Generic and \(q\)-rational representation theory (Q1974442)

This paper is mainly devoted to the study of generic aspects of representation theory and cohomology of general linear groups in non-describing characteristic. To this goal, the authors develop in the first part of the paper a general theory of orders in non-semisimple algebras, and formalize the notion of a generic property in representation theory. Let \(\mathcal O\) be a domain (commutative and Noetherian) with field of fractions \(K\) and let \(\mathfrak p\) be a prime ideal of \(\mathcal O\). Denote by \({\mathcal O}_{\mathfrak p}\) the localization of \(\mathcal O\) at \(\mathfrak p\), and by \(k({\mathfrak p})\) the corresponding residue field \({\mathcal O}_{\mathfrak p}/{\mathfrak p}{\mathcal O}_{\mathfrak p}\). If \(M\) is an \(\mathcal O\)-module, and \(R\) a commutative \(\mathcal O\)-algebra, then let \(M_R:=M\otimes_{\mathcal O}R\). Consider a property \(P\) of finite-dimensional algebras over fields. Then \(P\) is called generic if whenever \(A\) is an \(\mathcal O\)-finite and \(\mathcal O\)-torsion-free \(\mathcal O\)-algebra, then \(A_K\) satisfies \(P\) implies \(A_{k({\mathfrak p})}\) satisfies \(P\) for all prime ideals \(\mathfrak p\) in some nonempty open subset of \(\text{Spec }{\mathcal O}\). This concept is then extended to properties of pairs of algebras, and the notion of integrally generic property is also defined. In the first part of the paper the authors prove the genericity of several properties, among them being the following: the property of an algebra of being split semisimple; the property of an algebra of having finite global dimension; the property of an algebra of being split quasi-hereditary; the property of a pair of algebras with the same ground field of being Morita equivalent; the property of a pair of algebras with the same ground field of being derived equivalent. In the second part of the paper, the authors use the general results to make new contributions to the representation theory of the finite general linear groups in the non-defining characteristic case. Let \(n\) be a positive integer, and let \(G(q)=\text{GL}_n(q)\) be the general linear group over the finite field with \(q\) elements. The authors investigate the \(p\)-modular representation theory of \(G(q)\), as \(p\) and \(q\) vary subject to the condition that \(p\) does not divide \(q\). They rely here on the parametrization of the irreducible modules due to the work of R.~Dipper and G.~D.~James, and on the distribution of the characters of \(G(q)\) into \(p\)-blocks, which is known by the work of P.~Fong and B.~Srinivasan. One of the main results is a new proof of the ``Geck-Gruber-Hiss very large prime result'' which essentially says that the \(p\)-decomposition numbers of the \(q\)-Schur algebra \(S_q(n)\) generically depend only on the multiplicative order of \(q\) modulo \(p\). Another main theorem establishes a Morita equivalence \(F\) between the quotient \({\mathcal O}G(q)/J(q)\) of the group algebra \({\mathcal O}G(q)\) and a direct sum of \(q^*\)-Schur algebras over \(\mathcal O\), where \(\mathcal O\) is a discrete valuation ring with residue field \(k\) of characteristic \(p\) not dividing \(q\), \(J(q)\) is the annihilator of a certain naturally defined \({\mathcal O}G(q)\)-module. This in turn induces a Morita equivalence \(\overline F\) between \(kG(q)/J(q)_k\) and a direct sum of \(q^*\)-Schur algebras over \(k\). The Morita equivalence \(\overline F\) is used to establish a cohomological connection between simple \(kG(q)\)-modules and modules for the \(q\)-Schur algebra. The stability results on the dimensions of \(\text{Ext}\)-groups obtained in the first part imply stability results for the \(H^1\)-groups of unipotent \(kG(q)\)-modules. The results on \(H^1\) are then extended to higher cohomology groups. For this, the authors construct a finite projective resolution of the quantum determinant representation \(\det_q\) of \(S_q(n)_k\) in the module category of \(S_q(n)_{\mathcal O}\). The equivalence \(F\) transports this resolution into a resolution of the trivial \({\mathcal O}G(q)\)-module \(\mathcal O\), which turns out to be a projective resolution if the order of \(q\) modulo \(p\) is large enough. When \(p\) divides neither \(n\) nor \(q-1\), the results on the cohomology of the general linear group carry over to the special linear group.