CROSSED PRODUCT TENSOR CATEGORIES

CESAR GALINDO
arXiv:0911.0881v2 [math.QA] 27 Jun 2010

Abstract. A graded tensor category over a group G will be called a crossed

product tensor category if every homogeneous component has at least one mul-
tiplicatively invertible object. Our main result is a description of the crossed
product tensor categories, graded monoidal functors, monoidal natural trans-
formations, and braidings in terms of coherent outer G-actions over tensor
categories.

1. Introduction
L
A G-graded ring A = G A is called a G-crossed product if each A has
an invertible element. Some important classes of rings like skew group-rings and
twisted group-rings are special cases of crossed product rings. One of the basic
examples is the group algebra of a group F , it is graded by a quotient group of
F , see [4, Subsection 11C]. In this case the theory of representation of F can be
analyzed using Clifford theory, see [4, Subsection 11A].
InLanalogy with graded rings, a G-graded tensor category (see Subsection 2.5)
C = G C will be called G-crossed product tensor category, if there is an in-
vertible object in each homogeneous component of C.
The graded tensor categories appear naturally in classification problems of fusion
categories and finite tensor categories [5], [8]. One of the most interesting examples
of G-crossed product tensor categories is the semi-direct product tensor category
associated to an action of a group over a tensor category, see [14]. Semi-direct tensor
product category has been used in order to solve an important open problem in
semisimple Hopf algebras theory [13].
In [9] was proposed a Clifford theory categorification for crossed product ten-
sor categories, in order to describe simple module categories in terms of subgroups
and induced module categories. We stress that in [9] crossed product tensor cate-
gories were called strongly graded tensor categories. Recently, G-graded fusion cate-
gories (that include crossed product fusion categories) were classified by Etingof,
Nikshych, and Ostrik in [7], using invertible bimodule categories over fusion cate-
gories.
The crossed product rings are commonly described using crossed systems [12].
Crossed systems can be interpreted in terms of monoidal functors in the following
way: if A is a ring, let we denote by Out(A) the monoidal category (in fact, it is a
categorical-group) of outer automorphisms, were the objects are automorphisms of
A, and the arrows between automorphisms and are invertible elements a A,
such a(x) = (x)a for all x A. Given a group G, the data that define G-crossed

Date: June 29, 2010.

1991 Mathematics Subject Classification. 16W30, 18D10.
1
2
CESAR GALINDO

systems are the same as the data that define monoidal functors from G to Out(A),
where G is the discrete monoidal category associated with G.
We develop crossed product system theory for crossed product tensor categories
using higher category theory. If C is a tensor category, the bicategory Bieq(C) is
a monoidal bicategory (in fact, it is a weak 3-group). So, a G-crossed system or
coherent outer G-action over C must be a trihomomorphism from G (the discrete
3-category associated to G, see Remark 3.5) to Bieq(C).
The main goal of this paper is to describe the 2-category of G-crossed pro-
duct tensor categories in terms of coherent outer G-actions over a tensor category
(Theorem 4.1), and describe the braidings of G-graded tensor categories.
The organization of the paper is as follows: in Section 2 we recall the main
definitions of bicategory theory, as well as the definitions of categorical-groups,
graded tensor categories, and the monoidal structure over Bieq(C). In Section 3 we
define incoherent and coherent outer G-actions over a tensor category C, and we
show an explicit bijective correspondence between equivalence classes of G-crossed
product tensor categories and coherent outer G-actions. In Section 4 we show a
biequivalence between the 2-category of crossed product tensor categories and the
2-category of coherent outer G-actions. Finally, in Section 5 we give some necessary
and sufficient conditions for the existence of a braiding over a crossed product tensor
category.

2. preliminaries
2.1. General conventions. Throughout this article we work over a field k. By a
tensor category (C, , , I) we understand a k-linear abelian category C, endowed
with a k-bilinear exact bifunctor : CC C, an object I C, and an associativity
constraint V,W,Z : (V W ) Z V (W Z), such that Mac Lanes pentagon
axiom holds [2], V I = I V = V , V,I,W = idV W for all V, W C and
dimk End C (I) = 1.
We shall considerer only monoidal categories which constraint of unit are identi-
ties. So, without loss of generality, we shall suppose that for every monoidal functor
(F, ) : C D, we have F (IC ) = ID and V,I = I,V = idV , since each monoidal
functor is monoidally equivalent to one with these properties.

2.2. Bicategories. In this section we review some definitions on bicategory theory

that we shall need later. We refer the reader to [3] for a detailed exposition on the
subject.

Definition 2.1. A bicategory B consists of the following data

a set Obj(B) (with elements A, B, . . . called 0-cells ),
for each pair A, B Obj(B), a category B(A, B) (with objects V, W, . . .
called 1-cells and morphisms f, g, . . . called 2-cells),
for each A, B, C Obj(B), a bifunctor

ABC : B(A, B) B(B, C) B(A, C),

for each 0-cell A Obj(B), a 1-cell IA B(A, A),

 CROSSED PRODUCT TENSOR CATEGORIES 3

for each A, B, C, D Obj(B), natural isomorphisms (constraint of associa-

tivity)

A,B,C,D : ABD (BCD ) (ABC )ACD :

B(A, B) B(B, C) B(C, D) B(A, D).

