You are on page 1of 158
ms Lecture Notes in Pure and Applied Mathematics COORDINATOR OF THE EDITORIAL BOARD S, Kobayashi UNIVERSITY OF CALIFORNIA AT BERKELEY 1. N. Jacobson, Exceptional Lie Algebras 2. LA. Lindahl and F. Poulsen, Thin Sets in Harmonic Analysis 3. Satake, Classification Theory of Semi-Simple Algebraic Groups Other volumes in preparation Classification Theory of Semi-Simple Algebraic Groups |. SATAKE University of California, Berkeley With an Appendix by M.SUGIURA University of Tokyo Notes Prepared by Doris Schattschneider Moravian College, Bethlehem, Pennsylvania MARCEL DEKKER, INC. New York 1971 Taese Lecture jot in Mathem cs are produced directly from the author's typewritten notes. They are intended to make available to a vide eudience new developments in mathematical research and teaching that would normally be restricted to the author's classes and associates. ‘The publishers feel that this series will provide rapid, wide distribution of important material at a low price. COPYRIGHT © 1971 by MAR ALL RIGHTS RESERVED No part of this work may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, micro- film, and recording, or by any information storage and retrieval system, without permission in writing from the publisher. MARCEL DEKKER, INC. 95 Madison Avenue, New York, New York 10016 LIBRARY OF CONGRESS CATALOG CARD NUMBER 70-169183 ISBN NO. 0-8247-1607-8 PRINTED IN THE UNITED STATES OF AMERICA Foreword (added on April, 1971) This is essentially a photographical reproduction of my Lecture-Notes issued at the University of Chicago in 1967. Teking this opportunity of revision, I tried to make it more readeble, eliminating misprints and add- ing a few foot-notes, a new bibliography, an index of terms, and a list of notations. TI elso included an Appendix written by M. Sugiura of the University of Tokyo, which gives a very efficient way of classifying real simple algebraic groups in simplification of Araki's method. I should like to express here my gratitude to Sugiura for this invaluable addition to my Notes. My thenks are also due to S. Kobayashi, who invited me to join to the new program of Marcel Dekker mathematics series, to a number of my friends for their kind suggestions for improvements of the Notes, especially to Mrs. Doris Schattschneider for her constent assistance, and finelly to Mrs. Laura Hurbace for her fine job in typing these intricate materials. Ereface These notes are based on my course on "Classification-theory of semi-simple algebrai groups" given at the University of Chicago in the winter quarter of 1967. Though its primary“ain was to give a general idea of the classification-theory, I thought it convenient to include an outline of the basic theory of algebraic groups, in view of the fact that no standard textbook is as yet available. In this part, proofs are often very sketchy, or completely omitted, tut references are given to indicate where a more complete proof is to be found. Thus it is hoped that the graduate student with a sound background in algebra can easily seize the main idea without going into too mch detail. I gratefully acknowledge my debt to Mrs, Doris Schattschneider who Kindly helped ne in teking notes, reading proofs, and elaborating them in the form presented here. Table of Contents I, PRELIMINARIES ON ALGEBRAIC GROUPS 2... . Odod ea 81, Affine algebraic sets... 1... . 20 ao 22 Ll Definitions ...........0, we eeeee 1 1.2 Rational mappings... ... oe eee 82, Affine algebraic groups... eee eee eet eee ene 6 2.1 Definitions ......... Gcogcoa seen 6 2.2 Rational homomorphisms and quotient groups ....... 9 2.3 linear algebraic groups ...... Oe eis os 12 CS ey 83. K/k forms... ee. ee ae Coe 6 ole forms) and aneecocydlee oe | ie 3.2 Examples of K/k forms... ee ee ee eee +20 B pag teen OfaPrope lel 2 ee ee Os) 84, The structure of algebraic groups... . . eed 41 Semicsimple and unipotent elements ....... Bog059 4.2 Solvable grups 2... ee. eae G6 on 403 Borel subgroups «2... . ee ce 4e4 Questions of rationality ..... pee eee es 36 II, GENERAL PRINCIPLES OF CLASSIFICATION. ..... eee B 81. Structure of semi-simple algebraic groups... 