Hostname: page-component-669899f699-b58lm Total loading time: 0 Render date: 2025-04-26T02:29:14.604Z Has data issue: false hasContentIssue false

Deformation theory for prismatic G-displays

Published online by Cambridge University Press:  21 March 2025

Kazuhiro Ito*
Affiliation:
Mathematical Institute, Tohoku University, 6-3, Aoba, Aramaki, Aoba-Ku, Sendai, 980-8578, Japan

Abstract

For a smooth affine group scheme G over the ring of p-adic integers and a cocharacter $\mu $ of G, we develop the deformation theory for G-$\mu $-displays over the prismatic site of Bhatt–Scholze, and discuss how our deformation theory can be interpreted in terms of prismatic F-gauges introduced by Drinfeld and Bhatt–Lurie. As an application, we prove the local representability and the formal smoothness of integral local Shimura varieties with hyperspecial level structure. We also revisit and extend some classification results of p-divisible groups.

Type
Number Theory
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.
Copyright
© The Author(s), 2025. Published by Cambridge University Press

1 Introduction

Let p be a prime number. Let G be a smooth affine group scheme over the ring of p-adic integers ${\mathbb Z}_p$ . Let $ \mu \colon {\mathbb G}_m \to G_{W(k)}:=G \times _{\mathrm {Spec} {\mathbb Z}_p} \mathrm {Spec} W(k) $ be a cocharacter defined over $W(k)$ , where k is a perfect field of characteristic p and $W(k)$ is the ring of p-typical Witt vectors of k. The theory of G- $\mu $ -displays was originally introduced by Bültel [Reference BültelBül08] and Bültel–Pappas [Reference Bültel and PappasBP20] to study integral models of Shimura varieties and Rapoport–Zink spaces. In [Reference ItoIto23], we studied the notion of G- $\mu $ -displays over the prismatic site of Bhatt–Scholze [Reference Bhatt and ScholzeBS22]. The purpose of this paper is to develop the deformation theory for G- $\mu $ -displays over the prismatic site, which has applications in the theory of integral models of Shimura varieties and local Shimura varieties. Our deformation theory for $G= {\mathrm {GL}} _N$ provides new perspectives on some classification results of p-divisible groups. We also discuss how our results can be interpreted in terms of prismatic F-gauges introduced by Drinfeld and Bhatt–Lurie.

1.1 Main results

Let $(A, I)$ be a bounded prism in the sense of [Reference Bhatt and ScholzeBS22], such that A is a $W(k)$ -algebra. We study the groupoid of G- $\mu $ -displays over $(A, I)$

which is introduced in [Reference ItoIto23] following the theory of G- $\mu $ -displays over higher frames developed by Lau [Reference LauLau21] and the work of Bartling [Reference BartlingBar22]. The groupoid is equivalent to the groupoid of G-Breuil–Kisin modules of type $\mu $ over $(A, I)$ , which may be more familiar to the reader. See Section 2.4 for details.

Let us summarize the main results of this paper. In the remainder of this introduction, we assume that $\mu $ is 1-bounded; that is, the weights of the action of ${\mathbb G}_m$ on the Lie algebra of $G_{W(k)}$ induced by $g \mapsto \mu (t)^{-1}g\mu (t)$ are $\leq 1$ ([Reference LauLau21, Definition 6.3.1]). If G is reductive, then $\mu $ is 1-bounded if and only if $\mu $ is minuscule.

Definition 1.1.1. For a p-adically complete ring R over $W(k)$ , the groupoid of prismatic G- $\mu $ -displays over R is defined to be

Here, is the absolute prismatic site of R, whose objects are the bounded prisms $(A, I)$ equipped with a homomorphism $g \colon R \to A/I$ . For a prismatic G- $\mu $ -display $\mathfrak {Q}$ over R and an object , the image of $\mathfrak {Q}$ in is denoted by $\mathfrak {Q}_{(A, I)}$ or $\mathfrak {Q}_{g}$ .

Let $\mathcal {C}_{W(k)}$ be the category of complete regular local rings R over $W(k)$ with residue field k. Let $R \in \mathcal {C}_{W(k)}$ . The prism $(W(k), (p))$ with the natural homomorphism $R \to k$ is an object of , which we consider as ‘the base point’. Let $\mathcal {Q}$ be a G- $\mu $ -display over $(W(k), (p))$ (or equivalently, a prismatic G- $\mu $ -display over k).

Definition 1.1.2. A deformation of $\mathcal {Q}$ over R is a prismatic G- $\mu $ -display $\mathfrak {Q}$ over R together with an isomorphism $\mathfrak {Q}_{(W(k), (p))} \overset {\sim }{\to } \mathcal {Q}$ of G- $\mu $ -displays over $(W(k), (p))$ . We say that $\mathfrak {Q}$ is a universal deformation of $\mathcal {Q}$ if, for any $R' \in \mathcal {C}_{W(k)}$ and any deformation $\mathfrak {Q}'$ of $\mathcal {Q}$ over $R'$ , there exists a unique local homomorphism $ h \colon R \to R' $ over $W(k)$ such that the base change of $\mathfrak {Q}$ along h is isomorphic to $\mathfrak {Q}'$ as a deformation of $\mathcal {Q}$ over $R'$ .

Our first main result concerns the existence of a universal deformation of $\mathcal {Q}$ . Let $U^{-}_{\mu } \subset G_{W(k)}$ be the closed subgroup scheme consisting of elements $g \in G_{W(k)}$ such that $\lim _{t \to 0} \mu (t)^{-1}g\mu (t)=1$ . Let $R_{G, \mu }$ be the completed local ring of $U^{-}_{\mu }$ at the identity element $1 \in U^{-}_{\mu }$ . Since $U^{-}_{\mu }$ is smooth, there exists an isomorphism over $W(k)$

$$\begin{align*}R_{G, \mu} \simeq W(k)[[t_1, \dotsc, t_r]], \end{align*}$$

and in particular, $R_{G, \mu } \in \mathcal {C}_{W(k)}$ .

Theorem 1.1.3 (Theorem 4.1.7).

Let $\mathcal {Q}$ be a G- $\mu $ -display over $(W(k), (p))$ . There exists a universal deformation $\mathfrak {Q}^{\mathrm {univ}}$ of $\mathcal {Q}$ over $R_{G, \mu }$ .

We will show that $\mathfrak {Q}^{\mathrm {univ}}$ possesses a similar universal property for a larger class of deformations of $\mathcal {Q}$ . To explain this, we need to introduce some notation.

We consider a prism of Breuil–Kisin type

$$\begin{align*}(\mathfrak{S}, (\mathcal{E})):=(W(k)[[t_1, \dotsc, t_n]], (\mathcal{E})), \end{align*}$$

where $\mathcal {E} \in W(k)[[t_1, \dotsc , t_n]]$ is a formal power series whose constant term is p and the Frobenius $\phi $ of $W(k)[[t_1, \dotsc , t_n]]$ is such that $\phi (t_i)=t^p_i$ for $1 \leq i \leq n$ . Here, n could be any nonnegative integer. We set $ R:=\mathfrak {S}/\mathcal {E}, $ which belongs to the category $\mathcal {C}_{W(k)}$ .

The following fact plays a crucial role in our work.

Remark 1.1.4. Let $\mathfrak {Q}$ be a deformation of $\mathcal {Q}$ over R. The associated G- $\mu $ -display $\mathfrak {Q}_{(\mathfrak {S}, (\mathcal {E}))}$ over $(\mathfrak {S}, (\mathcal {E}))$ is a deformation of $\mathcal {Q}$ (i.e., it is equipped with an isomorphism between $\mathcal {Q}$ and the base change of $\mathfrak {Q}_{(\mathfrak {S}, (\mathcal {E}))}$ along the map $(\mathfrak {S}, (\mathcal {E})) \to (W(k), (p))$ defined by $t_i \mapsto 0$ ). The construction $\mathfrak {Q} \mapsto \mathfrak {Q}_{(\mathfrak {S}, (\mathcal {E}))}$ induces an equivalence from the category of deformations of $\mathcal {Q}$ over R to that of deformations of $\mathcal {Q}$ over $(\mathfrak {S}, (\mathcal {E}))$ . This is a consequence of [Reference ItoIto23, Theorem 1.2.1] (see also Theorem 2.6.5).

Let $\mathfrak {Q}$ be a deformation of $\mathcal {Q}$ over $R_{G, \mu }$ . Let $ {\mathrm {Hom}} (R_{G, \mu }, R)_e$ be the set of local homomorphisms $R_{G, \mu } \to R$ over $W(k)$ . Given a homomorphism $g \in {\mathrm {Hom}} (R_{G, \mu }, R)_e$ , we can regard $(\mathfrak {S}, (\mathcal {E}))$ as an object of . The construction $g \mapsto \mathfrak {Q}_g$ induces a map of sets

$$\begin{align*}{\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R_{G, \mu}, R)_e \to \left\{ \begin{array}{c} \displaystyle \text{isomorphism classes of} \\ \displaystyle \text{deformations of } \mathcal{Q} \text{ over } (\mathfrak{S}, (\mathcal{E})) \end{array} \right\}, \end{align*}$$

which we call the evaluation map. If $\mathfrak {Q}$ is a universal deformation, then the map $ {\mathrm {ev}} _{\mathfrak {Q}}$ is bijective by Remark 1.1.4. (In fact, $\mathfrak {Q}$ is a universal deformation if and only if for every prism $ (\mathfrak {S}, (\mathcal {E})) $ of Breuil–Kisin type, the map $ {\mathrm {ev}} _{\mathfrak {Q}}$ is bijective.)

We also consider the following classes of deformations:

Definition 1.1.5. For every integer $m \geq 1$ , we set

$$\begin{align*}(\mathfrak{S}_m, (\mathcal{E})):=(W(k)[[t_1, \dotsc, t_n]]/(t_1, \dotsc, t_n)^m, (\mathcal{E})), \end{align*}$$

which is naturally a bounded prism. Let $ {\mathrm {Hom}} (R_{G, \mu }, R/\mathfrak {m}^{m}_R)_e$ be the set of local homomorphisms $R_{G, \mu } \to R/\mathfrak {m}^{m}_R$ over $W(k)$ , where $\mathfrak {m}_R \subset R$ is the maximal ideal. We have the following map defined by $g \mapsto \mathfrak {Q}_g$ :

(1.1) $$ \begin{align} {\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R_{G, \mu}, R/\mathfrak{m}^{m}_R)_e \to \left\{ \begin{array}{c} \displaystyle \text{isomorphism classes of} \\ \displaystyle \text{deformations of } \mathcal{Q} \text{ over } (\mathfrak{S}_m, (\mathcal{E})) \end{array} \right\}. \end{align} $$

Definition 1.1.6. Let S be a perfectoid ring over $W(k)$ ([Reference Bhatt, Morrow and ScholzeBMS18, Definition 3.5]). Let $a^\flat \in S^\flat $ be an element such that $a:=\theta ([a^\flat ]) \in S$ is a nonzerodivisor and $p=0$ in $S/a$ . (See Section 2.1 for the notation used here.) For every integer $m \geq 1$ , the pair

$$\begin{align*}(W(S^\flat)/[a^\flat]^m, I_S) \end{align*}$$

is naturally a bounded prism, where $I_S$ is the kernel of the natural homomorphism $\theta \colon W(S^\flat )/[a^\flat ]^m \to S/a^m$ . Let $ {\mathrm {Hom}} (R_{G, \mu }, S/a^m)_{e}$ be the set of homomorphisms $R_{G, \mu } \to S/a^m$ over $W(k)$ lifting the composition $R_{G, \mu } \to k \to S/a$ . We regard $\mathcal {Q}$ as a G- $\mu $ -display over $(W(S^\flat )/[a^\flat ], I_S)$ by base change. We have the following map defined by $g \mapsto \mathfrak {Q}_g$ :

(1.2) $$ \begin{align} {\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R_{G, \mu}, S/a^m)_{e} \to \left\{ \begin{array}{c} \displaystyle \text{isomorphism classes of } \\ \displaystyle \text{deformations of } \mathcal{Q} \text{ over } (W(S^\flat)/[a^\flat]^m, I_S) \end{array} \right\}. \end{align} $$

We can now state our second main result:

Theorem 1.1.7 (Theorem 4.1.7).

If $\mathfrak {Q}$ is a universal deformation of $\mathcal {Q}$ over $R_{G, \mu }$ , then the maps (1.1) and (1.2) are bijective.

Using the prism $(W(k)[[t]]/t^2, (p))$ , we have the following characterization of universal deformations of $\mathcal {Q}$ .

Theorem 1.1.8 (Theorem 4.1.8).

Let $\mathfrak {Q}$ be a deformation of $\mathcal {Q}$ over $R_{G, \mu }$ . If the map

$$\begin{align*}{\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R_{G, \mu}, k[[t]]/t^2)_{e} \to \left\{ \begin{array}{c} \displaystyle \text{isomorphism classes of} \\ \displaystyle \text{deformations of } \mathcal{Q} \text{ over } (W(k)[[t]]/t^2, (p)) \end{array} \right\} \end{align*}$$

is surjective, then $\mathfrak {Q}$ is a universal deformation of $\mathcal {Q}$ .

The map $ {\mathrm {ev}} _{\mathfrak {Q}}$ as in Theorem 1.1.8 is called the Kodaira–Spencer map in this paper.

Remark 1.1.9. In fact, by using ${\mathcal O}_E$ -prisms as introduced in [Reference MarksMar23] and [Reference ItoIto23], we will formulate and prove our results for a smooth affine group scheme G over the ring of integers ${\mathcal O}_E$ of any finite extension E of ${\mathbb Q}_p$ . (In [Reference MarksMar23], ${\mathcal O}_E$ -prisms are called E-typical prisms.) The argument for general ${\mathcal O}_E$ is the same as that for the case where ${\mathcal O}_E={\mathbb Z}_p$ . For simplicity, the reader may assume that ${\mathcal O}_E={\mathbb Z}_p$ on a first reading.

Let R be a quasisyntomic ring over $W(k)$ in the sense of [Reference Bhatt, Morrow and ScholzeBMS19, Definition 4.10]. In [Reference ItoIto23, Section 8.2], we defined the groupoid of prismatic G-F-gauges of type $\mu $ over R and constructed a fully faithful functor

following the theory of prismatic F-gauges introduced by Drinfeld and Bhatt–Lurie (cf. [Reference DrinfeldDri22], [Reference Bhatt and LurieBL22a], [Reference Bhatt and LurieBL22b], [Reference BhattBha22]). If R belongs to $\mathcal {C}_{W(k)}$ , then the above functor is an equivalence (Proposition 7.1.1). Thus, Theorem 1.1.3 can be rephrased as follows:

Theorem 1.1.10 (Theorem 7.2.1).

Let be a prismatic G-F-gauge of type $\mu $ over k. There exists a deformation of $\mathscr {Q}$ over $R_{G, \mu }$ which is universal among deformations of $\mathscr {Q}$ over R with $R \in \mathcal {C}_{W(k)}$ .

Remark 1.1.11. In light of Theorem 1.1.7, one can expect that $\mathscr {Q}^{\mathrm {univ}}$ is also universal among deformations of $\mathscr {Q}$ over $R/\mathfrak {m}^{m}_R$ with $R \in \mathcal {C}_{W(k)}$ such that $\dim R =1$ and over $S/a^m$ as in Definition 1.1.6. We note that $R/\mathfrak {m}^{m}_R$ (with $\dim R =1$ ) and $S/a^m$ are quasisyntomic. In principle, it should be possible to develop the deformation theory for prismatic G-F-gauges of type $\mu $ in a more conceptual (but rather abstract) way, and the results presented in this paper should provide a practical way to understand it by using G- $\mu $ -displays over prisms. We will only make some conjectures in Section 7 and will not pursue this direction here. See also Remark 1.1.13.

Remark 1.1.12. Assume that G is reductive. In [Reference Bültel and PappasBP20], Bültel–Pappas introduced the notion of G- $\mu $ -displays over $ \mathrm {Spec} R$ for any $W(k)$ -algebra R by using the ring $W(R)$ of p-typical Witt vectors. (In [Reference Bültel and PappasBP20], G- $\mu $ -displays are called $(G, \mu )$ -displays.) The groupoid of G- $\mu $ -displays over $ \mathrm {Spec} k$ in their sense is equivalent to . Let $\mathcal {Q}$ be a (banal) G- $\mu $ -display over $ \mathrm {Spec} k$ . Under the assumption that $\mathcal {Q}$ is adjoint nilpotent in the sense of [Reference Bültel and PappasBP20, Definition 3.4.2], they proved that the deformation functor of $\mathcal {Q}$ defined on the category $ {\mathrm {Art}} _{W(k)}$ of artinian local rings over $W(k)$ with residue field k is pro-representable by $R_{G, \mu }$ . See [Reference Bültel and PappasBP20, 3.5.9] for details. The deformation theory developed in this paper applies to complete regular local rings $R \in \mathcal {C}_{W(k)}$ and their quotients $R/\mathfrak {m}^m_R$ without imposing any additional conditions on G- $\mu $ -displays, such as the adjoint nilpotent condition.

Remark 1.1.13. After this work was completed, and during the refereeing process, Gardner–Madapusi [Reference Gardner and MadapusiGM24] developed the deformation theory for prismatic G-F-gauges of type $\mu $ using the stacky approach of Drinfeld and Bhatt–Lurie, from which the conjectures stated in Section 7 will follow. More precisely, using their result, we can extend the definition of prismatic G-F-gauges of type $\mu $ to (not necessarily quasisyntomic) p-adically complete rings and then generalize our results to arbitrary artinian local rings in $ {\mathrm {Art}} _{W(k)}$ . Compared to [Reference Gardner and MadapusiGM24], the method developed in this paper cannot be applied to arbitrary artinian local rings, but it does not require the use of some powerful results on derived algebraic geometry, such as Lurie’s generalization of Artin’s representability theorem. See also [Reference Imai, Kato and YoucisIKY23] for the relation between our results and those in [Reference Gardner and MadapusiGM24].

1.2 Applications

We give two applications of our results. Assume that G is a connected reductive group scheme over ${\mathbb Z}_p$ (or over ${\mathcal O}_E$ ). The first application concerns the local representability and the formal smoothness of the integral local Shimura variety $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ with hyperspecial level structure, which is introduced in [Reference Scholze and WeinsteinSW20] as a v-sheaf on the category of perfectoid spaces of characteristic p. More precisely, we prove in Theorem 5.3.5 that the formal completions $\widehat {\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }}_{/ x}$ of $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ introduced in [Reference GleasonGle22b] are representable by the formal scheme $ \mathrm {Spf} R_{G, \mu }$ (or rather $ \mathrm {Spf} R_{G, \mu ^{-1}}$ , depending on the sign convention) by relating $\widehat {\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }}_{/ x}$ to universal deformations of prismatic G- $\mu $ -displays.

Remark 1.2.1. The above result (Theorem 5.3.5) implies that a conjecture of Pappas–Rapoport [Reference Pappas and RapoportPR24, Conjecture 3.3.5] in the hyperspecial case, which was originally proposed by Gleason [Reference GleasonGle21, Conjecture 1], holds true. In [Reference Pappas and RapoportPR22], Pappas–Rapoport proved their conjecture (in the more general case where G is parahoric) under the assumption that $p \geq 3$ and $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ is of abelian type. If $p=2$ , the same result was obtained under the additional assumption that $G_{{\mathbb Q}_p}$ is of type A or C. Our approach is completely different from that of [Reference Pappas and RapoportPR22], and we do not need to impose these assumptions. Our result also generalizes a result of Bartling [Reference BartlingBar22], in which the adjoint nilpotent condition of [Reference Bültel and PappasBP20] is imposed. See Section 5.3 for details.

Remark 1.2.2. In [Reference Imai, Kato and YoucisIKY23], our universal deformations of prismatic G- $\mu $ -displays are used to give a prismatic characterization of integral canonical models of (global) Shimura varieties of abelian type with hyperspecial level structure, which is inspired by [Reference Pappas and RapoportPR24, Theorem 4.2.4].

As a second application, we revisit and extend some classification results of p-divisible groups.

Remark 1.2.3. Let R be a quasisyntomic ring. In [Reference Anschütz and Le BrasALB23], Anschütz–Le Bras constructed a contravariant functor from the category of p-divisible groups over $ \mathrm {Spec} R$ to the category of admissible prismatic Dieudonné crystals over R ([Reference Anschütz and Le BrasALB23, Definition 4.5]), which is called the prismatic Dieudonné functor. They proved that the prismatic Dieudonné functor is an anti-equivalence; see [Reference Anschütz and Le BrasALB23, Theorem 4.74]. For any $R \in \mathcal {C}_{W(k)}$ , our prismatic $ {\mathrm {GL}} _N$ - $\mu $ -displays over R can be viewed as admissible prismatic Dieudonné crystals over R (with some additional conditions). Then Theorem 1.1.3 and Theorem 1.1.8 follow immediately from [Reference Anschütz and Le BrasALB23, Theorem 4.74], together with the deformation theory for p-divisible groups. However, Theorem 1.1.7 is a new result as far as we know. The proofs of Theorem 1.1.3, Theorem 1.1.7 and Theorem 1.1.8 presented in this paper do not rely on p-divisible groups.

By using our results, we can give an alternative proof of the classification result of Anschütz–Le Bras mentioned above for any $R \in \mathcal {C}_{W(k)}$ . In fact, Theorem 1.1.7 enables us to prove the following results:

  1. (1) (Theorem 6.3.2). Let the notation be as in Definition 1.1.5. For every $m \geq 1$ , the prismatic Dieudonné functor induces an anti-equivalence

    $$\begin{align*}\left. \left\{ \begin{array}{c} \displaystyle p\text{-divisible groups} \\ \text{over } \mathrm{Spec} R/\mathfrak{m}^m_R \end{array} \right\} \right. \overset{\sim}{\to} \left\{ \begin{array}{c} \displaystyle \text{minuscule Breuil--Kisin modules} \\ \displaystyle \text{over } (\mathfrak{S}_m, (\mathcal{E})) \end{array} \right\}. \end{align*}$$
  2. (2) (Theorem 6.3.7). Let ${\mathcal O}_C$ be a p-adically complete valuation ring of rank $1$ with algebraically closed fraction field C. Let $\varpi \in {\mathcal O}_C$ be a pseudo-uniformizer. For every $m \geq 1$ , the prismatic Dieudonné functor induces an anti-equivalence

    $$\begin{align*}\left. \left\{ \begin{array}{c} \displaystyle p\text{-divisible groups} \\ \text{ over } \mathrm{Spec} {\mathcal O}_C/\varpi^m \end{array} \right\} \right. \overset{\sim}{\to} \left\{ \begin{array}{c} \displaystyle \text{minuscule Breuil--Kisin modules} \\ \displaystyle \text{over } (W({\mathcal O}_{C^\flat})/[\varpi^\flat]^{m}, I_{{\mathcal O}_C}) \end{array} \right\}. \end{align*}$$

See Example 2.4.7 for the definition of minuscule Breuil–Kisin modules.

Remark 1.2.4. An equivalence as in (1) was previously obtained by Lau [Reference LauLau14] using Dieudonné displays and crystalline Dieudonné theory, where special efforts were made in the case of $p=2$ . (This result plays an essential role in the construction of $2$ -adic integral canonical models of Shimura varieties of abelian type with hyperspecial level structure in [Reference Kim and Madapusi PeraKMP16].)

By combining (2) with a theorem of Fargues [Reference Scholze and WeinsteinSW20, Theorem 14.1.1], we obtain an alternative proof of a result of Scholze–Weinstein [Reference Scholze and WeinsteinSW13, Theorem B], which says that, in the case where C is of characteristic $0$ , there exists an equivalence between the category of p-divisible groups over $ \mathrm { Spec} {\mathcal O}_C$ and the category of free ${\mathbb Z}_p$ -modules T of finite rank together with a C-subspace of $T \otimes _{{\mathbb Z}_p} C$ . See Section 6.3 for details.

1.3 Strategy of the proof

We remark that if there exists a deformation $\mathfrak {Q}$ of $\mathcal {Q}$ over $R_{G, \mu }$ such that the maps (1.1) and (1.2) for $\mathfrak {Q}$ are bijective, then $\mathfrak {Q}$ is a universal deformation, and moreover, Theorem 1.1.8 follows immediately. We briefly explain how to construct such a deformation $\mathfrak {Q}$ . We choose a certain prism $(\mathfrak {S}^{ {\mathrm {univ}} }, (\mathcal {E}^{ {\mathrm {univ}} }))$ of Breuil–Kisin type with an isomorphism $R_{G, \mu } \simeq \mathfrak {S}^{ {\mathrm {univ}} }/\mathcal {E}^{ {\mathrm {univ}} }$ over $W(k)$ . By Remark 1.1.4, giving a deformation of $\mathcal {Q}$ over $R_{G, \mu }$ is equivalent to giving a deformation of $\mathcal {Q}$ over $(\mathfrak {S}^{ {\mathrm {univ}} }, (\mathcal {E}^{ {\mathrm {univ}} }))$ . In practice, the latter is much easier. We will construct a deformation $\mathscr {Q}$ of $\mathcal {Q}$ over $(\mathfrak {S}^{ {\mathrm {univ}} }, (\mathcal {E}^{ {\mathrm {univ}} }))$ such that the maps (1.1) and (1.2) for the corresponding deformation $\mathfrak {Q}$ over $R_{G, \mu }$ are bijective.

The key ingredient is the Grothendieck–Messing deformation theory for G- $\mu $ -displays developed in Section 3. Let $\mathcal {Q}'$ be a G- $\mu $ -display over $(\mathfrak {S}_m, (\mathcal {E}))$ (resp. $(W(S^\flat )/[a^\flat ]^m, I_S)$ ). Then the Grothendieck–Messing deformation theory says that deformations of $\mathcal {Q}'$ over $(\mathfrak {S}_{m+1}, (\mathcal {E}))$ (resp. $(W(S^\flat )/[a^\flat ]^{m+1}, I_S)$ ) are classified by lifts of the Hodge filtration of $\mathcal {Q}'$ ; see Theorem 3.3.4 for the precise statement. Here, the 1-boundedness of $\mu $ is essential. In view of this result, it is not difficult to find a candidate for $\mathscr {Q}$ . In fact, our construction is an analogue of the construction of universal p-divisible groups given by Faltings [Reference FaltingsFal99, Section 7].

The proof of the bijectivity of the maps (1.1) and (1.2) for $\mathfrak {Q}$ goes as follows. We fix a homomorphism $g \in {\mathrm {Hom}} (R_{G, \mu }, S/a^{m})_{e}$ and let $ {\mathrm {Hom}} (R_{G, \mu }, S/a^{m+1})_{g}$ be the set of homomorphisms $h \colon R_{G, \mu } \to S/a^{m+1}$ over $W(k)$ which are lifts of g. To prove that the map (1.2) is bijective, it suffices to show that (for all m and g) the map

(1.3) $$ \begin{align} {\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R_{G, \mu}, S/a^{m+1})_{g} \to \left\{ \begin{array}{c} \displaystyle \text{isomorphism classes of} \\ \displaystyle \text{deformations of } \mathfrak{Q}_g \text{ over } (W(S^\flat)/[a^\flat]^{m+1}, I_S) \end{array} \right\} \end{align} $$

defined by $h \mapsto \mathfrak {Q}_h$ is bijective. After choosing a lift $h \in {\mathrm {Hom}} (R_{G, \mu }, S/a^{m+1})_{g}$ , we can endow the set $ {\mathrm {Hom}} (R_{G, \mu }, S/a^{m+1})_{g}$ with the structure of an $S/a$ -module such that h is the zero element. The target of the map (1.3) can be also regarded as an $S/a$ -module by the Grothendieck–Messing deformation theory. In fact, we will show that the map (1.3) is not only a bijection, but also an $S/a$ -linear isomorphism (cf. Theorem 4.4.2). For this, we may assume that any element of S admits a p-power root by André’s flatness lemma, and then we can prove the assertion by explicit computation. From this more precise statement, we can deduce that the map (1.1) is bijective (cf. Proposition 4.3.10).

1.4 The structure of this paper

This paper is organized as follows. In Section 2, we review the theory of ${\mathcal O}_E$ -prisms and G- $\mu $ -displays over (bounded) ${\mathcal O}_E$ -prisms. In Section 3, we prove some basic results on deformations of G- $\mu $ -displays and establish the Grothendieck–Messing deformation theory for G- $\mu $ -displays. In Section 4, we state and prove the main results of this paper.

In Section 5 and Section 6, we give applications of our results to integral local Shimura varieties and p-divisible groups, respectively. Finally, in Section 7, we discuss some consequences of our deformation theory on prismatic F-gauges, or more precisely, on prismatic G-F-gauges of type $\mu $ .

Notation

This paper is a continuation of our previous work [Reference ItoIto23]. Unless explicitly stated otherwise, we use definitions and notation from [Reference ItoIto23].

We recall some notation from [Reference ItoIto23]. All rings are commutative and unital. For a scheme X over $ \mathrm {Spec} R$ and a ring homomorphism $f \colon R \to R'$ , the base change $X \times _{ \mathrm {Spec} R} \mathrm {Spec} R'$ is denoted by $X_{R'}$ or $f^*X$ . We use similar notation for the base change of group schemes, p-divisible groups, etc. All actions of groups will be right actions, unless otherwise stated. A groupoid is a category whose morphisms are all invertible. For a category $\mathcal {C}$ , let $\mathcal {C}^{\simeq }$ be the largest groupoid contained in $\mathcal {C}$ .

2 Review of prismatic G- $\mu $ -displays

Throughout this paper, we fix a prime number p. Let E be a finite extension of ${\mathbb Q}_p$ with ring of integers ${\mathcal O}_E$ and residue field ${\mathbb F}_q$ . Here, ${\mathbb F}_q$ is a finite field with q elements. We fix a uniformizer $\pi \in {\mathcal O}_E$ for simplicity. Let k be a perfect field containing ${\mathbb F}_q$ , and we set $ {\mathcal O} := W(k) \otimes _{W({\mathbb F}_q)} {\mathcal O}_E. $ Let G be a smooth affine group scheme over ${\mathcal O}_E$ and let $ \mu \colon {\mathbb G}_m \to G_{{\mathcal O}}:=G \times _{ \mathrm {Spec} {\mathcal O}_E} \mathrm {Spec} {\mathcal O} $ be a cocharacter.

In this section, we first recall the notion of ${\mathcal O}_E$ -prisms as introduced in [Reference MarksMar23] and [Reference ItoIto23]. We also provide some preliminary results on certain specific ${\mathcal O}_E$ -prisms which will be used in this paper. Then we review the definition and basic properties of G- $\mu $ -displays over bounded ${\mathcal O}_E$ -prisms developed in [Reference ItoIto23]. More details can be found in [Reference ItoIto23].

2.1 ${\mathcal O}_E$ -prisms

Let A be an ${\mathcal O}_E$ -algebra. A $\delta _E$ -structure on A is a map $\delta _E \colon A \to A$ of sets with the following properties:

  • $\delta _E(xy)=x^q\delta _E(y)+y^q\delta _E(x)+\pi \delta _E(x)\delta _E(y)$ .

  • $\delta _E(x+y)=\delta _E(x)+\delta _E(y)+(x^q+y^q-(x+y)^q)/\pi $ .

  • For an element $x \in {\mathcal O}_E$ , we have $\delta _E(x)=(x-x^q)/\pi $ .

A $\delta _E$ -ring is an ${\mathcal O}_E$ -algebra A equipped with a $\delta _E$ -structure. We define

$$\begin{align*}\phi_A \colon A \to A, \quad x \mapsto x^q+\pi\delta_E(x). \end{align*}$$

Then $\phi _A$ is a homomorphism of ${\mathcal O}_E$ -algebras and is a lift of the q-th power Frobenius $A/\pi \to A/\pi $ , $x \mapsto x^q$ . The homomorphism $\phi _A$ is called the Frobenius of the $\delta _E$ -ring A. When there is no ambiguity, we omit the subscript and simply write $\phi =\phi _A$ .

An ${\mathcal O}_E$ -prism is a pair $(A, I)$ of a $\delta _E$ -ring A and a Cartier divisor $I \subset A$ such that A is derived $(\pi , I)$ -adically complete and $\pi \in I + \phi (I)A$ . We say that $(A, I)$ is bounded if $A/I$ has bounded $p^\infty $ -torsion (i.e., $(A/I)[p^n]=(A/I)[p^\infty ]$ for some integer $n \geq 1$ ). In this case, A is $(\pi , I)$ -adically complete ([Reference ItoIto23, Remark 2.3.2]). If ${\mathcal O}_E={\mathbb Z}_p$ , then bounded ${\mathcal O}_E$ -prisms are the same as bounded prisms in the sense of [Reference Bhatt and ScholzeBS22]. We say that $(A, I)$ is orientable if I is principal. An ${\mathcal O}_E$ -prism $(A, I)$ with a homomorphism ${\mathcal O} \to A$ of $\delta _E$ -rings is called an ${\mathcal O}_E$ -prism over ${\mathcal O}$ . We refer to [Reference MarksMar23] and [Reference ItoIto23] for details.

We give some examples of ${\mathcal O}_E$ -prisms, which play a central role in this paper. We set

$$\begin{align*}\mathfrak{S}_{\mathcal O}:={\mathcal O}[[t_1, \dotsc, t_n]] \end{align*}$$

for $n \geq 0$ , which admits a unique $\delta _E$ -structure such that the Frobenius $\phi \colon \mathfrak {S}_{\mathcal O} \to \mathfrak {S}_{\mathcal O}$ is given by $\phi (t_i)=t^q_i$ ( $1 \leq i \leq n$ ). For every integer $m \geq 1$ , the quotient

$$\begin{align*}\mathfrak{S}_{{\mathcal O}, m}:={\mathcal O}[[t_1, \dotsc, t_n]]/(t_1, \dotsc, t_n)^m \end{align*}$$

admits a unique $\delta _E$ -structure that is compatible with the one on $\mathfrak {S}_{\mathcal O}$ .

Example 2.1.1. Let $\mathcal {E} \in \mathfrak {S}_{\mathcal O}$ be a formal power series whose constant term is a uniformizer of ${\mathcal O}$ . The pair $ (\mathfrak {S}_{\mathcal O}, (\mathcal {E})) $ is a bounded ${\mathcal O}_E$ -prism over ${\mathcal O}$ ([Reference ItoIto23, Proposition 2.3.8]), which we call an ${\mathcal O}_E$ -prism of Breuil–Kisin type in this paper. Here, n could be any nonnegative integer. For any $m \geq 1$ , the pair $ (\mathfrak {S}_{{\mathcal O}, m}, (\mathcal {E})) $ is also a bounded ${\mathcal O}_E$ -prism over ${\mathcal O}$ . Here, we denote the image of $\mathcal {E}$ in $\mathfrak {S}_{{\mathcal O}, m}$ by the same symbol.

Our next example is related to (integral) perfectoid rings. We refer to [Reference Bhatt, Morrow and ScholzeBMS18, Section 3] and [Reference Česnavičius and ScholzeČS24, Section 2] for the definition and basic properties of perfectoid rings.

Let S be a perfectoid ring over ${\mathcal O}$ (i.e., S is a perfectoid ring with a ring homomorphism ${\mathcal O} \to S$ ). Let $ S^\flat :=\varprojlim _{x \mapsto x^p} S/p $ be the tilt of S, which is a perfect k-algebra. Then

$$\begin{align*}W_{{\mathcal O}_E}(S^\flat):=W(S^\flat) \otimes_{W({\mathbb F}_q)} {\mathcal O}_E \end{align*}$$

is naturally an ${\mathcal O}$ -algebra. Let $\phi \colon W_{{\mathcal O}_E}(S^\flat ) \to W_{{\mathcal O}_E}(S^\flat )$ be the base change of the q-th power Frobenius of $W(S^\flat )$ . Since $W_{{\mathcal O}_E}(S^\flat )$ is $\pi $ -torsion free, we obtain the corresponding $\delta _E$ -structure on $W_{{\mathcal O}_E}(S^\flat )$ . Moreover, for any element $x \in S^\flat $ , the quotient $W_{{\mathcal O}_E}(S^\flat )/[x]$ admits a unique $\delta _E$ -structure that is compatible with the one on $W_{{\mathcal O}_E}(S^\flat )$ . Here, $[-]$ denotes the Teichmüller lift.

We note the following fact.

Lemma 2.1.2. For any element $x \in S^\flat $ , the quotient $W_{{\mathcal O}_E}(S^\flat )/[x]$ is $\pi $ -torsion free (or equivalently, p-torsion free).

Proof. Since $W({\mathbb F}_q) \to {\mathcal O}_E$ is flat, it is enough to show that $W(S^\flat )/[x]$ is p-torsion free. Let $y \in W(S^\flat )$ be an element such that $py=[x]z$ for some $z \in W(S^\flat )$ . We want to show that $y \in ([x])$ . For the Witt vector expansions $y=(y_0, y_1, \dotsc )$ and $z=(z_0, z_1, \dotsc )$ , the equality $py=[x]z$ implies that $y^p_i=x^{p^{i+1}}z_{i+1}$ for every $i \geq 0$ . Since $S^\flat $ is perfect, we have $y_i=x^{p^{i}}z^{1/p}_{i+1}$ for every $i \geq 0$ , and thus, $ y=(y_0, y_1, \dotsc )=[x]\cdot (z^{1/p}_{1}, z^{1/p}_{2}, \dotsc ) \in ([x]). $

Let $ \theta \colon W(S^\flat ) \to S $ be the unique ring homomorphism whose reduction modulo p is the projection map $S^\flat \to S/p$ , $(x_0, x_1, \dotsc ) \mapsto x_0$ . Let $\theta _{{\mathcal O}_E} \colon W_{{\mathcal O}_E}(S^\flat ) \to S$ be the homomorphism induced from $\theta $ . We write $I_S:= {\mathrm {Ker}} \theta _{{\mathcal O}_E}$ for the kernel of $\theta _{{\mathcal O}_E}$ .