Subject to the following axioms

coherence of the associativity: if (S, T, U, V ) is an object in B(A, B)
B(B, C) B(C, D) B(D, E), the next diagram commutes

S(T (U V ))
t JJ
tt JJ
S,T ,U V ttt JJ idT ,U,V
t JJ
tt JJ
t JJ
ttt JJ
zt $
(ST )(U V ) S((T U )V )

ST ,U,V S,T U,V

 S,T ,U id

((ST )U )V o (S(T U ))V

coherence of the unity

S,IB ,T = idST .

If is the identity, we have (ST )U = S(T U ) and similarly for morphisms,

in this case we shall say that B is a 2-category.
A monoidal category (C, , , I) is the same as a bicategory with only one 0-cell,
and in this case = .

Definition 2.2. Let B = (, I, ) and B = (, I , ) be bicategories. A pseudo-

functor = (F, ) from B to B consists of the following data
a function F : Obj(B) Obj(B ), A 7 F (A),
for each pair A, B Obj(B), functors

FAB : B(A, B) B (F (A), F (B)), S 7 F (S), f 7 F (f ),

for each triple A, B, C Obj(B), a natural isomorphism

ABC : FAC (ABC ) FAB ()F (A)F (B)F (C) FBC (),

Subject to the following axioms

(i) FAA (IA ) = IF (A) ,
4
CESAR GALINDO

(ii) if (S, T, U ) is an object in B(A, B) B(B, C) B(C, D), the following

diagram commutes (where the indexes have been omitted)
F ()
F (S(T U )) / F ((ST )U )

 
F (S)F (T U ) F (ST )F (U )

id id
 
/ (F (S)F (T ))F (U )
F (S)(F (T )F (U ))

(iii) if S is an object in B(A, B), then S,IB = idF (S) and IA ,S = F (S), for
each pair of 0-cells A, B Obj(B).
Remark 2.3. (1) The notion of pseudofunctor can be in some manner dualized
by reversing the direction of the 2-cells FA,B , this notion will be called
op-pseudofunctor.
(2) A pseudofunctor between monoidal categories is just a monoidal functor
and an op-pseudofunctor is an op-monoidal functor.
Definition 2.4. Let F, G : B0 B1 be pseudofunctors between bicategories B0
and B1 . A pseudonatural transformation : F G, consists of the following data
for each A Obj(B0 ), 1-cells A B1 (F (A), G(A)),
for each pair A, B Obj(B0 ), and each 1-cell V B0 (A, B) a natural
isomorphism
V : F AB (V )F (A)F (B)G(B) B A F (A)G(A)G(B) GAB (V ),
such that IA = idIA for all A Obj(B0 ) and for all S B0 (A, B), T B0 (B, C),
the following diagram
ST
F (ST ) / G(ST )

F id id G

 idF (S) T id

F (S)F (T ) / F (S)G(T ) S G(T/ )G(S)G(T )

commutes (where associativity constraint, indexes, and the symbols between ob-
jects have been omitted as a space-saving measure).

Remark 2.5. Again, the notion of pseudonatural transformation can be dualized

by reversing the order of the natural isomorphisms V , this notion will be called
op-pseudonatural transformation.
Pseudonatural transformations may be composed in the obvious way. If :
F G, and : G H are pseudonatural transformations, then we define a new
pseudonatural transformation : F H by ( )A = A A , and ( )V is
defined by the commutativity of the diagram:
 CROSSED PRODUCT TENSOR CATEGORIES 5

( )V
F (V )(A A ) / (B B )H(V )
O
F (V ),A ,A B ,B ,H(V )

(F (V )A )A B (B H(V ))
O
V idA idB V
 1
B ,G(V ),A
(B G(V ))A / B (G(V )A )

where the index have been omitted.

A modification between two pseudonatural transformations : e, consists
of 2-cells A : A
eA in B1 (F (A), G(A)), such that for all 1-cell V B0 (A, B)
the diagram
V
F AB (V )B / A GAB (V )

idB A id

 
eV

F AB (V )e
B eA GAB (V )
/

commutes.
2.3. The Monoidal bicategory Bieq(C) of a tensor category. Given a pair of
bicategories B and B , we can define the functor bicategory [B, B ], whose 0-cells
are pseudofunctors B B , whose 1-cells are pseudonatural equivalence, and whose
2-cells are invertible modifications.
The bicategory [B, B ] is not usually a 2-category, because composition of 1-cells
in [B, B ] involves composition of 1-cells in B , but in the case that B is a 2-category,
[B, B ] is a 2-category.
When B = B , the bicategory [B, B] will be denoted by Bieq(B), and it has
a monoidal structure in the sense of [10]. Now, we shall describe the monoidal
bicategory Bieq(C) associated to a tensor category (C, , I).
The tensor product of monoidal endofunctors is defined by the composition
of monoidal functors. If ( , ()

) : K K , (, () ) : H H are pseudonatural
transformations, the tensor product ( , ()
)(, () ) : KH K H is defined as
( , () )(, () ) := (K() , ), where is given by the commutativity

of the following diagram

( )V
KH(V ) K() / K() K H (V )
O O
K
H(V ), id
idK() H (V )

K(V )id K (V ) id
K(H(V ) ) / K( H (V )) ,H / K() K(H (V ))

The tensor product of modifications g and f is defined as gf := K(f ) g.

