CHAPTER 1

Lebesgue Measure and Integration

1.1. SET FUNCTIONS

If A and B are any two sets, we write A − B for the set of all elements x such that x ∈ A,
x∈/ B. The notation A − B does not imply that B ⊂ A. We denote the empty set by ∅, and say
that A and B are disjoint if A ∩ B = ∅.
Definition 1. A family R of sets is called a ring if A, B ∈ R implies
(1.1.1) A ∪ B ∈ R, and A − B ∈ R
Remark 1. Note that
A ∩ B = A − (A − B) = B − (B − A)
so for any A, B ∈ R we also have A ∩ B ∈ R if R is a ring.
The set F(N) = {A ⊂ N : A is finite } is an example of ring.
Definition 2. A ring R is called a σ-ring if
[∞
(1.1.2) An ∈ R
n=1
whenever An ∈ R for all n = 1, 2, 3, . . ..
Since
∞ ∞ ∞ ∞
!
\ \ \ [
(1.1.3) An = A1 ∩ An = A1 − A1 − An = A1 − (A1 − An ) ,
n=1 n=2 n=2 n=1
∞
T
for any An ∈ R, n = 1, 2, . . ., we also have An ∈ R if R is a σ-ring.
n=1
An Example of σ-ring is P (X) the set of all subsets of and set X.
∞
S
Note that F(N) is not a σ-ring, because E = {2n} the set of all even numbers is not finite.
n=1

Definition 3. We say that φ is a set function defined on σ-ring R if φ assigns to every A ∈ R

a number φ(A) of the extended real number system. φ is additive if A ∩ B = ∅ implies
(1.1.4) φ(A ∪ B) = φ(A) + φ(B),
and φ is countably additive if Ai ∩ Aj = ∅ for i 6= j, (in this case we say that the family Ai is
pairwise disjoint) implies
∞ ∞
!
[ X
(1.1.5) φ An = φ(An ).
n=1 n=1

i
 ii 1. LEBESGUE MEASURE AND INTEGRATION

Remark 2. Here we will assume that φ is not the constant functions whose only value is +∞
or −∞, and that the range does not contain both +∞ and −∞, because if it did, the right side of
(1.1.4) could lose meaning.
Remark 3. Note that the left side of (1.1.5) is independent of the order in which the An ’s
are arranged. Hence, by the rearrangement theorem for series, if the right hand side of (1.1.5))
converges, it converges absolutely. Otherwise, the partial sums tend to +∞ or −∞.
Theorem 1. If φ is additive, then
(1) φ(∅)
 k= 0. 
S Pk
(2) φ An = n=1 φ(An ), if Ai ∩ Aj = ∅ for i 6= j.
n=1
(3) φ (A1 ∪ A2 ) + φ (A1 ∩ A2 ) = φ (A1 ) + φ (A2 ).
(4) If φ(A) ≥ 0 for all A, and A ⊂ B, then φ (A) ≤ φ (B).
added — PS (5) if A ⊂ B,and |φ (A)| < ∞, then φ (B − A) = φ (B) − φ (A).

(1) Note that for all A ∈ R, A = A ∪ ∅, and A ∩ ∅ = ∅. Since φ is additive, then

φ (A) + 0 = φ (A) = φ (A ∪ ∅) = φ (A) + φ (∅)
so 0 = φ (∅)
(2) The case n = 2 is the definition of additive function. Suppose inductively that
φ (A1 ∪ A2 ∪ A3 ∪ · · · ∪ An ) = φ (A1 ) + φ (A2 ) + φ (A3 ) + · · · + φ (An ) ,
if Ai ∩ Aj = ∅ for i 6= j, and let A1 , A2 , A3 , . . . , An+1 so that Ai ∩ Aj = ∅ for i 6= j, and
define B1 = A1 ∪ A2 ∪ A3 ∪ · · · ∪ An , B2 = An+1 , then B1 ∩ B2 = ∅ and
φ (A1 ∪ A2 ∪ A3 ∪ · · · ∪ An+1 ) = φ (B1 ∪ B2 ) = φ (B1 ) + φ (B2 )
= φ (A1 ∪ A2 ∪ A3 ∪ · · · ∪ An ) + φ (An+1 )
= φ (A1 ) + φ (A2 ) + φ (A3 ) + · · · + φ (An ) + φ (An+1 )
where in the last equality we use the inductive hypothesis.
(3) Note that
A1 = (A1 − A2 ) ∪ (A1 ∩ A2 )
A2 = (A2 − A1 ) ∪ (A1 ∩ A2 )
A1 ∪ A2 = (A1 − A2 ) ∪ (A2 − A1 ) ∪ (A1 ∩ A2 )

(see figure below), so

φ (A1 ∪ A2 ) + φ (A1 ∩ A2 ) = (φ (A1 − A2 ) + φ (A2 − A1 ) + φ (A1 ∩ A2 )) + φ (A1 ∩ A2 )

= (φ (A1 − A2 ) + φ (A1 ∩ A2 )) + (φ (A2 − A1 ) + φ (A1 ∩ A2 ))
= φ (A1 ) + φ (A2 )
 1.1. SET FUNCTIONS iii

(4) Since A ⊂ B, then B = A∪(B − A) and A∩(B − A) = ∅, so φ (B) = φ (A)+φ (B − A) ≥

φ (A), because φ (B − A) ≥ 0
(5) By item 4 we have φ (B) = φ (A) + φ (B − A), since |φ (A)| < ∞, then
φ (B) − φ (A) = (φ (A) + φ (B − A)) − φ (A) = φ (B − A)
Note that non-negative additive set functions satisfy item 4, because this fact these functions
are called monotonic.
Theorem 2. Suppose φ is countably additive on a ring R. Suppose An ∈ R for n = 1, 2, 3,
. . ., A1 ⊂ A2 ⊂ A3 ⊂ · · · ⊂ A ∈ R and
[∞
A= An .
n=1
Then,
φ(A) = lim φ(An ).
n→∞

Let B1 = A1 , and Bn = An − An−1 for n = 2, 3, . . . . Then Bi ∩ Bj = ∅, for i 6= j. Indeed,

since i 6= j we can suppose i < j, then by hypothesis
Ai−1 ⊂ Ai ⊆ Aj−1 ⊂ Aj .
So
Bi = Ai − Ai−1 ⊂ Ai ⊆ Aj−1
Since Aj−1 ∩ Bj = Aj−1 ∩ (Aj − Aj−1 ) = ∅, then Bi ∩ Bj = ∅.
n
S
On the other hand, using induction we show that An = Bj . added — PS
j=1
For n = 1, the above equality is A1 = B1 , and suppose that this equality is true for n = k,
since
Ak+1 = (Ak+1 − Ak ) ∪ Ak ,
then
k
[ k+1
[
Ak+1 = Bk+1 ∪ Ak = Bk+1 ∪ Bj = Bj .
j=1 j=1
n
S n
S
Furthermore, A1 ⊂ A2 ⊂ A3 ⊂ · · · ⊂ An , implies Aj = An = Bj . Then
j=1 j=1
∞
[ ∞
[
A= Aj = Bj .
j=1 j=1

The figure below illustrates this fact

Since φ is countably additive, then
 
n
[ n
X
φ(An ) = φ  Bj  = φ(Bj ),
j=1 j=1

and  
∞
[ ∞
X
φ(A) = φ  Bj  = φ(Bj ).
j=1 j=1
iv 1. LEBESGUE MEASURE AND INTEGRATION

So
n
X ∞
X
lim φ(An ) = lim φ(Bj ) = φ(Bj ) = φ(A)
n→∞ n→∞
j=1 j=1

as the theorem states.

Theorem 3. Suppose {φn : n = 1, 2, 3, . . .} is sequence of countably additive functions on a
ring R. Suppose φn ≤ φn+1 for n = 2, 3, . . ., then
φ = sup {φn : n = 1, 2, 3, . . .}
is a countably additive function on R.
∞
S
Note that for each B ∈ R,φn (B) ≤ φ (B) for all n ∈ N. Hence if A = Aj , with Ai ∩ Aj = ∅,
j=1
then for each n ∈ N we have
∞
X ∞
X
φn (A) = φn (Aj ) ≤ φ (Aj ) .
j=1 j=1

So
∞
X
(1.1.6) φ (A) = sup φn (A) ≤ φ (Aj )
n∈N j=1

k
S
On the other hand, Since Bk = Aj ⊂ A, then
j=1

φ (Bk ) = sup {φn (Bk )} ≤ sup {φn (A)} = φ (A)

n∈N n∈N

Fix k ∈ N and ε > 0, for each j = 1, 2, . . . , k there exists nj so that φ (Aj ) < φnj (Aj ) + ε/k.
Since {φn (Aj ) : n ∈ N} is increasing, taking n = max {nj : j = 1, 2, . . . , k}, and using the countable
 1.2. CONSTRUCTION OF THE LEBESGUE MEASURE v

aditivity of φn we have
     
Xk Xk k
[ k
[
φ (Aj ) ≤  φn (Aj ) + ε = φn  Aj  + ε ≤ φ  Aj  + ε ≤ φ (A) + ε.
j=1 j=1 j=1 j=1

Since ε was arbitrary, then

k
X
(1.1.7) φ (Aj ) ≤ φ (A) ,
j=1

and taking limit when k → ∞, we obtain???

X∞
(1.1.8) φ (Aj ) ≤ φ (A)
j=1

Equations (1.1.6) and (1.1.8) imply

∞
X
φ (Aj ) = φ (A) .
j=1

1.2. CONSTRUCTION OF THE LEBESGUE MEASURE

Definition 4. Let Rp denote p-dimensional space. By an interval in Rp we mean the set of
points x = (x1 , . . . , xp ) such that
(1.2.1) ai ≤ xi ≤ bi
for i = 1, . . . , p, or the set of points which is characterized by (1.2.1) with any or all of the signs
≤ replaced by <. The possibility that ai = bi , for any value of i is not ruled out; in particular, the
empty set is included among the intervals.
Note that an interval in Rp is the cartesian product of finite intervals (closed, open, semiopen
or degenarate) of R.
Definition 5. If A is the finite union of intervals, A is said to be an elementary set.
If I is an interval, we define the measure of I by
p
Y
m (I) = (bi − ai )
i=1

no matter whether equality is included or excluded in any of the inequalities (1.2.1).

Remark 4. If I = I1 × I2 × · · · × Ip and J = J1 × J2 × · · · × Jp where I1 , . . . , and J1 ,
. . . ,I ∩ J = (I1 ∩ J1 ) × (I2 ∩ J2 ) × · · · × (Ip ∩ Jp )
Remark 5. If I, J are two finite intervals of R, then I − J can be written as the union of two
(possible empty) intervals. Indeed, let a ≤ b be the extreme points of I and c ≤ d be the extreme
points of J, then we can have
(a) a ≤ b ≤ c ≤ d, in this case I − J = I.
(b) a ≤ c ≤ b ≤ d, in this case I − J is the interval with extreme points a, c.
(c) a ≤ c ≤ d ≤ b, in this case I − J is the union of the intervals with extreme points a, c and
d, b.
 vi 1. LEBESGUE MEASURE AND INTEGRATION

(d) c ≤ a ≤ d ≤ b, in this case I − J is the interval with extreme points d, b.

