Bourbaki

Proposition 14

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

Integration, chapter IX, § 1. Premeasures and measures on a topological space

Let f∈F+(T)f \in \mathscr{F}_+(T). If ff is μ\mu-moderated, then μ∗(f)=μ∙(f)\mu^*(f) = \mu^\bullet(f); if ff is not μ\mu-moderated, then μ∗(f)=+∞\mu^*(f) = +\infty.

If μ∗(f)<+∞\mu^*(f) < +\infty, there exists a lower semi-continuous function g≥fg \geq f such that μ∙(g)<+∞\mu^\bullet(g) < +\infty. For every n∈Nn \in \mathbf{N}, let GnG_n be the set of t∈Tt \in T such that g(t)>1/ng(t) > 1/n; the set GnG_n is open, one has μ∙(Gn)≤nμ∙(g)<+∞\mu^\bullet(G_n) \leq n \mu^\bullet(g) < +\infty, and ff is zero outside the union of the GnG_n: the function ff is therefore moderated.

Next, let us show that μ∗\mu^* and μ∙\mu^\bullet have the same value for moderated functions. Since μ∗\mu^* and μ∙\mu^\bullet are encumbrances, it suffices to establish the relation μ∗(f)=μ∙(f)\mu^*(f) = \mu^\bullet(f) when ff is a positive function, bounded above by a constant M, and zero outside an open set G of finite measure, which we shall now do.

The measure μ\mu is the supremum, in M(T)\mathscr{M}(T), of an increasing directed family (μi)i∈I(\mu_i)_{i \in I} of measures with compact support: this follows at once from Prop. 9 of No. 8. Let gg be a lower semi-continuous function on T, between ff and the lower semi-continuous function MφGM \varphi_G. Set νi=μ−μi\nu_i = \mu - \mu_i; then μ∙=μi∙+νi∙\mu^\bullet = \mu_i^\bullet + \nu_i^\bullet (No. 2, Remark 1), consequently

μ∙(g)−μ∙(f)=(μi∙(g)−μi∙(f))+(νi∙(g)−νi∙(f)) \mu^\bullet(g) - \mu^\bullet(f) = (\mu_i^\bullet(g) - \mu_i^\bullet(f)) + (\nu_i^\bullet(g) - \nu_i^\bullet(f))
≤(μi∙(g)−μi∙(f))+νi∙(MφG). \leq (\mu_i^\bullet(g) - \mu_i^\bullet(f)) + \nu_i^\bullet(M \varphi_G).
One has νi∙(MφG)=μ∙(MφG)−μi∙(MφG)\nu_i^\bullet(M \varphi_G) = \mu^\bullet(M \varphi_G) - \mu_i^\bullet(M \varphi_G) and μ∙(MφG)=sup⁡μi∙(MφG)\mu^\bullet(M \varphi_G) = \sup \mu_i^\bullet(M \varphi_G) (No. 7, Prop. 6); the number νi∙(MφG)\nu_i^\bullet(M \varphi_G) may therefore be made arbitrarily small by a suitable choice of ii. Thus everything comes down to showing that one can find, for any number c>0c > 0 and any index i∈Ii \in I, a lower semi-continuous function gg between ff and MφGM \varphi_G, such that μi∙(g)−μi∙(f)≤c\mu_i^\bullet(g) - \mu_i^\bullet(f) \leq c. Now, let L be the compact support of the measure μi\mu_i, and let λ\lambda be the measure (μi)L(\mu_i)_L; since μi\mu_i is concentrated on L, one has μi∙(h)=μi∙(hφL)=λ∙(hL)\mu_i^\bullet(h) = \mu_i^\bullet(h \varphi_L) = \lambda^\bullet(h_L) for every function h∈F+(T)h \in \mathscr{F}_+(T) (No. 1, Lemma 1 and No. 2, Prop. 2); therefore
μi∙(g)−μi∙(f)=λ∙(gL)−λ∙(fL). \mu_i^\bullet(g) - \mu_i^\bullet(f) = \lambda^\bullet(g_L) - \lambda^\bullet(f_L).

¹ The cited appendix in TG does not appear in GT (which translated an earlier edition of Ch. IX). Lindelöf spaces are defined in GT, I, §9, Exer. 14; this exercise and the definition of Souslin space (GT, IX, §6, No. 2, Def. 2) cover the material needed here.

But L is compact; therefore λ∗=λ∗\lambda^* = \lambda^*, consequently there exists a lower semi-continuous function hh defined on L, that is ≥fL\geq f_L and is such that λ∗(h)≤λ∗(fL)+c\lambda^*(h) \leq \lambda^*(f_L) + c. Since the set L is closed in T, the function k equal to h on L, and to +∞+\infty on T−LT - L, is lower semi-continuous on T and is ≥f\geq f, and λ∗(kL)=λ∗(h)≤λ∗(fL)+c\lambda^*(k_L) = \lambda^*(h) \leq \lambda^*(f_L) + c. It remains only to set g=inf⁡(k,MφG)g = \inf(k, M \varphi_G): gg is lower semi-continuous, g≥fg \geq f and

μi∗(g)≤μi∗(k)=λ∗(kL)≤λ∗(fL)+c=μi∗(f)+c. \mu_i^*(g) \leq \mu_i^*(k) = \lambda^*(k_L) \leq \lambda^*(f_L) + c = \mu_i^*(f) + c.

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.