Cocommutative Hopf Algebras with Antipode by ... - Semantic Scholar

Report 2 Downloads 23 Views
J-

Cocommutative Hopf Algebras with Antipode by Moss Eisenberg Sweedler

B.S.,

Massachusetts Institute of Technology

(1963) SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY

August, 1965

Signature of Author .

.-.

. .

Department of Mathematics, August 31, 1965

Certified by ....-..

.

Thesis Supervisqr

Accepted by .......... 0................................... Chairman, Departmental Committee on Graduate Students v/I

2.

Cocommutative Hopf Algebras

19

with Antipode by Moss Eisenberg Sweedler Submitted to the Department of Mathematics on August 31, 1965, in partial fulfillment of the requirements for the degree of Doctor of Philosophy. Abstract In the first chapter the preliminaries of the theory of Hopf algebras are presented. The notion and properties of the antipode are developed. An important filtration is induced in the Hopf algebra by its dual when the Hopf algebra is split. It is shown conilpotence and an algebraically closed field insure a Hopf algebra is split. The monoid of grouplike elements is studied. In the second chapter conditions for an algebra A -which is a comodule for a Hopf algebra H --to be of the form A 'E B ® H (linear isomorphism) are given. The dual situation is studied. The graded Hopf algebra associated with a split Hopf algebra decomposes in the above manner. Chapter III contains the cohomology theory of a commutative algebra which is a module for a cocommutative Hopf algebra. There is extension theory and specialization to the situation the Hopf algebra is a group algebra. Chapter IV is dual to chapter III. Chapter V is devoted to coconnected cocommutative Hopf algebras, mostly in characteristic p > 0 . There, the notion of divided powers is developed and shown to characterize the coalgebra structure of a class of Hopf algebras. The Hopf algebras are shown to be extensions of certain sub Hopf algebras by their primitive elements. Thesis Supervisor: Title:

Bertram Kostant Professor of Mathematics

3.

Contents

Introduction

4

............................................

7

Chapter I.

Preliminaries .......................... ..

Chapter II.

Decompositions ......................... . 36

Chapter III.

Cohomology ....................----.

. 47

Chapter IV.

Cohomology ..........................

. 87

Chapter V.

Cocommutative Coconnected Hopf Algebras. . 93

Bibliography . ......................... Index ........ ......................... Biography ....

... *0..........................-.....-

...... 0...... 000............

162

166

I

Introduction

A Hopf algebra as considered herein is simultaneously an algebra and a coalgebra where the algebra structure morphisms are morphisms of the coalgebra structure, (or vice versa).

This differs from the graded Hopf algebras

of [2] Milnor and Moore, except in characteristic 2.

The

problem is to determine the structure of cocommutative Hopf algebras. Our first approach lies in a cohomology theory.

We

have constructed abelian cohomology groups Hi(HA) , where

H

"C.H.A."

H(CH)i

is a Hopf algebra, H-module,

H-comodule.

C

0 < i E Z

,

A

an algebra which is a left

a coalgebra which is a right "C.H.A."

We then determine the structures of algebras

(coalgebras) which are extensions of

H

(C) by

B

(H)

This theory applies to the algebra structure and the coalgebra structure--separately--of coconnected cocommutative Hopf algebras.

We hope to develop an extension theory

where the extension is a Hopf algebra and is an extension of one Hopf algebra by another. Our cohomology theory gives the familiar group cohomology in case G .

If

Ar

H =p-(G)

the group algebra of the group

is the group of regular (invertible) elements

5.

of

A

then Hi(H,A)

Furthermore, if underlying field

A

Hi(G,Ar)

=

is a finite Galois extension of the

k and

G

is the Galois group of

then the isomorphism classes of extensions of by

A

A/k

H =F(G)

form a subgroup of the Brauer Group.

Kostant has shown--the results are unpublished--that a split cocommutative Hopf algebra with antipode is a smash product of a group algebra and a coconnected cocommutative Hoof algebra; and that in characteristic zero a coconnected cocommutative Hopf algebra is a universal enveloping algebra of the Lie algebra of primitive elements.

We present

proofs of these results and study coconnected cocommutative Hopf algebras in characteristic- p > 0

.

For a cer-

tain class (including all where the restricted Lie algebra of primitive elements is finite dimensional) of coconnected cocommutative Hopf algebras we are able to determine the coalgebra structure.

The coalgebra structure is described

in a generalization of the Poincare-Birkhoff-Witt theorem. We now outline the generalization. In characteristic zero let for the Lie algebra algebra,

L ,

a Hopf algebra. I=

LC U

(x

be an ordered basis

its universal enveloping

If

i = 0,1,2,...

6. then dx(

=

z i=o

x2=0n

The Poincare-Birkhoff-Witt theorem is equivalent to:

a i)

x

-

a 0 , we show the Hopf algebra obtained

by factoring out the ideal generated by the primitive elements is isomorphic to a sub Hopf algebra of the original Hopf algebra, when the vector space structure on the quotient is altered.

We also show the original Hopf algebra

is an extension--as an algebra and coalgebra--of the quotient by the primitively generated sub Hopf algebra.

Chapter I

The study of Hopf algebras is a self-dual theory. For this reason diagram notation is useful, as it makes dual definitions and proofs evident. For all time we fix the field for all vector spaces. over

k ,

n

If

k

which is the base

X1 ,...,Xn

are vector spaces

the permutation group on

a e

n-letters

we consider

a:

Often

a

Xe

-

X

->

x1@

-0 -@

x0

->x.

will be written

i ,...1

X

(1 ,...

...

ex . (& x,

in)

where

1,2,...n) ; in this case X@--

111n

OX

'''

0y

... e

X

1 If

X1

X2 =

n ,

@Xn

.

n is a left G-

module. If

X, Y

are vector spaces

a linear map from

X

to

"f: X -Y"

means

f

is

Y

Once we define a "right" object such as module or comodule, we consider the "left" object to be defined with the mirror definition.

Similarly for "left" objects.

