Hochster's theta invariant and the Hodge-Riemann bilinear relations (Q616900): Difference between revisions

Let \(R\) be an isolated hypersurface singularity. \textit{M. Hochster}'s theta invariant \(\theta(M,N)\) of any two finitely generated \(R\)-modules \(M\), \(N\) is defined to be \(\text{length}(\text{Tor}_{2i}^R(M,N))-\text{length}(\text{Tor}_{2i+1}^R(M,N))\) for \(i\gg 0\) [Algebraic geometry, Proc. Conf., Chicago Circle 1980, Lect. Notes Math. 862, 93--106 (1981; Zbl 0472.13005)]. In a recent preprint Dao has conjectured that the pairing \(\theta\) vanishes if the dimension of \(R\) is even and \(R\) contains a field. The main result of the paper under review is a proof of this conjecture when moreover \(R\) is graded with its irrelevant maximal ideal giving the isolated singularity. The paper's main technique is to factorize the \(\theta\)-pairing through étale or singular cohomology via the Chern character. This technique leads to interesting results also when \(\dim(R)\) is odd: if moreover the characteristic of \(R\) is zero the authors show that the pairing \((-1)^{\frac{n+1}{2}}\theta\) is positive semi-definite, they identify its kernel using the Hodge-Riemann bilinear relations and they finally show that an \(R\)-module \(M\) is Tor-rigid if \(\theta(M,M)=0\). In the final chapter the authors extend their results to hypersurfaces that are homogeneous with respect to a non-standard grading.