5
$\begingroup$

Let $G$ be a finite group, and consider the composition $X * Y$ on $\mathbb{C}G$ defined by $$(\sum_g \alpha_g u_g) * (\sum_g \beta_g u_g) = \sum_g \alpha_g \beta_g u_g.$$ This composition can be reformulated as a convolution using the Fourier transform.

Question: Let $X,Y \in \mathbb{C}G$ be positive elements. Is $X * Y$ also positive?

Remark: It is an application of Theorem 4.1 in http://dx.doi.org/10.1090/tran/6582 to finite groups theory, but we are here interested in a purely group-theoretic reference or proof.

It is like to Schur product Theorem because $X * Y$ is an entrywise product like the Schur product.

Bonus question: Can we extend to any locally compact group?

$\endgroup$
2
  • 3
    $\begingroup$ What means positive here? $\endgroup$ Commented Aug 6, 2017 at 13:22
  • $\begingroup$ @AlexanderChervov: Positive means of the form $a^{\star}a$, with $(\sum_g \alpha_g u_g)^{\star} = \sum_g \overline{\alpha}_g u_{g^{-1}}$. $\endgroup$ Commented Aug 6, 2017 at 13:26

1 Answer 1

4
$\begingroup$

The answer is positive. For ease of notation, let me write $a=\sum_g \alpha_g u_g$ and $b = \sum_g \beta_g u_g$.

Let's equip $\mathbb{C}G$ with the positive inner product induced by the standard basis. In terms of the usual normalized trace on a group algebra, this inner product is given by $(x,y)\mapsto\mathrm{tr}(x^*y)$. So when you decompose $\mathbb{C}G$ into a direct sum of matrix algebras, this is just the usual Hilbert-Schmidt inner product. In particular, the cone of positive elements is nicely self-dual: we have $y\geq 0$ if and only if $\mathrm{tr}(xy)\geq 0$ for all $x\geq 0$. Hence it is enough to show that $\mathrm{tr}\left(x(a \ast b)\right)\geq 0$ for all $x\geq 0$.

Now your composition operation $*:\mathbb{C}G\otimes\mathbb{C}G\to\mathbb{C}G$ is the adjoint of the usual comultiplication $\Delta:\mathbb{C}G\to\mathbb{C}G\otimes\mathbb{C}G$, meaning that $$ \mathrm{tr}\left(x(a \ast b)\right) = \mathrm{tr}(\Delta(x)(a\otimes b)). $$ Finally, we have $\Delta(x)\geq 0$ since $\Delta$ is a $*$-homomorphism, and $a\otimes b\geq 0$ since $a\geq 0$ and $b\geq 0$ by assumption.

I don't know about the generalization to the locally compact case.

$\endgroup$
4
  • $\begingroup$ Your nice proof extends immediately to any finite dimensional Kac algebra. $\endgroup$ Commented Aug 6, 2017 at 15:34
  • $\begingroup$ So, $\forall a,b \in \mathbb{C}G$, $\exists c \in \mathbb{C}G$ such that $(a^{\star}a) * (b^{\star}b) = c^{\star}c$. Is there a formula for $c$? $\endgroup$ Commented Aug 6, 2017 at 17:23
  • $\begingroup$ I agree that the argument works for any finite-dimensional Hopf C*-algebra. Are these the same as finite-dimensional Kac algebras? $\endgroup$ Commented Aug 6, 2017 at 18:06
  • 1
    $\begingroup$ Yes, Kac algebra and Hopf C*-algebra are synonymous. $\endgroup$ Commented Aug 6, 2017 at 18:21

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .