ON THE CLASSIFICATION OF GALOIS OBJECTS FOR

FINITE GROUPS
C

ESAR GALINDO AND MANUEL MEDINA

Abstract. We classify Galois objects for the dual of a group algebra
of a nite group over an arbitrary eld.
1. Introduction
Let H be a Hopf algebra. A right H-comodule algebra A is called a right
H-Galois object if the map can: A A A H, a b ab
(0)
b
(1)
is
bijective. In this paper we shall study Galois objects when H is the dual
of a group algebra of a nite group. In this case Galois objects are the G-
Galois extensions of non-commutative rings introduced by Chase, Harrison
and Rosenberg [3].
Let H be a Hopf algebra; H-Galois objects are interesting for several
reasons, e.g., they are in bijective correspondence with ber functors on
the category of nite dimensional H-comodules [17], they are used for the
construction of new Hopf algebras [14], and they are commutative algebras
in the Drinfeld center of the category of H-comodules, which has important
applications in Rational Conformal Field Theory [8, 9, 5].
Let
H
/ be the category of nite dimensional left H-comodules. Recall
that a ber functor F :
H
/ Vec
k
is an exact and faithful monoidal
functor that commutes with arbitrary colimits. Ulbrich dened in [17] a ber
functor F
A
associated with each H-Galois object A, by F
A
(V ) = A
H
V ,
where A
H
V is the cotensor product over H of the right H-comodule A
and the left H-comodule V . In loc. cit. a bijective correspondence between
isomorphism classes of H-Galois objects and isomorphism classes of ber
functors of
H
/ was dened.
Given a Hopf algebra H and a left H-Galois object A, one can associate
a new Hopf algebra L(A, H), see [14], which is the Tannaka-Krein recons-
truction from the ber functor associated to A, [14, Theorem 5.5].
Let A B be an H-Galois extension. The Miyashita-Ulbrich action of H
on the centralizer A
B
of B in A, makes A
B
into a commutative algebra in
2000 Mathematics Subject Classication. 16W30, 20C05.
Key words and phrases. Hopf-Galois objects, Hopf algebras, G-algebras.
This work was partially supported by the Project 003568 of the Department of Mathe-
matics, Ponticia Universidad Javeriana.
1
2 C

ESAR GALINDO AND MANUEL MEDINA

the category }T
H
H
of right Yetter-Drinfeld modules, and }T
H
H
is equivalent
to the Drinfeld center of the tensor category
H
/ (see [12, Chapter 10]). If
H is a split semisimple and cosemisimple Hopf algebra, the category }T
H
H
is modular and each Galois object denes a modular invariant [5].
Throughout this article we work over an arbitrary eld k, and we will
denote by k
G
the dual of the group algebra for a nite group G.
It is well known that the classication of H

-Galois objects for a nite

dimensional Hopf algebra H is equivalent to the classication of twists in
H, see [1, Section 4]. In [13], Movshev classied twists for complex group
algebras, same get a classication of Galois objects for the dual of a complex
group algebra of a nite group.
The aim of this paper is to classify k
G
-Galois objects, generalizing the
classication of Movshev [13], and Davydov [4].
To formulate our main result we rst dene:
Denition 1.1. AGalois datum associated to k
G
is a collection (S, K, N, , )
such that
i) S is a subgroup of G and N is a normal subgroup of S.
ii) K k is a Galois extension with Galois group S/N.
iii) char(k) [N[.
iv) : N N K

is a non-degenerate 2-cocycle.
v) : S N K

satises the equations (C1), (C2), and (C3) in Propo-

sition 4.7.
A 2-cocycle Z
2
(N, K

) is called non-degenerate if (s, t) = (t, s) for

all t C
N
(s) implies s = 1. Equivalently, is non-degenerate if the center
of the twisted group algebra K

N coincides with K. Also, by Maschkes

theorem for twisted group algebras [10, Theorem 2.10, pag 85], if char(k)
[N[ and is non-degenerate then K

N is a central simple algebra over K.

The function is uniquely determined by a Hochschild 1-cocycle

HZ
1
(S, C
1
(N, K

)), where C
1
(N, K

) has an S-bimodule structure given

by the equations (11), and (10), see Section 7. In Section 7 we give some
cohomological obstructions which establish necessary conditions for the exis-
tence of .
Let (S, K, N, , ) be a Galois datum associated to k
G
. The group S acts
on K through the canonical projection to Gal(K[k) = S/N. We will denote
by A(K

N, ) the twisted group algebra K

N with S-action dened by

g u
x
= g()(g, x)ug
x
,
for g S, x N, and K.
In section 3.1, we study the induction functor from the category of S-
algebras to the category of G-algebras. Using the induction functor we
dene the G-algebra Ind
G
S
(A(K

N, )).
ON THE CLASSIFICATION OF GALOIS OBJECTS FOR FINITE GROUPS 3
Now we can formulate our main result.
Theorem 1.2. Let G be a nite group and let k be a eld.
i) Let (S, K, N, , ) be a Galois datum associated to k
G
. Then the G-
algebra Ind
G
S
(A(K

N, )) is a k
G
-Galois object.
ii) Let A be a k
G
-Galois object. Then A Ind
G
S
(A(K

