LOCAL DEFORMATIONS OF WILD GROUP ACTIONS
By
Gregory Sulisz

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Mathematics
2012

ABSTRACT
LOCAL DEFORMATIONS OF WILD GROUP ACTIONS
By
Gregory Sulisz
In this dissertation, we study deformations of actions of a cyclic group of order p on the
formal power series ring k[[u1 , . . . , un ]], where k is a field of characteristic p > 0. We draw
upon work of B. Peskin in [20] to reduce, under certain hypotheses, the task of determining
the tangent space of the deformation functor D to a problem in invariant theory. When
n = 2 and p = 3, we use these results to explicitly compute the tangent space of D and then
generalize results of M´zard and Bertin for smooth curves to smooth surfaces. In particular,
e
we compute the prorepresentable hull of the equicharacteristic local deformation functor
D of a smooth surface with finite, cyclic group action at a point of wild ramification in
characteristic 3.

To all those of you who have made this possible: especially my parents, my wife, and my
adviser. Thank you!

iii

TABLE OF CONTENTS

1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

2

Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Schemes with Group Action . . . . . . . . . . . . . . . . . . . . . . .
Functors of Artin Rings . . . . . . . . . . . . . . . . . . . . . . . . .

6
6
7

3

The Deformation Theory of Schemes with Group Action . . . . .

9

4

The Existence of a Hull . . . . . . . . . . . . . . . . . . . . . . . . . . .

15

5

Computation of the Hull . . . . . . . . . . . . . . . . . . . . . . . . . .
The Single Jordan Block Case . . . . . . . . . . . . . . . . . . . . . .
The Case p = 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

19
20
24

An Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

28

2.1
2.2

5.1
5.2
6

Bibliography . . . . . . . . . . . . . . .

iv

. . . . . . . . . . . . . . . . . . .

34

1

Introduction

The main idea of deformation theory is to learn how an object can be continuously varied
in correspondence with the points of some parameter space. These objects can be chosen
to be subschemes of a fixed scheme, line bundles, nonsingular varieties, coherent sheaves, or
singularities, among other things. This theory has applications in a wide range of different
fields in mathematics. In this dissertation we study the local deformation functor associated
to a scheme with certain wild group actions. Along with this functor there often exists a
ring, called the (uni)versal deformation ring, that determines what local deformations can
arise. The typical goal is to determine this (uni)versal deformation ring for as many group
actions as possible.
In the last half century, research in deformation theory has thrived in algebraic geometry. The foundation of this large subject was laid by Kodaira and Spencer, Grothendieck,
Schlessinger, Illusie, M. Artin, Deligne, and others. Recently, the deformation theory of
Galois representations has found important applications in number theory in the work of
Wiles, Taylor-Wiles, and others ([26], [25]). M. Artin and others have studied deformations
of singularities. We focus on deformations of schemes with an action by a finite group G
over an algebraically closed field of characteristic p > 0. Most of the results in this area thus
far concern actions on smooth curves. (See [5], [6], and [18] for such results on deformations and local lifting problems). Grothendieck showed that at points of tame ramification,
meaning that p does not divide the order |G| of the group G, there are no obstructions
to lifting infinitesimal deformations and thus the versal deformation ring is a formal power
series ring ([11]). Bertin, Green, and Matignon demonstrate nontrivial obstructions when
the ramification is wild, i.e., when p divides |G| ([1], [10]). Progress toward determining the
1

(uni)versal deformation ring at points of wild ramification has been made exclusively in the
case of smooth curves. However, even in this case, explicit computations have only been
successful in certain special cases. M´zard and Bertin have computed the versal deformation
e
ring when the conductor m of the automorphism σ defining the cyclic group action is m = 1
and p ≥ 3 ([2]). In the case that m > 1 and (m, p) = 1, they are able to determine a quotient
of the hull and compute the hull’s Krull dimension. They also prove a local-global principle
for deformations. We work to prove local results similar to those of M´zard and Bertin for
e
higher-dimensional schemes. Let us now make the above statements more precise and give
an overview of the contents of this dissertation.
In section 2 we introduce some of the preliminary definitions and results that will be
required. Particularly, we set up notations for group actions on schemes, give a needed fact
for G-torsors, state a result from descent theory that we will use, and recall Schlessinger’s
theory of functors of Artin rings.
After this we start looking at equivariant deformations in section 3. The focus of our
attention will be the local equivariant deformation theory of a smooth scheme X with
dim(X) = n over a field k of prime characteristic p > 0 at a point of wild ramification
x. Particularly, we consider the completion of the stalk at x, OX,x ∼ k[[u1 , . . . , un ]], and
=
fix an action ρ : G → Autk (k[[u1 , . . . , un ]]) where G = σ is assumed to be cyclic of order
p. We will see that with this data one can define a local deformation functor D : C → Sets to
which Schlessinger’s results are applicable. The functor D takes a ring A from the category
C of local Artinian k-algebras with residue field k to the set of local deformations of the
base action ρ to A modulo some equivalence. We end this section by making explicit the
fundamental result that shows that the tangent space of this local deformation functor is iso-

2

∂
morphic to H 1 (G, Θ), the first group cohomology of the space Θ = ⊕n k[[u1 , . . . , un ]]
j=1
∂uj
of k-derivations of OX,x .
The objective of section 4 is to show that Schlessinger’s criterion for the existence of
hull are satisfied for the functor D. We then move on to the task of computing the tangent
space of this functor in section 5. By the results in section 3 this amounts to computing the
first cohomology group H 1 (G, Θ). This is very difficult in general, but in this dissertation,
we will exhibit a family of cases in which we can reduce this calculation to a calculation in
invariant theory.
These are cases when n = p − 1 and the action can be put in the following form:

σu1 = u1 + f (u1 , . . . , un )
σu2 = u2 + u1
.
.
.

(1)

σun = un + un−1 ,
where ord(f ) ≥ 2 and f is invariant. Actually, under the assumption that the linear terms
of the action ρ form a single Jordan block when in Jordan form, B. Peskin [20, p. 77] shows
that there exists a change of coordinates that puts the action in the form (1). Additionally,
we assume that f is invariant under this action and that n = p − 1.
The quotient k[[u0 , . . . , un ]]/(u0 − f ), where f ∈ k[[u1 , . . . , un ]] is the power series
given above, can be endowed with a G-action by defining σu0 = u0 and σui = ui + ui−1
for 1 ≤ i ≤ n. Then k[[u1 , . . . , un ]] ∼ k[[u0 , u1 , . . . , un ]]/(u0 − f ) as G-modules. Set
=
R = k[[u0 , . . . , un ]]/(u0 − f ). Our result that simplifies the calculation of H 1 (G, Θ) is the
following:
3

THEOREM 6. Suppose that the ring k[[u1 , . . . , un ]] has G-action given by (1) above
with f invariant and n = p − 1. Then H 1 (G, Θ) ∼ H 0 (G, R) and H 2 (G, Θ) ∼ H 1 (G, R).
=
=

Here H 0 (G, R) is the Tate cohomology group of G with coefficients in R. The remaining
steps to determine the tangent space H 1 (G, Θ) are admittedly still challenging, but Theorem
6 greatly simplifies these computations and allows us to make progress in certain cases.
Particularly, since H 0 (G, R) ∼ RG /Im(T r), the computation of the tangent space of D is
=
reduced to determining the invariant ring RG and the image of the trace map T r. Typically,
the more challenging part is computing RG . Even when the action of G = σ on the
polynomial ring k[u1 , . . . , un ] is linear in the sense that σui contains only terms of degree
one for all i, this is not trivial. With the aid of computer algebra systems, there is a good
deal of current research in computational invariant theory dealing with computing RG for
such wild group actions: for example, see the work of Campbell and Hughes [4], Shank
[23], Shank-Wehlau [24], and Campbell-Fodden-Wehlau [3]. Much of this work has been
summarized nicely in [7] by Derksen and Kemper.
Using the above and a result of Peskin [20, p. 96] that gives the invariant ring RG , we
then obtain a complete determination of the tangent space when n = 2, the characteristic p
is 3, and the action is free off the closed point. Also, by the main result of Peskin [20, p. 88],
there is an s ≥ 1 such that after a choice of variables the action (1) above can be written in
the form
σu1 = u1 + y s

(2)

σu2 = u2 + u1 ,
where y = N u2 is the norm of u2 under this action. Note that the norm y and the action
4

itself are defined recursively here. We can then prove the following result.

