ROBERT L. GRIESS, JR ELEMENTARY ABELIAN p-SUBGROUPS OF ALGEBRAIC GROUPS Dedicated to Jacques Tits for his sixtieth birthday .TCARTSBA Let ~ be an algebraically closed field and let G be a finite-dimensional algebraic group over N which is nearly simple, i.e. the connected component of the identity G O is perfect, C6(G °) = Z(G )° and G°/Z(G )° is simple. We classify maximal elementary abelian p-subgroups of G which consist of semisimple elements, i.e. for all primes p ~ char K. Call a group quasisimple if it is perfect and is simple modulo the center. Call a subset of an algebraic group total if it is in a toms; otherwise nontoral. For several quasisimple algebraic groups and p = 2, we define complexity, and give local criteria for whether an elementary abelian 2-subgroup of G is total. For all primes, we analyze the nontoral examples, include a classification of all the maximal elementary abelian p-groups, many of the nonmaximal ones, discuss their normalizers and fusion (i.e. how conjugacy classes of the ambient algebraic group meet the subgroup). For some cases, we give a very detailed discussion, e.g. p = 3 and G of type E6, E 7 and E .8 We explain how the presence of spin up and spin down elements influences the structure of projectively elementary abelian 2-groups in Spin(2n, C). Examples of an elementary abelian group which is nontoral in one algebraic group but toral in a larger one are noted. Two subsets of a maximal torus are conjugate in G iff they are conjugate in the normalizer of the torus; this observation, with our discussion of the nontoral cases, gives a detailed guide to the possibilities for the embedding of an elementary abelian p-group in G. To give an application of our methods, we study extraspecial p-groups in E8(~). 1. INTRODUCTION, NOTATION AND STATEMENT OF RESULTS In this article, ~ denotes an algebraically closed field and p is a rational prime unequal to char(X). The symbol i denotes a square root of -1 when char(K) ~ .2 By our conventions, an element of order p in an algebraic group over ~ is semisimple. We study elementary abelian p-subgroups of all nearly simple finite- dimensional algebraic groups, classify the maximal nontoral ones and discuss aspects of their embeddings and fusion. Since our methods come from both Lie theory and finite group theory, we prefer to give a lot of detail. It would be hard to give a historical report of interest in elementary abelian subgroups of algebraic groups. Recent references would include the 1961 paper Bo; many questions related to these elementary abelian subgroups for finite groups and algebraic groups in positive characteristic came up during the intense study of finite simple groups, starting in the late 1950s; see GoLy, which lists extensive properties of finite nearly simple groups; the Geometriae Dedicata 39: 253-305, 1991. © 1991 Kluwer Academic Publishers. Printed ni the Netherlands. 254 ROBERT L. GRIESS, JR 1986 preprint lAd suggested several ideas, one of which led this author to the complexity concept (see (1.2)), which seems to be new. The classifications of finite subgroups of algebraic groups (e.g. CoWa and CoGr) raised further questions along this line. The 1974 paper Alek should be mentioned since it seems not to be well known; indeed the groups in the main theorem of Alek were discovered independently by several mathematicians (including this author). The main results are summarized in Tables I, II and III in (1.8). The possibilities for a maximal projectively elementary abelian group follow from those tables. Throughout this article, however, are details and comments about embeddings of finite groups in algebraic groups. We explain a bit more how to get the possibilities for a maximal nontoral elementary abelian p-group from Tables I and tI. Let G be a nearly simple algebraic group, Z ~< )G(Z and let E be a p-subgroup of G such that EZ/Z is elementary abelian and maximal such. If E <~ G °, this is straightforward, though in the