Full download solution manual or testbank at testbankdeal.

com

Linear Algebra with Applications 5th Edition

Bretscher Solutions Manual

https://testbankdeal.com/product/linear-algebra-with-
applications-5th-edition-bretscher-solutions-manual/

OR CLICK HERE

DOWNLOAD NOW

Download more solution manual or test bank from https://testbankdeal.com

Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Linear Algebra with Applications 2nd Edition Bretscher

Solutions Manual

https://testbankdeal.com/product/linear-algebra-with-applications-2nd-
edition-bretscher-solutions-manual/

testbankdeal.com

Linear Algebra With Applications 8th Edition Leon

Solutions Manual

https://testbankdeal.com/product/linear-algebra-with-applications-8th-
edition-leon-solutions-manual/

testbankdeal.com

Linear Algebra with Applications 2nd Edition Holt

Solutions Manual

https://testbankdeal.com/product/linear-algebra-with-applications-2nd-
edition-holt-solutions-manual/

testbankdeal.com

Financial Accounting 15th Edition Williams Test Bank

https://testbankdeal.com/product/financial-accounting-15th-edition-
williams-test-bank/

testbankdeal.com
Strategic Management 3rd Edition Rothaermel Test Bank

https://testbankdeal.com/product/strategic-management-3rd-edition-
rothaermel-test-bank/

testbankdeal.com

Psychology and Your Life with POWER Learning 3rd Edition

Feldman Solutions Manual

https://testbankdeal.com/product/psychology-and-your-life-with-power-
learning-3rd-edition-feldman-solutions-manual/

testbankdeal.com

Business Analysis and Valuation Asia Pacific 2nd Edition

Palepu Solutions Manual

https://testbankdeal.com/product/business-analysis-and-valuation-asia-
pacific-2nd-edition-palepu-solutions-manual/

testbankdeal.com

Principles of Macroeconomics 6th Edition Frank Solutions

Manual

https://testbankdeal.com/product/principles-of-macroeconomics-6th-
edition-frank-solutions-manual/

testbankdeal.com

Managerial Accounting 4th Edition Braun Solutions Manual

https://testbankdeal.com/product/managerial-accounting-4th-edition-
braun-solutions-manual/

testbankdeal.com
Management Asia-Pacific 6th Edition Samson Test Bank

https://testbankdeal.com/product/management-asia-pacific-6th-edition-
samson-test-bank/

testbankdeal.com
 Section 5.1

Chapter 5
Section 5.1
√ √ √
5.1.1 k~v k = 72 + 112 = 49 + 121 = 170 ≈ 13.04
√ √ √
5.1.2 k~v k = 22 + 32 + 42 = 4 + 9 + 16 = 29 ≈ 5.39
√ √ √
5.1.3 k~v k = 22 + 32 + 42 + 52 = 4 + 9 + 16 + 25 = 54 ≈ 7.35

·~ √7+11 √18
5.1.4 θ = arccos k~u~ukk~
v
v k = arccos 2 170 = arccos 340 ≈ 0.219 (radians)
√

·~ 2+6+12
5.1.5 θ = arccos k~u~ukk~
v
v k = arccos 14 29 ≈ 0.122 (radians)
√ √

·~ 2−3+8−10
5.1.6 θ = arccos k~u~ukk~
v
v k = arccos
√ √
10 54
≈ 1.700 (radians)

5.1.7 Use the fact that ~u · ~v = k~ukk~v k cos θ, so that the angle is acute if ~u · ~v > 0, and obtuse if ~u · ~v < 0. Since
~u · ~v = 10 − 12 = −2, the angle is obtuse.

5.1.8 Since ~u · ~v = 4 − 24 + 20 = 0, the two vectors enclose a right angle.

5.1.9 Since ~u · ~v = 3 − 4 + 5 − 3 = 1, the angle is acute (see Exercise 7).

5.1.10 ~u · ~v = 2 + 3k + 4 = 6 + 3k. The two vectors enclose a right angle if ~u · ~v = 6 + 3k = 0, that is, if k = −2.

·~ √1
5.1.11 a θn = arccos k~u~ukk~
v
v k = arccos n

θ2 = arccos √12 = π
4 (= 45◦ )

θ3 = arccos √13 ≈ 0.955 (radians)

θ4 = arccos 12 = π
3 (= 60◦ )

b Since y = arccos(x) is a continuous function,

 
lim θn = arccos lim √1n = arccos(0) = π2 (= 90◦ )
n→∞ n→∞

~ 2 = (~v + w)
5.1.12 k~v + wk ~ · (~v + w)
~ (by hint)

= k~v k2 + kwk
~ 2 + 2(~v · w)
~ (by definition of length)

≤ k~v k2 + kwk
~ 2 + 2k~v kkwk
~ (by Cauchy-Schwarz)

~ 2 , so that
= (k~v k + kwk)

~ 2 ≤ (k~v k + kwk)
k~v + wk ~ 2

Taking square roots of both sides, we find that k~v + wk

~ ≤ k~v k + kwk,
~ as claimed.

227
Copyright c 2013 Pearson Education, Inc.
Chapter 5

5.1.13 Figure 5.1 shows that kF~2 + F~3 k = 2 cos kF~2 k = 20 cos θ2 .
θ



2

It is required that kF~2 + F~3 k = 16, so that 20 cos θ2 = 16, or θ = 2 arccos(0.8) ≈ 74◦ .

Figure 5.1: for Problem 5.1.13.

5.1.14 The horizontal components of F~1 and F~2 are −kF~1 k sin β and kF~2 k sin α, respectively (the horizontal compo-
nent of F~3 is zero).

Since the system is at rest, the horizontal components must add up to 0, so that −kF~1 k sin β + kF~2 k sin α = 0 or
~
kF~1 k sin β = kF~2 k sin α or kF1 k = sin α .
~2 k
kF sin β

EA EA tan α sin α cos β ~1 k cos β

kF
To find EB
, note that EA = ED tan α and EB = ED tan β so that EB
= tan β = sin β · cos α = ~2 k cos α .
kF
Since
EA ~1 k
kF
α and β are two distinct acute angles, it follows that EB
6= ~2 k ,
kF
so that Leonardo was mistaken.

5.1.15 The subspace consists of all vectors ~x in R4 such that

x1 1
   
 x2   2 
~x · ~v =   ·   = x1 + 2x2 + 3x3 + 4x4 = 0.
x3 3
x4 4
−2r −3s −4t −2 −3 −4
       
 r  1  0  0
These are vectors of the form   = r  + s  + t .

s 0 1 0
t 0 0 1
The three vectors to the right form a basis.

5.1.16 You may be able to find the solutions by educated guessing. Here is the systematic approach: we first find
all vectors ~x that are orthogonal to ~v1 , ~v2 , and ~v3 , then we identify the unit vectors among them.

Finding the vectors ~x with ~x · ~v1 = ~x · ~v2 = ~x · ~v3 = 0 amounts to solving the system
 
x1 + x2 + x3 + x4 = 0
 x1 + x2 − x3 − x4 = 0 
x1 − x2 + x3 − x4 = 0

we can omit all the coefficients 12 .

x1 t
   
 x2   −t 
The solutions are of the form ~x =   =  .
x3 −t
x4 t

228
Copyright c 2013 Pearson Education, Inc.
 Section 5.1

1
Since k~xk = 2|t|, we have a unit vector if t = 2 or t = − 21 . Thus there are two possible choices for ~v4 :
1
− 21
   
2
 − 12   1
2

  and  .

 −1   1
2 2
 
1
2 − 21

5.1.17 The orthogonal complement W ⊥ of W consists of the vectors ~x in R4 such that

x1 1 x1 5
       
 x2   2   x2   6 
  ·   = 0 and   ·   = 0.
x3 3 x3 7
x4 4 x4 8
 
x1 + 2x2 + 3x3 + 4x4 = 0
Finding these vectors amounts to solving the system .
5x1 + 6x2 + 7x3 + 8x4 = 0

The solutions are of the form

x1 s + 2t 1 2
       
 x2   −2s − 3t   −2   −3 
 =  = s  + t .
x3 s 1 0
x4 t 0 1

The two vectors to the right form a basis of W ⊥ .

1 1 1 1 4 1
5.1.18 a k~xk2 = 1 +


4 + 16 + 64 + ··· = 1− 41
= 3 use the formula for a geometric series, with a = 4 , so that
k~xk = √2 ≈ 1.155.
3

b If we let ~u = (1, 0, 0, . . .) and ~v = 1, 12 , 41 , · · · , then

√
·~ 1 3
θ = arccos k~u~ukk~
v
v k = arccos √2 = arccos 2 = π
6 (= 30◦ ).
3

 
c ~x = 1, √12 , √13 , · · · , √1n , · · · does the job, since the harmonic series 1 + 1
2 + 1
3 + · · · diverges (a fact discussed in
introductory calculus classes).
√
3
d If we let ~v = (1, 0, 0, . . .), ~x = 1, 21 , 41 , · · · and ~u = ~
x
1, 12 , 41 , · · · then



k~
xk = 2

3
1, 21 , 41 , · · · .


projL~v = (~u · ~v )~u = 4

5.1.19 See Figure 5.2.

5.1.20 On the line L spanned by ~x we want to find the vector m~x closest to ~y (that is, we want km~x − ~y k to be
minimal). We want m~x − ~y to be perpendicular to L (that is, to ~x), which means that ~x · (m~x − ~y ) = 0 or
·~
m(~x · ~x) − ~x · ~y = 0 or m = ~x~x·~
y 4182.9
x ≈ 198.532 ≈ 0.106.

x·~
~ y k~
yk
Recall that the correlation coefficient r is r = k~
xkk~yk , so that m = xk r.
k~ See Figure 5.3.

229
Copyright c 2013 Pearson Education, Inc.
Chapter 5

Figure 5.2: for Problem 5.1.19.

Figure 5.3: for Problem 5.1.20.

5.1.21 Call the three given vectors ~v1 , ~v2 , and ~v3 . Since ~v2 is required to be a unit vector, we must have b = g = 0.
Now ~v1 · ~v2 = d must be zero, so that d = 0.

Likewise, ~v2 · ~v3 = e must be zero, so that e = 0.

√
1 3
Since ~v3 must be a unit vector, we have k~v3 k2 = c2 + 4 = 1, so that c = ± 2 .
√
3
Since we are asked to find just one solution, let us pick c = 2 .
√
3
√
The condition ~v1 · ~v3 = 0 now implies that 2 a + 12 f = 0, or f = − 3a.

Finally, it is required that k~v1 k2 = a2 + f 2 = a2 + 3a2 = 4a2 = 1, so that a = ± 12 .

√
3
Let us pick a = 12 , so that f = − 2 .

230
Copyright c 2013 Pearson Education, Inc.
 Section 5.1

Summary:
  √ 
1   3
2 0 2
~v1 =  √0  , ~v2 =  1  , ~v3 =  0 
   
− 3 0 1
2 2

There are other solutions; some components will have different signs.

5.1.22 Let W = {~x in Rn : ~x · ~vi = 0 for all i = 1, . . . , m}. We are asked to show that V ⊥ = W , that is, any ~x in
V ⊥ is in W , and vice versa.

If ~x is in V ⊥ , then ~x · ~v = 0 for all ~v in V ; in particular, x · ~vi = 0 for all i (since the ~vi are in V ), so that ~x is in
W.

Conversely, consider a vector ~x in W . To show that ~x is in V ⊥ , we have to verify that ~x · ~v = 0 for all ~v in V .
Pick a particular ~v in V . Since the ~vi span V , we can write ~v = c1~v1 + · · · + cm~vm , for some scalars ci . Then
~x · ~v = c1 (~x · ~v1 ) + · · · + cm (~x · ~vm ) = 0, as claimed.

5.1.23 We will follow the hint. Let ~v be a vector in V . Then ~v · ~x = 0 for all ~x in V ⊥ . Since (V ⊥ )⊥ contains all
vectors ~y such that ~y · ~x = 0, ~v is in (V ⊥ )⊥ . So V is a subspace of (V ⊥ )⊥ .

Then, by Theorem 5.1.8c, dim (V ) + dim(V ⊥ ) = n and dim(V ⊥ ) + dim((V ⊥ )⊥ ) = n, so dim (V ) + dim(V ⊥ ) =
dim(V ⊥ ) + dim((V ⊥ )⊥ ) and dim (V ) = dim((V ⊥ )⊥ ). Since V is a subspace of (V ⊥ )⊥ , it follows that V = (V ⊥ )⊥ ,
by Exercise 3.3.61.

5.1.24 Write T (~x) = projV (~x) for simplicity.

To prove the linearity of T we will use the definition of a projection: T (~x) is in V , and ~x − T (~x) is in V ⊥ .

To show that T (~x + ~y ) = T (~x) + T (~y ), note that T (~x) + T (~y ) is in V (since V is a subspace), and ~x + ~y − (T (~x) +
T (~y )) = (~x − T (~x)) + (~y − T (~y )) is in V ⊥ (since V ⊥ is a subspace, by Theorem 5.1.8a).

To show that T (k~x) = kT (~x), note that kT (~x) is in V (since V is a subspace), and k~x − kT (~x) = k(~x − T (~x)) is
in V ⊥ (since V ⊥ is a subspace).

5.1.25 a kk~v k2 = (k~v ) · (k~v ) = k 2 (~v · ~v ) = k 2 k~v k2

√
Now take square roots of both sides; note that k 2 = |k|, the absolute value of k (think about the case when k
is negative). kk~v k = |k|k~v k, as claimed.

1 1
b k~uk = vk ~
k~ v = v k k~
k~ vk = 1, as claimed.

by part a

5.1.26 The two given vectors spanning the subspace are orthogonal, but they are not unit vectors: both have length
7. To obtain an orthonormal basis ~u1 , ~u2 of the subspace, we divide by 7:
   
2 3
~u1 = 17  3  , ~u2 = 17  −6  .
6 2

231
Copyright c 2013 Pearson Education, Inc.
Chapter 5

 
49
Now we can use Theorem 5.1.5, with ~x =  49 :
49
     
2 3 19
projV ~x = (~u1 · ~x)~u1 + (~u2 · ~x)~u2 = 11  3  −  −6  =  39 .
6 2 64

5.1.27 Since the two given vectors in the subspace are orthogonal, we have the orthonormal basis
2 −2
   
2  2
~u1 = 13   , ~u2 = 13  .
1 0
0 1

Now we can use Theorem 5.1.5, with ~x = 9~e1 : projV ~x = (~u1 · ~x)~u1 + (~u2 · ~x)~u2

2 −2 8
     
 2  2  0
= 2   − 2   =  .
1 0 2
0 1 −2

5.1.28 Since the three given vectors in the subspace are orthogonal, we have the orthonormal basis
1 1 1
     
1  1  −1 
~u1 = 12   , ~u2 = 12   , ~u3 = 21  .
1 −1 −1
1 −1 1

3
 
 1
Now we can use Theorem 5.1.5, with ~x = ~e1 : projV ~x = (~u1 · ~x)~u1 + (~u2 · ~x)~u2 + (~u3 · ~x)~u3 = 14  .
−1
1

5.1.29 By the Pythagorean theorem (Theorem 5.1.9),

k~xk2 = k7~u1 − 3~u2 + 2~u3 + ~u4 − ~u5 k2

= k7~u1 k2 + k3~u2 k2 + k2~u3 k2 + k~u4 k2 + k~u5 k2
.
= 49 + 9 + 4 + 1 + 1
= 64, so that k~xk = 8.

