0

I need one last lemma for the proof of finite dimensional subspaces are closed:

Is it true that if a subset is open so is its balanced core??

C-star-W-star
  • 16,275

1 Answers1

2

Yes, the balanced core of an open set is open.

Let $x$ be a point in the balanced core $U^\ast$ of $U$.

Then $K = \overline{\mathbb{D}}\cdot x = \{\lambda\cdot x : \lvert\lambda\rvert \leqslant 1\}$ is a subset of $U$, and $K$ is balanced, so $K\subset U^\ast$.

Furthermore, as the image of the compact set $\overline{\mathbb{D}}$ under the continuous map $\lambda \mapsto \lambda\cdot x$, $K$ is quasicompact.


Interlude: Some facts about topological vector spaces we need

In a topological vector space, Hausdorff or not, the balanced open neighbourhoods of $0$ form a neighbourhood basis at $0$: Let $N$ be a neighbourhood of $0$. By the continuity of scalar multiplication in $(0,0)$, there is an $\varepsilon > 0$ and a neighbourhood $N_1$ of $0$ such that $N_2 := D_\varepsilon(0)\cdot N_1 \subset N$. $N_2$ is balanced, since $D_\varepsilon(0)$ is balanced.

The interior of a balanced set $B$ is balanced if and only if it is empty or $0\in\overset{\large\circ}{B}$, for $A = \bigcup\limits_{0 < \lvert\lambda\rvert\leqslant 1}\lambda\cdot \overset{\large\circ}{B}$ is open as the union of open sets, and contained in $B$ since $B$ is balanced, hence $A\subset \overset{\large\circ}{B}$, and, if the interior of $B$ is not empty, $\overline{\mathbb{D}}\cdot \overset{\large\circ}{B} = A\cup \{0\}$ is therefore a subset of $\overset{\large\circ}{B}$ if and only if $0\in \overset{\large\circ}{B}$.

Thus $\overset{\large\circ}{N}_2$ is an open balanced neighbourhood of $0$ contained in $N$. Since $N$ was arbitrary, the claim about the neighbourhood basis is proved.


Back to the main proof:

Since $U$ is open, for every $k\in K$, there is a neighbourhood $V_k$ of $0$ such that $k + V_k \subset U$. By the continuity of addition, there is a neighbourhood $W_k$ of $0$ with $W_k + W_k \subset V_k$. By the interlude, we may assume $W_k$ as open and balanced.

Then $\mathfrak{W} = \{ k + W_k : k \in K\}$ is an open cover of $K$. Since $K$ is quasicompact, there is a finite subset $F = \{k_1,\dotsc,k_n\}$ of $K$ such that $\mathfrak{F} = \{ k + W_k : k\in F\}$ is also an open cover of $K$. Define

$$W := \bigcap_{k\in F} W_k.$$

As a finite intersection of balanced open sets, $W$ is also a balanced open set. And we have

$$K + W \subset U.$$

Namely, for every $y \in K$, there is a $k\in F$ such that $y \in k + W_k$, since $\mathfrak{F}$ is a cover of $K$, and therefore

$$y + W \subset (k+ W_k) + W \subset k + W_k + W_k \subset k + V_k \subset U.$$

Since $K$ and $W$ are both balanced, $K+W$ is balanced, therefore $K+W\subset U^\ast$.

Since $W$ is open, $K+W$ is open, and hence $K+W \subset \left(U^\ast\right)^{\Large\circ}$, thus, finally,

$$x \in K \subset K+W \subset \left(U^\ast\right)^{\Large\circ}$$

shows that $U^\ast$ is open, since $x\in U^\ast$ was arbitrary.

Daniel Fischer
  • 206,697
  • I guess you're assuming Hausdorffness in here in by using the separation property on the closed complement $S^\complement$ and the compact $K$ with $K\cap S^\complement=\varnothing$, do you? – C-star-W-star Jun 24 '14 at 22:27
  • Yes and no. Yes: I assumed Hausdorff because otherwise, finite-dimensional subspaces are not always closed. No: let $N = \overline{{0}}$, and $F = E/N$, $\pi\colon E \to F$ the canonical projection. Then $\pi(K)$ is closed and compact, $\pi(S)$ is open and contains $\pi(K)$. Take $\tilde{U}\subset F$ open, balanced, small enough that $\pi(K) + \tilde{U} \subset \pi(S)$. Then $U = \pi^{-1}(\tilde{U})$ is a balanced open neighbourhood of $0$ in $E$, and $K + U = \pi^{-1}(\pi(K) + \tilde{U}) \subset \pi^{-1}(\pi(S)) = S$. The latter holds because $S$ is open and we factored out $N$, which is – Daniel Fischer Jun 24 '14 at 22:50
  • the intersection of all neighbourhoods of $0$, so $S = S + N$. – Daniel Fischer Jun 24 '14 at 22:51
  • Hmm that seems quite hard proof - I hoped there will be an easy answer =/ Is it easier to prove that the interior of a balanced set is balanced if it is a neigborhood of zero... – C-star-W-star Jun 25 '14 at 00:18
  • You can also use the set $K + N$. That is a quasicompact closed subset of $S$ also in the non-Hausdorff case, and the remainder of the proof is exactly like in the Hausdorff case. – Daniel Fischer Jun 25 '14 at 08:37
  • My add the details on constructing $U$ to your answer (I did work it out now explicitely) Besides you were right there's no need for having a Hausdorff space a hint on that seems hidden in the fact that we only need an open for $K$ not also for the complement of $S$ – C-star-W-star Jun 25 '14 at 14:00
  • Ok the editing got a little bit bigger - I'm sorry for that and I hope you don't mind. I was nethertheless trying not to make a whole new answer out of yours and trying to keep the main procedure of your proof... – C-star-W-star Jun 26 '14 at 18:28
  • Let me know if the edit is ok... – C-star-W-star Jun 26 '14 at 18:30
  • I'm too slow, already rejected (changes too much). Let me see how it looks. – Daniel Fischer Jun 26 '14 at 18:39
  • Shall I redo it here? – C-star-W-star Jun 26 '14 at 22:06
  • I've edited myself and added much detail. I don't think it needs another edit. – Daniel Fischer Jun 26 '14 at 22:08
  • Yeah I read through it and I think so too - now it's really good ;) also nice to read... Thanks! – C-star-W-star Jun 26 '14 at 22:19