Proposition 2.1.3. Let S be a perfectoid ring over ${\mathcal O}$ .

  1. (1) The pair $ (W_{{\mathcal O}_E}(S^\flat ), I_S) $ is an orientable and bounded ${\mathcal O}_E$ -prism over ${\mathcal O}$ .

  2. (2) Let $a^\flat \in S^\flat $ be an element such that $a:=\theta ([a^\flat ]) \in S$ is a nonzerodivisor and we have $\pi \in (a)$ in S. Let $m \geq 1$ be an integer. Then

    $$\begin{align*}(W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^m, I_S) \end{align*}$$
    is an orientable and bounded ${\mathcal O}_E$ -prism over ${\mathcal O}$ . Here, we denote the image of $I_S$ in $W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m$ by the same notation.

Proof. (1) See [Reference ItoIto23, Proposition 2.4.3].

(2) We write $A_m:=W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m$ . Let $\xi \in I_S$ be a generator. We shall show that $\xi $ is a nonzerodivisor in $A_m$ . Let $x, y \in W_{{\mathcal O}_E}(S^\flat )$ be elements satisfying $\xi x=[a^\flat ]^m y$ . Since $a \in S$ is a nonzerodivisor, we have $y \in I_S=(\xi )$ . Since $\xi $ is a nonzerodivisor in $W_{{\mathcal O}_E}(S^\flat )$ , we obtain $x \in ([a^\flat ]^m)$ . This proves that $\xi $ is a nonzerodivisor in $A_m$ . Moreover, since $\pi \in (\xi , \phi (\xi ))$ in $W_{{\mathcal O}_E}(S^\flat )$ by (1), we have $\pi \in (\xi , \phi (\xi ))$ in $A_m$ as well.

The above argument, together with the fact that $W_{{\mathcal O}_E}(S^\flat )$ is $\xi $ -adically complete, implies that $W_{{\mathcal O}_E}(S^\flat )$ is $[a^\flat ]$ -torsion free. Since $W_{{\mathcal O}_E}(S^\flat )$ is derived $(\pi , \xi )$ -adically complete, so is $A_m$ . It is clear that $A_m/\xi =S/a^m$ has bounded $p^\infty $ -torsion (since $p^m$ =0 in $S/a^m$ ). In conclusion, $(A_m, I_S)$ is an orientable and bounded ${\mathcal O}_E$ -prism.

For the purposes of this paper, it will be convenient to introduce the following (slightly nonstandard) definition:

Definition 2.1.4. A pair $(S, a^\flat )$ consisting of a perfectoid ring S over ${\mathcal O}$ and an element $a^\flat \in S^\flat $ is called a perfectoid pair over ${\mathcal O}$ if $a:=\theta ([a^\flat ]) \in S$ is a nonzerodivisor and we have $\pi \in (a)$ in S.

2.2 Prismatic sites

We say that a map $(A, I) \to (A', I')$ of bounded ${\mathcal O}_E$ -prisms is a flat map (resp. a faithfully flat map) if $A \to A'$ is $(\pi ,I)$ -completely flat (resp. $(\pi ,I)$ -completely faithfully flat) in the sense of [Reference Bhatt and ScholzeBS22, Notation 1.2].

Let R be a $\pi $ -adically complete ${\mathcal O}_E$ -algebra. As in [Reference ItoIto23, Definition 2.5.3], let

be the category of bounded ${\mathcal O}_E$ -prisms $(A, I)$ together with a homomorphism $R \to A/I$ of ${\mathcal O}_E$ -algebras. We endow the opposite category with the flat topology – that is, the topology generated by the faithfully flat maps. It follows from [Reference ItoIto23, Remark 2.5.4] that is a site. If ${\mathcal O}_E={\mathbb Z}_p$ , then is the same as the category introduced in [Reference Bhatt and ScholzeBS22, Remark 4.7].

Remark 2.2.1. The functors

form sheaves with respect to the flat topology. Here, $\mathrm {Set}$ is the category of sets.

Let $(A, I)$ be a bounded ${\mathcal O}_E$ -prism. We write

$$\begin{align*}(A, I)_{\mathrm{\acute{e}t}} \end{align*}$$

for the category of $(\pi , I)$ -completely étale A-algebras (in the sense of [Reference Bhatt and ScholzeBS22, Notation 1.2]). We endow $ (A, I)^{ {\mathrm {op}} }_{\mathrm {\acute {e}t}} $ with the $(\pi , I)$ -completely étale topology – that is, the topology generated by the $(\pi , I)$ -completely étale coverings $B \to B'$ (i.e., $B \to B'$ is $(\pi , I)$ -completely étale and $(\pi , I)$ -completely faithfully flat). Every $B \in (A, I)_{\mathrm {\acute {e}t}}$ admits a unique $\delta _E$ -structure compatible with that on A, and the pair $(B, IB)$ is a bounded ${\mathcal O}_E$ -prism by [Reference ItoIto23, Lemma 2.5.10].

We recall the following fact (see [Reference ItoIto23, Example 2.5.11] and the references therein):

Example 2.2.2. Let S be a perfectoid ring over ${\mathcal O}_E$ and let $S \to S'$ be a $\pi $ -completely étale homomorphism. Then $S'$ is a perfectoid ring, and the induced homomorphism $W_{{\mathcal O}_E}(S^\flat ) \to W_{{\mathcal O}_E}(S^{\prime \flat })$ is $(\pi , I_S)$ -completely étale.

Lemma 2.2.3. Let $(S, a^\flat )$ be a perfectoid pair over ${\mathcal O}$ .

  1. (1) For a $\pi $ -completely flat S-algebra $S'$ , the element a is a nonzerodivisor in $S'$ . In particular, if $S'$ is a perfectoid ring, then $(S', a^\flat )$ is a perfectoid pair over ${\mathcal O}$ .

  2. (2) For a $\pi $ -completely étale S-algebra $S'$ , the homomorphism

    $$\begin{align*}W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^m \to W_{{\mathcal O}_E}(S^{\prime\flat})/[a^\flat]^m \end{align*}$$
    is $(\pi , I_S)$ -completely étale for every $m \geq 1$ . Conversely, any $(\pi , I_S)$ -completely étale $W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m$ -algebra is of the form $W_{{\mathcal O}_E}(S^{\prime \flat })/[a^\flat ]^m$ for some $\pi $ -completely étale S-algebra $S'$ .

Proof. (1) Since $\pi =0$ in $S/a$ , we see that $S' \otimes ^{{\mathbb L}}_S S/a$ is concentrated in degree $0$ . This means that a is a nonzerodivisor in $S'$ .

(2) This immediately follows from (1). See also the proof of [Reference ItoIto23, Lemma 2.5.9].

2.3 Display groups

Let A be an ${\mathcal O}$ -algebra and $I \subset A$ an ideal generated by a nonzerodivisor $d \in A$ . We assume that A is I-adically complete. We set $A[1/I]:=A[1/d]$ . As in [Reference ItoIto23, Section 4], the display group $G_\mu (A, I)$ is defined by

$$\begin{align*}G_\mu(A, I):=\{ \, g \in G(A) \, \vert \, \mu(d)g\mu(d)^{-1} \, \, \text{lies in} \, \, G(A) \subset G(A[1/I]) \, \}. \end{align*}$$

We shall recall a structural result about $G_\mu (A, I)$ .

Let $P_\mu , U^{-}_{\mu } \subset G_{\mathcal O}$ be the closed subgroup schemes defined by, for every ${\mathcal O}$ -algebra R,

$$ \begin{align*} P_\mu(R)&=\{ \, g \in G(R) \, \vert \, \lim_{t \to 0} \mu(t)g\mu(t)^{-1} \, \text{exists} \, \},\\ U^{-}_{\mu}(R)&=\{ \, g \in G(R) \, \vert \, \lim_{t \to 0} \mu(t)^{-1}g\mu(t)=1 \, \}. \end{align*} $$

The group schemes $P_\mu $ and $U^{-}_{\mu }$ are smooth over ${\mathcal O}$ .

Definition 2.3.1 [Reference LauLau21, Definition 6.3.1].

The cocharacter $\mu \colon {\mathbb G}_m \to G_{{\mathcal O}}$ is called 1-bounded if the weights of the action of ${\mathbb G}_m$ on the Lie algebra $ {\mathrm {Lie}} (G_{\mathcal O})$ induced by $g \mapsto \mu (t)^{-1}g\mu (t)$ are $\leq 1$ .

Remark 2.3.2. If G is a reductive group scheme over ${\mathcal O}_E$ , then $\mu $ is 1-bounded if and only if $\mu $ is minuscule.

Proposition 2.3.3.

  1. (1) We have $P_\mu (A) \subset G_\mu (A, I)$ , and the image of $G_\mu (A, I)$ under the the projection $G(A) \to G(A/I)$ is equal to $P_\mu (A/I)$ .

  2. (2) The multiplication map

    $$\begin{align*}(U^{-}_{\mu}(A) \cap G_\mu(A, I)) \times P_\mu(A) \to G_\mu(A, I) \end{align*}$$
    is bijiective.
  3. (3) Assume that $\mu $ is 1-bounded. Then $G_\mu (A, I)$ coincides with the inverse image of $P_\mu (A/I)$ in $G(A)$ under the projection $G(A) \to G(A/I)$ . Moreover, we have the following bijection:

    $$\begin{align*}G(A)/G_\mu(A, I) \overset{\sim}{\to} G(A/I)/P_\mu(A/I). \end{align*}$$

Proof. The assertion (1) follows from [Reference ItoIto23, Lemma 4.2.2]. The assertion (2) is [Reference ItoIto23, Proposition 4.2.8]. For the assertion (3), see [Reference ItoIto23, Proposition 4.2.9].

2.4 G- $\mu $ -displays

Let $(A, I)$ be a bounded ${\mathcal O}_E$ -prism over ${\mathcal O}$ . We recall the definition of G- $\mu $ -displays over $(A, I)$ .

We assume for a moment that $(A, I)$ is orientable. The results in Section 2.3 apply to $(A, I)$ . In particular, we have the display group $G_\mu (A, I)$ . For each generator $d \in I$ , we define the following homomorphism:

$$\begin{align*}\sigma_{\mu, d} \colon G_\mu(A, I) \to G(A), \quad g \mapsto \phi(\mu(d)g\mu(d)^{-1}). \end{align*}$$

Let $G(A)_d$ be the set $G(A)$ together with the following action of $G_\mu (A, I)$ :

$$\begin{align*}G(A) \times G_\mu(A, I) \to G(A), \quad (X, g) \mapsto g^{-1}X\sigma_{\mu, d}(g). \end{align*}$$

For another generator $d' \in I$ , we have $d=ud'$ for a unique $u \in A^\times $ . The map $G(A)_d \to G(A)_{d'}$ defined by $X \mapsto X\phi (\mu (u))$ is $G_\mu (A, I)$ -equivariant. Thus, we can define

$$\begin{align*}G(A)_I := {\varprojlim}_{d} G(A)_d, \end{align*}$$

where d runs over the set of generators $d \in I$ . The set $G(A)_I$ carries a natural action of $G_\mu (A, I)$ . The projection map $G(A)_I \to G(A)_d$ is a $G_\mu (A, I)$ -equivariant bijection. For an element $X \in G(A)_I$ , let

(2.1) $$ \begin{align} X_d \in G(A)_d \end{align} $$

denote the image of X.

Let $G_{\mu , A, I}$ and be the sheaves on $(A, I)^{ {\mathrm {op}} }_{\mathrm {\acute {e}t}}$ defined by

respectively. The sheaf is equipped with a natural action of the group sheaf $G_{\mu , A, I}$ . In fact, we can extend these definitions to (not necessarily orientable) bounded ${\mathcal O}_E$ -prisms $(A, I)$ over ${\mathcal O}$ ; see [Reference ItoIto23, Section 4.3] for details.

Definition 2.4.1 (G- $\mu $ -display).

Let $(A, I)$ be a bounded ${\mathcal O}_E$ -prism over ${\mathcal O}$ . A G- $\mu $ -display over $(A, I)$ is a pair

$$\begin{align*}(\mathcal{Q}, \alpha_{\mathcal{Q}}), \end{align*}$$

where $\mathcal {Q}$ is a $G_{\mu , A, I}$ -torsor and is a $G_{\mu , A, I}$ -equivariant map of sheaves on $(A, I)^{ {\mathrm {op}} }_{\mathrm {\acute {e}t}}$ . When there is no possibility of confusion, we write $\mathcal {Q}$ instead of $(\mathcal {Q}, \alpha _{\mathcal {Q}})$ . We say that $(\mathcal {Q}, \alpha _{\mathcal {Q}})$ is banal if $\mathcal {Q}$ is a trivial $G_{\mu , A, I}$ -torsor.

Isomorphisms between G- $\mu $ -displays over $(A, I)$ are defined in the obvious way. We write

for the groupoid of G- $\mu $ -displays over $(A, I)$ and the groupoid of banal G- $\mu $ -displays over $(A, I)$ , respectively.

Remark 2.4.2. Assume that $(A, I)$ is orientable. Let $ [G(A)_I/G_\mu (A, I)] $ be the groupoid whose objects are the elements $X \in G(A)_I$ and whose morphisms are defined by

$$\begin{align*}{\mathrm{Hom}}(X, X')=\{\, g \in G_\mu(A, I) \, \vert \, X'\cdot g=X \, \}. \end{align*}$$

Here, $(-)\cdot g$ denotes the action of $g \in G_\mu (A, I)$ . To each $X \in G(A)_I$ , we attach a banal G- $\mu $ -display

$$\begin{align*}\mathcal{Q}_X:=(G_{\mu, A, I}, \alpha_X) \end{align*}$$

over $(A, I)$ , where is given by $1 \mapsto X$ . This construction gives an equivalence of groupoids.

For a map $f \colon (A, I) \to (A', I')$ of bounded ${\mathcal O}_E$ -prisms over ${\mathcal O}$ , we have a base change functor

More precisely, we have a fibered category

over .

Remark 2.4.3. Assume that $(A, I)$ is orientable. For an element $X \in G(A)_I$ , let $f(X) \in G(A')_{I'}$ be the unique element such that for any generator $d \in I$ , we have $f(X)_{f(d)}=f(X_d)$ in $G(A')$ . Then we have $f^*(\mathcal {Q}_X)=\mathcal {Q}_{f(X)}$ for every $X \in G(A)_I$ .

In the following, we recall more concrete descriptions of G- $\mu $ -displays. Let $(A, I)$ be a bounded ${\mathcal O}_E$ -prism over ${\mathcal O}$ .

Definition 2.4.4. A G-Breuil–Kisin module over $(A, I)$ is a G-torsor $\mathcal {P}$ over $ \mathrm {Spec} A$ (with respect to the étale topology) with an isomorphism $ F_{\mathcal {P}} \colon (\phi ^*\mathcal {P})[1/I] \overset {\sim }{\to } \mathcal {P}[1/I] $ of G-torsors over $ \mathrm {Spec} A[1/I]$ . We call $F_{\mathcal {P}}$ the Frobenius of $\mathcal {P}$ . We say that $\mathcal {P}$ is of type $\mu $ if $(\pi , I)$ -completely étale locally on A, there exists some trivialization $\mathcal {P} \simeq G_A$ under which the isomorphism $F_{\mathcal {P}}$ is given by $g \mapsto Yg$ for an element Y in the double coset

$$\begin{align*}G(A)\mu(d)G(A) \subset G(A[1/I]), \end{align*}$$

where $d \in I$ is a generator.

Proposition 2.4.5. There is an equivalence $\mathcal {Q} \mapsto \mathcal {Q}_{\mathrm {BK}}$ from the groupoid to the groupoid of G-Breuil–Kisin modules of type $\mu $ over $(A, I)$ . This equivalence is compatible with base change along any map $(A, I) \to (A', I')$ .

Proof. See [Reference ItoIto23, Proposition 5.3.8]. For the construction of the functor $\mathcal {Q} \mapsto \mathcal {Q}_{\mathrm {BK}}$ , see [Reference ItoIto23, Definition 5.3.6].

Example 2.4.6. Assume that $(A, I)$ is orientable. Let $d \in I$ be a generator. For an element $X \in G(A)_I$ , the trivial G-torsor $G_A$ with the isomorphism

$$\begin{align*}(\phi^*G_A)[1/I]=G_A[1/I] \overset{\sim}{\to} G_A[1/I], \quad g \mapsto (\mu(d)X_d)g \end{align*}$$

is a G-Breuil–Kisin module of type $\mu $ over $(A, I)$ , and this is isomorphic to the one $(\mathcal {Q}_X)_{\mathrm {BK}}$ associated with .

Example 2.4.7 (Minuscule Breuil–Kisin module).

We assume that $G= {\mathrm {GL}} _N$ . Let $\mu \colon {\mathbb G}_m \to {\mathrm {GL}} _{N}$ be the cocharacter defined by

$$\begin{align*}t \mapsto {\mathrm{diag}}{(\underbrace{t, \dotsc, t}_s, \underbrace{1, \dotsc, 1}_{N-s})}, \end{align*}$$

which is 1-bounded. A minuscule Breuil–Kisin module over $(A, I)$ is a finite projective A-module M together with an A-linear homomorphism

$$\begin{align*}F_M \colon \phi^*M=A \otimes_{\phi, A} M \to M \end{align*}$$

whose cokernel is killed by I. We set $ {\mathrm {Fil}} ^1(\phi ^*M):= \{ \, x \in \phi ^*M \, \vert \, F_M(x) \in IM \, \} $ and let $ P^1 \subset (\phi ^*M)/I(\phi ^*M) $ be the image of $ {\mathrm {Fil}} ^1(\phi ^*M)$ . By [Reference ItoIto23, Proposition 3.1.6], we see that $P^1$ is a direct summand of $(\phi ^*M)/I(\phi ^*M)$ . We say that M is of type $\mu $ if the rank of M (resp. $P^1$ ) is constant and equal to N (resp. s). Let $ \mathrm {BK}_\mu (A, I)^{\simeq } $ be the groupoid of minuscule Breuil–Kisin modules over $(A, I)$ of type $\mu $ . In [Reference ItoIto23, Example 5.3.3, Corollary 5.3.11], we constructed a natural equivalence of groupoids:

We set $ {\mathrm {Fil}} ^1_\mu := A^s \oplus I^{N-s} \subset A^N$ . Then the underlying $G_{\mu , A, I}$ -torsor of $\mathcal {Q}(M)$ is the functor

$$\begin{align*}\underline{\mathrm{Isom}}_{{\mathrm{Fil}}}(A^N, \phi^*M) \colon (A, I)_{\mathrm{\acute{e}t}} \to \mathrm{Set} \end{align*}$$

defined by sending $B \in (A, I)_{\mathrm {\acute {e}t}}$ to the set of isomorphisms $B^N \overset {\sim }{\to } (\phi ^*M) \otimes _A B$ under which $ {\mathrm {Fil}} ^1_\mu $ agrees with $ {\mathrm {Fil}} ^1(\phi ^*M)$ .

2.5 Hodge filtrations and G- $\phi $ -modules

Let $(A, I)$ be an orientable and bounded ${\mathcal O}_E$ -prism over ${\mathcal O}$ . Here, we recall the definitions of the Hodge filtration $P(\mathcal {Q})_{A/I}$ and the underlying G- $\phi $ -module $\mathcal {Q}_\phi $ of a G- $\mu $ -display $\mathcal {Q}$ over $(A, I)$ introduced in [Reference ItoIto23].

Let , , and be the sheaves on $(A, I)^{ {\mathrm {op}} }_{\mathrm {\acute {e}t}}$ defined by

respectively. Let be the inclusion. The composition of $\tau $ with the projection is denoted by $\overline {\tau }$ , which factors through a homomorphism by Proposition 2.3.3. We have a commutative diagram

Remark 2.5.1. For a G-torsor $\mathcal {P}$ over $ \mathrm {Spec} A$ (with respect to the étale topology), the sheaf on $(A, I)^{ {\mathrm {op}} }_{\mathrm {\acute {e}t}}$ defined by $B \mapsto \mathcal {P}(B)$ is a -torsor. By [Reference ItoIto23, Proposition 4.3.1], this construction induces an equivalence of categories

We will make no distinction between a G-torsor over $ \mathrm {Spec} A$ and the corresponding -torsor. Similarly, the category of G-torsors over $ \mathrm {Spec} A/I$ (resp. $P_\mu $ -torsors over $ \mathrm {Spec} A/I$ ) is equivalent to that of -torsors (resp. -torsors).

Let . We write

$$\begin{align*}\mathcal{Q}_{A}:=\mathcal{Q}^{\tau} \quad (\text{resp.}\ \mathcal{Q}_{A/I}:=\mathcal{Q}^{\overline{\tau}},\, \text{resp.}\ P(\mathcal{Q})_{A/I}:=\mathcal{Q}^{\overline{\tau}_P}) \end{align*}$$

for the pushout of $\mathcal {Q}$ along $\tau $ (resp. $\overline {\tau }$ , resp. $\overline {\tau }_P$ ).

Definition 2.5.2. We regard $P(\mathcal {Q})_{A/I}$ as a $P_\mu $ -torsor over $ \mathrm {Spec} A/I$ . We call $P(\mathcal {Q})_{A/I}$ (or the morphism $P(\mathcal {Q})_{A/I} \to \mathcal {Q}_{A/I}$ ) the Hodge filtration of $\mathcal {Q}_{A/I}$ . We also say that $P(\mathcal {Q})_{A/I}$ is the Hodge filtration of $\mathcal {Q}$ .

We recall the following useful fact:

Proposition 2.5.3. A G- $\mu $ -display $\mathcal {Q}$ over $(A, I)$ is banal if and only if the Hodge filtration $P(\mathcal {Q})_{A/I}$ is a trivial $P_{\mu }$ -torsor over $ \mathrm {Spec} A/I$ .

Proof. See [Reference ItoIto23, Proposition 5.4.5].

A G- $\phi $ -module over $(A, I)$ is a G-torsor $\mathcal {P}$ over $ \mathrm {Spec} A$ with an isomorphism

$$\begin{align*}\phi_{\mathcal{P}} \colon (\phi^*\mathcal{P}) \times_{\mathrm{Spec} A} \mathrm{Spec} A[1/\phi(I)] \overset{\sim}{\to} \mathcal{P} \times_{\mathrm{Spec} A} \mathrm{Spec} A[1/\phi(I)] \end{align*}$$

of G-torsors over $ \mathrm {Spec} A[1/\phi (I)]$ . Here, we set $A[1/\phi (I)]:=A[1/\phi (d)]$ for a generator $d \in I$ . We call $\phi _{\mathcal {P}}$ the Frobenius of $\mathcal {P}$ .

In [Reference ItoIto23, Section 5.5], we constructed a functor

If $\mathcal {Q}=\mathcal {Q}_X$ is the banal G- $\mu $ -display over $(A, I)$ associated with an element $X \in G(A)_I$ (see Remark 2.4.2), then the isomorphism $\phi _{\mathcal {Q}_A}$ is defined as follows. We set

(2.2) $$ \begin{align} X_\phi := X_d\phi(\mu(d)) \in G(A[1/\phi(I)]), \end{align} $$

which is independent of the choice of a generator $d \in I$ . (See (2.1) for the notation $X_d$ .) We have $\mathcal {Q}_A=G_A$ , and hence, $\phi ^*(\mathcal {Q}_A)=G_A$ . The isomorphism $\phi _{\mathcal {Q}_A}$ is then given by $ g \mapsto X_\phi g. $

Definition 2.5.4. The G- $\phi $ -module $ \mathcal {Q}_{\phi }=(\mathcal {Q}_A, \phi _{\mathcal {Q}_A}) $ over $(A, I)$ is called the underlying G- $\phi $ -module of $\mathcal {Q}$ .

Example 2.5.5. We retain the notation of Example 2.4.7. Let $\mathcal {Q}:=\mathcal {Q}(M)$ be the $ {\mathrm {GL}} _N$ - $\mu $ -display over $(A, I)$ associated with a minuscule Breuil–Kisin module M over $(A, I)$ of type $\mu $ . The underlying $ {\mathrm {GL}} _N$ - $\phi $ -module $\mathcal {Q}_{\phi }$ is the one naturally associated with the pair

$$\begin{align*}(\phi^*M, \phi^*(F_M)). \end{align*}$$

For the Hodge filtration $P(\mathcal {Q})_{A/I}$ , we have

$$\begin{align*}P(\mathcal{Q})_{A/I} \simeq \underline{\mathrm{Isom}}_{{\mathrm{Fil}}}((A/I)^N, (\phi^*M)/I(\phi^*M)), \end{align*}$$

where the right-hand side is the scheme of isomorphisms $f \colon (A/I)^N \overset {\sim }{\to } (\phi ^*M)/I(\phi ^*M)$ such that $f((A/I)^s \oplus 0)=P^1$ .

2.6 Prismatic G- $\mu $ -displays over complete regular local rings

In this subsection, we recall the main result of [Reference ItoIto23].

Definition 2.6.1. Let R be a $\pi $ -adically complete ${\mathcal O}$ -algebra. We define the following groupoid:

A prismatic G- $\mu $ -display over R is an object of .

For a homomorphism $h \colon R \to R'$ of $\pi $ -adically complete ${\mathcal O}$ -algebras, we have a base change functor

Remark 2.6.2. Giving a prismatic G- $\mu $ -display $\mathfrak {Q}$ over R is equivalent to giving a section of the fibered category over . We thus have the associated G- $\mu $ -display $\mathfrak {Q}_{(A, I)}$ over $(A, I)$ for each and a natural isomorphism

$$\begin{align*}\gamma_f \colon f^*(\mathfrak{Q}_{(A, I)}) \overset{\sim}{\to} \mathfrak{Q}_{(A', I')} \end{align*}$$

for each morphism $f \colon (A, I) \to (A', I')$ in . We call $\mathfrak {Q}_{(A, I)}$ the value of $\mathfrak {Q}$ at . Recall that an object of is given by a bounded ${\mathcal O}_E$ -prism $(A, I)$ over ${\mathcal O}$ together with a homomorphism $g \colon R \to A/I$ over ${\mathcal O}$ . We also write $ \mathfrak {Q}_{g}:=\mathfrak {Q}_{(A, I)} $ in order to emphasize that it depends on the homomorphism g. For a homomorphism $h \colon R \to R'$ and an object , we have

$$\begin{align*}(h^*\mathfrak{Q})_g=\mathfrak{Q}_{g \circ h}. \end{align*}$$

Example 2.6.3. The functor

is an equivalence since is an initial object ([Reference ItoIto23, Example 2.5.8]).

Let $\mathcal {C}_{\mathcal O}$ be the category of complete regular local rings R with a local homomorphism ${\mathcal O} \to R$ which induces an isomorphism on the residue fields. The morphisms in $\mathcal {C}_{\mathcal O}$ are the local homomorphisms over ${\mathcal O}$ .

An important feature of the category $\mathcal {C}_{\mathcal O}$ is the following:

Remark 2.6.4. Let $(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))$ be an ${\mathcal O}_E$ -prism of Breuil–Kisin type over ${\mathcal O}$ . Then the quotient $\mathfrak {S}_{\mathcal O}/\mathcal {E}$ belongs to $\mathcal {C}_{\mathcal O}$ . Conversely, any $R \in \mathcal {C}_{\mathcal O}$ is of the form $ R \simeq \mathfrak {S}_{\mathcal O}/\mathcal {E} $ for some $(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))$ of Breuil–Kisin type. (See, for example, [Reference ChengChe18, Section 3.3].)

Let $R \in \mathcal {C}_{\mathcal O}$ . We choose an ${\mathcal O}_E$ -prism $(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))$ of Breuil–Kisin type with an isomorphism $ R \simeq \mathfrak {S}_{\mathcal O}/\mathcal {E} $ over ${\mathcal O}$ . The following result is a key ingredient in our deformation theory:

Theorem 2.6.5. The functor

is an equivalence if the cocharacter $\mu $ is 1-bounded.

Proof. See [Reference ItoIto23, Theorem 6.1.3].

3 The Grothendieck–Messing deformation theory for G- $\mu $ -displays

In this section, we establish an analogue of the Grothendieck–Messing deformation theory for G- $\mu $ -displays. The contents of this section can be summarized as follows. In Section 3.1, we collect some basic results on deformations of G- $\mu $ -displays. In Section 3.2, we introduce a class of maps $(A', I') \to (A, I)$ of orientable and bounded ${\mathcal O}_E$ -prisms over ${\mathcal O}$ , called special nilpotent thickenings, for which we formulate the analogue of the Grothendieck–Messing deformation theory. This class includes the maps $(\mathfrak {S}_{{\mathcal O}, m+1}, (\mathcal {E})) \to (\mathfrak {S}_{{\mathcal O}, m}, (\mathcal {E}))$ and $(W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m+1}, I_S) \to (W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m, I_S)$ . For a G- $\mu $ -display $\mathcal {Q}$ over $(A, I)$ and its deformation $\mathscr {Q}$ over $(A', I')$ , we construct a period map $ {\mathrm {Per}} _{\mathscr {Q}}$ from the set $ {\mathrm {Def}} (\mathcal {Q})_{(A', I')}$ of isomorphism classes of deformations of $\mathcal {Q}$ over $(A', I')$ to the set of isomorphism classes of lifts of the Hodge filtration $P(\mathcal {Q})_{A/I}$ in $\mathscr {Q}_{A'/I'}$ , and prove that $ {\mathrm {Per}} _{\mathscr {Q}}$ is bijective when $\mu $ is 1-bounded; see Section 3.3 for details. In Section 3.4, under a mild additional assumption on $(A', I') \to (A, I)$ , we show that the set $ {\mathrm {Def}} (\mathcal {Q})_{(A', I')}$ has a natural structure of a torsor under an $A/I$ -module when $\mu $ is 1-bounded. This structure plays an important role in Section 4.

3.1 Deformations

Let $f \colon (A', I') \to (A, I)$ be a map of orientable and bounded ${\mathcal O}_E$ -prisms over ${\mathcal O}$ such that $f \colon A' \to A$ is surjective.

We can define a deformation of a G- $\mu $ -display in the usual way:

Definition 3.1.1. Let $\mathcal {Q}$ be a G- $\mu $ -display over $(A, I)$ .

  1. (1) A deformation of $\mathcal {Q}$ over $(A', I')$ is a pair $ (\mathscr {Q}, h) $ consisting of a G- $\mu $ -display $\mathscr {Q}$ over $(A', I')$ and an isomorphism $ h \colon f^*\mathscr {Q} \overset {\sim }{\to } \mathcal {Q} $ of G- $\mu $ -displays over $(A, I)$ . If there is no ambiguity, we simply write $\mathscr {Q}$ instead of $(\mathscr {Q}, h)$ .

  2. (2) An isomorphism $ g \colon (\mathscr {Q}, h) \overset {\sim }{\to } (\mathscr {R}, i) $ of deformations of $\mathcal {Q}$ over $(A', I')$ is an isomorphism $g \colon \mathscr {Q} \overset {\sim }{\to } \mathscr {R}$ of G- $\mu $ -display over $(A', I')$ such that $i \circ (f^*g)=h$ .

The set of isomorphism classes of deformations of $\mathcal {Q}$ over $(A', I')$ is denoted by

$$\begin{align*}{\mathrm{Def}}(\mathcal{Q})_{(A', I')}. \end{align*}$$

The groupoid of deformations of $\mathcal {Q}$ over $(A', I')$ is denoted by $ \mathbf {Def}(\mathcal {Q})_{(A', I')}, $ using boldface letters.

We collect some preliminary results on deformations which will be used in the sequel. Let $J \subset A'$ (resp. $K \subset A'/I'$ ) be the kernel of $A' \to A$ (resp. $A'/I' \to A/I$ ).

We first note the following fact:

Lemma 3.1.2. Assume that $A'/I'$ is K-adically complete. Let $\mathcal {Q}$ be a banal G- $\mu $ -display over $(A, I)$ . Then any deformation $\mathscr {Q}$ of $\mathcal {Q}$ over $(A', I')$ is banal.

Proof. By Proposition 2.5.3, it suffices to prove that the Hodge filtration $P(\mathscr {Q})_{A'/I'}$ is trivial as a $P_\mu $ -torsor over $ \mathrm {Spec} A'/I'$ . The base change $P(\mathscr {Q})_{A'/I'} \times _{ \mathrm {Spec} A'/I'} \mathrm {Spec} A/I$ is trivial since $\mathcal {Q}$ is banal. Since $P(\mathscr {Q})_{A'/I'}$ is smooth over $ \mathrm {Spec} A'/I'$ and $A'/I'$ is K-adically complete, we see that $P(\mathscr {Q})_{A'/I'}$ is trivial, as desired.

The next lemma is an analogue of [Reference LauLau21, Lemma 7.1.4 (b)].

Lemma 3.1.3. If $A'$ is J-adically complete, then the homomorphism $f \colon G_\mu (A', I') \to G_\mu (A, I)$ is surjective.

Proof. Since $A'$ is J-adically complete and $P_\mu $ is smooth, the homomorphism $P_\mu (A') \to P_\mu (A)$ is surjective. It is easy to show that $U^{-}_{\mu }(A') \cap G_\mu (A', I') \to U^{-}_{\mu }(A) \cap G_\mu (A, I)$ is surjective; see [Reference ItoIto23, Remark 4.2.7]. The claim then follows from Proposition 2.3.3.

We have the following useful description of deformations for banal G- $\mu $ -displays.

Proposition 3.1.4. Assume that $A'$ is J-adically complete and $A'/I'$ is K-adically complete. Let $X \in G(A)_I$ be an element and $\mathcal {Q}_X$ the banal G- $\mu $ -display over $(A, I)$ associated with X. Let

$$\begin{align*}\mathbf{Def}(X)_{(A', I')} \end{align*}$$

be the groupoid whose objects are the elements $Y \in G(A')_{I'}$ such that $f(Y)=X$ , and whose morphisms are given by

$$\begin{align*}{\mathrm{Hom}}_{\mathbf{Def}(X)_{(A', I')}}(Y, Z)=\{\, g \in G_\mu(A', I') \, \vert \, Z \cdot g =Y \, \, \mathrm{and} \, \, f(g)=1 \, \}. \end{align*}$$

For each $Y \in \mathbf {Def}(X)_{(A', I')}$ , we have a natural isomorphism $f^*(\mathcal {Q}_Y) \overset {\sim }{\to } \mathcal {Q}_X$ given by $1 \in G_{\mu }(A, I)$ . The construction $Y \mapsto (\mathcal {Q}_Y, 1)$ gives an equivalence

$$\begin{align*}\mathbf{Def}(X)_{(A', I')} \overset{\sim}{\to} \mathbf{Def}(\mathcal{Q}_X)_{(A', I')}. \end{align*}$$

Proof. For the map $f \colon G(A')_{I'} \to G(A)_I$ , see Remark 2.4.3. It is clear that $Y \mapsto (\mathcal {Q}_Y, 1)$ gives a fully faithful functor. We shall prove that this functor is essentially surjective. Let $ (\mathscr {Q}, h) \in \mathbf {Def}(\mathcal {Q}_X)_{(A', I')}. $ It follows from Lemma 3.1.2 that $\mathscr {Q}$ is banal. We can identify $\mathscr {Q}$ with $\mathcal {Q}_Y$ for some $Y \in G(A')_{I'}$ , and then h is an element of $G_{\mu }(A, I)$ satisfying the equality $ X\cdot h=f(Y). $ By Lemma 3.1.3, there exists an element $h' \in G_\mu (A', I')$ with $f(h')=h$ . We define $Y':=Y \cdot (h')^{-1} \in G(A')_{I'}$ . Then $Y' \in \mathbf {Def}(X)_{(A', I')}$ and $h'$ induces $ (\mathcal {Q}_Y, h) \overset {\sim }{\to } (\mathcal {Q}_{Y'}, 1). $

For later use, we also record the following result.

Proposition 3.1.5. Let $(A, I)$ be an orientable and bounded ${\mathcal O}_E$ -prism over ${\mathcal O}$ . Let $J \subset A$ be an ideal satisfying the following conditions:

  1. (a) A is J-adically complete.

  2. (b) J is stable under $\delta _E$ . In particular, $A/J^m$ admits a $\delta _E$ -structure that is compatible with the one on A for every $m \geq 1$ .

  3. (c) For every $m \geq 1$ , the pair $(A/J^m, I)$ is an orientable and bounded ${\mathcal O}_E$ -prism over ${\mathcal O}$ . Here, we abuse notation and denote the image of I in $A/J^m$ by the same symbol.

Then the following assertions hold:

  1. (1) We have the following equivalence of groupoids:

  2. (2) We further assume that for any G- $\mu $ -display $\mathcal {Q}$ over $(A, I)$ (resp. $(A/J, I)$ ), there is a finite étale covering $A \to B$ such that the base change of $\mathcal {Q}$ to $(B, IB)$ (resp. $(B/JB, I(B/JB))$ ) is banal. Then we have

Proof. (1) We claim that

(3.1) $$ \begin{align} G(A)_I \overset{\sim}{\to} {\varprojlim}_{m} G(A/J^m)_I \quad \text{and} \quad G_\mu(A, I)\overset{\sim}{\to}{\varprojlim}_{m} G_\mu(A/J^m, I). \end{align} $$

Indeed, the first equality is clear since G is affine. The second equality follows from Proposition 2.3.3 (2).

