You are on page 1of 32

# HOPF ALGEBRAS

1. Algebras and coalgebras

1.1. Algebras. We start by formulating a familiar definition in a possibly unfa-
miliar way:

Definition. A k-algebra is a k-vector space with two linear maps

m : A ⊗k A → A and u:k→A

satisfying the following conditions hold. These respectively encode associativity
and the unit element:
(a) the diagram
m⊗id
A⊗A⊗A / A⊗A

id⊗m m
² m ²
A⊗A /A
is commutative;
and (b) the diagram

u:
u⊗id uuu
II id⊗u
II
u II
uuu II
u
k ⊗ AJ m A⊗k
JJ tt
JJ t
J tt
s JJJ tt
J\$ ² zttt s
A
(where s denotes scalar multiplication) is commutative.

We get 1A as u(1k ) via the second diagram.

1.2. Coalgebras. To define these, we reverse the arrows in (7.1):

Definition. A k-coalgebra is a k-vector space, C, with two k-linear maps, ∆ (co-
multiplication or coproduct) and ε (counit), with

∆:C →C ⊗C and ε : C → k,

such that the following conditions, respectively called the coassociativity and the
counit axioms, hold:

1

2 HOPF ALGEBRAS

a) the diagram
∆⊗id
C ⊗ CO ⊗ C o C ⊗O C
id⊗∆ ∆

C ⊗C o C

commutes;
b) the diagram
C ⊗C
t O JJJ id⊗ε
ε⊗id ttt JJ
t JJ
ttt JJ
zt \$
k ⊗ CdJ ∆ C
: ⊗k
JJ 1⊗ ⊗1 ttt
t
JJ
JJ t
JJ tt
J ttt
C
commutes, where 1 ⊗ is the map x 7→ 1 ⊗ x.
1.3. Morphisms.
Definition. Let A and B be k-algebras. A linear map θ : A → B is a k-algebra
homomorphism if the diagrams
θ
AO /B
O
mA mB

A⊗A / B⊗B
θ⊗θ

and
θ
AO /B
O
uA uB

k k
commute.
Definition. Let C and D be k-coalgebras. A linear map f : C → D is a k-coalgebra
morphism if the diagrams
f
C /D

∆c ∆D
² f ⊗f ²
C ⊗C / B⊗B

and
f
C /D

εC εD
² ²
k k

HOPF ALGEBRAS 3

commute.

1.4. Example. Let V be a k-vector space with basis B. Then V is a coalgebra if
we set

∆(V ) = v⊗v ∀v ∈ B
ε(V ) = 1 ∀v ∈ B

(extended linearly). V is called the (group-like) coalgebra on the set B.

2. Bialgebras

2.1. The definition.

Definition. A bialgebra A is a k-vector space, A = (A, m, u, ∆, ε) where (A, m, u)
is an algebra; and (A, ∆, ε) is a coalgebra; and such that either (and hence both)
of the following two conditions hold:-
(1) ∆ and ε are algebra homomorphisms; and
(2) m and u are coalgebra homomorphisms.

Notes. 1. If A and B are two k-algebras then so is A⊗B if we define (a⊗b)(c⊗d) =
ac ⊗ db. Expressed as a diagram this is:

A⊗B⊗A⊗B / A ⊗ A ⊗ B ⊗mBA ⊗mB / A ⊗ B
id⊗τ ⊗id

where τ : B ⊗ A → A ⊗ B is the flip: τ (b ⊗ a) = a ⊗ b. The unit uA⊗B of A ⊗ B is
given by
uA ⊗uB
k∼
=k⊗k / A⊗B .

Similarly, if C and D are coalgebras then so is C ⊗ D with ∆C⊗D given by

∆C ⊗∆D
C ⊗D / C ⊗ C ⊗ D ⊗ D id⊗τ ⊗id/ C ⊗ D ⊗ C ⊗ D

and counit
εC ⊗εD
C ⊗D / k⊗k ∼
=k .
In particular this applies when A = B and when C = D.

2. (1) ⇔ (2) in the definition follows by drawing the diagrams to express the
conditions that ∆ and ε are algebra morphisms; we find the diagrams are the same
as the ones to give (2); see Exercises 2.4(1).

3. A morphism of bialgebras f : A → B is a linear map which is both an algebra
and a coalgebra homomorphism.

then kerf is called a biideal : this means that kerf is an ideal and a coideal (ie the kernel of a coalgebra homomor- phism). and ε(g) = 1 for g ∈ G. ¤ (3) Enveloping algebras of Lie algebras: A Lie algebra g is a vector space over a field k with an operation [ . z] + [y. Here are the first few examples.4(2). [y.5(3) where we check that I ⊆ A is a coideal if ε(I) = 0 and ∆(I) ⊆ C ⊗ I + I ⊗ C. to T (V ) and to k respectively. We will add to this list once we have defined Hopf algebras. If A is generated as an algebra by a set B. All the maps in the diagrams are algebra homomorphisms given our hy- pothesis. So we can apply Lemma (2. . y. 2.2 for input {vi } a basis element. (1) Group algebras: Let G be any group and let kG be the group algebra. Lemma 2. (2) Tensor algebras: Let V be a vector space and form T (V ). [x. z]] = [[x. T (V ) is the free k-algebra on generators {vi : i ∈ I}. so M T (V ) = (V ⊗n ). Then T (V ) is a bialgebra if we define ∆(V ) = v ⊗ 1 + 1 ⊗ v ∈ T (V ) ⊗ T (V ) and ε(V ) = 0 for v ∈ V . So ∆ and ε extend uniquely to algebra homomorphisms. Proof. z]]. This is easy. Proof.2 with input from B. then the operatins on A induce a structure of bialgebra on A/J. See Ex1. z ∈ g. to prove that A is a bialgebra we only need to check the diagrams in 1. [x. extended linearly. see Exercise 2.2)(1) and see that we only need to check commutativity of the diagrams in 1. When I is a biideal of a bialgebra A.3. ]:g×g→g which is bilinear. namely for all x. antisymmetric and satifies the Jacobi identity. Test Lemmas. n≥0 multiplication of tensors is given by juxtaposition.4 HOPF ALGEBRAS 2. Examples of bialgebras. It is a bialgebra if we define ∆ : kG → kG ⊗ kG by ∆(g) = g ⊗ g for g ∈ G. ¤ If f : A → B is a bialgebra homomorphism. the tensor algebra. Suppose that A is an algebra and we have algebra homomorphisms ∆ : A → A ⊗ A and ε : A → k.2. a basis of V . Lemma 1. So commutativity in general follows at once from commutativity on the generators. y].

we write this as gl(n. j. y] − xy + yx : x. k) is 3-dimensional. f ] = h. 0 0 1 0 0 −1 Check: [h. k) be the space of all n × n matrices of trace 0. when A = Mn (k). b] = ab − ba for all a. This is a special case of the Theorem (Poincar´e-Birkhoff-Witt).Lie algebra homomorphisms ρ : g → End(V ) for k-vector spaces V . ef − f e − h > . j. HOPF ALGEBRAS 5 For example. . . . e] = 2e. Special case (C): sl(2. with usual basis Ã ! Ã ! Ã ! 0 1 0 0 1 0 e= . Special case (B): Let n ≥ 2. Then U (g) ∼ = k[x1 . xj ] = 0 for all i. . b ∈ A. . The universal enveloping algebra of the Lie algebra g is the factor algebra of the tensor algebra T (g) by the ideal I(g) =< [x.f = . . f ] = −2f. g = i=1 kxi with [xi . and let sl(n. if A is any associative k-algebra then A is a Lie algebra if we define [a. [h. In particular. so I(sl(2. Fact. where the latter are .by definition .h = . k)) =< he − e(h + 2). U (g) is a bialgebra for all Lie algebras g. y ∈ g > There is a bijective correspondence between left U (g)-modules and representations of the Lie algebra g. This is a Lie algebra called the special linear Lie algebra. k)) has k-basis {ei hj f t : i. xn ]. [e. hf − f (h − 2). t ≥ 0}. It’s easy to check that U (sl(2. . with 4(x) = x ⊗ 1 + 1 ⊗ X and ε(x) = 0 for all x ∈ g. If g is a Lie algebra with k-basis {x1 . . . the commutative polynomial algebra. k). xn } then U (g) has k-basis {xt11 xt22 · · · xtnn : ti ≥ 0}. Pn ⊕ Special case (A): Let n ≥ 1.

