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

Finite slope Triple product pp-adic LL-functions over totally real number fields

Santiago Molina
Abstract.

We construct pp-adic LL-functions associated with triples of finite slope pp-adic families of quaternionic automorphic eigenforms over totally real fieldson Shimura curves. These results generalize a previous construction, joint work with D.Barrera, performed in the ordinary setting.

1. Introduction

The theory of pp-adic L-functions has grown tremendously during the last decades due to its important arithmetic applications. In particular, it has been essential for recent developments towards the Birch and Swinnerton-Dyer conjecture and its generalizations. The seminal work of Kato provided deep results on the Birch and Swinnerton-Dyer conjecture in rank 0 for twists of elliptic curves over \operatorname{\mathbb{Q}} by Dirichlet characters. More recently, similar results with twists by certain Artin representations of dimension 2 and 4 have been obtained by Bertolini-Darmon-Rotger in [4] and Darmon-Rotger in [9] using analogous methods. In [17], [16], [18], [1] and [5] the development of such methods has been extended in different directions, as for example bounding certain Selmer groups and treating finite slope settings. In all these situations the pp-adic LL-function attached to a triple of pp-adic families of modular forms has played a very important role. Such triple product pp-adic LL-functions were constructed in [12], [8], [13] in the ordinary case and in [1] for Coleman Families, namely, in the finite slope situation.

This paper deals with the construction of triple product pp-adic L-functions associated with families of quaternionic automorphic forms in Shimura curves over totally real fields in the finite slope situation. In a previous paper with D. Barrera [3], we constructed such pp-adic L-functions in the ordinary setting. In fact, we were able to construct pp-adic families of finite slope quaternionic automorphic forms over totally real fields, but we had to restrict ourselves to the ordinary setting for the construction of the corresponding triple product pp-adic L-functions. The reason for such a restriction was the inability to iterate the Gauss-Manin connection in the finite slope situation. In this article we manage to overcome these obstacles by following the strategy of Andreatta and Iovita in [1].

Our setting is in many aspects analogous to the setting of elliptic modular forms treated in [1], hence many of the techniques will be similar. The main novelty of this paper with respect to [1] is the extension of the χ\chi-iteration of the Gauss-Manin connection to the universal character. This allows us to extend the definition of triple product pp-adic L-function to the whole weight space, removing the crucial [1, Assumption 4.7]. This also improves and complete the results in [1] when F=F=\operatorname{\mathbb{Q}}.

Our strategy relies on defining the space of nearly overconvergent families using the theory of formal vector bundles also introduced in [1]. These formal vector bundles have been also used in [11] to perform similar constructions in case of Hilbert modular schemes. However, we define a new type of formal vector bundle, already used in [3] to define pp-adic families of finite slope quaternionic automorphic forms (§4.2). During the elaboration of this paper, Andreatta and Iovita reported to the author that in parallel they had reached the same new definition of formal vector bundle. This space of nearly overconvergent families admits a good definition of a Gauss-Manin connection (Theorem 5.5). We use the explicit description of the universal character given in §4.3.1 to define a formal power series of endomorphisms liable to be the universal iteration of the Gauss-Manin connection. The choice of this new type of formal vector bundle ensures the convergence of such power series, as verified using qq-expansions on Serre-Tate coordinates. As mentioned before, these computations also apply in the setting over \operatorname{\mathbb{Q}}, hence as a consequence one can define triple product pp-adic L-function in the finite slope situation over \operatorname{\mathbb{Q}} without any restriction on the input weights ([1, Assumption 5.4]), improving the results known so far.

1.1. Main result

Let FF be a totally real number field with integer ring 𝒪F\mathcal{O}_{F} and fix a real embedding τ0\tau_{0}. Let p>2p>2 be a prime number and let 𝔭0\mathfrak{p}_{0} the prime over pp associated to τ0\tau_{0} under a fixed embedding ¯p\bar{\operatorname{\mathbb{Q}}}\hookrightarrow\operatorname{\mathbb{C}}_{p}. During all this paper we will suppose that F𝔭0=pF_{\mathfrak{p}_{0}}=\operatorname{\mathbb{Q}}_{p}.

Let BB be a quaternion algebra over FF that splits at any prime over pp, and it is ramified at any real place but τ0\tau_{0}. In [3], we construct pp-adic families of automorphic forms on (B𝔸)×(B\otimes\operatorname{\mathbb{A}})^{\times} as global sections of certain pp-adic sheaves on certain unitary Shimura curves, extending the previous work of [6]. Such families are parametriced by the (d+1)(d+1)-dimensional Iwasawa algebra Λ=p[[(𝒪Fp)××p×]]\Lambda=\operatorname{\mathbb{Z}}_{p}[[(\mathcal{O}_{F}\otimes\operatorname{\mathbb{Z}}_{p})^{\times}\times\operatorname{\mathbb{Z}}_{p}^{\times}]].

Let μ1,μ2,μ3\mu_{1},\mu_{2},\mu_{3} be three finite slope pp-adic families of automorphic eigenforms for BB and we denote by Λ1\Lambda_{1}, Λ2\Lambda_{2} and Λ3\Lambda_{3} the rings over which they are defined, respectively. Let x,y,zx,y,z be a triple of classical points corresponding to ((k¯1,ν1),(k¯2,ν2),(k¯3,ν3))((\underline{k}_{1},\nu_{1}),(\underline{k}_{2},\nu_{2}),(\underline{k}_{3},\nu_{3})), where μi\mu_{i}\in\operatorname{\mathbb{N}} and k¯i[F:]\underline{k}_{i}\in\operatorname{\mathbb{N}}^{[F:\operatorname{\mathbb{Q}}]}. We denote by πx\pi_{x} the automorphic representation of (B𝔸F)×(B\otimes\operatorname{\mathbb{A}}_{F})^{\times} generated by the automorphic form obtained from the specialization of μ1\mu_{1} at xx, and Πx\Pi_{x} the corresponding cuspidal automorphic representation of GL2(𝔸F)\mathrm{GL}_{2}(\operatorname{\mathbb{A}}_{F}). We also denote by αx𝔭\alpha_{x}^{\mathfrak{p}} and βx𝔭\beta_{x}^{\mathfrak{p}} the roots of the Hecke polynomial at 𝔭\mathfrak{p}. In the same way we obtain πy\pi_{y}, Πy\Pi_{y}, αy𝔭\alpha_{y}^{\mathfrak{p}}, βy𝔭\beta_{y}^{\mathfrak{p}}, πz\pi_{z}, Πz\Pi_{z}, αz𝔭\alpha_{z}^{\mathfrak{p}}, βz𝔭\beta_{z}^{\mathfrak{p}}.

We have (see Lemma 6.8 and Theorem 6.10 for more details):

Theorem 1.1.

There exists p(μ1,μ2,μ3)Λ1^Λ2^Frac(Λ3)\mathcal{L}_{p}(\mu_{1},\mu_{2},\mu_{3})\in\Lambda_{1}\hat{\otimes}\Lambda_{2}\hat{\otimes}\mathrm{Frac}(\Lambda_{3}) such that for each classical point (x,y,z)(x,y,z) corresponding to a triple ((k¯1,ν1),(k¯2,ν2),(k¯3,ν3))((\underline{k}_{1},\nu_{1}),(\underline{k}_{2},\nu_{2}),(\underline{k}_{3},\nu_{3})) we have:

p(μ1,μ2,μ3)(x,y,z)=K(𝔭p𝔭(x,y,z)𝔭,1(z))L(1ν1ν2ν32,ΠxΠyΠz)12,\mathcal{L}_{p}(\mu_{1},\mu_{2},\mu_{3})(x,y,z)=K\cdot\left(\prod_{\mathfrak{p}\mid p}\frac{\mathcal{E}_{\mathfrak{p}}(x,y,z)}{\mathcal{E}_{\mathfrak{p},1}(z)}\right)\cdot L\left(\frac{1-\nu_{1}-\nu_{2}-\nu_{3}}{2},\Pi_{x}\otimes\Pi_{y}\otimes\Pi_{z}\right)^{\frac{1}{2}},

where KK is a non-zero constant depending of (x,y,z)(x,y,z), 𝔭(x,y,z)=\mathcal{E}_{\mathfrak{p}}(x,y,z)=

{(1βx𝔭βy𝔭αz𝔭ϖ𝔭m¯𝔭2¯)(1αx𝔭βy𝔭βz𝔭ϖ𝔭m¯𝔭2¯)(1βx𝔭αy𝔭βz𝔭ϖ𝔭m¯𝔭2¯)(1βx𝔭βy𝔭βz𝔭ϖ𝔭m¯𝔭2¯),𝔭𝔭0(1αx𝔭0αy𝔭0βz𝔭0p1m0)(1αx𝔭0βy𝔭0βz𝔭0p1m0)(1βx𝔭0αy𝔭0βz𝔭0p1m0)(1βx𝔭0βy𝔭0βz𝔭0p1m0),𝔭=𝔭0,\left\{\begin{array}[]{lc}\mbox{\small$(1-\beta_{x}^{\mathfrak{p}}\beta_{y}^{\mathfrak{p}}\alpha_{z}^{\mathfrak{p}}\varpi_{\mathfrak{p}}^{-\underline{m}_{\mathfrak{p}}-\underline{2}})(1-\alpha_{x}^{\mathfrak{p}}\beta_{y}^{\mathfrak{p}}\beta_{z}^{\mathfrak{p}}\varpi_{\mathfrak{p}}^{-\underline{m}_{\mathfrak{p}}-\underline{2}})(1-\beta_{x}^{\mathfrak{p}}\alpha_{y}^{\mathfrak{p}}\beta_{z}^{\mathfrak{p}}\varpi_{\mathfrak{p}}^{-\underline{m}_{\mathfrak{p}}-\underline{2}})(1-\beta_{x}^{\mathfrak{p}}\beta_{y}^{\mathfrak{p}}\beta_{z}^{\mathfrak{p}}\varpi_{\mathfrak{p}}^{-\underline{m}_{\mathfrak{p}}-\underline{2}})$},&\mathfrak{p}\neq\mathfrak{p}_{0}\\ \mbox{\small$(1-\alpha_{x}^{\mathfrak{p}_{0}}\alpha_{y}^{\mathfrak{p}_{0}}\beta_{z}^{\mathfrak{p}_{0}}p^{1-m_{0}})(1-\alpha_{x}^{\mathfrak{p}_{0}}\beta_{y}^{\mathfrak{p}_{0}}\beta_{z}^{\mathfrak{p}_{0}}p^{1-m_{0}})(1-\beta_{x}^{\mathfrak{p}_{0}}\alpha_{y}^{\mathfrak{p}_{0}}\beta_{z}^{\mathfrak{p}_{0}}p^{1-m_{0}})(1-\beta_{x}^{\mathfrak{p}_{0}}\beta_{y}^{\mathfrak{p}_{0}}\beta_{z}^{\mathfrak{p}_{0}}p^{1-m_{0}})$},&\mathfrak{p}=\mathfrak{p}_{0}\end{array}\right.,
𝔭,1(z):={(1(βz𝔭)2ϖ𝔭k¯3,𝔭2¯)(1(βz𝔭)2ϖ𝔭k¯3,𝔭1¯),𝔭𝔭0,(1(βz𝔭0)2pk3,τ0)(1(βz𝔭0)2p1k3,τ0),𝔭=𝔭0,\mathcal{E}_{\mathfrak{p},1}(z):=\left\{\begin{array}[]{lc}(1-(\beta_{z}^{\mathfrak{p}})^{2}\varpi_{\mathfrak{p}}^{-\underline{k}_{3,\mathfrak{p}}-\underline{2}})\cdot(1-(\beta_{z}^{\mathfrak{p}})^{2}\varpi_{\mathfrak{p}}^{-\underline{k}_{3,\mathfrak{p}}-\underline{1}}),&\mathfrak{p}\neq\mathfrak{p}_{0},\\ (1-(\beta_{z}^{\mathfrak{p}_{0}})^{2}p^{-k_{3,\tau_{0}}})\cdot(1-(\beta_{z}^{\mathfrak{p}_{0}})^{2}p^{1-k_{3,\tau_{0}}}),&\mathfrak{p}=\mathfrak{p}_{0},\end{array}\right.

m0=k1,τ0+k2,τ0+k3,τ020m_{0}=\frac{k_{1,\tau_{0}}+k_{2,\tau_{0}}+k_{3,\tau_{0}}}{2}\geq 0, m¯𝔭=k¯1,𝔭+k¯2,𝔭+k¯3,𝔭2=(k1,τ+k2,τ+k3,τ2)τ𝔭\underline{m}_{\mathfrak{p}}=\frac{\underline{k}_{1,\mathfrak{p}}+\underline{k}_{2,\mathfrak{p}}+\underline{k}_{3,\mathfrak{p}}}{2}=\left(\frac{k_{1,\tau}+k_{2,\tau}+k_{3,\tau}}{2}\right)_{\tau\sim\mathfrak{p}} and τ𝔭\tau\sim\mathfrak{p} means real embeddings τ\tau corresponding to embeddings F𝔭pF_{\mathfrak{p}}\hookrightarrow\operatorname{\mathbb{C}}_{p} through ιp\iota_{p}.

As mentioned before this result is analogous to [3, Theorem 1.2] where the triple product pp-adic L-function p(μ1,μ2,μ3)\mathcal{L}_{p}(\mu_{1},\mu_{2},\mu_{3}) is constructed in the ordinary setting. We have been able to extend this result to the finite slope setting thanks to the strategy to pp-adically interpolate of the integral powers of the Gauss-Manin connexion. We perform such pp-adic interpolations as power series of endomorphisms in a well chosen space of nearly overconvergent modular forms.

Acknowledgements.

The author is supported in part by DGICYT Grant MTM2015-63829-P. This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 682152).

2. Notations

Let 𝔸\mathbb{A} be the adeles of \operatorname{\mathbb{Q}} and 𝔸f\operatorname{\mathbb{A}}_{f} the finite adeles. Let FF be a totally real field of degree d=[F:]d=[F:\operatorname{\mathbb{Q}}], 𝒪F\mathcal{O}_{F} its ring of integers and ΣF\Sigma_{F} the set of real embeddings of FF. In all this paper we fix an embedding τ0ΣF\tau_{0}\in\Sigma_{F}. We denote by 1¯[ΣF]\underline{1}\in\operatorname{\mathbb{Z}}[\Sigma_{F}] the element with each coordinate equals to 11. For xF×x\in F^{\times} and k¯[ΣF]\underline{k}\in\operatorname{\mathbb{Z}}[\Sigma_{F}] we put xk¯=τΣFτ(x)kτx^{\underline{k}}=\prod_{\tau\in\Sigma_{F}}\tau(x)^{k_{\tau}}.

We fix a prime number p>2p>2, and we identinfy ΣF\Sigma_{F} with the set of the embeddings of FF in ¯p\overline{\operatorname{\mathbb{Q}}}_{p} once we fix an embedding ιp:¯p\iota_{p}:\bar{\operatorname{\mathbb{Q}}}\hookrightarrow\operatorname{\mathbb{C}}_{p}. For each prime of 𝔭p\mathfrak{p}\mid p let F𝔭F_{\mathfrak{p}} be the completion of FF at 𝔭\mathfrak{p}, Σ𝔭\Sigma_{\mathfrak{p}} the set of its embeddings in p\operatorname{\mathbb{C}}_{p}, 𝒪𝔭\mathcal{O}_{\mathfrak{p}} its ring of integers, κ𝔭\kappa_{\mathfrak{p}} its residue field, q𝔭=κ𝔭q_{\mathfrak{p}}=\sharp\kappa_{\mathfrak{p}} and e𝔭e_{\mathfrak{p}} the ramification index. We also fix uniformizers ϖ𝔭𝒪𝔭\varpi_{\mathfrak{p}}\in\mathcal{O}_{\mathfrak{p}}. Moreover we use the notation 𝒪:=𝒪Fp\mathcal{O}:=\mathcal{O}_{F}\otimes\operatorname{\mathbb{Z}}_{p} which naturally decompose as 𝒪=𝔭𝒪𝔭\mathcal{O}=\prod_{\mathfrak{p}}\mathcal{O}_{\mathfrak{p}}. We will denote by 𝔭0\mathfrak{p}_{0} the prime corresponding to ιp\iota_{p} and τ0\tau_{0}. We suppose the following hypothesis:

Hypothesis 2.1.

[F𝔭0:p]=1[F_{\mathfrak{p}_{0}}:\operatorname{\mathbb{Q}}_{p}]=1 and FF unramified at pp.

We denote 𝒪τ0:=𝔭𝔭0𝒪𝔭\mathcal{O}^{\tau_{0}}:=\prod_{\mathfrak{p}\neq\mathfrak{p}_{0}}\mathcal{O}_{\mathfrak{p}}. Thus we have the following decomposition 𝒪=𝒪𝔭0×𝒪τ0=p×𝒪τ0\mathcal{O}=\mathcal{O}_{\mathfrak{p}_{0}}\times\mathcal{O}^{\tau_{0}}=\operatorname{\mathbb{Z}}_{p}\times\mathcal{O}^{\tau_{0}}.

We also fix a quaternion algebra over FF denoted by BB such that:

  • (i)

    split at τ0\tau_{0} and at each 𝔭p\mathfrak{p}\mid p,

  • (ii)

    is ramified at each τΣF{τ0}\tau\in\Sigma_{F}\setminus\{\tau_{0}\}.

We choose from now on λ\lambda\in\operatorname{\mathbb{Q}} such that λ<0\lambda<0 and pp split in (λ)\operatorname{\mathbb{Q}}(\sqrt{\lambda}). Let E:=F(λ)E:=F(\sqrt{\lambda}) and denote zz¯z\mapsto\overline{z} the not-trivial automorphism of E/FE/F.

We denote D:=BFED:=B\otimes_{F}E which is a quaternion algebra over EE and we consider the involution on DD defined by l=bzl¯:=b¯z¯l=b\otimes z\mapsto\overline{l}:=\bar{b}\otimes\overline{z} where b¯\bar{b} is the canonical involution of BB. We fix δD×\delta\in D^{\times} such that δ¯=δ\overline{\delta}=-\delta and define a new involution on DD by ll:=δ1l¯δl\mapsto l^{\ast}:=\delta^{-1}\overline{l}\delta. We denote by VV to the underlying \operatorname{\mathbb{Q}}-vector space of DD endowed with the natural left action of DD. We have a symplectic bilinear form on VV:

Θ:V×V,(v,w)TrE/(TrD/E(vδw)).\Theta:V\times V\rightarrow\operatorname{\mathbb{Q}},\ \ \ (v,w)\mapsto\mathrm{Tr}_{E/\operatorname{\mathbb{Q}}}(\mathrm{Tr}_{D/E}(v\delta w^{\ast})).

Let GG^{\prime} be the reductive group over \operatorname{\mathbb{Q}} such that for each \operatorname{\mathbb{Q}}-algebra RR we have:

G(R)={Dlinearsymplecticsimilitudesof(VR,ΘR)}.G^{\prime}(R)=\left\{\mathrm{D-linear\ symplectic\ similitudes\ of}\ (V\otimes_{\operatorname{\mathbb{Q}}}R,\Theta\otimes_{\operatorname{\mathbb{Q}}}R)\right\}.

Let ABA\subset B be a 𝒪F\mathcal{O}_{F}-order, and let AD:=A𝒪F𝒪EA_{D}:=A\otimes_{\mathcal{O}_{F}}\mathcal{O}_{E}. We introduce a way to cut certain modules endowed with an action of ADA_{D}. We denote by ψ:(λ)D\psi:\operatorname{\mathbb{Q}}(\sqrt{\lambda})\longrightarrow D the natural embedding given by z1zz\longmapsto 1\otimes z and fix an extension R/𝒪ER/\mathcal{O}_{E} such that A𝒪FR=M2(R)A\otimes_{\mathcal{O}_{F}}R=\mathrm{M}_{2}(R). For any RR-module MM endowed with a linear action of ADA_{D}, we define

M+:={vM:ψ(e)v=ev, for all e(λ)},M^{+}:=\left\{v\in M:\;\psi(e)\ast v=ev,\mbox{ for all }e\in\operatorname{\mathbb{Z}}(\sqrt{\lambda})\right\},
M:={vM:ψ(e)v=e¯v, for all e(λ)}.M^{-}:=\left\{v\in M:\;\psi(e)\ast v=\bar{e}v,\mbox{ for all }e\in\operatorname{\mathbb{Z}}(\sqrt{\lambda})\right\}.

Each M±M^{\pm} is equipped with an action of AR=M2(R)𝒪FA\otimes_{\operatorname{\mathbb{Z}}}R=\mathrm{M}_{2}(R)\otimes_{\operatorname{\mathbb{Z}}}\mathcal{O}_{F} and we put M±,1:=(1000)M±M^{\pm,1}:=(\begin{smallmatrix}1&0\\ 0&0\end{smallmatrix})M^{\pm} and M±,2:=(0001)M±.M^{\pm,2}:=(\begin{smallmatrix}0&0\\ 0&1\end{smallmatrix})M^{\pm}. Note that both are isomorphic R𝒪FR\otimes_{\operatorname{\mathbb{Z}}}\mathcal{O}_{F}-modules through the matrix (0110)(\begin{smallmatrix}0&1\\ 1&0\end{smallmatrix}). Moreover, by construction we have:

MM+M=M+,1M,1M+,2M,2,M\supseteq M^{+}\oplus M^{-}=M^{+,1}\oplus M^{-,1}\oplus M^{+,2}\oplus M^{-,2},

and the inclusion is an equality if disc((λ))R×{\rm disc}(\operatorname{\mathbb{Q}}(\sqrt{\lambda}))\in R^{\times}.

3. Unitary modular forms

We introduce unitary Shimura curves associated with GG^{\prime} and their moduli interpretation. We define the sheaves which give rise to modular forms. See [3, §4] for further details.

3.1. Unitary Shimura curves

Let 𝒪BB\mathcal{O}_{B}\subset B be a maximal order and let 𝒪D:=𝒪B𝒪F𝒪E\mathcal{O}_{D}:=\mathcal{O}_{B}\otimes_{\mathcal{O}_{F}}\mathcal{O}_{E}. The order 𝒪D\mathcal{O}_{D} is maximal (locally) at every prime not dividing disc(B)disc(D)1{\rm disc}(B){\rm disc}(D)^{-1}. We denote by GDG_{D} the algebraic group attached to 𝒪D\mathcal{O}_{D}, namely, GD(R):=(𝒪DR)×G_{D}(R):=(\mathcal{O}_{D}\otimes_{\operatorname{\mathbb{Z}}}R)^{\times} for any \operatorname{\mathbb{Z}}-algebra RR.

Let 𝔫\mathfrak{n} be an integral ideal of FF prime to disc(B)\mathrm{disc}(B) and consider the open compact subgroups of GD(^)G_{D}(\hat{\operatorname{\mathbb{Z}}}) and G(𝔸f)G^{\prime}(\operatorname{\mathbb{A}}_{f}):

K1D(𝔫):={(abcd)GD(^):cd10mod𝔫𝒪E},K1(𝔫):=K1D(𝔫)G(𝔸f).K_{1}^{D}(\mathfrak{n}):=\left\{\left(\begin{smallmatrix}a&b\\ c&d\end{smallmatrix}\right)\in G_{D}(\hat{\operatorname{\mathbb{Z}}}):\;c\equiv d-1\equiv 0\;{\rm mod}\;\mathfrak{n}\mathcal{O}_{E}\right\},\qquad K_{1}^{{}^{\prime}}(\mathfrak{n}):=K_{1}^{D}(\mathfrak{n})\cap G^{\prime}(\operatorname{\mathbb{A}}_{f}).

We define the unitary Shimura curve over \operatorname{\mathbb{C}} of level K1(𝔫)K_{1}^{\prime}(\mathfrak{n}) as:

(1) X():=G()+\(×G(𝔸f)/K1(𝔫)).X(\operatorname{\mathbb{C}}):=G^{\prime}(\operatorname{\mathbb{Q}})_{+}\backslash(\mathfrak{H}\times G^{\prime}(\operatorname{\mathbb{A}}_{f})/K_{1}^{{}^{\prime}}(\mathfrak{n})).

Let 𝒪B,𝔪𝒪B\mathcal{O}_{B,\mathfrak{m}}\subseteq\mathcal{O}_{B} be an Eichler order with a well chosen level 𝔪𝔫\mathfrak{m}\mid\mathfrak{n} so that the conditions of [3, Lemma 4.2] are satisfied. Let L/EL/E be a finite extension such that BFL=M2(L)B\otimes_{F}L=\mathrm{M}_{2}(L). By [7, §2.3] the Riemann surface X()X(\operatorname{\mathbb{C}}) has a model (also denoted XX) defined over LL. This curve solves the following moduli problem: if RR is a LL-algebra then X(R)X(R) corresponds to the set of the isomorphism classes of tuples (A,ι,θ,α)(A,\iota,\theta,\alpha) where:

  • (i)(i)

    AA is an abelian scheme over RR of relative dimension 4d4d.

  • (ii)(ii)

    ι:𝒪𝔪EndR(A)\iota:\mathcal{O}_{\mathfrak{m}}\rightarrow\mathrm{End}_{R}(A) gives an action of the ring 𝒪𝔪\mathcal{O}_{\mathfrak{m}} on AA such that Lie(A),1{\rm Lie}(A)^{-,1} is of rank 1 and the action of 𝒪F\mathcal{O}_{F} factors through 𝒪FEL\mathcal{O}_{F}\subset E\subseteq L.

  • (iii)(iii)

    θ\theta is a 𝒪𝔪\mathcal{O}_{\mathfrak{m}}-invariant homogeneous polarization of AA such that the Rosati involution sends ι(d)\iota(d) to ι(d)\iota(d^{\ast}).

  • (iv)(iv)

    α\alpha is a class modulo K1(𝔫)K_{1}^{{}^{\prime}}(\mathfrak{n}) of 𝒪𝔪\mathcal{O}_{\mathfrak{m}}-linear symplectic similitudes α:T^(A)𝒪^𝔪\alpha:\hat{T}(A)\stackrel{{\scriptstyle\simeq}}{{\rightarrow}}\hat{\mathcal{O}}_{\mathfrak{m}}.

Remark 3.1.

Since pdisc(B)p\nmid{\rm disc}(B), a class α\alpha of 𝒪𝔪\mathcal{O}_{\mathfrak{m}}-linear symplectic similitudes modulo K1(𝔫)K_{1}^{{}^{\prime}}(\mathfrak{n}) is decomposed as α=αp×αp\alpha=\alpha_{p}\times\alpha^{p}. By [3, Remark 4.5] to provide a αp\alpha_{p} modulo K1(𝔫)pK_{1}^{{}^{\prime}}(\mathfrak{n})_{p} amounts to giving a point PA[𝔫𝒪Fp],1P\in A[\mathfrak{n}\mathcal{O}_{F}\otimes\operatorname{\mathbb{Z}}_{p}]^{-,1} that generates a subgroup isomorphic to (𝒪Fp)/(𝔫𝒪Fp)(\mathcal{O}_{F}\otimes\operatorname{\mathbb{Z}}_{p})/(\mathfrak{n}\mathcal{O}_{F}\otimes\operatorname{\mathbb{Z}}_{p}). We have an analogous description in case of Γ0(𝔫)\Gamma_{0}(\mathfrak{n})-structures.

3.2. Modular sheaves

We introduce the sheaves which give rise to the modular forms for GG^{\prime}. Let L0/FL_{0}/F be an extension such that BFL0=M2(L0)B\otimes_{F}L_{0}=\mathrm{M}_{2}(L_{0}), write L=L0(λ)EL=L_{0}(\sqrt{\lambda})\supset E, and denote by XLX_{L} the base change to LL of the unitary Shimura curve XX. Using the universal abelian variety π:AXL\pi:A\rightarrow X_{L} we define the following coherent sheaves on XLX_{L}:

ω:=(πΩA/XL1)+,2ω:=((R1π𝒪A)+,2):=(1πΩA/XL)+,2.\omega:=\left(\pi_{\ast}\Omega^{1}_{A/X_{L}}\right)^{+,2}\qquad\omega_{-}:=\left(\left(R^{1}\pi_{\ast}\mathcal{O}_{A}\right)^{+,2}\right)^{\vee}\qquad\mathcal{H}:=\left(\mathcal{R}^{1}\pi_{\ast}\Omega^{\bullet}_{A/X_{L}}\right)^{+,2}.

The sheaf \mathcal{H} is endowed with a Gauss-Manin connection :ΩXL1\bigtriangledown:\mathcal{H}\rightarrow\mathcal{H}\otimes\Omega^{1}_{X_{L}}. The natural 𝒪D\mathcal{O}_{D}-equivariant exact sequence:

(2) 0πΩA/XL11πΩA/XLR1π𝒪A0,0\longrightarrow\pi_{\ast}\Omega^{1}_{A/X_{L}}\longrightarrow\mathcal{R}^{1}\pi_{\ast}\Omega^{\bullet}_{A/X_{L}}\longrightarrow R^{1}\pi_{\ast}\mathcal{O}_{A}\longrightarrow 0,

induces the Hodge exact sequence (see [10, §2.3.1]) 0ωω00\rightarrow\omega\rightarrow\mathcal{H}\rightarrow\omega_{-}^{\vee}\rightarrow 0.

If LL contains the Galois closure of FF then the natural decomposition FLLΣFF\otimes_{\operatorname{\mathbb{Q}}}L\cong L^{\Sigma_{F}} induces: ω=τΣFωτ\omega=\bigoplus_{\tau\in\Sigma_{F}}\omega_{\tau} and =τΣFτ\mathcal{H}=\bigoplus_{\tau\in\Sigma_{F}}\mathcal{H}_{\tau}. As the sheaves (ΩA/XL1)+,1(\Omega^{1}_{A/X_{L}})^{+,1} and (ΩA/XL1)+,2(\Omega^{1}_{A/X_{L}})^{+,2} are isomorphic then condition (ii)(2)(ii)(2) of the moduli problem of XX imply that ωτ0\omega_{\tau_{0}} is locally free of rank 1, while ωτ\omega_{\tau} is of rank 22 for ττ0\tau\neq\tau_{0}. Moreover, ω\omega_{-} is locally free of rank 1. Thus, ωτ=τ\omega_{\tau}=\mathcal{H}_{\tau} for each ττ0\tau\neq\tau_{0}, and we have the exact sequence

(3) 0ωτ0τ0ϵω0.0\longrightarrow\omega_{\tau_{0}}\longrightarrow\mathcal{H}_{\tau_{0}}\stackrel{{\scriptstyle\epsilon}}{{\longrightarrow}}\omega_{-}^{\vee}\longrightarrow 0.

Let k¯=(kτ)[ΣF]\underline{k}=(k_{\tau})\in\operatorname{\mathbb{N}}[\Sigma_{F}], we introduce the modular sheaves over XLX_{L} considered in this paper:

ωk¯:=ωτ0kτ0ττ0Symkτωτ.\omega^{\underline{k}}:=\omega_{\tau_{0}}^{\otimes k_{\tau_{0}}}\otimes\bigotimes_{\tau\neq\tau_{0}}\mathrm{Sym}^{k_{\tau}}\omega_{\tau}.
Definition 3.2.

A modular form of weight k¯\underline{k} and coefficients in LL for GG^{\prime} is a global section of ωk¯\omega^{\underline{k}}, i.e. an element of H0(XL,ωk¯)\mathrm{H}^{0}(X_{L},\omega^{\underline{k}}).

In [3, §4.3] an isomorphism φτ0:ωτ0ω\varphi_{\tau_{0}}:\omega_{\tau_{0}}\stackrel{{\scriptstyle\simeq}}{{\longrightarrow}}\omega_{-} is provided by the polarization. Thus, the Kodaira-Spencer isomorphism (see [10, Lemme 2.3.4]) induces the isomorphism:

KS:ΩXL1ωτ0ωφτ01ωτ02.KS:\Omega^{1}_{X_{L}}\stackrel{{\scriptstyle\simeq}}{{\longrightarrow}}\omega_{\tau_{0}}\otimes\omega_{-}\stackrel{{\scriptstyle\varphi_{\tau_{0}}^{-1}}}{{\longrightarrow}}\omega_{\tau_{0}}^{\otimes 2}.