In order to prove (1), it suffices to show that the natural functor

$$\begin{align*}[G(A)_I/G_\mu(A, I)] \to \mathcal{C}:={2-\varprojlim}_{m} [G(A/J^m)_I/G_\mu(A/J^m, I)] \end{align*}$$

is an equivalence (cf. Remark 2.4.2). The fully faithfulness follows from (3.1). We shall prove that the functor is essentially surjective. An object of $\mathcal {C}$ can be identified with a family of objects $\{ X_m \}_{m \geq 1}$ , where $X_m \in G(A/J^m)_I$ , together with a family of isomorphisms $\{ g_m \}_{m \geq 1}$ , where $g_m \in G_\mu (A/J^m, I)$ is such that the equality $X_m\cdot g_m=X_{m+1}$ holds in $G(A/J^m)_I$ . Applying Proposition 3.1.4 to the natural map $p_m \colon (A/J^{m+1}, I) \to (A/J^{m}, I)$ repeatedly, we can find an element $(X^{\prime }_m)_{m \geq 1} \in \varprojlim _{m} G(A/J^m)_I$ such that the corresponding object of $\mathcal {C}$ is isomorphic to the object $\{ X_m \}_{m \geq 1}$ . Let $X \in G(A)_I$ be the element corresponding to $(X^{\prime }_m)_{m \geq 1}$ . Then the image of X under the above functor is isomorphic to the object $\{ X_m \}_{m \geq 1}$ .

(2) An object of can be identified with a family $\{ \mathcal {Q}_m \}_{m \geq 1}$ , where $\mathcal {Q}_m$ is a G- $\mu $ -display over $(A/J^m, I)$ , together with isomorphisms $p^*_m(\mathcal {Q}_{m+1}) \overset {\sim }{\to } \mathcal {Q}_{m}$ ( $m \geq 1$ ). Our assumption and Lemma 3.1.2 imply that there is a finite étale covering $A \to B$ such that for any $m \geq 1$ , the base change of $\mathcal {Q}_m$ to $(B/J^mB, I(B/J^mB))$ is banal. We note that for a finite étale covering $A \to B$ , the ideal $JB$ satisfies the above conditions (a), (b), (c). Using these observations and finite étale descent for G- $\mu $ -displays, we see that (2) follows from (1).

Example 3.1.6. Let $(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))$ be an ${\mathcal O}_E$ -prism of Breuil–Kisin type as in Example 2.1.1, where $\mathfrak {S}_{\mathcal O}={\mathcal O}[[t_1, \dotsc , t_n]]$ . Let $\mathcal {Q}$ be a G- $\mu $ -display over $(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))$ . By Proposition 2.5.3, there is a finite extension $\widetilde {k}$ of k such that the base change of $\mathcal {Q}$ to $(\mathfrak {S}_{\widetilde {{\mathcal O}}}, (\mathcal {E}))$ is banal, where $\widetilde {{\mathcal O}}:= W(\widetilde {k}) \otimes _{W({\mathbb F}_q)} {\mathcal O}_E$ and $\mathfrak {S}_{\widetilde {{\mathcal O}}}:=\mathfrak {S}_{\mathcal O} \otimes _{\mathcal O} \widetilde {{\mathcal O}} = \widetilde {{\mathcal O}}[[t_1, \dotsc , t_n]]$ . Therefore, by Proposition 3.1.5, we have

where $\mathfrak {S}_{{\mathcal O}, m}={\mathcal O}[[t_1, \dotsc , t_n]]/(t_1, \dotsc , t_n)^m$ .

3.2 Special nilpotent thickenings

Let $f \colon (A', I') \to (A, I)$ be a map of orientable and bounded ${\mathcal O}_E$ -prisms over ${\mathcal O}$ such that $f \colon A' \to A$ is surjective. Let $J \subset A'$ be the kernel of f. We assume that $J^n=0$ for some $n \geq 1$ . In this case, we say that f is a nilpotent thickening. The functor

$$\begin{align*}(A', I')_{\mathrm{\acute{e}t}} \to (A, I)_{\mathrm{\acute{e}t}}, \quad B' \mapsto B, \end{align*}$$

where B is the $(\pi , I)$ -adic completion of $B' \otimes _{A'} A$ , is an equivalence (see [Reference ItoIto23, Lemma 2.5.9] for example).

Lemma 3.2.1. Let the notation be as above. Then the induced homomorphism $f_B \colon B' \to B$ is surjective, and we have $J^n_B=0$ for the kernel $J_B$ of $f_B$ . If furthermore $\phi _{A'}(J) = 0$ , then $\phi _{B'}(J_B) = 0$ .

Proof. We have the following exact sequence for every $m \geq 1$ :

$$\begin{align*}0 \to (J+(\pi, I')^m)/(\pi, I')^m \to A'/(\pi, I')^m \to A/(\pi, I)^m \to 0. \end{align*}$$

Since $A'/(\pi , I')^m \to B'/(\pi , I')^m$ is flat, the induced sequence

$$\begin{align*}0 \to (JB'+(\pi, I')^m)/(\pi, I')^m \to B'/(\pi, I')^m \to B/(\pi, I)^m \to 0 \end{align*}$$

is also exact. By taking the limit over m, we see that $B' \to B$ is surjective, and

$$\begin{align*}J_B \simeq {\varprojlim}_m (JB'+(\pi, I')^m)/(\pi, I')^m. \end{align*}$$

Using this isomorphism, we see that $J^n_B=0$ , and that $\phi _{B'}(J_B) = 0$ if $\phi _{A'}(J) =0$ .

Definition 3.2.2 (Special nilpotent thickening).

Let $ f \colon (A', I') \to (A, I) $ be a nilpotent thickening of orientable and bounded ${\mathcal O}_E$ -prisms over ${\mathcal O}$ . Let $J \subset A'$ be the kernel of $f \colon A' \to A$ . We say that f is a special nilpotent thickening if it satisfies the following two conditions:

  1. (1) We have $\phi _{A'}(J) = 0$ .

  2. (2) Let $d \in I'$ be a generator. For any $(\pi , I')$ -completely étale $A'$ -algebra $B'$ , the element $\phi _{B'}(d)$ is a nonzerodivisor in $B'$ .

Remark 3.2.3. It follows from Lemma 3.2.1 that if $f \colon (A', I') \to (A, I)$ is a nilpotent thickening (resp. a special nilpotent thickening), then for any $B' \in (A', I')_{\mathrm {\acute {e}t}}$ with corresponding $B \in (A, I)_{\mathrm {\acute {e}t}}$ , the induced map $(B', I'B') \to (B, IB)$ is also a nilpotent thickening (resp. a special nilpotent thickening).

Example 3.2.4. If $(A, I)$ is either $(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))$ or $(\mathfrak {S}_{{\mathcal O}, m}, (\mathcal {E}))$ as in Example 2.1.1, then $(A, I)$ satisfies the condition (2) in Definition 3.2.2. Indeed, it is clear that $\phi (\mathcal {E})$ is a nonzerodivisor in A. Since A is noetherian, any $(\pi , I)$ -completely étale A-algebra B is flat over A, which in turn implies that $\phi (\mathcal {E})$ is a nonzerodivisor in B.

Example 3.2.5. Let $(S, a^\flat )$ be a perfectoid pair over ${\mathcal O}$ (see Definition 2.1.4). Then the ${\mathcal O}_E$ -prism $ (W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m, I_S) $ as in Proposition 2.1.3 satisfies the condition (2) in Definition 3.2.2. Indeed, by Lemma 2.2.3, it suffices to show that $\phi (\xi )$ is a nonzerodivisor in $W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m$ . Since $ \phi \colon W_{{\mathcal O}_E}(S^\flat )/[(a^\flat )^{1/q}]^m \to W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m $ is bijective and $\xi $ is a nonzerodivisor in $W_{{\mathcal O}_E}(S^\flat )/[(a^\flat )^{1/q}]^m$ (by Proposition 2.1.3), the assertion follows.

Example 3.2.6. Let the notation be as in Example 3.2.4 and Example 3.2.5. In this paper, we say that a nilpotent thickening $f \colon (A', I') \to (A, I)$ is of Breuil–Kisin type (resp. of perfectoid type) if it is of the form

$$ \begin{align*} (\mathfrak{S}_{{\mathcal O}, m+1}, (\mathcal{E})) &\to (\mathfrak{S}_{{\mathcal O}, m}, (\mathcal{E})) \\ (\text{resp.\ } (W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^{m+1}, I_S) &\to (W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^{m}, I_S)). \end{align*} $$

In either case, f is a special nilpotent thickening.

Deformations of G- $\phi $ -modules are defined in the usual way. If $\mathcal {Q}$ is a G- $\mu $ -display over $(A, I)$ and $\mathscr {Q}$ is a deformation of $\mathcal {Q}$ over $(A', I')$ , then $\mathscr {Q}_\phi $ is naturally a deformation of $\mathcal {Q}_\phi $ over $(A', I')$ . Here, $\mathscr {Q}_\phi $ and $\mathcal {Q}_\phi $ are the underlying G- $\phi $ -modules; see Definition 2.5.4. The following proposition is the key ingredient in the construction of period maps introduced in Section 3.3 below.

Proposition 3.2.7. Assume that $f \colon (A', I') \to (A, I)$ is a special nilpotent thickening. Let $\mathcal {Q}$ be a G- $\mu $ -display over $(A, I)$ . Let $(\mathscr {Q}, h)$ and $(\mathscr {R}, i)$ be two deformations of $\mathcal {Q}$ over $(A', I')$ . Then the following assertions hold:

  1. (1) There exists a unique isomorphism $ \psi \colon \mathscr {Q}_{\phi } \overset {\sim }{\to } \mathscr {R}_{\phi } $ of deformations of $\mathcal {Q}_{\phi }$ over $(A', I')$ .

  2. (2) There is at most one isomorphism between two deformations $(\mathscr {Q}, h)$ and $(\mathscr {R}, i)$ .

Proof. (1) The problem is $(\pi , I)$ -completely étale local on A by $(\pi , I)$ -completely étale descent for G- $\phi $ -modules (which follows from the same argument as in [Reference ItoIto23, Remark 5.1.3]) and the uniqueness assertion. Thus, we may assume that $\mathscr {Q}$ , $\mathscr {R}$ and $\mathcal {Q}$ are banal. We can identify $\mathcal {Q}$ with $\mathcal {Q}_X$ for some $X \in G(A)_I$ . By Proposition 3.1.4, we may assume that $(\mathscr {Q}, h)$ and $(\mathscr {R}, i)$ are of the form $(\mathcal {Q}_Y, 1)$ and $(\mathcal {Q}_Z, 1)$ , respectively, for some $Y, Z \in \mathbf {Def}(X)_{(A', I')}$ . It suffices to prove that there exists a unique element $\psi \in G(A')$ such that $f(\psi )=1$ in $G(A)$ and

(3.2) $$ \begin{align} \psi^{-1}Z_\phi \phi(\psi) =Y_\phi \quad \text{in} \quad G(A'[1/\phi(I')]). \end{align} $$

Here, $Y_\phi , Z_\phi \in G(A'[1/\phi (I')])$ are the elements defined in (2.2).

We fix a generator $d \in I'$ . We shall show that the element

$$\begin{align*}\psi:=Z_d(Y_d)^{-1} \in G(A') \end{align*}$$

satisfies $f(\psi )=1$ and (3.2). Indeed, since $X=f(Y)=f(Z)$ , we have $f(\psi )=1$ . This implies that $f(\mu (d)\psi \mu (d)^{-1})=1$ . By our assumption that $\phi _{A'}(J)=0$ for the kernel $J \subset A'$ of $f \colon A' \to A$ , it then follows that $\phi (\mu (d)\psi \mu (d)^{-1})=1$ , whence (3.2) holds.

It remains to show the uniqueness of $\psi $ . Let $\psi \in G(A')$ be an element satisfying $f(\psi )=1$ and (3.2). As explained above, the equality $f(\psi )=1$ implies $\phi (\mu (d)\psi \mu (d)^{-1})=1$ . Then it follows from (3.2) that $\psi =Z_d(Y_d)^{-1}$ in $G(A'[1/\phi (I')])$ . Since the natural homomorphism $G(A') \to G(A'[1/\phi (I')])$ is injective by the condition (2) in Definition 3.2.2, we conclude that $\psi =Z_d(Y_d)^{-1}$ in $G(A')$ .

(2) We note that the functor $\mathcal {Q} \mapsto \mathcal {Q}_{\phi }$ from to the groupoid of G- $\phi $ -modules over $(A, I)$ is faithful (since the functor from the groupoid of $G_{\mu , A, I}$ -torsors to that of -torsors obtained by pushout along the inclusion is faithful). Then (2) follows from the uniqueness assertion in (1).

Remark 3.2.8. Let the notation be as in Proposition 3.1.4. Let $Y, Z \in \mathbf {Def}(X)_{(A', I')}$ . For the deformations $(\mathcal {Q}_Y, 1)$ and $(\mathcal {Q}_Z, 1)$ of the G- $\mu $ -display $\mathcal {Q}_X$ , the proof of Proposition 3.2.7 shows that for any generator $d \in I'$ ,

$$\begin{align*}(\mathcal{Q}_Y, 1) \simeq (\mathcal{Q}_Z, 1) \quad \text{if and only if} \quad Z_d(Y_d)^{-1} \in G_{\mu}(A', I'). \end{align*}$$

If this is the case, the unique isomorphism $(\mathcal {Q}_Y, 1) \overset {\sim }{\to } (\mathcal {Q}_Z, 1)$ is given by $Z_d(Y_d)^{-1}$ .

Corollary 3.2.9. Assume that $f \colon (A', I') \to (A, I)$ is a special nilpotent thickening. Let $\mathcal {Q}$ be a G- $\mu $ -display over $(A, I)$ . For each $B' \in (A', I')_{\mathrm {\acute {e}t}}$ with corresponding $B \in (A, I)_{\mathrm {\acute {e}t}}$ , we denote the base change of $\mathcal {Q}$ to $(B, IB)$ by $\mathcal {Q}_{(B, IB)}$ . Then the functor

$$\begin{align*}\underline{{\mathrm{Def}}}(\mathcal{Q}) \colon (A', I')_{\mathrm{\acute{e}t}} \to \mathrm{Set}, \quad B' \mapsto \underline{{\mathrm{Def}}}(\mathcal{Q})(B')={\mathrm{Def}}(\mathcal{Q}_{(B, IB)})_{(B', I'B')} \end{align*}$$

forms a sheaf with respect to the $(\pi , I')$ -completely étale topology.

Proof. This is an immediate consequence of Proposition 3.2.7 (2) and $(\pi , I)$ -completely étale descent for G- $\mu $ -displays.

Corollary 3.2.10. Let the notation be as in Example 3.2.4 and Example 3.2.5.

  1. (1) Let $\mathcal {Q}$ be a G- $\mu $ -display over $({\mathcal O}, (\pi ))=(\mathfrak {S}_{{\mathcal O}, 1}, (\mathcal {E}))$ . Let $m \geq 1$ be an integer. Any deformation of $\mathcal {Q}$ over $(\mathfrak {S}_{{\mathcal O}, m}, (\mathcal {E}))$ has no nontrivial automorphisms. Moreover, any deformation of $\mathcal {Q}$ over $(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))$ has no nontrivial automorphisms.

  2. (2) Let $m \geq 1$ be an integer. For a G- $\mu $ -display $\mathcal {Q}$ over $(W_{{\mathcal O}_E}(S^\flat )/[a^\flat ], I_S)$ , any deformation of $\mathcal {Q}$ over $(W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m, I_S)$ has no nontrivial automorphisms.

Proof. (1) By Example 3.1.6, it suffices to prove the first assertion. This follows by applying Proposition 3.2.7 (2) to special nilpotent thickenings $(\mathfrak {S}_{{\mathcal O}, l+1}, (\mathcal {E})) \to (\mathfrak {S}_{{\mathcal O}, l}, (\mathcal {E}))$ $(1 \leq l \leq m-1)$ repeatedly.

(2) This also follows from Proposition 3.2.7 (2).

3.3 Period maps

Let $f \colon (A', I') \to (A, I)$ be a special nilpotent thickening of orientable and bounded ${\mathcal O}_E$ -prisms over ${\mathcal O}$ . Let $\mathcal {Q}$ be a G- $\mu $ -display over $(A, I)$ . Recall the Hodge filtration $P(\mathcal {Q})_{A/I} \to \mathcal {Q}_{A/I}$ from Section 2.5.

Definition 3.3.1. Let $\mathscr {Q}$ be a G-torsor over $ \mathrm {Spec} A'/I'$ with an isomorphism

$$\begin{align*}f^*\mathscr{Q}:=\mathscr{Q} \times_{\mathrm{Spec} A'/I'} \mathrm{Spec} A/I \overset{\sim}{\to} \mathcal{Q}_{A/I} \end{align*}$$

of G-torsors over $ \mathrm {Spec} A/I$ . A lift of the Hodge filtration $P(\mathcal {Q})_{A/I} \to \mathcal {Q}_{A/I}$ in $\mathscr {Q}$ is a pair $(\mathscr {P}, \iota )$ of a $P_\mu $ -torsor $\mathscr {P}$ over $ \mathrm {Spec} A'/I'$ and a $P_\mu $ -equivariant morphism $\iota \colon \mathscr {P} \to \mathscr {Q}$ such that the isomorphism $f^*\mathscr {Q} \overset {\sim }{\to } \mathcal {Q}_{A/I}$ restricts to an isomorphism $f^*\mathscr {P} \overset {\sim }{\to } P(\mathcal {Q})_{A/I}$ .

There is an obvious notion of isomorphism of lifts of the Hodge filtration. We note that there is at most one isomorphism between two lifts. Let

$$\begin{align*}{\mathrm{Lift}}(P(\mathcal{Q})_{A/I}, \mathscr{Q}) \end{align*}$$

be the set of isomorphism classes of lifts of the Hodge filtration $P(\mathcal {Q})_{A/I} \to \mathcal {Q}_{A/I}$ in $\mathscr {Q}$ .

Example 3.3.2. For a deformation $\mathscr {Q}$ of $\mathcal {Q}$ over $(A', I')$ , the pair

$$\begin{align*}(P(\mathscr{Q})_{A'/I'}, P(\mathscr{Q})_{A'/I'} \to \mathscr{Q}_{A'/I'}) \end{align*}$$

is a lift of the Hodge filtration $P(\mathcal {Q})_{A/I}$ in $\mathscr {Q}_{A'/I'}$ .

Definition 3.3.3 (Period map).

Let $\mathscr {Q}$ be a deformation of $\mathcal {Q}$ over $(A', I')$ . We define a map of sets

$$\begin{align*}{\mathrm{Per}}_{\mathscr{Q}} \colon {\mathrm{Def}}(\mathcal{Q})_{(A', I')} \to {\mathrm{Lift}}(P(\mathcal{Q})_{A/I}, \mathscr{Q}_{A'/I'}) \end{align*}$$

as follows. Let $\mathscr {R} \in {\mathrm {Def}} (\mathcal {Q})_{(A', I')}$ be a deformation of $\mathcal {Q}$ over $(A', I')$ . By Proposition 3.2.7, there exists a unique isomorphism $ \psi \colon \mathscr {R}_{\phi } \overset {\sim }{\to } \mathscr {Q}_{\phi } $ of deformations of the underlying G- $\phi $ -module $\mathcal {Q}_{\phi }$ . In particular, $\psi $ induces an isomorphism $\mathscr {R}_{A'/I'} \overset {\sim }{\to } \mathscr {Q}_{A'/I'}$ of G-torsors over $ \mathrm {Spec} A'/I'$ . We define the map $ {\mathrm {Per}} _{\mathscr {Q}}$ by sending $\mathscr {R}$ to the lift

$$\begin{align*}(P(\mathscr{R})_{A'/I'}, \, P(\mathscr{R})_{A'/I'} \to \mathscr{R}_{A'/I'} \overset{\sim}{\to} \mathscr{Q}_{A'/I'}) \in {\mathrm{Lift}}(P(\mathcal{Q})_{A/I}, \mathscr{Q}_{A'/I'}). \end{align*}$$

We call $ {\mathrm {Per}} _{\mathscr {Q}}$ the period map associated with the deformation $\mathscr {Q}$ .

The main result of this section is the following theorem, which can be seen as an analogue of the Grothendieck–Messing deformation theory.

Theorem 3.3.4. Let $f \colon (A', I') \to (A, I)$ be a special nilpotent thickening of orientable and bounded ${\mathcal O}_E$ -prisms over ${\mathcal O}$ . Let $\mathcal {Q}$ be a G- $\mu $ -display over $(A, I)$ , and we fix a deformation $\mathscr {Q}$ of $\mathcal {Q}$ over $(A', I')$ . If the cocharacter $\mu $ is 1-bounded, then the period map

$$\begin{align*}{\mathrm{Per}}_{\mathscr{Q}} \colon {\mathrm{Def}}(\mathcal{Q})_{(A', I')} \to {\mathrm{Lift}}(P(\mathcal{Q})_{A/I}, \mathscr{Q}_{A'/I'}) \end{align*}$$

is bijective.

Proof. By Corollary 3.2.9, we may work $(\pi , I')$ -completely étale locally on $A'$ . We may thus assume that $\mathcal {Q}=\mathcal {Q}_X$ for some $X \in G(A)_I$ . We shall prove that $ {\mathrm {Per}} _{\mathscr {Q}}$ is injective. Let $\mathscr {R}$ and $\mathscr {S}$ be two deformations of $\mathcal {Q}$ over $(A', I')$ such that $P(\mathscr {R})_{A'/I'} = P(\mathscr {S})_{A'/I'}$ in $ {\mathrm {Lift}} (P(\mathcal {Q})_{A/I}, \mathscr {Q}_{A'/I'})$ . By Proposition 3.1.4, we may further assume that

$$\begin{align*}\mathscr{Q}=(\mathcal{Q}_{X'}, 1), \quad \mathscr{R}=(\mathcal{Q}_Y, 1), \quad \text{and} \quad \mathscr{S}=(\mathcal{Q}_Z, 1) \end{align*}$$

for some $X', Y, Z \in \mathbf {Def}(X)_{(A', I')}$ . Then $\mathscr {Q}_{A'/I'}$ is naturally identified with $G_{A'/I'}$ , and the isomorphism $f^*(\mathscr {Q}_{A'/I'}) \overset {\sim }{\to } \mathcal {Q}_{A/I}$ agrees with the canonical isomorphism $f^*G_{A'/I'} \overset {\sim }{\to } G_{A/I}$ . We fix a generator $d \in I'$ . The period map $ {\mathrm {Per}} _{\mathscr {Q}}$ sends $(\mathcal {Q}_Y, 1)$ and $(\mathcal {Q}_Z, 1)$ to

$$\begin{align*}((P_{\mu})_{A'/I'}, \iota_{X^{\prime}_d(Y_d)^{-1}}) \quad \text{and} \quad ((P_{\mu})_{A'/I'}, \iota_{X^{\prime}_d(Z_d)^{-1}}), \end{align*}$$

respectively, where $\iota _{X^{\prime }_d(Y_d)^{-1}} \colon (P_{\mu })_{A'/I'} \hookrightarrow G_{A'/I'}$ is defined by $g \mapsto X^{\prime }_d(Y_d)^{-1}g$ , and similarly for $\iota _{X^{\prime }_d(Z_d)^{-1}}$ . (See (2.1) for the notation $X_d$ .) Since the two lifts are isomorphic, the image of $Z_d(Y_d)^{-1}$ in $G(A'/I')$ belongs to $P_\mu (A'/I')$ . This implies that $Z_d(Y_d)^{-1} \in G_\mu (A', I')$ by Proposition 2.3.3 since $\mu $ is 1-bounded. Then $Z_d(Y_d)^{-1}$ gives $(\mathcal {Q}_Y, 1) \overset {\sim }{\to } (\mathcal {Q}_Z, 1)$ by Remark 3.2.8.

We next prove that $ {\mathrm {Per}} _{\mathscr {Q}}$ is surjective. We may assume that $\mathscr {Q}=(\mathcal {Q}_{X'}, 1)$ as above. Let $(\mathscr {P}, \iota ) \in {\mathrm {Lift}} (P(\mathcal {Q})_{A/I}, \mathscr {Q}_{A'/I'})$ . By the same argument as in the proof of Lemma 3.1.2, we see that $\mathscr {P}$ is trivial as a $P_\mu $ -torsor. Then, with the notation as above, this lift can be identified with $ ((P_{\mu })_{A'/I'}, \iota _{x}) $ for some $x \in G(A'/I')$ whose image in $G(A/I)$ is $1$ . By Lemma 3.3.5 below, after replacing x by $x t$ for some $t \in P_\mu (A'/I')$ , we can find an element $\widetilde {x} \in G(A')$ such that $f(\widetilde {x})=1$ in $G(A)$ and the image of $\widetilde {x}$ in $G(A'/I')$ coincides with x. We then define $Y \in \mathbf {Def}(X)_{(A', I')}$ to be the unique object such that $Y_d=\widetilde {x}^{-1} X^{\prime }_d$ . By construction, the period map $ {\mathrm {Per}} _{\mathscr {Q}}$ sends $(\mathcal {Q}_Y, 1)$ to $((P_{\mu })_{A'/I'}, \iota _{x})$ . This concludes the proof of the surjectivity of $ {\mathrm {Per}} _{\mathscr {Q}}$ , and hence that of Theorem 3.3.4.

The following lemma is used in the proof of Theorem 3.3.4.

Lemma 3.3.5. Let the notation be as in Theorem 3.3.4. Let $x \in G(A'/I')$ be such that $f(x)=1$ in $G(A/I)$ . Then there exist $t \in P_\mu (A'/I')$ and $\widetilde {x} \in G(A')$ such that $f(\widetilde {x})=1$ in $G(A)$ and the image of $\widetilde {x}$ in $G(A'/I')$ is equal to $x t$ .

Proof. Since G is smooth and $A'$ is $I'$ -adically complete, the map $G(A') \to G(A'/I')$ is surjective. Let $\widetilde {x}_1 \in G(A')$ be an element which is mapped to $x \in G(A'/I')$ . By Proposition 2.3.3, we have $f(\widetilde {x}_1) \in G_\mu (A, I)$ . By Lemma 3.1.3, there is an element $\widetilde {x}_2 \in G_\mu (A', I')$ such that $f(\widetilde {x}_2)=f(\widetilde {x}_1)$ in $G_\mu (A, I)$ . The image $t \in G(A'/I')$ of $\widetilde {x}^{-1}_2$ belongs to $P_\mu (A'/I')$ . Then t and $\widetilde {x}:=\widetilde {x}_1\widetilde {x}^{-1}_2$ have the desired properties.

Remark 3.3.6. A banal G- $\mu $ -display over $(A, I)$ admits a deformation over $(A', I')$ since $G(A') \to G(A)$ is surjective. However, for a G- $\mu $ -display $\mathcal {Q}$ over $(A, I)$ which is not necessarily banal, the existence of a deformation of $\mathcal {Q}$ over $(A', I')$ is not clear. In Proposition 3.4.6 below, under a mild additional assumption on $f \colon (A', I') \to (A, I)$ , we will prove that a deformation of $\mathcal {Q}$ over $(A', I')$ exists if $\mu $ is 1-bounded.

3.4 Normalized period maps

Let $f \colon (A', I') \to (A, I)$ be a special nilpotent thickening of orientable and bounded ${\mathcal O}_E$ -prisms over ${\mathcal O}$ . In this subsection, we assume that f satisfies the following assumption:

Assumption 3.4.1. The kernel K of $A'/I' \to A/I$ satisfies $K^2=0$ . Moreover, there exists an integer $n \geq 1$ such that $\pi ^n=0$ in $A'/I'$ .

Remark 3.4.2. A special nilpotent thickening of Breuil–Kisin type or of perfectoid type (Example 3.2.6) satisfies Assumption 3.4.1.

Let $\mathcal {Q}$ be a banal G- $\mu $ -display over $(A, I)$ . The quotient (fppf) sheaf

$$\begin{align*}X_\mu(\mathcal{Q}):=\mathcal{Q}_{A/I}/(P_\mu)_{A/I} \end{align*}$$

is representable by a scheme over $ \mathrm {Spec} A/I$ , which is also denoted by $X_\mu (\mathcal {Q})$ . (To see this, it suffices to show that the quotient sheaf $G_{{\mathcal O}/\pi ^n}/(P_\mu )_{{\mathcal O}/\pi ^n}$ is representable, which follows from [Reference Demazure and GrothendieckDG70, Expose $V$ , Théorème 3.2]. We note that $\mathcal {Q}_{A/I} \simeq G_{A/I}$ since $\mathcal {Q}$ is banal.) The closed immersion $P(\mathcal {Q})_{A/I} \hookrightarrow \mathcal {Q}_{A/I}$ induces a section

$$\begin{align*}s_{\mathcal{Q}} \colon \mathrm{Spec} A/I \hookrightarrow X_\mu(\mathcal{Q}). \end{align*}$$

The $A/I$ -module corresponding to the conormal sheaf of $s_{\mathcal {Q}}$ on $ \mathrm {Spec} A/I$ is denoted by $C(\mathcal {Q})$ . We call $C(\mathcal {Q})$ the conormal module.

Definition 3.4.3. Let ${\mathcal O}_{U^{-}_{\mu }, 1}$ be the local ring of $U^{-}_{\mu }$ at $1 \in U^{-}_{\mu }(k)$ and let

$$\begin{align*}R_{G, \mu}:= \widehat{{\mathcal O}}_{U^{-}_{\mu}, 1} \end{align*}$$

be the completion of ${\mathcal O}_{U^{-}_{\mu }, 1}$ with respect to the maximal ideal. We denote by $ J_{G, \mu } \subset R_{G, \mu } $ the kernel of the homomorphism $\epsilon \colon R_{G, \mu } \to {\mathcal O}$ induced by the unit morphism $ \mathrm {Spec} {\mathcal O} \to U^{-}_{\mu }$ . We set

$$\begin{align*}C(R_{G, \mu}):=J_{G, \mu}/J^2_{G, \mu}, \end{align*}$$

which is a free ${\mathcal O}$ -module of finite rank.

Example 3.4.4. An isomorphism $\mathcal {Q} \overset {\sim }{\to } \mathcal {Q}_{X}$ of G- $\mu $ -displays over $(A, I)$ for an element $X \in G(A)_I$ induces $ X_\mu (\mathcal {Q}) \overset {\sim }{\to } G_{A/I}/(P_\mu )_{A/I} $ and $ C(R_{G, \mu }) \otimes _{\mathcal O} A/I \overset {\sim }{\to } C(\mathcal {Q}). $

In the following, for modules M and N over a ring R, the set of R-linear homomorphisms $M \to N$ is denoted by $ {\mathrm {Hom}} _R(M, N)$ .

Lemma 3.4.5. Let $\mathcal {Q}$ be a banal G- $\mu $ -display over $(A, I)$ , and let $\mathscr {Q}$ be a deformation of $\mathcal {Q}$ over $(A', I')$ . Then there exists a natural bijection

$$\begin{align*}{\mathrm{Lift}}(P(\mathcal{Q})_{A/I}, \mathscr{Q}_{A'/I'}) \overset{\sim}{\to} {\mathrm{Hom}}_{A/I}(C(\mathcal{Q}), K). \end{align*}$$

(Since $K^2=0$ , the ideal $K \subset A'/I'$ can be naturally regarded as an $A/I$ -module.)

Proof. As above, the Hodge filtration $P(\mathscr {Q})_{A'/I'} \to \mathscr {Q}_{A'/I'}$ induces a section

$$\begin{align*}s_{\mathscr{Q}} \colon \mathrm{Spec} A'/I' \hookrightarrow X_{\mu}(\mathscr{Q}):=\mathscr{Q}_{A'/I'}/(P_\mu)_{A'/I'}. \end{align*}$$

Let $C(\mathscr {Q})$ be the conormal module of $s_{\mathscr {Q}}$ . We have $C(\mathscr {Q})\otimes _{A'/I'} A/I \simeq C(\mathcal {Q})$ . Let $s \colon \mathrm {Spec} A/I \hookrightarrow X_\mu (\mathscr {Q})$ be the composition of $s_{\mathscr {Q}}$ with $ \mathrm {Spec} A/I \hookrightarrow \mathrm {Spec} A'/I'$ , and let $C_s$ be the conormal module of s. We have a natural homomorphism $ C(\mathscr {Q})\otimes _{A'/I'} A/I \to C_s $ of $A/I$ -modules.

Let $\iota \colon \mathscr {P} \to \mathscr {Q}_{A'/I'}$ be a lift of $P(\mathcal {Q})_{A/I}$ . In the same way as above, this induces a section $ s_{\mathscr {P}} \colon \mathrm {Spec} A'/I' \to X_\mu (\mathscr {Q}). $ Since the composition of $s_{\mathscr {P}}$ with $ \mathrm {Spec} A/I \hookrightarrow \mathrm {Spec} A'/I'$ is equal to s, the morphism $s_{\mathscr {P}}$ induces a homomorphism $t_{\mathscr {P}} \colon C_s \to K$ of $A/I$ -modules (defined by $x \mapsto s^*_{\mathscr {P}}(x)$ ). Associating to $\iota \colon \mathscr {P} \to \mathscr {Q}_{A'/I'}$ the composition

$$\begin{align*}C(\mathcal{Q}) \simeq C(\mathscr{Q})\otimes_{A'/I'} A/I \to C_s \overset{t_{\mathscr{P}}}{\to} K, \end{align*}$$

we obtain a map

$$\begin{align*}{\mathrm{Lift}}(P(\mathcal{Q})_{A/I}, \mathscr{Q}_{A'/I'}) \to {\mathrm{Hom}}_{A/I}(C(\mathcal{Q}), K). \end{align*}$$

We shall prove that this map is bijective. For this, we may assume that $\mathcal {Q}=\mathcal {Q}_X$ for some $X \in G(A)_I$ and $\mathscr {Q}=(\mathcal {Q}_{X'}, 1)$ for some $X' \in \mathbf {Def}(X)_{(A', I')}$ as in the proof of Theorem 3.3.4. Then we have $P(\mathcal {Q})_{A/I}=(P_\mu )_{A/I}$ and $\mathscr {Q}_{A'/I'}=G_{A'/I'}$ , and by the argument in the proof of Theorem 3.3.4, we can identify $ {\mathrm {Lift}} (P(\mathcal {Q})_{A/I}, \mathscr {Q}_{A'/I'})$ with the set of $P_\mu (A'/I')$ -orbits $ x P_\mu (A'/I') \subset G(A'/I') $ of elements x of the kernel of $G(A'/I') \to G(A/I)$ . We note that for such elements x, there are elements $u \in U^-_{\mu }(A'/I')$ and $t \in P_\mu (A'/I')$ such that $x=ut$ since the multiplication map

$$\begin{align*}U^{-}_{\mu} \times_{\mathrm{Spec} {\mathcal O}} P_\mu \to G_{\mathcal O} \end{align*}$$

is an open immersion; see also the proof of [Reference ItoIto23, Proposition 4.2.8]. Thus, we can identify $ {\mathrm {Lift}} (P(\mathcal {Q})_{A/I}, \mathscr {Q}_{A'/I'})$ with the set of $P_\mu (A'/I')$ -orbits $ x P_\mu (A'/I') \subset G(A'/I') $ of elements x of the kernel of $U^{-}_{\mu }(A'/I') \to U^{-}_{\mu }(A/I)$ . Using this observation and using the fact that $U^{-}_{\mu } \times _{ \mathrm {Spec} {\mathcal O}} P_\mu \to G_{\mathcal O}$ is an open immersion again, we can identify $ {\mathrm {Lift}} (P(\mathcal {Q})_{A/I}, \mathscr {Q}_{A'/I'})$ with the kernel of $U^{-}_{\mu }(A'/I') \to U^{-}_{\mu }(A/I)$ . However, by Example 3.4.4, we have

$$\begin{align*}{\mathrm{Hom}}_{A/I}(C(\mathcal{Q}), K) = {\mathrm{Hom}}_{\mathcal O}(C(R_{G, \mu}), K). \end{align*}$$

The map $ {\mathrm {Lift}} (P(\mathcal {Q})_{A/I}, \mathscr {Q}_{A'/I'}) \to {\mathrm {Hom}} _{A/I}(C(\mathcal {Q}), K) $ sends an element $x \colon \mathrm {Spec} A'/I' \to U^{-}_{\mu }$ of the kernel of $U^{-}_{\mu }(A'/I') \to U^{-}_{\mu }(A/I)$ to the map $x^* \colon C(R_{G, \mu }) \to K$ . Using this description, it is easy to see that this map is bijective.

The proof of the following result is inspired by that of [Reference Bültel and PappasBP20, Theorem 3.5.11].

Proposition 3.4.6. Assume that $\mu $ is 1-bounded. Let $\mathcal {Q}$ be a (not necessarily banal) G- $\mu $ -display over $(A, I)$ . If $f \colon (A', I') \to (A, I)$ satisfies Assumption 3.4.1, then the set $ {\mathrm {Def}} (\mathcal {Q})_{(A', I')}$ of isomorphism classes of deformations of $\mathcal {Q}$ over $(A', I')$ has the structure of a torsor under an $A/I$ -module. In particular, $ {\mathrm {Def}} (\mathcal {Q})_{(A', I')}$ is nonempty.

Proof. We recall that the functor $ (A', I')_{\mathrm {\acute {e}t}} \to (A, I)_{\mathrm {\acute {e}t}} $ defined by $ B' \mapsto B, $ where B is the $(\pi , I)$ -adic completion of $B' \otimes _{A'} A$ , is an equivalence. Moreover, any $\pi $ -completely étale homomorphism $A/I \to C$ is étale since $\pi ^n=0$ in $A/I$ , and thus, the category $(A, I)_{\mathrm {\acute {e}t}}$ is equivalent to the category $(A/I)_{\mathrm {\acute {e}t}}$ of étale $A/I$ -algebras (see [Reference ItoIto23, Lemma 2.5.9]).