5.1.30 Since ~y = projV ~x, the vector ~x − ~y is orthogonal to ~y , by definition of an orthogonal projection (see Theo-
rem 5.1.4): (~x − ~y ) · ~y = 0 or ~x · ~y − k~y k2 = 0 or ~x · ~y = k~y k2 . See Figure 5.4.

Figure 5.4: for Problem 5.1.30.

232
Copyright c 2013 Pearson Education, Inc.
 Section 5.1

5.1.31 If V = span(~u1 , . . . , ~um ), then projV ~x = (~u1 · ~x)~u1 + · · · + (~um · ~x)~um , by Theorem 5.1.5, and kprojV ~xk2 =
(~u1 ·~x)2 +· · ·+(~um ·~x)2 = p, by the Pythagorean theorem (Theorem 5.1.9). Therefore p ≤ k~xk2 , by Theorem 5.1.10.
The two quantities are equal if (and only if) ~x is in V .

5.1.32 By Theorem 2.4.9a, the matrix G is invertible if (and only if) (~v1 · ~v1 )(~v2 · ~v2 ) − (~v1 · ~v2 )2
= k~v1 k2 k~v2 k2 − (~v1 · ~v2 )2 6= 0. The Cauchy-Schwarz inequality (Theorem 5.1.11) tells us that k~v1 k2 k~v2 k2 −
(~v1 · ~v2 )2 ≥ 0; equality holds if (and only if) ~v1 and ~v2 are parallel (that is, linearly dependent).

   
x1 1
5.1.33 Let ~x =  · · ·  be a vector in Rn whose components add up to 1, that is, x1 + · · · + xn = 1. Let ~y =  · · · 
xn 1
(all n components are √ 1). The Cauchy-Schwarz inequality (Theorem 5.1.11) tells us that |~x · ~y | ≤ k~xkk~y k, or,
|x1 + · · · + xn | ≤ k~xk n, or k~xk ≥ √1n . By Theorem 5.1.11, the equation k~xk = √1n holds if (and only if)
the vectors ~x and ~y are parallel, that is, x1 = x2 = · · · = xn = n1 . Thus the vector of minimal length is
 1 
n
~x =  · · ·  all components are n1 .


1
n

Figure 5.5 illustrates the case n = 2.

x2

1 1
→ = 2
x
X 1
2

x1
1

x1 + x2 = 1

Figure 5.5: for Problem 5.1.33.

 
1
5.1.34 Let ~x be a unit vector in Rn , that is, k~xk = 1. Let ~y =  . . .  (all n components are 1). The Cauchy-Schwarz
1
√ √
inequality (Theorem 5.1.11) tells us that |~x · ~y | ≤ k~xkk~y k, or, |x1 + . . . + xn | ≤ k~xk n = n. By Theorem 
5.1.11,

√ k
the equation x1 +. . .+xn = n holds if ~x = k~y for positive k. Thus ~x must be a unit vector of the form ~x =  . . . 
k
 1 
√
n  
for some positive k. It is required that nk 2 = 1, or, k = √1n . Thus ~x =  . . .  all components are √1n .
√1
n

Figure 5.6 illustrates the case n = 2.

233
Copyright c 2013 Pearson Education, Inc.
Chapter 5

x2

1
→ √2
x =
X 1
√2

x1
1
x1 + x2 =√ 2

Figure 5.6: for Problem 5.1.34.

   
x 1
5.1.35 Applying the Cauchy-Schwarz inequality to ~u =  y  and ~v =  2  gives |~u · ~v | ≤ k~ukk~v k, or |x + 2y + 3z| ≤
z 3
√ √
14. The minimal value x + 2y + 3z = − 14 is attained when ~u = k~v for negative k. Thus ~u must be a unit
− √114
 
 
k
vector of the form ~u =  2k , for negative k. It is required that 14k 2 = 1, or, k = − √114 . Thus ~u =  √2 
 
 − 14 .
3k − √3 14

   
a 0.2
5.1.36 Let ~x =  b  and ~y =  0.3 . It is required that ~x · ~y = 0.2a + 0.3b + 0.5c = 76. Our goal is to minimize
c 0.5
quantity ~x · ~x = a2 + b2 + c2 . The Cauchy-Schwarz inequality (squared) tells us that (~x · ~y )2 ≤ k~xk2 k~y k2 , or
762
762 ≤ (a2 + b2 + c2 )(0.22 + 0.32 + 0.52 ) or a2+ b2 +  c2 ≥ 0.38 2 2 2
 . The quantity a + b + c is minimal when
a 0.2k
762
a2 + b2 + c2 = 0.38 . This is the case when ~x =  b  =  0.3k  for some positive constant k. It is required that
c 0.5k
0.2a + 0.3b + 0.5c = (0.2)2 k + (0.3)2 k + (0.5)2 k = 0.38k = 76, so that k = 200. Thus a = 40, b = 60, c = 100: The
student must study 40 hours for the first exam, 60 hours for the second, and 100 hours for the third.

5.1.37 Using Definition 2.2.2 as a guide, we find that ref V ~x = 2(projV~x) − ~x = 2(~u1 · ~x)~u1 + 2(~u2 · ~x)~u2 − ~x.

5.1.38 Since ~v1 and ~v2 are unit
vectors, the condition ~v1 · ~v2 = k~v1 kk~v2 k cos(α) = cos(α) = 12 implies that ~v1 and ~v2
enclose an angle of 60◦ = π3 . The vectors ~v1 and ~v3 enclose an angle of 60◦ as well.
In the case n = 2 there are two possible scenarios: either ~v2 = ~v3 , or ~v2 and ~v3 enclose an angle of 120◦ . Therefore,
either ~v2 · ~v3 = 1 or ~v2 · ~v3 = cos(120◦ ) = − 21 .
In the case n = 3, the vectors ~v2 and ~v3 could enclose any angle between 0◦ (if ~v2 = ~v3 ) and 120◦ , as illustrated
in Figure 5.7. We have − 12 ≤ ~v2 · ~v3 ≤ 1.
 √  
3

0
 cos θ
  √2 
 
0
 √3 

For example, consider ~v1 = 0 , ~v2 =  2 , ~v3 =  3
 
sin θ
  
2
1 1
 
2 1
2

3 1
could be anything between − 12 (when sin θ = −1) and 1 (when sin θ = 1), as


Note that ~v2 · ~v3 = 4 sin θ + 4
claimed.

234
Copyright c 2013 Pearson Education, Inc.
 Section 5.1

Figure 5.7: for Problem 5.1.38.

If n exceeds three, we can consider the orthogonal projection w

~ of ~v3 onto the plane E spanned by ~v1 and ~v2 .

Since proj~v1 w ~ v1 = 21 ~v1 , and since kwk

~ = (~v1 · w)~ ~ ≤ k~v3 k = 1, (by Theorem 5.1.10), the tip of w
~ will be on the
line segment in Figure 5.8. Note that the angle φ enclosed by the vectors ~v2 and w ~ is between 0◦ and 120◦ , so
that cos φ is between − 12 and 1.

Therefore, ~v2 · ~v3 = ~v2 · w ~ cos φ is between − 12 and 1.

~ = kwk

This implies that ∠(~v2 , ~v3 ) is between 0◦ and 120◦ as well. To see that all these values are attained, add (n − 3)
zeros to the three vectors ~v1 , ~v2 , ~v3 in R3 given above.

v1

v2
φ

Figure 5.8: for Problem 5.1.38.

5.1.39 No! By definition of a projection, the vector ~x − projL ~x is perpendicular to projL ~x, so that
(~x − projL ~x) · (projL ~x) = ~x · projL ~x − kprojL ~xk2 = 0 and ~x · projL ~x = kprojL ~xk2 ≥ 0. (See Figure 5.9.)

Figure 5.9: for Problem 5.1.39.

√ √
5.1.40 ||~v2 || = ~v2 · ~v2 = a22 = 3.

·~ 20
5.1.41 θ =arccos( ||~v~v22||||~
v3 √ a23
v3 || ) =arccos( a22 a33 ) =arccos( 21 ) ≈ 0.31 radians.
√

p √ √
5.1.42 ||~v1 + ~v2 || = (~v1 + ~v2 ) · (~v1 + ~v2 ) = a11 + 2a12 + a22 = 22.

235
Copyright c 2013 Pearson Education, Inc.
Chapter 5

~
v2 ~
v2
5.1.43 Let ~u = ||~
v2 || =3 . Then, ~ u is an orthonormal basis for span(~v2 ). Using Theorem 5.1.5, proj~v2 (~v1 ) =
(~u · ~v1 )~u = ( 3 · ~v1 ) 3 = 3 (~v2 · ~v1 ) ~v32 = 31 (a12 ) ~v32 = 95 ~v2 .
~
v2 ~
v2 1

20
5.1.44 One method to solve this is to take ~v = ~v2 − proj~v3 ~v2 = ~v2 − 49 ~
v3 .

5.1.45 Write the projection as a linear combination of ~v2 and ~v3 , c2~v2 + c3~v3 . Now you want ~v1 − c2~v2 − c3~v3 to be
perpendicular to V , that is, perpendicular to both ~v2 and ~v3 . Using dot products, this boils down to two linear
25 1
equation in two unknowns, 9c2 + 20c3 = 5, and 20c2 + 49c3 = 11, with the solution c2 = 41 and c3 = − 41 . Thus
25 1
the answer is 41 ~v2 − 41 ~v3 .

5.1.46 Write the projection as a linear combination of ~v1 and ~v2 : c1~v1 + c2~v2 . Now we want ~v3 − c1~v1 + c2~v2 to be
perpendicular to V , that is, perpendicular to both ~v1 and ~v2 . Using dot products, this boils down to two linear
equations in two unknowns, 11 = 3c1 + 5c2 and 20 = 5c1 + 9c2 , with the solution c1 = − 12 , c2 = 52 . Thus, the
answer is − 21 ~v1 + 52 ~v2 .

Section 5.2
In Exercises 1–14, we will refer to the given vectors as ~v1 , . . . , ~vm , where m = 1, 2, or 3.
 
2
5.2.1 ~u1 = k~v11 k ~v1 = 31  1 
−2

 
6
5.2.2 ~u1 = k~v11 k ~v1 = 71  3 
2
 
2
v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1
~u2 = v2⊥ k
k~
= k~
v2 −(~
u1 ·~
v2 )~
u1 k = 17  −6 
3

Note that ~u1 · ~v2 = 0.

 
4
5.2.3 ~u1 = k~v11 k ~v1 = 51  0 
3
 
3
v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1
~u2 = v2⊥ k
k~
= k~
v2 −(~
u1 ·~
v2 )~
u1 k = 15  0 
−4

   
4 3
5.2.4 ~u1 = 51  0  and ~u2 = 1 
5 0  as in Exercise 3.
3 −4
 
0
Since ~v3 is orthogonal to ~u1 and ~u2 , ~u3 = k~v13 k ~v3 =  −1 .
0

236
Copyright c 2013 Pearson Education, Inc.
 Section 5.2

 
2
5.2.5 ~u1 = k~v11 k ~v1 = 31  2 
1
   
−1 −1
v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1
~u2 = v2⊥ k
k~
= k~
v2 −(~
u1 ·~
v2 )~
u1 k = √118  −1  = 1 
√
3 2
−1 
4 4
 
1
5.2.6 ~u1 = k~v11 k ~v1 =  0  = ~e1
0
 
0
v2⊥
~
~u2 = v2⊥ k
k~
= k~~vv22 −(~
u1 ·~
−(~
v2 )~
u1 ·~
u1
v2 )~
u1 k =  1  = ~e2
0
 
0
v3⊥
~ ~ v3 −(~ u1 ·~
v3 )~
u1 −(~u2 ·~v3 )~u2
~u3 = v3⊥ k
k~
= k~v3 −(~u1 ·~v3 )~u1 −(~u2 ·~v3 )~u2 k =  0  = ~e3
1
   
2 −2
5.2.7 Note that ~v1 and ~v2 are orthogonal, so that ~u1 = k~v11 k ~v1 = 13  2  and ~u2 = 1
v2 = 31  1 . Then
v2 k ~
k~
    1 2
2 1
~v⊥
~u3 = k~v3⊥ k = k~~vv33 −(~
u1 ·~
−(~
u
v3 )~
·~
v
1 3
u1 −(~
)~
u 1
u2 ·~
−(~
u
v3 )~
·~
v
2 3
u2
)~
u 2 k = √1  −4  = 1  −2 .
36 3
3
4 2

5
 
4
5.2.8 ~u1 = k~v11 k ~v1 = 71  

2
2
−2
 
 2
~u2 = k~v12 k ~v2 = 71 
5

−4

1
 
1
5.2.9 ~u1 = k~v11 k ~v1 = 21  

1
1
−1
 
v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1 1  7
~u2 = v2⊥ k
= k~
v2 −(~
u1 ·~
v2 )~
u1 k = 10
k~ −7
 
1

1
 
1
5.2.10 ~u1 = k~v11 k ~v1 = 21  

1
1

237
Copyright c 2013 Pearson Education, Inc.
Chapter 5

1
 
v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1  −1 
~u2 = v2⊥ k
= k~
v2 −(~
u1 ·~
v2 )~
u1 k = 12 
k~ 1

−1

4
 
1 1 0
5.2.11 ~u1 = k~v1 k ~v1 = 5  
0
3

−3 −3

  
v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1 1  2  1  2 
~u2 = v2⊥ k
= k~
v2 −(~
u1 ·~
v2 )~
u1 k = √225 = 15  14 
k~ 14

4 4

2
 
3
5.2.12 ~u1 = 71  

0
6

0
 
v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1  −2 
~u2 = v2⊥ k
= k~
v2 −(~
u1 ·~
v2 )~
u1 k = 13 
k~ 2

1

1
 
1
5.2.13 ~u1 = k~v11 k ~v1 = 21  

1
1
 1 
2
 −1 
v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1  2
~u2 = = =

v2⊥ k
k~ k~
v2 −(~
u1 ·~
v2 )~
u1 k  1
 −2


1
2
 1 
2
v3⊥
~ v3 −(~
~ u1 ·~
v3 )~
u1 −(~
u2 ·~
v3 )~
u2
 12 
~u3 = v3⊥ k
k~
= k~
v3 −(~
u1 ·~
v3 )~
u1 −(~
u2 ·~
v3 )~
u2 k = 
 −1


2
− 12

1
 
1 1 7
5.2.14 ~u1 = k~v1 k ~v1 = 10  
1
7

−1
 
v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1  0
~u2 = v2⊥ k
= k~
v2 −(~
u1 ·~
v2 )~
u1 k = √12 
k~ 1

0

238
Copyright c 2013 Pearson Education, Inc.
 Section 5.2

0

v3⊥
~ v3 −(~
~ u1 ·~
v3 )~
u1 −(~
u2 ·~
v3 )~
u2  1
~u3 = v3⊥ k
= k~
v3 −(~
u1 ·~
v3 )~
u1 −(~
u2 ·~
v3 )~
u2 k = √12 
k~ 0

−1

In Exercises 15–28, we will use the results of Exercises 1–14 (note that Exercise k, where k = 1, . . . , 14, gives the
QR factorization of the matrix in Exercise (k + 14)). We can set Q = [~u1 . . . ~um ]; the entries of R are

r11 = k~v1 k
r22 = k~v2⊥ k = k~v2 − (~u1 · ~v2 )~u1 k
r33 = k~v3⊥ k = k~v3 − (~u1 · ~v3 )~u1 − (~u2 · ~v3 )~u2 k
rij = ~ui · ~vj , where i < j.

