On the existence of \(({\mathfrak g},{\mathfrak k})\)-modules of finite type (Q1766226)

Let \(\mathfrak g\) be a reductive Lie algebra over an algebraically closed field of characteristic \(0\). If \(M\) is an irreducible \(\mathfrak g\)-module then the set of elements \(g\in \mathfrak g\) which act locally finitely on \(M\) is a subalgebra of \(\mathfrak g\), the Fernando-Kac subalgebra \(\mathfrak g[M]\) associated to \(M\). (An element \(g\) acts locally finitely on \(M\) if for each \(m\in M\), \(\langle g\rangle\cdot m\) is a finite dimensional subspace of \(M\); here \(\langle g\rangle\) denotes the subalgebra of the universal enveloping algebra \(\mathcal U(\mathfrak g)\) generated by \(g\).) Now let \(\mathfrak k\) be a subalgebra of \(\mathfrak g\) which is reductive in \(\mathfrak g\). A \((\mathfrak g,\mathfrak k)\)-module is a \(\mathfrak g\)-module \(M\) such that \(\mathfrak k\) acts locally finitely on \(M\). (If \(M\) is irreducible this says that \(\mathfrak k\subset\mathfrak g[M]\).) In particular, all irreducible \(\mathfrak k\)-submodules of \(M\) are finite dimensional. The reductive subalgebra \(\mathfrak k\) is called primal if there exists an irreducible \((\mathfrak g,\mathfrak k)\)-module \(M\) of finite type (i.e., with finite \(\mathfrak k\) multiplicities) which is not a \((\mathfrak g,\mathfrak k')\)-module for a strictly larger reductive in \(\mathfrak g\) subalgebra \(\mathfrak k'\). The main result of the paper is a complete description of all primal subalgebras of a reductive Lie algebra \(\mathfrak g\); it states that \(\mathfrak k\) is primal if and only if it contains its centralizer in \(\mathfrak g\). The heart of the proof consists of a geometric construction of a family of irreducible \((\mathfrak g,\mathfrak k)\)-modules of finite type over \(\mathfrak k\), a \(\mathcal D\)-module version of cohomologial induction. If \(\mathfrak k\) is a primal subalgebra and \(M\) an irreducible \((\mathfrak g,\mathfrak k)\)-module as in the definition of `primal', then \(\mathfrak k\) is maximal reductive in the Fernando-Kac subalgebra \(\mathfrak g[M]\); in general it is the reductive part \(\mathfrak g[M]=\mathfrak k\oplus\mathfrak n\), where \(\mathfrak n\) is a nilpotent ideal in \(\mathfrak g[M]\). For \(\mathfrak g=\mathfrak g\mathfrak l(n)\), the authors show that if \(\mathfrak k\) contains its centralizer then the module \(M\) can be chosen so that \(\mathfrak k\) is the full Fernando-Kac subalgebra associated to \(M\). Finally, the authors consider root subalgebras of \(\mathfrak g\mathfrak l(n)\) (subalgebras which contain a Cartan subalgebra); if \(\mathfrak k\) is a root subalgebra of \(\mathfrak g\mathfrak l(n)\), they describe all Fernando-Kac subalgebras containing \(\mathfrak k\).