Let u be a continuous and positive endomorphism of E, and F the set of all finite rank orthoprojectors on E. Then for every orthonormal basis (ei)i∈I of E, we have (in R+) the equality
i∈I∑⟨ei∣u(ei)⟩=p∈Fsupτ(pup).
For every finite subset J of I, put pJ=∑i∈Jeiei∗; this is the orthoprojector from E onto the vector subspace generated by the vectors ei, where i ranges over J. We have
pJupJ=i∈J,j∈J∑⟨ei∣u(ej)⟩eiej∗, hence
τ(pJupJ)=∑i∈J⟨ei∣u(ei)⟩. Since
pJ∈F,
i∈J∑⟨ei∣u(ei)⟩≤p∈Fsupτ(pup); and so we conclude that
i∈I∑⟨ei∣u(ei)⟩=Jsupi∈J∑⟨ei∣u(ei)⟩≤p∈Fsupτ(pup).
Let v be a finite rank continuous and positive endomorphism of E and let p∈F. By th. 2 of V, p. 23 there exists an orthonormal basis (fα)α∈A of E and a finite subset B of A such that (fα)α∈B is an orthonormal basis of the image of p. Then we have p=∑α∈Bfαfα∗, and so, as above, the relation τ(pvp)=∑α∈B⟨fα∣v(fα)⟩. By lemma 2 (V, p. 48) we have τ(v)=∑α∈A⟨fα∣v(fα)⟩, which gives the formula
α∈B∑⟨fα∣v(fα)⟩≤τ(v). Applying this inequality to the case where
v=pJ.upJ and where J is a finite subset of I, we get
α∈B∑⟨pJ(fα)∣upJ(fα)⟩≤i∈J∑⟨ei∣u(ei)⟩. For every
x∈E, we have
pJ(x)=∑i∈J⟨ei∣x⟩ei, and so
x=limJpJ(x) with respect to the ordered directed set of finite subsets J of I. Passing to the limit over J in (23), we get
τ(pup)=α∈B∑⟨fα∣u(fα)⟩≤i∈I∑⟨ei∣u(ei)⟩, and this completes the proof of lemma 3.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.
Nothing.