TheoremBase

Finite measures have real values and convergent real series, which gives the first two clauses; restrictions inherit countable additivity because a sub-σ-algebra is closed under countable unions; and countable additivity of the indefinite integral follows from dominated convergence.

Proof

Throughout, the field and order laws of R\mathbb{R} are used freely (The Structure of ℝ as a Field and an Ordered Ring (i), The Structure of ℝ as a Field and an Ordered Ring (ii), Rules of Arithmetic and Order in an Ordered Field (iv)).

Step 1: finite measures. Let μ\mu be a finite measure on (X,F)(X,\mathcal{F}). For A∈FA\in\mathcal{F} we have μ(A)≤μ(X)<∞\mu(A)\le\mu(X)<\infty (Basic Properties of a Measure (iii)), so μ(A)∈R\mu(A)\in\mathbb{R} (Small Facts about [0,∞]: Zero, Finiteness and Sums of Zeros (iii)); thus μ\mu is a map F→R\mathcal{F}\to\mathbb{R}, and μ(∅)=0\mu(\emptyset)=0. Let (Am)(A_{m}) be a pairwise disjoint sequence in F\mathcal{F} with union AA. Countable additivity of μ\mu says that the sum ∑m∈Nμ(Am)\sum_{m\in\mathbb{N}}\mu(A_{m}) in [0,∞][0,\infty] equals the real number μ(A)\mu(A); as its terms are real, the real series ∑m=1∞μ(Am)\sum_{m=1}^{\infty}\mu(A_{m}) converges with sum μ(A)\mu(A) (Sums of Sequences in [0,∞]: Real Series, Linearity, Comparison, Rearrangement, Double Series and Monotone Convergence (vi)). This proves (ii).

Step 2: differences of finite measures. Let μ1,μ2\mu_{1},\mu_{2} be as in (i). By Step 1 their values are real, so μ1−μ2\mu_{1}-\mu_{2} is a map F→R\mathcal{F}\to\mathbb{R}, and (μ1−μ2)(∅)=0(\mu_{1}-\mu_{2})(\emptyset)=0. Let (Am)(A_{m}) be a pairwise disjoint sequence in F\mathcal{F} with union AA. By Step 1 the series ∑mμi(Am)\sum_{m}\mu_{i}(A_{m}) converge with sums μi(A)\mu_{i}(A) (i=1,2i=1,2), hence (Elementary Properties of Series of Real Numbers: Linearity, Null Terms, the Cauchy Criterion, Index Shifts, Tails, Order and Telescoping (i))

∑m=1∞(μ1(Am)−μ2(Am))=μ1(A)−μ2(A).\sum_{m=1}^{\infty}\bigl(\mu_{1}(A_{m})-\mu_{2}(A_{m})\bigr)=\mu_{1}(A)-\mu_{2}(A).

Step 3: restriction to G\mathcal{G}. Let μ\mu be a measure and ν\nu a finite signed measure on (X,F)(X,\mathcal{F}). Since G⊆F\mathcal{G}\subseteq\mathcal{F}, the restrictions are maps μ∣G:G→[0,∞]\mu|_{\mathcal{G}}:\mathcal{G}\to[0,\infty] and ν∣G:G→R\nu|_{\mathcal{G}}:\mathcal{G}\to\mathbb{R} with the same values as μ\mu and ν\nu on G\mathcal{G} (Domains and Restrictions of Maps (ii), Domains and Restrictions of Maps (iii)). As ∅,X∈G\emptyset,X\in\mathcal{G}, we get μ∣G(∅)=0=ν∣G(∅)\mu|_{\mathcal{G}}(\emptyset)=0=\nu|_{\mathcal{G}}(\emptyset) and μ∣G(X)=μ(X)\mu|_{\mathcal{G}}(X)=\mu(X), so μ∣G\mu|_{\mathcal{G}} is finite if μ\mu is, and a probability measure if μ\mu is one. A pairwise disjoint sequence (Am)(A_{m}) in G\mathcal{G} is one in F\mathcal{F}, and its union lies in G\mathcal{G} (σ-Algebra and Measurable Space (i)); since all values involved agree, countable additivity of μ\mu and of ν\nu at (Am)(A_{m}) is countable additivity of μ∣G\mu|_{\mathcal{G}} and of ν∣G\nu|_{\mathcal{G}} at (Am)(A_{m}). This proves (iii).

