Review of Measure Theory (507/420), Winter 2010, Term 1

Let X be a nonempty set. We denote by P(X) the set of all subsets of X.

Denition (Algebras and -algebras)
(a) An algebra is a nonempty collection A of subsets of X that is closed under nite unions and
complements.
Main example: Let A
h
denote the algebra consisting of nite unions of h-intervals, i.e., intervals
of the form (a, b] or (a, ) where a < , < b < . Equivalently, A
h
is the collection
of nite disjoint unions of h-intervals.
(b) An -algebra is a nonempty collection A of subsets of X that is closed under countable unions
and complements.
Note: Every -algebra is an algebra.
(c) If E P(X), then the algebra generated by EE E is
M(E) =

is a algebra containing E
_
(d) If X is a metric space (or, more generally, a topological space) then the Borel algebra on
X, denoted B
X
, is the algebra generated by the family of open subsets of X.
Main example: B
IR
= M(E), where E is A
h
or any one of the families of open sets, open intervals
or closed intervals.
(e) A pair (X, A) where A is a -algebra of subsets of X is called a measurable space.
Denition (Measures)
(a) A nitely additive measure on an algebra A P(X) is a function : A [0, ] such that
i) () = 0
ii) (nite additivity) If {E
1
, , E
n
} is a nite collection of disjoint subsets of X with
{E
1
, , E
n
} A, then

_
n
_
j=1
E
j
_
=
n

j=1
(E
j
)
(b) A measure on a algebra M P(X) is a function : M [0, ] such that
i) () = 0
ii) (countable additivity) If {E
j
} is a countable collection of disjoint subsets of X with
{E
j
} M, then

_
_
j
E
j
_
=

j
(E
j
)
If is a measure on the algebra M P(X), then (X, M, ) is called a measure space.
October 22, 2010 Review of Measure Theory 1
(c) A measure on an algebra M P(X) is called
i) probability if (X) = 1
ii) nite if (X) <
iii) nite if there is a countable collection {E
n
}
nIN
M of subsets with X =

n=1
E
n
and with (E
n
) < for all n IN.
iv) seminite if for each E M with (E) = , there is an F M with 0 < (F) <
and F E.
Note: Probability Finite -nite semi-nite
Theorem. Let (X, M, ) be a measure space and E, F, E
1
, E
2
, M.
(a) (Monotonicity) If E F, then (E) (F).
(b) (Subadditivity)
_

n=1
E
n
_

n=1
(E
n
)
(c) (Continuity from below) If E
1
E
2
E
3
, then
_

n=1
E
n
_
= lim
n
(E
n
).
(d) (Continuity from above) If (E
1
) < and E
1
E
2
E
3
, then
_

n=1
E
n
_
=
lim
n
(E
n
).
Denition. A measure is complete if
N M, (N) = 0, Z N = Z M
Theorem (Completion) Let (X, M, ) be a measure space. Set
N =
_
N M

(N) = 0
_
M =
_
E Z

E M, Z N for some N N
_
: M [0, ] with (E Z) = (E) for all E M and Z N for some N N
Then
(a) M is a algebra.
(b) is a welldened, complete measure on

M, called the completion of .
(c) is the unique extension of to

M.
Denition. An outer measure on X is a function

: P(X) [0, ] such that

i) (monotonicity)

() = 0
ii) (countable subadditivity) If E F, then

(E)

(F).
iii) If {A
j
} is a countable collection of subsets of X, then

_
_
j
A
j
_

(A
j
)
October 22, 2010 Review of Measure Theory 2
Proposition. Let E P(X) and : E [0, ] be such that {, X} E and () = 0. Dene,
for all A X,

(A) = inf
_

n=1
(E
n
)

_
E
n
_
nIN
E, A

n=1
E
n
_
Then

is an outer measure (induced by , E).

Let

be an outer measure on X. A subset A X is said to be

measurable if

(E) =

(E A) +

(E A
c
) for all E X
Theorem (Caratheodory) Let

be an outer measure on X and M

be the set of

measurable
subsets of X. Then
(a) M

is a algebra.
(b) The restriction,

M, of

to M

is a complete measure.
(c) Any set with outer measure zero is

-measurable.
Denition. A premeasure on an algebra A P(X) is a function : A [0, ] such that
i) () = 0
ii) If {E
j
} is a countable collection of disjoint subsets of X with {E
j
} A and

E
j
A,
then

_
_
j
E
j
_
=

j
(E
j
)
Theorem. Let
i) A P(X) be an algebra,
ii)
0
be a premeasure on A,
iii)

be the outer measure induced by

0
, A
(iv) M

be the collection of

measurable sets.
(v) =

|
M
.
Then
(a) (X, M

, ) is a complete measure space.

