Suppose $A,B$ are $C^*$ algebras, $\alpha$ is some $C^*$-norm on the algebraic tensor product of $A$ and $B$. If $A\otimes_{\alpha}B$ is nuclear, can we conclude that $A$ and $B$ are nuclear? What about the converse?



Let us assume that $A$ and $B$ are unital.

Fix a $C^*$-norm $\alpha$ on $A \odot B$, by which I denote the algebraic tensor product. Next, we note that for any state $\varphi \in S(A)$ and $\psi \in S(B)$ there exists, by [Prop. 3.4.7., $C^*$-algebras and Finite-dimensional Approximation; Brown and Ozawa], a state $$ \varphi \otimes \psi \colon A \otimes_\alpha B \to \mathbb C: a \otimes b \mapsto \varphi(a)\psi(b). $$

In particular, the slice map $$ \eta_\psi \colon A \odot B \to A : a \otimes b \mapsto a\psi(b) $$ is $\alpha$-continuous. To see this, let $x \in A \odot B$ and $\varphi \in S(A) $ such that $$ \varphi(\eta_\psi(x)) = \lVert \eta_\psi(x) \rVert_A. $$ Now, by the previous discussion, we get $$ \lVert \eta_\psi(x) \rVert_A = \varphi(\eta_\psi(x)) = \lvert (\varphi \otimes \psi)(x) \rvert \leq \lVert x \rVert_{\alpha}. $$

In particular, $\lVert \eta_\psi \rVert \leq 1$, whenever $\psi \in S(B)$.

Now, assume that $A \otimes_{\alpha} B$ is nuclear. We show that $A$ is nuclear (and similarly you can show that $B$ is nuclear).

Let $\mathscr F \subset A$ be finite and $\varepsilon > 0$. Denote by $\iota_A \colon A \to A \otimes_{\alpha} B$ the obvious inclusion. Since $A \otimes_{\alpha} B$ is nuclear, there are a finite dimensional $C^*$-algebra $F$ and c.p.c. maps $\alpha \colon A \otimes_{\alpha} B \to F$ and $\beta \colon F \to A \otimes_\alpha B$ such that for all $a \in \mathscr F$, $$ \lVert (\beta \circ \alpha)(\iota_A(a)) - \iota_A(a) \rVert < \varepsilon . $$ Now, let $\psi$ be any state on $B$. Then

\begin{align*} \lVert \eta_\psi((\beta \circ \alpha)(\iota_A(a))) - a \rVert \\ = \lVert \eta_\psi((\beta \circ \alpha)(\iota_A(a)) - \iota_A(a)) \rVert \\ \leq \lVert (\beta \circ \alpha)(\iota_A(a)) - \iota_A(a) \rVert \\ < \varepsilon \end{align*}

It follows that $A$ is nuclear.

If $A$ or $B$ are not unital, you have to look at $\iota_A \colon A \to A \otimes_\alpha B$ given by $\iota_A(a) = A \otimes h$, where $h$ is a positive element of norm one.

The converse direction can be easily proved for the minimal and maximal tensor prodcut. However, if $\alpha$ is just some $C^*$-norm I am not sure if the tensor product of c.p.c. maps with finite dimensional domain or codomain might be $\alpha$-continuous.


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.