JOURNAL OF PURE AND ELSEVIER APPLIED ALGEBRA Journal of Pure and Applied Algebra 99 ( 1995) 53-l 11 Analogue of vertex operator triality for ternary codes * Xiaoping Xu l Department of Mathematics, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong Communicated by C.A. Weibel; received 28 May 1993; revised 4 November 1993 Abstract For each self-orthogonal ternary code C of length k, one can obtain an untwisted positive definite integral even lattice L(C) by gluing k copies of the root lattice of type A2 with C as the glue code. Under the assumption that C contains ( 1,. . . , l), we construct an untwisted space and two twisted spaces associated with a special lattice containing L(C) as an index 3 sublattice. All nine untwisted and twisted subspaces associated with the cosets of L(C) in L are irreducible modules of the vertex operator algebra associated with L(C), and they provide different realizations of sl( 3, @). Using these realizations of sl( 3, @)^,we construct an analogue of “vertex operator triality”: a beautiful and important phenomenon found by Frenkel, Lepowsky and Meurman in their work on construction of the moonshine representation of the Monster group. 1. Introduction The existence of an infinite-dimensional module for Monster group (the largest finite simple sporadic group), satisfying a number of remarkable properties known collectively as “moonshine”, was conjectured by McKay, Thompson, Conway and Norton (see, e.g., [2] ). Frenkel, Lepowsky and Meurman [ 8,9] constructed an example of such a representation, the so-called “moonshine module”, as a certain special type of vertex operator algebra on which the Monster acts, and proved that the Monster is in fact the full automorphism group for this algebra. Some of important features of the moonshine module are essentially hidden in what [9] called “vertex operator triality” for the binary Golay code. There, such triality was * This paper is a revised version of an earlier work done when the author was a graduate student in the Department of Mathematics,Rutgers University, New Brunswick, NJ 08903, USA. I Email: [email protected]. 0022-2049/95/$09.50 @ 1995 Elsevier Science B.V. All rights reserved SSDt0022-2049(93)E0185-7
59
Embed
Analogue of vertex operator triality for ternary codes · 2017. 2. 6. · Analogue of vertex operator triality for ternary codes * Xiaoping Xu l ... abelian group of finite rank with
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
JOURNAL OF PURE AND
ELSEVIER
APPLIED ALGEBRA
Journal of Pure and Applied Algebra 99 ( 1995) 53-l 11
Analogue of vertex operator triality for ternary codes *
Xiaoping Xu l Department of Mathematics, The Hong Kong University of Science and Technology,
Clear Water Bay, Kowloon, Hong Kong
Communicated by C.A. Weibel; received 28 May 1993; revised 4 November 1993
Abstract
For each self-orthogonal ternary code C of length k, one can obtain an untwisted positive definite integral even lattice L(C) by gluing k copies of the root lattice of type A2 with C as the glue code. Under the assumption that C contains ( 1,. . . , l), we construct an untwisted space and two twisted spaces associated with a special lattice containing L(C) as an index 3 sublattice. All nine untwisted and twisted subspaces associated with the cosets of L(C) in L are irreducible modules of the vertex operator algebra associated with L(C), and they provide different realizations of sl( 3, @). Using these realizations of sl( 3, @)^, we construct an analogue of “vertex operator triality”: a beautiful and important phenomenon found by Frenkel, Lepowsky and Meurman in their work on construction of the moonshine representation of the Monster group.
1. Introduction
The existence of an infinite-dimensional module for Monster group (the largest finite
simple sporadic group), satisfying a number of remarkable properties known collectively as “moonshine”, was conjectured by McKay, Thompson, Conway and Norton (see, e.g., [2] ). Frenkel, Lepowsky and Meurman [ 8,9] constructed an example of such a representation, the so-called “moonshine module”, as a certain special type of vertex operator algebra on which the Monster acts, and proved that the Monster is in fact the full automorphism group for this algebra.
Some of important features of the moonshine module are essentially hidden in what [9] called “vertex operator triality” for the binary Golay code. There, such triality was
* This paper is a revised version of an earlier work done when the author was a graduate student in the
Department of Mathematics, Rutgers University, New Brunswick, NJ 08903, USA. I Email: [email protected].
0022-2049/95/$09.50 @ 1995 Elsevier Science B.V. All rights reserved SSDt0022-2049(93)E0185-7
54 X. Xu/Joumal of Pure and Applied Algebra 99 (1995) 53-111
constructed for any binary code whose codewords have even “weights”. In this paper, we
study an analogue of triality for any self-orthogonal ternary code containing ( 1, . . . , 1) .
This analogue is useful for studying intertwining vertex operators; for example, it sug-
gests a certain way to define and construct intertwining operators between certain twisted
modules of a vertex operator algebra of a special type. We shall systematically study
intertwining operators of twisted modules in a separating paper. Moreover, we use it
also ( [ 151, in preparation) to construct “ternary moonshine vertex operator algebras”,
which form a class of vertex operator algebras analogous to the moonshine module. The
ternary moonshine algebra corresponding to the ternary Golay code is expected to be the
moonshine module. Therefore, we would obtain another realization of of the moonshine
module and of the Monster.
