Defect of projective hypersurfaces with isolated singularities

TL;DR

This paper investigates the defect of projective hypersurfaces with isolated singularities, relating it to cohomological properties, spectral sequences, and providing explicit computational methods, including examples with positive defect.

Contribution

It introduces a new approach to compute the defect using spectral sequences and demonstrates explicit calculations for hypersurfaces with isolated singularities.

Findings

Defect relates to cohomology and monodromy of hypersurfaces.

Explicit computation of defect via spectral sequences.

Examples include hypersurfaces with positive defect and a single singularity.

Abstract

Let $X$ be a hypersurface with isolated singularities defined by $f$ in ${\bf P^{n+1}}$ with $n>1$. The difference ${\rm def}(X):=h^{n+1}(X)-h^{n-1}(X)$ is called the defect of $X$ (for self-duality of the cohomology of $X$). It is known that its vanishing is closely related to ${\bf Q}$-factoriality of $X$ in the rational singularity case with $n=3$. This number coincides with the dimension of the cokernel of the inclusion $H^{n-1}(X)\to{\rm IH}^{n-1}(X)$, the rank of the morphism from the vanishing cohomologies of $X$ to $H^{n+1}(X)$ for a one-parameter smoothing of $X$ with total space smooth, and also with the dimension of the unipotent monodromy part of the Milnor fiber cohomology of $f$ with degree $n$. In the case $X$ has only weighted homogeneous isolated singularities, the defect ${\rm def}(X)$ is then given by the $E_2$-term of the spectral sequence of the double complex with differentials ${\rm d}f\wedge$ and $\rm d$ by the $E_2$-degeneration of the pole order spectral sequence. It can be calculated explicitly using a computer even for analogues of the Hirzebruch quintic threefold with more than one hundred ordinary double points found by B.\ van Geemen and J.\ Werner in a compatible way with their computation. We give also an example with ${\rm def}(X)>0$ and $|{\rm Sing}\,X|=1$ where $n=3$.