Math 410/510 — Solved Problems

1. Prove Lemma 2 of Section 1: namely,

Lemma. Let A = A1 × · · · × An, B = B1 × · · · × Bn , and C = C1 × · · · × Cn.
Then:
(1) B ∩ C = (B1 ∩ C1 ) × · · · × (Bn ∩ Cn)
(2) If B and C are disjoint, then for some k, Bk and Ck are disjoint; moreover,
(3) if also A = B ∪ C and B, C are non-empty then, for the k of part (2),
Ak = Bk ∪ Ck , while for i 6= k, Ai = Bi = Ci.

(1) By definition of the sets involved and elementary rules of logic,

x ∈ B ∩ C ⇐⇒ x ∈ B and x ∈ C
⇐⇒ (∀i = 1, . . . , n, xi ∈ Bi ) and (∀i = 1, . . . , n, xi ∈ Ci )
⇐⇒ ∀i = 1, . . . , n, xi ∈ Bi ∩ Ci
⇐⇒ x ∈ (B1 ∩ C1 ) × . . . × (Bn ∩ Cn)

Thus, by the definition of set equality, B ∩ C = (B1 ∩ C1) × . . . × (Bn ∩ Cn ),

as required.
(2) If, Bk ∩ Ck 6= ∅, for all k = 1, . . . , n, then we may choose x1, . . . , xn such
that xk ∈ Bk ∩ Ck , for all k = 1, . . . , n. But then the point x = (x1, . . . , xn) ∈
(B1 ∩ C1) × . . . × (Bn ∩ Cn) = B ∩ C, so that B ∩ C 6= ∅.
Thus, if for all k = 1, . . ., n, Bk and Ck are not disjoint, B and C are not
disjoint. The contrapositive therefore must hold: If B and C are disjoint, there
exists k ∈ {1, . . ., n} such that Bk and Ck are disjoint.
(3) Let πi : Rn −→ R denote the projection onto the first coordinate; namely
πi(x) = xi. Recall that in general if W = W1 × . . . × Wn 6= ∅, for all t ∈ Wi ,
there exists x ∈ W with xi = t. (That is, πi maps W onto Wi ). Indeed, if
w = (w1, . . . , wn) is one point in W , then the point x obtained by changing the
ith coordinate to t, namely

wj , j 6= i
xj =
t, j=i

is still in W .
We are given that A = B ∪ C. Even without the assumption that B and C are
disjoint, we have

Ai = πi (A) = πi (B) ∪ πi(C) = Bi ∪ Ci

.
Now add the assumption that B and C are disjoint, and choose k such that
Bk ∩ Ck = ∅, by part (2) and let i 6= k. We are left to show that Ai ⊂ Bi and

14/12/2005 1273 mam

1
2 MATH 410/510 ––– SOLVED PROBLEMS

Ai ⊂ Ci . Let t ∈ Ai .
Choose any point b ∈ B. Then b ∈ A, and the point a obtained by changing bi
to t is still in A. However, a ∈
/ C, because ak = bk ∈
/ Ck . Therefore a ∈ B, so
t = ai ∈ Bi . Thus, Ai ⊂ Bi . Similary, Ai ⊂ Ci. 
2. A family P of sets is a prering iff it is closed under intersection and E, F ∈ P
implies E \ F is a the union of a finite disjoint family of members of P
If P1 and P2 are prerings, then so is {A1 × A2 : A1 ∈ P1 , A2 ∈ P2}

Soln. . We follow the example of intervals.

Let W = {A1 × A2 : A1 ∈ P1, A2 ∈ P2 }. If E = E1 × E2 and F = F1 × F2
belong to W, then
(i) E ∩ F = (E1 ∩ F1) × (E2 ∩ F2 ) ∈ W, since E1 ∩ F1 ∈ P1 , and F1 ∩ F2 ∈ P2 ,
by definition of prering.
(ii) Now,
(E1 × E2 ) \ (F1 × F2 ) = [(E1 \ F1) × E2 ] ∪ [(E1 ∩ F1 ) × (E2 \ F2)]
S
Since P1 is a prering, E1 \F1 = A∈H1 A where H1 is a finite disjoint subfamily
of P1. So [
(E1 \ F1) × E2 = A × E2
A∈H1
S
Also, E1 ∩ F1 ∈ P, and E2 \ F2 = B∈H2 B, where H2 is a finite disjoint
subfamily of P, and hence
[
(E1 ∩ F1) × (E2 \ F2) = (E1 ∩ F1 ) × B
B∈H2

Thus,
(E1 × E2) \ (F1 × F2)
is the union of 2 disjoint sets, each a finite disjoint union of elements of W, so
is one also.

