You are on page 1of 14

EXTENSION OF RELATIVE RIGID HOMOMORPHISMS FROM

THE FORMAL MULTIPLICATIVE GROUP

MARTIN ORR

Abstract. A theorem of Lütkebohmert states that a rigid group homomor-


phism from the formal multiplicative group to a smooth commutative rigid
group G, with relatively compact image, can be extended to a homomorphism
from the full rigid multiplicative group to G. In this paper, we prove a relative
arXiv:2307.14095v1 [math.AG] 26 Jul 2023

version of this theorem, with any rigid space as a base. This theorem is a step
in the uniformisation of proper smooth rigid groups.

1. Introduction
Let K be a non-archimedean field and let R be the valuation ring of K. Through-
out this paper, rigid spaces will always mean rigid spaces over K and formal schemes
will always mean formal schemes over Spf(R). Let Gm (1) denote the formal rigid
multiplicative group over K, that is, the following open subgroup of the rigid
K-group Gm :
Gm (1) = {x : |x| = 1}.
The aim of this paper is to prove a relative version of the following theorem of
Lütkebohmert.
Proposition 1.1. [Lüt95, Prop. 3.1], [Lüt16, Prop. 7.3.1] Let G be a smooth
connected commutative rigid group over K. Let ϕ̄ : Gm (1) → G be a homomorphism
of rigid groups. Suppose that:
(i) ϕ̄ factors through an admissible open subgroup H ⊂ G, which admits a
smooth formal R-model Hfor ;
(ii) the image of ϕ̄ is relatively compact in a quasi-compact open subspace G′
of G.
Then there exists a unique homomorphism of rigid groups ϕ : Gm → G whose
restriction to Gm (1) is equal to ϕ̄.
Our main theorem is as follows.

1
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 2

Proposition 1.2. Let S be a rigid space over K. Let G → S be a smooth


commutative rigid S-group with connected fibres. Let ϕ̄ : Gm (1) × S → G be a
homomorphism of rigid S-groups. Suppose that ϕ̄ factors through an admissible
open S-subgroup H ⊂ G satisfying:

(i) H ∼
= Gm (1)g × S;
(ii) There is a quasi-compact open subspace G′ of G such that H is a relatively
S-compact subspace of G′ .

Then there exists a unique homomorphism of rigid S-groups ϕ : Gm × S → G whose


restriction to Gm (1) × S is equal to ϕ̄.

For the definition of “relatively S-compact”, see section 3.


Proposition 1.2 is more restricted than the naı̈ve relative version of Proposi-
tion 1.1: Proposition 1.2(i) requires H to be a split formal torus, while Proposi-
tion 1.1(i) allows more general H, and Proposition 1.2(ii) requires H to be relatively
compact, while Proposition 1.1(ii) only requires the image of ϕ̄ to be relatively
compact. The naı̈ve relative version of Proposition 1.1 may still be true, but there
are several steps in our proof of Proposition 1.2 where these additional restric-
tions simplify the argument. Further work is required to determine whether these
restrictions can be removed from Proposition 1.2. The proposition as stated is
sufficient for the application to the Zilber–Pink conjecture which motivated this
paper, namely [DO].
The proof of Proposition 1.2 is based on the proof of Proposition 1.1 in [Lüt16],
relying on Lütkebohmert’s approximation theorem for rigid analytic morphisms
(which is already relative). The most substantial difficulty in relativizing this proof
appears to be establishing a relative version of [Lüt16, Thm. 7.2.3] (subgroup
generated by the image of a smooth formal scheme). Under condition (i) of
Proposition 1.2, we avoid needing to relativize this theorem by explicitly writing
down a subgroup U of H isomorphic to a relative ball, which contains the image
of a certain morphism ū (see Step 2 in the proof of Proposition 1.2). Meanwhile,
Proposition 1.2(ii) allows us to skip the first step in the proof of [Lüt16, Prop. 7.3.1]
(reducing to the case where H is relatively compact)
We will also prove the following corollary of Proposition 1.2, which is sometimes
more convenient to apply. In particular, Corollary 1.3 applies whenever G → S is
the analytification of a group scheme G → S of finite type over K, since a covering
of S by affine schemes Si induces an admissible covering of S, and the resulting
rigid spaces G|Si = (G|Si )an have no boundary by [Lüt90, Cor. 5.11].
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 3

