Measure and Integration Full

MEASURE and INTEGRATION
Problems with Solutions

Anh Quang Le, Ph.D.
October 8, 2013
www.MATHVN.com - Anh Quang Le, PhD

1
NOTATIONS
A(X): The -algebra of subsets of X.
(X, A(X), ) : The measure space on X.
B(X): The -algebra of Borel sets in a topological space X.
ML : The -algebra of Lebesgue measurable sets in R.
(R, ML , L ): The Lebesgue measure space on R.
L : The Lebesgue measure on R.
L : The Lebesgue outer measure on R.
1E or E : The characteristic function of the set E.
www.MathVn.com - Math Vietnam

2
Contents
Contents
1 Measure on a -Algebra of Sets
2 Lebesgue Measure on R
21
3 Measurable Functions
33
4 Convergence a.e. and Convergence in Measure
45
5 Integration of Bounded Functions on Sets of Finite Measure
53
6 Integration of Nonnegative Functions
63
7 Integration of Measurable Functions
75
8 Signed Measures and Radon-Nikodym Theorem
97
9 Differentiation and Integration
109
10 Lp Spaces
121
11 Integration on Product Measure Space
141
12 Some More Real Analysis Problems
151
www.MathVn.com3 - Math Vietnam

4
CONTENTS
Chapter 1
Measure on a -Algebra of Sets
1. Limits of sequences of sets
Definition 1 Let (An )nN be a sequence of subsets of a set X.
(a) We say that (An ) is increasing if An An+1 for all n N, and decreasing if An An+1 for
all n N.
(b) For an increasing sequence (An ), we define
lim An :=
An .
n=1
For a decreasing sequence (An ), we define

lim An :=
An .
n=1
Definition 2 For any sequence (An ) of subsets of a set X, we define

[ \
Ak
lim inf An :=
n
nN kn
lim sup An :=
n
\ [
Ak .
nN kn
Proposition 1 Let (An ) be a sequence of subsets of a set X. Then

(i)
(ii)
(iii)
lim inf An = {x X : x An for all but finitely many n N}.

n
lim sup An = {x X : x An for infinitely many n N}.

n
lim inf An lim sup An .

n
2. -algebra of sets
www.MathVn.com5 - Math Vietnam

6
CHAPTER 1. MEASURE ON A -ALGEBRA OF SETS
Definition 3 (-algebra)
Let X be an arbitrary set. A collection A of subsets of X is called an algebra if it satisfies the
following conditions:
1. X A.
2. A A Ac A.
3. A, B A A B A.
An algebra A of a set X is called a -algebra if it satisfies the additional condition:
S
4. An A, n N nN An n N.
Definition 4 (Borel -algebra)
Let (X, O) be a topological space. We call the Borel -algebra B(X) the smallest -algebra of X
containing O.
It is evident that open sets and closed sets in X are Borel sets.
3. Measure on a -algebra
Definition 5 (Measure)
Let A be a -algebra of subsets of X. A set function defined on A is called a measure if it
satisfies the following conditions:
1. (E) [0, ] for every E A.
2. () = 0.
3. (En )nN A, disjoint
S
nN
P
En = nN (En ).
Notice that if E A such that (E) = 0, then E is called a null set. If any subset E0 of a null set
E is also a null set, then the measure space (X, A, ) is called complete.
Proposition 2 (Properties of a measure)
A measure on a -algebra A of subsets of X has the following
properties:
Sn
Pn
(1) Finite additivity: (E1 , E2 , ..., En ) A, disjoint = ( k=1 Ek ) = k=1 (Ek ).
(2) Monotonicity: E1 , E2 A, E1 E2 = (E1 ) m(E2 ).
(3) E1 , E2 A, E1 E2 , (E1 ) < = (E
S2 \ E1 ) =
(E
P2 ) (E1 ).
(4) Countable subadditivity: (En ) A = nN En nN (En ).
Definition 6 (Finite, -finite measure)

Let (X, A, ) be a measure space.
1. is called finite if (X) < .
2. is called -finite if there exists a sequence (En ) of subsets of X such that
[
En and (En ) < , n N.
X=
nN

7
4. Outer measures
Definition 7 (Outer measure)
Let X be a set. A set function defined on the -algebra P(X) of all subsets of X is called an
outer measure on X if it satisfies the following conditions:
(i) (E) [0, ] for every E P(X).
(ii) () = 0.
(iii) E, F P(X), E F (E) (F ).
(iv) countable subadditivity:
(En )nN P(X),
nN
!
En
(En ).
nN
Definition 8 (Caratheodory condition)

We say that E P(X) is -measurable if it satisfies the Caratheodory condition:
(A) = (A E) + (A E c ) for every A P(X).
We write M( ) for the collection of all -measurable E P(X). Then M( ) is a -algebra.
Proposition 3 (Properties of )
(a) If E1 , E2 M( ), then E1 E2 M( ).
(b) is additive on M( ), that is,
E1 , E2 M( ), E1 E2 = = (E1 E2 ) = (E1 ) + (E2 ).

8
Problem 1
Let A be a collection of subsets of a set X with the following properties:
1. X A.
2. A, B A A \ B A.
Show that A is an algebra.
Solution
(i) X A.
(ii) A A Ac = X \ A A (by 2).
(iii) A, B A A B = A \ B c A since B c A (by (ii)).
Since Ac , B c A, (A B)c = Ac B c A. Thus, A B A.
Problem 2
(a) ShowSthat if (An )nN is an increasing sequence of algebras of subsets of a set
X, then nN An is an algebra of subsets of X.
(b) Show by example that even if An in (a) is a -algebra for every n N, the
union still may not be a -algebra.
Solution S
(a) Let A = nN An . We show that A is an algebra.
(i) Since X An , n N, so X A.
(ii) Let A A. Then A An for some n. And so Ac An ( since An is an
algebra). Thus, Ac A.
(iii) Suppose A, B A. We shall show A B A.
S
Since {An } is increasing, i.e., A1 A2 ... and A, B nN An , there is
some n0 N such that A, B A0 . Thus, A B A0 . Hence, A B A.
(b) Let X = N, An = the family of all subsets of {1, 2, ...,Sn} and their complements.
Clearly, An is a -algebra and A1 A2 .... However, nN An is the family of all
finite and co-finite subsets of N, which is not a -algebra.

9
Problem 3
Let X be an arbitrary infinite set. We say that a subset A of X is co-finite if its
complement Ac is a finite subset of X. Let A consists of all the finite and the
co-finite subsets of a set X.
(a) Show that A is an algebra of subsets of X.
(b) Show that A is a -algebra if and only if X is a finite set.
Solution
(a)
(i) X A since X is co-finite.
(ii) Let A A. If A is finite then Ac is co-finite, so Ac A. If A co-finite then Ac
is finite, so Ac A. In both cases,
A A Ac A.
(iii) Let A, B A. We shall show A B A.
If A and B are finite, then A B is finite, so A B A. Otherwise, assume
that A is co-finite, then A B is co-finite, so A B A. In both cases,
A, B A A B A.
(b) If X is finite then A = P(X), which is a -algebra.
To show the reserve, i.e., if A is a -algebra then X is finite, we assume that X
is infinite. So we can find an infinite sequence (a1 , a2 , ...) of distinct elements of X
such that
S X \ {a1 , a2 , ...} is infinite. Let An = {anS}. Then An A for any n N,
while nN An is neither finite nor co-finite. So nN An
/ A. Thus, A is not a
-algebra: a contradiction!
Note:
For an arbitrary collection C of subsets of a set X, we write (C) for the smallest
-algebra of subsets of X containing C and call it the -algebra generated by C.
Problem 4
Let C be an arbitrary collection of subsets of a set X. Show that for a given
A (C), there exists a countable sub-collection CA of C depending on A such
that A (CA ). (We say that every member of (C) is countable generated).

10
Solution
Denote by B the family of all subsets A of X for which there exists a countable
sub-collection CA of C such that A (CA ). We claim that B is a -algebra and
that C B.
The second claim is clear, since A ({A}) for any A C. To prove the first one,
we have to verify that B satisfies the definition of a -algebra.
(i) Clearly, X B.
(ii) If A B then A (CA ) for some countable family CA (C). Then
Ac (CA ), so Ac B.
(iii) Suppose S
{An }nN B. Then An (CAn ) for some countable family CAn C.
Let E = nN CAn then E isS
countable and E C andSAn (E) for all n N.
By definition of -algebra, nN An (E), and so nN An B.
Thus, B is a -algebra of subsets of X and E B. Hence,
(E) B.
By definition of B, this implies that for every A (C) there exists a countable
E C such that A (E).
Problem 5
Let a set function defined on a -algebra A of subsets of X. Show that it is
additive and countably subadditive on A, then it is countably additive on A.
Solution
We first show that the additivity of implies its monotonicity. Indeed, let A, B A
with A B. Then
B = A (B \ A) and A (B \ A) = .
Since is additive, we get
(B) = (A) + (B \ A) (A).
Now let (En ) be a disjoint sequence in A. For every N N, by the monotonicity
and the additivity of , we have
!
N
!
N
[
[
X
En
En =
(En ).
nN
n=1
n=1

11
Since this holds for every N N, so we have
!
[
X
(i)
En
(En ).
nN
nN
On the other hand, by the countable subadditivity of , we have
!
X
[
(En ).
(ii)
En
nN
nN
From (i) and (ii), it follows that
!
En
nN
This proves the countable additivity of .
(En ).
nN
Problem 6
Let X be an infinite set and A be the algebra consisting of the finite and co-finite
subsets of X (cf. Prob.3). Define a set function on A by setting for every
A A:
0
if A is finite
(A) =
1 if A is co-finite.
(a) Show that is additive.
(b) Show that when X is countably infinite, is not additive.
(c) Show that when X is countably infinite, then X is the limit of an increasing
sequence {An : n N} in A with (An ) = 0 for every n N, but (X) = 1.
(d) Show that when X is uncountably, the is countably additive.
Solution
(a) Suppose A, B A and A B = (i.e., A B c and B Ac ).
If A is co-finite then B is finite (since B Ac ). So A B is co-finite. We have
(A B) = 1, (A) = 1 and (B) = 0. Hence, (A B) = (A) + (B).
If B is co-finite then A is finite (since A B c ). So A B is co-finite, and we have
the same result. Thus, is additive.
(b) Suppose X is countably infinite. We can then put X under this form: X =
{x1 , x2 , ...}, xi 6= xj if i 6= j. Let AP
n = {xn }. Then the family {An }nN is disjoint
and (An ) = 0 for every n N. So nN (An ) = 0. On the other hand, we have

12
S
nN

An = X, and (X) = 1. Thus,
!
[
X
An 6=
(An ).
nN
nN
Hence, is not additive.

(c) Suppose X is countably infinite, and X = {x1 , x2 , ...}, xi 6= xj if i 6= j as in
(b). Let Bn = {x1 , x2 , ..., xn }. Then (Bn ) = 0 for every n N, and the sequence
(Bn )nN is increasing. Moreover,
[
Bn = X and (X) = 1.
lim Bn =
n
nN
(d) Suppose XS is uncountably. Consider the family of disjoint sets {Cn }nN in A.
Suppose C = nN Cn A. We first claim: At most one of the Cn s can be co-finite.
Indeed, assume there are two elements Cn and Cm of the family are co-finite. Since
Cm Cnc , so Cm must be finite: a contradiction.
Suppose Cn0 is the co-finite set. Then since C Cn0 , C is also co-finite. Therefore,
!
[
(C) =
Cn = 1.
nN
On the other hand, we have

(Cn0 ) = 1 and (Cn ) = 0 for n 6= n0 .
Thus,
If all Cn are finite then
nN
!
Cn
(Cn ).
nN
Cn is finite, so we have
!
[
X
0=
Cn =
(Cn ).
nN
nN
nN
Problem 7
Let (X, A, ) be a measure space. Show that for any A, B A, we have the
equality:
(A B) + (A B) = (A) + (B).

13
Solution
If (A) = or (B) = , then the equality is clear. Suppose (A) and (B) are
finite. We have
A B = (A \ B) (A B) (B \ A),
A = (A \ B) (A B)
B = (B \ A) (A B).
Notice that in these decompositions, sets are disjoint. So we have
(1.1)
(1.2)
(A B) = (A \ B) + (A B) + (B \ A),
(A) + (B) = 2(A B) + (A \ B) + (B \ A).
From (1.1) and (1.2) we obtain

(A B) (A) (B) = (A B).
The equality is proved.
Problem 8
The symmetry difference of A, B P(X) is defined by
A 4 B = (A \ B) (B \ A).
(a) Prove that
A, B, C P(X), A 4 B (A 4 C) (C 4 B).
(b) Let (X, A, ) be a measure space. Show that
A, B, C A, (A 4 B) (A 4 C) + (C 4 B).
Solution
(a) Let x A 4 B. Suppose x A \ B. If x C then x C \ B so x C 4 B. If
x
/ C, then x A \ C, so x A 4 C. In both cases, we have
x A 4 B x (A 4 C) (C 4 B).
The case x B \ A is dealt with the same way.
(b) Use subadditivity of and (a).

14
Problem 9
Let X be an infinite set and the counting measure on the -algebra A = P(X).
Show that there exists a decreasing sequence (En )nN in A such that
lim En = with
lim (En ) 6= 0.
Solution
Since X is a infinite set, we can find an countably infinite set {x1 , x2 , ...} X with
xi 6= xj if i 6= j. Let En = {xn , xn+1 , ...}. Then (En )nN is a decreasing sequence in
A with
lim En = and lim (En ) = 0.
n
Problem 10 (Monotone sequence of measurable sets)

Let (X, A, ) be a measure space, and (En ) be a monotone sequence in A.
(a) If (En ) is increasing, show that
lim (En ) = lim En .

n
(b) If (En ) is decreasing, show that
lim (En ) = lim En ,
provided that there is a set A A satisfying (A) < and A E1 .
Solution
S
Recall that if (En ) is increasing
then
lim
E
=
n
n
nN En A, and if (En ) is
T
decreasing then limn
E
=
E
A.
Note
also
that if (En ) is a monotone
n
nN n
sequence in A, then (En ) is a monotone sequence in [0, ] by the monotonicity

of , so that limn (En ) exists in [0, ].
(a) Suppose (En ) is increasing. Then the sequence (En ) is also increasing.
Consider the first case where (En0 ) = for some En0 . In this case we have
limn (En ) = . On the other hand,
En0
[
nN
En = lim En = lim En (En0 ) = .

n

15
Thus
lim En = = lim (En ).

n
Consider the next case where (En ) < for all n N. Let E0 = , then consider
the disjoint sequence (Fn ) in A defined by Fn = En \En1 for all n N. It is evident
that
[
[
En =
Fn .
nN
nN
Then we have
lim En =
n
!
En
nN
!
Fn
nN
(Fn ) =
nN
(En \ En1 )
nN
(En ) (En1 )
nN
=
=
lim
n
X
(Ek ) (Ek1 )
k=1
lim (En ) (E0 ) = lim (En ).
(b) Suppose (En ) is decreasing and assume the existence of a containing set A with
finite measure. Define a disjoint sequence (Gn ) in A by setting Gn = En \ En+1 for
all n N. We claim that
\
[
(1) E1 \
En =
Gn .
nN
nN
T
T
To show this, let x E1 \ nN En . Then x E1 and x
/ nN En . Since the
sequence (En ) is decreasing, there exists the first set En0 +1 in the sequence not
containing x. Then
[
x En0 \ En0 +1 = Gn0 = x
Gn .
nN
Conversely, if x nN Gn , then x Gn0 = ET

N. Now
n0 \ En0 +1 for some n0 T
x En0 E1 . Since x
/ En0 +1 , we have x
/ nN En . Thus x E1 \ nN En .
Hence (1) is proved.
Now by (1) we have
!
\
[
(2) E1 \
En =
Gn .
nN
nN

16
Since

T
E
(E1 ) (A) < , we have
n
nN
!
\
\
(3) E1 \
En
= (E1 )
En
nN
nN
= (E1 ) ( lim En ).
n
By the countable additivity of , we have
!
[
X
X
(4)
Gn
=
(Gn ) =
(En \ En+1 )
nN
nN
nN
(En ) (En+1 )
nN
=
=
lim
n
X
(Ek ) (Ek+1 )
k=1
lim (E1 ) (En+1 )
= (E1 ) lim (En+1 ).

n
Substituting (3) and (4) in (2), we have

(E1 ) ( lim En ) = (E1 ) lim (En+1 ).
n
Since (E1 ) < , we have

( lim En ) = lim (En ).
n

17
Problem 11 (Fatous lemma for )

Let (X, A, ) be a measure space, and (En ) be a sequence in A.
(a) Show that
lim inf En lim inf (En ).

n
(b) If there exists A A with En A and (A) < for every n N, then
show that
lim sup En lim sup (En ).

n
Solution
(a) Recall that
[ \
lim inf En =
n
by the fact that

we have
()
T
kn
Ek
nN kn
nN
Ek = lim
Ek ,
kn
is an increasing sequence in A. Then by Problem 9a
lim inf En = lim

n
!
Ek
= lim inf
n
kn
!
Ek
kn
since the
is equal to the limit inferior of the sequence.
T limit of a sequence, if it exists,
T
Since kn Ek En , we have kn Ek (En ) for every n N. This implies
that
!
\
Ek lim inf (En ).
lim inf
n
kn
Thus by () we obtain
lim inf En lim inf (En ).

n
(b) Now
lim sup En =
n
\ [
Ek = lim
nN kn
Ek ,
kn
by the fact that

Ek nN is an decreasing sequence in A. Since En A for all
kn
S
n N, we have kn Ek A for all n N. Thus by Problem 9b we have
lim sup En =
n
lim
[
kn
!
Ek
= lim
n
kn
!
Ek

18
Now
lim
!
Ek
= lim sup
n
kn
!
Ek
kn
since the S
limit of a sequence, if it exists, is equal to the limit superior of the sequence.
Then by kn Ek En we have
!
[
Ek (En ).
kn
Thus
lim sup
n
It follows that
!
Ek
lim sup (En ).

n
kn

n
Problem 12
Let be an outer measure on a set X. Show that the following two conditions
are equivalent:
(i) is additive on P(X).
(ii) Every element of P(X) is -measurable, that is, M( ) = P(X).
Solution
Suppose is additive on P(X). Let E P(X). Then for any A P(X),
A = (A E) (A E c ) and (A E) (A E c ) = .
By the additivity of on P(X), we have
(A) = (A E) + (A E c ).
This show that E satisfies the Caratheodory condition. Hence E M( ). So
P(X) M( ). But by definition, M( ) P(X). Thus
M( ) = P(X).
Conversely, suppose M( ) = P(X). Since is additive on M( ) by Proposition 3, so is additive on P(X).

19
Problem 13
Let be an outer measure on a set X.
(a) Show that the restriction of on the -algebra M( ) is a measure on
M( ).
(b) Show that if is additive on P(X), then it is countably additive on P(X).
Solution
(a) By definition, is countably subadditive on P(X). Its restriction on M( )
is countably subadditive on M( ). By Proposition 3b, is additive on M( ).
Therefore, by Problem 5, is countably additive on M( ). Thus, is a measure
on M( ). But is the restriction of on M( ), so we can say that is a
measure on M( ).
(b) If is additive on P(X), then by Problem 11, M( ) = P(X). So is a
measure on P(X) (Problem 5). In particular, is countably additive on P(X).

20
Chapter 2
Lebesgue Measure on R
1. Lebesgue outer measure on R
Definition 9 (Outer measure)
Lebesgue outer measure on R is a set function L : P(R) [0, ] defined by
)
(
X
[
`(Ik ) : A
Ik , Ik is open interval in R .
L (A) = inf
k=1
k=1
Proposition 4 (Properties of L )
1. L (A) = 0 if A is at most countable.
2. Monotonicity: A B L (A) L (B).
3. Translation invariant: L (A + x) = L (A), x R.
S
P
4. Countable subadditivity: L ( n=1 An ) n=1 L (An ).
5. Null set: L (A) = 0 L (A B) = L (B) and L (B \ A) = L (B)
for all B P(R).
6. For any interval I R, L (I) = `(I).
7. Regularity:
E P(R), > 0, O open set in R : O E and L (E) L (O) L (E) + .
2. Measurable sets and Lebesgue measure on R
Definition 10 (Caratheodory condition)
A set E R is said to be Lebesgue measurable (or L -measurable, or measurable) if, for all A R,
we have
L (A) = L (A E) + L (A E c ).
www.MathVn.com21- Math Vietnam

22
CHAPTER 2. LEBESGUE MEASURE ON R
Since L is subadditive, the sufficient condition for Caratheodory condition is

L (A) L (A E) + L (A E c ).
The family of all measurable sets is denoted by ML . We can see that ML is a -algebra. The
restriction of L on ML is denoted by L and is called Lebesgue measure.
Proposition 5 (Properties of L )
1. (R, ML , L ) is a complete measure space.

2. (R, ML , L ) is -finite measure space.
3. BR ML , that is, every Borel set is measurable.
4. L (O) > 0 for every nonempty open set in R.
5. (R, ML , L ) is translation invariant.
6. (R, ML , L ) is positively homogeneous, that is,
L (E) = ||L (E), R, E ML .
Note on F and G sets:
Let (X, T ) be a topological space.
A subset E of X is called a F -set if it is the union of countably many closed sets.
A subset E of X is called a G -set if it is the intersection of countably many open sets.
If E is a G -set then E c is a F -set and vice versa. Every G -set is Borel set, so is every F -set.

Problem 14
If E is a null set in (R, ML , L ), prove that E c is dense in R.
Solution
For every open interval I in R, L (I) > 0 (property of Lebesgue measure). If
L (E) = 0, then by the monotonicity of L , E cannot contain any open interval as
a subset. This implies that
Ec I =
for any open interval I in R. Thus E c is dense in R.

23
Problem 15
Prove that for every E R, there exists a G -set G R such that
G E and L (G) = L (E).
Solution
We use the regularity property of L (Property 7).
For = n1 , n N, there exists an open set On R such that
On E and L (E) L (On ) L (E) +
Let G =
we have
T
nN
1
.
n
On . Then G is a G -set and G E. Since G On for every n N,

L (E) L (G) L (On ) L (E) +
1
.
n
This holds for every n N, so we have

L (E) L (G) L (E).
Therefore
(G) = (E).
Problem 16
Let E R. Prove that the following statements are equivalent:
(i) E is (Lebesgue) measurable.
(ii) For every > 0, there exists an open set O E with L (O \ E) .
(iii) There exists a G -set G E with L (G \ E) = 0.
Solution
(i) (ii) Suppose that E is measurable. Then
> 0, open set O : O E and L (E) L (O) L (E) + . (1)
Since E is measurable, with O as a testing set in the Caratheodory condition satisfied
by E, we have
L (O) = L (O E) + L (O E c ) = L (E) + L (O \ E). (2)

24
If L (E) < , then from (1) and (2) we get

L (O) L (E) + = L (O) L (E) = L (O \ E) .
If L (E) = , let En = E (n 1, n] for n Z. Then (En )nZ is a disjoint sequence
in ML with
[
En = E and L (En ) L (n 1, n] = 1.
nZ
Now, for every > 0, there is an open set On such that
Let O =
1
On En and L (On \ En ) . |n| .
3 2
nZ )On ,
then O is open and O E, and
!
!
!
!c
[
[
[
[
O\E =
On \
En =
On
En
nZ
"
On
nZ
!c #
En
"
On \
nZ
!#
En
nZ
nZ
nZ
nZ
nZ
(On \ En ).
nZ
Then we have
L (O \ E) L
!
(On \ En )
nZ
L (On \ E)
nZ
X1
X1
1
. |n| = + 2
. n
3
2
3
3
2
nZ
nN
=
1
2
+ = .
3
3
This shows that (ii) satisfies.

