The supercuspidal representations of \(p\)-adic classical groups (Q2482719)

The paper under review proves a major result: a construction of all supercuspidal representations of \(p\)-adic classical groups. Suppose that \(F\) is a nonarchimedean local field and \(G\) the group of \(F\)-rational points of a connected reductive \(F\)-group. A longstanding problem is to construct the irreducible, admissible, complex representations of \(G\). The supercuspidal representations of \(G\) and of its Levi subgroups are in some sense the building blocks for the representation theory of \(G\), so constructing them is an important subproblem. All known supercuspidal representations can be constructed as induced representations from open subgroups that are compact modulo the center of \(G\). In general, it is difficult to construct supercuspidal representations, and even more difficult to show that one has found all of them. If one assumes that the residual characteristic \(p\) of \(F\) avoids certain primes associated to the root system or the splitting field of \(G\), then the task can be greatly simplified. For example, R. Howe constructed a family of supercuspidal representations of \(\text{GL}_n(F)\), and A. Moy showed that Howe's construction is exhaustive when \(p\) is coprime to \(n\). Howe's construction has several descendants, the most general of which is due to J.-K. Yu. When \(F\) has characteristic zero and \(p\) is very large, then J.-L. Kim has shown that Yu's construction is exhaustive. If we drop the assumptions on \(p\) that underlie Yu's construction then the problem becomes much harder. (However, we have had for some time complete information concerning the so-called ``depth-zero'' supercuspidal representations, which were constructed independently by L. Morris, and by A. Moy and G. Prasad.) P. Kutzko constructed all supercuspidal representations of \(\text{GL}_2(F)\) without making any assumptions on \(p\). A series of advances eventually culminated in the construction by C. Bushnell and P. Kutzko of all supercuspidal representations of \(\text{GL}_n(F)\). Using this last construction as a tool, Bushnell and Kutzko have also constructed the supercuspidal representations of \(\text{SL}_n(F)\), and V. Sécherre and the author have done the same for \(\text{GL}_n(D)\), where \(D\) is a central division algebra over \(F\). The paper under review, the culmination of a series of papers by the author, advances the theory further. Suppose that \(G\) is symplectic, special orthogonal, or unitary, and that \(p\) is odd. Then the author constructs all irreducible, supercuspidal representations of \(G\) (many of which had never been constructed before), and shows that they are all induced. In outline, the construction for \(G\) is similar to that for \(\text{GL}_n(F)\), but some of the technical details (for example, those concerning so-called ``\(\beta\)-extensions'') are more difficult. As in the case of \(\text{SL}_n(F)\), the construction is not intrinsic, in that it relies on an understanding of the analogous construction for a general linear group in which \(G\) embeds. The proof of exhaustion uses the Bushnell-Kutzko theory of covers.