Corollary 1.3. Let S be a rigid space over K. Let G → S be a smooth commutative


rigid S-group with connected fibres. Let ϕ̄ : Gm (1)g × S → G be a homomorphism
of rigid S-groups. Suppose that:
(i) ϕ̄ is an isomorphism from Gm (1)g × S to an admissible open S-subgroup
H ⊂ G.
(ii) There exists an admissible covering of S by affinoids Si such that, for each i,
G|Si has no boundary, in the sense of [Lüt90, Def. 5.9].
Then there exists a unique homomorphism of rigid S-groups ϕ : Ggm × S → G whose
restriction to Gm (1)g × S is equal to ϕ̄.
1.A. Notation. If S is an affinoid space, let O(S) denote the ring of holomorphic
functions on S and let
O◦ (S) = {f ∈ O(S) : |f | ≤ 1},
where |·| is the sup norm on O(S).
Let D(r) denote the closed disc of radius r centred at the origin (as a K-rigid
space) and let D = D(1).
Acknowledgements. I am very grateful to Werner Lütkebohmert for patient
answers to my questions about relativizing Proposition 1.1, in particular for sug-
gesting that the argument in [Lüt95] should relativize, for helping me to understand
the approximation theorem, and for suggesting how to prove Lemma 3.2. I am
also grateful to Christopher Daw for helpful discussions.

2. Group laws on relative rigid balls


Let S be an affinoid space. Let U be a rigid S-group whose underlying S-space
is isomorphic to Dg × S, with the identity section S → U being equal to the zero
section S → Dg × S.
Then the group law U ×S U → U = Dg × S can be described by a g-tuple of
holomorphic functions
F ∈ O◦ (U ×S U )g = O◦ (Dg × Dg × S)g = O◦ (S)⟨X1 , . . . , Xg , Y1 , . . . , Yg ⟩g .
Likewise the inverse U → U = Dg × S is described by a g-tuple of holomorphic
functions
i ∈ O◦ (U )g = O◦ (Dg × S)g = O◦ (S)⟨X1 , . . . , Xg ⟩g .
Because U is a group over S, F satisfies the usual conditions to be a g-dimensional
formal group law over O◦ (S) i.e.
(1) F (X, 0) = X and F (0, Y ) = Y (the zero section is the identity section);
(2) F (X, F (Y , Z)) = F (F (X, Y ), Z) (associativity);
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 4

(3) F (X, i(X)) = F (i(X), X) = 0 (inverse).


Let π ∈ R with 0 < ϵ = |π| < 1.
Lemma 2.1. Let λ ∈ Z≥0 . Let x, y ∈ D(ϵλ )g . Then F (x, y) ≡ x + y mod π 2λ .
Proof. From condition (1) for a formal group law, we obtain that
F (X, Y ) = X + Y + higher order terms
where each higher order term is divisible by at least one Xi variable and by at
least one Yj variable. Hence if |x|, |y| ≤ ϵλ , then the higher order terms all have
absolute value at most ϵ2λ , as required. □
Lemma 2.2. Let R be an adic ring and let I ⊂ R be an ideal of definition. Let
F ∈ R[[X1 , . . . , Xg , Y1 , . . . , Yg ]] be a g-dimensional (commutative) formal group law,
with inverse i ∈ R[[X1 , . . . , Xg ]]. If x, y ∈ Rg with x − y ∈ I g , then F (x, i(y)) ∈ I g .
Proof. Fix k ∈ {1, . . . , g} and x ∈ Rg . Consider Fk (x, i(Y )) ∈ R[[Y1 , . . . , Yg ]]. By
“Taylor expansion near Y = x”, x−y ∈ I g implies that Fk (x, i(y))−Fk (x, i(x)) ∈ I g .
Since Fk (x, i(x)) = 0 (definition of inverse of a formal group law), this proves the
lemma. □

3. Relative S-compactness
The following notion is essential to Proposition 1.2.
Definition. [Lüt16, Def. 3.6.1] Let X → S be a morphism of quasi-separated,
quasi-compact rigid spaces. An open subspace U ⊂ X is said to be relatively
S-compact, written U ⋐S X, if there exists a formal R-model Xfor → Sfor of
X → S such that:
(1) U is induced by an open subscheme Ufor ⊂ Xfor ;
(2) the schematic closure of U0 in X0 is proper over S0 .
A general subset of X is said to be relatively S-compact if it is contained in
an S-relatively compact open subspace of X.
Remark 3.1. If X, S, U are all affinoid, then the above definition is equivalent to:
U is relatively S-compact in X if there exists a closed immersion X → B n (1)×S
which maps U into B n (r) × S for some r < 1.
For 0 < r ≤ 1, define the following rigid spaces over K:
A(r) = {x : r ≤ |x| ≤ r−1 },
A≤ (r) = {x : r ≤ |x| ≤ 1},
A≥ (r) = {x : 1 ≤ |x| ≤ r−1 }.
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 5

Note that A(r) is the same as what we called Gm (r) in the introduction. In this
section, we don’t care about the multiplicative group structure, so I am using the
notation A(r) instead.

Lemma 3.2. Fix r > 1. Let S be an affinoid space. Let Y ⊂ A(r) × S be an


admissible open subset such that A(1) × S is an S-relatively compact subset of Y .
Then Y contains A(r′ ) × S for some r′ < 1.

