Sei sulla pagina 1di 7

REAL ANALYSIS I HOMEWORK 1

CİHAN BAHRAN

The questions are from Tao’s text.


Exercise
X 0.0.1. If (xα )α∈A is a collection of numbers xα ∈ [0, +∞] such that
xα < ∞, show that xα = 0 for all but at most countably many α ∈ A, even if
α∈A
A itself is uncountable.

We want to show that the subset A0 = {α ∈ A : xα > 0} of A is countable if the sum


is finite. For every n ∈ N, let

An = {α ∈ A : xα > 1/n} .
[
Then clearly A0 = An . Now arguing contrapositively, suppose A0 is uncountable.
n∈N
Then at least one of the An ’s should be uncountable, as countable unions of countable
sets is countable1. Fix such an n. Then for every M , An has a finite subset FM of size
M ; hence
X X X X
xα ≥ xα ≥ xα > 1/n = M/n
α∈A α∈An α∈FM α∈FM

for every M (the first two inequalities hold since xα ’s are nonnegative). But by varying
M
X
we can make M/n arbitrarily large (note that n does not depend on M ). Hence
xα can’t be finite.
α∈A

Exercise 0.0.2. (Tonelli’s theorem for series over arbitrary sets). Let A, B be sets
(possibly infinite or uncountable), and (xn,m )n∈A,m∈B be a doubly infinite sequence
of extended non-negative reals xn,m ∈ [0, +∞] indexed by A and B. Show that
X X X X X
xn,m = xn,m = xn,m .
(n,m)∈A×B n∈A m∈B m∈B n∈A

(Hint: althoughX not strictly necessary, you may find it convenient to first establish
the fact that if xn is finite, then xn is non-zero for at most countably many n.)
n∈A

First, assume that the left hand side is infinite. Then for every positive real number R,
there is a finite subset F ⊆ A × B such that
X
xn,m > R .
(n,m)∈F

1Note that the proof of this innocent assertion uses the axiom of (countable) choice! This fact is usually
swept under the rug.
1
REAL ANALYSIS I HOMEWORK 1 2

Note that A has a finite subset G and B has a finite subset H such that F ⊆ G × H.
Now observe that
X X X X
R< xn,m ≤ xn,m = xn,m .
(n,m)∈F (n,m)∈G×H n∈G m∈H

Note that, for fixed n ∈ G, the sum of the collection (xn,m )m∈B is defined as
X
sup xn,m .
H 0 ⊆B is finite m∈H 0

Thus for every n ∈ G we have


X X
xn,m ≤ xn,m
m∈H m∈B

and hence
X X X X
R< xn,m ≤ xn,m
n∈G m∈B n∈A m∈B

Since R > 0 was arbitrary, we conclude that the middle term in the asserted equality
is also infinite. In a similar fashion, the right hand side is infinite.
We handled the infinite case. So now assume that the left hand side is finite. Then by
Exercise 0.0.1, the set
S = {(n, m) ∈ A × B : xn,m > 0}
is countable. Let πA : A × B → A and πB : A × B → B to be the projection maps.
Write A0 = πA (S) and B0 = πB (S) and note that A0 ⊆ A and B0 ⊆ B are countable
sets. Since S ⊆ A0 × B0 and xn,m = 0 if (n, m) ∈
/ S, we get
X X X
xn,m = xn,m = xn,m
(n,m)∈A×B (n,m)∈S (n,m)∈A0 ×B0
X X
= xn,m
n∈A0 m∈B0

where we X
used Tonelli’s
X
theorem for series in the last equality. Since xn,m = 0 if m ∈
/ B,
we have xn,m = xn,m for every n ∈ A0 . And A0 has a similar property, so
m∈B0 m∈B
X X X X X X
xn,m = xn,m = xn,m .
n∈A0 m∈B0 n∈A0 m∈B n∈A m∈B

This establishes the first equality in the finite sum case. The second equality is similar.

Exercise 1.2.4. Let E, F ⊆ Rd be disjoint closed sets, with at least one of E, F


being compact. Show that dist(E, F ) > 0. Give a counterexample to show that
this claim fails when the compactness hypothesis is dropped.

The statement to prove is actually true in the setting of an arbitrary metric space.
Lemma 1. Let (X, d) be a metric space and A ⊆ X. Then A = {x ∈ X : dist(x, A) =
0}.
REAL ANALYSIS I HOMEWORK 1 3

Proof. Suppose x ∈ X such that dist(x, A) = 0. Then for every  > 0, there exists a ∈ A
such that d(x, a) < , hence B(x, ) ∩ A 6= ∅. Since the collection {B(x, ) :  > 0}
forms a basis of open neighborhoods around x, we conclude that x ∈ A. Conversely
if x ∈ A, then for every  > 0 we have B(x, ) ∩ A 6= ∅. Hence there exists a ∈ A (a
depends on ) such that d(x, a) < . Thus by the definition of infimum, dist(x, A) < .
As  > 0 was arbitrary, we conclude that dist(x, A) = 0. 
Proposition 2. Let (X, d) be a metric space and E, F ⊆ X be disjoint sets such that
E is compact and F is closed. Then dist(E, F ) > 0.

