Chapter 2.

Integration

2.1 Measurable functions

2.2 Integration of Nonnegative Functions

2.3 Integration of general functions

1
Review

Let (X, M, µ) be a given measure space

Let L+ = {f : X ! [0, 1] | f is (M, BR̄ )-measurable}.

If =a E, where E 2 M, a 2 [0, 1),

Z
dµ := aµ(E)
X

is called the integral of over X.

P
If = nj=1 aj Ej , aj 2 [0, 1), ai 6= aj , Ei \ Ej = ; for i 6= j,
Z n
X
dµ := aj µ(Ej )
X j=1

is called the integral of over X.

If is a nonnegative simple function with the standard representation

P
= nj=1 aj Ej and A 2 M, then A is a simple function and

n
X
A = aj Ej \A .
1

Z Z n
X
dµ := A dµ = aj µ(Ej \ A)
A X 1
is called the integral of over A.

Let f : X ! [0, 1] be measurable, i.e., f 2 L+ .

R R
X f dµ := sup{ X dµ |, 0   f, 2 L+ , is simple } is called
the integral of f over X.

2
 The Monotone Convergence Theorem.

If {fn } is a sequence in L+ such that

fj  fj+1 8j and f (x) = lim fn (x),

n!1

then Z Z
lim fn dµ = f dµ.
n!1 X X

If fs E Lt then

Sita 点 sfnf.ge Lt
S f19 5ft59
Fatou’s Lemma. If {fn } is any sequence in L+ , then
Z Z
lim inf fn  lim inf fn .
n!1 n!1

R
If f 2 L+ , then f dµ = 0 i↵ f = 0 a.e.
R
If f 2 L+ and f < 1, then {x | f (x) = 1} is a null set (i.e. measure
zero set), and {x | f (x) > 0} is -finite.

3
2.3 Integration of general functions

Let (X, M, µ) be a measure space.

Definition. Let f : X ! R̄ be (M, BR̄ )-measurable.

(f = f + f , f + , f : X ! [0, 1] are measurable).
R R
1) If f + or f is finite, we define
Z Z Z
f = f+ f
R ㄨ X X
and call f the integral of f (on X).

R R
2) f is called integrable if both f + and f are finite.

Remark.

+
J If I J ft5 Sf t
+
ff
1) |f | = f + f . f measurable =) |f |, f , f R measurable. Suppose
that f is measurable. Then f is integrable i↵ |f | < 1.

2) We can define the integral of a complex function f : X ! C as follows:

R R R
f = Ref + iImf . f = Ref + i Imf .

3) We can define the integral of f on E 2 M as follows:

R R
E f = X f E.

4) Suppose that f, g : X ! R̄ are measurable and f = g a.e.

R R
If f 2 L1 (X), then g 2 L1 (X) and f = g.

i
i fg a.e Sit
EEM.ME o

fg on E If1 4191 a e
 S1f1 1191 CD 9E L IX

L1 (X, M, µ) or L1 (X, µ) or L1 (X) or L1 (µ):

L1 (X) := {f : X ! R̄ measurable | f is integrable on X}.

Proposition 2.21 L1 (X) is a real vector space (i.e. 8 f, g 2 L1 (X)

R
and a, b 2 R, af + bg 2 L1 (X)) and f 7! f dµ is a linear functional
R R R
on L1 (X) (i.e. (af +bf ) = a f +b g for any f, g 2 L1 (X) and a, b 2 R).

pnof.tn iiiu.afain
flafl Jlallfl lalSlflCD
a f E L'IX
of 9ELIX want ftg E L'Lx
J1ft91 Slf1 191 J IfIt5191 ceo

L XI
i
fig E
L Xl is a vectorspace

Now we prove
fcaftbg afftbfg
a
ER.ftL'M want faf as f
5
If a o nothing need to prove

If 970 la f aflt afi

af S afit flafi Slaftj fcaf
afft aff
af flfao.la
laf fi ca

af af la fi aft

f af t af al ft _affitafftaff
fig E L X want
J If 9 f 19
t
Let h f g
t n h h
ftfigtg
htfigthtftgt
19
lhtgluhtff
fht_f hE f f t sf fgtfg

fh ff t fg

6
 R R
Proposition 2.22 If f 2 L1 (X), then | f|  |f |.

Proof 1St1 15ft1 1 1f1 S1f 1 S1f1 1If I

7
 Proposition 2.23

a) If f 2 L1 (X), then {x 2 X | f (x) = +1 or f (x) = 1} is a null set

and {x 2 X | f (x) 6= 0} is -finite.
R R R
b) If f, g 2 L1 (X), then E f = E g for all E 2 M i↵ |f g| = 0 i↵
f = g a.e.

