Notes and problems on the Lebesgue integral

Let X ∈ M be a measurable subset of Rn . A function f : X → R is measurable if

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

Problem 1 Prove that the following are equivalent:

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

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, b)) is measurable for all a < b in R;

6. f −1 (U ) is measurable for all open subsets of R.

Problem 2 Let f and g be measurable functions. Show that the set

{x ∈ X|f (x) ≥ g(x)}

is measurable.

Theorem 1 Let fi be a countable collection of measurable functions. Then

1. inf fi is measurable;

2. sup fi is measurable;

3. lim inf fi is measurable;

4. lim sup fi is measurable.

Proof. Let h = inf fi . Then h−1 ((∞, a]) = fi−1 ((∞, a])) is a countable of union of
S
measurable sets and is therefore measurable. This implies that h is measurable by (3) of
Problem 1. A similar argument for sup fi where replace the sets (∞, a] with sets [a, ∞).
For (3) if we let gk = inf fi then g = sup gk = lim inf fi . All of the gk are measurable
i≥k
by (1) so g is measurable by (2). Reversing the roles of sup and inf we also get (4). 1

1
Corollary 2 1. A limit of measurable functions is measurable.

2. If f and g are measurable then min{f, g} and max{f, g} are measurable.

Proof. The limit of fi exists if and only if lim sup fi = lim inf fi so (1) follows from
(3) or (4) of Theorem 1.
(2) is a special case of (1) and (2) of Theorem 1. 2

Theorem 3 Let Φ : R2 → R be continuous and let u, v : X → R be measurable. Then

the function f (x) = Φ(u(x), v(x)) is measurable.

Proof. Let h(x) = (u(x), v(x)) and let R = I1 × I2 be a rectangle in R2 . Then

h−1 (R) = u−1 (I1 ) ∩ v −1 (I2 ) is measurable since u−1 (I1 ) and v −1 (I2 ) are measurable.
Since every open set in R2 is a countable union of rectangles this implies that h−1 (U ) is
measurable for any open set U .
To see that f is measurable we observe that for any open interval (a, b) we have
f −1 ((a, b)) = h−1 (Φ−1 (a, b)) is measurable since Φ−1 ((a, b)) is open by the continuity of
Φ and h−1 of any open set is measurable. 3

Corollary 4 If u, v : X → R are measurable then u + v and u · v are measurable.

Proof. Apply Theorem 3 to the functions (x, y) 7→ x + y and (x, y) 7→ x · y. 4

For E ⊂ X we define the characteristic function by


1 x∈E
χE (x) =
0 x 6∈ E.

Problem 3 Show that χE is measurable if and only if E ∈ M.

APsimple function is a finite linear combination of characteristic functions. That is

s = ki=1 ci χEi where ci ∈ R. Corollary 4 and Problem 3 implies that if all the Ei are
in M then s is measurable.

1. Let s = ki=1 ci χEi be a simple, measurable function and assume that

P
Problem 4
all of the ci are distinct (ci = cj if and only i = j). Show that all of the Ei are
measurable.

2. Show that every simple measurable function can be written as a finite linear com-
bination of characteristic functions of disjoint measurable sets.

2
 For a simple function we define the Lebesgue integral by
Z k
X
s(x)dm(x) = ci m(Ei ∩ X).
X i=1

Lemma 5 Let s and r be simple, meaurable functions, c a real number and A and B
disjoint measurable sets. Then
1. Z Z
cs(x)dm(x) = c s(x)dm(x);
A A

2. Z Z Z
s(x)dm(x) + r(x)dm(x) = (s + r)(x)dm(x);
A A A

3. Z Z Z
s(x)dm(x) + s(x)dm(x) = s(x)dm(x).
A B A∪B

Pk Proof. We’ll only prove (3) and the leave (1) and (2) to the reader. Let s =
i=1 ci χEi where the Ei are measurable sets. Then
Z Z k
X k
X
s(x)dm(x) + s(x)dm(x) = ci m(Ei ∩ A) + ci m(Ei ∩ B)
A B i=1 i=1
Xk
= ci (m(Ei ∩ A) + m(Ei ∩ B))
i=1
Xk
= ci m(Ei ∩ (A ∪ B))
i=1
Z
= s(x)dm(x).
A∪B

Let f : X → R be a non-negative function and let S(f ) be the set of simple, mea-
surable functions with s ≥ 0 and s ≤ f . Then we define the Lebesgue integral of f
by Z Z
f (x)dm(x) = sup s(x)dm(x).
X s∈S(f ) X

3
Theorem 6 Let f : X → R be a measurable, non-negative function. Then there exist
simple, measurable functions 0 ≤ s1 ≤ s2 ≤ . . . such that si (x) → f (x) for all x ∈ X.