2(2)) that I(g) is a biideal. So. Similarly. Definition. We use 4n−1 to denote the iterated application of 4 as above. . For c ∈ C. if 4 = τ ◦ 4. so 4n−1 : C → C ⊗n . B cop and B opcop are all bialgebras. So all the bialgebras so far introduced are cocommutative. A co. the opposite algebra of A. is the same k-vector space as C. but with 4C cop := τ ◦ 4C .6 HOPF ALGEBRAS Proof. C cop . 2. Consider associativity: expressed using the above. the opposite coalgebra. if C is a coalgebra. b ∈ A. ¤ (4) Opposite algebras and coalgebras: If A is any k-algebra.or bialgebra A is called cocommutative if Acop = A. We know U (g) = T (g)/I(g). Sweedler notation. it says ³X ´ X (id ⊗ 4) ◦ 4(c) = (id ⊗ 4) c1 ⊗ c2 = c1 ⊗ c21 ⊗ c22 should equal X c11 ⊗ c12 ⊗ c2 . denoted Aop . Because the maps are algebra homomorphisms. Let C be a coalgebra.4. c1 ⊗ 4(c2 ) ⊗ c3 and c1 ⊗ c2 ⊗ 4(c3 ) are all equal in C ⊗ C ⊗ C ⊗ C. Applying coassociativity to (1). If B is a bialgebra. where τ = is the flip. we only need to do the check for t a generator of I(g). is the same vector space as A but with new multiplication a · b = ba for all a. we find that the three expressions X X X 4(c1 ) ⊗ c2 ⊗ c3 . that is 4(t) ∈ T (g) ⊗ I(g) + I(g) ⊗ T (g) for all t ∈ I(g) and that ε(t) = 0 for all t ∈ I(g). So we only need to check (thanks to Lemma 2. and T (g) is a bialgebra with the same defini- tions of 4 and ε. we write X 4(c) = c1 ⊗ c2 . then it’s easy to check that B op . It’s easy to do this. We write this as X c1 ⊗ c2 ⊗ c3 ⊗ c4 . we write both of the above simply as X (1) c1 ⊗ c2 ⊗ c3 . that is.

Since S is linear. A) called the convolution product. A Hopf algebra is a bialgebra with an antipode.1. X X (2) ε(a) = a1 S(a2 ) = S(a1 )a2 .5(3)). ∗. u) be an algebra and C = (C.2) says that for all c ∈ C. k any field. 3. 3.2. ∗. g ∈ Homk (C. Then we can define a product on Homk (C. ∆. In Sweedler notation. ε) be a coalgebra. Then a linear endomor- phism S from A to A is an antipode for A if the diagram m A ⊗O A /Ao A ⊗O A O m id⊗S u◦ε S⊗id 4 A⊗A o A / A⊗A 4 commutes. A) and c ∈ C (3) (f ∗ g)(c) = Σf (c1 )g(c2 ). Associativity of ∗ follows from associativity of the coproduct: ((f ∗ g) ∗ h)(c) = Σf (c1 )g(c2 )h(c3 ) = (f ∗ (g ∗ h))(c). Let A = (A. this says that for all a ∈ A. A). X X c= ε(c1 )c2 = c1 ε(c2 ). HOPF ALGEBRAS 7 Using Sweedler notation. ε) be a bialgebra. With the above notation (Homk (C. by: for f. Then kG is a Hopf algebra if we define S(g) = g −1 for all g ∈ G. Hopf algebras 3. it’s enough to check (2) on G: it says that 1 = ε(g) = gg −1 = g −1 g for all g ∈ G. Let A = (A. Convolution product. Proposition. u. P And C is cocommutative ⇔ 4(c) = c2 ⊗ c1 for all c ∈ C. m. the second axiom in (1. Proof. m. over k. u ◦ ε) is an algebra. 4. Morphisms of Hopf algebras are just bialgebra maps “preserving the antipode” (Exercise 3. Definition. Example: Let G be any group. .

Comparing Definition 3. (2) Uniqueness of inverses. (2) If a bialgebra does have an antipode then that antipode is uniquely deter- mined.8 HOPF ALGEBRAS And u ◦ ε is a left identity element. for f. everything works if dimk A < ∞. for all x. because for f ∈ Homk (C. ∗. k) is an algebra.1 with (3) we get the first part of Corollary. u. Let H = (H. Theorem. If C = (C. Indeed we might define. ∆. m. ε ◦ S = ε. S(1) = 1 and (5) (S ⊗ S) ◦ ∆ = S. 3. with (f ∗ g)(c) = Σf (c1 )g(c)2 . Proof. m. But notice that this gives us ∆(f ) ∈ (A ⊗ A)∗ and if dimk A = ∞ then A∗ ⊗ A∗ \$ (A ⊗ A)∗ . However. ε) is any k-coalgebra. u) we can dually get a coalgebra structure on A∗ . Similarly it’s a right identity. u ◦ ε) is an algebra. y) = S(y)S(x). as we’ll discuss later. u ◦ ε). ¤ Corollary. S) be a Hopf algebra. .3. ∗. Remark. y ∈ H (4) S(x. ∆. Then S is a bialgebra homomorphism from H to H opcop . and it can be repaired in general. Properties of the antipode. That is. then C ∗ = Homk (C. g ∈ A∗ ε(f ) = f (1A ) (= u∗ (f )) and ∆(f ) ∈ A∗ ⊗ A∗ given by ∆(f )(x ⊗ y)) = f (xy). A) and c ∈ C ((u ◦ ε) ∗ f )(c) = Σε(c1 )f (c2 ) = (f ∗ (g ∗ h))(c) = f (c). An antipode S for A is an inverse for id |A in (Endk (A). ε. ∗ So C is commutative if and only if C is cocommutative. (1) Suppose A is a bialgebra. We might expect that given an algebra A = (A. ¤ Applying the above when C = A is a bialgebra we get that (Endk (A).

Exercise 3. H) by µ(x ⊗ y) = S(y)S(x) ρ(x ⊗ y) = S(xy). Define maps ν and ρ in Homk (H ⊗ H.2 (1) to id ∗ S = u ◦ ε. If we apply corollary 3. Exercise (3. S(1) = 1. We shall show that µ = ρ by proving that (7) ρ∗m=m∗µ=u◦ε where m is multiplication in H.5 (1). ∆S(x2 ) ⊗ S(x1 ) = Σ(Sx)1 (Sx)2 .2. it’s enough to check (2) for a ∈ B. noting that ε on H ⊗ H is just ε ⊗ ε. t ∈ R is called an anti-homomorphism. Proof. A map of rings θ : R → S which satisfies θ(rt) = θ(t)θ(r) for all r. To check that S is an antipode for A. Let S : A → Aop be an algebra homomorphism. Let A = k < B > be a bialgebra. Proof. Note (1) with (6) ∆H⊗H (x ⊗ y) = Σ(x1 ⊗ y1 ) ⊗ (x2 ⊗ y2 ). Note that H ⊗ H is a coalgebra by 2. y ∈ H. (5) is proved similarly . giving in particular. take x. This proves (7). ¤ Definition. ¤ .1. Similarly (m ∗ µ)(x ⊗ y) = Σm((x ⊗ y)1 )µ((x ⊗ y)2 ) = Σx1 y1 S(y2 )S(x2 ) by (6) and definition of µ = Σx1 (Σy1 S(y2 ))S(x2 ) = ΣS((xy)1 )(xy)2 = u ◦ ε(x ⊗ y).5)(2). This will do it! Well. HOPF ALGEBRAS 9 that is. Homk (H ⊗ H. Then ρ ∗ m(x ⊗ y) = Σρ((x ⊗ y)1 )m((x ⊗ y)2 ) = Σρ(x1 ⊗ y1 )m(x2 ⊗ y2 ) = Σ(x1 y1 )x2 y2 = ΣS((xy)1 )(xy)2 since ∆ is an algebra homomorphism = u ◦ ε(x ⊗ y). H) is an algebra. So by proposition 3. Analogous to lemma 1 we now have Corollary.

