Let $H$ be a Hilbert space, and
let $A_1,A_2,A_3\subset B(H)$ be three commuting von Neumann algebras.

We write $\odot$ for the algebraic tensor product, and $\bar\otimes$ for the spatial tensor product of von Neumann algebras.

Suppose that for every $i,j\in\{1,2,3\}$, the map $A_i\odot A_j\to B(H)$ extends to a map $A_i\,\bar\otimes\, A_j\to B(H)$.

Does it follow that the map $A_1\odot A_2\odot A_3\to B(H)$ extends to a map $A_1\,\bar\otimes\, A_2\,\bar\otimes\, A_3\to B(H)$?

  • $\begingroup$ Can you show that this implies that $(A_1\vee A_2) \odot A_3$ extends to a map $(A_1\vee A_2) \mathbin{\bar\otimes} A_3$? $\endgroup$ Jul 4, 2016 at 7:06
  • $\begingroup$ If I could show that, then I could solve my problem. Indeed, $A_1\vee A_2=A_1\,\bar\otimes\, A_2$ by assumption, and $(A_1\,\bar\otimes\, A_2)\,\bar\otimes\, A_3=A_1\,\bar\otimes\, A_2\,\bar\otimes\, A_3$ by definition. So, to answer your question: unfortunately no, I don't know. $\endgroup$ Jul 4, 2016 at 9:27
  • $\begingroup$ Exactly. Why is naively intersecting the type I factors not working? $\endgroup$ Jul 4, 2016 at 16:23
  • $\begingroup$ Because I don't know that the intersection still type I. $\endgroup$ Jul 4, 2016 at 20:26

1 Answer 1


No. Consider the case where $G$ is an ICC group, $H=\ell_2(G\times G)$, $A_1=(\lambda\otimes\lambda)(G)''$, $A_2=(1\otimes\rho)(G)''$, and $A_3=(\rho\otimes1)(G)''$. By the Fell absorption principle $(g,h)\leftrightarrow (g,gh)$, one has $A_1\vee A_2 \cong (\lambda\otimes1)(G)''\vee(1\otimes\rho)(G)''\cong A_1\mathbin{\bar{\otimes}}A_2$, and similarly for other $A_i \vee A_j$. But since the diagonal $(\lambda\otimes\lambda)(g)\cdot(1\otimes\rho)(g)\cdot(\rho\otimes1)(g)$ fixes $\delta_e\otimes\delta_e$, one has $A_1\vee A_2\vee A_3 \not\cong A_1\mathbin{\bar{\otimes}}A_2\mathbin{\bar{\otimes}}A_3$.

  • $\begingroup$ Beautiful example! $\endgroup$ Jun 22, 2017 at 10:04
  • 1
    $\begingroup$ The construction can even be generalised to yield $n$ commuting actions such that any $n-1$ of them extend to the spatial tensor product, but this does not happen for all actions at once. Take $H=\ell_2(G^{n-1})$ and the $n$ actions to be given by $\lambda^{\otimes n-1}$ and $1\otimes...\otimes \rho \otimes...\otimes 1$. $\endgroup$ Jun 22, 2017 at 10:11

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.