I do not know too much about linear operators so forgive me if this doesn't make much sense, but what would the space of linear operators on $C^\infty[a,b]$ be defined as?

If we denote this space as $L(C^\infty[a,b])$ what can we say about it? Since $C^\infty[a,b]$ is a Frechet space is $L(C^\infty[a,b])$ also a Frechet space? What would be a useful norm (seminorms?) on $L(C^\infty[a,b])$? Given a norm or seminorms could you then talk about sequences of operators and their convergence?

Also given that the space $L(C^\infty[a,b])$ exists would the space of linear differential operators on $C^\infty[a,b]$ be a subspace of $L(C^\infty[a,b])$?


The space $L(C^\infty[a,b])$ of continuous linear operators on $C^\infty[a,b]$ is - excepting the trivial case $a = b$ - not a Fréchet space (in the usual topologies).

The usual topologies on $L(C^\infty[a,b])$ are

  • the topology of uniform convergence on all bounded subsets, often (well, sometimes, at least ;) denoted $L_b(C^\infty[a,b])$,
  • the topology of uniform convergence on all compact subsets, often denoted $L_c(C^\infty[a,b])$,
  • and the topology of uniform convergence on all finite subsets (or pointwise convergence), also denoted by $L_p(C^\infty[a,b])$.

The seminorms generating the respective topology would be

$$p_{q,M}(T)=\sup\{q(Tx):x\in M\},$$

where $q$ is a continuous seminorm on $C^\infty[a,b]$, and $M$ ranges over the family of subsets of $C^\infty[a,b]$ under consideration, bounded, compact, or finite.

These topologies are not metrizable, since there is no countable basis of bounded/compact/finite sets in $C^\infty[a,b]$, hence no countable subset of seminorms generates the topology.

The space of linear differential operators (with smooth coefficients) is a subspace of $L(C^\infty[a,b])$, since such an operator is a continuous linear operator on $C^\infty[a,b]$.


Adding a little to @DanielFischer's answer: the space $C^\infty[a,b]$, being a nested intersection, is a projective limit of Banach spaces. Thus, any (continuous linear) map to it from a TVS is exactly induced (uniquely) from a collection of (continuous linear) maps to the limitands $C^k[a,b]$. In particular, this is true of self-maps, so every such just comes from a compatible family of maps $C^\infty[a,b]\to C^k[a,b]$. Since the latter is Banach, it is an easy exercise to see that such a map factors through some $C^\ell[a,b]\to C^k[a,b]$ (with $\ell$ depending on $k$). Maps between Banach spaces have a natural selection of topologies on them, for example the operator-norm, but also the strong and weak, as delineated by DanielFischer.

Further, possibly relevantly depending on one's situation, there is a Schwartz kernel theorem valid for maps $C^\infty[a,b]\to (C^\infty[a,b])^*$ from smooth functions to distributions... namely, that every such is given by a distribution in two variables.


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.