(e) c ≤ d ≤ a ≤ b, in this case I − J = I.
added — PS (f) c ≤ a ≤ b ≤ d, in this case I − J = ∅.
Note that, if I = I1 × I2 × · · · × Ip and J = J1 × J2 × · · · × Jp , then
(1.2.2)
I − J = (I1 − J1 )×I2 ×· · ·×Ip ∪(I1 − J1 )×(I2 − J2 )×· · ·×Ip ∪· · ·∪(I1 − J1 )×(I2 − J2 )×· · ·×(Ip − Jp )
Thus, by Remark 5 I − J is the union of intervals disjoint in Rp .
k
S
Definition 6. If If the intervals Ij are pairwise disjoint, then for A = Ij , we set
j=1

k
X
(1.2.3) m (A) = m(Ij )
j=1

We denote by E the family of all elementary subsets of Rp . Note that E satisfies the following
properties.
k
S l
S
E 1 E is a ring, but not a σ-ring. Clearly the if A = In and B = Jm , then A ∪ B =
n=1 m=1
k
S l
S
In ∪ Jm and
n=1 m=1

k l
! k l
!
[ [ [ \
A−B = In − Jm = (In − Jm )
n=1 m=1 n=1 n=1

but In − Jm is the union of at most 2p intervals in Rp and by Remark 4 the intersection

of intervals is an interval, then A − B is a finite union of intervals in Rp .
added — PS Finally note that E is not a σ-ring: if Rp is an element of E, since Rp can not be
written as a finite union of intervals in Rp then E is not a σ-ring
E 2 If A ∈ E, then A is the union of a finite number of disjoint intervals. If A is an interval
this fact is obvious. Now suppose that all unions of k intervals is the union of a finite
k+1
S
number of disjoint intervals, and let A = In , then
n=1

k+1
[ k
[ l
[
A= In = Ik+1 ∪ In = Ik+1 ∪ Jn
n=1 n=1 n=1
l
! l
[ [
= Ik+1 − Jn ∪ Jn
n=1 n=1
l
! l
\ [
= (Ik+1 − Jn ) ∪ Jn
n=1 n=1

where the Jn are disjoint intervals, since Ik+1 − Jn is the finite union of disjoint intervals
(see 5) and the intersection of intervals is an interval, then A is the union of a finite number
of disjoint intervals.
 1.2. CONSTRUCTION OF THE LEBESGUE MEASURE vii

E 3 If A ∈ E, m(A) is well defined by (1.2.3); that is, if two different decompositions of A

into disjoint intervals are used, each gives rise to the same value of m(A). Indeed, If
k
S Sl
A = Ir = Jq , where the intervals Ir and Jq are pairwise disjoint, then for each
r=1 q=1
r = 1, 2, . . . , k and q = 1, 2, . . . , l we have
l
[
Ir = A ∩ In = ( Jq ∩ I r )
q=1
k
[
Jq = A ∩ Jq = ( Jq ∩ Ir )
r=1

since the family {Bqr = Jq ∩ Ir : r = 1, 2, . . . , k and q = 1, 2, . . . , l} is pairwise disjoint,

then
k k l
! k l
!
X X X X X
m(A) = m(Ir ) = m (Jq ∩ Ir ) = m (Bqr )
r=1 r=1 q=1 n=1 q=1
l k
! l k
! l
X X X X X
= m (Bqr ) = m (Jq ∩ Ir ) = m(Jq )
q=1 r=1 q=1 r=1 q=1

k
S l
S
E 4 m is additive on E. Indeed, If A = Ir and B = Jq , where the intervals Ir and Jq
r=1 q=1
are pairwise disjoint, and A ∩ B = ∅, then Ir ∩ Jq = ∅ for each r = 1, 2, . . . , k and q = 1,
2, . . . , l.
Sk Sl
Since A ∪ B = Ir ∪ Jq , then
r=1 q=1

k l
! k l
[ [ X X
m (A ∪ B) = m Ir ∪ Jq = m (Ir ) + m (Jq ) = m (A) + m (B)
r=1 q=1 r=1 q=1

Note that if p = 1, 2, 3, then m is length, area, and volume, respectively.

Definition 7. A non-negative additive set function φ defined on E is said to be regular if the
following is true: To every A and to every ε > 0 there exist sets F , G ∈ E such that F is closed, G
is open, F ⊂ A ⊂ G, and
(1.2.4) φ (G) − ε ≤ φ (A) ≤ φ (F ) + ε
k
S
Note that by 1.2 we have A = In where In are intervals pairwise disjoint. So for each
n=1
n = 1, 2, . . . , k, if Fn is a closed set and Gn is an open set, such that Fn ⊂ In ⊂ Gn and
ε ε
φ (Gn ) − ≤ φ (In ) ≤ φ (Fn ) +
k k
k
S k
S
Then F = Fn and G = Gn , satisfy requirement in Definition 7 for A. Thus, to show that
n=1 n=1
φ is regular on E It is sufficient to verify the conditions of Definition 7 only in the intervals of Rp .
viii 1. LEBESGUE MEASURE AND INTEGRATION

Exercise 1. [Exercise 11.15] Let R be the ring of all elementary subsets of (0, 1]. If 0 < a <
b ≤ 1, define
φ ((a, b)) = φ ((a, b]) = φ ([a, b)) = φ ([a, b]) = b − a
but define
φ ((0, b)) = 1 + b
if 0 < b ≤ 1. Show that this gives an additive set function φ on R, which is not regular and which
cannot be extended to a countably additive set function on a σ-ring.
Solution 1. Here as in Definition 6 we define
Xk
(1.2.5) φ (A) = φ(Ij ),
j=1
k
S
if A = Ij , and the intervals Ij are pairwise disjoint.
j=1
First, if A is an elementary set, φ(A) is well defined by (1.2.5); that is, if two different decom-
positions of A into disjoint intervals are used, each gives rise to the same value of φ(A). Indeed, if
k
S Sl
A= In = Jm , where the intervals In and Jm are pairwise disjoint, then for each n = 1, 2,
n=1 m=1
. . . , k and m = 1, 2, . . . , l we have
l
[
In = A ∩ In = (Jm ∩ In )
m=1
k
[
Jm = A ∩ Jm = (Jm ∩ In )
n=1
since the family {Bmn = Jm ∩ In : n = 1, 2, . . . , k and m = 1, 2, . . . , l} is pairwise disjoint, then
k k l
! k l
!
X X X X X
φ(A) = φ(In ) = φ (Jm ∩ In ) = φ (Bmn )
n=1 n=1 m=1 n=1 m=1
l k
! l k
! l
X X X X X
= φ (Bmn ) = φ (Jm ∩ In ) = φ(Jm )
m=1 n=1 m=1 n=1 m=1
k
S
Recall that if A is an elementary set, then A = Ij is the union of a finite number of disjoint
j=1
intervals (see 1.2). So
( Pk
j=1 l(Ij ) if 0 is not the end point of any interval in A
φ (A) = P k
1 + j=1 l(Ij ) if 0 is the end point of any interval in A
where l(Ij ) is the length of the interval Ij . In particular, φ (A) < 1 if A is a closed set of (0, 1].
Indeed, note that 0 ∈ / A for any subset A ⊂ (0, 1].
Now, if 0 is the endpoint of any interval in A, since A is closed then 0 is limit point of A, and
0 ∈ A. which is clearly a contradiction.
If two elementary sets A and B are disjoint, at most one of them can have the point 0 as
the endpoint of one of its intervals. Then φ (A ∪ B) is the sum of the lengths of the intervals in
A ∪ B if neither set contains an interval having 0 as the endpoint, and 1 plus this sum if one of
 1.2. CONSTRUCTION OF THE LEBESGUE MEASURE ix

them does contain an interval with 0 as endpoint. In either case φ (A ∪ B) = φ (A) + φ (B) when
A ∩ B = ∅.Thus, the function φ is additive.
The function φ is not regular, because by definition φ 0, 21 
= 1 + 12 = 32 , but φ(A)
< 1 if A


is closed, so taking ε = 13 , we have φ(A) + 31 < 1 + 12 = φ 0, 21 for all closed A ⊂ 0, 12 . Thus,

φ does not satisfy Definition 7.
The function also cannot be extended to a countably additive set function on a σ-ring, because
  ∞  
1 [ 1 1
0, = ,
2 n=1
2n+1 2n
the intervals in this union are pairwise disjoint, but
  ∞ ∞  
1 3 1 X 1 X 1 1
φ 0, = > = = φ , .
2 2 2 n=1 2n+1 n=1
2n+1 2n

Example 1.

(a) The set function m is regular. If A is an interval, i.e., A = I1 × I2 × · · · × In , if ai , bi are

the extreme points of Ii , then by the continuity of the volume function on Rp , we can
choose r such that
G = (a1 − r, b1 + r) × (a2 − r, b2 + r) × · · · × (an − r, bn + r)
F = [a1 + r, b1 − r] × [a2 + r, b2 − r] × · · · × [an + r, bn − r]
and
n
Y n
Y
φ (G) = (bj − aj + 2r) ≤ (bj − aj ) + ε = φ (A) + ε
j=1 j=1
Yn n
Y
φ (A) − ε = (bj − aj ) − ε ≤ (bj − aj − 2r) = φ (F )
j=1 j=1

(b) Take p = 1, and let α be a monotonically increasing function defined for all real x. Put
µ ([a, b)) = α (b−) − α (a−) = sup α (t) − sup α (t)
t<b t<a
µ ([a, b]) = α (b+) − α (a−) = inf α (t) − sup α (t)
b<t t<a
µ ((a, b]) = α (b+) − α (a+) = inf α (t) − inf α (t)
b<t a<t
µ ((a, b)) = α (b−) − α (a+) = sup α (t) − inf α (t)
t<b a<t

Recall that if α is a monotonically increasing function then the set of points of discontinuity of α
is at most countable, and if α is continuous at x, then α (x−) = α (x) = α (x+). Also for each a,
b ∈ R with a < b, by the definition of infimum and supremum given ε > 0 there exists c, x, y, d
with c < a < x < y < b < d, so that α is continuous at c, x, y, d and
ε ε
α (a−) − < α (c) α (x) < α (a+) +
2 2
ε ε
α (b−) − < α (y) α (d) < α (b+) +
2 2
x 1. LEBESGUE MEASURE AND INTEGRATION

or equivalently
ε ε
−α (c) − < −α (a−) − α (a+) < −α (x) +
2 2
ε ε
α (b−) < α (y) + α (d) − < α (b+) .
2 2
The behavior of a monotonically increasing function around a discontinuity point x is sketched in
the figure below

Now we show that µ is regular on E. Here c, x, y, d are as above

(1) In the case A = [a, b) consider F = [a, y] and G = (c, b), then
ε ε
φ (G) − ε < φ (G) − = α (b−) − α (c) − ≤ α (b−) − α (a−) = φ (A)
2 2
ε ε
φ (A) = α (b−) − α (a−) ≤ α (y) + − α (a−) = φ (F ) + < φ (F ) + ε
2 2
(2) In the case A = [a, b] consider F = [a, b] and G = (c, d), then
ε ε
φ (G) − ε = α (d) − − α (c) − ≤ α (b+) − α (a−) = φ (A)
2 2
φ (A) = φ (F ) < φ (F ) + ε
(3) In the case A = (a, b] consider F = [x, b] and G = (a, d), then
ε ε
φ (G) − ε < φ (G) − = α (d) − − α (a+) ≤ α (b+) − α (a+) = φ (A)
2 2
ε ε
φ (A) = α (b+) − α (a+) ≤ α (b+) − α (x) + = φ (F ) + < φ (F ) + ε
2 2
(4) In the case A = (a, b) consider F = [x, y] and G = (a, b), then
φ (G) − ε < φ (G) = φ (A)
ε ε
φ (A) = α (b−) − α (a+) ≤ α (y) + − α (x) + = φ (F ) + ε
2 2

Now we show that every regular set function on E can be extended to a countably additive set
function on a σ-ring which contains E.
Definition 8. Let µ be additive, regular, non-negative, and finite on E. Consider countable
coverings of any set E ⊂ Rp by open elementary sets An .
[∞
E⊂ An .
n=1
 1.2. CONSTRUCTION OF THE LEBESGUE MEASURE xi

Define
∞
X
(1.2.6) µ∗ (E) = inf µ (An ) ,
n=1
where the infimum is taken over all countable coverings of E by open elementary sets.
µ∗ is called the outer measure of E, corresponding to µ.
It is clear that µ∗ (E) ≥ 0 for all E and that if E1 ⊂ E2 , then any countable coverings of E2
by open elementary sets is a countable coverings of E1 by open elementary sets and by properties
of infimum we have
(1.2.7) µ∗ (E1 ) ≤ µ∗ (E2 ) .
Theorem 4.

(a) For every A ∈ E, µ∗ (E) = µ (E)

∞
S
(b) if E ⊂ En , then
n=1
∞
X
(1.2.8) µ∗ (E) ≤ µ∗ (En )
n=1
∗
Remark 6. (a) implies that µ is an extension of µ from E to P (Rp ). The property (1.2.8) is
called subadditivity.
Choose A ∈ E and ε > 0. The regularity of µ shows that A is contained in an open elementary
set G such that
µ(G) ≤ µ(A) + ε.
∗
Since µ (A) ≤ µ(G) and ε is arbitrary, we have

(1.2.9) µ∗ (A) ≤ µ(A)

By the properties of the infimum there is a sequence of open elementary sets whose union contains
A, such that
∞
X ε
µ (An ) ≤ µ∗ (A) +
n=1
2
The regularity of µ implies that A contains a closed elementary F such that µ (A) ≤ µ (F ) + 2ε ;
and note that F is bounded, because it is the finite union of finite intervals, so F is compact, and
we have
[k
F ⊂ An ,
n=1
for some k. Since µ is additive, using Theorem 1, item 2 we have
∞
k
! k
ε [ ε X ε X ε
µ(A) ≤ µ (F ) + ≤ µ An + ≤ µ (An ) + ≤ µ (An ) + ≤ µ∗ (A) + ε,
2 n=1
2 n=1
2 n=1
2
since ε is arbitrary we obtain
(1.2.10) µ(A) ≤ µ∗ (A) ,
xii 1. LEBESGUE MEASURE AND INTEGRATION

Equations (1.2.9) and (1.2.10) prove (a).

