betaFeedback

WorksMathematics

On the ring of bounded functions on the character variety

Abstract

Let X\mathfrak{X} denote the character variety associated to the group (oL,+)(o_L, +), where oLo_L is the ring of integers in a finite extension L/QpL/\mathbb{Q}_p. X\mathfrak{X} is a rigid analytic variety parametrizing locally LL-analytic characters. Let ΛL(X)Λ_L\left(\mathfrak{X}\right) denote the global sections bounded by 11. In this paper we prove that ΛL(X)=oL[[oL]]Λ_L\left(\mathfrak{X}\right)= o_L[[o_L]], answering the question posed in [AB24].

Introduction

Let LL be a finite extension of Qp\mathbb{Q}_p with completed algebraic closure Cp\mathbb{C}_p. The character variety X\mathcal{X} was first defined by Schneider and Teitelbaum [8] in the context of pp-adic Fourier theory. X\mathcal{X} is a rigid analytic variety defined over LL whose set of KK-points X(K)\mathcal{X}(K), for L≤K≤CpL \leq K \leq\mathbb{C}_p complete, is the set of locally LL-analytic characters χ:(OL,+)→(K×,⋅)\chi:(\mathcal{O}_L,+)\to(K^\times,\cdot).

The character variety appears in the approach of Berger, Schneider and Xie to the Lubin-Tate generalization of cyclotomic (φ,Γ)(\varphi,\Gamma)-modules in [5]. This approach is motivated by the possibility of generalizing the pp-adic local Langlands correspondence from GL2(Qp)\mathrm{GL}_2(\mathbb{Q}_p) to GL2(L)\mathrm{GL}_2(L). One of the reasons that X\mathcal{X} appears is that the ring of global bounded functions, denoted Ob(X)\mathcal{O}^{\mathrm{b}}(\mathcal{X}), has a natural action of the monoid OL−{0}\mathcal{O}_L-\{0\} analogous to that in the classical setting. Lubin-Tate analogues of cyclotomic (φ,Γ)(\varphi,\Gamma)-modules are defined over the rings that we get by localizing and completing Ob(X)\mathcal{O}^{b}(\mathcal{X}).

In this paper we consider the subring ΛL(X)≤Ob(X)\Lambda_L(\mathcal{X}) \leq\mathcal{O}^{b}(\mathcal{X}) consisting of global functions bounded by 11 on the character variety. Ardakov and Berger in [1] study, among other things, the question whether every distribution bounded by 11 comes from an integral measure, or equivalently, whether ΛL(X)=oL[[oL]]\Lambda_L(\mathcal{X}) = o_L[[o_L]], see [1 Question 1.1.1]. Here oL[[oL]]o_L[[o_L]] denotes the completed group ring and the map oL[[oL]]⟶ΛL(X)o_L[[o_L]] \longrightarrow\Lambda_L(\mathcal{X}) is given by integration against the measure.

In [1] several different criteria that imply ΛL(X)=oL[[oL]]\Lambda_L(\mathcal{X}) = o_L[[o_L]] are proved. The question was further studied in the quadratic unramified case by Berger and Sprang in [4]. In this paper we prove

Theorem 1.1. It holds that ΛL(X)=oL[[oL]]\Lambda_L(\mathcal{X}) = o_L[[o_L]].

In independent work, Berger, Sprang and Steingart prove the same result for quadratic extensions of Qp\mathbb{Q}_p using different methods.

The starting point of the proof of Theorem 1.1 is a reduction to the integrality of a trace operator by using [1 Theorem 3.1.9], see Lemma 3.1. We prove the integrality of this operator using a version of analytic boundary results from [2] in combination with a modification of Tate’s classical result on absolute values of traces in deeply ramified field extensions given in Lemma 3.6. A key step in the proof is the connection between partial trace operators, which are used to factorize the trace operator from [1 Theorem 3.1.9], and a normalized arithmetic trace, see Proposition 3.4. For more details, see the sketch of the proof of Theorem 1.1 in § 2.6.

Theorem 1.1 has the following interesting corollary for integer valued polynomials, see [1 Theorem 4.2.21]. Let Int⁡(oL,oL)\operatorname{Int}(o_L,o_L) denote the ring of integer valued polynomials on oLo_L. Denote the formal Lubin-Tate action1 by [a]X[a]X. Recall from [1 Section 1.5] the definition Pol⁡(oL[[X]])=∑i,j≥0oLσi,j\operatorname{Pol}(o_L[[X]]) = \sum_{i,j\geq0} o_L\sigma_{i,j} where ([a]X)σi,j=∑jσi,j(a)Xj([a]X)^{\sigma_{i,j}} = \sum_j \sigma_{i,j}(a)X^j. Here σi,j∈Int⁡(oL,oL)\sigma_{i,j} \in\operatorname{Int}(o_L,o_L).

Corollary 1.2. The following equality holds: Pol⁡(oL[[X]])=Int⁡(oL,oL)\operatorname{Pol}(o_L[[X]]) = \operatorname{Int}(o_L,o_L).

Related results were studied in [7 Tat99], where it was proven that the space of coefficients {σi,j}i,j≥0\{\sigma_{i,j}\}_{i,j\geq0} generates Int⁡(oL,oL)\operatorname{Int}(o_L,o_L) as an algebra over oLo_L. Theorem B from [4] implies that Pol⁡(oL[[X]])\operatorname{Pol}(o_L[[X]]) is pp-adically dense in Int⁡(oL,oL)\operatorname{Int}(o_L,o_L). Corollary 1.2 strengthens these results by showing that {σi,j}i,j≥0\{\sigma_{i,j}\}_{i,j\geq0} generate all integer valued polynomials as an oLo_L-module.

Note that when L=QpL = \mathbb{Q}_p, the Lubin-Tate formal group can be taken to be the formal multiplicative group G^m\widehat{\mathbb{G}}_m. In this case Corollary 1.2 recovers the classical fact that all integer valued polynomials over Qp\mathbb{Q}_p are Zp\mathbb{Z}_p-linear combinations of binomial coefficients {(xn)}n≥0\left\{\binom{x}{n}\right\}_{n\geq0}. In this regard, Corollary 1.2 can be interpreted as a statement that the “generalized” binomial coefficients span all integer valued polynomials.