By Corollary 3.2.9, the functor

$$\begin{align*}\underline{{\mathrm{Def}}}(\mathcal{Q}) \colon (A', I')_{\mathrm{\acute{e}t}} \to \mathrm{Set}, \quad B' \mapsto \underline{{\mathrm{Def}}}(\mathcal{Q})(B')={\mathrm{Def}}(\mathcal{Q}_{(B, IB)})_{(B', I'B')} \end{align*}$$

forms a sheaf. Let $B' \in (A', I')_{\mathrm {\acute {e}t}}$ be such that $\mathcal {Q}_{(B, IB)}$ is banal. We choose a deformation $\mathscr {Q}$ of $\mathcal {Q}_{(B, IB)}$ over $(B', I'B')$ , which exists since $G(B') \to G(B)$ is surjective. To simplify the notation, the base change $\mathcal {Q}_{(B, IB)}$ is also denoted by $\mathcal {Q}$ . By Theorem 3.3.4, the period map

$$\begin{align*}{\mathrm{Per}}_{\mathscr{Q}} \colon \underline{{\mathrm{Def}}}(\mathcal{Q})(B')\to {\mathrm{Lift}}(P(\mathcal{Q})_{B/IB}, \mathscr{Q}_{B'/I'B'}) \end{align*}$$

is bijective. By Lemma 3.4.5, we have

$$\begin{align*}{\mathrm{Lift}}(P(\mathcal{Q})_{B/IB}, \mathscr{Q}_{B'/I'B'}) \overset{\sim}{\to} {\mathrm{Hom}}_{B/IB}(C(\mathcal{Q}), K \otimes_{A/I} B/IB). \end{align*}$$

We note that $K \otimes _{A/I} B/IB$ is the kernel of $B'/I'B' \to B/IB$ . We endow the set $\underline { {\mathrm {Def}} }(\mathcal {Q})(B')$ with an action of the $B/IB$ -module $ {\mathrm {Hom}} _{B/IB}(C(\mathcal {Q}), K \otimes _{A/I} B/IB)$ via these identifications. One can check that this action does not depend on the choice of $\mathscr {Q}$ . By étale descent for quasi-coherent sheaves, the $B/IB$ -modules $ {\mathrm {Hom}} _{B/IB}(C(\mathcal {Q}), K \otimes _{A/I} B/IB)$ give rise to a quasi-coherent sheaf $\mathcal {F}$ on $(A/I)^{ {\mathrm {op}} }_{\mathrm {\acute {e}t}}$ , and $\underline { {\mathrm {Def}} }(\mathcal {Q})$ is an $\mathcal {F}$ -torsor on $(A/I)^{ {\mathrm {op}} }_{\mathrm {\acute {e}t}}$ . Since any $\mathcal {F}$ -torsor on $(A/I)^{ {\mathrm {op}} }_{\mathrm {\acute {e}t}}$ is trivial, the result follows.

In the following, the maximal ideal of a local ring R is denoted by $\mathfrak {m}_R$ .

Remark 3.4.7. For a local ring R with a local homomorphism ${\mathcal O} \to R$ , we define

$$\begin{align*}\mathfrak{t}_R:=(\mathfrak{m}_R/(\mathfrak{m}^2_R+\pi R))^\vee:={\mathrm{Hom}}_k(\mathfrak{m}_R/(\mathfrak{m}^2_R+\pi R), k), \end{align*}$$

which is called the tangent space of R over ${\mathcal O}$ . When $R=R_{G, \mu }$ , we write $ \mathfrak {m}_{G, \mu }:=\mathfrak {m}_{R_{G, \mu }} $ and $ \mathfrak {t}_{G, \mu }:=\mathfrak {t}_{R_{G, \mu }} $ for simplicity. The natural homomorphism

$$\begin{align*}C(R_{G, \mu}) \otimes_{\mathcal O} k \to \mathfrak{m}_{G, \mu}/(\mathfrak{m}^2_{G, \mu}+\pi R_{G, \mu})=\mathfrak{t}^\vee_{G, \mu} \end{align*}$$

is an isomorphism.

Remark 3.4.8 (Normalized period map).

We assume that K is killed by $\pi $ , so that K is a k-vector space. Let $\mathcal {Q}$ be a banal G- $\mu $ -display over $(A, I)$ and $\mathscr {Q}$ a deformation of $\mathcal {Q}$ over $(A', I')$ . We choose an isomorphism $\beta \colon \mathcal {Q} \overset {\sim }{\to } \mathcal {Q}_{X}$ of G- $\mu $ -displays over $(A, I)$ for some $X \in G(A)_{I}$ . The isomorphism $\beta $ induces

$$\begin{align*}{\mathrm{Hom}}_{A/I}(C(\mathcal{Q}), K) \overset{\sim}{\to} {\mathrm{Hom}}_{\mathcal O}(C(R_{G, \mu}), K)=\mathfrak{t}_{G, \mu} \otimes_k K. \end{align*}$$

See Example 3.4.4 and Remark 3.4.7. By Lemma 3.4.5, we then obtain the following isomorphism:

$$\begin{align*}{\mathrm{Lift}}(P(\mathcal{Q})_{A/I}, \mathscr{Q}_{A'/I'}) \overset{\sim}{\to} \mathfrak{t}_{G, \mu} \otimes_k K. \end{align*}$$

We define

$$\begin{align*}{\mathrm{Per}}^{\beta}_{\mathscr{Q}} \colon {\mathrm{Def}}(\mathcal{Q})_{(A', I')} \to \mathfrak{t}_{G, \mu} \otimes_k K \end{align*}$$

as the composition of $ {\mathrm {Per}} _{\mathscr {Q}}$ with the above isomorphism. We call $ {\mathrm {Per}} ^{\beta }_{\mathscr {Q}}$ the normalized period map with respect to the isomorphism $ \beta \colon \mathcal {Q} \overset {\sim }{\to } \mathcal {Q}_{X}. $

If $\beta ' \colon \mathcal {Q} \overset {\sim }{\to } \mathcal {Q}_{X'}$ is another isomorphism, then there exists an $A/I$ -linear isomorphism $h \colon \mathfrak {t}_{G, \mu } \otimes _k K \overset {\sim }{\to } \mathfrak {t}_{G, \mu } \otimes _k K$ such that $ {\mathrm {Per}} ^{\beta '}_{\mathscr {Q}}=h \circ {\mathrm {Per}} ^{\beta }_{\mathscr {Q}}$ .

4 Universal deformations

This section is the main part of this paper. Throughout this section, we will assume that $\mu $ is 1-bounded. For a G- $\mu $ -display $\mathcal {Q}$ over $({\mathcal O}, (\pi ))$ , we construct a universal deformation of $\mathcal {Q}$ as a prismatic G- $\mu $ -display over $R_{G, \mu }$ , where $R_{G, \mu }$ is the local ring defined in Definition 3.4.3. The precise definition of a universal deformation is explained below. We also give some characterizations of universal deformations.

4.1 Definitions and main results

Recall that $\mathcal {C}_{\mathcal O}$ is the category of complete regular local rings R over ${\mathcal O}$ with residue field k. Let $R \in \mathcal {C}_{\mathcal O}$ . The ${\mathcal O}_E$ -prism $ ({\mathcal O}, (\pi )) $ with the projection $e \colon R \to k$ is an object of the category defined in Section 2.2.

Definition 4.1.1. Let $\mathcal {Q}$ be a G- $\mu $ -display over $({\mathcal O}, (\pi ))$ . Let $R \in \mathcal {C}_{\mathcal O}$ . A deformation of $\mathcal {Q}$ over R is a pair $(\mathfrak {Q}, \gamma )$ of a prismatic G- $\mu $ -display $\mathfrak {Q}$ over R (Definition 2.6.1) and an isomorphism $ \gamma \colon \mathfrak {Q}_{({\mathcal O}, (\pi ))} \overset {\sim }{\to } \mathcal {Q} $ of G- $\mu $ -displays over $({\mathcal O}, (\pi ))$ . An isomorphism $(\mathfrak {Q}, \gamma ) \overset {\sim }{\to } (\mathfrak {Q}', \gamma ')$ of deformations is an isomorphism $\mathfrak {Q} \overset {\sim }{\to } \mathfrak {Q}'$ of prismatic G- $\mu $ -displays over R such that the induced isomorphism $\mathfrak {Q}_{({\mathcal O}, (\pi ))} \overset {\sim }{\to } \mathfrak {Q}^{\prime }_{({\mathcal O}, (\pi ))}$ makes the following diagram commute:

When there is no danger of confusion, we write $\mathfrak {Q}$ instead of $(\mathfrak {Q}, \gamma )$ .

Remark 4.1.2. Let $(\mathfrak {Q}, \gamma )$ be a deformation of $\mathcal {Q}$ over R. Let $h \colon R \to R'$ be a morphism in $\mathcal {C}_{\mathcal O}$ . Then the base change $h^*\mathfrak {Q}$ with the isomorphism $ \gamma \colon (h^*\mathfrak {Q})_{({\mathcal O}, (\pi ))} = \mathfrak {Q}_{({\mathcal O}, (\pi ))} \overset {\sim }{\to } \mathcal {Q} $ is a deformation of $\mathcal {Q}$ over $R'$ .

Definition 4.1.3 (Universal deformation).

Let $\mathcal {Q}$ be a G- $\mu $ -display over $({\mathcal O}, (\pi ))$ . Let $R \in \mathcal {C}_{\mathcal O}$ . We say that a deformation $(\mathfrak {Q}, \gamma )$ of $\mathcal {Q}$ over R is a universal deformation if it satisfies the following property:

  • $(\ast )$ Let $R' \in \mathcal {C}_{\mathcal O}$ and let $(\mathfrak {Q}', \gamma ')$ be a deformation of $\mathcal {Q}$ over $R'$ . Then there exists a unique local homomorphism $ h \colon R \to R' $ over ${\mathcal O}$ such that $(h^*\mathfrak {Q}, \gamma )$ is isomorphic to $(\mathfrak {Q}', \gamma ')$ as a deformation of $\mathcal {Q}$ over $R'$ .

In order to state our main results, we need to introduce some more notation. We consider a map $f \colon (A', I') \to (A, I)$ of orientable and bounded ${\mathcal O}_E$ -prisms over ${\mathcal O}$ such that $f \colon A' \to A$ is surjective. Let $R \in \mathcal {C}_{\mathcal O}$ and let $ \mathfrak {Q} $ be a prismatic G- $\mu $ -display over R. Given a homomorphism $g \colon R \to A/I$ over ${\mathcal O}$ , we can regard $(A, I)$ as an object of , and we have the associated G- $\mu $ -display $\mathfrak {Q}_{g}=\mathfrak {Q}_{(A, I)}$ over $(A, I)$ . Let

$$\begin{align*}{\mathrm{Hom}}(R, A'/I')_{g} \end{align*}$$

be the set of homomorphisms $g' \colon R \to A'/I'$ over ${\mathcal O}$ which are lifts of g. For each $g' \in {\mathrm {Hom}} (R, A'/I')_{g}$ , the G- $\mu $ -display $\mathfrak {Q}_{g'}$ over $(A', I')$ is equipped with the isomorphism $ \gamma _f \colon f^*(\mathfrak {Q}_{g'}) \overset {\sim }{\to } \mathfrak {Q}_{g}; $ namely, it is a deformation of $\mathfrak {Q}_{g}$ . In conclusion, we have a map

$$\begin{align*}{\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R, A'/I')_{g} \to {\mathrm{Def}}(\mathfrak{Q}_g)_{(A', I')}, \quad g' \mapsto \mathfrak{Q}_{g'} \end{align*}$$

of sets. We call $ {\mathrm {ev}} _{\mathfrak {Q}}$ the evaluation map.

Example 4.1.4. Let $\mathcal {Q}$ be a G- $\mu $ -display over $({\mathcal O}, (\pi ))$ . Let $R \in \mathcal {C}_{\mathcal O}$ and let $ \mathfrak {Q} $ be a deformation of $\mathcal {Q}$ over R. We consider the nilpotent thickening

$$\begin{align*}f \colon ({\mathcal O}[[t]]/t^2, (\pi)) \to ({\mathcal O}, (\pi)) \end{align*}$$

of Breuil–Kisin type. By the above construction, we have a map

$$\begin{align*}{\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R, k[[t]]/t^2)_{e} \to {\mathrm{Def}}(\mathcal{Q})_{({\mathcal O}[[t]]/t^2, (\pi))}, \quad g \mapsto \mathfrak{Q}_{g}. \end{align*}$$

The set $ {\mathrm {Hom}} (R, k[[t]]/t^2)_{e}$ can be naturally identified with the tangent space $\mathfrak {t}_{R}$ of R over ${\mathcal O}$ (cf. Remark 3.4.7). We write

$$\begin{align*}{\mathrm{KS}}(\mathfrak{Q}) \colon \mathfrak{t}_{R} \to {\mathrm{Def}}(\mathcal{Q})_{({\mathcal O}[[t]]/t^2, (\pi))} \end{align*}$$

for the above map $ {\mathrm {ev}} _{\mathfrak {Q}}$ .

Definition 4.1.5 (Kodaira–Spencer map).

The map $ {\mathrm {KS}} (\mathfrak {Q})$ defined in Example 4.1.4 is called the Kodaira–Spencer map of $\mathfrak {Q}$ . We say that $\mathfrak {Q}$ is versal if the Kodaira–Spencer map $ {\mathrm {KS}} (\mathfrak {Q})$ is surjective.

Definition 4.1.6. Let $R \in \mathcal {C}_{\mathcal O}$ and let $ \mathfrak {Q} $ be a prismatic G- $\mu $ -display over R. We say that $\mathfrak {Q}$ has the property $(\mathrm {Perfd})$ if it satisfies the following condition:

  • $(\ast )$ Let $(S, a^\flat )$ be a perfectoid pair over ${\mathcal O}$ . Let $e \colon R \to S/a$ be the composition $R \to k \to S/a$ . Then the evaluation map

    $$\begin{align*}{\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R, S/a^{m})_{e} \to {\mathrm{Def}}(\mathfrak{Q}_e)_{(W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^{m}, I_S)} \end{align*}$$
    is bijective for any $m \geq 1$ .

We say that $\mathfrak {Q}$ has the property $(\mathrm {BK})$ if it satisfies the following condition:

  • $(\ast )$ Let $\widetilde {k}$ be a perfect field containing k. We set $\widetilde {{\mathcal O}}:=W(\widetilde {k}) \otimes _{W({\mathbb F}_q)} {\mathcal O}_E$ . Let $(\mathfrak {S}_{\widetilde {{\mathcal O}}}, (\mathcal {E}))$ be an ${\mathcal O}_E$ -prism of Breuil–Kisin type over $\widetilde {{\mathcal O}}$ , where $\mathfrak {S}_{\widetilde {{\mathcal O}}}=\widetilde {{\mathcal O}}[[t_1, \dotsc , n]]$ for an integer $n \geq 0$ . We set $\widetilde {R}:=\mathfrak {S}_{\widetilde {{\mathcal O}}}/\mathcal {E}$ and $\widetilde {R}_m:=\widetilde {R}/\mathfrak {m}^{m}_{\widetilde {R}}=\mathfrak {S}_{\widetilde {{\mathcal O}}, m}/\mathcal {E}$ for an integer $m \geq 1$ . Let $e \colon R \to \widetilde {k}$ be the natural homomorphism. Then the evaluation map

    $$\begin{align*}{\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R, \widetilde{R}_m)_{e} \to {\mathrm{Def}}(\mathfrak{Q}_e)_{(\mathfrak{S}_{\widetilde{{\mathcal O}}, m}, (\mathcal{E}))} \end{align*}$$
    is bijective for any $m \geq 1$ .

The goal of this section is to prove the following two results. The first one is about the existence of universal deformations and their properties. Recall that $R_{G, \mu }$ is the completion of ${\mathcal O}_{U^{-}_{\mu }, 1}$ with respect to the maximal ideal; see Definition 3.4.3. There exists an isomorphism $R_{G, \mu } \simeq {\mathcal O}[[t_1, \dotsc , t_r]]$ over ${\mathcal O}$ , so that $R_{G, \mu } \in \mathcal {C}_{\mathcal O}$ .

Theorem 4.1.7. Assume that $\mu $ is 1-bounded. For a G- $\mu $ -display $\mathcal {Q}$ over $({\mathcal O}, (\pi ))$ , there exists a universal deformation $(\mathfrak {Q}^{ {\mathrm {univ}} }, \gamma )$ of $\mathcal {Q}$ over $R_{G, \mu }$ . Moreover, $\mathfrak {Q}^{ {\mathrm {univ}} }$ has the properties $(\mathrm {Perfd})$ and $(\mathrm {BK})$ . In particular, $(\mathfrak {Q}^{ {\mathrm {univ}} }, \gamma )$ is versal.

The second one gives some characterizations of universal deformations. Let ${\mathcal O}_C$ be the $\pi $ -adic completion of the integral closure of ${\mathcal O}$ in an algebraic closure of ${\mathcal O}[1/\pi ]$ . Then ${\mathcal O}_C$ is a $\pi $ -adically complete valuation ring of rank $1$ with algebraically closed fraction field. We note that ${\mathcal O}_C$ is a perfectoid ring. We denote by ${\mathcal O}_{C^\flat }$ the tilt of ${\mathcal O}_C$ . There is an element $\pi ^\flat \in {\mathcal O}_{C^\flat }$ with $\theta ([\pi ^\flat ])=\pi $ , so that $({\mathcal O}_C, \pi ^\flat )$ is a perfectoid pair over ${\mathcal O}$ .

Theorem 4.1.8. Assume that $\mu $ is 1-bounded. Let $\mathcal {Q}$ be a G- $\mu $ -display over $({\mathcal O}, (\pi ))$ . Let $R \in \mathcal {C}_{\mathcal O}$ and let $(\mathfrak {Q}, \gamma )$ be a deformation of $\mathcal {Q}$ over R. Then the following statements are equivalent:

  1. (1) $(\mathfrak {Q}, \gamma )$ is a universal deformation.

  2. (2) $\mathfrak {Q}$ has the property $(\mathrm {BK})$ .

  3. (3) $\mathfrak {Q}$ has the property $(\mathrm {Perfd})$ .

  4. (4) Let $e \colon R \to {\mathcal O}_C/\pi $ be the composition $R \to k \to {\mathcal O}_C/\pi $ . The evaluation map

    $$\begin{align*}{\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R, {\mathcal O}_C/\pi^{2})_{e} \to {\mathrm{Def}}(\mathfrak{Q}_e)_{(W_{{\mathcal O}_E}({\mathcal O}_{C^\flat})/[\pi^\flat]^{2}, I_{{\mathcal O}_C})} \end{align*}$$
    is bijective.
  5. (5) The equalities $\dim R=\dim R_{G, \mu }$ and $\dim _k \mathfrak {t}_R=\dim _k \mathfrak {t}_{G, \mu }$ hold, and $(\mathfrak {Q}, \gamma )$ is versal. (Here, $\dim R$ is the dimension of the ring R and $\dim _k \mathfrak {t}_R$ is the dimension of the k-vector space $\mathfrak {t}_R$ .)

The rest of this section is devoted to the proofs of Theorem 4.1.7 and Theorem 4.1.8. In Section 4.2, we present a few technical results on maps of ${\mathcal O}_E$ -prisms. In Section 4.3, we investigate the relation between $\mathrm {(Perfd)}$ , $\mathrm {(BK)}$ and the property of being a universal deformation. For this, we discuss the linearity of evaluation maps, which plays an important role in our discussion. Finally, in Section 4.4, we complete the proofs of Theorem 4.1.7 and Theorem 4.1.8.

4.2 Preliminaries on maps of ${\mathcal O}_E$ -prisms

Let $R \in \mathcal {C}_{\mathcal O}$ . We fix an ${\mathcal O}_E$ -prism $ (\mathfrak {S}_{\mathcal O}, (\mathcal {E})) $ of Breuil–Kisin type, where $\mathfrak {S}_{\mathcal O}={\mathcal O}[[t_1, \dotsc , t_n]]$ , with an isomorphism $R \simeq \mathfrak {S}_{\mathcal O}/\mathcal {E}$ over ${\mathcal O}$ . We note that R is of dimension n. (Here, n is allowed to be zero.)

Let $(S, a^\flat )$ be a perfectoid pair over ${\mathcal O}$ . Let $m \geq 1$ be an integer. The following notion of primitive map will play an important role in our proof of Theorem 4.1.7.

Definition 4.2.1 (Primitive map).

A map

$$\begin{align*}f \colon (\mathfrak{S}_{\mathcal O}, (\mathcal{E})) \to (W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^m, I_S) \end{align*}$$

of ${\mathcal O}_E$ -prisms over ${\mathcal O}$ is called primitive with respect to $a^\flat $ (or simply, primitive) if for every $1 \leq i \leq n$ , the map f sends $t_i$ to $[a^\flat b^\flat _i]$ for some element $b^\flat _i \in S^\flat $ . (Here, we denote the image of $[a^\flat b^\flat _i] \in W_{{\mathcal O}_E}(S^\flat )$ in $W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m$ by the same symbol.)

We note that a primitive map $(\mathfrak {S}_{\mathcal O}, (\mathcal {E})) \to (W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m, I_S)$ factors through a map $(\mathfrak {S}_{{\mathcal O}, m}, (\mathcal {E})) \to (W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m, I_S)$ .

Lemma 4.2.2. Assume that the map $S \to S$ , $x \mapsto x^p$ is surjective. Then, for a homomorphism $g \colon R \to S/a^m$ over ${\mathcal O}$ which is a lift of the composition $R \to k \to S/a$ , there exists a primitive map $ f \colon (\mathfrak {S}_{\mathcal O}, (\mathcal {E})) \to (W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m, I_S) $ which induces g.

Proof. For every $1 \leq i \leq n$ , the image of $t_i$ under the composition $\mathfrak {S}_{\mathcal O} \to R \to S/a^m$ can be written as $a b_i$ for some element $b_i \in S$ . (Again, we abuse notation and denote the image of $a b_i \in S$ in $S/a^m$ by the same symbol.) Since the map $S \to S$ , $x \mapsto x^p$ is surjective, there exists an element $b^\flat _i \in S^\flat $ with $\theta ([b^\flat _i])=b_i$ (cf. [Reference Bhatt, Morrow and ScholzeBMS18, Lemma 3.3]). We define a homomorphism $f \colon \mathfrak {S}_{\mathcal O} \to W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m$ over ${\mathcal O}$ by $t_i \mapsto [a^\flat b^\flat _i]$ . Since $\phi ([a^\flat b^\flat _i])=[a^\flat b^\flat _i]^q$ , it follows that f is $\phi $ -equivariant, which in turn implies that f is a homomorphism of $\delta _E$ -rings since both the target and the source are $\pi $ -torsion free (see Lemma 2.1.2). By construction, we see that f is a lift of g. In particular, f sends $\mathcal {E}$ into the kernel of $W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m \to S/a^m$ . Thus, we have a primitive map $ f \colon (\mathfrak {S}_{\mathcal O}, (\mathcal {E})) \to (W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m, I_S) $ which induces g.

Remark 4.2.3. Any perfectoid ring S admits a p-completely faithfully flat homomorphism $S \to S'$ of perfectoid rings such that the map $S' \to S'$ , $x \mapsto x^p$ is surjective. This is a consequence of André’s flatness lemma; see [Reference Bhatt and ScholzeBS22, Theorem 7.14].

Lemma 4.2.4. Let $ f \colon (\mathfrak {S}_{\mathcal O}, (\mathcal {E})) \to (W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m, I_S) $ be a primitive map and let $g \colon R \to S/a^m$ be the induced homomorphism. Suppose that we are given a homomorphism $g' \colon R \to S/a^{m+1}$ over ${\mathcal O}$ which is a lift of g. Then there exists a map $ f' \colon (\mathfrak {S}_{\mathcal O}, (\mathcal {E})) \to (W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m+1}, I_S) $ which induces f and $g'$ .

Proof. Since f is primitive, it maps $t_i$ to $[a^\flat b^\flat _i] \in W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^m$ for some element $b^\flat _i \in S^\flat $ for any i. Let $b_i:=\theta ([b^\flat _i]) \in S$ . The image of $t_i$ under the composition $\mathfrak {S}_{\mathcal O} \to R \to S/a^{m+1}$ can be written as $ab_i+a^m y_i$ for some element $y_i \in S/a^{m+1}$ . Let $x_i \in W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m+1}$ be such that $\theta _{{\mathcal O}_E}(x_i)=y_i$ in $S/a^{m+1}$ . We define a homomorphism $f' \colon \mathfrak {S}_{\mathcal O} \to W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m+1}$ over ${\mathcal O}$ by $t_i \mapsto [a^\flat b^\flat _i]+[a^\flat ]^m x_i$ . By construction, it is a lift of $g'$ . We claim that $f'$ is a homomorphism of $\delta _E$ -rings. Indeed, we have

$$\begin{align*}\phi([a^\flat b^\flat_i]+[a^\flat]^m x_i)=[a^\flat b^\flat_i]^q + [a^\flat]^{qm} \phi(x_i)= [a^\flat b^\flat_i]^q \end{align*}$$

in $W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m+1}$ since $qm \geq m+1$ . Similarly, we have $ ([a^\flat b^\flat _i]+[a^\flat ]^m x_i)^q=[a^\flat b^\flat _i]^q. $ It follows that $f'$ is $\phi $ -equivariant, which in turn implies that it is a homomorphism of $\delta _E$ -rings. The resulting map $ f' \colon (\mathfrak {S}_{\mathcal O}, (\mathcal {E})) \to (W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m+1}, I_S) $ has the desired properties.

Lemma 4.2.5. Assume that k is algebraically closed. Let $m \geq 1$ be an integer and $x \in \mathfrak {m}^m_R/\mathfrak {m}^{m+1}_R$ a nonzero element. Then there exists a complete regular local ring $R'$ of dimension $1$ with a surjection $R \to R'$ such that the induced homomorphism $ \mathfrak {m}^m_R/\mathfrak {m}^{m+1}_R \to \mathfrak {m}^m_{R'}/\mathfrak {m}^{m+1}_{R'} $ sends x to a nonzero element.

Proof. Let $x_1, \dotsc , x_n \in \mathfrak {m}_R$ be a regular system of parameters. The homomorphism of graded k-algebras

$$\begin{align*}k[x_1, \dotsc, x_n] \to \bigoplus_{l \geq 0} \mathfrak{m}^l_R/\mathfrak{m}^{l+1}_R \end{align*}$$

defined by $x_i \mapsto x_i \in \mathfrak {m}_R/\mathfrak {m}^{2}_R$ is an isomorphism. Since $\mathfrak {m}^m_R/\mathfrak {m}^{m+1}_R$ is nonzero, we have $n \geq 1$ . If $n = 1$ , then the assertion holds trivially. We assume that $n \geq 2$ .

We claim that, for a nonzero homogeneous polynomial $f(x_1, \dotsc , x_n)$ of degree m, there exist two distinct integers $i, j \in \{ 1, \dotsc , n \}$ and an element $b \in k$ such that the image of $f(x_1, \dotsc , x_n)$ in $k[x_1, \dotsc , x_n]/(x_i-bx_j)$ is nonzero. Indeed, since k is infinite, there is a nonzero element $(b_1, \dotsc , b_n) \in k^n$ such that $f(b_1, \dotsc , b_n) \neq 0$ . Without loss of generality, we may assume that $b_{n-1} \neq 0$ . We set $b:=b_n/b_{n-1}$ . Then the polynomial $f(x_1, \dotsc , x_{n-1}, bx_{n-1}) \in k[x_1, \dotsc , x_{n-1}]$ is nonzero since its evaluation at $(b_1, \dotsc , b_{n-1})$ is nonzero.

The claim implies that there exist two distinct integers $i, j \in \{ 1, \dotsc , n \}$ and an element $b \in R$ such that, for the quotient $R":=R/(x_i-bx_j)$ , the induced homomorphism $ \mathfrak {m}^m_R/\mathfrak {m}^{m+1}_R \to \mathfrak {m}^m_{R"}/\mathfrak {m}^{m+1}_{R"} $ sends x to a nonzero element. Since $R"$ is regular, the assertion follows by repeating this procedure.

Corollary 4.2.6. Let $m \geq 1$ be an integer and $x \in \mathfrak {m}^m_R/\mathfrak {m}^{m+1}_R$ a nonzero element. Then there exists a perfectoid pair $(V, a^\flat )$ over ${\mathcal O}$ with the following properties:

  1. (1) V is a $\pi $ -adically complete valuation ring of rank $1$ with algebraically closed fraction field.

  2. (2) There exists a primitive map

    $$\begin{align*}(\mathfrak{S}_{\mathcal O}, (\mathcal{E})) \to (W_{{\mathcal O}_E}(V^\flat)/[a^\flat]^{m+1}, I_V) \end{align*}$$
    such that the induced homomorphism $ \mathfrak {m}^m_R/\mathfrak {m}^{m+1}_R \to a^mV/a^{m+1}V $ sends x to a nonzero element.

Proof. We may assume that k is algebraically closed. By Lemma 4.2.5, there exists a complete regular local ring $R'$ of dimension $1$ with a surjection $R \to R'$ such that the induced homomorphism $ \mathfrak {m}^m_R/\mathfrak {m}^{m+1}_R \to \mathfrak {m}^m_{R'}/\mathfrak {m}^{m+1}_{R'} $ sends x to a nonzero element.

Let $a \in \mathfrak {m}_{R'}$ be a uniformizer. Let V be the $\pi $ -adic completion of the integral closure of $R'$ in an algebraic closure of $R'[1/a]$ . Then V is a $\pi $ -adically complete valuation ring of rank $1$ with algebraically closed fraction field. The natural homomorphism $g \colon R \to V/a^{m+1}V$ sends $\mathfrak {m}_R$ into $(a)$ , and the induced homomorphism $ \mathfrak {m}^m_R/\mathfrak {m}^{m+1}_R \to a^mV/a^{m+1}V $ sends x to a nonzero element. There exists an element $a^\flat \in V^\flat $ with $\theta ([a^\flat ])=a$ , so that $(V, a^\flat )$ is a perfectoid pair over ${\mathcal O}$ . By Lemma 4.2.2, there exists a primitive map $ (\mathfrak {S}_{\mathcal O}, (\mathcal {E})) \to (W_{{\mathcal O}_E}(V^\flat )/[a^\flat ]^{m+1}, I_V) $ which induces g. This concludes the proof.

4.3 Linearity of evaluation maps

Let $\mathcal {Q}$ be a G- $\mu $ -display over $({\mathcal O}, (\pi ))$ . For deformations of $\mathcal {Q}$ , we investigate the relation between $\mathrm {(Perfd)}$ , $\mathrm {(BK)}$ , and the property of being a universal deformation.

We first note the following consequence of Theorem 2.6.5.

Remark 4.3.1. Let $R \in \mathcal {C}_{\mathcal O}$ . We choose an ${\mathcal O}_E$ -prism $(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))$ of Breuil–Kisin type, where $\mathfrak {S}_{\mathcal O}={\mathcal O}[[t_1, \dotsc , t_n]]$ , with an isomorphism $R \simeq \mathfrak {S}_{\mathcal O}/\mathcal {E}$ over ${\mathcal O}$ . Let $ f \colon (\mathfrak {S}_{\mathcal O}, (\mathcal {E})) \to ({\mathcal O}, (\pi )) $ be the morphism defined by $t_i \mapsto 0$ . For a deformation $\mathfrak {Q}$ of $\mathcal {Q}$ over R, the value $\mathfrak {Q}_{(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))}$ is naturally a deformation of $\mathcal {Q}$ over $(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))$ . By Theorem 2.6.5, this construction induces an equivalence from the groupoid of deformations of $\mathcal {Q}$ over R to that of deformations of $\mathcal {Q}$ over $(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))$ .

Corollary 4.3.2. Let $R \in \mathcal {C}_{\mathcal O}$ . There is at most one isomorphism between two deformations of $\mathcal {Q}$ over R.

Proof. This follows from Remark 4.3.1 and Corollary 3.2.10.

Corollary 4.3.3. Let $R \in \mathcal {C}_{\mathcal O}$ and let $\mathfrak {Q}$ be a deformation of $\mathcal {Q}$ over R. If $\mathfrak {Q}$ has the property $\mathrm {(BK)}$ , then $\mathfrak {Q}$ is a universal deformation of $\mathcal {Q}$ .

Proof. Let $R' \in \mathcal {C}_{\mathcal O}$ and let $\mathfrak {Q}'$ be a deformation of $\mathcal {Q}$ over $R'$ . We want to show that there exists a unique local homomorphism $ h \colon R \to R' $ over ${\mathcal O}$ such that $h^*\mathfrak {Q}$ is isomorphic to $\mathfrak {Q}'$ as a deformation of $\mathcal {Q}$ . We fix an ${\mathcal O}_E$ -prism $(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))$ of Breuil–Kisin type with an isomorphism $ R' \simeq \mathfrak {S}_{\mathcal O}/\mathcal {E} $ over ${\mathcal O}$ . By Example 3.1.6 and Corollary 3.2.10, we have

$$\begin{align*}{\mathrm{Def}}(\mathcal{Q})_{(\mathfrak{S}_{\mathcal O}, (\mathcal{E}))} \overset{\sim}{\to} {\varprojlim}_m {\mathrm{Def}}(\mathcal{Q})_{(\mathfrak{S}_{{\mathcal O}, m}, (\mathcal{E}))}. \end{align*}$$

Since $\mathfrak {Q}$ has the property $\mathrm {(BK)}$ , we see that the evaluation map

$$\begin{align*}{\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R, R')_{e} \to {\mathrm{Def}}(\mathcal{Q})_{(\mathfrak{S}_{\mathcal O}, (\mathcal{E}))}, \quad h \mapsto \mathfrak{Q}_h \end{align*}$$

is bijective. (Here, $e \colon R \to k$ is the projection.) For a local homomorphism $h \colon R \to R'$ over ${\mathcal O}$ , it follows from Remark 4.3.1 that $h^*\mathfrak {Q} \simeq \mathfrak {Q}'$ if and only if $\mathfrak {Q}_h \simeq \mathfrak {Q}^{\prime }_{(\mathfrak {S}_{\mathcal O}, (\mathcal {E}))}$ . Since the above map $ {\mathrm {ev}} _{\mathfrak {Q}}$ is bijective, such a homomorphism h exists and is unique.

Remark 4.3.4. If a deformation $ \mathfrak {Q} $ of $\mathcal {Q}$ over R for some $R \in \mathcal {C}_{\mathcal O}$ satisfies the property $(\mathrm {BK})$ , then the Kodaira–Spencer map $ {\mathrm {KS}} (\mathfrak {Q})$ is bijective, so that $\mathfrak {Q}$ is versal.

Remark 4.3.5. Assume that there exists a deformation $\mathfrak {Q}$ of $\mathcal {Q}$ over R for some $R \in \mathcal {C}_{\mathcal O}$ such that $\mathfrak {Q}$ has the property $\mathrm {(BK)}$ . It follows from Corollary 4.3.3 that $\mathfrak {Q}$ is a universal deformation. Let $\mathfrak {Q}'$ be another deformation of $\mathcal {Q}$ over $R'$ for some $R' \in \mathcal {C}_{\mathcal O}$ . There exists a unique local homomorphism $ h \colon R \to R' $ over ${\mathcal O}$ such that $h^*\mathfrak {Q} \simeq \mathfrak {Q}'$ by the universal property. For the homomorphism $ \mathfrak {t}_{R'} \to \mathfrak {t}_{R} $ induced from h, the following diagram commutes:

Since $ {\mathrm {KS}} (\mathfrak {Q})$ is bijective, we see that $\mathfrak {Q}'$ is versal if and only if $\mathfrak {t}_{R'} \to \mathfrak {t}_{R}$ is surjective.

To proceed further, we need to introduce two more conditions (Perfd-lin) and (BK-lin).

Remark 4.3.6. We retain the notation from Definition 4.1.6 with $R=R_{G, \mu }$ . Let $m \geq 1$ be an integer. Let $g \in {\mathrm {Hom}} (R_{G, \mu }, S/a^{m})_{e}$ . For any $h \in {\mathrm {Hom}} (R_{G, \mu }, S/a^{m+1})_{g}$ , the map

$$\begin{align*}\mathrm{Diff}_{h} \colon {\mathrm{Hom}}(R_{G, \mu}, S/a^{m+1})_{g} \to {\mathrm{Hom}}_k(\mathfrak{t}^\vee_{G, \mu}, a^{m}S/a^{m+1}S)=\mathfrak{t}_{G, \mu} \otimes_k (a^{m}S/a^{m+1}S) \end{align*}$$

defined by $g' \mapsto g'-h$ is bijective.