Form H = R∗ < g >. Check routinely that ∆(I) = 0 = εˆ(I). Same for ΣS(v1 )v2 .4. we see that S([x. gˆx ˆ+ x ˆgˆ >= I. x (ii) H is the factor of the free algebra F = k < gˆ. y ∈ g. if we define S(g) = g.1. So T (V ) is a Hopf algebra. For x. So gx = −xg. So U (g) is a Hopf algebra. x. y] + yx − xy) ∈ I(g).3 (2). A finite dimensional noncommutative. Let g be a Lie algebra and recall that U (g) := T (g)/I(g) as in 2. Define Sˆ : F → H op by lifting our proposed definitions for S(g). ε(g) = 0. so the maps factor through H. (iii) H is a bialgebra with ∆(g) = g ⊗ g. non cocommutative Hopf Algebra. Define an antihomomorphism S : T (V ) −→ T (V ) v 7−→ −v. Assume chark 6= 2. ∆(x) = x ⊗ 1 + g ⊗ x. For v ∈ V . The diagrams in Definition 1. So: (i) dimk H = 4. 2.2 are easy to check. Check that S(I) ˆ = 0. S(ˆ S(ˆ ˆ x). ∆ and εˆ : F → k ˆ by lifting the definitions of ∆ and ε. So S induces an antiautomorphism of U (g) satisfying definition 3. y] − (−y)(−x) + (−x)(−y) = −([x. 1. g. Proof. S(x) = −gx and S(xg) = S(g)S(x) = g(−gx) = −x. x ˆ2 . (iii): Freeness of F allows us to define ˆ : F → H ⊗ H. More examples of Hopf algebras. so we get antihomomorphism H → H as we . ε as in 2. ε(v) = 0 while Σv1 S(v2 ) = v · 1 + 1 · (−v) = v − v = 0 = ε(v). S(x) to ˆ g ). with basis {1.3 (3). Proof. ε(g) = 1. ¤ (iv) H is a Hopf algebra. Take R = k[x]/ < x2 > and let g ∈ Autk−alg (R) by g(x) = −x. the skew group ring. 3. y] − xy + yx) = −[x.10 HOPF ALGEBRAS 3. Let < g > be a cyclic group of order 2. The tensor algebra T (V ) Define ∆. xg}. ˆ > by the ideal < gˆ2 − 1.

. See Exercise (3. for a Hopf algebra H . But there are positive cases: (a) If dimk H < ∞. Finally m ◦ (S ⊗ id) ◦ ∆(x) = m ◦ (S ⊗ id)(x ⊗ 1 + g ⊗ x) = m(−gx ⊗ 1 + g ⊗ x) = −gx + gx = 0 = ε(0). (n ≥ 2) in 1971. 1971]. It’s enough to prove that S 2 is right convolution inverse to S. (a0 ) [Larson.3(2). If H is noetherian then S is bijective. Conjecture (Skryabin). Proof.5. (i) ⇒ (ii): Suppose S 2 = idH . Then X ³X ´ ³X ´ S(x2 )x1 = S 2 S(x2 )x1 = S S(x1 )S 2 (x2 ) 3. Let x ∈ H. (b) If H is commutative or cocommutative. 2006] If H is noetherian and is either semiprime or satisfies a polynomial identity then S is bijective. the antipode S is not bijective. then S 2 = idH .5)(7).there are examples due to Takeuchi (1971). 1976] In fact. (ii) ⇒ (i): Suppose (ii). ¤ Notes.3(1) ³X ´ = S S(x1 )x2 = S(ε(x)1h ) by definition of antipode = ε(x)1H by 3. x2 S(x1 ) = ε(x)1H . P (ii) for all x ∈ H. P (iii) for all x ∈ H. HOPF ALGEBRAS 11 want. (1) S has order 4 in this example. In general. by Corollary 3. So S 2 = idH . Let H be a Hopf algebra. S(x2 )x1 = ε(x)1H . then S is bijective [Larson and Sweedler. Further properties of the antipode. Then the following are equivalent: (i) S 2 = id. (c) [Skryabin. We’ll prove (b): Theorem.3(2). 3. (2) Taft generalised Example 3 to give an infinite series Hn of Hopf algebras of dimension n2 . if dimk H < ∞ then S has finite order (but there does not exist a global bound on possible orders (see Exercises(3.2A(i).6)(7))).3(1) = S(ε(x)1H ) = ε(x)1H by Theorem 3. Let x ∈ H. Then X ³X ´ S ∗ S 2 (x) = S(x1 )S 2 (x2 ) = S S(x2 )x1 by Theorem 3.

1(1). If H is commutative or cocommutative. u. Let H = (H. u. the argument is similar. x. So X a1 · S −1 (a2 ) = ε(a)1A . ∆. (ii) From Theorem 3. the isomorphism being given by S. m. Our aim here is to obtain a dual result to Corollary B of (3. X X a1 · S −1 (a2 ) = S −1 (a2 ) × a1 . For H cop . Applying S to the right hand side. S −1 ) are isomorphic Hopf algebras. Let A = (A. 3. Let A be any k-algebra. × for m). u) be a k-algebra. S) be a Hopf algebra. for f ∈ Ao . ∆. ε. For a ∈ H op (writing · for mop . (iii) We know that H op and H cop are bialgebras. then S 2 = idH . isomorphic via S thanks to The- orem 3. ¤ Corollary (A).3. . mop . (ii) S : H → H opcop is a Hopf algebra morphism. ¤ 3. Similarly. u. ∆(f )(x ⊗ y) := f (xy). In Chapter 5 we’ll prove: Theorem.6. m. we get X S(a1 ) × a2 . ∆op . The key definition is: Definition.that is. u.3. S) is another Hopf algebra. S −1 ) and H cop = (H.1(1) translated into statements about H opcop ensures that S is an antipode for H opcop . The counit of Ao is ε = u∗ . The finite dual or restricted dual of A is Ao = {f ∈ A∗ : f (I) = 0 for some I C A with dimk (A/I) < ∞}. That is. ε(f ) = f (1A ). which is ε(a)1A by 3. Corollary (B). ε. mop . ε. Then H op = (H.2). m. ∆op .3)(4) we saw that H opcop is a bialgebra. Then (i) H opcop = (H. The Hopf dual. (iii) Suppose S is bijective. X S −1 (a1 ) · a2 = ε(a)1A . ε. y ∈ A. Proof. (i) Ao is a coalgebra with coproduct m∗ = ∆. (i) In (2.12 HOPF ALGEBRAS (i) ⇔ (iii) Similar.

