4
$\begingroup$

Let $X,Y$ be smooth compact manifolds. Let $C^\infty(X)$ and $C^{-\infty}(X)$ denote the spaces of smooth and generalized functions on $X$ respectively. We have the obvious canonical linear map $$T\colon C^{-\infty}(X\times Y)\to Bil(C^\infty(X),C^\infty(Y)),$$ where the target is the space of continuous bilinear functionals $C^\infty(X)\times C^\infty(Y)\to \mathbb{C}$. By definition $(T\Phi)(f,g)=\Phi(f\otimes g)$.

It is well known that $T$ is isomorphism of vector spaces. For which standard topologies on the source and the target $T$ is an isomorphism of topological vector spaces?

For example let us choose on $C^{-\infty}(X\times Y)$ the strong topology, and on the target the topology given by seminorms $$||B||_{K,L}:=\sup_{k\in K,l\in L}|B(k,l))|,$$ where $K\subset C^\infty(X), L\subset C^\infty(Y)$ are arbitrary bounded subset. Then $T$ will be a continuous map for these topologies. But is $T$ a topological isomoprhism?

$\endgroup$
  • $\begingroup$ This true and one of the manifold formulations of the celecbrated kernel $\endgroup$ – priel Jul 21 '15 at 8:03
  • $\begingroup$ celebrated kernel theorem of L. Schwartz. Primary references: Grothendieck's thesis and Schwartz' articles on vector-vaued distributions and smooth functions which are easy to trace. Secondary: the books of F. Treves. $\endgroup$ – priel Jul 21 '15 at 8:10
2
$\begingroup$

$T$ is indeed a topological isomorphism. Indepently of the concrete situation, this can be shown by a suitable version of the closed graph theorem, e.g. the one of de Wilde: the domain has to be ultrabornological an the range should have a web (see, e.g., the book Introduction to Functional Analysis of Meise and Vogt, chapter 24). Here $T$ acts from the dual of the nuclear Frechet space $C^\infty(X\times Y)$ to $Bil(C^\infty(X),C^\infty(Y))\cong L(C^\infty(X),C^{-\infty}(Y))$. It follows e.g. from Grothendieck's work that both spaces are ultrabornological, and both spaces also have web. Therefore, one can apply the closed graph theorem fot $T$ and its inverse. (One can even show that the domain and range space are both isomorphic to $s'$, the space of slowly increasing sequences.)

| cite | improve this answer | |
$\endgroup$

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.