8.

Algebras (A,m,-q) space over

k

is

an algebra (over k) where

,

m: A @ A -+ A

T:

A

is a vector

k -+ A , if

the follow-

ing diagrams are commutative.

Igm A®A A

3 A @A

-

1S

m®I

I) AA

II) -m

A A (1k

A AA

is equivalent to associativity.

II) k

) A

A®1@A

A

m

m

I)

Ti@I k(DA~-

is equivalent to

TI(1),

a

is an algebra where

(rI)

is a unit.

is the identity,

m

is the usual

multiplication. An algebra

A

AO A

is commutative if

m

(2,1)

A

is commutative.

Ae A If then

A

is an algebra, -+

A

x

-+

l@ f(x)

If where

A

and

B

vector spaces

f: X ->Y

1 E A

always denotes

'QA(l)

Y

kD @f: X

where for an algebra

X, Y

A ,

are algebras

A& B

is

an algebra

9.

(A

B) q

(A @ B)

(>

A

A

BOB

m A mB mA® B and

nA e B = k @ 11B f: A ->B

A(

B

AD k .

is a morphism of algebras if the following

diagrams are commutative:

A

A

A

ff

ff

f

k B

B

B

4

B -$A

k

A

If space

A M

will often be written

is an algebra a left

with a map

?P: A®D M -. *M

k

k

A

A-module is a vector satisfying:

k@I A (A

M

M

)A

M

> A (M

M A0 M

If

M, N

phism of left

'f M

are left

A-modules

A-modules if

f: M -+ N

is a mor-

I 10.

M

M-)

A

is commutative .

jf SI@ f N > N A@N & An augmentation eA : A -+ k

morphism

.

A

of an algebra

EA

is an algebra

An augmented algebra is

an algebra

with a fixed augmentation. If

A

right)

is ar augmented algebra

k

is

a left (or

A-module by: EI

N @A M

m 3 k

k

k

-+

N

A-module and

is a left

M

If

I

-

A0 k

a right

A-module

is a vector space such that *N

NO A(

I

M

-

(9 ?M ) NO M

-

N (AM

_+ 0

is an exact sequence of vector spaces. If left

is an augmented algebra,

A

A-module, then

A+ = Ker e A,

k ®AM = Coker (A+O

) A0

M

i@I or if

A+ - M denotes

Im(A+

) A

M

M

a M) , ?P

M - M)

i®I k

A M = M/(A+ - M)

Coalgebras (C,d,E) space over

is a coalgebra over

k ,

k

where

C

is a vector

11.

d: C

-+

k

s: C -

CtC

if the following diagrams are commutative: k

d

C

I

II)

II)

d C SC 0C

I

C (k

I)

C

d

dO I

d C®C

C

C O----C

C I)

C '

C@ C

E

is equivalent to coassociativity E

is equivalent to k

is an augmentation of a coalgebra. d = k® I

is a coalgebra where

and

e = the

identity.

A coalgebra is cocommutative if C@ C C

(2,1) C OC If

C

is commutative.

d X, Y

is a coalgebra and

are vector spaces

f: X - Y , then E

C D X If

C, D

f:

c

C®X

k

Y

are coalgebras

-

Y

® C @) D

aYc is a coalgebra where

12.

dC(

dD

>C(C(DeD

CQD

2D 4) (1,3,

dC (

D (CO D) 0

f: C -* D

C

eD

'C

ECOD

(CO D)

is a morphism of coalgebras if the diagrams,

CO C

C f

C

f

dD D'-

D

k

f

f

D

D

6D

are commutative. If space

C M

is a coalgebra, a right $: M -. M @ C

with a map

C-comodule is a vector satisfying: M

M

$C MO C

>MP C %

I

>MD C

()®d C I (

If

M, N

are right

morphism of right

C-comodules

C-comodules if

E

f: M -* N

is a

13.

M-)

m® C

If

N

,

fI

N

is commutative.

C

A unit of a coalgebra

C

is a coalgebra morphism

k - C. A coalgebra with unit is a coalgebra with a fixed unit. If left)

C

is a coalgebra with unit

If

O

is a right (or

C-comodule where k

M DN

k

N

N

-%

d

kk

is a left

I

TIC

C-comodule

M a right

C-comodule,

is a vector space such that

-+ M 0

N

-+ M(DN

)

M

A

N

is an exact sequence of vector spaces. If 7: C -+ J

C

is a coalgebra with unit then

M D

k = Ker (M

-

JC = Coker M

C

iC

'

) M®J

)

We shall freely use facts such as kernels and cokernels of comodule morphisms are subcomodules or quotient comodules, when the "dual " fact is well known. If

X, Y

are vector spaces

Hom(X,Y) = ff: X -+Y.

14.

Hom(X,Y)

is a vector space.

Often

Hom(X,Y)

additional structure, for example if and

A

an algebra then

Hom(C,A)

C

carries

is a coalgebra

has a natural algebra

structure as follows Hom(C,A)

C,A)

) Hom(C D

Hom(C,A)

®

a

Hom(CA))

where

Hom(C,A)

g) = mA o f

a(f

g

d: C -+ C & C .

Multiplication in

denoted by

thus

in

*

Hom(C,A)

f*

-+

.

Hom(C,A)

X*

g = mA o f(& g o d .

(T)A 0

-C

'

X

is a coalgebra

-c. (X

0

X)*

a as above. a is injective for if n SAxi* y * s @ ,We can assume i=1 linearly independent. .

X*

Consider X*@0 X*

> =

The unit

Hom(Xk) , as usual we consider

denote

We have just shown if

is an algebra.



= (dim X) 2 = dim(X® X)*

is an isomorphism.

a

case

Xj0

is injective.

a

which shows

Y,*) i*

induce transpose mappings:

A* -+ k* = k , d = tim: A* -+ (A

a finite dimensional coalgebra

A*

A)*

= A*@

A*

A

is

Similarly if

is a coalgebra.

