Chapter 2.6

Complex Measures

[13/0%]
Show
LaTeX
Problem 2.6.1

If μ\mu is a complex measure on a σ\sigma-algebra M\mathfrak {M}, and if E∈ME \in \mathfrak {M}, define

λ(E)=sup⁡∑∣μ(Ei)∣ \lambda (E) = \sup \sum \left|\mu \left(E_{i}\right)\right|

the supremum being taken over all finite partitions {Ei}\left\{ E_{i}\right\} of EE. Does it follow that λ=∣μ∣\lambda = \left|\mu \right|?

?
Problem 2.6.2

Prove that the example given at the end of Sec. 6.10 has the stated properties.

?
Problem 2.6.3

Prove that the vector space M(X)M(X) of all complex regular Borel measures on a locally compact Hausdorff space XX is a Banach space if ∥μ∥=∣μ∣(X)\left\| \mu \right\| = \left|\mu \right|(X).

?
Problem 2.6.4

Suppose 1≤p≤∞1 \leq p \leq \infty, and qq is the exponent conjugate to pp. Suppose μ\mu is a positive σ\sigma-finite measure and gg is a measurable function such that fg∈L1(μ)f g \in L^{1}(\mu ) for every f∈Lp(μ)f \in L^{p}(\mu ). Prove that then g∈Lq(μ)g \in L^{q}(\mu ).

?
Problem 2.6.5

Suppose XX consists of two points aa and bb; define μ({a})=1,μ({b})=μ(X)=∞\mu (\left\{ a\right\} ) = 1, \mu (\left\{ b\right\} ) = \mu (X) = \infty, and μ(∅)=0\mu (\varnothing ) = 0. Is it true, for this μ\mu, that L∞(μ)L^{\infty }(\mu ) is the dual space of L1(μ)L^{1}(\mu )?

?
Problem 2.6.6

Suppose 1<p<∞1 < p < \infty and prove that Lq(μ)L^{q}(\mu ) is the dual space of Lp(μ)L^{p}(\mu ) even if μ\mu is not σ\sigma-finite. (As usual, 1/p+1/q=11 / p+1 / q = 1.)

?
Problem 2.6.7

Suppose μ\mu is a complex Borel measure on [0,2π[0,2 \pi) (or on the unit circle TT), and define the Fourier coefficients of μ\mu by

μ^(n)=∫e−intdμ(t)(n=0,±1,±2,…) \hat{\mu }(n) = \int e^{-i n t} d \mu (t) \quad (n = 0, \pm 1, \pm 2, \ldots )

Assume that μ^(n)→0\hat{\mu }(n) \rightarrow 0 as n→+∞n \rightarrow +\infty and prove that then μ^(n)→0\hat{\mu }(n) \rightarrow 0 as n→−∞n \rightarrow -\infty. Hint: The assumption also holds with fdμf d \mu in place of dμd \mu if ff is any trigonometric polynomial, hence if ff is continuous, hence if ff is any bounded Borel function, hence if dμd \mu is replaced by d∣μ∣d\left|\mu \right|.

?
Problem 2.6.8

In the terminology of Exercise 7, find all μ\mu such that μ^\hat{\mu } is periodic, with period kk.

?
Remark.
?

This means that μ^(n+k)=μ^(n)\hat{\mu }(n+k) = \hat{\mu }(n) for all integers nn; of course, kk is also assumed to be an integer.

Problem 2.6.9

Suppose that {gn}\left\{ g_{n}\right\} is a sequence of positive continuous functions on I=[0,1]I = [0,1], that μ\mu is a positive Borel measure on II, and that

(i) lim⁡n→∞gn(x)=0\lim_{n \rightarrow \infty } g_{n}(x) = 0 a.e. [m][m],

(ii) ∫Igndm=1\int_{I} g_{n} d m =1 for all nn,

(iii) lim⁡n→∞∫Ifgndm=∫Ifdμ\lim_{n \rightarrow \infty } \int_{I} f g_{n} d m = \int_{I} f d \mu for every f∈C(I)f \in C(I).

Does it follow that μ⊥m\mu \perp m?

?
Problem 2.6.10

Let (X,M,μ)(X, \mathfrak {M}, \mu ) be a positive measure space. Call a set Φ⊂L1(μ)\Phi \subset L^{1}(\mu ) uniformly integrable if to each ϵ>0\epsilon > 0 corresponds a δ>0\delta > 0 such that

∣∫Efdμ∣<ϵ \left|\int _{E} f d \mu \right| < \epsilon

whenever f∈Φf \in \Phi and μ(E)<δ\mu (E) < \delta.

?
(a)

Prove that every finite subset of L1(μ)L^{1}(\mu ) is uniformly integrable.

(b)

Prove the following convergence theorem of Vitali:

 If (i) μ(X)<∞, (ii) {fn} is uniformly integrable, (iii) fn(x)→f(x) a.e. as n→∞, and (iv) ∣f(x)∣ \text{ If (i) } \mu (X) < \infty \text{, (ii) }\left\{ f_{n}\right\} \text{ is uniformly integrable, (iii) } f_{n}(x) \rightarrow f(x) \text{ a.e. as } n \rightarrow \infty \text{, and (iv) }\left|f(x)\right|

<∞< \infty a.e., then f∈L1(μ)f \in L^{1}(\mu ) and

lim⁡n→∞∫x∣fn−f∣dμ=0 \lim _{n \rightarrow \infty } \int _{x}\left|f_{n}-f\right| d \mu = 0

Suggestion: Use Egoroff's theorem.

