An exceptional isomorphism between level 13 modular curves via Torelli's theorem (Q2255329)

Let \(C^+_{\text{s}} (13)\), respectively \(C^+_{\text{ns}} (13)\), be the normalizer of a split, respectively non-split, Cartan subgroup of \(\mathrm{GL}_2 (\mathbb{Z} / 13 \mathbb{Z})\), and consider the corresponding (geometrically connected) quotients \(X_{\text{s}} (13)\) and \(X_{\text{ns}} (13)\) of the modular curve \(X (13)\) with full level \(13\) structure, as studied in Serre's uniformity conjecture. The author shows that these curves are isomorphic over \(\mathbb{Q}\). This statement is proved by an ingenious detour through the Jacobians \(J_{\text{s}} (13)\) and \(J_{\text{ns}} (13)\) of these curves; it is shown that there exists an isomorphism between these abelian varieties that respects their canonical principal polarizations, after which Torelli's theorem is invoked. The details of the proof are as follows. First the existence of an isogeny \(J_{\text{s}} (13) \sim J_{\text{ns}} (13)\) is derived by building on results from [\textit{I. Chen}, Proc. Lond. Math. Soc., III. Ser. 77, No. 1, 1--38 (1998; Zbl 0903.11019)] and [\textit{B. de Smit} and \textit{B. Edixhoven}, Math. Res. Lett. 7, No. 2--3, 147--153 (2000; Zbl 0968.14024)]. Moreover, it is shown that for both of these curves the endomorphism ring is generated by Hecke correspondences and isomorphic to the ring of integers \(\mathcal{O}_K\) of the maximal totally real subfield \(K = \mathbb{Q} (\zeta_7)^+\) of \(\mathbb{Q} (\zeta_7)\), and that the isogeny between them can be taken to be compatible with the respective Hecke actions, or in other words, to be \(\mathcal{O}_K\)-linear. The author then studies the polarizations on the curves in question via complex analytic theory to show that it suffices to refine this \(\mathcal{O}_K\)-linear isogeny into an \(\mathcal{O}_K\)-linear isomorphism, since she shows that such an isomorphism, in turn, can be refined to respect the canonical principal polarizations. Following this, it is shown that the \(\mathcal{O}_K\)-linear isogeny in question can be taken to have degree coprime to any specified finite set of primes that do not divide \(12\). This is a corollary of a more general statement on the existence of similar isogenies \(J_{\text{s}} (p) \sim J_{\text{ns}} (p) \times J_0 (p)\) (coprime with a given finite set of primes not dividing \(p - 1\)) for arbitary \(p\). After this, it is shown that if we choose such an isogeny \(J_{\text{s}} (13) \to J_{\text{ns}} (13)\) of degree coprime to both \(7\) and \(13\), then its kernel is of the form \(J_{\text{s}} [I]\) for an ideal \(I\) of \(\mathcal{O}_K\). Since the class group of \(\mathcal{O}_K\) is trivial, this statement follows by induction from a study of the Galois modules \(J_{\text{s}} [\mathfrak{p}]\), which are shown to be irreducible as long as \(\mathfrak{p}\) does not divide \(7\) or \(13\). Another argument using the triviality of the class group of \(\mathcal{O}_K\) then allows us to obtain the desired refinement of the chosen \(\mathcal{O}_K\)-linear isogeny into an \(\mathcal{O}_K\)-linear isomorphism, which by the above proves the main theorem. It should be noted that the isomorphism found is unlikely to have a simple modular explanation, because \(X_{\text{s}} (13)\) has a \(\mathbb{Q}\)-rational cusp, whereas \(X_{\text{ns}} (13)\) does not. This makes the result all the more remarkable.