In terms of physics, ternary moonshine algebras and their representation theory might
be called “&-twisted conformal field theories” (cf. [4] ). We feel that the study of
these algebras could be very useful in the classification of conformal field theories.
The smallest example of our analogue of vertex operator triality is that for the affine
Lie algebra $, where the corresponding ternary code is Z3 ( 1 , 1,l). Since & arises
naturally in theoretical physics in connection with compactifications of the heterotic
string, for instance, our result in this paper could be directly useful in this area.
Let us recall some basic definitions. Let Z be the ring of integers. For a positive
integer n, set Z,, = Z/(n). A code C of length k over Z,, is a Z,-submodule of Zi.
The elements of C are called codewords. The weight of a codeword is the number of its
self-orthogonal (respectively, self-dual) if C c Cl (respectively, C = C’ ) . When n = 2 or 3, C is called a binary or a ternary code, respectively. A (rational) lattice is a free
abelian group of finite rank with a Q-valued symmetric Z-bilinear form (-, -). A lattice
L is called integral (respectively, even) if (L, L) C Z (respectively, (Ly, a) E 22 for
any Ly E L).
Let k be a positive integer. Throughout this paper, we take the index set notation
O(k) ={l,...,k}. (1.2)
Let
Hz&F&j (1.3) j=1
be a vector space with basis (aj} and the symmetric bilinear form (-, -) such that
(aj,Lyl) = 26j.l for j,l E L?(k). (1.4)
Define the section map 72 : Z2 + Z by 72(O) = 0, 772( 1) = 1. Set
X. Xu/Joumal of Pure and Applied Algebra 99 (1995) 53-111
k k
an = c ‘yi, ff, =
c vz(cj)aj fOrc=(cl,...,Ck) E@.
j=l j=l
For a binary code C, whose codes have even weights, define
55
(1.5)
(1.6)
In [ 8,9], the authors constructed an “untwisted space” VL and “twisted space” VTL
associated the lattice L, equipped with an “untwisted vertex operator map” Y( -, z) :
VL -+ (EndVL)[[z1/2,z- ‘I21 ] and a “twisted vertex operator map” 8(-, z) : VL -+
(EndVT,)[Ez1/2,z -1/2] ] satisfying the following Jacobi identities:
where all the variables are formal and commute with each other; 6(z) = xnEZ z” and whenever (zj - zi)“’ for m E Q appears, it is understood to be the binomial expansion in nonnegative powers of zi; VL = eTEL ( VL)~ is L-graded and u E (VL>~,LJ E (V,>p with
(cr,~u), (LY,~), (/3,/3) E Z; z” = z(“.Y) on (VL)~; 0 : VL + VL is an involution induced by - IdL; @a E End Vr,.
Let Vh, VTb and VL, , VT~, be the untwisted and twisted subspaces associated with La and Li, respectively. Now the vertex operator triality for C is an order-2 linear automorphism u of VL @ VT, such that
v, - VI, VL, - VT %’ Vii, + %, (1.9)
and
uY(u,z)a-’ = Y(uu,z) on Vh,
aE(u, z)u_ l = Y(uu,z> on VL,,
d(u, z)a_ ’ = B(cu,z) on VTL,,
(1.10)
(1.11)
for any u E V,.
56 X. Xu/Joumal of Pure and Applied Algebra 99 (1995) 53-111
Our work in this paper is as follows. We always use @ and R to denote the fields of complex numbers and real numbers, respectively. Define
H=Ck, (h,h’) =&*jTi;,
j=l
(h,h') = @(h,h') (1.12)
Xj=(O ,... ,O,i,O,...,O) EH. (1.13)
Then (-, -) is a hermitian form and (-, -) is a symmetric R-bilinear form. Define the section map
v:Zs+Z by~(Z+(3))=1forZ=O,fl. (1.14)
Let w = e2’ri/3 and R3 = Z[ w] . Then R3 is called the ring of Risenstein integers. We set
k
w= (co,... ,w)/(l -WI E H, Xc = c T(cj>Xj 6 H (1.15) j=l
for c = (cl,. . . , ck) E Zi. For a length-k self-orthogonal ternary code C containing
(l,..., l), we let
L,=RjlW+CRjx,+~R3(l_w)Xj,
I
Z=O,fl; L= u Ll. (1.16) CEC j=l I=-1
Then L is a lattice in H, & is an integral even sublattice of index 3 in L, and L-1, Lo, L1
are cosets of L/b (cf. [ 121). We construct an untwisted space VL and two twisted spaces VT,, Vii associated to L, equipped with an untwisted vertex operator map Y (-, z ) :
where VL is again L-graded and u E (V,),, u E (V,)p with (n,p), (cu, cy), (p, /3) E Z; @a E End VT, or End VF~; D : VL -+ VL is a map induced by v = w Idl;.
