arXiv:math/0104160v1 [math.QA] 16 Apr 2001 filearXiv:math/0104160v1 [math.QA] 16 Apr 2001...
Transcript of arXiv:math/0104160v1 [math.QA] 16 Apr 2001 filearXiv:math/0104160v1 [math.QA] 16 Apr 2001...
arX
iv:m
ath/
0104
160v
1 [
mat
h.Q
A]
16
Apr
200
1
Decomposition of the Moonshine Module
with respect to a code over Z2k
Hiroki SHIMAKURA
Graduate School of Mathematical Sciences,
University of Tokyo, Komaba 3-8-1, Tokyo 153-8914, Japan
e-mail: [email protected]
Abstract
In this paper, we give a decomposition of the moonshine module V with re-spect to an extremal Type II code over Z2k for an integer k ≥ 2. Then we obtainautomorphisms of V , some 4A and 2B elements of the Monster with respect to thedecomposition. We give examples of such a decomposition for some k and give theMcKay-Thompson series for a 4A element.
1 Introduction
The aim of this paper is to give a decomposition of the moonshine module V with respect
to an extremal Type II code over Z2k = Z/2kZ.
The notation of a vertex operator algebra (VOA) is introduced in [Bo, FLM]. One
of the most interesting examples of a vertex operator algebra is the moonshine module
V constructed in [FLM]. The automorphism group of V is the Monster, the largest
sporadic finite simple group.
V contains a subVOA isomorphic to the tensor product of 48 copies of L(1/2, 0),
called a Virasoro frame in [DGH], and the VOA structure of V was reconstructed from
a Virasoro frame in [Mi2]. By using a Virasoro frame, the McKay-Thompson series for
some elements of the Monster are explicitly computed.
Since L(1/2, 0) is isomorphic to W2 algebra at c = 1/2, we should study the first
member of the unitary series of Wn algebras. In particular, W4 algebra at c = 1 is
realized as the fixed-point subspace V +L of the lattice VOA VL corresponding to the rank
one lattice L = Zα with 〈α, α〉 = 6 with respect to the −1 automorphism of the lattice.
Its fusion rules have a nice symmetry according to [Ab] and for each embedding of V +L
into V , we get a 4A element of the Monster [Ma].
1
In this paper, we consider more general case L = Zα with 〈α, α〉 = 2k for an integer
k ≥ 2. Since the Leech lattice Λ contains many elements of squared length 2k and V
contains V +Λ as a subVOA, V contains many copies of V +
L . We consider a set of 24
mutually orthogonal pairs of opposite vectors of Λ with squared length 2k. We call such a
set a 2k-frame of Λ. We have a natural bijection between equivalence classes of 2k-frames
and extremal Type II codes of length 24 over Z2k. By using a 2k-frame, we show that V
contains a subVOA isomorphic to the tensor product of 24 copies of V +L . Then we give
the decomposition of V as a (V +L )⊗24-module, which is described in terms of the extremal
Type II code of length 24 over Z2k corresponding to the 2k-frame. By the fusion rules of
V +L , we define automorphisms of V with respect to this decomposition. When k is odd,
those are 4A elements of the Monster, and when k is even, those are 2B elements of the
Monster. We give examples of a 2k-frame for some k and the corresponding decomposition
of V . Then we compute the McKay-Thompson series for a 4A element of the Monster.
Professor Ching Hung Lam studies the subject independently.
Throughout the paper, we will work over the field C of complex numbers unless oth-
erwise stated. We denote the set of integers by Z and the ring of the integers modulo k
by Zk.
Acknowledgements. The author wishes to thank Professor Atsushi Matsuo, my re-
search supervisor, for his advice and warm encouragement. He also thanks Professor
Masaaki Kitazume and Professor Ching Hung Lam for helpful advice.
2 Preliminaries
In this section, we recall or give some definitions and facts necessary in this paper.
2.1 Vertex operator algebra V +L and its irreducible modules
In this section, we give the vertex operator algebra V +L and its irreducible modules. Note
that the description of V +L in this paper is slightly different from those given in [DN, Ab].
Let L = Zα be an even lattice of rank one with 〈α, α〉 = 2k. Set H = C ⊗Z L
and we regard L as a subgroup of H . We extend the form 〈·, ·〉 to a C-bilinear form
on H . Let L = h ∈ H|〈h, α〉 ∈ Z = L/2k be the dual lattice of L. Let L be the
trivial extension of L by the order 2 cyclic group 〈−1〉. Form the induced L-module
CL = C[L] ⊗C[±1] C, where C[·] denotes the group algebra and −1 acts on C as
multiplication by −1. Note that CL ∼= C[L] as an algebra and we denote an element
of CL corresponding to µ ∈ L by eµ. Set VL = S(H ⊗ t−1C[t−1]) ⊗ CL. For
2
λ + L ∈ L/L, we set Vλ+L = S(H ⊗ t−1C[t−1])⊗ Cλ + L, where λ + L is regarded as
a subset of L and Cλ + L is the subspace of CL spanned by eµ|µ ∈ λ + L. It iswell known that VL has a VOA structure and Vλ+L has a VL-module structure for each
λ+ L ∈ L/L (cf. [FLM]). Note that all the irreducible VL-modules are given by the set
Vλ+L|λ + L ∈ L/L (cf. [D1]). For convenience, we denote h ⊗ tn by h(n) for h ∈ H
and n ∈ Z.
Let ρ: L → C× be a group homomorphism such that ρ(2µ) = (−1)〈µ,µ〉/2 for µ ∈ L.
Let θL be the automorphism of L given by θL(eµ) = ρ(2µ)e−µ for µ ∈ L and θL(−1) = −1.
Let θVL be the unique commutative algebra automorphism of VL such that θVL(h(n)⊗1) =
−h(n)⊗ 1 and θVL(1⊗ eµ) = 1⊗ θL(eµ) for h ∈ H and µ ∈ L.
Let λ be a representative of a coset in L/L. Let θλ: Vλ+L → V−λ+L be the linear
map defined by θλ(x ⊗ eλ+µ) = ρ(2µ)θH(x) ⊗ e−λ−µ, where θH acts on S(H ⊗ t−1C[t−1])
as the unique algebra automorphism such that θH(h(n)) = −h(n) for h ∈ H . Then
θλ is an isomorphism of VL-modules. When λ ∈ L/2, the linear map Vλ+L → V−λ+L,
x 7→ ρ(−2λ)x is an isomorphism of VL-modules and we have the automorphism of Vλ+L
θλ(x ⊗ eλ+µ) = ρ(2λ + 2µ)θH(x) ⊗ e−λ−µ. By abuse of notation, we denote θλ by θVL ,
and we view θVL as an automorphism of VL . For each θVL-stable subspace M of VL , let
M± denote the ±1 eigenspaces of M with respect to θVL respectively. Note that V +L is a
subVOA of VL.
Let KL = θL(a)a−1|a ∈ L be the subgroup of L. Let T0 and T1 be irreducible
L/KL-modules over C such that eα acts by ρ(α) and eα acts by −ρ(α) respectively. Set
V TiL = S(H⊗t−1/2C[t−1])⊗Ti. Then it has an irreducible θVL-twisted VL module structure.
Moreover, V T0L and V T1
L give all the inequivalent irreducible θVL-twisted VL-modules (cf.
[D2]).
