You are on page 1of 9

REAL ANALYSIS I HOMEWORK 3

CİHAN BAHRAN

The questions are from Stein and Shakarchi’s text.

Chapter 1

18. Prove the following assertion: Every measurable function is the limit a.e. of a
sequence of continuous functions.

We first suppose that f : E → R is a measurable function (finite valued) with m(E) <
∞. Then for every n ∈ N, by Lusin’s theorem there exists a closed set Fn ⊆ E such
that m(E − Fn ) ≤ 1/n and f |Fn is continuous.
For each n ∈ N, write Cn = nk=1 Fn and define gn = f |Cn . By the pasting lemma every
S

gn is continuous (the continuous f |Fk ’s are pasted on finitely many closed sets).
Now since each Cn is closed, by the Tietze extension theorem there exists a continuous
function hn : E → R such that hn |Cn = gn . We claim that hn → f a.e. For, letting
B = n∈N Cn , since Cn ’s are increasing, every x ∈ B eventually falls in the Cn ’s (that is,
S

there exists N ∈ N such that x ∈ Cn whenever n ≥ N ); therefore limn→∞ hn (x) = f (x).


So hn → f on B and
\ \
m(E − B) = m( (E − Cn )) ≤ m( (E − Fn )) = 0 .
n∈N n∈N

20. Show that there exists closed sets A and B with m(A) = m(B) = 0, but
m(A + B) > 0:
(a) In R, let A = C (the Cantor set), B = C/2. Note that A + B ⊇ [0, 1].
(b) In R2 , observe that if A = I × {0} and B = {0} × I (where I = [0, 1]), then
A + B = I × I.

(a) Let x ∈ [0, 1]. We know that x has a ternary expansion

an 3−n .
X
x=
n∈N

Then if we define

a
n if an 6= 1
bn = 
0 otherwise

and

2 if an = 1
cn = 
0 otherwise
1
REAL ANALYSIS I HOMEWORK 3 2

then we have
X cn −n
bn 3−n +
X
x= 3
n∈N n∈N 2
1X
bn 3−n + cn 3−n ∈ C + C/2
X
=
n∈N 2 n∈N
since (bn ) and (cn ) are sequences of 0’s and 2’s. As x was arbitrary above, we obtain
[0, 1] ⊆ A + B. Hence m(A + B) ≥ 1, but A and B are closed sets of measure zero.
(b) Given (x, y) ∈ I × I, (x, y) = (x, 0) + (0, y) ∈ A + B. So I × I ⊆ A + B and the
reverse containment is similar so we have A+ B = I × I. Therefore m(A + B) = 1,
however both A and B can be covered by rectangles of area ε for any given ε > 0,
therefore m(A) = m(B) = 0.
21. Prove that there is a continuous function that maps a Lebesgue measurable
set to a non-measurable set.
[Hint: Consider a non-measurable subset of [0, 1], and its inverse image in C by the
function F in Exercise 2.]

We know that [0, 1] contains a non-measurable set N . So if we let F : C → [0, 1] to be


the function in Exercise 2, since A := F −1 (N ) ⊆ C, we have m∗ (A) = 0 and hence A is
measurable. But as F is surjective, F (A) = N , yet N is not measurable.
22. Let χ[0,1] be the characteristic function of [0, 1]. Show that there is no every-
where continuous function f on R such that
f (x) = χ[0,1] (x)
almost everywhere.

Write g = χ[0,1] and suppose such an f exists. Since g = 0 on a set of positive measure
(namely Rr [0, 1]), there exists a ∈ R such that f (a) = 0. Similarly, there exists b ∈ R
such that f (b) = 1. WLOG we may assume a < b. Consider the set
S = {x ∈ [a, b] : f (x) = 1} .
First, S 6= Ø since b ∈ S. And S is bounded below from a, hence has an infimum, say
d ≥ a. So there exists a sequence (xn ) in S such that f (xn ) = 1 and limn→∞ xn = d.
Since f is continuous, we get f (d) = 1. So a < d ≤ b and by the definition of c, for every
x ∈ [a, d) we have f (x) 6= 1. In a similar fashion, if we define c to be the supremum of
the set
{x ∈ [a, d] : f (x) = 0}
then f (c) = 0, c < d and for every x ∈ (c, d] we have f (x) 6= 0. Therefore 0, 1 ∈
/ f ((c, d)).
But g is always either 0 or 1, so
m ({x ∈ R : f (x) 6= g(x)}) ≥ m((c, d)) = d − c > 0
is a contradiction.
26. Suppose A ⊆ E ⊆ B, where A and B are measurable sets of finite measure.
Prove that if m(A) = m(B), then E is measurable.

Note that m(B) = m(A)+m(BrA) and since m(A) = m(B) < ∞ we have m(BrA) = 0.
Now E r A ⊆ B r A, so E r A is a null set and in particular measurable. Thus
REAL ANALYSIS I HOMEWORK 3 3

E = A ∪ (E r A) is measurable.

