CLASSIFICATION OF ISOMORPHISM TYPES OF HOPF ALGEBRAS IN A CLASS OF ABELIAN 2 EXTENSIONS 1 0 2 LEONID KROP AND YEVGENIA KASHINA v o Abstract. We study Hopf algebras which are extensions of the N groupalgebrakCpofacyclicgroupCpofprimeorderpbytheHopf 3 algebradualkG ofthegroupalgebrakGofafiniteabelianp-group 2 G over an algebraically closed field of characteristic 0. The main resultsare: structuretheoremforthe2nd Hopfcohomologygroups A] H2(⊳), a criterionforisomorphismoftwoextensions, enumeration c of isomorphism classes by certain orbits in H2(⊳), and a complete Q c classification of commutative extensions for any odd p. . h t a m Keywords Hopf algebras, Abelian extensions, Crossed products, [ Cohomology Groups Mathematics Subject Classification (2000) 16W30 - 16G99 1 v 1 Introduction 2 6 The purpose of this article is to find a general criterion for isomor- 5 . phismoftwo Hopfalgebrasincertainclasses ofabelianextensions. The 1 classes in question consist of Hopf algebras which are extensions of the 1 2 group algebra kC by kG for a prime p, where C is a cyclic group of p p 1 prime order p, G a finite abelian p-group, and k an algebraically closed : v field of characteristic zero. The customary definition of equivalence i X of extensions [16] gives rise to the set Ext(kC ,kG) of all equivalence p r classes of extensions. a Every algebra from any of our classes is semisimple by [18, 7.4.2]. By a fundamental result of D. Stefan [21] the number of isomorphism types in any of our classes is finite. Our main results are similar to Ste- fan’s inasmuch as both involve an orbit space. In our case this space is constructed as follows. The group G is a right C -module under p the action denoted by ‘⊳’. The set Ext(kC ,kG) splits up into a dis- p joint union of groups Opext(kC ,kG,⊳) of classes of extensions with p the same C -action on G. Cohomology theory of abelian extensions p Date: 10/20/12. Research by the first author partially supported by a grant from the College of Liberal Arts and Sciences at DePaul University. 1 2 LEONID KROPANDYEVGENIAKASHINA of finite-dimensional Hopf algebras [16] supplies us with the abelian group H2 (kC ,kG,⊳) of Hopf 2-cocycles parametrizing algebras in Hf p Opext(kC ,kG,⊳). Two groups play a distinguished rˆole in the pa- p per. The first one A(⊳) = Aut (G) is the group of C -automorphisms Cp p of G. In the second place comes the group A = Aut(C ) of au- p p tomorphisms of C . A(⊳) acts on H2 (kC ,kG,⊳) turning it into an p Hf p A(⊳)-module, while A twists the action ‘⊳’ into p−1 actions ⊳α,α ∈ p A . Let [G,⊳] denote the class of C -modules isomorphic to [G,⊳α] p p for some α ∈ A . Furthermore, we let Ext (kC ,kG) stand for p [G,⊳] p the set of equivalence classes of extensions whose C -module G lies p in [G,⊳]. The main result of the paper is the statement that two Hopf algebras H,H′ ∈ Ext(kC ,kG) are isomorphic implies they lie p in Ext (kC ,kG) for some ‘⊳’, and moreover there exists a bijection [G,⊳] p between the orbits of A(⊳) in H2 (kC ,kG,⊳) generated by a single Hf p noncoboundary and the isomorphism classes of noncocommutative al- gebras in Ext (kC ,kG). Under a stronger hypothesis on G, namely [G,⊳] p G elementary abelian, the theorem is strengthened to a bijection be- tweenallorbitsgeneratedbyasingle2-cocycleandisomorphismclasses of algebras. Elements ofExt(kC ,kG)possess two special features. Every algebra p H there is equivalent to a smash product kG#kC , and kCp is central p in the dual Hopf algebra H∗. In consequence, it is only the coalge- bra structure of H that is deformed by a 2-cococycle. These types of abelian extensions have been studied by M. Mastnak [12]. We adopt his notation H2(kC ,kG,⊳) for the group of Hopf 2-cocycles. A major c p result of the paper is structure theorem for H2(kC ,kG,⊳). It states c p that if G is any finite abelian p-group with p > 2, or a finite elementary 2-group then there is a C -isomorphism p (0.1) H2(kC ,kG,⊳) ≃ H2(C ,G)×H2(G,k•) c p p N where G is the dual group of G, H2(C ,bG) is the second cohomology p group of C over G with respect to the action ⊳, and H2(G,k•) is p N the kernbel of the norm mapping in the Schur multiplier of G. The for- mula shows an interbesting interplay between Hopf algebras and groups. For, in the case ⊳ is trivial, we have by Proposition 5.1 that the sub- group of commutative extensions in Ext(kC ,kG) is identical with the p group of central extensions of G by C , so that H2(kC ,kG,triv) equals p c p H2(G,C ). Thus, (0.1) can be seen as a Hopf algebraic generalization p of a structure theorem for the 2nd cohomology group H2(G,A) of cen- tral extensions of a finite abelian group by a group A [24, Thm. 2.2]. ISOMORPHISM TYPES OF HOPF ALGEBRAS 3 Another notable result is the computation of the number of isomor- phism types of commutative algebras in Ext(kC ,kG) for G elementary p abelian and an odd p. We conclude the paper with an illustration of our technique in the case G = C ×C completely described by p p A. Masuoka [15]. The paper is organized in six sections. In section 1 we review the necessary facts of the theory of abelian extension with the emphasis on duality inherent in the theory and we derive the formula for comulti- plication in the dual of a crossed product from first principles. Each abelian extension is associated with a pair of groups (G,F). In section 2 we narrow our focus to the so-called cocentral extensions [8] with the additional property of the trivial 2-cohomology of F in kG. In sec- tion 3 our setting is even more restrictive. There we study extensions associated with pairs (G,F) where G is a p-group and F = C is a p cyclic group of order p and we calculate the second Hopf cohomology groups. Section 4 contains the main results of the paper. In Section 5 we classify commutative extensions and compute their number up to isomorphism. In the last section we rederive a key result of [15] using our technique. 1. Background Review 1.1. Extensions of Hopf Algebras. Let k be a ground field and H denote the category of Hopf algebras over k. For a Hopf algebra H we let ǫ and u denote the augmentation H → k and the unit k → H H H mappings, respectively. k itself can be viewed as a Hopf algebra with ∆(1 ) = 1 ⊗ 1 , S(1 ) = 1 and ǫ(1 ) = 1 . It is easy to check that k k k k k k k for every H ∈ H, Hom (H,k) = {ǫ } and Hom (k,H) = {u }. H H H H From the point of view of the category theory k is the zero object in H. Applying the categorical definition of kernel to H we say that a Hopf algebra (K,ι) is the kernel of a Hopf morphism φ : F → G if ι : K → F is a Hopf morphism and linear monomorphism fitting into the commutative diagram K −−ǫ−K→ k (1.1) ι uG φ F −−−→ G y y and universal