Generalized (anti) Yetter-Drinfeld modules as components of a braided T-category

Florin PanaiteResearch partially supported by the programme CERES of the Romanian Ministry of Education and Research, contract no. 4-147/2004. Institute of Mathematics of the Romanian Academy PO-Box 1-764, RO-014700 Bucharest, Romania e-mail: Florin.Panaite@imar.ro

Mihai D. StaicPermanent address: Institute of Mathematics of the Romanian Academy, PO-Box 1-764, RO-014700 Bucharest, Romania. SUNY at Buffalo Amherst, NY 14260-2900, USA e-mail: mdstaic@buffalo.edu

Abstract
If H   is a Hopf algebra with bijective antipode and α , β   ∈ A u t H o p f ( H )   , we introduce a category H Y D H ( α , β )   , generalizing both Yetter-Drinfeld modules and anti-Yetter-Drinfeld modules. We construct a braided T-category Y D ( H )   having all the categories H Y D H ( α , β )   as components, which if H   is finite dimensional coincides with the representations of a certain quasitriangular T-coalgebra D T ( H )   that we construct. We also prove that if ( α , β )   admits a so-called pair in involution, then H Y D H ( α , β )   is isomorphic to the category of usual Yetter-Drinfeld modules H Y D H   .
Introduction Let H   be a Hopf algebra with bijective antipode S   and α , β   ∈ A u t H o p f ( H )   . We introduce the concept of an ( α , β )   -Yetter-Drinfeld module, as being a left H   -module right H   -comodule M   with the following compatibility condition:
( h ⋅ m ) ( 0 ) ⊗ ( h ⋅ m ) ( 1 ) = h 2 ⋅ m ( 0 ) ⊗ β ( h 3 ) m ( 1 ) α ( S − 1 ( h 1 ) ) .
This concept is a generalization of three kinds of objects appeared in the literature. Namely, for α = β = i d H   , we obtain the usual Yetter-Drinfeld modules; for α = S 2   , β = i d H   , we obtain the so-called anti-Yetter-Drinfeld modules, introduced in [7] , [8] , [10] as coefficients for the cyclic cohomology of Hopf algebras defined by Connes and Moscovici in [5] , [6] ; finally, an ( i d H , β )   -Yetter-Drinfeld module is a generalization of the object H β   defined in [4] , which has the property that, if H   is finite dimensional, then the map β ↦ E n d ( H β )   gives a group anti-homomorphism from A u t H o p f ( H )   to the Brauer group of H   .
It is natural to expect that ( α , β )   -Yetter-Drinfeld modules have some properties resembling the ones of the three kinds of objects we mentioned. We will see some of these properties in this paper (others will be given in a subsequent one), namely the ones directed to our main aim here, which is the following: if we denote by H Y D H ( α , β )   the category of ( α , β )   -Yetter-Drinfeld modules and we define Y D ( H )   as the disjoint union of all these categories, then we can organize Y D ( H )   as a braided T-category (or braided crossed group-category, in the original terminology of Turaev, see [16] ) over the group G = A u t H o p f ( H ) × A u t H o p f ( H )   with multiplication ( α , β ) * ( γ , δ ) = ( α γ , δ γ − 1 β γ )   . We also prove that the subcategory Y D ( H ) f d   consisting of finite dimensional objects has left and right dualities, and that, if H   is finite dimensional, then Y D ( H )   coincides with the representations of a certain quasitriangular T-coalgebra D T ( H )   that we construct.
Our second aim is to prove that, if α , β   ∈ A u t H o p f ( H )   such that there exists a so-called pair in involution ( f , g )   corresponding to ( α , β )   , then H Y D H ( α , β )   is isomorphic to H Y D H   . This result is independent on the theory concerning Y D ( H )   , but we can give it a very short proof using the results obtained during the construction of Y D ( H )   .

1 Preliminaries

We work over a ground field k   . All algebras, linear spaces, etc. will be over k   ; unadorned ⊗   means ⊗ k   . Unless otherwise stated, H   will denote a Hopf algebra with bijective antipode S   . We will use the versions of Sweedler's sigma notation: Δ ( h ) = h 1 ⊗ h 2   or Δ ( h ) = h ( 1 ) ⊗ h ( 2 )   . For unexplained concepts and notation about Hopf algebras we refer to [11] , [12] , [13] , [15] . By α , β , γ . . .   we will usually denote Hopf automorphisms of H   .
Let A   be an H   -bicomodule algebra, with comodule structures A → A ⊗ H   , a ↦ a < 0 > ⊗ a < 1 >   and A → H ⊗ A   , a ↦ a [ − 1 ] ⊗ a [ 0 ]   , and denote, for a ∈ A   ,
a { − 1 } ⊗ a { 0 } ⊗ a { 1 } = a < 0 > [ − 1 ] ⊗ a < 0 > [ 0 ] ⊗ a < 1 > = a [ − 1 ] ⊗ a [ 0 ] < 0 > ⊗ a [ 0 ] < 1 >
as an element in H ⊗ A ⊗ H   . We can consider the Yetter-Drinfeld datum ( H , A , H )   as in [3] (the second H   is regarded as an H   -bimodule coalgebra), and the Yetter-Drinfeld category A Y D ( H ) H   , whose objects are k   -modules M   endowed with a left A   -action (denoted by a ⊗ m ↦ a ⋅ m   ) and a right H   -coaction (denoted by m ↦ m ( 0 ) ⊗ m ( 1 )   ) satisfying the equivalent compatibility conditions
( a ⋅ m ) ( 0 ) ⊗ ( a ⋅ m ) ( 1 ) = a { 0 } ⋅ m ( 0 ) ⊗ a { 1 } m ( 1 ) S − 1 ( a { − 1 } ) , (1.1)
a < 0 > ⋅ m ( 0 ) ⊗ a < 1 > m ( 1 ) = ( a [ 0 ] ⋅ m ) ( 0 ) ⊗ ( a [ 0 ] ⋅ m ) ( 1 ) a [ − 1 ] , (1.2)
for all a ∈ A   and m ∈ M   .