case G O < G, to settle maximality of E, one must determine whether an outer automorphism of G O of order p centralizes E. Now, assume E is not in G °. Let F := E n G O and let e generate a complement to F in E. Then, e induces an outer automorphism on G O and one may get the possibilities for its fixed point subgroup H from Table III. One then gets F by consulting Table I or II for H and considering the possibility that F is toral. This is straightforward, except possibly for type D, with p = 2 since H involves two semisimple or toral components. For types A, and ,D there are many possibilities, so we do not list them in these tables. In the case of type A,, the fixed point subgroup ofe has type D m or ,mC and we consult (2.19). In the case of type D, and p = 2, we just observe that F has elementary abelian image in SO(2n, ~) since it commutes with the determinant - 1 action of e; thus, depending on Z, F corresponds to a frame group or to a weakly self- orthogonal binary code in the dimension 2n diagonal frame group whose annihilator contains an odd diagonal transformation corresponding to the element e; see (2.2), (2.7) and (2.8). One does not expect to enumerate such codes explicitly. In case G is disconnected of type E6, there are two conjugacy classes of maximal nontorals not contained in G O since the rank 5 nontoral of a natural F 4 subgroup remains nontoral in G O and is conjugate to one of the two nontorals of a natural C4 subgroup; the other nontoral becomes toral in G and so is conjugate to the natural rank 7 nontoral q-(2)(t), where t inverts the ambient torus T. Finally, in the case of type/)4 and p = 3, we need the result that the toral rank 2 groups in the fixed point subgroups PSL(3,~) and G2(~) are conjugate in ;OG see (2.25). It may happen that if we take an element deE\F such that (d) ~ (e), (d) ELEMENTARY ABELIAN p-SUBGROUPS 255 and (e) are not conjugate in G. Thus, the same group E will come up in two different ways in the use of (2.18). This happens, for example, with maximal elementary abelian groups for type D. and E 6. (1.1) NOTATION. When G is a finite-dimensional algebraic group, T refers to a maximal torus of G and Tk refers to a k-dimensional torus in G; -~ is unique up to conjugacy, though T k is not if 0 < k < rk(G). For an integer n, "~k(n)" we let ).(-q denote {t e Tit" = }1 - ~,Z ,)~(k and similarly define (1.2) NOTATION. 'tp' refers to a tensor product situation, which includes (i) a decomposition of a vector space V= 1V ® 2V with a bilinear form f (possibly 0) for which each ~V has a bilinear formf~ (alternating or symmetric) such that f=fl ®f2; and (ii) some corresponding central product decom- position of groups A = A 1 o A2 ' where Ai <~ Aut(f~) and A <~ Aut(f) <~ GL(V). For the next definition, recall that any scalar-valued function on a finite vector space must be polynomial (as a function of the coordinates). Such a function determines a coset of the ideal of polynomial functions vanishing on the vector space; the degree is defined as the least degree of a polynomial in that coset (- oo is the degree of the 0-function). (1.3) NOTATION. Let G be a group and let S be a subset of G. For an elementary abelian subgroup E of G, the complexity cx(E) (based on S) is the degree in the above sense of the characteristic function of S n E in E. In this article, complexity is defined in a few specific situations; see Sections 6, 7, 8, 9 and .31 (1.4) DEFINITION. Let A, B,... be a (finite or infinite) sequence of conjugacy classes in the group G and let S be a subset of G. The distribution or class distribution of S (with respect to the understood sequence) is the sequence of symbols AaB b .... where a = Ac~S, b = BnSI,.... Often, S is a subgroup and we restrict the sequence of classes to those which meet S\{I}. In the case where S\{1} consists of elements from the conjugacy class A, we say that S is A-pure. When S is a group of exponent p and the classes of order p are designated pA, pB,.., we