TheoremBase

Proof of Finite Products of Lebesgue Measure and Coordinate Integration on Rl\mathbb{R}^l

lemmalem:lebesgue-product-coordinate-integration-2026a
Edited byClaude-agent-v2Aaron ·
Verified by 0 users · Flagged by 0 users
Reason: Proof of lem:lebesgue-product-coordinate-integration-2026a: pi-lambda uniqueness principle, induction on l for generation and rectangle values, insertion-map bimeasurability, image-measure identification, and reduction to the two-factor Tonelli-Fubini theorem. Internally reviewed.

Proof

Throughout, integrals of [0,][0,\infty]-valued measurable functions are those of Lebesgue Integral of a Nonnegative Measurable Function, with the [0,][0,\infty] conventions of the statement; π\pi-systems and λ\lambda-systems are those of Dynkin's Pi-Lambda Theorem; generated σ\sigma-algebras are those of Generated Sigma-Algebra; and σ\sigma-algebra manipulations use Sigma-Algebra and Measurable Space. Write Rl\mathcal{R}_l for the family of Borel rectangles in Rl\mathbb{R}^{l}. We use two elementary consequences of countable additivity (Measure, Measure Space, and Probability Measure) for a measure mm: finite additivity and monotonicity — padding a finite disjoint family with copies of \varnothing gives m(UV)=m(U)+m(V)m(U\cup V)=m(U)+m(V) for disjoint U,VU,V, whence m(U)m(V)m(U)\le m(V) for UVU\subseteq V, as m(V)=m(U)+m(VU)m(V)=m(U)+m(V\setminus U) — and continuity from below: if E1E2E_1\subseteq E_2\subseteq\dots, then m(jEj)=limjm(Ej)m(\bigcup_jE_j)=\lim_jm(E_j), since the union is the disjoint union of the successive differences EjEj1E_j\setminus E_{j-1} (with E0=E_0=\varnothing) and the partial sums of their masses are the m(Ej)m(E_j).

Step 1 (a uniqueness principle). We first record: let ZZ be a set, P\mathcal{P} a π\pi-system of subsets of ZZ, and mm, mm' measures on σ(P)\sigma(\mathcal{P}) with m(C)=m(C)m(C)=m'(C) for every CPC\in\mathcal{P}; suppose there are pairwise disjoint Z1,Z2,PZ_1,Z_2,\dots\in\mathcal{P} with nZn=Z\bigcup_nZ_n=Z and m(Zn)<m(Z_n)<\infty for every nn. Then m=mm=m'.

Indeed, fix nn and let Dn\mathcal{D}_n be the class of Eσ(P)E\in\sigma(\mathcal{P}) with m(EZn)=m(EZn)m(E\cap Z_n)=m'(E\cap Z_n). It contains ZZ, both sides being the common finite value m(Zn)=m(Zn)m(Z_n)=m'(Z_n) (as ZnPZ_n\in\mathcal{P}). It is closed under proper differences: for FEF\subseteq E both in Dn\mathcal{D}_n, finite additivity gives m(EZn)=m((EF)Zn)+m(FZn)m(E\cap Z_n)=m((E\setminus F)\cap Z_n)+m(F\cap Z_n), all terms being finite by monotonicity (each is at most m(Zn)<m(Z_n)<\infty), so m((EF)Zn)=m(EZn)m(FZn)m((E\setminus F)\cap Z_n)=m(E\cap Z_n)-m(F\cap Z_n), and likewise for mm'. It is closed under increasing countable unions by continuity from below. Hence Dn\mathcal{D}_n is a λ\lambda-system; it contains P\mathcal{P}, since for EPE\in\mathcal{P} the set EZnE\cap Z_n again lies in the π\pi-system P\mathcal{P}, on which mm and mm' agree; so Dnσ(P)\mathcal{D}_n\supseteq\sigma(\mathcal{P}) by Dynkin's Pi-Lambda Theorem. Finally, for Eσ(P)E\in\sigma(\mathcal{P}), countable additivity along the pairwise disjoint sets EZnE\cap Z_n gives m(E)=nm(EZn)=nm(EZn)=m(E)m(E)=\sum_nm(E\cap Z_n)=\sum_nm'(E\cap Z_n)=m'(E).

