Directions of uniform rotundity in direct sums of normed spaces (Q1818249)

Two definitions (the second one complicated) are needed. A normed space is said to be uniformly rotund (uniformly convex) in the direction \(z\) (\(\|z\|= 1\)) if \[ \delta(z,\varepsilon):=\inf\{1- \|(x+y)/2\|: \|x\|\leq 1,\;\|y\|\leq 1,\;x-y = \lambda z,\;|\lambda|\geq \varepsilon\} \] is strictly positive for all \(\varepsilon \in (0,2]\). Secondly, let \(\{(X_i,\|. \|_i)\}_{i \in I}\) be an arbitrary family of normed spaces. Let \((E,\|. \|_E)\) be a space such that (i) \(E\) is an (order) ideal in \(\mathbb R^I\) and (ii) the norm is monotonic (if \(f(i) \in E\) and if \(|g(i)|\leq |f(i)|\forall i\) then \(g(i) \in E\) and \(\|g\|_E \leq \|f\|_E\)). Then the \(E\) direct sum of the \(X_i\)'s, \(E(X_i)\), is that subspace of \(\prod X_i\) such that \(x = (x_i) \in E(X_i)\) iff the function \(f_x : i \mapsto \|x_i\|_i\) is in \(E\) and then the norm on \(E(X_i)\) is defined by \(\|x\|= \|(x_i)\|:= \|f_x\|_E\). The paper gives a sufficient condition for \(E(X_i)\) to be uniformly rotund in a direction based on the uniform rotundity directions of \(E\) and the \(X_i\)'s. The condition is necessary if \(E = \ell_{\infty}\) and not so if \(E = \ell_1\).