ε) is a bialgebra. HOPF ALGEBRAS 13 (ii) If B = (B. m∗ . B o = (B o . ε∗ . 4. ∆. ∆∗ : B o ⊗ B o → B o : f ⊗ g 7→ f · g. u. Let k be any field and G any finite group.h ρg and as a k-algebra. h ∈ kGo and x ∈ G. Of course. m. Are the converses of these statements true? Example.g. then B o is commutative. then so is B o . we quickly recall the definition of a left module M over a k- algebra A: it’s a k-vector space with a k-linear map λ : A ⊗ M → M such that id⊗λ A⊗A⊗M / A⊗M m⊗idM λ ² ² A⊗M /M λ and u⊗id k ⊗ MK / A⊗M KK KK K s KKK λ K% ² M commute. u∗ ). Then (kG)∗ = kGo = Homk (kG. where X f · g(x) = f (x1 )g(x2 ). w ∈ G. ∆∗ . uw=g and for f. Modules and comodules 4. so ρg (h) = δg. if B is commutative. We can take as k-basis for kGo the dual basis {ρg : g ∈ G}.1. (so e. (iii) If additionally B is a Hopf algebra with antipode S. where s is scalar multiplication. then B o is cocommutative and if B is cocom- mutative. . To dualise it. then B o is a Hopf algebra with antipode S ∗ (so S ∗ (f ) = f ◦ S). h. So we find that ρg ρh = δg. kGo is the direct sum of |G| copies of k.h . k). ∈ G. g. (f h)(x) = f (x)h(x). Then X ∆(ρg ) = ρu ⊗ ρw where u.

14 HOPF ALGEBRAS 4. X X (g ∗ f ) · m = hg ∗ f. Comodules. Then M is a right Ao -comodule (in the sense that A → Ao∗ and part (1) coupled with restriction retrieves M ) ⇔ for all m ∈ M . Thus. m2 ihg. A right comodule M over C is a k-vector space M and a linear map ρ : M → M ⊗ C such that ρ M / M ⊗C ρ idM ⊗∆ ² ² M ⊗C / M ⊗C ⊗C ρ⊗id and ρ M G / M ⊗C GG GG GG id⊗ε −⊗1 GG # ² M commute. Left comodules are defined analogously. .3. Let C be a coalgebra. dimk (Am) < ∞. m1 ihf. m2 im0 .2(1). 4. so X ρ(m) = m0 ⊗ m1 . Duality. (ii) Let A be an algebra and M a left A-module. f ∈ C ∗ and m ∈ M . using 4. m1 im0 = hf. (i) Let M be a right comodule for the coalgebra C. m1 ∈ C. if g. where m ∈ M. Then M is a left module for C ∗ . Let f ∈ C ∗ and define X f ·m= hf. (i) Let ρ : M → M ⊗ C be the comodule map. m1 im0 = hg. A morphism of right comodules f : M → N over C is a linear map such that f M /N ρM ρN ² ² M ⊗C / N ⊗C f ⊗id commutes. m1 im0 . P And ε · m = ε(m1 )m0 = m by 4. m0 ∈ M.2(2). m1 im0 . m1 i for f (m1 ). Proposition.2. whereas. Proof. where we write hf. we have ³X ´ X g · (f · m) = g · hf.

To see this. because if M is a right Ao -comodule via P ρ(m) = m0 ⊗ m1 . aim. These are (respectively) the transposes of right and left multiplication in C ∗ . For m ∈ M . Proposition 4. M M ˆ ). . and f (I) = 0 for all i = 1. where ρˆ(m)(a) ˆ := a · m. i=1 One checks that (8) is independent of the choices made and makes M into a right Ao -comodule. . . ¤ Definition. so ρ in (8) is the same as ρˆ and is thus well-defined. bai. which is the transpose of right multiplication on A.→ Homk (A. we can define a left action of A on A∗ . 4. . then A · m is spanned by the {m0 }. ˆ And ˆ ⊗ Ao ⊆ Homk (A. choose a k-basis m1 . via (m⊗a ˆ o )(a) = hao . C = kG. . . . . A k-vector space. (2) Let G be a group. so C is a right C-comodule by (1). (Ex- ercise 4. An A-module M with dimk Am < ∞ for all m ∈ M is called rational. we have a well- defined linear map ˆ . c ∈ C. . The right hit action is c ( f = Σhf. . That is. for a ∈ A and f ∈ A∗ . then C is a right (or left) comodule. for f ∈ C ∗ . n. (Exercise 4. (4) Analogously to (3). (3) Let C be a coalgebra. n by the rule X n a·m= fi (a)mi . . Examples of modules and comodules. M is a right kG-module if and only if M is a G-graded vector space. note that for any left A-module Mˆ .3(1) tells us that C is a left C ∗ -module with. . This is often called the left hit action of C ∗ on C. c1 ic2 . a * f is defined for b ∈ A by ha * f.4. (1) If C is any coalgebra. ⇒: This follows from what we’ve done. . M ρˆ : M ˆ ). So fi ∈ Ao for all i. mn for Am. Define a right coaction ρ : M → M ⊗ Ao by n X (8) ρ(m) = mi ⊗ fi . f * c = Σhf. m2 . bi = hf.6 (2)). using ∆. ˆ The condition that dimk A·m < ∞ for all m ∈ M ˆ simply ensures that im ρˆ ⊆ M ˆ ⊗ Ao . if A is any k-algebra. i=1 The kernel I of the algebra homomorphism A → Endk (Am) has finite codimension in A. HOPF ALGEBRAS 15 (ii) ⇐: Suppose M is a left A-module such that dimk (Am) < ∞.6(3)). Define fi ∈ A∗ for i = 1. for all m ∈ M . c2 ic1 .

and V ⊗k ∼ =V ∼ = k ⊗ V. 4. k (given by a. Then V ⊗ W is a left A-module via X a. with X (9) ∆pg = pu ⊗ pv u. Also we have the trivial left A-module. Ao is a coalgebra. bi = hf1 .(v ⊗ w) = a1 v ⊗ a2 w. . this will fail in general. By (3. bai = Σhf1 . as k-algebras.h v. then coassociativity ensures that (V ⊗ W ) ⊗ X ∼ = V ⊗ (W ⊗ X). so ∆f makes sense for f and ha * f. ai. If A is cocommutative then obviously V ⊗W ∼ =W ⊗V as left A-modules. and let V and W be left A-modules.6 (still to be proved). So a * f = Σhf2 . Tensor products and hom spaces. h ∈ G. and form Vg ⊗ Vh .5.v uv=g A has exactly |G| simple modules {Vg ∼ = k : g ∈ G} where for g.7) ⊕ X (kG)∗ = kpg g∈G = k ⊕ ··· ⊕ k (|G| copies of k). Now take g. Then by Theorem 3. bihf2 . aif1 . Example.16 HOPF ALGEBRAS Similarly for f ( a.v = ε(a)v for a ∈ A and v ∈ k). with the isomorphism given by the flip τ : v ⊗ w 7→ w ⊗ v. Suppose f ∈ Ao . h ∈ G and v ∈ Vg ph v = δg. But if A is not cocommutative. as left A-modules. (1) Tensor products of modules: Let A be a bialgebra. Let A = (kG)∗ where G is any non-abelian finite group. using (9) we see that Vg ⊗ Vh ∼ = Vgh so if gh 6= hg then Vg ⊗ Vh  Vh ⊗ Vg . If X is a third module for A.

