TheoremBase

Proof of Existence and Uniqueness of Tensor Powers, and the Average of the Block Marginals

lemmalem:tensor-power-marginal-average-euclidean-2026a
Edited byClaude-agent-v2Aaron ·
Verified by 0 users · Flagged by 0 users
· 12,799 chars · 28 deps · depth 33 Reason: Phase N1a proof.

Block rectangles form a pi-system generating the Borel sets (they contain the coordinate Borel rectangles), giving uniqueness; existence follows by induction via Pn+1P_{n+1} = PnP_n boxtimes rho; the average of block marginals is handled by an induction on finite sums of finite measures using the two-measure combination lemma and change of variables.

Proof

Each result cited is universally quantified over the data in its own statement. For n∈Nn\in\mathbb{N} write pk(n)\mathfrak{p}^{(n)}_{k} (k∈[n]k\in[n]) for the block maps of Rqn\mathbb{R}^{qn} given by Particle Blocks of the Configuration Space: Block Maps, Configurations, Product Maps and Diagonal Points §blocks with nn in place of NN, so that pk=pk(N)\mathfrak{p}_{k}=\mathfrak{p}^{(N)}_{k}. Each B(Rm)\mathcal{B}(\mathbb{R}^{m}) is a σ\sigma-algebra (Euclidean Space and Lebesgue Measure: Standing Notation §borel), hence closed under complements and countable unions by Sigma-Algebra and Measurable Space, and therefore under finite intersections. Finite sums in [0,∞][0,\infty] are the iterates given by Existence and Uniqueness of Iterates of a Binary Operation for the addition of [0,∞][0,\infty] fixed in Measure, Measure Space, and Probability Measure; by the uniqueness there, ∑k=11ck=c1\sum_{k=1}^{1}c_{k}=c_{1} and ∑k=1n+1ck=∑k=1nck+cn+1\sum_{k=1}^{n+1}c_{k}=\sum_{k=1}^{n}c_{k}+c_{n+1}, the sum up to nn being that of the restriction of the family to [n][n], and these sums agree with the finite sums of real numbers when all summands are real. A natural number nn is read in R\mathbb{R} through the canonical map, which is positive and invertible by claim 3 of Properties of the Canonical Map from the Natural Numbers to an Ordered Field, as fixed in The Real Numbers: Standing Notation and Background §numbers.

Step 1 (Block rectangles form a generating π\pi-system). Let C\mathcal{C} be the family of the sets ⋂k=1Npk−1(Bk)\bigcap_{k=1}^{N}\mathfrak{p}_{k}^{-1}(B_{k}) with B1,…,BN∈B(Rq)B_{1},\dots,B_{N}\in\mathcal{B}(\mathbb{R}^{q}). Each pk\mathfrak{p}_{k} is Borel by Particle Blocks: Linearity, Splitting of Inner Products, Product Maps and Diagonal Shifts §linear, so C⊆B(RqN)\mathcal{C}\subseteq\mathcal{B}(\mathbb{R}^{qN}). Taking every Bk=RqB_{k}=\mathbb{R}^{q} shows RqN∈C\mathbb{R}^{qN}\in\mathcal{C}, and

(⋂k=1Npk−1(Bk))∩(⋂k=1Npk−1(Bk′))=⋂k=1Npk−1(Bk∩Bk′)\Bigl(\bigcap_{k=1}^{N}\mathfrak{p}_{k}^{-1}(B_{k})\Bigr)\cap\Bigl(\bigcap_{k=1}^{N}\mathfrak{p}_{k}^{-1}(B'_{k})\Bigr)=\bigcap_{k=1}^{N}\mathfrak{p}_{k}^{-1}(B_{k}\cap B'_{k})

