Small zeros of quadratic congruences modulo \(pq\). II (Q1345283): Difference between revisions

Let \(Q({\mathbf x})\) be an integral quadratic form in \(n\) variables. It is conjectured that if \(n\geq 4\), the congruence \[ Q({\mathbf x}) \equiv 0\pmod m \tag \(*\) \] has a non-zero integral solution with \(\max|x_i|\ll m^{1/2+ \varepsilon}\) for any fixed \(\varepsilon> 0\), or even with \(\max |x_i |\ll m^{1/2}\). Bounds of the first type have been established for \(m\) prime, or a product of two primes, by the reviewer [Glasg. Math. J. 27, 87-93 (1985; Zbl 0581.10008)]\ and [Mathematika 38, 264-284 (1991; Zbl 0739.11014)]\ respectively. The second of these results improves upon the author's bounds in the first paper of this series [Mathematika 37, 261-272 (1990; Zbl 0713.11030)]. In the case in which \(m\) is prime the exponent \(\varepsilon\) was removed by the author [J. Number Theory 37, 92-99 (1991; Zbl 0713.11031)]. The present paper now does the same for the case in which \(m\) is a product of two primes. As in the reviewer's work, a slightly more general result is established. Thus if \(B\) is any box, centred on the origin, with sides parallel to the coordinate axes, then \((*)\) has a non-trivial solution in \(B\), providing that \(B\) contains more than \(2^{50n} m^{n/2}\) integral points. The proof is an amalgam of methods from the reviewer's work on the case \(m= pq\) and the author's paper on the case \(m=p\).