Algebra Abstrata Ufrj Luciane

C HAPTER
Basic Theory of Algebraic Groups

The emphasis in this chapter is on afne algebraic groups over a base eld, but, when it requires no extra effort, we often study more general objects: afne groups (not of nite type); base rings rather than elds; afne algebraic monoids rather than groups; afne algebraic supergroups (very briey); quantum groups (even more briey). The base eld (or ring) is alway denoted k, and R is always a commutative k-algebra.
Basic Theory of Algebraic Groups 1 Introductory overview . . . . . . . . . . . . . . . . . . . 1a The building blocks . . . . . . . . . . . . . . . . 1b Extensions . . . . . . . . . . . . . . . . . . . . 1c Summary . . . . . . . . . . . . . . . . . . . . . 1d Exercises . . . . . . . . . . . . . . . . . . . . . 2 Denitions; examples . . . . . . . . . . . . . . . . . . . 2a Denition . . . . . . . . . . . . . . . . . . . . . 2b Examples . . . . . . . . . . . . . . . . . . . . . 2c Afne monoids . . . . . . . . . . . . . . . . . . 2d Afne supergroups . . . . . . . . . . . . . . . . 2e Relaxing the hypothesis on k . . . . . . . . . . . 2f Terminology . . . . . . . . . . . . . . . . . . . 3 Some basic constructions . . . . . . . . . . . . . . . . . 3a Products of afne groups . . . . . . . . . . . . . 3b Fibred products of afne groups . . . . . . . . . 3c Extension of the base ring (extension of scalars) . 3d Restriction of the base ring (restriction of scalars) 3e The Greenberg functor . . . . . . . . . . . . . . 4 Afne groups and Hopf algebras . . . . . . . . . . . . . 4a Algebras . . . . . . . . . . . . . . . . . . . . . 4b Co-algebras . . . . . . . . . . . . . . . . . . . . 4c The duality of algebras and co-algebras . . . . . 4d Bi-algebras . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
1 5 5 7 8 9 10 10 14 18 19 20 22 23 23 23 24 24 27 28 28 28 29 30
This is Chapter 1 of Algebraic Groups, Lie Groups, and their Arithmetic Subgroups, available at www.jmilne. org/math/. Version 2.21, April 27, 2010. Copyright c 2005, 2006, 2009, 2010 J.S. Milne.
4e Afne monoids and bi-algebras . . . . . . . . . . . . . . . . . . . . . 4f Afne groups and Hopf algebras . . . . . . . . . . . . . . . . . . . . . 4g Abstract restatement . . . . . . . . . . . . . . . . . . . . . . . . . . . 4h Explicit description of , , and S . . . . . . . . . . . . . . . . . . . . 4i Commutative afne groups . . . . . . . . . . . . . . . . . . . . . . . . 4j Finite at algebraic groups; Cartier duality . . . . . . . . . . . . . . . 4k Galois descent of afne groups . . . . . . . . . . . . . . . . . . . . . . 4l Quantum groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4m Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4n Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Algebraic groups and afne algebraic schemes . . . . . . . . . . . . . . . . . . 5a Afne k-algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5b The max spectrum of a ring . . . . . . . . . . . . . . . . . . . . . . . 5c Afne algebraic schemes . . . . . . . . . . . . . . . . . . . . . . . . . 5d Properties of afne algebraic schemes . . . . . . . . . . . . . . . . . . 5e Algebraic groups as groups in the category of afne algebraic schemes 5f Homogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5g Reduced algebraic groups . . . . . . . . . . . . . . . . . . . . . . . . 5h Smooth algebraic schemes . . . . . . . . . . . . . . . . . . . . . . . . 5i Smooth algebraic groups . . . . . . . . . . . . . . . . . . . . . . . . . 5j Algebraic groups in characteristic zero are smooth (Cartiers theorem) . 5k Transporters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5l Relaxing the conditions on k and G . . . . . . . . . . . . . . . . . . . 5m Appendix: The faithful atness of Hopf algebras . . . . . . . . . . . . 5n Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Group theory: subgroups and quotient groups. . . . . . . . . . . . . . . . . . . 6a A criterion to be an isomorphism . . . . . . . . . . . . . . . . . . . . . 6b Subgroups; injective homomorphisms . . . . . . . . . . . . . . . . . . 6c Kernels of homomorphisms . . . . . . . . . . . . . . . . . . . . . . . 6d Dense subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6e Normalizers; centralizers; centres . . . . . . . . . . . . . . . . . . . . 6f Quotient groups; surjective homomorphisms . . . . . . . . . . . . . . 6g Existence of quotients . . . . . . . . . . . . . . . . . . . . . . . . . . 6h Semidirect products . . . . . . . . . . . . . . . . . . . . . . . . . . . 6i Algebraic groups as sheaves . . . . . . . . . . . . . . . . . . . . . . . 6j Limits of afne groups . . . . . . . . . . . . . . . . . . . . . . . . . . 6k Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Representations of afne groups . . . . . . . . . . . . . . . . . . . . . . . . . 7a Denition of a representation . . . . . . . . . . . . . . . . . . . . . . . 7b Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7c Comodules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7d The category of comodules over C . . . . . . . . . . . . . . . . . . . . 7e Representations and comodules . . . . . . . . . . . . . . . . . . . . . 7f The category of representations of G . . . . . . . . . . . . . . . . . . 7g Afne groups are inverse limits of algebraic groups . . . . . . . . . . . 7h Algebraic groups admit nite-dimensional faithful representations . . . 7i The regular representation contains all . . . . . . . . . . . . . . . . . . 7j Every faithful representation contains all . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31 32 33 34 35 36 36 37 37 38 40 40 40 41 43 43 45 46 47 47 48 50 52 52 55 56 56 56 58 60 62 65 68 68 69 72 72 74 74 76 76 79 79 82 83 84 85 87
10
11
12
13
14
7k Stabilizers of subspaces . . . . . . . . . . . . . . . . . . . . . 7l Sub-coalgebras and subcategories . . . . . . . . . . . . . . . . 7m Quotient groups and subcategories . . . . . . . . . . . . . . . . 7n Normal subgroups and subcategories . . . . . . . . . . . . . . 7o Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Group theory: the isomorphism theorems . . . . . . . . . . . . . . . . 8a Review of abstract group theory . . . . . . . . . . . . . . . . . 8b The existence of quotients . . . . . . . . . . . . . . . . . . . . 8c The homomorphism theorem . . . . . . . . . . . . . . . . . . . 8d The isomorphism theorem . . . . . . . . . . . . . . . . . . . . 8e The correspondence theorem . . . . . . . . . . . . . . . . . . . 8f The Schreier renement theorem . . . . . . . . . . . . . . . . . 8g The category of commutative algebraic groups . . . . . . . . . 8h Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Recovering a group from its representations; Jordan decompositions . . 9a Recovering a group from its representations . . . . . . . . . . . 9b Application to Jordan decompositions . . . . . . . . . . . . . . Characterizations of categories of representations . . . . . . . . . . . . 10a Characterization of categories of comodules . . . . . . . . . . . 10b Characterization of categories of representations of afne groups Finite algebraic groups . . . . . . . . . . . . . . . . . . . . . . . . . . 11a Etale groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 11b Finite group schemes in characteristic p 0 . . . . . . . . . . 11c Cartier duality revisited . . . . . . . . . . . . . . . . . . . . . . 11d Finite at group schemes over commutative rings . . . . . . . . 11e Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The connected components of an algebraic group . . . . . . . . . . . . 12a Some algebraic geometry . . . . . . . . . . . . . . . . . . . . . 12b Separable subalgebras . . . . . . . . . . . . . . . . . . . . . . 12c The group of connected components of an algebraic group . . . 12d Connected algebraic groups . . . . . . . . . . . . . . . . . . . 12e Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12f Afne groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 12g Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12h Where we are . . . . . . . . . . . . . . . . . . . . . . . . . . . Groups of multiplicative type; tori . . . . . . . . . . . . . . . . . . . . 13a Group-like elements . . . . . . . . . . . . . . . . . . . . . . . 13b The characters of an algebraic group . . . . . . . . . . . . . . . 13c The algebraic group D.M / . . . . . . . . . . . . . . . . . . . . 13d Characterizing the groups D.M / . . . . . . . . . . . . . . . . 13e Diagonalizable groups . . . . . . . . . . . . . . . . . . . . . . 13f Diagonalizable groups are diagonalizable . . . . . . . . . . . . 13g Split tori and their representations . . . . . . . . . . . . . . . . 13h Groups of multiplicative type . . . . . . . . . . . . . . . . . . 13i Rigidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13j Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Solvable algebraic groups . . . . . . . . . . . . . . . . . . . . . . . . . 14a Commutative groups are triangulizable . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
88 90 91 91 94 95 95 95 96 98 99 99 101 101 102 102 105 110 110 113 115 115 118 119 120 120 121 121 122 123 125 126 127 127 128 129 129 129 130 131 132 133 134 135 136 137 138 138 3
15
16 17
18
14b Decomposition of a commutative algebraic group 14c The derived group of algebraic group . . . . . . 14d Solvable algebraic groups . . . . . . . . . . . . 14e Independence of characters . . . . . . . . . . . . 14f The Lie-Kolchin theorem . . . . . . . . . . . . . 14g Unipotent groups . . . . . . . . . . . . . . . . . 14h Structure of solvable groups . . . . . . . . . . . 14i Tori in solvable groups . . . . . . . . . . . . . . 14j Exercises . . . . . . . . . . . . . . . . . . . . . The classical almost-simple groups . . . . . . . . . . . . 15a Quadratic spaces . . . . . . . . . . . . . . . . . 15b Theorems of Witt and Cartan-Dieudonn . . . . e 15c The orthogonal group . . . . . . . . . . . . . . . 15d Super algebras . . . . . . . . . . . . . . . . . . 15e Brief review of the tensor algebra . . . . . . . . 15f The Clifford algebra . . . . . . . . . . . . . . . 15g The Spin group . . . . . . . . . . . . . . . . . . 15h The Clifford group . . . . . . . . . . . . . . . . 15i Action of O.q/ on Spin.q/ . . . . . . . . . . . . 15j Restatement in terms of algebraic groups . . . . The exceptional almost-simple groups . . . . . . . . . . 16a The group G2 . . . . . . . . . . . . . . . . . . . General algebraic groups . . . . . . . . . . . . . . . . . 17a The radical of an algebraic group . . . . . . . . 17b Semisimple and reductive groups . . . . . . . . 17c Reductive groups . . . . . . . . . . . . . . . . . 17d Structure of a general algebraic group . . . . . . Tannakian categories . . . . . . . . . . . . . . . . . . . 18a Properties of G versus those of Repk .G/ . . . . 18b Tensor categories . . . . . . . . . . . . . . . . . 18c Neutral tannakian categories . . . . . . . . . . . 18d Gradations on tensor categories . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
139 140 142 143 144 145 148 149 149 150 150 151 154 154 156 156 160 161 163 163 164 164 165 165 166 167 168 169 169 171 172 174 177 179
Bibliography Index of denitions
Introductory overview
Loosely speaking, an algebraic group over a eld k is a group dened by polynomials. Before giving the precise denition in the next section, we look at some examples of algebraic groups. Consider the group SLn .k/ of n n matrices with entries in a eld k and with determinant 1. The determinant of a matrix .aij / is a polynomial in the entries aij of the matrix, namely, X det.aij / D sign. / a1 .1/ an .n/ (Sn D symmetric group),
2Sn
and so SLn .k/ is the subset of Mn .k/ D k n dened by the polynomial condition det.aij / D 1. The entries of the product of two matrices are polynomials in the entries of the two matrices, namely, .aij /.bij / D .cij / with cij D ai1 b1j C C ai n bnj ;
and Cramers rule realizes the entries of the inverse of a matrix with determinant 1 as polynomials in the entries of the matrix, and so SLn .k/ is an algebraic group (called the special linear group). The group GLn .k/ of n n matrices with nonzero determinant is also an algebraic group (called the general linear group) because its elements can be identied with the n2 C1-tuples ..aij /1i;j n ; d / such that det.aij / d D 1. More generally, for a nite-dimensional vector space V , we dene GL.V / (resp. SL.V /) to be the group of automorphisms of V (resp. automorphisms with determinant 1). These are again algebraic groups. To simplify the statements, for the remainder of this section, we assume that the base eld k has characteristic zero.
1a
The building blocks
We now list the ve types of groups from which all others can be constructed by successive extensions: the nite groups, the abelian varieties, the semisimple groups, the tori, and the unipotent groups.
F INITE GROUPS
Every nite group can be realized as an algebraic group, and even as an algebraic subgroup of GLn .k/. Let be a permutation of f1; : : : ; ng and let I. / be the matrix obtained from the identity matrix by using to permute the rows. For any n n matrix A, the matrix I. /A is obtained from A by using to permute the rows. In particular, if and 0 are two permutations, then I. /I. 0 / D I. 0 /. Thus, the matrices I. / realize Sn as a subgroup of GLn . Since every nite group is a subgroup of some Sn , this shows that every nite group can be realized as a subgroup of GLn , which is automatically dened by polynomial conditions. Therefore the theory of algebraic groups includes the theory of nite groups. The algebraic groups dened in this way by nite groups are called constant nite algebraic groups. An algebraic group is connected if it has no constant nite group as a quotient, even over the algebraic closure of k.
A BELIAN VARIETIES
Abelian varieties are connected algebraic groups that are projective when considered as algebraic varieties. An abelian variety of dimension 1 is an elliptic curve, which can be described by a homogeneous equation Y 2 Z D X 3 C bXZ 2 C cZ 3 : 5
BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1)
Therefore, the theory of algebraic groups includes the theory of abelian varieties. We shall ignore this aspect of the theory. In fact, we shall study only algebraic groups that are afne when considered as algebraic varieties.
S IMPLE GROUPS ; SEMISIMPLE GROUPS

A connected algebraic group G is simple if it is not commutative and has no normal algebraic subgroups, and it is almost-simple1 if its centre Z is nite and G=Z is simple. For example, SLn is almost-simple for n > 1 because its centre ( ZD
0
! :: :
0
)
n
D1
is nite and the quotient PSLn D SLn =Z is simple. An isogeny of algebraic groups is a surjective homomorphism G ! H with nite kernel. Two algebraic groups H1 and H2 are isogenous if there exist isogenies H1 G ! H2 :
This is an equivalence relations. When k is algebraically closed, every almost-simple algebraic group is isogenous to exactly one algebraic group on the following list: An .n Bn .n 1/; the special linear group SLnC1 I 2/; the special orthogonal group SO2nC1 consisting of all 2n C 1 2n C 1 matrices A such that At A D I and det.A/ D 1; Cn .n 3/; the symplectic group Sp2n consisting of all invertible 2n 2n matrices A such that 0 I t J A D J where J D A ; I 0 Dn .n 4/; the special orthogonal group SO2n ; E6 ; E7 ; E8 ; F4 ; G2 the ve exceptional groups. We say that an algebraic group G is an almost-direct product of its algebraic subgroups G1 ; : : : ; Gr if the map .g1 ; : : : ; gr / 7! g1 gr W G1 Gr ! G
is an isogeny. In particular, this means that each Gi is a normal subgroup of G and that the Gi commute with each other. For example, G D SL2 SL2 =N; N D f.I; I /; . I; I /g (1)
is the almost-direct product of SL2 and SL2 , but it is not a direct product of two almost-simple algebraic groups. A connected algebraic group is semisimple if it is an almost-direct product of almost-simple subgroups. For example, the group G in (1) is semisimple. Semisimple algebraic groups will be our main interest.
1 Other
authors say quasi-simple or simple.
1. Introductory overview
G ROUPS OF MULTIPLICATIVE TYPE ; ALGEBRAIC TORI

An algebraic subgroup T of GL.V / is said to be of multiplicative type if, over k al , there exists a basis of V relative to which T is contained in the group Dn of all diagonal matrices 0 1 0 0 0 B0 0 0C B C B : : :: : :C B: : : : : C: : :C B: : @0 0 0A 0 0 0 In particular, the elements of an algebraic torus are semisimple endomorphisms of V . A connected algebraic group of multiplicative type is a torus.
U NIPOTENT GROUPS
An algebraic subgroup G of GL.V / is unipotent if there exists a basis of V relative to which G is contained in the group Un of all n n matrices of the form 0 1 1 C B0 1 B C B: : :: : :C (2) B: : : : : C: : :C B: : @0 0 A 1 0 0 0 1 In particular, the elements of a unipotent group are unipotent endomorphisms of V .
1b
Extensions
We now look at some algebraic groups that are nontrivial extensions of groups of the above types.
S OLVABLE GROUPS
An algebraic group G is solvable if it there exists a sequence of algebraic subgroups G D G0 Gi Gn D 1
such that each Gi C1 is normal in Gi and Gi =Gi C1 is commutative. For example, the group Un is solvable, and the group Tn of upper triangular n n matrices is solvable because it contains Un as a normal subgroup with quotient isomorphic to Dn . When k is algebraically closed, a connected subgroup G of GL.V / is solvable if and only if there exists a basis of V relative to which G is contained in Tn (Lie-Kolchin theorem 14.21).
R EDUCTIVE GROUPS
A connected algebraic group is reductive if it has no nontrivial connected normal unipotent subgroups. According to the table below, they are extensions of semisimple groups by tori. For example, GLn is reductive. It is an extension of the simple group PGLn by the torus Gm , 1 ! Gm ! GLn ! PGLn ! 1: Here Gm D GL1 , and the map sends it onto the subgroup of nonzero scalar matrices.
N ONCONNECTED GROUPS
We give some examples of naturally occurring nonconnected algebraic groups.
The orthogonal group. For an integer n 1, let O.n/ denote the group of n n matrices A such that At A D I . Then det.A/2 D det.At / det.A/ D 1, and so det.A/ 2 f1g. The matrix diag. 1; 1; : : :/ lies in O.n/ and has determinant 1, and so O.n/ is not connected: it contains SO.n/ D Ker O.n/ ! f1g as a normal algebraic subgroup of index 2 with quotient the constant nite group f1g.
def
det
The monomial matrices. Let M be the group of monomial matrices, i.e., those with exactly one nonzero element in each row and each column. This group contains both the algebraic subgroup Dn and the algebraic subgroup Sn of permutation matrices. Moreover, for any diagonal matrix diag.a1 ; : : : ; an /; I. / diag.a1 ; : : : ; an / I. /
1
D diag.a
.1/ ; : : : ; a .n/ /.
(3)
As M D Dn Sn , this shows that Dn is normal in M . Clearly D \ Sn D 1, and so M is the semi-direct product M D Dn where W Sn ! Aut.Dn / sends
Sn
to the automorphism in (3).
1c
Summary
Recall that we are assuming that the base eld k has characteristic zero. Every algebraic group has a composition series whose quotients are respectively a nite group, an abelian variety, a semisimple group, a torus, and a unipotent group. More precisely: (a) An algebraic group G contains a unique normal connected subgroup G such that G=G is nite (see 12.9). (b) A connected algebraic group G contains a largest normal afne algebraic subgroup N ; the quotient G=N is an abelian variety (Chevalley, Barsotti, Rosenlicht).2 (c) A connected afne algebraic group G contains a largest connected normal solvable subgroup N (see 17a); the quotient G=N semisimple. (d) A connected afne group G contains a largest connected normal unipotent subgroup N ; when G is solvable, G=N is of multiplicative type (see 17.2; 14.35). In the following tables, the group at left has a composition series whose quotients are the groups at right.
2 Rosenlicht (1956) notes that this was proved by Chevalley in 1953, who only published his proof in Chevalley 1960.
Barsotti (1955) and Rosenlicht (ibid.) independently gave different proofs. A modern proof can be found in Conrad 2002.
1. Introductory overview
General algebraic group

general
Afne algebraic group

afne
Reductive algebraic groups
j
connected
nite etale abelian variety semisimple torus unipotent
j j
connected
nite etale reductive semisimple torus torus f1g unipotent
connected afne
j j
solvable
j j
semisimple torus
solvable
j j
unipotent
unipotent
j j
f1g
f1g
1d
Exercises
E XERCISE 1-1 Let f .X; Y / 2 RX; Y . Show that if f .x; e x / D 0 for all x 2 R, then f is zero (as an element of RX; Y ). Hence the subset f.x; e x / j x 2 Rg of R2 is not the zero-set of a family of polynomials. E XERCISE 1-2 Let T be a commutative subgroup of GL.V / consisting of diagonalizable endomorphisms. Show that there exists a basis for V relative to which T Dn . E XERCISE 1-3 Let be a positive denite bilinear form on a real vector space V , and let SO. / be the algebraic subgroup of SL.V / of such that .x; y/ D .x; y/ for all x; y 2 V . Show that every element of SO. / is semisimple (but SO. / is not diagonalizable because it is not commutative). E XERCISE 1-4 Let k be a eld of characteristic zero. Show that every element of GLn .k/ of nite order is semisimple. (Hence the group of permutation matrices in GLn .k/ consists of semisimple elements, but it is not diagonalizable because it is not commutative).
Denitions; examples
What is an algebraic group? For example, what is SLn ? We know what SLn .R/ is for any commutative ring R, namely, it is the group of n n matrices with entries in R and determinant 1. Moreover, we know that a homomorphism R ! R0 of rings denes a homomorphism of groups SLn .R/ ! SLn .R0 /. So what is SLn without the .R/? Obviously, it is a functor from the category of rings to groups. We adopt this as our denition of an afne algebraic group together with the requirement that the functor be dened by polynomials. For simplicity, in 2a2d we require k to be a eld.
2a
Denition
M OTIVATING DISCUSSION
We rst explain how a set of polynomials denes a functor. Let S be a subset of kX1 ; : : : ; Xn . For any k-algebra R, the zero-set of S in Rn is S.R/ D f.a1 ; : : : ; an / 2 Rn j f .a1 ; : : : ; an / D 0 for all f 2 S g: A homomorphism of k-algebras R ! R0 denes a map S.R/ ! S.R0 /, and these maps make R S.R/ into a functor from the category of k-algebras to the category of sets. This suggest dening an afne algebraic group to be a functor from the category of k-algebras to groups that is is isomorphic (as a functor to sets) to the functor dened by a set of polynomials in a nite number of symbols. For example, the functor R SLn .R/ satises this condition because it is isomorphic to the functor dened by the polynomial det.Xij / 1 where X det.Xij / D sign. / X1 .1/ Xn .n/ 2 kX11 ; X12 ; : : : ; Xnn : (4)
2Sn
The condition that G can be dened by polynomials is very strong: it excludes, for example, the functor with ( Z=2Z if R D k G.R/ D f1g otherwise. Let S be a subset of kX1 ; : : : ; Xn . The ideal a generated by S consists of the nite sums X gi fi ; gi 2 kX1 ; : : : ; Xn ; fi 2 S: Clearly S and a have the same zero-sets for any k-algebra R. According to the Hilbert basis theorem (CA 3.6), every ideal in kX1 ; : : : ; Xn can be generated by a nite set of polynomials, and so an afne algebraic group is isomorphic (as a functor to sets) to the functor dened by a nite set of polynomials. We have just observed that an afne algebraic group G is isomorphic to the functor dened by an ideal a of polynomials in some polynomial ring kX1 ; : : : ; Xn . Let A D kX1 ; : : : ; Xn =a. For any k-algebra R, a homomorphism A ! R is determined by the images ai of the Xi , and the ntuples .a1 ; : : : ; an / that arise from a homomorphism are exactly those in the zero-set of a. Therefore the functor R a.R/ sending a k-algebra R to the zero-set of a in Rn is canonically isomorphic to the functor R Homk-alg .A; R/: Since the k-algebras that can be expressed in the form kX1 ; : : : ; Xn =a are exactly the nitely generated k-algebras, we conclude that the functors Algk ! Set dened by a set of polynomials 10
2. Denitions; examples
11
in a nite number of symbols are exactly the functors R Homk-alg .A; R/ dened by a nitely generated k-algebra A. Before continuing, it is convenient to review some category theory.
T HE YONEDA LEMMA
An object A of a category A denes a functor hA W A ! Set by hA .R/ D Hom.A; R/; R 2 ob.A/; hA .f /.g/ D f g; f W R ! R0 ; g 2 hA .R/ D Hom.A; R/:
A morphism W A ! A0 of objects denes a map f 7! f W hA0 .R/ ! hA .R/ which is natural in R (i.e., it is a natural transformation hA0 ! hA ): Let F W A ! Set be a functor from A to the category of sets, and let A be an object of A. A natural transformation T W hA ! F denes an element aT D TA .idA / of F .A/. 2.1 (YONEDA LEMMA ) For any functor F W A ! Set and object A of A, the map T 7! aT is a bijection Hom.hA ; F / ' F .A/ which is natural in both A and F . In particular, when F D hB , Hom.hA ; hB / ' Hom.B; A/: P ROOF. Let T be a natural transformation hA ! F . For any morphism f W A ! R, the commutative diagram hA .A/
TA hA .f /
(5)
hA .R/
TR
idA
F .A/ shows that
F .f /
F .R/
aT
F .f /.aT / D TR .f /
TR .f / D F .f /.aT /:
(6)
Therefore T is determined by aT , and so the map T 7! aT is injective. On the other hand, an element a of F .A/ denes a natural transformation Ta W hA ! F by the rule .Ta /R .f / D F .f /.a/; f 2 hA .R/ D Hom.A; R/:
2
As .Ta /A .idA / D F .idA /.a/ D a, the map T 7! aT is surjective. The proof of the naturality of (5) is left as an (easy) exercise for the reader.
2.2 A functor F W A ! Set is said to be representable if it is isomorphic to hA for some object A. A pair .A; a/, a 2 F .A/, is said to represent F if Ta W hA ! F is an isomorphism. Note that, if F is representable, say F hA , then the choice of an isomorphism T W hA ! F determines an element aT 2 F .A/ such that .A; aT / represents F in fact, T D TaT and so we sometimes say that .A; T / represents F . 2.3 Let F1 and F2 be functors A ! Set. In general, the natural transformations F1 ! F2 will form a proper class (not a set), but the Yoneda lemma shows that Hom.F1 ; F2 / is a set if F1 is representable (because it is isomorphic to a set).
12
D EFINITIONS
After these preliminaries, we make the following denition. D EFINITION 2.4 An afne group (resp. afne algebraic group) over k is a functor GW Algk ! Grp such that the composite
Algk ! Grp ! Set
G forget
is representable by a k-algebra (resp. a nitely generated k-algebra). A pair .A; a/ representing .forget/G is called the coordinate ring (or coordinate algebra) of G, and A is denoted O.G/ or kG. A homomorphism G ! H of afne groups over k is a natural transformation, i.e., a family of homomorphisms .R/W G.R/ ! H.R/ of groups such that, for every homomorphism of k-algebras W R ! R0 , the diagram G.R/ ? ? yG. G.R0 / commutes. Less formally, an afne group can be regarded as a functor from k-algebras to groups that is isomorphic (as a functor to sets) to the functor dened by a (possibly innite) set of polynomials in a (possibly innite) set of symbols. Later (7.21) we shall see that the afne groups are exactly the functors that can be expressed as inverse limits of afne algebraic groups. 2.5 The coordinate ring of an afne group is uniquely determined up to a unique isomorphism. Below we shall see that there is a canonical choice for it. It is often convenient to regard the coordinate ring .A; a/ of an afne group G as a k-algebra A together with an isomorphism aW hA ! G of functors to sets (see 2.2). 2.6 Statement 2.3 shows that if G and H are afne groups over k, then Hom.G; H / is a set. Therefore, the afne groups over k form a category with the afne algebraic groups as a full subcategory. 2.7 The group structure of an afne algebraic group is also described by polynomials. We explain briey how this follows from the Yoneda lemma. Let A and B be nitely generated k-algebras. The Yoneda lemma shows that every natural transformation hA ! hB arises from a unique homomorphism of k-algebras B ! A. When we write A and B as quotients of polynomial rings, say, A D kX1 ; : : : ; Xm =a and B D kY1 ; : : : ; Yn =b, we see that a homomorphism B ! A can be described by polynomials, namely, by a set ff1 ; : : : ; fn g of representatives for the images of Y1 ; : : : ; Yn . Let G be an algebraic group. The map g 7! g 1 W G.R/ ! G.R/ is natural in R, and so arises from a unique homomorphism S W O.G/ ! O.G/, which can be described by polynomials. The functor R G.R/ G.R/ is represented by O.G/ k O.G/ (CA (2), p.27), and so the multiplication map G.R/ G.R/ ! G.R/ arises from a unique homomorphism W A ! A k A, which again can be described by polynomials. We discuss the maps and S in detail in Section 4.
.R/
! H.R/ ? ? yH. ! H.R0 /
.R0 /
13
2.8 Let G0 be a functor Algk ! Set. To give a functor GW Algk ! Grp such that G0 D .forget/ G is the same as giving a natural transformation G0 G0 ! G0 that makes G0 .R/ into a group for all k-algebras R. Thus, to give an afne group over k is the same as giving a representable functor G0 W Algk ! Set and a natural transformation G0 G0 ! G0 that makes G.R/ into a group for all k-algebras R. 2.9 Let A be a k-algebra, and let W A ! A k A be a homomorphism. Given homomorphisms f1 ; f2 W A ! R, let .f1 ; f2 / be the unique homomorphism A k A ! R such that .f1 ; f2 /.a 1/ D f1 .a/ and .f1 ; f2 /.1 a/ D f2 .a/ for all a 2 A. Now denes a map f1 ; f2 7! f1 f2 D .f1 ; f2 / W hA .R/ hA .R/ ! hA .R/, which is natural in R. If these maps make hA .R/ into a group for all R, then they make hA into an afne group over k (by 2.8). We saw in (2.7) that all afne groups arise in this way, and so, to give an afne group over k is essentially the same as giving a k-algebra A and a homomorphism W A ! A k A such that makes hA .R/ into a group for all R. (More precisely, there is an equivalence of categories . . . .)
def
T HE CANONICAL COORDINATE RING

Let A1 be the functor sending a k-algebra R to its underlying set, A1 W Algk ! Set; .R; ; C; 1/ R;
and let G be a functor from the category of k-algebras to groups, GW Algk ! Grp. Let G0 D .forget/G be the underlying functor to Set, and let A be the set of natural transformations from G to A1 , A D Hom.G; A1 /: Thus an element f of A is a family of maps of sets fR W G.R/ ! R; R a k-algebra,
such that, for every homomorphism of k-algebras R ! R0 , the diagram G.R/ ? ? y G.R0 / commutes. For f; f 0 2 A and g 2 G.R/, dene
0 .f f 0 /R .g/ D fR .g/ fR .g/ 0 .ff 0 /R .g/ D fR .g/fR .g/: fR
! R ? ? y ! R0
fR 0
With these operations, A becomes a commutative ring, and even a k-algebra because each c 2 k denes a constant natural transformation cR W G.R/ ! R; cR .g/ D c for all g 2 G.R/:
An element g 2 G.R/ denes a homomorphism f 7! fR .g/W A ! R of k-algebras. In this way, we get a natural transformation W G0 ! hA of functors to sets.
14
P ROPOSITION 2.10 The functor G is an afne group if and only if is an isomorphism (in which case it is an algebraic group if and only if A is nitely generated). P ROOF. If is an isomorphism, then G is an afne group (by denition). Conversely, let G be an afne group, and let .B; aW hB ! G0 / be a coordinate ring for G. Then, because A1 D hkX , A D Hom.G0 ; A1 / ' Hom.hB ; A1 / ' A1 .B/ ' B. Thus A ' B, and one checks that W hB ! hA is the natural transformation dened by this isomorphism; therefore is an isomorphism. This proves the rst part of the statement (and the parenthetical statement is obvious). 2 Thus, for an afne group G, O.G/ D Hom.G0 ; A1 / is the coordinate ring of G. N OTATION 2.11 Let A be a coordinate ring for G. When we regard A as Hom.G; A1 /, an element f 2 A is a family of maps fR W G.R/ ! R (of sets) indexed by the k-algebras R. On the other hand, when we regard A as a k-algebra representing G, an element g 2 G.R/ is a homomorphism of k-algebras gW A ! R. The two points of views are related by the equation fR .g/ D g.f /; f 2 A; g 2 G.R/: (7)
def
def
Yoneda
According to the Yoneda lemma, a homomorphism W G ! H denes a homomorphism of rings W O.H / ! O.G/. Explicitly, . f /R .g/ D fR .R g/; f 2 O.H /; g 2 G.R/: (8)
2b
Examples
E XAMPLE 2.12 Let Ga be the functor sending a k-algebra R to itself considered as an additive group, i.e., Ga .R/ D .R; C/. To give an element of R is the same as giving a k-algebra homomorphism kX ! R, and so Ga .R/ ' Homk-alg .kX ; R/: Therefore Ga is an afne algebraic group with O.Ga / D kX ; it is called the additive group. E XAMPLE 2.13 Let Gm be the functor R inverse, and so R (multiplicative group). Each a 2 R has a unique 1/; R/:
Gm .R/ ' f.a; b/ 2 R2 j ab D 1g ' Homk-alg .kX; Y =.X Y
Therefore Gm is an afne algebraic group, called the multiplicative group. Let k.X / be the eld of fractions of kX , and let kX; X 1 be the subring of polynomials in X and X 1 . The homomorphism kX; Y ! kX; X 1 ; X 7! X; Y 7! X 1 denes an isomorphism kX; Y =.X Y 1/ ' kX; X
1 ,
and so
1
Gm .R/ ' Homk-alg .kX; X Thus O.Gm / D kX; X

1 ;
; R/:
for f 2 kX; X
and a 2 Gm .R/ D R ,
1
fR .a/ D f .a; a
/:
2. Denitions; examples E XAMPLE 2.14 Let G be the functor such that G.R/ D f1g for all k-algebras R. Then G.R/ ' Homk-alg .k; R/;
15
and so G is an afne algebraic group, called the trivial algebraic group. More generally, for any Q nite group G, let O.G/ D g2G kg (product of copies of k indexed by the elements of G). Then R Homk-alg .O.G/; R/ is an afne algebraic group .G/k over k such that .G/k .R/ D G for any k-algebra R with no nontrivial idempotents (see 4.13 below). Such an afne algebraic group is called a constant nite algebraic group. E XAMPLE 2.15 For an integer n 1,
n .R/
D fr 2 R j r n D 1g
is a multiplicative group, and R

n .R/
n .R/
is a functor. Moreover, 1/; R/; 1/.
' Homk-alg .kX =.X n

n/
and so
is an afne algebraic group with O.
D kX =.X n
E XAMPLE 2.16 In characteristic p 0, the binomial theorem takes the form .a C b/p D ap C b p . Therefore, for any k-algebra R over a eld k of characteristic p 0, p .R/ D fr 2 R j r p D 0g is an additive group, and R p .R/ is a functor. Moreover, p .R/ ' Homk-alg .kT =.T p /; R/; and so p is an afne algebraic group with O.p / D kT =.T p /. E XAMPLE 2.17 For any k-vector space V , the functor of k-algebras3 D a .V /W R Homk-lin .V; R/ (additive group) (9)
is represented by the symmetric algebra Sym.V / of V : Homk-alg .Sym.V /; R/ ' Homk-lin .V; R/, R a k-algebra,
(see CA 8). Therefore D a .V / is an afne group over k (and even an afne algebraic group when V is nite dimensional). Now assume that V is nite dimensional. Then Homk-lin .V _ ; R/ ' R k V , and so R Moreover,
3 Notations
R k V (additive group) is an afne algebraic group with coordinate ring Sym.V _ /. EndR-lin .R k V / ' R k Endk-lin .V / ' R k .V _ k V /,
suggested by those in DG II, 1, 2.1.
16 and so the functor R
EndR-lin .R k V / (additive group)
is an algebraic group Va with coordinate ring Sym.V k V _ /. The choice of a basis e1 ; : : : ; en for V denes isomorphisms EndR-lin .R k V / ' Mn .R/ and Sym.V k V _ / ' kX11 ; X12 ; : : : ; Xnn (polynomial algebra in the n2 symbols .Xij /1i;j n ). For f 2 kX11 ; X12 ; : : : ; Xnn and a D .aij / 2 Mn .R/, fR .a/ D f .a11 ; a12 ; : : : ; ann /. E XAMPLE 2.18 For n n matrices M and N with entries in a k-algebra R, det.MN / D det.M / det.N / and adj.M / M D det.M / I D M adj.M / where I denotes the identity matrix and adj.M / D . 1/i Cj det Mj i 2 Mn .R/ with Mij the matrix obtained from M by deleting the i th row and the j th column. These formulas can be proved by the same argument as for R a eld, or by applying the principle of permanence of identities (Artin 1991, 12.3). Therefore, there is a functor SLn sending a k-algebra R to the group of n n matrices with entries in R and with determinant 1. Moreover, kX11 ; X12 ; : : : ; Xnn SLn .R/ ' Homk-alg ;R ; .det.Xij / 1/ where det.Xij / is the polynomial (4), and so SLn is an afne algebraic group with O.SLn / D kX11 ;X12 ;:::;Xnn . It is called the special linear group. For f 2 O.SLn / and a D .aij / 2 SLn .R/, .det.X / 1/
ij
(10)
(Cramers rule)
(11)
fR .a/ D f .a11 ; : : : ; ann /: E XAMPLE 2.19 Similar arguments show that the n n matrices with entries in a k-algebra R and with determinant a unit in R form a group GLn .R/, and that R GLn .R/ is a functor. Moreover, kX11 ; X12 ; : : : ; Xnn ; Y ;R ; GLn .R/ ' Homk-alg .det.Xij /Y 1/
11 ;X12 and so GLn is an afne algebraic group with coordinate ring4 kX.det.X ;:::;Xnn ;Y . It is called the ij /Y 1/ general linear group. For f 2 O.GLn / and a D .aij / 2 GLn .R/,
fR .aij / D f .a11 ; : : : ; ann ; det.aij /
/:
Alternatively, let A be the k-algebra in 2n2 symbols, X11 ; X12 ; : : : ; Xnn ; Y11 ; : : : ; Ynn modulo the ideal generated by the n2 entries of the matrix .Xij /.Yij / I . Then Homk-alg .A; R/ D f.A; B/ j A; B 2 Mn .R/; AB D I g:
The map .A; B/ 7! A projects this bijectively onto fA 2 Mn .R/ j A is invertibleg (because a right inverse of a square matrix is unique if it exists, and is also a left inverse). Therefore A ' O.GLn /.
other words, O.GLn / is the ring of fractions of kX11 ; X12 ; : : : ; Xnn for the multiplicative subset generated by det.Xij /, O.GLn / D kX11 ; X12 ; : : : ; Xnn det.Xij / : See CA, Example 6.2.
4 In
2. Denitions; examples E XAMPLE 2.20 Let C be an invertible n n matrix with entries in k, and let G.R/ D fT 2 GLn .R/ j T t C T D C g. If C D .cij /, then G.R/ consists of the invertible matrices .tij / such that X tj i cj k tkl D ci l ; i; l D 1; : : : ; n;
j;k
17
and so G.R/ ' Homk-alg .A; R/ with A equal to the quotient of kX11 ; X12 ; : : : ; Xnn ; Y by the ideal generated by the polynomials det.Xij /Y 1 P j;k Xj i cj k Xkl ci l ; i; l D 1; : : : ; n:
Therefore G is an afne algebraic group. When C D I , it is the orthogonal group On , and when 0 C D I I , it is the symplectic group Spn . 0 E XAMPLE 2.21 There are abstract versions of the last groups. Let V be a nite-dimensional vector space over k, let be a nondegenerate symmetric bilinear form V V ! k, and let be a nondegenerate alternating form V V ! k. Then there are afne algebraic groups with SLV .R/ D fR-linear automorphisms of R k V with determinant 1g, GLV .R/ D fR-linear automorphisms of R k V g, O. /.R/ D f 2 GLV .R/ j .v; w/ D .v; w/ for all v; w 2 R k V g; Sp. /.R/ D f 2 GLV .R/ j .v; w/ D .v; w/ for all v; w 2 R k V g.
A choice of a basis for V denes an isomorphism of each of these functors with one of those in (2.18), (2.19), or (2.20), which shows that they are afne algebraic groups. E XAMPLE 2.22 Let K be a separable k-algebra of degree 2. There is a unique k-automorphism a 7! a of K such that a D a if and only if a 2 k. There are only two possibilities: N N (a) K is a separable eld extension of k of degree 2 and a 7! a is the nontrivial element of the N Galois group, or (b) K D k k and .a; b/ D .b; a/: N For an n n matrix A D .aij / with entries in K, dene A to be .aij / and A to be the transpose N of A. Then there is an algebraic group G over k such that G.k/ D fA 2 Mn .K/ j A A D I g: More precisely, for a k-algebra R, dene a r D a r for a r 2 K k R, and, with the obvious N notation, let G.R/ D fA 2 Mn .K k R/ j A A D I g: Note that A A D I implies det.A/ det.A/ D 1. In particular, det.A/ is a unit, and so G.R/ is a group. In case (b), G.R/ D f.A; B/ 2 Mn .R/ j AB D I g
18
and so .A; B/ 7! A is an isomorphism of G with GLn . In case (a), let e 2 K k. Then e satises a quadratic polynomial with coefcients in k. Assuming char.k/ 2, we can complete the square and choose e so that e 2 2 k and e D e. A matrix N with entries in K k R can be written in the form A C eB with A; B 2 Mn .R/. It lies in G.R/ if and only if .At eB t /.A C eB/ D I i.e., if and only if At A AB
t
e 2 BB t D I; B A D 0:
t
and
Evidently, G is represented by a quotient of k: : : ; Xij ; : : : k k: : : ; Yij ; : : :. E XAMPLE 2.23 There exists an afne algebraic group G, called the group of monomial matrices, such that, when R has no nontrivial idempotents, G.R/ is the group of invertible matrices in Mn .R/ having exactly one nonzero element in each row and column. For each 2 Sn (symmetric group), let A D O.GLn /=.Xij j j .i // Q and let O.G/ D 2Sn A . Then A ' kX1 and so G.R/ ' G Homk-alg .A ; R/ ' Homk-alg .O.G/; R/:
.1/ ; : : : ; Xn .n/ ; Y =.sign.
/ X1
.1/
Xn
.n/ Y
1/;
E XAMPLE 2.24 The determinant denes a homomorphism of algebraic groups detW GLn ! Gm : E XAMPLE 2.25 The homomorphisms R ! SL2 .R/; 1 a a 7! ; 0 1
dene a homomorphism of algebraic groups Ga ! SL2 .
2c
Afne monoids
Recall that a monoid is a set M together with an associative binary operation M M ! M and an identity element (usually denoted 0, 1, or e). In other words, it is a group without inverses. A homomorphism of monoids is a map 'W M ! M 0 such that (a) '.eM / D eM 0 , and (b) '.xy/ D '.x/'.y/ for all x; y 2 M . When M 0 is a group, (a) holds automatically because a group has only one element such that ee D e. For any monoid M , the set M of elements in M with inverses is a group (the largest subgroup of M ). An afne monoid (resp. afne algebraic monoid) over k is a functor from the category of k-algebras to monoids that is representable (as a functor to sets) by a k-algebra (resp. a nitely generated k-algebra).
2. Denitions; examples E XAMPLE 2.26 For a k-vector space V , let EndV be the functor R .EndR-lin .R k V /; /:
19
When V is nite dimensional, we saw in (2.17) that, as a functor to sets, EndV is represented by Sym.V k V _ /, and so it is an algebraic monoid. On choosing a basis e1 ; : : : ; en for V , we obtain an isomorphism of EndV with the functor R .Mn .R/; / (multiplicative monoid of n n matrices),
which is represented by the polynomial ring kX11 ; X12 ; : : : ; Xnn . P ROPOSITION 2.27 For any afne monoid M over k, the functor R M over k; when M is algebraic, so also is M . P ROOF. For an abstract monoid M , let M1 D f.a; b/ 2 M M ' f..a; b/; .a0 ; b 0 // 2 M1 M.R/ is an afne group
M j ab D 1g; then M1 j a D b 0 g:
This shows that M can be constructed from M by using only bred products: M1 ? ? y M M ! f1g ? ? y M ? ? y M1 ! M1 ? ? y.a;b/7!b
.a;b/7!a
.a;b/7!ab
! M
! M:
It follows that, for an afne monoid M , the functor R M.R/ can be obtained from M by forming bre products, which shows that it is representable (see 3b below). 2 E XAMPLE 2.28 An associative k-algebra B with identity (not necessarily commutative) denes a functor R 7! .R k B; / from the category of k-algebras to monoids. If B is a nite k-algebra, then this is an afne algebraic monoid: when we choose a basis for B as a k-vector space, then (as a functor to sets) it becomes R 7! Rdimk B , which is represented by kX1 ; : : : ; Xdimk B . We let GB m denote the corresponding afne algebraic group R 7! .R B/ : For example, if B D Endk-lin .V /, then GB D GLV . m
2d
Afne supergroups
The subject of supersymmetry was introduced by the physicists in the 1970s as part of their search for a unied theory of physics consistent with quantum theory and general relativity. Roughly speaking, it is the study of Z=2Z-graded versions of some of the usual objects of mathematics. We explain briey how it leads to the notion of an afne supergroup. A superalgebra over a eld k is a Z=2Z-graded associative algebra R over k. In other words, R is an associative k-algebra equipped with a decomposition R D R0 R1 (as a k-vector space) such that k R0 and Ri Rj Ri Cj (i; j 2 Z=2Z). An element a of R is said to be even, and have parity p.a/ D 0, if it lies in R0 ; it is odd, and has parity p.a/ D 1, if it lies in R1 . The homogeneous
20
elements of R are those that are either even or odd. A homomorphism of super k-algebras is a homomorphism of k-algebras preserving the parity of homogeneous elements. A super k-algebra R is said to be commutative if ba D . 1/p.a/p.b/ ab for all a; b 2 R. Thus even elements commute with all elements, but for odd elements a; b, ab C ba D 0. The commutative super k-algebra kX1 ; : : : ; Xm ; Y1 ; : : : ; Yn in the even symbols Xi and the odd symbols Yi is dened to be the quotient of the k-algebra of noncommuting polynomials in X1 ; : : : ; Yn by the relations Xi Xi 0 D Xi 0 Xi ; X i Yj D Yj X i ; Y j Yj 0 D Yj 0 Yj ; 1 i; i 0 m; 1 j; j 0 n:
When n D 0, this is the polynomial ring in the commuting symbols X1 ; : : : ; Xm , and when m D 0, it is the exterior algebra of the vector space with basis fY1 ; : : : ; Yn g provided 2 0 in k. A functor from the category of commutative super k-algebras to groups is an afne supergroup if it is representable (as a functor to sets) by a commutative super k-algebra. For example, for m; n 2 N, let GLmjn be the functor AB R B 2 Mm;n .R1 /; C 2 Mn;m .R1 /; D 2 GLn .R0 / : C D A 2 GLm .R0 /;
A B It is known that such a matrix C D is invertible (Varadarajan 2004, 3.6.1), and so GLmjn is a functor to groups. It is an afne supergroup because it is represented by the commutative super k-algebra obtained from the commutative super k-algebra kX11 ; X12 ; : : : ; XmCn;mCn ; Y; Z in the even symbols Y; Z; Xij .1 i; j m; m C 1 i; j m C n/
and the odd symbols Xij by setting Y .det.Xij /1i;j m D 1; Z det.Xij /mC1i:j mCn D 1: Much of the theory of afne groups extends to afne supergroups (see, for example, Fioresi and Gavarini 2008). .remaining pairs .i; j /)
2e
Relaxing the hypothesis on k
Everything in this section holds with only minor changes when k is not a eld. Let k be a commutative ring. A functor from commutative k-algebras to groups is an afne group (resp. an afne algebraic group) if it is representable (as a functor to sets) by a k-algebra (resp. by a nitely presented k-algebra). Recall that a k-algebra is nitely presented if it is isomorphic to the quotient of a polynomial ring kX1 ; : : : ; Xn by a nitely generated ideal. Thus, as a functor to sets, an afne algebraic group G is described by a nite set of polynomials in a nite number of variables with coefcients in k. The only other changes are that vector space over k must be replaced by module over k and nite-dimensional vector space over k by nitely generated projective module over k. The explanation for the second change is provided by the statements (2.30) and (2.31) below.
21
In more detail, no changes to the text through Example 2.16 are needed. In the second paragraph of Example 2.17, one should take V to be a nitely generated projective k-module,5 and the nal two sentences make sense only for free k-modules. In Example 2.21 one should take V to be a nitely generated projective k-module, and the proof that functors are afne algebraic groups is a little more complicated. In Example 2.26, one should take V to be a nitely generated projective k-module, and the nal sentence makes sense only when V is free. In order for GB to be an afne algebraic group in (2.28), m one should require B to be nitely generated and projective as a k-module (DG II, 1, 2.3, p.149). E XAMPLE 2.29 Let k D k1 kn , and write 1 D e1 C set of orthogonal idempotents in k. For any k-algebra R, R D R1 Rn C en . Then fe1 ; : : : ; en g is a complete
where Ri is the k-algebra Rei . To give an afne group G over k is the same as giving an afne group Gi over each ki . If G $ .Gi /1i n , then Y G.R/ D Gi .Ri /
i
for all k-algebras R D R1
Rn .
F INITELY GENERATED PROJECTIVE MODULES OVER COMMUTATIVE RINGS