3. Let P be a prering (that is, P is closed under intersection and A, B ∈ P implies

A \ B is the disjoint union of members of P). Prove that the set of finite unions
of sets in P is the smallest ring containing P.

Proof. Let P0 be the set of finite disjoint unions from P. Then,

(a) If A ∈ P0 , B ∈ P, then A \ B ∈ E0 .
(b) If A ∈ P0 , B ∈ Ps , then A \ B ∈ P0 .
(c) If A ∈ Ps , then A ∈ P0, so Ps = P0 .
S
Indeed, (a) If H is a finite disjoint family in P, A = H and B ∈ P, then
!
[ [
A\B = I \B = (I \ B).
I∈H I∈H

14/12/2005 1273 mam

 MATH 410/510 ––– SOLVED PROBLEMS 3

But I \ B ∈ P0 by definition of prering and the (I \ B) are disjoint, so A \ B is

in P0 .
(b) This follows from (a) by induction.
(c) If A = A1 ∪ . . . ∪ Am , where each Ai ∈ P, put B1 = A1 , B2 = A2 \ A1 ,
B3 = A3 \ (A1 ∪ A2 ), . . . , Bm = Am \ (A1S∪ . . . ∪ Am−1
Sm). Then the Bi are
m
disjoint, they belong to P0 by (b), and A = i=1 Ai = i=1 Bi , which belongs
to P0 . This shows Ps ⊂ P0, so they are equal.
Now Ps is clearly closed under finite union, and by (b), P0 = Ps is closed under
difference, so Ps is a ring. Since every element A of P is the union of the family
{A}, Ps contains P.
Finally, if R is a ring containing P, it must contain all finite unions from P, so
R ⊃ Ps . Thus, Ps is the smallest ring containing P. 

4. Let Q be the set of all rationals and IQ be its set of bounded intervals.
Let `(I) = b − a, if I ∈ IQ has endpoints a ≤ b.
Prove ` is not countably additive on IQ .
Why does the proof in the discussion on Lebesgue volume not work here to
produce countable additivity?

Soln. Consider
S A = [0, S 1] ∩ Q. Thus, A ∈ IQ . Since Q is countable, so is A.
Thus, A = a∈A a = a∈A [a, a], the union of countably many closed intervals.
For each a, `({a}) = `([a, a]) = a − a = 0, so
X
`({a}) = 0 6= 1 = `(A).
a∈A

Thus, ` is not countably additive on IQ . 

Just as with Lebesgue measure, ` is finitely additive on IQ , hence finitely
subadditive and finitely superadditive. As in that case, we can approximate
an arbitrary bounded interval outside with open ones and inside with closed
ones. But the non-degenerate closed intervals [a, b] ∩ Q are not compact, so we
cannot pass from countable covers to finite ones, and thus we are unable to
prove countable additivity.

5. The graph of any continuous function f : [a, b] → R has Lebesgue 2-dimensional

outer measure 0.

Proof. Since f is continuous on [a, b], it is uniformly continuous. That is, for
all ε > 0, there exists δ > 0 such that |f(x) − f(y)| < ε, whenever |x − y| < δ.
Divide [a, b] into intervals, [a0, a1], [a1, a2], . . . [ak−1, ak ] of length < δ. Then,
for each i = 1, . . ., k, x ∈ [ai, bi] implies f(x) ∈ [f(ai ) − ε, f(ai + ε], so the
graph of f, namely Gf = {(x, f(x)) : x ∈ [a, b]} is contained in

[
k
[ai−1, ai] × [f(ai ) − ε, f(ai ) + ε];
i=1

14/12/2005 1273 mam

4 MATH 410/510 ––– SOLVED PROBLEMS

hence,
k
X
λ∗ (Gf ) ≤ (ai − ai−1 )(2ε) = 2ε(b − a)
i=1

Since ε > 0 was arbitrary, λ∗ (Gf ) = 0.

