Quadratic forms and systems of forms in many variables (Q1656698): Difference between revisions

Let \(F_1(\mathbf{x}),\ldots,F_R(\mathbf{x})\) be forms of degree \(d\ge 2\), defined over \(\mathbb{Z}\). The famous theorem of \textit{B. J. Birch} [Proc. R. Soc. Lond., Ser. A 265, 245--263 (1962; Zbl 0103.03102)] gives the standard Hardy-Littlewood asymptotic formula for the counting function for rational points on the variety \(V:\,F_1=\ldots=F_R=0\), under the condition \[ n>\dim(S)+1+(d-1)2^{d-1}R(R+1). \] Here \(S\) is the (projective) ``Birch singular locus''. In particular one requires \(n>(d-1)2^{d-1}R(R+1)\) when \(V\) is smooth. The purpose of the present paper is to obtain a condition in which the lower bound is linear in \(R\) rather than quadratic. In the case \(d=2\), that is to say for systems of quadratic forms, the paper establishes the Hardy-Littlewood asymptotic formula when \[ n>\dim(S)+1+8R,\tag{\(\ast\)} \] recovering Birch's result when \(R=3\) and improving it as soon as \(R\ge 4\). Given the central role that Birch's theorem plays in the analytic approach to Diophantine geometry, this is a quite remarkable achievement. In addition to handling the case \(d=2\) the paper lays the groundwork for future work on forms of degree \(d\ge 3\). In fact the paper gives a slight sharpening of \((\ast)\), in which \(\dim(S)\) is replaced by the maximum of \(\dim\mathrm{Sing}(W)\), as \(W\) varies over all hypersurfaces defined by real linear combinations of \(F_1,\dots,F_R\). This is very much in the spirit of the papers by \textit{R. Dietmann} [Q. J. Math. 66, No. 1, 97--110 (2015; Zbl 1375.11034)] and \textit{D. Schindler} [Fields Inst. Commun. 77, 207--218 (2015; Zbl 1375.11035)], which relate to the original Birch theorem. The argument builds on ideas of \textit{V. Bentkus} and \textit{F. Götze} [Acta Arith. 80, No. 2, 101--125 (1997; Zbl 0871.11069); Ann. Math. (2) 150, No. 3, 977--1027 (1999; Zbl 0979.11048)] and of \textit{W. Müller} [Monatsh. Math. 153, No. 3, 233--250 (2008; Zbl 1184.11011)]. The key idea is that points where the generating function is large must either be very close, or must be well-spaced. This allows one to control the minor arc contribution, without requiring any reference to the height of rational approximations to the argument of the generating function. In this respect the philosophy is very different from that of other circle method papers. In order to push this approach through one needs to study systems of multilinear inequalities, replacing the multilinear equations occurring in Birch's analysis.