X. Xu/Journal of Pure and Applied Algebra 99 (1995) 53-111 57
Let VL,, V-TL, and VFL, be the untwisted and twisted subspaces associated with LI. The following is what we call an “analogue of vertex operator triality” which is our main theorem of this paper:
Main Theorem. There exists an order-4 linear automophism u of V, @ V, @ VFL such that
u : v, - v,, uY(u, z)a_l = Y(du),z), u E vi.,.), (1.19)
aY”(u, z)a -’ = Y(au, z) on vL_I,
aY(u, z>a-l = Y*(uu, z) on v- T&l ’
(Ty*cu, z)a -’ = Y(uu, z.) on VL,,
aY(u, z>u-’ = Y”(uu, z) on VT,+,
aY”(u,z)a- ’ =Y"(uu,Z) on bL_,,
aY"(u,z)a-' = Y*(au,z) on FL --I ’
uY*(u,z)cfl = Y*(uu,z) on VT~ , 1
uY*(u, z)u_’ =YN(uu,z) on bL,,
( 1.20)
(1.21)
(1.22)
(1.23)
( 1.24)
foranyuEVh.
Our work is organized as follows: In Section 2, we study the lattice L given in ( 1.16)) certain central extensions of L
and some irreducible modules of the central extension groups. In Section 3, we construct an untwisted space and two twisted spaces with corresponding vertex operators associ- ated to L. We study nine vertex operator realizations of sl( 3, Cc)- and their isomorphic relationships in Section 4. In Section 5, we construct the u in ( 1.19). The CT of the first diagram in ( 1.20) with properties ( 1.21) and ( 1.22) is constructed in Section 6. Finally, in Section 7, we construct the u of second diagram in ( 1.20) with properties (1.23) and (1.24).
58 X. X&Journal of Pure and Applied Algebra 99 (1995) 53-l 11
2. Lattices and modules
In this section, we first present some properties of the lattices L and Le. Then we
define the central extensions e and i of L. Moreover, we construct a e-module C(L)
and two i-modules T and F.
Let H, (-,-), {LI}, L and related notations be the same as in (1.12)-( 1.16). We also
use the notation n(k) defined in ( 1.2). In order to study the lattices L and Z& we set
Yj=(l-tO)Xj, Wj=-Xj l-w
for j E L?(k); CA=&Rsyj. j=l
Then W = Cjzl Wj.
We now list some important formulae which were proved in [ 121.
Lemma 2.1.
(1) (W’Xj, Xp) =
(2) (JYj, Xp) =
2s. 3 hP if2 - 0,
-isj,p tw’Yj 1 Yp
i.fl$O,
-sjsp ifZ E -1, (W'yj, Wp) =
0 ifZe 1,
{
1
3 ifl=l,
(3) (Wj, W’Xp) =
-i ifZ=O, (Wj9 Wp) = gsj.p;
2sj.p iflEO,
-sj,p if If 0;
-2s. 3 J,p if 1 = O9
I*. 3 J>P ifl$Oo;
(2.1)
(2.2)
(2.3)
(2.4)
where “3” means “(mod 3)“. 0
Since C is assumed to be a ternary self-orthogonal code containing ( 1, . . . , 1) ,
v(cj) E 0 (mod3). (2.5) j=l
In particular, k E 32. By the above lemma, (L, L) E Z/3 and (( 1 - o) L, L) E Z.
Therefore, we can define 4 : L x L -+ Z by
S(cqp) = ((w - w-‘)(Y,@ for cu,p E L. (2.6)
Then 6 is a Z-bilinear skew symmetric form. One can easily see that for an integral
sublattice M,
IV&P) = & wj,,p) (mod3) for cr,p E M. (2.7) j=l
X. Xu/Journal of Pure and Applied Algebra 99 (1995) 53-111
Furthermore, by [ 61,
59
8((1 -w)M,M) -0 (mod3).
The following is Lemma 2.2 in [ 131.
(2.8)
Lemma 2.2.
~(xe,xct) z 0 (mod3) for c,c’ EC.
6 E 0 (mod3) on QA x QA. Cl
We can assume that C has a generator matrix of the form
lk
(
1 1s lk-s-l Cl
or, c =;’
4 es
(2.9)
(2.10)
(2.11)
where s 5 (k/2) - 1. In fact, s = (k/2) - 1 if and only if C is self-dual. Set
We extend Y (-, z ) , W” (-, z ) and W* (-, z ) on VL linearly. Notice that
(ql)dq-1)) = &v,
by (2.2). In particular,
(Yj(l)vYI(-1)) =6j,I7 i,l E fi(k),
(3.28)
(3.29)
(3.30)
where yj was defined in (2.1). One can easily see that {yjct),yj(_t) 1 j E 0(k)} is a basis of H. Set
AZ = C k{ (W-l - l)Cn,mYj(l)tn)Yj(-~)(m)
O<n,mE.Z j=l
+ (0 - l)Cn,mYj(l)tm)Yj(-l)(n)}Z-'"-"~ (3.31)
where
c ~~,~.“c”y~ = In (l+x)‘/s-~-‘(l+y)1/s
O<fl,t?lE% l-w-’
(3.32)
One can check that our AZ, is exactly the same as that in [ 6,101. Switching the indices (1) and (-1) of yj in (3.31), we get another operator A*z. Now we define
Using the fact that each element of [ ( 1 - ti)*L]” acts as the identity on VT and VT,
and using the proof of Theorem 4.3 in [ 141, we have:
Theorem 3.1. For u = u’ 8 L(&), u = u’ @J L(B) E v, with a, p E L sufisjjhg
(a, a), (a, p), (p. p) E Z, the Jacobi identifies hold:
ZI -z2 ZO-‘s( - )YbZI)Y(U,Z2) - z$p3Y(u,z22)y(u,zI)
= z2q3$J) Y(Y(u, zo)u, 22) (57,
Zl -z* %?6(---_
)y”(u,z,)Y”(u,z2) -~-‘s(~)Y”(u,z*)Y”(u,zl)
= ~~6(o’(~)1’3)Y-(Y~~p~,~)u,z2)~
P=o
(3.34)
(3.35)
and
X. Xu/Joumaf of Pure and Applied Algebra 99 (1995) 53-111 65
Zl -z2
%3( T )Y’(U,Zl)r(U,Z2) - ~-‘s(~)Y*(,,~~)Y*(~,z~)
(3.36)
Remark 3.2. In this paper, we use some different notions from [ 141. Here are the relations: in this paper we use
F=&, EC&” and F*=(p)-’ (3.37)
to construct an untwisted vertex operator map Y (-, z ) and twisted vertex operator maps Y” (-. z ) and Y* (-, z ), respectively. Furthermore, the F here is same as (4.89) in [ 141. Thus the (T for Y” in (4.38) of [ 141 is trivial. The CT for Y* is:
(P(CY,p)~(P,(Y))-l(-l)(a,B). (3.38)
That is why we defined f* differently from f" in (3.20).
