HOPF ALGEBRAS ASSOCIATED TO TRANSITIVE PSEUDOGROUPS IN CODIMENSION 2
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the Degree Doctor of
Philosophy in the Graduate School of the Ohio State University
By
Jos´eRodrigo Cervantes Polanco , MS in Mathematics
Graduate Program in Mathematics
The Ohio State University
2016
Dissertation Committee:
Henri Moscovici, Advisor
James Cogdell
Thomas Kerler c Copyright by
Jos´eRodrigo Cervantes Polanco
2016 ABSTRACT
We associate two different Hopf algebras to the same transitive but not primitive pseudogrup of local diffeomorphisms on R2 leaving invariant the trivial foliation where we identify R2 as a product of lines R1 × R1. Their construction is based on ideas used to build the Hopf algebras associated to primitive Lie pseudogroups by Connes-Moscovici and Moscovici-Rangipour. Each of the two Hopf algebras is first defined via its action on the respective crossed product algebra associated to the pseudogroup, and then it is realized as a bicrossed product of a universal enveloping algebra of a Lie algebra and a Hopf algebra of regular functions on a formal group. Using the bicrossed product structure we prove that, although the two Hopf algebras are not isomorphic, they have the same periodic Hopf cyclic cohomology. More precisely, for each of them the periodic Hopf cyclic cohomology is canonically isomorphic to the Gelfand-Fuks cohomology of the infinite dimensional
Lie algebra related with the pseudogroup.
ii To my family and friends
iii ACKNOWLEDGMENTS
I would like to thank my advisor Henri Moscovici for his continuous support and advice during the completion of my Ph.D. degree. He has been a great academic men- tor and a wonderful friend for me. This dissertation grew out of numerous conversa- tions with him and all the team members from the research group of noncommutative geometry. I would like to thank my Ph.D. committee members, James Cogdell and
Thomas Kerler for their help during my time in the Department of Mathematics.
I also want to express my thanks to my friends who helped me many times during this process: Weitao Chen, Yang Liu, Angelo Nasca, Donald Robertson, Xiaoyue Xia,
Tao Yang and Kun Wang.
Finally, I wish to thank my parents Jos´eJavier Cervantes and Nilsa Concepci´on
Polanco Trujeque, and my girlfriend Evelyn Rodriguez for their support. It is their love and encouragement that lead me to this point of my life.
This project would have been impossible without the support of The Ohio State
University and CONACyT.
iv VITA
2005 ...... B.S. in Mathematics, University of Guanajuato, M´exico.
2007 ...... M.S. in Mathematics, CIMAT, M´exico.
2008-Present ...... Graduate Teaching Associate, The Ohio State University, USA.
FIELDS OF STUDY
Major Field: Mathematics
Specialization: Noncommutative Geometry
v TABLE OF CONTENTS
Abstract...... ii
Dedication...... ii
Acknowledgments...... iv
Vita...... v
CHAPTER PAGE
1 Introduction...... 1
2 Preliminaries...... 5
2.1 Basic concepts...... 5
3 Construction via Hopf actions...... 19
3.1 The Hopf algebra HH ...... 19 4 Bicrossed product realization...... 46
4.1 The bicrossed product Hopf algebra FH ICU(gh)...... 46 5 Hopf cyclic cohomology...... 64
∗ 5.1 HP (HH )...... 64 Bibliography...... 80
vi CHAPTER 1
INTRODUCTION
Hopf algebras were found in algebraic topology in 1941. The first example was constructed in the paper of Heinz Hopf in his computation of the rational cohomol- ogy of compact connected Lie groups [18]. The first book on Hopf algebras was written by Moss E. Sweedler in 1969 [30]. A large class of Hopf algebras are the
“quantum groups”. This term was coined by Vladimir Drinfel’d in his address to the
International Congress of Mathematicians in Berkeley on 1986 [10]. It stands for cer- tain special noncommutative noncocommutative Hopf algebras which are non-trivial deformations of the universal enveloping algebras of classical Lie algebras or of the algebra of regular functions on the corresponding Lie groups.
Hopf algebras appear in many mathematical fields such as algebraic geometry, Lie theory, quantum mechanics, etc.
On the other hand, cyclic cohomology was discovered by Alain Connes in 1981; one of his main motivations came from index theory on foliated spaces [4]. Independently, cyclic homology was shown to be the primitive part of the Lie algebra homology of matrices by Boris Tsygan [31] and also by Jean-Louis Loday and Daniel Quillen [21].
This (co)homology is closely related to K-theory and has many interesting rela- tionships with several branches of mathematics. It can be seen as an extension of de
Rham theory, Hochschild (co)homology, group (co)homology.
1 The Hopf algebras Hn associated to the pseudogroup of local diffeomorphisms
n of R were found by Alain Connes and Henri Moscovici in their work on the local index formula for transversely hypoelliptic operators on foliations [5]. Extending the construction of Hn, Henri Moscovici and Bahram Rangipour defined a Hopf algebra
HΠ for each infinite primitive Lie-Cartan pseudogroup Π of local diffeomorphisms of
n R [25]. As reference for primitive pseudogroups we are using Singer and Sternberg [29], and Guillemin [13].
In this dissertation we are implementing a similar construction for the transitive
2 but not primitive pseudogrup of local diffeomorphisms on R leaving invariant the 2 1 1 trivial foliation where we identify R as a product of lines R × R , but in this case we are able to define two different Hopf algebras.
The reason for which one obtains two different Hopf algebras is because the con- struction relies on splitting the group G of globally defined diffeomorphims of the pseudogroup as a set-theoretical product of two groups
G = G · N,G ∩ N = { Id } .
For example, in each flat primitive pseudogroup, there is a canonical splitting
n where G is the subgroup consisting of the affine transformations of R that are in G, while N is the subgroup consisting of those diffeomorphisms in G that preserve the origin and its tangent map at zero is the identity matrix.
For our case, G is the group of globally defined diffeomorphism of the form
1 2 ϕ(x1, x2) = (ϕ (x1, x2), ϕ (x2)) and one can consider two different splittings
G = GH · NH ,GH ∩ NH = { Id } . where H is either the subgroup of upper triangular matrices or the subgroup of diagonal matrices of GL2(R). For each of this splittings, GH is the subgroup of 2 “affine H-motions” transformations of the form Lx ◦ y where Lx are translations by
2 x ∈ R and y ∈ H, and NH is the subgroup of elements of G preserving the origin and with “H-tangent map” at zero the identity matrix. We will denoted by HH the Hopf algebra which depends on the group H.
For all primitive pseudogroups Π, it is proved that there is a quasi-isomorphism between the periodic Hopf cyclic cohomology of HΠ and the continuous cohomology of the Lie algebra of formal vector fields aΠ related to the pseudogroup [5, 26]. This type of cohomology is also called Gelfand-Fuks cohomology [11, 12].
The infinite dimensional Lie algebra a related to the pseudogrop discussed on this dissertation is given by all formal vector fields v of the form
1 ∂ 2 ∂ v = p (x1, x2) + p (x2) ∂x1 ∂x2
1 2 where p (x1, x2) and p (x2) are formal polynomials in their respective variables. In this dissertation, it is shown that the periodic Hopf cyclic cohomology for each
Hop algebra HH is canonically isomorphic to the same continuous cohomology of the infinite dimensional Lie algebra a. In this way, we have obtained two non-isomorphic
Hop algebras with the same periodic Hopf cyclic cohomology.
This dissertation is organized as follows.
In chapter §2, we introduce some background material on Hopf algebras, coho- mology and pseudogroups.
In chapter §3, the algebra HH is defined via its natural action on the crossed
∞ product algebra AH = Cc (GH ) o G. It is shown that every element h ∈ HH satisfies a “Leibniz rule” of the form
h(ab) = h(1)(a)h(2)(b) , for all a, b ∈ AH .
This property together with an “invariant” trace τ gives us the coproduct structure and the antipode map. Hence, we equipp HH with a Hopf algebra structure. 3 In chapter §4, the Hop algebra HH is realized as a bicrossed product Hopf algebra. The components of the bicrossed product realization are the universal enveloping algebra of the Lie algebra gh corresponding to GH , and the Hopf algebra of functions on NH given by the non-trivial coefficients of the Taylor expansion of ψ ∈ NH at zero. It is worth mentioning that in this chapter in order to prove the compatibility conditions of the two Hopf algebras, we employed the Lie-Hopf algebra techniques developed in [27], instead of the more geometric techniques as in [25]. The reason for this is to provide a detailed account for both proofs in written form.
In chapter §5, by using the bicrossed product structure of HH , we prove that the periodic Hopf cyclic cohomology of HH is the same for both H, either the subgroup of upper triangular matrices or the subgroup of diagonal matrices. In fact, for each of them it is canonically isomorphic to the continuous cohomology of the infinite dimensional algebra a. We use the techniques developed in [5, 26] to prove this isomorphism.
4 CHAPTER 2
PRELIMINARIES
The basic concepts related to Hopf algebras, cohomology, and pseudogroups are given through definitions and examples.
2.1 Basic concepts.
We begin this chapter with a resume of concepts related to Hopf algebras [30,1,8].
Let F be a field and G a finite group. All unadorned tensor products are over F. An algebra is a triple (A, µ, η) with A a vector space, µ : A ⊗ A → A a linear map called the product, η : F → A a linear map called the unit, and such that the following diagrams commute
µ ⊗ Id η ⊗ Id Id ⊗ η A ⊗ A ⊗ A A ⊗ A F ⊗ A A ⊗ A A ⊗ F
Id ⊗ µ µ =∼ µ =∼
µ A ⊗ A A A
We have denoted by Id the identity map of A, and the unnamed arrows from the second diagram are the canonical isomorphisms.
op Let (A, µ, η) be an algebra an the map µ := µ ◦ τ where τ(a1 ⊗ a2) = a2 ⊗ a1, then (Aop = A, µop, η) is an algebra which we call the opposite algebra.