N, )) for a Galois
datum (S, K, N, , ).
Remark 1.3. In [5, Theorem 3.5.1], Davydov classied in terms of group-
theoretical datum, indecomposable commutative algebras in the Drinfeld
center of the category of nite dimensional complex representations of a
nite group G. As explained above, each k
G
-Galois object denes a com-
mutative algebra in the Drinfeld center of Rep
k
(G), so there is a connec-
tion between the Galois data and the data appeared in loc. cit.. Indeed, if
(S, K, N, , ) is a Galois datum associated to k
G
such that (x, y), (s, x)
k for all x, y N, s S, then the conditions (C1), (C2), and (C3) are the
same as the conditions (8), (9), and (10) in [5, Proposition 3.4.2].
In section 5.1, we dene an equivalence relation on the set of all Galois
data so that equivalent data dene isomorphic Galois objects.
The present paper is organized as follows: in Section 2 we give a brief
review of the results on Hopf-Galois extensions that we shall need. In section
3 we show that every k
G
-Galois object is the induced from a simple Galois
object. In order to prove this result we rst dene the concept of imprimitive
system for G-algebras. In section 4 we show that every simple k
G
-Galois
object is a twisted group algebra. In section 5 we dene k
G
-Galois objects
associated to some group-theoretical data. In section 6 we prove our main
result, Theorem 1.2 and we give some non-trivial examples of Galois objects.
In Section 7 we present some cohomological obstructions for the existence
of a function , see Denition 1.1.
Acknowledgement. We thank Erik Backelin and the referee for very help-
ful remarks on this paper.
2. Preliminaries on Hopf Galois extensions
In this section we review some results on Hopf Galois extensions that
we will be needed later. We refer the reader to [15] for a more detailed
exposition.
Denition 2.1. Let H be a Hopf algebra. Let also A be a right H-comodule
algebra with structure map : A A H, (a) = a
(0)
a
(1)
, and let
B = A
co H
. The extension A B is called a right Hopf Galois extension,
or a right H-Galois extension if the canonical map
can : A
B
A AH, a b ab
(0)
b
(1)
,
is bijective.
4 C

ESAR GALINDO AND MANUEL MEDINA

A right H-Galois extension of the base eld k will be called a right H-
Galois object. Left H-Galois extensions and left H-Galois objects are dened
similarly.
Example 2.2. Let the Hopf algebra H coact on itself through the comulti-
plication . Then the canonical map HH HH, xy xy
(1)
y
(2)
is bijective with inverse x y xS(y
(1)
) y
(2)
.
Assume from now on that H is a nite dimensional Hopf algebra.
We can characterize H

-Galois objects A in terms of the natural H-action

on A:
Proposition 2.3. A is a right H

-Galois object if and only if A is nite

dimensional and the map : A#H End A, (a#h)(b) = a(h b) is an
isomorphism.
Proof. See [12, 8.3.3 Theorem].
Proposition 2.4. Every right H-Galois object is isomorphic to a crossed
product H

= k#

H, where : H H k is an invertible 2-cocycle.

Moreover, if H is semisimple then H

is semisimple.
On the other hand, H

is a right H-Galois object for all such .

Proof. Let A be a right H-Galois object. Since A#H

End A is simple
artinian, it follows from [12, Proposition 8.3.6] that there is : H H k
an invertible 2-cocycle such that
A k#

H.
Now, by [12, Theorem 7.4.2], if H is semisimple and nite dimensional then
A k#

H is semisimple.
Remark 2.5. It follows from Proposition 2.4 that for an H-Galois object A,
dim
k
(A) = dim
k
H.
2.1. Miyashita-Ulbrich action. Let A B be an H-Galois extension.
The Miyashita-Ulbrich action of H on the centralizer A
B
of B in A, makes
A
B
into a commutative algebra in the category }T
H
H
of right Yetter-Drinfeld
modules.
Denition 2.6. Let H be a Hopf algebra, and A an H-Galois extension of
B. The Miyashita-Ulbrich action of H on A
B
is dened by a h = h
[1]
ah
[2]
,
a A
B
, h H, where can(h
[1]
h
[2]
) = 1 h.
The Miyashita-Ulbrich action of H on the H-Galois object A is the unique
map AH A, ah a h, such that ab = b
(0)
(a b
(1)
), for all a, b A.
Example 2.7. Consider the H-Galois object A = H as in Example 2.2.
In this case the Miyashita-Ulbrich action coincides with the right adjoint
action of H on itself: a h = o(h
(1)
)ah
(2)
, a, h H.
Therefore a Hopf subalgebra H

H is a normal Hopf subalgebra if and

only if it is stable under the Miyashita-Ulbrich action.
ON THE CLASSIFICATION OF GALOIS OBJECTS FOR FINITE GROUPS 5
The next theorem will be useful to reduce the classication of H-Galois
object in the case that we have an exact succession k K H Q k
of Hopf algebras.
Theorem 2.8. Let H be a nite dimensional Hopf algebra, K H a nor-
mal Hopf subalgebra, and Q = H/K
+
H the quotient Hopf algebra. Let
: A A H be an H-comodule algebra. Then the following conditions
are equivalent:
i) A is an H-Galois objet.
ii) A
coQ
is a K-Galois object, and A
coQ
A is a faithfully at Q-Galois
extension.
Proof. See [15, Theorem 4.5.1]
In the case that H = k
G
for a nite group G and A is a simple algebra
we have:
Corollary 2.9. Let A be a k
G
-comodule algebra, where A is a simple algebra
with center K. Then A is a k
G
-Galois object if and only if A is a K
N
-Galois
object (where N = x G[x = , K) and K is a k
G/N
-Galois
object.
Proof. Suppose that A is a k
G
-Galois object. By Theorem 2.8, we need to
show that A
N
= K.
Since G acts on A and K = :(A), then we have a homomorphism : G
Aut
k
(K), such that ker = N. Thus induces an injective homomorphism
: G/N Aut
k
(K). Now, A
G
= k implies K
G/N
= k. Hence by [2,
Theorem 13] we have [K : k] [G/N[.
By Theorem 2.8 A
N
is a k
G/N
-Galois object, so dim
k
A
N
= [G/N[. Since
K A
N
, and [K : k] [G/N[, thus A
N
= K.
The converse follows from Theorem 2.8.

