Fundamental theorems of Doi-Hopf modules in a nonassociative setting
J.N. ALONSO ÁLVAREZ1, J.M. FERNÁNDEZ VILABOA2, R.
GONZÁLEZ RODRÍGUEZ3
1 Departamento de Matemáticas, Universidad
de Vigo, Campus Universitario Lagoas-Marcosende, E-36280 Vigo, Spain
(e-mail: [email protected])
2 Departamento de Álxebra, Universidad de
Santiago de Compostela. E-15771 Santiago de Compostela, Spain
(e-mail: [email protected])
3 Departamento de Matemática Aplicada II,
Universidad de Vigo, Campus Universitario Lagoas-Marcosende, E-36310
Vigo, Spain (e-mail: [email protected])
Abstract In this paper we introduce the notion of weak non-asssociative Doi-Hopf module and give the Fundamental Theorem of Hopf modules in this setting. Also we prove that there exists a categorical equivalence that admits as particular instances the ones constructed in the literature for Hopf algebras, weak Hopf algebras, Hopf quasigroups, and weak Hopf quasigroups.
Keywords. Hopf algebra, Weak Hopf
algebra, Hopf quasigroup, Weak Hopf
quasigroup, Doi-Hopf module, Fundamental Theorem.
MSC 2010: 18D10, 16T05, 17A30, 20N05.
1. introduction
Let F be a field and C=F−Vect. Let H be a Hopf algebra in C and let B be a right H-comodule algebra with coaction ρB:B→B⊗H, ρB(b)=b(0)⊗b(1), where the unadorned tensor product is the tensor product over F and for ρB(b) we used the Sweedler notation. In [15], Doi introduced the notion of (H,B)-Hopf module, as a generalization of the classical notion of Hopf module, defined by Larson and Sweedler in [24], in the following way: Let M be a right B-module and a right H-comodule. If, for all m∈M and b∈B, we write m.b for the action and ρM(m)=m[0]⊗m[1] for the coaction, we will say that M is an (H,B)-Hopf module if the equality
[TABLE]
holds, where m[1]b(2) is the product in H of m[1] and b(2). A morphism between two (H,B)-Hopf modules is an F-linear map that is B-linear and H-colinear. Hopf modules and morphisms of Hopf modules constitute the category of (H,B)-Hopf modules denoted by MBH. If there exists a right H-comodule map h:H→B which is an algebra map, and McoH={m∈M∣ρM(m)=m⊗1H}, BcoH={b∈B∣ρB(b)=b⊗1H} are the subobjects of coinvariants, for any m∈McoH and b∈BcoH we have that m.b∈McoH and then McoH is a right BcoH-module. Using this property Doi proved in Theorem 3 of [15] that M is isomorphic to McoH⊗BcoHB as (H,B)-Hopf modules. Moreover, if N is a right BcoH-module, the tensor product N⊗BcoHB, with the action and coaction induced by the product of B and the coproduct of H, is an (H,B)-Hopf module. This construction is functorial and then we have a functor, called the induction functor, F=−⊗BcoHB:CBcoH→MBH. Also, for all M∈MBH, the construction of McoH is functorial and we have a functor of coinvariants G=()coH:MBH→CBcoH such that F⊣G. Moreover, F and G induce a categorical equivalence between
MBH and the category of right BcoH-modules. This categorical equivalence was called by Doi and Takeuchi in [16], the strong structure theorem for MBH, and, for B=H and h=idH, contains as a particular instance the equivalence derived of the Fundamental Theorem of Hopf modules proved by Larson and Sweedler (see [24], [34]).
The categorical equivalence of the previous paragraph remains valid for weak Hopf algebras. For a weak Hopf algebra H, Böhm introduced in [10] the category of Hopf modules, denoted by MHH, in the same way that in the Hopf algebra setting. If M∈MHH, the subobject of coinvariants is defined by McoH={m∈M∣ρM(m)=m[0]⊗ΠHL(m[1])}, where ΠHL is the target morphism associated to H. In [10] we can find the weak version of the Fundamental Theorem of Hopf modules, i.e.: For all Hopf module M, McoH⊗HLH is isomorphic to M as Hopf modules, where HL is the image of ΠHL. Moreover, if CHL is the category of right HL-modules, there exist two functors F=−⊗HLH:CHL→MHH and G=()coH:MHH→CHL such that F is left adjoint of G and they induce a pair of inverse equivalences. Therefore, in the weak setting, MHH is equivalent to CHL. In this case, the following property is a relevant fact for subsequent generalizations: there is an isomorphism of Hopf modules between the tensor product McoH⊗HLH and McoH×H, where McoH×H is the image of a suitable idempotent morphism ∇M:McoH⊗H→McoH⊗H. Later, in [11], Böhm introduced the notion of weak Doi-Hopf module (or weak (H,B)-Doi-Hopf module), associated to a weak Hopf algebra H and a right H-comodule algebra B, and the category of weak Doi-Hopf modules denoted as in the non-weak setting by MBH.
In 2004, Zhang and Zhu proved that for any weak Doi-Hopf module M (also called by these authors weak (H,B)-Doi-Hopf module), if there exists a right H-comodule map h:H→B which is an algebra map, the objects M and McoH⊗BcoHB are isomorphic as (H,B)-Doi-Hopf modules. In this case BcoH={b∈B∣ρB(b)=b(0)⊗ΠHL(b(1))} and, if B=H
and h=idH, they recover the isomorphism constructed by Böhm in [10]. As in the Hopf setting, it is possible to construct the induction functor F=−⊗BcoHB:CBcoH→MBH and the functor of coinvariants G=()coH:MBH→CBcoH. These functors satisfy that F⊣G and F and G is a pair of inverse equivalences. Therefore, MBH is equivalent to the category of right BcoH-modules (see [18]).
In the two previous paragraphs we wrote about categorical equivalences for categories of Hopf modules connected to associative algebraic structures like Hopf algebras and weak Hopf algebras. An interesting generalization of Hopf algebras are nonassociative Hopf algebras. As in the quasi-Hopf setting, nonassociative Hopf algebras are not associative, but the lack of this property is compensated in this case by some axioms involving the division operation. The notion of nonassociative Hopf algebra in a category of vector spaces was introduced by Pérez- Izquierdo [29] with the aim of to construct the universal enveloping algebra for Sabinin algebras, prove a Poincaré-Birkhoff-Witt Theorem for Sabinin algebras and give a nonassociative version of the Milnor-Moore theorem. Later, Klim and Majid [23], in order
to understand the structure and relevant properties of the algebraic 7-sphere, introduced the notion of Hopf quasigroup.
Hopf quasigroups are examples of nonassociative Hopf algebras and in recent years, interesting research about its specific structure and its dual has been developed ([12], [19], [13], [14], [20], [21], [17], [2], [3]). Moreover, nonassociative Hopf algebras arise naturally related with other structures
in various nonassociative contexts like, for example quantum quasigroups in the sense of Smith ([30], [31], [32], [33]). Nonassociative Hopf algebras include the example of an enveloping algebra U(L) of a Malcev algebra (see [28], [23], [35]) as well as the notion of the loop algebra RL of a loop L (see [9], [26]). Then, nonassociative Hopf algebras unify Moufang loops and Malcev algebras, and, more generally, formal loops and Sabinin algebras, in the same way that Hopf algebras unify groups and Lie algebras.
For a of Hopf quasigroup in the sense of [23], Brzeziński defined in [12] the notion of Hopf module obtaining a categorical equivalence as in the associative context. In this case, the main difference appears in the definition of the category of Hopf modules MHH. Firstly, because the notion of Hopf module reflects the non-associativity of the product defined on H. Secondly, the morphisms are H-quasilinear and H-colinear (see Definition 3.4 of [12]). In Lemma 3.5 of [12], we can find that, if M∈MHH and McoH is defined like in the Hopf algebra setting, M is isomorphic to McoH⊗H as Hopf modules. Therefore the Fundamental Theorem of Hopf modules also holds for Hopf quasigroups. Moreover, there exist two functors F=−⊗H:C→MHH and G=()coH:MHH→C such that F⊣G, and they induce a pair of inverse equivalences. Thus, as it occurs in the Hopf algebra ambit, MHH is equivalent to the category of F-vector spaces.
Hopf quasigroups admit a generalization to the weak seetting. The new notion, called weak Hopf quasigroup, was introduced in [4] in a monoidal context and a family of non trivial examples can be obtained by working with bigroupoids, i.e., bicategories where every 1-cell is an equivalence and every 2-cell is an isomorphism (see Example 2.3 of [4]). In [6] we described these algebraic objects in terms of fusion morphisms and in [4], for a weak Hopf quasigroup H in a braided monoidal category C with tensor product ⊗, using the ideas proposed by Brzeziński for Hopf quasigroups, we introduce the notion of Hopf module and the category of Hopf modules MHH. In this case, if we define McoH in the same way that in the weak Hopf algebra setting, we obtain the weak nonassociative version of the Fundamental Theorem of Hopf modules in the following way: every Hopf module M is isomorphic to McoH×H as Hopf modules, where McoH×H is the image of the same idempotent ∇M used for Hopf modules associated to a weak Hopf algebra. Moreover, in [5] we proved that HL, the image of the target morphism, is a monoid, and then it is possible to take into consideration the category CHL to construct the tensor product McoH⊗HLH, and, if the functor −⊗H preserves coequalizers, to endow this object with a Hopf module structure. Unfortunately, unlike the case of weak Hopf algebras, it is not possible to assure in general that McoH⊗HLH is isomorphic to McoH×H. In order to find
sufficient conditions under which these objects are isomorphic in MHH, we introduce in [7] the category of strong Hopf modules, denoted by SMHH and obtain that there exist two functors F=−⊗HLH:CHL→SMHH and G=()coH:SMHH→CHL such that F is left adjoint of G and they induce a pair of inverse equivalences. In the Hopf quasigroup setting every Hopf module is strong, and then our results are the ones proved by Brzeziński in [12]. The same happens in the weak Hopf case and then we generalize the theorem proved by Böhm, Nill and Szlachányi in [10].
Let C be a braided monoidal category with tensor product ⊗. Then for a weak Hopf quasigroup H in C and a right H-comodule magma B (see [5] for the definition), a question naturally arises: Is it possible to define a general category of (H,B)-Hopf modules and to prove a general theorem that permit to recover as particular instances the categorical equivalences cited in the previous paragraphs? The main contribution of this paper is to give a positive answer to this question.
