Past exam of the mathematics course of the University of Cambridge 2010 iii Paper 16 5 ii Solution Created 2026-10-03 Updated 2026-10-14
Put and . A homogeneous polynomial of positive degree is nonzero, and is an integral domain. Multiplication by therefore gives an exact sequence of coherent sheaveswhere is the closed immersion.
We compute the necessary cohomology of twisting sheaves on projective space. The standard three affine open subsets and all their intersections are affine schemes. For a quasi-coherent sheaf, vanishing of quasi-coherent cohomology on an affine scheme and the acyclic cover theorem identify its sheaf cohomology with the Čech cohomology of this cover. In degree , the Čech cochain complex isDecompose it into subcomplexes indexed by Laurent monomials with . Let . Such a monomial occurs in the summand for a nonempty index set precisely when , since only the variables in are inverted.
If is empty, the monomial contributes the cochain complex of a full two-simplex: its degree-zero cohomology is and its higher cohomology vanishes. If , the dimensions in degrees are ; the first map has rank one and the second map has rank one with the same kernel as the first map's image. If , there is one copy of in each of degrees and and the map between them is an isomorphism. These two cases are acyclic over every field, including characteristic two. If , only the degree-two term occurs, contributing one copy of . It follows thatand that , , and . Moreover the Laurent-monomial description of top cohomology on projective space gives a basis of consisting ofFor , subtracting one from each counts these as . For there are no such triples, and the formula is also zero.
Use the long exact sequence in sheaf cohomology of the initial sequence, and cohomology under a closed immersion to identify the terms for . The vanishings just proved giveThereforeThis proves the cohomology of a projective plane hypersurface without imposing smoothness, irreducibility or reducedness. In particular, the second number is an arithmetic genus, not a claim about the genus of a normalization.