MATH 3033 Real Analysis

Lebesgue Integration
Dr. Albert Ku

1 Motivation
In the last chapter, we learned that if we have a sequence of Riemann integrable functions {fn } that converges
uniformly to f on [a, b], then
Z b Z b
lim fn = f.
n→∞ a a
If the convergence is not uniform, the limit function may not even be Riemann integrable. The following is
one such example:
Example 1.1. It is well-known that Q ∩ [0, 1] is countable i.e. we can enumerate all ratonal numbers in
[0, 1] as a sequence {rn }. Then for n ∈ N, we define
(
1 if x ∈ {r1 , . . . , rn }
fn (x) =
0 if x ∈
/ {r1 , . . . , rn }
Z 1
It is easy to see that for any n ∈ N, fn is Riemann integrable and fn = 0. However, {fn } converges
0
pointwise to f , where (
1 if x ∈ Q ∩ [0, 1]
f (x) =
0 if x ∈ [0, 1] \ Q
This is called Dirichlet function and we already showed that f is not Riemann integrable.

Example 1.2. Consider the sequence of functions {fn } on [0, 1] such that for n ∈ N, fn (x) = sin(xn ) on
[0, 1]. Suppose we want to evaluate the following limit:
Z 1
lim fn (x) dx
n→∞ 0

Notice that {fn } converges pointwise but not uniformly to f on [0, 1], where
(
0, if x ∈ [0, 1)
f (x) =
sin(1), if x = 1

It would
Z 1 be easier to Zcompute the limit if we can take the limit into the integral. In other word, we have
1
lim fn (x) dx = f (x) dx = 0. Unfortunately, we cannot do this because the convergence is not
n→∞ 0 0
uniform.

1
Question: Can we define a more general kind of integrals such that
1. the integral of functions like Dirichlet function is well-defined, and
2. a sequence of the integrals of functions converges to the integral of the limit function under some
convergence conditions that are less restrictive than uniform convergence?

The answer is yes. First of all, for any set S ⊆ R, we define the characteristic function χS by
(
χS (x) = 1 if x ∈ S
0 if x ∈ /S

Recall the definition of Riemann integral of a function g : [a, b] → R. For any partition P = {x0 , . . . , xn } of
[a, b], we can define the following “step function”:
n
mi χ[xi−1 ,xi ) ,
X
sn =
i=1

where mi = inf g([xi−1 , xi ]). And the integral of g on [a, b] can be approximated by the lower Riemann sum
n
X
L(g, P ) = mi l([xi−1 , xi )),
i=1

where l([u, v)) = v − u is the length of the interval [u, v). The French mathematician Henri Lebesgue
came up with the brilliant idea that instead of partitioning the interval [a, b], we can partition the range of
g: Suppose g is bounded i.e. there exists M > 0 such that |g(x)| ≤ M for all x ∈ [a, b]. For n ∈ N, let
yj = −M + nj M for j = 0, 1, . . . , 2n. Then we define the following function
2n
yj χEj
X
hn =
j=1

where Ej = g −1 ([yj , yj+1 )) for j = 0, 1, 2, . . . , 2n − 2 and E2n−1 = g −1 ([y2n−1 , y2n ]). Therefore, naturally
the integral of g on [a, b] can be approximated by
2n−1
X
yj µ(Ej ),
j=0

where µ(Ej ) is the “length” of Ej . In order for the above sum to make sense, we need to find a way to
extend the concept of “length” from intervals to more general sets in R. Lebesgue successsfully constructed
the Lebesgue measure, which assign lengths to a very large class of sets in R. As a consequence, we can
define integrals for a much wider class of functions using the above method. Also, such integrals behave
much better than Riemann integrals when taking limit of a sequence of functions, which is important in the
study of Fourier series and other topics in analysis.

2
2 Lebesgue Measure
Let P(R) be the set of all subsets of R. We want to define a measure µ : P(R) → [0, ∞] which has the
following desired properties:
1. (Interval length) µ(I) = l(I) for every interval I ⊆ R.
2. (Countably additive) For every sequence E1 , E2 , . . . of pairwise disjoint subsets of R,
∞ ∞
!
[ X
µ En = µ(En )
n=1 n=1

3. (Translation invariant) For any c ∈ R and any E ⊆ R, µ(E + c) = µ(E), where E + c =

{x + c ∈ R | x ∈ E}.

Remarks 2.1.
1. The range of µ includes ∞ i.e. it is possible for the measure of a subset of R to be infinity. For example,
µ((0, ∞)) = ∞.
2. It can be shown that for any A ⊆ B ⊆ R, µ(A) ≤ µ(B). This property is called monotonicity.
3. Unfortunately, we will see that it is impossible to construct such a measure that satisfies all the above
properties. The following example of a set that cannot be measured by µ.

Example 2.1. Define the following equivalence relation ∼ in [0, 1) by

x ∼ y if x − y ∈ Q

By the Axiom of choice, we can choose one real number from each equivalence class defined by ∼ to form
a set E. Now, for any x, y ∈ [0, 1), we define the sum modulo 1 of x and y by
(
x+y if x + y < 1
(x + y) mod 1 =
x+y−1 if x + y ≥ 1

As we know, [0, 1) ∩ Q is countable. So we can enumerate all the rational numbers in [0, 1] as a sequence
{rn }. Now we define
En = {(x + rn ) mod 1 ∈ [0, 1) | x ∈ E}
Then it can be shown that E1 , E2 , . . . are pairwise disjoint subsets of [0, 1) (Why?) and
∞
[
[0, 1) = En
n=1

By the countable additivity and translation invariance of µ, µ(En ) = µ(E) for any n ∈ N and
∞
X ∞
X
µ([0, 1)) = µ(En ) = µ(E)
n=1 n=1

If µ(E) > 0, then µ([0, 1)) = ∞, but µ([0, 1)) = 1 i.e. a contradiction.
If µ(E) = 0, then µ([0, 1)) = 0, but µ([0, 1)) = 1, which is again a contradiction.

