On the Eisenstein functoriality in cohomology for maximal parabolic subgroups (Q2674694)

The cohomology of an arithmetic group can be expressed either as the de Rham cohomology of the associated locally symmetric space, or as the relative Lie algebra cohomology of the space of automorphic forms with respect to the arithmetic group. The standard reference for these facts is the book by \textit{A. Borel} and \textit{N. Wallach} [Continuous cohomology, discrete subgroups, and representations of reductive groups. Providence, RI: American Mathematical Society (AMS) (2000; Zbl 0980.22015)]. The two descriptions of the same cohomology space give rise to two different approaches to its calculation: the topological approach and the approach by the theory of automorphic forms, which in turn can be used to exchange the ideas and results between geometry of the locally symmetric space and the arithmetic and analytic properties of automorphic forms. In the case of congruence arithmetic groups, this is best understood in the adèlic setting. Let \(\mathbb{A}\) be the ring of adèles of the field \(\mathbb{Q}\) of rational numbers, and let \(\mathbb{A}_f\) be the subring of finite adèles. Let \(G\) be a reductive group defined over \(\mathbb{Q}\). Then the adèlic variant of the locally symmetric space associated to a congruence subgroup \(\Gamma\) of \(G\) is the quotient \[ S_K = G(\mathbb{Q})\backslash G(\mathbb{A})\slash A_GK_\infty K, \] where \(A_G\) is the identity component of the group of real points of a maximal split torus in the center of \(G\), \(K_\infty\) is a maximal compact subgroup of \(G(\mathbb{R})\), and \(K\) is an open compact subgroup of \(G(\mathbb{A}_f)\). The cohomology \(H^\bullet (S_K,\mathbb{C})\) of the adèlic space \(S_K\) encapsulates the cohomology of the associated congruence arithmetic subgroups. From the analytic point of view, according to the work of \textit{A. Borel} [Duke Math. J. 50, 605--623 (1983; Zbl 0528.22010)] and \textit{J. Franke} [Ann. Sci. Éc. Norm. Supér. (4) 31, No. 2, 181--279 (1998; Zbl 0938.11026)], the cohomology \(H^\bullet(S_K,\mathbb{C})\) of \(S_K\) can be expressed as the relative Lie algebra cohomology of the space \(\mathcal{A}(G_K)\) of automorphic forms on \[ G_K = G(\mathbb{Q})\backslash G(\mathbb{A})\slash A_G K, \] that is, the space of automorphic forms on \(G(\mathbb{A})\) of level \(K\). The space of automorphic forms exhibits a direct sum decomposition along the cuspidal support, which was first observed by \textit{R. P. Langlands} [On the functional equations satisfied by Eisenstein series. Berlin-Heidelberg-New York: Springer-Verlag (1976; Zbl 0332.10018)], which gives rise to the corresponding decomposition in cohomology. The summand of \(\mathcal{A}(G_K)\) consisting of automorphic forms supported in the associate class of a parabolic \(\mathbb{Q}\)-subgroup \(P\) is denoted by \(\mathcal{A}_P(G_K)\), and the corresponding summand in the relative Lie algebra cohomology is \[ H^\bullet (\mathfrak{g},K_\infty; \mathcal{A}_P(G_K)), \] where \(\mathfrak{g}\) is the Lie algebra of \(G(\mathbb{R})\). The summand associated to the parabolic subgroup \(P=G\) is called the cuspidal cohomology, because \(\mathcal{A}_G(G_K)\) is the space of cuspidal automorphic forms on \(G(\mathbb{A})\) of level \(K\). According to the work of \textit{G. Harder} [in: Discrete Subgroups of Lie Groups Appl. Moduli, Pap. Bombay Colloq. 1973, 129--160 (1975; Zbl 0317.57022); Invent. Math. 89, 37--118 (1987; Zbl 0629.10023); Lect. Notes Math. 1447, 85--153 (1990; Zbl 0719.11034); in: Cycles, motives and Shimura varieties. Proceedings of the international colloquium, Mumbai, India, January 3--12, 2008. New Delhi: Narosa Publishing House/Published for the Tata Institute of Fundamental Research. Studies in Mathematics. Tata Institute of Fundamental Research 21, 131-190 (2010; Zbl 1294.11078)] and \textit{J. Schwermer} [Kohomologie arithmetisch definierter Gruppen und Eisensteinreihen. Berlin-Heidelberg-New York-Tokyo: Springer-Verlag (1983; Zbl 0506.22015); Invent. Math. 116, No. 1--3, 481--511 (1994; Zbl 0807.11031)], the remaining summands, i.e., those associated to \(P\neq G\), can be constructed, in principle, using the theory of Eisenstein series from the cuspidal cohomology classes (of appropriate level) in the cohomology of the Levi factor of \(P\). Hence, the sum of these summands is referred to as the Eisenstein cohomology. Although there is a general approach to the construction of Eisenstein classes, the actual construction of non-trivial Eisenstein classes depends on the subtle combination of geometric and arithmetic conditions, which are in many cases rather difficult to explicitly check. This is pointed out in a series of papers of \textit{J. Schwermer} and the reviewer [C. R., Math., Acad. Sci. Paris 348, No. 11--12, 597--600 (2010; Zbl 1203.11043); Int. Math. Res. Not. 2011, No. 7, 1654--1705 (2011; Zbl 1298.11050); Forum Math. 26, No. 6, 1635--1662 (2014; Zbl 1317.11055); Forum Math. 31, No. 5, 1225--1263 (2019; Zbl 1439.11133); Adv. Math. 376, Article ID 107438, 49 p. (2021; Zbl 1459.11128)], summarized in the survey paper [\textit{N. Grbac}, Springer Proc. Math. Stat. 245, 35--50 (2018; Zbl 1483.11102)]. From the topological point of view, the cohomology \(H^\bullet(S_K,\mathbb{C})\) of \(S_K\) can be approached through the Borel-Serre compactification. In the case of the cuspidal support in a maximal parabolic subgroup, \textit{P. Scholze} [Ann. Math. (2) 182, No. 3, 945--1066 (2015; Zbl 1345.14031)] proposed a topological method for construction of Eisenstein classes. The main purpose of the paper under review is to apply Scholze's method beyond the special cases of the Siegel maximal parabolic subgroup of the symplectic group and the even quasi-split unitary group associated to a CM quadratic extension, which are considered in [loc. cit.], and thus show that the method works in general. However, in the final section of the paper, the reasons why the method fails in the case of non-cuspidal maximal proper parabolic \(\mathbb{Q}\)-subgroup \(P\) are explained. The possibility for finding a topological method in that case is discussed and a related conjecture on certain property of the Borel-Serre compactification is formulated. The main result of the paper is the Scholze's construction in the following two cases. The first case is the case with a local Artin ring of coefficients, which is for technical reasons assumed to be Gorenstein, in particular, in characteristic \(\ell\), with respect to the local system arising from the trivial representation. The second case is the case of complex cohomology, but not limited to the local system arising from the trivial representation. In both cases the main result impies the non-vanishing of certain cohomology spaces for the full group \(G\) under the assumption that the so-called inner cohomology for the Levi factor of a maximal proper parabolic \(\mathbb{Q}\)-subgroup \(P\) is non-trivial. The comparison of the topological method of P. Scholze and the Eisenstein series method of G. Harder and J. Schwermer is also carried out. In particular, the example of the general linear group \(G=\mathrm{GL}(n)\), where \(n=2m\) is even, and its parabolic subgroup with the Levi factor isomorphic to \(\mathrm{GL}(m)\times \mathrm{GL}(m)\), is treated.