MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Back here I was asking for a coalgebraic characterisation of the hyperfinite $II_1$ factor. Recall the latter's construction by forming the inductive limit of a chain of matrix algebras $R \to M_2(R) \to M_{2^2}(R) \to ...$, where a matrix is sent to two copies of itself placed in the diagonal blocks, zero elsewhere. Then completion in the weak topology gives the hyperfinite factor.

The trace is halved each step along the chain. Traces for projections in the inductive limit are dyadic rationals in [0, 1], while in the hyperfinite factor they are the whole real interval [0, 1]. Now, the dyadic rationals are the initial algebra for the endofunctor on Bipointed Set, $X \mapsto X \vee X$, identifying the second point of the first copy with the first point of the second copy. The [0, 1] interval is the terminal coalgebra for the same functor and the Dedekind and Cauchy completion of the initial algebra. Perhaps results such as Adamek's Final Coalgebras are Ideal Completions of Initial Algebras may be extended here.

This put me on the quest of characterising the hyperfinite factor coalgebraically. So now the question: what relevant facts are known about the endofunctor on algebras over the reals: $X \to M_2(X)$? I believe the hyperfinite factor is a fixed point. Presumably there is a need to be clear over isomorphism versus Morita equivalence. Might it be that the hyperfinite factor is the greatest fixed point up to isomorphism? Maybe I should be looking in the category of a certain kind of algebra.

share|cite|improve this question
I suppose one useful check would be to make sure that of all those automorphisms of the hyperfinite factor only the identity function is compatible with a chosen isomorphism with 2x2 matrices over it. – David Corfield Nov 12 '09 at 14:46
I noticed that periods are allowed in tag names. As an experiment, I retagged with the arXiv code followed by a period followed by the category name. I think it looks good! My suggestion is to eventually tag most of mathoverflow this way. – Greg Kuperberg Nov 12 '09 at 20:09
@Greg: I'd really like you to join the tagging discussion at meta.MO ( – Anton Geraschenko Nov 12 '09 at 20:30
up vote 6 down vote accepted

This is an interesting question, but the motivation is a bit misaligned. $C^*$ algebras are a non-commutative or quantum generalization of compact Hausdorff spaces and von Neumann algebras are a non-commutative or quantum generalization of (not too unreasonable) measurable spaces. However, both of these generalizations are contravariant. Your motivation is a covariant comparison between von Neumann algebras and topological spaces, which is problematic.

The von Neumann algebra or $C^*$-algebra $M_2(\mathbb{C})$ is now famously known as a "qubit"; it is a great non-commutative analogue of $\mathbb{C} \oplus \mathbb{C}$, which is of course the complex functional algebra of a classical bit. The endofunctor that you ask about is a geometric product of $X$ and a qubit, and the morphism in your question is a geometric projection back to $X$ with a qubit fiber. So the completion that you ask about is thus a geometric product with a quantum Cantor set. I forget what the fiber is called in the $C^*$-algebra setting, but I remember that, unlike a classical Cantor set, its isomorphism type depends on the sizes of the matrices. In the von Neumann case, this quantum Cantor set is interpreted as a measurable space, and then it is always the hyperfinite $II_1$ factor and does not depend on the matrix sizes.

I think that the hyperfinite factor is not the only fixed point of tensoring with a qubit. Let $S$ be any set, let $F$ be the set of functions from $S$ to a bit (or any finite set), and then let $M$ be the von Neumann closure of the local operators on $\ell^2(F)$. Here a local operator is one that affects only finitely many values of $f \in F$. If $S$ is an infinite set of any cardinality, then $M$ goes to itself when you tensor it with a qubit.

share|cite|improve this answer
Thanks! John Baez noted to me that Jean-Yves Girard is working on using the type II_1 hyperfinite factor in a souped-up version of linear logic, and he called it a quantum version of the Cantor set ( – David Corfield Nov 13 '09 at 10:40
Von Neumann's initial motivation to study $II_1$-factors was to obtain a souped-up version of linear logic. There is a lot of work of von Neumann himself clarifying this viewpoint. He stopped his studies when he understood that it cannot be a sufficient framework for basic quantum mechanics. – Andreas Thom Apr 12 '12 at 19:14

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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