Topological Jordan decompositions (Q932856)

In [On lifting. Lie group representations II, Proc. Spec. Year, Univ. Md., College Park 1982-83, Lect. Notes Math. 1041, 209--249 (1984; Zbl 0538.20014)], \textit{D. Kazhdan} introduced a decomposition \(k=su\) of an element \(k\) in \(\mathrm{GL}(n,R)\), where \(R\) is the ring of integers in a \(p\)-adic field \(F\), as a commuting product of an ``absolutely semisimple'' element \(s\) (whose order is a power of \(q\), the residual cardinality of \(F\)), and a ``topologically unipotent'' element \(u\) (with \(u^{q^m}\to 1\) as \(m\to \infty\)). This he used to reduce the computation of orbital integrals at \(k\) to such integrals at \(u\) in the centralizer of \(s\), as part of his proof of the lifting of characters of the multiplicative group of a cyclic extension \(E\) of \(F\) of degree \(n\), to representations of \(\mathrm{GL}(n,R)\), locally and globally, and to characterize the representations of \(\mathrm{GL}(n,R)\) invariant under twisting with a character of order \(n\). The technique was used later to compare twisted orbital integrals of \(\mathrm{PGL}(3)\) with those of \(\mathrm{SL}(2)\), as needed for the symmetric square lifting [\textit{Y. Z. Flicker}, Pac. J. Math. 175, No. 2, 507--526 (1996; Zbl 0865.11045), and in Mem. Am. Math. Soc. 655, 112 p. (1999; Zbl 0916.11031)], to match orbital integrals on \(\mathrm{GL}(4)\) and \(\mathrm{GSp}(2)\), thus proving the fundamental lemma in these two cases. The author, with J. Adler, introduced a notion of a normal approximation, a refinement of the topological Jordan decomposition, but needed the properties of the topological Jordan decomposition over a discretely valued field \(F\) which is not necessarily locally compact. This paper establishes the needed refinement, as well as an abstract generalization to profinite groups. An important ingredient in the proof is an analogue of a fixed point result of Prasad and Yu.