∞
En . If µ∗ (En ) = +∞ for some n, then the right side of (1.2.8) is equal
S
Now, suppose E =
n=1
to +∞, and (1.2.8) is trivial, so we can assume that µ∗ (En ) < +∞ for all n.
Given ε > 0, there are coverings {Ank } for k = 1, 2, 3, · · · of En by elementary sets, such that
∞
X ε
µ (Ank ) ≤ µ∗ (En ) +
2n
k=1
Since
∞
[ ∞ [
[ ∞
E= En ⊂ Ank
n=1 n=1 k=1
then
∞ [
∞ ∞ X
∞ ∞  ∞
! !
∗ ∗
[ X X ε  X
µ (E) ≤ µ Ank ≤ µ (Ank ) ≤ µ∗ (En ) + = ε + µ∗ (En )
n=1 k=1 n=1 k=1 n=1
2n n=1

and since ε is arbitrary (1.2.8) follows.

Definition 9. For any A, B ⊂ Rp We define

(1.2.11) S (A, B) = (A − B) ∪ (B − A)

(1.2.12) d (A, B) = µ∗ (S (A, B))

S(A, B) is called symmetric difference of A and B. Now we will see some properties of
S (A, B) and d (A, B)
Lemma 1. For any A, B, C, A1 , A2 , B1 , B2 in Rp we have;
S1 S (A, B) = S (B, A) , S (A, A) = ∅.
S2 S (A, B) ⊂ S (A, C) ∪ S (C, B) .
S3 
S (A1 ∪ A2 , B1 ∪ B2 ) 
S (A1 ∩ A2 , B1 ∩ B2 ) ⊂ S (A1 , B1 ) ∪ S (A2 , B2 ) .
S (A1 − A2 , B1 − B2 )


S1 Since X ∪ Y = Y ∪ X for all X, Y ⊆ Rp , so

S (A, B) = (A − B) ∪ (B − A) = (B − A) ∪ (A − B) = S (B, A) ,
and X − X = ∅ for all X ⊆ Rp , implies
S (A, A) = (A − A) ∪ (A − A) = ∅ ∪ ∅ = ∅.
S2 Note that A ⊂ A ∪ C = (A − C) ∪ C, so
A − B ⊂ ((A − C) − B) ∪ (C − B) ⊂ (A − C) ∪ (C − B) ,
interchanging A and B we get
B − A ⊂ (B − C) ∪ (C − A)
 1.2. CONSTRUCTION OF THE LEBESGUE MEASURE xiii

Thus,
S (A, B) = (A − B) ∪ (B − A)
⊂ ((A − C) ∪ (C − B)) ∪ ((B − C) ∪ (C − A))
= ((A − C) ∪ (C − A)) ∪ ((C − B) ∪ (B − C))
= S (A, C) ∪ S (C, B)

S3

(a) If Y c = Rp − Y (the complement of Y in Rp ), then X − Y = X ∩ Y c , so

c
(A1 ∪ A2 ) − (B1 ∪ B2 ) = (A1 ∪ A2 ) ∩ (B1 ∪ B2 )
= (A1 ∪ A2 ) ∩ (B1c ∩ B2c )
= (A1 ∩ (B1c ∩ B2c )) ∪ (A2 ∩ (B1c ∩ B2c ))
⊂ (A1 ∩ B1c ) ∪ (A2 ∩ B2c )
= (A1 − B1 ) ∪ (A2 − B2 )
and
(B1 ∪ B2 ) − (A1 ∪ A2 ) ⊂ (B1 − A1 ) ∪ (B2 − A2 )
so
S (A1 ∪ A2 , B1 ∪ B2 ) = ((A1 ∪ A2 ) − (B1 ∪ B2 )) ∪ ((B1 ∪ B2 ) − (A1 ∪ A2 ))
⊂ ((A1 − B1 ) ∪ (A2 − B2 )) ∪ ((B1 − A1 ) ∪ (B2 − A2 ))
= ((A1 − B1 ) ∪ (B1 − A1 )) ∪ ((A2 − B2 ) ∪ (B2 − A2 ))
= S (A1 , B1 ) ∪ S (A2 , B2 ) .
(b) Since
S (A, B) = (A ∩ B c ) ∪ (B ∩ Ac )
c c
= ((Ac ) ∩ B c ) ∪ ((B c ) ∩ Ac )
c c
= (Ac ∩ (B c ) ) ∪ (B c ∩ (Ac ) )
= S (Ac , B c ) ,
then
c c
S (A1 ∩ A2 , B1 ∩ B2 ) = S ((A1 ∩ A2 ) , (B1 ∩ B2 ) )
= S (Ac1 ∪ Ac2 , B1c ∪ B2c )
⊂ S (Ac1 , B1c ) ∪ S (Ac2 , B2c )
= S (A1 , B1 ) ∪ S (A2 , B2 )
(c)
S (A1 − A2 , B1 − B2 ) = S (A1 ∩ Ac2 , B1 ∩ B2c )
⊂ S (A1 , B1 ) ∪ S (Ac2 , B2c )
= S (A1 , B1 ) ∪ S (A2 , B2 )
These properties of S(A, B) imply
xiv 1. LEBESGUE MEASURE AND INTEGRATION

Lemma 2. Forany A, B, C, A1 , A2 , B1 , B2 in Rp we have;

D1 d (A, B) = d (B, A), d (A, A) = 0.
D2 d (A, B) ≤ d (A, C) + d (C, B).
D3

d (A1 ∪ A2 , B1 ∪ B2 ) 
(1.2.13) d (A1 ∩ A2 , B1 ∩ B2 ) ≤ d (A1 , B1 ) + d (A2 , B2 ) .
d (A1 − A2 , B1 − B2 )


D1
d (A, B) = µ∗ (S (A, B)) = µ∗ (S (B, A)) = d (B, A)
d (A, A) = µ∗ (S (A, A)) = µ∗ (∅) = 0
D2 Since S (A, B) ⊂ S (A, C) ∪ S (C, B), by (1.2.7) and (1.2.8)
d (A, B) = µ∗ (S (A, B)) ≤ µ∗ (S (A, C) ∪ S (C, B))
≤ µ∗ (S (A, C)) + µ∗ (S (C, B)) = d (A, C) + d (C, B)
D3 As in the proof of D2, the inclusions in S3 and (1.2.7) and (1.2.8) imply

d (A1 ∪ A2 , B1 ∪ B2 ) 
d (A1 ∩ A2 , B1 ∩ B2 ) ≤ d (A1 , B1 ) ∪ d (A2 , B2 ) .
d (A1 − A2 , B1 − B2 )


The relations D1 and D2 show that d(A, B) satisfies the requirements of definition for a distance
except that d(A, B) = 0 does not imply A = B. For instance, if p = 1, µ = m, A = {an ∈ R : n ∈ N}
is countable, and B = ∅, then
d(A, B) = m∗ ((A − ∅) ∪ (∅ − A)) = m∗ (A)
 ε ε  ∞
S
If ε > 0, taken In = an − n+1 , an + n+1 , then In are elementary open sets and A ⊂ In
2 2 n=1
and
∞ ∞
X X ε
m∗ (A) ≤ m (In ) = n
=ε
n=1 n=1
2
Since ε is arbitrary, then m∗ (A) = 0.
If B = ∅, then D2 tells us that
µ∗ (A) = d(A, ∅) ≤ d(A, C) + d(C, ∅) = d(A, C) + µ∗ (C)
interchanging A and C we get
µ∗ (C) ≤ d(A, C) + µ∗ (A)
So if at least one of µ∗ (A) , µ∗ (C) is finite, then
(1.2.14) |µ∗ (A) − µ∗ (C)| ≤ d(A, C)
We write An → A, if
lim d(A, An ) = 0.
n→∞
If there is a sequence {An } of elementary sets such that An → A, we say that A is finitely
µ-measurable and write A ∈ MF (µ) .
 1.2. CONSTRUCTION OF THE LEBESGUE MEASURE xv

If A is the union of a countable collection of finitely µ-measurable sets, we say that A is

µ-measurable and write A ∈ M (µ) .
Theorem 5. M (µ) is a σ-ring, and µ∗ is countably additive on M (µ) .
Let A, B ∈ MF (µ), and {An }, {Bn } be sequences of elementary sets such that An → A and
Bn → B, then by 1.2.13 and 1.2.14 we have
d (An ∪ Bn , A ∪ B) ≤ d (An , A) + d (Bn , B)
d (An ∩ Bn , A ∩ B) ≤ d (An , A) + d (Bn , B)
d (An − Bn , A − B) ≤ d (An , A) + d (Bn , B)
|µ∗ (An ) − µ∗ (A)| ≤ d (An , A) .

Since µ∗ (An ) < ∞, taking limit when n → ∞,we have

(1.2.15a) An ∪ Bn → A∪B
(1.2.15b) An ∩ Bn → A∩B
(1.2.15c) An − B n → A−B
(1.2.15d) µ (An ) → µ∗ (A)
∗

So µ∗ (A) < ∞. By (1.2.15a) and (1.2.15c), MF (µ) is a ring. Since µ is additive by item 3 in
Theorem 1 we have µ (An ) + µ (Bn ) = µ (An ∪ Bn ) + µ (An ∩ Bn ) taking limit when n → ∞, by
(1.2.15d) and Theorem 2 (a) we obtain
µ∗ (A) + µ∗ (B) = µ∗ (A ∪ B) + µ∗ (A ∩ B) .
Since A ∩ B = ∅, implies µ∗ (A ∩ B) = 0, then if A ∩ B = ∅ we have
µ∗ (A) + µ∗ (B) = µ∗ (A ∪ B) .
And µ∗ is additive on MF (µ).
Now let
∞
[
A= An
n=1
with An ∈ MF (µ). for n = 1, 2, 3, . . . , Then Bn ∈ MF (µ) because Bn is a finite union of elements
∞
S
of MF (µ) , B1 ⊂ B1 ⊂ · · · ⊂ A, and A = Bn , if C1 = B1 and Cn = Bn − Bn−1 for n = 2, 3, · · · ,
n=1
then Cn ∈ MF (µ) because Cn is the difference of elements of MF (µ) , and as we see in the proof
∞
S
of Theorem 2 Ci ∩ Cj = ∅ for i 6= j and A = Ck .
k=1
By (1.2.8)
∞
X
(1.2.16) µ∗ (A) ≤ µ∗ (Ck ) .
k=1
n
Ck ⊂ A. the additivity of µ∗ implies
S
On the other hand, with the above notation Bn =
k=1
n n
!
X [
(1.2.17) µ∗ (Ck ) = µ∗ Ck = µ∗ (Bn ) ≤ µ∗ (A)
k=1 k=1
xvi 1. LEBESGUE MEASURE AND INTEGRATION

Equations (1.2.16) and (1.2.17) imply

∞
X
(1.2.18) µ∗ (Cn ) = µ∗ (A) .
n=1
∗
Suppose now that µ (A) < ∞. Then (1.2.17) shows
n
X ∞
X
lim µ∗ (Bn ) = lim µ∗ (Ck ) = µ∗ (Ck ) = µ∗ (A) .
n→∞ n→∞
k=1 k=1

Hence Bn → A; and since Bn ∈ MF (µ), there are sequences {Enk } for k = 1, 2, 3, . . . of

elementary sets, such that
lim µ∗ (Enk ) = µ∗ (Bn ) ,
k→∞
and the double sequence {Enk : n, k ∈ N} satisfies
lim lim µ∗ (Enk ) = µ∗ (A) .
n→∞ k→∞

So A ∈ MF (µ) if A ∈ M (µ) and µ∗ (A) < ∞.

∞
To see that µ∗ is countably additive on M (µ), note that if A =
S
Cn ,where {Cn } is a sequence
n=1
of disjoint sets of M (µ). Then
If µ∗ (Ck ) = ∞ for some k ∈ N, since Ck ⊂ A, then
∞
X
∗ ∗
(1.2.19) ∞ = µ (Ck ) ≤ µ (A) ≤ µ∗ (Cn ) = ∞
n=1

therefore the inequalities in (1.2.19) are all equalities.