Now, we describe the paper in detail. After this introduction, for a weak Hopf quasigroup H and a right H-comodule magma B in a strict braided monoidal category C where every idempotent morphism splits, in the second section we introduce the notion of anchor morphism h:H→B as an H-comodule morphism such that it is a morphism of unital magmas satisfying two suitable conditions. For an anchor morphism h, in Definition 2.10, we define the notion of strong (H,B,h)-Hopf module and prove some properties of these modules. We also find the condition under which the subobject of coinvariants of B, defined as in the weak Hopf algebra context, i.e., BcoH={b∈B∣ρB(b)=b(0)⊗ΠHL(b(1))}, is a monoid, and construct the new category of strong (H,B,h)-Hopf modules, denoted by SMBH. Moreover, if the category C admits coequalizers and the functors −⊗B and −⊗H preserve coequalizers, we prove in Theorem 2.24 that the Fundamental Theorem of Hopf Modules holds. In other words, for any strong (H,B,h)-Hopf module the objects M and McoH⊗BcoHB are isomorphic in SMBH. This result admits as particular instances the results with the same name cited in the previous paragraphs for associative and nonassociative (weak) Hopf structures. Finally, in the last section, we define the induction functor F=−⊗BcoHB:CBcoH→SMBH and the functor of coinvariants G=()coH:SMBH→CBcoH, proving that F⊣G. Also, F and G is a pair of inverse equivalences and, therefore, SMBH is equivalent to the category of right BcoH-modules.
Throughout this paper C denotes a strict braided monoidal category with tensor product ⊗, unit object K and braid c. Without loss of generality, by the coherence theorems, we can assume the monoidal structure of C strict. Then, in this paper, we omit explicitly the associativity and unit constraints. For each object M in C, we denote the identity morphism by idM:M→M and, for simplicity of notation, given objects M, N and P in C and a morphism f:M→N, we write P⊗f for idP⊗f and f⊗P for f⊗idP. We also assume that every idempotent morphism in C splits, i.e., if ∇:Y→Y is such that ∇=∇∘∇, there exist an object Z, called the image of p, and morphisms i:Z→Y and p:Y→Z such that ∇=i∘p and p∘i=idZ. The morphisms p and i will be called a factorization of q. Note that Z, p and i are unique up to isomorphism. The categories satisfying this property constitute a broad class that includes, among others, the categories with epi-monic decomposition for morphisms and categories with equalizers or coequalizers. For example, complete bornological spaces is a symmetric monoidal closed category that is not abelian, but it does have coequalizers (see [27]). On the other hand, let Hilb be the category whose objects are complex Hilbert spaces and whose morphisms are the continuous linear maps. Then Hilb is not an abelian and closed category but it is a symmetric monoidal category (see [22]) with coequalizers.
As for prerequisites, the reader is expected to be familiar with the notions of (co)unital (co)magma, (co)monoid, and morphism of (co)unital (co)magmas. By a unital magma in C we understand a triple A=(A,ηA,μA) where A is an object in C and ηA:K→A (unit), μA:A⊗A→A (product) are morphisms in C such that μA∘(A⊗ηA)=idA=μA∘(ηA⊗A). If μA is associative, that is, μA∘(A⊗μA)=μA∘(μA⊗A), the unital magma will be called a monoid in C. Given two unital magmas
(monoids) A=(A,ηA,μA) and B=(B,ηB,μB), f:A→B is a morphism of unital magmas (monoids) if μB∘(f⊗f)=f∘μA and f∘ηA=ηB.
By duality, a counital comagma in C is a triple D=(D,εD,δD) where D is an object in C and εD:D→K (counit), δD:D→D⊗D (coproduct) are morphisms in C such that (εD⊗D)∘δD=idD=(D⊗εD)∘δD. If δD is coassociative, that is, (δD⊗D)∘δD=(D⊗δD)∘δD, the counital comagma will be called a comonoid. If D=(D,εD,δD) and E=(E,εE,δE) are counital comagmas
(comonoids), f:D→E is a morphism of counital comagmas (comonoids) if (f⊗f)∘δD=δE∘f and εE∘f=εD.
If A, B are unital magmas (monoids) in C, the object A⊗B is a unital magma (monoid) in C where ηA⊗B=ηA⊗ηB and μA⊗B=(μA⊗μB)∘(A⊗cB,A⊗B). In a dual way, if D, E are counital comagmas (comonoids) in C, D⊗E is a counital comagma (comonoid) in C where εD⊗E=εD⊗εE and δD⊗E=(D⊗cD,E⊗E)∘(δD⊗δE).
Let A be a monoid. The pair
(M,ϕM) is a right A-module if M is an object in
C and ϕM:M⊗A→M is a morphism
in C satisfying ϕM∘(M⊗ηA)=idM, ϕM∘(ϕM⊗A)=ϕM∘(M⊗μA). Given two right A-modules (M,ϕM)
and (N,ϕN), f:M→N is a morphism of right
A-modules if ϕN∘(f⊗A)=f∘ϕM. If D is a comonoid, the pair
(M,ρM) is a right D-comodule if M is an object in
C and ρM:M→M⊗D is a morphism
in C satisfying (M⊗εD)∘ρM=idM, (ρM⊗D)∘ρM=(M⊗δD)∘ρM. Given two right D-comodules (M,ρM)
and (N,ρN), f:M→N is a morphism of right
D-comodules if (f⊗D)∘ρM=ρN∘f.
Finally, if D is a comagma and A a magma, given two morphisms f,g:D→A we will denote by f∗g its convolution product in C, that is
[TABLE]
2. Doi-Hopf modules for weak Hopf quasigroups
We begin this section by recalling the notion of weak Hopf quasigroup in a braided monoidal category introduced in [4]. In this reference the interested reader can find an exhaustive list of properties of weak Hopf quasigroups, that we will need along the paper.
Definition 2.1**.**
A weak Hopf quasigroup H in C is a unital magma (H,ηH,μH) and a comonoid (H,εH,δH) such that the following axioms hold:
δH∘μH=(μH⊗μH)∘δH⊗H.
εH∘μH∘(μH⊗H)=εH∘μH∘(H⊗μH)
- =((εH∘μH)⊗(εH∘μH))∘(H⊗δH⊗H)
- =((εH∘μH)⊗(εH∘μH))∘(H⊗(cH,H−1∘δH)⊗H).
- (a3)
(δH⊗H)∘δH∘ηH=(H⊗μH⊗H)∘((δH∘ηH)⊗(δH∘ηH))
- =(H⊗(μH∘cH,H−1)⊗H)∘((δH∘ηH)⊗(δH∘ηH)).
- (a4)
There exists λH:H→H in C (called the antipode of H) such that, if we denote the morphisms idH∗λH by ΠHL (target morphism) and λH∗idH by ΠHR (source morphism),
ΠHL=((εH∘μH)⊗H)∘(H⊗cH,H)∘((δH∘ηH)⊗H).
ΠHR=(H⊗(εH∘μH))∘(cH,H⊗H)∘(H⊗(δH∘ηH)).
λH∗ΠHL=ΠHR∗λH=λH.
μH∘(λH⊗μH)∘(δH⊗H)=μH∘(ΠHR⊗H).
μH∘(H⊗μH)∘(H⊗λH⊗H)∘(δH⊗H)=μH∘(ΠHL⊗H).
μH∘(μH⊗λH)∘(H⊗δH)=μH∘(H⊗ΠHL).
μH∘(μH⊗H)∘(H⊗λH⊗H)∘(H⊗δH)=μH∘(H⊗ΠHR).
Note that, if in the previous definition the triple (H,ηH,μH) is a monoid, we obtain the notion of weak Hopf algebra in a symmetric monoidal category. Then, if C is the category of vector spaces over a field F, we have the original definition of weak Hopf algebra introduced by Böhm, Nill and Szlachányi in [10]. On the other hand, under these conditions, if εH and δH are morphisms of unital magmas (equivalently, ηH, μH are morphisms of counital comagmas), ΠHL=ΠHR=ηH⊗εH. As a consequence, conditions (a2), (a3), (a4-1)-(a4-3) trivialize, and we get the notion of Hopf quasigroup defined by Klim and Majid in [23]. More concretely, a Hopf quasigroup H in
C is a unital magma (H,ηH,μH) and a comonoid (H,εH,δH) satisfying that εH and δH are morphisms of unital magmas (equivalently, ηH and μH are morphisms of counital comagmas), and such that there exists a morphism λH:H→H
in C, called the antipode of H, for which
[TABLE]
and
[TABLE]
hold. Then, as a consequence, we have (a1) and the following identities
[TABLE]
By Proposition 3.2 of [4] we know that the antipode of a weak Hopf quasigroup is unique, and satisfies that λH∘ηH=ηH, εH∘λH=εH. Also, by Theorem 3.19 of [4], we have that it is antimultiplicative and anticomultiplicative. Moreover, if we define the morphisms ΠHL and ΠHR by
[TABLE]
we proved in Proposition 3.4 of [4], that ΠHL, ΠHR, ΠHL and
ΠHR are idempotent.
Lemma 2.2**.**
Let H be a weak Hopf quasigroup and Π∈{ΠHL,ΠHR,ΠHL,ΠHR} . The following identities hold:
[TABLE]
[TABLE]
[TABLE]
Proof.
The proof for ΠHL is in Proposition 2.4 of [5] and in a similar way we can prove the result for ΠHR (see also Proposition 2.3 of [6]). The equalities for ΠHL and ΠHR follow from Proposition 3.11 of [4].
∎
If HL is the image of the idempotent morphism ΠHL, and
pL:H→HL, iL:HL→H are the
morphisms such that ΠHL=iL∘pL and pL∘iL=idHL, by Proposition 3.13 of [4], iL is the equalizer of δH and (H⊗ΠHL)∘δH and pL is the coequalizer of μH and μH∘(H⊗ΠHL). Then the triple (HL,εHL=εH∘iL,δH=(pL⊗pL)∘δH∘iL) is a comonoid in C, and as a consequence of Lemma 2.2, (HL,ηHL=pL∘ηH,μHL=pL∘μH∘(iL⊗iL)) is a monoid in C. Following Remark 3.15 of [4], we have similar results for the image of the idempotent morphism ΠHR denoted by HR.
Definition 2.3**.**
Let H be a weak Hopf quasigroup
and let B be a unital magma, which is also a right
H-comodule with coaction ρB:B→B⊗H such that
[TABLE]
We will say that (B,ρB) is a right H-comodule
magma if any of the following equivalent conditions hold:
(ρB⊗H)∘ρB∘ηB=(B⊗(μH∘cH,H−1)⊗H)∘((ρB∘ηB)⊗(δH∘ηH)).
