While studying the $C^*$-algebraic formulation of the recently solved Kadison-Singer problem, I was wondering about maximal abelian subalgebras:
Let $\mathcal{A}$ be a unital C* algebra. There seems to be a theorem (see e.g. Kadison/Singer "Extension of Pure States", however, I couldn't find a version not behind the paywall) that there are only three types of maximal abelian subalgebras (+ direct sums)
- the continuous maximal abelian subalgebra $\mathcal{A}_c$, which I believe is isomorphic to $L^{\infty}([0,1])$ acting on the Hilbert space $L^2([0,1])$.
- the discrete maximal abelian subalgebra $\mathcal{A}_d$, which I believe is isomorphic to $\mathcal{D}(\ell_2)$, the diagonal matrices of square integrable functions.
- finite dimensional versions of the latter algebra $\mathcal{A}_n$ for all $n\in\mathbb{N}$ (basically, diagonal matrices).
This theorem seems to be general knowledge, however I could not find a reference to a proof. Could someone either provide one or point me to the main ideas and/or intuitions to prove this? I can only see how this should be the case for finite dimensional systems (where only the third case survives).
Furthermore, this may be a long shot, but such a decomposition reminds me of spectral decompositions into continuous and pure point spectrum, so I would like to find out, whether there is maybe a relationship?
It is not hard to see that it is as you say when $\mathcal A$ is a von Neumann algebra.
But there are C$^*$-algebras with few to no projections, and so none of those masas can live there. Consider for instance $C^*_r(\mathbb F_2)$, the reduced C$^*$-algebra of the free group on two generators $\mathbb F=\langle a,b\rangle$. This algebra is known to be projectionless. But you can consider $C^*(a)\subset C^*_r(\mathbb F_2)$. I think this is a masa; but even if it isn't, it is contained in some masa by an easy application of Zorn's Lemma, and this masa is a proper subalgebra as the full algebra is non-abelian. So we have a masa in a projectionless C$^*$-algebra, which cannot be any kind of direct sums like the examples in the question, as all these have many projections.