This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

On the algebraicity of polyquadratic plectic points

Michele Fornea [email protected] Columbia University, New York, USA.    Lennart Gehrmann [email protected] Universität Duisburg-Essen, Essen, Germany.
Abstract

We establish direct evidence of the arithmetic significance of plectic Stark–Heegner points for elliptic curves of arbitrarily large rank. The main contribution is a proof of the algebraicity of plectic points associated to polyquadratic CM extensions of totally real number fields. Moreover, we relate the non-vanishing of plectic points to analytic and algebraic ranks of elliptic curves.

:
11F41, 11F67, 11G05, 11G40.

1 Introduction

In previous work, plectic Stark–Heegner points were associated to quadratic extensions E/FE/F of number fields and modular elliptic curves A/FA_{/F} under some technical assumptions. The construction generalizes the description of classical Heegner points on Shimura curves admitting a pp-adic uniformization, and combines Nekovárˇ\check{\text{r}}–Scholl’s plectic insights ([NS16], [Nek16]) with Darmon’s pioneering work on Stark-Heegner points [Dar01]. Moreover, Conjectures 1.3, 1.5 of [FG21] predict the algebraicity of plectic Stark–Heegner points and their significance for elliptic curves of higher rank, for which some numerical evidence was provided in [FGM21]. The aim of this paper is to establish direct evidence for the aforementioned conjectures in the polyquadratic CM case.

1.1 Conjectures on plectic Stark–Heegner points

Even though the formulation of the conjectures does does not require any restriction on the possible signatures of the fields EE and FF, we only consider CM extensions in this article; that is, FF is a totally real number field and E/FE/F a is totally complex quadratic extension. All prime divisors of the conductor 𝔣\mathfrak{f} of the elliptic curve AA are assumed to be unramified in E/FE/F. We fix a rational prime pp and a set S={𝔭1,,𝔭r}S=\{\mathfrak{p}_{1},\ldots,\mathfrak{p}_{r}\} of pp-adic primes of FF such that

  • \bullet

    A/FA_{/F} has multiplicative reduction at every 𝔭S\mathfrak{p}\in S,

  • \bullet

    the primes in SS are all inert in EE.

Given 𝔭S\mathfrak{p}\in S we also denote by 𝔭\mathfrak{p} the unique prime of EE above 𝔭\mathfrak{p} . Furthermore, we let a𝔭{±1}a_{\mathfrak{p}}\in\{\pm 1\} equal +1+1 (resp. 1-1) when AA has split (resp. non-split) multiplicative reduction at 𝔭\mathfrak{p}. The quantity a𝔭a_{\mathfrak{p}} is closely related to the local root number ε𝔭(A/F)\varepsilon_{\mathfrak{p}}(A/F) of AA:

a𝔭=ε𝔭(A/F).a_{\mathfrak{p}}=-\varepsilon_{\mathfrak{p}}(A/F).

By setting pS:=𝔭1𝔭rp_{S}:=\mathfrak{p}_{1}\cdots\mathfrak{p}_{r}, we may write

𝔣=pS𝔫𝔫,\mathfrak{f}=p_{S}\cdot\mathfrak{n}^{\sharp}\cdot\mathfrak{n}^{\flat},

for coprime ideals 𝔫,𝔫𝒪F\mathfrak{n}^{\sharp},\mathfrak{n}^{\flat}\in\mathcal{O}_{F} such that 𝔫\mathfrak{n}^{\sharp} is divisible by every prime divisor of 𝔣\mathfrak{f} split in E/FE/F.

Assumption 1.1 ((Plectic Heegner hypothesis for (A,E,S)(A,E,S))).

We require that

  • \bullet

    𝔫\mathfrak{n}^{\flat} is square-free,

  • \bullet

    the number of prime factors of 𝔫\mathfrak{n}^{\flat} is congruent to [F:][F:\mathbb{Q}] modulo 22.

Since E/FE/F is a CM extension, Assumption 1.1 implies that the sign of the functional equation of the LL-function L(A/E,s)L(A/E,s) of the base-change of A/FA_{/F} to EE is equal to ε(A/E)=(1)r\varepsilon(A/E)=(-1)^{r}.

We introduce the following useful notations. First, if HH is any commutative prodiscrete group, we denote by H^\widehat{H} the torsion-free part of its pro-pp completion. Second, if MM is an abelian group and Ω\Omega a field of characteristic zero, we use MΩM_{\Omega} and r(MΩ)\mathop{\mathchoice{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptscriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}}(M_{\Omega}) to respectively denote the tensor product MΩM\otimes_{\mathbb{Z}}\Omega, and the rr-th exterior power of the Ω\Omega-vector space MΩM_{\Omega}.

Consider the tensor product

A^(ES):=A^(E𝔭1)ppA^(E𝔭r).\widehat{A}(E_{S}):=\widehat{A}(E_{\mathfrak{p}_{1}})\otimes_{\mathbb{Z}_{p}}\ldots\otimes_{\mathbb{Z}_{p}}\widehat{A}(E_{\mathfrak{p}_{r}}).

Under the running assumptions, the plectic Stark–Heegner point (for the trivial character) is

PA,SA^(ES)\mathrm{P}_{A,S}\in\widehat{A}(E_{S})_{\mathbb{Q}}

as constructed in ([FG21], Section 4.4), where it was denoted PA𝟙\mathrm{P}^{\mathbbm{1}}_{A}. The notion of algebraicity for plectic Stark–Heegner points is formulated in terms of a determinant map. Writing ι𝔭:EE𝔭\iota_{\mathfrak{p}}\colon E\hookrightarrow E_{\mathfrak{p}} for the canonical embedding for every 𝔭S\mathfrak{p}\in S, we can consider the homomorphism

det:r(A(E))A^(ES)\operatorname{det}\colon\mathop{\mathchoice{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptscriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}}(A(E)_{\mathbb{Q}})\longrightarrow\widehat{A}(E_{S})_{\mathbb{Q}}\\

given by

det(P1Pr)=det(ι𝔭1(P1)ι𝔭r(P1)ι𝔭1(Pr)ι𝔭r(Pr)).\operatorname{det}\big{(}P_{1}\wedge\dots\wedge P_{r}\big{)}=\operatorname{det}\begin{pmatrix}\iota_{\mathfrak{p}_{1}}(P_{1})&\dots&\iota_{\mathfrak{p}_{r}}(P_{1})\\ \vdots&\ddots&\vdots\\ \iota_{\mathfrak{p}_{1}}(P_{r})&\dots&\iota_{\mathfrak{p}_{r}}(P_{r})\end{pmatrix}.

As usual, ralg(A/E)r_{\mathrm{alg}}(A/E) denotes the rank of the finitely generated abelian group A(E)A(E), and ran(A/E)r_{\mathrm{an}}(A/E) the order of vanishing of the LL-function L(A/E,s)L(A/E,s) at s=1s=1.

Conjecture 1.2.

If ralg(A/E)rr_{\mathrm{alg}}(A/E)\geq r, then there exists an element wA,Sr(A(E))w_{A,S}\in\mathop{\mathchoice{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptscriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}}(A(E)_{\mathbb{Q}}) such that

PA,S=det(wA,S).\mathrm{P}_{A,S}=\operatorname{det}(w_{A,S}).

Moreover, if PA,S0\mathrm{P}_{A,S}\neq 0, then ralg(A/E)=rr_{\mathrm{alg}}(A/E)=r.

Remark 1.3.

When SS consists of a single prime, Conjecture 1.2 follows from the generalization of the Gross–Zagier–Kolyvagin theorem for totally real number fields by Nekovárˇ\check{\text{r}} and Zhang. Indeed, Čerednik–Drinfeld’s uniformization of Shimura curves implies that the plectic point PA,S\mathrm{P}_{A,S} is a Heegner point when |S|=1\lvert S\rvert=1 (see Section 3.2 for more details).

There are also plectic Stark–Heegner points associated to non-trivial anticyclotomic characters of E/FE/F. Aiming for clarity in the introduction, we discuss them only in the body of the paper.

1.1.1 Eigenspaces for partial Frobenii.

Let σ𝔭\sigma_{\mathfrak{p}} be the generator of the Galois group of E𝔭/F𝔭E_{\mathfrak{p}}/F_{\mathfrak{p}}. It naturally acts on A^(E𝔭)\widehat{A}(E_{\mathfrak{p}}), and thus also on A^(ES)\widehat{A}(E_{S}) via its action on the 𝔭\mathfrak{p}-th factor. We set

A^(E𝔭)±:=A^(E𝔭)σ𝔭=±a𝔭andA^(ES)±:=𝔭SA^(E𝔭)±.\widehat{A}(E_{\mathfrak{p}})^{\pm}:=\widehat{A}(E_{\mathfrak{p}})^{\sigma_{\mathfrak{p}}=\pm a_{\mathfrak{p}}}_{\mathbb{Q}}\quad\mbox{and}\quad\widehat{A}(E_{S})^{\pm}:=\otimes_{\mathfrak{p}\in S}\widehat{A}(E_{\mathfrak{p}})^{\pm}.

There are two eigenspace projections

prS±:A^(ES)A^(ES)±,prS±=𝔭S(1±a𝔭σ𝔭).\operatorname{pr}^{\pm}_{S}\colon\widehat{A}(E_{S})_{\mathbb{Q}}\rightarrow\widehat{A}(E_{S})^{\pm},\qquad\operatorname{pr}^{\pm}_{S}=\prod_{\mathfrak{p}\in S}(1\pm a_{\mathfrak{p}}\cdot\sigma_{\mathfrak{p}}).

The main results of this article (Theorems A & B) establish the first cases of the minus part of the following conjecture, a direct consequence of Conjecture 1.2.

Conjecture 1.4.

If ralg(A/E)rr_{\mathrm{alg}}(A/E)\geq r, then there exists an element wA,Sr(A(E))w_{A,S}\in\mathop{\mathchoice{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptscriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}}(A(E)_{\mathbb{Q}}) such that

prS±(PA,S)=prS±(det(wA,S)).\operatorname{pr}^{\pm}_{S}(\mathrm{P}_{A,S})=\operatorname{pr}^{\pm}_{S}(\operatorname{det}(w_{A,S})).

Moreover, if prS±(PA,S)0\operatorname{pr}^{\pm}_{S}(\mathrm{P}_{A,S})\neq 0, then ralg(A/E)=rr_{\mathrm{alg}}(A/E)=r.

The special cases that we treat in Theorems A, B and C are singled out precisely to leverage the known properties of classical Heegner points. The key idea is to further suppose that FF is a polyquadratic extension of another totally real number field FF_{\circ}, the elliptic curve is the base change of an elliptic curve AA_{\circ} defined over FF_{\circ}, and EE is the compositum of FF with a quadratic CM extension E/FE_{\circ}/F_{\circ}. Then, under an appropriate Heegner hypothesis, we use Heegner points for A/EA_{\circ/E_{\circ}} and its twists by characters of Gal(F/F)\operatorname{Gal}(F/F_{\circ}), plus a factorization of anticyclotomic pp-adic LL-functions to establish our results.

