You are on page 1of 4

Homework 4

Real Analysis
Joshua Ruiter
March 23, 2018

Lemma 0.1 (for Exercise 21). Let f : X → Y be a surjective mapping, and let B ⊂ Y .
Then f (f −1 (B)) = B.
Proof. Let b ∈ B. Then there exists a ∈ f −1 (B) such that f (a) = b. Thus b ∈ f (f −1 (B)).
Hence B ⊂ f (f −1 (B)).
Let b ∈ f (f (−1 B)). Then there exists a ∈ f −1 (B) such that f (a) = b. Since a ∈ f −1 (B),
f (a) = b ∈ B. Hence f (f −1 (B)) ⊂ B.
We have two way containment of sets, hence f (f −1 (B)) = B.
Proposition 0.2 (Exercise 21). There exists a continuous function that maps a Lebesgue
measurable set to a non-measurable set.
Proof. Let C be the middle-thirds Cantor set and let F : C → [0, 1] be the Cantor-Lebesgue
function defined in Exercise 2. We showed that F is continuous and surjective in part (b).
Let N be the non-measurable subset of [0, 1] constructed in Theorem 3.6. We claim that the
preimage of N under F is measurable. We know that F −1 (N ) ⊂ C, so by monotonicity of
outer measure,
m∗ (F −1 (N )) ≤ m∗ (C) = m(C) = 0
thus F −1 (N ) is measurable (and has measure zero) by property 2 of Lebesgue measure. Since
F is surjective, by the above lemma, F (F −1 (N )) = N . Thus the continuous map F maps
the Lebesgue measurable set F −1 (N ) to the non-measurable set N .
Proposition 0.3 (Exercise 22). There is no function f : R → R such that f is continuous
everywhere and f (x) = χ[0,1] (x) almost everywhere.
Proof. Because f (x) = χ[0,1] (x) almost everywhere, f (x) = 0 on (−∞, 0) except on a set of
measure zero. Then there is a sequence xn ∈ (−∞, 0) and xn → x = 0 such that f (xn ) = 0.
Because f is continuous,
lim f (xn ) = f (x) =⇒ lim 0 = 0 = f (0)
n→∞ n→∞

so f (0) = 0. Since f = χ[0,1] on [0, 1] except on a set of measure zero, there is a sequence yn
with yn ∈ [0, 1] and yn → y = 0, and f (yn ) = 1. Then because f is continuous,
lim f (yn )f (y) =⇒ lim 1 = 1 = f (0)
n→∞ n→∞

so f (0) = 1. This is a contradiction, so we conclude that no such continuous f exists.

1
Proposition 0.4 (Exercise 24). There exists an enumeration {rn }∞
n=1 of the rationals so
that the complement of
∞  
[ 1 1
rn − , rn +
n=1
n n
in R is nonempty.
Proof. Let S = {n2 : n ∈ N} = {1, 4, 9, . . .}, let T = N \ S. Both S and T are countable,
so we can find a bijection T ↔ Q ∩ [0, 1] and a bijection S ↔ Q \ [0, 1]. Together, these
bijections give an enumeration {rn } of Q, such that for rn ∈ Q \ [0, 1], we have n = kn2 for
some kn ∈ N. We can the write our union as
[ 1 1
 [
1 1

rn − , rn + ∪ rn − 2 , rn + 2
n∈S
n n n∈T
kn kn

Consider the measures of these unions. For n ∈ S, rn ∈ [0, 1] and 1/n ≤ 1 so each interval
(rn − 1/n, rn + 1/n) is contained in [−1, 2], so their union is contained in [−1, 2]. That is,
[ 1 1