Remark 1.3. In [6 Remark 2.5], Colmez showed that a locally LL-analytic Cp\mathbb{C}_p-valued distribution on oLo_L with bounded Fourier transform has order at most d−1d-1, where [L:Qp]=d[L:\mathbb{Q}_p] = d. If such a distribution λ\lambda is defined over LL, Theorem 1.1 implies that λ\lambda has order 00.

Structure of the paper

1.1. Structure of the paper. Section 2 collects the necessary background and provides a sketch of the proof of Theorem 1.1. Section 3 contains the proof of Theorem 1.1. Section 4 contains the proof of a result about unions of fibers being an analytic boundary. Section 5 contains the proof of our reformulation of a classical result of Tate about the behaviour of normalised traces in deeply ramified extensions.

Acknowledgements

I want to thank Konstantin Ardakov for many valuable discussions, interesting insights and invaluable guidance and support. I want to thank Laurent Berger and Andrew Graham for the discussion related to the question. I want to thank Konstantin Ardakov, Laurent Berger, James Newton, Branislav Prvulović and Vukašin Stojisavljević for reading the first draft of the paper.

AI Declaration

AI assistance consisted of suggesting the relation with the arithmetic trace and pointing to the operators ψi\psi_i. No AI was used during the writing of the paper, but it was used to check the proofs and review the literature.

Notation and background

Fix a prime pp. Let CpC_p denote the completed algebraic closure of Qp\mathbb{Q}_p. Let the norm on CpC_p be normalized such that ∣p∣=p−1|p| = p^{-1}. Let L/QpL/\mathbb{Q}_p be a finite extension of degree dd with ramification index ee and a uniformizer π\pi. Denote by qq the cardinality of the residue field. Let {μpn}n=1∞\{\mu_{p^n}\}_{n=1}^{\infty} be a system of primitive pnp^n-th roots of unity and denote Qp,n=Qp(μpn)\mathbb{Q}_{p,n} = \mathbb{Q}_p(\mu_{p^n}), Ln=L(μpn)L_n = L(\mu_{p^n}), L∞=lim→⁡nLnL_\infty= \varinjlim_n L_n.

The character variety

The character variety X\mathcal{X} is a rigid analytic variety defined over LL whose set of KK-points X(K)\mathcal{X}(K), for L≤K≤CpL \leq K \leq C_p complete, is the set of locally LL-analytic characters χ:(oL,+)→(K×,⋅)\chi: (o_L,+) \to(K^\times,\cdot). Denote the global sections on X\mathcal{X} by O(X)\mathcal{O}(\mathcal{X}), the bounded global sections by Ob(X)\mathcal{O}^{b}(\mathcal{X}), and the power-bounded global sections by ΛL(X)\Lambda_L(\mathcal{X}). For the precise definition of the character variety see [8 Section 2].

The set X(Cp)\mathcal{X}(C_p) is a group under pointwise multiplication of characters and we denote this multiplication by ξ⋅η\xi\cdot\eta for ξ,η∈X(Cp)\xi,\eta\in\mathcal{X}(C_p). For ξ∈X(Cp)\xi\in\mathcal{X}(C_p) we will use additive notation kξk\xi to mean ξk\xi^k. There is an action of oLo_L on X(Cp)\mathcal{X}(C_p) defined by aζ(x)=ζ(ax)a\zeta(x) = \zeta(ax). Let a∈oLa \in o_L and denote the aa-torsion under this action by X[a]\mathcal{X}[a]. Define X[p∞]=lim→⁡mX[pm]\mathcal{X}[p^\infty] = \varinjlim_m \mathcal{X}[p^m] and let X[π>M]\mathcal{X}[\pi^{>M}] denote the set of characters in X[p∞]\mathcal{X}[p^\infty] whose π\pi-order is bigger than MM.

Remark 2.1. There exists a noncanonical isomorphism X[p∞]≅L/oL\mathcal{X}[p^\infty] \cong L/o_L as oLo_L-modules. One can use [5 Lemma 2.13], or argue directly as follows. Take a family of characters {ξm}m=0∞\{\xi_m\}_{m=0}^{\infty} such that πξm+1=ξm\pi\xi_{m+1} = \xi_m and ξm\xi_m has π\pi-order equal to mm, and consider the family of oLo_L-module isomorphisms θm:oL/πmoL→X[πm]\theta_m : o_L/\pi^m o_L \to\mathcal{X}[\pi^m] defined by sending a↦aξma \mapsto a\xi_m. These maps are compatible, so we can take the inductive limit to get an oLo_L-module isomorphism θ=lim→⁡mθm:L/oL→X[p∞]\theta= \varinjlim_m \theta_m : L/o_L \to\mathcal{X}[p^\infty]. Note that we view L/oL=lim→⁡moL/πmL/o_L = \varinjlim_m o_L/\pi^m where the transition maps are given by multiplication by π\pi.

p-adic Fourier theory and Lubin-Tate isomorphism

Let D(oL,L)\mathrm{D}(o_L,L) denote the set of locally LL-analytic distributions on the additive group oLo_L. These carry an action of oL∖{0}o_L \setminus\{0\} defined by (a∗λ)(f(x))=λ(f(ax))(a_*\lambda)(f(x)) = \lambda(f(ax)). One of the main results of [8 Theorem 2.3] is an isomorphism

D(oL,L)≅O(X).(1)\mathrm{D}(o_L,L) \cong\mathcal{O}(\mathcal{X}). \tag*{(1)}

This enables us to work with distributions, and we will be using this isomorphism implicitly2. Note that if λ∈D(oL,L)\lambda\in\mathrm{D}(o_L,L), we can evaluate λ\lambda at a locally LL-analytic CpC_p-valued function on oLo_L.

Let Db(oL,L)\mathrm{D}^{b}(o_L,L) denote the distributions corresponding to Ob(X)\mathcal{O}^{b}(\mathcal{X}) under this isomorphism. We call these bounded distributions. There is a multiplicative norm on Db(oL,L)\mathrm{D}^{b}(o_L,L) given by

∣λ∣=sup⁡ξ∈X(Cp)∣λ(ξ)∣.|\lambda|=\sup_{\xi\in\mathcal{X}(\mathbb{C}_p)}|\lambda(\xi)|.

Distributions that have norm at most one under the isomorphism (1) correspond to ΛL(X)\Lambda_L(\mathcal{X}). For more details see [5 p. 3 Remark, Corollary 1.23].