Let θH be the unique commutative algebra automorphism of S(H ⊗ t−1/2C[t−1]) such
that θH(h(n)) = −h(n) for h ∈ H and n ∈ 1/2+Z. We consider the linear automorphism
of V TiL given by θH ⊗ 1Ti , where 1Ti is the identity operator of Ti. By abuse of notation,
we denote both automorphisms of V T0L and V T1
L by θV TL . Since α = θVL(α) on Ti, θV TL is
an automorphism as a VL-module. We denote the ±1 eigenspaces of V TiL with respect to
θV TL by V Ti,±L respectively. Then V Ti,±
L becomes an irreducible V +L -module.
Note that V +L and its irreducible modules defined above are isomorphic to those given
in [DN]. By Theorem 5.13 of [DN], the set V ±L , V
±α/2+L, V
Ti,±L , Vrα/2k+L|i = 0, 1, r =
1, . . . , k − 1gives all inequivalent irreducible V +L -modules.
2.2 Moonshine module
Let us review the moonshine module V from [FLM] for what we need in this paper.
3
Let Λ be the Leech lattice with the positive-definite Z-bilinear form 〈· , · 〉. It is a
unique positive-definite even unimodular lattice of rank 24 without roots.
Let Λ be the central extension of Λ by the cyclic group 〈−1〉 ∼= Z2:
1 → 〈−1〉 → Λ → Λ → 0 (2.1)
with the commutator map c0(β, γ) = 〈β, γ〉+ 2Z for β, γ ∈ Λ.
Let θΛ be the automorphism of Λ given by θΛ(a) = a−1(−1)〈a,a〉/2 for a ∈ Λ. We denote
the center of Λ by Cent Λ. Set KΛ = θΛ(a)a−1|a ∈ Λ ⊂ Cent Λ. Then KΛ is a normal
subgroup of Λ; consider the group Λ/KΛ. Note that the center of Λ/KΛ is 〈−KΛ〉 ∼= Z2.
By Theorem 5.5.1 of [FLM], we have the following proposition.
Proposition 2.1. ([FLM]) Set G = Λ/KΛ and let χ be a character of Cent G = 〈−KΛ〉such that χ(−KΛ) = −1. Let A be a maximal abelian subgroup of G and let ψ: A →C× be a character with ψ|Cent G = χ. Then T = IndG
ACψ is the unique irreducible G-
module on which x ∈ Cent G acts by χ(x)1T , where Cψ is the one-dimensional A-module
corresponding to ψ. Moreover, T ∼= ⊕Cϕ, where ϕ ranges over the characters of A whose
restriction to Cent G is χ, and dim T = 212.
Set the induced Λ-module CΛ = C[Λ]⊗C[±1]C, where −1 acts on C as multiplication
by −1. We extend θΛ to C[Λ] linearly. Since θΛ fixes −1, we view the automorphism θΛ
as an automorphism of CΛ.Set h = C ⊗Z Λ. We regard Λ as a subgroup of h. We extend 〈·, ·〉 to a C-bilinear
form on h. Let T be the irreducible Λ/KΛ-module given in Proposition 2.1. Set VΛ =
S(h ⊗ t−1C[t−1]) ⊗ CΛ and V TΛ = S(h ⊗ t−1/2C[t−1]) ⊗ T . For convenience, we also
denote h⊗ tn by h(n) for h ∈ h and n ∈ Z/2. Let θVΛ be the unique commutative algebra
automorphism of VΛ such that θVΛ(h(n)⊗1) = −h(n)⊗1 and θVΛ(1⊗b) = 1⊗θΛ(b), whereh ∈ h, n ∈ Z<0 and b ∈ CΛ. Let θh be the unique commutative algebra automorphism
of S(h ⊗ t−1/2C[t−1]) such that θh(h(n)) = −h(n) for h ∈ h, n ∈ 1/2 + Z<0. Then
θV TΛ = θh ⊗ (−1T ) is an automorphism of V TΛ , where 1T is the identity operator on T .
Thus we have V = V +Λ ⊕ V T,+
Λ , where V +Λ is the θVΛ-fixed-point subspace of VΛ and V T,+
Λ
is the θV TΛ -fixed-point subspace of V TΛ .
2.3 2k-frames and codes over Z2k
In this subsection, we give some terminology on a code over Z2k (cf. [DHS]) and formulate
the inverse of the generalized Construction A relating lattices and codes.
4
A (linear) code C of length n over Z2k is a Z2k-submodule of Zn2k. We fix an ordered
basis of Zn2k and denote i-th element of this basis by (0, . . . , 0, 1, 0, . . . , 0), where 1 only
appears in the i-th position. An element of C is called a codeword. The Euclidean
weight is Ewt(·) on Zn2k is given by Ewt(c) =∑24
i=1m2i , where c = (m1, . . . , mn) ∈ Zn2k
and −k < mi ≤ k for i = 1, . . . , n. We define the inner product of x and y in Zn2k by
〈x, y〉 =∑n
i=1 xiyi (mod 2k), where x = (x1, . . . , xn) and y = (y1, . . . , yn). The dual code
C⊥ of C is defined as C⊥ = x ∈ Zn2k|〈x, y〉 = 0for all y ∈ C. C is self-orthogonal if
C ⊂ C⊥ and C is self-dual if C = C⊥. We define a Type II code over Z2k to be a self-dual
code with all codewords having Euclidean weight divisible by 4k.
The symmetric group Sn acts on Zn2k by the permutation of the coordinate positions.
Let εi be the operator which changes the signs of the i-th position of Zn2k. Set Π = 〈εi|i =1, . . . , n〉, and Sn = 〈Sn,Π〉 = Sn ⋉ Π, since (i j)εi(i j) = εj for i, j ∈ 1, . . . , n. Two
codes C0 and C1 over Z2k are called equivalent if they both have length n and if there
exists σ ∈ Sn such that C0 = σ(C1).
Definition 2.2. A 2k-frame of a lattice of rank n is a set of n mutually orthogonal pairs
of opposite vectors of squared length 2k.
We denote the group of isometries of a lattice Λ by Aut(Λ). Since Aut(Λ) preserves
the inner product, Aut(Λ) acts on the set of 2k-frames of Λ. We say that 2k-frames S0,
S1 of Λ are equivalent if there exists σ ∈ Aut(Λ) such that S0 = σ(S1).
It is well known that one of the constructions of a lattice from a code over Z2k is the
generalized Construction A and such a lattice contains a 2k-frame (cf. [CS, BDHO]). Let
us consider the inverse map of the correspondence.
First, we recall the generalized Construction A. Let π: Zn → Zn2k be the canonical
map. Let C be a code of length n over Z2k. By the generalized Construction A, a positive-
definite lattice of rank n is given by A2k(C) = π−1(C)/√2k. Let (e1, . . . , e24) be the
standard orthogonal basis of Z24. Then Λ has a 2k-frame FC = ±√2ke1, . . . ,±
√2ke24.
Let C0, C1 be equivalent codes over Z2k and let σ be an element of Sn such that C0 =
σ(C1). Then we have A2k(C0) = σ(A2k(C1)) and FC0 = σ(FC1), where we regard σ as an
element of O(24). Therefore the equivalent codes correspond to the same lattice (up to
isomorphism), and the corresponding 2k-frames of such a lattice are equivalent. Hence
we obtain the map from the equivalence class of codes over Z2k to the equivalence class
of 2k-frames of the lattice.
We define an extremal Type II code of length 24 over Z2k to be a Type II code with
minimum Euclidean weight 8k. Note that for an extremal Type II code C of length 24
over Z2k, A2k(C) is an even unimodular lattice with no roots, namely, A2k(C) is the Leech
lattice.