3
To construct the Lebesgue measure µ for a large class of sets in R, we begin by defining the (Lebesgue)
outer measure for all subsets in R as follows:
Definition 2.1. Suppose S ⊆ R, the (Lebesgue) outer measure of S, denoted by µ∗ (S), is defined by
( )
X
∗
µ (S) = inf l(I) C is a collection of countable open intervals that covers S
I∈C

Obviously, the outer measure cannot satisfy all the properties of a measure we have mentioned before. In
fact, we will prove later that it has the weaker property of countable subadditivity:
∞ ∞
!
[ X
∗
µ Sn ≤ µ∗ (Sn )
n=1 n=1

for any sequence {Sn } (not necessarily pairwise disjoint) of subsets of R. But it does satisfy all other
properties:
Theorem 2.1. µ∗ is translation invariant and µ∗ (I) = l(I) for any interval I ⊆ R.
Proof. It is obvious that µ∗ is translation invariant because for any S ⊆ R and any c ∈ R,
( )
X
∗
µ (S + c) = inf l(I) C is a collection of countable open intervals that covers S + c
I∈C
( )
X
= inf l(I + c) B is a collection of countable open intervals that covers S
I∈B
( )
X
= inf l(I) B is a collection of countable open intervals that covers S = µ∗ (S)
I∈B

To prove that µ∗ (I) = l(I) for any interval I ⊆ R, we begin with the case when I = [a, b]. Since for any
ε > 0, [a, b] ⊆ (a − ε, b + ε) i.e. an open interval covering of I. Hence µ∗ (I) ≤ l((a − ε, b + ε)) = b − a + 2ε,
which implies µ∗ (I) ≤ b − a. Then we have to show that µ∗ (I) ≥ b − a, or equivalently, if C is any collection
of countable open intervals that covers I, then
X
l(I) ≥ b − a.
I∈C

By Heine-Borel theorem, I is compact i.e. any such [ collection has a finite subcollection that also covers
I. Consider such finite subcollection. Since a ∈ In , there must be one of the In ’s that contains a. Let
this interval be (a1 , b1 ) and hence a1 < a < b1 . If b1 ≤ b, then b1 ∈ [a, b] and is contained in another
interval (a2 , b2 ) in the finite subcollection i.e. a2 < b1 < b2 . Repeat this process, we obtain a sequence
(a1 , b1 ), (a2 , b2 ), . . . , (ak , bk ) such that ai < bi−1 < bi . Since the subcollection is finite, the process must
terminate with some interval (ak , bk ) such that b ∈ (ak , bk ). Hence, we have
X X
l(In ) ≥ l((ai , bi ))
= (bk − ak ) + (bk−1 − ak−1 ) + · · · + (b1 − a1 )
= bk − (ak − bk−1 ) − (ak−1 − bk−2 ) − · · · − (a2 − b1 ) − a1
> bk − a1 > b − a
If I is any finite interval, then given any ε > 0, there is a closed interval J ⊆ I such that l(J) > l(I) − ε.
Hence
l(I) − ε < l(J) = µ∗ (J) ≤ µ∗ (I) ≤ µ∗ (I) = l(I) = l(I)
Thus for each ε > 0, l(I) − ε < µ∗ (I) ≤ l(I), which implies µ∗ (I) = l(I).

If I is any infinite interval, then for any give positive real number r, there is closed interval J ⊆ I with
l(J) = r. Hence µ∗ (I) ≥ µ∗ (J) = l(J) = r. This implies µ∗ (I) = ∞.

4
Theorem 2.2. (Countable subadditivity) Let {Sn } be a countable collection of sets in R. Then
!
[ X
µ∗ Sn ≤ µ∗ (Sn )
n n

∗
Proof. If one of the sets Sn has infinite outer measure, the inequality holds trivially. If µ[ (Sn ) is finite, then
given any ε > 0, there is a countable collection {In,i }i of open intervals such that Sn ⊆ In,i and
i
X ε
I(In,i ) < µ∗ (Sn ) +
i
2n
[
Now the collection {In,i }n,i is also countable and covers Sn . Hence
n
!
∗
[ X XX X ε 
µ Sn ≤ l(In,i ) = l(In,i ) < µ∗ (Sn ) +
n n,i n i n
2n
X
∗
= µ (Sn ) + ε
n

Since ε is an arbitrary positive number,

!
[ X
µ∗ Sn ≤ µ∗ (Sn ).
n n

Now we are ready to define a class of sets in R such that µ∗ is countable additive. The following definition
is due to Carathédory and is sometimes called the Carathédory’s criterion:
Definition 2.2. A set E ⊆ R is said to be measurable if for each set A ⊆ R,

µ∗ (A) = µ∗ (A ∩ E) + µ∗ (A ∩ E c )

where E c = R \ E is the complement of E. The set of all measurable sets is denoted by M.

Remarks 2.2.
1. Since we always have µ∗ (A) ≤ µ∗ (A ∩ E) + µ∗ (A ∩ E c ), E is measurable if and only if for each A ⊆ R,
we have
µ∗ (A) ≥ µ∗ (A ∩ E) + µ∗ (A ∩ E c )

2. As the definition of measurability is symmetric in E and E c , E c is measurable whenever E is.

3. Obviously, ∅ and R are measurable.

5
Theorem 2.3. If µ∗ (E) = 0, then E is measurable and is said to have measure zero.
Proof. Let A ⊆ R be any set. Then A ∩ E ⊆ E and so µ∗ (A ∩ E) ≤ µ∗ (E) = 0. Also A ∩ E c ⊆ A, and so

µ∗ (A) ≥ µ∗ (A ∩ E c ) = µ∗ (A ∩ E) + µ(A ∩ E c )

Therefore, E is measurable.
Remarks 2.3.
1. Obviously, any singleton in R has measure zero. Then by countable subadditivity of µ∗ , any countable
set in R also has measure zero.

2. By countable subadditivity of µ∗ again, any union of countable collection of sets of measure zero has
measure zero.

Definition 2.3. Let C ⊆ P(R). We say that C is a σ-algebra if it satisfies the following conditions:
ˆ ∅∈C

ˆ If A ∈ C, then Ac ∈ C.
[
ˆ If {An } is a countable collection of sets in C, then An ∈ C.
n

Remarks 2.4. The above definition implies the following:

ˆ R ∈ C.
\
ˆ If {An } is a countable collection of sets in C, then An ∈ C.
n

The following are the two main theorems about measurable sets and µ∗ . The proof of them are in the next
section.

Theorem 2.4. M is a σ-algebra.

Theorem 2.5. Let µ be the outer measure µ∗ restricted on the σ-algebra M. Then µ : M → [0, ∞] is a
measure satisfying the three desired properties:
1. µ(I) = l(I) for any interval I ⊆ R.

2. µ is countably additive.
3. µ is translation invariant.
The measure µ on R is called the Lebesgue measure.

