TheoremBase

The coordinate projections are continuous for the product metric, hence Borel, so every measurable rectangle is Borel. Conversely the projections are measurable for the product sigma-algebra, so by the pairing criterion for separable metric spaces the identity map is measurable from the product sigma-algebra to the Borel sigma-algebra.

Proof

Each result cited is universally quantified over the data in its own statement.

Write π1:Y×Z→Y\pi_{1}:Y\times Z\to Y and π2:Y×Z→Z\pi_{2}:Y\times Z\to Z for the coordinate projections, π1(y,z)=y\pi_{1}(y,z)=y and π2(y,z)=z\pi_{2}(y,z)=z, and write P\mathcal{P} for the family of measurable rectangles A×BA\times B with A∈B(Y)A\in\mathcal{B}(Y) and B∈B(Z)B\in\mathcal{B}(Z). By Product Sigma-Algebra, B(Y)⊗B(Z)=σ(P)\mathcal{B}(Y)\otimes\mathcal{B}(Z)=\sigma(\mathcal{P}), and by Generated Sigma-Algebra the family σ(P)\sigma(\mathcal{P}) is a σ\sigma-algebra on Y×ZY\times Z which contains P\mathcal{P} and is contained in every σ\sigma-algebra on Y×ZY\times Z containing P\mathcal{P}. Each of B(Y)\mathcal{B}(Y), B(Z)\mathcal{B}(Z) and B(Y×Z)\mathcal{B}(Y\times Z) is, by Borel Sigma-Algebra of a Metric Space, the σ\sigma-algebra generated by the open subsets of the corresponding metric space, hence a σ\sigma-algebra on YY, ZZ, respectively Y×ZY\times Z, by claim 2 of Intersections of Sigma-Algebras and Minimality of the Generated Sigma-Algebra.

Step 1: B(Y)⊗B(Z)⊆B(Y×Z)\mathcal{B}(Y)\otimes\mathcal{B}(Z)\subseteq\mathcal{B}(Y\times Z). We first show that π1\pi_{1} is continuous on Y×ZY\times Z in the sense of Continuous Map Between Metric Spaces. Let p=(y1,z1)∈Y×Zp=(y_{1},z_{1})\in Y\times Z and let ε\varepsilon be a positive real number; put δ=ε\delta=\varepsilon. Let q=(y2,z2)∈Y×Zq=(y_{2},z_{2})\in Y\times Z satisfy dY×Z(p,q)<δd_{Y\times Z}(p,q)<\delta. Since dY×Zd_{Y\times Z} is a metric by claim 1 of The Product Metric is a Metric, it is symmetric, so dY×Z(q,p)=dY×Z(p,q)d_{Y\times Z}(q,p)=d_{Y\times Z}(p,q); and claim 2 of The Product Metric is a Metric, applied to the points qq and pp in that order, gives

dY(π1(q),π1(p))=dY(y2,y1)≤dY×Z(q,p)=dY×Z(p,q)<ε.d_{Y}\bigl(\pi_{1}(q),\pi_{1}(p)\bigr)=d_{Y}(y_{2},y_{1})\le d_{Y\times Z}(q,p)=d_{Y\times Z}(p,q)<\varepsilon .

Thus π1\pi_{1} is continuous at every point of Y×ZY\times Z, and in the same way, using the second inequality of claim 2 of The Product Metric is a Metric, so is π2\pi_{2}. By claim 3 of Borel Measurability and Bounded Integration on a Metric Space, π1\pi_{1} is measurable with respect to B(Y×Z)\mathcal{B}(Y\times Z) and B(Y)\mathcal{B}(Y), and π2\pi_{2} is measurable with respect to B(Y×Z)\mathcal{B}(Y\times Z) and B(Z)\mathcal{B}(Z), in the sense of Measurable Function and Real-Valued Measurable Function. Now let A∈B(Y)A\in\mathcal{B}(Y) and B∈B(Z)B\in\mathcal{B}(Z). Then

A×B=π1−1(A)∩π2−1(B),A\times B=\pi_{1}^{-1}(A)\cap\pi_{2}^{-1}(B),

and both sets on the right belong to B(Y×Z)\mathcal{B}(Y\times Z), which is closed under finite intersections by Sigma-Algebra and Measurable Space. Hence P⊆B(Y×Z)\mathcal{P}\subseteq\mathcal{B}(Y\times Z), and since B(Y×Z)\mathcal{B}(Y\times Z) is a σ\sigma-algebra on Y×ZY\times Z, the minimality recorded in Generated Sigma-Algebra gives σ(P)⊆B(Y×Z)\sigma(\mathcal{P})\subseteq\mathcal{B}(Y\times Z).