28. Let E be a subset of R with m∗ (E) > 0. Prove that for each 0 < α < 1, there
exists an open interval I so that
m∗ (E ∩ I) ≥ αm∗ (I).
Loosely speaking, the estimate shows that E contains almost a whole interval.
[Hint: Choose an open set O that contains E, and such that m∗ (E) ≥ αm∗ (O).
Write O as the countable union of disjoint open intervals, and show that one of
these intervals must satisfy the desired property.]
1
Since m∗ (E) > 0 and α
> 1, we have m∗ (E) < α1 m∗ (E). So
1
m∗ (E) = inf{m(O) : E ⊆ O-open} < m∗ (E)
α
therefore there exists an open set O ⊆ R containing E such that
1
m(O) < m∗ (E) .
α
Write O = n∈N In where In ’s are disjoint (nonempty) open intervals. Then
S
[
E =E∩O = (E ∩ In )
n∈N

so we get
X X
αm(In ) = αm(O) < m∗ (E) ≤ m∗ (E ∩ In ) .
n∈N n∈N

Hence the statement “for every n ∈ N, αm(In ) ≥ m∗ (E ∩ In )” would yield a contradic-


tion. Thus there exists n ∈ N such that αm(In ) < m∗ (E ∩ In ).
37. Suppose Γ is a curve y = f (x) in R2 , where f is continuous. Show that
m(Γ) = 0.

Note that since the map x 7→ −x preserves areas of rectangles, Γ has the same measure
with the curve given by y = |f (x)|. Therefore we may assume f is nonnegative. Also
since
[
Γ= {(x, f (x)) : x ∈ [n, n + 1]}
n∈N

and measure is countably subadditive, it suffices to show that each term in the above
union has measure zero. Thus we may assume that f : [a, b] → R where [a, b] ⊆ R is a
finite interval. Moreover by replacing f with f + 1, we may assume that f (x) ≥ 1 for
every x ∈ [a, b].
Then given 0 < ε < 1, the set
Eε = {(x, y) : a ≤ x ≤ b , f (x) − ε ≤ y ≤ f (x) + ε}
contains Γ. But since f ≥ 1 > ε, both f + ε and f − ε are nonnegative and continuous,
therefore the measure of Eε can be calculated by a definite Riemann integral as
Z b Z b Z b
m(Eε ) = (f (x) + ε)dx − (f (x) − ε)dx = 2ε dx = 2ε(b − a) ,
a a a
REAL ANALYSIS I HOMEWORK 3 4

So m(Γ) ≤ 2ε(b − a) for arbitrarily small ε. As a, b is independent from ε, this shows


that m(Γ) = 0.

Chapter 2

4. Suppose f is integrable on [0, b], and


Z b
f (t)
g(x) = dt for 0 ≤ x ≤ b.
x t
Prove that g is integrable on [0, b] and
Z b Z b
g(x)dx = f (t)dt .
0 0

First, we show the claim assuming f is non-negative. Let


E = {(x, t) ∈ R2 : 0 ≤ x < t ≤ b } .
Note that E is measurable and contained in [0, b] × (0, b]. Now define
h : [0, b] × (0, b] → R
χE (x, t)f (t)
(x, t) 7→
t
which is non-negative and measurable (it is obtained by algebraic operations on mea-
surable functions). Using Tonelli’s theorem we get the following: first, for almost every
x ∈ [0, b], the function
hx : (0, b] → R
t 7→ h(x, t)
is measurable. Second, the function
[0, b] → R
Z
x 7→ hx
(0,b]

is measurable. Note that this function is nothing but g because


Z b Z b
f (t)
Z
hx = h(x, t)dt = dt .
(0,b] 0 x t
Third, we have
Z Z Z b
h= g= g(x)dx .
[0,b]×(0,b] [0,b] 0

Also using Tonelli for the other variable, we get the following: First, for almost every
t ∈ (0, b), the function
ht : [0, b] → R
x 7→ h(x, t)
is measurable. Second, the function
(0, b] → R
Z
t 7→ ht
[0,b]
REAL ANALYSIS I HOMEWORK 3 5

is measurable. Note that this function is nothing but f , since


Z b Z t
f (t)
Z
ht = h(x, t)dx = dx = f (t) .
[0,b] 0 0 t
Third, we have
Z Z Z b
h= f= f (t)dt .
[0,b]×(0,b) (0,b] 0

Thus we get
Z b Z b
g(x)dx = f (t)dt .
0 0

As f is integrable on [0, b], the integral above is finite; hence g is also integrable on
[0, b].
For the general case, let P = {t ∈ (0, b] : f (t) ≥ 0} and N = (0, b] − P and define

f+ : (0, b] → R

f (t) if t ∈ P
t 7→ 
0 if t ∈ N ,
f− : (0, b] → R

0 if t ∈ P
t 7→ 
−f (t) if t ∈ N .