Recall now from [9] the construction of the (left) diagonal crossed product H * ⋈ A   , which is an associative algebra constructed on H * ⊗ A   , with multiplication given by
( p ⋈ a ) ( q ⋈ b ) = p ( a { − 1 } ⇀ q ↼ S − 1 ( a { 1 } ) ) ⋈ a { 0 } b , (1.3)
for all a , b ∈ A   and p , q ∈ H *   , and with unit ɛ H ⋈ 1 A   . Here ⇀   and ↼   are the regular actions of H   on H *   given by ( h ⇀ p ) ( l ) = p ( l h )   and ( p ↼ h ) ( l ) = p ( h l )   for all h , l ∈ H   and p ∈ H *   .
If H   is finite dimensional, we can consider the Drinfeld double D ( H )   , which is a quasitriangular Hopf algebra realized on H * ⊗ H   ; its coalgebra structure is H * c o p ⊗ H   and the algebra structure is just H * ⋈ H   , that is
( p ⋈ h ) ( q ⋈ l ) = p ( h 1 ⇀ q ↼ S − 1 ( h 3 ) ) ⋈ h 2 l , (1.4)
for all p , q ∈ H *   and h , l ∈ H   .
The diagonal crossed product H * ⋈ A   becomes a D ( H )   -bicomodule algebra, with structures
H * ⋈ A → ( H * ⋈ A ) ⊗ D ( H ) , p ⋈ a ↦ ( p 2 ⋈ a < 0 > ) ⊗ ( p 1 ⊗ a < 1 > ) ,
H * ⋈ A → D ( H ) ⊗ ( H * ⋈ A ) , p ⋈ a ↦ ( p 2 ⊗ a [ − 1 ] ) ⊗ ( p 1 ⋈ a [ 0 ] ) ,
for all p ∈ H *   and a ∈ A   , see [9] .
In the case when H   is finite dimensional, by results in [1] , [3] it follows that the category A Y D ( H ) H   is isomorphic to the category H * ⋈ A ℳ   of left modules over H * ⋈ A   .

2 ( α , β )   -Yetter-Drinfeld modules

Definition 2.1 Let α , β   ∈ A u t H o p f ( H )   . An ( α , β )   -Yetter-Drinfeld module over H   is a vector space M   , such that M   is a left H   -module (with notation h ⊗ m ↦ h ⋅ m   ) and a right H   -comodule (with notation M → M ⊗ H   , m ↦ m ( 0 ) ⊗ m ( 1 )   ) with the following compatibility condition:
( h ⋅ m ) ( 0 ) ⊗ ( h ⋅ m ) ( 1 ) = h 2 ⋅ m ( 0 ) ⊗ β ( h 3 ) m ( 1 ) α ( S − 1 ( h 1 ) ) , (2.1)
for all h ∈ H   and m ∈ M   . We denote by H Y D H ( α , β )   the category of ( α , β )   -Yetter-Drinfeld modules, morphisms being the H   -linear H   -colinear maps.
Remark 2.2 As for usual Yetter-Drinfeld modules, one can see that ( 2.1 ) is equivalent to
h 1 ⋅ m ( 0 ) ⊗ β ( h 2 ) m ( 1 ) = ( h 2 ⋅ m ) ( 0 ) ⊗ ( h 2 ⋅ m ) ( 1 ) α ( h 1 ) . (2.2)
Example 2.3 For α = β = i d H   , we have H Y D H ( i d , i d ) =   H Y D H   , the usual category of (left-right) Yetter-Drinfeld modules.
Example 2.4 For α = S 2   , β = i d H   , the compatibility condition ( 2.1 ) becomes
( h ⋅ m ) ( 0 ) ⊗ ( h ⋅ m ) ( 1 ) = h 2 ⋅ m ( 0 ) ⊗ h 3 m ( 1 ) S ( h 1 ) , (2.3)
hence H Y D H ( S 2 , i d )   is the category of anti-Yetter-Drinfeld modules defined in [7] , [8] , [10] .
Example 2.5 For β   ∈ A u t H o p f ( H )   , define H β   as in [4] , that is H β = H   , with regular right H   -comodule structure and left H   -module structure given by h ⋅ h ′ = β ( h 2 ) h ′ S − 1 ( h 1 )   , for all h , h ′ ∈ H   . It was noticed in [4] that H β   satisfies a certain compatibility condition, which actually says that H β ∈   H Y D H ( i d , β )   . More generally, if α , β   ∈ A u t H o p f ( H )   , define H α , β   as follows:
H α , β = H   , with regular right H   -comodule structure and left H   -module structure given by h ⋅ h ′ = β ( h 2 ) h ′ α ( S − 1 ( h 1 ) )   , for h , h ′ ∈ H   . Then one can check that H α , β ∈   H Y D H ( α , β )   .
Example 2.6 Take l   an integer and define α l = S 2 l   ∈ A u t H o p f ( H )   . The compatibility in H Y D H ( S 2 l , i d )   becomes
( h ⋅ m ) ( 0 ) ⊗ ( h ⋅ m ) ( 1 ) = h 2 ⋅ m ( 0 ) ⊗ h 3 m ( 1 ) S 2 l − 1 ( h 1 ) . (2.4)
An object in H Y D H ( S 2 l , i d )   will be called an l − Y D   -module. Hence, a 0 − Y D   -module is a Yetter-Drinfeld module and a 1 − Y D   -module is an anti-Yetter-Drinfeld module. The right-left version of l − Y D   -modules has been introduced in [14] .
Example 2.7 Let α   , β   ∈ A u t H o p f ( H )   and assume that there exist an algebra map f : H → k   and a group-like element g ∈ H   such that
α ( h ) = g − 1 f ( h 1 ) β ( h 2 ) f ( S ( h 3 ) ) g , ∀ h ∈ H . (2.5)
Then one can check that k ∈   H Y D H ( α , β )   , with structures h ⋅ 1 = f ( h )   and 1 ↦ 1 ⊗ g   . More generally, if V   is any vector space, then V ∈   H Y D H ( α , β )   , with structures h ⋅ v = f ( h ) v   and v ↦ v ⊗ g   , for all h ∈ H   and v ∈ V   .