 
2
5.2.15 Q = 13  1  , R = [3]
−2

 
6 2  
7 0
5.2.16 Q = 17  5 −6 , R =

0 7
2 3

 
4 3  
5 5
5.2.17 Q = 15  0 0 ,R =

0 35
5 −4

   
4 3 0 5 5 0
5.2.18 Q = 15  0 0 −5  , R =  0 35 0 
5 −4 0 0 0 2

− √12
 
2  
1 − √12 , R = 3 1 √1
 
5.2.19 Q = 2
3   0 2
1 √4
2

 
2 3 5
5.2.20 Q = I3 , R = [ ~v1 ~v2 ~v3 ] =  0 4 6
0 0 7

   
2 −2 1 3 0 12
5.2.21 Q = 31  2 1 −2 , R =  0 3 −12 
1 2 2 0 0 6

5 −2
 
 
1 4 2 7 7
5.2.22 Q = 7  , R =
2 5 0 7

2 −4

239
Copyright c 2013 Pearson Education, Inc.
Chapter 5

0.5 −0.1
 
 
 0.5 0.7  2 4
5.2.23 Q =  , R =
0.5 −0.7 0 10

0.5 0.1

1 1
 
 
1 −1  2 10
5.2.24 Q = 12  , R =

1 1 0 2
1 −1

12 −3
 
 
1  0 2 5 10
5.2.25 Q = 15 , R =
0 14 0 15

9 4

2
0

7
3
− 23 
 
7 14

7
5.2.26 Q =  , R =
 
0 2 

0 3
3
6 1
7 3

1 1 1
   
2 1 1
1 1 −1 1
5.2.27 Q = 2  , R =  0 1 −2 
1 −1 −1

0 0 1
1 1 −1
 1

10 − √12 0  
 7
0 √1
 10 10 10
√
10 2
 
5.2.28 Q =  1
, R =  0 2 √0 
 √1 0 

 10 2 0 0 2
7
10 0 − √12

   
1 1 −3 v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1 1 4
5.2.29 ~u1 = v1 k ~
k~ v1 = 5 ~u2 = v2⊥ k
= k~
v2 −(~
u1 ·~
v2 )~
u1 k = 5 . (See Figure 5.10.)
4 k~ 3

Figure 5.10: for Problem 5.2.29.

240
Copyright c 2013 Pearson Education, Inc.
 Section 5.2

5.2.30 See Figure 5.11.

Figure 5.11: for Problem 5.2.30.

 
1
5.2.31 ~u1 = k~v11 k ~v1 =  0  = ~e1
0
       
b b 0 0
v2⊥
~
~v2⊥ = ~v2 − projV1 ~v2 =  c  −  0  =  c , so that ~u2 = v2⊥ k
k~
=  1  = ~e2
0 0 0 0

Here V1 = span(~e1 ) = x axis.

       
d d 0 0
v3⊥
~
~v3⊥ = ~v3 − projV2 ~v3 =  e  −  e  =  0 , so that ~u3 = v3⊥ k
k~
=  0  = ~e3 .
f 0 f 1

Here V2 = span(~e1 , ~e2 ) = x-y plane. (See Figure 5.12.)

Figure 5.12: for Problem 5.2.31.


  
−1 −1
5.2.32 A basis of the plane is ~v1 =  1 , ~v2 =  0 .
0 1

241
Copyright c 2013 Pearson Education, Inc.
Chapter 5

Now apply the Gram-Schmidt process.

 
−1
1
~u1 = v1 k ~
k~ v1 = √12  1 
0
 
−1
v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1
~u2 = v2⊥ k
k~
= k~
v2 −(~
u1 ·~
v2 )~
u1 k = √16  −1 
2

Your solution may be different if you start with a different basis ~v1 , ~v2 of the plane.
 
1 0 0 1
5.2.33 rref(A) =
0 1 1 0

−1 0
   
 0  −1 
A basis of ker(A) is ~v1 =  , ~v2 =  .
0 1
1 0

−1 0
   
 0  −1 
Since ~v1 and ~v2 are orthogonal already, we obtain ~u1 = √12  , ~u2 = √1  .
0 2 1
1 0
 
1 0 −1 −2
5.2.34 rref(A) =
0 1 2 3

1 2
   
 −2   −3 
A basis of ker(A) is ~v1 =  , ~v2 =  .
1 0
0 1
We apply the Gram-Schmidt process and obtain

1
 
1  −2 
~u1 = v1 k ~
k~ v1 = √16 
1

0
2
 
v2⊥
~ v2 −(~
~ u1 ·~
v2 )~
u1  −1 
~u2 = v2⊥ k
= k~
v2 −(~
u1 ·~
v2 )~
u1 k = √130 
k~ −4

3

1
1 0
 
3
1
5.2.35 rref(A) = 0
 1 3

0 0 0

The non-redundant columns of A give us a basis of im(A):

242
Copyright c 2013 Pearson Education, Inc.
 Section 5.2

   
1 2
~v1 =  2 , ~v2 =  1 
2 −2
   
1 2
Since ~v1 and ~v2 are orthogonal already, we obtain ~u1 = 13  2 , ~u2 = 1 
3 1 .
2 −2

1 1 1
   
2 3 5
1 −1 −1 
5.2.36 Write M = 21   0 −4 6

1 −1 1

0 0 7
1 1 −1
↑ ↑
Q0 R0
This is almost the QR factorization of M : the matrix Q0 has orthonormal columns and R0 is upper triangular;
the only problem is the entry −4 on the diagonal of R0 . Keeping in mind how matrices are multiplied, we can
change all the signs in the second column of Q0 and in the second row of R0 to fix this problem:
1 −1 1
   
2 3 5
1 1 −1  
M = 12  0 4 −6 
1 1 −1

0 0 7
1 −1 1
↑ ↑
Q R

1 1 1 1 3 4
   
1  1 −1 −1 1 0 5
5.2.37 Write M = 2 
1 −1 1 −1 0 0
  
1 1 −1 −1 0 0
↑ ↑
Q0 R0
Note that the last two columns of Q0 and the last two rows of R0 have no effect on the product Q0 R0 ; if we drop
them, we have the QR factorization of M :
1 1
 
 
1  1 −1  3 4
M = 2
1 −1 0 5

1 1
↑ ↑
Q R

5.2.38 Since ~v1 = 2~e3 , ~v2 = −3~e1 and ~v3 = 4~e4 are orthogonal, we have
0 −1 0
     
k~v1 k 0 0 2 0 0
 ~v1 ~
v2 ~
v3  0 0 0
Q = k~v1 k k~v2 k k~v3 k =  and R =  0 k~
v 2 k 0  =  0 3 0.
1 0 0

0 0 k~v3 k 0 0 4
0 0 1

243
Copyright c 2013 Pearson Education, Inc.
Chapter 5

     
1 1 −5
5.2.39 ~u1 = √114  2 , ~u2 = √1  1 , ~
3
u3 = ~u1 × ~u2 = √1  4 
42
3 −1 −1
 
k~v1 k 0
5.2.40 If ~v1 , . . . , ~vn are the columns of A, then Q =
 ~
v1
··· ~
vn 
and R = 
 .. .

k~
v1 k k~
vn k .
0 k~vn k

(See Exercise 38 as an example.)

5.2.41 If all diagonal entries of A are positive, then we have Q = In and R = A. A small modification is necessary
if A has negative entries on the diagonal: if aii < 0 we let rij = −aij for all j, and we let qii = −1; if aii > 0 we
let rij = aij and qii = 1. Furthermore, qij = 0 if i 6= j (that is, Q is diagonal).
     
−1 2 3 −1 0 0 1 −2 −3
For example,  0 4 5 =  0 1 0  0 4 5
0 0 −6 0 0 −1 0 0 6
↑ ↑ ↑
A Q R

5.2.42 We have r11 = k~v1 k and r22 = k~v2⊥ k = k~v2 − projL~v2 k, so that r11 r22 is the area of the parallelogram defined
by ~v1 and ~v2 . See Figure 5.13.

Figure 5.13: for Problem 5.2.42.

5.2.43 Partition the matrices Q and R in the QR factorization of A as follows:

 
R1 R2
[ A1 A2 ] = A = QR = [ Q1 Q2 ] = [ Q1 R 1 Q1 R 2 + Q2 R 3 ] ,
0 R3
where Q1 is n × m1 , Q2 is n × m2 , R1 is m1 × m1 , and R3 is m2 × m2 .

Then, A1 = Q1 R1 is the QR factorization of A1 : note that the columns of A1 are orthonormal, and R1 is upper
triangular with positive diagonal entries.

5.2.44 No! If m exceeds n, then there is no n × m matrix Q with orthonormal columns (if the columns of a matrix
are orthonormal, then they are linearly independent).

5.2.45 Yes. Let A = [ ~v1 · · · ~vm ]. The idea is to perform the Gram-Schmidt process in reversed order, starting
with ~um = k~v1m k ~vm .

Then we can express ~vj as a linear combination of ~uj , . . . , ~um , so that [ ~v1 · · · ~vj · · · ~vm ] = [ ~u1 · · · ~uj · · · ~um ] L
for some lower triangular matrix L, with

244
Copyright c 2013 Pearson Education, Inc.
 Section 5.3

l1j
 
 ··· 
~vj = [ ~u1 · · · ~uj · · · ~um ]  ljj  = ljj ~uj + · · · + lmj ~um .
 
···
 
lmj

Section 5.3
5.3.1 Not orthogonal, the column vectors fail to be perpendicular to each other.

5.3.2 This matrix is orthogonal. Check that the column vectors are unit vectors, and that they are perpendicular
to each other.

5.3.3 This matrix is orthogonal. Check that the column vectors are unit vectors, and that they are perpendicular
to each other.

5.3.4 Not orthogonal, the first and third column vectors fail to be perpendicular to each other.

5.3.5 3A will not be orthogonal, because the length of the column vectors will be 3 instead of 1, and they will fail
to be unit vectors.

5.3.6 −B will certainly be orthogonal, since the columns will be perpendicular unit vectors.

5.3.7 AB is orthogonal by Theorem 5.3.4a.

5.3.8 A + B will not necessarily be orthogonal, because the columns may not be unit vectors. For example, if
A = B = In , then A + B = 2In , which is not orthogonal.

5.3.9 B −1 is orthogonal by Theorem 5.3.4b.

5.3.10 This matrix will be orthogonal, by Theorem 5.3.4.

5.3.11 AT is orthogonal. AT = A−1 , by Theorem 5.3.7, and A−1 is orthogonal by Theorem 5.3.4b.

5.3.13 3A is symmetric, since (3A)T = 3AT = 3A.

5.3.14 −B is symmetric, since (−B)T = −B T = −B.

5.3.15 AB is not necessarily symmetric, since (AB)T = B T AT = BA, which is not necessarily the same as AB.
(Here we used Theorem 5.3.9a.)

5.3.16 A + B is symmetric, since (A + B)T = AT + B T = A + B.

5.3.17 B −1 is symmetric, because (B −1 )T = (B T )−1 = B −1 . In the first step we have used 5.3.9b.

5.3.18 A10 is symmetric, since (A10 )T = (AT )10 = A10 .

245
Copyright c 2013 Pearson Education, Inc.
Chapter 5

5.3.19 This matrix is symmetric. First note that (A2 )T = (AT )2 = A2 for a symmetric matrix A. Now we can use
the linearity of the transpose, (2In + 3A − 4A2 )T = 2InT + 3AT − (4A2 )T = 2In + 3A − 4(AT )2 = 2In + 3A − 4A2 .

5.3.20 AB 2 A is symmetric, since (AB 2 A)T = (ABBA)T = (BA)T (AB)T = AT B T B T AT = AB 2 A.

5.3.21 Symmetric. (AT A)T = AT (AT )T = AT A.

5.3.22 BB T is symmetric: (BB T )T = (B T )T B T = BB T .

5.3.23 Not necessarily symmetric. (A − AT )T = AT − A = −(A − AT ).

5.3.24 Not necessarily symmetric. (AT BA)T = AT (AT B)T = AT B T A.

5.3.25 Symmetric, because (AT B T BA)T = AT B T (B T )T (AT )T = AT B T BA.

5.3.26 Symmetric, since (B(A + AT )B T )T = ((A + AT )B T )T B T = B(A + AT )T B T

= B(AT + A)T B T = B((AT )T + AT )B T = B(A + AT )B T .

~ = (A~v )T w
5.3.27 Using Theorems 5.3.6 and 5.3.9a, we find that (A~v ) · w ~ = ~v T AT w
~ = ~v · (AT w),
~ as claimed.

5.3.28 We will follow the hint.

(iv) ⇒ (vi) : If AT A = In , then (A~x) · (A~y ) = (A~x)T (A~y ) = ~xT AT A~y = ~xT In ~y = ~x · ~y for all ~x and ~y .
p √
(vi) ⇒ (ii) : If (A~x) · (A~y ) = ~x · ~y for all ~x and ~y , then kA~xk = (A~x) · (A~x) = ~x · ~x = k~xk for all ~x.
Recall that the equivalence of statements (i) through (v) is proven in the text.

5.3.29 We will use the fact that L preserves length (by Definition 5.3.1) and the dot product, by Summary 5.3.8
(vi).
v )·L(w)
L(~ ~ ·w
∠(L(~v ), L(w))
~ = arccos kL(~
v )kkL(w)k
~ = arccos k~v~vkk~
~ = ∠(~
wk v , w).
~

5.3.30 If L(~x) = ~0, then kL(~x)k = k~xk = 0, so that ~x = ~0. Therefore, ker(L) = {~0}.

By Theorem 3.3.7, dim(im(L)) = m − dim(ker(L)) = m.

Since Rn has an m-dimensional subspace (namely, im(L)), the inequality m ≤ n holds.

The transformation L preserves right angles (the proof of Theorem 5.3.2 applies), so that the columns of A are
orthonormal (since they are L(~e1 ), . . . , L(~em )).

Therefore, we have AT A = Im (the proof of Theorem 5.3.7 applies).

Since the vectors ~v1 , . . . , ~vm form an orthonormal basis of im(A), the matrix AAT represents the orthogonal
projection onto im(A), by Theorem 5.3.10.
   
1 0   x1
x 1
A simple example of such a transformation is L(~x) =  0 1  ~x, that is, L =  x2 .
x2
0 0 0

246
Copyright c 2013 Pearson Education, Inc.
 Section 5.3

5.3.31 Yes! If A is orthogonal, then so is AT , by Exercise 11. Since the columns of AT are orthogonal, so are the
rows of A.