6
Next, we will show that all open sets and closed sets in R are measurable.
Lemma 2.1. For any a ∈ R, (a, ∞) is measurable.
Proof. Let A ⊆ R be any set, A1 = A ∩ (a, ∞), and A2 = A ∩ (−∞, a]. It suffices to show that

µ∗ (A1 ) + µ∗ (A2 ) ≤ µ∗ (A)

If µ∗ (A) = ∞, the above inequality holds trivially. If µ∗ (A) < ∞, then given any ε > 0, there exists a
countable collection {In } of open intervals which covers A and
X
l(In ) ≤ µ∗ (A) + ε
n

Let In,1 = In ∩ (a, ∞) and In,2 = In ∩ (−∞, a]. Then In,1 and In,2 are intervals and

l(In ) = l(In,1 ) + l(In,2 ) = µ∗ (In,1 ) + µ∗ (In,2 )

[
Since Ai ⊆ In,i for i = 1, 2, we have
n
!
[ X
∗ ∗
µ (Ai ) ≤ µ In,i ≤ µ∗ (In,i ) (i = 1, 2)
n n

Thus
X
µ∗ (A1 ) + µ∗ (A2 ) ≤ (µ∗ (In,1 ) + µ∗ (In,2 ))
n
X
≤ l(In ) ≤ µ∗ (A) + ε
n

Since ε is an arbitrary positive number, we have

µ∗ (A1 ) + µ∗ (A2 ) ≤ µ∗ (A)

Theorem 2.6. All open sets and closed sets in R are measurable.
Proof. First, we prove that all open intervals are measurable.

By Lemma 2.1, (a, ∞) is measurable for any a ∈ R. Then (−∞, a] = (a, ∞)c is also measurable. Moreover,
for any b ∈ R,
∞  
[ 1
(−∞, b) = −∞, b − ,
n=1
n

then (−∞, b) is measurable. For a < b, (a, b) = (−∞, b) ∩ (a, ∞), which implies that (a, b) is measurable.

For any open set in R, we already proved that it is a countable union of open intervals. Therefore, any open
set in R is measurable. As any closed set is, by definition, a complement of an open set. Therefore, any
closed set in R is also measurable.

Remarks 2.5. Let B be the collection of Borel sets, the smallest σ-algebra containing all open sets of R.
Then by the above theorem, B ⊆ M.

7
3 Proof of the Theorem 2.4 and Theorem 2.5 (Optional)
Lemma 3.1. If E1 and E2 are measurable, so are E1 ∪ E2 and E1 ∩ E2 .
Proof. Let A ⊆ R be any set. Since E2 is measurable, we have

µ∗ (A ∩ E1c ) = µ∗ (A ∩ E1c ∩ E2 ) + µ∗ (A ∩ E1c ∩ E2c ) = µ∗ (A ∩ E1c ∩ E2 ) + µ∗ (A ∩ (E1 ∪ E2 )c )

and since A ∩ (E1 ∪ E2 ) = (A ∩ E1 ) ∪ (A ∩ E2 ∩ E1c ), we have

µ∗ (A ∩ (E1 ∪ E2 )) ≤ µ∗ (A ∩ E1 ) + µ∗ (A ∩ E2 ∩ E1c ).

Therefore, we get

µ∗ (A ∩ (E1 ∪ E2 )) + µ∗ (A ∩ (E1 ∪ E2 )c )
≤ µ∗ (A ∩ E1 ) + µ∗ (A ∩ E2 ∩ E1c ) + µ∗ (A ∩ (E1 ∪ E2 )c )
= µ∗ (A ∩ E1 ) + µ∗ (A ∩ E1c ) = µ∗ (A)

by the measurability of E1 . Hence E1 ∪ E2 is measurable.

Since E1 ∩ E2 = (E1c ∪ E2c )c , E1 ∩ E2 is also measurable.

Remarks 3.1. The above lemma implies that any finite union and intersection of measurable sets are
measurable.

Lemma 3.2. (Finite additivity) Let A ⊆ R be any set, and E1 , . . . , En be a finite sequence of disjoint
measurable sets. Then " n #!
[ Xn
∗
µ A∩ Ei = µ∗ (A ∩ Ei )
i=1 i=1

Proof. We will prove the lemma by induction. It is clearly true for n = 1. Assume it is true for n = k.
Consider E1 , . . . , Ek+1 disjoint measurable sets. We have
"k+1 #
[
A∩ Ei ∩ Ek+1 = A ∩ Ek+1
i=1
"k+1 # " k
#
[ [
c
A∩ Ei ∩ Ek+1 =A∩ Ei
i=1 i=1

Hence by the measurability of Ek+1 , we have

"k+1 #! " k
#!
[ [
∗
µ A∩ Ei = µ∗ (A ∩ Ek+1 ) + µ∗ A∩ Ei
i=1 i=1
k
X
= µ∗ (A ∩ Ek+1 ) + µ∗ (A ∩ Ei ) (by induction hypothesis)
i=1

Hence, the lemma holds for n = k + 1.

8
(Proof of Theorem 2.4). We already know that empty set is measurable and the complement of a mea-
surable set is measurable. To prove that M is a σ-algebra, it remains to prove that the union of a countable
∞
[
collection of measurable sets is measurable. Suppose E = Dk . Then we define the sets E1 , E2 , . . . by
k=1

E1 = D1
E2 = D2 \ E1
E3 = D3 \ (E1 ∪ E2 )
.. ..
. .
Ek = Dk \ (E1 ∪ · · · ∪ Ek−1 )
.. ..
. .
∞
[
It is not difficult to see that {Ek } are pairwise disjoint and E = Ek . Moreover, by the previous lemmas,
k=1
Ek is measurable for any k ∈ N.

n
[
Let A ⊆ R be any set and Fn = Ek . Then Fn is measurable and E c ⊆ Fnc . Hence, we have
k=1

µ∗ (A) = µ∗ (A ∩ Fn ) + µ∗ (A ∩ Fnc )
≥ µ∗ (A ∩ Fn ) + µ∗ (A ∩ E c )
Xn
= µ∗ (A ∩ Ek ) + µ∗ (A ∩ E c ) (by Lemma 3.2)
k=1

Since the left side of the above inequality is independent of n, we have

∞
X
µ∗ (A) ≥ µ∗ (A ∩ Ek ) + µ∗ (A ∩ E c )
k=1
≥ µ∗ (A ∩ E) + µ∗ (A ∩ E c ) (Countable subadditivity)

Therefore, E is measurable.

9
(Proof of Theorem 2.5). Since µ is a restriction of the outer measure µ∗ and thanks to Theorem 2.1,
µ(I) = µ∗ (I) = l(I) for any interval I ⊆ R and µ is translation invariant. Therefore, it remains to prove
that µ is countably additive.

Let {Ek } be a countable collection of disjoint measurable sets. By Lemma 3.2, we already have the finite
additivity (set A = R) i.e. for any n ∈ N,
n
! n
[ X
∗
µ Ek = µ∗ (Ek )
k=1 k=1

n
[ ∞
[
Since Ek ⊆ Ek , we have
k=1 k=1

∞
! n
! n
[ [ X
µ∗ Ek ≥ µ∗ Ek = µ∗ (Ek )
k=1 k=1 k=1

Since the left side of the above inequality is independent of n, we have

∞ ∞
!
[ X
∗
µ Ek ≥ µ∗ (Ek )
k=1 k=1

Combining with the countable subadditivity of µ∗ , we get

∞ ∞
!
[ X
∗
µ Ek = µ∗ (Ek )
k=1 k=1

Because all the sets above are measurable, we have

∞ ∞
!
[ X
µ Ek = µ(Ek )
k=1 k=1

Hence µ is countably additive.