Step 2: the case of an empty factor. If Y=∅Y=\varnothing or Z=∅Z=\varnothing, then Y×Z=∅Y\times Z=\varnothing, whose only subset is ∅\varnothing. Every σ\sigma-algebra on ∅\varnothing contains the whole set ∅\varnothing by property 1 of Sigma-Algebra and Measurable Space, so both B(Y×Z)\mathcal{B}(Y\times Z) and B(Y)⊗B(Z)\mathcal{B}(Y)\otimes\mathcal{B}(Z) are equal to {∅}\{\varnothing\}, and the claim holds. From now on assume that YY and ZZ are both nonempty.

Step 3: separable metric data. By Separable Metric Space there is a countable subset DY⊆YD_{Y}\subseteq Y that is dense in YY for the topology TdY\mathcal{T}_{d_{Y}} of open subsets of (Y,dY)(Y,d_{Y}). We check that DYD_{Y} is nonempty. Choose y∈Yy\in Y. The set YY is open in (Y,dY)(Y,d_{Y}) by Open Subset of a Metric Space, since every open ball in YY is a subset of YY; so Y∈TdYY\in\mathcal{T}_{d_{Y}}, y∈Yy\in Y and Y∩∅=∅Y\cap\varnothing=\varnothing, whence yy does not belong to the closure of ∅\varnothing by Closure of a Subset of a Topological Space. Since DYD_{Y} is dense, its closure is YY by Dense Subset of a Topological Space, and this closure contains yy. If DYD_{Y} were empty, its closure would be the closure of ∅\varnothing, which does not contain yy, a contradiction; hence DY≠∅D_{Y}\ne\varnothing. Thus ((Y,dY),DY)\bigl((Y,d_{Y}),D_{Y}\bigr) is a separable metric datum in the sense of Borel Sets and Measurable Maps in a Separable Metric Space, and in the same way there is a countable dense subset DZD_{Z} of ZZ with ((Z,dZ),DZ)\bigl((Z,d_{Z}),D_{Z}\bigr) a separable metric datum.

Step 4: B(Y×Z)⊆B(Y)⊗B(Z)\mathcal{B}(Y\times Z)\subseteq\mathcal{B}(Y)\otimes\mathcal{B}(Z). Put Ω=Y×Z\Omega=Y\times Z and F=σ(P)=B(Y)⊗B(Z)\mathcal{F}=\sigma(\mathcal{P})=\mathcal{B}(Y)\otimes\mathcal{B}(Z), so that (Ω,F)(\Omega,\mathcal{F}) is a measurable space. The map π1\pi_{1} is measurable with respect to F\mathcal{F} and B(Y)\mathcal{B}(Y): for A∈B(Y)A\in\mathcal{B}(Y) we have π1−1(A)=A×Z\pi_{1}^{-1}(A)=A\times Z, and Z∈B(Z)Z\in\mathcal{B}(Z) by property 1 of Sigma-Algebra and Measurable Space, so A×Z∈P⊆FA\times Z\in\mathcal{P}\subseteq\mathcal{F}. Likewise π2−1(B)=Y×B∈P⊆F\pi_{2}^{-1}(B)=Y\times B\in\mathcal{P}\subseteq\mathcal{F} for B∈B(Z)B\in\mathcal{B}(Z), so π2\pi_{2} is measurable with respect to F\mathcal{F} and B(Z)\mathcal{B}(Z). Apply claim 4 of Borel Sets and Measurable Maps in a Separable Metric Space to the measurable space (Ω,F)(\Omega,\mathcal{F}), the separable metric data ((Y,dY),DY)\bigl((Y,d_{Y}),D_{Y}\bigr) and ((Z,dZ),DZ)\bigl((Z,d_{Z}),D_{Z}\bigr) of Step 3, and the maps π1\pi_{1} and π2\pi_{2} in the roles of the two measurable maps of that claim: the map ω↦(π1(ω),π2(ω))\omega\mapsto(\pi_{1}(\omega),\pi_{2}(\omega)) from Ω\Omega to Y×ZY\times Z is measurable with respect to F\mathcal{F} and the Borel σ\sigma-algebra of (Y×Z,dY×Z)(Y\times Z,d_{Y\times Z}), which is B(Y×Z)\mathcal{B}(Y\times Z). This map is the identity map of Y×ZY\times Z, since (π1(y,z),π2(y,z))=(y,z)(\pi_{1}(y,z),\pi_{2}(y,z))=(y,z). Hence every E∈B(Y×Z)E\in\mathcal{B}(Y\times Z) satisfies E=id−1(E)∈FE=\mathrm{id}^{-1}(E)\in\mathcal{F}, that is, B(Y×Z)⊆B(Y)⊗B(Z)\mathcal{B}(Y\times Z)\subseteq\mathcal{B}(Y)\otimes\mathcal{B}(Z).

Steps 1 and 4 together give B(Y×Z)=B(Y)⊗B(Z)\mathcal{B}(Y\times Z)=\mathcal{B}(Y)\otimes\mathcal{B}(Z) when YY and ZZ are nonempty, and Step 2 covers the remaining case.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…