Diophantine Approximation of Ternary Linear Forms

From MaRDI portal





The paper gives an efficient method for finding arbitrarily many solutions in integers \(x,y,z\) of the Diophantine inequality \[ \vert x+\alpha y + \beta z\vert \max(y^2, z^2 ) < c, \] where the numbers \(1,\alpha,\beta\) form an integral basis for a totally real cubic field over the rationals. If \(c\) is small enough, the method generates all solutions of the inequality. The algorithm introduced in the paper has features similar to those of the simple continued fraction algorithm for a single real number, and so can be used to obtain detailed information about the solutions of the Diophantine inequality under consideration. For example, in another paper [``Diophantine approximation of ternary linear forms. II. Math. Comput. 26, 977--993 (1972; Zbl 0258.10016)] the algorithm is used in the exact evaluation of the constant \(\limsup_{N\to\infty} \min M^2 \vert x + \theta y + \theta^2 z\vert\), where the \(\min\) is taken over all integers \(x,y,z\) satisfying \(1\le \max(\vert y\vert,\vert z\vert)\le M\) and where \(\theta\) denotes the positive root of the equation \(x^3 + x^2 - 2x - 1= 0\), that is, \(\theta = 2 \cos(2\pi/7)\); the value of the constant is \((2\theta + 3)/7 \approx .78485\). The interest of constants of this type in the theory of Diophantine approximation is discussed in a paper of \textit{H. Davenport} and \textit{W. M. Schmidt} [Symp. Math. 4, 113--132 (1970; Zbl 0226.10032)], where estimates for the particular constant just mentioned are given. The method of the paper under review can be extended to linear forms in \(n\) variables with coefficients \(1, \alpha_1, \ldots, \alpha_{n-1}\) which form a basis (not necessarily integral) for a real (not necessarily totally real) number field of degree \(n\).











This page was built for publication: Diophantine Approximation of Ternary Linear Forms

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5620661)