A new approach to Kostant's problem (Q969273)

Let \(\mathfrak g\) be a complex semisimple finite-dimensional Lie algebra with a fixed triangular decomposition and \(U(\mathfrak g)\) be its universal enveloping algebra. Then for every two \(\mathfrak g\)-modules \(M\) and \(N\) the space \(\text{Hom}_{\mathbb C}(M,N)\) is a bimodule and a \(\mathfrak g\)-module under the adjoint action of \(\mathfrak g\). The bimodule \(\text{Hom}_{\mathbb C}(M,N)\) has a sub-bimodule denoted by \(\mathcal{L}(M,N)\), which consists of all elements on which the adjoint action of \(U(\mathfrak g)\) is locally finite. Since \(U(\mathfrak g)\) itself consists of locally finite elements under the adjoint action, it naturally maps to \({\mathcal L}(M,M)\) for every \(\mathfrak g\)-module \(M\), and the kernel of this map is the annihilator \(\text{Ann}(M)\) of \(M\) in \(U(\mathfrak g)\). The classical problem of Kostant asks for which \(\mathfrak g\)-modules \(M\) the natural injection \(U(\mathfrak g)/\text{Ann}(M)\to {\mathcal L}(M,M)\) is surjective? For every involution \(w\) of the symmetric group \(S_n\) the authors establish, in terms of a special canonical quotient of the dominant Verma module associated with \(w\), an effective criterion to verify whether the universal enveloping algebra \(U(\text{sl}_n)\) surjects onto the space of all ad-finite linear transformations of the simple highest weight module \(L(w)\). An easy sufficient condition derived from this criterion admits a straightforward computational check (using a computer, for example). All this is applied to get some old and many new results, which answer the classical question of Kostant in special cases; in particular the authors give a complete answer for simple highest weight modules in the regular block of \(\text{sl}_n\), \(n < 5\).