Classification of restricted simple Lie algebras (Q1102366)

This paper gives the proof that every finite-dimensional restricted simple Lie algebra \(L\) over an algebraically closed field of characteristic \(p>7\) is of either classical or Cartan type. This result was conjectured by \textit{A. I. Kostrikin} and \textit{I. R. Shafarevich} [Dokl. Akad. Nauk SSSR 168, 740--742 (1966; Zbl 0158.03803)] and previously announced by the authors [Proc. Natl. Acad. Sci. USA 81, 5271--5274 (1984; Zbl 0542.17003)]. The paper contains extensive background material concerning algebras of Cartan type and the history of the classification problem. By the Recognition Theorem [\textit{V. G. Kac}, Izv. Akad. Nauk SSSR, Ser. Mat. 38, 800--834 (1974; Zbl 0317.17002); \textit{R. L. Wilson}, J. Algebra 40, 418--465 (1976; Zbl 0355.17012)] this classification follows if \(L\) can be shown to contain a maximal subalgebra \(L_ 0\) such that when \(L_{- 1}/L_ 0\) is an irreducible \(L_ 0\)-submodule of \(L/L_ 0\), \(L_{i- 1}=[L_ i,L_{-1}]+L_ i\) for \(i<0\), and \(L_{i+1}=\{x\in L_ i|\) \([x,L_{-1}]\subseteq L_ i\}\) for \(i\geq 0\), then for the associated graded algebra \(G=\sum G_ i\) with \(G_ i=L_ i/L_{i+1}\), the following hold: \(G_ 0\) is a restricted Lie algebra and a direct sum of restricted ideals each of which is classical simple, \(\mathfrak{gl}(n)\), \(\mathfrak{sl}(n)\), or \(\mathfrak{pgl}(n)\) with \(p| n\), or abelian; the action of \(G_ 0\) on \(G_{-1}\) is restricted; and for \(i\leq 0\) and \(x\in G_ i\), \([x,G_ 1]=0\) only if \(x=0\). The proof is achieved by demonstrating the existence of such a subalgebra \(L_ 0.\) The description of \(L_ 0\) is based on an appropriate root space decomposition of \(L\). For a maximal torus \(T\) in a restricted Lie algebra \(A\) and root space decomposition \(A={\mathfrak z}_ A(T)+\sum_{\gamma}A_{\gamma}\), let \(A^{(\alpha)}=\sum_{i\in {\mathbb Z}}A_{i\alpha}\) and \(A^{(\alpha,\beta)}=\sum_{i,j\in {\mathbb Z}}A_{i\alpha +j\beta}\). Then \(A^{(\alpha)}\) and \(A^{(\alpha,\beta)}\) modulo their respective solvable radicals are rank-one and rank-two sections of \(A\), respectively. The authors succeed in showing that a torus \(T\) of maximal dimension exists with respect to which all roots are proper, \(\gamma\) being called proper if \(\gamma ([A_{i\gamma},A_{-i\gamma}])=0\) for some \(i\in {\mathbb Z}_ p-\{0\}\). The procedure for replacing another torus by one with respect to which all roots are proper can readily be carried out for \(L\) if it can be accomplished in the rank-two sections of \(L\). To achieve this step a large portion of this very long paper is devoted to determining all possible rank-two sections of \(L\) and the necessary preliminary results on toral rank-one algebras. For every root \(\alpha\) with respect to such a torus \(T\) \(L^{(\alpha)}\) has a unique subalgebra \(U^{(\alpha)}\) of maximal dimension with all composition factors classical simple or abelian. Let \(Q\) be the sum of all such subalgebras \(U^{(\alpha)}\). If \(L\) is not classical, then for \(L_ 0\) a maximal subalgebra of \(L\) containing \(Q\), the hypotheses of the Recognition Theorem hold, and the classification follows.