10
$\begingroup$

Does there exists a triple $(G, X, \pi)$, where $G$ is a compact group, $X$ an infinite dimensional Banach space over $\mathbf{C}$, and $\pi : G \to B(X)$ a strongly continuous representation of $G$, such that if $Y$ is a non-zero closed invariant subspace of $X$ in the sense that $\pi(g)Y \subseteq Y$ for all $g \in G$, then $Y = X$ (we call nonzero strongly representations of $G$ satisfying this property topologically irreducible)?

I am not sure that if this is settled already, as the question seems natural, yet I am ignorant of the relevant literature, nor could I come up with a proof or disproof. Let me be (perhaps overly) precise about the question asked in the title. By compact group, I mean Hausdorff plus quasi-compact, and topologically irreducible means that the only invariant closed subspaces are either $0$ or the whole space. More generally, can we find such an example by replacing infinite dimensional Banach spaces by complete infinite dimensional locally convex Hausdorff spaces instead? While we are on the subject, I'd also appreciate some recommendation of books, papers/surveys on representations of topological groups on Banach spaces that is not focused on unitary representations on Hilbert spaces?

$\endgroup$
2
  • $\begingroup$ Could you write quantifiers to make the question clear? Do you mean: does there exist a compact group and an infinite-dim Banach space and an strongly continuous irreducible representation of this group in this space? $\endgroup$
    – YCor
    Commented Mar 3, 2022 at 22:51
  • $\begingroup$ @YCor Done, and I also added the requirement of completeness in the locally convex case. $\endgroup$
    – Hua Wang
    Commented Mar 3, 2022 at 23:17

2 Answers 2

9
$\begingroup$

For Banach spaces, the question is no : there is a form of the Peter-Weyl theorem, due to Shiga, which implies that in every Banach space representation of a compact group, the finite-dimensional sub-representations span a dense subspace. In particular, strongly continuous irreducible representations on a Banach space are finite-dimensional. I am not sure about arbitrary locally convex topological vector space.

Shiga's paper is here. If the link does not work, the precise reference is:

K. Shiga, Representations of a compact group on a Banach space, Journal of the Mathematical Society of Japan 7 (1955), 224–248.

$\endgroup$
1
  • $\begingroup$ This is exactly what I was looking for, thanks! For the locally convex case, I added the requirement that the space should be complete. $\endgroup$
    – Hua Wang
    Commented Mar 3, 2022 at 23:17
5
$\begingroup$

Another reference is

Hofmann, Morris, "The structure of compact groups", Zbl 1277.22001

(There is a 4th edition, but my library has access to this 3rd edition which I'll give references to).

Chapter 3 considers this question, and the "Big Peter and Weyl Theorem", Theorem 3.51, is exactly the result which Mikael de la Salle mentions, but is here stated and proved for any "$G$-complete locally convex" space. In particular, this applies to any complete locally convex topological space, see page 72.

(There are some very brief references given, but no bibliographic notes, so I don't know where this result was first proved.)

$\endgroup$
2
  • 4
    $\begingroup$ I also knew this as Shiga's theorem. (Of course, the original case, namely Peter-Weyl from the 1920s, deserves credit too). Looking at Shiga's paper, it seems to be inspired by a weaker 1949 result of Kaplansky asserting that there's no infinite-dimensional irreducible Banach $G$-module for $G$ compact (I. Kaplansky Primary ideals in group algebras, Proc. Nat. Acad. Sci. U.S.A. 35 (1949), 133-136.). $\endgroup$
    – YCor
    Commented Mar 4, 2022 at 8:18
  • $\begingroup$ Thanks a lot for this nice reference! $\endgroup$
    – Hua Wang
    Commented Mar 4, 2022 at 10:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.