5 Let C(G) denote the set of all functions defined on G with values in F. It is a vector space via
(λf1)(g) = λf1(g) , (f1 + f2)(g) = f1(g) + f2(g) , f1, f2 ∈ C(G) , λ ∈ F .
It becomes an algebra when we define the product and the unit of C(G) by
µ(f1, f2)(g) = f1(g)f2(g) , η(1) = 1 , f1, f2 ∈ C(G) , where 1 ∈ C(G) is the function defined by 1(g) = 1 ∈ F.
A coalgebra is a triple (C, ∆, ε) with C a vector space, ∆ : C → C ⊗ C a linear map called the coproduct, ε : C → C a linear map called the counit, and such that the following diagrams commute
∆ ε ⊗ Id Id ⊗ ε C C ⊗ C F ⊗ C C ⊗ C C ⊗ F
∆ Id ⊗ ∆ =∼ ∆ =∼
∆ ⊗ Id C ⊗ C C ⊗ C ⊗ C C
cop Let (C, ∆, ε) be a coalgebra an the map ∆ := τ ◦ ∆ where τ(c1 ⊗ c2) = c2 ⊗ c1, then (Ccop = C, ∆cop, ε) is a coalgebra which we call the coopposite coalgebra.
If we identify C(G) ⊗ C(G) with C(G × G), which is fine since G is finite, then
C(G) becomes a coalgebra when we define the coproduct and the counit via
∆(f)(g1 ⊗ g2) = f(g1g2) ε(f) = f(e) , f ∈ C(G) , where e is the unit element in G.
We will use Sweedler notation [30] and omit the summation to denote the comul- tiplication by
∆(c) = c(1) ⊗ c(2) .
6 Using this notation we can write the commutative diagrams for a coalgebra as follows
c(1)(1) ⊗ c(1)(2) ⊗ c(2) = c(1) ⊗ c(2)(1) ⊗ c(2)(2) ,
ε(c(1))c(2) = c = c(1)ε(c(2)) .
The first element is usually written in the literature as c(1) ⊗ c(2) ⊗ c(3).
Let (A, µA, ηA), (B, µB, ηB) be two algebras. The linear map i : A → B is a morphism of algebras if the following diagrams are commutative
i ⊗ i i A ⊗ A B ⊗ B A B
µA µB ηA ηB
i A B F
The tensor product A ⊗ B is an algebra with µ :(A ⊗ B) ⊗ (A ⊗ B) → A ⊗ B and
η : F → A ⊗ B defined by
µ = (µA ⊗ µB)(Id ⊗ τ ⊗ Id) , where τ(b ⊗ a) = (a ⊗ b) ,
η = (ηA ⊗ ηB) .
Let (C, ∆C , εC ), (D, ∆D, εD) be two coalgebras. The linear map j : C → D is a morphism of coalgebras if the following diagrams are commutative
j j C D C D
∆C ∆D εC εD
j ⊗ j C ⊗ C D ⊗ D F
The commutativity of the first diagram may be written in the Sweedler notation as follows
j(c)(1) ⊗ j(c)(2) = j(c(1)) ⊗ j(c(2)) . 7 The tensor product C ⊗ D is a coalgebra with ∆ : C ⊗ D → (C ⊗ D) ⊗ C ⊗ D and
ε : C ⊗ D → F defined by
∆ = (Id ⊗ τ ⊗ Id)(∆A ⊗ ∆B) , where τ(c ⊗ d) = (d ⊗ c) ,
ε = εC · εD .
A bialgebra is a quintuple (H, µ, η, ∆, ε), with (H, µ, η) an algebra, (H, ∆, ε) a coalgebra and the linear maps µ and η are morphisms of coalgebras (or equivalently, the maps ∆ and ε are morphisms of algebras), i.e., the diagrams below commute
µ µ H ⊗ H H H ⊗ H H
(Id ⊗ τ ⊗ Id)(∆ ⊗ ∆) ∆ ε · ε ε
µ ⊗ µ (H ⊗ H) ⊗ (H ⊗ H) H ⊗ H F
η η F H F H
ε ∆F ∆ F ε η ⊗ η F ⊗ F H ⊗ H F
where F is a Hopf algebra via ∆F(1) = 1 ⊗ 1 and εF(1) = 1.
Let (A, µ, η) be an algebra and (C, ∆, ε) a coalgebra. We define on the set
HomF(C,A) an algebra structure in which the product, denoted by ? and called the convolution product in Hom(C,A), is given as follows
α ? β(c) = α(c(1))β(c(2)) , α, β ∈ Hom(C,A) , and the unit is given by 1 → η ◦ ε ∈ Hom(C,A).
Let H be a bialgebra. A linear map S : H → H is called an antipode of the bialgebra H if S is the inverse of the identity map Id : H → H with respect to the convolution product in Hom(H,H). 8 Definition 2.1.1. A Hopf algebra is a sixtuple (H, µ, η, ∆, ε, S) with (H, µ, η, ∆, ε) a bialgebra and S is the antipode of H.
The linear maps µ and η for C(G) are morphisms of coalgebras, then C(G) is a bialgebra. Furthermore, it has an antipode which is defined by
S(f)(g) = f(g−1) , f ∈ C(G) , since
−1 −1 f(1)S(f(2)) (g) = f(1)(g)S(f(2))(g) = f(1)(g)f(2)(g ) = f(gg ) = f(e) = ε(f)η(1)(g) ,
similarly S(f(1))f(2) = ε(h)η(1). In conclusion, C(G) is a Hopf algebra.
In a Hopf algebra, the antipode is unique, being the inverse of the element Id in the algebra Hom(H, H). The fact that S is the antipode is written as S∗Id = η◦ε = Id ∗S, and using the Sweedler notation
S(h(1))h(2) = ε(h)η(1) = h(1)S(h(2)) , h ∈ H.
Let H1 and H2 be two Hopf algebras. A map α : H1 → H2 is called a morphism of Hopf algebras if it is a morphism of bialgebras, i.e. , a morphism of algebras and a morphism of coalgebras.
Proposition 2.1.2. Let H1 and H2 be two Hopf algebras with antipodes S1 and S2.
If α : H1 → H2 is a morphism of bialgebras, then S2 ◦ α = α ◦ S1.
Proof. It follows by the uniqueness of the antipode on a Hopf algebra.
Let A be an algebra. A left A-module is a pair (M,.) with M a vector space, and . : A ⊗ M → M a morphism of vector spaces such that the following diagrams are commutative
9 Id ⊗ . η ⊗ Id A ⊗ A ⊗ M A ⊗ M F ⊗ M A ⊗ M
µ ⊗ Id . =∼ .
. A ⊗ M M M
Similarly, one can define right A-modules, the only difference being that the structure map of the right module M is of the form / : M ⊗ A → M. An A-bimodule M is a vector space with both a left and a right module structure.
We will use the following notation for left and right modules, respectively,
a . m := .(a ⊗ m) , and m / a := /(m ⊗ a) , a ∈ A , m ∈ M.
Let C be a coalgebra. A left C-comodule is a pair (N, H) with N a vector space, and H : N → C ⊗ N a morphism of vector spaces such that the following diagrams are commutative
H ε ⊗ Id N C ⊗ N C ⊗ N F ⊗ N
∼ H ∆ ⊗ Id H =
Id ⊗ C ⊗ N H C ⊗ C ⊗ N N
Simarly, one can define right C-comodules, the difference being that the structure of a right C-comodule N is of the form H : N → N ⊗ C. An C-bicomodule N is a vector space with both a left and a right comodule structure.
We will also use Sweedler notation for left and right comodules, respectively,
H(n) := n<−1> ⊗ n<0> , and H(n) := n<0> ⊗ n<1>, n ∈ N. where n<0> always stay in the same vector space N while n<−1>, n<1> are in the coalgebra C and the −/+ sign tell the left/right direction of the coaction, respectively. 10 Next, we introduce a quick summary on complexes and bicomplexes [3, 23, 32].
Let k be an associative ring.
∗ n A cochain complex { C , d } of k-modules, is a family { C }n∈Z of k-modules, together with k-module homomorphism dn : Cn → Cn+1 such that each composite dn−1 ◦ dn = 0. The maps dn are called the differentials of C. The kernel of dn is the module of n-cocycles of C, denoted by Zn(C). The image of dn−1 is the module of n-coboundaries of C, denoted by Bn(C). Because dn−1 ◦ dn = 0, we have
n n n B ⊂ Z ⊂ C , n ∈ Z .
The nth-cohomology module of { C∗, d } is the subquotient of Cn given by
Hn(C) = Zn(C)/Bn(C) .
A map of complexes f : C → D is a family of k-module homomorphism f n : Cn →
Dn commuting with dn in the sense that f n+1 ◦ dn = dn ◦ f n. It induces a map fˆn : Hn(C) → Hn(D). The map of complexes f is called a quasi-isomorphism if fˆn is an isomorphism for all n.
Let A be a algebra over F and M an A-bimodule. Consider the family of F-modules n ⊗n C (A, M) = HomF(A ,M) and define the face operators
δi : Cn(A, M) → Cn+1(A, M) , 0 ≤ i ≤ n ,
0 δ (f)(a0 ⊗ · · · ⊗ an) = a0 . f(a1 ⊗ · · · ⊗ an) ,
i δ (f)(a0 ⊗ · · · ⊗ an) = f(a1 ⊗ · · · ⊗ ai−1ai ⊗ · · · ⊗ an) , 1 ≤ i ≤ n − 1 ,
n δ (f)(a0 ⊗ · · · ⊗ an) = f(a0 ⊗ · · · ⊗ an−1) / an .