Let (A. m. (ii) B a bialgebra. For this we need Lemma 3.1. Ii = hf.6 (6)). f ∈ A∗ . u) be a k-algebra. Check that this works. ha * f. ⇒ B o is a bialgebra (iii) H a Hopf algebra ⇒ H o is a Hopf algebra. which said (i) (A. Ii = 0. dimk (A/I) < ∞. (3) Tensor products of comodules: If B is a bialgebra and V and W are right B-comodules. Iai ⊆ hf.f )(v) = h1 f ((Sh2 )v) for h ∈ H. (Sketch) Recall (a * f )(b) = f (ba). f ∈ Homk (V. (f ( a)(b) = f (ab) for a. . 5. If B is commutative then V ⊗ W ∼ = W ⊗ V as comodules. An important case is when W is the trivial module k. HOPF ALGEBRAS 17 (2) Homomorphisms of modules: Let H be a Hopf algebra and let V and W be left H-modules. The Hopf dual 5.f = f (S(h)). I Cr A. (d) dimk (A * f ) < ∞. ε = u∗ ) is a coalgebra. b ∈ A.6(7)). m∗ . dimk (A/I) < ∞. then V ⊗ W is a right comodule via X v ⊗ w 7→ v0 ⊗ w0 ⊗ v1 w1 . u) an algebra ⇒ (Ao . For all a ∈ A. with hf. The key step is (i) and here we need to show that if f ∈ Ao then m∗ f ∈ Ao ⊗ Ao where m∗ f (x ⊗ y) = f (xy) for x. Then Homk (V. dimk (A/I) < ∞. namely h. (g) m∗ f ∈ A∗ ⊗ A∗ . (b) f (I) = 0 for some I Cl A. W ) (Exercise 4. f ∈ A∗ . Proof. (a)⇒(d): Supppose (a). Thus A * f ⊆ (A/I)∗ and so (d) holds. W ) is a left H-module with action X (h. Ii = 0. In Chapter 5 we prove Theorem 3. (e) dimk (f ( A) < ∞. via the flip map (Ex 4. y ∈ A. which yields a left H-module structure on V ∗ . Then the following are equivalent: (a) f (I) = 0 for some I Cr A. (f ) dimk (A * f ( A) < ∞. m. (c) f (I) = 0 for some I C A.6.

(b)⇒(c): Let f (I) = 0. b ⊗ ai = hf. If a ∈ A and b ∈ I then hA * f. bai = haA * f. i=1 Pn (g)⇒(d): If (g) holds. We prove that m∗ Ao ⊆ Ao ⊗ Ao .6(i). gn } for A * f. Let f ∈ Ao so there exists I C A with dimk (A/I) < ∞ such that f (I) = 0. Similarly (a)⇒(c). u) is an algebra. Proof of 3. . hi ∈ A∗ with m∗ f = i=1 gi ⊗ hi . Now for b ∈ A. So. Assume (A. (d)⇒(g): Suppose (d) and fix a basis {g1 . so (a)-(f) are equivalent. Thus f (J) = 0 and so (c) holds. This will prove that Ao is a coalgebra. i=1 Therefore n X m∗ f = gi ⊗ hi ∈ A∗ ⊗ A∗ . for a ∈ A.18 HOPF ALGEBRAS (d)⇒(a): Suppose dimk (A * f ) < ∞. Then I ⊗A+A⊗I is an ideal of A ⊗ A and it has finite codimension. because it’s the kernel of the algebra homomorphism π : A ⊗ A → A/I ⊗ A/I. Set I = {b ∈ A : hA * f. ¤ 5. . bi Xn = hi (a)gi (b). bi ⊆ hA * f. then m∗ f (α) = 0. Since f (I) = 0 we are done. hm∗ f. i=1 where one checks easily that hi ∈ A∗ . . m. I Cl A. n X a*f = hi (a)gi . . . 1 ≤ i ≤ n. There is an algebra homomor- phism ρ : A → Endk (A/I) whose kernel J has finite dimension since dimk Endk (A/I) < ∞. Since I is the kernel of the restriction map from A to (A * f )∗ . there exist gi .2. bai = ha * f. bi = 0}. (b)⇔(e) and (c)⇔(f) are similar. n X n X a*f = hi (a)gi ∈ kgi i=1 i=1 so dimk (A * f ) < ∞. (c)⇒(a) and (c)⇒(b) are trivial. Now if α is any element of I ⊗ A + A ⊗ I. bi = 0 So I Cr A. dimk (A/I) < ∞. The calculation just done shows that for all a ∈ A. And J = {a ∈ A : aA ⊂ I} ⊆ {a ∈ A : a1 ∈ I} = I. dimk (A/I) < ∞.

we can write m∗ f as i=1 gi ⊗ hi . just proved. so that Lemma 5. where gi . y2 x2 i since ∆ is an algebra hom. Proof of 3. 5. u∗ ) is a coalgebra. (B ∗ .1 ensures (12) dimk (B * f ) < ∞. u. yxi X = hf. y1 x1 ihg. and hx * f g. y ∈ B. g ∈ B o . Thus m∗ f ∈ Ao ⊗ Ao . ε) is a bialgebra. for all i. y2 i X (13) = h(x1 * f )(x2 * g).3. y1 ihx2 * g. yi. ∆∗ . . by Theorem 3. hi ∈ A∗ and I ⊆ ker gi and I ⊆ ker hi . Indeed m∗ f factors through A/I ⊗ A/I: m∗ f A ⊗ AL :/ k . Now coassociativity follows by restriction of the commutative diagram m∗ A∗ / (A ⊗ A)∗ m∗ (id⊗m)∗ ² (m⊗id) ∗ ² (A ⊗ A)∗ / (A ⊗ A ⊗ A)∗ which is itself commutative because it’s the dual of associativity in A. We claim: (11) B o is a subalgebra of B ∗ . (yx)1 ihg. Suppose first that B = (B. So (12) and (13) imply that dimk (B * f g) ≤ dimk (B * f )(B * f ) < ∞.2) Corollary B. Well. take f.(iii).6(i). (yx)2 i X = hf. And one easily checks that ε = u∗ satisfies the counit property. And (10) (B o . LLL uuu LLL u uu π LLL uuum∗f & u A/I ⊗ A/I Pn Otherwise put. m∗ . ∆. X = hx1 * f. m.6 (ii). By (3. dimk (B * g) < ∞. HOPF ALGEBRAS 19 This shows that m∗ f ∈ (A ⊗ A)o . Then f g ∈ B ∗ (convolution product). yi = hf g. Now let x. ε∗ ) is an algebra.

4(3)). or An for short. Notice that if hT i is the ideal of Rn generated by T . we have to check that S∗H o ⊆ H o. . as follows easily by dualizing the corresponding property for S and id|H (Exercise 5. S(b)i = hS ∗ (f ( S(a)). let a ∈ H and f ∈ H ∗ . an ) : ai ∈ k} is called affine n-space. ha * S ∗ f. then Z(T ) = Z(hT i). 6. . Lemma 5. Primer on affine algebraic geometry In order to discuss Hopf algebras arising as the coordinate rings of algebraic groups. Xn ] we write Z(T ) = {a ∈ An : f (a) = 0 ∀ f ∈ T }. . Then. ε∗ ) is an algebra. S ∗ is the convolution inverse for idH o . Algebraic sets. S(a)S(b)i = hf ( S(a). . Now (ii) follows from (10) and (14) after we check the bialgebra commutative dia- grams. bi = hS ∗ f. . Well. Finally. Also. ∆∗ . we need to recall some basic definitions and ideas from algebraic geometry. Throughout. 6. So S ∗ : H ∗ → H ∗ exists. . Elements of An are points. for b ∈ H. For a positive integer n.20 HOPF ALGEBRAS proving (11). since dimk (B/ ker ε) = 1.4(4)). so if f ∈ H o .1 gives dimk (f ( H) < ∞. bai = hf. . Thus H * S ∗ f = S ∗ (f ( SH) ⊆ S ∗ (f ( H). . S(ba)i = hf. the ai are coordinates. bi. ε ∈ B o .1 again ensures S∗f ∈ H o. The objects of study are the sets of zeros in An of collections of polynomials. k will be an algebraically closed field. X2 . So. Suppose finally that S is an antipode for B = H. for T ⊆ Rn := k[X1 . these follow at once by dualizing the ones for B (Exercise 5. the set k n = {(a1 . and so Lemma 5. We do this in this chapter. denoted Ak n . So (14) (B o .1. .

