Skip to content

Zhu Algebras, Characters, and Modular Invariance

Zhu’s associative algebra extracts the lowest-energy action of a VOA, and genus-one trace functions lift that algebraic information to modularly transforming holomorphic functions. For the Ising VOA, the Zhu algebra is a three-point quotient whose simple modules have lowest weights 00, 1/21/2, and 1/161/16; their characters close under explicit SS and TT matrices. Neither conclusion survives unchanged for a general logarithmic VOA, where ordinary characters may require pseudotraces to form a modularly closed space.

Required background. Vertex Operator Algebras: Axioms, Grading, and Locality supplies modes and conformal weights. Modules, Intertwining Operators, and Tensor Categories supplies admissible and ordinary modules.

Helpful background. Torus Partition Functions as Bootstrap Data gives the physical trace interpretation. Chiral Blocks, Sewing, and Modular Invariance supplies the genus-one continuation problem.

For homogeneous uVu\in V and arbitrary vVv\in V, define

uv=Resz(1+z)wtuzY(u,z)v,uv=Resz(1+z)wtuz2Y(u,z)v.u*v=\operatorname{Res}_z\frac{(1+z)^{\operatorname{wt}u}}{z}Y(u,z)v, \qquad u\circ v=\operatorname{Res}_z\frac{(1+z)^{\operatorname{wt}u}}{z^2}Y(u,z)v.

Let O(V)O(V) be the span of the circle products and A(V)=V/O(V)A(V)=V/O(V). Zhu proves that * descends to an associative product, [1][\mathbf1] is its unit, and [ω][\omega] is central Zhu 1996, Theorem 2.1.1, pp. 247–251. If M=n0M(n)M=\bigoplus_{n\geq0}M(n) is admissible, its top level M(0)M(0) is an A(V)A(V)-module under the zero mode

o(u)=uwtu1.o(u)=u_{\operatorname{wt}u-1}.

Conversely, induction from an A(V)A(V)-module produces an admissible module after quotienting the maximal submodule that misses its top level. Consequently, irreducible admissible VV-modules correspond to irreducible A(V)A(V)-modules Zhu 1996, Theorems 2.2.1–2.2.2, pp. 254–257. This is a classification of simple top-level data, not a claim that every module is semisimple. Rationality makes A(V)A(V) finite-dimensional and semisimple; C2C_2-cofiniteness alone gives finite-dimensionality but can leave a radical.

For an ordinary module MM, the character is

χM(τ)=trMqL0c/24,q=e2πiτ,Imτ>0.\chi_M(\tau)=\operatorname{tr}_M q^{L_0-c/24}, \qquad q=e^{2\pi i\tau},\quad \operatorname{Im}\tau>0.

Zhu’s modular-invariance theorem assumes rationality and his finiteness condition, now recognized in the relevant setting as C2C_2-cofiniteness. It proves convergence and finite-dimensional SL2(Z)SL_2(\mathbb Z) covariance of trace functions; the character case is Zhu 1996, Theorem 5.3.3, pp. 299–300. The proof changes coordinates from the plane to the torus, derives modular differential equations, and uses semisimplicity of A(V)A(V) to identify leading coefficients as traces on simple top levels.

The first application connects to Chiral Blocks, Sewing, and Modular Invariance. For V=L(12,0)V=L(\tfrac12,0), the Zhu algebra is generated by x=[ω]x=[\omega]. The vacuum singular vector imposes

A(V)C[x]x(x12)(x116).A(V)\cong \frac{\mathbb C[x]}{x(x-\tfrac12)(x-\tfrac1{16})}.

The three roots are exactly the lowest conformal weights of 1\mathbf1, ε\varepsilon, and σ\sigma. They are distinct, so this finite algebra is semisimple. Evaluating xx on a top level independently recovers o(ω)=L0o(\omega)=L_0, confirming the root interpretation.

In the order (1,ε,σ)(\mathbf1,\varepsilon,\sigma), the character vector transforms by

χ(1/τ)=Sχ(τ),S=12(112112220),\boldsymbol\chi(-1/\tau)=S\boldsymbol\chi(\tau), \quad S=\frac12 \begin{pmatrix} 1&1&\sqrt2\\ 1&1&-\sqrt2\\ \sqrt2&-\sqrt2&0 \end{pmatrix},

and

χ(τ+1)=Tχ(τ),Taa=e2πi(hac/24).\boldsymbol\chi(\tau+1)=T\boldsymbol\chi(\tau), \qquad T_{aa}=e^{2\pi i(h_a-c/24)}.

Two checks separate normalization from theorem use. Direct multiplication gives S2=IS^2=I because all three Ising sectors are self-contragredient. Also T1=e2πi/48T_{\mathbf1}=e^{-2\pi i/48} follows from c=1/2c=1/2 and h1=0h_{\mathbf1}=0. These finite matrices describe chiral characters; constructing a full torus partition function additionally requires left–right pairing and sewing consistency.

The Zhu algebra and the C2C_2 quotient are finite shadows with different jobs. A(V)A(V) acts on module top levels and can distinguish their lowest conformal weights; V/C2(V)V/C_2(V) is a commutative Poisson algebra controlling spanning sets and singular support. There is a natural flow of finiteness results between them under the theorem’s grading hypotheses, but there is no canonical isomorphism. For the Ising VOA both happen to have dimension three, yet their products differ: the Zhu polynomial has three distinct roots, whereas the C2C_2 polynomial is nilpotent. Confusing them would falsely infer semisimplicity from the C2C_2 quotient.

The modular theorem likewise concerns trace functions with insertions, not only the three vacuum characters. The zero-point functions form the easiest invariant subspace, but the coordinate-change argument and differential equations are proved for the larger family. This is why convergence and C2C_2 control are theorem hypotheses rather than observations made after writing a formal qq series.

Apply the same recipe to a C2C_2-cofinite but nonrational triplet VOA. Generalized modules have a nilpotent part of L0L_0, and ordinary irreducible characters need not span an SL2(Z)SL_2(\mathbb Z)-invariant space. Miyamoto’s theorem replaces missing ordinary traces by pseudotraces on suitable generalized modules Miyamoto 2004, Theorem 5.5, pp. 80–84. Thus the adversarial inference “finite Zhu algebra implies a finite modular representation on simple characters” fails. What survives is a finite-dimensional generalized trace space under the theorem’s C2C_2 and grading hypotheses; semisimplicity and a basis of ordinary characters do not.

  1. Construct the primitive idempotent of the Ising Zhu algebra projecting to the h=1/16h=1/16 top level.
Solution Lagrange interpolation gives $$e_\sigma(x)=\frac{x(x-1/2)}{(1/16)(1/16-1/2)}=\frac{128}{7}x(1/2-x).$$ It equals one at $x=1/16$ and zero at $x=0,1/2$, so $e_\sigma^2=e_\sigma$ in the quotient.
  1. Why does [ω][\omega] act centrally on every top level?
Solution Zhu's theorem makes $[\omega]$ central in $A(V)$, and its action is $o(\omega)=L_0$. On an irreducible top level, Schur's lemma therefore makes it scalar, equal to the lowest conformal weight.
  • Miyamoto, Masahiko. “Modular Invariance of Vertex Operator Algebras Satisfying C2C_2-Cofiniteness.” Duke Mathematical Journal 122 (2004), 51–91. DOI.
  • Zhu, Yongchang. “Modular Invariance of Characters of Vertex Operator Algebras.” Journal of the American Mathematical Society 9 (1996), 237–302. DOI.