Solutions of Homework #2
1. A = {A ⊂ X|A or Ac is finite}.
(a)
(A is an algebra)
(i) ∵ ∅ = Xc is finite, ∴ ∅, X ∈ A.
(ii) Let A ∈ A.
If A is finite ⇒ (Ac)c = A is finite ⇒ Ac ∈ A.
If Ac is finite ⇒ Ac ∈ A.
(iii) Let {An}kn=1 ⊂ A.
If A0ns are finite. ⇒Sk
n=1An is finite. ∴Sk
n=1An ∈ A.
If there is An0 ∈ A such that Acn0 is finite.
⇒ (Sk
n=1An)c =Tk
n=1Acn ⊂ Acn
0, ∴ (Sk
n=1An)c is finite.
∴Sk
n=1An∈ A.
(A is not a σ-algebra)
Let X = {xn}∞n=1 and Ak = {x2k} for k = 1, 2, . . ..
Then S∞
k=1Ak is infinite and (S∞
k=1Ak)c = {x2k−1}∞k=1 is also in- finite. Therefore S∞
k=1Ak doesn’t belong to A and A is not a σ-algebra.
(b) Let {An}kn=1 ⊂ A be such that An∩ Am = ∅, ∀m 6= n. Then (i) If A0ns are finite, then Sk
n=1An is finite.
⇒ m(Sk
n=1An) = 0 =Pk
n=1m(An).
(ii) If there is An0 such that (An0)c is finite, then m(An0) = 1.
⇒ (Sk
n=1An)c =Tk
n=1(An)c ⊂ An0 is finite.
∴ m(Sk
n=1An) = 1.
On the other hand, ∵ An∩ Am = ∅, ∀m 6= n
∴S
n6=n0An⊂ (An0)c.
∴S
n6=n0An is finite. ⇒ m(S
n6=n0An) = 0 =P
n6=n0m(An).
Hence m(Sk
n=1An) = 1 = m(An0)+0 = m(An0)+S
n6=n0m(An) = Pk
n=1m(An).
(c) If X is uncountable, then m can be extended to a countably ad- ditive measure on a σ-algebra.
Claim: m is countably additive on A.
pf: Let {An}∞n=1⊂ A be such that
An∩ Am = ∅, ∀n 6= m and
∞
[
n=1
An∈ A.
(i) If An is finite, ∀n and S∞
n=1An ∈ A, then we will claim S∞
n=1An is finite.
if not, (S∞
n=1An)c is finite.
Because X is uncountable, we know thatS∞
n=1An is also un- countable.
Then we get a contradiction by the fact that the A0ns are finite.
(ii) If there is An0 such that (An0)c is finite, then (S∞
n=1An)c ⊂ (An0)c is finite.
∴ m(S∞
n=1An) = 1.
On the other hand, ∵ An∩ An0 = ∅, ∀n 6= n0.
∴S
n6=n0An⊂ Acn0 ⇒S
n6=n0An is finite.
∴ m(S
n6=n0An) = 0 =S
n6=n0m(An).
So m(S∞
n=1An) = 1 = m(An0) = P∞
n=1m(An).
Since m is countably additive on the algebra A, m can be extended to a countably additive measure on a σ-algebra.
If fact, if we set
S = {A ⊂ X|A or Ac is countable } and
˜
m(A) = 1, if Ac is countable 0, if A is countable
Then we can check that S is a σ-algebra and ˜m is a measure on S such that ˜m = m on A.
2.
µ∗([
j
Aj) ≤ µ∗([
j
Aj∪[
j
Bj)
= µ∗([
j
(Aj ∪ Bj)) = µ∗([
j
((Aj 4 Bj) ∪ (Aj ∩ Bj)))
= µ∗(([
j
(Aj 4 Bj)) ∪ ([
j
(Aj∩ Bj)))
≤ µ∗(([
j
(Aj 4 Bj)) + µ∗([
j
(Aj ∩ Bj)))
≤X
j
µ∗(Aj4 Bj) + µ∗(([
j
Aj) ∩ ([
j
Bj)) = µ∗(([
j
Aj) ∩ ([
j
Bj))
≤ µ∗([
j
Aj)
Therefore µ∗(S
jAj) = µ∗((S
jAj) ∩ (S
jBj)).
Similarly, µ∗(S
jBj) = µ∗((S
jAj) ∩ (S
jBj)).
Hence, µ∗(S
jAj) = µ∗(S
jBj).
3. Let F be a closed subset of X. We want to show that ∀A ∈ 2X µ∗(A) = µ∗(A ∩ F ) + µ∗(A \ F ).
Consider An= {x ∈ A \ F |d(x, F ) ≥ 1n}.
Since ρ(An, A ∩ F ) = inf {d(x, y)|x ∈ An, y ∈ A ∩ F } > 0, by assump- tion we have
µ∗(An∪ (A ∩ F )) = µ∗(An) + µ∗(A ∩ F ) Since An∪ (A ∩ F ) ⊂ A, therefore
µ∗(An) + µ∗(A ∩ F ) ≤ µ∗(A) (1)
Claim(1): S∞
n=1An= A \ F . Obviously,S∞
n=1An⊂ A \ F . If not, ∃x ∈ (A \ F ) \S∞
n=1An. ⇒ d(x, F ) < 1n, ∀n.