If µ∗ (Cn ) < ∞ for all n ∈ N, above we see that An ∈ MF (µ) and by (1.2.18) we have
∞
X
µ∗ (A) = µ∗ (Cn ) .
n=1
∞
S
Finally, if A = An , with An ∈ M (µ) , then for each n there exists{Bnk } for k = 1, 2, 3, . . .
n=1
∞
S
with Bnk ∈ MF (µ) and An = Bnk , so
k=1
∞
[ ∞ [
[ ∞
A= An = Bnk ,
n=1 n=1 k=1

Note that A is a countable union of elements in MF (µ) , then A ∈ M (µ).

Now suppose that A, B ∈ M (µ) and
∞
[ ∞
[
A= An , B= Bm ,
n=1 m=1

where An , Bn ∈ MF (µ) for all n ∈ N. Note that

∞
[ ∞
[
An ∩ B = An ∩ Bm = (An ∩ Bm ) .
m=1 m=1
 1.2. CONSTRUCTION OF THE LEBESGUE MEASURE xvii

So An ∩ Bm ∈ MF (µ) , and An ∩ B ∈ M (µ) for all n ∈ N. Since An ∩ B ⊂ An , then

µ∗ (An ∩ B) ≤ µ∗ (An ) < ∞.
Thus, An ∩ B ∈ MF (µ) , so An − B = An − (An ∩ B) ∈ MF (µ) for all n ∈ N. Hence
∞ ∞
!
[ [
A−B = An − B = (An − B) ∈ M (µ)
n=1 n=1

We now replace µ∗ (A) by µ(a), if A ∈ M (µ). So µ, initially defined on E, is extended to a

countably additive set function on the σ-ring M (µ). This extended set function is called a measure.
The case µ = m is called the Lebesgue measure on Rp .
Remark 7.

(a) If A is open, then A ∈ M (µ). Because every open set in Rp is the union of a countable
collection of intervals. To see this, using the density of Q in R, we can see that β =
{I = (a1 , b1 ) × (a2 , b2 ) × · · · × (ap , bp ) : ai , bi ∈ Q} is a countable base whose elements are
open intervals. Since Rp is an open set, taking complements we obtain that closed set is
in M (µ)
(b) If A ∈ M (µ) and ε > 0, there exist sets F and G that F ⊂ A ⊂ G, F is closed, G is open,
and
(1.2.20) µ (G − A) < ε, µ (A − F ) < ε.
Indeed, if µ (A) < ∞, i.e., A ∈ MF (µ) by (1.2.6), there exists a sequence {An } of
open elementary sets, so that
∞
[ ∞
X
A⊂ An and µ (An ) < µ (A) + ε,
n=1 n=1
∞
S
taking G = An , then
n=1
∞
X
µ (G − A) = µ (G) − µ (A) ≤ µ (An ) − (µ (A)) < ε.
n=1

If µ (A) = ∞ there exists a sequence {An } with An ∈ MF (µ) for all n ∈ N, so

∞
S
that A = An . Now, for each n ∈ N by show above there exists an open set Gn with
n=1
An ⊂ Gn so that
ε
µ (Gn − An ) < ,
2n
∞
S ∞
S
taking G = Gn , then, G is open, A ⊂ G, G − A = (Gn − An ), and
n=1 n=1
∞ ∞
!
[ X
µ (G − A) ≤ µ (Gn − An ) ≤ µ (Gn − An ) < ε.
n=1 n=1

For the second inequality, since M (µ) is a σ-ring, then Ac ∈ M (µ) , so there exists G
µ (G − Ac ) < ε.
xviii 1. LEBESGUE MEASURE AND INTEGRATION

Taking F = Gc , then F is closed A ⊂ F , using Remark 1 we have

µ (A − F ) = µ (A − Gc ) = µ (A ∩ G) = µ (G − Ac ) < ε
(c) We say that E is a Borel set if E can be obtained by a countable number of operations,
starting from open sets, each operation consisting in taking unions, intersections, or com-
plements. The collection B of all Borel sets in Rp is a σ-ring; in fact, it is the smallest
σ-ring which contains all open sets. By Remark (a) if B ⊂ M (µ).
(d) By (b) If A ∈ M (µ), for each n ∈ N, there exist Borel sets Fn so that that Fn ⊂ A, Fn is
closed for all n ∈ N, and
1
µ (A − Fn ) < ,
n
∞
S
If F = Fn , then F is a Borel set, F ⊂ A, A − F ⊂ A − Fn for all n ∈ N, in consequence
n=1
we have
1
µ (A − F ) ≤ µ (A − Fn ) < for all n ∈ N.
n
Thus,
(1.2.21) µ (A − F ) = 0.
Since A = F ∪ (A − F ), we see that every A ∈ M (µ) is the union of a Borel set and a set
of measure zero.
The Borel sets are always µ-measurable for all µ. But the sets of measure zero, i.e.,
the sets E for which µ∗ (E) = 0 may be different for different measures µ’s.
(e) For every µ, the sets of measure zero form a σ-ring, Indeed, recall that E has measure
zero, if for a given ε > 0, there exists a sequence {An } of open elementary sets, so that
∞
[ ∞
X
E⊂ An and µ (An ) < ε, i
n=1 n=1

Since E1 − E2 ⊂ E1 for all E1 , E2 , then

0 ≤ µ∗ (E1 − E2 ) ≤ µ∗ (E1 ) = 0
∞
En , with µ∗ (En ) = 0, for all n ∈ N. For each n ∈ N let
S
On the other hand, if E =
n=1
{Ank : k ∈ N} a sequence of open elementary sets, so that
∞ ∞
[ X ε
En ⊂ Ank and µ (Ank ) < ,
2n
k=1 k=1
∞
S ∞ S
S ∞
Thus, E = En ⊂ Ank , and
n=1 n=1 k=1
∞ X
∞ ∞
X X ε
µ (Ank ) < n
= ε,
n=1 k=1 n=1
2

(f) In case of the Lebesgue measure, every countable set has measure zero. In effect, for
each ε > 0, the interval centered on x and of volume 2ε is a covering of A = {x} by
elementary sets, with measure less than ε. As any countable set B is the countable union
of its elements, the remark 6 (e) shows that B has zero measurement. But there are
 1.2. CONSTRUCTION OF THE LEBESGUE MEASURE xix

uncountable sets of measure zero. The Cantor set P is an example: Recall that P is
defined as
∞
\
P = En ,
n=0
where E0 = [0, 1]. Removing the the middle thirds ibterval of this intervals we obtain
E1 = E0 − ( 31 , 23 ), so E1 is [0, 13 ] ∪ [ 23 , 1]. Removing the middle thirds interval of these
intervals we obtain E2 and continuing in this way, we obtain a sequence of compact sets
En , such that
E0 ⊃ E1 ⊃ E2 · · ·
2 n
and En is the union of 2 intervals, each of length 3−n . So m (En ) =
n


3 .
Below we show the E1 , E2 , E3 and E4 .

Now P can be identified with the set of the sequences (a0 a1 . . . an . . .) where an = 0 or an = 2,
and using the same argument as the one that shows that [0, 1] is uncountable, we get that P is
uncountable.
n
Since P ⊂ En for all n ∈ N, then m (P ) ≤ m (En ) = 32 , and taking limit when n → ∞ we
obtain m (P ) = 0.
Moreover, if we denote by c the cardinality of R, we obtain that cardinality of P is equal to c.
Because every subset of the set of measure 0 is the set of measure 0, we have at least 2c measurable
sets on R. Because the cardinality of the family of all subset of R is also 2c , the natural question
is: ‘are there unmeasurable sets?’. added — PS

(g) A Vitali set is a subset V of the interval [0, 1] of real numbers such that, for each real
number r, there is exactly one number v ∈ V such that v − r is a rational number. Vitali
sets are a set of representative of the group R/Q in [0, 1]. added — PS
Every Vitali set V is uncountable, and
(1.2.22) v − u is irrational for any u, v ∈ V, u 6= v.
A Vitali set is non-measurable. Indeed, assume that V is measurable and let q1 , q2 ,
. . . be an enumeration of the rational numbers in [−1, 1]. And let Vn be the translated
sets defined by
Vn = V + qn = {v + qn : v ∈ V },
Note that Vn ∩ Vm = ∅, because if y ∈ Vn ∩ Vm , then v + qn = y = u + qm , implies v − u
is rational in contradiction with (1.2.22).
∞
S
Also note that [0, 1] ⊆ Vn ⊆ [−1, 2].
n=1
xx 1. LEBESGUE MEASURE AND INTEGRATION

To see the first inclusion, consider any real number r ∈ [0, 1] and let v be the repre-
sentative in V for the equivalence class [r]; then r − v = qn for some rational number qn
∈ [−1, 1] which implies that r ∈ Vn
Since the Lebesgue measure is countably additive, then
∞
X
1≤ m(Vn ) ≤ 3.
n=1

Because the Lebesgue measure is translation invariant, we have m(Vn ) = m(V ) for all
n ∈ N, and therefore
X ∞
1≤ m(V ) ≤ 3.
n=1

But this is impossible. Summing infinitely many copies of the constant m(V ) yields either
zero or infinity, according to whether the constant is zero or positive. In neither case is the
sum in [1, 3]. So V cannot bee measurable. An adequate change of the above argument
shows that, for all measurable set A with m(A) > 0 there exists a non-measurable set B
with B ⊂ A.

1.3. MEASURE SPACES

Definition 10. Suppose X is a set, not necessarily a subset of an Euclidean space, or indeed
of any metric space, X is said to be a measure space if there exists a σ-ring M of subsets of X,
which are called measurable sets, and a non-negative countably additive set function µ which is
called a measure, defined on M.
If, in addition, X ∈ M then X is said to be a measurable space.
For example, we can take X = Rn , and M. the collection of all Lebesgue measurable subsets
of Rn , and µ = m the Lebesgue measure.
Or, let X = N the set of all positive integers, M the collection of all subsets of X, and µ (E) is
the number of elements of E. µ is know as the counting measure
Another example is provided by probability theory, where events may be considered as sets,
and the probability of the occurrence of events is an additive (or countably additive) set function.

1.4. MEASURABLE FUNCTIONS

Definition 11. Let f be a function defined on the measurable space X, with values in the
extended real number system. The function f is said to measurable if the set
(1.4.1) {x : f (x) > a}
is measurable for every real a.
Example 2. If X = Rp and M = M (µ) as in Theorem 5, every continuous f is measurable,
because then (1.4.1) is an open set, and the open sets belong to M.
Theorem 6. Let f be a function defined on the measurable space X. The followings conditions
are equivalent:
 1.4. MEASURABLE FUNCTIONS xxi

(1.4.2) {x : f (x) > a} is measurable for every real a

(1.4.3) {x : f (x) ≥ a} is measurable for every real a

(1.4.4) {x : f (x) < a} is measurable for every real a

(1.4.5) {x : f (x) ≤ a} is measurable for every real a

Let a ∈ R.
1
(1) If f (x) ≥ a, then f (x) > a − nfor all n ∈ N. So
\ 1

{x : f (x) ≥ a} ⊂ x : f (x) > a − .
n
n∈N

Reciprocally, if f (x) > a − n1 for all n ∈ N, then f (x) is an upper bound of the set
a − n : n ∈ N , since a = sup a − n1 : n ∈ N we have f (x) ≥ a and
1


\ 1

x : f (x) > a − ⊂ {x : f (x) ≥ a}.
n
n∈N

{x : f (x) > a − n1 }. Since the intersections of measurable sets is

T
So {x : f (x) ≥ a} =
n∈N
a measurable set, then 1.4.1 implies 1.4.2
(2) Note that
{x : f (x) ≥ a}c = X − {x : f (x) ≥ a} = {x : f (x) < a}
Since the complement of measurable sets is a measurable set, then (1.4.2) implies (1.4.3)
(3) If f (x) ≤ a, then f (x) < a + n1 for all n ∈ N. So
\ 1
{x : f (x) ≤ a} ⊂ {x : f (x) < a + }.
n
n∈N

Reciprocally, If f (x) < a + n1for all n ∈ N, then f (x) is lower bound of the set
1
a + n : n ∈ N , since a = inf a + n1 : n ∈ N we have f (x) ≥ a and


\ 1

x : f (x) < a + ⊂ {x : f (x) ≤ a}.
n
n∈N

{x : f (x) < a + n1 }. Since the intersections of measurable sets is

T
So {x : f (x) ≤ a} =
n∈N
a measurable set, then (1.4.3) implies (1.4.4).
(4) Note that
{x : f (x) ≤ a}c = X − {x : f (x) ≤ a} = {x : f (x) > a}
Since the complement of measurable sets is a measurable set, then (1.4.4) implies (1.4.1)
Corollary 1. Let f be a measurable function on a measurable space X, then −f is a mea-
surable function on X.
xxii 1. LEBESGUE MEASURE AND INTEGRATION

Note that
{x : −f (x) > a} = {x : f (x) < −a}
Hence {x : −f (x) > a} is measurable for every real a if f is a measurable function.
Hence any of these conditions may be used instead of (1.2.21) to define measurability
Theorem 7. Let f be a measurable function on a measurable space X, then |f | is a measurable
function on X.
If a < 0, then {x : |f (x)| > a} = X, hence measurable.
Recall that |f (x)| > a, if and only if f (x) > a or −f (x) < −a. So
{x : |f (x)| > a} = {x : f (x) > a} ∪ {x : −f (x) < −a} ,
and by Theorem 6 and corollary 1 {x : |f (x)| > a} is measurable for every real a if f is a measurable
function.
Theorem 8. Let {fn } be a sequence of measurable functions on a measurable space X. Then
g(x) = sup {fn (x) : n ∈ N}
h(x) = inf {fn (x) : n ∈ N}
k(x) = lim sup {fn (x)}
n→∞
l(x) = lim inf {fn (x)}
n→∞

are measurable functions.