6. Every translate of a Lebesgue-measurable set in Rn is measurable.

Proof. Let A be Lebesgue measurable, and x ∈ Rn . To prove x + A = {x + a :

a ∈ A} is also measurable, let T ⊂ Rn , arbitrary. Then, T ∩ (x + A) =
x + ((T − x) ∩ A). Since A is measurable, we obtain

λ∗ (T − x) = λ∗ ((T − x) ∩ A) + λ∗ ((T − x) ∩ Ac )

Since λ∗ is tranlation invariant this yields

λ∗ (T ) = λ∗ (x + (T − x) ∩ A) + λ∗ (x + ((T − x) ∩ Ac ))
= λ∗ (T ∩ (x + A)) + λ(T ∩ (x + A)c )

7. Every proper vector subspace of Rn has Lebesgue measure 0.

Proof. Let V be a proper subspace vector subspace of Rn. Let a be any non-
zero vector orthogonal to V . Then,

V ⊂ H = {x : a1x1 + a2x2 + · · · + an xn = 0},

a hyperplane in Rn . It will be enough to show that H has measure 0.

Without loss of generality we assume a1 is not 0.

S m ∈ N, let Im be the n-dimensional cube [−m, m] × . . . × [−m, m], so

For each
that m Im = Rn. Fix m ∈ N and consider the part of H in Im , A = H ∩ Im .
Let e1 be the unit vector (1, 0, . . ., 0). For any real t with 0 ≤ t ≤ 1, te1 + A ⊂
te1 + Im ⊂ [−m, m + 1] × [−m, m] × . . . × [−m, m], a bounded interval K.
Now, choose a sequence tk ∈ [0, 1] of distict
P numbers. Since a1 6= 0, the sets
tk e1 + A are disjoint subsets of K. thus k∈N λ(tk e1 + A) ≤ λ(K) < ∞. But,
since λ is translation invariant, all these tk e1 + A have the same measure as A,
so λ(A) = 0.
S
This proves that λ(H ∩ Im ) = 0, for all m. Since H = m (H ∩ Im , λ(H) is
also 0. 

14/12/2005 1273 mam

 MATH 410/510 ––– SOLVED PROBLEMS 5

8. Sets A and B in Rn are said to be separated if the closure of A does not

intersect B and the closure of B does not intersect A:

cl(A) ∩ B = A ∩ cl(B) = ∅.

Using the concept of measurability, show that if A and B are separated, then
λ∗ (A ∪ B) = λ∗ (A) + λ∗ (B).

Proof. The closure cl(A) of the set A is closed, hence Lebesgue measurable.
Use T = A ∪ B in the definition of measurable, to get

λ∗ (A ∪ B) = λ∗ ((A ∪ B) ∩ cl(A)) + λ∗ (A ∪ B) ∩ cl(A)c

Since A ⊂ cl(A) and cl(A) ∩ B = ∅, we obtain

λ∗ (A ∪ B) = λ∗ (cl(A)) + λ∗ (B),

as required. 

9. Prove that Lebesgue outer measure in Rn could have been defined using only
the intervals of the type (a, b] = (a1 , b1] × . . . (an, bn]. That is if Ir is the family
of all these intervals,
X [
λ∗ (A) = inf{ λ(I) : H ⊂ Ir , H is countable, and H ⊃ A},
I∈H

for all A ⊂ R.

Proof. Let λ(A) be the number given by this new formula. Since each cover by
open elements of Ir is a cover by elements of I, we have λ∗ (A) ≤ λ(A).
For the reverse, recall that for each bounded interval I, and each ε > 0, there is
an open interval U = (a, b) = (a1 , b1)×. . .×(an , bn) ⊃ I, with λ(U )−λ(I) < ε.
But the interval J = (a, b] = (a1 , b1] × . . . × (an , bn] ∈ Ir has the same measure
as U , so also λ(J) − λ(I) < ε.
So fix ε > 0. If H = {Ii : i ∈ N} is a countable I-cover of A, for each
i
iPthere existsPan open interval JP i with λ(Ji ) < λ(Ii ) + ε/2 . Thus λ(A) ≤
i
i λ(Ji ) ≤ i (λ(Ii ) + ε/2 ) = i λ(Ii ) + ε. Taking infimum over such H
gives λ(A) ≤ λ (A) + ε. Since ε is arbitrary, λ(A) ≤ λ∗ (A), so equality is
∗

obtained. 