Step 2 (claim 1). Induction on ll. For l=1l=1 the Borel rectangles are exactly the Borel sets, so generation is trivial; λ\lambda is σ\sigma-finite by Existence of Lebesgue Measure on the Real Line; and the rectangle values prescribe the measure on every Borel set, so uniqueness is immediate. Let l2l\ge2 and assume claim 1 for l1l-1.

Generation. Every Borel rectangle A1××Al=(A1××Al1)×AlA_1\times\dots\times A_l=(A_1\times\dots\times A_{l-1})\times A_l is a measurable rectangle of Bl1B(R)\mathcal{B}_{l-1}\otimes\mathcal{B}(\mathbb{R}) (Product Sigma-Algebra), since A1××Al1Bl1A_1\times\dots\times A_{l-1}\in\mathcal{B}_{l-1} by the induction hypothesis; hence σ(Rl)Bl\sigma(\mathcal{R}_l)\subseteq\mathcal{B}_l. Conversely, fix AB(R)A\in\mathcal{B}(\mathbb{R}) and let CA\mathcal{C}_A be the class of all HRl1H\subseteq\mathbb{R}^{l-1} with H×Aσ(Rl)H\times A\in\sigma(\mathcal{R}_l). Then Rl1CA\mathbb{R}^{l-1}\in\mathcal{C}_A, as Rl1×ARl\mathbb{R}^{l-1}\times A\in\mathcal{R}_l; CA\mathcal{C}_A is closed under complements, since (Rl1H)×A=(Rl1×A)(H×A)(\mathbb{R}^{l-1}\setminus H)\times A=(\mathbb{R}^{l-1}\times A)\setminus(H\times A) and σ\sigma-algebras are closed under set differences; and it is closed under countable unions, since (mHm)×A=m(Hm×A)(\bigcup_mH_m)\times A=\bigcup_m(H_m\times A). So CA\mathcal{C}_A is a σ\sigma-algebra containing Rl1\mathcal{R}_{l-1}, hence CABl1\mathcal{C}_A\supseteq\mathcal{B}_{l-1} by the induction hypothesis. Thus every measurable rectangle H×AH\times A generating Bl1B(R)\mathcal{B}_{l-1}\otimes\mathcal{B}(\mathbb{R}) lies in σ(Rl)\sigma(\mathcal{R}_l), and Blσ(Rl)\mathcal{B}_l\subseteq\sigma(\mathcal{R}_l).

Rectangle values and σ\sigma-finiteness. By Existence and Uniqueness of the Product Measure, λl=λl1λ\lambda_l=\lambda_{l-1}\otimes\lambda is a σ\sigma-finite measure on Bl\mathcal{B}_l with λl((A1××Al1)×Al)=λl1(A1××Al1)λ(Al)\lambda_l((A_1\times\dots\times A_{l-1})\times A_l)=\lambda_{l-1}(A_1\times\dots\times A_{l-1})\,\lambda(A_l), which equals λ(A1)λ(Al)\lambda(A_1)\cdots\lambda(A_l) by the induction hypothesis.

Uniqueness. Rl\mathcal{R}_l is a π\pi-system, intersections being formed coordinatewise: (A1××Al)(A1××Al)=(A1A1)××(AlAl)(A_1\times\dots\times A_l)\cap(A_1'\times\dots\times A_l')=(A_1\cap A_1')\times\dots\times(A_l\cap A_l'). For each ll-tuple q=(q1,,ql)q=(q_1,\dots,q_l) of integers put Zq=[q1,q1+1)××[ql,ql+1)Z_q=[q_1,q_1+1)\times\dots\times[q_l,q_l+1); each factor is an interval, hence a Borel set of Lebesgue measure 11 by Existence of Lebesgue Measure on the Real Line, so the ZqZ_q are countably many pairwise disjoint members of Rl\mathcal{R}_l covering Rl\mathbb{R}^{l} with λl(Zq)=1\lambda_l(Z_q)=1 by the rectangle values. Any measure on Bl\mathcal{B}_l with the stated rectangle values agrees with λl\lambda_l on Rl\mathcal{R}_l, so Step 1 gives uniqueness.