3. Galois extensions of finite groups

Let G be a nite group and let A be a G-algebra with G-action kGA
A, g a g a. Equivalently, A is a k
G
-comodule algebra with respect to
the coaction : A Ak
G
dened by
(a) =

gG
g a t
g
,
where (t
g
)
gG
denotes the basis of k
G
consisting of the canonical idempotents
t
g
(h) =
g,h
(Kroneckers delta).
The next proposition is a restatement of [4, Proposition 3.1] when k is a
not necessarily algebraically closed.
Proposition 3.1. Let A be a G-algebra. Then A is a k
G
-Galois object if
and only if A satises the following conditions:
6 C

ESAR GALINDO AND MANUEL MEDINA

i) dim(A) = [G[.
ii) A has no no-trivial G-invariant left ideals.
iii) A
G
= k.
Proof. Let J be a G-invariant left ideal of A. By [15, Theorem 2.3.9] the
map A(A/J)
G
A/J, a m a m is an isomorphism. Now, if J ,= 0,
then A/J = 0, so A = J.
Conversely, we can consider A as a left AG-module with structure map .
By (ii), A is a simple AG-module. Let D = End
AG
(A). Applying the
Jacobson Density Theorem (see [11, F20, pag 139]), we set that the natural
homomorphism
: AG End
D
(A) = End
A
G(A) = End
k
(A)
is surjective. Thus, by (i), it is bijective.
Remark 3.2. The condition A
G
= k in Proposition 3.1 is not necessary if k
is algebraically closed.
The following corollary provides an alternative proof for Proposition 2.4
for the case H = k
G
.
Corollary 3.3. Every k
G
-Galois object is a semisimple algebra.
Proof. Let A be a k
G
-Galois object. Since the Jacobson radical rad(A) of
the algebra A is a left G-invariant ideal, we conclude that rad(A) = 0.
3.1. Imprimitive algebras and induced algebras. In this subsection
we introduce the notion of imprimitive algebra.
Denition 3.4. Let A be a G-algebra and e
1
, . . . , e
n
a set of orthogonal
central idempotents in A, such that 1
A
= e
1
+ + e
n
. We will say that
e
1
, . . . , e
n
is an imprimitive system if g e
i
e
1
, . . . , e
n
for all g
G, i = 1, . . . , n.
A G-algebra is called imprimitive if there is a non-trivial imprimitive
system in A.
Let S be a subgroup of G and (B, ) an S-algebra. The vector space
kG
kS
B has a G-algebra structure given by:
g (h x) = gh x
(g x)(h y) =
_
h (h
1
g x)y if h
1
g S
0 if h
1
g / S,
for g, h G and x, y B
Also, the algebra of functions
A
S
(G, B) = r : G B[r(sg) = s r(g) s S, g G,
is a G-algebra with G-action dened by (g r)(x) = r(xg).
Lemma 3.5. A
S
(G, B) kG
kS
B as G-algebras.
ON THE CLASSIFICATION OF GALOIS OBJECTS FOR FINITE GROUPS 7
Proof. See [6, Proposition 3.3].
Denition 3.6. Let us denote by Ind
G
S
(B) the G-algebra A
S
(G, B)
kG
kS
B and call it the induced algebra from the S-algebra B.
Remark 3.7. Induction is a covariant functor: where to each homomor-
phism of S-algebras f : A B is send to the homomorphism of G-algebras
Ind
G
S
(f) : Ind
G
S
(A) Ind
G
S
(B), Ind
G
S
(f)(r) = f r.
Proposition 3.8. Let A be an imprimitive algebra, such that G acts transi-
tively on the set e
1
, . . . , e
n
. If S is the stabilizer of e
1
then A Ind
G
S
(e
1
A)
as G-algebras.
Proof. Let g
1
, . . . , g
n
be a set of representatives for the left cosets G/S
such that g
i
e
1
= e
i
. Then Ind
G
S
(e
1
A) = kG
kS
e
1
A =

n
i=1
g
i
e
1
A.
Let consider the map
: kG
kS
e
1
A A
n

i=1
g
i
e
1
a
i

n

i=1
e
i
(g
i
a
i
).
It is straightforward to check that the map is a well dened G-algebra
isomorphism.
Lemma 3.9. Let S be a subgroup of G, and (B, ) be an S-algebra. Then
there is an algebra isomorphism between the induced algebra Ind
G
S
(B) and
B B
. .
[G:S]
. Moreover, (Ind
G
S
(B))
G
B
S
.
Proof. Let n = [G : S] and 1
G
, g
2
, . . . , g
n
be a set of representatives for
the left cosets G/S. Let consider the functions e
i
Ind
G
S
(B) dened by
(1) e
i
(x) =
_
1
B
, si x Sg
i
,
0, si x / Sg
i
.
The set e
1
, . . . , e
n
is an imprimitive system in Ind
G
S
(B), thus
Ind
G
S
(B) e
1
Ind
G
S
(B) e
n
Ind
G
S
(B).
It is clear that B e
i
Ind
G
S
(B), so Ind
G
S
(B) B B
. .
[G:S]
.
Now, for the last claim, the map
B
S
(Ind
G
S
(B))
G
b [x b],
is an algebra isomorphism.
Remark 3.10. We will say that the system e
i
, dened in (1), is the cano-
nical imprimitive system associated to Ind
G
S
(B).
8 C