5
Let Λ be the Leech lattice and suppose k ≥ 2. Let S be a 2k-frame of Λ and let N be
the sublattice of Λ generated by S. We choose an ordered basis (x1, . . . , x24) of N from
S. Since N ⊂ Λ ⊂ N = N/2k, we have the isomorphism
N/N → Z242k,
24∑
i=1
cixi/2k +N 7→ (c1, . . . , c24).
By the above isomorphism, we view Λ/N ⊂ N/N as a code of length 24 over Z2k. We
denote the codeword corresponding to v ∈ Λ by cv. For v, w ∈ Λ, we have Ewt(cv) =
2k ×min〈u, u〉|u ∈ v +N and 〈cv, cw〉 = 2k〈v, w〉 (mod 2k). Thus Λ/N is an extremal
Type II code of length 24 over Z2k. Note that the code Λ/N depends on the choice of the
ordered basis. However, it is easy to see that its equivalence class is uniquely determined
by the 2k-frame S. Therefore we denote the equivalence class of Λ/N by ϕ(S). For any
σ ∈ Aut(Λ), we have Λ/N = σ(Λ)/σ(N) = Λ/σ(N). Hence the equivalence class of codes
is uniquely determined by the equivalence class of the 2k-frame. Then ϕ is a map from
the set of the equivalence classes of extremal Type II codes over Z2k to the set of the
equivalence classes of 2k-frames of the Leech lattice.
Proposition 2.3. The map ϕ is bijective. Moreover, ϕ is the inverse map of the map
given by the generalized Construction A.
Proof. Let C be an extremal Type II code of length 24 over Z2k and let FC be the
2k-frame of Λ given by the generalized Construction A. It is easy to see that C is an
element of the equivalence class ϕ(FC). Therefore ϕ is surjective. Next, we will show
that ϕ is injective. Let F = ±x1, . . . ,±x24 be a 2k-frame of Λ and let C be a code
of ϕ(F ). Let FC be the 2k-frame corresponding to a code C. It is suffices to show that
F and FC are equivalent. Note that Λ = ∑ cixi/2k|(c1, . . . , c24) ∈ C and A2k(C) =
∑
ciei/√2k|(c1, . . . , c24) ∈ C. Then we have the isomorphism of the lattice
A2k(C) → Λ,∑
ciei/√2k 7→
∑
cixi/2k,
and this map sends FC to F . Therefore FC is equivalence to F .
Proposition 2.4. Let S be a 2k-frame of Λ and let N be the sublattice of Λ generated by
S. Let C = Λ/N be a code over Z2k and let C2 = (Λ ∩ (N/2))/N be a binary code. Set
m = dimC2 over Z2.
(1) |Λ/N | = (2k)12.
(2) m ≥ 12.
6
(3) |N/(2Λ ∩N)| is an elementary abelian 2-group with order 224−m.
(4) If k is odd, then C2 is a Type II code.
Proof. For any lattice R, we denote the dual lattice of R by R. Consider the inclusion
N ⊂ Λ = Λ ⊂ N = N/2k. It is well known that |Λ/N | = |N/Λ|. Since N/N ∼= Z242k,
we have |Λ/N | = |N/Λ| = (2k)12. Thus we obtain (1). Since for each element of C,
its order is divisible by 2k, we obtain (2). Consider the inclusion 2N ⊂ 2Λ ∩ N ⊂ N
and we obtain |N/(2Λ ∩ N)||(2Λ ∩ N)/2N | = |N/2N | = 224. Since (Λ ∩ (N/2))/N ∼=(2Λ∩N)/2N ∼= Zm2 , we have |N/(2Λ∩N)| = 224−m and we obtain (3). Suppose k is odd.
Let v =∑
cixi/2, w =∑
dixi/2 be elements of Λ∩ (N/2) and let cv = (c1, . . . , c24), cw =
(d1, . . . , d24) be corresponding codewords of C2. Then we have 〈v, w〉 = (k/2)∑
cidi ∈ Z,
namely 〈cv, cw〉 =∑
cidi ∈ 2Z. Thus C2 is self-orthogonal. Since C2 is a Syolw 2-group,
the dimension of C2 is 12, and C2 is self-dual. Since 〈v, v〉 ∈ 2Z, we have 〈cv, cv〉 ∈ 4Z.
Hence we obtain (4).
3 Decomposition of V and Z2k codes
In this section, for an integer k ≥ 2, we give the decomposition of V as a (V +L )⊗24-module
associated with an extremal Type II code over Z2k.
For convenience, we use the following notation. For c = (c1, . . . , c24) ∈ Z242 , we set
M(c)+ =⊕
ei∈±∏
ei=+
24⊗
i=1
V eiciα/2+L
,
MT (c)− =⊕
ei∈±∏
ei=−
24⊗
i=1
VTci ,ei,
L ,
where we regard ci ∈ Z2 as an element of Z.
Remark 3.1. Let c = (c1, . . . , c24) be a codeword of Z242 .
(1) M(c)+ is the fixed-point subspace of ⊗Vciα/2+L with respect to ⊗θVL .
(2) MT (c)− is the −1 eigenspace of ⊗V TciL with respect to ⊗θV TL .
For c = (c1, . . . , c24) ∈ Z242k, we set
M(c) =
24⊗
i=1
Vciα/2k+L.
Note that M(c) ∼= M(σ(c)) as a (V +L )⊗24-module for c ∈ Z24
2k and σ ∈ Π, where Π is the
set of the operators which change signs of some positions.
7
3.1 Main result
The following is our main theorem.
Theorem 3.2. Let C be an extremal Type II code of length 24 over Z2k and set C2 =
(c1, . . . , c24) ∈ C|ci = 0 (mod k) for all i, which is binary code. Set m = dimC2 over
Z2.
(1) There is an embedding
V ⊃ V +Λ ⊃ (V +
L )⊗24
of subVOAs.
(2) V +Λ is decomposed as a (V +
L )⊗24-module as follows:
V +Λ
∼=⊕
c∈C2
M(c)+ ⊕ 1
2
⊕
c∈C\C2
M(c).
(3) V T,+Λ is decomposed as a (V +
L )⊗24-module as follows:
V T,+Λ
∼=⊕
c∈C⊥2
2m−12MT (c)−.
Remark 3.3. This decomposition is uniquely determined by the extremal Type II code
of length 24 over Z2k, up to the permutation of the 24 tensor products for each components
of V .
Remark 3.4. [Ch, GH] For any integer k ≥ 2, there exists an extremal Type II code of
length 24 over Z2k. However, the classification of extremal Type II codes of length 24
over Z2k is still missing.
The rest of this section is devoted to the proof of the theorem.
3.2 Existence of (V +L )⊗24 in V
In this subsection, we will show Theorem 3.2 (1). In order to do it, we have to construct
an appropriate 2-cocycle of Λ.
Let S be a 2k-frame corresponding to C and let N be the sublattice of Λ generated
by S.
8
Lemma 3.5. There exists a set I of representatives of Λ/N satisfying the following con-
ditions:
(H1) (Λ ∩ (N/2)) ∩ I = ∑m
i=1 dihi|di ∈ 0, 1, where hi ∈ Λ ∩ (N/2).
(H2) For any λ ∈ I \ ((Λ ∩N/2) ∩ I), −λ ∈ I.
Proof. Let h1 + N, · · · , hm + N be a set of generators of (Λ ∩ (N/2))/N . Then
we choose ∑m
i=1 dihi|di ∈ 0, 1 as the set of representatives of (Λ ∩ (N/2))/N . For
λ ∈ b + N ∈ Λ/N with b /∈ N/2, we choose the representative of −b +N ∈ Λ/N as −λ.Therefore we obtain the set of representatives satisfying (H1) and (H2).