Let LT\mathrm{LT} denote a Lubin-Tate group associated to π∈oL\pi\in o_L, and let B\mathbb{B} denote the open unit ball over Cp\mathbb{C}_p. Denote the base change of the character variety to Cp\mathbb{C}_p by XCp\mathcal{X}_{\mathbb{C}_p}. In [8 Theorem 3.6] it is proved that

B≅XCp.\mathbb{B}\cong\mathcal{X}_{\mathbb{C}_p}.

Operators φ\varphi and ψ\psi

We recall the operators φ,ψq\varphi,\psi_q used in [1], and transport them to distributions via pp-adic Fourier theory and the LT isomorphism. Define the operator φ:O(B)⟶O(B)\varphi:\mathcal{O}(\mathbb{B})\longrightarrow\mathcal{O}(\mathbb{B}) by Z↦[π]ZZ\mapsto[\pi]Z, where [π][\pi] denotes the Lubin-Tate action. This is a finite locally free map of constant rank, so we can define the normalized trace ψq:O(B)⟶O(B)\psi_q:\mathcal{O}(\mathbb{B})\longrightarrow\mathcal{O}(\mathbb{B}) as the unique map such that

φψq=1qTr⁡O(B)/φ(O(B)).\varphi\psi_q=\frac{1}{q}\operatorname{Tr}_{\mathcal{O}(\mathbb{B})/\varphi(\mathcal{O}(\mathbb{B}))}.

From now on, by transport and descent, we view these two operators as acting on distributions. We call them φ,ψ:D(oL,L)⟶D(oL,L)\varphi,\psi:\mathrm{D}(o_L,L)\longrightarrow\mathrm{D}(o_L,L). After transport and descent, we get that φ\varphi acts on distributions as multiplication by π\pi, or in other words φ=π∗\varphi=\pi_*. The action of ψ\psi is given by

(ψλ)(f)=λ(Mf).(\psi\lambda)(f)=\lambda(M_f).

where Mf(x)M_f(x) is defined as 00 on oL−πoLo_L-\pi o_L and as f(xπ)f\left(\frac{x}{\pi}\right) otherwise. For the proof of these claims see [9 Remark 4.1.4, Section 4.1.4].

Analytic boundary

We slightly modify the definition of an analytic boundary from [2] by restricting the family of distributions to LL-valued ones. We fix the norm on X(Cp)\mathcal{X}(\mathbb{C}_p) by the LT isomorphism from Section 2.2.

Definition 2.2. A subset {an}n=0∞\{a_n\}_{n=0}^{\infty} of X(Cp)\mathcal{X}(\mathbb{C}_p) that satisfies ∣an∣→1|a_n|\to1 with n→∞n\to\infty is called an analytic boundary if for all λ∈Db(oL,L)\lambda\in\mathrm{D}^{b}(o_L,L) we have

∣λ∣=sup⁡n∣λ(an)∣.|\lambda|=\sup_n|\lambda(a_n)|.

The following result will play an important role.

Theorem 2.3 (Berger). Let NN be a nonnegative integer. Then X[π>N]\mathcal{X}[\pi^{>N}] is an analytic boundary. In particular, for λ∈Db(oL,L)\lambda\in\mathrm{D}^{b}(o_L,L) we have ∣λ∣=sup⁡ξ∈X[π>N]∣λ(ξ)∣|\lambda|=\sup_{\xi\in\mathcal{X}[\pi^{>N}]}|\lambda(\xi)|.

The proof follows by combining [2 Theorem D] and [2 Lemma 1.1.3]. We record one useful lemma here.

Lemma 2.4. Let λ∈D(oL,L)\lambda\in\mathrm{D}(o_L,L) and a∈oLa\in o_L with a≠0a\ne0. Then λ∈Db(oL,L)\lambda\in\mathrm{D}^{b}(o_L,L) if and only if a∗λ∈Db(oL,L)a_*\lambda\in\mathrm{D}^{b}(o_L,L). Furthermore, if λ∈Db(oL,L)\lambda\in\mathrm{D}^{b}(o_L,L), then

∣λ∣=∣a∗λ∣.|\lambda|=|a_*\lambda|.

Specifically, if ∣φλ∣≤1|\varphi\lambda|\le1 then ∣λ∣≤1|\lambda|\le1.

Proof. The proof is immediate since multiplication by aa is surjective on X(Cp)\mathcal{X}(\mathbb{C}_p) by [9 Lemma 4.1.10]. □\square

Partial trace operators

Here we define partial trace operators that will be used to factorize ψ\psi. Since X[p]≅oL/poL\mathcal{X}[p] \cong\mathfrak{o}_L/p\mathfrak{o}_L is a Z/pZ\mathbb{Z}/p\mathbb{Z}-module of rank dd, there exist dd elements ζ1,…,ζd\zeta_1,\ldots,\zeta_d such that they form a Z/pZ\mathbb{Z}/p\mathbb{Z}-basis of X[p]\mathcal{X}[p]. Let ker⁡ζi=Ki≤oL\operatorname{ker}\zeta_i=K_i\leq\mathfrak{o}_L.

Definition 2.5. For 1≤i≤d1\leq i\leq d define

ψi:D(oL,L)⟶D(oL,L)\psi_i:\mathcal{D}(\mathfrak{o}_L,L)\longrightarrow\mathcal{D}(\mathfrak{o}_L,L)

by $\psi_i(\lambda)=\lambda(1_{K_i}).

Sketch of the proof

Recall that {μpn}n=1∞\{\mu_{p^n}\}_{n=1}^{\infty} is a system of primitive pnp^n-th roots of unity.

We want to prove that ψ(ΛL(X))≤ΛL(X)\psi(\Lambda_L(\mathcal{X}))\leq\Lambda_L(\mathcal{X}), i.e. that the normalized trace operator is integral, see [1 Theorem 3.1.9]. The reduction to the integrality of the partial trace operators ψi\psi_i is formal, see Lemma 3.1.

Fix from now on one index ii and a distribution λ∈ΛL(X)\lambda\in\Lambda_L(\mathcal{X}). To prove the integrality of the operator ψi\psi_i, i.e. that ∣ψiλ∣≤1\lvert\psi_i\lambda\rvert\leq1, we can use the theorem of Berger [2 Theorem D] that says that the norm of ψiλ\psi_i\lambda can be seen as the supremum of ∣(ψiλ)(ξ)∣\lvert(\psi_i\lambda)(\xi)\rvert over all ξ∈X[p∞]\xi\in\mathcal{X}[p^\infty].