In many situations, the correct generalization of nite-dimensional vector space is not nitely generated module but nitely generated projective module. P ROPOSITION 2.30 Let A be a commutative ring. The following conditions on an A-module M are equivalent: (a) M is nitely generated and projective; (b) M is nitely presented and Mm is a free Am -module for every maximal ideal m; (c) there exists a nite family .fi /i 2I of elements of A generating the ideal A and such that, for all i 2 I , the Afi -module Mfi is free of nite rank; (d) M is nitely presented and at. P ROOF. See CA 10.4.
2
P ROPOSITION 2.31 For any A-modules M , S , T with M nitely generated and projective, the canonical maps HomA-lin .S; T A M / ! HomA-lin .S A M _ ; T / M T
_ _
(12) (13) (14)
! .M T /
__
M !M are isomorphisms.
P ROOF. When M is free of nite rank the statements are easy to prove. For example, when M D Ar and (13) becomes the isomorphism .T _ /r ! .T r /_ . Since the nitely generated projective modules are exactly the direct summands of free modules of nite rank, the general case follows. 2
fact, the functor R R k V is representable only if V is nitely generated and projective (cf. EGA I (1971) 9.4.10, where this is stated without proof).
5 In
22
2f
Terminology
From now on algebraic group will mean afne algebraic group and algebraic monoid will mean afne algebraic monoid.
Some basic constructions
Throughout this section, k is a commutative ring.
3a
Products of afne groups
Let G and H be afne groups over k. The functor R is an afne group G G.R/ H.R/
H over k with coordinate ring O.G H / D O.G/ k O.H /; (15)
because, for any k-algebras A, B, R, Homk-alg .A k B; R/ ' Homk-alg .A; R/ Homk-alg .B; R/ (16)
(CA (2), p.27). More generally, let .Gi /i 2I be a (possibly innite) family of afne groups over k, and let G be the functor Y R Gi .R/: i2I N Then G is an afne group with coordinate ring i 2I O.Gi / (in the innite case, apply Bourbaki A, III 5, Prop. 8). Moreover, G together with the projection maps is the product of the Gi in the Q category of afne groups. If each Gi is an algebraic group and I is nite, then i 2I Gi is algebraic.
3b
Fibred products of afne groups
Let G1 , G2 , and H be functors from the category of k-algebras to sets, and let G1 ! H G2 (17)
be natural transformations. We dene the bred product functor G1 H G2 to be the functor sending a k-algebra R to the set of pairs .g1 ; g2 / 2 G1 .R/ G2 .R/ such that g1 and g2 have the same image in H.R/, i.e., .G1
H
G2 / .R/ D G1 .R/
H.R/ G2 .R/:
Obviously G1 H G2 is the bred product of (17) in the category of functors from Algk to Set. Let B be a k-algebra, and let A1 and A2 be B-algebras. The k-algebra A1 B A2 is the largest quotient of A1 k A2 such that the diagram B ? ? y A1 ! A2 ? ? ya7!1a
a7!a1
! A1 B A2
commutes. More precisely, for any k-algebra R, Homk-alg .A1 B A2 ; R/ ' Homk-alg .A1 ; R/ 23
Homk-alg .B;R/ Homk-alg .A2 ; R/:
(18)
24
Therefore, if the functors G1 , G2 , and H in (17) are represented by k-algebras A1 , A2 , and B, then G1 H G2 is represented by the k-algebra A1 B A2 . When the natural transformations G1 ! H G2 are homomorphisms of afne groups, G1 H G2 is a functor to Grp, and the above remark shows that it is an afne group with coordinate ring O.G1
H
G2 / D O.G1 / O.H / O.G2 /.
(19)
If G1 and G2 are algebraic groups, so also is G1 H G2 . For example, let H be an afne group and let ! H be the unique homomorphism from the trivial group to H . For any homomorphism W G ! H; .G The afne group .G
H H
/ .R/ D Ker..R/W G.R/ ! H.R//:
/ is called the kernel of , and is denoted Ker./. Note that O.Ker.G ! H // D O.G/ O.H / k: (20)
3c
Extension of the base ring (extension of scalars)
Let G be an afne group over k, and let k 0 be a k-algebra. A k 0 -algebra R can be regarded as a kalgebra through k ! k 0 ! R, and so G.R/ is dened. For any k-algebra A, Homk 0 -alg .k 0 k A; R/ is the subset of Homk-alg .k 0 k A; R/ consisting of the homomorphisms that restrict to the given homomorphism k 0 ! R on k 0 . Therefore (18) shows that Homk 0 -alg .k 0 k A; R/ ' Homk-alg .A; R/. In particular, Homk 0 -alg .k 0 k O.G/; R/ ' Homk-alg .O.G/; R/ .' G.R// ;
and so, on restricting the functor G to k 0 -algebras, we obtain an afne group Gk 0 over k 0 with coordinate ring O.Gk 0 / D k 0 k O.G/. The afne group Gk 0 is said to have been obtained from G by extension of the base ring or by extension of scalars. If G is an algebraic group, so also is Gk 0 . Homomorphisms of afne groups over k give homomorphisms of afne groups over k 0 , and so G Gk 0 is a functor. E XAMPLE 3.1 Let V be a nitely generated projective k-module, and let Vk 0 D k 0 k V ; then .Va /k 0 ' .Vk 0 /a . E XAMPLE 3.2 Let G be the unitary group dened by a separable k-algebra K of degree 2 (see 2.22). For any eld extension k ! k 0 , Gk 0 is the unitary group dened by the k 0 -algebra K k k 0 , and so, for example, Gk al ' GLn .
3d
Restriction of the base ring (restriction of scalars)
Throughout this subsection, k 0 is a k-algebra that is nitely generated and projective as a k-module. We shall show that there is a right adjoint to the functor G Gk 0 . We rst explain this for functors to sets. As above, a functor F W Algk ! Set denes a functor Fk 0 W Algk 0 ! Set with Fk 0 .R/ D F .R/. On the other hand, a functor F 0 W Algk 0 ! Set denes a functor .F 0 /k 0 =k W Algk ! Set with .F 0 /k 0 =k .R/ D F 0 .k 0 k R/. A natural transformation 'W Fk 0 ! F 0 denes homomorphisms F .R/
F .r7!1r/
F .k 0 k R/
'.k 0 k R/
F 0 .k 0 k R/ D .F 0 /k 0 =k .R/
def
3. Some basic constructions
25
which are natural in the k-algebra R, and so form a natural transformation F ! .F 0 /k 0 =k . Thus, we have a morphism Hom.Fk 0 ; F 0 / ! Hom.F; .F 0 /k 0 =k /: This has an obvious inverse6 , and so it is a bijection. We have shown that the extension of scalars functor F Fk 0 has a right adjoint F 0 .F 0 /k 0 =k : Hom.Fk 0 ; F 0 / ' Hom.F; .F 0 /k 0 =k /: (21)
Because it is a right adjoint, F 0 .F 0 /k 0 =k preserves inverse limits. In particular, it takes (bred) products to (bred) products. This can also be checked directly. L EMMA 3.3 If F W Algk 0 ! Set is represented by a (nitely-generated) k-algebra, then so also is .F /k 0 =k . P ROOF. We prove this only in the case that k 0 is free as a k-module, say, k 0 D ke1 R, R0 D k 0 k R ' Re1 and so there is a bijection .ai /1i n 7! .bij / 1i n W R0n ! Rnd
1j d
def
ked ;
ei 2 k 0 :
Consider rst the case that F D An , so that F .R/ D Rn for all k 0 -algebras R. For any k-algebra Red ;
which sends .ai / to the family .bij / dened by the equations ai D Pd

j D1 bij ej ;
i D 1; : : : ; n.
(22)
The bijection is natural in R, and shows that .F /k 0 =k And (the isomorphism depends only on the choice of the basis e1 ; : : : ; ed ). Now suppose that F is the subfunctor of An dened by a polynomial f .X1 ; : : : ; Xn / 2 k 0 X1 ; : : : ; Xn . On substituting P Xi D dD1 Yij ej j into f , we obtain a polynomial g.Y11 ; Y12 ; : : : ; Ynd / with the property that f .a1 ; : : : ; an / D 0 g.b11 ; b12 ; : : : ; bnd / D 0 when the as and bs are related by (22). The polynomial g has coefcients in k 0 , but k 0 Y11 ; : : : D kY11 ; : : :e1 and so we can write it as a sum g D g1 e1 C
6 Given
kY11 ; : : :ed ;
C gd ed ;
gi 2 kY11 ; Y12 ; : : : ; Ynd :
F ! .F 0 /k 0 =k , we need Fk 0 ! F 0 . Let R be a k 0 -algebra, and let R0 be R regarded as a k-algebra. The given k-algebra map k 0 ! R and the identity map R0 ! R dene a map k 0 k R0 ! R (of k 0 -algebras). Hence we have F .R0 / ! F 0 .k 0 k R0 / ! F 0 .R/; and F .R0 / D Fk 0 .R/:
26 Clearly,
g.b11 ; b12 ; : : : ; bnd / D 0 gi .b11 ; b12 ; : : : ; bnd / D 0 for i D 1; : : : ; d , and so .F /k 0 =k is isomorphic to the subfunctor of And dened by the polynomials g1 ; : : : ; gd . This argument extends in an obvious way to the case that F is the subfunctor of An dened by a nite set of polynomials, and even to the case that it is a subfunctor of an innite dimensional afne space dened by innitely many polynomials. 2
A SIDE 3.4 In order to extend the proof to the general case it sufces to show: Let f1 ; : : : ; fr 2 k be such that .f1 ; : : : ; fr / D k, and let F be a functor Algk ! Set. If the restriction of F to Algkf is representable for all i , then F itself is representable.
i
We sketch a proof of this. Let Ai represent F jAlgkf . Then kfi fj kf Ai ' kfi fj kf Aj for all i; j i j i because they both represent the restriction of F to Algkf f . The Ai patch together (on Spec k) to give a i j k-algebra A such that Afi ' Ai for all i, and that A represents F . This is obvious in terms of sheaves. (Add details.)
If G is a functor Algk 0 ! Grp, then .G/k 0 =k is a functor Algk ! Grp. The lemma shows that if G is an afne (algebraic) group, then so also is .G/k 0 =k , and (21) shows that the functor G 0 .G 0 /k 0 =k is right adjoint to the functor extension of scalars: Hom.G; .G 0 /k 0 =k / ' Hom.Gk 0 ; G 0 /: The afne group .G/k 0 =k is said to have been obtained from G by (Weil) restriction of scalars (or by restriction of the base ring). It is sometimes denoted Resk 0 =k G or k 0 =k G, and called the Weil restriction of G. 3.5 For any homomorphisms k ! k 0 ! k 00 of rings such that k 0 (resp. k 00 ) is nitely generated and projective over k (resp. k 0 ), k 0 =k k 00 =k 0 ' k 00 =k . Indeed, for any afne group G over k 00 and k-algebra R, k 0 =k k 00 =k 0 .G/ .R/ D k 0 =k .k 00 =k 0 G/ .R/ D G.k 00 k 0 k 0 k R/ ' G.k 00 k R/ D k 00 =k G .R/ because k 00 k 0 k 0 k R ' k 00 k R. 3.6 For any k-algebra K and any afne group G over k 0 , k 0 =k G
K
' k 0 k K=K .GK /I
(23)
in other words, Weil restriction commutes with base extension. Indeed, for a K-algebra R, k 0 =k G
K
.R/ D G.k 0 k R/ ' G.k 0 k K K R/ D k 0 k K=K .GK /.R/
def
def
because k 0 k R ' k 0 k K K R.
3. Some basic constructions
27
0 0 kn . Let G be the afne group over k 0 corresponding to a family .Gi /i of 3.7 Let k 0 D k1 afne groups over the ki0 (see 2.29). Then
0 .G/k 0 =k ' .G1 /k1 =k 0 .Gn /kn =k .
(24)
Indeed, for any k-algebra R, .G/k 0 =k .R/ D G.k 0 R/

0 D G1 .k1 R/
0 D .G1 /k1 =k
def
def
0 Gn .kn R/ 0 .Gn /kn =k .R/
0 because k 0 R ' k1 R
0 kn R.
3.8 There is a homomorphism iW G ! .k 0 =k G/k 0 of afne groups over k 0 such that, for all k 0 algebras R, i.R/ is the map G.R/ ! k 0 =k G .R/ D G.k 0 R/ dened by a 7! 1 aW R ! k 0 R. Then i is injective (obviously), and has the following universal property: let H be an afne group over k; then any homomorphism G ! Hk 0 (over k 0 ) factors uniquely through i . 3.9 Let k 0 be a nite separable eld extension of a eld k, and let K be a eld containing all k-conjugates of k 0 , i.e., such that jHomk .k 0 ; K/j D k 0 W k. Then Y G k 0 =k G K ' 0
Wk !K
def
where G is the afne group over K obtained by extension of scalars W k 0 ! K. Indeed k 0 =k G because k 0 K ' K Homk .k ;K/ . 3.10 (Add a discussion of G .G/k 0 =k when k 0 is purely inseparable over k.)
0
(23) K
' k 0 K=K GK '
(24)
Y
Wk 0 !K
3e
The Greenberg functor
Let A be a local artinian ring with residue eld k. For example, A could be the ring Wm .k/ of Witt vectors of length m. In general, A is a Wm .k/-module for some m. For an afne group G over A, consider the functor G.G/: R G.A Wm .k/ Wm .R//: Then G.G/ is an afne group over k. See Greenberg 1961, Greenberg 1963. (To be extended.)
Afne groups and Hopf algebras

Un principe g n ral: tout calcul relatif aux e e cog` bres est trivial et incompr hensible. e e Serre 1993, p39..
In this section, we examine the extra structure that the coordinate ring of an afne group G acquires from the group structure on G. Throughout k is a commutative ring. Unadorned tensor products are over k.
4a
Algebras
Recall that an associative algebra over k with identity is a module A over k together with a pair of k-linear maps7 mW A A ! A eW k ! A such that the following diagrams commute:
Am
AAA
mA
/ AA
m
k AM
AA
/A
eA Ae / AA o Ak MMM oo MMM ooo o MM m ooo ' MMM ooo ' MM& oo wo
(25)
associativity
A identity
On reversing the directions of the arrows, we obtain the notion of a co-algebra.
4b
Co-algebras
D EFINITION 4.1 A co-associative co-algebra over k with co-identity (henceforth, a co-algebra over k) is a module C over k together with a pair of k-linear maps WC ! C C such that the diagrams C C C o O
C C
WC ! k
C C O
k CfMo
C C o
C co-associativity
MMM MMM MM ' MMM MM
C C O
C co-identity
/ C k oo7 oo ooo ooo oo ' ooo
(26)
commute, i.e., such that .C / .C /

7 Warning:
D . C/ D idC D . C / :
(27)
I sometimes use e for the neutral element of G.k/ and also for the homomorphism k ! O.G/ that makes O.G/ into a k-algebra.
28
4. Afne groups and Hopf algebras A homomorphism of co-algebras over k is a k-linear map f W C ! D such that the diagrams C C x ? ? C C commute, i.e., such that ( .f f /
C f f
29
! D D x ? ? D ! D D D
D f C.
C ? ? y k
! D ? ? yD k
D f
Let D be a subspace of a co-algebra C such that .D/ D D. Then .D; jD; jD/ is a co-algebra (obvious), and the inclusion D ,! C is a co-algebra homomorphism. Such a D is called a sub-co-algebra of C . Let .C; C ; C / and .D; D ; D / be co-algebras over k; dene C D to be the composite C D
C
! C C k D D
C D
C t D
'
C D k C D to be the composite
where t is the transposition map c d 7! d c, and dene C D then .C D; C D ; co-algebra over k.

C D /
C D
! k k ' k;
is a co-algebra over k. On taking D D C , we see that C C is a
4c
The duality of algebras and co-algebras

.f g/.v w/ D f .v/ g.w/; f 2 V _ , g 2 W _ , v 2 V , w 2 W; (28)
Recall that V _ denotes the dual of a k-module V . If V and W are k-modules, then the formula
denes a linear map V _ W _ ! .V W /_ which is always injective, and is an isomorphism when at least one of V or W is nitely generated and projective (2.31). If .C; ; / is a co-associative co-algebra over k with a co-identity, then C _ becomes an associative algebra over k with the multiplication C _ C _ ,! .C C /_ ! C _ and the identity k ' k _ ! C _ . Similarly, if .A; m; e/ is an associative algebra over k with an identity and A is nitely generated and projective as a k-module, then A_ becomes a co-associative co-algebra over k with the co-multiplication A_ ! .A A/_ ' A_ A_ and the co-identity k ' k _ ! A_ . These statements are proved by applying the functor _ to one of the diagrams (25) or (26). E XAMPLE 4.2 Let X be a set, and let C be the free k-module with basis X . The k-linear maps W C ! C C; W C ! k; .x/ D x x; .x/ D 1; x 2 X; x 2 X;
m_
_ _ _
endow C with the structure of co-algebra over k. The dual algebra C _ can be identied with the k-module of maps X ! k endowed with the k-algebra structure m.f; g/.x/ D f .x/g.x/ e.c/.x/ D cx:
30
4d
Bi-algebras
For k-algebras A and B, A B becomes a k-algebra with the maps mAB ..a b/ .a0 b 0 // D mA .a a0 / mA .b b 0 / eAB .c/ D eA .c/ 1 D 1 eB .c/: D EFINITION 4.3 A bi-algebra over k is a k-module with compatible structures of an associative algebra with identity and of a co-associative co-algebra with co-identity. In detail, a bi-algebra over k is a quintuple .A; m; e; ; / where (a) .A; m; e/ is an associative algebra over k with identity e; (b) .A; ; / is a co-associative co-algebra over k with co-identity ; (c) W A ! A A is a homomorphism of algebras; (d) W A ! k is a homomorphism of algebras. The next proposition shows that the denition is self dual. P ROPOSITION 4.4 For a quintuple .A; m; e; ; / satisfying (a) and (b) of (4.3), the following conditions are equivalent: (a) and are algebra homomorphisms; (b) m and e are co-algebra homomorphisms. P ROOF Consider the diagrams: AA

At A m
/A
/ AA O
mm
(29)
AAAA
/ AAAA /A
AA o O
ee
A O
e '
AA
k k o
k k
'
/k
A3 E 333 e 33 33 id /k k
(30)
The rst and second diagrams commute if and only if is an algebra homomorphism, and the third and fourth diagrams commute if and only if is an algebra homomorphism. On the other hand, the rst and third diagrams commute if and only if m is a co-algebra homomorphism, and the second and fourth commute if and only if e is a co-algebra homomorphism. Therefore, each of (a) and (b) is equivalent to the commutativity of all four diagrams. 2 A bi-algebra is said to be commutative (resp. nitely presented) if its underlying k-algebra is commutative (resp. nitely presented). Note that these notions are not self dual. A homomorphism of bi-algebras .A; m; : : :/ ! .A0 ; m0 ; : : :/ is a k-linear map A ! A0 that is both a homomorphism of k-algebras and a homomorphism of k-co-algebras.
4. Afne groups and Hopf algebras An inversion (or antipode) for a bi-algebra A is a k-linear map S W A ! A such that A x ?e ? k commutes, i.e., such that m .S id/ D e D m .id S / .
m.Sid/
31
AA x ? ? A
m.id S /
! A x ?e ? k
(31)
D EFINITION 4.5 A Hopf algebra over k is a bi-algebra over k together with an inversion. A homomorphism of Hopf algebras .A; m; : : :/ ! .A0 ; m0 ; : : :/ is a homomorphism f of bi-algebras such that f S D S 0 f . E XAMPLE 4.6 Let X be a monoid, and let A be the k-module with basis X. The k-linear maps mW A A ! A; eW k ! A; m.x x 0 / D xx 0 ; e.c/ D c1X ; x; x 0 2 X; c 2 k;
endow A with the structure of a k-algebra. When combined with co-algebra structure in (4.2), this makes A into a bi-algebra over k. When X is a group, the map SW A ! A; is an inversion. The reader encountering bi-algebras for the rst time should do Exercise 4-1 below before continuing.
A SIDE 4.7 To give a k-bialgebra that is nitely generated and projective as a k-module is the same as giving a pair of nitely generated projective k-algebras A and B together with a nondegenerate k-bilinear pairing h ; iW B A!k
.Sf /.x/ D f .x
satisfying compatibility conditions that we leave to the reader to explicate.
4e
Afne monoids and bi-algebras
An afne monoid over k can be regarded as a representable functor from the category of commutative k-algebras to sets together with natural transformations mW M M !M eW ! M (32)
such that the following diagrams commute: M M

idM
/M
M
m
MM M
e idM
m idM
/M
MMM ooo MMM ooo o m M ooo ' MMM MM& ooooo ' wo
/M
M o
idM
M (33)
32
(cf. 2.8). Here denotes the trivial monoid, which is represented by k. If A represents M , then A A represents the functor M M , and so the maps in (32) can be rewritten as mW hAA ! hA eW hk ! hA :
The Yoneda lemma shows that these maps arise from unique homomorphisms of k-algebras WA ! AA W A ! k;
and that the commutativity of the diagrams (33) is equivalent to that of the diagrams (26) (with A for C ). In other words, the structure .m; e/ of a monoid on the set-valued functor M denes the structure of a bi-algebra on the k-algebra A representing it. Conversely, for a commutative k-algebra A, the structure of a bi-algebra on A denes the structure of an afne monoid on hA . Explicitly, the product of two elements g; h 2 G.R/ D Homk-alg .A; R/ is the composite of A ! A A ! R, and the identity element of G.R/ is the composite of A ! k ! R: A homomorphism M ! N of afne monoids denes a homomorphism O.N / ! O.M / of kbi-algebras (by the Yoneda lemma again). Therefore, to give an afne monoid over k is essentially the same as giving a commutative bi-algebra over k. More precisely, the functor M O.M / is a contravariant equivalence from the category of afne monoids over k to the category of commutative k-bi-algebras. The afne monoid M is algebraic if and only if the k-bi-algebra O.M / is nitely presented.
.g;h/
(34)
4f
Afne groups and Hopf algebras
An afne monoid M is an afne group if and only if there exists a natural transformation invW M ! M such that M ? ? y
.inv;id/
! M
M ? ?m y
.id;inv/
M ? ? y
(35)
commutes. Here is the trivial monoid and .id; inv/ denotes the morphism whose composites with the projection maps are id and inv. Such a natural transformation corresponds to a k-linear map SW A ! A satisfying (31), i.e., to an inversion for A. We conclude that to give an afne group over k is essentially the same as giving a commutative bi-algebra over k with an inversion. More precisely, M O.M / is a contravariant equivalence from the category of afne groups over k to the category of commutative k-bi-algebras with inversion. The afne group M is algebraic if and only if the k-bi-algebra O.M / is nitely presented. P ROPOSITION 4.8 (a) For any commutative k-algebra .A; m; e/ and homomorphism W A ! A A, there exists at most one pair . ; S / such that .A; m; e; ; ; S / is a Hopf algebra. (b) Let .A; m; : : :/ and .A0 ; m0 ; : : :/ be commutative Hopf algebras over k. A homomorphism f W .A; m; e/ ! .A0 ; m0 ; e 0 / is a homomorphism of Hopf algebras if .f f / D 0 f .
4. Afne groups and Hopf algebras
33
P ROOF. (a) follows from (b) applied to the identity map. (b) Let G and G 0 be the afne groups over k with Hopf algebras .A; m; : : :/ and .A0 ; m0 ; : : :/. A homomorphism f W .A; m; e/ ! .A0 ; m0 ; e 0 / denes a morphism of functors G ! G 0 . If .f f / D 0 f , then this morphism sends products to products, and therefore is a morphism of group valued functors (because any map of abstract groups sending products to products is a homomorphism). Therefore f is a homomorphism of Hopf algebras. 2 In view of the proposition, it would be more natural to dene a commutative Hopf algebra to be a quadruple .A; m; e; / such that .A; m; e/ is a commutative k-algebra and is a homomorphism A ! A A such that there exists a pair . ; S / making .A; m; e; ; ; S / a bi-algebra with involution. R EMARK 4.9 (a) For any commutative k-algebra .A; m; e/ and homomorphism W A ! A A, there exists at most one such that .A; m; e; ; / is a bi-algebra (because an abstract monoid can have at most one neutral element). (b) There can exist more than one Hopf algebra structure on a commutative k-algebra (see 7.31).
4g
Abstract restatement
Let C be a category with nite products and, in particular, a nal object (the product over the empty set). A monoid object in C is an object M together with morphisms mW M M ! M and eW ! M such that the diagrams (33) commute. A morphism of monoid objects is a morphism of the objects compatible with the additional data. Let A be a category, and let A_ be the category of functors A ! Set. For any nite family Q _ .Fi /i 2I of functors, the functor A i 2I Fi .A/ is the product of the Fi , and so A has nite products. To give the structure of a monoid object on a functor M W A ! Set is to give a factorization of M through Mon. Now assume that A has nite direct sums. It follows from the denitions of direct sums and products, that the functor A hA sends direct sums to direct products. According to the Yoneda lemma (2.1), A hA W Aopp ! A_ is fully faithful. Its essential image is (by denition) the subcategory of representable functors. Therefore A hA is an equivalence from the category of monoid objects in Aopp to the category of monoid objects in A_ whose underlying functor to sets is representable (equivalently, to the category of functors A ! Mon whose underlying functor to sets is representable). Now take A D Algk . Tensor products in this category are direct sums (in the sense of category theory), and so the above remarks show that A hA is an equivalence from the category of monoid opp objects in Algk to the category of afne monoids over k. On comparing the diagrams (26) and opp (33), we see that a monoid object in Algk is just a commutative bi-algebra. Similarly, a group object in a category C with nite products is dened to be an object M together with morphisms m, e, and invW M ! M such that the diagrams (33) and (35) commute.8 The same arguments as above show that A hA is an equivalence from the category of group opp opp objects in Algk to the category of afne groups over k. Moreover, a group object in Algk is just a commutative bi-algebra with an inversion. In summary: the functor A hA denes an equivalence from the category of commutative bi-algebras (resp. commutative Hopf algebras) to the category of afne monoids (resp. groups). Under the equivalence, nitely presented bi-algebras (resp. Hopf algebras) correspond to algebraic monoids (resp. groups).
any object T of C, the maps m, e, and inv dene a group structure on Hom.T; M /. The Yoneda lemma shows that inv is uniquely determined by m and e. Thus, one can also dene a group object to be a monoid object for which there exists a morphism inv such that the diagram (35) commutes.
8 For
34
4h
Explicit description of
, , and S
Let G be an afne group over k. Recall (2.11) that an element f of the coordinate ring O.G/ is a family of functions fR W G.R/ ! R of sets compatible with homomorphisms of k-algebras. An element f1 f2 of O.G/ O.G/ denes a function .f1 f2 /R W G.R/ G.R/ ! R by the rule: .f1 f2 /R .a; b/ D .f1 /R .a/ .f2 /R .b/: In this way, O.G/ O.G/ becomes identied with the coordinate ring of G G. For f 2 O.G/, .f / is the (unique) element of O.G/ O.G/ such that . f /R .a; b/ D fR .ab/; Moreover, f D f .1/ (constant function), and Sf is the element of O.G/ such that .Sf /R .a/ D fR .a
1
for all R and all a; b 2 G.R/:
(36)
(37)
/;
for all R and all a 2 G.R/:
(38)
E XAMPLE 4.10 Recall (2.12) that Ga has coordinate ring kX with f .X / 2 kX acting as a 7! f .a/ on Ga .R/ D R. The ring kX kX is a polynomial ring in X1 D X 1 and X2 D 1 X, kX kX ' kX1 ; X2 ; and so Ga Ga has coordinate ring kX1 ; X2 with F .X1 ; X2 / 2 kX1 ; X2 acting as .a; b/ 7! F .a; b/ on G.R/ G.R/. As . f /R .a; b/ D fR .a C b/ (see (36)), we nd that . f /.X1 ; X2 / D f .X1 C X2 /; f 2 O.Ga / D kX I
in other words, is the homomorphism of k-algebras kX ! kX kX sending X to X 1 C 1 X. Moreover, f D f .0/ .D constant term of f /; and .Sf /R .a/ D fR . a/, so that .Sf /.X / D f . X /: E XAMPLE 4.11 For G D Gm , O.G/ D kX; X 1 , is the homomorphism of k-algebras kX; X 1 ! kX; X 1 kX; X 1 sending X to X X, is the homomorphism kX ! k sending f .X; X 1 / to f .1; 1/, and S is the homomorphism kX; X 1 ! kX; X 1 sending X to X 1 . E XAMPLE 4.12 For G D GLn , O.G/ D and ( xi k D P
j D1;:::;n
kX11 ; X12 ; : : : ; Xnn ; Y D kx11 ; : : : ; xnn ; y .Y det.Xij / 1/
xij xj k
y D y y
8 < .xi i / D 1 .xij / D 0, i j : .y/ D 1
S.xij / D yaj i S.y/ D det.xij /
4. Afne groups and Hopf algebras where aj i is the cofactor of xj i in the matrix .xj i /. Symbolically, we can write the formula for as .x/ D .x/ .x/ where .x/ is the matrix with ij th entry xij . We check the formula for . xi k /R .aij /; .bij / D .xi k /R .aij /.bij / P D j aij bj k P D . j D1;:::;n xij xj k /R .aij /; .bij / .xi k /: denition (36) as .xkl /R ..cij // D ckl as claimed.
35
E XAMPLE 4.13 Let F be a nite group, and let A be the set of maps F ! k with its natural kalgebra structure. Then A is a product of copies of k indexed by the elements of F . More precisely, let e be the function that is 1 on and 0 on the remaining elements of F . The e s are a complete system of orthogonal idempotents for A: e2 D e ; The maps .e / D X
; with D
e e D 0 for
e D 1.
e e ;
.e / D
1 if D 1 ; 0 otherwise
S.e / D e
dene a bi-algebra structure on A with inversion S. Let .F /k be the associated algebraic group, so that .F /k .R/ D Homk-alg .A; R/: If R has no idempotents other than 0 or 1, then a k-algebra homomorphism A ! R must send one e to 1 and the remainder to 0. Therefore, .F /k .R/ ' F , and one checks that the group structure provided by the maps ; ; S is the given one. For this reason, .F /k is called the constant algebraic group dened by F (even though for k-algebras R with nontrivial idempotents, .F /k .R/ may be bigger than F ).
4i
Commutative afne groups
A monoid or group G (resp. an algebra A) is commutative if and only if the diagram at left (resp. the middle diagram) commutes, and a co-algebra or bi-algebra C is co-commutative if the diagram at right commutes: G G ?
t
?? ?? m ? ?
/G G m
AA ?
?? ?? m ? ?
/ AA m
C C o _?
? ? ? ? ? ?

C ? k C
(39)
In each diagram, t is the transposition map .x; y/ 7! .y; x/ or x y 7! y x. On comparing the rst and third diagrams and applying the Yoneda lemma, we see that an afne monoid or group is commutative if and only if its coordinate ring is co-commutative.
36
4j
Finite at algebraic groups; Cartier duality
If .A; m; e; ; / is a bi-algebra over k and A is nitely generated and projective as a k-module, then .A_ ; _ ; _ ; m_ ; e _ / is also a k-bialgebra (see 4c and Proposition 4.4). If moreover .A; m; e; ; / is commutative (resp. co-commutative), then .A_ ; _ ; _ ; m_ ; e _ / is co-commutative (resp. commutative). An algebraic group G is said to be nite and at if O.G/ is nitely generated and projective as a k-module. The coordinate ring O.G/ of a commutative nite at algebraic monoid is a commutative co-commutative bi-algebra, and so its dual O.G/_ is the coordinate ring of a commutative nite at algebraic monoid G _ , called the Cartier dual of G. If O.G/ admits an inversion S , then S _ is an algebra homomorphism, and so G _ is an algebraic group. To check that S _ is an algebra homomorphism, we have to check that _ .S _ S _ / D S _ _ , or, equivalently, that S D .S S/ . In other words, we have check the diagram at left below commutes. This corresponds (under a category equivalence) to the diagram at right, which commutes precisely because G is commutative (the inverse of a product is the product of the inverses): O.G/ ? ? yS O.G/ Note that G __ ' G. ! O.G/ O.G/ ? ? yS S ! O.G/ O.G/ G x ? ?inv G
m
G x ? ?inv G:
inv
4k
Galois descent of afne groups
In this subsection, k is a eld. Let be a Galois extension of the eld k, and let D Gal.=k/. When is an innite extension of k, we endow with the Krull topology. By an action of on an -vector space V we mean a homomorphism ! Autk .V / such that each 2 acts -linearly, i.e., such that .cv/ D .c/ .v/ for all 2 , c 2 , and v 2 V . We say that the action is continuous if every element of V is xed by an open subgroup of [ 0 V D V (union over the open subgroups 0 of ). 0 P ROPOSITION 4.14 For any -vector space V equipped with a continuous action of P is an isomorphism. P ROOF. See AG, 16.15 or V, 1.2 (the proof is quite elementary). For any vector space V over k, the group acts continuously on V according to rule: 2 , c 2 , and v 2 V:
2 i ci
, i.e., if
, the map
vi 7!
i ci vi W k
!V
.c v/ D c v for all
P ROPOSITION 4.15 The functor V k V from vector spaces over k to vector spaces over equipped with a continuous action of is an equivalence of categories.
4. Afne groups and Hopf algebras
37
P ROOF. When we choose bases for V and V 0 , then Homk-lin .V; V 0 / and Hom-lin . V; V 0 / become identied with with certain sets of matrices, and the fully faithfulness of the functor follows from the fact that D k. That the functor is essentially surjective follows from (4.14). 2 Let G be an afne group over k. By a continuous action of on G we mean a continuous action of on O.G/ preserving its Hopf algebra structure .m; e; ; ; S /. For example, we require that . .f; g// D . f; g/ for all 2 , f; g 2 O.G/: P ROPOSITION 4.16 The functor G equipped with a continuous action of G from afne groups over k to afne groups over is an equivalence of categories.
2
P ROOF. Immediate consequence of Proposition 4.15.
4l
Quantum groups
Until the mid-1980s, the only Hopf algebras seriously studied were either commutative or cocommutative. Then Drinfeld and Jimbo independently discovered noncommutative Hopf algebras in the work of physicists, and Drinfeld called them quantum groups. There is, at present, no definition of quantum group, only examples. Despite the name, a quantum group does not dene a functor from the category of noncommutative k-algebras to groups. One interesting aspect of quantum groups is that, while semisimple algebraic groups cant be deformed (they are determined up to isomorphism by a discrete set of invariants), their Hopf algebras can be. For q 2 k , dene Aq to be the free associative (noncommutative) k-algebra on the symbols a; b; c; d modulo the relations ba D qab; db D qbd; bc D cb; ca D qac; q
1
dc D qcd; ad D q
1
da D ad C .q dened by
/bc;
bc D 1:
This becomes a Hopf algebra with a b c d
a b a b D , i.e., c d c d :
8 <
.a/ .b/ .c/ .d /
D D D D
aaCb c ab Cb d ; c aCd c c b Cd d
and with suitable maps and S . When q D 1, Aq becomes O.SL2 /, and so the Aq can be regarded as a one-dimensional family of quantum groups that specializes to SL2 when q ! 1. The algebra Aq is usually referred to as the Hopf algebra of SLq .2/: For bi-algebras that are neither commutative nor cocommutative, many statements in this section become more difcult to prove, or even false. For example, while it is still true that a bi-algebra admits at most one inversion, the composite of an inversion with itself need not be the identity map (Kassel 1995, III 3).
4m
Terminology
From now on, bialgebra will mean commutative bi-algebra and Hopf algebra will mean commutative bi-algebra that admits an inversion (antipode) (necessarily unique). Thus, the notion of a bialgebra is not self dual.9
the literature, there are different denitions for Hopf algebra. Bourbaki and his school (Dieudonn , Serre, . . . ) e use cog` bre and big` bre for co-algebra and bi-algebra. e e
9 In
38
4n
Exercises
To avoid possible problems, in the exercises assume k to be a eld. E XERCISE 4-1 For a set X, let R.X / be the k-algebra of maps X ! k. For a second set Y , let R.X/ R.Y / act on X Y by the rule (f g/.x; y/ D f .x/g.y/. (a) Show that the map R.X / R.Y / ! R.X Y / just dened is injective. (Hint: choose a P basis fi for R.X / as a k-vector space, and consider an element fi gi .) (b) Let be a group and dene maps W R. / ! R. W R. / ! k; SW R. / ! R. /; /; . f /.g; g 0 / D f .gg 0 / f D f .1/ .Sf /.g/ D f .g
1
/: , , and S dene
Show that if maps R. / into the subring R. / R. / of R. /, then on R. / the structure of a Hopf algebra. (c) If is nite, show that always maps R. / into R. / R. /.
E XERCISE 4-2 We continue the notations of the last exercise. Let be an arbitrary group. From a homomorphism W ! GLn .k/, we obtain a family of functions g 7! .g/i;j , 1 i; j n, on G. Let R0 . / be the k-subspace of R. / spanned by the functions arising in this way for varying n. (The elements of R0 . / are called the representative functions on .) (a) Show that R0 . / is a k-subalgebra of R. /. (b) Show that maps R0 . / into R0 . / R0 . /. (c) Deduce that , , and S dene on R0 . / the structure of a Hopf algebra. (Cf. Abe 1980, Chapter 2, 2; Cartier 2007, 3.1.1.) E XERCISE 4-3 Let G be the constant algebraic group over k dened by a nite commutative group . Let n be the exponent of , and assume that k contains n distinct nth roots of 1 (so, in particular, n is not divisible by the characteristic of k). Show that the Cartier dual of G is the constant algebraic group dened by the dual group Hom. ; Q=Z/.
_ E XERCISE 4-4 If k has characteristic p 0, show that p ' p and .Z=pZ/_ ' p (hence k .Z=pZ/k ) (here .Z=pZ/k , p , and p are the groups in (2.14), (2.15), and (2.16)). _ p
'
E XERCISE 4-5 Let A be a Hopf algebra. Prove the following statements by interpreting them as statements about afne groups. (a) S S D idA . (b) S D t S S where t .a b/ D b a: (c) S D : (d) The map a b 7! .a 1/ .b/W A A ! A A is a homomorphism of k-algebras. Hints: .a
1/ 1
D e; .ab/
Db
1a 1;
D e.
E XERCISE 4-6 Show that there is no algebraic group G over k such that G.R/ has two elements for every k-algebra R. E XERCISE 4-7 Verify directly that O.Ga / and O.Gm / satisfy the axioms to be a Hopf algebra.
4. Afne groups and Hopf algebras E XERCISE 4-8 Verify all the statements in 4.13. E XERCISE 4-9 A subspace V of a k-coalgebra C is a coideal if C .V / D 0.
C .V /
39
V C C C V and
(a) Show that the kernel of any homomorphism of coalgebras is a coideal. (b) Let V be a coideal in a k-coalgebra C . Show that the quotient vector space C =V has a unique k-coalgebra structure for which C ! C =V is a homomorphism. Show that any homomorphism of k-coalgebras C ! D whose kernel contains V factors uniquely through C ! C =V . Hint: show that if f W V ! V 0 and gW W ! W 0 are homomorphisms of k-vector spaces, then Ker.f g/ D Ker.f / W C V Ker.g/: E XERCISE 4-10 (cf. Sweedler 1969, 4.3.1). A k-subspace a of a k-bialgebra A is a bi-ideal if it is both an ideal and a co-ideal. When A admits an inversion S , a bi-ideal a is a Hopf ideal if S.a/ a. In other words, an ideal a A is a bi-ideal if .a/ a A C A a and
.a/ D 0; and it is a Hopf ideal if, in addition, S.a/ a: (a) Show that the kernel of any homomorphism of bialgebras (resp. Hopf algebras) is a bi-ideal (resp. Hopf ideal). (b) Let a be a bi-ideal in a k-bialgebra A. Show that the quotient vector space A=a has a unique k-bialgebra structure for which A ! A=a is a homomorphism. Show that any homomorphism of k-bialgebras A ! B whose kernel contains a factors uniquely through A ! A=a. Show that an inversion on A induces an inversion on A=a provided that a is a Hopf ideal. (c) Show that the image of a homomorphism W A ! B of bialgebras is a sub-bialgebra A of B, and that denes an isomorphism of A= Ker./ onto A. If A and B admit inversions, show that A does also, and that denes an isomorphism of Hopf algebras A= Ker./ ! A (hint: apply 4.8).
Algebraic groups and afne algebraic schemes
In the last section, we saw that algebraic groups over a eld k correspond to group objects in the opposite of the category of nitely presented algebras over k (see 4g). In this section we give a geometric interpretation of this opposite category as the category of afne algebraic schemes over k. Thus algebraic groups over k correspond to group objects in the category of afne algebraic schemes over k, or, in other words, to afne algebraic group schemes. This geometric interpretation provides additional insights. Throughout this section, k is a eld. We shall need to use some statements from commutative algebra (proved in CA).
5a
Afne k-algebras
An afne k-algebra is a nitely generated k-algebra A such that k al k A is reduced. If A is afne, then K k A is reduced for all elds K containing k; in particular, A itself is reduced (CA 18.3). When k is perfect, every reduced nitely generated k-algebra is an afne k-algebra (CA 18.1). The tensor product of two afne k-algebras is again an afne k-algebra (CA 18.4):
5b
The max spectrum of a ring
Let A be commutative ring, and let V be the set of maximal ideals in A. For an element f of A, let D.f / D fm 2 V j f mg; and for an ideal a in A, let V .a/ D fm 2 V j m ag: The sets D.f / are a base for a topology on V whose closed sets are exactly the sets V .a/. This is the Zariski topology, and the set V endowed with the Zariski topology is the max spectrum specm.A/ of A (see CA 12). A topological space V is noetherian if every ascending chain of open subsets U1 U2 eventually becomes constant. A nonempty topological space is irreducible if it is not the union of two proper closed subsets. Every noetherian topological space V can be expressed as the union of a nite collection I of irreducible closed subsets: [ V D fW j W 2 I g: Among such collections I there is exactly one such that no subset in I contains a second (CA 12.10). The elements of this I are called the irreducible components of V . When A is a nitely generated k-algebra, specm.A/ is noetherian, and the map a 7! V .a/ denes one-to-one correspondences radical ideals $ closed subsets prime ideals $ irreducible closed subsets maximal ideals $ one-point sets: T The ideal corresponding to a closed set W is I.W / D fm j m 2 W g. The nilradical N of A is the smallest radical ideal, and so it corresponds to the whole space specm.A/. Therefore specm.A/ is irreducible if and only if N is prime. 40
5. Algebraic groups and afne algebraic schemes
41
Let 'W A ! B be a homomorphism of nitely generated k-algebras. For any maximal ideal m in B, ' 1 .m/ is maximal in A, and therefore ' denes a map specm.'/W specm B ! specm A; m 7! '
1
.m/;
which is continuous because the inverse image of D.f / is D.'.f //. In this way, specm becomes a contravariant functor from the category of nitely generated k-algebras to topological spaces (CA p.48).
VARIANT:
THE PRIME SPECTRUM
The (prime) spectrum spec.A/ of a commutative ring A is the set V of prime ideals in A endowed with the topology for which the closed sets are those of the form V .a/ D fp 2 V j p ag; a an ideal in A:
Clearly V .p/ is the smallest closed subset containing the point p 2 V , and so specm.A/ is the subspace of spec.A/ consisting of the closed points. When A is a nitely generated k-algebra, the map U 7! U \ specm.A/ is a bijection from the set of open subsets of spec.A/ onto the set of open subsets of specm.A/; therefore specm.A/ and spec.A/ have the same topologies only the underlying sets differ. The advantage of the functor A spec.A/ is that it is dened on all commutative rings: the inverse image of a prime ideal is again prime, but the inverse image of a maximal ideal need not be maximal when the rings are not nitely generated algebras over a eld. The advantage of specm.A/ is that it is more geometric.
5c
Afne algebraic schemes
Let A be a commutative ring, and let V D spec A. Recall that the set of principal open subsets of V , B D fD.f / j f 2 Ag, is a base for the topology. Moreover, B is closed under nite intersections because D.f1 fr / D D.f1 / \ : : : \ D.fr /:
For a principal open subset D of V , dene OA .D/ D SD1 A where SD is the multiplicative subset S A p2D p of A. If D D D.f /, then SD is the smallest saturated multiplicative subset of A containing f , and so OA .D/ ' Af (see CA 6). If D D 0 , then SD SD 0 , and so there is a canonical restriction homomorphism OA .D/ ! OA .D 0 /. These restriction maps make D S D.U / into a functor on B satisfying the sheaf condition: for any covering D D i 2I Di of a D 2 B by Di 2 B, the sequence Y Y OA .D/ ! OA .Di / OA .Di \ Dj /
i 2I .i;j /2I I
is exact. For an open subset U of V , dene OA .U / by the exactness of Y Y OA .U / ! OA .D/ OA .D \ D 0 / 0

D2I .D;D /2I I
(40)
where I D fD 2 B j D U g. Clearly, U OA .U / is a functor on the open subsets of V , and it is not difcult to check that it is a sheaf. The set I in (40) can be replaced by any subset of B
42
covering U without changing OA .U /. In particular, if U D D.f /, then OA .U / ' OA .D.f // ' Af . Therefore, the stalk of OA at a point p 2 V is A ' Ap O .D.f // ' lim O .U / D lim Op D lim !f p f !f p A !U 3p A (for the last isomorphism, see CA 6). The ringed space Spec.A/ D .spec A; OA / is called the spectrum of A. An afne scheme .V; OV / is a ringed space isomorphic to Spec.A/ for some commutative ring A; note that A D OV .V /. Let k be a eld. If OV .V / is a nitely generated k-algebra (resp. an afne k-algebra), then we say that V is an afne algebraic scheme over k (resp. an afne algebraic variety over k), and we replace Spec.A/ with Specm.A/ (so the points of an afne algebraic scheme are closed). When k is algebraically closed, this denition of an afne algebraic variety over k agrees with that in AG, Chapter 3. A morphism of afne schemes is morphism of ringed spaces, and a morphism of afne algebraic schemes over k is a morphism of spaces ringed by k-algebras. In the second case, we often call it a regular map. P ROPOSITION 5.1 The functor .V; O/ O.V / is a contravariant equivalence from the category of afne schemes to the category commutative rings (resp. from afne algebraic schemes over k to nitely generated commutative k-algebras; resp. from afne algebraic varieties over k to afne k-algebras). P ROOF. The functor A Spec.A/ is a quasi-inverse.
2
def
E XAMPLE 5.2 Let k be an algebraically closed eld, and endow k n with the topology for which the closed sets are the zero-sets of families of polynomials. Let V be a closed subset of k n , let a be the set of polynomials that are zero on V , and let kV D kX1 ; : : : ; Xn =a D kx1 ; : : : ; xn : A pair of elements g; h 2 kV with h 0 denes a function P 7!
g.P / W D.h/ h.P /
!k
on the open subset D.h/ of V where h is nonzero. A function f W U ! k on an open subset U of V is said to be regular if it is of this form in a neighbourhood of each point of U . Let O.U / be the set of regular functions on U . Then U O.U / is a sheaf of k-algebras on V , and .V; O/ is an afne algebraic scheme over k with O.V / D kV . See AG 3.4 the map .a1 ; : : : ; an / 7! .x1 a1 ; : : : ; xn an /W V ! specm .kV / is a bijection because of the Nullstellensatz. When V D k n , the scheme .V; O/ is afne n-space An . E XAMPLE 5.3 Let k be an algebraically closed eld. The afne algebraic scheme Specm.kX; Y =.Y // can be identied with the scheme attached to the closed subset Y D 0 of k k in (5.2). Now consider Specm.kX; Y =.Y 2 //. This has the same underlying topological space as before (namely, the x-axis in k k) but it should now be thought of as having multiplicity 2, or being a line thickened in another dimension. We often write V for .V; O/, and we call O.V / the coordinate ring of V . The reader should think of an afne scheme as being a topological space V together with the structure provided by the ring O.V /.
43
S CHEMES IN GENERAL .
A scheme is a locally ringed space that admits an open covering by afne schemes.
5d
Properties of afne algebraic schemes
Let K be a eld containing k. An afne algebraic scheme V over k denes an afne algebraic scheme VK over K with O.VK / D K k O.V /. An afne algebraic scheme V over k is reduced if O.V / is reduced, and it is geometrically reduced of VK is reduced for all elds K containing k. Thus V is geometrically reduced if and only if O.V / is an afne k-algebra, and so a geometrically reduced afne algebraic scheme is an afne algebraic variety. Let N be the nilradical of O.V /. Then V is reduced N D 0I V is irreducible N is prime; V is reduced and irreducible O.V / is an integral domain. The rst statement follows from the denitions, the second statement has already been noted (p.40), and the third statement follows from the rst two. Recall (CA 3.11) that the height ht.p/ of a prime ideal p in a noetherian ring A is the greatest length d of a chain of distinct prime ideals p and that the Krull dimension of A is supfht.m/ j m 2 specm.A/g. The dimension of an afne algebraic scheme V is the Krull dimension of O.V / this is nite (CA 13.11). When V is irreducible, the nilradical N of O.V / is prime, and so O.V /=N is an integral domain. In this case, the dimension of V is the transcendence degree over k of the eld of fractions of O.V /=N, and every maximal chain of distinct prime ideals in O.V / has length dim V (CA 13.8). Therefore, every maximal chain of distinct irreducible closed subsets of V has length dim V . For example, the dimension of An is the transcendence degree of k.X1 ; : : : ; Xn / over k, which is n. p1 pd ,
5e
Algebraic groups as groups in the category of afne algebraic schemes
Let V and W be afne algebraic schemes over k, and let V W be an afne algebraic scheme such that O.V W / D O.V / k O.W /. Then (16), p.23, shows that Hom.T; V W / ' Hom.T; V / Hom.T; W /; all afne algebraic schemes T;
and so V W is the product of V and W in the category of afne algebraic schemes. Similarly, D Specm.k/ is a nal object. Therefore the category of afne algebraic schemes has nite products, and so monoid objects and group objects are dened. A monoid (resp. group) in the category of afne algebraic schemes over k will be called a monoid scheme (resp. group scheme) over k. As the tensor product of two afne k-algebras is again afne, the category of afne algebraic varieties also has products. A monoid (resp. group) in the category of afne algebraic varieties will
44
be called a monoid variety (resp. group variety). Equivalently, a monoid (resp. group) variety is a monoid (resp. group) scheme G such that O.G/ is an afne k-algebra. An afne algebraic scheme V denes a functor R V .R/ D Homk-alg .O.V /; R/;
def
(41)
from k-algebras to sets. For example, An .R/ ' Rn for all k-algebras R. Let V 0 be the functor dened by V . It follows from (5.1) and the Yoneda lemma that V V 0 is an equivalence from the category of algebraic schemes over k to the category of functors from k-algebras to sets representable by nitely generated k-algebras. Monoid (resp. group) structures on V correspond to factorizations of V 0 through the category of monoids (resp. groups). Thus V V 0 is an equivalence from the category of monoid (resp. group) schemes to the category of algebraic monoids (resp. groups), with quasi-inverse G Specm.O.G//. We saw in (5.1) that V O.V / is an equivalence from the category of algebraic schemes over k to the category of nitely generated k-algebras. Monoid (resp. group) structures on V correspond to bialgebra (resp. Hopf algebra) structures on O.V /. Thus V O.V / is a contravariant equivalence from the category of monoid (resp. group) schemes to the category of nitely generated bialgebras (resp. nitely generated Hopf algebras). S UMMARY 5.4 There are canonical equivalences between the following categories: (a) the category of algebraic monoids (resp. groups) over k; (b) the opposite of the category of nitely generated bialgebras (resp. nitely generated Hopf algebras) over k; (c) the category of monoids (resp. groups) in the category of algebraic schemes over k. For an algebraic group G, we let .jGj; O.G//, or just jGj, denote the corresponding group scheme (or group variety); thus jGj D Specm.O.G//. The dimension of G is dened to be the dimension of jGj.
I S THE SET jGj A GROUP ?

