Let $V_1$ and $V_2$ be Fréchet spaces. Let $\{ \lVert \cdot \rVert_{1,n} \}_{n \in \mathbb{N}}$ be a family of seminorms for $V_1$ and similarly $\{ \lVert \cdot \rVert_{2,n} \}_{n \in \mathbb{N}}$ for $V_2$.

Now, let $L(V_1, V_2)$ be the space of continuous linear maps from $V_1$ into $V_2$. Then, $L(V_1, V_2)$ is a vector space itself. However,

I cannot come up with a "sensible" (or "natural") topology for $L(V_1,V_2)$, making it a topological vector space.

If $V_1$ and $V_2$ are Banach spaces, then $L(V_1,V_2)$ itself is a Banach space with the operator norm.

I have checked if it is possible to apply the strong dual topology to $L(V_1,V_2)$. However, the strong dual topology requires a dual system to start with, so I don't think it is suitable for $L(V_1, V_2)$.

Could anyone help me?


1 Answer 1


For $V_2=\mathbb K$, the field of real or complex numbers, $L(V_1,V_2)=V_1'$ is the continuous dual of $V_1$. The generalization of the strong topology on $V_1'$ to $L(V_1,V_2)$ is the topology of uniform convergence on bounded sets which is described by the seminorms $p_{B,n}(T)=\sup\{\|T(x)\|_{2,n}: x\in B\}$ for $n\in\mathbb N$ and $B\subseteq V_1$ bounded. This makes $L(V_1,V_2)$ a complete locally convex space. Instead of all bounded sets one can consider smaller classes, e.g., uniform convergence on all finite or on all compact subsets of $V_1$.

None of these topologies makes $L(V_1,V_2)$ again a Fréchet space. However, this space of operators has the natural structure of countable projective limit of countable inductive limits of Banach spaces, a so-called PLB-space. This structure is used for splitting theorems for Fréchet spaces via the functor Ext$^1(V_1,V_2)$. If house advertisement is allowed, I would recommend my Springer Lecture Notes Derived Functors in Functional Analysis.

  • 3
    $\begingroup$ I won't edit in a link in case you prefer not to link to your own work, but here it is in a comment: Wengenroth - Derived functors in functional analysis. $\endgroup$
    – LSpice
    Commented Jul 9 at 14:44
  • $\begingroup$ Thank you for your clarification and reference! $\endgroup$
    – Isaac
    Commented Jul 9 at 14:55
  • $\begingroup$ Just one more. If $V_1$ and $V_2$ are Montel spaces (e.g. nuclear Frechet), then is $L(V_1,V_2)$ also a Montel space? You said it is a PLB-space, which involves Banach spaces. So, I am a bit confused here. $\endgroup$
    – Isaac
    Commented Jul 11 at 21:44
  • 1
    $\begingroup$ At least, for nuclear Fréchet spaces, $L(V_1,V_2)$ is a PLB-space where the LB-spaces are (can be chosen to be) nuclear. Then $L(V_1,V_2)$ is also nuclear and hence Montel (this is somewhat implicitly alreqady contained in Grothendieck's thèse since $L(V_1,V_2)\cong V_1' \hat\otimes V_2$). $\endgroup$ Commented Jul 12 at 8:36
  • $\begingroup$ @JochenWengenroth I am aware that a nuclear Frechet space is Montel, but you said $L(V_1,V_2)$ is not a Frechet space. I guess it is still Montel somehow? $\endgroup$
    – Isaac
    Commented Jul 12 at 13:10

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.