Next let us take a look at eAz. First of all
( (l+x)‘/s-CU-‘(l+y)1’s
1-w-’ > =1+ x-w-‘y X2 - ,-iy2
3(1 -C0-’ ) - 9( 1 - o-1) +..*. (3.39)
Hence
ln (1 +x)1/3 - w-‘( 1 + y)‘/3
1-W’ > x - w-‘y
=3(1-w-‘) + (-2+3w)x2+ (-2+3w-‘)y2 1
54 - zxy +. . . . (3.40)
Thus
k
C[Yjcl)(l)Yjc-1,(0) +2Yj(l,(0)Yj(-*)(1)lZ-’ j=l
k k
+ i CYjcl,(0)Yjc-1,(2)Z-2 + f CYj(l,(2)Yj(-1,(“)Z-2
j=l J4 k
+ i CYjcl,(l)YjC-l,(1)Z-20(Z-2). J=I
(3.41)
Therefore,
66 X. Xu/Journal of Pure and Applied Algebra 99 (1995) 53-111
k
e A’ = 1 - i C[YjcI)(l)Yjc-I)(O) + 2Yj(l)(0)Yj(-1)(1)lZ-’ j=l k
68 X. Xu/Jownal of Pure and Applied Algebra 99 (1995) S3-Iii
4. Representations of the affie Lie algebra A;’
In this section, we study nine vertex operator realizations of sl(3, Cc)” and their iso- morphic relations.
Let M c L be a subset and set
(cf. (2.23) ). Denote
(4.1)
Vc = the weight-l subspace of Vl. (4.2)
We define the operation “[ -, -1” and the C-bilinear form “(-, -)a” by
[u, ul = uou, (u, u)oa9 = UIU, for u, u E VO. (4.3)
Then Vi is a Lie algebra, and (-, -)a is a symmetric &invariant form. Moreover,
[hmhll =hdJn--dh = [~,Ulm+n+~~m+n,O(UIU)O (4.4)
for U, u E VO and n, m E Z (cf. Section 8.9 of [9] or Section 3 of [ 141). This implies
that
$= c a&@@ (4.5) UCv& nEZ
is an affine Lie algebra associated with the Lie algebra Vo with respect to (-, -)o. Recall that D acts on Vb as v 8 P. One can verify that
DY(u,r)D-’ = Y(PU,Z) (4.6)
for u E Vh . Thus by (4.3)) D is a Lie automorphisms of Vo and an isometry with respect to (-,-)a. So it is also an automorphism of Vo.
