James' space $J$ is subprojective; i.e., every infinite dimensional (closed) subspace of $J$ contains an infinite dimensional subspace which is complemented in $J$. This fact can be found in Corollary 11 of [P.G. Casazza et al. Proc. Amer. Math. Soc. 26 (1977), 294-305].

I would like to know if the dual space $J^*$ is subprojective. It is known that a non-reflexive subspace of $J^*$ contains an infinite dimensional subspace complemented in $J^*$ (Theorem 2 in [A. Andrew. Israel J. Math. 38 (1981), 276-282]), but I do not know what happens for infinite dimensional reflexive subspaces.


Yes, $J^*$ is subprojective.

Suppose that $M$ is a closed subspace of $J^*$. By Mazur's theorem, without loss of generality we additionally suppose that $M$ has a normalised basis $(f_k)_{k=1}^\infty$. Let $\iota\colon M\to J^*$ be the inclusion map and let $Q\colon J^{**}\to M^*$ be its (surjective) adjoint. Denote by $(f^*_k)_{k=1}^\infty$ the coordinate functionals of $(f_k)_{k=1}^\infty$.

Choose a bounded sequence $(z_n)_{n=1}^\infty$ in $J^{**}$ such that $Qz_n = f_n^*$ ($n\in \mathbb{N}$). As $J$ is quasi-reflexive, by Rosenthal's $\ell_1$-theorem, $(z_n)_{n=1}^\infty$ contains a weakly Cauchy subsequence, call it $(z^\prime_n)_{n=1}^\infty$. As $(z^\prime_n)_{n=1}^\infty$ fails to have a convergent subsequence, we may suppose that the sequence $(z^\prime_{2n+1}-z^\prime_{2n})_{n=1}^\infty$ is semi-normalised. However, by a result of Andrew [1, Theorem 2.1], every semi-normalised weakly null sequence in $J$ (so also in $J^{**}$ as $J^{**}\cong J$) contains a sequence spanning a complemented copy of $\ell_2$, so $(z^\prime_{2n+1}-z^\prime_{2n})_{n=1}^\infty$ contains such subsequence, say $(w_n)_{n=1}^\infty$. By reflexivity of $\ell_2$, $(w_n)_{n=1}^\infty$ is shrinking so the adjoint of a projection from $J^{**}$ onto $[w_n]_{n=1}^\infty$ is a projection onto a subspace of $M$. As $M\subset J^* \subset J^{***}$, the conclusion follows.


  1. A. Andrew, Spreading basic sequences and subspaces of James' quasi-reflexive space, Math. Scand. 48 (1981), 276–282.
| cite | improve this answer | |

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.