Not usually. The problem is that the functor specm does not send sums to products. For example, when k1 and k2 are nite eld extensions of k, the set specm.k1 k k2 / may have many points whereas specm.k1 / specm.k2 / has only one. For an algebraic group G, there is a canonical map jG Gj ! jGj jGj, but the map jG Gj ! jGj dened by m need not factor through it. However, jGj is a group when k is algebraically closed. Then the Nullstellensatz shows that jGj ' G.k/, and so jGj inherits a group structure from G.k/. To put it another way, for nitely generated algebras A1 and A2 over an algebraically closed eld k; specm.A1 k A2 / ' specm.A1 / specm.A2 / (42)
(as sets, not as topological spaces), and so the forgetful functor .V; O/ V sending an afne algebraic scheme over k to its underlying set preserves nite products,10 and hence also monoid objects and group objects.
regarded as a functor to topological spaces, .V; O/ V does not preserve nite products: the topology on V W is not the product topology. For an afne algebraic group G, the map mW jGj jGj ! jGj is not usually continuous relative to the product topology.
10 When
45
Assume k is perfect, and let D Gal.k al =k/. Then jGj ' nG.k al / and G.k/ ' G.k al / . In other words, jGj can be identied with the set of -orbits in G.k al / and G.k/ with the set of orbits consisting of a single point. While the latter inherits a group structure from G.k/, the former need not. The situation is worse with spec. For example, (42) fails for spec even when k is algebraically closed.
5f
Homogeneity
Let G be an algebraic group over a eld k. An element a of G.k/ denes an element of G.R/ for each k-algebra R, which we denote aR (or just a). Let e denote the identity element of G.k/. P ROPOSITION 5.5 For each a 2 G.k/, the natural map La W G.R/ ! G.R/; is an isomorphism of set-valued functors. Moreover, Le D idG and La Lb D Lab ; Here e is the neutral element in G.k/. P ROOF. The second statement is obvious, and the rst follows from it, because the equalities La La show that La is an isomorphism. The homomorphism O.G/ ! O.G/ dened by La is the composite of the homomorphisms O.G/ ! O.G/ k O.G/
aO.G/
1
g 7! aR g;
all a; b 2 G.k/:
D Le D idG
2
! k k O.G/ ' O.G/.
(43)
For a 2 G.k/, we let ma denote the kernel of aW O.G/ ! k; thus ma D ff 2 O.G/ j fk .a/ D 0g (see the notations 2.11). Then O.G/=ma ' k, and so ma is a maximal ideal in O.G/. Note that O.G/ma is the ring of fractions obtained from O.G/ by inverting the elements of the multiplicative set ff 2 O.G/ j fk .a/ 0g: P ROPOSITION 5.6 For each a 2 G.k/, O.G/ma ' O.G/me : P ROOF. The isomorphism à W O.G/ ! O.G/ corresponding (by the Yoneda lemma) to La is dened by à .f /R .g/ D fR .aR g/, all g 2 G.R/. Therefore, à 1 me D ma , and so à extends to an isomorphism O.G/ma ! O.G/me (because of the universal property of rings of fractions; CA 6.1). 2 C OROLLARY 5.7 When k is algebraically closed, the local rings O.G/m at maximal ideals m of O.G/ are all isomorphic.
46
P ROOF. When k is algebraically closed, the Nullstellensatz (CA 11.6) shows that all maximal ideals in O.G/ are of the form ma for some a 2 G.k/. 2 5.8 The corollary fails when k is not algebraically closed. For example, for the algebraic group 3 over Q, p kX kX kX O. 3 / D ' ' Q Q 3; .X 3 1/ .X 1/ .X 2 C X C 1/ p and so the local rings are Q and Q 3.
5g
Reduced algebraic groups
An algebraic group G is reduced if jGj is reduced, i.e., O.G/ has no nilpotents. P ROPOSITION 5.9 Let G be a reduced algebraic group over a eld k. If G.K/ D f1g for some algebraically closed eld K containing k, then G is the trivial algebraic group, i.e., O.G/ D k. P ROOF. Every maximal ideal of O.G/ arises as the kernel of a homomorphism O.G/ ! K (Nullstellensatz, CA 11.5), and so O.G/ has only one maximal ideal m. As O.G/ is reduced, the intersection of its maximal ideals is zero (CA 11.8), and so m D 0. Therefore O.G/ is a eld. It contains k, and the identity element in G is a homomorphism O.G/ ! k, and so O.G/ D k. 2
5.10 The proposition is false for nonreduced groups. For example, p .K/ D f1g for every eld K containing k, but p is not the trivial group. P ROPOSITION 5.11 Let G be an algebraic group over a perfect eld k, and let N be the nilradical of O.G/. There is a unique Hopf algebra structure on O.G/=N such that O.G/ ! O.G/=N is a homomorphism of Hopf algebras. Let Gred ! G be the corresponding homomorphism of algebraic groups. Any homomorphism H ! G with H a reduced algebraic group factors uniquely through Gred ! G. P ROOF. Let A D O.G/ and Ared D O.G/=N. Then Ared is a nitely generated reduced algebra over a perfect eld, and so it is an afne k-algebra. Hence Ared k Ared is also an afne k-algebra. In particular, it is reduced, and so the map A ! A k A ! Ared k Ared factors through Ared . Similarly, S and are dened on Ared , and it follows that there exists a unique structure of a Hopf algebra on Ared such that A ! Ared is a homomorphism of Hopf algebras. Any homomorphism from A to a reduced k-algebra factors uniquely through A ! Ared , from which the nal statement follows. 2 The algebraic group Gred is called the reduced algebraic group attached to G.
5.12 When k is not perfect, a Hopf algebra structure on A need not pass to the quotient A=N. For example, let k be a eld of characteristic 2, and let a be a nonsquare in k. Then R G.R/ D fx 2 R j x 4 D ax 2 g is an additive commutative algebraic group, but O.G/=N is not a Hopf algebra quotient of O.G/ (see Exercise 12-6 below).
47
5h
Smooth algebraic schemes
We review some denitions and results in commutative algebra. 5.13 Let m be a maximal ideal of a noetherian ring A, and let n D mAm be the maximal ideal of the local ring Am ; for all r n, the map a C mn 7! a C nn W mr =mn ! nr =nn is an isomorphism (CA 6.7). 5.14 Let A be a local noetherian ring with maximal ideal m and residue eld k. Then m=m2 is a k-vector space of dimension equal to the minimum number of generators of m (Nakayamas lemma, CA 3.9). Moreover, ht.m/ dimk .m=m2 / (CA 16.5), and when equality holds A is said to be regular. Every regular noetherian local ring is an integral domain (CA 17.3). 5.15 A point m of an afne algebraic scheme V is said to be regular if the local ring O.V /m is regular, and V is said to be regular if all of its points are regular. A regular afne algebraic scheme is reduced. To see this, let f be a nilpotent element of O.V /; as f maps to zero in O.V /m , sf D 0 for some s 2 O.V / m; therefore the annihilator of f is an ideal O.V / not contained in any maximal ideal, and so it equals O.V /. 5.16 An afne algebraic scheme V over k is said to be smooth if Vk al is regular. If V is smooth, then VK is regular for all elds K containing k; in particular, V itself is regular (CA 18.14). If V is smooth, then it follows from (5.15) that O.V / is an afne k-algebra, and so V is an algebraic variety. Every afne algebraic variety contains a regular point (CA 18.15).
5i
Smooth algebraic groups
An algebraic group G is said to be smooth if jGj is smooth. P ROPOSITION 5.17 Let H be an algebraic subgroup of an algebraic group G. Then dim H dim G, and dim H < dim G if G is smooth and connected and H G. P ROOF. Because O.H / is a quotient of O.G/, dim.O.H // dim.O.G//. If G is smooth and connected, then O.G/ is an integral domain; if H G, then dim H < dim G by (CA 13.3). 2 P ROPOSITION 5.18 An algebraic group G over an algebraically closed eld k is smooth if and only if O.G/me is regular, where me D Ker. W O.G/ ! k/. P ROOF. If O.G/m is regular for m D me , then O.G/m is regular for all m by homogeneity (5.6). Hence G is smooth. 2 P ROPOSITION 5.19 An algebraic group G is smooth if and only if jGj is geometrically reduced (i.e., an algebraic variety). In particular, an algebraic group G over an algebraically closed eld is smooth if and only if jGj is reduced. P ROOF. If G is smooth, then jGj is an algebraic variety by (5.16). For the converse, we have to show that Gk al is regular. According to (5.16), Gk al has a regular point, and so, by homogeneity (5.6), all of its points are regular. 2
48
C OROLLARY 5.20 An algebraic group G over an algebraically closed eld k is smooth if every nilpotent element of O.G/ is contained in m2 . e N P ROOF. Let G be the reduced algebraic group attached to G (see 5.11), and let e be the neutral N N N D O.G/=N where N is the nilradical of O.G/. Every prime ideal element of G.k/. Then O.G/ N of O.G/ contains N, and so the prime ideals of O.G/ and O.G/ are in natural one-to-one correN N spondence. Therefore me and me have the same height, and so O.G/me and O.G/me have the same N Krull dimension. The hypothesis implies that me =m2 ! me =m2 N e e N N N N N is an isomorphism of k-vector spaces. Because jGj is a reduced, G is smooth; in particular, O.G/me is regular, and so N N dimk .me =m2 / D Krull dim.O.G/me /. N e N Therefore dimk .me =m2 / DKrull dim.O.G/me /, and so O.G/me is regular. By homogeneity (5.6), e O.G/m is regular for all maximal ideals m in O.G/. 2
5.21 A reduced algebraic group over a nonperfect eld need not be smooth. For example, let k be such a eld, so that char.k/ D p 0 and there exists an element a of k that is not a pth power. Then the subgroup G of Ga Ga dened by Y p D aX p is reduced but not smooth. Indeed, O.G/ D kX; Y =.Y p which is an integral domain because Y p aX p /;
aX p is irreducible in kX; Y , but aX p / D k al x; y
O.Gk al / D k al X; Y =.Y p
1
contains the nilpotent element y a p x. The reduced subgroup .Gk al /red of Gk al is the subgroup of 1 Ga Ga is dened by Y D a p X , which is not dened over k (as a subgroup of Ga Ga ). Note that G is the kernel of .x; y/ 7! y p ax p W Ga Ga ! Ga . Therefore, although Ker./ is (of course) dened over k, Ker./red is not.
5j
Algebraic groups in characteristic zero are smooth (Cartiers theorem)
We rst prove two lemmas. L EMMA 5.22 Let V and V 0 be vector spaces, and let W be a subspace of V . For x 2 V , y 2 V 0 , x y 2 W V 0 x 2 W or y D 0: P ROOF. Choose a decomposition V D W W 0 . Then x has a unique decomposition x D xW CxW 0 with xW 2 W and xW 0 2 W 0 , and x lies in W if and only if xW 0 D 0. As V V 0 ' .W V 0 / .W 0 V 0 /; x y D xW y C xW 0 y;
we see that x y 2 W V 0 if and only if xW 0 y D 0, which holds if and only if xW 0 D 0 or y D 0. 2 L EMMA 5.23 Let .A; ; / be a Hopf algebra over k, and let I D Ker. /.
5. Algebraic groups and afne algebraic schemes (a) As a k-vector space, A D k I . (b) For any a 2 I , .a/ D a 1 C 1 a
49
mod I I .
P ROOF. (a) The maps k ! A ! k are k-linear, and compose to the identity. Therefore A D k I and a 2 A decomposes as a D .a/ C .a .a// 2 k I . (b) For a 2 A, write a D a0 C a00 with a0 D .a/ 2 k and a00 2 I: Let .a/ D P b c; b; c 2 A:
From the commutativity of the second diagram in (26), p.28, we nd that 1a D a1 D Therefore .a/ a1 1a D P .b c b 0 c b c 0 / P 00 D .b c 00 b 0 c 0 / P 0 b c 0 mod I I . P P b0 c b c0 in k k A in A k k.
Now (cf. 2.9) .. ; / / .a/ D . ; /. .. ; / / .a/ D . and so P b 0 c 0 D 0 if a 2 I . P b c/ D P b0 c0
/.a/ D .a/;
2
T HEOREM 5.24 (C ARTIER 1962) Every algebraic group over a eld of characteristic zero is smooth. P ROOF. We may replace k with its algebraic closure. Thus, let G be an algebraic group over an algebraically closed eld k of characteristic zero, and let A D O.G/. Let m D me D Ker. /. Let a be a nilpotent element of A; according to (5.20), it sufces to show that a lies in m2 . If a maps to zero in Am , then it maps to zero in Am =.mAm /2 , and therefore in A=m2 by (5.13), and so a 2 m2 . Thus, we may suppose that there exists an n 2 such that an D 0 in Am but an 1 0 in Am . Now san D 0 in A for some s m. On replacing a with sa, we nd that an D 0 in A but an 1 0 in Am . Now a 2 m (because A=m D k has no nilpotents), and so (see 5.23) .a/ D a 1 C 1 a C y Because is a homomorphism of k-algebras, 0D .an / D . a/n D .a 1 C 1 a C y/n . (44) with y 2 m k m.
When expanded, the right hand side becomes a sum of terms an 1; n.an
1
1/ .1 a C y/;
.a 1/h .1 a/i y j
.h C i C j D n, i C j
2/:
50 As an D 0 and the terms with i C j nan

1
BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1) 2 lie in A k m2 , equation (44) shows that
1
a 2 an
m k A C A k m2
(inside A k A).
In the quotient A k A=m2 this becomes nan

1
a 2 an N
m k A=m2
(inside A k A=m2 ).
(45)
As k has characteristic zero, n is a nonzero element of k, and hence it is a unit in A. On the other hand, an 1 an 1 m, because if an 1 D an 1 m with m 2 m, then .1 m/an 1 D 0; as 1 m is a unit in Am , this would imply an 1 D 0 in Am , which is a contradiction. Hence nan 1 an 1 m, and so (see 5.22), a 2 m2 , as required. 2 C OROLLARY 5.25 Let G be an algebraic group over a eld of characteristic zero. If G.K/ D f1g for some algebraically closed eld K, then G is the trivial algebraic group. P ROOF. According to the theorem, G is reduced, and so we can apply Proposition 5.9.
2
T HEOREM 5.26 An algebraic group G over an algebraically closed eld k of characteristic p 0 is smooth if O.G/ has the following property: a 2 O.G/; ap D 0 H) a D 0: (46)
P ROOF. Let a be a nilpotent element of O.G/. As in the proof of Theorem 5.24, we may suppose n n that an D 0 in O.G/ but an 1 0 in O.G/m , m D me . If pjn, then .a p /p D 0, and so a p D 0, which is a contradiction. Therefore n is nonzero in k, and the argument in the proof of Theorem 5.24 shows that a 2 m2 . 2 R EMARK 5.27 Let G be an afne algebraic group over an algebraically closed eld k of characteristic p 0. For each a in the nilradical N of O.G/, there exists an exponent, which we may take to r be a power p r of p, such that ap D 0. As N is nitely generated, there exists a single r that works for all a 2 N. With this r, let O.G/p D ff p j f 2 O.G/g
r r
O.G/:
r
The map f 7! f p W O.G/ ! O.G/ is a homomorphism of Hopf algebras, and so O.G/p is a Hopf subalgebra of O.G/. As it satises (46), it is reduced.
5k
Transporters
Recall that an action of a monoid G on a set X is a map .g; x/ 7! gxW G such that (a) .g1 g2 /x D g1 .g2 x/ for all g1 ; g2 2 G, x 2 X, and (b) ex D x for all x 2 X (here e is the identity element of G). X !X
51
Now let G be an afne monoid over k, and let X be a functor from the category of k-algebras to sets. An action of G on X is a natural transformation G X ! X such that G.R/ X.R/ ! X.R/ is an action of the monoid G.R/ on the set X.R/ for all k-algebras R. Let Z Y be subfunctors of X . The transporter TG .Y; Z/ of Y into Z is the functor R fg 2 G.R/ j gY Zg;
where the condition gY Z means that gY .R0 / Z.R0 / for all R-algebras R0 , i.e., that gY Z as functors on the category of R-algebras. We say that a map Z ! Y in Alg_ is a closed immersion if, for every map hA ! Y , the bred k product hA Y Z is represented by a quotient of A. For example, if Y D hB , then Z ! Y is a closed immersion if and only if Z itself is represented by a quotient of A. To say that Z ! Y is a closed immersion means that, for any k-algebra B and 2 Y .B/, there exists an ideal b B such that, for a homomorphism 'W B ! R of k-algebras, R 2 Z.R/ '.b/ D 0 where R D Y .'/./ 2 Y .R/:
P ROPOSITION 5.28 Let G X ! X be an action of G on X , and let Z Y be subfunctors of X such that Z ! Y is a closed immersion. If X is representable, then TG .Y; Z/ is represented by a quotient of O.G/. This follows from the next three lemmas. For a k-algebra A and functor XW AlgA ! Set, we let A=k X denote the functor R X.A R/ (cf. 3d). L EMMA 5.29 If Z ! Y is a closed immersion, then, for any k-algebra A, so also is A=k Z ! A=k Y . P ROOF. To prove that A=k Z ! A=k Y is a closed immersion we have to show that, for any kalgebra B and 2 Y .A B/, there exists an ideal b0 B such that, for a homomorphism 'W B ! R of k-algebras, R 2 Z.A B/ '.b/ D 0: Because Z ! Y is a closed immersion, there exists an ideal c 2 A B such that R 2 Z.A B/ if T and only if .A '/.c/ D 0. Let b0 D Ab c b. Then .A '/.c/ D 0 '.b0 / D 0, and so b0 has the required property.
2
L EMMA 5.30 If Z ! Y is a closed immersion, then, for any map T ! Y of functors, so also is T Y Z !T. P ROOF. Let hA ! T be a map of functors. Then hA is obvious.
T
Y ' hA
Y , and so the statement

2
L EMMA 5.31 In the situation of the proposition, let B D O.X /; then TG .Y; Z/ ' G
B=k Y
B=k Z:
2
P ROOF. This follows directly from the denitions.
52
5l
Relaxing the conditions on k and G
In this subsection, the reader is assumed to be familiar with the basic theory of schemes (Hartshorne 1977, Chapter II). Let k be a commutative ring. The functor Spec from k-algebras to afne k-schemes takes nite direct sums (i.e., tensor products) to nite products, and so sends bialgebras and Hopf algebras over k to monoid objects and group objects in the category of afne k-schemes. Thus there is an equivalence of categories between afne groups over k and afne group schemes over k (group objects in the category of afne k-schemes). Algebraic groups over k correspond to nitely presented afne group schemes over k (which are called algebraic afne group schemes over k). A functor F W Algk ! Set is said to be formally smooth if, for any k-algebra A and nilpotent ideal n in A, the map F .A/ ! F .A=n/ is surjective. A k-scheme X is smooth over k if it is locally def nitely presented and the functor A X.A/ D Homk .Spec A; X / is formally smooth. There is the following criterion (SGA1, II): a nitely presented morphism is smooth if it is at and its geometric bres are nonsingular algebraic varieties. Therefore, when k contains a eld of characteristic zero, Cartiers theorem (5.24) shows that every at algebraic afne group scheme over k is smooth. The lemmas 5.29, 5.30, and 5.31 remains true with k a commutative ring provided A is free as a k-module. Therefore Proposition 5.28 is true provided O.G/ is free as a k-module. (DG I 2, 7).
5m
Appendix: The faithful atness of Hopf algebras
In this subsection, we prove a very important technical result (Theorem 5.32). T HEOREM 5.32 For any Hopf algebras A B over a eld k, B is faithfully at over A.
In the proof, we shall make use of the following criteria. (a) A homomorphism 'W A ! B is at if A' 1 .n/ ! Bn is at for all maximal ideals n in B.11 (b) A at homomorphism 'W A ! B is faithfully at if every maximal ideal of A is of the form ' 1 .n/ for some maximal ideal n of B (CA 9.9). For any eld k 0 k, the homomorphism A ! k 0 A is faithfully at, and so it sufces to show that k 0 B is faithfully at over k 0 A (CA 9.4). Therefore we may suppose that k is algebraically closed. Let WH ! G be a homomorphism of afne groups such that O.H ! G/ D B A.
N 0 ! N be an injective homomorphism of A-modules, and let n be a maximal ideal of B. Then p D ' a prime ideal in A, and Ap A .N 0 ! N / is injective (CA 9.8). Therefore,
11 Let 1 .n/
is
Bn A .N 0 ! N / ' Bn Ap .Ap A .N 0 ! N // is injective, and so the kernel M of B A .N 0 ! N / has the property that Mn D 0. Let x 2 M , and let a D fb 2 B j bx D 0g. For each maximal ideal n of B, x maps to zero in Mn , and so a contains an element not in n. Hence a D B, and so x D 0.
53
C ASE THAT A IS REDUCED AND A AND B
ARE FINITELY GENERATED .
In this case, jGj is a smooth algebraic variety. According to the above criterion, to show that 'W O.G/ ! O.H / is at, it sufces to show that, for each b 2 H.k/, the map O.G/ma ! O.H /mb ; a D '.b/, is at. Homogeneity, more precisely, the existence of commutative diagrams H ? ? y G
Lb
! H ? ? y
O.H /me x ? ? O.G/me
'
O.H /mb x ? ? O.G/ma
La
! G
'
(see 5f), implies that it sufces to prove this for a single b. When we replace G with its identity connected component, the ring O.G/ becomes an integral domain, and we can apply the generic atness (or freeness) theorem (CA 9.11; CA 16.9) to nd an open set of bs for which O.H /mb is faithfully at over O.G/ma . To show that O.G/ ! O.H / is faithfully at, it sufces to show that the map (of sets) jH j ! jGj is surjective. Because O.G/ ! O.H / is injective, the image of jH j ! jGj is dense in jGj; in particular, it meets every connected component of jGj. Because G is smooth, its identity component G is irreducible. According to (CA 12.13), the image of jH j ! jGj contains a nonempty open subset U of jG j. For any g 2 jG j, the sets U 1 and Ug 1 have nonempty intersection. This means that there exist u; v 2 U such that u 1 D vg 1 , and so g D vu 2 U . Thus U D jG j, and the translates of U by points in the image cover jGj.
E LEMENTARY DIGRESSION
For any homomorphism W H ! G of abstract groups, the map .n; h/ 7! .nh; h/W Ker./ H ! H
G
(47)
is an bijection this just says that two elements of H with the same image in G differ by an element of the kernel. Similarly, for any homomorphism W H ! G of afne groups, there is an isomorphism Ker./ H ! H
G
(48)
which becomes the map (47) for each k-algebra R. Because of the correspondence between afne groups and Hopf algebras, this implies that, for any homomorphism A ! B of Hopf algebras, the map b b 7! b bW B A B ! .B=IA B/ k B is an isomorphism (IA is the augmentation ideal of A). More generally, for any homomorphisms of abstract groups H ? ? y G the map .n; h/ 7! .n .h/; h/W Ker./ H ! G is a bijection. This implies a similar statement for afne groups: Ker.G ! G 0 / H ' G
G0 G0
(49)
! G 0; H
H:
(50)
54
C ASE THAT THE AUGMENTATION IDEAL OF A IS NILPOTENT

Assume that the augmentation ideal I D Ker. W A ! k/ of A is nilpotent, say I n D 0. Choose a family .ej /j 2J of elements in B whose image in B=IB is a k-basis and consider the map P .aj /j 2J 7! j aj ej W A.J / ! B (51) where A.J / is a direct sum of copies of A indexed by J . We shall show that (51) is an isomorphism (hence B is even free as an A-module). Let C be the cokernel of (51). A diagram chase in A.J / ? ? y .A=I /.J / ! B ? ? y ! C ! 0
onto
! B=IB
shows that every element of C is the image of an element of B mapping to zero in B=IB, i.e., lying in IB. Hence C D IC , and so C D IC D I 2 C D D I n C D 0. Hence A.J / ! B is surjective. For the injectivity, consider diagrams A.J / ? ? y B .J /
onto
B ? ? y
k .J / ? ? y .B=IB/.J /
'
B=IB ? ? y
onto
! B A B
'
! .B=IB/ k .B=IB/
in which the bottom arrows are obtained from the top arrows by tensoring on the left with B and B=IB respectively. If b 2 B .J / maps to zero in B A B, then it maps to zero in B=IB k B=IB, which implies that it maps to zero in .B=IB/.J / . Therefore the kernel M of B .J / ! B A B is contained in .IB/.J / D I B .J / . Recall (49) that B A B ' B k B=IB. As B=IB is free as a k-module (k is a eld), B k B=IB is free as a left B-module, and so B A B is free (hence projective) as a left B-module. Therefore there exists a B-submodule N of B .J / mapping isomorphically onto B A B, and B .J / D M N (direct sum of B-submodules). We know that M I B .J / D IM I N; and so M IM . Hence M D IM D I 2 M D D 0. We have shown that B .J / ! B A B is injective, and this implies that A.J / ! B is injective because A.J / B .J / .
C ASE THAT A IS FINITELY GENERATED

After Theorem 5.24, we may suppose that k has characteristic p 0. According to (5.27), there n exists an n such that O.G/p is a reduced Hopf subalgebra of O.G/. Let G 0 be the algebraic group n such that O.G 0 / D O.G/p , and consider the diagrams 1 ! N ? ? y ! M ! H ? ? y ! G ! G0 O.N / x ? ? O.M / O.H / at x ?injective ? O.G/
faithfully
O.G 0 /
! G0
O.G 0 /
55
where N and M are the kernels of the homomorphisms H ! G 0 and G ! G 0 respectively. Because O.G 0 / is reduced, the homomorphism O.G 0 / ! O.H / is faithfully at, and so O.G/ ! O.H / injective H) .O.G/ ! O.H // O.G 0 / O.H / injective.
As k is a direct summand of O.H /, this implies that .O.G/ ! O.H // O.G 0 / k is injective. From the diagram O.N / x ? ? O.M /
(20)
'
O.H / O.G 0 / k x ? ? O.G/ O.G 0 / k
(20)
'
we see that O.M / ! O.N / is injective, and hence is faithfully at (because the augmentation ideal of O.M / is nilpotent). From the diagrams N ?H ? y M H
(48)
'
H ?0H G ? y G
G0
(50)
'
O.N / O.H / ' O.H / O.G 0 / O.H / x x ? ? ? ? O.M / O.H / ' O.G/ O.G 0 / O.H /:
we see that .O.G/ ! O.H // O.G 0 / O.H / is faithfully at. As O.G 0 / ! O.H / is faithfully at, this implies that O.G/ ! O.H / is faithfully at (CA 9.4).
G ENERAL CASE
We show in (7.23) below that A and B are directed unions of nitely generated Hopf subalgebras Ai and Bi such that Ai Bi , and so the general case follows from the nitely generated case. C OROLLARY 5.33 Let A B be Hopf algebras with B an integral domain, and let K elds of fractions. Then B \ K D A; in particular, A D B if K D L. L be their
P ROOF. Because B is faithfully at over A, cB \ A D cA for any c 2 A. Therefore, if a=c 2 B, a; c 2 A, then a 2 cB \ A D cA, and so a=c 2 A. 2
5n
Terminology
From now on group scheme and algebraic group scheme will mean afne group scheme and afne algebraic group scheme.
A SIDE 5.34 Some statements have easy geometric proofs for smooth algebraic groups. In extending the proof to all algebraic groups, one often has to make a choice between a nonelementary (sometimes difcult) proof using algebraic geometry, and an elementary but uninformative proof using Hopf algebras. In general, we sketch the easy geometric proof for smooth algebraic groups, and give the elementary Hopf algebra proof in detail N OTES In most of the literature, for example, Borel 1991, Humphreys 1975, and Springer 1998, algebraic group means smooth algebraic group in our sense. Our denition of algebraic group is equivalent to afne group scheme algebraic over a eld. The approach through functors can be found in Demazure and Gabriel 1970 and Waterhouse 1979. The important Theorem 5.24 was announced in a footnote to Cartier 1962; the proof given here follows Oort 1966. Theorem 5.32 is proved entirely in the context of Hopf algebras in Takeuchi 1972; the proof given here follows Waterhouse 1979, Chapter 14.
Group theory: subgroups and quotient groups.
In this section and in Section 8, we show how the basic denitions and theorems in the theory of abstract groups can be extended to afne groups. Throughout, k is a eld.
6a
A criterion to be an isomorphism
P ROPOSITION 6.1 A homomorphism of afne groups W H ! G is an isomorphism if and only if (a) the map .R/W H.R/ ! G.R/ is injective for all k-algebras R, and (b) the homomorphism W O.G/ ! O.H / is injective. P ROOF. The conditions are obviously necessary. For the sufciency, note that the maps H
G
H !G
give rise to homomorphisms of Hopf algebras O.G/ ! O.H / In particular, the homomorphisms x 7! x 1 x 7! 1 x W O.H / ! O.H / O.G/ O.H / (52) O.H / O.G/ .H /:
agree on O.G/, and so dene elements of H.O.H / O.G/ O.H // mapping to the same element in G.O.H / O.G/ O.H //. Now, condition (a) with R D O.H / O.G/ O.H / implies that the two homomorphisms (52) are equal, and condition (b) implies that O.H / is a faithfully at O.G/-algebra (see 5.32), and so the subset of O.H / on which the two homomorphisms (52) agree is .O.G// by (CA 9.6). On combining these statements, we nd that that is surjective, and so it is an isomorphism.
2
6b
Subgroups; injective homomorphisms
D EFINITION 6.2 (a) A homomorphism H ! G of afne groups is injective if the map H.R/ ! G.R/ is injective for all k-algebras R. An injective homomorphism is also called an embedding. (b) A subgroup (resp. normal subgroup) of an afne group G is a subfunctor12 H of G such that H.R/ is a subgroup (resp. normal subgroup) of G.R/ for all R. P ROPOSITION 6.3 A homomorphism W H ! G of afne groups is injective if and only if the map W O.G/ ! O.H / is surjective. In other words, W H ! G is injective if and only if jjW jH j ! jGj is a closed immersion.
12 In
particular, this means that H.R/ is a subset of G.R/ for all k-algebras R.
56
6. Group theory: subgroups and quotient groups. P ROOF. ): The homomorphism factors into homomorphisms of Hopf algebras O.G/ .O.G// ,! O.H /
57
(see Exercise 4-10). Let H 0 be the afne group whose Hopf algebra is .O.G//. Then factors into H ! H 0 ! G; and the injectivity of implies that H.R/ ! H 0 .R/ is injective for all k-algebras R. Because O.H 0 / ! O.H / is injective, Proposition 6.1 shows that the map H ! H 0 is an isomorphism, and so .O.G// D O.H /. (: If is surjective, then any two homomorphisms O.H / ! R that become equal when composed with must already be equal, and so H.R/ ! G.R/ is injective. 2 C OROLLARY 6.4 A subgroup of an algebraic group is an algebraic group. P ROOF. If H ! G is an embedding, then O.G/ ! O.H / is surjective, and so, if O.G/ is nitely generated as a k-algebra, so also is O.H /. 2 P ROPOSITION 6.5 Let W H ! G be a homomorphism of afne groups. If is injective, then so also is k 0 W Hk 0 ! Gk 0 for any eld k 0 containing k. Conversely, if k 0 is injective for one eld k 0 containing k, then is injective. P ROOF. For any eld k 0 containing k, the map O.G/ ! O.H / is surjective if and only if the map k 0 k O.G/ ! k 0 k O.H / is surjective (this is simply a statement about vector spaces over elds). 2 6.6 When k is a perfect eld, Gred is a subgroup of G (see 5.11). However, it need not be normal. For example, over a eld k of characteristic 3, let G D 3 .Z=2Z/k for the nontrivial action of .Z=2Z/k on 3 ; then Gred D .Z=2Z/k , which is not normal in G (see SGA3 VIA 0.2). P ROPOSITION 6.7 The subgroups of an afne group G are in natural one-to-one correspondence with the Hopf ideals on O.G/. P ROOF. For a subgroup H of G, I.H / D ff 2 O.G/ j fR .h/ D 1 for all h 2 H.R/ and all Rg is a Hopf ideal in G (it is the kernel of O.G/ ! O.H /; see Exercise 4-10). Conversely, if a is a Hopf ideal in G, then the functor R fg 2 G.R/ j fR .g/ D 0 for all f 2 ag
is an afne subgroup G.a/ of O.G/ (it is represented by O.G/=a). The maps H 7! I.H / and a 7! G.a/ are inverse. 2 C OROLLARY 6.8 Every set of algebraic subgroups of an algebraic group G has a minimal element (therefore every descending chain of algebraic subgroups becomes stationary).
58

2
P ROOF. The ring O.G/ is noetherian (Hilbert basis theorem, CA 3.6).
P ROPOSITION 6.9 For any algebraic subgroup H of an algebraic group G, the algebraic scheme jH j is closed in jGj. P ROOF. If a is the kernel of O.G/ ! O.H /, then jH j is the subspace V .a/ D fm j m
def
ag of jGj :2
P ROPOSITION 6.10 For any family .Hj /j 2J of subgroups of an afne group G, the functor R is a subgroup ideals I.Hj /. T
j 2J
\
j 2J
Hj .R/ (intersection inside G.R/)
Hj of G, with coordinate ring O.G/=I where I is the ideal generated by the
P ROOF. We have Hj .R/ D fg 2 G.R/ j fR .g/ D 0 for all f 2 I.Hj /g: Therefore, H.R/ D fg 2 G.R/ j fR .g/ D 0 for all f 2 D Hom.O.G/=I; R/: [ I.Hj /g
2
E XAMPLE 6.11 The intersection of the algebraic subgroups SLn and Gm (scalar matrices) of GLn is n (matrices diag.c; : : : ; c/ with c n D 1). We sometimes loosely refer to an injective homomorphism W H ! G as a subgroup of G.
6.12 In the realm of (not necessarily afne) group schemes over a eld, there can exist nonafne (necessarily nonclosed) subgroup schemes of an algebraic group. For example, the constant subgroup scheme .Z/k of Ga over Q is neither closed nor afne.
6c
Kernels of homomorphisms
The kernel of a homomorphism W H ! G of afne groups is the functor R N.R/ D Ker..R/W H.R/ ! G.R//.
def
Let W O.G/ ! k be the identity element of G.k/. Then an element hW O.H / ! R of H.R/ lies in N.R/ if and only if its composite with W O.G/ ! O.H / factors through : O.H /
h
O.G/
k:
6. Group theory: subgroups and quotient groups.
59
Let IG be the kernel of W O.G/ ! k (this is called the augmentation ideal), and let IG O.H / denote the ideal generated by its image in O.H /. Then the elements of N.R/ correspond to the homomorphisms O.H / ! R that are zero on IG O.H /, i.e., N.R/ D Homk-alg .O.H /=IG O.H /; R/: We have proved: P ROPOSITION 6.13 For any homomorphism H ! G of afne groups, there is an afne subgroup N of H (called the kernel of the homomorphism) such that N.R/ D Ker.H.R/ ! G.R// for all R; its coordinate ring is O.H /=IG O.H /. Alternatively, note that the kernel of is the bred product of H ! G algebraic group with coordinate ring O.H / O.G/ .O.G/=IG / ' O.H /=IG O.H / (see 3b). E XAMPLE 6.14 Consider the map g 7! g n W Gm ! Gm . This corresponds to the map on Hopf algebras Y 7! X n W kY; Y 1 ! kX; X 1 because X n .g/ D g n D Y .g n / (cf. (8), p.14). The map W kY; Y 1 ! k sends f .Y / to f .1/, and so the augmentation ideal for Gm is .Y 1/. Thus, the kernel has coordinate ring kX; X
1
f1G g, and so it is an
=.X n
1/ ' kX =.X n
n,
1/:
In other words, the kernel is the algebraic group
as we would expect.
E XAMPLE 6.15 Let N be the kernel of the determinant map detW GLn ! Gm . This corresponds to the map on Hopf algebras X 7! det.Xij /W kX; X because det.Xij /.aij / D det.aij / D X.det.aij //: As we just noted, the augmentation ideal for Gm is .X O.N / D k: : : ; Xij ; : : : ; det.Xij / .det.Xij / 1/ 1/, and so
1 1
! k: : : ; Xij ; : : : ; det.Xij /
'
k: : : ; Xij ; : : : : .det.Xij / 1/
In other words, the kernel of det is the algebraic group SLn , as we would expect. P ROPOSITION 6.16 When k has characteristic zero, a homomorphism G ! H is injective if and only if G.k al / ! H.k al / is injective.
60
P ROOF. If G.k al / ! H.k al / is injective, the kernel N of the homomorphism has the property that N.k al / D 0, and so it is the trivial algebraic group (by 5.25). 2 6.17 Proposition 6.16 is false for elds k of characteristic p 0. For example, the homomorphism x 7! x p W Ga ! Ga has kernel p , and so is not injective, but the map x 7! x p W Ga .R/ ! Ga .R/ is injective for every reduced k-algebra R. R EMARK 6.18 Let A be an object of some category A. A morphism uW S ! A is a monomorphism if f 7! u f W Hom.T; S / ! Hom.T; A/ is injective for all objects T . Two monomorphisms uW S ! A and u0 W S 0 ! A are said to be equivalent if each factors through the other. This is an equivalence relation on the monomorphisms with target A, and an equivalence class of monomorphisms is called a subobject of A. A homomorphism of afne groups is a injective if and only if it is a monomorphism in the category of afne groups. To see this, let W H ! G be a homomorphism of afne groups. If is injective and the homomorphisms ; W H 0 ! H agree when composed with , then (6.3a) with R D O.H 0 / shows that D . Suppose, on the other hand, that is not injective, so that its kernel N is nontrivial. Then the homomorphisms n 7! 1, n 7! nW N ! N are distinct, but they agree when composed with , and so is not a monomorphism. Let G be an afne group. Two monomorphisms uW H ! G and uW H 0 ! G are equivalent if and only if Im.uR / D Im.u0 / for all k-algebras R. It follows that, in each equivalence class of R monomorphisms with target G, there is exactly one with H an afne subgroup of G and with u the inclusion map.
A SIDE 6.19 In any category, the equalizer of a pair of morphisms is a monomorphism. A monomorphism that arises in this way is said to be regular. In Grp, every monomorphism is regular (see, for example, van Oosten, Basic Category Theory, Exercise 42, p.21). For example, the centralizer of an element a of a group A (which is not a normal subgroup in general) is the equalizer of the homomorphisms x 7! x, x 7! axa 1 W A ! A. Is it true that every monomorphism in the category of afne (or algebraic) groups is regular?
6d
Dense subgroups
Let G be an algebraic group over a eld k. By denition, a point a 2 G.k/ is a homomorphism O.G/ ! k, whose kernel we denote ma (a maximal ideal in O.G/). As we discussed 5e, the map a 7! ma W G.k/ ! jGj is injective with image the set of maximal ideals m of O.G/ such that O.G/=m D k. We endow G.k/ with the subspace topology. P ROPOSITION 6.20 Let G be an algebraic group over a eld k, and let be a subgroup of G.k/. There exists an algebraic subgroup H of G such that H.k/ D if and only if is closed, in which case there exists a unique smallest H with this property. When k is algebraically closed, every smooth algebraic subgroup of G arises in this way. P ROOF. If D H.k/ for an algebraic subgroup H of G, then D jH j \ G.k/, which is closed by (6.9). With the notations of Exercise 4-1, there is a commutative diagram O.G/ ? ? y R. /
G
! O.G/ O.G/ ? ? y R. /;
61
which shows that maps into R. / R. /, and so .R. /; / is a Hopf algebra (ibid.). If def is closed, then it is the zero set of a D Ker.O.G/ ! R. //, and the subgroup H of G with O.H / D O.G/=a ' R. / has H.k/ D ; clearly, it is the smallest subgroup of G with this property. When k is algebraically closed and H is a smooth subgroup of G, then the group attached to D H.k/ is H itself. 2 R EMARK 6.21 When k is not algebraically closed, then not every smooth algebraic subgroup of G arises from an closed subgroup of G.k/. Consider, for example, the algebraic subgroup n Gm over Q. If n is odd, then n .Q/ D f1g, and the algebraic group attached to f1g is the trivial group. D EFINITION 6.22 Let G be an algebraic group over a eld k, and let k 0 be a eld containing k. We say that a subgroup of G.k 0 / is dense in G if the only algebraic subgroup H of G such that H.k 0 / is G itself. 6.23 It follows from the proof of (6.20) that G.k/ is dense in G if and only if f 2 O.G/, f .P / D 0 for all P 2 G.k/ H) f D 0: (53)
Equivalently, G.k/ is dense in G if noT nonzero element of O.G/ maps to zero under all homomorphisms of k-algebras O.G/ ! k, i.e., Kerk-alg .O.G/; k/ D 0. 6.24 If G.k/ is dense in G, then G is reduced (hence smooth if k is perfect). 6.25 If G is smooth, then G.k al / is dense in G by the Nullstellensatz (CA 11.5); in fact, G.k sep / is dense in G (AG 11.15). [In detail, assume that G is smooth, and let H be a subgroup of G. By denition, this means that O.H / D O.G/=a for some ideal a in O.G/. The elements of H.k sep / are the k-algebra homomorphisms O.H / ! k sep . To say that H.k sep / D G.k sep / means that every k-algebra homomorphism O.G/ ! k sep factors through O.G/=a, but this implies that the elements of a lie in every maximal ideal m of O.G/ such that O.G/=m is separable over k, and hence they are zero by (AG 11.15).] 6.26 If G.k/ is nite, for example, if the eld k is nite, and dim G > 0, then G.k/ is never dense in G. P ROPOSITION 6.27 If k is innite, then G.k/ is dense in G when G D Ga , GLn , or SLn . P ROOF. We use the criteria in (6.23). Recall (e.g., FT, proof of 5.18) that, because k is innite, no nonzero polynomial in kX1 ; : : : ; Xn can vanish on all of k n . This implies that no nonzero polynomial f can vanish on a set of the form D.h/ D fa 2 k n j h.a/ 0g; because otherwise hf would vanish on k n . The proposition for GLn follows from this, because GLn .k/ D fa 2 k n j det.a/ 0g:
2
def
h 0;
62
The proposition is obvious for Ga , and it can be proved for SLn by realizing O.SLn / as a subalgebra of O.GLn /. Specically, the natural bijection A; r 7! A diag.r; 1; : : : ; 1/W SLn .R/ Gm .R/ ! GLn .R/ (of set-valued functors) denes an isomorphism of k-algebras O.GLn / ' O.SLn / O.Gm /; and the algebra on the right contains O.SLn /.
2
P ROPOSITION 6.28 Let G be an algebraic group over a perfect eld k, and let D Gal.k al =k/. Then acts on G.k al /, and H $ H.k al / is a one-to-one correspondence between the smooth subgroups of G and the Zariski-closed subgroups of G.k al / stable under . P ROOF. Combine (6.20) with (4.16). (More directly, both correspond to radical ideals a in the k al -bialgebra k al O.G/ stable under the action of ; see AG 16.7, 16.8.) 2
A SIDE 6.29 (a) Let G be a smooth connected algebraic group over an innite eld k. It is known that, if either k is perfect or G is reductive, then G.k/ is dense in G (Borel 1991, 18.3, p.220). However, there exists a one-dimensional (smooth connected) unipotent group U over an innite eld k such that U.k/ is nite and hence not dense (Rosenlicht 1957, p.46). (b) If G is unirational and k is innite, then G.k/ is dense in G. Borel 1991, 18.2, in fact proves that if k is perfect and G is reduced, or G is perfect, then G is unirational over k. In SGA3, IV 6.11 one nds: One knows (Rosenlicht) examples of forms of Ga over a nonperfect eld, which have only nitely many rational points, and therefore a fortiori are not unirational. Moreover Chevalley has given an example of a torus over a eld of characteristic zero which is not a rational variety. On the other hand, it follows from the Chevalleys theory of semisimple groups that over an algebraically closed eld, every smooth connected afne algebraic group is a rational variety. For a nonunirational nonconnected algebraic group, Rosenlicht gives the example of the group of matrices a b over R with a2 C b 2 D 1. For a nonunirational connected algebraic group, Rosenlicht gives the b a example of the subgroup of pa Ga dened by Y p Y D tX p over the eld k D k0 .t / (t transcendental). p G On the other hand, if k a; b has degree 4 over k, then the norm torus associated with this extension is a three-dimensional torus that is not a rational variety. [To be rewritten.] A SIDE 6.30 Before Borel introduced algebraic geometry into the theory of algebraic groups in a more systematic way, Chevalley dened algebraic groups to be closed subsets of k n endowed with a group structure dened by polynomial maps. Hence, effectively he studied smooth algebraic groups G with the property that G.k/ is dense in G.
6e
Normalizers; centralizers; centres
For a subgroup H of an abstract group G, we let NG .H / (resp. CG .H /) denote the normalizer (resp. centralizer) of H in G, and we let Z.G/ denote the centre of G. In this subsection, we extend these notions to a subgroup H of an afne group G over k. For g 2 G.R/, let g H be the functor of R-algebras R0 g H.R0 / g
1
(subset of G.R0 /):
Dene N to be the functor of k-algebras R fg 2 G.R/ j g H D H g:
6. Group theory: subgroups and quotient groups. Thus, for any k-algebra R, N.R/ D fg 2 G.R/ j g H.R0 / g 1 D H.R0 / for all R-algebras R0 g \ D G.R/ \ NG.R0 / .H.R0 //: 0
R
63
P ROPOSITION 6.31 The functor N is an afne subgroup of G. P ROOF. Clearly N.R/ is a subgroup of G.R/, and so it remains to show that N is representable. Clearly g H.R0 / g
1
D H.R0 / g H.R0 / g
H.R0 / and g
H.R0 / g
H.R0 /;
and so, when we let G act on itself by conjugation, N D TG .H; H / \ TG .H; H /

