Virasoro Minimal Models

• Virasoro minimal models are rational two-dimensional conformal field theories (RCFTs) characterized by central charge c_{p, q}=1-(6(p-q)^2/(pq)) and governed by two coprime integers p and q, defining highest-weight representations and modular-invariant properties, with unitary series examples of Ising and Tricritical Ising models.
• they are applied in various contexts, such as rational vertex operator algebras, vertex-operator orbifolds, conformal-block recursion, and three-dimensional realizations.
• significant examples include the Yang-Lee theory (M(2, 5)) with a non-unitary central charge and consistent behavior in gravitational and topological realizations.

Virasoro minimal models are rational two-dimensional conformal field theories whose chiral symmetry is generated by the Virasoro algebra and whose representation theory is controlled by two coprime integers. For coprime $p,q\geq 2$, the model $M(p,q)$ has central charge

$c_{p,q}=1-\frac{6(p-q)^2}{pq},$

finitely many irreducible highest-weight representations, explicitly computable characters, fusion rules, and modular-invariant torus partition functions. The unitary series is characterized by $|p-q|=1$, conventionally $M(p,p+1)$; generic coprime pairs give non-unitary models, including the Yang–Lee theory $M(2,5)$ with $c=-22/5$. Virasoro minimal models occur in coset constructions, rational vertex operator algebras, lattice scaling limits, conformal-block recursion, and topological and three-dimensional realizations (Beltaos et al., 2010, Koshida, 2018, Gang et al., 2024).

1. Definition and parametrization

The Virasoro algebra has generators $L_n$ and central element $C$ satisfying

$$[L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}m(m^2-1)\delta_{m+n,0}.$$

A highest-weight representation is generated by a vector $M(p,q)$0 obeying

$M(p,q)$1

A minimal model is specified by coprime integers $M(p,q)$2. Its central charge is

$M(p,q)$3

The same model is obtained upon exchanging $M(p,q)$4 and $M(p,q)$5. The model is unitary precisely when

$M(p,q)$6

The standard unitary parametrization is

$M(p,q)$7

The unitary sequence contains the Ising model $M(p,q)$8 and the tricritical Ising model $M(p,q)$9. For non-unitary models the primary of smallest conformal weight need not be the vacuum. The effective central charge,

$c_{p,q}=1-\frac{6(p-q)^2}{pq},$0

is consequently often more useful in finite-size and thermodynamic analyses. For example, $c_{p,q}=1-\frac{6(p-q)^2}{pq},$1 has $c_{p,q}=1-\frac{6(p-q)^2}{pq},$2 and $c_{p,q}=1-\frac{6(p-q)^2}{pq},$3, while $c_{p,q}=1-\frac{6(p-q)^2}{pq},$4 has $c_{p,q}=1-\frac{6(p-q)^2}{pq},$5 and $c_{p,q}=1-\frac{6(p-q)^2}{pq},$6 (Tanaka et al., 2024, Katsevich et al., 21 Oct 2025).

The irreducible Virasoro representations are labelled by Kac-table pairs

$c_{p,q}=1-\frac{6(p-q)^2}{pq},$7

subject to

$c_{p,q}=1-\frac{6(p-q)^2}{pq},$8

The number of independent primary representations is

$c_{p,q}=1-\frac{6(p-q)^2}{pq},$9

Their conformal weights are

$|p-q|=1$0

Equivalent conventions interchange $|p-q|=1$1 and $|p-q|=1$2 and correspondingly relabel the Kac indices. The vacuum is $|p-q|=1$3, although in a non-unitary model it need not have the smallest conformal weight.

The Verma module associated with a Kac weight is reducible. The two standard singular-vector levels of the irreducible minimal-model representation labelled by $|p-q|=1$4 are

$|p-q|=1$5

The irreducible module is obtained by quotienting by the submodules generated by these null vectors. This null-state structure is responsible for the finite character formula, degenerate BPZ equations, and truncated fusion rules.

2. Characters, modular data, and fusion

A standard expression for a Virasoro minimal-model character is

$|p-q|=1$6

where one convention for the theta function is

$|p-q|=1$7

The field-identification quotient is understood. Equivalent theta-function normalizations and Kac-table representatives give equivalent formulas.