1 0
5.3.32 a No! As a counterexample, consider A =  0 1  (see Exercise 30).
0 0

b Yes! More generally, if A and B are n × n matrices such that BA = In , then AB = In , by Theorem 2.4.8c.

 
cos(φ)
5.3.33 Write A = [ ~v1 ~v2 ]. The unit vector ~v1 can be expressed as ~v = , for some φ. Then ~v2 will be one
   sin(φ) 
− sin(φ) sin(φ)
of the two unit vectors orthogonal to ~v1 : ~v2 = or ~v2 = . (See Figure 5.7.)
cos(φ) − cos(φ)
   
cos(φ) − sin(φ) cos(φ) sin(φ)
Therefore, an orthogonal 2 × 2 matrix is either of the form A = or A = ,
sin(φ) cos(φ) sin(φ) − cos(φ)
representing a rotation or a reflection. Compare with Exercise 2.2.24. See Figure 5.14.

Figure 5.14: for Problem 5.3.33.

 
a b
5.3.34 Since the first two columns are orthogonal to the third, we have c = d = 0. Then is an ortho-
 e f 
cos(φ) − sin(φ) 0
gonal 2 × 2 matrix; By Exercise 33, the 3 × 3 matrix A is either of the form A =  0 0 1  or
  sin(φ) cos(φ) 0
cos(φ) sin(φ) 0
A= 0 0 1 .
sin(φ) − cos(φ) 0

5.3.35 Let us first think about the inverse L = T −1 of T .

247
Copyright c 2013 Pearson Education, Inc.
Chapter 5

2
3
Write L(~x) = A~x = [ ~v1 ~v2 ~v3 ] ~x. It is required that L(~e3 ) = ~v3 =  23 .
1
3

− 23
 
1
Furthermore, the vectors ~v1 , ~v2 , ~v3 must form an orthonormal basis of R3 . By inspection, we find ~v1 =  3
.
2
3

− 13
   
−2 −1 2
Then ~v2 = ~v1 × ~v3 =  23  does the job. In summary, we have L(~x) = 13  1 2 2  ~x.
− 23 2 −2 1

Since the matrix of L is orthogonal, the matrix of T = L−1 is the transpose of the matrix of L:
 
−2 1 2
T (~x) = 31  −1 2 −2  ~x.
2 2 1

There are many other answers (since there are many choices for the vector ~v1 above).

2
√1 √1

3 2 18
2
5.3.36 Let the third column be the cross product of the first two: A = 

3 − √12 √1
18 .

1
3 0 − √418

There is another solution, with the signs in the last column reversed.
       
2 −3 3 2
5.3.37 No, since the vectors  3  and  2  are orthogonal, whereas  0  and  −3  are not (see Theorem 5.3.2).
0 0 2 0

 
0 a b
5.3.38 a The general form of a skew-symmetric 3 × 3 matrix is A =  −a 0 c , with
−b −c 0
 2
−a − b2

−bc ac
A2 =  −bc −a2 − c2 −ab , a symmetric matrix.
ac −ab −b2 − c2

b By Theorem 5.3.9.a, (A2 )T = (AT )2 = (−A)2 = A2 , so that A2 is symmetric.

5.3.39 By Theorem 5.3.10, the matrix of the projection is ~u~uT ; the ij th entry of this matrix is ui uj .

0.5 −0.1
   
 0.5   0.7 
5.3.40 An orthonormal basis of W is ~u1 =   , ~u2 =   (see Exercise 5.2.9).
0.5 −0.7
0.5 0.1

By Theorem 5.3.10, the matrix of the projection onto W is QQT , where Q = [ ~u1 ~u2 ].

248
Copyright c 2013 Pearson Education, Inc.
 Section 5.3

26 18 32 24
 
T 1  18 74 −24 32 
QQ = 100 
32 −24 74 18

24 32 18 26

1
 
.
5.3.41 A unit vector on the line is ~u = √1n  .. .
1
1
The matrix of the orthogonal projection is ~u~uT , the n × n matrix whose entries are all n (compare with Exer-
cise 39).

5.3.42 a Suppose we are projecting onto a subspace W of Rn . Since A~x is in W already, the orthogonal projection
of A~x onto W is just A~x itself: A(A~x) = A~x, or A2 ~x = A~x.
Since this equation holds for all ~x, we have A2 = A.

b A = QQT , for some matrix Q with orthonormal columns ~u1 , . . . , ~um . Note that QT Q = Im , since the ij th entry
of QT Q is ~ui · ~uj . Then A2 = QQT QQT = Q(QT Q)QT = QIm QT = QQT = A.

5.3.43 Examine how A acts on ~u, and on a vector ~v orthogonal to ~u:

A~u = (2~u~uT − I3 )~u = 2~u~uT ~u − ~u = ~u, since ~uT ~u = ~u · ~u = k~uk2 = 1.
A~v = (2~u~uT − I3 )~v = 2~u~uT ~v − ~v = −~v , since ~uT ~v = ~u · ~v = 0.
Since A leaves the vectors in L = span(~u) unchanged and reverses the vectors in V = L⊥ , it represents the
reflection about L.
Note that B = −A, so that B reverses the vectors in L and leaves the vectors in V unchanged; that is, B
represents the reflection about V .

5.3.44 Note that AT is an m × n matrix. By Theorems 3.3.7 and 5.3.9c we have

dim(ker(AT )) = n − rank(AT ) = n − rank(A).
By Theorem 3.3.6, dim(im(A)) = rank(A), so that dim(im(A)) + dim(ker(AT )) = n.

5.3.45 Note that AT is an m × n matrix. By Theorems 3.3.7 and 5.3.9c, we have

dim(ker(A)) = m − rank(A) and dim(ker(AT )) = n − rank(AT ) = n − rank(A),
so that dim(ker(A)) = dim(ker(AT )) if (and only if) A is a square matrix.

5.3.46 By Theorem 5.2.2, the columns ~u1 , . . . , ~um of Q are orthonormal. Therefore, QT Q = Im , since the ij th entry
of QT Q is ~ui · ~uj .
If we multiply the equation M = QR by QT from the left then QT M = QT QR = R, as claimed.

5.3.47 By Theorem 5.2.2, the columns ~u1 , . . . , ~um of Q are orthonormal. Therefore, QT Q = Im , since the ij th entry
of QT Q is ~ui · ~uj .
By Theorem 5.3.9a, we now have AT A = (QR)T QR = RT QT QR = RT R.

249
Copyright c 2013 Pearson Education, Inc.
Chapter 5

5.3.48 As suggested, we consider the QR factorization

AT = P R
of AT , where P is orthogonal and R is upper triangular with positive diagonal entries.
By Theorem 5.3.9a, A = (P R)T = RT P T .
Note that L = RT is lower triangular and Q = P T is orthogonal.

5.3.49 Yes! By Exercise 5.2.45, we can write AT = P L, where P is orthogonal and L is lower triangular.
By Theorem 5.3.9a, A = (P L)T = LT P T .
Note that R = LT is upper triangular, and Q = P T is orthogonal (by Exercise 11).

5.3.50 a If an n×n matrix A is orthogonal and upper triangular, then A−1 is both lower triangular (since A−1 = AT )
and upper triangular (being the inverse of an upper triangular matrix; compare with Exercise 2.4.35c).
Therefore, A−1 = AT is a diagonal matrix, and so is A itself. Since A is orthogonal with positive diagonal entries,
all the diagonal entries must be 1, so that A = In .

b Using the terminology suggested in the hint, we observe that Q−1 −1

2 Q1 is orthogonal (by Theorem 5.3.4) and R2 R1
−1 −1
is upper triangular with positive diagonal entries. By part a, the matrix Q2 Q1 = R2 R1 is In , so that Q1 = Q2
and R1 = R2 , as claimed.

5.3.51 a Using the terminology suggested in the hint, we observe that Im = QT1 Q1 = (Q2 S)T Q2 S = S T QT2 Q2 S =
S T S, so that S is orthogonal, by Theorem 5.3.7.

b Using the terminology suggested in the hint, we observe that R2 R1−1 is both orthogonal (let S = R2 R1−1 in part
a) and upper triangular, with positive diagonal entries. By Exercise 50a, we have R2 R1−1 = Im , so that R1 = R2 .
Then Q1 = Q2 R2 R1−1 = Q2 , as claimed.
 
a b c
5.3.52 Applying the strategy outlined in Summary 4.1.6 to the general element  b d e  of V , we find the basis
           c e f
1 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0
0 0 0,1 0 0,0 0 0,0 1 0 · 0 0 1  ,  0 0 0 , so that dim(V ) = 6.
0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 1
 
0 b c
5.3.53 Applying the strategy outlined in Summary 4.1.6 to the general element  −b 0 e  of V , we find the
      −c −e 0
0 1 0 0 0 1 0 0 0
basis  −1 0 0  ,  0 0 0,0 0 1 , so that dim(V ) = 3.
0 0 0 −1 0 0 0 −1 0

5.3.54 To write the general form of a skew-symmetric n × n matrix A, we can place arbitrary constants above the
diagonal, the opposite entries below the diagonal (aij = −aji ), and zeros on the diagonal (since aii = −aii ). See
Exercise 53 for the case n = 3. Thus the dimension of the space equals the number of entries above the diagonal
of an n × n matrix. In Exercise 55 we will see that there are (n2 − n)/2 such entries. Thus dim(V ) = (n2 − n)/2.

250
Copyright c 2013 Pearson Education, Inc.
 Section 5.3

5.3.55 To write the general form of a symmetric n × n matrix A, we can place arbitrary constants on and above
the diagonal, and then write the corresponding entries below the diagonal (aij = aji ). See Exercise 52 for the
case n = 3. Thus the dimension of the space equals the number of entries on and above the diagonal of an n × n
matrix. Now there are n2 entries in the matrix, n2 − n off the diagonal, and half of them, (n2 − n)/2, above the
diagonal. Since there are n entries on the diagonal, we have dim(V ) = (n2 − n)/2 + n = (n2 + n)/2.

5.3.56 Yes and yes (see Exercise 57).

5.3.57 Yes, L is linear, since L(A + B) = (A + B)T = AT + B T = L(A) + L(B) and L(kA) = (kA)T = kAT = kL(A).
Yes, L is an isomorphism; the inverse is the transformation R(A) = AT from Rn×m to Rm×n .

5.3.58 Adapting the solution of Exercise 59, we see that the kernel consists of all skew-symmetric matrices, and the
image consists of all symmetric matrices.

1
5.3.59 The kernel consists of all matrixes A such that L(A) = 2 (A − AT ) = 0, that is, AT = A; those are the
symmetric matrices.

Following the hint, let’s apply L to a skew-symmetric matrix A, with AT = −A. Then L(A) = (1/2)(A − AT ) =
(1/2)2A = A, so that A is in the image of L. Conversely, if A is any 2 × 2 matrix, then L(A) will be skew-
symmetric, since (L(A))T = (1/2)(A − AT )T = (1/2)(AT − A) = −L(A). In conclusion: The kernel of L consists
of all symmetric matrices, and the image consists of all skew-symmetric matrices.

1 0 0 0
 
0 1 0 0
5.3.60 Using Theorem 4.3.2, we find the matrix  .
0 0 1 0
0 0 0 −1

5.3.61 Note that the first three matrices of the given basis B are symmetric, so that L(A) = A − AT = 0, and the
coordinate vector [L(A)]B is ~0 for all three of them. The last matrix of the basis is skew-symmetric, so that
0 0 0 0
 
0 0 0 0
L(A) = 2A, and [L(A)]B = 2~e4 . Using Theorem 4.3.2, we find that the B-matrix of L is  .
0 0 0 0
0 0 0 2
   
1 0 0 1 0 0 1 3 2
5.3.62 By Theorem 5.3.9a, AT =  1 1 0   0 −1 0   0 1 0 .
−1 2 1 0 0 2 0 0 1

5.3.63 By Exercise 2.4.94b, the given LDU factorization of A is unique.

By Theorem 5.3.9a, A = AT = (LDU )T = U T DT LT = U T DLT is another way to write the LDU factorization
of A (since U T is lower triangular and LT is upper triangular). By the uniqueness of the LDU factorization, we
have U = LT (and L = U T ), as claimed.

A −B T C −DT A + C −B T − DT
     
5.3.64 a + =
 B AT D  CT B+D AT + C T
A + C −(B + D)T
= is of the required form.
B+D (A + C)T

251
Copyright c 2013 Pearson Education, Inc.
Chapter 5

−B T −kB T −(kB)T
     
A kA kA
b k = = is of the required form.
B AT kB kAT kB (kA)T

p −q −r −s

q p s −r 
c The general element of H is M =  , with four arbitrary constants, r,s,p, and q. Thus
r −s p q
s r −q p
dim(H) = 4; use the strategy outlined in Summary 4.1.6 to construct a basis.

A −B T C −DT AC − B T D −ADT − B T C T AC − B T D −(BC + AT D)T

      
d = =
B AT D CT BC + AT D −BDT + AT C T BC + AT D (AC − B T D)T
is of the required form.
Note that A, B, C, D, and their transposes are rotation-dilation matrices, so that they all commute.
T
−B T AT BT
  
A
e = is of the required form.
B AT −B (AT )T
p
f Note that the columns ~v1 , ~v2 , ~v3 , ~v4 or M are orthogonal, and they all have length p2 + q 2 + r2 + s2 . Now
M T M is the 4 × 4 matrix whose ij th entry is ~vi · ~vj , so that M T M = (p2 + q 2 + r2 + s2 )I4 .

1
g If M 6= 0, then k = p2 + q 2 + r2 + s2 > 0, and T


kM M = I4 , so that M is invertible, with
1
M −1 = T
p2 +q 2 +r 2 +s2 M .

By parts b and e, M −1 is in H as well.

0 −1 0 0 0 0 −1 0
   
1 0 0 0 0 0 0 −1 
h No! A =   and B =   do not commute (AB = −BA).
0 0 0 1 1 0 0 0
0 0 −1 0 0 1 0 0
  a b

a b
5.3.65 Write 10A = ; it is required that a, b, c and d be integers. Now A = 10c
10
d must be an orthogonal
c d 10 10
a 2 c 2
matrix, implying that ( 10 ) + ( 10 ) = 1, or a2 + c2 = 100. Checking the squares of all integers from 1 to 9, we
see that there are only two ways to write 100 as a sum of two positive perfect squares: 100 = 36 + 64 = 64 + 36.
Since a and c are required to be positive, we have either a = 6 and c = 8 or a = 8 and c = 6. In each case
we have two options for the second column of A, namely, the two unit vectors perpendicular to the first column
vector. Thus we end up with four solutions:
       
.6 −.8 .6 .8 .8 −.6 .8 .6
A= , , or .
.8 .6 .8 −.6 .6 .8 .6 −.8
 
0.8 −0.6
5.3.66 One approach is to take one of the solutions from Exercise 65, say, the rotation matrix B = ,
  0.6 0.8
0.28 −0.96
and then let A = B 2 = . Matrix A is orthogonal by Theorems 5.3.4a.
0.96 0.28

5.3.67 a We need to show that AT A~c = AT ~x, or, equivalently, that AT (~x − A~c) = ~0. But AT (~x − A~c) = AT (~x −
c1~v1 − · · · − cm~vm ) is the vector whose ith component is

252
Copyright c 2013 Pearson Education, Inc.
 Section 5.3

(~vi )T (~x − c1~v1 − · · · − cm~vm ) = ~vi · (~x − c1~v1 − · · · − cm~vm ), which we know to be zero.

b The system AT A~c = AT ~x has a unique solution ~c for a given ~x, since ~c is the coordinate vector of projV ~x with
respect to the basis ~v1 , . . . , ~vm . Thus the coefficient matrix AT A must be invertible, so that we can solve for ~c
and write ~c = (AT A)−1 AT ~x. Then projV ~x = c1~v1 + · · · + cm~vm = A~c = A(AT A)−1 AT ~x.