1
(notations as in 5k). Proposition 5.28 shows that TG .H; H / is representable, and it follows from (6.10) that N is representable. 2 The subgroup N of G is called the normalizer NG .H / of H in G. It is obvious from its denition that the formation of NG .H / commutes with extension of the base eld, i.e., for any eld k 0 k, .NG .H //k 0 ' NGk0 .Hk 0 /. C OROLLARY 6.32 Let G be an algebraic group, and let k 0 k be such that G.k 0 / is dense in G. A subgroup H of G is normal in G if H.k 0 / is normal in G.k 0 /. P ROOF. The condition says that NG .H /.k 0 / D G.k 0 /, and so NG .H / D G. P ROPOSITION 6.33 If H is algebraic and H.k 0 / is dense in H for some eld k 0 NG .k/ D G.k/ \ NG.k 0 / .H.k 0 //: P ROOF. Let g 2 G.k/ \ NG.k 0 / .H.k 0 //. Because g 2 G.k/, g H is an algebraic subgroup of G, and so g H \ H is an algebraic subgroup of H . Because g 2 NG.k 0 / .H.k 0 //,
g 2
k, then
H .k 0 / D H.k 0 /;
(54)
and so .g H \ H / .k 0 / D H.k 0 /. As H.k 0 / is dense in H , this implies that g H \ H D H , and so gH D H. 2 P ROPOSITION 6.34 Let H be a subgroup of an afne group G. For any eld k 0 containing k; NGk0 .Hk 0 / ' NG .H /k 0 CGk0 .Hk 0 / ' CG .H /k 0 : In other words, the formation of normalizers and centralizers commutes with extension of the base eld. P ROOF. It follows directly from the denitions that the statements are true for the functors; therefore they are true for the afne groups. 2
64
BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1) Let N be the normalizer of H . Each n 2 N.R/ denes a natural transformation in h 7! nhn
1
W H.R0 / ! H.R0 /
of H regarded as a functor from the category of R-algebras to sets, and we dene C to be the functor of k-algebras R fn 2 N.R/ j in D idH g: Thus, C.R/ D G.R/ \ \
R0
CG.R0 / .H.R0 //:
P ROPOSITION 6.35 The functor C is an afne subgroup of G. P ROOF. When we let G act on itself by conjugation, \ CD TG .h; h/;
h2jH j
and so Propositions 5.28 and 6.10 imply that C is representable.
The subgroup C of G is called the centralizer CG .H / of H in G. It is obvious from its denition that the formation of CG .H / commutes with extension of the base eld. C OROLLARY 6.36 A subgroup H of a smooth algebraic group G is central in G if H.k al / is central in G.k al /. P ROOF. The condition says that CG .H /.k al / D G.k al /, and we know that G.k al / is dense in G (6.25). 2 P ROPOSITION 6.37 If H is algebraic and H.k 0 / is dense in H for some eld k 0 CG .H /.k/ D G.k/ \ CG.k 0 / .H.k 0 //: P ROOF. Let n 2 G.k/ \ CG.k 0 / .H.k 0 //. Then n 2 NG .H / and the automorphisms in and idH of H agree on an algebraic subgroup Z of H , which has the property that Z.k 0 / D H.k 0 /. Therefore Z D H , and so in 2 CG .H /. 2 The centre Z.G/ of an afne group G is dened to be CG .G/. It is an afne subgroup of G, and if G is algebraic and G.k 0 / is dense in G, then Z.G/.k/ D G.k/ \ Z.G.k 0 //: k, then
6.38 Even when G and H are smooth, CG .H / need not be smooth. For example, it is possible for CG .H / to be nontrivial but for CG .H /.k 0 / to be trivial for all elds k 0 k. To see this, let G be the functor R R R with the multiplication .a; u/.b; v/ D .a C bup ; uv/; here 0 p D char.k/: This is an algebraic group because, as a functor to sets, it is isomorphic to Ga Gm . For a pair .a; u/ 2 R R , .a; u/.b; v/ D .b; v/.a; u/ for all .b; v/ if and only if up D 1. Therefore, the centre of G is p , and so Z.G/.k 0 / D 1 for all elds k 0 containing k. Another example is provided by SLp over a eld of characteristic p. The centre of SLp is p , which is not smooth.
6. Group theory: subgroups and quotient groups. E XAMPLE 6.39 Let G D GLn over a eld k. For an integer N , let HN be the subgroup R of G. Thus HN ' .
N HN .R/ D fdiag.a1 ; : : : ; an / 2 GLn .R/ j a1 D n N/ . N D an D 1g.
65
For N sufciently large CG .HN / D Dn
(group of diagonal matrices) (see (13.23)). We consider two cases. (a) k D Q and N odd. Then HN .k/ D f1g, and CG.k/ .HN .k// D GLn .k/ Dn .k/ D CG .HN /.k/: (b) k is algebraic closed of characteristic p 0 and N is a power of p. Then HN .k/ D 1 and CG.k/ .HN .k// D GLn .k/ Dn .k/ D CG .HN /.k/: An afne subgroup H is said to normalize (resp. centralize) an afne subgroup N of G if H.R/ normalizes (resp. centralizes) N.R/ for all k-algebras R; equivalently, if H NG .N / (resp. H CG .N /).
6f
Quotient groups; surjective homomorphisms
What does it mean for a homomorphism of algebraic groups G ! Q to be surjective? One might guess that it means that G.R/ ! Q.R/ is surjective for all R, but this condition is too stringent. For example, it would say that x 7! x n W Gm ! Gm is not surjective even though x 7! x n W Gm .k/ ! n Gm .k/ is surjective whenever k is algebraically closed. In fact, Gm ! Gm is surjective according to the following denition. D EFINITION 6.40 A homomorphism G ! Q is said to be surjective (and Q is called a quotient of G) if for every k-algebra R and q 2 Q.R/, there exists a faithfully at R-algebra R0 and a g 2 G.R0 / mapping to the image of q in Q.R0 /: G.R0 / O G.R/
/ Q.R0 / O / Q.R/
9g
/
q:
In other words, a homomorphism G ! Q is surjective if every q 2 Q.R/ lifts to G after a faithfully at extension. A surjective homomorphism is also called a quotient map. T HEOREM 6.41 A homomorphism G ! Q is surjective if and only if O.Q/ ! O.G/ is injective. P ROOF. ): Consider the universal element idO.Q/ 2 Q.O.Q//. If G ! Q is surjective, there exists a g 2 G.R0 / with R0 faithfully at over O.Q/ such that g and idO.Q/ map to the same element of Q.R0 /, i.e., such that the diagram O.G/ ? ?g y R0
faithfully at
O.Q/ ? ?id y O.Q/ O.Q/
66
commutes. The map O.Q/ ! R0 , being faithfully at, is injective (CA 9.6), which implies that O.Q/ ! O.G/ is injective. (: According to (5.32) O.Q/ ! O.G/ is faithfully at. Let q 2 Q.R/. Regard q as a homomorphism O.Q/ ! R, and form the tensor product R0 D O.G/ O.Q/ R: O.G/ o
gD1q faithfully at
R0 D O.G/ O.Q/ R o
nnq vnnn
? _ O.Q/ n n nnn 0n
q
(55)
? _R
Then R0 is a faithfully at R-algebra because O.G/ is a faithfully at O.H /-algebra (apply CA 9.7). The commutativity of the square in (55) means that g 2 G.R0 / maps to the image q 0 of q in Q.R0 /. 2 P ROPOSITION 6.42 Let W H ! G be a homomorphism of afne groups. If is surjective, then so also is k 0 W Hk 0 ! Gk 0 for any eld k 0 containing k. Conversely, if k 0 is surjective for one eld k 0 containing k, then is surjective. P ROOF. Because k ! k 0 is faithfully at, the map O.G/ ! O.H / is injective if and only if k 0 k O.G/ ! k 0 k O.H / is injective (see CA 9.2). 2 P ROPOSITION 6.43 A homomorphism of afne groups that is both injective and surjective is an isomorphism. P ROOF. The map on coordinate rings is both surjective and injective, and hence is an isomorphism. 2 P ROPOSITION 6.44 Let G ! Q be a homomorphism of algebraic groups. If G ! Q is a quotient map, then G.k al / ! Q.k al / is surjective; the converse is true if Q is smooth. P ROOF. Let q 2 Q.k al /. For some nitely generated k al -algebra R, the image of q in Q.R/ lifts to an element g of G.R/. Zariskis lemma (CA 11.1) shows that there exists a k al -algebra homomorphism R ! k al , and the image of g in G.k al / maps to q 2 Q.k al /:
id
k al
/R
/(
G.R/ k al Q.k al /
/ Q.R/
/ G.k al / / Q.k al /
g _ q
/ qR
/ gk al _ /q
For the converse, we may suppose that k is algebraically closed. Recall (2.11) that an element f of O.Q/ is a family .fR /R with fR a map Q.R/ ! R. Because Q is smooth, O.Q/ is reduced, and so f is determined by fk (CA 11.8). As G.k/ ! Q.k/ is surjective, f is determined by the composite G.k/ ! Q.k/ ! k, and so O.Q/ ! O.G/ is injective.
fk 2
k0
More generally, a homomorphism W G ! H of algebraic groups is surjective if, for some eld containing k, the image of G.k 0 / in H.k 0 / is dense in H (8.9 below).
67
6.45 The smoothness condition in the second part of the proposition is necessary. Let k be a eld of characteristic p 0, and consider the homomorphism 1 ! p where 1 denotes the trivial algebraic group. The map 1.k al / ! p .k al / is f1g ! f1g, which is surjective, but 1 ! p is not a quotient map because the map on coordinate rings is kX =.X p / ! k, which is not injective. T HEOREM 6.46 Let G ! Q be a quotient map with kernel N . Then any homomorphism G ! Q0 whose kernel contains N factors uniquely through Q: N G
0
Q0 .
P ROOF. Note that, if g and g 0 are elements of G.R/ with the same image in Q.R/, then g 1 g 0 lies in N and so maps to 1 in Q0 .R/. Therefore g and g 0 have the same image in Q0 .R/. This shows that the composites of the homomorphisms G
QG
G ! Q0
are equal. Therefore, the composites of the homomorphisms O.G/ O.Q/ O.G/ O.G/ O.Q0 /
are equal. The subring of O.G/ on which the two maps coincide is O.Q/ (CA 9.6), and so the map O.Q0 / ! O.G/ factors through uniquely through O.Q/ ,! O.G/. Therefore G ! Q0 factors uniquely through G ! Q. 2 C OROLLARY 6.47 If W G ! Q and 0 W G ! Q0 are quotient maps with the same kernel, then there is a unique homomorphism W Q ! Q0 such that D 0 ; moreover, is an isomorphism. P ROOF. From the theorem, there are unique homomorphisms W Q ! Q0 and 0 W Q0 ! Q such that D 0 and 0 0 D . Now 0 D idQ , because both have the property that D . Similarly, 0 D idQ0 , and so and 0 are inverse isomorphisms. 2 When G ! Q is a surjective map with kernel N , the afne group Q is called the quotient of G by N , and is denoted G=N . It is uniquely determined up to a unique isomorphism by the universal property in (6.46). We shall see later (7.59) that quotients by normal subgroups always exist. D EFINITION 6.48 A sequence 1!N !G!Q!1 is exact if G ! Q is a quotient map with kernel N .
A SIDE 6.49 A morphism uW A ! B in some category A is said to be an epimorphism if Hom.B; T / ! Hom.A; T / is injective for all objects T .
68
It is obvious from Theorem 6.41 that a surjective homomorphism of afne groups is an epimorphism. The converse is true for groups (MacLane 1971, Exercise 5 to I 5), but it is false for afne groups. For example, the embedding BD 0 ,! D GL2 is a nonsurjective epimorphism (any two homomorphisms from GL2 that agree on B are equal).13
6g
Existence of quotients
P ROPOSITION 6.50 Let G be an algebraic group, and let H be a subgroup. There exists a surjective homomorphism G ! Q containing H in its kernel and universal among homomorphisms with this property. P ROOF. ForT nite family .G ! Qi /i 2I of surjective morphisms such that H Ker.qi / all any i , let HI D i 2I Ker.qi /. According to (6.8), there exists a family for Q which HI is minimal. For such a family, I claim that the map from G to the image of .qi /W G ! i 2I Qi is universal. If it isnt, then there exists a homomorphism qW G ! Q containing H in its kernel but not HI . But then HI [fqg D HI \ Ker.q/ is properly contained in HI . 2 T HEOREM 6.51 Let N be a normal subgroup of an algebraic group G. The universal surjective homomorphism G ! Q containing N in its kernel has kernel exactly N . This is a consequence of the next lemma, which will be proved in (7.55) below. L EMMA 6.52 Let N N 0 be normal subgroups of an algebraic group G. If N N 0 , then there exists a homomorphism qW G ! Q such that q.N / D 1 but q.N 0 / 1. We briey sketch the proof. According to a theorem of Chevalley (7.46), there exists a representation G ! GLV and a line L in V such that N is the stabilizer of L in V . Let W be the largest subspace of V V _ on which N acts trivially. Then N 0 acts nontrivially on W because it acts nontrivially on L L_ W .
qi
6h
Semidirect products
D EFINITION 6.53 An afne group G is said to be a semidirect product of its subgroups N and Q, denoted G D N Q, if N is normal in G and the map .n; q/ 7! nqW N.R/ Q.R/ ! G.R/ is a bijection of sets for all k-algebras R. In other words, G is a semidirect product of the subgroups N and Q if G.R/ is a semidirect product of its subgroups N.R/ and Q.R/ for all k-algebras R (cf. GT 3.7). For example, let Tn be the algebraic group of upper triangular matrices, so Tn .R/ D f.aij / 2 GLn .R/ j aij D 0 for i > j g: Then Tn is the semidirect product of its (normal) subgroup Un and its subgroup Dn .
follows from the fact that GL2 =B ' P1 . Let f; f 0 be two homomorphisms GL2 ! G. If f jB D f 0 jB, then g 7! f f .g/ 1 denes a map P1 ! G, which has image 1G because G is afne and P1 is complete (see AG 7.5). Alternatively, in characteristic zero, one can show that any homomorphism of B \ SL2 has at most one extension to SL2 because any nite dimensional representation of sl2 can be reconstructed from the operators h and y. Specically, if hv D mv and y mC1 v D 0, then xv D 0; if hv D mv and u D y n v, then xy n v can be computed as usual using that x; y D h.
13 This 0 .g/
69
P ROPOSITION 6.54 Let N and Q be afne subgroups of an algebraic group G. Then G is the semidirect product of N and Q if and only if there exists a homomorphism G ! Q whose restriction to Q is the identity map and whose kernel is N . P ROOF. )W By assumption, the product map is a bijection of functors N Q ! G. The composite of the inverse of this map with the projection N Q ! Q has the required properties. (W Let 'W G ! Q be the given homomorphism. For each k-algebra R, '.R/ realizes G.R/ as a semidirect product G.R/ D N.R/ Q.R/ of its subgroups N.R/ and Q.R/. 2 Let G be an afne group and X a functor from the category of k-algebras to sets. Recall 5k that an action of G on X is a natural transformation W G X ! X such that each map G.R/ X.R/ ! X.R/ is an action of the group G.R/ on the set X.R/. Now let N and Q be algebraic groups and suppose that there is given an action of Q on N .q; n/ 7! R .q; n/W Q.R/ N.R/ ! N.R/ such that, for each q, the map n 7! R .q; n/ is a group homomorphism. Then the functor R N.R/
R
Q.R/
(cf. GT 3.9) is an afne group because, as a functor to sets, it is N Q, which is represented by O.N / k O.G/. We denote it by N Q, and call it the semidirect product of N and Q dened by .
6i
Algebraic groups as sheaves
Some of the above discussion simplies when regard afne groups as sheaves. P ROPOSITION 6.55 Let F be a functor from the category of k-algebras to sets. If F is representable, then Q Q (F1) for every nite family of k-algebras .Ri /i 2I , the canonical map F . i Ri / ! i F .Ri / is bijective; (F2) for every faithfully at homomorphism R ! R0 of k-algebras, the sequence F .R/ ! F .R0 / F .R0 R R0 /
is exact (i.e., the rst arrow realizes F .R/ as the equalizer of the pair of arrows). P ROOF. (F1). For any k-algebra A, it follows directly from the denition of product that Q Q Hom.A; i 2I Ri / ' i 2I Hom.A; Ri /; (F2). If R ! R0 is faithfully at, then it is injective, and so Hom.A; R/ ! Hom.A; R0 / is injective for any k-algebra A. According to (CA 9.5), the sequence R ! R0 is exact, and it follows that Homk-alg .A; R/ ! Homk-alg .A; R0 / is exact. Homk-alg .A; R0 R R0 /
2
R0 R R0
70
BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1) A functor satisfying the conditions (F1) and (F2) is said to be a sheaf for the at topology14 .
P ROPOSITION 6.56 A functor F W Algk ! Set is a sheaf if and only if it satises the following condition: Q (S) for every k-algebra R and nite family .Ri /i 2I of k-algebras such that R ! i Ri is faithfully at, the sequence F .R/ ! is exact. P ROOF. Easy exercise (cf. Milne 1980, II 1.5).
2
i 2I
F .Ri /
.i;i 0 /2I I
F .Ri k Ri 0 /
Q We sometimes use (S1) to denote the condition that F .R/ ! i 2I F .Ri / is injective and (S2) for the condition that its image is subset on which the pair of maps agree. P ROPOSITION 6.57 For any functor F W Algk ! Set, there exists a sheaf aF and a natural transformation F ! aF that is universal among natural transformations from F to sheaves. P ROOF. For a; b 2 F .R/, set a b if a and b have the same image in F .R0 / for some faithfully at R-algebra R0 . Then is an equivalence relation on F .R/, and the functor R F .R/= satises (S1). Moreover, any natural tranformation from F to a sheaf will factor uniquely through F ! F= . Now let F be a functor satisfying (S1). For any k-algebra R, dene F 0 .R/ D lim Ker.F .R0 / ! F .R0 R R0 //:
where R0 runs over the faithfully at R-algebras. One checks easily that F 0 is a sheaf, and that any natural transformation from F to a sheaf factors uniquely through F ! F 0 . 2 The sheaf aF is called the associated sheaf of F . P ROPOSITION 6.58 Let S be a sheaf, and let F be a subfunctor of S. If [ S.R/ D S.R/ \ F .R0 / 0
R a faithfully at R-algebra
(intersection inside S.R0 /), then S is the sheaf associated with F . P ROOF. Obviously any natural transformation F ! F 0 with F 0 a sheaf extends uniquely to S .
2
Let P be the category of functors Algk ! Set, and let S be the full subcategory of P consisting of the sheaves. P ROPOSITION 6.59 The inclusion functor i W S ! P preserves inverse limits; the functor aW P ! S preserves direct limits and nite inverse limits.
14 Strictly,
for the fpqc (d` lement plat quasi-compacte) topology. e
71
P ROOF. By denition Hom.a. /; / ' Hom. ; i. //, and so a and i are adjoint functors. This implies (immediately) that i preserves inverse limits and a preserves direct limits. To show that a preserves nite inverse limits, it sufces to show that it preserves nite products and equalizers, which follows from the construction of a. 2 P ROPOSITION 6.60 Let G ! Q be a surjective homomorphism of afne groups with kernel N . Then Q represents the sheaf associated with the functor R P ROOF. Let P be the functor R show: G.R/=N.R/:
G.R/=N.R/. Then P commutes with products, and we shall
(a) For any injective homomorphism R ! R0 of k-algebras, the map P .R/ ! P .R0 / is injective. (b) Let P 0 .R/ D lim Ker.P .R0 / ! 0
R
P .R0 R R0 //
where the limit is over all faithfully at R-algebras; then P 0 ' Q: For (a), we have to prove that N.R/ D N.R0 / \ G.R/ (intersection inside G.R0 /).
For some index set I , N.R/ is the subset of RI dened by some polynomial conditions fj .: : : ; Xi ; : : :/ D 0 and N.R0 / is the subset of R0I dened by the same polynomial conditions. But if an element of RI satises the conditions when regarded as an element of R0I , then it already satises the conditions in RI (because R ! R0 is injective). For (b), consider the diagram K.R0 / ! P .R0 / ? ? y Q.R/ ! Q.R0 / P .R0 R R0 / ? ? y Q.R0 R R0 /
where K.R0 / is the equalizer of the top pair of maps we know that Q.R/ is the equalizer of the bottom pair of maps. For any k-algebra R0 , the map P .R0 / ! Q.R0 / is injective, and so the two vertical arrows induce an injective homomorphism K.R0 / ! Q.R/. When we pass to the limit over R0 , it follows directly from the denition of surjective (see 6.40) that this map becomes surjective. 2
A SIDE 6.61 Discuss why it is useful to regard the afne groups as sheaves rather than presheaves.
72
6j
Limits of afne groups
Recall (MacLane 1971, III 4, p.68) that, for a functor F W I ! C from a small category I to a category C, there is the notion of an inverse limit of F (also called a projective limit, or just limit). This generalizes the notions of a limit over a directed set and of a product. T HEOREM 6.62 Let F be a functor from a small category I to the category of afne groups over k; then the functor R lim F .R/ (56)
is an afne group, and it is the inverse limit of F in the category of afne groups. P ROOF. Denote the functor (56) by F ; thus F .R/ is the inverse limit of the functor i Fi .R/ from I to the category of (abstract) groups. It is easy to see that F D lim F in the category of functors from k-algebras to groups, and it will follow that F is the inverse limit in the category of afne groups once we show that it is an afne group. But F is equal to the equalizer of two homomorphisms Y Y Fi Ftarget.u/ (57)
i 2ob.I / u2ar.I /
(MacLane 1971, V 2 Theorem 2, p.109). Both products are afne groups, and the equalizer of the two homomorphisms is equal to their bre product, and so that also is an afne group (see 3b). 2 In particular, inverse limits of algebraic groups exist as afne groups. Later (7.21) we shall see that every afne group arises in this way. T HEOREM 6.63 Let F be a functor from a nite category I to the category of algebraic groups over k; then the functor R lim Fi .R/ (58)
is an algebraic group, and it is the inverse limit of F in the category of algebraic groups. P ROOF. Both products in (57) are algebraic groups.
2
Direct limits, even nite direct limits, are more difcult. For example, the sum of two groups is their free product, but when G1 and G2 are algebraic groups, the functor R G1 .R/ G2 .R/ will generally be far from being an algebraic group. Moreover, the functor R lim Fi .R/ need !I not be a sheaf. Roughly speaking, when the direct limit of a system of afne groups exists, it can be constructed by forming the naive direct limit in the category of functors, and then forming the associated sheaf. For example, when N is a normal subgroup of an afne group G, the quotient afne group G=N is the sheaf associated with the functor R G.R/=N.R/ (cf. 6.60).
6k
Exercises
E XERCISE 6-1 Let A and B be subgroups of the afne group G, and let AB be the sheaf associated with the subfunctor R A.R/ B.R/ of G.
73
(a) Show that AB is representable by O.G/=a where a is the kernel of homomorphism O.G/ ! O.A/ O.B/ dened by the map a; b 7! abW A B ! G (of set-valued functors). (b) Show that, for any k-algebra R, an element G.R/ lies in .AB/ .R/ if and only if its image in G.R0 / lies in A.R0 / B.R0 / for some faithfully at R-algebra R0 , i.e., \ .AB/ .R/ D G.R/ \ A.R0 / B.R0 / . 0
R
(c) Show that AB is a subgroup of G if B normalizes A. E XERCISE 6-2 Let A, B, C be subgroups of an afne group G such that A is a normal subgroup of B. Show: (a) C \ A is a normal subgroup of C \ B; (b) CA is a normal subgroup of CB. E XERCISE 6-3 (Dedekinds modular laws). Let A, B, C be subgroups of an afne group G such that A is a subgroup of B. Show: (a) B \ AC D A.B \ C /I (b) if G D AC , then B D A.B \ C /.
Representations of afne groups
One of the main results in this section is that all algebraic groups can be realized as subgroups of GLn for some n. At rst sight, this is a surprising result, because it says that all possible multiplications in algebraic groups are just matrix multiplication in disguise. Before looking at the case of algebraic groups, we should review how to realize a nite group as a matrix group. Let G be a nite group. A representation of G on a k-vector space V is a homomorphism of groups G ! Autk-lin .V /, or, in other words, an action G V ! V of G on V in which each 2 G acts k-linearly. Let X G ! X be a right action of G on a nite set X. Dene V to be the k-vector space of maps X ! k, and let G act on V according to the rule: .gf /.x/ D f .xg/ for g 2 G, f 2 V , x 2 X:
This denes a representation of G on V , which is faithful if G acts effectively on X. The vector space V has a basis consisting of the maps that send one element of X to 1 and the remainder to 0, and so this gives a homomorphism G ! GLn .k/ where n is the order of X. For example, for the symmetric group Sn acting on f1; 2; : : : ; ng, this gives the map 7! I. /W Sn ! GLn .k/ in 1. When we take X D G, the representation we get is called the regular representation, and the map G ! Autk-lin .V / is injective. In this section, we often work with algebraic monoids rather than groups since this forces us to distinguish between left and right correctly. Note that for a commutative ring R, the only difference between a left module and a right module is one of notation: it is simply a question of r whether we write rm or mr (or better m). In this section, it will sometimes be convenient to regard R-modules as right modules, and write V k R instead of R k V . Throughout, k is a eld, and tensor products are over k unless otherwise noted.
7a
Denition of a representation
Let V be a vector space over k. For a k-algebra R, we let V .R/ D V R; EndV .R/ D EndR-lin .V .R//; AutV .R/ D AutR-lin .V .R//; (R-module) (monoid under composition) (group under composition).
Then R EndV .R/ is a functor from the category of k-algebras to monoids and R AutV .R/ is a functor from the category of k-algebras to groups. With the terminology of (2.27), AutV D EndV . Let G be an afne monoid or group over k. A linear representation of G on a k-vector space V is a natural transformation rW G ! EndV of functors Algk ! Mon. In other words, it is a family of homomorphisms of monoids rR W G.R/ ! EndR-lin .V .R//; R a k-algebra, (59)
such that, for every homomorphism R ! R0 of k-algebras, the diagram G.R/ ? ? y G.R0 /
rR
! EndR-lin .V .R// ? ? y ! EndR0 -lin .V .R0 // 74
rR0
7. Representations of afne groups
75
commutes. When G is an afne group, r takes values in AutV and is a natural transformation of group-valued functors. A linear representation is said to be nite-dimensional when V is nitedimensional as a k-vector space, and it is faithful if all the homomorphisms rR are injective. A subspace W of V is a subrepresentation if rR .g/.W .R// W .R/ for all k-algebras R and all g 2 G.R/. A homomorphism of linear representations .V; r/ ! .V 0 ; r 0 / is a k-linear map W V ! V 0 such that V .R/ ! V 0 .R/ ? ? ? ?r 0 .g/ rR .g/ y yR V .R/
.R/ .R/
! V 0 .R/
commutes for all g 2 G.R/ and all k-algebras R. We write V also for the functor R V .R/ dened by V . Then a linear representation of G on V can also be dened as an action of G on V; G V ! V; (60)
such that each g 2 G.R/ acts R-linearly on V .R/. When V D k n , EndV is the monoid R .Mn .R/; / and AutV D GLn . A linear representation of an afne monoid (resp. group) G on V is a homomorphism G ! .Mn ; / (resp. G ! GLn ). E XAMPLE 7.1 Let G D Ga . Let V be a nite-dimensional k-vector space, and let 0 ; : : : ; i ; : : : be a sequence of endomorphisms V such that all but a nite number are zero. For t 2 R, let X i rR .t / D i t 2 End.V .R//
i 0
(so rR .t/.v c/ D
i .v/ ct
i ).
If
0
D idV D i Cj i
i Cj
all i; j
0;
(61)
then rR .t C t 0 / D rR .t / C rR .t 0 / for all t; t 0 2 R; and so rR is a representation. We shall see later (7.14) that all nite-dimensional representations of n Ga are of this form. Note that (61) implies that i 1 D .i C 1/ i C1 , and so 1 D n n . When k n has characteristic zero, this implies that 1 is nilpotent and that n D 1 =n, and so rR .t / D X . 1 t /n =n D exp. 1 t /:
When k has nonzero characteristic, there are more possibilities. See Abe 1980, p185. E XAMPLE 7.2 Let G D GLn , and let Mn denote the vector space of all n n matrices with entries in k. The actions .P; A/ 7! PAP 1 W G.R/ Mn .R/ ! Mn .R/ dene a linear representation of G on Mn . The orbits of G.k/ acting on Mn .k/ are the similarity classes, which are represented by the Jordan matrices when k is algebraically closed.
76
E XAMPLE 7.3 There is a unique linear representation r of G on O.G/ (regarded as a k-vector space) such that .gf /R .x/ D fR .xg/; for all g 2 G.R/, f 2 O.G/, x 2 G.R/: (62)
This is called the regular representation. In more detail: the formula (62) denes a map G.R/ O.G/ ! R O.G/, which extends by linearity to a map G.R/ R O.G/ ! R O.G/.
7b
Terminology
From now on, representation will mean linear representation.
7c
Comodules
Let .A; m; e/ be a k-algebra, not necessarily commutative. Recall that a left A-module is a k-vector space V together with a k-linear map W A V ! V such that the diagrams V x ? ? AV
A
AV x ? ?mV AAV
AV x ? ?eV
'
(63)
k V
commute. On reversing the directions of the arrows, we obtain the notion of comodule over a coalgebra. D EFINITION 7.4 Let .C; ; / be a k-coalgebra. A right C -comodule15 is a k-linear map W V ! V C (called the coaction of C on V ) such that the diagrams V ? ? y V C commute, i.e., such that .V / .V / D . C/ D V: ! V C ? ? yV V ! V C ? ? yV
'
(64)
! V C C
! V k
A homomorphism W .V; / ! .V 0 ; 0 / of C -comodules is a k-linear map W V ! V 0 such that the diagram V ? ? y V C
V0 ? ? 0 y
! V 0 C
commutes. A comodule is said to be nite-dimensional if it is nite-dimensional as a k-vector space.

would be more natural to consider left comodules, except that it is right comodules that correspond to left representations of monoids. Because we consider right comodules we are more-or-less forced to write V R where elsewhere we write R V .
15 It
77
E XAMPLE 7.5 (a) The pair .C; / is a right C -module (compare (26), p.28, with (64)). More generally, for any k-vector space V , V WV C ! V C C is a right C -comodule (called the cofree comodule on V ). (b) Let .V1 ; 1 / and .V2 ; 2 / be comodules over coalgebras C1 and C2 respectively. The map V1 V2
1 2
! V1 C1 V2 C2 ' V1 V2 C1 C2
provides V1 V2 with the structure of a C1 C2 -comodule. (c) Let .V; / be a right C -comodule, and let W C ! C 0 be a homomorphism of coalgebras. The map V ! V C
V
! V C0
provides V with the structure of a right C 0 -comodule. (d) Let V be a k-vector space, and let W V ! V C be a k-linear map. Choose a basis .ei /i 2I for V , and write X .ej / D ei cij ; cij 2 C; (65)
i 2I
(nite sum, so, for each j , almost all cij s are zero). Then .V; / is a right comodule if and only if P .cij / D k2I ci k ckj all i; j 2 I: (66) .cij / D ij For a module V over an algebra A, there is a smallest quotient of A, namely, the image of A in Endk .V //, through which the action of A on V factors. The next remark shows that for a comodule V over a coalgebra C , there is a smallest subcoalgebra of C through which the co-action of C on V factors. R EMARK 7.6 Let .V; / is a C -comodule. (a) When we choose a k-basis .ei /i 2I for V , the equations (66) show that the k-subspace spanned by the cij is a subcoalgebra of C , which we denote CV . Clearly, it is the smallest subspace of C such that .V / V CV , and so it is independent of the choice of the basis. (b) Recall that for a nite-dimensional k-vector space V , Homk-lin .V; V C / ' Homk-lin .V V _ ; C /: If $
0
under this isomorphism, then X ei ci H) .v/ D

i 2I
.v f / D
X
i 2I
f .ei /ci :
_ In particular, 0 .ej ei / D cij (notation as in (65)). Therefore CV is the image of 0 W V V _ ! C . (c) If .V; / is a sub-comodule of .C; /, then V CV . To see this, note that the restriction of the co-identity of C to V is an element V of V _ and that 0 .v V / D v for all v 2 V because X 0 .ej V / D .ej /cij i2I
D . idC / .ej / D .idC / .ej / X D ej .cij /

i 2I
(by (26), p.28)
D ej
(by (66)).
78
R EMARK 7.7 Recall (4c) that the linear dual of a coalgebra .C; ; / is an associative algebra with identity .C _ ; _ ; _ /. Let V be a k-vector space, and let W V ! V C be a k-linear map. Dene to be the composite of C_ V
C _
! C_ V C ' V C_ C
V ev
! V k ' V
where evW C _ C ! k is the evaluation map. One can check that .V; / is a right C -comodule if and only if .V; / is a left C _ -module. When C and V are nite-dimensional, 7! is a bijection Homk-lin .V; V C / ' Homk-lin .C _ V; V /, and so there is a one-to-one correspondence between the right C -comodule structures on V and the left C _ -module structures on V . In the general case, not every C _ -module structure arises from a C -comodule structure, but it is known which do (Sweedler 1969, 2.1). A k-subspace W of V is a subcomodule if .W / comodule. W C . Then .W; jW / is itself a C -
P ROPOSITION 7.8 Every comodule .V; / is a ltered union of its nite-dimensional sub-comodules. P ROOF. As a sum of sub-comodules is a sub-comodule, it sufces to show that each element v of V is contained in a nite-dimensional sub-comodule. Let .ei /i 2I be a basis for C as a k-vector space, and let X .v/ D vi ei ; vi 2 V;
i
(nite sum, i.e., only nitely many vi are nonzero). Write X .ei / D rij k .ej ek /;
j;k
rij k 2 k.
We shall show that .vk / D X

i;j
rij k vi ej
(67)
from which it follows that the k-subspace of V spanned by v and the vi is a subcomodule containing v. Recall from (64) that .V / D . C / : On applying each side of this equation to v, we nd that X X rij k .vi ej ek / D .vk / ek
i;j;k k
(inside V C C /:
2
On comparing the coefcients of 1 1 ek in these two expressions, we obtain (67).
C OROLLARY 7.9 A coalgebra C is a union of its sub-coalgebras CV , where V runs over the nitedimensional C -comodules. P ROOF. For any nite-dimensional sub-comodule V of C , V CV C
2
(see 7.6), and so this follows from the proposition.
A SIDE 7.10 The main denitions in this subsection require only that k be a commutative ring. When k is noetherian, every comodule over a k-coalgebra C is a ltered union of nitely generated subcoalgebras (Serre 1993, 1.4).
79
7d
The category of comodules over C
Let C be a coalgebra over k. With the obvious denitions, the standard isomorphism theorems (cf. 8.1, 8.2, 8.3, 8.4) hold in comodules over C . For example, if .W; W / is a sub-comodule of .V; V /, then the quotient vector space V =W has a (unique) comodule structure V =W for which .V; V / ! .V =W; V =W / is a homomorphism. Therefore every sub-comodule arises as the kernel of a homomorphism of comodules. The category of comodules over C is abelian and the forgetful functor to k-vector spaces is exact. A bialgebra structure .m; e/ on C denes a tensor product structure on the category of comodules over C . Specically, as we observed (7.5b), when .V1 ; 1 / and .V2 ; 2 / are C -comodules, V1 V2 has a natural structure of C C -comodule; the homomorphism of coalgebras mW C C ! C turns this into a C -comodule structure (7.5c). Moreover, the homomorphism of coalgebras eW k ! C ' k C denes a C -comodule structure on k this is the trivial comodule. Assume that V is nite dimensional. Under the canonical isomorphisms Homk-lin .V; V C / ' Homk-lin .V V _ ; C / ' Homk-lin .V _ ; C V _ /; right comodule structures on V correspond to left comodule structures on V _ . When C is a bialgebra with an inversion, the inversion can be used to turn the left comodule structure on V _ into a right comodule structure. The space V _ with this structure is called the dual or contragredient of .V; /. S UMMARY 7.11 Let Comod.C / be the category of nite-dimensional comodules over a k-coalgebra C . Then Comod.C / is an abelian category; a bialgebra structure on C endows Comod.C / with tensor products; a Hopf algebra structure on C endows Comod.C / with tensor products and duals.
7e
Representations and comodules
A comodule over a bialgebra .A; m; e; ; / is dened to be a comodule over the coalgebra .A; ; /. P ROPOSITION 7.12 Let G be an afne monoid over k. For any k-vector space V , there is a natural one-to-one correspondence between the linear representations of G on V and the O.G/-comodule structures on V . We rst describe the correspondence in the case that V is nite dimensional. The choice of a basis .ei /i 2I for V identies EndV with Mn and the natural transformations of set-valued functors rW G ! EndV with the families of .rij /.i;j /2I I of regular functions on G: rR .g/ ej D P
i 2I
rij
.g/ ei
for all k-algebras R and g 2 G.R/:
The natural transformation r is a morphism of afne monoids if and only if P rij D k2I ri k rkj all i; j 2 I: rij .1G.R/ / D ij (Kronecker delta)
(68)
On the other hand, to give a k-linear map W V ! V O.G/ is the same as giving a family P .rij /1i;j n of elements of O.G/ (write .ej / D ei rij ), and is a co-action if and only if (66), p.77, holds. On reviewing the denition of and (4h), one sees that the conditions (66) and (68) correspond. Therefore r $ .rij / $
80
gives a one-to-one correspondence between the linear representations of G on V and the O.G/comodule structures on V . In the proof of Proposition 7.12 below, we construct a canonical correspondence between the representations and the comodule structures, and in Proposition 7.17 we show that, once a basis is chosen, the correspondence becomes that just described. P ROOF ( OF P ROPOSITION 7.12) Let A D O.G/. We prove the following more precise result: Let rW G ! EndV be a representation; the universal element u D idA in G.A/ ' def Homk-alg .A; A/ maps to an element of EndV .A/ D EndA-lin .V .A// whose restriction to V V .A/ is a comodule structure W V ! V A on V . Conversely, a comodule structure on V determines a representation r such that, for R a k-algebra and g 2 G.R/, the restriction of rR .g/W V .R/ ! V .R/ to V V .R/ is V ! V A ! V R: These operations are inverse. Let V be a vector space over k, and let rW G ! EndV be a natural transformation of set-valued functors. Let g 2 G.R/ D Homk-alg .A; R/, and consider the diagram:
/ V A V NNN NNN NNN rA .u/ NN def DrA .u/jV N' V A
v7!v1 V g V g
/ V R / V R:
rR .g/
V g
The k-linear map determines rR .g/ because rA .u/ is the unique A-linear extension of to V A and rR .g/ is the unique R-linear map making the right hand square commute. Thus the map determines the natural transformation r. Moreover, the diagram can be used to extend any k-linear map W V ! V A to a natural transformation r of set-valued functors, namely, for g 2 G.R/ D Homk-alg .A; R/ and dene rR .g/ to be the linear map V .R/ ! V .R/ whose restriction to V is .V g/ . Thus, rR .g/.v c/ D .V g/.c .v//; for all g 2 G.R/, v 2 V , c 2 R: (69)
In this way, we get a one-to-one correspondence r $ between natural transformations of setvalued functors r and k-linear maps , and it remains to show that r is a representation of G if and only if is a comodule structure on V . Recall that the identity element 1G.k/ of G.k/ is A ! k. To say that rk .1G.k/ / D idV k means that the following diagram commutes,
v7!v1
V NNNv7!v1 / V A NNN NNN NNN & V A
+/
V k

V k
/ V k;
i.e., that the right hand diagram in (64) commutes. Next consider the condition that rR .g/rR .h/ D rR .gh/ for g; h 2 G.R/. By denition (see (34)), gh is the map A ! A A ! R;
.g;h/
7. Representations of afne groups and so rR .gh/ acts on V as V ! V A On the other hand, rR .g/rR .h/ acts as V ! V A i.e., as V ! V A
A V h V
81
! V AA
V .g;h/
! V R:
(70)
! V R
! V AR
V .g;id/
! V R;
! V AA
V .g;h/
! V R:
(71)
2
The maps (70) and (71) agree for all g; h if and only if the rst diagram in (64) commutes.
E XAMPLE 7.13 Recall (7.5) that, for any k-bialgebra A, the map W A ! A A is a comodule structure on A. When A D O.G/, this comodule structure on A corresponds to the regular representation of G on O.G/ (7.3). E XAMPLE 7.14 Let W V ! V O.Ga / be a nite-dimensional O.Ga /-comodule. The k-vector space O.Ga / ' kX has basis 1; X; X 2 ; : : : and so we can write X i .v/ D i .v/ X , v 2 V:
i 0
As is k-linear, so also is each map v 7! i .v/, and as the sum is nite, for each v; i .v/ is zero except for a nite number of i . As V is nite-dimensional, this implies that only a nite number of the i are nonzero. It follows that the representations constructed in (7.1) form a complete set. P ROPOSITION 7.15 Let rW G ! EndV be the representation corresponding to a comodule .V; /. A subspace W of V is a subrepresentation if and only if it is a subcomodule. P ROOF. Routine checking.
2
P ROPOSITION 7.16 Every representation of G is a union of its nite-dimensional subrepresentations. P ROOF. In view of (7.12) and (7.15), this is simply a restatement of Proposition 7.8.
2
P ROPOSITION 7.17 Let rW G ! EndV be the representation corresponding to a comodule .V; /. Choose a basis .ei /i 2I for V , and write X .ej / D ei aij ; aij 2 O.G/: (72)
i
Then, for each g 2 G.R/, rR .g/.ej 1/ D P

i 2I ei
g.aij / D
i 2I ei
aijR .g/
(73)
(equality in V .R/; recall that aijR is a map G.R/ ! R and that rR .g/ is a map V .R/ ! V .R/).
82 P ROOF. According to (69),
rR .g/.ej 1/ D .idV g/. .ej // P D .idV g/. i ei aij / P D i ei g.aij / P D i ei aijR .g/: In the last step, we used that g.f / D fR .g/ for f 2 O.G/ and g 2 G.R/ (see 2.11).
2
The coordinate ring of the monoid Mn is O.Mn / D kX11 ; X12 ; : : : ; Xnn , and so the proposition says that the homomorphism O.Mn / ! O.G/ dened by r sends Xij to aij . C OROLLARY 7.18 Let rW G ! EndV be the representation corresponding to a comodule .V; /. Let H be a subgroup of G, and let O.H / D O.G/=a. The following conditions on a vector v 2 V are equivalent: (a) for all k-algebras R and all g 2 H.R/, rR .g/.vR / D vR I (b) .v/ v 1 mod V a. P ROOF. We may suppose that v 0, and so is part of a basis for V , say v D ej (as in the statement of the proposition). Conditions (a) and (b) each hold for ej if and only aij ij 2 a for all i . 2 We say that v is xed by H if it satises the equivalent conditions of the corollary, and we let V H denote the subspace of xed vectors in V . If H.k/ is dense in H , then v 2 V H if and only if r.g/v D v for all g 2 H.k/ (because there is a largest subgroup of G xing v). L EMMA 7.19 Let G, r, V , , and H be as in the corollary, and let R be a k-algebra. The following submodules of V .R/ are equal: (a) V H R; (b) fv 2 V .R/ j rR0 .g/.vR0 / D vR0 for all R-algebras R0 and g 2 H.R0 /gI (c) fv 2 V .R/ j .v/ v 1 mod V a Rg. P ROOF. Nothing in this section requires that k be a eld (provided one assumes V to be free). Therefore the equality of the sets in (b) and (c) follows by taking k D R in Corollary 7.18. The condition .v/ v 1 mod V a is linear in v, and so if W is the solution space over k, then W k R is the solution space over R. This proves the equality of the sets in (a) and (c). 2
7f
The category of representations of G
Let G be an afne monoid, and let Repk .G/ be the category of representations of G on nitedimensional vector spaces. As this is essentially the same as the category of nite-dimensional O.G/-comodules (see 7.12), it is an abelian category and the forgetful functor to k-vector spaces is exact and faithful. The tensor product of two representations .V; r/ and .V 0 ; r 0 / is dened to be .V V; r r 0 / 0 where .r r 0 /R .g/ D rR .g/ rR .g/.
83
When G is a group, a representation .V; r/ has contragredient (or dual) is dened to be .V _ ; r _ / where, _ rR .g/.f / .v/ D f .rR .g 1 /v/; g 2 G.R/; f 2 V _ .R/; v 2 V .R/ (more succinctly, .gf /.v/ D f .g
1 v/).
P ROPOSITION 7.20 Let .V; r/ and .V 0 ; r 0 / be representations of G, and let and 0 be the corresponding comodule structures on V and V 0 . The comodule structures on V V 0 and V _ dened by r r 0 and r _ are those described in 7d. P ROOF. Easy exercise for the reader.
2
7g
Afne groups are inverse limits of algebraic groups
It is convenient at this point to prove the following theorem. T HEOREM 7.21 Every afne monoid (resp. group) is an inverse limit of its algebraic quotients. In particular, every afne monoid (resp. group) is an inverse limit of algebraic monoids (resp. groups) in which the transition maps are quotient maps. We prove Theorem 7.21 in the following equivalent form. T HEOREM 7.22 Every bialgebra (resp. Hopf algebra) over k is a directed union of its nitely generated sub-bialgebras (resp. Hopf subalgebras) over k. P ROOF. Let A be a k-bialgebra. By (7.8), every nite subset of A is contained in a nite-dimensional P k-subspace V such that .V / V A. Let .ei / be a basis for V , and write .ej / D i ei aij . P Then .aij / D k ai k akj (see (66), p.77), and the subspace L of A spanned by the ei and aij 0, satises .L/ L L. The k-subalgebra A0 generated by L satises .A0 / A0 AS and so it is a nitely generated sub-bialgebra of A. It follows that A is the directed union A D A0 of its nitely generated sub-bialgebras. P P Suppose that A has an inversion S . If .a/ D bi ci , then .Sa/ D Sci S bi (Exercise 4-5b). Therefore, the k-subalgebra A0 generated by L and SL satises S.A0 / A0 , and so it is a nitely generated Hopf subalgebra of A. It follows that A is the directed union of its nitely generated Hopf subalgebras. 2 C OROLLARY 7.23 Let A be a Hopf subalgebra of the Hopf algebra B. Then A and B are directed unions of nitely generated Hopf subalgebras Ai and Bi such that Ai Bi . C OROLLARY 7.24 Let A be a Hopf algebra over a eld k. If A is an integral domain and its eld of fractions is nitely generated (as a eld) over k, then A is nitely generated. P ROOF. Any nite subset S of A is contained in a nitely generated Hopf subalgebra A0 of A. When S is chosen to generate the eld of fractions of A, then A0 and A have the same eld of fractions, and so they are equal (5.33). 2 C OROLLARY 7.25 A Hopf algebra whose augmentation ideal is nitely generated is itself nitely generated.
84
P ROOF. Let A be a Hopf algebra. If IA is nitely generated, then there exists a nitely generated Hopf subalgebra A0 of A containing a set of generators for IA . The inclusion A0 ! A corresponds to a quotient map G ! G 0 whose kernel has Hopf algebra A A0 A0 =IA0 ' A=IA0 A D A=IA ' k. Proposition 6.1 shows that G ' G 0 , and so A0 ' A. 2 P ROPOSITION 7.26 Every quotient of an algebraic group is itself an algebraic group. P ROOF. We have to show that a Hopf subalgebra A of a nitely generated Hopf algebra B is nitely generated. Because B is noetherian, the ideal IA B is nitely generated, and because B is at over A, the map IA A B ! A A B ' B is an isomorphism of IA A B onto IA B. Therefore IA B is a nitely generated as a B-module, and as B is faithfully at over A, this implies that IA is nitely generated.16 2
A SIDE 7.27 There are many subalgebras of kX1 ; : : : ; Xn that are not nitely generated; for example, the subring kX; X Y; X Y 2 ; : : : of kX; Y is not even noetherian. Hence (7.26) is nontrivial,
7h
Algebraic groups admit nite-dimensional faithful representations
In fact, every sufciently large nite-dimensional subrepresentation of the regular representation will be faithful. T HEOREM 7.28 Let G be an algebraic group. For some n, there exists an injective homomorphism G ! GLn . P ROOF. Let A D O.G/, and let V be a nite-dimensional subcomodule of A containing a set of P generators for A as a k-algebra. Let .ei /1i n be a basis for V , and write .ej / D i ei aij . According to (7.17), the image of O.GLV / ! A contains the aij . But, because W A ! k is a co-identity (see (26), p.28), X ej D . idA / .ej / D .ei /aij ;
i
and so the image contains V ; it therefore equals A. We have shown that O.GLV / ! A is surjective, which means that G ! GLV is injective. 2 C OROLLARY 7.29 Every afne group admits a faithful family of nite-dimensional representations. P ROOF. Write G as an inverse limit G D lim Gi of algebraic groups, and, for each i 2 I , choose i 2I a faithful nite-dimensional representation .Vi ; ri / of Gi . Each .Vi ; ri / can be regarded as a representation of G, and the family is faithful. 2 P ROPOSITION 7.30 Let .V; r/ be a faithful representation of an algebraic group G. Then V is a union of nite-dimensional faithful representations.
16 As
a B-module, IA A B has a nite set of generators fc1 b1 ; : : : ; cm bm g, and the map .a1 ; : : : ; am / 7! ai ci W Am ! IA
is surjective because it becomes surjective when tensored with B.
85
P P ROOF. Let .ei /i 2I be a basis for V , and write .ej / D i 2I ei aij , aij 2 A. Because .V; r/ is faithful, the k-algebra A is generated by the aij (7.17 et seq.). Because A is nitely generated as a k-algebra, only nitely many aij s are need to generate it, and so there exists a nitely subset J of I such that the aij s appearing in .ej / for some j 2 J generate A. Every nite-dimensional subrepresentation of .V; r/ containing fej j j 2 J g is faithful. 2 The theorem says that every algebraic group can be realized as an algebraic subgroup of GLn for some n. This does not mean that we should consider only subgroups of GLn because realizing an algebraic group in this way involves many choices. E XAMPLE 7.31 Let G be the functor sending a k-algebra R to R R R with the (noncommutative) group structure .x; y; z/ .x 0 ; y 0 ; z 0 / D .x C x 0 ; y C y 0 ; z C z 0 C xy 0 /: This is an algebraic group because it is representable by kX; Y; Z. The map 0 1 1 x z .x; y; z/ 7! @0 1 y A 0 0 1 is an embedding of G into GL3 . Note that the functor R R R R also has an obvious commutative group structure (componentwise addition), which shows that the k-algebra kX; Y; Z has more than one Hopf algebra structure.
A SIDE 7.32 Does Theorem 7.28 still hold when k is allowed to be a commutative ring? Apparently, this is not known even when k is the ring of dual numbers over a eld and G is smooth (mo22078, Brian Conrad). Using (7.10), one sees by the above arguments that an afne group scheme G of nite type over a noetherian ring k has a faithful representation on a nitely generated submodule M of the regular representation. If M is at over k, then it is projective, and hence a direct summand of a free nitely generated k-module L, and so G ,! GLrank.L/ . When k is a Dedekind domain and G is at, the module M is torsion-free, and hence automatically at. Thus, every at group scheme of nite type over a Dedekind domain admits an embedding into GLn for some n. As every split reductive group scheme over a ring k arises by base change from a similar group over Z (Chevalley), such group schemes admit embeddings into GLn . Since every reductive group splits over a nite etale extension of the base ring (SGA3), an argument using restriction of scalars proves the statement for any reductive group (mo22078).
7i
The regular representation contains all
Let .V; rV / be a representation of G. For v 2 V .R/ and u 2 V _ .R/, let hu; vi D u.v/ 2 R. For a xed v 2 V and u 2 V _ , the maps x 7! hu; rV .x/viW G.R/ ! R are natural in R, and so dene an element of O.G/, i.e., there exists a
u .v/R .x/ u .v/
2 O.G/ such that
D hu; rV .x/vi (in R) for all x 2 G.R/:
Let A D O.G/, and let rA be the regular representation of G on A. P ROPOSITION 7.33 The map
u
is a homomorphism of representations .V; rV / ! .A; rA /.
86 P ROOF. We have to show that .

