On central extensions and simply laced Lie algebras (Q2215796)

The main aim of this paper is to construct a \(\mathbb Z_d\)-graded semisimple Lie algebra attached to each \textit{input datum}. The author was lead to this purely algebraic problem when studying some families of algebraic curves, which provide naturally such an input datum. Note that this work is part of a program trying to use representation theory to extend results in arithmetic statistics, although precisely in this occasion the work is more focused on the algebraic part than on the number-theoretic applications, which are relegated. For \(k\) a field of characteristic zero containing a primitive \(d\)th root of the unit \(\zeta\), the \textit{input datum} is a quadruple \((\Lambda, w,\mathcal H, \varepsilon)\) with \begin{itemize} \item A \textit{simply laced} root lattice \(\Lambda\), that is, a finitely generated free \(\mathbb Z\)-module with a symmetric positive-definite bilinear form \((\cdot,\cdot)\colon\Lambda\times \Lambda\to\mathbb Z\) generated by \(\{\alpha\in\Lambda:(\alpha,\alpha)=2\}\); \item \(w\colon \Lambda\to\Lambda\) a lattice automorphism of order \(d\) which is \textit{elliptic}, that is, the group of \textit{coinvariants} \(\Lambda_w:=\Lambda/(1-w)\Lambda\) is finite; \item A central extension \(1\to\langle\zeta\rangle\to\mathcal H\to \Lambda_w\to 1\) with \(\langle\cdot,\cdot\rangle\colon\Lambda_w\times \Lambda_w\to \langle\zeta\rangle\) its commutator pairing; \item A \textit{bilinear} map \(\varepsilon\colon \Lambda\times \Lambda\to k^\times\) (that is, \(\langle \alpha,\beta+\gamma \rangle=\langle \alpha,\beta \rangle\langle \alpha,\gamma \rangle\) and similarly for the other variable), satisfying \(\varepsilon(w\alpha, w\beta)=\varepsilon(\alpha, \beta)\) and \(\varepsilon(\alpha, \beta)/\varepsilon( \beta,\alpha)=-\langle \beta,\alpha \rangle\) whenever \((\alpha, \beta)=-1\). \end{itemize} For each quadruple as above, the author constructs a Lie algebra with Dynkin type given by \(\Lambda\), endowed with a grading produced by an order \(d\) automorphism which is a lifting of \(w\). The author claims that the input data are easy to find: first, if one has \((\Lambda, w)\), then it is always possible to give a group \(\mathcal H\) and a bilinear map \(\varepsilon\) such that \((\Lambda, w,\mathcal H, \varepsilon)\) is a suitable quadruple as required above. Also, and this is important for the applications, input data arise naturally from some algebraic curves. For instance, an interesting collection of examples coming from a simple singularity of type \(E_8\) is described when \(k\) is algebraically closed. From a semiuniversal deformation of the surface \(X^5+Y^3+Z^2=0\), the Picard group of each smooth fiber is a root lattice of type \(E_8\), and this lattice has elliptic automorphisms of orders \(2\), \(3\) and \(5\). Note that, for some usual choice of \(\varepsilon\), representation theory of \(\mathcal H\) is closely related with representation theory of the Lie subalgebra fixed by the automorphism. Finally, new proofs of some well-known results on lifts of automorphisms on simply laced Lie algebras are given as a consequence of the construction here developed.