Rational points of bounded height on projective surfaces (Q2475965)

Let \(X\subset\mathbb{P}^n\) be a geometrically integral projective surface over \(\mathbb{Q}\), of degree \(d\); and let \(N(B)\) count the number of rational points on \(X\) of height at most \(B\), not lying on any line contained in \(X\). The reviewer [Ann. Math. (2) 155, No. 2, 553--598 (2002; Zbl 1039.11044)] established bounds of the form \(N(B)= O_d(B^{2-\delta})\) (for the case \(n= 3\)), which hold uniformly in \(X\). These were improved by the author [Ann. Sci. Éc. Norm. Supér. 38, No. 1, 93--115 (2005; Zbl 1110.14020)]. The present paper shows that \(N(B)= O_{n,d,\theta}(B^\theta)\) for \[ \theta>{2\over 3}+{73\over 36\sqrt{3}}\;(d= 3),\quad \theta> {1\over 2}+ {9\over 4\sqrt{d}}\quad (4\leq d\leq 2), \] and \(\theta> 1\) \((d\geq 21)\). The exponent is best possible for \(d\geq 21\), since we can have \(N(B)\gg B\) when the surface contains a conic. As an example, these results give new bounds for the number of non-diagonal solutions to the Diophantine equation \(x^d_1+ x^d_2= x^d_2+ x^d_4\). In recent work, yet to appear, the author has improved these results further. The author uses a development of the reviewer's ``determinant method'' (loc. cit.), which leads one to consider points which lie on low degree curves on \(X\). When \(X\) is non-singular the number of such curves is bounded, but in general there may be infinite families. One new idea of particular interest introduced in this paper, is to consider the height (in the Hilbert scheme) of a family of conics on \(X\), and to use the fact that curves of large height have relatively few points. In this way it is shown that, even though the argument may produce a large number of conics, the total contribution is essentially no larger then that from a single conic.