(ρB⊗H)∘ρB∘ηB=(B⊗μH⊗H)∘((ρB∘ηB)⊗(δH∘ηH)).
(B⊗ΠHR)∘ρB=(μB⊗H)∘(B⊗(ρB∘ηB)),
(B⊗ΠHL)∘ρB=((μB∘cB,B−1)⊗H)∘(B⊗(ρB∘ηB)).
(B⊗ΠHR)∘ρB∘ηB=ρB∘ηB.
(B⊗ΠHL)∘ρB∘ηB=ρB∘ηB.
This definition is similar to the notion of right H-comodule monoid in the weak Hopf algebra setting and the proof for the equivalence of (b1)-(b6) also follows in a similar way.
Note that, if H is a Hopf quasigroup and B is a unital magma which is also a right
H-comodule with coaction ρB:B→B⊗H, we will say that (B,ρB) is a right H-comodule magma if it satisfies (7) and ρB∘ηB=ηH⊗ηB. In this case (b1)-(b6) trivialize.
Example 2.4**.**
-
If H is a (weak) Hopf quasigroup, (H,δH) is a right H-comodule magma.
-
Let H be a cocommutative weak Hopf quasigrop and assume that C is symmetric. Then, cH,H∘δH=δH and, by Theorem 3.22 of [4], λH2=idH. If we denote by Hop the unital magma Hop=(H,ηHop=ηH,μHop=μH∘cH,H), we have that (Hop,ρHop=(H⊗λH)∘δH) is an example of right H-comodule magma. Indeed, first note that Hop is a unital magma. Also (H⊗εH)∘ρHop=idH because λH presevers the counit. On the other hand, if H is cocommutative, the equality
[TABLE]
holds. Then, by the coassociativity of δH and (8), we obtain that (ρHop⊗H)∘ρHop=(H⊗δH)∘ρHop, an we have that (Hop,ρHop) is a right H-comodule. Finally,
- ρHop∘μHop
- =(μH⊗(λH∘μH))∘(cH,H⊗cH,H)∘δH⊗H (naturality of c and c2=idH)
- =(μHop⊗(μH∘(λH⊗λH)))∘δH⊗H ((53) of [4] and c2=idH)
- =(μHop⊗μH)∘(H⊗cH,H⊗H)∘(ρHop⊗ρHop) (naturality of c)
and
- ρHop∘ηHop
- =(H⊗(λH∘ΠHL))∘δH∘ηH ((21) of [4])
- =(H⊗(λH∘ΠHL))∘δH∘ηH (if H is cocommutative ΠHL=ΠHL)
- =(H⊗ΠHL)∘δH∘ηH ((39) of [4])
- =(H⊗(ΠHL∘ΠHL))∘δH∘ηH (ΠHL is idempotent)
- =(H⊗ΠHL)∘ρHop∘ηH ((21) of [4]).
Therefore, (Hop,ρHop) is a right H-comodule magma. Note that, if H is a Hopf quasigroup we have the same example.
- By Example 3.1 of [3] we have the following. Let H be a Hopf quasigroup and A a unital magma in C. If there exists a morphism φA:H⊗A→A such that
[TABLE]
[TABLE]
hold, then the smash product A♯H=(A⊗H,ηA♯H,μA♯H) defined by
[TABLE]
where
[TABLE]
is a right H-comodule magma with comodule structure given by
[TABLE]
- Let H, B two Hopf quasigroups. Assume that there exists a morphism of Hopf quasigroups g:B→H, i.e., a morphism of unital magmas and comonoids. Then, (B,ρB=(B⊗g)∘δB) is an example of right H-comodule magma.
Definition 2.5**.**
Let H be a weak Hopf quasigroup and let (B,ρB) be a right H-comodule
magma. We will say that h:H→B is an integral if it is a morphism of right H-comodules. The integral will be called total if h∘ηH=ηB.
Proposition 2.6**.**
Let H be a weak Hopf quasigroup and let (B,ρB) be a right H-comodule
magma. Let h:H→B be a total integral. The endomorphism qB:=μB∘(B⊗(h∘λH))∘ρB:B→B satisfies
[TABLE]
[TABLE]
and, as a consequence, qB is an idempotent morphism. Moreover, if BcoH (object of coinvariants) is the image of qB and pB:B→BcoH, iB:BcoH→B are the morphisms such that qB=iB∘pB and
idBcoH=pB∘iB,
[TABLE]
[TABLE]
are equalizer diagrams.
Proof.
First note that, by the naturality of c, the condition of right H-comodule morphism for h, and the equality h∘ηH=ηB, we obtain that
[TABLE]
holds. Then, as a consequence, by (7) and the properties of εH and ηB, we have
[TABLE]
Also,
- ρB∘qB
- =μB⊗H∘(ρB⊗(ρB∘h∘λH))∘ρB
((7))
- =(μB⊗H)∘(B⊗((h⊗μH)∘(cH,H⊗H)∘(H⊗(δH∘λH))∘δH))∘ρB (comodule condition for B and
- comodule morphism condition for h)
- =(μB⊗H)∘(B⊗((h⊗μH)∘(cH,H⊗H)∘(H⊗cH,H)∘(((H⊗λH)∘δH)⊗λH)∘δH))∘ρB
- (anticomultiplicativity of λH and coassociativity of δH)
- =((μB∘(B⊗h))⊗ΠHL)∘(B⊗(cH,H∘(H⊗λH)∘δH))∘ρB (naturality of c),
and then using that ΠHL is an idempotent morphism we obtain (11). Therefore, by (34) of [4], we have (12) because
[TABLE]
Then, qB is an idempotent morphism. Indeed,
- qB∘qB
- =μH∘(B⊗(h∘λH∘ΠHR))∘ρB∘qB((12))
- =μB∘(B⊗(h∘ΠHR))∘ρB∘qB ((40) of [4])
- =qB ((14)).
On the other hand, by (40) of [4] and (14),
[TABLE]
Then,
[TABLE]
is a split cofork (see [25]) and thus an equalizer diagram. As a consequence, by (33) of [4], we have
[TABLE]
and qB equalizes ρB and (B⊗ΠHL)∘ρB. Therefore, iB equalizes ρB and (B⊗ΠHL)∘ρB because pB∘iB=idBcoH. Moreover, if t:D→B is such that (B⊗ΠHL)∘ρB∘t=ρB∘t, composing with B⊗ΠHR, we obtain that (B⊗ΠHR)∘ρB∘t=ρB∘t. Thus, there exists a unique morphism t′:D→BcoH such that iB∘t′=t. Then,
[TABLE]
is an equalizer diagram.
∎
Remark 2.7*.*
Note that, under the conditions of the previous proposition, the object of coinvariants is independent of the total integral h because it is the equalizer object of ρB and (B⊗ΠHR)∘ρB (or equivalently, ρB and (B⊗ΠHL)∘ρB).
Moreover, (see [5]) the triple (BcoH,ηBcoH,μBcoH) is a unital magma (the submagma of coinvariants of B), where ηBcoH:K→BcoH, μBcoH:BcoH⊗BcoH→BcoH are the factorizations through iB of the morphisms ηB and μB∘(iB⊗iB), respectively. Therefore, ηBcoH is the unique morphism such that
[TABLE]
and μBcoH is the unique morphism satisfying
[TABLE]
Thus,
[TABLE]
and
[TABLE]
In what follows, the object of coinvariants BcoH will be called the submagma of coinvariants of B.
Note that, if B=H and ρB=δH, the submagma of coinvariants is HcoH=HL and then, in this case, it is a monoid.
Definition 2.8**.**
Let H be a weak Hopf quasigroup and let (B,ρB) be a right H-comodule
magma. We will say that h:H→B is an anchor morphism if it is a multiplicative total integral (i.e., a right H-comodule morphism such that it is a morphism of unital magmas) and the following equalities hold:
μB∘((μB∘(B⊗h))⊗(h∘λH))∘(B⊗δH)=μB∘(B⊗(h∘ΠHL)).
μB∘((μB∘(B⊗(h∘λH)))⊗h)∘(B⊗δH)=μB∘(B⊗(h∘ΠHR)).
Note that, if the product on B is associative, every multiplicative total integral h satisfies (c1)-(c2) and therefore is an anchor morphism. Also, using that h is a comodule morphism, the condition (c1) can be rewritten as
[TABLE]
Finally, if H is a Hopf quasigroup, (c1) and (c2) are
[TABLE]
or, in an equivalent way, B is a cleft right H-comodule algebra in the sense of Definition 3.1 of [3].
Example 2.9**.**
-
By the definition of weak Hopf quasigroup and (40) of [4], the identity morphism idH is an anchor morphism for the right H-comodule magma (H,δH). Also, if H is a Hopf quasigrup, the identity of H is an anchor morphism.
-
By the second point of Example 2.4 we know that, if H is a cocommutative weak Hopf quasigroup and C is symmetric, (Hop,ρHop=(H⊗λH)∘δH) is an example of right H-comodule magma. Note that in this case we have that λH∘ηH=ηH and
- ρHop∘λH
- =(λH⊗λH2)∘δH((8))
- =(λH⊗H)∘δH (Theorem 3.22 of [4], i.e., λH2=idH)
Therefore λH is a total integral. Moreover, by (52) of [4], we obtain that
μHop∘(λH⊗λH)=λH∘μH. Finally,
- μHop∘((μHop∘(H⊗λH))⊗(λH∘λH))∘(H⊗δH)
- =μH∘(H⊗μH)∘(cH,H⊗H)∘(H⊗cH,H)∘(cH,H⊗H)∘(H⊗cH,H)∘(H⊗((H⊗λH)∘δH))
- (Theorem 3.22 of [4], naturality of c and coassociativity of δH)
- =μH∘(H⊗μH)∘(H⊗λH⊗H)∘(δH⊗H)∘cH,H (naturality of c and c2=idH)
- =μH∘(ΠHL⊗H)∘cH,H ((a4-5) of Definition 2.1)
- =μHop∘(H⊗ΠHL)(naturality of c)
- =μHop∘(H⊗(λH∘ΠHL)) ((39) of [4])
- =μHop∘(H⊗(λH∘ΠHL)) (H is cocommutative, ΠHL=ΠHL),
and, by similar arguments,
[TABLE]
Therefore, λH is an anchor morphism for (Hop,ρHop). Of course, the same result holds for cocommutative Hopf quasigroups.
