For a Banach space $X$, let $A \subseteq \mathcal{B}(X^*)$ denote the closed subspace of those operators $T : X^* \to X^*$ such that $T = S^*$ for some $S \in \mathcal{B}(X)$, that is, $T$ is the adjoint of some bounded linear operator on $X$. It can be shown that this condition is equivalent to $T$ being weak*-continuous. Certainly $A=\mathcal{B}(X^*)$ for a Hilbert space. Do there exist Banach spaces $X$ for which $A$ is of finite positive codimension in $\mathcal{B}(X^*)$, i.e. the quotient space $\mathcal{B}(X^*)/A$ is finite-dimensional (but nontrivial)?


If $X$ is reflexive, then $A= B(X^*)$.

If $X^{**}/X$ is infinite dimensional, let $(x_n^{**})$ be a sequence in $X^{**}$ linearly independent over $X$. Fix a nonzero $x^*\in X^*$. If a finite linear combination $\sum a_nx^{**}_n\otimes x^*\in A$, then $\sum a_nx^{**}_n\in X$ and hence $a_n =0$ for all $n$. Thus $(x^{**}_n\otimes x^*)$ is linearly independent over $A$. This shows that $B(X^*)/A$ is infinite dimensional.

This leaves the quasi-reflexive spaces ($X^{**}/X$ finite dimensional). I'm pretty sure that $B(X^*)/A$ is finite dimensional and nontrivial for $X=$ James Space.


The answer is no.

As noted in Denny Leung's answer, if $X$ is reflexive then $A=B(X^\ast)$.

A dual version of Denny Leung's subsequent argument for a certain non-reflexive case actually gives the complete answer in the nonreflexive case. Indeed, suppose $X$ is nonreflexive, fix $x^{\ast\ast}\in X^{\ast\ast}\setminus X$ and let $(x_n^\ast)$ be a linearly independent sequence in $X^\ast$. Since $x^{\ast\ast}$ is not weak$^\ast$-continuous, the sequence $(x^{\ast\ast}\otimes x_n^\ast)$ is linearly independent over $A$, hence $B(X^\ast)/A$ is infinite dimensional.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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