A*

With these maps

@

is an algebra; however,

this is the same algebra structure as deduced above. If

X, Y

are vector spaces -+

a: X* ® Y*

Y)*

(x

x E X, y E Y



=

<x*,x> x* E X*, y* E Y*

a

is injective.

same proof works.) a

in

(X @ Y)*

coalgebra then

(We proved this, above, for Identify If

M

X*

@

is a left

Y*

Y = X , the

with its image under C-comodule,

C

a

16.

C*@ M* ,

) (C@ M)*

IPM*

defines left

4'M* ,

and this gives

C*-module.

algebra,

M*

Similarly if,

M a left

A

the structure of a is a finite dimensional

A-module by

AQ M then

t M

-M

tPM: M* -+ (A

M

M)*

comodule structure on

A*Q M*

defines a left

M*

Hopf Algebras

(H,m,i3,d,E)

is

1)

(H,m,TI)

is an algebra with augmentation

2)

(H,d,e)

is a coalgebra with unit

3)

the diagram

H

a Hopf algebra when

m -

H

d H

)H

d)d HOH

is commutative.

L

H

Tj

H mDm

H

(1,3,2,4)1 Hf

HO H@ H

e ,

17.

3)

and

"

is equivalent to

3)

and

(H,m,Tj)"

"E

is a unit for the coalgebra "d:H -*H ® H

(H,d,E)"

is an algebra morphism".

is an augmentation for the algebra

is equivalent to

"m: H

DH

-+ H

is a coalgebra

morphism".

Since a Hopf algebra Hom(H,H)

H

is a coalgebra and algebra

has the structure of an algebra.

algebra structure and if

H

H* has an

is finite dimensional

a coalgebra structure, as well.

In this case

H*

has

H* is a

Hopf algebra. In

Hom(H,H)

consider

T

o s

k = k - 1 C H

is the unit.

so

e

We shall often

will denote the unit in

Hom(H,H) .

The identity

I: H -+H

Hom(H,H) .

S E Hom(H,H)

is an antipode for

I * S = E = S * I , that is

S

is an element of H

if

is the inverse to

I.

Example:

Let

G

be a monoid (group without inverses).

the "monoid" algebra =

n

7(G)

L

Q) k - g g E G

m

n,m

j=1

i,j=l

is an algebra where

where

F(G)

=

18.

-ri: k

F(G)

a Hopf algebra where

is

F(G)

d:

-+

g s: F(G)

'(G) @ F(G)

-+

g@g

-

k

+1

g

r(G)

(G)

-+

has an antipode if G

Suppose

. and only if

G

S: F(G)

is a group define

is

F(G)

-

-1

g then

S

is an antipode.

If

S

a group.

is an antipode,

n e = 1 = I - S(g) S(g) = g~

implies

gS(g)

=

,

,

hence

We now show that if

H

X g1 which i=l has inverses and is a group.

where G

S(g)

d o S = (2,1) o S @ S o d

2)

S o m = m o S @ S o (2,1)

3)

E o S = S o

4)

If

=s

is commutative (as an algebra) or co-

commutative (as a coalgebra) then We prove

1)

inverse to

52

=

S o S

=

as follows:

(2,1) o S 0 S o d e Hom(H, H@ H)

which is

d e Hom(H, HO H)

is a left

d o S , an algebra.

S , then

has an antipode,

1)

H

=

d

and

we show

(2,1) o S 0 S o d

d o S

is a right inverse

19.

to

d

This proves they are equal.

.

following: f: C -+ D

If

C, D

are coalgebras,

a coalgebra morphism,

phism and (h o g

We shall use the

g1 , g 2 E Hom(D,A)

A, B

h: A -+ B

are algebras, an algebra mor-

then

o f) * (h o g 2 o f)

= h o (g1 * g 2 ) 0 f

The proof is clear from the definitions. Consider

(d o S)* d = d o(S * I)= d o e = k (d o e = d We let

because

m's

.

d(l) = 1 0 1 ) .

mn =mo m@lo---om@

by associativity

mn

e E

O ...

T

n -1 is independent of the order of the

Similarly we let

dn = d

T

o---o o d

T o d

n -1 dn

is independent of the order of the

tivity.

If

h e H

Z hi'@ h "

d's

by coassocia-

we often let - --

hi(n+l) = dnh

Then

)

d * [(2,1) o S Q S o d](h) = Z h 'S(h

hi "S(h '") I*s

=Z h

'S (h

'")

@

E)

=

20.

Z hi'S(hi")@ 1 = E(h) i

1

I * S= which verifies For

2)

1) . we show

m o S @ S o (2,1)

S o m

is left inverse to

m * [m a S® S o (2,1)](h&

Z h

For

and

4)

g) = Zh

") =

E(g)S(h

= e

E = I *

For

Then

3)

a o I * S

=

'g'

H

m a S ® S

=

S(g")S(h ")

(e o I) * (E a

o E = (I o E) * (S o e)

suppose

E .

E(g)hi' S(h") =

=

S)

(g)E(h) S o 5

=

*E

=

o S

* S o E = S

a

is commutative so m o S ( S o (2,1)

=

S o m

S * S2 (h) = Z S(h i)S2(h"') = z S(hi' S(hi"))

i

i =

which shows H

and

is right inverse.

(S o m)* M = (S * I) a m = E O m = 6

=

m

S2

is

is cocommutative

right inverse to

S

®S

o d

S * 32 (h) = Z S(hi')S 2 (hi") i = m o I

m o I®

S (Z [S(h)]

j

S ,

so

(2,1) o S

= =

S o e(h) = e(h) S2

®S

.

If

o d = d o S

S(h ')@ S(hi")

S( i

[S(h)]J")

.

21.

I * S o S(h) = s o S(h) = s(h) which shows

S2

is right inverse to

H

We have

of

H*

on

as an

S

H*-module

H** ; under this action

H

is a submodule.

