6
$\begingroup$

Let $\{e_{n}(x)\}_{n=0}^{\infty}$ be orthnormal basis of Hilbert space $L^2(\mathbb{R})$. If $\{e_{n}(x)\}_{n=0}^{\infty} \subset L^p(\mathbb{R})$ for some $p\geq 1$, is the $\{e_{n}(x)\}_{n=0}^{\infty}$ Schauder basis for $L^p(\mathbb{R})$?

Any reference?

$\endgroup$

2 Answers 2

2
$\begingroup$

One can derive the answer "Not in general" working out suitable variations of the following lemma.

Lemma. Let $X$ be a separable Banach space which is not isomorphic to a Hilbert space, but is continuously and injectively embedded into a Hilbert space $H$. Then $X$ contains a sequence $\{x_i\}_{i=1}^\infty$, which is linearly independent, has dense span in $X$, is an orthonormal sequence in $H$, but is not a Schauder basis in $X$ because the linear spans of finite subsets $\{x_i\}_{i=1}^n$ have indefinitely increasing projection constants in $X$.

Proof: By the Lindenstrauss-Tzafriri theorem [Israel J. Math. 9 (1971), 263–269] we can find a sequence $\{y_i\}_{i=1}^\infty$ in $X$ with the dense linear span satisfying the last condition. Applying the Gram-Schmidt ortonormalization, we get the desired sequence $\{x_i\}_{i=1}^\infty$.

$\endgroup$
5
  • $\begingroup$ Mikhail, good answer for a student who asks a question that the student probably should have worked on before asking. More challenging is to have no reordering be a Schauder basis. $\endgroup$ Jan 3, 2018 at 23:32
  • $\begingroup$ Hm, is it just orthonormal sequence or an orthonormal basis in $H$? $\endgroup$ Jan 3, 2018 at 23:50
  • $\begingroup$ @Fedor Petrov In this setting it will be an orthonormal sequence only, since I did not require that $X$ is dense in $H$. To get an answer to the Question one has to work because the "continuously embedded" assumption is not satisfied for $L^2(\mathbb{R})$ and $L^p(\mathbb{R})$, but it is satisfied for certain subspaces. I just wanted to show one of the ways to achieve non-Schauderness. $\endgroup$ Jan 4, 2018 at 0:41
  • $\begingroup$ @BillJohnson It seems that it's not true in general. But does there exist an orthonormal basis in $L^2(\mathbb{R})$ that is also a Schauder basis in $L^p(\mathbb{R})$? How about the case when $ L^2(\mathbb{R})$ and $L^p(\mathbb{R})$ are replaced by $L^2(\mathcal{O})$ and $L^p(\mathcal{O})$ , where $\mathcal{O}$ is a domain with smooth boundary? $\endgroup$ Sep 22, 2023 at 14:31
  • $\begingroup$ Sure. The Haar system is a basis in $L^p(0,1)$ for all $1\le p < \infty$. Piece Haar systems on disjoint intervals to get a basis on $L^p(-\infty,\infty)$. $\endgroup$ Sep 22, 2023 at 19:20
1
$\begingroup$

I am afraid that trigonometric system (multiplied by characteristic functions of the segments $[2\pi k,2\pi(k+1)]$ to become a basis in $L^2$) is not a basis in $L^1$.

$\endgroup$
3
  • $\begingroup$ And for some p>1? $\endgroup$ Jan 3, 2018 at 21:38
  • $\begingroup$ Trigonometric system is a Schauder basis for all $L^p,p>1$. Probably we may check other specific example like Walsh. $\endgroup$ Jan 3, 2018 at 23:19
  • $\begingroup$ A uniformly bounded orthonormal system that is unconditional in $L^p$ must be equivalent to the unit vector basis of $\ell^2$ (e.g., use type-ecotype theory), so a basis of characters cannot be unconditional. Therefore there is a permutation of any character basis for $L^p$ that is not a basis for $L^p$ when $p\not=2$. $\endgroup$ Jan 4, 2018 at 13:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.