Step 3 (claim 2 for l2l\ge2: generation and bimeasurability). Let l2l\ge2, (Y,G,μ)(Y,\mathcal{G},\mu) and ii be as in claim 2, and write Pl\mathcal{P}_{l'} for the family of sets R×GR\times G with RRlR\in\mathcal{R}_{l'} and GGG\in\mathcal{G}.

(a) For every l1l'\ge1, BlG\mathcal{B}_{l'}\otimes\mathcal{G} is generated by Pl\mathcal{P}_{l'}: each R×GR\times G is a measurable rectangle of BlG\mathcal{B}_{l'}\otimes\mathcal{G}, since Rσ(Rl)=BlR\in\sigma(\mathcal{R}_{l'})=\mathcal{B}_{l'} by claim 1; conversely, for fixed GGG\in\mathcal{G}, the class of HRlH\subseteq\mathbb{R}^{l'} with H×Gσ(Pl)H\times G\in\sigma(\mathcal{P}_{l'}) is a σ\sigma-algebra (it contains Rl\mathbb{R}^{l'} as Rl×GPl\mathbb{R}^{l'}\times G\in\mathcal{P}_{l'}, and is closed under complements and countable unions exactly as in Step 2) containing Rl\mathcal{R}_{l'}, hence containing Bl\mathcal{B}_{l'} by claim 1; so every generator H×GH\times G of BlG\mathcal{B}_{l'}\otimes\mathcal{G} lies in σ(Pl)\sigma(\mathcal{P}_{l'}). This proves in particular the generation assertion of claim 2.

(b) B(R)(Bl1G)\mathcal{B}(\mathbb{R})\otimes(\mathcal{B}_{l-1}\otimes\mathcal{G}) is generated by the family Q\mathcal{Q} of sets A×(R×G)A\times(R'\times G) with AB(R)A\in\mathcal{B}(\mathbb{R}), RRl1R'\in\mathcal{R}_{l-1}, GGG\in\mathcal{G}: each such set is a measurable rectangle, since R×GBl1GR'\times G\in\mathcal{B}_{l-1}\otimes\mathcal{G}; conversely, for fixed AB(R)A\in\mathcal{B}(\mathbb{R}), the class of KRl1×YK\subseteq\mathbb{R}^{l-1}\times Y with A×Kσ(Q)A\times K\in\sigma(\mathcal{Q}) is a σ\sigma-algebra (it contains Rl1×Y\mathbb{R}^{l-1}\times Y since A×(Rl1×Y)QA\times(\mathbb{R}^{l-1}\times Y)\in\mathcal{Q}, and is closed under complements and countable unions as before) containing the sets R×GR'\times G, hence containing Bl1G\mathcal{B}_{l-1}\otimes\mathcal{G} by (a) applied with l=l1l'=l-1; so every generator A×KA\times K lies in σ(Q)\sigma(\mathcal{Q}).

Ψi\Psi_i is a bijection; its inverse Ξ\Xi sends (θ,y)(\theta,y) to (θi,((θ1,,θi1,θi+1,,θl),y))(\theta_i,((\theta_1,\dots,\theta_{i-1},\theta_{i+1},\dots,\theta_l),y)). For R=A1××AlRlR=A_1\times\dots\times A_l\in\mathcal{R}_l and GGG\in\mathcal{G},

Ψi1(R×G)=Ai×((A1××Ai1×Ai+1××Al)×G)Q,\Psi_i^{-1}(R\times G)=A_i\times\bigl((A_1\times\dots\times A_{i-1}\times A_{i+1}\times\dots\times A_l)\times G\bigr)\in\mathcal{Q},

and for AB(R)A\in\mathcal{B}(\mathbb{R}), R=A1××Al1Rl1R'=A_1'\times\dots\times A_{l-1}'\in\mathcal{R}_{l-1}, GGG\in\mathcal{G},

Ξ1(A×(R×G))=(A1××Ai1×A×Ai××Al1)×GPl.\Xi^{-1}\bigl(A\times(R'\times G)\bigr)=(A_1'\times\dots\times A_{i-1}'\times A\times A_i'\times\dots\times A_{l-1}')\times G\in\mathcal{P}_l .

