Solving algebraic equations in roots of unity.
This paper had its origin in the following problem: Let \(f(x,y)\) be an irreducible polynomial with complex coefficients of degree \(d_1\) in \(x\) and \(d_2\) in \(y\). What can we say about the solutions of \(f(\zeta_1,\zeta_2)=0\) where \(\zeta_1\) and \(\zeta_2\) are roots of unity? Ihara, Serre and Tate [cf. \textit{S. Lang}, Ann. Mat. Pura Appl. (4) 70, 229--234 (1965; Zbl 0151.27401)] have shown that infinitely many such solutions exist if and only if \(f\) has the form \(x^{d_1} - \zeta y^{d_2}\) or \(x^{d_1}y^{d_2}-\zeta\) where \(\zeta\) is a root of unity. What can be said if there are only finitely many solutions? The roots of unity are the torsion points of the group \(\mathbb{C}^\times\) which we can identify with the algebraic group \(G_m\). The \(n\)-tuples of roots of unity \((\zeta_1,\ldots,\zeta_n)\) are then just the torsion points of \(G_m\times\cdots\times G_m=(G_m)^n\). Let \(V\) be an irreducible subvariety of \((G_m)^n\) and denote by \(V_{\text{tor}}\) the Zariski closure of the torsion points which are contained in \(V\). The problem is then to investigate \(V_{\text{tor}}\) and its relationship to \(V\). The irreducible components of \(V_{\text{tor}}\) satisfy \(W=W_{\text{tor}}\). We remark that the problem was also studied when the torsion points are replaced by a group \(\Gamma\) such that \(\text{rk}_\mathbb{Q}(\Gamma\otimes\mathbb{Q})<\infty\) and in the case when \((G_m)^n\) is replaced by an abelian variety. Suppose that \(V\subseteq(G_m)^n\) is irreducible of dimension \(d\) and satisfies \(V=V_{\text{tor}}\). A conjecture by \textit{S. Lang} [cf. ``Fundamentals of diophantine geometry. New York etc.: Springer-Verlag (1983; Zbl 0528.14013); p. 220] implies that \(V\) should be the translation of a subgroup of \((G_m)^n\) by a torsion point. This was proved (in even greater generality) by \textit{M. Laurent} [C. R. Acad. Sci., Paris, Sér. I 296, 945--947 (1983; Zbl 0533.10011)]. In section 1 we give a direct proof of this result by showing that there exist \(a_{ij}\in\mathbb{Z}\) and roots of unity \((\zeta_1,\ldots,\zeta_n)\) such that \(\varphi:(G_m)^d\to (G_m)^n\) with \(\varphi(t_1,\ldots,t_d)=(\zeta_1\prod t_ i^{a_{i1}},\dots,\zeta_ n\prod t_i^{a_{in}})\) induces an isomorphism \((G_n)^d\cong V\). The second section deals with equations \(f(x,y)=0\). The basic result is that the torsion points of \(f=0\) lie also on curves which are associated with \(f\). This means that the torsion points can be effectively determined. Bézout's theorem then gives that the number of torsion points is bounded. --- The third section studies in detail curves of degre (1,1) in \(G_ m\times G_ m\). The basic result is that modulo isomorphism there are exactly two curves of degree (1,1) with 14 torsion points. All other curves have \(\leq 12\) torsion points or infinitely many. The results depend heavily on calculations with the computer algebra system MAPLE. If \(C\) is a curve in \((G_m)^n\) of degree \((d_1, \ldots, d_n)\) with only finitely many torsion points (and \(d_i\ge 1)\) then we show in section 4 that the number of torsion points is \(\le 22\min(d_i)\max(d_i)\). This is done by induction on \(n\) and the results of section 2. Examples are given that indicate that the bound is reasonable. Then we look at surfaces \(S\subseteq G_m\times G_m\times G_m\). For technical reasons we restrict us to `nice' surfaces. The constructive methods for curves can be directly translated to give results for surfaces. \(S_{\text{tor}}\) consists of isolated points and of curves. We give estimates for the number and the degree of curves in \(S_{\text{tor}}\) and also for the isolated torsion points. --- Finally section 6 deals with varieties of higher dimensions. We present an algorithm how \(V_{\text{tor}}\) can be effectively determined if \(V\) is given (at least in principle). Then we make a conjecture for varieties \(V\) of codimension 1 in \((G_ m)^ n\): let \(V_{\text{tor},i}\) be the \(i\)-dimensional part of \(V_{\text{tor}}\) and denote by \([V]\) the class of the Zariski closure of \(V\) in \((\mathbb{P}^1)^n\). Then the conjecture says: there are absolute constants \(c_i\) (independent of \(n)\) such that \([V_{\text{tor},n-i}]\le c_i[V]^i\) in the Chow ring of \((\mathbb{P}^1)^n\). (The conjecture is true for \(i=1\) and \(i=2\) with \(c_1=1\) and some \(c_2\) with \(7\le c_2\le 11\).) We conclude with some examples and remarks that support the conjecture.
- Betti numbers of congruence groups. (Appendix: On representations of compact \(p\)-adic groups by Ze'ev Rudnick)
- Limit distribution of small points on algebraic tori
- Roots algebraic equations and Clifford algebra
- The number of maximal torsion cosets in subvarieties of tori
- DIFFERENTIALS OF THE FAMILYY2−X3−tX2
- Solving algebraic equations in roots of unity
- Computing the torsion points of a variety defined by lacunary polynomials
- Torsion Points on an Algebraic Subset of an Affine Torus
- scientific article; zbMATH DE number 1014500 (Why is no real title available?)
- Uniform bound for generic points of curves in tori.
- Cyclotomic points and algebraic properties of polygon diagonals
- Efficiently detecting torsion points and subtori
- Root Locus of Algebraic Equations
- Good reduction of certain covers \(\mathbb{P}^1\to\mathbb{P}^1\)
- Singular intersections of subgroups and character varieties
- Khovanskii-finite rational curves of arithmetic genus 2
- A bound for the torsion on subvarieties of abelian varieties
- Fractal uncertainty for discrete two-dimensional Cantor sets
- Small points on subvarieties of a torus
This page was built for publication: Solving algebraic equations in roots of unity.
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4025265)