Proof. For every x ∈ E, since x ∈/ F = F , by Lemma 1 we have dist(x, F ) > 0. In


other words, the continuous function
f :E→R
x 7→ dist(x, F )
satisfies f (E) ⊆ (0, +∞). But E is compact, so f attains a minimum, say  > 0. Hence
for every x ∈ E, dist(x, F ) ≥ . Therefore dist(E, F ) ≥  > 0. 

For a counter-example in the non-compact situation, let E = N and F = {n + 1/n : n ∈


N} as subsets of R. Note that for every a, b ∈ F |a − b| > 1/2. So the only convergent
sequences in F are eventually constant ones; thus F is closed. Similarly E is closed.
However, for every n ∈ N we have n ∈ E and n + 1/n ∈ F and hence
1/n = |n − (n + 1/n)| = d(n, n + 1/n) ≥ dist(E, F ) .
Thus dist(E, F ) = 0.

Exercise 1.2.6 Give an example to show that the reverse statement


m∗ (E) = sup m∗ (U )
U ⊆E,U open

is false.

Note that For every open set U contained in E, we have m∗ (U ) ≤ m∗ (E), that is,
m∗ (E) is an upper bound for the set
SE = {m∗ (U ) : E ⊇ U -open }
of real numbers. Thus by the definition of sup we obtain
sup SE ≤ m∗ (E) .
So a counterexample to the claim must yield sup SE < m∗ (E).
Consider the Lebesgue outer measure for R. Let E = R − Q. Since every interval
contains a rational number, the only open set contained in E is the empty set! So we
have SE = {0} whose supremum is (surprisingly) 0. However by subadditivity we have
∞ = m∗ (R) ≤ m∗ (Q) + m∗ (E) = m∗ (E)
which gives m∗ (E) = ∞.

Exercise 1.2.7 (Criteria for measurability) Let E ⊆ Rd . Show that the following
are equivalent:
(i) E is Lebesgue measurable.
REAL ANALYSIS I HOMEWORK 1 4

(ii) (Outer approximation by open) For every ε > 0, one can contain E in an
open set U with m∗ (U r E) ≤ ε.
(iii) (Almost open) For every ε > 0, one can find an open set U such that
m∗ (U 4E) ≤ ε.
(iv) (Inner approximation by closed) For every ε > 0, one can find a closed set
F contained in E with m∗ (E r F ) ≤ ε.
(v) (Almost closed) For every ε > 0, one can find a closed set F such that
m∗ (F 4E) ≤ ε.
(vi) (Almost measurable) For every ε > 0, one can find a Lebesgue measurable
set Eε such that m∗ (Eε 4E) ≤ ε.
(ii) was our definition of measurability, so (i) and (ii) are the same statement in our
setup.
(ii) ⇒ (iii) is trivial because if U contains E then U 4E = U r E.
For (iii) ⇒ (ii), let ε > 0. Then there exists an open set V such that m∗ (V 4E) ≤ ε/3.
In particular m∗ (E r V ) ≤ ε/3 < 2ε/3 by monotonicity. Then by the outer regularity
of m∗ , there exists an open set W containing E r U such that m∗ (W ) < 2ε/3. So
U = V ∪ W is an open set containing E and
m∗ (U r E) = m∗ ((V r E) ∪ (W r E))
≤ m∗ (V r E) + m∗ (W r E)
≤ m∗ (V 4E) + m∗ (W )
≤ ε/3 + 2ε/3 = ε .
For (i) ⇒ (iv), let ε > 0. Since E is measurable, so is E c . Therefore by assumption
there exists an open set U containing E c such that m∗ (U r E c ) ≤ ε. Note that F := U c
is contained in E and
m∗ (E r F ) = m∗ (U ∩ E) = m∗ (U r E c ) ≤ ε .
(iv) ⇒ (v) is trivial, similar to (ii) ⇒ (iii).
(v) ⇒ (vi) holds because closed sets are measurable.
For (vi) ⇒ (iii), let ε > 0. Then there exists a measurable set D such that m∗ (D4E) ≤
/2. And there exists an open set U containing D such that m∗ (U r D) ≤ /2. Now
U 4E = (U r E) ∪ (E r U )
= (D r E) ∪ ((U r D)r E) ∪ (E r U )
⊆ (D r E) ∪ (U r D) ∪ (E r D) = (D4E) ∪ (U r D)
and hence by subadditivity and monotonicity we have
m∗ (U 4E) ≤ m∗ (D4E) + m∗ (U r D) ≤  .

