Let $N$ be a type $II_{1}$-factor with trace $\tau$, and $B$ a von Neumann subalgebra. The existence of the semifinite trace on the Jones basic construction $\langle N, e_{B} \rangle$ is reasonably easy to establish if $B$ is a subfactor of $N$, but appears not to be so easy in general.

Question: What is the shortest known proof of the existence of the trace on the basic construction for a von Neumann subalgebra inclusion?

Two complete proofs of this appear in the excellent book Finite von Neumann algebras and Masas by Sinclair and Smith. I (and others far more adept than I) am curious if any other proofs of this exist in the literature.

EDIT: The previous question, as asked, is embarrassingly silly. What seems less than obvious is how to construct the trace and verify that what you have written down is actually a faithful, NORMAL, semifinite trace. If someone could indicate how to do this, I'd very much appreciate it.


1 Answer 1


Perhaps I am missing some hypothesis, but I think the proof is just about the same whether or not $B$ is a factor. Here is the proof from Jones' original paper, and I believe it does not use factoriality of B (and not even facoriality of $N$?)

Lemma. Let $J:L^2(N)\to L^2(N)$ be the modular conjugation. Then $\langle N, e_B\rangle$ is $J B' J$.

Indeed, clearly $N=JN'J \subset JB'J$ and $e_B=Je_BJ\subset JB'J$ since $L^2(B)$ is $B$-invariant; thus $\langle N,e_B\rangle \subset JN'J$. On the other hand, $J\langle N, e_B\rangle J ' \subset JNJ ' \cap \{e_B\}' =\{e_B\}'\cap N$ since $JNJ'=N$ and $Je_BJ=e_B$. If $x\in N$ commutes with $e_B$ then, denoting by $1\in L^2(N)$ the trace vector, $x e_B 1 = e_B x 1 \in L^2(B)$ so that $x\in L^2(B)$. Thus $x=E_B(x)$ (where $E_B$ is the trace-preserving conditional expectation onto $B$) and so $x\in B$. Thus $\{e_B'\}\cap N = B$ and as a result $J\langle N, e_B\rangle J' \subset B$. By the bicommutant theorem you then get that $\langle N, e_B\rangle \subset JB'J$.

Now, given the Lemma, it is clear that $JB'J$ has a semi-finite trace, since $B'$ has a semi-finite trace (since $B$ has a finite trace).

  • $\begingroup$ You didn't miss anything, and have answered the question I (am embarrassed to) have asked. What I meant to ask is if there is a short and easy explicit construction of the trace, and verification that it is, in fact, a normal, semifinite trace. I would still very much like to see this, if you should know how to do it. $\endgroup$
    – Jon Bannon
    Apr 2, 2011 at 11:09
  • 2
    $\begingroup$ I think your question is: how to construct a semifinite trace on $B'$ if $B\subset B(H)$ is a subalgebra for which some vector $\xi$ gives a tracial state $x\mapsto \langle x\xi,\xi\rangle$ (think for example of $B=\mathbb{C}\subset B(\mathbb{C}^N$ with trace vector $e_1=(1,0,...)$). You need to find isometries $v_i\in B'$ with $\sum v_iv_i^*=1$ and $v_i^* v_i\leq P$, where $P$ is the projection onto the closure of $B\xi\subset H$. Then $Tr(x)=\sum \langle v_i^* x v_i \xi ,\xi \rangle$ is the semifinite trace you want. This is a classical construction going back to Murray-von Neumann. $\endgroup$ Apr 2, 2011 at 16:39
  • $\begingroup$ Thanks, Dima. I'll give it another try! (Trying to "learn and relearn" the subject.) $\endgroup$
    – Jon Bannon
    Apr 2, 2011 at 19:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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