6
CESAR GALINDO

If 1 = (, () ) : F F and 2 = ( , ()
) : H H are pseudonatural trans-

formations, where F, F , H, H : C C are monoidal functors, then the comparison
constraint
F: HJ
1 idH ttt JJ idF 2
t JJ
t JJ
tt $
F HJ c1 ,2 F: H
JJ tt
JJ tt
J t
idF 2 J$ tt 1 idH
F H
is given by
c1 ,2 := 1
: F ( ) F ( ) .
The constraint of associativity af,g,h : (f g) h f (g h) of the tensor
product of pseudonatural transformations f : K K , g : H H , h : G G is
given by the modification
K
H(h),g idf : KH(h) K(g) f K(H(h) g) f,
and it is easy to see that a satisfies the pentagonal identity.
Remark 2.6. The data (TD6), (TD7), and (TD8) of [10], in the monoidal bicategory
Bieq(C) are trivial, since we only considerer monoidal functor (F, ) : C C such
that F (I) = I and V,I = I,V = idV , for all V C.
The category Bieq(C)(idC , idC ) is exactly the center of C, i.e., the braided monoidal
category Z(C), see [11, pag. 330].
2.4. Categorical-groups. A categorical-group G is a monoidal category where
every object, and every arrow is invertible, i.e. for all X Obj(G) there is X
Obj(G), such that X X = X X = I. We refer the reader to [1] for a detailed
exposition on the subject.
A trivial example of a categorical-group is the discrete categorical-group G, as-
sociated to a group G. The objects of G are the elements of G, the arrows are only
the identities, and the tensor product is the multiplication of G. A more interesting
examples is the following.
Example 2.7. Let G be a group, A a G-module, and Z 3 (G, A) a normalized
3-cocycle. We shall define the category C(G, A, ) by:
(1) Obj(C(G, A, )) = G,
A, if g = h
(2) HomC(G,A,) (g, h) =
, if g 6= h.
We define a monoidal structure in C(G, A, ) as follows:
Let g End (a) and h End (b), a, b A, g, h G. Thus, a b = a + g b and
g h = gh. We define the associator as g,h,k = (g, h, k).
The 3-cocycle condition is equivalent to the pentagon axiom, and the condition
of normality implies that e is the unit object for this category.
Complete invariants of a categorical-group G with respect to monoidal equiva-
lences are
0 (G), 1 (G), (G) H 3 (0 (G), 1 (G)),
where 0 (G) is the group of isomorphism classes of objects, 1 (G) is the abelian
group of automorphisms of the unit object. The group 1 (G) is a 0 (G)-module in
 CROSSED PRODUCT TENSOR CATEGORIES 7

the natural way, and (G) is a third cohomology class given by the associator, see
[1, Subsection 8.3] for details on how to obtain (G).
If G is a categorical group by [1, Theorem 43] there is an equivalence of monoidal
categories between G and C(0 (G), 1 (G), ), where is a 3-cocycle in the class (G).
Also, it is easy to see that there is a bijective correspondence between monoidal
functors
F : C(G, A, ) C(G , A , )
and triples (0 (F ), 1 (F ), (F )) that consist of:
a group morphism 0 (F ) : G G ,
a G-module morphism 1 (F ) : A A ,
a normalized 2-cochain k(F ) : G2 A , such that dk(F ) = 1 (F )
0 (F )3 .
For monoidal functors F, F : C(G, A, ) C(G , A , ), there is bijective cor-
respondence between monoidal natural isomorphisms : F F and normalized
1-cochains p() : G A , where dp() = k(F ) k(F ).
The next result follows from the last discussion or from [1, Theorem 43].
Proposition 2.8. Let G be a categorical group and let f : G 0 (G) be a morphism
of groups. Then there is a monoidal functor F : G G, such that f = 0 (F ) if
and only if the cohomology class of f 3 is zero, where is a 3-cocycle in the class
(G).
If f 3 is zero, the classes of equivalence of monoidal functors F : G G such
that 0 (F ) = f are in one to one correspondence with H 2 (G, 1 ((G))).

