← Modular Forms

Part 12 ended with Thompson's reading: the coefficients of J look like the dimensions of the graded pieces of one Monster representation. Conway and Norton turned that into a precise list of properties the representation must have, and Frenkel, Lepowsky and Meurman (FLM) built it, the moonshine module V^♮, announcing it in 1984 and giving the full construction in their 1988 book. Their tool was the vertex operator algebra (VOA), a framework drawn from string theory and conformal field theory in which the states of a graded space act on one another under very rigid rules. This part states the specification, introduces VOAs through the free boson, and follows the construction from the Leech lattice.

A note on notation: j(τ) = q⁻¹ + 744 + … is the classical modular invariant, while J(τ) = j(τ) − 744 = q⁻¹ + 0 + 196884q + … is the moonshine-normalised graded dimension. The moonshine module is built so that its graded dimension is J, not the unshifted j.

The specification

A ℤ-graded vector space is a direct sum of finite-dimensional pieces indexed by integers:

Each V_n is a grade. The whole space is infinite-dimensional, but linear algebra can be done one grade at a time. The graded dimension packages all the dimensions into one power series:

If a group G acts on V and each g ∈ G sends V_n into itself, then g has a trace on each grade, and its graded trace is

For g = 1 each trace is a dimension, so the graded trace of the identity is the graded dimension. For other elements the coefficients are character values, so knowing T_g for every g tells you how each grade splits into irreducible representations. The modular language of the earlier parts then translates line by line:

The Conway–Norton conjecture (1979) says there is a graded Monster module V^♮ = ⊕_{n ≥ −1} V_n^♮ with graded dimension J, and that for every g ∈ 𝕄 the graded trace T_g is the Hauptmodul of a genus-zero group. Three grades carry most of the structure:

Grade −1 has dimension 1: a copy of ℂ with trivial Monster action, giving the q⁻¹ term. Grade 0 is empty. In the vertex-algebra construction below it would form a Lie algebra, and exponentiating that would give V^♮ a continuous group of symmetries; an empty grade 0 leaves room for a finite automorphism group. That is why J has constant term 0, and the construction gets it from the fact that the Leech lattice has no vectors of norm 2. Grade 1 has dimension 196 884. It carries the commutative non-associative product Griess used in 1982 and splits as 1 ⊕ 196 883 under the Monster, which is McKay's observation.

Higher grades also split into Monster irreducibles. Thompson worked out the first few from the character table:

Dimensions alone leave the Monster action underdetermined; the traces pin it down. A Hauptmodul for a group Γ commensurable with SL₂(ℤ) is a single modular function that generates the whole field of Γ-invariant functions, and one exists exactly when the compactified quotient of ℍ by Γ has genus zero. j is the Hauptmodul for SL₂(ℤ). Conway and Norton proposed a specific Hauptmodul for each of the Monster's 194 conjugacy classes. Algebraically conjugate classes, such as 23A and 23B, share a series, so these amount to 171 distinct functions.

Vertex operators

A VOA lets each state v ∈ V act on every other through its vertex operator, a formal power series of linear maps in a complex variable z:

Each v_(n) is a linear endomorphism of V. In physics Y(v, z) is a field, an operator-valued distribution on the punctured complex line, and z is the worldsheet coordinate of a string.

The four axioms

A vertex operator algebra is a graded vector space V = ⊕_n≥0 V_n, graded by weight, with a vertex operator map Y, a vacuum vector |0⟩ ∈ V₀ and a conformal vector ω ∈ V₂, satisfying four axioms.

1. Vacuum: Y(|0⟩, z) = id_V, and Y(v, z)|0⟩ → v as z → 0.

2. Translation: there is an operator T with [T, Y(v, z)] = ∂_z Y(v, z) and T|0⟩ = 0, which is translation covariance on the worldsheet.

3. Locality: for any u, v, the operators Y(u, z) and Y(v, w) commute up to a pole,

for some N. This is the operator product expansion of conformal field theory reduced to algebra: colliding insertions give at worst finite-order poles.

4. Virasoro: ω generates a Virasoro algebra via Y(ω, z) = Σ L_n z⁻ⁿ⁻²,

c is the central charge. For a lattice VOA it equals the rank of the lattice, so V^♮, built from the 24-dimensional Leech lattice, has c = 24. The weight grading and the J-grading used above differ by one: the graded dimension is Σ dim V_k q^k−1, so the vacuum (weight 0) gives q⁻¹, the empty grade 0 is the weight-1 space, and the Griess algebra is the weight-2 space.

V^♮ is generated as a VOA by its weight-2 space, so an automorphism is fixed by what it does there, which is how FLM pin Aut(V^♮) down to the Monster.