(ii) (iii) Assume that E satisfies (ii). Then for = n1 , n N, there is an open
set On such that
On En and L (On \ En )
Let G =
T
nN
1
, n N.
n
On . Then G is a G -set containing E. Now

G O = L (G \ E) L (On \ E)
1
, n N.
n

25
Thus L (G \ E) = 0. This shows that E satisfies (iii).
(iii) (i) Assume that E satisfies (iii). Then there exists a G -set G such that
G E and L (G \ E) = 0.
Now L (G \ E) = 0 implies that G \ E is (Lebesgue) measurable. Since E G,
we can write E = G \ (G \ E). Then the fact that G and G \ E are (Lebesgue)
measurable implies that E is (Lebesgue) measurable.
Problem 17(Similar problem)

Let E R. Prove that the following statements are equivalent:
(i) E is (Lebesgue) measurable.
(ii) For every > 0, there exists an closed set C E with L (E \ C) .
(iii) There exists a F -set F E with L (E \ F ) = 0.
Problem 18
Let Q be the set of all rational numbers in R. For any > 0, construct an open
set O R such that
O Q and L (O) .
Solution
Since Q is countable, we can write Q = {r1 , r2 , ...}. For any > 0, let

In = rn n+1 , rn + n+1 , n N.
2
2
S
Then In is open and O = n=1 In is also open. We have, for every n N, rn In .
Therefore O Q.
Moreover,
!
[
X
In
L (In )
L (O) = L
n=1
n=1
X
2
=
2n+1
n=1
X
1
= .
n
2
n=1

26
Problem 19
Let Q be the set of all rational numbers in R.
(a) Show that Q is a null set in (R, BR , L ).
(b) Show that Q is a F -set.
(c) Show that there exists a G -set G such that G Q and L (G) = 0.
(d) Show that the set of all irrational numbers in R is a G -set.
Solution
(a) Since Q is countable, we can write Q = {r1 , r2 , ...}. Each {rn }, n N is closed,
so {rn } BR . Since BR is a -algebra,
Q=
{rn } BR .
n=1
Since L ({rn }) = 0, we have

L (Q) =
L ({rn }) = 0.
n=1
Thus, Q is a null set in (R, BR , L ).

(b) Since {rn } is closed and Q =
n=1 {rn },
Q is a F -set.
(c) By (a), L (Q) = 0. This implies that, for every n N, there exists an open set
Gn such that
1
Gn Q and L (Gn ) < .
n
T
If G = n=1 Gn then G is a G -set and G Q. Furthermore,
L (G) L (Gn ) <
1
, n N.
n
This implies that L (G) = 0.

(d) By (b), Q is a F -set, so R \ Q, the set of all irrational numbers in R, is a
G -set.
Problem 20
Let E ML with L (E) > 0. Prove that for every (0, 1), there exists a
finite open interval I such that
L (I) L (E I) L (I).

27
Solution
Consider first the case where 0 < L (E) < . For any (0, 1), set 1 = 1 + a.
Since a > 0, 0 < = aL (E) < . By the regularity property of L (Property 7),
there exists an open set O E such that1
L (O) L (E) + aL (E) = (1 + a)L (E) =
1
L (E) < .
(i)
Now since O is an open set in R, it is union of a disjoint sequence (In ) of open

intervals in R:
X
[
O=
In = L (O) =
L (In ). (ii)
nN
nN
Since E O, we have
L (E) = L (E O) = L
[
nN
!
In
L (E In ).
(iii)
nN
From (i), (ii) and (iii) it follows that

X
nN
L (In )
1X
L (E In ).
nN
Note that all terms in this inequality are positive, so that there exists at least one
n0 N such that
1
L (In0 ) L (E In0 ).
Since L (O) is finite, all intervals In are finite intervals in R. Let I := In0 , then I is
a finite open interval satisfying conditions:
L (I) L (E I) L (I).
Now consider that case L (E) = . By the -finiteness of the Lebesgue measure
space, there exists a measurable subset E0 of E such that 0 < L (E0 ) < . Then
using the first part of the solution, we obtain
L (I) L (E0 I) L (E I) L (I).
Recall that for (Lebesgue) measurable set A, L (A) = L (A).

28
Problem 21
Let f be a real-valued function on (a, b) such that f 0 exists and satisfies
|f 0 (x)| M for all x (a, b) and for some M 0.
Show that for every E (a, b) we have
L (f (E)) M L (E).
Solution
If M = 0 then f 0 (x) = 0, x (a, b). Hence, f (x) = y0 , x (a, b). Thus, for any
E (a, b) we have
L (f (E)) = 0.
The inequality holds. Suppose M > 0. For all x, y (a, b), by the Mean Value
Theorem, we have
|f (x) f (y)| = |x y||f 0 (c)|, for some c (a, b)
M |x y|. ()
By definition of the outer measure, for any E (a, b) we have
L (E)
X
= inf
(bn an ),
n=1
where {In = (an , bn ), n N} is a covering class of E. By (*) we have
|f (bn ) f (an )| M
|bn an |
n=1
n=1
M inf
|bn an |
n=1
M L (E).
Infimum takes over all covering classes of E. Thus,
L (f (E))
= inf
|f (bn ) f (an )| M L (E).
n=1

29
Problem 22
(a) Let E R. Show that F = {, E, E c , R} is the -algebra of subsets of R
generated by {E}
(b) If S and T are collections of subsets of R, then
(S T ) = (S) (T ).
Is the statement true? Why?
Solution
(a)It is easy to check that F is a -algebra.
Note first that {E} F . Hence
({E}) F .
(i)
On the other hand, since ({E}) is a -algebra, so , R ({E}). Also, since

E ({E}), so E c ({E}). Hence
F ({E}).
(ii)
From (i) and (ii) it follows that

F = ({E}).
(b) No. Here is why.
Take S = {(, 1]} and T = {(1, 2]}. Then, by part (a),
(S) = {, (0, 1], (0, 1]c , R} and (T ) = {, (1, 2], (1, 2]c , R}.
Therefore
(S) (T ) = {, (0, 1], (0, 1]c , (1, 2], (1, 2]c , R}.
We have
(0, 1] (1, 2] = (0, 2]
/ (S) (T ).
Hence (S) (T ) is not a -algebra. But, by definition, (S T ) is a -algebra.
And hence it cannot be equal to (S) (T ).
Problem 23
Consider F = {E R : either E is countable or E c is countable}.
(a) Show that F is a -algebra and F is a proper sub--algebra of the -algebra
BR .
(b) Show that F is the -algebra generated by {x} : x R .

(c) Find a measure : F [0, ] such that the only -null set is .

30
Solution
(a) We check conditions of a -algebra:
It is clear that is countable, so F.
Suppose E F . Then E R and E is countable or E c is countable. This is
equivalent to E c R and E c is countable or E is countable. Thus,
E F E c F.
S
Suppose
E
,
E
,
...
F.
Either
all
E
s
are
countable,
so
1
2
n
n=1 En is countable.
S
Hence
E
F
.
Or
there
exists
some
E
F
which
is
not countable. By
n0
n=1 n
c
definition, En0 must be countable. Now
!c
[
\
En =
Enc En0 .
n=1
This implies that (
S
n=1
n=1
En ) is countable. Thus
En F.
n=1
Finally, F is a -algebra.
Recall that BR is the -algebra generated by the family of open sets in R. It is also
generated by the family of closed sets in R. Now suppose E F . If E is countable
then we can write
[
E = {x1 , x2 , ...} =
{xn }.
n=1
Each {xn } is a closed set in R, so belongs to BR . Hence E BR . Therefore,

F BR .
F is a proper subset of BR . Indeed, [0, 1] BR and [0, 1]
/ F.
(b) Let S = {x} : x R . Clearly, S F , and so
(S) F .
Now take E F and E 6= . If E is countable then we can write
E=
[
n=1
{xn } (S).
| {z }
S
Hence
F (S).

31
Thus
(S) = F.
(c) Define the set function : F [0, ] by
(
|E| if E is finite
(E) =
otherwise.
We can check that is a measure. If E 6= then (E) > 0 for every E F.
Problem 24
For E ML with L (E) < , define a real-valued function E on R by setting
E (x) := L (E (, x]) for x R.
(a) Show that E is an increasing function on R.
(b) Show that E satisfies the Lipschitz condition on R, that is,
|E (x0 ) E (x00 )| |x0 x00 | for x0 , x00 R.
Solution
(a) Let x, y R. Suppose x < y. It is clear that (, x] (, y]. Hence,
E (, x] E (, y] for E ML . By the monoticity of L we have
E (x) = L (E (, x]) L (E (, y]) = E (y).
Thus E is increasing on R.
(b) Suppose x0 < x00 we have
E (x0 , x00 ] = (E (, x00 ]) \ (E (, x0 ]).
It follows that
E (x00 ) E (x0 ) = L (E (, x00 ]) L (E (, x0 ])
= L (E (x0 , x00 ])
L (x0 , x00 ] = x00 x0 .

32
Problem 25
Let E be a Lebesgue measurable subset of R with L (E) = 1. Show that there
exists a Lebesgue measurable set A E such that L (A) = 12 .
Solution
Define the function f : R [0, 1] by
f (x) = L E (, x] , x R.
By Problem 23, we have
|f (x) f (y)| |x y|, x, y R.
Hence f is (uniformly) continuous on R. Since
lim f (x) = 0 and
lim f (x) = 1,
by the Mean Value Theorem, we have

1
x0 R such that f (x0 ) = .
2
Set A = E (, x0 ]. Then we have
1
A E and L (A) = .
2
Chapter 3
Measurable Functions
Remark:
From now on, measurable means Lebesgue measurable. Also measure means Lebesgue measure,
and we write instead of L for Lebesgue measure.
1. Definition, basic properties
Proposition 6 (Equivalent conditions)

Let f be an extended real-valued function whose domain D is measurable. Then the following
statements are equivalent:
1. For each real number a, the set {x D : f (x) > a} is measurable.
2. For each real number a, the set {x D : f (x) a} is measurable.
3. For each real number a, the set {x D : f (x) < a} is measurable.
4. For each real number a, the set {x D : f (x) a} is measurable.
Definition 11 (Measurable function)
An extended real-valued function f is said to be measurable if its domain is measurable and if it
satisfies one of the four statements of Proposition 6.
Proposition 7 (Operations)
Let f, g be two measurable real-valued functions defined on the same domain and c a constant.
Then the functions f + c, cf, f + g, and f g are also measurable.
Note:
A function f is said to be Borel measurable if for each R the set {x : f (x) > } is a Borel set.
Every Borel measurable function is Lebesgue measurable.
2. Equality almost everywhere
A property is said to hold almost everywhere (abbreviated a.e.) if the set of points where it fails
to hold is a set of measure zero.
We say that f = g a.e. if f and g have the same domain and {x D : f (x) 6= g(x)} = 0.
Also we say that the sequence (fn ) converges to f a.e. if the set {x : fn (x) 9 f (x)} is a null set.

34
CHAPTER 3. MEASURABLE FUNCTIONS
Proposition 8 (Measurable functions)

If a function f is measurable and f = g a.e., then g is measurable.
3. Sequence of measurable functions
Proposition 9 (Monotone sequence)

Let (fn ) be a monotone sequence of extended real-valued measurable functions on the same measurable domain D. Then limn fn exists on D and is measurable.
Proposition 10 Let (fn ) be a sequence of extended real-valued measurable functions on the same
measurable domain D. Then max{f1 , ..., fn }, min{f1 , ..., fn }, lim supn fn , lim inf n fn , supnN , inf nN
are all measurable.
Proposition 11 If f is continuous a.e. on a measurable set D, then f is measurable.
Problem 26
Let D be a dense set in R. Let f be an extended real-valued function on R such
that {x : f (x) > } is measurable for each D. Show that f is measurable.
Solution
Let be an arbitrary real number. For each n N, there exists n D such that
< n < + n1 by the density of D. Now
{x : f (x) > } =

[
n=1
1
x : f (x) +
n
{x : f (x) > n }.
n=1
S
Since
n=1 {x : f (x) > n } is measurable (as countable union of measurable sets),
{x : f (x) > } is measurable. Thus, f is measurable.
Problem 27
Let f be an extended real-valued measurable function on R. Prove that {x :
f (x) = } is measurable for any R.

35
Solution
For R, we have
{x : f (x) = } = {x : f (x) } \ {x : f (x) < } .
|
{z
} |
{z
}
measurable
measurable
Thus {x : f (x) = } is measurable.

For = , we have
{x : f (x) = } = R \ {x : f (x) < } = R \
[
nN
{x : f (x) n} .
|
{z
}
measurable

For = , we have
{x : f (x) = } = R \ {x : f (x) > } = R \
[
nN
{x : f (x) n} .
{z
}
|
measurable
Problem 28
(a). Let D and E be measurable sets and f a function with domain D E. Show
that f is measurable if and only if its restriction to D and E are measurable.
(b). Let f be a function with measurable domain D. Show that f is measurable
if and only if the function g defined by
(
f (x) for x D
g(x) =
0
for x
/D
is measurable.
Solution
(a) Suppose that f is measurable. Since D and E are measurable subsets of D E,
the restrictions f |D and f |E are measurable.
Conversely, suppose f |D and f |E are measurable. For any R, we have
{x : f (x) > } = {x D : f |D (x) > } {x E : f |E (x) > }.
Each set on the right hand side is measurable, so {x : f (x) > } is measurable.
Thus, f is measurable.

36
(b) Suppose that f is measurable. If 0, then {x : g(x) > } = {x : f (x) > },

which is measurable. If < 0, then {x : g(x) > } = {x : f (x) > } Dc , which
is measurable. Hence, g is measurable.
Conversely, suppose that g is measurable. Since f = g|D and D is measurable, f is
measurable.
Problem 29
Let f be measurable and B a Borel set. Then f 1 (B) is a measurable set.
Solution
Let C be the collection of all sets E such that f 1 (E) is measurable. We show that
C is a -algebra. Suppose E C. Since
c
f 1 (E c ) = f 1 (E) ,
which is measurable, so E c C. Suppose (En ) is a sequence of sets in C. Since
!
[
[
f 1 (En ),
En =
f 1
n
S
which is measurable, so n En C. Thus, C is a -algebra.
Next, we show that all intervals (a, b), for any extended real numbers a, b with
a < b, belong to C. Since f is measurable, {x : f (x) > a} and {x : f (x) < b} are
measurable. It follows that (a, ) and (, b) C. Furtheremore, we have
(a, b) = (, b) (a, ),
so (a, b) C. Thus, C is a -algebra containing all open intervals, so it contains all
Borel sets. Hence f 1 (B) is measurable.
Problem 30
Show that if f is measurable real-valued function and g a continuous function
defined on R, then g f is measurable.
Solution
For any R,

{x : (g f )(x) > } = (g f )1 ((, )) = f 1 g 1 (, ) .

37
By the continuity of g, g 1 (, ) is an open set, so a Borel set. By Problem 24,

the last set is measurable. Thus, g f is measurable.
Problem 31
Let f be an extended real-valued function defined on a measurable set D R.
(a) Show that if {x D : f (x) < r} is measurable in R for every r Q, then
f is measurable on D.
(b) What subsets of R other than Q have this property?
(c) Show that if f is measurable on D, then there exists a countable sub-collection
C ML , depending on f , such that f is (C)-measurable on D.
(Note: (C) is the -algebra generated by C.)
Solution
(a) To show that f is measurable on D, we show that {x D : f (x) < a} is
measurable for every a R. Let I = {r Q : r < a}. Then I is countable , and
we have
[
{x D : f (x) < a} = {x D : f (x) < r}.
rI
Since {x D : f (x) < r} is measurable,

Thus, {x D : f (x) < a} is measurable.
rI {x
D : f (x) < r} is measurable.
(b) Here is the answer to the question:

Claim 1 : If E R is dense in R, then E has the property in (a), that is, if
{x D : f (x) < r} is measurable for every r E then f is measurable on D.
Proof.
Given any a R, the interval (a 1, a) intersects E since E is dense. Pick some
r1 E (a 1, a). Now the interval (r1 , a) intersects E for the same reason. Pick
some r2 E (r1 , a). Repeating this process, we obtain an increasing sequence (rn )
in E which converges to a.
By assumption, {x D : f (x) < rn } is measurable, so we have
{x D : f (x) < a} =
{x D : f (x) < rn } is measurable .
nN
Thus, f is measurable on D.
Claim 2 : If E R is not dense in R, then E does not have the property in (a).
Proof.
Since E is not dense in R, there exists an interval [a, b] E. Let F be a non

38
measurable set in R. We define a function f as follows:

(
a if x F c
f (x) =
b if x F .
For r E, by definition of F , we observe that
If r < a then f 1 ([, r)) = .
If r > b then f 1 ([, r)) = R.
If r =
a+b
2
then f 1 ([, r)) = F c .
Since F is non measurable, F c is also non measurable. Through the above observation, we see that
a+b
non measurable.
x D : f (x) <
2
Thus, f is not measurable.
Conclusion : Only subsets of R which are dense in R have the property in (a).
(c) Let C = {Cr }rQ where Cr = {x D : f (x) < r} for every r Q. Clearly, C is
a countable family of subsets of R. Since f is measurable, Cr is measurable. Hence,
C ML . Since ML is a -algebra, by definition, we must have (C) ML . Let
a R. Then
[
[
{x D : f (x) < a} =
{x D : f (x) < r} =
Cr .
r<a
r<a
It follows that {x D : f (x) < a} (C).

Thus, f is (C)-measurable on D.
Problem 32
Show that the following functions defined on R are all Borel measurable, and
hence Lebesgue measurable also on R:
(
(
0 if x is rational
x
if x is rational
(a) f (x) =
(b) g(x) =
1 if x is irrational.
x if x is irrational.
(
(c) h(x) =
sin x
cos x
if x is rational
if x is irrational.

39
Solution
(a) For any a R, let E = {x D : f (x) < a}.
If a > 1 then E = R, so E BR (Borel measurable).
If 0 < a 1 then E = Q, so E BR (Borel measurable).
If a 0 then E = , so E BR (Borel measurable).
Thus, f is Borel measurable.
(b) Consider g1 defined on Q by g1 (x) = x, then g|Q = g1 . Consider g2 defined on
R\Q by g(x) = x, then g|R\Q = g2 . Notice that R, R\Q BR (Borel measurable).
For any a R, we have
{x D : f1 (x) < a} = [, a) Q BR (Borel measurable),
and
{x D : f2 (x) < a} = [, a) (R \ Q) BR
(Borel measurable).
Thus, g is Borel measurable.

(c) Use the same way as in (b).
Problem 33
Let f be a real-valued increasing function on R. Show that f is Borel measurable,
and hence Lebesgue measurable also on R.
Solution
For any a R, let E = {x D : f (x) a}. Let = inf E. Since f is increasing,
if Im(f ) (, a) then E = .
if Im(f ) * (, a) then E is either (, ) or [, ).
Since , (, ), [, ) are Borel sets, so f is Borel measurable.
Problem 34
If (fn ) is a sequence of measurable functions on D R, then show that
{x D : lim fn (x) exists} is measurable.
n

40
Solution
Recall that if fn s are measurable, then lim supn fn , lim inf n fn and g(x) =
lim supn fn lim inf n fn are also measurable, and if h is measurable then
{x D : h(x) = } is measurable (Problem 22).
Now we have
E = {x D : lim fn (x) exists} = {x D : g(x) = 0}.
n
Thus, E is measurable.
Problem 35
(a) If g : R R is continuous and f : R R is measurable then g f is
measurable.
(b) If f is measurable then |f | is measurable. Does the converse hold?
Solution
(a) For any a R, then
E = {x : (g f )(x) < a} = (g f )1 (, a)
= f 1 g 1 (, a) .
Since g is continuous, g 1 (, a) is open.SThen there is a family of open disjoint
intervals {In }nN such that g 1 (, a) = nN In . Hence,
!
[
[
E = f 1
In =
f 1 (In ).
nN
nN
Since f is measurable, f 1 (In ) is measurable. Hence E is measurable. This tells us

that g f is measurable.
(b) If g(u) = |u| then g is continuous. We have
(g f )(x) = g(f (x)) = |f (x)|.
By part (a), g f = |f | is measurable.
The converse is not true.
Let E be a non-measurable subset of R. Consider the function:
(
1
if x E
f (x) =
1 if x
/ E.
Then f 1 ( 12 , ) = E, which is not measurable. Since ( 21 , ) is open, so f is not
measurable, while |f | = 1 is measurable.

41
Problem 36
Let (fn : n N) and f be an extended real-valued measurable functions on a
measurable set D R such that limn fn = f on D. Then for every R
prove that:
(i) {x D : f (x) > } lim inf {x D : fn (x) }
n
(ii) {x D : f (x) < } lim inf {x D : fn (x) }.

n
Solution
Recall that, for any sequence (En )nN of measurable sets,
(lim inf En ) lim inf (En ), ()
n
n
[ \
\
Ek = lim
Ek .
lim inf En =
n
nN kn
kn
Now for every R, let Ek = {x D : fk (x) } for each k N. Then

\
lim inf En = lim
Ek
n
lim
kn
{x D : fk (x) }
kn
= {x D : f (x) > } since fk (x) f (x) on D.

Using (*) we get
{x D : f (x) > } lim inf {x D : fn }.
n
For the second inequality, we use the similar argument.

Let Fk = {x D : fk (x) } for each k N. Then
\
lim inf En = lim
Fk
n
lim
kn
{x D : fk (x) }
kn
= {x D : f (x) < } since fk (x) f (x) on D.

Using (*) we get
{x D : f (x) < } lim inf {x D : fn }.
n

42
Simple functions
Definition 12 (Simple function)
A function : X R is simple if it takes only a finite number of different values.
Definition 13 (Canonical representation )
Let be a simple function on X. Let {a1 , ..., an } the set of distinct valued assumed
by on D. Let Di = {x X : (x) = ai } for i = 1, ..., n. Then the expression
=
n
X
ai Di
i=1
is called the canonical representation of .

It is evident that Di Dj = for i 6= j and
Sn
i=1
Di = X.
Problem 37
(a). Show that
AB = A B
AB = A + B A B
Ac = 1 A .
(b). Show that the sum and product of two simple functions are simple functions.
Solution
(a). We have
AB (x) = 1 x A and x B
A (x) = 1 = B (x).
Thus,
AB = A B .
We have
AB (x) = 1 x A B.