THEOREM 7. Suppose char(k) = 3 and the action of G = σ on k[[u1 , u2 ]] is free off
the closed point. Then H 1 (G, Θ) ∼ k[y]/(y s ).
=

With this result in hand we can prove one of the main results of this dissertation. Particularly, we are able to determine the hull of the local equicharacteristic deformation functor D,
i.e., the functor restricted to the subcategory C3 of C consisting of k-algebras of characteristic
3.

THEOREM 8. Suppose char(k) = 3 and the action ρ : G → Autk (k[[u1 , u2 ]]) is free
off the closed point. Then the hull of the deformation functor D|C is k[[x0 , x1 , . . . , xs−1 ]],
3
where s ≥ 1 is the integer given in (2) above.

In the final section, we consider a specific example of an action of Z/3Z = σ on the
Fermat quartic X ⊂ P3 given by the equation x4 + x4 + x4 + x4 = 0. Namely, we look at
0
1
2
3
k
the action given by σ(x0 , x1 , x2 , x3 ) = (x0 , x2 , x3 , x1 ). This example was considered
in a paper of Dolgachev and Keum [8, p. 114]. There is a unique fixed point for this action
in the case that char(k) = 3. By Peskin’s results, we know that the induced local action
at this fixed point can be put in the form (2) after some change of coordinates. Applying
the above results, we show that the action can be put in this form with s = 1. In doing so
we demonstrate an example where the local wild action can be lifted to characteristic zero.
This also provides a specific example where the local actions we consider in this dissertation
arise from a global action.
5

2

Preliminaries

2.1

Schemes with Group Action

We now work in the category Sch/k of schemes over a fixed algebraically closed field k.
Let G be a group. If X is a scheme in Sch/k, an action of G on X is given by a group
homomorphism ρ : G → Autk (X). The image of g ∈ G under ρ will be denoted by ρg , or
by g when no confusion could arise. If T is a set on which G acts, we will denote the set of
fixed points by T G . The functor of fixed points X G of X ∈ Sch/k with G-action is given by

X G : Sch/k → Sets, (T → k) → X(T )G ,

where X(T ) = {T → X} denotes the set of T -valued points of X. It can be shown that
X G is represented by a subscheme of X [9, p. 293]. The action of G on X is free if the set
of P -valued points X G (P ) is empty for all P ∈ Sch/k. If S is a local ring with maximal
ideal mS , we say that the action of G on X = Spec S is free off the closed point if X G is
supported on {mS } ∈ Spec S = X.
One result related to actions of a group G on a scheme X that we will require in section
4 is the following result, which can be found in [13, p. 216].
THEOREM. Let R be a ring, A an R-algebra, and G a finite group acting on A by Ralgebra automorphisms. Suppose that G acts freely on A in the sense that (Spec A)G = ∅.
Then A is a finite ´tale G-torsor over AG , and the natural map
e

A⊗ G A→
A

A, x ⊗ y → (· · · , x ⊗ g(y), · · · )
g∈G
6

is an isomorphism of left A-algebras.
In the above theorem, we say A is a G-torsor over AG when the map Spec(A) ×
Spec(AG )
G → Spec(A) ×
Spec(A) given by (a, g) → (a, ga) is an isomorphism. We will
Spec(AG )
also say that Spec A is a G-principal homogeneous space over Spec(AG ). Note here that
Spec(AG ) is the quotient Spec(A)/G.
In the sections that follow we will also require a well-known fact from descent theory. If
A is a ring, M is an A-module, and the group G acts on both A and M , we say that M has
compatible G-action if g · (am) = (ga) · (gm). As in [16, p. 18], one can prove that when G
acts freely on A the following map gives an equivalence of categories:


 A-modules with





←→ AG -modules ,



 compatible G-action 
M −→ M G
A ⊗ G N →− N.
A
Under this correspondence, we note that M is finitely generated if and only if M G is finitely
generated.

2.2

Functors of Artin Rings

Let Λ be a local Noetherian ring with maximal ideal mΛ with residue field k = Λ/mΛ .
Denote by C the category of Artinian local Λ-algebras having residue field k. Let C be the
category consisting of complete local Noetherian Λ-algebras A for which A/mn ∈ C for all n
A
and having local Λ-algebra homomorphisms as morphisms. Note that C is a full subcategory
of C. Suppose F : C → Sets is a functor such that F (k) is a singleton set. A couple (A, ε) is
7

a pair such that A ∈ C and ε ∈ F (A). We extend the definition of F from C to C by setting
F (A) = lim F (A/mn ) for A ∈ C. With this definition, we can consider pro-couples (A, ε),
A
←
−
n
where A ∈ C and ε ∈ F (A).
Let us now sketch an argument showing that F (A) ∼ Hom(hA , F ), where hA (R) =
=
Hom(A, R) for R ∈ C. In order to set up this isomorphism, start with ξ = (ξj ) ∈ F (A). Let
u : A → R be a homomorphism in C. Since R is Artinian, there is some l such that ml = 0.
R
Since u(mA ) = mR , it follows that ml ⊂ Ker(u). Thus u factors through ul : A/ml → R.
A
A
To build the desired homomorphism in Hom(hA , F ) from ξ we send u : A → R to F (ul )(ξl ).
It is easily checked that this gives an isomorphism. A pro-couple (A, ε) therefore naturally
induces a morphism of functors hA → F .
DEFINITION. A morphism F → H of functors is said to be smooth if F (S) → F (R)×H(R)
H(S) is surjective for any surjection S → R in C.
As discussed in Schlessinger, it suffices to check the surjectivity of this map for all small
extensions S → R in C in order to show that F → H is smooth. A map S → R in C is
called a small extension if its kernel is a nonzero principal ideal (t) such that t · mS = 0. We
define the tangent space of a functor F by F (k[ε]), and denote this by tF . We will use the
special notation tA to denote the tangent space of the functor hA .
DEFINITION. A pro-couple (A, ε) is a prorepresentable hull if the induced map hA → F
is smooth and the map tA → tF of tangent spaces is bijective. The pair (A, ε) is a universal
deformation ring if the induced map hA (R) → F (R) is an isomorphism for all R in C.
In [21, p. 212], Schlessinger provides necessary and sufficient conditions for a functor
F : C → Sets, with F (k) a singleton set, to have a prorepresentable hull and a universal
deformation ring.

8

THEOREM. Suppose φ : A → A and φ : A → A are maps in C and consider the
natural map
f : F (A ×A A ) → F (A ) ×F (A) F (A ).
Then F has a prorepresentable hull if and only if the conditions (H1 ), (H2 ), and (H3 ) below
are satisfied:
(H1 ) If φ is a small extension, then f is surjective.
(H2 ) If A = k and A = k[ε], then f is a bijection.
(H3 ) F (k[ε]) is a finite-dimensional vector space over k.
F has a universal deformation ring if and only if F satisfies conditions (H1 ) − (H3 ) above
and the condition:
∼
(H4 ) For any small extension A → A, f : F (A ×A A ) → F (A )×F (A) F (A ) is a bijection.
For more information concerning functors of Artin rings, one can consult Schlessinger’s
famous paper [21] where these notions were first introduced.

