Sie sind auf Seite 1von 3

MATH 271 Homework 9 Solutions

1. Let D ⊆ R be a measurable set with m(D) < ∞ and f a measurable function defined on D that is finite a.e. on
D. Prove that for all  > 0, there exists a measurable subset A of D such that m(A) <  and f is bounded on
D \ A.

Proof : Let E = {x ∈ D : |f (x)| = ∞}. Then m(E) = 0 since f is finite a.e. on D.


For n ∈ N, put Dn = {x ∈ D : |f (x)| ≤ n}. Then Dn is measurable for all n ≥ 1 since |f | is measurable.
Moreover
+∞
[
D\E = Dn
n=1

Indeed, let x ∈ D \ E. Then |f (x)| is finite. Hence |f (x)| ≤ k for some k ∈ N. So x ∈ Dk ⊆ ∪n≥1 Dn . Let
x ∈ ∪n≥1 Dn . Then x ∈ Dk for some k ≥ 1. So |f (x)| ≤ k. Hence x ∈ / E.

Note that {Dn }n≥1 is increasing: if x ∈ Dn then |f (x)| ≤ n < n + 1 and so x ∈ Dn+1 .

By the continuity of the measure and excision, we get that

lim m(Dn ) = m(∪n≥1 Dn ) = m(D \ E) = m(D) − m(E) = m(D)


n→∞

Let  > 0. Since m(D) is finite, we get that there exists N ∈ N such that |m(Dn ) − m(D)| <  for all n ≥ N .
Since DN ⊆ D, we get that 0 ≤ m(D) − m(DN ) < . Put A = D \ DN . Then

m(A) = m(D \ DN ) = m(D) − m(DN ) < 

Note that D \ A = D \ (D \ DN ) = DN . So |f (x)| ≤ N for all x ∈ D \ A. Hence f is bounded on D \ A (by N ).


2

 1− x

if 0≤x≤n
2. For n ≥ 1, define fn : [0, +∞) → R : x → n

0 if x>n
(a) Put f (x) = lim fn (x) for all x ≥ 0. Find (with proof) f .
n→+∞
(b) Prove that there exists an  > 0 with the following property :
For all subsets A of [0, +∞) with m∗ (A) < ∞ and for all n ≥ 1, there exists x ∈ [0, +∞) \ A with
|fn (x) − f (x)| ≥ ε.
(c) Prove that Egoroff’s Theorem fails for this sequence of measurable functions.
This shows that the condition ‘m(D) is finite’ is needed in Egoroff’s Theorem.

Proof : (a) We will show that {fn (x)}n≥1 → 1 for all x ≥ 0. Indeed, pick x ≥ 0. Pick  > 0. Let N ∈ N with
n xo
N > max x, . Pick n ≥ N . Then n ≥ N > x and so

 x x x
|fn (x) − 1| = 1 − − 1 = ≤ <

n n N
1
(b) Put  = . Let A ⊆ [0, +∞) with m ∗ (A) < +∞. Let n ≥ 1. If (n, +∞) ⊆ A, then +∞ = m∗ ((n, +∞)) ≤
2
m∗ (A), a contradiction. So (n, +∞) 6⊆ A. Hence (n, +∞) \ A 6= ∅. Let x ∈ (n, +∞) \ A. Then
1
|fn (x) − f (x)| = |0 − 1| = 1 ≥ =
2

(c) Let δ = 1 (in fact, any δ > 0 will do the job). Let A be a measurable subset of [0, +∞) with m(A) < 1.
Suppose that the sequence {fn }n≥1 converges uniformly to f on [0, +∞) \ A. Let  be as in (b). Because of
uniform convergence, we have

∃N ∈ N : ∀n ≥ N, ∀x ∈ [0, +∞) \ A : |fn (x) − f (x)| < 

1
By (b), there exists x0 ∈ [0, +∞) \ A with |fN (x0 ) − f (x0 )| ≥ , a contradiction.
Hence the sequence {fn }n≥1 does not converge uniformly to f on [0, +∞) \ A. 2


 1 − nx 1
if 0≤x≤

3. For n ≥ 1, define fn : [0, 1] → R : x → n
1
0 if x>


n
(a) Put f (x) = 0 for all x ∈ [0, 1]. Prove that {fn }n≥1 converges to f a.e. on [0, 1].
(b) Prove that {fn }n≥1 does not converge uniformly to f a.e. on [0, 1] (so you have to show that {fn }n≥1 does
not converge uniformly to f on [0, 1] \ A for any A ⊆ [0, 1] with m(A) = 0).
This shows that we cannot ‘improve’ Egoroff’s Theorem to uniform convergence a.e.