Definition 2.1.3. The Hochschild cohomology H∗(A, M) of A with coefficients in M is defined to be the nth-cohomology modules of { C∗(A, M), b }, where n X bn = (−1)iδi . i=0 11 Let C be a coalgebra over F and N a C-bicomodule. Consider the family of n ⊗n F-modules C (C,N) = N ⊗ C and define the face operators
δi : Cn(C,M) → Cn+1(C,M) , 0 ≤ i ≤ n
0 δ (m ⊗ c1 ⊗ · · · ⊗ cn) = m<0> ⊗ m<1> ⊗ c1 ⊗ · · · ⊗ cn ,
i δ (m ⊗ c1 ⊗ · · · ⊗ cn) = m ⊗ c1 ⊗ · · · ⊗ ∆(ci) ⊗ · · · ⊗ cn , 1 ≤ i ≤ n − 1 ,
n δ (m ⊗ c1 ⊗ · · · ⊗ cn) = m<0> ⊗ c1 ⊗ · · · ⊗ cn ⊗ m<−1> .
Definition 2.1.4. The Hochschild cohomology H∗(C,N) of C with coefficients in N is defined to be the nth-cohomology modules of { C∗(C,N), b }, where
n X bn = (−1)iδi . i=0
Let g a Lie algebra over F and V a finite dimensional left g-module. Consider the n family of F-modules C (g,V ) of antisymmetric n-linear maps on g with values in V , n n i.e. , C (g,V ) = HomF(∧ g,V )
Definition 2.1.5. The cohomology H∗(g,V ) of the algebra g with coefficients in V is defined to be the nth-cohomology modules of { C∗(g,V ), d }, where
X s+t−1 d(ω)(x1, . . . , xn+1) = (−1) ω([xs, xt], x1,..., xˆs,..., xˆt, . . . , xn+1) 1≤s X s + (−1) xi ω(x1,..., xˆs, . . . , xn+1) , 1≤s≤n+1 where the sign ˆ indicates that the argument below it must be omitted. If g is a topological Lie algebra and the vector space V is a topological space, the action will be continuous and Cn(g,V ) is restricted to continuous maps. The ∗ cohomology is denoted by Hcont(g,V ). 12 Examples of Lie algebras: • Finite dimensional. Let gln(R) be the algebra of n × n matrices with coeffi- cients in R with Lie bracket given by [X,Y ] = XY − YX,X,Y ∈ gln(R) . n • Infinite dimensional. Let an(R) be the space of formal vector fields in R with Lie bracket defined by " # ! X ∂ X ∂ X X ∂gi ∂fi ∂ f , g = f − g , i ∂x i ∂x j ∂x j ∂x ∂x i i i i i j j j i in which f and g denote formal power series in x1, . . . , xn. It is a topological Lie algebra when it is equipped with the projective limit topology. A cochain bicomplex is a family of k-modules Cp,q, together with maps dh : Cp,q → Cp+1,q , and dv : Cp,q → Cp,q+1 , such that dh ◦ dh = dv ◦ dv = dv ◦ dh + dh ◦ dv = 0. It is useful to picture the bicomplex C•,• as a lattice ...... dh dh ··· Cp−1,q+1 Cp,q+1 Cp+1,q+1 ··· dv dv dv dh dh ··· Cp−1,q Cp,q Cp+1,q ··· dv dv δv dh dh ··· Cp−1,q−1 Cp,q−1 Cp+1,q−1 ··· ...... 13 in which the maps dh go horizontally, the maps dv go vertically, and each square anticommutes. Each row C•,q and each column Cp,• is a chain complex. Define the total complexes Y M TotΠ(C)n = Cp,q and Tot⊕(C)n = Cp,q . p+q=n p+q=n The formula d = dh + dv defines maps such that d ◦ d = 0, making TotΠ(C) and Tot⊕(C) into chain complexes. Note that in some special cases they are equal, for example if C is a first quadrant double complex. n ⊗(n+1) Let A be an algebra over F. Let C (A) = HomF(A , F) and the face, degen- eracy and cyclic operators n−1 n δi : C (A) → C (A) , 0 ≤ i ≤ n , n+1 n σi : C (A) → C (A) , 0 ≤ i ≤ n . n n τn : C (A) → C (A) . defined by δi(c)(a0 ⊗ · · · ⊗ an) = c(a0 ⊗ aiai+1 ⊗ an) , 0 ≤ i ≤ n − 1 , δn(c)(a0 ⊗ · · · ⊗ an) = c(ana0 ⊗ · · · ⊗ an−1) , σi(c)(a0 ⊗ · · · ⊗ an) = c(a0 ⊗ · · · ⊗ ai ⊗ 1 ⊗ ai+1 ⊗ · · · ⊗ an) , 0 ≤ i ≤ n . τn(c)(a0 ⊗ · · · ⊗ an) = c(an ⊗ a0 ⊗ · · · ⊗ an−1) . Consider the family of F-modules q−p C (A), if q ≥ p, Cp,q(A) = 0, if q < p, 14 Definition 2.1.6. The cyclic cohomology HC∗(A) of A is defined to be the nth co- homology modules of the first quadrant bicomplex Cp,q(A) ...... b b b B B C2(A) C1(A) C0(A) b b B C1(A) C0(A) b C0(A) where b : Cn(A) → Cn+1(A),B : Cn+1(A) → Cn(A) , n n X i X ni i n b = (−1) δi ,B = ( (−1) τn)σnτn+1(1 − (−1) τn+1) . i=0 i=0 A variant of the cyclic cohomology is the periodic cyclic cohomology HP ∗(A) which is Z2-graded. It is defined as the even and odd cohomology of the total complex shown above, but extended indefinitely to the left and down. For completeness, we also define the concept of pseudogroup [20, 29]. Definition 2.1.7. A pseudogroup of transformations on a topological space S is a set Γ of transformations satisfying the following axioms: 1. Each ψ ∈ Γ is a homeomorphism of an open set (called the domain of ψ) of S onto another open set (called the range of φ) of S; 2. If ψ ∈ Γ, then the restriction of ψ to an arbitrary open subset of the domain of ψ is in Γ; 15 [ 3. Let U = Ui where each Ui is an open set of S. A homeomorphism ψ of U i onto an open set of S belongs to Γ if the restrictions of ψ to Ui is in Γ for all i; 4. For every open set U of S, the identity transformation of U is in Γ; 5. If ψ ∈ Γ, then ψ−1 ∈ Γ; 6. If ψ ∈ Γ is a homeomorphism of U onto V and ψ0 ∈ Γ is a homeomorphism of U 0 onto V 0 and if V ∩ U 0 is non-empty, then the homeomorphism ψ0 ◦ ψ of ψ−1(V ∩ U 0) onto ψ0(V ∩ U 0) is in Γ. n n A diffeomorphism ψ of an open set U ⊂ R onto an open set V ∈ R is called n a local diffeomorphism of R . We will be interested in certain pseudogroups defined n n on R and all pseudogroups appearing in this dissertation will be defined on R and n the elements in Γ will be local diffeomorphisms on R , not only homeomorphisms. A pseudogroup Γ of local diffeomorphisms is Lie if its elements ψ form the general solution of a set of PDE’s in the jet of ψ of some finite order. It is clear from the definition that if Γα is a collection of pseudogroups defined on n \ n R then Γα is also a pseudogroup on R . It therefore make sense to speak of the α n pseudogroup generated by a family of diffeomorphisms defined locally on R . Let L = { Xr }r∈R be a collection of differentiable vector fields each defined on n some Ur ⊂ R . We then get a collection of diffeomorphisms defined as follows: each x ∈ Ur has a neighborhood over which we can integrate Xr for sufficiently small t to obtain a (local) one parameter group ψt. The set of all such ψt is a family of diffeomorphisms and thus generate a pseudogroup. We shall called it the pseudogroup generated by L and denote by Γ(L). On the other hand, the vector fields whose local flows are in a pseudogroup Γ (for small time) form a Lie algebra L(Γ). If dim L(Γ) < ∞ then Γ comes from an action of a Lie group. 16 A pseudogroup Γ is called transitive if for every pair of points x and y in S there exists a ψ ∈ Γ with ψ(x) = y. It is called locally transitive if every point x in S has a neighborhood U such that for any point y in U there is a ψ ∈ Γ with ψ(x) = y. A pseudogroup Γ is called primitive if it leaves no nontrivial foliation invariant. Note that if Γ is primitive then it is automatically transitive. E. Cartan [2] has classified the complex primitive infinite dimensional pseudogroups, finding 6 classes. Gaps in his proof rectified by Singer-Sternberg [29]; complete proof by Guillemin-Quillen-Sternberg [14] in the complex case and Shnider [28] in the real case. Here is the list for the real case: n • The pseudogroup of local diffeormorphism of R . n • The pseudogroup of local diffeormorphism preserving a volume form of R . 