Lemma 3.6. Let ρ: 2Λ ∩ N → 〈−1〉 be a group homomorphism such that ρ(2γ) =
(−1)〈γ,γ〉 for γ ∈ N . There exist a set-theoretic section Λ → Λ, β 7→ eβ and a set I of
representatives of Λ/N satisfying
eβeγ = eβ+γ , (3.1)
θΛ(eλ+β) =
ρ(2λ+ 2β)e−λ−β if λ ∈ I ∩N/2,ρ(2β)e−λ−β if λ ∈ I ∩ (Λ \ N/2), (3.2)
for any β ∈ N , γ ∈ Λ.
Proof. Let (γ1, . . . , γ24) be an ordered basis of Λ. Consider the Z-bilinear form ε0
Λ× Λ → Z2 defined by
ε0(γi, γj) =
〈γi, γj〉 if i > j,〈γi, γi〉/2 if i = j,
0 if i < j,
for i, j ∈ Ω. It is easy to see that ε0 satisfies
ε0(β, β) = 〈β, β〉/2, (3.3)
ε0(β, γ)− ε0(γ, β) = 〈β, γ〉, (3.4)
for β, γ ∈ Λ. In other words, ε0 is a 2-cocycle corresponding to (2.1).
Let s be a set-theoretic section Λ → Λ corresponding to ε0. Then s(β)s(γ) = s(β +
γ)(−1)ε0(β,γ) for β, γ ∈ Λ. By (3.3), we have θΛ(s(β)) = s(−β)(−1)ε0(β,β)(−1)〈β,β〉/2 =
s(−β) for β ∈ Λ. Since 〈N,N〉 ⊂ 2Z, there exists a 2-coboundary η: N → Z2 such that
ε0(β, γ) = η(β + γ)− η(β)− η(γ) (3.5)
for β, γ ∈ N .
Let I be a set of representatives of Λ/N satisfying (H1) and (H2) of Lemma 3.5. Set
ρ(β) = (−1)ρ0(β) for β ∈ 2Λ ∩N . Let us show that we can choose η so that it satisfies
η(2λ) = ρ0(2λ) (3.6)
9
for any λ ∈ I ∩ (Λ ∩N/2).Set Ω0 = 1, . . . , m. Let hi|i ∈ Ω0 be a subset of I satisfying (H1), namely,
∑m
i=1 dihi|di ∈ 0, 1 = I ∩ (Λ ∩ (N/2)). Note that the set hi +N |i ∈ Ω0 is a set of
generators of (Λ ∩ (N/2))/N , namely, 2hi + 2N is a set of generators of (2Λ ∩N)/2N .
Let δ: (2Λ ∩ N)/2N → Z2 be the linear map defined by δ(2hi + 2N) = η(2hi) + ρ0(2hi)
for i ∈ Ω0. Since η + ρ0 is linear on 2Λ ∩ N , we have δ(2h + 2N) = η(2h) + ρ0(2h) for
h ∈ I ∩ (Λ∩ (N/2)). Note that η+ ρ0 is not zero on 2N , and so δ depend on a choice of a
set satisfying (H1). By Proposition 2.4, (2Λ∩N)/2N ∼= Zm2 is a subgroup of N/2N ∼= Z
242 .
Thus we can extend the linear map δ to N/2N . Therefore there exists a linear map δ:
N → Z2 such that δ(2h) = η(2h) + ρ(2h) for h ∈ I ∩ (Λ∩ (N/2)). Hence we may assume
that η satisfies (3.6), since η + δ satisfies (3.5) and (3.6).
Since ε0 is Z-bilinear, we can extend η to Λ so that
ε0(β, γ) = η(β + γ)− η(β)− η(γ) (3.7)
for β ∈ N , γ ∈ Λ. In fact, for λ ∈ I, we choose a value of η(λ), and we define η(β + λ) =
ε0(β, λ) + η(β) + η(λ) for β ∈ N . It is easy to see that η satisfies (3.7).
Now, let us modify the section s. Set s0(β) = (−1)η(β)s(β) for β ∈ Λ. Then s0 satisfies
(3.1), and we have
θΛ(s0(λ+ β)) = s0(−λ− β)(−1)η(λ+β)+η(−λ−β) = s0(−λ− β)(−1)η(λ)+η(−λ)+〈β,β〉/2 ,
for λ ∈ I and β ∈ N . By (H2) of Lemma 3.5, we can choose a value of η(λ) satisfying
η(λ) + η(−λ) = 0, for λ ∈ I \ (I ∩ N/2). Then s0 satisfies (3.2) on Λ \ (Λ ∩ N/2). Note
that η(β) + η(−β) = η(2β) for β ∈ Λ ∩N/2. By (3.6), we have the results.
Let ι: C[Λ] → CΛ be the canonical homomorphism. For a subset M of Λ, let CMdenote the subspace of CΛ generated by ι(a)|a ∈M. Let s: Λ → Λ be a section and
let I be a set of representatives of Λ/N satisfying (3.1). Consider the linear isomorphism
CΛ ∼= C[Λ], ι(s(β)) 7→ eβ for β ∈ Λ. By (3.1), the restriction to CN is an algebra
isomorphism and the restriction to Cλ+N is an isomorphism of those algebra modules
with respect to the left multiplication.
Let (x1, . . . , x24) be the ordered basis of N consisting of the 2k-frame S. For a repre-
sentative λ + N of Λ/N , we set Vλ+N = S(h ⊗ t−1C[t−1]) ⊗ Cλ + N ⊂ VΛ, where we
regard λ+N as a subset of Λ. Note that VN is a subVOA of VΛ. Let Li be the rank one
sublattice of N generated by xi. It is well known that VN ∼=⊗24
i=1 VLi∼= V ⊗24
L . Let ρi:
Li → C be the group homomorphism such that ρi(2µ) = (−1)〈µ,µ〉/2 corresponding to VLiand set ρ = ⊕24
i=1ρi. For x ∈ 2Λ ∩N , we have ρ(2x) = 1, namely, ρ(x) ∈ 〈−1〉. By (3.2),
we obtain θVΛ = ⊗θVL on VN . Hence we have V +Λ ⊃ V +
N ⊃ (V +L )⊗24. It is obvious that
(V +L )⊗24 is a subVOA of V +
Λ . Thus we have obtained Theorem 3.2 (1).
10
Remark 3.7. Let V +L,R and V
Rbe the real forms of V +
L and V as constructed in [FLM].
By [FLM], those have the positive-definite symmetric bilinear forms. By Theorem 3.2 (1),
we have the embedding V +L,R ⊂ V
R. In this embedding, the positive-definite symmetric
bilinear form is invariant.
3.3 Decomposition of the untwisted space V +Λ
In this subsection, we give the decomposition of V +Λ and show Theorem 3.2 (2).
From [D1], it follows that VΛ is decomposed as
VΛ =⊕
λ+N∈Λ/N
Vλ+N (3.8)
as a VN -module. Let (x1, . . . , x24) be the ordered basis of N given in Section 3.2. Let
λ =∑24
i=1 cixi/2k be an element of I. It is well known that
Vλ+N ∼=24⊗
i=1
Vciα/2k+L =M(c) (3.9)
as a (V +L )⊗24-module, where c = (c1, . . . , c24) ∈ C = Λ/N . By (3.2), we have θVΛ = ⊗θVL
on VΛ.
Next, we consider the θVΛ-fixed-point subspace of VΛ. We obtain the following lemma.