2.5. Crossed product tensor categories. Let G be a group and let C be a tensor
category. We shall say that C is G-graded, if there is a decomposition
C = xG Cx
into a direct sum of full abelian subcategories, such that for all , x G, the
bifunctor maps C Cx to Cx , see [5].
DefinitionL 2.9. Let C be a tensor category graded over a group G. We shall say
that C = G C is a crossed product tensor category over G, if every C has a
multiplicatively invertible object.
Given a group G, we define the 2-category of G-crossed product tensor categories.
The 0-cells are crossed product tensor categories over G, 1-cells are graded op-
monoidal functors, i.e., op-monoidal functors F : C D such that F maps C to
D for all G, and 2-cells are monoidal natural transformations between the
graded op-monoidal functors. The composition of 1-cells and 2-cells is the obvious.
Remark 2.10. The existence of some extra properties of a crossed product tensor
category C, can be deduced from the tensor subcategory Ce . For example C is
semisimple or rigid if and only if Ce is semisimple or rigid. However, if Ce is a
braided tensor category, not necessary C is braided, see Section 5.
A crossed product tensor category C is a fusion category [5] or finite tensor
category [8], if and only if G is finite, and Ce is a fusion category or a finite tensor
category, respectively.
8
CESAR GALINDO

3. Outer G-actions over tensor categories

3.1. Incoherent outer G-actions. Let C be a tensor category. We define the
categorical-group 2Out (C), where objects are monoidal autoequivalences of C,
and arrows are equivalence classes of invertible pseudonatural isomorphisms up to
invertible modifications. The composition of arrows in 2Out (C) is the equivalence
class of pseudonatural isomorphisms composition, and the tensor product is the
composition of monoidal functors and pseudonatural transformations.
Definition 3.1. Let G be a group and let C be a monoidal category. An inco-
herent outer G-action over C, is an op-monoidal functor : G 2Out (C). Two
incoherent outer G-actions are equivalent if the associated monoidal functors are
monoidally equivalent.
We shall analyze the incoherent outer G-action using the Subsection 2.4. Com-
plete invariants for the categorical group 2Out (C) are 0 (2Out (C)) the equiva-
lences classes of monoidal functor under invertible modification, 1 (2Out (C)) =
Inv(Z(C)) the abelian group of isomorphisms classes of invertible objects of the cen-
ter of C, and a third cohomology class (2Out (C)) H 3 (0 (2Out (C)), Inv(Z(C))).
Every incoherent outer G-action over a tensor category induces a group mor-
phism f : G 0 (2Out (C)). We shall say that a group morphism f : G
0 (2Out (C)) is realizable if there is some incoherent outer G-action such that the
induced group morphism coincides with f .
Proposition 3.2. Let G be a group and let f : G 0 (Out (C)) be a group
morphism. Then there is an incoherent outer G-action over C that realize the
morphism f if and only if the cohomology class of f 3 is zero, where is some
3-cocycle in the class of (2Out (C)).
If f 3 is zero, the classes of equivalence of monoidal functors F : G 2Out (C)
such that 0 (F ) = f are in one to one correspondence with H 2 (G, Inv(Z(C))).
Proof. See Proposition 2.8. 
3.2. Coherent outer G-actions. Let C be a monoidal category and let F : G
2Out (C) be an incoherent outer G-action. We define a crossed system associated
to F as the following data
monoidal functors ( , ) : C C for all G,
pseudonatural isomorphisms (U, , , ) : ( ) for all , G,
invertible modifications ,, : , (id , ) , (, id ) for
all , , G.
that realize the incoherent outer G-action F (recall that the symbol is the com-
position of 1-cells in the bicategory Bieq(C)).
Remark 3.3. (1) By abuse of notation, we write , instead of pseudonatural
transformation (U, , , ), when no confusion can arise.
(2) By the definition of 2Out (C), there are several crossed systems that realize
an incoherent G-action.
(3) For every crossed system, without loss of generality, we can and shall assume
that
e = idC the monoidal identity functor,
e, = ,e = (I, id ) the identity pseudonatural isomorphism,
 CROSSED PRODUCT TENSOR CATEGORIES 9

,e, = id, the identity modification,

for all , G.
Definition 3.4. Let C be a tensor category and let F : G 2Out (C) be an
incoherent outer G-action. A coherent outer G-action associated to F , is a crossed
}G , , ) associated to F , such that for the pseudonatural isomor-
system ({e
phisms , and the invertible modifications ,, , the diagram

(3.1)
(,, )idU, U , ,U ,
idU,
( (U, )U, )U, / (U, U , )U, / (U, ) (U , )U,

(U idU, id (U, ) , ,
, ),U,
 
( (U, )) (U, )U, (U, )U, U ,

id ( (U, )) ,, ,, idU ,
 , (U, )idU, idU ,, 
( (U, ))U, U, / U, ( ) (U, )U, , / U, U, U ,

commutes for all , , , G (where tensor symbols among objects and arrows
have been omitted as a space-saving measure).
Remark 3.5. For every group G, we can associate a discrete 3-category G, where
objects are elements of G, and

{}, if g = h
G(g, h) =
, if g 6= h.

