I'm working through Li's and Barlak's Cartan Subalgebras and the UCT Problem but I'm stuck at one of the simpler proofs of the paper. On page 9 they deal with masas (maximal abelian subalgebras) of a C*-algebra $A$ and look at the crossed products $A \rtimes_\alpha \mathbb{Z}_n$ where $\alpha\in Aut(A)$ with $\alpha^n=id$. I now have two questions:

1.) Before prop. 4.1 the paper says "let $t$ be the unitary in (the multiplier algebra of) $A\rtimes_\alpha \mathbb{Z}_n$ implementing the $\mathbb{Z}_n$-action." I'm not quite familiar with theory of crossed products. What does this mean and why does such a unitary exist? (Do you have literature recommendations?)

2.) In the proof they compute $[t^k x_k, b]$ and the result seems to be $t^k(x_kb-\alpha^{-k}(b)x_k)$. By construction of crossed products I would've guessed $[t^k x_k, b]=t^k\left(x_{k}\alpha^{k}\left(b\right)-bx_{k}\right)$. Where's my mistake?


The unitary exists by the definition of the crossed product. In the case of a $\mathbb{Z}/n\mathbb{Z}$-action you can think of an element of the crossed product as a linear combination $$ \sum_{n = 0}^{n-1} a_n t^n $$ with $a_n \in A$ and where $t$ is a unitary (i.e. $t^* = t^{-1}$), such that the following relations hold $$ tat^* = tat^{-1} = \alpha(a) $$ In particular $t^kat^{-k} = \alpha^k(a)$ and $t^{-k}at^{k} = \alpha^{-k}(a)$. In the case $A = \mathbb{C}$ and $\alpha$ trivial, you obtain the group $C^*$-algebra of $\mathbb{Z}/n\mathbb{Z}$ this way, which you might be more familiar with. Concerning the second question we have $$ [t^kx_k, b] = t^kx_kb - bt^kx_k = t^kx_kb - t^k(t^{-k}bt^k)x_k = t^k(x_kb - \alpha^{-k}(b)x_k)\ . $$ Therefore their calculation seems to be correct. A good source of information about crossed products is this book by Dana P. Williams.

Here are some more details about crossed products. Formally the crossed product is defined as follows: Let $G = \mathbb{Z}/n\mathbb{Z}$. Consider the algebra $C(G,A)$ of functions on $G$ with values in $A$ equipped with the following twisted multiplication $$ (f \cdot h)(x) = \sum_{g \in G} f(g)\alpha_g(h(g^{-1}x)) $$ and the $*$-operation $$ f^*(x) = \alpha_x(f(x^{-1})^*). $$ This $*$-algebra has several interesting faithful representations, for example a regular and a maximal one. These correspond to (possibly different) completions to $C^*$-algebras, e.g. the reduced or maximal crossed product.

To obtain the picture described above, we can then identify the unitary $t$ with the function $$ t(x) = \begin{cases} 1_A & if x = 1 \\ 0 & else \end{cases} $$ where $1 \in \mathbb{Z}/n\mathbb{Z}$ refers to the generator on the right hand side and $1_A$ is the unit of $A$. Strictly speaking this only works if $A$ is unital. Otherwise $t$ will not be an element in the crossed product, but in its multiplier algebra. Then we have $$ t\cdot a \cdot t^*(x) = \sum_{g,h} t(g)\alpha_g(a(h)\alpha_h(t^{-1}((gh)^{-1}x))) = \alpha(a(x)) $$ which corresponds to the relation mentioned above.

The crossed product therefore is a $C^*$-algebra that contains the original algebra $A$ (in the unital case) and in which the action becomes inner.

  • $\begingroup$ Thanks for your answer! I see that if we have a $t$ such that $tat^*=\alpha(a)$ elements of the above polynomial form are exactly the elements of the crossed product. But why is every automorphism of this form? Is this a general property of the multiplier algebra? $\endgroup$ Aug 20 '17 at 15:54
  • $\begingroup$ Note that $t$ is not an element of $A$, i.e. the automorphism does not have to be inner from the start. However, the crossed product is defined in such a way that the automorphism becomes inner by adding another element $t$ that satisfies $tat^* = \alpha(a)$. $\endgroup$ Aug 20 '17 at 15:58
  • $\begingroup$ All right! I think I get it now. Tanks a lot! $\endgroup$ Aug 20 '17 at 16:19

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.