Proof : (a) We will prove that {fn (x)}n≥1 → 0 for all x ∈ (0, 1]. Indeed, pick x ∈ (0, 1]. Pick  > 0. Let N ∈ N
1 1 1
with N > . Pick n ≥ N . Then x > ≥ and so |fn (x) − 0| = |0 − 0| = 0 < . So {fn }n≥1 → f on (0, 1].
x N n
Since m({0}) = 0, we get that {fn }n≥1 → f a.e. on [0, 1].

(b) Let A ⊆ [0, 1] with m(A) = 0. We have to prove :


¬(∀ > 0 : ∃N ∈ N : ∀n ≥ N, ∀x ∈ [0, 1] \ A : |fn (x) − f (0)| < )
So
∃ > 0 : ∀N ∈ N : ∃n ≥ N, ∃x ∈ [0, 1] \ A : |fn (x) − 0| ≥ 
   
1 1 1 1
Put  = . Pick N ∈ N. Put n = N . If 0, ⊆ A, then = m 0, ≤ m(A), a contradiction.
 2   2N   2N 2N
1 1 1 1 1 1
So 0, 6⊆ A. Hence 0, \ A 6= ∅. Let x ∈ 0, \ A. Then 0 ≤ x ≤ < and 0 ≤ N x ≤ . So
2N 2N 2N 2N N 2
1
|fn (x) − 0| = |1 − N x| ≥ = . 2
2

4. Prove there exists a measurable subset D ⊆ [0, 1] such χD ◦ ψ −1 is non-measurable.

This shows that the composition of two measurable functions does not need to be measurable (in fact, measurable
after continuous does not need to be measurable).

Proof : By Proposition 3.20(d), there exists a subset D of the Cantor set C such that ψ(D) is non-measurable.
Note that D is measurable (since m∗ (D) ≤ m(C) = 0) and that D ⊆ [0, 1]. For x ∈ [0, 2], we have that

1 if x ∈ ψ(D)
(χD ◦ ψ −1 )(x) = χD ψ −1 (x) =

0 if x ∈/ ψ(D)
So
{x ∈ [0, 2] : (χD ◦ ψ −1 )(x) = 1} = ψ(D)
Since ψ(D) is not measurable, it follows from Proposition 3.3 that χD ◦ ψ −1 is not measurable. 2

5. Let F ⊆ R be closed and f : F → R continuous. Prove there exists a continuous function g : R → R with f = g
on F .

Proof : We may assume that ∅ = 6 F 6= R. Then Fe is open and hence the countable union of disjoint open
intervals, say Fe = ∪·n≥1 (an , bn ). Note that it is possible that an = −∞ or bn = +∞ for some n ≥ 1 but not
both since F 6= ∅. Moreover, if an ∈ R then an ∈ F (indeed, if an ∈ Fe, then an ∈ (am , bm ) for some m 6= n; so
(an − r, an + r) ⊆ (am , bm ) for some r > 0; thus (an , bn ) ∩ (am , bm ) 6= ∅, a contradiction). Similarly, if bn ∈ R
then bn ∈ F .

We define a function g : R → R as follows:

2
• on F : g(x) = f (x) for all x ∈ F .
• on (an , bn ) if an = −∞: g(x) = f (bn ) for all x ∈ (−∞, bn ).
• on (an , bn ) if bn = +∞: g(x) = f (an ) for all x ∈ (an , +∞).
f (bn ) − f (an )
• on (an , bn ) if an , bn ∈ R: g(x) = f (an ) + (x − an ) for all x ∈ (an , bn ).
bn − an
We prove that g is right-continuous on R. The proof for left-continuous is similar.
Let x0 ∈ R. We consider three cases:
CASE 1: x0 ∈ Fe
Then x0 ∈ (an , bn ) for some n ≥ 1. Let  > 0. Since g is linear on (an , bn ), we get