Proof. Choose formal models Sfor , Xfor , Yfor , Ufor for S, X := A(r) × S, Y , U :=
A(1) × S respectively. By the definition of A(1) × S ⋐S Y , we can choose these
formal models such that Ufor is an open formal subscheme of Yfor , and the schematic
closure U 0 of U0 in Y0 is proper over S0 . We can also choose the formal models
such that Yfor is an open formal subscheme of Xfor .
Let Z0 = X0 \ U 0 , which is an open subscheme of X0 . Thus Z0 corresponds to
an open subdomain Z = Red−1 (Z0 ) ⊂ X. Let
Z ≤ = {(x, s) ∈ Z : |x| ≤ 1}, Z ≥ = {(x, s) ∈ Z : |x| ≥ 1}.
Let
r≤ = sup{|x| : (x, s) ∈ Z ≤ }, r≥ = sup{|x|−1 : (x, s) ∈ Z ≥ }.
(If Z ≤ or Z ≥ empty, set r≤ = r or r≥ = r respectively.)
We claim that r≤ < 1. If Z ≤ is empty, then this is immediate. Otherwise,
assume that Z ≤ is non-empty. Since Z is the rigid generic fibre of Zfor , it is quasi-
compact. Since Z ≤ is a rational subdomain of Z, Z ≤ is quasi-compact. Hence,
by the maximum principle, the absolute value of the first co-ordinate |x| attains
its supremum r≤ on Z ≤ . But Z is disjoint from U = A(1) × S, so |x| = ̸ 1 on Z ≤ .
Hence r≤ < 1.
Similarly, we obtain r≥ < 1.
Choose r′ such that max{r≤ , r≥ } < r′ < 1. We claim that A(r′ ) × S ⊂ Y .
We shall show that A≤ (r′ ) × S ⊂ Y ; a very similar argument will show that
A≥ (r′ ) × S ⊂ Y , completing the proof of the lemma.
Indeed, if (x, s) ∈ A≤ (r′ ) × S, then r≤ < r′ ≤ |x|. Hence by the definition of r≤ ,
(x, s) ̸∈ Z ≤ . Since |x| ≤ 1, it follows that (x, s) ̸∈ Z. Hence
Red(x) ∈ X0 \ Z0 = U 0 ⊂ Y0 .
Thus (x, s) ∈ Y , as required. □

4. Approximation theorem
The following theorem is the form of the approximation theorem which we shall
use. Fix π ∈ K × with |π| < 1.
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 6

Theorem 4.1. Let S be an affinoid space over K. Let X → S and Z → S be


quasi-compact quasi-separated rigid S-spaces. Let U be an admissible open subspace
of Z and let ϕ : U → X be a morphism of rigid S-spaces. Suppose that:
(i) Z is affinoid and U is a Weierstrass domain in Z;
(ii) ϕ(U ) is contained in an affinoid V which is relatively S-compact in X;
(iii) X → S is smooth;
(iv) U ⋐S Z.
Let λ ∈ Z>0 . Then there exists an open subdomain U ′ ⊂ Z and an S-morphism
ϕ′ : U ′ → X with the following properties:
(1) U ⋐S U ′ ⋐S Z;
(2) ϕ′ |U ≡ ϕ mod π λ with respect to given formal R-models.
Theorem 4.1 is a mild generalisation of [Lüt09, Thm. 6.3] because, by replacing
X by X ×S Z and S by Z, [Lüt09, Thm. 6.3] implies Theorem 4.1 under the
additional assumption that U ⋐ Z.
We deduce Theorem 4.1 from the following theorem of Lütkebohmert. In the
following theorem statement, S denotes Raynaud’s rigid generic fibre of Sfor , etc.
Theorem 4.2. [Lüt95, Thm. 7.4], [Lüt16, Thm. 3.6.7] Let Sfor be an affine ad-
missible formal scheme over Spf(R). Let Xfor be a separated admissible formal
Sfor -scheme such that X → S is smooth. Let Zfor → Sfor be an admissible formal
scheme and let Ufor be an open subscheme of Zfor . Let ϕ : Ufor → Xfor be a morphism
of formal schemes over Sfor . Suppose that:
(i) Z is affinoid and U is a Weierstrass domain in Z;
(ii) ϕ(U ) is contained in an affinoid V which is relatively S-compact in X.

