Norm-square localization and the quantization of Hamiltonian loop group spaces (Q2295695): Difference between revisions

Let $G$ be a compact connected Lie group, which is assumed to be simple and simply connected, $LG$ the loop group of $G$, $\Phi_{\mathcal{M}}:\mathcal{M}\rightarrow L_{\mathfrak{g}^{\ast}}$ a Hamiltonian $LG$-space, with level $k>0$ prequantum line bundle $L$ and $T\subset G$ a maximal torus. In [``Spinor modules for Hamiltonian loop group spaces'', Preprint, \url{arXiv:1706.07493}], the authors, together with \textit{E. Meinrenken}, constructed a finite-dimensional \textit{global transversal} $\mathcal{Y}\subset\mathcal{M}$ as well as a canonical spinor module $S_{0}\rightarrow\mathcal{Y}$. The submanifold $\mathcal{Y}$ is a small \textit{thickening} of the singular subset $\mathcal{X}=\Phi_{M}^{-1}(t^{\ast}) \subset\mathcal{M}$. The submanifold $\mathcal{Y}$ is equipped with a moment map $\phi:\mathcal{Y}\rightarrow\mathfrak{t}$, which is proper on the support of the Bott-Thom element. In [Math. Ann. 374, No. 1--2, 681--722 (2019; Zbl 1416.81068)], the authors investigated a Dirac-type operator $\mathcal{D}$ on $\mathcal{Y}$ acting on sections of $\wedge\mathfrak{n}_{-}\widehat{\otimes}S$, where $S=S_{0}\otimes L$ and $\mathfrak{n}_{-}\subset\mathfrak{g}_{\mathbb{C}}$ denotes the sum of the negative root spaces of $G$. The operator $\mathcal{D}$ was shown to represent an index pairing between a spin-$\mathrm{c}$ Dirac operator for $S$ and the pullback of a Bott-Thom element for $\mathfrak{g}/\mathfrak{t}$, the latter formally playing the role of a Poincaré dual to $\mathcal{X}$ in $\mathcal{Y}$. This paper, consisting of 6 sections, studies a deformation of the above operator in the vein of [\textit{Y. Tian} and \textit{W. Zhang}, Invent. Math. 132, No. 2, 229--259 (1998; Zbl 0944.53047); \textit{X. Ma} and \textit{W. Zhang}, Acta Math. 212, No. 1, 11--57 (2014; Zbl 1380.53102); \textit{X. Ma} and \textit{W. Zhang}, Prog. Math. 297, 299--315 (2012; Zbl 1270.58013)]. \S 2 is a brief introduction to elliptic boundary value problems, following mostly [\textit{C. Bär} and \textit{W. Ballmann}, Prog. Math. 319, 43--80 (2016; Zbl 1377.58022); Surv. Differ. Geom. 17, 1--78 (2012; Zbl 1331.58022)]. \S 3 summarizes some results from [Math. Ann. 374, No. 1--2, 681--722 (2019; Zbl 1416.81068); \url{arXiv:1706.07493}]. \S 4 introduces a deformation $\mathcal{D}_{t}$ of $\mathcal{D}$ defined by \[ \mathcal{D}_{t}=\boldsymbol{D}+(1+t) f\theta\widehat{\otimes}1-it\widehat{\otimes}\mathrm{c}(v_{\mathcal{Y}}). \] Inspired by [\textit{X. Ma} and \textit{W. Zhang}, Acta Math. 212, No. 1, 11--57 (2014; Zbl 1380.53102)], \S 5 establishes a formula for $\mathrm{index}(\mathcal{D})$, which is expressed as a sum of contributions indexed by the components $Z_{\beta}=\mathcal{Y}^{\beta}\cap\phi^{-1}(\beta)$ of $Z=\{v_{\mathcal{Y}}=0\}$: \[ \mathrm{index}(\mathcal{D}) =\sum_{\beta\in W\cdot\mathcal{B}}\lim_{t\rightarrow\infty}\mathrm{index}_{\mathrm{APS}}(\mathcal{D}_{t}\upharpoonright U_{\beta}) \] Each constribution is a limit in $R^{-\infty}(T)$ as $t\rightarrow\infty$, of the index of an Atiyah-Patodi-Singer (APS) boundary value problem on a compact neighborhood $U_{\beta}$ of $Z_{\beta}\cap\mathcal{X}$. Following the lines of [Zbl 1380.53102; Zbl 1270.58013; \textit{M. Braverman}, \(K\)-Theory 27, No. 1, 61--101 (2002; Zbl 1020.58020)], \S 6 establishes a formula for the contributions in the above formula in terms of transversally elliptic operators in the spirit of [\textit{P.-E. Paradan}, J. Funct. Anal. 187, No. 2, 442--509 (2001; Zbl 1001.53062)]: \[ \mathrm{index}(\mathcal{D}) =\sum_{\beta\in W\cdot\mathcal{B}}\mathrm{index}(\sigma_{\beta,\theta}\otimes\mathrm{Sym}(\upsilon_{\beta})) \] where $\sigma_{\beta,\theta}$ is a transversally elliptic symbol on the fixed-point set $\mathcal{Y}^{\beta}$, and $\upsilon_{\beta}$ is the normal bundle to $\mathcal{Y}^{\beta}$ in $\mathcal{Y}$. endowed with a $\beta$\textit{-polarized} complex structure. The formula is sometimes called a \textit{norm-square localization} formula. \S 6.4 is concerned with the $[Q,R]=0$ theorem, which follows from the main result of this paper together with a relatively small part of [arXiv:1604.01965], though its complete proof is postponed to a subsequent paper. The most important application of the $\left[ Q,R\right] =0$ theorem for Hamiltonian loop group spaces is to the Verlinde formula, concerning which the reader is referred to [Zbl 0997.53066] for a symplectic approach and to [\textit{E. Meinrenken}, Contemp. Math. 583, 175--210 (2012; Zbl 1365.53080); \textit{E. Meinrenken}, Int. Math. Res. Not. 2012, No. 20, 4563--4618 (2012; Zbl 1260.53143)] for the relationship with $[Q,R]=0$ for Hamiltonian loop group spaces and quasi-Hamiltonian $G$-spaces.