Definition 2.8 If α   , β   ∈ A u t H o p f ( H )   such that there exist f , g   as in Example  2.7 , we will say that ( f , g )   is a pair in involution corresponding to ( α , β )   (in analogy with the concept of modular pair in involution due to Connes and Moscovici) and the ( α , β )   -Yetter-Drinfeld modules k   and V   constructed in Example  2.7 will be denoted by f k g   and respectively f V g   .
As an example, if α   ∈ A u t H o p f ( H )   , then ( ɛ , 1 )   is a pair in involution corresponding to ( α , α )   .
Let α , β   ∈ A u t H o p f ( H )   . We define an H   -bicomodule algebra H ( α , β )   as follows: H ( α , β ) = H   as algebra, with comodule structures
H ( α , β ) → H ⊗ H ( α , β ) , h ↦ h [ − 1 ] ⊗ h [ 0 ] = α ( h 1 ) ⊗ h 2 ,
H ( α , β ) → H ( α , β ) ⊗ H , h ↦ h < 0 > ⊗ h < 1 > = h 1 ⊗ β ( h 2 ) .
Then we can consider the Yetter-Drinfeld datum ( H , H ( α , β ) , H )   and the Yetter-Drinfeld modules over it, H ( α , β ) Y D ( H ) H   .
Proposition 2.9 H Y D H ( α , β ) =   H ( α , β ) Y D ( H ) H   .
Proof. It is easy to see that the compatibility conditions for the two categories are the same. □  
In particular, the category of anti-Yetter-Drinfeld modules coincides with H ( S 2 , i d ) Y D ( H ) H   , which improves the remark in [7] that anti-Yetter-Drinfeld modules are entwined modules.
Consider now the diagonal crossed product A ( α , β ) = H * ⋈ H ( α , β )   , whose multiplication is
( p ⋈ h ) ( q ⋈ l ) = p ( α ( h 1 ) ⇀ q ↼ S − 1 ( β ( h 3 ) ) ) ⋈ h 2 l , (2.6)
for all p , q ∈ H *   and h , l ∈ H   . For α = β = i d   we get A ( i d , i d ) = D ( H )   ; for α = S 2   and β = i d   , the multiplication in A ( S 2 , i d )   is
( p ⋈ h ) ( q ⋈ l ) = p ( S 2 ( h 1 ) ⇀ q ↼ S − 1 ( h 3 ) ) ⋈ h 2 l , (2.7)
hence A ( S 2 , i d )   coincides with the algebra A ( H )   defined in [7] .
Assume now that H   is finite dimensional; then A ( α , β )   becomes a D ( H )   -bicomodule algebra, with structures
H * ⋈ H ( α , β ) → ( H * ⋈ H ( α , β ) ) ⊗ D ( H ) , p ⋈ h ↦ ( p 2 ⋈ h 1 ) ⊗ ( p 1 ⊗ β ( h 2 ) ) ,
H * ⋈ H ( α , β ) → D ( H ) ⊗ ( H * ⋈ H ( α , β ) ) , p ⋈ h ↦ ( p 2 ⊗ α ( h 1 ) ) ⊗ ( p 1 ⋈ h 2 ) .
In particular, A ( H )   becomes a D ( H )   -bicomodule algebra, improving the remark in [7] that A ( H )   is a right D ( H )   -comodule algebra. Since H   is finite dimensional, we have an isomorphism of categories H ( α , β ) Y D ( H ) H ≃   H * ⋈ H ( α , β ) ℳ   , hence H Y D H ( α , β ) ≃   H * ⋈ H ( α , β ) ℳ   (for α = S 2   , β = i d   we recover the result in [7] that the category of anti-Yetter-Drinfeld modules is isomorphic to A ( H ) ℳ   ). The correspondence is given as follows. If M ∈   H Y D H ( α , β )   , then M ∈   H * ⋈ H ( α , β ) ℳ   with structure
( p ⋈ h ) ⋅ m = p ( ( h ⋅ m ) ( 1 ) ) ( h ⋅ m ) ( 0 ) .
Conversely, if M ∈   H * ⋈ H ( α , β ) ℳ   , then M ∈   H Y D H ( α , β )   with structures
h ⋅ m = ( ɛ ⋈ h ) ⋅ m ,
m ↦ m ( 0 ) ⊗ m ( 1 ) = ( e i ⋈ 1 ) ⋅ m ⊗ e i ,
where { e i }   , { e i }   are dual bases in H   and H *   .

3 A braided T-category Y D ( H )  

Let α , β   ∈ A u t H o p f ( H )   and consider the objects H α , H β   as in Example  2.5 . In [4] was considered the object M = H α ⊗ H β   , with the following structures:
h ⋅ ( x ⊗ y ) = h 1 ⋅ x ⊗ α ( h 2 ) ⋅ y ,
x ⊗ y ↦ ( x 1 ⊗ y 1 ) ⊗ y 2 x 2 ,
for all h , x , y ∈ H   , where by ⋅   we denoted both the actions of H   on H α   and H β   given as in Example  2.5 . Then it was noticed in [4] that M   satisfies a compatibility condition which says that M ∈   H Y D H ( i d , β α )   .
On the other hand, it was noticed in [7] that the tensor product between an anti-Yetter-Drinfeld module and a Yetter-Drinfeld module becomes an anti-Yetter-Drinfeld module.
The next result can be seen as a generalization of both these facts.
Proposition 3.1 If M ∈   H Y D H ( α , β )   , N ∈   H Y D H ( γ , δ )   , then M ⊗ N ∈   H Y D H ( α γ , δ γ − 1 β γ )   , with structures:
h ⋅ ( m ⊗ n ) = γ ( h 1 ) ⋅ m ⊗ γ − 1 β γ ( h 2 ) ⋅ n ,
m ⊗ n ↦ ( m ⊗ n ) ( 0 ) ⊗ ( m ⊗ n ) ( 1 ) = ( m ( 0 ) ⊗ n ( 0 ) ) ⊗ n ( 1 ) m ( 1 ) .