1.1.2 Plectic pp-adic invariants.

Anticyclotomic pp-adic LL-functions come in to play in the proofs of our theorems because of the pp-adic Gross-Zagier formula ([FG21], Theorem A) relating higher order derivatives to plectic pp-adic invariants. These invariants, denoted QA,S\mathrm{Q}_{A,S}, are canonical lifts of the points prS(PA,S)\operatorname{pr}^{-}_{S}(\mathrm{P}_{A,S}) with respect to a “plectic” Tate parametrization: as the elliptic curve A/FA_{/F} has multiplicative reduction at every 𝔭S\mathfrak{p}\in S, Tate’s pp-adic uniformization results provides surjections ϕ𝔭Tate:E𝔭×A(E𝔭)\phi^{\mathrm{Tate}}_{\mathfrak{p}}\colon E_{\mathfrak{p}}^{\times}\twoheadrightarrow A(E_{\mathfrak{p}}) whose kernels are generated by Tate periods q𝔭F𝔭×𝒪F𝔭×q_{\mathfrak{p}}\in F_{\mathfrak{p}}^{\times}\setminus\mathcal{O}_{F_{\mathfrak{p}}}^{\times}. If we denote by E^𝔭:=(E^𝔭×)σ𝔭=1\widehat{E}_{\mathfrak{p}}^{-}:=(\widehat{E}_{\mathfrak{p}}^{\times})^{\sigma_{\mathfrak{p}}=-1} the subgroup of E^𝔭×\widehat{E}_{\mathfrak{p}}^{\times} on which σ𝔭\sigma_{\mathfrak{p}} acts via inversion, and we set

E^S,:=E^𝔭1ppE^𝔭r,\widehat{E}_{S,\otimes}^{-}:=\widehat{E}_{\mathfrak{p}_{1}}^{-}\otimes_{\mathbb{Z}_{p}}\ldots\otimes_{\mathbb{Z}_{p}}\widehat{E}_{\mathfrak{p}_{r}}^{-},

then the plectic pp-adic invariant QA,S\mathrm{Q}_{A,S} is the unique element of E^S,\widehat{E}_{S,\otimes}^{-} satisfying

ϕSTate(QA,S)=prS(PA,S).\phi^{\mathrm{Tate}}_{S}(\mathrm{Q}_{A,S})=\operatorname{pr}^{-}_{S}(\mathrm{P}_{A,S}).

Here ϕSTate:E^S,×A^(ES)\phi^{\mathrm{Tate}}_{S}\colon\widehat{E}_{S,\otimes}^{\times}\to\widehat{A}(E_{S}) denotes the tensor product of Tate’s local uniformizations. Since the restriction of ϕSTate\phi^{\mathrm{Tate}}_{S} to E^S,\widehat{E}_{S,\otimes}^{-} is injective, we have that QA,S0\mathrm{Q}_{A,S}\neq 0 if and only if prS(PA,S)0\operatorname{pr}^{-}_{S}(\mathrm{P}_{A,S})\neq 0.

1.2 The polyquadratic setup

For the rest of the introduction we suppose that the totally real number field FF is a polyquadratic extension of degree 2t2^{t} of a number field FF_{\circ}, i.e., F/FF/F_{\circ} is a Galois extension with Galois group 𝔊:=Gal(F/F)(/2)t.\mathfrak{G}:=\operatorname{Gal}(F/F_{\circ})\cong(\mathbb{Z}/2\mathbb{Z})^{t}. Further, we assume that EE is the compositum of FF with a quadratic CM extension E/FE_{\circ}/F_{\circ} and that the following technical assumptions are satisfied.

Assumption 1.5.

We require that

  • \bullet

    every non-trivial subextension of F/FF/F_{\circ} is ramified,

  • \bullet

    all primes of FF_{\circ} that ramify in FF split in EE_{\circ},

  • \bullet

    the elliptic curve A/FA_{/F} is the base change of a modular elliptic curve AA_{\circ} defined over FF_{\circ}, whose conductor 𝔣\mathfrak{f}_{\circ} is unramified in F/FF/F_{\circ},

  • \bullet

    the set SS consists of all primes of FF lying above a single prime \wp of FF_{\circ}, totally split in FF.

Remark 1.6.

The need for the first assumption is explained in Remark 4.2. The splitting in EE_{\circ} of the primes ramified in F/FF/F_{\circ} is necessary to construct Heegner points on twists of A/FA_{\circ/F} by characters of 𝔊\mathfrak{G}, while the total splitting of the prime \wp in F/FF/F_{\circ} is just a simplifying hypothesis for the proof of Proposition 4.5.

Assumption 1.5 implies that the elliptic curve A/FA_{/F} is modular by quadratic base change for Hilbert modular forms. Moreover, we deduce that the cardinality r=|S|r=\lvert S\rvert equals 2t2^{t}, and that the prime \wp is inert in E/FE_{\circ}/F_{\circ}. By a small abuse of notation, we denote by \wp the unique prime of EE_{\circ} lying above \wp. Since \wp is completely split in F/FF/F_{\circ}, the elliptic curve A/FA_{\circ/F_{\circ}} has multiplicative reduction at \wp. Furthermore, if we set a=1a_{\wp}=1 (resp. a=1a_{\wp}=-1) in case A/FA_{\circ/F_{\circ}} has split (resp. non-split) multiplicative reduction, we have

a𝔭=a𝔭S.a_{\mathfrak{p}}=a_{\wp}\quad\forall\ \mathfrak{p}\in S. (1)

Now, write the conductor 𝔣\mathfrak{f}_{\circ} of A/FA_{\circ/F_{\circ}} as

𝔣=𝔫𝔫\mathfrak{f}_{\circ}=\wp\cdot\mathfrak{n}_{\circ}^{\sharp}\cdot\mathfrak{n}_{\circ}^{\flat}

where 𝔫\mathfrak{n}_{\circ}^{\sharp} is divisible by every prime divisor of 𝔣\mathfrak{f}_{\circ} split in E/FE_{\circ}/F_{\circ}.

Assumption 1.7 ((Generalized Heegner hypothesis for (A,E,)(A_{\circ},E_{\circ},\wp))).

We require that

  • \bullet

    𝔫\mathfrak{n}_{\circ}^{\flat} is square-free,

  • \bullet

    the number of prime factors of 𝔫\mathfrak{n}_{\circ}^{\flat} is congruent to [F:][F_{\circ}:\mathbb{Q}] modulo 22.

Under Assumption 1.7, the sign of the functional equation for A/EA_{\circ/E_{\circ}} equals ε(A/E)=1\varepsilon(A_{\circ}/E_{\circ})=-1. Hence, the BSD–conjecture predicts that A(E)A_{\circ}(E_{\circ}) is non-torsion. Moreover, for any character η:𝔊{±1}\eta\colon\mathfrak{G}\to\{\pm 1\}, the twist A/FηA^{\eta}_{\circ/F_{\circ}} also fulfils the generalized Heegner hypothesis and we have

ε(Aη/E)=1\varepsilon(A_{\circ}^{\eta}/E_{\circ})=-1

because every prime ramified in F/FF/F_{\circ} splits in E/FE_{\circ}/F_{\circ} by Assumption 1.5. Thus, we expect that

ralg(A/E)?[F:F]=r,r_{\mathrm{alg}}(A/E)\stackrel{{\scriptstyle?}}{{\geq}}[F:F_{\circ}]=r,

with equality if and only if ralg(Aη/E)=1r_{\mathrm{alg}}(A^{\eta}_{\circ}/E_{\circ})=1 for every character η:𝔊{±1}\eta\colon\mathfrak{G}\to\{\pm 1\}.

Remark 1.8.

Under the running assumptions 1.5 and 1.7, the plectic Heegner hypothesis for (A,E,S)(A,E,S) holds if and only the number of prime divisors of 𝔫\mathfrak{n}^{\flat} is even. For example, this is the case when the conductor 𝔣\mathfrak{f}_{\circ} of A/FA_{\circ/F_{\circ}} is completely split in F/FF/F_{\circ}.

Remark 1.9.

Simple examples satisfying all our hypotheses can be found by considering F=F_{\circ}=\mathbb{Q}, E/E_{\circ}/\mathbb{Q} an imaginary quadratic field, F/F/\mathbb{Q} a real quadratic field, and AA_{\circ} a rational elliptic curve of conductor 𝔣=pq\mathfrak{f}_{\circ}=p\cdot q for two rational primes both inert in E/E_{\circ}/\mathbb{Q}, and with pp split in F/F/\mathbb{Q}.

1.3 Main results

As \wp is completely split in FF, we have for every 𝔭S\mathfrak{p}\in S canonical identifications F,=F𝔭F_{\circ,\wp}=F_{\mathfrak{p}} and E,=E𝔭E_{\circ,\wp}=E_{\mathfrak{p}}. The resulting identifications A^(E,)=A^(E𝔭)\widehat{A_{\circ}}(E_{\circ,\wp})=\widehat{A}(E_{\mathfrak{p}}) are used to define the norm map

NS/:A^(ES)\xlongrightarrowA^(E,)rSympr(A^(E,)),\operatorname{N}_{S/\wp}\colon\widehat{A}(E_{S})\xlongrightarrow{\sim}\widehat{A_{\circ}}(E_{\circ,\wp})^{\otimes r}\longrightarrow\operatorname{Sym}^{r}_{\mathbb{Z}_{p}}\big{(}\widehat{A_{\circ}}(E_{\circ,\wp})\big{)},

where the second arrow is the canonical projection.

Remark 1.10.

If \wp is of degree one, the restriction of the norm map NS/\operatorname{N}_{S/\wp} to A^(ES)±\widehat{A}(E_{S})^{\pm} is injective.

Under our running assumptions (1.1, 1.5, 1.7), we deduce the following theorems about plectic points from the known properties of Heegner points, the pp-adic uniformization of Shimura curves, and a factorization of anticyclotomic pp-adic LL-functions (Corollary 4.4).

Theorem A ((Arithmetic significance)).

The following implication holds:

NS/(prS(PA,S))0ralg(A/E)=r&ran(A/E)=r.\operatorname{N}_{S/\wp}\big{(}\operatorname{pr}^{-}_{S}(\mathrm{P}_{A,S})\big{)}\neq 0\quad\implies\quad r_{\mathrm{alg}}(A/E)=r\quad\&\quad r_{\mathrm{an}}(A/E)=r.
Theorem B ((Algebraicity)).

There is a quadratic extension Ω/\Omega/\mathbb{Q} and wA,Sr(A(E)Ω)w_{A,S}\in\mathop{\mathchoice{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptscriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}}(A(E)_{\Omega}) s.t.

NS/(prS(PA,S))=NS/(prS(det(wA,S))).\operatorname{N}_{S/\wp}\big{(}\operatorname{pr}^{-}_{S}(\mathrm{P}_{A,S})\big{)}=\operatorname{N}_{S/\wp}\big{(}\operatorname{pr}^{-}_{S}(\operatorname{det}(w_{A,S}))\big{)}.
Remark 1.11.

Aside from the quadratic extension Ω/\Omega/\mathbb{Q}, Theorems A and B provide a proof of the minus part of Conjecture 1.4 in the polyquadratic setup when \wp is of degree one. Using the main theorem of [HM22] one can apply the same strategy to prove the plus part of the conjecture. This will be explained in more detail in future work.

Remark 1.12.

The quadratic extension Ω/\Omega/\mathbb{Q} is generated by the square-root of a rational number that is the product of various explicit terms: Petersson norms, discriminants, Euler factors and special values of Dedekind zeta functions. It would be interesting to know whether that rational number is in fact a square. Similar questions were raised and shown to be implied by the Birch–Swinnerton-Dyer conjecture in [Mok10].

Now, set A/F+=A/FA^{+}_{/F}=A_{/F} and denote by A/FA^{-}_{/F} the quadratic twist of A/FA_{/F} with respect to the extension E/FE/F. We partition S=S+SS=S^{+}\cup S^{-} by declaring that the subset S+SS^{+}\subseteq S contains all the primes in SS of split multiplicative reduction for A/F+A^{+}_{/F}, and define ϱA(S):=max{ralg(A±/F)+|S±|}.\varrho_{A}(S):=\max\big{\{}r_{\mathrm{alg}}(A^{\pm}/F)+\lvert S^{\pm}\rvert\big{\}}.

Theorem C.