Proof. We define each si as follows. Define

Ej = f −1 ([j2−1 , (j + 1)2−i )

for j = 0, . . . , i2i − 1 and define

Ei2i = f −1 ([i, ∞)).

Since f is measurable all of the Ej are measurable and

i
i2
X
si = j2−i χEj .
j=0

Note that all the Ej are disjoint and their union is X. Therefore if x ∈ Ej , si (x) =
j2−i ≤ f (x). Also if f (x) < i we f (x) − si (x) < 2−1 so si (x) → f (x) as i → ∞. 4

Lemma 7 For each E ∈ M define a function φE : M → [0, ∞] by

Z
φE (A) = χE (x)dm(x).
A
S∞
If Ai ∈ M are disjoint and A = i=1 Ai then
∞
X
φE (A) = φE (Ai ).
i=1

Proof. For anySB ∈ M we have φE (B) = m(E ∩ B). Since Ai ∩ E are disjoint sets
in M and A ∩ E = ∞ i=1 Ai ∩ E we have

∞
X ∞
X
φE (A) = m(E ∩ A) = m(E ∩ Ai ) = φE (Ai )
i=1 i=1

by the countable additivity of the measure m. 7

4
Proposition 8 Let s be a non-negative, simple, measurable function and define φs :
M → [0, ∞] by Z
φs (A) = s(x)dm(x).
A
S∞
If Ai ∈ M are disjoint and A = i=1 Ai then
∞
X
φs (A) = φs (Ai ).
i=1

Proof. By (2) of Problem 4 there are disjoint Ei ∈ M with s = ki=1 ci χEi . Since
P
s is non-negative the ci are also non-negative. We can even assume they are positive
since if ciP
= 0 we can clearly drop that term from thePsum. Then for any set B we have
φs (B) = ki=1 ci φEi (B). Therefore if φs (A) < ∞ or ∞ i=1 φs (Ai ) < ∞ we have
 
k k ∞ ∞ k ∞
!
X X X X X X
φs (A) = ci φEi (A) = ci  φEi (Aj ) = ci φEi (Aj ) = φs (Aj ).
i=1 i=1 j=1 j=1 i=1 j=1

Note that we are allowed to switch the order of summation (the 3rd equality) since the
total sum is finite and all terms are positive. In particular if one side of the equality is
finite so is the other side. 8

S∞
Corollary 9 Let A1 ⊂ A2 . . . be nested measurable sets with A = i=1 Ai and let s be
a measurable simple function. Then
Z Z
lim s(x)dm(x) = s(x)dm(x).
i→∞ A A
i

Proof. Let B1 = A1 and let Bi = Ai − Ai−1 for i ≥ 2. Note that the Bi are disjoint

5
 S∞ Sk
and measurable, A = i=1 Bi and Ak = i=1 Ai . Then
Z
s(x)dm(x) = φs (A)
A
∞
X
= φs (Bi ) by Proposition 8
i=1
k
X
= lim φs (Bi )
k→∞
i=1
Xk Z
= lim s(x)dm(x)
k→∞ Bi
Zi=1
= lim s(x)dm(x).
k→∞ Ak

Theorem 10 (Lebesgue’s Monotone Convergence Theorem) Let fi ≥ 0 be a se-

quence of non-negative measurable functions. Assume that fi ≤ fi+1 and fi → f , point-
wise, for some function f . The f is measurable and
Z Z
lim fi (x)dm(x) = f (x)dm(x).
i→∞ X X

Proof. By (1) of Corollary 2 the function f is measurable.

Since the fi are increasing we have fi ≤ fi+1 ≤ f and
Z Z Z
fi dm(x) ≤ fi+1 dm(x) ≤ f (x)dm(x).
X X X
R
ThereforeR X fi (x)dm(x) is an increasing sequence so it must converge to some λ ∈ [0, ∞]
with λ ≤ X f (x)dm(x). We need to prove the reverse inequality.
For all  > 0 we can find a simple, measurable function g with 0 ≥ g ≤ f and
Z Z
f (x)dm(x) − g(x)dm(x) < .
X X

For c ∈ (0, 1) define

Ei = {x ∈ X|fi (x) ≥ cg(x)}.