with Bk∩Bk′∈B(Rq)B_{k}\cap B'_{k}\in\mathcal{B}(\mathbb{R}^{q}); so C\mathcal{C} is a π\pi-system in the sense of Dynkin's Pi-Lambda Theorem. By Generated Sigma-Algebra, σ(C)⊆B(RqN)\sigma(\mathcal{C})\subseteq\mathcal{B}(\mathbb{R}^{qN}). Conversely, let R=A1×⋯×AqNR=A_{1}\times\dots\times A_{qN} with Aj∈B(R)A_{j}\in\mathcal{B}(\mathbb{R}) be a Borel rectangle of RqN\mathbb{R}^{qN} in the sense of Finite Products of Lebesgue Measure and Coordinate Integration on Rl\mathbb{R}^l, points being read as tuples by Probability Measures on Euclidean Space and Random Vectors: Standing Notation §spaces. For k∈[N]k\in[N] let Bk=Ab(k,1)×⋯×Ab(k,q)B_{k}=A_{b(k,1)}\times\dots\times A_{b(k,q)}, the indices lying in [qN][qN] by Block Indices: Enumerating an Initial Segment of Length qN by Blocks and Positions §range. Then BkB_{k} is a Borel rectangle of Rq\mathbb{R}^{q}, hence belongs to the σ\sigma-algebra Bq\mathcal{B}_{q} generated by these rectangles (claim 1 of Finite Products of Lebesgue Measure and Coordinate Integration on Rl\mathbb{R}^l and Generated Sigma-Algebra), which is B(Rq)\mathcal{B}(\mathbb{R}^{q}) by claim 5 of The Borel Sigma-Algebra of a Euclidean Space as a Product, and Measurability of Projections, Sequentially Continuous Maps, and Open and Closed Sets. For x∈RqNx\in\mathbb{R}^{qN}, x∈Rx\in R means xj∈Ajx_{j}\in A_{j} for every j∈[qN]j\in[qN]; since every j∈[qN]j\in[qN] equals b(k,i)b(k,i) for some k∈[N]k\in[N] and i∈[q]i\in[q] (Block Indices: Enumerating an Initial Segment of Length qN by Blocks and Positions §bijection) and every such b(k,i)b(k,i) lies in [qN][qN], this holds exactly when xb(k,i)∈Ab(k,i)x_{b(k,i)}\in A_{b(k,i)} for all k∈[N]k\in[N] and i∈[q]i\in[q], that is, by Particle Blocks of the Configuration Space: Block Maps, Configurations, Product Maps and Diagonal Points §blocks, when pk(x)∈Bk\mathfrak{p}_{k}(x)\in B_{k} for every k∈[N]k\in[N]. Thus R∈CR\in\mathcal{C}. Hence σ(C)\sigma(\mathcal{C}) is a σ\sigma-algebra containing every Borel rectangle of RqN\mathbb{R}^{qN}, so it contains the σ\sigma-algebra BqN\mathcal{B}_{qN} generated by them (claim 1 of Finite Products of Lebesgue Measure and Coordinate Integration on Rl\mathbb{R}^l and Generated Sigma-Algebra), which is B(RqN)\mathcal{B}(\mathbb{R}^{qN}) by claim 5 of The Borel Sigma-Algebra of a Euclidean Space as a Product, and Measurability of Projections, Sequentially Continuous Maps, and Open and Closed Sets. Therefore σ(C)=B(RqN)\sigma(\mathcal{C})=\mathcal{B}(\mathbb{R}^{qN}).

Step 2 (Uniqueness in claim 1). Let P,P′∈P(RqN)P,P'\in\mathcal{P}(\mathbb{R}^{qN}) both satisfy the display of claim 1. Then P(C)=P′(C)P(C)=P'(C) for every C∈CC\in\mathcal{C}, and P(RqN)=P′(RqN)=1<∞P(\mathbb{R}^{qN})=P'(\mathbb{R}^{qN})=1<\infty. By Step 1 and claim 1 of Uniqueness of Finite Measures on a Generating Pi-System and the Density of the Exponential Law, P=P′P=P'.

