Torsion points, Pell's equation, and integration in elementary terms (Q2663756)

The paper under review deals generally with local-global principles. In particular, it shows that if a given phenomenon does not happen globally, then it does not happen locally, except for finitely many exceptions. This is the case in particular for the solubility of Pell's equations, or for the elementary integrability of algebraic functions. More precisely, Theorem 1.1 of the paper under review studies the Pell equation \(A^2 - D B^2 = 1\) over the ring \(\overline{\mathbb{Q}}(C)[x]\), where \(C\) is a fixed base curve defined over \(\overline{\mathbb{Q}}\) and \(D \in \overline{\mathbb{Q}}(C)[x]\) is fixed. If \(\deg(D) = 2 g + 2 \geq 6\), then the equation \(y^2 = D(x)\) defines a curve over \(\overline{\mathbb{Q}}(C)\), whose Jacobian will be denoted by \(J_D\), with two distinguished points at infinity, whose difference yields a point \(P_D\) inside \(J_D\). One knows that the aforementioned Pell equation is solvable, for \(A,B \in \overline{\mathbb{Q}}(C)[x]\), if and only if the point \(P_D\) is torsion. On the other hand, one can also specialize the previous Pell equation by evaluating the coefficients of \(D\) at any given point \(\mathbf{c} \in C(\mathbb{C})\). The authors show in Theorem 1.1 that only finitely many of these specialized Pell equations are solvable, if \(P_D\) is not torsion (i.e. if the global Pell equation is not solvable) and no positive integer multiple of \(P_D\) is contained in an elliptic curve \(E \subseteq J_D\). Indeed, if the latter happens, the authors prove that either infinitely many of the specialized Pell equations are solvable, or none of them are. This second case occurs precisely when there exists an isogeny \(\iota \colon E \to E_0\), where \(E_0\) is an elliptic curve defined over \(\overline{\mathbb{Q}}\) such that \(n P_D \in E\) and \(\iota(n P_D) \in E_0(\overline{\mathbb{Q}})\) is not torsion. Finally, if \(P_D\) is torsion on \(J_D\), then the localized Pell equation is solvable for all but finitely many \(\mathbf{c} \in C(\mathbb{C})\). One of the main ingredients in the proof of Theorem 1.1 is Theorem 1.7, which proves a special case of the relative Manin-Mumford problem. More precisely, the authors prove that for every variety \(X\) defined over \(\mathbb{C}\), and every relative abelian surface \(\mathcal{A} \to X\) with an irreducible closed curve \(\mathcal{V} \hookrightarrow \mathcal{A}\), the intersection of \(\mathcal{V}\) with the set \(\mathcal{A}^{[2]}\) of all torsion points in the fibers of \(\mathcal{A} \to X\), is contained in a finite union of abelian subschemes of \(\mathcal{A}\) of positive codimension. Then, one gets Theorem 1.1 out of Theorem 1.7 using techniques similar to the ones already appearing in [\textit{D. Masser} and \textit{U. Zannier}, J. Eur. Math. Soc. (JEMS) 17, No. 9, 2379--2416 (2015; Zbl 1328.11068)]. The other main result of the paper under review is Theorem 1.3, which deals with integration in elementary terms. More precisely, the authors take a base curve \(C\) defined over \(\overline{\mathbb{Q}}\), and an irreducible, smooth curve \(X\) defined over \(\mathbb{K} := \overline{\mathbb{Q}}(C)\) with a fixed function \(x \in \mathbb{K}\), which makes \(\mathbb{K}\) into a differential field, whose derivation operator is given by \(\delta := d/dx\). As was the case for Pell's equation, one can also specialize this global situation to local ones, corresponding to the choice of a point \(\mathbf{c} \in C(\mathbb{C})\). Then, the authors prove that, if \(f\) is not integrable in elementary terms, then the specialization of \(f\) is also not integrable in elementary terms for all but at most finitely many \(\mathbf{c} \in C(\mathbb{C})\), unless \(f\) belongs to a special class of functions, that the authors call ``elusive''. This is a technical definition, which is explained in detail in Section 16, and gives rise to a class of functions which are not integrable in elementary terms, but admit infinitely many specializations which are. This disproves a conjecture of [\textit{J. H. Davenport}, On the integration of algebraic functions. Berlin Heidelberg New York: Springer-Verlag (1981; Zbl 0471.14009)]. Note however that these specializations are quite sparse. For instance, Proposition 1.4 specifies that there are only finitely many points \(\mathbf{c} \in C(\overline{\mathbb{Q}})\) of bounded degree, for which the corresponding specialization of \(f\) is integrable in elementary terms. The connection with Pell's equation is made by Corollary 1.2, which shows that Theorem 1.1 implies that the specializations of functions of the form \(E/\sqrt{D}\), with \(E \in \mathbb{C}[x] \setminus \{0\}\), are not integrable in elementary terms, for all but finitely many \(\mathbf{c} \in C(\mathbb{C})\), if \(P_D\) is not a torsion point of \(J_D\). The paper under review provides thus an important achievement in problems of Manin-Mumford type, and their applications to Pell's equations and integrability questions. Its study is thus highly recommended to everyone who wants to be up to date with the finest achievements in the field.