Cotangent bundle to the flag variety. I (Q2416407)

For a semi-simple algebraic group \(G\) on \(\mathbb C\), the affine Weyl group \(\widehat W\) is defined as \(\widehat W = N(\mathbb C[t,t^{-1}])/T\), where \(T\) is a maximal torus in \(G\) and \(N\) is the normalizer of \(T\). Elements \(w\in \widehat W\) define \textit{affine Schubert varieties} \[ X(w)= \mathop{\dot{\bigcup}}_{\tau \leq w} \mathcal{B}\tau\mathcal{B} \pmod{\mathcal{B}}, \] where \(\mathcal{B}\) is the inverse image of a Borel subgroup \(B\supset T\) in the projection \(G(\mathbb C[[t]])\to G\). These varieties \(X(w)\) are indeed projective. In the paper [Transform. Groups 21, No. 2, 519--530 (2016; Zbl 1390.14148)], the first author showed that there is an embedding of the cotangent bundle \(T^*G_{d,n}\) of a Grassmannian \(G_{d,n}\) into a Schubert variety \(X(\iota)\) (over \(G=\mathrm{SL}(n)\)), that provides a compactification of the cotangent bundle. The result has been generalized by the authors to the cotangent bundles of cominuscule Grassmannian varieties \(G/P\). The aim of the paper under review is to explore extensions of the previous embeddings to the contangent bundle of quotients \(G/B\), where \(B\) is a Borel subgroup. The authors indeed determine an embedding of \(T^*G/B\) into a suitable variety \(X(\kappa_0)\), which identifies a certain \(\mathrm{SL}_n\)-stable closed subvariety of \(X(\kappa_0)\) as a compactification of \(T^*G/B\). On the other hand, the authors prove that one cannot realize an affine Schubert variety in \(\widehat{\mathrm{SL}}_3/\mathcal B\) as a compactification of \(T^*\mathrm{SL}_3/B\).