2n • The pseudogroup of local diffeormorphism preserving a symplectic form of R . 2n+1 • The pseudogroup of local diffeormorphism preserving a contact form of R . n • The pseudogroup of local diffeormorphism preserving the volume form of R up to a constant. 2n • The pseudogroup of local diffeormorphism preserving a symplectic form of R up to a constant. Finally, let us recall the Generalized Fa´adi Bruno’s formula [9, 17, 22]. n Let α = (α1, α2, . . . , αn) ∈ N , where N is the set of all nonnegative integers. For n every x = (x1, x2, . . . , xn) ∈ R , we set n n N α Y αi X Y x = xi , |α| = αi , α! = αi! . i=1 i=1 i=1 17 n m Let us consider two open subsets U ⊂ R , V ⊂ R , and two differentiable maps of class greater or equal than r, f g U −→ V −→ R . If (x1, . . . , xn), (y1, . . . , ym) are the standard coordinates in U, V , respectively, then we set fi = yi ◦ f and y0 = f(x0), and x0 ∈ U. Theorem 2.1.8. With the previous notation and assumptions, the following formula holds: e |α| r |σ| m |αi| ! iαi ∂ (g ◦ f) X ∂ (g) X Y Y 1 1 ∂ (fi) (x ) = α! (y ) (x ) , α 0 σ 0 i αi 0 ∂x ∂y eiαi ! α ! ∂x |σ|=1 Eσ i=1 Aα where r = |α|, ( r ) i X Eσ = (e1α1 , . . . , emαm ): eiαi ∈ N, 1 ≤ |α | ≤ r, eiαi = σi, 1 ≤ i ≤ m , |αi|=1 and ( m r ) 1 m i X X i Aα = (α , . . . , α ) : 1 ≤ |α | ≤ r, 1 ≤ i ≤ m, eiαi · α = α . i=1 |αi|=1 18 CHAPTER 3 CONSTRUCTION VIA HOPF ACTIONS The Hopf algebra HH associated to the group G of globally defined diffeomor- 2 2 phisms ϕ : R → R of the form 1 2 ϕ(x1, x2) = (ϕ (x1, x2), ϕ (x2)) is defined as an algebra of linear operators acting over an algebra AH . The coalgebra structure and antipode of HH are obtained by using the properties of these operators. 3.1 The Hopf algebra HH . 2 2 2 Let F R be the frame bundle on R , which we identify to R × GL2(R) in the 2 2 usual way: the first jet at 0 ∈ R of a germ of a local diffeomorphism φ on R viewed as the pair 2 x := φ(0), y := (φ)∗,0 ∈ R × GL2(R). 2 i The flat connection on F R is given by the matrix-valued 1-form ω = (ωj) where, with the usual summation convention, i −1 i −1 i µ ωj := (y dy)j = (y )µdyj , i, j = 1, 2 , and the canonical form is the vector-valued 1-form θ = (θk) where k −1 k −1 k µ µ θ := (y dx) = (y )µd , k = 1, 2 , where d := dxµ . 19 j Let Yi be the fundamental vertical vector fields associated to the standard basis of j gl2(R) formed by the elementary matrices Ei . Thus we have j µ j j ∂ Yi = yi ∂µ , i, j = 1, 2 , where ∂µ := µ . ∂yj Let Xk be the basic horizontal vector fields. They are given by µ ∂ Xk = yk ∂µ , k = 1, 2 , where ∂µ := . ∂xµ 2 2 Besides the usual tangent space TxR for x ∈ R , it will be convenient to introduce 2 the graded version gr TxR given by the following gradation 2 2 gr−1 TxR := spanh ∂1 i gr−2 TxR := spanh ∂2 i . 2 2 2 2 If φ ∈ G, then its tangent map at x ∈ R ,(φ)∗,x : TxR → Tφ(x)R , leaves gr−1 TxR 2 2 invariant, and hence induces a graded tangent map gr(φ)∗,x : gr TxR → gr Tφ(x)R . In fact, we have that 1 1 1 ∂1 φ ∂2 φ ∂1 φ 0 (φ)∗,x = ∈ P and gr(φ)∗,x = ∈ D 2 2 0 ∂2 φ 0 ∂2 φ where P and D are the subgroup of upper triangular matrices and the subgroup of diagonal matrices of GL2(R), respectively. To simplify notation, we define 2 Tx if H = P. H 2 R Tx R := 2 gr TxR if H = D. (φ)∗,x if H = P. H (φ)∗,x := gr(φ)∗,x if H = D. Throughout the rest of this dissertation H will be either the subgroup P or D. 2 2 Definition 3.1.1. Let FH R be the sub-bundle consisting of the “H-frames” on R . 2 It consists of “first H-jet” at 0 ∈ R of the germs of local diffeomorphism φ ∈ G H 2 x = φ(0), y := (φ)∗,0 ∈ R × H. We denote this identification by (x, y) 'H φ. 20 2 2 The flat connection on F R restricts to a connection form on FH R with values 2 in the lie algebra h ⊂ gl2(R ) of H, ω = y−1dy ∈ h, y ∈ H, and the canonical form restricts to a vector-valued 1-form: θ := (y−1dx) y ∈ H. 2 The fundamental vertical vector fields are restrictions of those on F R : j µ j Yi = yi ∂µ , y ∈ H, and the basic horizontal vector fields are: µ Xk = yk ∂µ , y ∈ H. j j Note that when Yi are assembled into a matrix-valued vector field, Y = (Yi ) takes j i value in the Lie subalgebra h. Therefore, the indexes of Yi (ωj) must be chosen depending of the subgroup H to be non-zero vector fields (non-zero 1-forms). 2 2 Let GH := R o H denote the group of “affine H-motions” of R , which are a subgroup of G. Note that 2 ∼ FH R = GH via (x, y) 'H Lx ◦ y . In this section, we will use the notation (x, y) for elements in GH . Proposition 3.1.2. The vector fields j { Yi ,Xk | { (i, j) } ∈ S and k = 1, 2 } where S is a set of indexes chosen depending of the subgroup H, form a basis of left- 2 invariant vector fields on the group GH associated to the standard basis of gh = R oh. 21 Proof. Represent GH as the subgroup of GL3(R) consisting of the matrices y x a = 0 1 2 j with y ∈ H and x ∈ R . Let { Ei , ek | { i, j } ∈ IH and k = 1, 2 } be the standard 2 j basis of the Lie algebra gh = R oh and denote by { Eei , ek | { i, j } ∈ IH and k = 1, 2 } j the corresponding left-invariant vector fields. By definition, at the point a, Eei is tangent to the curve tEj tEj y x e i 0 ye i 0 t → = 0 1 0 1 0 1 j µ j and therefore coincides with Yi = yi ∂µ with y ∈ H, while the ek is tangent to the curve y x 1 te y tye + x k k t → = 0 1 0 1 0 1 µ which is precisely Xk = yk ∂µ with y ∈ H. The group G acts on GH by the following formula H ϕe(x, y) := ( ϕ(x), (ϕ)∗,x · y ) , for ϕ ∈ G which is derived by computing the “first H-jet” of the composition of ϕ ◦ φ where (x, y) 'H φ , i.e. , ϕe(x, y) 'H ϕ ◦ φ . Definition 3.1.3. Viewing here G as a discrete group, we form the crossed product algebra ∞ AH := Cc (GH ) o G. As a vector space is spanned by monomials of the form ∗ ∞ ∗ ∗ fUϕ, where f ∈ Cc (GH ) and Uϕ stands for ( ϕe ) , 22 while the product is given by the multiplicative rule f U ∗ · f U ∗ := f (f ◦ ϕ )U ∗ . 