3

The Deformation Theory of Schemes with Group Action

Suppose X is a connected n-dimensional, finite type separated scheme over a fixed algebraically closed field k of characteristic p > 0. We will assume that X is smooth over k.
Let G ⊂ Autk (X) be a finite subgroup and denote this inclusion by ρ : G → Autk (X). A
deformation of the pair (X, ρ) to an object R in the category C is a triple (X, ρ, φ) consisting of a smooth scheme X of finite type over R, an injective morphism ρ : G → AutR (X),
σ → ρσ , and an isomorphism φ : X ⊗Spec R Spec k → X of schemes over k such that ρ = ρ
via this map. Two deformations (X1 , ρ1 , φ1 ) and (X2 , ρ2 , φ2 ) are said to be isomorphic if
there exists an isomorphism ψ : X1 → X2 of schemes over R such that φ1 ◦ (ψ ⊗R k) = φ2
9

and ψ ◦ (ρ2 )σ = (ρ1 )σ ◦ ψ for all σ ∈ G. We say that a closed point x ∈ X(k) is a
point of wild ramification when the stabilizer subgroup Gx ⊂ Autk (X) is of order divisible by p. Further, OX,x ∼ k[[u1 , . . . , un ]] and it follows that there is an injective mor=
phism Gx → Autk (k[[u1 , . . . , un ]]). A representative of a deformation of (X, ρ) to R gives
a similar morphism Gx → AutR (R[[u1 , . . . , un ]]), which reduces by R → R/mR ∼ k
=
to the initial representation of Gx over k[[u1 , . . . , un ]]. Henceforth, fix a representation
ρ : G → Autk (k[[u1 , . . . , un ]]), and denote the image of σ under ρ by ρσ . One is led to
define the local deformation functor

D = DG,ρ : C → Sets,

R→



 liftings G → Aut (R[[u , . . . , un ]]) of ρ

1

R











 modulo conjugation by an element




 of ker Aut (R[[u , . . . , u ]]) → Aut (k[[u , . . . , u ]])
n
n
1
1
R
k








Progress toward determining the (uni)versal deformation ring has been made by M´zard
e
and Bertin [2] for the case of smooth projective curves, which corresponds to n = 1. We
work to obtain similar results for schemes of higher dimension.
We henceforth assume that G = σ is cyclic of order p. So σ is an automorphism of
k[[u1 , . . . , un ]] of order p. The proof of a result given by Cornelissen and Kato for curves
[6, p. 442] applies also in the case of higher-dimensional schemes because their argument
∗
is purely formal. Namely, letting ΘA = Ω1
A/k , we have the following fundamental result. Note that when A = k[[u1 , . . . , un ]], we will denote Θk[[u ,...,u ]] by Θ and we have
n
1
∂
Θ = ⊕n k[[u1 , . . . , un ]]
.
j=1
∂uj

10

.

PROPOSITION 1. The map D(k[ε]) → H 1 (G, Θ) given by ρ → dρ , where dρ is the
1-cocycle given by
dρ (σ) =

ρσ ◦ ρ−1 − Id
σ
for σ ∈ G,
ε
n

is a bijection. Here we identify the k-derivation dρ (σ) with

dρ (σ)(uj )
j=1

∂
∈ Θ.
∂uj

Proof. The G-action on Θ is the adjoint action. Namely, for a k-derivation δ ∈ Θ and
σ ∈ G,
σ · δ = ρσ ◦ δ ◦ ρ−1 .
σ
For a lifting ρ of ρ, write ρσ (x) = ρσ (x) + ρσ (x)ε for σ ∈ G and x ∈ k[[u1 , . . . , un ]]. We
first show that the cocycle dρ is determined by ρσ . First note that

ρσ (x + yε) = ρσ (x) + (ρσ (x) + ρσ (y))ε.
ρ ◦ρ−1 −Id
Therefore, dρ (σ) = σ σ
= ρσ ◦ ρ−1 . Next note that for x, y ∈ k[[u1 , . . . , un ]],
σ
ε
ρσ (x)ρσ (y) + ρσ (xy)ε = ρσ (xy)
= ρσ (x)ρσ (y)
= (ρσ (x) + ρσ (x)ε)(ρσ (y) + ρσ (y)ε)
= ρσ (x)ρσ (y) + (ρσ (x)ρσ (y) + ρσ (y)ρσ (x))ε,

which implies ρσ (xy) = ρσ (x)ρσ (y) + ρσ (y)ρσ (x). Thus for a, b ∈ k[[u1 , . . . , un ]], we have

11

dρ (σ)(ab) = ρσ (ρ−1 (ab))
σ
= ρσ (ρ−1 (a)ρ−1 (b))
σ
σ
= ρσ (ρ−1 (a))ρσ (ρ−1 (b)) + ρσ (ρ−1 (a))ρσ (ρ−1 (b))
σ
σ
σ
σ
= a(dρ (b)) + (dρ (a))b

and so dρ (σ) is a k-derivation. Next, for σ, τ ∈ G,

ρστ (x) + (ρστ (x) + ρστ (y))ε = ρστ (x + yε)
= ρσ (ρτ (x + yε))
= ρσ (ρτ (x) + (ρτ (x) + ρτ (y))ε)
= ρστ (x) + (ρσ (ρτ (x)) + ρσ (ρτ (x)) + ρσ (ρτ (y)))ε.

This implies that ρστ = ρσ ◦ ρτ + ρσ ◦ ρτ , and so

dρ (στ )(x) = ρστ (ρ−1 (x))
στ
= (ρσ ◦ ρτ + ρσ ◦ ρτ )(ρ−1 (x))
στ
= ρσ (ρ−1 (x)) + ρσ (ρτ (ρ−1 (ρ−1 (x))))
σ
τ
σ
= dρ (σ)(x) + σ · (dρ (τ ))(x).

Thus dρ is a cocycle.
To show the map d is well-defined, consider two isomorphic liftings ρ and ρ of ρ. That

12

is, there exists ψ ∈ Autk[ε] (k[ε][[u1 , . . . , un ]]) such that ψ = Idk[[u ,...,u ]] and ψ ◦ ρσ =
n
1
ρσ ◦ ψ. Since ψ = Id, we can write ψ(x) = x + δ(x)ε for x ∈ k[[u1 , . . . , un ]] where δ ∈ Θ.
We then have

ψ(ρσ (x + yε)) = ρσ (ψ(x + yε))
⇒ρσ (x) + (ρσ (x) + δ(ρσ (x)) + ρσ (y))ε = ρσ (x) + (ρσ (x) + ρσ (δ(x)) + ρσ (y))ε
⇒ρσ (x) + ρσ (y) + δ(ρσ (x)) = ρσ (x) + ρσ (δ(x)) + ρσ (y)
⇒(ρσ − ρσ )(x) = ρσ (δ(x)) − δ(ρσ (x)).

Thus

(d σ − dρ σ)(x) = (ρσ ◦ ρ−1 − ρσ ◦ ρ−1 )(x)
σ
σ
ρ
= (ρσ − ρσ )(ρ−1 (x))
σ
= (ρσ ◦ δ − δ ◦ ρσ )(ρ−1 (x))
σ
= ρσ (δ(ρ−1 (x))) − δ(x)
σ
= (σ · δ − δ)(x)
= (σ − Id)(δ)(x).