Proof. Obviously M ⊗ N   is a left H   -module and a right H   -comodule. We check now the compatibility condition. We compute:
( h ⋅ ( m ⊗ n ) ) ( 0 ) ⊗ ( h ⋅ ( m ⊗ n ) ) ( 1 )  
= ( γ ( h 1 ) ⋅ m ⊗ γ − 1 β γ ( h 2 ) ⋅ n ) ( 0 ) ⊗ ( γ ( h 1 ) ⋅ m ⊗ γ − 1 β γ ( h 2 ) ⋅ n ) ( 1 )
= ( ( γ ( h 1 ) ⋅ m ) ( 0 ) ⊗ ( γ − 1 β γ ( h 2 ) ⋅ n ) ( 0 ) ) ⊗ ( γ − 1 β γ ( h 2 ) ⋅ n ) ( 1 ) ( γ ( h 1 ) ⋅ m ) ( 1 )
= ( γ ( h 1 ) 2 ⋅ m ( 0 ) ⊗ γ − 1 β γ ( h 2 ) 2 ⋅ n ( 0 ) ) ⊗
⊗ δ ( γ − 1 β γ ( h 2 ) 3 ) n ( 1 ) γ ( S − 1 ( γ − 1 β γ ( h 2 ) 1 ) ) β ( γ ( h 1 ) 3 ) m ( 1 ) α ( S − 1 ( γ ( h 1 ) 1 ) )
= ( γ ( h 2 ) ⋅ m ( 0 ) ⊗ γ − 1 β γ ( h 5 ) ⋅ n ( 0 ) ) ⊗ δ γ − 1 β γ ( h 6 ) n ( 1 ) β γ ( S − 1 ( h 4 ) ) β γ ( h 3 ) m ( 1 ) α ( S − 1 ( γ ( h 1 ) ) )
= ( γ ( h 2 ) ⋅ m ( 0 ) ⊗ γ − 1 β γ ( h 3 ) ⋅ n ( 0 ) ) ⊗ δ γ − 1 β γ ( h 4 ) n ( 1 ) m ( 1 ) α γ ( S − 1 ( h 1 ) )
= h 2 ⋅ ( m ( 0 ) ⊗ n ( 0 ) ) ⊗ δ γ − 1 β γ ( h 3 ) n ( 1 ) m ( 1 ) α γ ( S − 1 ( h 1 ) )
= h 2 ⋅ ( m ⊗ n ) ( 0 ) ⊗ δ γ − 1 β γ ( h 3 ) ( m ⊗ n ) ( 1 ) α γ ( S − 1 ( h 1 ) ) ,
that is M ⊗ N ∈   H Y D H ( α γ , δ γ − 1 β γ )   . □  
Note that, if M ∈   H Y D H ( α , β )   , N ∈   H Y D H ( γ , δ )   and P ∈   H Y D H ( μ , ν )   , then ( M ⊗ N ) ⊗ P = M ⊗ ( N ⊗ P )   as objects in H Y D H ( α γ μ , ν μ − 1 δ γ − 1 β γ μ )   .
Denote G = A u t H o p f ( H ) × A u t H o p f ( H )   , a group with multiplication
( α , β ) * ( γ , δ ) = ( α γ , δ γ − 1 β γ ) (3.1)
(the unit is ( i d , i d )   and ( α , β ) − 1 = ( α − 1 , α β − 1 α − 1 ) )   .
Proposition 3.2 Let N ∈   H Y D H ( γ , δ )   and ( α , β ) ∈ G   . Define ( α , β ) N = N   as vector space, with structures
h ⇀ n = γ − 1 β γ α − 1 ( h ) ⋅ n ,
n ↦ n < 0 > ⊗ n < 1 > = n ( 0 ) ⊗ α β − 1 ( n ( 1 ) ) .
Then ( α , β ) N ∈   H Y D H ( α γ α − 1 , α β − 1 δ γ − 1 β γ α − 1 ) =   H Y D H ( ( α , β ) * ( γ , δ ) * ( α , β ) − 1 )   .
Proof. Obviously ( α , β ) N   is a left H   -module and right H   -comodule, so we check the compatibility condition. We compute:
( h ⇀ n ) < 0 > ⊗ ( h ⇀ n ) < 1 >  
= ( γ − 1 β γ α − 1 ( h ) ⋅ n ) ( 0 ) ⊗ α β − 1 ( ( γ − 1 β γ α − 1 ( h ) ⋅ n ) ( 1 ) )
= γ − 1 β γ α − 1 ( h 2 ) ⋅ n ( 0 ) ⊗ α β − 1 ( δ γ − 1 β γ α − 1 ( h 3 ) n ( 1 ) γ γ − 1 β γ α − 1 ( S − 1 ( h 1 ) ) )
= γ − 1 β γ α − 1 ( h 2 ) ⋅ n ( 0 ) ⊗ α β − 1 δ γ − 1 β γ α − 1 ( h 3 ) α β − 1 ( n ( 1 ) ) α γ α − 1 ( S − 1 ( h 1 ) )
= h 2 ⇀ n ( 0 ) ⊗ α β − 1 δ γ − 1 β γ α − 1 ( h 3 ) n < 1 > α γ α − 1 ( S − 1 ( h 1 ) ) ,
that is ( α , β ) N ∈   H Y D H ( α γ α − 1 , α β − 1 δ γ − 1 β γ α − 1 )   . □  
Remark 3.3 Let M ∈   H Y D H ( α , β )   , N ∈   H Y D H ( γ , δ )   and ( μ , ν ) ∈ G   . Then we have
( α , β ) * ( μ , ν ) N = ( α , β ) ( ( μ , ν ) N )
as objects in H Y D H ( α μ γ μ − 1 α − 1 , α β − 1 μ ν − 1 δ γ − 1 ν μ − 1 β μ γ μ − 1 α − 1 )   , and
( μ , ν ) ( M ⊗ N ) = ( μ , ν ) M ⊗ ( μ , ν ) N
as objects in H Y D H ( μ α γ μ − 1 , μ ν − 1 δ γ − 1 β α − 1 ν α γ μ − 1 )   .
Proposition 3.4 Let M ∈   H Y D H ( α , β )   and N ∈   H Y D H ( γ , δ )   . Define M N = ( α , β ) N   as object in H Y D H ( ( α , β ) * ( γ , δ ) * ( α , β ) − 1 )   . Define the map
c M , N : M ⊗ N → M N ⊗ M , c M , N ( m ⊗ n ) = n ( 0 ) ⊗ β − 1 ( n ( 1 ) ) ⋅ m .