1 ϕ1 2 ϕ2 1 2 e1 ϕ2ϕ1 ∗ We will also use Uϕ := Uϕ−1 to simplify notation afterwards. Alternatively, AH can be regarded as the subalgebra of the endomorphism algebra ∞ End(Cc (GH )), generated by the multiplication and the translation operators ∞ Mf (ξ) = fξ, f, ξ ∈ Cc (GH ), ∗ ∗ ∞ Uϕ(ξ) = ( ϕe ) ξ = ξ ◦ ϕ,e ϕ ∈ G, ξ ∈ Cc (GH ). We proceed to associate to the pseudogroup G a Hopf algebra HH , realized via its Hopf action on the crossed product algebra AH following the construction given by Moscovici and Rangipour [25]. One starts with a fixed basis { Zr }r∈R for the Lie algebra gH of GH . Each Z ∈ gH is a left invariant vector field Z on GH , which is then extended to a linear operator on AH , ∗ ∗ ∞ Z(fUϕ) = Z(f)Uϕ, f ∈ Cc (GH ), ϕ ∈ G. One has ∗ p Uϕ Zq Uϕ = Γ q(ϕ) Zp , p, q ∈ R. p ∞ p with Γ q(ϕ) ∈ C (GH ). The matrix of functions Γ(ϕ) = Γ q(ϕ) automatically p,q∈R satisfies the cocycle identity Γ(ϕ1 ◦ ϕ2) = Γ(ϕ1) ◦ ϕe2 · Γ(ϕ2) , ϕ1, ϕ2 ∈ G. p We next denote by Ξ q the following multiplication operator on AH p ∗ p ∗ Ξ q(fUϕ) = Γq(ϕ)fUϕ , p, q ∈ R. p p −1 where Γq(ϕ) := Γ(ϕ) . q 23 Definition 3.1.4. Let HH be the unital subalgebra of linear operators on AH gener- p ated by the operators Zr’s and Ξ q’s with p, q, r ∈ R. In particular, HH contains all iterated commutators p p Ξ q;r1···rs = [ Zrs , ··· [ Zr1 , Ξ q ] ··· ], which act as multiplication operators by the function p p −1 Γq;r ...r (ϕ) := Zrs ··· Zr1 Γ(ϕ) ϕ ∈ G. 1 s q To understand better the algebra HH , i.e. relations between elements, and to provide it with a Hopf algebra structure, first we are going to give a precise expression for • the operators Ξ •;•···•. Note that for Z ∈ hH −1 ∗ ( ϕg )∗(Z) = Uϕ ZUϕ , this implies that the matrix function Γ(ϕ) expressed in terms of a basis of gH is equal −1 to the matrix function ( ϕg )∗ expressed in the same basis. Since the definition of • −1 Ξ •;•···• depends on Γ(ϕ) we need to find ( ϕe )∗ which is the transpose of the matrix ∗ ∗ function ( ϕe ) expressed in terms of a basis of gH . j Let us consider the basis { Z j = Y ,Zk = Xk } given by Proposition 3.1.2 as well (i ) i i k j as the canonical and connection forms { ωj , θ }. At each point of GH , { Yi ,Xk } i k and { ωj , θ } form basis of the tangent and cotangent space, respectively, dual to each other: j r j r r h Yi , ωs i = δs δi , h Xk , ωs i = 0, j t t t h Yi , θ i = 0, h Xk , θ i = δk. 24 H If ( xe , ye ) = ( ϕ(x), (ϕ)∗,x · y ), then −1 −1 H H ye d ye = (ϕ)∗,ϕ(x) · y d (ϕ)∗,x · y −1 −1 H H H = y · (ϕ )∗,ϕ(x) d(ϕ)∗,x · y + (ϕ)∗,x · dy −1 −1 H H k −1 = y · (ϕ )∗,ϕ(x) · Xk((ϕ)∗,x) · y θ + y d y , and −1 −1 H ye d xe = (ϕ)∗,x · y d ϕ(x) −1 −1 H = y · (ϕ )∗,ϕ(x) · (ϕ)∗,x dx −1 −1 H −1 = y · (ϕ )∗,ϕ(x) · (ϕ)∗,x · y y dx . −1 −1 H H −1 −1 H Note that the matrices y ·(ϕ )∗,ϕ(x)·Xk((ϕ)∗,x)·y and y ·(ϕ )∗,ϕ(x)·(ϕ)∗,x·y are upper triangular matrix with the latest having ones on its diagonal, i.e., unipotent. i k This can be written more precisely in terms of the basis { ωj , θ } as ( ϕ )∗(ω1) = ω1 + γ1 (ϕ) θ1 + γ1 (ϕ) θ2 e 1 1 11 12 ∗ 1 1 1 1 1 2 ( ϕ ) (ω ) = ω + γ (ϕ) θ + γ (ϕ) θ e 2 2 21 22 ∗ 2 2 2 2 for H = P ( ϕe ) (ω2) = ω2 + γ22(ϕ) θ ∗ 1 1 ( ϕ ) (θ ) = θ e ∗ 2 2 ( ϕe ) (θ ) = θ and ∗ 1 1 1 1 1 2 ( ϕe ) (ω1) = ω1 + γ11(ϕ) θ + γ12(ϕ) θ ∗ 2 2 2 2 ( ϕe ) (ω2) = ω2 + γ22(ϕ) θ for H = D ( ϕ )∗(θ1) = θ1 + γ1(ϕ) θ2 e 2 ∗ 2 2 ( ϕe ) (θ ) = θ 25 with i i −1 −1 H H γjk(ϕ)(x, y) := y · (ϕ )∗,ϕ(x) · Xk((ϕ)∗,x) · y j i λ −1 −1 H ρ = y (ϕ )∗,ϕ(x) XjXk(ϕ ) , λ ρ 1 1 −1 −1 H γ2 (ϕ)(x, y) := y · (ϕ )∗,ϕ(x) · (ϕ)∗,x · y 2 1 1 −1 −1 H 1 = y (ϕ )∗,ϕ(x) X2(ϕ ) , 1 1 j i (i ) 1 1 where γjk(ϕ) := Γ k (ϕ) and γ2 (ϕ) := Γ2(ϕ). Remark 3.1.5. Therefore, we obtain j • (Definition 3.1.4) The algebra HH is generated by the operators Z j = Y , (i ) i Zk = Xk and i ∗ i ∗ ξjk(fUϕ) := γjk(ϕ)fUϕ , 1 ∗ 1 ∗ ξ2 (fUϕ) := γ2 (ϕ)fUϕ , j i (i ) 1 1 where ξjk := Ξ k and ξ2 := Ξ2. • Note that for the other combination of valid indexes (depending on H) we have j (i ) j i k k k Ξ (s) = δsδr Id , Ξ j = 0 Id , Ξ l = δl Id , r (i ) where Id is the identity operator. i i i i • Furthermore, we have that ξjk = ξkj because γjk(ϕ) = γjk(ϕ). j The change of notation was made to simplify the indexes of the form (i ) and because the order of the indexes j, k is immaterial. With the notation of Definition 3.1.3 we have the following “Leibniz rules” which are important to prove relations between the iterated commutators. 26 Lemma 3.1.6. For any a, b ∈ A, one has p Zq(ab) = Zq(a)b + Ξ q(a)Zp(b) , p p µ Ξ q(ab) = Ξ µ(a)Ξ q (b) . ∗ ∗ Proof. Let a = f1Uϕ1 and b = f2Uϕ2 , one has Z (ab) = Z (f U ∗ f U ∗ ) = Z f (f ◦ ϕ ) U ∗ q q 1 ϕ1 2 ϕ2 q 1 2 e1 ϕ2◦ϕ1 = Z (f )(f ◦ ϕ )U ∗ + f Z (f ◦ ϕ )U ∗ q 1 2 e1 ϕ2◦ϕ1 1 q 2 e1 ϕ2◦ϕ1 = Z (f )U ∗ f U ∗ + f U ∗ Z (f ◦ ϕ ) ◦ ϕg−1 U ∗ q 1 ϕ1 2 ϕ2 1 ϕ1 q 2 e1 1 ϕ2 ∗ ∗ ∗ ∗ ∗ = Zq(f1)Uϕ1 f2Uϕ2 + f1Uϕ1 Uϕ1 ZqUϕ1 (f2) Uϕ2 ∗ ∗ ∗ p −1 ∗ = Zq(f1)Uϕ1 f2Uϕ2 + f1Uϕ1 Γq(ϕ1 )Zp(f2) Uϕ2 = Z (f )U ∗ f U ∗ + f Γp(ϕ−1) ◦ ϕ U ∗ Z (f )U ∗ q 1 ϕ1 2 ϕ2 1 q 1 e1 ϕ1 p 2 ϕ2 p ∗ ∗ −1 ∗ ∗ = Zq(f1)Uϕ f2Uϕ + f1 Γ(ϕ1) Uϕ Zp(f2)Uϕ 1 2 q 1 2 p = Zq(a)b + Ξ q(a)Zp(b) . p by using the definition of Γq(ϕ) in line 5 and the cocycle identity in line 7. We also have p p p ∗ ∗ −1 ∗ Ξ q(ab) = Ξ q(f1Uϕ f2Uϕ ) = Γ(ϕ2 ◦ ϕ1) f1(f2 ◦ ϕ1)Uϕ ◦ϕ 1 2 q e 2 1 p −1 ∗ = Γ(ϕ2) ◦ ϕ1 · Γ(ϕ1) f1(f2 ◦ ϕ1)Uϕ ◦ϕ e q e 2 1 p −1 ∗ = Γ(ϕ1) · Γ(ϕ2) ◦ ϕ1 f1(f2 ◦ ϕ1)Uϕ ◦ϕ e q e 2 1 p µ −1 ∗ = Γ(ϕ1) f1 Γ(ϕ2) ◦ ϕ1 (f2 ◦ ϕ1)Uϕ ◦ϕ µ e q e 2 1 p µ −1 ∗ ∗ = Γ(ϕ1) f1Uϕ Γ(ϕ2) f2Uϕ µ 1 q 2 p µ = Ξ µ(a)Ξ q (b) . where we are using the cocycle identity in line 2. 27 • Proposition 3.1.7. The operators Ξ •;• satisfy the structure constants identities: p p p λ µ ν p Ξ q;r − Ξ r;q = cλµ Ξ q Ξ r − cqr Ξ ν , ν where cqr are the structure constants of g, ν [ Zq,Zr ] = cqr Zν . Proof. Applying the first “Leibniz rule” from Lemma 3.1.6, one has for any a, b ∈ AH , µ ZqZr(ab) = Zq Zr(a)b + Ξ r (a)Zµ(b) λ µ λ µ = Zq(Zr(a))b + Ξ q (Zr(a))Zλ(b) + Zq(Ξ r (a))Zµ(b) + Ξ q (Ξ r (a))Zλ(Zµ(b)) , and thus the commutator can be expressed as follows λ µ [ Zq,Zr ](ab) = [ Zq,Zr ](a)b + Ξ q (Zr(a))Zλ(b) − Ξ r (Zq(a))Zµ(b) µ λ + Zq(Ξ r (a))Zµ(b) − Zr(Ξ q (a))Zλ(b) λ µ µ λ + Ξ q (Ξ r (a))Zλ(Zµ(b)) − Ξ r (Ξ q (a))Zµ(Zλ(b)) p p λ µ p = [ Zq,Zr ](a)b − Ξ q,r(a) − Ξ r,q(a) Zp(b) + (Ξ q Ξ r )(a)cλµZp(b) . ν On the other hand, by using [ Zq,Zr ] = cqr Zν, the left hand side equals, ν ν ν p ν p cqr Zν = cqr Zν(a)b + cqr Ξ ν(a)Zp(b) = [ Zq,Zr ](a)b + cqr Ξ ν(a)Zp(b) . Equating the two expressions one obtains after cancelation p p p λ µ ν p − Ξ q,r(a) − Ξ r,q(a) Zp(b) + cλµ(Ξ q Ξ r )(a)Zp(b) = cqr Ξ ν(a)Zp(b) . Since a, b ∈ AH are arbitrary and the Zp’s are linearly independent, this gives the claimed identity. j Now, using the basis { Z j = Y ,Zk = Xk } and the notation from Remark 3.1.5, (i ) i we rewrite the previous Lemma and Proposition. 