Lecture Notes on Measure Theory and Integration

3 — Measurable Functions

Joel Feldman

University of British Columbia

October 16, 2019

1
3 Measurable Functions
Rb
We’re almost ready to define a f (x) dx or, for a more general measure µ,
Z X  
m −1
 m m+1


f (x) dµ(x) = lim n
µ f n
, n
∩ [a, b]
[a,b] n→∞
m

y = f (x)
(m+1)/n

m/n

a b x

m


f −1 n
, m+1
n
∩ [a, b]

−1
 m m+1


For this
 m m+1
to make sense, the set, f n
, n , of points where f takes values in
n
, n
had better be in the domain of µ, i.e. in the σ-algebra on which µ is
defined.

Definition 3.1 (Measurable Functions). Let X and Y be nonempty sets and MX

and MY be σ-algebras of subsets of X and Y respectively.
(a) A function f : X → Y is said to be (MX , MY )-measurable if

E ∈ MY =⇒ f −1 (E) ≡ x ∈ X f (x) ∈ E ∈ MX

(The (MX , MY ) in the notation is often suppressed.)

(b) A function f : X → R or C is said to be MX -measurable if

E ∈ BR or BC =⇒ f −1 (E) ≡ x ∈ X f (x) ∈ E ∈ MX

Measurable Functions 2 October 16, 2019

(c) A function f : R → R is said to be Borel-measurable if it is (BR , BR )-
measurable and is said to be Lebesgue-measurable if it is (L, BR )-measurable,
where L is the set of all Lebesgue measurable subsets of R. Similarly, a function
f : R → C is said to be Borel-measurable if it is (BR , BC )-measurable and is
said to be Lebesgue-measurable if it is (L, BC )-measurable.
(d) A function f : X → [0, ∞] is said to be MX -measurable if

E ∈ BR =⇒ f −1 (E) ∈ MX

where R = R ∪ {−∞, ∞} is the extended real line and BR is the σ-algebra1


BR = E ⊂ R E ∩ R ∈ BR

Lemma 3.2. Let X and Y be nonempty sets and MX and MY be σ-algebras of

subsets of X and Y respectively. Let f : X → Y
(a) Then
 −1
f (A) ⊂ X A ∈ MY ⊂ P(X)

and A ⊂ Y f −1 (A) ∈ MX ⊂ P(Y )

are σ-algebras.
(b) If MY is generated by E, then

f is (MX , MY )-measurable ⇐⇒ f −1 (E) ∈ MX for all E ∈ E


Proof. (a) That f −1 (A) ⊂ X A ∈  MY is a σ-algebra is part of Problem Set 6,
#2. We still have to show that M = A ⊂ Y f −1 (A) ∈ MX is a σ-algebra.
• Proof that M is not empty:
Since ∅ = f −1 (∅) ∈ MX , we have ∅ ∈ M.
• Proof that M is closed under complements:
If A ∈ M, then

f −1 (A) ∈ MX =⇒ f −1 (Y \ A) = X \ f −1 (A) ∈ MX
=⇒ Y \ A ∈ M

1
That BR is a σ-algebra is Problem Set 6, #8.

Measurable Functions 3 October 16, 2019

 • Proof that M is closed under countable unions:

If An n∈N ⊂ M, then

S  ∞
[
−1 −1 ∞
f (An ) ∈ MX for all n ∈ N =⇒ f n=1 An = f −1 (An ) ∈ MX
n=1
∞
[
=⇒ An ∈ M
n=1

(b) The direction “ =⇒ ” isobvious. For the other direction, if f −1 (E) ∈ MX for
all E ∈ E then, by part (a), A ⊂ Y f −1 (A) ∈ MX is a σ-algebra that contains
E and hence contains M(E) = MY . Hence f −1 (A) ∈ MX for all A ∈ MY and f is
(MX , MY )-measurable.

Corollary 3.3 (Continuous Functions). If X and Y are metric spaces (topological

spaces) and f : X → Y is continuous, then f is (BX , BY )-measurable.

Proof. If A ⊂ Y is open, then f −1 (A) ⊂ X is open, since f is continuous. Apply

part (b) of Lemma 3.2.