Then c M , N   is H   -linear H   -colinear and satisfies the conditions (for P ∈   H Y D H ( μ , ν )   ):
c M ⊗ N , P = ( c M , N P ⊗ i d N ) ∘ ( i d M ⊗ c N , P ) , (3.2)
c M , N ⊗ P = ( i d M N ⊗ c M , P ) ∘ ( c M , N ⊗ i d P ) . (3.3)
Moreover, if M ∈   H Y D H ( α , β )   , N ∈   H Y D H ( γ , δ )   and ( μ , ν ) ∈ G   , then c ( μ , ν ) M , ( μ , ν ) N = c M , N   .
Proof. We prove that c M , N   is H   -linear. We compute:
c M , N ( h ⋅ ( m ⊗ n ) )  
= c M , N ( γ ( h 1 ) ⋅ m ⊗ γ − 1 β γ ( h 2 ) ⋅ n )
= ( γ − 1 β γ ( h 2 ) ⋅ n ) ( 0 ) ⊗ β − 1 ( ( γ − 1 β γ ( h 2 ) ⋅ n ) ( 1 ) ) γ ( h 1 ) ⋅ m
= γ − 1 β γ ( h 2 ) 2 ⋅ n ( 0 ) ⊗ β − 1 ( δ ( γ − 1 β γ ( h 2 ) 3 ) n ( 1 ) γ ( S − 1 ( ( γ − 1 β γ ( h 2 ) ) 1 ) ) ) γ ( h 1 ) ⋅ m
= γ − 1 β γ ( h 3 ) ⋅ n ( 0 ) ⊗ β − 1 δ γ − 1 β γ ( h 4 ) β − 1 ( n ( 1 ) ) γ ( S − 1 ( h 2 ) ) γ ( h 1 ) ⋅ m
= γ − 1 β γ ( h 1 ) ⋅ n ( 0 ) ⊗ β − 1 δ γ − 1 β γ ( h 2 ) β − 1 ( n ( 1 ) ) ⋅ m ,
h ⋅ c M , N ( m ⊗ n ) = h ⋅ ( n ( 0 ) ⊗ β − 1 ( n ( 1 ) ) ⋅ m )
= α ( h 1 ) ⇀ n ( 0 ) ⊗ α − 1 α β − 1 δ γ − 1 β γ α − 1 α ( h 2 ) β − 1 ( n ( 1 ) ) ⋅ m
= γ − 1 β γ α − 1 α ( h 1 ) ⋅ n ( 0 ) ⊗ β − 1 δ γ − 1 β γ ( h 2 ) β − 1 ( n ( 1 ) ) ⋅ m
= γ − 1 β γ ( h 1 ) ⋅ n ( 0 ) ⊗ β − 1 δ γ − 1 β γ ( h 2 ) β − 1 ( n ( 1 ) ) ⋅ m ,
so the two terms are equal. The fact that c M , N   is H   -colinear is similar and left to the reader. We prove now ( 3.2 ). First note that, due to Remark  3.3 , we have M ( N P ) =   M ⊗ N P   and M ( N ⊗ P ) =   M N ⊗ M P   . We compute:
( c M , N P ⊗ i d N ) ∘ ( i d M ⊗ c N , P ) ( m ⊗ n ⊗ p ) = c M , N P ( m ⊗ p ( 0 ) ) ⊗ δ − 1 ( p ( 1 ) ) ⋅ n
= p ( 0 ) < 0 > ⊗ β − 1 ( p ( 0 ) < 1 > ) ⋅ m ⊗ δ − 1 ( p ( 1 ) ) ⋅ n
= p ( 0 ) ( 0 ) ⊗ β − 1 γ δ − 1 ( p ( 0 ) ( 1 ) ) ⋅ m ⊗ δ − 1 ( p ( 1 ) ) ⋅ n
= p ( 0 ) ⊗ β − 1 γ δ − 1 ( p ( 1 ) 1 ) ⋅ m ⊗ δ − 1 ( p ( 1 ) 2 ) ⋅ n ,
c M ⊗ N , P ( m ⊗ n ⊗ p ) = p ( 0 ) ⊗ γ − 1 β − 1 γ δ − 1 ( p ( 1 ) ) ⋅ ( m ⊗ n )
= p ( 0 ) ⊗ γ γ − 1 β − 1 γ δ − 1 ( p ( 1 ) 1 ) ⋅ m ⊗ γ − 1 β γ γ − 1 β − 1 γ δ − 1 ( p ( 1 ) 2 ) ⋅ n
= p ( 0 ) ⊗ β − 1 γ δ − 1 ( p ( 1 ) 1 ) ⋅ m ⊗ δ − 1 ( p ( 1 ) 2 ) ⋅ n ,
and we are done. The proof of ( 3.3 ) is easier and left to the reader, and similarly the last statement of the Proposition. □  
Note that c M , N   is bijective with inverse c M , N − 1 ( n ⊗ m ) = β − 1 ( S ( n ( 1 ) ) ) ⋅ m ⊗ n ( 0 )   .
We are ready now to introduce the desired braided T-category (we use terminology as in [18] ; for the subject of Turaev categories, see also the original paper of Turaev [16] and [2] , [17] ).
Define Y D ( H )   as the disjoint union of all H Y D H ( α , β )   , with ( α , β ) ∈ G   (hence the component of the unit is just H Y D H   ). If we endow Y D ( H )   with tensor product as in Proposition  3.1 , then it becomes a strict monoidal category with unit k   as object in H Y D H   (with trivial structures).
The group homomorphism φ : G → a u t ( Y D ( H ) )   , ( α , β ) ↦ φ ( α , β )   , is given on components as
φ ( α , β ) : H Y D H ( γ , δ ) → H Y D H ( ( α , β ) * ( γ , δ ) * ( α , β ) − 1 ) , φ ( α , β ) ( N ) = ( α , β ) N ,
and the functor φ ( α , β )   acts as identity on morphisms. The braiding in Y D ( H )   is given by the family { c M , N }   . As a consequence of the above results, we obtain:
Theorem 3.5 Y D ( H )   is a braided T-category over G   .