We then concern ourselves with the action of H**

a left

h**= h e H,

Then if

1

H** D H , we define a left action

H , H*

Consider

S2

hence

on

H

H*-module by



dh =Z hi'

hi"

h =Z h ' e H C H**

b*.

i

-

This shows

H

the action of module.

is a submodule of H*

on

H

.

H**

We consider

(We could have begun with

=

Again

H*H*H

would be a submodule where h - a* = Z 2) 1=1 where g E G are distinct ki . By the minimality of

n

a* E H*

g 2 'g3 ' 'n where

are linearly independent.

= 1

,

= 0

Choose i=3,.. .n

26.

n Then

X1 -

-

iA =-2

1=2

n

n

0 = a* - 0 = a* -(

%g) X

g

=

1=1i=

= -A2 g1 + XA 2 92 = This implies

X2 = 0

minimality of

n

X2 (g2

g1 = g2

or

d(gh) = (dg)(dh) = (g ) g)(h

which contradicts the

g, h e G

Observe if

.

g1 )

-

QDh)

= gh(® gh

Thus

.

G

a monoid and since it is a linearly independent set in the space spanned by

G

T(G)

as an algebra.

,

is

H

We consider

T(G) C H . If

H

dS(g) = (2,1) o S 0 S o dg gS(g) so

S(g)

S , g e G

has an antipode

=

I

=

S(g)(D S(g)

*

S(g) = E(g) = 1

is the inverse to

g

in

Filtration of

G

H

is split.

so that

and

G

S(g) E G

; is a group.

H

We exhibit a natural filtration on exhaustive when

then

H

which is

The multiplication and dia-

gonal map respect the filtration as does an antipode when present.

The filtration leads to a graded Hopf algebra.

It is also an invaluable basis for induction. Within

H*

let

J = F(G)

.

It is a straight

forward verification that for any subcoalgebra

C C H

27.

Thus

C4

C C H @H)

(do CC® J

is a

We assume

il= (J1+1L

1)

H = UHi

2)

H H

3)

dH

let

c

Z

D: then

.

.

J i+l

a**b* E J +

b*,

x>

b*

x = 0

let

x

x> = 0

= 0

;

and we are done. H* 0

H*

O+ z E H 0 H

is

"dense " in

there is

In fact there is + 0 prcve

(H

z* E H*

a*@ b* E H* .

H)*

in

H*

that if

where

$ 0 where

Thus to

d(a*

.

h) = Z h i'

a* - h "

we need

28.

only show

d(a* - h)> = B

B T1

The left diagram means

B -+ B

m: B

morphism; the right diagram means H-module morphism.

B

H

A

Regn (C,A)

Regn(C,A) B

0 < n E Z

For a coalgebra

C n n n Hom (C,A) = Hom(C ® '-.-C,A)

B -.B

Regn (H,B)

Homo(C,A) = Hom(k,A) = A

Now we assume

@

let

,

H-module

regular elements of

s: H

is a commutative algebra

is a multiplicative subgroup of

We let

is a left

regular (invertible) elements of

=

is a left C.H.A.

group.

B

a cocommutative Hopf algebra.

and algebra Let

k -+ B

I:

H-module

is called commutative Hopf algebra

H-module (C.H.A. H-module) if and

is a left

A B

is a left C.H.A.

.

If

is an abelian and

Rego(C,A) =

H-module where

Regn(H,B)

Define

,n: Regn(HB)

Homn(C, A)

= Ar

is the module action.

is an abelian group.

Homn(C,A).

-+ Homn+1(HB)

0 < n E Z

49.

n

f(-)

on7

o (I)

---

I 0I -.m

)

~1=1 i - 1

n - i

(-1)n+l

where

n > 0 ; for

f E Regn(HB) h

b E Br

H

E

We shall let

6 0 b(h) = (h-b)b~

n = 0

F[a]

denote

a F

Q)

---

unit.

In

Q)F

n > 0 ,

Regn(H,B)

e .

This is often denoted

and given

to

morphism.

6 01 (h) = (h - 1)1

h - 1 = e(h)

H.A.

Regn+1(H,B)

6n

6n so

E(h)

is a group homo601 = E .

is one of the conditions for

B

Note

to be an

H-module, we now use the other:

6 0 (bc) (h) = (h - (bc))c~Ab~A =

Z (hi 'I - b)b~

(hi

=Z (hi i

For the case ,n(E)

- c)(c

1b 1

i

= [6 0 (b) * 60(c)](h)

.

n > 0 n T

?P 0

E ()i

n E *

E *

i=1

k

as a map from

and that =

on 6 n() = 6 n(f) * ,n(g)

* g)

These imply that we can consider Regn(H, B)

the

We shall show

6 n (f

f,g e Regn(H,B)

is

E[n]

E =

E

o

[i

[n-i

E

)

50.

Note

o I

&

e = e

because

-

and

(f

(1)~

* g

o

bc(-1)us [i-] 1 because

I

is an H.A.

H-module.

?P o I()(f * g)= (7P o I

This also guarantees (f * g)(~l)

B

g(-l)i

*

m

I

is clear

~ (f * g)(l)

and thus

6n(f *

g) =

coogb (-)

[1-1 p

=

*

(f(l)n+ll

I n- ]

m

is a coalgebra morphism.

E

6 n(f)

I[n-i]

[n-i] I

mi1 M0in-il I) 1

It

(-1)n+1

n(g)

We have the sequence of abelian groups and homomorphisms 0 Reg (H,1B)

60 ) Reg (H,B)

1

2

-' Reg (H,B) -

62

)

We shall now show 6 n+1 6 n

= E: Regn (H,B) -+Regn± (HB)

We do this by expanding -0

* term *

---

41 *

term *

6 n+1 6 nf

.

In our notation

expands to

42 term * term *

The explanations of the expansions follow the computation and a typical note is of the form: "texplanation".

g)

because the group is abelian

m

I

f) * (p o I

4 -+ 41 * 42 * 43:

51.

