Вы находитесь на странице: 1из 6

Jeffrey Hellrung

Wednesday, October 12, 2005


Math 245A, Homework 01
Chapter 1, # 3, 4, 5, 7, 8, 12, 14, 15

3. Let M be an infinite σ-algebra.

a. M contains an infinite sequence of disjoint sets.


b. card(M) ≥ c.

Solution

a. Construct the sequences


{En }∞
n=0

and
{Mn }∞
n=0 ;

where Ei ⊃ Ei+1 (and the containment is strict) and Mn is an infinite σ-algebra on En ; as follows.
Set E0 = X and M0 = M.
Now given En and Mn as above, let A ∈ Mn with A 6= ∅ and A 6= En (possible since Mn is
infinite). Consider the following subsets of Mn :

EA = {A ∩ B | B ∈ Mn } ;

EEn \A = {(En \A) ∩ B | B ∈ Mn } .


Notice that EY is a σ-algebra on Y for Y = A, En \A.
Now since Mn is infinite, at least one of the EY ’s must be infinite. For notice that every set in
Mn is the union of a set in EA and a set in EEn \A , hence if both EY ’s were finite, Mn would be
finite. Thus, without loss of generality, we may assume that EA is infinite. We then set En+1 = A
and Mn+1 = EA , and note that both satisfy the desired properties.
We thus obtain an infinite strictly nested sequence {En } of sets in M, hence the sequence
{En \En+1 }∞n=0 is an infinite sequence of nonempty pairwise-disjoint sets in M.
b. Let N = {0, 1, 2, . . .}. We can provide a bijection between P(N) and M({En }) ⊂ M, which will,
of course,
S show that card(M) ≥ card(P(N)) = c. Indeed, we can associate to each N ⊂ N the
set n∈N En ∈ M({En }), and due to the pairwise-disjointness of the En ’s, this association is
bijective.

4. An algebra A is a σ-algebra iff A is closed under countable increasing unions (i.e., if {Ej }∞
j=1 ⊂ A
S∞
and E1 ⊂ E2 ⊂ · · ·, then j=1 Ej ∈ A).
Solution
One direction is clear; that is, if A is a σ-algebra, A is closed under countable increasing unions.
Sn
Conversely, let {En }∞
S S
n=1 ⊂ A, and set Fn = i=1 En . Then Fn ⊂ Fn+1 , hence n En = n Fn ∈ A,
and A is closed under countable unions. A is an algebra, hence closed under complements, so it follows
that A is a σ-algebra.
5. If M is the σ-algebra generated by E, then M is the union of the σ-algebras generated by F as F
ranges over all countable subsets of E. (Hint: Show that the latter object is a σ-algebra.)
Solution
Let [
N = M(F),
F ⊂E

1
where each F is countable. Certainly N ⊂ M, as each M(F) ⊂ M(E) = M.
To show the opposite inclusion, we need only show that N is a σ-algebra. Indeed, given some E ∈ N ,
E ∈ M(F) for some countable subset F of E, hence E c ∈ M(F) and E c ∈ N . Thus, N is closed
under complements.
Let {En }∞
S
n=1 ⊂ N . Then each En ∈ M(Fn ) for some countable subset Fn of E. Let F = n Fn ; then
F is a countable subset of E, so S M(F) ⊂ N . Further, we have that M(F n ) ⊂ M(F), hence each
En ∈ M(F), so it follows that n En ∈ M(F) ⊂ N . Therefore, N is closed under countable unions,
and so N is a σ-algebra.
Pn
7. If µ1 , . . . , µn are measures on (X, M) and a1 , . . . , an ∈ [0, ∞), then j=1 aj µj is a measure on (X, M).
Solution
Let
n
X
µ= aj µj .
j=1

Then clearly µ : M → [0, ∞], µ(∅) = 0, and, for any sequence {Ei }∞
i=1 ⊂ M of disjoint sets,
à ! n
à ! n n
[ X [ X X XX X
µ Ei = aj µj Ei = aj µj (Ei ) = aj µj (Ei ) = µ(Ei ),
i j=1 i j=1 i i j=1 i

where the rearrangement of the sum is justified by the nonnegativity of all terms. Thus µ is countably
additive, hence a measure on (X, M).
8. If (X, M, µ) is a measure space and {Ej }∞ ⊂ M, then µ(lim inf Ej ) ≤ lim inf µ(Ej ). Also, µ(lim sup Ej ) ≥
³S ´j=1

lim sup µ(Ej ) provided that µ j=1 Ej < ∞.