The main idea of the proof is that for sufficiently large integers nn and characters ξ∈X[p∞]\xi\in\mathcal{X}[p^\infty] satisfying pnξ=1p^n\xi=1, pn−1ξ=ζip^{n-1}\xi=\zeta_i, we have that the value ψiλ(ξ)\psi_i\lambda(\xi) can be seen as the normalized arithmetic trace for the totally ramified field extension L(μpn)/L(μpn−1)L(\mu_{p^n})/L(\mu_{p^{n-1}}), see Proposition 3.4 (compare to [3 p. 1023]). There are good estimates for these traces due to Tate, see [10 3.1, Corollary 2], which allow us to bound ∣(ψiλ)(ξ)∣\lvert(\psi_i\lambda)(\xi)\rvert giving us the integrality of the operators ψi\psi_i.

Two issues remain. The first one is that extensions L(μpn)/L(μpn−1)L(\mu_{p^n})/L(\mu_{p^{n-1}}) for small nn need not satisfy the hypothesis of Tate’s theorem (for example, these fields could be equal), so we need to be able to take nn large enough. The second issue is that for the trace formula to hold we need to restrict to characters ξ∈X[p∞]\xi\in\mathcal{X}[p^\infty] satisfying pn−1ξ=ζip^{n-1}\xi=\zeta_i. This is where Proposition 3.7 comes in, which is a version of [2 Theorem D] that restricts X[p∞]\mathcal{X}[p^\infty] to the family of characters such that pn−1ξ=ζip^{n-1}\xi=\zeta_i for large enough nn.

The geometric interpretation of the equality between the evaluation of ψiλ\psi_i\lambda at a torsion point and the normalized arithmetic trace is that the trace operator for finite étale (or finite locally free) maps commutes with base change. We apply this to the following unramified cover of degree pp, where Xker⁡ζi\mathcal{X}_{\operatorname{ker}\zeta_i} is the character variety associated to the subgroup ker⁡ζi\operatorname{ker}\zeta_i and the vertical arrow on the right is the restriction.

Sp⁡L(μpn)→ξX↓↓Sp⁡L(μpn−1)→ξ∣ker⁡ζiXker⁡ζi\begin{CD} \operatorname{Sp} L(\mu_{p^n}) @>{\xi}>> \mathcal{X} \\ @VVV @VVV \\ \operatorname{Sp} L(\mu_{p^{n-1}}) @>{\xi\mid_{\operatorname{ker}\zeta_i}}>> \mathcal{X}_{\operatorname{ker}\zeta_i} \end{CD}

To elaborate further, take a large enough integer nn and take a character ξ∈X(Cp)\xi\in\mathcal{X}(\mathbb{C}_p) of order pnp^n such that pn−1ξ=ζip^{n-1}\xi=\zeta_i. Then ξ\xi has values in L(μpn)L(\mu_{p^n}), and ξ∣ker⁡ζi\xi\mid_{\operatorname{ker}\zeta_i} has values in L(μpn−1)L(\mu_{p^{n-1}}). The fiber of the image of ξ\xi in Xker⁡ζi(Cp)\mathcal{X}_{\operatorname{ker}\zeta_i}(\mathbb{C}_p) is equal to {ξ⋅ζik}k=0p−1\{\xi\cdot\zeta_i^k\}_{k=0}^{p-1}. But this set is exactly the set of Galois conjugates of ξ\xi under the action of Gal⁡(L(μpn)/L(μpn−1))\operatorname{Gal}(L(\mu_{p^n})/L(\mu_{p^{n-1}})). Now the partial trace operator ψi\psi_i is equal to averaging over the fiber, see Lemma 3.2. However, the fiber consists of Galois conjugates as noted, so the averaging over the fiber is actually the normalized arithmetic trace.

The idea to connect arithmetic trace to evaluation at torsion points is already in the literature, see for example [3 p. 1023].

Proof of Theorem 1.1

We start the proof of Theorem 1.1 by reducing the statement to the integrality of the partial trace operators ψi\psi_i.

Lemma 3.1. To prove Theorem 1.1 it suffices to prove that ψiΛL(X)≤ΛL(X)\psi_i\Lambda_L(\mathcal{X}) \leq\Lambda_L(\mathcal{X}) for every 1≤i≤d1 \leq i \leq d.

Proof. We have

∏i=1dψi=φeψe,(2)\prod_{i=1}^{d}\psi_i=\varphi^e\psi^e, \tag*{(2)}

where the product means composition of operators. Indeed, we have that ⋂i=1dKi=poL=πoLe\bigcap_{i=1}^{d}K_i=p_{o_L}=\pi^{e}_{o_L}, and therefore

(∏i=1dψi)λ=λ(1poL⋅)=φeψe(λ).\left(\prod_{i=1}^{d}\psi_i\right)\lambda=\lambda(1_{p_{o_L}\cdot})=\varphi^e\psi^e(\lambda).

By [1 Theorem 3.1.9], to prove Theorem 1.1 it suffices to prove ψ(ΛL(X))≤ΛL(X)\psi(\Lambda_L(\mathcal{X}))\leq\Lambda_L(\mathcal{X}). By the assumption we have

(∏i=1dψi)ΛL(X)≤ΛL(X),\left(\prod_{i=1}^{d}\psi_i\right)\Lambda_L(\mathcal{X})\leq\Lambda_L(\mathcal{X}),

which, combined with (2), gives us that

φeψeΛL(X)≤ΛL(X).\varphi^e\psi^e\Lambda_L(\mathcal{X})\leq\Lambda_L(\mathcal{X}).

Lemma 2.4 gives us that

ψeΛL(X)≤ΛL(X).\psi^e\Lambda_L(\mathcal{X})\leq\Lambda_L(\mathcal{X}).

Since φΛL(X)≤ΛL(X)\varphi\Lambda_L(\mathcal{X})\leq\Lambda_L(\mathcal{X}) again by Lemma 2.4, we get that

ψeφe−1ΛL(X)≤ΛL(X).\psi^e\varphi^{e-1}\Lambda_L(\mathcal{X})\leq\Lambda_L(\mathcal{X}).

Since ψφ=1\psi\varphi=1, this finishes the proof. □

Now we study partial trace operators further. First we give a different interpretation of ψi\psi_i as an average over the fiber under restriction X→XKi\mathcal{X}\to\mathcal{X}_{K_i}.