5.3.68 If A = QR, then A(AT A)−1 AT = QR(RT QT QR)−1 RT QT = QR(RT R)−1 RT QT = QRR−1 (RT )−1 RT QT =
QQT , as in Theorem 5.3.10. The equation QT Q = Im holds since the columns of Q are orthonormal.

5.3.69 We will use the terminology introduced in Theorem √ 5.3.10. Since the two given vectors ~v1 and ~v2 are
orthogonal (verify that ~v1 ·~v2 = 0), with k~v1 k = k~v2 k = 3, we have the orthonormal basis ~u1 = √13 ~v1 , ~u2 = √13 ~v2
, so
     
1 0 1 0   1 1 −1 0
 1 1  1  1 1 
Q = √13   1 1 −1 0 1  1 2 0 −1 
T
 
 −1 1  and PW = QQ = 3  −1 1  0 1 1 −1 = 3  −1 0
 .
2 −1 
0 −1 0 −1 0 −1 −1 1

5.3.70 a. We will use the terminology introduced in Theorem 5.3.10. Consider the basis B = (~v1 , ~v2 ) of V presented √
in Exercise 4.3.73c. Since the two vectors ~v1 and ~v2 are orthogonal (verify that ~v1 ·~v2 = 0), with k~v1 k = k~v2 k = 6,
we have the orthonormal basis ~u1 = √16 ~v1 , ~u2 = √16 ~v2 , so
     
0 2 0 2   4 −2 2 0
 1 −1 
Q = √16   and PV = QQT = 1  1 −1  0 1 1 2 = 1  −2 2 0 2  .
   
 1 1  6  1 1  2 −1 1 0 6  2 0 2 2 
2 0 2 0 0 2 2 4
       
1 x1 0 x1
 1   x2   1   x2 
 −1  ·  x3  = x1 + x2 − x3 = 0 and  1  ·  x3  = x2 + x3 − x4 = 0,
b. If ~x is any vector in V , then        

0 x4 −1 x4
by definition of V . Since the given basis vectors of W are orthogonal to V , the space W is the orthogonal
compliment of V , meaning that W = V ⊥ and V = W ⊥ . By definition of a projection, we have ~x = projW ~x +
projV ~x for all ~x in R4 , so that PW + PV = I4 .
Also, projW (projV ~x) = projV (projW ~x) = ~0 for all ~x in R4 , meaning that PV PW = PW PV = 0, the zero matrix
of size 4 × 4.

5.3.71 If A and B are Hankel matrices of size n × n, and C = A + B, then cij = aij + bij = ai+1,j−1 + bi+1,j−1 =
ci+1,j−1 for all i = 1, ..., n − 1 and for all j = 2, ..., n, showing that C is a Hankel matrix as well. An analogous
argument shows that the Hankel matrices are closed under scalar multiplication.
Now, what is the dimension of the space Hn of the Hankel matrices of size n×n? In the case n = 4, the dimension
is 7, since there are 7 free variables in the matrix
 
a b c d
 b c d e 
A=  c d e f .


d e f g
For an arbitrary n, we can choose the entries in the first column and those in the last row freely; the other entries
are then determined by those choices. Because an1 belongs both to the first column and to the last row, we have
dim Hn = 2n − 1.

253
Copyright c 2013 Pearson Education, Inc.
Chapter 5

5.3.72 We will use the terminology introduced in Theorem 5.3.10. Let

a2 an−1
   
1 1 a ...
2
 a   a a a3 ... an 
1 1  1
 2
  2 3

Q = ~u = k~vk ~v = k~vk  a T
 and P = QQ =

 a
v k2 
k~
a a4 ... an+1 ,

 ...   ... ... ... ... ... 
an−1 an−1 an an+1 ... a2n−2
2 1
where k~v k = 1 + a2 + a4 + ... + a2n−2 . Note that the ij th entry of P is pij = v k2
k~
ai+j−2 so that P is a Hankel
1 1
matrix since pij = v k2
k~
ai+j−2 = v k2
k~
a(i+1)+(j−1)−2 = pi+1,j−1 .
   
1 1 2 4
2 1 
In the case of ~v =  2  we have k~v k = 1 + 4 + 16 = 21, so P = 21
2 4 8 . This is indeed a Hankel
4 4 8 16
matrix.
√ √
(1+ 5)(1− 5)
5.3.73 a. Note that ab = 4 = −1. Now

~a · ~b = 1 + ab + (ab)2 + (ab)3 + ... + (ab)n−2 + (ab)n−1 = 1 − 1 + 1 − 1 + ... + 1 − 1 = 0,

so that ~a is orthogonal to ~b, as claimed.

b. If we let ~u1 = k~a1k ~a and ~u2 = ~1b ~b , then P = ~u1 ~uT1 + ~u2 ~uT2 ; see the paragraph preceding Theorem 5.3.10.
kk
In Exercise 72 we saw that both ~u1 ~uT1 and ~u2 ~uT2 are Hankel matrices, and in Exercise 71 we saw that the set of
Hankel matrices is closed under addition. Thus P is a Hankel matrix as well.

5.3.74 a.
m -5 -4 -3 -2 -1 0 1 2 3 4 5
fm 5 -3 2 -1 1 0 1 1 2 3 5
b. We conjecture that fm = f−m for odd m and fm = −f−m for even m.
We will prove this conjecture by induction, with the base case, f1 = f−1 , being established in the table above.
Assuming the result for positive integers < m, we will prove it for m.
For odd m, we have fm = fm−1 + fm−2 = −f1−m + f2−m = −f1−m + (f1−m + f−m) = f−m
For even m, we have fm = fm−1 + fm−2 = f1−m − f2−m = f1−m − (f1−m + f−m) = −f−m .
   
0 5
 1   −3 
   
 1   2 
c. ~v =   , with k~v k2 = kwk 2
, w~ = ~ = 40 = 5 · 8 = f5 f6 and ~v · w
~ = 0.
  
 2   −1 

 3   1 
5 0
d. If k is an odd integer between 0 and n, then the kth summand of ~v · w
~ is fk−1 f−n+k = fk−1 fn−k , while the
(n + 1 − k)th summand is fn−k f1−k = −fn−k fk−1 ; these two summands add up to 0. As we sum over all odd
integers from 1 to n − 1, we see that ~v · w
~ = 0.
1
e. We have P = k~v1k2 ~v~v T + kwk
~ 2
w
~w 1
~ T = fn−1 v~v T + w
fn (~ ~w~ T ); see the paragraph preceding Theorem 5.3.10.
Considering the first and last components of ~v and w,~ we can write the first and the last columns of P :
       
1  ~0 ... 1 1
P = fn−1~v  +  fn−1 w
~ ... ~0  =  fn−1 w
~ .... fn−1~v  =  w
~ .... ~v  .
fn−1 fn fn−1 fn fn

254
Copyright c 2013 Pearson Education, Inc.
 Section 5.4

Since P is a Hankel matrix, it is determined by its first and last columns, so that P has the form given in the
exercise.
2 2
f. Using the result k~v k = kwk
~ = 40 = 5 · 8 = f5 f6 from part c, we find
   
0 5 25 −15 10 −5 5 0

 1 −3 
 

 −15 10 −5 5 0 5 

1  1 2  0 1 1 2 3 5 1  10 −5 5 0 5 5 
P =  =  
40 
  5
2 −1  −3 2 −1 0 0 40 
 −5 5 0 5 5 10 

 3 1   5 0 5 5 10 15 
5 0 0 5 5 10 15 25

 
5 −3 2 −1 1 0

 −3 2 −1 1 0 1 

1 2 −1 1 0 1 1 
=  .
8
 −1 1 0 1 1 2 

 1 0 1 1 2 3 
0 1 1 2 3 5

Section 5.4
 
T −3
5.4.1 A basis of ker(A ) is . (See Figure 5.15.)
2

Figure 5.15: for Problem 5.4.1.

 
1
5.4.2 A basis of ker(AT ) is  −2 . im(A) is the plane perpendicular to this line.
1

255
Copyright c 2013 Pearson Education, Inc.
Chapter 5

5.4.3 We will first show that the vectors ~v1 , . . . , ~vp , w ~ q span Rn . Any vector ~v in Rn can be written as
~ 1, . . . , w
~v = ~vk + ~v⊥ , where ~vk is in V and ~v⊥ is in V ⊥ (by definition of orthogonal projection, Theorem 5.1.4).

Now ~vk is a linear combination of ~v1 , . . . , ~vp , and ~v⊥ is a linear combination of w
~ 1, . . . , w
~ q , showing that the
vectors ~v1 , . . . , ~vp , w ~ q span Rn .
~ 1, . . . , w

Note that p + q = n, by Theorem 5.1.8c; therefore, the vectors ~v1 , . . . , ~vp , w ~ q form a basis of Rn , by
~ 1, . . . , w
Theorem 3.3.4d.

5.4.4 By Theorem 5.4.1, the equation (im B)⊥ = ker(B T ) holds for any matrix B. Now let B = AT . Then
(im(AT ))⊥ = ker(A). Taking transposes of both sides and using Theorem 5.1.8d we obtain im(AT ) = (kerA)⊥ ,
as claimed.
 
1 1 1 1
5.4.5 V = ker(A), where A = .
1 2 5 4

Then V ⊥ = (kerA)⊥ = im(AT ), by Exercise 4.

The two columns of AT form a basis of V ⊥ :

1 1
   
1 2
 , 
1 5
1 4

5.4.6 Yes! For any matrix A,

im(A) = (ker(AT ))⊥ = (ker(AAT ))⊥ = (ker(AAT )T )⊥ = im(AAT ).
↑ ↑ ↑
Theorem 5.4.1 Theorem 5.4.2a Theorems 5.4.1 and 5.1.8d.

5.4.7 im(A) and ker(A) are orthogonal complements by Theorem 5.4.1:

(imA)⊥ = ker(AT ) = ker(A)

5.4.8 a By Theorem 5.4.6, L+ (~y ) = (AT A)−1 AT ~y .

The transformation L+ is linear since it is “given by a matrix,” by Definition 2.1.1.

b If L (and therefore A) is invertible, then L+ (~y ) = A−1 (AT )−1 AT ~y = A−1 ~y = L−1 ~y , so that L+ = L−1 .

c L+ (L(~x)) =


the unique least-squares solution ~u of L(~u) = L(~x) = ~x.

d L(L+ (~y )) = A(AT A)−1 AT ~y = projV ~y , where V = im(A), by Theorem 5.4.7.

 
1 0  
1 0 0
e Here A =  0 1 . Then L+ (~y ) = (AT A)−1 AT ~y = ~y .
0 1 0
0 0

5.4.9 ~x0 is the shortest of all the vectors in S. (See Figure 5.16.)

256
Copyright c 2013 Pearson Education, Inc.
 Section 5.4

Figure 5.16: for Problem 5.4.9.

5.4.10 a If ~x is an arbitrary solution of the system A~x = ~b, let ~xh = projV ~x, where V = ker(A), and ~x0 = ~x − projV ~x.
Note that ~b = A~x = A(~xh + ~x0 ) = A~xh + A~x0 = A~x0 , since ~xh is in ker(A).

b If ~x0 and ~x1 are two solutions of the system A~x = ~b, both from (kerA)⊥ , then ~x1 − ~x0 is in the subspace (kerA)⊥
as well. Also, A(~x1 − ~x0 ) = A~x1 − A~x0 = ~b − ~b = ~0, so that ~x1 − ~x0 is in ker(A). By Theorem 5.1.8b, it follows
that ~x1 − ~x0 = ~0, or ~x1 = ~x0 , as claimed.

c Write ~x1 = ~xh + ~x0 as in part a; note that ~xh is orthogonal to ~x0 . The claim now follows from the Pythagorean
Theorem (Theorem 5.1.9).

5.4.11 a Note that L+ (~y ) = AT (AAT )−1 ~y ; indeed, this vector is in im(AT ) = (kerA)⊥ , and it is a solution of
L(~x) = A~x = ~y .

b L(L+ (~y )) = ~y , by definition of L+ .

c L+ (L(~x)) = AT (AAT )−1 A~x = projV ~x, where V = im(AT ) = (kerA)⊥ , by Theorem 5.4.7.

d im(L+ ) = im(AT ), by part c, and ker(L+ ) = {~0} (if ~y is in ker(L+ ), then ~y = L(L+ (~y )) = L(~0) = ~0, by part b).
 
  1 0
1 0 0
e Let A = ; then the matrix of L+ is AT (AAT )−1 = AT =  0 1 .
0 1 0
0 0

5.4.12 By Theorem 5.4.5, the least-squares solutions of the linear system A~x = ~b are the exact solutions of the
(consistent) system AT A~x = AT ~b. The minimal solution of this normal equation (in the sense of Exercise 10) is
called the minimal least-squares solution of the system A~x = ~b.
Equivalently, the minimal least-squares solution of A~x = ~b can be defined as the minimal solution of the consistent
system A~x = projV ~b, where V = im(A).

257
Copyright c 2013 Pearson Education, Inc.
Chapter 5

5.4.13 a Suppose that L+ (~y1 ) = ~x1 and L+ (~y2 ) = ~x2 ; this means that ~x1 and ~x2 are both in (kerA)⊥ = im(AT ), AT A~x1 =
AT ~y1 , and AT A~x2 = AT ~y2 . Then ~x1 + ~x2 is in im(AT ) as well, and AT A(~x1 + ~x2 ) = AT (~y1 + ~y2 ), so that
L+ (~y1 + ~y2 ) = ~x1 + ~x2 .
The verification of the property L+ (k~y ) = kL+ (~y ) is analogous.

b L+ (L(~x)) is the orthogonal projection of ~x onto (kerA)⊥ = im(AT ).

c L(L+ (~y )) is the orthogonal projection of ~y onto im(A) = (kerAT ))⊥ .

d im(L+ ) = im(AT ) and ker(L+ ) = ker(AT ), by parts b and c.

 y1  1 
  2 2 0
y1
e L+ =  0 , so that the matrix of L+ is  0 0 .
y2
0 0 0

Figure 5.17: for Problem 5.4.14.

5.4.14 L+ (w~ 1 ) is the minimal solution of the system L(~x) = w

~ 1 . The line S in Figure 5.17 shows all solutions of
the system
~ 1 (compare with Exercise 9). The minimal solution, L+ (w
L(~x) = w ~ 1 ), is perpendicular to ker(L).
L+ ( w ~ 2 ) = L+ (~0) = ~0
~ 2 ) = L+ (projim(L) w

L+ ( w
~ 3 ) = L+ (projim(L) w
~ 3 ) ≈ L+ (0.55w
~ 1 ) = 0.55L+ (w
~ 1)

5.4.15 Note that (AT A)−1 AT A = In ; let B = (AT A)−1 AT .

5.4.16 If A is an m × n matrix, then

dim(imA)⊥ = m − dim(imA) = m − rank(A)
↑ ↑

Theorem 5.1.8c Theorem 3.3.6

and dim(ker(AT )) = m − rank(AT ).

↑

Theorem 3.3.7
It follows that rank(A) = rank(AT ), as claimed.