ESAR GALINDO AND MANUEL MEDINA

3.2. Galois Objects as Induced Algebras. In this section we show that
every Galois object of a nite group G is isomorphic to the induced algebra
of a simple Galois object of a subgroup of G (see Theorem 3.14).
Lemma 3.11. Let A be a semisimple G-algebra such that A
G
= k. Then
there is an imprimitive system e
1
, . . . , e
n
in A, such that G acts transi-
tively on this system, and e
i
A is a simple algebra, for all i.
Proof. Applying Wedderburns Theorem for semisimple algebras (see [11,
Theorem 4, pag 157]) we conclude that there is a set e
1
, , e
n
of primitive
orthogonal central idempotents in A such that 1
A
= e
1
+ + e
n
, and e
i
A
is a simple algebra. It is immediate that e
i
is an imprimitive system in
A. Suppose that G does not act transitivity on e
i
. Renumbering if is
necessary, we can suppose that the orbit of e
1
is e
1
, . . . , e
t
and the orbit
of e
t+1
is e
t+1
, . . . , e
r
, where 1 t < r n. Then x
1
= e
1
+ + e
t
and
x
2
= e
t+1
+ +e
r
are linearly independent elements in A
G
. However, this
contradicts the fact that dim
k
A
G
= 1.
Remark 3.12. Suppose k = C and keep the hypothesis of Lemma 3.11. Then,
we have that e
1
A = End
k
(V ) for some complex vector space V . Using the
Skolem-Noether theorem, it is easy to see that every S-action on End
k
(V ),
is dened by a projective representation : S PLG(V ),
(g f)(v) = (g)(f((g)
1
(v))),
for f End
k
(V ), g G, v V . Note that End
k
(V )
S
= k if and only if V is
an irreducible projective representation.
The following result is proved in [4]. We provide an alternative proof.
Proposition 3.13. Let S be a subgroup of G, and (B, ) be an S-algebra.
Then B is a k
S
-Galois object if and only if the induced algebra Ind
G
S
(B) is
a k
G
-Galois object.
Proof. For Lemma 3.9 and Proposition 3.1, it is clear that B
S
= k if and only
if Ind
G
S
(B)
G
= k. Moreover, dim(B) = [S[ if and only if dim(Ind
G
S
(B)) =
[G[.
Let e
1
, . . . , e
n
be the canonical imprimitive system in Ind
G
S
(B), see Re-
mark 3.10.
Suppose that B is a k
S
-Galois object. Let J be a G-invariant left ideal
in Ind
G
S
(B). Then e
1
J is an S-invariant left ideal of e
1
Ind
G
S
(B) B. Since,
g
i
(e
1
J) = e
i
J then J =

n
i=1
e
i
J is a trivial ideal.
Conversely, suppose that Ind
G
S
(B) is a k
G
-Galois object. Let J be an
S-invariant left ideal in B e
1
Ind
G
S
(B), since

J = g
1
J g
n
J is a
G-invariant left ideal in Ind
G
S
(B). Then J is a trivial ideal.
Theorem 3.14. Every k
G
-Galois object is isomorphic as a G-algebra to the
induced algebra Ind
G
S
(B), where B is a simple k
S
-Galois object and S is a
subgroup of G.
ON THE CLASSIFICATION OF GALOIS OBJECTS FOR FINITE GROUPS 9
Proof. Suppose that A is a k
G
-Galois object. By Corollary 3.3 the algebra
A is semisimple, so by Lemma 3.11 and Proposition 3.8 the k
G
-Galois object
A is the induced of a simple Galois object. Thus the theorem follows from
Proposition 3.13.
Remark 3.15. Keep the assumptions in Remark 3.12. By Proposition 3.1 we
see that if V is an irreducible projective representation of the group S, the
S-algebra End
C
(V ), is a C
S
-Galois object if and only dim
C
(V ) = [S[
1/2
. The
condition dim
C
(V ) = [S[
1/2
implies that the 2-cocycle of S, associated to
the projective representation is non-degenerated, see Denition 4.4. Hence,
in the complex case, the Galois objects of a group G are classied by pairs
(S, ), where S G is a subgroup, and Z
2
(S, C

) is a non-degenerated
2-cocycle. This is the main result of [13].
4. Simple Galois objects
In this section we show a structure theorem for simple k
G
-Galois objects.
We use the Miyashita-Ulbrich action to prove that every simple Galois object
is isomorphic as a graded algebra to a twisted group algebra of a normal
subgroup of G.
4.1. Simple Galois objects as twisted group algebras. Recall that a
k
G
-module algebra B is the same as a G-graded algebra B =

gG
B
g
,
where B
g
= B e
g
. Thus, if A is a k
G
-Galois object, the Miyashita-Ulbrich
action of k
G
on A denes a structure of G-graded algebra A =

gG
A
g
,
where
A
g
= a A[ab = (g b)a b A.
Proposition 4.1. Let A be a k
G
-Galois object, then
i) A
g
A = AA
g
is bilateral ideal of A,
ii) A
g
is a :(A)-submodule of A,
iii) A
e
= :(A),
iv) g A
x
= A
gxg
1.
Proof. Straightforward.
From now on A will denote a central simple algebra over K which is also
a k
G
-Galois object.
Let consider the normal subgroup of G, given by
(2) N = x G[x = , K.
Lemma 4.2. Let x G. The following conditions are equivalent:
i) x N.
ii) A
x
,= 0.
iii) A
g
A
x
= A
gx
for all g G.
iv) A
x
A
x
1 = A
x
1A
x
= K.
v) There is an invertible element in A
x
.
10 C