10
4 Measurable Functions
Definition 4.1. A function f : R → R is said to be measurable if f −1 (U ) is measurable for any open set
U ⊆ R.
Example 4.1.
1. All continuous functions from R to R are measurable.

2. Suppose f, g : R → R such that f is measurable and g is continuous. Then g ◦ f is measurable.

3. The set E in Example 2.1 is a non-measurable sets. Therefore, it can be shown that χE is a non-
measurable function.

Theorem 4.1. Let f : R → R be a function. The following statements are equivalent:

1. f is measurable.
2. f −1 ((a, ∞)) is measurable for all a ∈ R.

3. f −1 ([a, ∞)) is measurable for all a ∈ R.

4. f −1 ((−∞, a)) is measurable for all a ∈ R.
5. f −1 ((−∞, a]) is measurable for all a ∈ R.

Proof. First of all, it is obvious that (1) ⇒ (2), (2) ⇔ (5), and (3) ⇔ (4).

∞  
\ 1
(2) ⇒ (3): f −1 ([a, ∞)) = f −1 a− ,∞ .
n=1
n
∞  
[ 1
(3) ⇒ (2): f −1 ((a, ∞)) = f −1 a+ ,∞ .
n=1
n

Combining above, we have (2) ⇔ (3) ⇔ (4) ⇔ (5) and (1) ⇒ (2). To prove (2) ⇒ (1), we need the following:
(2)+(4)⇒(1): Since any open set in R is a countable union of open intervals, it suffices to prove that the
inverse image of f of any open interval is measurable. By (2) and (4), it holds for infinte open intervals. As
for finite intervals, we have
f −1 ((a, b)) = f −1 ((a, ∞)) ∩ f −1 ((−∞, b))

11
Theorem 4.2. Suppose c ∈ R and f, g : R → R are measurable functions. f + c, cf , f ± g, f g, max {f, g},
min {f, g}, and |f | are measurable.
Proof. For any a ∈ R, we have
ˆ (f + c)−1 ((−∞, a)) = f −1 ((−∞, a − c))

ˆ Ifc > 0, (cf )−1 ((−∞, a)) = f −1 −∞,

a

c
Ifc < 0, (cf )−1 ((−∞, a)) = f −1 ac , ∞
Ifc = 0 and a ≥ 0, (cf )−1 ((−∞, a)) = R.
Ifc = 0 and a < 0, (cf )−1 ((−∞, a)) = ∅.
[
ˆ (f + g)−1 ((−∞, a)) = f −1 ((−∞, r)) ∩ g −1 ((−∞, a − r))

r∈Q

ˆ f − g = f + (−1)g.

ˆ If a ≤ 0, (f 2 )−1 ((−∞, a)) = ∅. √ √

If a > 0, (f 2 )−1 ((−∞, a)) = f −1 ((−∞, a)) ∩ f −1 ((− a, ∞)).
1h 2
i
Then f g = (f + g) − f 2 − g 2 .
2
For max {f, g}, min {f, g}, and |f |, the proofs are left as exercise.

Exercise 4.1. Let A ⊆ R and {fn } be a sequence of functions such that fn : A → R is measurable. Define
for any x ∈ A,

(sup fn )(x) = sup {fn (x) | n ∈ N}

(inf fn )(x) = inf {fn (x) | n ∈ N}
(lim sup fn )(x) = lim sup fn (x)
n→∞
(lim inf fn )(x) = lim inf fn (x)
n→∞
(lim fn )(x) = lim f (x)
n→∞

If sup fn , inf fn , lim sup fn , lim inf fn and lim fn exist (in R) everywhere on A, prove that they are measurable.

12
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. For example, we say that for two functions f, g : R → R, f = g a.e. if

µ ({x ∈ R | f (x) ̸= g(x)}) = 0.

Similarly, we say that a sequence of functions {fn } converges to f a.e. if there is a set E ⊆ R such that
µ(E) = 0 and {fn (x)} converges to f (x) for all x ∈
/ E.

Theorem 4.3. If f is a measurable function and f = g a.e., then g is also measurable.

Proof. Let E = {x ∈ R | f (x) ̸= g(x)}. By assumption, E has measure zero. We already proved that all sets
that have measure zero are measurable. Hence E is measurable. Then for any a ∈ R,

g −1 ((−∞, a)) = f −1 ((−∞, a)) ∩ E c ∪ g −1 ((−∞, a)) ∩ E .

First of all, since f is measurable, f −1 ((−∞, a)) ∩ E c is obviously measurable. Moreover, as E has measure
zero, its subseteq g −1 ((−∞, a))∩E also has measure zero and is measurable. Therefore, g is measurable.

Theorem 4.4. Let A ⊆ R. Then χA is a measurable function if and only if A is measurable.

Proof. Left as exercise.

Definition 4.2. A function f : R → R is called a simple function if there exists finitely many measurable
sets A1 , A2 , . . . , An ⊆ R and a1 , a2 , . . . , an ∈ R such that
n
ai χAi
X
f=
i=1

Remarks 4.1.
1. Any simple function is a measurable function.
2. If f, g are simple functions, then f + g, f − g, f g, max {f, g}, min {f, g}, and |f | are also simple.

13
Theorem 4.5. For any non-negative measureable function f : R → [0, ∞), there exists a monotonically
increasing sequence {sn } of simple functions i.e. 0 ≤ s1 ≤ s2 ≤ · · · such that {sn } converges pointwise to f
(Notation: sn ↗ f on R).
Proof. For n, k ∈ N, define the following sets:
   
k k+1 −1 k k+1
An,k = x ∈ R | n ≤ f (x) < =f ,
2 2n 2n 2n

Fn = {x ∈ R | f (x) ≥ n} = f −1 ([n, ∞))

Obviously, An,k and Fn are measurable sets because f is measurable. Then we define the simple function
sn by
n
n2 −1
k
χ + n χF n
X
sn =
2n An,k
k=0

It is easy to see that sn ≤ f and {sn } is a monotonically increasing sequence. Moreover, Fix any x ∈ R, for
1
any ε > 0, we choose N ∈ N such that f (x) < N and N < ε. Then for n > N , we have
2
1
|f (x) − sn (x)| ≤ <ε
2n
Hence {sn (x)} converges to f (x).
Remarks 4.2.
1. If f is a bounded measurable function, then the choice of N in the above proof is independent of x.
That is to say, {sn } converges uniformly to f on R.