43
If x A B then A (x) + B (x) A (x) B (x) = 1 + 1 1 = 1.
If x
/ A B, then x A \ B or x B \ A. Then A (x) + B (x) = 1 and
A B A (x) + B (x) = 0.
Also,
AB (x) = 0 x
/ A B.
Then
A (x) = B (x) = A (x) B (x) = 0.
Thus,
AB = A + B A B .
If Ac (x) = 1, then x
/ A, so A (x) = 0.
If Ac (x) = 0, then x A, so A (x) = 1. Thus,
Ac = 1 A .
(b). Let be a simple function having values a1 , ..., an . Then

n
X
ai Ai where Ai = {x : (x) = ai }.
i=1
Similarly, if is a simple function having values b1 , ..., bm . Then

=
m
X
bj Bj where Bj = {x : (x) = bj }.
j=1
Define Cij := Ai Bj . Then

Ai X =
m
[
Bj and so Ai = Ai
j=1
m
[
Bj =
j=1
Similarly, we have
Bj =
n
[
m
[
j=1
Cij .
i=1
Since the Cij s are disjoint, this means that (see part (a))
Ai =
m
X
Cij and Bj =
j=1
Thus
=
n X
m
X
i=1 j=1
ai Cij and =
n
X
Cij .
i=1
n X
m
X
bj Cij .
i=1 j=1
Cij .

44
Hence

n X
m
n X
m
X
X
+ =
(ai + bj )Cij and =
ai bj Cij .
i=1 j=1
They are simple function.
i=1 j=1
Problem 38
Let : R R be a simple function defined by
n
X
ai Ai where Ai = {x R : (x) = ai }.
i=1
Prove that is measurable if and only if all the Ai s are measurable.

Solution
Assume that Ai is measurable for all i = 1, ..., n. Then for any c R, we have
[
{x : (x) > c} =
Ai .
ai >c
S
Since every Ai is measurable, ai >c Ai is measurable. Thus {x : (x) > c} is
measurable. By definition, is measurable.
Conversely, suppose is measurable. We can suppose a1 < a2 < ... < an . Given
j {1, 2, ..., n}, choose c1 and c2 such that aj1 < c1 < aj < c2 < aj+1 . (If j = 1 or
j = n, part of this requirement is empty.) Then
!
!
[
[
Aj =
Ai \
Ai
ai >c1
ai >c2
= {x : (x) > c1 } \ {x : (x) > c2 } .

|
{z
} |
{z
}
measurable
Thus, Aj is measurable for all j {1, 2, ..., n}.
measurable
Chapter 4
Convergence a.e. and Convergence
in Measure
1. Convergence almost everywhere
Definition 14 Let (fn ) be a sequence extended real-valued measurable functions on a measurable
set D R.
1. We say that limn fn exists a.e. on D if there exists a null set N such that N D and
limn fn (x) exists for every x D \ N .
2. We say that (fn ) converges a.e. on D if limn fn (x) exists and limn fn (x) R for every
x D \ N.
Proposition 12 (Uniqueness)
Let (fn ) be a sequence extended real-valued measurable functions on a measurable set D R. Let
g1 and g2 be two extended real-valued measurable functions on D. Then
h
i
lim fn = g1 a.e. on D and lim fn = g2 a.e. on D = g1 = g2 a.e. on D.
n
Theorem 1 (Borel-Cantelli Lemma)

For any sequence (An ) of measurable subsets in R, we have
X
(An ) < = lim sup An = 0.

nN
Definition 15 (Almost uniform convergence)

Let (fn ) be a sequence extended real-valued measurable functions on a measurable set D R and
f a real-valued measurable functions on D. We say that (fn ) converges a.u. on D to f if for every
> 0 there exists a measurable set E D such that (E) < and (fn ) converges uniformly to f
on D \ E.
Theorem 2 (Egoroff )
Let D be a measurable set with (D) < . Let (fn ) be a sequence extended real-valued measurable
functions on D and f a real-valued measurable functions on D. If (fn ) converges to f a.e. on D,
then (fn ) converges to f a.u. on D.

46 CHAPTER 4. CONVERGENCE A.E. AND CONVERGENCE IN MEASURE
2. Convergence in measure
Definition 16 Let (fn ) be a sequence extended real-valued measurable functions on a measurable
set D R. We say that (fn ) converges in measure on D if there exists a real-valued measurable
function f on D such that for every > 0 we have
lim {D : |fn f | } := lim {x D : |fn (x) f (x)| } = 0.
That is,
> 0, > 0, N (, ) N : {D : |fn f | } < for n N (, ).
We write fn
f on D for this convergence.
f and g be two real-valued measurable functions on D. Then
[fn
f on D and fn
g on D] = f = g a.e. on D.
Proposition 14 (Equivalent conditions)
(1) [fn
f on D] > 0, N () N : {D : |fn f | } < for n N ().
n
1
1o
<
for m N (m).
(2) [fn
f on D] m N, N (m) : D : |fn f |
m
m
3. Convergence a.e. and convergence in measure
Theorem 3 (Lebesgue)
f be a real-valued measurable functions on D. Suppose
1. fn f a.e. on D,
2. (D) < .
Then fn
f on D.
Theorem 4 (Riesz)
f be a real-valued measurable functions on D. If fn

f on D, then there exists a subsequence
(fnk ) which converges to f a.e. on D.

47
Problem 39(An exercise to warn up.)

1. Consider the sequence (fn ) defined on R by fn = [n,n+1] , n N and the
function f 0. Does (fn ) converge to f a.e.? a.u.? in measure?
2. Same questions with fn = n[0, 1 ] .
n
(Note: A is the characteristic function of the set A. Try to write your solution.)
Problem 40
Let (fn ) be a sequence of extended real-valued measurable functions on X and
let f be an extended real-valued function which is finite a.e. on X. Suppose
limn fn = f a.e. on X. Let [0, (X)) be arbitrarily chosen. Show that
for every > 0 there exists N N such that {X : |fn f | < } for
n N.
Solution
Let Z be a null set such that f is finite on X \ Z. Since fn f a.e. on X, fn f
a.e. on X \ Z. For every > 0 we have1
(lim sup{X \ Z : |fn f | }) = 0
n
lim sup {X \ Z : |fn f | } = 0

n
lim {X \ Z : |fn f | } = 0
n
The last condition is equivalent to

lim {X \ Z : |fn f | < } = (X \ Z) = (X)
> 0, N N : (X) {X \ Z : |fn f | < } for all n N.

Let us take = (X) > 0. Then we have
N N : {X \ Z : |fn f | < } for all n N.
Since {X : |fn f | < } {X \ Z : |fn f | < }, so we have
> 0, N N : n N ({X : |fn f | < }) .
1
See Problem 11b. We have

n

Problem 41
(a) Show that the condition
lim {x D : |fn (x) f (x)| > 0} = 0
implies that fn
f on D.
(b) Show that the converse is not true.
(c) Show that the condition in (a) implies that for a.e. x D we have fn (x) =
f (x) for infinitely many n N.
Solution
(a) Given any > 0, for every n N, let
En = {x D : |fn (x) f (x)| > };
Fn = {x D : |fn (x) f (x)| > 0}.
Then we have for all n N

x En |fn (x) f (x)| >
|fn (x) f (x)| > 0
x Fn .
Consequently, En Fn and (En ) (Fn ) for all n N. By hypothesis, we have
that limn (Fn ) = 0. This implies that limn (En ) = 0. Thus, fn

f.
(b) The converse of (a) is false.
Consider functions:
1
fn (x) = , x [0, 1] n N.
n
f (x) = 0, x [0, 1].
Then fn f (pointwise) on [0, 1]. By Lebesgue Theorem fn

f on [0, 1]. But for
every n N
1
|fn (x) f (x)| = > 0, x [0, 1].
n
In other words,
{x D : |fn (x) f (x)| > 0} = [0, 1].
Thus,
lim {x D : |fn (x) f (x)| > 0} = 1 6= 0.
(c) Recall that (Problem 11a)

(lim inf En ) lim inf (En ).
n
()

49
Let En = {x D : fn (x) 6= f (x)} and E = lim inf n En . By (a),
lim inf (En ) = lim (En ) = 0.
n
Therefore, by (), (E) = 0. By definition, we have

[ \
E=
Ek .
nN kn
Hence, x
/ E whenever x Enc for infinitely many ns, that is fn (x) = f (x) a.e. in
D for infinitely many ns.
Problem 42
Suppose fn (x) fn+1 (x) for all n N and x D \ Z with (Z) = 0. If fn

f
on D , then prove that fn f a.e. on D.
Solution
Let B = D \ Z. Since fn
f on D, fn
f on B. Then, By Riesz theorem, there
exists a sub-sequence (fnk ) of (fn ) such that fnk f a.e. on B.
Let C = {x B : fnk 9 f }. Then (C) = 0 and fnk f on B \ C.
From fn (x) fn+1 (x) for all n N, and since nk k, we get fk fnk for all k N.
Therefore
|fk f | |fnk f |.
This implies that fk f on B \ C. Since B \ C = D \ (Z C) and (Z C) = 0,
it follows that fn f a.e. on D .
Problem 43
Show that if fn
f on D and gn
g on D then fn + gn
f + g on D.
Solution
Since fn
f and gn
g on D, for every > 0,
(4.1)
(4.2)
lim {D : |fn f | } = 0
n
2
lim {D : |gn g| } = 0.
n
2
Now
|(fn + gn ) (f + g)| |fn f | + |gn g|.

By the triangle inequality above, if |(fn + gn ) (f + g)| is true, then at least
one of the two following inequalities must be true:
|fn f |
or |gn g| .
2
2
Hence
n
o n
o
{D : |(fn + gn ) (f + g)| } D : |fn f |
D : |gn g|
.
2
2
Therefore,
n
n
o
o
{D : |(fn + gn ) (f + g)| } D : |fn f |
+ D : |gn g|
.
2
2
From (4.1) and (4.2) we obtain
lim {D : |(fn + gn ) (f + g)| } = 0.
That is, by definition, fn + gn

f + g on D.
Problem 44
Show that if fn
f on D and gn
g on D and (D) < , then fn gn
fg
on D.
(Assume that both fn and gn are real-valued for every n N so that the multiplication fn gn
is possible.)
Solution
For every > 0 and > 0, we want {|fn gn f g| } < for n large enough.
Notice that
()
|fn gn f g| |fn gn f gn | + |f gn f g| |fn f ||gn | + |f ||gn g|.
For any N N, let

EN = {D : |f | > N } {D : |g| > N }.
It is clear that EN EN +1 for every N N. Since (D) < , we have
!
\
lim (EN ) =
EN = () = 0.
N
N N

51
It follows that, we can take N large enough to get, for every > 0,
()
Observe that
< 1 and (EN ) < .

2N
3
n
o
{D : |gn | > N + 1} D : |gn g|
EN
2N
(since |gn | |gn g| + |g|). Now if we have

|fn f |
; |gn | > N + 1; |gn g|

, and |f | > N,
2(N + 1)
2N
then (*) implies

n
{D : |fn gn f g| }
D : |fn f |
EN
2(N + 1)
n
o
D : |gn g|
{D : |gn | > N + 1}.
2N
By assumption, given > 0, > 0, for n > N , we have
n
o
D : |fn f |
<
2(N + 1)
3
n
o
D : |gn g|
< .
2N
3
From these results, from (*), and (**) we get
{D : |fn gn f g| } <

+ + = .
3 3 3
Problem 45
(a) Definition of Almost uniform convergence (a.u).
(b) Show that if fn f a.u on D then fn

f on D.
(c) Show that if fn f a.u on D then fn f a.e. on D.
Solution
(a) > 0, E D such that (E) < and fn f uniformly on D \ E.
(b) Suppose that fn f a.u on D and fn does not converges to f in measure on
D. Then there exists an 0 > 0 such that
{x D : |fn (x) f (x)| > 0 } 9 0 as n .

We can choose n1 < n2 < ... such that
{x D : |fnk (x) f (x)| > 0 } r for some r > 0 and k N.
Now since fn f a.u on D,
E D such that (E) <
r
and fn f uniformly on D \ E.
2
Let C = {x D : |fnk (x) f (x)| > 0 } k N. Then (C) r. Since

fn f uniformly on D \ E,
N : n N |fn (x) f (x)| 0 , x D \ E.
Thus,
C (D \ E)c = E.
Hence,
r
0 < r (C) (E) < .
2
This is a contradiction.
(c) Since fn f a.u. on D, for every n N, there exists
En D such that
T
(En ) < n1 and fn f uniformly on D \ En . Let E = nN En , then (E) = 0.
Since fn f on D \ En for every n N, fn f on
[
\
(D \ En ) = D \
En = D \ E.
nN
Since (E) = 0, fn f a.e. on D
nN
Chapter 5
Integration of Bounded Functions
on Sets of Finite Measure
In this chapter we suppose (D) < .
1. Integration of simple functions
Definition 17 (Lebesgue integral of simple
Pn functions)
Let be a simple function on D and = i=1 ai Di be its canonical representation. The Lebesgue
integral of on D is defined by
Z
(x)(dx) =
D
n
X
ai (Di ).
i=1
We usually use simple notations for the integral of :

Z
Z
Z
d,
(x)dx or
.
D
If
R
D
d < , then we say that is integrable on D.
Proposition 15 (properties of integral of simple functions)

1. (D) = 0
2.
3.
4.
5.
6.
7.
d = 0.
R
R
0, E D E d D d.
R
R
cd = c D d.
D
R
Pn R
d = i=1 Di d.
D
R
R
cd = c D d (c is a constant).
D
R
R
R
(1 + 2 )d = D 1 d + D 2 d.
D
R
R
1 = 2 a.e. on D D 1 d = D 2 d.
D

54CHAPTER 5. INTEGRATION OF BOUNDED FUNCTIONS ON SETS OF FINITE MEASURE
2. Integration of bounded functions
Definition 18 (Lebesgue integral of bounded functions)

Let f be a bounded real-valued measurable function on D. Let be the collection of all simple
functions on D. We define the Lebesgue integral of f on D by
Z
Z
Z
f d = inf
d = sup
d where , .
D
If
R
D
f d < , then we say that f is integrable on D.
Proposition 16 (properties of integral of bounded functions)
1.
2.
3.
4.
5.
6.
cf d = c
f d.
R
R
(f + g)d = D f d + D gd.
D
R
R
f = g a.e. on D D f d = D gd.
R
R
f g on D D f d D gd.
R
R
|f | M on D D f d D |f |d M (D).
R
f 0 a.e. on D and D f d = 0 f = 0 a.e. on D.
D
7. If (Dn ) be a disjoint sequence of measurable subset Dn D with

Z
fd =
D
XZ
nN
S
nN
Dn = D then
f d.
Dn
Theorem 5 (Bounded convergence theorem)

Suppose that (fn ) is a uniformly bounded sequence of real-valued measurable functions on D, and
f is a bounded real-valued measurable function on D. If fn f a.e. on D, then
Z
lim
|fn f |d = 0.
n
In particular,
Z
lim
Z
fn d =
f d.
D

55
Problem 46
Let f be an extended real-valued measurable function on a measurable set D. For
M1 , M2 R, M1 < M2 , let the truncation of f at M1 and M2 be defined by
if f (x) < M1
M1
g(x) = f (x) if M1 f (x) M2
M2
if f (x) > M2 .
Show that g is measurable on D.
Solution
Let a R. We need to show that the set E = {x D : g(x) > a} is measurable.
There are three cases to consider:
1. If a M2 then E = which is measurable.
2. If a < M1 then E = D which is measurable.
3. If M1 a < M2 then E = {x D : f (x) > a} which is measurable.
Thus, in all three cases E is measurable, so g is measurable.
Problem 47
Given a measure space (X, A, ). Let f be a bounded real-valued A-measurable
function on D A with (D) < . Suppose |f (x)| M, x D for some
constant M > 0. R
(a) Show that if D f d = M (D), then f = M a.e. on D.
R
(b) Show that if f < M a.e. on D and if (D) > 0, then D f d < M (D).
Solution
(a) For every n N, let En = {x D : f (x) < M n1 }. Then, since f M on
D \ En , we have
Z
Z
Z
f d =
f d +
f d
D
En
D\En
M
(En ) + M (D \ En ).
n

Since En D, we have
(D \ En ) = (D) (En ).
Therefore,
Z
f d
D
1
M
n
(En ) + M (D) M (En )
1
(En ).
n
R
By assumption D f d = M (D), it follows that
= M (D)
1
0 (En ) 0, n N,
n
which implies (E
S n ) = 0, n N.
Now let E = n=1 En then E = {x D : f (x) < M }. We want to show that
(E) = 0. We have
X
0 (E)
(En ) = 0.
n=1
Thus, (E) = 0. Since |f | M , the last result implies f = M a.e. on D.

R
(b)
First
we
note
that
|f
|
M
on
D
implies
that
f d M (D). Assume that
D
R
f
d
=
M
(D).
By
part
(a)
we
have
f
=
M
a.e.
on
D. This contradicts the fact
D
R
that f < M a.e. on D. Thus D f d < M (D).
Problem 48
Consider a sequence of functions (fn )nN defined on [0, 1] by
fn (x) =
nx
for x [0, 1].
1 + n 2 x2
(a) Show that (fn ) is uniformly bounded on [0, 1] and evaluate

Z
nx
lim
d.
n [0,1] 1 + n2 x2
(b) Show that (fn ) does not converge uniformly on [0, 1].
Solution
(a) For all n N, for all x [0, 1], we have 1 + n2 x2 2nx 0 and 1 + n2 x2 > 0,
hence
1
nx
.
0 fn (x) =
2
2
1+n x
2

57
Thus, (fn ) is uniformly bounded on [0, 1].
Since each fn is continuous on [0, 1], f is Riemann integrable on [0, 1]. In this case,
Lebesgue integral and Riemann integral on [0, 1] coincide:
Z
Z 1
nx
nx
d
=
dx
2 2
2 2
[0,1] 1 + n x
0 1+n x
Z 1+n2
1
1
=
dt (with t = 1 + n2 x2 )
2n 1
t
ln(1 + n2 )
1
ln(1 + n2 ) =
.
=
2n
2n
Using LHospital rule we get limx
Z
lim
n
ln(1+x2 )
2x
[0,1]
= 0. Hence,
nx
d = 0.
1 + n 2 x2
(b) For each x [0, 1],
nx
= 0.
n 1 + n2 x2
Hence, fn f 0 pointwise on [0, 1]. To show fn does not converge to f 0
uniformly on [0, 1], we find a sequence (xn ) in [0, 1] such that xn 0 and fn (xn ) 9
f (0) = 0 as n . Indeed, take xn = n1 . Then fn (x) = 12 . Thus,
lim
lim fn (xn ) =
1
6= f (0) = 0.
2
Problem 49
Let (fn )nN and f be extended real-valued measurable functions on D ML with
(D) < and assume that f is real-valued a.e. on D. Show that fn

f on D
if and only if
Z
|fn f |
lim
d = 0.
n D 1 + |fn f |
Solution
Suppose fn
f on D. By definition of convergence in measure, for any > 0,
there exists an N N such that for n N ,
En D : (En ) <
and |fn f | <

on D \ En .
2
2(D)

For n N we have
Z
Z
Z
|fn f |
|fn f |
|fn f |
()
d =
d +
d.
D 1 + |fn f |
En 1 + |fn f |
D\En 1 + |fn f |
Note that for all n N, we have 0
0
|fn f |
1+|fn f |
1 on En and
|fn f |
1
= |fn f |
|fn f |
on D \ En .
1 + |fn f |
1 + |fn f |
2(D)
So for n N , we can write (*) as

Z
Z
Z
|fn f |
0
d
1 d +
d
D 1 + |fn f |
En
D\En 2(D)
= (En ) +
(D \ En )
2(D)

(En ) + < + = .
2
2 2
R |fn f |
Thus, limn D 1+|fn f | (dx) = 0.
R |fn f |
Conversely, suppose limn D 1+|f

d = 0. We show fn
f on D. For any
n f |
> 0, for n N, let En = {x D : |fn f | }. We have
|fn f |
|fn f |
1 + |fn f |
1+
x
, x > 0 is increasing).
( since the function (x) = 1+x
It follows that
Z
Z
Z
|fn f |
|fn f |
0
d
d
d.
En 1 +
En 1 + |fn f |
D 1 + |fn f |
Hence,
Since limn
0
(En )
1+
|fn f |
d
D 1+|fn f |
Z
D
|fn f |
d.
1 + |fn f |
= 0, limn (En ) = 0. Thus, fn

f on D.
Problem 50
Let (X, A, ) be a finite measure space. Let be the set of all extended realvalued A-measurable function on X where we identify functions that are equal
a.e. on X. Let
Z
|f g|
d for f, g .
(f, g) =
X 1 + |f g|

59
(a) Show that is a metric on .

(b) Show that is complete w.r.t. the metric .
Solution
(a) Note that (X) is finite and 0
|f g|
1+|f g|
< 1, so 0 < .
R |f g|
(f, g) = 0 X 1+|f
d = 0 f g = 0 f = g. (We identify
g|
functions that are equal a.e. on X.)
It is clear that (f, g) = (g, f ).
We make use the fact that the function (x) =
f, g, h ,
x
,
1+x
x > 0 is increasing. For
|f h|
|f g| + |g h|
1 + |f h|
1 + |f g| + |g h|
|f g|
|g h|
=
+
1 + |f g| + |g h| 1 + |f g| + |g h|
|f g|
|g h|
+
.
1 + |f g| 1 + |g h|
Integrating over X we get
Z
X
|f h|
d
1 + |f h|
Z
X
|f g|
d +
1 + |f g|
Z
X
|g h|
d.
1 + |g h|
That is
(f, g) (f, h) + (h, g).
Thus, is a metric on .
(b) Let (fn ) be a Cauchy sequence in . We show that there exists an f such
that (fn , f ) 0 as n .
First we claim that (fn ) is a Cauchy sequence w.r.t. convergence in measure. Let
> 0. For n, m N, define Am,n = {X : |fn fm | }. For every > 0, there
exists an N N such that
()
n, m N (fn , fm ) <
.
1+

While we have that
(fn , fm ) =
X
|fn fm |
d
1 + |fn fm |
Z
Am,n
|fn fm |
d
1 + |fn fm |
(Am,n ).
1+
For n, m N , from (*) we get
>
(Am,n ).
1+
1+
This implies that (Am,n ) < . Thus, (fn ) is Cauchy in measure. We know that if
(fn ) is Cauchy in measure then (fn ) converges in measure to some f .
Next we prove that (fn , f ) 0. Since fn

f , for any > 0 there exists E A
and an N N such that
(E) <
and |fn f | <
on X \ E whenever n N.
2
2(X)
On X \ E, for n N , we have
Z
Z
|fn f |
d
|fn f |d <
(X \ E) .
2(X)
2
X\E 1 + |fn f |
X\E
On E, for all n, we have
Z
E
|fn f |
d
1 + |fn f |
1 d = (E) < .
2
E
Hence, for n N , we have

Z
Z
Z
|fn f |
|fn f |
|fn f |
(fn , f ) =
d =
d +
d < .
X 1 + |fn f |
E 1 + |fn f |
X\E 1 + |fn f |
Thus, (fn ) converges to f . And hence, (, ) is complete
Problem 51(Bounded convergence theorem under convergence in measure)

Suppose that (fn ) is a uniformly bounded sequence of real-valued measurable funcw
f
tions on D, and f is a bounded real-valued measurable function on D. If fn
on D, then
Z
lim
|fn f |d = 0.
n

61
Solution
We will use this fact:
Let (an ) be a sequence of real numbers. If there exists a real number a such that
every subsequence (ank ) has a subsequence (ankl ) converging to a, then the sequence
(an ) converges to a.
Consider the sequence of real numbers
Z
an =
|fn f |d, n N.
D
Take an arbitrary subsequence (ank ). Consider the sequence (fnk ). Since (fn ) converges to f in measure on D, the subsequence (fnk ) converges to f in measure on
D too. By Riesz theorem, there exists a subsequence (fnkl ) converging to f a.e. on
D. Thus by the bounded convergence theorem, we have
Z
lim
|fnkl f |d = 0.
n
That is, the subsequence (ankl ) of the arbitrary subsequence (ank ) of (an ) converges
to 0. Therefore the sequence (an ) converges to 0. Thus
Z
lim
|fn f |d = 0.