258
Copyright c 2013 Pearson Education, Inc.
 Section 5.4

5.4.17 Yes! By Theorem 5.4.2, ker(A) = ker(AT A). Taking dimensions of both sides and using Theorem 3.3.7, we
find that n − rank(A) = n − rank(AT A); the claim follows.

5.4.18 Yes! By Exercise 17, rank(A) = rank(AT A). Substituting AT for A in Exercise 17 and using Theorem 5.3.9c,
we find that rank(A) = rank(AT ) = rank(AAT ). The claim follows.
 
∗ T −1 T~ 1
5.4.19 ~x = (A A) A b= , by Theorem 5.4.6.
1
 
  −1
2
5.4.20 Using Theorem 5.4.6, we find ~x∗ = and ~b − A~x∗ =  1 .
2
1

Note that ~b − A~x∗ is perpendicular to the two columns of A.

 
  −12
−1
5.4.21 Using Theorem 5.4.6, we find ~x∗ = and ~b − A~x∗ =  36 , so that k~b − A~x∗ k = 42.
2
−18
 
3
5.4.22 Using Theorem 5.4.6, we find ~x∗ = and ~b − A~x∗ = ~0. This system is in fact consistent and ~x∗ is the
−2
exact solution; the error k~b − A~x∗ k is 0.

5.4.23 Using Theorem 5.4.6, we find ~x∗ = ~0; here ~b is perpendicular to im(A).

5.4.24 Using Theorem 5.4.6, we find ~x∗ = [2].

    
5 15 x1 5
5.4.25 In this case, the normal equation AT A~x = AT ~b is = , which simplifies to x1 + 3x2 = 1,
 15 45
 x 2 15
1 − 3t
or x1 = 1 − 3x2 . The solutions are of the form ~x∗ = , where t is an arbitrary constant.
t
 
t − 67
    
66 78 90 x1 1
5.4.26 Here, the normal equation AT A~x = AT ~b is  78 93 108   x2  =  2 , with solutions ~x∗ =  1 − 2t ,
 
90 108 126 x3 3 t
where t is an arbitrary constant.

5.4.27 The least-squares solutions of the system SA~x = S~b are the exact solutions of the normal equation (SA)T SA~x =
(SA)T S~b.
Note that S T S = In , since S is orthogonal; therefore, the normal equation simplifies as follows: (SA)T SA~x =
AT S T SA~x = AT A~x and (SA)T S~b = AT S T S~b = AT ~b, so that the normal equation is AT A~x = AT ~b, the same
as the normal equation of the  system
 A~x = ~b. Therefore, the systems A~x = ~b and SA~x = S~b have the same
7
least-squares solution, ~x∗ = .
11

5.4.28 The least-squares solutions of the system A~x = ~un are the exact solutions of A~x = projim(A) ~un . Note that
~un is orthogonal to im(A), so that projim(A) ~un = ~0, and the unique least-squares solution is ~x∗ = ~0.

259
Copyright c 2013 Pearson Education, Inc.
Chapter 5

 −1    1 
1+ε 1 1+ε
1+ε
5.4.29 By Theorem 5.4.6, ~x∗ = (AT A)−1 AT ~b = = ≈ 21 , where ε = 10−20 .
1
2+ε
1 1+ε 1+ε
1+ε 2
    
1 1 x1 1
If we use a hand-held calculator, due to roundoff errors we find the normal equation = , with
1 1 x2 1
infinitely many solutions.

5.4.30 We attempt to solve the system

   
c0 + 0c1 = 0 1 0   0
c
c0 + 0c1 = 1 , or  1 0  0 =  1  .
c1
c0 + 1c1 = 1 1 1 1

1
c∗0
 
2
This system cannot be solved exactly; the least-squares solution is = . The line that fits the data
c∗1 1
2
1
points best is f ∗ (t) = 2 + 12 t.

Figure 5.18: for Problem 5.4.30.

The line goes through the point (1, 1) and “splits the difference” between (0, 0) and (0, 1). See Figure 5.18.
 
c0
5.4.31 We want such that
c1    
3 = c0 + 0c1 1 0   3
c
3 = c0 + 1c1 or  1 1  1 =  3  .
c2
6 = c0 + 1c1 1 1 6
    −1  
1 0  ∗   1 0   3
~ c 0 1 1 1 1 1 1  
Since ker 1 1 = {0},
  =   1 1   3
c1 0 1 1 0 1 1
1 1 1 1 6
−1    
3

3 2 12
= = 3 so f ∗ (t) = 3 + 23 t. (See Figure 5.19.)
2 2 9 2
 
c0
5.4.32 We want  c1  of f (t) = c0 + c1 t + c2 t2 such that
c2
27 = c0 + 0c1 + 0c2 1 0 0   27
   
c0
0 = c0 + 1c1 + 1c2  1 1 1     0 
or   c1 =  
0 = c0 + 2c1 + 4c2 1 2 4 0
c2
0 = c0 + 3c1 + 9c2 1 3 9 0

260
Copyright c 2013 Pearson Education, Inc.
 Section 5.4

Figure 5.19: for Problem 5.4.31.

If we call the coefficient matrix A, we notice that ker(A) = {~0} so

27
 ∗    
c0 25.65
 0
 c1  = (AT A)−1 AT   =  −28.35  so f ∗ (t) = 25.65 − 28.35t + 6.75t2 .

0
c2 6.75
0



c0
5.4.33 We want  c1  such that
c2

0 = c0 + sin(0)c1 + cos(0)c2 1 0 1 0
    
c
1 = c0 + sin(1)c1 + cos(1)c2 1 sin(1) cos(1)   1   1 
or   c2 =   .
2 = c0 + sin(2)c1 + cos(2)c2 1 sin(2) cos(2) 2
c3
3 = c0 + sin(3)c1 + cos(3)c2 1 sin(3) cos(3) 3
∗ 
c1
Since the coefficient matrix has kernel {~0}, we compute  c2  using Theorem 5.4.6, obtaining
c3
 ∗  
c0 1.5
 c1  ≈  0.1  so f ∗ (t) ≈ 1.5 + 0.1 sin t − 1.41 cos t.
c2 −1.41

c0


 c1 
5.4.34 We want  c2  such that
 
c3
 
c4

1 0 1 0 1 0
   
1 sin(0.5) cos(0.5) sin(1) cos(1)  c0  0.5 
 
1 sin(1) cos(1) sin(2) cos(2)   c1   1 
   
1 sin(1.5) cos(1.5) sin(3) cos(3)   c2  =  1.5 
    
1 sin(2) cos(2) sin(4) cos(4)  c3 2 
    
1 sin(2.5) cos(2.5) sin(5) cos(5) c4 2.5
   
1 sin(3) cos(3) sin(6) cos(6) 3

Since the columns of the coefficient matrix are linearly independent, its kernel is {~0}. We can use Theorem 5.4.6

261
Copyright c 2013 Pearson Education, Inc.
 Another Random Document on
Scribd Without Any Related Topics
raiment, wearing on his bold front a crown, and sallying forth on a
red steed.
Some infernal Duke would appear in his proper character, quietly
seated on a griffin; another spirit of a similar rank would display the
three heads of a serpent, a man, and a cat; he would also bestride a
viper, and carry in his hand a firebrand; another of the same stamp,
would appear like a duchess, encircled with a fiery zone, and
mounted on a camel; a fourth would wear the aspect of a boy, and
amuse himself on the back of a two-headed dragon. A few spirits,
however, would be content with the simple garbs of a horse, a
leopard, a lion, an unicorn, a night-raven, a stork, a peacock, or a
dromedary; the latter animal speaking fluently the Egyptian
language. Others would assume the more complex forms of a lion or
of a dog, with a griffin’s wings attached to each of their shoulders; or
of a bull equally well gifted; or of the same animal, distinguished by
the singular appendage of a man’s face; or of a crow clothed with
human flesh; or of a hart with a fiery tail. To certain other noble
devils were assigned such shapes as those of a dragon with three
heads, one of these being human; of a wolf with a serpent’s tail,
breathing forth flames of fire; of a she wolf exhibiting the same
caudal appendage, together with a griffin’s wings, and ejecting
hideous matter from the mouth. A lion would appear either with the
head of a branded thief, or astride upon a black horse, and playing
with a viper, or adorned with the tail of a snake, and grasping in his
paws two hissing serpents. These were the varied shapes assumed by
devils of rank. To those of an inferior order were consigned upon
earth, the duty of carrying away condemned souls. These were
described as blacker than pitch: as having teeth like lions, nails on
their fingers like those of the wild boar, on their forehead horns,
through the extremities of which, poison was emitted, having wide
ears flowing with corruption, and discharging serpents from their
nostrils, and having cloven feet[54]. But this last appendage, as Sir
Thomas Brown has learnedly proved, is a mistake, which has arisen
from the devil frequently appearing to the Jews in the shape of a
rough and hairy goat, this animal being the emblem of sin-
offerings[55].
It is worthy of farther remark, says Dr. Hibbert, that the forms of
the demons described by St. Bernard, differs little from that which is
no less carefully pourtrayed by Reginald Scott, 350 years later, and,
perhaps, by the Demonologists of the present day. “In our
childhood,” says he, “our mothers’ maids have so terrified us with an
ouglie devell having hornes on his head, fier in his mouth, and a taile
in his breech, eies like a bason, fangs like a dog, clawes like a bear, a
skin like a tiger, and a voice roaring like a lion,—whereby we start
and are afraid when we heare one cry bough.”
It is still an interesting matter of speculation worth noticing—why,
after the decay of the regular systems of demonology taught in the
middle ages, the same hideous form should still be attached to the
devil? The learned Mede has remarked, “that the devil could not
appear in human shape while man was in his integrity; because he
was a spirit fallen from his first glorious perfection; and, therefore,
must appear in such a shape which might argue his imperfection and
abasement, which was the shape of a beast; otherwise, no reason can
be given, why he should not rather have appeared to Eve in the shape
of a woman than of a serpent. But since the fall of man, the case is
altered: now we know he can take upon him the shape of man. He
appears, it seems, in the shape of man’s imperfection, either for age
or deformity, as like an old man (for so the witches say); and perhaps
it is not altogether false, which is vulgarly affirmed, that the devil
appearing in human shape, has always a deformity of some uncouth
member or other, as though he could not yet take upon him human
shape entirely, for that man himself is not entirely and utterly fallen
as he is.” Grose, with considerable less seriousness, observes, that
“although the devil can partly transform himself into a variety of
shapes, he cannot change his cloven feet, which will always mark him
under every appearance.
The late Dr. Ferriar took some trouble to trace to their real source
spectral figures, which have been attributed to demoniacal visits. In
his observations on the works of Remy, the commissioner in
Lorraine, for the trial of witches, he makes the following remark:
—“My edition of this book was printed by Vincente, at Lyons, in
1595; it is entitled Dæmonolatria. The trials appear to have begun in
1583. Mr. Remy seems to have felt great anxiety to ascertain the
exact features and dress of the demons, with whom many people
supposed themselves to be familiar. Yet nothing transpired in his
examinations, which varied from the usual figures exhibited by the
gross sculptures and paintings of the middle age. They are said to be
black faced, with sunk but fiery eyes, their mouths wide and swelling
of sulphur, their hands hairy, with claws, their feet horny and
cloven.” In another part of Dr. Ferriar’s, the following account is also
given of a case which passed under his own observation:—“I had
occasion,” he observes, “to see a young married woman, whose first
indication of illness was a spectral delusion. She told me that her
apartment appeared to be suddenly filled with devils, and that her
terror impelled her to quit the house with great precipitation. When
she was brought back, she saw the whole staircase filled with
diabolical forms, and was in agonies of fear for several days. After the
first impression wore off, she heard a voice tempting her to self
destruction, and prohibiting her from all exercises of piety. Such was
the account given by her when she was sensible of the delusion, yet
unable to resist the horror of the impression. When she was newly
recovered, I had the curiosity to question her, as I have interrogated
others, respecting the forms of the demons with which she had been
claimed; but I never could obtain any other account, than that they
were very small, very much deformed, and had horns and claws like
the imps of our terrific modern romances.” To this illustration of the
general origin of the figures of demoniacal illusions, I might observe,
that, in the case of a patient suffering under delirium tremens, which
came under my notice, the devils who flitted around his bed were
described to me as exactly like the forms that he had recently seen
exhibited on the stage in the popular drama of Don Giovanni.
With the view of illustrating other accounts of apparitions, I shall
now return to the doctrine of demonology which was once taught.
Although the leading tenets of this occult science may be traced to
the Jews and early Christians, yet they were matured by our early
communication with the Moors of Spain, who were the chief
philosophers of the dark ages, and between whom and the natives of
France and Italy, a great communication subsisted. Toledo, Seville,
and Salamanca, became the greatest schools of magic. At the latter
city, prelections on the black art were, from a consistent regard to the
solemnity of the subject, delivered within the walls of a vast and
gloomy cavern. The schoolmen taught, that all knowledge might be
obtained from the assistance of the fallen angels. They were skilled in
the abstract sciences, in the knowledge of precious stones, in
alchymy, in the various languages of mankind and of the lower
animals, in the belles lettres, in moral philosophy, pneumatology,
divinity, magic, history, and prophecy. They could controul the
winds, the waters, and the influence of the stars; they could raise
earthquakes, induce diseases, or cure them, accomplish all vast
mechanical undertakings, and release souls out of purgatory. They
could influence the passions of the mind—procure the reconciliation
of friends or foes—engender mutual discord—induce mania and
melancholy—or direct the force and objects of the sexual affections.
Such was the object of demonology, as taught by its most orthodox
professors. Yet other systems of it were devised, which had their
origin in causes attending the propagation of Christianity. For it
must have been a work of much time to eradicate the universal belief,
that the Pagan deities, who had become so numerous as to fill every
part of the universe, were fabulous beings. Even many learned men
were induced to side with the popular opinion on the subject, and did
nothing more than endeavour to reconcile it with their acknowledged
systems of demonology. They taught that such heathen objects of
reverence were fallen angels in league with the prince of darkness,
who, until the appearance of our Saviour, had been allowed to range
on the earth uncontrolled, and to involve the world in spiritual
darkness and delusion. According to the various ranks which these
spirits held in the vast kingdom of Lucifer, they were suffered, in
their degraded state, to take up their abode in the air, in mountains,
in springs, or in seas. But, although the various attributes ascribed to
the Greek and Roman deities, were, by the early teachers of
Christianity, considered in the humble light of demoniacal delusions,
yet for many centuries they possessed great influence over the minds
of the vulgar. In the reign of Adrian, Evreux, in Normandy, was not
converted to the Christian faith, until the devil, who had caused the
obstinacy of the inhabitants, was finally expelled from the temple of
Diana. To this goddess, during the persecution of Dioclesian,
oblations were rendered by the inhabitants of London. In the 5th
century, the worship of her existed at Turin, and incurred the rebuke
of St. Maximus. From the ninth to the fifteenth century, several
denunciations took place of the women who, in France and Germany,
travelled over immense spaces of the earth, acknowledging Diana as
their mistress and conductor. In rebuilding St. Paul’s cathedral, in
London, remains of several of the animals used in her sacrifices were
found; for slight traces of this description of reverence, subsisted so
late as the reign of Edward the First, and of Mary. Apollo, also, in an
early period of Christianity, had some influence at Thorney, now
Westminster. About the 11th century, Venus formed the subject of a
monstrous apparition, which could only have been credited from the
influence which she was still supposed to possess. A young man had
thoughtlessly put his ring around the marble finger of her image.
This was construed by the Cyprian goddess as a plighted token of
marriage; she accordingly paid a visit to her bridegroom’s bed at
night, nor could he get rid of his bed-fellow until the spells of an
exorcist had been invoked for his relief. In the year 1536, just before
the volcanic eruption of Mount Etna, a Spanish merchant, while
travelling in Sicily, saw the apparition of Vulcan attended by twenty
of his Cyclops, as they were escaping from the effects which the over
heating of his furnace foreboded[56].
To the superstitions of Greece and Rome, we are also indebted for
those subordinate evil spirits called genii, who for many centuries
were the subject of numerous spectral illusions. A phantasm of this
kind appeared to Brutus in his tent, prophesying that he should be
again seen at Philippi. Cornelius Sylla had the first intimation of the
sudden febrile attack with which he was seized, from an apparition
who addressed him by his name; concluding, therefore, that his
death was at hand, he prepared himself for the event, which took
place the following evening. The poet Cassius Severus, a short time
before he was slain by order of Augustus, saw, during the night, a
human form of gigantic size,—his skin black, his beard squalid, and
his hair dishevelled. The phantasm was, perhaps, not unlike the evil
genius of Lord Byron’s Manfred:—
“I see a dusk and awful figure rise
Like an infernal god from out the earth;
His face wrapt in a mantle, and his form
Robed as with angry clouds; he stands between
Thyself and me—but I do fear him not.”

