7
$\begingroup$

Let $H$ be a complex Hilbert space and $\mathcal{L}(H)$ be the algebra of all bounded linear operators on $E$.

If $A,B\in \mathcal{L}(H)$, It is true that $\overline{\text{Im}(A)}\otimes \overline{\text{Im}(B)}\subset \overline{\text{Im}(A\otimes B)}$?

I try as follows:

Let $z\in \overline{\text{Im}(A)}\otimes \overline{\text{Im}(B)}$, then $z=\displaystyle\sum_{i=1}^dx_i\otimes y_i$, where $x_i\in\overline{\text{Im}(A)},\;y_i\in\overline{\text{Im}(B)}$ and $d\in\mathbb{N}$. So, there exists $(\alpha_i(n))_{n\in\mathbb{N}}\subset \text{Im}(A)$ and $(\beta_i(n))_{n\in\mathbb{N}}\subset \text{Im}(B)$ such that $x_i=\lim_{n\rightarrow\infty}\alpha_i(n)$ and $y_i=\displaystyle\lim_{n\rightarrow\infty}\beta_i(n)$. Moreover, $\alpha_i(n)=A\tilde{\alpha}_i(n),\;\beta_i(n)=B\tilde{\beta}_i(n)$ and $$\alpha_i(n)\otimes\beta_i(n)=(A\otimes B)(\tilde{\alpha}_i(n)\otimes \tilde{\beta}_i(n))\subset \text{Im}(A\otimes B),$$ where $(\tilde{\alpha}_i(n))_{n\in\mathbb{N}}\subset \mathcal{H}$ and $(\tilde{\beta}_i(n))_{n\in\mathbb{N}}\subset \mathcal{H}$. Let $$z_n=\displaystyle\sum_{i=1}^d\alpha_i(n)\otimes \beta_i(n)\in \text{Im}(A\otimes B).$$ Since $\|x\otimes y\|=\|x\|\|y\|$, then the map $(x,y)\longmapsto x\otimes y$ is continuous. So $z_n\longrightarrow z$ as $n\longrightarrow \infty$, this implies that $z\in \overline{\text{Im}(A\otimes B)}$. As a consequence, $\overline{\text{Im}(A)}\otimes \overline{\text{Im}(B)}\subset \overline{\text{Im}(A\otimes B)}$.

$\endgroup$
2
  • 2
    $\begingroup$ The proof is okay. You did not specify the on the tensor product $H\otimes H$ in which you take the closure. Most naturally this would be the Hilbert-space or $\ell^2$ tensor product. But the proof works for all reasonable ones. $\endgroup$ Jan 18, 2018 at 7:31
  • $\begingroup$ @PeterMichor Dear Professor please do we have $Im(A\otimes B)=Im(A)\otimes Im(B)$? thanks $\endgroup$
    – Student
    Oct 3, 2023 at 18:36

1 Answer 1

7
$\begingroup$

The proof is okay. You did not specify the norm on the tensor product $H⊗H$ in which you take the closure. I do not know which one you want. Most naturally this would be the Hilbert-space or $\ell^2$ tensor product (which also describes the space of Hilbert-Schmidt operators). But the proof works for all reasonable norms. Another one describes the space of compact operators. Some of these are called Schatten-norms. See the beginnig of here, or also section 7 of here (and references therein) for more information.

Added:

Details of the proof:

The Hilbert-Schmidt norm on $\mathcal H\otimes\mathcal H$ can be defined as: $$ \|z\|_2 = \inf \Big\{ \|(\|x_i\|)\|_{\ell^2}.\sup_{\|y'\|\le1}\|(\langle y_i,y'\rangle)\|_{\ell^2} : z = \sum x_i\otimes y_i \Big\} $$ Here the infimum is taken over all (finite) representations of $z$ in $\mathcal H\otimes\mathcal H$. This can be seen from section 7 of the second reference I gave above (quite complicated there); best are the references Chevet or Saphar therein which I do not have available now. Given $z$ as in your proof above, choose a finite representation $z = \sum_{i=1}^{N(n)} x_i(n)\otimes y_i(n)$ such that $$ \|(\|x_i(n)\|)\|_{\ell^2}.\sup_{\|y'\|\le1}\|(\langle y_i(n),y'\rangle)\|_{\ell^2} \le \|z\|_2 +\frac{\epsilon}n\,. $$ Now choose your $\alpha_i(n,m)$ and $\beta_i(n,m)$ in such a way that $$ \|(\|x_i(n)-\alpha_i(n,m)\|)\|_{\ell^2}<\frac{\epsilon}{nm}\,,\qquad \|(\|y_i(n)-\beta_i(n,m)\|)\|_{\ell^2}<\frac{\epsilon}{nm}\,. $$ Then you can prove convergence in the $\ell^2$-tensor product.

$\endgroup$
4
  • $\begingroup$ It seems that both links point to the same document. $\endgroup$ Jan 18, 2018 at 23:12
  • $\begingroup$ Link corrected. Sorry. So the tensor product is the one describing Hilbert-Schmidt operators. $\endgroup$ Jan 19, 2018 at 9:29
  • $\begingroup$ I don't understand (both here and in the OP) why $z$ would have to be a finite sum of simple tensors. That is true in the algebraic tensor product, but normally one defines the Hilbert space tensor product as the completion of the algebraic version. $\endgroup$ Oct 4, 2023 at 14:11
  • $\begingroup$ Of course, the overall argument is still fine: one first shows that the algebraic tensor product of the closed images is contained in the RHS, and then one can take closures. $\endgroup$ Oct 4, 2023 at 14:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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