6 n+1 6 n

=

(where

n > 0)

1

o 1O(p

0

f-

r

1 @f

0(1[i-

®m

[1-i])

* f (_)

*

2

/"--, n+1

-----------------

i

n (

o I e f

*

o(I[i-e

f

m

1=1

j=1

I[n-l)

(-1) J f (-1)n+1

0

( 1 i]

0

m

0 1 [n+1-j])

3

"K * (*

o I

rf

0([1-110 Me

f 1=1

(-+1) _n2

*

11

=*0 I

1 n-i3)

12

(* o

) f)

o (11-ilomer i=1

13

*

4

I

f

-

))

52.

21

n+-1 j1:1

'i)

(*, oIOf(i

o -1 (I 0 M®@ I[n-1j])

22

-------------[1-]

*(-1)

[ -1

o (

1[J1]l(Dm®

I[n+1-j

*

I [n+l-i])

1D

H

PD=D-)D D

D

kE.

D

The diagram: P

P

I

E

T -- >D

B

IPD

T

)D DH

It can be rewritten

is commutative.

) D

T

m oP

k

P

PD H k.

JD

B

TID

H Since

Im P

Im TD = D

are left

DH k ,

D H~~

Clearly

EH k.

B-module morphisms

is a factoring

Q

m oP @ P

is surjective, it follows

I C Ker m o P (FP, IT C Ker m o P

,

T

Thus

T -+ D

D

@

since P ,

P,

P

or there

86.

(R@k B

)

-

T

D

)

B

is commutative. have shown

The horizontal composite is is

TrD

injective. 11D

B

VD

hence we

Thus

$D

D

D

-

)H

is a split exact sequence. The second condition for extensions is satisfied. Thus

D If

is an extension of

y: A -+ A, ,

H

by

y: A -+ A1

B

are equivalences of exten-

sions then: ,y() Y: A®() A

y y@ ~y|T

factors to

of extensions.

~yIT: T

A -T

yD: D-+ D

and

yD

is

an equivalence

Thus the method of producting extensions

gives the isomorphism classes of extensions a product struc-

f ture.

Direct calculation shows the product of

f

f B

H

is extension isomorphic to

B

B

H ,

f' H ,

so that

under the correspondence of theorem 3, the multiplicative structures correspond.

This shows the isomorphism classes

form a group and the correspondence is a group isomorphism.

pop--

87.

Chapter IV Cohomology

We treat the dual situation to the previous chapter. We outline the definition of abelian cohomology groups H(BH)i

i=0,1,2... , when

algebra,

B

H

is a commutative Hopf

a cocommutative coalgebra which is a right

H-comodule satisfying certain conditions.

We also outline

the dual extension theory, defining when a. coalgebra is an extension of

B

by

H

and when extensions are isomorphic.

The isomorphism classes of extension again correspond to H(O2 , we define the dual extension to B f) H of the H(B,H) previous chapter; this indicates the correspondence beH(B,H) 2

tween B

is

is

a right

and the isomorphism classes of extensions.

42.

and

d: B

-BO)B

s: B

->k , B

5: B B ®B

H.A.

is a right C.H.A.

H-comodule,

is a right H.A. H

B .

is a right

B

are comodule morphisms. B

a Hopf algebra

H-comodule structure for

H-comodule under

if

H

a coalgebra,

B

-* B(

H

is an

H-comodule if

H-comodule

is cocommutative

is commutative.

For a coalgebra

C

and an algebra

0 < n e Z

let

88.

Hom(C,A)n = Hom(C,A En)

Then C.H.A.

f

Reg(B,H)n

B

Let

B

B

is a

Reg(B,H)O = Reg(B,k).

be a right C.H.A.

Reg(BH)

E

Reg(C,A)n = (Hom(C,A)n)r

is an abelian group when

H-comodule.

Let

Let

.

H-comodule.

For

consider

4k@H= >B@H

H.

Define 60

Reg(B,H)

=

6 0f we are considering

-+ Reg(B,H) 1

f: B

-+

f * fI

k C H .

'I When

o

),

n > 0

define

6 n: Reg(B,H)n -+ Reg(B,H)n+l

6nf = k O f *

o fn+1 i *

I

d

I i=1

Io Then

6n+1an = E,

6

b

.

is a group homomorphism so we have

a complex 61

6n Reg(B,H) 0

Reg(B,H)

) -

and define the cohomology groups H(B,H)n

=

Ker 6 n/Im

H (B, H)O

=

Ker 6 0

6

n-1

n > 1

89.

As in the previous theory we have a normal complex: n > 0

= ff 6 Reg(B,H )n

Reg(B,H)

B

[n]

-

k

is commutative

Sn+ =

n|iReg(B,H)+n: Reg(B,H)+n -*Reg(B,H)+n+1

Reg(B,H)+ 0

Reg(B,H) 0

=

60+ = 60: Reg(B,H)+ 0

-+ Reg(B,H)+ 1

The cohomology computed by this complex is the same.

Extension Theory

H

is a Hopf algebra

A, B

H-module (as a coalgebra), structure.

called

A

a left

*: H(@ A -+ A , being the module

The sequence H(

is

coalgebras,

A

) A -- -B

,

right exact if

1)

r

2)

Coim r = k OH A

is a surjective coalgebra morphism

Ker Tr = H

i.e.

A

The sequence is called split exact when:

H

has an antipode,

90.

the sequence is right exact, and there is a regular left H-module morphism d A

)A

A

P: A -+ H

P g r -

H(

comodules and left give A

B

B

In this case

.

is an isomorphism of right

H-modules by theorem 2.

the structure of a right C.H.A.

is called an extension of

4

r

A

-+ B

B

by

H

H-comodule, then when:

HD @A

2)

the following diagram is commutative, d

A->

is a split exact sequence,

I@7r A@A

A@B

-

--4

d

A

$: B -+B

Let

1)

-+

B-

A

B

B

H

H

where the

A*

A ir~I Given two extensions

A, A'

of

B

by

split exact sequences are

4 H@ A

HwA'

-+

ri

Al -+

B

A'a

B ,

we say they are isomorphic if there is a coalgebra morphism 'y: A -+ A' ,

such that,

H

'1 91.