Lemma 3.2. For λ∈D(oL,L)\lambda\in\mathrm{D}(o_L,L) and f∈CL-an(oL,Cp)f\in\mathrm{C}^{L\text{-an}}(o_L,\mathbb{C}_p) we have3^3

ψiλ(f)=1p∑k=0p−1λ(f⋅kζi).\psi_i\lambda(f)=\frac{1}{p}\sum_{k=0}^{p-1}\lambda(f\cdot k\zeta_i).

Proof. The lemma is equivalent to the statement that

λ(1Ki⋅f)=1p∑k=0p−1λ(f⋅kζi),\lambda(1_{K_i}\cdot f)=\frac{1}{p}\sum_{k=0}^{p-1}\lambda(f\cdot k\zeta_i),

which follows because ζi\zeta_i is a character of order pp and this implies

1Ki=1p∑k=0p−1kζi.1_{K_i}=\frac{1}{p}\sum_{k=0}^{p-1}k\zeta_i.
□\square

Remark 3.3. In the Lubin-Tate coordinate, ψi\psi_i is given by ψiF(Z)=1p∑k=0p−1F(Z⊕[k]ζi)\psi_i F(Z) = \frac{1}{p}\sum_{k=0}^{p-1} F(Z \oplus[k]\zeta_i).

As a consequence of Lemma 3.2 we have that ψi\psi_i preserves bounded distributions

ψi:Db(oL,L)⟶Db(oL,L),\psi_i: D^b(o_L,L) \longrightarrow D^b(o_L,L),

and that these operators are bounded on Db(oL,L)D^b(o_L,L).

The idea for proving the integrality of the operators ψi\psi_i is testing their values against ξ∈X[p∞]\xi\in\mathcal{X}[p^\infty] by using Theorem 2.3. Since such ξ∈X[p∞]\xi\in\mathcal{X}[p^\infty] are torsion, their values lie in cyclotomic extensions and hence they inherit a natural Galois action which we inspect. To bound the norm of ψiλ\psi_i\lambda, for λ∈Db(oL,L)\lambda\in D^b(o_L,L), we will interpret the value ψiλ(ξ)\psi_i\lambda(\xi) as the value of the normalized trace of the cyclotomic extension. This equality is the key step in the entire proof. Recall that {μpn}n=1∞\{\mu_{p^n}\}_{n=1}^{\infty} is a system of primitive pnp^n-th roots of unity and Ln=L(μpn)L_n = L(\mu_{p^n}), Ki=ker⁡ζi≤oLK_i = \ker\zeta_i \leq o_L.

Proposition 3.4. Let ξ∈X[p∞]\xi\in\mathcal{X}[p^\infty] and let n≥2n \geq2 be an integer. Suppose that pn−1ξ=ζip^{n-1}\xi= \zeta_i and that Ln/Ln−1L_n/L_{n-1} is cyclic of degree pp. For λ∈D(oL,L)\lambda\in D(o_L,L) we have λ(ξ)∈Ln\lambda(\xi) \in L_n and

ψiλ(ξ)=1pTr⁡Ln/Ln−1λ(ξ).\psi_i\lambda(\xi) = \frac{1}{p}\operatorname{Tr}_{L_n/L_{n-1}}\lambda(\xi).

Proof. Note that pnξ=pζi=1p^n\xi= p\zeta_i = 1, and hence ξ\xi takes values in LnL_n. Since torsion characters are locally constant, there exists a decomposition

ξ=∑aˉ∈oL/pnoLξ(a)1a+pnoL.\xi= \sum_{\bar{a}\in o_L/p^n o_L} \xi(a)1_{a+p^n o_L}.

But because λ\lambda is defined over LL, this shows that λ(ξ)∈Ln\lambda(\xi) \in L_n.

The group G=Gal⁡(Ln/Ln−1)G = \operatorname{Gal}(L_n/L_{n-1}) acts on X[pn]\mathcal{X}[p^n], and we are interested in the orbit of ξ\xi. First we prove that the orbit of ξ\xi has cardinality pp. To do this, it is enough to prove that the stabilizer of ξ\xi has cardinality 1. Therefore let g∈Gg \in G be such that gξ=ξg\xi= \xi. Note that since ξ\xi has exact order pnp^n, there exists an a∈oLa \in o_L such that ξ(a)=μpn\xi(a) = \mu_{p^n}. But we have that g(μpn)=gξ(a)=ξ(a)=μpng(\mu_{p^n}) = g\xi(a) = \xi(a) = \mu_{p^n}, which implies that g=1g = 1.

We claim that gξξ(Ki)=1\frac{g\xi}{\xi}(K_i) = 1 for all g∈Gg \in G. To prove this, it suffices to prove that GG acts trivially on ξ(Ki)\xi(K_i). From the assumption that pn−1ξ=ζip^{n-1}\xi= \zeta_i, we get that ξ(Ki)pn−1=(pn−1ξ)(Ki)=ζi(Ki)=1\xi(K_i)^{p^{n-1}} = (p^{n-1}\xi)(K_i) = \zeta_i(K_i) = 1. This implies that ξ(Ki)≤Ln−1\xi(K_i) \leq L_{n-1}, and so GG acts trivially on ξ(Ki)\xi(K_i).

We have proved that the set {gξξ}g∈G\left\{\frac{g\xi}{\xi}\right\}_{g\in G} gives us pp different elements of Hom⁡(oL/Ki,Cp×)\operatorname{Hom}(o_L/K_i,\mathbb{C}_p^\times). On the other hand, since oL/Kio_L/K_i is cyclic of order pp, we have that Hom⁡(oL/Ki,Cp×)={kζi}k=0p−1\operatorname{Hom}(o_L/K_i,\mathbb{C}_p^\times) = \{k\zeta_i\}_{k=0}^{p-1}. This shows that {gξ}g∈G={ξ⋅kζi}k=0p−1\{g\xi\}_{g\in G} = \{\xi\cdot k\zeta_i\}_{k=0}^{p-1}.

Note that g(λ(ξ))=λ(gξ)g(\lambda(\xi)) = \lambda(g\xi) because λ\lambda is defined over LL and ξ\xi is torsion and hence locally constant. Therefore

1pTr⁡Ln/Ln−1λ(ξ)=1p∑g∈Gal⁡(Ln/Ln−1)g(λ(ξ))=1p∑g∈Gal⁡(Ln/Ln−1)λ(gξ)=1p∑k=0p−1λ(ξ⋅kζi).\frac{1}{p}\operatorname{Tr}_{L_n/L_{n-1}}\lambda(\xi) = \frac{1}{p}\sum_{g\in\operatorname{Gal}(L_n/L_{n-1})}g(\lambda(\xi)) = \frac{1}{p}\sum_{g\in\operatorname{Gal}(L_n/L_{n-1})}\lambda(g\xi) = \frac{1}{p}\sum_{k=0}^{p-1}\lambda(\xi\cdot k\zeta_i).

