MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

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

Suppose I have a Banach space $E$ (which may be finite dimensional if you wish), a Hilbert space $H$ and a tensor $\tau \in H\otimes E$ in the algebraic tensor product. There are lots of ways to choose a measure $\mu$ and an isometry (not assumed surjective) $\theta:H\rightarrow L^2(\mu)$. Then $(\theta\otimes\iota)\tau \in L^2(\mu)\otimes E \subseteq L^2(\mu;E)$, and so I can compute the norm of $(\theta\otimes\iota)\tau$ in the vector-valued space $L^2(\mu;E)$.

Is there an intrinsic (or simple, etc.) characterisation of the infimum (over all choices of $\theta$ and $\mu$) of this norm?

(The infimum is non-zero, assuming $\tau\not=0$, as it's always larger than the injective tensor norm. But it's not obvious to me that you actually get a norm on $H\otimes E$ from this).

If $E$ is a Hilbert space, then the norm is independent of the choice of $\mu$ and $\theta$; you just get the Hilbert space tensor product norm. But what if, say, $E$ is a finite-dimensional $\ell^\infty$ space?

share|cite|improve this question
Hi Matt! Are you placing any kind of restriction on $\mu$? Are you allowing any measure space at all? It's rather a long shot, but you know the Nagy-Foias functional model for contraction operators between Hilbert spaces? That might change your question into an equivalent question about Hankel and Toeplitz operators - but it probably will be equally difficult. – Zen Harper Jun 7 '11 at 1:10
@Zen: Well, I did originally mean any measure. But as Pietro correctly says, you can obviously approximate and work just in $\ell^2$... – Matthew Daws Jun 7 '11 at 9:17

Here is a first step. Let $\tau=\sum_{i=1}^n h_i\otimes u_i$ with $h_1,\dots h_n\in H$ and $u_1,\dots u_n\in E$, where w.l.o.g. $h_i$ are orthonormal.

Then, in your infimum, you may fix $\mu$ to be the counting measure on $\mathbb{N}$, so that $L^2(\mu)=\ell^2$ (this follows from a simple argument using the density of simple functions and the Gram-Schmidt orthonormalization process), and the infimum writes

$$\inf\Bigg( \, \sum_{k=0}^\infty \, \Bigg \| \, \sum_{i=1}^n \lambda_{k,i}u_i \Bigg\|_E^2 \, \Bigg)^{1/2}$$

taken over all $\lambda_{k,i}$ with $\sum_{k=0}^\infty \lambda_{k,i} \lambda_{k,j}=\delta_{i,j} $. Then, it is not clear to me how to make a further reduction, even in the case of $n=2$ vectors.

share|cite|improve this answer

Your Answer


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.