A chiral Borel-Weil-Bott theorem (Q409631)

Let \(G\) be a simple complex Lie group, \({\mathfrak g}={\mathfrak n}_+\oplus{\mathfrak h}\oplus{\mathfrak n}_-\) its Lie algebra and \(X= G/({\mathfrak h}\oplus{\mathfrak n}_-)\) the corresponding flag manifold. Then there is a localization functor \[ \Delta:{\mathfrak g}_-\text{Mod}\to{\mathcal D}^\lambda_X{_-\text{Mod}}, \] where \({\mathcal D}^\lambda_X\) is the algebra of wisted differential operators acting on \({\mathcal L}_\lambda\); \({\mathcal L}_\lambda\) is the invertible \(G\)-equivariant sheaf of \({\mathcal O}_X\)-modules corresponding to an integral weight \(\lambda\in{\mathfrak h}^*\). Let \({\mathcal L}_{\nu_0}= \Delta(V_{\nu_0})\), \(V_{\nu_0}\) is the simple (finite-dimensional) \({\mathfrak g}\)-module with highest weight \(\nu_0\). Then the Borel-Weil-Bott theorem asserts \(H^0(X,{\mathcal L}_{\nu_0})= V_{\nu_0}\) and \(H^i(X,{\mathcal L}_{\nu_0})= 0\), \(i\geq 1\). The authors and Chebotarev introduced an analogue of localization functor \[ {\mathcal Z}hu_{\nu(z)}\circ \Delta:{\mathfrak g}_-\text{Mod}\to{\mathcal D}^{ch,tw}_X{_-\text{Mod}_{\nu(z)}}, \] in the case of the affine Lie algebra \(\widehat{\mathfrak g}\), a universal central extension of \({\mathfrak g}\otimes \mathbb C[t, t^{-1}]\) [\textit{T. Arakawa}, \textit{D. Chebotarev} and \textit{F. Malikov}, Sel. Math., New Ser. 17, No. 1, 1--46 (2011; Zbl 1233.17021)], hereafter referred to as [1]. Here, \({\mathcal D}^{ch,tw}_X\) is the sheaf of twisted chiral differential operators proposed in [1], and \(\nu(z)= \nu_0/z+ \nu_{-1}+ \nu_{-2}z+\cdots\in{\mathfrak h}^*((z))\). \({\mathcal L}^{ch}_{\nu(z)}={\mathcal Z}hu_{\nu(z)}\circ\Delta(V_{\nu_0})\) is an analogue of \({\mathcal L}_{\nu_0}\) in the case of \(\widehat{\mathfrak g}\). In this paper, formal character \(\text{ch\,}H^i(X,{\mathcal L}^{ch}_{\nu(z)})\) is defined (\S3,(3.12)) and set \[ \chi({\mathcal L}^{ch}_{\nu(z)})= \sum^{\dim X}_{i=0} (-1)^i\text{ch\,}H^i(X,{\mathcal L}^{ch}_{\nu(z)}). \] Then the following chiralization of the Borel-Weil-Bott theorem is proved. Theorem 1.1. Let \(W\) be the Weyl group of \({\mathfrak g}\), \(\ell(w)\) is the length of \(w\in W\), then \[ \chi({\mathcal L}^{ch}_{\nu(z)})= \sum_{w\in W} (-1)^{\ell(w)} e^{w\circ\nu_0}\times \prod_{w\in\widehat\Delta\, re_+} (1- e^{-\alpha})^{-1}, \] \[ H^i(X,{\mathcal L}^{ch}_{\nu(z)})= \bigoplus_{w\in W,\ell(w)= i} \mathbb V_{nu(z)}[\langle\nu_0- w\circ\nu_0, \rho^\nu\rangle]. \] Here \(\mathbb V_{\nu(z)} [m]\) stands for \(\mathbb V_{nu(z)}\) as a \(\widehat{\mathfrak g}\)-module with conformal filtration shifted by \(m\) (explained in \S3.1.5). Corollary 1.2. \[ \text{ch\,}\mathbb V_{\nu(z)}= {\sum_{w\in W}(-1)^{\ell(w)} e^{w\circ\nu_0}\over \prod_{\alpha\in\Delta_+} (1- e^{-\langle \nu_0+ \rho,\alpha^\vee\rangle\delta}) \prod_{\alpha\in \widehat\Delta^{re}_+} (1- e^{-\alpha})}. \] In the homogeneous grading specialization (\(e^\alpha\to 1\), \(e^{-\delta}\to q\)), we have simpler formulas \[ \chi({\mathcal L}^{ch}_{\nu(z)}, q)= \dim V_{\nu_0} \prod^\infty_{j=1} (1- q^j)^{-2\dim X}, \] \[ \dim_Q\mathbb V_{\nu(z)}= \dim V_{\nu_0} \prod^\infty_{j=1} (1- q^j)^{-2\dim X} \prod_{\alpha\in\Delta_+} (1- q^{\langle\nu_0+ \rho, \alpha^\vee\rangle})^{-1}. \] The authors remark this character formula is not new [\textit{A. Arakawa}, Characters of representations of affine Kac-Moody-Lie algebras at the critical level, \url{arXiv:0706.1817}, \textit{E. Frenkel} and \textit{D. Gaitsgory}, Weyl modules and opers without monodromy. Arithmetic and geometry around quantization. Basel: Birkhäuser. Prog. Math. 279, 101--121 (2010; Zbl 1231.17007)], and say the point is not so much the formula itself but the fact that it nicely fits in and follows the proposed geometric framework. This character formula can be applied to the computation of elliptic genus of \(X\). Because introducing generating function \[ \begin{multlined} {\mathcal E}_\lambda={\mathcal L}_\lambda\otimes \Biggl(\bigoplus^\infty_{n=1} \Biggl(\bigoplus^\infty_{m=0} q^{nm} S^m{\mathcal T}_X\Biggr)\Biggr)\otimes \Biggl(\bigotimes^\infty_{n=1} \Biggl(\bigotimes^\infty_{m=0} q^{nm} S^m \Omega_X\Biggr)\Biggr)=\\ {\mathcal L}_\lambda+q{\mathcal E}_{\lambda,1}+ q^2{\mathcal E}_{\lambda,2}+\cdots,\end{multlined} \] and define \(\chi({\mathcal E}_\lambda, q)= \chi({\mathcal L}_\lambda)+ q\chi({\mathcal E}_{\lambda, 1})+\cdots\), it is known \(\chi({\mathcal E}_\lambda, q)= \chi({\mathcal L}_\lambda)+ q\chi({\mathcal E}_{\lambda, 1})+\cdots\), is the elliptic genus \(g_Q(X, q)\) of \(X\) [\textit{L. Borisov} and \textit{A. Libgober}, Invent. Math. 140, No. 2, 453--485 (2000; Zbl 0958.14033)], where \[ Q(x)= {x\over 1-e^{-x}} \prod^\infty_{n=1} (1- q^n e^{-x})^{-1}(1- q^n e^x)^{-1}. \] In [\textit{E. Frenkel} and \textit{D. Gaitsgory}, Local geometric Langlands correspondence: the spherical case. Algebraic analysis and around. Adv. Stud. Pure Math. 54, 167--186 (2009; Zbl 1192.17012)], it is shown \[ H^{\infty/2+k}_{DS}(L{\mathfrak n}_+, H^i(X,{\mathcal L}^{ch}_{\nu(z)}))= \bigoplus^{m_i}_{j=1} \mathbb C[n_{ij}], \] if \(k=0\), and \(=0\) otherwise. It also follows from this paper \(H^i(X,{\mathcal L}^{ch}_{\nu(z)})= \bigoplus^{m_i}_{j=1} \mathbb V_{\nu(z)}[n_{ij}]\) (meaning of \(\mathbb V_{\nu(z)}[n_{ij}]\) is explained in \S4.3). Here the Drinfeld-Sokolov reduction functor \(H^{\infty/2+\bullet}_{DS}(L{\mathfrak n}_+,M)\) is the cohomology of \((M\otimes C\ell({\mathfrak n}_+), d)\), \(C\ell({\mathfrak n}_+)\) is the vertex Clifford algebra defined in \S2.1.5 [cf. \textit{E. Frenkel} and \textit{D. Ben-Zvi}, Vertex algebras and algebraic curves. 2nd ed. Providence, RI: American Mathematical Society (2004; Zbl 1106.17035)]. Therefore main part of this paper is detailed studies of Drinfeld-Sokolov reduction at the critical level, which are given in \S3. They contain the following results, which are interesting in themselves 1. the functor \(H^{\infty/2+i}_{DS}(L{\mathfrak n}_+,?)= 0\) if \(i>0\); 2. the functor \(H^{\infty/2+i}_{DS}(L{\mathfrak n}_+,?)\) is right exact, and the class of modules with Verma filtration is adapted to this functor; 3. \(H^{\infty/2+0}_{DS}(L{\mathfrak n}_+,?)\), \(i>0\) is isomorphic to the derived functor \(L^i H^{\infty/2+0}_{DS}(L{\mathfrak n}_+,?)\), (Th.3.5). The authors say this result shows somewhat unexpectedly, that the Drinfeld-Sokolov reduction, \(H^{\infty/2+\bullet}_{DS}(L{\mathfrak n}_+,?)\), is more like homology. Then Theorem 1.1 and Corollary 1.2 are proved in \S4. Definitions and examples of vertex algebras and chiral differential operators, including definition of \({\mathcal Z}hu\), are given in \S2.