Correspondence of models under local theta lifting (Q2636365)

The paper under review studies model transitions under local theta correspondence over non-Archimedean fields. The main results are Theorem 3 and Theorem 4. Let \(m\), \(n\) be two integers such that \(m\geq n\). In Theorem 3, the author considered the type II dual pair \((G',G) := (\mathrm{GL}(2n), \mathrm{GL}(m))\). In Theorem 4, the author considered the type I dual pair \((G',G) := (\mathrm{U}(2n), \mathrm{U}(m))\) such that \(G'=\mathrm{U}(2n)\) is a split unitary group. The author showed that: Suppose that an irreducible smooth representation \(\pi'\) of \(G'\) is the theta lift of an irreducible smooth representation \(\pi\) of \(G\), then \(\pi'\) has a nonzero Shalika model only if \(\pi\) has a nonzero linear model of type \((n,m-n)\) (see Definition 4 and Definition 6). As a collorally, the author obtained the uniqueness of the Shalika model for the unitary group representation \(\pi'\) when \(m=n+1\) in Corollary 1. These results are generalizations of results of Jacquet-Shalika, Nien, Hanzer and Liu. On the other hand, all such kind of model transition results can be understood in terms of the transition of generalized Whittaker models discussed in Theorem 3.7 [\textit{C.-B. Zhu}, ``Local theta correspondence and nilpotent invariants'', Preprint, \url{arXiv:1802.01774}] due to Gomez-Zhu. We now illustrate how to obtain Theorem 3 from Theorem 3.7 of [loc. cit.]. First, note that, although it is stated for type I dual pairs, the results also hold for type II dual pairs with the same prove (at least in the non-Archimedean case). The nilpotent orbit \(\mathcal O'\) attached to the Shalika model is given by the Young diagram with two columns which are both of length \(n\). From Theorem 3.7 (a), we see that \(\pi'\) would not have a Shalika model if \(m< n\). Now assume \(m\geq n\). The generalized decent of \(\mathcal O'\) is the trivial orbit \(\mathcal O=\{0\}\). Moreover, under a suitable choice of \(X'\), one has: (1) \(M'_{X'} = M_{X,X'} = \mathrm{GL}(n)\) is identified with the diagnal group \(\{s(g,0)| g\in \mathrm{GL}(n)\}\) in the Shalika group \(S\); (2) \(L'\) is trivial; (3) \(L = \mathrm{GL}(n-m)\); and (4) \(\tau'\) and \(\Theta(\tau')\) are trivial representations. Now the space of the Shalika model \(\mathrm{Hom}_S(\pi', 1)\) is a quotient of \[ \begin{aligned} \mathrm{Hom}_{M'_{X'}}(\mathrm{Wh}_{\mathcal{O}',\tau'}(\Theta(\pi))) &= \mathrm{Hom}_{\mathrm{GL}(n)}(\mathrm{Wh}_{\mathcal{O}, \Theta(\tau')}(\pi), 1) \\ &= \mathrm{Hom}_{\mathrm{GL}(n)\times \mathrm{GL}(n-m)}(\pi, 1\boxtimes 1) \end{aligned} \] since \(\mathrm{Wh}_{\mathcal{O}, \Theta(\tau')}(\pi) = \mathrm{Hom}_{L}(\pi, 1)\). Theorem 4 could be deduced similarly and the character \(\xi\) (defined on line 3 of page 198) comes into the picture since the natural action of \(\widetilde{M'_{X'}}\) on the Schrödinger model is nonlinear. The proof of the main results in this paper is in fact the same as that of Gomez-Zhu. The key integredient is a formula on the twisted Jacquet module of the Weil representation, which are Propositions 1 and 4 in this paper and Proposition 3.8 in [loc. cit.]. The formula is proved in two steps: 1. find a suitable (mixed) model/polarization of the Weil representation (compare the paragraph after Theorem~3 and the middle part of page 195 with equation (13--15) in [loc. cit.]); 2. under that polarization, show that the twisted Jacquet module is realized as a space of functions supported on a certain \(M'_X\times M_X\)-orbit (note that \(\Omega(\alpha)=\mathcal O_{\gamma, \gamma'}\) and compare Lemma 2 and Proposition 3 with Lemma 3.4 in [loc. cit.]).