Representations of the Poisson algebra in prime characteristic (Q1566363)

Let \(k\) be an algebraically closed field of prime characteristic \(p\). There are two main classes of simple Lie algebras over \(k\), namely, the Lie algebras of classical type which are analogues of the simple Lie algebras in characteristic zero and the Lie algebras of Cartan type which are finite-dimensional analogues of infinite-dimensional Lie algebras arising in differential geometry. In the last decade the representation theory of Lie algebras of classical type has been studied by several people and a lot progress has been made but at the moment there is no similar general approach for studying the representations of Lie algebras of Cartan type. The Poisson algebra \(B_{2n}=k[x_1,\dots,x_{2n}]\), \(x_i^p=0\) for \(1\leq i\leq 2n\), is the associative algebra of truncated polynomials in \(2n\) commuting variables over \(k\) with a Poisson bracket defined by \[ [f,g]=\sum_{j=1}^n(\partial_j(f)\partial_{n+j}(g)- \partial_{n+j}(f)\partial_j(g)), \] where \(\partial_i\) denotes the partial derivative with respect to \(x_i\) for \(1\leq i\leq 2n\). The center of the Lie algebra \((B_{2n},[\cdot,\cdot])\) coincides with the scalars \(k\cdot 1\) and the factor algebra \([B_{2n},B_{2n}]/k\cdot 1\) is isomorphic to a Hamiltonian Lie algebra \(H_{2n}\) which form one family of the simple Lie algebras of Cartan type. The Hamiltonian Lie algebra \(H_{2n}\) as well as \(B_{2n}\) are restricted Lie algebras, i.e., both have a \([p]\)-mapping compatible with their Lie algebra structures. Note that the \([p]\)-mapping of \(H_{2n}\) is unique whereas the \([p]\)-mapping of \(B_{2n}\) is only unique modulo its center. The aim of the paper under review is to develop the representation theory of \(B_{2n}\) similarly to the representation theory of Lie algebras of reductive groups. One parallel between \(B_{2n}\) and the latter is the existence of a non-degenerate symmetric invariant bilinear form which allows the identification of linear forms on the Lie algebra with elements in the Lie algebra. The first section of the paper is of a more general nature and elaborates on a Morita equivalence theorem proved by \textit{A. Premet} [Adv. Math. 170, No. 1, 1--55 (2002; Zbl 1005.17007)]. Let \(\mathfrak{g}\) be an arbitrary restricted Lie algebra over \(k\), let \(\xi\) be a linear form on \(\mathfrak{g}\), and let \(U_{\xi}(\mathfrak{g})\) denote the factor algebra of the universal enveloping algebra of \(\mathfrak{g}\) modulo the two-sided ideal generated by \(\{x^p-x^{[p]}-\xi(x)^p\mid x\in\mathfrak{g}\}\). A restricted subalgebra \(\mathfrak{n}\) of \(\mathfrak{g}\) is called \(\xi\)-admissible if every element of \(\mathfrak{n}\) is \([p]\)-nilpotent, \(\xi\) vanishes on the restricted subalgebra generated by \([\mathfrak{n},\mathfrak{n}]\), and every finite-dimensional \(U_{\xi}(\mathfrak{g})\)-module is \(U_{\xi}(\mathfrak{n})\)-free. It follows that \(U_{\xi}(\mathfrak{n})\) is local with the (up to isomorphism) unique simple module \(k_{\xi}\). Premet proves in the paper cited above that the induced module \(Q_{\mathfrak{n}}:=U_{\xi} (\mathfrak{g})\otimes_{U_{\xi}(\mathfrak{n})}k_{\xi}\) is a projective generator in the category of finite-dimensional \(U_{\xi}(\mathfrak{g})\)-modules and \(U_{\xi}(\mathfrak{g}) \cong\text{Mat}_{p^d}(H_{\mathfrak{n}})\), where \(d:=\dim_k\mathfrak{n}\) and \(H_{\mathfrak{n}}:=\text{End}_{\mathfrak{g}}(Q_{\mathfrak{n}})^{\text{op}}\). In particular, \(U_{\xi}(\mathfrak{g})\) and \(H_{\mathfrak{n}}\) are Morita equivalent. In general, it is difficult to describe the algebra \(H_{\mathfrak{n}}\). Let \(\text{Nor}_{\mathfrak{g}}(\mathfrak{n})\) denote the normalizer of \(\mathfrak{n}\) in \(\mathfrak{g}\). If \(\dim_k\text{Nor}_{\mathfrak{g}}(\mathfrak{n})\geq\dim_k \mathfrak{g}/\mathfrak{n}\) and \(\xi_{|\mathfrak{n}}=0\), then Premet shows that \(H_{\mathfrak{n}}\cong U_{\xi}(\text{Nor}_{\mathfrak{g}}(\mathfrak{n})/ \mathfrak{n})\). This generalizes a result of \textit{E. M. Friedlander} and \textit{B. J. Parshall} [Am. J. Math. 110, No. 6, 1055--1093 (1988; Zbl 0673.17010)] for Lie algebras of reductive groups. The author generalizes Premet's results further by considering restricted subalgebras \(\mathfrak{n}\subseteq\tilde{\mathfrak{n}} \subseteq\mathfrak{p}\) of \(\mathfrak{g}\) where \(\mathfrak{n}\) is \(\xi\)-admissible with \(\dim_k\mathfrak{p}=\dim_k\mathfrak{g}/\mathfrak{n}\) and \(\tilde{\mathfrak{n}}\) is an ideal of \(\mathfrak{p}\) satisfying certain conditions. Then the induction functor \(U_{\xi}(\mathfrak{g})\otimes_{U_{\xi}(\mathfrak{p})}-\) yields an equivalence between the category of finite-dimensional \(U_{\xi}(\mathfrak{p})\)-modules annihilated by the Jacobson radical of \(U_{\xi}(\tilde{\mathfrak{n}})\) and a certain full subcategory of the category of finite-dimensional \(U_{\xi}(\mathfrak{g})\)-modules which contains all simple modules. If \(\mathfrak{s}\) is a restricted subalgebra of \(\mathfrak{p}\) such that \(\mathfrak{p}=\mathfrak{s}\oplus\tilde{\mathfrak{n}}\), then there is a bijection between the isomorphism classes of simple \(U_{\xi}(\mathfrak{g})\)-modules and the isomorphism classes of simple \(U_{\xi}(\mathfrak{p}/\tilde{\mathfrak{n}})\)-modules. If \(\mathfrak{n} =\tilde{\mathfrak{n}}\) and \(\mathfrak{p}=\mathfrak{s}\oplus\mathfrak{n}\), then \(U_{\xi}(\mathfrak{g})\) and \(U_{\xi}(\mathfrak{p}/\mathfrak{n})\) are Morita equivalent. The author also gives sufficient conditions for the \(\xi\)-admissibility of a restricted subalgebra \(\mathfrak{n}\) of \(\mathfrak{g}\) without using support varieties. In the remaining sections the paper is only concerned with the representation theory of the Poisson algebra \(B_{2n}\). In the second section the author introduces some notation and discusses various technical tools. He fixes the \([p]\)-mapping of \(B_{2n}\) and makes some observations on the restricted nullcone of \(B_{2n}\). It is well-known that \([B_{2n},B_{2n}]\) is a subspace of codimension one in \(B_{2n}\). Let \(\varphi\) denote the linear form on \(B_{2n}\) with kernel \([B_{2n},B_{2n}]\). Then \(\varphi\) turns \(B_{2n}\) into a symmetric Frobenius algebra. Moreover, the associated non-degenerate symmetric bilinear form \((f,g)\mapsto\varphi(fg)\) is invariant with respect to the Lie algebra structure on \(B_{2n}\) and allows the identification of an element \(f\) in \(B_{2n}\) with a linear form \(\xi_f\) on \(B_{2n}\) via \(\xi_f(g)= \varphi(fg)\) for every \(g\in B_{2n}\). The author also introduces so-called good systems of generators for ideals of the associative algebra \(B_{2n}\) and associated restricted filtrations of the Lie algebra \(B_{2n}\) which will be an important tool for constructing polarizations in the last section. The purpose of the third section is to establish a Morita equivalence between \(U_{\xi}(B_{2n})\) and \(U_{\xi}(\mathfrak{s})\) for the centralizer \(\mathfrak{s}\) of a \([p]\)-semisimple element in \(B_{2n}\) and certain \(p\)-characters \(\xi\). More precisely, let \(\mathfrak{s}\) be the centralizer of an element \(s\) in the torus generated by the \([p]\)-semisimple part of \(f\in B_{2n}\). If \(\xi=\xi_f\) and \(p\geq 5\) (or \(s\) satisfies some technical condition for \(p=2\) or \(p=3\)), then \(U_{\xi}(B_{2n})\) and \(U_{\xi}(\mathfrak{s})\) are Morita equivalent. In the fourth section the author defines good associative subalgebras and