We also have

NS/(prS(PA,S))0ran(A/E)=r&ϱA(S)=r.\operatorname{N}_{S/\wp}\big{(}\operatorname{pr}^{-}_{S}(\mathrm{P}_{A,S})\big{)}\neq 0\quad\iff\quad r_{\mathrm{an}}(A/E)=r\quad\&\quad\varrho_{A}(S)=r.
Remark 1.13.

If \wp is a prime of degree one, Theorem C establishes ([FGM21], Conjecture 1.5) in the polyquadratic CM case.

We note that in the main body of this article, we prove generalizations of Theorems A, B for plectic Stark–Heegner points associated to anticyclotomic characters of E/FE/F that are restrictions of anticyclotomic characters of E/FE_{\circ}/F_{\circ}.

Acknowledgements.
We warmly thank Jan Vonk for the numerous conversations related to our work on plectic Stark–Heegner points. While working on this article, the first named author was a Simons Junior Fellow.

2 Preliminaries

We gather some basic results on symmetric powers and completed group algebras.

2.1 Symmetric powers

Let us fix a commutative ring RR. Given an RR-module MM and an integer n0n\geq 0 we write

Mn:=MRRMntimes.M^{\otimes n}:=\underbrace{M\otimes_{R}\ldots\otimes_{R}M}_{n\ \text{times}}.

Recall that the symmetric algebra SymR(M)\operatorname{Sym}_{R}(M) of MM is the quotient of the tensor algebra

TR(M):=n0Mn\operatorname{T}_{R}(M):=\bigoplus_{n\geq 0}M^{\otimes n}

by the ideal generated by xyyxx\otimes y-y\otimes x for x,yMx,y\in M. As this ideal is graded, the natural grading on TR(M)\operatorname{T}_{R}(M) induces a grading on SymR(M)\operatorname{Sym}_{R}(M):

SymR(M)=n0SymRn(M).\operatorname{Sym}_{R}(M)=\bigoplus_{n\geq 0}\operatorname{Sym}^{n}_{R}(M).

We denote the image of an element mMnm\in M^{\otimes n} in SymRn(M)\operatorname{Sym}_{R}^{n}(M) by [m][m]. Given a homomorphism f:MNf\colon M\to N of RR-modules we write

SymRn(f):SymRn(M)SymRn(N)\operatorname{Sym}_{R}^{n}(f)\colon\operatorname{Sym}_{R}^{n}(M)\longrightarrow\operatorname{Sym}_{R}^{n}(N)

for the induced homomorphism. If MM and NN are RR-modules, there is a canonical isomorphism

SymR(NM)=SymR(M)RSymR(N)\displaystyle\operatorname{Sym}_{R}(N\oplus M)=\operatorname{Sym}_{R}(M)\otimes_{R}\operatorname{Sym}_{R}(N) (2)

of RR-algebras. Now, suppose MM is a finitely generated free RR-module with generators m1,,mm_{1},\ldots,m_{\ell}, then there is an isomorphism of graded RR-algebras

R[x1,,x]\xlongrightarrowSymR(M),ximi.\displaystyle R[x_{1},\ldots,x_{\ell}]\xlongrightarrow{\sim}\operatorname{Sym}_{R}(M),\qquad x_{i}\mapsto m_{i}. (3)
Lemma 2.1.

Let RR be an integrally closed domain, MM a finitely generated free RR-module, and

()2:SymRn(M)SymR2n(M)(-)^{2}\colon\operatorname{Sym}^{n}_{R}(M)\longrightarrow\operatorname{Sym}^{2n}_{R}(M)

the squaring map. If x,yx,y are elements of SymRn(M)\operatorname{Sym}^{n}_{R}(M) and CR{0}C\in R\setminus\{0\} is a non-zero constant satisfying x2=Cy2x^{2}=C\cdot y^{2}, then there exists a square-root CR\sqrt{C}\in R such that x=Cyx=\sqrt{C}\cdot y.

Proof.

As RR is an integrally closed domain, equation (3) implies that the RR-algebra SymR(M)\operatorname{Sym}_{R}(M) is one as well. Thus, the equality C=(x/y)2C=(x/y)^{2} in the fraction field of SymR(M)\operatorname{Sym}_{R}(M) implies that C:=x/y\sqrt{C}:=x/y is an element of SymR(M)\operatorname{Sym}_{R}(M). Moreover, CR\sqrt{C}\in R because its square belongs to RR. ∎

The following lemma can be easily deduced from (2) and (3).

Lemma 2.2.

Let M1,,MnM_{1},\ldots,M_{n} and MM be finitely generated free RR-modules.

  1. (a)

    The canonical map

    μ:M1RRMnSymRn(M1Mn),m1mn[m1mn]\mu\colon M_{1}\otimes_{R}\ldots\otimes_{R}M_{n}\to\operatorname{Sym}^{n}_{R}(M_{1}\oplus\ldots\oplus M_{n}),\qquad m_{1}\otimes\ldots\otimes m_{n}\mapsto[m_{1}\otimes\ldots\otimes m_{n}]

    is injective.

  2. (b)

    The following diagram is commutative

    Mn\textstyle{M^{\otimes n}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}μ\scriptstyle{\mu}[]\scriptstyle{[-]}SymRn(Mn)\textstyle{\operatorname{Sym}_{R}^{n}(M^{\oplus n})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}SymRn(idMn)\scriptstyle{\operatorname{Sym}_{R}^{n}(\operatorname{id}_{M}^{\oplus n})}SymRn(M).\textstyle{\operatorname{Sym}_{R}^{n}(M).}

2.2 Completed group algebras

Let G=limiGiG=\varprojlim_{i}G_{i} be a topologically finitely generated commutative profinite group. Recall that the completed group algebra of GG with coefficients in a commutative ring RR is defined as

RG=limR[Gi].R\llbracket G\rrbracket=\varprojlim\hskip 1.42262ptR[G_{i}].

We denote by ():RGRG(-)^{\vee}\colon R\llbracket G\rrbracket\to R\llbracket G\rrbracket the involution induced by inversion on GG. In the rest of this section we always consider the coefficient ring R=pR=\mathbb{Z}_{p}. Let I(G)pGI(G)\subseteq\mathbb{Z}_{p}\llbracket G\rrbracket be the augmentation ideal, i.e. the kernel of the natural map pGp\mathbb{Z}_{p}\llbracket G\rrbracket\twoheadrightarrow\mathbb{Z}_{p}. More generally, if QQ is a quotient of GG by an open subgroup, the relative augmentation ideal is defined as

IQ(G):=ker(pG-↠p[Q]).I_{Q}(G):=\ker(\mathbb{Z}_{p}\llbracket G\rrbracket\relbar\joinrel\twoheadrightarrow\mathbb{Z}_{p}[Q]).

Note that the quotients IQ(G)n/IQ(G)n+1I_{Q}(G)^{n}/I_{Q}(G)^{n+1} are modules over the group ring p[Q]\mathbb{Z}_{p}[Q]. If Θ\Theta is an element of IQ(G)nI_{Q}(G)^{n}, we write Qn(Θ)\partial^{n}_{Q}(\Theta) for its image in IQ(G)n/IQ(G)n+1I_{Q}(G)^{n}/I_{Q}(G)^{n+1}. The map GI(G)G\to I(G), g[g]1g\mapsto[g]-1 induces an isomorphism of p\mathbb{Z}_{p}-modules

G^p\xlongrightarrowI(G)/I(G)2,\displaystyle G\otimes_{\widehat{\mathbb{Z}}}\mathbb{Z}_{p}\xlongrightarrow{\sim}I(G)/I(G)^{2}, (4)

and for every integer n1n\geq 1 a surjection of p\mathbb{Z}_{p}-modules

Sympn(G^p)-↠I(G)n/I(G)n+1.\operatorname{Sym}_{\mathbb{Z}_{p}}^{n}(G\otimes_{\widehat{\mathbb{Z}}}\mathbb{Z}_{p})\relbar\joinrel\twoheadrightarrow I(G)^{n}/I(G)^{n+1}.

When GG is a finitely generated free p\mathbb{Z}_{p}-module, a choice of topological generators {g1,,gs}\{g_{1},\ldots,g_{s}\} determines an isomorphism

pG\xlongrightarrowpt1,,ts,[gi]ti+1,\displaystyle\mathbb{Z}_{p}\llbracket G\rrbracket\xlongrightarrow{\sim}\mathbb{Z}_{p}\llbracket t_{1},\ldots,t_{s}\rrbracket,\quad[g_{i}]\longmapsto t_{i}+1, (5)

mapping the augmentation ideal to the ideal (t1,,ts)(t_{1},\ldots,t_{s}). It follows that the surjective maps

Sympn(G)I(G)n/I(G)n+1\displaystyle\operatorname{Sym}_{\mathbb{Z}_{p}}^{n}(G)\overset{\sim}{\longrightarrow}I(G)^{n}/I(G)^{n+1} (6)

are isomorphisms for all n1n\geq 1. Furthermore, when GG is a product G=H×QG=H\times Q with QQ finite, it is easy to see that the canonical map

I(H)n/I(H)n+1ppI(G)n/I(G)n+1ppI(H)^{n}/I(H)^{n+1}\otimes_{\mathbb{Z}_{p}}\mathbb{Q}_{p}\longrightarrow I(G)^{n}/I(G)^{n+1}\otimes_{\mathbb{Z}_{p}}\mathbb{Q}_{p}

is an isomorphism for all n1n\geq 1. The following lemma gives a slight generalization of this fact.

Lemma 2.3.

Let GG be a finitely generated commutative profinite group, HGH\leq G an open subgroup that is a finitely generated p\mathbb{Z}_{p}-module, and QQ a finite quotient of GG such that Hker(GQ)H\subseteq\ker(G\twoheadrightarrow Q). Then, the canonical p[Q]\mathbb{Q}_{p}[Q]-linear map

I(H)n/I(H)n+1pp[Q]IQ(G)n/IQ(G)n+1ppI(H)^{n}/I(H)^{n+1}\otimes_{\mathbb{Z}_{p}}\mathbb{Q}_{p}[Q]\longrightarrow I_{Q}(G)^{n}/I_{Q}(G)^{n+1}\otimes_{\mathbb{Z}_{p}}\mathbb{Q}_{p}

is injective for all n1n\geq 1.

Now, both the augmentation ideal I(G)I(G) and the relative versions IQ(G)I_{Q}(G) are clearly stable under ()(-)^{\vee}. Equation (4) implies that ()(-)^{\vee} induces multiplication with 1-1 on I(G)/I(G)2I(G)/I(G)^{2} and, thus, it induces multiplication with (1)n(-1)^{n} on I(G)n/I(G)n+1I(G)^{n}/I(G)^{n+1}. This observation readily implies the following relative statement: under the assumptions of Lemma 2.3 the following diagram commutes

I(H)n/I(H)n+1pp[Q]\textstyle{I(H)^{n}/I(H)^{n+1}\otimes_{\mathbb{Z}_{p}}\mathbb{Z}_{p}[Q]\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}(1)nid()\scriptstyle{(-1)^{n}\operatorname{id}\otimes(-)^{\vee}}IQ(G)n/IQ(G)n+1\textstyle{I_{Q}(G)^{n}/I_{Q}(G)^{n+1}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}()\scriptstyle{(-)^{\vee}}I(H)n/I(H)n+1pp[Q]\textstyle{I(H)^{n}/I(H)^{n+1}\otimes_{\mathbb{Z}_{p}}\mathbb{Z}_{p}[Q]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}IQ(G)n/IQ(G)n+1.\textstyle{I_{Q}(G)^{n}/I_{Q}(G)^{n+1}.} (7)

3 Plectic points and pp-adic LL-functions

We begin by explaining how the construction of plectic Stark-Heegner points recovers the pp-adic uniformization of classical Heegner points as a special case. Then, we recall the pp-adic Gross–Zagier formula ([FG21], Theorem A) relating plectic points to derivatives of certain anticyclotomic pp-adic LL-functions, for which we also state precise interpolation formulas. Note that in this section we work in the setup of Subsection 1.1, that is, E/FE/F is an arbitrary quadratic CM extension and Assumption 1.1 is supposed to hold. In particular, we never assume that we are in a polyquadratic situation.