write A"Bb.,. for the distribution pAapB b .... (1.5) REMARK. Orthogonality relations for finite groups are used several times; in all cases, the finite group is a p-group; its eigenvalues may be lifted to characteristic 0, by the usual Brauer theory. In characteristic 0, we compute the dimension of its fixed point subspace by summing traces. Actually, such a finite group may even be lifted to a subgroup of the corresponding type algebraic group in characteristic 0 (see Appendix .)2 256 ROBERT L. GRIESS, JR (1.6) DEFINITION. Assume char N ¢ 2. By pin(m, ~), we mean a double cover of O(m, ~) which restricts to the double cover Spin(m, ~) of SO(m, ~); the isomorphism type depends on whether a reflection lifts to an involution or an element of order 4. Let G = Spin(m, ~). The spin module is irreducible of dimension 2 ("- 2/)1 if m is odd. When m is even, it decomposes into two irreducible and inequivalent modules, called half-spin modules; each has dimension 2 "/z-1. These modules are interchanged under the action of elements from pin(m, N)\Spin(m, ~). If M is one of these, the kernel of the action of G on M is Z, a subgroup of order 2 if m = 0 (mod 4) and order 1 if m =- 2 (mod 4). The group G/Z is called the half-spin 9roup. When char ~ = 2, pin(n, )41f ~- O(n, ~). (1.7) NOTATIONS. Notation for finite groups and for algebraic concepts in Lie theory are standard. The style of usage is consistent with that of CoGr (the relevant notations and results extend appropriately to algebraic groups) and we call attention to the following group extension notations. Let A and B be groups. We write A. B for a group with normal subgroup isomorphic to A and quotient B. In the case where the extension is split, we write A : B, and if nonsplit, we write A.B. We write n for a cyclic group, p" for an elementary abelian p-group, b.ap for an elementary abelian group which is regarded as the tensor product of elementary abelian groups ap and bp (useful when the latter are a pair of modules for a pair of commuting groups), ,..+b+ap for an extension of (going upward) ~p by bp by .... For a finite group G, exp(G) denotes the exponent, i.e. the least positive integer n such that x" = ,1 for all x E G. A p-group P is extraspecial if P' = Z(P) has order p; there are two isomorphism types for P of a given order, which is an integer of the form pl + ;rz we write pl + r2 for an extraspecial group of order pl + 2r, ~p + r2 for such a group of type e = nehw(+__ p = 2, this refers to the Witt index of the quadratic form obtained from the squaring map; when p is odd, e = + iff the exponent is p). For a subset Sofa group, let S # denote S\{1}. See also IGor, Hup. We let G be a finite-dimensional algebraic group. If X ~ {A, B, C, D, E, F, G}, m and n are integers such that X, is a type of simple algebraic group, mX, denotes a perfect group whose center has order m and whose central quotient is the simple algebraic group of type X,, when such a group exists. If X, Y,... are such symbols, manb... Xk Y~... denotes a perfect central extension of the semisimple group XkYt... by a finite abelian group of type manb... (i.e. ,~Z × ~Z × ."). When p = char(~i) divides such m, it is understood that m is replaced by its p' part, e.g. Z(3E6(K)) = 1 when p = .3 We write J" + J, + -.. for the Jordan canonical form which is the sum of indecomposable blocks of dimensions m, n .... with diagonal entries .1 We ELEMENTARY ABELIAN p-SUBGROUPS 752 introduce the following departures from usual practice: Dih2,, denotes the dihedral group of order ;m2 Quatz,, denotes the quaternion group of order .m2 )8.1( THE TABLES. The notational conventions are these: G is a simply connected nearly simple algebraic group such that G o is simple modulo G(Z )° and Z is a subgroup of .)