Exercise 1.2.9 (Middle thirds Cantor set). Let I0 := [0, 1] be the unit interval, let
I1 = [0, 1/3] ∪ [2/3, 1] be I0 with the interior of the middle third interval removed,
let I2 := [0, 1/9]∪[2/9, 1/3]∪[2/3, 7/9]∪[8/9, 1] be I1 with the interior of the middle
third of each of the two intervals of I1 removed, and so forth. More formally, write
" n n
#
[ X ai X ai 1
In := , + n .
a1 ,...,an ∈{0,2} i=1 3i i=1 3i 3
REAL ANALYSIS I HOMEWORK 1 5

Let C := ∞n=1 In be the intersection of all the elementary sets In . Show that C is
T

compact, uncountable, and a null set.


By definition, each In is the union of finitely many closed intervals, so is closed. And
since for any choice of a1 , . . . an ∈ {0, 2} we have
n n n Ç åi Ç å
X ai X 2 X 1 1 − (1/3)n+1
0≤ i
≤ i
= 2 =2 −1
i=1 3 i=1 3 i=1 3 1 − 1/3
Ç å
1
=3 1− −2
3n+1
1
=1−
3n
hence In ⊆ [0, 1]. So being an intersection of closed and bounded sets, C is closed and
bounded; hence compact.
Let’s observe that the family (In )n∈N is decreasing. Indeed, given x ∈ In there exists
a1 , . . . , an ∈ {0, 2} such that
n n
X ai X ai 1
≤x≤ +
i=1 3i i=1 3i 3n
Since
an 1 1
n
+ n ≤ n−1 ,
3 3 3
we have
n−1 n−1
X ai X ai 1
i
≤ x ≤ i
+ n−1
i=1 3 i=1 3 3
so x ∈ In−1 .
Consider the set P = {0, 2}N , that is the set of sequences in {0, 2}. Define a map
λ:P→C
X an
(an )n∈N 7→ n
n∈N 3

To see that this map is well-defined, given (an )n∈N with an ∈ {0, 2} write
n
X ai
Sn = .
i=1 3i
For every m ∈ N we claim that Sn ∈ Im . We may assume m ≥ n. Let bi = ai for
1 ≤ i ≤ n and bi = 0 for i > m, then we have
m
X bi
Sn = ∈ Im .
i=1 3i
And since Im is closed, the limit
Xan
lim Sn = n
n∈N 3
n→∞

(which definitely exists by comparison with geometric series) lies in Im for every m.
X an
Hence n
is in C.
n∈N 3
REAL ANALYSIS I HOMEWORK 1 6

Next, we show that λ is injective. So suppose (an ) and (bn ) are two sequences where
an , bn ∈ {0, 2} such that
X an X bn
n
= n
.
n∈N 3 n∈N 3

Suppose that the sequences do not coincide. So let k be the first number such that
ak 6= bk . WLOG we may assume ak = 2 and bk = 0.
X an X b n
0= n
− n
n∈N 3 n∈N 3
X an − b n
=
n∈N 3n

2 X an − b n
= k
+
3 n=k+1 3n
Hence
∞ ∞
2 X |an − bn | X 2
k
= n
≤ n
.
3 n=k+1 3 n=k+1 3

2
Cancelling yields
3k

X 1 1 1
1≤ n
= 1 −1=
n=1 3 1− 3
2
which is nonsense. Thus λ must be injective and as a result we conclude that C is
uncountable as P = {0, 2}N is uncountable.
Finally we deal with the outer measure of C. We first show that" given two distinct#
n n
X ai X ai 1
n-tuples a1 , . . . , an and b1 , . . . , bn with an , bn ∈ {0, 2}, the intervals i
, i
+
i=1 3 i=1 3 3n
" n n
#
X bi X bi 1
and i
, i
+ are disjoint. Similar to above, let k be the first number among
i=1 3 i=1 3 3n
1, . . . , n where ak 6= bk and then we may assume ak = 2 and bk = 0. Then
n n n n
X ai X bi 2 X ai − b i 2 X −2
− = + ≥ +
i=1 3i i=1 3i 3k i=k+1 3i 3k i=k+1 3i
n
2 X 1
= k
− 2 i
3 i=k+1 3
1 1
!
2 1 − 3n+1 1 − 3k+1
= k −2 −
3 1 − 13 1 − 13
1 1
2 3k+1
− 3n+1
= − 2 · 2
3k 3
Ç å
2 1 1
= k − 3 k+1 − n+1
3 3 3
1 1 1
= k+ n > n
3 3 3
REAL ANALYSIS I HOMEWORK 1 7

thus
n n
X ai X bi 1
> + .
i=1 3i i=1 3i 3n
We established that the defining intervals of In are disjoint closed intervals of length
1
. And the number of these intervals is the number of n-tuples taking values in {0, 2},
3n
which is 2n . Thus by countable additivity of Lebesgue measure, we conclude that
2n
m(In ) = n .
3
Finally, for every n ∈ N we have C ⊆ In therefore
Ç ån
2
m(C) ≤
3
for every n ∈ N. Thus m(C) = 0.

Exercise 1.2.10. Show that the half-open interval [0, 1) cannot be expressed as
the countable union of disjoint closed intervals. For an additional challenge, show
that [0, 1) cannot be expressed as the countable union of disjoint closed sets.

Potrebbero piacerti anche