Similarly, let $g \in {\mathrm {Hom}} (R_{G, \mu }, \widetilde {R}_m)_{e}$ . For any $h \in {\mathrm {Hom}} (R_{G, \mu }, \widetilde {R}_{m+1})_{g}$ , we have the following bijection:

$$\begin{align*}\mathrm{Diff}_{h} \colon {\mathrm{Hom}}(R_{G, \mu}, \widetilde{R}_{m+1})_{g} \overset{\sim}{\to} \mathfrak{t}_{G, \mu} \otimes_k (\mathfrak{m}^{m}_{\widetilde{R}}/\mathfrak{m}^{m+1}_{\widetilde{R}}), \quad g' \mapsto g'-h. \end{align*}$$

Definition 4.3.7. We retain the notation of Definition 4.1.6. Let $ \mathfrak {Q} $ be a prismatic G- $\mu $ -display over $R_{G, \mu }$ . We say that $\mathfrak {Q}$ has the property (Perfd-lin) if it satisfies the following condition:

  • $(\ast )$ Let $(S, a^\flat )$ be a perfectoid pair over ${\mathcal O}$ such that $\mathfrak {Q}_e$ is a banal G- $\mu $ -display over $(W_{{\mathcal O}_E}(S^\flat )/[a^\flat ], I_S)$ . Let $m \geq 1$ be an integer and let $g \in {\mathrm {Hom}} (R_{G, \mu }, S/a^{m})_{e}$ . Then, for any $h \in {\mathrm {Hom}} (R_{G, \mu }, S/a^{m+1})_{g}$ and any isomorphism $\beta \colon \mathfrak {Q}_g \overset {\sim }{\to } \mathcal {Q}_X$ , the following composition

    $$ \begin{align*} \begin{split} \mathfrak{t}_{G, \mu} \otimes_k &(a^{m}S/a^{m+1}S) \overset{\mathrm{Diff}^{-1}_{h}}{\longrightarrow} {\mathrm{Hom}}(R_{G, \mu}, S/a^{m+1})_{g} \\ &\overset{{\mathrm{ev}}_{\mathfrak{Q}}}{\longrightarrow} {\mathrm{Def}}(\mathfrak{Q}_g)_{(W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^{m+1}, I_S)} \overset{{\mathrm{Per}}^{\beta}_{\mathfrak{Q}_h}}{\longrightarrow} \mathfrak{t}_{G, \mu} \otimes_k (a^{m}S/a^{m+1}S) \end{split} \end{align*} $$
    is an $S/a$ -linear isomorphism. See Remark 3.4.8 for the normalized period map $ {\mathrm {Per}} ^{\beta }_{\mathfrak {Q}_h}$ , which is bijective by Theorem 3.3.4. The above composition is denoted by $ \mathrm {L}^{\beta }_{\mathfrak {Q}, h}. $

We say that $\mathfrak {Q}$ has the property (BK-lin) if it satisfies the following condition:

  • $(\ast )$ Let $\widetilde {k}$ be a perfect field containing k such that $\mathfrak {Q}_{(\widetilde {{\mathcal O}}, (\pi ))}$ is banal. Let $(\mathfrak {S}_{\widetilde {{\mathcal O}}}, (\mathcal {E}))$ be an ${\mathcal O}_E$ -prism of Breuil–Kisin type over $\widetilde {{\mathcal O}}$ . Let $m \geq 1$ be an integer and let $g \in {\mathrm {Hom}} (R_{G, \mu }, \widetilde {R}_m)_{e}$ . Then, for any $h \in {\mathrm {Hom}} (R_{G, \mu }, \widetilde {R}_{m+1})_{g}$ and any isomorphism $\beta \colon \mathfrak {Q}_g \overset {\sim }{\to } \mathcal {Q}_X$ , the following composition

    $$ \begin{align*} \begin{split} \mathfrak{t}_{G, \mu} \otimes_k &(\mathfrak{m}^{m}_{\widetilde{R}}/\mathfrak{m}^{m+1}_{\widetilde{R}}) \overset{\mathrm{Diff}^{-1}_{h}}{\longrightarrow} {\mathrm{Hom}}(R_{G, \mu}, \widetilde{R}_{m+1})_{g} \\ &\overset{{\mathrm{ev}}_{\mathfrak{Q}}}{\longrightarrow} {\mathrm{Def}}(\mathfrak{Q}_g)_{(\mathfrak{S}_{\widetilde{{\mathcal O}}, m+1}, (\mathcal{E}))} \overset{{\mathrm{Per}}^{\beta}_{\mathfrak{Q}_h}}{\longrightarrow} \mathfrak{t}_{G, \mu} \otimes_k (\mathfrak{m}^{m}_{\widetilde{R}}/\mathfrak{m}^{m+1}_{\widetilde{R}}) \end{split} \end{align*} $$
    is a $\widetilde {k}$ -linear isomorphism. This composition is denoted by $ \mathrm {L}^{\beta }_{\mathfrak {Q}, h}. $

Remark 4.3.8. Let the notation be as in Definition 4.3.7. The property (Perfd-lin) can be checked $\pi $ -completely flat locally on S. More precisely, let $S \to S'$ be a $\pi $ -completely faithfully flat homomorphism of perfectoid rings. Let $S"$ be the $\pi $ -adic completion of $S' \otimes _S S'$ , which is a perfectoid ring (see [Reference Česnavičius and ScholzeČS24, Proposition 2.1.11]). If the condition in Definition 4.3.7 is satisfied for $(S', a^\flat )$ and $(S", a^\flat )$ (cf. Lemma 2.2.3 (1)), then it is also satisfied for $(S, a^\flat )$ . Indeed, since $\pi =0$ in $S/a$ , we see that $S/a \to S'/a$ is faithfully flat and $S"/a = S'/a \otimes _{S/a} S'/a$ . Therefore, the sequence

$$\begin{align*}\mathfrak{t}_{G, \mu} \otimes_k (a^{m}S/a^{m+1}S) \to \mathfrak{t}_{G, \mu} \otimes_k (a^{m}S'/a^{m+1}S') \rightrightarrows \mathfrak{t}_{G, \mu} \otimes_k (a^{m}S"/a^{m+1}S") \end{align*}$$

is exact by flat descent, from which our claim follows. (The same holds if we replace $S"$ by a perfectoid ring which is $\pi $ -completely faithfully flat over the $\pi $ -adic completion of $S' \otimes _S S'$ .)

Lemma 4.3.9. Let $ \mathfrak {Q} $ be a prismatic G- $\mu $ -display over $R_{G, \mu }$ .

  1. (1) Assume that $\mathfrak {Q}$ has the property . Then $\mathfrak {Q}$ has the property $\mathrm {(Perfd)}$ .

  2. (2) Assume that $\mathfrak {Q}$ has the property . Then $\mathfrak {Q}$ has the property $\mathrm {(BK)}$ .

Proof. (1) Let $(S, a^\flat )$ be a perfectoid pair over ${\mathcal O}$ . We want to prove that the map

$$\begin{align*}{\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R_{G, \mu}, S/a^{m})_{e} \to {\mathrm{Def}}(\mathfrak{Q}_e)_{(W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^{m}, I_S)} \end{align*}$$

is bijective for any $m \geq 1$ . It suffices to prove that for any $m \geq 1$ and any $g \in {\mathrm {Hom}} (R_{G, \mu }, S/a^{m})_{e}$ , the map

$$\begin{align*}{\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R_{G, \mu}, S/a^{m+1})_{g} \to {\mathrm{Def}}(\mathfrak{Q}_g)_{(W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^{m+1}, I_S)} \end{align*}$$

is bijective. By Corollary 3.2.9, we may assume that $\mathfrak {Q}_e$ is banal. Then the assertion follows from the property and Theorem 3.3.4.

(2) This follows by the same argument as in (1).

The following proposition is the main reason for introducing (Perfd-lin) and (BK-lin).

Proposition 4.3.10. Let $ \mathfrak {Q} $ be a prismatic G- $\mu $ -display over $R_{G, \mu }$ . Assume that $\mathfrak {Q}$ has the property . Then $\mathfrak {Q}$ has the property .

Proof. With the notation of Definition 4.3.7, we want to prove that for any homomorphism $h \in {\mathrm {Hom}} (R_{G, \mu }, \widetilde {R}_{m+1})_{g}$ and any isomorphism $\beta \colon \mathfrak {Q}_g \overset {\sim }{\to } \mathcal {Q}_X$ , the map

$$\begin{align*}\mathrm{L}^{\beta}_{\mathfrak{Q}, h} \colon \mathfrak{t}_{G, \mu} \otimes_k (\mathfrak{m}^{m}_{\widetilde{R}}/\mathfrak{m}^{m+1}_{\widetilde{R}}) \to \mathfrak{t}_{G, \mu} \otimes_k (\mathfrak{m}^{m}_{\widetilde{R}}/\mathfrak{m}^{m+1}_{\widetilde{R}}) \end{align*}$$

is a $\widetilde {k}$ -linear isomorphism. We first prove that $\mathrm {L}^{\beta }_{\mathfrak {Q}, h}$ is a $\widetilde {k}$ -linear homomorphism. For two elements $g_1, g_2 \in \mathfrak {t}_{G, \mu } \otimes _k (\mathfrak {m}^{m}_{\widetilde {R}}/\mathfrak {m}^{m+1}_{\widetilde {R}})$ , we claim that

$$\begin{align*}\mathrm{L}^{\beta}_{\mathfrak{Q}, h}(g_1+g_2)=\mathrm{L}^{\beta}_{\mathfrak{Q}, h}(g_1)+\mathrm{L}^{\beta}_{\mathfrak{Q}, h}(g_2). \end{align*}$$

Indeed, by virtue of Corollary 4.2.6, it suffices to prove that for any perfectoid pair $(V, a^\flat )$ over $\widetilde {{\mathcal O}}$ and any map $ f \colon (\mathfrak {S}_{\widetilde {{\mathcal O}}}, (\mathcal {E})) \to (W_{{\mathcal O}_E}(V^\flat )/[a^\flat ]^{m+1}, I_V) $ over $\widetilde {{\mathcal O}}$ , the induced homomorphism

$$\begin{align*}\mathfrak{t}_{G, \mu} \otimes_k (\mathfrak{m}^{m}_{\widetilde{R}}/\mathfrak{m}^{m+1}_{\widetilde{R}}) \to \mathfrak{t}_{G, \mu} \otimes_k (a^mV/a^{m+1}V) \end{align*}$$

sends the element $ \mathrm {L}^{\beta }_{\mathfrak {Q}, h}(g_1+g_2)-\mathrm {L}^{\beta }_{\mathfrak {Q}, h}(g_1)-\mathrm {L}^{\beta }_{\mathfrak {Q}, h}(g_2) $ to $0$ . Let $h' \colon R_{G, \mu } \to V/a^{m+1}$ be the composition of h with the induced homomorphism $\widetilde {R}_{m+1} \to V/a^{m+1}$ , and let $\beta '$ be the base change of $\beta $ along the map $ (\mathfrak {S}_{\widetilde {{\mathcal O}}, m}, (\mathcal {E})) \to (W_{{\mathcal O}_E}(V^\flat )/[a^\flat ]^{m}, I_V) $ induced from f. The following diagram is commutative:

Since $\mathrm {L}^{\beta '}_{\mathfrak {Q}, h'}$ is a $V/a$ -linear homomorphism by our assumption, the claim follows. We can prove that $ \mathrm {L}^{\beta }_{\mathfrak {Q}, h}(bg)=b\mathrm {L}^{\beta }_{\mathfrak {Q}, h}(g) $ for any $b \in \widetilde {k}$ and any $g \in \mathfrak {t}_{G, \mu } \otimes _k (\mathfrak {m}^{m}_R/\mathfrak {m}^{m+1}_R)$ in the same way. Thus, we conclude that $\mathrm {L}^{\beta }_{\mathfrak {Q}, h}$ is a $\widetilde {k}$ -linear homomorphism.

It remains to prove that $\mathrm {L}^{\beta }_{\mathfrak {Q}, h}$ is bijective. Since we have shown that $\mathrm {L}^{\beta }_{\mathfrak {Q}, h}$ is a $\widetilde {k}$ -linear endomorphism of the finite dimensional $\widetilde {k}$ -vector space $\mathfrak {t}_{G, \mu } \otimes _k (\mathfrak {m}^{m}_{\widetilde {R}}/\mathfrak {m}^{m+1}_{\widetilde {R}})$ , it suffices to prove that its kernel is zero. This follows from the same argument as above (since $\mathrm {L}^{\beta '}_{\mathfrak {Q}, h'}$ is injective by our assumption).

Remark 4.3.11. Let $\mathfrak {Q}$ be a deformation of $\mathcal {Q}$ over $R_{G, \mu }$ . In summary, we have the following implications for $\mathfrak {Q}$ :

We will prove in Section 4.4 below that there exists a deformation of $\mathcal {Q}$ over $R_{G, \mu }$ with the property . This in turn implies that all these conditions are equivalent for deformations of $\mathcal {Q}$ over $R_{G, \mu }$ .

4.4 Constructions of universal deformations

Since $\mu $ is 1-bounded, there is an isomorphism

$$\begin{align*}U^{-}_{\mu} \simeq {\mathbb G}^r_a=\mathrm{Spec} {\mathcal O}[t_1, \dotsc, t_r] \end{align*}$$

of group schemes over ${\mathcal O}$ ; see [Reference ItoIto23, Lemma 4.2.6]. We fix such an isomorphism. This induces an identification $ R_{G, \mu } \simeq {\mathcal O}[[t_1, \dotsc , t_r]] $ over ${\mathcal O}$ . We set

$$\begin{align*}\mathfrak{S}^{{\mathrm{univ}}}_{\mathcal O}:={\mathcal O}[[t_1, \dotsc, t_{r+1}]] \quad \text{and} \quad \mathcal{E}:=\pi-t_{r+1}, \end{align*}$$

so that $(\mathfrak {S}^{ {\mathrm {univ}} }_{\mathcal O}, (\mathcal {E}))$ is an ${\mathcal O}_E$ -prism of Breuil–Kisin type over ${\mathcal O}$ . We obtain the following isomorphism defined by $t_{r+1} \mapsto \pi $ :

$$\begin{align*}\mathfrak{S}^{{\mathrm{univ}}}_{\mathcal O}/\mathcal{E} \simeq R_{G, \mu}. \end{align*}$$

For a G- $\mu $ -display $\mathcal {Q}$ over $({\mathcal O}, (\pi ))$ , we will construct a deformation of $\mathcal {Q}$ over $(\mathfrak {S}^{ {\mathrm {univ}} }_{\mathcal O}, (\mathcal {E}))$ such that the corresponding deformation over $R_{G, \mu }$ has the property .

We need a lemma which allows us to reduce the problem to the case where $\mathcal {Q}$ is banal. Let $\widetilde {k}$ be a finite Galois extension of k, and we set $\widetilde {{\mathcal O}}:=W(\widetilde {k}) \otimes _{W({\mathbb F}_q)} {\mathcal O}_E$ . Let $\widetilde {\mu } \colon {\mathbb G}_m \to G_{\widetilde {{\mathcal O}}}$ be the base change of $\mu $ . We define $R_{G, \widetilde {\mu }} \in \mathcal {C}_{\widetilde {{\mathcal O}}}$ in the same way as $R_{G, \mu }$ ; see Definition 3.4.3. We have $ R_{G, \mu }\otimes _{\mathcal O} \widetilde {{\mathcal O}} \overset {\sim }{\to } R_{G, \widetilde {\mu }}. $

Lemma 4.4.1. Let $\mathcal {Q}$ be a G- $\mu $ -display over $({\mathcal O}, (\pi ))$ . Assume that the base change $\widetilde {\mathcal {Q}}$ of $\mathcal {Q}$ to $(\widetilde {{\mathcal O}}, (\pi ))$ is banal. If there exists a deformation $\widetilde {\mathfrak {Q}}$ of $\widetilde {\mathcal {Q}}$ over $R_{G, \widetilde {\mu }}$ having the property , then the same holds for $\mathcal {Q}$ (i.e., there exists a deformation of $\mathcal {Q}$ over $R_{G, \mu }$ having the property ).

Proof. The proof is divided into four parts.

Step 1. Let $ {\mathrm {Gal}} (\widetilde {k}/k)$ be the Galois group of $\widetilde {k}$ over k. For an ${\mathcal O}$ -algebra A, each $s \in {\mathrm {Gal}} (\widetilde {k}/k)$ induces an automorphism of $A \otimes _{{\mathcal O}} \widetilde {{\mathcal O}}$ over ${\mathcal O}$ , which we denote by the same symbol s. In particular, since $R_{G, \widetilde {\mu }}=R_{G, \mu }\otimes _{\mathcal O} \widetilde {{\mathcal O}}$ , we have an automorphism $ s \colon R_{G, \widetilde {\mu }} \overset {\sim }{\to } R_{G, \widetilde {\mu }} $ over ${\mathcal O}$ . Since $s^*(\widetilde {\mathcal {Q}})=\widetilde {\mathcal {Q}}$ , the base change $s^*(\widetilde {\mathfrak {Q}})$ is a deformation of $\widetilde {\mathcal {Q}}$ over $R_{G, \widetilde {\mu }}$ . By Remark 4.3.11, we see that $\widetilde {\mathfrak {Q}}$ , and hence, $s^*(\widetilde {\mathfrak {Q}})$ is a universal deformation of $\widetilde {\mathcal {Q}}$ . Thus, there exists a unique isomorphism $ h_s \colon R_{G, \widetilde {\mu }} \overset {\sim }{\to } R_{G, \widetilde {\mu }} $ over $\widetilde {{\mathcal O}}$ such that $ h^*_s(\widetilde {\mathfrak {Q}}) \simeq s^*(\widetilde {\mathfrak {Q}}) $ as deformations of $\widetilde {\mathcal {Q}}$ . We have

(4.1) $$ \begin{align} h_{ss'}=s(h_{s'}) \circ h_s \end{align} $$

for all $s, s' \in {\mathrm {Gal}} (\widetilde {k}/k)$ . Here, $s(h_{s'}) \colon R_{G, \widetilde {\mu }} \to R_{G, \widetilde {\mu }}$ is the base change of $h_{s'}$ along s.

Step 2. Since

$$\begin{align*}R_{G, \mu}/(\mathfrak{m}^2_{G, \mu}+\pi R_{G, \mu}) = k[t_1, \dotsc, t_r]/(t_1, \dotsc, t_r)^2, \end{align*}$$

the automorphisms $h_s$ induce automorphisms

$$\begin{align*}h^{\prime}_s \colon \widetilde{k}[t_1, \dotsc, t_r]/(t_1, \dotsc, t_r)^2 \overset{\sim}{\to} \widetilde{k}[t_1, \dotsc, t_r]/(t_1, \dotsc, t_r)^2 \end{align*}$$

over $\widetilde {k}$ satisfying the same relation as (4.1). The set of automorphisms of $\widetilde {k}[t_1, \dotsc , t_r]/(t_1, \dotsc , t_r)^2$ over $\widetilde {k}$ can be identified with $ {\mathrm {GL}} _r(\widetilde {k})$ . Since the Galois cohomology $H^1( {\mathrm {Gal}} (\widetilde {k}/k), {\mathrm {GL}} _r(\widetilde {k}))$ has only one element, there exists an automorphism $g'$ of $\widetilde {k}[t_1, \dotsc , t_r]/(t_1, \dotsc , t_r)^2$ over $\widetilde {k}$ such that $h^{\prime }_s=s(g^{\prime -1}) \circ g'$ for any $s \in {\mathrm {Gal}} (\widetilde {k}/k)$ . Let g be an automorphism of $R_{G, \widetilde {\mu }}$ over $\widetilde {{\mathcal O}}$ lifting $g'$ . By replacing $\widetilde {\mathfrak {Q}}$ by $g^*(\widetilde {\mathfrak {Q}})$ , we may assume that the automorphisms $h_s$ induce the identity on $\widetilde {k}[t_1, \dotsc , t_r]/(t_1, \dotsc , t_r)^2$ .

Step 3. The ${\mathcal O}_E$ -prism $ (\widetilde {{\mathcal O}}[[t_1, \dotsc , t_r]]/(t_1, \dotsc , t_r)^2, (\pi )) $ is naturally an object of the category . Since $h_s$ is a lift of the identity of $\widetilde {k}[t_1, \dotsc , t_r]/(t_1, \dotsc , t_r)^2$ , we have

$$\begin{align*}(h^*_s(\widetilde{\mathfrak{Q}}))_{(\widetilde{{\mathcal O}}[[t_1, \dotsc, t_r]]/(t_1, \dotsc, t_r)^2, (\pi))}=\widetilde{\mathfrak{Q}}_{(\widetilde{{\mathcal O}}[[t_1, \dotsc, t_r]]/(t_1, \dotsc, t_r)^2, (\pi))}; \end{align*}$$

see also Remark 2.6.2. However, we have

$$\begin{align*}(s^*(\widetilde{\mathfrak{Q}}))_{(\widetilde{{\mathcal O}}[[t_1, \dotsc, t_r]]/(t_1, \dotsc, t_r)^2, (\pi))}=s^*(\widetilde{\mathfrak{Q}}_{(\widetilde{{\mathcal O}}[[t_1, \dotsc, t_r]]/(t_1, \dotsc, t_r)^2, (\pi))}). \end{align*}$$

Thus, the isomorphism $h^*_s(\widetilde {\mathfrak {Q}}) \simeq s^*(\widetilde {\mathfrak {Q}})$ induces an isomorphism

$$\begin{align*}\widetilde{\mathfrak{Q}}_{(\widetilde{{\mathcal O}}[[t_1, \dotsc, t_r]]/(t_1, \dotsc, t_r)^2, (\pi))} \simeq s^*(\widetilde{\mathfrak{Q}}_{(\widetilde{{\mathcal O}}[[t_1, \dotsc, t_r]]/(t_1, \dotsc, t_r)^2, (\pi))}) \end{align*}$$

of deformations of $\widetilde {\mathcal {Q}}$ . Then, by Galois descent (see Corollary 3.2.9), we see that

$$\begin{align*}\widetilde{\mathfrak{Q}}_{(\widetilde{{\mathcal O}}[[t_1, \dotsc, t_r]]/(t_1, \dotsc, t_r)^2, (\pi))} \end{align*}$$

arises from a deformation $\mathscr {Q}'$ of $\mathcal {Q}$ over $({\mathcal O}[[t_1, \dotsc , t_r]]/(t_1, \dotsc , t_r)^2, (\pi ))$ .

The homomorphism $\mathfrak {S}^{ {\mathrm {univ}} }_{\mathcal O} \to {\mathcal O}[[t_1, \dotsc , t_r]]/(t_1, \dotsc , t_r)^2$ , defined by $t_i \mapsto t_i$ ( $1 \leq i \leq r$ ) and $t_{r+1} \mapsto 0$ , induces a morphism

$$\begin{align*}(\mathfrak{S}^{{\mathrm{univ}}}_{\mathcal O}, (\mathcal{E})) \to ({\mathcal O}[[t_1, \dotsc, t_r]]/(t_1, \dotsc, t_r)^2, (\pi)) \end{align*}$$

in . By Proposition 3.1.5 and Proposition 3.4.6, there exists a deformation $\mathscr {Q}$ of $\mathscr {Q}'$ over $(\mathfrak {S}^{ {\mathrm {univ}} }_{\mathcal O}, (\mathcal {E}))$ .

Step 4. The deformation $\mathfrak {Q}$ of $\mathcal {Q}$ over $R_{G, \mu }$ corresponding to $\mathscr {Q}$ has the property . Indeed, let $\iota \colon R_{G, \mu } \to R_{G, \widetilde {\mu }}$ be the inclusion. It is enough to show that $\iota ^*\mathfrak {Q}$ has the property . (To check that the condition in Definition 4.3.7 is satisfied for a perfectoid pair $(S, a^\flat )$ over ${\mathcal O}$ , it suffices to check it for the induced perfectoid pair $(S \otimes _{{\mathcal O}} \widetilde {{\mathcal O}}, a^\flat )$ ; see Remark 4.3.8.) Let $h \colon R_{G, \widetilde {\mu }} \to R_{G, \widetilde {\mu }}$ be the unique local homomorphism over $\widetilde {{\mathcal O}}$ such that $h^*(\widetilde {\mathfrak {Q}}) \simeq \iota ^*\mathfrak {Q}$ as deformations of $\widetilde {\mathcal {Q}}$ . By construction, we have

$$\begin{align*}(\iota^*\mathfrak{Q})_{(\widetilde{{\mathcal O}}[[t_1, \dotsc, t_r]]/(t_1, \dotsc, t_r)^2, (\pi))} \simeq \widetilde{\mathfrak{Q}}_{(\widetilde{{\mathcal O}}[[t_1, \dotsc, t_r]]/(t_1, \dotsc, t_r)^2, (\pi))}. \end{align*}$$

Since $\widetilde {\mathfrak {Q}}$ has the property by Proposition 4.3.10, it follows that h induces the identity on $\widetilde {k}[t_1, \dotsc , t_r]/(t_1, \dotsc , t_r)^2$ . In particular, h is an isomorphism, which in turn implies that $\iota ^*\mathfrak {Q}$ has the property . This concludes the proof.

Theorem 4.4.2. Let $\mathcal {Q}$ be a G- $\mu $ -display over $({\mathcal O}, (\pi ))$ . There exists a deformation $\mathfrak {Q}^{ {\mathrm {univ}} }$ of $\mathcal {Q}$ over $R_{G, \mu }$ having the property . In particular, $\mathfrak {Q}^{ {\mathrm {univ}} }$ is a universal deformation of $\mathcal {Q}$ .

Proof. By Lemma 4.4.1, we may assume that $\mathcal {Q}=\mathcal {Q}_X$ for some $X \in G({\mathcal O})_{(\pi )}$ . Let $ U \in G(R_{G, \mu }) $ be the $R_{G, \mu }$ -valued point corresponding to the composition $ \mathrm {Spec} R_{G, \mu } \to U^{-}_{\mu } \to G_{\mathcal O}. $ We define

$$\begin{align*}U^{{\mathrm{univ}}} \in G(\mathfrak{S}^{{\mathrm{univ}}}_{\mathcal O}) \end{align*}$$

as the image of $U \in G(R_{G, \mu })$ under the homomorphism $G(R_{G, \mu }) \to G(\mathfrak {S}^{ {\mathrm {univ}} }_{\mathcal O})$ induced by $t_i \mapsto t_i$ ( $1 \leq i \leq r$ ). Let $X^{ {\mathrm {univ}} } \in G(\mathfrak {S}^{ {\mathrm {univ}} }_{\mathcal O})_{(\mathcal {E})}$ be the element such that

$$\begin{align*}(X^{{\mathrm{univ}}})_{\mathcal{E}}=(U^{{\mathrm{univ}}})^{-1}X_\pi \in G(\mathfrak{S}^{{\mathrm{univ}}}_{\mathcal O}). \end{align*}$$

(Here, the image of $X_\pi \in G({\mathcal O})$ in $G(\mathfrak {S}^{ {\mathrm {univ}} }_{\mathcal O})$ is denoted by the same symbol. See also (2.1) for $(X^{ {\mathrm {univ}} })_{\mathcal {E}}$ and $X_\pi $ .) Then the G- $\mu $ -display $\mathcal {Q}_{X^{ {\mathrm {univ}} }}$ over $(\mathfrak {S}^{ {\mathrm {univ}} }_{\mathcal O}, (\mathcal {E}))$ is naturally a deformation of $\mathcal {Q}_X$ . Let $ \mathfrak {Q}^{ {\mathrm {univ}} } $ be the deformation of $\mathcal {Q}_X$ over $R_{G, \mu }$ such that

$$\begin{align*}(\mathfrak{Q}^{{\mathrm{univ}}})_{(\mathfrak{S}^{{\mathrm{univ}}}_{\mathcal O}, (\mathcal{E}))} \simeq \mathcal{Q}_{X^{{\mathrm{univ}}}}. \end{align*}$$

We shall prove that $\mathfrak {Q}^{ {\mathrm {univ}} }$ has the property . In other words, with the notation of Definition 4.3.7, we want to show that for any $\widetilde {g} \in {\mathrm {Hom}} (R_{G, \mu }, S/a^{m+1})_{g}$ and any isomorphism $\beta \colon \mathfrak {Q}^{ {\mathrm {univ}} }_g \overset {\sim }{\to } \mathcal {Q}_Y$ , the map

$$\begin{align*}\mathrm{L}^{\beta}_{\mathfrak{Q}^{{\mathrm{univ}}}, \widetilde{g}} \colon \mathfrak{t}_{G, \mu} \otimes_k (a^{m}S/a^{m+1}S) \to \mathfrak{t}_{G, \mu} \otimes_k (a^{m}S/a^{m+1}S) \end{align*}$$

is an $S/a$ -linear isomorphism. By Remark 4.2.3 and Remark 4.3.8, we are reduced to the case where the map $S \to S$ , $x \mapsto x^p$ is surjective. By Lemma 4.2.2, the homomorphism $\widetilde {g}$ then lifts to a primitive map

$$\begin{align*}\widetilde{f} \colon (\mathfrak{S}^{{\mathrm{univ}}}_{\mathcal O}, (\mathcal{E})) \to (W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^{m+1}, I_S). \end{align*}$$

Let $f \colon (\mathfrak {S}^{ {\mathrm {univ}} }_{\mathcal O}, (\mathcal {E})) \to (W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m}, I_S)$ be the map induced by $\widetilde {f}$ . Without loss of generality, we may assume that $Y=f(X^{ {\mathrm {univ}} })$ in $G(W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m})_{I_S}$ and $\beta $ is the following composition:

$$\begin{align*}\mathfrak{Q}^{{\mathrm{univ}}}_{g} \overset{\gamma^{-1}_{f}}{\to} {f}^*(\mathcal{Q}_{X^{{\mathrm{univ}}}}) \simeq \mathcal{Q}_{f(X^{{\mathrm{univ}}})}; \end{align*}$$

see the last paragraph of Remark 3.4.8. In this case, we claim that $\mathrm {L}^{\beta }_{\mathfrak {Q}^{ {\mathrm {univ}} }, \widetilde {g}}$ is the identity map. Let $g' \in {\mathrm {Hom}} (R_{G, \mu }, S/a^{m+1})_{g}$ . By Lemma 4.2.4, there exists a map

$$\begin{align*}f' \colon (\mathfrak{S}^{{\mathrm{univ}}}_{\mathcal O}, (\mathcal{E})) \to (W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^{m+1}, I_S) \end{align*}$$

of ${\mathcal O}_E$ -prisms which induces $g'$ and f. It suffices to show that the period map

$$\begin{align*}{\mathrm{Per}}_{\mathcal{Q}_{\widetilde{f}(X^{{\mathrm{univ}}})}} \colon {\mathrm{Def}}(\mathcal{Q}_{f(X^{{\mathrm{univ}}})})_{(W_{{\mathcal O}_E}(S^\flat)/[a^\flat]^{m+1}, I_S)} \to \mathfrak{t}_{G, \mu} \otimes_k (a^{m}S/a^{m+1}S) \end{align*}$$

sends the deformation $\mathcal {Q}_{f'(X^{ {\mathrm {univ}} })}$ to $\mathrm {Diff}_{\widetilde {g}}(g')$ . Here, as in Remark 3.4.8, we identify

$$\begin{align*}{\mathrm{Lift}}(P(\mathcal{Q}_{f(X^{{\mathrm{univ}}})})_{S/a^m}, (\mathcal{Q}_{\widetilde{f}(X^{{\mathrm{univ}}})})_{S/a^{m+1}}) \end{align*}$$

with $\mathfrak {t}_{G, \mu } \otimes _k (a^{m}S/a^{m+1}S)$ . We set $d:=\widetilde {f}(\mathcal {E})$ . We have $f'(\mathcal {E})=u d$ for a unique $u \in (W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m+1})^\times $ . Since the image of u in $W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m}$ is equal to $1$ , the equality $\phi (\mu (u))=1$ holds in $G(W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m+1})$ . Then we have

$$\begin{align*}\widetilde{f}(X^{{\mathrm{univ}}})_d=\widetilde{f}(U^{{\mathrm{univ}}})^{-1} X_\pi \quad \text{and} \quad f'(X^{{\mathrm{univ}}})_d=f'(U^{{\mathrm{univ}}})^{-1} X_\pi \end{align*}$$

in $G(W_{{\mathcal O}_E}(S^\flat )/[a^\flat ]^{m+1})$ . By the proof of Proposition 3.2.7, the unique isomorphism

$$\begin{align*}(\mathcal{Q}_{f'(X^{{\mathrm{univ}}})})_\phi \overset{\sim}{\to} (\mathcal{Q}_{\widetilde{f}(X^{{\mathrm{univ}}})})_\phi \end{align*}$$

of deformations of the underlying G- $\phi $ -module $(\mathcal {Q}_{f(X^{ {\mathrm {univ}} })})_\phi $ is given by $\widetilde {f}(U^{ {\mathrm {univ}} })^{-1}f'(U^{ {\mathrm {univ}} })$ . Thus, the lift $ {\mathrm {Per}} _{\mathcal {Q}_{\widetilde {f}(X^{ {\mathrm {univ}} })}}(\mathcal {Q}_{f'(X^{ {\mathrm {univ}} })})$ of the Hodge filtration can be identified with the $P_\mu (S/a^{m+1})$ -orbit

$$\begin{align*}\widetilde{g}(U)^{-1}g'(U)P_\mu(S/a^{m+1}) \subset G(S/a^{m+1}), \end{align*}$$

and this corresponds to $\mathrm {Diff}_{\widetilde {g}}(g') \in \mathfrak {t}_{G, \mu } \otimes _k (a^{m}S/a^{m+1}S)$ as desired. This concludes the proof of Theorem 4.4.2.

We can now complete the proofs of Theorem 4.1.7 and Theorem 4.1.8.

Proof of Theorem 4.1.7.

This follows from Theorem 4.4.2 and Remark 4.3.11.

Proof of Theorem 4.1.8.

We have already proved that (1) implies (2), (3), (4) and (5); see Theorem 4.1.7. By Corollary 4.3.3, we have $(2) \Rightarrow (1)$ .

Let $\mathfrak {Q}^{ {\mathrm {univ}} }$ be a universal deformation of $\mathcal {Q}$ over $R_{G, \mu }$ , which has the properties $\mathrm {(Perfd)}$ and $\mathrm {(BK)}$ by Theorem 4.1.7. Let $h \colon R_{G, \mu } \to R$ be the unique local homomorphism over ${\mathcal O}$ such that $h^*(\mathfrak {Q}^{ {\mathrm {univ}} }) \simeq \mathfrak {Q}$ .

We assume that (5) holds. By Remark 4.3.5, the induced homomorphism $\mathfrak {t}_R \to \mathfrak {t}_{G, \mu }$ is surjective. Since $\dim _k \mathfrak {t}_R = \dim _k \mathfrak {t}_{G, \mu }$ , we then have $\mathfrak {t}_R \overset {\sim }{\to } \mathfrak {t}_{G, \mu }$ . It follows that $h \colon R_{G, \mu } \to R$ is surjective. Since $\dim R=\dim R_{G, \mu }$ , we conclude that h is an isomorphism, and hence, $\mathfrak {Q}$ is a universal deformation. Thus, we have $(5) \Rightarrow (1)$ .

It is obvious that (3) implies (4). It remains to show that (4) implies (1). We assume that the evaluation map

$$\begin{align*}{\mathrm{ev}}_{\mathfrak{Q}} \colon {\mathrm{Hom}}(R, {\mathcal O}_C/\pi^{2})_{e} \to {\mathrm{Def}}(\mathcal{Q})_{(W_{{\mathcal O}_E}({\mathcal O}_{C^\flat})/[\pi^\flat]^{2}, I_{{\mathcal O}_C})} \end{align*}$$

is bijective, where the base change of $\mathcal {Q}$ to $(W_{{\mathcal O}_E}({\mathcal O}_{C^\flat })/[\pi ^\flat ], I_{{\mathcal O}_C})$ is denoted by the same symbol. Since $ {\mathrm {Def}} (\mathcal {Q})_{(W_{{\mathcal O}_E}({\mathcal O}_{C^\flat })/[\pi ^\flat ]^{2}, I_{{\mathcal O}_C})}$ is nonempty (by Remark 3.3.6 or Proposition 3.4.6), there exists a homomorphism $g \colon R \to {\mathcal O}_C/\pi ^{2}$ which is a lift of e. In particular, it follows that $\pi $ is not contained in $\mathfrak {m}^2_R \subset R$ . The following diagram commutes:

where the vertical arrows are induced from h. It follows that

$$\begin{align*}\mathfrak{t}_{R} \otimes_k (\pi {\mathcal O}_C/\pi^2 {\mathcal O}_C) \to \mathfrak{t}_{G, \mu} \otimes_k (\pi {\mathcal O}_C/\pi^2 {\mathcal O}_C), \end{align*}$$

and hence, $\mathfrak {t}_R \to \mathfrak {t}_{G, \mu }$ , is an isomorphism. This implies that $h \colon R_{G, \mu } \to R$ is surjective. Since $\pi $ is not contained in $\mathfrak {m}^2_R$ , we see that $R/\pi $ is regular, or equivalently, $\dim R/\pi = \dim _k \mathfrak {t}_R$ . We then obtain

$$\begin{align*}\dim R = \dim_k \mathfrak{t}_R +1 = \dim_k \mathfrak{t}_{G, \mu} +1 = \dim R_{G, \mu}. \end{align*}$$

It follows that $h \colon R_{G, \mu } \to R$ is an isomorphism.

The proof of Theorem 4.1.8 is now complete.

5 Integral local Shimura varieties with hyperspecial level structure

In this section, as an application of our deformation theory, we prove the local representability and the formal smoothness of integral local Shimura varieties with hyperspecial level structure; see Theorem 5.3.5 for the precise statement. Throughout this section, we assume that G is a connected reductive group scheme over $ \mathrm {Spec} {\mathcal O}_E$ . We assume that k is an algebraic closure of ${\mathbb F}_q$ .

In Section 5.1, we recall the definition of G-shtukas (with one leg) over perfectoid spaces over k introduced by Scholze, and we discuss their relation to G- $\mu $ -displays. In Section 5.2, we study the notion of quasi-isogeny for G-shtukas. Integral local Shimura varieties are defined as moduli spaces of G-shtukas together with some quasi-isogeny. In Section 5.3, we prove the main result of this section.

We assume that the reader is familiar with the theory of perfectoid spaces, v-sheaves, and ‘the curve’ $\mathcal {Y}^{[0, \infty )}_S$ . Our basic references are [Reference ScholzeSch22], [Reference Scholze and WeinsteinSW20] and [Reference Fargues and ScholzeFS21].

5.1 G-shtukas

Let $ {\mathrm {Perf}} _k $ be the category of perfectoid spaces over k. We endow $ {\mathrm {Perf}} _k$ with the v-topology introduced in [Reference ScholzeSch22, Definition 8.1]. A sheaf on $ {\mathrm {Perf}} _k$ with respect to the v-topology is called a v-sheaf.

Remark 5.1.1 (The notation in this section).