2. For a measurable function f in general i.e. not necessarily non-negative, we can write

f = f+ − f−

where f + = max {f, 0} and f − = max {−f, 0} are non-negative measurable functions. Then by the
above theorem, there exists sequences of simple functions {un } and {vn } such that un ↗ f + and
vn ↗ f − on R. Let sn = un − vn . Then {sn } is a sequence of simple functions that converges to f .

14
Remarks 4.3. (A note on extended real numbers)
1. It is often more convenient to extend the system of real numbers by the addition of two elements ∞
and −∞. The enlarged set is called the set of extended real numbers, denoted by R∗ (or [−∞, ∞]).
For −∞ < x < ∞, we define

x+∞ = ∞
x−∞ = −∞
(
∞ if x > 0
x·∞=∞·x =
−∞ if x < 0
(
−∞ if x > 0
x · −∞ = −∞ · x =
∞ if x < 0

Moreover, we have
∞+∞=∞ −∞ − ∞ = −∞
∞ · (±∞) = ±∞ −∞ · (±∞) = ∓∞
The operation ∞ − ∞ is left undefined. By convention, 0 · ∞ = ∞ · 0 = 0.

2. We can also extend the topology of R to R∗ by adding intervals of the forms (a, ∞] and [−∞, a) as
open sets.
3. The definition of measurable functions can be readily extended to the extended real-valued functions
i.e. f : R → R∗ is said to be measurable if f −1 (U ) is measurable for any open set U ⊆ R∗ .
4. The theorems about measurable functions still hold for extended real-valued functions (with slight
modification if necessary).
5. It is not difficult to see that if {fn } is a sequence of extended real-valued measurable functions, then
sup fn , inf fn , lim sup fn , and lim inf fn are also extended real-valued measurable functions. In partic-
n n n→∞ n→∞
ular, if {fn } converges to f a.e., f is also measurable.

15
5 Lebesgue Integrals
First we define the Lebesgue integrals for non-negative simple functions:
n
ai χAi be a non-negative simple function i.e. ai ≥ 0 for all i. Then the
X
Definition 5.1. Let s =
i=1
Lebesgue integral of s is
Z n
X
s dµ = ai µ(Ai )
i=1

Morevoer, if E ⊆ R is measurable, the Lebesgue integral of s over E is

Z Z n
sχE dµ =
X
s dµ = ai µ(Ai ∩ E)
E i=1

(Convention: If ai µ(Ai ∩ E) is of the form 0 · ∞, we take this to be 0.)

Remarks 5.1.
Z
1. s dµ may have the value ∞ if ai > 0 and µ(Ai ) = ∞ for some i.
Z
2. It is easy to see that the value of s dµ is independent of the representation of s which we use.

3. If s and t are simple functions and a, b are positive real numbers, then
Z Z Z
(as + bt) dµ = a s dµ + b t dµ

Now we are ready to define the Lebesgue integrals for non-negative measurable functions:
Definition 5.2. Let f : R → [0, ∞) be a non-negative measurable function and let E ⊆ R be a measurable
set. Then the Lebesgue integral of f on E is
Z Z 
f dµ = sup s dµ 0 ≤ s ≤ f and s is a simple function
E E
Z
Moreover, if f dµ is finite, f is said to be Lebesgue integrable over E.
E

Remarks 5.2.

1. It is obviousZfrom the definition

Z that for any two non-negative measurable functions f, g such that
f ≤ g on E, f dµ ≤ g dµ.
E E

2. By Theorem 4.5, there exists an increasing sequence of simple functions {sn } such that sn ↗ f . Later,
we will prove that Z Z
f dµ = lim sn dµ.
n→∞

16
Exercise 5.1. Suppose f : R → [0, ∞) is a non-negative measurable function and E ∈ M.
Z Z
1. Prove that for any c ∈ [0, ∞), cf dµ = c f dµ.
E E
Z
2. Show that if f = 0 a.e., f dµ = 0.
E
Z
3. Show that if µ(E) = 0, f dµ = 0.
E

In order to derive various properties of Lebesgue integrals, we need the following important theorem:
Theorem 5.1. (Fatou’s Lemma) Suppose {fn } is a sequence of non-negative measurable functions, then
Z Z
lim inf fn dµ ≤ lim inf fn dµ
n→∞ n→∞

Proof. (Optional) First of all, let f = lim inf fn . Then f : R → [0, ∞] is measurable. It suffices to prove
n→∞
that for any simple function s ≤ f , Z Z
s dµ ≤ lim inf fn dµ
n→∞
Z Z Z
because f dµ = sup s dµ. If s dµ = ∞, then there is a measurable set A such that µ(A) = ∞ and
s≤f
s > a > 0 on A. For n ∈ N, define
\
An = {x ∈ R | fn (x) > a for all k ≥ n} = fk−1 ((a, ∞])
k≥n

Notice that An ∈ M and A1 ⊆ A2 ⊆ · · · . Since x ∈ A ⇒ s(x) > a ⇒ f (x) > a, then lim inf fn (x) > a, which
n→∞
implies that fk (x) > a for large enough k i.e. x ∈ Ak for large enough k. Therefore, we have
∞
[
A⊆ An
n=1

∞
!
[
Then ∞ = µ(A) ≤ µ An = lim µ(An ) (Why?).
n→∞
n=1

Therefore, we have Z Z Z
fn dµ ≥ fn dµ ≥ a dµ = aµ(An ) → ∞ as n → ∞
An An
Z Z
Hence lim inf fn dµ = ∞ = s dµ.
n→∞
Z
If s dµ < ∞, then there exists A ∈ M such that µ(A) < ∞ and s = 0 on Ac , s > 0 on A.
n
ai χAi . Then A = union of the Ai such that ai > 0.)
X
(Suppose s =
i=1

17
Let M = max s(x) and let ε > 0 be any positive small real number. For any n ∈ N, define
x∈R
\
An = {x ∈ R | fn (x) > (1 − ε)s(x) for all k ≥ n} = gk−1 ((0, ∞])
k≥n

where gk = fk − (1 − ε)s. It is clear that An ∈ M.