Solution
Let

\
Fj = Ei ;
i=j

then

[
lim inf Ej = Fj .
j=1

Now since Fj ⊂ Fj+1 , Theorem 1.8c gives that

µ(lim inf Ej ) = lim µ(Fj ).


j→∞

But each Fj ⊂ Ej , hence µ(Fj ) ≤ µ(Ej ) and

lim µ(Fj ) ≤ lim inf µ(Ej ),


j→∞ j→∞

which shows the former claim.


Similarly, if we let

[
Gj = Ei ,
i=j

then

\
lim sup Ej = Gj .
j=1

2
Since Gj ⊃ Gj+1 and G1 < ∞, Theorem 1.8d gives that
µ(lim sup Ej ) = lim µ(Gj ),
j→∞

and since Gj ⊃ Ej , µ(Gj ) ≥ µ(Ej ) and


lim µ(Gj ) ≥ lim sup µ(Ej ),
j→∞ j→∞

which shows the latter claim.


12. Let (X, M, µ) be a finite measure space.

a. If E, F ∈ M and µ(E∆F ) = 0, then µ(E) = µ(F ).


b. Say that E ∼ F if µ(E∆F ) = 0; then ∼ is an equivalence relation on M.
c. For E, F ∈ M, define ρ(E, F ) = µ(E∆F ). Then ρ(E, G) ≤ ρ(E, F ) + ρ(F, G), and hence ρ
defines a metric on the space M/ ∼ of equivalence classes.

Solution

a.
µ(E∆F ) = µ((E\F ) ∪ (F \E)) = µ(E\F ) + µ(F \E),
so if µ(E∆F ) = 0, it must be that
µ(E\F ) = 0 = µ(F \E).
Now
µ(E\F ) = µ(E ∪ F ) − µ(F ),
µ(F \E) = µ(E ∪ F ) − µ(E),
hence
µ(E) = µ(E ∪ F ) = µ(F ).
b. Since E∆E = ∅, ∼ is reflexive.
E∆F = F ∆E, so ∼ is symmetric.
Suppose µ(E∆F ) = µ(F ∆G) = 0. Again,
µ(E∆G) = µ(E\G) + µ(G\E).
Now
µ(E\G) = µ(((E\F ) ∪ (E ∩ F ))\G)
= µ((E\F \G) ∪ ((E ∩ F )\G))
= µ(E\F \G) + µ((E ∩ F )\G) ,
≤ µ(E\F ) + µ(F \G)
= 0
and similarly µ(G\E) = 0, hence µ(E∆G) = 0 and ∼ is transitive.
c. We first establish that ρ is well-defined. Suppose E ∼ E ′ and F ∼ F ′ ; then
ρ(E, F ) = µ(E∆F ) = µ(E\F ) + µ(F \E).
Now E\F ∼ E ′ \F , so µ(E\F ) = µ(E ′ \F ). Further,
µ((F \E)∆(F \E ′ )) = µ((F \E)\(F \E ′ )) + µ((F \E ′ )\(F \E))
= µ(F ∩ (E ′ \E)) + µ(F ∩ (E\E ′ ))
≤ µ(E ′ \E) + µ(E\E ′ ) ,
= µ(E∆E ′ )
= 0

3
so F \E ∼ F \E ′ and µ(F \E) = µ(F \E ′ ). It follows that

ρ(E, F ) = µ(E ′ \F ) + µ(F \E ′ ) = ρ(E ′ , F ),

and a similarly argument shows this is then equal to ρ(E ′ , F ′ ). Therefore, ρ is well-defined.
Using the expansion in b.,

ρ(E, G) = µ(E∆G)
= µ(E\G) + µ(G\E)
,
≤ µ(E\F ) + µ(F \G) + µ(G\F ) + µ(F \E)
= µ(E∆F ) + µ(F ∆G)

hence ρ satisfies the triangle inequality. Defining ρ on equivalence classes of M is necessary


to obtain positive definiteness, and ρ is clearly symmetric, hence ρ is a metric on the space of
equivalence classes.