We consider now the problem of existence of left and right dualities.
Proposition 3.6 Let M ∈   H Y D H ( α , β )   and assume that M   is finite dimensional. Then M * = H o m ( M , k )   becomes an object in H Y D H ( α − 1 , α β − 1 α − 1 )   , with ( h ⋅ f ) ( m ) = f ( ( β − 1 α − 1 S ( h ) ) ⋅ m )   and f ( 0 ) ( m ) ⊗ f ( 1 ) = f ( m ( 0 ) ) ⊗ S − 1 ( m ( 1 ) )   . Moreover, the maps b M : k → M ⊗ M *   , b M ( 1 ) = ∑ i e i ⊗ e i   (where { e i }   and { e i }   are dual bases in M   and M *   ) and d M : M * ⊗ M → k   , d M ( f ⊗ m ) = f ( m )   , are morphisms in H Y D H   and we have ( i d M ⊗ d M ) ( b M ⊗ i d M ) = i d M   and ( d M ⊗ i d M * ) ( i d M * ⊗ b M ) = i d M *   .
Proof. We first prove that M *   is indeed an object in H Y D H ( α − 1 , α β − 1 α − 1 )   . We compute:
( h ⋅ f ) ( 0 ) ( m ) ⊗ ( h ⋅ f ) ( 1 ) = ( h ⋅ f ) ( m ( 0 ) ) ⊗ S − 1 ( m ( 1 ) )
= f ( ( β − 1 α − 1 S ) ( h ) ⋅ m ( 0 ) ) ⊗ S − 1 ( m ( 1 ) ) ,
( h ( 2 ) ⋅ f ( 0 ) ) ( m ) ⊗ ( α β − 1 α − 1 ) ( h ( 3 ) ) f ( 1 ) ( α − 1 S − 1 ) ( h ( 1 ) )  
= f ( 0 ) ( β − 1 α − 1 S ( h ( 2 ) ) ⋅ m ) ⊗ ( α β − 1 α − 1 ) ( h ( 3 ) ) f ( 1 ) ( α − 1 S − 1 ( h ( 1 ) ) )
= f ( ( ( β − 1 α − 1 S ) ( h ( 2 ) ) ⋅ m ) ( 0 ) ) ⊗ ( α β − 1 α − 1 ( h ( 3 ) ) )
S − 1 ( ( ( β − 1 α − 1 S ) ( h ( 2 ) ) ⋅ m ) ( 1 ) ) ( α − 1 S − 1 ( h ( 1 ) ) )
= f ( ( β − 1 α − 1 S ) ( h ( 3 ) ) ⋅ m ( 0 ) ) ⊗ ( α β − 1 α − 1 ( h ( 5 ) ) )
S − 1 ( ( α − 1 S ) ( h ( 2 ) ) m ( 1 ) ( α β − 1 α − 1 ) ( h ( 4 ) ) ) ( α − 1 S − 1 ( h ( 1 ) ) )
= f ( ( β − 1 α − 1 S ) ( h ( 3 ) ) ⋅ m ( 0 ) ) ⊗ ( α β − 1 α − 1 ) ( h ( 5 ) S − 1 ( h ( 4 ) ) ) S − 1 ( m ( 1 ) ) α − 1 ( h ( 2 ) S − 1 ( h ( 1 ) ) )
= f ( ( β − 1 α − 1 S ) ( h ) ⋅ m ( 0 ) ) ⊗ S − 1 ( m ( 1 ) ) ,
which means that
( h ⋅ f ) ( 0 ) ⊗ ( h ⋅ f ) ( 1 ) = ( h ( 2 ) ⋅ f ( 0 ) ) ⊗ ( α β − 1 α − 1 ) ( h ( 3 ) ) f ( 1 ) ( α − 1 S − 1 ) ( h ( 1 ) ) , q . e . d .
On k   we have the trivial module and comodule structure, and with these k ∈   H Y D H   . We want to prove that b M   and d M   are H   -module maps. We compute:
( h ⋅ b M ( 1 ) ) ( m ) = ( h ⋅ ( ∑ i e i ⊗ e i ) ) ( m )
= ∑ i α − 1 ( h ( 1 ) ) ⋅ e i ⊗ ( ( α β α − 1 ) ( h ( 2 ) ) ⋅ e i ) ( m )
= α − 1 ( h ( 1 ) ) ⋅ e i ⊗ e i ( ( β − 1 α − 1 S α β α − 1 ) ( h ( 2 ) ) ⋅ m )
= ∑ i α − 1 ( h ( 1 ) ) ⋅ e i ⊗ e i ( ( α − 1 S ) ( h ( 2 ) ) ⋅ m )
= α − 1 ( h ( 1 ) S ( h ( 2 ) ) ) ⋅ m
= ɛ ( h ) ∑ i e i ⊗ e i ( m )
= ( ɛ ( h ) b M ( 1 ) ) ( m ) ,
d M ( h ⋅ ( f ⊗ m ) ) = d M ( α ( h ( 1 ) ) ⋅ f ⊗ β − 1 ( h ( 2 ) ) ⋅ m )
= ( α ( h ( 1 ) ) ⋅ f ) ( β − 1 ( h ( 2 ) ) ⋅ m )
= f ( ( β − 1 α − 1 S α ( h ( 1 ) ) ) β − 1 ( h ( 2 ) ) ⋅ m )
= f ( β − 1 ( S ( h ( 1 ) ) h ( 2 ) ) ⋅ m )
= ɛ ( h ) d M ( f ⊗ m ) .