Chapter 6
Integration of Nonnegative
Functions
Definition 19 Let f be a nonnegative extended real-valued measurable function on a measurable
D R. We define the Lebesgue integral of f on D by
Z
f d = sup d,
0f
where the supremum is on the collection of all nonnegative simple function on D.

If the integral is finite, we say that f is integrable on D.
Proposition 17 (Properties)
Let f, f1 and f2 be nonnegative extended real-valued measurable functions on D. Then
1.
2.
3.
4.
5.
6.
R
D
f d < f < a.e. on D.
f d = 0 f = 0 a.e. on D.
R
R
D0 D D0 f d D f d.
R
f > 0 a.e. on D and D f d = 0 (D) = 0.
R
R
f1 f2 on D D f1 d D f2 d.
R
R
f1 = f2 a.e. on D D f1 d D f2 d.
D
Theorem 6 (Monotone convergence theorem)

Let (fn ) be an increasing sequence of nonnegative extended real-valued measurable functions on D.
If fn f on D then
Z
Z
lim
fn d =
f d.
n
Remark: The conclusion is not true for a decreasing sequence.

64
CHAPTER 6. INTEGRATION OF NONNEGATIVE FUNCTIONS
Proposition 18 Let (fn ) be an increasing sequence of nonnegative extended real-valued measurable functions on D. Then we have
Z X !
XZ
fn d =
fn d.
D
nN
nN
Theorem 7 (Fatous Lemma)

Let (fn ) be a sequence of nonnegative extended real-valued measurable functions on D. Then we
have
Z
Z
lim inf fn d lim inf
fn d.
D n
In particular, if limn fn = f exists a.e. on D, then

Z
Z
f d lim inf
fn d.
n
Proposition 19 (Uniform absolute continuity of the integral)

Let f be an integrable nonnegative extended real-valued measurable functions on D. Then for every
> 0, there exists > 0 such that
Z
f d <
E
for every measurable E D with (E) < .
Problem 52
Let f1 and f2 be nonnegative extended real-valued measurable functions on a
measurable set D R. Suppose f1 f2 and f1 is integrable on D. Prove that
f2 f1 is defined a.e. on D and
Z
Z
Z
(f2 f1 )d =
f2 d
f1 d.
D
Solution
Since f1 is integrable on D, f1 is real-valued a.e. on D. Thus there exists a null set
N D such that 0 f1 (x) < , x D \ N . Then f2 f1 is defined on D \ N .
That is f2 f1 is defined a.e. on D. On the other hand, since f2 = f1 + (f2 f1 ),
we have
Z
Z
Z
Z
f1 + (f2 f1 ) d =
f1 d + (f2 f1 )d.
f2 d =
D

65
Since
R
D
f1 d < , we have
Z
(f2 f1 )d =
f2 d
Remark: If
R
D
f1 d = ,
R
D
f2 d
R
D
f1 d.
f1 d may have the form .
Problem 53
Let f be a non-negativeR real-valued measurable function on a measure space
(X, A, ). Suppose that E f d = 0 for every E A. Show that f = 0 a.e.
Solution
Since f 0, A = {x X : f (x) > 0} = {x X : f (x) 6= 0}. We shall show that
(A) = 0.
S
Let An = {x X : f (x) n1 } for every n N. Then A = nN An . Now on An we
have
Z
1
1
f
f d (An )
n
n
An
Z
(An ) n
f d = 0 (by assumption)
An
Thus, 0 (A)
a.e.
P
nN
(An ) = 0 for every n N.

(An ) = 0. Hence, (A) = 0. This tells us that f = 0
Problem 54
Let (fn : n N) be a sequence of non-negative real-valued measurable functions
on R such that fnR f a.e. on
R R.
Suppose limn R fn d = R f d < . Show that for each measurable set
E R we have
Z
Z
fn d =
f d.
lim
n
Solution

66
Since gn = fn fn E 0, n N and fn f a.e., we have, by Fatous lemma,

Z
Z
lim gn d lim inf gn d
n
n
Z R
ZR
(f f E )d lim inf (fn fn E )d
n
Z R
Z R
Z
Z
f d
f d lim
fn d lim sup fn d.
R
From the last inequation and assumption we get

Z
Z
(6.1)
f d lim sup fn d.
n
Let hn = fn fn E 0. Using the similar calculation, we obtain

Z
Z
(6.2)
f d lim inf
fn d.
n
From (6.1) and (6.2) we have
lim
fn d =
E
f d.
Problem 55
Given a measure space (X, A, ). Let (fn ) and f be extended real-valued Ameasurable functions on D A and assume that f is real-valued a.e. on D.
Suppose there exists a sequence of positive numbers (n ) such that
1.
2.
P
R
D
nN n
< .
|fn f |p d < n for every n N for some fixed p (0, ).
Show that the sequence (fn ) converges to f a.e. on D. (Note that no integrability
of fn , f, |f |p on D is assumed).
Solution
P
N
p
|f
f
|
Since |fn f |p is non-negative measurable for every n N, the sequence
n=1 n
is an increasing sequence of non-negative measurable functions. By the Monotone
Convergence Theorem, we have
N
!
Z
Z X
N
X
p
lim
|fn f | d = lim
|fn f |p d.
D N
n=1
D n=1
N N

67
Using assumptions we get
Z X
|fn f |p d =
D n=1
N Z
X
lim
N
Z
X
n=1
n=1
|fn f |p d
D
|fn f |p d
D
n < .
n=1
This means that the function under the integral symbol in the left hand side is finite
a.e. on D. We have
X
|fn f |p < a.e. on D lim |fn f |p = 0 a.e. on D
n
n=1
lim |fn f | = 0 a.e. on D
fn f a.e. on D. .
Problem 56
Given a measure space (X, A, ). Let (fn ) and f be extended real-valued measurable functions
on D A and assume that f is real-valued a.e. on D. Suppose
R
limn D |fn f |p d = 0 for some fixed p (0, ). Show that
fn
f on D.
Solution
Given any > 0. For every n N, let An = {D : |fn f | }. Then
Z
Z
Z
p
p
|fn f | d =
|fn f | d +
|fn f |p d
D
A
D\An
Z n
|fn f |p d
An
Since limn
R
D
p (An ).
|fn f |p d = 0, limn (An ) = 0. This means that
fn
f on D.

68
Problem 57
Let (X, A, ) be a measure space and
R letp f be an extended real-valued Ameasurable function on X such that X |f | d < for some fixed p (0, ).
Show that
lim p {X : |f | } = 0.
Solution
For n = 0, 1, 2, ..., let En S
= {D : n |f | < n + 1}. Then En A and the En s are
disjoint. Moreover, X =
n=0 En . We have
Z
Z
X
X
p
p
>
|f | d =
|f | d
np (En ).
X
n=0
En
n=0
Since n=0 np (En ) < , for any > 0, there exists N N such that for n N
we have
X
np (En ) < .
n=N
p
Note that n N since p > 0. So we have

N
But
S
n=N
(En ) < .
n=N
En = {X : |f | N }. So with the above N , we have

!
[
N p
En = N p {X : |f | N } < .
n=N
Thus,
lim p {X : |f | } = 0.
Problem 58
Let (X, A, ) be a -finite measure space. Let f be an extended real-valued Ameasurable function on X. Show that for every p (0, ) we have
Z
Z
p
|f | d =
pp1 {X : |f | > }L (d). ()
X
[0,)

69
Solution
We may suppose f 0 (otherwise we set g = |f | 0).
1. If f = E , E A, then
Z
Z
p
f d =
(E )p d = (E).
X
ZX
Z 1
p1
p {X : E > }L (d) =
pp1 (E)d = (E).
[0,)
Thus, the equality () holds.

P
2. If f = ni=1 ai Ei (simple function), with ai 0, Ei A, i = 1, ..., n., then the
equality () holds because of the linearity of the integral.
3. If f 0 measurable, then there is a sequence (n ) of non-negative measurable
simple functions such that n f . By the Monotone Convergence Theorem we have
Z
Z
p
f d = lim
pn d
n
X
ZX
= lim
pp1 {X : n > }L (d)
n [0,)
Z
=
pp1 {X : f > }L (d).
[0,)
Notes:
1. A = {X : E > } = {x X : E (x) > }.

If 0 < 1 then A = E.
If 1 then A = .
2. Why -finite measure?
Problem 59
Given a measure space (X, A, ). Let f be a non-negative extended real-valued
A-measurable function on D A with (D) < .
Let Dn = {x D : f (x) n} for n N. Show that
Z
X
f d <
(Dn ) < .
D
nN

70
Solution
From the expression Dn = {x D : f (x) n} with f A-measurable, we deduce
that Dn A and
D := D0 D1 D2 ... Dn Dn+1 ...
Moreover, all the sets Dn \ Dn+1 = {D : n f < n + 1, n N} are disjoint and
[
(Dn \ Dn+1 ).
D=
nN
It follows that
n(Dn \ Dn+1 )
f d (n + 1)(Dn \ Dn+1 )
Z
n(Dn \ Dn+1 )
f d
n=0
Dn \Dn+1
nN (Dn \Dn+1 )
n[(Dn ) (Dn+1 )]
f d
D
n=0
(n + 1)(Dn \ Dn+1 )
n=0
(n + 1)[(Dn ) (Dn+1 )]. (i)
n=0
Some more calculations:
X
n[(Dn ) (Dn+1 )] = 1[(D1 ) (D2 )] + 2[(D2 ) (D3 )] + ...
n=0
(Dn ),
n=1
and
X
(n + 1)[(Dn ) (Dn+1 )] = 1[(D0 ) (D1 )] + 2[(D1 ) (D2 )] + ...
n=0
= (D) +
(Dn ).
n=1
With these, we rewrite (i) as follows

Z
X
X
(Dn )
f d (D) +
(Dn ).
n=1
n=1
Since (D) < , we have

Z
X
f d <
(Dn ) < .
D
nN

71
Problem 60
Given a measure space (X, A, ) with (X) < . Let f be a non-negative
extended real-valued A-measurable function on X. Show that f is -integrable
on X if and only if
2n {x X : f (x) > 2n } < .
n=0
Solution
Let En = {X : f > 2n } for each n = 0, 1, 2, ... Then it is clear that
E0 E1 ... En En+1 ...
En \ En+1 = {X : 2n < f 2n+1 } and are disjoint
X \ E0 = {X : 0 f 1}
[
X = (X \ E0 )
(En \ En+1 ).
n=0
Now we have
Z
f d =
X
f d +
Z
X\E0
f d
n=0 (En \En+1 )
f d +
X\E0
Z
X
n=0
f d.
En \En+1
This implies that

Z
X
(6.3)
n=0
f d =
En \En+1
f d
X
f d.
X\E0
On the other hand, for n = 0, 1, 2, ..., we have

Z
n
2 (En \ En+1 )
f d 2n+1 (En \ En+1 ).
En \En+1
Therefore,
X
n=0
2 (En \ En+1 )
Z
X
n=0
f d
En \En+1
2n+1 (En \ En+1 ).
n=0

72
From (6.3) we obtain

Z
2 (En \ En+1 ) +
f d
f d
X\E0
n=0
Since
Z
n+1
(En \ En+1 ) +
f d.
X\E0
n=0
Z
0
f d (X \ E0 ) (X) < ,
X\E0
we get
(6.4)
Z
n
2 (En \ En+1 )
f d
X
n=0
2n+1 (En \ En+1 ) + (X).
n=0
2 (En \ En+1 ) =
n=0
2n [(En ) (En+1 )]
n=0
= (E0 ) (E1 ) + 2[(E1 ) (E2 )] + 4[(E2 ) (E3 )] + ...

= (E0 ) + (E1 ) + 2(E2 ) + 4(E3 ) + ...
1
1X n
=
(E0 ) +
2 (En ),
2
2 n=0
and
n+1
(En \ En+1 ) =
n=0
2n+1 [(En ) (En+1 )]
n=0
= 2[(E0 ) (E1 )] + 4[(E1 ) (E2 )] + 8[(E2 ) (E3 )] + ...

= (E0 ) + [(E0 ) + 2(E1 ) + 4(E2 ) + 8(E3 ) + ...]
X
= (E0 ) +
2n (En ).
n=0
With these, we rewrite (6.4) as follows
1
1X n
(E0 ) +
2 (En )
2
2 n=0
Z
f d (E0 ) +
X
2n (En ) + (X).
n=0
This implies that
1X n
2 (En )
2 n=0
Z
f d
X
2n (En ) + 2(X).
n=0

73
Since (X) < , we have
Z
X
f d <
2n {x X : f (x) > 2n } < .
X
n=0
Problem 61
(a) Let {cn,i : n, i N} be an array of non-negative extended real numbers.
Show that
X
X
cn,i
lim inf cn,i .
lim inf
n
iN
iN
(b) Show that if (cn,i : n N) is an increasing sequence for each i N, then

X
X
lim
cn,i =
lim cn,i .
n
iN
iN
Solution
(a) Let : N [0, ] denote the counting measure. Consider the space (N, P(N), ).
It is a measure space in which every A N is measurable. Let i 7 b(i) be any
function on N. Then
Z
X
bd =
b(i).
N
iN
For the array {cn,i }, for each i N, we can write cn,i = cn (i), n N. Then cn is a
non-negative -measurable function defined on N. By Fatous lemma,
Z
Z
lim inf cn d lim inf cn d,
N n
that is
lim inf cn,i lim inf
iN
cn,i .
iN
(b) If (cn,i : n N) is an increasing sequence for each i N, then the sequence of

functions (cn ) is non-negative increasing. By the Monotone Convergence Theorem
we have
Z
Z
cn (i)d =
lim cn (i)d,
lim
n
that is
lim
N n
X
iN
cn,i =
X
iN
lim cn,i .

74
Chapter 7
Integration of Measurable
Functions
Given a measure space (X, A, ). Let f be a measurable function on a set D A. We define the
positive and negative parts of f by
f + := max{f, 0} and f := max{f, 0}.
Then we have
f = f + f and |f | = f + + f .
Definition 20 Let f be an extended real-valued measurable function on D. The function f is said

to be integrable on D if f + and f are both integrable on D. In this case we define
Z
Z
Z
+
f d =
f d
f d.
D
Proposition 20 (Properties)
1. f is integrable on D if and only if |f | is integrable on D.
2. If f is integrable on D then cf is integrable on D, and we have
c is a constant in R.
R
D
cf d = c
3. If
R f and gRare integrable on D then f + g are integrable on D, and we have
f d + D gd.
D
R
R
4. f g D f d D gd.
R
D
R
D
f d, where
(f + g)d =
5. If f is integrable on D then |f | < a.e. on D, that is, f is real-valued a.e. on D.

6. If {D1 , ..., Dn } is a disjoint collection in A, then
Z
n Z
X
f
d
=
S
n
i=1
Di
i=1
f d.
Di

76
CHAPTER 7. INTEGRATION OF MEASURABLE FUNCTIONS
Theorem 8 (generalized monotone convergence theorem)

Let (fn ) be a sequence of integrable extended real-valued functions on D.
1. If (fn ) is increasing and there is a extended real-valued measurable function g such that fn g
for every n N, then
Z
lim fn d =
gd.
D
2. If (fn ) is decreasing and there is a extended real-valued measurable function g such that fn g
for every n N, then
Z
gd.
lim fn d =
n
Theorem 9 (Lebesgue dominated convergence theorem theorem - D.C.T)

Let (fn ) be a sequence of integrable extended real-valued functions on D and g be an integrable
nonnegative extended real-valued function on D such that |fn | g on D for every n N. If
limn fn = f exists a.e. on D, then f is integrable on D and
Z
Z
Z
|fn f |d = 0.
f d and lim
lim
fn d =
n
Problem 62
Prove this statement:
Let f be extended real-valued measurable function on a measurable set D. If f
is integrable on D, then the set {D : f 6= 0} is a -finite set.
Solution
For every n N set
n
1o
.
Dn = x D : |f (x)|
n
Then we have
{x D : f (x) 6= 0} = {x D : |f (x)| > 0} =
[
nN
Now for each n N we have

1
(Dn )
n
Z
|f |d
Dn
|f |d < .
D
Thus
(Dn ) = < , n N,
that is, the set {x D : f (x) 6= 0} is -finite.
Dn .

77
Problem 63
Let f be extended real-valued measurable function on a measurable set D. If (En )
is an increasing sequence of measurable sets such that limn En = D, then
Z
Z
f d = lim
f d.
n
En
Solution
Since (En ) is an increasing sequence with limit D, so by definition, we have
D=
En .
n=1
Let
D1 = E1 and Dn = En \ En+1 , n 2.
Then {D1 , D2 , ...} is a disjoint collection of measurable sets, and we have
n
[
Di = En and
Dn =
n=1
i=1
En = D.
n=1
Hence
Z
f d =
D
Z
X
n=1
lim
f d = lim
Dn
Sn
i=1 Di
n Z
X
i=1
f d
Di
f d = lim
f d.
En
Problem 64
Let (X, A, ) be a measure space. Let f and g be extended real-valued
measurable
R
R
functions on X. Suppose that f and g are integrable on X and E f d = E gd
for every E A. Show that f = g a.e. on X.
Solution
Case 1: f and g are two real-valued integrable functions on X.
Assume that the statement f = g a.e. on X is false. Then at least one of the two

78
sets E = {X : f < g} and F = {X : f > g} has a positive measure. Consider the

case (E) > 0. Now since both f and g are real-valued, we have
n
E=
Ek
kN
Then 0 < (E)

that
Therefore
P
kN
1o
where Ek = E = X : g f
.
k
(Ek ). Thus there exists k0 N such that (Ek0 ) > 0, so

Z
1
(g f )d (Ek0 ) > 0.
k0
Ek0
Z
1
f d + (Ek0 ) >
gd
k0
Ek0
Ek0
Z
f d.
Ek0
This is a contradiction. Thus (E) = 0. Similarly, (F ) = 0. This shows that f = g

a.e. on X.
Case 2: General case, where f and g are two extended real-valued integrable
functions on X. The integrability of f and g implies that f and g are real-valued
a.e. on X. Thus there exists a null set N X such that f and g are real-valued on
X \ N . Set
(
(
f
on
X
\
N,
g on X \ N,
f =
and
g =
0 on N.
0 on N.
Then f and g are real-valued on X, and so on every E A we have
Z
Z
Z
Z
f d =
f d =
gd =
gd.
E
By the first part of the proof, we have f = g a.e. on X. Since f = f a.e. on X

and g = g a.e. on X, we deduce that
f = g a.e. on X.
Problem 65
Let (X, A, ) be a -finite measure space
R and let Rf, g be extended real-valued
measurable functions on X. Show that if E f d = E gd for every E A then
f = g a.e. on X. (Note that the integrability of f and g is not assumed.)

79
Solution
The space (X, A, ) is -finite :
X=
Xn , (Xn ) < , n N and {Xn : n N} are disjoint.
nN
To show f = g a.e. on X it suffices to show f = g a.e. on each Xn (since countable

union of null sets is a null set).
Assume that the conclusion is false, that is if E = {Xn : f < g} and F = {Xn :
f > g} then at least one of the two sets has a positive measure. Without lost of
generality, we may assume (E) > 0.
Now, E is composed of three disjoint sets:
E (1) = {Xn : < f < g < },
E (2) = {Xn : < f < g = },
E (3) = {Xn : = f < g < }.
Since (E) > 0, at least one of these sets has a positive measure.
1. (E (1) ) > 0. Let
(1)
Em,k,l = {Xn : m f ; f +
Then
E (1) =
[ [[
1
g ; g l}.
k
(1)
Em,k,l .
nN kN lN
By assumption and the subadditivity of we have

X
(1)
0 < (E (1)
(Em,k,l ).
m,k,lN
This implies that there are some m0 , k0 , l0 N such that

(Em0 ,k0 ,l0 ) > 0.
Let E = Em0 ,k0 ,l0 then we have
Z
Z
Z
1
(g f )d (E ) > 0 so
gd >
f d.
k0
E
E
E
This is a contradiction.

80

2. (E (2) ) > 0. Let
(2)
El
= {Xn : < f l; g = }.
Then
E (2) =
(2)
El .
lN

X
(2)
0 < (E (2) )
(El ).
lN
This implies that there is some l0 N such that

(2)
(El0 ) > 0.
(2)
Let E = El0 . Then
Z
gd = >
E
This contradicts the assumption that
f d.
E
R
E
f d =
R
E
gd for every E A.
3. (E (3) ) > 0. Let

(2)
Em
= {Xn : = f ; m g < }.
Then
E (3) =
(3)
Em
.
mN

X
(3)
0 < (E (3) )
(Em
).
mN
This implies that there is some m0 N such that

(3)
(Em
) > 0.
0
(3)
Let E = Em0 . Then

Z
Z
gd m(E ) > =
E
f d :
E
This contradicts the assumption.

Thus, (E) = 0. Similarly, we get (F ) = 0. That is f = g a.e. on X.

81
Problem 66
Given a measure space (X, A, ). Let f be extended real-valued measurable and
integrable function on X.
1. Show that for any > 0 there exists > 0 such that if A A with (A) <
then
f d < .
2. Let R(En ) be a sequence in A such that limn (En ) = 0. Show that

limn En f d = 0.
Solution
1. For every n N, set
(
fn (x) =
f (x)
n
if f (x) n
otherwise.
Then the sequence (fn ) is increasing. Each fn is bounded and fn f pointwise.

By the Monotone Convergence Theorem,
Z

> 0, N N such that fN d
f d < .
2
X
X
Take =
.
2N
If (A) < , we have

Z
Z
Z
f d (fN f )d + fN d
A
A
ZA
(fN f )d + N (A)
X
+ = .
<
2 2
2. Since limn (En ) = 0, with and as above, there exists n0 N such that
for n n0 , (En ) < . Then we have
< .
f
d
En
This shows that limn
R
En
f d = 0.