Corollary 3.4 (Real Valued Functions). Let X be a nonempty set and M be a

σ-algebra of subsets of X. Let f : X → R. Then


f is M-measurable ⇐⇒ f −1 (a, ∞) ∈ M ∀ a ∈ R


⇐⇒ f −1 [a, ∞) ∈ M ∀ a ∈ R


⇐⇒ f −1 (−∞, a) ∈ M ∀ a ∈ R


⇐⇒ f −1 (−∞, a] ∈ M ∀ a ∈ R

For f : X → [0, ∞] Then

f is M-measurable ⇐⇒ f −1 (a, ∞] ∈ M ∀ a > 0


⇐⇒ f −1 [a, ∞) ∈ M ∀ a ≥ 0

Proof. Apply part (b) of Lemma 3.2 and Proposition 2.9, for the first half. For the
second half tweak Problem Set 6, #8.

Example 3.5.
(a) If f : R → R is continuous, then f is Borel-measurable, by Corollary 3.3.
(b) If f : R → R is monotone, then f is Borel-measurable, by Problem Set 6, #5.

Measurable Functions 4 October 16, 2019

(c) If X is a set, M is a σ-algebra of subsets of X and E ∈ M, then the characteristic
function χE is M-measurable, since
 

 X if {0, 1} ⊂ A 


 

−1 E if 1 ∈ A, 0 ∈
/ A
χE (A) = ∈M


 X \ E if 0 ∈ A, 1 ∈
/ A 


 
∅ if 0, 1 ∈
/A
(c) If X is a set, M is a σ-algebra of subsets of X and E ∈
/ M, then the characteristic
function χE is not M-measurable, since


χ−1
E {1} = E ∈ /M
Theorem 3.6 (Measurable Function Toolbox). Let X be a nonempty set and M be
a σ-algebra of subsets of X. Let f, g : X → R be M-measurable and c ∈ R, or let
f, g : X → [0, ∞] be M-measurable and c > 0. Then
(a) f + c and cf are M-measurable.
(b) f + g is M-measurable.
(c) f g is M-measurable. (As usual, we use the convention that 0 × ∞ = 0.)
(d) max{f, g} is M-measurable. Here max{f, g} is the function from X to R defined
by max{f, g}(x) = max{f (x), g(x)}.
(e) min{f, g} is M-measurable.
(f ) Let, for each n ∈ N, the function fn : X → R be M-measurable and let the
function h : X → R. Or let, for each n ∈ N, the function fn : X → [0, ∞] be
M-measurable and let h : X → [0, ∞]. If
h(x) = lim fn (x)
n→∞
or h(x) = inf fn (x)
n∈N
or h(x) = sup fn (x)
n∈N
or h(x) = lim inf fn (x)
n→∞
or h(x) = lim sup fn (x)
n→∞

for all x ∈ X, then h is M-measurable.

(g) If h : R → R is Borel measurable and g : X → R is M-measurable, then h ◦ g
is
M measurable. Here h◦ g : X → R is the function defined by h◦ g(x) = h g(x) .
In particular, if h : R → R is continuous and g : X → R is M-measurable, then
h ◦ g is M measurable.

Measurable Functions 5 October 16, 2019

Remark 3.7. If, in part (f) ofTheorem 3.6, we have h(x) = supn fn (x) : X → R,
then, for each x ∈ X, the set fn (x) n ∈ N is bounded above. We shall see
shortly
 that it suffices for there to exist a set E ∈ M with µ(E c ) = 0 and the set
fn (x) n ∈ N bounded above for each x ∈ E.
Warning 3.8. In contrast to part (g) of Theorem 3.6, if h is Lebesgue measurable
and g is continuous, then the composition h ◦ g need not be Lebesgue measurable. See
#9 on page 48 of Folland.
Proof of Theorem 3.6. We’ll give the proof for the case of R-valued functions.
(a) The function f + c is measurable because

−1


f +c (a, ∞) = f −1 (a − c, ∞) ∈ M

If c > 0, cf and −cf are measurable because

−1


cf (a, ∞) = f −1 (a/c, ∞) ∈ M

−1


− cf (a, ∞) = f −1 (−∞, −a/c) ∈ M

The case c = 0 is trivial.