(b) A M

(c)

A =
0
.
If
0
is -nite, then
(d) is the unique extension of
0
to any -algebra U such that M(A) U M

.
(e) (X, M

, ) is the completion of (X, M(A), |

M(A)
)
October 22, 2010 Review of Measure Theory 3
Proposition. Let A
h
be the algebra consisting of nite unions of h-intervals. Let F : IR IR be
increasing (i.e., nondecreasing) and right continuous. Then

0
((a, b]) F(b) F(a)
extends to a well dened premeasure on A
h
Corollary.
(a) Let F : IR IR be increasing and right continuous. There is a unique Borel measure
F
on
IR such that
F
_
(a, b]
_
= F(b) F(a) for all a, b IR with a < b.
(b) Let F, G : IR IR be increasing and right continuous. Then
F
=
G
if and only if F G is
a constant function.
(c) If is a Borel measure on IR that is nite on all bounded Borel sets, then =
F
for the
increasing, right continuous function
F(x) =
_
_
_

_
(0, x]
_
if x > 0
0 if x = 0

_
(x, 0]
_
if x < 0
Denition (LebesgueStieltjes measures): Let F : IR IR be increasing and right continuous.
The LebesgueStieltjes measure,
F
, associated to F is the complete measure determined (by
Caratheodorys Theorem ) from the outer measure

F
which is, in turn, determined from the
premeasure
0
that is associated to F.
Note that I have not used the term LebesgueStieltjes measure in class, but this is the
terminology used in Folland.
Lebesgue measure, m, is the LebesgueStieltjes measure associated to the function F(x) =
x. Recall that in class, we rst dened Lebesgue outer measure m

as the outer measure induced by

length on open intervals, showed that this is the same as the outer measure induced by length on A
h
and then dened Lebesgue measure m to be the complete measure obtained from Caratheodorys
Theorem . The collection of Lebesgue measurable sets, L, is the set of m

-measurable sets
Theorem (Lebesgue measure properties)
(a) L = {E Z : E B
IR
, m

(Z) = 0}
(b) Lebesgue measure m is the unique measure dened on L such that the measure of an interval
is its length.
(c) (IR, L, m) is the completion of (IR, B
IR
, m|
B
IR
)
(d) Lebesgue measure m is translation-invariant.
(e) There does not exist a translation-invariant measure on P(IR) such that the measure of an
interval is its length (because of the famous set N, which consists of exactly one element from
every coset of Q in the group [0, 1) (with addition modulo 1 as the group operation)).
October 22, 2010 Review of Measure Theory 4
Example 1 (Examples of Lebesgue measure)
(a) Lebesgue measurable sets with measure zero: any nite or countable set, Q, the Cantor set C
(b) Lebesgue measurable subsets of [0, 1] with measure one: [0, 1], (0, 1), [0, 1] \ Q, [0, 1] \ C,
(c) The generalized Cantor set C

, where = (
1
,
2
, . . .), with
i
(0, 1) has measure

i
(1
i
).
And this is strictly positive i

i
converges. All generalized Cantor sets are compact, nowhere
dense, uncountable subsets of [0.1]. For every (0, 1), there is a generalized Cantor set with
measure .
(d) If
(m)
satises m(C

(m) ) > 1 1/m, then K =

m
C

(m) is contained in [0, 1], has Lebesgue

measure one and is a countable union of nowhere dense closed sets. And [0, 1] \ K has Lebesgue
measure zero and is a countable intersection of dense open sets in [0, 1].
Theorem (Regularity) Let
F
be a LebesgueStieltjes measure,

F
be the corresponding outer
measure and M

be the set of all

F
measurable sets.
(a) For all E IR,

F
(E) = inf
_

n=1

F
((a
n
, b
n
))

n=1
_
a
n
, b
n
_
_
(b) For all E IR,

F
(E) = inf
_

F
(O)

O open , E O
_
(c) For all E M

F
(E) = sup
_

F
(K)

K compact, K E
_
(d) Let E IR. The following are equivalent.
(i) E M

(ii) E = V \ N
1
where V is G

(a countable intersection of open sets) and

F
(N
1
) = 0
(iii) E = H N
2
where H is F

(a countable union of closed sets) and

F
(N
2
) = 0
Denition (Product -algebras) Let be an index set. For each , let X

be a set and A

a -algebra of subsets of X

. Let X =

and

: X X

be the coordinate projection.

Dene the Product -algebra:

M(

(A

)).
Example (Examples of product -algebras)
(a) B
IR
n =
n
i=1
B
IR
(b) For i = 1, 2, . . ., let X
i
= {0, 1} and A
i
= P(X
i
). Let C be the collection of all cylinder sets in
X, i.e., sets of the form:
C
i
1
,i
2
,...,i
n
;a
1
,a
2
,...,a
n
= {x X : x
i
1
= a
1
, x
i
2
= a
2
, . . . , x
i
n
= a
n
};
October 22, 2010 Review of Measure Theory 5
here, 1 i
1
< i
2
< . . . < i
n
and each a
j
{0, 1} for j = 1, . . . , n. Then

i=1
A
i
= M(C) = B
X
where B
X
denotes the Borel sets in X corresponding to a specic metric on X, where the open
sets turn out to be precisely the unions of cylinder sets.
Facts:
(i) X is a compact metric space.
(ii) A subset of X is closed i it is compact.
(iii) A subset of X is compact i every open cover of the subset has a nite subcover (Heine-Borel
property)
Note: In Homework #3, Problem #7, you constructed probability measures on (X, B
X
) (Markov
chains).
October 22, 2010 Review of Measure Theory 6