— Let μ be a positive measure on X, F a real or complex Banach space, F' its strong dual, and (z,z′)↦⟨z,z′⟩ the canonical bilinear form on F×F′.
1° For every function f∈LFp (1≤p≤+∞),
Np(f)=sup∫⟨f,g⟩dμ
as g runs over the set of functions in LF′q, such that Nq(g)≤1.
2° For every function g∈LF′q (1≤q≤+∞),
Nq(g)=sup∫⟨f,g⟩dμ
as f runs over the set of functions in LFp such that Np(f)≤1.
Let us first prove the relation (9); we distinguish between two cases.
(i) 1≤p<+∞. The relation (9) being trivial when Np(f)=0 (because f and ⟨f,g⟩ are then negligible), we can always suppose, on multiplying f by a scalar, that Np(f)=1. Suppose first that f is an integrable step function, f=∑k=1nakφAk, where the Ak are pairwise disjoint (§ 4, No. 9, Lemma). Thus ∑k=1n∣ak∣pμ(Ak)=1 by hypothesis. For every ε>0, there exists (for every index k) a vector ak′∈F′ such that ∣ak′∣q=∣ak∣p if p>1 (resp. ∣ak′∣=1 if p=1) and ⟨ak,ak′⟩≥(1−ε)∣ak∣⋅∣ak′∣ (TVS, IV,
§1, No. 3, Prop. 8). Setting g=∑k=1nak′φAk, we have ∑k=1n∣ak′∣qμ(Ak)=1 if p>1 (resp. sup1≤k≤n∣ak′∣=1 if p=1), thus Nq(g)=1; on the other hand,
and, since ∣ak′∣=∣ak∣p/q=∣ak∣p−1 if p>1 (resp. ∣ak′∣=1 if p=1), we have
∫⟨f,g⟩dμ≥(1−ε)k=1∑n∣ak∣pμ(Ak)=(1−ε)Np(f)=1−ε,
which proves the relation (9) in this case.
Let us pass to the case that f is any element of LFp such that Np(f)=1. For every ε>0, there exists a step function f1∈LFp such that Np(f−f1)≤ε (§ 4, No. 10, Cor. 1 of Prop. 19). By what we have just seen, there exists a function g∈LFq, such that Nq(g)=1 and
∫⟨f1,g⟩dμ≥Np(f1)−ε≥1−2ε.
Now,
∫⟨f,g⟩dμ=∫⟨f1,g⟩dμ+∫⟨f−f1,g⟩dμ
and, by (6),
∫⟨f−f1,g⟩dμ≤Np(f−f1)Nq(g),
whence
∫⟨f,g⟩dμ≥1−3ε,
which proves (9).
(ii) p=+∞. We may again restrict ourselves to the case that N∞(f)>0. Let α be any number such that 0<α<N∞(f); by hypothesis, the set of x∈X such that ∣f(x)∣>α is measurable and is not locally negligible, therefore it contains a compact set K of measure >0. Since f is measurable, there exists a compact set K1⊂K of measure >0, such that the restriction of f to K1 is continuous. It follows that, for every ε>0, there exists a partition of K1 into a finite number of integrable sets, in each of which the oscillation of f is ≤ε; at least one of these sets A has measure > 0. Let a be one of the values of f in A; then ∣a∣>α and ∣f(x)−a∣≤ε for all x∈A. There exists a vector a′∈F′ such that ∣a′∣=1 and ∣⟨a,a′⟩∣≥∣a∣−ε; the function g=φA⋅a′/μ(A) is integrable and N1(g)=1; on the other hand,
∫⟨f,g⟩dμ=μ(A)1∫⟨f,a′⟩φAdμ.
Now, one can write
∫⟨f,a′⟩φAdμ=⟨a,a′⟩μ(A)+∫⟨f−a,a′⟩φAdμ,
and since
∣⟨f−a,a′⟩φA∣≤εφA,
we see that
∫⟨f,g⟩dμ≥∣⟨a,a′⟩∣−ε≥∣a∣−2ε>α−2ε;
since ε is arbitrary and α is any number < N∞(f), the relation (9) is also verified in this case.
One argues in exactly the same manner to prove the relation (10), on considering separately the case 1≤q<+∞ and the case q=+∞, and using the fact that for every z′∈F′, ∣z′∣=sup∣z∣≤1∣⟨z,z′⟩∣ by the definition of the norm in F′.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.