Applying Lemma 3.2 finishes the proof. □

Remark 3.5. Proposition 3.4 can also be deduced from the cyclotomic case of the trace formula in [3 p. 1023] by passing to the finite cyclic quotient defined by ξ\xi and extending scalars. The proof we give is direct.

The next step is to bound this trace. It turns out this is a direct corollary of the classic result by Tate [10 Section 3.1, Corollary 2]. We prove it in Section 5.

Lemma 3.6. For any ε>0\varepsilon> 0 there exists an integer N=N(ε)N = N(\varepsilon) such that for every n>Nn > N we have that Ln/Ln−1L_n/L_{n-1} is Galois cyclic totally ramified of degree pp and

∣1pTr⁡Ln/Ln−1(x)∣≤pε,\left|\frac{1}{p}\operatorname{Tr}_{L_n/L_{n-1}}(x)\right| \le p^{\varepsilon},

for all x∈oLnx \in o_{L_n}.

Recall that we want to prove that ∣ψiλ∣≤1|\psi_i\lambda| \le1 for λ∈ΛL(X)\lambda\in\Lambda_L(\mathcal{X}). At this point we would like to apply a description of the norm as a supremum of ∣(ψiλ)(ξ)∣|(\psi_i\lambda)(\xi)| over ξ∈X[π∞]\xi\in\mathcal{X}[\pi^\infty], as given in Theorem 2.3, and then apply Proposition 3.4 and Lemma 3.6 to bound these values. However, in order to apply Proposition 3.4 and Lemma 3.6, we need to restrict to the family of ξ∈X[p∞]\xi\in\mathcal{X}[p^\infty] such that pn−1ξ=ζip^{n-1}\xi= \zeta_i for large enough nn. To this end, we prove the following strengthening of Theorem 2.3.

Proposition 3.7. Let ζ∈X[p∞]\zeta\in\mathcal{X}[p^\infty] and let NN be a positive integer. The set

Bζ,N={ξ∈X[p∞]:(∃m∈Z, m>N), pmξ=ζ}B_{\zeta,N} = \{\xi\in\mathcal{X}[p^\infty] : (\exists m \in\mathbb{Z},\ m > N),\ p^m\xi= \zeta\}

is an analytic boundary. In particular, for λ∈Db(oL,L)\lambda\in\mathrm{D}^{b}(o_L,L) we have ∣λ∣=sup⁡ξ∈Bζ,N∣λ(ξ)∣|\lambda| = \sup_{\xi\in B_{\zeta,N}} |\lambda(\xi)|.

The proof of this proposition uses Theorem 2.3 and is given in Section 4.

We can now prove that the operators ψi\psi_i are integral by combining Proposition 3.4, Lemma 3.6 and Proposition 3.7.

Theorem 3.8. The operators ψi:D(oL,L)→D(oL,L)\psi_i : \mathrm{D}(o_L,L) \to\mathrm{D}(o_L,L) are integral. In other words, ψiΛL(X)≤ΛL(X)\psi_i\Lambda_L(\mathcal{X}) \le\Lambda_L(\mathcal{X}).

Proof. For λ∈ΛL(X)\lambda\in\Lambda_L(\mathcal{X}) it suffices to prove that ∣ψiλ∣≤pε|\psi_i\lambda| \le p^\varepsilon for every 0<ε0 < \varepsilon. Therefore, fix a nonzero λ∈ΛL(X)\lambda\in\Lambda_L(\mathcal{X}) and ε>0\varepsilon> 0, and take NN to be large enough such that Lemma 3.6 holds. By Proposition 3.7 we have

∣ψiλ∣=sup⁡ξ∈Bζi,N∣ψiλ(ξ)∣,|\psi_i\lambda| = \sup_{\xi\in B_{\zeta_i,N}} |\psi_i\lambda(\xi)|,

and therefore it is enough to prove that

∣ψiλ(ξ)∣≤pε|\psi_i\lambda(\xi)| \le p^\varepsilon

for every ξ∈Bζi,N\xi\in B_{\zeta_i,N}. In this regard, take ξ∈Bζi,N\xi\in B_{\zeta_i,N} and a positive integer nn such that we have pn−1ξ=ζip^{n-1}\xi= \zeta_i. By the definition of Bζi,NB_{\zeta_i,N}, we have that n>Nn > N.

By Proposition 3.4

ψiλ(ξ)=1pTr⁡Ln/Ln−1λ(ξ),\psi_i\lambda(\xi) = \frac{1}{p}\operatorname{Tr}_{L_n/L_{n-1}}\lambda(\xi),

and λ(ξ)∈Ln\lambda(\xi) \in L_n. Note also that since ∣λ∣≤1|\lambda| \le1 it follows that λ(ξ)∈oLn\lambda(\xi) \in o_{L_n}. Now, by the way we chose NN to satisfy Lemma 3.6, and because n>Nn > N we have

∣ψiλ(ξ)∣=∣1pTr⁡Ln/Ln−1λ(ξ)∣≤pε.|\psi_i\lambda(\xi)| = \left|\frac{1}{p}\operatorname{Tr}_{L_n/L_{n-1}}\lambda(\xi)\right| \le p^\varepsilon.

This finishes the proof. □

Remark 3.9. The fact that Ln/Ln−1L_n/L_{n-1} is totally ramified is important only for Tate’s bound.

The proof of Theorem 1.1 follows by combining Theorem 3.8 and Lemma 3.1.

A union of fibers in X[p∞]\mathfrak{X}[p^\infty] is an analytic boundary

In this section we prove Proposition 3.7. Fix a positive integer NN throughout this section.

Let x∈L/OLx \in L/\mathcal{O}_L be an element of π\pi-order equal to mm and define

Bx,N={y∈L/OL:(∃n∈Z,n>N),pny=x}.B_{x,N}=\{y\in L/\mathcal{O}_L:(\exists n\in\mathbb{Z},n>N),p^ny=x\}.

Let (L/OL)[π>M](L/\mathcal{O}_L)[\pi^{>M}] be the set of elements of L/OLL/\mathcal{O}_L whose π\pi-order is bigger than MM. The following lemma is just a linear algebra fact about the OL\mathcal{O}_L-module L/OLL/\mathcal{O}_L.