The torus partition function is a sesquilinear combination

$|p-q|=1$8

where

$|p-q|=1$9

and modular invariance requires

$M(p,p+1)$0

The characters transform under

$M(p,p+1)$1

with $M(p,p+1)$2 diagonal:

$M(p,p+1)$3

Up to convention-dependent signs and normalization, the Virasoro modular $M(p,p+1)$4-matrix is

$M(p,p+1)$5

The field identification is implemented in this expression by passing to identification classes. The modular data determine the fusion coefficients through the Verlinde formula,

$M(p,p+1)$6

The fusion rules have truncated $M(p,p+1)$7 form in both Kac-table directions. For admissible triples, the fusion multiplicity is $M(p,p+1)$8; otherwise it is $M(p,p+1)$9. In a common notation, the allowed labels in the first direction run from $M(2,5)$0 to the appropriate truncated upper bound in steps of two, and analogously for the second direction. The fusion rules can therefore be described as two coupled $M(2,5)$1-type truncations subject to the Kac identification (Krauel et al., 2016, Tanaka et al., 2024).

3. Cosets, affine correspondences, and modular invariants

The unitary Virasoro series has the coset realization

$M(2,5)$2

with

$M(2,5)$3

More generally, the coset construction at admissible level realizes a minimal model with labels $M(2,5)$4 when

$M(2,5)$5

At the vertex-algebra level, the Virasoro algebra is the commutant of the diagonal affine algebra:

$M(2,5)$6

The coset stress tensor is the difference of Sugawara stress tensors,

$M(2,5)$7

so that

$M(2,5)$8

The coset selection rule is a parity constraint relating the affine labels to the Virasoro Kac labels (Koshida, 2018).

Virasoro minimal models are the $M(2,5)$9 member of the $c=-22/5$0 minimal-model family. The $c=-22/5$1 algebra is generated by fields of spins

$c=-22/5$2

so $c=-22/5$3 is precisely the Virasoro algebra. For $c=-22/5$4, the relevant affine theory is

$c=-22/5$5

A theorem relating $c=-22/5$6 modular invariants to affine product invariants gives, in the Virasoro case, a correspondence between Virasoro modular invariants and suitable pairs of affine $c=-22/5$7 modular invariants satisfying a vacuum-entry condition. Because $c=-22/5$8 and $c=-22/5$9 are coprime, the surviving product invariants are essentially factorized, with the Virasoro field-identification restriction (Beltaos et al., 2010).

The affine $L_n$0 modular invariants are classified by the ADE list. At height $L_n$1, the diagonal invariant is

$L_n$2

The $L_n$3-type invariants are simple-current extensions and require the relevant divisibility conditions, including $L_n$4 in the standard extension formulation. Exceptional invariants occur at heights

$L_n$5

corresponding respectively to $L_n$6, $L_n$7, and $L_n$8. Their Coxeter numbers satisfy

$L_n$9

The Cappelli–Itzykson–Zuber classification expresses Virasoro modular invariants through pairs of ADE graphs with Coxeter numbers $C$0 and $C$1:

$C$2

The diagonal invariant corresponds to an $C$3-type pair, $C$4-type invariants to simple-current extensions, and exceptional Virasoro invariants to the affine $C$5, $C$6, and $C$7 invariants. The coprimality condition strongly restricts which graph pairs can occur. The affine correspondence provides a shorter route to the classification than constructing directly the full commutant of the Virasoro modular matrices (Beltaos et al., 2010).

4. Vertex-operator algebras and permutation orbifolds

The vacuum sector of a Virasoro minimal model is represented by the simple Virasoro vertex operator algebra

$C$8

It is obtained from the universal vacuum Virasoro module by quotienting by its maximal proper submodule. The vacuum vector is $C$9, the Virasoro vector is $$[L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}m(m^2-1)\delta_{m+n,0}.$$0, and

$$[L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}m(m^2-1)\delta_{m+n,0}.$$1

For coprime $$[L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}m(m^2-1)\delta_{m+n,0}.$$2, $$[L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}m(m^2-1)\delta_{m+n,0}.$$3 is $$[L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}m(m^2-1)\delta_{m+n,0}.$$4-cofinite and is strongly generated by $$[L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}m(m^2-1)\delta_{m+n,0}.$$5.

For a VOA $$[L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}m(m^2-1)\delta_{m+n,0}.$$6,

$$[L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}m(m^2-1)\delta_{m+n,0}.$$7

The VOA is $$[L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}m(m^2-1)\delta_{m+n,0}.$$8-cofinite if

$$[L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}m(m^2-1)\delta_{m+n,0}.$$9

The quotient carries a commutative associative Poisson-algebra structure. $M(p,q)$00-cofiniteness is related to finite-dimensional Zhu-type quotients, modular-invariance results for trace functions, fusion products, and finiteness properties of module categories, but it does not by itself imply rationality or regularity.

