Demazure character formula for semi-infinite flag varieties (Q1659935)

The representation theory of affine Lie algebras can be encoded by semi-infinite flag varieties. These can also be interpreted as spaces of rational maps, and so play a major role in the computation of the quantum \(K\)-theory of flag varieties. \textit{A. Braverman} and \textit{M. Finkelberg} [J. Am. Math. Soc. 27, No. 4, 1147--1168 (2014; Zbl 1367.17011); Math. Ann. 359, No. 1--2, 45--59 (2014; Zbl 1367.17010); J. Lond. Math. Soc., II. Ser. 96, No. 2, 309--325 (2017; Zbl 1384.17009)] used this last interpretation to give fundamental properties, such as normality, rationality of the singularities, an analogue of the Borel-Weil theorem, the computation of quantum \(J\) functions (in the line of Givental-Lee) and its connection with \(q\)-Whittaker functions. This article has two main objectives. The first is to extend a cohomology formula for line bundles given by Braverman and Finkelberg to include a special class of twisted sheaves. The other main objective is to generalize the other results to all Schubert varieties giving more applications to representation theory. This last extension provides a natural realization of specializations of non-symmetric Macdonald polynomials together with some difference equations characterizing them. This generalizes the link of non-symmetric Macdonald polynomials to representation theory of algebras currently defined by \textit{A. Braverman} and \textit{M. Finkelberg} [Zbl 1367.17010], \textit{C. Lenart} et al. [Int. Math. Res. Not. 2015, No. 7, 1848--1901 (2015; Zbl 1394.05143); ``A uniform model for Kirillov-Reshetikhin crystals. II: Alcove model, path model, and \(P=X\)'', Int. Math. Res. Not. 2017, No. 14, 4259--4319 (2017; \url{doi:10.1093/imrn/rnw129}); Transform. Groups 22, No. 4, 1041--1079 (2017; Zbl 1428.05325)], \textit{I. Cherednik} and \textit{D. Orr} [Math. Z. 279, No. 3--4, 879--938 (2015; Zbl 1372.20009)], \textit{S. Naito} et al. [Trans. Am. Math. Soc. 370, No. 4, 2739--2783 (2018; Zbl 1432.17015)], and \textit{E. Feigin} and \textit{I. Makedonskyi} [Sel. Math., New Ser. 23, No. 4, 2863--2897 (2017; Zbl 1407.17028)]. Let \(G\) be a simply connected simple algebraic group and let \(W\) be its Weyl group with set of simple reflections \(\{s_i\}.\) Let \(\Lambda\) be the weight lattice, \(\Lambda_+\) the set of dominant weights, let \(Q^\vee\) be the co-root lattice of \(G\). Consider the space \(\mathcal Q\) of rational maps from \(\mathbb P^1\) to \(G/B\) with subspace \(\mathcal Q(w)\) defined as the closure of the set of rational maps whose value at \(0\) is in a Schubert variety corresponding to \(w\in W\). This space has a natural line bundle \(\mathcal O(\lambda)\) for each \(\lambda\in\Lambda\). Associated to \(G\), the current algebra is defined as \(\mathfrak g[z]=\text{Lie}G\otimes_{\mathbb C}\mathbb C[z]\), and \(G\) comes with the associated Iwahori subalgebra \(\mathcal J\). For each \(\lambda\in\Lambda_+\), \(\mathfrak g[z]\) has a representation \(W(\lambda)\) called a global Weyl module. \textit{M. Kashiwara} [Publ. Res. Inst. Math. Sci. 41, No. 1, 223--250 (2005; Zbl 1147.17306)] defined its Demazure submodule \(W(\lambda)_w\) for \(w\in W\) as the cyclic \(\mathcal J\)-module generated by a vector with weight \(w\lambda\in\Lambda\). As these last modules are graded, this defines their character \(\text{ch}W(\lambda)_w\in\mathbb C(\!(q)\!)[\Lambda].