ESAR GALINDO AND MANUEL MEDINA

Proof. i) ii) Since A is simple, by [16, Example 2.4] we have a category
equivalence
A
K
() :
K
/
A
/
A
, V AV,
with quasi-inverse functor given by
()
A
:
A
/
A

K
/, M M
A
= m M[am = ma, a A.
For every x N, we dene an A-bimodule A
x
, where the right action is the
multiplication and left action is given by ba = (x b)a for all b A, a A
x
.
Thus, (A
x
)
A
= A
x
= a A[ab = (xb)a, b A, and since A
x
,= 0,
then A
x
,= 0.
ii) iii) Recall that A
x
A = AA
x
is a bilateral ideal. Since A is simple,
0 A
x
A
x
A = A. The equation

gG
A
g
= A = A
x
A =

gG
A
x
A
g
,
implies A
x
A
g
= A
xg
.
iii) iv) v) Straightforward.
v) i) Let u
x
A
x
be a unit, and let K. By the characterization of
Miyashita-Ulbrich, we have
[(x) ]u
x
= 0,
hence, (x) = 0.
Remark 4.3. Let S be a nite group, K a eld, and Z
2
(S, K

) a 2-
cocycle. For each s S, we will use the notation u
s
K

S to indicate the
corresponding element in the twisted group algebra K

S. Thus (u
s
)
sS
is a
K-basis of K

S, and in this basis u

s
u
t
= (s, t)u
st
.
Recall that an element s S is called -regular if (s, t) = (t, s) for all
t C
S
(s). This denition depends only on the class of s under conjugation.
Denition 4.4. Let Z
2
(S, K

) be a 2-cocycle. The 2-cocycle is called

non-degenerate if and only if 1 is the only -regular class in S.
Remark 4.5. The 2-cocycle is non-degenerate if and only if dim
K
:(K

S) =
1 (see [10, Theorem 9.3, pag 410]).
Proposition 4.6. Let A be a k
G
-Galois object, where A is a central simple
algebra over K. Then A is isomorphic to a twisted group algebra K

N,
where N = Stab
G
(K) and : N N K

is a non-degenerated 2-cocycle.
Moreover, if we suppose A = K

N, then the action of x N on a A is

given by
xa = u
x
au
1
x
.
ON THE CLASSIFICATION OF GALOIS OBJECTS FOR FINITE GROUPS 11
Proof. By Lemma 4.2, A is a crossed product ring over N. Since A
e
=
:(A) = K, A is isomorphic to a twisted group algebra K

N, thus by Remark
4.5 is a non-degenerate 2-cocycle.
Finally, if x N, then
y A
x
= Ku
x
ya = (xa)y a A.
Taking y = u
x
, we have xa = u
x
au
1
x
for all a A.
4.2. The function . The group G acts on K through the natural projec-
tion,
(3) G G/N Gal(K[k), g g.
By (iv) in Proposition 4.1, the action of G on A denes a function :
GN K

determined by the equation

(4) g u
x
= (g, x)ug
x
(
g
x := gxg
1
).
Proposition 4.7. For all g, h G and x, y N, the function : GN
K

dened in (4) satises the following equations:

Condition C1:
(x, y)(x, x
1
) = (x, y)(xy, x
1
). (C1)
Condition C2:
g((x, y))(g, xy) = (
g
x,
g
y)(g, x)(g, y). (C2)
Condition C3:
(gh, x) = g((h, x))(g,
h
x). (C3)
Proof. The conditions (C1), (C2), and (C3) are equivalent to
xu
y
= u
x
u
y
u
1
x
,
g u
x
u
y
= (g u
x
)(g u
y
),
g (hu
x
) = ghu
x
,
respectively.
Remark 4.8. If N :(G), the condition (C1) is equivalent to
(x, y) =
(x, y)
(y, x)
=: Alt

(x, y).
Moreover, if (N, N), (G, N) k

, the conditions (C2) and (C3) are equi-

valent to
(g, xy) = (g, x)(g, y) (5)
(gh, x) = (g, x)(h, x) (6)
for all g, h G and x, y N. A function that satises (5) and (6) is called
a pairing.
12 C

ESAR GALINDO AND MANUEL MEDINA

5. Construction of k
G
-Galois objects from Group-Theoretical
data
5.1. Data associated with Galois objects. Let G be a nite group and
k be a eld.
Denition 5.1. AGalois datum associated to k
G
is a collection (S, K, N, , )
such that
i) S is a subgroup of G and N is a normal subgroup of S.
ii) K k is a Galois extension with Galois group S/N.
iii) char(k) [N[.
iv) : N N K

is a non-degenerate 2-cocycle.
v) : S N K

satises the equations (C1), (C2), and (C3).