Lemma 3.8. Let c be a codeword of C = Λ/N .
(1) If c ∈ C2 = (Λ ∩N/2)/N , then M(c) is θVΛ-invariant, and M(c)+ is the θVΛ-fixed-
point subspace.
(2) If c ∈ C \C2, then M(c)⊕M(−c) is θVΛ-invariant, and the θVΛ-fixed-point subspace
(M(c)⊕M(−c))+ ∼=M(c) as a (V +L )⊗24-module.
Proof. By Remark 3.1, we obtain (1). Suppose c ∈ C\C2. Note that θVΛ is an involution
and maps M(c) toM(−c). Then M(c)⊕M(−c) is θVΛ-invariant and (M(c)⊕M(−c))+ is
generated by x+ θVΛ(x) for x ∈M(c). Since θVΛ commutes with the action of (V +L )⊗24 on
M(c) ⊕M(−c), we get an isomorphism of (V +L )⊗24-modules (M(c) ⊕M(−c))+ → M(c)
by x+ θVΛ(x) 7→ x. Hence we get (2).
By Lemma 3.8, (3.8) and (3.9), we obtain Theorem 3.2 (2).
11
3.4 Decomposition of the twisted space V T,+Λ
In this subsection, we give the decomposition of V T,+Λ and show Theorem 3.2 (3).
Set K(N) = KΛ ∩ N and K(2N) = KΛ ∩ ˆ2N . Since T is a Λ/KΛ-module, T is
a NKΛ/KΛ-module. Note that −1 ∈ Λ acts on T as multiplication by −1. Let M
be a maximal abelian subgroup of Λ/KΛ such that M ⊃ NKΛ/KΛ. Note that |M | =213. By Proposition 2.1, T decomposes into the direct sum of all the irreducible M-
modules on which −KΛ acts by −1. Therefore T decomposes into the direct sum of
irreducible NKΛ/KΛ-modules on which −KΛ acts by −1. By Proposition 2.4 (3), we
have |NKΛ/KΛ| = 2|N/(N ∩ 2Λ)| = 225−m, and any multiplicities of the irreducible
NKΛ/KΛ-module is 2m−12. Since NKΛ/KΛ∼= N/K(N), T is a N/K(N)-module. Thus
we obtain
T ∼=⊕
ψ∈Irr(N/K(N))ψ(−K(N))=−1
2m−12Tψ, (3.10)
where Irr(N/K(N)) is the set of the characters for N/K(N) and Tψ is the one-dimensional
module C with a character ψ.
By using the canonical map π: N/K(2N) → N/K(N), we view T as a N/K(2N)-
module. Note that the kernel of π is K(N)/K(2N). Therefore, as a N/K(2N)-module,
T is decomposed
T ∼=⊕
χ∈Irr(N/K(2N))χ(−K(2N))=−1
mχTχ, (3.11)
where mχ is the multiplicity. By (3.10), we have mχ ∈ 0, 2m−12. By (3.11), V TΛ has a
decomposition,
V TΛ
∼=⊕
χ∈Irr(N/K(2N))χ(−K(2N))=−1
mχVTχN , (3.12)
where VTχN is the θVΛ-twisted VN -module corresponding to N/K(2N)-module Tχ.
As Section 3.2, we fix the ordered basis of N as (x1, . . . , x24). Let Li = Zxi be a
sublattice of N and set K(Li) = KΛ ∩ Li. For χ ∈ Irr(N/K(2N)), we set the character
of Li/K(Li) χi(exi) = χ(exi). For a character χ = ⊗24
i=1χi such that χ(−K(2N)) =
−1, we set P (χ) = (c1, . . . , c24) ∈ Z242 , where χi(e
xi) = ρi(xi)(−1)ci . For exK(2N) =
(∏
edixi)K(2N) ∈ N/K(2N), we set Q(x) = (d1, . . . , d24) ∈ Z242 . Then we have χ(x) =
ρ(∑
dixi)(−1)∑
cidi = ρ(x)(−1)〈P (χ),Q(x)〉.
Now, let us determine the multiplicities mχ. Let χ be an element of Irr(N/K(2N))
such that χ(−K(2N)) = −1. Set n = 2m−12. Then mχ = n if and only if there exists χ ∈
12
Irr(N/K(N)) such that χ = χ π. If χ(x) = 1 for any x ∈ Kerπ = K(N)/K(2N), then
χ is regarded as a character of N/K(N) and mχ = n. Note that K(N) = θΛ(a)a−1|a ∈Λ ∩ N/2 = ρ(−β)eβ|β ∈ 2Λ ∩ N. Therefore mχ = n if and only if χ(x) = 1 for
any x ∈ K(N)/K(2N), namely, 〈P (χ), Q(x)〉 = 0 for any x ∈ K(N)/K(2N). Note that
K(N)/K(2N) ∼= (N ∩ 2Λ)/2N ∼= (Λ ∩ (N/2))/N = C2. Hence mχ = n if and only if
P (χ) ∈ C⊥2 . Then we have an isomorphism
V TΛ
∼= 2m−12⊕
χ∈Irr(N/K(2N))χ(−K(2N))=−1
P (χ)∈C⊥2
VTχN . (3.13)
Next, we fix a character χ and consider the space VTχN . If P (χ) = (c1, . . . , c24), then
Tχ = ⊗Tci. Thus, we get an isomorphism VTχN
∼= ⊗V TciL . Note that we have θV TΛ =
−(⊗θV TL ), because θV TΛ acts by −1 on Tχ and θV TL acts by 1 on Tci. Therefore, for χ ∈Irr(N/K(2N)) with χ(−K(2N)) = −1, we have an isomorphism of (V +
L )⊗24-modules
VTχ,+N
∼=MT (P (χ))−. (3.14)
By (3.13) and (3.14), we get Theorem 3.2 (3).
4 Character and automorphisms of the moonshine
module
In this section, we give the character of V and give some automorphisms of V .
4.1 Character of the moonshine module
In order to give the character of V , we consider the symmetrized weight enumerator of
a code over Z2k. The symmetrized weight enumerator of a code C over Z2k is defined as:
sweC(p0, . . . , pk) =∑
c∈C
pn0(c)0 p
n1(c)1 . . . p
nk(c)k ,
where ni(c) denotes the number of j such that cj = ±i. Note that the symmetrized weight
enumerators of equivalence codes are same. For 0 ≤ i ≤ k, we set
ai = chViα/2k+L =1
φ(q)
∑
j∈Z
qk(j+i/2k)2
,
b = (chV +L − chV −
L ) =1
φ(q)
∑
j∈Z
(−1)jqj2
,
where φ(q) =∏
n≥1(1− qn).
By Theorem 3.2, we obtain the following corollary.
13
Corollary 4.1. Let C be an extremal Type II code of length 24 over Z2k. Then we have
ch(V ) = q(J(q)− 744) =1
2sweC(a0, a1, . . . , ak) +
1
2b24 + 211q3/2
( φ(q)24
φ(q1/2)24− φ(q2)24φ(q1/2)24
φ(q)48)
.
Remark 4.2. We have sweC(a0, a1, . . . , ak) = ΘΛ(q)/(φ(q))24, where ΘΛ(q) is the theta
function associated with Λ. Since b = φ(q)/φ(q2), we obtain
ch(V ) =1
2
(ΘΛ(q)
φ(q)24+
φ(q)24
φ(q2)24)
+ 211q3/2( φ(q)24
φ(q1/2)24− φ(q2)24φ(q1/2)24
φ(q)48)
.
This equation is given in Remark 10.5.8 of [FLM].
