SPIN VERLINDE SPACES AND PRYM THETA FUNCTIONS (Q4487545)

Let \(G\) be a semi-simple algebraic group and \((g,\ell) \in \mathbb{Z}^2\) a pair of integers with \(g\geq 0\). Then the Verlinde formula is a trigonometric expression which yields a natural number \(N_\ell(G,g)\) associated with these data \((G,g,\ell)\). In physics, the Verlinde number \(N_\ell(G,g)\) is interpreted as the dimension of the space of conformal blocks at level \(\ell\) in the Wess-Zumino-Witten model of conformal field theory on a compact Riemann surface \(C\) of genus \(g\). In algebraic geometry, the Verlinde number \(N_\ell (G,g)\) has been proved to be the dimension of the space \(H^0({\mathcal M}_C(G)\), \({\mathcal L}^{\otimes 1})\), where \({\mathcal M}_C(G)\) is the moduli space of semistable principal \(G\)-bundles over the curve \(C\), and \({\mathcal L}\) is the ample generator of its Picard group. The paper under review is concerned with value distribution of the numbers \(N_\ell(G,g)\) for the special classical group \(G= \text{Spin}_m\). More precisely, the author derives interesting identities linking the complex spin groups \(\text{Spin}_m\) with the configuration of principally polarized Prym varieties associated to the curve \(C\) via its unramified double covers. While this connection was first established in the author's foregoing paper ``Prym varieties and the moduli of spin bundles'' [\textit{W. M. Oxbury}, in: Algebraic geometry, Pap. presented EUROPROJ Conf., Catania 1993 and Barcelona 1994, Lect. Notes Pure Appl. Math. 200, 351-376 (1998; Zbl 0940.14020)] for the odd spin groups \(\text{Spin}_{2n+1}\), the present paper provides an extension of these results to the remaining case of the even spin groups \(\text{Spin}_{2n}\). The main results in this article are given by constructing certain homomorphisms between spaces of sections of theta bundles over \({\mathcal M}^\pm_C (\text{Spin}_{2n})\) and some associated Prym varieties \(P_\eta\) to \(C\), where \(\eta\) runs over the 2-torsion points in the Jacobian \(J(C)\) of the base curve \(C\). These natural homomorphisms do link spaces of the same dimension (via comparing Verlinde numbers and dimensions of spaces of theta characteristics), and so they are conjectured to be isomorphisms. As the author points out, in the course of this paper, this conjecture leads to various other, related conjectures, which are then verified, individually, in some special cases. One of those related conjectures gives a twisted Verlinde formula for Clifford bundles of odd degree, generalizing the twisted Verlinde formula for rank-2 vector bundles by \textit{M. Thaddeus} [J. Differ. Geom. 35, 131-149 (1992; Zbl 0772.53013)]. All together, this excellent paper provides a very deep and clear analysis of the subject, with many impulses for further research on the mystery of Verlinde numbers.