∞
[
Claim: A ⊆ An .
n=1

x∈A ⇒ s(x) > 0 ⇒ lim inf fk (x) ≥ s(x) > (1 − ε)s(x)

n→∞
⇒ fk (x) > (1 − ε)s(x) for large enough k
⇒ x ∈ Ak for large enough k
∞
[
⇒ x∈ An
n=1

∞
\
Since A \ An ∈ M for n ∈ N, A \ A1 ⊃ A \ A2 ⊃ · · · , and (A \ An ) = ∅, we have
n=1

∞
!
\
lim µ(A \ An ) = µ (A \ An ) = µ(∅) = 0. (Why?)
n→∞
n=1

Therefore, there exists N ∈ N such that µ(A \ Ak ) < ε for all k ≥ N . Then, for k ≥ N , we have
Z Z Z Z Z
fk dµ ≥ fk dµ ≥ (1 − ε) s dµ = (1 − ε) s dµ − (1 − ε) s dµ
Ak Ak A\Ak
Z Z Z
≥ s dµ − ε s dµ − s dµ
A\Ak
Z Z 
≥ s dµ − ε s dµ + M

Since ε is an arbitrary positive real number, we get

Z Z
lim inf fk dµ ≥ s dµ.
k→∞

18
An immediate consequence of Fatou’s lemma is the following very useful result:
Theorem 5.2. (Monotone Convergence Theorem MCT) Let {fn } be an increasing sequence of non-
negative measurable functions which converges a.e. to f . Then
Z Z
lim fn dµ = f dµ.
n→∞

Proof. Without loss of generality, we may assume that fn → f everywhere as the set of points that {fn }
does not converge to f has measure zero, which will not affect the Lebesgue integrals.

Since {fn } is an increasing sequence, fn ≤ f for all n. Hence, for any n ∈ N,

Z Z
fn dµ ≤ f dµ
Z Z
which implies lim sup f dµ ≤ f dµ. Then by Fatou’s lemma,
n→∞
Z Z Z Z
f dµ ≤ lim inf fn dµ ≤ lim sup f dµ ≤ f dµ.
n→∞ n→∞
Z Z
⇒ lim fn dµ = f dµ.
n→∞

Corollary 5.1. Given any non-negative measurable function f : R → [0, ∞), let {sn } be the increasing
sequence of simple functions defined in the proof of Theorem 4.5 i.e. sn ↗ f on R. Then
Z Z
f dµ = lim sn dµ.
n→∞

Proof. Apply MCT directly.

Using the above corollary, we can prove the following properties of Lebesgue integrals:
Theorem 5.3. Suppose f, g : R → [0, ∞) are non-negative measurable functions and a, b > 0. Then
Z Z Z
(a) (af + bg) dµ = a f dµ + b g dµ;
Z
(b) f dµ = 0 if and only if f = 0 a.e..

19
Proof. For (a), by Corollary 5.1, there exist {sn } and {tn } sequences of simple functions such that sn ↗ f
and tn ↗ g as n → ∞ and
Z Z Z Z
lim sn dµ = f dµ, lim tn dµ = g dµ.
n→∞ n→∞

Then {asn + btn } is an increasing sequence of simple function such that asn + btn ↗ af + bg as n → ∞. By
MCT, we have
Z Z
(af + bg) dµ = lim (asn + btn ) dµ
n→∞
 Z Z 
= lim a sn dµ + b tn dµ
n→∞
Z Z
= a f dµ + b g dµ

For
Z (b), the “if” part was already proved in the exercise. Now we consider the “only if” part. Suppose
f dµ = 0. For any n ∈ N, define
 
1
Dn = x ∈ R f (x) ≥
n
Z Z
1
Then fn ≥ χDn and µ(Dn ) = χDn dµ ≤ n f dµ = 0. Therefore, µ(Dn ) = 0 for all n ∈ N.
n
∞
[
Let S = {x ∈ R | f (x) > 0} = Dn . Then we have
n=1

∞
!
[
µ(S) = µ Dn = lim µ(Dn ) = 0
n→∞
n=1

Hence, f = 0 a.e..
Remarks 5.3. The theorem still holds for Lebesgue integrals over any measurable set E ⊆ R.

Theorem 5.4. If fn : R → [0, ∞) is a measurable function for all n ∈ N, then

∞
Z X ∞ Z
X
fn dµ = fn dµ.
n=1 n=1

m
X ∞
X
Proof. Define gm = fn and g = fn . It is clear that gm ↗ g as m → ∞. By MCT, the result
n=1 n=1
follows.

20
Definition 5.3. Let f : R → R be a measurable function and E ∈ M. We say that f is Lebesgue Z
integrable over E if f + = max {f, 0} and f − = max {−f, 0} are Lebesgue integrable over E i.e. f + dµ
Z E

and f − dµ are both finite. Then we define

E
Z Z Z
f dµ = f + dµ − f − dµ.
E E E

Exercise 5.2. Suppose f, g are Lebesgue integrable functions and a, b ∈ R.

Z Z Z
(a) Show that (af + bg) dµ = a f dµ + b g dµ.

(b) Prove that if h is a measurable function such that |h| ≤ |f |, then h is Lebesgue integrable.
Z Z
(c) Prove that if f ≥ g a.e., then f dµ ≥ g dµ.
Z Z
(d) Show that |f | is Lebesgue integrable and f dµ ≤ |f | dµ.

21
The following is arguably the most useful convergence theorem about Lebesgue integrals:
Theorem 5.5. (Lebesgue’s Dominated Convergence Theorem LDCT) Let E ∈ M. Suppose g :
E → [0, ∞] is Lebesgue integrable on E and let {fn } be a sequence of measurable functions such that
|fn (x)| ≤ |g(x)| (n ∈ N, x ∈ E) and fn → f a.e. on E. Then f is Lebesgue integrable on E and
Z Z
f dµ = lim fn dµ.
E n→∞ E

Proof. Since |fn | = fn+ + fn− , fn+ ≤ g and fn− ≤ g. Hence, g is Lebesgue integrable implies that fn is also
Lebesgue integrable for all n ∈ N. Moreover, |f | ≤ g a.e. and so f is Lebesgue integrable.

Consider g + fn and g − fn . They are both non-negative measurable functions. By Fatou’s lemma, we have
Z Z
(g + f ) dµ ≤ lim inf (g + fn ) dµ
E n→∞ E
Z Z
⇒ f dµ ≤ lim inf fn dµ
E n→∞ E
Z Z
(g − f ) dµ ≤ lim inf (g − fn ) dµ
E n→∞ E
Z  Z 
⇒− f dµ ≤ lim inf − fn dµ
E n→∞ E
Z Z
⇒ f dµ ≥ lim sup fn dµ
E n→∞ E
Combine, we have Z Z Z Z
f dµ ≤ lim inf fn dµ ≤ lim sup fn dµ ≤ f dµ
E n→∞ E n→∞ E E
Z Z
⇒ f dµ = lim fn dµ.
E n→∞ E

22
Next, we will show that Lebesgue integral is indeed a generalization of Riemann integral:
Theorem 5.6. Suppose f : [a, b] → R is a bounded real-valued function such that it is Riemann integrable
Z b
on [a, b] i.e. f (x) dx exists and is finite. Then f is Lebesgue integrable on [a, b] and
a
Z b Z
f (x) dx = f dµ.
a [a,b]