T
Note g(x) ≤ a, if and only if fn (x) ≤ a for all n ∈ N. Thus, since {x : g(x) ≤ a} = {x :
n∈N
fn (x) ≤ a}, {x : g(x) ≤ a} is measurable for every real a if fn are measurable functions for all n.
If fn are measurable functions for all n, then −fn are measurable functions, so p(x) =
sup {−fn (x)} is a measurable function. Since h(x) = −p(x), then h is also a measurable function.
Note that
k(x) = inf gm (x)
l(x) = sup hm (x) ,
where hm (x) = inf {fn (x) : n > m} and gm (x) = sup {fn (x) : n > m}. Then by the show above,
gm , hm are measurable functions, in consequence k, l are also measurable functions.
Corollary 2. Let X be a measurable space, then
(1) If f , g are measurable functions on X, then
max {f, g} , min {f, g} , f +, f−
are also measurable functions.
(2) The limit of a convergent sequence of measurable functions on X is measurable on X.
(a) Recall that
max {f, g} = sup {f, g} ,
min {f, g} = inf {f, g} ,
f + = max {f, 0} ,
f − = max {−f, 0} ,
 1.4. MEASURABLE FUNCTIONS xxiii

where 0(x) = 0.
(b) If f (x) = lim fn (x), then
n→∞

f (x) = lim fn (x) = lim sup {fn (x)} .

n→∞ n→∞

Theorem 9. Let f and g be measurable real valued functions defined on X, let F be a real
valued and continuous function on R2 , and put
h(x) = F (f (x), g(x))
then h is measurable.
In particular, f + g and f · g are measurable.
Let
Ga = {(u, v) : F (u, v) > a} = F −1 (a, +∞)
Since F is continuous, then Ga is an open subset of R2 , recall that every open set in R2 is the union
of a countable collection of intervals (because both open squares and open disks form a topological
∞
basis in R2 ), so we can write Ga =
S
In , where {In } is a sequence of open intervals: added — PS
n=1

In = (an , bn ) × (cn , dn ) .
Since f and g are measurable, then the sets
{x : an < f (x) < bn } = f −1 (an , bn ) ,
{x : cn < g(x) < dn } = g −1 (cn , dn ) ,
are measurable. So
{x : (f (x), g(x)) ∈ In } = {x : an < f (x) < bn } ∩ {x : cn < g(x) < dn }
is measurable, Hence the same is true for
∞
[
{x : (f (x), g(x)) ∈ In } = {x : (f (x), g(x)) ∈ Ga } = {x : h(x) = F (f (x), g(x)) > a} .
n=1

Above we see that all ordinary operations of analysis when we apply to measurable functions,
lead to measurable functions. but the next example shows that there exits continuous function g
and a measurable function h such that h ◦ g is non-measurable.
Exercise 2. [Exercise 11.3] If {fn } is a sequence of measurable functions, prove that the set
of points at which {fn (x)} converges is measurable.
Solution 2. Let A = {x : {fn (x)} converges}, if x ∈ A then {fn (x)} is a Cauchy sequence in
R. Thus, given n ∈ N, there existsn0 ∈ N such that |fp (x) − fq (x)| < n1 if p, q > n0 .
Note that for each n ∈ N. we have
\  1
 [ \  1

x∈ z : |fp (z) − fq (z)| < ⊂ z : |fp (z) − fq (z)| <
p,q>n
n p,q>n
n
0 n0 ∈N 0

and x belong to the set

\ [ \  1

C= z : |fp (z) − fq (z)| <
n
n∈N n0 ∈N p,q>n0
xxiv 1. LEBESGUE MEASURE AND INTEGRATION

Clearly by construction if x ∈ C, then {fn (x)} is a Cauchy sequence in R, so {fn (x)} converges
and x ∈ A. in consequence A = C.
 since fn is measurable for all n, then |fp − fq | is measurable for all p, q, in consequence
Finally,
the sets z : |fp (z) − fq (z)| < n1 are measurable for all p, q and n. Since the measurable sets form
a σ-ring then A = C is a measurable set.
Example 3. Consider the following sequence {fn } of functions defined by
0, 0 ≤ x < 31




1 1 2
f0 (x) =
 2, 3 ≤x< 3

1, 23 ≤ x ≤ 1


1
0 ≤ x < 31



 2 fn−1 (3x),
1 1 2
fn (x) =
 2, 3 ≤x< 3

 1 2
2 (1 + fn−1 (3x − 2)) , 3 ≤x≤1

This sequence of functions converges to the continuous function f know as the devil’s stair-
case function of Cantor, below we can see the graph of the first three functions fn and a sketch
of the graph of f .

Consider the function g(x) = f (x) + x, where f is the devil’s staircase function of Cantor. The
function g is strictly increasing homeomorphism from [0, 1] onto [0, 2] and note that the measure
 1.5. SIMPLE FUNCTIONS xxv

of [0, 2] − g(C), where C is the Cantor set, is equal to two minus the sum of the measures of the
images of the middle thirds of the intervals in En , that we use to defined C, i.e.,
∞ 
2n
X 
m([0, 2] − g(C)) = 2 − =1
n=0
3n+1
in the figure below we can see a sketch of the graph of g,

If A is a non-measurable A ⊂ g(C). Note that B = g −1 (A) is measurable, because B ⊂ C, so

m(B) = 0. Since g is a homeomorphism, then g −1 is continuous. If h = χB , then h is measurable
on [0, 1] but
1 g −1 (x) ∈ B
 
1 x∈A
h(g −1 (x)) = = = χA ,
0 g −1 (x) ∈/B 0 x∈ /A
is the non-measurable characteristic function of the non-measurable set A.
Note that measure µ has not been mentioned in above discussion of measurable functions. In
fact, the class of measurable functions on X depends only on the σ-ring M.
For instance, we speak of Borel-measurable functions on R, that is, of function f for which
{x : f (x) < a}
is always a Borel set, without reference to any particular measure.

1.5. SIMPLE FUNCTIONS

Definition 12. Let s be a real-valued function defined on X. If the range of s is finite, we say
that s a simple function.
Let E ⊂ X, and put χE : X → R

1 x∈E
(1.5.1) χE =
0 x∈
/E
χE is called the characteristic function of E. Note that χE = χ2E = |χE | .
xxvi 1. LEBESGUE MEASURE AND INTEGRATION

Suppose the range of s consists of the distinct numbers c1 , . . ., cr . Let

Ei = {x : s(x) = ci } ,
r
S
then, X = Ei , Ei ∩ Ej = ∅ if i 6= j and
i=1
r
X
(1.5.2) s= ci χEi
i=1

Remark 8. So every simple function is a finite linear combination of characteristic functions.

It is noted that if s is measurable, then
Ei = {x : s(x) = ci } = {x : s(x) ≤ ci } ∩ {x : s(x) ≥ ci }
are measurable sets. And reciprocally, if Ei are measurable sets for all i = 1, 2, . . . , r, then
[
{x : s(x) < a} = Ei
ci <a

So s is measurable, iff Ei are measurable sets for all i = 1, 2, . . . , r.

Now we see that every function can be approximated by simple functions
Theorem 10. Let f be a real function on X. There exists a sequence {sn } of simple functions
such that
(1.5.3) lim sn (x) = f (x) for every x ∈ X
n→∞
If f is measurable, sn may be chosen to be measurable functions. If f ≥ 0, then {sn } may be chosen
to be a monotonically increasing sequence.
If f ≥ 0, for each n ∈ N define
(1.5.4) Fn = {x : f (x) ≥ n} = f −1 ([n, +∞)) ,
and for k = 1, 2, . . . , n2n define
   
k−1 k −1 k−1 k
(1.5.5) En,k = x : ≤ f (x) ≤ n = f , ,
2n 2 2n 2n
and let
n2n
X k−1
sn (x) = χEn,k + nχFn
2n
k=1
Note that Fn+1 ⊂ Fn , so if x ∈ Fn+1 , then
sn+1 (x) = n + 1 > n = sn (x).
Since
     
2k − 1 2k 2k − 1 k k−1 k
En+1,2k = f −1 , = f −1
, ⊂ f −1
, = En,k
2n+1 2n+1 2n+1 2n 2n 2n
     
2k 2k + 1 k 2k + 1 k k+1
En+1,2k+1 = f −1 , = f −1
, ⊂ f −1
, = En,k+1
2n+1 2n+1 2n 2n+1 2n 2n
so if x ∈ En+1,2k , then
2k − 1 2k − 2 k−1
sn+1 (x) = n+1
> n+1 = = sn (x),
2 2 2n
 1.6. INTEGRATION xxvii

and if x ∈ En+1,2k+1 , then

2k k
sn+1 (x) = = = sn (x).
2n+1 2n
So {sn } is a monotonically increasing sequence.
On the other hand, let x ∈ X, since f (x) ∈ R, then there existsm ∈ N so that f (x) < m, and
let n, k ∈ N so that n > m and k−1 k n
2n < f (x) < 2n , for some k ∈ {1, 2, . . . , n2 }. Then, x ∈ En,k and
by definition of sn (x) we have
1
(1.5.6) 0 ≤ f (x) − sn (x) <
2n
taking limit when n→ ∞, we obtain
lim sn (x) = f (x)
n→∞

In the general case, consider f = f + − f − and apply the preceding construction to f + and f − .
Note that if f is measurable, then Fn , En,k are measurable sets, so sn are measurable functions.
Note that if f ≥ 0, then functions sn in the proof satisfy sn ≥ 0, and if f is bounded, i.e.,
|f (x)| < M for all x ∈ X and some M ∈ R, then equation (1.5.6) tells us that the sequence {sn }
converges uniformly to f .