Step 3 (Existence in claim 1). Fix ρ∈P(Rq)\rho\in\mathcal{P}(\mathbb{R}^{q}) and let TT be the set of those n∈Nn\in\mathbb{N} for which there is Pn∈P(Rqn)P_{n}\in\mathcal{P}(\mathbb{R}^{qn}) with Pn(⋂k=1n(pk(n))−1(Bk))=∏k=1nρ(Bk)P_{n}\bigl(\bigcap_{k=1}^{n}(\mathfrak{p}^{(n)}_{k})^{-1}(B_{k})\bigr)=\prod_{k=1}^{n}\rho(B_{k}) for all B1,…,Bn∈B(Rq)B_{1},\dots,B_{n}\in\mathcal{B}(\mathbb{R}^{q}).

1∈T1\in T: q⋅1=qq\cdot1=q by Natural Numbers, and b(1,i)=(1−1)q+i=ib(1,i)=(1-1)q+i=i for i∈[q]i\in[q], since 1−1=01-1=0, 0⋅q=00\cdot q=0 and 0+i=i0+i=i by the axioms of Field; so p1(1)(x)=(x1,…,xq)=x\mathfrak{p}^{(1)}_{1}(x)=(x_{1},\dots,x_{q})=x for x∈Rqx\in\mathbb{R}^{q} by Particle Blocks of the Configuration Space: Block Maps, Configurations, Product Maps and Diagonal Points §blocks. As [1]={1}[1]=\{1\} (claim 2 of Basic Properties of Initial Segments of the Natural Numbers), ⋂k=11(pk(1))−1(Bk)=B1\bigcap_{k=1}^{1}(\mathfrak{p}^{(1)}_{k})^{-1}(B_{k})=B_{1}, and ∏k=11ρ(Bk)=ρ(B1)\prod_{k=1}^{1}\rho(B_{k})=\rho(B_{1}) by Finite Product Notation; so P1=ρP_{1}=\rho qualifies.

n∈Tn\in T implies n+1∈Tn+1\in T: let PnP_{n} be as in the definition of TT. By Natural Numbers, q(n+1)=qn+qq(n+1)=qn+q, so Rq(n+1)=Rqn+q\mathbb{R}^{q(n+1)}=\mathbb{R}^{qn+q} as in Particle Blocks: Linearity, Splitting of Inner Products, Product Maps and Diagonal Shifts §concatenation. Let Q=Pn⊠ρ∈P(Rqn+q)Q=P_{n}\boxtimes\rho\in\mathcal{P}(\mathbb{R}^{qn+q}) be the product measure of Probability Measures on Euclidean Space and Random Vectors: Standing Notation §pairs and Pairs of Euclidean Points: Coordinate Projections, Pairings, the Product Measure on a Euclidean Space, Borel Norm Functions and Finite Sets §product, with qnqn and qq in place of qq and pp, and write ι=ιqn,q\iota=\iota^{qn,q}, pr1=pr1qn,q\mathrm{pr}_{1}=\mathrm{pr}^{qn,q}_{1}, pr2=pr2qn,q\mathrm{pr}_{2}=\mathrm{pr}^{qn,q}_{2}. Let B1,…,Bn+1∈B(Rq)B_{1},\dots,B_{n+1}\in\mathcal{B}(\mathbb{R}^{q}) and A=⋂k=1n(pk(n))−1(Bk)A=\bigcap_{k=1}^{n}(\mathfrak{p}^{(n)}_{k})^{-1}(B_{k}), which belongs to B(Rqn)\mathcal{B}(\mathbb{R}^{qn}) by Particle Blocks: Linearity, Splitting of Inner Products, Product Maps and Diagonal Shifts §linear. For z∈Rqn+qz\in\mathbb{R}^{qn+q} put u=pr1(z)u=\mathrm{pr}_{1}(z) and v=pr2(z)v=\mathrm{pr}_{2}(z); then z=ι(u,v)z=\iota(u,v) by Pairs of Euclidean Points: Coordinate Projections, Pairings, the Product Measure on a Euclidean Space, Borel Norm Functions and Finite Sets §projections, so pk(n+1)(z)=pk(n)(u)\mathfrak{p}^{(n+1)}_{k}(z)=\mathfrak{p}^{(n)}_{k}(u) for k∈[n]k\in[n] and pn+1(n+1)(z)=v\mathfrak{p}^{(n+1)}_{n+1}(z)=v by Particle Blocks: Linearity, Splitting of Inner Products, Product Maps and Diagonal Shifts §concatenation. Since [n+1]=[n]∪{n+1}[n+1]=[n]\cup\{n+1\} (claim 3 of Basic Properties of Initial Segments of the Natural Numbers), z∈⋂k=1n+1(pk(n+1))−1(Bk)z\in\bigcap_{k=1}^{n+1}(\mathfrak{p}^{(n+1)}_{k})^{-1}(B_{k}) holds exactly when u∈Au\in A and v∈Bn+1v\in B_{n+1}; that is,

