TheoremBase

For a set D of finite ν-measure, the sets E for which x ↦ ν(ExE_x ∩ D) is measurable form a λ-system containing the measurable rectangles, so by the π-λ theorem they are all of F⊗G; cutting Y into disjoint pieces of finite measure and integrating term by term gives measurability, countable additivity and the value on rectangles.

Proof

Throughout, Ex∈GE_{x}\in\mathcal{G} for E∈F⊗GE\in\mathcal{F}\otimes\mathcal{G} and x∈Xx\in X (The Product σ-Algebra: Rectangles, Generators, Sections, Projections, Pairings and the Swap (iv)), and sections commute with the set operations (Rectangles, Sections and the Swap: Set Identities (ii), Rectangles, Sections and the Swap: Set Identities (i)): for E⊆X×YE\subseteq X\times Y, a sequence (En)(E_{n}) of subsets of X×YX\times Y, A∈FA\in\mathcal{F}, B∈GB\in\mathcal{G} and x∈Xx\in X,

((X×Y)∖E)x=Y∖Ex,(⋃n∈NEn)x=⋃n∈N(En)x,(A×B)x={B,x∈A,∅,x∉A,\bigl((X\times Y)\setminus E\bigr)_{x}=Y\setminus E_{x},\qquad\Bigl(\bigcup_{n\in\mathbb{N}}E_{n}\Bigr)_{x}=\bigcup_{n\in\mathbb{N}}(E_{n})_{x},\qquad(A\times B)_{x}=\begin{cases}B,&x\in A,\\\emptyset,&x\notin A,\end{cases}

and disjoint subsets of X×YX\times Y have disjoint sections.

Step 1: a set of finite measure. Fix D∈GD\in\mathcal{G} with ν(D)<∞\nu(D)<\infty. For E∈F⊗GE\in\mathcal{F}\otimes\mathcal{G} let gE:X→[0,∞]g_{E}:X\to[0,\infty] be the function

gE(x)=ν(Ex∩D);g_{E}(x)=\nu(E_{x}\cap D);

its values are real, as gE(x)≤ν(D)<∞g_{E}(x)\le\nu(D)<\infty (Basic Properties of a Measure (iii)). Let L\mathcal{L} be the set of those E∈F⊗GE\in\mathcal{F}\otimes\mathcal{G} with gE∈M+(X,F)g_{E}\in\mathcal{M}^{+}(X,\mathcal{F}). We show that L\mathcal{L} is a λ\lambda-system on X×YX\times Y containing the set Rect(F,G)\mathrm{Rect}(\mathcal{F},\mathcal{G}) of measurable rectangles.

Rectangles. Let A∈FA\in\mathcal{F} and B∈GB\in\mathcal{G}. By the section of A×BA\times B displayed above, and as ν(∅)=0\nu(\emptyset)=0,

gA×B(x)=ν(B∩D) 1A(x)(x∈X).g_{A\times B}(x)=\nu(B\cap D)\,\mathbf{1}_{A}(x)\qquad(x\in X).

Since ν(B∩D)\nu(B\cap D) is a nonnegative real number and 1A∈M+(X,F)\mathbf{1}_{A}\in\mathcal{M}^{+}(X,\mathcal{F}) with ∫X1A dμ=μ(A)\int_{X}\mathbf{1}_{A}\,d\mu=\mu(A) (Basic Properties of the Integral of Nonnegative Measurable Maps (ii)), Basic Properties of the Integral of Nonnegative Measurable Maps (iv) gives gA×B∈M+(X,F)g_{A\times B}\in\mathcal{M}^{+}(X,\mathcal{F}) and

∫XgA×B dμ=ν(B∩D) μ(A).(1)\int_{X}g_{A\times B}\,d\mu=\nu(B\cap D)\,\mu(A).\tag{1}

Thus Rect(F,G)⊆L\mathrm{Rect}(\mathcal{F},\mathcal{G})\subseteq\mathcal{L}; in particular X×Y∈LX\times Y\in\mathcal{L}, as X×Y∈Rect(F,G)X\times Y\in\mathrm{Rect}(\mathcal{F},\mathcal{G}) (The Product σ-Algebra: Rectangles, Generators, Sections, Projections, Pairings and the Swap (i)).