1.6. INTEGRATION
Now we define integration on a measurable space X, in which M is the σ-ring of measurable
sets, and µ is the measure.
Definition 13. Suppose s ≥ 0
n
X
s(x) = ck χEk (x) x ∈ X
k=1

is measurable simple function, and suppose E ∈ M, we define the integral of s on E by

n
X
(1.6.1) IE (s) = ck µ (E ∩ Ek )
k=1

In the above definition we need s ≥ 0, i.e., ck ≥ 0 for k = 1, 2, . . . , n, to avoid indefinitions in

the sum of the form +∞ − ∞, For example, if µ = m is the Lebesgue measure on R, and we do not
have this restrictions, consider s(x) = 1 · χ(−∞,0) + (−1) · χ[0,+∞) , then IR (s) is not defined.
Now we see some properties of this integral
Theorem 11. Let s ≥ 0 be a measurable simple function on X, and suppose E ∈ M, with
∞
[
E= Fj
j=1

and Fj ∈ M are pairwise disjoint. Then

∞
X
(1.6.2) IE (s) = IFj (s)
j=1
xxviii 1. LEBESGUE MEASURE AND INTEGRATION

Pn
Suppose that s(x) = k=1 ck χEk (x) x ∈ X, then
 
∞
n n
!
X X [
IE (s) = ck µ (E ∩ Ek ) = ck µ  Fj ∩ Ek 
k=1 k=1 j=1
n
X ∞
[  Xn ∞
X
= ck µ (Fj ∩ Ek ) = ck µ (Fj ∩ Ek )
k=1 j=1 k=1 j=1
∞ ∞
n X  X n
!
X X
= ck µ (Fj ∩ Ek ) = ck µ (Fj ∩ Ek )
k=1 j=1 j=1 k=1
∞
X
= IFj (s) .
j=1

Remark 9. The above theorem tells us that if φs (E) = IE (s) for each E ∈ M, then φs is
countably additive.
Theorem 12. Let s1 , s2 ≥ 0 be measurable simple functions on X,
(1) if s1 ≤ s2 , then for each E ∈ M we have
(1.6.3) IE (s1 ) ≤ IE (s2 )
(2) s1 + s2 ≥ 0 is a measurable simple function on X, and for each E ∈ M we have
(1.6.4) IE (s1 + s2 ) = IE (s1 ) + IE (s2 )
Suppose that
n
X k
X
s1 (x) = ck χEk , s2 (x) = dj χFj
k=1 j=1

where ck ≥ 0, dj ≥ 0.
Sn k
S
Since X = Ei = Fj , then for 1 ≤ i ≤ n, 1 ≤ j ≤ k, let Aij = Fj ∩ Ei ∩ E
i=1 j=1

k
[ n
[
Ei ∩ E = Aij , Fj ∩ E = Aij
j=1 i=1
n
[ n [
[ k
E= (Ei ∩ E) = Aij
i=1 i=1 j=1

(1) Note that s1 (x) = ci ≤ dj = s2 (x) for x ∈ Aij . So

IAij (s1 ) = ci µ (Aij ) ≤ dj µ (Aij ) = IAij (s2 )
By theorem (11)
n X
X k n X
X k
IE (s1 ) = IAij (s1 ) ≤ IAij (s2 ) = IE (s2 )
i=1 j=1 i=1 j=1
 1.6. INTEGRATION xxix

(2) Note that s1 (x) + s2 (x) = ci + dj for x ∈ Aij . So

IAij (s1 + s2 ) = (ci + dj ) µ (Aij ) = ci µ (Aij ) + dj µ (Aij ) = IAij (s1 ) + IAij (s2 ) ;
By theorem 11
n X
X k n X
X k


IE (s1 + s2 ) = IAij (s1 + s2 ) = IAij (s1 ) + IAij (s2 ) = IE (s1 ) + IE (s2 ) .
i=1 j=1 i=1 j=1

Note that if 0 ≤ s1 ≤ s2 are measurable simple functions and Aij are as the above proof, then
m
X
s2 − s1 = (dj − ci ) χAij ≥ 0
j=1

and
(1.6.5) IE (s2 − s1 ) = IE (s2 ) − IE (s1 )
Definition 14. If f is measurable and non-negative, we define
Z
(1.6.6) f dµ = sup IE (s) ,
E
where the supremum is taken over all measurable simple functions s such that 0 ≤ s ≤ f .
The left member of (1.6.6) is called the Lebesgue integral of f , with respect to the measure
µ, over the set E. Note that the integral may have the value +∞.
Note that if 0 ≤ s is a measurable simple function, then s ≤ s, and by Theorem 12 (1)
IE (s1 ) ≤ IE (s) for all measurable simple functions s1 such that 0 ≤ s1 ≤ s so
Z
(1.6.7) sdµ = IE (s) .
E

Definition 15. Let f be measurable, and consider the two integrals

Z Z
(1.6.8) +
f dµ, f − dµ,
E E

where f = f + − f − and f + and f − are defined as Corollary 2. If at least one of the integrals (1.6.8)
is finite, we define
Z Z Z
(1.6.9) f dµ = f + dµ − f − dµ.
E E E
If both integrals in (1.6.8) are finite, then the integral in (1.6.9) is finite, and we say that f is
integrable (or summable) on E in the Lebesgue sense, with respect to µ and we write f ∈ L (µ)
on E. If µ = m the Lebesgue measure, we say that f is Lebesgue integrable on E and that
f ∈ L on E.
Note that if the integral in (1.6.9) is +∞ or −∞, then the integral of f over E is defined,
although f is not integrable on E in the above sense; f is integrable on E only if its integral over
E is finite.
All properties in Remark 11.23 of Rudin’s books are true, but not are evidently as he say, here
we only show those we need, and lead to later the rest
Remark 10. The integral satisfies the following properties:
 xxx 1. LEBESGUE MEASURE AND INTEGRATION

(a) If f ∈ L (µ) on E, and λ ∈ R, then

Z Z
(1.6.10) λf dµ = λ f dµ
E E
Pn
We first see the case 0 ≤ s = k=1 ck χEk
R
1. If λ ≥ 0, then 0 ≤ λs is a measurable simple functionR and E λs dµ = IE (λs) =
Pn Pn
k=1 λck µ (E ∩ Ek ) = λ k=1 ck µ (E ∩ Ek ) = λIE (s) = λ E s dµ
−
2. If λ < 0, then λs ≤ 0, so (λs) = (−λ) s
Z Z Z Z Z
−
λs dµ = − (λs) dµ = − (−λ) s dµ = − (−λ) sdµ = λ s dµ
E E E E E

In the case 0 ≤ f ;
1. If λ ≥ 0, then s1 is a simple function with 0 ≤ s1 ≤ λf , if and only if s1 = λs where s is
a simple function with 0 ≤ s ≤ f . So
Z Z   Z  Z  Z
λf dµ = sup {IE (λs)} = sup λs dµ = sup λ s dµ = λ sup s dµ = λ f dµ
E E E E E
and consequently f dµ
to the end −
2. If λ < 0, then λf ≤ 0, so (λf ) = (−λ) f
Z Z Z Z Z
−
λf dµ = − (λf ) dµ = − (−λ) f dµ = − (−λ) f dµ = λ f dµ
E E E E E

In the general case;

+ −
1. If λ ≥ 0, then (λf ) = λf + , (λf ) = λf − and
Z Z Z Z Z Z Z Z Z  Z
+ − + − + − + −
λf dµ = (λf ) − (λf ) = λf − λf = λ f −λ f =λ f − f =λ f dµ
E E E E E E E E E E
+ −
2. If λ < 0, then (λf ) = (−λ) f − , (λf ) = (−λ) f +
Z Z Z Z Z Z Z
+ −
λf dµ = (λf ) − (λf ) = (−λ) f − − (−λ) f + = (−λ) f − − (−λ) f+
E E E E E E E
Z Z  Z
+ −
=λ f − f =λ f dµ
E E E

(b) If f is measurable and bounded on E, and if µ (E) < ∞, then f ∈ L (µ) on E.

Since f is bounded,
Pn then there existsP M , so that |f (x)| ≤ M , f + (x) ≤ M and f − (x) ≤ M.
m +
If s1 = k=1 ck χEk and s2 = k=1 dk χFk are simple functions so that 0 ≤ s1 ≤ f and
−
0 ≤ s2 ≤ f , then 0 ≤ ck , dk ≤ M and
Z Xn X n
s1 dµ = ck µ (E ∩ Ek ) ≤ M · µ (E ∩ Ek ) ≤ M · µ (E)
E k=1 k=1
Z Xm m
X
s2 dµ = dk µ (E ∩ Fk ) ≤ M · µ (E ∩ Fk ) ≤ M · µ (E)
E k=1 k=1
 1.6. INTEGRATION xxxi

So
Z
f + dµ ≤ M · µ (E) < ∞
E
Z
f + dµ ≤ M · µ (E) < ∞
E

and f ∈ L (µ) on E.
(c) If f, g ∈ L (µ) , and 0 ≤ f (x) ≤ g(x), then
Z Z
(1.6.11) f dµ ≤ gdµ.
E E

Note that 0 ≤ f (x) ≤ g(x) implies that

{s : s is a simple function, 0 ≤ s ≤ f } ⊂ {s : s is a simple function, 0 ≤ s ≤ g} .
So Z Z
f dµ = sup IE (s) ≤ sup IE (s) = f dµ.
E 0≤s≤f 0≤s≤g E
(d) If µ (E) = 0, and f is measurable, then
Z
(1.6.12) f dµ = 0.
E
Pn
Note that E ∩ A ⊂ E implies µ (E ∩ A) = 0,So if 0 ≤ f, and s = k=1 ck χEk is a simple
function, 0 ≤ s ≤ f on E, then
n
X n
X
IE (s) = ck µ (E ∩ Ek ) = ck · 0 = 0.
k=1 k=1

Thus, Z
f dµ = sup IE (s) = 0.
E 0≤s≤f

f + dµ = E f − dµ = 0, so
R R
In the general case, we have E
Z Z Z
f dµ = f + dµ − f − dµ = 0.
E E E

(e) If f ∈ L (µ) on E, and A ∈ M, A ⊂ E, then f ∈ L (µ) on A.

Pn
if 0 ≤ f, and s = k=1 ck χEk is a simple function, 0 ≤ s ≤ f , since A ⊂ E, then A ∩ Ek ⊂ E ∩ Ek
and µ (A ∩ Ek ) ≤ µ (E ∩ Ek ). Thus,
n
X n
X
IA (s) = ck µ (A ∩ Ek ) ≤ ck µ (A ∩ Ek ) = IE (s) .
k=1 k=1

and taking supremun we have,

Z Z
f dµ = sup IA (s) ≤ sup IA (s) = f dµ.
A 0≤s≤f 0≤s≤f E
−
+ +
f − dµ < ∞, so f ∈ L (µ)
R R R R
In the general case, we have A
f dµ ≤ E
f dµ < ∞, A
f dµ ≤ E
on A.
 xxxii 1. LEBESGUE MEASURE AND INTEGRATION

Theorem 13 (Lebesgue’s monotone convergence theorem). Suppose E ∈ M. Let {fn } be a

sequence of real, measurable functions such that
(1.6.13) 0 ≤ f1 (x) ≤ f2 (x) ≤ . . .
for all x ∈ E. Let f be defined by
(1.6.14) f (x) = lim fn (x),
n→∞

then
Z Z
(1.6.15) f dµ = lim fn dµ
E n→∞ E

First note that the sequence {fn (x)} is an increasing sequence of real functions so
f (x) = lim fn (x) = sup {fn (x)}
n→∞ n∈N

and fRis a measurable non-negative function with fn (x) ≤ f (x) for all x ∈ E and n ∈ N. Also we
end of the proof not here have E fn (x)dµ is an increasing sequence of real extended numbers, so
Z Z 
lim fn dµ = sup fn dµ = α
n→∞ E n∈N E

for some α, and since fn (x) ≤ f (x) for all x ∈ E and n ∈ N, we have
Z
(1.6.16) α≤ f dµ.
E
Choose c such that 0 < c < 1, and let s be a simple measurable function such that 0 ≤ s ≤ f
and let
En = {x : fn (x) ≥ cs (x)} ,
for n ∈ N. Since
By (1.6.13), E1 ⊂ E2 ⊂ · · · ⊂ En ⊂ · · · . If x ∈ E, since 0 < c < 1 then
cs (x) < f (x) = sup {fn (x)} ,
n∈N

so there existsn ∈ N, so that cs (x) ≤ fn (x). Thus,

∞
[
(1.6.17) E= En .
n=1

Remark 10 (c) implies that for every n,

Z Z Z Z
(1.6.18) fn dµ ≥ fn dµ ≥ csdµ = c sdµ = cIEn (s) = cφs (En )
E En En En

Since the φs is a countably additive set function (see Remark 9) by (1.6.17 ) we can apply
Theorem 11 and Remark 10 (a) to φs in (1.6.18), and taking limit when n → ∞ in(1.6.18), we
obtain
Z Z
(1.6.19) α = lim fn dµ ≥ lim cφs (En ) = cφs (E) = c sdµ
n→∞ E n→∞ E
 1.6. INTEGRATION xxxiii

Taking limit when c → 1, we see that

Z
(1.6.20) α≥ sdµ
E
so
Z Z
(1.6.21) α ≥ sup sdµ = f dµ
0≤s≤f E E

Inequalities (1.6.16) and (1.6.21) and definition of α imply

Z Z
f dµ = α = lim fn dµ.
E n→∞ E

Theorem 14. (a) Suppose f is measurable and non-negative on X. For A ∈ M, define

Z
(1.6.22) φf (A) = f dµ
A
then φf is countably additive.
(b) The same conclusion holds if f ∈ L (µ) on X.
Note that if f ∈ L (µ) on X, then φf + (A) = A f + dµ and φf − (A) = A f − dµ are both finite,
R R

then write
Z Z Z
(1.6.23) φf (A) = f dµ = f + dµ − f − dµ = φf + (A) − φf − (A)
A A A
and apply (a) to φf + and φf − we obtain (b).
To prove (a), by Theorem (10) there exists a sequence {sn } of simple functions such that
(1.6.24) 0 ≤ s1 (x) ≤ s2 (x) ≤ ... ≤ f
for all x ∈ X. So that
(1.6.25) f = lim sn = sup {sn : n = 1, 2, 3, . . .}
n→∞
Remark 10 (c) shows that φsn ≤ φsn+1 for n = 1, 2, 3, . . . . So {φsn : n = 1, 2, 3, . . .} is a
sequence of countably additive functions on M. By Lebesgue’s monotone convergence theorem
Z Z Z 
(1.6.26) φf (A) = f dµ = lim sn = sup sn dµ = sup {φsn (A)} ,
A n→∞ A n∈N A n∈N
and by Theorem 3, φf is a countably additive function on M.
Corollary 3. If A, B ∈ M, B ⊂ A and µ (B − A) = 0, then
Z Z
(1.6.27) f dµ = f dµ
A B

R Since B ⊂ A, then A = B ∪ (A − B), B ∩ (A − B) = ∅. The condition µ (B − A) = 0 implies

A−B
f dµ = 0 (see Remark 10 (d)), so
Z Z Z Z
(1.6.28) f dµ = φf (A) = φf (B) + φf (A − B) = f dµ + f dµ = f dµ.
A B A−B B

Remark 11. The preceding corollary shows that sets of measure zero are negligible in integra-
tion
 xxxiv 1. LEBESGUE MEASURE AND INTEGRATION

Let us write f ∼ g on E if the set {x : f (x) 6= g(x)} ∩ E has measure zero.

