Let Ei(1≤i≤n) and F be normed spaces over a non-discrete valued division ring K, and let f be a multilinear mapping of ∏i=1nEi into F. Then f is continuous on ∏i=1nEi if and only if there exists a real number a>0 such that, for all xi∈Ei (1≤i≤n), we have
∣∣f(x1,x2,…,xn)∣∣≤a.∣∣x1∣∣.∣∣x2∣∣…∣∣xn∣∣
(3)
The condition is necessary. For if f is continuous at the point (0,0,…,0) there exists a number b>0 such that the relations ∣∣xi∣∣≤b (1≤i≤n) imply ∣∣f(x1,…,xn)∣∣≤1. Let t0 be an element of K such that 0<∣t0∣<1; then for every point (xi)∈∏i=1nEi such that none of the xi is zero, there exist n rational integers ki such that b∣t0∣<∣∣t0kixi∣∣≤b; consequently we have
∣t0∣k1+k2+⋯+kn∣∣f(x1,x2,…,xn)∣∣≤1;
on the other hand we have ∣t0∣ki1≤b∣t0∣1∣∣xi∣∣, and the relation (3) therefore follows, with a=(1/b∣t0∣)n. This relation is evidently still valid when one of the xi is zero.
The condition is sufficient. We shall show that, if it is satisfied, f is continuous at every point (ai) of ∏i=1nEi. We can write
f(x1,…,xn)−f(a1,…,an)=i=1∑nf(a1,…,ai−1,xi−ai,xi+1,…,xn).
Now, using (3), the conditions ∣∣xi−ai∣∣≤r (1≤i≤n) imply that
∣∣f(a1,…,ai−1,xi−ai,xi+1,…,xn)∣∣≤ark=i∏n(∣∣ak∣∣+r);
hence, if c is the maximum of the numbers ∣∣ai∣∣ (1≤i≤n), we have
∣∣f(x1,…,xn)−f(a1,…,an)∣∣≤nar(c+r)n−1.
Since the right-hand side of this inequality is a polynomial in r with zero constant term, it tends to 0 as r tends to 0; hence f is continuous.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.
Nothing.