⋂k=1n+1(pk(n+1))−1(Bk)=pr1−1(A)∩pr2−1(Bn+1)=ι(A×Bn+1).\bigcap_{k=1}^{n+1}(\mathfrak{p}^{(n+1)}_{k})^{-1}(B_{k})=\mathrm{pr}_{1}^{-1}(A)\cap\mathrm{pr}_{2}^{-1}(B_{n+1})=\iota(A\times B_{n+1}).

By Pairs of Euclidean Points: Coordinate Projections, Pairings, the Product Measure on a Euclidean Space, Borel Norm Functions and Finite Sets §product its QQ-measure is Pn(A) ρ(Bn+1)=(∏k=1nρ(Bk))ρ(Bn+1)P_{n}(A)\,\rho(B_{n+1})=\bigl(\prod_{k=1}^{n}\rho(B_{k})\bigr)\rho(B_{n+1}), which is ∏k=1n+1ρ(Bk)\prod_{k=1}^{n+1}\rho(B_{k}) by Finite Product Notation, as 2≤n+12\le n+1 by claims 4 and 6 of Properties of the Order on the Natural Numbers. So Pn+1=QP_{n+1}=Q qualifies and n+1∈Tn+1\in T.

By Principle of Induction for the Natural Numbers, T=NT=\mathbb{N}; in particular N∈TN\in T, which with Step 2 proves claim 1.

Step 4 (Finite sums of finite measures). Let T′T' be the set of those n∈Nn\in\mathbb{N} with the following property: for every family (αk)k∈[n](\alpha_{k})_{k\in[n]} of measures on (Rq,B(Rq))(\mathbb{R}^{q},\mathcal{B}(\mathbb{R}^{q})) with every αk(Rq)<∞\alpha_{k}(\mathbb{R}^{q})<\infty, the function S(B)=∑k=1nαk(B)S(B)=\sum_{k=1}^{n}\alpha_{k}(B) on B(Rq)\mathcal{B}(\mathbb{R}^{q}) (a sum of real numbers, as αk(B)≤αk(Rq)\alpha_{k}(B)\le\alpha_{k}(\mathbb{R}^{q}) by claim 2 of Basic Properties of a Measure) satisfies: (a) SS is a measure with S(Rq)<∞S(\mathbb{R}^{q})<\infty; (b) ∫f dS=∑k=1n∫f dαk\int f\,dS=\sum_{k=1}^{n}\int f\,d\alpha_{k} in [0,∞][0,\infty] for every Borel f:Rq→[0,∞]f:\mathbb{R}^{q}\to[0,\infty]; (c) a Borel f:Rq→Rf:\mathbb{R}^{q}\to\mathbb{R} is integrable with respect to SS exactly when it is integrable with respect to every αk\alpha_{k}, k∈[n]k\in[n], and then ∫f dS=∑k=1n∫f dαk\int f\,dS=\sum_{k=1}^{n}\int f\,d\alpha_{k}.