H@ 10

A

A

-

B

y

Y H@

)A'

Being isomorphic is an equivalence rela-

is commutative.

tion and we can form isomorphism classes of extensions. An example: Given

g e Reg(B,HH)

2

a

2-cocycle, form

H

®B

which

g H®@ B

is

H-module.

as a left

I

E

7r: H

B BB

'

g I@ P: H

E

B

0 H

.

g d@ g) g B

H

B

)H

$

I

B

B

g

(1,3,5,2,4,6,7) H 0H

H

H

B

H

B

H(

B

H

H

H

B

'I

m2 (

m H

(H

defines the coalgebra structure on

H

@ g

H

H

B

g

B

-H

g

B

B .

( B) g

Then

(H

@ B) g

92.

is split exact, the second condition for extensions is satisfied so

H a B is an extension of B by H . Two g such extensions are isomorphic if and only if the 2-cocycles are homologous.

There is such an extension in each isomor-

phism class of extensions. tive correspondence between classes of extensions.

This yields the natural bijecH(B,H) 2

and the isomorphism

93.

Chapter V Cocommutative, Coconnected Hopf Algebras

We have shown a split conilpotent Hopf algebra with antipode is a smash product

E

B (D H

where

Hopf algebra acting as automorphisms of split Hopf algebra where

G(B) = 1j.

B

H .

is a group B

is a

We shall study such

Hopf algebras which-are cocommutative.

In characteristic

zero we prove such a Hopf algebra is a universal enveloping algebra--u.e.a.--of a Lie algebra, in characteristic p > 0

it contains a restricted u.e.a.--r.u.e.a.--of a res-

tricted Lie algebra.

(We assume familiarity with univer-

sal enveloping algebras and restricted universal enveloping algebras as can be found in

[2]

Lie Algebras.)

We define divided powers as sequences of elements

1

=

I 0 '2l''''

N

n

n < N

din = Z and show that in a i=0 class of cocommutative split Hopf algebras H , where where for

G(H) =

[1

the coalgebra structure of

H

is characterized

by divided powers. A Hopf algebra G(H) =1j. phisms of

Since k

to

H G(H)

H

,

is coconnected if it is split and corresponds to the coalgebra mor-

coconnectivity is equivalent to:

94.

being split, and morphism of

k

TI:

k -+ H

is the unique coalgebra

H .

to

all coconnected Hopf algebras have antipodes.

By

All Hopf Algebras Henceforth Are Assumed To Be Coconnected Unless Otherwise Specified. A primitive element

x

in a Hopf algebra

H

is one

where dx Then

x E H1

=

xEl

+

1@ x

e(x) = 0 .

and

Let

under

[

,

] ; if characteristic =

dxp

(binomially expand

+

1@ x

L(H)

H .

space of primitive elements in

(dx)P), so

.

L

denote the

(=L)

forms a Lie algebra

p > 0 x

L

1

forms a restricted Lie

algebra. Suppose

x E H dx E H dx

I * E = I

E * E = E

I

=

,

E(x) = 0 . @ H1

+ H

10y

+

H0 , z

HO =

so

l E(y) +

imply

k

z

= X

+

E(z)

= x

E (y) +

E(Z)

=

y

E(x)

=

0

95.

dx = 1QD (y + e(z)) + (z + E(y))Y

Thus

and

x

is primitive.

U

and

L C U .

U

H

x&l,

= L(q H 0 = L(0 k

denote the u.e.a. of

and the r.u.e.a. of

E(Z)

This shows

Ker efH 1= L Let

+

l®x

=

1 - 19

L

in

L

in

char p > 0 .

.

char p =0 We consider

is a (coconnected) Hopf algebra where the co-

algebra structure is induced by specifying the elements of L

are primitive.

Then

L(U) = L .

By the defining property of

U

there is an algebra

morphism : U -H which is induced by the identification of L C H .

y

L C U

with

is also a coalgebra morphism; hence, is a mor-

phism of Hopf algebras. That

y is injective follows from the next lemma.

Lemma 4I:) bra, if

V

v: A -+ C

,

A

a Hopf algebra

a coalgebra morphism.

vjL(A) Proof:

v

C a coalge-

is injective if and only

is injective. Suppose

a coalgebra morphism

v|L(A)

is injective.

Since

V

is

96.

EC(V(L(A))) = 0 thus

vA 1

VIAn

is injective

is injective.

injective.

C(V(l)) = 1

,

By induction we can assume

a e A n+

Suppose

v(a) = 0

@A

VI An

is

then

da=l0a+a®l +Y, Suppose

V

n > 1 ; hence,

YEAn

A

.

then

®

0 = dv(a) = v @ voda = v(1)

v(a) + v(a)

) v(1) + (v

@

v)(y)

= (v&@v)(Y) . Since

v

VIA |A

n

is injective it follows

a E L(A) ; which contradicts

VIL

is injective.

The converse is obvious. This implies

'y: U -+ H

Q.E.D. is injective.

with its image which is the subalgebra of L .

(Thus

U

Y = 0,

is the subalgebra of

H

We identify H

U

generated by

generated by

L

both as algebra and coalgebra.) We shall show if

where

J = k

C H*

H

is cocommutative and

char p = 0

U = H ,

char p > 0

U = (H*p(J))£

,

p(J) =

the ideal generated by

p(J)

a*P|a* s J] .

and

(H*p(J))

We do this by developing

is

97.

a technique which "picks-out" subspaces of associate with subspaces of the polynomial algebra on (Note:

H , subspaces of

fxaJ a basis for

work with

f

k[ f xa

k[ I xJ ] L(H)

.

