As stated, I know of course weakly closed implies weakly sequentially closed, but is the converse also true? In other words, I want to know if in weak topology, a subset is weakly closed if and only if it is sequentially closed. aka they are equivalent definitions just analogous to the case in norm topology. I also know that it is not true to say sequentially closed imply closed in ALL topological spaces. But is this true specifically in weak topology? Answers appreciated.
Does weakly sequentially closed imply weakly closed?
1.5k Views Asked by Bumbble Comm https://math.techqa.club/user/bumbble-comm/detail AtThere are 2 best solutions below
On
It isn't true in general.
Let $H$ be an infinite-dimensional Hilbert space and let $(e_n)_n$ be an orthonormal sequence in $H$. Consider $S = \{\sqrt{n}e_n : n \in \mathbb{N}\} \subseteq H$.
$S$ is weakly sequentially closed because every weakly convergent sequence in $S$ has to be bounded, so the sequence has to have at most finitely many distinct terms, and hence it stabilizes to an element of $S$.
On the other hand, $0$ is in the weak closure of $S$. Indeed, consider a basic weakly open neighbourhood $U(0, a_1,\ldots, a_k, \varepsilon) = \{x \in H : |\langle x, a_i\rangle| < \varepsilon, 1 \le i \le k\}$ with $a_1, \ldots, a_k \in H$ and $\varepsilon > 0$.
Since $\sum_{n=1}^\infty |\langle a_i, e_n\rangle|^2 \le \|a_i\| < +\infty$, the series $\sum_{n=1}^\infty\sum_{i=1}^k |\langle a_i, e_n\rangle|^2$ also converges so there exists $n\in\mathbb{N}$ such that $$\sum_{i=1}^k |\langle a_i, e_n\rangle|^2 < \frac{\varepsilon^2}n$$ (otherwise we would have $\sum_{n=1}^\infty\sum_{i=1}^k |\langle a_i, e_n\rangle|^2 \ge \sum_{n=1}^\infty \frac{\varepsilon^2}n = +\infty$.)
For this $n$ we have $|\langle a_i, e_n\rangle| < \frac{\varepsilon}{\sqrt{n}}, \forall 1 \le i \le k$ so $\sqrt{n}e_n \in U(0, a_1,\ldots, a_k, \varepsilon)$.
Therefore every basic weakly open neighbourhood of $0$ intersects $S$ so $0 \in \overline{S}^w$. We conclude that $S$ is not weakly closed.
The converse is false, even in Hilbert space. See Example 3.33 in Bauschke-Combettes's Convex Analysis and Monotone Operator Theory in Hilbert Spaces (second edition), which is a more general version of @mechanodroid Example.
The converse is true in finite-dimensional Banach spaces or when the set considered is convex.