u /R rV .g/
D rA .g/ .
u /R
for all g 2 G.R/ and all R:
For any v 2 V .R/ and x 2 G.R/, .LHS.v// .x/ D

u .rV .g/v/R .x/
D hu; rV .x/rV .g/vi D hu; rV .xg/vi D

u .v/R .xg/ u .v//R .x/
D .rA .g/ as required.
D .RHS.v// .x/,
2
P ROPOSITION 7.34 If u1 ; : : : ; un span V _ , then the k-linear map v 7! . is injective. P ROOF. Note that
u .v/.1/ u1 .v/; : : : ; un .v//W V
! An
(74)
D hu; vi, and so the composite V .R/ ! An .R/ ! Rn
of (74) with the map evaluate at 1 is v 7! .hu1 ; vi; : : : ; hun ; vi/; which is injective by our choice of the ui s.
2
Thus, V embeds into a nite sum of copies of the regular representation. We give a second proof of this. P ROPOSITION 7.35 Let V be a nite-dimensional representation of G, and let V0 denote the underlying vector space with the trivial representation. Let be the coaction of O.G/ on V . Then W V ! V0 O.G/ is an injective homomorphism of representations. Here G acts on V0 O.G/ through the second factor, and so V0 O.G/ is a direct sum of copies of O.G/. P ROOF. Let A D O.G/. The k-vector space V0 A becomes a comodule (isomorphic to a direct sum of copies of A) with the coaction17 V W V A ! V A k A: The commutative diagram (see (64), p.76) V ? ? y V A
17 This
V0 A ? ? yV0
An ? ? y
! V0 A A
! .A A/n
says exactly that the map W V ! V0 A is homomorphism of comodules. It is obviously injective.2

is the cofree comodule on V0 see (7.5a).
87
7j
Every faithful representation contains all
Recall that, for a comodule .V; / over a coalgebra (or bialgebra) C , we dened CV to be the smallest subspace of C such that .V / V CV , and we showed that it is a sub-coalgebra of C (see 7.6). Any CV -comodule .W; W / can be regarded as a C -comodule with the coaction W ! W CV
W
W ,!
! W C:
L EMMA 7.36 Let V be a nite-dimensional C -comodule. Every nite-dimensional CV -comodule (considered as a C -comodule) is isomorphic to a quotient of a sub-comodule of V n for some n. P ROOF. We may replace C with CV . By duality (7.6), it sufces to prove the following statement: let A be a nite k-algebra and let V be a nite-dimensional faithful left A-module; then every nitedimensional A-module W is isomorphic to a quotient of a submodule of V n for some n. But W is isomorphic to a quotient of the free module Am for some m, and so it sufces to prove that A is isomorphic to a submodule of V n for some n. If e1 ; : : : ; en span V as a k-vector space, then a 7! .ae1 ; : : : ; aen /W A ! V n is injective because V is faithful. 2 L EMMA 7.37 Let A be a bialgebra, and let V and V 0 be nite-dimensional A-comodules. Then AV V 0 D AV AV 0 .
0 P ROOF. Choose k-bases .ei /i 2I and .ei /i 2I 0 for V and V 0 , and write X X 0 0 0 ei aij ; ei aij : V .ej / D V 0 .ej / D i 2I i 2I 0
Then .ei ei 0 /.i;i 0 /2I I 0 is a basis for V k V 0 , and P 0 V V 0 .ej ej 0 / D i;i 0 .ei ei 0 / .aij ai 0 j 0 / (see 7d). As AV D haij j i; j 2 I i AV 0 D haij j i; j 2 I 0 i
0 AV V 0 D haij ai 0 j 0 j i; j 2 I;
i 0 ; j 0 2 I 0 i;
the statement is clear. (Alternatively, note that AV AV 0 is the sub-coalgebra attached to the AAcomodule V V 0 , and that AV V 0 is the image of this by the multiplication map mW A A ! A.)2 L EMMA 7.38 Let C be a Hopf algebra (with inversion S). For any nite-dimensional C -comodule V , CV _ D S CV . P ROOF. This follows easily from the denition of the contragredient V _ (see 7d). P ROPOSITION 7.39 Let A be a bialgebra, and let V be a nite-dimensional A-comodule. Let X A.V / D AV n A:
n 0 2
Then A.V / is a sub-bialgebra of A containing AV and 1, and it is the smallest such sub-bialgebra; it is nitely generated.
88 P ROOF. It follows from the lemma that AV n D AV
AV
(n factors),
2
and so it is clear that A.V / is the subalgebra of A generated by AV and 1. S S Note that A D V A.V / because A D V AV (see 7.9). Now assume A has admits an inversion S.
P ROPOSITION 7.40 The algebra A.V V _ / is a sub-bialgebra of A containing AV and 1 and stable under S, and it is the smallest such sub-bialgebra; it is nitely generated. P ROOF. This follows easily from (7.38) and (7.40). D EFINITION 7.41 An A-comodule V is said to be faithful if A.V V _ / D A. P ROPOSITION 7.42 A representation rW G ! GLV is faithful if and only if V is a faithful O.G/comodule. P ROOF. Let G 0 be the image of G in GLV . Then O.G 0 / is a sub-bialgebra of O.G/ containing O.G/V and 1 and stable under S ; therefore if V is a faithful comodule, then Proposition 7.40 shows that it O.G 0 / D O.G/, and so r is a faithful representation. The converse is similar. 2 P ROPOSITION 7.43 Let G ! GLV be a representation of G. If V is faithful, then every nitedimensional representation of G is a quotient of a sub-representation of a direct sum of representaN N N tions n .V V _ / (and hence also of a direct sum of representations n V m det.V / 1 ). P ROOF. This follows from Lemma 7.36.
2 2
When M is an afne monoid with coordinate ring O.M / D A, we let MV denote the quotient afne monoid of M with coordinate ring A.V /. Similarly, when G is an afne group, we let GV denote the quotient afne group of G with coordinate ring A.V V _ /. Both MV and GV act faithfully on V . Moreover, M D lim MV ; G D lim GV S because A D A.V /. It may happen that GV is closed in EndV (and not only in GLV ), i.e., that C.V / D C.V V _ /. This is the case if GV SLV then det.V / is trivial.
7k
Stabilizers of subspaces
P ROPOSITION 7.44 Let G ! GLV be a representation of G, and let W be a subspace of V . The functor def R GW .R/ D fg 2 G.R/ j g.W R/ D W Rg is an afne subgroup of G.
7. Representations of afne groups P ROOF. Let .ei /i 2J be a basis for W , and extend it to a basis .ei /J tI for V . Write .ej / D P
i 2J tI ei
89
aij ;
aij 2 O.G/:
Let g 2 G.R/ D Homk-alg .O.G/; R/. Then (see 7.17) gej D P

i 2J tI ei
g.aij /:
Thus, g.W R/ W R if and only if g.aij / D 0 for j 2 J , i 2 I . Hence GW is represented by the quotient of O.G/ by the ideal generated by faij j j 2 J; i 2 I g (recall that g.aij / D .aij /R .g/ (7)). 2 The afne group GW is called the stabilizer of W in G. P ROPOSITION 7.45 Let G ! GLV be a representation of G, and let v 2 V . The functor R is an afne subgroup of G. P ROOF. If v D 0, then Gv D G and there is nothing to prove. Otherwise, choose a basis .ei /i 2I for V with ei0 D v for some i0 2 I . Write .ej / D P
i 2J tI ei
Gv .R/ D fg 2 G.R/ j g.v 1/ D v 1g
def
aij ;
aij 2 O.G/:
An element g 2 G.R/ xes v 1 if and only if g.ai i0 / D 1 if i D i0 0 otherwise. i i0 j i 2 I g.2
Therefore Gv is represented by the quotient of O.G/ by the ideal generated by fai i0 The afne group Gv is called the isotropy or stablity group of v in G.
T HEOREM 7.46 (C HEVALLEY ) Every algebraic subgroup of an algebraic group G is the stabilizer of a subspace in some nite-dimensional representation of G; the subspace can even be taken to be one-dimensional and the representation faithful. P ROOF. Let H be an algebraic subgroup of G, and let a be the kernel of O.G/ ! O.H /. Then H is the stabilizer of a in the regular representation of G on O.G/. According to (7.16), there exists a nite-dimensional k-subspace V of O.G/ that is G-stable and contains a generating set for a as an ideal and a generating set for O.G/ as a k-algebra. Then H is the stabilizer of the subspace def W D a \ V in V . That W can be taken to be one-dimensional follows from the next lemma. 2 L EMMA 7.47 Let V be a G-module, and let W be a G-submodule of V of dimension d . Then the def V stabilizer of W in G is equal to the stabilizer of D D d W in G.
90
V P ROOF. Let g 2 GL.V /. If gW D W , then certainly . d g/.D/ D. For the converse, choose a basis e1 ; : : : ; en for V such that e1 ; : : : ; em is a basis for W \ gW , e1 ; : : : ; em ; : : : ed is a basis for W , and e1 ; : : : ; em ; ed C1 ; : : : ; e2d m is a basis for gW . Then V . d g/.e1 ^ ^ em ^ ^ ed / D c e1 ^ ^ em ^ ed C1 ^ ^ ed
2
V for some c 2 k and so, if . d g/.D/ D D, then m D d and gW D W .
C OROLLARY 7.48 A subgroup H of an algebraic group G is the subgroup of G xing a vector in some faithful nite-dimensional representation of G in each of the following two cases: (a) all the representations of H are semisimple; (b) every character of H dened over k extends to a similar character of G. P ROOF. According to Chevalleys theorem, H is the stabilizer of a line D in a nite-dimensional representation V of G. Let D _ be the dual of D with H acting contragrediently. If we can nd a representation V 0 of G containing D _ as an H -stable subspace, then H will be the subgroup of G xing any nonzero vector in D D _ V V 0 . Certainly D _ occurs as a quotient of V _ , and so, in case (a), it also occurs as a direct summand of V _ (regarded as an H -module). In this case, we can take V 0 D V _ . The action of H on D denes a character of H , which in case (b) extends to a character of G. In this case, we can take V 0 D D _ . 2
7l
Sub-coalgebras and subcategories
Let C be a coalgebra over k, and let Comod.C / denote the category of nite-dimensional right C -comodules. Let D be a sub-coalgebra of C . As we noted earlier, if .V; / is a comodule over D, then V becomes C -comodule with the coaction V ! V D ,! V C: We get in this way an exact fully faithful functor Comod.D/ ! Comod.C /. We let D _ denote the full subcategory of Comod.C / whose objects are isomorphic to a comodule in the image of this functor. D EFINITION 7.49 A full subcategory of an abelian category is replete if it is closed under the formation of nite direct sums, subobjects, and quotient objects. In particular, every object isomorphic to an object in a replete subcategory also lies in the subcategory. A replete subcategory is an abelian category, and the inclusion functor is exact. T HEOREM 7.50 The map D 7! D _ is a bijection from the set of sub-coalgebras of C onto the set of replete subcategories. P ROOF. It is obvious that D _ is replete. Let S be a replete subcategory of Comod.C /, and let X C.S/ D CV (sub-coalgebra of C ).
V 2S
To prove the theorem, we have to show that:
7. Representations of afne groups C.D _ / D D for all sub-coalgebras D of C , and C.S/_ D S for all replete subcategories S of Comod.C /.
91
The rst statement follows from the fact that D is a union of its subcoalgebras DV (see 7.9), and the second follows from (7.36).
P ROPOSITION 7.51 Let A be a bialgebra over k. (a) A sub-coalgebra D of A is a sub-bialgebra of A with 1 if and only if D _ is stable under tensor products and contains the trivial comodule k. (b) Assume A has an inversion S. A sub-bialgebra D is stable under S if and only if D _ is stable under the contragredient functor. P ROOF. (a) If D is a subalgebra of A, S certainly D _ is stable under tensor products and contains then k. For the converse, recall that D D DV and that DV DV 0 D DV V 0 (see 7.37), and so D is closed under products. Because D _ contains V0 D k, D contains DV0 D k. (b) Use the formula AV _ D S.AV /. 2
7m
Quotient groups and subcategories
For an afne group G over k, Rep.G/ denotes the category of nite-dimensional G-modules. Let r G ! Q be a quotient of G. A representation rW Q ! GLV denes a representation G ! Q ! GLV of G. We get in this way an exact fully faithful functor Rep.Q/ ! Rep.G/. The essential image of the functor consists of the representations of G containing Ker.G ! Q/ in their kernel. We let Q_ denote this subcategory of Rep.G/. T HEOREM 7.52 The map Q 7! Q_ is a bijection from the set of isomorphism classes of quotients of G to the set of replete subcategories of Rep.G/ closed under the formation of tensor products (including the empty tensor product) and under passage to the contragredient. P ROOF. Obvious from (7.50), (7.51), and the dictionary between Hopf algebras and their comodules and afne groups and their representations. 2
7n
Normal subgroups and subcategories
L EMMA 7.53 Let and be endomorphisms of vector spaces V and W respectively, and let v and w be nonzero vectors in V and W . If v w is xed by , then there exists a c 2 k such that .v/ D cv and .v/ D c 1 v. P ROOF. Let v and w be nonzero elements of vector spaces V and W , and write V D hvi V 0 ; Then V W D hv wi hvi W 0 V 0 hwi V 0 W 0 ; where hv wi D hvi hwi. In particular, we see that v w 0 (because hvi hwi 0). Now write v D av C v 0 ; w D bw C w 0 ; v 0 2 V 0 ; w 0 2 W 0 . W D hwi W 0 :
92 Then
. /.v w/ D ab.v w/ C av w 0 C v 0 bw C v 0 w 0 : If . /.v w/ D ab.v w/, then ab D 1 and av w 0 D 0 D v 0 bw: As av 0 bw, (75) implies that w 0 D 0 D v 0 , as required. (75)
2
L EMMA 7.54 Let N be a normal afne subgroup of an afne group G. Let .V; r/ be a representation of G, and let V N be the subspace on which N acts trivially. Then V N is stable under G (i.e., the stabilizer of V N is G). P ROOF. Let w 2 V N , and let g 2 G.k/. For any n 2 N.k/, r.n/.r.g/w/ D r.ng/w D r.gn0 /w D r.g/r.n0 /w D r.g/w: When N.k/ is Zariski dense in N , this implies that r.g/v 2 V N , which implies that G stabilizes V when G.k/ is Zariski dense in G. In the general case, we let w 2 V N .R/ and g 2 G.R/ for some k-algebra R. The same argument shows that, for any R-algebra R0 and n 2 N.R0 /, rR0 .n/.rR .g/w/R0 D .rR .g/w/R0 ; and so (7.19) shows that rR .g/w 2 V N .R/, as required.
2
By a character of an afne group G, we mean a homomorphism G ! Gm . A one-dimensional representation of G denes a character of G. A one-dimensional representation is obviously simple. A standard argument (GT 7.11) shows that if V is a sum of simple representations, then it is a direct sum of such representations. P ROPOSITION 7.55 For any distinct normal afne subgroups N N 0 of an afne group G, there exists a representation of G on which N acts trivially but N 0 acts nontrivially. P ROOF. According to Chevalleys theorem (7.46), there exists a representation .V; r/ of G and a line L in V such that N is the stabilizer of L. Let U be the smallest subspace of V stable under G. I claim that N 0 acts nontrivially on W D .V U _ /N (which Lemma 7.54 shows to be a representation of G). It sufces to prove this after an extension of the base eld k, and so we may assume that k is algebraically closed. Suppose rst that G is smooth. Let N act on the line L D hvi through the character 0 . For any g 2 G.k/, N acts on the line gL through the conjugage character g 0 . As U is spanned by the L lines gL, we see that U decomposes into a direct sum U D 2 U in which U is a subspace of U on which N acts through the character . Correspondingly, U _ decomposes into a direct sum L _ 1 _ _ U_ D 2 U 1 . In particular, U contains a line L D hui on which N acts through 0 . Now v u is xed by N , but it is not xed by N 0 (since the line L is not xed by N 0 and we can apply Lemma 7.53). If G is algebraic but not smooth, then the characteristic of k is p 0, and there exists an n n n n such that O.G/p D O.G 0 / is reduced (see 5.27). Replace V by V p and v by v p . Lemma n def 7.53 shows that the stabilizer of L0 D hv p i is still N . The action of N on L0 factors through its
93
quotient in G 0 (cf. Waterhouse 1979, 16.3, p123). The previous argument shows that U _ contains n a line L0_ D hu0 i such that v p u0 is xed by N but not by N 0 . When G is not algebraic, there will be an algebraic quotient in which the images of N and N 0 are distinct, and to which we can apply the previous argument. 2 L EMMA 7.56 Let N1 and N2 be normal afne subgroups of G. If Rep.G/N1 D Rep.G/N2 then N1 D N2 . P ROOF. If N1 N2 , then Lemma 7.55 shows that there exists a representation .V; r/ of G and a v 2 V xed by N1 but not by N1 N2 . Then V N1 is an object of Rep.G/N1 but not of Rep.G/N2 , which contradicts the hypothesis. 2 T HEOREM 7.57 Let N be a normal afne subgroup of an afne group G, and let Q be a quotient of G. Then N D Ker.G ! Q/ if and only if Rep.G/N D Q_ . P ROOF. ): According to Theorem 6.46, a representation rW G ! GLV factors through Q (and so lies in Q_ ) if and only if r maps N to 1 (and so lies in Rep.G/N ). 0 0 (: Let N 0 be the kernel of G ! Q. Then Rep.G/N D Q_ , and so Rep.G/N D Rep.G/N . This implies that N D N 0 . 2 C OROLLARY 7.58 The map N 7! Rep.G/N is a bijection from the set of normal afne subgroups of G to the set of replete subcategories of Rep.G/ closed under tensor products and passage to the contragredient. P ROOF. Let S be a replete subcategory of Rep.G/ closed under tensor products and passage to the contragredient. The S D Q_ for some quotient Q of G, well-dened up to isomorphism, and the kernel N of G ! Q is a normal subgroup of G. The maps S 7! N and N 7! Rep.G/N are inverse. 2 T HEOREM 7.59 For any normal afne subgroup N of an afne group G, there exists a quotient map with kernel N . P ROOF. The subcategory Rep.G/N of Rep.G/ is replete and closed under tensor products and passage to the contragredient. Therefore Rep.G/N D Q_ for some quotient Q of G, and the theorem implies that N is the kernel of G ! Q. 2
A SIDE 7.60 Perhaps prove: let H be a subgroup of G, and let Q be the quotient of G such that Q_ D Rep.G/H ; the kernel of G ! Q is the normalizer of H . This implies the existence of normalizers (and perhaps centralizers). A SIDE 7.61 Add a discussion of the correspondence between normal subgroups of an afne group G and normal Hopf algebras (Abe 1980, p179), and also of the correspondence between normal Hopf ideals and Hopf subalgebras (ibid. 4.4.7, p207, in the case that k is algebraically closed and the Hopf algebras are assumed to be reduced). A SIDE 7.62 This subsection needs to be rewritten. In particular, the proof of (7.55) needs to be rethought. Cf. the sketch of proof of (6.52). A SIDE 7.63 For a representation G ! GLV , add a denition of V G . Show that, for any eld k 0 G 0 .V G /k 0 ' Vk 0 k . k,
94
7o
Exercises
EXERCISE 7-1 Let .V; / be a comodule over a bialgebra A. Show that, for a suitably dene determinant element V of A, A.V V _ / D A.V / 1 (cf. Serre 1993, 3.5, p.61). V
Group theory: the isomorphism theorems
In this section, we show that the isomorphism theorems in abstract group theory extend to afne groups.
8a
Review of abstract group theory
For a group G (in the usual sense), we have the notions of subgroup, a normal subgroup, an embedding (injective homomorphism), and of a quotient map (surjective homomorphism). Moreover, there are the following basic results, which are often referred to collectively as the isomorphisms theorems.18 8.1 (Existence of quotients). The kernel of a quotient map G ! Q is a normal subgroup of G, and every normal subgroup N of G arises as the kernel of a quotient map G ! G=N . 8.2 (Homomorphism theorem). The image of a homomorphism W G ! G 0 is a subgroup G of G 0 , and denes an isomorphism from G= Ker./ onto G; in particular, every homomorphism is the composite of a quotient map and an embedding. 8.3 (Isomorphism theorem). Let H and N be subgroups of G such that H normalizes N ; then HN is a subgroup of G, N is a normal subgroup of HN , H \ N is a normal subgroup of H , and the map h.H \ N / 7! hN W H=H \ N ! HN=N is an isomorphism. 8.4 (Correspondence theorem). Let N be a normal subgroup of G. The map H 7! H=N denes a one-to-one correspondence between the set of subgroups of G containing N and the set of subgroups of G=N . A subgroup H of G containing N is normal if and only if H=N is normal in G=N , in which case the map G=H ! .G=N /=.H=N / dened by the quotient map G ! G=N is an isomorphism. In this section, we shall see that, appropriately interpreted, all these notions and statements extend to afne groups (in particular, to algebraic groups).
8b
The existence of quotients
T HEOREM 8.5 For every normal subgroup N of an afne group G, there exists a quotient map with kernel N . P ROOF. This is a restatement of Theorem 7.59. Thus the normal subgroups of an afne group are exactly the kernels of homomorphisms.
(8.2), (8.3), and (8.4) are sometimes called the rst, second, and third isomorphism theorems, but the numbering varies. In Noether 1927, the rst isomorphism theorem is (8.4) and the second is (8.3).
18 Statements
95
96
E XAMPLE 8.6 Let PGLn be the quotient of GLn by its centre, and let PSLn be the quotient of SLn by its centre: PGLn D GLn =Gm ; PSLn D SLn = n : The homomorphism SLn ! GLn ! PGLn contains phism
n
in its kernel, and so denes a homomor-
PSLn ! PGLn : Is this an isomorphism? Note that SLn .k/=

n .k/
(76)
! GLn .k/=Gm .k/
(77)
is injective, but not in general surjective: not every invertible n n matrix can be written as the product of a matrix with determinant 1 and a scalar matrix (such a matrix has determinant in k n ). Nevertheless, I claim that (76) is an isomorphism of algebraic groups. In characteristic zero, this follows from the fact that (77) is an isomorphism when k D k al (apply 6.16 and 6.44). In the general case, we have to check the conditions (6.2a) and (6.40). Let q 1 2 PSLn .R/. For some faithfully at R-algebra R0 , there exists a g 2 SLn .R0 / mapping to q in PSLn .R0 /. The image of g in GLn .R0 / is not in Gm .R0 / (because q 1/; therefore, the image of g in PGLn .R0 / is 1, which implies that the image of q in PGL.R/ is 1: PSLn .R0 / x ? ? PSLn .R/ ! PGLn .R0 / x ?injective ? ! PGLn .R/:
We have checked condition (6.2a). Let q 2 PGLn .R/. For some faithfully at R-algebra R0 , there exists a g 2 GLn .R0 / mapping def to q. If a D det.g/ is an nth power, say a D t n , then g D g0 t with det.g0 / D 1, and the image of g in GLn .R0 /=Gm .R0 / is in the image of SLn .R0 /= n .R0 /. Hence, the image of q in PGLn .R0 / is in the image of PSLn .R0 /. If a is not an nth power in R0 , we replace R0 by the faithfully at (even free) algebra R0 T =.T n a/ in which it does become an nth power. We have checked condition (6.40).
8c
The homomorphism theorem
A homomorphism W G ! G 0 of afne groups denes a homomorphism W O.G 0 / ! O.G/ of Hopf algebras, whose kernel a is a Hopf ideal in O.G 0 /.19 Thus a D ff 2 O.G 0 / j fR .R .P // D 0 for all k-algebras R and all P 2 G.R/g: The afne subgroup H of G 0 corresponding to a (see 6.7) is called the image of (and often denoted G). Thus H.R/ D fg 2 G.R/ j fR .g/ D 0 for f 2 ag.
19 In
fact, we dont need to use that a is a Hopf ideal, just that it is an ideal.
8. Group theory: the isomorphism theorems
97
T HEOREM 8.7 For any homomorphism W G ! G 0 of afne groups, the kernel N of is a normal subgroup of G, the image G of is a subgroup of G 0 , and factors in a natural way into the composite of a surjection, an isomorphism, and an injection: G ? ? surjectivey G=N
! G0 x ?injective ? ! G:
isomorphism
If G is an algebraic group, then so also are G=N and G. P ROOF. The factorization O.G/ of denes a factorization G ! G ! G 0 of into a surjection followed by an injection. As G ! G=N and G ! G are both quotient maps with kernel N , there is a unique isomorphism G=N ! G such that the composite G ! G=N ! G is G ! G (apply 6.47). The nal statement follows from (7.26). C OROLLARY 8.8 For any k-algebra R, [ .G/.R/ D G.R/ \ Im .R0 / 0
R 2
O.G 0 /=a
O.G 0 /
(R0 runs over the R-algebras).
Therefore G represents the sheaf associated with R Im..R//: H.R/
Moreover, G is the intersection of the subgroups H of G 0 with the property that Im .R/ for all k-algebras R.
P ROOF. The map G ! G is a quotient map, and so the rst statement follows from (6.60). If H is an afne subgroup of G 0 such that H.R/ Im .R/ for all k-algebras R, then, for any xed k-algebra R, [ H.R/ G.R/ \ Im .R0 / D .G/.R/; 0
R
and so H
G.
C OROLLARY 8.9 A homomorphism W G ! G 0 of algebraic groups is surjective if, for some eld k 0 containing k, the image of G.k 0 / in G 0 .k/ is dense in G 0 . P ROOF. As .G.k 0 // .G/.k 0 / G.k 0 /, the condition implies that G D G.
2
Let W G ! G 0 be a homomorphism of algebraic groups. Then G.k al / ! .G/.k al / is surjective (see 6.44), and so .G/.k/ D G 0 .k/ \ .G/.k al / D G 0 .k/ \ Im.G.k al / ! G 0 .k al //:
.k al /
98
8d
The isomorphism theorem
Let H and N be algebraic subgroups of G such that H normalizes N . The natural action of H.R/ on N.R/ denes an action of H on N by group homomorphisms, and multiplication denes a homomorphism N H ! G. We dene NH D HN to be the image of this homomorphism. The following statements are obvious from 8c. 8.10 For any k-algebra R, .HN /.R/ consists of the elements of G.R/ that lie in H.R0 /N.R0 / for some nitely generated faithfully at R-algebra R0 . Therefore HN represents the sheaf associated with the functor R H.R/ N.R/ G.R/: Moreover, HN is the interesection of the subgroups G 0 of G such that, for all k-algebras R, G 0 .R/ contains both H.R/ and N.R/. 8.11 We have .HN /.k al / D H.k al / N.k al /; and so .HN /.k/ D G.k/ \ .H.k al / N.k al //:
8.12 It is not true that .HN /.R/ D H.R/N.R/ for all k-algebras R. For example, consider the algebraic subgroups SLn and Gm (nonzero scalar matrices) of GLn . Then GLn D SLn Gm , but a matrix A 2 GLn .R/ whose determinant is not an nth power is not the product of a scalar matrix with a matrix of determinant 1. T HEOREM 8.13 Let H and N be algebraic subgroups of the algebraic group G such that H normalizes N . The natural map H=H \ N ! HN=N is an isomorphism. P ROOF. We have an isomorphism of group-valued functors H.R/=.H \ N /.R/ ! H.R/N.R/=N.R/ .HN /.R/=N.R/:
2
(78)
The statement now follows from (6.60), or by passing to the associated sheaves. E XAMPLE 8.14 Let G D GLn , H D SLn , and N D Gm (scalar matrices in G). Then N \ H D (obviously), HN D GLn (by the arguments in 8.6), and (78) becomes the isomorphism SLn =
n
! GLn =Gm :
99
8e
The correspondence theorem
T HEOREM 8.15 (Third isomorphism theorem). Let N be a normal algebraic subgroup of G. The map H 7! H=N denes a one-to-one correspondence between the set of algebraic subgroups of G containing N and the set of algebraic subgroups of G=N . An algebraic subgroup H of G containing N is normal if and only if H=N is normal in G=N , in which case the map G=H ! .G=N /=.H=N / dened by the quotient map G ! G=N is an isomorphism. P ROOF. The rst statement follows from the fact that the analogous statement holds for Hopf algebras (cf. Exercise 4-10). For the second statement, note that the map G.R/=H.R/ ! .G.R/=N.R//=.H.R/=N.R// dened by the quotient map G.R/ ! G.R/=N.R/ is an isomorphism. This isomorphism is natural in R, and when we pass to the associated sheaves, we obtain the isomorphism (79). 2
A SIDE 8.16 Let qW G ! G=N be the quotient map. For any subgroup H of G, qH is a subgroup of G=N , which corresponds to HN . Deduce that if H 0 is normal in H , then H 0 N is normal in HN .
(79)
8f
The Schreier renement theorem
L EMMA 8.17 (B UTTERFLY L EMMA ) Let H1 N1 and H2 N2 be algebraic subgroups of an algebraic group G with N1 and N2 normal in H1 and H2 . Then N1 .H1 \ N2 / and N2 .N1 \ H2 / are normal algebraic subgroups of the algebraic groups N1 .H1 \ H2 / and N2 .H2 \ H1 / respectively, and there is a canonical isomorphism of algebraic groups N1 .H1 \ H2 / N2 .H1 \ H2 / ' N1 .H1 \ N2 / N2 .N1 \ H2 / P ROOF. The algebraic group H1 \ N2 is normal in H1 \ H2 and so N1 .H1 \ H2 / is normal in N1 .H1 \ N2 / (see Exercise 6-2). Similarly, N2 .H2 \ N1 / is normal in N2 .H2 \ H1 /. The subgroup H1 \ H2 of G normalizes N1 .H1 \ N2 /, and so the isomorphism Theorem 8.13 shows that .H1 \ H2 / N1 .H1 \ N2 / H1 \ H2 ' : .H1 \ H2 / \ N1 .H1 \ N2 / N1 .H1 \ N2 / As H1 \ N2 H1 \ H2 , we have that H1 \ H2 D .H1 \ H2 / .H1 \ N2 /, and so N1 .H1 \ H2 / D N1 .H1 \ H2 / .H1 \ N2 /. The rst of Dedekinds modular laws (Exercise 6-3a) with A D H1 \ N2 , B D H1 \ H2 , and C D N1 becomes .H1 \ H2 / \ N1 .H1 \ N2 / D .H1 \ N2 /.H1 \ H2 \ N1 / D .H1 \ N2 / .N1 \ H2 /. (80)
100 Therefore (80) is an isomorphism
H1 \ H2 N1 .H1 \ H2 / ' : .H1 \ N2 / .N1 \ H2 / N1 .H1 \ N2 / A symmetric argument shows that H1 \ H2 N2 .H1 \ H2 / ' ; .H1 \ N2 / .N1 \ H2 / N2 .H2 \ N1 / and so N1 .H1 \ H2 / N2 .H1 \ H2 / ' : N1 .H1 \ N2 / N2 .H2 \ N1 /
2
A subnormal series in an algebraic group G is a nite sequence of subgroups, beginning with G and ending with 1, such that each subgroup is normal in the preceding subgroup. Two subnormal sequences G D G0 G D H0 G1 H1 Gs D f1g Ht D f1g of f1; 2; : : : ; sg such that Gi =Gi C1
are said to be equivalent if s D t and there is a permutation H .i / =H .i /C1 .
T HEOREM 8.18 Any two subnormal sequences in an algebraic group have equivalent renements. P ROOF. Let Gij D Gi C1 .Hj \ Gi / and let Hj i D Hj C1 .Gi \ Hj /. According to the buttery lemma Gij =Gi;j C1 ' Hj i =Hj;i C1 , and so the renement .Gij / of .Gi / is equivalent to the renement .Hj i / of .Hi /.
2
A subnormal series is a composition series if no quotient group Gi has a proper nontrivial normal subgroup. T HEOREM 8.19 For any two composition series G D G0 G D H0 s D t and there is a permutation for each i. G1 H1 Gs D f1g Ht D f1g;
.i / =H .i /C1
of f1; 2; : : : ; sg such that Gi =Gi C1 is isomorphic to H
P ROOF. Use that, for each i , only one of the quotients Gi C1 .Hj \ Gi /=Gi C1 .Hj C1 \ Gi / is nontrivial 2
101
8g
The category of commutative algebraic groups
T HEOREM 8.20 The commutative algebraic groups over a eld form an abelian category. P ROOF. The Hom sets are commutative groups, and the composition of morphisms is bilinear. Moreover, the product G1 G2 of two commutative algebraic groups is both a product and a sum of G1 and G2 . Thus the category of commutative algebraic groups over a eld is additive. Every morphism in the category has both a kernel and cokernel (6.13; 8.5), and the canonical morphism from the coimage of the morphism to its image is an isomorphism (homomorphism theorem, 8.7). Therefore the category is abelian. 2 C OROLLARY 8.21 The nitely generated co-commutative Hopf algebras over a eld form an abelian category.
A SIDE 8.22 Theorem 8.20 is generally credited to Grothendieck but, as we have seen, it is a fairly direct consequence of allowing the coordinate rings to have nilpotent elements. See SGA3, VIA , 5.4; DG III 3, 7.4, p.355. Corollary 8.21 is proved purely in the context of Hopf algebras in Sweedler 1969, Chapter XVI, for nite-dimensional co-commutative Hopf algebras, and in Takeuchi 1972, 4.16, for nitely generated cocommutative Hopf algebras.
8h
Exercises
E XERCISE 8-1 Let H and N be subgroups of the algebraic group G such that H normalizes N . Show that the kernel of O.G/ ! O.HN / is equal to the kernel of the composite O.G/ ! O.G/ k O.G/ ! O.H / k O.N /:
N OTES As noted earlier, in much of the expository literature (e.g., Humphreys 1975, Borel 1991, Springer 1998), algebraic group means smooth algebraic group. With this terminology, many of the results in this section become false.20,21 Fortunately, because of Theorem 5.24, this is only a problem in nonzero characteristic. The importance of allowing nilpotents was pointed out by Cartier (1962) more than forty years ago, but, except for Demazure and Gabriel 1970 and Waterhouse 1979, this point-of-view has not been adopted in the expository literature. Contrast our treatment of the isomorphism theorems and the Schreier renement theorem with that in Rosenlicht 1956.
20 For example, in the category of smooth groups, the homomorphism H=H \ N ! HN=N is a purely inseparable isogeny of degree q where q is the multiplicity of H \ N in the intersection product H N . 21 The situation is even worse, because these books use a terminology based on Weils Foundations, which sometimes makes it difcult to understand their statements. For example, in Humphreys 1975, p.218, one nds the following statement: for a homomorphism 'W G ! G 0 of k-groups, the kernel of ' need not be dened over k. By this, he means 0 the following: form the kernel N of 'k al W Gk al ! Gk al (in our sense); then Nred need not arise from a smooth algebraic group over k. Of course, with our (or any reasonable) denitions, the kernel of a homomorphism of algebraic groups over k is certainly an algebraic group over k.
Recovering a group from its representations; Jordan decompositions
By a character of a topological group, I mean a continuous homomorphism from the group to the circle group fz 2 C j z z D 1g. A nite abelian group G can be recovered from its group G _ of N characters because the canonical homomorphism G ! G __ is an isomorphism. More generally, a locally compact abelian topological group G can be recovered from its character group because, again, the canonical homomorphism G ! G __ is an isomorphism (Pontryagin duality). Moreover, the dual of a compact abelian group is a discrete abelian group, and so, the study of compact abelian topological groups is equivalent to that of discrete abelian groups. Clearly, abelian is required in the above statements, because any character will be trivial on the derived group. However, Tannaka showed that it is possible to recover a compact nonabelian group from its category of unitary reprsesentations. In this section, I discuss an analogue of this for algebraic groups.
9a
Recovering a group from its representations
Let G be an algebraic monoid with coordinate ring A, and let rA W G ! EndA be the regular representation. Recall that g 2 G.R/ acts on f 2 A according to the rule: .gf /R .x/ D fR .x g/ all x 2 G.R/: (81)
L EMMA 9.1 Let G be an afne monoid over a ring k, and let A D O.G/. Let be a k-linear map A ! A for which the three diagrams k
e
! A ? ? y ! A
A k A ? ? y A k A
! A ? ? y ! A;
A ? ? y A
! AA ? ? y1 ! AA
commute. Then there exists a unique g 2 G.k/ such that D rA .g/. P ROOF. The commutativity of the rst two diagrams says that is a homomorphism of A as a k-algebra, and so (by the Yoneda lemma) there exists morphism W G ! G of set-valued functors such that .f /R .x/ D fR .
R x/
all f 2 A, x 2 G.R/:
(82)
The commutativity of the third diagram says that, for f 2 A, . /.f / D ..1 / /.f /: Recall that . f /R .x; y/ D fR .x y/ for f 2 A (see 4h). Therefore, for x; y 2 G.R/, .LHS/R .x; y/ D .f /R .x y/ D fR . .RHS/R .x; y/ D . f /R .x; 102
R y/ R .x
y//
R y/: 22
D fR .x
9. Recovering a group from its representations; Jordan decompositions Hence

R .x
103
y/ D x
R .y/;
all x; y 2 G.R/:
R .x/ .81)
On setting y D e in the last equation, we nd that f 2 A and x 2 G.R/,

.82/
D x g with g D
R .e/.
Therefore, for
.f /R .x/ D fR .x g/ D .gf /R .x/. Hence D rA .g/. The uniqueness of g follows from the faithfulness of the regular representation.
2
T HEOREM 9.2 Let G be an afne monoid (or group) over a eld k, and let R be a k-algebra. Suppose that we are given, for each nite-dimensional representation rV W G ! EndV of G, an Rlinear map V W R k V ! R k V . If the family . V / satises the conditions, (a) for all representations V; W ,
V W
W;
(b) 1 D id1 (here 1 D k with the trivial action) 1 1 1 (c) for all G-equivariant maps W V ! W ,
W
.idR / D .idR /
V
V;
then there exists a unique g 2 G.R/ such that
D rV .g/ for all V .
P ROOF. Recall (7.16) that every representation of G is a ltered union of nite-dimensional representations. It follows that, for each representation rV W G ! GLV of G (possibly innite dimensional), there exists a unique R-linear endomorphism V of R k V such that V jW D W for each nite-dimensional subrepresentation W V . The conditions (a,b,c) will continue to hold for the enlarged family. Let A D R k O.G/, and let A W A ! A be the R-linear map corresponding to the regular representation r of G on O.G/. The map mW A k A ! A is equivariant for the representations r r and r,23 and so the rst two diagrams in (9.1) commute with and replaced by A and AA D A A respectively. Similarly, the map W A ! A A is equivariant for the representation 1 r on A k A, and so the third diagram in (9.1) commutes with replaced by A . Now Lemma 9.1, applied to the afne monoid GR over R, shows that there exists a g 2 G.R/ such A D r.g/.
22 In
detail, let f D
fi gi ; then .RHS/R .x; y/ D D D D P P P

i i
fi gi
R .x; y/
fiR .x/ .gi /R .y/ fiR .x/ giR .

i fi gi R y/: R y/ R y/ R .x;
D . f /R .x;
23 We
check that, for x 2 G.R/; .r.g/ m/ .f f 0 /.x/ D .r.g/.ff 0 //.x/ D .ff 0 /.xg/ D f .xg/ f 0 .xg/ .m r.g/ r.g// .f f 0 /.x/ D ..r.g/f / .r.g/f 0 /.x/ D f .xg/ f 0 .xg/:
104
BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1) Let .V; rV / be a nite-dimensional representation of G, and let u 2 V _ . Then
uW V
! O.G/;
u .v/
D hu; vi
is equivariant (see 7.33), and so

u V
A u
D r.g/
u rV .g/:
As this holds for all u 2 V _ , this implies that V D rV .g/. This proves the existence of g, and the uniqueness follows the fact that G admits a faithful family of nite-dimensional representations (see 7.29). 2 We close this subsection with a series of remarks. 9.3 Each g 2 G.R/ of course denes such a family. Thus, from the category Repk .G/ of representations of G on nite-dimensional k-vector spaces we can recover G.R/ for any k-algebra R, and hence the group G itself. For this reason, Theorem 9.2 is often called the reconstruction theorem. 9.4 Let . V / be a family satisfying the conditions (a,b,c) of Theorem 9.4. When G is an afne group (rather than just a monoid), each V is an isomorphism, and the family satises the condition _ V _ D . V / (because this is true of the family .rV .g//). 9.5 Let !R be the forgetful functor RepR .G/ ! ModR , and let End .!R / be the set of natural transformations W !R ! !R commuting with tensor products the last condition means that satises conditions (a) and (b) of the theorem. The theorem says that the canonical map G.R/ ! End .!R / is an isomorphism. Now let End .!/ denote the functor R 7! End .!R /; then G ' End .!/. When G is a group, this can be written G ' Aut .!/. 9.6 Suppose that k is algebraically closed and that G is reduced, so that O.G/ can be identied with a ring of k-valued functions on G.k/. It is possible to give an explicit description description of O.G/ in terms of the representations of G. For each representation .V; rV / of G (over k/ and u 2 V _ , we have a function u on G.k/,
u .g/
D hu; rV .g/i 2 k:
Then u 2 O.G/, and every element of O.G/ arises in this way (cf. Springer 1998, p.39, and Exercise 4-2). 9.7 Let H be a subgroup of an algebraic group G. For each k-algebra R, let H 0 .R/ be the subgroup of G.R/ xing all tensors in all representations of G xed by H . The functor R H 0 .R/ is representable by a subgroup H 0 of G, which clearly contains H . It follows from the theorem that H0 D H. 9.8 In (9.7), instead of all representations of G, it sufces to choose a faithful representation V and take all quotients of subrepresentations of a direct sum of representations of the form n .V V _ / (by 7.43).
9. Recovering a group from its representations; Jordan decompositions
105
9.9 In general, we cant omit quotients of from (9.8). 24 However, we can omit it if some nonzero multiple of every homomorphism H ! Gm extends to a homomorphism G ! Gm . 9.10 Lemma 9.1 and its proof are valid with k a commutative ring. Therefore (using 7.10), one sees that Theorem 9.2 holds with k a noetherian ring and Repk .G/ the category of representations of G on nitely generated k-modules, or with k a Dedekind domain, G a at group scheme, and Repk .G/ the category of representations of G on nitely generated projective k-modules (or even nitely generated free k-modules).
9b
Application to Jordan decompositions
J ORDAN DECOMPOSITION OF A LINEAR MAP

In this subsubsection, we review some linear algebra. Recall that an endomorphism of a vector space V is diagonalizable if V has a basis of eigenvectors for , and that it is semisimple if it becomes diagonalizable after an extension of the base eld k. For example, the linear map x 7! AxW k n ! k n dened by an n n matrix A is diagonalizable if and only if there exists an invertible matrix P with entries in k such that PAP 1 is diagonal, and it is semisimple if and only if there exists such a matrix P with entries in some eld containing k. From linear algebra, we know that is semisimple if and only if its minimum polynomial m .T / has distinct roots; in other words, if and only if the subring k ' kT =.m .T // of Endk .V / generated by is separable. Recall that an endomorphism of a vector space V is nilpotent if m D 0 for some m > 0, and that it is unipotent if idV is nilpotent. Clearly, if is nilpotent, then its minimum polynomial divides T m for some m, and so the eigenvalues of are all zero, even in k al . From linear algebra, we know that the converse is also true, and so is unipotent if and only if its eigenvalues in k al all equal 1. Let be an endomorphism of a nite-dimensional vector space V over k. We say that has all of its eigenvalues in k if the characteristic polynomial P .T / of splits in kX : P .T / D .T a1 /n1 .T ar /nr ; ai 2 k:
For each eigenvalue a of in k, the generalized eigenspace is dened to be: Va D fv 2 V j . a/N v D 0; N sufciently divisible25 g:
P ROPOSITION 9.11 If has all of its eigenvalues in k, then V is a direct sum of its generalized eigenspaces: M V D V ai .
i
for example, the subgroup B D 0 of GL2 acting on V D k k and suppose that a vector v 2 .V V _ /n is xed by B. Then g 7! gv is a regular map GL2 =B ! .V V _ /n of algebraic varieties (not afne). But GL2 =B ' P1 , and so any such map is trivial. Therefore, v is xed by GL2 , and so B 0 D B. Cf 6.49. 24 By this I mean that there exists an N such that the statement holds for all positive integers divisible by N , i.e., that 0 0 N is sufciently large for the partial ordering M N M divides N:
24 Consider
106
P ROOF. Let P .T / be a polynomial in kT such that P ./ D 0, and suppose that P .T / D Q.T /R.T / with Q and R relatively prime. Then there exist polynomials a.T / and b.T / such that a.T /Q.T / C b.T /R.T / D 1: For any v 2 V , a./Q./v C b./R./v D v, (83)
which implies immediately that Ker.Q.// \ Ker.R.// D 0. Moreover, because Q./R./ D 0, (83) expresses v as the sum of an element of Ker.R.// and an element of Ker.Q.//. Thus, V is the direct sum of Ker.Q.// and Ker.P .//. On applying this remark repeatedly, we nd that M V D Ker.T a1 /n1 Ker..T a2 /n2 .T ar /nr / D D Ker.T ai /ni ;
i
as claimed.
T HEOREM 9.12 Let V be a nite-dimensional vector space over a perfect eld. For any automorphism of V , there exist unique automorphisms s and u of V such that (a) D s u D u s , and (b) s is semisimple and u is unipotent. Moreover, each of s and u is a polynomial in . P ROOF. Assume rst that hasL of its eigenvalues in k, so that V is a direct sum of the generalall ized eigenspaces of , say, V D 1i m Vai where the ai are the distinct roots of P . Dene s to be the automorphism of V that acts as ai on Vai for each i. Then s is a semisimple automorphism of V , and u D s 1 commutes with s (because it does on each Va ) and is unipotent (because its eigenvalues are 1). Thus s and u satisfy (a) and (b). Because the polynomials .T ai /ni are relatively prime, the Chinese remainder theorem shows that there exists a Q.T / 2 kT such that Q.T / a1 mod .T Q.T / a2 mod .T : Then Q./ acts as ai on Vai for each i , and so s D Q./, which is a polynomial in . Similarly, s 1 2 k, and so u D s 1 2 k. It remains to prove the uniqueness of s and u . Let D s u be a second decomposition satisfying (a) and (b). Then s and u commute with , and therefore also with s and u (because they are polynomials in ). It follows that s 1 s is semisimple and that u u 1 is unipotent. Since they are equal, both must equal 1. This completes the proof in this case. In the general case, because k is perfect, there exists a nite Galois extension k 0 of k such that has all of its eigenvalues in k 0 . Choose a basis for V , and use it to attach matrices to endomorphisms of V and k 0 k V . Let A be the matrix of . The rst part of the proof allows us to write A D As Au D Au As with As a semisimple matrix and Au a unipotent matrix with entries in k 0 ; moreover, this decomposition is unique.
def def
a1 /n1 a2 /n2
107
Let 2 Gal.k 0 =k/, and for a matrix B D .bij /, dene B D . bij /. Because A has entries in k, A D A. Now A D . As /. Au / is again a decomposition of A into commuting semisimple and unipotent matrices. By the uniqueness of the decomposition, As D As and Au D Au . Since this is true for all 2 Gal.K=k/, the matrices As and Au have entries in k. Now D s u , where s and u are the endomorphisms with matrices As and Au , is a decomposition of satisfying (a) and (b). Finally, the rst part of the proof shows that there exist ai 2 k 0 such that As D a0 C a1 A C C an
n 1 1A
.n D dim V /:
The ai are unique, and so, on applying , we nd that they lie in k. Therefore, s D a0 C a1 C Similarly, u 2 k. The automorphisms s and u are called the semisimple and unipotent parts of , and D s u D u s is the (multiplicative) Jordan decomposition of . P ROPOSITION 9.13 Let and be automorphisms of vector spaces V and W over a perfect eld, and let 'W V ! W be a linear map. If ' D ', then ' s D s ' and ' u D u '. P ROOF. It sufces to prove this after an extension of scalars, and so we may suppose that both and have all of their eigenvalues in k. Recall that s acts on each generalized eigenspace Va , a 2 k, as multiplication by a. As ' obviously maps Va into Wa , it follows that ' s D s '. Similarly, ' s 1 D s 1 ', and so ' u D u '. 2 C OROLLARY 9.14 Every subspace W of V stable under is stable under s and u , and jW D s jW u jW is the Jordan decomposition of jW: P ROOF. It follows from the proposition that W is stable under s and u , and it is obvious that the decomposition jW D s jW u jW has the properties to be the Jordan decomposition. 2 P ROPOSITION 9.15 For any automorphisms and of vector spaces V and W over a perfect eld, . /s D s s . /u D u u : P ROOF. It sufces to prove this after an extension of scalars, and so we may suppose that both and have all of their eigenvalues in k. For any a; b 2 k, Va k Wb .V k W /ab , and so s s and . /s both act on Va k Wb as multiplication by ab. This shows that . /s D s s . Similarly, .s 1 s 1 / D . /s 1 , and so . /u D u u . 2 C an
1 n 1
2 k:
2
108
9.16 Let k be a nonperfect eld of characteristic 2, so that there exists an a 2 k that is not a square p 0 in k, and let M D a 1 . In k a, M has the Jordan decomposition 0 p p a 0 0 1= a p p : MD 0 a a 0 Since these matrices do not have coefcients in k, the uniqueness shows that M does not have a Jordan decomposition in M2 .k/.
I NFINITE - DIMENSIONAL VECTOR SPACES

Let V be a vector space, possibly innite dimensional, over a perfect eld k. An endomorphism of V is locally nite if V is a union of nite-dimensional subspaces stable under . A locally nite endomorphism is semisimple (resp. locally nilpotent, locally unipotent) if its restriction to each stable nite-dimensional subspace is semisimple (resp. nilpotent, unipotent). Let be a locally nite automorphism of V . By assumption, every v 2 V is contained in a nitedimensional subspace W stable under , and we dene s .v/ D .jW /s .v/. According to (9.12), this is independent of the choice of W , and so in this way we get a semisimple automorphism of V . Similarly, we can dene u . Thus: T HEOREM 9.17 For any locally nite automorphism of V , there exist unique automorphisms s and u such that (a) D s u D u s ; and (b) s is semisimple and u is locally unipotent. For any nite-dimensional subspace W of V stable under , jW D .s jW / .u jW / D .u jW / .s jW / is the Jordan decomposition of jW .
J ORDAN DECOMPOSITIONS IN ALGEBRAIC GROUPS

Finally, we are able to prove the following important theorem. T HEOREM 9.18 Let G be an algebraic group over a perfect eld k. For any g 2 G.k/ there exist unique elements gs ; gu 2 G.k) such that, for all representations .V; rV / of G, rV .gs / D rV .g/s and rV .gu / D rV .g/u . Furthermore, g D gs gu D gu gs : (84)
P ROOF. In view of (9.13) and (9.15), the rst assertion follows immediately from (9.2) applied to the families .rV .g/s /V and .rV .g/u /V . Now choose a faithful representation rV . Because rV .g/ D rV .gs /rV .gu / D rV .gu /rV .gs /; (84) follows.
2
The elements gs and gu are called the semisimple and unipotent parts of g, and g D gs gu is the Jordan decomposition of g.
109
9.19 To check that a decomposition g D gs gu is the Jordan decomposition, it sufces to check that r.g/ D r.gs /r.gu / is the Jordan decomposition of r.g/ for a single faithful representation of G. 9.20 Homomorphisms of groups preserve Jordan decompositions. To see this, let W G ! G 0 be a homomorphism and g D gs gu a Jordan decomposition in G.k/. For any representation 'WG 0 ! GLV , ' is a representation of G, and so .' /.g/ D ..' /.gs // ..' /.gu // is the Jordan decomposition in GL.V /. If we choose ' to be faithful, this implies that .g/ D .gs / .gu / is the Jordan decomposition of .g/.
N OTES Our proof of the existence of Jordan decompositions (Theorem 9.18) is the standard one, except that we have made Lemma 9.1 explicit. As Borel has noted (Borel 1991, p88; Borel 2001, VIII 4.2, p169), the result essentially goes back to Kolchin 1948, 4.7.
10
Characterizations of categories of representations
Pontryagin duality has two parts. First it shows that a locally compact abelian group G can be recovered from its dual. This it does by showing that the canonical map G ! G __ is an isomorphism. Secondly, it characterizes the abelian groups that arise as dual groups. For example, it shows that the duals of discrete abelian groups are exactly the compact abelian groups, and that the duals of locally compact abelian groups are exactly the locally compact abelian groups. In 9 we showed how to recover an algebraic group G from its dual Rep.G/. In this section, we characterize the categories that arise as the category of representations of an algebraic or afne group. Throughout, k is a eld. In Theorems 10.1, 10.5, and 10.8, C is a small category (or, at least, admits a set of representatives for its isomorphism classes of objects).
10a
Characterization of categories of comodules
An additive category C is said to be k-linear if the Hom sets are k-vector spaces and composition is k-bilinear. Functors of k-linear categories are required to be k-linear, i.e., the maps Hom.a; b/ ! Hom.F a; F b/ dened by F are required to be k-linear. Recall that Veck denotes the category of nite-dimensional vector spaces over k. T HEOREM 10.1 Let C be a k-linear abelian category, and let !W C ! Veck be an exact faithful k-linear functor. Then there exists a coalgebra C such that C is equivalent to the category of C comodules of nite dimension.
For the proof, we follow Serre 1993, 2.5. For a slightly different proof, see Deligne and Milne 1982, 2, or Saavedra Rivano 1972. We shall frequently make use of the fact that, because ! is faithful, if !./ is a monomorphism (resp. an epimorphism), then so also is . For objects X , Y of C, Hom.X; Y / is a subspace of Hom.!X; !Y /, and hence has nite dimension. For monomorphisms X ! Y and X 0 ! Y with the same target, write x x 0 if there exists a morphism X ! X 0 (necessarily unique) giving a commutative triangle. The lattice of subobjects of Y is obtained from the collection of monomorphisms by identifying two monomorphisms x and x 0 if x x 0 and x 0 x. The functor ! maps the lattice of subobjects of Y injectively26 to the lattice of subspaces of !Y . Hence X has nite length. Similarly ! maps the lattice of quotient objects of Y injectively to the lattice of quotient spaces of !Y . For X in C, we let hX i denote the full subcategory of C whose objects are the quotients of subobjects of direct sums of copies of X. Let X be an object of C. For any subset S of !.X /, there exists a smallest subobject Y of X such that !.Y / S , which we call the subobject of X generated by S . An object Y is monogenic if it is generated by a single element, i.e., there exists a y 2 !.Y / such that the only subobject Y 0 of Y such y 2 !.Y 0 / is Y itself.
26 If
x0
!.X / D !.X 0 /, then the kernel of
x x0
WX
X0 ! Y
projects isomorphically onto each of X and X 0 (because it does after ! has been applied).
110
10. Characterizations of categories of representations
111
P ROOF OF T HEOREM 10.1

The case that C D hXi for some object X. Let n D dimk !.X /. L EMMA 10.2 For any monogenic object Y of C, dimk !.Y / n2 : P ROOF. By hypothesis, Y D Y1 =Y2 where Y1 is isomorphic to a subobject of X m for some m. Let y 2 !.Y / generate Y , and let y1 be an element of !.Y1 / whose image in !.Y / is y. Let Z be the subobject of Y1 generated by y1 . The image of Z in Y D Y1 =Y2 is Y , and so it sufces to prove the lemma for Z, i.e., we may suppose that Y X m for some m. If m n, the lemma is obvious.. Suppose therefore that m > n. We have y 2 !.Y / !.X m / D !.X/m . Let y D .z1 ; : : : ; zm / in !.X /m . Since m > n, there exist ai 2 k, not all zero, such that P ai zi D 0. But the ai dene a surjective morphism X m ! X; if N is the kernel of this morphism, then one sees easily27 that N X m 1 . On the other hand, y1 2 !.N /, and so Y N because y1 generates Y . Therefore Z embeds into X m 1 . Continue in this fashion until Z X m with m n.2 As dimk !.Y / can take only nitely many values when Y is monogenic, there exists a monogenic P for which dimk !.P / has its largest possible value. Let p 2 !.P / generate P . L EMMA 10.3 (a) The pair .P; p/ represents the functor !. (b) The object P is a projective generator for C. P ROOF. (a) Let X be an object of C, and let x 2 !.X /; we have to prove that there exists a unique morphism f W P ! X such that !.f / sends p to x. The uniqueness follows from the fact p generates P . To prove the existence, let Q be the smallest subobject of P X such that !.Q/ contains .p; x/. The morphism Q ! P dened by the projection map is surjective, because P is generated by p. Therefore, dimk !.Q/ dimk !.P /; but because dimk .!.P // is maximal, equality must hold, and so Q ! P is an isomorphism. The composite of its inverse with the second projection Q ! X is a morphism P ! X sending p to x. (b) The object P is projective because ! is exact, and it is a generator because ! is faithful. 2 Let A D End.P / it is a k-algebra of nite dimension as a k-vector space (not necessarily commutative). L EMMA 10.4 The functor X Hom.P; X / is an equivalence from C to the category of right Amodules of nite dimension over k; its composite with the forgetful functor is !. P ROOF. Because P is a generator, the functor is fully faithful, and because P is projective, it is exact. It remains to prove that it is essentially surjective. Let M be a nite-dimensional right A-module, and choose a nite presentation for M , Am ! An ! M ! 0
27 Let
A be an m
1
of N onto X m
a1 : : : a m is invertible. Then AW X m ! X m A (because it does after ! has been applied). 1 m matrix such that
denes an isomorphism
112
where is an m n matrix with coefcients in A. This matrix denes a morphism P m ! P n whose cokernel X has the property that Hom.P; X / M . We have !.X / ' Hom.P; X / ' Hom.hP .P /; hP .X // D Hom.A; hP .X // ' hP .X /:
2
As A is a nite k-algebra, its linear dual C D A_ is a k-coalgebra, and the category of right A-modules is equivalent to the category of left C -comodules (see 7.7). Together with (10.4), this completes the proof in this case. The general case. For an object X of C, let AX be algebra of endomorphisms of !jhX i, and let CX D A_ . Then ! denes an equivalence of categories X hXi ! Comod.CX /: We dene a partial ordering on the set of isomorphism classes of objects in C by the rule: X Y if hX i hY i.
Note that X ; Y X Y , so that we get a directed set, and that if X Y , then restriction denes a homomorphism AY ! AX . When we pass to the limit over the isomorphism classes, we obtain the following more precise form of the theorem. T HEOREM 10.5 Let C be a k-linear abelian category and let !W C ! Veck be a k-linear exact faithful functor. Let B D lim End.!jhXi/_ . Then ! denes an equivalence of categories C ! ! Comod.B/ carrying ! into the forgetful functor. E XAMPLE 10.6 Let A be a nite k-algebra (not necessarily commutative), and let R be a commutative k-algebra. Consider the functors
Mod.A/
! forget
! Vec.k/
V 7!Rk V
! Mod.R/:
R !/
For M 2 ob.Mod.A//, let M0 D !.M /. An element

M W R k
of End.
is a family of R-linear maps
M0 ! R k M0 ,
functorial in M . An element of R k A denes such a family, and so we have a map W R k A ! End.

