Heegner points and exceptional zeros of Garrett \(p\)-adic \(L\)-functions (Q2044722)

The aim of the paper under review is to provide evidence for a \(p\)-adic analogue of the equivariant Birch and Swinnerton-Dyer conjecture, formulated in a recent preprint by the same authors [\textit{M. Bertolini} et al., \url{https://www.esaga.uni-due.de/f/massimo.bertolini/publications/BSV-TPR.pdf}]). This is done by restricting to a situation where the corresponding \(L\)-function factors as a product of two Rankin-Selberg \(L\)-functions. This fact allows the authors to adapt the techniques of the third author's PhD thesis (see [\textit{R. Venerucci}, Invent. Math. 203, No. 3, 923--972 (2016; Zbl 1406.11060)] and [\textit{R. Venerucci}, Comment. Math. Helv. 91, No. 3, 397--444 (2016; Zbl 1406.11061)]) to prove that their conjecture holds true in this specific setting. Let us be more precise. First of all, recall that the classical conjecture of Birch and Swinnerton-Dyer relates the arithmetic of an elliptic curve \(A\) defined over a number field \(F\) to the analytic behaviour of its \(L\)-function \(L(A,s)\) around the point \(s = 1\). It is then interesting to see how the invariants appearing in this conjecture behave under base change. To do so, assume that \(F = \mathbb{Q}\), and take a finite extension \(K/\mathbb{Q}\). Then, the \(L\)-function of the base change \(A_K\) factors as products of \(L\)-functions of the form \(L(A,\rho;s)\), where \(\rho\) is an Artin representation. The order of vanishing at \(s = 1\) of these \(L\)-functions is conjectured to be related to the \(\rho\)-part of the group of rational points \(A(K)\) (see [\textit{D. E. Rohrlich}, in: Automorphic forms and analytic number theory, Proc. Conf., Montréal/Can, 1989, 123--133 (1990; Zbl 0737.11014)]). Moreover, their special values at \(s = 1\) are predicted by the specialization of the equivariant Tamagawa number conjecture at the motive \(H^1(A)(1)\) (see [\textit{D. Burns} and \textit{M. Flach}, Doc. Math. 6, 501--570 (2001; Zbl 1052.11077)]), or by more recent conjectures (see [\textit{V. Dokchitser} et al., J. Reine Angew. Math. 773, 199--230 (2021; Zbl 1485.11107)]). The recent preprint [loc. cit.] by the authors of the paper under review gives a \(p\)-adic analogue of these equivariant versions of the BSD conjecture. In order to do so, they assume that \(K = K_\rho\) is the number field cut out by a Galois representation \(\rho = \rho_1 \otimes \rho_2\), where \(\rho_1 \colon G_\mathbb{Q} \to \mathrm{GL}_2(V_{\rho_1})\) and \(\rho_2 \colon G_\mathbb{Q} \to \mathrm{GL}_2(V_{\rho_2})\) are odd, two-dimensional Artin representations of the absolute Galois group \(G_\mathbb{Q} := \mathrm{Gal}(\overline{\mathbb{Q}}/\mathbb{Q})\) with coefficients in a number field \(\mathbb{Q}(\rho)\), satisfying the self-duality assumption \(\det(\rho_1) = \det(\rho_2)^{-1}\). Then, the modularity theorem (see [Modular forms and Fermat's last theorem. Papers from a conference, Boston, MA, USA, August 9--18, 1995. New York, NY: Springer (1997; Zbl 0878.11004)]) associates to \(A\) a modular form \(f = \sum_n a_n(f) q^n\) of weight \(2\), and the theorem of Khare and Winterberger (see [\textit{C. Khare} and \textit{J.-P. Wintenberger}, Invent. Math. 178, No. 3, 485--504 (2009; Zbl 1304.11041)]) associates to \(\rho_1\) and \(\rho_2\) two modular forms \(g\) and \(h\) of weight \(1\). Fix a prime \(p\) such that \(A\) has either good ordinary or bad multiplicative reduction at \(p\). In the first case, let \(\beta_f\) be the unique root of the polynomial \(x^2 - a_p(f) x + p\) which is not a \(p\)-adic unit, and in the second case set \(\beta_f = 0\). Then, the \(p\)-stabilization \(f_\alpha(q) := f(q) - \beta_f f(q^p)\) is the specialization at weight two of a unique cuspidal Hida family \(\mathbf{f} \in \mathcal{O}(U_\mathbf{f})[[q]]\), and similarly the \(p\)-stabilizations of \(g\) and \(h\) are specializations at weight one of two other cuspidal Hida families \(\mathbf{g} \in \mathcal{O}(U_\mathbf{g})[[q]]\) and \(\mathbf{h} \in \mathcal{O}(U_\mathbf{h})[[q]]\). Here, \(U_\mathbf{f}\) is connected open disc centered at \(2\) in the weight space \(\mathcal{W}\) over \(\mathbb{Q}_p\), and \(U_\mathbf{g}, U_\mathbf{h}\) are open discs centered at \(1\) in \(\mathcal{W} \otimes_{\mathbb{Q}_p} L\), where \(L\) is a suitable finite extension of \(\mathbb{Q}_p\). Then, the recent work of \textit{M.