∃δ1 > 0 : ∀x ∈ (an , bn ) : |x − x0 | < δ1 =⇒ |g(x) − g(x0 )| < 

Since (an , bn ) is open, we have that (x0 − δ2 , x0 + δ2 ) ⊆ (an , bn ) for some δ2 > 0. Put δ = min{δ1 , δ2 }. Let x ∈ R
with |x − x0 | < δ. Then |x − x0 | < δ2 and so x ∈ (an , bn ). Since |x − x0 | < δ1 and x ∈ (an , bn ), we get that
|g(x) − g(x0 )| < . So g is continuous at x0 .

CASE 2: x0 ∈ F and (x0 , x0 + r) ⊆ Fe for some r > 0.


Then there exists n ≥ 1 such that x0 = an and (x0 , x0 + r) ⊆ (an , bn ). Indeed, since (x0 , x0 + r) ⊆ Fe, we have
that (x0 , x0 + r) ∩ (an , bn ) 6= ∅ for some n ≥ 1. Let t ∈ (x0 , x0 + r) ∩ (an , bn ). Then x0 < t < x0 + r and
an < t < bn . If x0 < an then x0 < an < t < x0 + r and so an ∈ (x0 , x0 + r) ⊆ Fe, a contradiction since an ∈ F .
Hence an ≤ x0 . If an < x0 then an < x0 < t < bn and so x0 ∈ (an , bn ) ⊆ Fe, a contradiction since x0 ∈ F . Thus
x0 = an . If bn < x0 + r then x0 = an < bn < x0 + r and so bn ∈ (x0 , x0 + r) ⊆ Fe, a contradiction since bn ∈ F .
Hence x0 + r ≤ bn and we get that (x0 , x0 + r) = (an , x0 + r) ⊆ (an , bn ).

Let  > 0. Since g is linear on [an , bn ), we get

∃δ1 > 0 : ∀x ∈ [an , bn ) : |x − an | < δ1 =⇒ |g(x) − g(an )| < 

Since (an , bn ) is open, we have that (an , an + δ2 ) ⊆ (an , bn ) for some δ2 > 0. Put δ = min{δ1 , δ2 }. Let x ∈ R
with 0 ≤ x − an < δ. Then 0 ≤ x − x0 < δ2 and so x ∈ [an , bn ). Since |x − an | < δ2 and x ∈ [an , bn ), we get that
|g(x) − g(an )| < . So g is right-continuous at x0 = an .

CASE 3: x0 ∈ F and (x0 , x0 + r) 6⊆ Fe for all r > 0.


Let  > 0. Since f is continuous at x0 , we have

∃δ1 > 0 : ∀x ∈ F : |x − x0 | < δ1 =⇒ |f (x) − f (x0 )| < 

By assumption, (x0 , x0 + δ1 ) ∩ F 6= ∅. Pick x1 ∈ (x0 , x0 + δ1 ) ∩ F 6= ∅. Put δ = x1 − x0 . Let x ∈ R with


0 ≤ x − x0 < δ. If x ∈ F then |g(x) − g(x0 )| = |f (x) − f (x0 )| <  since 0 ≤ x − x0 < δ < δ1 . So we may assume
that x ∈ Fe. Then x ∈ (an , bn ) for some n ≥ 1. Since x0 , x1 ∈ F , we must have that x0 ≤ an < x < bn ≤ x1 . By
construction of g on (an , bn ), we have that g(x) is between f (an ) and f (bn ). So g(x) = λf (an ) + (1 − λ)f (bn )
for some 0 ≤ λ ≤ 1. Note that f (x0 ) = λf (x0 ) + (1 − λ)f (x0 ). Hence

|g(x) − g(x0 )| = |(λf (an ) + (1 − λ)f (bn )) − (λf (x0 ) + (1 − λ)f (x0 ))|
≤ λ|f (an ) − f (x0 )| + (1 − λ)|f (bn ) − f (x0 )|
< λ + (1 − λ)
= 

since an , bn ∈ F with |an − x0 |, |bn − x0 | ≤ δ < δ1 . So g is right-continuous at x0 . 2

Das könnte Ihnen auch gefallen