3.1 Plectic Stark–Heegner points

We fix an 𝒪F\mathcal{O}_{F}-ideal 𝔠\mathfrak{c} coprime to the conductor of A/FA_{/F}. Let E𝔠/EE_{\mathfrak{c}}/E denote the anticyclotomic extension of conductor 𝔠\mathfrak{c} defined in ([FG21], Section 4.2.1) with Galois group 𝒢𝔠=Gal(E𝔠/E)\mathcal{G}_{\mathfrak{c}}=\operatorname{Gal}(E_{\mathfrak{c}}/E). Recall that for any character χ:𝒢𝔠×\chi\colon\mathcal{G}_{\mathfrak{c}}\to\mathbb{C}^{\times} its conductor is the maximal divisor 𝔠χ\mathfrak{c}_{\chi} of 𝔠\mathfrak{c} such that χ\chi factors through 𝒢𝔠𝒢𝔠χ\mathcal{G}_{\mathfrak{c}}\twoheadrightarrow\mathcal{G}_{\mathfrak{c}_{\chi}}. We write χ\mathbb{Q}_{\chi} for the extension of \mathbb{Q} generated by the values of χ\chi. With a small change of notation compared to ([FG21], Section 4.4), we denote the plectic Stark–Heegner point associated to an anticyclotomic character χ\chi by

PA,SχA^(ES)χ.\mathrm{P}_{A,S}^{\chi}\in\widehat{A}(E_{S})_{\mathbb{Q}_{\chi}}.

It follows easily from the construction of plectic Stark–Heegner points that there is an element

PA,S𝒢𝔠A^(ES)[𝒢𝔠]\mathrm{P}_{A,S}^{\mathcal{G}_{\mathfrak{c}}}\in\widehat{A}(E_{S})\otimes_{\mathbb{Z}}\mathbb{Q}[\mathcal{G}_{\mathfrak{c}}]

such that the equality χ(PA,S𝒢𝔠)=PA,Sχ\chi(\mathrm{P}_{A,S}^{\mathcal{G}_{\mathfrak{c}}})=\mathrm{P}_{A,S}^{\chi} holds for any character χ:𝒢𝔠×\chi\colon\mathcal{G}_{\mathfrak{c}}\to\mathbb{C}^{\times} of conductor 𝔠\mathfrak{c}. A similar statement also holds for plectic pp-adic invariants ([FG21], Section 4.2): there is an element

QA,S𝒢𝔠E^S,[𝒢𝔠]\mathrm{Q}_{A,S}^{\mathcal{G}_{\mathfrak{c}}}\in\widehat{E}_{S,\otimes}^{-}\otimes_{\mathbb{Z}}\mathbb{Z}[\mathcal{G}_{\mathfrak{c}}]

such that the equality χ(QA,S𝒢𝔠)=QA,Sχ\chi(\mathrm{Q}_{A,S}^{\mathcal{G}_{\mathfrak{c}}})=\mathrm{Q}_{A,S}^{\chi} holds for any character χ:𝒢𝔠×\chi\colon\mathcal{G}_{\mathfrak{c}}\to\mathbb{C}^{\times} of conductor 𝔠\mathfrak{c}. Moreover, the two elements are related by the following equation

ϕSTate(QA,S𝒢𝔠)=prS(PA,S𝒢𝔠).\phi^{\mathrm{Tate}}_{S}(\mathrm{Q}_{A,S}^{\mathcal{G}_{\mathfrak{c}}})=\operatorname{pr}_{S}^{-}(\mathrm{P}_{A,S}^{\mathcal{G}_{\mathfrak{c}}}). (8)

3.2 Plectic points are Heegner points when |S|=1\lvert S\rvert=1

The construction of plectic Stark–Heegner points generalizes the pp-adic description of classical Heegner points given in ([BD98], [Mok11]). In this subsection, we recall the precise relation between the two constructions when the set SS consists of a single prime {𝔭}\{\mathfrak{p}\} and χ:𝒢𝔠×\chi\colon\mathcal{G}_{\mathfrak{c}}\to\mathbb{C}^{\times} is a character of conductor 𝔠\mathfrak{c}.

We fix embeddings ι𝔭:¯¯p\iota_{\mathfrak{p}}\colon\overline{\mathbb{Q}}\hookrightarrow\overline{\mathbb{Q}}_{p} and ιν:¯\iota_{\nu}\colon\overline{\mathbb{Q}}\hookrightarrow\mathbb{C} respectively inducing the pp-adic prime 𝔭\mathfrak{p} and an Archimedean place ν\nu of FF. Since 𝔭\mathfrak{p} is inert in E/FE/F, it splits completely in the anticyclotomic extension E𝔠/EE_{\mathfrak{c}}/E. Thus, the embedding ι𝔭\iota_{\mathfrak{p}} restricts to an embedding ι:E𝔠E𝔭\iota_{\wp}\colon E_{\mathfrak{c}}\hookrightarrow E_{\mathfrak{p}} inducing an injective homomorphism

ιA,𝔭:A(E𝔠)χA^(E𝔭)χ.\iota_{A,\mathfrak{p}}\colon A(E_{\mathfrak{c}})_{\mathbb{Q}_{\chi}}\longrightarrow\widehat{A}(E_{\mathfrak{p}})_{\mathbb{Q}_{\chi}}.

We write ralg(A/E,χ)r_{\mathrm{alg}}(A/E,\chi) for the χ\mathbb{Q}_{\chi}-dimension of the χ\chi-component A(E𝔠)χχA(E_{\mathfrak{c}})^{\chi}_{\mathbb{Q}_{\chi}} and define ran(A/E,χ)r_{\mathrm{an}}(A/E,\chi) to be the order of vanishing of the LL-function L(A/E,χ,s)L(A/E,\chi,s) at s=1s=1. As explained in ([GMM20], Appendix A.1.2), there exists a Heegner point PA,𝔠A(E𝔠)P_{A,\mathfrak{c}}\in A(E_{\mathfrak{c}}) of conductor 𝔠\mathfrak{c}, arising from a Shimura curve associated to the FF-quaternion algebra ramified exactly at 𝔭𝔫\mathfrak{p}\mathfrak{n}^{\flat} and all Archimedean places different from ν\nu, such that the image of

PAχ:=σ𝒢𝔠χ1(σ)σ(PA,𝔠)A(E𝔠)χχP_{A}^{\chi}:=\sum_{\sigma\in\mathcal{G}_{\mathfrak{c}}}\chi^{-1}(\sigma)\cdot\sigma(P_{A,\mathfrak{c}})\ \in A(E_{\mathfrak{c}})^{\chi}_{\mathbb{Q}_{\chi}} (9)

under ιA,𝔭\iota_{A,\mathfrak{p}} is a non-zero rational multiple of the plectic Stark–Heegner point associated to the triple (A,E,{𝔭})(A,E,\{\mathfrak{p}\}), i.e., there exists kA,𝔭χ×k_{A,\mathfrak{p}}^{\chi}\in\mathbb{Q}^{\times} such that:

kA,𝔭χPA,{𝔭}χ=ιA,𝔭(PAχ).k_{A,\mathfrak{p}}^{\chi}\cdot\mathrm{P}_{A,\{\mathfrak{p}\}}^{\chi}=\iota_{A,\mathfrak{p}}\big{(}P_{A}^{\chi}\big{)}. (10)
Proposition 3.1.

Let χ:𝒢𝔠×\chi\colon\mathcal{G}_{\mathfrak{c}}\to\mathbb{C}^{\times} be a character of conductor 𝔠\mathfrak{c}. We have

PA,{𝔭}χ0ran(A/E,χ)=1,\mathrm{P}_{A,\{\mathfrak{p}\}}^{\chi}\neq 0\quad\iff\quad r_{\mathrm{an}}(A/E,\chi)=1,

and both statements imply ralg(A/E,χ)=1r_{\mathrm{alg}}(A/E,\chi)=1. Moreover, if χ\chi is not quadratic,

ran(A/E,χ)=1QA,{𝔭}χ0.r_{\mathrm{an}}(A/E,\chi)=1\quad\implies\quad\mathrm{Q}_{A,\{\mathfrak{p}\}}^{\chi}\neq 0.
Proof.

By equation (10), the equivalence between the non-triviality of PA,{𝔭}χ\mathrm{P}_{A,\{\mathfrak{p}\}}^{\chi} and the analytic rank one statement follows from ([Zha01], Theorem 1.2.1), while the relation with the algebraic rank is a consequence of the main theorem in [Nek07]. For the second claim, we begin by noting that the involution A(E𝔠)A(E𝔠),PP¯,A(E_{\mathfrak{c}})\to A(E_{\mathfrak{c}}),\ P\mapsto\overline{P}, induced by the complex conjugation associated to the Archimedean place ν\nu, yields an isomorphism

A(E𝔠)χχ\xlongrightarrowA(E𝔠)χχ1.A(E_{\mathfrak{c}})^{\chi}_{\mathbb{Q}_{\chi}}\xlongrightarrow{\sim}A(E_{\mathfrak{c}})^{\chi^{-1}}_{\mathbb{Q}_{\chi}}.

Then, we observe that equations (8) and (10) imply the equality

kA,𝔭χϕTate(QA,{𝔭}χ)=ιA,𝔭(PAχaPAχ¯)k_{A,\mathfrak{p}}^{\chi}\cdot\phi^{\mathrm{Tate}}_{\wp}\big{(}\mathrm{Q}_{A,\{\mathfrak{p}\}}^{\chi}\big{)}=\iota_{A,\mathfrak{p}}\big{(}P_{A}^{\chi}-a_{\wp}\cdot\overline{P_{A}^{\chi}}\big{)} (11)

because σιA,𝔭(PAχ)=ιA,𝔭(PAχ¯)\sigma_{\wp}\circ\iota_{A,\mathfrak{p}}\big{(}P_{A}^{\chi}\big{)}=\iota_{A,\mathfrak{p}}\big{(}\overline{P_{A}^{\chi}}\big{)} by ([BD98], Theorem 4.7). Now, our assumption is that ran(A/E,χ)=1r_{\mathrm{an}}(A/E,\chi)=1, and ([Zha01], Theorem 1.2.1) implies that PAχ0P_{A}^{\chi}\neq 0. When χ\chi is not quadratic, the intersection A(E𝔠)χχA(E𝔠)χχ1A(E_{\mathfrak{c}})^{\chi}_{\mathbb{Q}_{\chi}}\cap A(E_{\mathfrak{c}})^{\chi^{-1}}_{\mathbb{Q}_{\chi}} is trivial and the claim follows. ∎

Remark 3.2.

Let χ=𝟙\chi=\mathbbm{1} be the trivial character and assume that ran(A/E)=1r_{\mathrm{an}}(A/E)=1. Using equation (11) and ([Mok11], Corollary 4.2), we deduce that pr(PA,{𝔭}𝟙)0\operatorname{pr}^{-}_{\wp}(\mathrm{P}_{A,\{\mathfrak{p}\}}^{\mathbbm{1}})\not=0 is equivalent to either

(a=+1&ralg(A/F)=0)or(a=1&ralg(A/F)=1).\Big{(}a_{\wp}=+1\quad\&\quad r_{\mathrm{alg}}(A/F)=0\Big{)}\quad\text{or}\quad\Big{(}a_{\wp}=-1\quad\&\quad r_{\mathrm{alg}}(A/F)=1\Big{)}.

