4

Rajendra Bhatia's Notes on Functional Analysis (Texts and Readings in Mathematics), Pg. $14$ states (without proof) that:

Every $\ell^p$ space ($1\le p \le \infty)$ with $p\ne 2$ has a subspace without a Schauder basis.

where $\ell^p$ denotes the sequence space with the $p$-norm.

  1. What is the proof of this fact? I have been trying for a while now, and it seems more difficult than I had imagined.

  2. Also, I would be interested to see why every subspace of $\ell^2$ has a Schauder basis.

P.S. If the proof is doable with some hints, then just hints would be great too!

Update: The proof of the main assertion is found in Lindenstrauss and Tzafriri's Classical Banach Spaces I and II, as mentioned by David Mitra in the comments.

Jose Avilez
  • 13,432
  • 1
    See the wikipedia article on Schauder basis and look for the section on "Relation to Fourier series". This might be a counter-example to the assertion. – Peter Aug 07 '21 at 04:30
  • I don't see how that is related. Could you elaborate? Note that it is enough to produce one subspace of $\ell^p$ without a Schauder basis. Many other subspaces may have a Schauder basis - doesn't matter. – stoic-santiago Aug 07 '21 at 04:33
  • 2
    It is not easy. Lindenstrauss and Tzafriri's Classical Banach Spaces I and II contain a proof. (The stronger assertion, in fact, that they have subspaces without the approximation property.) – David Mitra Aug 07 '21 at 05:01
  • Thanks, David. I do not have access to that book right now. Would you be kind enough to sketch a proof outline, or give some idea on how to proceed? Also, what is the approximation property? – stoic-santiago Aug 07 '21 at 05:08
  • 2
    You can google AP. Here is a link to a proof for $1\le p <2$. As I said, the proof is (far from) easy; I don't have the time to attempt to outline it here. – David Mitra Aug 07 '21 at 05:22

1 Answers1

3

The answer to question (1) is in the comment by @DavidMitra: You can refer to Lindenstrauss' and Tzafriri's Classical Banach Spaces I and II for a proof that every $\ell^p$ space with $p \neq 2$ admits a subspace without a Schauder basis.

The answer to question (2) is far easier: note that every subspace of $\ell^2$ is a separable inner product space. Since every separable inner product space admits a (countable) orthonormal basis (e.g. @mechanodroid answer here), you can take this orthonormal basis as your Schauder basis.

Jose Avilez
  • 13,432
  • Hi, I searched the book but couldn't find a proof of the statement. Can anyone pinpoint the exact chapter and/or page? –  Nov 04 '24 at 05:24
  • @DwaipayanSharma see Theorem 2.d.6, which proves it for $p>2$; a remark at the bottom of the page points you to the case for $1 \leq p < 2$ – Jose Avilez Nov 05 '24 at 06:07