Towards a classification of entanglements of Galois representations attached to elliptic curves (Q6172744)

The article under review deals with the problem of classifying the possible entanglements between division fields of elliptic curves, which is related to Mazur's celebrated \textit{Program B}, whose aim is to classify the possible images of the adelic Galois representations attached to elliptic curves defined over a given number field. In particular, the paper under review classifies so-called abelian entanglements in four possible types, plus a set of exceptions which can be understood in terms of three explicit infinite families of \(j\)-invariants, and also an unexplicited finite set of \(j\)-invariants. Let us be more precise. If \(E\) is an elliptic curve defined over a number field \(F\), the absolute Galois group \(G_F\) acts on the torsion points \(E_{\text{tors}} := E(\overline{F})_{\text{tors}}\), yielding a Galois representation \(\rho_E \colon G_F \to \mathrm{Aut}_{\mathbb{Z}}(E_{\text{tors}}) \cong \mathrm{GL}_2(\widehat{\mathbb{Z}})\). A celebrated theorem of \textit{J. P. Serre}'s [Invent. Math. 15, 259--331 (1972; Zbl 0235.14012)] shows that \(\rho_E(G_F)\) is open inside \(\mathrm{Aut}_{\mathbb{Z}}(E_{\text{tors}})\) (endowed with the profinite topology coming from the isomorphism with \(\mathrm{GL}_2(\widehat{\mathbb{Z}})\)) whenever \(E\) does not have extra endomorphisms, \textit{i.e.} if \(\mathrm{End}_{\overline{F}}(E) \cong \mathbb{Z}\). However, describing explicitly the possible images \(\rho_E(G_F)\), when \(E\) varies among all possible elliptic curves defined over \(F\), is a challenging question, which was firstly posed by \textit{B. Mazur} [Lect. Notes Math. 601, 107--148 (1977; Zbl 0357.14005)]. In general, the representation \(\rho_E\) can fail to be surjective for two reasons. First of all, because one of the \(\ell\)-adic representations \(\rho_{E,\ell^\infty} \colon G_F \to \mathrm{Aut}_{\mathbb{Z}}(E[\ell^\infty]) \cong \mathrm{GL}_2(\mathbb{Z}_\ell)\) is not surjective, or because the different \(\ell\)-adic representations are \textit{entangled}. This means that the adelic image \(\rho_E(G_F)\) is strictly smaller than the product of \(\ell\)-adic images \(\prod_\ell \rho_{E.\ell^\infty}(G_F)\). In other words, if one considers the division fields \(F(E[n])\), this means that there exist \(a,b \in \mathbb{N}\) such that \(F(E[\gcd(a,b)]) \subsetneqq F(E[a]) \cap F(E[b])\). In this case, the Galois group \(T_{a,b} := \mathrm{Gal}(F(E[a]) \cap F(E[b])/F(E[\gcd(a,b)]))\) is the type of entanglement that we are witnessing, following a terminology introduced recently by the first and last authors of the paper under review [Trans. Am. Math. Soc., Ser. B 9, 827--858 (2022; Zbl 1517.11059)]. Sometimes, the entanglement occurs already when one intersects the division fields with the maximal abelian extension \(F^{\text{ab}}\). In this case, we say that \(E\) has abelian entanglements, of types \(S_{a,b} := \mathrm{Gal}(F(E[a]) \cap F(E[b]) \cap F^{\text{ab}}/F(E[\gcd(a,b)]) \cap F^{\text{ab}})\). When \(F = \mathbb{Q}\), the paper under review classifies these abelian entanglements in four families. First of all, if the discriminant \(\Delta_E\) is odd and is not a square, \(\mathbb{Q}(E[2])\) is entangled with the field \(\mathbb{Q}(E[\Delta_E])\), because both contain the quadratic extension \(\mathbb{Q}(\sqrt{\Delta_E})\). This notion can be generalized to the cases when \(\Delta_E\) is even and still not a square, as the authors of the paper under review do in Definition 3.7. On the other hand, if \(\Delta_E\) is a square then the \(2\)-adic image \(\rho_{E,2^\infty}(G_\mathbb{Q})\) is