The emperor Julian was struck with a spectre clad in rags, yet
bearing in his hands a horn of plenty, which was covered with a linen
cloth. Thus emblematically attired, the spirit walked mournfully past
the hangings of the apostate’s tent[57].
 We may now advert to the superstitious narratives of the middle
ages, which are replete with the notices of similar marvellous
apparitions. When Bruno, the Archbishop of Wirtzburg, a short
period before his sudden death, was sailing with Henry the Third, he
descried a terrific spectre standing upon a rock which overhung the
foaming waters, by whom he was hailed in the following words:
—“Ho! Bishop, I am thy evil genius. Go whether thou choosest, thou
art and shalt be mine. I am not now sent for thee, but soon thou shalt
see me again.” To a spirit commissioned on a similar errand, the
prophetic voice may be probably referred, which was said to have
been heard by John Cameron, the Bishop of Glasgow, immediately
before his decease. He was summoned by it, says Spottiswood, “to
appear before the tribunal of Christ, there to atone for his violence
and oppressions.”
“I shall not pursue the subject of Genii much farther. The notion of
every man being attended by an evil genius, was abandoned much
earlier than the far more agreeable part of the same doctrine, which
taught that, as an antidote to this influence, each individual was also
accompanied by a benignant spirit. “The ministration of angels,” says
a writer in the Athenian oracle, “is certain, but the manner how, is
the knot to be untied.” ’Twas generally thought by the ancient
philosophers, that not only kingdoms had their tutelary guardians,
but that every person had his particular genius, or good angel, to
protect and admonish him by dreams, visions, &c. We read that
Origin, Hierome, Plato, and Empedocles, in Plutarch, were also of
this opinion; and the Jews themselves, as appears by that instance of
Peter’s deliverance out of prison. They believed it could not be Peter,
but his angel. But for the particular attendance of bad angels, we
believe it not; and we must deny it, till it finds better proof than
conjecture.”
Such were the objects of superstitious reverence, derived from the
Pantheon of Greece and Rome, the whole synod of which was
supposed to consist of demons, who were still actively bestirring
themselves to delude mankind. But in the West of Europe, a host of
other demons, far more formidable, were brought into play, who had
their origin in Celtic, Teutonic, and even Eastern fables; and as their
existence, as well as influence, was not only by the early Christians,
but even by the reformers, boldly asserted, it was long before the
rites to which they had been accustomed were totally eradicated.
Thus in Orkney, for instance, it was customary, even during the last
century, for lovers to meet within the pale of a large circle of stones,
which had been dedicated to the chief of the ancient Scandinavian
deities. Through a hole in one of the pillars, the hands of contracting
parties were joined, and the faith they plighted, was named the
promise of Odin, to violate which was infamous. But the influence of
the Dii Majores of the Edda was slight and transient, in comparison
with that of the duergar or dwarfs, who figure away in the same
mythology, and whose origin is thus recited. Odin and his brothers
killed the giant Ymor, from whose wound ran so much blood that all
the families of the earth were drowned, except one that saved himself
on board a bark. These gods then made, of the giant’s bones of his
flesh and his blood, the earth, the waters, and the heavens. But in the
body of the monster, several worms had in the course of putrefaction
been engendered, which, by order of the gods, partook of both
human shape and reason. These little beings possessed the most
delicate figures, and always dwelt in subterraneous caverns or clefts
in the rocks. They were remarkable for their riches, their activity, and
their malevolence[58]. This is the origin of our modern faries, who, at
the present day, are described as a people of small stature, gaily drest
in habiliments of green[59]. They possess material shapes, with the
means, however, of making themselves invisible. They multiply their
species; they have a relish for the same kind of food that affords
sustenance to the human race, and when, for some festal occasion,
they would regale themselves with good beef or mutton, they employ
elf arrows to bring down their victims. At the same time, they delude
the shepherds with the substitution of some vile substance, or
illusory image, possessing the same form as that of the animal they
had taken away. These spirits are much addicted to music, and when
they make their excursions, a most exquisite band of music never
fails to accompany them in their course. They are addicted to the
abstraction of the human species, in whose place they leave
substitutes for living beings, named Changelings, the unearthly
origin of whom is known by their mortal imbecility, or some wasting
disease. When a limb is touched with paralysis, a suspicion often
arises that it has been touched by these spirits, or that, instead of the
sound member, an insensible mass of matter has been substituted in
its place.
 In England, the opinions originally entertained relative to the
duergar or dwarfs, have sustained considerable modifications, from
the same attributes being assigned to them as to the Persian peris, an
imaginary race of intelligences, whose offices of benevolence were
opposed to the spightful interference of evil spirits. Whence this
confusion in proper Teutonic mythology has originated, is doubtful;
conjectures have been advanced, that it may be traced to the
intercourse the Crusaders had with the Saracens; and that from
Palestine was imported the corrupted name, derived from the peris,
of faries; for under such a title the duegar of the Edda are now
generally recognized; the malevolent character of the dwarfs being
thus sunk in the opposite qualities of the peris, the fairies. Blessing
became in England, proverbial: “Grant that the sweet fairies may
nightly put money in your shoes, and sweep your house clean.” In
more general terms, the wish denoted, “Peace be to the house[60].
Fairies, for many centuries, have been the objects of spectral
impressions. In the case of a poor woman of Scotland, Alison
Pearson, who suffered for witchcraft in the year 1586, they probably
resulted from some plethoric state of the system, which was followed
by paralysis. Yet, for these illusive images, to which the popular
superstition of the times had given rise, the poor creature was
indicted for holding communication with demons, under which light
fairies were then considered, and burnt at a stake. During her illness,
she was not unfrequently impressed with sleeping and waking
visions, in which she held an intercourse with the queen of the
Elfland and the good neighbours. Occasionally, these capricious
spirits would condescend to afford her bodily relief; at other times,
they would add to the severity of her pains. In such trances or
dreams, she would observe her cousin, Mr. William Sympsoune, of
Stirling, who had been conveyed away to the hills by the fairies, from
whom she received a salve that would cure every disease, and of
which the Archbishop of St. Andrews deigned himself to reap the
benefit. It is said in the indictment against her, that “being in Grange
Muir with some other folke, she, being sick, lay downe; and, when
alone, there came a man to her clad in green, who said to her, if she
would be faithful, he would do her good; but she being feared, cried
out; but nae bodie came to her, so she said, if he came in God’s name,
and for the gude of her soul, it was all well; but he gaed away; he
appeared another tyme like a lustie man, and many men and women
with him—at seeing him she signed herself, and pray and past with
them, and saw them making merrie with pypes, and gude cheir and
wine;—she was carried with them, and when she telled any of these
things, she was sairlie tormented by them, and the first time she gaid
with them, she gat a sair straike frae one of them, which took all the
poustie (power) of her side frae her, and left an ill-far’d mark on her
side.
“She saw the gude neighbours make their saws (salves) with panns
and fyres, and they gathered the herbs before the sun was up, and
they cam verie fearful sometimes to her, and flaire (scared) her very
sair, which made her cry, and threatened they would use her worse
than before; and at last, they tuck away the power of her haile syde
frae her, and made her lye many weeks. Sometimes they would come
and sit by her, and promise that she should never want if she would
be faithful, but if she would speak and telle of them, they would
murther her. Mr. William Sympsoune is with them who healed her,
and telt her all things;—he is a young man, not six yeares older than
herself, and he will appear to her before the court comes;—he told
her he was taken away by them; and he bid her sign herself that she
be not taken away, for the teind of them are tane to hell every
yeare[61].”
Another apparition of a similar kind may be found on the
pamphlet which was published A. D. 1696, under the patronage of
Dr. Fowler, Bishop of Glocester, relative to Ann Jefferies, “who was
fed for six months by a small sort of airy people, called fairies.” There
is every reason to suppose, that this female was either affected with
hysteria, or with that highly excited state of nervous irritability,
which, as I have shewn, gives rise to ecstatic illusions. The account of
her first fit is the only one which relates to the present subject. In the
year 1695, says her historian, “she then being nineteen years of age,
and one day knitting in an arbour in the garden, there came over the
hedges to her (as she affirmed) six persons of small stature, all
clothed in green, and which she called fairies: upon which she was so
frightened, that she fell into a kind of convulsive fit: but when we
found her in this condition, we brought her into the house, and put
her to bed, and took great care of her. As soon as she was recovered
out of the fit, she cries out, ‘they are just gone out of the window;
they are just gone out of the window. Do you not see them?’ And
thus, in the height of her sickness, she would often cry out, and that
with eagerness; which expressions we attributed to her distemper,
supposing her light-headed.” This narrative of the girl seemed highly
interesting to her superstitious neighbours, and she was induced to
relate far more wonderful stories, upon which not the least
dependance can be placed, as the sympathy she excited eventually
induced her to become a rank impostor[62].
But besides fairies, or elves, which formed the subject of many
spectral illusions, a domestic spirit deserves to be mentioned, who
was once held in no small degree of reverence. In most northern
countries of Europe there were few families that were without a
shrewd and knavish sprite, who, in return for the attention or neglect
which he experienced, was known to
——“sometimes labour in the quern,
And bootless make the breathless housewife churn;
And sometimes make the drink to bear no barm!”

Mr. Douce, in his Illustrations of Shakspeare, has shewn, that the

Samogitæ, a people formerly inhabiting the shores of the Baltic, who
remained idolatrous so late as the 15th century, had a deity named
Putseet, whom they invoked to live with them, by placing in the barn,
every night, a table covered with bread, butter, cheese, and ale. If
these were taken away, good fortune was to be expected; but if they
were left, nothing but bad luck. This spirit is the same as the goblin-
groom, Puck, or Robin Good-fellow of the English, whose face and
hands were either of a russet or green colour, who was attired in a
suit of leather, and armed with a flail. For a much lesser fee than was
originally given him, he would assist in threshing, churning, grinding
malt or mustard, and sweeping the house at midnight[63]. A similar
tall “lubbar fiend,” habited in a brown garb, was known in Scotland.
Upon the condition of a little wort being laid by for him, or the
occasional sprinkling, upon a sacrificial stone, of a small quantity of
milk, he would ensure the success of many domestic operations.
According to Olaus Magnus, the northern nations regarded domestic
spirits of this description, as the souls of men who had given
themselves up during life to illicit pleasures, and were doomed, as a
punishment, to wander about the earth, for a certain time, in the
peculiar shape which they assumed, and to be bound to mortals in a
sort of servitude. It is natural, therefore, to expect, that these familiar
spirits would be the subjects of many apparitions, of which a few
relations are given in Martin’s Account of the Second Sight in
Scotland. “A spirit,” says this writer, “called Browny, was frequently
seen in all the most considerable families in the isles and the north of
Scotland, in the shape of a tall man; but within these twenty or thirty
years, he is seen but rarely.”
It is useless to pursue this subject much farther: in the course of a
few centuries, the realms of superstition were increased to almost an
immeasurable extent; the consequence was, that the air, the rocks,
the seas, the rivers, nay, every lake, pool, brook, or spring, were so
filled with spirits, both good and evil, that of each province it might
be said, in the words of the Roman satirist, “Nosiba regio tam plena
est numinibus, ut facilius possis deum quam hominem invenire.”
Hence the modification which took place of systems of demonology,
so as to admit of the classification of all descriptions of devils,
whether Teutonic, Celtic, or Eastern systems of mythology. “Our
schoolmen and other divines,” says Burton in his Anatomy of
Melancholy, “make nine kinds of bad devils, as Dionysius hath of
angels. In the first rank, are those false gods of the Gentiles, which
were adored heretofore in several idols, and gave oracles at Delphos
and elsewhere, whose prince is Beelzebub. The second rank is of liars
and equivocators, as Apollo, Pythias, and the like. The third are those
vessels of anger, inventors of all mischief, as that of Theutus in Plato.
Esay calls them vessels of fury: their prince is Belial. The fourth are
malicious, revengeful devils, and their prince is Asmodeus. The fifth
kind are coseners, such as belong to magicians and witches; their
prince is Satan. The sixth are those aërial devils that corrupt the air,
and cause plagues, thunders, fires, &c. spoken of in Apocalypse and
Paule; the Ephesians name them the prince of the air: Meresin is
their prince. The seventh is a destroyer, captaine of the furies,
causing wars, tumults, combustions, uproares, mentioned in the
Apocalypse, and called Abaddon. The eighth is that accusing or
calumniating devil, whom the Greeks call Διάβολος, that drives us to
despair. The ninth are those tempters in several kindes, and their
prince is Mammon.”
But this arrangement was not comprehensive enough; for, as
Burton adds, “no place was void, but all full of spirits, devils, or other
inhabitants; not so much as an haire-breadth was empty in heaven,
earth, or waters, above or under the earth; the earth was not so full of
flies in summer as it was at all times of invisible devils.”
Pneumatologists, therefore, made two grand distinctions of demons;
there were celestial demons, who inhabited the regions higher than
the moon; while those of an inferior rank, as the Manes or Lemures,
were either nearer the earth, or grovelled on the ground. Psellus,
however, “a great observer of the nature of devils,” seems to have
thought, that such a classification destroyed all distinction between
good and evil spirits: he, therefore, denied that the latter ever
ascended the regions above the moon, and contending for this
principle, founded a system of demonology, which had for its basis
the natural history and habitations of all demons. He named his first
class fiery devils. They wandered in the region near the moon, but
were restrained from entering into that luminary; they displayed
their power in blazing stars, in fire-drakes, in counterfeit suns and
moons, and in the euerpo santo, or meteoric lights, which, in vessels
at sea, flit from mast to mast, and forebode foul weather. It was
supposed that these demons occasionally resided in the furnaces of
Hecla, Etna, or Vesuvius. The second class consisted of aërial devils.
They inhabited the atmosphere, causing tempests, thunder and
lightning; rending asunder oaks, firing steeples and houses, smiting
men and beasts, showering down from the skies, stones[64], wool, and
even frogs; counterfeiting in the clouds the battles of armies, raising
whirlwinds, fires, and corrupting the air, so as to induce plagues. The
third class was terrestrial devils, such as lares, genii, fawns, satyrs,
wood-nymphs, foliots, Robin good-fellows, or trulli. The fourth class
were aqueous devils; as the various description of water-nymph, or
mermen, or of merwomen. The fifth were subterranean devils, better
known by the name dæmones itallici, metal-men, Getuli or Cobals.
They preserved treasure in the earth, and prevented it from being
suddenly revealed; they were also the cause of horrible earthquakes.
Psellus’s sixth class of devils were named lucifugi; they delighted in
darkness; they entered into the bowels of men, and tormented those
whom they possessed with phrenzy and the falling sickness. By this
power they were distinguished from earthly and aërial devils; they
could only enter into the human mind, which they either deceived or
provoked with unlawful affections.
 Nor were speculations wanting with regard to the common nature
of these demons. Psellus conceived that their bodies did not consist
merely of one element, although he was far from denying that this
might have been the case before the fall of Lucifer. It was his opinion,
that devils possessed corporeal frames capable of sensation; that
they could both feel and be felt; they could injure and be hurt; that
they lamented when they were beaten, and that if struck into the fire,
they even left behind them ashes,—a fact which was demonstrated in
a very satisfactory experiment made by some philosophers upon the
borders of Italy; that they were nourished with food peculiar to
themselves, not receiving the aliment through the gullet, but
absorbing it from the exterior surface of their bodies, after the
manner of a sponge; that they did not hurt cattle from malevolence,
but from mere love of the natural and temperate heat and moisture
of these animals; that they disliked the heat of the sun, because it
dried too fast; and, lastly, that they attained a great age. Thus,
Cardan had a fiend bound to him twenty-eight years, who was forty-
two years old, and yet considered very young. He was informed, from
this very authentic source of intelligence, that devils lived from two
to three hundred years, and that their souls died with their bodies.
The very philosophical statement was, nevertheless, combated by
other observers. “Manie,” says Scot, “affirmed that spirits were of
aier, because they had been cut in sunder and closed presentlie
againe, and also because they vanished away so suddenlie.”
“The Narrative of the Demon of Tedworth,
or the disturbances at Mr. Monpesson’s
house, caused by Witchcraft and Villainy
of the Drummer.”
“In winter’s tedious nights, sit by the fire
With good old folks; and let them tell the tales
Of woeful ages long ago betid.”

