3
$\begingroup$

Hi, I am interested in the following question:

Fix $n$. Let $M_n$ be matrix algebra over the field of complex numbers with normalized trace $tr_n$. Let $M_n^{\omega}$ be an ultrapover of $M_n$, namely we consider the algebra of all bounded (in norm) sequences in $M_n$,say $l_{\infty}(M_n)$, and take a quotient of this space by sequences $(a_i)_{i\in\mathbb{N}}$ with $lim_{\omega} tr_n(a_i^*a_i)=0$. Is $M_n^{\omega}$ is finite-dimensional? What is the structure of $M_n$?

Also, if $F_i$ is $n$-dimensional Banach space, what is the dimension of the Banach space ultraproduct of $\{F_i\}$.

$\endgroup$
1

1 Answer 1

4
$\begingroup$

If the space $F$ is $k$-dimensional with basis $b_1,\ldots b_k$ then the ultraproduct with respect to the ultrafilter $U$ will also be $k$-dimensional. To see this let $x_i\in \ell_\infty(F_i)$ and let $q_{i,j}$ be scalars such that $\sum_{j=1}^kq_{i,j}b_j =x_i $. By the compactness of the ball in finite dimensional spaces the $q_{i,j}$ converge to $q_{j}$ with respect to $U$ and so $\sum_{j=1}^kq_{j}b_j = x$ modulo $U$.

The same argument works of different $k$-dimensional Banach spaces as long as there is a uniform bound on the norm of the identity map from one space to the other to allow the compactness argument to be used --- in other words, there is a single ball containing the unit balls of all the $F_i$.

$\endgroup$
3
  • 1
    $\begingroup$ The condition on your second paragraph is always true, as a k-dim Banach space is always at most $\sqrt{k}$ distant from $\ell^2_k$ in the Banach-Mazur distance-- see en.wikipedia.org/wiki/Banach%E2%80%93Mazur_compactum $\endgroup$ Jul 6, 2011 at 12:37
  • $\begingroup$ Thanks Matthew. I had had in mind a sequence of 2-dimensional spaces with increasingly oblong unit balls. I will have to think about why the bounded Banach-Mazur distance saves the day for such a product. $\endgroup$ Jul 6, 2011 at 12:58
  • $\begingroup$ @Juris-- But then the "correct" unit length basis vectors also become enlongated, and so the distortions sort of cancel out... $\endgroup$ Jul 6, 2011 at 15:04

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.