Hence (d ) − (dρ ) is a coboundary.
ρ
If d σ − dρ σ = (σ − Id)(δ) for some δ ∈ Θ, then we can define a map ψ(x + yε) =
ρ
x + (y + δ(x))ε. ψ is clearly additive since δ is additive, and

13

ψ ((x1 + y1 ε)(x2 + y2 ε)) = ψ (x1 x2 + (x1 y2 + x2 y1 ) ε)
= x1 x2 + (x1 y2 + x2 y1 + δ(x1 x2 ))ε
= x1 x2 + (x1 y2 + x2 y1 + x1 δ(x2 ) + x2 δ(x1 ))ε
= (x1 + (y1 + δ(x1 ))) (x2 + (y2 + δ(x2 )))
= ψ(x1 + y1 ε)ψ(x2 + y2 ε).

Further, ψ(x + yε) = 0 =⇒ x + (y + δ(x))ε = 0 =⇒ x = y = 0, so ψ is injective. ψ is also
obviously surjective. Noting that d σ − dρ σ = (σ − Id)(δ) implies that ρσ (x) − ρσ (x) =
ρ
ρσ (δ(x)) − δ(ρσ (x)), we have

ψ ◦ ρ(x + yε) = ψ ρσ (x) + (ρσ (x) + ρσ (y))ε
= ρσ (x) + ρσ (x) + ρσ (y) + δ(ρσ (x)) ε
= ρσ (x) + ρσ (δ(x)) − δ(ρσ (x)) + ρσ (x) + ρσ (y) + δ(ρσ (x)) ε
= ρσ (x) + ρσ (x) + ρσ (y + δ(x)) ε
= ρ (x + (y + δ(x))ε)
= ρ ◦ ψ(x + yε).

So ψ is an automorphism that shows ρ and ρ are isomorphic. Thus d is injective.
Lastly, we show that d is surjective. Starting with a 1-cocycle f : G → Θ, we get an
automorphism ρσ by defining ρσ (x) = ρσ (x) + (f σ)(ρσ (x))ε for each σ ∈ G. This gives the
desired lifting of ρ corresponding to the cocycle f .

14

4

The Existence of a Hull

We will now show that conditions (H1 ) − (H3 ) are satisfied for the functor D defined in the
previous section.
PROPOSITION 2. Suppose the action of G on S = k[[u1 , . . . , un ]] is free off the closed
point of SpecS. Then the local deformation functor D : C → Sets defined in section 3 has a
prorepresentable hull.
Proof. Let φ : A → A and φ : A → A be morphisms in the category C.
(H1 ) :

Suppose that the map φ : A → A is a small extension with kernel (t). In order

to verify the (H1 ) property, we must show that f : D(A ×A A ) → D(A ) ×D(A) D(A ) is
surjective. Let (ε , ε ) ∈ D(A ) ×D(A) D(A ), where D(φ )(ε ) = D(φ )(ε ) = ε. Suppose
that ε = [ρ ] and ε = [ρ] are representatives such that ρ : G → Aut A [[u1 , . . . , un ]] is a lift
of ρ : G → Aut A[[u1 , . . . , un ]] via φ . Let ρ be a representative of ε . Since D(φ )(ε ) = ε,
there is an automorphism ψ ∈ AutA (A[[u1 , . . . , un ]]) showing that D(φ )(ρ ) and ρ are
equivalent, i.e., ψ = Id and ψ ◦ D(φ )(ρ ) ◦ ψ −1 = ρ. Suppose that ψ is the automorphism
given by
ui →

n c u +d
i
j=1 i,j j

for 1 ≤ i ≤ n, where ci,j ∈ A and di ∈ A[[u1 , . . . , un ]].
Since φ is surjective, there are lifts of the elements ci,j ∈ A and the coefficients of di
to A : call these lifts ci,j and di , respectively. Next we show that the map ψ given by
ui →

n c u +d
j=1 i,j j
i

is an A -automorphism. Since the ci,j and di are lifts under the small extension A → A, we
have that ci,j = ci,j + tci,j and di = di + tdi for some ci,j ∈ A and di ∈ A [[u1 , . . . , un ]].
Let ai denote the constant term of di and ai the constant term of di . Since ψ is an
15

automorphism, ai ∈ mA . Thus ai ∈ φ −1 (ai ) ⊆ φ −1 (mA ) ⊆ mA . Also, we know that
the determinant of the matrix of coefficients of the linear terms for ψ is a unit in A. Then

=
det ci,j
i,j

τ ∈Sn

sgn(τ )c1,τ (1) · · · cn,τ (n)
sgn(τ )(c1,τ (1) + tc1,τ (1) ) · · · (cn,τ (n) + tcn,τ (n) )

=
τ ∈Sn

n
sgn(τ )c1,τ (1) · · · cn,τ (n) + t

=
τ ∈Sn

sgn(τ )c1,τ (1) · · · cj,τ (j) · · · cn,τ (n)
j=1 τ ∈Sn

n
+t
= det ci,j
i,j

sgn(τ )c1,τ (1) · · · cj,τ (j) · · · cn,τ (n)
j=1 τ ∈Sn

is a unit in A = A ⊕ At, since det ci,j
is a unit in A. Hence ψ is an automorphism.
i,j
Thus the action given by ψ ◦ ρ ◦ ψ −1 : G → Aut A [[u1 , . . . , un ]] is equivalent to that of ρ
and, by construction, we have that D(φ )(ψ ◦ ρ ◦ ψ −1 ) = ρ. We now use the representative
ψ ◦ ρ ◦ ψ −1 for ε .
Lastly, suppose that the actions ψ ◦ ρ ◦ ψ −1 , ρ , and ρ are determined by the following
automorphisms, respectively:
ui →

n r u +s , u →
i
j=1 i,j j
i

n r u +s , u →
i
j=1 i,j j
i

n r u +s .
i
j=1 i,j j

with ri,j ∈ A , si ∈ A [[u1 , . . . , un ]], ri,j ∈ A , si ∈ A [[u1 , . . . , un ]], ri,j ∈ A, and
si ∈ A[[u1 , . . . , un ]]. Define ρ : G → Aut A ×A A [[u1 , . . . , un ]] by sending σ to the
automorphism
ui →

n
j=1 ri,j , ri,j uj + si , si .

Note that the determinant of interest for this map will be a unit because the corresponding
determinants in A and A are units and since multiplication/addition is done componentwise in fiber products. The equivalence class of this automorphism will map to (ε , ε ) under
16

f and so the (H1 ) condition is verified.
(H2 ) :

A sketch of the proof of this criterion can be found at [15, p. 390]. For each

ρ ∈ E = Hom(G, Aut A [[u1 , . . . , un ]]), define

C(ρ ) = {ψ ∈ Ker(Aut A [[u1 , . . . , un ]] → Aut k[[u1 , . . . , un ]]) | ψρ (σ) = ρ (σ)ψ}.