-L. Hsieh} [Am. J. Math. 143, No. 2, 411--532 (2021; Zbl 1470.11115)] associates to the triple \((\mathbf{f},\mathbf{g},\mathbf{h})\) a \textit{Garrett-Hida \(p\)-adic \(L\)-function} \(L^{\alpha \alpha}_p(A,\rho) := \mathcal{L}_p(\mathbf{f},\mathbf{g},\mathbf{h})^2 \in \mathscr{O}_{\mathbf{fgh}} := \mathscr{O}(U_\mathbf{f}) \hat{\otimes} \mathscr{O}(U_\mathbf{g}) \hat{\otimes} \mathscr{O}(U_\mathbf{h})\), where \(\mathscr{O}(U)\) denotes the ring of bounded analytic functions on an open \(U\), which interpolates between special values of the \(L\)-functions associated to the products of the specializations of \(\mathbf{f}, \mathbf{g}\) and \(\mathbf{h}\) at classical points. The authors of the paper under review propose two conjectures concerning the behaviour of \(L^{\alpha \alpha}_p(A,\rho)\) at the point \(w_0 = (2,1,1)\) in the weight space. To state them, set \(\mathscr{I} \subseteq \mathscr{O}_{\mathbf{fgh}}\) to be the ideal of functions vanishing at \(w_0\). Then, given a function \(\varphi \in \mathscr{O}_{\mathbf{fgh}}\), we say that its order of vanishing at \(w_0\) is the unique \(r \in \mathbb{N}\) such that \(\varphi \in \mathscr{I}^r \setminus \mathscr{I}^{r+1}\), and the special value \(\varphi^\ast\) of \(\varphi\) at \(w_0\) is the image of \(\varphi\) in the quotient \(\mathscr{I}^r/\mathscr{I}^{r+1}\). Moreover, let us define the \textit{\(p\)-extended Mordell-Weil group of \((A,\rho)\)} as \(A^\dagger(K_\rho)^\rho := (A(K_\rho) \otimes_\mathbb{Z} (V_{\rho_1} \otimes_{\mathbb{Q}(\rho)} V_{\rho_2}))^{\mathrm{Gal}(K_\rho/\mathbb{Q})} \oplus \mathcal{Q}_p(A,\rho)\), where \(\mathcal{Q}_p(A,\rho)\) is a suitable \(\mathbb{Q}(\rho)\)-vector space depending only on the base change of \(A\) to \(\mathbb{Q}_p\), and on the restriction of \(\rho\) to \(\mathrm{Gal}(\overline{\mathbb{Q}}_p/\mathbb{Q}_p)\). Then, the authors of the paper under review conjecture in the recent preprint [loc. cit.] that \(\mathrm{ord}_{w_0}(L_p^{\alpha \alpha}(A,\rho)) = r^\dagger := \dim_{\mathbb{Q}(\rho)}(A^\dagger(K_\rho)^\rho)\) and \(L_p^{\alpha \alpha}(A,\rho)^\ast = \det(\ll P_i, P_j \gg)_{i,j}\), where \(P_1,\dots,P_{r^\dagger}\) is a \(\mathbb{Q}(\rho)\)-basis of \(A^\dagger(K_\rho)^\rho\), and \(\ll \cdot, \cdot \gg \colon A^\dagger(K_\rho)^\rho \otimes A^\dagger(K_\rho)^\rho \to \mathscr{I}/\mathscr{I}^2\) is a suitable height pairing. Finally, the paper under review proves these conjectures under the additional assumptions that \(p \geq 5\) is a prime of bad multiplicative reduction for \(A\), and the representations \(\rho_1\) and \(\rho_2\) are induced by two characters of the absolute Galois group of a given imaginary quadratic field \(K\) in which \(p\) is inert. Moreover, the authors assume that these characters are not induced by Dirichlet characters and do not ramify at the divisors of \(\Delta_K \cdot N_A\), where \(\Delta_K\) is the discriminant of \(K\), and \(N_A\) is the conductor of \(A\). Finally, the authors assume in the paper under review that there are an even number of rational primes which are inert in \(K\) and of bad reduction for \(A\). Under all these assumptions, they manage to prove the conjecture, using their recent construction of a big diagonal class satisfying certain reciprocity laws (see [\textit{M. Bertolini} et al., \url{https://www.uni-due.de/~ade847f/BSV.MO1.pdf} and \url{https://www.uni-due.de/~ade847f/BSV.MO2.pdf}]). The paper under review provides thus an important motivation to believe in the validity of the general \(p\)-adic analogue of the Birch and Swinnerton-Dyer conjecture put forward by the authors in their recent preprint [loc. cit.]. Thus, we recommend the interested reader to study the latter before approaching the paper under review.