In this section, we let S stand for a perfectoid space over k (rather than a perfectoid ring). We write

$$\begin{align*}{\mathcal O}_{\breve{E}}:=W(k) \otimes_{W({\mathbb F}_q)} {\mathcal O}_E \quad \text{and} \quad \breve{E}:={\mathcal O}_{\breve{E}}[1/\pi]. \end{align*}$$

Let $ {\mathrm {Spd}} {\mathcal O}_{\breve {E}} $ be the v-sheaf on $ {\mathrm {Perf}} _k$ which sends a perfectoid space S over k to the set of isomorphism classes of untilts $S^\sharp $ of S over ${\mathcal O}_{\breve {E}}$ ; see [Reference Scholze and WeinsteinSW20, Section 18.1]. If $S= \mathrm {Spa}(R, R^+)$ is an affinoid perfectoid space over k, then an untilt $S^\sharp $ is also an affinoid perfectoid space, and we use the following notation:

$$\begin{align*}S^\sharp=\mathrm{Spa}(R^\sharp, R^{\sharp+}). \end{align*}$$

The ring $R^{\sharp +}$ is a perfectoid ring (in the sense of [Reference Bhatt, Morrow and ScholzeBMS18]) with $(R^{\sharp +})^\flat = R^+$ . Let $\varpi ^\flat \in R^+$ be a pseudo-uniformizer (i.e., it is a topologically nilpotent unit in R). Then $\varpi :=\theta ([\varpi ^\flat ]) \in R^{\sharp +}$ is also a pseudo-uniformizer. After replacing $\varpi ^\flat $ by $(\varpi ^\flat )^{1/p^n}$ for a large enough n, we may assume that $\pi \in (\varpi )$ in $R^{\sharp +}$ . The pair $(R^{\sharp +}, \varpi ^\flat )$ is a perfectoid pair in the sense of Definition 2.1.4.

Example 5.1.2 (Product of points).

Let $\{ (C_i, C^{+}_i) \}_{i \in I}$ be a set of analytic affinoid fields over k such that $C_i$ is algebraically closed for any $i\in I$ . We choose a pseudo-uniformizer $\varpi ^\flat _i \in C^{+}_i$ for each $i \in I$ . Let $\varpi ^\flat := (\varpi ^\flat _i)_{i \in I} \in \prod _{i \in I} C^{+}_i$ . We endow $\prod _{i \in I} C^{+}_i$ with the $\varpi ^\flat $ -adic topology. Then

$$\begin{align*}S:=\mathrm{Spa}((\prod_{i \in I} C^{+}_i)[1/\varpi^\flat], \prod_{i \in I} C^{+}_i) \end{align*}$$

is an affinoid perfectoid space over k. Following [Reference GleasonGle22b], we call such a perfectoid space a product of points. Products of points over k form a basis of $ {\mathrm {Perf}} _k$ with respect to the v-topology; see [Reference GleasonGle22b, Example 1.1]. Let $S^\sharp $ be an untilt of S over ${\mathcal O}_{\breve {E}}$ . By [Reference ScholzeSch22, Corollary 3.20], we have the corresponding untilt $ \mathrm {Spa}(C^{\sharp }_i, C^{\sharp +}_i)$ of $ \mathrm {Spa}(C_i, C^{+}_i)$ over $S^\sharp $ . Let $\varpi _i:=\theta ([\varpi ^\flat _i]) \in C^{\sharp +}_i$ and let $\varpi :=(\varpi _i)_{i \in I} \in \prod _{i \in I} C^{\sharp +}_i$ . Then the untilt $S^\sharp $ is isomorphic to $ \mathrm {Spa}((\prod _{i \in I} C^{\sharp +}_i)[1/\varpi ], \prod _{i \in I} C^{\sharp +}_i). $

We shall recall the definition of G-shtukas, following [Reference Scholze and WeinsteinSW20]. In the rest of this subsection, we assume that $S= \mathrm {Spa}(R, R^+)$ is an affinoid perfectoid space over k for simplicity. Let $S^\sharp = \mathrm {Spa}(R^\sharp , R^{\sharp +})$ be an untilt of S over ${\mathcal O}_{\breve {E}}$ .

We consider the following adic space:

$$\begin{align*}\mathcal{Y}^{[0, \infty)}_S:=\mathrm{Spa}(W_{{\mathcal O}_E}(R^+))\backslash V([\varpi^\flat]), \end{align*}$$

where $\varpi ^\flat \in R^+$ is a pseudo-uniformizer and $W_{{\mathcal O}_E}(R^+)$ is equipped with the $(\pi , [\varpi ^\flat ])$ -adic topology. The adic space $\mathcal {Y}^{[0, \infty )}_S$ is called the curve and denoted by $\mathcal {Y}_S$ in [Reference Fargues and ScholzeFS21, Section II.1.1]. The curve $\mathcal {Y}^{[0, \infty )}_S$ is a sousperfectoid adic space over ${\mathcal O}_{\breve {E}}$ in the sense of [Reference Scholze and WeinsteinSW20, Definition 6.3.1, Appendix to Lecture 19]; see the proof of [Reference Fargues and ScholzeFS21, Proposition II.1.1]. The Frobenius of $W_{{\mathcal O}_E}(R^+)$ induces an isomorphism $ {\mathrm {Frob}} \colon \mathcal {Y}^{[0, \infty )}_S \to \mathcal {Y}^{[0, \infty )}_S $ over ${\mathcal O}_{E}$ . The untilt $S^\sharp $ can be regraded as a (closed) Cartier divisor of $\mathcal {Y}^{[0, \infty )}_S$ ; see [Reference Fargues and ScholzeFS21, Proposition II.1.4]. This Cartier divisor is induced by the ideal $I_{R^{\sharp +}} \subset W_{{\mathcal O}_E}(R^+)$ .

Remark 5.1.3. For a sousperfectoid adic space X over ${\mathcal O}_E$ , let $ {\mathrm {Vect}} (X)$ be the category of vector bundles on X. A G-torsor over X is an exact tensor functor $ \mathrm {Rep}_{{\mathcal O}_E}(G) \to {\mathrm {Vect}} (X), $ where $\mathrm {Rep}_{{\mathcal O}_E}(G)$ is the category of algebraic representations of G on free ${\mathcal O}_E$ -modules of finite rank. We refer to [Reference Scholze and WeinsteinSW20, Theorem 19.5.2] and [Reference Fargues and ScholzeFS21, Definition/Proposition III.1.1] for equivalent definitions of G-torsors.

Definition 5.1.4. A G-shtuka over S with one leg at $S^\sharp $ is a G-torsor $\mathscr {P}$ over $\mathcal {Y}^{[0, \infty )}_S$ with an isomorphism

$$\begin{align*}\phi_{\mathscr{P}} \colon ({\mathrm{Frob}}^*\mathscr{P})_{ \vert \mathcal{Y}^{[0, \infty)}_S \backslash S^{\sharp}} \overset{\sim}{\to} \mathscr{P}_{ \vert \mathcal{Y}^{[0, \infty)}_S \backslash S^{\sharp}} \end{align*}$$

of G-torsors over the open subspace $\mathcal {Y}^{[0, \infty )}_S \backslash S^{\sharp } \subset \mathcal {Y}^{[0, \infty )}_S$ which is meromorphic along the Cartier divisor $S^{\sharp } \hookrightarrow \mathcal {Y}^{[0, \infty )}_S$ . The meromorphic condition on $\phi _{\mathscr {P}}$ is defined as in [Reference Scholze and WeinsteinSW20, Definition 5.3.5] (via the Tannakian formalism).

Remark 5.1.5. Let $ {\mathrm {Vect}} (W_{{\mathcal O}_E}(R^+))$ be the category of finite projective $W_{{\mathcal O}_E}(R^+)$ -modules. We have a natural morphism $ \mathcal {Y}^{[0, \infty )}_S \to \mathrm {Spec}(W_{{\mathcal O}_E}(R^+)) $ of locally ringed spaces. This induces a functor

$$\begin{align*}{\mathrm{Vect}}(W_{{\mathcal O}_E}(R^+)) \to {\mathrm{Vect}}(\mathcal{Y}^{[0, \infty)}_S), \end{align*}$$

which in turn induces a functor from the category of G-torsors over $ \mathrm {Spec} W_{{\mathcal O}_E}(R^+)$ to the category of G-torsors over $\mathcal {Y}^{[0, \infty )}_S$ .

In the following, a G-Breuil–Kisin module over $(W_{{\mathcal O}_E}(R^+), I_{R^{\sharp +}})$ in the sense of Definition 2.4.4 is also called a G-Breuil–Kisin module for $R^{\sharp +}$ .

Example 5.1.6. Let $\mathcal {P}_{ {\mathrm {sht}} }$ be the G-torsor over $\mathcal {Y}^{[0, \infty )}_S$ associated with a G-Breuil–Kisin module $\mathcal {P}$ for $R^{\sharp +}$ . The isomorphism $ \phi _{\mathcal {P}_{ {\mathrm {sht}} }} \colon ( {\mathrm {Frob}} ^*\mathcal {P}_{ {\mathrm {sht}} })_{ \vert \mathcal {Y}^{[0, \infty )}_S \backslash S^{\sharp }} \overset {\sim }{\to } (\mathcal {P}_{ {\mathrm {sht}} })_{ \vert \mathcal {Y}^{[0, \infty )}_S \backslash S^{\sharp }} $ induced by the Frobenius $F_{\mathcal {P}}$ is meromorphic along $S^{\sharp } \hookrightarrow \mathcal {Y}^{[0, \infty )}_S$ . With this isomorphism, we view $\mathcal {P}_{ {\mathrm {sht}} }$ as a G-shtuka over S with one leg at $S^\sharp $ .

Following [Reference Scholze and WeinsteinSW20] and [Reference Pappas and RapoportPR24], we introduce some boundedness condition on G-shtukas. First, we fix some notation:

Remark 5.1.7. We assume that S is a geometric point of rank 1; that is, $S= \mathrm {Spa}(C, {\mathcal O}_C)$ for an algebraically closed nonarchimedean field C over k with ring of integers ${\mathcal O}_C$ . Let $S^\sharp = \mathrm {Spa}(C^\sharp , {\mathcal O}_{C^\sharp })$ be an untilt of S over ${\mathcal O}_{\breve {E}}$ . The completed local ring of $\mathcal {Y}^{[0, \infty )}_S$ at the closed point $ \mathrm {Spa}(C^\sharp , {\mathcal O}_{C^\sharp }) \hookrightarrow \mathcal {Y}^{[0, \infty )}_S$ is naturally isomorphic to

$$\begin{align*}B^+_{{\mathrm{dR}}}(C^\sharp):={\varprojlim}_{m} W_{{\mathcal O}_E}({\mathcal O}_C)[1/[\varpi^\flat]]/\xi^m, \end{align*}$$

where $\xi \in I_{{\mathcal O}_{C^\sharp }}$ is a generator. (See also [Reference Scholze and WeinsteinSW20, Example 15.2.10] and the discussion before it.) We recall that $B^+_{ {\mathrm {dR}}}(C^\sharp )$ is a complete discrete valuation ring. The maximal ideal of $B^+_{ {\mathrm {dR}}}(C^\sharp )$ is generated by $\xi $ , and the homomorphism $\theta $ induces $B^+_{ {\mathrm {dR}}}(C^\sharp )/\xi \simeq C^\sharp $ . We set $ B_{ {\mathrm {dR}}}(C^\sharp ):=B^+_{ {\mathrm {dR}}}(C^\sharp )[1/\xi ]. $

Let $\mathscr {P}$ be a G-shtuka over S with one leg at $S^\sharp $ . We write $\mathscr {P}_{\vert B^+_{ {\mathrm {dR}}}(C^\sharp )}$ for the restriction of $\mathscr {P}$ to $B^+_{ {\mathrm {dR}}}(C^\sharp )$ , which we regard as a G-torsor over $ \mathrm {Spec} B^+_{ {\mathrm {dR}}}(C^\sharp )$ . Since $\phi _{\mathscr {P}}$ is meromorphic, it induces an isomorphism

$$\begin{align*}({\mathrm{Frob}}^*\mathscr{P})_{ \vert B^+_{{\mathrm{dR}}}(C^\sharp)} \times_{\mathrm{Spec} B^+_{{\mathrm{dR}}}(C^\sharp)} \mathrm{Spec} B_{{\mathrm{dR}}}(C^\sharp) \overset{\sim}{\to} \mathscr{P}_{ \vert B^+_{{\mathrm{dR}}}(C^\sharp)} \times_{\mathrm{Spec} B^+_{{\mathrm{dR}}}(C^\sharp)} \mathrm{Spec} B_{{\mathrm{dR}}}(C^\sharp) \end{align*}$$

of G-torsors over $ \mathrm {Spec} B_{ {\mathrm {dR}}}(C^\sharp )$ , which is denoted by $(\phi _{\mathscr {P}})_{ \vert B_{ {\mathrm {dR}}}(C^\sharp )}$ . We note that since $C^\sharp $ is algebraically closed, the G-torsor $\mathscr {P}_{\vert B^+_{ {\mathrm {dR}}}(C^\sharp )}$ is trivial.

Let $\mu \colon {\mathbb G}_m \to G_{{\mathcal O}_{\breve {E}}}$ be a minuscule cocharacter.

Definition 5.1.8. Let $\mathscr {P}$ be a G-shtuka over S with one leg at $S^\sharp $ .

  1. (1) We assume that $S = \mathrm {Spa}(C, {\mathcal O}_C)$ is a geometric point of rank 1 with $S^\sharp = \mathrm {Spa}(C^\sharp , {\mathcal O}_{C^\sharp })$ . We say that $\mathscr {P}$ is bounded by $\mu $ if for some (and hence all) trivializations $\mathscr {P}_{ \vert B^+_{ {\mathrm {dR}}}(C^\sharp )} \simeq G_{B^+_{ {\mathrm {dR}}}(C^\sharp )}$ and $( {\mathrm {Frob}} ^*\mathscr {P})_{ \vert B^+_{ {\mathrm {dR}}}(C^\sharp )} \simeq G_{B^+_{ {\mathrm {dR}}}(C^\sharp )}$ , the isomorphism $(\phi _{\mathscr {P}})_{ \vert B_{ {\mathrm {dR}}}(C^\sharp )}$ (see Remark 5.1.7) is given by $g \mapsto Yg$ for some element Y in the double coset

    $$\begin{align*}G(B^+_{{\mathrm{dR}}}(C^\sharp))\mu(\xi)^{-1}G(B^+_{{\mathrm{dR}}}(C^\sharp)) \subset G(B_{{\mathrm{dR}}}(C^\sharp)). \end{align*}$$
  2. (2) In general, we say that $\mathscr {P}$ is bounded by $\mu $ if, for any morphism $ \mathrm {Spa}(C, {\mathcal O}_C) \to S$ with $ \mathrm {Spa}(C, {\mathcal O}_C)$ a geometric point of rank 1, the pull-back of $\mathscr {P}$ to $\mathcal {Y}^{[0, \infty )}_{ \mathrm {Spa}(C, {\mathcal O}_C)}$ is bounded by $\mu $ in the sense of (1).

Notice that we have chosen $\mu (\xi )^{-1}$ rather than $\mu (\xi )$ ; our convention agrees with that of [Reference Pappas and RapoportPR24, Definition 2.4.3].

Example 5.1.9. Let $\mu ^{-1}$ be the inverse of $\mu $ . If $\mathcal {P}$ is a G-Breuil–Kisin module for $R^{\sharp +}$ of type $\mu ^{-1}$ in the sense of Definition 2.4.4, then the associated G-shtuka $\mathcal {P}_{ {\mathrm {sht}} }$ over S with one leg at $S^\sharp $ is bounded by $\mu $ .

The following result can be viewed as a partial converse to Example 5.1.9.

Proposition 5.1.10. Assume that $S= \mathrm {Spa}(R, R^+)$ is a product of points over k. Let $\mathcal {P}$ be a G-Breuil–Kisin module for $R^{\sharp +}$ . Then $\mathcal {P}$ is of type $\mu ^{-1}$ if and only if the associated G-shtuka $\mathcal {P}_{ {\mathrm {sht}} }$ over S with one leg at $S^\sharp $ is bounded by $\mu $ .

Proof. We write $ (R, R^+)=((\prod _{i \in I} C^{+}_i)[1/\varpi ^\flat ], \prod _{i \in I} C^{+}_i). $ Then, as in Example 5.1.2, $(R^{\sharp }, R^{\sharp +})$ is of the form $((\prod _{i \in I} C^{\sharp +}_i)[1/\varpi ], \prod _{i \in I} C^{\sharp +}_i)$ . It suffices to prove the ‘if’ direction. By the proof of [Reference ItoIto23, Proposition 5.6.11], it is enough to show that the base change of $\mathcal {P}$ to $(W_{{\mathcal O}_E}(C^{+}_i), I_{C^{\sharp +}_i})$ is of type $\mu ^{-1}$ for each $i \in I$ . Thus, we may assume that the set I is a singleton. We write $(R, R^+)=(C, C^+)$ and $(R^{\sharp }, R^{\sharp +}) =(C^{\sharp }, C^{\sharp +})$ .

We note that $\mathcal {P}$ is trivial as a G-torsor since $W_{{\mathcal O}_E}(C^{+})$ is strictly henselian. We fix a trivialization $\mathcal {P} \simeq G_{W_{{\mathcal O}_E}(C^{+})}$ , and we set $Y:=F_{\mathcal {P}}(1) \in G(W_{{\mathcal O}_E}(C^{+})[1/\xi ])$ , where $\xi \in I_{C^{\sharp +}}$ is a generator. Since $\mathcal {P}_{ {\mathrm {sht}} }$ is bounded by $\mu $ , the image of Y under the homomorphism $ G(W_{{\mathcal O}_E}(C^{+})[1/\xi ]) \to G(B_{ {\mathrm {dR}}}(C^\sharp )) $ lies in $G(B^+_{ {\mathrm {dR}}}(C^\sharp ))\mu (\xi )^{-1}G(B^+_{ {\mathrm {dR}}}(C^\sharp ))$ . We want to prove that Y is contained in

$$\begin{align*}G(W_{{\mathcal O}_E}(C^{+}))\mu(\xi)^{-1}G(W_{{\mathcal O}_E}(C^{+})) \subset G(W_{{\mathcal O}_E}(C^{+})[1/\xi]). \end{align*}$$

To simplify the notation, we write $A:=W_{{\mathcal O}_E}(C^{+})$ and $B^+:=B^+_{ {\mathrm {dR}}}(C^\sharp )$ . The homomorphism $A/\xi \to B^+/\xi $ can be identified with $C^{\sharp +} \to C^\sharp $ , and hence, it is injective. It follows that $A \to B^+$ is injective, and we have $A=A[1/\xi ] \cap B^+$ . This implies that the natural map $ G(A[1/\xi ])/G(A) \to G(B^+[1/\xi ])/G(B^+) $ is injective. We claim that this map induces the following bijection:

$$\begin{align*}G(A)\mu(\xi)^{-1}G(A)/G(A) \overset{\sim}{\to} G(B^+)\mu(\xi)^{-1}G(B^+)/G(B^+). \end{align*}$$

This claim implies that $Y \in G(A)\mu (\xi )^{-1}G(A)$ , as desired.

We shall prove the claim. The map $ G(A)\mu (\xi )^{-1}G(A)/G(A) \to G(A)/G_\mu (A, (\xi )) $ defined by $g\mu (\xi )^{-1}G(A) \mapsto gG_\mu (A, (\xi ))$ is bijective. By Proposition 2.3.3, we have $G(A)/G_\mu (A, (\xi )) \overset {\sim }{\to } G(C^{\sharp +})/P_\mu (C^{\sharp +})$ . Similarly, we have

$$\begin{align*}G(B^+)\mu(\xi)^{-1}G(B^+)/G(B^+) \overset{\sim}{\to} G(B^+)/G_\mu(B^+, (\xi)) \overset{\sim}{\to} G(C^{\sharp})/P_\mu(C^{\sharp}). \end{align*}$$

(We note that the results of Section 2.3 apply to the pair $(B^+, (\xi ))$ .) It then suffices to prove that the map

$$\begin{align*}G(C^{\sharp+})/P_\mu(C^{\sharp+}) \to G(C^{\sharp})/P_\mu(C^{\sharp}) \end{align*}$$

is bijective. This follows from the valuative criterion and the properness of the scheme $G_{{\mathcal O}_{\breve {E}}}/P_\mu $ over $ \mathrm {Spec} {\mathcal O}_{\breve {E}}$ . (We also use that the field $C^\sharp $ is algebraically closed and that the valuation ring $C^{\sharp +}$ is strictly henselian.)

We define an open subset

$$\begin{align*}\mathcal{Y}^{[0, \infty]}_S:=\mathrm{Spa}(W_{{\mathcal O}_E}(R^+))\backslash V(\pi, [\varpi^\flat]) \subset \mathrm{Spa}(W_{{\mathcal O}_E}(R^+)). \end{align*}$$

We collect some results on $\mathcal {Y}^{[0, \infty ]}_S$ that will be needed in the sequel.

Remark 5.1.11. As explained in [Reference Scholze and WeinsteinSW20, Proposition 13.1.1] and [Reference KedlayaKed20] (see also [Reference Pappas and RapoportPR24, Section 2.1]), we can regard $\mathcal {Y}^{[0, \infty ]}_S$ as a sousperfectoid adic space. More precisely, we consider the following rational open subsets of $ \mathrm {Spa}(W_{{\mathcal O}_E}(R^+))$ :

$$ \begin{align*} U&:= \{ \, x \in \mathrm{Spa}(W_{{\mathcal O}_E}(R^+)) \, \vert \, \vert [\varpi^\flat](x) \vert \leq \vert \pi(x) \vert \neq 0 \, \},\\ V&:= \{ \, x \in \mathrm{Spa}(W_{{\mathcal O}_E}(R^+)) \, \vert \, \vert \pi(x) \vert \leq \vert [\varpi^\flat](x) \vert \neq 0 \, \}. \end{align*} $$

By the proof of [Reference KedlayaKed20, Proposition 3.6], both U and V are sousperfectoid adic spaces. We can glue them along $U\cap V$ and endow $\mathcal {Y}^{[0, \infty ]}_S$ with the structure of a sousperfectoid adic space. Strictly speaking, only the case where ${\mathcal O}_E={\mathbb Z}_p$ is considered in [Reference KedlayaKed20]. However, it is easy to see that the same argument works for general ${\mathcal O}_E$ .

Remark 5.1.12. Let $ j^* \colon {\mathrm {Vect}} (W_{{\mathcal O}_E}(R^+)) \to {\mathrm {Vect}} (\mathcal {Y}^{[0, \infty ]}_S) $ be the functor induced by the natural morphism $ j \colon \mathcal {Y}^{[0, \infty ]}_S \to \mathrm {Spec}(W_{{\mathcal O}_E}(R^+)) $ of locally ringed spaces. We claim that $j^*$ is fully faithful. Indeed, the morphism j factors through the open subscheme

$$\begin{align*}\mathcal{Y}^{[0, \infty], {\mathrm{alg}}}_S:=\mathrm{Spec}(W_{{\mathcal O}_E}(R^+)) \backslash V(\pi, [\varpi^\flat]) \subset \mathrm{Spec}(W_{{\mathcal O}_E}(R^+)). \end{align*}$$

Thus, $j^*$ can be written as the composition

$$\begin{align*}{\mathrm{Vect}}(W_{{\mathcal O}_E}(R^+)) \to {\mathrm{Vect}}(\mathcal{Y}^{[0, \infty], {\mathrm{alg}}}_S) \to {\mathrm{Vect}}(\mathcal{Y}^{[0, \infty]}_S), \end{align*}$$

where $ {\mathrm {Vect}} (\mathcal {Y}^{[0, \infty ], {\mathrm {alg}} }_S)$ is the category of vector bundles on the scheme $\mathcal {Y}^{[0, \infty ], {\mathrm {alg}} }_S$ . Since we have $W_{{\mathcal O}_E}(R^+) = {\mathcal O}_{\mathcal {Y}^{[0, \infty ], {\mathrm {alg}} }_S}(\mathcal {Y}^{[0, \infty ], {\mathrm {alg}} }_S)$ , the first functor is fully faithful. The second functor is an equivalence by [Reference KedlayaKed20, Theorem 3.8] (again, the result holds for general ${\mathcal O}_E$ by the same argument), and hence, the claim follows. The second equivalence is also an exact equivalence of exact categories (cf. [Reference GleasonGle22a, Remark 2.3]).

Let $\xi \in I_{R^{\sharp +}}$ be a generator. For finite projective $W_{{\mathcal O}_E}(R^+)$ -modules M and N, the functor $j^*$ induces a bijection between the set of isomorphisms $M[1/\xi ] \overset {\sim }{\to } N[1/\xi ]$ and the set of isomorphisms $ (j^*M)_{\vert \mathcal {Y}^{[0, \infty ]}_S \backslash S^{\sharp }} \overset {\sim }{\to } (j^*N)_{\vert \mathcal {Y}^{[0, \infty ]}_S \backslash S^{\sharp }} $ which are meromorphic.

The following result is obtained in [Reference GleasonGle21, Section 2.1.2] (in the more general case where G is parahoric).

Proposition 5.1.13. Assume that $S= \mathrm {Spa}(R, R^+)$ is a product of points over k. Then the functor $j^* \colon {\mathrm {Vect}} (W_{{\mathcal O}_E}(R^+)) \to {\mathrm {Vect}} (\mathcal {Y}^{[0, \infty ]}_S)$ induces an equivalence from the category of G-torsors over $ \mathrm {Spec} W_{{\mathcal O}_E}(R^+)$ to that of G-torsors over $\mathcal {Y}^{[0, \infty ]}_S$ .

Proof. We briefly recall the argument for the convenience of the reader. By Remark 5.1.12, it suffices to prove that the restriction functor from the category of G-torsors over $ \mathrm {Spec} W_{{\mathcal O}_E}(R^+)$ to that of G-torsors over $\mathcal {Y}^{[0, \infty ], {\mathrm {alg}} }_S$ is an equivalence.

By the proof of [Reference GleasonGle21, Proposition 2.1.17], we see that the functor $ {\mathrm {Vect}} (W_{{\mathcal O}_E}(R^+)) \to {\mathrm {Vect}} (\mathcal {Y}^{[0, \infty ], {\mathrm {alg}} }_S) $ is an equivalence. (This is proved in [Reference KedlayaKed20, Theorem 2.7] if $S= \mathrm {Spa}(C, C^+)$ for an algebraically closed nonarchimedean field C over k, and the general case can be deduced from this special case.) Then the assertion follows from [Reference AnschützAns22, Proposition 8.5] since we assume that G is reductive here.

The Frobenius of $W_{{\mathcal O}_E}(R^+)$ induces an isomorphism $ {\mathrm {Frob}} \colon \mathcal {Y}^{[0, \infty ]}_S \to \mathcal {Y}^{[0, \infty ]}_S $ over ${\mathcal O}_{E}$ . It will be convenient to make the following definition:

Definition 5.1.14. An extended G-shtuka over S with one leg at $S^{\sharp }$ is a G-torsor $\widetilde {\mathscr {P}}$ over $\mathcal {Y}^{[0, \infty ]}_S$ with an isomorphism

$$\begin{align*}\phi_{\widetilde{\mathscr{P}}} \colon ({\mathrm{Frob}}^*\widetilde{\mathscr{P}})_{ \vert \mathcal{Y}^{[0, \infty]}_S \backslash S^{\sharp}} \overset{\sim}{\to} \widetilde{\mathscr{P}}_{ \vert \mathcal{Y}^{[0, \infty]}_S \backslash S^{\sharp}}, \end{align*}$$

which is meromorphic along $S^{\sharp } \hookrightarrow \mathcal {Y}^{[0, \infty ]}_S$ .

As in Example 5.1.6, to a G-Breuil–Kisin module $\mathcal {P}$ for $R^{\sharp +}$ , we can attach an extended G-shtuka $\widetilde {\mathcal {P}}_{ {\mathrm {sht}} }$ over S with one leg at $S^{\sharp }$ .

Corollary 5.1.15. The functor $\mathcal {P} \mapsto \widetilde {\mathcal {P}}_{ {\mathrm {sht}} }$ from the groupoid of G-Breuil–Kisin modules for $R^{\sharp +}$ to that of extended G-shtukas over S with one leg at $S^{\sharp }$ is fully faithful. If S is a product of points over k, then the functor is an equivalence.

Proof. The first assertion follows from Remark 5.1.12. The second assertion follows from Remark 5.1.12 and Proposition 5.1.13.

Proposition 5.1.16 (cf. [Reference Pappas and RapoportPR24, Proposition 2.4.6]).

Assume that $S= \mathrm {Spa}(C, {\mathcal O}_C)$ is a geometric point of rank $1$ with $S^{\sharp }= \mathrm {Spa}(C^{\sharp }, {\mathcal O}_{C^\sharp })$ . Then the construction $\mathcal {P} \mapsto \mathcal {P}_{ {\mathrm {sht}} }$ induces an equivalence of groupoids:

$$\begin{align*}\left\{ \begin{array}{c} \displaystyle G\text{-Breuil--Kisin modules for } {\mathcal O}_{C^\sharp} \end{array} \right\} \overset{\sim}{\to} \left\{ \begin{array}{c} \displaystyle G\text{-shtukas over } \mathrm{Spa}(C, {\mathcal O}_C) \\ \displaystyle \text{with one leg at } \mathrm{Spa}(C^{\sharp}, {\mathcal O}_{C^\sharp}) \end{array} \right\}. \end{align*}$$

Moreover, this equivalence induces

$$\begin{align*}\left\{ \begin{array}{c} \displaystyle G\text{-}\mu^{-1}\text{-displays} \\ \text{over } (W_{{\mathcal O}_E}({\mathcal O}_C), I_{{\mathcal O}_{C^\sharp}}) \end{array} \right\} \overset{\sim}{\to} \left\{ \begin{array}{c} \displaystyle G\text{-shtukas over } \mathrm{Spa}(C, {\mathcal O}_C) \\ \displaystyle \text{with one leg at } \mathrm{Spa}(C^{\sharp}, {\mathcal O}_{C^\sharp}) \\ \displaystyle \text{which are bounded by } \mu \end{array} \right\}. \end{align*}$$

Proof. Although this result is not used in the rest of this paper, we include a proof for completeness. By Proposition 2.4.5 and Proposition 5.1.10, it suffices to prove the first assertion. By [Reference Pappas and RapoportPR24, Proposition 2.2.7] (which holds for general ${\mathcal O}_E$ by the same argument), the functor is fully faithful. As in [Reference Pappas and RapoportPR24, Proposition 2.4.6], we can extend a G-shtuka $\mathscr {P}$ over S with one leg at $S^\sharp $ to an extended G-shtuka $\widetilde {\mathscr {P}}$ , by using Fargues’ classification [Reference FarguesFar20] of G-torsors over the Fargues–Fontaine curve (see also [Reference AnschützAns19]). By Corollary 5.1.15, it then follows that the functor is essentially surjective.

5.2 Quasi-isogenies

Let $S= \mathrm {Spa}(R, R^+)$ be an affinoid perfectoid space over k, and let $S^\sharp = \mathrm {Spa}(R^\sharp , R^{\sharp +})$ be an untilt of S over ${\mathcal O}_{\breve {E}}$ . We fix a pseudo-uniformizer $\varpi ^\flat \in R^+$ .

For an integer $n \geq 1$ , we define a rational open subset

$$\begin{align*}\mathcal{Y}^{[n, \infty]}_S:= \{ \, x \in \mathrm{Spa}(W_{{\mathcal O}_E}(R^+)) \, \vert \, \vert [\varpi^\flat](x) \vert \leq \vert \pi(x) \vert^{n} \neq 0 \, \} \subset \mathrm{Spa}(W_{{\mathcal O}_E}(R^+)). \end{align*}$$

We set $ \mathcal {Y}^{[n, \infty )}_S:=\mathcal {Y}^{[0, \infty )}_S \cap \mathcal {Y}^{[n, \infty ]}_S. $ For the isomorphism $ {\mathrm {Frob}} \colon \mathcal {Y}^{[0, \infty ]}_S \to \mathcal {Y}^{[0, \infty ]}_S$ , the equality $ {\mathrm {Frob}} (\mathcal {Y}^{[n, \infty ]}_S)=\mathcal {Y}^{[qn, \infty ]}_S$ holds. Similarly, we have $ {\mathrm {Frob}} (\mathcal {Y}^{[n, \infty )}_S)=\mathcal {Y}^{[qn, \infty )}_S$ .

Definition 5.2.1. Let $\mathcal {P}_0$ be a G-Breuil–Kisin module over $({\mathcal O}_{\breve {E}}, (\pi ))$ and $\mathscr {P}$ a G-shtuka over S with one leg at $S^\sharp $ . A quasi-isogeny from $\mathscr {P}$ to $\mathcal {P}_0$ is an element of the set

$$\begin{align*}{\varinjlim}_n\{ \, \text{Frobenius equivariant isomorphisms} \ \mathscr{P}_{\vert \mathcal{Y}^{[n, \infty)}_S} \overset{\sim}{\to} (\mathcal{P}_0)_{\vert \mathcal{Y}^{[n, \infty)}_S} \ \text{over} \ \mathcal{Y}^{[n, \infty)}_S \, \}, \end{align*}$$

where n runs over all integers $n \geq 1$ such that $\mathcal {Y}^{[n, \infty )}_S$ does not intersect with $S^\sharp $ . Here, $(\mathcal {P}_0)_{\vert \mathcal {Y}^{[n, \infty )}_S}$ is the pull-back of $\mathcal {P}_0$ along the morphism $\mathcal {Y}^{[n, \infty )}_S \to \mathrm {Spec} {\mathcal O}_{\breve {E}}$ of locally ringed spaces. Similarly, for an extended G-shtuka $\widetilde {\mathscr {P}}$ over S with one leg at $S^{\sharp }$ , we define a quasi-isogeny from $\widetilde {\mathscr {P}}$ to $\mathcal {P}_0$ as an element of the set

$$\begin{align*}{\varinjlim}_n\{ \, \text{Frobenius equivariant isomorphisms} \ \widetilde{\mathscr{P}}_{\vert \mathcal{Y}^{[n, \infty]}_S} \overset{\sim}{\to} (\mathcal{P}_0)_{\vert \mathcal{Y}^{[n, \infty]}_S} \ \text{over} \ \mathcal{Y}^{[n, \infty]}_S \, \}, \end{align*}$$

where n runs over all integers $n \geq 1$ such that $\mathcal {Y}^{[n, \infty ]}_S$ does not intersect with $S^\sharp $ .

Remark 5.2.2. Let $\mathscr {P}$ be a G-shtuka over S with one leg at $S^\sharp $ . Using a quasi-isogeny from $\mathscr {P}$ to $\mathcal {P}_0$ , we can extend $\mathscr {P}$ to an extended G-shtuka over S with one leg at $S^{\sharp }$ . It follows that the restriction functor from the groupoid of extended G-shtukas $\widetilde {\mathscr {P}}$ over S with one leg at $S^\sharp $ together with a quasi-isogeny from $\widetilde {\mathscr {P}}$ to $\mathcal {P}_0$ to the groupoid of G-shtukas $\mathscr {P}$ over S with one leg at $S^\sharp $ together with a quasi-isogeny from $\mathscr {P}$ to $\mathcal {P}_0$ is an equivalence.

We want to relate quasi-isogenies of (extended) G-shtukas to quasi-isogenies of G-Breuil–Kisin modules, which we define as follows:

Definition 5.2.3. Let $(A, (\pi ))$ be an ${\mathcal O}_E$ -prism. Let $\mathcal {P}$ and $\mathcal {P}'$ be G-Breuil–Kisin modules over $(A, (\pi ))$ . A quasi-isogeny from $\mathcal {P}$ to $\mathcal {P}'$ is a Frobenius equivariant isomorphism

$$\begin{align*}\eta \colon \mathcal{P}[1/\pi] \overset{\sim}{\to} \mathcal{P}'[1/\pi] \end{align*}$$

of G-torsors over $ \mathrm {Spec} A[1/\pi ]$ . If $\eta $ arises from a (unique) isomorphism $\mathcal {P} \overset {\sim }{\to } \mathcal {P}'$ of G-Breuil–Kisin modules over $(A, (\pi ))$ , we say that $\eta $ is effective. In this case, the isomorphism $\mathcal {P} \overset {\sim }{\to } \mathcal {P}'$ is denoted by the same symbol $\eta $ .

We assume that $\pi =0$ in $R^{\sharp +}/\varpi $ , or equivalently, $\pi \in I_{R^{\sharp +}}$ , where $\varpi :=\theta ([\varpi ^\flat ]) \in R^{\sharp +}$ . By [Reference ItoIto23, Lemma 2.3.3], we have $I_{R^{\sharp +}}=(\pi )$ in $W_{{\mathcal O}_E}(R^+)/[\varpi ^\flat ]$ . We use the following notation. For an integer $m \geq 0$ , we set

$$\begin{align*}(A_m, (\pi)):=(W_{{\mathcal O}_E}(R^+)/[\varpi^\flat]^{1/q^m}, (\pi)), \end{align*}$$

where $[\varpi ^\flat ]^{1/q^m}:=[(\varpi ^\flat )^{1/q^m}]$ . We write $A=A_0$ . Let

$$\begin{align*}(A_{{\mathrm{red}}}, (\pi)):=(W_{{\mathcal O}_E}(R^+_{{\mathrm{red}}}), (\pi)), \end{align*}$$

where $ R^+_{ {\mathrm {red}} }:= \varinjlim _{m} R^+/(\varpi ^\flat )^{1/q^m}. $ For a G-Breuil–Kisin module $\mathcal {P}$ over $(A, (\pi ))$ , the base change of $\mathcal {P}$ to $(A_m, (\pi ))$ (resp. $(A_{ {\mathrm {red}} }, (\pi ))$ ) is denoted by $\mathcal {P}_m$ (resp. $\mathcal {P}_{ {\mathrm {red}} }$ ). We use the same notation for quasi-isogenies (or isomorphisms).

Lemma 5.2.4. Let $\mathcal {P}$ and $\mathcal {P}'$ be G-Breuil–Kisin modules over $(A, (\pi ))$ . The map

$$\begin{align*}\{ \, \text{quasi-isogenies from}\ \mathcal{P} \ \text{to} \ \mathcal{P}' \,\} \to \{ \, \text{quasi-isogenies from}\ \mathcal{P}_m \ \text{to} \ \mathcal{P}^{\prime}_m \,\}, \quad \eta \mapsto \eta_m \end{align*}$$

is bijective for every $m \geq 1$ .