Clearly f ∼ f .and f ∼ g implies g ∼ f ; and note that
(1.6.29) y ∈ {x : f (x) = g(x)} ∩ {x : g(x) = h(x)}
then
y ∈ {x : f (x) = h(x)}
so
c
{x : f (x) 6= h(x)} ∩ E = {x : f (x) = h(x)} ∩ E
c c
⊂ ({x : f (x) = g(x)} ∪ {x : g(x) = h(x)} ) ∩ E
= ({x : f (x) 6= g(x)} ∩ E) ∪ ({x : g(x) 6= h(x)} ∩ E)

Thus, µ ({x : f (x) 6= h(x)} ∩ E) = 0 if µ ({x : f (x) 6= g(x)} ∩ E) = 0 and µ ({x : g(x) 6= h(x)} ∩ E) =
0. Hence f ∼ g and g ∼ h implies f ∼ h. In conclusion, ∼ is an equivalence relation on E
If f ∼ g on E, then f ∼ g on A, for all A ⊂ E. let, B = {x : f (x) = h(x)} ∩ A and
C = {x : f (x) 6= h(x)} , then A = B ∪ C, B ∩ C = ∅. So if f and g are integrable on A then
Z Z Z Z
(1.6.30) f dµ = f dµ = gdµ = gdµ.
A B B A

If a property p holds for x ∈ E − A and if µ(A) = 0, we say that p holds for almost all x ∈ E,
or that p holds almost everywhere on E.
n
Note that if µ (A) > 0, where A = {x : f (x) = +∞} , then sn (x) = µ(A) χA ≤ f (x), so
Z Z
(1.6.31) n= sn dµ ≤ f dµ.
A A
R
Thus, A f dµ = +∞ and f ∈ / L(µ) on E, in consequence if f ∈ L(µ) on E, then f (x) must be finite
almost everywhere on E. In most cases we therefore do not lose generality if we assume the given
functions to be finite-valued from the outset.
R
Exercise 3. Let f be a measurable function on E. Then, if f ≥ 0 and E f dµ = 0, prove that
= 0 almost everywhere on E. Hint: Let En be the subset of E on which f (x) > n1 . Write
f (x) S
∞
A = n=1 En . Then µ(A) = 0 if and only if µ(En ) = 0 for every n.
Solution 3. Note that for every n, En ⊂ A , so if µ(A) = 0, then µ(En ) = 0 also. Conversely,
Not labelled if µ(En ) = 0 for every n, by Corollary ?? we have
∞
[ ∞
X
µ(A) = µ( En ) ≤ µ(En ) = 0
n=1 n=1

On the other hand, if µ(En ) > 0 for any n, since f ≥ n1 χEn on E,then
Z Z
1 1 1
0= f dµ ≥ χEn dµ = µ(E ∩ En ) = µ(En ) > 0
E E n n n
which is a clearly contradiction.
R
Exercise 4. [Exercise 11.2] Let f be a measurable function on E. Then, if A f dµ = 0 for
every measurable subset A of a measurable set E, then f (x) = 0 almost everywhere on E.
 1.6. INTEGRATION xxxv

Solution 4. Consider A+ = {x ∈ E : f (x) ≥ 0} and A− = {x ∈ E : f (x) < 0} , since f is

measurable, then A+ and A− are measurable sets and

A+ ∩ A− = ∅, E = A+ ∪ A−

Note that
f + (x) x ∈ A+

f (x) = ,
−f − (x) x ∈ A−

f (x) x ∈ A+
f + (x) =
0 x ∈ A−

0 x ∈ A+
f − (x) =
−f (x) x ∈ A−

If A is a measurable subset of E, then

A = (A+ ∩ A) ∪ (A− ∩ A)

Since A+ ∩ A and A− ∩ A are measurable sets of E, we have

Z Z
0= f dµ = f + dµ
A+ ∩A A+ ∩A
Z Z Z
0= f dµ = −f − dµ = − f − dµ
A− ∩A A− ∩A A− ∩A

Since
Z Z Z Z Z
+ + + +
f dµ = f dµ + f dµ = f dµ + 0dµ = 0 + 0 = 0
A A+ ∩A A− ∩A A+ ∩A A− ∩A
Z Z Z Z Z
f − dµ = f − dµ + f − dµ = 0dµ + f − dµ = 0 + 0 = 0,
A A+ ∩A A− ∩A A+ ∩A A− ∩A

Since f + and f − are non-negative measurable functions, then exercise 11.1 implies that f + (x) =
0 and f − (x) = 0 almost everywhere on E, so

f (x) = f + (x) − f − (x) = 0

almost everywhere on E.

Theorem 15. If f ∈ L(µ) on E, then |f | ∈ L(µ) on E and

Z Z
(1.6.32) f dµ ≤ |f | dµ
E E

Write E = A ∪ B, where A = {x ∈ E : f (x) ≥ 0} and B = {x ∈ E : f (x) < 0}, then A and B

are measurable and A ∩ B = ∅, and note that
(1) If x ∈ A, then |f | (x) = f (x) = f + (x), f − (x) = 0.
(2) If x ∈ B, then |f | (x) = −f (x) = f − (x), f + (x) = 0.
xxxvi 1. LEBESGUE MEASURE AND INTEGRATION

By Theorem 14 we have
Z Z Z Z
(1.6.33) f + dµ = φf + (E) = φf + (A) + φf + (B) = f + dµ + f + dµ = f + dµ
ZE ZA ZB ZA
− − −
(1.6.34) f dµ = φf − (E) = φf − (A) + φf − (B) = f dµ + f dµ = f . dµ
E A B B
and,
Z
|f | dµ = φ|f | (E) = φ|f | (A) + φ|f | (B)
E
Z Z Z Z
= |f | dµ + |f | dµ = +
f dµ + f − dµ
A B A B
Z Z
= f + dµ + f − dµ < ∞.
E E

Since 0 ≤ f + , and 0 ≤ f − we have 0 ≤ E f + dµ, and 0 ≤ E f − dµ. So

R R
Z Z Z Z Z
−
f dµ = +
f dµ − f dµ ≤ +
f dµ + f − dµ
E E E E E
Z Z Z
+ −
= f dµ + f dµ = |f | dµ.
E E E

Since the integrability of f implies that |f | s integrable, the Lebesgue integral is called an
absolutely convergent integral.
Theorem 16. Let f be a measurable function on E and g ∈ L (µ) on E. If |f | ≤ g, then
f ∈ L (µ) on E.
Note that 0 ≤ f + , f − ≤ |f | ≤ g so
Z Z
+
0≤ f dµ ≤ gdµ < ∞
ZE ZE
0≤ f − dµ ≤ gdµ < ∞
E E

and f ∈ L (µ) on E.
Exercise 5. [Exercise 11.4] If f ∈ L(µ) on E and g is bounded and measurable on E, then
f g ∈ L(µ) on E.
Solution 5. Since g is bounded, let M a positive constant such that |g (x)| ≤ M for all x ∈ E.
Since f ∈ L(µ) on E, by Remark 10 (a) and Theorem ?? we have M |f | ∈ L(µ) on E. Since f and
g are measurable, f g is also measurable. Now
|f g| (x) = |g (x)| |f (x)| ≤ M |f (x)| = M |f | (x)
so by Theorem ?? we have f g ∈ L(µ) on E.
Theorem 17. If f , g ∈ L (µ) on E, then
Z Z Z
(1.6.35) (f + g) dµ = f dµ + gdµ
E E E
 1.6. INTEGRATION xxxvii

In the case 0 ≤ f , g let {sn }, {pn } be sequences of simple functions such that
0 ≤ s1 (x) ≤ s2 (x) ≤ . . . ≤ f,
0 ≤ p1 (x) ≤ p2 (x) ≤ . . . ≤ g,
for all x ∈ E. and
(1.6.36) f (x) = lim sn (x), g(x) = lim pn (x) x ∈ E.
n→∞ n→∞

Then
0 ≤ s1 (x) + p1 (x) ≤ s2 (x) + p2 (x) ≤ . . . ≤ f + g,
for all x ∈ E and
(1.6.37) (f + g) (x) = lim (sn + pn ) (x). x ∈ E.
n→∞

By Lebesgue’s monotone convergence theorem and item 2 in Theorem 12 we have

Z Z Z Z 
(f + g) dµ = lim (sn + pn ) dµ = lim sn dµ + pn dµ
E n→∞ E n→∞ E
Z Z Z ZE
= lim sn dµ + lim pn dµ = f dµ + gdµ
n→∞ E n→∞ E E E

In the general case, since f, g ∈ L (µ) we can suppose that f (x) and g(x) are finite for all x ∈ E.
Let h = f + g, then
h+ − h− = h = f + g = f + − f − + g + − g −
or equivalently
(1.6.38) h+ + f − + g − = f + + g + + h−
Since both sides in equation (1.6.38) are the sum of three positive functions of L (µ) we conclude
Z Z Z Z
− −
+
h+ + f − + g − dµ


h dµ + f dµ + g dµ =
E E E
ZE
f + + g + + h− dµ


=
ZE Z Z
= f + dµ + g + dµ + h− dµ
E E E

Thus,
Z Z Z Z Z Z Z Z
(f + g) dµ = hdµ = h+ dµ − h− dµ = f + dµ − f − dµ + g + dµ − g − dµ
E E E E E E E E
Z Z
= f dµ + gdµ.
E E

Theorem 18 (Absolute Continuity). Let E be a measurable set and f ∈ L(µ) on E. Then,

given any ε > 0, there is a δ > 0 such that for any A ⊂ E measurable with µ (A) < δ, we have
Z
φ|f | (A) = |f | dµ < ε.
A
xxxviii 1. LEBESGUE MEASURE AND INTEGRATION

ε
If f is bounded, i.e., |f (x)| ≤ M for some real positive M and all x ∈ E, then if µ (A) < δ = M,
since µ (A) < µ (E) we have
Z Z
φ|f | (A) = |f | dµ < M dµ = µ (A) M < ε.
A A
Otherwise, define 
|f (x)| f (x) ≤ n
fn (x) =
n f (x) > n
Then {fn } is an increasing sequence of bounded, measurable functions with f = lim fn on E. By
n→∞
Lebesgue’s monotone convergence theorem, we can find N ∈ N so that
Z Z
ε
fN dµ > |f | dµ −
E E 2
and hence Z Z Z
ε
> |f | dµ − fN dµ = (|f | − fN ) dµ.
2 E E E
ε
If 0 < δ < 2N and µ (A) < δ, since 0 < fN (x) < N for all x ∈ E, then
Z Z Z Z
|f | dµ = (|f | − fn ) + fn dµ = (|f | − fn ) + fn dµ
A
ZA Z A A

≤ (|f | − fn ) + N dµ
E A
ε ε ε
< + N µ (A) < + = ε,
2 2 2
and the proof is complete.
Theorem 19. Suppose E ∈ M. Let {fn } be a sequence of non-negative measurable functions
on E. Let f be defined by
X∞
(1.6.39) f (x) = fk (x),
k=1
then
Z ∞ Z
X
(1.6.40) f dµ = fk dµ
E k=1 E
Pn
Note that the sequence {gn = k=1 fk } of the partial sums of (1.6.39) form a monotonically
increasing sequence such that
f = lim gn ,
n→∞
and by Theorem 17
Z n Z
X
(1.6.41) gn dµ = fk dµ
E k=1 E

By Lebesgue’s monotone convergence theorem we have

Z Z Xn Z ∞ Z
X
f dµ = lim gn dµ = lim fk dµ = fk dµ.
E n→∞ E n→∞ E E
k=1 k=1
 1.6. INTEGRATION xxxix

Theorem 20 (Fatou’s Lemma). Suppose E ∈ M. Let {fn } be a sequence of non-negative

measurable functions on E. Let f defined by
(1.6.42) f (x) = lim inf fn (x),
n→∞
then
Z Z
(1.6.43) f dµ ≤ lim inf fn dµ.
E n→∞ E

Consider the sequence {gn = inf k≥n fk } , by Theorem 8 the functions gn are measurable func-
tions on E. On the other hand: or comma

(1.6.44) 0 ≤ g1 (x) ≤ g2 (x) ≤ . . .