Let λ ∈ Z≥0 . Then there exists an admissible formal blowing-up Zfor → Zfor ,
′ ′
which is finite over Ufor , and an open subscheme Ufor ⊂ Zfor , such that:

(a) the schematic closure of (Zfor ×Zfor Ufor )0 (in Z0′ ) is contained in U0′ ;

(b) the schematic closure U 0 (in Z0′ ) is proper over U 0 (closure in Z0 );
(c) there exists a morphism ϕ′ : Ufor

→ Xfor such that ϕ′ |Ufor coincides with ϕ
up to the level λ.
To deduce Theorem 4.1 from Theorem 4.2: Choose formal R-models Sfor , Xfor ,
Zfor , Ufor , such that Ufor is an open subscheme of Zfor [Bos14, Lemma 8.4/5] and
such that X → S, Z → S, ϕ : U → X come from morphisms of formal schemes.
Conditions (i) and (ii) of Theorem 4.2 are the same as conditions (i) and (ii) of

Theorem 4.1. Now Theorem 4.2 gives us a new formal model Zfor of Z and an
′ ′ ′
open subscheme Ufor . Then the rigid generic fibre U of Ufor is an open subdomain

of Z. By Theorem 4.2(b), we have U 0 proper over U 0 , while U 0 is proper over S0
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 7


by Theorem 4.1(iv). Hence U 0 is proper over S0 , so U ′ ⋐S Z. By Theorem 4.2(a),

the schematic closure of (Zfor ×Zfor Ufor )0 in U0′ is equal to its schematic closure
′ ′
in Z0′ , and hence is Zariski closed in U 0 . Since (as we just showed), U 0 is proper

over S0 , we conclude that the schematic closure of (Zfor ×Zfor Ufor )0 in U0′ is proper
over S0 . Hence U ⋐S U ′ . Finally, Theorem 4.2(c) is equivalent to Theorem 4.1(2),
shifting λ by 1.

5. Main proof
For 0 < ϵ ≤ 1, define Gm (ϵ) to be the following affinoid space over K:
Gm (ϵ) = {x : ϵ ≤ |x| ≤ ϵ−1 }.
This is consistent with our earlier definition of Gm (1), but note that Gm (ϵ) is not
closed under multiplication when 0 < ϵ < 1.
Before beginning the proof of Proposition 1.2, we prove the following lemma.
Lemma 5.1. Let S be an affinoid space with a formal R-model Sfor . Let G → S be
a smooth commutative rigid S-group. Let U ⊂ G be an admissible open S-subgroup
whose underlying rigid S-space is isomorphic to Dg × S, via an isomorphism
ζ : U → Dg × S. Write ζ1 , . . . , ζg : U → D for the composition of ζ with the
projection onto the i-th copy of D.
Let 0 < ϵ < 1 and let π ∈ K with |π| = ϵ.
Let ᾱ, ϕ̄ : Gm (1) × S → G be morphisms of rigid S-spaces such that:
(i) ϕ̄ is a homomorphism of S-groups;
(ii) ᾱ extends to a morphism of S-spaces α : Gm (ϵ2 ) × S → G;
(iii) the S-morphism wα : Gm (ϵ) × Gm (ϵ) × S → G defined by wα (x1 , x2 ) =
α(x−1 −1
1 x2 ) · α(x1 ) · α(x2 ) has image contained in U ;
(iv) the S-morphism ū = ᾱ · ϕ̄−1 : Gm (1) × S → G has image contained in U .
Write the Laurent series ūi for ζi ◦ ū : Gm (1) × S → D as
ūij (s)ξ j ∈ O◦ (S)⟨ξ, ξ −1 ⟩.
X
ūi (ξ, s) =
j∈Z

Let λ ∈ Z≥1 . If im(ū) ⊂ π λ U (that is, if ζi (ū(x)) ∈ D(ϵλ ) × S for all x ∈


Gm (1) × S and all i = 1, . . . , g), then
|uij |S ≤ ϵ2 min{λ,|j|}
for all i = 1, . . . , g and all j ∈ Z.
Proof. Write the Laurent series wi for ζi ◦ wα : Gm (ϵ) × Gm (ϵ) × S → D as
wijk (s)ξ1j ξ2k ∈ O◦ (S)⟨πξ1 , πξ1−1 , πξ2 , πξ2−1 ⟩,
X
wi (ξ1 , ξ2 , s) =
j,k∈Z
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 8

where wijk ∈ O◦ (S). Since wi maps Gm (ϵ) × Gm (ϵ) × S into D, we have

|wijk |S ≤ ϵ|j|+|k| (1)

for all j, k ∈ Z.
The S-group structure on U induces an S-group structure on Dg × S. As
described in section 2, the S-group structure on Dg × S can be described by a
formal group law F ∈ O◦ (S)⟨X1 , . . . , Xg , Y1 , . . . , Yg ⟩g . Stating this precisely using
the notation of the current lemma, we have

ζi (x · y) = Fi (ζ1 (x), . . . , ζg (x), ζ1 (y), . . . , ζg (y))

for all (x, y) ∈ U ×S U . By Lemma 2.1, we obtain that, if x, y ∈ D(ϵλ ) × S, then