For u E VO, set
1 U(f) = - 3 c
w-ljpi, (4.7) jG3
Then
DU(f) = JU(f). (U(f)9U(j))Oco (lfj), for U,U E VI, I, j E iZ3. (4.8)
Set
Gj = Span Of {L(~),L(-~~),(Y*yj)(-1) 11 E Z3}, G=eGj9 (4.9) j=l
X. Xu/Joumal of Pure and Applied Algebra 99 (1995) 53-l I1 69
where yi = ( 1 - o)xi. Then 0 is a Lie subalgebra of VO. We also denote &? as the associated affine subalgebra. Let
G(I) = {U(1) I u E G). (4.10)
For 1, j, p E Zs, we define the linear map Vj,l : B -+ 6 by
Vj,[(L(Z&) = U*'L(ZiiYi), Vj,J(V'Yi) = Yj+'Yi, for i E .0(k). (4.11)
We define twisted afline Lie algebras
o”jJ = e C [ C WYf’))(p) @ t” i=l pE5 nE(pjfl)/3+Z
+ c cYi(p) 63 tm 1 + CC + CD
mE(pjP)+Z
with the operations
(4.12)
[u 8 tm, u 63 f"l = [u, ul @ tm+n + m(u, U)ly3m+n,oC,
[D,u@fy =nuG3rn (4.13)
for u @ tm u ~3 t” E CP and 7
[pJ,c] =o. (4.14)
Now by Theorem 3.4. of [ 141 and (2.3), we have the following proposition:
Proposition 4.1. Z70.~t : u 8 tm H u,, D H -L(O), c I--+ Id”+ (c$ (3.50), (3.58) and (3.61) ) define representations of the twisted afine Lie algebras gVa*l on I$,. El
The following proposition is by (2.17), (2.27), (3.16) and Theorem 4.5 of [ 141:
Proposition 4.2. ZIr,l : u ~3 tm H u;, D H -L” (0) , c I+ IdbL_, (cf. (3.50), (3.59)
and (3.61)) define representations of the twisted a&e Lie algebras &P on VT~_, . l7_,,[ : u @ tm H u;, D H -L*(O), c H Id” fL_, (c$ (3.50), (3.60) and (3.61))
give representations of the twisted ajke Lie algebras G”-lJ on V&_,. q
Next we study the relationships among @‘jJ for 0 # (j, I) E 23 x Zs. Denote
ap(L(bc,c*)) = j~(cp)&3--‘/* c 0-9(cP’cP+‘)‘(q;~,)
9E5
(5.27)
and dually
(5.28)
Now by induction on p E n(k), we have the following:
Theorem 5.5. For any c E C and c* E A(c),
c(L(bc,ce)) = js(~)~~(c*)3-~*/* C W-~(~‘)-c’~~*~~L(~c,c,),
C’Ed(C)
where
‘Cc) = kv(Cj)v C* = ev(Cj)*, j=l j=l
k
(5.29)
(5.30)
(5.31)
c’ . c* . c = c ~(C~)~(Ci*)T(Cj). 0 j=l
X. Xu/Journal of Pure and Applied Algebra 99 (1995) 53-111 77
Since {&. 1 c E C} and B generate Vh for any fixed map c H c* E d(c), the concrete action of c on V, is clear. The following result is useful in the next two sections.
Corollary 5.6. Let
b: = kc.
Then we have
a*‘( ‘(b:)) = j*s(c)3-c2/2 c &“~(b&
c'EA(c)
and
cr*‘(b(bz)) = i*S(c)3-c2/2 c b(be,e,). 0 c’EA(cf
(5.32)
(5.33)
(5.34)
6. Linear automorphism u of VI., $ VrL, @ VL_~ @3 VpL,
In this section, we shall use Propositions 4.1-4.4 to construct a linear automorphism 0 of V,, @ VT~, @ VL, @ VTk satisfying the Iirst diagram in ( 1.20) and the equations ( 1.21) and ( 1.22). Our work IS divided into the following four steps.
6.1. Step 1. Construction of the linear automorphism u of VL, @ VL_, $ VT% ~3 VT%
Set
x; = *tow! b; =x;.
Then by (3.14) and (5.32),
(6.1)
k
c u2 = [~(Cj)W*"' f -1xj 1-W j=l
k = c
~(Cj)td*"j( 1 - 0) f W-IX,
1-W J j=l
k
(6.2)
for c E C. Thus
(b&,b,f) = (w: w) = $ (6.3)
78
Let
X. Xu/Journal of Pure and Applied Algebra 99 (1995) 53-111
Lc, = b:b;, + Q‘s,.
Then
L*1 =(JLk is a disjoint union.
Set
Then
VL,, = 63 VC &I. CEC
For any c, c*, c’ E I!$, we have
’ j=l k
= Cr](Cj)v(C;)(OcT - l,W-’ -W)
j=l k
= c -~(Cj)~(C$)~(Ci*) = -C ’ C' ’ C* j=l
by ( 5.16). Therefore,
Y
Xc,c’ = K -S(X..C:=lrl(C,)(Oc~-l)xj) v
Xc ’ [e,(Cj)(U'i - l)Xj] j=l
=K
(6.4)
(6.5)
(6.6)
(6.7)
(6.8)
(6.9)
by (2.22) and (5.8). Hence
S(ke,e’) = w -“““g( xc>, Q(&,cJ) = w “‘c”Q( kc). (6.10)
Set
Then
(6.11)
(6.12)
X. Xu/Journal of Pure and Applied Algebra 99 (199.5) 53-111 79
By our definitions of the vertex operators in Section 3 and Theorems 2.3.1-2, one can
prove that V&, are irreducible @‘~~L-modules, respectively, and V+b, V;b are irreducible
Q’*l,o-modules, respectively. Now we need to use the notations in (4.36). First we have:
1 (x,:a,)o”G,) = - c
9a lE&
g[ (&;‘)X,] (by (3.22))
w (f”‘(o-o-‘)y,.~;=, v(c.)x.) I,_(: ) E
(by (2 33))
1 = -W(f(w-o-‘)Yj,cjxj)~(XE) (by (2.8))
3& wZCj
= -g(X,) (by (2.3)). 3ti
Similarly by (2.23) and (5.8), we get
(X,:,))0*G(_Q = -Q(k,). 3&
(6.13)
(6.14)
Furthermore, by (2.3), (3.17) and (6.2),
eUj(*l)) (O)L(b,b*_) = ,+l,F(Cj+l)
fi 3& L(bib*-) =
aFcj -L(b,bT). 3&
(6.15)
and
= __wF(-Cj+l)L(be+b;> = _ &j
34 -L(b,+bf_). 3J3
(6.16)
Now by (4.42), Propositions 4.1-4.4, (6.13-6.16) and the highest weight represen-
tation theory, each pair of the following irreducible representations are equivalent for
each c E C:
the representations I7t.a of Gyl.O on V& and ZIc,t o u of @‘I,0 on Vz_, , the representa-
tions ZIa,t of GVoJ on Vi_, and II-t,0 o u of POJ on V;, , the representations ZZ_t,c of
G”-‘,” on VYb and I7s,_t o u of G’“-‘” on V&, the representations ZIe,_t of P-I on VL,
and I71 ,o o ff of P-l on V$b.