82
Problem 67
Given a measure space (X, A, ). Let f be extended real-valued A-measurable
and integrable function on X. Let En = {x X : |f (x)| n} for n N. Show
that limn (En ) = 0.
Solution
First we note that X = E0 . For each n N, we have
En \ En+1 = {X : n |f | < n + 1}.
Moreover, the collection {En \ En+1 : n N} A consists of measurable disjoint
sets and
[
(En \ En+1 ) = X.
n=0
By the integrability of f we have

Z
Z
X
>
|f |d =
X
n=0
|f |d
En \En+1
n(En \ En+1 ).
n=0
Some more calculations for the last summation:
n(En \ En+1 ) =
n=0
n[(En ) (En+1 )]
n=0
= (E1 ) (E2 ) + 2[(E2 ) (E3 )] + 3[(E3 ) (E4 )] + ...
X
=
(En ) < .
n=1
Since the series
P
n=1
(En ) converges, limn (En ) = 0.
Problem 68
Let (X, A, ) be a measure space.
(a) Let {En : n N} be a disjoint collection
S in A. Let f be an extended realvalued A-measurable
R function defined on nN En . If f is integrable on En for
every n N, does S En f d exist?
nN
(b) Let (Fn : n N) be an increasing sequence
in A. Let f be an extended realS
valued A-measurable function defined on R nN Fn . Suppose f is integrable
on En
R
for every n N and moreover limn Fn f d exists in R. Does S Fn f d
nN
exist?

83
Solution
(a) NO.
X = [1, ), En = [n, n + 1), n = 1, 2, ..., {En } disjoint.
A = ML , L .
[
X=
En , f (x) = 1, x X.
nN
f d = 1, n N,
En
nN
f d =
En
1d = .
[1,)
(b) NO.
X = R, Fn = (n, n), n = 1, 2, ..., (Fn : n N) increasing
A = ML , L .
[
X=
Fn , f (x) = 1 for x 0, f (x) = 1 for x < 0
nN
f d =
Z
(1)d +
Z
Z
f d =
f d =
Fn
S
nN
(n,0)
Fn
[0,n)
1d = 0 lim
f d = 0
n F
n
Z
(1)d +
1d does not exist.
(,0)
(0,)
Problem 69
Let f is a real-valued uniformly continuous function on [0, ). Show that if f
is Lebesgue integrable on [0, ), then
lim f (x) = 0.
Solution
Suppose NOT. Then there exists > 0 such that for each n N, there is xn > n
such that |f (xn )| . W.L.O.G. we may choose (xn ) such that
xn+1 > xn + 1 for all n N.
Since f is uniformly continuous on [0, ), with the above ,
1
(0, ) : |x y| < |f (x) f (y)| < .

2
2

84
In particular, for x In = (xn , xn + ), we have
|f (xn ) f (x)| < , n N.

2
This implies
|f (x)| > |f (xn )| = .

2
2
2
2
S
Since xn+1 xn > 1 and 0 < < 12 , In In+1 = . Moreover,
n=1 In [0, ).
By assumption, f is integrable on [0, ), so we have
|f (xn )| |f (x)| <
Z
>
f d
Z
X
[0,)
n=1
f d >
In
Z
X
n=1
In
d = .
2
This is a contradiction. Thus,

lim f (x) = 0.
Problem 70
Let (X, A, ) be a measure space and let (fn )nN , and f, g be extended real-valued
A-measurable and integrable functions on D A. Suppose that
1. limn fn = f a.e. on D.
R
R
2. limn D fn d = D f d.
3. either fn g on D for all n N or fn g on D for all n N.
Show that, for every E A and E D, we have
Z
Z
lim
fn d =
f d.
n
Solution
(a) First we solve the problem in the case the condition 3. is replaced by fn 0 on
D for all n N.
Let hn = fn fn E for every E A and E D. Then hn 0 and A-measurable

85
and integrable on D. Applying Fatous lemma to hn and using assumptions, we get
Z
Z
Z
Z
f d
f d = (f f E )d lim inf (fn fn E )d
n
D
E
D
Z D
Z
= lim
fn d lim sup
fn E d
n D
n
D
Z
Z
=
f d lim sup fn d.
n
Since f is integrable on D,
()
f d < . From the last inequality we obtain,

Z
Z
lim sup fn d
f d.
n
Let kn = fn + fn E for every E A and E D. Using the same way as in the

previous paragraph, we get
Z
Z
()
f d lim inf
fn d.
n
From (*) and (**) we get

Z
lim
fn d =
E
f d.
Next we are coming back to the problem. Assume fn g on D for all n N. Let
n = fn g. Using the above result for n 0 we get
Z
Z
lim
n d =
d.
n
That is
Z
lim
ZE
lim
(fn g)d = (f g)d

E Z
Z
Z
fn d
gd =
f d
gd.
Since g is integrable on E,
gd < . Thus, we have

Z
Z
fn d =
f d.
lim
E

86
Problem 71(An extension of the Dominated Convergence Theorem)

Let (X, A, ) be a measure space and let (fn )nN , (gn )nN , and f, g be extended
real-valued A-measurable functions on D A. Suppose that
1. limn fn = f and limn gn = g a.e. on D.
R
R
2. (gn ) and g are all integrable on D and limn D gn d = D gd.
3. |fn | gn on D for every n N.
Prove that f is integrable on D and limn
R
D
fn d =
R
D
f d.
Solution
Consider the sequence (gn fn ). Since |fn | gn , and (fn ) and (gn ) are sequences
of measurable functions, the sequence (gn fn ) consists of non-negative measurable
functions. Using the Fatous lemma we have
Z
Z
lim inf (gn fn )d lim inf (gn fn )d
n
n
D
D
Z
Z
Z
lim (gn fn )d lim
gn d lim sup
fn d
n
n D
n
D
D
Z
Z
Z
Z
gd
f d
gd lim sup
fn d
n
D
D
ZD
Z D
Z
f d lim sup
fn d. () (since
gd < ).
D
Using the same process for the sequence (gn + fn ), we have

Z
Z
f d lim inf
fn d. ().
n
From (*) and (**) we obtain
lim
fn d =
D
f d.
D
The fact that f is integrable comes from gn is integrable:

Z
Z
|fn | gn
fn d
gn d <
D
ZD
f d < .
D

87
Problem 72
Given a measure space (X, A, ). Let (fn )nN and f be extended real-valued
A-measurable and integrabe functions on D A. Suppose that
lim fn = f a.e. on D.
R
R
R
R
(a) Show that if limn D |fn |d = D |f |d, then limn D fn d = D f d.
(b) Show that the converse of (a) is false by constructing a counter example.
Solution
(a) We will use Problem 71 for
gn = 2(|fn | + |gn |) and hn = |fn f | + |fn | |f |, n N.
We have
hn 0 a.e. on D,
gn 4|f | a.e. on D,
|hn | = hn 2|fn | gn ,
Z
Z
Z
Z
lim
gn d = 2 lim
|fn |d + 2
|f |d =
4|f |d.
n
So all conditions of Problem 71 are satisfied. Therefore,

Z
Z
lim
hn d =
hd = 0 (h = 0).
n D
D
Z
Z
Z
lim
|fn f |d + lim
|fn |d
|f |d = 0.
n
Since limn
R
D
|fn |d
R
D
|f |d = 0 by assumption, we have
Z
lim
|fn f |d = 0.
n
This implies that
Z
Z
f d = 0.
lim fn d
n
D
D
R
R
Hence, limn D fn d = D f d.
(b) We will give an example showing that it is not true that
Z
Z
Z
Z
lim
fn d =
f d lim
|fn |d =
|f |d.
n

88
if 0 x < n1
n
fn (x) = 0
if n1 x 1 n1
n if 1 n1 < x 1.
And so
(
|fn |(x) =
n if 0 x < n1 or 1
0 if n1 x 1 n1 .
1
n
<x1
Then we have
Z
fn 0 0 and
fn d = 0 0 =
[0,1]
while
0d
[0,1]
Z
|fn |d = 2 2 6= 0.
[0,1]
Problem 73
Given a measure space (X, A, ).
(a) Show that an extended real-valued integrable function is finite a.e. on X.
(b)
functions defined on X such that
P If R(fn )nN is a sequence of measurable
P
|f
|d
<
,
then
show
that
n
nN X
nN fn converges a.e. to an integrable
function f and
Z X
Z
XZ
fn d =
f d =
fn d.
X nN
nN
Solution
(a) Let E = {X : |f | = }. We want to show that (E) = 0. Assume that
(E) > 0. Since f is integrable
Z
Z
>
|f |d
|f |d = .
X
This is a contradiction. Thus, (E) = 0.

P
(b) First we note that N
n=1 |fn | is measurable since fn is measurable for n N.
Hence,
N
X
X
lim
|fn | =
|fn |
N
n=1
n=1

89
is measurable. Recall that (for nonnegative measurable functions)
Z X
|fn |d =
X n=1
By assumption,
n=1
Z
X
n=1
hence,
Z
X
|fn |d.
X
|fn |d < ,
X
Z X
|fn |d < .
X n=1
P
Since
n=1 |fn | is integrable on X, by part (a), it is finite a.e. on X. Define a
function f as follows:
(P
P
where
n=1 fn
n=1 |fn | <
f (x) =
0
otherwise.
P
So f is everywhere defined and f = limN N
n=1 fn a.e. Hence, f is measurable
on X. Moreover,
Z
Z
Z X
Z X
f d
|f
|d
=
f
d
|fn |d < .
X
X
X n=1
X n=1
Thus, f is integrable and hN =
PN
n=1
|hN |
fn converges to f a.e. and
N
X
|fn |
n=1
|fn |
n=1
which is integrable. By the D.C.T. we have

Z
Z
Z
f d =
lim hN d = lim
hN
X
X N
=
=
lim
Z X
N
N
Z
X
n=1
fn d = lim
X n=1
fn d.
N Z
X
n=1
fn d
X

90
Problem 74
Let f be a real-valued Lebesgue measurable function on [0, ) such that
1. f is Lebesgue integrable on every finite subinterval of [0, ).
2. limx f (x) = c R.
Show that
1
lim
a a
Z
f dL = c.
[0,a]
Solution
By assumption 2. we can write
() > 0, N : x > N |f (x) c| < .
Now, for a > N we have
Z
Z
1
f dL c =
(f c)dL
a
a [0,a]
[0,a]
Z
1
|f c|dL
a [0,a]
Z
Z
1
=
|f c|dL +
|f c|dL .
a
[0,N ]
[N,a]
By (*) we have
x [N, a] |f (x) c| < .
Therefore,
()
Z
1
1
(a N )
f dL c
|f c|dL +
.
a
a [0,N ]
a
[0,a]
It is evident that
(a N )
= .
a
a
R
By assumption 1., |f c| is integrable on [0, N ], so [0,N ] |f c|dL is finite and does
not depend on a. Hence
Z
1
lim
|f c|dL = 0.
a a [0,N ]
lim

91
Thus, we can rewrite (**) as
Z
lim
f dL c .
a a [0,a]
Since > 0 is arbitrary, this implies that
Z
lim
f dL c = 0.
a a [0,a]
Problem 75
Let
P f be a non-negative real-valued Lebesgue measurable on R. Show that if
n=1 f (x + n) is Lebesgue integrable on R, then f = 0 a.e. on R.
Solution
Recall these two facts:
R
P
P R
(
f
)
d
=
n
n=1
n=1 D fn d.
D
R
R
2. If f is defined and measurable on R then R f (x + h)d = R f (x)d.
1. If fn 0 is measurable on D then
From these two facts we have

!
Z X
Z
X
f (x + n) dL =
f (x + n)dL
R
n=1
n=1 R
Z
X
n=1
Since
P
n=1
f (x + n) is Lebesgue integrable on R,
!
Z X
f (x + n) dL < .
R
Therefore,
()
Since
f (x)dL .
n=1
Z
X
n=1
f (x)dL 0, (*) implies that

R
f (x)dL < .
R
f (x)dL = 0. Thus, f = 0 a.e. on R.

92
Problem 76
Show that the Lebesgue Dominated Convergence
a.e. convergence is replaced by convergence in measure.
Theorem
holds
if
Solution
We state the theorem:
Given a measure space (X, A, ). Let (fn : n N) be a sequence of extended realvalued A-measurable functions on D A such that |fn | g on D for every n N
for some integrable non-negative extended real-valued A-measurable function g on
D. If fn
f on D, then f is integrable on D and
Z
Z
lim
fn d =
f d.
n
Proof:
Let (fnk ) be any subsequence of (fn ). Then fnk

f since fn
f . By Riesz
theorem, there exists a subsequence (fnkl ) of (fnk ) such that fnkl f a.e. on D.
And we have also |fnkl | g on D. By the Lebesgue D.C.T. we have
Z
Z
fnkl d.
()
f d = lim
Let an =
R
D
fn d and a =
R
D
f d. Then (*) can be written as

lim ankl = a.
Hence we can say that any subsequence (ank ) of (an ) has a subsequence (ankl )
converging to a. Thus, the original sequence, namely (an ), converges to the same
limit (See Problem 51): limn an = a. That is,
Z
Z
lim
fn d =
f d.
n
Problem 77
measurable and integrable
functions on D A.
R
Suppose that limn D |fn f |d = 0. Show that
(a) fn
f Ron D.
R
(b) limn D |fn |d = D |f |d.

93
Solution
(a) Given any > 0, for each n N, let En = {D : |fn f | }. Then
Z
Z
|fn f |d
|fn f |d (En ).
D
Since limn
R
D
En
|fn f |d = 0, limn (En ) = 0. That is fn

f on D.
(b) Since fn and f are integrable

Z
Z
Z
Z
(|fn | |f |)d =
|fn |d
|f |d
|fn f |d.
D
By this and the assumption, we get

Z
Z
Z
lim
|fn |d
|f |d lim
|fn f |d = 0.
n
This implies
lim
|fn |d =
D
|f |d.
Problem 78
measurable and
R integrableR functions on D A. Assume that fn f a.e. on D
and limn D |fn |d = D |f |d. Show that
Z
lim
|fn f |d = 0.
n
Solution
For each n N, let hn = |fn | + |f | |fn f |. Then hn 0 for all n N.
Since fn f a.e. on D, hn 2|f | a.e on D. By Fatous lemma,
Z
Z
Z
2
|f |d lim inf (|fn | + |f |)d lim sup
|fn f |d
n
n
D
D
D
Z
Z
Z
= lim
|fn |d + lim
|f |d lim sup
|fn f |d
n D
n D
n
D
Z
Z
= 2
|f |d lim sup
|fn f |d.
D

94
Since |f | is integrable, we have

Z
lim sup
|fn f |d 0. (i)
Now for each n N, let gn = |fn f | (|fn | |f |). Then hn 0 for all n N.
Since fn f a.e. on D, gn 0 a.e on D. By Fatous lemma,
Z
Z
Z
0 = lim
gn d lim inf
|fn f |d lim sup (|fn | |f |)d
n D
n
n
D
Z
Z D
Z
lim inf
|fn f |d lim
|fn |d + lim
|f |d) .
n
n D
n D
D
|
{z
}
=0
Hence
Z
lim inf
|fn f |d 0. (ii)
From (i) and (ii) it follows that

Z
lim
|fn f |d = 0.
Problem 79
Let (R, ML , L ) be the Lebesgue space. Let f be an extended real-valued Lebesgue
measurable function on R. Show that if f is integrable on R then
Z
lim |f (x + h) f (x)|dx = 0.
h0
Solution
Since f is integrable,
Z
lim
|f |dx +
|f |dx
= 0 for M R.
Given any > 0, we can pick an M > 0 such that

Z
|f |dx +
|f |dx < .
4

95
Since Cc (R) is dense in L1 (R), we can find a continuous function vanishing outside
[M, M ] such that
Z M
|f |dx < .
4
M
Then we have
Z
kf k1 :=
|f |dx
R
Z M
Z
|f |dx +
<
|f |dx +
+ = .
4 4
2
|f |dx
M
(Recall: = 0 outside [M, M ] ). Now for any h R we have

kf (x + h) f (x)k1 kf (x) (x)k1 + k(x) (x + h)k1 + k(x + h) f (x + h)k1 .
Because of Cc (R) and translation invariance, we have
lim k(x) (x + h)k1 = 0 and k(x + h) f (x + h)k1 = kf (x) (x)k1 .
h0
It follows that
lim kf (x + h) f (x)k1 kf k1 + lim k(x) (x + h)k1 + kf k1
h0
2 + 0 = .
2
h0
Since > 0 is arbitrary, we have

Z
lim kf (x + h) f (x)k1 = lim
h0
h0
|f (x + h) f (x)|dx = 0.
R

96
Chapter 8
Signed Measures and
Radon-Nikodym Theorem
1. Signed measure
Definition 21 (Signed measure)
A signed measure on a measurable space (X, A) is a function : A [, ] such that:
(1) () = 0.
(2) assumes at most one of the values .
(3) is countably additive. That is, if {En }nN A is disjoint, then
!
[
X
En =
(En ).
nN
nN
Definition 22 (Positive, negative, null sets)

Let (X, A, ) be a signed measure space. A set E A is said to be positive (negative, null) for the
signed measure if
F A, F E = (F ) 0 ( 0, = 0).
Proposition 21 (Continuity)
Let (X, A, ) be a signed measure space.
1. If (En )nN A is an increasing sequence then
[
lim (En ) = lim
En = lim En .
n
nN
2. If (En )nN A is an decreasing sequence and (E1 ) < , then
\
lim (En ) = lim
En = lim En .
n
nN

98 CHAPTER 8. SIGNED MEASURES AND RADON-NIKODYM THEOREM
Proposition 22 (Some more properties)
Let (X, A, ) be a signed measure space.
1. Every measurable subset of a positive (negative, null) set is a positive (negative, null) set.
2. If E is a positive set and F is a negative set, then E F is a null set.
3. Union of positive (negative, null) sets is a positive (negative, null) set.
Theorem 10 (Hahn decomposition theorem)

Let (X, A, ) be a signed measure space. Then there is a positive set A and a negative set B such
that
A B = and A B = X.
Moreover, if A0 and B 0 are another pair, then A 4 A0 and B 4 B 0 are null sets.
{A, B} is called a Hahn decomposition of (X, A, ).
Definition 23 (Singularity)
Two signed measure 1 and 2 on a measurable space (X, A) are said to be mutually singular and
we write 1 2 if there exist two set E, F A such that E F = , E F = X, E is a null set
for 1 and F is a null set for 2 .
Definition 24 (Jordan decomposition)
Given a signed measure space (X, A, ). If there exist two positive measures and , at least one
of which is finite, on the measurable (X, A) such that
and = ,
then {, } is called a Jordan decomposition of .
Theorem 11 (Jordan decomposition of signed measures)
Given a signed measure space (X, A, ). A Jordan decomposition for (X, A, ) exists and unique,
that is, there exist a unique pair {, } of positive measures on (X, A), at least one of which is
finite, such that and = .
Moreover, with any arbitrary Hahn decomposition {A, B} of (X, A, ), if we define two set functions
and by setting
(E) = (E A) and (E) = (E B) for E A,
then {, } is a Jordan decomposition for (X, A, ).
2. Lebesgue decomposition, Radon-Nikodym Theorm
Definition 25 (Radon-Nikodym derivative)
Let be a positive measure and be a signed measure on a measurable space (X, A). If there exists
an extended real-valued A-measurable function f on X such that
Z
(E) =
f d for every E A,
E
then f is called a Radon-Nikodym derivative of with respect to , and we write
d
d
for it.

99
Let be a -finite positive measure and be a signed measure on a measurable space (X, A). If
two extended real-valued A-measurable functions f and g are Radon-Nikodym derivatives of with
respect to , then f = g -a.e. on X.
Definition 26 (Absolute continuity)
Let be a positive measure and be a signed measure on a measurable space (X, A). We say that
is absolutely continuous with respect to and write if
E A), (E) = 0 = (E) = 0.
Definition 27 (Lebesgue decomposition)
Let be a positive measure and be a signed measure on a measurable space (X, A). If there exist
two signed measures a and s on (X, A) such that
a , s and = a + s ,
then we call {a , s } a Lebesgue decomposition of with respect to . We call a and s the
absolutely continuous part and the singular part of with respect to .
Theorem 12 (Existence of Lebesgue decomposition)
Let be a -finite positive measure and be a -finite signed measure on a measurable space
(X, A). Then there exist two signed measures a and s on (X, A) such that
Z
a , s , = a + s and a is defined by a (E) =
f d, E A,
E
where f is an extended real-valued measurable function on X.

Theorem 13 (Radon-Nikodym theorem)
Let be a -finite positive measure and be a -finite signed measure on a measurable space
(X, A). If , then the Radon-Nikodym derivative of with respect to exists, that is, there
exists an extended real-valued measurable function on X such that
Z
(E) =
f d, E A.
E
Problem 80
Given a signed measure space (X, A, ). Suppose that {, } is a Jordan decomposition of , and E and F are two measurable subsets of X such that
E F = , E F = X, E is a null set for and F is a null set for .Show
that {E, F } is a Hahn decomposition for (X, A, ).

Solution
We show that E is a positive set for and F is a negative set for . Since {, } is
a Jordan decomposition of , we have
(E) = (E) (E), E A.
Let E0 A, E0 E. Since E is a null set for , E0 is also a null set for . Thus
(E0 ) = 0. Consequently, (E0 ) = (E0 ) 0. This shows that E is a positive set
for .
Similarly, let F0 A, F0 E. Since F is a null set for , F0 is also a null set for
. Thus (F0 ) = 0. Consequently, (F0 ) = (F0 ) 0. This shows that F is a
negative set for .
We conclude that {E, F } is a Hahn decomposition for (X, A, ).
Problem 81
Consider a measure space ([0, 2], ML [0, 2], L ). Define a signed measure
on this space by setting
Z
(E) =
sin xdL , for E ML [0, 2].
E
Let C = [ 43 , 53 ]. Let > 0 be arbitrary given. Find a measurable set C 0 C

such that (C 0 ) (C) and (E) > for every measurable subset E of C 0 .
Solution
Let X = [0, 2], f (x) = sin x. Then f is continuous on X, so f is Lebesgue
(=Riemann) integrable on X. Given > 0, let = min{ 2 , 3 }. Let C 0 = [ 34 , 43 +],
then
C 0 C and f (x) = sin x < 0, x C 0 .
We have
(C ) =
sin xdL
C0
sin xdL = (C).

C
Now for any E C 0 and E ML [0, 2], since (E) (C 0 ) and f (x) 0 on
C 0 , we have
Z
Z
Z
(E) =
sin xdL
sin xdL
(1)dL = (C 0 ) = .
C0
C0
By the choice of , we have

<
> > .
2
2

101
Thus, for any E ML [0, 2] with E C 0 we have (E) > .
Problem 82
Given a signed measure space (X, A, ).
(a) Show that if E A and (E) > 0, then there exists a subset E0 E which
is a positive set for with (E0 ) (E).
(b) Show that if E A and (E) < 0, then there exists a subset E0 E which
is a negative set for with (E0 ) (E).
Solution
(a) If E is a positive set for then were done (just take E0 = E).
Suppose E is a not positive set for . Let {A, B} be a Hahn decomposition of
(X, A, ). Let E0 = E A. Since A is a positive set, so E0 is also a positive set (for
E0 A). Moreover,
(E) = (E A) + (E B) = (E0 ) + (E B).
Since (E B) 0, 0 < (E) (E0 ). Thus, E0 = E A is the desired set.
(b) Similar argument. Answer: E0 = E B.
Problem 83
Let and two positive measures on a measurable space (X, A). Suppose for
every > 0, there exists E A such that (E) < and (E c ) < . Show that
.
Solution
Recall: For positive measures and
A A : (A) = 0 and (Ac ) = 0.
By hypothesis, for every n N, there exists En A such that
(En ) <
Hence,
X
nN
(En )
1
1
and (Enc ) < 2 .
2
n
n
X 1
X
X 1
c
)
<
and
(E
< .
n
2
2
n
n
nN
nN
nN

By Borel-Cantellis lemma we get
c
lim sup En = 0 and lim sup En = 0.
n
Let A = lim supn En . Then (A) = 0.

