The $\mathcal{L}_p$-spaces ($1\leq p \leq \infty$) are Banach spaces $X$ such that there exists a constant $\lambda$ so that every finite dimensional subspace $E$ of $X$ is contained in another finite dimensional subspace $F$ which is $\lambda$-isomorphic to $\ell_p^n$.

Let us say that a Banach space $X$ locally admits a basis if there exists a constant $\lambda$ so that every finite dimensional subspace $E$ of $X$ is contained in another finite dimensional subspace $F$ admitting a basis with basis constant $\leq \lambda$.

Each $\mathcal{L}_p$-space locally admits a basis, and it is not difficult to show that if $X$ has a basis then it locally admits a basis.


  1. Suppose that $X$ is separable and locally admits a basis. Does $X$ have a basis?
  2. Suppose that $X$ is separable. Does $X$ locally admit a basis?

The answer to Question 1 is positive for $\mathcal{L}_p$-spaces.

  • $\begingroup$ Professor Gonzalez, as far as I can remember, both questions have a negative answer in Szarek's paper where he provided a space with BAP & without a basis. $\endgroup$
    – Onur Oktay
    Mar 6 at 23:50
  • $\begingroup$ Thank you for your comment. $\endgroup$ Mar 7 at 9:12
  • $\begingroup$ I have looked at Szarek's paper. I think you should write an answer, trying to say something about the second question using Kalton's review in MathScinet to one of Pujara's papers. $\endgroup$ Mar 7 at 9:13
  • $\begingroup$ I'm afraid I don't have an access to MathSciNet, and I don't have a copy of Kalton's review in another form. Presently I can record here the exact reference to Szarek's article doi.org/10.1007/BF02392555 Corollary 1.6 within nicely sums up the answers to both questions -- for future visitors to this MO post. $\endgroup$
    – Onur Oktay
    Mar 9 at 1:20
  • $\begingroup$ Szarek's result aside, perhaps Enflo's space is another example of a separable space with local basis structure (by Proposition 1.3 in doi.org/10.1007/BF02392555) & without basis? $\endgroup$
    – Onur Oktay
    Mar 9 at 1:28

1 Answer 1


The purpose of this post is to make an expanded summary of the comments to OP on request of Professor Gonzalez.

Local basis structure (LBS) was considered by Pujara in 1975 under a different name as pointed out by Prof. Gonzalez in his comment . To the best of my knowledge, both questions were answered negatively by Szarek in 1987. The snippet below is from [Szarek1987].

enter image description here

Imho, the main actor in Szarek's construction is the space $(\bigoplus Y_{p_k}^{n_k})_{\ell^2}$ that lacks LBS, where $Y_p^n$ are carefully chosen $n$-dimensional subspaces of $L^p$ (Proposition 3.1) and $p_k\to 2$, $n_k\to\infty$ are carefully chosen sequences (Proposition 4.1). It is worth to note Remark 5.1 that points out other nonreflexive examples, cf. Professor Bill Johnson's comment.

enter image description here

Perhaps Szarek's construction should be read along with the asymptotically Hilbertian spaces in [CasazzaGarciaJohnson2001].

For the future visitors to this post: interested reader should pay attention to the type and cotype of the constructed spaces in both articles.


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.