28 (Lemma 3.1.6) For any a, b ∈ AH we have the following expressions: Y 1(ab) = Y 1(a)b + aY 1(b) 1 1 1 Y 2(ab) = Y 2(a)b + aY 2(b) 1 1 1 2 2 2 Y2 (ab) = Y2 (a)b + aY2 (b) 1 2 1 1 X1(ab) = X1(a)b + aX1(b) + ξ (a)Y (b) + ξ (a)Y (b) 21 1 11 1 2 2 1 2 1 1 X2(ab) = X2(a)b + aX2(b) + ξ22(a)Y2 (b) + ξ22(a)Y1 (b) + ξ12(a)Y1 (b) for H = P 1 1 1 ξ11(ab) = ξ11(a)b + aξ11(b) ξ1 (ab) = ξ1 (a)b + aξ1 (b) 12 12 12 ξ1 (ab) = ξ1 (a)b + aξ1 (b) 21 21 21 1 1 1 ξ22(ab) = ξ22(a)b + aξ22(b) 2 2 2 ξ22(ab) = ξ22(a)b + aξ22(b) Y 1(ab) = Y 1(a)b + aY 1(b) 1 1 1 2 2 2 Y2 (ab) = Y2 (a)b + aY2 (b) 1 1 X1(ab) = X1(a)b + aX1(b) + ξ (a)Y (b) 11 1 1 2 2 1 1 X2(ab) = X2(a)b + aX2(b) + ξ2 (a)X1(b) + ξ22(a)Y2 (b) + ξ12(a)Y1 (b) for H = D 1 1 1 ξ11(ab) = ξ11(a)b + aξ11(b) ξ1 (ab) = ξ1 (a)b + aξ1 (b) + ξ1 (a)ξ1(b) 12 12 12 11 2 ξ2 (ab) = ξ2 (a)b + aξ2 (b) 22 22 22 1 1 1 ξ2 (ab) = ξ2 (a)b + aξ2 (b) 29 j (Proposition 3.1.7) The basis { Z j = Y ,Zk = Xk } satisfy the following com- (i ) i mutation relations j s j s s j j j [Yi ,Yr ] = δrYi − δi Yr , [Yi ,Xk] = δk Xi , [Xk,Xl] = 0 , where we are using just valid indexes depending on H. Therefore, the structure constant identities are given explicitly by µ w q Case 1: Top index (λ) and bottom indexes (v ) and (p). µ µ µ µ (λ) (λ) (λ) r s l (λ) Ξ w q − Ξ q w = crs Ξ w Ξ q − c w q Ξ , (v );(p) (p);(v ) (v ) (p) (v )(p) l reduce to µ µ (λ) (λ) 0 = c w q − c w q . (v )(p) (v )(p) µ q Case 2: Top index (λ) and bottom indexes v and (p). µ µ µ µ (λ) (λ) (λ) r s l (λ) Ξ q − Ξ q = crs Ξ Ξ q − c q Ξ , v;(p) (p);v v (p) v(p) l reduce to q λ q λ λ q q λ [Yp , ξµv] = δµ ξpv − δp ξµv + δv ξµp . µ Case 3: Top index (λ) and bottom indexes v and p. µ µ µ µ (λ) (λ) (λ) r s l (λ) Ξ v;p − Ξ p;v = crs Ξ v Ξ p − cvp Ξ l , reduce to λ λ s λ s λ [Xp, ξµv] − [Xv, ξµp] = ξµp ξsv − ξµv ξsp . w q Case 4: Top index λ and bottom indexes (v ) and (p). λ λ λ r s l λ Ξ w q − Ξ q w = c Ξ w Ξ q − c w q Ξ , (v );(p) (p);(v ) rs (v ) (p) (v )(p) l reduce to 0 = 0 30 q Case 5: Top index λ and bottom indexes v and (p). λ λ λ r s l λ Ξ q − Ξ q = c Ξ Ξ q − c q Ξ , v;(p) (p);v rs v (p) v(p) l reduce to 0 = 0 for H = P, q 1 q p 1 q p 1 [Yp , ξ2 ] = −δ1 δ1 ξ2 + δ2 δ2 ξ2 for H = D. Case 6: Top index λ and bottom indexes v and p. λ λ λ r s l λ Ξ v;p − Ξ p;v = crs Ξ v Ξ p − cvp Ξ l , reduce to 0 = 0 for H = P, 1 1 1 1 [X1, ξ2 ] = ξ12 − ξ2 ξ11 and 0 = 0 for H = D. Remark 3.1.8. The previous non-trivial relations give us the following information: q µ 1 • Since the commutators of Yp with ξλv and ξ2 can be written as a linear combi- nation of generators (Case 2 and 5) and because of the Jacobi identity is true for the commutator of operators, the only possible new operators on HH are (j ) ξi := Ξ i = [X ,... [X , ξi ] ... ] , jkr1···rs k;r1···rs rs r1 jk 1 1 1 ξ2r1···rs := Ξ 2;r1···rs = [Xrs ,... [Xr1 , ξ2 ] ... ] , which are the multiplication operators by the function i i γjkr1··· ,rs := Xrs ...Xr1 (γjk) , 1 1 γ2r1··· ,rs := Xrs ...Xr1 (γ2 ) . 1 Note that γ2•···• is defined in a different way depending on the subgroup H. 31 • The order of the last s-indexes is immaterial for these operators because the Xk’s commute and the Jacobi identity. λ λ 1 • The operators ξµvp − ξµpv and ξ21 can be written as a sum of products of gener- ators (Case 3 and 6). In view of Remark 3.1.5 and 3.1.8, the algebra HH admits a basis as a vector space (similar to the Poincar´e-Birkhoff-Witt basis of a universal enveloping algebra). The notation needed to specify such a basis involves two kind of multi-indices. The first kind are of the form ( ) j1 jq I = k1 ≤ · · · ≤ kp; ≤ · · · ≤ , i1 iq and the second kind are of the form ( ) it K = { κ1 ≤ · · · ≤ κs } , where κt = . t t jt ≤ kt ≤ r1 ≤ · · · ≤ rpt In both cases, the indexes jλ and iλ depend on the group H and the inner multi- indices are ordered lexicographically top to bottom. We then denote Z := X ··· X Y j1 ··· Y jq , I k1 kp i1 iq i1 is ξ := ξ ··· ξ s s . K j k r1···r1 jsksr ···r 1 1 1 p1 1 ps Proposition 3.1.9. The monomials ξK ZI form a linear basis of HH as a vector space. Before we go to the proof of this Proposition, we are going to provide some definitions and lemmas which will be useful not just for this proof but also for next chapter. H Definition 3.1.10. Let NH := {ψ ∈ G | ψ(0) = 0 and (ψ)∗,0 = Id} and let p p αq ···q (ψ) := ∂xq ··· ∂xq (ψ ) for ψ ∈ NH , 1 s s 1 x=0 32 • where the indexes are taken such that α•···• are non-trivial functions on NH . The algebra of functions on NH generated by these functions will be denoted by FH . Fur- p thermore, we say that the length of αq1···qs is s. Note that these functions are multiples of the non-trivial coefficients of the Taylor expansion of ψ at 0 and they are symmetric in the lower indexes but otherwise arbitrary. Lemma 3.1.11. The monomials αK form a basis for FH as a vector space. Proof. Suppose that we have cK ∈ C such that X cK αK (ψ) = 0 , for all ψ ∈ NH , K • we need to prove that cK = 0 for any K. Let l be the largest length of the α•···• on all αK ’s. For this l consider all ψ ∈ NH such that the taylor coefficients with size greater than l are zero, and let α p := αp (ψ) ∈ , e q1···qs q1···qs C for all s ≤ l. For example, if l = 2 and H = D then 1 1 1 1 1 ψ (x1, x2) = x1 + αe 2 x2 + αe 11 x1x1 + αe 12 x1x2 + αe 22 x2x2 2 2 ψ (x1, x2) = x2 + αe 22 x2x2 , By evaluating at this ψ, one obtains X cK αeK = 0 K where the left hand side is a polynomial with variables α p . This implies that e q1···qs cK = 0 for any K, because it is the only way to obtain the zero polynomial. 33 Definition 3.1.12. For any ψ ∈ NH , define the functions p p ηq1···qs (ψ) := γq1q2q3···qs (ψ)(0, Id) , • where the indexes are taken such that η•···• are non-trivial functions on NH . Lemma 3.1.13. The monomials ηK form a basis for FH as a vector space. i 1 Proof. First, by evaluating the explicit formulas for γjk(ψ) and γ2 (ψ) for ψ ∈ NH at the point (0, Id) ∈ GH , we obtain i i −1 H ρ γjk(ψ)(0, Id) = (ψ )∗,ψ(x) ∂xj ∂xk ψ , ρ x=0 1 1 −1 H 1 γ2 (ψ)(0, Id) = (ψ )∗,ψ(x) ∂x2 ψ . 1 x=0 H −1 H Furthermore, since ψ(0) = 0 and (ψ)∗,0 = Id we get (ψ )∗,0 = Id, thus i i i ηjk(ψ) = γjk(ψ)(0, Id) = αjk(ψ) 1 1 1 η2(ψ) = γ2 (ψ)(0, Id) = α2(ψ) where the indexes are taking depending on the subgroup H. Next, one has h i i i −1 H ρ γjkr ···r (ψ)(0, Id) = ∂xr ··· ∂xr (ψ )∗,ψ(x) ∂xj ∂xk ψ 1 s s 1 ρ x=0 h 1 i 1 −1 H 1 γ2r ···r (ψ)(0, Id) = ∂xr ··· ∂xr (ψ )∗,ψ(x) ∂x2 ψ 1 s s 1 1 x=0 34 2 note that γ2···2 is different depending on H. The product rule implies h i i −1 H ρ ∂xr ··· ∂xr (ψ )∗,ψ(x) ∂xj ∂xk ψ s 1 ρ x=0 i −1 H ρ = ∂xr ··· ∂xr (ψ )∗,ψ(x) ∂xj ∂xk ψ s 1 ρ x=0 x=0 + ...... i −1 H ρ + (ψ )∗,ψ(x) ∂xr ··· ∂xr ∂xj ∂xk ψ ρ x=0 s 1 x=0 h 1 i −1 H 1 ∂xr ··· ∂xr (ψ )∗,ψ(x) ∂x2 ψ s 1 1 x=0 1 −1 H 1 = ∂xr ··· ∂xr (ψ )∗,ψ(x) ∂x2 ψ s 1 1 x=0 x=0 + ...... 