14/12/2005 1273 mam

6 MATH 410/510 ––– SOLVED PROBLEMS

10. Let Ic be the family of bounded closed intervals of R.

Let α be the real valued function defined on R by

1 , if x < 0
α(x) = .
2 , if x ≥ 0

Let
τ ([a, b]) = α(b) − α(a).
Let τ ∗ be the Caratheodory outer measure generated by τ and Ic.
Find τ ∗ (A), for all A ⊂ R.

Soln. The entire real line R can be written as

[ [
R= [−n, −1/n] ∪ [0, n].
n∈N n∈N

Since, for a ≤ b < 0 we have τ ([a, b]) = α(b) − α(a) = 1 − 1 = 0, and for
0 ≤ a ≤ b, we also have τ ([a, b]) = 2 − 2 = 0, we see that, for all A ⊂ R,
X X
0 ≤ τ ∗ (A) ≤ τ ([−n, −1/n]) + τ ([0, n]) = 0.
n∈N n∈N

Thus, τ (A) = 0, for all A ⊂ R.

11. If µ is a countably additive extended real-valued function on a ring R, then for
a sequence (Ak ) in R, if Ak & A ∈ R, then µ(Ak ) −→ µ(A), provided some
µ(Ak ) is finite.

Proof. We may assume µ(A1 ) < ∞. For each k, let Bk = A1 \ Ak .

Since AS
k ⊃ Ak+1 , for T
all k, we have Bk = A1 \ Ak ⊂ A1 \ Ak+1 = Bk+1 , for all
k, and k Bk = A1 \ k Ak = A1 \ A. Thus,

µ(A1 ) − µ(Ak ) = µ(Bk ) −→ µ(A1 \ A) = µ(A1 ) − µ(A)

Hence, µ(Ak ) −→ µ(A). (The subtractions were possible by the finiteness

assumpion.) 

14/12/2005 1273 mam

 MATH 410/510 ––– SOLVED PROBLEMS 7

12. If C is a family of sets, prove that every element of σ(C)) is covered by a

countable subfamily of C.

Proof. Let W be the family of all sets that can be covered by a countable
subfamily of C. If A, B ∈ W, then A is covered by a countable family H ⊂ C,
so A \ B is covered by the same family. Hence A \ B ∈ W.

S i ∈ N, and Hi is a corresponding countable subfamily which

If Ai ∈ W, for all
covers Ai , then i Hi is a countable subfamily of C and
[ [ [
Ai ⊂ C,
i i C∈Hi

S
so i Ai ∈ W.
Thus W is closed under countable union and set difference, so it is a σ-ring. 

13. B(Rn ) is the σ-ring generated by the family of compact sets in Rn.

Proof. There are many possible solutions.

Let K be the family of compact sets of Rn and G, the family of open sets.
Since B(Rn ) contains all the open sets, it contains the complements of these,
the closed sets. But each compact set is closed, so

B(Rn ) ⊃ K and hence B(Rn ) ⊃ σ(K)).

Now every open set G is the union of countably many compact intervals of the
form [a, b] = [a1, b1] × . . . × [an, bn], where the ai and bi are rational. Thus,
G ∈ σ(K). Thus,

σ(K) ⊃ G and hence σ(K) ⊃ G = B(Rn ).

Thus, B(Rn ) = σ(K). 

14. Let A be a set in Rn of finite outer Lebesgue measure such that

λ∗ (A) = sup{λ(K) : K is compact, K ⊂ A}.

Prove that A is λ-measurable.

Proof. For each n ∈ N, λ∗ (A) − 1/n < sup{λ(K) : K is compact, K ⊂ A},

hence we may choose a compact set Kn ⊂ A with λ(Kn ) > λ∗ (A) − 1/n.
S
Let K = n Kn . Then, K ⊂ A and, K is a Borel set, hence λ∗ -measurable.
Therefore,

λ∗ (A \ K) = λ∗ (A) − λ∗ (A ∩ K) = λ∗ (A) − λ(K) < 1/n.

