability and Algebra are
available now.

The new book [Introduction to Counting & Probability] is great. I have started to use it in my
classes on a regular basis. I can see the improvement in my kids over just a short period.
-Jeff Boyd, coach of the 2007 MATHCOUNTS National Championship team from Texas

• Classes
The Art of Problem Solving offers online classes on topics such as number theory, counting,
geometry, algebra, and more at beginning, intermediate, and Olympiad levels.

All the children were very engaged. It's the best use of technology I have ever seen.
- Mary Fay-Zenk, coach of National Champion California MATHCOUNTS teams

• Forum
As of March 2008, the Art of Problem Solving Forum has over 41,000 members who have posted
over 1,000,000 messages on our discussion board. Members can also participate in any of our free
"Math Jams."
Γd just like to thank the coordinators of this site for taking the time to set it up... I think this is
a great site, and I bet just about anyone else here would say the same...
- AoPS Community Member

• Resources
We have links to summer programs, book resources, problem sources, national and local compe­
titions, scholarship listings, a math wiki, and a UTgX tutorial.

I'd like to commend you on your wonderfid site. It's informative, welcoming, and supportive of
the math community. I wish it had been around when I was growing up.
- AoPS Community Member

• ... and more!

Membership is FREE! Come join the Art of Problem Solving community today!