4.2 Automorphisms of the moonshine module
By the fusion rules of V +L determined in [Ab], we have the following proposition (cf. [Ma]).
Proposition 4.3. (1) If k is odd then the linear map τ defined by setting
1 on V ±L , Viα/k+L|1 ≤ i ≤ (k − 1)/2,
−1 on V ±α/2+L, V(2i−1)α/2k+L|1 ≤ i ≤ (k − 1)/2,
√−1 on V T0,±
L ,−√−1 on V T1,±
L
is an automorphism of the fusion algebra of V +L .
(2) If k is even then the linear map τ defined by setting
1 on V ±L , V
±α/2+L, Viα/2k+L|1 ≤ i ≤ (k − 1),
−1 on V T0,±L , V T1,±
L
is an automorphism of the fusion algebra of V +L .
Let Γ be the set of inequivalent irreducible V +L -modules. Suppose given a decomposi-
tion
V =⊕
(W1,...,W24)∈Γ24
mW1,...W24W1 ⊗ · · · ⊗W24
of the moonshine module as a (V +L )⊗24-module, where mW1,...,W24 is the multiplicity. Con-
sider the map µ : Γ → C× defined by
τ(W ) = µ(W )W
14
for W ∈ Γ. For each i ∈ Ω, we define a linear automorphism σi of V by
σi(x) = µ(Wi)x
for x ∈⊗24
j=1Wj . Then we have the following proposition (cf. [Mi1]).
Proposition 4.4. σi is a VOA automorphism of V .
Proof. By the definition, σi fixes any element of (V +L )⊗24. In particular, σi fixes the
Virasoro element and the vacuum vector. Note that the fusion rules of tensor products
of modules are the tensor products of the fusion rules of those modules. Let Γ24 be the
set of inequivalent irreducible modules for (V +L )⊗24. Let µi: Γ24 → C× be a map such
that µi(W ) = µ(Wi) for W = ⊗Wj ∈ Γ24. Let W1, W 2 be elements of Γ24. For v ∈ W 1,
w ∈ W 2, we have Y (v, z)w ∈ (W 1 ×W 2)[z, z−1], where Y (·, z) is the vertex operator.
Therefore we have Y (σi(v), z)σi(w) = µi(W1)µi(W
2)Y (v, z)w = µi(W1×W 2)Y (v, z)w =
σi(Y (v, z)w).
Remark 4.5. If k = 3, then σi is a 4A element of the Monster (cf.[Ma]) and the McKay-
Thompson series T4A is given by
T4A(q) = q−1∑
(W1,...,W24)∈Γ24
mW1,...,W24 µ(Wi)
24∏
i=1
chWi.
Proposition 4.6. Suppose k is odd. In the decomposition of V given by Theorem 3.2, σi
is a 4A element of the Monster. More precisely, σi ∈ 21+24+ ⊂ 21+24
+ .Conway1 ⊂ Monster ,
where 21+24+ .Conway1 is a non-split extension of Conway1 (largest simple Conway group)
by the extra-special group 21+24+ .
Proof. By Theorem 3.2 and Proposition 4.3, σ2i acts by 1 on V +
Λ and acts by −1 on
V T,+Λ . By [FLM], the centralizer of σ2
i in the Monster simple group is 21+24+ .Conway1 .
Then we have σi ∈ 21+24+ .Conway1 . Since σi preserves the space Ceβ for any β ∈ Λ, we
have σi ∈ 21+24+ . It is well known that there are the four types of Monster elements 1A,
2A, 2B and 4A contained in 21+24+ . Therefore σi is a 4A element of the Monster.
Remark 4.7. Suppose k is even. In the decomposition of V given by Theorem 3.2, σi
is a 2B element of the Monster. More precisely, σi acts by 1 on VΛ and acts by −1 on
V TΛ , namely, σi = z given in (10.4.48) of [FLM].
5 Examples of a decomposition
In this section, we give examples of a 2k-frame for some k and the corresponding de-
composition of V . Accordingly, we give an expression of the McKay-Thompson series
T4A.
15
5.1 Examples of a 2k-frame
In this subsection, we recall the Golay code and the Leech lattice. Then we give examples
of a 2k-frame for some k.
Let G24 ⊂ P(Ω) (the power set of Ω) be the binary Golay code. Then G24 is a vector
space over F2 of dimension 12. Let αi|i ∈ Ω be a basis of R24 such that 〈αi, αj〉 = 2δi,j,
where 〈·, ·〉 is the symmetric bilinear form on R24. For B ⊂ Ω, we set αB =∑
i∈B αi.
Proposition 5.1. ([CS]) The Leech lattice Λ can be realized as follows:
Λ =∑
B∈G24
Z1
2αB +
∑
i∈Ω
Z(1
4αΩ − αi).
Proof. Since G24 is a Type II code, the lattice Λ1 =∑
B∈G24Z(1/2)αB +
∑
i∈Ω Zαi is
even unimodular. Λ is unimodular if and only if Λ1 is, since the lattice Λ0 = Λ1 ∩ Λ =∑
B∈G24Z(1/2)αB +
∑
k,l∈Ω Z(αk +αl) has index 2 in both Λ1 and Λ. One can check that
Λ is even and has no roots.
Let p be a prime such that p = 3 (mod 4). In order to construct a 2k-frame, let us
consider the quadratic residues in Fp. Let Q be the set of quadratic residues in Fp. Note
that |Q| = (p − 1)/2 For i ∈ Fp, we set the subset Qi = i + j ∈ Fp|j ∈ Q = x ∈Fp|x− i ∈ Q ⊂ Fp.
Proposition 5.2. (1) |Qi ∩Qj | = (p− 3)/4, for any i, j ∈ Fp with i 6= j.
(2) For any i, j ∈ Fp such that i 6= j, we have exactly one of the following:
(I) i ∈ Qj, (II) j ∈ Qi.
Proof. Since |F×p | is not divisible by 4, Q does not contain β = −1. We view Q as a
subgroup of F×p , and let γ be a generator of Q. Note that β, γ are generators of F×
p .
By the definition, we have Qi ∩ Qj = x ∈ Fp|x − j, x − i ∈ Q and Q ∩ Qi−j = x ∈Fp|x, x− (i − j) ∈ Q for i, j ∈ Fp. Therefore we have |Qi ∩ Qj| = |Qi−j ∩ Q|. We have
to show that |Qi ∩Q| = |Qj ∩ Q| for i, j ∈ F×p . Suppose j = βsγti. Let x be an element
of Qi ∩ Q. If s = 0, then γtx, γtx − j ∈ Q and if s = 1, then γtx, γtx + j ∈ Q, where
s ∈ 0, 1 such that s = s (mod 2). Then we have the map ϕi,j: Qi ∩ Q → Qj ∩ Q,
x 7→ γtx + δs,1j. Note that ϕi,j is injective. Let us show that ϕj,lϕi,j = ϕi,l. Suppose
j = βs1γt1i and l = βs2γt2j. Then l = βs1+s2γt1+t2i. Let x be an element of Qi ∩Q. Thenϕi,jϕj,l(x) = γt1+t2x+ δs1,1γ
t2j + δs2,1l = γt1+t2x+ (δs1,1βs2 + δs2,1)l. It is easy to see that
δs1,1βs2 + δs2,1 = δs1+s2,1.