The free boson

The simplest VOA is the free boson, or Heisenberg VOA, of central charge c = 1: a polynomial algebra in α₋₁, α₋₂, … with α_−n of degree n. Its graded dimensions are partition numbers, since dim V_n counts the multi-boson states at energy n:

From free bosons to lattice VOAs

d independent free bosons give central charge d. Compactifying on a lattice Λ ⊂ ℝ^d gives the lattice VOA V_Λ, with central charge d and graded dimension the theta series divided by η^d. Physically this is a bosonic string on ℝ^d/Λ.

For the Leech lattice Λ₂₄, the lattice VOA V_Λ24 has central charge 24:

V₁ has dimension 24, one boson per lattice direction, generating an abelian Lie algebra of continuous symmetries. J has constant term 0, so the weight-1 space of V^♮ must be empty; hence V_Λ₂₄ ≠ V^♮, and V^♮ is built from it as an orbifold instead.

In physics terms, a VOA axiomatises the chiral half of a two-dimensional conformal field theory, and the central charge counts target dimensions. Here c = 24 is the number of transverse directions of the bosonic string in its critical dimension 26. Borcherds imported the no-ghost theorem as a pure-algebra lemma.

Constructing V^♮: the FLM orbifold

FLM start from V_Λ24 and must remove its 24-dimensional weight-1 space.

The ℤ₂ orbifold

The Leech lattice has the involution θ: x ↦ −x. This lifts to an automorphism of V_Λ24 that acts as −1 on each oscillator αⁱ_−n (so as +1 on states with an even number of oscillators and −1 on odd) and sends each momentum state e^α to ±e^−α. The invariant subspace

has no weight-1 states, since all 24 oscillators are odd under θ, but it also loses too many states at higher weights: its graded dimension alone does not match j(τ) − 744.

The twisted sector

In orbifold conformal field theory, quotienting by a symmetry creates a twisted sector, a new space of strings that close only up to the action of the group element. For the θ-orbifold, the twisted sector V_Λ24^T consists of strings on the 24-torus whose boundary conditions are anti-periodic. This twisted module also decomposes under θ into ±1 eigenspaces, and we take the +1 part:

The moonshine module is the direct sum, a VOA of central charge 24:

The graded dimension is j(τ) − 744

Each piece has a graded dimension that can be written down and expanded. The Leech lattice VOA gives Θ_Λ/η²⁴, and since Θ_Λ = E₄³ − 720Δ this is j − 720. Averaging with the trace of θ keeps the invariant half. The twisted sector starts at weight 3/2 with 2¹² ground states, and its θ-invariant part is the integer-weight part. The figure below expands all of these and adds them up.

At weight 1 the Leech VOA has only the 24 oscillator states, because the Leech lattice has no roots, and θ negates all 24 of them. The twisted sector starts at weight 3/2, so it adds nothing at weight 1 either. That is where the empty grade 0 of J comes from. At weight 2 the two sectors contribute 98 580 and 98 304 = 2¹²·24, which add to 196 884.

The Monster as automorphism group

FLM's main theorem is Aut(V^♮) = 𝕄. The weight-2 space (grade 1 in the J-indexing) inherits a commutative non-associative product from the vertex operator, the v₍₁₎w coefficient, and becomes the 196884-dimensional Griess algebra. Griess built the Monster as a group of automorphisms of this algebra (published 1982), and Tits proved in 1983–84 that it is the full automorphism group. An automorphism of V^♮ is determined by what it does on the weight-2 space, so Aut(V^♮) embeds in the automorphism group of the Griess algebra, which is 𝕄. FLM then show every Monster element extends to a VOA automorphism, so the embedding is onto.

The twisted sector is essential here. On its own, V_Λ24⁺ has the wrong graded dimension, and its automorphism group is 2²⁴·Co₁, where Co₁ = Co₀/{±1} and Co₀ is the automorphism group of the Leech lattice. The symmetries of V^♮ that preserve the split into untwisted and twisted sectors form 2¹⁺²⁴·Co₁, which is the centraliser of an involution of class 2B in 𝕄. The extension does not split, so Co₁ appears here only as the quotient of that subgroup by 2¹⁺²⁴, not as a subgroup of it. FLM add one more automorphism that mixes the two sectors, and together with 2¹⁺²⁴·Co₁ it generates the whole Monster.

What FLM proved, and what remained

FLM established that V^♮ is a VOA with the correct graded dimension and automorphism group, but they did not prove the full Conway–Norton conjecture. They could compute graded traces for elements of the involution centraliser 2¹⁺²⁴·Co₁, but had no method that reached every one of the 194 conjugacy classes. Proving that every T_g(τ) is the predicted Hauptmodul took a Lie algebra large enough to encode the j-function's product expansion, with a denominator identity rigid enough to constrain the graded traces. That is Borcherds' work, the subject of Part 14.