A note on a Griffiths-type ring for complete intersections in Grassmannians (Q2664653)

Let \(X=\mathbf{V}(f)\subset\mathbb P^{n+1}\) be a smooth hypersurface of degree \(\deg f=d\). A classical result of [\textit{P. A. Griffiths}, Am. J. Math. 90, 805--865 (1968; Zbl 0183.25501)] identifies the primitive part of \(H^{n-p+1,p-1}(X)\) with the \(pd-n-2\) homogeneous component of \(\mathbb C[x_0,\ldots,x_{n+1}]/J_f\), where \(J_f=(f_0,\ldots,f_{n+1})\) is the Jacobian ideal generated by the partial derivatives of \(f\). There exist various generalizations of this statement. For instance to complete intersections in projective space by [\textit{A. Dimca}, Duke Math. J. 78, No. 1, 89--100 (1995; Zbl 0839.14009)], or toric varieties by [\textit{V. V. Batyrev} and \textit{D. A. Cox}, Duke Math. J. 75, No. 2, 293--338 (1994; Zbl 0851.14021)], [\textit{K. Konno}, Compos. Math. 78, No. 3, 271--296 (1991; Zbl 0737.14002)], [\textit{A. R. Mavlyutov}, Pac. J. Math. 191, No. 1, 133--144 (1999; Zbl 1032.14013)], or hypersurfaces of high degree in arbitrary projective manifolds by [\textit{M. L. Green}, Compos. Math. 55, 135--156 (1985; Zbl 0588.14004)], or recently to zero loci of homogeneous vector bundles by [\textit{A. Huang} et al., ``Jacobian rings for homogenous vector bundles and applications'', Preprint, \url{arXiv:1801.08261}]. The main results of this paper are generalizations of Griffiths' result to complete intersections in Grassmann varieties. They do not use primitive cohomology, but a refinement, the vanishing cohomology. For hypersurfaces in projective space the distinction is irrelevant. For instance if \(X\subset G=\mathrm{Gr}(k,n)\) is a smooth hypersurface of degree \(d\geq n-1\) and dimension \(N-1=k(n-k)-1\), and \(N\) is odd, then the vanishing part of \(H^{N-1-p,p}(X)\) is isomorphic to the degree \((p+1)d-n\) part of \(R^G_f\). The latter is what the authors call the Griffiths ring of \(X\). It is the quotient of the Plücker coordinate ring of \(G\) by the homogeneous ideal generated by \(f\) and its \(\mathfrak{sl}_n\)-orbit. Here \(\mathfrak{sl}_n\) acts by derivations in a concrete way. When \(N\) is even, then we may no longer obtain an isomorphism, but the difference is explicit, determined only by \(G\) and \(p\). The authors prove a similar result for complete intersections of more than one hypersurface in \(G\). The Griffiths-type ring is now denoted \(\mathcal U\). Its definition is cohomological, motivated by a Cayley trick to reduce from complete intersections in \(G\) to hypersurfaces in a projective bundle over \(G\). The main advantage of the results of this paper over other generalizations is their explicit nature, mirroring the original result of Griffiths. The authors give concrete presentations of their Griffiths-type rings and use them to compute several explicit examples of Hodge groups. For instance they carry out these computations for Fano 5-folds and 4-folds of genus 6 and degree 10, a Calabi-Yau section of \(\mathrm{Gr}(2,7)\), and for Fano varieties of K3-type. Some of these computations of Hodge groups were known, but the ring structure is new. Apart from classical methods such as chasing exact sequences in cohomology, the paper makes great use of the Cayley trick for reducing complete intersections to hypersurfaces in a projective bundle, and of the connection between the Hodge theory of a particular class of projectively normal \(X\) and the infinitesimal first-order deformation module of the affine cone \(A_X\).