If ττ0\tau\neq\tau_{0}, also the polarization provides an isomorphism:

(4) φτ:2(1πΩA/XL)τ+,2=2ωτ𝒪XL\varphi_{\tau}:\bigwedge^{2}\left(\mathcal{R}^{1}\pi_{\ast}\Omega^{\bullet}_{A/X_{L}}\right)_{\tau}^{+,2}=\bigwedge^{2}\omega_{\tau}\stackrel{{\scriptstyle\simeq}}{{\longrightarrow}}\mathcal{O}_{X_{L}}

3.3. Katz Modular forms

Let R0R_{0} be a LL-algebra and let fH0(X/R0,ωk¯)f\in H^{0}(X/R_{0},\omega^{\underline{k}}). If RR is a R0R_{0}-algebra, (A,ι,θ,α)(A,\iota,\theta,\alpha) is a tuple corresponding to a point of X(Spec(R))X(\mathrm{Spec}(R)) and w=(f0,(fτ,eτ)ττ0)w=(f_{0},(f_{\tau},e_{\tau})_{\tau\neq\tau_{0}}) is a RR-basis of ωA=(ΩA/R1)+,2\omega_{A}=\left(\Omega^{1}_{A/R}\right)^{+,2}, then there exists f(A,ι,θ,α,w)ττ0Symkτ(R2)f(A,\iota,\theta,\alpha,w)\in\bigotimes_{\tau\neq\tau_{0}}\mathrm{Sym}^{k_{\tau}}\left(R^{2}\right)^{\vee} such that

f(A,ι,θ,α)=f(A,ι,θ,α,w)(P)f0kτ0 with P((xτ,yτ)ττ0)=ττ0|fτeτxτyτ|kτφτ(fτeτ)kτf(A,\iota,\theta,\alpha)=f(A,\iota,\theta,\alpha,w)(P)f_{0}^{\otimes k_{\tau_{0}}}\mbox{ with }P((x_{\tau},y_{\tau})_{\tau\neq\tau_{0}})=\bigotimes_{\tau\neq{\tau_{0}}}|\begin{smallmatrix}f_{\tau}&e_{\tau}\\ x_{\tau}&y_{\tau}\end{smallmatrix}|^{k_{\tau}}\varphi_{\tau}(f_{\tau}\wedge e_{\tau})^{-k_{\tau}}

Thus a section fH0(X/R0,ωk¯)f\in H^{0}(X/R_{0},\omega^{\underline{k}}) is characterized as a rule that assigns to any R0R_{0}-algebra RR and (A,ι,θ,α,w)(A,\iota,\theta,\alpha,w) over RR a linear form

f(A,ι,θ,α,w)ττ0Symkτ(R2)f(A,\iota,\theta,\alpha,w)\in\bigotimes_{\tau\neq{\tau_{0}}}\mathrm{Sym}^{k_{\tau}}\left(R^{2}\right)^{\vee}

satisfying:

  • (A1)

    The element f(A,ι,θ,α,w)f(A,\iota,\theta,\alpha,w) depends only on the RR-isomorphism class of (A,ι,θ,α)(A,\iota,\theta,\alpha).

  • (A2)

    Formation of f(A,ι,θ,α,w)f(A,\iota,\theta,\alpha,w) commutes with extensions RRR\rightarrow R^{\prime} of R0R_{0}-algebras.

  • (A3)

    If (t,g¯)R××GL2(R)ΣF{τ0}(t,\underline{g})\in R^{\times}\times\operatorname{\mathrm{GL}}_{2}(R)^{\Sigma_{F}\setminus\{\tau_{0}\}}:

    f(A,ι,θ,α,w(t,g¯))=tkτ0(g¯1f(A,ι,θ,α,w)).f(A,\iota,\theta,\alpha,w(t,\underline{g}))=t^{-k_{\tau_{0}}}\cdot\left(\underline{g}^{-1}f(A,\iota,\theta,\alpha,w)\right).

    where (t,g¯)(t,\underline{g}) acts on ww in a natural way (see [3, §4.4] for further details).

4. Overconvergent families of modular forms

In this section we construct families of modular sheaves that pp-adically interpolate the modular sheaves introduced in 3.2. Again, for further details see [3, §5].

4.1. Integral models, Hasse invariants and canonical subgroups

Let 𝔫\mathfrak{n} be an integral ideal of FF prime to pp and disc(B)\mathrm{disc}(B). Write K1(𝔫,𝔭𝔭0𝔭):=K1(𝔫)K0(𝔭𝔭0𝔭)K_{1}^{\prime}\left(\mathfrak{n},\prod_{\mathfrak{p}\neq\mathfrak{p}_{0}}\mathfrak{p}\right):=K_{1}^{\prime}\left(\mathfrak{n}\right)\cap K_{0}^{\prime}\left(\prod_{\mathfrak{p}\neq\mathfrak{p}_{0}}\mathfrak{p}\right), and denote by XX the unitary Shimura curve of level K1(𝔫,𝔭𝔭0𝔭)K_{1}^{\prime}\left(\mathfrak{n},\prod_{\mathfrak{p}\neq\mathfrak{p}_{0}}\mathfrak{p}\right) similarly as in §3.1. Since pp is coprime to the discriminant of BB, by [7, §5.3] XX admits a canonical model XintX_{\mathrm{int}} over 𝒪𝔭0\mathcal{O}_{\mathfrak{p}_{0}}, representing the analogue moduli problem described in §3.1 but exchanging an EE-algebra by an 𝒪𝔭0\mathcal{O}_{\mathfrak{p}_{0}}-algebra RR. Namely, it classifies quadruples (A,ι,θ,α𝔭0)(A,\iota,\theta,\alpha^{\mathfrak{p}_{0}}) over RR, where α𝔭0\alpha^{\mathfrak{p}_{0}} is a class of 𝒪D\mathcal{O}_{D}-linear symplectic similitudes outside 𝔭0\mathfrak{p}_{0}. By [7, §5.4], XintX_{\mathrm{int}} has good reduction. Let π:𝐀Xint\pi:{\bf A}\rightarrow X_{\mathrm{int}} be the universal abelian variety, whose existence is guaranteed by the moduli interpretation of XintX_{\mathrm{int}}. Notice that the universal abelian variety AXLA\rightarrow X_{L} introduced in §3.2 is the generic fibre of 𝐀Xint{\bf A}\rightarrow X_{\mathrm{int}}. Since we have added Γ0(𝔭)\Gamma_{0}(\mathfrak{p})-structure for all 𝔭𝔭0\mathfrak{p}\neq\mathfrak{p}_{0}, 𝐀{\bf A} is endowed with a subgroup C𝔭𝐀[𝔭],1C_{\mathfrak{p}}\subset{\bf A}[\mathfrak{p}]^{-,1} isomorphic to 𝒪𝔭/𝔭\mathcal{O}_{\mathfrak{p}}/\mathfrak{p} by Remark 3.1.

Let 𝔛\mathfrak{X} be denote the formal scheme over Spf(𝒪𝔭0)\mathrm{Spf}(\mathcal{O}_{\mathfrak{p}_{0}}) obtained as the completion of XintX_{\mathrm{int}} along its special fiber which is denoted by X¯int\bar{X}_{\mathrm{int}} .

The pp-divisible group 𝐀[p]{\bf A}[p^{\infty}] over XintX_{\mathrm{int}} is decomposed as:

𝐀[p]=𝐀[𝔭0]+[𝔭𝔭0𝐀[𝔭]+]𝐀[𝔭0][𝔭𝔭0𝐀[𝔭]],{\bf A}[p^{\infty}]={\bf A}[\mathfrak{p}_{0}^{\infty}]^{+}\oplus\left[\bigoplus_{\mathfrak{p}\neq\mathfrak{p}_{0}}{\bf A}[\mathfrak{p}^{\infty}]^{+}\right]\oplus{\bf A}[\mathfrak{p}_{0}^{\infty}]^{-}\oplus\left[\bigoplus_{\mathfrak{p}\neq\mathfrak{p}_{0}}{\bf A}[\mathfrak{p}^{\infty}]^{-}\right],

We are interested in the pp-divisible groups 𝒢0:=𝐀[𝔭0],1\mathcal{G}_{0}:={\bf A}[\mathfrak{p}_{0}^{\infty}]^{-,1} and 𝒢𝔭:=𝐀[𝔭],1\mathcal{G}_{\mathfrak{p}}:={\bf A}[\mathfrak{p}^{\infty}]^{-,1} if 𝔭𝔭0\mathfrak{p}\neq\mathfrak{p}_{0}, which are defined over XintX_{\mathrm{int}} and endowed with actions of 𝒪𝔭0\mathcal{O}_{\mathfrak{p}_{0}} and 𝒪𝔭\mathcal{O}_{\mathfrak{p}} respectively. The sheaves of invariant differentials of the corresponding Cartier dual pp-divisible groups are denoted by ω0:=ω𝒢0D\omega_{0}:=\omega_{\mathcal{G}_{0}^{D}} and ω𝔭:=ω𝒢𝔭D\omega_{\mathfrak{p}}:=\omega_{\mathcal{G}_{\mathfrak{p}}^{D}} if 𝔭𝔭0\mathfrak{p}\neq\mathfrak{p}_{0}. This fits with the definition of ω\omega given in §3.2 since, as explained in [3, §5.1], (πΩ𝐀/Xint1)+,2(πΩ𝐀/Xint1),1\left(\pi_{\ast}\Omega^{1}_{{\bf A}/X_{\rm int}}\right)^{+,2}\simeq\left(\pi_{\ast}\Omega^{1}_{{\bf A}^{\vee}/X_{\rm int}}\right)^{-,1} by means of the polarization.

As explained in [3, §5.1], the universal polarization provides an isomorphism of sheaves of invariant differentials ω0ω𝒢0\omega_{0}\simeq\omega_{\mathcal{G}_{0}} compatible with φτ0\varphi_{\tau_{0}}. Moreover, for any ring 𝒪^\hat{\mathcal{O}} containing all the pp-adic embeddings of 𝒪F𝒪p\mathcal{O}_{F}\hookrightarrow\mathcal{O}_{\operatorname{\mathbb{C}}_{p}}, if we extend our base ring 𝒪𝔭0\mathcal{O}_{\mathfrak{p}_{0}} to 𝒪^\hat{\mathcal{O}} then we have a decomposition ω𝔭=τΣ𝔭ω𝔭,τ\omega_{\mathfrak{p}}=\oplus_{\tau\in\Sigma_{\mathfrak{p}}}\omega_{\mathfrak{p},\tau}, where each ω𝔭,τ\omega_{\mathfrak{p},\tau} has rank 22 and we have an isomorphism φτ:2ω𝔭,τ𝒪Xint\varphi_{\tau}:\bigwedge^{2}\omega_{\mathfrak{p},\tau}\stackrel{{\scriptstyle\simeq}}{{\rightarrow}}\mathcal{O}_{X_{\rm int}}. The sheaves ω𝔭,τ\omega_{\mathfrak{p},\tau} and ω0\omega_{0} are the integral models of ωτ\omega_{\tau} and ωτ0\omega_{\tau_{0}}, respectively.

There is a dichotomy in XintX_{\mathrm{int}} which says that any point in the generic fiber X¯int\bar{X}_{\mathrm{int}} is ordinary or supersingular (with respect to 𝒢0\mathcal{G}_{0}), and there are finitely many supersingular points in X¯int\bar{X}_{\mathrm{int}}. From [14, Proposition 6.1] there exists HaH0(X¯int,ω0p1)\mathrm{Ha}\in H^{0}\left(\bar{X}_{\mathrm{int}},\omega_{0}^{p-1}\right) that vanishes exactly at supersingular geometric points and these zeroes are simple. This is called the Hasse invariant.

We denote by Hdg¯\overline{\rm Hdg} the locally principal ideal of 𝒪X¯int\mathcal{O}_{\bar{X}_{\mathrm{int}}} described as follows: for each U=Spec(R)X¯intU=\mathrm{Spec}(R)\subset\bar{X}_{\mathrm{int}} if ω0U=Rw\omega_{0}\mid_{U}=Rw and HaU=Hw(p1)\mathrm{Ha}\mid_{U}=Hw^{\otimes(p-1)} then Hdg¯U=HRR\overline{\rm Hdg}\mid_{U}=HR\subseteq R. Let Hdg\rm Hdg the inverse image of Hdg¯\overline{\rm Hdg} in 𝒪𝔛\mathcal{O}_{\mathfrak{X}}, which is also a locally principal ideal. Note that Hapn\mathrm{Ha}^{p^{n}} extends canonically to a section of H0(𝔛,ω0pn(p1)/pn+1)H^{0}(\mathfrak{X},\omega_{0}^{p^{n}(p-1)}\otimes\operatorname{\mathbb{Z}}/p^{n+1}\operatorname{\mathbb{Z}}), indeed, for any two extensions Ha1\mathrm{Ha}_{1} and Ha2\mathrm{Ha}_{2} of Ha\mathrm{Ha} we have Ha1pn=Ha2pn\mathrm{Ha}_{1}^{p^{n}}=\mathrm{Ha}_{2}^{p^{n}} modulo pn+1p^{n+1} by the binomial formula.

Remark 4.1.

From [6, Prop. 3.4] there exists a (p1)(p-1)-root of the principal ideal Hdg\mathrm{Hdg}. This ideal is denoted δ¯\underline{\delta} or Hdg1/(p1)\mathrm{Hdg}^{1/(p-1)}.

We introduce the corresponding neighborhoods of the ordinary locus. For each integer r1r\geq 1 we denote by 𝔛r\mathfrak{X}_{r} the formal scheme over 𝔛\mathfrak{X} which represents the functor that classifies for each pp-adically complete 𝒪^\hat{\mathcal{O}}-algebra RR:

𝔛r(R)={[(h,η)]|h𝔛(R),ηH0(Spf(R),h(ω𝒢(1p)pr+1)),ηHapr+1=pmodp2},\mathfrak{X}_{r}(R)=\left\{[(h,\eta)]\ \ |\ \ h\in\mathfrak{X}(R),\;\eta\in H^{0}(\mathrm{Spf}(R),h^{\ast}(\omega_{\mathcal{G}}^{(1-p)p^{r+1}})),\;\eta\cdot{\rm Ha}^{p^{r+1}}=p\mod p^{2}\right\},

here the brackets means the equivalence class given by (h,η)(h,η)(h,\eta)\equiv(h^{\prime},\eta^{\prime}) if h=hh=h^{\prime} and η=η(1+pu)\eta=\eta^{\prime}(1+pu) for some uRu\in R (see [2, Definition 3.1]).

Proposition 4.2.

[2, Corollaire A.2] There exists a canonical subgroup CnC_{n} of 𝒢0[𝔭0n]\mathcal{G}_{0}[\mathfrak{p}_{0}^{n}] for nrn\leq r over 𝔛r\mathfrak{X}_{r}. This is unique and satisfy the compatibility Cn[𝔭0n1]=Cn1C_{n}[\mathfrak{p}_{0}^{n-1}]=C_{n-1}. Moreover, if we denote Dn:=𝒢0[𝔭0n]/CnD_{n}:=\mathcal{G}_{0}[\mathfrak{p}_{0}^{n}]/C_{n} then ωDnω𝒢0[𝔭n]/Hdgpn1p1\omega_{D_{n}}\simeq\omega_{\mathcal{G}_{0}[\mathfrak{p}^{n}]}/\mathrm{Hdg}^{\frac{p^{n}-1}{p-1}}.

If we write Ω0ω0\Omega_{0}\subseteq\omega_{0} for the subsheaf generated by the lifts of the image of the Hodge-Tate map, by [2, Proposition A.3] we have that Ω0=δ¯ω0\Omega_{0}=\underline{\delta}\omega_{0}. Thus, we obtain a morphism

(5) dlog0:Dn(𝔛r)Ω0𝒪𝔛r(𝒪𝔛r/n)ω0/pnHdgpn1p1,{\rm dlog}_{0}:D_{n}(\mathfrak{X}_{r})\longrightarrow\Omega_{0}\otimes_{\mathcal{O}_{\mathfrak{X}_{r}}}(\mathcal{O}_{\mathfrak{X}_{r}}/\mathcal{I}_{n})\subset\omega_{0}/p^{n}\mathrm{Hdg}^{-\frac{p^{n}-1}{p-1}},

where n:=pnHdgpnp1\mathcal{I}_{n}:=p^{n}\mathrm{Hdg}^{-\frac{p^{n}}{p-1}}.

By the moduli interpretation, the pp-divisible group 𝔭𝔭0𝒢𝔭𝔛r\prod_{\mathfrak{p}\neq\mathfrak{p}_{0}}\mathcal{G}_{\mathfrak{p}}\rightarrow\mathfrak{X}_{r} is étale isomorphic to 𝔭𝔭0(F𝔭/𝒪𝔭)2\prod_{\mathfrak{p}\neq\mathfrak{p}_{0}}(F_{\mathfrak{p}}/\mathcal{O}_{\mathfrak{p}})^{2}. Assume that rnr\geq n. We denote by 𝔛r,n𝔛r\mathfrak{X}_{r,n}\rightarrow\mathfrak{X}_{r} the formal scheme obtained by adding to the moduli interpretation a point of order 𝔭n\mathfrak{p}^{n} for each 𝔭𝔭0\mathfrak{p}\neq\mathfrak{p}_{0} whose multiples generate 𝒢𝔭[𝔭]/C𝔭\mathcal{G}_{\mathfrak{p}}[\mathfrak{p}]/C_{\mathfrak{p}} (see Remark 3.1). It is clear that the extension 𝔛r,n𝔛r\mathfrak{X}_{r,n}\rightarrow\mathfrak{X}_{r} is étale and its Galois group contains 𝔭𝔭0(𝒪𝔭/pn𝒪𝔭)×\prod_{\mathfrak{p}\neq\mathfrak{p}_{0}}(\mathcal{O}_{\mathfrak{p}}/p^{n}\mathcal{O}_{\mathfrak{p}})^{\times} as a subgroup. Moreover, 𝔛r,n\mathfrak{X}_{r,n} has also good reduction (see [7, §5.4]). Now we trivialize the subgroup DnD_{n}: Let 𝒳r,n\mathcal{X}_{r,n} be the adic generic fiber of 𝔛r,n\mathfrak{X}_{r,n}. By [2, Corollaire A.2], the group scheme Dn𝒳r,nD_{n}\rightarrow\mathcal{X}_{r,n} is also étale isomorphic to pn𝒪𝔭0/𝒪𝔭0p^{{-n}}\mathcal{O}_{\mathfrak{p}_{0}}/\mathcal{O}_{\mathfrak{p}_{0}}. We denote by 𝒢r,n\mathcal{IG}_{r,n} the adic space over 𝒳r,n\mathcal{X}_{r,n} that trivialize DnD_{n}. Then the map 𝒢r,n𝒳r,n\mathcal{IG}_{r,n}\rightarrow\mathcal{X}_{r,n} is a finite étale with Galois group (𝒪𝔭0/pn𝒪𝔭0)×(\mathcal{O}_{\mathfrak{p}_{0}}/p^{n}\mathcal{O}_{\mathfrak{p}_{0}})^{\times}. We denote by 𝔊r,n\mathfrak{IG}_{r,n} the normalization 𝒢r,n\mathcal{IG}_{r,n} in 𝔛r,n\mathfrak{X}_{r,n} which is finite over 𝔛r,n\mathfrak{X}_{r,n} and it is also endowed with an action of (𝒪𝔭0/pn𝒪𝔭0)×(\mathcal{O}_{{\mathfrak{p}_{0}}}/p^{n}\mathcal{O}_{{\mathfrak{p}_{0}}})^{\times}. These constructions are captured by the following tower of formal schemes:

𝔊r,n𝔛r,n𝔛r,\mathfrak{IG}_{r,n}\longrightarrow\mathfrak{X}_{r,n}\longrightarrow\mathfrak{X}_{r},

endowed with a natural action of (𝒪/pn𝒪)×𝔭(𝒪𝔭/pn𝒪𝔭)×(\mathcal{O}/p^{n}\mathcal{O})^{\times}\simeq\prod_{\mathfrak{p}}(\mathcal{O}_{\mathfrak{p}}/p^{n}\mathcal{O}_{\mathfrak{p}})^{\times}.

Let us consider the morphism

η:𝔊r,n𝔛r,n𝔛r𝔛.\eta:\mathfrak{IG}_{r,n}\longrightarrow\mathfrak{X}_{r,n}\longrightarrow\mathfrak{X}_{r}\longrightarrow\mathfrak{X}.

The following result is completely analogous to [1, Lemma 3.3]:

Lemma 4.3.

The induced map η(Ω𝔛/𝒪^1)Ω𝔊r,n/𝒪^1\eta^{*}\left(\Omega_{\mathfrak{X}/\hat{\mathcal{O}}}^{1}\right)\longrightarrow\Omega_{\mathfrak{IG}_{r,n}/\hat{\mathcal{O}}}^{1} has kernel and cokernel annihilated by a power of δ¯\underline{\delta} and, in particular, by a power of pp, depending on nn.

4.2. Formal vector bundles

We briefly recall in this subsection constructions performed in [1, §2] and [3, §6]. Let SS be a formal scheme, \mathcal{I} its (invertible) ideal of definition and \mathcal{E} a locally free 𝒪S\mathcal{O}_{S}-module of rank nn. We write S¯\bar{S} the scheme with structural sheaf 𝒪S/\mathcal{O}_{S}/\mathcal{I} and put ¯\overline{\mathcal{E}} the corresponding 𝒪S¯\mathcal{O}_{\overline{S}}-module. We fix marked sections s1,,sms_{1},\cdots,s_{m} of ¯\overline{\mathcal{E}}, namely, the sections s1,,sms_{1},\cdots,s_{m} define a direct sum decomposition ¯=𝒪S¯mQ\overline{\mathcal{E}}=\mathcal{O}_{\bar{S}}^{m}\oplus Q, where QQ is a locally free 𝒪S¯\mathcal{O}_{\bar{S}}-module of rank nmn-m.

Let SSchS-\mathrm{Sch} be the category of the formal SS-schemes. There exists a formal scheme 𝕍()\operatorname{\mathbb{V}}(\mathcal{E}) over SS called the formal vector bundle attached to \mathcal{E} which represents the functor, denoted by the same symbol, SSchSetsS-\mathrm{Sch}\rightarrow\mathrm{Sets}, given by 𝕍()(t:TS):=H0(T,t())=Hom𝒪T(t(),𝒪T)\operatorname{\mathbb{V}}(\mathcal{E})(t:T\rightarrow S):=H^{0}(T,t^{\ast}(\mathcal{E})^{\vee})={\rm Hom}_{\mathcal{O}_{T}}(t^{\ast}(\mathcal{E}),\mathcal{O}_{T}). Crucial in [1] is the construction of the so called formal vector bundles with marked sections which is the formal scheme 𝕍0(,s1,,sm)\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{m}) over 𝕍()\operatorname{\mathbb{V}}(\mathcal{E}) that represents the sub-functor SSchSetsS-\mathrm{Sch}\rightarrow\mathrm{Sets}

𝕍0(,s1,,sm)(t:TS)={ρH0(T,t())|ρ¯(t(si))=1,i=1,,m},\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{m})(t:T\rightarrow S)=\left\{\rho\in H^{0}(T,t^{\ast}(\mathcal{E})^{\vee})\ |\ \bar{\rho}(t^{\ast}(s_{i}))=1,\ i=1,\cdots,m\right\},

here ρ¯\bar{\rho} is the reduction of ρ\rho modulo \mathcal{I}.

Given the fixed decomposition ¯=Qsii\overline{\mathcal{E}}=Q\oplus\langle s_{i}\rangle_{i}, let us consider now the sub-functor 𝕍Q(,s1,,sm)\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{m}) that associates to any formal SS-scheme t:TSt:T\rightarrow S the subset of sections ρ𝕍0(,s1,,sm)(T)\rho\in\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{m})(T) whose reduction ρ¯\bar{\rho} modulo \mathcal{I} also satisfies ρ¯(t(m))=0\bar{\rho}(t^{\ast}(m))=0 for every mQm\in Q.

Lemma 4.4.

[3, Lemma 6.3] The morphism 𝕍Q(,s1,,sm)𝕍0(,s1,,sm)\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{m})\rightarrow\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{m}) is represented by a formal subscheme.

Remark 4.5.

Notice that this construction is also functorial with respect to (,Q,s1,,sm)(\mathcal{E},Q,s_{1},\cdots,s_{m}). Indeed, given a morphism φ:\varphi:\mathcal{E}^{\prime}\rightarrow\mathcal{E} of locally free 𝒪S\mathcal{O}_{S}-modules of finite rank and marked sections s1,,sm¯s_{1},\cdots,s_{m}\in\overline{\mathcal{E}}, s1,,sm¯s_{1}^{\prime},\cdots,s_{m}^{\prime}\in\overline{\mathcal{E}^{\prime}} such that φ¯(si)=si\bar{\varphi}(s_{i}^{\prime})=s_{i} and φ¯(Q)Q\bar{\varphi}(Q^{\prime})\subseteq Q, we have the morphisms making the following diagram commutative

𝕍Q(,s1,,sm)\textstyle{\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{m})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝕍0(,s1,,sm)\textstyle{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{m})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝕍()\textstyle{\operatorname{\mathbb{V}}(\mathcal{E})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝕍Q(,s1,,sm)\textstyle{\operatorname{\mathbb{V}}_{Q}(\mathcal{E}^{\prime},s_{1}^{\prime},\cdots,s_{m}^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝕍0(,s1,,sm)\textstyle{\operatorname{\mathbb{V}}_{0}(\mathcal{E}^{\prime},s_{1}^{\prime},\cdots,s_{m}^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝕍()\textstyle{\operatorname{\mathbb{V}}(\mathcal{E}^{\prime})}
Remark 4.6.

In fact, 𝕍Q(,s1,,sm)\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{m}) depends on Q¯Q\subset\overline{\mathcal{E}} and the image of sis_{i} in ¯/Q\overline{\mathcal{E}}/Q. Indeed, given miQm_{i}\in Q and t:TSt:T\rightarrow S, since ρ¯(t(mi))=0\bar{\rho}(t^{\ast}(m_{i}))=0,

𝕍Q(,(si+mi)i)(T)={ρH0(T,t());ρ¯(t(si+mi))=1,ρ¯(t(Q))=0}=𝕍Q(,(si)i)(T).\operatorname{\mathbb{V}}_{Q}(\mathcal{E},(s_{i}+m_{i})_{i})(T)=\left\{\rho\in H^{0}(T,t^{\ast}(\mathcal{E})^{\vee});\quad\bar{\rho}(t^{\ast}(s_{i}+m_{i}))=1,\;\bar{\rho}(t^{\ast}(Q))=0\right\}=\operatorname{\mathbb{V}}_{Q}(\mathcal{E},(s_{i})_{i})(T).

4.2.1. Filtrations

Let \mathcal{E} be a locally free 𝒪S\mathcal{O}_{S}-module of rank hh and assume that there exists an 𝒪S\mathcal{O}_{S}-submodule \mathcal{F}\subset\mathcal{E}, locally free as 𝒪S\mathcal{O}_{S}-module of rank dd, which is a locally direct summand in \mathcal{E}. Assume also that we have marked sections s1,,sds_{1},\cdots,s_{d} of ¯\overline{\mathcal{E}} that define a 𝒪S¯\mathcal{O}_{\bar{S}}-basis of ¯\overline{\mathcal{F}}. Assume as above that we fix a direct summand ¯=Qsi=Q¯\overline{\mathcal{E}}=Q\oplus\langle s_{i}\rangle=Q\oplus\overline{\mathcal{F}}. The following commutative diagram is obtained by functionality

𝕍Q(,s1,,sd)\textstyle{\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{d})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}gQ\scriptstyle{g_{Q}}𝕍0(,s1,,sd)\textstyle{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝕍()\textstyle{\operatorname{\mathbb{V}}(\mathcal{E})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝕍0(,s1,,sd)\textstyle{\operatorname{\mathbb{V}}_{0}(\mathcal{F},s_{1},\cdots,s_{d})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝕍()\textstyle{\operatorname{\mathbb{V}}(\mathcal{F})}

By [1, Lemma 2.5], the diagram on the right hand side is cartesian and the vertical morphisms are principal homogeneous spaces under the group of affine transformations 𝔸Shd\operatorname{\mathbb{A}}_{S}^{h-d}. As a corollary, if f:𝕍()Sf:\operatorname{\mathbb{V}}(\mathcal{E})\rightarrow S and f0:𝕍0(,s1,sd)Sf_{0}:\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1}\cdots,s_{d})\rightarrow S are the structural morphisms, f0,𝒪𝕍0(,s1,,sd)f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})} is endowed with increasing filtrations Filf0,𝒪𝕍0(,s1,,sd){\rm Fil}_{\bullet}f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})} with graded pieces

(6) Grnf0,𝒪𝕍0(,s1,,sd)f0,𝒪𝕍0(,s1,,sd)𝒪SSymn(/).{\rm Gr}_{n}f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})}\simeq f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{F},s_{1},\cdots,s_{d})}\otimes_{\mathcal{O}_{S}}{\rm Sym}^{n}(\mathcal{E}/\mathcal{F}).

If we consider the structural morphism fQ=f0gQ:𝕍Q(,s1,sd)Sf_{Q}=f_{0}\circ g_{Q}:\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1}\cdots,s_{d})\rightarrow S, this defines a filtration FilfQ,𝒪𝕍Q(,s1,,sd){\rm Fil}_{\bullet}f_{Q,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{d})} in fQ,𝒪𝕍Q(,s1,,sd)f_{Q,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{d})},

FilnfQ,𝒪𝕍Q(,s1,,sd):=fQ,𝒪𝕍Q(,s1,,sd)Filnf0,𝒪𝕍0(,s1,,sd).{\rm Fil}_{n}f_{Q,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{d})}:=f_{Q,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{d})}\cap{\rm Fil}_{n}f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})}.

The filtration is characterized by its local description: Over U=Spf(R)SU={\rm Spf}(R)\subset S, an open affine subscheme such that \mathcal{F}, \mathcal{E} are free over UU, we have that

𝕍0(,s1,,sd)U=Spf(RZ1,,Zd),𝕍0(,s1,,sd)U=Spf(RZ1,,Zd,Y1,,Yhd).\operatorname{\mathbb{V}}_{0}(\mathcal{F},s_{1},\cdots,s_{d})\mid_{U}={\rm Spf}(R\langle Z_{1},\cdots,Z_{d}\rangle),\quad\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})\mid_{U}={\rm Spf}(R\langle Z_{1},\cdots,Z_{d},Y_{1},\cdots,Y_{h-d}\rangle).

Then Filnf0,𝒪𝕍0(,s1,,sd)(U){\rm Fil}_{n}f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})}(U) consists of the polynomials of degree n\leq n in the variables Y1,,YhdY_{1},\cdots,Y_{h-d} with coefficients in RZ1,,ZdR\langle Z_{1},\cdots,Z_{d}\rangle. Similarly, if α\alpha\in\mathcal{I} is a generator, we have that 𝕍Q(,s1,,sd)=Spf(RZ1,,Zd,T1,,Thd)\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{d})={\rm Spf}(R\langle Z_{1},\cdots,Z_{d},T_{1},\cdots,T_{h-d}\rangle) and

gQ:RZ1,,Zd,Y1,,YhdRZ1,,Zd,T1,,Thd,gQ(Yi)=αTi.g_{Q}^{\ast}:R\langle Z_{1},\cdots,Z_{d},Y_{1},\cdots,Y_{h-d}\rangle\rightarrow R\langle Z_{1},\cdots,Z_{d},T_{1},\cdots,T_{h-d}\rangle,\qquad g_{Q}^{\ast}(Y_{i})=\alpha T_{i}.