: G G G,
g,h : G(, ) G(, ) G(, ).
The definition of a coherence outer G-action over C, is equivalent to the definition of
a trihomorphism from G to Bieq(C) (see [10] for the definition of trihomomorphism).
Given a crossed system associated to an incoherent outer G-action, we can define
a monoidal bicategory. In order to describe the monoidal bicategory in a simple
way we can suppose, without loss of generality that C is skeletal, so for every pair
, G there is only one pseudonatural transformation 1 , : ( ) ,
such that 1
, , = id
, and , 1
, = id( )
for all , G. Let
< G > Bieq(C) be the full sub-bicategory where objects are { }G . We define
a homomorphism of bicategories G :< G > < G >< G > by G = ( ) ,
and the commutativity of the diagram
f g
/
 
 
,   ,
 
 
f G g
( ) / ( )

where f Bieq( , ), g Bieq(C)( , ).

10
CESAR GALINDO

We define a pseudonatural equivalence in the bicategory [< G > < G > <
G >, < G >] by the commutativity of the diagram
a,,
( ) / ( )
O O
 
,   ,
 
 
( ) ( )
_?? ?
?? 
?? 
?? 
id , ?? 
??  , id


The diagram (3.1) define a modification

: a,, a,, a,, G id a, , id G a,, ,
in the bicategory [< G > < G > < G > < G >, < G >].
Since C is skeletal for every invertible object U C, we can identify AutC (U )
with Aut(I) = k , so the modification is identified by a map
: G G G k .
The modification defines a map
: G G G G k ,
given by
(, , , ) = ()(, , , )(U, ),U, , (U, )(U

, ,U ,
)1 .

It is straightforward to see that is a 4-cocycle, see [7, Subsection 8.4]. It is

also possible to see the 4-cocycle condition directly for the nonabelian 4-cocycle
condition [10, (TA1)] in the monoidal bicategory < G >. It is clear that if the
chosen modification is changed, the 4-cocycle only change for a 4-coboundary, so
an incoherent outer G-action defines a fourth cohomology class.
Proposition 3.6. Let C be a tensor category. An incoherent outer G-action over
C is coherent if and only if the associated fourth cohomology class is trivial.
Proof. If an outer G-action is coherent, the diagram (3.1) commutes, so the map
is trivial. Conversely, if there is a 3-coboundary : G G G G, such that
() = , then the modification defined by the map 1 defines a coherent outer
G-action. 

3.3. The crossed product tensor category associated to a coherent outer

G-action. If a group G acts over a monoidal category C, we shall define a G-
crossed product
L tensor category associated to this action, denoted as C G. We set
C G = C as an abelian category,Lwhere C = C. L We shall denote by [V, ] the
object
L V C , and a morphism from G [V , ] to G [W , ] is expressed as
G [f , ] with f : V W a morphism in C.
 CROSSED PRODUCT TENSOR CATEGORIES 11

The tensor product : C G C G C G is defined by

[V, ] [W, ] := [V (W ) U, , ] for objects, and
[f, ] [g, ] := [f (g) idU, , ] for morphisms.
It is easy to see that the unit object is (I, e). The associativity is given by

[V, ] ([W, ] [Z, ]) [V (W (Z) U, ) U, , ]

[V,],[W, ],[Z,] [CG (V,,W,,Z,), ]

 
([V, ] [W, ]) [Z, ] [V (W ) U, ( ) (Z) U, , ]

where CG (V, , W, , Z, ) is the composition

V (W (Z) U, ) U,

idV W, idU,
(Z)U,

V (W ) ( (Z) U, ) U,

idV (W ) (Z),U idU,

,

V (W ) ( (Z)) (U, ) U,

idV (W ) ( (Z)) ,,

V (W ) ( (Z)) U, U,

idV (W ) , idU,

V (W ) U, ( ) (Z) U,

The associativity constraint have been omitted as a space-saving measure. As we

shall see, the coherence condition over an outer G-action, is exactly the pentagonal
identity for C G.

3.3.1. Pentagonal identity for C G. For a category D with a bifunctor : DD

D and natural isomorphisms A,B,C : A (B C) (A B) C, we shall denote
by P (A, B, C, D) the following pentagonal diagram

A (B (C D))
o OOO
A,B,CD ooo
ooo OOO
OOOidB,C,D
ooo OOO
ooo OOO
wooo O'
(A B) (C D) A ((B C) D)
 
 
AB,C,D   A,BC,D
 
 A,B,C id

((A B) C) D o (A (B C)) D
12
CESAR GALINDO

Remark 3.7. From now on, we shall denote [V ] := [V, e] and [] := [I, ],
for all V C, G. Analogously, [f ] := [f, e] : [V ] [W ] for all arrow
f : V W in C. Note that [V ] [] = [V, ] and [] [V ] = [ (V ), ] for all
V C, G.
In order to prove the coherence of C G, is sufficient to see the pentagonal
identity for the [V ], [], V C, G, since every object in C G is a direct sum
of tensor products of [V ], [].
First, see the next equality

