Bourbaki

Proposition 4

Tag 07AZ, permanent. Cite it as [Bourbaki, Tag 07AZ].

Topological Vector Spaces, chapter II, § 1. Semi-norms

Let EiE_i (1≤i≤n1 \leq i \leq n) and FF be topological vector spaces over KK; we suppose that, for every ii, the topology of EiE_i is defined by a directed set of semi-norms Γi\Gamma_i, and that the topology of FF is defined by a set of semi-norms Γ\Gamma. Then a set HH, of multilinear mappings of ∏i=1nEi\prod_{i=1}^n E_i in F is equicontinuous if, and only if, for each semi-norm q∈Γq \in \Gamma, and each index i, there exists a semi-norm pi∈Γip_i \in \Gamma_i, and a number a>0a > 0, such that for each function u∈Hu \in \mathbf{H} and point (xi)∈∏i=1nEi(x_i) \in \prod_{i=1}^n E_i,

q(u(x1,x2,...,xn))≤a⋅p1(x1) p2(x2)...pn(xn). q(u(x_1, x_2, ..., x_n)) \leq a \cdot p_1(x_1)\ p_2(x_2) ... p_n(x_n) .

The condition is sufficient since it implies that H is equicontinuous at (0, 0, ..., 0) and therefore everywhere (I, p. 9, prop. 6).

We show that the condition is necessary. By hypothesis, for every semi-norm q∈Γq \in \Gamma and every number β>0\beta > 0, we have q(u(x1,x2,...,xn))≤βq(u(x_1, x_2, ..., x_n)) \leq \beta for every function u∈Hu \in \mathbf{H} provided that pi(xi)≤αip_i(x_i) \leq \alpha_i are true for each index ii, 1≤i≤n1 \leq i \leq n, and certain appropriately chosen numbers αi>0\alpha_i > 0 and semi-norms pi∈Γip_i \in \Gamma_i. As K is non-discrete, we can also suppose that, for every ii, we have αi=∣λi∣<1\alpha_i = |\lambda_i| < 1 where λi∈K\lambda_i \in \mathbf{K}. Then let (x1,x2,...,xn)(x_1, x_2, ..., x_n) be any point of ∏i=1nEi\prod_{i=1}^n E_i, and for each index ii, let mi∈Zm_i \in \mathbf{Z} be an integer such that pi(xi)≤∣λi∣mi+1p_i(x_i) \leq |\lambda_i|^{m_i + 1}; this can be written as pi(λi−mixi)≤∣λi∣p_i(\lambda_i^{-m_i} x_i) \leq |\lambda_i| (1≤i≤n1 \leq i \leq n), therefore, by hypothesis, we have

q(u(x1,x2,...,xn))≤β∣λ1∣m1∣λ2∣m2...∣λn∣mn. q(u(x_1, x_2, ..., x_n)) \leq \beta |\lambda_1|^{m_1} |\lambda_2|^{m_2} ... |\lambda_n|^{m_n} .

Suppose firstly that one of the pi(xi)p_i(x_i) is zero, then we can take mi∈Nm_i \in \mathbf{N} arbitrarily large, therefore q(u(x1,x2,...,xn))=0q(u(x_1, x_2, ..., x_n)) = 0. If, on the contrary, all the pi(xi)p_i(x_i) are ≠0\neq 0, take the integer mim_i such that ∣λi∣mi+2<pi(xi)≤∣λi∣mi+1|\lambda_i|^{m_i + 2} < p_i(x_i) \leq |\lambda_i|^{m_i + 1} for each ii; then we have ∣λi∣mi<∣λi∣−2pi(xi)|\lambda_i|^{m_i} < |\lambda_i|^{-2} p_i(x_i), from which, by (6), the relation (5) follows with

a=β(∣λ1∣⋅∣λ2∣...∣λn∣)−2. a = \beta (|\lambda_1| \cdot |\lambda_2| ... |\lambda_n|)^{-2} .

Q.E.D.

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Nothing.

Cites

Nothing.

Transcribed from the printed volume. Not translated.