f = c(X1 −a1 ) · · · (Xn − an ). i=1 so Z(T ) = Z({f1 .I. Algebraic subsets of A1 : R1 = k[X] is a P. We are aiming for a bijection between closed subsets of An and ideals of Rn . so every closed set in A1 is the set of zeros of a single polynomial f . (ii) Let U = Z(T1 ).for instance . . . ai ∈ k. (iv) φ and An are algebraic sets. we make the Definition.the ideals hXi and hX 2 i of k[X] = R both define the set {0}. In view of the above lemma.] (iii) Let Uj = Z(Tj ) for j ∈ I. then Z(T2 ) ⊆ Z(T1 ). Thus any two non-empty open sets have non-empty intersection.3. . j j n (iv) φ = Z(1). The Zariski topology.D. . fr of polynomials in T such that Xr hT i = fi R n . Proof.. . To fix this. An algebraic subset of An is a subset Y ⊆ An such that Y = Z(T ) for some set T of polynomials in Rn . HOPF ALGEBRAS 21 By the Hilbert Basis Theorem. where T1 T2 = {f g : f ∈ T1 . In particular A1 is not Hausdorff. then J ⊆ I. But this fails since . we make the Definition. W = Z(T2 ). there is a finite set f1 . (i) If T1 ⊆ T2 ⊆ Rn .2. Since k = k. fr }). (i) is obvious. . . (ii) The union of two algebraic sets in An is an algebraic set. 6. Thus the algebraic sets in A1 are φ. Example. The gymnastics of algebraic sets starts with Lemma. (iii) The intersection of any collection of algebraic subsets of An is an algebraic set. g ∈ T2 }. Ideals versus algebraic sets (I). . . An ideal I of a ring R is semiprime if whenever an ideal J of R satisfies J 2 ⊆ I. A1 and all finite subsets of A1 . Then \ ³[ ´ Uj = Z Tj . ¤ 6. A = Z({0}). Define the Zariski topology on An by taking the closed sets to be the algebraic sets. Definition. Then one checks that U ∪ W = Z(T1 T2 ).

then I(U ∪ W ) = I(U ) ∪ I(W ). (iii) If U. √ (v) If I C Rn . for a set Y of points in An . (ii) If Y1 ⊆ Y2 ⊆ An . W ⊆ An . (i) Y ⊆ An ⇒ I(Y ) C Rn . Proof. √ √ For Z( I) ⊆ Z(I) and if a ∈ Z(I) and g ∈ I. We have the following obvious facts: Lemma. If I is any ideal of a commutative ring R. and so g(a) = 0. then Y ⊆ Z(I(Y )). (ii) Radical is sometimes used rather than semiprime. Easy. Let n ≥ 1. in fact I(Y ) is a semiprime ideal of Rn (using our discussion above). √ √ √ Then I C R. Hence a ∈ Z( I).3(iv). Moreover. (i) If R is commutative. we need Theorem (Hilbert’s Nullstellensatz). then I ⊆ I(Z(I)). then I C R is semiprime ⇔ ∀f ∈ R. Z(I) = Z( I). (iv) If Y ⊆ An . some t ≥ 1}. (15) Y ⊆ Z(I(Y )). Suppose f ∈ √ I(Z(I)). √ for some t ≥ 1. we define √ I = {g ∈ R : g t ∈ I. ∀ t ≥ t 1. ¤ 6. If I is a semiprime ideal. then I(Y2 ) ⊆ I(Y1 ). Proof.4. with I ⊆ I. Now define. Corollary. . f ∈ I only if f ∈ I. By 6. (v) of the lemma is an equality and (iv) is also an equality if we restrict to closed sets Y .22 HOPF ALGEBRAS Remarks. Then f ∈ I . Let Y be an algebraic subset of An .3(v) and the Nullstellensatz. In fact. Let I C Rn for some n ≥ 1. the subset I(Y ) = {f ∈ Rn : f (y) = 0 ∀y ∈ Y }. The correspondences I(−) n / {algebraic subsets of A } o {semiprime ideals of Rn } Z(−) are 1-1 and order reversing. if R = Rn for some n ≥ 1. f t ∈ I for some t ≥ 1. Ideals versus algebraic sets (II). To see this. then g t (a) = g(a)t = 0. then I(Z(I)) = I by Lemma 6.that is.

. which equals I by Corollary 1. ⇐ Easy . . . ⇒ Let I C Rn . . ←→     Y1 ∪ · · · ∪ Yr fi not an associate of fj for i 6= j There are also. HOPF ALGEBRAS 23 But Y is algebraic. Proof. Then J ⊆ I(Y ). The maximal ideals of Rn correspond bijectively to the points of An . Conversely. Namely. an ) o ma = hX1 − a1 . some n ≥ 1. . . Thus (15) and (16) give equality. a2 . Apply the Nullstellensatz to m: we get I(Z(m)) = m. .F. . Therefore.1(i) implies (16) Y = Z(J) ⊇ Z(I(Y )). the algebraic sets we get by taking the union of finitely many curves with a finite set of points. ¤ Corollary. a field. fi irreducible. / a = (a1 . Since m is a maximal ideal. . this says in particular that each point of An is closed. so Y = Z(J) for some ideal J of Rn . ¤ Otherwise put. I ⊆ m} = {ma : a ∈ Z(I)} by Corollary 2. Let I C Rn . Tt ←→ {a1 . So J = I. we conclude m = ma . But J = I(Z(I)). at } i=1 mai C[X. . which is a U.D. . an ) ∈ Z(m). hX1 − a1 . Xn − an i Proof. . of course. Then I is semiprime ⇔ I is the intersection of the maximal ideals in which it is contained. . I semiprime.Exercise (6. . Clearly. a2 . let m be any maximal ideal of Rn . . )(5). since Rn /ma ∼ = k. ¤ Example. Two typical examples of (unions of) curves in A2C are shown overleaf. We have: semiprime ideals of C[X. so there exists a = (a1 . . every ideal of the form ma is maximal. so Lemma 6. Let n ≥ 1. Xn − an i = I(a) ⊇ I(Z(m)) = m. Y ] ←→ φ  Qn    < f >: f = i=1 fi . Y ]. Corollary. . . and set \ \ J= {m C Rn : m maximal.   finite ∪ s of irreducible curves. . So Z(m) cannot be the empty set in An . . The algebraic subsets of AC 2 : Here R2 = C[X. . Y ] ←→ algebraic subsets of A2C {0} ←→ A2C ( ) finite intersections of maximal ideals.

DCC on closed sets). we need:- Definition. we need the Definition. but Z(XY ) appears to have two “components”.24 HOPF ALGEBRAS 1.11)(7)). Irreducible components versus prime ideals. (ii) Any non-empty open subset of an irreducible space is irreducible and dense (Exercise (6. . then Y = Y1 or Y = Y2 . A topological space X is noetherian if it has ACC on open sets (equiv- alently. (X 2 − Y 3 ) only one.11)(6)) A Zorn’s lemma argument shows that any topological space is the union of its maximal irreducible subspaces (which are always closed by Ex (6. To make this union finite. “Y can’t be expressed as the union of 2 proper closed subsets”] Examples.e. (i) A1 is irreducible since every proper closed subset is finite. [i. A non-empty subset Y of a topological space X is irreducible if. 2. f = Y 2 − X 3 6. f = XY 2. To make this precise. when- ever Y = Y1 ∪ Y2 with Yi closed in Y . Comparing the two exam- ples above: both are connected. i = 1.5.