among all diagrams 1.1. That is, for every (K′,ι′) sat- isfying 1.1 there is a unique λ : K′ → K with ιλ = ι′. Since (k,u ) F satisfies 1.1, the set of subHopfalgebras of F satisfying 1.1 is nonempty. Moreover, the subalgebra generated by two Hopf subalgebras satisfying 4 LEONID KROPANDYEVGENIAKASHINA 1.1 is a Hopf subalgebra satisfying 1.1 hence there exists a unique Hopf subalgebra satisfying 1.1. Following [1] we denote that Hopf algebra by HKerφ. Dually, a Hopf morphism and a linear epimorphism π : G → Q is called the Hopf cokernel of φ, notation Q = HCokerφ, if π fits into the commutative diagram φ F −−−→ G (1.2) ǫF π k −−u−Q→ Q y y and for every π′ : G → Q′ satisfying 1.2 there is a unique µ : Q → Q′ satisfying µπ = π′. It is immediate that π : G → G/Gι(F+)G is HCokerφ. We can make Definition 1.1. A Hopf algebra C is called a weak extension of a Hopf algebra B by a Hopf algebra A if there is a sequence of Hopf mappings (S) A ι C ։π B. with A = HKerπ and B = HCokerι. In what follows we assume the antipode of C is invertible. We point out that our definition of Hopf kernels coincides with the one in [19] and [1]. Let Fcoφ and coφF be the subalgebras of coinvari- ants [18, 3.4], also denoted by RKerφ and LKerφ, respectively, in [3]. A characterization of HKerφ in [19] and [1] yields immediately inclu- sions HKerφ ⊂ LKerφ ∩ RKerφ. On the other hand, the equalities HKerφ = RKerφ = LKerφ, hold provided RKerφ or LKerφ are Hopf subalgebras of F by the arguments [3, 4.19] or [1, 1.1.4]. A major drawback of Definition 1.1 is that it does not guarantee, in general, that ι(A) coincides with either one of subalgebras of coin- variants. That is, HKerφ is, in general, not equal to RKerφ. This is already easy to see for any Taft Hopf algebra [22] with respect to the standard epimorphism π : A → kG, G a cyclic group generated by T in the notation of [19, Ex.1.2]. In this case HKerπ is zero, i.e. equals to k, while RKerπ is |G|-dimensional. An example of a weak extension (S) with ι(A) $ RKerπ ∩ LKerπ is harder and can be found in [19, Ex.1.2]. We adopt the following definition of a Hopf algebra extension. Definition 1.2. A Hopf algebra C is said to be an extension of B by A if C is a weak extension of B by A and HKerφ = RKerφ. ISOMORPHISM TYPES OF HOPF ALGEBRAS 5 We add some comments about the definition. The equality HKerφ = RKerφ is equivalent to HKerφ = LKerφ and thanks to [3, 4.13] we have that ι(A) is normal in C. It then follows from [19] that HCokerι = ι(A)+C = Cι(A)+. In consequence our definition coincides with the definition of extension in [2]. If A is finite-dimensional, it is equivalent to H.-J. Schneider definition of strict extension in [19]. To see the latter,use[19,L.2.1(2)]toobtainthatC isfaithfullyflatA. Conversely, assuming A is normal andC is faithfully flat over A, apply [20, Remark 1.2] or [18, 3.4.3] to derive ι(A) = RKerπ. Inthecaseofafinite-dimensionalC weakerconditionssuffice. Namely, a sequence (S) with ι linear monomorphism, π linear epimorphism is an extension if either ι(A) = Ccoπ or Kerπ = ι(A)+C. For details see [2, 3.3.1]. 