Proof. Since f is Riemann integrable on [a, b], there exists a sequence of partitions {Pn } such that P1 ⊆
P2 ⊆ · · · and
U (f, Pn ) − L(f, Pn ) → 0 as n → ∞
Z b
f (x) dx = U (f ) = L(f ) = lim U (f, Pn ) = lim L(f, Pn ).
a n→∞ n→∞

For each partition Pn = {t0 , t1 , . . . , tm }, define the simple functions

m
mj χ[tj−1 ,tj )
X
un =
j=1

m
Mj χ[tj−1 ,tj )
X
vn =
j=1

where mj = inf f ([tj−1 , tj ]) and Mj = sup f ([tj−1 , tj ]) for j = 1, . . . m. Then it is clear that
Z Z
L(f, Pn ) = un dµ , U (f, Pn ) = vn dµ.
[a,b] [a,b]

It is clear that vn ↘ and un ↗. Moreover, u1 ≤ un ≤ f ≤ vn ≤ v1 on [a, b] for all n ∈ N. Hence lim vn

n→∞
and lim un exists. Define v(x) = lim vn and u(x) = lim un .
n→∞ n→∞ n→∞

Since f is bounded, there exists M > 0 such that |f | ≤ M on [a, b]. Obviously, the constant function M on
[a, b] is Lebesgue integrable on [a, b]. Therefore, f is Lebesgue integrable on [a, b]. Moreover,

|vn | ≤ M, |un | ≤ M

By LDCT, u and v are Lebesgue integrable on [a, b] and

Z Z Z b
u dµ = lim un dµ = lim L(f, Pn ) = L(f ) = f (x) dx
[a,b] n→∞ [a,b] n→∞ a

Z Z Z b
v dµ = lim vn dµ = lim U (f, Pn ) = U (f ) = f (x) dx
[a,b] n→∞ [a,b] n→∞ a

Also, we have Z Z Z
un ≤ f ≤ v n ⇒ un dµ ≤ f dµ ≤ vn dµ
[a,b] [a,b] [a,b]

Taking n → ∞, we get
Z b Z
f (x) dx = f dµ.
a [a,b]

23
Example 5.1. The followingZ ∞is an example of a real-valued continuous function on [0, ∞) such that the
improper Riemann integral f (x) dx converges but f is NOT Lebesgue integrable on [0, ∞).
0

Let n ∈ N. Define


 0 when x ∈ (−∞, 2n) ∪ (2n + 2, ∞)
 4 (x − 2n)

when x ∈ 2n, 2n + 21
 
fn (x) = n+1
−4
x ∈ 2n + 12 , 2n + 23
 
 (x − (2n + 1)) when
 n+1


4
x ∈ 2n + 32 , 2n + 2
 
n+1 (x − (2n + 2)) when

It is clear that for n ∈ N, fn is continuous on [0, ∞) and we have

Z 2n+1 Z 2n+2
1
fn (x) dx = , fn (x) dx = 0
0 n+1 0

Then define
∞
X
f (x) = fn (x)
n=0

Obviously, f is also continuous on [0, ∞). For any y > 0, there exists k ∈ N such that 2k < y < 2k + 2.
Therefore, we have
Z y Z 2k Z y Z y
f (x) dx = f (x) dx + f (x) dx = f (x) dx
0 0 2k 2k
2
For any ε > 0, there exists N ∈ N such that < ε. Then for y > 2N , let k ∈ N such that 2k < y < 2k+2,
N +1
Z y Z y Z 2k+2
2 2
f (x) dx = f (x) dx ≤ |f (x)| dx = ≤ <ε
0 2k 2k 2k + 1 2N +1
Z ∞
Hence f (x) dx converges and equals zero.
0

Claim: f is not Lebesgue integrable on [0, ∞).

Assume the contrary i.e. f is Lebesgue integrable on [0, ∞). Then |f | is also Lebesgue integrable on [0, ∞).
However, by Theorem 5.4, we have
Z Z ∞ ∞ Z ∞
X X X 2
|f | dµ = |fn | dµ = |fn | dµ = =∞
[0,∞) [0,∞) n=0 n=0 [0,∞) n=0
n + 1

i.e. a contradiction.

24
Theorem 5.7. Suppose
Z ∞ f : [0, ∞) → [0, ∞) is a nonnegative measurable function. Then the improper
Riemann integral f (x) dx converges implies that f is Lebesgue integrable and
0
Z Z ∞
f dµ = f (x) dx.
[0,∞) 0

Proof. Use the fact that f χ[0,n] ↗ f as n → ∞, MCT and Theorem ??, we have
Z ∞ Z n Z Z Z
f (x) dx = lim f (x) dx = lim f dµ = lim f χ[0,n] dµ = f dµ
0 n→∞ 0 n→∞ [0,n] n→∞ [0,∞) [0,∞)

More generally, we have the following result:

Theorem 5.8. Let I be an interval (possibly unbounded) and f ∈ Lloc (I) i.e. f is Riemann integrable
on all [s, t] ⊆ I. If |f | is improper Riemann integrable on I, then f is improper Riemann integrable and
Lebesgue integrable on I and two integrals give the same value.
Proof. By absolute convergence test, |f | is improper Riemann integrable on I implies that f is improper
Riemann integrable on I and therefore,

|f | + f |f | − f
f+ = f− =
2 2
are both improper Riemann integrable on I. It suffices to prove the theorem for f + and f − .

For the case when I is an unbounded interval, the proof is similar to the one in the previous theorem.

For the case when I = (a, b] such that f + is unbounded as x → a+ , we use the fact that f + χ[a+ 1 ,b] ↗ f +
n
as n → ∞, MCT and Theorem ??.
Z b Z b Z Z Z
+
f (x) dx = lim
n→∞ 1
+
f (x) dx = lim
n→∞ 1
f +
dµ = lim
n→∞
f +
χ[a+ 1
n ,b]
dµ = f + dµ
a a+ n [a+ n ,b] [a,b] [a,b]

The proof for f − and other types of interval are similar.