Observe that f+ and f− are integrable non-negative functions such that f = f+ − f− .


Now define

g+ : [a, b] → R
Z b
f+ (t)
x 7→ dt ,
x t
g− : [a, b] → R
Z b
f− (t)
x 7→ dt .
x t
By the first part g+ , g− are integrable and
Z b Z b
g+ (x)dx = f+ (t)dt ,
0 0
Z b Z b
g− (x)dx = f− (t)dt .
0 0

Observe that
f (t)
Z
g+ (x) = dt ,
t∈[x,b]∩P t
−f (t)
Z
g− (x) = dt .
t∈[x,b]∩N t
REAL ANALYSIS I HOMEWORK 3 6

Thus
f (t) f (t)
Z Z
g+ (x) − g− (x) = dt + dt
t∈[x,b]∩P t t∈[x,b]∩N t
f (t)
Z
= dt = g(x)
t∈[x,b] t

and hence since everything is integrable we can do subtraction to get


Z b Z b Z b
g(x)dx = g+ (x)dx − g− (x)dx
0 0 0
Z b Z b
= f+ (t)dt − f− (t)dt
0 0
Z b
= f (t)dt < ∞ .
0

6. Integrability of f on R does not necessarily imply the convergence of f (x) to 0


as x → ∞.
(a) There exists a positive continuous function f on R so that f is integrable on R,
but yet lim supx→∞ f (x) = ∞.
(b) However, if we assume that f is uniformly continuous on R and integrable, then
lim|x|→∞ f (x) = 0.
[Hint: For (a), construct a continuous version of the function equal to n on the
segment [n, n + 1/n3 ) , n ≥ 1.]

(a) To patch the holes of the hinted function to extend it to a continuous function in a
nice way, we use a lemma.
Lemma 1. Let a, b, c, d ∈ R such that a < b and c, d > 0. Then
®Z ´
b
+
inf f (x)dx f : [a, b] → R is continuous, f (a) = c, f (b) = d = 0 .
a

Proof. Let ε > 0 such that ε < min{(b − a)/2, c, d}. Then the function
f : [a, b] → R

ε−c
 ε (x − a) + c if a ≤ x ≤ a + ε


x 7→ ε if a + ε ≤ x ≤ b − ε
 d−ε
(x − b) + d if b − ε ≤ x ≤ b

ε
is well-defined, positive and continuous. Its graph looks like the following:

So we see that
Z b
1 1
f (x)dx = ε(c − ε) + ε(b − a) + ε(d − ε)
a 2Ç 2å
c−ε d−ε
=ε +b−a+
2 2
Ç å
c+d
=ε +b−a−ε
2
≤ ε(c + d + b − a) .
REAL ANALYSIS I HOMEWORK 3 7

d y = f (x)

ε
a a+ε b−ε b x

Hence indeed, the integral of a positive continuous function f with f (a) = c and
f (b) = d can be made arbitrarily small. 
Corollary 2. Let C be a closed subset of R and f : C → R be a positive continuous
function. Then given
R
εR> 0, f can be extended to a positive continuous function g :
R → R such that g ≤ C f + ε.

Proof. Since U := R − C is an open subset of R, it is a (countable) union of disjoint


open intervals, that is, there exists sequences (an ), (bn ) such that
a1 < b1 < a2 < b2 < a3 < b3 < · · ·
and
[
U= (an , bn ) .
n∈N
Therefore each an and bn belongs to C. So by the above proposition, for every n ∈ N
there exists a positive continuous function gn : [an , bn ] → R with gn (an ) = f (an ),
gn (bn ) = f (bn ) such that
ε
Z
gn ≤ n .
[an ,bn ] 2
So the function
g:R→R

f (x) if x ∈ C
x 7→ 
gn (x) if x ∈ [an , bn ]
is well-defined and continuous.Z And Z Z
g= g+ g
ZC U
XZ
= f+ gn
C n∈N [an ,bn ]
Z
≤ f + ε.
C

