Calculating ``small'' solutions of inhomogeneous relative Thue inequalities (Q2075017)

A relative Thue equation is one of the form \(N_{K/M}(X-\alpha Y)=m\), where: \(K/M\) is a (finite) extension of number-fields, \(N_{K/M}(\cdot)\) is the norm relative to this extension, \(X,Y\in\mathbb{Z}_M\) (= ring of integers of \(M\)), \(K=M(\alpha)\) with \(\alpha\) is an algebraic number of degree \(\geq 3\) over \(M\) and \(m\in\mathbb{Z}_M\). Such equations appeared in Siegel's work as early as 1921. An inhomogeneous relative Thue equation has the form \(N_{K/M}(X-\alpha Y+\lambda)=m\), where \(M,K,\alpha, m\) are as above and \(\lambda\in\mathbb{Z}_K\) may be either a variable or (as in this paper) fixed. In the present paper the author is interested in effectively solving inhomogeneous relative Thue inequalities (obviously this study includes that of inhomogeneous relative Thue equations) which have the form \[ |N_{K/M}(X-\alpha Y+\lambda)| \leq c_0Z^k, \quad Z=\max(\operatorname{house}{X},\operatorname{house}{Y})\leq Z_0 \] where, for an algebraic number \(\gamma\), the symbol ``\(\operatorname{house}{\gamma}\)'' denotes the maximum of the absolute values of the conjugates of \(\gamma\), \(c_0\) is a given positive constant, \(k\geq 0\) is a fixed positive integer and \(Z_0\) is a given huge number, say of the size of \(10^{100}\). Actually, the author considers an even more general problem, special case of which is the above mentioned problem, namely, the following: Let \(\alpha_1,\ldots,\alpha_n\) be given non-zero distinct complex numbers and \(\lambda_1,\ldots,\lambda_n\) any given complex numbers with \(\max_j |\lambda_j| \leq c_{\lambda}\) (a given positive constant). Let also \(c_0>\) be a given constant and \(k\geq 0\) a given integer. Then, the author develops a method for the explicit solution of the inequality \[ \left|\prod_{j=1}^n(X-\alpha_jY+\lambda_j)\right| \leq c_0Z^k, \quad X,Y\in\mathbb{Z}_M,\; Z=\max(\operatorname{house}{X}\,,\operatorname{house}{Y})\leq Z_0. \] His method is based on theorems proved in this paper (too technical to be reproduced in this review) and techniques developed in earlier works of him. Two explicit examples are discussed. First example. Let \(\alpha_i\), \(i=1,\ldots,9\) the roots of \(x^9-9x^7+24x^5-2x^4-20x^3+3x^2+5x-1\) which is viewed as a polynomial over \(M=\mathbb{Q}(\sqrt{2})\), and \(\lambda_i=\alpha_i^2+2\alpha_i\) (\(1\leq i\leq 9\)). All solutions of \[ \left|\prod_{j=1}^9(X-\alpha_jY+\lambda_j)\right| \leq10 \quad X,Y\in\mathbb{Z}_M,\; \max(\operatorname{house}{X}\,,\operatorname{house}{Y})\leq 10^{100}. \] are computed; it turns out that they are totally 138. Second example. Let \(Μ=\mathbb{Q}(\sqrt{-2})\), \(f(t)=t^5-t^4-4t^3+3t^2+3t-1\in M[t]\). The problem is to compute all polynomials \(g(t)=t^2-Yt+X\in\mathbb{Z}_M[t]\) with \(\max(\operatorname{house}{X}\,,\operatorname{house}{Y})\leq 10^{100}\) and \(|\mathrm{Res}_M(f,g)|\leq 25\). According to this paper's arguments, this problem is reduced to solving \(|N_{K/M}(X-\alpha Y+\alpha^2)|\leq 25\) in \(X,Y\) as above, where \(K=\mathbb{Q}(\alpha,\sqrt{-2})\) and \(f(\alpha)=0\). According to the author's computations, there are exactly 39 solutions of the last inequality. Remark. A misprint in the title ``How to apply Lemma 3?'' of Subsection 3.4: ``Lemma'' should be replaced by ``Theorem''.