On Kostant root systems of Lie superalgebras (Q827058)

\textit{B. Kostant} [Prog. Math. 278, 129--152 (2010; Zbl 1226.17008)] considered the following generalization of the root decomposition of \(\mathfrak{ g}\) a simple complex Lie algebra. For any \(\mathfrak{ t}\) a toral subalgebra of \(\mathfrak{ g}\) (not necessarily a Cartan subalgebra), consider the decomposition \(\mathfrak{g }=\mathfrak{c }_{\mathfrak{ g}}(\mathfrak{ t})\oplus(\oplus_{\nu\in R}\mathfrak{ g}_{\nu})\) where \(\mathfrak{ g}_\nu=\{x\in\mathfrak{ g}:[t,x]=\nu(t)x \textrm{ for every }t\in\mathfrak{ t}\}\) and \(R=\{\nu\in\mathfrak{ t}^*\setminus\{0\}:\mathfrak{ g}_\nu\ne0\}\). The elements in \(R\) (here called \(\mathfrak{ t}\)-roots or \emph{Kostant roots}) satisfy similar properties to the usual roots in case that \(\mathfrak{ t}\) equals the center of \(\mathfrak{c }_{\mathfrak{ g}}(\mathfrak{ t})\). Namely: \(\mathfrak{ g}_{\nu}\) is an irreducible \(\mathfrak{c }_{\mathfrak{ g}}(\mathfrak{ t})\)-module, \([\mathfrak{ g}_{\mu},\mathfrak{ g}_{\nu}]=\mathfrak{ g}_{\mu+\nu}\) whenever \(\mu,\nu,\mu+\nu\in R\) and \(R\) admits bases of simple roots. The aim of the paper under review is to extend these results to reductive Lie superalgebras. A finite-dimensional Lie superalgebra \(\mathfrak{ g}\) is called a \emph{reductive superalgebra} if \(\mathfrak{ g}=\mathfrak{ a}\oplus \mathfrak{ g}_1\oplus\dots\oplus\mathfrak{ g}_l\) for \(\mathfrak{ a}\) an abelian Lie algebra and each \(\mathfrak{ g}_i\) either a simple Lie algebra or a superalgebra isomorphic to one in the following list: \(\mathfrak{ sl}(m|n)\) with \(m\ne n\), \(\mathfrak{ gl}(m|m)\) with \(m\ge 1\), \(\mathfrak{ osp}(m|2n)\), \(D(2,1;\alpha)\), \(G(3)\) and \(F(4)\) (all of them simple except for \(\mathfrak{ gl}(m|m)\), which is considered instead of \(\mathfrak{ psl}(m|m)\) by some technical details). If \(\mathfrak{ g}=\mathfrak{ g}_{\bar0}\oplus\mathfrak{ g}_{\bar1}\) is a reductive superalgebra, \(\mathfrak{ t}\subset \mathfrak{ g}_{\bar0}\) is said a \emph{toral subalgebra} if it is abelian and acts semisimply on \(\mathfrak{ g}\). It is proved that the centralizer \(\mathfrak{m }=\mathfrak{c }_{\mathfrak{ g}}(\mathfrak{ t})=\{x\in \mathfrak{ g}:[t,x]=0 \textrm{ for every }t\in\mathfrak{ t}\}\) is a reductive Lie superalgebra too and, under the assumption that \(\mathfrak{t }\) equals the center of \(\mathfrak{m }\), it is also proved that \begin{itemize} \item \(\mathfrak{g}_\nu\) is an irreducible \(\mathfrak{m }\)-module for any \(\nu\in R\); \item \([\mathfrak{ g}_{\mu},\mathfrak{ g}_{\nu}]=\mathfrak{ g}_{\mu+\nu}\) whenever \(\mu,\nu,\mu+\nu\in R\); \item if \(\mu,\nu, \mu+k\nu\in R\) for some positive integer \(k\), then \(\mu+j\nu\in R\) for all \(0\le j\le k\); \item \(R\) admits a base, that is, a linearly independent subset \(\Sigma\subset R\) such that any element in \(R\) is an integral linear combination of elements in \(\Sigma\) with either all coefficients not negative or all coefficients not positive.(Moreover, they describe every base of \(R\).) \end{itemize} To summarize, Kostant root systems also inherit the main properties of the classical root systems in the case of reductive Lie superalgebras. The approach is new: the main technique used by Kostant was the Killing form, while here the authors use graphs (which behave nicely with respect to the projection \(\pi\colon \mathfrak{h }^*\to\mathfrak{t }^*\) for \(\mathfrak{h}\) a Cartan subalgebra containing \(\mathfrak{t }\)). As application, Hermitian symmetric pairs of Lie superalgebras are studied.