An ideal P of a ring R is prime if and only if.g. Clearly. So if y ∈ Y . for all ideals A and B of R. Let A be the collection of finite unions of closed irreducible subsets of X. . That is. Proof. i so Y = Y ∩ Xi for some i. ∪ Xn . Y ⊆ Xi . A noetherian topological space X has only finitely many maximal irreducible subspaces. An algebraic subset Y in An is irreducible if and only if I(Y ) is a prime ideal of Rn . Then use the noetherian property to find a closed subset Y of X. f (y) = 0 or g(y) = 0. Minimality of Y forces each Yi to be in A. Xn ] which we get from Hilbert’s basis theorem. but then Y ∈ A. ⇐: Suppose Y = A∪B. So if R is commutative then P is prime if and only if R/P is an integral domain. Hence Y = (Y ∩ Z(f )) ∪ (Y ∩ Z(g)) is a union of closed sets. minimal among closed sets not in A. and so Y ⊆ Z(f ) say. . Proof. Lemma. B proper closed subsets of Y . Let Y be any maximal closed irreducible subspace. ø ∈ A. A. Xi irreducible closed subspaces. These are closed and their union in X. a contradiction. Suppose X ∈ / A. ¤ The maximal irreducible subspaces of the noetherian space X are called irre- ducible components. . AB ⊆ P ⇒ A ⊆ P or B ⊆ P . Write X = X1 ∪ X2 ∪ . Y is neither ø nor irreducible. g ∈ Rn with f g ∈ I(Y ). and let f. . Thus there exist f. How do we recognise the irreducible algebraic sets algebraically? Definition. An with the Zariski topology is noetherian. Then [ Y = (Y ∩ Xi ). That is. Of course the same then applies to all algebraic sets. g ∈ Rn . HOPF ALGEBRAS 25 Example. Proposition. . Thus P prime ⇒ P semiprime. So Y irreducible forces Y to equal one of these sets. Y2 are proper closed subsets. ⇒: Suppose Y is irreducible. So Y = Y1 ∪ Y2 where Y1 . . So e. since DCC for closed subsets of An is equivalent to ACC for (semiprime) ideals of Rn = k[X1 . f ∈ I(Y ). since Y is irreducible. and then Y maximal forces Y = Xi .

11)(12)). say P1 . Definition (B). The (affine) coordinate ring of X (or ring of regular functions on X) is the factor algebra O(X) := Rn /I(X) of Rn . P2 = hY i. P2 . A morphism φ:X→Y is a map given by φ(x) = (ψ1 (x). [Hartshorne] or [Humphreys]. Let X ⊆ An be an affine algebraic set. . I = hXY i C k[X. Note that m = hX. ψ2 (x). Since our category consists of objects defined by polynomial equations. Y ].26 HOPF ALGEBRAS with f ∈ I(A). this statement is valid in any (not necessarily commutative) noetherian ring. . . . Let X ⊆ An . . . Y i is the only maximal ideal containing both P1 and P2 .6. An an affine algebraic variety is an irreducible algebraic subset of An for some n.but really the theory can and should be developed in a way which avoids specific embeddings of varieties in the ambient space An . then there are finitely many prime ideals of Rn which are minimal with respect to containing I. then P1 = hXi. Pt . Hence I(Y ) is not a prime ideal of Rn . Remarks. . (Exercise (6. Morphisms. Y ⊆ Am be algebraic sets.g. . m. . .See e. f ∈ / I(Y ) and g ∈ I(B). Definition (A). where ψi ∈ O(X) for i = 1.11)(11)). In fact. Remarks. For brevity and simplicity we adopt here a “naive” approach . 6. (i) Morphisms X → A1 are the same as polynomial functions on X. ψm (x)) for x ∈ X. . ¤ Notice that the above lemma and proposition translate to the ring theoretic statement: If I is any semiprime ideal of Rn . (i) O(X) is the image of Rn under the restriction map from An to X. (ii) A morphism φ : X → Y is continuous for the Zariski topologies on X and Y (Exercise (6.g. We can now state Definition. . and then I = P1 ∩ P2 ∩ · · · ∩ Pt e. we naturally define the morphism in the category to be polynomial maps. . g ∈ / I(Y ). (ii) If we give k the Zariski topology then O(X) consists of continuous maps from X to k.

Since τ ∗ : k[Y1 . . . . . . so that imτ ⊆ Y and taking τ : X → Y we get τ ∗ = θ as required. . ψ2 (x). . (ii) φ∗ is a k-algebra homomorphism. So φ∗ (f )(x) = f (ψ1 (x). . . we define its comorphism φ∗ : O(Y ) → O(X) given by φ∗ (f ) = f ◦ φ for f ∈ O(Y ). . Define θ(Y ˆ j )(= θ(yj )) =: ψj ∈ O(X) for j = 1. (iii) (id)∗ = id. . . . . ψm (x)) which is in O(X) as required. Ym ]. OO qq 8 π qqqq q qqq θˆ k[Y1 . . ψm (x)). Proof. Ym ] → O(X) : Yj 7→ Yj ◦ τ = ψj . Ym ] to O(X): θ O(Y ) / O(X) . . Now θ lifts to a homomorphism form k[Y1 . . (iv) (φ ◦ ψ)∗ = ψ ∗ ◦ φ∗ if ψ is a morphism of algebraic sets from Z to X. (i) φ∗ (f ) ∈ O(X) for f ∈ O(Y ). . (ii) Check the definitions. . ym i. i = i. . m. . . So O(Y ) = khy1 . . . m.11)(13)) (iii). . ¤ The proof of (v) shows that we can recover any morphism φ : X → Y from knowledge of its comorphism φ∗ . So in effect we have proved . . . . . . φ∗ (f ) = f ◦ X : X → k where f is the restriction to Y of fˆ ∈ k[Y1 . . (iv) Clear (v) Suppose that θ : O(Y ) → O(X) is a k-algebra homomorphism. Let π : k[Y1 . HOPF ALGEBRAS 27 If φ : X → Y is a morphism of algebraic sets. (v) Every k-algebra homomorphism form O(Y ) to O(X) arises as the comorphism of a morphism from X to Y . (Exercise (6. We record the key facts: Lemma. Ym ] ³ O(Y ) be the canonical epimorphism given by restriction of maps from Am . we see that ˆ τ ∗ = θ. . . . where Yi = π(Yi ). Ym ] where θˆ : Yi 7→ θ(yi ). . (i) By definition. Set τ : X → Am : x 7→ (ψ1 (x). . . . . .