Given a VOA $M(p,q)$01, the transposition

$M(p,q)$02

defines the two-cycle permutation orbifold

$M(p,q)$03

Its fixed-point space is spanned by

$M(p,q)$04

and it is convenient to define

$M(p,q)$05

The orbifold is strongly generated by the vectors $M(p,q)$06.

The general criterion for $M(p,q)$07-cofiniteness reduces the problem to the $M(p,q)$08-subspace. If $M(p,q)$09 is $M(p,q)$10-cofinite, of CFT type, and strongly generated by a finite set $M(p,q)$11, then $M(p,q)$12 is $M(p,q)$13-cofinite if and only if

$M(p,q)$14

is finite-dimensional in $M(p,q)$15.

For the Virasoro VOA, $M(p,q)$16. Virasoro-mode identities imply that

$M(p,q)$17

and hence the relevant $M(p,q)$18-subspace is finite-dimensional. Consequently,

$M(p,q)$19

is $M(p,q)$20-cofinite for every coprime $M(p,q)$21. This theorem concerns only $M(p,q)$22-cofiniteness; it does not establish rationality, regularity, modularity of all orbifold trace functions, or a classification of untwisted and twisted irreducible modules (Abe, 2010).

5. Conformal blocks, recursion, and modular representations

Virasoro conformal blocks are determined by the Virasoro Ward identities and the inverse Shapovalov form. A descendant is labelled by a partition

$M(p,q)$23

and the Gram matrix at level $M(p,q)$24 is

$M(p,q)$25

The descendant coefficients satisfy

$M(p,q)$26

At generic central charge and conformal weight, the Gram matrix is nonsingular. At degenerate weights it has a kernel generated by null descendants. The ambiguity in $M(p,q)$27 does not affect the conformal block because null vectors have zero pairing with the Ward-identity vector.

Degenerate Virasoro fields satisfy BPZ differential equations. If the lowest null vector occurs at level $M(p,q)$28, the corresponding four-point block satisfies an order-$M(p,q)$29 Fuchsian differential equation. For example, the Ising spin field satisfies a second-order equation, while the $M(p,q)$30 field in the tetracritical Ising model satisfies a third-order equation. For higher-level null vectors, accessory parameters can be determined directly from the Shapovalov-computed block series rather than by explicitly deriving the null vector. The number of accessory parameters for an order-$M(p,q)$31 equation is

$M(p,q)$32

The solutions around $M(p,q)$33 and $M(p,q)$34 form Frobenius bases. The braiding or $M(p,q)$35-matrix is the connection matrix between these bases. In rational theories with null-state differential equations, the physical conformal blocks may form only a subspace of the full solution space; the remaining solutions are spurious. The physical braiding matrix is obtained by restricting the connection matrix to the physical block subspace (Govindarajan et al., 17 Sep 2026).

Crossing symmetry relates the braiding matrix to products of three-point structure constants. If $M(p,q)$36 and $M(p,q)$37 denote the products of structure constants in the $M(p,q)$38- and $M(p,q)$39-channels, respectively, then

$M(p,q)$40

After diagonal rescaling,

$M(p,q)$41

the matrix is orthogonal or unitary. Numerical connection matrices agree with Dotsenko–Fateev Coulomb-gas results. The squared entries of the unitary matrix are conjectured to lie in cyclotomic extensions of the rational numbers.

Intertwining operators provide a different modular representation. For an irreducible module $M(p,q)$42, one-point functions of intertwining operators span spaces of vector-valued modular forms. If $M(p,q)$43 is a highest-weight vector, the full one-point-function space is a cyclic module over the ring of modular differential operators:

$M(p,q)$44

For selected modules $M(p,q)$45, the representation dimension is

$M(p,q)$46

These modular representations can have congruence or noncongruence kernels. Low-dimensional classifications exhibit both finite-image congruence representations and infinite-image noncongruence representations. Arithmetic criteria based on the prime-power factors of the representation level imply that, for many interior Kac labels in prime-power minimal models, the associated modular representation is noncongruence (Krauel et al., 2016).

At rational central charge, Zamolodchikov recursion relations require special care. Distinct Kac labels can have coincident conformal dimensions, producing resonant denominators in individual recursion terms. These poles are artifacts of the recursive decomposition rather than singularities of the final conformal block or character. Correlated regulators for the central charge, internal dimension, and external dimensions cause the resonant contributions to cancel. For torus characters, the resulting cancellation reproduces the inclusion–exclusion structure associated with the two null submodules (Javerzat et al., 2018).