1... 1 ee 2B 1.1 The root system and the Weyl group... ee... eee 3 1,2 Fundanentel systems and Weyl chambers... 1... .. 46 1.3. The fundamental theorem of Chevalley ..... : + SL 1.4 Some consequences of the fundamental theoren of Che VeL Oy ees ee sy 82, Structure of semi-simple algebraic groups defined over kk (k perfect)... eee ee ee ee te aed) Restricted: roots emer. see ee 2.2 Structure of parabolic subgroups»... - ee. ee B 2.3 The set 7 is a root system in a wider sense, ..... - 80 2.4 The fundamental theorem of classification. ....... 8b vii vidi 53. Classification of semi-simple algebraic groups defined Over & (kh perfect) 0c. eee 3.1 Investigation of T-diagrams...... 2.000. 3.2 Some special cases (finite fields) ........ 3.3 The invariant Y(G) madi fields and the classification over 3.4 Classification over the real field... 1.1... 3.5 Open questions (algebraic number fields) .. . Bibliography. Appendix: Classification over the real Foot-notes . List of notation Index of terms field, by Mitsuo Sugiura - 98 - 98 ol 10 123 126 127 128 ks aT 8 This chapter is an exposition of definitions and kmovn results. The bibliographical references following theorems, or titles of sections indicate where more details and proofs can be found. In a few instances, @ proof has been sketched here. | 81. Affine algebraic sets 1.1 Definitions ((4] Chap. IT and III; [1] Chap. I, §1) Notation: () : universal domain (i.e., a sufficiently large algebraically closed field) ON: Nedinensional affine space over 1 X= (ay y+00935)) = (x4)? @ Point in a O[x]= DL» Aye algebra of polynomials in N variables with coefficients in OMX): quotient field of O[X]. Definition: A subset Ac" is catied an (affine) algebraic set if there exists a subset 7 C [x] such that A = {xe ON | £(x) = 0 for all f eM}. (A is denoted as A(M), the algebraic set determined by ML we also write A< 71). Tet A = A(TL) be an algebraic set, and put O(a) = ff © Q(x] |t(x) = 0 for all xe ap. Clearly OL(A) is an ideal in ()[X], containing M. Since ()[X] is Noetherian, there exists a finite set of polynomials f),.-.,f, in Ox] which generate OL(A); then A = A(OL(A)) =, A(fyye+05f,)+ ‘Thus the cor= respondence between an algebraic set and its corresponding ideal OL(A) is one-to-one; we will use the notation A <—> Ol (& = OL(A)). | Tt is easy to see that if A, end Ay are algebraic sets in (N, with 1 2 Ay Ms Ay > Nyy then AU AL <> HLA HL, and AM Ay <— 1 * My (<> doos not hold for the latter). In general, if A <> OL.) 4 running through any set of indices, then nace O,+ Also, if A and B are algebraic sets in" ang Q™ respectively, with A <—> OL COE), B<> YC NI), then AxB is an algebraic set in O™ aoter- mined by UO[Y] + & [x]. Definition: An algebraic sot is called irreducible if A= Ay U Ay (AyyAy non-empty algebraic sets) implies A = A, or Aj. (An irreducible algebraic set is somstimes called a "variety.") Tt follows from the remarks above that an algebraic set A is irre- ducible if and only if O(a) is a prime ideal, Also, every algebraic set can be decomposed uniquely as a finite union of irreducible algebraic sets: n A v A;, A, irreducible (all i), and AHA, wif}. Now, let k be a subfield of }. if Mt is a subset of OLX], denote mt, = mn ux). Definition: An algebraic set A is keclosed if and only if there exists @ subset MC X{X] such that A OL, then A <> M%, If A is k-closed, then A° depends only on the restriction of 6 to k. Notation: For k CK, subfields of (), let k = algebraic closure of k; x! = inseparable closure of kj k° = separable closure of kj Aut(K/k) (or Gal(K/k)) the group of automorphisms of K leaving k pointwise fixed. opos: 1.1.1: For an algebraic set A, the following conditions are equivalent: 1) A is defined over x 2) A is keclosed. - 3) A =A for allo © Aut(O/k). - 3") A is Keclosed, and A” = A for allo € Gal(k/k). (1) => 2) => 3) is almost trivial; 3) => 1) follows from the Lemma of Weil on field of definition.) Corollary: If A is