0
$\begingroup$

Let $X$ be a separable Banach space of dimensional $>2$. When does there exist a sequence positive integers $\{N_n\}_{n \in \mathbb{n}}$ such that

  • For any sequence of distinct finite-dimensional subspaces $\{L_n\}_{n \in \mathbb{N}}$ of $X$ we have: $$ N_n\leq \dim(L_n) (\forall n \in \mathbb{N})\,\Rightarrow \,\overline{\bigcup_{n \in \mathbb{N}} L_n} =X. $$ By distict, I simply mean that $L_n\neq L_m$ if $n\neq m$ but we can have containement.
  • There exists a sequence of subspaces $\{\tilde{L}_n\}$ of $X$ for which $$ Dim(\tilde{L}_n)<N_n \mbox{ and } \overline{\bigcup_{n \in \mathbb{N}} \tilde{L}_n} \neq X. $$ Let's call such $\{N_n\}$ critical for $X$.

Finite-Dimensional Intuition:

If $X$ is finite-dimensional with dimension at-least $2$, then WLOG assume $X=\mathbb{R}^D$ for some positive integer $D>2$. Then $N_n=\min\{n,D\}$ works. But if $\tilde{L}_n$ are rotations of $\mathbb{R}^{D-1}\subseteq \mathbb{R}^D$ by $\pi_n$ about the origin, then $\tilde{L}_n$ satisfy the second condition. So $N_n=\min\{n,D\}$ is critical in the above sense.

Note: We can always construct a dense linearly independent subset of $X$ by from a topological basis (since X is infinite-dimensional). So we can find $L_n$ with $1$-dimension each. However, here, I'm looking for a "critical" integer $N$ above which any collection of $\{L_n\}$ must satisfy $\overline{\bigcup_{n \in \mathbb{N}} L_n} =X$.


**Edit: ** I removed the requirement of each $L_n$ to be distinct so that I can give my (updated) intuition in the finite-dimensional setting.

$\endgroup$
5
  • $\begingroup$ Thanks. But your question should be more concrete anyway. Because you can just take a collection of 1-dimensional subspaces with dense union, in which case the second condition holds and the first does not. Also $N$ can be taken to be equal to 1 and then the first part is trivial. $\endgroup$ Commented Aug 9, 2020 at 8:43
  • $\begingroup$ @TarasBanakh Yes I just noticed the bad formulation. I meant to ask if there is a "critical dimension" above which any collection of subspaces of that dimension must have dense union. $\endgroup$
    – ABIM
    Commented Aug 9, 2020 at 8:44
  • $\begingroup$ I think nothing like you wish cannot hold because of the existence of basic sequences in any infinite-dimensional Banach space. Using a basic sequence you can form many finite-dimensional subspaces with desirable properties. $\endgroup$ Commented Aug 9, 2020 at 8:48
  • $\begingroup$ @TarasBanakh Yes, instead if we ask if there is a sequence of dimensions of these subspaces must grow at then I think its possible. I put down the example in the finite-dimensional case also. $\endgroup$
    – ABIM
    Commented Aug 9, 2020 at 9:08
  • 4
    $\begingroup$ How can the first condition be satisfied, if X is infinite dimensional? We can pick each $L_n$ within a fixed closed hyperplane, with any finite dimension $N_n$ we like. $\endgroup$ Commented Aug 9, 2020 at 11:03

0

You must log in to answer this question.