Quadratic diophantine inequalities (Q5942984)

The theme of this paper is to investigate certain systems of Diophantine inequalities on real diagonal quadratic forms. First, let \(Q_1\) and \(Q_2\) be real diagonal quadratic forms in \(s\) variables, with \(s\geq 10\), and suppose that whenever \(\alpha\) and \(\beta\) are real numbers with \((\alpha,\beta)\neq(0,0)\), then the form \(\alpha Q_1+\beta Q_2\) has at least five nonzero coefficients, at least one irrational coefficient, at least one negative coefficient, and at least one positive coefficient. It is proved here that for any \(\varepsilon>0\), there exists a nonzero integral vector \({\mathbf x}\in{\mathbb Z}^s\) such that one has \(|Q_1({\mathbf x})|<\varepsilon\) and \(|Q_2({\mathbf x})|<\varepsilon\). The above condition that some coefficient of the form \(\alpha Q_1+\beta Q_2\) is irrational may be regarded as a natural extension of the requirement adopted in the work of Margulis on the corresponding research of single quadratic form. This condition is, moreover, weaker than the assumption of the similar theorem of \textit{R. J. Cook} for nine variable quadratic forms [Acta Arith. 25, 337-346 (1974; Zbl 0278.10021)]. In fact, this paper contains a more general theorem which implies the above result. Let \(R\geq 2\), and \(Q_1,\dots,Q_R\) be \(R\) real diagonal quadratic forms in \(s\) variables, and write \(Q_i({\mathbf x})=\lambda_{i1}x_1^2+\cdots+\lambda_{is}x_s^2\). Let \({\mathbf v}_j\) be the column vector of the coefficients of \(x_j^2\), that is, the transposition of \([\lambda_{1j},\dots,\lambda_{Rj}]\). For every subset \(J\) of \(\{1,\dots,s\}\), define \(A_J\) to be the matrix consisting of the columns \({\mathbf v}_j\) with \(j\in J\), that is, when \(J=\{j_1,\dots,j_l\}\), we have \(A_J=[{\mathbf v}_{j_1},\dots, {\mathbf v}_{j_l}]\). And write \(r(A_J)\) for the rank of \(A_J\). Now suppose that \(s\geq 5R\), and that the following three conditions are satisfied: (i) there exists a real nonsingular solution of the system \(Q_1({\mathbf x})=\cdots=Q_R({\mathbf x})=0\) (``nonsingular'' means that the matrix composed of \(\partial Q_i/\partial x_j\) \((1\leq i\leq R\), \(1\leq j\leq s)\) is of full rank), (ii) For every subset \(J\subseteq\{1,\dots,s\}\), one has \(|J|\leq s-5(R-r(A_J))\), (iii) Whenever \(\beta_1,\dots,\beta_R\) are real numbers not all zero, then at least one coefficient of \(\beta_1Q_1+\cdots+\beta_RQ_R\) is irrational. Under these assumptions, the main theorem asserts that for any \(\varepsilon>0\), there exists a nonzero integral vector \({\mathbf x}\in{\mathbb Z}^s\) such that \(|Q_i({\mathbf x})|<\varepsilon\) for all \(i\) with \(1\leq i\leq R\). It may be said, in short, that the proof is constructed by combining the Davenport-Heilbronn method with the ideas of recent work of \textit{V. Bentkus} and \textit{F. Götze} [Ann. Math. (2) 150, 977-1027 (1999; Zbl 0979.11048)].