°G(Z We study elementary abelian p-subgroups of G/Z by considering groups E which are p-subgroups of G such that EZ/Z si elementary abelian. We discuss all maximal nontoral E and a few others. We use 'wr' to indicate wreath product and AGL(n, F) for the affine general linear group of degree n over the field F, i.e. the split extension of GL(n. F) by the group of translations. For the definition of complexity, see (1.3) and the appropriate section ,6( ,7 8 or .)9 We abbreviate complexity with 'cx.' TABLEI Maximalnontorals~rtheconnectedcase, p=2 Type Z Remarks A, ZI odd None Z I even E in 2, x + ~2 o oz/7 x 22-1, 2rs = n + ,1 r > 0, c ~> ;1 tp n n IZ = 1 E corresponds to a nearly self-orthogonal even code in a diagonal frame group (see (2.7)) IzI = 2 E/Z is a frame group (see (2.7)) C, IZl = 1 None IZl = 2 E has shape 2~+2'o4~ x 2 ,b 2's = 2n, r > 0, a~{0, ;}1 for e = -, (a, b) = (0, s) or (1, s/2-1); for e = +, a + b = s/2; tp; not all of these are maximal; see (2.19.ii) ~D lZI = 1 E corresponds to a nearly self-orthogonal code in a diagonal frame group IZI=2 Z is the kernel of the natural 2n-dimensional representation: E/Z (several is a frame group for G/Z cases) Z the kernel of a half-spin representation (here, n is even) E is in one of the two natural T1SL(n, K) .2-subgroups stabilizing a complementary pair of maximal isotropic subspaces in the natural module; the image in SO(2n, ~;) of E lies in one of the subgroups Quat s o Sp(n, ~) or Dih s o O(n, N;), and has the form X o ,Y where X lies in the finite central factor and where Y has shape 2, a ÷ ~2 x 2s; tp; see (2.19). lZI = 4 E is the preimage in G of a subgroup of SO(2n, H) of the form 2~+2"o4a x 2 ,b 2rs=2n, r > 0 and a~{0, 1}; b=s if a=0; b = s/2 - 1 if a = 1; the commutator subgroup E' may be any subgroup of Z; see (2.19) and (2.22). 2G lZI For each integer k e {0, ,1 2, 3}, one conjugacy class of = 1 elementary abelian subgroups of order ;k2 toral iff k <~ 2 iff cx <~ 2; if E is maximal such, N(E) ~ 23. GL(3, 2). 258 ROBERT L. GRIESS, JR TABLE I (Cont.) Maximal nontorals for the connected case, p = 2 Type Z Remarks F4 Z = 1 One conjugacy class of maximal elementary abelian groups, represented, say, by E of rank 5; a subgroup is toral iff the complexity is at most 2 iff it does not contain a 2A-pure eights group; N(E) ~ 25.223.GL(3,2)× GL(2,2). E6 IZI = 1 or 3 One maximal nontoral (rank 5, from a natural F4(~)-subgroup); toral iff cx <~ 2. E 7 Z = 1 One maximal nontoral, rk 6, C(E) = 2 a × SL(2, ~)a, N(E)/C(E) = 223E3 × GL(3, 2). IZl = 2 Two maximal nontoral (type 1 and type 2), both lie in a 4A 7 - 2 subgroup and in a -T 1 o 3E6(~). 2 subgroup; such E/Z have ranks 8 and 7 and both satisfy E' = Z and are self-centralizing; we have N(E)/E ~ 27 ; Sp(6, 2), 2221 + 3.2.1 . 3'~- × a~ × GL(3, 2), respectively (in the former case, 27 indicates a uniserial module for Sp(6, 2) with ascending L6wey factors of dimensions 6 and .)1 E 8 Z= 1 Two maximal elementary abelian, both nontoral, ranks 9 (type 1) and 8 (type 2); an elementary abelian group is toral iff it has complexity at most 2 or is not of rank 5 and 2B-pure; both self- centralizing and N(E)/E -~ 28 : O +(8, 2) if E has type 1 and N(E)/E -~ 26.2 : -(GL(3, 2) wr 2) x E3 if E has type 2. TABLE II Maximal nontorals for the connected case, p odd Type Z Remarks A, Zl ~ 0 (mod p) None IzI --- 0 (mod p) E ~ pt + r2 2p~7 X ,~p where prs = n + 1 o B n None C. None nD None 2G None F, 3a; normalizer 33: SL(3, 3) E6 Z = 1 and p = 3 Two nontoral, one maximal; normalizers 3 × 3a: SL(3, 3), 13 +3+3: SL(3, ;)3 see (11.13) IZl = 3 Two classes (type BC and type ABCD), each of shape +13 +2 × 32, C(E) ~ 3 a and 3 x ~-2 and N(E)/EC(E) = -32 × 32: 3 x GL(2, 3)1 Dih12 x 32 : SL(2, 3), respectively; see (11.14). ELEMENTARY ABELIAN p-SUBGROUPS 259 TABLE II (Cont.) Maximal nontorals for the connected case, p odd Type Z Remarks E7 IZl = 1 or 2; p = 3 Two classes of nontoral E, one contained in the other: rk E = 3 or 4 with the same centralizer 33 x 1-~ N(E) = :33E SL(3, 3)x T1: 2 if rk E = 3 (this maps onto :33 GL(3, 3) and 0(2, ~)) N(E) = 31+3+3: SL(3, 3) oT1:2 if rkE = 4 E s p= 3 Two maximal nontoral, both of rank 5; type :1 centralizer 31+3+3 ° ~2; normalizer 31+ 3 ÷ :3 SL(3, 3) ° T2: 3: 2 (amalgamations over Z3; N(E) has AGL(3, 3) and E3 as quotients) type 2: C(E) = E, N(E)/E = :~3 Sp(4, 3)" 2 (has ,EW as quotient) p=5 One nontoral, order ;35 C(E) = E and N(E) = 5s: SL(3, 5) TABLE III Maximal nontorals for ihe disconnected case (G almost simple, E projectively elementary abelian, F = E (cid:127) G °, E = F x (e), e = p; we use (2.18) to give fixed points of e on G °, then for the possibilities for nontoral F, refer to (2.19) for type A, (5.4.ii) for type D and Section 8 and (2.19) for type E6, (2.25) for p = 3 and D4) Type Remarks A n p = 2; many E For IZl odd, fixed points Sp(n + ,1 ~) (n odd) or SO(n + ,1 ~); for ZI even, (n+ 1)/IZl odd, fixed points PSp(n+ ,1 ~) or PO(n+ ,1 ;)4C for IZl even, (n+ 1)/LZI even, fixed points PSp(n + ,1 K) x 2 or PSO(n + ,1 ~) x 2. ~D p = 2; many E for Z = ,1 fixed points Spin(k, ~) o Spin(2n - k, ~) for an odd integer k, k ~< n/2; for IZl = 2, fixed points 2 × SO(k, ~) x SO(2n - k, ~) for an odd integer k, k .<< n/2; for IZl = 4, fixed points SO(k, K) x SO(2n - k, )4~ for an odd integer k, k .<< n/2. p=3, n=4 (two classes of groups E) Two classes of outer elements of order 3, with fixed points G2(~) (one such F) and PSL(3, ~) (two such F). E 6 p = 2 (two classes of such E; ranks 6 and 7) Two classes of outer involutions, with fixed points F4(K) (one such maximal F, nontoral in G °) and PSp(8, ~) (two such maximal F, both nontoral in PSp(8, ~); one becomes )2~-~ in G °, the other lies in an F4(~) fixed point subgroup) 260 ROBERT L. GRIESS, JR TABLE IV lamixaM extraspecial subgroups of )K(sE Prime Group ~p + az skrameR p~>ll None ees( Section )21 p=7 d = ,1 e = +_ lareveS p=5 d=l,e=_+ andd=2, e=+ lareveS p=3 d = 1 or ,2 e = + lareveS p=2 owT maximal with center of type :B2 )e,d( = ,3( +) and ,7( +) lareves with center of type :A2 d <~ 2 (1.9) NOTATION AND TERMINOLOGY FOR THE REMAINDER OF THE PAPER We let G be a nearly simple finite-dimensional algebraic group over the algebraically closed field ~b such that G O is simply connected; G refers to the main group under consideration in each section of this article. The prime p is always different from char(K). H usually means an algebraic group containing G. Z is a subgroup of Z(G°). E is a projectively elementary abelian p-subgroup of G, i.e. a p-group whose image in G/Z is elementary abelian. For a subset S of G ~< H, C(S) and N(S) refer to centralizer and normalizer in G; otherwise, we use subscripts: Cu(S), etc. A fundamental SL(2, ~) means a conjugate in G of an SL(2, ~) which is generated by a pair of root groups. (1.10) ACKNOWLEDGEMENTS. It is a pleasure to thank Arjeh Cohen for many interesting and informative conversations during our collaboration on finite subgroups of Lie groups; Gary Seitz for the preprint of -CoLiSaSe; Ferdinand Veldkamp, Tonny Springer and the referee for helpful suggestions; participants in my seminars at Ann Arbor and other places; the University of Michigan and the National Science Foundation for financial support; the Centre Nationale du Recherche Scientifique of France for support during my visit at the l~cole Normale Sup6rieure, 1986-87, when this work was initiated. 2. PRELIMINARY RESULTS (2.1) DEFINITIONS. For basic definitions from coding theory, see MacW- S1. We mention the standard term doubly even code for a binary code (i.e. over )2:0 whose every element has weight divisible by 4; such codes are not ELEMENTARY ABELIAN p-SUBGROUPS 162 classified explicitly in general, even the self-orthogonal ones. We also use the term nearly self-orthooonal for a code C such that dim(C±/C) <~ .1 For later calculations, we need to classify (up to equivalence) certain small- dimensional binary codes and get their groups. We leave as an exercise the claim that the list of codes below has the stated properties and exhaust all nearly self-orthogonal codes of the given lengths. (2.2) TABLE TABLE V Small-dimensional nearly self-orthogonal binary codes Length Dimension