- In the third point of Example 2.4 we saw that if H is a Hopf quasigroup, A is a unital magma in C, and there exists a morphism φA:H⊗A→A satisfying (9), (10), the smash product A♯H is a right H-comodule magma with coaction ρA♯H=A⊗δH. By Example 3.1 of [3], we have that h=ηA⊗H:H→A♯H is a total integral. On the other hand,
- μA♯H∘(h⊗h)
- =((φA∘(H⊗ηA))⊗μH)∘(δH⊗H)(unit properties and naturality of c)
- =h∘μH ((10) and counit properties)
and then h is multiplicative. Also, by similar arguments, we have
[TABLE]
Thus, by (20) and (2), the identities
[TABLE]
hold. Similarly, by (20) and (1) we obtain that
[TABLE]
Therefore, h=ηA⊗H is an anchor morphism.
- Assume that H and B are Hopf quasigrous in C. Let g:B→H, f:H→B be morphisms of Hopf quasigroups such that g∘f=idH. Consider the right H-comodule structure on B defined in fourth point of Example 2.4. Then, f is an anchor morphism. Indeed, first note that f∘ηH=ηB, f∘μH=μB∘(f⊗f) hold because f is a morphism of unital magmas. Also, f is a comodule morphism, i.e.,
ρB∘f=(f⊗H)∘δH, because f is a comonoid morphism and g∘f=idH. On the other hand, by Lemma 1.4 of [2], we have that
[TABLE]
holds, and then, using that f is a comonoid morphism, we get
[TABLE]
[TABLE]
Similarly, by the same arguments,
[TABLE]
and this implies that f is an anchor morphism.
As a consequence, the examples of strong projections that we can find in [2] and [8] provide examples of anchor morphisms.
If (B,ρB) is a right H-comodule monoid the following identity
[TABLE]
holds. Indeed:
- (B⊗ΠHR)∘ρB∘μB
- =(μB⊗(εH∘μH∘(μH⊗H))⊗H)∘(B⊗cH,B⊗H⊗(δH∘ηH))∘(ρB⊗ρB)((7) and definition of ΠHR)
- =(μB⊗(((εH∘μH)⊗(εH∘μH))∘(H⊗δH⊗H))⊗H)∘(B⊗cH,B⊗H⊗(δH∘ηH))∘(ρB⊗ρB)
- ((a2) of Definition 2.1)
- =(B⊗εH⊗εH⊗H)∘(μB⊗H⊗μH⊗H)∘(ρB⊗ρB⊗H⊗(δH∘ηH))∘(B⊗ρB) (comodule condition
- for B)
- =(((B⊗εH)∘ρB∘μB)⊗ΠHR))∘(B⊗ρB) ((7) and definition of ΠHR)
- =(μB⊗ΠHR)∘(B⊗ρB) (comodule condition for B).
As a consequence, if h:H→B is a total integral, the equality
[TABLE]
holds because
- μB∘(μB⊗(h∘ΠHR))∘(B⊗ρB)
- =μB∘(μB⊗(h∘λH∘ΠHR))∘(B⊗ρB) ((40) of [4])
- =μB∘(μB⊗(h∘λH∘ΠHR))∘ρB∘μB ((22))
- =μB∘(μB⊗(h∘ΠHR))∘ρB∘μB ((40) of [4])
- =μB ((14)).
Also, if h:H→B is a multiplicative morphism of right H-comodules, the following identity holds
[TABLE]
because
- h∘ΠHL
- =μB∘(h⊗(h∘λH))∘δH (h is multiplicative)
- =qB∘h (h is a comodule morphism).
If h:H→B is an anchor morphism we have that the equality
[TABLE]
holds. Indeed:
- μB∘((μB∘(B⊗(h∘ΠHL)))⊗h)∘(B⊗δH)
- =μB∘((μB∘(μB⊗(h∘λH))∘(B⊗ρB))⊗h)∘(B⊗(ρB∘h)) ((19) and comodule morphism condition
- (19) for h)
- =μB∘((μB∘(B⊗(h∘λH)))⊗h)∘(B⊗δH))∘(μB⊗H)∘(B⊗(ρB∘h)) (comodule condition for B)
- =μB∘((μB∘(h∘ΠHR))∘(B⊗(ρB∘h)) ((c2) of Definition 2.8)
- =μB∘(B⊗h) ((23)).
Then, by (24) and the comodule morphism condition for h, (25) is equivalent to
[TABLE]
Finally, for an anchor morphism h:H→B the equality
[TABLE]
holds, because
- μB∘(qB⊗h)∘ρB
- =μB∘((μB∘(B⊗(h∘λH)))⊗h)∘(B⊗δH)∘ρB (comodule condition for B)
- =μB∘(B⊗(h∘ΠHR))∘ρB ((c2) of Definition 2.8)
- =idB ((14)).
Definition 2.10**.**
Let H be a weak Hopf quasigroup and let (B,ρB) be a right H-comodule
magma. Let h:H→B be an anchor morphism and let M be an object in C. We say that (M,ϕM,ρM) is a strong (H,B,h)-Hopf module if the following axioms hold:
The pair (M,ρM) is a right H-comodule.
The morphism ϕM:M⊗B→M satisfies:
ϕM∘(M⊗ηB)=idM.
ϕM∘((ϕM∘(M⊗iB))⊗B)=ϕM∘(M⊗(μB∘(iB⊗B))).
ρM∘ϕM=(ϕM⊗μH)∘(M⊗cH,B⊗H)∘(ρM⊗ρB).
ϕM∘((ϕM∘(M⊗h))⊗(h∘λH))∘(M⊗δH)=ϕM∘(M⊗(h∘ΠHL)).
ϕM∘((ϕM∘(M⊗(h∘λH)))⊗h)∘(M⊗δH)=ϕM∘(M⊗(h∘ΠHR)).
For example, the triple (H,μH,δH) is a strong (H,H,idH)-Hopf module. Also, if the following equality
[TABLE]
holds, the triple (B,μB,ρB) is a strong (H,B,h)-Hopf module.
Let (M,ϕM,ρM) be a strong (H,B,h)-Hopf module. In a similar way to (22), but using (d2-3) of Definition 2.10 instead of (7), it is easy to see that
[TABLE]
holds. Moreover, by (13), (d2-3), (d2-1), and the condition of comodule for M, we also have the equality
[TABLE]
As a consequence, we can obtain the Hopf module version of (23), i.e.,
[TABLE]
Moreover, in this setting, the equality
[TABLE]
holds because
- ϕM∘((ϕM∘(M⊗(h∘ΠHL)))⊗h)∘(M⊗δH)
- =ϕM∘((ϕM∘(M⊗(qB∘h)))⊗h)∘(M⊗δH) ((24))
- =ϕM∘(M⊗(μB∘((qB∘h)⊗h)∘δH)) ((d2-2) of Definition 2.10)
- =ϕM∘(M⊗(μB∘((h∘ΠHL)⊗h)∘δH)) ((24))
- =ϕM∘(M⊗(h∘(ΠHL∗idH))) (h is multiplicative)
- =ϕM∘(M⊗h) ((4) of [4]).
Finally, by (24) the equality (32) is equivalent to
[TABLE]
Proposition 2.11**.**
Let H be a weak Hopf quasigroup and let (B,ρB) be a right H-comodule
magma. Let h:H→B be an anchor morphism and let (M,ϕM,ρM) be a strong (H,B,h)-Hopf module. The endomorphism qM:=ϕM∘(M⊗(h∘λH))∘ρM:M→M satisfies
[TABLE]
[TABLE]
and, as a consequence, qM is an idempotent. Moreover, if McoH (object of coinvariants) is the image of qM and pM:M→McoH, iM:McoH→M are the morphisms such that qM=iM∘pM and
idMcoH=pM∘iM,
[TABLE]
[TABLE]
are equalizer diagrams.
Proof.
The proof is similar to the one developed in Proposition 2.6 but using (d2-3) instead of (7) and the comodule condition for M instead of the comodule condition for B.
∎
Remark 2.12*.*
Note that, as in the case of B, the object of coinvariants McoH is independent of the anchor morphism h because it is the equalizer object of ρM and (M⊗ΠHR)∘ρM (or ρM and (M⊗ΠHL)∘ρM).
Lemma 2.13**.**
Let H be a weak Hopf quasigroup and let (B,ρB) be a right H-comodule
magma. Let h:H→B be an anchor morphism and let (M,ϕM,ρM) be a strong (H,B,h)-Hopf module. The following equalities hold:
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
Proof.
The proof for the first equality is the following:
- ρM∘ϕM∘(iM⊗B)
- =(ϕM⊗μH)∘(M⊗cH,B⊗H)∘((ρM∘iM)⊗ρB) ((d2-3) of Definition (2.10))
- =(ϕM⊗μH)∘(M⊗cH,B⊗H)∘(((M⊗ΠHR)∘ρM∘iM)⊗ρB) ((35))
- =(ϕM⊗(μH∘(ΠHR⊗H)))∘(M⊗cH,B⊗H)∘((ρM∘iM)⊗ρB) (naturality of c)
- =(ϕM⊗(εH∘μH)⊗H)∘(M⊗cH,B⊗H⊗H)∘((ρM∘iM)⊗((B⊗δH)∘ρB)) ((10) of [4])
- =(((ϕM⊗(εH∘μH))∘(M⊗cH,B⊗H)∘(ρM⊗ρB))⊗H)∘(iM⊗ρB) (comodule condition for B)
- =(((M⊗εH)∘ρM∘ϕM)⊗H)∘(iM⊗ρB) ((d2-3) of Definition (2.10))
- =(ϕM⊗H)∘(iM⊗ρB) (comodule condition for B).
The equality (37) follows by
- qM∘ϕM∘(iM⊗B)
- =ϕM∘(ϕM⊗(h∘λH))∘(iM⊗ρB) ((36))
- =ϕM∘(iM⊗qB) ((d2-4) of Definition (2.10)).
Thus, composing in (36) with qM⊗H, we obtain (38). Also, composing in (38) with pM⊗B, we have (39). Moreover, composing with McoH⊗iB in (37), we obtain (40), and doing the same with
pM we get (41). Finally, (42) holds because
- ϕM∘(qM⊗h)∘ρM
- =ϕM∘((ϕM∘(M⊗(h∘λH)))⊗h)∘(M⊗δH)∘ρM (comodule condition for M)
- =ϕM∘(M⊗(h∘ΠHR))∘ρM ((d2-5) of Definition (2.10))
- =idM ((30)).
∎
Proposition 2.14**.**
Let H be a weak Hopf quasigroup and let (B,ρB) be a right H-comodule
magma. Let h:H→B be an anchor morphism. If (28) holds, the submagma of coinvariants
(BcoH,ηBcoH,μBcoH) is a monoid.