ζi (x · y) ≡ ζi (x) + ζi (y) mod ϵ2λ (2)

for each i = 1, . . . , g.
Since ϕ̄ is a homomorphism of commutative S-groups, we have

wα (x1 , x2 , s) = ū(x−1 −1
1 x2 ) · ū(x1 ) · ū(x2 )

for all (x1 , x2 , s) ∈ Gm (1) × Gm (1) × S. Using the hypothesis that im(ū) ⊂ π λ U
and applying (2), we obtain the following congruence for all ξ1 , ξ2 ∈ Gm (1):

wi (ξ1 , ξ2 , s) ≡ ūi (ξ1−1 ξ2−1 , s) + ūi (ξ1 , s) + ūi (ξ2 , s) mod π 2λ . (3)

The above is valid as a congruence of Laurent series in O◦ (S)⟨ξ1 , ξ2 , ξ1−1 , ξ2−1 ⟩.


Comparing the coefficients of ξ1−j ξ2−j , we obtain the following congruence:

wijj ≡ ūi,−j mod π 2λ .

By (1), we have |wijj |S ≤ ϵ2|j| , so this proves the lemma. □

Now we prove Proposition 1.2.

Proof of Proposition 1.2. First note that it suffices to construct ϕ locally on an


admissible covering of S and glue, so we can and do assume that S is affinoid.
In the notation of the first paragraphs of [Lüt16, Prop. 7.3.1], we have B̃ = Ũ =
{1}, T̃ = L̃ = H̃ (where ˜· denotes reduction), r = 0, L(c) = H, t = dim(G/S),
s = 0. In particular, the step where H is replaced by L(c) does not change
anything.
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 9

Step 1: Approximation. We apply Theorem 4.1 to our given S, X = G′ , Z =


Gm (a) × S for some a with 0 < a < 1, U = Gm (1) × S and ϕ = ϕ̄. Theorem 4.1(i)
and (iv) hold by our explicit description of Z and U . Theorem 4.1(ii) holds with
V = H, by hypotheses (i) and (ii) of Proposition 1.2. Theorem 4.2(iii) holds
because the hypotheses of Proposition 1.2 include that G → S is smooth.
By Theorem 4.1, there exist an open subdomain U ′ ⊂ Z ⊂ Gm × S and a
morphism of rigid S-spaces α : U ′ → G such that
(1) Gm (1) × S ⋐S U ′ ⋐S Gm × S;
(2) letting ᾱ = α|Gm (1)×S , we have ᾱ0 = ϕ̄0 .
By Lemma 3.2, U ′ contains Gm (ϵ2 ) × S for some ϵ < 1, so we may assume that
U ′ = Gm (ϵ2 ) × S. Increasing ϵ (and shrinking Gm (ϵ2 )) if necessary, we may assume
that there exists π ∈ K × such that ϵ = |π| < 1, and that π has a square root
π 1/2 ∈ K.
Step 2: Set up of U . Let
ū = ᾱ · ϕ̄−1 : Gm (1) × S → H.
(This means multiplication and inverse in the S-group law on H, not functional
composition and inverse.) Let Z ⊂ H denote the preimage under the reduction
map H → H0 of the image of the zero section S0 → H0 . By (2), ū0 is the zero
homomorphism of S0 -group schemes Gm,S0 → H0 , so im(ū) ⊂ Z.
Let µ : H → Ag × S denote the isomorphism H → Gm (1)g × S given by Propo-
sition 1.2(ii), composed with the translation sending (1, . . . , 1) to (0, . . . , 0). Then
g
µ restricts to an isomorphism Z → D+ × S, where D+ denotes the open unit disc.
Let v̄ denote the following composition:
ū g µ
Gm (1) × S −→ Z −→ D+ × S.
Let pi : Ag × S → A1 denote projection onto the i-th coordinate in Ag . Since
Gm (1) × S is affinoid, by the maximum principle, each function pi ◦ v̄ attains
its maximum absolute value, which must be less than 1. Thus the image of v̄
is contained in D(r)g × S for some r < 1, where D(r) ⊂ D+ is the closed disc
around 0 of radius r.
Increasing either r or ϵ (and adjusting π so that we still have |π| = ϵ), we may
assume that r = ϵ2 < 1.
Let U = µ−1 (D(ϵ)g × S) ⊂ Z. Thanks to our explicit formula for µ, we have
U = D(1, ϵ)g × S ⊂ Gm (1)g × S ∼ = H, where D(1, ϵ) denotes the closed disc of
radius ϵ and centre 1. Thus U is an S-subgroup of H.
Let ζ : U → Dg × S denote the following composition:
µ ∼
U −→ D(ϵ)g × S −→ Dg × S,
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 10

where the second arrow is a rescaling of each factor. Via ζ, we can view Agfor × Sfor
as a formal model for U . Explicitly, ζ −1 : Dg × S → U is given by
ζ −1 (x, s) = (1 + πx, s).
Since the image of v̄ is contained in D(ϵ2 )g × S, while U maps to D(ϵ)g × S, the
image of ū reduces to the zero section in U0 . Indeed, the image of ū is contained
in πU .
Step 3: Measuring how far α is from being a group homomorphism. Define a
morphism of rigid S-spaces wα : Gm (ϵ) × Gm (ϵ) × S → G by
wα (x1 , x2 ) = α(x−1 −1
1 x2 )α(x1 )α(x2 ).