This implies that FilnfQ,𝒪𝕍Q(,s1,,sd)(U){\rm Fil}_{n}f_{Q,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{d})}(U) consists of the polynomials of degree n\leq n in the variables T1,,ThdT_{1},\cdots,T_{h-d} with coefficients in RZ1,,ZdR\langle Z_{1},\cdots,Z_{d}\rangle, and gQg_{Q}^{\ast} provides an isomorphism

(7) Filnf0,𝒪𝕍0(,s1,,sd)(U)[α1]FilnfQ,𝒪𝕍Q(,s1,,sd)(U)[α1].{\rm Fil}_{n}f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})}(U)[\alpha^{-1}]\simeq{\rm Fil}_{n}f_{Q,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{d})}(U)[\alpha^{-1}].

Finally, both constructions of the filtrations are functorial in sense of [1, Corollary 2.7].

4.2.2. Connections

The following part is a brief summary of [1, §2.4]. Suppose that we have a p\operatorname{\mathbb{Z}}_{p}-algebra A0A_{0} and an element ϖA0\varpi\in A_{0} such that A0A_{0} is ϖ\varpi-adically complete and separated. Let SS be a formal scheme locally of finite type over Spf(A0){\rm Spf}(A_{0}) such that the topology of SS is the ϖ\varpi-adic topology. We let ΩS/A01\Omega_{S/A_{0}}^{1} be the 𝒪S\mathcal{O}_{S}-module of continuous Khäler differentials.

Consider a free 𝒪S\mathcal{O}_{S}-module \mathcal{E} endowed with an integrable connection :𝒪SΩS/A01\bigtriangledown:\mathcal{E}\longrightarrow\mathcal{E}\otimes_{\mathcal{O}_{S}}\Omega_{S/A_{0}}^{1}. Assume that we have fixed marked sections s1,,sd¯s_{1},\cdots,s_{d}\in\overline{\mathcal{E}} which are horizontal for the reduction of \bigtriangledown modulo =ϖ𝒪S\mathcal{I}=\varpi\mathcal{O}_{S}. Then by [1, §2.4] \bigtriangledown defines an integrable connection

0:𝕍0(,s1,,sd)𝕍0(,s1,,sd)^𝒪SΩS/A01\bigtriangledown_{0}:\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})\longrightarrow\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})\hat{\otimes}_{\mathcal{O}_{S}}\Omega^{1}_{S/A_{0}}

where ^\hat{\otimes} denotes the completed tensor product.

Moreover, if we assume that we are in the situation of the previous section with a locally free 𝒪S\mathcal{O}_{S}-module and a direct summand \mathcal{F}\subset\mathcal{E}, we consider the filtrations Filf0,𝒪𝕍0(,s1,,sd){\rm Fil}_{\bullet}f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})}. By [1, Lemma 2.9] the connection 0\bigtriangledown_{0} satisfies Griffith’s transversality with respect to the filtration Filf0,𝒪𝕍0(,s1,,sd){\rm Fil}_{\bullet}f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})}, namely for every integer nn we have

0(Filnf0,𝒪𝕍0(,s1,,sd))Filn+1f0,𝒪𝕍0(,s1,,sd)^𝒪SΩS/A01.\bigtriangledown_{0}\left({\rm Fil}_{n}f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})}\right)\subset{\rm Fil}_{n+1}f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})}\hat{\otimes}_{\mathcal{O}_{S}}\Omega_{S/A_{0}}^{1}.

Furthermore the induced map

(8) grn(0):Grnf0,𝒪𝕍0(,s1,,sd)Grn+1f0,𝒪𝕍0(,s1,,sd)^𝒪SΩS/A01{\rm gr}_{n}(\bigtriangledown_{0}):{\rm Gr}_{n}f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})}\longrightarrow{\rm Gr}_{n+1}f_{0,*}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{E},s_{1},\cdots,s_{d})}\hat{\otimes}_{\mathcal{O}_{S}}\Omega_{S/A_{0}}^{1}

is 𝒪S\mathcal{O}_{S}-linear and, via the identification (6), the morphism gr(0){\rm gr}_{\bullet}(\bigtriangledown_{0}) is Sym(/){\rm Sym}^{\bullet}(\mathcal{E}/\mathcal{F})-linear.

4.3. Weight space

We fix a decomposition:

𝒪×𝒪0×H,\mathcal{O}^{\times}\cong\mathcal{O}^{0}\times H,

where HH is the torsion subgroup of 𝒪×\mathcal{O}^{\times} and 𝒪01+p𝒪\mathcal{O}^{0}\simeq 1+p\mathcal{O} is a free p\operatorname{\mathbb{Z}}_{p}-module of rank dd. We put ΛF:=p[[𝒪×]]\Lambda_{F}:=\operatorname{\mathbb{Z}}_{p}[[\mathcal{O}^{\times}]] and ΛF0:=p[[𝒪0]]\Lambda_{F}^{0}:=\operatorname{\mathbb{Z}}_{p}[[\mathcal{O}^{0}]]. The choice of a basis {e1,,ed}\{e_{1},...,e_{d}\} of 𝒪0\mathcal{O}^{0} furnishes an isomorphism ΛF0p[[T1,.,Td]]\Lambda_{F}^{0}\cong\operatorname{\mathbb{Z}}_{p}[[T_{1},....,T_{d}]] given by 1+Ti=ei1+T_{i}=e_{i} for i=1,,di=1,...,d. Moreover, for each nn\in\operatorname{\mathbb{N}} we consider the algebras:

Λn:=ΛFT1pn1p,,Tdpn1pΛn0:=ΛF0T1pn1p,,Tdpn1p\Lambda_{n}:=\Lambda_{F}\left\langle\frac{T_{1}^{p^{n-1}}}{p},\cdots,\frac{T_{d}^{p^{n-1}}}{p}\right\rangle\qquad\qquad\Lambda_{n}^{0}:=\Lambda_{F}^{0}\left\langle\frac{T_{1}^{p^{n-1}}}{p},\cdots,\frac{T_{d}^{p^{n-1}}}{p}\right\rangle

The formal scheme 𝔚=Spf(ΛF)\mathfrak{W}=\mathrm{Spf}(\Lambda_{F}) is our formal weight space for GG^{\prime}. Thus for each complete p\operatorname{\mathbb{Z}}_{p}-algebra RR we have:

𝔚(R)=Homcont(𝒪×,R×).\mathfrak{W}(R)={\rm Hom}_{\mathrm{cont}}(\mathcal{O}^{\times},R^{\times}).

We also consider the following formal schemes 𝔚0=Spf(ΛF0)\mathfrak{W}^{0}=\mathrm{Spf}(\Lambda_{F}^{0}), 𝔚n:=Spf(Λn)\mathfrak{W}_{n}:=\mathrm{Spf}(\Lambda_{n}) and 𝔚n0:=Spf(Λn0)\mathfrak{W}^{0}_{n}:=\mathrm{Spf}(\Lambda_{n}^{0}). By construction we have 𝔚=n𝔚n\mathfrak{W}=\bigcup_{n}\mathfrak{W}_{n} and 𝔚0=n𝔚n0\mathfrak{W}^{0}=\bigcup_{n}\mathfrak{W}^{0}_{n}. Moreover, we have the following explicit description:

𝔚n0(p)={kHomcont(𝒪0,p×):|k(ei)1|ppn+1,i=1,,d}\mathfrak{W}_{n}^{0}(\operatorname{\mathbb{C}}_{p})=\{k\in{\rm Hom}_{\mathrm{cont}}(\mathcal{O}^{0},\operatorname{\mathbb{C}}_{p}^{\times}):\;|k(e_{i})-1|\leq p^{-p^{-n+1}},\;i=1,\cdots,d\}

We denote by 𝐤:𝒪×ΛF×{\bf k}:\mathcal{O}^{\times}\rightarrow\Lambda_{F}^{\times} the universal character of 𝔚\mathfrak{W}, which decomposes as 𝐤=𝐤0𝐤f{\bf k}={\bf k}^{0}\otimes{\bf k}_{f} where:

𝐤f:Hp[H]×𝐤0:𝒪0(ΛF0)×.{\bf k}_{f}:H\longrightarrow\operatorname{\mathbb{Z}}_{p}[H]^{\times}\qquad\qquad{\bf k}^{0}:\mathcal{O}^{0}\longrightarrow(\Lambda_{F}^{0})^{\times}.

Let 𝐤n0:𝒪0(Λn0)×{\bf k}_{n}^{0}:\mathcal{O}^{0}\rightarrow(\Lambda^{0}_{n})^{\times} be given by the composition of 𝐤0{\bf k}^{0} with the inclusion (ΛF0)×(Λn0)×(\Lambda_{F}^{0})^{\times}\subseteq(\Lambda^{0}_{n})^{\times} and we put 𝐤n:=𝐤n0𝐤f:𝒪×Λn×{\bf k}_{n}:={\bf k}_{n}^{0}\otimes{\bf k}_{f}:\mathcal{O}^{\times}\rightarrow\Lambda_{n}^{\times}. The following Lemma can be found in [3, Lemma 6.4], but it also can be deduced from the computations below.

Lemma 4.7.

Let RR be a pp-adically complete Λn0\Lambda^{0}_{n}-algebra. Then 𝐤n0{\bf k}_{n}^{0} extends locally analytically to a character 𝒪0(1+pnλ1𝒪FR)R×\mathcal{O}^{0}(1+p^{n}\lambda^{-1}\mathcal{O}_{F}\otimes R)\rightarrow R^{\times}, for any λR\lambda\in R such that λp1pp2mR\lambda^{p-1}\in p^{p-2}m_{R}, where mRm_{R} is the maximal of RR. In particular, 𝐤n0{\bf k}_{n}^{0} is analytic on 1+pn𝒪1+p^{n}\mathcal{O}.

Recall that hypothesis 2.1 imply a decomposition of rings 𝒪=𝒪𝔭0×𝔭𝔭0𝒪𝔭=p×𝒪τ0\mathcal{O}=\mathcal{O}_{\mathfrak{p}_{0}}\times\prod_{\mathfrak{p}\neq\mathfrak{p}_{0}}\mathcal{O}_{\mathfrak{p}}=\operatorname{\mathbb{Z}}_{p}\times\mathcal{O}^{\tau_{0}} and put 𝒪τ0,0:=1+p𝒪τ0\mathcal{O}^{\tau_{0},0}:=1+p\mathcal{O}^{\tau_{0}}. Analogously as above we introduce:

Λτ0:=p[[p×]]Λτ00:=p[[1+pp]]p[[T]]Λτ0:=p[[𝒪τ0×]]Λτ0,0:=p[[𝒪τ0,0]]p[[T2,,Td]]Λτ0,n:=Λτ0Tpn1pΛτ0,n0:=Λτ00Tpn1pΛnτ0:=Λτ0T2pn1p,,Tdpn1pΛnτ0,0:=Λτ0,0T2pn1p,,Tdpn1p.\displaystyle\begin{gathered}\Lambda_{\tau_{0}}:=\operatorname{\mathbb{Z}}_{p}[[\operatorname{\mathbb{Z}}_{p}^{\times}]]\qquad\qquad\Lambda_{\tau_{0}}^{0}:=\operatorname{\mathbb{Z}}_{p}[[1+p\operatorname{\mathbb{Z}}_{p}]]\simeq\operatorname{\mathbb{Z}}_{p}[[T]]\\ \Lambda^{\tau_{0}}:=\operatorname{\mathbb{Z}}_{p}[[\mathcal{O}^{\tau_{0}\times}]]\qquad\qquad\Lambda^{\tau_{0},0}:=\operatorname{\mathbb{Z}}_{p}[[\mathcal{O}^{\tau_{0},0}]]\simeq\operatorname{\mathbb{Z}}_{p}[[T_{2},\cdots,T_{d}]]\\ \Lambda_{\tau_{0},n}:=\Lambda_{\tau_{0}}\left\langle\frac{T^{p^{n-1}}}{p}\right\rangle\qquad\qquad\Lambda_{\tau_{0},n}^{0}:=\Lambda_{\tau_{0}}^{0}\left\langle\frac{T^{p^{n-1}}}{p}\right\rangle\\ \Lambda^{\tau_{0}}_{n}:=\Lambda^{\tau_{0}}\left\langle\frac{T_{2}^{p^{n-1}}}{p},\cdots,\frac{T_{d}^{p^{n-1}}}{p}\right\rangle\qquad\qquad\Lambda^{\tau_{0},0}_{n}:=\Lambda^{\tau_{0},0}\left\langle\frac{T_{2}^{p^{n-1}}}{p},\cdots,\frac{T_{d}^{p^{n-1}}}{p}\right\rangle.\end{gathered}

Thus, we have decompositions Λn0=Λτ0,n0^Λnτ0,0\Lambda_{n}^{0}=\Lambda_{\tau_{0},n}^{0}\hat{\otimes}\Lambda_{n}^{\tau_{0},0} and Λn=Λτ0,n^Λnτ0\Lambda_{n}=\Lambda_{\tau_{0},n}\hat{\otimes}\Lambda_{n}^{\tau_{0}}. We denote by 𝐤τ0,n0:(1+pp)Λτ0,n0{\bf k}_{\tau_{0},n}^{0}:(1+p\operatorname{\mathbb{Z}}_{p})\longrightarrow\Lambda_{\tau_{0},n}^{0}, 𝐤nτ0,0:𝒪τ0,0Λnτ0,0{\bf k}_{n}^{\tau_{0},0}:\mathcal{O}^{\tau_{0},0}\longrightarrow\Lambda_{n}^{\tau_{0},0}, 𝐤τ0,n:p×Λτ0,n{\bf k}_{\tau_{0},n}:\operatorname{\mathbb{Z}}_{p}^{\times}\rightarrow\Lambda_{\tau_{0},n} and 𝐤nτ0:𝒪τ0×Λnτ0{\bf k}_{n}^{\tau_{0}}:\mathcal{O}^{\tau_{0}\times}\rightarrow\Lambda^{\tau_{0}}_{n} the universal characters. Then we have 𝐤n0=𝐤τ0,n0𝐤nτ0,0{\bf k}_{n}^{0}={\bf k}_{\tau_{0},n}^{0}\otimes{\bf k}_{n}^{\tau_{0},0} and 𝐤n=𝐤τ0,n𝐤nτ0{\bf k}_{n}={\bf k}_{\tau_{0},n}\otimes{\bf k}_{n}^{\tau_{0}}. Moreover,

𝐤τ0,n𝐤nτ0=𝐤n=𝐤n0𝐤f=𝐤τ0,n0𝐤nτ0,0𝐤f.{\bf k}_{\tau_{0},n}\otimes{\bf k}_{n}^{\tau_{0}}={\bf k}_{n}={\bf k}_{n}^{0}\otimes{\bf k}_{f}={\bf k}_{\tau_{0},n}^{0}\otimes{\bf k}_{n}^{\tau_{0},0}\otimes{\bf k}_{f}.

4.3.1. The universal character 𝐤τ0,n0{\bf k}_{\tau_{0},n}^{0}

Recall that the weight space 𝔚τ0,n0=Spf(Λτ0,n0)\mathfrak{W}_{\tau_{0},n}^{0}={\rm Spf}(\Lambda_{\tau_{0},n}^{0}) classifies characters such that

(10) 𝔚τ0,n0(p)={kHomcont(1+pp,p×):|k(exp(p))1|ppn+1}.\mathfrak{W}_{\tau_{0},n}^{0}(\operatorname{\mathbb{C}}_{p})=\left\{k\in{\rm Hom}_{\mathrm{cont}}(1+p\operatorname{\mathbb{Z}}_{p},\operatorname{\mathbb{C}}_{p}^{\times}):\;|k(\exp(p))-1|\leq p^{-p^{-n+1}}\right\}.

In this part we will describe the universal character 𝐤τ0,n0{\bf k}_{\tau_{0},n}^{0}. By the above lemma 𝐤τ0,n0{\bf k}_{\tau_{0},n}^{0} is analytic when restricted to 1+pnp1+p^{n}\operatorname{\mathbb{Z}}_{p}. Moreover, it is given by

1+pp𝐤τ00Λτ0×Λτ0,n×,𝐤τ00(exp(αp))=(1+T)α,αp.1+p\operatorname{\mathbb{Z}}_{p}\stackrel{{\scriptstyle{\bf k}_{\tau_{0}}^{0}}}{{\longrightarrow}}\Lambda_{\tau_{0}}^{\times}\hookrightarrow\Lambda_{\tau_{0},n}^{\times},\qquad{\bf k}_{\tau_{0}}^{0}(\exp(\alpha p))=(1+T)^{\alpha},\quad\alpha\in\operatorname{\mathbb{Z}}_{p}.

Notice that (1+T)α=exp(αun)(1+T)^{\alpha}=\exp(\alpha u_{n}), where un:=log(1+T)pn+2Λτ0,n0u_{n}:=\log(1+T)\in p^{-n+2}\Lambda_{\tau_{0},n}^{0} since

log(1+T)=k1(T)kk=pn+2kpn1(1)k+1Tk(v(k)n+2)pn1kpv(k)(Tpn1p)v(k)n+2kpn1(T)kk,\log(1+T)=-\sum_{k\geq 1}\frac{(-T)^{k}}{k}=p^{-n+2}\sum_{k\in p^{n-1}\operatorname{\mathbb{N}}}(-1)^{k+1}\frac{T^{k-(v(k)-n+2)p^{n-1}}}{kp^{-v(k)}}\left(\frac{T^{p^{n-1}}}{p}\right)^{v(k)-n+2}-\sum_{k\not\in p^{n-1}\operatorname{\mathbb{N}}}\frac{(-T)^{k}}{k},

where v:v:\operatorname{\mathbb{Z}}\rightarrow\operatorname{\mathbb{N}} is the pp-adic valuation. This implies that

(1+T)α\displaystyle(1+T)^{\alpha} =\displaystyle= j=0pn11(1+T)j1j+pn1p(α)exp((αj)un)\displaystyle\sum_{j=0}^{p^{n-1}-1}(1+T)^{j}\cdot 1_{j+p^{n-1}\operatorname{\mathbb{Z}}_{p}}(\alpha)\cdot\exp((\alpha-j)u_{n})
=\displaystyle= j=0pn11(1+T)j1j+pn1p(α)i0(p1uni)(exp(αp)exp(jp)exp(jp))i,\displaystyle\sum_{j=0}^{p^{n-1}-1}(1+T)^{j}\cdot 1_{j+p^{n-1}\operatorname{\mathbb{Z}}_{p}}(\alpha)\cdot\sum_{i\geq 0}\binom{p^{-1}u_{n}}{i}\left(\frac{\exp(\alpha p)-\exp(jp)}{\exp(jp)}\right)^{i},

where the last equality comes from the following equality in [[X,Y]]\operatorname{\mathbb{Q}}[[X,Y]]

exp(Xlog(1+Y))=i0(Xi)Yi,(Xi):=X(X1)(Xi+1)i!.\exp(X\log(1+Y))=\sum_{i\geq 0}\binom{X}{i}Y^{i},\qquad\binom{X}{i}:=\frac{X(X-1)\cdots(X-i+1)}{i!}.
Remark 4.8.

Notice that v(h!)hp1v(h!)\leq\frac{h}{p-1} (see [1, Lemma 4.12]), hence we have

(p1uni)pi(npp1)Λτ0,n0.\binom{p^{-1}u_{n}}{i}\in p^{-i\left(n-\frac{p}{p-1}\right)}\Lambda_{\tau_{0},n}^{0}.

This implies that i0(p1uni)(exp(αp)exp(jp)exp(jp))i\sum_{i\geq 0}\binom{p^{-1}u_{n}}{i}\left(\frac{\exp(\alpha p)-\exp(jp)}{\exp(jp)}\right)^{i} converges since (exp(αp)exp(jp)exp(jp))ipinp\left(\frac{\exp(\alpha p)-\exp(jp)}{\exp(jp)}\right)^{i}\in p^{in}\operatorname{\mathbb{Z}}_{p}.

Thus, we obtain that the locally analytic expression of 𝐤τ0,n0{\bf k}_{\tau_{0},n}^{0} is given by

(11) 𝐤τ0,n0(β)=j=0pn11(1+T)j1exp(jp)+pnp(β)i0(p1uni)(βexp(jp)exp(jp))i.{\bf k}_{\tau_{0},n}^{0}(\beta)=\sum_{j=0}^{p^{n-1}-1}(1+T)^{j}\cdot 1_{\exp(jp)+p^{n}\operatorname{\mathbb{Z}}_{p}}(\beta)\cdot\sum_{i\geq 0}\binom{p^{-1}u_{n}}{i}\left(\frac{\beta-\exp(jp)}{\exp(jp)}\right)^{i}.

The following examples help us to visualize that the above expression can be specialized at both analytic character and locally constant characters that factor through (1+pp)/(1+pn+1p)(1+p\operatorname{\mathbb{Z}}_{p})/(1+p^{n+1}\operatorname{\mathbb{Z}}_{p}).

Exemple 4.9.

Let k𝔚τ0,n0(p)k\in\mathfrak{W}_{\tau_{0},n}^{0}(\operatorname{\mathbb{C}}_{p}) be the point corresponding to the character ββk\beta\mapsto\beta^{k}, for some kk\in\operatorname{\mathbb{Z}}. It clearly satisfies the characterization of Equation (10). It corresponds to the maximal ideal (Texp(kp)+1)(T-\exp(kp)+1). We check that

k𝐤τ0,n0(β)\displaystyle k^{\ast}{\bf k}_{\tau_{0},n}^{0}(\beta) =\displaystyle= j=0pn11exp(pk)j1exp(jp)+pnp(β)i=0k(ki)(βexp(jp)exp(jp))i\displaystyle\sum_{j=0}^{p^{n-1}-1}\exp(pk)^{j}\cdot 1_{\exp(jp)+p^{n}\operatorname{\mathbb{Z}}_{p}}(\beta)\cdot\sum_{i=0}^{k}\binom{k}{i}\left(\frac{\beta-\exp(jp)}{\exp(jp)}\right)^{i}
=\displaystyle= j=0pn11exp(jp)k1exp(jp)+pnp(β)(βexp(jp))k=βk.\displaystyle\sum_{j=0}^{p^{n-1}-1}\exp(jp)^{k}\cdot 1_{\exp(jp)+p^{n}\operatorname{\mathbb{Z}}_{p}}(\beta)\left(\frac{\beta}{\exp(jp)}\right)^{k}=\beta^{k}.
Exemple 4.10.

Let χ𝔚τ0,n0(p)\chi\in\mathfrak{W}_{\tau_{0},n}^{0}(\operatorname{\mathbb{C}}_{p}) be the point corresponding to the character χ(exp(αp))=ξnα\chi(\exp(\alpha p))=\xi_{n}^{\alpha}, for a fix (n1)(n-1)-th root of unity ξn\xi_{n}. It also satisfies the characterization of Equation (10) and corresponds to the maximal ideal (Tξn+1)(T-\xi_{n}+1). We check that

χ𝐤τ0,n0(exp(αp))\displaystyle\chi^{\ast}{\bf k}_{\tau_{0},n}^{0}(\exp(\alpha p)) =\displaystyle= j=0pn11ξnj1exp(jp)+pnp(exp(αp))i0(0i)(exp(αp)exp(jp)exp(jp))i\displaystyle\sum_{j=0}^{p^{n-1}-1}\xi_{n}^{j}\cdot 1_{\exp(jp)+p^{n}\operatorname{\mathbb{Z}}_{p}}(\exp(\alpha p))\cdot\sum_{i\geq 0}\binom{0}{i}\left(\frac{\exp(\alpha p)-\exp(jp)}{\exp(jp)}\right)^{i}
=\displaystyle= j=0pn11ξnj1j+pn1p(α)=ξnα.\displaystyle\sum_{j=0}^{p^{n-1}-1}\xi_{n}^{j}\cdot 1_{j+p^{n-1}\operatorname{\mathbb{Z}}_{p}}(\alpha)=\xi_{n}^{\alpha}.

4.4. Overconvergent modular sheaves

We fix LL a finite extension of p\operatorname{\mathbb{Q}}_{p} containing all the pp-adic embedding of FF and let us work over the ring of integers of LL. Let rnr\geq n. As in §4.1, we consider the ideal of 𝒪𝔊r,n\mathcal{O}_{\mathfrak{IG}_{r,n}} given by n:=pnHdgpnp1\mathcal{I}_{n}:=p^{n}\mathrm{Hdg}^{-\frac{p^{n}}{p-1}}, which is our ideal in order to perform the construction of §4.2. Then using notations from §4.2 we put 𝔊r,n¯\overline{\mathfrak{IG}_{r,n}} for the corresponding reduction modulo n\mathcal{I}_{n}. From Equation (5), we have an isomorphism

(12) dlog0:Dn(𝔊r,n)p(𝒪𝔊r,n/n)Ω0𝒪𝔊r,n(𝒪𝔊r,n/n),\mathrm{dlog}_{0}:D_{n}(\mathfrak{IG}_{r,n})\otimes_{\operatorname{\mathbb{Z}}_{p}}(\mathcal{O}_{\mathfrak{IG}_{r,n}}/\mathcal{I}_{n})\stackrel{{\scriptstyle\simeq}}{{\longrightarrow}}\Omega_{0}\otimes_{\mathcal{O}_{\mathfrak{IG}_{r,n}}}(\mathcal{O}_{\mathfrak{IG}_{r,n}}/\mathcal{I}_{n}),

where Ω0\Omega_{0} is the 𝒪𝔊r,n\mathcal{O}_{\mathfrak{IG}_{r,n}}-submodule of ω0\omega_{0} generated by all the lifts of dlog0(Dn)\mathrm{dlog}_{0}(D_{n}). By construction, there exist on 𝔊r,n\mathfrak{IG}_{r,n} a universal canonical generator P0,nP_{0,n} of DnD_{n}, and universal points P𝔭,nP_{\mathfrak{p},n} of order pnp^{n} in 𝒢𝔭[pn]\mathcal{G}_{\mathfrak{p}}[p^{n}]. We put:

(13) Ω:=Ω0𝔭𝔭0ω𝔭=Ω0Ω0\Omega:=\Omega_{0}\oplus\bigoplus_{\mathfrak{p}\neq\mathfrak{p}_{0}}\omega_{\mathfrak{p}}=\Omega_{0}\oplus\Omega^{0}

where Ω0=𝔭𝔭0ω𝔭\Omega^{0}=\bigoplus_{\mathfrak{p}\neq\mathfrak{p}_{0}}\omega_{\mathfrak{p}}, and we denote Ω¯\overline{\Omega} the associated 𝒪𝔊r,n¯\mathcal{O}_{\overline{\mathfrak{IG}_{r,n}}}-module. Now we produce marked sections in Ω¯\overline{\Omega} as follows. Let 𝔭p\mathfrak{p}\mid p and we consider two cases:

  • if 𝔭=𝔭0\mathfrak{p}=\mathfrak{p}_{0} we denote by s0Ω¯s_{0}\in\overline{\Omega} the image dlog0(P0,n)\mathrm{dlog}_{0}(P_{0,n}) using the isomorphism (12).

  • if 𝔭𝔭0\mathfrak{p}\neq\mathfrak{p}_{0} we firstly consider the decomposition ω𝔭=τΣ𝔭ω𝔭,τ\omega_{\mathfrak{p}}=\oplus_{\tau\in\Sigma_{\mathfrak{p}}}\omega_{\mathfrak{p},\tau} over 𝔊r,n\mathfrak{IG}_{r,n} and the dlog map:

    (14) dlog𝔭:𝒢𝔭[pn]ω𝒢𝔭[pn]D=ω𝔭/pnω𝔭=τΣ𝔭ω𝔭,τ/pnω𝔭,τ.\mathrm{dlog}_{\mathfrak{p}}:\mathcal{G}_{\mathfrak{p}}[p^{n}]\longrightarrow\omega_{\mathcal{G}_{\mathfrak{p}}[p^{n}]^{D}}=\omega_{\mathfrak{p}}/p^{n}\omega_{\mathfrak{p}}=\bigoplus_{\tau\in\Sigma_{\mathfrak{p}}}\omega_{\mathfrak{p},\tau}/p^{n}\omega_{\mathfrak{p},\tau}.

    Hence the image of P𝔭,nP_{\mathfrak{p},n} through dlog𝔭\mathrm{dlog}_{\mathfrak{p}} provides a set of sections {s𝔭,τ}τΣ𝔭\{s_{\mathfrak{p},\tau}\}_{\tau\in\Sigma_{\mathfrak{p}}} of ω𝔭,τ/nω𝔭,τω𝔭/nω𝔭\omega_{\mathfrak{p},\tau}/\mathcal{I}_{n}\omega_{\mathfrak{p},\tau}\subseteq\omega_{\mathfrak{p}}/\mathcal{I}_{n}\omega_{\mathfrak{p}}.

By [3, Lemma 6.5], the set 𝐬:={s0}𝔭𝔭0{s𝔭,τ}τΣ𝔭{\bf s}:=\{s_{0}\}\cup\bigcup_{\mathfrak{p}\neq\mathfrak{p}_{0}}\{s_{\mathfrak{p},\tau}\}_{\tau\in\Sigma_{\mathfrak{p}}} define marked sections for the locally free 𝒪𝔊r,n\mathcal{O}_{\mathfrak{IG}_{r,n}}-module Ω\Omega. Hence we construct the formal scheme 𝕍0(Ω,𝐬)\operatorname{\mathbb{V}}_{0}(\Omega,{\bf s}) over 𝔊r,n\mathfrak{IG}_{r,n}. By construction we have the following tower of formal schemes:

𝕍0(Ω,𝐬)\textstyle{\operatorname{\mathbb{V}}_{0}(\Omega,{\bf s})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝔊r,n\textstyle{\mathfrak{IG}_{r,n}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}gn\scriptstyle{g_{n}}𝔛r.\textstyle{\mathfrak{X}_{r}.}

For any 𝔛r\mathfrak{X}_{r}-scheme TT we have:

𝕍0(Ω,𝐬)(T)={(ρ,φ)𝔊r,n(T)×Γ(T,ρΩ);φ(ρsi)1modn}\operatorname{\mathbb{V}}_{0}(\Omega,{\bf s})(T)=\{(\rho,\varphi)\in\mathfrak{IG}_{r,n}(T)\times\Gamma(T,\rho^{\ast}\Omega^{\vee});\quad\varphi(\rho^{\ast}s_{i})\equiv 1\mod\mathcal{I}_{n}\}

Let 𝐬0:=𝔭𝔭0{s𝔭,τ}τΣ𝔭{\bf s}^{0}:=\bigcup_{\mathfrak{p}\neq\mathfrak{p}_{0}}\{s_{\mathfrak{p},\tau}\}_{\tau\in\Sigma_{\mathfrak{p}}} then from (13) we have:

𝕍0(Ω,𝐬)=𝕍0(Ω0,s0)×𝔊r,n𝕍0(Ω0,𝐬0).\operatorname{\mathbb{V}}_{0}(\Omega,{\bf s})=\operatorname{\mathbb{V}}_{0}(\Omega_{0},s_{0})\times_{\mathfrak{IG}_{r,n}}\operatorname{\mathbb{V}}_{0}(\Omega^{0},{\bf s}^{0}).