Let (S, K, N, , ) be a Galois datum associated to k
G
. We will denote
by A(K

N, ) the twisted group algebra K

N with S-action dened by

g u
x
= (g )(g u
x
) = g()(g, x)ug
x
,
for g S, x N, and K.
We will denote by Ind
G
S
(A(K

N, )) the induced G-algebra from the S-

algebra A(K

N, ).
Proposition 5.2. Let (S, K, N, , ) be a Galois datum associated to k
G
.
The S-algebra A(K

N, ) is a simple k
S
-Galois object, with :(A(K

N, )) =
K.
Proof. Since char(k) [N[, it follows from Maschkes Theorem for twisted
group algebra (see [10, 2.10 Theorem]), that the algebra K

N is semisimple.
Moreover, since is non-degenerate we have :(K

N) = K. Hence K

N is
a central simple algebra over K.
Let us denote by A the S-algebra A(K

N, ). If we can prove that A

is a K
N
-Galois object, then it will follows that A is a k
S
-Galois object by
Corollary 2.9.
Since A is a central simple algebra over K, the map
A
K
A
op
End
K
(A)
a b [t atb]
is an isomorphism (see [7, pag. 32]). The map
N
: A#N End
K
(A),

N
(a#x)(b) = a(xb) is surjective, since that

N
(u
x
u
y
#y
1
)(a) = u
x
au
y
.
Now, dim
K
(A#N) = dim
K
(End
K
(A)), hence
N
is bijective. Thus, A is an
K
N
-Galois object by Proposition 2.3.
5.2. Equivalence of Galois data.
Denition 5.3. We will say that (S, K, N, , ) and (S

, K

, N

) are
equivalent Galois data associated to k
G
if there exists a G-algebra isomor-
phism Ind
G
S
(A(K

N, ))

= Ind
G
S
(A(K

)).
ON THE CLASSIFICATION OF GALOIS OBJECTS FOR FINITE GROUPS 13
Denition 5.4. Let G be a nite group, S a subgroup of G and (A, ) an
S-algebra. For each g G, we consider the g
1
Sg-algebra (A
(g)
,
g
), such
that A
(g)
= A as algebras and g
1
Sg-action given by
h
g
a = (ghg
1
) a,
for all h g
1
Sg and a A
(g)
.
Note that for all g G, the map

g
: Ind
G
S
(A) Ind
G
g
1
Sg
(A
(g)
) (7)
f
g
(f) = [h f(gh)].
denes a G-algebra isomorphism Ind
G
S
(A)

= Ind
G
g
1
Sg
(A
(g)
).
Recall that two transitive G-sets G/S and G/S

are isomorphic if and

only if there exists g G such that S

= g
1
Sg.
Lemma 5.5. Let S, S

be two subgroups of G, A be a simple S-algebra,

and B be a simple S

-algebra. Then there exists a G-algebra isomorphism

between Ind
G
S
(A) and Ind
G
S
(B) if and only if S = g
1
S

g and A B
(g)
as
S-algebras.
Proof. Let e
i

1i[G:S]
, r
i

1i[G:S

]
and e

1i[G:g
1
S

g]
be the canoni-
cal imprimitive systems associated to Ind
G
S
(A), Ind
G
S
(B) and Ind
G
g
1
S

g
(B
(g)
)
(see Remark 3.10).
Suppose that : Ind
G
S
(A) Ind
G
S
(B) is a G-algebra isomorphism.
Since that A and B are simple algebras, the elements in e
i

1i[G:S]
,
and r
i

1i[G:S

]
are the central primitive idempotents for Ind
G
S
(A), and
Ind
G
S
(B), respectively. Then induce a G-set isomorphism between e
i

G/S and r
i
G/S

. Hence, there exists g G such that S = g

1
S

g.
Now, by (7) we have the following G-algebra isomorphism
Ind
G
S
(A)

Ind
G
S
(B)
g
Ind
G
g
1
S

g
(B
(g)
),
then we have that
A e
1
Ind
G
S
(A)
g
e

1
Ind
G
S
(B
(g)
) B
(g)
.
Conversely, suppose that S = g
1
S

g and A B
(g)
as S-algebras. We
can construct a G-algebra isomorphism between Ind
G
S
(A) and Ind
G
S
(B
(g)
),
using the Remark 3.7 and (7).
Let f : A A

be an S-algebra isomorphism between A = A(K

N, )
and A

= A(K

). By the denition of the Miyashita-Ulbrich action

f is an S-graded algebra isomorphism. Since N = x S[A
x
,= 0 (see
Lemma 4.2), we have N = N

. Moreover, f[
K
is a eld isomorphism, so we
can suppose without loss of generality that K = K

.
14 C

ESAR GALINDO AND MANUEL MEDINA

Proposition 5.6. The Galois data (S, K, N, , ) and (S, K, N,

) as-
sociated to k
S
are equivalent if and only if there are :(Gal(K[k)), and
: N K

, such that
((x, y))(xy) = (x)(y)

(x, y), (8)

((s, x))(
s
x) = s((x))

(s, x), (9)

for all x, y N, s S.
Proof. Let u
x
and u

x
be a K-basis of A and A

, see Remark 4.3.

Suppose that f : A(K

N, ) A(K

N,

) is an S-algebra isomorphism.
Consider := f[
K
:(Gal(K[k)) and let : N K

be determined by
the equation
f(u
x
) = (x)u

x
(x N).
A straightforward computation shows that and satisfy the equations (8),
and (9).
Conversely, it is easy to check that if and satisfy the equations (8),
(9), then f(u
x
) := ()(x)u

x
is an S-algebra isomorphism.
Lemma 5.7. Let (S, K, N, , ) be a Galois datum associated to k
G
. For
every g G, the Galois datum associated to A(K

N, )
(g)
is given by
(g
1
Sg, K, N,
(g)
,
(g)
), where
(g)
(x, y) = (gxg
1
, gyg
1
), and
(g)
(h, x) =
(ghg
1
, gxg
1
), for all x, y N, h g
1
Sg.
Proof. Let A = A(K