1.2. Abelian Extensions. We assume in what follows the ground field k to be an algebraically closed field of characteristic 0 and C to be a finite-dimensional Hopf algebra. An extension (S) is called abelian if A is commutative and B is cocommutative. It is well-known [10, Theorem 1] and [18, 2.3.1] that in this case A = kG and B = kF for some finite groups G and F. Below we consider only extensions of this kind and we use the notation (A) kG ι H ։π kF. Of the crucial importance to the sequel is a theorem [20, 2.4] and [14, 3.5]asserting H isa kF-crossed product over kG 1. Thetheorem entails the existence of a mapping called section ( see, e.g. [2, 3.1.13]) (1.3) χ : kF → H giving rise to the crossed product structure on H. Thus H = kGχ(F) with the multiplication (1.4) (αχ(x))(α′χ(y)) = α(χ(x)α′χ−1(x))χ(x)χ(y) (1.5) = α(χ(x)α′χ−1(x))[χ(x)χ(y)χ−1(xy)]χ(xy) forα,α′ ∈ kG, x,y ∈ F The mapping x ⊗ α 7→ x.α := χ(x)αχ−1(x) defines a module-algebra action of F on kG and the function σ : F × F → kG,σ(x,y) = χ(x)χ(y)χ−1(xy) is a left, normalized 2-cocycle for that action [18, 7.2.3]. The definition of action is independent of a choice of section. This can be made more precise. Let us write Sec(kF,H) for the set of all sections of kF in H. Next define the group R = Reg (kF,kG) 1,ǫ of all convolution invertible mapping preserving the unit and counit. 1 A short independent proof for abelian extension is given in the Appendix 6 LEONID KROPANDYEVGENIAKASHINA Identifying ι(A) with A, Sec(B,C) becomes an R-set under multiplica- tion by convolution, viz. f.χ = f ∗χ, f ∈ R,χ ∈ Sec(kF,kG), and by [18, 7.3.5] Sec(kF,H) is a transitive R-set. The claim follows, for f ∗χ evidently induces the same action as χ. The theory of extensions depends fundamentally on the fact that for each sequence (S) its companion sequence (S∗) B∗ π∗ C∗ ։ι∗ A∗ isalsoanextension, see[4, 4.1]or[2, 3.3.1]. Intheabeliancase(kF)∗ = kF and also (kG)∗ = kG via the Hopf isomorphism ev : kG → kG,ev(g)(α) = α(g) for allα ∈ kG Thus every diagram (A) induces a diagram (A∗) kF H∗ ։ kG A crossed product structure on H∗ is effected by a section (1.6) ω : kG → H∗ We choose to write H∗ = ω(G)kF with the multiplication (1.7) (ω(a)β)(ω(b)β′) = ω(ab)τ(a,b)(β.b)β′, where for a,b ∈ G,β,β′ ∈ kF (1.8) β.b = ω−1(b)βω(b),and (1.9) τ(a,b) = ω−1(ab)ω(a)ω(b). We note that τ : G×G → kF is a right, normalized 2-cocycle for the action β ⊗b 7→ β.b. Below we write x for χ(x) and a for ω(a). The left action of F on kG induces a right action of F on G by set permutations of G. Namely, we define the element a⊳x by the equality (1.10) (x.f)(a) = f(a⊳x) for allf ∈ kG Noting that an algebra action permutes minimal central idempotents, the action ‘⊳’ is related to the action of F on kG in the basis {p |a ∈ G} a [18] by (1.11) x.pa = pa⊳x−1 For, x.p = p iff (x.p )(b) = 1. Now, (x.p )(b) = p (b ⊳ x) = 1 iff a b a a a b⊳x = a, hence b = a⊳x−1. Similarly, the right action of G on kF induces a left action of G on F by permutations denoted by a⊲x satisfying the property (1.12) px.a = pa−1⊲x ISOMORPHISM TYPES OF HOPF ALGEBRAS 7 We fuse both actions into the definition of a product on F ×G via (1.13) (xa)(yb) = x(a⊲y)(a⊳y)b It was noted by M.Takeuchi [23] that composition 1.13 defines a group structure on F ×G provided the actions ⊲,⊳ satisfy the conditions (1.14) ab⊳x = (a⊳(b⊲x))(b⊳x) (1.15) a⊲xy = (a⊲x)((a⊳x)⊲y) We use the standard notation F ⊲⊳ G for the set F ×G endowed with multiplication (1.13). 