3.3 The anticyclotomic Gross–Zagier formula

Let E𝔠,SE_{\mathfrak{c},S} be the union of the anticyclotomic extensions of EE of conductor 𝔠𝔭S𝔭n\mathfrak{c}\cdot\prod_{\mathfrak{p}\in S}\mathfrak{p}^{n} for n0n\geq 0. We put 𝒢𝔠,S=Gal(E𝔠,S/E)\mathcal{G}_{\mathfrak{c},S}=\operatorname{Gal}(E_{\mathfrak{c},S}/E) and denote by

I𝔠p𝒢𝔠,SI_{\mathfrak{c}}\subseteq\mathbb{Z}_{p}\llbracket\mathcal{G}_{\mathfrak{c},S}\rrbracket

the relative augmentation ideal with respect to the quotient map 𝒢𝔠,S𝒢𝔠\mathcal{G}_{\mathfrak{c},S}\twoheadrightarrow\mathcal{G}_{\mathfrak{c}}. Restriction of the global Artin homomorphism to the local components at 𝔭S\mathfrak{p}\in S induces the homomorphism

recS:𝔭SE𝔭𝒢𝔠,S.\operatorname{rec}_{S}\colon\bigoplus_{\mathfrak{p}\in S}E_{\mathfrak{p}}^{-}\longrightarrow\mathcal{G}_{\mathfrak{c},S}.

Since E/FE/F is a quadratic CM extensions, the kernel of recS\operatorname{rec}_{S} is torsion and its image is an open subgroup of ker(𝒢𝔠,S𝒢𝔠)\ker(\mathcal{G}_{\mathfrak{c},S}\twoheadrightarrow\mathcal{G}_{\mathfrak{c}}). Moreover, Lemma 2.2 (a) and Lemma 2.3 imply that the map

drecS:E^S,pp[𝒢𝔠]Sympr(E^𝔭1ppE^𝔭r)pp[𝒢𝔠](I𝔠r/I𝔠r+1)\textnormal{d}\operatorname{rec}_{S}\colon\widehat{E}_{S,\otimes}^{-}\otimes_{\mathbb{Z}_{p}}\mathbb{Z}_{p}[\mathcal{G}_{\mathfrak{c}}]\longrightarrow\operatorname{Sym}^{r}_{\mathbb{Z}_{p}}(\widehat{E}_{\mathfrak{p}_{1}}^{-}\otimes_{\mathbb{Z}_{p}}\ldots\otimes_{\mathbb{Z}_{p}}\widehat{E}_{\mathfrak{p}_{r}}^{-})\otimes_{\mathbb{Z}_{p}}\mathbb{Z}_{p}[\mathcal{G}_{\mathfrak{c}}]\longrightarrow\left(I_{\mathfrak{c}}^{r}/I_{\mathfrak{c}}^{r+1}\right)_{\mathbb{Q}}

is injective. Associated to the quadruple (A,S,E,𝔠)(A,S,E,\mathfrak{c}) there is the square-root pp-adic LL-functions

S(A/E)𝔠𝒢𝔠,S\mathscr{L}_{S}(A/E)_{\mathfrak{c}}\in\mathbb{Z}\llbracket\mathcal{G}_{\mathfrak{c},S}\rrbracket

constructed in ([BG18], Definition 5.2) and also in ([FG21], Definition 5.7). On the one hand, the values of this pp-adic LL-function at finite order characters not satisfying certain ramification conditions are always equal to zero (see [FG21], Theorem 5.10), while the (squares of the) non-trivial values are explicitly calculated in ([BG18], Theorem 5.8). We recall the interpolation formula in Theorem 3.8 below. For the rest of this subsection we consider S(A/E)𝔠\mathscr{L}_{S}(A/E)_{\mathfrak{c}} as an element of p𝒢𝔠,S\mathbb{Z}_{p}\llbracket\mathcal{G}_{\mathfrak{c},S}\rrbracket. Recall that ([BG18], Theorem 5.5) shows that

S(A/E)𝔠I𝔠r.\mathscr{L}_{S}(A/E)_{\mathfrak{c}}\in I_{\mathfrak{c}}^{r}.

The following is a reformulation of the main theorem of [FG21].

Theorem 3.3 ((Anticyclotomic Gross-Zagier formula)).

The equality

2r𝒢𝔠r(S(A/E)𝔠)=drecS((QA,S𝒢𝔠))2^{r}\cdot\partial_{\mathcal{G}_{\mathfrak{c}}}^{r}(\mathscr{L}_{S}(A/E)_{\mathfrak{c}})=\textnormal{d}\operatorname{rec}_{S}\big{(}(\mathrm{Q}_{A,S}^{\mathcal{G}_{\mathfrak{c}}})^{\vee}\big{)}

holds in (I𝔠r/I𝔠r+1)(I_{\mathfrak{c}}^{r}/I_{\mathfrak{c}}^{r+1})_{\mathbb{Q}}.

Interestingly, ([BG18], Proposition 5.6) allows us to describe the behaviour of S(A/E)𝔠\mathscr{L}_{S}(A/E)_{\mathfrak{c}} under the involution ()(-)^{\vee} in terms of the global root number ε(A/F)\varepsilon(A/F) of A/FA_{/F} and a product of local root numbers εS(A/F):=𝔭Sε𝔭(A/F)\varepsilon_{S}(A/F):=\prod_{\mathfrak{p}\in S}\varepsilon_{\mathfrak{p}}(A/F).

Proposition 3.4.

The equality

(S(A/E)𝔠)=ε(A/F)εS(A/F)S(A/E)𝔠(\mathscr{L}_{S}(A/E)_{\mathfrak{c}})^{\vee}=\varepsilon(A/F)\cdot\varepsilon_{S}(A/F)\cdot\mathscr{L}_{S}(A/E)_{\mathfrak{c}}

holds up to multiplication with an element in 𝒢𝔠,S\mathcal{G}_{\mathfrak{c},S}.

Corollary 3.5.

There exists an element g𝒢𝔠g\in\mathcal{G}_{\mathfrak{c}} such that the equality

QA,Sχ1=χ(g)ε(A/F)εS(A/F)(1)rQA,Sχ\mathrm{Q}_{A,S}^{\chi^{-1}}=\chi(g)\cdot\varepsilon(A/F)\cdot\varepsilon_{S}(A/F)\cdot(-1)^{r}\cdot\mathrm{Q}_{A,S}^{\chi}

holds. In particular, if χ=𝟙\chi=\mathbbm{1} is the trivial character, we have

QA,S𝟙0(1)r=ε(A/F)εS(A/F).\mathrm{Q}_{A,S}^{\mathbbm{1}}\neq 0\quad\implies\quad(-1)^{r}=\varepsilon(A/F)\cdot\varepsilon_{S}(A/F).
Proof.

Thanks to the commutative diagram (7) and Theorem 3.3 we have

χ1(drecS(QA,S𝒢𝔠))\displaystyle\chi^{-1}(\textnormal{d}\operatorname{rec}_{S}(\mathrm{Q}_{A,S}^{\mathcal{G}_{\mathfrak{c}}})) =(1)rχ(drecS(QA,S𝒢𝔠))\displaystyle=(-1)^{r}\chi(\textnormal{d}\operatorname{rec}_{S}(\mathrm{Q}_{A,S}^{\mathcal{G}_{\mathfrak{c}}})^{\vee})
=(2)rχ(𝒢𝔠r(S(A/E)𝔠)).\displaystyle=(-2)^{r}\chi(\partial_{\mathcal{G}_{\mathfrak{c}}}^{r}(\mathscr{L}_{S}(A/E)_{\mathfrak{c}})).

Proposition 3.4 implies that there is g𝒢𝔠g\in\mathcal{G}_{\mathfrak{c}} such that

χ(𝒢𝔠r(S(A/E)𝔠))=χ(g)ε(A/F)εS(A/F)χ(𝒢𝔠rS(A/E)𝔠),\chi(\partial_{\mathcal{G}_{\mathfrak{c}}}^{r}(\mathscr{L}_{S}(A/E)_{\mathfrak{c}}))=\chi(g)\cdot\varepsilon(A/F)\cdot\varepsilon_{S}(A/F)\cdot\chi(\partial_{\mathcal{G}_{\mathfrak{c}}}^{r}\mathscr{L}_{S}(A/E)_{\mathfrak{c}}^{\vee}),

and applying Theorem 3.3 once more we obtain the equality

χ1(drecS(QA,S𝒢𝔠))=χ(g)ε(A/F)εS(A/F)(1)rχ(drecS(QA,S𝔠)).\chi^{-1}(\textnormal{d}\operatorname{rec}_{S}(\mathrm{Q}_{A,S}^{\mathcal{G}_{\mathfrak{c}}}))=\chi(g)\cdot\varepsilon(A/F)\cdot\varepsilon_{S}(A/F)\cdot(-1)^{r}\cdot\chi(\textnormal{d}\operatorname{rec}_{S}(\mathrm{Q}_{A,S}^{\mathfrak{c}})).

The claim now follows from the injectivity of drecS\textnormal{d}\operatorname{rec}_{S}. ∎

Remark 3.6.

Corollary 3.5 proves one implication of ([FGM21], Conjecture 1.6) in the CM case.

3.4 Interpolation formula

The square-root pp-adic LL-function S(A/E)𝔠\mathscr{L}_{S}(A/E)_{\mathfrak{c}} does not interpolate values of complex LL-functions but – as the name suggests – a choice of their square-roots.

Definition 3.7.

The anticyclotomic pp-adic LL-function is defined as the product

𝔏S(A/E)𝔠:=S(A/E)𝔠S(A/E)𝔠𝒢𝔠,S.\mathfrak{L}_{S}(A/E)_{\mathfrak{c}}:=\mathscr{L}_{S}(A/E)_{\mathfrak{c}}\cdot\mathscr{L}_{S}(A/E)_{\mathfrak{c}}^{\vee}\ \in\mathbb{Z}\llbracket\mathcal{G}_{\mathfrak{c},S}\rrbracket.

It is this pp-adic LL-function that interpolates special values of complex LL-functions. We introduce some notation to state the interpolation property: the normalized special value of the complex LL-function L(A/E,χ,s)L(A/E,\chi,s) at s=1s=1 is given by

Lnm(A/E,χ):=L(A/E,χ,1)L(πA,Ad,1)νCν(E,πA,χ)L^{\operatorname{nm}}(A/E,\chi):=\frac{L(A/E,\chi,1)}{L(\pi_{A},\operatorname{Ad},1)}\cdot\prod_{\nu\mid\infty}C_{\nu}(E,\pi_{A},\chi)

where L(πA,Ad,s)L(\pi_{A},\operatorname{Ad},s) is the adjoint LL-function of the GL2,F\operatorname{GL}_{2,F}-automorphic representation πA\pi_{A} attached to A/FA_{/F}, and Cν(E,πA,χ)C_{\nu}(E,\pi_{A},\chi) is the Archimedean factor defined in ([FMP17], Section 7B). Since the quadratic extension E/FE/F is CM, the factors Cν(E,πA,χ)C_{\nu}(E,\pi_{A},\chi) are all equal to a fixed constant CC_{\infty} independent of ν\nu, A/FA_{/F}, EE and χ\chi. Note that we always include the Archimedean Euler factors in the definition of the LL-functions and ζ\zeta-functions that occur.

By our assumptions, the automorphic representation πA\pi_{A} admits a Jacquet–Langlands lift to group of units of the totally definite quaternion algebra B/FB_{/F} that is ramified exactly at those finite primes that divide 𝔫\mathfrak{n}^{\flat}. As B/FB_{/F} is totally definite, we may normalize the newform of the Jacquet-Langlands lift such that it takes rational values. Let fABf_{A}^{B} be the rational newform involved in the construction of the pp-adic LL-function in [BG18].

Theorem 3.8.