\) The first main theorem is stated more or less verbatim, as it is stated in the category of ind-schemes: Theorem A. For each \(\lambda\in\Lambda\) and \(w\in W\); {\parindent=0.8cm \begin{itemize}\item[1.] The ind-scheme \(\mathcal Q(w)\) is normal, and projectively normal; \item[2.] There is an isomorphism of \(\mathcal J\)-modules \(H^i(\mathcal Q(w),\mathcal O_{\mathcal Q(w)}(\lambda))^\ast=\begin{cases} W(\lambda)_w,\;(i=0,\lambda\in\Lambda_+)\\\{0\}\text{ otherwise }\end{cases};\) \item[3.] For each \(i\in\mathsf{I}\) so that \(s_iw>w\), \(\text{ch}W(\lambda)_{s_i w}=D_i(\text{ch} W(\lambda)_w)\), where \(D_i\) is a Demazure operator acting on \(\mathbb C (\!(q)\!)[\Lambda]\); \item[4.] There exists a Demazure operator \(D_{t\beta}\) for each \(\beta\in\mathcal Q^\vee\) such that \(\langle\beta,w\alpha\rangle>0\) for every positive root \(\alpha\). This operator is mutually commutative, and \(D_{t\beta}(\text{ch}W(\lambda)_w)=q^{-\langle\beta,w\lambda\rangle}\cdot\text{ch}W(\lambda)_w\). \end{itemize}} The author remarks that (2) to (4) above are regarded as semi-infinite analogues of the Demazure character formula due to Demazure-Joseph-Kumar in the ordinary setting. We state the second main result (more or less) verbatim: Theorem B. For each \(w\in W\) and \(\lambda\in\Lambda_+\), the module \(W(\lambda)_w\) admits a free action of a certain polynomial ring and its specialization to \(\mathbb C\) gives the Feigin-Makedonskyi module \(W_{w\lambda}\). In particular, \(\Gamma(\text{FI}_G^\frac{\infty}{2})(w),\mathcal O_{\text{FI}_G^\frac{\infty}{2}(w)}(\lambda))^\ast\cong W_{w\lambda},\) where \(\text{FI}_G^\frac{\infty}{2}(w)\) is a variant of \(Q(w)\). The third theorem is a result of the comparison of Cerednik-Orr's recursive formula for non-symmetric Macdonald polynomials specialized to \(t=\infty\) with the authors construction, verbatim: Theorem C. For each \(\lambda\in\Lambda_+\) and \(w\in W\), there exists an \((\mathbf{I}\rtimes\mathbb G_m)\)-equivariant sheaf \(\mathcal E_w(\lambda)\) such that \(\text{ch}H^0(Q(w),\mathcal E_w(\lambda))^\ast=(\prod_{i\in I}\prod_{k=1}^{\langle\alpha_i^\vee,\lambda_w\rangle}\frac{1}{1-q^k})\cdot E^\dagger_{-w\lambda}(q^{-1},\infty),\) where \(\lambda_w\) is a dominant weight determined by \(\lambda\) and \(w\), and \(E^\dagger_{-w\lambda}(q,t)\) is the bar-conjugate of a non-symmetric Macdonald polynomial, in addition, \(H(Q(w),\mathcal E_w(\lambda))=\{0\}\) for \(i>0\). To prove the results, an analogue of the Kodaira vanishing theorem is presented. This leads to the fourth main result of the article, which gives relations between different specializations of non-symmetric Macdonald polynomials. The article starts off with two sections on the preliminary material regarding current algebra representations and semi-infinte flag varieties. This involves many very complicated definitions, and includes a lot of high-end results which the author proves because of lack of references. Then the necessary observation that the semi-infinite flag variety is projectively normal is proved, so that the proof of Theorem A can be given by algebraic manipulations. The last sections contain the proofs of the remaining main results. These results are simple observations from the author's point of view, but it would take a lot of labour to understand them fully, at least for this reviewer. We understand that the theory of ind-schemes, limiting to derived schemes and a touch of deformation theory are important ingredients, and they are put together in a concise and detailed manner to prove these very important results.