R !/;
which we shall show to be an isomorphism by dening an inverse . Let . / D A .1 1/. Clearly D id, and so we only have to show D id. The A-module Ak M0 is a direct sum of copies of A, and the additivity of implies that AM0 D A idM0 . The map a m 7! amW A k M0 ! M is A-linear, and hence R k A k M0 ! R k M ? ? ? id ? y A M0 y M R k A k M0 commutes. Therefore
M .1 m/
! R k M
A .1/ m
D . . //M .1 m/ for 1 m 2 R M;
10. Characterizations of categories of representations i.e., D id.
113
In particular, A ! End.!/, and it follows that, if in (10.5) we take C D Mod.A/, so that C D hAi, then the equivalence of categories obtained is the canonical equivalence (7.7).
A SIDE 10.7 Need to rewrite this section to obtain the following result: Let C be a k-linear abelian category for which there exists an exact faithful k-linear functor. Assume C has a tensor product. Then there exists a coalgebra C in C together with a coaction of C on each object of A such that every exact faithful k-linear functor ! to Veck denes an equivalence A ! Comod!.C / ; plus a uniqueness statement.
'
10b
Characterization of categories of representations of afne groups
Let !W A ! B be a faithful functor. We say that a morphism !X ! !Y lives in A if it lies in Hom.X; Y / Hom.!X; !Y /. For k-vector spaces U; V; W , there are canonical isomorphisms
U;V;W W U
k .V k W / ! .U k V / k W; u .v w/ 7! .u v/ w W U k V ! V U; u v 7! v u: U;V T HEOREM 10.8 Let C be a k-linear abelian category, let be a k-bilinear functor C C ! C, and let ! be an exact faithful k-linear functor C ! Veck satisfying the following conditions: (a) (b) (c) (d) is an isomorphism if !./ is; for all X; Y , !.X Y / D !.X / k !.Y /I for all X; Y; Z; the isomorphisms !X;!Y;!Z and !X;!Y live in C; there exists an object 1 in C such that !.1 D k and the canonical isomorphisms 1 1/ !.1 !.X / ' !.X / ' !.X / !.1 1/ 1/ live in C; (e) if !.X / has dimension 1, then there exists an object X For each k-algebra R, let G.R/ be the set of families . such that D V W for all V; W 2 ob.C/, D id!.1 for every identity object of 1 of C, and 1 1 1 1/ W !./R D !./R V for all arrows in C.
V W V /V 2ob.C/ ; V 1
in C such that X X
D1 1.
2 EndR-linear .!.V /R /;
Then G is an afne group over k, and ! denes an equivalence of tensor categories over k,
C ! Rep.G/:
We have already dened the coalgebra of G in Theorem 10.5. The tensor structure will enable us to dene an algebra structure on A. With the notations of the previous subsection, let BX D A_ . Then ! denes an equivalence of X categories from hX i to Comod.BX /. Let B be a coalgebra over k and let ! be the forgetful functor Comod.B/ ! Vec.k/. The discussion in Example 10.6 shows that B D lim End.!jhXi/_ . We deduce easily that every functor !
114
unique homomorphism B ! B 0 . Again, let B be a coalgebra over k. A coalgebra homomorphism uW B k B ! B denes a functor u W Comod.B/ Comod.B/ ! Comod.B/ sending .X; Y / to X k Y with the coaction X Y (see 7.7).
X Y
Comod.B/ ! Comod.B 0 / carrying the forgetful functor into the forgetful functor arises from a
! X B Y B ' X Y B B
X Y u
! X Y B
P ROPOSITION 10.9 The map u 7! u denes a one-to-one correspondence between the set of homomorphisms B k B ! B and the set of functors W Comod.B/ Comod.B/ ! Comod.B/ such that .X; Y / D X k Y as k-vector spaces. The natural associativity and commutativity constraints on Veck induce similar constraints on .Comod.B/; u / if and only if the multiplication dened by u on B is associative and commutative; there is an identity object in .Comod.B/; u / with underlying vector space k if and only if B has an identity element. P ROOF. The pair .Comod.B/ Comod.B/; ! !/, with .! !/.X Y / D !.X / !.Y / (as a k-vector space), satises the conditions of (10.5), and lim End.! !jh.X; Y /i/_ D B B. Thus ! the rst statement of the proposition follows from (10.6). The remaining statements are easy. 2 Let .C; !/ and B be as in (10.5) except now assume that C is a tensor category and ! is a tensor functor. The tensor structure on C induces a similar structure on Comod.B/, and hence, because of (10.9), the structure of an associative commutative k-algebra with identity element on B. Thus B lacks only a coinverse map S to be a k-bialgebra (in our sense) and G D Spec B is an afne monoid scheme. Using (10.6) we nd that, for any k-algebra R, End.!/.R/ D End.
def
R !/
D lim Homk-lin .BX ; R/ D Homk-lin .B; R/.
An element 2 Homk-lin .BX ; R/ corresponds to an element of End.!/.R/ commuting with the tensor structure if and only if is a k-algebra homomorphism; thus End .!/.R/ D Homk-alg .B; R/ D G.R/: We have shown that, if in the statement of Theorem 10.8, condition (e) is omitted, then one can conclude that End .!/ is representable by an afne monoid G D Spec B and that ! denes an equivalence of tensor categories
C ! Comod.B/ ! Repk .G/.
It is easy to see that (e) implies that End .!/ D Aut .!/ (apply (e) to the highest exterior power of an object), which completes the proof. R EMARK 10.10 Let .C; !/ be .Repk .G/;forget/. On following through the proof of (10.8) in this case one recovers Theorem 9.2: Aut .! G / is represented by G.
A SIDE 10.11 Add discussion of how much of this section extends to base rings k. (Cf. mo3131.)
11
Finite algebraic groups
Throughout this section, k is a eld and all k-algebras are commutative. For a nite k-algebra A, we let AW k denote the degree of A (dimension of A as a k-vector space). An algebraic group G is nite if O.G/ is a nite k-algebra, in which case O.G/W k is called the order of G. We say that an algebraic group is a p-group if it is nite and its order is a power of p.
11a
Etale groups
S EPARABLE k- ALGEBRAS
Let A be a nite k-algebra. For any nite set S of maximal ideals in A, the Chinese remainder theQ T orem (CA 2.12) says that the map A ! m2S A=m is surjective with kernel m2S m. In particular, jS j AW k, and so A has only nitely many maximal ideals. If S is the set of all maximal ideals in T A, then m2S m is the nilradical N of A (CA 11.8), and so A=N is a nite product of elds. P ROPOSITION 11.1 The following conditions on a nite k-algebra A are equivalent: (a) A is a product of separable eld extensions of k; (b) A k k al is a product of copies of k al ; (c) A k k al is reduced. P ROOF. (a))(b). We may suppose that A itself is a separable eld extension of k. From the primitive element theorem (FT 5.1), we know that A D k for some . Because k is separable over k, the minimum polynomial f .X / of is separable, which means that Y f .X / D .X i /; i j for i j; in k al X. Now A k k al ' .kX =.f // k k al ' k al X =.f /, and, according to the Chinese remainder theorem (CA 2.12), Y k al X =.f / ' k al X =.X i / ' k al
i
k al .
(b))(c). Obvious. (c))(a). The map a 7! a 1W A ! A k k al is injective, and so A is reduced. Therefore the above discussion shows that it is a nite product of elds. Let k 0 one of the factors of A. If k 0 is not separable over k, then k has characteristic p 0 and there exists an element of k 0 whose minimum polynomial is of the form f .X p / with f 2 kX (see FT 3.6, et seq.). Now k k k al ' .kX =.f .X p // k k al ' k al X =.f .X p //; which is not reduced because f .X p / is a pth power in k al X . Hence A k k al is not reduced.
2
D EFINITION 11.2 A nite k-algebra satisfying the equivalent conditions of the proposition is said to be separable. P ROPOSITION 11.3 Finite products, tensor products, and quotients of separable k-algebras are separable. 115
116

2
P ROOF. This is obvious from the condition (b).
C OROLLARY 11.4 The composite of any nite set of separable subalgebras of a k-algebra is separable. P ROOF. Let Ai be separable subalgebras of B. Then A1 a1 and so is a quotient of A1 k an 7! a1 k An . an W A1 k An is the image of the map k An ! B;
2
P ROPOSITION 11.5 If A is separable over k, then A k k 0 is separable over k 0 for any eld extension k 0 of k. P ROOF. Let k 0al be an algebraic closure of k 0 , and let k al be the algebraic closure of k in k al . Then k0 x ? ? k is commutative, and so .A k k 0 / k 0 k 0al ' A k k al k al k 0al ' .k al ' k 0al k al / k al k 0al k 0al :
2
! k 0al x ? ? ! k al
C LASSIFICATION OF SEPARABLE k- ALGEBRAS

Let k sep be the composite of the subelds k 0 of k al separable over k. If k is perfect, for example, of characteristic zero, then k sep D k al . Let be the group of k-automorphisms of k sep . For any subeld K of k sep , nite and Galois over k, an easy Zorns lemma argument28 shows that 7! jKW ! Gal.K=k/ ,
is surjective. Let X be a nite set with an action of
X ! X: We say that the action is29 continuous if it factors through k sep nite and Galois over k. For a separable k-algebra A, let ! Gal.K=k/ for some subeld K of
F .A/ D Homk-alg .A; k al / D Homk-alg .A; k sep /: Then acts on F .A/ through its action on k sep : . f /.a/ D .f .a//; 2 , f 2 F .A/, a 2 A:
The images of all homomorphisms A ! k sep will lie in some nite Galois extension of k, and so the action of on F .A/ is continuous.
2 Gal.K=k/. Apply Zorns lemma to the set of all pairs .E; / where E is a subeld of k sep containing k and is homomorphism E ! k sep whose restriction to K is 0 . 29 Equivalently, the action is continuous relative to the discrete topology on X and the Krull topology on (FT 7).
0 28 Let
11. Finite algebraic groups
117
T HEOREM 11.6 The map A 7! F .A/ is a contravariant equivalence from the category separable k-algebras to the category of nite sets with a continuous action of . P ROOF. This is a restatement of the fundamental theorem of Galois theory (FT 3), and is left as an exercise to the reader (the indolent may see Waterhouse 1979, 6.3). 2 Suppose that A is generated by a single element, so A D kX =.f .X // D kx. Then A is separable if and only if f .X / has distinct roots in k al . Assume this, and (for simplicity) that f .X / is monic. A k-algebra homomorphism A ! k sep is determined by the image of x, which can be any root of f in k sep . Therefore, F .A/ can be identied with the set of roots of f . Suppose F .A/ decomposes into r orbits under the action of , and let f1 ; : : : ; fr be the monic polynomials whose roots are the orbits. Then each fi is stable under , and so has coefcients in k (FT 7.8). It follows that f D f1 fr is the decomposition of f into its irreducible factors over k, and that Q A D 1i r kX =.fi .X // is the decomposition of A into a product of elds.
C LASSIFICATION OF E TALE ALGEBRAIC GROUPS

Recall that an algebraic group G is said to be nite if O.G/ is nite-dimensional as a k-vector space. We say G is etale if in addition O.G/ is a separable k-algebra. R EMARK 11.7 (a) When k has characteristic zero, Theorem 5.24 says that every nite algebraic group is etale. (b) Algebraic geometers will recognize that an algebraic group G is etale if and only if the morphism of schemes jGj ! Spec k is etale. Let A be the category of separable k-algebras. The functor G O.G/ is an equivalence from the category of etale algebraic groups over k to the category of group objects in the category Aopp (see 4g). On combining this statement with Theorem 11.6, we obtain the following theorem. T HEOREM 11.8 The functor G G.k sep / is an equivalence from the category of etale algebraic groups over k to the category of nite groups endowed with a continuous action of . Let K be a subeld of k sep containing k, and let 0 be the subgroup of xing the elements of K. Then K is the subeld of k sep of elements xed by it follows that G.K/ is the subgroup G.k sep / of elements xed by 0 : consisting of the (see FT 7.10), and
E XAMPLES
For an etale algebraic group G, the order of G is the order of the (abstract) group G.k al /. Since Aut.X / D 1 when X is a group of order 1 or 2, there is exactly one etale algebraic group of order 1 and one of order 2 over k (up to isomorphism). Let X be a group of order 3. Such a group is cyclic and Aut.X / D Z=2Z. Therefore the etale algebraic groups of order 3 over k correspond to homomorphisms ! Z=2Z factoring through Gal.K=k/ for some nite Galois extension K of k. A separable quadratic extension K of k denes such a homomorphism, namely, 7! jKW ! Gal.K=k/ ' Z=2Z
and all nontrivial such homomorphisms arise in this way (see FT 7). Thus, up to isomorphism, there is exactly one etale algebraic group G K of order 3 over k for each separable quadratic extension
118
K of k, plus the constant group G0 . For G0 , G0 .k/ has order 3. For G K , G K .k/ has order 1 butp K .K/ has order 3. There are innitely many distinct quadratic extensions of Q, for example, G p p p p Q 1, Q 2, Q 3, : : :, Q p, : : :. Since 3 .Q/ D 1 but 3 .Q 3 1/ D 3, 3 must be the p group corresponding to Q 3 1.
A SIDE 11.9 Should replace separable k-algebra with tale k-algebra; align more with Bourbaki A, V e 6 (a k-algebra A is etale if A k K K n for some integer n 0). Add proof of Theorem 11.6.
11b
Finite group schemes in characteristic p 0
P ROPOSITION 11.10 Let G be a nite group scheme over a eld k of characteristic p 0, and 2 suppose that x p D 0 for all x 2 O.G/. For any basis x1 ; : : : ; xr of IG =IG , the monomials
m x1 1 m xr r ;
0 mi < p
form a basis for O.G/ as a k-vector space (and so O.G/W k D p r ). P ROOF. Omitted for the moment. The proposition says that O.G/ ' kx1 ; : : : ; xr =.x1 ; : : : ; xr /. This generalizes. T HEOREM 11.11 Let G be a nite group scheme over a eld k of characteristic p 0. For any 2 basis x1 ; : : : ; xr of IG =IG , there exist integers e1 ; : : : ; er 1 such that
p O.G/ ' kx1 ; : : : ; xr =.x1 ; : : : ; xr /: p e1
er
2 p p
P ROOF. Omitted for the moment (Waterhouse 1979, 14.4).
C LASSIFICATION
Let k be a perfect eld of characteristic p, and let W be the ring of Witt vectors with entries in k. Thus W is a complete discrete valuation ring with maximal ideal generated by p D p1W and residued eld k. For example, if k D Fp , then W D Zp . The Frobenius automorphism of W is the unique automorphism such that a ap .mod p/: T HEOREM 11.12 There exists a contravariant equivalence G M.G/ from the category of nite commutative algebraic p-groups to the category of triples .M; F; V / in which M is a W -module of nite length and F and V are endomorphisms of M satisfying the following conditions (c 2 W , m 2 M ): F .c m/ D c F m V . c m/ D c V m F V D p idM D VF: The order of G is p length.M.G// . For any perfect eld k 0 containing k, there is functorial isomorphism M.Gk 0 / ' W .k 0 / W .k/ M.G/: P ROOF. Demazure 1972, III 7, p.69.
2
11. Finite algebraic groups For example: M.Z=pZ/ D W =pW; M.

p/
119
F D 1; F D 0; F D 0;
V D 0I V D pI V D 0:
D W =pW;
M.p / D W =pW;
11c
Cartier duality revisited
Let G be a nite commutative algebraic group with bialgebra .O.G/; m; e; ; /. Recall (4j) that the Cartier dual G _ of G is the algebraic group with bialgebra .O.G/_ ; _ ; _ ; m_ ; e _ /. In this subsection, we describe G _ as a functor. Let A D O.G/. Let R be a k-algebra. Under the inclusion Homk-lin .A_ ; R/ ' HomR-lin .A_ R; R/ ,! HomR-lin .R; A R/ ' A R; k-algebra homomorphisms correspond to elements f 2 A R such that (a) R .f / D f f , and (b) eR .f / D 1. On the other hand, under the inclusion Hom.GR ; GmR / ,! Homk-alg .kX; X
1
; A R/;
homomorphisms GR ! GmR correspond to elements f 2 A R such that (a) R .f / D f f , and (c) f 2 .A R/ . L EMMA 11.13 For an element f of A R satisfying (a), the conditions (b) and (c) are equivalent. P ROOF. Omitted for the moment (easy calculation). We conclude that there is canonical bijection G _ .R/ ' Hom.GR ; GmR /: Let G_ G ! Gm (86) (85)
2
be the natural transformation corresponding to the k-algebra homomorphism ! A_ A P _ sending X to the element of A_ A such that D ei ei for any basis .ei /i 2I of A. In other words, (86) is the pairing kX; X
_ .y; x/ 7! ei .y/ei .x/W G _ .R/ G.R/ ! R : 1
T HEOREM 11.14 The pairing (86) is bi-multiplicative; for any k-algebra R, it denes the bijection (85).
120 In other words, G _ represents the functor R
Hom.GR ; GmR /:
2
P ROOF. Omitted for the moment (easy calculation).
E XAMPLE 11.15 Let G D p , so that O.G/ D kX =.X p / D kx. Let 1; y; y2 ; : : : ; yp 1 be the basis of O.G _ / D O.G/_ dual to 1; x; : : : ; x p 1 . Then y i D i yi ; in particular, y p D 0. In fact, G _ ' p , and the pairing is a; b 7! exp.ab/W p .R/ p .R/ ! R where exp.ab/ D 1 C ab .ab/2 C C 1 2 .ab/p 1 . .p 1/
A SIDE 11.16 The omitted proofs can be found in the notes on Cartier duality on Ching-Li Chais webpage.
11d
Finite at group schemes over commutative rings
Include a discussion. For the present, see: Tate, John, Finite at group schemes. Modular forms and Fermats last theorem (Boston, MA, 1995), 121154, Springer, New York, 1997.
11e
Exercises
E XERCISE 11-1 Show that A is separable if and only if there are no nonzero k-derivations DW A ! k. [Regard A as a left A-module by left multiplication. Let A be a k-algebra and M an A-module. A k -derivation is a k-linear map DW A ! M such that D.fg/ D f D.g/ C g D.f / (Leibniz rule).] E XERCISE 11-2 How many nite algebraic groups of orders 1; 2; 3; 4 are there over R (up to isomorphism)?
12
The connected components of an algebraic group
Recall that a topological space X is disconnected if it is a disjoint union of two nonempty open subsets; equivalently, if it contains a nonempty proper closed-open subset. Otherwise, it is connected. The maximal connected subspaces of X are called the connected components of X, and X is a disjoint union of them. Write 0 .X / for the set of connected components of X (for compact spaces it is nite). For a topological group G, 0 .G/ is again a group, and the kernel of G ! 0 .G/ is a normal connected subgroup G of G, called the identity (connected) component of G. For example, GL2 .R/ has two connected components, namely, the identity component consisting of the matrices with determinant > 0 and another connected component consisting of the matrices with determinant < 0. In this section, we extend these statements to afne and algebraic groups. Throughout, k is a eld.
12a
Some algebraic geometry
Throughout this subsection, A is a commutative ring. An element e of A is idempotent if e 2 D e. For example, 0 and 1 are both idempotents they are called the trivial idempotents. Idempotents e1 ; : : : ; en are orthogonal if ei ej D 0 for i j . Any sum of orthogonal idempotents is again idempotent. A set fe1 ; : : : ; en g of orthogonal idempotents is complete if e1 C C en D 1. Any set of orthogonal idempotents fe1 ; : : : ; en g can be completed by adding the idempotent e D 1 .e1 C C en /. If A D A1 An (direct product of rings), then the elements e1 D .1; 0; : : :/, e2 D .0; 1; 0; : : :/,. . . , en D .0; : : : ; 0; 1/ form a complete set of orthogonal idempotents. Conversely, if fe1 ; : : : ; en g is a complete set of orthogonal idempotents in A, then Aei becomes a ring with the addition and multiplication induced by that of A, and A ' Ae1 Aen . L EMMA 12.1 The space spec.A/ is disconnected if and only if A contains a nontrivial idempotent. P ROOF. Let e be a nontrivial nilpotent, and let f D 1 e. For a prime ideal p, the map A ! A=p must send exactly one of e or f to a nonzero element. This shows that spec A is a disjoint union of the sets D.e/ and D.f /, each of which is open. If D.e/ D spec A, then e is nilpotent (CA 2.5), and hence 0. Similarly, D.f / spec A. Conversely, suppose that specm.A/ is disconnected, say, the disjoint union of two nonempty sets V .a/ and V .b/. Because the union is disjoint, no prime ideal contains both a and b, and so a C b D A. Thus a C b D 1 for some a 2 a and b 2 b. As ab 2 a \ b, all prime ideals contain ab, which is therefore nilpotent, say .ab/m D 0. Any prime ideal containing am contains a; similarly, any prime ideal containing b m contains b; thus no prime ideal contains both am and b m , which shows that the ideal they generate is A. Therefore, we can write 1 D ram C sb m for some r; s 2 A. Now .ram /.sb m / D 0; .ram /2 D .ram /.1 sb m / D ram , ram C sb m D 1; and so fram ; sb m g is a complete set of orthogonal idempotents. Moreover, V .a/ V .ram / and V .b/ V .sb m /. As V .ram / \ V .sb m / D ;, we see that V .a/ D V .ram / and V .b/ D V .sb m /, and so neither ram nor sb m is zero. 2 P ROPOSITION 12.2 Let fe1 ; : : : ; en g be a complete set of orthogonal idempotents in A. Then spec A D D.e1 / t : : : t D.en / 121
122
is a decomposition of spec A into a disjoint union of open subsets. Moreover, every such decomposition arises in this way. P ROOF. After the above discussion, it remains to prove that every decomposition specm A D U1 t : : : t Un , each Ui open, corresponds to a decomposition A D A1 An of A into a product of rings. When n D 2 this follows from the lemma, and for n > 2 it can be proved by induction. 2 R EMARK 12.3 Recall that a ring A is said to be Jacobson if every prime ideal is an intersection of maximal ideals, and that every nitely generated algebra over a eld is Jacobson (see CA 12.3 et seq.). In a Jacobson ring, the nilradical is an intersection of maximal ideals. When A is Jacobson, prime ideal can be replaced by maximal ideal and spec with specm in the above discussion. In particular, for a Jacobson ring A, there are natural one-to-one correspondences between the decompositions of specm.A/ into a nite disjoint unions of open subspaces, the decompositions of A into a nite direct products of rings, and the complete sets of orthogonal idempotents in A.
A SIDE 12.4 On Cn there are two topologies: the Zariski topology, whose closed sets are the zero sets of collections of polynomials, and the complex topology. Clearly Zariski-closed sets are closed for the complex topology, and so the complex topology is the ner than the Zariski topology. It follows that a subset of Cn that is connected in the complex topology is connected in the Zariski topology. The converse is false. For example, if we remove the real axis from C, the resulting space is not connected for the complex topology but it is connected for the topology induced by the Zariski topology (a nonempty Zariski-open subset of C can omit only nitely many points). Thus the next result is a surprise: If V Cn is closed and irreducible for the Zariski topology, then it is connected for the complex topology. For the proof, see Shafarevich, Basic Algebraic Geometry, 1994, VII 2.
12b
Separable subalgebras
L EMMA 12.5 Let A be a nitely generated k-algebra. The degrees of the separable subalgebras of A are bounded (in fact, by the number of irreducible components of specm A k al ). P ROOF. A separable subalgebra of A will give a separable subalgebra of the same degree of A k k al , and so we can assume that k is algebraically closed. Then a separable subalgebra is of the form k k. For such a subalgebra, the elements e1 D .1; 0; : : :/; : : : ; er D .0; : : : ; 0; 1/ form a complete set of orthogonal idempotents, and so D.e1 /; : : : ; D.er / are disjoint open subsets covering specm A. Because specm A is noetherian, it is a nite union of its irreducible components (CA 12.10). Similarly, each open subsets D.ei / is a nite union of its irreducible components, each of which is an irreducible component of A. Therefore r is at most the number of irreducible components of specm A. 2 Let A be a nitely generated k-algebra. The composite of two separable subalgebras of A is separable (11.4), and so, because of the lemma, there is a separable subalgebra 0 .A/ of A containing all other separable subalgebras. Let K be a eld containing k. Then 0 .A/ k K is a separable subalgebra of A k K (see 11.5). We shall need to know that it is the largest separable subalgebra.
12. The connected components of an algebraic group
123
P ROPOSITION 12.6 Let A be a nitely generated k-algebra, and let K be a eld containing k. Then
0 .A/ k
KD
0 .A k
K/:
P ROOF. If 0 .A/ K is not maximal in A K, then 0 .A/ L will not be maximal in A L for any eld L containing K. Therefore, it sufces to prove the proposition with K algebraically closed. However, we rst prove the statement with K D k sep . The subalgebra 0 .A k sep / of A k sep is stable under the action of Gal.k sep =k/, and hence equals 0 .A k sep / k k sep (cf. 11.6). Since sep sep D sep 0 .A k /. 0 .A/ 0 .A k / , we have that 0 .A/ k k We next prove the statement when k is separably closed and K D k al . It k al k, then k has characteristic p 0 and k al is purely inseparable over it. Let e1 ; : : : ; em be a basis of idempotents P pr for 0 .A k al /. Write ej D ai ci with ai 2 A and ci 2 k al . For some r, all the elements ci P pr pr pr pr lie in k, and then ej D ai ci 2 A. But ej D ej , and so 0 .A k al / has a basis in A. Finally, we prove the statement when k and K are both algebraically closed. We may suppose that 0 .A/ D k. Since specm A D specm A=N we may suppose that A is reduced. Hence specm A is an algebraic variety over an algebraically closed eld, and a connected algebraic variety remains connected under extension of the algebraically closed base eld. 2 Let A and A0 be nitely generated k-algebras. Then 0 .A/ k 0 .A0 / is a separable subalgebra of A k A0 (see 11.3). We shall need to know that it is the largest separable subalgebra. P ROPOSITION 12.7 Let A and A0 be nitely generated k-algebras. Then
0 .A k
A0 / D
0 .A/ k
0 .A
/:
P ROOF. After (12.6) we may suppose that k is algebraically closed, in which case the proposition follows from the fact that a product of connected algebraic varieties over an algebraically closed eld is connected. 2
A SIDE 12.8 Need to rethink this. Cf. the result (AG 11.5): let A be a nitely generated k-algebra; assume that A is an integral domain and that 0 .A/ D k; then A k k sep is an integral domain (in particular, 0 .A k k sep / D k sep ).
12c
The group of connected components of an algebraic group
Let G be an algebraic group with coordinate ring A D O.G/. The map W A ! A k A is a k12.7 algebra homomorphism, and so sends 0 .A/ into 0 .A k A/ D 0 .A/ k 0 .A/. Similarly, S W A ! A sends 0 .A/ into 0 .A/, and we can dene on 0 .A/ to be the restriction of on A. With these maps 0 .A/ becomes a Hopf subalgebra of A. T HEOREM 12.9 Let G ! 0 .A/ ! A.
0G
be the quotient map corresponding to the inclusion of bialgebras
(a) Every quotient map from G to an etale algebraic group factors uniquely through G ! 0 .G/. D Ker.G ! (b) Let G 0 G/. Then G is the unique normal algebraic subgroup of G such that i) 0 G D 1, ii) G=G is etale.
124
P ROOF. (a) A quotient map G ! H corresponds to an injective homomorphism O.H / ! O.G/ of k-bialgebras. If H is etale, then O.H / is separable, and so the image of the homomorphism is def contained in 0 .O.G// D O. 0 G/. This proves (a). (b) The k-algebra homomorphism W O. 0 G/ ! k decomposes O. 0 G/ into a direct product O.
0 G/
Dk
0 G/
B. is .1 e/, and
Let e D .1; 0/. Then the augmentation ideal of O.
O.G/ D eO.G/ .1
e/O.G/
with eO.G/ ' O.G/=.1 e/O.G/ D O.G / (see 6.13). Clearly, k D 0 .eO.G// ' 0 .O.G //. Therefore 0 G D 1, and (by denition) G=G ' 0 G, which is etale. Suppose H is a second normal algebraic subgroup of G. If G=H is etale, then (by (a)) the homomorphism G ! G=H factors through 0 .G/, and so we get a commutative diagram 1 ! G ? ? y ! H ! G !
0G
! 1
? ? y ! G ! G=H ! 1
with exact rows. The similar diagram with each an exact sequence
replaced with .R/ gives, for each k-algebra R,
1 ! G .R/ ! H.R/ ! .
0 G/ .R/:
(87)
Since this functorial in R, it gives a sequence of algebraic groups 1 ! G ! H !

0 G: 0 G.
The exactness of (87) shows that G is the kernel of H ! and so if 0 H D 1, its kernel is H : therefore G ' H .
This map factors through
0H , 2
D EFINITION 12.10 The subgroup G is called identity component of G. P ROPOSITION 12.11 For any eld extension k 0 In other words, the functors G
0G
k,
0 .Gk 0 /
'
0 .G/k 0
and .Gk 0 / ' .G /k .
and G
G commute with extension of the base eld.

2
P ROOF. As O.Gk 0 / ' O.G/ k k 0 , this follows from (12.6). P ROPOSITION 12.12 For any algebraic groups G and G 0 , G 0 / ' G G 0 . P ROOF. The coordinate ring O.G
0 .G
G 0/ '
0 .G/
0 .G
0 / and .G
G 0 / ' O.G/ O.G 0 /, and so this follows from (12.7).
125
12d
Connected algebraic groups

0 .G/
D EFINITION 12.13 An algebraic group G is connected if
D 1 (i.e.,
0 .O.G//
D k).
Theorem 12.9 says that, for any algebraic group G, there is a unique exact sequence 1 ! G ! G ! such that G is connected and
0 .G/ 0 .G/
!1
is etale.
R EMARK 12.14 (a) Let G be an algebraic group over k. For any eld k 0 containing k, Proposition 12.11 shows that G is connected if and only if Gk 0 is connected. (b) Let G and G 0 be algebraic groups over k. Proposition 12.12 shows that G G 0 is connected if and only if both G and G 0 are connected. R EMARK 12.15 In particular, if an algebraic group G over a eld is connected, then so also is Gk al . In other words, a connected algebraic group is geometrically connected. This is false for algebraic varieties: for example, X2 C Y 2 D 0 is connected over R (even irreducible), but becomes a disjoint union of the two lines X C iY D 0 over C the ring RX; Y =.X 2 C Y 2 / is an integral domain, but CX; Y =.X 2 C Y 2 / ' CX; Y =.X C iY / CX; Y =.X iY /.
The reason for the difference is the existence of the homomorphism W O.G/ ! k (the neutral element of G.k/). An integral afne algebraic variety V over a eld k is geometrically connected if and only if k is algebraically closed in O.V /, which is certainly the case if there exists a k-algebra homomorphism O.V / ! k (AG 11.5). T HEOREM 12.16 The following four conditions on an algebraic group G are equivalent: (a) (b) (c) (d) G is connected; the topological space specm.O.G// is connected; the topological space specm.O.G// is irreducible; the ring O.G/=N is an integral domain.
P ROOF. (b))(a). Remark 12.3 implies that 0 O.G/ has no nontrivial idempotents, and so is a eld. The existence of the k-algebra homomorphism W O.G/ ! k implies that 0 .O.G// D k. (c))(b). Trivial. (d),(c). In general, specm A is irreducible if and only if the nilradical of A is prime (see 5d). (a))(d). If G is connected, so also is Gk al . Write specm O.Gk al / as a union of its irreducible components. No irreducible component is contained in the union of the remainder. Therefore, there exists a point that lies on exactly one irreducible component. By homogeneity (5.5), all points have this property, and so the irreducible components are disjoint. As specm O.Gk al / is connected, there must be only one, and so Gk al is irreducible. Let N0 be the nilradical of O.Gk al / D k al k O.G/ we have shown that O.Gk al /=N0 is an integral domain. As the canonical map O.G/=N ! O.Gk al /=N0 is injective, we obtain (d). 2
126 P ROPOSITION 12.17 Let
1!N !G!Q!1 be an exact sequence of algebraic groups. If N and Q are connected, so also is G; conversely, if G is connected, so also is Q. P ROOF. Assume N and Q are connected. Then N is contained in the kernel of G ! 0 .G/, so this map factors through G ! Q (see 6.46), and therefore has image f1g. Conversely, since G maps onto 0 .Q/, it must be trivial if G is connected. 2
12e
Examples
E XAMPLE 12.18 The groups Ga , GLn , Tn (upper triangular), Un (strictly upper triangular), Dn are connected because in each case O.G/ is an integral domain. For example, kTn D kGLn =.Xij j i > j /; which is isomorphic to the polynomial ring in the symbols Xij , 1 i j n, with the product X11 Xnn inverted. E XAMPLE 12.19 For the group G of monomial matrices (2.23), 0 .O.G// is a product of copies of k indexed by the elements of Sn . Thus, 0 G D Sn (regarded as a constant algebraic group (4.13)), and G D Dn . E XAMPLE 12.20 The group SLn is connected. As we noted in the proof of (6.27), the natural isomorphism A; r 7! A diag.r; 1; : : : ; 1/W SLn .R/ Gm .R/ ! GLn .R/ (of set-valued functors) denes an isomorphism of k-algebras O.GLn / ' O.SLn / O.Gm /; and the algebra on the right contains O.SLn /. In particular, O.SLn / is a subring of O.GLn /, and so is an integral domain. E XAMPLE 12.21 Assume char.k/ 2. For any nondegenerate quadratic space .V; q/, the algebraic group SO.q/ is connected. It sufces to prove this after replacing k with k al , and so we may suppose 2 2 that q is the standard quadratic form X1 C C Xn , in which case we write SO.q/ D SOn . The latter is shown to be connected in the exercise below. The determinant denes a quotient map O.q/ ! f1g with kernel SO.q/. Therefore O.q/ D SO.q/ and 0 .O.q// D f1g (constant algebraic group). E XAMPLE 12.22 The symplectic group Sp2n is connected (for some hints on how to prove this, see Springer 1998, 2.2.9).
A SIDE 12.23 According to (12.4) and (12.16), an algebraic group G over C is connected if and only if G.C/ is connected for the complex topology. Thus, we could for example deduce that GLn is a connected algebraic group from knowing that GLn .C/ is connected for the complex topology. However, it is easier to deduce that GLn .C/ is connected from knowing that GLn is connected (of course, this requires the serious theorem stated in (12.4)).
127
12.24 For an algebraic group G over R, G may be connected without G.R/ being connected, and conversely. For example, GL2 is connected as an algebraic group, but GL2 .R/ is not connected for the real topology, and 3 is not connected as an algebraic group, but 3 .R/ D f1g is certainly connected for the real topology.
12f
Afne groups
For an afne group G, we write G D lim Gi where .Gi /i 2I is the family of algebraic quotients i 2I of G (see 7.21), and we dene G D lim
0G i 2I i 2I
Gi ;
0 Gi :
D lim
12g
Exercises
E XERCISE 12-1 What is the map O.SLn / ! O.GLn / dened in example 12.20? E XERCISE 12-2 Prove directly that
0 .O.On //
Dk
k.
E XERCISE 12-3 (Springer 1998, 2.2.2). Assume k has characteristic 2. For any k-algebra R, let V .R/ be the set of skew-symmetric matrices, i.e., the matrices A such that At D A. (a) Show that the functor R 7! V .R/ is represented by a nitely generated k-algebra A, and that A is an integral domain. (b) Show that A 7! .In C A/ 1 .In A/ denes a bijection from a nonempty open subset of SOn .k al / onto an open subset of V .k al /. (c) Deduce that SOn is connected. E XERCISE 12-4 Let A be a product of copies of k indexed by the elements of a nite set S . Show that the k-bialgebra structures on A are in natural one-to-one correspondence with the group structures on S. E XERCISE 12-5 For a nite algebraic group G, show that the following conditions are equivalent: (a) O.Gred / is separable; (b) Gred is an algebraic subgroup of GI (c) G is isomorphic to the semi-direct product of G and
0 G.
E XERCISE 12-6 Let k be a nonperfect eld of characteristic 2, so that there exists an a 2 k that is def not a square. Show that the functor R G.R/ D fx 2 R j x 4 D ax 2 g becomes a nite commutative algebraic group under addition. Show that G.k/ has only one element but 0 .G/ has two. Deduce that G is not isomorphic to the semi-direct product of G and 0 .G/. (Hence 12-5 shows that O.G/=N is not a Hopf algebra.) E XERCISE 12-7 Let k be a eld of characteristic p. Show that the extensions 0!
p
! G ! Z=pZ ! 0
with G a nite commutative algebraic group are classied by the elements of k =k p (the split extension G D p Z=pZ corresponds to the trivial element in k =k p ). Show that Gred is not a subgroup of G unless the extension splits
128
12h
Where we are
As discussed in the rst lecture, every afne algebraic group has a composition series with the quotients listed at right: afne G j nite etale connected G j semisimple
solvable
j
unipotent
torus
j unipotent f1g We have constructed the top segment of this picture. Next we look at tori and unipotent groups. Then we study the most interesting groups, the semisimple ones, and nally, we put everything together.
13
Groups of multiplicative type; tori
In this section we study the algebraic groups that become diagonalizable over an algebraically closed extension eld. We state for reference: Gm .R/ D R O.Gm / D kX; X .X / D X X .X / D 1 S.X / D X 1
1
D f 2 R j n D 1g O. n / D kX =.X n 1/ D kx .x/ D x x .x/ D 1 S.x/ D x n 1

n .R/
13a
Group-like elements
.a/ D a a and
D EFINITION 13.1 An element a of a k-coalgebra .A; ; / is group-like if .a/ D 1. Let A be a Hopf algebra. If a is a group-like element in A, then 1 D .e / .a/ D .mult .S idA / /.a/ D S.a/a, and so a is a unit in A with a
1 (31)
D S.a/. Conversely, if a is a unit in A such that a D .. ; idA / /.a/ D .a/a;

(27)
.a/ D a a, then
and so .a/ D 1; hence a is group-like. Thus the group-like elements of A are exactly the units such that .a/ D a a. Let A be a bialgebra. Because is a k-algebra homomorphism, .ab/ D .c
1
.a/ .b/ D .a a/.b b/ D ab ab .c/

1
/D
D .c c/
Dc
when a; b; c are group-like elements in A and c is a unit. Therefore the group-like elements form a submonoid of .A; /, and a subgroup of A if A is a Hopf algebra.
13b
The characters of an algebraic group
D EFINITION 13.2 A character of an algebraic group G is a homomorphism G ! Gm . P ROPOSITION 13.3 There is a canonical one-to-one correspondence between the characters of G and the group-like elements of O.G/. P ROOF. According to (4.8), characters of G correspond to homomorphisms of k-algebras O.Gm / ! O.G/ respecting . To give a homomorphism of k-algebras O.Gm / ! O.G/ amounts to giving a unit a in O.G/ (the image of X), and the homomorphism respects if and only if a is group-like.2 For characters ; 0 , dene C by . C
0 0 0
W G.R/ ! R
0
/.g/ D .g/
.g/:
Then C is again a character, and the set of characters is an abelian group, denoted X.G/. The correspondence in the proposition is an isomorphism of groups. 129
130
13c
The algebraic group D.M /
Let M be a nitely generated abelian group (written multiplicatively), and let kM be the k-vector space with basis M . Thus, the elements of kM are nite sums P
i
ai mi ;
ai 2 k;
mi 2 M;
and kM becomes a k-algebra (called the group algebra of M ) when we set P

i ai mi
P bj nj D i;j ai bj mi nj : j
It becomes a Hopf algebra when we set .m/ D m m; .m/ D 1; S.m/ D m

1
.m 2 M /
because, for m an element of the basis M , .id /. .m// D m .m m/ D .m m/ m D . id/. .m//, . id/. .m// D 1 m; .id /. .m// D m 1; .mult .S id//.m m/ D 1 D .mult .id S //.m m/: Note that kM is generated as a k-algebra by any set of generators for M , and so it is nitely generated. E XAMPLE 13.4 Let M be a cyclic group, generated by e. P (a) Case e has innite order. Then the elements of kM are the nite sums i 2Z ai e i with the obvious addition and multiplication, and .e/ D e e, .e/ D 1, S.e/ D e 1 . Therefore, kM ' kGm . (b) Case e is of order n. Then the elements of kM are sums a0 C a1 e C C an 1 e n 1 with the obvious addition and multiplication (using e n D 1), and .e/ D e e, .e/ D 1, and S.e/ D e n 1 . Therefore, kM ' k n . E XAMPLE 13.5 If W and V are vector spaces with bases .ei /i 2I and .fj /j 2J , then W k V is a vector space with basis .ei fj /.i;j /2I J . This shows that if M D M1 M2 , then .m1 ; m2 / $ m1 m2 W kM $ kM1 k kM2 is an isomorphism of k-vector spaces, and one checks easily that it respects the Hopf k-algebra structures. P ROPOSITION 13.6 For any nitely generated abelian M , the functor D.M / R Hom.M; R / (homomorphisms of abelian groups)
is an algebraic group, with coordinate ring kM . The choice of a basis for M determines an isomorphism of D.M / with a nite product of copies of Gm and various n s.
13. Groups of multiplicative type; tori
131
P ROOF. To give a k-linear map kM ! R is the same as giving a map M ! R. The map kM ! R is a k-algebra homomorphism if and only if M ! R has image in R and is a homomorphism M ! R . This shows that D.M / is represented by kM , and it is therefore an algebraic group. A decomposition of abelian groups M Z Z Z=n1 Z Z=nr Z;
denes a decomposition of k-bialgebras kM kGm k k kGm k k

n1 k
k k
nr
(13.4,13.5). Since every nitely generated abelian group M has such a decomposition, this proves the second statement. 2 R EMARK 13.7 The group D.M / is smooth except when k has characteristic p 0 and M has p-torsion (because Gm is smooth, and n is etale unless pjn).
13d
Characterizing the groups D.M /
L EMMA 13.8 The group-like elements in any k-coalgebra A are linearly independent. P ROOF. If not, it will be possible to express one group-like element e as a linear combination of other group-like elements ei e: P e D i ci ei , ci 2 k: We may even suppose that the ei are linearly independent. Now P .e/ D e e D i;j ci cj ei ej P P .e/ D i ci .ei / D i ci ei ei : The ei ej are also linearly independent, and so this implies that ci cj D We also know that .e/ D 1 P P .e/ D ci .ei / D ci : On combining these statements, we see that the ci form a complete set of orthogonal idempotents in the eld k, and so one of them equals 1 and the remainder are zero, which contradicts our assumption that e is not equal to any of the ei . 2 L EMMA 13.9 The group-like elements of kM are exactly the elements of M . P ROOF. Let a 2 kM be group-like. Then P a D ci mi for some ci 2 k, mi 2 M: The argument in the above proof shows that the ci form a complete set of orthogonal idempotents in k, and so a D mi for some i . 2 ci 0 if i D j otherwise.
132 Thus
X.D.M // ' M: The character of D.M / corresponding to m 2 M is D.M /.R/ D Hom.M; R /

def
f 7!f .m/
! R D Gm .R/:
def
P ROPOSITION 13.10 An algebraic group G is isomorphic to D.M / for some M if and only if the group-like elements in O.G/ span it as a k-vector space. P ROOF. The group-like elements of kM span it by denition. Conversely, suppose the group-like elements M span O.G/. Then they form a basis for O.G/ (as a k-vector space), and so the inclusion M ,! O.G/ extends to an isomorphism kM ! O.G/ of vector spaces. That this isomorphism is compatible with the bialgebra structures .m; e; ; / can be checked on the basis elements m 2 M , where it is obvious. 2
13e
Diagonalizable groups
D EFINITION 13.11 An algebraic group G is diagonalizable if O.G/ is spanned by group-like elements. T HEOREM 13.12 (a) The functor M D.M / is a contravariant equivalence from the category of nitely generated abelian groups to the category of diagonalizable algebraic groups (with quasiinverse G 7! X.G/). (b) If 1 ! M 0 ! M ! M 00 ! 1 is an exact sequence of nitely generated abelian groups, then 1 ! D.M 00 / ! D.M / ! D.M 0 / ! 1 is an exact sequence of algebraic groups. (c) Subgroups and quotient groups of diagonalizable algebraic groups are diagonalizable. P ROOF. (a) Certainly, we have a contravariant functor DW fnitely generated abelian groupsg We rst show that D is fully faithful, i.e., that Hom.M; M 0 / ! Hom.D.M 0 /; D.M // (88) fdiagonalizable groupsg:
is an isomorphism for all M; M 0 . As D sends direct sums to products, it sufces to do this when M; M 0 are cyclic. If, for example, M and M 0 are both innite cyclic groups, then Hom.M; M 0 / D Hom.Z; Z/ ' Z; Hom.D.M 0 /; D.M // D Hom.Gm ; Gm / D fX i j i 2 Zg ' Z; and (88) is an isomorphism. The remaining cases are similarly easy. Proposition 13.10 shows that the functor is essentially surjective, and so it is an equivalence.
133
(b) The map kM 0 ! kM is injective, and so D.M / ! D.M 0 / is a quotient map (by denition). Its kernel is represented by kM =IkM 0 , where IkM 0 is the augmentation ideal of kM 0 (see 6.13). But IkM 0 is the ideal generated the elements m 1 for m 2 M 0 , and so kM =IkM 0 is the quotient ring obtained by putting m D 1 for all m 2 M 0 . Therefore M ! M 00 denes an isomorphism kM =IkM 0 ! kM 00 . (c) If H is an algebraic subgroup of G, then O.G/ ! O.H / is surjective, and so if the grouplike elements of O.G/ span it, the same is true of O.H /. Let D.M / ! Q be a quotient map, and let H be its kernel. Then H D D.M 00 / for some quotient M 00 of M . Let M 0 be the kernel of M ! M 00 . Then D.M / ! D.M 0 / and D.M / ! Q are quotient maps with the same kernel, and so are isomorphic (6.47). 2
13f
Diagonalizable groups are diagonalizable
Recall that Dn is the group of invertible diagonal n n matrices; thus Dn ' Gm Gm ' D.Zn /:
n copies
T HEOREM 13.13 Let V be a nite-dimensional vector space, and let G be an algebraic subgroup of GLV . There exists a basis of V for which G Dn if and only if G is diagonalizable. In more down-to-earth terms, the theorem says that an algebraic subgroup G of GLn is diagonalizable (in the sense of denition 13.11) if and only if there exists an invertible matrix P in Mn .k/ such that, for all k-algebras R and all g 2 G.R/, 80 19 0 > < = B C :: P gP 1 2 @ A : : > : ; 0 P ROOF. ): As Dn is diagonalizable, so also is any subgroup of it (13.12c). (W Let W V ! V k O.G/ be the comodule corresponding to the representation G ,! GLV (see 7.12). We have to show that V is a direct sum of one-dimensional representations or, equivalently, that there exists a basis for V consisting of vectors u such that .u/ 2 hui k O.G/. The family .ei /i 2I of group-like elements in O.G/ is a basis. Let v 2 V , and write P .v/ D i ui ei : On applying the identities (p.76) .idV / .idV / to v, we nd that P
i
D . idA / D idV :
ui e i e i D vD
X
i
.ui / ei
ui :
The rst equality shows that .ui / D ui ei 2 hui i k A; and the second shows that the set of ui s arising in this way span V (and hence include a basis).
2
134
13g
Split tori and their representations
D EFINITION 13.14 An algebraic group is a split torus if it is isomorphic to a product of copies of Gm , and it is a torus if it becomes a split torus over k al . In other words, the split tori are the diagonalizable groups D.M / with M torsion-free. The functor M 7! D.M / is a contravariant equivalence from the category of free abelian groups of nite rank to the category of split tori, with quasi-inverse T 7! X.T /. For example, let T D Gm Gm . Then X.T / ' Z Z, and the character corresponding to .m1 ; m2 / 2 Z Z is m m .t1 ; t2 / 7! t1 1 t2 2 W T .R/ ! Gm .R/. A quotient group of a torus is again a torus (because it corresponds to a subgroup of a free abelian group of nite rank), but a subgroup of a torus need not be a torus. For example, n is a subgroup of Gm (the map n ! Gm corresponds to Z ! Z=nZ). A character W T ! Gm denes a representation of T on any nite-dimensional space V : let t 2 T .R/ act on R k V as multiplication by .t / 2 R . For example, denes a representation of T on k n by 0 1 .t / 0 B C :: t 7! @ A: : 0 .t / Let W T ! GLV be a representation of T . We say that T acts on V through if .t /v D .t /v all t 2 T .R/, v 2 R k V: More precisely, this means that the image of composite of is contained in the centre Gm of GLV and is the
T ! Gm ,! GLV : If V is 1-dimensional, then GLV D Gm , and so T always acts on V through some character. T HEOREM 13.15 Let rW T ! GL .V / be a representation of a split torus on a nite dimensional vector space V . For each character , let V be the largest subspace of V on which T acts through the character . Then M V : V D
2X.T /
P P ROOF. Theorem 13.13 shows that V D 2X.T / V . In proving that the sum is direct, we may replace k by its algebraic closure. Let 1 ; : : : ; m be distinct characters of T such that V i is nonzero, and consider V 1 C C V m . If the sum is not direct, then there exists a relation v1 C C vm D 0, vi 2 V i , vi 0. (89)
On applying t 2 T .k/ to (89), we get a new relation