Proof.
Firstly, remember that if (28) holds, the triple (B,μB,ρB) is a strong (H,B,h)-Hopf module. Then by (37),
[TABLE]
As a consequence, BcoH is a monoid because,
- iB∘μBcoH∘(BcoH⊗μBcoH)
- =qB∘μB∘(iB⊗(qB∘μB∘(iB⊗iB))) ((18))
- =qB∘μB∘(iB⊗(μB∘(iB⊗iB))) ((43))
- =qB∘μB∘((μB∘(iB⊗iB))⊗iB) ((28))
- =qB∘μB∘((qB∘μB∘(iB⊗iB))⊗iB) ((43))
- =iB∘μBcoH∘(μBcoH⊗BcoH) ((18)).
∎
Proposition 2.15**.**
Let H be a weak Hopf quasigroup and let (B,ρB) be a right H-comodule
magma. Let h:H→B be an anchor morphism. If (28) holds, for all strong (H,B,h)-Hopf module (M,ϕM,ρM), the object of coinvariants McoH is a right BcoH-module.
Proof.
First note that, by Proposition 2.14, the object BcoH is a monoid. Now we will show that there exists an action ϕMcoH:McoH⊗BcoH→McoH such that ϕMcoH∘(McoH⊗ηBcoH)=idMcoH, and ϕMcoH∘(ϕMcoH⊗BcoH)=ϕMcoH∘(McoH⊗μBcoH). To define the action, we begin by proving that
[TABLE]
Indeed,
- ρM∘ϕM∘(iM⊗iB)
- =(ϕM⊗H)∘(iM⊗(ρB∘iB)) ((36))
- =(ϕM⊗ΠHL)∘(iM⊗(ρB∘iB)) ((11))
- =(M⊗ΠHL)∘ρM∘ϕM∘(iM⊗iB) ((36))
Then, there exists a unique morphism ϕMcoH:McoH⊗BcoH→McoH such that
[TABLE]
Therefore,
[TABLE]
The pair (McoH,ϕMcoH) satisfies the conditions of right BcoH-module because
[TABLE]
and
- ϕMcoH∘(ϕMcoH⊗BcoH)
- =pM∘ϕM∘((qM∘ϕM∘(iM⊗iB))⊗iB) ((46))
- =pM∘ϕM∘((ϕM∘(iM⊗iB))⊗iB) ((40))
- =pM∘ϕM∘(iM⊗(μB∘(iB⊗iB))) ((d2-2) of Definition (2.10))
- =pM∘ϕM∘(iM⊗(iB∘μBcoH)) ((16))
- =ϕMcoH∘(McoH⊗μBcoH) ((46)).
∎
Proposition 2.16**.**
Let H be a weak Hopf quasigroup and let (B,ρB) be a right H-comodule
magma. Let h:H→B be an anchor morphism. Assume that (28) and
[TABLE]
hold. Then if the category C admits coequalizers and the functors −⊗B and −⊗H preserve coequalizers, for all strong (H,B,h)-Hopf module (M,ϕM,ρM), the object McoH⊗BcoHB, defined by the coequalizer of TM1=ϕMcoH⊗B and TM2=McoH⊗(μB∘(iB⊗B)), is a strong (H,B,h)-Hopf module. Moreover,
McoH⊗BcoHB and M are isomorphic as right H-comodules.
Proof.
In the first step, we begin by proving the existence of an action and a coaction for McoH⊗BcoHH. If the object McoH⊗BcoHH is defined by the coequalizer diagram
[TABLE]
we have that
[TABLE]
is also a coequalizer diagram because the functor −⊗B preserves coequalizers. Consider the morphism nMcoH∘(McoH⊗μB):McoH⊗B⊗B→McoH⊗BcoHB. Then,
[TABLE]
and, as a consequence, there exists a unique morphism
[TABLE]
such that
[TABLE]
On the other hand, consider the morphism
[TABLE]
Then, taking into account that, by (28), (B,μB,ρB) is a strong (H,B,h)-Hopf module,
- (nMcoH⊗H)∘(ϕMcoH⊗ρB)
- =(nMcoH⊗H)∘(McoH⊗((μB⊗H)∘(iB⊗ρB))) ((48))
- =(nMcoH⊗H)∘(McoH⊗(ρB∘μB∘(iB⊗B))) ((36)).
Thus, there exists a unique morphism
[TABLE]
such that
[TABLE]
We proceed to show that (McoH⊗BcoHB,ϕMcoH⊗BcoHB,ρMcoH⊗BcoHB) is a strong (H,B,h)-Hopf module. Indeed, first note that by (51) and the comodule condition for B,
[TABLE]
As a consequence, and using that nMcoH is a coequalizer, we obtain that
[TABLE]
By similar arguments we have
[TABLE]
[TABLE]
Therefore, the pair (McoH⊗BcoHB,ρMcoH⊗BcoHB) is a right H-comodule and we have (d1) of Definition 2.10.
On the other hand, by (50),
[TABLE]
and then, (d2-1) of Definition 2.10, i.e., ϕMcoH⊗BcoHB∘(McoH⊗BcoHB⊗ηB)=idMcoH⊗BcoHB, holds. Moreover,
- ϕMcoH⊗BcoHB∘((ϕMcoH⊗BcoHB∘(nMcoH⊗qB))⊗B)
- =nMcoH∘(McoH⊗(μB∘((μB∘(B⊗qB))⊗B))) ((50))
- =nMcoH∘(McoH⊗(μB∘(B⊗(μB∘(qB⊗B))))) ((28))
- =ϕMcoH⊗BcoHB∘(nMcoH⊗(μB∘(qB⊗B))) ((50))
and, as a consequence,
[TABLE]
[TABLE]
because nMcoH⊗B⊗B is a coequalizer. Therefore (d2-2) of Definition 2.10 follows because pB is a projection.
The proof for (d2-3) of Definition 2.10 is the following: Composing with the coequalizer nMcoH⊗B we obtain that
- ρMcoH⊗BcoHB∘ϕMcoH⊗BcoHB∘(nMcoH⊗B)
- =(nMcoH⊗H)∘(McoH⊗(ρB∘μB)) ((50), (51))
- =(nMcoH⊗H)∘(McoH⊗(μB⊗H∘(ρB⊗ρB))) ((7))
- =(ϕMcoH⊗BcoHB⊗μH)∘(McoH⊗BcoHB⊗cH,B⊗H)∘((ρMcoH⊗BcoHB∘nMcoH)⊗ρB) ((50), (51))
and thus
[TABLE]
Using that −⊗H preserves coequalizers, the equality (d2-4) of Definition 2.10 follows from
- ϕMcoH⊗BcoHB∘(ϕMcoH⊗BcoHB⊗B)∘(nMcoH⊗((h⊗(h∘λH))∘δH))
- =nMcoH∘(McoH⊗(μB∘(μB⊗B)∘(B⊗((h⊗(h∘λH))∘δH)))) ((50))
- =nMcoH∘(McoH⊗(μB∘(B⊗(h∘ΠHL)))) ((c1) of Definition 2.8)
- =ϕMcoH⊗BcoHB∘(nMcoH⊗(h∘ΠHL)) ((50)).
Also, we have
- ϕMcoH⊗BcoHB∘((ϕMcoH⊗BcoHB∘(McoH⊗BcoHB⊗(h∘λH)))⊗h)∘(nMcoH⊗δH)
- =nMcoH∘(McoH⊗(μB∘((μB∘(B⊗(h∘λH)))⊗h)∘(B⊗δH))) ((50))
- =nMcoH∘(McoH⊗(μB∘(B⊗(h∘ΠHR)))) ((c2) of Definition 2.8)
- =ϕMcoH⊗BcoHB∘(nMcoH⊗(h∘ΠHR)) ((50)),
and then (d2-5) of Definition 2.10 holds.
Consider the morphism ϕM∘(iM⊗B):McoH⊗B→M. By (45) and (d2-2) of Definition 2.10, we obtain that ϕM∘(iM⊗B)∘TM1=ϕM∘(iM⊗B)∘TM2 and, as a consequence, there exists a unique morphism ωM:McoH⊗BcoHB→M such that
[TABLE]
The morphism ωM is a morphism of right H-comodules because
[TABLE]
[TABLE]
and then (ωM⊗H)∘ρMcoH⊗BcoHB=ρM∘ωM.
Finally, ωM is an isomorphism with inverse ωM′=nMcoH∘(pM⊗h)∘ρM. Indeed, firstly note that, by (42), we have that
ωM∘ωM′=idM. On the other hand, composing with nMcoH we have
- ωM′∘ωM∘nMcoH
- =nMcoH∘(pM⊗h)∘ρM∘ϕM∘(iM⊗B) ((52))
- =nMcoH∘((pM∘ϕM∘(M⊗qB))⊗h)∘(iM⊗ρB) ((39))
- =nMcoH∘((ϕMcoH∘(McoH⊗pB))⊗h)∘(McoH⊗ρB) ((46))
- =nMcoH∘(McoH⊗(μB∘(qB⊗h)∘ρB)) ((48))
- =nMcoH ((27)),
and, as a consequence, ωM′∘ωM=idMcoH⊗BcoHB.
∎
Remark 2.17*.*
Note that in the previous proposition the existence of the comodule structure on McoH⊗BcoHB does not depend on the preservation of coequalizers by the functors −⊗B and −⊗H.
Proposition 2.18**.**
Let H be a weak Hopf quasigroup and let (B,ρB) be a right H-comodule
magma. Let h:H→B be an anchor morphism and let (P,ϕP,ρP), (Q,ϕQ,ρQ) be strong (H,B,h)-Hopf modules. If there exists a right H-comodule isomorphism ω:Q→P, the triple (P,ϕPω=ω∘ϕQ∘(ω−1⊗B),ρP), called the ω-deformation of (P,ϕP,ρP), is a strong (H,B,h)-Hopf module.
Proof.
The proof follows easily because, if ω is a right H-comodule isomorphism, ρP=(ω⊗H)∘ρQ∘ω−1.
∎
Definition 2.19**.**
Let H be a weak Hopf quasigroup and let (B,ρB) be a right H-comodule
magma. Let h:H→B be an anchor morphism. Assume that (28) and (47)
hold. Then if the category C admits coequalizers and the functors −⊗B and −⊗H preserve coequalizers, we define the category of strong (H,B,h)-Hopf modules as the one whose objects are strong (H,B,h)-Hopf modules, and whose morphisms f:M→N are morphisms of right H-comodules and
B-quasilinear, i.e.
