Bourbaki

Theorem 2

Tag 09C7, permanent. Cite it as [Bourbaki, Tag 09C7].

Integration, chapter VI, § 3. Disintegration of measures

— *Let T and B be two locally compact spaces having countable bases, μ\mu a positive measure on T, pp a μ\mu-measurable mapping of T into B, and ν\nu a pseudo-image measure of μ\mu under pp. Then there exists a ν\nu-adequate family b↦λbb \mapsto \lambda_b (b∈Bb \in B) of positive measures on T, having the following properties: a) λb≠0\lambda_b \neq 0 for b∈p(T)b \in p(T); b) λb\lambda_b is concentrated on the set p−1‾(b)\overline{p^{-1}}(b) for every b∈Bb \in B; in particular, λb=0\lambda_b = 0 for b∉p(T)b \notin p(T); c) μ=∫λb dν(b)\mu = \int \lambda_b \, d\nu(b).

Moreover, if ν′=r⋅ν\nu' = r \cdot \nu is a second pseudo-image measure of μ\mu under pp, and if b↦λb′b \mapsto \lambda'_b is a ν′\nu'-adequate family of positive measures on T having the properties b) and c) with respect to ν′\nu', then λb=r(b)λb′\lambda_b = r(b)\lambda'_b almost everywhere in B (for ν\nu or ν′\nu').

There exists a continuous and finite numerical function ff defined on T, such that f(t)>0f(t) > 0 for every t∈Tt \in T and such that μ′′=f⋅μ\mu'' = f \cdot \mu is bounded (Ch. V, §5, No. 6, Prop. 11). Let ν′′=p(μ′′)\nu'' = p(\mu''), which is equivalent to ν\nu, and write ν′′=g⋅ν\nu'' = g \cdot \nu, with gg finite and locally ν\nu-integrable; one can suppose, moreover, that g(b)>0g(b) > 0 for all b∈Bb \in B (Ch. V, §5, No. 6, Prop. 10). Th. 1 of No. 1, applied to μ′′\mu'' and ν′′\nu'', shows that there exists a ν′′\nu''-adequate family b↦λb′′b \mapsto \lambda''_b (b∈Bb \in B) of positive measures on T, such that:

1) ∥λb′′∥=1\| \lambda''_b \| = 1 for b∈p(T)b \in p(T); 2) λb′′\lambda''_b is concentrated on p−1‾(b)\overline{p^{-1}}(b) for every b∈Bb \in B; 3) μ′′=∫λb′′ dν′′(b)\mu'' = \int \lambda''_b \, d\nu''(b).

For every b∈Bb \in B, let us define a positive measure λb\lambda_b on T by the formula λb=(1/f)⋅(g(b)λb′′)\lambda_b = (1/f) \cdot (g(b)\lambda''_b). It is clear that the family b↦λbb \mapsto \lambda_b has the properties a) and b) of the statement. On the other hand, for every function h∈K(T)h \in \mathscr{K}(T), h/fh/f belongs to K(T)\mathscr{K}(T), therefore

∫h(t) dμ(t)=∫(h(t)/f(t)) dμ′′(t)=∫dν′′(b)∫(h(t)/f(t)) dλb′′(t). \int h(t) \, d\mu(t) = \int (h(t)/f(t)) \, d\mu''(t) = \int d\nu''(b) \int (h(t)/f(t)) \, d\lambda''_b(t).

But since the function b↦∫(h(t)/f(t)) dλb′′(t)b \mapsto \int (h(t)/f(t)) \, d\lambda''_b(t) is ν′′\nu''-integrable, the function b↦g(b)∫(h(t)/f(t)) dλb′′(t)b \mapsto g(b) \int (h(t)/f(t)) \, d\lambda''_b(t) is ν\nu-integrable (Ch. V, §5, No. 3, Th. 1).

By the definition of λb\lambda_b, this function is b↦∫h(t)dλb(t)b \mapsto \int h(t) d\lambda_b(t), whence (loc. cit.) ∫h(t)dμ(t)=∫dν(b)∫h(t)dλb(t)\int h(t) d\mu(t) = \int d\nu(b) \int h(t) d\lambda_b(t), which proves that μ=∫λbdν(b)\mu = \int \lambda_b d\nu(b).

To establish the second part of the statement, we remark that one can suppose that r(b)>0r(b) > 0 for all b∈Bb \in B (Ch. V, §5, No. 6, Prop. 10); set λb′′′=f⋅((r(b)/g(b))λb′)\lambda'''_b = f \cdot \left( (r(b)/g(b)) \lambda'_b \right); one shows, as above, that for every function h∈K(T)h \in \mathscr{K}(T), the relation

∫h(t)dμ(t)=∫dν′(b)∫h(t)dλb′(t) \int h(t) d\mu(t) = \int d\nu'(b) \int h(t) d\lambda'_b(t)

implies

∫h(t)dμ(t)=∫dν′′(b)∫(h(t)/f(t))dλb′′′(t). \int h(t) d\mu(t) = \int d\nu''(b) \int (h(t)/f(t)) d\lambda'''_b(t) .

Therefore Th. 1 of No. 1, applied to μ′′\mu'' and ν′′\nu'', implies that for almost every b∈Bb \in B, λb′′′=λb′′\lambda'''_b = \lambda''_b, whence λb=r(b)λb′\lambda_b = r(b)\lambda'_b.

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.