The size of infinite-dimensional representations (Q2407998)

This paper is based on the 18th Takagi Lectures that the author delivered for general audience on November 5--6, 2016. Thus the paper is a reorganization of the author's longstanding ingenious works on the associated varieties and Gelfand-Kirillov dimensions of (unitary) infinite dimensional representations of a real reductive group \( G \), and it will serve as a very good introduction to these subjects. In spite of its introductory nature, there are various new points of view to understand the results in a simpler and advanced way. Let \( G \) be a real reductive Lie group and \( K \) its maximal compact subgroup. Pick an (infinite dimensional) admissible representation \( \pi \) of \( G \). In many cases we are interested in the ``size'' of \( \pi \) imposing some measure which tells us how big \(\pi\) is. For compact groups, since irreducible representations are all finite dimensional, just the dimension of the representation space will serve as such a measurement. However, for noncompact \( G \), general representations are infinite dimensional, and a most popular notion of size is the Gelfand-Kirillov dimension, which was first introduced by the author in the late 1970's. Shortly after this, it was recognized as the (complex) dimension of the associated variety of \( \pi \) (Theorem 6.6). Roughly speaking, associated varieties are the base spaces on which representations live. So it is natural to expect the size should be the dimension of this base space. One of the most fundamental invariants of \( \pi \) is its global (or distributional) character. From this, we can define a wave front set first introduced by Roger Howe, which is a union of (real) nilpotent orbits in \( \mathrm{lie }{g} = \mathrm{Lie }G \). The associated variety of \( \pi \) is its algebraic counterpart which is a union of nilpotent \( K_\mathbb C \)-orbits in \( (\mathrm{lie }{g}_{\mathbb C}/\mathrm{lie }{k}_\mathbb C)^{\ast} \). These two invariants are easier to handle and give astonishingly precise information on \( \pi \). It is known that wave front sets and associated varieties are related by Sekiguchi correspondence and, in fact, open dense parts of them are \( K \)-equivariantly diffeomorphic (thus have the same real dimension). The author introduced a more subtle notion of associated cycle in the 1990's, which is now understood as an equivariant \( K \)-theory over the nilpotent cone (\S 9). The comprehensive details (and many interesting examples) are given in this paper without proofs. (An expert can fill all the details of the story or give proofs based on the arguments given here, I guess.) There is one exception of this style, which is Proposition 5.2 with a full proof. This proposition is of independent interest and I will summarize the statement here: for any Lie subalgebra \(\mathrm{lie }{h} \) of \(\mathrm{lie }{g} \), there is a Lie subalgebra \( \mathrm{lie }e{h}^1 \) of a parabolic subalgebra \( \mathrm{lie }{p} \) of \( \mathrm{lie }{g} \), such that \( \mathrm{lie }{h}^1 \in \overline{\mathrm{Ad}(G) \mathrm{lie }{h}} \) and the normalizer of \( \mathrm{lie }{h}^1 \) in \( \mathrm{lie }{g} \) is precisely \( \mathrm{lie }{p} \). I suppose this is a reply to one of the questions from the audience at the times of the lectures.