[TABLE]
where ωM:McoH⊗BcoHB→M, ωN:NcoH⊗BcoHB→N are the isomorphisms of right H-comodules obtained in the proof of Proposition 2.16. This category will be denoted by SMBH(h).
Example 2.20**.**
As particular instances of SMBH(h) we will describe in detail some interesting examples of Hopf module categories associated to Hopf algebras, weak Hopf algebras, Hopf quasigroups and weak Hopf quasigroups.
Examples for Hopf algebras: If H is a Hopf algebra and B=H, ρB=δH, h=idH, the equalities (28) and (47) hold trivially because the product on H is associative. In this case the category SMHH(idH) is the one whose objects are triples (M,ϕM,ρM) where:
The pair (M,ρM) is a right H-comodule.
The morphism ϕM:M⊗H→M satisfies:
ϕM∘(M⊗ηH)=idM,
ρM∘ϕM=(ϕM⊗μH)∘(M⊗cH,H⊗H)∘(ρM⊗δH),
ϕM∘(ϕM⊗λH))∘(M⊗δH)=M⊗εH,
ϕM∘((ϕM∘(M⊗λH))⊗H)∘(M⊗δH)=M⊗εH,
because in this case ΠHL=ΠHR=qH=ηH⊗εH, HcoH=K, and iH=ηH. Then, if (M,ϕM,ρM) is a classical Hopf module in the sense of Larson and Sweedler [24] (see also [34]), we have that (M,ϕM,ρM) is an object in SMHH(idH) because the identity
[TABLE]
holds. Moreover, the morphisms of SMHH(idH) are morphisms of right H-comodules and H-quasilinear, i.e., satisfying (53), where ωM=ϕM∘(iM⊗H):McoH⊗H→M is the associated isomorphism with inverse ωM−1=(pM⊗H)∘ρM, and ϕMcoH⊗H=McoH⊗μH and ρMcoH⊗H=McoH⊗δH. Then any morphism f:M→N of right H-comodules and linear in the classical sense, i.e., such that f∘ϕM=ϕN∘(f⊗H), satisfies (53). Therefore, the Larson-Sweedler category of Hopf modules, denoted by MHH, is a subcategory of SMHH(idH).
Also, in the Hopf algebra setting we have the following more general example: Let (B,ρB) be a right H-comodule monoid such that the functors −⊗B, −⊗H preserve coequalizers. In this case, the equalities (28) and (47) hold trivially because the product on B is associative. For any h multiplicative total integral, therefore an anchor morphism, the category SMBH(h) is the one whose objects are triples (M,ϕM,ρM) where:
The pair (M,ρM) is a right H-comodule.
The morphism ϕM:M⊗B→M satisfies:
ϕM∘(M⊗ηB)=idM,
ϕM∘((ϕM∘(M⊗iB))⊗B)=ϕM∘(M⊗(μB∘(iB⊗B))),
ρM∘ϕM=(ϕM⊗μH)∘(M⊗cH,B⊗H)∘(ρM⊗ρB),
ϕM∘((ϕM∘(M⊗h))⊗(h∘λH))∘(M⊗δH)=M⊗εH,
ϕM∘((ϕM∘(M⊗(h∘λH)))⊗h)∘(M⊗δH)=M⊗εH,
because in this case ΠHL=ΠHR=ηH⊗εH. The morphisms in SMBH(h) are morphisms of right H-comodules and B-quasilinear where ωM:M∘H⊗BcoHH→M is the associated isomorphism of right H-comodules defined in the proof of Proposition 2.16. Then, if (M,ϕM,ρM) is an (H,B)-Hopf module in the sense of Doi [15], we have that (M,ϕM,ρM) is an object in SMBH(h) because the identity
[TABLE]
holds. The morphisms in SMBH(h) are colinear morphisms satisfying (53), for the action ϕM∘H⊗BcoHH and the coaction ρM∘H⊗BcoHH defined in the proof of Proposition 2.16. Then any morphism f:M→N of right H-comodules and right B-linear, i.e., such that f∘ϕM=ϕN∘(f⊗B), satisfies (53). Therefore, the category of right (H,B)-Hopf modules, denoted by MBH, is a subcategory of SMBH(h) (for all multiplicative total integral h).
Examples for weak Hopf algebras: Let H be a weak Hopf algebra. Let (B,ρB) be a right H-comodule monoid such that the functors −⊗B, −⊗H preserve coequalizers. As in the Hopf algebra setting, the equalities (28) and (47) hold trivially because the product on B is associative. Also, in this case, a multiplicative total integral h is an anchor morphism because the product on B is associative, (13) holds and ΠHL∗idH=idH. Then, the category SMBH(h) is the one whose objects are triples (M,ϕM,ρM) where:
The pair (M,ρM) is a right H-comodule.
The morphism ϕM:M⊗B→M satisfies:
ϕM∘(M⊗ηB)=idM,
ϕM∘((ϕM∘(M⊗iB))⊗B)=ϕM∘(M⊗(μB∘(iB⊗B))),
ρM∘ϕM=(ϕM⊗μH)∘(M⊗cH,B⊗H)∘(ρM⊗ρB),
ϕM∘((ϕM∘(M⊗h))⊗(h∘λH))∘(M⊗δH)=ϕM∘(M⊗(h∘ΠHL)),
ϕM∘((ϕM∘(M⊗(h∘λH)))⊗h)∘(M⊗δH)=ϕM∘(M⊗(h∘ΠHR)).
Then, if (M,ϕM,ρM) is an (H,B)-Hopf module in the sense Böhm [11] (see also [37]), the triple (M,ϕM,ρM) is an object in SMBH(h) because the identity (55) holds.
As in the Hopf case, the morphisms in SMBH(h) are colinear morphisms satisfying (53), for ϕM∘H⊗BcoHH and ρM∘H⊗BcoHH the action and the coaction defined in the proof of Proposition 2.16. Then any morphism f:M→N of right H-comodules and right B-linear satisfies (53). Therefore, as in the classical context, the category of right (H,B)-Hopf modules, denoted by MBH, is a subcategory of SMBH(h) (for all multiplicative total integral h). As a consequence, if H=B, ρB=δH and h=idH, we obtain that the category MHH, i.e., the category of Hopf modules associated to H, is a subcategory of SMHH(idH). Note that in this case the objects of SMHH(idH) are triples
(M,ϕM,ρM) where:
The pair (M,ρM) is a right H-comodule.
The morphism ϕM:M⊗H→M satisfies:
ϕM∘(M⊗ηH)=idM,
ϕM∘((ϕM∘(M⊗iL))⊗H)=ϕM∘(M⊗(μH∘(iL⊗H))),
ρM∘ϕM=(ϕM⊗μH)∘(M⊗cH,H⊗H)∘(ρM⊗δH),
ϕM∘(ϕM⊗λH)∘(M⊗δH)=ϕM∘(M⊗ΠHL).
ϕM∘((ϕM∘(M⊗λH))⊗H)∘(M⊗δH)=ϕM∘(M⊗ΠHR),
because in this case BcoH=HL and iB=iL.
Examples for Hopf quasigroups: The following example comes from the nonassociative setting. Let H be a Hopf quasigroup and assume that B=H, ρB=δH, h=idH. In this case the equalities (28) and (47) hold because qH=ηH⊗εH, HcoH=K, and iH=ηH. Then the category SMHH(idH) is the one whose objects are triples (M,ϕM,ρM) where:
The pair (M,ρM) is a right H-comodule.
The morphism ϕM:M⊗H→M satisfies:
ϕM∘(M⊗ηH)=idM,
ρM∘ϕM=(ϕM⊗μH)∘(M⊗cH,H⊗H)∘(ρM⊗δH),
ϕM∘(ϕM⊗(M∘λH))∘(M⊗δH)=M⊗εH,
ϕM∘(ϕM⊗λH)∘(M⊗δH)=M⊗εH.
Note that in this setting the equality ϕM∘((ϕM∘(M⊗iH))⊗H)=ϕM∘(M⊗(μH∘(iH⊗H))) holds because iH=ηH. The morphisms of SMHH(idH) are morphisms of right H-comodules and H-quasilinear, i.e., satisfying (53), where ωM=ϕM∘(iM⊗H):McoH⊗H→M is the associated isomorphism with inverse ωM−1=(pM⊗H)∘ρM, and ϕMcoH⊗H=McoH⊗μH and ρMcoH⊗H=McoH⊗δH. Therefore SMHH(idH) is the category of Hopf modules introduced by Brzeziński in [12].
In the previous nonassociative setting, let (B,ρB) be a right H-comodule magma such that the functors −⊗B and −⊗H preserve coequalizers, and such that the equalities (28) and (47) hold. For any anchor morphism h, the category SMBH(h) is the one whose objects are triples (M,ϕM,ρM) where:
The pair (M,ρM) is a right H-comodule.
The morphism ϕM:M⊗B→M satisfies:
ϕM∘(M⊗ηB)=idM,
ϕM∘((ϕM∘(M⊗iB))⊗B)=ϕM∘(M⊗(μB∘(iB⊗B))),
ρM∘ϕM=(ϕM⊗μH)∘(M⊗cH,B⊗H)∘(ρM⊗ρB),
ϕM∘((ϕM∘(M⊗h))⊗(h∘λH))∘(M⊗δH)=M⊗εH,
ϕM∘((ϕM∘(M⊗(h∘λH)))⊗h)∘(M⊗δH)=M⊗εH,
because in this case ΠHL=ΠHR=ηH⊗εH. The morphisms of SMBH(h) are morphisms of right H-comodules and B-quasilinear where ωM:McoH⊗BcoHH→M is the associated isomorphism of right H-comodules defined in the proof of Proposition 2.16.
Examples for weak Hopf quasigroups: If H is a weak Hopf quasigroup SMHH(idH) is the category of strong Hopf modules defined in [7] and denoted by SMHH. Note that, as a consequence of (32), we obtain that
[TABLE]
is a superfluous identity in the definition of strong Hopf module introduced in [7].
Proposition 2.21**.**
Assume that the conditions of Proposition 2.16 hold. Let (M,ϕM,ρM) be an object in SMBH(h). Let ωM be the isomorphism of right H-comodules between McoH⊗BcoHB and M. Then the triple (M,ϕMωM,ρM) is a strong (H,B,h)-Hopf module with the same object of coinvariants that (M,ϕM,ρM). Also, the identity
[TABLE]
holds and
[TABLE]
where qMωM=ϕMωM∘(M⊗(h∘λH))∘ρM is the idempotent morphism associated to the
Hopf module (M,ϕMωM,ρM). Moreover, for (M,ϕMωM,ρM), the associated isomorphism of right H-comodules between McoH⊗BcoHB and M is ωM, and the equality
[TABLE]
holds. Finally, there exists an idempotent functor
[TABLE]
called the ω-deformation functor, defined on objects by D((M,ϕM,ρM))=(M,ϕMωM,ρM) and on morphisms by the identity.