1 .t /v1 C
m 1 .t /vm 1 C
m .t /vm
D 0:
(90)
As m m 1 and T is smooth, there exists a t 2 G.k/ such that m .t / m 1 .t /. Multiply (90) by m .t/ 1 and subtract it from (89). This will give us a new relation of the same form but with fewer terms. Continuing in this fashion, we arrive at a contradiction. 2
135
For example, let T D Gm Gm , and let rW T ! GL.V / be a representation of T on a nitedimensional vector space V . Then V decomposes into a direct sum of subspaces V.m1 ;m2 / , .m1 ; m2 / 2 m m Z Z, such that .t1 ; t2 / 2 T .k/ acts on V.m1 ;m2 / as t1 1 t2 2 (of course, all but a nite number of the subspaces V.m1 ;m2 / are zero).
13h
Groups of multiplicative type
D EFINITION 13.16 An algebraic group G is of multiplicative type if Gk sep is diagonalizable. Let M be a nitely generated abelian group, and let D Gal.k sep =k/. A continuous action of on M is a homomorphism ! Aut.M / factoring through Gal.K=k/ for some nite Galois extension K of k contained in k al . For an algebraic group G, we dene X .G/ D Hom.Gk sep ; Gm /: Then acts continuously on X .G/, because X .G/ is nitely generated, and each of its generators is dened over a nite extension of k. T HEOREM 13.17 The functor G X .G/ is a contravariant equivalence from the category of algebraic groups of multiplicative type over k to the category of nitely generated abelian groups with a continuous action of . P ROOF. To give a continuous action of on M is the same as giving a continuous action of on M , and so this follows from Theorem 13.12 and Proposition 4.16. Let G be a group of multiplicative type over k. For any K G.K/ D Hom.X .G/; k sep / k sep ,
K
where K is the subgroup of of elements xing K, and the notation means the G.K/ equals the group of homomorphisms X .G/ ! k sep commuting with the actions of K . E XAMPLE 13.18 Take k D R, so that is cyclic of order 2, and let X .G/ D Z. Then Aut.Z/ D Z D f1g, and so there are two possible actions of on X .G/: (a) Trivial action. Then G.R/ D R , and G ' Gm . (b) The generator of acts on Z as m 7! m. Then G.R/ D Hom.Z; C / elements of C xed under the following action of , z D z N
1
consists of the
Thus G.R/ D fz 2 C j z z D 1g, which is compact. N E XAMPLE 13.19 Let K be a nite extension of k. Let T be the functor R 7! .R k K/ . Then T is an algebraic group, in fact, the group of multiplicative type corresponding to the -module al ZHomk .K;k / (families of elements of Z indexed by the k-homomorphisms K ! k al ).
A SIDE 13.20 Need to add a proof that if G becomes diagonalizable over some eld containing k, then it becomes diagonalizable over a separable extension (see Waterhouse 1979, Chapter 7). Then can replace k al with k sep everywhere.
136
13i
Rigidity
Later we shall need the following result. T HEOREM 13.21 Every action of a connected algebraic group G on a group H of multiplicative type is trivial. P ROOF. We may suppose that k is algebraically closed. When H D n , an action of G on M denes a map G ! Aut.
n/
Hom.
n;
n/
' Hom.
n ; Gm /
' Z=nZ
(see 11c), which is trivial, because G is connected. A similar argument proves the theorem when H is nite (hence a nite product of groups of the form n ). We next prove the theorem when G.k/ is dense in G (e.g., if G is smooth). We may suppose that H is a torus T . The kernel of x 7! x m W T ! T is a product of copies of m , and so G acts trivially on it. Because of the category equivalence T X.T /, it sufces to show that g 2 G.k/ acts trivially on the X.T /, and because g acts trivially on the kernel of mW T ! T it acts trivially on X.T /=mX.T /. We can now apply the following elementary lemma. 2 L EMMA 13.22 Let M be a free abelian group of nite rank, and let W M ! M be a homomorphism such that M ! M ? ? ? ? y y M=mM commutes for all m. Then D id. P P ROOF. Choose a basis ei for M , and write .ej / D i aij ei , aij 2 Z. The hypothesis is that, for every integer m, .aij / In mod m; i.e., that mjaij for i j and mjai i 1. Clearly, this implies that .aij / D In .
2 id
! M=mM
P ROPOSITION 13.23 Let T be a torus, and let TN be the kernel of N W T ! T . Let W T ! T be a homomorphism whose restriction to TN is the identity map. Then is the identity map. P ROOF. It sufces to show that X ./W X .T / ! X .T / is the identity map, but the hypothesis says that X ./ induces the identity map on the quotient X .T /=NX .T / D X .TN / for all N , S and so this follows from the lemma.(Add that TN is dense in T for every algebraic group T of multiplicative type; for 6.39.) 2
A SIDE 13.24 Need to add the Hopf algebra proof, which is simpler and doesnt require G to be smooth (Waterhouse 1979, 7.7).
137
13j
Exercises
E XERCISE 13-1 Show that Aut. m / ' .Z=mZ/ (constant group dened by the group of invertible elements in the ring Z=mZ). Hint: To recognize the elements of Aut. m /.R/ as complete systems of orthogonal idempotents, see the proof of (13.8). E XERCISE 13-2 Let k 0 =k be a cyclic Galois extension of degree n with Galois group by , and let G D .Gm /k 0 =k . (a) Show that X .G/ ' Z (group algebra Z C Z C C Z n 1 of ). (b) Show that 80 9 1 a1 a2 : : : an > > B > < an a1 : : : an C = B C End .X .G// D B : ai 2 Z : : : C : : A @ : > : : : > > : ; a2 a3 a1 generated
14
Solvable algebraic groups

1
Let G be an abstract group. Recall that the commutator of x; y 2 G is x; y D xyx y

1
D .xy/.yx/
Thus, x; y D 1 if and only if xy D yx, and G is commutative if and only if every commutator equals 1. The (rst) derived group G 0 (or DG) of G is the subgroup generated by commutators. Every automorphism of G maps commutators to commutators, and so G 0 is a characteristic subgroup of G (in particular, it is normal). In fact, it is the smallest normal subgroup such that G=G 0 is commutative. The map (not a group homomorphism) .x1 ; y1 ; : : : ; xn ; yn / 7! x1 ; y1 xn ; yn W G 2n ! G
has image the set of elements of G that can be written as a product of at most n commutators, and so DG is the union of the images of these maps. Note that the map G 2n 2 ! G factors through G 2n ! G, .x1 ; y1 ; : : : ; xn
1 ; yn 1 /
7! .x1 ; y1 ; : : : ; xn
1 ; yn 1 ; 1; 1/
7! x1 ; y1
xn
1 ; yn 1 :
A group G is said to be solvable if the derived series G DG D2 G
terminates with 1. For example, if n 5, then Sn (symmetric group on n letters) is not solvable because its derived series Sn An terminates with An . In this section we extend this theory to algebraic groups. Throughout, k is a eld.
14a
Commutative groups are triangulizable
We rst prove a result in linear algebra. L EMMA 14.1 Let V be a nite-dimensional vector space over an algebraically closed eld k, and let S be a set of commuting endomorphisms of V . There exists a basis of V for which S is contained in the group of upper triangular matrices, i.e., a basis e1 ; : : : ; en such that .he1 ; : : : ; ei i/ he1 ; : : : ; ei i for all i: (91)
In more down-to-earth terms, let S be a set of commuting n n matrices; then there exists an invertible matrix P such that PAP 1 is upper triangular for all A 2 S . P ROOF. We prove this by induction on the dimension of V . If every 2 S is a scalar multiple of the identity map, then there is nothing to prove. Otherwise, there exists an 2 S and an eigenvalue a for such that the eigenspace Va V . Because every element of S commutes with , Va is stable under the action of the elements of S : for 2 S and x 2 Va , .x/ D .x/ D .ax/ D a.x/: The induction hypothesis applied to S acting on Va and V =Va shows that there exist bases e1 ; : : : ; em for Va and emC1 ; : : : ; en for V =Va such that N N .he1 ; : : : ; ei i/ .hemC1 ; : : : ; emCi i/ N N he1 ; : : : ; ei i for all i m m:
2
hemC1 ; : : : ; emCi i for all i n N N
Let emCi D emCi C Va with emCi 2 V . Then e1 ; : : : ; en is a basis for V satisfying (91): N 138
14. Solvable algebraic groups
139
P ROPOSITION 14.2 Let V be a nite-dimensional vector space over an algebraically closed eld k, and let G be a commutative smooth algebraic subgroup of GLV . There exists a basis of V for which G is contained in Tn . P ROOF. According to the lemma, there exists a basis for V such that the map GL.V / ! GLn .k/ dened by it sends G.k/ into Tn .k/: G.k/ GL.V /
Tn .k/
GLn .k/
Now G \ Tn is an algebraic subgroup of G such that .G \ Tn /.k/ D G.k/, and so G \ Tn D G (see 6d). Therefore G Tn . 2
14b
Decomposition of a commutative algebraic group
D EFINITION 14.3 Let G be an algebraic group over a perfect eld k. An element g of G.k/ is semisimple (resp. unipotent) if g D gs (resp. g D gu ). Thus, g is semisimple (resp. unipotent) if and only if .g/ is semisimple (resp. unipotent) for all representations of G. Theorem 9.18 shows that G.k/ D G.k/s G.k/u (cartesian product of sets) (92)
where G.k/s (resp. G.k/u ) is the set of semisimple (resp. unipotent) elements in G.k/. However, this will not in general be a decomposition of groups, because Jordan decompositions do not respect products, for example, .gh/u gu hu in general. However, if G is commutative, then G G
multiplication
!G
is a homomorphism of groups, and so it does respect the Jordan decompositions (9.20). Thus, in this case (92) realizes G.k/ as a product of subgroups. We can do better. T HEOREM 14.4 Every commutative smooth algebraic group G over an algebraically closed eld is a direct product of two algebraic subgroups G ' Gs such that Gs .k/ D G.k/s and Gu .k/ D G.k/u . P ROOF. First note that the subgroups Dn and Un of Tn have trivial intersection, because Dn .R/ \ Un .R/ D fIn g for all R. (inside Tn .R/) Gu
140
Choose an embedding G ,! Tn for some n, and let Gs D G \ Dn and Gu D G \ Un . Because G is commutative, Gs Gu ! G (93)
is a homomorphism with kernel Gs \ Gu (cf. 6). Because Dn \ Un D f1g as algebraic groups, Gs \ Gu D f1g, and so (93) is injective; because Gs .k/Gu .k/ D G.k/ and G is smooth, (93) is surjective (6.41); therefore it is an isomorphism. 2 R EMARK 14.5 Let G be a smooth algebraic group over an algebraically closed eld k. In general, G.k/s will not be closed for the Zariski topology. However, G.k/u is closed. To see this, embed G in GLn for some n. A matrix A is unipotent if and only if 1 is its only eigenvalue, i.e., if and only if its characteristic polynomial is .T 1/n . But the coefcients of the characteristic polynomial of A are polynomials in the entries of A, and so this is a polynomial condition.
A SIDE 14.6 In fact every commutative afne algebraic group over a perfect eld decomposes into a product of a group of multiplicative type and a unipotent group (Waterhouse 1979, 9.5)
14c
The derived group of algebraic group
Let G be an algebraic group over a eld k. D EFINITION 14.7 The derived group DG (or G 0 or G der ) of G is the intersection of the normal algebraic subgroups N of G such that G=N is commutative. P ROPOSITION 14.8 The quotient G=DG is commutative (hence DG is the smallest normal subgroup with this property). P ROOF. Let N1 and N2 be normal subgroups of G. Then N1 .R/ and N2 .R/ are normal subgroups of G.R/ and .N1 \ N2 /.R/ D N1 .R/ \ N2 .R/. Therefore G.R/=.N1 \ N2 /.R/ ' G.R/=N1 .R/ G.R/=N2 .R/ (GT 1.50). On varying R and passing to the associated sheaves, we obtain an isomorphism G=N1 \ N2 ' G=N1 G=N2 :
Therefore, G=N1 \N2 is commutative if G=N1 and G=N2 are commutative. Similarly, for any nite T family .Ni /i 2I of G= Ni is commutative if each quotient G=Ni is commutative, and it sufces to consider nite families by (6.8). 2 We shall need another description of DG, which is analogous to the description of the derived group as the subgroup generated by commutators. As for abstract groups, there exist maps of functors G 2 ! G 4 ! ! G 2n ! G: Let In be the kernel of the homomorphism O.G/ ! O.G 2n / of k-algebras (not Hopf algebras) dened by G 2n ! G: Then I1 I2 In T and we let I D In .
14. Solvable algebraic groups P ROPOSITION 14.9 The coordinate ring of DG is O.G/=I . P ROOF. From the diagram of set-valued functors G 2n ? ? y G we get a diagram of k-algebras O.G/=In O.G/=In x x ? ? ? ? O.G/ O.G/ O.G/=I2n x ? ? O.G/ G 2n ? ? y G !
mult
141
G 4n ? ? y G
(because O.G/=In is the image of O.G/ in O.G 4n / ). It follows that W O.G/ ! O.G/=I O.G/=I factors through O.G/ ! O.G/=I , and denes a Hopf algebra structure on O.G/=I , which corresponds to the smallest algebraic subgroup G 0 of G such that G 0 .R/ contains all the commutators for all R. Clearly, this is also the smallest normal subgroup such that G=G 0 is commutative. 2 C OROLLARY 14.10 For any eld K k, DGK D .DG/K :
2
P ROOF. The denition of I commutes with extension of the base eld. C OROLLARY 14.11 If G is connected (resp. smooth), then DG is connected (resp. smooth).
P ROOF. Recall that an algebraic group G is connected (resp. smooth) if and only if O.G/ has no nontrivial idempotents (resp. nilpotents). If O.G/=I had a nontrivial idempotent (resp. nilpotent), then so would O.G/=In for some n, but (by denition) the homomorphism of k-algebras O.G/=In ,! O.G 2n / is injective. If G is connected (resp. smooth), then so also is G 2n , and so O.G 2n / has no nontrivial idempotents (resp. nilpotents). 2 C OROLLARY 14.12 Let G be a smooth connected algebraic group. Then O.DG/ D O.G/=In for some n, and .DG/.k al / D D.G.k al //. P ROOF. As G is smooth and connected, so also is G 2n (5.1, 12.12). Therefore, each ideal In is prime, and a descending sequence of prime ideals in a noetherian ring terminates. This proves the rst part of the statement (CA 16.5). Let Vn be the image of G 2n .k al / in G.k al /. Its closure in G.k al / is the zero-set of In . Being the image of a regular map, Vn contains a dense open subset U of its closure (CA 12.13). Choose n as in the rst part, so that the zero-set of In is DG.k al /. Then [ U U 1 Vn Vn V2n D.G.k al // D Vm DG.k al /:
m
It remains to show that U U D Let g 2 DG.k al /. Because U is open and dense DG.k al /, so is gU 1 , which must therefore meet U , forcing g to lie in U U . 2
DG.k al /.
142
14.13 Let G D PGLn . Then DG D G but D.G.k// G.k/ when k is a nonperfect separably closed eld of characteristic p dividing n.
14d
Solvable algebraic groups
Write D2 G for the second derived group D.DG/, D3 G for the third derived group D.D2 G/, and so on. D EFINITION 14.14 An algebraic group G is solvable if the derived series G terminates with 1. L EMMA 14.15 An algebraic group G is solvable if and only if it admits a sequence of algebraic subgroups G D G0 G1 Gn D f1g (94) DG D2 G
with Gi C1 normal in Gi for each i , and Gi =Gi C1 commutative. P ROOF. If G is solvable, then the derived series is such a sequence. Conversely, given a sequence as in (94), G1 DG, so G2 D2 G, . . . , so Gn Dn G. 2 A sequence of algebraic subgroups (94) such that Gi C1 is normal in Gi for each i and Gi =Gi C1 is commutative is called solvable series. E XAMPLE 14.16 Let G be a nite group, and let .G/k be the algebraic group such that .G/k .R/ D G for any k-algebra R with no nontrivial idempotents. Then D.G/k D .DG/k , D2 .G/k D .D2 G/k , and so on. Therefore .G/k is solvable if and only if G is solvable. In particular, the theory of solvable algebraic groups includes the theory of solvable nite groups, which is already quite complicated. E XAMPLE 14.17 The group Tn of upper triangular matrices is solvable. For example, 1 10 f. 0 /g 0 1 01 and n
0 0 0
1 0 1 0 0 1
10 01 0 00 1
100 010 001
o Gm and Ga , and
demonstrate that T2 and T3 are solvable. In the rst case, the quotients are Gm in the second the quotients are Gm Gm Gm , Ga Ga , and Ga . More generally, the functor R G0 .R/ D f.aij / j ai i D 1 for all i g
def
is an algebraic subgroup of Tn because it is represented by O.Tn /=.T11 1; : : : ; Tnn 1/. Similarly, there is an algebraic subgroup Gr of G0 of matrices .aij / such that aij D 0 for 0 < j i r. The functor .aij / 7! .a1;rC2 ; : : : ; ai;rCi C1 ; : : :/
143
is a homomorphism from Gr onto Ga Ga with kernel GrC1 . Thus the sequence of algebraic subgroups Tn G0 G1 Gn D f1g exhibits Tn as a solvable group. Alternatively, we can work abstractly. A full ag F in a vector space V of dimension n is a sequence of subspaces V D Vn Vi Vi
1
V1
f0g
with Vi of dimension i . Let T be the algebraic subgroup of GLV such that T.R/ consists of the automorphisms preserving the ag, i.e., such that .Vi R/ Vi R for all k-algebras R. When we take F to be the obvious ag in k n , G D Tn . Let G0 be the algebraic subgroup of G of acting as id on the quotients Vi =Vi i ; more precisely, Y G0 D Ker.G ! GLVi =Vi i /: Then G0 is a normal algebraic subgroup of T with quotient isomorphic to Gn . Now dene Gr to be m the algebraic subgroup of G0 of elements acting as id on the quotients Vi =Vi r 1 : Again, GrC1 is a normal algebraic subgroup of Gr with quotient isomorphic to a product of copies of Ga . E XAMPLE 14.18 The group of n n monomial matrices is solvable if and only if n 4 (because Sn is solvable if and only if n 4; GT 4.33).
14e
Independence of characters
Let Gm be the subgroup of GLn of scalar matrices, i.e., it is the subgroup dened by the equations Xij D 0 for i j I X11 D X22 D D Xnn : Then a 2 Gm .R/ D R acts on Rn as .x1 ; : : : ; xn / 7! .ax1 ; : : : ; axn /. Similarly, GLV contains a subgroup Gm such that a 2 Gm .R/ acts on R k V by the homothety v 7! av. Under the isomorphism GLV ! GLn dened by any basis of V , the Gm s correspond. In fact, Gm is the centre of GLV . Now let rW G ! GLV be a representation of G on V . If r factors through the centre Gm of GLV , G ! Gm
r
GLV
then r is a character of G, and we say that G acts on V through the character r (cf. p.134). More generally, we say that G acts on a subspace W of V through a character if W is stable under G and G acts on W through . Note that this means, in particular, that the elements of W are common eigenvectors for the g 2 G.k/: if w 2 W , then for every g 2 G.k/, r.g/w is a scalar multiple of w. For this reason, we also call V an eigenspace for G with character . Let rW G ! GLV be a representation of G on V . If G acts on subspaces W and W 0 through a character , then it acts on W C W 0 through . Therefore, there is a largest subspace V of V on which G acts through . P ROPOSITION 14.19 For each character of G, let V be the largest subspace of V on which G acts through . If V is a sum of spaces V , then it is a direct sum. In other words, vectors lying in eigenspaces corresponding to different s are linearly independent.
144
P ROOF. As we saw in 13, characters of G correspond to group-like elements of O.G/. If $ a. /, then the coaction of O.G/ on V is given by .v/ D v a. /. P Suppose V D i 2I V i with the i (i 2 I ) distinct characters of G. If the sum is not direct, then there exists a relation P vi 2 V i ; vi 0; J a nite subset of I . (95) i 2J vi D 0; Then 0D P
i 2J
.vi / D
i 2J
vi a. i /
2
which contradicts the linear independence of the a. i / (see 13.8).
We saw in 13 that if G is a split torus, V is always a sum of the eigenspaces V . In general, this will be far from true. For example, SLn has no nontrivial characters.
A SIDE 14.20 This section duplicates part of 13g. It should be moved somewhere earlier.
14f
The Lie-Kolchin theorem
T HEOREM 14.21 Let G be an algebraic subgroup of GLV . If G is connected, smooth, and solvable, and k is algebraically closed, then there exists a basis for V such that G Tn . Before proving this, it will be useful to see that the hypotheses are really needed. solvable As Tn is solvable (14.17) and a subgroup of a solvable group is obviously solvable, this condition is necessary. k algebraically closed If G.k/ Tn .k/, then the elements of G.k/ have a common eigenvector, namely, e1 D .1 0 : : : 0/t . Unless k is algebraically closed, an endomorphism need not have an eigenvector, and, for example, ab a; b 2 R; a2 C b 2 D 1 b a is an commutative algebraic group over R that is not triangulizable over R. connected The group G of monomial 2 2 matrices is solvable but not triagonalizable. The only common eigenvectors of D2 .k/ G.k/ are e1 D 1 and e2 D 0 , but the monomial matrix 0 1 0 1 interchanges e and e , and so there is no common eigenvector for the elements of 1 2 10 G.k/. P ROOF. It sufces to show that there exists a basis for V such that G.k/ Tn .k/ (because then .G \ Tn / .k/ D G.k/, and so G \ Tn D G, which implies that G T). Also, it sufces to show that the elements of G.k/ have a common eigenvector, because then we can apply induction on the dimension of V (cf. the proof of 14.1). We prove this by induction on the length of the derived series G. If the derived series has length zero, then G is commutative, and we proved the result in (14.2). Let N D DG. Its derived series is shorter than that of G, and so we can assume that the elements of N have a common eigenvector, i.e., for some character of N , the space V (for N ) is nonzero. Let W be the sum of the nonzero eigenspaces V for N . According to (14.19), the sum is direct, M W D V and so the set fV g of nonzero eigenspaces for N is nite.
14. Solvable algebraic groups Let x be a nonzero element of V for some , and let g 2 G.k/. For n 2 N.k/, ngx D g.g
1
145
ng/x D g
.g
ng/x D .g
ng/ gx
For the middle equality we used that N is normal in G. Thus, gx lies in the eigenspace for the character 0 D .n 7! .g 1 ng// of N . This shows that G.k/ permutes the nite set fV g. Choose a and let H be the stabilizer of V in G.k/. Thus, H is a subgroup of nite index in G.k/. Moreover, it is closed for the Zariski topology on G.k/ because it is the set where the characters and 0 coincide. But every closed subgroup of nite index of a topological group is open30 , and so H is closed and open in G.k/. But G.k/ is connected for the Zariski topology (12.16), and so G.k/ D H . This shows that W D V , and so G.k/ stabilizes V . An element n 2 N.k/ acts on V as the homothety x 7! .n/x, .n/ 2 k: But each element n of N.k/ is a product of commutators x; y of elements of G.k/ (see 14.12), and so n acts on V as an automorphism of determinant 1. This shows that .n/dim V D 1, and so the image of W G ! Gm is nite. Because N is connected, this shows that N.k/ in fact acts trivially31 on V . Hence G.k/ acts on V through the quotient G.k/=N.k/, which is commutative. In this case, we know there is a common eigenvalue (14.1). 2
A SIDE 14.22 Is smoothness needed?
14g
Unipotent groups
We begin by proving two results from linear algebra. L EMMA 14.23 Let G ! End.W / be a simple representation of an abstract group G on a nitedimensional vector space W over an algebraically closed eld k. Let G act on End.W / by the rule: .gf /.w/ D g.f .w//; g 2 G; f 2 End.W /; w 2 W: Then every nonzero G-subspace X of End.W / contains an element f0 W W ! W such that f0 .W / has dimension one. P ROOF. We may suppose that X is simple. Then the k-algebra of G-endomorphisms of X is a division algebra (Schurs lemma, GT 7.23 or II, 5.3), and hence equals k (GT 7.28). For any w 2 W , the map 'w W X ! W sending f to f .w/ is a G-homomorphism. As both X and W are simple, it is either zero or an isomorphism. Since X 0, there exists a w0 2 W such that 'w0 is an isomorphism. Let f0 2 X be such that 'w0 .f0 / D w0 . 1 Let w 2 W . Then 'w0 'w is a G-endomorphism of X, and so 'w D c.w/'w0 for some c.w/ 2 k. On applying this equality to f0 , we nd that f0 .w/ D c.w/w0 , and so f0 .W / hw0 i. 2 Recall that an endomorphism of a vector space is unipotent if its characteristic polynomial is 1/dim V .
.T
P ROPOSITION 14.24 Let V be a nite-dimensional vector space, and let G be a subgroup of GL.V / consisting of unipotent endomorphisms. Then there exists a basis of V for which G is contained in Un (in particular, G is solvable).
it is the complement of nite set of cosets, each of which is also closed. more detail, the argument shows that the character takes values in m Gm where m D dim V . If k has characteristic zero, or characteristic p and p 6 jm, then m is etale, and so, because N is connected, is trivial. If pjm, the argument only shows that takes values in p r for p r the power of p dividing m. But p r .k/ D 1, and so the action of N.k/ on V is trivial, as claimed.
31 In 30 Because
146
P ROOF. It sufces to show that there exists a nonzero subspace of W on which G acts trivially, because then we can apply induction on the dimension of V to the vector spaces W and V =W to obtain a basis of V with the required property (cf. the proof of 14.1).P Choose a basis .ei /1i n for V . The condition that a vector v D ai ei is xed by all g 2 G is linear in the ai , and so has a solution in k n if and only if it has a solution in .k al /n . Therefore, we may suppose that k is algebraically closed. Let W be a nonzero subspace of V of minimal dimension among those stable under G. Clearly W is simple. For each g 2 G, TrW .g/ D dim W , and so TrW .g.g 0 1// D TrW .gg 0 / TrW .g/ D 0:
Let U D ff 2 End.W / j TrW .gf / D 0 for all g 2 Gg. Then .g 0 1/jW 2 U for all g 0 2 G, and so U is nonzero if G acts nontrivially on W . The lemma then shows that U contains an element f0 such that f0 .W / has dimension one. Such an f0 has TrW f0 0, which contradicts the fact that f0 2 U . We conclude that G acts trivially on W . 2 T HEOREM 14.25 The following conditions on an algebraic group G are equivalent: (a) every nonzero representation of G has a nonzero xed vector (i.e., a nonzero v 2 V such that .v/ D v 1 when V is regarded as a O.G/-comodule); (b) for every representation G ! End.V / of G, there exists a basis of V for which the image of G is contained in Un ; (c) G is isomorphic to an algebraic subgroup of Un for some n. When G.k/ is dense in G, the conditions are equivalent to: (d) all elements of G.k/ are unipotent. P ROOF. We rst prove the equivalence of (c) and (d) when G.k/ is dense in G. (c))(d). Obvious. (d))(c). Choose a faithful representation G ! End.V / of G. According to Proposition 14.24, there exists a basis of V for which G.k/ Udim V .k/. Because G.k/ is dense in G, this implies that G Udim V . We now prove the equivalence of (a), (b), and (c). (a))(b). Prove (b) by induction on the dimension of V . (b))(c). Apply (b) to a faithful representation of G. (c))(a). (Following Waterhouse 1979, 8.3.). A Hopf algebra A is said to be coconnected if it satises the following condition: there exists a ltration C0 C1 C2 of A by subspaces Ci such that [ X C0 D k, Cr D A, and .Cr / Ci Cr i : (96)
r 0 0i r
We rst show that (c) implies that O.G/ is coconnected, and then show that (a) holds for any algebraic group G such that O.G/ is coconnected. Any quotient of a coconnected Hopf algebra is coconnected (the image of a ltration satisfying (96) will still satisfy (96)), and so it sufces to show that O.Un / is coconnected. Recall that O.Un / D kXij j i < j , and that X .Xij / D Xij 1 C 1 Xij C Xi r Xrj :
i <r<j
147
Q n P Assign a weight of j i to Xij , so that a monomial Xijij will have weight Sij .j i /, and let n Cm be the subspace spanned by the monomials of weight m. Clearly, C0 D k, r 0 Cr D A, and Ci Cj Ci Cj . It sufces to check the third condition in (96) on the monomials. For the Xij it is obvious. We proceed by induction on weight of a monomial. If the condition holds for monomials P , Q of weights r, s, then .PQ/ D .P / .Q/ lies in X X X Cj Cr j Ci Cj Cr i Cs j Ci Cr i X Ci Cj CrCs i j . Now assume that A is a coconnected Hopf algebra, and let W V ! V A be a comodule. Then V is a union of the subspaces Vr D fv 2 V j .v/ 2 V Cr g. If V0 contains a nonzero vector v, then .v/ D v 0 1 for some vector v 0 ; on applying , we nd that v D v 0 , and so v is xed. We complete the proof by showing that Vr D 0 H) VrC1 D 0: By denition, .VrC1 / V CrC1 , and so .id / .VrC1 / V X
i
Ci Cr
i:
Hence VrCi maps to zero in V A=Cr A=Cr . We now use that .id / D . id/ . The map V ! V A=Cr dened by is injective because Vr D 0, and on applying id we nd that V ! .V A=Cr / A=Cr is injective. Hence VrC1 D 0. 2 D EFINITION 14.26 An algebraic group G is unipotent if it satises the equivalent conditions of (14.25). C OROLLARY 14.27 Every subgroup and quotient group of a unipotent group is unipotent. P ROOF. If G can be realized as a subgroup of Un , then so also can any subgroup of G. If every representation of G contains a nonzero xed vector, then the same is true of every quotient of G. 2 C OROLLARY 14.28 Let k 0 be a eld containing k. An algebraic group G over k is unipotent if and only if Gk 0 is unipotent. P ROOF. We have (see the proof of 14.25): G unipotent H) O.G/ coconnected H) k 0 O.G/ coconnected H) Gk 0 unipotent. Conversely, if Gk 0 satises (a), so also does G, because .v/ D v 1 is a linear condition on v (or, rather, it is a linear condition on the coefcients of v with respect to some basis for V ). 2 C OROLLARY 14.29 A smooth algebraic group G is unipotent if G.k al / consists of unipotent elements.
148

2
P ROOF. The condition implies that Gk al is unipotent, which implies that G itself is unipotent.
P ROPOSITION 14.30 An algebraic group that is both unipotent and of multiplicative type is trivial. P ROOF. We may suppose that the base eld is algebraically closed. A nontrivial algebraic group of multiplicative type contains a subgroup ` for some prime `, but it is easy to see that there is no nontrivial homomorphism ` ! Ga . 2
A A
14.31 A unipotent group need not be smooth. For example, in characteristic p, the subgroup of U2 consisting of matrices 1 a with ap D 0 is not smooth (it is isomorphic to p ). 0 1 14.32 For an algebraic group G, all elements of G.k al / can be unipotent without G being unipotent. For example, in characteristic p, the algebraic group p has p .k al / D 1, but it is not unipotent.
A SIDE 14.33 In DG IV, 2, 2.1, p.485, an algebraic group G is dened to be unipotent if, for every nontrivial subgroup H , there exists a nontrivial homomorphism H ! Ga . In SGA3 XVIII, 1.3, an algebraic group G is dened to be unipotent if admits a composition series over k al whose quotients are isomorphic to subgroups of Ga . Both these conditions are equivalent to the condition that G be isomorphic to an afne algebraic subgroup of Un for some n (DG IV, 2, 2.5, p.487; SGA3 XVII, 3.5(i),(v)). A SIDE 14.34 The unipotent algebraic groups over a eld of characteristic p 0 are complicated, but those over a eld of characteristic zero are classied by their Lie algebras; more precisely, over a eld k of characteristic zero, the functor G Lie.G/ is an equivalence from the category of unipotent algebraic groups over k to the category of nilpotent Lie algebras over k (see II 3.27).
14h
Structure of solvable groups
T HEOREM 14.35 Let G be a connected solvable smooth group over a perfect eld k. There exists a unique connected normal algebraic subgroup Gu of G such that (a) Gu is unipotent; (b) G=Gu is of multiplicative type. The formation of Gu commutes with change of the base eld. P ROOF. We rst prove this when k D k al . Embed G into Tn for some n, and construct 1 ! Un x ? ? ! Gu ! Tn x ? ? ! G ! Dn x ? ? ! T ! 1
! 1
where Gu D Un \ G and T is the image of G in Dn . Certainly Gu is a normal algebraic subgroup of G satisfying (a) and (b). We next prove that Gu is connected. Let Q D G=DG. It is commutative, so that (14.4) Q ' Qu Qs .
149
This shows that Qu is connected (if it had an etale quotient, so would Q). As G=Gu is commutative, DG Gu , and the diagram 1 ! DG ! Gu ? ? y ! G ? ? y T ? ? y 1 shows that T ' Q=
0 .Gu /.
0 .Gu /
! 1
? ? y ! Q ? ? y ! 1
! DG
! Q= Gu ? ? y 1
0 .Gu /
Since .Gu /
Qu , this shows that
D Qu , and so (12.17)
Qu , DG connected H) Gu connected. For the uniqueness, note that Gu is the largest connected normal unipotent subgroup of G, or that Gu .k al / consists of the unipotent elements of G.k al / (and apply a previous result). When k is only perfect, the uniqueness of .Gk al /u implies that it is stable under , and hence arises from a unique algebraic subgroup Gu of G (6.28), which clearly has the required properties.2
14i
Tori in solvable groups
P ROPOSITION 14.36 Let G be a connected smooth solvable group over an algebraically closed eld. If T and T 0 are maximal tori in G, then T 0 D gT g 1 for some g 2 G.k/. P ROOF. Omitted for the present (cf. Humphreys 1975, 19.2).
2
P ROPOSITION 14.37 The centralizer of any torus in a connected smooth solvable group G is connected. P ROOF. Omitted for the present (cf. Humphreys 1975, 19.4).
2
14j
Exercises
E XERCISE 14-1 Give a geometric proof that G connected implies DG connected. [Show that the image of connected set under a continuous map is connected (for the Zariski topology, say), the closure of a connected set is connected, and a nested union of connected sets is connected sets is connected; then apply the criterion (12.16).] E XERCISE 14-2 Show that if 1 ! N ! G ! Q ! 1 is exact, so also is 0 .N / ! 0 .G/ ! 0 .Q/ ! 1 (in an obvious sense). Give an example to show that 0 .N / ! 0 .G/ need not be injective.
15
The classical almost-simple groups
In this section, we construct the split classical almost-simple algebraic groups, and study their basic properties.
A SIDE 15.1 This section needs to be rewritten. At the moment it only (partially) treats the spin group, which is, in fact, the most important case.
A smooth connected algebraic group is semisimple if it has no smooth connected normal commutative subgroup other than the identity. This is the most important class of algebraic groups, and the next chapter is largely devoted to their study and classication. In this section, we conne ourselves to constructing an important class of semisimple groups. Let be a nondegenerate bilinear form on a k-vector space V . The special orthogonal group SO. / is connected and almost-simple, and it has a 2-fold covering Spin. / which we now construct. Throughout this section, k is a eld not of characteristic 2 and k-algebra means associative (not necessarily commutative) k-algebra containing k in its centre. For example, the n n matrices with entries in k become such a k-algebra Mn .k/ once we identify an element c of k with the scalar matrix cIn .
15a
Quadratic spaces
Let k be a eld not of characteristic 2, and let V be a nite-dimensional k-vector space. A quadratic form on V is a mapping qW V ! k such that q.x/ D
q .x; x/
for some symmetric bilinear form q.x C y/ D q.x/ C q.y/ C 2
qW V q .x; y/,
V ! k. Note that (97)
and so q is uniquely determined by q. A quadratic space is a pair .V; q/ consisting of a nitedimensional vector space and a quadratic form q. Often Ill write (rather than q ) for the associated symmetric bilinear form and denote .V; q/ by .V; q / or .V; /. A nonzero vector x in V is isotropic if q.x/ D 0 and anisotropic if q.x/ 0. Note that q is zero (i.e., q.V / D 0) if and only if is zero (i.e., .V; V / D 0). The discriminant of .V; q/ is the determinant of the matrix . .ei ; ej // where e1 ; : : : ; en is a basis of V . The choice of a different basis multiplies det. .ei ; ej // by a nonzero square, and so the discriminant is an element of k=k 2 . Let .V1 ; q1 / and .V2 ; q2 / be quadratic spaces. An isometry is an injective k-linear map W V1 ! V2 such that q2 . x/ D q1 .x/ for all x 2 V (equivalently, . x; y/ D .x; y/ for all x; y 2 V ). By .V1 ; q1 / .V2 ; q2 / we mean the quadratic space .V; q/ with V D V1 V2 q.x1 C x2 / D q.x1 / C q.x2 /; x 1 2 V1 , x2 2 V2 :
Let .V; q/ be quadratic space. A basis e1 ; : : : ; en for V is said to be orthogonal if .ei ; ej / D 0 for all i j . P ROPOSITION 15.2 Every quadratic space has an orthogonal basis (and so is an orthogonal sum of quadratic spaces of dimension 1). 150
15. The classical almost-simple groups
151
P ROOF. If q.V / D 0, then every basis is orthogonal. Otherwise, let e 2 V be such that q.e/ 0, and extend it to a basis e; e2 ; : : : ; en for V . Then e; e2 is again a basis for V , and the last n induction to W . .e; e2 / e; : : : ; en q.e/ .e; en / e q.e/
2
1 vectors span a subspace W for which .e; W / D 0. Apply
An orthogonal basis denes an isometry .V; q/ ! .k n ; q 0 /, where

2 q 0 .x1 ; : : : ; xn / D c1 x1 C 2 C cn xn ;
ci D q.ei / 2 k:
If every element of k is a square, for example, if k is algebraically closed, we can even scale the ei so that each ci is 0 or 1.
15b
Theorems of Witt and Cartan-Dieudonn e
A quadratic space .V; q/ is said to be regular32 (or nondegenerate,. . . ) if for all x 0 in V , there exists a y such that .x; y/ 0. Otherwise, it is singular. Also, .V; q/ is isotropic if it contains an isotropic vector, i.e., if q.x/ D 0 for some x 0; totally isotropic if every nonzero vector is isotropic, i.e., if q.x/ D 0 for all x, and anistropic if it is not isotropic, i.e., if q.x/ D 0 implies x D 0. Let .V; q/ be a regular quadratic space. Then for any nonzero a 2 V , hai? D fx 2 V j .a; x/ D 0g is a hyperplane in V (i.e., a subspace of dimension dim V reection in the hyperplane orthogonal to a is dened to be Ra .x/ D x 2 .a; x/ a. q.a/
def
def
1). For an anisotropic a 2 V , the
Then Ra sends a to a and xes the elements of W D hai? . Moreover, q.Ra .x// D q.x/ 2 2 .a; x/ 4 .a; x/2 q.a/ D q.x/; .a; x/ C q.a/ q.a/2
and so Ra is an isometry. Finally, relative to a basis a; e2 ; : : : ; en with e2 ; : : : ; en a basis for W , its matrix is diag. 1; 1; : : : ; 1/, and so det.Ra / D 1. T HEOREM 15.3 Let .V; q/ be a regular quadratic space, and let be an isometry from a subspace W of V into V . Then there exists a composite of reections V ! V extending . P ROOF. Suppose rst that W D hxi with x anisotropic, and let x D y. Geometry in the plane suggests that we should reect in the line x C y. In the plane this is the line orthogonal to x y, and, if x y is anisotropic, then Rx y .x/ D y
32 With
the notations of the last paragraph, .V; q/ is regular if c1 : : : cn 0.
152 as required. To see this, note that .x because q.x/ D q.y/, and so .x .x hence Rx If x
y .x/
y; x/ D
.x
y; y/
y; x C y/ D 0 y; x y/ D 2 .x y; x/I
Dx
2 .x y; x/ .x .x y; x y/
y/ D x
.x
y/ D y.
y is isotropic, then 4q.x/ D q.x C y/ C q.x y/ D q.x C y/
and so x C y is anisotropic. In this case, RxCy Rx .x/ D Rx We now proceed33 by induction on m.W / D dim W C 2 dim.W \ W ? /: C ASE W NOT TOTALLY ISOTROPIC : In this case, the argument in the proof of (15.2) shows that there exists an anisotropic vector x 2 W , and we let W 0 D hxi? \ W . Then, for w 2 W , w .w;x/ x 2 W 0 ; and so W D hxi W 0 (orthogonal decomposition). As m.W 0 / D m.W / 1, we q.x/ can apply induction to obtain a composite 0 of reections such that 0 jW 0 D jW 0 . From the denition of W 0 , we see that x 2 W 0? ; moreover, for any w 0 2 W 0 , . 0
def 1 . y/ .
x/ D y:
x; w 0 / D .x;
0 w 0 / D .x; w 0 / D 0;
and so y D 0 1 x 2 W 0? . By the argument in the rst paragraph, there exist reections (one or two) of the form Rz , z 2 W 0? , whose composite 00 maps x to y. Because 00 acts as the identity on W 0 , 0 00 is the map sought: . 0 00 /.cx C w 0 / D 0 .cy C w 0 / D c x C w 0 : C ASE W TOTALLY ISOTROPIC : Let V _ D Homk-lin .V; k/ be the dual vector space, and consider the surjective map W V
x7! .x; /
!V_
f 7!f jW
!W_
(so x 2 V is sent to the map y 7! .x; y/ on W ). Let W 0 be a subspace of V mapped isomorphically onto W _ . Then W \ W 0 D f0g and we claim that W C W 0 is a regular subspace of V . Indeed, if x C x 0 2 W C W 0 with x 0 0, then there exists a y 2 W such that 0 .x 0 ; y/ D .x C x 0 ; y/; if x 0, there exists a y 2 W 0 such that .x; y/ 0.
33 Following
Scharlau 1985, Chapter 1, 5.5.
15. The classical almost-simple groups Endow W W _ with the symmetric bilinear form .x; f /; .x 0 ; f 0 / 7! f .x 0 / C f 0 .x/. Relative to this bilinear form, the map x C x 0 7! .x; .x 0 //W W C W 0 ! W W _ is an isometry. The same argument applied to W gives a subspace W 00 and an isometry x C x 00 7! .x; : : :/W W C W 00 ! W . W /_ : Now the map W CW 0 ! W W _ is an isometry extending . As m.W W 0 / D 2 dim W < 3 dim W D m.W / we can apply induction to complete the proof. C OROLLARY 15.4 Every isometry of .V; q/ is a composite of reections. P ROOF. This is the special case of the theorem in which W D V . C OROLLARY 15.5 (W ITT CANCELLATION ) Suppose .V; q/ has orthogonal decompositions
0 0 0 0 .V; q/ D .V1 ; q1 / .V2 ; q2 / D .V1 ; q1 / .V2 ; q2 / 0 0 0 0 with .V1 ; q1 / and .V1 ; q1 / regular and isometric. Then .V2 ; q2 / and .V2 ; q2 / are isometric. 0 P ROOF. Extend an isometry V1 ! V1 0 0? onto V2 D V1 . (98)
_ 1
153
(98)
(99)
! W . W /_ ! W C W 00
(99)
? V to an isometry of V . It will map V2 D V1 isometrically 2
C OROLLARY 15.6 All maximal totally isotropic subspace of .V; q/ have the same dimension. P ROOF. Let W1 and W2 be maximal totally isotropic subspaces of V , and suppose that dim W1 dim W2 . Then there exists an injective linear map W W1 ! W2 V , which is automatically an 1 W is a isometry. Therefore, by Theorem 15.3 it extends to an isometry W V ! V . Now 2 1 W , and so totally isotropic subspace of V containing W1 . Because W1 is maximal, W1 D 2 dim W1 D dim 1 W2 D dim W2 . 2 R EMARK 15.7 In the situation of Theorem 15.3, Witts theorem says simply that there exists an isometry extending to V (not necessarily a composite of reections), and the Cartan-Dieudonn e theorem says that every isometry is a composite of at most dim V reections. When V is anisotropic, the proof of Theorem 15.3 shows this, but the general case is considerably more difcult see Artin 1957.
154
D EFINITION 15.8 The (Witt) index of a regular quadratic space .V; q/ is the maximum dimension of a totally isotropic subspace of V . D EFINITION 15.9 A quadratic space .V; q/ is a hyperbolic plane if it satises one of the following equivalent conditions: (a) .V; q/ is regular and isotropic of dimension 2I (b) for some basis of V , the matrix of the form is 0 1 ; 10 (c) V has dimension 2 and the discriminant of q is 1 (modulo squares). T HEOREM 15.10 (W ITT DECOMPOSITION ) A regular quadratic space .V; q/ with Witt index m has an orthogonal decomposition V D H1 Hm Va (100)
with the Hi hyperbolic planes and Va anisotropic; moreover, Va is uniquely determined up to isometry. P ROOF. Let W be a maximal isotropic subspace of V , and let e1 ; : : : ; em be a basis for W . One easily extends the basis to a linearly independent set e1 ; : : : ; em ; emC1 ; : : : ; e2m such that .ei ; emCj / D ij (Kronecker delta) and q.emCi / D 0 for i m. Then V decomposes as (100) with34 Hi D hei ; emCi i and Va D he1 ; : : : ; e2m i? . The uniqueness of Va follows from the Witt cancellation theorem (15.5). 2
15c
The orthogonal group
Let .V; q/ be a regular quadratic space. Dene O.q/ to be the group of isometries of .V; q/. Relative to a basis for V , O.q/ consists of the automorphs of the matrix M D . .ei ; ej //, i.e., the matrices T such that T t M T D M: Thus, O.q/ is an algebraic subgroup of GLV (see 2.20), called the orthogonal group of q (it is also called the orthogonal group of , and denoted O. /). Let T 2 O.q/. As det M 0, det.T /2 D 1, and so det.T / D 1. The subgroup of isometries with det D C1 is an algebraic subgroup of SLV , called the special orthogonal group SO.q/.
15d
Super algebras
Recall (2d) that a superalgebra (or Z=2Z-graded algebra) over k is k-algebra C together with a decomposition C D C0 C1 of C as a k-vector space such that k C0 ; C0 C0 C0 ; C0 C1 C1 ; C1 C0 C1 ; C1 C1 C0 :
Note that C0 is a k-subalgebra of C . A homomorphism of super k-algebras is a homomorphism 'W C ! D of algebras such that '.Ci / Di for i D 0; 1:
34 We
often write hS i for the k-space spanned by a subset S of a vector space V .
155
E XAMPLE 15.11 Let c1 ; : : : ; cn 2 k. Dene C.c1 ; : : : ; cn / to be the k-algebra with generators e1 ; : : : ; en and relations 2 ei D ci ; ej ei D ei ej (i j ).
i i As a k-vector space, C.c1 ; : : : ; cn / has basis fe11 : : : enn j ij 2 f0; 1gg, and so has dimension 2n . When we set C0 and C1 equal to the subspaces i i C0 D he11 : : : enn j i1 C
C in eveni C in oddi;
C1 D
i i he11 : : : enn
j i1 C
of C.c1 ; : : : ; cn /, then it becomes a superalgebra. Let C D C0 C1 and D D D0 D1 be two super k-algebras. The super tensor product of C O and D; C D, is dened to be the k-vector space C k D endowed with the superalgebra structure O C D O C D
0 .ci dj /.ck 0
D .C0 D0 / .C1 D1 / D .C0 D1 / .C1 D0 /

0 D . 1/j k .ci ck dj dl0 /
1 0 dl /
ci 2 Ci , dj 2 Dj etc..
The maps O iC W C ! C D; O iD W D ! C D; c 7! c 1 d 7! 1 d
have the following universal property: for any homomorphisms of k-superalgebras f W C ! T; whose images anticommute in the sense that f .ci /g.dj / D . 1/ij g.dj /f .ci /; ci 2 Ci ; dj 2 Dj ; gW D ! T
O there is a unique superalgebra homomorphism hW C D ! T such that f D h iC , g D h iD . O E XAMPLE 15.12 As a k-vector space, C.c1 /C.c2 / has basis 1 1, e 1, 1 e, e e, and .e 1/2 D e 2 1 D c1 1 1 .1 e/2 D 1 e 2 D c2 1 1 .e 1/.1 e/ D e e D .1 e/.e 1/: Therefore, O C.c1 /C.c2 / ' C.c1 ; c2 / e 1 $ e1 1 e $ e2 : Similarly, C.c1 ; : : : ; ci and so, by induction, O C.c1 / O C.cn / ' C.c1 ; : : : ; cn /: O 1 /C.ci / ' C.c1 ; : : : ; ci /,
156
E XAMPLE 15.13 Every k-algebra A can be regarded as a k-superalgebra by setting A0 D A and O A1 D 0. If A; B are both k-algebras, then A k B D Ak B.
def L E XAMPLE 15.14 Let X be a manifold. Then H.X /L i H i .X; R/ becomes an L D R-algebra under cup-product, and even a superalgebra with H.X /0 D i H 2i .X; R/ and H.X /1 D i H 2i C1 .X; R/. If Y is a second manifold, the K nneth formula says that u
H.X (super tensor product).
O Y / D H.X /H.Y /
15e
Brief review of the tensor algebra

L
n 0V n ,
Let V be a k-vector space. The tensor algebra of V is T .V / D V 0 D k; V 1 D V; V n D V
where
V .n copies of V /
with the algebra structure dened by juxtaposition, i.e., .v1 vm / .vmC1 vmCn / D v1 vmCn :
It is a k-algebra. If V has a basis e1 ; : : : ; em , then T .V / is the k-algebra of noncommuting polynomials in e1 ; : : : ; em . L There is a k-linear map V ! T .V /, namely, V D V 1 ,! n 0 V n , and any other k-linear map from V to a k-algebra R extends uniquely to a k-algebra homomorphism T .V / ! R.
15f
The Clifford algebra

be the corresponding bilinear form on V .
Let .V; q/ be a quadratic space, and let
D EFINITION 15.15 The Clifford algebra C.V; q/ is the quotient of the tensor algebra T .V / of V by the two-sided ideal I.q/ generated by the elements x x q.x/ .x 2 V /. Let W V ! C.V; q/ be the composite of the canonical map V ! T .V / and the quotient map T .V / ! C.V; q/. Then is k-linear, and35 .x/2 D q.x/, all x 2 V: Note that if x is anisotropic in V , then .x/ is invertible in C.V; q/, because (101) shows that .x/ .x/ D 1. q.x/ (101)
35 For a k-algebra R, we are regarding k as a subeld of R. When one regards a k-algebra R as a ring with a k ! R, it is necessary to write (101) as .x/2 D q.x/ 1C.V;q/ :
157
E XAMPLE 15.16 If V is one-dimensional with basis e and q.e/ D c, then T .V / is a polynomial algebra in one symbol e, T .V / D ke, and I.q/ D .e 2 c/. Therefore, C.V; q/ C.c/. E XAMPLE 15.17 If q D 0, then C.V; q/ is the exterior algebra on V , i.e., C.V; q/ is the quotient of T .V / by the ideal generated by all squares x 2 , x 2 V . In C.V; q/, 0 D . .x/ C .y//2 D .x/2 C .x/ .y/ C .y/ .x/ C .y/2 D .x/ .y/ C .y/ .x/ and so .x/ .y/ D .y/ .x/.
P ROPOSITION 15.18 Let r be a k-linear map from V to a k-algebra D such that r.x/2 D q.x/. Then there exists a unique homomorphism of k-algebras rW C.V; q/ ! D such that r D r: N N V
r
C.V; q/
r N
D: P ROOF. According to the universal property of the tensor algebra, r extends uniquely to a homomorphism of k-algebras r 0 W T .V / ! D, namely, r 0 .x1 As r 0 .x x r 0 factors uniquely through C.V; q/. q.x// D r.x/2 q.x/ D 0;
2
xn / D r.x1 /
r.xn /.
As usual, .C.V; q/; / is uniquely determined up to a unique isomorphism by the universal property in the proposition.
T HE MAP C.c1 ; : : : ; cn / ! C.V; q/