Let w̄α denote the restriction of wα to Gm (1) × Gm (1) × S. Since ϕ̄ is a homomor-


phism of commutative S-groups, we have
w̄α (x1 , x2 ) = ū(x−1 −1
1 x2 )ū(x1 )ū(x2 ). (4)
Since the image of ū is contained in πU , so is the image of w̄α .
We claim that, after enlarging ϵ, we may assume that wα factorises through U .
Indeed, consider the set

A = {(x1 , x2 , s) ∈ Gm (ϵ) × Gm (ϵ) × S :


|pi (µ(wα (x1 , x2 , s)))| ≥ ϵ for some i = 1, . . . , g and |x1 |, |x2 | ≤ 1}.
Assume that A is non-empty. If some point (x1 , x2 , s) ∈ A satisfies |x1 x2 | =
1, then it must satisfy |x1 | = |x2 | = 1, i.e. x1 , x2 ∈ Gm (1). Since w̄α maps
Gm (1) × Gm (1) × S into πU , it follows that |pi (µ(wα (x1 , x2 , s)))| ≤ ϵ2 < ϵ for all i,
contradicting (x1 , x2 , s) ∈ A. Thus |x1 x2 | < 1 on A.
The set A is a finite union of affinoids (one for each i). Hence, by the maximum
principle, |x1 x2 | attains its maximum on A. Thus there exists ϵ′ < 1 such that
|x1 x2 | ≤ ϵ′ on A. Now if (x1 , x2 , s) ∈ Gm (ϵ′1/3 ) × Gm (ϵ′1/3 ) × S, we have |x1 x2 | ≥
ϵ′2/3 > ϵ′ so (x1 , x2 , s) ̸∈ A. But |x1 |, |x2 | ≤ 1, so such (x1 , x2 , s) must satisfy that
|pi (µ(wα (x1 , x2 , s)))| < ϵ for all i, so wα (x1 , x2 , s) ∈ U .
Thus, replacing ϵ by max{ϵ, ϵ′1/3 } (which makes Gm (ϵ) smaller and U bigger),
we obtain that wα : Gm (ϵ) × Gm (ϵ) × S → G factors through U .
Step 4: Formal power series for ū. Let ūi : Gm (1)×S → D denote the composition
ū ζ p
Gm (1) × S −→ U −→ Dg × S −→
i
D.
Then ūi ∈ O◦ (Gm (1) × S) = O◦ (S)⟨ξ, ξ −1 ⟩. Thus we can write
ūij ξ j
X
ūi = (5)
j∈Z
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 11

where ūij ∈ O◦ (S) and |uij |S → 0 as |j| → ∞.


The aim of the next step of this proof is to show that the Laurent series (5)
converge on Gm (ϵ2 ) × S. In other words, we will show that ϵ−2|j| |uij |S → 0 as
|j| → ∞.
Step 5: Induction. We claim that
ϵ−2 min{λ,|j|} |ūij |S ≤ 1 (6)
for all j ∈ Z and all λ ∈ Z≥1 . We shall establish this by induction on λ.
For the base case λ = 1, we saw at the end of Step 2 that the image of ū is
contained in πU . Hence, by Lemma 5.1, (6) holds for λ = 1.
For the inductive step, assume that (6) holds for some λ ≥ 1. We shall show
that it holds for λ + 1.
For i = 1, . . . , g, define uλi to be the truncated Laurent series:
λ
ūij ξ j ∈ O◦ (S)[ξ, ξ −1 ].
X
uλi (ξ) =
j=−λ

This defines a holomorphic function uλi ∈ O(Gm (ϵ2 ) × S). By the inductive
hypothesis (6), all non-zero coefficients of uλi satisfy ϵ−2|j| |ūij |S ≤ 1. Hence
uλi (Gm (ϵ2 ) × S) ⊂ D.
We can therefore define a morphism of rigid S-spaces
uλ = ζ −1 ◦ (uλ1 , . . . , uλg , s) : Gm (ϵ2 ) × S → U.
By the inductive hypothesis (6), the coefficients of ūi which have been removed
in uλi satisfy ϵ−2(λ+1) |ūij |S ≤ 1, so
ūi ≡ uλi mod π 2(λ+1) (7)
as Laurent series, and hence as holomorphic functions on Gm (1) × S.
Define
βλ = α · u−1 2
λ : Gm (ϵ ) × S → G,

