TheoremBase

Boxes are Borel, closed under intersection and generate the open sets; the cubes cover ℝⁿ, so the π-λ uniqueness theorem gives the determination by boxes. Existence is by induction on the dimension: the image of Lebesgue measure on ℝ in dimension one, and the image of a product measure under concatenation in the step.

Proof

Steps 1 to 4 are proved for an arbitrary dimension n∈Nn\in\mathbb{N}, and are used in Step 6 in every dimension.

Step 1: open boxes are Borel. Let a,b∈Rna,b\in\mathbb{R}^{n}. For i∈[n]i\in[n], the set {x∈Rn:ai<xi<bi}\{x\in\mathbb{R}^{n}:a_{i}<x_{i}<b_{i}\} is the preimage of the open interval (ai,bi)∈B(R)(a_{i},b_{i})\in\mathcal{B}(\mathbb{R}) under the ii-th coordinate function, so it lies in B(Rn)\mathcal{B}(\mathbb{R}^{n}) (Intervals Are Borel, and Rays, Half-Open Intervals and Rational Intervals Generate the Borel Sigma-Algebra of the Real Line (ii), The Borel Sets of ℝⁿ and of the Real Matrices Are Generated by the Coordinates (iii), Closure Properties of σ-Algebras, Intersections and Traces of σ-Algebras, and the Generated σ-Algebra (iii)). The open box R(a,b)R(a,b) is the intersection of these nn sets, a finite intersection of Borel sets, so R(a,b)∈B(Rn)R(a,b)\in\mathcal{B}(\mathbb{R}^{n}) (Closure Properties of σ-Algebras, Intersections and Traces of σ-Algebras, and the Generated σ-Algebra (ix), by induction on the number of sets). Hence Rn⊆B(Rn)\mathcal{R}_{n}\subseteq\mathcal{B}(\mathbb{R}^{n}).

Step 2: a π\pi-system. Let a,b,a′,b′∈Rna,b,a',b'\in\mathbb{R}^{n}. By Open Boxes: Emptiness, Intersections and Preimages under Concatenation (ii), the points c,d∈Rnc,d\in\mathbb{R}^{n} with components ci=max⁡{ai,ai′}c_{i}=\max\{a_{i},a'_{i}\} and di=min⁡{bi,bi′}d_{i}=\min\{b_{i},b'_{i}\} satisfy

R(a,b)∩R(a′,b′)=R(c,d)∈Rn,R(a,b)\cap R(a',b')=R(c,d)\in\mathcal{R}_{n},

and Rn\mathcal{R}_{n} is a π\pi-system on Rn\mathbb{R}^{n}.

Step 3: the boxes generate B(Rn)\mathcal{B}(\mathbb{R}^{n}). By Step 1, σ(Rn)⊆B(Rn)\sigma(\mathcal{R}_{n})\subseteq\mathcal{B}(\mathbb{R}^{n}) (Closure Properties of σ-Algebras, Intersections and Traces of σ-Algebras, and the Generated σ-Algebra (iv)). Conversely, let U⊆RnU\subseteq\mathbb{R}^{n} be open. By The Borel Sets of ℝⁿ and of the Real Matrices Are Generated by the Coordinates (i), U=⋃k∈NWkU=\bigcup_{k\in\mathbb{N}}W_{k}, where each WkW_{k} is a rational open box R(a,b)R(a,b), a,b∈Qna,b\in\mathbb{Q}^{n}; reading the components of aa and bb as real numbers (The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness (v)), WkW_{k} is an open box in the sense of Open Boxes in ℝⁿ (i), so Wk∈RnW_{k}\in\mathcal{R}_{n} and U∈σ(Rn)U\in\sigma(\mathcal{R}_{n}). Thus every open subset of Rn\mathbb{R}^{n} lies in σ(Rn)\sigma(\mathcal{R}_{n}), and B(Rn)⊆σ(Rn)\mathcal{B}(\mathbb{R}^{n})\subseteq\sigma(\mathcal{R}_{n}) (Borel Sigma-Algebra of a Metric Space (i), Closure Properties of σ-Algebras, Intersections and Traces of σ-Algebras, and the Generated σ-Algebra (iv)).

Step 4: the cubes cover Rn\mathbb{R}^{n}. Let x∈Rnx\in\mathbb{R}^{n}. There is k∈Nk\in\mathbb{N} with ∥x∥<k\lVert x\rVert<k (The Integers in the Real Numbers: Greatest Elements, the Integer Part, and Archimedean Consequences (ii), with ε=1\varepsilon=1), and ∣xi∣≤∥x∥|x_{i}|\le\lVert x\rVert for every i∈[n]i\in[n] (Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square (iv)). So −k<xi<k-k<x_{i}<k for every i∈[n]i\in[n], that is, x∈Qkx\in Q_{k}. Steps 2 and 3 prove (i), and Step 4 proves (ii).

