Suppose $H_1$ and $H_2$ are two Hilbert spaces with dimension $n$ and $m$, for $ x \in H_1 \otimes H_2$

consider $$\|x\|_\pi = \inf \left\{ \sum_{i=1}^n \|a_i\| \|b_i\| : x = \sum_{i} a_i \otimes b_i \right\}$$ where $x \in A \otimes B$.

$ \|a\otimes b\|_{2}^2 =\langle a \otimes b,a\otimes b\rangle = \langle a,a\rangle_1 \, \langle b,b\rangle_2 \quad \mbox{for all } a \in H_1 \mbox{ and } b \in H_2 $ respectively.

Questions: 1. We know norms on finite dimensional spaces are equivalent. Hence there are $c_1$ and $c_2$ such that $$ c_1\|x\|_2 \leq \|x\|_\pi \leq c_2\|x\|_2 $$ since $\|x\|_\pi$ is largest cross norm so $c_1$ can be taken equal to $1$. what about $c_2$

2.What about when $H_1$ and $H_2$ are arbitrary Hilbert spaces, are these two norm equivalent?

  1. Where can I know more about such comparing norms on tensor of Hilbert spaces. (any book or article)

thanks, and pardon me for poor English.


The norms are equivalent if and only if $\min(\dim(H_1),\dim(H_2))<\infty$.

One way to think about this is to dualize. The space of bounded linear functionals on $(H_1 \otimes H_2, \Vert\cdot\Vert_\pi)$ can be identified naturally with the space of bounded linear maps from $H_1^*$ to $H_2$. On the other hand, the space of bounded linear functionals on $(H_1 \otimes H_2, \Vert\cdot\Vert_2)$ can be identified naturally with the space of Hilbert-Schmidt operators $H_1^*\to H_2$. If both $H_1$ and $H_2$ are infinite-dimensional then there are bounded linear maps $H_1^*\to H_2$ that are not Hilbert-Schmidt, e.g. the "identity map on $\ell_2$" if both $H_1$ and $H_2$ are separable.

A good source to learn about tensor norms on Banach spaces is Ryan's book Introduction to Tensor Products of Banach Spaces, or if you want something more condensed and more advanced, the classic book of Defant and Floret.

| cite | improve this answer | |
  • $\begingroup$ Excuse me @Yemon choi, It is my question but I cannot access my account, your answer is great and accepted. thank you very much. $\endgroup$ – reza Nov 4 '15 at 17:12
  • $\begingroup$ Just a minor detail, but googling suggests that the precise title of the book by Ryan is Introduction to Tensor Products of Banach Spaces (unless it's a different book). $\endgroup$ – M.G. Nov 4 '15 at 18:22
  • $\begingroup$ @July Oops, you are completely correct. I was looking at my bookshelf and somehow conflated two titles of adjacent books :) $\endgroup$ – Yemon Choi Nov 5 '15 at 1:33

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.