16
On the other hand, ϕi,l(x) = γt1+t2x + δs1+s2,1l. Therefore we have ϕj,lϕi,j = ϕi,l. In
particular, we have ϕj,iϕi,j = ϕi,i, namely, ϕi,j is bijective. Hence, we have |Qi ∩ Q| =|Qj ∩ Q|. Let us show that |Qi ∩ Q| = (p − 3)/4. Set λ = |Qi ∩ Q|. For i ∈ F×
p ,
we consider the injective map ψi: Qi ∩ Q → Q × Q, x 7→ (x, x − i). Since a − b = i for
(a, b) ∈ ψi(Qi∩Q), we have ψ(Qi∩Q)∩ψ(Qj∩Q) = φ. Thus we obtain⋃
i∈F×pψi(Qi∩Q) =
(a, b) ∈ Q × Q|a 6= b. Therefore we have λ(p − 1) = (p − 1)/2 × (p − 3)/2, namely,
λ = (p − 3)/4. Hence we obtain (1). By the definition, i ∈ Qj if and only if i − j ∈ Q.
Similarly, j ∈ Qi if and only if j − i ∈ Q. Q contains only one of i − j, j − i, since−1 /∈ Q. Thus we obtain (2).
First, we consider the case when p = 23. Consider the following generators of G24 (cf.
[CS]):
A1 = Ω,
Ai = 1 ∪ 2 + j|j ∈ Qi−2, (i = 2, · · · , 24),
where we regard 2 + j|j ∈ Qi−2 as a subset of Ω. For i, j ≥ 2, j ∈ Ai if and only if
j − 2 ∈ Qp,i−2. By Proposition 5.2 (1), |Qi ∩Qj| = 5 if i 6= j. Thus we have the following
remark.
Remark 5.3. (1) |Ai ∩ Aj| = 6 for i 6= j ≥ 2.
(2) For i, j ≥ 2, we have exactly one of the following:
(I) i ∈ Aj , (II) j ∈ Ai.
Proposition 5.4. For l ∈ Z, the set ±x1, . . . ,±x24 defined by
x1 =1
4αΩ + (2l − 1)α1,
xi = −1
2αAi +
1
4αΩ + (2l − 1)αi, (i = 2, · · · , 24)
is a 2(4l2 − 3l + 2)-frame.
Proof. Set m = 2(4l2 − 3l + 2). By Proposition 5.1, Λ contains ±xi. We have to
check that 〈xi, xj〉 = mδi,j . By direct calculation, we have 〈x1, x1〉 = m and 〈x1, xi〉 = 0
for i ∈ Ω\1. Since i /∈ Qi, we have i /∈ Ai. Thus, for i ∈ Ω\1, we get 〈xi, xi〉 = m. By
Remark 5.3, we obtain 〈αAi, αj〉+〈αAj , αi〉 = 2 and |Ai∩Aj | = 6 , namely, 〈αAi, αAj〉 = 12
for i, j ∈ Ω\1 with i 6= j. Therefore we have 〈xi, xj〉 = 0 for i, j ∈ Ω\1 with i 6= j.
Next, we consider the case when p = 11. Let D be an element of G24 with |D| = 12
and set D = d1, . . . , d12. By Proposition 5.2 (1), |Qi ∩ Qj| = 2 if i 6= j. We consider
17
the bijictive map F11 → D \ d1, i 7→ di+2, and we regard Qi as a subset of D. Then we
set the following subsets of D
D1 = D,
Di+2 = d1 ∪Qi. (i = 0, 1, . . . , 10)
Set Ω \D = d13, . . . , d24. Consider the bijective map F11 → D \ d13, i 7→ di+14, and
we regard Qi as a subset of Ω \D. Then we set the following subsets of Ω \D
D13 = Ω \D,Di+14 = d13 ∪Qi. (i = 0, 1, . . . , 10)
Note that |Di| = 6 for i ∈ Ω \ 1, 13. By direct calculation, we have the following
proposition.
Proposition 5.5. For l ∈ Z, the set ±x1, . . . ,±x24 defined by
x1 =1
2αD1 + 2lαd1 ,
xi =1
2αD1 − αDi + 2lαdi , (i = 2, 3, . . . , 12)
x13 =1
2αD13 + 2lαd13 ,
xi =1
2αD13 − αDi + 2lαdi, (i = 14, 15, . . . , 24)
is a 2(4l2 + 2l + 3)-frame.
Remark 5.6. As above case, we can consider the case when p = 7. Then we obtain a
4(2l2 + l + 1)-frame.
For example, we consider the case l = 1 in Proposition 5.4 and we get the 6-frameand the corresponding extremal Type II code C of length 24 over Z6. Then we obtain its
18
symmetrized weight enumerator:
sweC(p, q, r, s) = s24 + 48r24 + 36432q2r16s6 + 97152q3r12s9
+ 36432q4r8s12 + 510048q5r16s3 + 3497472q6r12s6 + 1603008q7r8s9
+ 36432q8r16 + 4048q9s15 + 6800640q9r12s3 + 8962272q10r8s6
+ 61824q12s12 + 123648q12r12 + 5100480q13r8s3 + 242880q15s9
+ 36432q16r8 + 198352q18s6 + 24288q21s3 + 48q24
+ 13248p1q1r11s11 + 971520p1q3r15s5 + 2914560p1q4r11s8 + 582912p1q5r7s11
+ 4080384p1q6r15s2 + 36140544p1q7r11s5 + 14572800p1q8r7s8 + 24482304p1q10r11s2
+ 39055104p1q11r7s5 + 8743680p1q14r7s2 + 145728p2q1r14s7 + 218592p2q2r10s10
+ 48576p2q3r6s13 + 9472320p2q4r14s4 + 24773760p2q5r10s7 + 3934656p2q6r6s10
+ 8743680p2q7r14s1 + 123868800p2q8r10s4 + 45029952p2q9r6s7 + 24482304p2q11r10s1
+ 57076800p2q12r6s4 + 4080384p2q15r6s1 + 4048p3r9s12 + 24288p3r21
+ 2185920p3q2r13s6 + 2963136p3q3r9s9 + 218592p3q4r5s12 + 34974720p3q5r13s3
+ 97637760p3q6r9s6 + 11075328p3q7r5s9 + 5100480p3q8r13 + 183941120p3q9r9s3
+ 63391680p3q10r5s6 + 6800640p3q12r9 + 34974720p3q13r5s3 + 510048p3q16r5
+ 36432p4r12s8 + 36432p4q1r8s11 + 14427072p4q3r12s5 + 14281344p4q4r8s8
+ 728640p4q5r4s11 + 57076800p4q6r12s2 + 187406208p4q7r8s5 + 15665760p4q8r4s8
+ 1238688004q10r8s2 + 42989760p4q11r4s5 + 9472320p4q14r4s2 + 582912p5q1r11s7
+ 437184p5q2r7s10 + 42989760p5q4r11s4 + 37014912p5q5r7s7 + 1068672p5q6r3s10
+ 39055104p5q7r11s1 + 187406208p5q8r7s4 + 11075328p5q9r3s7 + 36140544p5q11r7s1
+ 14427072p5q12r3s4 + 971520p5q15r3s1 + 198352p6r18 + 3934656p6q2r10s6
+ 1603008p6q3r6s9 + 63391680p6q5r10s3 + 50276160p6q6r6s6 + 728640p6q7r2s9
+ 8962272p6q8r10 + 97637760p6q9r6s3 + 3934656p6q10r2s6 + 3497472p6q12r6
+ 2185920p6q13r2s3 + 36432p6q16r2 + 24288p7r9s8 + 11075328p7q3r9s5
+ 3133152p7q4r5s8 + 45029952p7q6r9s2 + 37014912p7q7r5s5 + 255024p7q8r1s8
+ 24773760p7q10r5s2 + 582912p7q11r1s5 + 145728p7q14r1s2 + 759p8s16
+ 255024p8q1r8s7 + 15665760p8q4r8s4 + 3133152p8q5r4s7 + 14572800p8q7r8s1
+ 14281344p8q8r4s4 + 24288p8q9s7 + 2914560p8q11r4s1 + 36432p8q12s4
+ 242880p9r15 + 728640p9q2r7s6 + 11075328p9q5r7s3 + 1603008p9q6r3s6
+ 1603008p9q8r7 + 2963136p9q9r3s3 + 97152p9q12r3 + 1068672p10q3r6s5
+ 3934656p10q6r6s2 + 437184p10q7r2s5 + 218592p10q10r2s2 + 728640p11q4r5s4
+ 582912p11q7r5s1 + 36432p11q8r1s4 + 13248p11q11r1s1 + 2576p12s12
+ 61824p12r12 + 218592p12q5r4s3 + 36432p12q8r4 + 4048p12q9s3
+ 48576p13q6r3s2 + 4048p15r9 + 759p16s8 + p24.
