TheoremBase

The upper bound on the dual norm comes from the bound for integrals of bounded functions; the lower bound tests against a difference of two Urysohn functions with disjoint supports concentrated near compact parts of a Hahn decomposition. Existence decomposes a functional into positive parts and applies the Riesz-Markov theorem to each; uniqueness holds because finite signed Borel measures are determined by their integrals of compactly supported continuous functions.

Proof

Throughout, continuous real functions on XX are Borel measurable, and every f∈Cc(X)f\in C_{c}(X) satisfies ∣f(x)∣≤∥f∥u|f(x)|\le\|f\|_{u} for every x∈Xx\in X; and ν↦Φ(ν)\nu\mapsto\Phi(\nu) is a function on M(X)M(X), the condition r=∫Xf dνr=\int_{X}f\,d\nu being predicative in (ν,f,r)(\nu,f,r) by The Simple, Nonnegative, Integrable and Signed Integrals Are Predicative in the Measure and the Function, at Any Variable Positions (iv) (and Maps and Relations Given by Formulas (i)).

The functional and the upper bound. Let ν∈M(X)\nu\in M(X). For f∈Cc(X)f\in C_{c}(X), Properties of the Integral with Respect to a Finite Signed Measure: Linearity, Bounds, Bounded Functions, Linearity in the Measure, Indicators and Finite Measures (iii), with C=∥f∥uC=\|f\|_{u}, shows that ff is integrable with respect to ν\nu and gives the bound below. The function Φ(ν)\Phi(\nu) is linear by Properties of the Integral with Respect to a Finite Signed Measure: Linearity, Bounds, Bounded Functions, Linearity in the Measure, Indicators and Finite Measures (i), and

∣Φ(ν)(f)∣≤∥ν∥ ∥f∥u(f∈Cc(X)).|\Phi(\nu)(f)|\le\|\nu\|\,\|f\|_{u}\qquad(f\in C_{c}(X)).

So the nonnegative number ∥ν∥\|\nu\| is a bound for Φ(ν)\Phi(\nu); hence Φ(ν)\Phi(\nu) is bounded, Φ(ν)∈Cc(X)∗\Phi(\nu)\in C_{c}(X)^{*}, and ∥ν∥\|\nu\| lies in the set of nonnegative bounds of Φ(ν)\Phi(\nu), whose infimum is ∥Φ(ν)∥∗\|\Phi(\nu)\|_{*} by The Dual Norm (i). Thus ∥Φ(ν)∥∗≤∥ν∥\|\Phi(\nu)\|_{*}\le\|\nu\|.

Compact parts of a Hahn decomposition. Let PP be a Hahn set for ν\nu, which exists by Hahn and Jordan Decompositions of a Finite Signed Measure (i) (using the axiom of dependent choice). By Properties of the Total Variation: a Finite Measure, the Jordan Decomposition, Linear Combinations, the Total Variation Norm, and Integrability (ii) with A=XA=X,

∥ν∥=∣ν∣(X)=ν(P)−ν(X∖P),\|\nu\|=|\nu|(X)=\nu(P)-\nu(X\setminus P),

and ∣ν∣|\nu| is a finite Borel measure by Properties of the Total Variation: a Finite Measure, the Jordan Decomposition, Linear Combinations, the Total Variation Norm, and Integrability (i). Let ε>0\varepsilon>0. By Finite Borel Measures on a Locally Compact Separable Metric Space Are Radon (i), applied to ∣ν∣|\nu| and the Borel sets PP and X∖PX\setminus P, there are compact sets K⊆PK\subseteq P and L⊆X∖PL\subseteq X\setminus P with ∣ν∣(K)>∣ν∣(P)−ε|\nu|(K)>|\nu|(P)-\varepsilon and ∣ν∣(L)>∣ν∣(X∖P)−ε|\nu|(L)>|\nu|(X\setminus P)-\varepsilon, that is,

∣ν∣(P∖K)<ε,∣ν∣((X∖P)∖L)<ε.|\nu|(P\setminus K)<\varepsilon,\qquad|\nu|\bigl((X\setminus P)\setminus L\bigr)<\varepsilon .

The sets KK and LL are disjoint.

