6
$\begingroup$

Suppose $A$ is a Hilbert-Schmidt operator on a Hilbert space $\mathcal{H}$, then when is it true that

$$\langle Ax, x \rangle_{\mathcal{H}} = \langle A, x \otimes x \rangle_{\text{HS}}, \quad \forall \, x \in \mathcal{H}$$

I know from the definition of Hilbert-Schmidt inner product and the definition of tensor product that

$$ \langle A, x \otimes x \rangle_{\text{HS}} = \sum_{j \in J} \langle A e_j, \langle x, e_j \rangle_{\mathcal{H}} x \rangle_{\mathcal{H}} $$ where $\{e_j\}_{j \in J}$ is an arbitrary ONB of $\mathcal{H}$. However, I am unable to simplify further. Any help would be appreciated.

If the above result is incorrect, in general, I am interested in going from an expression of the form $\langle Ax, x \rangle_{\mathcal{H}}$ to an expression of the form $\langle A, x \otimes x \rangle_{\text{HS}}$. How could I do that?

Also, I would really appreciate some references where I can look into these kind of results in more detail.

$\endgroup$
3
  • 1
    $\begingroup$ Shouldn't this follow simply from a resolution of the identity? You have outer products of $e_j$ in there which sum to $I$. $\endgroup$ Apr 4 at 4:08
  • $\begingroup$ I am sorry, but I am very new to functional analysis. Could you be more detailed about what you mean by the outer product of $e_j$ and how it follows from resolution of the identity? Let me read up on resolution of the identity till you reply. $\endgroup$ Apr 4 at 4:15
  • $\begingroup$ @CameronWilliams forgot to tag you. $\endgroup$ Apr 4 at 4:28
4
$\begingroup$

What you need to do is use an orthonormal basis whose first element is $x/\|x\|$. Then $$ \langle A,x\otimes x\rangle_{\rm HS}=\operatorname{Tr}(A(x\otimes x))=\langle A(x\otimes x)\tfrac{x}{\|x\|},\tfrac{x}{\|x\|}\rangle=\langle Ax,x\rangle, $$ since $(x\otimes x)x=\|x\|^2\,x$.

$\endgroup$
4
  • $\begingroup$ This is brilliant! Just to confirm one thing though: $\langle A , x \otimes x \rangle_{\text{HS}}$ should equal $\text{Tr}(A^* (x \otimes x))$, right? and therefore we have $\langle A , x \otimes x \rangle_{\text{HS}} = \langle A^* x , x \rangle$. Or is it true that $A$ being Hilbert-Schmidt must also be self-adjoint? $\endgroup$ Apr 4 at 4:45
  • 1
    $\begingroup$ No, $\langle A,x\otimes x\rangle$ is $\operatorname{Tr}((x\otimes x)^*A)$. And $x\otimes x$ is selfadjoint. $\endgroup$ Apr 4 at 4:55
  • $\begingroup$ Okay, must be different conventions then? The Wikipedia page says $\langle A, B \rangle_{\text{HS}} = \text{Tr}(A^* B)$. $\endgroup$ Apr 4 at 4:59
  • 1
    $\begingroup$ It's about convention. If you want to use the physicist's convention, then you also need to use it in the definition of the trace. $\endgroup$ Apr 4 at 5:05

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.