1.3. Co-crossed Coproducts. Fundamental to the theory of abelian extensions is an explicit description of the coalgebra structure of the dual coalgebra of a crossed product algebra H = A# B [18], A and σ B finite-dimensional. The coalgebra H∗ is an example of a co-crossed co-product of the coalgebra A∗ with bialgebra B∗. Coalgebras of this type were introduced in [11], generalizing an earlier construction in [17] and studied in [1]. The formula below is [11, p.9] and [1, 2.15], though we derive it rather than postulate. We need some preliminaries. We identify H∗ with A∗⊗B∗ via thepairing ha∗ ⊗b∗,a⊗bi = ha∗,aihb∗,bi. Aweak actionr : B⊗A → Ainduces a weak coactionρ : A∗ → B∗⊗A∗ via (1.16) ρ(a∗) = (a∗)(1) ⊗(a∗)(2) ⇔ a∗(b.a) = (a∗)(1)(b)(a∗)(2)(a). Next, the 2-cocycle σ : B ⊗B → A induces a 2-co-cocycle θ : A∗ → B∗ ⊗B∗ as follows (1.17) θ(a∗) = (a∗)h1i ⊗(a∗)h2i ⇔ a∗(σ(b,b′)) = (a∗)h1i(b)(a∗)h2i(b′). Finally, we use notation (a∗) ⊗ (a∗) for the coproduct in A∗ and a 1 2 similar notation in B∗. We agree to write a∗θ♯ b∗ for a∗⊗b∗ viewed as an element of the co-crossed co-product coalgebra H∗. Lemma 1.3. Let ∆H∗ and ǫH∗ be the structure maps of H∗. There hold the formulas: ∆H∗(a∗θ♯ b∗) = (a∗)1θ♯ (a∗)(21)(a∗)h31i(b∗)1 ⊗(a∗)(22)θ♯ (a∗)h32i(b∗)2, ǫH∗(a∗θ♯ b∗) = a∗(1A)b∗(1B). Proof:By definition of ∆H∗ and multiplication in H [18, 7.1.1] ∆H∗(a∗θ♯ b∗)(a#σb⊗a′#σb′) = ha∗ ⊗b∗,(a#σb)(a′#σb′)i = a∗(a(b .a′)σ(b ,b′))b∗(b b′). 1 2 1 3 2 8 LEONID KROPANDYEVGENIAKASHINA Expanding the right-hand side of the last equality using (1.16) and (1.17) and reshuffling factors we get (a∗) (a)(a∗)(1)(b )(a∗)(2)(a′)(a∗)h1i(b )(a∗)h2i(b′)(b∗) (b )(b∗) (b′) 1 2 1 2 3 2 3 1 1 3 2 2 = [(a∗) (a)((a∗)(1)(b )(a∗)h1i(b )(b∗) (b )][(a∗)(2)(a′)((a∗)h2i(b′)(b∗) (b′)] 1 2 1 3 2 1 3 2 3 1 2 2 = (a∗) (a)[(a∗)(1)(a∗)h1i(b∗) ](b)(a∗)(2)(a′)[(a∗)h2i(b∗) ](b′), 1 2 3 1 2 3 2 which is the the first formula. The formula for the counit is just ha∗ ⊗b∗,1 ⊗1 i. (cid:3) A B We return to abelian extensions. We introduce some notation. For a group Γ we let Γn be the direct product of n copies of Γ. Let F and G be groups. We abbreviate (x ,...,x ) ∈ Fn and (a ,...,a ) ∈ Gm 1 n 1 m to x and a, respectively. We note that every α : Fn → (kGm)• can be uniquely represented in the standard basis {p |a ∈ Gm} as a (1.18) α(x) = α(x)(a)p . a a X There α(x,a) := α(x)(a) is a mapping Fn×Gm → k•. Conversely, ev- ery φ : Fn×Gm → k• gives rise to two mappings φ′ : Fn → (kGm)• and φ′′ : Gm → (kFn)• via(1.18)forφ′ andtheformulaφ′′(a) = φ(x,a)p x for φ′′. Let Fun(X,Y) be the set of mappings from a set X to a set Y. We arrive at the conclusion that the groups Fun(FPn,(kGm)•), Fun(Gm,(kFn)•) and Fun(Fn ×Gm,k•) can be identified. Proposition 1.4. Suppose H = kG# kF. The structure mappings of σ coalgebra H∗ are given by the formulas (1.19) ∆H∗(a⊗β) = σ(x,y,a)a⊗pxβ1 ⊗(a⊳x)⊗pyβ2, a ∈ G,β ∈ kF, x,y∈F X ǫ(a⊗β) = β(1 ) F Proof:By the preceeding Lemma we need to know the coaction ρ : kG → kF ⊗ kG dual to the action of F on kG or, equivalently, to the action of F on G (see (1.11)). As well, we need a formula for the co-cocycle θ : kG → kF ⊗kF. These formulas are as follows: (1.20) ρ(a) = p ⊗a⊳x x x∈F X (1.21) θ(a) = σ(x,y,a)p ⊗p x y x,y∈F X ISOMORPHISM TYPES OF HOPF ALGEBRAS 9 To show the first formula, begin with the expression ρ(a) = p ⊗ y∈F y r with r ∈ kG. Apply (1.16) and calculate y y P hρ(a),x⊗pbi = ev(a)(x.pb) = (x.pb)(a) = pb⊳x−1(a) = δb,a⊳x On the other hand hρ(a),x⊗p i = p (x)p (r ) = p (r ). Say r = b y b y b x x kc