β̄λ = βλ |Gm (1)×S ,


v̄λ = β̄λ · ϕ̄−1 : Gm (1) × S → G.
We can calculate v̄λ = ū · u−1
λ , so the image of v̄λ is contained in U . Hence we can
define v̄λi : Gm (1) × S → D to be the i-th projection of ζ ◦ vλ .
Let ι denote the inverse for the formal group law F . By (7) and Lemma 2.2, we
have
  
v̄λi (ξ) = Fi ū1 (ξ), . . . , ug (ξ), ι uλ1 (ξ), . . . , uλg (ξ) ≡ 0 mod π 2(λ+1) .
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 12

Thus v̄λ maps Gm (1) × S into π 2(λ+1) U . A fortiori, v̄λ maps Gm (1) × S into π (λ+1) U .
If we define wβλ : Gm (ϵ) × Gm (ϵ) × S → G by
wβλ (x1 , x2 ) = βλ (x−1 −1
1 x2 ) · βλ (x1 ) · βλ (x2 ),

then we can calculate


wβλ (x1 , x2 ) = wα (x1 , x2 ) · uλ (x−1 −1
1 x2 ) · uλ (x1 ) · uλ (x2 ).

Hence the image of wβλ is contained in U .


Therefore, we can apply Lemma 5.1 to β̄λ instead of ᾱ (thus replacing ū by v̄λ )
to deduce that
ϵ−2 min{λ+1,|j|} |v̄λij |S ≤ 1 (8)
for all j ∈ Z.
The components of the formal group law F have the form
Fi (X, Y , s) = Xi + Yi + qi (X, Y , s) (9)
where qi (X, Y , s) ∈ O◦ (S)⟨X, Y ⟩ with every monomial in qi divisible by at least
one Xi′ and by at least one Yi′ .
Since ū = v̄λ · uλ , we have
 
ūi (ξ) = Fi v̄λ1 (ξ), . . . , v̄λg (ξ), uλ1 (ξ), . . . , uλg (ξ)
 
= v̄λi (ξ) + uλi (ξ) + qi v̄λ1 (ξ), . . . , v̄λg (ξ), uλ1 (ξ), . . . , uλg (ξ) . (10)
Write  
qλij ξ j
X
qi v̄λ1 (ξ), . . . , v̄λg (ξ), uλ1 (ξ), . . . , uλg (ξ) =
j∈Z

where qλij ∈ O◦ (S). The non-zero coefficients of uλi satisfy |ūij |S ≤ ϵ2|j| , and the co-
efficients of v̄λi satisfy |v̄ij |S ≤ ϵ2|j| . Since O◦ (S)⟨π 2 ξ, π 2 ξ −1 ⟩ is a ring, it follows that
every monomial in v̄λ1 (ξ), . . . , v̄λg (ξ), uλ1 (ξ), . . . , uλg (ξ) lies in O◦ (S)⟨π 2 ξ, π 2 ξ −1 ⟩,
hence that |qλij |S ≤ ϵ2|j| for all j.
Since also |ūij |S ≤ ϵ2|j| ≤ 1 and |v̄λij |S ≤ ϵ2(λ+1) for all i, j, and every term of qi
is divisible by some v̄λi′ , we obtain |qλij |S ≤ ϵ2(λ+1) . Thus we get
|qλij |S ≤ ϵ2 min{λ+1,|j|}
for all j ∈ Z.
Using (10), (8), and the fact that non-zero coefficients of uλi (namely, ūij for
|j| ≤ λ) satisfy |ūij |S ≤ ϵ2|j| = ϵ2 min{λ+1,|j| , we obtain
|ūij |S ≤ ϵ2 min{λ+1,|j|}
for all j ∈ Z, that is, (6) holds for λ + 1, completing the inductive step.
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 13

Step 6: Conclusion. We now know that (6) holds for all λ ∈ Z≥1 . Thus

ϵ−2|j| |ūij |S ≤ 1

for all i = 1, . . . , g and all j ∈ Z. Hence ūi converges on

{(x, s) ∈ Gm × S : ϵ2 < |x| < ϵ−2 }.

Thus ūi defines a morphism of rigid spaces Gm (ϵ) × S → D for each i, so ū


defines a morphism of rigid S-spaces Gm (ϵ) × S → U . Define

ψ = α · ū−1 : Gm (ϵ) × S → G.

By the definition of ū, ψ|Gm (1)×S = ϕ̄. Since ϕ̄ is an S-group homomorphism, the
identity principle implies that

ψ(ξ1 ) · ψ(ξ2 ) = ψ(ξ1 · ξ2 ) (11)

for all ξ1 , ξ2 ∈ Gm (ϵ1/2 ) × S.