defined over k” and A is keclosed, then A is defined over ke From this proposition we see immediately that if k is a perfect field (i.e. k= 3), then the terms "k-closed" and “defined over k" for an algebraic set are synonymous. Later we will only be concerned with the case of k a perfect field. From the proposition (or the definition), it is easy to check that if Ay, Ay are knclosed algebraic sots in CW, then A, U dp and Ay Ay are also; in general, if fad is any collection of k-closed sets, then AA, 4s k-closed. Thus k-closed algebraic sots satisfy the usual topo- logical conditions of closed sets. The topology on (" having as its closed sets the k-closed algebraic sets is called the “Zariski-k~topology" (or "Zarisii topology" when k = 1). Unless otherwise specified, in all that follows, by “k-open" (resp., “open") and "k-closed" (resp., "closed") sets in at, we will always mean with respect to the Zariski-k (resp., Zariski) topology. It should be noted that the Zariski topology on ao does not satisfy the Hausdorff separation axiom; in fact, if 0) and 0, are any non-empty keopen subsets of fone then 0, 1 0, is also a non-empty k-open subset. In addition, any (relatively) open subset of an irreducible set A is necessarily dense in A. 1.2 Rational mappings ([4] Chap. IV; [1] Chap. I, 61) Definition: Let A be an algebraic set in On A polynomial function (defined over k) on A is the restriction to A of a function defined by a polynomial in [X] (resp., k{X]). A rational function (defined over k) on A is the restriction to A of a function defined by a rational quotient £/g in ()(X) (respe, k(X)), with g nob vanishing identically on each irreducible component of A. (This last condition is equivalent to: if A= UA, is the decomposition of A into irreducible components, and A, <—> fj, then g £ fon i). Notation: We denote by ()[4] (resp., K{A]) the ring of polynomial func- tions on A (defined over k), and by (1(A) (resp., k(A)) the ring of rational functions on A (defined over k). The ring Q[A] can be canonically identified with Q[X]/Ol(A), so that it is an integral domain when A is irreducible, and in that case, QA) is just the quotient field of OLA]. Definition: Let A be an irreducible algebraic set. The dimension of A is the transcendence degree of the field extension )(A)/Q. (We write dim A = dim((4)/0).) When A is irreducible, and A/k, one has ()[4] = x[a] @,4,; so that dim A = dim(k(A)/k). Definition: Let A and B te algebraic sets in A" ana O™ respectively. A polynomial (resp., rational) mp 9 from A to B is a mapping given by P= (PpreerGys % € OLA] (resp., g € AA), LS ism. Hise rational map fron A to B and each 9, is represented by £,/g, € OA), and x € A satisfies 8,(x) #0, 1<¢4M, then we say that p is defined at x, and the value of p at x is (x) = (p,(x),.--,q(x)) = (25-23) eB. We say that 9 is defined over k (we write 9/k) af 9, © KA] (resp., k(A)), Lg i gM. From the definitions, we see that a rational function on an alge- braic set A is a rational map fron A to oa =). It also follows that any rational map ? from A to B is defined on a non-empty open set in each irreducible component of A. In fact, if we denote by Ap the subset of points of A at which 9 is defined, then Ay - Vay (ay the irreducible components of A), and if @/k, then Ay 48 2 k-open set in A, (for all i). Proposition 1.2.]: A rational map p from A to B is a polynomial map if and only if &y = Definition: We say a polynomial map 9 is a birational isomorphism if p is bijective and gt is also a polynomial map. Notation: If 9 is e rational map from A to B, and M is any subset of A, then denote by 9(M) the set-theoretic image of My = 4p AM by p in Bs and denote by g(M) the Zariski-closure af @(M) in B. (9(M) is called the algebraic image of M by 9.) 