(1.6.45) gn ≤ fn
(1.6.46) f = sup gn = lim gn ,
n∈N n→∞

Inequality (1.6.45) and Remark 10 (c) imply

Z Z
gn dµ ≤ fn dµ
E E
So
Z Z
(1.6.47) lim inf gn dµ ≤ lim inf fn dµ
n→∞ E n→∞ E
Equations (1.6.44) and (1.6.46) imply that {gn } satisfies the hypothesis of Lebesgue’s monotone
convergence theorem, so using ( 1.6.47) we have
Z Z Z Z
(1.6.48) f dµ = lim gn dµ = lim inf gn dµ ≤ lim inf fn dµ.
E n→∞ E n→∞ E n→∞ E

Exercise 6. [Exercise 11.5] Put

0 ≤ x < 21

0
g (x) = 1 ,
1 2 ≤x≤1
f2k (x) = g(x),
f2k+1 (x) = g(1 − x).
Show that
lim inf fn (x) = 0,
n→∞
but Z 1
1
fn (x)dm =
0 2
[Compare with (1.6.43).]
Solution 6. Since for each x with 0 ≤ x < 12 , we have f2k (x) = g(x) = 0 for all k, then for
these x, by the definition of the inferior limit we have lim inf fn (x) = 0.
n→∞
Analogously, for each x with 21 ≤ x ≤ 1, we have f2k+1 (x) = g(1 − x) = 0 for all k, thus
lim inf fn (x) = 0 for 21 ≤ x ≤ 1.
So lim inf fn (x) = 0 for all 0 ≤ x ≤ 1.
n→∞
 xl 1. LEBESGUE MEASURE AND INTEGRATION

Note that g(x) = χ[ 1 ,1] (x), so

2

f2k (x) = g(x) = χ[ 1 ,1] (x)

2

f2k+1 (x) = g(1 − x) = χ[0, 1 ) (x)

2

Thus,
R1 R1
f (x)dm = 0 χ[ 1 ,1] (x)dm = m 12 , 1 = 21
Z 1 (  

0 2k
n = 2k
fn (x)dm = R1 R1 2
,
(x)dm = 0 χ[0, 1 ) (x)dm = m 0, 12 = 21



0 f
0 2k+1
n = 2k + 1
2

This exercise show that we can obtain strict inequality in equation (1.6.43 ) in Fatou’s Lemma.
Theorem 21 (Lebesgue dominated convergence theorem). Suppose E ∈ M. Let {fn } be a
sequence of measurable functions on E and f a measurable function on E such that
(1.6.49) f = lim fn in E
n→∞

If there exists a function g ∈ L (µ) on E, such that

(1.6.50) |fn (x)| ≤ g (x)
too early end of the the- for all n ∈ N and x ∈ E, then f ∈ L (µ) on E and
orem Z Z
(1.6.51) f dµ = lim fn dµ
E n→∞ E

Note that (1.6.50) implies

|f (x)| = lim |fn (x)| ≤ f (x) in E
n→∞

and by Theorem 16 we have f , fn ∈ L (µ) on E, (1.6.50) also implies

(1.6.52) −fn (x) ≤ g (x)
(1.6.53) fn (x) ≤ g (x)
So the sequences {g + fn } , {g − fn } are measurable non-negative functions such that
g + f = lim (g + fn ) = lim inf (g + fn )
n→∞ n→∞
g − f = lim (g − fn ) = lim inf (g − fn )
n→∞ n→∞
By Fatou’s theorem and Theorem 17 we have
Z Z Z Z Z Z
gdµ + f dµ = (g + f ) dµ ≤ lim inf (g + fn ) dµ = gdµ + lim inf fn dµ
E E E n→∞ E E n→∞ E
Z Z Z Z Z Z
gdµ − f dµ = (g − f ) dµ ≤ lim inf (g − fn ) dµ = gdµ − lim inf fn dµ
E E E n→∞ E E n→∞ E

So
Z Z
(1.6.54) f dµ ≤ lim inf fn dµ
n→∞
ZE E
Z
(1.6.55) − f dµ ≤ −lim inf fn dµ
E n→∞ E
 1.6. INTEGRATION xli

Note that (1.6.55) can be rewritten as

Z Z
(1.6.56) lim sup fn dµ ≤ f dµ
n→∞ E E

Thus,
Z Z Z Z
(1.6.57) lim sup fn dµ ≤ f dµ ≤ lim inf fn dµ ≤ lim sup fn dµ
n→∞ E E n→∞ E n→∞ E

So all inequalities in equation (1.6.57) are equalities which implies that the limit in (1.6.51)
exist and the equality in (1.6.51) is holds.
Corollary 4. If µ (E) < ∞, and {fn } is uniformly bounded on E, and f (x) = lim fn (x)
n→∞
on E, then
Z Z
(1.6.58) f dµ = lim fn dµ.
E n→∞ E

Since {fn } is uniformly bounded

R on E, then there exists M such that |fn (x)| ≤ M = M χE (x)
for all n ∈ N and x ∈ E. note that E M χE (x) dµ = M µ (E) < ∞, So M χE (x) ∈ L (µ) on E, and
we can apply the Lebesgue’s dominated convergence theorem with g = M χE .
A uniformly bounded convergent sequence is said to be boundedly convergent.
Exercise 7. [Exercise 11.6] Let
1

n |x| ≤ n
fn (x) = .
0 |x| > n

Then fn (x) → 0 uniformly on R1 , but

Z ∞
fn dm = 2 (n = 1, 2, 3, . . .).
−∞
R∞ R
(We write −∞ in place of R1 ). Thus uniform convergence does not imply dominated convergence in
the sense of Lebesgue dominated convergence theorem. However, on sets of finite measure, uniformly
convergent sequences of bounded functions do satisfy the Lebesgue dominated convergence theorem.
Solution 7. Recall that {fn } converges to g uniformly on R1 if for given ε > 0 there exist a
natural number N , such that |fn (x) − g(x)| < ε for all x ∈ R1 and n > N .
In this case given ε > 0. Let N be natural number such that N1 < ε. Note that for all x ∈ R1 and
all integers n > N , we have

1 1
|fn (x) − 0| ≤ < < ε.
n N
So {fn } converges to 0 uniformly on R1 .
On the other hand, we have that fn = n1 χ[−n,n] . So for all n ∈ N
Z ∞ Z ∞
1 1 1
fn (x) dm = χ[−n,n] (x) dm = m ([−n, n]) = 2n = 2.
−∞ −∞ n n n
and the proof is complete.
 xlii 1. LEBESGUE MEASURE AND INTEGRATION

Example 4. In this example we have a sequence {fk } of continuous function that converge to
a continuous function f , but
Z b Z b
lim fk dx 6= f dx
k→∞ a a
In this case consider  1
 k2 x 0≤x≤
k




  
fk (x) = 2 1 2
k2 −x <x≤
k k k





2

0 k <x≤1
0, not o This sequence punctually converge to the function f (x) = 0 for all x ∈ [0, 1], below we can see the
graph of fk , this graph shows that the integral of fk is 1 equal to the area of the triangle in red on
the graph, so
Z b Z b
lim fk dx = 1 6= 0 = f dx.
k→∞ a a

The last example shows that the condition |fk (x)| ≤ g(x), in the dominated converge theorem
is necessary to obtain the conclusion.
Exercise 8. [Exercise 11.12] Suppose
(a) |f (x, y)| ≤ 1, for 0 ≤ x ≤ 1, 0 ≤ y ≤ 1
(b) for fixed x, f (x, y) is a continuous function of y,
(c) for fixed y, f (x, y) is a continuous function of x.
Put Z 1
g(x) = f (x, y)dy (0 ≤ x ≤ 1)
0
Is g continuous?
Solution 8. The answer is yes, g is continuous. We only need to show that if {xn } is a
sequence so that xn → x, then f (xn ) → f (x).
 1.7. COMPARISON WITH THE RIEMANN INTEGRAL xliii

Indeed, first note that the continuity condition (b) implies that the right side on the definition
of g(x) is well defined (really we only need that for fixed x, fx (y) = f (x, y) is integrable on [0, 1]).
On the other hand, by (c), f (xn , y) → f (x, y) for each y ∈ [0, 1], in particular for almost every
y.
Since the set [0, 1] has finite measure, then the function h = χ[0,1] ∈ L (m)
Let fn (y) = f (xn , y) and fx (y) = f (x, y), then fn (y) → fx (y) and by (a)
|fn (y)| = |f (xn , y)| ≤ 1 = h (y)
for all n and y ∈ [0, 1], using the dominated convergence theorem we have
Z 1 Z 1 Z 1
lim g(xn ) = lim f (xn , y)dy = lim f (xn , y)dy = f (x, y)dy = g(x).
n→∞ n→∞ 0 0 n→∞ 0
This concludes the proof.

1.7. COMPARISON WITH THE RIEMANN INTEGRAL

Our next theorem will show that every function which is Riemann-integrable on an interval with hyphen?
is also Lebesgue-integrable, and that Riemann-integrable functions are subject to rather stringent
continuity conditions. Quite apart from the fact that the Lebesgue theory therefore enables us to
integrate a much larger class of functions, its greatest advantage lies perhaps in the ease with which
many limit operations can be handled; from this point of view, Lebesgue’s convergence theorems
may well be regarded as the core of the Lebesgue theory.
One of the difficulties which is encountered in the Riemann theory is that limits of Riemann-
integrable functions, for example, If {qn ∈ [a, b] : n ∈ N} is an enumeration of the rationals number
in the interval [a, b] and Ek = {q1 , q2 , . . . , qk } consider fk = χEk , f = χQ∩[a,b] , then the discon-
tinuities of fk are the points of Ek , since Ek is finite, then fk are Riemann-integrable functions,
also lim fk = f and note that f is not Riemann-integrable, because if P = {x1 , x2 , . . . , xn } is a
k→∞
partition of [a, b] , L (f, P ) = 0, U (f, P ) = 1, So
Z b Z b
f dx = sup {L (f, P )} 0 < 1 = inf {U (f, P )} = f dx.
a a
This difficulty is eliminated with the Lebesgue theory, because limits of measurable functions are
always measurable.
Let the measure space X be the interval [a, b] of the real line, with µ = m (the Lebesgue
measure), and M the family of Lebesgue-measurable subsets of [a, b]. Instead of
Z
f dm
X
it is customary to use the familiar notation
Z b
f dx
a
for the Lebesgue integral of f over [a, b]. The notation f ∈ R on [a, b] means that f is Riemann
integrable on [a, b], To distinguish Riemann integrals from Lebesgue integrals, we shall now denote
the former by
Z b
R f dx.
a
xliv 1. LEBESGUE MEASURE AND INTEGRATION

Theorem 22. (a) If f ∈ R on [a, b], then f ∈ L on [a, b] and

Z b Z b
(1.7.1) f dx = R f dx.
a a
(b) Suppose f is bounded on [a, b]. Then f ∈ R on [a, b] if and only if f is continuous almost
everywhere on [a, b].