Quadratic integral solutions to double Pell equations (Q661348): Difference between revisions

\textit{P. Corvaja} and \textit{U. Zannier} [J. Reine Angew. Math. 565, 27--42 (2003; Zbl 1153.11315)] proved a general result on quadratic integral points on algebraic curves which is as follows. Let \(\widetilde{C}\) be a projective, non-singular curve defined over a number field \(k\) and \(C=\widetilde{C}\setminus\{ P_1,\ldots , P_4\}\) with \(P_1,\ldots , P_4\in C(k)\) an open affine subset. Then there are finitely many rational maps \(\psi_1,\ldots ,\psi_s:\, \widetilde{C}\to {\mathbb P}^1\) of degree \(2\) such that \(C\) has only finitely many quadratic integral points outside the families \(\psi_i^{-1}({\mathbb P}^1(k))\) (\(i=1,\ldots , s\)). Their proof uses the Subspace Theorem. The author proves a more precise result for the affine curve \(C\) in \({\mathbb A}^3\) given by the two Pellian equations \(y^2=ax^2+c\), \(z^2=bx^2+d\). We may view \(C\) as \(\widetilde{C}\setminus\{ P_1,\ldots ,P_4\}\), where \(\widetilde{C}\) is the curve in \({\mathbb P}^3\) given by \(Y^2=aX^2+cW^2\), \(Z^2=bX^2+dW^2\), and \(P_1,\ldots ,P_4\) are the points with \(W=0\). Let \(k\) be a number field containing \(\sqrt{a}\), \(\sqrt{b}\), \(c\), \(d\), and \(S\) a finite set of places of \(k\) containing the infinite places and the prime ideals dividing \(cd(ad-bc)\). Notice that the maps \((x,y,z)\mapsto (x,y)\), \((x,y,z)\mapsto (y,z)\), \((x,y,z)\mapsto (x,z)\) map \(C\) to Pellian curves, each having infinitely many points with coordinates from \(O_S\), and each of the inverse images of these maps forms an infinite family of quadratic \(S\)-integral points on \(C\). The author shows that \(C\) has only finitely many quadratic \(S\)-integral points \((x,y,z)\) (i.e., with coefficients of the minimal polynomial over \(k\) in \(O_S\)) outside these families, and he gives an explicit upper bound for the number of these points. In his proof the author uses an idea of Vojta, specialized to the situation he is considering. Let \((x,y,z)\) be a quadratic \(S\)-integral point of \(C\), and \((x',y',z')\) its conjugate. Define \(C=(y-x\sqrt{a})(y'-x'\sqrt{a})/c\), \(A=(z-x\sqrt{b})(z'-x'\sqrt{b})/d\), \(A=(y\sqrt{b}+z\sqrt{a})(y'\sqrt{b}+z'\sqrt{a})/(ad-bc)\), \(\alpha =BC\), \(\beta = CA\), \(\gamma =AB\). Then \(\alpha ,\beta ,\gamma\) are \(S\)-units, and \(\alpha +\beta +\gamma =1\). This \(S\)-unit equation has only finitely many solutions with \(\alpha ,\beta ,\gamma\) all different from \(1\), and these lead to finitely many points on \(C\). The points for which at least one of \(\alpha ,\beta ,\gamma\) is \(1\) are either \(S\)-integral points on certain curves of genus \(1\), of which there are only finitely many, or points in the three infinite families mentioned above.