N, ). Note that (A
(g)
)
x
= A
gxg
1, then u
gxg
1[x
N is a basis of A
(g)
, where u
gxg
1 (A
(g)
)
x
. Hence,
(g)
(x, y) = (gxg
1
, gyg
1
),
and
(g)
(h, x) = (ghg
1
, gxg
1
), for all x, y N, h g
1
Sg.
Theorem 5.8. The data (S, K, N, , ) and (S

, K, N,

) are equivalent
if and only if there exists g G, :(Gal(K[k)) and : N K

such
that S = g
1
S

g and
((x, y))(xy) = (x)(y)

(gxg
1
, gyg
1
),
((s, x))(
s
x) = s((x))

(gsg
1
, gxg
1
),
for all x, y N, s S.
Proof. Since A(K

N, ) and A(K

N,

) are simple algebras, the proof of

the theorem follows from Lemma 5.7, Lemma 5.5, and Proposition 5.6.
6. Classification of k
G
-Galois objects
In this section we shall prove Theorem 1.2.
Lemma 6.1. Let (S, K, N, , ) be a Galois datum associated to k
S
. If
A(K

N, ) is a k
S
-Galois object then char(K) [N[. In particular, char(k)
[N[.
ON THE CLASSIFICATION OF GALOIS OBJECTS FOR FINITE GROUPS 15
Proof. By Corollary 2.9, the algebra K

N is a K
N
-Galois object, so the
map : K

N N End(K

N) is an isomorphism. Note that

(

yN
u
e
#y)(u
x
) =

yN
u
y
u
x
u
1
y
lies in the center of the twisted group algebra. Then (

yN
u
e
#y)(u
x
) = 0
if x ,= e and (

yN
u
e
#y)(u
e
) = [N[u
e
. Hence, [N[ , = 0.
Theorem 6.2. Let S be a nite group, k a eld, and A a simple S-algebra.
Then A is a k
S
-Galois object if and only if there exists (S, K, N, , ) a
Galois datum associated to k
S
, such that
A A(K

N, ),
as S-algebras.
Proof. Suppose that A is a k
S
-Galois object. Let K = :(A) and N = g
S[g = K. A is isomorphic to the twisted group algebra K

N,
where is non-degenerate (see Proposition 4.6). Now, char(k) [N[ by
Lemma 6.1. The function dened by the equation
g u
x
= g()(g, x)ug
x
, (g S, x N, K),
completes a Galois datum (S, K, N, , ), such that A A(K

N, ).
Proof of Theorem 1.2. It follows from Theorem 6.2, and Theorem 3.14.
6.1. Two families of Examples. Let G = Z
n
Z
n
Z
n
and N = 0
Z
n
Z
n
G with n an odd prime. Consider the elds k = Q[] and
K = Q[q], where q = e
2i
n
2
is a primitive nth root of unity and = q
n
. Now,
dene a 2-cocycle : N N k

of N with values in

, by
(x, y) =
x
2
y
1
x
1
y
2
(x = (x
1
, x
2
), y = (y
1
, y
2
) N).
Since 0 = x N[(x, y) = (y, x) y N, is a non-degenerate
2-cocycle.
In this case, Remark 4.8 gives a function : GN k

satises the
conditions (C1), (C2) and (C3) if and only if is a pairing such that [
NN
=
Alt

=
2
. Hence, it is easy to see that the collection (G, K, N, , ) is a
Galois datum associated to k
G
. Moreover, if

is another paring such that

[
NN
=
2
, then the Galois data (G, K, N, , ) and (G, K, N, ,

) are
equivalent if and only if =

.
We can obtain another family of Galois data by taking G, N, k, and K
as above, and dening the 2-cocycle

: N N k

by

(x, y) =
x
2
y
1
(x = (x
1
, x
2
), y = (y
1
, y
2
) N),
and a pairing : GN k

, such that
(x, y) = Alt

(x, y) =
x
2
y
1
x
1
y
2
,
for all x, y N.
16 C

ESAR GALINDO AND MANUEL MEDINA

7. Obstruction theory for Galois data
In this section we present some necessary conditions for the existence of
a function : G N K

, such that (S, N, K, , ) is a Galois datum

associated to k
G
.
7.1. Hochschild cohomology for groups. Recall that the Hochschild
complex (C
n
(G, A), d
n
) is dened as follows.
Let G be a group and A be a G-bimodule. Dene C
0
(G, A) = A, and for
n 1
C
n
(G, A) = f : G G
. .
ntimes
A[f(x
1
. . . , x
n
) = 0, if x
i
= 1
G
for some i.
Let consider the following cochain complex
0 C
0
(G, A)
d
0
C
1
(G, A)
d
1
C
2
(G, A) C
n
(G, A)
dn
C
n+1
(G, A)
where
d
n
(f)(x
1
, x
2
, . . . , x
n+1
) = x
1
f(x
2
, . . . , x
n+1
)
+
n

i=1
(1)
i
f(x
1
, . . . , x
i1
, x
i
x
i+1
, x
i+2
, . . . , x
n+1
)
+ (1)
n+1
f(x
1
, . . . , x
n
) x
n+1
.
Then put HZ
n
(G, A) := ker(d
n
), HB
n
(G, A) := Im(d
n1
) and HH
n
(G, A) :=
HZ
n
(G, A)/HB
n
(G, A) (n 0) is the Hochschild cohomology of G with co-
ecients in A.
Remark 7.1. A left G-module A is a G-bimodule with the trivial right action,
in this case the dierential maps will be denoted by
n
, and the Hochschild
cohomology of G with coecients in A, is the ordinary group cohomology.
Let G be a nite group, N a normal subgroup of G, K a Galois extension
of k with Galois group G/N, and Z
2
(N, K

) a non-degenerate 2-cocycle.
Thus the abelian group
C
1
(N, K

) = f : N K

[f(e) = 1,
is a G-bimodule with left action
(10) (g f)(x) = g(f(x)),
and right action
(11) (f g)(x) = f(
g
x) (
g
x = gxg
1
),
for all g G, x N, and f C
1
(N, K

).
By abuse of notation we will identify a function : G C
1
(N, K

) with
its adjoint function : GN K

, and vice versa.