Given a vector space 9-.module.

n > 0

SnX =

n = 0

S0 X

xa]

x

we have chosen to

,

]--which makes the basis clear--rather L .)

than the symmetric algebra on

Let

We

.

since the method of associating subspaces with sub-

spaces depends upon the basis

left

H

X

n

is a

x =®X

,

The symmetric tensors of degree Y E X[n]aj=

Y

Y

=

are:

J

aE

k .

be an ordered basis for k[

all

n

x

X ,

is a graded algebra.

]I

CO

k[

x a]

SX =

is isomorphic to

e

SnX

as follows:

n=0

a Let (i.e.

(x 1 )1

e 1 +--+

a Let Let

---

Y =

1el x

em

(tm) em =n,


=

-..

L

H

except

a

ES



This proves:

0

h E Hn - Hn-l

For

h e H

if

0 := E[n] o dn-lh E SnL .

let

- H

K(h) =P 0oE

Then we have

and only if

n]o dn-l(h)

Kn

H-J/yJVK h

-+ K(h)

which is not linear. morphism we have,

But by

and since

0

is an iso-

100.

if

h E H

, g E H

and

K(h) - K(g) = 0 h -g

For a subspace K(X)

let

the subspace of

K(H) I K .

In general

Xn = X

Recall for a subspace K(Xn) = K(X) n (Kn a

Lemma 5:)

- --

0 + K(h)

fh

C Xn ci

n - 1 .

? 1K(h 1 ) +---

then it is clear

K0 )

X, Y

We show by induction

then

=

($

1

n Hn

K(X) = K(Y)

n = 0 , suppose true for

7 1(p o

K

Given subspaces

X 0 =YO ; then

Proof:

0

spanned by

K

S X.

(K(x)|x

X C Y,

and

H n-1*

E

X C H =

g e Hn

then

of

H

where X = Y.

implies

X D Yn . Let

True for

h e Yn -n-l'

+ ?mK(hm) ,

where

A

E k ,

Thus [n] o d nl(h )) + +

oo E[n] o d n-(?h

which implies

g = N h

+

+ Am(P o E[n] o dn-l(hM + Nmhm)

-

+

mhm

E Xn - Xn-1

and

101.

K(g) x

K(h)

.

- g E Y n-

=h

=

h

Observe if

=

X, Y

By so that

C X

Q.E.D.

x + g e X .

are both subalgebras, subcoalgebras, XO =

ideals or coideals the condition

is automatically

satisfied.

0)

If

x e Hn

[

o dn-l(X)

p

Y

where

Hq (xa)

a

1

a E

]o d

f

. 0 ( x ' m)

1

(o

0)

emae-

2

2e

a

e 1 (x

m

2

1

] o dn

E

n - 1 , then if

---

B*

is a right

B-module where

= If

x E L(B)

then

x

b* e B* , a, b E B

acts as a derivation of

a basis of monomials in

, within

xi

basis (not necessarily a basis for basis for

B

x aim

a i*]

Let

.

so

(x M)



=69

e .

E

a a * 2)



105.

= 0

by induction on the length of the monomial;

U C (H*p(J))

hence,

UO = (H*p(J))0 Letting f xa) K = k[ rx"I]

Let

V

fal

xe

k .

be an ordered basis for

L

Q

then by

b tD K(U)

=

HO0

m = 0,1,...

am p. --

a

p

m

0Then

Let

a

CJ

be a

1 =1

a* = al*

m

e

H*p(J)

and



=

t

106.

Xt = 0

which implies

and

K(h) E V .

Thus we are done

by lemma 5. We can draw the corollary that in characteristic p > 0

if

U = H

if and only if for each

H

is cocommutative and coconnected then

follows because Since

U

H*p(J) = 0

a*

H

generated by the primitives in

We shall show

is

J

a*P = 0 .

This

p(J) = 0

if and only if

may not equal

we examine the ideal H .

=a*p Ia*

LH = HL = p(H*)

E

e

LH = p(H*)

H*] HL = p(H*)

the proof

the same (up to reflection). HL c p(H*)

:

Let

=

2

act on

H*

from the right by

a* e H* ,

Then for

.

=

=

E L , h E H ,

0

acts as a derivation.

If fx

is an ordered basis for

L

K = k[ { xl] ,

K D W = the space spanned by

e 1 ) a (x

m 00(a

(x

m

=

1,2,...

a1 < - -< am for some

then

K(p(H*)

) C W .

Since if

e. p &ei

h E p(H*)

107.

K(h) =

Z

A (x

i = t

Suppose for

a

a1 eii

plet

m

C J

, then letting

...pet 1

a*

eim

)

m x i m

be a dual basis to

i=1 et

et

a* = a

1

*--x

implies

t

=

am* 0

= At

and

K(LH) D W n K(H) .

Suppose

K(h) s W , we shall show there is

K(g) = K(h) .

which

K(p(H*) ) C W .

and

Next we show

where

m e p(H*)

g E LH

h s H

- H

where

Let ei

K(h)

=

a M I A (xa) i=1 0

a

a*(x) = .)

1 < n(i)

_
o

xa f

1

o ds-2

a

@ - - - @ xa s-

Y

H+[S-1]

E

f +.-+f s-1 We show

= Ker e)

(H

Ker(E @ E o d) E

= H

Y e L

H+[s-2]

Since

.

it suffices to show

E 0 I[s-2] o d (

I[s-2](Y) = 0

or it suffices to show a

E @ E @ I[s-2] o d 0I[s-2]( Z Xa f > 0 i f +- -± --

t

E ® E ® I[s-2] o d 0 I[s-2]( xaf f i>0 1 f +- --

+ Y)

s-1 =

s

)

t.

The left hand side (top) =

E

E0I

[s2

o d

I[s-2] o E[Sl1

(since =

Els]

E0E od oE =EOEod) a

o ds-1(Yas)

Xfi

f i> 0 f +- . - - fs= t

a

which is the right hand side, (bottom). ing

Y E H

* .*

o ds-2 (as

L

- --.@ H+

so

By similar reason-

Y e L[s-1]

127.