REAL ANALYSIS I HOMEWORK 3 8

Now let ñ ô

[ 1
C= n, n + 3
n=2 n
which, being a union of uniformly distant closed sets, is closed. And the function
f :C→R
¶ î ó
1
x 7→ n if x ∈ n, n + n3

is well-defined and continuous, with


∞ ∞
1 1
Z X X
f= 3
n = 2
< ∞.
C n=2 n n=2 n

So byR the corollary, f can be extended to a continuous function g : R → R such


that g < ∞, that is, g is integrable. However, since lim supx→∞ f (x) = ∞ we have
lim supx→∞ g(x) = ∞.
(b) We prove the contrapositive, that is, assuming that f is uniformly continuous and
lim|x|→∞ f (x) 6= 0 we show that |f | is not integrable (it follows that f is not integrable).
By uniform continuity, there exists 0 < δ < 1 such that |f (x) − f (y)| ≤ ε/2 whenever
|x − y| < δ.
Also since lim|x|→∞ f (x) 6= 0 there exists x1 such that |f (x1 )| ≥ ε. Again since
lim|x|→∞ f (x) 6= 0, we may choose x2 > x1 + 1 such that |f (x2 )| ≥ ε. Then we
choose x3 > x2 + 1 such that |f (x3 )| ≥ ε and proceed in this way. Since δ < 1, by our
construction the collection of intervals In := (xn − δ, xn + δ) -as n varies- is disjoint.
Note that for fixed n, if y ∈ In then |f (xn ) − f (y)| ≤ ε/2. But |f (xn )| ≥ ε, so
|f (y)| ≥ ε/2 by triangle inequality. Therefore
Z Z XZ X
|f | ≥ S |f | = |f | ≥ ε/2 = ∞ ,
I
n∈N n n∈N In n∈N

as desired.
(b) For every k ∈ N write Ek = {x ∈ R : |f (x)| > 1/k}. It suffices to show that each
Ek is bounded. Suppose not, that is, for some k the set Ek is unbounded. Then we can
find a sequence {xn } in Ek such that |xn+1 | > |xn | + 1 for every n. Now by uniform
continuity there exists δ ∈ (0, 1) such that |x − y| < δ implies |f (x) − f (y)| < 1/2k.
Note that the intervals In := (xn − δ, xn + δ) are disjoint. And given y ∈ In since
|f (xn )| > 1/k we have |f (y)| > 1/2k. Thus
Z Z XZ X
|f | ≥ S |f | = |f | ≥ 1/2k = ∞ ,
I
n∈N n n∈N In n∈N

contradicting that f ∈ L1 (R).


Rx
8. If f is integrable on R, show that F (x) = −∞ f (t)dt is uniformly continuous.

Given x, y ∈ R with x ≤ y, by additivity of the Lebesgue integral we have


Z Z Z
f+ f= f
(−∞,x] [x,y] (−∞,y]
Z x Z y Z y
f (t)dt + f (t)dt = f (t)dt .
−∞ x −∞
REAL ANALYSIS I HOMEWORK 3 9

Since f is integrable on R, the above integrals are all finite. Therefore we can perform
usual algebra to get
Z y Z y Z x
f (t)dt = f (t)dt − f (t)dt
x −∞ −∞
= F (y) − F (x) .
Given ε > 0, by Proposition 1.12 part (ii) in Stein & Shakarchi’s text, there exists δ > 0
such that Z y
Z y

|F (y) − F (x)| =
f (t)dt ≤ |f (t)|dt < ε
x x
whenever |x − y| < δ (taking E = [x, y] in the statement of the proposition). This is
precisely uniform continuity for F .
11. Prove that if f is integrable on Rd , real-valued, andR E f (x)dx ≥ 0 for every
R

measurable E, then f (x) ≥ 0 a.e. x. As a result, if E f (x)dx = 0 for every


measurable E, then f (x) = 0 a.e.

Write En = {x ∈ Rd : f (x) < −1/n}. En ’s are measurable. Note that


{x ∈ Rd : f (x) < 0} =
[
En ,
n∈N

So it is enough to show that every En has measure zero. Suppose not, so m(En ) > 0
for some n. So using the assumption on En , we have
−1 −1
Z Z
0≤ f≤ = m(En ) < 0 ,
En En n n
a contradiction.
Now let’s
R
doR the second part. By the first part, we have f ≥ 0 a.e. Writing g = −f ,
since E g = E (−f ) = 0 for every measurable E, again by the first part we deduce that
−f = g ≥ 0 a.e. Thus f ≤ 0 a.e. and hence f = 0 a.e.
12. Show that there are f ∈ L1 (Rd ) and a sequence {fn } with fn ∈ L1 (Rd ) such
that
||f − fn ||L1 → 0.
but fn (x) → f (x) for no x.
[Hint: In R, let fn = χIn , where In is an appropriately chosen sequence of intervals
with m(In ) → 0.]

For every n ∈ N there exists unique


î k,ój ∈ N such that n + 2k + j. Note that k → ∞
as n → ∞. So defining In = 2jk , j+1 2k
, by binary expansions every x ∈ [0, 1] belong
in infinitely many In ’s and also every x ∈ [0, 1] belong in the complement of infinitely
many In ’s. So given x ∈ [0, 1], if we define fn = χIn , the limit lim fn (x) does not exist.
However Z
||fn || = |fn | = m(In ) = 2−k → 0
as n → ∞. So taking f = 0 gives a counterexample.

You might also like