Let there be given an inverse system of topological spaces T=(Ti,pij) indexed by I, a topological space T, a coherent and separating family of continuous mappings pi:T→Ti (for i∈I) and a sub-inverse system (μi)i∈I of measures on T. For every compact subset K of T, set
J(K)=i∈Iinfμi∙(pi(K)).
Then there exists a bounded measure π on T, and only one, such that π∙(K)=J(K) for every compact subset K of T. One has μi≥pi(π) for all i∈I, and π is the largest measure on T satisfying this condition.
Let us first prove that J(K) is the limit of μi∙((pi(K))) with respect to the section filter F of the directed preordered set I: for this, it suffices (GT, IV, §5, No. 2, Th. 2) to show that μi∙(pi(K))≥μj∙(pj(K)) for i≤j; now, setting μij′=pij(μj), we have μij′≤μi and pj(K)⊂pij−1(pi(K)), whence
μj∙(pj(K))≤μj∙(pij−1(pi(K)))=(μij′)∙(pi(K))≤μi∙(pi(K)).
Let us now pass to the study of the properties of the function J: 1) It is clear that J(K)≤J(L) when K⊂L. 2) Let K and L be two compact subsets of T. For every i∈I, we have pi(K∪L)=pi(K)∪pi(L), whence
μi∙(pi(K∪L))≤μi∙(pi(K))+μi∙(pi(L));
passing to the limit with respect to the filter F, we obtain J(K∪L)≤J(K)+J(L). 3) Suppose that the compact sets K and L are disjoint. By Prop. 2 of No. 1, there exists an i∈I such that pj(K)∩pj(L)=∅ for j≥i. For j≥i we therefore have
μj∙(pj(K∪L))=μj∙(pj(K))+μj∙(pj(L)),
whence J(K∪L)=J(K)+J(L) on passing to the limit with respect to the filter F. 4) Let (Kα)α∈A be a decreasing directed family of compact subsets of T, with intersection K. By Prop. 1 of No. 1, pi(K)=⋂α∈Api(Kα) and therefore μi∙(pi(K))=infα∈Aμi∙(pi(Kα)) for all i∈I (§ 1, No. 6, Cor. of Prop. 5). From this, one deduces
J(K)=i∈Iinfμi∙(pi(K))=i∈Iinfα∈Ainfμi∙(pi(Kα))=α∈Ainfi∈Iinfμi∙(pi(Kα))=α∈AinfJ(Kα). 5) Let us choose an
i∈I and set
c=μi∙(Ti). Then
c is finite and
J(K)≤μi∙(pi(K))≤μi∙(Ti), thus
J(K)≤c for every compact set
K in
T.
The preceding properties permit applying Th. 1 of § 3, No. 1; we conclude that there exists one and only one bounded measure π on T such that π∙(K)=J(K) for every compact subset K of T. For every i∈I, let us denote by νi the measure on Ti that is the image of π under pi. Let i∈I, A a compact subset of Ti, and L the set of compact subsets of pi−1(A). By Remark 3 of § 1, No. 2, we have π∙(pi−1(A))=supK∈Lπ∙(K); moreover, νi∙(A)=π∙(pi−1(A)) and J(K)=π∙(K) for K∈L, whence νi∙(A)=supK∈LJ(K).
For K∈L, we have pi(K)⊂A, whence J(K)≤μi∙(pi(K))≤μi∙(A) and finally νi∙(A)≤μi∙(A). Since A is an arbitrary compact set in Ti, we conclude that νi≤μi. The last assertion of the proposition is obvious.
Q.E.D.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.