Define An ⊂ Gm to be the annulus

An = π n/2 Gm (ϵ1/2 ) = {x ∈ Gm : ϵ(n+1)/2 ≤ |x| ≤ ϵ(n−1)/2 }.

The sets An for n ∈ Z form an admissible cover of Gm , with non-empty intersections

An ∩ An+1 = {x ∈ Gm : ϵ(n+1)/2 ≤ |x| ≤ ϵn/2 }

and
An ∩ An+2 = {x ∈ Gm : |x| = ϵ(n+1)/2 }.
Define ψn : An × S → G by

ψn (π n/2 ξ) = ψ(π 1/2 )n ψ(ξ)

for all ξ ∈ Gm (ϵ1/2 ). If (y, s) ∈ (An ∩ An+1 ) × S, with y = π n/2 ξ = π (n+1)/2 π −1/2 ξ,
then
ψn (y) = ψ(π 1/2 )n ψ(ξ) = ψ(π 1/2 )n ψ(π 1/2 )ψ(π −1/2 ξ) = ψn+1 (y),
where the middle equality is (11), valid because π 1/2 , π −1/2 ξ ∈ Gm (ϵ1/2 ). Applying
this twice shows that ψn and ψn+2 agree on An ∩ An+2 .
Thus we can glue the ψn together into a morphism of rigid S-spaces ϕ : Gm ×
S → G. As a consequence of (11) and the definition of ψn , ϕ is an S-group
homomorphism.
Uniqueness of ϕ holds by the identity principle for morphisms of rigid spaces. □
EXTENSION OF RELATIVE RIGID HOMOMORPHISMS 14

6. Proof of corollary
Now we prove Corollary 1.3. It suffices to assume that S is affinoid and that G
has no boundary, since we could prove the corollary on each affinoid subspace Si
from Corollary 1.3(ii), and glue the resulting homomorphisms ϕ.
Write ξ1 , . . . , ξg : Gm × S → Ggm × S for the inclusions of the direct factors Gm .
Write ξ¯i for the restriction of ξi to Gm (1) × S.
We apply Proposition 1.2 to ϕ̄i := ϕ̄ ◦ ξ¯i : Gm (1) × S → G. By the hypotheses
of Corollary 1.3, ϕ̄i factors through H, which is isomorphic to Gm (1)g × S (via ϕ̄).
Since H ∼ = Gm (1)g × S is affinoid and G has no boundary, there exists an open
affinoid subspace G′ ⊂ G such that H ⋐ G′ . It follows that H ⋐S G′ (choose
a formal R-model G′for → Sfor → Spf(R) such that H is induced by an open
subscheme Hfor ⊂ G′for ; then H 0 → Spec(k) is proper, so H 0 → S0 is proper by

[Har75, Cor. II.4.8(e)], where H 0 denotes the schematic closure of H 0 in G0 .) Thus
ϕ̄i and H satisfy conditions (i) and (ii) of Proposition 1.2.
Thus, by Proposition 1.2, for each i = 1, . . . , g, we obtain a unique homomor-
phism of rigid S-groups ϕi : Gm × S → G whose restriction to Gm (1) × S is equal to
ϕ̄i . By the universal property of fibre products, there is a unique ϕ : Ggm × S → G
satisfying ϕ ◦ ξi = ϕi for each i, completing the proof of Corollary 1.3.

References
[Bos14] S. Bosch, Lectures on formal and rigid geometry, Lecture Notes in Mathematics, vol.
2105, Springer, Cham, 2014.
[DO] C. Daw and M. Orr, The large Galois orbits conjecture under multiplicative degeneration,
preprint, available at arxiv.org/abs/2306.13463.
[Har75] R. Hartshorne, On the De Rham cohomology of algebraic varieties, Inst. Hautes Études
Sci. Publ. Math. (1975), no. 45, 5–99.
[Lüt90] W. Lütkebohmert, Formal-algebraic and rigid-analytic geometry, Math. Ann. 286 (1990),
no. 1-3, 341–371.
[Lüt95] , The structure of proper rigid groups, J. Reine Angew. Math. 468 (1995), 167–
219.
[Lüt09] , From Tate’s elliptic curve to abeloid varieties, Pure Appl. Math. Q. 5 (2009),
no. 4, Special Issue: In honor of John Tate. Part 1, 1385–1427.
[Lüt16] , Rigid geometry of curves and their Jacobians, Ergebnisse der Mathematik
und ihrer Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics, vol. 61,
Springer, Cham, 2016.

Orr: Department of Mathematics, The University of Manchester, Alan Turing


Building, Oxford Road, Manchester M13 9PL, United Kingdom
Email address: martin.orr@manchester.ac.uk

You might also like