We can similarly define C(ρ) for ρ ∈ E = Hom(G, Aut A[[u1 , . . . , un ]]), where ρ is the image
of ρ under φ . For ease of reference, set G0 = Ker(Aut A[[u1 , . . . , un ]] → Aut k[[u1 , . . . , un ]]),
G1 = Ker(Aut A [[u1 , . . . , un ]] → Aut k[[u1 , . . . , un ]]), G2 = Ker(Aut A [[u1 , . . . , un ]] →
Aut k[[u1 , . . . , un ]]), and G3 = Ker(Aut A ×A A [[u1 , . . . , un ]] → Aut k[[u1 , . . . , un ]]). We
will need the following lemma in order to verify (H2 ).
LEMMA 3. If C(ρ ) → C(ρ) is surjective for all ρ ∈ E , then the map f is injective.
Proof. Suppose that ρ1 , ρ2 ∈ E = Hom(G, Aut A ×A A [[u1 , . . . , un ]]) are such that
f ([ρ1 ]) = f ([ρ2 ]). Let f ([ρ1 ]) = [φ1 ] ×[φ ] [φ2 ] and f ([ρ2 ]) = [φ1 ] ×
[φ ]. Since [φ2 ] =
[φ0 ] 2
0
−1
[φ2 ], by definition it follows that there exists ψ2 ∈ G2 such that ψ2 φ2 ψ2 = φ2 . Applying
−1
the map φ to this equation gives a ψ0 ∈ G0 such that ψ0 φ0 ψ0 = φ0 , i.e, ψ0 is the image
of ψ2 under the natural map G2 → G0 . Next [φ1 ] = [φ1 ] implies that there exists ψ1 ∈ G1
−1
such that ψ1 φ1 ψ1 = φ1 . Applying the map φ to this last equation gives a ψ0 ∈ G0 such
that ψ0 φ0 ψ0 −1 = φ0 ⇒ φ0 = ψ0 −1 φ0 ψ0 . Thus ψ0 −1 ψ0 φ0 (ψ0 −1 ψ0 )−1 = φ0 and so
ψ0 −1 ψ0 ∈ C(φ0 ). So, by assumption, there exists β2 ∈ C(φ2 ) such that β2 → ψ0 −1 ψ0
−1
−1
via C(φ2 ) → C(φ0 ). Then note that ψ2 = ψ2 β2 ∈ G2 is such that ψ2 φ2 ψ2
=
−1
−1
−1
ψ2 β2 φ2 β2 ψ2 = ψ2 φ2 ψ2 = φ2 , using that β2 ∈ C(φ2 ). Also, when we reduce ψ2 via
−1
A → A, we get ψ0 (ψ0 −1 ψ0 )−1 = ψ0 . Therefore, ψ2 φ2 ψ2
reduces to ψ0 φ0 ψ0 −1 = φ0
on the level of A. Next we define the map g = ψ1 × ψ2 , which can be viewed as an
ψ0
17

element of G3 . This map is well-defined on the fiber product and will have unit determinant
of its linear coefficients because this is also true for both ψ1 and ψ2 . Then, by construction,
we have that gρ1 g −1 = ρ2 . Thus [ρ1 ] = [ρ2 ] and so f is injective.
The (H2 ) condition is that f is a bijective when A

= k[ ] and A = k. By the already

proved (H1 ), we have surjectivity of f and so we only have to show injectivity. However,
since C(ρ ) → C(ρ) is clearly surjective when A

= k[ ] and A = k, this follows from

Lemma 3.
(H3 ) :

Since D(k[ ]) ∼ H 1 (G, Θ) by Proposition 1, it suffices to show that H 1 (G, Θ) is a
=

finite k-vector space. As in [19, p. 622], start by noting that H 1 (G, Θ) is finitely generated as
an R = S G -module. Thus H 1 (G, Θ) is a coherent sheaf on SpecR. We now show that this
sheaf is supported at the closed point of SpecR. Namely, that H 1 (G, Θ)P = H 1 (G, ΘP ) = 0
for all P ∈ SpecR with P = mR . Consider the diagram
S ⊗R RP

.
..
.....................
....................
.
.

.
.
...
.
....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

RP

.
..
...............................
..............................
.
.

S
.
.
...
.
....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

R

Here ΘP is an S ⊗R RP -module. Since the action of G on S is free off the closed point,
d
by the descent result stated in section 2.1, ΘP ∼ S ⊗R RP for some d and S ⊗R RP is a
=
G-torsor. Thus (S ⊗R RP ) ⊗R (S ⊗R RP ) ∼ S ⊗R RP [G] and so
=
P

18

H 1 (G, S ⊗R RP ) ⊗R (S ⊗R RP ) ∼ H 1 (G, (S ⊗R RP ) ⊗R (S ⊗R RP ))
=
P
P
∼ H 1 (G, S ⊗ R [G])
=
R P
∼ (0).
=

Since S ⊗R RP is faithfully flat over RP , this implies that H 1 (G, S ⊗R RP ) = 0 by [14,
d =
p. 47]. Therefore, H 1 (G, ΘP ) = H 1 (G, S ⊗R RP ) ∼ H 1 (G, S ⊗R RP )d ∼ (0), as desired.
=
Therefore, H 1 (G, Θ) is a vector space over k. It contains a copy of k, since R → S is totally
ramified at mS and so σ − id ≡ 0 modulo mS . Hence H 1 (G, Θ) is also finite-dimensional.
Since (H1) − (H3) are satisfied, D has a prorepresentable hull.

5

.

Computation of the Hull

As stated earlier, we have fixed a base action ρ : G → Autk (k[[u1 , . . . , un ]]). Peskin shows
in [20, p. 77] that such an action can be transformed by a change of coordinates so that it
consists of blocks of the form

σui = ui + fi (u1 , . . . , un )
σui+1 = ui+1 + ui
.
.
.
σui+j = ui+j + ui+j−1 ,
where fi has order ≥ 2 and j + 1 ≤ p.
19

(3)

5.1

The Single Jordan Block Case

We will focus on the case where the linear terms of the G-action consist of a single Jordan
block in its Jordan form and fi is fixed by the action. Namely, we assume that ρ is given by

σu1 = u1 + f (u1 , . . . , un )
σu2 = u2 + u1

(4)

.
.
.
σun = un + un−1 ,

where ord(f ) ≥ 2 and f is invariant. Under the further assumption n = p − 1, Peskin shows
[20, p. 88] that there is a coordinate change so that f can be realized as f = (N up−1 )s , a
power of the norm of up−1 for some s ≥ 1. In this case, note that this means that the norm
N up−1 and the action will be defined recursively. To simplify notation, we will denote the
norm N up−1 by y.
As in Peskin’s paper, we will use the following clever trick to simplify computations.
By introducing an extra variable u0 and taking an appropriate “slice”, we can produce a
linear model for our action. Particularly, begin with an action G → Autk (k[[u1 , . . . , un ]])
in the form (4) above. Consider f ∈ k[[u1 , . . . , un ]] as an element of k[[u0 , u1 , . . . , un ]]. The
power series ring k[[u0 , u1 , . . . , un ]] has a G-action given by σu0 = u0 , σui = ui + ui−1 for
1 ≤ i ≤ n. We can then define a surjective G-equivariant map ψ by

k[[u0 , u1 , . . . , un ]] → k[[u1 , . . . , un ]]
u0 → f
ui → ui
20

for i ≥ 1.

The kernel of ψ is generated by the element u0 − f . Thus there is a G-isomorphism
k[[u0 , u1 , . . . , un ]]/(u0 − f ) ∼ k[[u1 , . . . , un ]], where the action on the quotient is given
=
by the same formulas in (5) above. We will use k[[u0 , u1 , . . . , un ]]/(u0 − f ) to do our future
computations. Henceforth, set S = k[[u0 , u1 , . . . , un ]] and R = k[[u0 , u1 , . . . , un ]]/(u0 − f ),
with the actions of G given above. Our first goal is to compute H 1 (G, ΘR ) and H 2 (G, ΘR ):
since these groups give the tangent space and the obstructions to infinitesimal local deformations, respectively.
Begin by noting that ΘS =

Ω1
S/k

∗
= HomS

Ω1 , S
S/k

and ΘR =

Ω1
R/k

∗
. By

basic algebra, we have the following short exact sequence

γ
I/I 2 → Ω1 ⊗S R → Ω1
S/k
R/k → 0,
where I = (u0 − f ). In [20, p. 82], it’s explained that when n = p − 1 the ring S G is
factorial and there exists a unit s ∈ S such that I is generated by the invariant element
(u0 − f )s ∈ S G . Thus I/I 2 is isomorphic to R as G-modules. The map γ : R → Ω1 ⊗S R
S/k
is given by r → rd(u0 − f ) and is clearly injective. Thus the above sequence of G-modules is
also left exact. Taking the dual of this short exact sequence gives the short exact sequence

0 → ΘR → ΘS ⊗S R → R → 0.