A test function. The compact set LL is closed, so X∖LX\setminus L is open and contains KK, and Thickenings, Neighbourhoods with Compact Closure, Urysohn Functions and Partitions of Unity in a Locally Compact Metric Space (iii) gives f1∈Cc(X)f_{1}\in C_{c}(X) with values in [0,1][0,1], f1=1f_{1}=1 on KK and supp⁡f1⊆X∖L\operatorname{supp}f_{1}\subseteq X\setminus L. The compact set supp⁡f1\operatorname{supp}f_{1} is closed, so X∖supp⁡f1X\setminus\operatorname{supp}f_{1} is open and contains LL, and the same lemma gives f2∈Cc(X)f_{2}\in C_{c}(X) with values in [0,1][0,1], f2=1f_{2}=1 on LL and supp⁡f2⊆X∖supp⁡f1\operatorname{supp}f_{2}\subseteq X\setminus\operatorname{supp}f_{1}. Let g=f1−f2∈Cc(X)g=f_{1}-f_{2}\in C_{c}(X). Since the supports of f1f_{1} and f2f_{2} are disjoint, at every point at least one of f1,f2f_{1},f_{2} vanishes, by Continuous Functions with Compact Support: the Zero Function, Closure under Operations, Vanishing off the Support, Existence of the Maximum Modulus, and Uniform Continuity (iii). Hence ∣g(x)∣≤1|g(x)|\le1 for every x∈Xx\in X, so ∥g∥u≤1\|g\|_{u}\le1; and g=1g=1 on KK, where f1=1≠0f_{1}=1\ne0, and g=−1g=-1 on LL, where f2=1≠0f_{2}=1\ne0.

The lower bound. Let h=1P−1X∖Ph=\mathbf{1}_{P}-\mathbf{1}_{X\setminus P}. By Properties of the Integral with Respect to a Finite Signed Measure: Linearity, Bounds, Bounded Functions, Linearity in the Measure, Indicators and Finite Measures (vi) and Properties of the Integral with Respect to a Finite Signed Measure: Linearity, Bounds, Bounded Functions, Linearity in the Measure, Indicators and Finite Measures (i), hh is integrable with respect to ν\nu and

∫Xh dν=ν(P)−ν(X∖P)=∥ν∥.\int_{X}h\,d\nu=\nu(P)-\nu(X\setminus P)=\|\nu\|.

The function g−hg-h vanishes on K∪LK\cup L, and ∣g−h∣≤2|g-h|\le2 everywhere, so

∣g(x)−h(x)∣≤2⋅1P∖K(x)+2⋅1(X∖P)∖L(x)(x∈X).|g(x)-h(x)|\le2\cdot\mathbf{1}_{P\setminus K}(x)+2\cdot\mathbf{1}_{(X\setminus P)\setminus L}(x)\qquad(x\in X).

By Properties of the Integral with Respect to a Finite Signed Measure: Linearity, Bounds, Bounded Functions, Linearity in the Measure, Indicators and Finite Measures (i) and Properties of the Integral with Respect to a Finite Signed Measure: Linearity, Bounds, Bounded Functions, Linearity in the Measure, Indicators and Finite Measures (ii), and monotonicity of the integral with respect to the measure ∣ν∣|\nu|,

∣∫Xg dν−∥ν∥∣=∣∫X(g−h) dν∣≤∫X∣g−h∣ d∣ν∣≤2 ∣ν∣(P∖K)+2 ∣ν∣((X∖P)∖L)<4ε.\Bigl|\int_{X}g\,d\nu-\|\nu\|\Bigr|=\Bigl|\int_{X}(g-h)\,d\nu\Bigr|\le\int_{X}|g-h|\,d|\nu|\le2\,|\nu|(P\setminus K)+2\,|\nu|\bigl((X\setminus P)\setminus L\bigr)<4\varepsilon .

Since the dual norm is a bound for Φ(ν)\Phi(\nu),

∥ν∥−4ε<∫Xg dν=Φ(ν)(g)≤∥Φ(ν)∥∗ ∥g∥u≤∥Φ(ν)∥∗.\|\nu\|-4\varepsilon<\int_{X}g\,d\nu=\Phi(\nu)(g)\le\|\Phi(\nu)\|_{*}\,\|g\|_{u}\le\|\Phi(\nu)\|_{*}.

As ε>0\varepsilon>0 was arbitrary, ∥ν∥≤∥Φ(ν)∥∗\|\nu\|\le\|\Phi(\nu)\|_{*}. With the upper bound this proves (i).

Linearity. Let ν,ν′∈M(X)\nu,\nu'\in M(X) and c∈Rc\in\mathbb{R}; then ν+ν′\nu+\nu' and cνc\nu lie in M(X)M(X) by Properties of the Total Variation: a Finite Measure, the Jordan Decomposition, Linear Combinations, the Total Variation Norm, and Integrability (v). For f∈Cc(X)f\in C_{c}(X), which is measurable and bounded by ∥f∥u\|f\|_{u}, Properties of the Integral with Respect to a Finite Signed Measure: Linearity, Bounds, Bounded Functions, Linearity in the Measure, Indicators and Finite Measures (v) gives

Φ(ν+ν′)(f)=Φ(ν)(f)+Φ(ν′)(f),Φ(cν)(f)=c Φ(ν)(f).\Phi(\nu+\nu')(f)=\Phi(\nu)(f)+\Phi(\nu')(f),\qquad\Phi(c\nu)(f)=c\,\Phi(\nu)(f).