Complements. Let E∈LE\in\mathcal{L} and E′=(X×Y)∖EE'=(X\times Y)\setminus E, which lies in F⊗G\mathcal{F}\otimes\mathcal{G}. For x∈Xx\in X we have Ex′∩D=D∖(Ex∩D)E'_{x}\cap D=D\setminus(E_{x}\cap D), so by Basic Properties of a Measure (iv)

gE′(x)=ν(D)−gE(x),g_{E'}(x)=\nu(D)-g_{E}(x),

a difference in R\mathbb{R}. As a real-valued function, gEg_{E} is measurable (Measurable Maps: Generators, Composition, Continuous Maps, Restrictions, and Criteria for Real- and [0,∞]-Valued Maps (vii)), and so is the constant function ν(D)\nu(D) (Measurable Maps: Generators, Composition, Continuous Maps, Restrictions, and Criteria for Real- and [0,∞]-Valued Maps (iii)); hence their difference gE′g_{E'} is measurable (Arithmetic of Measurable Real- and [0,∞]-Valued Maps (i)), and since gE′≥0g_{E'}\ge0, it lies in M+(X,F)\mathcal{M}^{+}(X,\mathcal{F}) (Measurable Maps: Generators, Composition, Continuous Maps, Restrictions, and Criteria for Real- and [0,∞]-Valued Maps (vii)). So E′∈LE'\in\mathcal{L}.

Disjoint unions. Let (En)(E_{n}) be a pairwise disjoint sequence in L\mathcal{L} and E=⋃n∈NEnE=\bigcup_{n\in\mathbb{N}}E_{n}, which lies in F⊗G\mathcal{F}\otimes\mathcal{G}. For x∈Xx\in X the sets (En)x∩D(E_{n})_{x}\cap D are pairwise disjoint with union Ex∩DE_{x}\cap D, so countable additivity of ν\nu (Measure, Measure Space, and Probability Measure (i)) gives

gE(x)=∑n∈NgEn(x)(x∈X),g_{E}(x)=\sum_{n\in\mathbb{N}}g_{E_{n}}(x)\qquad(x\in X),

and gE∈M+(X,F)g_{E}\in\mathcal{M}^{+}(X,\mathcal{F}) by Linearity and Monotonicity of the Lebesgue Integral, and the Integrability Criterion (ii). So E∈LE\in\mathcal{L}.

As L⊆F⊗G⊆P(X×Y)\mathcal{L}\subseteq\mathcal{F}\otimes\mathcal{G}\subseteq\mathcal{P}(X\times Y), this shows that L\mathcal{L} is a λ\lambda-system on X×YX\times Y. Since Rect(F,G)\mathrm{Rect}(\mathcal{F},\mathcal{G}) is a π\pi-system (The Product σ-Algebra: Rectangles, Generators, Sections, Projections, Pairings and the Swap (i)) contained in L\mathcal{L}, the π\pi-λ\lambda theorem (Dynkin's Pi-Lambda Theorem (ii)) gives F⊗G=σ(Rect(F,G))⊆L\mathcal{F}\otimes\mathcal{G}=\sigma(\mathrm{Rect}(\mathcal{F},\mathcal{G}))\subseteq\mathcal{L} (Measurable Rectangles and the Product σ-Algebra (ii)). Hence gE∈M+(X,F)g_{E}\in\mathcal{M}^{+}(X,\mathcal{F}) for every E∈F⊗GE\in\mathcal{F}\otimes\mathcal{G}.

Step 2: a disjoint cover of YY by sets of finite measure. Since ν\nu is σ\sigma-finite, fix a sequence (Yk)(Y_{k}) in G\mathcal{G} with ⋃k∈NYk=Y\bigcup_{k\in\mathbb{N}}Y_{k}=Y and ν(Yk)<∞\nu(Y_{k})<\infty for every k∈Nk\in\mathbb{N} (Measure, Measure Space, and Probability Measure (iv)), and let (Dk)(D_{k}) be its disjointing sequence, the sequence (Bn)(B_{n}) of The Disjointing Sequence of a Sequence of Sets. Then Dk∈GD_{k}\in\mathcal{G} (The Disjointing Sequence of a Sequence of Sets (v)), the DkD_{k} are pairwise disjoint (The Disjointing Sequence of a Sequence of Sets (iii)) with union YY (The Disjointing Sequence of a Sequence of Sets (iv)), and Dk⊆YkD_{k}\subseteq Y_{k} (The Disjointing Sequence of a Sequence of Sets (ii)), so ν(Dk)≤ν(Yk)<∞\nu(D_{k})\le\nu(Y_{k})<\infty. For every C∈GC\in\mathcal{G} the sets C∩DkC\cap D_{k} are pairwise disjoint with union CC (Monotone Sequences of Sets, Unions of Traces, and Disjointness of Smaller Sets (iii), Monotone Sequences of Sets, Unions of Traces, and Disjointness of Smaller Sets (v)), so countable additivity gives