Proof. By the Tannakian formalism, we are reduced to proving the following claim (by applying it to internal homs): Let M be a finite projective A-module with an isomorphism $F_M \colon (\phi ^*M)[1/\pi ] \overset {\sim }{\to } M[1/\pi ]$ . Then the natural homomorphism

$$\begin{align*}M[1/\pi]^{F_M={\mathrm{id}}} \to (M \otimes_A A_m[1/\pi])^{F_{M}={\mathrm{id}}} \end{align*}$$

is bijective. This is easy to see since the kernel of $A \to A_m$ is killed by $\phi ^m$ .

Lemma 5.2.5. Let $\mathcal {P}$ and $\mathcal {P}'$ be G-Breuil–Kisin modules over $(A, (\pi ))$ such that $\mathcal {P}$ and $\mathcal {P}'$ are trivial as G-torsors over $ \mathrm {Spec} A$ . Let $\eta \colon \mathcal {P}[1/\pi ] \overset {\sim }{\to } \mathcal {P}'[1/\pi ]$ be a quasi-isogeny. If the base change $\eta _{ {\mathrm {red}} } \colon \mathcal {P}_{ {\mathrm {red}} }[1/\pi ] \overset {\sim }{\to } \mathcal {P}^{\prime }_{ {\mathrm {red}} }[1/\pi ]$ is effective, then there exists an integer $m \geq 1$ such that $ \eta _m \colon \mathcal {P}_{m}[1/\pi ] \overset {\sim }{\to } \mathcal {P}^{\prime }_{m}[1/\pi ] $ is effective.

Proof. We fix trivializations $\mathcal {P} \simeq G_A$ and $\mathcal {P}' \simeq G_A$ . Then $\eta $ is identified with an element $g \in G(A[1/\pi ])$ whose image in $G(A_{ {\mathrm {red}} }[1/\pi ])$ lies in $G(A_{ {\mathrm {red}} })$ . We want to show that the image of g in $G(A_m[1/\pi ])$ belongs to $G(A_m)$ for some $m \geq 1$ . Since G is an affine scheme of finite type over ${\mathcal O}_E$ , it suffices to prove that for an element $x \in A$ whose image $x_{ {\mathrm {red}} } \in A_{ {\mathrm {red}} }$ is divisible by $\pi $ , there exists an integer $m \geq 1$ such that the image $x_{m} \in A_m$ is also divisible by $\pi $ . This is clear since $R^+_{ {\mathrm {red}} }= \varinjlim _{m} R^+/(\varpi ^\flat )^{1/q^m}$ .

We define $B^{[n, \infty ]}_S:={\mathcal O}_{\mathcal {Y}^{[n, \infty ]}_S}(\mathcal {Y}^{[n, \infty ]}_S)$ , which is a $W_{{\mathcal O}_E}(R^+)[1/\pi ]$ -algebra. The universal property of the rational open subset $\mathcal {Y}^{[n, \infty ]}_S$ (by applying it to the affinoid ring $(A[1/\pi ], A[1/\pi ]^{\circ })$ ) induces a natural homomorphism

(5.1) $$ \begin{align} B^{[n, \infty]}_S \to A[1/\pi]=(W_{{\mathcal O}_E}(R^+)/[\varpi^\flat])[1/\pi] \end{align} $$

such that the composition $ W_{{\mathcal O}_E}(R^+)[1/\pi ] \to B^{[n, \infty ]}_S \to (W_{{\mathcal O}_E}(R^+)/[\varpi ^\flat ])[1/\pi ] $ coincides with the quotient map.

Let $\mathcal {P}_0$ be a G-Breuil–Kisin module over $({\mathcal O}_{\breve {E}}, (\pi ))$ and let $\mathcal {P}$ be a G-Breuil–Kisin module over $(W_{{\mathcal O}_E}(R^+), I_{R^{\sharp +}})$ . We assume that $\mathcal {P}$ is trivial as a G-torsor. We denote by $\mathcal {P}_A$ (resp. $(\mathcal {P}_0)_A$ ) the base change of $\mathcal {P}$ (resp. $\mathcal {P}_0$ ) to A. Let $\widetilde {\mathcal {P}}_{ {\mathrm {sht}} }$ be the extended G-shtuka over S with one leg at $S^\sharp $ associated with $\mathcal {P}$ . A quasi-isogeny

$$\begin{align*}\iota \colon (\widetilde{\mathcal{P}}_{{\mathrm{sht}}})_{\vert \mathcal{Y}^{[n, \infty]}_S} \overset{\sim}{\to} (\mathcal{P}_0)_{\vert \mathcal{Y}^{[n, \infty]}_S} \end{align*}$$

from $\widetilde {\mathcal {P}}_{ {\mathrm {sht}} }$ to $\mathcal {P}_0$ induces a quasi-isogeny $ \eta _\iota \colon \mathcal {P}_A[1/\pi ] \overset {\sim }{\to } (\mathcal {P}_0)_A[1/\pi ] $ by base change along (5.1). The following proposition is implicitly proved in [Reference GleasonGle22a].

Proposition 5.2.6 [Reference GleasonGle22a].

The above construction $\iota \mapsto \eta _\iota $ induces a bijection between the set of quasi-isogenies $\iota $ from $\widetilde {\mathcal {P}}_{ {\mathrm {sht}} }$ to $\mathcal {P}_0$ such that $(\eta _\iota )_{ {\mathrm {red}} }$ is effective and the set of quasi-isogenies $\eta $ from $\mathcal {P}_A$ to $(\mathcal {P}_0)_A$ such that $\eta _{ {\mathrm {red}} }$ is effective.

Proof. We fix trivializations $\mathcal {P} \simeq G_{W_{{\mathcal O}_E}(R^+)}$ and $\mathcal {P}_0 \simeq G_{{\mathcal O}_{\breve {E}}}$ . We first prove the injectivity of the map. Let $\iota $ and $\iota '$ be quasi-isogenies from $\widetilde {\mathcal {P}}_{ {\mathrm {sht}} }$ to $\mathcal {P}_0$ defined over $\mathcal {Y}^{[n, \infty ]}_S$ such that $(\eta _\iota )_{ {\mathrm {red}} }$ and $(\eta _{\iota '})_{ {\mathrm {red}} }$ are effective. We assume that $\eta _{\iota }=\eta _{\iota '}$ . We identify $\iota $ and $\iota '$ with elements $g, g' \in G(B^{[n, \infty ]}_S)$ , respectively. Arguing as in [Reference GleasonGle22a, Lemma 2.14], we can find elements $h, h' \in G(W_{{\mathcal O}_E}(R^+))$ such that for some $r \geq n$ and $m \geq 1$ , we have $h=g$ and $h'=g'$ in $G(B^{[r, \infty ]}_S/[\varpi ^\flat ]^{1/q^m})$ . Since $\eta _{\iota }=\eta _{\iota '}$ , we see that $h=h'$ in $G(W_{{\mathcal O}_E}(R^+)/[\varpi ^\flat ]^{1/q^{m}})$ . This implies that $g=g'$ in $G(B^{[r, \infty ]}_S/[\varpi ^\flat ]^{1/q^{m}})$ . It now follows from [Reference GleasonGle22a, Lemma 2.15] (which also holds for general ${\mathcal O}_E$ ) that $g=g'$ in $G(B^{[r, \infty ]}_S)$ , whence $\iota =\iota '$ .

Let $\eta \colon \mathcal {P}_A[1/\pi ] \overset {\sim }{\to } (\mathcal {P}_0)_A[1/\pi ]$ be a quasi-isogeny such that $\eta _{ {\mathrm {red}} }$ is effective. By applying [Reference GleasonGle22a, Lemma 2.15] to the base change of $\eta $ along $A[1/\pi ] \to B^{[n, \infty ]}_S/[\varpi ^\flat ]$ , we can find a quasi-isogeny $\iota $ from $\widetilde {\mathcal {P}}_{ {\mathrm {sht}} }$ to $\mathcal {P}_0$ with $\eta _\iota =\eta $ . This proves that the map is surjective.

5.3 Formal completions of integral local Shimura varieties

Let $\mu \colon {\mathbb G}_m \to G_{{\mathcal O}_{\breve {E}}}$ be a minuscule cocharacter. Let $B(G)$ be the set of $\phi $ -conjugacy classes of $G(\breve {E})$ and let $B(G, \mu ^{-1}) \subset B(G)$ be the subset of neutral acceptable elements for $\mu ^{-1}$ ; see [Reference Pappas and RapoportPR24, Section 2.4.1]. Let $b \in G(\breve {E})$ be an element such that its $\phi $ -conjugacy class belongs to $B(G, \mu ^{-1})$ . Let $\mathcal {E}_b=G_{{\mathcal O}_{\breve {E}}}$ be the G-Breuil–Kisin module over $({\mathcal O}_{\breve {E}}, (\pi ))$ with the Frobenius given by $g \mapsto bg$ .

Definition 5.3.1 [Reference Scholze and WeinsteinSW20, Definition 25.1.1], [Reference Pappas and RapoportPR24, Definition 3.2.1].

The integral moduli space of local shtukas

$$\begin{align*}\mathcal{M}^{{{\mathrm{int}}}}_{G, b, \mu} \end{align*}$$

associated with the triple $(G, b, \mu )$ is the v-sheaf on $ {\mathrm {Perf}} _k$ which sends an affinoid perfectoid space $S= \mathrm {Spa}(R, R^+)$ over k to the set of isomorphism classes of tuples

$$\begin{align*}(S^{\sharp}, \mathscr{P}, \iota) \end{align*}$$

consisting of an untilt $S^{\sharp }= \mathrm {Spa}(R^{\sharp }, R^{\sharp +})$ of S over ${\mathcal O}_{\breve {E}}$ , a G-shtuka $\mathscr {P}$ over S with one leg at $S^\sharp $ which is bounded by $\mu $ , and a quasi-isogeny $\iota $ from $\mathscr {P}$ to $\mathcal {E}_b$ . We also call $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ the integral local Shimura variety (with hyperspecial level structure). To see that $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ is a v-sheaf, we can use [Reference Scholze and WeinsteinSW20, Proposition 19.5.3]. (Although only the case of ${\mathcal O}_E={\mathbb Z}_p$ is considered in [Reference Scholze and WeinsteinSW20, Proposition 19.5.3], the same argument works for general ${\mathcal O}_E$ .)

We have a morphism $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu } \to {\mathrm {Spd}} {\mathcal O}_{\breve {E}}$ defined by $(S^{\sharp }, \mathscr {P}, \iota ) \mapsto S^{\sharp }$ . Namely, $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ is a v-sheaf over $ {\mathrm {Spd}} {\mathcal O}_{\breve {E}}$ . The generic fiber of $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ is (the v-sheaf associated with) the local Shimura variety $\mathcal {M}_{G_E, b, \mu , K}$ over $ \mathrm {Spa} (\breve {E}, {\mathcal O}_{\breve {E}})$ , where $K:=G({\mathcal O}_E)$ .

Remark 5.3.2. Let $S= \mathrm {Spa}(R, R^+)$ be a product of points over k and let $(S^{\sharp }, \mathscr {P}, \iota ) \in \mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }(S)$ . By Remark 5.2.2, we can uniquely extend $\mathscr {P}$ to an extended G-shtuka $\widetilde {\mathscr {P}}$ over S with one leg at $S^\sharp $ and regard $\iota $ as a quasi-isogeny from $\widetilde {\mathscr {P}}$ to $\mathcal {E}_b$ . By Corollary 5.1.15, Proposition 5.1.10 and Proposition 2.4.5, there is a unique (up to isomorphism) G- $\mu ^{-1}$ -display $\mathcal {Q}$ over $(W_{{\mathcal O}_E}(R^+), I_{R^{\sharp +}})$ such that $ \widetilde {(\mathcal {Q}_{\mathrm {BK}})}_{ {\mathrm {sht}} } \simeq \widetilde {\mathscr {P}}. $ We note $\mathcal {Q}$ is banal by Proposition 2.5.3. (One can show that the Hodge filtration $P(\mathcal {Q})_{R^{\sharp +}}$ is trivial as a $P_\mu $ -torsor over $ \mathrm {Spec} R^{\sharp +}$ arguing as in the proof of [Reference ItoIto23, Proposition 5.6.11].) Let $\mathcal {Q}_{ {\mathrm {red}} }$ be the base change of $\mathcal {Q}$ along $(W_{{\mathcal O}_E}(R^+), I_{R^{\sharp +}}) \to (W_{{\mathcal O}_E}(R^+_{ {\mathrm {red}} }), (\pi ))$ . Then $\iota $ induces a quasi-isogeny

$$\begin{align*}(\eta_\iota)_{{\mathrm{red}}} \colon (\mathcal{Q}_{{\mathrm{red}}})_{\mathrm{BK}}[1/\pi] \overset{\sim}{\to} \mathcal{E}_b[1/\pi] \end{align*}$$

as in Proposition 5.2.6. Here, the base change of $\mathcal {E}_b$ to $(W_{{\mathcal O}_E}(R^+_{ {\mathrm {red}} }), (\pi ))$ is denoted by the same symbol.

Let $\mathcal {Q}_0$ be a G- $\mu ^{-1}$ -display over $({\mathcal O}_{\breve {E}}, (\pi ))$ together with a quasi-isogeny

$$\begin{align*}\iota_0 \colon (\mathcal{Q}_0)_{\mathrm{BK}}[1/\pi] \overset{\sim}{\to} \mathcal{E}_b[1/\pi]. \end{align*}$$

We note that $\mathcal {Q}_0$ is banal by Proposition 2.5.3. In particular, $(\mathcal {Q}_0)_{\mathrm {BK}}$ is trivial as a G-torsor over $ \mathrm {Spec} {\mathcal O}_{\breve {E}}$ . We fix a trivialization $(\mathcal {Q}_0)_{\mathrm {BK}} \simeq G_{{\mathcal O}_{\breve {E}}}$ . Then $\iota _0$ can be regarded as an element $g \in G(\breve {E})$ such that $g^{-1}b\phi (g) \in G({\mathcal O}_{\breve {E}})\mu (\pi )^{-1} G({\mathcal O}_{\breve {E}})$ . The image of g in $G(\breve {E})/G({\mathcal O}_{\breve {E}})$ does not depend on the choice of $(\mathcal {Q}_0)_{\mathrm {BK}} \simeq G_{{\mathcal O}_{\breve {E}}}$ . In this way, we obtain a point

$$\begin{align*}x \in X_{G}(b, \mu^{-1}):= \{ \, g \in G(\breve{E})/G({\mathcal O}_{\breve{E}}) \, \vert \, g^{-1}b\phi(g) \in G({\mathcal O}_{\breve{E}})\mu(\pi)^{-1} G({\mathcal O}_{\breve{E}}) \, \} \end{align*}$$

of the affine Deligne–Lusztig variety $X_{G}(b, \mu ^{-1})$ .

Definition 5.3.3 [Reference GleasonGle22b].

The formal completion

$$\begin{align*}\widehat{\mathcal{M}^{{{\mathrm{int}}}}_{G, b, \mu}}_{/ x} \end{align*}$$

of $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ at x is the subsheaf of $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ which sends a product of points $S= \mathrm {Spa}(R, R^+)$ over k to the set of isomorphism classes of tuples $ (S^{\sharp }, \mathscr {P}, \iota ) \in \mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }(S) $ such that, for the associated G- $\mu ^{-1}$ -display $\mathcal {Q}$ over $(W_{{\mathcal O}_E}(R^+), I_{R^{\sharp +}})$ , there exists an isomorphism $\mathcal {Q}_{ {\mathrm {red}} } \overset {\sim }{\to } \mathcal {Q}_{0}$ of G- $\mu ^{-1}$ -displays over $(W_{{\mathcal O}_E}(R^+_{ {\mathrm {red}} }), (\pi ))$ which makes the following diagram commute:

See Remark 5.3.2 for the notation used here. As in Remark 5.3.2, the base change of $(\mathcal {Q}_{0})_{\mathrm {BK}}$ to $(W_{{\mathcal O}_E}(R^+_{ {\mathrm {red}} }), (\pi ))$ is denoted by the same symbol.

Remark 5.3.4. It is easy to see that the subsheaf $\widehat {\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }}_{/ x}$ exists and is uniquely determined by the above condition (since products of points over k form a basis of $ {\mathrm {Perf}} _k$ with respect to the v-topology). In [Reference GleasonGle22b], Gleason gives a more conceptual definition of $\widehat {\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }}_{/ x}$ . With the notation of [Reference GleasonGle22b], $\widehat {\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }}_{/ x}$ is the formal neighborhood of $x \in (\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu })^{\mathrm {red}}$ on the prekimberlite $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ ; see [Reference GleasonGle22b, Definition 4.18]. We note that $(\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu })^{\mathrm {red}}$ can be identified with the affine Deligne–Lusztig variety $X_{G}(b, \mu ^{-1})$ by [Reference GleasonGle22a, Proposition 2.30]. See also [Reference GleasonGle22a, Lemma 2.31] for the fact that $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ is a prekimberlite (in the sense of [Reference GleasonGle22b, Definition 4.15]).

We consider the local ring $R_{G, \mu ^{-1}}$ over ${\mathcal O}_{\breve {E}}$ defined as in Definition 3.4.3. There is an isomorphism $ R_{G, \mu ^{-1}} \simeq {\mathcal O}_{\breve {E}}[[t_1, \dotsc , t_r]] $ over ${\mathcal O}_{\breve {E}}$ . We endow $R_{G, \mu ^{-1}}$ with the $\mathfrak {m}_{G, \mu ^{-1}}$ -adic topology. Let $ {\mathrm {Spd}} R_{G, \mu ^{-1}} $ be the v-sheaf (over $ {\mathrm {Spd}} {\mathcal O}_{\breve {E}}$ ) on $ {\mathrm {Perf}} _k$ which sends a perfectoid space S over k to the set of isomorphism classes of untilts $S^\sharp $ of S over ${\mathcal O}_{\breve {E}}$ equipped with a morphism of adic spaces $S^\sharp \to \mathrm {Spa} R_{G, \mu ^{-1}}$ over ${\mathcal O}_{\breve {E}}$ ; see [Reference Scholze and WeinsteinSW20, Section 18.4].

The main result of this section is the following theorem.

Theorem 5.3.5. There exists an isomorphism

$$\begin{align*}\widehat{\mathcal{M}^{{{\mathrm{int}}}}_{G, b, \mu}}_{/ x} \simeq {\mathrm{Spd}} R_{G, \mu^{-1}} \end{align*}$$

of v-sheaves over $ {\mathrm {Spd}} {\mathcal O}_{\breve {E}}$ .

Proof. Let $S= \mathrm {Spa}(R, R^+)$ be a product of points over k and let $S^{\sharp }= \mathrm {Spa}(R^{\sharp }, R^{\sharp +})$ be an untilt of S over ${\mathcal O}_{\breve {E}}$ . Then the inverse image of $S^{\sharp } \in ( {\mathrm {Spd}} {\mathcal O}_{\breve {E}})(S)$ under the map $( {\mathrm {Spd}} R_{G, \mu ^{-1}})(S) \to ( {\mathrm {Spd}} {\mathcal O}_{\breve {E}})(S)$ can be identified with the set of continuous homomorphisms $R_{G, \mu ^{-1}} \to R^{\sharp +}$ over ${\mathcal O}_{\breve {E}}$ .

Let $\varpi ^{\flat } \in R^+$ be a pseudo-uniformizer such that $\pi \in (\varpi )$ in $R^{\sharp +}$ , where $\varpi :=\theta ([\varpi ^\flat ]) \in R^{\sharp +}$ . The inverse image of $S^{\sharp } \in ( {\mathrm {Spd}} {\mathcal O}_{\breve {E}})(S)$ under the map $\widehat {\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }}_{/ x}(S) \to ( {\mathrm {Spd}} {\mathcal O}_{\breve {E}})(S)$ can be identified with the set of isomorphism classes of pairs $ (\mathcal {Q}, \eta ) $ consisting of a banal G- $\mu ^{-1}$ -display $\mathcal {Q}$ over $(W_{{\mathcal O}_E}(R^+), I_{R^{\sharp +}})$ and an element $\eta $ of the set

where $(A_m, (\pi ))=(W_{{\mathcal O}_E}(R^+)/[\varpi ^\flat ]^{1/q^m}, (\pi ))$ , and $\mathcal {Q}_m$ (resp. $(\mathcal {Q}_0)_m$ ) is the base change of $\mathcal {Q}$ (resp. $\mathcal {Q}_0$ ) to $(A_m, (\pi ))$ . This follows by combining Proposition 2.4.5, Proposition 5.1.10, Corollary 5.1.15, Lemma 5.2.4, Lemma 5.2.5 and Proposition 5.2.6.

By Theorem 4.1.7, there exists a universal deformation $\mathfrak {Q}^{ {\mathrm {univ}} }$ of $\mathcal {Q}_0$ over $R_{G, \mu ^{-1}}$ . Moreover, $\mathfrak {Q}^{ {\mathrm {univ}} }$ has the property $(\mathrm {Perfd})$ . Let $g \colon R_{G, \mu ^{-1}} \to R^{\sharp +}$ be a continuous homomorphism over ${\mathcal O}_{\breve {E}}$ . We regard $(W_{{\mathcal O}_E}(R^+), I_{R^{\sharp +}})$ as an object of via the homomorphism g, and let $\mathfrak {Q}^{ {\mathrm {univ}} }_g$ be the associated G- $\mu ^{-1}$ -display over $(W_{{\mathcal O}_E}(R^+), I_{R^{\sharp +}})$ . For some $m \geq 1$ , we have $g(\mathfrak {m}_{G, \mu ^{-1}}) \subset (\varpi ^{1/q^m})$ in $R^{\sharp +}$ , where $\varpi ^{1/q^m}:=\theta ([\varpi ^\flat ]^{1/q^m}) \in R^{\sharp +}$ . Hence, there is a natural isomorphism $ \eta _g \colon (\mathfrak {Q}^{ {\mathrm {univ}} }_g)_m \overset {\sim }{\to } (\mathcal {Q}_0)_m $ over $(A_m, (\pi ))$ . We define a map

$$\begin{align*}({\mathrm{Spd}} R_{G, \mu^{-1}})(S) \times_{({\mathrm{Spd}} {\mathcal O}_{\breve{E}})(S)} \{ S^\sharp \} \to \widehat{\mathcal{M}^{{{\mathrm{int}}}}_{G, b, \mu}}_{/ x}(S) \times_{({\mathrm{Spd}} {\mathcal O}_{\breve{E}})(S)} \{ S^\sharp \} \end{align*}$$

by $g \mapsto (\mathfrak {Q}^{ {\mathrm {univ}} }_g, \eta _g)$ . Using the property $(\mathrm {Perfd})$ and using Proposition 3.1.5, we see that this map is bijective. Then we obtain a bijection $ ( {\mathrm {Spd}} R_{G, \mu ^{-1}})(S) \overset {\sim }{\to } \widehat {\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }}_{/ x}(S) $ over $( {\mathrm {Spd}} {\mathcal O}_{\breve {E}})(S)$ , which is independent of the choice of $\varpi ^\flat $ and is functorial in S. Since products of points over k form a basis of $ {\mathrm {Perf}} _k$ , we finally get the desired isomorphism $ {\mathrm {Spd}} R_{G, \mu ^{-1}} \overset {\sim }{\to } \widehat {\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }}_{/ x} $ of v-sheaves over $ {\mathrm {Spd}} {\mathcal O}_{\breve {E}}$ .

In the following, we assume that ${\mathcal O}_E={\mathbb Z}_p$ for simplicity.

Remark 5.3.6. Theorem 5.3.5 implies that a conjecture of Pappas–Rapoport [Reference Pappas and RapoportPR24, Conjecture 3.3.5] in the hyperspecial case, which was originally proposed by Gleason [Reference GleasonGle21, Conjecture 1], holds true. This result was already known in the following cases.

  1. (1) In [Reference Pappas and RapoportPR22], Pappas–Rapoport proved Theorem 5.3.5 under the assumption that $p \geq 3$ and the pair $(G_{{\mathbb Q}_p}, \mu )$ is of abelian type in the sense of [Reference Pappas and RapoportPR22, Definition 2.1.3]. In fact, the result is formulated and proved in the more general case where G is parahoric. If $p=2$ , the same result is obtained under the additional assumption that $G_{{\mathbb Q}_p}$ is of type A or C.

  2. (2) In [Reference BartlingBar22], Bartling proved Theorem 5.3.5 under the assumption that the element b satisfies the adjoint nilpotent condition introduced in [Reference Bültel and PappasBP20, Definition 3.4.2] and $p \geq 3$ . He studied the relation between $\widehat {\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }}_{/ x}$ and the deformation space constructed by Bültel–Pappas [Reference Bültel and PappasBP20, 3.5.9].

Remark 5.3.7. Scholze conjectured that there exists a (unique) normal formal scheme $\mathscr {M}_{G, b, \mu }$ which is flat and locally formally of finite type over $ \mathrm {Spf} W(k)$ such that the v-sheaf associated with $\mathscr {M}_{G, b, \mu }$ is isomorphic to $\mathcal {M}^{{ {\mathrm {int}} }}_{G, b, \mu }$ . (This conjecture is stated in the more general case where G is parahoric.) If $(G, b, \mu )$ is of EL type or PEL type, then this is proved in [Reference Scholze and WeinsteinSW20, Lecture 25]. Moreover, in the case (1) of Remark 5.3.6, this is proved in [Reference Pappas and RapoportPR22] using Theorem 5.3.5. Since we have proved Theorem 5.3.5 for any p (including $p=2$ ), we can prove that if $(G_{{\mathbb Q}_p}, \mu )$ is of abelian type, then $\mathscr {M}_{G, b, \mu }$ exists for any p in the same way as in [Reference Pappas and RapoportPR22].

If $\mathscr {M}_{G, b, \mu }$ exists, then it is formally smooth over $ \mathrm {Spf} W(k)$ by Theorem 5.3.5.

6 Comparison with universal deformations of p-divisible groups

In this section, we assume that ${\mathcal O}_E={\mathbb Z}_p$ . Let k be a perfect field of characteristic p. Let $\mathcal {G}$ be a p-divisible group over $ \mathrm {Spec} k$ of height N and of dimension d. Let $\mu \colon {\mathbb G}_m \to {\mathrm {GL}} _N$ be the cocharacter over $W(k)$ defined by

$$\begin{align*}t \mapsto {\mathrm{diag}}{(\underbrace{t, \dotsc, t}_{N-d}, \underbrace{1, \dotsc, 1}_{d})}. \end{align*}$$

Let $\mathcal {Q}$ be the $ {\mathrm {GL}} _N$ - $\mu $ -display over the prism $(W(k), (p))$ associated with the Dieudonné module of $\mathcal {G}$ . For a universal deformation $\mathcal {G}^{ {\mathrm {univ}} }$ of $\mathcal {G}$ , we prove that the prismatic Dieudonné crystal of $\mathcal {G}^{ {\mathrm {univ}} }$ , as introduced in [Reference Anschütz and Le BrasALB23], induces a universal deformation $\mathfrak {Q}^{ {\mathrm {univ}} }$ of $\mathcal {Q}$ ; see Theorem 6.2.1. Then, by relating the properties $\mathrm {(Perfd)}$ and $\mathrm {(BK)}$ of the universal deformation $\mathfrak {Q}^{ {\mathrm {univ}} }$ to the universal property of $\mathcal {G}^{ {\mathrm {univ}} }$ , we establish some classification results of p-divisible groups; see Section 6.3 for details.

6.1 Prismatic Dieudonné crystals of p-divisible groups

In this section and Section 7 below, we need the notion of quasisyntomic rings introduced in [Reference Bhatt, Morrow and ScholzeBMS19, Definition 4.10]. We refer to [Reference Bhatt, Morrow and ScholzeBMS19, Section 4] for basic properties of quasisyntomic rings.

Example 6.1.1.

  1. (1) Let $R \in \mathcal {C}_{W(k)}$ be a complete regular local ring over $W(k)$ with residue field k. Then R is a quasisyntomic ring. Moreover, $R/\mathfrak {m}^m_R$ is a quasisyntomic ring if $\dim R \leq 1$ . See [Reference Anschütz and Le BrasALB23, Example 3.17].

  2. (2) Let $(S, a^\flat )$ be a perfectoid pair (Definition 2.1.4). Then S and $S/a^m$ are quasisyntomic rings. More precisely, S and $S/a^m$ are quasiregular semiperfectoid rings in the sense of [Reference Bhatt, Morrow and ScholzeBMS19, Definition 4.20]; see [Reference Anschütz and Le BrasALB23, Example 3.20].

Let R be a p-adically complete ring. Let $\mathscr {G}$ be a p-divisible group over $ \mathrm {Spec} R$ . Following [Reference Anschütz and Le BrasALB23], we consider the sheaf

on the site , where $\underline {\mathscr {G}}$ is the sheaf defined by $(A, I) \mapsto \mathscr {G}(A/I)$ and is defined by $(A, I) \mapsto A$ . In [Reference Anschütz and Le BrasALB23, Proposition 4.69], it is proved that is a finite projective A-module for any and its formation commutes with base change along any morphism $(A, I) \to (A', I')$ in . By [Reference ItoIto23, Remark 7.3.1], there is a canonical isomorphism

for any , where we view the site as the localization of at $(A, I)$ , and the restriction of $\underline {\mathscr {G}}$ to is denoted by the same symbol. We set

In [Reference Anschütz and Le BrasALB23, Theorem 4.71], it is proved that if R is a quasisyntomic ring, then is a prismatic Dieudonné crystal on in the sense that for any , the A-module with the A-linear homomorphism

induced by the Frobenius is a minuscule Breuil–Kisin module over $(A, I)$ , and its formation commutes with base change along any morphism $(A, I) \to (A', I')$ in . (See also [Reference ItoIto23, Section 7.3].)

Remark 6.1.2. We assume that R is a quasisyntomic ring. By [Reference Anschütz and Le BrasALB23, Proposition 4.4] (see also [Reference Bhatt and ScholzeBS23, Proposition 2.7]), the category of prismatic Dieudonné crystals on is equivalent to that of prismatic Dieudonné crystals over R in the sense of [Reference Anschütz and Le BrasALB23, Definition 4.5], which are defined as sheaves on the quasisyntomic site of R. Let $\mathrm {DM}(R)$ denote the category of prismatic Dieudonné crystals on . We have an identification

where $ \mathrm {BK}_{\mathrm {min}}(A, I) $ is the category of minuscule Breuil–Kisin modules over $(A, I)$ .

Example 6.1.3. Let $\mathfrak {S}:=W(k)[[t_1, \dotsc , t_n]]$ and $ \mathfrak {S}_m:=W(k)[[t_1, \dotsc , t_n]]/(t_1, \dotsc , t_n)^{m} $ . We assume that $n \leq 1$ (so that $\mathfrak {S}_m/p$ is a quasisyntomic ring by Example 6.1.1). Let $\mathscr {G}$ be a p-divisible group over $ \mathrm {Spec} \mathfrak {S}_m/p$ . Then there exists a natural isomorphism

of minuscule Breuil–Kisin modules over $(\mathfrak {S}_m, (p))$ , where $\mathbb {D}(\mathscr {G})(\mathfrak {S}_m \to \mathfrak {S}_m/p)$ is the evaluation on the divided power extension $\mathfrak {S}_m \to \mathfrak {S}_m/p$ of the contravariant Dieudonné crystal $\mathbb {D}(\mathscr {G})$ defined in [Reference Berthelot, Breen and MessingBBM82, Définition 3.3.6]. This is a special case of [Reference Anschütz and Le BrasALB23, Lemma 4.45].

6.2 Universal deformations of p-divisible groups

Let $\mathcal {G}$ be a p-divisible group over $ \mathrm {Spec} k$ . Let N be the height of $\mathcal {G}$ and let d be the dimension of $\mathcal {G}$ .

Let $ {\mathrm {Art}} _{W(k)}$ be the category of artinian local rings R with a local homomorphism $W(k) \to R$ which induces an isomorphism on the residue fields. By [Reference IllusieIll85, Corollaire 4.8] (or [Reference LauLau14, Proposition 3.11]), the functor $ {\mathrm {Art}} _{W(k)} \to \mathrm {Set}$ sending R to the set of isomorphism classes of deformations of $\mathcal {G}$ over $ \mathrm {Spec} R$ is pro-representable by

$$\begin{align*}R^{{\mathrm{univ}}}:=W(k)[[t_1, \dotsc, t_{d(N-d)}]]. \end{align*}$$

Let $\mathcal {G}^{ {\mathrm {univ}} }$ be a universal deformation of $\mathcal {G}$ over $ \mathrm {Spec} R^{ {\mathrm {univ}} }$ .

Let $\mu \colon {\mathbb G}_m \to {\mathrm {GL}} _N$ be the cocharacter over $W(k)$ defined as in the beginning of this section. The minuscule Breuil–Kisin module is of type $\mu $ for every . Via the equivalence in Example 2.4.7, the prismatic Dieudonné crystal induces a prismatic $ {\mathrm {GL}} _N$ - $\mu $ -display $ \mathfrak {Q}(\mathcal {G}^{ {\mathrm {univ}} }) $ over $R^{ {\mathrm {univ}} }$ . We set $\mathcal {Q}:=\mathfrak {Q}(\mathcal {G}^{ {\mathrm {univ}} })_{(W(k), (p))}$ , which corresponds to .

Theorem 6.2.1. The prismatic $ {\mathrm {GL}} _N$ - $\mu $ -display $ \mathfrak {Q}(\mathcal {G}^{ {\mathrm {univ}} }) $ over $R^{ {\mathrm {univ}} }$ is a universal deformation of $\mathcal {Q}$ . Moreover, $ \mathfrak {Q}(\mathcal {G}^{ {\mathrm {univ}} }) $ has the properties $\mathrm {(Perfd)}$ and $\mathrm {(BK)}$ .

Proof. Since $R_{ {\mathrm {GL}} _N, \mu } \simeq R^{ {\mathrm {univ}} }$ , it suffices to prove that $\mathfrak {Q}(\mathcal {G}^{ {\mathrm {univ}} })$ is versal by Theorem 4.1.8. We consider the period map

$$\begin{align*}{\mathrm{Per}}_{s^*\mathcal{Q}} \colon {\mathrm{Def}}(\mathcal{Q})_{(W(k)[[t]]/t^2, (p))} \to {\mathrm{Lift}}(P(\mathcal{Q})_{k}, (s^*\mathcal{Q})_{k[[t]]/t^2}) \end{align*}$$

where $s^*\mathcal {Q}$ is the base change of $\mathcal {Q}$ along the natural map

$$\begin{align*}s \colon (W(k), (p)) \to (W(k)[[t]]/t^2, (p)). \end{align*}$$

The period map $ {\mathrm {Per}} _{s^*\mathcal {Q}}$ is bijective by Theorem 3.3.4. Let $ {\mathrm {Def}} (\mathcal {G})_{k[[t]]/t^2}$ be the set of isomorphism classes of deformations of $\mathcal {G}$ over $ \mathrm { Spec} k[[t]]/t^2$ . By the universal property of $\mathcal {G}^{ {\mathrm {univ}} }$ , we may identify $ {\mathrm {Def}} (\mathcal {G})_{k[[t]]/t^2}$ with $\mathfrak {t}_{R^{ {\mathrm {univ}} }}$ . It thus suffices to show that the following composition is bijective:

(6.1) $$ \begin{align} {\mathrm{Def}}(\mathcal{G})_{k[[t]]/t^2} \to {\mathrm{Def}}(\mathcal{Q})_{(W(k)[[t]]/t^2, (p))} \overset{\sim}{\to} {\mathrm{Lift}}(P(\mathcal{Q})_{k}, (s^*\mathcal{Q})_{k[[t]]/t^2}). \end{align} $$

Let $\mathscr {G} \in {\mathrm {Def}} (\mathcal {G})_{k[[t]]/t^2}$ . By Example 6.1.3, we have

and

It follows from Example 2.5.5 and Proposition 3.2.7 that there exists a unique Frobenius equivariant isomorphism

$$\begin{align*}\psi \colon \mathbb{D}(\mathscr{G})(W(k)[[t]]/t^2 \to k[[t]]/t^2) \overset{\sim}{\to} \mathbb{D}(\mathcal{G})(W(k)) \otimes_{W(k)} W(k)[[t]]/t^2 \end{align*}$$

which is a lift of the identity of $\mathbb {D}(\mathcal {G})(W(k))$ . We shall give a different description of $\psi $ . The kernel of $W(k)[[t]]/t^2 \to k$ has a unique divided power structure which is compatible with the usual divided power structure on $(p)$ and the trivial divided power structure on $(t)$ . Then we have a chain of canonical isomorphisms

$$ \begin{align*} \mathbb{D}(\mathscr{G})(W(k)[[t]]/t^2 \to k[[t]]/t^2) &\simeq \mathbb{D}(\mathscr{G})(W(k)[[t]]/t^2 \to k) \\ & \simeq \mathbb{D}(\mathcal{G})(W(k)[[t]]/t^2 \to k) \\ & \simeq \mathbb{D}(\mathcal{G})(W(k)) \otimes_{W(k)} W(k)[[t]]/t^2, \end{align*} $$

where the first and third isomorphisms are induced by the crystals $\mathbb {D}(\mathscr {G})$ and $\mathbb {D}(\mathcal {G})$ and the second one is induced by base change. The composition, denoted by $\psi '$ , is Frobenius equivariant and is a lift of the identity of $\mathbb {D}(\mathcal {G})(W(k))$ . Then the uniqueness part of Proposition 3.2.7 ensures that $\psi =\psi '$ . In particular, we see that the reduction modulo p of $\psi $ coincides with the composition of canonical isomorphisms

$$\begin{align*}\mathbb{D}(\mathscr{G})(k[[t]]/t^2 \to k) \simeq \mathbb{D}(\mathcal{G})(k[[t]]/t^2 \to k) \simeq \mathbb{D}(\mathcal{G})(k) \otimes_k k[[t]]/t^2, \end{align*}$$

where $\mathbb {D}(\mathcal {G})(k):=\mathbb {D}(\mathcal {G})(k \to k)$ .