(b) follows from

 
(f + g)−1 (a, ∞) = x ∈ X f (x) + g(x) > a = x ∈ X f (x) > a − g(x)
[
= x ∈ X f (x) > r > a − g(x)
r∈Q
[


= f −1 (r, ∞) ∩ g −1 (a − r, ∞)
r∈Q

(c) First, f 2 is measurable because

(

−1
X if a < 0
f2 (a, ∞) = √
√

f −1 ( a, ∞) ∪ f −1 (−∞, − a) if a ≥ 0

Then f g is measurable because

f g = 41 (f + g)2 − 14 (f − g)2

(d) follows from

max{f, g} (a, ∞) = f −1 (a, ∞) ∪ g −1 (a, ∞)

Measurable Functions 6 October 16, 2019

(e) follows from min{f, g} = − max{−f, −g} or from





min{f, g} (a, ∞) = f −1 (a, ∞) ∩ g −1 (a, ∞)

(f) follows from

−1
[ −1

sup fn (a, ∞) = fn (a, ∞)
n∈N
n∈N



inf fn (x) = − sup −fn (x)
n∈N n∈N

−1
[ \ −1

or inf fn (a, ∞) = fn (a + r, ∞)
n∈N
r∈Q n∈N
r>0
n o
lim inf fn (x) = sup inf fm (x)
n→∞ n∈N m>n
n o
lim sup fn (x) = inf sup fm (x)
n→∞ n∈N m>n

and the fact that if limn∈N fn (x) exists, then limn∈N fn (x) = lim inf n→∞ fn (x).
(g) follows from



(h ◦ g)−1 (a, ∞) = x ∈ X h g(x) > a


= x∈X g(x) ∈ h−1 (a, ∞)
∈B
z }|R {

= g −1 h−1 (a, ∞) ∈ M
| {z }
∈BR

Definition 3.9 (Almost Everywhere2 ). Let (X, M, µ) be a measure space and let
f, g, fn : X → R for all n ∈ N. Then
(a) f = g a.e. if there is a set E ∈ M such that µ(E) = 0 and f (x) = g(x) for all
x∈ / E.
(b) f (x) = lim fn (x) a.e. if there is a set E ∈ M such that µ(E) = 0 and f (x) =
n→∞
lim fn (x) for all x ∈
/ E.
n→∞

2
“a.e.” is read “almost everywhere”.

Measurable Functions 7 October 16, 2019

Lemma 3.10 (Almost Everywhere). Let (X, M, µ) be a complete measure space.
(a) Let f, g : X → R. If f is measurable and f = g µ-a.e., then g is measurable.
(b) Let f : X → R and fn
: X → R for all n ∈ N. If fn is measurable, for all n ∈ N,
and the sequence fn n∈N converges µ-a.e. to f , then f is measurable.

Proof. The proof is (part of) Problem Set 6, #9.

Example 3.11. Here is an example which illustrates the need for the hypothesis,
in Lemma 3.10, that (X, M, µ) be complete. Let (X, M, µ) be any measure space
which is not complete. Pick any subsets Z, N ⊂ X with ∅ =
6 Z ⊂ N, µ(N) = 0 and
Z∈ / M. Define

f (x) = 1 for all x ∈ X

(
1 if x ∈ X \ Z
g(x) =
0 if x ∈ Z

Then
• f is measurable and
• f = g a.e. because f (x) = g(x) for all x ∈/ N , but
−1


• g is not measurable because g {0} = Z ∈ /M
In practice, the completeness requirement is usually not a big problem. If (X, M, µ)
is not complete and one ends up with a g (in part (a)) or an f (in part (b)) which
is not measurable, then one can tweak their definitions on a set of measure zero to
get measurable replacements — if Z is the set of x’s for which f (x) 6= g(x) (in part
(a)) or {fn (x)}n∈N does not converge to f (x) (in part (b)), and N is a null set which
contains Z, then just multiply g (in part (a)) or f (in part (b)) by the characteristic
function of the set X \ N.

Measurable Functions 8 October 16, 2019