Lemma 6.1. For any (0 # Aj,c 1 j E L!(4), c E C} c C, there exists a unique linear aufomolphism L+ of VL., $ V,_, @ VT-% @ VT~ satisfying
(6.17)
80 X. Xu/Journal of Pure and Applied Algebra 99 (1995) 53-1 I I
and
(T(s(Xc)) = h,&(b,-K),
cr(G(X,)) = A3,&,+Q > 1
and
cry-(u,z)F’ = Y(au,z)
aY(u, z)a -I = Y*(cTu, z)
gY*(u,z)cT-’ = Y(au,z)
(TY(u,z)& = Y”(au,z)
for any u E 6.
(6.18)
(6.19)
(6.20)
(6.21)
Proof. By the above equivalences of irreducible representations, there exists a unique linear automorphism u of VT_, $ V’_, $ VT% $ Vrb such that
CT satisfies (6.18)-(6.19) and
aI&@+ = ZIO,] on VL-,. aI70,la-’ = II_],0 on V- % ’
(6.22)
an- 1 ,og -’ = 170,-l on VL,, ~&,.-~c+-’ = nl.0 on vTb. (6.23)
Now (6.20) -( 6.2 1) are equivalent to (6.22) -( 6.23) by Propositions 4.1 and 4.2. Cl
6.2. Sfep 2. Required conditions for {A,,, 1 j E L!(4), c E C}
First we need some important formulae. Let c, c* E C and c’ E A(c). We have
j=l k
j=l
k
=Fi C9(Cj)9(C; f Ci*>
J=l
= qx!- (c’fc’)
3 (6.24)
X. X&Journal of Pure and Applied Algebra 99 (1995) 53-l 11 81
by (2.4), where for any c’,G E C, we take the notion c’ -c* = ‘& q(cj)v(c,?). NOW for c,c* EC and c’ E A(c).cO E A(c*),
j=l I=1
= B&ll(CiH ,‘p - 1)x1) (by (5.8)) I=1
G c . c* . co (by (6.8))
and
~(Xc,c’rX c*,cO)
k j=’ I=1
= C~(Cj)rl(cj*)((W- Ol-‘)OJ”;*Oc~Xj)
j=l k
j=l
Thus
~oO(&,c’rxc*,co) =4c*c* .co+c.c*. (8-c’).
For c, c* E C, we assume ct E A(c) satisfying
F c. (c+ *c*) = -c*.
Then ct is unique for each case. In fact, if cj # 0, then
i
Cj = fCj cJ= 0
if Cj* = 0, ifcJr =Cj,
Fj if cy = -Cj.
(6.25)
(6.26)
(6.27)
(6.28)
(6.29)
We find that
82 X. Xu/Journal of Pure and Applied Algebra 99 (1995) 53-l 11
-%,ct + xc* ,fC’
k
c t = [rl(Cj)W"j + r](C;)w*‘T]xj
j=l
+ C [r](Cj) +7)(Ci*)]W"jXj
= c TCcj + c~)u*(C'+C')Xj + C q(cj + cJ~)~~(C,+C;)~~ CjCy=O
90 X. Xu/Joumal of Pure and Applied Algebra 99 (1995) 53-111
VL, = S(M+_)a(%,,), 6% = ~S(M344b,+b;)). CEC
Furthermore, by (6.73-80))
[M-,Z’(b,z)l =0 on VL_,, [M*,Z*‘(b,z)] =0 on V- . rL0 ’
[M+,Z’(b,z)l =0 on K,, [M”,Z”‘(b,z)] =0 on VT~.