Also they are H   -comodule maps:
( ( b M ( 1 ) ) ( 0 ) ⊗ ( b M ( 1 ) ) ( 1 ) ) ( m ) = ∑ i ( e i ) ( 0 ) ⊗ ( e i ) ( 0 ) ( m ) ⊗ ( e i ) ( 1 ) ( e i ) ( 1 )
= ∑ i ( e i ) ( 0 ) ⊗ ( e i ) ( m ( 0 ) ) ⊗ S − 1 ( m ( 1 ) ) ( e i ) ( 1 )
= m ( 0 ) ⊗ S − 1 ( m ( 1 ) 2 ) m ( 1 ) 1
= ( b M ( 1 ) ⊗ 1 ) ( m ) ,
d M ( ( f ⊗ m ) ( 0 ) ) ⊗ ( f ⊗ m ) ( 1 ) = f ( 0 ) ( m ( 0 ) ) ⊗ m ( 1 ) f ( 1 )
= f ( m ( 0 ) ) ⊗ m ( 1 ) 2 S − 1 ( m ( 1 ) 1 )
= d M ( f ⊗ m ) ⊗ 1 .
Finally, the last two identities ( i d M ⊗ d M ) ( b M ⊗ i d M ) = i d M   and ( d M ⊗ i d M * ) ( i d M * ⊗ b M ) = i d M *   are trivial. □  
Similarly, one can prove:
Proposition 3.7 Let M ∈   H Y D H ( α , β )   and assume that M   is finite dimensional. Then * M = H o m ( M , k )   becomes an object in H Y D H ( α − 1 , α β − 1 α − 1 )   , with ( h ⋅ f ) ( m ) = f ( ( α − 1 β − 1 S − 1 ( h ) ) ⋅ m )   and f ( 0 ) ( m ) ⊗ f ( 1 ) = f ( m ( 0 ) ) ⊗ S ( m ( 1 ) )   . Moreover, the maps b M : k →   * M ⊗ M   , b M ( 1 ) = ∑ i e i ⊗ e i   and d M : M ⊗   * M → k   , d M ( m ⊗ f ) = f ( m )   , are morphisms in H Y D H   and we have ( d M ⊗ i d M ) ( i d M ⊗ b M ) = i d M   and ( i d * M ⊗ d M ) ( b M ⊗ i d * M ) = i d * M   .
Consequently, if we consider Y D ( H ) f d   , the subcategory of Y D ( H )   consisting of finite dimensional objects, we obtain:
Theorem 3.8 Y D ( H ) f d   is a braided T-category with left and right dualities over G   , the left (respectively right) duals being given as in Proposition  3.6 (respectively Proposition  3.7 ).
Assume now that H   is finite dimensional. We will construct a quasitriangular T-coalgebra over G   , denoted by D T ( H )   , with the property that the T-category R e p ( D T ( H ) )   of representations of D T ( H )   is isomorphic to Y D ( H )   as braided T-categories. For ( α , β ) ∈ G   , the ( α , β )   -component D T ( H ) ( α , β )   will be the diagonal crossed product algebra H * ⋈ H ( α , β )   . Define
Δ ( α , β ) , ( γ , δ ) : H * ⋈ H ( ( α , β ) * ( γ , δ ) ) → ( H * ⋈ H ( α , β ) ) ⊗ ( H * ⋈ H ( γ , δ ) ) ,
Δ ( α , β ) , ( γ , δ ) ( p ⋈ h ) = ( p 2 ⋈ γ ( h 1 ) ) ⊗ ( p 1 ⋈ γ − 1 β γ ( h 2 ) ) .
One can check, by direct computation, that these maps are algebra maps, satisfying the necessary coassociativity conditions.
The counit ɛ   is just the counit of D T ( H ) ( i d , i d ) = D ( H )   , the Drinfeld double of H   .
For ( α , β ) , ( γ , δ ) ∈ G   , define now
φ ( α , β ) ( γ , δ ) : H * ⋈ H ( γ , δ ) → H * ⋈ H ( ( α , β ) * ( γ , δ ) * ( α , β ) − 1 ) ,
φ ( α , β ) ( γ , δ ) ( p ⋈ h ) = p ∘ β α − 1 ⋈ α γ − 1 β − 1 γ ( h ) .
Then one can check by direct computation that these are algebra isomorphisms giving a conjugation (that is they are multiplicative and compatible with the comultiplications and the counit).
The antipode is given, for ( α , β ) ∈ G   , by
S ( α , β ) : H * ⋈ H ( α , β ) → H * ⋈ H ( ( α , β ) − 1 ) ,
S ( α , β ) ( p ⋈ h ) = ( ɛ ⋈ α β ( S ( h ) ) ) ⋅ ( S * − 1 ( p ) ⋈ 1 ) ,
where the multiplication ⋅   in the right hand side is made in H * ⋈ H ( ( α , β ) − 1 )   .
Finally, the universal R   -matrix is given by
R ( α , β ) , ( γ , δ ) = ∑ i ( ɛ ⋈ β − 1 ( e i ) ) ⊗ ( e i ⋈ 1 ) ∈ ( H * ⋈ H ( α , β ) ) ⊗ ( H * ⋈ H ( γ , δ ) ) ,
for all ( α , β ) , ( γ , δ ) ∈ G   , where { e i }   , { e i }   are dual bases in H   and H *   .
Thus, we have obtained:
Theorem 3.9 D T ( H )   is a quasitriangular T-coalgebra over G   , with structure as above.
Moreover, the structure of D T ( H )   was constructed in such a way that, via the isomorphisms H * ⋈ H ( α , β ) ℳ ≃   H Y D H ( α , β )   from Section  2 , we obtain:
Theorem 3.10 R e p ( D T ( H ) )   and Y D ( H )   are isomorphic as braided T-categories over G   .
Remark 3.11 From Y D ( H )   (respectively D T ( H )   ) we can obtain, by pull-back along the group morphism A u t H o p f ( H ) → G   , α ↦ ( α , i d )   , a braided T-category (respectively a quasitriangular T-coalgebra) over A u t H o p f ( H )   .
If π   is a group together with a group morphism π → A u t H o p f ( H )   , by pull-back along it we obtain a braided T-category (respectively a quasitriangular T-coalgebra) over π   . Quasitriangular T-coalgebras over such π   have been studied by Virelizier in [17] .

4 An isomorphism of categories H Y D H ( α , β ) ≃   H Y D H   in the presence of a pair in involution

The aim of this section is to prove the following result.
Theorem 4.1 Let α , β   ∈ A u t H o p f ( H )   and assume that there exists ( f , g )   a pair in involution corresponding to ( α , β )   . Then the categories H Y D H ( α , β )   and H Y D H   are isomorphic.