(3.2) [V,],[W, ],[Z,] = id[V ] [I,],[W, ],[Z,] ,

so [V,e],[W, ],[Z,] = id. The eight pentagonal identities of the form P ([V ], . . .)
follows from (3.2).
The pentagon P ([], [V ], [W ], [ ]) commutes because
[],[W ],[Z,] = [],[W ],[Z] id[] .
The pentagons P ([], [V ], [ ], [W ]), P ([], [ ], []), P ([], [ ], [V ], []) commute by
the definition of .
The following table explains the commutativity of the other pentagons
Pentagons Pentagonal identity equivalence
P ([], [V ], [W ], [Z]) ( , ) is a monoidal functor
P ([], [ ], [V ], [W ]) (, , U, ) is a pseudonatural equivalence
P ([], [ ], [], [V ]) ,, is a modification
P ([], [ ], [], []) commutativity of the diagram (3.1).

3.4. The coherent outer G-action associated to a G-crossed product ten-

sor category. Let C be a G-crossed product tensor category. In order to show
more clearly the associated coherent outer G-action, we shall make some reduc-
tions. Let we choose a family {N }G of homogeneous invertible objects, where
Ne = I. The family {N }G defines the equivalences of categories
N () : Ce C
V 7 V N
f 7 f idN .
Using these equivalences, we have an equivalence of categories
M M
C= C Ce ,
G G

where Ce = Ce , for all G.

L
Now, we can transport the monoidal structure of C to G Ce . Then,Lwithout
loss of generality we can suppose that the graded tensor category C = G C
has the following properties:
C = Ce for all G (so we can and will use the same notations of the
Remark 3.7),
the objects [] C are invertible for all G,
[V ] [W, ] = [V W, ], for all V, W Ce , G.
 CROSSED PRODUCT TENSOR CATEGORIES 13

For each pair , G, we have that [][ ] C , so there is an unique invertible

object U Ce , such that [] [ ] = [U, , ]. Analogously, the objects [] define
functors : Ce Ce , V 7 (V ) by the rule [] [V ] = [ (V ), ] for all V Ce ,
and id[] [f ] = [ (f ), ], for all arrow f in Ce .
Lemma 3.8. If the category (Ce , , I) is skeletal, then
[V, ] [W, ] = [V (W ) U, , ]
for all V, W Ce , , G.
L
Proof. Since Ce is skeletal, the category C = C is skeletal. Then we do not
need to parenthesize tensor products for objects in C. Also, recall that [] [V ] =
[ (V ), ] = [ (V )] [], for all V Ce , G.
Hence,
[V, ] [W, ] = [V ] [] [W ] [ ]
= [V ] [ (W )] [] [ ]
= [W (W )] [U, , ]
= [W (W )] [U, ] [ ]
= [W (W ) U, ] [ ]
= [W (W ) U, , ]

for all V, W Ce , G. 

Under this reduction and using the Lemma 3.8 we can describe the coherent
outer G-action as the reciprocal construction of the Subsection 3.3. Suppose that
Ce is skeletal, then the data that define the coherent outer G-action associated to
C are the following:
monoidal equivalences: ( , ) : Ce Ce , where

[W,Z , ] := [],[W ],[Z] : [ (W Z), ] [ (W ) (Z), ],
pseudonatural transformations: (U, , , ) : ( ) , where
[, (Z), ] := [],[ ],[Z] : [ ( (Z)) U, , ] [U, ( ) (Z), ],
modifications , : , (id , ) , (, id ), where
[,, , ] := [],[ ],[] : [ (U, ) U, , ] [U, U, , ].

4. Classification of crossed product tensor category

In this section, we shall see that the 2-category of crossed product tensor category
over a fixed group G, is equivalent to the 2-category of all coherent outer G-actions.
The 0-cells of the 2-category of coherent outer G-actions are coherent outer G-
action over a tensor category.
Let ({e }G , , ) be coherent outer G-actions over tensor
}G , , ) and ({b
categories C and D, respectively. An arrow from ({e }G , , ),
}G , , ) to ({b
is a triple (H, , ), where (H, ) : C D is an op-monoidal functor, ( , ) :
H

b H H e
is a pseudonatural equivalence for each G, and is a modification
14
CESAR GALINDO

b? H e?