We will need the following results.
LEMMA 4. H i (G, ΘS ) = 0 for i > 0.
Proof. The key to the proof of this fact is that ΘS ∼ k[G]⊗k S as G-modules, so let’s first
= 

n
n
n
∂ 
∂
∂ ∼
k
⊗k S. Set V =
k
. Note
prove this. First, we have ΘS =
S
=
∂ui
∂ui
∂ui
i=0
i=0
i=0
21

that V =

∗
m/m2 , where m = (u0 , . . . , un ) is the maximal ideal of k[[u0 , u1 , . . . , un ]].

The action of G on m/m2 is given by σu0 = u0 , σui = ui + ui−1 for 1 ≤ i ≤ n. Define a
k-linear map ψ : m/m2 → k[G] by ui → (σ − 1)n−i and note that this is clearly bijective.
ψ is G-equivariant because

ψ σui = ψ ui + ui−1
= (σ − 1)n−i + (σ − 1)n−i+1
= (σ − 1)n−i (1 + σ − 1)
= σ (σ − 1)n−i = σψ ui

for 0 ≤ i ≤ n. Thus ψ is an isomorphism of G-modules. Since k[G] is self-dual as a Gmodule, it follows that V = (m/m2 )∗ ∼ (k[G])∗ ∼ k[G]. Hence, we get that ΘS ∼ k[G]⊗k S
=
=
=
where the action on this latter module is the diagonal action. We can change this diagonal
action to an action on the left factor only using Frobenius reciprocity. Particularly, we have
∼
k[G] ⊗k S → k[G] ⊗k S by g ⊗ s → g ⊗ g −1 s. Hence we have ΘS ∼ k[G] ⊗k S with G acting
=
on the left factor of this tensor product. As a result, for i > 0, we get

H i G, ΘS ∼ H i G, k[G] ⊗k S ∼ H i (G, k[G]) ⊗k S = 0.
=
=

As a result of this, we can now prove:
LEMMMA 5. H i (G, ΘS ⊗S R) = 0 for i > 0.
Proof. Since ΘS ⊗S R ∼ ΘS /(u0 − f ), we can form the short exact sequence
=

0 → ΘS → ΘS → ΘS ⊗S R → 0.
22

The map ΘS → ΘS is given by multiplication by u0 − f . Now if we consider the corresponding long exact sequence of cohomology (as explained in [22, p. 111])

0 → ΘS G → ΘS G → ΘS ⊗S R G
→ H 1 (G, ΘS ) → H 1 (G, ΘS ) → H 1 (G, ΘS ⊗S R)
→ H 2 (G, ΘS ) → H 2 (G, ΘS ) → H 2 (G, ΘS ⊗S R)
→ H 3 (G, ΘS ) → · · ·

and use that H i (G, ΘS ) = 0 for i > 0 by Lemma 4, it follows that H i (G, ΘS ⊗S R) = 0 for
i > 0.
Using these results, we can now say something about H i (G, ΘR ) for i = 1, 2.
THEOREM 6. Suppose that the G-action ρ : G → Autk (k[[u1 , . . . , un ]]) with n = p − 1
is such that the linear terms form a single Jordan block when in its Jordan form and f is
invariant when the action is put in the form (4) by [20, p. 77]. Then H 1 (G, ΘR ) ∼ H 0 (G, R)
=
and H 2 (G, ΘR ) ∼ H 1 (G, R).
=
Proof. The short exact sequence 0 → ΘR → ΘS ⊗S R → R → 0 gives us the following
long exact sequence (as in [22, p. 128])

· · · → H 0 (G, ΘR ) → H 0 (G, ΘS ⊗S R) → H 0 (G, R)
→ H 1 (G, ΘR ) → H 1 (G, ΘS ⊗S R) → H 1 (G, R)
→ H 2 (G, ΘR ) → H 2 (G, ΘS ⊗S R) → H 2 (G, R)
→ H 3 (G, ΘR ) → H 3 (G, ΘS ⊗S R) → · · · .

23

Applying Lemma 5 to this sequence, it follows that H 3 (G, ΘR ) ∼ H 2 (G, R) and H 2 (G, ΘR ) ∼
=
=
H 1 (G, R). Since G is cyclic, it is a standard fact that the group cohomology will be 2periodic. Thus H 1 (G, ΘR ) ∼ H 3 (G, ΘR ) ∼ H 2 (G, R) ∼ H 0 (G, R). So the desired result
=
=
=
holds.

5.2

The Case p = 3

We now assume that the characteristic of k is p = 3 and the action ρ : G → Autk (k[[u1 , u2 ]])
is free off the closed point. By our comments in the last section and [20, p. 88], there is an
s ≥ 1 such that after a change of coordinates the action ρ can be put in the form
σu1 = u1 + y s

(5)

σu2 = u2 + u1
with y = N u2 and this action defined recursively. Further, we can take advantage of a
result of Peskin in [20, p. 96] that states that the invariant ring for this G-action on R =
k[[u0 , u1 , u2 ]]/(u0 − y s ) is

RG = k[[x, y, z]]/(z 3 + y 2s z 2 − y 3s+1 − x2 ),

where x = N u1 , y = N u2 , and z = u2 − y s u1 + y s u2 . In general positive characteristics,
1
it is very difficult to compute this invariant ring or to do the other computations necessary
to explicitly compute the cohomology groups H 1 (G, ΘR ) and H 2 (G, ΘR ). However, with
this result in hand, we can prove the following.

24

THEOREM 7. Suppose char(k) = 3 and the action of G on k[[u1 , u2 ]] is free off the
closed point. Then H 1 (G, ΘR ) ∼ k[y]/(y s ).
=
Proof. It follows from Theorem 6 and standard results of group cohomology that

H 1 (G, ΘR ) ∼ H 0 (G, R) = RG /Im(T r).
=

Therefore, to prove the desired result we need to compute the image of the trace map
T r : R → RG . As explained in [20, p. 94], R is a free R = k[[x, y]]-module of rank 9
with basis

j
ui u2
1

0≤i,j≤2

. Thus to compute Im(T r) we compute the image of these ba-

sis elements under T r as power series in x, y, and z. Particularly, T r(1) = T r(u1 ) = 0,
T r(u2 ) = y s , T r(u2 ) = T r(u1 u2 ) = −y 2s , T r(u2 ) = y 2s − z, T r(u2 u2 ) = y 3s − y s z,
1
2
1
T r(u1 u2 ) = −y 3s − x + y s z, and T r(u2 u2 ) = y 4s − z 2 − y s x − y 2s z. Thus Im(T r) is
2
1 2
generated by x, y s , z, and z 2 as an R -module. Since RG is of rank 3 as an R -module,
generated by 1, z, and z 2 , it follows that H 1 (G, ΘR ) ∼ k[y]/(y s ).
=

Consider the following action of G on k[ε][[u1 , . . . , un ]] that lifts the base action ρ:

σu1 = u1 + y s + εy k
σu2 = u2 + u1 ,

where both this action and the norm y = N u2 are defined recursively here. A simple
computation shows that y = y + εy for some power series y ∈ k[[u1 , u2 ]] and that σy = y.
∼
Under the bijection D(k[ε]) − H 1 (G, Θ) from Proposition 1 we have
→
25

σ(σ −1 (u1 )) − u1
ε
σ(u1 − y s ) − u1
=
ε
s + εy k − y s − u
u +y
1
= 1
ε
= yk

dρ σ (u1 ) =

and
σ(σ −1 (u2 )) − u2
dρ σ (u2 ) =
ε
σ(u2 − u1 + y s ) − u2
=
ε
u + u1 − u1 − y s − εy k + y s − u2
= 2
ε
= −y k .