1 −1 H 1 + (ψ )∗,ψ(x) ∂xr ··· ∂xr ∂x2 ψ 1 x=0 s 1 x=0 i −1 H where the summands at the middle are product of higher derivatives of (ψ )∗,ψ(x) ρ ρ 1 and higher derivatives of ∂xj ∂xk ψ or ∂x2 ψ all of them evaluated at x = 0 and with the number of derivatives taken less or equal than s. −1 H H Furthermore, by differentiating Id = (ψ)∗,x (ψ)∗,x and using induction, the −1 −1 H H higher derivatives of (ψ )∗,ψ(x) = (ψ)∗,x can be expressed as product of higher H derivatives of (ψ)∗,x . For example: −1 h i h H H i 0 = ∂xr Id = ∂xr (ψ)∗,x (ψ)∗,x −1 −1 h H i H H h H i = ∂xr (ψ)∗,x (ψ)∗,x + (ψ)∗,x ∂xr (ψ)∗,x . H After evaluating at x = 0 and using that (ψ)∗,0 = Id, we obtain −1 h H i h H i ∂xr (ψ)∗,0 = −∂xr (ψ)∗,0 , which implies that i i i ν ηjkr(ψ) = αjkr(ψ) − αrν(ψ)αjk(ψ) 35 In the general case, we have proved that p p p λ λ ηq1···qs = αq1···qs + Pq1···qs (αµ, ··· , αµ1···µs−1 ) p where Pq1···qs is a polynomial. This implies that X ηK = αK + cK0 αK0 K0 0 λ where K are made of αµ1···µs of length strictly less than the ones on K. λ Therefore, the transformations sending αK to ηK , with αµ1···µs of length less or equal than a fixed s, will be invertible linear maps sending basis to basis on these finite subspaces of FH . This proves that the monomials ηK are also a basis for FH because we are always using finite linear combination and we can pick a finite subspace with λ a fixed maximum length for αµ1···µs to check the independence of them. Next, we prove Proposition 3.1.9: Proof. We need to prove that if cI,K ∈ C are such that X cI,K ξK ZI (a) = 0 , for all a ∈ AH , I,K then cI,K = 0 for any (I,K). To this end, we evaluate the previous expression on all elements of the form ∗ a = fUϕ ∈ AH at the point (x, y) ∈ GH . In particular, for any fixed but arbitrary ϕ ∈ G, one obtains: ! X X ∞ cI,K γK (ϕ)(x, y) (ZI f)(x, y) = 0 , for all f ∈ Cc (GH ) , I K Since the ZI ’s form a PBW basis for U(gh), which can be viewed as the algebra of left invariant differential operators on GH , the validity of the previous equality for 36 ∞ any f ∈ Cc (GH ) implies the vanishing for each I of the corresponding coefficient. One therefore obtains, for any fixed I, X cI,K γK (ϕ)(x, y) = 0 , for all ϕ ∈ G . K By using ψ ∈ NH and the identity element (0, Id) ∈ GH , we get X cI,K ηK (ψ) = 0 , K which by Lemma 3.1.13 implies cI,K = 0. 1 p ⊗p We can associate to any element h ⊗ · · · ⊗ h ∈ HH a multi-differential operator, ⊗p acting on AH , by the following formula T (h1 ⊗ · · · ⊗ hp)(a1 ⊗ · · · ⊗ ap) := h1(a1) ··· hp(ap), 1 p 1 p where h , . . . , h ∈ HH and a , . . . , a ∈ AH , the linear extension of this assignment will be denoted by the same letter. Proposition 3.1.14. For each p ∈ N, the linear transformation ⊗p ⊗p T : HH → L(AH , AH ) is injective. Proof. For p = 1, T gives the standard action of HH on AH , which was just shown to be faithful. To prove that Ker T = 0 for an arbitrary p ∈ N, assume that X 1 p H = hρ ⊗ · · · ⊗ hρ ∈ Ker T ρ j By using the basis as above, we may uniquely express each hρ in the form j X hρ = Cρ,Ij ,Kj ξKj ZIj . Ij ,Kj 37 ∗ ∗ Evaluating T (H) on elementary tensors of the form f1Uϕ1 ⊗ · · · ⊗ fpUϕp , one obtains X ∗ ∗ Cρ,I1,K1 ··· Cρ,Ip,Kp ξK1 (ZI1 (f1)Uϕ1 ) ··· ξKp (ZIp (fp)Uϕp ) = 0 . ρ,I,K Evaluating further at a point u1 = (x, y) ∈ GH , and denoting u2 := ϕe(u1) , . . . , up := ϕep−1(up−1) the above identity gives X Cρ,I1,K1 ··· Cρ,Ip,Kp · γK1 (ϕ1)(u1) ··· γKp (ϕp)(up) ρ,I,K · ZI1 (f1)(u1) ··· ZIp (fp)(up) = 0 . Since ϕj are arbitrary then the points uj can be chosen arbitrarily, and using again the fact that ZI form a PBW basis for U(gh), we can infer that for each p-tuple of indices of the first kind (I1,...,Ip) one has X Cρ,I1,K1 ··· Cρ,Ip,Kp · γK1 (ϕ1)(u1) ··· γKp (ϕp)(up) = 0 . ρ,K Finally, by taking all points xj = (0, Id) and ϕk = ψ ∈ NH we have X Cρ,I1,K1 ··· Cρ,Ip,Kp · ηK1 (ψ) ··· ηKp (ψ) = 0 . ρ,K which implies that for any (K1,...,Kp) X Cρ,I1,K1 ··· Cρ,Ip,Kp = 0 . ρ This proves H = 0. ∞ Definition 3.1.15. The algebra AH = Cc (GH ) o G carries a canonical trace, uniquely determined up to a scaling factor. It is defined as the linear functional τ : AH → C, Z f ω, if ϕ = Id, ∗ G τ fUϕ = H 0, otherwise. 38 i k j Here ω is the volume form attached to the dual basis { ωj , θ } of { Yi ,Xk } on GH ^ i ^ k ω := ωj ∧ θ . (i,j) k=1,2 The tracial property τ(ab) = τ(ba), ∀ a, b ∈ AH . is a consequence of the G- invariance of the volume form ω. In turn, the later follows from the fact that ∗ i i i k ∗ k k X p q ( ϕe ) (ωj) = ωj + γj,k(ϕ)θ and ( ϕe ) (θ ) = θ + γq (ϕ) θ q>k indeed, ! ! ∗ ^ i i k ^ k X p q ( ϕe ) (ω) = ωj + γj,k(ϕ)θ ∧ θ + γq (ϕ) θ (i,j) k=1,2 q>k ^ i ^ k = ωj ∧ θ = ω . (i,j) k=1,2 Therefore, we have Z Z ∗ ∗ −1 ∗ τ f1Uϕ · f2Uϕ−1 = f1(f2 ◦ ϕg) ω = (f1 ◦ ϕe)f2 ( ϕe ) (ω) GH GH Z ∗ ∗ = f2(f1 ◦ ϕe) ω = τ f2Uϕ−1 · f1Uϕ , GH which proves the tracial property for the only non-trivial case. Lemma 3.1.16. The pairing (a, b) → τ(ab) is non-degenerate. Proof. Because of the definition and the tracial property of τ, it is enough to show that Z f1( f2 ◦ ϕe) ω = 0 for all f2 implies f1 = 0, GH which is the same that Z f1f2 ω = 0 for all f2, implies f1 = 0 . GH 39 Z If we assume f1f2 ω = 0 for all f2, but f1(g) 6= 0 at some g ∈ GH , without loss of GH generality suppose f1(g) > 0. Then there exits some small neighborhood U of g such that f1 is strictly positive, say f > c > 0 on U. We can choose a compact set V1 ⊂ U and a smaller compact set V2 $ V1. Take f2 to be a smooth cutoff function which is the identity on V2 and vanish on V1, then Z Z Z Z f1f2 = f1f2 ω > f1f2 $ > c $ > 0 , GH V1 V2 V2 hence a contradiction. This trace satisfies an invariant property relative to a character of HH which is defined as follows: Definition 3.1.17. The infinitesimal trace character δ(Z) = Tr(ad(Z)), Z ∈ gh, extends uniquely to a character δ : HH → C. It is defined on the algebra generators as follows: j j i 1 δ(Yi ) := δi , δ(Xk) = 0 , δ(ξjk) := 0 , δ(ξ2 ) := 0 , This definition is compatible with all the relations between the generators of HH and therefore extends to a character of the algebra HH . Proposition 3.1.18. The trace τ is Hn-invariant relative to the character δ, i.e. τ( h(a) ) = δ( h )τ( a ), a ∈ AH , h ∈ HH . Proof. It suffices to verify the stated identity when ϕ = Id on the generators of HH . 