good elements of the Poisson algebra. Let \(\mathfrak{m}=\{f\in B_{2n}\mid f^p=0\}\) denote the unique maximal ideal of the associative algebra \(B_{2n}\). An associative subalgebra \(B\) of \(B_{2n}\) is called good if \(B\) can be generated by elements in \(\mathfrak{m}\) which are linearly independent modulo \(\mathfrak{m}^2\). An element \(f\in B_{2n}\) is called good if the associative subalgebra generated by the \([p]\)-powers \(\{f^{[p]^i}\mid i\geq 0\}\) is good. Let \(B\) be a subspace of \(B_{2n}\) which is closed under the associative multiplication, the Poisson bracket, and the \([p]\)-mapping of \(B_{2n}\), let \(\mathfrak{z}(B)\) denote the Lie centralizer of \(B\) in \(B_{2n}\), and let \(m\) denote the dimension of \(B\cap\mathfrak{m}/B\cap \mathfrak{m}^2\). Then \(\dim_k\mathfrak{z}(B)\leq p^{2n-m}\) and \(B\) is good if and only if \(\dim_k\mathfrak{z}(B)=p^{2n-m}\). As a consequence, an element \(f\in B_{2n}\) is good if and only if \(\dim_k\mathfrak{z}(f)=p^{2n-m}\) where \(m\) is the smallest integer such that \(\{f^{[p]^i}\mid 0\leq i\leq m\}\) is linearly dependent modulo \(k+\mathfrak{m}^2\). In the fifth section the author classifies the simple \(U_{\xi}(B_{2n})\)-modules for linear forms \(\xi\) corresponding to certain good elements. Let \(\xi\) be a linear form on \(B_{2n}\). Then a Lie subalgebra \(\mathfrak{p}\) of \(B_{2n}\) is called polarization of \(\xi\) if \(\mathfrak{p}\) is a maximal totally isotropic subspace with respect to the bilinear form \(\beta\) given by \(\beta(f,g):=\xi([f,g])\) for any \(f,g\in B_{2n}\). The main result of this section reads as follows. Let \(f\in B_{2n}\) be a good element. If \(\xi=\xi_f\) and \(p\geq 5\) (for \(p=2\) assume in addition that \(\{f,f^{[p]},f^{[p]^2}\}\) is linearly independent modulo \(k+\mathfrak{m}^2\) and for \(p=3\) assume in addition that \(\{f,f^{[p]}\}\) is linearly independent modulo \(k+\mathfrak{m}^2\)), then there exists a polarization \(\mathfrak{p}\) of \(\xi\) such that the induction functor \(U_{\xi}(B_{2n}) \otimes_{U_{\xi}(\mathfrak{p})}-\) yields a bijection between the isomorphism classes of simple \(U_{\xi}(\mathfrak{p})\)-modules and the isomorphism classes of simple \(U_{\xi}(B_{2n})\)-modules. Moreover, the restriction functor yields a bijection between the isomorphism classes of simple \(U_{\xi}(\mathfrak{p})\)-modules and the isomorphism classes of simple \(U_{\xi} (\mathfrak{z}(f))\)-modules where \(\mathfrak{z}(f)\) denotes the Lie centralizer of \(f\) in \(B_{2n}\). It was shown previously by the author in [J. Algebra 256, No. 1, 146--179 (2002; Zbl 1039.17024)] that \(U:=\{f\in B_{2n}\mid f,f^{[p]},\dots,f^{[p]^{n-1}}\) are linearly independent modulo \(k+\mathfrak{m}^2\}\) is a non-empty Zariski open subset of \(B_{2n}\) and \(\dim_k\mathfrak{z}(f)=p^n\) for every \(f\in U\). Note that every \(f\in U\) is good. As an application of the main result of this section the author proves the following. If \(\xi=\xi_f\) with \(f\in U\) and \(p\geq 5\) (for \(p=2\) assume that \(n\geq 3\) and for \(p=3\) assume that \(n\geq 2\)), then there exists a solvable polarization \(\mathfrak{p}\) of \(\xi\) such that every simple \(U_{\xi}(B_{2n})\)-module is induced from a one-dimensional \(U_{\xi}(\mathfrak{p})\)-module. Consequently, all simple \(U_{\xi}(B_{2n})\)-modules have the same dimension \(p^{\frac{1}{2}(p^{2n}-p^n)}\) and there are exactly \(p^{\dim_k \mathfrak{t}}\) isomorphism classes of simple \(U_{\xi}(B_{2n})\)-modules where \(\mathfrak{t}\) denotes the torus generated by the \([p]\)-semisimple part of \(f\). In particular, this verifies the analogue of the Kac-Weisfeiler conjecture on the maximum dimension of the simple modules in the case \(B_{2n}\).