ON THE CLASSIFICATION OF GALOIS OBJECTS FOR FINITE GROUPS 17
7.2. Obstructions. The abelian groups C
n
(N, K

), Z
n
(G, K

) are G-bimodules
with left action (g )(x
1
, . . . , x
n
) = g((x
1
, . . . , x
n
)), and right action
(
g
)(x
1
, . . . , x
n
) = (
g
x
1
, . . . ,
g
x
n
). Analogously, the abelian group Hom(N, K

) =

N is a G-bimodule.
Proposition 7.2 (First obstruction). There exists a function : GN
K

that satises (C2) if and only if the second cohomology class of

g

g
is
zero for all g G.
Proof. This follows immediately from condition (C2).
Suppose that the rst obstruction is zero, then for all g G there exists
a
g
: N K

, such that

1
(
g
) =
g

g
B
2
(N, K

).
This denes a function : G C
1
(N, K

), g
g
.
Lemma 7.3. For all g, h G, d
1
()(g, h)

N, the function d
1
() : GG

N is a Hochschild 2-cocycle of G with values in

N.
Proof. If g, h G, then

1
(d
1
()(g, h)) =
1
[(g
h
)(
gh
)
1
(
h
g
)]
=
_
g
h

h
__

gh
gh
__
g
h
(
g
)
h
_
= 1.

A straightforward computation shows that if

: G C
1
(N, K

), g

g
,
is another function such that
1
(

g
) =
g

g
for all g G, then the Hochschild
2-cocycles d
1
(

) and d
1
() are cohomologous.
Proposition 7.4 (Second obstruction). There exists a function : GN
K

that satises (C3) if and only if the second Hochschild cohomology class
of d
1
() is zero.
Proof. If there is a function : G N K

that satises (C3), then

d
1
() = 0, so its cohomology is zero too.
Conversely, if C
1
(G, C
1
(N, K

)) is a function, such that there exists

C
1
(G,

N) with d
1
() = d
1
(), then the function : GN K

dened
by (g, n) =
g
(n)
1
(g, n) for all g G, n N, satises (C3).
References
[1] E. Aljade, P. Etingof, S. Gelaki, and D. Nikshych. On twisting of nite-dimensional
Hopf algebras. J. Algebra, (256), pages 484501, 2002.
[2] E. Artin. Galois Theory. University of Notre Dame Press, 1944.
[3] S. U. Chase, D. K. Harrison, and A. Rosenberg. Galois theory and cohomology of
commutative rings. Mem. AMS 52, 1965.
18 C

ESAR GALINDO AND MANUEL MEDINA

[4] A. A. Davydov. Galois algebras and monoidal functors between categories of repre-
sentations of nite groups. J. Algebra, (244), pages 273301, 2001.
[5] A. A. Davydov. Modular invariants for group-theoretical modular data. I. J. Algebra,
(323), pages 13211348, 2010.
[6] C. Galindo and S. Natale. Normal Hopf subalgebras in cocycle deformations of nite
groups. Manuscripta Math., (125) pages 501504, 2008.
[7] P. Gille and T. Szamuely. Central Simple Algebras and Galois Cohomology. Cambrigde
University Press, 2006.
[8] Y.-Z.Huang, L. Kong. Open-string vertex algebras, tensor categories and operads.
Comm. Math. Phys. (250), pages 433471, 2004.
[9] Y.-Z.Huang, L. Kong. Full eld algebras. Comm. Math. Phys., (272), pages 345396,
2007.
[10] G. Karpilovsky. Projective Representation of Finite Groups, volume 94. Pure and
Applied Mathematics, Marcel Dekker, New York-Basel, 1985.
[11] F. Lorenz. Algebra: Fields with Structure, Algebras and Advanced Topics, volume II.
Springer, 2008.
[12] S. Montgomery. Hopf Algebras and Their Action on Rings, volume 82. Am. Math.
Soc., Providence, Rhode Island, 1993.
[13] M. Movshev. Twisting in group algebras of nite groups. Func. Anal Appl., (27),
pages 240244, 1994.
[14] P. Schauenburg. Hopf Bigalois extensions. Comm. in Algebra, (24), pages 37973825,
1996.
[15] P. Schauenburg and H.-J. Schneider. Galois Type Extensions and Hopf Algebras.
Banach Center Publ., in press, 2004.
[16] M. Takeuchi. Formal ring laws and monoidal equivalences of categories of bimodules.
J. Math. Soc. Japan, pages 301336, 1987.
[17] K.-H. Ulbrich. Fiber functor of nite dimensional comodules. Manuscripta Math.,
pages 3946, 1989.
Departamento de Matem aticas, Universidad de los Andes, Bogot a, Colom-
bia
E-mail address: cn.galindo1116@uniandes.edu.co, cesarneyit@gmail.com
UFR de Math ematiques, Universit e Paris 7 - Denis Diderot, Paris, France
E-mail address: manuel.medina@etu.univ-paris-diderot.fr