Proof a x EX If ix or fix1 03
x EX1 1fX 1 03
YJHaoi.ME 0

Let F X1f 1 0
x1 Ifix 1 03
SHI CD in F is a finite

b f g ae ㄑ f g oa.ec If 9lo ae

J If 91 4

prove
JE f SE g for any E Em fgae
Assume fgae lf.gl o ae

fx If 91 0

EEMIJEf fgl llElf.gl

8

SElf.gl Sxlf
 Assume SE f JE g EEM.Want ME 0

Let E xlfxisgcx E X1fix 29 x

f9 o n f 95 Eco 0

JE f E 9 JE f 9 o fg so on Ft ME
0

M É
Similarly 0

f g a e

9
Remarks.

1) If f, gR 2 L1 (X),
R N 2 M with µ(N ) = 0, f (x) = g(x) for x 2 X \ N ,
then E f = E g for any E 2 M.

2) If f, g 2 L1 (X) and f = g a.e., we may write f = g.

Hence, for f, g 2 L1 (X), f = g really means that f = g a.e.
Let [f ] = {g 2 L1 (X) | f = g a.e.}. We usually write [f ] as f .

3) Suppose f 2 L1 (X), g : X ! R̄, N 2 M with µ(N ) = 0, f (x) = g(x)

for x 2 X \ N (g may not be (M, BR̄ )-measurable).
For any A 2 BR̄ ,
g 1 (A) = {x | g(x) 2 A}
= {x 2 X \ N | g(x) 2 A} [ {x 2 N | g(x) 2 A}
1
= f (A) \ (X \ N ) [ {x 2 N | g(x) 2 A} 2 M̄

Hence g is (M̄, BR̄ )-measurable.

4) If f 2 L1 (X, M, µ), then f 2 L1 (X, M̄, µ̄) and

R R
X f dµ = X f dµ̄.

5) If f 2 L1 (X, M, µ) and g = f a.e., then g 2 L1 (X, M̄, µ̄) and

R R
X gdµ̄ = X f dµ.

10
6) View f = g if f = g a.e.
L1 (X, M, µ) ⇠
= L1 (X, M̄, µ̄).
When we deal with integrals, we may assume (X, M, µ) is complete.
For otherwise, we can work on (X, M̄, µ̄).

7) (L1 (X), ⇢) is a metric space, where

R
⇢(f, g) = X |f g|dµ.

9If191 flf hltfch g

Y f ff 0

9If 91 919 f1
8) If {fn } ⇢ L1 (X) and fn ! f a.e., we may think f is measurable when
dealing with integrals.

9) For {fn } ⇢ L1 (X) and f 2 L1 (X), we have the following two conver-
gence:
fn ! f a.e.
R
fn ! f in L1 (i.e. ⇢(fn , f ) ! 0 or |fn f | ! 0)

11
Questions:
R R no
fn ! f a.e. =) fn ! f ? in general
fn ! f a.e. =) fn ! f in L1 ? in general no

fn ! f in L1 =) fn ! f a.e. ? in general no
R R
fn ! f in L1 =) fn ! f ?
Yes

12
 The Dominated Convergence Theorem:
Suppose that {fn } ⇢ L1 (X) satisfies

a) fn ! f a.e.
b) 9g 2 L1 \ L+ such that |fn |  g a.e. for all n.
R R
Then f 2 L1 and limn fn = limn fn .

Proof
lilfnlcga.li n

En EM MIEn
0 sit I fulx 1 gcxl on En

Let E U En Then ME 0 and lfnlxll g x for

i If I ga e any n 1 XEE

1151
fgcn
f
in E L X

Let gn g fn.hu gtfn

Fatou's lemma
By
flisn.fign liminffgn Jliminfhn

liminffhJ g fliminfcg
fgf.nl limsupfn flimsupfn
13
lininf 19 f
 fg_limsupffn

Sgfsfgumsupffn
limsupffn

f fliminfhn fgtfliminff
fliminflgtfn7

fgtffc.li minffgtfnl fgtliminfffn

Jf liminfff

limffn limfn Sf

14
 P R P1
Theorem 2.25 , 1
If {fj } ⇢ RL1P j=1 |f j |
P1 R < 1, then j=1 fj converges
1 1
a.e. to a function in L , and j=1 ff = j=1 fj .

Proof
iflfjl Jilfilc.co I By M.C.T

let gn
ffi9ilf.az
Thengcx ccoa.e 9n x

ifcxsae lg.nl
g

By Di CT 1点 9i'ㄨ 1
㗊 9i ㄨ

15
16
Remark.