Let $P^1 \subset \mathbb {D}(\mathcal {G})(k)$ be the Hodge filtration. We may identify $ {\mathrm {Lift}} (P(\mathcal {Q})_{k}, (s^*\mathcal {Q})_{k[[t]]/t^2})$ with the set of lifts $\mathscr {P} \subset \mathbb {D}(\mathcal {G})(k) \otimes _k k[[t]]/t^2$ of the Hodge filtration $P^1$ . The above argument shows that, under this identification, the composition (6.1) coincides with the usual period map in the Grothendieck–Messing deformation theory for p-divisible groups. In particular, (6.1) is bijective; see [Reference MessingMes72, Chapter V, Theorem 1.6].

The proof of Theorem 6.2.1 is complete.

6.3 Classifications of p-divisible groups

For a ring R, let $\mathrm {BT}(R)$ be the category of p-divisible groups over $ \mathrm {Spec} R$ . Recall that $ \mathrm {BK}_{\mathrm {min}}(A, I) $ is the category of minuscule Breuil–Kisin modules over a bounded prism $(A, I)$ .

Let $R \in \mathcal {C}_{W(k)}$ and let $ (\mathfrak {S}, (\mathcal {E})):=(W(k)[[t_1, \dotsc , t_n]], (\mathcal {E})) $ be a prism of Breuil–Kisin type equipped with an isomorphism $R \simeq \mathfrak {S}/\mathcal {E}$ over $W(k)$ . We set $ \mathfrak {S}_m:=W(k)[[t_1, \dotsc , t_n]]/(t_1, \dotsc , t_n)^{m} $ and $R_m:=R/\mathfrak {m}^m_R$ . We remark that $R_m$ is not a quasisyntomic ring if $m \geq 2$ and $\dim R \geq 2$ . However, we still have the following result.

Lemma 6.3.1. Let $R \in \mathcal {C}_{W(k)}$ and let $\mathscr {G}$ be a p-divisible group over $ \mathrm {Spec} R_m$ for some $m \geq 1$ . Then is a minuscule Breuil–Kisin module over $(\mathfrak {S}_m, (\mathcal {E}))$ .

Proof. There is a p-divisible group $\mathscr {G}'$ over $ \mathrm {Spec} R$ such that $\mathscr {G}' \times _{ \mathrm {Spec} R} \mathrm {Spec} R_m \simeq \mathscr {G}$ . (See [Reference IllusieIll85, Corollaire 4.8] for example.) Then is a minuscule Breuil–Kisin module over $(\mathfrak {S}, (\mathcal {E}))$ since R is quasisyntomic and the base change of along $\mathfrak {S} \to \mathfrak {S}_m$ is isomorphic to as explained in Section 6.1. The assertion follows from this fact.

We can deduce the following result from Theorem 6.2.1.

Theorem 6.3.2. For every integer $m \geq 1$ , the contravariant functor

is an anti-equivalence of categories.

Proof. If $m=1$ , then the assertion follows from the classical Dieudonné theory (see also Example 6.1.3). For a general $m \geq 1$ , it is enough to show that the functor induces an equivalence of groupoids; see, for example, the proof of [Reference LauLau10, Theorem 3.2]. Since the assertion holds for $m=1$ , we are reduced to proving that for any p-divisible group $\mathcal {G}$ over $ \mathrm {Spec} k$ and , the functor induces an equivalence from the groupoid of deformations of $\mathcal {G}$ over $ \mathrm {Spec} R_m$ to that of deformations of M over $(\mathfrak {S}_m, (\mathcal {E}))$ . By the fact that any deformation of $\mathcal {G}$ over $ \mathrm {Spec} R_m$ has no nontrivial automorphisms and by Corollary 3.2.10, it is enough to check that the construction induces a bijection between the sets of isomorphism classes of objects. Let $\mathcal {G}^{ {\mathrm {univ}} }$ be a universal deformation of $\mathcal {G}$ . Since $\mathfrak {Q}(\mathcal {G}^{ {\mathrm {univ}} })$ has the property (BK) by Theorem 6.2.1, the assertion follows.

As a consequence, we obtain an alternative proof of the following result of Anschütz–Le Bras. A detailed comparison will be discussed at the end of this subsection.

Corollary 6.3.3 [Reference Anschütz and Le BrasALB23, Theorem 5.12].

The contravariant functor

is an anti-equivalence of categories.

Proof. The assertion follows from Theorem 6.3.2 since we have

$$\begin{align*}\mathrm{BT}(R) \overset{\sim}{\to} {2-\varprojlim}_{m} \mathrm{BT}(R_m) \quad \text{and} \quad \mathrm{BK}_{\mathrm{min}}(\mathfrak{S}, (\mathcal{E})) \overset{\sim}{\to} {2-\varprojlim}_{m} \mathrm{BK}_{\mathrm{min}}(\mathfrak{S}_m, (\mathcal{E})). \end{align*}$$

Remark 6.3.4. Assume that $\mathfrak {S}=W(k)[[t]]$ and $\mathcal {E}$ is an Eisenstein polynomial. In this case, a classification result as in Corollary 6.3.3 was initiated by Breuil and Kisin. More precisely, Breuil conjectured that there exists an equivalence between the two categories in Corollary 6.3.3, and Kisin proved this conjecture when $p \geq 3$ ; see [Reference KisinKis06] and [Reference KisinKis09].

For general R and $(\mathfrak {S}, (\mathcal {E}))=(W(k)[[t_1, \dotsc , t_n]], (\mathcal {E}))$ as above, an equivalence between the two categories in Theorem 6.3.2 was previously obtained by Lau, using crystalline Dieudonné theory and Dieudonné displays; see [Reference LauLau10] for $p \geq 3$ and [Reference LauLau14, Corollary 5.4, Theorem 6.6] for any p. This result in particular implies that Breuil’s conjecture holds true for any p; see also [Reference KimKim12] and [Reference LiuLiu13].

Let ${\mathcal O}_C$ be a p-adically complete valuation ring of rank $1$ with algebraically closed fraction field C. The ring ${\mathcal O}_C$ is a perfectoid ring. We write ${\mathcal O}_{C^\flat }$ for the tilt of ${\mathcal O}_C$ . Let $\varpi ^\flat \in {\mathcal O}_{C^\flat }$ be a pseudo-uniformizer such that $p=0$ in ${\mathcal O}_C/\varpi $ , where $\varpi :=\theta ([\varpi ^\flat ])$ . Assume that k is the residue field of ${\mathcal O}_C$ . There exists a local homomorphism $s \colon W(k) \to {\mathcal O}_C$ which induces the identity $ {\mathrm {id}} _k \colon k \to k$ ; this can be proved by the same argument as in the proof of [Reference Bhatt, Morrow and ScholzeBMS18, Theorem 13.19]. We fix such a local homomorphism s.

Proposition 6.3.5 [Reference Fargues and FontaineFF18, Théorème 11.1.7].

Let M be a minuscule Breuil–Kisin module over $(W({\mathcal O}_{C^\flat })/[\varpi ^\flat ], (p))$ . Let $M_{ {\mathrm {red}} }$ be the base change of M along the map $(W({\mathcal O}_{C^\flat })/[\varpi ^\flat ], (p)) \to (W(k), (p))$ . Then there exist an integer $n \geq 1$ and an isomorphism

$$\begin{align*}M_{{\mathrm{red}}} \otimes_{W(k)} W({\mathcal O}_{C^\flat})/[\varpi^\flat]^{1/p^{n}} \simeq M \otimes_{W({\mathcal O}_{C^\flat})/[\varpi^\flat]} W({\mathcal O}_{C^\flat})/[\varpi^\flat]^{1/p^{n}} \end{align*}$$

of minuscule Breuil–Kisin modules over $(W({\mathcal O}_{C^\flat })/[\varpi ^\flat ]^{1/p^{n}}, (p))$ which is a lift of the identity of $M_{ {\mathrm {red}} }$ .

Proof. We set $\overline {B}:= (\varinjlim _n W({\mathcal O}_{C^\flat })/[\varpi ^\flat ]^{1/p^{n}})[1/p]$ , which agrees with the ring defined in [Reference Fargues and FontaineFF18, Définition 1.10.14] (for ${\mathcal O}_F={\mathcal O}_{C^\flat }$ and $E={\mathbb Q}_p$ ). By [Reference Fargues and FontaineFF18, Théorème 11.1.7], there exists a finite dimensional $W(k)[1/p]$ -vector space N together with an isomorphism $\phi ^*N \overset {\sim }{\to } N$ and a Frobenius equivariant isomorphism

$$\begin{align*}\eta \colon N \otimes_{W(k)[1/p]} \overline{B} \overset{\sim}{\to} M \otimes_{W({\mathcal O}_{C^\flat})/[\varpi^\flat]} \overline{B}. \end{align*}$$

Let $\eta _{ {\mathrm {red}} } \colon N \overset {\sim }{\to } M_{ {\mathrm {red}} }[1/p]$ be the base change of $\eta $ along $\overline {B} \to W(k)[1/p]$ . By composing the base change of $\eta ^{-1}_{ {\mathrm {red}} }$ along $W(k)[1/p] \to \overline {B}$ with $\eta $ , we then obtain a Frobenius equivariant isomorphism

$$\begin{align*}M_{{\mathrm{red}}} \otimes_{W(k)} \overline{B} \overset{\sim}{\to} M \otimes_{W({\mathcal O}_{C^\flat})/[\varpi^\flat]} \overline{B} \end{align*}$$

which is a lift of the identity of $M_{ {\mathrm {red}} }[1/p]$ . Since $\overline {B}= (\varinjlim _n W({\mathcal O}_{C^\flat })/[\varpi ^\flat ]^{1/p^{n}})[1/p]$ , there is a Frobenius equivariant isomorphism

$$\begin{align*}M_{{\mathrm{red}}} \otimes_{W(k)} (W({\mathcal O}_{C^\flat})/[\varpi^\flat]^{1/p^{n}})[1/p] \overset{\sim}{\to} M \otimes_{W({\mathcal O}_{C^\flat})/[\varpi^\flat]} (W({\mathcal O}_{C^\flat})/[\varpi^\flat]^{1/p^{n}})[1/p] \end{align*}$$

for some integer $n \geq 1$ which is a lift of the identity of $M_{ {\mathrm {red}} }[1/p]$ . Now the assertion follows from Lemma 5.2.5.

Theorem 6.3.6 [Reference LauLau18, Theorem 5.7].

The contravariant functor

is an anti-equivalence of categories.

Proof. This is a special case of [Reference LauLau18, Theorem 5.7]. Indeed, the contravariant functor $\Phi _A$ associated with $A=W({\mathcal O}_{C^\flat })/[\varpi ^\flat ]$ in [Reference LauLau18, (5.2)] can be identified with by [Reference Anschütz and Le BrasALB23, Lemma 4.45] and [Reference Cais and LauCL17, Lemma 2.1.16]. (The proof of [Reference LauLau18, Theorem 5.7] relies on Gabber’s classification of p-divisible groups over perfect rings of characteristic p, which is also proved in [Reference LauLau13, Theorem D]. In fact, Berthelot’s classification [Reference BerthelotBer80, Corollaire 3.4.3] of p-divisible groups over perfect valuation rings of characteristic p is enough for the proof of Theorem 6.3.6.)

By combining Theorem 6.2.1, Proposition 6.3.5 and Theorem 6.3.6, we obtain the following result:

Theorem 6.3.7. For every integer $m \geq 1$ , the contravariant functor

is an anti-equivalence of categories.

Proof. If $m=1$ , then this is Theorem 6.3.6. For a general $m \geq 1$ , arguing as in the proof of Theorem 6.3.2, we are reduced to proving that for any p-divisible group $\mathcal {G}$ over $ \mathrm {Spec} {\mathcal O}_C/\varpi $ and

the construction induces a bijection between the set $ {\mathrm {Def}} (\mathcal {G})_{{\mathcal O}_C/\varpi ^m}$ of isomorphism classes of deformations of $\mathcal {G}$ over $ \mathrm {Spec} {\mathcal O}_C/\varpi ^m$ and the set of isomorphism classes of deformations of M over $(W({\mathcal O}_{C^\flat })/[\varpi ^\flat ]^{m}, I_{{\mathcal O}_C})$ .

By Proposition 6.3.5, Theorem 6.3.6 and the classical Dieudonné theory, we may assume that $\mathcal {G}$ is the base change of a p-divisible group $\mathcal {G}_0$ over $ \mathrm {Spec} k$ along the section $k \to {\mathcal O}_C/\varpi $ after replacing $\varpi ^\flat $ by $(\varpi ^\flat )^{1/p^n}$ for some n. Let $\mathcal {G}^{ {\mathrm {univ}} }_0$ be a universal deformation of $\mathcal {G}_0$ over $ \mathrm {Spec} R$ for some $R \in \mathcal {C}_{W(k)}$ . Let $e \colon R \to {\mathcal O}_C/\varpi $ be the composition $R \to k \to {\mathcal O}_C/\varpi $ . By [Reference LauLau14, Remark 3.12], the following map is bijective:

$$\begin{align*}{\mathrm{Hom}}(R, {\mathcal O}_C/\varpi^m)_{e} \to {\mathrm{Def}}(\mathcal{G})_{{\mathcal O}_C/\varpi^m}, \quad g \mapsto g^*\mathcal{G}^{{\mathrm{univ}}}_0. \end{align*}$$

Then the assertion follows since $\mathfrak {Q}(\mathcal {G}^{ {\mathrm {univ}} }_0)$ has the property (Perfd) by Theorem 6.2.1.

To the best of our knowledge, Theorem 6.3.7 is a new result. As a consequence, we can give an alternative proof of (the first part of) the following result.

Corollary 6.3.8 [Reference BerthelotBer80], [Reference LauLau18], [Reference Scholze and WeinsteinSW20], [Reference Anschütz and Le BrasALB23].

  1. (1) The contravariant functor

    is an anti-equivalence of categories.
  2. (2) Let S be a perfectoid ring. The contravariant functor

    is an anti-equivalence of categories.

Proof. (1) As in the proof of Corollary 6.3.3, the assertion follows from Theorem 6.3.7.

(2) As explained in [Reference Scholze and WeinsteinSW20, Theorem 17.5.2] and [Reference ItoIto21, Section 5], the assertion follows from (1) (and the classical Dieudonné theory) by using p-complete $ {\mathrm {arc}} $ -descent for finite projective modules over perfectoid rings S and over $W(S^\flat )$ .

We compare our results with previous studies in more detail.

Remark 6.3.9. Let the notation be as in Corollary 6.3.8. An equivalence

$$\begin{align*}\mathrm{BT}({\mathcal O}_C) \overset{\sim}{\to} \mathrm{BK}_{\mathrm{min}}(W({\mathcal O}_{C^\flat}), I_{{\mathcal O}_C}) \quad (\text{resp.\ } \mathrm{BT}(S) \overset{\sim}{\to} \mathrm{BK}_{\mathrm{min}}(W(S^\flat), I_{S})) \end{align*}$$

was originally constructed by Scholze–Weinstein in [Reference Scholze and WeinsteinSW20, Theorem 14.4.1] (resp. [Reference Scholze and WeinsteinSW20, Theorem 17.5.2], see also [Reference ItoIto21, Section 5]). If $p \geq 3$ , then such equivalences were also obtained by Lau [Reference LauLau18, Theorem 9.8] independently. Corollary 6.3.8 itself was proved in [Reference Anschütz and Le BrasALB23, Corollary 4.49] by showing that the dual of the contravariant functor coincides with the equivalence constructed in [Reference Scholze and WeinsteinSW20].

Remark 6.3.10. Assume that C is of characteristic $0$ . We consider the following categories:

  • $\mathcal {C}_1:=\mathrm {BT}({\mathcal O}_C)$ .

  • $\mathcal {C}_2:=\mathrm {BK}_{\mathrm {min}}(W({\mathcal O}_{C^\flat }), I_{{\mathcal O}_C})$ .

  • The category $\mathcal {C}_3$ of free ${\mathbb Z}_p$ -modules T of finite rank together with a C-subspace of $T \otimes _{{\mathbb Z}_p} C$ .

It is a theorem of Fargues that $\mathcal {C}_2$ is equivalent to $\mathcal {C}_3$ ; see [Reference Scholze and WeinsteinSW20, Theorem 14.1.1]. Scholze–Weinstein proved that $\mathcal {C}_1$ is equivalent to $\mathcal {C}_3$ in [Reference Scholze and WeinsteinSW13, Theorem B]. Then, by using Fargues’ theorem, they obtained that $\mathcal {C}_1$ is equivalent to $\mathcal {C}_2$ ; see [Reference Scholze and WeinsteinSW20, Theorem 14.4.1]. Our proof of Corollary 6.3.8 (1) does not use [Reference Scholze and WeinsteinSW13, Theorem B]. Consequently, together with Fargues’ theorem, we obtain an alternative proof of [Reference Scholze and WeinsteinSW13, Theorem B]. If C is of characteristic p, then Corollary 6.3.8 (1) follows from [Reference BerthelotBer80, Corollaire 3.4.3] (and [Reference Anschütz and Le BrasALB23, Lemma 4.45]), and our proof does not give any new information.

Remark 6.3.11. Let S be a quasisyntomic ring. Recall the category $\mathrm {DM}(S)$ of prismatic Dieudonné crystals on from Remark 6.1.2. Let

$$\begin{align*}\mathrm{DM}^{\mathrm{adm}}(S) \subset \mathrm{DM}(S) \end{align*}$$

be the full subcategory of admissible prismatic Dieudonné crystals over S in the sense of [Reference Anschütz and Le BrasALB23, Definition 4.5]. By [Reference Anschütz and Le BrasALB23, Theorem 4.74], we have the following anti-equivalence:

(6.2)

Its proof uses [Reference Scholze and WeinsteinSW20, Theorem 17.5.2] mentioned in Remark 6.3.9.

In [Reference Anschütz and Le BrasALB23, Section 5.2] and [Reference ItoIto23, Proposition 7.1.1], the following equivalences are obtained (without using p-divisible groups):

$$\begin{align*}\mathrm{DM}^{\mathrm{adm}}(R) \overset{\sim}{\to} \mathrm{DM}(R) \overset{\sim}{\to} \mathrm{BK}_{\mathrm{min}}(\mathfrak{S}, (\mathcal{E})), \end{align*}$$

where the second functor is defined by $\mathcal {M} \mapsto \mathcal {M}(\mathfrak {S}, (\mathcal {E}))$ . In [Reference Anschütz and Le BrasALB23, Theorem 5.12], Anschütz–Le Bras proved Corollary 6.3.3 by combining this result and (6.2).

On the other hand, by using the classification results of p-divisible groups obtained in [Reference Anschütz and Le BrasALB23] and in this paper, one can show the following:

Corollary 6.3.12. Let $m \geq 1$ be an integer. We assume that $\dim R =1$ . We set $S:=R_m$ (resp. $S:={\mathcal O}_C/\varpi ^m$ ) and $(A, I):=(\mathfrak {S}_m, (\mathcal {E}))$ (resp. $(A, I):=(W({\mathcal O}_{C^\flat })/[\varpi ^\flat ]^{m}, I_{{\mathcal O}_C})$ ). Then the functor

$$\begin{align*}\mathrm{DM}^{\mathrm{adm}}(S) \to \mathrm{BK}_{\mathrm{min}}(A, I), \quad \mathcal{M} \mapsto \mathcal{M}(A, I) \end{align*}$$

is an equivalence.

Proof. We note that S is a quasisyntomic ring by Example 6.1.1. The assertion then follows from (6.2), Theorem 6.3.2 and Theorem 6.3.7.

It would be interesting to give a direct proof of Corollary 6.3.12, without using p-divisible groups. This will be achieved in [Reference Gardner and MadapusiGM24]. See also Conjecture 7.1.2 below.

7 Consequences on prismatic F-gauges

In this section, we assume that ${\mathcal O}_E={\mathbb Z}_p$ . Let G be a smooth affine group scheme over ${\mathbb Z}_p$ and let $ \mu \colon {\mathbb G}_m \to G_{W(k)} $ be a cocharacter where k is a perfect field of characteristic p. Let R be a quasisyntomic ring over $W(k)$ . In [Reference ItoIto23, Section 8.2], we introduced the groupoid

of prismatic G-F-gauges of type $\mu $ over R, following the theory of prismatic F-gauges introduced by Drinfeld and Bhatt–Lurie (cf. [Reference DrinfeldDri22], [Reference Bhatt and LurieBL22a], [Reference Bhatt and LurieBL22b], [Reference BhattBha22]) and the work of Guo–Li [Reference Guo and LiGL23]. We refer to [Reference ItoIto23, Section 8.2] for details.

Here, we discuss some consequences of our deformation theory on prismatic G-F-gauges of type $\mu $ and make some conjectures. These conjectures should follow from the results in [Reference Gardner and MadapusiGM24]; see also Remark 1.1.13.

7.1 Prismatic G-F-gauges of type $\mu $

Let R be a quasisyntomic ring over $W(k)$ . The relation between prismatic G-F-gauges of type $\mu $ over R and prismatic G- $\mu $ -displays over R can be described as follows:

Proposition 7.1.1. There exists a fully faithful functor

(7.1)

This functor is an equivalence if R is a perfectoid ring over $W(k)$ or a complete regular local ring over $W(k)$ with residue field k. In particular, we have

Proof. See [Reference ItoIto23, Proposition 8.2.11, Corollary 8.2.12].

Let $R \in \mathcal {C}_{W(k)}$ be a complete regular local ring over $W(k)$ with residue field k. Let $ (\mathfrak {S}, (\mathcal {E})):=(W(k)[[t_1, \dotsc , t_n]], (\mathcal {E})) $ be a prism of Breuil–Kisin type with an isomorphism $R \simeq \mathfrak {S}/\mathcal {E}$ over $W(k)$ . By Theorem 2.6.5 and Proposition 7.1.1, we have the following equivalences:

when $\mu $ is 1-bounded. Similarly to what we have seen in Section 4, these equivalences enable us to study the groupoid using . The former is conceptually important, while the latter is useful from a practical viewpoint. The following analogous statement should hold true.

Conjecture 7.1.2. We assume that $\mu $ is 1-bounded.

  1. (1) We assume that $\dim R =1$ . For every integer $m \geq 1$ , the natural functor

    is an equivalence.
  2. (2) Let $(S, a^\flat )$ be a perfectoid pair over $W(k)$ . For every integer $m \geq 1$ , the natural functor

    is an equivalence.

We assume that $G= {\mathrm {GL}} _{N}$ . Let R be a quasisyntomic ring over $W(k)$ . A prismatic $ {\mathrm {GL}} _{N}$ -F-gauge of type $\mu $ over R can be viewed as a prismatic F-gauge in vector bundles over R (with some additional condition) in the sense of Drinfeld and Bhatt–Lurie. By [Reference Guo and LiGL23, Theorem 2.54], the category $\mathrm {DM}^{\mathrm {adm}}(R)$ of admissible prismatic Dieudonné crystals over R (see Remark 6.3.11) is equivalent to that of prismatic F-gauges in vector bundles ‘of weight $[0, 1]$ ’ over R. In particular, we have the following conclusion:

Example 7.1.3. Let $\mu \colon {\mathbb G}_m \to {\mathrm {GL}} _{N}$ be a cocharacter defined as in Example 2.4.7. We define

which is a full subcategory of the category $\mathrm {DM}(R)$ of prismatic Dieudonné crystals on . By Example 2.4.7, we have Let

$$\begin{align*}\mathrm{DM}^{\mathrm{adm}}_\mu(R):=\mathrm{DM}_\mu(R) \cap \mathrm{DM}^{\mathrm{adm}}(R) \subset \mathrm{DM}_\mu(R) \end{align*}$$

be the full subcategory of those objects that are admissible. Then we have

and the functor (7.1) can be identified with the inclusion

$$\begin{align*}\mathrm{DM}^{\mathrm{adm}}_\mu(R)^{\simeq} \hookrightarrow \mathrm{DM}_\mu(R)^{\simeq}. \end{align*}$$

This follows from [Reference ItoIto23, Example 8.2.9] and [Reference Guo and LiGL23, Theorem 2.54].

Remark 7.1.4. Assume that $G= {\mathrm {GL}} _N$ . Then the following assertions hold:

  • Conjecture 7.1.2 (1) is true.

  • Conjecture 7.1.2 (2) is true for a p-adically complete valuation ring $S={\mathcal O}_C$ of rank $1$ over $W(k)$ with algebraically closed fraction field.

This follows from Corollary 6.3.12 and Example 7.1.3.

7.2 Deformations of prismatic G-F-gauges of type $\mu $

We conclude this paper by discussing how our deformation theory should be interpreted in terms of deformations of prismatic G-F-gauges of type $\mu $ . We assume that $\mu $ is 1-bounded.

Let

be a prismatic G-F-gauge of type $\mu $ over k. For a complete noetherian local ring R over $W(k)$ with residue field k which is quasisyntomic, the set of isomorphism classes of deformations of $\mathscr {Q}$ over R (defined in the usual way) is denoted by $ {\mathrm {Def}} (\mathscr {Q})_R. $ Recall that $\mathcal {C}_{W(k)}$ is the category of complete regular local rings over $W(k)$ with residue field k. The following result can be deduced from our deformation theory.

Theorem 7.2.1. Assume that $\mu $ is 1-bounded. Then the functor

$$\begin{align*}{\mathrm{Def}}(\mathscr{Q}) \colon \mathcal{C}_{W(k)} \to \mathrm{Set}, \quad R \mapsto {\mathrm{Def}}(\mathscr{Q})_R \end{align*}$$

is representable by $R_{G, \mu }$ .

Proof. This follows from Theorem 4.1.7 and Proposition 7.1.1.

Let be a deformation of $\mathscr {Q}$ which represents the functor $ {\mathrm {Def}} (\mathscr {Q}). $ We expect that $\mathscr {Q}^{\mathrm {univ}}$ is universal in a larger class of deformations of $\mathscr {Q}$ . For example, we have the following conjecture:

Conjecture 7.2.2.

  1. (1) Let R be a complete noetherian local ring over $W(k)$ with residue field k which is quasisyntomic. Let $ {\mathrm {Hom}} (R_{G, \mu }, R)_{e}$ be the set of local homomorphisms $R_{G, \mu } \to R$ over $W(k)$ . Then the map

    $$\begin{align*}{\mathrm{Hom}}(R_{G, \mu}, R)_{e} \to {\mathrm{Def}}(\mathscr{Q})_R \end{align*}$$
    induced by $\mathscr {Q}^{\mathrm {univ}}$ is bijective.
  2. (2) Let $(S, a^\flat )$ be a perfectoid pair over $W(k)$ . We regard $\mathscr {Q}$ as a prismatic G-F-gauge of type $\mu $ over $S/a$ by base change. Then the map

    $$\begin{align*}{\mathrm{Hom}}(R_{G, \mu}, S/a^m)_{e} \to {\mathrm{Def}}(\mathscr{Q})_{S/a^m} \end{align*}$$
    induced by $\mathscr {Q}^{\mathrm {univ}}$ is bijective for every $m \geq 1$ . Here, $ {\mathrm {Def}} (\mathscr {Q})_{S/a^m}$ is the set of isomorphism classes of deformations of $\mathscr {Q}$ .

Remark 7.2.3. It follows from [Reference Anschütz and Le BrasALB23, Theorem 4.74] (see (6.2)) and the deformation theory for p-divisible groups that Conjecture 7.2.2 holds true for $G= {\mathrm {GL}} _N$ .

Remark 7.2.4.

  1. (1) Let $R \in \mathcal {C}_{W(k)}$ and we assume that $\dim R=1$ . Since the prismatic G- $\mu $ -display $\mathfrak {Q}^{\mathrm {univ}}$ over $R_{G, \mu }$ corresponding to $\mathscr {Q}^{\mathrm {univ}}$ has the property $(\mathrm {BK})$ by Theorem 4.1.7, we see that Conjecture 7.1.2 (1) holds true for $R/\mathfrak {m}^{m}_R$ if and only if Conjecture 7.2.2 (1) holds true for all and their deformations over $R/\mathfrak {m}^{m}_R$ .

  2. (2) Similarly, since $\mathfrak {Q}^{\mathrm {univ}}$ has the property $(\mathrm {Perfd})$ by Theorem 4.1.7, we see that Conjecture 7.1.2 (2) implies Conjecture 7.2.2 (2).

Acknowledgements

The author would like to thank Zachary Gardner, Kentaro Inoue, Tetsushi Ito, Teruhisa Koshikawa, Arthur-César Le Bras, Yuta Takaya and Alex Youcis for helpful discussions and comments. The author also thanks the referee for many helpful comments and for pointing out a mistake in an earlier version of this paper.

Competing interest

The authors have no competing interest to declare.

Financial support

The work of the author was supported by JSPS KAKENHI Grant Numbers 22K20332, 24K16887, and 24H00015.

References

Anschütz, J., ‘Reductive group schemes over the Fargues-Fontaine curve’, Math. Ann. 374(3–4) (2019), 12191260.CrossRefGoogle Scholar
Anschütz, J., ‘Extending torsors on the punctured’, J. Reine Angew. Math. 783 (2022), 227268.CrossRefGoogle Scholar
Anschütz, J. and Le Bras, A.-C., ‘Prismatic Dieudonné theory’, Forum Math. Pi 11 (2023), Paper No. e2, 92.CrossRefGoogle Scholar
Bartling, S., ‘ $\mathcal{G}$ - $\mu$ -displays and local shtuka’, Preprint, 2022, arXiv:2206.13194.Google Scholar
Berthelot, P., ‘Théorie de Dieudonné sur un anneau de valuation parfait’, Ann. Sci. École Norm. Sup. (4) 13(2) (1980), 225268.CrossRefGoogle Scholar
Berthelot, P., Breen, L. and Messing, W., Théorie de Dieudonné cristalline. II. (Lecture Notes in Mathematics) vol. 930 (Springer-Verlag, Berlin, 1982).CrossRefGoogle Scholar
Bhatt, B. and Lurie, J., ‘Absolute prismatic cohomology’, Preprint, 2022, arXiv:2201.06120.CrossRefGoogle Scholar
Bhatt, B. and Lurie, J., ‘The prismatization of $p$ -adic formal schemes’, Preprint, 2022, arXiv:2201.06124.Google Scholar
Bhatt, B., Morrow, M. and Scholze, P., ‘Integral $p$ -adic Hodge theory’, Publ. Math. Inst. Hautes Études Sci. 128 (2018), 219397.CrossRefGoogle Scholar
Bhatt, B., Morrow, M. and Scholze, P., ‘Topological Hochschild homology and integral $p$ -adic Hodge theory’, Publ. Math. Inst. Hautes Études Sci. 129 (2019), 199310.CrossRefGoogle Scholar
Bhatt, B. and Scholze, P., ‘Prisms and prismatic cohomology’, Ann. of Math. (2) 196(3) (2022), 11351275.CrossRefGoogle Scholar
Bhatt, B. and Scholze, P., ‘Prismatic $F$ -crystals and crystalline Galois representations’, Camb. J. Math. 11(2) (2023), 507562.CrossRefGoogle Scholar
Bültel, O., ‘PEL moduli spaces without $\mathbb{C}$ -valued points’, Preprint, 2008, arXiv:0808.4091.Google Scholar
Bültel, O. and Pappas, G., ‘ $\left(G,\mu \right)$ -displays and Rapoport-Zink spaces’, J. Inst. Math. Jussieu 19(4) (2020), 12111257.CrossRefGoogle Scholar
Cais, B. and Lau, E., ‘Dieudonné crystals and Wach modules for $p$ -divisible groups’, Proc. Lond. Math. Soc. (3) 114(4) (2017), 733763.CrossRefGoogle Scholar
Česnavičius, K. and Scholze, P., ‘Purity for flat cohomology’, Ann. of Math. (2) 199 (1) (2024), 51180.Google Scholar
Cheng, C., ‘Breuil $\mathcal{O}$ -windows and $\pi$ -divisible $\mathcal{O}$ -modules’, Trans. Amer. Math. Soc. 370(1) (2018), 695726.Google Scholar
Demazure, M. and Grothendieck, A., Schémas en groupes. I: Propriétés générales des schémas en groupes (Lecture Notes in Mathematics) vol. 151 (Springer-Verlag, Berlin-New York, 1970, séminaire de Géométrie Algébrique du Bois Marie 1962/64 (SGA 3)).Google Scholar
Drinfeld, V., ‘Prismatization’, Preprint, 2022, arXiv:2005.04746.Google Scholar
Faltings, G., ‘Integral crystalline cohomology over very ramified valuation rings’, J. Am. Math. Soc. 12(1) (1999), 117144.Google Scholar
Fargues, L., ‘ $G$ -torseurs en théorie de Hodge $p$ -adique’, Compos. Math. 156(10) (2020), 20762110.Google Scholar
Fargues, L. and Fontaine, J.-M., ‘Courbes et fibrés vectoriels en théorie de Hodge $p$ -adique’, Astérisque 406 (2018), xiii+382, with a preface by Pierre Colmez.Google Scholar
Fargues, L. and Scholze, P., ‘Geometrization of the local Langlands correspondence’, 2021, Preprint, arXiv:2102.13459.Google Scholar
Gardner, Z. and Madapusi, K., ‘An algebraicity conjecture of Drinfeld and the moduli of $p$ -divisible groups’, 2024, Preprint, arXiv:2412.10226.Google Scholar
Gleason, I., ‘Specialization maps for Scholze’s category of diamonds’, 2021, $\mathrm{P}$ hD thesis, University of California, Berkeley.Google Scholar
Gleason, I., ‘On the geometric connected components of moduli spaces of p-adic shtukas and local Shimura varieties’, Preprint, 2022, arXiv:2107.03579.Google Scholar
Gleason, I., ‘Specialization maps for Scholze’s category of diamonds’, Preprint, 2022, arXiv:2102.05483.Google Scholar
Guo, H. and Li, S., ‘Frobenius height of prismatic cohomology with coefficients’, Preprint, 2023, arXiv:2309.06663.Google Scholar
Illusie, L., ‘Déformations de groupes de Barsotti-Tate (d’après A. Grothendieck)’, Astérisque 127 (1985), 151198, seminar on arithmetic bundles: the Mordell conjecture (Paris, 1983/84).Google Scholar
Imai, N., Kato, H. and Youcis, A., ‘The prismatic realization functor for Shimura varieties of abelian type’, Preprint, 2023, arXiv:2310.08472.Google Scholar
Ito, K., ‘ $p$ -complete arc-descent for perfect complexes over integral perfectoid rings’, 2021, arXiv:2112.02443, to appear in Mathematical Research Letters.Google Scholar
Ito, K., ‘Prismatic $G$ -displays and descent theory’, 2023, arXiv:2303.15814, to appear in Algebra and Number Theory.Google Scholar
Kedlaya, K. S., ‘Some ring-theoretic properties of, in $p$ -adic hodge theory’, Simons Symp. (Springer, Cham, 2020), 129141.Google Scholar
Kim, W., ‘The classification of $p$ -divisible groups over 2-adic discrete valuation rings’, Math. Res. Lett. 19(1) (2012), 121141.CrossRefGoogle Scholar
Kim, W. and Madapusi Pera, K., ‘2-adic integral canonical models’, Forum Math. Sigma 4 (2016), Paper No. e28, 34.Google Scholar
Kisin, M., ‘Crystalline representations and $F$ -crystals’, in Algebraic Geometry and Number Theory (Progr. Math.) vol. 253 (Birkhäuser Boston, Boston, MA, 2006), 459496.CrossRefGoogle Scholar
Kisin, M., ‘Moduli of finite flat group schemes, and modularity’, Ann. of Math. (2) 170(3) (2009), 10851180.CrossRefGoogle Scholar
Lau, E., ‘Frames and finite group schemes over complete regular local rings’, Doc. Math. 15 (2010), 545569.CrossRefGoogle Scholar
Lau, E., ‘Smoothness of the truncated display functor’, J. Amer. Math. Soc. 26(1) (2013), 129165.CrossRefGoogle Scholar
Lau, E., ‘Relations between Dieudonné displays and crystalline Dieudonné theory’, Algebra Number Theory 8(9) (2014), 22012262.CrossRefGoogle Scholar
Lau, E., ‘Dieudonné theory over semiperfect rings and perfectoid rings’, Compos. Math. 154(9) (2018), 19742004.CrossRefGoogle Scholar
Lau, E., ‘Higher frames and $G$ -displays’, Algebra Number Theory 15(9) (2021), 23152355.CrossRefGoogle Scholar
Liu, T., ‘The correspondence between Barsotti-Tate groups and Kisin modules when $p=2$ ’, J. Théor. Nombres Bordeaux 25(3) (2013), 661676.Google Scholar
Marks, S., ‘Prismatic $F$ -crystals and Lubin-Tate $\left({\phi}_q,\varGamma \right)$ -modules’, Preprint, 2023, arXiv:2303.07620.Google Scholar
Messing, W., The Crystals Associated to Barsotti-Tate Groups: With Applications to Abelian Schemes (Lecture Notes in Mathematics) vol. 264 (Springer-Verlag, Berlin-New York, 1972).CrossRefGoogle Scholar
Pappas, G. and Rapoport, M., ‘On integral local Shimura varieties’, Preprint, 2022, arXiv:22204.02829.Google Scholar
Pappas, G. and Rapoport, M., ‘ $p$ -adic shtukas and the theory of global and local Shimura varieties’, Camb. J. Math. 12(1) (2024), 1164.CrossRefGoogle Scholar
Scholze, P., ‘Etale cohomology of diamonds’, Preprint, 2022, arXiv:1709.07343.Google Scholar
Scholze, P. and Weinstein, J., ‘Moduli of $p$ -divisible groups’, Camb. J. Math. 1(2) (2013), 145237.Google Scholar
Scholze, P. and Weinstein, J., Berkeley Lectures on $p$ -adic Geometry vol. 389 (Annals of Mathematics Studies) (Princeton University Press, Princeton, NJ, 2020).Google Scholar