14/12/2005 1273 mam

8 MATH 410/510 ––– SOLVED PROBLEMS

(The subtraction is possible, since the outer measures involved are finite.)
Since this is true for all n, λ∗ (A \ K) = 0.
Since sets of outer measure 0 are always measurable, and since K is measurable,

A = K ∪ (A \ K)

is measurable.
15. Prove that every D-system closed under finite intersection is a σ-ring.
Let D be a D-system closed under finite intersection. If A, B ∈ D, then A∩B ∈
D, by hypothesis. Thus, A \ B = A \ A ∩ B ∈ D, since D is closed under proper
differences.
S
Now, suppose (Ai ) is a sequence in D. Put Bi = Ai \ j<i. Then the Bi are
T S S
disjoint and Bi = j<i (Ai \ Aj ) ∈ D. Thus, i Ai = i Bi ∈ D, since D is
closed under countable disjoint unions.
16. Let µ(B) = λ2 ({x ∈ R × [0, 1] : x1 + x2 ∈ B}), for each B ∈ B(R).
a) Prove that µ is a measure (that is, a non-negative countably additive func-
tion) on B(R).

Proof. Let T : R2 −→ R be defined by T (x1 , x2) = (x1 +x2). Then T is a linear

transformation, hence continuous. Thus, for each Borel set B ⊂ R, T −1(B) is
a Borel subset of R2 . Let B = B(R). Let X = {x ∈ R × [0, 1] : x1 + x2 ∈ [0, 1]})
For each B ∈ B, we see that

µ(B) = λ2 (X ∩ T −1(B)).

Now, for each B ∈ B, λ(T −1 (B)) ≥ 0, and if (Bn ) is a disjoint sequence in B,

then (T −1 (Bn )) is a disjoint sequence in B(R2 ). Indeed, if i 6= j,

T −1 (Bi ) ∩ T −1(Bj ) = T −1 (Bi ∩ Bj ) = T −1(∅) = ∅.

Moreover, !
[ [
−1
T Bn = T −1 (Bn ),
n∈N n∈N

hence,
!!
[ [ X X
µ( Bn ) = λ2 X ∩ T −1 Bn = λ(X ∩ T −1(Bn )) = µ(Bn ).
n n∈N n∈N n

This shows µ is a measure

14/12/2005 1273 mam

 MATH 410/510 ––– SOLVED PROBLEMS 9

b) Show that µ is a constant multiple of λ1 (on B(R)).

We need only show µ is translation invariant and has a finite value on [0, 1].
For the translation invariance,

µ(B + a) = λ2 ({x ∈ R × [0, 1] : x1 + x2 ∈ (B + a)})

= λ2 ({x ∈ R × [0, 1] : x1 + x2 ∈ B} + (a, 0))
= λ2 ({x ∈ R × [0, 1] : x1 + x2 ∈ B}) by translation invariance of λ2
= µ(B).

Now, X ∩ T −1 ([0, 1]) = {(x1, x2) : x1 + x2 ∈ [0, 1], x2 ∈ [0, 1]} ⊂ [−1, 1] × [0, 1],
so µ([0, 1]) ≤ λ2 ([−1, 1] × [0, 1] < ∞.
Thus, µ is a constant multiple of Lebesgue measure.
Although the question did not ask it, we can do better:
X ∩ T −1 ([0, 1]) = {(x1, x2) : x1 + x2 ∈ [0, 1], x2 ∈ [0, 1]}, and the linear
transformation with matrix  
1 −1
0 1
has determinant 1 and maps [0, 1] × [0, 1] onto this set, so

µ([0, 1]) = λ2 ({(x1, x2) : x1 +x2 ∈ [0, 1], x2 ∈ [0, 1]} = λ2 ([0, 1]×[0, 1]) = 1 = λ1([0, 1]).

Thus, µ actually is exactly Lebesgue measure on the Borel sets of [0, 1].
17. Let P be set of all intervals in Q of the form (a, b), with a < b, a, b ∈ Q. Show
that σ(P) is the set of all subsets of Q. Let µ(B) =the number of points in
B, for all B ∈ σ(P), and let ν(B) = 2µ(B), for all such B. Show that µ = ν
on P, but not on σ(P). Why does this not contradict the Unique Extension
Theorem?

Soln.
For each rational a, {a} = (a − 1, a + 1) \ [(a − 1, a) ∪ (a, a + 1)]. Thus, {a} ∈ P.
But every subset B of Q is the union
[
{a}
a∈B

of countably many such singletons, so B ∈ σ(P). Thus, P = {B : B ⊂ Q}.

Since the number of points in each interval (a, b) of Q is infinite, µ(a, b) =
∞ = 2µ(a, b) = ν(a, b), but for a finite set of m > 0 elements µ(B) = m and
ν(B) = 2m 6= m.
Even though P is closed under finite intersection, this does not contradict the
Unique Extension Theorem, because µ and ν are not finite on P.

14/12/2005 1273 mam

10 MATH 410/510 ––– SOLVED PROBLEMS

18. By using the theorem on change of measure under a linear transformation, find
the area of a a triangle determined by 3 points a, b, c of R2.
Let A be the triangle determined by a, b, c. Then, A − c is the triangle deter- 
a − c1 b1 − c1
mined by 0, a−c, b−c. The linear transformation T with matrix 1
a2 − c2 b2 − c2
maps the triangle B = {x : x1 ≥ 0, x2 ≥ 0, x1 + x2 ≤ 1} onto A − c, so
  
 a − c1 b1 − c1 
λ(A) = λ(A − c) = λ(T (B)) = | det(T )|λ(B) = (1/2) det 1 .
a2 − c2 b2 − c2 

By the way,
 
  a1 b1 c1
a − c1 b1 − c1
det 1 = det  a2 b2 c2  ,
a2 − c2 b2 − c2
1 1 1

which is one way the formula is often seen.

Here we used the fact that the triangle B has measure 1/2. To prove this, let
R be the linear
 transformation
 which rotates all vectors through an angle π. It
−1 0
has matrix .
0 −1
Then R(B) is the triangle {x = (x1, x2) : x1 ≤ 0, x2 ≤ 0, x1 + x2 ≥ −1} and
R(B) + (1, 1) is the triangle C = {x : x1 ≤ 1, x2 ≤ 1, x1 + x2 ≥ 1}.
Thus, B ∪ C = {x : 0 ≤ x1 ≤ 1, 0 ≤ x2 ≤ 1}, the closed unit cube K =
[0, 1] × [0, 1].
Now,
λ(C) = λ(R(B)) = | det(R)|λ(B) = λ(B).
The set B ∩ C is the line segment √
from (0, 1) to (1, 0). It has measure 0, since
it may be obtained from {0} × [0, 2], by rotation and translation. Thus,

1 = λ(K) = λ(B ∪ C) = λ(B) + λ(C) − λ(B ∩ C) = λ(B) + λ(B) = 2λ(B).

Hence λ(B) = 1/2.

19. The family B(Rn ) of Borel sets of Rn is the smallest family containing the open
sets which is closed under countable unions and countable intersections.

Soln. Actually the result is true in a general metric space, so we will write the
solution that way.
Let G be the family of open sets in the metric space E.
Since B(E) contains G and is closed under countable union and intersection,
we need only show that it is the smallest such family.
Let C be a family of sets closed under countable union and countable intersec-
tion such that C ⊃ G.
Let W = {A : A ∈ C and Ac ∈ C}.
Then, W is a σ-algebra.

14/12/2005 1273 mam

 MATH 410/510 ––– SOLVED PROBLEMS 11

Indeed, if A ∈ W, then Ac ∈ W, so W is closed under

S complementation.
If H is a countable subfamily of W, then H ⊂ C, so A∈H A ∈ C and for each
S
c T
A ∈ H, Ac ∈ C, so A∈H A = A∈H Ac ∈ C.
Thus, all we need to prove is that W ⊃ G, for then W ⊃ σ(G) = B (E), so
C ⊃ B(E).
If G is open, then G ∈ C, by hypothesis.
Moreover, Gc is closed, so will belong to C, if we can show that each closed set
F in a metric space is the intersection of a countable family of open sets. To
see this, define the distance from a point x to F by

d(x, F ) = inf{d(x, y) : y ∈ F }.