For all locally constant characters χ:𝒢𝔠,S×\chi\colon\mathcal{G}_{\mathfrak{c},S}\to\mathbb{C}^{\times} we have

χ(𝔏S(A/E)𝔠)=\displaystyle\chi(\mathfrak{L}_{S}(A/E)_{\mathfrak{c}})= 12fAB,fABBLΣA(1,ωE/F)LΣAadd(πA,Ad,1)𝔮ΣASe𝔮(E/F)\displaystyle\frac{1}{2}\cdot\langle f_{A}^{B},f_{A}^{B}\rangle_{B}\cdot L_{\Sigma_{A}}(1,\omega_{E/F})\cdot L_{\Sigma_{A}^{\operatorname{add}}}(\pi_{A},\operatorname{Ad},1)\cdot\prod_{\mathfrak{q}\in\Sigma_{A}\setminus S}e_{\mathfrak{q}}(E/F)
×ζFΣA(2)ΔFΔELnm(A/E,χ,1)𝔮pS𝔠Cord𝔭(𝔠),χ𝔭,A𝔭\displaystyle\times\zeta_{F}^{\Sigma_{A}}(2)\cdot\sqrt{\frac{\Delta_{F}}{\Delta_{E}}}\cdot L^{\operatorname{nm}}(A/E,\chi,1)\cdot\prod_{\mathfrak{q}\mid p_{S}\mathfrak{c}}C_{\operatorname{ord}_{\mathfrak{p}}(\mathfrak{c}),\chi_{\mathfrak{p}},A_{\mathfrak{p}}}

where

  • \bullet

    ,B\langle\cdot,\cdot\rangle_{B} denotes the Petersson inner product on automorphic forms of the unit group of B/FB_{/F},

  • \bullet

    ΣA\Sigma_{A} denotes the set of primes of FF at which A/FA_{/F} has bad reduction, and ΣAaddΣA\Sigma_{A}^{\operatorname{add}}\subset\Sigma_{A} the subset of primes of additive reduction,

  • \bullet

    ωE/F\omega_{E/F} is the quadratic character associated to E/FE/F,

  • \bullet

    e𝔮(E/F)e_{\mathfrak{q}}(E/F) denotes the ramification degree of 𝔮\mathfrak{q} in E/FE/F,

  • \bullet

    ΔF\Delta_{F} are ΔE\Delta_{E} the absolute values of the discriminants of FF and EE respectively,

  • \bullet

    Cord𝔭(𝔠),χ𝔭,AF𝔭χC_{\operatorname{ord}_{\mathfrak{p}}(\mathfrak{c}),\chi_{\mathfrak{p}},A_{F_{\mathfrak{p}}}}\in\mathbb{Q}_{\chi} are constants that only depend on the the 𝔭\mathfrak{p}-adic valuation of 𝔠\mathfrak{c}, the restriction of χ𝔭\chi_{\mathfrak{p}} to a decomposition group at 𝔭\mathfrak{p}, and the base change A/F𝔭A_{/F_{\mathfrak{p}}}.

Proof.

In case the conductor of χ\chi is exactly 𝔠\mathfrak{c} this follows from ([BG18], Theorem 5.8), and the explicit Waldspurger formula of [FMP17]. The general case follows from the norm relations of ([BG18], Theorem 5.8). ∎

4 Artin formalism for plectic points

We keep the same notation as in Subsection 3.3. In addition, we assume that we are in the polyquadratic setup of Subsection 1.2, and that 𝔠=c𝒪F\mathfrak{c}=c\cdot\mathcal{O}_{F} for some ideal cc of FF_{\circ}. Let E,cE_{\circ,c} denote the ring class field of conductor cc and E,c,{}E_{\circ,c,\{\wp\}} the union of the ring class fields of EE_{\circ} of conductor cnc\wp^{n} for n0n\geq 0. We put Gc=Gal(E,c/E)G_{c}=\operatorname{Gal}(E_{\circ,c}/E_{\circ}) and Gc,=Gal(E,c,{}/E)G_{c,\wp}=\operatorname{Gal}(E_{\circ,c,\{\wp\}}/E_{\circ}), so that there are natural maps 𝒢𝔠Gc\mathcal{G}_{\mathfrak{c}}\to G_{c} and 𝒢𝔠,SGc,.\mathcal{G}_{\mathfrak{c},S}\to G_{c,\wp}.

Assumption 4.1.

We require that

  • \bullet

    every prime divisor of cc is completely split in F/FF/F_{\circ} and unramified in E/FE/F_{\circ},

  • \bullet

    the character χ:𝒢𝔠×\chi\colon\mathcal{G}_{\mathfrak{c}}\to\mathbb{C}^{\times} is the restriction of an anticyclotomic character ξ:Gc×\xi\colon G_{c}\to\mathbb{C}^{\times}.

Remark 4.2.

Since every non-trivial subextension of F/FF/F_{\circ} is ramified by Assumption 1.5, requiring the ideal cc to be unramified in E/FE/F_{\circ} implies that the fields EE and E,cE_{\circ,c} are linearly disjoint over EE_{\circ}. We impose the total splitting in F/FF/F_{\circ} of the prime divisors of cc just as a simplifying hypothesis for the proof of Proposition 4.5.

4.1 Factorization of complex LL-functions

We write 𝔊:=Homgr(𝔊,{±1})\mathfrak{G}^{\star}:=\operatorname{Hom}_{\mathrm{gr}}(\mathfrak{G},\{\pm 1\}) for the Pontryagin dual of 𝔊=Gal(F/F)Gal(E/E)\mathfrak{G}=\operatorname{Gal}(F/F_{\circ})\cong\operatorname{Gal}(E/E_{\circ}). For any number field Ω\Omega we write GΩ:=Gal(¯/Ω)G_{\Omega}:=\operatorname{Gal}(\overline{\mathbb{Q}}/\Omega) for its absolute Galois group.

Lemma 4.3.

There exists an isomorphism of GFG_{F_{\circ}}-representations:

IndGEGF(χ)η𝔊IndGEGF(ξη).\operatorname{Ind}_{G_{E}}^{G_{F_{\circ}}}(\chi)\cong\bigoplus_{\eta\in\mathfrak{G}^{\star}}\operatorname{Ind}_{G_{E_{\circ}}}^{G_{F_{\circ}}}(\xi\cdot\eta).
Proof.

First, we claim that

IndGEGE(χ)η𝔊ξη.\operatorname{Ind}_{G_{E}}^{G_{E_{\circ}}}(\chi)\cong\bigoplus_{\eta\in\mathfrak{G}^{\star}}\xi\cdot\eta.

To prove the claim note that all the characters ξη\xi\cdot\eta, for η𝔊\eta\in\mathfrak{G}^{\star}, have the same restriction to GEG_{E}, namely the character χ\chi. By Frobenius reciprocity we have

HomGE(ξη,IndGEGE(χ))HomGE(χ,χ)0\operatorname{Hom}_{G_{E_{\circ}}}\big{(}\xi\cdot\eta,\operatorname{Ind}_{G_{E}}^{G_{E_{\circ}}}(\chi)\big{)}\cong\operatorname{Hom}_{G_{E}}\big{(}\chi,\chi\big{)}\not=0

and, therefore, the claim follows from semi-simplicity of representations of finite groups in characteristic zero. The statement of the lemma follows from transitivity of induction. ∎

Corollary 4.4.

The following equality of complex LL-functions holds:

Lnm(A/E,χ,s)=η𝔊Lnm(Aη/E,ξ,s).L^{\operatorname{nm}}(A/E,\chi,s)=\prod_{\eta\in\mathfrak{G}^{\star}}L^{\operatorname{nm}}(A^{\eta}_{\circ}/E_{\circ},\xi,s).
Proof.

This is a direct consequence of Lemma 4.3 and Artin formalism for complex LL-functions. ∎

4.2 Factorization of pp-adic LL-functions

For each η𝔊\eta\in\mathfrak{G}^{\star}, the quadruple (Aη,E,,c)(A_{\circ}^{\eta},E_{\circ},\wp,c) fulfils the conditions of Subsection 3.3 and thus, we can define the pp-adic L-functions

{}(Aη/E)c,𝔏{}(Aη/E)cGc,.\mathscr{L}_{\{\wp\}}(A_{\circ}^{\eta}/E_{\circ})_{c},\ \mathfrak{L}_{\{\wp\}}(A_{\circ}^{\eta}/E_{\circ})_{c}\in\mathbb{Z}\llbracket G_{c,\wp}\rrbracket.

Let RR be a commutative ring. The homomorphism 𝒢𝔠,SGc,\mathcal{G}_{\mathfrak{c},S}\longrightarrow G_{c,\wp} induces the restriction map

resS,:R𝒢𝔠,SRGc,\operatorname{\operatorname{res}}_{S,\wp}\colon R\llbracket\mathcal{G}_{\mathfrak{c},S}\rrbracket\longrightarrow R\llbracket G_{c,\wp}\rrbracket

between completed group algebras.

Proposition 4.5.

There is a constant C×C\in\mathbb{Q}^{\times} such that the equality

resS,(𝔏S(A/E)𝔠)=Cη𝔊𝔏{}(Aη/E)c.\operatorname{\operatorname{res}}_{S,\wp}\left(\mathfrak{L}_{S}(A/E)_{\mathfrak{c}}\right)=C\cdot\prod_{\eta\in\mathfrak{G}^{\star}}\mathfrak{L}_{\{\wp\}}(A_{\circ}^{\eta}/E_{\circ})_{c}.

holds in Gc,\mathbb{Z}\llbracket G_{c,\wp}\rrbracket.

Proof.

To prove the statement it is enough to use Theorem 3.8 to show equality of both sides after evaluation at every finite order character of Gc,G_{c,\wp}. We note that the first line of the interpolation formula is a non-zero rational number and, thus, we may neglect it. By Corollary 4.4, the normalized special values on both sides cancel out. In addition, the local constants Cord𝔭(𝔠),χ𝔭,A𝔭C_{\operatorname{ord}_{\mathfrak{p}}(\mathfrak{c}),\chi_{\mathfrak{p}},A_{\mathfrak{p}}} cancel out as well because we assumed that all the primes of FF_{\circ} dividing cc\wp are totally split in F/FF/F_{\circ}. As we are taking a product over r=2tr=2^{t} factors on the right hand side, we see that it suffices to show that

ΔFΔEζF(2)ζF(2)r\sqrt{\frac{\Delta_{F}}{\Delta_{E}}}\cdot\frac{\zeta_{F}(2)}{\zeta_{F_{\circ}}(2)^{r}}

is a rational number. Using the functional equation of the Dedekind zeta function, we may rewrite this term as

ΔFΔEζF(1)ΔF3/2ζF(1)rΔF3r/2.\sqrt{\frac{\Delta_{F}}{\Delta_{E}}}\cdot\frac{\zeta_{F}(-1)\Delta_{F}^{-3/2}}{\zeta_{F_{\circ}}(-1)^{r}\Delta_{F_{\circ}}^{-3r/2}}.

By the Klingen–Siegel Theorem ([Sie37], [Kli62]), the special values of the Dedekind zeta function (excluding the Archimedean Euler factors) at negative integers are rational. The Archimedean factors of ζF(s)\zeta_{F}(s) and ζF(s)r\zeta_{F_{\circ}}(s)^{r} agree, so we are left to show that ΔE×\Delta_{E}\in\mathbb{Q}^{\times} is a square. Let ΔE/F\Delta_{E/F_{\circ}} denote the relative discriminant of E/FE/F_{\circ}. The formula for relative discriminants in towers gives

ΔE=NF/(ΔE/F)ΔF[E:F].\Delta_{E}=\operatorname{N}_{F_{\circ}/\mathbb{Q}}(\Delta_{E/F_{\circ}})\cdot\Delta_{F_{\circ}}^{[E:F_{\circ}]}.