As ( 11.3.22) in [9], we have:
(6.89)
(6.90)
(6.91)
Z’(b,z) = z;(O)z’ = z;(O) on VL*, ,
Z*‘(b,z) = z*L(O)z’ = z*;(O) on VT%; (6.92)
Z”“(b) z) = z”L(O)z’ = z”:(O) on VT,+, (6.93)
where z~(O),z*~(O) and z “i( 0) are operators of degree 0 on VL*, , VT% and VT,+, respectively,
Now (6.20) and (6.21) for u = c(b) are equivalent to
uXYL(b),z)cr-’ = X’(b,z)
uX(db),z)a -’ = X*‘(b, z)
aX*(L(b),z)a-’ = X’(b,z)
aX(l(b),z)+ = X’--‘(b,z)
on VL-,, (6.94)
on V- . T%’
on VL,, (6.95)
on vTh;
or by (3.21)-(3.23), (6.81)-(6.83), and (6.92) and (6.93), equivalently,
‘TX& (O)a-t = z;(O) = .h(b) (0) on VL-, , (6.96)
a&(b)(O)& = z*:(O) = xTrL(b) (0) on v&v (6.97)
cxf@) (O)a-’ = z;(O) = h(b)(o) on VL,, (6.98)
(7&(b) (O)a-’ = z”:(O) =x&)(O) on %%. (6.99)
Now the lemma follows from (6.44), (6.52)) (6.57), (6.64)) (6.88) and (6.89) and
(6.96)-(6.99). 0
6.4. Step 4. Proof of (6.20) and (6.21) for any u E Vh and the statement of the Main theorem
Let
B = {x,‘:,,, a,(,) I P E a(k),4 E Z3) (6.100)
X. Xu/Joumal of Pure and Applied Algebra 99 (1995) 53-l 11
(cf. (4.36) ) . For any u E L3, we can write
u= c u;@&(b) withubES(&_), b=(l,p)~&. b
91
(6.101)
By (2.3) and (4.42), we have:
Lemma 6.4. y u c Gtl) n 23 and (TU = xb 24; 63 b(b), tkn (6, w) E ~(t)/3(~20dZ). Conversefy, if u = xbub @ r(b) E a(G) and (6,W) s 2/3(modZ), then uu E
G(-I). 0
Now we assume that (6.20) and (6.21) are true for some u E Vr, and let u E Q.
Then,
zo-‘s( y, (TYN(U, z1 )Y"(u, 22)d
- q’pI..$) aY"(u,z2)Yyu,z,)d
= q’s(y) Y(au, Zl)Y(W 22)
- z$s( 2_z’)Y(,ru, z*)Y(au, a)
Zl - z2 %-‘qQ-) aY(u, z, )Y(u, z2)Kl
- z(-‘a( Z2_Z’)UY(V, z2)Y(u, z&7-1
= zO-‘q!L$) Y*(CTu,z1)Y*(Uu,z2)
- qQ( z$L) Y*((+u,z2)Y*(C7u,zI)
on V- T% ;
Zl - z2 0( -) (+Y*(u,z,)Y*(u,z2)(T-’
_ zO-+J (+Y*(u,z2)Y*(u,z1)(T-1
= ~-‘~(~) Y(W Zl )Y(cru, 22)
- ~-‘s(~)Y(~u,z2)Y((TII,zL)
(6.102)
(6.103)
(6.104)
on VL,;
92 X. Xu/Joumal of Pure and Applied Algebra 99 (1995) 53-111
21 -z2 %?S( -) aY(u, Zl )Y(u, zzw’
_ gqz+, aY(v, zL!)Y(u, z1 Ia-’
= g_Ls( K$) Yw((TU,Z1)Ym((TU,z2)
- ~l~(~)Y-(,,,z,)Y-(,,,z*) (6.105)
Furthermore, we assume that u E $1, f-la(G). By (2.8), (3.16) and Theorem 3.1,
aY(Y(u,zo)o, z2)(+ -’ = Y*(aY(u,z0)u,z2) on VT% (6.118)
aY(Y(u, zolu, z2)(T -’ = Y”(c+Y(u,za)u,z2) on VTL,. (6.119)
This tells us that (6.20) and (6.21) are true for the coefficient of ,$ in Y( u, zg)u, for n E Z. Viewing VI.,, as a module of c c PO (cf. (4.5) ), we can see that
{U E Vb 1 (6.20) and (6.21) hold} (6.120)
is a &ubmodule. On the other hand
&J = u b,& = u be+&‘& (6.121)
CEC CEC
As a &module,
V, is generated by {I 1 c E C} or {I 1 c E C}. (6.122)
Thus by Lemma 6.3, (6.20) and (6.21) are true for any u E Vh. By (3.47)-(3.49), (6.3), (6.18) and (6.19), u is grading preserving. Moreover, if nft Aj.0 = 1, then ai = 1 by (6.68) and (6.69). Here we summarize them as our main theorem in this
section:
Theorem 6.5. For any {Aj,o 1 j E a(4)) c C} with n;=, Aj.0 = 1, there exists a unique order-4 grading preserving linear automorphism (+ of VL, 03 VL_, @ VT~ @ VTb satisfying that
Therefore, {xTE, 1 p = f 1, E E ZJ} is a set of semisimple elements of sl( 3, Cc). For
each p = fl, the span of {x(~~,x&)} is a Cartan subalgebra of sl(3,C). Hence
the span of {xj(p),xgLp),C,D 1 j E a(k)} (7.20)
is a Cartan subalgebra of GY*l+p (cf. (4.12) ) . By the highest weight representation the- ory and (7.10-17), each pair of the following irreducible representations are equivalent for each c E C:
the representations IIt,_, of LW-’ on V&, and ZZt,t o u of @‘1,-1 on l$_, , the representations Z7t.t of P.1 on V+L_, and Z7_1.1 o u of PI on ViL_, , the representations
ZI_r,t of G”-‘vl on V! TL-,
and l7_t,_t o u of G’-‘J on $YL , the representations II-1 ,_ 1 I
X. Xu/Journal of Pure and Applied Algebra 99 (1995) 53-l 11 103
Condition 4:
(7.56)
_ i2c2/3~c~c~c* ( n W_")pC-&4,c+c. = i2s(c)~-c’c*‘c* ( n wCj)
Cj=C' I
Cj=C;
(by (7.33) and (7.37)) (7.57)
* ru4,c = 1 ~2s~c’+2c2’3~4,0 for any c E c (7.58)
by Condition (2) and the fact that 2c2/3 E -2c2/3(mod4).