40 If h = Z ∈ gh, it is just the restatement of the invariance property of the left Haar measure on GH with respect to right translation at the level of the Lie algebra, i.e. Z ∗ ∗ τ Z(fUId) = τ Z(f)UId = Z(f) ω GH Z d = f((∗) exp(tZ)) ω dt GH t=0 Z d = f((∗) exp(tZ)) ω dt t=0 GH Z d = f det(Ad(exp(tZ))) ω dt t=0 GH Z d = det(Ad(exp(tZ))) f ω dt t=0 GH Z ∗ = Tr(ad(Z)) f ω = δ(Z)τ fUId . GH i On the other hand, if h = ξjk we have i ∗ i ∗ ∗ τ ξjk(fUId) = τ γjk(Id)fUId = τ 0 · fUId ∗ i ∗ = 0 = 0 · τ fUId = δ( ξjk) τ fUId . 1 Similarly if h = ξ2 when H = D. Proposition 3.1.19. There exist a unique anti-automorphism Sδ : HH → HH such that τ( h(a)b ) = τ( aSδ(h)(b)) , for any h ∈ HH and a, b ∈ AH . Moreover, Sδ is involutive: 2 Sδ = Id . Proof. The idea behind the proof of this Proposition is to use the different types of “Leibniz rule” (Lemma 3.1.6) and the invariance property (Proposition 3.1.18). i 1 First, let us consider the generators h = ξj,k or h = ξ2 which satisfy h(ab) = h(a)b + ah(b) 41 by using the invariance property, one obtains 0 = δ( h )τ( ab ) = τ( h(ab) ) = τ( h(a)b ) + τ( ah(b)) , which implies τ( h(a)b ) = −τ( ah(b)) . 1 Note that for H = D, the generator ξ12 satisfies 1 1 1 1 1 ξ12(ab) = ξ12(a)b + aξ12(b) + ξ11(a)ξ2 (b) , and for this case we get 1 1 0 = δ( ξ12 )τ( ab ) = τ( ξ12(ab)) 1 1 1 1 = τ( ξ12(a)b ) + τ( aξ12(b) ) + τ(ξ11(a)ξ2 (b)) 1 1 1 1 = τ( ξ12(a)b ) + τ( aξ12(b)) − τ(aξ11ξ2 (b)) 1 1 where the last equality is obtain by using that τ( ξ11(a)b ) = −τ( aξ11(b) ), then 1 1 1 1 τ( ξ12(a)b ) = −τ( aξ12(b) ) + τ(aξ11ξ2 (b)) . j Second, let us consider the generators Yi which satisfy j j j Yi (ab) = Yi (a)b + aYi (b) , for this generators we have j j j j j δi τ( ab ) = δ( Yi )τ( ab ) = τ( Yi (ab) ) = τ( Yi (a)b ) + τ( aYi (b)) , thus j j j τ( Yi (a)b ) = −τ( aYi (b) ) + δi τ( ab ) . Finally, for the operators Xk satisfying i j Xk(ab) = Xk(a)b + aXkb + ξjk(a)Yi (b) 42 i the invariance property and the identities for ξjk give 0 = δ( Xk )τ( ab ) = τ( Xk(ab)) i j = τ( Xk(a)b ) + τ( aXk(b) ) + τ(ξjk(a)Yi (b)) i j = τ( Xk(a)b ) + τ( aXk(b)) − τ(aξjkYi (b)) , which can be written as i j τ( Xk(a)b ) = −τ( aXk(b) ) + τ(aξj,kYi (b)) Note that for H = D, the generator X2 satisfies 1 2 2 1 1 X2(ab) = X2(a)b + aX2(b) + ξ2 (a)X1(b) + ξ22(a)Y2 (b) + ξ12(a)Y1 (b) , and by following the same procedure as before, for this case we have 0 = δ( X2 )τ( ab ) = τ( X2(ab)) 1 = τ( X2(a)b ) + τ( aX2(b) ) + τ(ξ2 (a)X1(b)) 2 2 1 1 + τ(ξ22(a)Y2 (b)) + τ(ξ12(a)Y1 (b)) 1 = τ( X2(a)b ) + τ( aX2(b)) − τ(aξ2 X1(b)) 2 2 1 1 1 1 1 − τ(aξ22Y2 (b)) − τ(aξ12Y1 (b)) + τ(aξ11ξ2 Y1 (b)) , therefore 1 τ( X2(a)b ) = − τ( aX2(b) ) + τ(aξ2 X1(b)) 2 2 1 1 1 1 1 + τ(aξ22Y2 (b)) + τ(aξ12Y1 (b)) − τ(aξ11ξ2 Y1 (b)) . Thus the generators of HH satisfies identities of the form τ( h(a)b ) = τ( aSδ(h)(b) ). By Lemma 3.1.16, this operators are uniquely determined. This rule, being multiplicative, extends from generators to all elements h ∈ HH , and uniquely defines a map Sδ : HH → HH satisfying the identity form our Proposi- tion. In turn, this identity implies that Sδ is anti-automorphism as well as the fact that Sδ is involutive. 43 Finally, we can equip HH with a canonical Hopf structure. Theorem 3.1.20. There exists a unique Hopf algebra structure on HH with respect to which AH is a left HH -module algebra. Proof. The “Leibniz rules” given by Lemma 3.1.6 extend multiplicatively to any element h ∈ HH h(ab) = h(1)(a)h(2)(b), h(1), h(2) ∈ HH , a, b ∈ AH , where we are using Sweedler notation. By Proposition 3.1.14, this property uniquely determines the coproduct map ∆ : HH → HH ⊗ HH , ∆(h) := h(1) ⊗ h(2), that satisfies T (∆h)(a ⊗ b) = h(ab). Furthermore, the coassociativity of ∆ becomes a consequence of the associativity of AH , because after applying T it amounts to the identity h((ab)c) = h(a(bc)), h ∈ HH , a, b, c ∈ AH . Similarly, the property that ∆ is an algebra homomorphism follows from the fact that AH is a left HH -module. By the very definition of the coproduct, AH is actually a left HH -module algebra (see Definition §4.1). The counit is defined by (h) = h(1) , when transported via T , its required properties amount to the obvious identities h(a1) = h(1a) = h(a), h ∈ HH , a ∈ AH . 44 It remains to show the existence of the antipode. We first check that the anti- automorphism Sδ is a twisted antipode, i.e., satisfies for any h ∈ HH , Sδ(h(1))h(2) = δ(h)1 = h(1)Sδ(h(2)) . Indeed, with a, b ∈ AH arbitrary, one has, τ aδ(h)b = δ(h)τ(ab) = τ h(ab) = τ h(1)(a)h(2)(b) = τ a Sδ(h(1))h(2) (b) which proves the left hand side equality. For the right hand side is checked case by case on generators, and hence for any h ∈ Hn. ˇ ∗ ∗ Now, let δ ∈ HH denote the convolution inverse of the character δ ∈ HH , which on generators is given by ˇ j j ˇ ˇ i ˇ 1 δ(Yi ) := −δi , δ(Xk) = 0 , δ(ξjk) := 0 , δ(ξ2 ) := 0 , ˇ Then S(h) := δH (h(1))Sδ(h(2)) is an algebra anti-homomorphism which satisfies the antipode requirement S(h(1))h(2) = (h)1 = h(1)S(h(2)) on the generators, and hence for any h ∈ Hn. 45 CHAPTER 4 BICROSSED PRODUCT REALIZATION. The Hopf algebra HH can be reconstructed as a bicrossed product Hopf algebra [24]. This structure arises naturally from the canonical splitting of the group G as a 2 set-theoretical product G = GH · NH of the group GH of “affine H-motions” of R and the group NH of elements of G with trivial “first H-jet”. 4.1 The bicrossed product Hopf algebra FH IC U(gh). We recall below the basic notions concerning the bicrossed product construction. Let U be a Hopf algebra. Suppose an algebra A is a left U-module with action . : U ⊗ A → A where . (u, a) = u . a, then A is a left H-module algebra if it satisfies the following conditions u . (ab) = (u(1) . a)(u(2) . b), u . (1A) = ε(u)1A. Then we can form a left cross product algebra A>CU built on the vector space A⊗U with the product structure (a>Cu)(b>Cv) = a(u(1) . b)>Cu(2)v, a, b ∈ A, u, v ∈ U 1A>C U = 1A ⊗ 1U . 46 Let F be a Hopf algebra. Suppose a coalgebra C is a right F-comodule with coaction H : C → C ⊗ F where H(c) = c<0> ⊗ c<1>, then C is a right F−comodule coalgebra if it satisfies the following conditions c<0>(1) ⊗ c<0>(2) ⊗ c<1> = c(1)<0> ⊗ c(2)<0> ⊗ c(1)<1>c(2)<1> ε(c<0>)c<1> = ε(c)1F . Then we can form a right cross coproduct coalgebra F I ∆(f I ε(f I Definition 4.1.1. Let F, U be Hopf algebras, let F be a left U-module algebra and U be a right F-comodule coalgebra such that the compatibility conditions ε(u . f) = ε(u)ε(f) , ∆(u . f) = u(1)<0> . f(1) ⊗ u(1)<1>(u(2) . f(2)) , H(1) = 1 ⊗ 1, H(uv) = u(1)<0>v<0> ⊗ u(1)<1>(u(2) . v<1>) , u(2)<0> ⊗ (u(1) . f)u(2)<1> = u(1)<0> ⊗ u(1)<1>(u(2) . f) , are satisfied for all u, v ∈ U and f ∈ F. Then F and U are said to form a matched pair, and the algebra F >C U and coalgebra F I