(L1 (X), ⇢) is a metric space.

It is separable if 9 a countable dense subset, i.e., 9 { n} ⇢ L1 (X) such

that for any f 2 L1 (X), 9 nk such that

in L1 (X). i.ee
nk !f
14nn fl 0

Is L1 (X) separable?

Are simple functions dense in L1 (X)?

17
Theorem 2.26

1
P that f 2 L (X). 8 ✏ > 0, 9 an integrable simple function
1) Suppose
= aj Ej such that
Z
|f |dµ < ✏.

(Hence simple integrable functions are dense in L1 ).

2) If (X, M, µ) = (R, L, m), the sets Ej in the definition of can be taken

to be finite unions of open intervals. Moreover, there is a continuous
function g that vanishes outside a bounded interval such that
Z
|f g|dµ < ✏.

(Hence integrable continuous functions are dene in L1 (R) and then

L1 (R) is separable).
dense

18
Prof D Assume
ft L a sequence of simplefunctions
41142 st o Ell il 1421 14nklfl

lfen.li lIfos n sn

andIf 4nlE21t1

nlisnSlfM 4nlxlldMP him.If M 4nlxlldn.JOduo

Eso IN sit n N.flf 4nl

First Eso by 1 4
iaiXEiaiER.EiE W.L
2

Gassumeaito.si t.SI
0 f 41

Ut MEi aol.iq is integrable

intervals sit
Ii Ii Ini open

MlEi
i c
朚
1
let nt
iaigXIj EIaiXI JH

41 E 2E
4lc.SIf 41 114 E

For each ai X Ii we have

19 Jlaixi gil
 Īigi
let 9 cont Vanishes outside a boundedinterval

519 41 引 aixi gil c E

flf.gl flf 41 114 91 3E

Ree L'IR is separable

b7
g cont.gofor
x a
Y

sit
poly
flgPgl
pglzlcE Apolyonnthorder.pcxi
aotaixt.it an x
with rational coeff are countable
All the poly
countable
0 order ao

aotaix countable
1st order
countable
2nd order aotaixtazx2
Question: fn 0 E L'IX
Suppose that f : X ⇥ [a, b] ! R̄, and f (·, t) 2 L1 (X) for each t 2 [a, b].
R
Let F (t) = X f (x, t)dµ(x).

timo Stix du
R
When limt!t0 F (t) = X limt!t0 f (x, t)dµ(x) ? t
R
When dFdt(t) = X @f @t (x, t)dµ(x) ?

Theorem 2.27
f f
l X it der
Suppose that f : X ⇥ [a, b] ! R ( 1 < a < b < 1)
and
R that f (·, t) : X ! R is integrable for each t 2 [a, b]. Let F (t) =
X f (x, t)dµ(x).

a) Suppose that there is g 2 L1 (X) such that |f (x, t)|  g(x) for all x, t.
If limt!t0 f (x, t) = f (x, t0 ) for all x, then
limt!t0 F (t) = F (t0 ); in particular, if f (x, ·) is continuous for each x,
then F is continuous.
@f
b) Suppose that @t exists and there is g 2 L1 \ L+ such that
@f
| (x, t)|  g(x) 8 x, t.
@t
Then F is di↵erentiable and
Z
0 @f
F (t) = (x, t)dµ(x).
@t

20
Proof a To prove
tito Ft tint fcxit du X

ffcx.to f time fix it dMIX

it is equivalent to prove forany tn to

nligltlx.tn dunn fnmnflx.tn dulxl

1 ㄨ E L x Dc.T
I flx.tn 9 g by

hint tidu fliin.fixtnl

F't F
want to prove
fyi
a
b
no n hi du
11
0
limo ffxitth
f t'd
h Mx

hu fxitffftlx.tn
his lfx.thnl 9x
sometn between
for
t and

tthn i.ByP
4Tbimnfflxitthnl
hu
flx.tl
dM

fǎmntlxithtdu
21
 ftx.nu
Fine 苦exit dux

22
Question: What is the relation between Riemann integral and Lebesgue
integral?
Riemann integral:

23
Lebesgue integral:

24
Theorem 2.28 Let f be a bounded real-valued function on [a, b].

a) If f is Riemann integrable, then f is Lebesgue measurable (hence it

Rb
is
R Lebesgue integrable on [a, b] since it is bounded) and a f (x)dx =
[a,b] f dm.

b) f is Riemann integrable i↵ {x 2 [a, b] | f is discontinuous at x} has

Lebesgue measure zero.

25
26
27