So dρ σ = y k ∂ − y k ∂ . Thus the following correspondence holds:
∂u1
∂u2


 σu = u + y s + εf (y)
1
1

 σu = u + u
2
2
1





←→




f (y)

∂
∂
− f (y)
∂u1
∂u2

,

∂ − ∂ , y ∂ − y ∂ , . . .,
∂u1
∂u2 ∂u1
∂u2
are linearly independent over k and H 1 (G, ΘR ) ∼ k[y]/(y s ) is s=

where f (y) ∈ k[y] is a polynomial of degree < s. Since

y s−1 ∂ − y s−1 ∂
∂u1
∂u2
dimensional by Theorem 6, we’re led to consider deformations of ρ to A ∈ C given by

26

σu1 = u1 + y s + a0 + a1 y + · · · + as−1 y s−1

(6)

σu2 = u2 + u1

where the ai lie in the maximal ideal mA of A and this action and the element y = N u2 are
defined recursively. As long as A is of characteristic 3, we note that such a map does indeed
give a local deformation because it is assumed to be an A-linear map and y is invariant
by definition: thus a simple calculation shows that σ 3 = id. Further, we know that the
deformations of this form will cover the tangent space. We can now get some information
about the prorepresentable hull of the local deformation functor D = Dρ : C → Sets. In fact,
if we restrict D to the subcategory C3 = {A ∈ C | char(A) = 3} of C, then we can compute
the prorepresentable hull of the restriction D|C .
3
THEOREM 8. Suppose char(k) = 3 and the action ρ : G → Autk (k[[u1 , u2 ]]) is free
off the closed point. Then the hull of the deformation functor D|C is k[[x0 , x1 , . . . , xs−1 ]],
3
where s ≥ 1 is the integer given by [20, p. 88] as explained above.
Proof. By Propositions 1 and Theorem 6, tR ∼ tD . So it remains to show that hR → D
=
is smooth, i.e., f : Hom(R, A ) → Hom(R, A) ×D(A) D(A ) is surjective for any small
extension φ : A → A in C3 . Begin with a homomorphism g : R → A that induces a
deformation [ξ] ∈ D(A). ξ is given by (7), where a0 , a1 , . . . , as−1 are the images of
x0 , x1 , . . . , xs−1 under g. Suppose that [ξ] ∈ D(A ) is a lifting of the deformation [ξ] to
A . Take any lifts a0 , a1 , . . . , as−1 ∈ A of a0 , a1 , . . . , as−1 via φ. Thus a0 = a0 + ta0 ,
a1 = a1 + ta1 , . . . as−1 = as−1 + tas−1 , where J = (t) is the kernel of φ. We lift g to a
homomorphism g ∈ Hom(R, A ) by sending xi → ai . It remains to show that g gets sent
to the desired element under f .

27

As explained in [21, p. 213], by Schlessinger’s (H2 ) condition for D and the isomorphism
A ×

A /J

∼
A = A ×A A − A ×k k[J], (a, b) → (a, ao + b − a) where a, b ∈ A and ao is
→

the k-residue of a, we obtain a map

D(A ) × (tD ⊗ J) → D(A ) ×D(A) D(A )
This map determines an action of tD ⊗ J on the fibers D(φ)−1 (η) for each η ∈ D(A).
This action is transitive by (H1 ). Similarly, we have another transitive action of tR ⊗ J
on each fiber hR (φ)−1 (b), for b ∈ hR (A), that is compatible with the previous action via
G : Hom(R, A ) → D(A ). Since the image of g under G is in D(φ)−1 ([ξ]), there exists an
element w ∈ tD ⊗ J such that wG(g ) = [ξ]. Thus the element w ∈ tR ⊗ J corresponding
∼
to w under the natural isomorphism tR ⊗ J − tD ⊗ J is such that the image of w g
→
under Hom(R, A ) → D(A ) is [ξ]. However, the image of w g under hR (φ) is still g: since
g ∈ hR (φ)−1 (g) and tR ⊗ J acts on hR (φ)−1 (g). Therefore, w g maps to the desired
element (g, [ξ]) and so f is surjective.

6

An Example

One problem in the area of deformation theory that has garnered a lot of interest is that of
lifting certain wild actions to characteristic zero. In the local setting where our results hold,
the question is whether or not the base action ρ : G → Autk (k[[u, v]]) can be lifted from k
to a ring of characteristic zero. One may also want to explore in what ways local actions can
be realized from global actions on smooth projective varieties.
Let X ⊂ P3 be the Fermat quartic given by the equation x4 + x4 + x4 + x4 = 0. Suppose
0
1
2
3
k
28

that the group G = σ ∼ Z/3Z acts on X via the automorphism σ(x0 , x1 , x2 , x3 ) =
=
(x0 , x2 , x3 , x1 ). This action was considered in a different context in [8, p. 114]. Note that
σ(x0 , x1 , x2 , x3 ) = (x0 , x1 , x2 , x3 ) implies that (x0 , x1 , x2 , x3 ) = (x0 , x2 , x3 , x1 )
in P3 . If x0 = 0, we can scale these tuples by 1/x0 and conclude that x1 = x2 = x3 since
k
x 4
we are then working in an open affine subset of P3 . So x4 + 3x4 = 0 =⇒ 3 x1
= −1. If
0
1
k
0
x
char(k) = 3, we arrive at a contradiction. If char(k) = 0, we find four solutions for x1 and
0
hence four fixed points of the action. Now suppose that x0 = 0. There must exist c ∈ k ∗
such that x1 = cx2 , x2 = cx3 , and x3 = cx1 . We can further assume that x1 = 0, since
x1 = 0 would imply that x0 = x1 = x2 = x3 = 0. The equation x4 + x4 + x4 + x4 = 0
0
1
2
3
then becomes (c8 + c4 + 1)x4 = 0, which gives that c8 + c4 + 1 = 0. The relations x1 = cx2 ,
1
x2 = cx3 , and x3 = cx1 also yield that x1 = c3 x1 =⇒ (c3 − 1)x1 = 0 =⇒ c3 = 1. If
char(k) = 3, this implies that c = 1 and thus (0, 1, 1, 1) is the only fixed point. However, if
char(k) = 0, then c8 + c4 + 1 = 0 and c3 = 1 imply that c2 + c + 1 = 0. Thus, we obtain
two more fixed points in the characteristic zero case, giving us a total of six fixed points.
We henceforth assume that char(k) = 3. If we consider the induced action on OX,x ∼
=
k[[u, v]], the completion of the stalk at the fixed point x = (x0 , x1 − x2 , x2 − x3 ), we get an
action of the type we have thus far been studying. By the results of Peskin [20, p. 88], this
action must be of the form

σu = u + (N v)s
σv = v + u

where s ≥ 1 and both the action and N v are defined recursively. Our goal is to determine
the value of s and thus obtain a specific example where the actions we have been studying
29

can be lifted to characteristic zero: and also an example where our local actions arise from
a global action on a surface. To get a handle on the action, let’s first consider the action on
k[x,y,z,y −1 ,z −1 ]
the open affine set U = {x1 = 0} {x2 = 0} {x3 = 0} = Spec
where
1+x4 +y 4 +z 4
x
x
x
x = x0 , y = x2 , and z = x3 . The action here is given by
1
1
1
x
σ: x= 0 →
x1
x
y= 2 →
x1
x
z= 3 →
x1
x
y −1 = 1
x2
x
z −1 = 1
x3

x0
= xy −1
x2
x3
= zy −1
x2
x1
= y −1
x2
x
→ 2 = yz −1
x3
x
→ 2 = y.
x1

Next consider the stalk at m = (x, y−1, z−1) and take the completion. In the completion
we note that the ideal m is generated by just the elements u = y − 1 and v = z − 1: since
1 + x4 + y 4 + z 4 = 0 =⇒ x4 = −y 4 − z 4 − 1 and we can solve this equation for x in terms of
u and v using the Binomial Theorem for rational exponents, which is allowed since we are
working in a power series ring once we take the completion. Therefore, the completed stalk
with the induced action is isomorphic to k[[u, v]] with the action