 ??
id
b  ?? ide
 ??
 ?

bbH
, Hee
, idH idH ,


c H / H f

such that: (e , e ) = (I, id), ,e = e, = id for all G, and the diagram

DG ( ,, ,, ,)
b( b( )U,
)U,
/ b( )U, c ( )U,

id b
(, )idU , idd
( )U
, ,
 

b(H(U, ) )U,
H(U, ) c ( )U,
b

id H(U, ),
idU idH(U, ) ,
,
 

b(H(U, ))b
( )U, H(U, )H(U, )
H
U ,
id
b ( )U U id
, , ,U,
 

H(e b( )U,
(U, )) H(U, U, )
O
idH(e
(U, )) , H(,, )
 H

e (U, ),U,
(U, )) H(U, )
H(e / H( (U, )U, )

commutes for all , , G (where tensor symbols among objects have been omitted
as a space-saving measure).
e ,
A 2-cell from (H, , ) to (H, e )
e consist of the data {m , m}G , where m :
H H e is a monoidal natural transformation and m : e are morphisms
in C. The previous data are subject to the following axioms: me = idI and the
diagrams
(m )idU
m b
(mV )
m b

b( )U,
/ e,
b(e )U,

b (H(V ))
/ e e
b (H(V ))

, e ,
V eV

   m

mU, m e (V ) m
H(U, ) e , )e
/ H(U (V ))
H(e e (V ))e
/ H(e

commute for all , G, V C (where tensor symbols among objects have been
omitted).
Theorem 4.1. There is a biequivalence between the 2-category of coherent outer
G-actions and the 2-category of G-crossed product tensor categories.
 CROSSED PRODUCT TENSOR CATEGORIES 15

Proof. The bijective correspondence between G-crossed product tensor categories

and coherent outer G-action was described in the Section 3.
If T = (H, , ) is a 1-cell between coherent outer G-action ({e }G , ) and
}G , , ) over C and D respectively, then we define an op-monoidal functor
({b
(T, T ) : C G D G as
T ([V, ]) = [H(V ) , ], T ([f, ]) = [H(f ) id , ] for all V C, G,
T : T ([V, ]) T ([W, ]) T ([V, ] [W, ]),
where

T ([V, ]) T ([W, ]) b (H(W ) )U,
H(V )

b

idH(V ) H(W ), idU ,



b(H(W ))b
H(V ) ( )U,

idH(V ) W id
b ( )U
,



T
[V,],[W, ] H(V )H(e b( )U,
(W ))

H
V,e
(W ) ,


e(W ))H(U, )
H(V

H
V e (W ),U, id

 
T ([V, ] [W, ]) e(W )U, )
H(V

(where tensor symbols among objects of C have been omitted as a space-saving

measure). Conversely, given a graded op-monoidal functor (T, T ) : C G
D G, we define a 1-cell (H, , ) as [H(V ), e] = T ([V, e]), [ , ] = T ([I, ]),
[V , ] = [I,],[V,e]
T T
, [, , ] = [I,],[I, ].

Given a 2-cell {m , m}G between 1-cells T = (H, , ) and T = (H , , ),

we define a monoidal natural isomorphism m : T T between the associated op-
monoidal functors by m[V,] = [mV m , ]. Conversely, given a monoidal natural
isomorphism m : T T , we define a 2-cell by [mV , e] = m[V,e] , [m , ] = m[I,] .
Finally, in order to see that the 2-categories are biequivalent, note that every
crossed product tensor category is equivalent to one of the form Ce G. So, every
functor between C G and D G is monoidally equivalent to one induced by
a 1-cell of the coherent outer G-action 2-category, and every monoidal natural
transformation is equal to one induced by a 2-cell. 

5. Braided crossed product tensor categories

Recall that a braiding for a monoidal category (C, , I, ) is a natural isomor-
phism c : , where : C C C C is the flip, and the hexagons
16
CESAR GALINDO

(5.1)
cU V,W
(U V ) W / W (U V )
aU,V,W jjj4 TTTT aW,U,V
jj TTTT
jjjj T*
U (V W ) (W U ) V
TTTT
TTTT jjjjjj4
idU cV,W T*
aU,W,V
jjj cU,W idV
U (W V ) / (U W ) V

(5.2)
cU,V W
U (V W ) / (V W ) U
a1 TTTT a1
U,V,W
jjjjjj4 TTV,W,U
TTT*
jjj
(U V ) W V (W U )
TTTT jjj4
TTTT j
jjjid
cU,V idW T* 1
aV,U,W jj V cU,W
(V U ) W / V (U W )

commute for all U, V, W C.

If a G-crossed product tensor category admits a braiding, the group G must be
abelian. So, from now on we shall suppose that G is abelian.
Let C be a tensor category with a coherent outer G-action, such that the tensor
category C G admits a braiding c. The braiding c[V ],[] : [V, ] [ (V ), ] defines
natural isomorphisms cV, : V (V ). The commutativity of the hexagon (5.1)
is equivalent to cV, is a monoidal natural isomorphism from idC to . For that
reason, if C G has a braiding, we can suppose that = idC for all G.

Definition 5.1. A coherent outer G-action shall be called central if = idC for
all G.

Remark 5.2. For a central coherent outer G-action, the pseudonatural transforma-
tions (, , U, ) are just elements in Z(C) (the center of C), and the modifications
are morphisms in Z(C).