We claim: Ac = lim inf n Enc . Recall:
lim inf An = {x X : x An
n
(*)
for all but finitely many n N}.
For every x X, for each n N, we have either x En or x Enc . If x En for

infinitely many n, then x lim supn En and vice versa. Otherwise, x En for
a finite numbers of n. But this is equivalent to x Enc for all but finitely many n.
That is x lim inf n Enc . Hence,
lim sup En lim inf Enc = X.
n
Now, if x lim supn En then x En for infinitely many n, so x

/ lim inf n Enc .
This shows that
lim sup En lim inf Enc = .
n
lim inf n Enc

c
Thus, A =
as required.
Last, we show that (A ) = 0. Since lim inf n Enc lim supn Enc and (lim supn Enc ) =
0 (by the first paragraph), we get
(lim inf Enc ) = (Ac ) = 0.
n
From (*) and (**) we obtain .
()
Problem 84
Consider the Lebesgue measure space (R, ML , L ). Let be the counting measure
on ML , that is, is defined by setting (E) to be equal to the numbers of elements
in E ML if E is a finite set and (E) = if E is infinite set.
L
does not exist.
(a) Show that L but d
d
(b) Show that does not have a Lebesgue decomposition with respect to L .
Solution
(a) Let E R with (E) = 0. Since be the counting measure, E = . Then
L (E) = L () = 0. Thus,
E R, (E) = 0 L (E) = 0.

103
Hence, L .
Suppose there exists a measurable function f such that
Z
mL (E) =
f d for every E ML .
E
Take E = {x}, x R then we have

E ML , L (E) = 0, and (E) = 1.
This implies that f 0. Then for every A ML we have
Z
L (A) =
0d = 0.
A
This is impossible.
(b) Assume that have a Lebesgue decomposition with respect to L . Then, for
every E R and some measurable function f ,
Z
= a + s , a L , s L , and a (E) =
f dL .
E
Since s L , there exists A ML such that L (Ac ) = 0 and A is a null set for s .
Pick a A then s ({a}) = 0. On the other hand,
Z
a ({a}) =
f dL and L ({a}) = 0.
{a}
It follows that a ({a}) = 0. Since = a + s , we get

1 = ({a}) = a ({a}) + s ({a}) = 0 + 0 = 0.
This is a contradiction. Thus, does not have a Lebesgue decomposition with
respect to L .
Problem 85
Let and be two positive measures on a measurable space (X, A).
(a) Show that if for every > 0 there exists > 0 such that (E) < for every
E A with (E) < , then .
(b) Show that if is a finite positive measure, then the converse of (a) holds.
Solution
(a) Suppose this statement is true:
(*):= for every > 0 there exists > 0 such

that (E) < for every E A with (E) < .
Take E A with (E) = 0. Then
> 0, (E) < .
It follows that (E) = 0. Hence .
(b) Suppose is a finite positive measure and is a positive measure such that
. We want to show (*) is true. Assume that (*) is false. that is
> 0 st > 0, E A st {(E) < and (E) } .

In particular,
i
h
1
n N, En A st {(En ) < 2 and (En ) }
n
P
P
Since nN (En ) nN n12 < , by Borel-Catelli lemma, we have
> 0 st
(lim sup En ) = 0.
n
Set E = lim supn En , then (E) = 0. Since , (E) = 0. Note that

(X) < , we have
(E) = (lim sup En ) lim sup (En ) (En ) .
n
This is a contradiction. Thus, (*) must be true.
Problem 86
d
Let and be two positive measures on a measurable space (X, A). Suppose d
exists so that .
d
> 0, -a.e. on X, then and thus, .
(a) Show that if d
d
exists
(b) Show that if d > 0, -a.e. on X and if and are -finite, then d
d
and
1
d
d
=
, a.e. and a.e. on X.
d
d
Solution
(a) For every E A, by definition, we have
Z
d
d.
(E) =
E d

105
Suppose (E) = 0. Since
d
d
> 0, -a.e. on X, we have

Z
E
d
d = 0.
d
Hence, (E) = 0. This implies that and so (since is given).

d
(b) Suppose d
> 0, -a.e. on X and if and are -finite. The existence of d
is
d
guaranteed by the Radon-Nikodym theorem (since by part a). Moreover,
d
> 0, a.e. on X.
d
By the chain rule,
d d
d
.
=
= 1, a.e. on X.
d d
d
d d
d
.
=
= 1, a.e. on X.
d d
d
Thus,
d
=
d
d
d
1
, a.e. and a.e. on X.
Problem 87
Let (X, A, ) be a measure space. Assume that there exists a measurable function
f : X (0, ) satisfying the condition that {x X : f (x) n} < for
every n N.
(a) Show that the existence of such a function f implies that is a -finite
measure.
(b) Define a positive measure on A by setting
Z
(E) =
f d for E A.
E
Show that is a -finite measure.

exists and
(c) Show that d
d
d
1
= , a.e. and a.e. on X.
d
f

Solution
(a)By
S assumption, {x X : f (x) n} < for every n N. Since 0 < f < ,
so n=1 {X : f n} = X. Hence is a -finite measure.
R
(b) Let (E) = E f d for E A.
Since f > 0, is a positive measure and if (E) = 0 then (E) = 0. Hence .
Conversely, if (E) = 0, since f > 0, (E) = 0. So . Thus, . Since is
-finite ( by (a)), is also -finite.
(c) Since is -finite,
d
d
exists. By part (b), f =
d
.
d
By chain rule,
d d
.
= 1, a.e. and a.e. on X.
d d
Thus,
d
1
= , a.e. and a.e. on X.
d
f
Problem 88
Let and be -finite positive measures on (X, A). Show that there exist A, B
A such that
A B = , A B = X, on (A, A A) and on (B, A B).
Solution
Define a -finite measure = + . Then and . By the RadonNikodym theorem there exist non-negative A-measurable functions f and g such
that for every E A,
Z
Z
(E) =
f d and (E) =
gd.
E
Let A = {x X : f (x)g(x) > 0} and B = Ac . Then . Indeed, f > 0 in A.

Thus, if (E) = 0, then (E) = 0, and therefore, (E) = 0. This implies .
We can prove in the same manner. Hence, .
Let C = {x B : f (x) = 0}, D = B \ C. For any measurable sets E C and
F D, (E) = (F ) = 0. Thus, on (B, A B).

107
Problem 89
Let and be -finite positive measures on (X, A). Show that there exists
a non-negative extended real-valued A-measurable function on X and a set
A0 A with (A0 ) = 0 such that
Z
(E) =
d + (E A0 ) for every E A.
E
Solution
By the Lebesgue decomposition theorem,
Z
= a + s , a , s and a (E) =
d for any E A,
E
where is a non-negative extended real-valued A-measurable function on X.

Now since s , there exists A0 A such that
(A0 ) = 0 and s (Ac0 ) = 0.
Hence
a and (A0 ) = 0 = a (A0 ) = 0. ()
On the other hand, since s (E) = s (E A0 ) for every E A, so we have

(E A0 ) = a (E A0 ) +s (E A0 ) = s (E A0 ) = s (E).
| {z }
=0 by ()
Finally,
Z
(E) = a (E) + s (E) =
d + (E A0 ) for every E A.
E

Chapter 9
Differentiation and Integration
The measure space in this chapter is the space (R, ML , L ). Therefore, we write
instead of L for the Lebesgue measure. Also, we say f is integrable (derivable)
instead of f is L -integrable (derivable).
1. BV functions and absolutely continuous functions
Definition 28 (Variation of f )
Let [a, b] R with a < b. A partition of [a, b] is a finite ordered set P = {a = x0 < x1 < ... < xn =
b}. For a real-valued function f on [a, b] we define the variation of f corresponding to a partition
P by
n
X
Vab (f, P) :=
|f (xk ) f (xk1 )| [0, ).
k=1
We define the total variation of f on [a, b] by

Vab (f ) := sup Vab (f, P) [0, ],
P
where the supremum is taken over all partitions of [a, b]. We say that f is a function of bounded
variation on [a, b], or simply a BV function, if Vab (f ) < .
We write BV ([a, b]) for the collection of all BV functions on [a, b].
Theorem 14 (Jordan decomposition of a BV function)
1. A function f is a BV function on [a, b] if and only if there are two real-valued increasing
functions g1 and g2 on [a, b] such that f = g1 g2 on [a, b].
{g1 , g2 } is called a Jordan decomposition of f .
2. If a BV function on [a, b] is continuous on [a, b], then g1 and g2 can be chosen to be continuous
on [a, b].
Theorem 15 (Derivability and integrability)
If f is a BV function on [a, b], then f 0 exists a.e. on [a, b] and integrable on [a, b].
109- Math Vietnam

www.MathVn.com

110
CHAPTER 9. DIFFERENTIATION AND INTEGRATION
Definition 29 (Absolutely continuous functions)

A real-valued function f on [a, b] is said to be absolutely continuous on [a, b] if, given any > 0,
there exists a > 0 such that
n
X
|f (bk ) f (ak )| <
k=1
for every finite collection {[ak , bk ]}1kn of non-overlapping intervals contained in [a, b] with
n
X
|bk ak | < .
k=1
Theorem 16 (Properties)
If f is an absolutely continuous on [a, b] then
1. f is uniformly continuous on [a, b],
2. f is a BV function on [a, b],
3. f 0 exists a.e. on [a, b],
4. f is integrable on [a, b].
Definition 30 (Condition (N))
Let f be a real-valued function on [a, b]. We say
thatf satisfies Lusins Condition (N) on [a, b] if
for every E [a, b] with L (E) = 0, we have f (E) = 0.
Theorem 17 (Banach-Zarecki criterion for absolute continuity)
Let f be a real-valued function on [a, b]. Then f is absolutely continuous on [a, b] if and only if it
satisfies the following three conditions:
1. f is continuous on [a, b].
2. f is of BV on [a, b].
3. f satisfies condition (N) on [a, b].
2. Indefinite integrals and absolutely continuous functions
Definition 31 (Indefinite integrals)
Let f be a extended real-valued function on [a, b]. Suppose that f is measurable and integrable on
[a, b]. By indefinite integral of f on [a, b] we mean a real-valued function F on [a, b] defined by
Z
F (x) =
f d + c, x [a, b] and c R is a constant.
[a,x]
Theorem 18 (Lebesgue differentiation theorem)

Let f be a extended real-valued, measurable and integrable function on [a, b]. Let F be an indefinite
integral of f on [a, b]. Then
1. F is absolutely continuous on [a, b],
2. F 0 exists a.e. on [a, b] and F 0 = f a.e. on [a, b],

111
Theorem 19 Let f be a real-valued absolutely continuous on [a, b]. Then
Z
f 0 d = f (x) f (a), x [a, b].
[a,x]
Thus, an absolutely continuous function is an indefinite integral of its derivative.

Theorem 20 (A characterization of an absolutely continuous function)
A real-valued function f on [a, b] is absolutely continuous on [a, b] if and only if it satisfies the
following conditions:
(i) f 0 exists a.e. on [a, b]
(ii) fR0 is measurable and integrable on [a, b].
(iii) [a,x] f 0 d = f (x) f (a), x [a, b].
3. Indefinite integrals and BV functions
Theorem 21 (Total variation of F )
Let f be a extended real-valued measurable and integrable function on [a, b]. Let F be an indefinite
integral of f on [a, b] defined by
Z
F (x) =
f d + c, x [a, b].
[a,x]
Then the total variation of F is given by

Z
Vab (F )
|f |d.
[a,b]
Problem 90
Let f BV ([a, b]). Show that if f c on [a, b] for some constant c > 0, then
1
BV ([a, b]).
f
Solution
Let P = {a = x0 < x1 < ... < xn = b} be a partition of [a, b]. Then
n
n
X
1
X
1
|f (xk ) f (xk1 )|
b 1
=
Va
,P =
f
f
(x
f
(x
|f
(x
k)
k1 )
k )f (xk1 )|
k=1
k=1
Since f c > 0,
|f (xk ) f (xk1 )|
|f (xk ) f (xk1 )|
.
|f (xk )f (xk1 |)
c2

112
It follows that
n
1 X
1
1
b 1
Va
,P 2
|f (xk ) f (xk1 )| = 2 Vab (f, P) 2 Vab (f ).
f
c k=1
c
c

Since Vab (f ) < , Vab f1 < .
Problem 91
Let f, g BV ([a, b]). Show that f g BV ([a, b]) and
Vab (f g) sup |f |.Vab (g) + sup |g|.Vab (f ).
[a,b]
[a,b]
Solution
Note first that f, g BV ([a, b]) implies that f and g are bounded on [a, b]. There
are some 0 < M < and 0 < N < such that
M = sup |f | and N = sup |g|.
[a,b]
[a,b]
For any x, y [a, b] we have

|f (x)g(x) f (y)g(y)| |f (x) f (y)||g(x)| + |g(x) g(y)||f (y)|
N |f (x) f (y)| + M |g(x) g(y)| ().
Now, let P = {a = x0 < x1 < ... < xn = b} be any partition of [a, b]. Then we have
Vab (f g, P) =
n
X
|f (xk )g(xk ) f (xk1 )g(xk1 )|
k=1
n
X
|g(xk ) g(xk1 )| + N
k=1
M Vab (g, P)
n
X
|f (xk ) f (xk1 )|
k=1
N Vab (f, P).
Since P is arbitrary,
sup Vab (f g, P) M sup Vab (g, P) + N sup Vab (f, P),
P
where the supremum is taken over all partitions of [a, b]. Thus,
Vab (f g) sup |f |.Vab (g) + sup |g|.Vab (f ).
[a,b]
[a,b]

113
Problem 92
Let f be a real-valued function on [a, b]. Suppose f is continuous on [a, b] and
satisfying the Lipschitz condition, that is, there exists a constant M > 0 such
that
|f (x0 ) f (x00 )| M |x0 x00 |, x0 , x00 [a, b].
Show that f BV ([a, b]) and Vab (f ) M (b a).
Solution
Let P = {a = x0 < x1 < ... < xn = b} be any partition of [a, b]. Then
Vab (f, P)
n
X
|f (xk ) f (xk1 )|
k=1
n
X
(xk xk1 )
k=1
M (xn x0 ) = M (b a).
This implies that
Vab (f ) = sup Vab (f, P) M (b a) < .
Problem 93
Let f be a real-valued function on [a, b]. Suppose f is continuous on [a, b] and
is differentiable on (a, b) with |f 0 | M for some constant M > 0. Show that
f BV ([a, b]) and Vab (f ) M (b a).
Hint:
Show that f satisfies the Lipschitz condition.
Problem 94
Let f be a real-valued function on [0, 2 ] defined by
(
sin x1 for x (0, 2 ]
f (x) =
0
for x = 0.
Show that f
/ BV ([0, 2 ]).

114
Solution
Let us choose a particular partition of [0, 2 ]:
x1 =
2
2
2
> x2 =
> ... > x2n1 =
> x2n = 0.
+ 2
+ 2n.2
Then we have
2
V0 (f, P) = |f (x1 ) f (x2 )| + |f (x2 ) f (x3 )| + ... + |f (x2n1 ) f (x2n )|

= 2| + 2 +
{z... + 2} +1 = (2n 1)2 + 1.
2n1
Therefore,
2
sup V0 (f, P) = ,
P
where the supremum is taken over all partitions of [0, 2 ]. Thus, f is not a BV
function.
Problem 95
Let f be a real-valued continuous and BV function on [0, 1]. Show that
2
n
X
i
i 1
lim
f n f
= 0.
n
n
i=1
Solution
Since f is continuous on [0, 1], which is compact, f is uniformly continuous on [0, 1].
Hence,
> 0, N > 0 : |x y|
1
|f (x) f (y)| , x, y [0, 1].
N
Partition of [0, 1]:

x0 = 0 < x 1 =
1
2
n
< x2 = < ... < xn = = 1.
n
n
n
= 1 . Hence,
For n N we have ni i1
n
n
N

i
i
1
f
, i = 1, 2, ...
f
n
n

115
Now we can write, for n N ,

1
i
f
=
n
n
i
i
1
i
i
1
f
. f
n
n
n
n

i
i 1
f
f
,
n
n
and so
n
n
X
X
f i f i 1 V01 (f ).
f i f i 1
n
n
n
n
i=1
i=1
Since V01 (f ) < , we can conclude that
2
n
X
i
i
1
f
= 0.
f
lim
n
n
n
i=1
Problem 96
Let (fi : i N) and f be real-valued functions on an interval [a, b] such that
limi fi (x) = f (x) for x [a, b]. Show that
Vab (f ) lim inf Vab (fi ).
i
Solution
Let Pn = {a = x0 < x1 < ... < xn = b} be a partition of [a, b]. Then
Vab (f, Pn )
Vab (fi , Pn ) =
n
X
|f (xk ) f (xk1 )|,
k=1
n
X
|fi (xk ) fi (xk1 )| for each i N.
k=1
Consider the counting measure space (N, P(N), ) where is the counting measure.
Let D = {1, 2, ..., n}. Then D P(N). Define
gi (k) = |fi (xk ) fi (xk1 )| 0,
g(k) = |f (xk ) f (xk1 )| for k D.

116
Since limi fi (x) = f (x) for x [a, b], we have

lim gi (k) = g(k) for every k D.
By Fatous lemma,
Z
Z
Z
g(k)d =
lim gi (k)d lim inf
gi (k)d. ()
Since D =
Fn
D i
k=1 {k}
(union of disjoint sets), we have

Z
n Z
X
g(k)d =
g(k)d
D
k=1
n
X
k=1
n
X
{k}
g(k)
|f (xk ) f (xk1 )|
k=1
Vab (f, Pn ).
=
Similarly, we get
Z
D
gi (k)d = Vab (fi , Pn ) for each i N.
With these, we can rewrite (*) as follows:

Vab (f, Pn ) lim inf Vab (fi , Pn ).
i
By taking all partitions Pn , we obtain

Vab (f ) lim inf Vab (fi ).
i
Problem 97
Let f be a real-valued absolutely continuous function on [a, b]. If f is never zero,
show that f1 is also absolutely continuous on [a, b].
Solution
The function f is continuous on [a, b], which is compact, so f has a minimum on it.
Since f is non-zero, there is some m (0, ) such that
min |f (x)| = m.
x[a,b]

117
Given any > 0 there exists > 0 such that for any finite family of non-overlapping
closed
= 1, ..., n} in [a, b] such that
Pn intervals {[ai , bi ] : i P
n
i=1 (bi ai ) < we have
i=1 |f (ai ) f (bi )| < . Now,
n
n
X
X
1
1
|f (ai ) f (bi )|
f (ai ) f (bi )
|f (ai )f (bi )|
i=1
i=1
n
1 X
|f (ai ) f (bi )|
m2 i=1
.
m2
Problem 98
Let f be a real-valued function on [a, b] satisfying the Lipschitz condition on [a, b].
Show that f is absolutely continuous on [a, b].
Solution
The Lipschitz condition on [a, b]:
K > 0 : x, y [a, b], |f (x) f (y)| K|x y|.
Given any > 0, let = K . Let {[c
Pi ,ndi ] : i = 1, ..., n} be a family of nonoverlapping subintervals of [a, b] with
i=1 (di ci ) < , then, by the Lipschitz
condition, we have
n
X
|f (ck ) f (dk )|
i=1
n
X
K(dk ck )
i=1
n
X
(dk ck )
i=1
< K
Thus f is absolutely continuous on [a, b].
= .
K
Problem 99
Show that if f is continuous on [a, b] and f 0 exists on (a, b) and satisfies
|f 0 (x)| M for x (a, b) with some M > 0, then f satisfies the Lipschitz
condition and thus absolutely continuous on [a, b].

118
(Hint: Just use the Intermediate Value Theorem.)
Problem 100
Let f be a continuous function on [a, b]. Suppose f 0 exists on (a, b) and satisfies
|f 0 (x)| M for x (a, b) with some M > 0. Show that for every E [a, b] we
have
L f (E) M L (E).
Solution
Recall:
(
L (E) = inf
`(In ) : In are open intervals and
n=1
)
In E
n=1
Let E [a, b]. Let {In = (a0n , b0n )} be a covering of E, where each (a0n , b0n ) [a, b].
Then
[
[
E (a0n , b0n ) f (E)

f (a0n , b0n ) .
Since f is continuous, f (a0n , b0n ) must be an interval. So
f (a0n , b0n ) = f (an ), f (bn ) for an , bn (a0n , b0n ).

Hence,
f (E)
f (an ), f (bn ) .
n
Therefore { f (an ), f (bn ) } is a covering of f (E). By the Mean Value Theorem,
` f (an ), f (bn )
= |f (bn ) f (an )|
= |f 0 (x)||bn an |, x (an , bn )
M |bn an |.
It follows that
X
X
X
` f (an ), f (bn )
M
|bn an | M
|b0n a0n |
n
M
` (a0n , b0n ) .

119
Thus,
inf
X
X
` f (an ), f (bn ) M inf

` (a0n , b0n ) .
n
The infimum is taken over coverings of f (E) and E respectively. By definition (at
the very first of the proof) we have
L f (E) M L (E).
Problem 101
Let f be a real-valued function on [a, b] such that f is absolutely continuous on
[a + , b] for every (0, b a). Show that if f is continuous and of bounded
variation on [a, b], then f is absolutely continuous on [a, b].
Solution
Using the Banach-Zaracki theorem, to show that f is absolutely continuous on [a, b],
we need to show that f has property (N) on [a, b]. Suppose E [a, b] such that
L (E) = 0. Given any > 0, since f is continuous at a+ , there exists (0, b a)
such that
a x a + |f (x) f (a)| < . ()

2
Let E1 = E [a, a + ] and E2 = E \ E1 . Then E = E1 E2 and so f (E) =
f (E1 ) f (E2 ). But E2 [a + , b) and f is absolutely continuous on [a + , b], so f
has property (N) on this interval. Since E2 E, we have L (E2 ) = 0. Therefore,
L (f (E2 )) = 0 = L (f (E2 )).
On the other hand,
x E1
x [a, a + )
f (a) f (x) f (a) +

by ()
2
2
f (E1 ) [ f (a) , f (a) + ]

2
2
L (f (E1 )) .
Thus,
L (f (E)) L (f (E1 )) + L (f (E2 )) .
Since > 0 is arbitrary, L (f (E)) = 0 and so L (f (E)) = 0.

120
Problem 102
Let f be a real-valued integrable function on [a, b]. Let
Z
F (x) =
f dL , x [a, b].
[a,x]
Show that F is continuous and of bounded variation on [a, b].