As we are interested in locally analytic distributions (rather than functions) we perform the following construction. Let t𝔭,τω¯𝔭,τt_{\mathfrak{p},\tau}\in\bar{\omega}_{\mathfrak{p},\tau} be any section such that φτ(s𝔭,τt𝔭,τ)=1\varphi_{\tau}(s_{\mathfrak{p},\tau}\wedge t_{\mathfrak{p},\tau})=1 and Q𝐬0Ω0¯Q_{{\bf s}^{0}}\subset\overline{\Omega^{0}} be the direct summand generated by the sections in 𝐬0{\bf s}^{0}. We put 𝐭0:=𝔭𝔭0{t𝔭,τ}τΣ𝔭{\bf t}^{0}:=\bigcup_{\mathfrak{p}\neq\mathfrak{p}_{0}}\{t_{\mathfrak{p},\tau}\}_{\tau\in\Sigma_{\mathfrak{p}}} and

𝕍Q𝐬0(Ω0,𝐭0)f0𝔊r,ngn𝔛r,𝕍0(Ω0,s0)f0𝔊r,ngn𝔛r.\operatorname{\mathbb{V}}_{Q_{{\bf s}^{0}}}(\Omega^{0},{\bf t}^{0})\stackrel{{\scriptstyle f^{0}}}{{\longrightarrow}}\mathfrak{IG}_{r,n}\stackrel{{\scriptstyle g_{n}}}{{\longrightarrow}}\mathfrak{X}_{r},\qquad\operatorname{\mathbb{V}}_{0}(\Omega_{0},s_{0})\stackrel{{\scriptstyle f_{0}}}{{\longrightarrow}}\mathfrak{IG}_{r,n}\stackrel{{\scriptstyle g_{n}}}{{\longrightarrow}}\mathfrak{X}_{r}.

The sections t𝔭,τt_{\mathfrak{p},\tau} are well defined modulo Q𝐬0Q_{{\bf s}^{0}} which is fine because remark 4.6. We have

𝕍Q𝐬0(Ω0,𝐭0)(T)={(ρ,φ)𝔊r,n(T)×Γ(T,ρ(Ω0));φ(ρt𝔭,τ)1,φ(ρs𝔭,τ)0modn}.\operatorname{\mathbb{V}}_{Q_{{\bf s}^{0}}}(\Omega^{0},{\bf t}^{0})(T)=\{(\rho,\varphi)\in\mathfrak{IG}_{r,n}(T)\times\Gamma(T,\rho^{\ast}(\Omega^{0})^{\vee});\quad\varphi(\rho^{\ast}t_{\mathfrak{p},\tau})\equiv 1,\;\varphi(\rho^{\ast}s_{\mathfrak{p},\tau})\equiv 0\mod\mathcal{I}_{n}\}.

The morphism gng_{n} is endowed with an action of (𝒪/pn𝒪)×(\mathcal{O}/p^{n}\mathcal{O})^{\times}, then both 𝕍Q𝐬0(Ω0,𝐭0)/𝔛r\operatorname{\mathbb{V}}_{Q_{{\bf s}^{0}}}(\Omega^{0},{\bf t}^{0})/\mathfrak{X}_{r} and 𝕍0(Ω0,s0)/𝔛r\operatorname{\mathbb{V}}_{0}(\Omega_{0},s_{0})/\mathfrak{X}_{r} are equipped with actions of 𝒪×(1+nRes𝒪F/𝔾a)Res𝒪/p𝔾m\mathcal{O}^{\times}(1+\mathcal{I}_{n}{\rm Res}_{\mathcal{O}_{F}/\operatorname{\mathbb{Z}}}\operatorname{\mathbb{G}}_{a})\subseteq{\rm Res}_{\mathcal{O}/\operatorname{\mathbb{Z}}_{p}}\operatorname{\mathbb{G}}_{m} (see [3, §6.3]).

Since rnr\geq n by Lemma 4.7 (with λ=Hdgpnp1\lambda={\rm Hdg}^{\frac{p^{n}}{p-1}}) the character 𝐤n0{\bf k}_{n}^{0} extends to a locally analytic character

𝐤n0:𝒪×(1+𝒪Fn𝒪𝔛rpΛn0)𝒪𝔛rpΛn0.{\bf k}_{n}^{0}:\mathcal{O}^{\times}(1+\mathcal{O}_{F}\otimes_{\operatorname{\mathbb{Z}}}\mathcal{I}_{n}\mathcal{O}_{\mathfrak{X}_{r}}\otimes_{\operatorname{\mathbb{Z}}_{p}}\Lambda_{n}^{0})\longrightarrow\mathcal{O}_{\mathfrak{X}_{r}}\otimes_{\operatorname{\mathbb{Z}}_{p}}\Lambda_{n}^{0}.
Definition 4.11.

We consider the following sheaves over 𝔛r×𝔚n\mathfrak{X}_{r}\times\mathfrak{W}_{n}:

n\displaystyle\mathcal{F}_{n} :=\displaystyle:= ((gnf0)𝒪𝕍Q𝐬0(Ω0,𝐭0)^Λn)[𝐤nτ0,0],Ω0𝐤τ0,n0:=((gnf0)𝒪𝕍0(Ω0,s0)^Λn)[𝐤τ0,n0],\displaystyle\left((g_{n}\circ f^{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q_{{\bf s}^{0}}}(\Omega^{0},{\bf t}^{0})}\hat{\otimes}\Lambda_{n}\right)[{\bf k}_{n}^{\tau_{0},0}],\qquad\Omega_{0}^{{\bf k}_{\tau_{0},n}^{0}}:=\left((g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\Omega_{0},s_{0})}\hat{\otimes}\Lambda_{n}\right)[{\bf k}_{\tau_{0},n}^{0}],
Ω𝐤f\displaystyle\Omega^{{\bf k}_{f}} :=\displaystyle:= (g1,(𝒪𝔊1)^Λn)[𝐤f].\displaystyle\left(g_{1,\ast}(\mathcal{O}_{\mathfrak{IG}_{1}})\hat{\otimes}\Lambda_{n}\right)[{\bf k}_{f}].

The formal overconvergent modular sheaf over 𝔛r×𝔚n\mathfrak{X}_{r}\times\mathfrak{W}_{n} is defined as

Ω𝐤n:=Ω0𝐤τ0,n0𝒪𝔛r×𝔚nn𝒪𝔛r×𝔚nΩ𝐤f,n:=\calligraom𝔛r×𝔚n(n,𝒪𝔛r×𝔚n).\Omega^{{\bf k}_{n}}:=\Omega_{0}^{{\bf k}_{\tau_{0},n}^{0}}\otimes_{\mathcal{O}_{\mathfrak{X}_{r}\times\mathfrak{W}_{n}}}\mathcal{F}_{n}^{\vee}\otimes_{\mathcal{O}_{\mathfrak{X}_{r}\times\mathfrak{W}_{n}}}\Omega^{{\bf k}_{f}},\qquad\mathcal{F}_{n}^{\vee}:=\operatorname{\mathscr{H}\text{\kern-3.0pt{\calligra\large om}}}_{\mathfrak{X}_{r}\times\mathfrak{W}_{n}}\left(\mathcal{F}_{n},\mathcal{O}_{\mathfrak{X}_{r}\times\mathfrak{W}_{n}}\right).
Definition 4.12.

A section in H0(𝔛r×𝔚n,Ω𝐤n)\mathrm{H}^{0}(\mathfrak{X}_{r}\times\mathfrak{W}_{n},\Omega^{{\bf k}_{n}}) is called a family of locally analytic Overconvergent modular forms.

4.5. Overconvergent modular forms à la Katz

Here we give a moduli description of the families of overconvergent modular forms introduced above.

4.5.1. Notations

Let RR be a complete local 𝒪^\hat{\mathcal{O}}-algebra.

Definition 4.13.

Let k:𝒪τ0×Rk:\mathcal{O}^{\tau_{0}\times}\rightarrow R be a character and nn\in\operatorname{\mathbb{N}}. We denote by Cnk(𝒪τ0,R)C^{k}_{n}(\mathcal{O}^{\tau_{0}},R) the RR-module of the functions f:𝒪τ0××𝒪τ0Rf:\mathcal{O}^{\tau_{0}\times}\times\mathcal{O}^{\tau_{0}}\rightarrow R such that:

  • f(tx,ty)=k(t)f(x,y)f(tx,ty)=k(t)\cdot f(x,y) for each t𝒪τ0×t\in\mathcal{O}^{\tau_{0}\times} and (x,y)𝒪τ0××𝒪τ0(x,y)\in\mathcal{O}^{\tau_{0}\times}\times\mathcal{O}^{\tau_{0}};

  • the function yf(1,y)y\mapsto f(1,y) is analytic on the disks y0+pn𝒪τ0y_{0}+p^{n}\mathcal{O}^{\tau_{0}} where y0y_{0} below to a system of representatives of 𝒪τ0/pn𝒪τ0\mathcal{O}^{\tau_{0}}/p^{n}\mathcal{O}^{\tau_{0}}.

The space of distributions is defined by Dnk(𝒪τ0,R):=HomR(Cnk(𝒪τ0,R),R).D^{k}_{n}(\mathcal{O}^{\tau_{0}},R):={\rm Hom}_{R}(C^{k}_{n}(\mathcal{O}^{\tau_{0}},R),R).

Remark 4.14.

Note that Cnk(𝒪τ0,R)Cn+1k(𝒪τ0,R)C^{k}_{n}(\mathcal{O}^{\tau_{0}},R)\subseteq C^{k}_{n+1}(\mathcal{O}^{\tau_{0}},R) and if k=k¯[ΣF{τ0}]k=\underline{k}\in\operatorname{\mathbb{N}}[\Sigma_{F}\smallsetminus\{\tau_{0}\}] is a classical weight then C0k¯(𝒪τ0,R)C^{\underline{k}}_{0}(\mathcal{O}^{\tau_{0}},R) is the module of analytic functions and naturally contains Symk¯(R2)\mathrm{Sym}^{\underline{k}}(R^{2}). We obtain a natural projection D0k¯(𝒪τ0,R)Symk¯(R2)D^{\underline{k}}_{0}(\mathcal{O}^{\tau_{0}},R)\rightarrow\mathrm{Sym}^{\underline{k}}(R^{2})^{\vee}.

We have a natural action of the subgroup K0(p)τ0GL2(𝒪τ0)K_{0}(p)^{\tau_{0}}\subset\operatorname{\mathrm{GL}}_{2}(\mathcal{O}^{\tau_{0}}) of upper triangular matrices modulo pp on Cnk(𝒪τ0,R)C^{k}_{n}(\mathcal{O}^{\tau_{0}},R) and Dnk(𝒪τ0,R)D^{k}_{n}(\mathcal{O}^{\tau_{0}},R) given by:

(gf)(x,y)=f((x,y)g)(gμ)(f):=μ(g1f),(g\ast f)(x,y)=f((x,y)g)\qquad\qquad(g\ast\mu)(f):=\mu(g^{-1}\ast f),

where gK0(p)τ0g\in K_{0}(p)^{\tau_{0}}, fCnk(𝒪τ0,R)f\in C^{k}_{n}(\mathcal{O}^{\tau_{0}},R) and μDnk(𝒪τ0,R)\mu\in D^{k}_{n}(\mathcal{O}^{\tau_{0}},R). Since yf(1,y)y\mapsto f(1,y) is analytic on the disks y0+pn𝒪τ0y_{0}+p^{n}\mathcal{O}^{\tau_{0}} this action extends to an action of K0(p)τ0(1+pnM2(Rp𝒪τ0))K_{0}(p)^{\tau_{0}}(1+p^{n}\mathrm{M}_{2}(R\otimes_{\operatorname{\mathbb{Z}}_{p}}\mathcal{O}^{\tau_{0}})).

4.5.2. Locally analytic distributions

We are going to describe the elements of H0(𝔛r×𝔚n,Ω𝐤n)H^{0}(\mathfrak{X}_{r}\times\mathfrak{W}_{n},\Omega^{{\bf k}_{n}}) as rules, extending the classical interpretation due to Katz. As always, we are assuming that rnr\geq n.

Now let RR be a Λn\Lambda_{n}-algebra. Recall that a tuple (A,ι,θ,α𝔭0)(A,\iota,\theta,\alpha^{\mathfrak{p}^{0}}) defined over RR corresponds to a point in 𝔛r(R)\mathfrak{X}_{r}(R). We will denote by w=(f0,{(fτ,eτ)}τ)w=(f_{0},\{(f_{\tau},e_{\tau})\}_{\tau}) the basis of Ω\Omega such that f0f_{0} is a basis of Ω0\Omega_{0} and {eτ,fτ}\{e_{\tau},f_{\tau}\} is a basis of ω𝔭,τ\omega_{\mathfrak{p},\tau}, where 𝔭𝔭0\mathfrak{p}\neq\mathfrak{p}_{0} and τΣ𝔭\tau\in\Sigma_{\mathfrak{p}}. As seen in [3, §6.5.2], any section μH0(𝔛r×𝔚n,Ω𝐤n)\mu\in H^{0}(\mathfrak{X}_{r}\times\mathfrak{W}_{n},\Omega^{{\bf k}_{n}}) is characterized by a rule that assigns to each tuple (A,ι,θ,α𝔭0,w)(A,\iota,\theta,\alpha^{\mathfrak{p}^{0}},w) over RR, a distribution μ(A,ι,θ,α𝔭0,w)Dn𝐤nτ0(𝒪τ0,R)\mu(A,\iota,\theta,\alpha^{\mathfrak{p}^{0}},w)\in D^{{\bf k}_{n}^{\tau_{0}}}_{n}(\mathcal{O}^{\tau_{0}},R). The rule (A,ι,θ,α𝔭0,w)μ(A,ι,θ,α𝔭0,w)(A,\iota,\theta,\alpha^{\mathfrak{p}_{0}},w)\mapsto\mu(A,\iota,\theta,\alpha^{\mathfrak{p}_{0}},w) satisfies:

  • (B1)

    μ(A,ι,θ,α𝔭0,w)\mu(A,\iota,\theta,\alpha^{\mathfrak{p}_{0}},w) depends only on the RR-isomorphism class of (A,ι,θ,α𝔭0)(A,\iota,\theta,\alpha^{\mathfrak{p}_{0}}).

  • (B2)

    The formation of μ(A,ι,θ,α𝔭0,w)\mu(A,\iota,\theta,\alpha^{\mathfrak{p}_{0}},w) commutes with arbitrary extensions of scalars RRR\rightarrow R^{\prime} of Λn\Lambda_{n}-algebras.

  • (B3-a)

    μ(A,ι,θ,α𝔭0,a1w)=knτ0(t)μ(A,ι,θ,α𝔭0,w)\mu(A,\iota,\theta,\alpha^{\mathfrak{p}_{0}},a^{-1}w)=k_{n}^{\tau_{0}}(t)\cdot\mu(A,\iota,\theta,\alpha^{\mathfrak{p}_{0}},w), for all tp×t\in\operatorname{\mathbb{Z}}_{p}^{\times}.

  • (B3-b)

    gμ(A,ι,θ,α𝔭0,wg)=μ(A,ι,θ,α𝔭0,w)g\ast\mu(A,\iota,\theta,\alpha^{\mathfrak{p}_{0}},wg)=\mu(A,\iota,\theta,\alpha^{\mathfrak{p}_{0}},w), for all gK0(p)τ0g\in K_{0}(p)^{\tau_{0}}.

Remark 4.15.

Note that this description fits with the classical setting of classical Katz modular forms explained in §3.3.

5. Nearly overconvergent modular forms

Let us consider the sheaf 1πΩ𝐀/𝔛\mathcal{R}^{1}\pi_{\ast}\Omega_{{\bf A}/\mathfrak{X}}^{\bullet}, where 𝐀𝔛{\bf A}\rightarrow\mathfrak{X} is the universal abelian variety. By the moduli interpretation of 𝔛\mathfrak{X}

:=(1πΩ𝐀/𝔛)+,2=0𝔭𝔭0ω𝔭.\mathcal{H}:=\left(\mathcal{R}^{1}\pi_{\ast}\Omega_{{\bf A}/\mathfrak{X}}^{\bullet}\right)^{+,2}=\mathcal{H}_{0}\oplus\bigoplus_{\mathfrak{p}\neq\mathfrak{p}_{0}}\omega_{\mathfrak{p}}.

Notice that 0\mathcal{H}_{0} is endowed with:

  • A connection :00𝒪𝔛Ω𝔛/𝒪^1\triangledown:\mathcal{H}_{0}\rightarrow\mathcal{H}_{0}\otimes_{\mathcal{O}_{\mathfrak{X}}}\Omega_{\mathfrak{X}/\hat{\mathcal{O}}}^{1}.

  • A Hodge filtration

    (15) 0ω00ϵω010,0\longrightarrow\omega_{0}\longrightarrow\mathcal{H}_{0}\stackrel{{\scriptstyle\epsilon}}{{\longrightarrow}}\omega_{0}^{-1}\longrightarrow 0,

    (see [3, §9.1]).

Fixing integers rnr\geq n as above and having the morphisms 𝔊r,n𝔛r,n𝔛r𝔛\mathfrak{IG}_{r,n}\rightarrow\mathfrak{X}_{r,n}\rightarrow\mathfrak{X}_{r}\rightarrow\mathfrak{X}, we can base-change the triple (0,,Fil)(\mathcal{H}_{0},\triangledown,{\rm Fil}^{\bullet}) over 𝔊r,n\mathfrak{IG}_{r,n} and denote it the same way.

Write 0:=Ω0+δ¯p0\mathcal{H}_{0}^{\sharp}:=\Omega_{0}+\underline{\delta}^{p}\mathcal{H}_{0}. Since δ¯\underline{\delta} is a locally free 𝒪𝔊r,n\mathcal{O}_{\mathfrak{IG}_{r,n}}-module of rank 1, 0\mathcal{H}_{0}^{\sharp} is a 𝒪𝔊r,n\mathcal{O}_{\mathfrak{IG}_{r,n}}-module of rank 2 with Hodge filtration

(16) 0Ω00ϵδ¯pω010.0\longrightarrow\Omega_{0}\longrightarrow\mathcal{H}_{0}^{\sharp}\stackrel{{\scriptstyle\epsilon}}{{\longrightarrow}}\underline{\delta}^{p}\omega_{0}^{-1}\longrightarrow 0.

This implies that s0s_{0} is also a marked section of 0\mathcal{H}_{0}^{\sharp}. Similarly as in [1, Lemma 6.1] one can prove that the construction of the exact sequence (16) is functorial.

Lemma 5.1.

For any lift s~0Ω0\tilde{s}_{0}\in\Omega_{0} of s0s_{0} and any generator DD of the space of derivations of 𝔛\mathfrak{X}, the subspace Q=Hdg(D)(s~0)(modn)Q={\rm Hdg}\langle\triangledown(D)(\tilde{s}_{0})\rangle\;({\rm mod}\,\mathcal{I}_{n}) defines a canonical direct summand 0¯=s0Q\overline{\mathcal{H}_{0}^{\sharp}}=\langle s_{0}\rangle\oplus Q.

Proof.

First, let me remark that the definition of QQ does not depend on the choice of DD nor the lift s~0\tilde{s}_{0}. Indeed, for any αn=pnδ¯pn\alpha\in\mathcal{I}_{n}=p^{n}\underline{\delta}^{-p^{n}}

Hdg(D)(s~0+αs)=Hdg((D)(s~0)+Dαs+α(D)(s))Hdg(D)(s~0)+HdgDαsmodn,{\rm Hdg}\triangledown(D)(\tilde{s}_{0}+\alpha s)={\rm Hdg}(\triangledown(D)(\tilde{s}_{0})+D\alpha s+\alpha\triangledown(D)(s))\equiv{\rm Hdg}\triangledown(D)(\tilde{s}_{0})+{\rm Hdg}D\alpha s\mod\,\mathcal{I}_{n},

But Dαp2nδ¯pn1=pnδ¯1nD\alpha\in p^{2n}\underline{\delta}^{-p^{n}-1}=p^{n}\underline{\delta}^{-1}\mathcal{I}_{n}, thus HdgDαs=δ¯p1Dαs0{\rm Hdg}D\alpha s=\underline{\delta}^{p-1}D\alpha s\equiv 0 modulo n\mathcal{I}_{n}.

In order to prove the claim, we will work locally. Let ρ:Spf(R)𝔊r,n\rho:\mathrm{Spf}(R)\rightarrow\mathfrak{IG}_{r,n} be a morphism of formal schemes without pp-torsion such that ρω0\rho^{\ast}\omega_{0}, and ρ0\rho^{\ast}\mathcal{H}_{0} are free RR-modules of rank 1 and 2, respectively. We choose basis ρω0=Rf\rho^{\ast}\omega_{0}=Rf, ρ0=Rf+Re\rho^{\ast}\mathcal{H}_{0}=Rf+Re. Recall that the Kodaira-Spencer isomorphism KSKS is obtained from restricting \triangledown to ω0\omega_{0} and composing with the morphism ϵ\epsilon of (15). Thus, if the derivation DD is dual to Θ=KS(f,e)Ω𝔛/p1\Theta=KS(f,e)\in\Omega_{\mathfrak{X}/\operatorname{\mathbb{Z}}_{p}}^{1}, we have that

(D)(f)=af+e,for some aR.\triangledown(D)(f)=af+e,\quad\mbox{for some }a\in R.

Assume also that ρδ¯=δR\rho^{\ast}\underline{\delta}=\delta R. Moreover, we can choose δ\delta so that s~0=δf\tilde{s}_{0}=\delta f. Hence we obtain

(D)(s~0)=(D)(δf)=(Dδ+δa)f+δe.\triangledown(D)(\tilde{s}_{0})=\triangledown(D)(\delta f)=(D\delta+\delta a)f+\delta e.

Since ρHdg=δp1R\rho^{\ast}{\rm Hdg}=\delta^{p-1}R, we obtain

Hdg(D)(s~0)=δp1(D)(s~0)R=((δp2Dδ+δp1a)δf+δpe)R0.{\rm Hdg}\langle\triangledown(D)(\tilde{s}_{0})\rangle=\delta^{p-1}\triangledown(D)(\tilde{s}_{0})R=((\delta^{p-2}D\delta+\delta^{p-1}a)\delta f+\delta^{p}e)R\subset\mathcal{H}_{0}^{\sharp}.

Since ρ0=δfRδpeE\rho^{\ast}\mathcal{H}_{0}^{\sharp}=\delta fR\oplus\delta^{p}eE, we obtain that 0¯=s0Q\overline{\mathcal{H}_{0}^{\sharp}}=\langle s_{0}\rangle\oplus Q. ∎

The above lemma provides all the ingredients to construct the formal vector bundle 𝕍Q(0,s0)𝔛r\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})\rightarrow\mathfrak{X}_{r}. Similarly as in [3, §6.3], we define an action of p×(1+n𝔾a)\operatorname{\mathbb{Z}}_{p}^{\times}(1+\mathcal{I}_{n}\operatorname{\mathbb{G}}_{a}) on 𝕍Q(0,s0)𝔛r\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})\rightarrow\mathfrak{X}_{r} as follows: For any 𝔛r\mathfrak{X}_{r}-scheme TT

𝕍Q(0,s0)(T)={(ρ,φ)𝔊r,n(T)×Γ(T,ρ(0));φ(ρs0)1,φ(ρQ)0modn},\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})(T)=\{(\rho,\varphi)\in\mathfrak{IG}_{r,n}(T)\times\Gamma(T,\rho^{\ast}(\mathcal{H}_{0}^{\sharp})^{\vee});\quad\varphi(\rho^{\ast}s_{0})\equiv 1,\;\varphi(\rho^{\ast}Q)\equiv 0\mod\mathcal{I}_{n}\},

and the action of λ(1+γ)p×(1+n𝔾a)\lambda(1+\gamma)\in\operatorname{\mathbb{Z}}_{p}^{\times}(1+\mathcal{I}_{n}\operatorname{\mathbb{G}}_{a}) is given by λ(1+γ)(ρ,φ):=(λρ,λ(1+γ)φ)\lambda(1+\gamma)\ast(\rho,\varphi):=(\lambda\rho,\lambda(1+\gamma)\ast\varphi), where λφ(w)=φ(λw)\lambda\ast\varphi(w)=\varphi(\lambda w).

Definition 5.2.

We consider the following sheaf over 𝔛r×𝔚n\mathfrak{X}_{r}\times\mathfrak{W}_{n}:

𝕎0𝐤τ0,n0:=((gnf0)𝒪𝕍Q(0,s0)^Λn)[𝐤τ0,n0],\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}}:=\left((g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}\hat{\otimes}\Lambda_{n}\right)[{\bf k}_{\tau_{0},n}^{0}],

corresponding to sections such that ts=𝐤τ0,n0(t)st\ast s={\bf k}_{\tau_{0},n}^{0}(t)s, for all tp×(1+n𝔾a)t\in\operatorname{\mathbb{Z}}_{p}^{\times}(1+\mathcal{I}_{n}\operatorname{\mathbb{G}}_{a}). The formal nearly overconvergent modular sheaf over 𝔛r×𝔚n\mathfrak{X}_{r}\times\mathfrak{W}_{n} is defined as

𝕎𝐤n:=𝕎0𝐤τ0,n0𝒪𝔛r×𝔚nn𝒪𝔛r×𝔚nΩ𝐤f.\operatorname{\mathbb{W}}^{{\bf k}_{n}}:=\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}}\otimes_{\mathcal{O}_{\mathfrak{X}_{r}\times\mathfrak{W}_{n}}}\mathcal{F}_{n}^{\vee}\otimes_{\mathcal{O}_{\mathfrak{X}_{r}\times\mathfrak{W}_{n}}}\Omega^{{\bf k}_{f}}.
Definition 5.3.

A section in H0(𝔛r×𝔚n,𝕎𝐤n)\mathrm{H}^{0}(\mathfrak{X}_{r}\times\mathfrak{W}_{n},\operatorname{\mathbb{W}}^{{\bf k}_{n}}) is called a family of locally analytic nearly overconvergent modular forms.

The inclusion Ω00\Omega_{0}\subset\mathcal{H}_{0}^{\sharp} provides a filtration of locally free sheaves with marked sections (Ω0,s0)(0,s0)(\Omega_{0},s_{0})\hookrightarrow(\mathcal{H}_{0}^{\sharp},s_{0}). By §4.2.1, the sheaf (gnf0)𝒪𝕍Q(0,s0)(g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})} has a canonical filtration Fil(gnf0)𝒪𝕍Q(0,s0):=(gnf0)Fil𝒪𝕍Q(0,s0){\rm Fil}_{\bullet}(g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}:=(g_{n}\circ f_{0})_{\ast}{\rm Fil}_{\bullet}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}.

Analogously as in [1, Theorem 3.11.], the action of p×(1+n𝔾a)\operatorname{\mathbb{Z}}_{p}^{\times}(1+\mathcal{I}_{n}\operatorname{\mathbb{G}}_{a}) on (gnf0)𝒪𝕍Q(0,s0)(g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})} preserves the filtration (gnf0)Fil𝒪𝕍Q(0,s0)(g_{n}\circ f_{0})_{\ast}{\rm Fil}_{\bullet}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}. Moreover, if we define

Filn𝕎0𝐤τ0,n0:=((gnf0)Filn𝒪𝕍Q(0,s0)^Λn)[𝐤τ0,n0]((gnf0)𝒪𝕍Q(0,s0)^Λn)[𝐤τ0,n0]=𝕎0𝐤τ0,n0.{\rm Fil}_{n}\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}}:=\left((g_{n}\circ f_{0})_{\ast}{\rm Fil}_{n}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}\hat{\otimes}\Lambda_{n}\right)[{\bf k}_{\tau_{0},n}^{0}]\subset\left((g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}\hat{\otimes}\Lambda_{n}\right)[{\bf k}_{\tau_{0},n}^{0}]=\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}}.

Similarly as in [1, §3.3.1] we have

  • (i)

    Filn𝕎0𝐤τ0,n0{\rm Fil}_{n}\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}} is a locally free 𝒪𝔛r×𝔚n\mathcal{O}_{\mathfrak{X}_{r}\times\mathfrak{W}_{n}}-module for the Zariski topology on 𝔛r×𝔚n\mathfrak{X}_{r}\times\mathfrak{W}_{n};

  • (ii)

    𝕎0𝐤τ0,n0\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}} is the pp-adic completion of limnFiln𝕎0𝐤τ0,n0\lim_{n}{\rm Fil}_{n}\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}};

  • (iii)

    Fil0𝕎0𝐤τ0,n0Ω0𝐤τ0,n0{\rm Fil}_{0}\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}}\simeq\Omega_{0}^{{\bf k}_{\tau_{0},n}^{0}}.

Moreover, if we define Filn𝕎𝐤n:=Filn𝕎0𝐤τ0,n0𝒪𝔛r×𝔚nn𝒪𝔛r×𝔚nΩ𝐤f{\rm Fil}_{n}\operatorname{\mathbb{W}}^{{\bf k}_{n}}:={\rm Fil}_{n}\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}}\otimes_{\mathcal{O}_{\mathfrak{X}_{r}\times\mathfrak{W}_{n}}}\mathcal{F}_{n}^{\vee}\otimes_{\mathcal{O}_{\mathfrak{X}_{r}\times\mathfrak{W}_{n}}}\Omega^{{\bf k}_{f}}, we obtain an increasing filtration {Filn𝕎𝐤n}n\{{\rm Fil}_{n}\operatorname{\mathbb{W}}^{{\bf k}_{n}}\}_{n} by directs summands such that claims (ii)(ii) and (iii)(iii) hold replacing Ω0𝐤τ0,n0\Omega_{0}^{{\bf k}_{\tau_{0},n}^{0}} with Ω𝐤n\Omega^{{\bf k}_{n}}.

Finally, if kk\in\operatorname{\mathbb{N}} is a classical weight then we have a canonical identification

Symk(0)[1/p]=Filk(𝕎0k)[1/p]:=k(Filk(𝕎0𝐤τ0,n0)𝒪𝔛r×𝔚nΩ𝐤f)[1/p].{\rm Sym}^{k}(\mathcal{H}_{0})[1/p]={\rm Fil}_{k}(\operatorname{\mathbb{W}}_{0}^{k})[1/p]:=k^{\ast}\left({\rm Fil}_{k}(\operatorname{\mathbb{W}}_{0}^{{\bf k}^{0}_{\tau_{0},n}})\otimes_{\mathcal{O}_{\mathfrak{X}_{r}\times\mathfrak{W}_{n}}}\Omega^{{\bf k}_{f}}\right)[1/p].

as sheaves on the corresponding adic space 𝒳r×𝒲n\mathcal{X}_{r}\times\mathcal{W}_{n} compatible with the Hodge filtration on Symk(0){\rm Sym}^{k}(\mathcal{H}_{0}).

Remark 5.4.

We omit the proof of the above facts because it is completely analogous to [1, §3.3.3] using results of [2] and §4.2.1.

5.1. Gauss-Manin connections

Consider the morphism of adic spaces 𝒢¯r,n𝒢r,n\overline{\mathcal{IG}}_{r,n}\rightarrow\mathcal{IG}_{r,n} defined by the trivializations 𝒢0[𝔭0n](/pn)2\mathcal{G}_{0}[\mathfrak{p}_{0}^{n}]\simeq(\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}})^{2} compatible with the trivializations Dn/pnD_{n}\simeq\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}}. Let 𝔊¯r,n𝔊r,n\overline{\mathfrak{IG}}_{r,n}\rightarrow\mathfrak{IG}_{r,n} be the normalization. It follows from [1, Proposition 6.3] and §4.2.2 that over 𝔊¯r,n\overline{\mathfrak{IG}}_{r,n} the sheaf

𝕎¯0𝐤τ0,n0:=((gnf0)𝒪𝕍0(0,s0)^Λn)[𝐤τ0,n0],\overline{\operatorname{\mathbb{W}}}_{0}^{{\bf k}_{\tau_{0},n}^{0}}:=\left((g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{H}_{0}^{\sharp},s_{0})}\hat{\otimes}\Lambda_{n}\right)[{\bf k}_{\tau_{0},n}^{0}],

admits an integrable connection relatively to Λτ0,n0\Lambda_{\tau_{0},n}^{0} for which Fil𝕎¯0𝐤τ0,n0{\rm Fil}_{\bullet}\overline{\operatorname{\mathbb{W}}}_{0}^{{\bf k}_{\tau_{0},n}^{0}} satisfies Griffiths’ transversality.