“Mr. John Monpesson of Tedworth, in the County of Wilts, being

about the middle of March, in the year 1661, at a neighbouring town
called Ludlow, and hearing a drummer beat there, he enquired of the
bailiff of the town at whose house he then was, what it meant. The
bailiff told him, that they had for some days past been annoyed by an
idle drummer, who demanded money of the constable by virtue of a
pretended pass, which he thought was counterfeited. On hearing
this, Mr. Monpesson sent for the fellow, and asked him by what
authority he went up and down the country in that manner with his
drum. The drummer answered, that he had good authority, and
produced his pass, with a warrant under the hands of Sir William
Cawley, and Colonel Ayliff, of Gretenham. Mr. Monpesson, however,
being acquainted with the hand-writing of these gentlemen,
discovered that the pass and warrant were counterfeit, upon which
he commanded the vagrant to lay down his drum, and at the same
time gave him in charge to a constable, to carry him before the next
justice of the peace, to be farther examined and punished. The fellow
then confessed that the pass and warrant were forged, and begged
earnestly to be forgiven and to have his drum restored: upon this Mr.
Monpesson told him, that if, upon enquiry from Colonel Ayliff,
whose drummer he represented himself to be, he should turn out to
be an honest man, he should listen to his entreaty and have the drum
back again; but that, in the mean time, he would take care of it. The
drum, therefore, was left in the bailiff’s hand; and the drummer went
off in charge of the constable, who, it appears, was prevailed upon, by
the fellow’s entreaties, to allow him to escape.
 About the middle of April following, at a time when Mr.
Monpesson was preparing for a journey to London, the bailiff sent
the drum to his house. On his return from his journey, his wife
informed him that they had been very much alarmed in the night by
thieves, and that the house had like to have been torn down. In
confirmation of this alarm, Mr. Monpesson had not been above three
nights at home, when the same noise was again heard which had
disturbed the family in his absence. It consisted of a tremendous
knocking at the doors, and thumping on the walls of the house; upon
which Mr. M. got out of bed, armed himself with a brace of pistols,
opened the street door to ascertain the cause, which he had no
sooner done, than the noise removed to another door, which he also
opened, went out, and walked round the house; but could discover
nothing, although he heard a strange noise and hollow sound. He
had no sooner returned and got into bed, than he was again
disturbed by a noise and drumming on the top of the house, which
continued for a length of time, and then gradually subsided, as if it
went off into the air.
The noise of thumping and drumming, after this, was very
frequent; usually for five nights together, when there would be an
intermission of three. The noise was on the outside of the house,
which principally consisted of board; and usually came on just as the
family was going to bed, whether that happened early or late. After
continuing these annoyances for a month on the outside of the
house, it at length made bold to come into the room where the drum
lay, four or five nights in every seven; coming always on after they
had got into bed, and continuing for two hours after. The signal for
the appearance of the noise was the hearing of a hurling of the air
over the house; and when it was about to retire, the drum would beat
the same as if a guard were being relieved. It continued in this room
for the space of two months, during which time Mr. Monpesson lay
there to observe it. In the early part of the night, it used to be very
troublesome, but after it had continued two hours, all would be quiet
again.
During the prevalence of this disturbance, Mrs. Monpesson was
brought to bed, and the night on which this occurrence took place,
there was but very little noise made, nor any at all for the three
subsequent weeks of her confinement. After this polite and well-
timed cessation, it returned in a sudden and more violent manner
than before; it followed and teased their youngest children, and beat
against their bedstead so violently that every moment they were
expected to be broken to pieces. On placing their hands upon them at
this time, no blows were felt, although they were perceived to shake
exceedingly. For an hour together the drum would beat roundheads
and cuckold, the tat-too, and several other martial pieces, as well as
any drummer could possibly execute them. After this, a scratching
would be heard under the children’s beds, as if something that had
iron claws were at work. It would lift the children up in their beds,
follow them from one room to another, and for a while only haunted
them, without playing any other pranks.
There was a cockloft in the house, which had not been observed to
be troubled; and to this place the children were removed; and were
always put to bed before daylight disappeared, but here they were no
sooner laid, than their disturber was at his work again with them.
On the fifth of November, 1661, a terrible noise was kept up; and
one of Mr. Monpesson’s servants observing two boards moving in the
children’s room, asked that one might be given to him; upon which a
board came (nothing moving it that he saw) within a yard of him; the
man said again, let me have it in my hand; when it was brought quite
close to him, and in this manner it was continued moving up and
down, to and fro, for at least twenty minutes together. Mr.
Monpesson, however, forbade his servant to take liberties with the
invisible and troublesome guest in future. This circumstance took
place in the day-time, and was witnessed by a whole room full of
people. The morning this occurred, it left a very offensive
sulphureous smell behind it. At night, the minister of the parish, one
Mr. Cragg, and several of the neighbours, paid Mr. M. a visit. The
minister prayed at the children’s bedside, when the demon was then
extremely troublesome and boisterous. During time of prayer it
retired into the cockloft, but as soon as prayers were over it returned;
when in the presence and sight of the company, the chairs began to
walk and strut about the room of their own accord, the children’s
shoes were thrown over their heads, and every thing loose moved
about the room. At the same time, a bedpost was thrown at the
minister, which struck him on the leg, but so gently that a lock of
wool could not have fallen more gently; and it was observed, that it
stopped just where it fell, without rolling or otherwise moving from
the place.
In consequence of the demon tormenting the children so
incessantly, he had them removed to a neighbour’s house, taking his
eldest daughter, who was about ten years of age, into his own
chamber, where it had not been for a month before; but, as soon as
she was in bed, the noise began there again, and the drumming
continued for three weeks with other noises; and if any particular
thing was called for to be beaten on the drum, it would perform it.
The children were brought home again, in consequence of the house
where they were placed being crowded with strangers. They were
now placed in the parlour, which, it was remarked, had hitherto not
been disturbed; but no sooner were they here, than their tormentor,
while they were in bed, amused himself with pulling their hair and
bedgowns, without offering any other violence.
It was remarked, that when the noise was loudest, and when it
came with the most sudden and surprising violence, no dog about
the house would move or bark, though the knocking and thumping
were often so boisterous and rude, that they were heard at a
considerable distance in the fields, and awakened the neighbours in
the village, some of whom lived very near this house. Not
unfrequently the servants would be lifted up, with their bed, to a
considerable height, and then let gently down again without harm; at
other times it would lie like a great weight upon their feet.
About the end of December, 1661, the drumming was less frequent,
but then a noise like the chinking of money was substituted for it,
occasioned, as it was thought, in consequence of something Mr.
Monpesson’s mother had said the day before to a neighbour, who
spoke about fairies leaving money behind them; viz. that she should
like it well, if it would leave them some to make them amends for the
trouble it had caused them. The following night, a great chinking and
jingling of money was heard all over the house. After this it left off its
ruder pranks, and amused itself in little apish and less troublesome
tricks. On Christmas morning, a little before daylight, one of the little
boys was hit, as he was getting out of bed, upon a sore place on his
heel, with the latch of the door, the pin of which, that fastened it to
the door, was so small, that it was a matter of no little difficulty for
any one else to pick it. The night after Christmas, it threw the old
gentlewoman’s clothes about the room, and hid her bible in the
ashes; with a number of other mischievous tricks of the same kind.
After this, it became very troublesome to one of Mr. Monpesson’s
servant men, a stout fellow, and of sober conversation. This man
slept in the house during the greater part of the disturbance; and for
several nights something would attempt to pull the bedclothes off
him, which he often, though not always, prevented by main force; his
shoes were frequently thrown at his head, and sometimes he would
find himself forcibly held, as it were, hand and feet; but he found that
when he could use a sword which he had by him, and struck with it,
the spirit let go his hold.
Some short time after these contests, a son of Mr. Thomas Bennet,
for whom the drummer had sometimes worked, came to the house,
and mentioned some words to Mr. Monpesson that the drummer
had spoken, which it seems were not well taken; for they were no
sooner in bed, than the drum began to beat in a most violent
manner: the gentleman got up and called his man, who was lying
with Mr. Monpesson’s servant just mentioned, whose name was
John. As soon as Mr. Bennet’s man was gone, John heard a rustling
noise in his chamber, as if a person in silks were moving up and
down; he immediately put out his hand for his sword, which he felt
was withheld by some one, and it was with difficulty and much
tugging, that he got it again into his possession, which he had no
sooner done, than the spectre left him; and it was always remarked it
avoided a sword. About the beginning of January, 1662, they used to
hear a singing in the chimney before it descended; and one night,
about this time, lights were seen in the house. One of them came into
Mr. Monpesson’s chamber, which appeared blue and glimmering,
and caused a great stiffness in the eyes of those who beheld it. After
the light disappeared, something was heard walking or creeping up
stairs, as if without shoes. The light was seen four or five times in the
children’s chamber; and the maids confidently affirm, that the doors
were at least ten times opened or shut in their presence; and that,
when they were opened, they heard a noise as if half a dozen had
entered together; some of which were afterwards heard to walk about
the room, and one rustled about as if it had been dressed in silk,
similar to that Mr. Monpesson himself heard.
 While the demon was in one of his knocking moods, and at a time
when many were present, a gentleman of the company said, “Satan,
if the drummer set thee to work, give three knocks and no more;”
which it did very distinctly, and stopped. The same gentleman then
knocked to hear if it would answer him as it was accustomed to do.
For further proof, he required it, if it actually were the drummer that
employed him as the agent of his malice, to give five knocks and no
more that night; which it did, and quietly left the house for the
remainder of the night. This was done in the presence of Sir Thomas
Chamberlaine of Oxfordshire, and many other creditable persons.
On Saturday morning, an hour before daylight, January 10, a drum
was heard beating upon the outside of Mr. Monpesson’s chamber,
from whence it went to the other end of the house, where some
gentlemen strangers lay, and commenced playing at their door four
or five different tunes; and at length flew off in the air. The next
night, a blacksmith in the village, and Mr. Monpesson’s man John,
who was lying with him, heard a noise in the room, as if somebody
were shoeing a horse; and something came with something like a
pair of pincers, and nipped at the blacksmith’s nose the whole of the
night.
Getting up one morning to go a journey, Mr. Monpesson heard a
great noise below, where the children lay; and on running down
instantly with a pistol in his hand, he heard a voice cry out, a witch! a
witch! similar to one they had heard on a former occasion. On his
entering the apartment, all became quiet again.
The demon having one night played some little pranks at the foot
of Mr. Monpesson’s bed, it went into another bed, where one of his
daughters lay, and passed from one side to the other, lifting her up as
it passed under her. At that time there were three kinds of noises in
the bed. They attempted to thrust at it with a sword, but it
continually evaded them. The following night it came panting like a
dog out of breath, when some one present took a bedpost to strike at
it, when it was immediately snatched out of her hand; and company
coming up stairs at the same time, the room was filled with a
nauseous stench, and very hot, although there was no fire on, and
during a very sharp winter’s night. It continued panting an hour and
a half, panting and scratching; and afterwards went into the
adjoining chamber, where it began to knock a little, and seemed to
rattle a chair; thus it continued for two or three nights in succession.
The old lady’s bible after this was found again among the ashes, with
the leaves downwards. It was taken up by Mr. Monpesson, who
observed that it lay open at the third chapter of St. Mark, where
mention is made of the unclean spirits falling down before our
Saviour, and of his giving power to the twelve Apostles to cast out
devils, and of the Scribes’ opinion, and that he cast them out through
Beelzebub.
The following morning ashes were scattered over the chamber
floor, to see what impressions would be left upon it; in the morning,
in one place they found the resemblance of a great claw in another
that of a smaller one, some letters in another, which could not be
decyphered, besides a number of circles and scratches in the ashes,
which no one understood except the demon itself.
About this time, the author of the narration went to the house to
enquire after the truth of the circumstances which made so much
noise in that part of the country. The demon had left off drumming,
and the terrible noises it was in the habit of making before he
arrived; but most of the remarkable facts already related, were
confirmed to him there by several of the neighbours, on whose
veracity he could depend, who had witnessed them. It now used to
haunt the children after they were gone to bed. On the night he was
there, the children went to bed about 8 o’clock; a maid servant
immediately came down and informed us that the spirit was come.
The neighbours then present went away, as well as two ministers
who had previously been some time in the house, but Mr.
Monpesson the author, and another gentleman who came with him,
went up to the room where the children were in bed. A scratching
was heard as they went up stairs, and just as they got into the room,
it was perceived just behind the bolster of the bed in which the
children lay, and appeared to be lying against the tick. The noise it
made was like that made with long nails upon the bolster. There were
two little girls, about seven or eight years of age, in the bed. Their
hands were outside the bedclothes, so that it was perfectly visible the
noise was not made by them which was behind their heads: they had
been so used to it of late, and always with some present in the
chamber, that they seemed to take very little notice of it. The
narrator, who was standing at the head of the bed, thrust his hand