Factorisation of generalised theta functions. II: The Verlinde formula (Q1918437)

In part I of this paper [Invent. Math. 114, No. 3, 565-623 (1993; Zbl 0815.14014)], \textit{M. S. Narasimhan} and the author had proved that the space of generalized theta functions on the moduli space of semi-stable rank-2 vector bundles of degree \(d\) over a complex curve of genus \(g\geq 4\), with precisely one node, is (non-canonically) isomorphic to the corresponding space of generalized theta functions for the desingularization of the base curve. Moreover, it was shown, and used for the proof of the isomorphism theorem, that (again for \(g\geq 4\)) the first cohomology group of the theta bundle on the moduli space \(U_X (d)\) of semi-stable rank-2 bundles of degree \(d\), over the nodal base curve \(X\), is trivial. Finally, the authors had announced that the restricting condition \(\geq 4\) for the genus of \(X\) can actually be dropped and, as to that fact, referred to a forthcoming paper. The present article, the second in a series of planned papers, provided the promised proof of the announced more general result. The proof of the vanishing theorem: ``\(H^1 (U_X (d), \Theta)= (0)\) for a given nodal curve of arithmetic genus \(g\geq 1\) and with precisely one node'', together with the derivation of a concrete Verlinde formula for the dimension of the space \(H^0 (U_X (d), \Theta)\) of generalized theta functions on \(U_X (d)\), is based on two new ingredients. One of them is a rather general result from geometric invariant theory and invariant cohomology, which is certainly of independent interest. More precisely, this auxiliary result states that, for a given projective variety acted on by a reductive algebraic group \(G\), the invariant cohomology (with coefficients in an ample \(G\)-linearized line bundle \(L\)) coincides with that of the open set of semi-stable points (with respect to \(G\)). The second basic ingredient is a statement about the effect of certain ``Hecke transformations'' on quasi-parabolic rank-2 sheaves over genus-zero curves with one node. The proof of the Verlinde formula for \(\dim_C H^0 (U_X (d), \Theta)\) then uses the vanishing theorem, \(H^1 (U_X (d), \Theta)= (0)\), and an induction argument with respect to the genus \(g\). As to the starting point \((g=0)\), the corresponding Verlinde formula was established by \textit{K. Gawȩdzki} and \textit{A. Kupiainen} in 1991 [cf.: Commun. Math. Phys. 135, No. 3, 531-546 (1991; Zbl 0722.53084)].