Therefore d(x, F ) = 0. Since F is closed, we have x ∈ F , and therefore we get a contradiction.
If we can show µ∗(S∞
n=1An) = limn→∞µ∗(An), then let n → ∞ in (1), we have
µ∗(A \ F ) + µ∗(A ∩ F ) ≤ µ∗(A)
Since the opposite inequality is always true, F is µ∗-measurable.
Claim(2): µ∗(S∞
n=1An) = limn→∞µ∗(An) There are two proof of this claim:
pf(a): We will show the general case, that is,
If Bn
∞
[
n=1
, then µ∗(
∞
[
n=1
Bn) = lim
n→∞µ∗(Bn)
step1 : If {Bn} are µ∗-measurable, then the equality holds.(We have showed this in the class.)
step2 : For each n, we can choose a measurable cover ˜Cn of Bn.
Let Cn = T∞
k=nC˜n, then Cn % and measurable. Moreover C˜k ⊃ Bk ⊃ Bn, ∀k ≥ n. Therefore Bn⊂ Cn.
From the following inequality,
µ∗(Bn) ≤ µ∗(Cn) ≤ µ∗( ˜Cn) = µ∗(Bn)
we have µ∗(Cn) = µ∗(Bn). ie. Cn is also a measurable cover of Bn.
By problem 4 of Homework#3,S∞
n=1Cnis a measurable cover of S∞
n=1Bn.
Then by step1, µ∗(S∞
n=1Cn) = limn→∞µ∗(Cn).
Hence µ∗(S∞
n=1Bn) = limn→∞µ∗(Bn).
By Taking An= Bn, we proves this claim.
pf(b): (多數同學的證明)
If µ∗(A) = ∞, then µ∗(A ∩ F ) + µ∗(A \ F ) ≥ µ∗(A) = ∞.
Therefore µ∗(A) = µ∗(A ∩ F ) + µ∗(A \ F ).
If µ∗(A) < ∞.
Let B1 = A1, Bn= An\ An−1. Then
ρ(B2k, B(2(k+1))), ρ(B2k−1, B2k+1) > 0, ∀k ∈ N By assumption,
µ∗(
∞
[
k=1
B2k) =
∞
X
k=1
µ∗(B2k)µ∗(
∞
[
k=1
B2k−1) =
∞
X
k=1
µ∗(B2k−1)
Since S∞
k=1B2k,S∞
k=1B2k−1⊂ A, we have
∞
X
k=1
µ∗(B2k),
∞
X
k=1
µ∗(B2k−1) < ∞.
Therefore ∀ε > 0, ∃ large integer k0 such that
∞
X
k=k0
µ∗(B2k),
∞
X
k=k0
µ∗(B2k−1) < ε 2. Hence ∀2n > k0,
µ∗(
∞
[
n=1
An) ≤ µ∗(A2n) +
∞
X
k=n+1
µ∗(B2k) +
∞
X
k=n+1
µ∗(B2k−1)
< µ∗(A2n) + ε 2+ ε
2 = µ∗(A2n) + ε.
Therefore supnµ∗(An) ≥ µ∗(S∞ n=1An).
∵ µ∗(An) % and µ∗(An) ≤ µ∗(A) < ∞,
∴ limn→∞µ∗(An) = supnµ∗(An).
µ∗(S∞
n=1An) ≤ limn→∞µ∗(An). On the other hand, ∵ An ⊂ S∞
n=1An, the opposite inequality is also true.
4. Let E ⊂ X.
If ν∗(E) = ∞ ⇒ ν∗(E) ≥ µ∗(E).
If ν∗(E) < ∞, given ε > 0, ∃Ai ∈ A such that E ⊂[
i
Ai and X
i
ν(Ai) ≤ ν∗(E) + ε.
Since µ = ν on A, we haveP
iν(Ai) = P
iµ(Ai) ≥ µ∗(E). ⇒ µ∗(E) ≤ ν∗(E) + ε, ∀ε.
⇒ µ∗(E) ≤ ν∗(E), ∀E.
∵ ν is σ-finite, countably additive, σ(A) is the σ-algebra generated by A and µ is the extension of ν. ∴ µ is the measure defined on σ(A) and µ = ν∗ on σ(A).
Let E ⊂ X.
If µ∗(E) = ∞ ⇒ µ∗(E) ≥ ν∗(E).
If µ∗(E) < ∞, given ε > 0, ∃Bi ∈ σ(A) such that E ⊂[
i
Bi and X
i
µ(Bi) ≤ µ∗(E) + ε.
Since µ = ν∗ on σ(A), we have P
iµ(Bi) =P
iν∗(Bi) ≤ ν∗(E).
⇒ ν∗(E) ≤ µ∗(E) + ε, ∀ε.
⇒ ν∗(E) ≤ µ∗(E), ∀E.
Hence µ∗ = ν∗.