6 Tf Ais a keclosed algebraic sot in Q", donote 4, = ANK. (Ay 4s called the set of kerational points of A.) If 9 is a rational mp fron A to B, with g/k, & and B keclosed, then clearly o(A,) C.B,j in particular, if 9 is an isomorphism, 9 gives 4 one~ to-ons correspondence between 4, and B,. The following proposition sums up some facts relating g(A) and 9(A)+ Proposition 1.22: Let A and B be algebraic sets, A irreducible, and 9 8 rational map from A to B. Thent 1) 9(A) contains a set which is relatively open in (A). In fact, if U is any non-empty (relatively) open subset of A, then p(U) contains a subset which is relatively open in p(A). 2) (A) is irreducible. 3) If A is defined over k andy is defined over k, then 9(A) is defined over ke If A and B are irreducible algebraic sets, and 9 is a surjective rational map from A to B, thon there is a natural injection (1(A) <— Q(B) given by ¥op — } (Pe (B)). Under this injection, (1(B) can be identified with a subfield of ()(A), and we make the following defi- nition. Definition: The degree of 9, denoted deg gp, is the degree [(1(4):(B)], Af this is finite; otherwise the degree of p is zero. We call 9 insep- erable (resp., separable) if ()(A)/(.(B) is @ purely inseparable (resp., separable) extension. $2. Affine algebraic groups ([1] Chap. 1; [2] exposé 3; [13] Chap. 1) 2e1 Definitions Definition: G is called an (affine) algsbraic group if 1) G is an abstract group; 2) G is an algebraic set in ar; 3) The mapping GxG >G is a polynomial map. (x,y) > xy G is defined over k (write G/k) if G as an algebraic set is defined over k, and the mapping in 3) is defined over ks If G is an algebraic group, then for any fixed a € G, the left (resp., right) translation Lyi x>ax, xed ,: is an automorphism of G with respect to the structure of an algebraic x-—>xa, xe G) set. Since left translations are trahsitive, G is a "homogeneous" al- gebraic set; in particular, G has no "singular" points. [1] These facts are used in the proofs of some of the properties of algebraic groups. If G is an algebraic group defined over k, then the identity ele- nent of G is k-rational, and it is easily seen that G, is an abstract group. If G is an algebraic group and G° is an irreducible component of G containing the identity element, 1, then it can be shown that G° is the ply irreducible component of G containing 1. Further, we havet Proposition 2.1.1: irreducible component of G containing 1, Then G° is a normal, algebraic let G be an algebraic group defined over k, G° the subgroup of G, defined over k, and G= UL) g,G°, the coset decomposition of G with respect to G°, is the decomposition of G into irreducible com ponents. From this proposition, we see that an algebraic group G is irre- gucible if and only if it is e connected set in the Zariski topology. 8 (Note: the words "connected" and "irreducible" are not interchangeable for en arbitrary algebraic set A.) Also from this proposition, we see that the dimension of each of the irreducible components of G is the same as dim G°; thus we have the following Definition: The dimension of an algebraic group G, denoted dim G, is equal to dim 6°, Examples of gleebraic groups Ex. 1. G= G, ={), the "additive" group of qn. G, is defined by the zero polynomial, i.e-, 6, = A(O). dim 6, Ex. 2, G= 6.2, the "mltiplicative" group of 1). 2 = = Gc’, and G= a0 - 1). din 6 =1. Ex. 3. G=SL(n), the "special linear group." Su(n) € OF, and Sita) = a(det(X,,) ~ 2). Ex. 4e G=GL(n), the "general linear group." GL(n) c a, and GL(n) = Adee (i, 5) -1). Al1 of these groups are connected (since their corresponding ideals, peing generated by an irreducible polynomial, are prime), and all are defined over the prine field. Note that when k is a topological field, then the group GL(n),, = GL(n,k) becomes a topological group with respect to the natural topology on "1, With respect to this natural topology, it can be shown that GL(n,€) is connected, GL(n, JR) has two connected components, and GL(n, Q,) is totally discomected. Thus, the Zariski (k-)topology and the natural topology should be carefully distinguished. Ex. 5. Let G), G, be algebraic groups in O” ana 0" respectively. Then GX Gy C CO" 45 also an algebraic group, and is called the direct 9 product of G, and G,, If G) and G, are both defined over k, then G, x G, is also defined