σ : u = y − 1 → zy −1 − 1 =

v−u
= (v − u)
1+u

−u
v = z − 1 → y −1 − 1 =
=
1+u

30

∞

∞

(−u)i
i=0

(−u)i+1 .
i=0

Proceeding as suggested by Peskin’s work, we make the change of coordinates u = u + v,
v = u to put the linear terms of this action in Jordan form:

σ : u → u − uv − v 2 + (u + v + 1)

∞

u
(−v)i =
1+v
i=2

∞

v → u + v + (u + v)

u+v
.
(−v)i =
1+v
i=1

Lastly, make a final change of coordinates v = v, u = (σ − 1)v to put the action in the
form:

σu = u + f
σv = u + v,

where f = (σ − 1)2 v = T r(v) and so is clearly invariant.
Using that v = v and u = (σ − 1)v = u+v − v =⇒ u = (u + v)(1 + v) − v, we can find a
1+v
closed form for f :
f = (σ − 1)2 v
= (σ 2 + σ + 1)v
v−u
u+v
+
+v
1+u−v 1+v
v − (u + v)(1 + v) + v
(u + v)(1 + v)
+
+v
=
1 + (u + v)(1 + v) − v − v
1+v
u2 + u2 v − v 3
=
.
(1 + v)(1 + u + v)
=

31

By Theorem 7, we know that the dimension of H 1 (G, Θ) as a k-vector space is s. Thus
we can achieve our goal by computing H 1 (G, Θ). By Theorem 6, H 1 (G, Θ) ∼ RG /Im(T r).
=
The results in [20, p. 96] show that
RG =

k[[x, y, z, f ]]
3 + y 2s z 2 − y 3s+1 − x2 )
(z

where x = Norm(u), y = Norm(v), and z = u2 − uf + vf . The same calculations done in
Theorem 7 to compute Im(T r) show that Im(T r) = f RG + zRG + xRG . Next

f=

u2 + u2 v − v 3
(1 + v)(1 + u + v)

= (u2 + u2 v − v 3 )(1 − u + v + · · · )
= u2 − u2 v − v 3 − u3 − u3 v − uv 3 + u2 v 2 − v 4 + · · ·
= z − y + h(x, y, z, f )

where ord(h) ≥ 2. Thus, since f ≡ z ≡ 0 in the quotient RG /Im(T r), we have 0 ≡ f =
z − y + h(x, y, z, f ) ≡ −y + h(x, y, z, f ). Hence y ≡ h(x, y, z, f ) ≡ g(y) for a power series
g with ord(g) ≥ 2, since all the terms of h containing x, z, or f are ≡ 0 in the quotient.
Iterating this equivalence, we have that y ≡ (g ◦ g ◦ · · · ◦ g )(y) for any l and so it follows
l
that y ≡ 0. Thus Im(T r) also contains y. Therefore, H 1 (G, Θ) ∼ k and so s = 1.
=

32

BIBLIOGRAPHY

33

BIBLIOGRAPHY
[1] J. Bertin, Compactification des sch´mas de Hurwitz, C. R. Acad. Sci. Paris S´r I Math.
e
e
332 (1996), no. 11, 1063-1066.
[2] J. Bertin and A. M´zard, D´formations formelles des revˆtements sauvagement ramifi´s
e
e
e
e
de courbes alg´briques, Invent. Math. 141 (2000), no. 1, 195-238.
e
[3] H.E.A. Campbell, B. Fodden, D.L. Wehlau, Invariants of the diagonal Cp -action on V3 ,
J. Algebra 303 (2006), no. 2, 501-513.
[4] H.E.A. Campbell and I.P. Hughes, Vector invariants of U2 (Fp ): a proof of a conjecture
of David Richman, Adv. Math. 126 (1997), no. 1, 1-20.
[5] T. Chinburg, R. Guralnick, D. Harbater, Oort groups and lifting problems. Compos.
Math. 144 (2008), no. 4, 849-866.
[6] G. Cornelissen and F. Kato, Equivariant deformation of Mumford curves and of ordinary
curves in positive characteristic, Duke Math. J. 116 (2003), no. 3, 431-470.
[7] H. Derksen and G. Kemper, Computational Invariant Theory, Invariant Theory and
Algebraic Transformation Groups, I, Springer-Verlag, Berlin, 2002, EMS 130.
[8] I. Dolgachev and J. Keum, Wild p-cyclic action on K3-surfaces, J. Algebraic Geom. 10
(2001), no. 1, 101-131.
[9] B. Edixhoven, N´ron models and tame ramification, Compositio Math. 81 (1992), no.
e
3, 291-306.
[10] B. Green and M. Matignon, Liftings of Galois covers of smooth curves. Compositio
Math. 113 (1998), no. 3, 237-272.
34

[11] A. Grothendieck, Revˆtements ´tales et groupe fondamental. SGA1, Lecture Notes in
e
e
Math. 224 (1971).
[12] H. Ito and S. Schr¨er, Wildly ramified actions and surfaces of general type arising from
o
Artin-Schreier curves. arXiv:1103.0088v1[math.AG] 1 Mar 2011.
[13] N. Katz and B. Mazur, Arithmetic moduli of elliptic curves, Annals of Mathematics
Studies, vol. 108, Princeton University Press, Princeton, N.J., 1985.
[14] H. Matsumura, Commutative Ring Theory, second ed., Cambridge Studies in Advanced
Mathematics, vol. 8, Cambridge University Press, Cambridge, 1989, Translated from
the Japanese by M. Reid.
[15] B. Mazur. Deforming Galois representations, Galois groups over Q (Berkeley, CA, 1987),
Math. Sci. Res. Inst. Publ., vol. 16, Springer, New York, 1989, 385437.
´
[16] J. Milne, Etale cohomology, Princeton Mathematical Series, vol. 33, Princeton University
Press, Princeton, N.J., 1980.
[17] F. Oort, Liftings of algebraic curves, abelian varieties and their endomorphisms to characteristic zero. Proceedings in Symposia in Pure Mathematics 46 (1987), 167-195.
[18] F. Oort, T. Sekiguchi, N. Suwa, On the deformation of Artin-Schreier to Kummer. Ann.
´
Sci. Ecole Norm. Sup. (4) 22 (1989), no. 3, 345-375.
[19] B. Peskin, On rings of invariants with rational singularities, Proc. Amer. Math. Soc.
87 (1983), no. 4, 621-626.
[20] B. Peskin, Quotient-singularities and wild p-cyclic actions, J. Algebra 81 (1983), no. 1,
72-99.
[21] M. Schlessinger, Functors of Artin rings, Trans. Amer. Math. Soc. 130 (1968), 208-222.
[22] J. P. Serre, Local fields, Graduate Texts in Mathematics, vol. 67, Springer-Verlag, New
York, 1979, Translated from the French by Marvin Jay Greenburg.
[23] R.J. Shank, S.A.G.B.I. bases for rings of formal modular semiinvariants, Comment.
Math. Helv. 73 (1998), no. 4, 548-565.
35

[24] R.J. Shank and D.L. Wehlau, Noether numbers for subrepresentations of cyclic groups
of prime order, Bull. London Math. Soc. 34 (2002), no. 4, 438-450.
[25] R. Taylor and A. Wiles, Ring-theoretic properties of certain Hecke algebras. Ann. of
Math. (2) 141 (1995), no. 3, 553-572.
[26] A. Wiles, Modular elliptic curves and Fermat’s last theorem. Ann. of Math. (2) 141
(1995), no. 3, 443-551.

36