We summarize the above as the following lemma:
Lemma 7.2. The following conditions are necessary to make (7.24) and (7.25) true for any u E Vb:
Pl,C = PI.09 p,2,c = i2s(c)+2c2/3p2,0, (7.59)
ru3,c = p3.0, p4,c = i2s(c)+2c2J3~4,0, (7.60)
for any c E C. Moreover, (7.38), (7.45), (7.53) and/ (7.56) are true under the above conditions. 0
7.3. Step 3. Proof of (7.24) and (7.25) for u = b(b,+) or u = l(b;)
Lemma 7.3. The first equation in (7.24) and the second equation in (7.25) are true for u = L( bf). The second equation in (7.24) and the first equation in (7.25) are true for u = r(b,).
aY*(Y(u, zo)u,z2)(+-1 = Y”(aY(u, ZO)U, 22) on bL,. (7.100)
This tells us that (7.24) and (7.25) are true for the coefficient of z,$’ in Y( U, zc)u, for n E Z. Viewing Vh as a module of 0 C f& (cf. (4.5)), we can see that
{n E Vb 1 (7.24) and (7.25) hold) (7.101)
is a &submodule. Thus by (6.122) and Lemma 7.3, (7.24) and (7.25) are true for any u E Vb. By (3.48), (3.49), (7.22) and (7.23), u is grading preserving. Moreover, if I$=, c~j,o = 1, then CT; = 1 by (7.59) and (7.60). Here we summarize them as our main theorem in this section:
Theorem 7.4. FOB U~Y {pj,o 1 j E O(4)) C Cc} with nfI pj,o = 1, there exists u unique order-4 grading preserving linear automorphism CT of VT~, $ VT~_, $ VTL, @ VTL_,
satisfying
ds(hW+)) = pl,os(~,W-1,
a(4(icW-)) = i2S”‘+2C2/3rU2,0~(~:eW-),
(7.103)
X. Xu/Joumal of Pure and Applied Algebra 99 (1995) 53-111
a(T(SW-) = ~3,O?(XcW+),
a( T( x, W+) ) = i*s(C)+*c2/3/Lq,0q( 2, w_ )
(cf. (5.31)) and
cTYN(u,z)g -’ = Y"(au,z) on VT_,*
aY"(u,z)(T -l = Y*(au,z) on V- TL_, ’
(+Y*(u,z)fT-’ = Y*((+u,z) on &L , 1
uY*(u, z)a -I = Y”(CTu,z) 011 bL,
foranyu E v,. 0
111
(7.104)
(7.105)
(7.106)
This completes the proof of the main theorem presented in the Introduction.
Acknowledgments
The author of this paper thanks Professors J. Lepowsky and R. Wilson for their advice
and encouragement.
References
[ 1 I R. Borcherds, Kac-Moody algebras, aad the monster, Proc. Nat. Acad. Sci. USA 83 (1986) 3068-3071. [2] J.H. Conway and S.F! Norton, Monstrous moonshine, Bull. London Math. Sot. 11 (1979) 308-339. [ 31 J.H. Conway and N.J.A. Sloane, Twenty-three constructions of the Leech lattice, Proc. Roy. Sot. London
A 381 (1982) 275-283. [4] L. Dolan, P. Goddard and P. Montague, Conformal field theory of twisted vertex operators, Nuclear
Phys. 338 (1990) 529-601. [ 51 L. Dolan, P. Goddard and P. Montague, Conformal field theory, triality and the Monster group, preprint,
1989. [ 61 C. Dong and J. Lepowsky, A Jacobi identity for relative twisted vertex operators, preprint, 1989. [ 71 I. Frenkel, Y. Huang and J. Lepowsky, On axiomatic approaches to vertex operator algebras and modules,
preprint, 1989. [ 8 I I.B. Frenkel, J. Lepowsky and A. Meurman, A natural representation of the Fischer-Griess Monster with
the modular function J as character, Proc. Nat. Acad. Sci. USA 81 (1984) 32.56-3260. [ 91 LB. Frenkel, J. Lepowsky and A. Meurman, Vertex Operator Algebras and the Monster, Pure and Applied
Mathematics (Academic Press, New York, 1988). [lo] J. Lepowsky, Calculus of twisted vertex operators, Proc. Nat. Acad. Sci. USA 82 (1985) 8295-8299. [ 11 I N.J.A. Sloane, Self-dual codes and lattices, Proc. Symp. Pure Math. 34 ( 1979) 273-307. [ 121 X. Xu, Untwisted and twisted gluing techniques for self-dual lattices, preprint, 1992. [ 131 X. Xu, Central extensions of self-dual type-A complex lattices, preprint, 1992. [ 141 X. XII, Lattice coset constructions of vertex operators, preprint, 1992. [ 151 X. XII, Ternary moonshine vertex operator algebras, in preparation.