(c)

Show that (b) fails if μ\mu is Lebesgue measure on (−∞,∞)(-\infty , \infty ), even if {∥fn∥1}\left\{ \left\| f_{n}\right\|_{1}\right\} is assumed to be bounded. Hypothesis (i) can therefore not be omitted in (b).

(d)

Show that hypothesis (iv) is redundant in (b) for some μ\mu (for instance, for Lebesgue measure on a bounded interval), but that there are finite measures for which the omission of (iv) would make (b) false.

(e)

Show that Vitali's theorem implies Lebesgue's dominated convergence theorem, for finite measure spaces. Construct an example in which Vitali's theorem applies although the hypotheses of Lebesgue's theorem do not hold.

(f)

Construct a sequence {fn}\left\{ f_{n}\right\}, say on [0,1][0,1], so that fn(x)→0f_{n}(x) \rightarrow 0 for every x,∫fn→0x, \int f_{n} \rightarrow 0, but {fn}\left\{ f_{n}\right\} is not uniformly integrable (with respect to Lebesgue measure).

(g)

However, the following converse of Vitali's theorem is true:

If μ(X)<∞,fn∈L1(μ)\mu (X) < \infty , f_{n} \in L^{1}(\mu ), and

lim⁡n→∞∫Efndμ \lim _{n \rightarrow \infty } \int _{E} f_{n} d \mu

exists for every E∈ME \in \mathfrak {M}, then {fn}\left\{ f_{n}\right\} is uniformly integrable.

Prove this by completing the following outline. Define ρ(A,B)=∫∣χA−χB∣dμ\rho (A, B) = \int \left|\chi_{A}-\chi_{B}\right| d \mu. Then (M,ρ)(\mathfrak {M}, \rho ) is a complete metric space (modulo sets of measure 0), and E→∫EfndμE \rightarrow \int_{E} f_{n} d \mu is continuous for each nn. If ϵ>0\epsilon > 0, there exist E0,δ,NE_{0}, \delta , N (Exercise 13, Chap. 5) so that

∣∫E(fn−fN)dμ∣<ϵ if ρ(E,E0)<δ,n>N(*) \left|\int _{E}\left(f_{n}-f_{N}\right) d \mu \right| < \epsilon \quad \text{ if } \rho \left(E, E_{0}\right) < \delta , \quad n > N \tag {*}

If μ(A)<δ,(∗)\mu (A) < \delta ,\left(^{*}\right) holds with B=E0−AB = E_{0}-A and C=E0∪AC = E_{0} \cup A in place of EE. Thus (*) holds with AA in place of EE and 2ϵ2 \epsilon in place of ϵ\epsilon. Now apply (a) to {f1,…,fN}\left\{ f_{1}, \ldots , f_{N}\right\} : There exists δ′>0\delta^{\prime } > 0 such that

∣∫Afndμ∣<3ϵ if μ(A)<δ′,n=1,2,3,… \left|\int _{A} f_{n} d \mu \right| < 3 \epsilon \quad \text{ if } \quad \mu (A) < \delta ^{\prime }, \quad n = 1,2,3, \ldots
Problem 2.6.11

Suppose μ\mu is a positive measure on X,μ(X)<∞,fn∈L1(μ)X, \mu (X) < \infty , f_{n} \in L^{1}(\mu ) for n=1,2,3,…,fn(x)→f(x)n = 1,2,3, \ldots , f_{n}(x) \rightarrow f(x) a.e., and there exists p>1p > 1 and C<∞C < \infty such that ∫x∣fn∣pdμ<C\int_{x}\left|f_{n}\right|^{p} d \mu < C for all nn. Prove that

lim⁡n→∞∫x∣f−fn∣dμ=0 \lim _{n \rightarrow \infty } \int _{x}\left|f-f_{n}\right| d \mu = 0

Hint: {fn}\left\{ f_{n}\right\} is uniformly integrable.

?
Problem 2.6.12

Let M\mathfrak {M} be the collection of all sets EE in the unit interval [0,1][0,1] such that either EE or its complement is at most countable. Let μ\mu be the counting measure on this σ\sigma-algebra M\mathfrak {M}. If g(x)=xg(x) = x for 0≤x≤10 \leq x \leq 1, show that gg is not M\mathfrak {M}-measurable, although the mapping

f→∑xf(x)=∫fgdμ f \rightarrow \sum x f(x) = \int f g d \mu

makes sense for every f∈L1(μ)f \in L^{1}(\mu ) and defines a bounded linear functional on L1(μ)L^{1}(\mu ). Thus (L1)∗≠L∞\left(L^{1}\right)^{*} \neq L^{\infty } in this situation.

?
Problem 2.6.13

Let L∞=L∞(m)L^{\infty } = L^{\infty }(m), where mm is Lebesgue measure on I=[0,1]I = [0,1]. Show that there is a bounded linear functional Λ≠0\Lambda \neq 0 on L∞L^{\infty } that is 0 on C(I)C(I), and that therefore there is no g∈L1(m)g \in L^{1}(m) that satisfies Λf=∫Ifgdm\Lambda f = \int_{I} f g d m for every f∈L∞f \in L^{\infty }. Thus (L∞)∗≠L1\left(L^{\infty }\right)^{*} \neq L^{1}.

?