Bourbaki

Theorem 2

Tag 0929, permanent. Cite it as [Bourbaki, Tag 0929].

Integration, V, p. 37, § 4. Integration of positive point measures

Let (π,g)(\pi, g) be a μ\mu-adapted pair, and let

ν=∫g(t)επ(t) dμ(t). \nu = \int g(t) \varepsilon_{\pi(t)} \, d\mu(t).
Let ff be a function defined on XX, with values in a Banach space FF or in R‾\overline{\mathbf{R}}. For ff to be essentially ν\nu-integrable, it is necessary and sufficient that t↦f(π(t))g(t)t \mapsto f(\pi(t))g(t) be essentially μ\mu-integrable, in which case
∫f(x) dν(x)=∫f(π(t))g(t) dμ(t). \int f(x) \, d\nu(x) = \int f(\pi(t))g(t) \, d\mu(t).
Suppose, moreover, that π\pi is continuous and proper, and that gg is continuous and such that g(t)>0g(t) > 0 for all t∈Tt \in T. Then, for ff to be ν\nu-integrable, it is necessary and sufficient that t↦f(π(t))g(t)t \mapsto f(\pi(t))g(t) be μ\mu-integrable.

A) We begin by treating the case that the measure μ\mu has compact support KK, on which gg is bounded. The measures μ\mu and ν\nu are then bounded, and one can replace 'essentially integrable' in the statement by 'integrable'. Suppose that ff is ν\nu-integrable: the function f(π(t))g(t)f(\pi(t))g(t) is then μ\mu-integrable, and the relation (9) is verified, by Th. 1 of §3, No. 3. Conversely, suppose that f(π(t))g(t)f(\pi(t))g(t) is μ\mu-integrable: ff is then ν\nu-measurable (No. 3, Prop. 3 and Remark), and

∫∙∣f(x)∣ dν(x)=∫∙∣f(π(t))∣g(t) dμ(t)<+∞ \int^\bullet |f(x)| \, d\nu(x) = \int^\bullet |f(\pi(t))|g(t) \, d\mu(t) < +\infty
(No. 2, Th. 1); ff is therefore essentially ν\nu-integrable (§1, No. 3, Prop. 9), hence ν\nu-integrable. Th. 1 of §3, No. 3 then implies (9).

B) Let us pass to the general case. Let K\mathscr{K} be the set of compact subsets KK of TT such that g∣Kg|K is continuous: K\mathscr{K} is μ\mu-dense (Ch. IV, §5, No. 10, Prop. 15), therefore the measure μ\mu is the sum of a family (μα)α∈A(\mu_\alpha)_{\alpha \in A} of measures whose supports are elements of K\mathscr{K} (§2, No. 3, Prop. 4). The pair (g,π)(g, \pi) is obviously μα\mu_\alpha-adapted for every α∈A\alpha \in A, and the measure ν\nu is the sum of the family of measures να=∫επ(t)g(t) dμα(t)\nu_\alpha = \int \varepsilon_{\pi(t)}g(t) \, d\mu_\alpha(t) (§3, No. 1, Cor. of Prop. 1). Since the argument of A) may be applied to the measures μα,να\mu_\alpha, \nu_\alpha, the first part of the statement then follows from Prop. 3 of §2, No. 2.

For the function ff (resp. t↦f(π(t))g(t)t \mapsto f(\pi(t))g(t)) to be integrable for ν\nu (resp. for μ\mu), it is necessary and sufficient that it be essentially integrable and that

∫∗∣f(x)∣dν(x)<+∞(resp. ∫∗∣f(π(t))∣g(t) dμ(t)<+∞). \int^* |f(x)| d\nu(x) < +\infty \quad (\text{resp. } \int^* |f(\pi(t))|g(t)\,d\mu(t) < +\infty ).

The second part of the statement therefore follows from the first part and Proposition 2.

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.