A pair of inverse functors ( F , G )   is given as follows. If M ∈   H Y D H ( α , β )   , then F ( M ) ∈   H Y D H   , where F ( M ) = M   as vector space, with structures
h → m = f ( β − 1 ( S ( h 1 ) ) ) β − 1 ( h 2 ) ⋅ m ,
m ↦ m < 0 > ⊗ m < 1 > = m ( 0 ) ⊗ m ( 1 ) g − 1 .
If N ∈   H Y D H   , then G ( N ) ∈   H Y D H ( α , β )   , where G ( N ) = N   as vector space, with structures
h ⇀ n = f ( h 1 ) β ( h 2 ) ⋅ n ,
n ↦ n ( 0 ) ⊗ n ( 1 ) = n ( 0 ) ⊗ n ( 1 ) g .
Both F   and G   act as identities on morphisms.
Proof. One checks, by direct computation, that F   and G   are functors, inverse to each other.
Alternatively, we can give a very short proof using results from the previous section. By Example  2.7 , we have f k g ∈   H Y D H ( α , β )   . By Proposition  3.6 , we get ( f k g ) * ∈   H Y D H ( ( α , β ) − 1 )   . Then, one can check that actually F ( M ) = ( f k g ) * ⊗ M ∈ H Y D H   and G ( N ) = ( f k g ) ⊗ N ∈ H Y D H ( α , β )   . Also, one can see that ( f k g ) * ⊗   f k g =   f k g ⊗ ( f k g ) * = k   as objects in H Y D H   , hence F ∘ G = G ∘ F = i d   , using the associativity of the tensor product. □  
As we have noticed before, for any α ∈ A u t H o p f ( H )   we have that ( ɛ , 1 )   is a pair in involution corresponding to ( α , α )   , hence we obtain:
Corollary 4.2 H Y D H ( α , α ) ≃   H Y D H   .
Also, as a consequence of the theorem, we obtain the following result (a right-left version was given in [14] ), which might be useful for the aria of applicability of anti-Yetter-Drinfeld modules:
Corollary 4.3 Assume that there exists a pair in involution ( f , g )   corresponding to ( S 2 , i d )   .
Then the category H Y D H ( S 2 , i d )   of anti-Yetter-Drinfeld modules is isomorphic to H Y D H   , and any anti-Yetter-Drinfeld module can be written as a tensor product f k g ⊗ N   , with N ∈ H Y D H   .
Let again α , β   ∈ A u t H o p f ( H )   such that there exists ( f , g )   a pair in involution corresponding to ( α , β )   , and assume that H   is finite dimensional. Then we know that H Y D H ( α , β ) ≃   H * ⋈ H ( α , β ) ℳ   , H Y D H ≃   D ( H ) ℳ   , and the isomorphism H Y D H ( α , β ) ≃   H Y D H   constructed in the theorem is induced by an algebra isomorphism between H * ⋈ H ( α , β )   and D ( H )   , given by
D ( H ) → H * ⋈ H ( α , β ) , p ⊗ h ↦ g − 1 ⇀ p ⋈ f ( β − 1 ( S ( h 1 ) ) ) β − 1 ( h 2 ) ,
H * ⋈ H ( α , β ) → D ( H ) , p ⋈ h ↦ g ⇀ p ⊗ f ( h 1 ) β ( h 2 ) .
References

  1. D. Bulacu, F. Panaite, F. Van Oystaeyen, Generalized diagonal crossed products and smash products for (quasi) Hopf algebras. Applications, in preparation.
  2. S. Caenepeel, M. De Lombaerde, A categorical approach to Turaev's Hopf group-coalgebras, arXiv:math.QA/0409600.
  3. S. Caenepeel, G. Militaru, S. Zhu, ”Frobenius and separable functors for generalized module categories and nonlinear equations”, Lecture Notes in Mathematics 1787, Springer Verlag, Berlin, 2002.
  4. S. Caenepeel, F. Van Oystaeyen, Y. Zhang, The Brauer group of Yetter-Drinfeld module algebras, Trans. Amer. Math. Soc. 349 (1997), 3737–3771.
  5. A. Connes, H. Moscovici, Hopf algebras, cyclic cohomology and the transverse index theorem, Comm. Math. Phys. 198 (1998), 199–264.
  6. A. Connes, H. Moscovici, Cyclic cohomology and Hopf algebra symmetry, Lett. Math. Phys. 52 (2000), 1–28.
  7. P. M. Hajac, M. Khalkhali, B. Rangipour, Y. Sommerhauser, Stable anti-Yetter-Drinfeld modules, C. R. Math. Acad. Sci. Paris 338 (2004), 587–590.
  8. P. M. Hajac, M. Khalkhali, B. Rangipour, Y. Sommerhauser, Hopf-cyclic homology and cohomology with coefficients, C. R. Math. Acad. Sci. Paris 338 (2004), 667–672.
  9. F. Hausser and F. Nill, Diagonal crossed products by duals of quasi-quantum groups, Rev. Math. Phys. 11 (1999), 553–629.
  10. P. Jara, D. Ş tefan, Cyclic homology of Hopf Galois extensions and Hopf algebras, arXiv:math.KT/0307099.
  11. C. Kassel, ”Quantum groups”, Graduate Texts in Mathematics 155, Springer Verlag, Berlin, 1995.
  12. S. Majid, ”Foundations of quantum group theory”, Cambridge Univ. Press, 1995.
  13. S. Montgomery, ”Hopf algebras and their actions on rings”, CBMS Regional Conference Series, Vol. 82, Amer. Math. Soc., Providence, RI, 1993.
  14. M. D. Staic, A note on anti-Yetter-Drinfeld modules, preprint 2004.
  15. M. E. Sweedler, ”Hopf algebras”, Benjamin, New York, 1969.
  16. V. Turaev, Homotopy field theory in dimension 3 and crossed group-categories, arXiv:math.GT/0005291.
  17. A. Virelizier, Graded quantum groups, arXiv:math.QA/0312330.
  18. M. Zunino, Yetter-Drinfeld modules for crossed structures, J. Pure Appl. Algebra 193 (2004), 313–343.