Let $H$ be a Hilbert space. Any unitary operator $u\in B(H)$ induces an spatial isomorphism $\phi_u(x)=uxu^*$ on $B(H)$.

Question: Let $\phi:B(H)\to B(H)$ be a surjective *-ismorphism. Is $\phi$ implemented by a unitary operator $u\in B(H)$ i,e. $\phi=\phi_u$?

Remark. As for this problem, let us review two facts:

i) It is well-known that any surjective *-isomorphism $\phi$ on $B(H)$ (and in general on any von Neumann algebra) is $w^*$-continuous.

ii) Let $\psi$ be the restriction of $\phi$ on compact operators $K(H)$. Assume $\psi$ forms an isomorphism from $K(H)$ onto $K(H)$ then $\psi$ is implemented by a unitary operator and then i) implies that $\phi$ is also implemented by a unitary operator.

This remark says that, my problem may be re-written in this form too: Does there exist any surjective $*$-isomorphism $\phi$ on $B(H)$ such that $K(H)\neq\phi(K(H))$.


Fix an orthonormal basis $\xi_j$, and consider the associated matrix units $\{e_{kj}\}$, i.e. $e_{kj}\xi=\langle \xi,\xi_j\rangle\,\xi_k$. The matrix units are characterized by $$ e_{kj}e_{st}=\delta_{js}\,e_{kt}.$$ It is easy to check that $\{\phi(e_{kj})\}$ is also a system of matrix units. If we let $\eta_j$ denote a unit vector in the range of the projector $\phi(e_{jj})$, then $\{\eta_j\}$ is an orthonormal basis (because $\sum\phi(e_{jj})=1$).

Let $u\in B(H)$ be the unitary that sends $\xi_j\longmapsto \eta_j$. Then, for any $\xi\in H$, $$ \phi(e_{kj})\xi=\langle\xi,\eta_j\rangle\,\eta_k =\langle\xi,u\xi_j\rangle\,u\eta_k=u\,\langle u^*\xi,\xi_j\rangle\,\xi_k =ue_{kj}u^*\xi. $$ So $\phi(e_{kj})=ue_{kj}u^*$. Now one can extend by linearity and continuity to all of $K(H)$ (and all of $B(H)$ by using wot-continuity.

  • $\begingroup$ Dear Martin, if $\phi(e_{jj})$ was not a rank one projection(or even be infinite rank), then why does the sequence $\{\eta_j\}$ form a basis? $\endgroup$ – Ali Bagheri Jan 29 '16 at 10:15
  • $\begingroup$ That's not possible. If $q\leq\phi(e_{jj})$, then $\phi^{-1}(q)\leq e_{jj}$. So $\phi(e_{jj})$ is a minimal projection. $\endgroup$ – Martin Argerami Jan 29 '16 at 13:34

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.