Recall that, by (11), the universal character 𝐤τ0,n0{\bf k}_{\tau_{0},n}^{0} is given by

𝐤τ0,n0(β)=j=0pn11(1+T)j1exp(jp)+pnp(β)i0(wni)(βexp(jp)exp(jp))i.{\bf k}_{\tau_{0},n}^{0}(\beta)=\sum_{j=0}^{p^{n-1}-1}(1+T)^{j}\cdot 1_{\exp(jp)+p^{n}\operatorname{\mathbb{Z}}_{p}}(\beta)\cdot\sum_{i\geq 0}\binom{w_{n}}{i}\left(\frac{\beta-\exp(jp)}{\exp(jp)}\right)^{i}.

where wn=p1log(1+T)pn+1Λτ0,n0w_{n}=p^{-1}\log(1+T)\in p^{-n+1}\Lambda_{\tau_{0},n}^{0}.

Theorem 5.5.

The above connection descends to an integrable connection

𝐤τ0,n:𝕎0𝐤τ0,n0𝕎0𝐤τ0,n0𝒪𝔛δ¯cnΩ𝔛/Λn1;𝔛:=𝔛r×𝔚n,\bigtriangledown_{{\bf k}_{\tau_{0},n}}:\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}}\longrightarrow\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}}\otimes_{\mathcal{O}_{\mathfrak{X}}}\underline{\delta}^{-c_{n}}\Omega^{1}_{\mathfrak{X}/\Lambda_{n}};\qquad\mathfrak{X}:=\mathfrak{X}_{r}\times\mathfrak{W}_{n},

for some cnc_{n}\in\operatorname{\mathbb{N}} depending on nn, such that the induced 𝒪𝔛\mathcal{O}_{\mathfrak{X}}-linear map on the nn graded piece

Grh(𝐤τ0,n):Grn(𝕎0𝐤τ0,n0)[1/p]Grh+1(𝕎0𝐤τ0,n0)𝒪𝔛Ω𝔛/Λn1[1/p]{\rm Gr}_{h}(\bigtriangledown_{{\bf k}_{\tau_{0},n}}):{\rm Gr}_{n}(\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}})[1/p]\longrightarrow{\rm Gr}_{h+1}(\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}})\otimes_{\mathcal{O}_{\mathfrak{X}}}\Omega^{1}_{\mathfrak{X}/\Lambda_{n}}[1/p]

is an isomorphism times wnhw_{n}-h.

Proof.

Let us work locally. Let ρ:Spf(R)𝔊r,n\rho:\mathrm{Spf}(R)\rightarrow\mathfrak{IG}_{r,n} be a morphism of formal schemes without pp-torsion such that ρω0\rho^{\ast}\omega_{0}, and ρ0\rho^{\ast}\mathcal{H}_{0} are free RR-modules of rank 1 and 2, respectively. We choose basis ρω0=Rf\rho^{\ast}\omega_{0}=Rf, ρ0=Rf+Re\rho^{\ast}\mathcal{H}_{0}=Rf+Re and assume that δ¯=δR\underline{\delta}=\delta R. Moreover, we assume that δfs0\delta f\equiv s_{0} and δpeQ\langle\delta^{p}e\rangle\equiv Q modulo n\mathcal{I}_{n}. Thus, if the derivation dd is dual to the image of the Kodaira-Spencer isomorphism Θ=KS(f,e)ΩR/p1\Theta=KS(f,e)\in\Omega_{R/\operatorname{\mathbb{Z}}_{p}}^{1}, we have that

(17) (d)(δf)=af+δe,for some an.\triangledown(d)(\delta f)=af+\delta e,\quad\mbox{for some }a\in\mathcal{I}_{n}.

Computing things in R[1/δ]R[1/\delta], the setting becomes ordinary. It can be deduced from [15, Main Theorem 4.3.2] that any section ss in the image of the map dlog:Tp𝒢0Ω0{\rm dlog}:T_{p}\mathcal{G}_{0}\rightarrow\Omega_{0} satisfies (d)2(s)=0\triangledown(d)^{2}(s)=0. Since δf\delta f is in the image of dlog{\rm dlog} modulo n\mathcal{I}_{n}, we have δf=s+s0\delta f=s+s_{0}, where sdlog(Tp𝒢0)s\in{\rm dlog}(T_{p}\mathcal{G}_{0}) and s0δannΩ0s_{0}\in\delta^{-a_{n}}\mathcal{I}_{n}\Omega_{0}, for some ana_{n}\in\operatorname{\mathbb{N}}. Since (d)δannΩ0δan1nΩ0\triangledown(d)\delta^{-a_{n}}\mathcal{I}_{n}\Omega_{0}\subseteq\delta^{-a_{n}-1}\mathcal{I}_{n}\Omega_{0}, we deduce that (d)2(δf)δbnn\triangledown(d)^{2}(\delta f)\in\delta^{-b_{n}}\mathcal{I}_{n}, for some bnb_{n}\in\operatorname{\mathbb{N}}. Applying the Leibniz rule, we compute that

(d)(δpe)=(δp3a(dδa)δp2da)δf+δ1((p1)dδa)δpe+δp1(d)2(δf).\triangledown(d)(\delta^{p}e)=(\delta^{p-3}a(d\delta-a)-\delta^{p-2}da)\delta f+\delta^{-1}((p-1)d\delta-a)\delta^{p}e+\delta^{p-1}\triangledown(d)^{2}(\delta f).

Since dapnδ1nda\in p^{n}\delta^{-1}\mathcal{I}_{n}, this implies that, for some cnc_{n}\in\operatorname{\mathbb{N}},

(δcnd)(δf)=Aδf+Bδpe,(δcnD)(δpe)=Cδf+Dδpe,\triangledown(\delta^{c_{n}}d)(\delta f)=A\delta f+B\delta^{p}e,\qquad\triangledown(\delta^{c_{n}}D)(\delta^{p}e)=C\delta f+D\delta^{p}e,

where A,CnA,C\in\mathcal{I}_{n} and B,DRB,D\in R.

Fix a generator αn\alpha\in\mathcal{I}_{n}. As explained in [3, §6.4], the sections ρ𝒪𝕍Q(0,s0)(Spf(R))\rho^{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}(\mathrm{Spf}(R)) are given by

ρ𝒪𝕍(0)=RX,YRZ,T=ρ𝒪𝕍Q(0,s0),X1+αZ,YαT,\rho^{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}(\mathcal{H}_{0}^{\sharp})}=R\langle X,Y\rangle\longrightarrow R\langle Z,T\rangle=\rho^{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})},\qquad X\mapsto 1+\alpha Z,\quad Y\mapsto\alpha T,

where XX corresponds to δf\delta f and YY corresponds to δpe\delta^{p}e. This implies that the connection on ρ𝒪𝕍(0)\rho^{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}(\mathcal{H}_{0}^{\sharp})} is given by (δcnd)(X)=AX+BY\triangledown(\delta^{c_{n}}d)(X)=AX+BY, (δcnd)(Y)=CX+DY\triangledown(\delta^{c_{n}}d)(Y)=CX+DY. We observe that this connection descends to ρ𝒪𝕍Q(0,s0)\rho^{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}, indeed we can define

(δcnd)(Z)=α1A(1+αZ)+BT,(δcnd)(T)=α1C(1+αZ)+DT,\triangledown(\delta^{c_{n}}d)(Z)=\alpha^{-1}A(1+\alpha Z)+BT,\qquad\triangledown(\delta^{c_{n}}d)(T)=\alpha^{-1}C(1+\alpha Z)+DT,

since α1A,α1CR\alpha^{-1}A,\alpha^{-1}C\in R. Thus, by Lemma 4.3, after possibly enlarging cnc_{n} we obtain

:(gnf0)𝒪𝕍Q(0,s0)(gnf0)𝒪𝕍Q(0,s0)𝒪𝔛δ¯cnΩ𝔛/Λn1.\bigtriangledown:(g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}\longrightarrow(g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}\otimes_{\mathcal{O}_{\mathfrak{X}}}\underline{\delta}^{-c_{n}}\Omega^{1}_{\mathfrak{X}/\Lambda_{n}}.

In [3, §6.4] we have also the local description of ρ𝕎0𝐤τ0,n0\rho^{\ast}\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}}: It is given by

ρ𝕎0𝐤τ0,n0=Pn(Z)RT1+αZ,Pn(Z)=i0(wni)αiZi,\rho^{\ast}\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}}=P_{n}(Z)\cdot R\left\langle\frac{T}{1+\alpha Z}\right\rangle,\qquad P_{n}(Z)=\sum_{i\geq 0}\binom{w_{n}}{i}\alpha^{i}Z^{i},

being PnP_{n} the analytic extension of 𝐤τ0,n0{\bf k}_{\tau_{0},n}^{0} around 1+nRZ1+\mathcal{I}_{n}R\langle Z\rangle. We compute using Leibniz rule:

(δcnd)(Pn(Z)(T1+αZ)m)=\displaystyle\triangledown(\delta^{c_{n}}d)\left(P_{n}(Z)\left(\frac{T}{1+\alpha Z}\right)^{m}\right)=
=Pn(Z)(T1+αZ)m1((wnm)αB(T1+αZ)2+((wnm)A+mD)(T1+αZ)+mα1C).\displaystyle=P_{n}(Z)\left(\frac{T}{1+\alpha Z}\right)^{m-1}\left((w_{n}-m)\alpha B\left(\frac{T}{1+\alpha Z}\right)^{2}+((w_{n}-m)A+mD)\left(\frac{T}{1+\alpha Z}\right)+m\alpha^{-1}C\right).

From this we deduce the first assertion.

Moreover, the above computation shows that the induced morphism in graded peaces is given by multiplication by (wnm)αB(w_{n}-m)\alpha B, and recall by (17) that B=δcn+1pB=\delta^{c_{n}+1-p}. Hence, once we invert pp (and therefore δ¯\underline{\delta}) we obtain the second assertion. ∎

Together with the connection on Ω𝐤f[1/p]\Omega^{{\bf k}_{f}}[1/p] provided by the universal derivation on 𝔊r,n\mathfrak{IG}_{r,n}, the above 𝐤τ0,n\bigtriangledown_{{\bf k}_{\tau_{0},n}} induces a connection on 𝕎𝐤τ0,n:=𝕎0𝐤τ0,n0𝒪𝔛r×𝔚τ0,nΩ𝐤f[1/p]\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}}:=\operatorname{\mathbb{W}}_{0}^{{\bf k}^{0}_{\tau_{0},n}}\otimes_{\mathcal{O}_{\mathfrak{X}_{r}\times\mathfrak{W}_{\tau_{0},n}}}\Omega^{{\bf k}_{f}}[1/p] compatible with the identification

Symk(0)[1/p]=k(Filk(𝕎0𝐤τ0,n0)𝒪𝔛r×𝔚τ0,nΩ𝐤f)[1/p],{\rm Sym}^{k}(\mathcal{H}_{0})[1/p]=k^{\ast}\left({\rm Fil}_{k}(\operatorname{\mathbb{W}}_{0}^{{\bf k}^{0}_{\tau_{0},n}})\otimes_{\mathcal{O}_{\mathfrak{X}_{r}\times\mathfrak{W}_{\tau_{0},n}}}\Omega^{{\bf k}_{f}}\right)[1/p],

and the Gauss-Manin connection on Symk(0)[1/p]{\rm Sym}^{k}(\mathcal{H}_{0})[1/p].

5.2. Hecke operator U𝔭0U_{\mathfrak{p}_{0}}

Let us consider the morphisms p1,p2:𝔛r+1𝔛rp_{1},p_{2}:\mathfrak{X}_{r+1}\rightarrow\mathfrak{X}_{r} defined on the universal abelian variety 𝐀𝐀{\bf A}\mapsto{\bf A} and 𝐀𝐀=𝐀/C~1{\bf A}\mapsto{\bf A}^{\prime}={\bf A}/\tilde{C}_{1}, where C~1𝐀[𝔭0]\tilde{C}_{1}\subset{\bf A}[\mathfrak{p}_{0}] is the 𝒪D\mathcal{O}_{D}-submodule generated by the canonical subgroup C1C_{1}. By [1, Proposition 3.24] and [1, Proposition 6.5], the dual isogeny λ:𝐀𝐀\lambda:{\bf A}^{\prime}\rightarrow{\bf A} induces morphisms of 𝒪𝔛\mathcal{O}_{\mathfrak{X}}-modules

𝒰:p2,p1(𝕎0𝐤τ0,n0)p2,p2(𝕎0𝐤τ0,n0),𝒰:p2,p1(𝕎𝐤τ0,n)p2,p2(𝕎𝐤τ0,n),\mathcal{U}:p_{2,\ast}p_{1}^{\ast}(\operatorname{\mathbb{W}}_{0}^{{\bf k}^{0}_{\tau_{0},n}})\longrightarrow p_{2,\ast}p_{2}^{\ast}(\operatorname{\mathbb{W}}_{0}^{{\bf k}^{0}_{\tau_{0},n}}),\qquad\mathcal{U}:p_{2,\ast}p_{1}^{\ast}(\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}})\longrightarrow p_{2,\ast}p_{2}^{\ast}(\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}}),

which commute with 𝐤τ0,n\bigtriangledown_{{\bf k}_{\tau_{0},n}} and preserve filtrations.

Similarly as in [2, Proposition 3.3], the morphism p2:𝔛r+1𝔛rp_{2}:\mathfrak{X}_{r+1}\rightarrow\mathfrak{X}_{r} is finite flat of degree pp hence there is a well defined trace map with respect to 𝒪𝔛rp2,(𝒪𝔛r+1)\mathcal{O}_{\mathfrak{X}_{r}}\rightarrow p_{2,\ast}(\mathcal{O}_{\mathfrak{X}_{r+1}}). We define the U𝔭0U_{\mathfrak{p}_{0}}-operator

U𝔭0:H0(𝔛,𝕎𝐤τ0,n)𝒰p1H0(𝔛,p2,p2(𝕎𝐤τ0,n))1pTrp2H0(𝔛,𝕎𝐤τ0,n)[1/p].U_{\mathfrak{p}_{0}}:H^{0}(\mathfrak{X},\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}})\stackrel{{\scriptstyle\mathcal{U}\circ p_{1}^{\ast}}}{{\longrightarrow}}H^{0}(\mathfrak{X},p_{2,\ast}p_{2}^{\ast}(\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}}))\stackrel{{\scriptstyle\frac{1}{p}{\rm Tr}_{p_{2}}}}{{\longrightarrow}}H^{0}(\mathfrak{X},\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}})[1/p].
Lemma 5.6.

The morphism induced by U𝔭0U_{\mathfrak{p}_{0}} on H0(𝔛,𝕎𝐤τ0,n/Film𝕎𝐤τ0,n)H^{0}(\mathfrak{X},\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}}/{\rm Fil}_{m}\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}}) is nilpotent modulo p[m/2]1p^{[m/2]-1}, for mpm\geq p where [][\cdot] denotes the integral part.

Proof.

Analogously to [1, Proposition 3.24] we will prove that the morphism 𝒰\mathcal{U} induces a morphism on the mmth graded part of the filtration that is zero modulo p[m/p]p^{[m/p]}, this automatically implies the claim.

By [1, Lemma 6.4] the map induced on de Rham cohomology λ:0𝒢0\lambda:\mathcal{H}_{0}^{\sharp}\rightarrow\mathcal{H}_{\mathcal{G}_{0}^{\prime}}^{\sharp} (where 𝒢0\mathcal{G}_{0}^{\prime} is the pp-divisible group associated with 𝐀{\bf A}^{\prime}) gives an isomorphism Ω0Ω0:=Ω𝒢01\Omega_{0}\simeq\Omega_{0}^{\prime}:=\Omega_{\mathcal{G}_{0}^{\prime}}^{1} and identifies 0/Ω0=δ¯pω0\mathcal{H}_{0}^{\sharp}/\Omega_{0}=\underline{\delta}^{p}\omega_{0}^{\vee} with pδ¯1p2𝒢0/Ω0p\underline{\delta}^{1-p^{2}}\mathcal{H}_{\mathcal{G}_{0}^{\prime}}^{\sharp}/\Omega_{0}^{\prime}. Using the description of Grm𝕎0𝐤τ0,n0{\rm Gr}_{m}\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}^{0}} provided in (6) we deduce that the induced morphism on graded parts is zero modulo pmδ¯(1p2)mp^{m}\underline{\delta}^{(1-p^{2})m}. By construction pHdgpr+2p{\rm Hdg}^{-p^{r+2}} is a well defined section of 𝔛r+1\mathfrak{X}_{r+1}, since r1r\geq 1,

(pmδ¯(1p2)m)2=p2mHdg2(p+1)mpm(pHdgp3)mpm𝒪𝔛r+1,\left(p^{m}\underline{\delta}^{(1-p^{2})m}\right)^{2}=p^{2m}{\rm Hdg}^{-2(p+1)m}\subset p^{m}(p{\rm Hdg}^{-p^{3}})^{m}\subset p^{m}\mathcal{O}_{\mathfrak{X}_{r+1}},

hence the result follows. ∎

Remark 5.7.

We note that the above definition of U𝔭0U_{\mathfrak{p}_{0}} fits with the one given in [3, §7.3] acting on H0(𝔛,Ω𝐤n)H^{0}(\mathfrak{X},\Omega^{{\bf k}_{n}}). Moreover, in [3, §7] one can also find the definition of the Hecke operators V𝔭0V_{\mathfrak{p}_{0}} and U𝔭U_{\mathfrak{p}}, for 𝔭𝔭0\mathfrak{p}\neq\mathfrak{p}_{0}, acting on H0(𝔛,Ω𝐤n)H^{0}(\mathfrak{X},\Omega^{{\bf k}_{n}}).

5.3. De Rham cohomology with coefficients in 𝕎0𝐤τ0,n0\operatorname{\mathbb{W}}_{0}^{{\bf k}^{0}_{\tau_{0},n}} and the overconvergent projection

The multiplicative law on (gnf0)𝒪𝕍0(0,s0)(g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{H}_{0}^{\sharp},s_{0})} provides a morphism

Fil2𝕎2×𝕎𝐤τ0,n𝕎𝐤τ0,n+2.{\rm Fil}_{2}\operatorname{\mathbb{W}}^{2}\times\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}}\longrightarrow\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}+2}.

Moreover, by the Kodaira-Spenser isomorphism we can identify Ω𝒳r1\Omega_{\mathcal{X}_{r}}^{1} with Fil2𝕎2{\rm Fil}_{2}\operatorname{\mathbb{W}}^{2} in the adic space 𝒳r\mathcal{X}_{r}. Hence, the connection 𝐤τ0,n\bigtriangledown_{{\bf k}_{\tau_{0},n}} can be reinterpreted as the morphism of de Rham complexes of sheaves on the adic space 𝒳r×𝒲n\mathcal{X}_{r}\times\mathcal{W}_{n}

𝕎𝐤τ0,n:𝕎𝐤τ0,n𝐤τ0,n𝕎𝐤τ0,n+2;Film(𝕎𝐤τ0,n):Film(𝕎𝐤τ0,n)𝐤τ0,nFilm+1(𝕎𝐤τ0,n+2).\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet}:\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}}\stackrel{{\scriptstyle\bigtriangledown_{{\bf k}_{\tau_{0},n}}}}{{\longrightarrow}}\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}+2};\qquad{\rm Fil}_{m}(\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet}):{\rm Fil}_{m}(\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}})\stackrel{{\scriptstyle\bigtriangledown_{{\bf k}_{\tau_{0},n}}}}{{\longrightarrow}}{\rm Fil}_{m+1}(\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}+2}).

Let us denote by HdRi(𝒳r×𝒲n,)H_{dR}^{i}(\mathcal{X}_{r}\times\mathcal{W}_{n},\bullet) the ii-th hypercohomology group of such de Rham complexes.

The same identification as above allows us to consider the Gauss-Manin operator given by Griffith’s transversality

m:ω0kmSymm0ω0km+1Symm+10.\bigtriangledown_{m}:\omega_{0}^{k-m}\otimes\mathrm{Sym}^{m}\mathcal{H}_{0}\longrightarrow\omega_{0}^{k-m+1}\otimes\mathrm{Sym}^{m+1}\mathcal{H}_{0}.

Notice that the sheaves ω0kmSymm0\omega_{0}^{k-m}\otimes\mathrm{Sym}^{m}\mathcal{H}_{0} define a filtration Film{\rm Fil}_{m} of Symk0\mathrm{Sym}^{k}\mathcal{H}_{0}. Moreover, on graded pieces the connection m\bigtriangledown_{m} defines an isomorphism times kmk-m. This implies that

k(Symk0)FilkSymk+20=Filk1Symk+20+Imk1==Fil0Symk+20+Imk1.\bigtriangledown_{k}(\mathrm{Sym}^{k}\mathcal{H}_{0})\subseteq{\rm Fil}_{k}\mathrm{Sym}^{k+2}\mathcal{H}_{0}={\rm Fil}_{k-1}\mathrm{Sym}^{k+2}\mathcal{H}_{0}+{\rm Im}\bigtriangledown_{k-1}=\cdots={\rm Fil}_{0}\mathrm{Sym}^{k+2}\mathcal{H}_{0}+{\rm Im}\bigtriangledown_{k-1}.

Since Grm(Symk0)ω0k2m{\rm Gr}_{m}(\mathrm{Sym}^{k}\mathcal{H}_{0})\simeq\omega_{0}^{k-2m}, we obtain that k\bigtriangledown_{k} defines a morphism

θk+1:H0(𝒳r,ω0k)H0(𝒳r,ω0k+2).\theta^{k+1}:H^{0}(\mathcal{X}_{r},\omega_{0}^{-k})\longrightarrow H^{0}(\mathcal{X}_{r},\omega_{0}^{k+2}).

As shown in [3, §8.2], the classical computations on qq-expansions also apply on this setting considering Serre-Tate coordinates. This implies that the exactly same proof of [1, Lemma 3.32] works showing that HdR1(𝒳r×𝒲n,Film(𝕎𝐤τ0,n))H_{dR}^{1}(\mathcal{X}_{r}\times\mathcal{W}_{n},{\rm Fil}_{m}(\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet})) lies in the exact sequence

(18) 0H0(𝒳r×𝒲n,Ω0𝐤τ0,n+2)HdR1(𝒳r×𝒲n,Film(𝕎𝐤τ0,n))i=0,ρmH0(𝒳r×𝒲n,jρ,(ω0)i)00\longrightarrow H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},\Omega_{0}^{{\bf k}_{\tau_{0},n}+2})\longrightarrow H_{dR}^{1}(\mathcal{X}_{r}\times\mathcal{W}_{n},{\rm Fil}_{m}(\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet}))\longrightarrow\bigoplus_{i=0,\rho}^{m}H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},j_{\rho,\ast}(\omega_{0})^{-i})\longrightarrow 0

where jρj_{\rho} is the closed immersion 𝒳r𝒳r×𝒲n\mathcal{X}_{r}\hookrightarrow\mathcal{X}_{r}\times\mathcal{W}_{n} defined by a point such that wn=iw_{n}=i. Moreover, the torsion free part HdR1(𝒳r×𝒲n,Film(𝕎𝐤τ0,n))tfH_{dR}^{1}(\mathcal{X}_{r}\times\mathcal{W}_{n},{\rm Fil}_{m}(\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet}))^{tf} lies in the exact sequence

0H0(𝒳r×𝒲n,Ω0𝐤τ0,n+2)HdR1(𝒳r×𝒲n,Film(𝕎𝐤τ0,n))tfi=0,ρmθi+1(H0(𝒳r×𝒲n,jρ,(ω0)i))00\longrightarrow H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},\Omega_{0}^{{\bf k}_{\tau_{0},n}+2})\longrightarrow H_{dR}^{1}(\mathcal{X}_{r}\times\mathcal{W}_{n},{\rm Fil}_{m}(\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet}))^{tf}\longrightarrow\bigoplus_{i=0,\rho}^{m}\theta^{i+1}\left(H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},j_{\rho,\ast}(\omega_{0})^{-i})\right)\longrightarrow 0

where the morphisms are equivariant for the action of U𝔭0U_{\mathfrak{p}_{0}} and the action of U𝔭0U_{\mathfrak{p}_{0}} on H0(𝒳r×𝒲n,jρ,(ω0)i)H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},j_{\rho,\ast}(\omega_{0})^{-i}) is divided by pi+1p^{i+1}.

Let ImI_{m} be the ideal i=0m(wni)\prod_{i=0}^{m}(w_{n}-i) and ΛIm\Lambda_{I_{m}} the localization of Λτ0,n\Lambda_{\tau_{0},n} at ImI_{m}. The above exact sequence (18) provides an isomorphism

(19) Hm:HdR1(𝒳r×𝒲n,Film(𝕎𝐤τ0,n))Λτ0,nΛImH0(𝒳r×𝒲n,Ω0𝐤τ0,n+2)Λτ0,nΛIm.H_{m}^{\dagger}:H_{dR}^{1}(\mathcal{X}_{r}\times\mathcal{W}_{n},{\rm Fil}_{m}(\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet}))\otimes_{\Lambda_{\tau_{0},n}}\Lambda_{I_{m}}\longrightarrow H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},\Omega_{0}^{{\bf k}_{\tau_{0},n}+2})\otimes_{\Lambda_{\tau_{0},n}}\Lambda_{I_{m}}.

Since Film𝕎𝐤n{\rm Fil}_{m}\operatorname{\mathbb{W}}^{{\bf k}_{n}} is coherent and U𝔭0U_{\mathfrak{p}_{0}} is compact, the usual discussion on slope decompositions applies to H0(𝒳r×𝒲n,Film𝕎𝐤n)H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},{\rm Fil}_{m}\operatorname{\mathbb{W}}^{{\bf k}_{n}}). Hence, given a finite slope h0h\geq 0 we have, locally on the weight space, a slope hh decomposition. This implies that HdRi(𝒳r×𝒲n,Film(𝕎𝐤τ0,n))H_{dR}^{i}(\mathcal{X}_{r}\times\mathcal{W}_{n},{\rm Fil}_{m}(\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet})) admits a slope hh decomposition. Due to Lemma 5.6, the operator U𝔭0U_{\mathfrak{p}_{0}} on HdRi(𝒳r×𝒲n,(𝕎/Film(𝕎)))H_{dR}^{i}(\mathcal{X}_{r}\times\mathcal{W}_{n},\left(\operatorname{\mathbb{W}}/{\rm Fil}_{m}(\operatorname{\mathbb{W}})\right)^{\bullet}) is divisible by ph+1p^{h+1} for large enough mm, where (𝕎/Film(𝕎))\left(\operatorname{\mathbb{W}}/{\rm Fil}_{m}(\operatorname{\mathbb{W}})\right)^{\bullet} is the quotient complex of 𝕎𝐤τ0,n\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet} and Film(𝕎𝐤τ0,n){\rm Fil}_{m}(\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet}). By the long exact sequence of de Rham complexes this implies that

(20) HdRi(𝒳r×𝒲n,Film(𝕎𝐤τ0,n))hHdRi(𝒳r×𝒲n,𝕎𝐤τ0,n)h.H_{dR}^{i}(\mathcal{X}_{r}\times\mathcal{W}_{n},{\rm Fil}_{m}(\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}})^{\bullet})^{\leq h}\simeq H_{dR}^{i}(\mathcal{X}_{r}\times\mathcal{W}_{n},\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet})^{\leq h}.

The above isomorphism together with (18) provides a morphism

(21) H:HdR1(𝒳r×𝒲n,𝕎𝐤τ0,n)hΛτ0,nΛImH0(𝒳r×𝒲n,Ω0𝐤τ0,n+2)hΛτ0,nΛIm,H^{\dagger}:H_{dR}^{1}(\mathcal{X}_{r}\times\mathcal{W}_{n},\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}}^{\bullet})^{\leq h}\otimes_{\Lambda_{\tau_{0},n}}\Lambda_{I_{m}}\longrightarrow H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},\Omega_{0}^{{\bf k}_{\tau_{0},n}+2})^{\leq h}\otimes_{\Lambda_{\tau_{0},n}}\Lambda_{I_{m}},

where mm is an natural number depending on hh.

The spectral theory developed in [2, Appendice B] provides a overconvergent projection ehe_{\leq h} onto slope h\leq h subspace, this gives rise to the composition:

(22) H0(𝒳r×𝒲n,𝕎𝐤τ0,n)Λτ0,nΛIm\textstyle{H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}})\otimes_{\Lambda_{\tau_{0},n}}\Lambda_{I_{m}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}H,h\scriptstyle{H^{\dagger,\leq h}}H0(𝒳r×𝒲n,Ω0𝐤τ0,n)hΛτ0,nΛIm\textstyle{H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},\Omega_{0}^{{\bf k}_{\tau_{0},n}})^{\leq h}\otimes_{\Lambda_{\tau_{0},n}}\Lambda_{I_{m}}}HdR1(𝒳r×𝒲n,𝕎𝐤τ0,n2)Λτ0,nΛIm\textstyle{H_{dR}^{1}(\mathcal{X}_{r}\times\mathcal{W}_{n},\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}-2}^{\bullet})\otimes_{\Lambda_{\tau_{0},n}}\Lambda_{I_{m}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}eh\scriptstyle{e_{\leq h}}HdR1(𝒳r×𝒲n,𝕎𝐤τ0,n2)hΛτ0,nΛIm\textstyle{H_{dR}^{1}(\mathcal{X}_{r}\times\mathcal{W}_{n},\operatorname{\mathbb{W}}_{{\bf k}_{\tau_{0},n}-2}^{\bullet})^{\leq h}\otimes_{\Lambda_{\tau_{0},n}}\Lambda_{I_{m}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}H\scriptstyle{H^{\dagger}}

where the left down arrow is the natural projection.

6. Triple products and pp-adic L-functions

6.1. Triple products at τ0\tau_{0}

In [3, Theorem 4.11], it is shown that there exists a morphism

tk1,k2,k3:H0(𝒳r,ω0k1)×H0(𝒳r,ω0k2)\displaystyle t_{k_{1},k_{2},k_{3}}:H^{0}(\mathcal{X}_{r},\omega_{0}^{k_{1}})\times H^{0}(\mathcal{X}_{r},\omega_{0}^{k_{2}}) \displaystyle\longrightarrow H0(𝒳r,ω0k3);\displaystyle H^{0}(\mathcal{X}_{r},\omega_{0}^{k_{3}});
tk1,k2,k3(s1,s2)\displaystyle t_{k_{1},k_{2},k_{3}}(s_{1},s_{2}) :=\displaystyle:= j=0m(1)j(mj)(M2k1+j1)k1j(s1)k2mj(s2),\displaystyle\sum_{j=0}^{m}(-1)^{j}\binom{m}{j}\binom{M-2}{k_{1}+j-1}\bigtriangledown_{k_{1}}^{j}(s_{1})\bigtriangledown^{m-j}_{k_{2}}(s_{2}),

where k1+k2+k3k_{1}+k_{2}+k_{3} is even, k3k1+k2k_{3}\geq k_{1}+k_{2}, M=k1+k2+k32M=\frac{k_{1}+k_{2}+k_{3}}{2}, and m=k3k1k22m=\frac{k_{3}-k_{1}-k_{2}}{2}. The aim of this section is to define a triple product

𝐭:H0(𝒳r×𝒲n1,Ω0𝐤τ0,n1)×H0(𝒳r×𝒲n2,Ω0𝐤τ0,n2)H0(𝒳b×𝒲n1×𝒲n2×𝒲n3,Ω0𝐤τ0,n3)h,{\bf t}:H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n_{1}},\Omega_{0}^{{\bf k}_{\tau_{0},n_{1}}})\times H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n_{2}},\Omega_{0}^{{\bf k}_{\tau_{0},n_{2}}})\longrightarrow H^{0}(\mathcal{X}_{b}\times\mathcal{W}_{n_{1}}\times\mathcal{W}_{n_{2}}\times\mathcal{W}_{n_{3}},\Omega_{0}^{{\bf k}^{\ast}_{\tau_{0},n_{3}}})^{\leq h},

where 𝐤τ0,n3:p×Λτ0,n1^Λτ0,n2^Λ^τ0,n3{\bf k}^{\ast}_{\tau_{0},n_{3}}:\operatorname{\mathbb{Z}}_{p}^{\times}\rightarrow\Lambda_{\tau_{0},n_{1}}\hat{\otimes}\Lambda_{\tau_{0},n_{2}}\hat{\otimes}\hat{\Lambda}_{\tau_{0},n_{3}} is given by 𝐤τ0,n3=𝐤τ0,n1𝐤τ0,n2𝐤τ0,n32{\bf k}^{\ast}_{\tau_{0},n_{3}}={\bf k}_{\tau_{0},n_{1}}\otimes{\bf k}_{\tau_{0},n_{2}}\otimes{\bf k}^{2}_{\tau_{0},n_{3}}, and b>rb>r is an integer depending on rmax{n1,n2,n3}r\geq{\rm max}\{n_{1},n_{2},n_{3}\}. Such triple product 𝐭{\bf t} when evaluated at (k1,k2,m)(3)(𝒲n1×𝒲n2×𝒲n3)(p)(k_{1},k_{2},m)\in(\operatorname{\mathbb{N}}^{3})\cap(\mathcal{W}_{n_{1}}\times\mathcal{W}_{n_{2}}\times\mathcal{W}_{n_{3}})(\operatorname{\mathbb{Q}}_{p}) must coincide with tk1,k2,k3t_{k_{1},k_{2},k_{3}} up to constant, where k3=k1+k2+2mk_{3}=k_{1}+k_{2}+2m.

