3
$\begingroup$

Let $X$ be a locally convex space, and $\{Y_i;\ i\in I\}$ a covariant system of locally convex spaces over a partially ordered set $I$, i.e. a system of linear continuous mappings is given $\sigma^j_i:Y_i\to Y_j$, $i\le j\in I$ such that $$ \sigma^k_j\circ\sigma^j_i=\sigma^k_i, \qquad i\le j\le k\in I. $$ (We assume that all locally convex spaces are over a common field, say, $\mathbb C$. If necessary, we can think that $X$ and $Y_i$ are complete.)

Let $Y$ be the injective limit of $\{Y_i;\ i\in I\}$ in the category of locally convex spaces $$ Y=\injlim_{i\in I}Y_i. $$ For each $i\in I$ we consider the space ${\mathcal L}(X,Y_i)$ of operators (i.e. linear continuous mappings) $\varphi:X\to Y_i$ endowed with the compact-open topology. Similarly, we endow the space ${\mathcal L}(X,Y)$ of operators $\varphi:X\to Y$ with the compact-open topology.

The family of spaces $\{{\mathcal L}(X,Y_i);\ i\in I\}$ is also a covariant system over $I$, and we can consider its injective limit in the category of locally convex spaces: $$ \injlim_{i\in I}{\mathcal L}(X,Y_i). $$ There is a natural mapping $$ \omega:\injlim_{i\in I}{\mathcal L}(X,Y_i)\to {\mathcal L}(X,Y). $$

Question:

Is it true that the topology of $\injlim_{i\in I}{\mathcal L}(X,Y_i)$ is inherited from ${\mathcal L}(X,Y)$?

In other words,

if $U$ is a neighbourhood of zero in $\injlim_{i\in I}{\mathcal L}(X,Y_i)$ do there always exist a compact set $T\subseteq X$ and a neighbourhood of zero $V$ in $Y$ such that $$ U\supseteq \omega^{-1}(\{\varphi\in{\mathcal L}(X,Y):\ \varphi(T)\subseteq V\} ) $$ ?

$\endgroup$
2
  • $\begingroup$ If I understand your question correctly, the answer is no in this generality (let $X$ be a large vector space regarded as the lcs inductive (=? injective) limit of its finite dimensional subspaces. Then the identity mapping on $X$ doesn‘t factor through a component). On the positive side, there was a considerable amount of work done in the golden age of lcs theory on special situations—Involving, e.g., $\cal F$- and $\cal {LF}$-spaces. For example, by Grothendieck in his thesis. $\endgroup$ Commented May 31, 2024 at 14:02
  • $\begingroup$ @crow I don't understand you. Do you mean that the mapping $\injlim_{i\in I}{\mathcal L}(X,Y_i)\to {\mathcal L}(X,Y)$ is not necessarily surjective? I understand this, my question is not about this. I'll edit it. $\endgroup$ Commented May 31, 2024 at 14:23

1 Answer 1

4
$\begingroup$

The answer to your question is certainly negative. Before giving a reference let me spell out what it means that the injection $\injlim L(X,Y_i) \to L(X,\injlim Y_i)$ is an isomorphism onto its range: $0$-neighbourhoods in $\injlim L(X,Y_i)$ are absolutely convex hulls of unions of the form $W_i=\{T\in L(X,Y_i): T(K_i)\subseteq V_i\}$ where $K_i$ are compact in $X$ and $V_i$ are $0$-neighbourhoods in $Y_i$ whereas a $0$-neighbourhood $W$ in $L(X,Y)$ is defined by a single compact set $K\subseteq X$ and a $0$-neighbourhood $V$ in $\injlim Y_i$. You need that, for all such $W_i$ there is one $W$ with $$W\cap \injlim L(X,Y_i) \subseteq \Gamma(\bigcup_i W_i).$$ Despite the difficulty that you have to decompose continuous linear operators in a uniform way, you first have to manufacture some candidates for $V$ depending on the given $V_i$ -- a natural first try would be the absolutely convex hull of the union of the $V_i$ -- and, in particular, for a compact set $K$ depending on the given $K_i$ -- and without further very special requirements for $X$ and/or $Y$ you have no chance even for countable inductive limits.

As a reference with much more information about such problems I recommend the work

Concrete counterexamples for $Y$ even a countable locally convex direct sum of the field are contained in chapter 2, $X$ can be taken as a product $\mathbb K^I$ for a set $I$ of cardinality at least that of $\mathbb R$.

$\endgroup$
3
  • $\begingroup$ Jochen, I need some time to find this paper by Susanne Dierolf. $\endgroup$ Commented Jun 9, 2024 at 12:40
  • 2
    $\begingroup$ @SergeiAkbarov I added a link $\endgroup$ Commented Jun 13, 2024 at 8:09
  • $\begingroup$ Thanks for adding the link. $\endgroup$ Commented Jun 13, 2024 at 9:21

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.