A new parametrization for ideal classes in rings defined by binary forms, and applications (Q6042647)

Let \(R\) be a principal ideal domain with fraction field \(K\). To any separable binary form \(F(x,y)=\sum_{i=0}^nf_ix^{n-i}y^i\) of degree \(n\geq 1\) over \(R\) with \(f_0\neq 0\), one can associate a subring \(R_F\) inside the étale \(K\)-algebra \(K[x]/(F(x,1))\) which, as an \(R\)-module, is free of rank \(n\) with basis \((1,f_0\theta,f_0\theta^2+f_1\theta,\ldots,f_0\theta^{n-1}+\ldots+f_{n-2}\theta)\), where \(\theta\) is the image of \(x\) in \(K_F\). Furthermore, one considers the fractional \(R_F\)-ideal \(I_F=(1,\theta)\) and its powers \(I^k_F=(I_F)^k\), \(0\leq k\leq n-1\). Geometrically, \(R_F\) can be thought of as the ring of global sections of the subscheme of \(\mathbb{P}_{\mathbb{Z}}^1\) cut out by \(F\). It turns out that when \(n>2\), all ideals \(I^k_F\) are invertible iff \(F\) is primitive. Of particular importance is the ideal \(I^{n-2}_F\) as its class is that of the inverse different of \(R_F\). A good account of the construction and alternative interpretations of \(R_F\) and its fractional ideals \(I^k_F\) can be found in [\textit{M. M. Wood}, J. Lond. Math. Soc., II. Ser. 83, No. 1, 208--231 (2011; Zbl 1228.11053)]. It was also [\textit{M. M. Wood}, J. Reine Angew. Math. 689, 169--199 (2014; Zbl 1317.11039)] (extending earlier work by Bhargava in the case \(n=3\)) who showed that, for any \(n\geq 3\), square roots of the class of \(I^{n-3}_F\) can be parametrized by \(\mathrm{GL}_n(\mathbb{Z})\)-orbits of pairs \((A,B)\in\mathbb{Z}^2\otimes_{\mathbb{Z}}\mathrm{Sym}_2\mathbb{Z}^n\) such that \(\det(xA+yB)=\pm F(x,y)\). These results have generated a series of applications by various authors concerning, among other things, results on the non-existence of rational points on hyperelliptic curves, and on bounds on the average size of the \(2\)-class groups of rings of algebraic integers. As for the latter, the results could only be obtained for \(n\) odd or for monic \(F\) (i.e., \(|f_0|=1\)), the obstruction being that in the other cases, the class of \(I^{n-3}_F\) may not be a square. In the present paper, this problem is overcome by considering \(I^{n-2}_F\) instead, which is always a square when \(n\) is even and also when \(n\) is odd provided \(R_F\) is a maximal order. The author shows that square roots of the class of \(I^{n-2}_F\) can be parametrized as above with the slight modification \(\det(xA+yB)=\pm f_0^{-1}F(x,f_0y)\), where the locus of pairs \((A,B)\) that arise in this way is cut out of the hypersurface \(\det A=\pm 1\) by congruence conditions modulo \(f_0^{n-1}\). As application, the author provides bounds on the average size of the \(2\)-class group in families of number fields defined by binary forms of degree \(n\geq 3\) using a uniform approach that also includes the non-monic even degree case. Other applications (to appear in forthcoming papers by the author) include a proof that most integral odd-degree binary forms fail to primitively represent a square because of the Brauer-Manin obstruction ([the author, ``Most integral odd-degree binary forms fail to properly represent a square'', Preprint, \url{arXiv:1910.12409}]) and (jointly with Bhargava and Shankar) a bound on the second moment of the size of the \(2\)-Selmer group of elliptic curves [the author et al., ``The second moment of the size of the \(2\)-Selmer group of elliptic curves'', Preprint, \url{arXiv:2110.09063}]). The introduction of the present paper provides a quite readable account on the history of the questions treated herein, on the methods used in their study and on the manifold applications of the results.