not maximal, because the aforementioned quadratic extension \(\mathbb{Q}(\sqrt{\Delta_E})\) collapses. This type of behaviour was already noticed by Serre, and implies that the index of \(\rho_E(G_{\mathbb{Q}})\) inside \(\mathrm{Aut}_{\mathbb{Z}}(E_{\text{tors}})\) is always even. It could also happen that an elliptic curve has a \textit{Weil entanglement}, which means that \(\mathbb{Q}(E[a]) \cap \mathbb{Q}(\zeta_b)\) or \(\mathbb{Q}(E[b]) \cap \mathbb{Q}(\zeta_a)\) are non-trivial extensions of \(\mathbb{Q}(\zeta_d)\), where \(d = \gcd(a,b)\). The other possible type of entanglement for elliptic curves without complex multiplication that the authors introduce is the so-called \textit{fake CM entanglement}, which occurs when there exists a prime \(p\) such that the image of \(\rho_{E,p} \colon G_{\mathbb{Q}} \to \mathrm{Aut}_{\mathbb{Z}}(E[p]) \cong \mathrm{GL}_2(\mathbb{Z}/p \mathbb{Z})\) is contained in the normalizer of a Cartan subgroup. In this case, there are three subfields of \(\mathbb{Q}(E[p])\) which are quadratic over \(\mathbb{Q}\), given by \(\mathbb{Q}(\sqrt{(-1)^{(p-1)/2} p})\), \(\mathbb{Q}(\sqrt{d})\) and \(\mathbb{Q}(\sqrt{(-1)^{(p-1)/2} p d})\), for some squarefree \(d\). Then, one says that \(E\) has a fake CM entanglement if there exists a prime \(q \neq p\) such that the intersection \(\mathbb{Q}(E[p]) \cap \mathbb{Q}(E[q])\) equals either \(\mathbb{Q}(\sqrt{d})\) or \(\mathbb{Q}(\sqrt{(-1)^{(p-1)/2} p d})\). Finally, another ``trivial'' type of entanglement happens when \(E\) has complex multiplication by an order \(\mathcal{O}\) inside an imaginary quadratic field \(K\). Indeed, in this case \(K \subseteq \mathbb{Q}(E[n])\) for every \(n > 3\), which yields a non-trivial entanglement between the different division fields. The authors of the paper under review show in Theorem A that, if \(E\) does not have any of the aforementioned type of entanglements, but still has some non-trivial abelian entanglement, then \(j(E)\) belongs either to three explicit one-parameter families, which correspond to modular curves of genus zero, or to a finite set \(J\), which is undetermined because we cannot classify all the possible elliptic curves with a \(13\)-isogeny such that \(\Delta_E\) is a square and \(\mathbb{Q}(E[2]) \subseteq \mathbb{Q}(E[13])\). Moreover, they show in Theorem C that there are infinte families of elliptic curves that exhibit several types of Weil entanglement, and they partially generalize this result to higher dimensional abelian varieties in Theorem E. Finally, they investigate the various types of entanglement occurring in elliptic curves \(E\) with complex multiplication, and they compute in particular the indices of the images \(\rho_E(G_\mathbb{Q})\) inside the normalizer of the specific Cartan subgroup of \(\mathrm{GL}_2(\widehat{\mathbb{Z}})\) in which these images are naturally contained. Note that more general results in this direction have been found by \textit{F. Campagna} and the reviewer in two recent papers [Pac. J. Math. 317, No. 1, 21--66 (2022; Zbl 1509.11043); Contemp. Math. 779, 41--56 (2022; Zbl 1525.11060)]. To conclude, the paper under review provides some interesting new contributions to the classifications of the possible abelian entanglements in the family of division fields of elliptic curves defined over \(\mathbb{Q}\), which complement the study of non-abelian entanglements carried out by \textit{N. Jones} and \textit{K. McMurdy} [New York J. Math. 28, 182--229 (2022; Zbl 1497.11150)] for non-abelian entanglements. The results obtained are rather technical, because they rely on the study of several different types of modular curves.