Step 5: boxes determine measures. Let μ\mu and ν\nu be as in (iv). The cubes QkQ_{k} are open boxes, so (Qk)k∈N(Q_{k})_{k\in\mathbb{N}} is a sequence in Rn\mathcal{R}_{n} with union Rn\mathbb{R}^{n} (Step 4) and μ(Qk)<∞\mu(Q_{k})<\infty for every kk. Since Rn\mathcal{R}_{n} is a π\pi-system generating B(Rn)\mathcal{B}(\mathbb{R}^{n}) on which μ\mu and ν\nu agree, μ=ν\mu=\nu by Uniqueness of Measures Agreeing on a Generating π-System (ii). This proves (iv).

Step 6: existence. For p∈Np\in\mathbb{N} let φ(p)\varphi(p) be the statement: there is a measure κ\kappa on (Rp,B(Rp))(\mathbb{R}^{p},\mathcal{B}(\mathbb{R}^{p})) with κ(R(a,b))=∏i=1p(bi−ai)\kappa(R(a,b))=\prod_{i=1}^{p}(b_{i}-a_{i}) for all a,b∈Rpa,b\in\mathbb{R}^{p} with ai<bia_{i}<b_{i} for every i∈[p]i\in[p]. We prove φ(p)\varphi(p) for every p∈Np\in\mathbb{N} by induction (Induction for Predicative Formulas, and the Order of ℕ below 1 and below n+1 (i); φ\varphi is predicative by Routine Facts Used Without Citation: Arithmetic, Maps, Predicativity and Indices (iv)). Throughout, products of nonnegative real numbers in [0,∞][0,\infty] are their real products (Addition and Multiplication on [0,∞] (ii)).

A measure κ\kappa as in φ(p)\varphi(p) is σ\sigma-finite: the cubes Qk(p)=R(−ku,ku)Q^{(p)}_{k}=R(-k\mathbf{u},k\mathbf{u}) of Rp\mathbb{R}^{p}, k∈Nk\in\mathbb{N}, with u=(1,…,1)∈Rp\mathbf{u}=(1,\dots,1)\in\mathbb{R}^{p}, lie in B(Rp)\mathcal{B}(\mathbb{R}^{p}) (Step 1), cover Rp\mathbb{R}^{p} (Step 4), and κ(Qk(p))=∏i=1p2k\kappa(Q^{(p)}_{k})=\prod_{i=1}^{p}2k is a real number.

Dimension one. Let τ:R→R1\tau:\mathbb{R}\to\mathbb{R}^{1} be the function with τ(t)1=t\tau(t)_{1}=t, which is B(R)/B(R1)\mathcal{B}(\mathbb{R})/\mathcal{B}(\mathbb{R}^{1})-measurable by Borel Sets of ℝ^{m+n} and the Product σ-Algebra of ℝ^m and ℝ^n (iii). Lebesgue measure λ\lambda is a measure on (R,B(R))(\mathbb{R},\mathcal{B}(\mathbb{R})) (Lebesgue Measure on the Real Line: Existence, Lengths of Intervals and Uniqueness (iv)), so its image measure κ1=λ∘τ−1\kappa_{1}=\lambda\circ\tau^{-1} is a measure on (R1,B(R1))(\mathbb{R}^{1},\mathcal{B}(\mathbb{R}^{1})) (Image Measures and Measures with a Density: Measures, Integrals and Change of Variables (i)). Let a,b∈R1a,b\in\mathbb{R}^{1} with a1<b1a_{1}<b_{1}. Since τ(t)1=t\tau(t)_{1}=t, we have τ−1(R(a,b))=(a1,b1)\tau^{-1}(R(a,b))=(a_{1},b_{1}), and

κ1(R(a,b))=λ((a1,b1))=b1−a1=∏i=11(bi−ai)\kappa_{1}\bigl(R(a,b)\bigr)=\lambda\bigl((a_{1},b_{1})\bigr)=b_{1}-a_{1}=\prod_{i=1}^{1}(b_{i}-a_{i})

(Lebesgue Measure on the Real Line: Existence, Lengths of Intervals and Uniqueness (i), Iterated Operations: Recursion, Splitting, Reordering, Termwise Combination and Homomorphisms (i)). So φ(1)\varphi(1) holds.

