Reduction type of smooth plane quartics (Q2239228)

A smooth projective irreducible curve of genus \(3\) over an algeraically closed field is either a smooth plane quartic or a hyperelliptic curve (i.e. has a degree \(2\) map into \(\mathbb{P}^1\)), depending on whether the canonical divisor is very ample or not. The authors study smooth projective quartics \(X\) over a field with a discrete valuation with a residue field of characteristic \(p>0\). In particular, they ask if \(X\) has a smooth projective model over the discrete valuation ring whose reduction is a smooth plane quartic (resp. hyperelliptic). Then they say that \(X\) has \textit{good quartic} (resp. \textit{good hyperelliptic}) reduction. Otherwise, \(X\) is said to have \textit{bad reduction}. For \(p\neq 2\), the authors introduce the key concept of a \textit{toggle model} which means that \(X\) is isomorphic over \(K\) to a quartic of the form \(Q^2+\pi^{2s}G=0\) where \(\pi\) is the uniformizing parameter, \(G\) is a primitive quartic form, \(Q\) is a primitive quadratic form (meaning that coefficients have gcd equal to \(1\)), and the reduction modulo \(\pi\) of \(Q\) is irreducible. A \textit{good toggle model} means that in addition, the zero loci of the reductions of \(G\) and \(Q\) intersect transversely at \(8\) distinct points of the algebraic closure of the residue field. One main result of the paper is that for \(p\neq 2\), having good hyperelliptic reduction is equivalent to having a good toggle model. There is also a version for characteristic \(p=2\). (Theorems 1.4, 2.8, 2.9 of the paper.) The authors also characterize the property of \(X\) having a good quartic reduction over some finite separable extension of the ground field. Consider a homogeneous system of radical generators \(I_i\) of ternary quartic forms over the discrete valuation ring which are invariant under the action of the special linear group. The authors prove that the condition for a smooth plane quartic \(X\) having a good quartic reduction over a finite separable extension is equivalent to the condition that for every invariant \(I_i\) of degree \(d_i\), the element \(I_i^{27}/D_{27}^{d_i}\) be in the discrete valuation ring, where \(D_{27}\) is the discriminant. (Theorems 1.9, 3.15 of the paper.) A number of interesting new results in geometric invariant theory is proved in the course of the proof of these theorems.