Lemma 4.1. There exists a finite set S=Sx,N≤OL−{0}S=S_{x,N}\leq\mathcal{O}_L-\{0\} and an integer M=M(m,N)M=M(m,N) such that for every y∈(L/OL)[π>M]y\in(L/\mathcal{O}_L)[\pi^{>M}] there exists s∈Ss\in S such that sy∈Bx,Nsy\in B_{x,N}.

Proof. For x=0x=0 the proof is trivial, so we assume that x≠0x\neq0. Let R≤OL−{0}R\leq\mathcal{O}_L-\{0\} be a finite set such that every element of (OL/πmOL)×(\mathcal{O}_L/\pi^m\mathcal{O}_L)^\times has a lift in RR, and take

S={rπi:r∈R,0≤i<e}.S=\{r\pi^i:r\in R,0\leq i<e\}.

We can take MM large enough, and we will see that it suffices to take M=2eN+mM=2eN+m. We take y∈(L/OL)[π>M]y\in(L/\mathcal{O}_L)[\pi^{>M}], and note that, because of our choice of MM, there exists a unique pair of integers (l,i)(l,i), with 0≤i≤e−10\leq i\leq e-1 and l>Nl>N, such that plπiyp^l\pi^iy has π\pi-order equal to mm. Since the action of OL×\mathcal{O}_L^\times on elements of π\pi-order mm in OL/πm\mathcal{O}_L/\pi^m is transitive, we can find an element r∈Rr\in R such that rplπiy=xrp^l\pi^iy=x. Therefore pl(rπi)y=xp^l(r\pi^i)y=x, which implies that rπiy∈Bx,Nr\pi^iy\in B_{x,N}. □

Remark 4.2. The set SS works for any NN and only depends on the π\pi-order of xx.

Proof of Proposition 3.7. For ζ=1\zeta=1 we have B1,N=X[p∞]B_{1,N}=\mathfrak{X}[p^\infty], so by Theorem 2.3 we are done. Henceforth assume that ζ≠1\zeta\neq1. By Remark 2.1 we have an isomorphism X[p∞]≅L/OL\mathfrak{X}[p^\infty]\cong L/\mathcal{O}_L as OL\mathcal{O}_L-modules, so we can take SS and M=M(m,N)M=M(m,N) as in Lemma 4.1, where mm is the π\pi-order of ζ\zeta. By Theorem 2.3, for any α∈Db(OL,L)\alpha\in D^b(\mathcal{O}_L,L), we have

∣α∣=sup⁡ξ∈X[π>M]∣α(ξ)∣.|\alpha|=\sup_{\xi\in\mathfrak{X}[\pi^{>M}]}|\alpha(\xi)|.

Let 0≠λ∈Db(OL,L)0\neq\lambda\in D^b(\mathcal{O}_L,L) and

λS=∏s∈S(s∗λ).\lambda_S=\prod_{s\in S}(s_*\lambda).

Take some η∈X[π>M]\eta\in\mathfrak{X}[\pi^{>M}]. Then from Lemma 4.1 it follows that there exists some s∈Ss\in S such that sη∈Bζ,Ns\eta\in B_{\zeta,N}. We have

∣λS(η)∣=∣λ(sη)∣∏s′∈S−{s}∣λ(s′η)∣≤(sup⁡ξ∈Bζ,N∣λ(ξ)∣)∣λ∣∣S∣−1.|\lambda_S(\eta)|=|\lambda(s\eta)|\prod_{s'\in S-\{s\}}|\lambda(s'\eta)|\leq\left(\sup_{\xi\in B_{\zeta,N}}|\lambda(\xi)|\right)|\lambda|^{|S|-1}.

Taking the supremum over η∈X[π>M]\eta\in\mathfrak{X}[\pi^{>M}] and noting that ∣λS∣=∣λ∣∣S∣|\lambda_S|=|\lambda|^{|S|} by multiplicativity of the norm (§ 2.2) and Lemma 2.4, we get

∣λ∣∣S∣=∣λS∣=sup⁡η∈X[π>M]∣λS(η)∣≤∣λ∣∣S∣−1sup⁡ξ∈Bζ,N∣λ(ξ)∣.|\lambda|^{|S|}=|\lambda_S|=\sup_{\eta\in\mathfrak{X}[\pi^{>M}]}|\lambda_S(\eta)|\leq|\lambda|^{|S|-1}\sup_{\xi\in B_{\zeta,N}}|\lambda(\xi)|.

Therefore we get ∣λ∣≤sup⁡ξ∈Bζ,N∣λ(ξ)∣|\lambda|\leq\sup_{\xi\in B_{\zeta,N}}|\lambda(\xi)|. The reverse inequality is obvious, showing that ∣λ∣=sup⁡ξ∈Bζ,N∣λ(ξ)∣|\lambda|=\sup_{\xi\in B_{\zeta,N}}|\lambda(\xi)|. Finally, if ξ∈Bζ,N\xi\in B_{\zeta,N} has π\pi-order nn, by the proof of [1 Theorem 3.5.4] we have that

∣ξ∣=∣π∣1qn−1(q−1),|\xi|=|\pi|^{\frac{1}{q^{n-1}(q-1)}},

and the left-hand side tends to 11 as nn tends to infinity, finishing the proof. □

Remark 4.3. For the argument in Theorem 3.8 we do not need that Bζ,NB_{\zeta,N} has points whose norm is tending to 11.

Normalized trace estimates for totally ramified Zp\mathbb{Z}_p-extensions

In this section we prove Lemma 3.6. Recall the following result of Tate [10 Section 3.1, Corollary 2]. Here KK is a complete discretely valued field of characteristic 00, with a perfect residue field of characteristic pp.

Theorem 5.1 (Tate). Let K∞/KK_\infty/K be a totally ramified Galois extension with Galois group Zp\mathbb{Z}_p. Let KnK_n correspond to the subgroup pnZpp^n\mathbb{Z}_p. Then for every integer n≥1n \ge1, Kn/KK_n/K is cyclic of degree pnp^n and there exists a constant aa, independent of nn, such that for any x∈Knx \in K_n

∣Tr⁡Kn/Kn−1(x)∣≤∣x∣∣p∣1−ap−n+1.\left|\operatorname{Tr}_{K_n/K_{n-1}}(x)\right| \le|x| |p|^{1-ap^{-n+1}}.