6
By Problem 2, the Ei are measurable. Note that ESi ⊂ Ei+1 and every x ∈ X is contained
in some Ei so Ei is a nested sequence with X = ∞ i=1 Ei . We then have
Z Z Z
fi (x)dm(x) ≥ fi (x)dm(x) ≥ c g(x)dm(x). (0.1)
X Ei Ei

The limit of the left hand side of this inequality is λ and by Corollary 9 the limit of the
right hand side is the integral of g over X. In particular we have
Z Z 
λ≥c g(x)dm(x) ≥ c f (x)dm(x) − 
X X

for all c ∈ (0, 1) and all  > 0. Therefore

Z
λ≥ f (x)dm(x)
X

and the proof is done. 10

Theorem 11 Let f ≥ 0 and g ≥ 0 be measurable, non-negative functions, c a real

number and A and B disjoint measurable sets. Then

1. Z Z
cf (x)dm(x) = c f (x)dm(x);
A A

2. Z Z Z
f (x)dm(x) + g(x)dm(x) = (f + g)(x)dm(x);
A A A

3. Z Z Z
f (x)dm(x) + f (x)dm(x) = f (x)dm(x).
A B A∪B

Proof. By Theorem 6 we can find non-decreasing sequences si and ri of simple,

non-negative, measurable functions with si → f and ri → g. Note that csi and si + ri
are also non-decreasing sequences of simple, non-negative, measurable functions with
csi → cf and si + ri → f + g. If we replace f and g by si and ri then (1)-(3) all hold by
Lemma 5. By the Theorem 10 the statements all hold in the limit also. 11

7
 We now define the Lebesgue integral of an arbitrary function. Given a function f set
f + = max{f, 0} and f − = max{−f, 0}. Then f = f + − f − . Define the integral of f by
Z Z Z
f (x)dm(x) = f + (x)dm(x) − f − (x)dm(x).
A A A

Note that if f is measurable so are both f + and f − .

Theorem 12 (Fatou’s Lemma) Let fi be non-negative measurable functions then

Z Z
(lim inf fi )(x)dm(x) ≤ lim inf fi (x)dm(x).
A i→∞ A

Proof. Let gk = inf fi . Then gk ≤ fi if i ≥ k and therefore

i≥k
Z Z
gk (x)dm(x) ≤ fi (x)dm(x)
A A

and Z Z
gk (x)dm(x) ≤ inf fi (x)dm(x).
A i≥k A
Also note that gk ≤ gk+1 and gk → lim inf fi so the Monotone Convergence Theorem
applies to the left hand side of the inequalities and as k → ∞ we get
Z  Z 
lim gk (x)dm(x) ≤ lim inf fi (x)dm(x)
k→∞ k→∞ i≥k A

which is exactly the desired inequality. 12

Theorem 13 (Lebesgue’s Dominated Convergence Theorem) Let fi be a sequence

Rof measurable functions with fi → f . Assume that |fi | ≤ g where g is measurable and
A g(x)dm(x) < ∞. Then
Z Z
lim fi (x)dm(x) = f (x)dm(x).
i→∞ A A

Proof. Since |fi | ≤ g, −fi ≤ g and 0 ≤ fi + g. By Fatou’s Lemma

Z Z
lim inf(fi + g)(x)dm(x) ≤ lim inf (fi + g)(x)dm(x).
A A

8
Note that lim inf(fi + g) = f + g since fi converges to f and g doesn’t depend on i.
Therefore the left hand side of the inequality becomes
Z Z Z
(f + g)(x)dm(x) = f (x)dm(x) + g(x)dm(x).
A A A

For the left hand side we have

Z Z Z 
lim inf (fi + g)(x)dm(x) = lim inf fi (x)dm(x) + g(x)dm(x)
A A A
Z Z
= lim inf fi (x)dm(x) + lim inf g(x)dm(x)
A A
 Z  Z
= lim inf fi (x)dm(x) + g(x)dm(x).
A A

Since the integral of g is on both sides of the inequality we can cancel it to get
Z Z
f (x)dm(x) ≤ lim inf fi (x)dm(x).
A A

Next we note that |fi | ≤ g implies that fi ≤ g and g − fi ≥ 0. Applying Fatou’s

Lemma to this sequence we get
Z Z
lim inf(g − fi )(x)dm(x) ≤ lim inf (fi + g)(x)dm(x).
A A

Following similar reasoning as above we see that the left hand side of the inequality
becomes Z Z
g(x)dm(x) − f (x)dm(x)
A A
and the right hand side is
Z  Z 
g(x)dm(x) + lim inf −fi (x)dm(x) .
A A

Again the integrals of g cancel out so we have

Z Z
− f (x)dm(x) ≤ lim inf −fi (x)dm(x)
A AZ

= − lim sup fi (x)dm(x).

A

This becomes Z Z
f (x)dm(x) ≥ lim sup fi (x)dm(x)
A A
which completes the proof. 13