25
6 Applications of Convergence Theorems
In this section, we will go through some applications of convergence theorems of Lebesgue integrals such as
MCT and LDCT.
Example 6.1. Suppose E ∈ M and f : E → [0, ∞) is a measurable function. Find
Z  
f
lim n ln 1 + dµ.
n→∞ E n

First we define    n
f f
fn = n ln 1 + = ln 1 + .
n n
 x n
Note that fn is nonnegative and fn ↗ f (because ex = lim 1 + ). By MCT, we have
n→∞ n
Z   Z Z
f
lim n ln 1 + dµ. = lim fn dµ = f dµ.
n→∞ E n n→∞ E E

∞
X 1
Example 6.2. Let ζ(s) = be the Riemann zeta function for s > −1. Prove that for any p ∈ N,
n=1
ns

xp
Z
1
ζ(p + 1) = dµ.
p! [0,∞) ex −1

First, we have
∞
1 X
= xn , |x| < 1
1 − x n=0
∞
xp X
⇒ = xp e−nx , x>0
1 − e−x n=0
∞ ∞
xp X
p −(n+1)x
X
⇒ = x e = xp e−nx , x>0
ex − 1 n=0 n=1

Therefore, by Theorem 5.4, we have

∞ Z
xp
Z X
x
dµ = xp e−nx dµ
[0,∞) e −1 n=1 [0,∞)
∞
X p!
= p+1
(by parts p times)
n=1
n
= p!ζ(p + 1)

Hence
xp
Z
1
ζ(p + 1) = dµ.
p! [0,∞) ex − 1

26
Exercise 6.1. Compute the following limits if they exist and justify the calculations:
Z 1
1. lim fn (x) dx, where fn (x) = sin(xn ) on [0, 1]. (Example 1.2)
n→∞ 0
Z 1
ln(x + n) −x
2. lim e cos x dx.
n→∞ 0 n
∞ 2 2
n2 xe−n x
Z
3. lim dx.
n→∞ 0 1+x

27
Theorem 6.1. (Continuity of integrals) Asumme f : R × R → R is such that x 7→ f [t] (x) = f (x, t) is
measurable for each t ∈ R and t 7→ f (x, t) is continuous for each x ∈ R. Assume also that |f (x, t)| ≤ g(x)
for each x, t ∈ R, where g : R → R is Lebesgue integrable. Then the function f [t] is Lebesgue integrable for
each t and the function F : R → R defined by
Z Z
[t]
F (t) = f dµ = f (x, t) dµ(x)

is continuous.

(Note: The notation “dµ(x)” indicates that it is a Lebesgue integral with respect to variable x.)
Z Z
[t] [t] [t]
Proof. Since f is measurable and |f | ≤ g, we have |f | dµ ≤ g dµ < ∞ and so f [t] is Lebesgue
integrable for each t and F (t) is well-defined.

To show that F is continnous at t0 ∈ R, it suffices to show that for each sequence {tn } with lim tn = t0 ,
n→∞
we have lim F (tn ) = F (t0 ). Define fn (x) = f [tn ] (x). By continuity of t 7→ f (x, t), we have
n→∞

lim fn (x) = lim f (x, tn ) = f (x, t0 ) = f [t0 ] (x).

n→∞ n→∞

Also, |fn (x)| ≤ g(x) for each x ∈ R and g is Lebesgue integrable. Therefore, by LDCT, we have
Z Z Z
[tn ]
lim F (tn ) = lim f dµ = lim fn dµ = f [t0 ] dµ = F (t0 )
n→∞ n→∞ n→∞

Exercise 6.2. Let f be a measurable function such that it is Lebesgue integrable on [0, ∞). Define
Z
F (t) = f (x) cos(x − t) dµ(x)
[0,∞)

(a) Show that F (t) is continuous for all t ∈ R.

(b) Show that F (t) is differentiable for all t ∈ R and

Z
F ′ (t) = f (x) sin(x − t) dµ(x), t ∈ R.
[0,∞)

F (tn ) − F (t0 )
(Hint: For any t0 , take any sequence {tn } converging to t0 and compute lim .)
n→∞ tn − t0

28
(Optional) The following result is of great importance in Fourier analysis:
Theorem 6.2. (Riemann-Lebesgue Lemma) Let f be a measurable function that is Lebesgue integrable
on R. Then Z
lim f (x) sin(nx) dµ(x) = 0.
n→∞

Proof. We first consider the special case: Suppose f = χ[a,b] . Then

Z Z b
1
lim f (x) sin(nx) dµ(x) = lim sin(nx) dx = lim (cos(na) − cos(nb)) = 0.
n→∞ n→∞ a n→∞ n

It is easy to see that the above result also holds if f is replaced by a step function i.e. a linear combination of
finite number of characteristic functions of disjoint finite intervals. Now we consider f = χA , where A ∈ M
and µ(A) < ∞.

Claim: For any ε > 0, there exists a finite collection of disjoint open intervals {I1 , I2 , . . . , Im } such that
Z
|f − s| dµ < ε

m
χI
X
where s = i
i.e. a step function.
i=1
∞
ε [
First, there exists an open set U ⊃ A such that µ(U \ A) = . As U ⊆ R can be written as U = Ii , where
2 i=1
{Ii } is a countable collection of disjoint open intervals. Hence, we have
∞
X
µ(U ) = µ(Ii ).
i=1

Choose m ∈ N large enough such that !

m
[ ε
µ U\ Ii < .
i=1
2
m
χI , we have
X
Then define s = i
i=1

Z m
! m
! m
!
[ [ [ ε ε
|f − s| dµ = µ A \ Ii +µ Ii \ A ≤µ U\ Ii + µ (U \ A) < + = ε.
i=1 i=1 i=1
2 2

From the claim, we can deduce that given any simple function ϕ, for any ε > 0, there exists a step function
s such that Z
|ϕ − s| dµ < ε.

(Why?)

29
Now, we are ready to prove the case: f is a nonnegative measurable function such that it is Lebesgue
integrable on R. Then there exists a sequence of simple functions {ϕn } such that ϕn ↗ f as n → ∞.
Therefore, given any ε > 0, there exist k ∈ N such that
Z
ε
|f − ϕk | dµ < .
3
Moreover, we already proved that there exists a step function s such that
Z
ε
|ϕk − s| dµ < .
3
Z
Since lim s(x) sin(nx) dµ(x) = 0, there exist N ∈ N such that for any n > N ,
n→∞

Z
ε
s(x) sin(nx) dµ(x) < .
3

Hence, combining these estimates, we have

Z Z Z
f (x) sin(nx) dµ(x) ≤ s(x) sin(nx) dµ(x) + (f (x) − s(x)) sin(nx) dµ(x)
Z
ε
≤ + |f − s| dµ
3
Z Z
ε
≤ + |f − ϕk | dµ + |ϕk − s| dµ
3
ε ε ε
< + + =ε
3 3 3
Finally, we prove the general case: Suppose f is a measurable function such that it is Lebesgue integrable
on R. Then Z Z
lim f + (x) sin(nx) dµ(x) = 0, lim f − (x) sin(nx) dµ(x) = 0.
n→∞ n→∞
+ −
As f = f − f , we have Z
lim f (x) sin(nx) dµ(x) = 0.
n→∞

30