rn − , rn + ⊂ [−1, 2]
n∈S
n n
and therefore by monotonicity of outer measure,
[ !
1 1
m∗ rn − , rn + ≤3
n∈S
n n
Now consider the union over T . By monotonicity of outer measure,
[ ! X ∞
1 1 2 X 2
m∗ rn − 2 , rn + 2 ≤ 2
=
n∈T
kn kn k
n∈T n k=1
k2
2
This infinite sum converges. (In particular, it has the value π3 , by the well-known argument
2
of Euler.) Hence the outer measure of these two unions is bounded above by π3 + 3. Because
it is finite, it cannot be the case that
∞  
[ 1 1
rn − , rn + =R
n=1
n n
Thus, this set’s complement in R is nonempty.
Proposition 0.5 (Exercise 25). A set E ⊂ Rd is measurable if and only if for every  > 0
there exists a closed set F with F ⊂ E and m∗ (E \ F ) < .
Proof. If E is measurable, and  > 0, by Theorem 3.4 there exists a closed set F with F ⊂ E
and m(E \ F ) ≤ /2 < , so m∗ (E \ F ) < .
Now suppose that E ⊂ Rd such that for every  > 0, there exists a closed set F with
F ⊂ E and m∗ (E \ F ) < . Let  > 0 and let F be as described. Then R \ F is open,
R \ E ⊂ R \ F , and (R \ F ) \ (R \ E) = E \ F . So then m∗ ((R \ F ) \ (R \ E)) = m∗ (E \ F ) < ,
so by the definition of measurable, R \ E is measurable. Then E is measurable (property 5
of measure).

2
Proposition 0.6 (Exercise 26). Suppose A ⊂ E ⊂ B where A, B are measurable subsets of
finite measure, and m(A) = m(B). Then E is measurable, and thus m(E) = m(A) = m(B).
Proof. Since A, B are measurable, B \ A is measurable, and everything here has finite mea-
sure, by additivity we have
m(A) + m(B \ A) = m(B)
so m(B \A) = m(B)−m(A) = 0. Since E ⊂ B, we have (E \A) ⊂ (B \A) so by monotonicity
of outer measure,
m∗ (E \ A) ≤ m∗ (B \ A) = m(B \ A) = 0
hence E \ A is measurable and m(E \ A) = 0 (property 2 of Lebesgue measure).
Because A is measurable, it differs from a Gδ set by a set of measure zero, that is, there
exists a Gδ set, call it G, such that G ⊂ A and m(A \ G) = 0 (by Corollary 3.5). We can
write A as A = G ∪ (A \ G). Now we write E as

E = A ∪ (E \ A) = G ∪ (A \ G) ∪ (E \ A) = G ∪ (A \ G) ∪ E(\A)
Since A \ G, E \ A are measurable and have measure zero, their union is a set of measure
zero. In particular, we have that E differs from the Gδ set G by a set of measure zero. Hence
E is measurable (using Corollary 3.5). Then m(E) = m(A) = m(B) follows immediately
from monotonicity of m.
Proposition 0.7 (Exercise 28). Let E be a subset of R with m∗ (E) > 0. For every α ∈ (0, 1),
there exists an open interval I so that
m∗ (E ∩ I) ≥ αm∗ (I)
Proof. Let α ∈ (0, 1). By observation 3 regarding exterior measure, m∗ (E) = inf{m∗ (O)}
where O is an open set containing E. Then for  > 0 there exists an open set O such that
E ⊂ O and
m∗ (E) ≥ m∗ (O) +  > m∗ (O) > αm∗ (O)
UsingSTheorem 1.3,S O can be written uniquely as a disjoint countable union of open intervals,
O= ∞ (a ,
i=1 i i b ) = ∞
i=1 Ii . We claim that for some Ii , we have

m∗ (E ∩ Ii ) ≥ αm∗ (Ii )
Suppose the reverse (strict) inequality holds for all i. Then

X ∞
X
m∗ (E ∩ Ii ) < αm∗ (Ii ) =⇒ m∗ (E ∩ Ii ) ≤ αm∗ (Ii )
i=1 i=1
S S
We know that E ⊂ O = i Ii , so E = ∩ Ii ), so by monotonicity,
i (E


! ∞
[ X
m∗ (E) = m∗ (E ∩ Ii ) ≤ m∗ (E ∩ Ii )
i=1 i=1

3
Since each Ii is measurable,

! ∞ ∞
[ X X
αm∗ (O) = αm(O) = αm Ii =α m(Ii ) = αm∗ (Ii )
i=1 i=1 i=1

Combining these inequalities, we get



X ∞
X
m∗ (E) ≤ m∗ (E ∩ Ii ) ≤ αm∗ (Ii ) = αm∗ (O)
i=1 i=1

That is, m∗ (E) ≤ αm∗ (O). But this is a direct contradiction with the inequality we already
established, m∗ (E) > αm∗ (O). So we reject our supposition that this property fails for every
Ii . Hence there exists an open interval Ii such that

m∗ (E ∩ Ii ) ≥ αm∗ (Ii )

You might also like