For each k ∈ N, put

[
Uk = {x : d(x, F ) < 1/k} = B(y, 1/k).
y∈F

T
Then each Uk is open and k∈N Uk = {x : d(x, F ) = 0} = F , since F is closed.

14/12/2005 1273 mam

12 MATH 410/510 ––– SOLVED PROBLEMS

20. Let f be a function on S to E. If E is a σ-ring in E, then {f −1 [B] : B ∈ E} is

a σ-ring in S.

Proof. Let S = {f −1 [B] : B ∈ E}. If A1 , A2, . . . ∈ S, then Ai = f −1 [Bi ], where

Bi ∈ E, for each i ∈ N. Thus, AS1 \ A2 = f −1
S [B−1 1] \ f
−1
[B2 ] = fS−1 [B1 \ B2 ] ∈ S,
since B1 \ B2 ∈ E. Moreover, i∈N Ai = i f (Bi ) = f −1 ( i Bi ) ∈ S, since
S
i Bi ∈ E. Thus, S is a σ-ring, as required. 

21. Recall: Let f : S −→ Rn; f = (f1 , . . . , fn). Then f is measurable iff each fi is
measurable.
Use this to prove that B(Rn ) = σ({B1 × . . . × Bn : Bi ∈ B(R), for i =
1, . . ., n}).

Proof.
Let C = {B1 × . . . × Bn : Bi ∈ B, for i = 1, . . ., n}). Let B = B(R). The
σ-ring σ(C) is called the product σ-ring of B with itself n-times, often
denoted B n .
Let G be the family of open sets of Rn. Take f to be the identity map on
Rn. Its coordinate maps are the projection maps: fi (x) = πi(x) = xi , for each
i. For each D ∈ B, π1−1 (D) = B1 × . . . × Bn , where Bj = R, for j 6= i, and
Bi = D, so πi is σ(C)-measurable. Hence, the identity map f is σ(C)-B(Rn )
measurable. Hence, for each B ∈ B(Rn ), B = f −1 (B) ∈ σ(C). Thus,

B(Rn ) ⊂ σ(C).

Now, the identity map is certainly B(Rn ) measurable, so each coordinate map
π
Ti is Borel measurable. Therefore, if Bi ∈ B,n for each i, then B1 × . . . × Bn =
i (πi ∈ Bi ) ∈ B(Rn), which shows C ⊂ B(R ). Thus,

σ(C) ⊂ B(Rn ),

as required to show equality. 

14/12/2005 1273 mam

 MATH 410/510 ––– SOLVED PROBLEMS 13

22. Let R be a σ-ring in a space S. Let f be an R measurable function on S to

R. Then, there exists a sequence (fn ) of R measurable functions such that for
each n, fn has countably many values and fn → f uniformly on S.

Proof. There is no need here to consider the non-negative case separately, and
in fact there is an advantage to not doing so. for, we can then also arrange that
(fn ) increases to f.
Let f be as in the hypothesis and for each n ∈ N, put

X
∞
k
fn = 1[k/2n <f ≤(k+1)/2n] .
2n
k=−∞

For each n, there exists a k ∈ Z such that k/2n < f(x) ≤ (k + 1)/2n and then
fn (x) = 2kn < f(x) and |f(x) − fn (x)| ≤ 21n −→ 0, so fn → f uniformly.
We also obtain fn ≤ fn+1 , for all n. Indeed, for each n, with k such that
k/2n < f(x) ≤ (k + 1)/2n. Then there are 2 possibilities: Either,
k 1 2k+1 k
(i) f(x) ≤ 2n + 2n+1 = 2n+1 , in which case, fn+1 (x) = fn (x) = 2n , or
2k+1 2k+1+1 k+1
(ii) 2n+1 < f(x) ≤ 2n+1 = 2n , in which case fn+1 (x) = fn (x)+1/2n+1 > fn (x).
Thus, in all cases fn (x) ≤ fn+1 (x), as promised. 

14/12/2005 1273 mam