Bourbaki

Proposition 13

Tag 091N, permanent. Cite it as [Bourbaki, Tag 091N].

Integration, chapter V, § 3. Integration of positive measures

— Let T,X,YT, X, Y be three locally compact spaces, Λ:t↦λt\Lambda : t \mapsto \lambda_t a bounded diffusion of TT in XX, and H:x↦ηxH : x \mapsto \eta_x a bounded diffusion of XX in YY. The mapping t↦λtHt \mapsto \lambda_t H is then a bounded diffusion of TT in YY, which is denoted by ΛH\Lambda H, and

∥ΛH∥≤∥Λ∥∥H∥. \| \Lambda H \| \leq \| \Lambda \| \| H \|.

Let ff be a universally measurable function ≥0\geq 0 defined on YY, and μ\mu a measure on TT. Suppose that μ\mu belongs to the domain of Λ\Lambda, and that μΛ belongs to the domain of H; then μ belongs to the domain of ΛH, and

⟨μ(ΛH),f⟩=⟨μΛ,Hf⟩=⟨μ,ΛHf⟩; \langle \mu(\Lambda H), f \rangle = \langle \mu \Lambda, Hf \rangle = \langle \mu, \Lambda Hf \rangle;
(15)
(μΛ)H=μ(ΛH);Λ(Hf)=(ΛH)f. (\mu \Lambda)H = \mu(\Lambda H); \quad \Lambda(Hf) = (\Lambda H)f.

Set γt=λtH\gamma_t = \lambda_t H; we shall denote by Γ\Gamma the mapping ΛH\Lambda H of T into M+(Y)\mathscr{M}_+(Y), and by Γf\Gamma f the function t↦⟨γt,f⟩t \mapsto \langle \gamma_t, f \rangle (an abuse of notation, since we do not yet know whether Γ\Gamma is a diffusion). Then ⟨γt,f⟩=⟨λtH,f⟩=⟨λt,Hf⟩\langle \gamma_t, f \rangle = \langle \lambda_t H, f \rangle = \langle \lambda_t, Hf \rangle by (13); since the function HfHf is positive and universally measurable on X (Cor. of Prop. 12), it follows first of all that Γf=Λ(Hf)\Gamma f = \Lambda(Hf), and then that Γf\Gamma f is universally measurable on T (same reference). It is clear that all of the measures γt\gamma_t have total mass at most equal to ∥Λ∥∥H∥\| \Lambda \| \| H \|. Consequently Γg\Gamma g is universally measurable and bounded for every function g∈K+(Y)g \in \mathscr{K}_+(Y); Γ\Gamma is therefore scalarly essentially integrable for every bounded measure on T, and in particular for every measure with compact support. More generally, if μ\mu is a measure in the domain of Λ\Lambda, such that μΛ\mu \Lambda belongs to the domain of H, then, for g∈K+(Y)g \in \mathscr{K}_+(Y),

⟨μ,Γg⟩=⟨μ,Λ(Hg)⟩=⟨μΛ,Hg⟩=⟨(μΛ)H,g⟩. \langle \mu, \Gamma g \rangle = \langle \mu, \Lambda(Hg) \rangle = \langle \mu \Lambda, Hg \rangle = \langle (\mu \Lambda)H, g \rangle.

Since the last quantity is finite, we see that Γ\Gamma is scalarly essentially μ\mu-integrable. Let us denote by μΓ\mu \Gamma the integral ∫γtdμ(t)\int \gamma_t d\mu(t) (an abuse of notation, since we do not yet know whether Γ\Gamma is a diffusion). The preceding relations may then be written

⟨μΓ,g⟩=⟨(μΛ)H,g⟩, \langle \mu \Gamma, g \rangle = \langle (\mu \Lambda)H, g \rangle,

or also μΓ=(μΛ)H\mu \Gamma = (\mu \Lambda)H since gg is arbitrary in K+(Y)\mathscr{K}_+(Y).

Consider anew the universally measurable function f⩾0f \geqslant 0. We have

⟨μΓ,f⟩=⟨(μΛ)H,f⟩=⟨μΛ,Hf⟩=⟨μ,Λ(Hf)⟩=⟨μ,Γf⟩. \langle \mu \Gamma, f \rangle = \langle (\mu \Lambda)H, f \rangle = \langle \mu \Lambda, Hf \rangle = \langle \mu, \Lambda(Hf) \rangle = \langle \mu, \Gamma f \rangle.

When ff is lower semi-continuous and μ\mu runs over the set of positive measures with compact support, these relations express that Γ\Gamma is a diffusion of T in Y. The assertion then does no more than make explicit the relations obtained in the course of the above proof.

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Cites

Transcribed from the printed volume. Not translated.