Recall that {μpn}n=1∞\{\mu_{p^n}\}_{n=1}^{\infty} is a system of primitive pnp^n-th roots of unity and that Qp,n=Qp(μpn)\mathbb{Q}_{p,n} = \mathbb{Q}_p(\mu_{p^n}), Ln=L(μpn)L_n = L(\mu_{p^n}), L∞=lim⁡⟶nLnL_\infty= \lim\limits_{\longrightarrow n} L_n.

Proof of Lemma 3.6. Denote the residue fields of LnL_n by knk_n, and note that the sequence [kn:Fp][k_n : \mathbb{F}_p] is bounded and nondecreasing, and hence it stabilises. The sequence of numbers [Ln:Qp,n][L_n : \mathbb{Q}_{p,n}] is nonincreasing and hence it stabilises. Take a large m≥2m \ge2 such that both of the sequences stabilise. Then we have [Ln:Ln−1]=p[L_n : L_{n-1}] = p for n>mn > m and these extensions are totally ramified. Next we prove that L∞/LnL_\infty/L_n is totally ramified Galois with Galois group Zp\mathbb{Z}_p for all n≥mn \ge m.

We have [Ln+s:Ln]=ps[L_{n+s} : L_n] = p^s and this extension is totally ramified. Since Ln+s=Ln(μpn+s)L_{n+s} = L_n(\mu_{p^{n+s}}), this is also Galois. Take σ∈Gal⁡(Ln+s/Ln)\sigma\in\operatorname{Gal}(L_{n+s}/L_n). Then σ\sigma fixes μpn\mu_{p^n} and is determined by what it does to μpn+s\mu_{p^{n+s}}. This proves that the natural restriction map

Gal⁡(Ln+s/Ln)⟶Gal⁡(Qp,n+s/Qp,n)\operatorname{Gal}(L_{n+s}/L_n) \longrightarrow\operatorname{Gal}(\mathbb{Q}_{p,n+s}/\mathbb{Q}_{p,n})

is an injection. Since both groups have cardinality psp^s, this has to be an isomorphism. Finally we get

Gal⁡(L∞/Ln)=lim⁡⟵sGal⁡(Ln+s/Ln)≅lim⁡⟵sGal⁡(Qp,n+s/Qp,n)→∼χcyc1+pnZp,\operatorname{Gal}(L_\infty/L_n) = \lim_{\longleftarrow s} \operatorname{Gal}(L_{n+s}/L_n) \cong\lim_{\longleftarrow s} \operatorname{Gal}(\mathbb{Q}_{p,n+s}/\mathbb{Q}_{p,n}) \xrightarrow[\sim]{\chi_{\mathrm{cyc}}} 1+p^n\mathbb{Z}_p,

and note that 1+pnZp≅Zp1+p^n\mathbb{Z}_p \cong\mathbb{Z}_p since n≥2n \ge2. This shows that L∞/LnL_\infty/L_n is a totally ramified Galois extension with Galois group Zp\mathbb{Z}_p for all n≥mn \ge m.

Finally let K=LmK = L_m and K∞=L∞K_\infty= L_\infty and apply Tate’s theorem to get

∣1pTr⁡Lm+n+1/Lm+n(x)∣≤∣x∣∣p∣−ap−n≤∣p∣−ap−n=pap−n\left|\frac{1}{p}\operatorname{Tr}_{L_{m+n+1}/L_{m+n}}(x)\right| \le|x||p|^{-ap^{-n}} \le|p|^{-ap^{-n}} = p^{ap^{-n}}

for all nn and all x∈OLm+n+1x \in\mathcal{O}_{L_{m+n+1}}. Now take nn large enough such that pap−n<pεp^{ap^{-n}} < p^\varepsilon. Taking N=n+m+1N = n + m + 1 finishes the proof. □\square

Uroš A. Colović, Mathematical Institute, University of Oxford

Email address: [email protected]

References

  1. [1]Konstantin Ardakov and Laurent Berger. “Bounded functions on the character variety”. In: Münster J. Math. 17.1 (2024), pp. 1–83. ISSN: 1867-5778,1867-5786. DOI: 10.17879/51958619387.DOI
  2. [2]Laurent Berger. “Lubin-Tate generalizations of the p-adic Fourier transform”. In: Bull. Soc. Math. France 153.2 (2025), pp. 559–573. ISSN: 0037-9484,2102-622X. DOI: 10.24033/bsmf.2906.DOI
  3. [3]Laurent Berger and Lionel Fourquaux. “Iwasawa theory and F-analytic Lubin-Tate (φ, Γ)-modules”. In: Doc. Math. 22 (2017), pp. 999–1030. ISSN: 1431-0635,1431-0643.DOI
  4. [4]Laurent Berger and Johannes Sprang. “Integer-valued polynomials and p-adic Fourier theory”. In: Rendiconti del Seminario Matematico della Università di Padova (2026). DOI: 10.4171/RSMUP/204.DOI
  5. [5]Laurent Berger, Peter Schneider, and Bingyong Xie. “Rigid character groups, Lubin–Tate theory, and (φ, Γ)-modules”. In: Mem. Amer. Math. Soc. 263.1275 (2020), pp. v+79. ISSN: 0065-9266,1947-6221. DOI: 10.1090/memo/1275.DOI
  6. [6]Pierre Colmez. “Représentations localement analytiques de GL₂(Qₚ) et (φ, Γ)-modules”. In: Representation Theory 20 (July 2016), pp. 187–248. DOI: 10.1090/ert/484.
  7. [7]Ehud de Shalit and Eran Iceland. “Integer valued polynomials and Lubin–Tate formal groups”. In: Journal of Number Theory 129.3 (2009), pp. 632–639.DOI
  8. [8]P. Schneider and J. Teitelbaum. “p-adic Fourier theory”. In: Doc. Math. 6 (2001), pp. 447–481. ISSN: 1431-0635,1431-0643.
  9. [9]Peter Schneider and Otmar Venjakob. Reciprocity laws for (φ_L, Γ_L)-modules over Lubin-Tate extensions. Vol. 24. Memoirs of the European Mathematical Society. EMS Press, Berlin, 2026. ISBN: 978-3-98547-100-3; 978-3-98547-600-8.DOI
  10. [10]J. T. Tate. “p-divisible groups”. In: Proc. Conf. Local Fields (Driebergen, 1966). Springer, Berlin-New York, 1967, pp. 158–183.DOI

Paper details

vlnc:W_yzwkk9ga
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents