*Let E be a prehilbertian space and let Φ be a non-empty, directed decreasing set of non-empty Hausdorff and complete convex subsets of E. For every x∈E and every subset H of E, put d(x,H)=infz∈H∥x−z∥. In order that the intersection M of the sets H belonging to Φ be non-empty, it is necessary and sufficient that there exists x0 in E such that supH∈Φd(x0,H) is finite. For every x∈E we then have
pM(x)=H∈ΦlimpH(x)
(limit with respect to the directed set Φ).
If M is non-empty, d(x,H)≤d(x,M) for all H∈Φ and all x∈E.
Conversely, suppose that there exists a point x0 in E and a real number C≥0 such that d(x0,H)≤C for all H∈Φ. Let x∈E; then
d(x,H)≤∥x−x0∥+Cfor all H∈Φ,
hence the number d=supH∈Φd(x,H) is finite. Let B be the set of all z∈E such that ∥x−z∥≤d. Since B is convex and closed in E, the sets H∩B, for H ranging over Φ, are convex, Hausdorff and complete. Let ε>0; there exists a set H∈Φ such that d(x,H)≥d−ε, and if ε<d/2, the diameter of H∩B is bounded by 12ε(d−ε) by prop. 5 (V, p. 9). In other words, for all H0∈Φ, the closed sets H∩B, for H∈Φ and H⊂H0, form a base of the Cauchy filter on the Hausdorff and complete space H0. Hence the intersection of the sets H∩B (for H∈Φ) reduces to a point y. We get y∈M and ∥x−y∥=d=d(x,M). Since M is closed in H0, it is a Hausdorff, convex and complete set in E, and so y=pM(x). For every H∈Φ, we have pH(x)∈H∩B, from which we get that pM(x)=limH∈ΦpH(x).
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.