Given s1H0(𝒳r×𝒲n1,Ω0𝐤τ0,n1)s_{1}\in H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n_{1}},\Omega_{0}^{{\bf k}_{\tau_{0},n_{1}}}), s2H0(𝒳r×𝒲n2,Ω0𝐤τ0,n2)s_{2}\in H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n_{2}},\Omega_{0}^{{\bf k}_{\tau_{0},n_{2}}}), and mm\in\operatorname{\mathbb{N}} we can define using the multiplicative structure of (gnf0)𝒪𝕍0(0,s0)(g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{H}_{0}^{\sharp},s_{0})}

𝐤τ0,n1m(s1)s2H0(𝒳r×𝒲n1,n2,Film𝕎𝐤τ0,n1+𝐤τ0,n2+2m),𝒲n1,n2:=𝒲n1×𝒲n2.\bigtriangledown_{{\bf k}_{\tau_{0},n_{1}}}^{m}(s_{1})\cdot s_{2}\in H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n_{1},n_{2}},{\rm Fil}_{m}\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n_{1}}+{\bf k}_{\tau_{0},n_{2}}+2m}),\qquad\mathcal{W}_{n_{1},n_{2}}:=\mathcal{W}_{n_{1}}\times\mathcal{W}_{n_{2}}.

Write 𝐤n1,n2,m:=𝐤τ0,n1+𝐤τ0,n2+2m{\bf k}_{n_{1},n_{2},m}:={\bf k}_{\tau_{0},n_{1}}+{\bf k}_{\tau_{0},n_{2}}+2m, Λn1,n2=Λτ0,n1^Λτ0,n2\Lambda_{n_{1},n_{2}}=\Lambda_{\tau_{0},n_{1}}\hat{\otimes}\Lambda_{\tau_{0},n_{2}} and

Θm(s1,s2)H0(𝒳r×𝒲n1,n2,Ω0𝐤n1,n2,m)Λn1,n2ΛIm\Theta_{m}(s_{1},s_{2})\in H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n_{1},n_{2}},\Omega_{0}^{{\bf k}_{n_{1},n_{2},m}})\otimes_{\Lambda_{n_{1},n_{2}}}\Lambda_{I_{m}}

for the image of 𝐤τ0,n1m(s1)s2\bigtriangledown_{{\bf k}_{\tau_{0},n_{1}}}^{m}(s_{1})\cdot s_{2} under the composition of the natural projection

H0(𝒳r×𝒲n1,n2,Film𝕎𝐤n1,n2,m)HdR1(𝒳r×𝒲n1,n2,Film1(𝕎𝐤n1,n2,m2))Λn1,n2ΛImH^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n_{1},n_{2}},{\rm Fil}_{m}\operatorname{\mathbb{W}}^{{\bf k}_{n_{1},n_{2},m}})\longrightarrow H_{dR}^{1}(\mathcal{X}_{r}\times\mathcal{W}_{n_{1},n_{2}},{\rm Fil}_{m-1}(\operatorname{\mathbb{W}}_{{\bf k}_{n_{1},n_{2},m}-2}^{\bullet}))\otimes_{\Lambda_{n_{1},n_{2}}}\Lambda_{I_{m}}

and the morphism Hm1H_{m-1}^{\dagger} of (19).

Lemma 6.1.

if ρ=(k1,k2)2Λn1,n2()\rho=(k_{1},k_{2})\in\operatorname{\mathbb{N}}^{2}\cap\Lambda_{n_{1},n_{2}}(\operatorname{\mathbb{Q}}) we have that

ρ(Θm(s1,s2))=(1)m(k32m+k21)1tk1,k2,k3(ρ(s1),ρ(s2)),\rho^{\ast}(\Theta_{m}(s_{1},s_{2}))=(-1)^{m}\binom{k_{3}-2}{m+k_{2}-1}^{-1}t_{k_{1},k_{2},k_{3}}(\rho^{\ast}(s_{1}),\rho^{\ast}(s_{2})),

where k3=k1+k2+2mk_{3}=k_{1}+k_{2}+2m.

Proof.

A laborious but straightforward computation shows that

ρ(𝐤τ0,n1m(s1)s2)=k1m(ρ(s1))ρ(s2)=\displaystyle\rho^{\ast}\left(\bigtriangledown_{{\bf k}_{\tau_{0},n_{1}}}^{m}(s_{1})\cdot s_{2}\right)=\bigtriangledown_{k_{1}}^{m}(\rho^{\ast}(s_{1}))\cdot\rho^{\ast}(s_{2})=
=(1)m(k32m+k21)1tk1,k2,k3(ρ(s1),ρ(s2))+(i=0m1aik1i(ρ(s1))k2m1i(ρ(s2))),\displaystyle=(-1)^{m}\binom{k_{3}-2}{m+k_{2}-1}^{-1}t_{k_{1},k_{2},k_{3}}(\rho^{\ast}(s_{1}),\rho^{\ast}(s_{2}))+\bigtriangledown\left(\sum_{i=0}^{m-1}a_{i}\bigtriangledown_{k_{1}}^{i}(\rho^{\ast}(s_{1}))\bigtriangledown^{m-1-i}_{k_{2}}(\rho^{\ast}(s_{2}))\right),

where

ai=(1)i+m+1(k32m+k21)1(j=0i(mj)(M2k1+j1)).a_{i}=(-1)^{i+m+1}\binom{k_{3}-2}{m+k_{2}-1}^{-1}\left(\sum_{j=0}^{i}\binom{m}{j}\binom{M-2}{k_{1}+j-1}\right).

This implies that

(1)m(k32m+k21)1tk1,k2,k3(ρ(s1),ρ(s2))=\displaystyle(-1)^{m}\binom{k_{3}-2}{m+k_{2}-1}^{-1}t_{k_{1},k_{2},k_{3}}(\rho^{\ast}(s_{1}),\rho^{\ast}(s_{2}))=
ρ(𝐤τ0,n1m(s1)s2𝐤n1,n2,m(i=0m1ai𝐤τ0,n1i(s1)𝐤τ0,n2m1i(s2))).\displaystyle\rho^{\ast}\left(\bigtriangledown_{{\bf k}_{\tau_{0},n_{1}}}^{m}(s_{1})\cdot s_{2}-\bigtriangledown_{{\bf k}_{n_{1},n_{2},m}}\left(\sum_{i=0}^{m-1}a_{i}\bigtriangledown_{{\bf k}_{\tau_{0},n_{1}}}^{i}(s_{1})\bigtriangledown^{m-1-i}_{{\bf k}_{\tau_{0},n_{2}}}(s_{2})\right)\right).

Hence the result follows from the fact that the class of 𝐤τ0,n1m(s1)s2\bigtriangledown_{{\bf k}_{\tau_{0},n_{1}}}^{m}(s_{1})\cdot s_{2} in the de Rham cohomology group coincides with the class of 𝐤τ0,n1m(s1)s2𝐤n1,n2,m(i=0m1ai𝐤τ0,n1i(s1)𝐤τ0,n2m1i(s2))\bigtriangledown_{{\bf k}_{\tau_{0},n_{1}}}^{m}(s_{1})\cdot s_{2}-\bigtriangledown_{{\bf k}_{n_{1},n_{2},m}}\left(\sum_{i=0}^{m-1}a_{i}\bigtriangledown_{{\bf k}_{\tau_{0},n_{1}}}^{i}(s_{1})\bigtriangledown^{m-1-i}_{{\bf k}_{\tau_{0},n_{2}}}(s_{2})\right). ∎

6.2. pp-adic interpolation of the operator Θm\Theta_{m}

We aim to construct the operator Θ𝐤τ0,n3\Theta_{{\bf k}_{\tau_{0},n_{3}}} for the universal character 𝐤τ0,n3:p×Λn3{\bf k}_{\tau_{0},n_{3}}:\operatorname{\mathbb{Z}}_{p}^{\times}\longrightarrow\Lambda_{n_{3}}. A strategy one can think of is to pp-adically iterate the operators Θm\Theta_{m} as mm varies. Following this strategy, the problem that one immediately finds is that as mm grows the ideal ImI_{m} grows as well. Since we have to invert ImI_{m} in order to define HmH_{m}^{\dagger}, we lose control of the denominators.

One can think of the strategy of defining Θ𝐤τ0,n3\Theta_{{\bf k}_{\tau_{0},n_{3}}} in the space of pp-adic modular forms, namely, sections on the ordinary locus. This is the strategy used in [3]. By equation (20), we have to restrict ourselves to the case of slope h\leq h in order to control the filtration where Θ𝐤τ0,n3\Theta_{{\bf k}_{\tau_{0},n_{3}}} lives. Using the interpolation of the Serre operator introduced in [3], one could define Θ𝐤τ0,n3\Theta_{{\bf k}_{\tau_{0},n_{3}}} provided by the existence of a slope h\leq h projector ehe_{\leq h} on the space of pp-adic modular forms. But there is no such projector acting on the space of pp-adic modular forms except for the ordinary case.

The strategy followed by Andreatta and Iovita in [1] relies on p-adically interpolate the operator 𝐤τ0,nm\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{m} acting on 𝕎:=k𝕎𝐤τ0,n+2k(gnf0)𝒪𝕍0(0,s0)g1,𝒪𝔊1^Λn\operatorname{\mathbb{W}}^{\prime}:=\bigoplus_{k}\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}+2k}\subset(g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{0}(\mathcal{H}_{0}^{\sharp},s_{0})}\otimes g_{1,\ast}\mathcal{O}_{\mathfrak{IG}_{1}}\hat{\otimes}\Lambda_{n}. Since on the space of nearly overconvergent modular forms we have the projector ehe_{\leq h}, we have the operator H,hH^{\dagger,\leq h} of the diagram (22), hence we can proceed with a well defined construction of Θ𝐤τ0,n3\Theta_{{\bf k}_{\tau_{0},n_{3}}} on the space of nearly overconvergent modular forms. In order to ensure the convergence of the series involved in the definition of 𝐤τ0,n𝐬\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\bf s}, for a universal character 𝐬{\bf s}, one has to control de valuation of the corresponding terms, and this can be done by computing the Gauss-Manin connection on the ordinary locus. There we can use Serre-Tate coordinates and the corresponding qq-expansions.

6.3. Twists by finite characters

This section is analogous to [1, §3.8]. As always rnr\geq n and let 𝒢0\mathcal{G}_{0} be the universal formal group over 𝒢2n,r+n\mathcal{IG}_{2n,r+n}. Let Cn𝒢0[pn]C_{n}\subset\mathcal{G}_{0}[p^{n}] be the canonical subgroup and let us consider

π:𝒢0𝒢0:=𝒢0/CnCn:=C2n/Cn.\pi:\mathcal{G}_{0}\longrightarrow\mathcal{G}_{0}^{\prime}:=\mathcal{G}_{0}/C_{n}\qquad C_{n}^{\prime}:=C_{2n}/C_{n}.

It is clear that CnC_{n}^{\prime} is the canonical subgroup of 𝒢0\mathcal{G}_{0}^{\prime} and the dual isogeny λ:𝒢0𝒢0\lambda:\mathcal{G}_{0}^{\prime}\rightarrow\mathcal{G}_{0} maps CnC_{n}^{\prime} to CnC_{n}. This construction defines a morphism t:𝒢2n,r+n𝒢n,rt:\mathcal{IG}_{2n,r+n}\rightarrow\mathcal{IG}_{n,r}, since a trivialization of D2nD_{2n} induces by multiplication by pnp^{n} a trivialization of Dn=(Cn)D_{n}^{\prime}=(C_{n}^{\prime})^{\vee}.

If Cn′′:=kerλC_{n}^{\prime\prime}:=\ker{\lambda}, we have that 𝒢0[pn]=Cn×Cn′′\mathcal{G}_{0}[p^{n}]=C_{n}^{\prime}\times C_{n}^{\prime\prime} as group schemes over 𝒢2n,r+n\mathcal{IG}_{2n,r+n}. The universal trivialization provides isomorphisms

s:/pnCn′′,s:Cnμpn,s:\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}}\longrightarrow C_{n}^{\prime\prime},\qquad s^{\vee}:C_{n}^{\prime}\longrightarrow\mu_{p^{n}},

since the Weil pairing and the polarization identifies Cn(Cn′′)C_{n}^{\prime}\simeq(C_{n}^{\prime\prime})^{\vee}.

A choice of a primitive pnp^{n}-root of unity ξ\xi provides an identification /pnHom(/pn,μpn)\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}}\simeq{\rm Hom}(\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}},\mu_{p^{n}}), j(1ξj)j\mapsto(1\mapsto\xi^{j}). Thus, we obtain a bijection

Hom(Cn′′,Cn)Hom(/pn,μpn)/pngsgs.{\rm Hom}(C_{n}^{\prime\prime},C_{n}^{\prime})\longrightarrow{\rm Hom}(\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}},\mu_{p^{n}})\simeq\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}}\,\qquad g\longmapsto s^{\vee}\circ g\circ s.

Write ρjHom(Cn′′,Cn)\rho_{j}\in{\rm Hom}(C_{n}^{\prime\prime},C_{n}^{\prime}) for the morphism corresponding to j/pnj\in\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}}, and let us consider tj:𝒢2n,r+n𝒢n,rt_{j}:\mathcal{IG}_{2n,r+n}\rightarrow\mathcal{IG}_{n,r} for the morphism induced by making quotient by (ρj×Id)(Cn′′)𝒢0[pn](\rho_{j}\times{\rm Id})(C_{n}^{\prime\prime})\subset\mathcal{G}_{0}^{\prime}[p^{n}], with the induced trivialization of (Cn,j)(C_{n,j})^{\vee}, where Cn,jC_{n,j} is the image of CnC_{n}^{\prime} in 𝒢0j:=𝒢0/(ρj×Id)(Cn′′)\mathcal{G}_{0}^{j}:=\mathcal{G}_{0}/(\rho_{j}\times{\rm Id})(C_{n}^{\prime\prime}) which coincides with the canonical subgroup. Write also tj:𝔊2n,r+n𝔊n,rt_{j}:\mathfrak{IG}_{2n,r+n}\rightarrow\mathfrak{IG}_{n,r} for the corresponding morphism taking normalizations. Over 𝔊2n,r+n\mathfrak{IG}_{2n,r+n} we have isogenies on the universal formal groups

𝒢0λ𝒢0λj𝒢0j\mathcal{G}_{0}\stackrel{{\scriptstyle\lambda}}{{\longleftarrow}}\mathcal{G}_{0}^{\prime}\stackrel{{\scriptstyle\lambda_{j}}}{{\longrightarrow}}\mathcal{G}_{0}^{j}

respecting canonical groups and trivializations. By functoriality, this gives rise to the corresponding maps

0λ(0)λj(0j).\mathcal{H}_{0}^{\sharp}\stackrel{{\scriptstyle\lambda^{\sharp}}}{{\longleftarrow}}(\mathcal{H}_{0}^{\prime})^{\sharp}\stackrel{{\scriptstyle\lambda_{j}^{\sharp}}}{{\longrightarrow}}(\mathcal{H}_{0}^{j})^{\sharp}.

By [1, Lemma 6.4] the images coincide Imλ=Imλj{\rm Im}\lambda^{\sharp}={\rm Im}\lambda_{j}^{\sharp}, hence it gives rise to a isomorphism fj:(0j)0f_{j}:(\mathcal{H}_{0}^{j})^{\sharp}\rightarrow\mathcal{H}_{0}^{\sharp} preserving Hodge exact sequences and Gauss-Manin connections. Again by functoriality, this gives rise to a morphism over 𝔊2n,r+n\mathfrak{IG}_{2n,r+n}

fj:tj(f0)𝒪𝕍Q(0,s0)(f0)𝒪𝕍Q(0,s0)f_{j}^{\ast}:t_{j}^{\ast}(f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}\longrightarrow(f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}

preserving filfrations and Gauss-Manin connections. Thus, after extending scalars in order to have the root ξ\xi, we obtain a morphism for any finite character χ:(/pn)ׯp×\chi:(\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}})^{\times}\rightarrow\bar{\operatorname{\mathbb{Q}}}_{p}^{\times}

θχ=gχ11jχ(j)1fjtj:H0(𝔛r,(gnf0)𝒪𝕍Q(0,s0))H0(𝔛r+n,(g2nf0)𝒪𝕍Q(0,s0)),\theta_{\chi}=g_{\chi^{-1}}^{-1}\sum_{j}\chi(j)^{-1}f_{j}^{\ast}\circ t_{j}^{\ast}:H^{0}\left(\mathfrak{X}_{r},(g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}\right)\longrightarrow H^{0}\left(\mathfrak{X}_{r+n},(g_{2n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}\right),

where gχ1:=nχ(n)1ξng_{\chi^{-1}}:=\sum_{n}\chi(n)^{-1}\xi^{n} is a Gauss sum.

The following result is completely analogous to [1, Lemma 3.31]:

Lemma 6.2.

The morphism θχ\theta_{\chi} provides a morphism

𝐤τ0,nχ:H0(𝔛r×𝔚n,𝕎𝐤τ0,n)H0(𝔛r+n×𝔚n,𝕎𝐤τ0,n+2χ).\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\chi}:H^{0}\left(\mathfrak{X}_{r}\times\mathfrak{W}_{n},\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}}\right)\longrightarrow H^{0}\left(\mathfrak{X}_{r+n}\times\mathfrak{W}_{n},\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}+2\chi}\right).

6.4. qq-expansions on unitary Shimura curves

Over the ordinary locus 𝔛ord\mathfrak{X}_{\rm ord} it is constructed in [3] an affine Spf(W)\mathrm{Spf}(W)-formal scheme 𝔛()\mathfrak{X}(\infty), where WW is the ring of Witt vectors of F¯p\bar{F}_{p}. 𝔛()\mathfrak{X}(\infty) represents ordinary abelian varieties (A,ι,θ,α𝔭0)(A,\iota,\theta,\alpha^{\mathfrak{p}_{0}}) together with frames:

(23) 0𝔾m[p]ıA[p],1πp/p×𝔭𝔭0(F𝔭/𝒪𝔭)20.0\longrightarrow\operatorname{\mathbb{G}}_{m}[p^{\infty}]\stackrel{{\scriptstyle\imath}}{{\longrightarrow}}A[p^{\infty}]^{-,1}\stackrel{{\scriptstyle\pi}}{{\longrightarrow}}\operatorname{\mathbb{Q}}_{p}/\operatorname{\mathbb{Z}}_{p}\times\prod_{\mathfrak{p}\neq\mathfrak{p}_{0}}(F_{\mathfrak{p}}/\mathcal{O}_{\mathfrak{p}})^{2}\longrightarrow 0.

Moreover, 𝔛()\mathfrak{X}(\infty) is equipped with a classifying morphism c:𝔛()/W𝔾m/Wc:\mathfrak{X}(\infty)/W\rightarrow\operatorname{\mathbb{G}}_{m}/W. The local coordinates given by the morphism cc are called Serre-Tate coordinates.

As explained in [3, §8.2], given a point P𝔛()P\in\mathfrak{X}(\infty) providing a frame

(24) 0𝔾m[p]ı𝒢0[p]πp/p00\longrightarrow\operatorname{\mathbb{G}}_{m}[p^{\infty}]\stackrel{{\scriptstyle\imath}}{{\longrightarrow}}\mathcal{G}_{0}[p^{\infty}]\stackrel{{\scriptstyle\pi}}{{\longrightarrow}}\operatorname{\mathbb{Q}}_{p}/\operatorname{\mathbb{Z}}_{p}\longrightarrow 0

by restriction of (23), its image is c(P)=q𝔾mc(P)=q\in\operatorname{\mathbb{G}}_{m} such that

𝒢0Eq:=(𝔾m,R[p]p)/((1+q)m,m),mp,0𝔾m,R[p]ıEqπqp/p0,\mathcal{G}_{0}\simeq E_{q}:=(\operatorname{\mathbb{G}}_{m,R}[p^{\infty}]\oplus\operatorname{\mathbb{Q}}_{p})/\langle((1+q)^{m},-m),\;m\in\operatorname{\mathbb{Z}}_{p}\rangle,\qquad 0\longrightarrow\operatorname{\mathbb{G}}_{m,R}[p^{\infty}]\stackrel{{\scriptstyle\imath}}{{\longrightarrow}}E_{q}\stackrel{{\scriptstyle\pi_{q}}}{{\longrightarrow}}\operatorname{\mathbb{Q}}_{p}/\operatorname{\mathbb{Z}}_{p}\longrightarrow 0,

where ı(a)=(a,0)\imath(a)=(a,0) and πq(a,b)=b\pi_{q}(a,b)=b. Moreover, we have that Cn={(ξn,0),n/pn}C_{n}=\{(\xi^{n},0),\;n\in\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}}\}. This implies that

𝒢0=E(q+1)pn1:=(𝔾m,R[p]p)/((1+q)pnm,m),mp,\mathcal{G}_{0}^{\prime}=E_{(q+1)^{p^{n}}-1}:=(\operatorname{\mathbb{G}}_{m,R}[p^{\infty}]\oplus\operatorname{\mathbb{Q}}_{p})/\langle((1+q)^{p^{n}m},-m),\;m\in\operatorname{\mathbb{Z}}_{p}\rangle,

and the morphism λ:𝒢0𝒢0\lambda:\mathcal{G}_{0}^{\prime}\rightarrow\mathcal{G}_{0} is given by the morphism

λ:E(q+1)pn1Eq;λ(a,x)=(a,pnx),Cn′′=ker(λ)={((1+q)m,mpn),mp}.\lambda:E_{(q+1)^{p^{n}}-1}\longrightarrow E_{q};\qquad\lambda(a,x)=(a,p^{n}x),\qquad C_{n}^{\prime\prime}=\ker(\lambda)=\left\{\left((1+q)^{m},-\frac{m}{p^{n}}\right),\;m\in\operatorname{\mathbb{Z}}_{p}\right\}.

For any j(/pn)×j\in(\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}})^{\times}, the morphism ρj:Cn′′Cn={(ξm,0),m/pn}\rho_{j}:C_{n}^{\prime\prime}\rightarrow C_{n}^{\prime}=\{(\xi^{m},0),m\in\operatorname{\mathbb{Z}}/p^{n}\operatorname{\mathbb{Z}}\} is given by

ρj:Cn′′Cn,((1+q)m,mpn)(ξmj,0),(ρj×Id)Cn′′={(ξmj(1+q)m,mpn),mp}.\rho_{j}:C_{n}^{\prime\prime}\longrightarrow C_{n}^{\prime},\;\left((1+q)^{m},-\frac{m}{p^{n}}\right)\longmapsto(\xi^{mj},0),\qquad(\rho_{j}\times{\rm Id})C_{n}^{\prime\prime}=\left\{\left(\xi^{mj}(1+q)^{m},-\frac{m}{p^{n}}\right),\;m\in\operatorname{\mathbb{Z}}_{p}\right\}.

This implies that (see [3, §8.2])

𝒢0j=Eξj(q+1)1=(𝔾m,R[p]p)/(ξmj(1+q)m,m),mp.\mathcal{G}_{0}^{j}=E_{\xi^{j}(q+1)-1}=(\operatorname{\mathbb{G}}_{m,R}[p^{\infty}]\oplus\operatorname{\mathbb{Q}}_{p})/\langle(\xi^{mj}(1+q)^{m},-m),\;m\in\operatorname{\mathbb{Z}}_{p}\rangle.

Thus, for any f(q)W[[q]]=𝒪𝔾mf(q)\in W[[q]]=\mathcal{O}_{\operatorname{\mathbb{G}}_{m}}, we have tjf(q)=f(ξj(q+1)1)t_{j}^{\ast}f(q)=f(\xi^{j}(q+1)-1).

Over 𝔛()\mathfrak{X}(\infty), the sheaf 0\mathcal{H}_{0} has a cannonical basis ω,η\omega,\eta such that ω\omega is a basis of ω0\omega_{0}. In Serre-Tate coordinates the Gauss-Manin connection acts as follows (see [3, proof Theorem 9.9]): Given f1(q),f2(q)W[[q]]=𝒪𝔾mf_{1}(q),f_{2}(q)\in W[[q]]=\mathcal{O}_{\operatorname{\mathbb{G}}_{m}},

(25) (f1(q)ω+f2(q)η)=f1ω3+(f1+f2)ηω2,fi:=(q+1)dfidq.\bigtriangledown(f_{1}(q)\omega+f_{2}(q)\eta)=\partial f_{1}\omega^{3}+\left(f_{1}+\partial f_{2}\right)\eta\omega^{2},\qquad\partial f_{i}:=(q+1)\frac{df_{i}}{dq}.

By the local description of 𝕎0𝐤τ0,n\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}}, for an small open subset U𝔛()U\subset\mathfrak{X}(\infty), the space 𝕎0𝐤τ0,n(U)\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}}(U) is generated by elements of the form

V𝐤τ0,n,m:=Pn(Z)Tm(1+pnZ)mFilm𝕎0𝐤τ0,n,Pn(Z)=i0(wni)pniZi,V_{{\bf k}_{\tau_{0},n},m}:=P_{n}(Z)\frac{T^{m}}{(1+p^{n}Z)^{m}}\in{\rm Fil}_{m}\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}},\qquad P_{n}(Z)=\sum_{i\geq 0}\binom{w_{n}}{i}p^{ni}Z^{i},

where (1+pnZ)(1+p^{n}Z) corresponds to ω\omega and pnTp^{n}T corresponds to η\eta. By functoriality, the morphism fjtjf_{j}^{\ast}\circ t_{j}^{\ast} leaves ZZ and TT invariant. Hence, for any finite character χ\chi of conductor at most pnp^{n} and any f(q)=nan(q+1)nf(q)=\sum_{n}a_{n}(q+1)^{n}, we obtain

θχ(f(q)V𝐤τ0,n,m)\displaystyle\theta_{\chi}(f(q)V_{{\bf k}_{\tau_{0},n},m}) =\displaystyle= gχ11jχ(j)1fjtj(f(q)V𝐤τ0,n,m)=gχ11jχ(j)1f(ξj(q+1)1)V𝐤τ0,n,m\displaystyle g_{\chi^{-1}}^{-1}\sum_{j}\chi(j)^{-1}f_{j}^{\ast}\circ t_{j}^{\ast}(f(q)V_{{\bf k}_{\tau_{0},n},m})=g_{\chi^{-1}}^{-1}\sum_{j}\chi(j)^{-1}f(\xi^{j}(q+1)-1)V_{{\bf k}_{\tau_{0},n},m}
=\displaystyle= gχ11jχ(j)1(nanξjn(1+q)n)V𝐤τ0,n,m=fχ(q)V𝐤τ0,n,m,\displaystyle g_{\chi^{-1}}^{-1}\sum_{j}\chi(j)^{-1}\left(\sum_{n}a_{n}\xi^{jn}(1+q)^{n}\right)V_{{\bf k}_{\tau_{0},n},m}=f_{\chi}(q)V_{{\bf k}_{\tau_{0},n},m},

where fχ(q):=nχ(n)an(q+1)nf_{\chi}(q):=\sum_{n}\chi(n)a_{n}(q+1)^{n} (here χ(n)=0\chi(n)=0 if gcd(n,p)1\gcd(n,p)\neq 1).

Remark 6.3.

Notice that the image of θχ\theta_{\chi} lies in the kernel of U𝔭0U_{\mathfrak{p}_{0}} by [3, §8.3].

6.5. Iterations of the Gauss-Manin operator

For any character (k:p×R)𝔚τ0,n(R)(k:\operatorname{\mathbb{Z}}_{p}^{\times}\longrightarrow R)\in\mathfrak{W}_{\tau_{0},n}(R), let 𝕎0k(gnf0)𝒪𝕍Q(0,s0)^R\operatorname{\mathbb{W}}_{0}^{k}\subset(g_{n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}_{0}^{\sharp},s_{0})}\hat{\otimes}R be the specialization of 𝕎0𝐤τ0,n\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}}.

Let Λτ0,n\Lambda_{\tau_{0},n}^{\prime} be isomorphic to Λτ0,n\Lambda_{\tau_{0},n}. We aim to define 𝐤τ0,n𝐬τ0,n:𝕎0𝐤τ0,n𝕎0𝐤τ0,n+2𝐬τ0,n\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{{\bf s}_{\tau_{0},n}}:\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}}\rightarrow\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}+2{\bf s}_{\tau_{0},n}}, where 𝐬τ0,n:p×Λτ0,n{\bf s}_{\tau_{0},n}:\operatorname{\mathbb{Z}}_{p}^{\times}\rightarrow\Lambda_{\tau_{0},n}^{\prime} is the universal character seen in 𝔚τ0,2n(Λτ0,n^Λτ0,n)\mathfrak{W}_{\tau_{0},2n}(\Lambda_{\tau_{0},n}\hat{\otimes}\Lambda_{\tau_{0},n}^{\prime}). Recall that by (11)

𝐬τ0,n(β)=γ𝐬(γ)1γ+pnp(β)i0(wni)(βγγ)i,γ=w(s)exp(jp), 0jpn11,s(/p)×{\bf s}_{\tau_{0},n}(\beta)=\sum_{\gamma}{\bf s}(\gamma)\cdot 1_{\gamma+p^{n}\operatorname{\mathbb{Z}}_{p}}(\beta)\cdot\sum_{i\geq 0}\binom{w_{n}^{\prime}}{i}\left(\frac{\beta-\gamma}{\gamma}\right)^{i},\quad\gamma=w(s)\exp(jp),\;0\leq j\leq p^{n-1}-1,\;s\in(\operatorname{\mathbb{Z}}/p\operatorname{\mathbb{Z}})^{\times}

where wn=p1log(1+T)p1nΛτ0,nw_{n}^{\prime}=p^{-1}\log(1+T)\in p^{1-n}\Lambda_{\tau_{0},n}^{\prime}, ww is the Teichmüller character and 𝐬(w(s)exp(jp))=(1+T)j(s)Λτ0=p[(/p)×][[T]]{\bf s}(w(s)\exp(jp))=(1+T)^{j}(s)\in\Lambda_{\tau_{0}}=\operatorname{\mathbb{Z}}_{p}[(\operatorname{\mathbb{Z}}/p\operatorname{\mathbb{Z}})^{\times}][[T]].

This implies that 𝐤τ0,n𝐬τ0,n\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{{\bf s}_{\tau_{0},n}} should be a limit of the form

(26) 𝐤τ0,n𝐬τ0,n=γ𝐬(γ)i0(wni)(𝐤τ0,nγIdγ)i𝐤τ0,nγ+pnp.\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{{\bf s}_{\tau_{0},n}}=\sum_{\gamma}{\bf s}(\gamma)\cdot\sum_{i\geq 0}\binom{w_{n}^{\prime}}{i}\left(\frac{\bigtriangledown_{{\bf k}_{\tau_{0},n}}-\gamma{\rm Id}}{\gamma}\right)^{i}\circ\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\gamma+p^{n}\operatorname{\mathbb{Z}}_{p}}.