Since the sum and the scalar multiples in Cc(X)∗C_{c}(X)^{*} are pointwise (The Dual of a Real Normed Space Is a Real Banach Space; Weak-Star Limits (i)), this is (ii).

Existence of a representing measure. Let Λ∈Cc(X)∗\Lambda\in C_{c}(X)^{*}. By Jordan Decomposition of a Bounded Linear Functional on the Continuous Functions with Compact Support (i) there are positive linear functionals Λ+,Λ−∈Cc(X)∗\Lambda^{+},\Lambda^{-}\in C_{c}(X)^{*} with Λf=Λ+f−Λ−f\Lambda f=\Lambda^{+}f-\Lambda^{-}f for every f∈Cc(X)f\in C_{c}(X). By The Riesz–Markov Representation Theorem on a Locally Compact Metric Space (i) there are Radon measures μ+\mu^{+} and μ−\mu^{-} on XX with

Λ+f=∫Xf dμ+,Λ−f=∫Xf dμ−(f∈Cc(X)),\Lambda^{+}f=\int_{X}f\,d\mu^{+},\qquad\Lambda^{-}f=\int_{X}f\,d\mu^{-}\qquad(f\in C_{c}(X)),

and by The Riesz–Markov Representation Theorem on a Locally Compact Metric Space (iii), μ±(X)=∥Λ±∥∗<∞\mu^{\pm}(X)=\|\Lambda^{\pm}\|_{*}<\infty (μ±\mu^{\pm} is the measure of The Riesz–Markov Representation Theorem on a Locally Compact Metric Space (i) for Λ±\Lambda^{\pm}, unique by The Riesz–Markov Representation Theorem on a Locally Compact Metric Space (ii)), so μ+\mu^{+} and μ−\mu^{-} are finite Borel measures. Their difference ν\nu, ν(B)=μ+(B)−μ−(B)\nu(B)=\mu^{+}(B)-\mu^{-}(B) for B∈B(X)B\in\mathcal{B}(X), is a finite signed Borel measure, so ν∈M(X)\nu\in M(X). For f∈Cc(X)f\in C_{c}(X), measurable and bounded by ∥f∥u\|f\|_{u}, Properties of the Integral with Respect to a Finite Signed Measure: Linearity, Bounds, Bounded Functions, Linearity in the Measure, Indicators and Finite Measures (iv) gives

∫Xf dν=∫Xf dμ+−∫Xf dμ−=Λ+f−Λ−f=Λf.\int_{X}f\,d\nu=\int_{X}f\,d\mu^{+}-\int_{X}f\,d\mu^{-}=\Lambda^{+}f-\Lambda^{-}f=\Lambda f .

Thus Λ=Φ(ν)\Lambda=\Phi(\nu).

Uniqueness. Let ν,ν′∈M(X)\nu,\nu'\in M(X) with Φ(ν)=Λ=Φ(ν′)\Phi(\nu)=\Lambda=\Phi(\nu'). Then ∫Xf dν=∫Xf dν′\int_{X}f\,d\nu=\int_{X}f\,d\nu' for every f∈Cc(X)f\in C_{c}(X), and since XX is locally compact and separable, Radon Measures and Finite Signed Borel Measures Are Determined by Their Integrals of Compactly Supported Continuous Functions (iii) gives ν(B)=ν′(B)\nu(B)=\nu'(B) for every B∈B(X)B\in\mathcal{B}(X), that is, ν=ν′\nu=\nu'. This proves (iii).

Isomorphism. By (iii), every Λ∈Cc(X)∗\Lambda\in C_{c}(X)^{*} is Φ(ν)\Phi(\nu) for exactly one ν∈M(X)\nu\in M(X), so Φ\Phi is a bijection of M(X)M(X) onto Cc(X)∗C_{c}(X)^{*}. Let ν,ν′∈M(X)\nu,\nu'\in M(X). The map Φ\Phi is linear ((ii)), so Φ(ν)−Φ(ν′)=Φ(ν−ν′)\Phi(\nu)-\Phi(\nu')=\Phi(\nu-\nu'), and by (i),

∥Φ(ν)−Φ(ν′)∥∗=∥Φ(ν−ν′)∥∗=∥ν−ν′∥.\|\Phi(\nu)-\Phi(\nu')\|_{*}=\|\Phi(\nu-\nu')\|_{*}=\|\nu-\nu'\| .

This proves (iv).

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…