1
$\begingroup$

Let $K$ and $L$ be compact spaces. Consider the Banach algebra $V(K,L)=C(K)\mathbin{\hat\otimes} C(L)$ , which is the completion of the $C(K)\otimes C(L)$ with respect to the projective tensor norm. It is known that $V(K,L)$ is a subalgebra of $C(K\times L)$ , not isometrically. I am trying to establish this result . Heres what I have figured out. Due to the universal property of tensor product , there is an algebra homomorphism $\theta : C(K)\otimes C(L)\to C(K\times L)$ such that $$\theta(\sum_{i=1}^nf_i\otimes g_i)(x,y)=\sum_{i=1}^nf_i(x)g_i(y)$$

I want to show that this map is injective. Once this is done then we can extend it to its completion.

$\endgroup$
  • $\begingroup$ But be careful: The extension of an injective uniformly continuous map to the completion needn't be injective! Consider a Banach space $(X,\|\cdot\|)$ and a strictly finer norm $\||\cdot\||$ on it (e.g., $\|x\|+|f(x)|$ for a discontinuous linear functional). Then the identity from $(X,\||\cdot\||)$ to $(X,\|\cdot\|)$ is continuous and injective but its extension to the completion is not injective (because otherwise its inverse would be continuous by the closed graph theorem). $\endgroup$ – Jochen Jan 15 at 13:47
  • $\begingroup$ @jochen Thanks for reminding , I was actually thinking the same . $\endgroup$ – Lav Kumar Jan 16 at 13:55
2
$\begingroup$

Suppose $h\in C(K)\otimes C(L)$ and $\theta(h)=0$. Write $h=\sum_{i=1}^nf_i\otimes g_i$, where the $g_i$ are linearly independent. But then this implies that for all $x\in K$, $$0=\theta(h)(x,\cdot)=\sum_{i=1}^nf_i(x)g_i(\cdot),$$ whence $f_i=0$ for all $i$, and thus $h=0$.

$\endgroup$
  • $\begingroup$ Thank you so much. $\endgroup$ – Lav Kumar Jan 14 at 16:59
  • 1
    $\begingroup$ You're welcome. Glad to help! $\endgroup$ – Aweygan Jan 14 at 17:00
  • $\begingroup$ Is the choice of representation of h with g's linearly independent comes from smallest representation of h as sums of elementary tensors ? $\endgroup$ – Lav Kumar Jan 14 at 17:02
  • $\begingroup$ I wouldn't think so, but possibly. You might be able to do it with the $f_i$ linearly independent and obtain a representation with less elementary tensors. $\endgroup$ – Aweygan Jan 14 at 17:09

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.