over k. 2.2 Rational homomorphisms and quotient groups. Definition: Let G be an algebraic group defined over kj HCG is a ke closed (resp., defined over k) subgroup if 1) H is an abstract subgroup of G, and 2) His a k-closed subset (resp. defined over k) of G. Definition: Let G, Gt be algetraic groups. A mapping 9: G > G! is called a (rational) homomorphism defined over k if 1) @ is a homomorphism of abstract groups, 2) p is a rational map defined over k. Such a mapping 9 is also called a k-homomorphisn of G into Gt; it / should be noted that conditions 1) and 2) imply that g is in facta \ polynomial map (since @ is necessarily defined on all of G). : Definition: A rational homomorphism from an algebraic group G into 6, is called a character of G. Proposition 2.2.1: Let G, G! be algebraic groups, with G/k, and pa ke homomorphism of G into Gt, 1) @ 22) = kor @ is a k-closed normal subgroup of G. 2) p(G) = Img is a closed subgroup of G!, defined over k. 3) dim(Im yp) = dim G - dim(ker 9). As a special case of part 3) above, we see that if G and G! are \\connected and have the same dimension, then @ is surjective if and only if ker p is finite. Definition: let G, G' be connected algebraic groups. 10 A rational homomorphism of G onto G! having finite kernel is called an isogeny!) If g is an isogeny from G onto G! then by definition, deg @ = [A(e): O(G!)]; af G and g are both defined over k, then it can be shown that deg 9 = [k(G): k(G)]. Proposition 2.2.2: Let 9 be an isogeny from G onto G!. The mmber of elements in ker 9 equals [k(G): k(G')],, the degree of the maximal sepa rable extension of k(G') contained in k(G). ([4], Chap. IV, 3) ‘Thus en isogeny @ is bijective if and only if the extension k(G)/k(G') is purely inseparable (i.e., @ is inseparable). Definition: Let G, G! be algebraic groups. A mapping 9: G—> G! is called a (rational) k-isomorphism if 9 is a bijective rational k-homo~ morphism having as inverse a rational k-homomorphism, If G = G', we call p a k-gutonorvhism of G; the group of k-automorphisms of @ is de~ noted Aut, (G). It 4s clear from remarks above that if p is an isomorphism, then deg 9 = 1. Remark: Not every inseparable isogeny is an isonorphism, as is shown in the following example ("Frobenius endomorphism"). let G be an algebraic group defined over k = IF, the finite field having q elements. let 9 be the rational mapping from G into itself de~ fined by x = (x,) > (x,5) = @)®, xeG. (p maps G into itself since for each x € Gand £¢ GL(G),, one has 0 = £(x) = (#(x))* = 2te9) 9 The image of k(G) under the injection induced ty @ is equal to k(G)*; in fact, if } e k(G), then (vop)(x) = (W(x) ]*, for all xe G. Clearly g is defined over k, and is an automorphism of G as an abstract group, but 9 n is not necessarily an isomorphism since [k(G): k(G)°} > 1 in general. Quotient groups ([1]-I, 85; [2], 3-05 and exposé 8; (13]-II) let G be an algebraic group, defined over k, and N a closed normal subgroup, also defined over k. The following proposition allows us to speak of "the quotient group" G/N, which is an algebraic group defined over ke Proposition 2.2.3: There exists an algebraic group G, defined over k, and a surjective homomorphism r, defined over k, 1: G—> G which satisfies: © i) ker n= ' 4i) if 9 is a rational homomorphism, defined over k, from G into an algetraic group G!, satisfying Nc ker g, then there exists a unique homomorphism 9, defined over k such that 9 = gon. By ‘the universal mapping property (ii) we see that G is unique up to k-isonorphism; the group G is denoted G/N. One of the key lemmas used to prove this result is due to Chevalley: Loma 2.2.4: Let G be an algebraic group, and N a closed invariant sub- group, both defined over k. Then there exists a rational representation @ (defined over k) of G having Nas kernel ([2], 4-04). If G is a comected algebraic group, then k(G/N) can be identified with a subfield of k(G) (since 1 is surjective). The group N acts on the field k(G) by ¥—> YoR,,aeN, The following proposition characterizes G/N. Prombsition 