Let $H$ be a Hilbert space. I'm looking for an example of a series $\sum x_n$ with the following two properties:
- $\sum x_n$ does not converge, i.e., partial sums do not have a limit in the norm topology of $H$;
- $\sum x_n$ converges unconditionally in the weak topology, that is for every bijection $\sigma:\mathbb{N}\to\mathbb{N}$ the partial sums of the rearranged series $\sum x_{\sigma(n)}$ converge weakly.
Progress so far: considered some weakly but not strongly convergent sequences, and corresponding series. In all cases the weak convergence turned out to be conditional. For example: let $e_n$ be an orthonormal basis, and consider $x_1=e_1$, $x_n = e_n - e_{n-1}$ for $n>1$. Then the partial sums of $\sum x_n$ are precisely the vectors $e_n$ which converge to zero weakly but not in the norm. However, this weak convergence is conditional: if we take only even-numbered elements $$x_2+x_4+x_6+\dots = (e_2-e_1) + (e_4-e_3) + (e_6-e_4)+\dots$$ the norm of partial sums tends to $\infty$; and including an odd-numbered $x_n$ once in a million terms will not change that. Since a weakly convergent sequence must be bounded, the rearranged series does not converge weakly.
There is no such example. Theorem 1, p. 80 Of Normed Linear Spaces by M. M. Day (Third edition) implies that if the series unconditionally convergent in the weak topology then $\sum x_n$ converges in the norm.