As [E:F][E:F_{\circ}] is a power of 22, it suffices to show that ΔE/F\Delta_{E/F_{\circ}} is a square. At this point we conclude the argument because ([Kha19], Theorem 1.2) implies

ΔE/F=ΔE/FrΔF/F2\Delta_{E/F_{\circ}}=\Delta_{E_{\circ}/F_{\circ}}^{r}\cdot\Delta_{F/F_{\circ}}^{2}

since the relative discriminants ΔE/F\Delta_{E_{\circ}/F_{\circ}} and ΔF/F\Delta_{F/F_{\circ}} are coprime by Assumption 1.5. ∎

4.3 Factorization of plectic invariants

Recall that by Assumption 4.1 the character χ\chi is the restriction to 𝒢𝔠\mathcal{G}_{\mathfrak{c}} of an anticyclotomic character ξ:Gc×\xi\colon G_{c}\to\mathbb{C}^{\times}. For every η𝔊\eta\in\mathfrak{G}^{\star} we can consider the plectic pp-adic invariant

QAη,{}ξE^,ξ\mathrm{Q}_{A_{\circ}^{\eta},\{\wp\}}^{\xi}\in\widehat{E}_{\circ,\wp}^{-}\otimes_{\mathbb{Z}}\mathbb{Q}_{\xi}

attached to the triple (Aη/F,ξ,{})(A_{\circ}^{\eta}/F_{\circ},\xi,\{\wp\}). As in Subsection 1.3, we use the fixed identifications E𝔭=E,E_{\mathfrak{p}}^{-}=E_{\circ,\wp}^{-} for 𝔭S\mathfrak{p}\in S to define the norm map

NS/:E^S,(E,)rSympr(E^,).\operatorname{N}_{S/\wp}\colon\widehat{E}_{S,\otimes}^{-}\longrightarrow(E_{\circ,\wp}^{-})^{\otimes r}\longrightarrow\operatorname{Sym}^{r}_{\mathbb{Z}_{p}}(\widehat{E}_{\circ,\wp}^{-}).
Corollary 4.6.

There exists a constant Cχχ×C_{\chi}\in\mathbb{Q}_{\chi}^{\times} such that the equality

NS/(QA,Sχ)2=Cχη𝔊(QAη,{}ξ)2\operatorname{N}_{S/\wp}\big{(}\mathrm{Q}_{A,S}^{\chi}\big{)}^{2}=C_{\chi}\cdot\prod_{\eta\in\mathfrak{G}^{\star}}\big{(}\mathrm{Q}_{A^{\eta}_{\circ},\{\wp\}}^{\xi}\big{)}^{2}

holds in Symp2r(E^,)χ\operatorname{Sym}^{2r}_{\mathbb{Z}_{p}}(\widehat{E}_{\circ,\wp}^{-})_{\mathbb{Q}_{\chi}}. Moreover, NS/(QA,Sχ)0\operatorname{N}_{S/\wp}(\mathrm{Q}_{A,S}^{\chi})\neq 0 implies QAη,{}ξ0\mathrm{Q}_{A^{\eta}_{\circ},\{\wp\}}^{\xi}\neq 0 for all η𝔊\eta\in\mathfrak{G}^{\star}.

Proof.

The second claim follows from the first because Symp(E^,)χ\operatorname{Sym}_{\mathbb{Z}_{p}}(\widehat{E}_{\circ,\wp}^{-})_{\mathbb{Q}_{\chi}} has no nilpotent elements. By Proposition 4.5 and Proposition 3.4 there exists C×C\in\mathbb{Q}^{\times} such that the equality

resS,(S(A/E)𝔠)2=Cη𝔊{}(Aη/E)c2\operatorname{\operatorname{res}}_{S,\wp}\left(\mathscr{L}_{S}(A/E)_{\mathfrak{c}}\right)^{2}=C\cdot\prod_{\eta\in\mathfrak{G}^{\star}}\mathscr{L}_{\{\wp\}}(A_{\circ}^{\eta}/E_{\circ})_{c}^{2}

holds up to multiplication by an element in Gc,G_{c,\wp}. Lemma 2.2 (b) implies that the diagram

E^S,\textstyle{\widehat{E}_{S,\otimes}^{-}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}drecS\scriptstyle{\textnormal{d}\operatorname{rec}_{S}}NS/\scriptstyle{\operatorname{N}_{S/\wp}}(I𝔠r/I𝔠r+1)\textstyle{(I_{\mathfrak{c}}^{r}/I_{\mathfrak{c}}^{r+1})_{\mathbb{Q}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}resS,\scriptstyle{\operatorname{\operatorname{res}}_{S,\wp}}E^,\textstyle{\widehat{E}_{\circ,\wp}^{-}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}drec\scriptstyle{\textnormal{d}\operatorname{rec}_{\wp}}(Icr/Icr+1)\textstyle{(I_{c}^{r}/I_{c}^{r+1})_{\mathbb{Q}}}

commutes. Thus, we may conclude by invoking Theorem 3.3. ∎

Assume that pp does not split in the field χ\mathbb{Q}_{\chi}, i.e., that χ,p:=χp\mathbb{Q}_{\chi,p}:=\mathbb{Q}_{\chi}\otimes_{\mathbb{Q}}\mathbb{Q}_{p} is a field. Then, Lemma 2.1 allows us to deduce the following corollary.

Corollary 4.7.

Assume that pp does not split in χ\mathbb{Q}_{\chi}, then there exists a square-root Cχχ,p×\sqrt{C_{\chi}}\in\mathbb{Q}_{\chi,p}^{\times} of CχχC_{\chi}\in\mathbb{Q}_{\chi} such that the equality

NS/(QA,Sχ)=Cχη𝔊QAη,{}ξ\operatorname{N}_{S/\wp}(\mathrm{Q}_{A,S}^{\chi})=\sqrt{C_{\chi}}\cdot\prod_{\eta\in\mathfrak{G}^{\star}}\mathrm{Q}_{A^{\eta}_{\circ},\{\wp\}}^{\xi}

holds in Sympr(E^,)χ\operatorname{Sym}^{r}_{\mathbb{Z}_{p}}(\widehat{E}_{\circ,\wp}^{-})_{\mathbb{Q}_{\chi}}. Moreover, NS/(QA,Sχ)0\operatorname{N}_{S/\wp}(\mathrm{Q}_{A,S}^{\chi})\neq 0 if and only if QAη,{}ξ0\mathrm{Q}_{A^{\eta}_{\circ},\{\wp\}}^{\xi}\neq 0 for all η𝔊\eta\in\mathfrak{G}^{\star}.

5 Algebraicity and arithmetic significance of plectic points

We keep the same notation as in Section 4. In particular, we are in the polyquadratic CM setting and Assumptions 1.1, 1.5, 1.7 and 4.1 hold. Moreover, we suppose that the conductor of the character ξ\xi is cc. One can always achieve this by shrinking cc. We have the equalities

r(A/E,χ)=η𝔊r(Aη/E,ξ),{an,alg},r_{\scalebox{0.75}{$\bullet$}}(A/E,\chi)=\sum_{\eta\in\mathfrak{G}^{\star}}r_{\scalebox{0.75}{$\bullet$}}(A_{\circ}^{\eta}/E_{\circ},\xi),\qquad\scalebox{0.75}{$\bullet$}\in\{\mathrm{an},\mathrm{alg}\}, (12)

which imply ran(A/E,χ)rr_{\mathrm{an}}(A/E,\chi)\geq r because ran(Aη/E,ξ)r_{\mathrm{an}}(A_{\circ}^{\eta}/E_{\circ},\xi) is odd for every character η𝔊\eta\in\mathfrak{G}^{\star}.

Theorem 5.1.

The following implication holds:

NS/(prS(PA,Sχ))0ralg(A/E,χ)=r&ran(A/E,χ)=r.\operatorname{N}_{S/\wp}\big{(}\operatorname{pr}^{-}_{S}(\mathrm{P}^{\chi}_{A,S})\big{)}\neq 0\quad\implies\quad r_{\mathrm{alg}}(A/E,\chi)=r\quad\&\quad r_{\mathrm{an}}(A/E,\chi)=r.

Suppose pp does not split in χ\mathbb{Q}_{\chi} and that the character χ\chi is not quadratic, then

ran(A/E,χ)=rNS/(prS(PA,Sχ))0&ralg(A/E,χ)=r.r_{\mathrm{an}}(A/E,\chi)=r\quad\implies\quad\operatorname{N}_{S/\wp}\big{(}\operatorname{pr}^{-}_{S}(\mathrm{P}^{\chi}_{A,S})\big{)}\neq 0\quad\&\quad r_{\mathrm{alg}}(A/E,\chi)=r.
Proof.

The non-vanishing of NS/(prS(PA,Sχ))\operatorname{N}_{S/\wp}\big{(}\operatorname{pr}^{-}_{S}(\mathrm{P}^{\chi}_{A,S})\big{)} implies that for every character η𝔊\eta\in\mathfrak{G}^{\star} the point prS(PAη,{}ξ)\operatorname{pr}^{-}_{S}(\mathrm{P}^{\xi}_{A_{\circ}^{\eta},\{\wp\}}) is non-zero by Corollary 4.6. Hence, the point PAη,{}ξ\mathrm{P}^{\xi}_{A^{\eta},\{\wp\}} is non-zero for every η𝔊\eta\in\mathfrak{G}^{\star} and, thus, Proposition 3.1 gives

ralg(Aη/E,ξ)=ran(Aη/E,ξ)=1η𝔊.r_{\mathrm{alg}}(A_{\circ}^{\eta}/E_{\circ},\xi)=r_{\mathrm{an}}(A_{\circ}^{\eta}/E_{\circ},\xi)=1\quad\forall\ \eta\in\mathfrak{G}^{\star}.

The first claim follows from (12). The second claim is proved similarly using Corollary 4.7. ∎

Set A/F+=A/FA^{+}_{/F}=A_{/F} and denote by A/FA^{-}_{/F} the quadratic twist of A/FA_{/F} with respect to the extension E/FE/F. We partition S=S+SS=S^{+}\cup S^{-} by declaring that the subset S+SS^{+}\subseteq S contains all the primes in SS of split multiplicative reduction for A/F+A_{/F}^{+}, and define ϱA(S):=max{ralg(A±/F)+|S±|}.\varrho_{A}(S):=\max\big{\{}r_{\mathrm{alg}}(A^{\pm}/F)+\lvert S^{\pm}\rvert\big{\}}.

Theorem 5.2.

We also have

NS/(pr(PA,S𝟙))0ran(A/E)=r&ϱA(S)=r.\operatorname{N}_{S/\wp}\big{(}\operatorname{pr}^{-}(\mathrm{P}^{\mathbbm{1}}_{A,S})\big{)}\neq 0\quad\iff\quad r_{\mathrm{an}}(A/E)=r\quad\&\quad\varrho_{A}(S)=r.
Proof.

Under our hypotheses we either have S=S+S=S^{+} or S=SS=S^{-}. For simplicity we assume that S=S+S=S^{+}, or equivalently that a=+1a_{\wp}=+1, since the other case is proved by similar arguments.