ν(C)=∑k∈Nν(C∩Dk).(2)\nu(C)=\sum_{k\in\mathbb{N}}\nu(C\cap D_{k}).\tag{2}

Step 3: proof of (i). Let E∈F⊗GE\in\mathcal{F}\otimes\mathcal{G}. By (2) with C=ExC=E_{x},

ν(Ex)=∑k∈Nν(Ex∩Dk)(x∈X).\nu(E_{x})=\sum_{k\in\mathbb{N}}\nu(E_{x}\cap D_{k})\qquad(x\in X).

By Step 1 with D=DkD=D_{k}, each function x↦ν(Ex∩Dk)x\mapsto\nu(E_{x}\cap D_{k}) lies in M+(X,F)\mathcal{M}^{+}(X,\mathcal{F}), so by Linearity and Monotonicity of the Lebesgue Integral, and the Integrability Criterion (ii) the function x↦ν(Ex)x\mapsto\nu(E_{x}) lies in M+(X,F)\mathcal{M}^{+}(X,\mathcal{F}), and

κ(E)=∑k∈N∫Xν(Ex∩Dk) dμ(x).(3)\kappa(E)=\sum_{k\in\mathbb{N}}\int_{X}\nu(E_{x}\cap D_{k})\,d\mu(x).\tag{3}

Step 4: proof of (ii). By (i), κ\kappa is a function from F⊗G\mathcal{F}\otimes\mathcal{G} to [0,∞][0,\infty]. As ∅x=∅\emptyset_{x}=\emptyset for every xx, κ(∅)=∫X0 dμ=0\kappa(\emptyset)=\int_{X}0\,d\mu=0 (Basic Properties of the Integral of Nonnegative Measurable Maps (v)). Let (En)(E_{n}) be a pairwise disjoint sequence in F⊗G\mathcal{F}\otimes\mathcal{G} with union EE. For x∈Xx\in X the sections (En)x(E_{n})_{x} are pairwise disjoint with union ExE_{x}, so ν(Ex)=∑n∈Nν((En)x)\nu(E_{x})=\sum_{n\in\mathbb{N}}\nu((E_{n})_{x}), and by Linearity and Monotonicity of the Lebesgue Integral, and the Integrability Criterion (ii)

κ(E)=∫X∑n∈Nν((En)x) dμ(x)=∑n∈N∫Xν((En)x) dμ(x)=∑n∈Nκ(En).\kappa(E)=\int_{X}\sum_{n\in\mathbb{N}}\nu\bigl((E_{n})_{x}\bigr)\,d\mu(x)=\sum_{n\in\mathbb{N}}\int_{X}\nu\bigl((E_{n})_{x}\bigr)\,d\mu(x)=\sum_{n\in\mathbb{N}}\kappa(E_{n}).

So κ\kappa is a measure on (X×Y,F⊗G)(X\times Y,\mathcal{F}\otimes\mathcal{G}).

Step 5: proof of (iii). Let A∈FA\in\mathcal{F} and B∈GB\in\mathcal{G}. By (3), then (1) with D=DkD=D_{k}, then Sums of Sequences in [0,∞]: Real Series, Linearity, Comparison, Rearrangement, Double Series and Monotone Convergence (ix) and (2) with C=BC=B,

κ(A×B)=∑k∈Nν(B∩Dk) μ(A)=μ(A)∑k∈Nν(B∩Dk)=μ(A) ν(B).\kappa(A\times B)=\sum_{k\in\mathbb{N}}\nu(B\cap D_{k})\,\mu(A)=\mu(A)\sum_{k\in\mathbb{N}}\nu(B\cap D_{k})=\mu(A)\,\nu(B).

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…