We may consider each $C^*$-algebra as a Banach space (by forgetting the multiplication and adjoint). I wonder how drastic this step is, i.e., which properties of the $C^*$-algebra are reflected by its Banach space structure.

Clearly, finite-dimensionality and other cardinality properties can be detected. Thus, to avoid these cases, let us restrict to separable, infinite-dimensional $C^*$-algebras.

For commutative $C^*$-algebras the problem has been solved by Milutin. In particular, $C(X)$ and $C(Y)$ are isomorphic as Banach spaces for any uncountable, compact, metric spaces $X$ and $Y$.

In the paper
Hamana. On linear topological properties of some $C^*$-algebras, Tohoku Math. J., II. Ser. 29, 157-163 (1977).
it is shown that the Banach space structure also reflects if all irreducible representations of the $C^*$-algebra are finite-dimensional.

Let us ask what happens at the 'opposite' end of the scale:

Question: Are all simple, separable, infinite-dimensional $C^*$-algebras isomorphic as Banach spaces?

  • 4
    $\begingroup$ Is an isomorphism of Banach spaces a bounded linear map with bounded linear inverse or an isometric isomorphism? $\endgroup$ – Qiaochu Yuan Feb 3 '14 at 23:45
  • 4
    $\begingroup$ I assume "isomorphic as Banach spaces" means linear homeomorphism. It's well known that there are non-isomorphic C*-algberas that are isometric as Banach spaces. (Google "C*-algebra not isomorphic to its opposite algebra.) The correct statement is: A and B are isomorphic as C*-algebras if and only if they are completely isometric as operator spaces. $\endgroup$ – Nik Weaver Feb 4 '14 at 3:14
  • 4
    $\begingroup$ Hi Hannes, Kirchberg showed in his subalgebras of CAR algebra paper that all separable nuclear, non Type I C*-algebras are isomorphic as operator spaces (much stronger than just isomorphic as Banach spaces). I think Christensen and others also did some things in this direction for von Neumann algebras: those results are in Pisier's operator space book, but my copy isn't with me. $\endgroup$ – Caleb Eckhardt Feb 4 '14 at 3:52
  • 7
    $\begingroup$ A complement to Caleb's remark: Being type I is invariant under a Banach space isomorphism (at least in the separable case). This follows from Haagerup--Rosenthal--Sukochev's classification of noncommutative $L^1$-spaces up to Banach isomorphism. $\endgroup$ – Narutaka OZAWA Feb 4 '14 at 8:49
  • 10
    $\begingroup$ By Szankowski (Acta Math 1981), there is a separable (simple, unital, etc.) $C^*$-algebra which does not have the approximation property. Such a $C^*$-algebra cannot be Banach isomorphic to a nuclear one. $\endgroup$ – Narutaka OZAWA Feb 4 '14 at 8:57

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.