6. Logarithmic extensions, limits, and deformations

Ordinary minimal models are rational, semisimple, and contain finitely many irreducible modules. Logarithmic minimal models retain the same central charges and Kac weights but allow the full infinite Kac table. Their representation categories contain reducible but indecomposable Virasoro modules.

Virasoro Kac modules $M(p,q)$47 are finitely generated submodules of Feigin–Fuchs modules. Their characters are

$M(p,q)$48

The embedding structure may be point-like, chain-like, island-like, or braid-like. A character determines the composition factors but not their extension structure or arrow orientation. Lattice boundary-seam algebras provide finite-dimensional approximations whose scaling limits are conjectured to produce these Kac modules. The basic Grothendieck fusion rule has $M(p,q)$49 form, but the actual tensor product can be a non-split logarithmic extension.

For the braided tensor category $M(p,q)$50 of finite-length modules, the modules satisfying

$M(p,q)$51

are rigid and self-dual. The category as a whole is not rigid. The fundamental fusion sequences are

$M(p,q)$52

which split if and only if $M(p,q)$53, and

$M(p,q)$54

which split if and only if $M(p,q)$55. In particular,

$M(p,q)$56

At divisibility boundaries, the products are indecomposable and $M(p,q)$57 acts non-semisimply (Morin-Duchesne et al., 2015, McRae et al., 2023).

The $M(p,q)$58 limit of the unitary series illustrates another enlargement of the minimal-model framework. In the $M(p,q)$59 limit of $M(p,q)$60 minimal models, with

$M(p,q)$61

the central charge approaches $M(p,q)$62. Keeping coset labels fixed produces the singlet sector of $M(p,q)$63 free bosons. This sector is not modular invariant by itself; it is interpreted as the untwisted sector of a continuous orbifold by

$M(p,q)$64

For $M(p,q)$65, the orbifold group is $M(p,q)$66, and its twisted sectors supply the light states omitted in the fixed-label limit. The resulting twisted-sector theory agrees with the Runkel–Watts $M(p,q)$67 theory, whose primaries are labelled by $M(p,q)$68 and have

$M(p,q)$69

The agreement extends to fusion, boundary spectra, and crossing-symmetry data (Gaberdiel et al., 2011).

Minimal models also serve as ultraviolet fixed points for deformations that leave the rational Virasoro framework. Coupling a relevant primary to a generalized free field produces long-range minimal models. The perturbation

$M(p,q)$70

has dimension $M(p,q)$71 in two dimensions and breaks local Virasoro symmetry while preserving global conformal invariance. For the $M(p,q)$72 family, the large-$M(p,q)$73 expansion is well behaved; for the $M(p,q)$74 family, the perturbative window shrinks as $M(p,q)$75 increases. These theories retain Kac-table data as ultraviolet organizing information but acquire new operators, anomalous dimensions, and generally nonlocal dynamics (Behan et al., 30 Sep 2025).

RG flows between minimal models can likewise be organized by non-invertible defects. A proposed infinite family is

$M(p,q)$76

induced by $M(p,q)$77. The perturbation preserves the $M(p,q)$78 fusion subcategory generated by the defect lines $M(p,q)$79. The preserved defects have matching quantum dimensions across the proposed flow, and their defect spin contents satisfy the expected ultraviolet-to-infrared containment property. The effective central charge decreases,

$M(p,q)$80

although the complete classification of such flows remains conjectural (Tanaka et al., 2024).

Virasoro minimal models also admit proposed bulk and gravitational realizations. In three dimensions, a model $M(p,q)$81 is associated through 3D–3D correspondence with a Seifert manifold

$M(p,q)$82

For $M(p,q)$83, the bulk theory is expected to flow to a unitary topological field theory; for $M(p,q)$84, it is expected to flow to a three-dimensional $M(p,q)$85 rank-0 SCFT whose topological twist supports the non-unitary chiral minimal model. Bethe vacua, Chern–Simons invariants, Reidemeister torsions, and supersymmetric partition functions reproduce the minimal-model modular data, although the required boundary condition has not been constructed in general (Gang et al., 2024).

Finally, Virasoro minimal-model data enter matrix-model and minimal-string constructions. Conventional $M(p,q)$86 minimal strings couple the matter model to Liouville theory and ghosts. More recent Virasoro minimal-string constructions instead involve continuous Liouville-type sectors, spectral curves, topological recursion, ZZ and FZZT branes, and resurgent completions. These constructions should not be identified with the rational minimal models themselves: they use the minimal-model terminology and degenerate-representation data but study continuous gravitational or matrix-model theories rather than finite rational Virasoro representation categories (Castro, 2024, Schwick, 25 Feb 2026).