The density of integer points on homogeneous varieties (Q2266039)

Let \(\mathcal F\) be a set of \(r\) rational forms of odd degrees \(\leq k\) in \(s\) variables. It was shown by \textit{B. J. Birch} [Mathematika 4, 102--105 (1957; Zbl 0081.04501)], by the Brauer reduction method, that there exists \(c_1(k,r)\) such that \(\mathcal F\) has a non-trivial rational zero whenever \(s\geq c_1\). The principal object of this paper is to prove this result by the Hardy-Littlewood circle method. In fact it is shown that \(Z_P({\mathcal F})\gg P^{s-c_2}\) for \(s\geq c_2(k,r)\), where \(Z_P(\mathcal F)\) denotes the number of common integer zeros \((x_1,\ldots, x_s)\) of the forms \(\mathcal F\) satisfying \(| x_i| \leq P\). The paper allows estimates for \(c_2(k,r)\) in principle, and these would probably be better than those obtained for \(c_1\) by Birch's method. However a reasonable bound for \(c_2(5,1)\), say, (i.e. a single quintic form) still seems out of reach, partly because the corresponding \(p\)-adic problem has not been settled for small \(p\). It should be noted that for systems where the singular locus is not too large, the analytic method of \textit{B. J. Birch} [Proc. R. Soc. Lond., Ser. A 265, 245--263 (1962; Zbl 0103.03102)] applies. In particular, if \(\mathcal F\) consists of a single non-singular form with zeros in every \(p\)-adic field, then we may take \(c_2=(k-1)2^k\). The paper follows \textit{H. Davenport} [ibid. 272, 285--303 (1963; Zbl 0107.04102)] and \textit{H. Davenport} and \textit{D. J. Lewis} [Am. J. Math. 84, 649--665 (1962; Zbl 0118.28103)] in using the invariant \(h(F)\) of the rational form \(F\), defined as the least \(h\) for which one can write \(F=G_1H_1+G_2H_2+\ldots+G_hH_h\), with rational forms \(G_i\), \(H_i\) of positive degree. The paper starts with a fairly routine presentation of the circle method, and proceeds to the estimation of multiple exponential sums via Weyl's inequality. The final section of the paper, which is the most novel and difficult, relates \(h(F)\) to the density of integer solutions of certain multilinear forms associated with \(F\). These multilinear forms are familiar from Davenport's work (loc. cit.) on the cubic case, and the kernel of the problem is the extension of his results to higher degrees. In fact a totally different method is required. In addition to the main theorem stated above, other more quantitative results are given. Moreover it is shown that one can allow forms of even degree in \(\mathcal F\) under certain circumstances. Finally an estimate for multiple exponential sums, which arises in the course of the work, is given. This important paper makes the first progress in the application of analytic methods to general systems of forms. It is to be hoped that, in time, the techniques begun here can be refined to yield ``realistic'' quantitative results.