Symmetric tensors and geometry of \({\mathbb{P}}^N\) subvarieties (Q1024661)

The authors study the vanishing of cohomology groups of twisted symmetric differentials over a projective manifold \(X\). The main tool in their analysis is a geometrical interpretation of these differentials as defined over fibers of a map between projective bundles. Using the Euler sequence, the authors produce an embedding of the projective bundle \(\mathbb P(\tilde\Omega^1_X)\), (\(\tilde\Omega^1_X\) being an extension of \(\Omega_X^1\)), in a product \(\mathbb P^N\times X\). The fibers of the resulting projection \(f: \mathbb P(\tilde\Omega^1_X)\to \mathbb P^N\) turn out to be the embedded tangent spaces \(T_xX\), so that \(f\) is called the \textit{tangent map}. When the dimension of a general fiber of \(f\) is positive (which clearly happens when \(n=\dim(X)\) is bigger than \(N/2\)), the authors prove that \(H^0(S^m(\Omega_X^1)\otimes L)\) vanishes for any \(m\geq 0\) and for any negative line bundle \(L\) on \(X\). As a result, the authors obtain a new proof of a vanishing theorem of M. Schneider, which says that \(n> N/2\) implies \(H^0(S^m(\Omega_X^1)\otimes O_X(k))=0\) for all \(k<m\). Then the authors turn to the more difficult case \(m=k\). Here the analysis requires that the map \(f\) is dominant, which happens to be true when \(n\geq 2/3(N-1)\), by Zak's result on linear normality. In Zak's range, the authors give a geometrical description of symmetric differentials in terms of some cones, corresponding to varieties of trisecant lines which are somewhere tangent to \(X\). Since trisecant lines lie in any quadric containing \(X\), results on symmetric differentials are linked to the existence of such quadrics. For instance, the authors show that, when \(X\) is a hypersurface, then \(H^0(S^m(\Omega^1_X(1)))=0\) unless \(\deg(X)=2\). Using this interpretation, the authors are also able to produce examples of smooth families in which the dimension of spaces of twisted symmetric differentials are not constant.