Quaternion distinguished generic representations of \(\mathrm{GL}_{2n}\) (Q2200061)

Let \(F\) be a non-Archimedean local field of characteristic \(0\) and \(E\) be a quadratic extension of \(F\). Fixing an embedding of \(E\) in the unique quaternion division algebra \(D\) over \(F\), \(\mathrm{GL}_{m}(D)\) can be identified as a subgroup of \(\mathrm{GL}_{2m}(E)\). The paper under review classified irreducible generic \(\mathrm{GL}_{2m}(E)\)-representations \(\pi\) distinguished by \(\mathrm{GL}_{m}(D)\), i.e., \(\Hom_{\mathrm{GL}_{m}(D)}(\pi,\mathbb C)\neq 0\) (The space has at most one dimension by Gelfand-Kazhdan criterion, see Proposition~3.4). We first recall Bernstein-Zelevinsky's classification theory. A segment \(\Delta\) is a set of the form \[ \left\{\rho{\left|\det\right|}^{\frac{-l+1}{2}},\rho{\left|\det\right|}^{\frac{-l+3}{2}},\cdots ,\rho{\left|\det\right|}^{\frac{l-1}{2}}\right\} \] where \(\rho\) is an irreducible supercuspidal representation of \(\mathrm{GL}_{n}(E)\) and \(l\) is a positive integer. There is a unique irreducible sub-representation \(\mathrm{St}_{l}(\rho)\) (call a Steinberg representation) in the (normalized) parabolically induced module \[ \rho{|\det|}^{\frac{-l+1}{2}}\times \rho{|\det|}^{\frac{-l+3}{2}} \times \cdots \times \rho{|\det|}^{\frac{l-1}{2}}. \] The \(\mathrm{GL}_{nl}(E)\)-module \(\mathrm{St}_{l}(\rho)\) is essentially square-integrable. Moreover, every irreducible essentially square-integrable module of general linear groups can be realized uniquely as a Steinberg representation. Two segments \(\Delta_{1}\) and \(\Delta_{2}\) are called unlinked if \(\Delta_1\cup \Delta_2\) is not a segment. Irreducible generic representations of the general linear groups can be parameterized by the set of pairwise unlinked segments. More precisely, each collection \(\left\{\Delta_{1}, \Delta_{2},\cdots, \Delta_{r}\right\}\) of pairwise unlinked segments gives rise to a generic representation: the irreducible parabolically induced module \[ \pi:=\mathrm{St}_{l_{1}}(\rho_{1})\times \mathrm{St}_{l_{2}}(\rho_{2})\times\cdots \times \mathrm{St}_{l_{r}}(\rho_{r}). \] (See ``Zelevinsky, Induced representations of reductive \({p}\)-adic groups. II. On irreducible representations of \(\mathrm{GL}(n)\), Theorem~9.7''). Therefore the classification problem can be reduced to the classification of distinguished Steinberg representations via Mackey theory following the method of Matringe, ``Distinguished generic representations of \(\mathrm{GL}(n)\) over \(p\)-adic fields''. This is the content of Theorem~5.2: The \(\mathrm{GL}_{2m}(E)\)-module \(\pi\) defined above is \(\mathrm{GL}_{m}(D)\)-distinguished if and only if there are an integer \(1\leq k\leq r/2\) and a reordering of \(\Delta_{i}\) such that \begin{itemize} \item \(\mathrm{St}_{l_{2i}}(\rho_{2i})\) is isomorphic to the conjugate dual of \(\mathrm{St}_{l_{2i-1}}(\rho_{2i-1})\) for \(1\leq i \leq k\); \item \(\mathrm{St}_{l_{i}}(\rho_i)\) is a \(\mathrm{GL}_{m_{i}}(D)\)-distinguished representation of \(\mathrm{GL}_{2m_{i}}(E)\) for \(2k+1\leq i \leq r\). \end{itemize} According to Beuzart-Plessis's result ``On distinguished square-integrable representations for Galois pairs and a conjecture of Prasad'' Theorem~1, an irreducible essentially square-integrable representation \(\Delta\) of \(\mathrm{GL}_{2m}(E)\) is \(\mathrm{GL}_{m}(D)\)-distinguished if and only if it is \(\mathrm{GL}_{2m}(F)\)-distinguished. Using Martinge's result ``Conjectures about distinction and local Asai \(L\)-functions'' Theorem~3.7, the author showed that the \(\mathrm{GL}_{2m}(F)\)-distinguishedness of \(\Delta\) is equivalent to that \(\Delta\) arises from the unstable base change lift from \(\mathrm{U}_{2m}\), see Lemma~6.1. At the end of the paper, the author showed that every irreducible generic \(\mathrm{GL}_{2m}(E)\)-representation arising from the unstable base change lifting is \(\mathrm{GL}_{m}(D)\)-distinguished. It is worth mentioning that a \(\mathrm{GL}_{m}(D)\)-distinguished generic representation is not necessarily coming from the unstable base change, see Remark~6.2.