Symmetric theta-structures (Q1173611)

Let \(X\) be a complex abelian variety of dimension \(g\). An ample line bundle \(L\) on \(X\) defines a polarization \(H\) on \(X\), i.e., by taking its first Chern class, a Hermitean form \(H\) on the universal covering space of \(X\), whose imaginary part is integer-valued on the lattice defining \(X\). Assume that \(H\) is of type \(D=\text{diag}(d_ 1,\ldots,d_ g)\in M_ \mathbb{Z}(g,g)\) with respect to a Frobenius basis of the lattice. According to these data, one has the theta group \({\mathcal G}(L)\) of the line bundle \(L\) and the Heisenberg group \(Heis(D)\) of type \(D\). A theta structure for the line bundle \(L\) is then an isomorphism \(b:{\mathcal G}(L)\to Heis(D)\) inducing the identity on the natural subgroups \(\mathbb{C}^*\). If \(L\) is assumed to be a symmetric line bundle, i.e. \((- 1)^*_ XL\cong L\), then a symmetric theta structure on \(L\) is defined to be an ordinary theta structure which is compatible with the induced involutions on \({\mathcal G}(L)\) and \(Heis(D)\). The aim of the present paper is to answer the natural question of whether a given line bundle admits a (symmetric) theta structure, and how many such structures exist. The authors give a complete and precise solution of this particular problem, which is of special importance for moduli problems in the theory of abelian varieties [cf. \textit{D. Mumford}, Invent. Math. 1, 287-354 (1966; Zbl 0219.14024)]. --- In section 1 of their paper, the authors introduce the concept of characteristics for an ample line bundle. These characteristics appear as elements in the universal covering space \(V\) of \(X\), and are generalizations of the classical theta characteristics for theta functions. Section 2 contains a slightly modified definition of the theta group of a line bundle, together with an explicit description of its elements via lifting to linear automorphisms of the (trivial) pull-back line bundle on the universal covering space \(V\). --- Heisenberg groups and theta structures for line bundles are briefly recalled in section 3. It is shown that every characteristic \(c\in V\) for a line bundle \(L\) determines a theta structure \(b_ c:{\mathcal G}(L)\to Heis(D)\) of \(L\), so that every ample line bundle on \(X\) really admits at least one theta structure. --- The precise number of theta structures on a given ample line bundle \(L\) is computed in section 4. More precisely, it is proved that this number is exactly \(h^ 0(X,L)^ 2\cdot\text{card}(Sp(D))\), where \(Sp(D)\) denotes the symplectic group of the polarization type \(D\). The proof is essentially based on the observation that the set of theta structures on \(L\) can be parametrized by the characteristics of \(L\), which had been introduced in section 1. --- The remaining two sections deal with symmetric line bundles and symmetric theta structures on them. The main result in setion 5 is the computation of the dimensions \(h^ 0(L)_ +\) and \(h^ 0(L)_ -\) of the eigenspaces of \(H^ 0(X,L)\) with respect to the involution \((-1)^*_ L\) induced by the normalized isomorphism \((- 1)_ L\) of the symmetric line bundle \(L\). The formula says that \(h^ 0(L)_ +={1\over 2}h^ 0(X,L)\) or \(h^ 0(L)_ -={1\over 2}h^ 0(X,L)\pm 2^{g-s-1}\), where \(s\) denotes the number of odd entries in \(D=\text{diag}(d_ 1,\ldots,d_ g)\), and in terms of the corresponding characteristic of \(L\) it is regulated which case occurs. In the concluding section 6, the authors compute the number of symmetric theta structures of a given symmetric ample line bundle \(L\). Depending on the characteristic of \(L\), the nunber \(n(L)\) of symmetric theta structures of \(L\) turns out to be equal to 0 or to \(\text{card}(Sp(D))\cdot 2^{2(g- s)}\). This means, in particular, there are exactly \(2^{2s}\) symmetric line bundles with polarization class \(H\), which admit symmetric theta structures. Moreover, putting the main results together, the following concluding theorem is obtained: An ample symmetric line bundle \(L\) on \(X\) admits a symmetric theta structure if and only if the eigenspace dimensions \(h^ 0(L)_ +\) or \(h^ 0(L)_ -\) are maximal, and exactly \(2^{2s}\) out of \(2^{2g}\) line bundles in a given polarization class have this property. The paper is very clearly and carefully written. The beautiful and important result is presented in a nearly self-contained way.