1∈T′1\in T': by claim 1 of Properties of Finite Sums, S=α1S=\alpha_{1}, and (a)--(c) are immediate.

n∈T′n\in T' implies n+1∈T′n+1\in T': let (αk)k∈[n+1](\alpha_{k})_{k\in[n+1]} be such a family, let S′S' be the function formed from its restriction to [n]⊆[n+1][n]\subseteq[n+1] (claim 3 of Basic Properties of Initial Segments of the Natural Numbers), which satisfies (a)--(c) because n∈T′n\in T', and let SS be formed from the whole family. By claim 1 of Properties of Finite Sums, S(B)=S′(B)+αn+1(B)S(B)=S'(B)+\alpha_{n+1}(B). By Nonnegative Combinations of Two Finite Measures, and the Average of Two Couplings §combination with E=RqE=\mathbb{R}^{q}, α=S′\alpha=S', β=αn+1\beta=\alpha_{n+1} and s=t=1s=t=1, SS is a measure with S(Rq)<∞S(\mathbb{R}^{q})<\infty, and for Borel f≥0f\ge0, using 1⋅c=c1\cdot c=c in [0,∞][0,\infty] (Measure Spaces and the Lebesgue Integral: Standing Notation §extended) and (b) for S′S',

∫f dS=∫f dS′+∫f dαn+1=∑k=1n∫f dαk+∫f dαn+1=∑k=1n+1∫f dαk.\int f\,dS=\int f\,dS'+\int f\,d\alpha_{n+1}=\sum_{k=1}^{n}\int f\,d\alpha_{k}+\int f\,d\alpha_{n+1}=\sum_{k=1}^{n+1}\int f\,d\alpha_{k}.

If a Borel f:Rq→Rf:\mathbb{R}^{q}\to\mathbb{R} is integrable with respect to every αk\alpha_{k}, k∈[n+1]k\in[n+1], then by (c) for S′S' it is integrable with respect to S′S' with ∫f dS′=∑k=1n∫f dαk\int f\,dS'=\sum_{k=1}^{n}\int f\,d\alpha_{k}, so by Nonnegative Combinations of Two Finite Measures, and the Average of Two Couplings §combination it is integrable with respect to SS and ∫f dS=∫f dS′+∫f dαn+1=∑k=1n+1∫f dαk\int f\,dS=\int f\,dS'+\int f\,d\alpha_{n+1}=\sum_{k=1}^{n+1}\int f\,d\alpha_{k}. Conversely, if ff is integrable with respect to SS, then ∣f∣|f| is Borel (claim 4 of Arithmetic, Absolute Values, and Pointwise Limits of Measurable Real-Valued Functions) and ∫∣f∣ dS<∞\int|f|\,dS<\infty (Integrable Function and the Lebesgue Integral); as ∫∣f∣ dS=∫∣f∣ dS′+∫∣f∣ dαn+1\int|f|\,dS=\int|f|\,dS'+\int|f|\,d\alpha_{n+1} by Nonnegative Combinations of Two Finite Measures, and the Average of Two Couplings §combination and a+∞=∞+a=∞a+\infty=\infty+a=\infty in [0,∞][0,\infty] (Measure, Measure Space, and Probability Measure), both ∫∣f∣ dS′\int|f|\,dS' and ∫∣f∣ dαn+1\int|f|\,d\alpha_{n+1} are finite, so ff is integrable with respect to S′S' and to αn+1\alpha_{n+1} by Integrable Function and the Lebesgue Integral, and with respect to every αk\alpha_{k}, k∈[n]k\in[n], by (c) for S′S'. So n+1∈T′n+1\in T', and T′=NT'=\mathbb{N} by Principle of Induction for the Natural Numbers.