222.5: k(G/N) is the subfield of k(G) consisting of all N- invariant functions; i.e., k(G/N) = {p € k(G)]¥oR, =p, all a c Nf. 12 2.3 Linear algebraic groups Definition: A (finite dimensional) vector space V is said to be defined over k (or have kestructure) if 1) V is a vector space over (}, 2) there is a given a subset vy of V which is a vector space over k, spanned over k by an %&basis of V , (hence one has V = V, a). Throughout, unless otherwise specified V will be defined over k; and we fix the kestructure once end for all. If K is another subfield of Q, kcKCO, then defining Vy = V,@,, X, we see that V has ‘the natural structure of a vector space defined over K. Notation: GL(V) is the group of non-singular linear transformations of “5 GL(V) is called the general linear group of V. If we take a basis (egy ++98,) of V, over k, then with respect to this basis, there is a natural isomorphism p: GL(V) ~ Gi(n). Ifa second basis of V, is taken and p! is the isomorphism of GL(V) into GL(n) with respect to then it is easy to see that p = pop! where 9 is an immer automorphism of GL(n), defined over k (given by the change of basis matrix). Thus the identification of GL(V) with GL(n) is unique up to an imer automorphism defined over k. Definition: A subgroup Gc GL(¥) is a linear algebraic group, defined over k (write G/k), if under the isomorphism p described above, p(G) is an affine algebraic group defined over ke Notation: If G is a linear algebraic group, denote by G, the inverse image of P(E), by p. (Clearly G, is uniquely determined.) We call the elenents of G, the k-rational points of G. 13 The following proposition, easily proved, shows that linear algebraic groups end affine algebraic groups are essentially the same. Proposition 2.3.]: Any affine algebraic group is isomorphic to a linear algebraic group. ([2], 4-03). Examples. 1. SL(V) ={g € GL(V)|det g = 1} (special Linear group of V). The sequence of mappings 1> suv) > av) 2s G6 > is exact, and it can be shown that G_ and det satisfy the universal mapping property (ii) of Prop. 2.2.3. Thus G. = GL(V)/SL(V). 2, Let 2 = center of GL(V) = scalar miltiplications. The quotient group GL(V)/2 is denoted PL(V), called the projective linear group of V. The realization of PL(V) as a matrix group is as follows: let V* = dual of V, and m the mapping of GL(V) > GL(V* @V) defined by ge> sg @g, ge GL(V). Then ker = Z, and the pair n(G), 7 satis- fies Gi)of Prop. 2.2.3, so that 0(G) = PL(V). 3. Let B be a non-degenerate bilinear form on VxV. (Recall that two bilinear forms B, B! are equivalent, BB, if and only if there exists an elenent g GL(V) such that BY (x,y) = Blgx,ey) for all (xyy) EVKV. If (e),++.,¢,) 48 a basis for V,, and B is identified with the matrix (Ble, 505), thén BB! if and only if BY = tape for some g € GL(n).) If B=S 4s a symmetric form, then define O(V,S) = {g GL(V)|S = tgs g}, the orthogonal group of V with respect to 8; SO(V,S) = O(V,8) 0 SL(V), the special orthogonal group of V with respect to S. If B = A is an alternating form, then define uk Sp(V,A) = {g © GL(V)|A = *gAg}, the symplectic group of V with respect to A. We say B is defined over k (write B/k) if the matrix (80,303) & GL(n),3 clearly this does not depend on the choice of the basis (ey y+++58,) of V- If B/K, then the three groups just defined ere all defined over k. If two bilinear forms B, B! are both defined over k, we will write BoyBt if Bt = “sBg with g k-rational, It is known that if char k # 2, then there is only one equivalence class over k of alternating forms. Thus if A./k is an alternating form, we can choose a basis of V,, such that A, has matrix (Ch win), n= an, (Wo write Sp(V,A,) = Sp(m).) Tf A~A,, then A=g Aig, and det(g) is 2 polynomial func— tion of A, uniquely determined by A; det(g) is called the "Pfaffian" of A relative to Aj. If g, © Sp(VyA), then A = ‘gage = *(ee)A (ee) 80 det(gg,) = det(g), hence det g, = 1. Tams Sp(V,A) ¢ suv). Finally, it is known that S0(V,S) and Sp(V,A) are connected algebraic groups'(and are simple groups except for SO(V,S) with dim V = 1,254). 