Definition 5.3. Let (C, c) be a braided tensor category, and let G be an abelian

group. A braiding for a central coherent G-action over C is a triple ( , , t, ), G ,
where

, : idC idC are monoidal natural isomorphisms,
t, : , , are isomorphisms in Z(C) for all , G,
e
such that e = = id, I = idI , t,e = te, = idI , and the diagrams

, (Z)
Z U, / U, Z
JJ
JJ
JJ cU, ,Z
(5.3)
JJ
((Z )1 Z Z )idU, JJ
J$ 
Z U,
 CROSSED PRODUCT TENSOR CATEGORIES 17

cU, ,Z
Z U, / U, Z
JJ
JJ
JJ
(5.4) 1
JJ
JJ
, (Z)
((Z ) Z Z )idU, J$ 
Z U,

U ,
t,
U, U, / U, U,
,,
oooo7 OOO w,,
OOO
ooo O'
(5.5) U, U, U, U,
OOO
OOO oooo7
t, idU, O'
,,
ooo t, idU,
U, U, / U, U,

U, t,
U, U, / U, U ,
1 1
,,
ooo7 OOO ,,
OOO
o
ooo O'
(5.6) U, U, U, U,
OOO
OOO ooo7
o
t, idU, O' 1
,, ooo t, idU,
U, U , / U, U,

commute for all , , G, Z C (where tensor symbols among objects have been
omitted as a space-saving measure).
Theorem 5.4. Let (C, c) be a braided tensor category with a coherent central outer
G-action. Then, there is a bijective correspondence between braidings over C G
and braidings over the central coherent outer G-action of C.
Proof. Let (, , t) be a braiding for a central coherent outer G-action (, ). Then,
we define a braiding over C G by

c[V,],[W, ] = (cV,W (V W )) t,

= (( W V ) cV,W ) t, .
Conversely, given a braiding c over C G, we define a braiding for the coherent
central outer G-action by

[V , ] := c[V ],[] , [V , ] := c[],[V ] , and [t, , ] := c[],[ ] .
Let we denote by H(U, V, W ) and H (U, V, W ) the hexagons (5.1) and (5.2),

respectively. Let , : idC idC be natural isomorphisms for each G, and
let t, : U, U, be isomorphisms in C. If we set the following definitions of
natural isomorphisms

c[V ],[] := [V , ], c[],[V ] := [V , ], c[],[ ] := [t, , ],
it is easy to see that the commutativity of H([V ], [W ], ) and H ([], [V ], [W ])

is equivalent to and be monoidal natural isomorphisms, respectively. The
commutativity of H ([], [ ], [Z]) and H([], [V ], [ ]) is equivalent to t, be a mor-
phism in Z(C). The commutativity of H([], [ ], [Z]) and H([Z], [], [ ]) is equiv-
alent to the commutativity of (5.3) and (5.4), respectively. The commutativity of
18
CESAR GALINDO

H([], [ ], []) and H ([], [ ], []) is equivalent to the commutativity of (5.5), (5.6),
respectively.


References
[1] J. Baez and A. D. Lauda, Higher-dimensional algebra V: 2-groups, Theor. and Appl. Cat.
12 No. 14 (2004), 423491.
[2] B. Bakalov and A. Kirrilov Jr., Lectures on Tensor categories and modular functors, AMS,
(2001).
[3] J. Benabou, Introduction to bicategories, Reports of the Midwest Category Seminar (1967),
Lecture notes in Math. 47, pp. 177, Springer, Berlin.
[4] C. Curtis and I. Reiner, Methods of representation theory, I, Wiley Interscience Publications,
New York (1990).
[5] P. Etingof, D. Nikshych and V. Ostrik, On fusion categories, Ann. Math. 162 (2005),
581642.
[6] P. Etingof, D. Nikshych and V. Ostrik, Weakly group-theoretical and solvable fusion cat-
egories, preprint arXiv:0809.3031.
[7] P. Etingof, D. Nikshych and V. Ostrik, Fusion categories and homotopy theory, preprint
arXiv:0909.3140.
[8] P. Etingof and V. Ostrik, Finite tensor categories, Mosc. Math. J. 4 (2004), 627654,
782783.
[9] C. Galindo, Cliford theory for tensor categories. to appear in J. London Math Soc., preprint
arXiv:0902.1088.
[10] R. Gordon, A. Power and R. Street Coherence for tricategories, Mem. Am. Math. Soc.
117 (1995), 558.
[11] C. Kassel, Quantum Groups, Graduate Texts in Math. 155, Springer (1995).
[12] C. Na sta
sescu and F. Van Oystaeyen, Methods of Graded Rings. Lecture Notes in Math-
ematics, 1836. Springer-Verlag, Berlin, (2004).
[13] D. Nikshych, Non group-theoretical semisimple Hopf algebras from group actions on fusion
categories. Sel. math., New ser. 14 (2008), 145161.
[14] D. Tambara, Invariants and semi-direct products for finite group actions on tensor cate-
gories, J. Math. Soc. Japan 53 (2001), 429456.

Departamento de Matema ticas

Pontificia Universidad Javeriana
Bogota , Colombia
E-mail address: cesar.galindo@javeriana.edu.co