From pp to p+1p+1. Assume φ(p)\varphi(p), let κp\kappa_{p} be a measure as there, and let κ1\kappa_{1} be the measure just constructed; both are σ\sigma-finite. Their product measure κp⊗κ1\kappa_{p}\otimes\kappa_{1} is a measure on the product σ\sigma-algebra B(Rp)⊗B(R1)\mathcal{B}(\mathbb{R}^{p})\otimes\mathcal{B}(\mathbb{R}^{1}) with

(κp⊗κ1)(A×B)=κp(A) κ1(B)(A∈B(Rp), B∈B(R1))(\kappa_{p}\otimes\kappa_{1})(A\times B)=\kappa_{p}(A)\,\kappa_{1}(B)\qquad(A\in\mathcal{B}(\mathbb{R}^{p}),\ B\in\mathcal{B}(\mathbb{R}^{1}))

(Existence and Uniqueness of the Product Measure (i)). The concatenation function ι:Rp×R1→Rp+1\iota:\mathbb{R}^{p}\times\mathbb{R}^{1}\to\mathbb{R}^{p+1} of Concatenation Identifies a Product of Euclidean Spaces with a Euclidean Space (ii) is B(Rp)⊗B(R1)/B(Rp+1)\mathcal{B}(\mathbb{R}^{p})\otimes\mathcal{B}(\mathbb{R}^{1})/\mathcal{B}(\mathbb{R}^{p+1})-measurable (Borel Sets of ℝ^{m+n} and the Product σ-Algebra of ℝ^m and ℝ^n (i)), so κp+1=(κp⊗κ1)∘ι−1\kappa_{p+1}=(\kappa_{p}\otimes\kappa_{1})\circ\iota^{-1} is a measure on (Rp+1,B(Rp+1))(\mathbb{R}^{p+1},\mathcal{B}(\mathbb{R}^{p+1})) (Image Measures and Measures with a Density: Measures, Integrals and Change of Variables (i)).

Let a,b∈Rp+1a,b\in\mathbb{R}^{p+1} with ai<bia_{i}<b_{i} for every i∈[p+1]i\in[p+1], and let (a′,a′′)(a',a'') and (b′,b′′)(b',b'') be the pairs that ι\iota maps to aa and bb (Concatenation Identifies a Product of Euclidean Spaces with a Euclidean Space (iii)): ai′=aia'_{i}=a_{i} and bi′=bib'_{i}=b_{i} for i∈[p]i\in[p], and a1′′=ap+1a''_{1}=a_{p+1}, b1′′=bp+1b''_{1}=b_{p+1}. By Open Boxes: Emptiness, Intersections and Preimages under Concatenation (iii),

ι−1(R(a,b))=R(a′,b′)×R(a′′,b′′),\iota^{-1}\bigl(R(a,b)\bigr)=R(a',b')\times R(a'',b''),

a product of Borel sets by Step 1. Here κp(R(a′,b′))=∏i=1p(bi−ai)\kappa_{p}(R(a',b'))=\prod_{i=1}^{p}(b_{i}-a_{i}) by φ(p)\varphi(p), and κ1(R(a′′,b′′))=b1′′−a1′′=bp+1−ap+1\kappa_{1}(R(a'',b''))=b''_{1}-a''_{1}=b_{p+1}-a_{p+1} by the dimension-one computation above. Therefore

κp+1(R(a,b))=κp(R(a′,b′)) κ1(R(a′′,b′′))=(∏i=1p(bi−ai))(bp+1−ap+1)=∏i=1p+1(bi−ai),\kappa_{p+1}\bigl(R(a,b)\bigr)=\kappa_{p}\bigl(R(a',b')\bigr)\,\kappa_{1}\bigl(R(a'',b'')\bigr)=\Bigl(\prod_{i=1}^{p}(b_{i}-a_{i})\Bigr)(b_{p+1}-a_{p+1})=\prod_{i=1}^{p+1}(b_{i}-a_{i}),

the last equality by Iterated Operations: Recursion, Splitting, Reordering, Termwise Combination and Homomorphisms (i). So φ(p+1)\varphi(p+1) holds, and by induction φ(n)\varphi(n) holds: a measure as in (iii) exists.

Step 7: uniqueness. Let κ\kappa and κ′\kappa' be measures as in (iii), and let a,b∈Rna,b\in\mathbb{R}^{n}. If ai<bia_{i}<b_{i} for every i∈[n]i\in[n], then κ(R(a,b))=∏i=1n(bi−ai)=κ′(R(a,b))\kappa(R(a,b))=\prod_{i=1}^{n}(b_{i}-a_{i})=\kappa'(R(a,b)), a real number. Otherwise bi≤aib_{i}\le a_{i} for some ii, so R(a,b)=∅R(a,b)=\emptyset (Open Boxes: Emptiness, Intersections and Preimages under Concatenation (i)) and κ(R(a,b))=0=κ′(R(a,b))\kappa(R(a,b))=0=\kappa'(R(a,b)). Thus κ\kappa and κ′\kappa' agree and are finite on every open box, and κ=κ′\kappa=\kappa' by Step 5.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…