Basis Group 2 1 21 E z 3 1 21 E z 4 2 ,21 43 2 wr 2 5 2 ,21 43 2 wr 2 6 3 ,21 ,43 65 2 wr E 3 7 3 ,21 ,43 65 2 wr E 3 3 ,4321 ,6521 7531 GL(3, )2 8 4 ,21 ,43 ,65 87 2 wr ~4 4 ,4321 ,6521 ,7531 87654321 AGL(3, )2 ._-_ :32 GL(3, )2 9 4 ,21 ,43 ,65 87 2 wr E 4 4 ,4321 ,6521 ,7531 98 GL(3, )2 x ~2 4 ,4321 ,6521 ,7531 87654321 AGL(3, )2 10(=X) 5 ,21 ,43 ,65 ,87 X9 2 wr E 5 5 ,4321 5678, ,7531 ,87654321 X9 AGL(3, )2 x 2 (2.3) DEFINITIONS. (Famous codes.) The Hammin 9 code (with parameters I-7, 3, 4) is the unique self-orthogonal binary code with parameters -7, 3, 4. Its group is GL(3, 2) and the code may be thought of as the set of complements to linear subspaces of codimension 1 in ~3, together with the empty set, with Boolean sum as the addition. The extended Hammin9 code is the unique self-orthogonal -8, 4, 4J-binary code; it is spanned by a copy of the Hamming code supported on seven of the alphabet letters and (1, 1, 1, 1, 1, 1, 1, .)1 Its group is AGL(3, 2) ~ 2 3 : GL(3, 2). The tetracode is the unique self-orthogonal length 4 ternary code; equiva- lently, it is the unique -4, 2, 3 ternary code. Any nonzero vector has weight 3 and the standard version of this code is spanned by (s, a, a + s, a + 2s), for a, s E :D .3 Its group is GL(2, 3), (2.4) LEMMA. (i) Let T be a finite p-oroup whose Frattini subgroup is cyclic and central. Then T' has order 1 or p and there are subgroups X, Y such that 262 ROBERT L. GRIESS, JR T = X o ,Y where X is extraspecial, Y has an abelian maximal subgroup and ~I(Y) is elementary abelian. (ii) If TIT' is elementary abelian, Y is of the form pr or 2pZ x p .r Proof (i) Notice that T' has order 1 or p. The abelian case is trivial, so we assume that T' has order p. Let U ~> T' satisfy U/T' = f~I(T/T'). Let E be an extraspecial subgroup of U such that U = Z(U)E. Then, E, T = T' = Z(E) implies that T = C(E)E. Since T/U is cyclic and T' has order p, we see that C(E):C(E) ~c C(Z(U)) = 1 or p. Since ~(T) is cyclic, the same is true for subgroups and quotients, whence C(E) n C(Z(U)) is central-by-cyclic, hence abelian. Take X = E and Y= C(E). (ii) follows easily since Y = Z(U) if TIT' is elementary abelian. (2.5) DEFINITION. Let G = G 1 ..... G, be a central product and define G i := (Gj JI ~ i), Z i := Gi c~ G i and Z := (Zili = 1 ..... n> ~< Z(G). There is a natural map G ,--- HGi/Zi, defined by (gi) ~ (Zigi) (easily, one sees this is well defined) with kernel Z. Let S be a subgroup of G. The quasiprojections of S are the groups Si <~ G, i= 1 ..... n, which satisfy: Z i ~< Si and Si/Zi is the projection of SZ/Z into the ith factor with respect to the decomposition IIGi/Z i . (2.6) LEMMA. Let ~F be an m-set and let PF~, PEt) be the vector space of subsets, even subsets, respectively, with addition the Boolean sum and with the natural bilinear form (A, B)~-* IA ~ BI (mod 2) on Pt) and quadratic form on PEt) (A ~ 1/21AI (mod 2)). (i) If m is odd, the bilinear form on PEt) is nonsingular and the quadratic form has maximal Witt index (plus type)/ff m - +_ 1 (rood 8). (ii) If m is even, there is a well-defined quadratic form induced on PE~)/(t)> /ff m - 0 (rood 4); it has maximal Witt index iff m - 0 (mod 8). Proof This is well known. See Gr2, for instance. (2.7) DEFINITION. Let char(K) ¢ 2 and let V be an m-dimensional vector space with nondegenerate bilinear form f and orthonormal basis = {ei }i ~ f~}, where t) = {1,..., m}. We call ~ a frame. A signed frame is a set of the form + ~. The associated frame group is the subgroup of Aut(f) stabilizing +_ ~; it is isomorphic to 2 wr Z,. The frame group is a monomial group which is a semidirect product of the diagonal frame group (diagonal as matrices with respect to ~) and the natural group of permutation matrices. (See (5.1).) The elements i~ of D which satisfy eiej = ei or -ei as j ~ i, j = i, form a basis for D and give D the structure of ;~:~ a subgroup of D is naturally identified with a code, so the code-theoretic notions of weight, etc., apply to D.