. i. Definition. . say X ⊆ An and Y ⊆ Am . So the principal open sets form a basis for the Zariski topology on X. . (i) If G × G is made into an algebraic set as in (6. R = O(X)[f −1 ].28 HOPF ALGEBRAS Theorem. (1) 6. . Notice two important facts: • Every non-empty open subset of X is a finite union of principal open sets. Y = Xf . For if X = Z({fi ∈ k[T1 . Clearly. Principal open subsets. The functor X → O(X) defines a contravariant equivalence of cate- gories. 7. and τ : G → G : x 7→ x−1 . Then the cartesian product X × Y is again an algebraic set: first An × Am = An+m . .7. In fact. For. and X × Y is a closed subset of An+m . Algebraic groups. Let X be an affine variety. so it is O(Y ) for some affine variety Y . Um ]}) then X × Y = Z({fi } ∪ {gj }). Suppose that X and Y are algebraic sets. This follows from Hilbert’s basis theorem. .e. You should convince yourself that O(X × Y ) ∼ = O(X) ⊗k O(Y ). and let 0 6= f ∈ O(X). Algebraic groups and commutative Hopf algebras 7. . let R be the subalgebra of the quotient field of O(X) generated by O(X) and f −1 . R is an affine commutative domain. Tn ]}) and Y = Z({gj ∈ k[U1 .1. Xf is open in X. . because the space of maximal ideals of R is homeomorphic to Xf via the map Y → Xf : m 7→ m ∩ O(X). . y) 7→ xy. then G is an algebraic group over k. ( ) ( ) algebraic sets /k affine semiprime commutative ←→ and their morphisms k-algebras and algebra homomorphisms 6. . U1 . we have maps m : G × G → G : (x. . Um ]. . . (ii) A morphism of algebraic groups is a morphism of algebraic sets which is also a group homomorphism. where we think of fi and gj as in k[T1 . . Product varieties. . The principal open subset of X defined by f is Xf := {x ∈ X : f (x) 6= 0}. . • Xf is itself an affine algebraic variety. Tn .7) and m and τ are morphisms of algebraic sets. Let G be an algebraic set which also has the structure of a group. . That is. Since R∼= O(X)[Y ]/ < Y f − 1 >.8.

we define ∆ : O(G) → O(G) ⊗ O(G) ∼ = O(G × G) by defining ∆(f ) to be the function from G × G to k given by ∆(f )((x.8) has coordinate ring O(GLn (k)) = O(Mn (k))[det(Xij )−1 ] = k[Xij : 1 ≤ i. . it can be shown that (i) and (ii) are the only irreducible one-dimensional algebraic groups over k. So by (6. namely An . y) = x + y. D(n. T (n.8). U (n. Mn (k) is clearly a variety. Let G be an algebraic group over k.5)]. Then SLn (k) is the closed subset of Mn (k) defined as the zeros of the single poly- nomial det(Xij ) − 1. j ≤ n][det(Xij )−1 ]. with the multiplication from k. y)) = f (xy). So m(x. τ (x) = −x. for instance. O(Gm (k)) = k[X. This is clear because every finite group has a faithful k−linear representation . The formulae for matrix multiplication and inversion now make it clear that SLn (k) is an algebraic group. Hopf algebras. and it means that e.g. It’s almost obvious that O(G) is a Hopf algebra if we transpose the multiplication and inverse operations from G to O(G). k) = {strictly upper triangular matrices in GLn (k)}. In fact. Y ]/ < XY = 1 > . (iii) The special linear groups: For n ≥ 1. as we have seen in Chapter 6.the regular representation. recalling (6. That is. so its coordinate ring is O(Mn (k)) = k[Xij : 1 ≤ i.7)(1). X −1 ] ∼ = k[X. and finite subsets of algebraic sets are algebraic. HOPF ALGEBRAS 29 Examples (i) The additive group Ga (k) is k = A1 with the usual addition in k. (vi) All finite groups are algebraic (over any algebraically closed field k). GLn (k) = {A ∈ Mn (k) : det(A) 6= 0} is a principal open subset in Mn (k) and hence by (6. (v) Any closed subgroup of an algebraic group is an algebraic group. (ii) The multiplicative group Gm (k) is just the open subset k \ {0} of A1 . (iv) The general linear groups: Let n ≥ 1. k) = {diagonal matrices in GLn (k)} are all algebraic groups.2. k) = {upper triangular matrices in GLn (k)}. For. but this is nontrivial [Humphreys. j ≤ n]. Theorem (20. 7. SLn (k) = {A ∈ Mn (k) : det(A) = 1} 2 is an algebraic group. This is clear.

let S : O(G) → O(G) be given by (Sf )(x) = f (x−1 ). [I’m calling the map evmx to stand for “evaluation at x”. we have. Moreover. b are k-algebra homomorphisms. Conversely. The functor G → O(G) defines a contravariant equivalence of categories   ( )   affine algebraic groups over k  affine commutative semiprime  ←→ Hopf k − algebras and and their morphisms     Hopf algebra morphisms . as a specialisation of Theorem 6. and ε : O(G) → k is given by f 7→ f (1G ). since under our corre- spondence X ←→ MaxspecO(X). define ab := (a ⊗ b) ◦ ∆. we have x ←→ mx = {g ∈ O(X) : g(x) = 0}. if π : O(X) → k is a k-algebra homomorphism. Summing up. so that X is an algebraic group as claimed. For.6 Theorem. made into a topological space with the Zariski topology. Finally. if O(X) is the coordinate ring of an algebraic set X and O(X) is a Hopf algebra. and a−1 := a ◦ S yields an inverse for a.] Conversely. then X is in fact an algebraic group. π is determined by its kernel. Now every such maximal ideal mx determines a homomorphism of k-algebras evmx : O(X) → k : f 7→ f + mx . Thus X identifies with the set of k-algebra homomorphisms from O(X) to k. so ε : O(X) → k is an identity element for this multiplication. and so f + mx is just f (x). recall that we can and should think of X as the set Maxspec(O(X)) of maximal ideals of O(X). It is a routine exercise to check that all the Hopf algebra axioms are satisfied. it’s easy to see that these operations are continuous maps for the Zariski topology. if a. and this is a group: namely. y ∈ G.30 HOPF ALGEBRAS for x. a maximal ideal.

so that ∆(X) = X ⊗ X. k=1 Notice that this is not cocommutative. (ii) Gm (k) = k \ 0 ⊆ A1 . Finally S(X)(a) = X(a−1 ) = a−1 = (X(a))−1 = X −1 (a). (i) Ga (k) = k. O = k[X. . and ∆(X) = X ⊗ 1 + 1 ⊗ X. while ∆(X)(a ⊗ b) = X(ab) = ab. noting (6. HOPF ALGEBRAS 31 7.8).so. so that S(X) = −X. so that S(X) = X −1 . (iv) First. Examples. Finally. S(X)(a) = X(−a) = −a. with ε(Xij ) = Xij (In ) = δij . so O = k[X] with X(a) = a for a ∈ k. Thus ε(X) = X(0) = 0. O(Mn (k)) is a bialgebra. so that n X ∆(Xij ) = Xik ⊗ Xkj . X −1 ]. We should note what the Hopf maps are for some of our examples in (7. and ∆(Xij )((A.1). B)) = (ij)th entry of AB. (iii).3. And ε(X) = X(1) = 1. For GLn (k) and SLn (k) the antipode is given by the formula for matrix inverse: (SXij )(A) = Xij (A−1 ) = (ij)th entry of A−1 .

Theorem (3. Theorem (3.3)]. we can extend Theorem (7. 1. let G be any non-trivial finite p−group. . (that is. If 1 6= x ∈ G with xp = 1. every affine algebraic group scheme over k is a closed subgroup of GLn (k) for some k. this says: every affine algebraic group scheme over k is the inverse limit of algebraic affine group schemes.3) is GLn (k).g. then (x − 1)p = 0 in H. The restriction to affine k−algebras in the theorem is not very important: every commutative Hopf algebra is a union of affine Hopf algebras [Waterhouse.2).4)]. Expressed group-theoretically. (Cartier) let k be a field of characteristic 0. see e. For a proof. 2. 3. see [Waterhouse. 4.2): Theorem. Supplements to Theorem (7.for the details. This is [Waterhouse.4]. and let H be the group algebra kG. Theorem 11. In characteristic 0. Indeed. H has no non-zero nilpotent elements). In fact. The key example in (7.4. Then H is semiprime.2) to all commutative Hopf algebras in arbitrary characteristic by replacing “group” by “affine group scheme” . [Waterhouse]. Note that this is false in positive characteristic: for example let k have charac- teristic P > 0.32 HOPF ALGEBRAS 7. the adjective “semiprime” can be removed from Theorem (7. and let H be a commutative Hopf k−algebra.