Let $A$ and $B$ be two finite-dimensional C*-algebras. Let $\gamma$ denote the projective Banach space tensor product norm on the algebraic tensor product $A\odot B$, so $\gamma(t)=\inf\{\sum_{i}\|a_i\|\|b_i\|:t=\sum_i a_i\otimes b_i\}$. Then $\gamma(ts)\leq\gamma(t)\gamma(s)$ and $\gamma(t^*)=\gamma(t)$, but does $\gamma$ also satisfy $\gamma(t^*t)=\gamma(t)^2$?

I know that $\gamma$ should be equivalent to the C*-algebra norm on $A\odot B$, since all norms on finite-dimensional spaces are equivalent, but I would like to have equality, since this would imply that the embedding of finite-dimensional C*-algebras with subunital completely positive maps in the category of Banach spaces with contractions and with the projective tensor product as a monoidal product is a strong monoidal functor.

  • 3
    $\begingroup$ I think the answer is no unless either $A$ or $B$ is one-dimensional, by looking at the diagonal copies of $D=\ell_\infty^2$ inside $A$ and $B$. The point is that the natural map $D\hat\otimes_\gamma D \to D \otimes_{\rm min} D$ is not isometric. If I have time later I'll try to check this and flesh out the details $\endgroup$ – Yemon Choi Oct 8 '18 at 18:59

I think there are various reasons from abstract tensor norm theory why this is impossible, but I'll try to give a concrete example. First, a $C^*$-norm on an algebra is uniquely determined; therefore, if $A=\ell_\infty(2)=\mathbf C^2$ with the max-norm, then the tensor norm on $A\otimes A$ is the injective tensor norm $\otimes_\varepsilon$, identifying $A\otimes A$ isometrically with $\mathbf{C}^4$ in the max-norm. (If $A=C(K)$, then $A\otimes_\varepsilon A = C(K\times K)$.) So we have to argue that $\otimes_\gamma$ is not $\otimes _\varepsilon$ in this example.

For this let us compare the duals; $(\ell_\infty(2)\otimes_\gamma \ell_\infty(2))^* = L(\ell_\infty(2), \ell_1(2)) =: X$ and $(\ell_\infty(2)\otimes_\varepsilon \ell_\infty(2))^* = (\ell_\infty(2; \ell_\infty(2))^* = \ell_1(2; \ell_1(2)) := Y$. Specifically, consider the element of the dual represented by $(\alpha,\beta)\mapsto \alpha (1 , 1) + \beta (1,-1)$. The norm of this element in $Y$ is $\|(1,1)\|_1 + \|(1,-1)\|_1 = 4$, but the norm in $X$ is $\max_{|\alpha|,\beta|\le 1} \|\alpha(1,1)+\beta(1,-1)\|_1= \max_{\alpha,\beta} (|\alpha+\beta|+|\alpha-\beta|)<4$.

So the norms are isometrically different.

PS: I read Yemon's comment, which practically says the same as mine, only after having typed this$\dots$.

| cite | improve this answer | |
  • 1
    $\begingroup$ I saw your answer while typing essentially the same thing -- the norm of the $2 \times 2$ Hadamard matrix is 2 as a bilinear form on $\ell^\infty(2) \times \ell^\infty(2)$ (and hence in the dual space of $\ell^\infty(2) \hat{\otimes} \ell^\infty(2)$) but 4 as an element of $\ell^\infty(2 \times 2)^*$. $\endgroup$ – Robert Furber Oct 8 '18 at 20:25

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.