Thus

~1

E

o dss-2]

as=

e.

a1

a

m

+ 2 Aa(x

z

f > 0 f1 +- -.+f

where

1 _

s-1

= t

e 11+- o*+

= S - 1

e

a1 p

r

qr =1

by the choice of an extending basis,

n(1 )+1

xa

Without loss we assume

so we can choose a dual basis ai*)m C J

where we assume

a

E

so

Since

ya

E

1

m

(Vna )+L -

(Vn(a)+1)

(a*) na

]

x

to

=

Ker p

)+1 = 0

let Vn-ItI(()

S(a*n

and

z E H

=

y .

)-n+|t|+

where

Let

128. eF -pnda)-n+ t|+1 a*= a1 *

1

I.

e1

*

am m

*

Note for

b* a H* ,

h e H

- h)

Vn(pn(b*)

m

b* - V"(h)

since



pn(b*)



= Consider

0 =



n(a 1 )-n+It|+1 =



=

X

a e

> --1-

k If H

H

is the

is finite

is connected H

is coconnected.

is cocommutative.

SupH*

is finite dimensional, connected, commutative; k

is split and

is perfect, then

is coconnected and

H*

cocommutative and has an extending basis by finile dimensionThus by theorem 6:

ality.

H* ~

as a coalgebra.

a

(in the notation of the theorem) a

m

By finite dimensionality

char p > 0

and

n(ai) < 00 . H*

has a basis of monomials as described in theorem 6.

Let {[a T**] C (H*)* for

H*

.

(a

to (x a

m=1

Let ta ,

be a dual basis to the monomial basis **

m

C

a

*

be the dual basis

(the extending basis for

L(H*).)

144.

k[x 1 ]

Then

(H*)*

n-l

k[Y ]

n~a p(

-@-m

)+1n(am)+1 3

p

as an algebra where an isomorphism is induced by -a1 aY **

By finite dimensionality

1=1,...0

m

H** = H

.

.

Thus we have proved

a finite dimensional commutative, connected, Hopf algebra over a perfect field where the dual is split is of the form:

(as an algebra) k[Y ]1

k[Y ] em

X1 >-2mp If

k

is algebraically closed then by the results of the

first chapter we can drop the condition that the dual is split since it is automatically satisfied, also

k

is

automatically perfect. We now conclude by applying results of previous chapters.

We show a coconnected cocommutative Hopf algebra

over a perfect field is an extension--as an algebra and a coalgebra--of

H/LH

by

U

characteristic zero

H = U

is trivial.

k

Assume

when and

U

is commutative.

LH = H+

In

so, the result

is perfect of characteristic

p > 0.

--4

145.

Lemma 6: and

K = k[

Let

be an ordered basis for

{ xa? If

xa ]I

where

h e V1

e

e =

K(h)

1 1

(a

A

a

m

i g

then there is

H

£

where pei

1 pe

a K(g)

=

Z 1

''

m

xm

and V(g)

V(f) = h .

Say

show we can choose choose

f

where

f

#(f)

K(f)

=

I

> pn

is minimal then

..

Pt

Suppose for some

be a dual basis t

i

,

f

)

C

x

Pt

We first

.

Suppose not

f

x1

S pn

+00+fit

pIf

90*

t x kf

k=1

a* = a*

t

Let

it

ak]

f

fj /p *.

*

1 /j

=

= = H

Observe

H/LH Dl k .

U C H

By

theorem 1 i

H/LH

(H

a

k H/LH

O

m )H

(D H

--

)H

H/LH is a linear isomorphism.

U-

is a split exact sequence. dition for

H

Hence,

U = H

H

H

®

0

k , and

H/LH

This verifies the first con-

to be an extension of

H/LH

by

U .

161.

Bibliography

[1]

E. Artin, C. Nesbitt, and R. Thrall, Rings with Minimum Condition, University of Michigan Press, 1944.

[2]

N. Jacobson,

[3]

J. Milnor and J. Moore, On the Structure of Hopf Algebras, Princeton University Notes.

Lie Algebras, Interscience, 1962.

162.

Index

Adjoint representation, Algebra, 8 111

A ,

commutative, 8 Antipode, 17 Augmentation, 10, 11 Brauer Group, 80 Coalgebra, 10 C.,

111

cocommutative, 11 Coconnected, 93 Cohomology, 47, 87

6, 48 6n

,

88

H(B,H)1 , 88

H' (H,B), 56 Hom(C,A)n, 88 Homn(C,A), 48 normal complex, 58 Reg(C,A)n

88

Regn (C,A), 48

155

r

163.

Coideal,

43

Comodule, 12 (C.)H.A. comodule, 87 comodule as an algebra, 36

$2,

structure on

X © X, 36

Connected, 143 Divided power, 122 Duality, 14 Extension, 61, 62, 89,

90, 155

f B 0) H, 67 Brauer Group, 80 H 9) g

B,

91

isomorphism of extensions, 62, 90 left exact sequence, 61 product of extensions, 81 right exact sequence, 89 smash product, 63 split exact sequence, 61, 89 Filtration, 26

r(G),

17

Grouplike element, 24 Hopf algebra, 16 coconnected, 93 conilpotent,

22

164.

filtration, 26 24

G(H), Hg,

25

L(H),

94

split, 21

U

95

,

J , 26 K(x),

98

L , 121 Module,

9

(C.)H.A. H

as

module, H*,

48

21

module as a coalgebra, 43 e3' -module, n 2

7

structure on

Poincare-Birkhoff-Witt

X 0 X, 42 theorem, 140

Primitive element, 94

p

,

113

Restricted Universal Enveloping Algebra, 93, 95 Smash product, 63 m He Tr(G) -+ H , 66 Symmetric tensor, 97 Universal Enveloping Algebra,

93, 95

165.

Unit, 8,

V V,

,

13

115 119

Vector space, 7 X , 110

166.

Biography

Moss Eisenberg Sweedler was born in Brooklyn, New York, April 29, 1942.

He was graduated from the Massachusetts

Institute of Technology with a B.S. in June 1963. tinued study there as a graduate student.

He con-