Because is linear, .x C y/2 D . .x/ C .y//2 D .x/2 C .x/ .y/ C .y/ .x/ C .y/2 : On comparing this with .x C y/2 D q.x C y/ D q.x/ C q.y/ C 2 .x; y/; we nd that .x/ .y/ C .y/ .x/ D 2 .x; y/: In particular, if f1 ; : : : ; fn is an orthogonal basis for V , then .fi /2 D q.fi /; .fj / .fi / D .fi / .fj / .i j /: (102)
(101)
Let ci D q.fi /. Then there exists a surjective homomorphism ei 7! .fi /W C.c1 ; : : : ; cn / ! C.V; /: (103)
158
T HE GRADATION ( SUPERSTRUCTURE ) ON THE C LIFFORD ALGEBRA

Decompose T .V / D T .V /0 T .V /1 M T .V /0 D V m
m even
T .V /1 D As I.q/ is generated by elements of T .V /0 ,
M
m odd
V m :
I.q/ D .I.q/ \ T .V /0 / .I.q/ \ T .V /1 / ; and so C.V; q/ D C0 C1 with Ci D T .V /i =I.q/ \ T .V /i :
Clearly this decomposition makes C.V; q/ into a super algebra. In more down-to-earth terms, C0 is spanned by products of an even number of vectors from V , and C1 is spanned by products of an odd number of vectors.
T HE BEHAVIOUR OF THE C LIFFORD ALGEBRA WITH RESPECT TO DIRECT SUMS

Suppose .V; q/ D .V1 ; q1 / .V2 ; q2 /: Then the k-linear map V x has the property that r.x/2 D . 1 .x1 / 1 C 1 D q.x/; because . 1 .x1 / 1/.1
2 .x2 // 2 2 .x2 //
D V1 V2 D .x1 ; x2 /
! 7!
O C.V1 ; q1 /C.V2 ; q2 / 1 .x1 / 1 C 1 2 .x2 /:
D .q.x1 / C q.x2 //.1 1/
1 .x1 / 2 .x2 /
D .1
2 .x2 //. 1 .x1 / 1//:
Therefore, it factors uniquely through C.V; q/:
O C.V; q/ ! C.V1 ; q1 /C.V2 ; q2 /.
(104)
159
E XPLICIT DESCRIPTION OF THE C LIFFORD ALGEBRA

T HEOREM 15.19 Let .V; q/ a quadratic space of dimension n. (a) For every orthogonal basis for .V; q/, the homomorphism (103) C.c1 ; : : : ; cn / ! C.V; q/ is an isomorphism. (b) For every orthogonal decomposition .V; q/ D .V1 ; q1 / .V2 ; q2 /, the homomorphism (104) O C.V; q/ ! C.V1 ; q1 /C.V2 ; q2 / is an isomorphism. (c) The dimension of C.V; q/ as a k-vector space is 2n . P ROOF. If n D 1, all three statements are clear from (15.16). Assume inductively that they are true for dim.V / < n. Certainly, we can decompose .V; q/ D .V1 ; q1 / .V2 ; q2 / in such a way that dim.Vi / < n. The homomorphism (104) is surjective because its image contains 1 .V1 / 1 and O 1 2 .V2 / which generate C.V1 ; q1 /C.V2 ; q2 /, and so dim.C.V; q// 2dim.V1 / 2dim.V2 / D 2n :
From an orthogonal basis for .V; q/, we get a surjective homomorphism (103). Therefore, dim.C.V; q// 2n : It follows that dim.C.V; q// D 2n . By comparing dimensions, we deduce that the homomorphisms (103) and (104) are isomorphisms. 2 C OROLLARY 15.20 The map W V ! C.V; q/ is injective. From now on, we shall regard V as a subset of C.V; q/ (i.e., we shall omit ). R EMARK 15.21 Let L be a eld containing k. Then
0
extends uniquely to an L-bilinear form
WV 0
V 0 ! L;
V 0 D L k V;
and C.V 0 ; q 0 / ' L k C.V; q/ where q 0 is quadratic form dened by

0.
T HE CENTRE OF THE C LIFFORD ALGEBRA

Assume that .V; q/ is regular, and that n D dim V > 0. Let e1 ; : : : ; en be an orthogonal basis for .V; q/, and let q.ei / D ci . Let D . 1/ We saw in (15.19) that C.c1 ; : : : ; cn / ' C.V; q/: Note that, in C.c1 ; : : : ; cn /, .e1 ei .e1 .e1 Therefore, e1 en /2 D . Moreover,
1
n.n 1/ 2
c1
cn D . 1/
n.n 1/ 2
det. .ei ; ej //.
en / D . 1/i
ci .e1
ei ei
1 ei C1 1 ei C1
en / en /.
en / ei D . 1/n i ci .e1
en lies in the centre of C.V; q/ if and only if n is odd.
160
P ROPOSITION 15.22 (a) If n is even, the centre of C.V; q/ is k; if n is odd, it is of degree 2 over k, generated by e1 en : In particular, C0 \ Centre.C.V; q// D k. (b) No nonzero element of C1 centralizes C0 . P ROOF. First show that a linear combination of reduced monomials is in the centre (or centralizes C0 ) if and only if each monomial does, and then nd the monomials that centralize the ei (or the ei ej ). 2 In Scharlau 1985, Chapter 9, 2.10, there is the following description of the complete structure of C.V; q/: If n is even, C.V; q/ is a central simple algebra over k, isomorphic to a tensor product of quaternion algebras. If n is odd, the centre of C.V; q/ is generated over k by the element e1 en whose square is , and, if is not a square in k, then C.V; q/ is a p central simple algebra over the eld k .
T HE INVOLUTION
An involution of a k-algebra D is a k-linear map W D ! D such that .ab/ D b a and a D 1. For example, M 7! M t (transpose) is an involution of Mn .k/. Let C.V; q/opp be the opposite k-algebra to C.V; q/, i.e., C.V; q/opp D C.V; q/ as a k-vector space but ab in C.V; q/opp D ba in C.V; q/. The map W V ! C.V; q/opp is k-linear and has the property that .x/2 D q.x/. Thus, there exists an isomorphism W C.V; q/ ! C.V; q/opp inducing the identity map on V , and which therefore has the property that .x1 for x1 ; : : : ; xr 2 V . We regard xr / D xr x1
as an involution of A. Note that, for x 2 V , x x D q.x/.
15g
The Spin group
Initially we dene the spin group as an abstract group. D EFINITION 15.23 The group Spin.q/ consists of the elements t of C0 .V; q/ such that (a) t t D 1; (b) tV t 1 D V , (c) the map x 7! txt
1W V
! V has determinant 1:
R EMARK 15.24 (a) The condition (a) implies that t is invertible in C0 .V; q/, and so (b) makes sense. (b) We shall see in (15.28) below that the condition (c) is implied by (a) and (b).
161
T HE MAP Spin.q/ ! SO.q/

Let t be an invertible element of C.V; q/ such that tV t V is an isometry, because q.txt
1 1
D V . Then the mapping x 7! txt

1
1W V
/ D .txt
1 2
/ D tx 2 t
D t q.x/t
D q.x/. SO.q/.
Therefore, an element t 2 Spin.q/ denes an element x 7! txt T HEOREM 15.25 The homomorphism Spin.q/ ! SO.q/
1 of
just dened has kernel of order 2, and it is surjective if k is algebraically closed. P ROOF. The kernel consists of those t 2 Spin.q/ such that txt 1 D x for all x 2 V . As V generates C.V; q/, such a t must lie in the centre of C.V; q/. Since it is also in C0 , it must lie in k. Now the condition t t D 1 implies that t D 1. For an anisotropic a 2 V , let Ra be the reection in the hyperplane orthogonal to a. According to Theorem 15.3, each element of SO.q/ can be expressed D Ra1 Ram for some ai . As det.Ra1 Ram / D . 1/m , we see that m is even, and so SO.q/ is generated by elements Ra Rb with a; b anisotropic elements of V . If k is algebraically closed, we can even scale a and b so that q.a/ D 1 D q.b/. Now axa
1
D . xa C 2 .a; x// a 2 .a; x/ D x a q.a/ D Ra .x/:
as .ax C xa D 2 .a; x/, see (102)) as a2 D q.a/
Moreover, .ab/ ab D baab D q.a/q.b/: Therefore, if q.a/q.b/ D 1, then Ra Rb is in the image of Spin.q/ ! SO.q/. As we noted above, such elements generate SO.q/ when k is algebraically closed. 2 In general, the homomorphism is not surjective. For example, if k D R, then Spin.q/ is connected but SO.q/ will have two connected components when is indenite. In this case, the image is the identity component of SO.q/.
15h
Write
The Clifford group

for the automorphism of C.V; q/ that acts as 1 on C0 .V; q/ and as 1 on C1 .V; q/.
D EFINITION 15.26 The Clifford group is .q/ D ft 2 C.V; q/ j t invertible and .t /V t For t 2 .q/, let .t / denote the homomorphism x 7! .t /xt
1
D V g: !V.
1W V
162 P ROPOSITION 15.27 For all t 2
BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1) .q/, .t / is an isometry of V , and the sequence 1!k ! .q/ ! O.q/ ! 1
is exact (no condition on k).
P ROOF. Let t 2 .q/. On applying and to .t /V D V t, we nd that .t /V D V t , and so t 2 .q/. Now, because and act as 1 and 1 on V , .t / x t and so .t / .t /x D xt t: We use this to prove that .t / is an isometry: q..t/.x// D ..t /.x// ..t /.x// D t
1 1
. .t / x t
/ D
.t
x .t // D .t
/xt ;
(105)
x .t /
.t /xt
1 .105/
D t
xxt t t
D q.x/:
As k is in the centre of .q/, k is in the kernel of . Conversely, let t D t0 C t1 be an invertible element of C.V; q/ such that .t /xt 1 D x for all x 2 V , i.e., such that t0 x D xt0 ; t1 x D xt1
for all x 2 V . As V generates C.V; q/ these equations imply that t0 lies in the centre of C.V; q/, and hence in k (15.22a), and that t1 centralizes C0 , and hence is zero (15.22b). We have shown that Ker./ D k : It remains to show that is surjective. For t 2 V , .t /.y/ D tyt 1 and so (see the proof of (15.25)), .t / D Rt . Therefore the surjectivity follows from Theorem 15.3. 2 C OROLLARY 15.28 For an invertible element t of C0 .V; q/ such that tV t of x 7! txt 1 W V ! V is one. P ROOF. According to the proposition, every element t 2 t D ca1 am Ram on V , and
2 1
D V , the determinant
.q/ can be expressed in the form
with c 2 k and the ai anisotropic elements of V . Such an element acts as Ra1 has determinant . 1/m . If t 2 C0 .V; q/, then m is even, and so det.t / D 1. Hence, the condition (c) in the denition of Spin .q/ is superuous.
163
15i
Action of O.q/ on Spin.q/
15.29 An element of O.q/ denes an automorphism of C.V; q/ as follows. Consider W V ! C.V; q/. Then . . .x//2 D . .x// 1 D .x/ 1 for every x 2 V . Hence, by the universal property, there is a unique homomorphism Q W C.V; q/ ! C.V; q/ rendering V ? ? y V ! C.V; q/ ? ? yQ ! C.V; q/
1 e commutative. Clearly 1 2 D e e and id D id, and so g D Q 1 , and so Q is an automor1 2 phism. If 2 SO. /, it is known that Q is an inner automorphism of C.V; q/ by an invertible element of C C .V; q/.
15j
Restatement in terms of algebraic groups
Let .V; q/ be quadratic space over k, and let qK be the unique extension of q to a quadratic form on K k V . As we noted in (15.21), C.V; qK / D K k C.V; q/. T HEOREM 15.30 There exists a naturally dened algebraic group Spin.q/ over k such that Spin.q/.K/ ' Spin.qK / for all elds K containing k. Moreover, there is a homomorphism of algebraic groups Spin.q/ ! SO.q/ giving the homomorphism in (15.25) for each eld K containing k. Finally, the action of O.q/ on C.V; q/ described in (15.25) denes an action of O.q/ on Spin.q/. P ROOF. Show that, when k is innite, the algebraic group attached to the subgroup Spin.q/ of GL.V / (see 6.20) has these properties. Alternatively, dene a functor R Spin.qR / that coincides with the previous functor when R is a eld. 2 In future, we shall write Spin.q/ for the algebraic group Spin.q/.
N OTES A representation of a semisimple algebraic group G gives rise to a representation of its Lie algebra g, and all representations of g arise from G only if G has the largest possible centre. When E. Cartan classied the simple representations of all simple Lie algebras, he discovered a new representation of the orthogonal Lie algebra [not arising from the orthogonal group]. But he did not give a specic name to it, and much later, he called the elements on which this new representation operates spinors, generalizing the terminology adoped by physicists in a special case for the rotation group of the three dimensional space (C. Chevalley, The Construction and Study of Certain Important Algebras, 1955, III 6). This explains the origin and name of the Spin group.
16
The exceptional almost-simple groups
A SIDE 16.1 Need to explain how to construct them explicitly (perhaps without giving all proofs). In characteristic zero, we can recover them from the Lie algebras (see Chapter II).
16a
The group G2
Let k be a eld of characteristic zero. A Hurwitz algebra over k is a nite k-algebra A (not necessarily commutative) together with a nondegenerate quadratic form N W A ! k such that N.xy/ D N.x/N.y/ for all x; y 2 A: The possible dimensions of A are 1, 2, 4, and 8. A Hurwitz algebra of dimension 8 is also known as an octonion or Cayley algebra. For such an algebra A, the functor R Autk .R k A/
is an algebraic group over k of type G2 . (To be continued).
164
17
General algebraic groups
Throughout this section, k is a eld.
17a
The radical of an algebraic group
P ROPOSITION 17.1 (a) Subgroups and quotient groups of solvable algebraic groups are solvable. (b) An algebraic group G is solvable if it contains a normal subgroup N such that N and G=N are solvable (i.e., extensions of solvable groups are solvable). (c) Let N and H be algebraic subgroups of G with N normal. If H and N are solvable (resp. connected), then HN is solvable (resp. connected). P ROOF. (a) Let G D G0 G1 Gn D f1g be a solvable series for G. Let H be a subgroup of G. The restriction of the quotient map Gi ! Gi =Gi C1 to H \ Gi ; H \ Gi ! Gi =Gi C1 ; has kernel .H \ Gi / \ Gi C1 D H \ Gi C1 . Therefore H \ Gi C1 is a normal subgroup of H \ Gi , and the quotient H \ Gi =H \ Gi C1 injects into Gi =Gi C1 , which is commutative. Therefore H H \ G1 H \ Gn D f1g
is a solvable series for H . N Let Q be a quotient group of G, and let Gi be the image of Gi in Q. Then it follows from the isomorphism theorems that N N N Q D G0 G1 Gn D f1g is a solvable series for Q. (b) Let G=N D Q0 N D N0 Q1 N1 Qm D f1g and Nn D f1g
be solvable series for G=N and N , and let Gi be the inverse image of Qi in G. Then Gi =Gi C1 ' Qi =Qi C1 (see 8.15), and so G G1 Gm .D N / N1 Nn
is a solvable series for G. (c) The quotient HN=N is solvable (resp. connected) because it is isomorphic to H=H \ N (see 8.13), and so this follows from (b) (resp. 12.17). 2 C OROLLARY 17.2 For any algebraic group G over a perfect eld k, there exists a unique largest connected normal smooth solvable subgroup (called the radical RG of G), and a unique largest connected normal smooth unipotent subgroup (called the unipotent radical Ru G of G). P ROOF. The rst statement follows from (c) of the proposition; for the second, take Ru G D .RG/u . 2
A SIDE 17.3 Discuss what happens when k is not perfect. The subgroup Ru G exists, but may not commute with extension of scalars. In other words, in general, .Ru G/k al Ru Gk al .
165
166
17b
Semisimple and reductive groups
D EFINITION 17.4 An algebraic group over an algebraically closed eld is reductive (resp. semisimple) if it is smooth and connected and has no smooth connected normal unipotent (resp. solvable) subgroup other than the identity. An algebraic group over a eld is reductive (resp. semisimple) if it becomes reductive (resp. semisimple) over the algebraic closure of the eld. For example, SLn , SOn , Spn are semisimple, and GLn is reductive. P ROPOSITION 17.5 Let G be a smooth connected algebraic group over a perfect eld k. (a) G is semisimple if and only if RG D 1. (b) G is reductive if and only if Ru G D 1. P ROOF. Obvious. P ROPOSITION 17.6 Let G be a smooth connected algebraic group over a eld k: (a) If G is semisimple, then it has no smooth connected normal commutative subgroup other than the identity; the converse is true if k is perfect. (b) If G is reductive, then the only smooth connected normal commutative subgroups it contains are tori; the converse is true if k is perfect. P ROOF. (a) Only the converse statement is nonobvious. For this, we use that, for any algebraic group G, RG and DG are characteristic subgroups, i.e., every automorphism of G maps RG onto RG and DG onto DG. This is obvious from their denitions: RG is the largest connected normal solvable algebraic subgroup and DG is the smallest normal algebraic subgroup such that G=DG is commutative. Therefore the chain G RG D.RG/ D2 .RG/ Dr .RG/ 1
2
is preserved by every automorphism of G. This remains true over k al , and so the groups are normal in G by (6.32). (b) Similar. 2 R EMARK 17.7 If one of the conditions, commutative, connected, normal, smooth, is dropped, then a semisimple group may have such a subgroup. For example, SL2 has the commutative normal subgroup fI g and the commutative connected subgroup U2 . Moreover, SL2 SL2 is semisimple, but it has the connected normal subgroup f1g SL2 . Finally, over a eld of characteristic 2, SL2 has the connected normal commutative subgroup 2 .
B E XAMPLE 17.8 Let G be the group of invertible matrices A C . The unipotent radical of G is the 0 subgroup of matrices I B . The quotient of G by Ru G is isomorphic to the reductive group of 0 I 0 invertible matrices of the form A C , i.e., to GLm GLn . The radical of this is Gm Gm . 0
A SIDE 17.9 A representation G ! GL.V / is said to be semisimple (or completely reducible) if every stable subspace W has a stable complement W 0 (so V is a direct sum V D W W 0 of representations), or,
17. General algebraic groups
167
equivalently, if V is a direct sum of simple (i.e., irreducible) representations (those with no proper nonzero subrepresentations). For example, the action of U2 on k 2 , 1 a x x C ay D ; 0 1 y y is not semisimple because the only stable one-dimensional subspace is the x-axis (the map is a shear). In general, representations of unipotent groups are not semisimple; nor should you expect the representations of a group containing a normal unipotent group to be semisimple. However, in characteristic zero, a connected algebraic group is reductive if and only if all of its representations are semisimple (see later). In characteristic p, the representations of a connected algebraic group are all semisimple if and only if it is a torus.
17c
Reductive groups
For simplicity, in this subsection I assume that k has characteristic zero. T HEOREM 17.10 If G is reductive, then the derived group G der of G is semisimple, the connected centre Z.G/ of G is a torus, and Z.G/ \ G der is the (nite) centre of G der ; moreover, Z.G/ G der D G. N P ROOF. It sufces to prove this with k D k. By denition, .RG/u D 0, and so (14.35) shows that RG is a torus T . Rigidity (13.21) implies that the action of G on RG by inner automorphisms is trivial, and so RG Z.G/ . Since the reverse inclusion always holds, this shows that R.G/ D Z.G/ D torus. We next show that Z.G/ \ G der is nite. Choose an embedding G ,! GLV , and write V as a direct sum V D V1 Vr of eigenspaces for the action of Z.G/ (see Z.G/ .k/ consists of the matrices 0 A1 B @0 0 13.15). When we choose bases for the Vi , then 0 :: : 0 0 1
C 0A Ar
with each Ai nonzero and scalar,36 and so its centralizer in GLV consists of the matrices of this shape with the Ai arbitrary. Since G der .k/ consists of commutators (14.12), it consists of such matrices with determinant 1. As SL.Vi / contains only nitely many scalar matrices, this shows that Z.G/ \ G der is nite. Note that Z.G/ G der is a normal algebraic subgroup of G such that G=.Z.G/ G der / is commutative (being a quotient of G=G der ) and semisimple (being a quotient of G=R.G/). Hence37 G D Z.G/ G der : Therefore G der ! G=R.G/ is surjective with nite kernel. As G=R.G/ is semisimple, so also is G der . Certainly Z.G/ \ G der Z.G der /, but, because G D Z.G/ G der and Z.G/ is commutative, Z.G der / Z.G/. 2
is, of the form diag.a; : : : ; a/ with a 0. G D DG if G is semisimple. In other words, a semisimple group has no commutative quotients. At the moment this is only proved at the end of II 4.
37 Because 36 That
168
R EMARK 17.11 From a reductive group G, we obtain a semisimple group G 0 (its derived group), a group Z of multiplicative type (its centre), and a homomorphism 'W Z.G 0 / ! Z. Moreover, G can be recovered from .G 0 ; Z; '/ as the quotient Z.G 0 /
z7!.'.z/
1 ;z/
!Z
G 0 ! G ! 1:
(106)
Clearly, every reductive group arises from such a triple .G 0 ; Z; '/ (and G 0 can even be chosen to be simply connected).
17d
Structure of a general algebraic group
Recall that algebraic group means afne algebraic group. T HEOREM 17.12 (a) An algebraic group G over a eld k contains a unique normal connected algebraic subgroup G such that G=G is an etale algebraic group. (b) If k is perfect, an algebraic group contains a unique largest smooth subgroup. (c) A smooth connected algebraic group G over a perfect eld contains a unique smooth connected normal solvable subgroup RG such that G=RG is a semisimple group. (d) A smooth connected solvable algebraic group G over a perfect eld contains a unique connected normal unipotent subgroup Gu such that G=Gu is of multiplicative type. P ROOF. (a) See 12.9. (b) The radical of G is such a subgroup (see the above discussion). (c) See 14.35.
A SIDE 17.13 Explain the connected components for a nonafne algebraic group, at least in the smooth case. Also discuss things over a ring k: A SIDE 17.14 Explain the Barsotti-Chevalley-Rosenlicht theorem. A SIDE 17.15 Somewhere explain anti-afne groups. A SIDE 17.16 Explain what is true when you drop smooth and perfect, and maybe even allow a base ring. In particular, discuss the work of Conrad, Gabber, and Prasad on pseudo-reductive groups (those groups G such Ru G D 0 but G is not reductive, i.e., Ru Gk al 0; in other words, G is pseudo-reductive if G has no smooth connected normal unipotent subgroup, but Gk al does).
18
18a
Tannakian categories
Properties of G versus those of Repk .G/
Since each of G and Repk .G/ determines the other, we should be able to see properties of one reected in the other. P ROPOSITION 18.1 An algebraic group G is nite if and only if there exists a representation .r; V / such that every representation of G is a subquotient38 of V n for some n 0. P ROOF. If G is nite, then the regular representation X of G is nite-dimensional, and (7.34) says that it has the required property. Conversely if, with the notations of (10a), Repk .G/ D hXi, then G D Spec B where B is the linear dual of the nite k-algebra AX . 2 P ROPOSITION 18.2 Assume k has characteristic zero. An algebraic group G is connected if and only if, for every representation V on which G acts nontrivially, the full subcategory of Rep.G/ of subquotients of V n , n 0, is not stable under . P ROOF. In characteristic zero, all nite groups are etale. Therefore, a group G is connected if and only if there is no non-trivial epimorphism G ! G 0 with G 0 nite. According to (7.52), this is equivalent to Repk .G/ having no non-trivial subcategory of the type described in (18.1). 2 P ROPOSITION 18.3 Let k be an algebraically closed eld. A smooth algebraic group over k is unipotent (resp. solvable) if and only if every nonzero representation of the group has a nonzero xed vector (resp. stable one-dimensional subspace). P ROOF. See (14.25) and (14.21).
2
P ROPOSITION 18.4 The identity component G of an algebraic group G over a eld of characteristic zero is reductive if and only if Repk .G/ is semisimple. P ROOF. See later.
2
P ROPOSITION 18.5 Let f W G ! G 0 be a homomorphism of afne groups over k, and let ! f be the corresponding functor Repk .G 0 / ! Repk .G/. (a) f is faithfully at if and only if ! f is fully faithful and every subobject of ! f .X /, for X 0 2 ob.Repk .G//, is isomorphic to the image of a subobject of ! f .X 0 /. (b) f is a closed immersion if and only if every object of Repk .G/ is isomorphic to a subquotient of an object of the form of ! f .X 0 /, X 0 2 ob.Repk .G 0 //. P ROOF. (a) If G ! G 0 is faithfully at, and therefore an epimorphism, then Repk .G 0 / can be identied with the subcategory of Repk .G/ of representations G ! GL.W / factoring through G 0 . It is therefore obvious that ! f has the stated properties. Conversely, if ! f is fully faithful, it denes an equivalence of Repk .G 0 / with a full subcategory of Repk .G/, and the second condition shows
V n is a direct sum of n copies of V , and subquotient means any representation isomorphic to a subrepresentation of a quotient (equivalently, to a quotient of a subrepresentation).
38 Here
169
170
that, for X 0 2 ob.Repk .G 0 //, hX 0 i is equivalent to h! f .X /i. Let G D Spec B and G 0 D Spec B 0 ; then (10b) shows that B 0 D lim End.! 0 jhX 0 i/_ D lim End.!jh! f .X 0 /i/_ ! ! lim End.!jhX i/_ D B; !
and B ! B 0 being injective implies that G ! G 0 is faithfully at (5.32). (b) Let C be the strictly full subcategory of Repk .G/ whose objects are isomorphic to subquotients of objects of the form of ! f .X 0 /. The functors
Repk .G 0 / ! C ! Repk .G/
correspond to homomorphisms of k-coalgebras B 0 ! B 00 ! B where G D Spec B and G 0 D Spec B 0 . An argument as in the above above proof shows that B 00 ! B is injective. Moreover, for X 0 2 ob.Repk .G 0 //, End.!jh! f .X /i/ ! End.! 0 jhX 0 i/ is injective, and so B 0 ! B 00 is surjective. If f is a closed immersion, then B 0 ! B is surjective and it follows that B 00 ! B, and C D Repk .G/. Conversely, if C D Repk .G/, B 00 D B and B 0 ! B is surjective. [Take a faithful representation of G 0 ; it is also a faithful representation of G, etc..] 2 P ROPOSITION 18.6 Let G and G 0 be algebraic groups over a eld k of characteristic zero, and assume G is reductive. Let f W G ! G 0 be a homomorphism, and let ! f W Rep.G 0 / ! Rep.G/ be the functor .r; V / 7! .r ; V /. Then: (a) f is a quotient map if and only if ! f is fully faithful; (b) f is an embedding if and if every object of Repk .G/ is isomorphic to a direct factor of an object of the form ! f .V /. P ROOF. Omitted for the present (Deligne and Milne 1982, 2.21, 2.29).
P ROPOSITION 18.7 Let G GLV be a connected algebraic group over a eld k of characteristic zero. The following conditions on a subgroup H of G are equivalent: (a) H is normal in G; (b) for every tensor space T D V m V _n and k-character of H , the subspace T on which H acts through is stable under G; (c) every H -isotypic component of any representation of G is stable under G. If, moreover, G is reductive, these conditions imply that Z.EndH .V // Z.EndG .V //:
P ROOF. Omitted for the present (Andr , Yves. Mumford-Tate groups of mixed Hodge structures e and the theorem of the xed part. Compositio Math. 82 (1992), no. 1, 124, Lemma 1). 2
18. Tannakian categories
171
18b
Tensor categories
18.8 A k-linear category is an additive category in which the Hom sets are nite-dimensional k-vector spaces and composition is k-bilinear. Functors between such categories are required to be k-linear, i.e., induce k-linear maps on the Hom sets. 18.9 A tensor category over k is a k-linear category together with a k-bilinear functor W C C ! C and compatible associativity and commutativity contraints ensuring that the tensor product of any unordered nite set of objects is well-dened up to a well-dened isomorphism. An associativity constraint is a natural isomorphism
U;V;W W U
.V W / ! .U V / W;
U; V; W 2 ob.C/;
and a commutativity constraint is a natural isomorphism

V;W W V
W ! W V;
V; W 2 ob.C/:
Compatibility means that certain diagrams, for example, U .V W / ? ?id y U V;W U .W V /

U;V;W
! .U V / W
U V;W
! W .U V / ? ? y W;U;V ! .W U / V;
U;W;V
! .U W / V
U;W
idV
commute, and that there exists a neutral object (tensor product of the empty set), i.e., an object U together with an isomorphism uW U ! U U such that V 7! V U is an equivalence of categories. For a complete denition, see Deligne and Milne 1982, 1. We use 1 to denote a neutral object of 1 C. 18.10 An object of a tensor category is trivial if it is isomorphic to a direct sum of neutral objects. E XAMPLE 18.11 The category of nitely generated modules over a ring R becomes a tensor category with the usual tensor product and the constraints u .v w/ 7! .u v/ wW U .V W / ! .U V / W v w ! w vW V W ! W V: (107)
Any free R-module U of rank one together with an isomorphism U ! U U (equivalently, the choice of a basis for U ) is a neutral object. It is trivial to check the compatibility conditions for this to be a tensor category. E XAMPLE 18.12 The category of nite-dimensional representations of a Lie algebra or of an algebraic (or afne) group G with the usual tensor product and the constraints (107) is a tensor category. The required commutativities follow immediately from (18.11). 18.13 Let .C; / and .C0 ; / be tensor categories over k. A tensor functor C ! C0 is a pair .F; c/ consisting of a functor F W C ! C0 and a natural isomorphism cV;W W F .V / F .W / ! F .V W / compatible the associativity and commutativity constraints and sending neutral objects to a neutral objects. Then F commutes with nite tensor products up to a well-dened isomorphism. See Deligne and Milne 1982, 1.8.
172
18.14 Let C be a tensor category over k, and let V be an object of C. A pair .V _ ; V _ V ! 1 1/ is called a dual of V if there exists a morphism V W 1 ! V V _ such that the composites 1 V V_
V V V ev ev
V V _ V
V _ V
! V _ V V _
ev V _
! V_
are the identity morphisms on V and V _ respectively. Then V is uniquely determined, and the dual .V _ ; ev/ of V is uniquely determined up to a unique isomorphism. For example, a nitedef dimensional k-vector space V has as dual V _ D Homk .V; k/ with ev.f v/ D f .v/ here V P is the k-linear map sending 1 to ei fi for any basis .ei / for V and its dual basis .fi /. More generally, a module M over a ring admits a dual if and only if M is nitely generated and projective (CA 10.8, 10.9). Similarly, the contragredient of a representation of a Lie algebra or of an algebraic group is a dual of the representation. 18.15 A tensor category is rigid if every object admits a dual. For example, the category Veck of nite-dimensional vector spaces over k and the category of nite-dimensional representations of a Lie algebra (or an algebraic group) are rigid.
18c
Neutral tannakian categories
18.16 A neutral tannakian category over k is an abelian k-linear category C endowed with a rigid tensor structure for which there exists an exact tensor functor !W C ! Veck . Such a functor ! is called a bre functor over k. T HEOREM 18.17 Let .C; !/ be a neutral tannakian category over k, i.e., C is a neutral tannakian category over k and ! is a bre functor over k. For each k-algebra R, let G.R/ be the set of families D. such that Then R k,
V W D V W for all V; W 2 ob.C/, 1 1 D id!.1 for every neutral object of 1 of 1 1/ W V /V 2ob.C/ ; V
2 EndR-linear .!.V /R /;
R D R
C, and for all arrows W V ! W in C.
G.R/ is an afne group over k, and ! denes an equivalence of tensor categories over
C ! Rep.G/:
P ROOF. This is a restatement of Theorem 10.8.
18.18 Let !R be the functor V !.V / R; then G.R/ consists of the natural transformations W !R ! !R such that the following diagrams commute !R .V / !R .W / ? ? y V W !R .V / !R .W /
cV;W
! !R .V W / ? ? y V W
!R .1 1/ ? ? y 1 1 !R .1 1/
!R .u/
! !R .1 1 1 1/ ? ? y 1 1 11 ! !R .1 1 1 1/
cV;W
! !R .V W /
!R .u/
for all objects V , W of C and all identity objects .1 u/. 1;
173
18.19 I explain the nal statement of (18.17). For each V in C, there is a representation rV W G ! GL!.V / dened by rV .g/v D
V .v/
if g D .
V/
2 G.R/ and v 2 V .R/:
The functor sending V to !.V / endowed with this action of G is an equivalence of categories C ! Rep.G/. 18.20 A tannakian category C is said to be algebraic if there exists an object V such that every other object is a subquotient of P .V; V _ / for some P 2 NX; Y . If G is an algebraic group, then (7.28) and (7.43) show that Rep.G/ is algebraic. Conversely, if Rep.G/ is algebraic, with generator V , then G is algebraic because G GLV . 18.21 It is usual to write Aut .!/ (functor of tensor automorphisms of !) for the afne group G attached to the neutral tannakian category .C; !/ we call it the Tannaka dual or Tannaka group of C. Alternatively, denote it by .C; !/. E XAMPLE 18.22 If C is the category of nite-dimensional representations of an algebraic group H over k and ! is the forgetful functor, then G.R/ ' H.R/ by (9.2), and C ! Rep.G/ is the identity functor. E XAMPLE 18.23 Let N be a normal subgroup of an algebraic group G, and let C be the subcategory of Rep.G/ consisting of the representations of G on which N acts trivially. The group attached to C and the forgetful functor is G=N (alternatively, this can be used as a denition of G=N ). 18.24 Let .C; !/ and .C0 ; ! 0 / be neutral tannakian categories with Tannaka duals G and G 0 . An exact tensor functor F W C ! C0 such that ! 0 F D ! denes a homomorphism G 0 ! G, namely, .
V /V 2ob.C0 /
7! .
F V /V 2ob.C/ W G
.R/ ! G.R/:
18.25 Let C D Rep.G/ for some algebraic group G. (a) For an algebraic subgroup H of G, let CH denote the full subcategory of C whose objects are those on which H acts trivially. Then CH is a neutral tannakian category whose Tannaka dual is G=N where N is the smallest normal algebraic subgroup of G containing H (intersection of the normal algebraic subgroups containing H ). (b) (Tannaka correspondence.) For a collection S of objects of C D Rep.G/, let H.S / denote the largest subgroup of G acting trivially on all V in S ; thus \ H.S / D Ker.rV W G ! Aut.V //:
V 2S
Then the maps S 7! H.S / and H 7! CH form a Galois correspondence fsubsets of ob.C/g falgebraic subgroups of Gg;
CH.S /
i.e., both maps are order reversing and S and H.CH / H for all S and H . It follows that the maps establish a one-to-one correspondence between their respective images. In this way, we get a natural one-to-one order-reversing correspondence ftannakian subcategories of Cg $ fnormal algebraic subgroups of Gg (a tannakian subcategory is a full subcategory closed under the formation of duals, tensor products, direct sums, and subquotients).
1W1
174
18d
Gradations on tensor categories
18.26 Let M be a nitely generated abelian group. An M -gradation on an object X of an abelian L category is a family of subobjects .X m /m2M such that X D m2M X m . An M -gradation on a tensor category C is an M -gradation on each object L of C compatible with all arrows in C and X with tensor products in the sense that .X Y /m D rCsDm X r X s . Let .C; !/ be a neutral tannakian category, and let G be its Tannaka dual. To give an M -gradation on C is the same as to give a central homomorphism D.M / ! G.!/: a homomorphism corresponds to the M -gradation such that X m is the subobject of X on which D.M / acts through the character m (Saavedra Rivano 1972; Deligne and Milne 1982, 5). 18.27 Let C be a semsimple k-linear tensor category such that End.X / D k for every simple object X in C, and let I.C/ be the set of isomorphism classes of simple objects in C. For elements x; x1 ; : : : ; xm of I.C/ represented by simple objects X; X1 ; : : : ; Xm , write x x1 xm if X is a direct factor of X1 Xm . The following statements are obvious. (a) Let M be a commutative group. To give an M -gradation on C is the same as to give a map f W I.C/ ! M such that x x1 x2 H) f .x/ D f .x1 / C f .x2 /:
A map from I.C/ to a commutative group satisfying this condition will be called a tensor map. For such a map, f .1 D 0, and if X has dual X _ , then f .X _ / D f .X /. 1/ (b) Let M.C/ be the free abelian group with generators the elements of I.C/ modulo the relations: x D x1 C x2 if x x1 x2 . The obvious map I.C/ ! M.C/ is a universal tensor map, i.e., it is a tensor map, and every other tensor map I.C/ ! M factors uniquely through it. Note that I.C/ ! M.C/ is surjective. 18.28 Let .C; !/ be a neutral tannakian category such that C is semisimple and End.V / D k for def every simple object in C. Let Z be the centre of G D Aut .!/. Because C is semisimple, G is reductive (II, 5.16 ), and so Z is of multiplicative type. Assume (for simplicity) that Z is split, so that Z D D.N / with N the group of characters of Z. According to (18.26), to give an M -gradation on C is the same as giving a homomorphism D.M / ! Z, or, equivalently, a homomorphism N ! M . On the other hand, (18.27) shows that to give an M -gradation on C is the same as giving a homomorphism M.C/ ! M . Therefore M.C/ ' N . In more detail: let X be an object of C; if X is simple, then Z acts on X through a character n of Z, and the tensor map X 7! nW I.C/ ! N is universal. 18.29 Let .C; !/ be as in (18.28), and dene an equivalence relation on I.C/ by a a0 there exist x1 ; : : : ; xm 2 I.C/ such that a; a0 x1 xm :
A function f from I.C/ to a commutative group denes a gradation on C if and only if f .a/ D f .a0 / whenever a a0 . Therefore, M.C/ ' I.C/= .
A SIDE 18.30 Discuss the prehistory: Tannaka (cf. Serre 1973, p71, remark), Krein (cf. Breen), Chevalley (book), Hochschild and Mostow 1969, 4 (AJM 91, 11271140).
175
E XERCISES
E XERCISE 18-1 Use the criterion (18.1) to show that the centralizer of a torus in a connected algebraic group is connected.
Bibliography
A BE , E. 1980. Hopf algebras, volume 74 of Cambridge Tracts in Mathematics. Cambridge University Press, Cambridge. Translated from the Japanese by Hisae Kinoshita and Hiroko Tanaka. A RTIN , E. 1957. Geometric algebra. Interscience Publishers, Inc., New York-London. A RTIN , M. 1991. Algebra. Prentice Hall Inc., Englewood Cliffs, NJ. BARSOTTI , I. 1955. Un teorema di struttura per le variet` gruppali. Atti Accad. Naz. Lincei. Rend. a Cl. Sci. Fis. Mat. Nat. (8) 18:4350. B OREL , A. 1991. Linear algebraic groups. Springer-Verlag, New York. C ARTIER , P. 1962. Groupes alg briques et groupes formels, pp. 87111. In Colloq. Th orie des e e Groupes Alg briques (Bruxelles, 1962). Librairie Universitaire, Louvain. e C ARTIER , P. 2007. A primer of Hopf algebras, pp. 537615. In Frontiers in number theory, physics, and geometry. II. Springer, Berlin. Preprint available at IHES. C HEVALLEY, C. 1960. Une d monstration dun th or` me sur les groupes alg briques. J. Math. e e e e Pures Appl. (9) 39:307317. C ONRAD , B. 2002. A modern proof of Chevalleys theorem on algebraic groups. J. Ramanujan Math. Soc. 17:118. D ELIGNE , P. AND M ILNE , J. S. 1982. Tannakian categories, pp. 101228. In Hodge cycles, motives, and Shimura varieties, Lecture Notes in Mathematics 900. Springer-Verlag, Berlin. D EMAZURE , M. 1972. Lectures on p-divisible groups. Springer-Verlag, Berlin. D EMAZURE , M. AND G ABRIEL , P. 1970. Groupes alg briques. Tome I: G om trie alg brique, e e e e g n ralit s, groupes commutatifs. Masson & Cie, Editeur, Paris. e e e F IORESI , R. AND G AVARINI , F. 2008. Chevalley supergroups. arXiv:0808.07851. G REENBERG , M. J. 1961. Schemata over local rings. Ann. of Math. (2) 73:624648. G REENBERG , M. J. 1963. Schemata over local rings. II. Ann. of Math. (2) 78:256266. H ARTSHORNE , R. 1977. Algebraic geometry. Springer-Verlag, New York. H UMPHREYS , J. E. 1975. Linear algebraic groups. Springer-Verlag, New York. K ASSEL , C. 1995. Quantum groups, volume 155 of Graduate Texts in Mathematics. SpringerVerlag, New York. 177
178
KOLCHIN , E. R. 1948. On certain concepts in the theory of algebraic matric groups. Ann. of Math. (2) 49:774789. M AC L ANE , S. 1971. Categories for the working mathematician. Springer-Verlag, New York. Graduate Texts in Mathematics, Vol. 5. M ILNE , J. S. 1980. Etale cohomology, volume 33 of Princeton Mathematical Series. Princeton University Press, Princeton, N.J. N OETHER , E. 1927. Abstrakter Aufbau der Idealtheorie in algebraischen Zahl- und Funktionenk rpern. Mathematische Annalen 96:2661. o O ORT, F. 1966. Algebraic group schemes in characteristic zero are reduced. Invent. Math. 2:7980. ROSENLICHT, M. 1956. Some basic theorems on algebraic groups. Amer. J. Math. 78:401443. ROSENLICHT, M. 1957. Some rationality questions on algebraic groups. Ann. Mat. Pura Appl. (4) 43:2550. S AAVEDRA R IVANO , N. 1972. Cat gories Tannakiennes. Springer-Verlag, Berlin. e S CHARLAU , W. 1985. Quadratic and Hermitian forms, volume 270 of Grundlehren der Mathematischen Wissenschaften. Springer-Verlag, Berlin. S ERRE , J.-P. 1993. G` bres. Enseign. Math. (2) 39:3385. e S PRINGER , T. A. 1998. Linear algebraic groups, volume 9 of Progress in Mathematics. Birkh user a Boston Inc., Boston, MA. S WEEDLER , M. E. 1969. Hopf algebras. Mathematics Lecture Note Series. W. A. Benjamin, Inc., New York. TAKEUCHI , M. 1972. A correspondence between Hopf ideals and sub-Hopf algebras. Manuscripta Math. 7:251270. VARADARAJAN , V. S. 2004. Supersymmetry for mathematicians: an introduction, volume 11 of Courant Lecture Notes in Mathematics. New York University Courant Institute of Mathematical Sciences, New York. WATERHOUSE , W. C. 1979. Introduction to afne group schemes, volume 66 of Graduate Texts in Mathematics. Springer-Verlag, New York.
Index of denitions
action, 51 continuous, 135 of a monoid, 50 acts through a character, 134 algebra afne, 40 Clifford, 156 etale, 117 graded, 154 opposite, 160 separable, 115 tensor, 156 algebraic group connected, 125 constant nite, 5, 15 diagonalizable, 132 general linear, 16 general linear GLn , 5 multiplicative type, 135 of monomial matrices, 18 orthogonal, 17, 154 reduced, 46 reduced attached to, 46 reductive, 7, 166 semisimple, 6, 166 smooth, 47 solvable, 7, 142 special linear, 16 special linear SLn , 5 special orthogonal, 154 symplectic, 17 unipotent, 7, 147 algebraic space regular, 47 smooth, 47 algebraic variety afne, 43 almost simple, 6 antipode, 31 automorphs, 154 basis orthogonal, 150 bi-algebra, 30 commutative, 30 bi-ideal, 39 category linear, 110, 171 neutral tannakian, 172 rigid tensor, 172 tensor, 171 centralize, 65 centre of an afne group, 64 character of an algebraic group, 129 closed immersion, 51 co-algebra, 28 co-commutative, 35 coaction, 76 coconnected, 146 coideal, 39 comodule cofree, 77 contragredient, 79 dual, 79 faithful, 88 nite-dimensional, 76 over a bialgebra, 79 right, 76 component identity, 124 components irreducible, 40 connected algebraic group, 5 contragredient of a representation, 83 correspondence Tannaka, 173 decomposition Jordan, 107, 108 dimension Krull, 43
179
180
of an algebraic group, 44 of an algebraic space, 43 dual Cartier, 36 of a representation, 83 Tannaka, 173 eigenspace generalized, 105 element group-like, 129 semisimple, 139 unipotent, 139 embedding, 56 endomorphism diagonalizable, 105 has all its eigenvalues, 105 locally nite, 108 nilpotent, 105 semisimple, 105 unipotent, 105 exact sequence, 67 ag full, 143 form quadratic, 150 function representative, 38 functor bre, 172 bred product, 23 representable, 11 tensor, 171 gradation, 174 graded, 154 group additive, 14 afne, 20 Clifford, 161 constant algebraic, 35 derived, 138, 140 multiplicative, 14 of monomial matrices, 8 solvable, 138 Tannaka, 173 trivial algebraic, 15 group algebra, 130 group space, 43 group variety, 44 height, 43 homomorphism injective, 56

of afne groups, 12 of bialgebras, 30 of co-algebras, 29 of comodules, 76 of monoids, 18 of superalgebras, 154 surjective, 65 homomorphisms of representations, 75 Hopf algebra coconnected, 146 ideal augmentation, 59 Hopf, 39 idempotent, 121 trivial, 121 idempotents complete set of orthogonal, 121 orthogonal, 121 index Witt, 154 inversion, 31 involution, 160 irreducible, 40 isogenous, 6 isogeny, 6 isometry, 150 kernel of a homomorphism, 58, 59 map lives in, 113 quotient, 65 tensor, 174 transposition, 29 monoid, 18 afne, 18 afne algebraic, 18 monoid space, 43 monoid variety, 44 noetherian, 40 normalize, 65 object dual, 172 monogenic, 110 trivial, 171 part semisimple, 107, 108 unipotent, 107, 108 plane
Index of denitions
hyperbolic, 154 point regular, 47 product almost direct, 6 semidirect, 68 semidirect dened by a map, 69 quadratic space anisotropic, 151 isotropic, 151 nondegenerate, 151 regular, 151 singular, 151 totally isotropic, 151 quotient by N , 67 of an algebraic group, 65 radical, 165 unipotent, 165 reduced, 43, 46 geometrically, 43 reection, 151 regular, 47 represent, 11 representation faithful, 75 nite-dimensional, 75 linear, 74 regular, 74, 76 semisimple, 166 simple, 167 ring coordinate, 42 regular, 47 series composition, 100 derived, 142 subnormal, 100 sheaf for the fpqc topology, 70 simple, 6 space quadratic, 150 spectrum max, 40 prime, 41 stabilizer, 89 subcategory replete, 90 subcoalgebra, 29 subcomodule, 78 subgroup afne, 56 algebraic, 56 subobject generated by, 110 subrepresentation, 75 superalgebra, 19, 154 commutative, 20 supergroup afne algebraic, 20 tannakian category algebraic, 173 TBA, 174 tensor product super, 155 topology Zariski, 40 torus, 7, 134 split, 134 transporter, 51 vector anisotropic, 150 isotropic, 150
181

Algebra Abstrata Ufrj Luciane

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Algebra Abstrata Ufrj Luciane

Uploaded by

Copyright:

Available Formats

C HAPTER

Basic Theory of Algebraic Groups

Bibliography Index of denitions

The building blocks

BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1)

S IMPLE GROUPS ; SEMISIMPLE GROUPS

authors say quasi-simple or simple.

G ROUPS OF MULTIPLICATIVE TYPE ; ALGEBRAIC TORI

BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1)

to the automorphism in (3).

General algebraic group

Afne algebraic group

Reductive algebraic groups

nite etale abelian variety semisimple torus unipotent

nite etale reductive semisimple torus torus f1g unipotent

F .A/ shows that

BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1)

! H.R/ ? ? yH. ! H.R0 /

T HE CANONICAL COORDINATE RING

BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1)

Gm .R/ ' f.a; b/ 2 R2 j ab D 1g ' Homk-alg .kX; Y =.X Y

Gm .R/ ' Homk-alg .kX; X Thus O.Gm / D kX; X

is a multiplicative group, and R

is a functor. Moreover, 1/; R/; 1/.

' Homk-alg .kX =.X n

is an afne algebraic group with O.

16 and so the functor R

BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1)

EndR-lin .R k V / (additive group)

fR .aij / D f .a11 ; : : : ; ann ; det.aij /

BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1)

dene a homomorphism of algebraic groups Ga ! SL2 .

BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1)

Relaxing the hypothesis on k

for all k-algebras R D R1

F INITELY GENERATED PROJECTIVE MODULES OVER COMMUTATIVE RINGS

(12) (13) (14)

BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1)

Some basic constructions

Throughout this section, k is a commutative ring.

Products of afne groups

H over k with coordinate ring O.G H / D O.G/ k O.H /; (15)

Fibred products of afne groups

BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1)

G2 / D O.G1 / O.H / O.G2 /.

/ .R/ D Ker..R/W G.R/ ! H.R//:

Extension of the base ring (extension of scalars)

Restriction of the base ring (restriction of scalars)

3. Some basic constructions

which sends .ai / to the family .bij / dened by the equations ai D Pd

gi 2 kY11 ; Y12 ; : : : ; Ynd :

BASIC T HEORY OF A LGEBRAIC G ROUPS (ALA1)

' k 0 k K=K .GK /I

.R/ D G.k 0 k R/ ' G.k 0 k K K R/ D k 0 k K=K .GK /.R/

3. Some basic constructions

Indeed, for any k-algebra R, .G/k 0 =k .R/ D G.k 0 R/

0 Gn .kn R/ 0 .Gn /kn =k .R/

' k 0 K=K GK '

The Greenberg functor

Afne groups and Hopf algebras

eA Ae / AA o Ak MMM oo MMM ooo o MM m ooo ' MMM ooo ' MM&  oo wo

On reversing the directions of the arrows, we obtain the notion of a co-algebra.

MMM MMM MM ' MMM MM

eA Ae / AA o Ak MMM oo MMM ooo o MM m ooo ' MMM ooo ' MM& oo wo