Density of quasismooth hypersurfaces in simplicial toric varieties (Q2011702)

A result of \textit{B. Poonen} [Ann. Math. (2) 160, No. 3, 1099--1127 (2005; Zbl 1084.14026)] states that the proportion of degree~\(k\) hypersurfaces in a smooth projective variety \(Y\) of dimension \(r\) over \({\mathbb F}_q\) that are smooth approaches \(\zeta_Y(r+1)^{-1}\) as \(k\to\infty\). Here \(\zeta_Y\) is the usual Hasse-Weil zeta function. The aim here is to extend Poonen's result to weighted projective spaces or more generally to projective simplicial toric varieties. In this context it is not sensible to ask for smoothness, either of \(Y\) or of its subvarieties: rather one should ask for quasismoothness. So let us fix a projective simplicial toric variety \(X\) over \({\mathbb F}_q\), and a quasismooth subscheme \(Y\) meeting the singular locus of \(X\) in finitely many points. As we no longer have a natural polarisation, we need to fix one, but picking an ample divisor is overly restrictive. Instead we fix an ample Cartier divisor \(E\) on \(X\) and a Weil divisor \(D\), and consider sections \(f\in H^0(X, {\mathcal O}_X(D+kE))\). Then the result is that the proportion of such \(f\) for which \(Y\cap V(f)\) is quasismooth tends to \(\prod_P (1-q^{-\nu_P(D)})\) as \(k\to \infty\), where the product is taken over all closed points \(P\in Y\) and the number \(\nu_P(D)\) is the rank of the restriction map \(H^0(X, {\mathcal O}_X(D+kE))\to H^0(Y_P, {\mathcal O}_X(D)|_{Y_P})\) for a suitable \(0\)-dimensional subscheme \(Y_P\). This number is equal to \((\deg P)(1+\dim Y)\) if \(Y\) is smooth at~\(P\). The main new contribution here is really the definition of \(Y_P\) and of \(\nu_P\). The point is that quasismoothness (i.e.\ smoothness of the affine cone) of \(Y\cap V(f)\) is a local condition on \(Y\), so can be detected by restriction to the correct \(0\)-dimensional subscheme. Once this idea has been turned into a precise statement and the technical details have been worked out, which takes some insight and care, the proof of the result follows Poonen. The last part of the paper derives some consequences, considering for instance what happens if one allows the intersection to fail to be quasismooth at some finite number of points. For the case where \(X\) is smooth, this allows one to calculate the proportion of hypersurfaces that have, for instance, at most a single node. For instance, if \(X={\mathbb P}^2\) the proportion of curves that are at worst \(1\)-nodal is computed completely explicitly. For \(q=5\) it is about \(0.93113\).