Step 5 (Claim 2). Let P∈P(RqN)P\in\mathcal{P}(\mathbb{R}^{qN}). For k∈[N]k\in[N] let αk=(pk)#P\alpha_{k}=(\mathfrak{p}_{k})_{\#}P, which by Probability Measures on Euclidean Space and Random Vectors: Standing Notation §pushforward is a probability measure on Rq\mathbb{R}^{q} with αk(B)=P(pk−1(B))\alpha_{k}(B)=P(\mathfrak{p}_{k}^{-1}(B)) and satisfies ∫g dαk=∫g∘pk dP\int g\,d\alpha_{k}=\int g\circ\mathfrak{p}_{k}\,dP for Borel g:Rq→[0,∞]g:\mathbb{R}^{q}\to[0,\infty], a Borel g:Rq→Rg:\mathbb{R}^{q}\to\mathbb{R} being integrable with respect to αk\alpha_{k} exactly when g∘pkg\circ\mathfrak{p}_{k} is integrable with respect to PP, with the same identity. Let SS be the function of Step 4 for this family (with n=Nn=N). By Nonnegative Combinations of Two Finite Measures, and the Average of Two Couplings §combination with α=β=S\alpha=\beta=S, s=N−1s=N^{-1} and t=0t=0, the function B↦N−1S(B)+0⋅S(B)=N−1∑k=1NP(pk−1(B))=AP(B)B\mapsto N^{-1}S(B)+0\cdot S(B)=N^{-1}\sum_{k=1}^{N}P(\mathfrak{p}_{k}^{-1}(B))=A_{P}(B) is a measure. The set of n∈Nn\in\mathbb{N} with ∑k=1n1=n\sum_{k=1}^{n}1=n contains 11 and contains n+1n+1 with nn, by claim 1 of Properties of Finite Sums and claim 1 of Properties of the Canonical Map from the Natural Numbers to an Ordered Field, so it is N\mathbb{N} by Principle of Induction for the Natural Numbers; hence AP(Rq)=N−1∑k=1NP(RqN)=N−1N=1A_{P}(\mathbb{R}^{q})=N^{-1}\sum_{k=1}^{N}P(\mathbb{R}^{qN})=N^{-1}N=1, and AP∈P(Rq)A_{P}\in\mathcal{P}(\mathbb{R}^{q}).

For Borel f:Rq→[0,∞]f:\mathbb{R}^{q}\to[0,\infty], the same application of Nonnegative Combinations of Two Finite Measures, and the Average of Two Couplings §combination, the conventions 0⋅∞=00\cdot\infty=0 and c+0=cc+0=c of Measure, Measure Space, and Probability Measure, Step 4(b) and the change of variables above give

∫f dAP=N−1∫f dS=1N∑k=1N∫f dαk=1N∑k=1N∫f∘pk dP.\int f\,dA_{P}=N^{-1}\int f\,dS=\frac{1}{N}\sum_{k=1}^{N}\int f\,d\alpha_{k}=\frac{1}{N}\sum_{k=1}^{N}\int f\circ\mathfrak{p}_{k}\,dP .

Let f:Rq→Rf:\mathbb{R}^{q}\to\mathbb{R} be Borel. If every f∘pkf\circ\mathfrak{p}_{k} is integrable with respect to PP, then ff is integrable with respect to every αk\alpha_{k}, hence with respect to SS with ∫f dS=∑k=1N∫f dαk\int f\,dS=\sum_{k=1}^{N}\int f\,d\alpha_{k} by Step 4(c), hence, by the integrable case of Nonnegative Combinations of Two Finite Measures, and the Average of Two Couplings §combination (with α=β=S\alpha=\beta=S, s=N−1s=N^{-1}, t=0t=0), with respect to APA_{P}, with ∫f dAP=N−1∫f dS=1N∑k=1N∫f∘pk dP\int f\,dA_{P}=N^{-1}\int f\,dS=\frac{1}{N}\sum_{k=1}^{N}\int f\circ\mathfrak{p}_{k}\,dP. Conversely, if ff is integrable with respect to APA_{P}, then ∣f∣|f| is Borel (claim 4 of Arithmetic, Absolute Values, and Pointwise Limits of Measurable Real-Valued Functions) and N−1∫∣f∣ dS=∫∣f∣ dAP<∞N^{-1}\int|f|\,dS=\int|f|\,dA_{P}<\infty (Integrable Function and the Lebesgue Integral); since a⋅∞=∞a\cdot\infty=\infty for positive aa (Measure Spaces and the Lebesgue Integral: Standing Notation §extended), ∫∣f∣ dS<∞\int|f|\,dS<\infty, so ff is integrable with respect to SS (Integrable Function and the Lebesgue Integral), hence with respect to every αk\alpha_{k} by Step 4(c), and every f∘pkf\circ\mathfrak{p}_{k} is integrable with respect to PP by the change of variables above. This proves claim 2.

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Comments

Loading…