Suppose NS/(prS(PA,S𝟙))\operatorname{N}_{S/\wp}\big{(}\operatorname{pr}^{-}_{S}(\mathrm{P}^{\mathbbm{1}}_{A,S})\big{)} is non-zero. By Theorem 5.1, it is enough to show that ϱS(A)=r\varrho_{S}(A)=r. Now, Corollary 4.7 implies that pr(PAη,𝟙)0\operatorname{pr}^{-}_{\wp}(\mathrm{P}^{\mathbbm{1}}_{A_{\circ}^{\eta},\wp})\neq 0 for every η𝔊\eta\in\mathfrak{G}^{\star}, and since a=+1a_{\wp}=+1, Remark 3.2 tells us that ralg(Aη/F)=0r_{\mathrm{alg}}(A_{\circ}^{\eta}/F_{\circ})=0 for every η𝔊\eta\in\mathfrak{G}^{\star}. We deduce ralg(A+/F)=0r_{\mathrm{alg}}(A^{+}/F)=0 and ralg(A/F)=rr_{\mathrm{alg}}(A^{-}/F)=r as required. For the converse implication, note that the assumptions imply that ralg(Aη/E)=1r_{\mathrm{alg}}(A_{\circ}^{\eta}/E_{\circ})=1 and ralg(Aη/F)=0r_{\mathrm{alg}}(A_{\circ}^{\eta}/F_{\circ})=0 for every η𝔊\eta\in\mathfrak{G}^{\star}. Then, Remark 3.2 shows that pr(PAη,𝟙)0\operatorname{pr}^{-}_{\wp}(\mathrm{P}^{\mathbbm{1}}_{A_{\circ}^{\eta},\wp})\neq 0 for every η𝔊\eta\in\mathfrak{G}^{\star} and the claim follows from Corollary 4.7. ∎

As in Subsection 3.2 we fix an embedding ι:E,cE,\iota_{\wp}\colon E_{\circ,c}\hookrightarrow E_{\circ,\wp} extending the canonical embedding EE,E_{\circ}\hookrightarrow E_{\circ,\wp}. Furthermore, we choose an embedding ι𝔭1:E𝔠E𝔭1\iota_{\mathfrak{p}_{1}}\colon E_{\mathfrak{c}}\hookrightarrow E_{\mathfrak{p}_{1}} that restricts to the chosen embedding ι\iota_{\wp} as well as to the canonical embedding EE𝔭1E\hookrightarrow E_{\mathfrak{p}_{1}}. Under our assumptions, the fields EE and E,cE_{\circ,c} are linearly disjoint over EE_{\circ}, thus for every 𝔭S\mathfrak{p}\in S we can choose an element τ𝔭Gal(E𝔠/E,c)\tau_{\mathfrak{p}}\in\operatorname{Gal}(E_{\mathfrak{c}}/E_{\circ,c}) whose restriction to Gal(E/E)𝔊\operatorname{Gal}(E/E_{\circ})\cong\mathfrak{G} sends 𝔭\mathfrak{p} to 𝔭1\mathfrak{p}_{1}, and such that ι𝔭:=ι𝔭1τ𝔭\iota_{\mathfrak{p}}:=\iota_{\mathfrak{p}_{1}}\circ\tau_{\mathfrak{p}} corresponds to the prime 𝔭\mathfrak{p}. Then, we define the determinant map det:r(A(E𝔠))A^(ES)\operatorname{det}\colon\mathop{\mathchoice{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptscriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}}(A(E_{\mathfrak{c}})_{\mathbb{Q}})\to\widehat{A}(E_{S})_{\mathbb{Q}} by setting

det(P1Pr)=det(ιA,𝔭1(P1)ιA,𝔭r(P1)ιA,𝔭1(Pr)ιA,𝔭r(Pr)).\operatorname{det}\big{(}P_{1}\wedge\dots\wedge P_{r}\big{)}=\operatorname{det}\begin{pmatrix}\iota_{A,\mathfrak{p}_{1}}(P_{1})&\dots&\iota_{A,\mathfrak{p}_{r}}(P_{1})\\ \vdots&\ddots&\vdots\\ \iota_{A,\mathfrak{p}_{1}}(P_{r})&\dots&\iota_{A,\mathfrak{p}_{r}}(P_{r})\end{pmatrix}.

For every η𝔊\eta\in\mathfrak{G}^{\star} we may view Aη(E,c)A_{\circ}^{\eta}(E_{\circ,c}) as a subgroup of A(E𝔠)A(E_{\mathfrak{c}}), and by construction we have

ιA,𝔭(P)=η(τ𝔭)ιAη,(P)PAη(E,c).\displaystyle\iota_{A,\mathfrak{p}}(P)=\eta(\tau_{\mathfrak{p}})\cdot\iota_{A_{\circ}^{\eta},\wp}(P)\qquad\forall\ P\in A_{\circ}^{\eta}(E_{\circ,c}). (13)
Theorem 5.3.

Suppose pp does not split in χ\mathbb{Q}_{\chi}. There exists a quadratic extension Ωχ/χ\Omega_{\chi}/\mathbb{Q}_{\chi}, in which pp splits, and an element wA,Sχr(A(E𝔠)Ωχχ)w^{\chi}_{A,S}\in\mathop{\mathchoice{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\textstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}{\raisebox{0.0pt}{$\scriptscriptstyle\bigwedge$}^{\mspace{-2.0mu}r}\kern-0.5pt}}(A(E_{\mathfrak{c}})^{\chi}_{\Omega_{\chi}}) such that

NS/(prS(PA,Sχ))=NS/(prS(det(wA,Sχ)))\operatorname{N}_{S/\wp}\big{(}\operatorname{pr}^{-}_{S}(\mathrm{P}^{\chi}_{A,S})\big{)}=\operatorname{N}_{S/\wp}\big{(}\operatorname{pr}^{-}_{S}(\operatorname{det}(w_{A,S}^{\chi}))\big{)}

in Sympr(A^(E,))χ.\operatorname{Sym}^{r}_{\mathbb{Z}_{p}}(\widehat{A}(E_{\circ,\wp}))_{\mathbb{Q}_{\chi}}.

Proof.

We choose an ordering η{1},,η{r}\eta\{1\},\ldots,\eta\{r\} of the elements of the character group 𝔊\mathfrak{G}^{\star}. For η𝔊\eta\in\mathfrak{G}^{\star} we consider the Heegner point PAηξP_{A^{\eta}_{\circ}}^{\xi} of equation (9), and set

w~A,Sχ:=PAη{1}ξPAη{r}ξ.\widetilde{w}^{\chi}_{A,S}:=P_{A^{\eta\{1\}}_{\circ}}^{\xi}\wedge\ldots\wedge P_{A^{\eta\{r\}}_{\circ}}^{\xi}.

Using (13), the formula for the determinant gives

NS/(det(w~A,Sχ))\displaystyle\operatorname{N}_{S/\wp}\big{(}\operatorname{det}(\widetilde{w}^{\chi}_{A,S})\big{)} =NS/(σSrsgn(σ)ι𝔭1(PAη{σ(1)}ξ)ι𝔭r(PAη{σ(r)}ξ))\displaystyle=\operatorname{N}_{S/\wp}\left(\sum_{\sigma\in S_{r}}\mathrm{sgn}(\sigma)\cdot\iota_{\mathfrak{p}_{1}}\big{(}P_{A^{\eta\{\sigma(1)\}}_{\circ}}^{\xi}\big{)}\otimes\ldots\otimes\iota_{\mathfrak{p}_{r}}\big{(}P_{A^{\eta\{\sigma(r)\}}_{\circ}}^{\xi}\big{)}\right)
=C𝔊η𝔊ιAη,(PAηξ),\displaystyle=C_{\mathfrak{G}}\cdot\prod_{\eta\in\mathfrak{G}^{\star}}\iota_{A_{\circ}^{\eta},\wp}\big{(}P_{A^{\eta}_{\circ}}^{\xi}\big{)},

where

C𝔊:=σSrsgn(σ)i=1rη{σ(i)}(τ𝔭i)C_{\mathfrak{G}}:=\sum_{\sigma\in S_{r}}\mathrm{sgn}(\sigma)\cdot\prod_{i=1}^{r}\eta\{\sigma(i)\}(\tau_{\mathfrak{p}_{i}})

is the determinant of the character table of the group 𝔊\mathfrak{G}. By orthogonality of characters, the determinant is non-zero. In fact, it is equal to ±rr/2\pm r^{r/2}. Let kAη,ξ×k_{A_{\circ}^{\eta},\wp}^{\xi}\in\mathbb{Q}^{\times} be the constants appearing in (10), and Cχχ,p×\sqrt{C_{\chi}}\in\mathbb{Q}_{\chi,p}^{\times} the constant appearing in Corollary 4.7. By setting

wA,Sχ:=CχC𝔊η𝔊kAη,ξw~A,Sχw^{\chi}_{A,S}:=\frac{\sqrt{C_{\chi}}}{C_{\mathfrak{G}}\cdot\prod_{\eta\in\mathfrak{G}^{\star}}k_{A_{\circ}^{\eta},\wp}^{\xi}}\cdot\widetilde{w}^{\chi}_{A,S}

we get that

NS/(det(wA,Sχ))=Cχη𝔊PAη,{}ξ.\operatorname{N}_{S/\wp}\big{(}\operatorname{det}(w^{\chi}_{A,S})\big{)}=\sqrt{C_{\chi}}\cdot\prod_{\eta\in\mathfrak{G}^{\star}}\mathrm{P}_{A^{\eta}_{\circ},\{\wp\}}^{\xi}.

The claim follows from Corollary 4.7 after applying the minus projector on both sides. ∎

References

  • [BD98] M. Bertolini and H. Darmon. Heegner points, pp-adic LL-functions, and the Cerednik-Drinfeld uniformization. Invent. Math., 131(3):453–491, 1998.
  • [BG18] F. Bergunde and L. Gehrmann. Leading terms of anticyclotomic Stickelberger elements and pp-adic periods. Trans. Amer. Math. Soc., 370(9):6297–6329, 2018.
  • [Dar01] H. Darmon. Integration on p×\mathcal{H}_{p}\times\mathcal{H} and arithmetic applications. Ann. of Math. (2), 154(3):589–639, 2001.
  • [FG21] M. Fornea and L. Gehrmann. Plectic Stark-Heegner points. Preprint, 2021.
  • [FGM21] M. Fornea, X. Guitart, and M. Masdeu. Plectic pp-adic invariants. preprint, 2021.
  • [FMP17] D. File, K. Martin, and A. Pitale. Test vectors and central LL-values for GL(2){\mathrm{G}L}(2). Algebra Number Theory, 11(2):253–318, 2017.
  • [GMM20] X. Guitart, M. Masdeu, and S. Molina. An automorphic approach to Darmon points. Indiana Univ. Math. J., 69(4):1251–1274, 2020.
  • [HM22] V. Hernández and S. Molina. Plectic points and Hida-Rankin p-adic L-functions. preprint, 2022.
  • [Kha19] S. K. Khanduja. The discriminant of compositum of algebraic number fields. Int. J. Number Theory, 15(2):353–360, 2019.
  • [Kli62] H. Klingen. Über die Werte der Dedekindschen Zetafunktion. Math. Ann., 145:265–272, 1961/62.
  • [Mok10] C. P. Mok. Special values of LL-functions of elliptic curves over \mathbb{Q} and their base change to real quadratic fields. J. Number Theory, 130(2):431–438, 2010.
  • [Mok11] C. P. Mok. Heegner points and pp-adic LL-functions for elliptic curves over certain totally real fields. Comment. Math. Helv., 86(4):867–945, 2011.
  • [Nek07] J. Nekovář. The Euler system method for CM points on Shimura curves, page 471–547. London Mathematical Society Lecture Note Series. Cambridge University Press, 2007.
  • [Nek16] J. Nekovár. Some remarks on the BSD conjecture. Rubinfest’s talk at Harvard, 2016.
  • [NS16] J. Nekovář and A. J. Scholl. Introduction to plectic cohomology. In Advances in the theory of automorphic forms and their LL-functions, volume 664 of Contemp. Math., pages 321–337. Amer. Math. Soc., Providence, RI, 2016.
  • [Sie37] C. L. Siegel. Über die analytische Theorie der quadratischen Formen. III. Ann. of Math. (2), 38(1):212–291, 1937.
  • [Zha01] S. Zhang. Gross-Zagier formula for GL2{\mathrm{G}L}_{2}. Asian J. Math., 5(2):183–290, 2001.