14. If µ is a semifinite measure and µ(E) = ∞, for any C > 0 there exists F ⊂ E with C < µ(F ) < ∞.
Solution
Let
C ∗ = sup {µ(F ) < ∞ | F ⊂ E} .
By the semifiniteness of µ, C ∗ > 0 (i.e., the set over which the supremum is taken is nonempty). Now
if C ∗ = ∞, the claim holds, so suppose C ∗ < ∞. Let Cn = C ∗ − 1/n, n = 1, 2, . . .. Then by the
definition of C ∗ , for each Cn , there exists Fn ⊂ E such that Cn < µ(Fn ) < ∞. Set F = n Fn ⊂ E.
S
Then
1
µ(F ) ≥ µ(Fn ) > Cn = C ∗ −
n
for any n, hence µ(F ) ≥ C ∗ .
Let -n−1
[
Gn = Fn Fi ⊂ Fn ;
i=1

then each of the Gn ’s are pairwise-disjoint, and their union is all of F , so


à !
[ X
µ(F ) = µ Gn = µ(Gn ).
n n

Now by the construction of C ∗ , either µ(F ) = C ∗ or µ(F ) = ∞. In the latter case, we see that, for
some N ,
XN
C∗ < µ(Gn ) < ∞;
n=1

but à !
N
[ N
X
µ Gn = µ(Gn ),
n=1 n=1
SN
contradicting the construction of C ∗ (since n=1 Gn ⊂ E).

On the other hand, in the case that µ(F ) = C , we have that µ(E\F ) = ∞ still, hence the semifiniteness
of µ provides G ⊂ E\F such that 0 < µ(G) < ∞, from which it follows that G ∪ F ⊂ E and
C ∗ < µ(G ∪ F ) < ∞, also contradicting the construction of C ∗ .
It follows that C ∗ cannot be finite, which proves the claim.

4
15. Given a measure µ on (X, M), define µ0 on M by µ0 (E) = sup {µ(F ) | F ⊂ E and µ(F ) < ∞}.

a. µ0 is a semifinite measure. It is called the semifinite part of µ.


b. If µ is semifinite, then µ = µ0 . (Use Exercise 14.)
c. There is a measure ν on M (in general, not unique) which assumes only the values 0 and ∞ such
that µ = µ0 + ν.

Solution

a. First, µ0 makes sense since the set over which the supremum is taken is nonempty (it includes
the empty set, for example). Also note that µ0 (E) ≤ µ(E), with equality if µ(E) < ∞.
We first show that µ0 is indeed a measure. Clearly, µ0 (∅) = 0.
Suppose {Ej }∞
S
j=1 ⊂ M are pairwise-disjoint, and set E = j Ej .
Let F ⊂ E such that µ(F ) < ∞. Then
 
[ X X
µ(F ) = µ  F ∩ Ej  = µ(F ∩ Ej ) ≤ µ0 (Ej ),
j j j

and taking the supremum over all such F yields


X
µ0 (E) ≤ µ0 (Ej ).
j

On the other hand, let Fj ⊂ Ej such that µ(Fj ) < ∞. Then, for any n,
 
n
[ n
X
µ0 (E) ≥ µ  Fj  = µ(Fj ),
j=1 j=1

since the sum is finite, hence the inequality is preserved in the limit:
X
µ0 (E) ≥ µ(Fj ),
j

and since this holds for all such choices of {Fj }, it follows that
X
µ0 (E) ≥ µ0 (Ej ),
j

which establishes that X


µ0 (E) = µ0 (Ej )
j

and µ0 is indeed a measure on M.


Now suppose µ0 (E) = ∞. Then the set over which the supremum is taken includes more than ∅,
and there exists some F ⊂ E such that 0 < µ(F ) < ∞. In such case, µ0 (F ) = µ(F ), hence µ0 is
semifinite.
b. If µ is semifinite, it follows from Exercise 14 that, given any set E such that µ(E) = ∞ and any
C > 0, there exists F ⊂ E with µ(F ) > C. It follows immediately that µ0 (E) = ∞, and since µ0
agrees with µ on sets of finite µ-measure, µ0 = µ on all of M.

5
c. Define ν : M → {0, ∞} by
ν(E) = 0 if E is µ-semifinite,
ν(E) = ∞ otherwise
for E ∈ M.
Then in the two cases for E, it is easy to verify that µ = µ0 +ν, since µ and µ0 agree on µ-semifinite
sets, by b.
We next show that ν is indeed a measure. Clearly ν(∅) = 0, while if we let {Ej }∞ j=1 ⊂ M
be³ pairwise-disjoint,
´ then if all E j ’s are µ-semifinite, their union is certainly µ-semifinite, so
S P
ν j Ej = 0 = j ν(Ej ). On the other hand, if one of the Ej ’s fails to be µ-semifinite, the
³S ´ P
union likewise fails, and we similarly arrive at ν j Ej = ∞ = j ν(Ej ). Therefore, ν is a
measure as claimed.

Вам также может понравиться