MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let's consider a separable Hilbert space $(\mathcal H, \langle\cdot, \cdot\rangle_{\mathcal H})$ with Norm $||\cdot||_{\mathcal H} := \langle\cdot, \cdot\rangle^{1/2}_{\mathcal H},$ orthonomal basis $(e_j)$ of $\mathcal H$ and let $s\colon \mathcal H \rightarrow \mathcal H$ be a Hilbert Schmidt operator, denoted by $s \in \mathcal{S_H}.$ It's known that $(\mathcal{S_H}, \langle\cdot, \cdot\rangle_{\mathcal{S_H}})$ is a (separable ??) Hilbert space and for $s_1, s_2 \in \mathcal{S_H}\colon$ $\langle s_1, s_2\rangle_{\mathcal{S_H}}=\sum_{j=1}^\infty\langle s_1(e_j), s_2(e_j)\rangle_{\mathcal H},$ such that $||s_1||^2_{\mathcal{S_H}} = \sum_{j=1}^\infty||s_1(e_j)||^2_{\mathcal H}.$ Now, assuming that $\mathcal{ S_H}$ indeed is separable, let $(\phi_j)$ be an orthonormal basis of $\mathcal S_{\mathcal H}.$ It easily can be shown, since $||\cdot||_{\mathcal{S_H}}$ is sub-multiplicative, that for $s \in \mathcal{S_H}\colon s\colon \mathcal{S_H} \rightarrow \mathcal{S_H}$ is a bounded operator as well, denoted by $s \in \mathcal{L_{S_H}}.$

Is $s\colon\mathcal{S_H} \rightarrow \mathcal{S_H}$ "Hilbert-Schmidt" as well ($s \in \mathcal{S_{S_H}}$)?

Thank you in advance!

share|cite|improve this question
5  
I don't know what you're talking about, but the title is lovely. – lentic catachresis Mar 23 at 9:12
    
Yeah ;). But it is how it is ;). – Obriareos Mar 23 at 9:16

No it's not.

If $s:H\to H$ is a rank one projection, then $s\otimes\mathrm{id}:H\otimes \overline H\to H\otimes \overline H$ is a projection of rank $\dim(H)$.

In particular, if $H$ is infinite dimensional, then $s\otimes\mathrm{id}:H\otimes \overline H\to H\otimes \overline H$ is a projection of infinite rank, and thus not Hilbert-Schmidt.

Here, I've identified the space of Hilbert-Schmidt operators with the Hilbert space tensor product $H\otimes \overline H$.

share|cite|improve this answer
    
Thank you very much André! Just to make sure: I work on hilbert spaces on $\mathbb R.$ I've got some questions: 1. I'm not familiar with the tensor product - I googled it, but why is it important? 2. What do you mean with $\bar{H}$ ? 3. Projections of infinite rank automatically are'nt Hilbert-Schmidt? I have an idea, but I am not 100% sure! – Obriareos Mar 24 at 9:59
    
The bar indicates that I take the complex conjugate. When working over the real numbers, $\overline H = H$. – André Henriques Mar 24 at 16:48

The answer is No, assuming of course that for you $s$ acts on $\mathcal{S}_\mathcal{H}$ by left-multiplication. There is no need to assume that $\mathcal{S}_\mathcal{H}$ is separable, since it actually is, with an explicit countable orthonormal basis given by the elementary operators $e_{ij}$ that act as $e_{ij}(e_k) = e_i \delta_{jk}$ on the orthonormal basis of $\mathcal{H}$.

The Hilbert-Schmidt norm of $s$ on $\mathcal{S}_\mathcal{H}$ is $\operatorname{tr}(s' s)$, where $s'$ is the adjoint of $s$ on $\mathcal{S}_\mathcal{H}$. It is not hard to tell that $s' = s^*$, that is, left-multiplication by $s^*$, where $s^*$ is the adjoint of $s$ on $\mathcal{H}$. Let $r = s^* s$ on $\mathcal{H}$ and denote by $r$ also the left-multiplication by $r$ on $\mathcal{S}_\mathcal{H}$. Its trace is $\operatorname{tr}(r) = \sum_{ij} (e_{ij}, r e_{ij})_{\mathcal{S}_\mathcal{H}}$. A quick calculation gives $(e_{ij}, r e_{ij})_{\mathcal{S}_\mathcal{H}} = (e_i, r e_i)_\mathcal{H}$. So, each of the diagonal elements $(e_i, r e_i)_\mathcal{H}$ occurs infinitely many times in the summation for $\operatorname{tr}(r)$, which hence cannot be finite unless all diagonal elements of $r$ vanish. But we know that that's impossible from $r = s^* s$.

This answer is essentially the same as the one by André Henriques, which appeared as I was typing, but goes into more explicit detail.

share|cite|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.