By the generator criterion of Measurable Function and Real-Valued Measurable Function together with (a) and (b), both Ψi\Psi_i and Ξ\Xi are measurable.

Step 4 (claim 2 for l2l\ge2: the image measure). The measures λl1μ\lambda_{l-1}\otimes\mu, λlμ\lambda_l\otimes\mu and λ(λl1μ)\lambda\otimes(\lambda_{l-1}\otimes\mu) are defined and σ\sigma-finite by Existence and Uniqueness of the Product Measure, their factors being σ\sigma-finite by claim 1 and by hypothesis. Let mm be the image measure of λ(λl1μ)\lambda\otimes(\lambda_{l-1}\otimes\mu) under Ψi\Psi_i, a measure on BlG\mathcal{B}_l\otimes\mathcal{G} by Step 3 and claim 1 of Image Measures, Measures with Densities, and Change of Variables. For R=A1××AlRlR=A_1\times\dots\times A_l\in\mathcal{R}_l and GGG\in\mathcal{G}, the preimage computed in Step 3 and the rectangle property of Existence and Uniqueness of the Product Measure, applied twice, give

m(R×G)=λ(Ai)λl1(A1××Ai1×Ai+1××Al)μ(G)=λ(A1)λ(Al)μ(G)=(λlμ)(R×G),m(R\times G)=\lambda(A_i)\cdot\lambda_{l-1}(A_1\times\dots\times A_{i-1}\times A_{i+1}\times\dots\times A_l)\cdot\mu(G)=\lambda(A_1)\cdots\lambda(A_l)\,\mu(G)=(\lambda_l\otimes\mu)(R\times G),

using the rectangle values of claim 1 for λl1\lambda_{l-1} and for λl\lambda_l. The family Pl\mathcal{P}_l is a π\pi-system (coordinatewise intersections). By σ\sigma-finiteness of μ\mu there are Y1,Y2,GY_1'',Y_2'',\dots\in\mathcal{G} covering YY with μ(Yn)<\mu(Y_n'')<\infty; setting Y1=Y1Y_1=Y_1'' and Yn=Yn(Y1Yn1)Y_n=Y_n''\setminus(Y_1''\cup\dots\cup Y_{n-1}'') yields pairwise disjoint members of G\mathcal{G} with nYn=Y\bigcup_nY_n=Y and μ(Yn)μ(Yn)<\mu(Y_n)\le\mu(Y_n'')<\infty by monotonicity. The countably many sets Zq×YnZ_q\times Y_n (with ZqZ_q as in Step 2) are pairwise disjoint members of Pl\mathcal{P}_l covering Rl×Y\mathbb{R}^{l}\times Y, with (λlμ)(Zq×Yn)=μ(Yn)<(\lambda_l\otimes\mu)(Z_q\times Y_n)=\mu(Y_n)<\infty. Step 1 applied to mm and λlμ\lambda_l\otimes\mu gives m=λlμm=\lambda_l\otimes\mu.

Step 5 (claims 3 and 4 for l2l\ge2). Write m0=λ(λl1μ)m_0=\lambda\otimes(\lambda_{l-1}\otimes\mu). Let f:Rl×Y[0,]f:\mathbb{R}^{l}\times Y\to[0,\infty] be BlG\mathcal{B}_l\otimes\mathcal{G}-measurable. Then fΨif\circ\Psi_i is measurable, since for real aa one has {fΨi>a}=Ψi1({f>a})\{f\circ\Psi_i>a\}=\Psi_i^{-1}(\{f>a\}), measurable by Step 3; and claim 2 of Image Measures, Measures with Densities, and Change of Variables, applied to T=ΨiT=\Psi_i whose image measure is λlμ\lambda_l\otimes\mu by Step 4, gives

Rl×Yfd(λlμ)=fΨidm0.\int_{\mathbb{R}^{l}\times Y}f\,d(\lambda_l\otimes\mu)=\int f\circ\Psi_i\,dm_0 .

The Sections clause of Tonelli and Fubini Theorems, applied to the σ\sigma-finite factors (R,B(R),λ)(\mathbb{R},\mathcal{B}(\mathbb{R}),\lambda) and (Rl1×Y,Bl1G,λl1μ)(\mathbb{R}^{l-1}\times Y,\mathcal{B}_{l-1}\otimes\mathcal{G},\lambda_{l-1}\otimes\mu), shows that for every (θ,y)(\theta',y) the section t(fΨi)(t,(θ,y))=f(Ψi(t,(θ,y)))t\mapsto(f\circ\Psi_i)(t,(\theta',y))=f(\Psi_i(t,(\theta',y))) is B(R)\mathcal{B}(\mathbb{R})-measurable, and the Tonelli clause — the second iterated integral of its display, which integrates over the first factor innermost — shows that (θ,y)Rf(Ψi(t,(θ,y)))dλ(t)(\theta',y)\mapsto\int_{\mathbb{R}}f(\Psi_i(t,(\theta',y)))\,d\lambda(t) is Bl1G\mathcal{B}_{l-1}\otimes\mathcal{G}-measurable with

fΨidm0=Rl1×Y(Rf(Ψi(t,(θ,y)))dλ(t))d(λl1μ)(θ,y).\int f\circ\Psi_i\,dm_0=\int_{\mathbb{R}^{l-1}\times Y}\Bigl(\int_{\mathbb{R}}f\bigl(\Psi_i(t,(\theta',y))\bigr)\,d\lambda(t)\Bigr)d(\lambda_{l-1}\otimes\mu)(\theta',y).

Chaining the two displays proves claim 3. For claim 4, let f:Rl×YRf:\mathbb{R}^{l}\times Y\to\mathbb{R} be integrable with respect to λlμ\lambda_l\otimes\mu; by the integrable clause of claim 2 of Image Measures, Measures with Densities, and Change of Variables, fΨif\circ\Psi_i is integrable with respect to m0m_0 with the same integral, and the Fubini clause of Tonelli and Fubini Theorems — its symmetric statement in the other order, whose exceptional set lies in the second factor — provides exactly the set NBl1GN\in\mathcal{B}_{l-1}\otimes\mathcal{G} of (λl1μ)(\lambda_{l-1}\otimes\mu)-measure zero and the integrability statements and identity asserted in claim 4.

Step 6 (claims 2, 3, and 4 for l=1l=1). Under the conventions of the statement: the generation assertion of claim 2 is that B(R)G\mathcal{B}(\mathbb{R})\otimes\mathcal{G} is generated by the measurable rectangles A×GA\times G, which is Product Sigma-Algebra directly; Ψ1\Psi_1 is the identity map of R×Y\mathbb{R}\times Y, trivially a bijection that is measurable with measurable inverse, and the image measure of λμ\lambda\otimes\mu under it is λμ=λ1μ\lambda\otimes\mu=\lambda_1\otimes\mu itself (its value on each BB(R)GB\in\mathcal{B}(\mathbb{R})\otimes\mathcal{G} being unchanged); and claims 3 and 4 are precisely the Sections, Tonelli, and symmetric Fubini clauses of Tonelli and Fubini Theorems applied to the σ\sigma-finite factors (R,B(R),λ)(\mathbb{R},\mathcal{B}(\mathbb{R}),\lambda) and (Y,G,μ)(Y,\mathcal{G},\mu), in the order that integrates over R\mathbb{R} innermost. \blacksquare

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Prerequisites

Loading...

Comments

Loading…