Proof.
By Proposition 2.18, (M,ϕMωM,ρM) is a strong (H,B,h)-Hopf module. Moreover, note that (56) follows by (50) and (52). Then qMωM=qM because
- qMωM
- =ϕM∘(qM⊗(μB∘(h⊗h)))∘(ρM⊗λH)∘ρM ((56))
- =ϕM∘(qM⊗(h∘ΠHL))∘ρM (comodule condition for M and multiplicative condition for h)
- =ϕM∘(ϕM⊗B)∘(qM⊗h⊗(h∘λH))∘(M⊗δH)∘ρM ((d2-4) of Definition 2.10)
- =ϕM∘((ϕM∘(qM⊗h)∘ρM)⊗(h∘λH))∘ρM (comodule condition for M)
- =qM((42)).
Therefore, (M,ϕMωM,ρM) has the same object of coinvariants that (M,ϕM,ρM). On the other hand,
- (ϕMωM)ωM
- =ϕM∘(qM⊗μB)∘(((M⊗h)∘ρM∘qM)⊗(μB∘(h⊗B)))∘(ρM⊗B) ((56))
- =ϕM∘(qM⊗μB)∘(((M⊗(h∘ΠHL))∘ρM∘qM)⊗(μB∘(h⊗B)))∘(ρM⊗B) ((34))
- =ϕM∘(qM⊗μB)∘(((M⊗(qB∘h))∘ρM∘qM)⊗(μB∘(h⊗B)))∘(ρM⊗B) ((24))
- =ϕM∘((ϕM∘(qM⊗(qB∘h))∘ρM∘qM)⊗(μB∘(h⊗B)))∘(ρM⊗B) ((d2-2) of Definition 2.10)
- =ϕM∘((ϕM∘(qM⊗(h∘ΠHL))∘ρM∘qM)⊗(μB∘(h⊗B)))∘(ρM⊗B) ((24))
- =ϕM∘((ϕM∘(qM⊗h)∘ρM∘qM)⊗(μB∘(h⊗B)))∘(ρM⊗B) ((34))
- =ϕMωM ((42)).
Finally, by (57) and (58), it is easy to show that D is a well defined idempotent endofunctor.
∎
Lemma 2.22**.**
Assume that the conditions of Proposition 2.16 hold. Let (M,ϕM,ρM) be an object in SMBH(h). The following identity holds:
[TABLE]
Proof.
Indeed,
- ϕMωM∘(iM⊗B)
- =ϕM∘(qM⊗μB)∘(((M⊗(h∘ΠHL))∘ρM∘iM)⊗B) ((34))
- =ϕM∘(qM⊗μB)∘(((M⊗(qB∘h))∘ρM∘iM)⊗B) ((24))
- =ϕM∘((ϕM∘(qM⊗(qB∘h))∘ρM∘iM)⊗B) ((d2-2) of Definition 2.10)
- =ϕM∘((ϕM∘(qM⊗(h∘ΠHL))∘ρM∘iM)⊗B) ((24))
- =ϕM∘((ϕM∘(qM⊗h)∘ρM∘iM)⊗B) ((34))
- =ϕM∘(iM⊗B) ((42)).
∎
Proposition 2.23**.**
Assume that the conditions of Proposition 2.16 hold. For any object (M,ϕM,ρM) in SMBH(h), the strong (H,B,h)-Hopf module (McoH⊗BcoHB,ϕMcoH⊗BcoHB,ρMcoH⊗BcoHB), constructed in Proposition 2.16, is invariant for the ω-deformation functor, i.e.,
[TABLE]
Proof.
To prove the proposition we only need to show that
[TABLE]
Indeed, composing with the coequalizer nMcoH⊗B we have
[TABLE]
[TABLE]
Therefore, (60) holds.
∎
The following result is the nonassociative general version of the Fundamental Theorem of Hopf Modules. The proof follows by the properties of the morphism ωM, obtained in Proposition 2.16, and by (60).
Theorem 2.24**.**
*(Fundamental Theorem of Hopf modules)
Assume that the conditions of Proposition 2.16 hold. Let (M,ϕM,ρM) be an object in SMBH(h). The objects (McoH⊗BcoHB,ϕMcoH⊗BcoHB,ρMcoH⊗BcoHB) and (M,ϕM,ρM) are isomorphic in SMBH(h).*
The previous theorem is a generalization of the one proved by Sweedler [34] for Hopf modules over an ordinary Hopf algebra. It also contains the Fundamental Theorem of relative Hopf modules (or (H,B)-Hopf modules, or Doi-Hopf modules) given by Doi and Takeuchi in [16]. On the other hand, in the weak setting, Theorem 2.24 is a generalization of the one obtained by Böhm, Nill and Szlachányi [10], for Hopf modules over a weak Hopf algebra H, and the one proved by Zhang and Zhu [37], for (H,B)-Hopf modules associated to a weak right H-comodule algebra B. Moreover, in the nonassociative context, it generalizes the result obtained by Brzeziński [12] for Hopf modules associated to a Hopf quasigroup. Finally, for weak Hopf quasigroups, Theorem 2.24 is a generalization of the Fundamental Theorem of Hopf modules proved in [4] (see also [7]).
3. Categorical equivalences for strong (H,B,h)-Hopf modules
As for prerequisites, in this section we will assume that the conditions of Proposition 2.16 hold. Then, in the following H is a weak Hopf quasigroup and (B,ρB) is a right H-comodule
magma. Also, h:H→B is an anchor morphism such that (28) and (47)
hold. Finally, the category C admits coequalizers and the functors −⊗B and −⊗H preserve coequalizers. With CBcoH we will denote the category of right BcoH-modules.
The main target of this section is to prove that there exists an equivalence between CBcoH and the category of strong (H,B,h)-Hopf modules.
Let (N,ψN) be an object in CBcoH and consider the coequalizer diagram
[TABLE]
Then,
[TABLE]
and, as a consequence, there exists a unique morphism ρN⊗BcoHB:N⊗BcoHB→N⊗BcoHB⊗H such that
[TABLE]
On the other hand, by (47), we have
[TABLE]
and then,using that the functor −⊗B preserves coequalizers, there exists a unique morphism
[TABLE]
such that
[TABLE]
By a similar proof to the one used for McoH⊗BcoHB in Proposition 2.16, we can prove that
[TABLE]
is a strong (H,B,h)-Hopf module such that (the proof follows the ideas given in Proposition 2.23 for McoH⊗BcoHB)
[TABLE]
On the other hand, if f:N→P is a morphism in CBcoH, we have
[TABLE]
and then there exists a unique morphism f⊗BcoHB:N⊗BcoHB→P⊗BcoHB such that
[TABLE]
The morphism f⊗BcoHB is a morphism in SMBH(h) because
[TABLE]
[TABLE]
and
[TABLE]
[TABLE]
Summarizing, we have the following proposition:
Proposition 3.1**.**
There exists a functor F:CBcoH→SMBH(h), called the induction functor, defined on objects by F((N,ψN))=(N⊗BcoHB,ϕN⊗BcoHB,ρN⊗BcoHB) and on morphisms by F(f)=f⊗BcoHB.
Now let (M,ϕM,ρM) be an object in SMBH(h). By Proposition
2.15 we have that the object of coinvariants McoH is a right BcoH-module where
ϕMcoH=pM∘ϕM∘(iM⊗iB) (see (46)). Let g:M→Q be a morphism in SMBH(h). Using the comodule morphism condition we obtain that ρQ∘g∘iM=(Q⊗ΠHR)∘ρQ∘g∘iM and this implies that there exists a unique morphism gcoH:McoH→QcoH such that
[TABLE]
Also, using that g is B-quasilinear, H colinear, and (57) we have
[TABLE]
Then,
[TABLE]
and, as a consequence,
[TABLE]
holds.
The morphism gcoH is a morphism in CBcoH because
- ϕQcoH∘(gcoH⊗BcoH)
- =pQ∘ϕQ∘((iQ∘gcoH)⊗iB) ((46))
- =pQ∘ϕQωQ∘((iQ∘gcoH)⊗iB) ((59))
- =pQ∘ϕQωQ∘((g∘iM)⊗iB) ((66))
- =pQ∘g∘ϕMωM∘(iM⊗iB) ((53))
- =pQ∘g∘ϕM∘(iM⊗iB) ((59))
- =gcoH∘ϕMcoH ((68)).
Thus, we have the following result.
Proposition 3.2**.**
There exists a functor G:SMBH(h)→CBcoH, called the functor of coinvariants, defined on objects by G((M,ϕM,ρM))=(McoH,ψMcoH) and on morphisms by G(g)=gcoH.
Theorem 3.3**.**
The categories SMBH(h) and CBcoH are equivalent.
Proof.
To prove the theorem, we firstly obtain that the induction functor F, introduced in Proposition 3.1, is left adjoint to the functor of coinvariants G introduced in Proposition 3.2. Later, we show that the unit and counit associated to this adjunction are natural isomorphisms. Then, we proceed as in the proof of Theorem 3.10 of [7] by dividing the proof in three steps.
Step 1: In this step we define the unit of the adjunction. For any right BcoH-module (N,ψN), consider
[TABLE]
as the unique morphism such that
[TABLE]
This morphism exists and is unique because
[TABLE]
[TABLE]
Also, αN is a morphism in CBcoH. Indeed: Composing with the equalizer iN⊗BcoHB we have
- iN⊗BcoHB∘ψ(N⊗BcoHB)coH∘(αN⊗BcoH)
- =qN⊗BcoHB∘ϕN⊗BcoHB∘((iN⊗BcoHB∘αN)⊗iB) ((46))
- =qN⊗BcoHB∘ϕN⊗BcoHB∘((nN∘(N⊗ηB))⊗iB) ((69))
- =qN⊗BcoHB∘nN∘(N⊗(μB∘(ηB⊗iB))) ((63))
- =nN∘(N⊗(qB∘iB)) (properties of ηB, (62), and (63))
- =nN∘(N⊗iB) (properties of iB)
- =nN∘(N⊗(μB∘(iB⊗ηB))) (properties of ηB)
- =nN∘(ψN⊗ηB) ((61))
- =iN⊗BcoHB∘αN∘ψN ((69))
and therefore ψN⊗BcoHB∘(αN⊗BcoH)=αN∘ψN. On the other hand, the morphism αN is natural in N because if f:N→P is a morphism in CBcoH
[TABLE]
[TABLE]
and then (f⊗BcoHB)coH∘αN=αP∘f.