First, using the definition of the operator 𝐤τ0,nχ\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\chi} of Lemma 6.2 we define

𝐤τ0,nγ+pnp:=1φ(pn)χχ(γ)1𝐤τ0,nχ,\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\gamma+p^{n}\operatorname{\mathbb{Z}}_{p}}:=\frac{1}{\varphi(p^{n})}\sum_{\chi}\chi(\gamma)^{-1}\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\chi},

where the sum is taken over all characters of conductor pnp^{n} and φ(pn)=(p1)pn1\varphi(p^{n})=(p-1)p^{n-1}. Notice that, locally over the ordinary locus and in Serre-Tate coordinates

𝐤τ0,nγ+pnp(an(1+q)nV𝐤τ0,n,m)\displaystyle\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\gamma+p^{n}\operatorname{\mathbb{Z}}_{p}}(a_{n}(1+q)^{n}V_{{\bf k}_{\tau_{0},n},m}) =\displaystyle= an(1+q)nV𝐤τ0,n,m1φ(pn)χχ(γ)1χ(n)\displaystyle a_{n}(1+q)^{n}V_{{\bf k}_{\tau_{0},n},m}\frac{1}{\varphi(p^{n})}\sum_{\chi}\chi(\gamma)^{-1}\chi(n)
=\displaystyle= {an(q+1)nV𝐤τ0,n,m,nγ(modpn)0,nγ(modpn).\displaystyle\left\{\begin{array}[]{lc}a_{n}(q+1)^{n}V_{{\bf k}_{\tau_{0},n},m},&n\equiv\gamma\;({\rm mod}\;p^{n})\\ 0,&n\not\equiv\gamma\;({\rm mod}\;p^{n}).\end{array}\right.

Once 𝐤τ0,nγ+pnp\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\gamma+p^{n}\operatorname{\mathbb{Z}}_{p}} is defined, the existence of 𝐤τ0,n𝐬τ0,n\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{{\bf s}_{\tau_{0},n}} will follow from the convergence of the series given in (26).

Assuming as always that rnr\geq n, after applying 𝐤τ0,nγ+pnp\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\gamma+p^{n}\operatorname{\mathbb{Z}}_{p}} to a section in 𝔛r×𝔚n\mathfrak{X}_{r}\times\mathfrak{W}_{n}, we obtain a section in 𝔛r+n×𝔚n\mathfrak{X}_{r+n}\times\mathfrak{W}_{n}. Over 𝔛r+n×𝔚n\mathfrak{X}_{r+n}\times\mathfrak{W}_{n} we will denote by 𝕎0𝐤τ0,n+2𝐬τ0,n\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}+2{\bf s}_{\tau_{0},n}} the specialization of 𝕎0𝐤τ0,2n\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},2n}} at 𝐤τ0,n+2𝐬τ0,n{\bf k}_{\tau_{0},n}+2{\bf s}_{\tau_{0},n}. Since any locally analytic function with radius of analycity pnp^{n} has also radius of analicity p2np^{2n}, we can always see a section in 𝕎0𝐤τ0,n\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}} as an specialization of a section in 𝕎0𝐤τ0,2n\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},2n}}.

To ensure the convergence of the series (26), we will prove in Proposition 6.6 that

γ𝐬(γ)(wni)(𝐤τ0,nγIdγ)i𝐤τ0,nγ+pnp(s)H0(𝔛ord×𝔚n×𝔚τ0,n,pipp1(g2nf0)𝒪𝕍Q(,s0)),\sum_{\gamma}{\bf s}(\gamma)\cdot\binom{w_{n}^{\prime}}{i}\left(\frac{\bigtriangledown_{{\bf k}_{\tau_{0},n}}-\gamma{\rm Id}}{\gamma}\right)^{i}\circ\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\gamma+p^{n}\operatorname{\mathbb{Z}}_{p}}(s)\in H^{0}(\mathfrak{X}_{\rm ord}\times\mathfrak{W}_{n}\times\mathfrak{W}_{\tau_{0},n}^{\prime},p^{i\frac{p}{p-1}}(g_{2n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}^{\sharp},s_{0})}),

for any sH0(𝔛r×𝔚n,𝕎0𝐤τ0,n)s\in H^{0}(\mathfrak{X}_{r}\times\mathfrak{W}_{n},\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}}) with 𝔚τ0,n:=Spf(Λτ0,n)\mathfrak{W}_{\tau_{0},n}^{\prime}:={\rm Spf}(\Lambda_{\tau_{0},n}^{\prime}). By Theorem 5.5 we have

Ai:=Hdgicnγ𝐬(γ)(wni)(𝐤τ0,nγIdγ)i𝐤τ0,nγ+pnpH0(𝔛r+n×𝔚n×𝔚τ0,n,(g2nf0)𝒪𝕍Q(,s0)).A_{i}:={\rm Hdg}^{ic_{n}}\sum_{\gamma}{\bf s}(\gamma)\cdot\binom{w_{n}^{\prime}}{i}\left(\frac{\bigtriangledown_{{\bf k}_{\tau_{0},n}}-\gamma{\rm Id}}{\gamma}\right)^{i}\circ\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\gamma+p^{n}\operatorname{\mathbb{Z}}_{p}}\in H^{0}(\mathfrak{X}_{r+n}\times\mathfrak{W}_{n}\times\mathfrak{W}_{\tau_{0},n}^{\prime},(g_{2n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}^{\sharp},s_{0})}).

Hence applying [1, Proposition 4.10], we deduce that there exists bb depending on rr and nn such that

Ai(s)p[i/2]H0(𝔛b×𝔚n×𝔚τ0,n,(g2nf0)𝒪𝕍Q(,s0)),A_{i}(s)\in p^{[i/2]}H^{0}(\mathfrak{X}_{b}\times\mathfrak{W}_{n}\times\mathfrak{W}_{\tau_{0},n}^{\prime},(g_{2n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}^{\sharp},s_{0})}),

for any sH0(𝔛r×𝔚n,𝕎0𝐤τ0,n)s\in H^{0}(\mathfrak{X}_{r}\times\mathfrak{W}_{n},\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}}). This provides the existence of the operator

𝐤τ0,n𝐬τ0,n:=i0Ai:H0(𝔛r×𝔚n,𝕎0𝐤τ0,n)H0(𝔛b×𝔚n×𝔚τ0,n,𝕎0𝐤τ0,n+2𝐬τ0,n)U𝔭0=0.\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{{\bf s}_{\tau_{0},n}}:=\sum_{i\geq 0}A_{i}:H^{0}(\mathfrak{X}_{r}\times\mathfrak{W}_{n},\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}})\longrightarrow H^{0}(\mathfrak{X}_{b}\times\mathfrak{W}_{n}\times\mathfrak{W}_{\tau_{0},n}^{\prime},\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}+2{\bf s}_{\tau_{0},n}})^{U_{\mathfrak{p}_{0}}=0}.

Moreover, by the action of U𝔭0U_{\mathfrak{p}_{0}} and V𝔭0V_{\mathfrak{p}_{0}} on qq-expansions described in [3, §8.3], for any integer m𝔚τ0,n(p)m\in\mathfrak{W}_{\tau_{0},n}^{\prime}(\operatorname{\mathbb{C}}_{p})\cap\operatorname{\mathbb{N}} and any section sH0(𝔛r×𝔚n,𝕎0𝐤τ0,n)s\in H^{0}(\mathfrak{X}_{r}\times\mathfrak{W}_{n},\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}}) we have

(28) m𝐤τ0,n𝐬τ0,n(s)=𝐤τ0,nm(s[p])=(𝐤τ0,nm(1V𝔭0U𝔭0))(s).m^{\ast}\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{{\bf s}_{\tau_{0},n}}(s)=\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{m}(s^{[p]})=\left(\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{m}\circ(1-V_{\mathfrak{p}_{0}}\circ U_{\mathfrak{p}_{0}})\right)(s).
Remark 6.4.

The existence of 𝐤τ0,n𝐬τ0,n\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{{\bf s}_{\tau_{0},n}} generalize the results of Andreatta-Iovita in [1] even for F=F=\operatorname{\mathbb{Q}}. Indeed, they were only able to construct 𝐤ns\bigtriangledown_{{\bf k}_{n}}^{s} under certain analycity conditions for ss ([1, Assumption 5.4]). The innovative tools that have allow us to construct 𝐤τ0,n𝐬τ0,n\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{{\bf s}_{\tau_{0},n}} for the universal character 𝐬τ0,n{\bf s}_{\tau_{0},n} and were not available in [1] are the locally analytic description of 𝐬τ0,n{\bf s}_{\tau_{0},n}, provided by Equation (11), and the construction of 𝕎0𝐤τ0,n\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}} by means of the new formal vector bundles 𝕍Q(,s1,,sm)\operatorname{\mathbb{V}}_{Q}(\mathcal{E},s_{1},\cdots,s_{m}) described in §4.2, that ensure the convergence of the power series AiA_{i} as we will see in the next section.

6.6. Computations in Serre-Tate coordinates

For any aa\in\operatorname{\mathbb{N}}, the local description of 𝕎0𝐤τ0,n+a\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}+a} over 𝔛()\mathfrak{X}(\infty) tells us that any section of 𝕎0𝐤τ0,n+a(U)\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}+a}(U), for some open U𝔛()U\subset\mathfrak{X}(\infty), is generated by elements of the form

V𝐤τ0,n+a,m:=P2n(Z)(1+p2nZ)aTm(1+p2nZ)m.V_{{\bf k}_{\tau_{0},n}+a,m}:=P_{2n}(Z)(1+p^{2n}Z)^{a}\frac{T^{m}}{(1+p^{2n}Z)^{m}}.

Applying the relations of (25), we obtain that

𝐤τ0,n+a(V𝐤τ0,n+a,m)=p2n(wn+rm)V𝐤τ0,n+a+2,m+1,\bigtriangledown_{{\bf k}_{\tau_{0},n}+a}(V_{{\bf k}_{\tau_{0},n}+a,m})=p^{2n}(w_{n}+r-m)\cdot V_{{\bf k}_{\tau_{0},n}+a+2,m+1},

and this implies that (see [1, Lemma 3.38])

(29) 𝐤τ0,ns(fV𝐤τ0,n,m)=i=0sp2ni(si)(wn+sm1i)i!sifV𝐤τ0,n+2s,m+i,s.\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{s}(f\cdot V_{{\bf k}_{\tau_{0},n},m})=\sum_{i=0}^{s}p^{2ni}\binom{s}{i}\binom{w_{n}+s-m-1}{i}i!\cdot\partial^{s-i}f\cdot V_{{\bf k}_{\tau_{0},n}+2s,m+i},\qquad s\in\operatorname{\mathbb{N}}.

Assume that f(q)=an(1+q)nf(q)=a_{n}(1+q)^{n}, where pnp\nmid n, and assume that nγ0n\equiv\gamma_{0} modulo pnp^{n}. Since in this case f=nf\partial f=nf, we compute that

γ(wni)(𝐤τ0,nγIdγ)i𝐤τ0,nγ+pnp(fV𝐤τ0,n,m)=(wni)(𝐤τ0,nγ0Idγ0)i(fV𝐤τ0,n,m)\displaystyle\sum_{\gamma}\binom{w_{n}^{\prime}}{i}\left(\frac{\bigtriangledown_{{\bf k}_{\tau_{0},n}}-\gamma{\rm Id}}{\gamma}\right)^{i}\circ\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\gamma+p^{n}\operatorname{\mathbb{Z}}_{p}}(f\cdot V_{{\bf k}_{\tau_{0},n},m})=\binom{w_{n}^{\prime}}{i}\left(\frac{\bigtriangledown_{{\bf k}_{\tau_{0},n}}-\gamma_{0}{\rm Id}}{\gamma_{0}}\right)^{i}(f\cdot V_{{\bf k}_{\tau_{0},n},m})
=\displaystyle= (wni)j=0i(1)ij(ij)γ0j𝐤τ0,nj(fV𝐤τ0,n,m)\displaystyle\binom{w_{n}^{\prime}}{i}\sum_{j=0}^{i}(-1)^{i-j}\binom{i}{j}\gamma_{0}^{-j}\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{j}(f\cdot V_{{\bf k}_{\tau_{0},n},m})
=\displaystyle= (wni)j=0i(ij)(1)ijγ0jk=0jp2nk(jk)(wn+jm1k)k!jkfV𝐤τ0,n+2j,m+k\displaystyle\binom{w_{n}^{\prime}}{i}\sum_{j=0}^{i}\binom{i}{j}(-1)^{i-j}\gamma_{0}^{-j}\sum_{k=0}^{j}p^{2nk}\binom{j}{k}\binom{w_{n}+j-m-1}{k}k!\cdot\partial^{j-k}f\cdot V_{{\bf k}_{\tau_{0},n}+2j,m+k}
=\displaystyle= (wni)fV𝐤τ0,n,mj=0ik=0j(ij)(jk)(M+jk)k!(1)ijAkBj,\displaystyle\binom{w_{n}^{\prime}}{i}fV_{{\bf k}_{\tau_{0},n},m}\sum_{j=0}^{i}\sum_{k=0}^{j}\binom{i}{j}\binom{j}{k}\binom{M+j}{k}k!(-1)^{i-j}A^{k}B^{j},

where M:=wnm1M:=w_{n}-m-1 A:=(p2nTn(1+p2nZ))A:=\left(\frac{p^{2n}T}{n(1+p^{2n}Z)}\right) and B:=(n(1+p2nZ)2γ0)B:=\left(\frac{n(1+p^{2n}Z)^{2}}{\gamma_{0}}\right)

Lemma 6.5.

We have that

j=0ik=0j(ij)(jk)(M+jk)k!(1)ijAkBj=\displaystyle\sum_{j=0}^{i}\sum_{k=0}^{j}\binom{i}{j}\binom{j}{k}\binom{M+j}{k}k!(-1)^{i-j}A^{k}B^{j}=
=(1)ik=0i(ik)(AB)kk!j=0k(M+kkj)(ikj)(1B)ikj(B)j.\displaystyle=(-1)^{i}\sum_{k=0}^{i}\binom{i}{k}(-AB)^{k}k!\sum_{j=0}^{k}\binom{M+k}{k-j}\binom{i-k}{j}(1-B)^{i-k-j}(-B)^{j}.
Proof.

We compute that

j=0k(M+kkj)(ikj)(1B)ikj(B)j\displaystyle\sum_{j=0}^{k}\binom{M+k}{k-j}\binom{i-k}{j}(1-B)^{i-k-j}(-B)^{j} =\displaystyle= j=0kr=0ikj(M+kkj)(ikj)(ikjr)(B)r+j=\displaystyle\sum_{j=0}^{k}\sum_{r=0}^{i-k-j}\binom{M+k}{k-j}\binom{i-k}{j}\binom{i-k-j}{r}(-B)^{r+j}=
=j=0is=jik(M+kkj)(ikj)(ikjsj)(B)s\displaystyle=\sum_{j=0}^{i}\sum_{s=j}^{i-k}\binom{M+k}{k-j}\binom{i-k}{j}\binom{i-k-j}{s-j}(-B)^{s} =\displaystyle= s=0ik(iks)(B)sj=0k(M+kkj)(sj)=\displaystyle\sum_{s=0}^{i-k}\binom{i-k}{s}(-B)^{s}\sum_{j=0}^{k}\binom{M+k}{k-j}\binom{s}{j}=
=\displaystyle= s=0ik(iks)(M+k+sk)(B)s.\displaystyle\sum_{s=0}^{i-k}\binom{i-k}{s}\binom{M+k+s}{k}(-B)^{s}.

Thus, we obtain

(1)ik=0i(ik)(AB)kk!j=0k(M+kkj)(ikj)(1B)ikj(B)j=\displaystyle(-1)^{i}\sum_{k=0}^{i}\binom{i}{k}(-AB)^{k}k!\sum_{j=0}^{k}\binom{M+k}{k-j}\binom{i-k}{j}(1-B)^{i-k-j}(-B)^{j}=
=\displaystyle= (1)ik=0is=0ik(ik)(iks)(M+k+sk)k!Ak(B)k+s=\displaystyle(-1)^{i}\sum_{k=0}^{i}\sum_{s=0}^{i-k}\binom{i}{k}\binom{i-k}{s}\binom{M+k+s}{k}k!A^{k}(-B)^{k+s}=
=\displaystyle= (1)ik=0ij=ki(ik)(ikjk)(M+jk)k!Ak(B)j,\displaystyle(-1)^{i}\sum_{k=0}^{i}\sum_{j=k}^{i}\binom{i}{k}\binom{i-k}{j-k}\binom{M+j}{k}k!A^{k}(-B)^{j},

hence the result follows from a change of variable and an easy equality with binomials. ∎

Since p2nAp^{2n}\mid A, pn(1B)p^{n}\mid(1-B) and k!(M+km)pnmΛτ0,nk!\binom{M+k}{m}\in p^{-nm}\Lambda_{\tau_{0},n} (see Remark 4.8), we deduce that

pn(i2k)k!(M+kkj)(ikj)(1B)ikj(B)j,\displaystyle p^{n(i-2k)}\mid k!\binom{M+k}{k-j}\binom{i-k}{j}(1-B)^{i-k-j}(-B)^{j},
pni(1)ik=0i(ik)(AB)kk!j=0k(M+kkj)(ikj)(1B)ikj(B)j.\displaystyle p^{ni}\mid(-1)^{i}\sum_{k=0}^{i}\binom{i}{k}(-AB)^{k}k!\sum_{j=0}^{k}\binom{M+k}{k-j}\binom{i-k}{j}(1-B)^{i-k-j}(-B)^{j}.

By Remark 4.8 (wni)pi(npp1)Λτ0,n\binom{w_{n}^{\prime}}{i}\in p^{-i\left(n-\frac{p}{p-1}\right)}\Lambda_{\tau_{0},n}^{\prime}, hence the above lemma provides the following result:

Proposition 6.6.

For any sH0(𝔛ord,𝕎0𝐤τ0,n)s\in H^{0}(\mathfrak{X}_{\rm ord},\operatorname{\mathbb{W}}_{0}^{{\bf k}_{\tau_{0},n}}),

γ(wni)(𝐤τ0,nγIdγ)i𝐤τ0,nγ+pnp(s)pipp1H0(𝔛ord×𝔚n×𝔚τ0,n,(g2nf0)𝒪𝕍Q(,s0)).\sum_{\gamma}\binom{w_{n}^{\prime}}{i}\left(\frac{\bigtriangledown_{{\bf k}_{\tau_{0},n}}-\gamma{\rm Id}}{\gamma}\right)^{i}\circ\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{\gamma+p^{n}\operatorname{\mathbb{Z}}_{p}}(s)\in p^{i\frac{p}{p-1}}H^{0}(\mathfrak{X}_{\rm ord}\times\mathfrak{W}_{n}\times\mathfrak{W}_{\tau_{0},n}^{\prime},(g_{2n}\circ f_{0})_{\ast}\mathcal{O}_{\operatorname{\mathbb{V}}_{Q}(\mathcal{H}^{\sharp},s_{0})}).
Remark 6.7.

This computation of the pp-adic valuation shows the necessity of working with radius of analicity p2np^{2n}. One could predict this fact by observing (29) and guessing that over 𝔛ord\mathfrak{X}_{\rm ord}

𝐤τ0,n𝐬τ0,n(fV𝐤τ0,n,m)=i0p2ni(wni)(wn+wnm1i)i!𝐬τ0,nifV𝐤τ0,n+2𝐬τ0,n,m+i,\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{{\bf s}_{\tau_{0},n}}(f\cdot V_{{\bf k}_{\tau_{0},n},m})=\sum_{i\geq 0}p^{2ni}\binom{w_{n}^{\prime}}{i}\binom{w_{n}+w_{n}^{\prime}-m-1}{i}i!\cdot\partial^{{\bf s}_{\tau_{0},n}-i}f\cdot V_{{\bf k}_{\tau_{0},n}+2{\bf s}_{\tau_{0},n},m+i},

where 𝐬τ0,nian(1+q)n=𝐬τ0,n(n)an(1+q)n\partial^{{\bf s}_{\tau_{0},n}-i}a_{n}(1+q)^{n}={\bf s}_{\tau_{0},n}(n)a_{n}(1+q)^{n}, if pnp\nmid n, and 𝐬τ0,nian(1+q)n=0\partial^{{\bf s}_{\tau_{0},n}-i}a_{n}(1+q)^{n}=0, otherwise.

6.7. Triple products for any weight

The existence of 𝐤τ0,n𝐬τ0,n\bigtriangledown_{{\bf k}_{\tau_{0},n}}^{{\bf s}_{\tau_{0},n}} together with the morphism

H,h:H0(𝒳r×𝒲n,𝕎𝐤τ0,n)H0(𝒳r×𝒲n,Ω0𝐤τ0,n)hΛτ0,nΛIm,H^{\dagger,\leq h}:H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},\operatorname{\mathbb{W}}^{{\bf k}_{\tau_{0},n}})\longrightarrow H^{0}(\mathcal{X}_{r}\times\mathcal{W}_{n},\Omega_{0}^{{\bf k}_{\tau_{0},n}})^{\leq h}\otimes_{\Lambda_{\tau_{0},n}}\Lambda_{I_{m}},

of (22) provides the triple product

Θ𝐤τ0,n3:H0(𝒳r,n1,Ω0𝐤τ0,n1)×H0(𝒳r,n2,Ω0𝐤τ0,n2)\displaystyle\Theta_{{\bf k}_{\tau_{0},n_{3}}}:H^{0}(\mathcal{X}_{r,n_{1}},\Omega_{0}^{{\bf k}_{\tau_{0},n_{1}}})\times H^{0}(\mathcal{X}_{r,n_{2}},\Omega_{0}^{{\bf k}_{\tau_{0},n_{2}}}) \displaystyle\longrightarrow H0(𝒳b,n1,n2,n3,Ω0𝐤τ0,n3)hΛn1,n2,n3ΛIm\displaystyle H^{0}(\mathcal{X}_{b,n_{1},n_{2},n_{3}},\Omega_{0}^{{\bf k}_{\tau_{0},n_{3}}^{\ast}})^{\leq h}\otimes_{\Lambda_{n_{1},n_{2},n_{3}}}\Lambda_{I_{m}}
𝐭τ0(s1,s2)\displaystyle{\bf t}_{\tau_{0}}(s_{1},s_{2}) :=\displaystyle:= H,h(𝐤τ0,n1𝐤τ0,n3s1s2),\displaystyle H^{\dagger,\leq h}\left(\bigtriangledown_{{\bf k}_{\tau_{0},n_{1}}}^{{\bf k}_{\tau_{0},n_{3}}}s_{1}\cdot s_{2}\right),

where 𝐤τ0,n3:=𝐤τ0,n1+𝐤τ0,n2+2𝐤τ0,n3{\bf k}_{\tau_{0},n_{3}}^{\ast}:={\bf k}_{\tau_{0},n_{1}}+{\bf k}_{\tau_{0},n_{2}}+2{\bf k}_{\tau_{0},n_{3}}, 𝒳r,ni:=𝒳r×𝒲ni\mathcal{X}_{r,n_{i}}:=\mathcal{X}_{r}\times\mathcal{W}_{n_{i}}, 𝒳b,n1,n2,n3:=𝒳b×𝒲n1×𝒲n2×𝒲n3\mathcal{X}_{b,n_{1},n_{2},n_{3}}:=\mathcal{X}_{b}\times\mathcal{W}_{n_{1}}\times\mathcal{W}_{n_{2}}\times\mathcal{W}_{n_{3}}, Λn1,n2,n3:=Λτ0,n1Λτ0,n2Λτ0,n3\Lambda_{n_{1},n_{2},n_{3}}:=\Lambda_{\tau_{0},n_{1}}\otimes\Lambda_{\tau_{0},n_{2}}\otimes\Lambda_{\tau_{0},n_{3}}, and the superindex h\leq h relies on the slope of the operator U𝔭0U_{\mathfrak{p}_{0}}. By Equations (28) and (22) we have that, for any m𝒲τ0,n3(p)m\in\mathcal{W}_{\tau_{0},n_{3}}(\operatorname{\mathbb{C}}_{p})\cap\operatorname{\mathbb{N}},

(30) mΘ𝐤τ0,n3(s1,s2)=ehΘm(s1[p],s2).m^{\ast}\Theta_{{\bf k}_{\tau_{0},n_{3}}}(s_{1},s_{2})=e_{\leq h}\Theta_{m}(s_{1}^{[p]},s_{2}).

Recall that sections on the sheaf Ω𝐤n\Omega^{{\bf k}_{n}} can be described using Katz interpretation (§4.5) as rules that assign to any tuple (A,ı,θ,α𝔭0,w)(A,\imath,\theta,\alpha^{\mathfrak{p}_{0}},w) over RR a locally analytic distribution μ(A,ı,θ,α𝔭0,w)Dn𝐤nτ0(𝒪τ0,R)\mu(A,\imath,\theta,\alpha^{\mathfrak{p}_{0}},w)\in D_{n}^{{\bf k}_{n}^{\tau_{0}}}(\mathcal{O}^{\tau_{0}},R). Recall also that w=(w0,wτ0)w=(w_{0},w^{\tau_{0}}) is a basis for Ω=Ω0𝔭𝔭0,τω𝔭,τ\Omega=\Omega_{0}\oplus\bigoplus_{\mathfrak{p}\neq\mathfrak{p}_{0},\tau}\omega_{\mathfrak{p},\tau}.

We denote by ΛF1\Lambda_{F}^{1} the (d+1)(d+1)-dimesional weight space ΛF1:=p[[𝒪××p×]]\Lambda_{F}^{1}:=\operatorname{\mathbb{Z}}_{p}[[\mathcal{O}^{\times}\times\operatorname{\mathbb{Z}}_{p}^{\times}]]. Similarly, we define

Λn1:=ΛF1T1pnp,,Tdpnp,Tpnp,𝔚1:=Spf(ΛF1),𝔚n1:=Spf(Λn1).\Lambda_{n}^{1}:=\Lambda_{F}^{1}\left\langle\frac{T_{1}^{p^{n}}}{p},\cdots,\frac{T_{d}^{p^{n}}}{p},\frac{T^{p^{n}}}{p}\right\rangle,\qquad\mathfrak{W}^{1}:={\rm Spf}(\Lambda_{F}^{1}),\qquad\mathfrak{W}_{n}^{1}:={\rm Spf}(\Lambda_{n}^{1}).

The morphism 𝒪×𝒪××p×\mathcal{O}^{\times}\rightarrow\mathcal{O}^{\times}\times\operatorname{\mathbb{Z}}_{p}^{\times} given by t(t2,N(t))t\mapsto(t^{-2},{\rm N}(t)), where N(t){\rm N}(t) is the norm of tt, provides a morphism of weight spaces k:𝔚1𝔚k:\mathfrak{W}^{1}\rightarrow\mathfrak{W}. By [3, Lemma 7.1] it gives rise to a morphism k:𝔚n1𝔚nk:\mathfrak{W}_{n}^{1}\rightarrow\mathfrak{W}_{n}.

We fix integers rn3n1,n21r\geq n_{3}\geq n_{1},n_{2}\geq 1. For i=1,2,3i=1,2,3 we denote by (𝐫ni,νni):𝒪××p×Λni1({\bf r}_{n_{i}},{\bf\nu}_{n_{i}}):\mathcal{O}^{\times}\times\operatorname{\mathbb{Z}}_{p}^{\times}\rightarrow\Lambda_{n_{i}}^{1} the universal characters of 𝔚ni1\mathfrak{W}^{1}_{n_{i}}. Then we put 𝐤n3:=k(𝐫n3,νn1+νn2){\bf k}_{n_{3}}:=k({\bf r}_{n_{3}},\nu_{n_{1}}+\nu_{n_{2}}) and 𝐤ni:=k(𝐫ni,νni){\bf k}_{n_{i}}:=k({\bf r}_{n_{i}},{\bf\nu}_{n_{i}}) for i=1,2i=1,2.

We put :=Λn11^Λn21^νn3=νn1+νn2Λn31Λn11^Λn21^Λn3\mathcal{R}:=\Lambda_{n_{1}}^{1}\hat{\otimes}\Lambda_{n_{2}}^{1}\hat{\otimes}_{{\bf\nu}_{n_{3}}=\nu_{n_{1}}+{\bf\nu}_{n_{2}}}\Lambda_{n_{3}}^{1}\simeq\Lambda_{n_{1}}^{1}\hat{\otimes}\Lambda_{n_{2}}^{1}\hat{\otimes}\Lambda_{n_{3}} and consider the characters:

𝐦1τ0:=𝐫n1τ0𝐫n3τ0𝐫n2τ0+νn2N:𝒪τ0×\displaystyle{\bf m}_{1}^{\tau_{0}}:={\bf r}_{n_{1}}^{\tau_{0}}-{\bf r}_{n_{3}}^{\tau_{0}}-{\bf r}^{\tau_{0}}_{n_{2}}+{\bf\nu}_{n_{2}}\circ N:\mathcal{O}^{\tau_{0}\times} \displaystyle\longrightarrow ,\displaystyle\mathcal{R},
𝐦2τ0:=𝐫n2τ0𝐫n1τ0𝐫n3τ0+νn1N:𝒪τ0×\displaystyle{\bf m}_{2}^{\tau_{0}}:={\bf r}^{\tau_{0}}_{n_{2}}-{\bf r}^{\tau_{0}}_{n_{1}}-{\bf r}^{\tau_{0}}_{n_{3}}+{\bf\nu}_{n_{1}}\circ N:\mathcal{O}^{\tau_{0}\times} \displaystyle\longrightarrow ,\displaystyle\mathcal{R},
𝐦3τ0:=𝐫n3τ0𝐫n1τ0𝐫n2τ0:𝒪τ0×\displaystyle{\bf m}_{3}^{\tau_{0}}:={\bf r}^{\tau_{0}}_{n_{3}}-{\bf r}^{\tau_{0}}_{n_{1}}-{\bf r}^{\tau_{0}}_{n_{2}}:\mathcal{O}^{\tau_{0}\times} \displaystyle\longrightarrow ,\displaystyle\mathcal{R},
𝐦3,τ0:=𝐫n1,τ0+𝐫n2,τ0𝐫n3,τ0:p×\displaystyle{\bf m}_{3,\tau_{0}}:={\bf r}_{n_{1},\tau_{0}}+{\bf r}_{n_{2},\tau_{0}}-{\bf r}_{n_{3},\tau_{0}}:\operatorname{\mathbb{Z}}_{p}^{\times} \displaystyle\longrightarrow ,\displaystyle\mathcal{R},

where N:𝒪τ0×p×N:\mathcal{O}^{\tau_{0}\times}\rightarrow\operatorname{\mathbb{Z}}_{p}^{\times} denotes the norm map. In the same way as in [3, §10.2] we denote Δ¯τ0Cn1𝐤n1τ0(𝒪τ0,)Cn2𝐤n2τ0(𝒪τ0,)C¯n3𝐤n3τ0(𝒪τ0,)\underline{\Delta}^{\tau_{0}}\in C_{n_{1}}^{{\bf k}_{n_{1}}^{\tau_{0}}}(\mathcal{O}^{\tau_{0}},\mathcal{R})\otimes_{\mathcal{R}}C_{n_{2}}^{{\bf k}_{n_{2}}^{\tau_{0}}}(\mathcal{O}^{\tau_{0}},\mathcal{R})\otimes_{\mathcal{R}}\bar{C}_{n_{3}}^{{\bf k}_{n_{3}}^{\tau_{0}}}(\mathcal{O}^{\tau_{0}},\mathcal{R}) the function defined by:

Δ¯τ0((x1,y1),(x2,y2),(x3,y3)):=𝐦1τ0(x3y2x2y3)𝐦2τ0(x3y1x1y3)𝐦3τ0(x1y2x2y1),\underline{\Delta}^{\tau_{0}}((x_{1},y_{1}),(x_{2},y_{2}),(x_{3},y_{3})):={\bf m}_{1}^{\tau_{0}}(x_{3}y_{2}-x_{2}y_{3})\cdot{\bf m}_{2}^{\tau_{0}}(x_{3}y_{1}-x_{1}y_{3})\cdot{\bf m}_{3}^{\tau_{0}}(x_{1}y_{2}-x_{2}y_{1}),

where C¯nkτ0(𝒪τ0,)\bar{C}_{n}^{k^{\tau_{0}}}(\mathcal{O}^{\tau_{0}},\cdot) denote the kτ0k^{\tau_{0}}-homogeneous locally analytic functions on p𝒪τ0×𝒪τ0×p\mathcal{O}^{\tau_{0}}\times\mathcal{O}^{\tau_{0}\times}, and the function is extended by 0 where 𝐦3τ0{\bf m}_{3}^{\tau_{0}} is not defined.