Solution
The continuity follows from Theorem 18 (absolute continuity implies continuity).
To show that F is of BV on [a, b], let a = x0 < x1 < ... < xn = b be any partition
of [a, b]. Then
n
X
n Z
X
|F (xi xi1 | =
i=1
[xi1 ,xi ]
i=1
Z
n
X
f dL
|f |dL
Zi=1
=
[xi1 ,xi ]
|f |dL .
[a,b]
Thus, since |f | is integrable,

Z
Vab (F )
|f |dL < .
[a,b]
Chapter 10
Lp Spaces
1. Norms
For 0 < p < :
Z
kf kp =
1/p
|f |p d
.
For p = :
kf k = inf M [0, ) : {x X : |f (x)| > M } = 0 .
Theorem 22 Let (X, A, ) be a measure space. Then the linear space Lp (X) is a Banach space
with respect to the norm k.kp for 1 p < or the norm k.k for p = .
2. Inequalities for 1 p <
1. Holders inequality: If p and q satisfy the condition p1 + 1q = 1, then for f Lp (X), g Lq (X),
we have
Z
1/p Z
1/q
Z
p
q
|f g|d =
|f | d
|g| d
,
or
kf gk1 kf kp kgkq .
In particular,
kf gk1 kf k2 kgk2 (Schwarzs inequality).
2. Minkowskis inequality: For f, g Lp (X), we have
kf + gkp kf kp + kgkp .
3. Convergence
Theorem 23 Let (fn ) be a sequence in Lp (X) and f an element in Lp (X) with 1 p < . If
fn f in Lp (X), i.e., kfn f kp 0, then
(1) kfn kp kf kp ,
(2) fn
f on X,
(3) There exists a subsequence (fnk ) such that fnk f a.e. on X.
121- Math Vietnam

www.MathVn.com

122
LP SPACES
CHAPTER 10.
fn f a.e. on X and kfn kp kf kp , then kfn f kp 0.
fn
f on X and kfn kp kf kp , then kfn f kp 0.
kfn f k 0, then
(1) kfn k kf k ,
(2) fn f uniformly on X \ E where E is a null set.
(3) fn
f on X.
Problem 103
Let f be a Lebesgue measurable function on [0, 1]. Suppose 0 < f (x) < for all
x [0, 1]. Show that
Z
Z
1
f d
d 1.
[0,1] f
[0,1]
Solution
The functions f and 1f are Lebesgue measurable since f is Lebesgue measurable

and 0 < f < . By Schwarzs inequality, we have
Z
1=
1d =
[0,1]
[0,1]
1
f d
f
Z
Z
p 2 1/2
f d
[0,1]
[0,1]
1/2 Z
f d
[0,1]
[0,1]
1
d
f
Squaring both sides we get

Z
Z
f d
[0,1]
[0,1]
1
d
f
1.
1/2
.
!1/2
2
d

123
Problem 104
Let (X, A, ) be a finite measure space. Let f Lp (X) with p (1, ) and q its
conjugate. Show that
Z
|f |d (X)
1
q
Z
p
|f | d
p1
Hint:
Write
f = f 1X
where 1X is the characteristic function of X, then apply the Holders inequality.
Problem 105
Let (X, A, ) be a finite measure space.
(1) If 1 p < show that L (X) Lp (X).
(2) If 1 p1 < p2 < show that Lp2 (X) Lp1 (X).
Solution
(1) Take any f L (X). Then kf k < . By definition, we have |f | kf k a.e.
on X. So we have
Z
Z
p
|f | d
kf kp d = (X)kf kp .
X
By assumption, (X) < . Thus
R
X
|f |p d < . That is f Lp (X).
(2) Consider the case 1 p1 < p2 < . Take any f Lp2 (X). Let := pp21 . Then
1 < < . Let (1, ) be the conjugate of , that is, 1 + 1 = 1. By the
Holders inequality, we have
Z
Z
p2 1/
p1
|f | d =
|f |
1X d
X
1/ Z
1/
p2
|f | d
X
kf kpp22 / (X)
|1X | d
X
< ,
since kf kp2 < and (X) < . Thus f Lp1 (X).

124
CHAPTER 10.
LP SPACES
Problem 106 (Extension of Holders inequality)

Let (X, A, ) be an arbitrary measure space. Let f1 , ..., fn be extended complexvalued measurable functions on X such that |f1 |, ..., |fn | < a.e. on X. Let
p1 , ..., pn be real numbers such that
p1 , ..., pn (1, ) and
1
1
+ ... +
= 1.
p1
pn
Prove that
kf1 ...fn k1 kf1 kp1 ...kfn kpn .
()
Hint:
Proof by induction. For n = 2 we have already the Holders inequality.
Assume that () holds for n 2. Let
q=
1
1
+ ... +
p1
pn
Then
q, pn+1 (0, ) and
1
.
1
1
+
= 1.
q
pn+1
Keep going this way.
Problem 107
Let (X, A, ) be an arbitrary measure space. Let f1 , ..., fn be extended complexvalued measurable functions on X such that |f1 |, ..., |fn | < a.e. on X. Let
p1 , ..., pn and r be real numbers such that
p1 , ..., pn , r (1, ) and
1
1
1
+ ... +
= .
p1
pn
r
Prove that
kf1 ...fn kr kf1 kp1 ...kfn kpn .
Solution
We can write (i) as follows:
1
1
+ ... +
= 1.
p1 /r
pn /r
(i)

125
From the extension of Holders inequality (Problem 105) we have
|f1 ...fn |r |f1 |r ...|fn |r

. (ii)
1
p1 /r
pn /r
Now we have
|f1 ...fn |r =
1
Z
X
|f1 ...fn |r d = kf1 ...fn krr ,
and for i = 1, ..., n we have

r
|fi |
Z
pi /r
|fi |
pi
r
r/pi
d
r/pi
pi
|fi | d
X
= kfi krpi .
By substituting these expressions into (ii), we have

kf1 ...fn krr kf1 krp1 ...kfn krpn .
Taking the r-th roots both sides of the above inequality we obtain (i).
Problem 108
Let (X, A, ) be a measure space. Let (0, 1) and let p, q, r 1 with p, q r
be related by
1
1
= +
.
r
p
q
Show that for every extended complex-valued measurable function on X we have
kf kr kf kp kf k1
q .
Solution
Recall: (Extension of Holders inequality)
1
1
1
=
+ ... +
kf1 ...fn kr kf kp1 ...kfn kpn .
r
p1
pn
For n = 2 we have
Now, we have
1 1
1
= +
kf gkr kf kp kgkq .
r
p q
1
1
1
1
= +
=
+
.
r
p
q
p/ q/(1 )

126
CHAPTER 10.
Applying the above formula we get

kf kr = |f | .|f |1 |f |

|f |
=
Z
=
And
p/
/p
/p
|f |
X
kf kp .
Z
q/(1)
|f |
1
|f |
1
|f |
.
. ()
p/
q/(1)
Z
p/
LP SPACES
1 q/1
|f |
=
Z
1/q
1/q
|f |
X
kf k1
q .
Plugging into (*) we obtain

kf kr kf kp .kf k1
q .
Problem 109
Let (X, A, ) be a measure space. Let p, q [1, ] be conjugate. Let (fn )nN
Lp (X) and f Lp (X) and similarly (gn )nN Lq (X) and g Lq (X). Show
that
h
i
lim kfn f kp = 0 and lim kgn gkq = 0 lim kfn gn f gk1 = 0.
n
Solution
We use Holders inequality:
kfn gn f gk1 =
|fn gn f g|d
Z
(|fn gn fn g| + |fn g f g|)d

Z
Z
|fn ||gn g|d +

|g||fn f |d
X
kfn kp .kgn gkq + kgkq .kfn f kp . ()

127
By Minkowskis inequality, we have
kfn kp kf kp + kfn f kp .
Since kf kp and kfn f kp are bounded (why?), kfn kp is bounded for every n N.
From assumptions we deduce that limn kfn kp .kgn gkq = 0.
Since kgkq is bounded, from assumptions we get limn kgkq .kfn f kp = 0. Therefore, from (*) we obtain
lim kfn gn f gk1 = 0.
Problem 110
Let (X, A, ) be a measure space and let p [1, ). Let (fn )nN Lp (X) and
f Lp (X) be such that limn kfn f kp = 0. Let (gn )nN be a sequence of
complex-valued measurable functions on X such that |gn | M for every n N
and let g be a complex-valued measurable function on X such that limn gn = g
a.e. on X. Show that
lim kfn gn f gkp = 0.
n
Solution
We first note that |g| M a.e. on X. Indeed, we have for all n N,
|g| |gn g| + |gn |.
Since |gn | M for every n N and |gn g| 0 a.e. on X by assumption. Hence
|g| M a.e. on X.
Now, by Minkowskis inequality, we have
kfn gn f gkp kfn gn f gn kp + kf gn f gkp
kgn (fn f )kp + kf (gn g)kp
kgn (fn
Z
f )kpp
|gn (fn f )|p d
=
ZX
|gn |p .|fn f |p d
X
p
M kfn f kpp .
()

128
CHAPTER 10.
LP SPACES
Since kfn f kp 0 by assumption, we have that kgn (fn f )kp 0.

Let hn = f gn f g for every n N. Then
|hn | |f |.|gn g| |f |(|gn | + |g|) 2M |f |
|hn |p 2p M p |f |p < .
Now, |hn |p is bounded and |hn |p |f |p .|gn g|p |hn |p 0 (since gn g a.e.).
By the Dominated Convergence Theorem, we have
Z
Z
p
0=
lim |hn | d = lim
|hn |p d
n
n
X
ZX
= lim
|f gn f g|p d
n
lim kf (gn g)kpp .
From these results, (*) gives that

lim kfn gn f gkp = 0.
Problem 111
RLet f bep an extended real-valued Lebesgue measurable function on [0, 1] such that
|f | d < for some p [1, ). Let q (1, ] be the conjugate of p. Let
[0,1]
a (0, 1]. Show that
Z
1
lim
|f |d = 0.
a0 a1/q [0,a]
Solution
p=1
Since q = , we have to show
Z a
lim
|f (s)|ds = 0 (Lebesgue integral = Riemann integral).
a0
This is true since f is integrable so

1<p<
Then 1 < q < . We have
Z a
Z
|f (s)|ds =
0
Ra
0
|f (s)|.1 ds
0
([0, a])
1/p
1/q
|f (s)|ds
0
Z
= a
|f (s)|ds is continuous with respect to a.
1/q
|f (s)|ds
(Problem 104)
1/p
.

129
Hence,
|f (s)|ds
a1/q
1/p
|f (s)|ds
()
Since |f | is integrable, we have1 (Problem 66)

Z
|f |d < p .
> 0, > 0 : ([0, a]) <

[0,a]
Equivalently,
Z
> 0, > 0 : 0 < a <
1/p
|f (s)|ds
< . ()

> 0, > 0 : 0 < a <
That is,
lim
a0
1
a1/q
1
a1/q
|f (s)|ds < .
0
|f (s)|ds = 0.
0
Problem 112
Let (X, A, ) be a finite measure space. Let fn , f L2 (X) for all n N such
that limn fn = f a.e. on X and kfn k2 M for all n N.
Show that limn kfn f k1 = 0.
Solution
We first claim: kf k2 M . Indeed, y Fatous lemma, we have
Z
Z
2
2
kf k2 =
|f | d lim inf
|fn |2 d M 2 .
X
Since (X) < , we can use Egoroffs theorem:

> 0, A A with (A) < 2 and fn f uniformly on X \ A.
Now we can write
kfn f k1 =
|fn f |d =
X
Z
|fn f |d +
A
|fn f |d.
X\A
This is called the uniform continuity of the integral with respect to the measure .

130
CHAPTER 10.
LP SPACES
R
On X \ A, fn f uniformly, so for large n, we have X\A |fn f |d < . On A we
have
Z
Z
|fn f |d =
|fn f |A d (A)1/2 .kfn f k2
A
(A)1/2 (kfn k2 + kf k2 )
2M (since (A) < 2 ).
Thus, for any > 0, for large n, we have
kfn f k1 (2M + 1).
This tells us that limn kfn f k1 = 0.
Problem 113
Let (X, A, ) be a finite measure space and let p, q (1, ) be conjugates. Let
fn , f Lp (X) for all n N such that limn fn = f a.e. on X and kfn kp M
for all n N. Show that
(a) kf kp M .
(b) limn kf
R n f kp =R0.
(c) limn RX fn gd =R X f gd for every g Lq (X).
(d) limn E fn d = E f d for every E A.
Hint:
(a) and (b): See Problem 112.
(c) Show kfn g f gk1 kfn f kp kgkq . Then use (b).
(d) Write
Z
Z
Z
fn g =
fn g1E =
fn (g1E ).
E
Then use (c).
Problem 114
Let (X, A, ) be a measure space. Let f be a real-valued measurable function on
X such that f L1 (X) L (X). Show that f Lp (X) for every p [1, ].
Hint:
If p = 1 or p = , there is nothing to prove.
Suppose p (1, ). Let f L1 (X) L (X). Write
|f |p = |f |1 |f |p1 |f |.kf kp1
.
Integrate over X, then use the fact that kf k1 and kf k are finite.

131
Problem 115
Let (X, A, ) be a measure space and let 0 < p1 < p < p2 . Show that
Lp (X) Lp1 (X) + Lp2 (X),
that is, if f Lp (X) then f = g +h for some g Lp1 (X) and some h Lp2 (X).
Solution
For any f Lp (X), let D = {X : |f | 1}. Let g = f 1D and h = f 1Dc . Then
g + h = f 1D + f 1Dc = f (1D + 1Dc ) = f (See Problem 37).
| {z }
=1DDc
We want to show g Lp1 (X) and h Lp2 (X).

On D we have : 1 |f |p1 |f |p |f |p2 . It follows that
Z
p1
p1
|g| d =
X
|f |p d < since f Lp (X).
|f | d
D
Hence, g Lp1 (X).

On Dc we have : |f |p1 |f |p |f |p2 . It follows that
Z
p2
p2
|h| d =
X
|f |p d < .
|f | d
Dc
Hence, h Lp2 (X). This completes the proof.
Problem 116
Given a measure space (X, A, ). For 0 < p < r < q , show that
Lp (X) Lq (X) Lr (X).
Hint:
Let D = {X : |f | 1}. On D we have |f |r |f |q , and on X \ D we have |f |r |f |p .

132
CHAPTER 10.
LP SPACES
Problem 117
Suppose f L4 ([0, 1]), kf k4 = C 1 and kf k2 = 1. Show that
1
kf k4/3 1.
C
Solution
First we note that 4 and 4/3 are conjugate. By assumption and by Holders inequality we have
Z
Z
2
2
1 = kf k2 =
|f | d =
|f |.|f |d
[0,1]
[0,1]
kf k4 .kf k4/3
C.kf k4/3 .
This implies that kf k4/3 C1 . ().
By Schwrarzs inequality we have
Z
Z
4/3
4/3
kf k4/3 =
|f | d =
[0,1]
|f |.|f |1/3 d
[0,1]
1/3
kf k2 .kf k2
= 1 since kf k2 = 1.
Hence, kf k4/3 1. ()
1
kf k4/3 1.
C
Problem 118
Let (X,
be a measure space with (X) (0, ). Let f L (X) and let
R A, )
n = X |f |n d for n N. Show that
n+1
= kf k .
n n
lim
Solution
We first note that if kf k = 0, the problem does not make sense. Indeed,
kf k = 0 f 0 a.e. on X
n = 0, n N.

133
Suppose that 0 < kf k < . Then n > 0, n N. We have
Z
Z
n+1
n+1 =
|f | d =
|f |n |f |d
X
X
Z
kf k .
|f |n d = kf k n .
X
This implies that

n+1
kf k , n N.
n
n+1
lim sup
kf k . ()
n
n
Notice that
n+1
n
and n + 1 are conjugate. Using again Holders inequality, we get

Z
Z
n
n =
|f | .1 d
X
(|f | )
Z
n+1
n
(|f |
n+1
)d
n Z
n+1
n
n+1
n+1
1
n+1
X
1
.(X) n+1
X
n
n+1
= n+1 .(X) n+1 .

With a simple calculation we get
1
1
n+1
n+1
n+1
.(X) n+1 , n N.
n
Given any > 0, let E = {X : |f | > kf k }, then, by definition of kf k , we

have (E) > 0. Now,
1
n+1
Z
1
n+1
n+1
(|f | )d
n+1 =
Z
(|f |n+1 )d
1
n+1
E
1
> (E) n+1 .(kf k ).

It follows that
1
(E) n+1
n+1
(kf k ).
.
n
(X)
n+1
kf k , > 0
lim inf
n
n
n+1
lim inf
kf k . ()
n
n

134
CHAPTER 10.
LP SPACES

n+1
= kf k .
n n
lim
Problem 119
Let (X, A, ) be a measure space and p [1, ).
Let f Lp (X) and (fn : n N) Lp (X). Suppose limn kfn f kp = 0.
Show that for every > 0, there exists > 0 such that for all n N we have
Z
|fn |p d < for every E A such that (E) < .
E
Solution
By assumption we have limn kfn f kpp = 0. Equivalently,
> 0, N N : n N kfn f kpp <
2p+1
. (1)
From triangle inequality we have2

|fn | |fn f | + |f |,
|fn |p (|fn f | + |f |)p 2p |fn f |p + 2p |f |p .
Integrating over E A and using (1), we get for n N ,
Z
Z
Z
p
p
p
p
|fn | d 2
|fn f | d + 2
|f |p d
E
E
E
Z
2p kfn f kpp + 2p
|f |p d
E
Z
2p . p+1 + 2p
|f |p d
2
E
Z
+ 2p
|f |p d.
(2)
=
2
E
2
In fact, for a, b 0 and 1 p < we have

(a + b)p 2p1 (ap + bp ).

135
Since |f |p is integrable, by the uniform absolute continuity of integral (Problem 66)
we have
Z
0 > 0 : (E) < 0

|f |p d < p+1 .
2
E
So, for n N , from (2) we get
Z
|fn |p d
0 > 0 : (E) < 0

E
+ 2p . p+1 = . (3)
2
2
Similarly, all |f1 |p , ..., |fN 1 |p are integrable, so we have

Z
j > 0 : (E) < j
|fj |p d < , j = 1, ..., N 1. (4)
E
Let = min{0 , 1 , .., N 1 }. From (3) and (4) we get for every n N,
Z
> 0 : (E) <
|fn |p d < .
E
Problem 120
Let
f be a bounded real-valued integrable function on [0, 1].
R
xn f d = 0 for n = 0, 1, 2, .... Show that f = 0 a.e. on [0, 1].
[0,1]
Suppose
Solution
Fix an arbitrary function C[0, 1]. By the Stone-Weierstrass theorem, there is a
sequence (pn ) of polynomials such that k pn k 0 as n . Then
Z
f
d
=
f
(
p
+
p
)d
n
n
[0,1]
[0,1]
Z
|f | | pn |d +
f pn d
[0,1]
Z [0,1]
kf k1 k pn k +
f pn d
[0,1]
{z
}
|
=0 by hypothesis
= kf k1 k pn k .
Since kf k1 < and k pn k 0, we have
Z
f d = 0, C[0, 1].
()
[0,1]

136
CHAPTER 10.
LP SPACES
Now, since C[0, 1] is dense in L1 [0, 1], there exists a sequence (n ) C[0, 1] such
that kn f k1 0 as n . Then
Z
0
f 2 d =
f (f n + n )d
[0,1]
Z [0,1]
Z
|f | |f n |d +
f n d
[0,1]
[0,1]
|
{z
}
=0 by (*)
kf k kf n k1 .
Since kf k < and kf n k1 0, we have
Z
f 2 d = 0.
[0,1]
Thus f = 0 a.e. on [0, 1].
Problem 121
Let (X, A, ) be a -finite measure space with (X) = .
(a)
S Show that there exists a disjoint sequence (En : n N) in A such that
nN En = X and (En ) [1, ) for every n N.
(b) Show that there exists an extended real-valued measurable function f on X
such that f
/ L1 (X) and f Lp (X) for all p (1, ].
Solution
(a) Since (X, A, ) is a -finite measure space, there exists a sequence (An : n N)
of disjoint sets in A such that
[
X=
An and (An ) < , n N.
nN
By the countable additivity and by assumption, we have

X
(X) =
(An ) = .
nN
It follows that
k1 N : 1
k1
X
(An ) = (A1 ... Ak1 ) < .
n=1

137
Let E1 = A1 ... Ak1 then we have
1 (E1 ) < and (Ak1 +1 Ak1 +2 ...) = (X \ E1 ) = .
Then there exists k2 k1 + 1 such that
1 (Ak1 +1 ... Ak2 ) < .
Let E1 = Ak1 +1 ... Ak2 then we have
1 (E2 ) < and E1 E1 = .
And continuing this process we are building a sequence (En : n N) of disjoint
subsets in A satisfying
[
[
En =
An = X and (En ) [1, ), n N.
nN
nN
(b) Define a real-valued function f on X =

f=
X
n=1
Then
f |A1 =
Hence,
nN
An by
An
.
n(An )
1
1
, ..., f |An =
, ...
1(A1 )
n(An )
Z
f d =
X
Z
X
n=1
X
1
f d =
= .
n
An
n=1
That is f
/ L (X).
We also have
f p |A1 =
1
1
p
,
...,
f
|
=
, ...(1 < p < )
A
n
1p (A1 )p
np (An )p
By integrating
Z
p
f d =
X
Z
X
n=1
Thus, f L (X).
An
np (An )p1
n=1
X
n=1
f p d
1
< . since (An )p1 1.
np

138
CHAPTER 10.
LP SPACES
Problem 122
Consider the space Lp ([0, 1]) where p (1, ].
(a) Prove that kf kp is increasing in p for any bounded measurable function f .
(b) Prove that kf kp kf k when p .
Solution
(a)
Suppose 1 < p < . We want to show kf kp kf k .
By definition, we have
|f | kf k a.e. on [0, 1].
Therefore,
|f |p kf kp a.e. on [0, 1].
Z
Z
p
kf kp d
|f | d
[0,1]
kf kpp
[0,1]
kf kp
([0, 1])
kf kp kf k .
Suppose 1 < p1 < p2 < . We want to show kf kp1 kf kp2 .
Notice that
p1 p2 p1
1
1
+
= 1 or
+
= 1.
p2
p2
p2 /p1 p2 /(p2 p1 )
By Holders inequality we have
Z
Z
p1
p1
kf kp1 =
|f | d =
|f |p1 .1.d
[0,1]
[0,1]
p1
k|f | kp2 /p1 .k1kp2 /(p2 p1 )

= kf kpp12 /p1 . ()
Now,
Z
kf kpp12 /p1
|f |
Z
p1 . p2
1
p1 /p2
d
[0,1]
p2
|f | d
[0,1]
p1 . p1
= kf kpp12 .
Finally, (*) implies that kf kp1 kf kp2 .

In both cases we have
1 < p1 < p2 = kf kp1 kf kp2 .

139
That is kf kp is increasing in p.
(b) By part (a) we get kf kp kf k . Then
lim sup kf kp kf k .
(i)
Given any > 0, let E = {X : |f | > kf k }. Then (E) > 0 and

Z
p
kf kp
|f |p d > (kf k )p (E).
E
kf kp (kf k )(E)1/p
lim inf kf kp kf k , > 0 (since
p
lim (E)1/p = 1).
lim inf kf kp kf k . (ii)

p
From (i) and (ii) we obtain

lim kf kp = kf k .