Remark 5.7. This symmetrized weight enumerator is equal to that of the lifted Golay
codes G246 given in [BDHO].
19
5.2 McKay-Thompson series for a 4A element of the Monster
Let us explicitly compute the decomposition with respect to the 6-frame obtained in the
previous subsection.
We set
ai = chViα/2k+L =1
φ(q)
∑
j∈Z
q3(j+i/6)2
(0 ≤ i ≤ 3),
b = (chV +L − chV −
L ) =1
φ(q)
∑
j∈Z
(−1)jqj2
.
By Remark 4.5, the McKay-Thompson series T4A is given by
T4A(q) =1
2a240 +
1
2b24 − 1
2a243 + 24a242 − 24a241
+253
2a160 a83 −
253
2a80a
163 + 2024a150 a92 − 2024a91a
153
+ 4048a70a92a
83 − 4048a80a
91a
73 + 6072a40a
122 a83 − 6072a41a
82a
123
+ 6072a120 a81a42 − 6072a80a
121 a43 + 6072a21a
162 a63 − 6072a60a
161 a22
+ 6072a81a162 − 6072a161 a82 + 8096a130 a61a
32a
23 − 8096a20a
31a
62a
133
+ 12144a30a212 − 12144a211 a33 + 24288a20a
11a
142 a73 − 24288a70a
141 a12a
23
+ 30912a120 a122 − 30912a121 a123 + 36432a120 a51a42a
33 − 36432a30a
41a
52a
123
+ 42504a80a11a
82a
73 − 42504a70a
81a
12a
83 + 85008a51a
162 a33 − 85008a30a
161 a52
+ 97152a50a11a
112 a73 − 97152a10a
51a
72a
113 + 97152a110 a71a
52a
13 − 97152a70a
111 a12a
53
+ 99176a60a182 − 99176a181 a63 + 121440a110 a41a
52a
43 − 121440a40a
51a
42a
113
+ 121440a90a21a
72a
63 − 121440a60a
71a
22a
93 + 121440a90a
152 − 121440a151 a93
+ 161920a10a31a
152 a53 − 161920a50a
151 a32a
13 + 178112a100 a31a
62a
53 − 178112a50a
61a
32a
103
+ 267168a90a81a
72 − 267168a71a
82a
93 + 364320a30a
21a
132 a63 − 364320a60a
131 a22a
33
+ 655776a100 a61a62a
23 − 655776a20a
61a
62a
103 + 655776a60a
21a
102 a63 − 655776a60a
101 a22a
63
+ 680064a10a61a
152 a23 − 680064a20a
151 a62a
13 + 850080a30a
81a
132 − 850080a131 a82a
33
+ 1457280a20a71a
142 a13 − 1457280a10a
141 a72a
23 + 1493712a60a
81a
102 − 1493712a101 a82a
63
+ 1578720a20a41a
142 a43 − 1578720a40a
141 a42a
23 + 1845888a90a
51a
72a
33 − 1845888a30a
71a
52a
93
+ 1845888a70a31a
92a
53 − 1845888a50a
91a
32a
73 + 2404512a40a
31a
122 a53 − 2404512a50a
121 a32a
43
+ 2428800a80a71a
82a
13 − 2428800a10a
81a
72a
83 + 2610960a80a
41a
82a
43 − 2610960a40a
81a
42a
83
+ 5829120a30a51a
132 a33 − 5829120a30a
131 a52a
33 + 6509184a50a
71a
112 a13 − 6509184a10a
111 a72a
53
+ 7164960a50a41a
112 a43 − 7164960a40a
111 a42a
53 + 7504992a70a
61a
92a
23 − 7504992a20a
91a
62a
73
+ 9512800a40a61a
122 a23 − 9512800a20a
121 a62a
43 + 10565280a60a
51a
102 a33 − 10565280a30a
101 a52a
63.
20
References
[Ab] T. Abe, Fusion Rules for the Charge Conjugation Orbifold, Preprint.
[BDHO] E. Bannai, S.T. Dougherty, M. Harada, and M. Oura, Type II Codes, Even
Unimodular Lattices, and Invariant Rings, IEEE Trans. Inform. Theory, 45,
(1999), 1194-1205.
[Bo] R. E. Borcherds, Vertex algebras, Kac-Moody algebras, and the Monster, Proc.
Nat’l. Acad. Sci. U.S.A, 83, (1986), 3068-3071.
[Ch] R. Chapman, Double circulant constructions of the Leech lattice, J. Aust. Math.
Soc. Ser. A, 69, (2000), 287-297.
[CS] J.H. Conway, N.J.A. Sloane, Sphere Packing, Lattices and Groups, 3rd Edition,
Springer, New York, 1999.
[D1] C. Dong, Vertex algebras associated with even lattices, J. Algebra, 161, (1993),
245-265.
[D2] C. Dong, Twisted modules for vertex algebras associated with even lattice, J.
Algebra, 165, (1994), 91-112.
[DGH] C. Dong, R. L. Griess Jr., and G. Hohn, Framed vertex operator algebras, codes
and moonshine module, Comm. Math. Phys., 193, (1998), 407-448.
[DHS] S. T. Dougherty, M. Harada and P. Sole, Self-dual codes over rings and the
Chinese remainder theorem. J. Math. Hokkaido Univ. 28 (1999), 253-283.
[DN] C. Dong and K. Nagatomo, Representations of Vertex operator algebra V +L for
rank one lattice L, Comm. Math. Phys., 202, (1999), 169-195.
[FLM] I. Frenkel, J. Lepowsky and A. Meurman, Vertex Operator Algebras and the
Monster, Pure and Appl. Math., Vol.134, Academic Press, Boston, (1988).
[GH] T. A. Gulliver, M. Harada, Orthogonal Frames in the Leech Lattice and a Type
II code over Z22, J. Combin. Theory. Ser. A, (to appear).
[Ma] A. Matsuo, Norton’s Trace Formulae for the Griess Algebra of a Vertex Operator
Algebra with Larger Symmetry. Preprint.
[Mi1] M. Miyamoto, Griess Algebras and Conformal Vectors in Vertex Operator Al-
gebras, J. Algebra, 179, (1996), 523-548.
21
[Mi2] M. Miyamoto, A new construction of the Moonshine vertex operator algebra
over real number field. Preprint.
[MN] A. Matsuo and K. Nagatomo, Axioms for a Vertex Algebra and the Locality of
Quantum Fields, MSJ-Memoirs 4, Mathematical Society of Japan, (1999).
22