Let μiH0(𝒳r,ni1,Ω𝐤ni)\mu_{i}\in H^{0}(\mathcal{X}_{r,n_{i}}^{1},\Omega^{{\bf k}_{n_{i}}}), where i=1,2i=1,2 and 𝒳r,ni1:=𝒳r×𝒲ni1\mathcal{X}_{r,n_{i}}^{1}:=\mathcal{X}_{r}\times\mathcal{W}_{n_{i}}^{1}, be a pair of global sections. The choice of a basis wτ0w^{\tau_{0}} of 𝔭𝔭0,τω𝔭,τ\bigoplus_{\mathfrak{p}\neq\mathfrak{p}_{0},\tau}\omega_{\mathfrak{p},\tau} in some affine open U=Spf(R)U={\rm Spf}(R), together with the function Δ¯τ0\underline{\Delta}^{\tau_{0}} and the map Θ𝐦3,τ0\Theta_{{\bf m}_{3,\tau_{0}}} above provide an element t(μ1,μ2)HomR(D¯n3𝐤n3τ0(𝒪τ0,R),Ω0𝐤τ0,n3(U))t(\mu_{1},\mu_{2})\in{\rm Hom}_{R}(\bar{D}_{n_{3}}^{{\bf k}_{n_{3}}^{\tau_{0}}}(\mathcal{O}^{\tau_{0}},R),\Omega_{0}^{{\bf k}_{\tau_{0},n_{3}}}(U)) as follows:

t(μ1,μ2)(μ):=𝒪τ0××𝒪τ0𝒪τ0××𝒪τ0p𝒪τ0×𝒪τ0×Δ¯τ0(v1,v2,v3)𝑑Θ𝐦3,τ0(μ1,μ2)(v1,v2)𝑑μ(v3).t(\mu_{1},\mu_{2})(\mu):=\int_{\mathcal{O}^{\tau_{0}\times}\times\mathcal{O}^{\tau_{0}}}\int_{\mathcal{O}^{\tau_{0}\times}\times\mathcal{O}^{\tau_{0}}}\int_{p\mathcal{O}^{\tau_{0}}\times\mathcal{O}^{\tau_{0}\times}}\underline{\Delta}^{\tau_{0}}(v_{1},v_{2},v_{3})\;d\Theta_{{\bf m}_{3,\tau_{0}}}(\mu_{1},\mu_{2})(v_{1},v_{2})d\mu(v_{3}).

By [3, §10.1], the space HomR(D¯n3𝐤n3τ0(𝒪τ0,R),Ω0𝐤τ0,n3(U)){\rm Hom}_{R}(\bar{D}_{n_{3}}^{{\bf k}_{n_{3}}^{\tau_{0}}}(\mathcal{O}^{\tau_{0}},R),\Omega_{0}^{{\bf k}_{\tau_{0},n_{3}}}(U)) is endowed with the action of the Hecke operators U𝔭U_{\mathfrak{p}}, for 𝔭𝔭0\mathfrak{p}\neq\mathfrak{p}_{0}. These are compact operators, hence we can define the usual finite slope projectors eh¯e_{\leq\underline{h}}, where h¯#{𝔭𝔭0}\underline{h}\in\operatorname{\mathbb{N}}^{\#\{\mathfrak{p}\neq\mathfrak{p}_{0}\}}, onto the spaces HomR(D¯n3𝐤n3τ0(𝒪τ0,R),Ω0𝐤τ0,n3(U))h¯{\rm Hom}_{R}(\bar{D}_{n_{3}}^{{\bf k}_{n_{3}}^{\tau_{0}}}(\mathcal{O}^{\tau_{0}},R),\Omega_{0}^{{\bf k}_{\tau_{0},n_{3}}}(U))^{\leq\underline{h}}. The same argument showed in [3, Lemma 10.2] implies that we have an isomorphism of R[U𝔭]𝔭R[U_{\mathfrak{p}}]_{\mathfrak{p}}-modules:

HomR(D¯n3𝐤n3τ0(𝒪τ0,R),Ω0𝐤τ0,n3(U))h¯Dn3𝐤n3τ0(𝒪τ0,Ω0𝐤τ0,n3(U))h¯.{\rm Hom}_{R}(\bar{D}_{n_{3}}^{{\bf k}_{n_{3}}^{\tau_{0}}}(\mathcal{O}^{\tau_{0}},R),\Omega_{0}^{{\bf k}_{\tau_{0},n_{3}}}(U))^{\leq\underline{h}}\simeq D_{n_{3}}^{{\bf k}_{n_{3}}^{\tau_{0}}}(\mathcal{O}^{\tau_{0}},\Omega_{0}^{{\bf k}_{\tau_{0},n_{3}}}(U))^{\leq\underline{h}}.

Thus, for any choice of horizontal basis wτ0w^{\tau_{0}} we obtain eh¯t(μ1,μ2)Dn3𝐤n3τ0(𝒪τ0,Ω0𝐤τ0,n3(U))h¯e_{\leq\underline{h}}t(\mu_{1},\mu_{2})\in D_{n_{3}}^{{\bf k}_{n_{3}}^{\tau_{0}}}(\mathcal{O}^{\tau_{0}},\Omega_{0}^{{\bf k}_{\tau_{0},n_{3}}}(U))^{\leq\underline{h}}, satisfying all the properties (B1), (B2), (B3) of §4.5.2. This implies that we obtain a global section

t(μ1,μ2)H0(𝒳b,n1,n2,n31,Ω𝐤n3)h¯,𝒳b,n1,n2,n31:=𝒳b×𝒲n11×𝒲n21×𝒲n3,t(\mu_{1},\mu_{2})\in H^{0}(\mathcal{X}_{b,n_{1},n_{2},n_{3}}^{1},\Omega^{{\bf k}_{n_{3}}})^{\leq\underline{h}},\qquad\mathcal{X}_{b,n_{1},n_{2},n_{3}}^{1}:=\mathcal{X}_{b}\times\mathcal{W}_{n_{1}}^{1}\times\mathcal{W}_{n_{2}}^{1}\times\mathcal{W}_{n_{3}},

where h¯#{𝔭p}\underline{h}\in\operatorname{\mathbb{N}}^{\#\{\mathfrak{p}\mid p\}} now stands for the finite slopes of the different U𝔭U_{\mathfrak{p}}. This construction defines a triple product

(31) t:H0(𝒳r,n11,Ω𝐤n1)×H0(𝒳r,n21,Ω𝐤n2)H0(𝒳b,n1,n2,n31,Ω𝐤n3)h¯ΛIm.t:H^{0}(\mathcal{X}_{r,n_{1}}^{1},\Omega^{{\bf k}_{n_{1}}})\times H^{0}(\mathcal{X}_{r,n_{2}}^{1},\Omega^{{\bf k}_{n_{2}}})\longrightarrow H^{0}(\mathcal{X}_{b,n_{1},n_{2},n_{3}}^{1},\Omega^{{\bf k}_{n_{3}}})^{\leq\underline{h}}\otimes\Lambda_{I_{m}}.

6.8. Triple product pp-adic L-functions: Finite slope case

As in the previous section, let us denote by (𝐫n,νn):𝒪××p×Λn1({\bf r}_{n},{\bf\nu}_{n}):\mathcal{O}^{\times}\times\operatorname{\mathbb{Z}}_{p}^{\times}\rightarrow\Lambda_{n}^{1} the universal character of 𝔚n1\mathfrak{W}^{1}_{n}, and we put 𝐤n:=k(𝐫n,νn){\bf k}_{n}:=k({\bf r}_{n},{\bf\nu}_{n}). Similarly, we denote by (𝐫,ν):𝒪××p×ΛF1({\bf r},{\bf\nu}):\mathcal{O}^{\times}\times\operatorname{\mathbb{Z}}_{p}^{\times}\rightarrow\Lambda_{F}^{1} the universal character of 𝔚F1\mathfrak{W}^{1}_{F}, and 𝐤:=k(𝐫,ν){\bf k}:=k({\bf r},{\bf\nu}). In [3, §7.2] it is described an action of the finite group Δ:=(𝒪F)+×/{u(𝒪F)+×,u1mod𝔫}\Delta:=(\mathcal{O}_{F})^{\times}_{+}/\{u\in(\mathcal{O}_{F})_{+}^{\times},\;u\equiv 1\;{\rm mod}\;\mathfrak{n}\} on the space H0(𝔛r×𝔚n1,Ω𝐤n)H^{0}(\mathfrak{X}_{r}\times\mathfrak{W}_{n}^{1},\Omega^{{\bf k}_{n}}): for any μH0(𝔛r×𝔚n1,Ω𝐤n)\mu\in H^{0}(\mathfrak{X}_{r}\times\mathfrak{W}_{n}^{1},\Omega^{{\bf k}_{n}}) and s(𝒪F)+×s\in(\mathcal{O}_{F})^{\times}_{+},

([s]μ)(A,ı,θ,α𝔭0,w):=𝐫n(sτ0(s)2)νn(τ0s)μ(A,ı,s1θ,γs1α𝔭0,w),([s]\ast\mu)(A,\imath,\theta,\alpha^{\mathfrak{p}_{0}},w):={\bf r}_{n}(s\tau_{0}(s)^{-2})\cdot\nu_{n}(\tau_{0}s)\cdot\mu(A,\imath,s^{-1}\theta,\gamma_{s}^{-1}\alpha^{\mathfrak{p}_{0}},w),

where γsK1(𝔫,p)\gamma_{s}\in K_{1}(\mathfrak{n},p) and detγs=s\det\gamma_{s}=s. Let us consider the space

M𝐤nr(Γ1(𝔫,p),Λn1):=𝔠Pic(𝒪F)H0(𝔛r𝔠×𝔚n1,Ω𝐤n)Δ,M^{r}_{{\bf k}_{n}}(\Gamma_{1}(\mathfrak{n},p),\Lambda_{n}^{1}):=\bigoplus_{\mathfrak{c}\in{\rm Pic}(\mathcal{O}_{F})}H^{0}(\mathfrak{X}_{r}^{\mathfrak{c}}\times\mathfrak{W}_{n}^{1},\Omega^{{\bf k}_{n}})^{\Delta},

where 𝔛r𝔠\mathfrak{X}_{r}^{\mathfrak{c}} is the overconvergent neighborhood of the unitary Shimura curve associated with 𝔠\mathfrak{c} defined as in [3, Remark 4.3]. Then the ΛF1\Lambda_{F}^{1}-modules

M𝐤(Γ1(𝔫,p),ΛF1):=limr,nM𝐤nr(Γ1(𝔫,p),Λn1),M𝐤(Γ1(𝔫,p),ΛF1)h¯:=limr,nM𝐤nr(Γ1(𝔫,p),Λn1)h¯,M^{\dagger}_{{\bf k}}(\Gamma_{1}(\mathfrak{n},p),\Lambda_{F}^{1}):=\varprojlim_{r,n}M^{r}_{{\bf k}_{n}}(\Gamma_{1}(\mathfrak{n},p),\Lambda_{n}^{1}),\qquad M^{\dagger}_{{\bf k}}(\Gamma_{1}(\mathfrak{n},p),\Lambda_{F}^{1})^{\leq\underline{h}}:=\varprojlim_{r,n}M^{r}_{{\bf k}_{n}}(\Gamma_{1}(\mathfrak{n},p),\Lambda_{n}^{1})^{\leq\underline{h}},

are the spaces of overconvergent families of automorphic forms. By [3, Proposition 7.8] the specialization of such families at sufficiently big classical points provide classical automorphic forms.

Let μ3M𝐤(Γ1(𝔫,p),ΛF1)h¯\mu_{3}\in M^{\dagger}_{{\bf k}}(\Gamma_{1}(\mathfrak{n},p),\Lambda_{F}^{1})^{\leq\underline{h}} be a eigenvector for the Hecke operators not dividing 𝔫\mathfrak{n} and such that U𝔭μ=α3𝔭μU_{\mathfrak{p}}\mu=\alpha_{3}^{\mathfrak{p}}\mu. Let us also consider μ1,μ2M𝐤(Γ1(𝔫,p),ΛF1)\mu_{1},\mu_{2}\in M^{\dagger}_{{\bf k}}(\Gamma_{1}(\mathfrak{n},p),\Lambda_{F}^{1}).

Let =ΛF1^ΛF1^ΛF\mathcal{R}^{\prime}=\Lambda_{F}^{1}\hat{\otimes}\Lambda_{F}^{1}\hat{\otimes}\Lambda_{F}^{\prime}, where ΛF\Lambda_{F}^{\prime} is the fraction field of ΛF\Lambda_{F}, thus \mathcal{R}^{\prime} can be viewed as rational functions on 𝔚1×𝔚1×𝔚\mathfrak{W}^{1}\times\mathfrak{W}^{1}\times\mathfrak{W} with poles at finitely many weights in 𝔚\mathfrak{W}. Write 𝐤1{\bf k}_{1}, 𝐤2{\bf k}_{2} and 𝐤3{\bf k}_{3} for the universal characters corresponding each copy of 𝔚1\mathfrak{W}^{1} and 𝔚\mathfrak{W}.

In the rest we use the following notation: If (x,y,z)𝔚1×𝔚1×𝔚(x,y,z)\in\mathfrak{W}^{1}\times\mathfrak{W}^{1}\times\mathfrak{W} we denote by μx\mu_{x}, μy\mu_{y} and μz\mu_{z} the specializations of the families μ1\mu_{1} at xx, μ2\mu_{2} at yy, and μ3\mu_{3} at zz respectively. If z𝔚z\in\mathfrak{W} is a classical weight then the specialization μz\mu_{z} is a classical automorphic form of weight kzk_{z}, where kzk_{z} is the specialization of 𝐤3{\bf k}_{3} at zz. Recall the triple product of (31) defines a triple product

t:M𝐤1(Γ1(𝔫,p),ΛF1)×M𝐤2(Γ1(𝔫,p),ΛF1)M𝐤3(Γ1(𝔫,p),)h¯.t:M^{\dagger}_{{\bf k}_{1}}(\Gamma_{1}(\mathfrak{n},p),\Lambda_{F}^{1})\times M^{\dagger}_{{\bf k}_{2}}(\Gamma_{1}(\mathfrak{n},p),\Lambda_{F}^{1})\longrightarrow M^{\dagger}_{{\bf k}_{3}}(\Gamma_{1}(\mathfrak{n},p),\mathcal{R}^{\prime})^{\leq\underline{h}}.

The following result is analogous to [8, Lemma 2.19] and [3, Lemma 10.3].

Lemma 6.8.

There exists p(μ1,μ2,μ3)\mathcal{L}_{p}(\mu_{1},\mu_{2},\mu_{3})\in\mathcal{R}^{\prime} such that for each classical point (x,y,z)𝔚1×𝔚1×𝔚(x,y,z)\in\mathfrak{W}^{1}\times\mathfrak{W}^{1}\times\mathfrak{W}, we have:

p(μ1,μ2,μ3)(x,y,z)=μz,t(μ1,μ2)(x,y,z)μz,μz\mathcal{L}_{p}(\mu_{1},\mu_{2},\mu_{3})(x,y,z)=\frac{\left\langle\mu_{z}^{*},t(\mu_{1},\mu_{2})_{(x,y,z)}\right\rangle}{\left\langle\mu_{z}^{*},\mu_{z}\right\rangle}

where ,\langle\cdot,\cdot\rangle is the Petersson inner product and μz\mu_{z}^{*} defines the dual basis of μz\mu_{z}.

Definition 6.9.

Let μ1,μ2,μ3\mu_{1},\mu_{2},\mu_{3} as above and assume that they are test vectors for three families of eigenvectors. The functions p(μ1,μ2,μ3)\mathcal{L}_{p}(\mu_{1},\mu_{2},\mu_{3})\in\mathcal{R}^{\prime} introduced in 6.8 is called the triple product pp-adic L-function of μ1,μ2,μ3\mu_{1},\mu_{2},\mu_{3}.

6.9. Interpolation formula

Let (r¯1,ν1)𝔚1(p)(\underline{r}_{1},\nu_{1})\in\mathfrak{W}^{1}(\operatorname{\mathbb{C}}_{p}), (r¯2,ν2)𝔚1(p)(\underline{r}_{2},\nu_{2})\in\mathfrak{W}^{1}(\operatorname{\mathbb{C}}_{p}) and r¯𝔚(p)\underline{r}\in\mathfrak{W}(\operatorname{\mathbb{C}}_{p}) be classical weights and put k¯1=k(r¯1,ν1),k¯2=k(r¯2,ν2)\underline{k}_{1}=k(\underline{r}_{1},\nu_{1}),\underline{k}_{2}=k(\underline{r}_{2},\nu_{2}) and k¯3=k(r¯3,ν1+ν2)\underline{k}_{3}=k(\underline{r}_{3},\nu_{1}+\nu_{2}). We suppose that (k¯1,k¯2,k¯3)(\underline{k}_{1},\underline{k}_{2},\underline{k}_{3}) is unbalanced at τ0\tau_{0} with dominant weight k¯3\underline{k}_{3} in the sense of [3, Definition 3.4].

We write (x,y,z)𝔚1×𝔚1×𝔚(x,y,z)\in\mathfrak{W}^{1}\times\mathfrak{W}^{1}\times\mathfrak{W} for the point corresponding to the triple (k¯1,ν1)(\underline{k}_{1},\nu_{1}), (k¯2,ν2)(\underline{k}_{2},\nu_{2}) and (k¯3,ν1+ν2)(\underline{k}_{3},\nu_{1}+\nu_{2}). Since μ3\mu_{3} has finite slope, its specialization μz\mu_{z} correspond to an automorphic form of weight (k¯3,ν1+ν2)(\underline{k}_{3},\nu_{1}+\nu_{2}). If k¯1\underline{k}_{1} and k¯2\underline{k}_{2} are big enough the the same is true for μx\mu_{x} and μy\mu_{y}, obtaining automorphic forms of weights (k¯1,ν1)(\underline{k}_{1},\nu_{1}) and (k¯2,ν2)(\underline{k}_{2},\nu_{2}) respectively. We denote by πx\pi_{x}, πy\pi_{y} and πz\pi_{z} the automorphic representations of (B𝔸F)×(B\otimes\operatorname{\mathbb{A}}_{F})^{\times} generated by these automorphic forms, and Πx\Pi_{x}, Πy\Pi_{y} and Πz\Pi_{z} the corresponding cuspidal automorphic representations of GL2(𝔸F)\mathrm{GL}_{2}(\operatorname{\mathbb{A}}_{F}).

Assume that μi\mu_{i} are eigenvectors for all the U𝔭U_{\mathfrak{p}} operators, namely U𝔭μi=αi𝔭μiU_{\mathfrak{p}}\mu_{i}=\alpha_{i}^{\mathfrak{p}}\cdot\mu_{i}, and write αx𝔭\alpha_{x}^{\mathfrak{p}}, αy𝔭\alpha_{y}^{\mathfrak{p}} and αz𝔭\alpha_{z}^{\mathfrak{p}} for the corresponding specializations at xx, yy and zz. Moreover, we assume that μx\mu_{x} is the test vector defined in [13] of the 𝔭\mathfrak{p}-stabilization of the newform μx\mu_{x}^{\circ} for each 𝔭p\mathfrak{p}\mid p, and similarly for μy\mu_{y} and μz\mu_{z}. Write βi𝔭\beta_{i}^{\mathfrak{p}} for the other eigenvalue of U𝔭U_{\mathfrak{p}} as usual, and write βx𝔭\beta_{x}^{\mathfrak{p}}, βy𝔭\beta_{y}^{\mathfrak{p}} and βz𝔭\beta_{z}^{\mathfrak{p}} for the corresponding specializations. The following result justify the name given to p(μ1,μ2,μ3)\mathcal{L}_{p}(\mu_{1},\mu_{2},\mu_{3}).

Theorem 6.10.

With the notations above we have:

p(μ1,μ2,μ3)(x,y,z)=K(μx,μy,μz)(𝔭p𝔭(x,y,z)𝔭,1(z))L(1ν1ν2ν32,ΠxΠyΠz)12μz,μz\mathcal{L}_{p}(\mu_{1},\mu_{2},\mu_{3})(x,y,z)=K(\mu_{x}^{\circ},\mu_{y}^{\circ},\mu_{z}^{\circ})\cdot\left(\prod_{\mathfrak{p}\mid p}\frac{\mathcal{E}_{\mathfrak{p}}(x,y,z)}{\mathcal{E}_{\mathfrak{p},1}(z)}\right)\cdot\frac{L\left(\frac{1-\nu_{1}-\nu_{2}-\nu_{3}}{2},\Pi_{x}\otimes\Pi_{y}\otimes\Pi_{z}\right)^{\frac{1}{2}}}{\langle\mu_{z}^{\circ},\mu_{z}^{\circ}\rangle}

here K(μ1,μ2,μ3)K(\mu_{1}^{\circ},\mu_{2}^{\circ},\mu_{3}^{\circ}) is a non-zero constant, 𝔭(x,y,z)=\mathcal{E}_{\mathfrak{p}}(x,y,z)=

{(1βx𝔭βy𝔭αz𝔭ϖ𝔭m¯𝔭2¯)(1αx𝔭βy𝔭βz𝔭ϖ𝔭m¯𝔭2¯)(1βx𝔭αy𝔭βz𝔭ϖ𝔭m¯𝔭2¯)(1βx𝔭βy𝔭βz𝔭ϖ𝔭m¯𝔭2¯),𝔭𝔭0(1αx𝔭0αy𝔭0βz𝔭0p1m0)(1αx𝔭0βy𝔭0βz𝔭0p1m0)(1βx𝔭0αy𝔭0βz𝔭0p1m0)(1βx𝔭0βy𝔭0βz𝔭0p1m0),𝔭=𝔭0,\left\{\begin{array}[]{lc}\mbox{\small$(1-\beta_{x}^{\mathfrak{p}}\beta_{y}^{\mathfrak{p}}\alpha_{z}^{\mathfrak{p}}\varpi_{\mathfrak{p}}^{-\underline{m}_{\mathfrak{p}}-\underline{2}})(1-\alpha_{x}^{\mathfrak{p}}\beta_{y}^{\mathfrak{p}}\beta_{z}^{\mathfrak{p}}\varpi_{\mathfrak{p}}^{-\underline{m}_{\mathfrak{p}}-\underline{2}})(1-\beta_{x}^{\mathfrak{p}}\alpha_{y}^{\mathfrak{p}}\beta_{z}^{\mathfrak{p}}\varpi_{\mathfrak{p}}^{-\underline{m}_{\mathfrak{p}}-\underline{2}})(1-\beta_{x}^{\mathfrak{p}}\beta_{y}^{\mathfrak{p}}\beta_{z}^{\mathfrak{p}}\varpi_{\mathfrak{p}}^{-\underline{m}_{\mathfrak{p}}-\underline{2}})$},&\mathfrak{p}\neq\mathfrak{p}_{0}\\ \mbox{\small$(1-\alpha_{x}^{\mathfrak{p}_{0}}\alpha_{y}^{\mathfrak{p}_{0}}\beta_{z}^{\mathfrak{p}_{0}}p^{1-m_{0}})(1-\alpha_{x}^{\mathfrak{p}_{0}}\beta_{y}^{\mathfrak{p}_{0}}\beta_{z}^{\mathfrak{p}_{0}}p^{1-m_{0}})(1-\beta_{x}^{\mathfrak{p}_{0}}\alpha_{y}^{\mathfrak{p}_{0}}\beta_{z}^{\mathfrak{p}_{0}}p^{1-m_{0}})(1-\beta_{x}^{\mathfrak{p}_{0}}\beta_{y}^{\mathfrak{p}_{0}}\beta_{z}^{\mathfrak{p}_{0}}p^{1-m_{0}})$},&\mathfrak{p}=\mathfrak{p}_{0}\end{array}\right.,
𝔭,1(z):={(1(βz𝔭)2ϖ𝔭k¯3,𝔭2¯)(1(βz𝔭)2ϖ𝔭k¯3,𝔭1¯),𝔭𝔭0,(1(βz𝔭0)2pk3,τ0)(1(βz𝔭0)2p1k3,τ0),𝔭=𝔭0,\mathcal{E}_{\mathfrak{p},1}(z):=\left\{\begin{array}[]{lc}(1-(\beta_{z}^{\mathfrak{p}})^{2}\varpi_{\mathfrak{p}}^{-\underline{k}_{3,\mathfrak{p}}-\underline{2}})\cdot(1-(\beta_{z}^{\mathfrak{p}})^{2}\varpi_{\mathfrak{p}}^{-\underline{k}_{3,\mathfrak{p}}-\underline{1}}),&\mathfrak{p}\neq\mathfrak{p}_{0},\\ (1-(\beta_{z}^{\mathfrak{p}_{0}})^{2}p^{-k_{3,\tau_{0}}})\cdot(1-(\beta_{z}^{\mathfrak{p}_{0}})^{2}p^{1-k_{3,\tau_{0}}}),&\mathfrak{p}=\mathfrak{p}_{0},\end{array}\right.

m0=k1,τ0+k2,τ0+k3,τ020m_{0}=\frac{k_{1,\tau_{0}}+k_{2,\tau_{0}}+k_{3,\tau_{0}}}{2}\geq 0, and m¯𝔭=k¯1,𝔭+k¯2,𝔭+k¯3,𝔭2[Σ𝔭]\underline{m}_{\mathfrak{p}}=\frac{\underline{k}_{1,\mathfrak{p}}+\underline{k}_{2,\mathfrak{p}}+\underline{k}_{3,\mathfrak{p}}}{2}\in\operatorname{\mathbb{Z}}[\Sigma_{\mathfrak{p}}].

Proof.

By Equation (30) and Lemma 6.1 we have:

p(μ1,μ2,μ3)(x,y,z)\displaystyle\mathcal{L}_{p}(\mu_{1},\mu_{2},\mu_{3})(x,y,z) =\displaystyle= μz,t(μ1,μ2)(x,y,z)μz,μz=μz,ehΘm3,τ0(μx[p],μy)(Δ¯(x,y,z)τ0)μz,μz\displaystyle\frac{\left\langle\mu_{z}^{*},t(\mu_{1},\mu_{2})_{(x,y,z)}\right\rangle}{\left\langle\mu_{z}^{*},\mu_{z}\right\rangle}=\frac{\left\langle\mu_{z}^{*},e_{\leq h}\Theta_{m_{3,\tau_{0}}}(\mu_{x}^{[p]},\mu_{y})(\underline{\Delta}_{(x,y,z)}^{\tau_{0}})\right\rangle}{\left\langle\mu_{z}^{*},\mu_{z}\right\rangle}
=\displaystyle= (1)m3,τ0(k3,τ02m3,τ0+k2,τ01)1μz,tk1,τ0,k2,τ0,k3,τ0(μx[p],μy)(Δ¯(x,y,z)τ0)μz,μz.\displaystyle(-1)^{m_{3,\tau_{0}}}\binom{k_{3,\tau_{0}}-2}{m_{3,\tau_{0}}+k_{2,\tau_{0}}-1}^{-1}\frac{\left\langle\mu_{z}^{*},t_{k_{1,\tau_{0}},k_{2,\tau_{0}},k_{3,\tau_{0}}}(\mu_{x}^{[p]},\mu_{y})(\underline{\Delta}_{(x,y,z)}^{\tau_{0}})\right\rangle}{\left\langle\mu_{z}^{*},\mu_{z}\right\rangle}.

At this point the proof of this result is identical to that of [3, Theorem 10.5]. ∎


Santiago Molina; Universitat Politècnica de Catalunya
Campus Nord, Calle Jordi Girona, 1-3, 08034 Barcelona, Spain

santiago-molina@upc.edu

References

  • [1] Fabrizio Andreatta and Adrian Iovita. Triple product pp-adic ll-functions associated to finite slope pp-adic families of modular forms. 2017. Preprint.
  • [2] Fabrizio Andreatta, Adrian Iovita, and Vincent Pilloni. Spectral halo. Annales scientifiques de l’École Normale Supérieure, 2017.
  • [3] Daniel Barrera and Santiago Molina. Triple product pp-adic l-functions for shimura curves over totally real number fields. Preprint, https://arxiv.org/abs/1908.00091.
  • [4] Massimo Bertolini, Henri Darmon, and Victor Rotger. Beilinson-Flach elements and Euler systems II: the Birch and Swinnerton-Dyer conjecture for Hasse-Weil LL-series. Journal of Algebraic Geometry, 24:569–604, 2015.
  • [5] Massimo Bertolini, Marco Seveso, and Rodolfo Venerucci. Diagonal classes and the Bloch-Kato conjecture. Munster Journal of Mathematics, Special volume in honor of Christophe.
  • [6] Riccardo Brasca. pp-adic modular forms of non integral weight over Shimura curves. Compos. Math., 149(1):32–62, 2013.
  • [7] Henri Carayol. Sur la mauvaise réduction des courbes de Shimura. Compos. Math., 59:151–230, 1986.
  • [8] Henri Darmon and Victor Rotger. Diagonal cycles and euler systems I: A pp-adic Gross-Zagier formula. Annales scientifiques de l’École Normale Supérieure, 47(4):779–832, 2014.
  • [9] Henri Darmon and Victor Rotger. Diagonal cycles and euler systems II: the Birch and Swinnerton-Dyer conjecture for Hasse-Weil-Artin LL-series. Journal of the AMS, 30(3):601–672, 2017.
  • [10] Yiwen Ding. Formes modulaires pp-adiques sur les courbes de Shimura unitaires et compatibilité local-global. Mémoires de la Soc. Math. France, 155, 2017.
  • [11] Giacomo Graziani. Modular sheaves of de rham classes on hilbert formal modular schemes for unramified primes. https://arxiv.org/pdf/2007.15997.pdf.
  • [12] Michael Harris and Jacques Tilouine. pp-adic measures and square roots of special values of triple product LL-functions. Math. Annalen, 320:127–147, 2001.
  • [13] Ming-Lun Hsieh. Hida families and pp-adic triple product LL-functions. American Journal of Mathematics, 2018.
  • [14] Payman Kassaei. 𝒫\mathcal{P}-adic modular forms over Shimura curves over totally real fields. Compos. Math., 140:359–395, 2004.
  • [15] N. Katz. Serre-tate local moduli. In Jean Giraud, Luc Illusie, and Michel Raynaud, editors, Surfaces Algébriques, pages 138–202, Berlin, Heidelberg, 1981. Springer Berlin Heidelberg.
  • [16] Guido Kings, David Loeffler, and Sarah Zerbes. Rankin–Eisenstein classes and explicit reciprocity laws. Cambridge J. Math., 5(1), 2017.
  • [17] Antonio Lei, David Loeffler, and Sarah Zerbes. Euler systems for Hilbert modular surfaces. 2016. Preprint.
  • [18] David Loeffler and Sarah Zerbes. Rankin–Eisenstein classes in Coleman families. Res. Math. Sci., 29 (Robert Coleman memorial volume)(3), 2016.