2.4 Tori ((1], Chap. I; [2], exposé 4; [5]) Definition: An algebraic group G is called a torus if, for some n, there exists an isouorphism @ of G onto (G,)". G is said to be k-triviel (or split over k) if G and p are both defined over ke Throughout this section, we will use T to denote a torus, and X = X(T) the group of characters of T. We use additive notation for X (i.e., ty definition, (A, + X4)(+) = Ay (e)A,(t) for A,,Ay © X, te T) so that X is a Z-module. The group of diagonal matrices D(n) in Gi(n) is a torus (w (6,)") 15 in OF", saith D(n) <> OL = (y+oK,¥ = 1), and clearly D(n) splits over the prime field, Thus we identify D(n) = (6,)". let T= (6) there are n "canonical" characters Ky of T defined by: if x = (xy ++6)%,) © T, then A,(x) = x,. Under the identification M{T] = Oly,- no the function Qy Ay) 1 is identified with ¥ (nod OX). From this it can be showm thet (1) Olt) = OLAS... AR, (2) 02) = OA, a purely transcendental extension of (), (3) K is just the subset of YO, the character X, is identified with X, (mod OL), and o,,) is in Q(T]. This last fact shows that as a Z-nodule, monomials _ xen, Now let T be any torus and X its character group. ~As might be ex= pected, there is "duality" between certain submodules of X and subgroups of Tj although we restrict to algebraic subgroups of T, most of the usual duality results hold. If T, is a closed subgroup of T, denote ty T} the submodule of X: Tr =(ex|At) =1, all te Th. Similarly, if X, de a submodule of X, denote ty Xj the closed subgroup of T defined by: XT = {te TAlt) = » all Xe Xs}. We have the following Proposition 2.4.1: YAP, A> x} define reciprocal bijections between the set of closed subgroups of T and the set of submodules of X satisfying the condition X/X, has no p-torsion (p = characteristic of the prime field; if p = 0 there is no condition on torsion). 2) A closed subgroup T; CT is connected if and only if X/K, has no tor- fo sion, where X) + 1, is a torus if end only if it is connected. 16 3) X(T) = X/K,, H(T/T,) =X, where X, = tT. Two key lemmas used to prove this proposition are: Lema 2.4.2 (Chevalley): Let H be a keclosed subgroup of an algebraic group G/k. Then there exists a rational representation p/k of G into GL(V) and a vector v, € V,, such that H = {g & Glp(e)vy=Ale)v, Ce) & OF. (A is @ character on H.) ([2] ,4-03) Lemma 2.423: Any rational representation of a torus T is completely reducible, Let T and T! be tori with character modules X, X' respectively. If @ is e homomorphism of T into 7’, then the mapping to of X! into X, de- fined by *e(Xt) = Xt, A! eK", is a module homomorphism. Conversely, if § is a module homomorphism of X' into X, then there exists a rational homonorphism @ of T into T! such that ~ = “p. To see this, we can iden tify T with (6.)" and T! with (ens then X = Pareeohle, and X! = {By's-++yAql}y » and so there corresponds to an integral matrix (m5) n where 9(X,') ? 1) ck (Prop. 2A), then T is defined over k if and only if X, is a T-submodule of X. 2) let T and T! be tori defined over k, and p a homomorphism from T into 3, 9 4s defined over k if and only if *p is a T-homonorphismy in par ticular, @ is a k-isonorphism if and only if “p is a I-iscuorphism. This proposition leads to the following main theorem on tori: Theorem 2.448: There is a one-to-one correspondence between the cate- gory of tori defined over k and the category of finitely generated torsion-free P-modules. To complete the theorem, it mst be shown that to each such module X, there corresponds a unique (up to k-isonorphism) torus T such that X(T) and X are isomorphic as T-modules. This leads us to the next topic. 