Weyl-Kac type character formula for admissible representations of Borcherds-Kac-Moody Lie superalgebras (Q2182433)

The Weyl-Kac character formula for a Kac-Moody Lie algebra \(\mathfrak{g}\) expresses the character of integrable irreducible highest-weight \(\mathfrak{g}\)-modules in terms of the easily-calculated characters of a set of Verma modules indexed by the Weyl group of \(\mathfrak{g}\). This formula has been generalized to wider classes of Lie (super)algebras and modules. On the one hand, \textit{R. Borcherds} [J. Algebra 115, No. 2, 501--512 (1988; Zbl 0644.17010)] proved a Weyl-Kac type character formula for integrable modules of Borcherds-Kac-Moody Lie algebras, which unlike Kac-Moody algebras may have imaginary simple roots. On the other hand, \textit{V. Kac} and \textit{M. Wakimoto} [Proc. Natl. Acad. Sci. USA 85, No. 14, 4956--4960 (1988; Zbl 0652.17010)] generalized the Weyl-Kac character formula to a class of non-integrable representations for Kac-Moody Lie algebras that they called admissible. In this paper, the author proves a Weyl-Kac type character formula for certain admissible representations of Borcherds-Kac-Moody Lie superalgebras. To describe this result in more detail, let \(\mathfrak{g}\) be a symmetrizable Borcherds-Kac-Moody Lie superalgebra with Cartan subalgebra \(\mathfrak{h}\) and bilinear form \((\cdot\,\vert\,\cdot)\) on \(\mathfrak{h}^*\) induced from the restriction to \(\mathfrak{h}\) of the natural bilinear form on \(\mathfrak{g}\). The author defines a modified shifted Tits cone \(K^g\) which is characterized as the set of weights \(\lambda\) in the \(\mathbb{R}\)-span of the simple roots such that: \begin{itemize} \item \((\lambda+\rho\,\vert\,\alpha)<0\) for only finitely many positive roots \(\alpha\) where \(\rho\) is a Weyl vector in \(\mathfrak{h}^*\). \item \((\lambda+\rho\,\vert\,\alpha)\neq 0\) for all positive roots \(\alpha\) such that \((\alpha\,\vert\,\alpha)=0\). \end{itemize} Since Borcherds-Kac-Moody algebras can have imaginary simple roots, the second condition implies that the author's shifted Tits cone may not contain all \(-\rho\)-shifts of dominant integral weights. Also, for any \(\lambda\in\mathfrak{h}^*\), the author defines \(W^{\lambda+\rho}\) to be the Weyl group of the root subsystem consisting of all roots \(\alpha\) of \(\mathfrak{g}\) such that \((\lambda+\rho)(\alpha^\vee)\in\mathbb{Z}\), where \(\alpha^\vee\) is the coroot associated to \(\alpha\). The author notes that \(W^{\lambda+\rho}\) indeed depends on \(\lambda+\rho\) rather than \(\lambda\) alone because \(\rho(\alpha^\vee)\) is not always an integer for general Borcherds-Kac-Moody superalgebras. Now the main result of this paper applies to \textit{admissible} weights \(\lambda\), which satisfy \begin{itemize} \item \((\lambda+\rho)(\alpha^\vee)\notin\mathbb{Z}_{\leq 0}\) for all positive real coroots \(\alpha^\vee\). \end{itemize} The author proves that if \(L(\lambda)\) is the irreducible highest-weight \(\mathfrak{g}\)-module with admissible highest weight \(\lambda\in K^g\), then \[ \mathrm{ch}\,L(\lambda)=\sum_{w\in W^{\lambda+\rho}} (-1)^{l(w)}\mathrm{ch}\,M(w\cdot\lambda) \] where \(l(w)\) is the length function on the Coxeter group \(W^{\lambda+\rho}\) and \(M(w\cdot\lambda)\) is the Verma module with highest weight \(w\cdot\lambda=w(\lambda+\rho)-\rho\). The proof uses the translation functors introduced for infinite-dimensional algebras in [\textit{V. Deodhar}, \textit{O. Gabber} and \textit{V. Kac}, Adv. Math. 45, 92--116 (1982; Zbl 0491.17008)].