Finally, we prove that αN is an isomorphism for all right BcoH-module N. First note that, under the conditions of this theorem, the triple (B,μB,δB) is a strong (H,B,h)-Hopf module, and then
[TABLE]
because
- ψN∘(ψN⊗pB)
- =ψN∘(N⊗(μBcoH∘(BcoH⊗pB))) (module condition for N)
- =ψN∘(N⊗(pB∘μB∘(iB⊗qB)) ((18))
- =ψN∘(N⊗(pB∘μB∘(iB⊗B))) ((41)).
Therefore, there exists a unique morphism mN:N⊗BcoHB→N such that
[TABLE]
Now define the morphism xN:(N⊗BcoHB)coH→N by xN=mN∘iN⊗BcoHB. Then,
[TABLE]
On the other hand, composing with
iN⊗BcoHB and pN⊗BcoHB∘nN we have
- iN⊗BcoHB∘αN∘xN∘pN⊗BcoHN∘nN
- =nN∘((mN∘qN⊗BcoHB∘nN)⊗ηB) ((69))
- =nN∘((ψN∘(N⊗(pB∘qB)))⊗ηB) ((62), (63), (70))
- =nN∘((ψN∘(N⊗pB))⊗ηB) (properties of qB)
- =nN∘(N⊗(μB∘(qB⊗ηB))) ((61))
- =nN∘(N⊗qB) (properties of ηB)
- =qN⊗BcoHB∘nN ((62), (63)).
Therefore, αN∘xN=id(N⊗BcoHB)coH and, as a consequence, αN is an isomorphism.
Step 2: For any (M,ϕM,ρM)∈SMBH(h) the counit is defined as
βM=ωM:McoH⊗BcoHB→M,
where ωM is the isomorphism satisfying (52). By Theorem 2.24, we know that ωM is an isomorphism in SMBH(h), and it is natural because if g:M→Q is a morphism in SMBH(h) we have
[TABLE]
[TABLE]
and thus βQ∘gcoH⊗BcoH⊗B=g∘βM.
Step 3: Now we prove the triangular identities for the unit and counit previously defined. Indeed: The first triangular identity holds because composing with nN we have
- βN⊗BcoHB∘(αN⊗BcoHB)∘nN
- =βN⊗BcoHB∘n(N⊗BcoHB)coH∘(αN⊗B) (by (65))
- =ϕN⊗BcoHB∘((iN⊗BcoHB∘αN)⊗B) (by (52))
- =ϕN⊗BcoHB∘((nN∘(N⊗ηB))⊗B) (by (69))
- =nN∘(N⊗(μB∘(ηB⊗B))) (by (63))
- =nN (by the unit properties).
Finally, if we compose with iM,
[TABLE]
and then βMcoH∘αMcoH=idMcoH.
∎
As a consequence, we obtain the following particular instances of our main theorem.
Corollary 3.4**.**
The following assertions hold:
Let H be a Hopf algebra and let (B,ρB) be a right H-comodule monoid such the functors −⊗B and −⊗H preserve coequalizers. Then, if there exists a multiplicative total integral h:H→B, the categories of right (H,B)-Hopf modules, denoted by MBH and introduced by Doi in **[15]**, the category SMBH(h) of strong (H,B,h)-Hopf modules, and the category of right BcoH-modules CBcoH are equivalent. In particular, if B=H and ρB=δH, the Sweedler category of Hopf modules MHH, the category SMHH(idH) of strong (H,H,idH)-Hopf modules, and the category C are equivalent.
Let H be a weak Hopf algebra and let (B,ρB) be a right H-comodule monoid such that the functors −⊗B and −⊗H preserve coequalizers. Then, if there exists a multiplicative total integral h:H→B, the categories of right (H,B)-Hopf modules, denoted by MBH and introduced by Böhm in **[11]** (see also **[37]** and **[18]** for the categorical equivalence), the category SMBH(h) of strong (H,B,h)-Hopf modules, and the category of right BcoH-modules CBcoH are equivalent. In particular, if B=H and ρB=δH, the category of Hopf modules MHH, the category SMHH(idH) of strong (H,H,idH)-Hopf modules, and the category CHL of right HL-modules are equivalent.
Let H be a Hopf quasigroup and let (B,ρB) be a right H-comodule magma such the functors −⊗B and −⊗H preserve coequalizers. Then, if there exists an anchor morphism h:H→B, the categories SMBH(h) of strong (H,B,h)-Hopf modules, and the category of right BcoH-modules CBcoH are equivalent. In particular, if B=H and ρB=δH, we obtain the result proved by Brzeziński in **[12]**: the category of Hopf modules MHH=SMHH(idH) and the category C are equivalent.
Let H be a weak Hopf quasigroup such the functor −⊗H preserves coequalizers. The category of strong Hopf modules SMHH=SMHH(idH), and the category CHL of right HL-modules are equivalent (this is the main result proved in **[7]**).
Example 3.5**.**
- Consider H a Hopf quasigroup, A a unital magma in C, and t φA:H⊗A→A a morphism satisfying (9), (10). By the third points of Examples 2.4 and 2.9 we know that the smash product A♯H is a right H-comodule magma with coaction ρA♯H=A⊗δH and h=ηA⊗H:H→A♯H is an anchor morphism.
Moreover, if A is a monoid and the equality
[TABLE]
holds, then so hold (28) and (47). Indeed, first note that
- qA♯H
- =A⊗(μH∘(H⊗λH)∘δH ((20))
- =A⊗ηH⊗εH. ((2)).
Therefore (A♯H)coH=A, pA♯H=A⊗εH and iA♯H=A⊗ηH. As a consequence, we have
- μA♯H∘((μA♯H∘(A⊗H⊗iA♯H))⊗A⊗H)
- =(μA⊗μH)∘(A⊗(((μA∘(φA⊗φA))⊗H)∘(H⊗A⊗H⊗cH,A)∘(H⊗A⊗δH⊗A)∘(H⊗cH,A⊗A)
- ∘(δH⊗A⊗A))⊗H) (unit properties and associativity of μA)
- =(μA⊗μH)∘(A⊗(((μA∘(φA⊗φA)∘(H⊗cH,A⊗A)∘(δH⊗A⊗A))⊗H)∘(H⊗A⊗cH,A)
- ∘(H⊗cH,A⊗A)∘(δH⊗A⊗A))⊗H) (coassociativity of δH and naturality of c )
- =(μA⊗μH)∘(A⊗(((φA∘(H⊗μA))⊗H)∘(H⊗A⊗cH,A)∘(H⊗cH,A⊗A)∘(δH⊗A⊗A))⊗H)
- ((71))
- =μA♯H∘(A⊗H⊗μA⊗H) (naturality of c)
- =μA♯H∘(A⊗H⊗(μA♯H∘(iA♯H⊗A⊗H))) ((3), naturality of c, unit properties, and (9)),
and, on the other hand,
- μA♯H∘(iA♯H⊗μA♯H)
- =(μA⊗H)∘(A⊗μA♯H) ((3), naturality of c, unit properties, and (9))
- =μA♯H∘(μA⊗H⊗A⊗H) (associativity of μA)
- =μA♯H∘((μA♯H∘(iA♯H⊗A⊗H))⊗A⊗H) ((3), naturality of c, unit properties, and (9)).
Therefore, if −⊗A and −⊗H preserve coequalizers, by (iii) of the previous Corollary, we have that the categories SMA♯HH(h) and CA are equivalent.
An interesting example of this case can be found using the theory developed in [23]. Let K be a field and let C be the symmetric monoidal category of vector spaces over K. Let G be the abelian group Z2n and let F:G×G→K∗ be a 2-cochain, i.e., F is a morphism such that F(θ,a)=F(a,θ)=1 for all a∈G where θ is the group identity. The group algebra of G, denoted by KG, is a K-vector space with basis {ea;a∈G} and also a unital magma with the product (see [1]):
[TABLE]
In what follows we will denote this magma by KFG. As was pointed in [23], this algebraic object lives in the symmetric monoidal category of G-graded spaces with associator defined by the 3-cocycle ϕ(a,b,c)=F(a,b)F(a+b,c)F(b,c)−1F(a,b+c)−1 and symmetry defined by R(a,b)=F(a,b)F(b,a)−1. For example, the choice of G=Z23 and certain F gives the octonions. Moreover, KFG is a composition algebra with respect to the Euclidean norm in basis G if two suitable conditions hold (see (2.1) and (2.2) of [23]). This means that the norm q(a∑uaea)=a∑ua2 is multiplicative. Then
[TABLE]
is closed under the product in KFG. By Proposition 3.6 of [23] we know that S2n−1 is an IP loop, that becomes an usual sphere if we work over R, and then its loop algebra, denoted by KS2n−1 is a cocommutative Hopf quasigroup (see Proposition 4.7 of [23]). Let H be KS2n−1 and let A be the group algebra of G. Then, A is a monoid (it is a cocommutative Hopf algebra) and we have an action φA:H⊗A→A, where ⊗=⊗K, defined by
[TABLE]
It is easy to see that φA satisfies (9), (10) and (71) and, as a consequence of the general theory, we have a categorical equivalence between SMKZ2n♯KS2n−1KS2n−1(h) and CKZ2n for h=ηKZ2n⊗idKS2n−1.
- By the second points of Examples 2.4 and 2.9 we know that, if H is a cocommutative weak Hopf quasigroup and C is symmetric, (Hop,ρHop=(H⊗λH)∘δH) is an example of right H-comodule magma and λH is an anchor morphism. Then, by Theorem 3.22 of [4] and the cocommutativity of H, we have the equality:
[TABLE]
Therefore, iHop=iL, pHop=pL and (Hop)coH=HL. On the other hand, by the naturality of c and (5) we obtain that
[TABLE]
and by the naturality of c and (6),
[TABLE]
Therefore, we have (28) and (47). As a consequence, if the category C admits coequalizers and the functor −⊗H preserves coequalizers, by Theorem 3.3 we obtain an equivalence between the categories SMHopH(λH) and CHL. If H is a Hopf quasigroup, we have a similar result that asserts the following: The categories SMHopH(λH) and C are equivalent.
Acknowledgements
The authors were supported by Ministerio de Economía y Competitividad (Spain), grant MTM2016-79661-P. AEI/FEDER, UE, support included (Homología, homotopía e invariantes categóricos en grupos y álgebras no asociativas).