Let $G$ be a finite abelian group (cosidered as a discrete topological group), $A$ a unital separable $C^*$-algebra. Let $T\colon G\to \operatorname{Aut}(A)$, $T_g(a)=a$ for all $g\in G$ the trivial action.

Let $Lt\colon G\to \operatorname{Aut}(C(G))$ the left translation, i.e. $Lt$ is given by $Lt_g(f)(h)=f(g^{-1}h)$ for a continuous function $f\in C(G)$ and $h\in G$.

My question: Is $T\otimes Lt\colon G\to \operatorname{Aut}(A\otimes C(G))$ the dual action of $T$, or is it possible to identify the dual action of $T$ with $T\otimes Lt$? ($\otimes$ here means the minimal tensor product of $C^*$-algebras, but note that $C(G)$ is nuclear)

First of all, the dual action is an action $\hat{T}\colon \hat{G}\to \operatorname{Aut}(A\rtimes_T G)$, defined by $$\hat{T}_\tau(\sum_{s\in G}a_s \delta_s)=\sum_{s\in G}a_s\tau(s) \delta_s$$ for $\tau\in\hat{G}$ and $\sum_sa_s \delta_s\in C_C(G,A)$. $\hat{G}$ denotes the dual group (the Pontryagin dual) of $G$. Since $T$ is the trivial action, there is an isomorphism $$A\rtimes_T G\to A\otimes_{max}C^*(G),$$ induced by the canonical map $\sum_sa_s\cdot \delta_s\mapsto \sum_sa_s\otimes \delta_s$ on the $*$-algebraic tensor products ($C^*(G)$ means the full group $C^*$-algebra of $G$). Since $C^*(G)$ is nuclear, $A\otimes_{max}C^*(G)\cong A\otimes C^*(G)$. Forthermore, there is an isomorphism $$ F\colon C^*(G)\to C(\hat{G}),$$ given by the Fourier transform $F(f)(\chi)=\sum\limits_{g\in G}f(g)\chi(g)$, where $f\in C_C(G)$ and $\chi\in \hat{G}$.

All in all, after composing these isomorphisms, we obtain an action $\hat{T}\colon \hat{G}\to \operatorname{Aut}(A\otimes C(\hat{G}))$ which does not look like $T\otimes Lt\colon \hat{G}\to \operatorname{Aut}(A\otimes C(\hat{G}))$ to me, or does it?

The last step is to identify $\hat{G}\cong G$ (but this is not canonically isomorphic). I appreciate your help.


Using the isomorphisms you wrote down, the dual action is the left translation action, composed with the automorphism $g\mapsto g^{-1}$.

To see this, first notice that the $C^*$-algebra $A$ plays no role in any of the definitions, so it can safely be disregarded in the computation. We're thus considering the action of $\widehat{G}$ on the group algebra $C^*(G)=\mathbb{C}[G]$ given by $\tau\cdot \delta_s = \tau(s)\delta_s$ for $\tau\in \widehat{G}$ and $s\in G$. The Fourier/Gelfand transform $F:\mathbb{C}[G]\to C(\widehat{G})$ sends $\delta_s$ to the function $F(\delta_s):\chi\mapsto \chi(s)$. We have $$ F(\tau\cdot \delta_s)(\chi) = \tau(s)\chi(s) = F(\delta_s)(\tau\chi), $$ showing that the action $\widehat{G}\rightarrow\operatorname{Aut}(C(\widehat{G}))$ is given by $(\tau\cdot f)(\chi) = f(\tau\chi)$. Of course, the same formula continues to hold if you choose an isomorphism $\widehat{G}\cong G$ and use that same isomorphism to identify $C(\widehat{G})$ with $C(G)$.

| cite | improve this answer | |

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.