Solutions of Homework #4
1. Claim1 : λ(E + s) = λ(E).
λ∗(E + s) = inf {Σλ∗(In)|[
In⊇ E + s, In: open}
= inf {Σλ∗(In− s)|[
(In− s) ⊇ E, In: open}
= inf {Σλ∗(In)|[
(In− s) ⊇ E, In: open}
= inf {Σλ∗(In)|[
In⊇ E, In: open} = λ∗(E) Therefore, λ(E + s) = λ(E).
Claim2 : E ∈ L ⇒ E + s ∈ L.
Let A ⊆ R and B = A − s.
λ∗(A ∩ (E + s)) + λ∗(A \ (E + s))
= λ∗((B + s) ∩ (E + s)) + λ∗((B + s) \ (E + s))
= λ∗((B ∩ E) + s) + λ∗((B \ E) + s)
= λ∗(B ∩ E) + λ∗(B \ E) = λ∗(B) = λ∗(A).
Therefore, E + s ∈ L.
Claim3 : λ(rE) = |r|λ(E), r ∈ R.
λ∗(rE) = inf {Σλ∗(In)|[
In ⊇ rE, In: open}
= inf {Σλ∗(rIn)|[
In⊇ rE, In: open}
= |r|inf {Σλ∗(In)|[
In ⊇ E, In: open} = |r|λ(E).
Therefore, λ(rE) = |r|λ(E).
Claim4 : E ∈ L ⇒ rE ∈ L.
If r = 0, then rE = {0} ∈ L.
For r 6= 0, let A ⊆ R and B = 1rA.
λ∗(A ∩ (rE)) + λ∗(A \ (rE))
= λ∗((rB) ∩ (rE)) + λ∗((rB) \ (rE))
= λ∗(r(B ∩ E)) + λ(r(B \ E))
= |r|λ∗(B ∩ E) + |r|λ(B \ E)
= |r|λ∗(B) = λ∗(A) Therefore rE ∈ L.
2. Let E be Lebesgue measurable with λ(E) < ∞. Then λ(E) = λ∗(E).
By the definition of λ∗, ∀ε > 0, ∃ open intervals {An} such that S∞
n=1An⊃ E and λ(
∞
[
n=1
An) ≤
∞
X
n=1
λ(An) < λ∗(E) + ε
2 = λ(E) + ε
2 (1)
Since λ(E) < ∞, P∞
n=1λ(An) converges.
Therefore with respect to this ε, ∃n0 ∈ N s.t.
∞
X
n=n0+1
λ(An) < ε
2. (2)
Let G =Sn0
n=1An, then because A0ns are open intervals , we can write G = SK0
k=1Ik, where Ik is a finite sequence of mutually disjoit open intervals. Thus, we have
G M E ⊂ (G \ E) ∪ (
∞
[
n=n0+1
An) ⊂ ((
∞
[
n=1
An) \ E) ∪ (
∞
[
n=n0+1
An)
Since E is Lebesgue measurable and by (1) (2), we have λ(G M E) < ε
2+ ε 2 = ε 3. Let B1 = [0.5, 0.6) and
Bn= {0.a1a2· · · an−15|ai ∈ I, ∀i = 1, . . . , n−1}, where I = 0, 1, 2, . . . , 9.
Let A1 = ([0, 1] \ B1), A2 = (A1\ B2), . . ., An= (An−1\ Bn).
Therefore,
A1 ⊇ A2 ⊇ · · · ⊇ An⊇ · · · .
⇒ A =T∞ n=1An.
Since λ(An) = (109)n, ⇒ λ(A1) = 109 < ∞. Thus we have
λ(A) = λ(
∞
\
n=1
An) = lim
n→∞λ(An) = lim
n→∞( 9
10)n = 0 4. There are two examples! Given 0 < ε < 1.
1o The first stage of the construction is to subdivide [0, 1] into three parts. The length of the middle part is 3ε, and the remaining two parts have equal length. We denote the middle part by A1. At the second stage, we subdivide each of the remaining two parts at the first stage into three parts. The length of each middle part is
ε
9. Then let A2 = A1∪(the two middle parts at the second stage).
Use similar algorithm to construct {Ak} for all k ∈ N and observe that Ak % and each Ak is measurable.
Set E =S∞ n=1An.
λE = lim
n→∞λAn= ε
3+ 2 ∗ ε
9+ 4 ∗ ε 33 + · · ·
= ε 3(1 + 2
3 + (2
3)2+ · · · )
= ε 3∗ 1
1 − 23 = ε Therefore λ(E) = ε.
Claim : E is dense in [0, 1].
Pf :∀x ∈ [0, 1]
case< 1 >: x ∈ E ⇒ ∀r > 0B(x, r) ∩ E 6= ∅.
case< 2 >: x ∈ ([0, 1] \ E) ⇒ x ∈ (T∞
n=1(An)c) ∩ [0, 1].
⇒ x ∈ Acn ∩ [0, 1], ∀n. We can estimate dist(x, An). For example,
dist(x, An) ≤ 1 − λ(An) 2n < 1
2n. Thus x is a limit point ofS An = E.
2o Let E0 = (0, ε), and E = E0∪ (Q ∩ [0, 1]).
Since Q is dense in [0, 1], E is also dense in [0, 1]. Moreover, ε = λ(0, ε) ≤ λ(E) ≤ λ(E0) + λ(Q) = ε + 0 = ε Therefore λ(E) = ε.