c,kc ∈ k,c ∈ G. From p (r ) = kb = δ we get the equality x x b x x b,a⊳x P r = a⊳x, which is the result. x P To see the second formula, we use (1.17) to compute hθ(a),x⊗yi = ev(a)(σ(x,y)) = σ(x,y,a) which gives the result. Substituting those expressions of ρ and θ in Lemma 1.3 gives the first statement. To compute the counit use the fact that 1 = ǫ . (cid:3) kG G Remark 1.5. Starting off with the crossed product H∗ = kG# kF, τ an argument similar to the one in Proposition 1.4 gives the coalgebra structure mappings for H [16, p.7], namely (1.22) ∆ (α⊗x) = τ(x,a,b)α p ⊗b⊲x⊗α p ⊗x H 1 a 2 b a,b∈G X ǫ (α⊗a) = α(1 ) H G 1.4. Compatible Cocycles and Equivalence Relation. The dis- cussion in §1.2 enables us to associate to every Hopf algebra H in a diagram (A) a datum {σ,τ,⊳,⊲} and we write H = H(σ,τ,⊳,⊲). Two data σ,⊳ and τ,⊲ appearing in the crossed products H and H∗ are compatible if the coalgebra structure induced in H by multiplication in H∗ (or equivalently in H∗ by multiplication in H) turns H (or H∗) into a bialgebra, hence a Hopf algebra [2, 3.1.12]. The compatibility conditions are derived using (1.19) or (1.22). They can be found in [16, 4.7] as follows. Theorem 1.6. (1) F ⊲⊳ G is a group, (2) σ(x,y,1) = 1 = τ(a,b,1), σ(x,y,ab)σ−1(b⊲x,(b⊳x)⊲y,a)σ−1(x,y,b) (3) = τ(a,b,x)τ(a⊳(b⊲x),b⊳x,y)τ−1(a,b,xy) The second key concept of the theory is equivalence of extensions. Following [6, 3.10] we call two extensions E = (A ι C ։π B) and E′ = (A ι′ C ։π′ B) equivalent, and write E ∼ E′, if there is a Hopf 10 LEONID KROPANDYEVGENIAKASHINA algebra map ψ : H → H′ making the following diagram commute ι π A −−−→ H −−−→ B (1.23) ψ A(cid:13)(cid:13) −−ι−′→ H′ −−π−′→ B(cid:13)(cid:13) (cid:13) y (cid:13) Welet Ext(kF,kG) stand fortheset ofequivalence classes ofextensions oftype(A).WedesignateOpext(kF,kG,⊳,⊲))fortheset ofequivalence classes of extensions with the fixed actions ⊳ and ⊲. We proceed to describe equivalent extensions precisely. As a preliminary we introduce groups Map(Fn × Gm,k•) of all normalized mappings, i.e. f : Fn × Gm → k• such that f(x ,...,x ,a ,...,a ) = 1 if some of x ,...,x 1 n 1 m 1 n or a ...,a are equal to 1 or 1 , respectively, under the pointwise 1 m F G multiplication. We define actions of F and G on Map(Fn×Gm,k•) as follows. For any f : Fn ×Gm → k•, (1.24) y.f(x ,...,x ,a ,...,a ) = f(x ,...,x ,a ⊳(a ···a ⊲y),...,a ⊳y) 1 n 1 m 1 n 1 2 m m (1.25) f(x ,...,x ,a ,...,a ).b = f(b⊲x ,...,(b⊳x ···x )⊲x ,a ,...,a ), 1 n 1 m 1 1 n−1 n 1 n where all x ,y ∈ F and a ,b ∈ G. i j The formulas below reflect the fact, standard in the theory of exten- sions, that an equivalence is just a diagonal change of basis. They are particular cases of [6, 3.11] and also [16, 5.2]. Combining [16, 3.4]or [18, 7.3.4] with (1.19) or (1.22) one can show Lemma 1.7. Two extensions defined by data {σ,τ,⊳,⊲} and {σ′,τ′,⊳′,′⊲} are equivalent if and only if ⊳ = ⊳′, ⊲ = ′⊲ and there exists ζ ∈ Map(F ×G,k•) satisfying (1.26) σ′(x,y,a) = σ(x,y,a)(x.ζ−1(y,a))ζ(xy,a)ζ−1(x,a), (1.27) τ′(x,a,b) = τ(x,a,b)(ζ(x,a).b)ζ(x,ab)−1ζ(x,b) The mapping ψ of (1.23) is given by ψ(αx) = αζ(x)x for all α ∈ kG,x ∈ F. We shall need the dual version of the above Lemma. We agree to write H∗(σ,τ,⊳,⊲) for H(σ,τ,⊳,⊲)∗. Lemma 1.8. H∗(σ,τ,⊳,⊲) and H∗(σ′,τ′,⊳′,′⊲) are equivalent exten- sions if and only if ⊳ = ⊳′, ⊲ = ′⊲ and there exists η ∈ Map(F ×G,k•)