83. K/e-Lorms, 3.1 K/keforms and one cocycles Definition: Let k and K ve subfields of 1, k CK, and G, an algebraic group defined over K. A pair (G,f) is called a K/k-form of G if G is 19 an algebraic group defined over k, and £ is an isomorphism of G onto G,, f defined over K, A pair (G,f) is called a keform of G, if (G,f) is a K/k-form of G, for some (sufficiently large) extension K of ke We are only interested in the case where k is perfect, and K/k is a finite extension, and make these assumptions for the remainder of this section. let P= Gal(k/k) (note k= 1°). If G, is an algebraic group defined over K and (G,f) is a K/k-form of G,, then any elenent ¢ € I acts on the coefficients of the polynomial mapping f; let # denote this new mapping. Since G/k, (7 = G, so thet £° is an isomorphism of G onto G9. Define 9 is = £% 27}; q, is then 6 K-isonorphion of G, onto its conjugate G,' «—> 4, ,, WN bo fy From the definition of 9,, it is easy to check that the system of iso- morphisms ag gp satisfies the condition (a) % oe, = Although [ is an infinite group, the system {Q)g er is essentially finite, for f is defined over K, and from the definition of %,, it is clear that (2) qq depends only on the restriction of 0 to K, and q, is defined over K' for some finite extension K! of ky A nain result on K/k-forms says that conversely, any such system of mappings Os oe p Satisfying (1) and (2) necessarily belongs to a K/k-form. 20 Proposition 3elel: Let G, be an algebraic group defined over K, and : A ey 6% i (ger 2 system of isomorphisms (9? G %> Gy '), satisfying (1) and (2). ‘Then there exists a K/k-form, (G,f), of G, such that q = Port for alloel. Wo will give a proof of this proposition later; it is true, in fact, Af G, is ony algebraic set, (See A. Weil, "The Field of Definition of a Variety" —A.M.S., Vol. 78) Notation: We will denote by (9%) a system of mappings as described in Prop. 3.1.2 (dropping the obvious index set T), and call it the system corresponding to the K/k-form (G,f).+ Definition: Let (G',2) and (G,£) be k-forms of G). We say that (a',£") are ismorphic if there is a k-isomorphism of G onto Gt. Let (G,£) and (G!,£") be isomorphic K/k-forms of G), (gg) and (95") their corresponding systems, and P a k~isomorphiom of G onto G'. Define a mapping } ty ¥ = topes}; ¥ is a K-autonorphism of G,. e 4>4, fo at >a, Tf we apply 6 € [ to the equation defining }, we have (since p/k) 1 = 9gte 929, 3 o _ -1 yar opot? =g,testopor oe Thus (3) Definition: Let G,/K. Two systems of isomorphisms (9%) (9515 where Gg and Gy! mAP Gy onto G,°, and satisfy (2) and (2) are said to be K-equivalent (resp., equivalent) if there exists 4 K-automor= phism (resp., automorphism) p of G) satisfying (3). 21 sition 3.1.2: Let G,/K, (G,f) and (G',2") K/k forms (resp., k- forms) of G. (G,f) and (G!,f1) are isomorphic if and only if their corresponding systems (95) and (9, Qg') are K-equivalent (resp., equivalent). We have just proved this statement in one direction. Conversely, if (95) and (%") are systems which are K-equivalent (or equivalent) under ¥ € Aut,(G,), we know there exist K/k-forms (G,f) and (G',f') of G, corresponding to these systems (Prop. 3.1.1). Defining p = tlopor, clearly p is an isomorphism of G onto G', and p° = 11 pos = (tr og Nolo ropog tol, of) = £17 yor py 50 p/k. ‘Thus we have: Gore! 1,3! The k-isomorphism classes of K/k-forms of G,/K are in one-to-one correspondence with the K-equivalence classes of systems (94) (et Gy > @,°) satisfying (1) and (2). Now assume G, is defined over k and K/k is @ Galois extension, Then G,9 = G, for allg eI, and the elenents 9, of a system (P,) corresponding to a K/k-form (G,f) of G, are ali K-automorphisms of Gy (since K/k is Galois, X= K, allg € I, so £/K implies £9K, hence For) = eft) Condition (1) implies that the system (9. -g) is a one-cocycle of Tr in duty (G)), and condition (3) shows that "cohomologous" and "K-equivalent"™ are synonymous terms for a pair of systems (5) and @y') Condition (2) implies that the system (p,) can be regarded as a one-cocycle of Gal(K/k) in Aut,(G,). If we denote by w1(Gal (K/«) 5 Aut, (G))) the first cohomology set of Gal(K/k) in Aut,(G,), thon Corollary 3.1.3 becones: Corollary 3.144: one-to-one correspondence with the elements of HE" (GaL(K/x) yAuty (05) « The k~isomorphism classes of K/k-forms of G,/k are in In the case of a k-form (G,f) of G,, we see that conditions (1) and

You might also like