APPENDIX
The Lp Spaces for 0 < p < 1
Let (X, A, ) be a measure space and p (0, 1). It is easy to check that Lp (X) is a linear space.
R
1/p
Exercise 1. If kf kp := X |f |p d
and 0 < p < 1, then k.kp is not a norm on X.
Hint:
Show that k.kp does not satisfy the triangle inequality:
Take X = [0, 1] with the Lebesgue measure on it. Let f = 1[0, 12 ) and g = 1[ 12 ,1) . Then show that
kf + gkp = 1.
and that
kf kp = 2 p and kgkp = 2 p .
It follows that
kf + gkp > kf kp + kgkp .
Exercise 2. If , C and 0 < p < 1, then
| + |p ||p + ||p .
Hint:
Consider the real-valued function (t) = (1 + t)p 1 tp , t [0, ). Show that it is strictly
||
> 0.
decreasing on [0, ). Then take t = ||

140
CHAPTER 10.
Exercise 3. For 0 < p < 1, k.kp is not a norm. However

Z
p (f, g) :=
|f g|p d, f, g Lp (X)
X
is a metric on Lp (X).
Proof.
We prove only the triangle inequality. For f, g, h Lp (X), we have
Z
|f g|p d
p (f, g) =
ZX
=
|(f h) + (h g)|p d
X
Z
(|f h| + |h g|)p d
X
Z
Z
|f h|p d +
|h g|p d (by Exercise 2)
X
p (f, h) + p (h, g).
LP SPACES
Chapter 11
Integration on Product Measure
Space
1. Product measure spaces
Definition 32 (Product measure)
Given n measure spaces (X1 , A1 , 1 ), ..., (Xn , An , n ). Consider the product measurable space
X1 ... Xn , (A1 ... An ) . A measure on (A1 ... An ) such that
(E) = 1 (A1 )...n (An ) for E = A1 ... An A1 ... An
with the convention .0 = 0 is called a product measure of 1 , ..., n and we write
= 1 ... n .
Theorem 27 (Existence and uniqueness)
For n arbitrary measure spaces (X1 , A1 , 1 ), ..., (Xn , An , n ), a product measure space X1 ...
Xn , (A1 ... An ), 1 ... n exists. Moreover, if the n measure spaces are all -finite, then
the product measure space is unique.
2. Integration
Definition
33 (Sections and
section functions)
Let X Y, (A B), be the product of two -finite measure spaces (X, A, ) and (Y, B, ).
Let E X Y , and f be an extended real-valued function on E.
(a) For x X, the set E(x, .) := {y Y : (x, y) E} is called the x-section of E.
For y Y , the set E(., y) := {x X : (x, y) E} is called the y-section of E.
(b) For x X, the function f (x, .) defined on E(x, .) is called the x-section of f .
For y Y , the function f (., y) defined on E(., y) is called the y-section of f .
Proposition 24 Let XY, (AB), be the product

of two -finite measure spaces (X, A, )
and
(Y,
B,
).
For
every
E
(A
B),
E(x,
.)
is
a
A-measurable function of x X and
E(., y) is a B-measurable function of y Y . Furthermore, we have

Z
Z
( )(E) =
E(x, .) (dx) =
E(., y) (dy).
X
141- Math Vietnam

www.MathVn.com

142
CHAPTER 11. INTEGRATION ON PRODUCT MEASURE SPACE
Theorem 28 (Tonellis Theorem)

Let (X Y, (A B), ) be product measure space of two -finite measure spaces. Let f be a
non-negative extended real-valued measurable on X Y . Then
(a) F 1 (x) :=
R
R
f (x, .)d is a A-measurable function of x X.
(b) F 2 (y) := X f (., y)d is a B-measurable function of y Y .

R
R
R
(c) XY f d( ) = X F 1 d = Y F 2 d, that is,
Z Z
f d( ) =
XY
Z Z
f (x, .)d d =
X
f (., y)d d.
Theorem 29 (Fubinis Theorem)

Let (X Y, (A B), ) be product measure space of two -finite measure spaces. Let f be a
-integrable extended real-valued measurable function on X Y . Then
(a) The B-measurable function f (x, .) is -integrable on Y for -a.e.
measurable function f (., y) is -integrable on X for -a.e. y Y .
R
(b) The function F 1 (x) := Y f (x, .)d is defined for -a.e. x X,
integrable on X.
R
The function F 2 (y) := X f (., y)d is defined for -a.e. y X,
integrable on Y .
R
R
R
(c) We have the equalities: XY f d( ) = X F 1 d = Y F 2 d, that
Z Z
f d( ) =
XY
Z Z
f (x, .)d d =
X
x X and the AA-measurable and B-measurable and is,
f (., y)d d.
Y
Problem 123
Consider the product measure space R R, (BR BR ), L L .

Let D = {(x, y) R R : x = y}. Show that
D (BR BR ) and (L L )(D) = 0.
Solution
Let = L L . Let D0 = {(x, y) [0, 1] [0, 1] : x = y}. For each n Z let

143
Dn = {(x, y) [n, n + 1] [n, n + 1] : x = y}. Then, by translation invariance of
Lebesgue measure, we have
(D0 ) = (Dn ), n N.
[
and D =
Dn .
nZ
To solve the problem, it suffices to prove D0 (BR BR ) and (D0 ) = 0.

For each n N, divide [0, 1] into 2n equal subintervals as follows:
1
1 2
2 1
In,1 = 0, n , In,2 = n , n , ..., In,2n =
,1 .
2
2 2
2n
Sn
Let Sn = 2k=1 (In,k In,k ), then D0 = limn Sn .
Now, for each n N and for k = 1, 2, ..., 2n , In,k BR . Therefore,
In,k In,k (BR BR ) and so Sn (BR BR ).
Hence, D0 (BR BR ).
It is clear that (Sn ) is decreasing (make a picture yourself), so
D0 = lim Sn =
n
Sn .
n=1
And we have
n
(Sn ) =
=
2
X
k=1
2n
X
k=1
(In,k In,k )
1 1
1
n 1
.
=
2
.
=
.
2n 2n
22n
2n
It follows that
(D0 ) (Sn ) =
Thus, (D0 ) = 0.
1
, n N.
2n
Problem 124
Consider the product measure space (R R, (BR BR ), L L ). Let f be a
real-valued function of bounded variation on [a, b]. Consider the graph of f :
G = {(x, y) R R : y = f (x) for x R}.
Show that G (BR BR ) and (L L )(G) = 0.

144
Hint:
Partition of [a, b]:
P = {a = x0 < x1 < ... < xn = b}.
Elementary rectangles:
Rn,k = [xk1 , xk ] [mk , Mk ], k = 1, ..., n,
where
mk =
inf
x[xk1 ,xk ]
Let
Rn =
n
[
f (x)
Rn,k
and Mk =
sup
f (x).
x[xk1 ,xk ]
and
kP k = max (xk xk1 ).
k=1
1kn
Let = L L . Show that

(Rn ) kP k
n
X
(Mk mk ) kP kVab (f ),
k=1
Problem 125
Let (X, A, ) and (Y, B, ) be the measure spaces given
X = Y = [0, 1]
A = B = B[0,1] , the -algebra of the Borel sets in [0, 1],
= L and is the counting measure.
Consider the product measurable space X Y, (A B) and a subset in it

defined by E = {(x, y) X Y : x = y}. Show that
(a) E (A B),
Z Z
Z Z
(b)
E d d 6=
E d d.
X
Why is Tonellis theorem not applicable?

Solution
(a) For each n N, divide [0, 1] into 2n equal subintervals as follows:
1 2
2 1
1
,1 .
In,1 = 0, n , In,2 = n , n , ..., In,2n =
2
2 2
2n
Sn
Let Sn = 2k=1 (In,k In,k ). It is clear that (Sn ) is decreasing, so
E = lim Sn =
n
Sn .
n=1

145
Now, for each n N and for k = 1, 2, ..., 2n , In,k B[0,1] . Therefore,
In,k In,k (B[0,1] B[0,1] ) and so Sn (B[0,1] B[0,1] ).
Hence, E (B[0,1] B[0,1] ).
(b) For any x X, 1E (x, .) = 1{x} (.). Therefore,
Z
Z
1E d =
1{x} d = {x} = 1.
Hence,
[0,1]
Z Z
1E d d =
X
1d = 1.
()
[0,1]
On the other hand, for every y Y, 1E (., y) = 1{y} (.). Therefore,

Z
Z
1E d =
1{y} d = {y} = 0.
Hence,
[0,1]
Z Z
1E d d =
Y
0d = 0.
()
[0,1]
Thus, from (*) and (**) we get
Z Z
Z Z
1E d d 6=
X
1E d d.
Y
Tonellis theorem requires that the two measures must be -finite. Here, the counting
measure is not -finite, so Tonellis theorem is not applicable.
Question: Why the counting measure on [0, 1] is not -finite?
Problem 126
Suppose g is a Lebesgue measurable real-valued function on [0, 1] such that the
function f (x, y) = 2g(x) 3g(y) is Lebesgue integrable over [0, 1] [0, 1]. Show
that g is Lebesgue integrable over [0, 1].

146
Solution
By Fubinis theorem we have
Z
Z
f (x, y)d(L (x) L (y)) =
[0,1][0,1]
Z
=
f (x, y)dxdy
0
1
=
=
=
=
0
1
[2g(x) 3g(y)]dxdy
Z 1Z 1
Z 1Z 1
2g(x)dxdy
3g(y)dxdy
0
0
0
0
Z 1
Z 1
Z 1
Z 1
2
g(x)
1.dy dx 3
g(y)
1.dx dy
0
0
0
0
Z 1
Z 1
2
g(x).1.dx 3
g(y).1.dy
0
0
Z 1
Z 1
g(y)dy
g(x)dx 3
2
0
0
Z 1
g(x)dx.
0
Since f (x, y) is Lebesgue integrable over [0, 1] [0, 1]:

Z
f (x, y)d(L (x) L (y)) < .
[0,1][0,1]
Therefore,
1
0
g(x)dx < .
That is g is Lebesgue (Riemann) integrable over [0, 1].
Problem 127
Let (X, M, ) be a complete measure space and let f be a non-negative integrable
function on X. Let b(t) = {x X : f (x) t}. Show that
Z
Z
f d =
b(t)dt.
X
Solution
Define F : [0, ) X R by
(
1 if 0 t f (x)
F (t, x) =
0 if t > f (x).

147
If Et = {x X : f (x) t}, then F (t, x) = 1Et (x). We have
Z
Z f (x)
Z
F (t, x)dt =
F (t, x)dt +
F (t, x)dt = f (x) + 0 = f (x).
0
f (x)
By Fubinis theorem we have

Z
Z Z
f d =
X
Z Z
f (x)
!
dt dx
F (t, x)dt dx
Z
=
F (t, x)dx dt
0
X
Z Z
=
1Et (x)dx dt
0
X
Z
=
b(t)dt. (since (Et ) = b(t) ).
ZX
Problem 128
Consider the function u : [0, 1] [0, 1] R defined by
( 2 2
x y
for (x, y) 6= (0, 0),
(x2 +y 2 )2
u(x, y) =
0
for (x, y) = (0, 0).
(a) Calculate
Z
u(x, y)dx dy.
u(x, y)dy dx and

0
Observation?
(b) Check your observation by using polar coordinates to show that
ZZ
|u(x, y)|dxdy = ,
D
that is, u is not integrable. Here D is the unit disk.

Answer.
(a) 4 and 4 .

148
Problem 129
Let
I[0, 1], R+ = [0, ),
1
f (u, v) =
,
1 + u2 v 2
g(x, y, t) = f (x, t)f (y, t), (x, y, t) I I R+ := J.
(a) Show that g is integrable on J (equipped with Lebesgue measure). Using
Tonellis theorem on R+ I I show that
Z
A =:
gdtdxdy =
J
R+
arctan t
t
2
dt.
(b) Using Tonellis theorem on I I R+ show that

Z
1
A=
dxdy.
2 II x + y
(c) Using Tonellis theorem again show that A = ln 2.
Solution
(a) It is clear that g is continuous on R3 , so measurable. Using Tonellis theorem on
R+ I I we have
Z Z
A =
f (x, t)f (y, t)dxdy dt
R+
II

Z
Z Z
f (y, t)dy dx dt
=
f (x, t)
R+
I
I
Z
Z Z
1
1
=
dx
dy
dt
2 2
2 2
R+
I 1+x t
I 1+y t
2
Z Z
1
dx dt
=
2 2
R+
I 1+x t
2
Z
arctan t
=
dt.
t
R+
Note that for all t R+ , 0 < arctan t < 2 and arctan t t as t 0, so
2
Z
arctan t
dt < .
A=
t
R+

149
Thus g is integrable on J.
(b) We first decompose g(x, y, t) = f (x, t)f (y, t) into simple elements:
1
1
.
2
2
1 + x t 1 + y 2 t2
1
x2
y2
= 2
.
x y 2 1 + x2 t2 1 + y 2 t2
g(x, y, t) = f (x, t)f (y, t) =
Using Tonellis theorem on I I R+ we have

Z Z
1
x2
y2
A =
dt dxdy
2
2 1 + x2 t 2
1 + y 2 t2
II
R+ x y
Z

Z
1
x
y
=
ds dxdy
2
2
2
1 + s2
II x y
R+ 1 + s
Z
Z
1
ds
=
dxdy
1 + s2
II x + y
0
Z
1
=
dxdy.
2 II x + y
(c) Using (b) and using Tonellis theorem again we get
Z Z 1
1
1
A =
dy dx
2 0
0 x+y
Z
1
=
[ln(x + 1) ln x]dx
2 0
=
[(x + 1) ln(x + 1) x ln x]x=1
x=0 = ln 2.
2

150
Chapter 12
Some More Real Analysis
Problems
Problem 130
Let (X, M, ) be a measure space where the measure is positive. Consider a
sequence (An )nN in M such that
(An ) < .
n=1
Prove that
[
\
!
Ak
= 0.
n=1 kn
Hint:
S
Let Bn = kn Ak . Then (Bn ) is a decreasing sequence in M with
(B1 ) =
(An ) < .
n=1
Problem 131
Let (X, M, ) be a measure space where the measure is positive.
Prove that (X, M, ) is -finite if and only if there exists a function f L1 (X)
and f (x) > 0, x X.
151- Math Vietnam
www.MathVn.com

152
CHAPTER 12. SOME MORE REAL ANALYSIS PROBLEMS
Hint:
Consider the function
f (x) =
1 (x)
Xn
.
n (X ) + 1
2
n
n=1
It is clear that f (x) > 0, x X. Just show that f is integrable on X.

Conversely, suppose that there exists f L1 (X) and f (x) > 0, x X. For every n N set
n
1 o
.
Xn = x X : f (x) >
n+1
Show that
Z
Xn = X and (Xn ) (n + 1)
f d.
X
n=1
Problem 132
Let (X, M, ) be a measure space whereR the measure is positive. Let f : X
R+ be a measurable function such that X f d < .
(a) Let N = {x X : f (x) = }. Show that N M and (N ) = 0.
(b) Given any > 0, show that there exists > 0 such that
Z
f d < for any E M with (E) .
E
Hint:
(a) N = f 1 ({}) and {} is closed.
For every n N, n1N f .
(b) Write
f d =
f d.
EN c
For every n N set gn := f 1f >n f 1N c . Show that gn (x) 0 for all x X.
Problem 133
Let > 0 be arbitrary. Construct an open set R which is dense in R and
such that L () < .
Hint:
Write Q = {x1 , x2 , ...}. For each n N let
In := xn
2n+2
, xn +

2n+2

153
Then the In s are open and :=
n=1 In
Q.
Problem 134
Let (X, M, ) be a measure space. Suppose is positive and (X) = 1 (so
(X, M, ) is a probability space). Consider the family
T := {A M : (A) = 0 or (A) = 1}.
Show that T is a -algebra.
Hint:
S
Let (An )nN M. Let A = nN An .
If (A) = 0, then A T .
If (An0 ) = 1 for some n0 N, then
1 = (An0 ) (A) (X) = 1.
Problem 135
For every n N, consider the functions fn and gn defined on R by
n
where , R and > 1
(|x| + n)
gn (x) = n en|x| where R.
fn (x) =
(a) Show that fn Lp (R) and compute kfn kp for 1 p .

(b) Show that gn Lp (R) and compute kgn kp for 1 p .
(c) Use (a) and (b) to show that, for 1 p < q , the topologies induced on
Lp Lq by Lp and Lq are not comparable.
Hint:
(a)
For 1 p < we have
kfn kp = 2 p (p 1) p n+ p .
For p = we have
kfn k = lim kfn kp = n .

p
(b)
For p = we have
kgn k = n .

154
For 1 p < we have

1
kgn kp = 2 p n p p p .
(c) If the topologies induced on Lp Lq by Lp and Lq are comparable, then, for n Lp Lq , we
must have
()
lim kn kp = 0 = lim kn kq = 0.
n
Find an example which shows that the above assumption is not true. For example:
1
n = n+ q gn .
Problem 136
(a) Show that any non-empty open set in Rn has strictly positive Lebesgue measure.
(b) Is the assertion in (a) true for closed sets in Rn ?
Hint:
(a) For any > 0, consider the open ball in Rn
B2 (0) = x = (x1 , .., xn ) : x21 + ... + x2n < 42 .
h
For each n R, let In (0) := n , k . Show that
I (0) := In (0) ... In (0) B2 (0).
|
{z
}
n
(b) No.
Problem 137
(a) Construct an open and unbounded set in R with finite and strictly positive Lebesgue measure.
(b) Construct an open, unbounded and connected set in R2 with finite and strictly positive
Lebesgue measure.
(c) Can we find an open, unbounded and connected set in R with finite and strictly positive
Lebesgue measure?
Hint:
(a) For each k = 0, 1, 2, ... let
1
1
Ik = k k , k + k .
2
2

155
S
Then show that I = k=0 Ik satisfies the question.
(b) For each k = 1, 2, ... let
1 1
Bk = k , k (k, k).
2 2
S
Then show that B = k=0 Bk satisfies the question.
(c) No. Why?
Problem 138
Given a measure space (X, A, ). A sequence (fn ) of real-valued measurable
functions on a set D A is said to be a Cauchy sequence in measure if given
any > 0, there is an N such that for all n, m N we have
{x : |fn (x) fm (x)| } < .
(a) Show that if fn

f on D, then (fn ) is a Cauchy sequence in measure on D.
(b) Show that if (fn ) is a Cauchy sequence in measure, then there is a function
f to which the sequence (fn ) converges in measure.
Hint:
(a) For any > 0, there exists N > 0 such that for n, m N we have
+ D : |fn f |
.
{D : |fm fn | } D : |fm f |
2
2
(b) By definition,
for =
n
1
1o 1
, n1 N : D : |fn1 +p fn1 |
<
for all p N.
2
2
2
In general,
for =
n
1
1o
1
,
n
N,
n
>
n
:
D
:
|f
f
|
< k for all p N.

k
k
k1
n
+p
n
k
k
k
k
2
2
2
Since nk+1 = nk + p for some p N, so we have

n
1
1o
D : |fnk+1 fnk | k < k for k N.
2
2
Let gk = fnk . Show that (gk ) converges a.e. on D. Let Dc := {x D : limk gk (x) R}.
Define f by f (x) = limk gk (x) for x Dc and f (x) = 0 for x D \ Dc . Then show that gk
f
on D. Finally show that fn

f on D.

156
Problem 139
Check whether the following functions are Lebesgue integrable :
(a) u(x) = x1 , x [1, ).
(b) v(x) = 1x , x (0, 1].
Hint:
(a) u(x) is NOT Lebesgue integrable on [1, ).
Z n
Z
Z
1
1
u(x)dL (x) = lim
1[1,n) (x)dL (x) = lim
dx.
n
n
x
x
1
[1,)
(b) v(x) is Lebesgue integrable on (0, 1].
We can write
1
1
1
v(x) = , x (0, 1] = 1(0,1] (x) = sup 1[ n1 ,1] (x).
x
x
x
n
1
Use the Monotone Convergence Theorem for the sequence x 1[ n1 ,1] nN .
Bibliography
[1] Wheeden, L.R.; Zygmund, A. Measure and Integral. Marcel Dekker. New York, 1977.
[2] Folland. G.B. Real Analysis. John Wiley and Sons. New York, 1985.
[3] Rudin. W. Real and Complex Analysis. Third edition. McGraw-Hill, Inc. New York, 1987.
[4] Royden. H.L. Real Analysis. Third edition. Prentice Hall. NJ, 1988.
[5] Rudin. W Functional Analysis. McGraw-Hill, Inc. New York, 1991.
[6] Hunter. J.K.; Nachtergaele. B. Applied Analysis. World Scientific. NJ, 2001.
[7] Yeh, J. Real Analysis. Theory of measure and integration. Second edition. World Scientific.
NJ, 2006.
157- Math Vietnam

www.MathVn.com

Measure and Integration Full

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Measure and Integration Full

Загружено:

Авторское право:

Доступные форматы

MEASURE and INTEGRATION

Problems with Solutions

www.MATHVN.com - Anh Quang Le, PhD

www.MathVn.com - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

www.MathVn.com - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

1 Measure on a -Algebra of Sets

4 Convergence a.e. and Convergence in Measure

5 Integration of Bounded Functions on Sets of Finite Measure

6 Integration of Nonnegative Functions

7 Integration of Measurable Functions

8 Signed Measures and Radon-Nikodym Theorem

9 Differentiation and Integration

11 Integration on Product Measure Space

12 Some More Real Analysis Problems

www.MathVn.com3 - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

www.MathVn.com - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

For a decreasing sequence (An ), we define

Definition 2 For any sequence (An ) of subsets of a set X, we define

Proposition 1 Let (An ) be a sequence of subsets of a set X. Then

lim inf An = {x X : x An for all but finitely many n N}.

lim sup An = {x X : x An for infinitely many n N}.

lim inf An lim sup An .

www.MathVn.com5 - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

CHAPTER 1. MEASURE ON A -ALGEBRA OF SETS

Definition 6 (Finite, -finite measure)

www.MathVn.com - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

(En )nN P(X),

Definition 8 (Caratheodory condition)

www.MathVn.com - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

CHAPTER 1. MEASURE ON A -ALGEBRA OF SETS

www.MathVn.com - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

www.MathVn.com - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

CHAPTER 1. MEASURE ON A -ALGEBRA OF SETS

www.MathVn.com - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

On the other hand, by the countable subadditivity of , we have

From (i) and (ii), it follows that

This proves the countable additivity of .

www.MathVn.com - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

CHAPTER 1. MEASURE ON A -ALGEBRA OF SETS

Hence, is not additive.

On the other hand, we have

If all Cn are finite then

www.MathVn.com - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

From (1.1) and (1.2) we obtain

www.MathVn.com - Math Vietnam

www.MATHVN.com - Anh Quang Le, PhD

CHAPTER 1. MEASURE ON A -ALGEBRA OF SETS

Problem 10 (Monotone sequence of measurable sets)

lim (En ) = lim En .

(b) If (En ) is decreasing, show that

lim (En ) = lim En ,

provided that there is a set A A satisfying (A) < and A E1 .