Step 4: indefinite integrals. Let μ\mu, ff and νf\nu_{f} be as in (iv); ff is a measurable map X→RX\to\mathbb{R} (The Borel Sets of ℝ Form a σ-Algebra, and Integrable Functions Are Measurable Real-Valued Maps (iii)). Let A∈GA\in\mathcal{G}. As A∈FA\in\mathcal{F}, ff is integrable on AA (Integrals over Measurable Subsets (vi)), so νf(A)\nu_{f}(A) is a real number and νf\nu_{f} is a map G→R\mathcal{G}\to\mathbb{R}. Moreover f1A∈L1(μ)f\mathbf{1}_{A}\in\mathcal{L}^{1}(\mu) and

∫Af dμ=∫Xf1A dμ\int_{A}f\,d\mu=\int_{X}f\mathbf{1}_{A}\,d\mu

(Integrals over Measurable Subsets (ii)); in particular f1Af\mathbf{1}_{A} is measurable (The Borel Sets of ℝ Form a σ-Algebra, and Integrable Functions Are Measurable Real-Valued Maps (iii)). As μ(∅)=0\mu(\emptyset)=0, νf(∅)=0\nu_{f}(\emptyset)=0 by Integrals over Measurable Subsets (ix). Let (Am)(A_{m}) be a pairwise disjoint sequence in G\mathcal{G} with union A∈GA\in\mathcal{G}, and put Bn=A1∪⋯∪An∈GB_{n}=A_{1}\cup\dots\cup A_{n}\in\mathcal{G}. By Integrals over Measurable Subsets (vii) and induction on nn, then by what was just shown with BnB_{n} in place of AA,

∑m=1n∫Amf dμ=∫Bnf dμ=∫Xf1Bn dμ.\sum_{m=1}^{n}\int_{A_{m}}f\,d\mu=\int_{B_{n}}f\,d\mu=\int_{X}f\mathbf{1}_{B_{n}}\,d\mu .

For each x∈Xx\in X the sequence (f1Bn(x))n(f\mathbf{1}_{B_{n}}(x))_{n} is constant from some index on, with value f1A(x)f\mathbf{1}_{A}(x): equal to f(x)f(x) from the index mm with x∈Amx\in A_{m} on if x∈Ax\in A, and 00 if x∉Ax\notin A. Moreover the measurable functions f1Bnf\mathbf{1}_{B_{n}} satisfy ∣f1Bn∣≤∣f∣|f\mathbf{1}_{B_{n}}|\le|f|, and ∣f∣∈L1(μ)|f|\in\mathcal{L}^{1}(\mu) by Linearity and Monotonicity of the Lebesgue Integral, and the Integrability Criterion (x). By dominated convergence (Dominated Convergence Theorem (iii)),

∫Xf1Bn dμ→∫Xf1A dμ=∫Af dμ,\int_{X}f\mathbf{1}_{B_{n}}\,d\mu\to\int_{X}f\mathbf{1}_{A}\,d\mu=\int_{A}f\,d\mu ,

so the series ∑mνf(Am)\sum_{m}\nu_{f}(A_{m}) converges with sum νf(A)\nu_{f}(A). Thus νf\nu_{f} is a finite signed measure on (X,G)(X,\mathcal{G}). Finally, μ∣G\mu|_{\mathcal{G}} is a measure on (X,G)(X,\mathcal{G}) by Step 3, and if A∈GA\in\mathcal{G} with μ∣G(A)=μ(A)=0\mu|_{\mathcal{G}}(A)=\mu(A)=0, then νf(A)=0\nu_{f}(A)=0 by Integrals over Measurable Subsets (ix); so νf≪μ∣G\nu_{f}\ll\mu|_{\mathcal{G}}.

Step 5: nonnegative integrands. Let moreover f(x)≥0f(x)\ge0 for every x∈Xx\in X. As ff is measurable, for A∈GA\in\mathcal{G} the number νf(A)\nu_{f}(A) is also the integral over AA of f∈M+f\in\mathcal{M}^{+} (Integrals over Measurable Subsets (iii)), an element of [0,∞][0,\infty]; so νf\nu_{f} is a map G→[0,∞]\mathcal{G}\to[0,\infty] with real values, and νf(∅)=0\nu_{f}(\emptyset)=0. For a pairwise disjoint sequence (Am)(A_{m}) in G\mathcal{G} with union AA, the real series ∑mνf(Am)\sum_{m}\nu_{f}(A_{m}) converges with sum νf(A)\nu_{f}(A) by Step 4, so νf(A)\nu_{f}(A) is also the sum ∑m∈Nνf(Am)\sum_{m\in\mathbb{N}}\nu_{f}(A_{m}) in [0,∞][0,\infty] (Sums of Sequences in [0,∞]: Real Series, Linearity, Comparison, Rearrangement, Double Series and Monotone Convergence (vi)). Hence νf\nu_{f} is a measure on (X,G)(X,\mathcal{G}); it is finite because νf(X)\nu_{f}(X) is real (Small Facts about [0,∞]: Zero, Finiteness and Sums of Zeros (ii)), and νf≪μ∣G\nu_{f}\ll\mu|_{\mathcal{G}} because νf\nu_{f} vanishes on the sets A∈GA\in\mathcal{G} with μ∣G(A)=0\mu|_{\mathcal{G}}(A)=0, as shown in Step 4.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…