Minimal Models and Fusion Rules

Virasoro minimal models turn null-vector decoupling into a finite, exactly solvable bootstrap. Their central charges and primary weights lie on a Kac table; null states impose BPZ differential equations; compatible local solutions determine fusion and correlators. The unitary Ising model $M(3,4)$ supplies a complete benchmark, including a normalized four-spin correlator rather than only a list of dimensions.

Required background. Highest-weight modules, null states, and the Kac determinant construct the irreducible Virasoro quotients. Crossing and positivity explain channel equivalence and the extra sign constraints present in a unitary theory.

Helpful background. Characters and multiplet counting connect null quotients to torus traces.

Minimal spectra and Kac identification

For coprime integers $2\le p<p'$, the Virasoro minimal model convention used here is

$$c_{p,p'}=1-\frac{6(p'-p)^2}{pp'}, \qquad h_{r,s}=\frac{(p'r-ps)^2-(p'-p)^2}{4pp'}.$$

Labels lie in $1\le r\le p-1$, $1\le s\le p'-1$, with the field identification

$$(r,s)\sim(p-r,p'-s).$$

The corresponding Verma module has singular vectors at levels $rs$ and $(p-r)(p'-s)$. Quotienting both submodules yields finitely many irreducible chiral modules. The full CFT still requires a compatible left–right pairing; the Kac table alone specifies only chiral representation data.

The unitary series has $p'=p+1$. Other coprime pairs are generally nonunitary: the algebraic formulas remain valid, but the Hermitian form has negative-norm states. The finite spectrum, null decoupling, and fusion construction were established in Belavin, Polyakov, and Zamolodchikov 1984, §§2–5, pp. 344–369.

For Ising, $p=3$, $p'=4$, and $c=1/2$. The complete Kac data are

Full field	Identified chiral labels	$h=\bar h$	$\Delta$	Generating singular-vector levels
$\mathbf1$	$(1,1)\sim(2,3)$	$0$	$0$	$1,6$
$\sigma$	$(1,2)\sim(2,2)$	$1/16$	$1/8$	$2,4$
$\epsilon$	$(1,3)\sim(2,1)$	$1/2$	$1$	$3,2$

The diagonal bulk theory pairs each holomorphic module with its antiholomorphic copy. With unit-normalized two-point functions, choose real fields and $C_{\sigma\sigma\epsilon}=1/2$; field rephasings can change signs of odd OPE coefficients, so the normalization and reality convention are part of the data. Relative to the course thermal convention, this page uses $\epsilon=-\varepsilon_{\mathrm{course}}$. Changing the sign of $\sigma$ alone cannot change $C_{\sigma\sigma\epsilon}$.

From a null state to the BPZ equation

The Ising spin state obeys

$$\left(L_{-2}-\frac43L_{-1}^2\right)|\sigma\rangle=0.$$

Insert this relation at $z$ in a correlator with primary fields $\phi_i(z_i)$. The stress-tensor Ward identity converts $L_{-2}\sigma$ into contour residues, while $L_{-1}\sigma=\partial\sigma$. Null-state decoupling therefore gives

$$\left[ \sum_i\left( \frac{h_i}{(z-z_i)^2} +\frac{1}{z-z_i}\partial_{z_i} \right) -\frac43\partial_z^2 \right] \left\langle\sigma(z)\prod_i\phi_i(z_i)\right\rangle=0.$$

This is a chiral equation on a configuration-space patch: collision points are removed, an ordering of insertions is fixed, and fractional powers require a branch. It does not by itself choose a single-valued full correlator.

Use the unit-normalized plane identity vacuum, and define $\sigma(\infty)=\lim_{\lvert R\rvert\to\infty}\lvert R\rvert^{1/4}\sigma(R,\bar R)$ inside full correlators. This is the full-field version of the normalized chiral insertion at infinity. For four spins, place the insertions at $(\infty,1,z,0)$ and initially take $0<z<1$. On principal square-root branches, two holomorphic solutions with the normalization used in the full pairing below are

$$\mathcal F_{\mathbf1}(z) =\frac{\sqrt{1+\sqrt{1-z}}} {\sqrt2\,[z(1-z)]^{1/8}},$$ $$\mathcal F_{\epsilon}(z) =\frac{\sqrt{1-\sqrt{1-z}}} {\sqrt2\,[z(1-z)]^{1/8}}.$$

Near $z=0$,

$$\mathcal F_{\mathbf1}(z)=z^{-1/8}(1+O(z)), \qquad \mathcal F_{\epsilon}(z)=\frac12z^{3/8}(1+O(z)).$$

The exponents identify the intermediate weights $0$ and $1/2$. In the full pairing below, the energy-channel leading term has coefficient $1/4$, fixing $\lvert C_{\sigma\sigma\epsilon}\rvert=1/2$ with unit two-point normalization. The positive sign is the declared energy-field convention; the four-spin correlator is unchanged under $\epsilon\mapsto-\epsilon$. The energy-channel block already includes this magnitude and is not normalized to a unit leading term. This magnitude-versus-sign distinction is explicit in Di Francesco, Mathieu, and Sénéchal 1997, §12.3.3, p. 451, Eqs. (12.63)–(12.64). The diagonal full correlator is

$$\left\langle \sigma(\infty)\sigma(1)\sigma(z,\bar z)\sigma(0) \right\rangle =|\mathcal F_{\mathbf1}(z)|^2 +|\mathcal F_{\epsilon}(z)|^2.$$

Analytic continuation of each chiral block has monodromy, but the displayed pairing is single-valued on the punctured sphere and invariant under the crossing transformations generated by $z\mapsto1-z$ and $z\mapsto1/z$ with their matching prefactors. Exact Ising blocks and normalization checks are worked out in Di Francesco, Mathieu, and Sénéchal 1997, §§8.2–8.4 and 12.2–12.4.

Fusion from allowed local exponents

Null equations restrict which intermediate highest weights can occur. In $M(p,p')$, the multiplicity-free Virasoro fusion rule is

$$(r_1,s_1)\times(r_2,s_2) =\sum_{r}\sum_s(r,s),$$

where $r$ advances by two from $|r_1-r_2|+1$ to

$$\min(r_1+r_2-1,\,2p-r_1-r_2-1),$$

and $s$ advances by two from $|s_1-s_2|+1$ to

$$\min(s_1+s_2-1,\,2p'-s_1-s_2-1),$$

followed by the Kac identification. Applied to Ising,

$$\sigma\times\sigma=\mathbf1+\epsilon, \qquad \sigma\times\epsilon=\sigma, \qquad \epsilon\times\epsilon=\mathbf1.$$

The fusion coefficient is an allowed-channel multiplicity, not an OPE coefficient. For example, $N_{\sigma\sigma}^{\epsilon}=1$ while $C_{\sigma\sigma\epsilon}=1/2$. Virasoro fusion gives $C_{\epsilon\epsilon\epsilon}=0$ because $\epsilon$ is absent from $\epsilon\times\epsilon$, even though three-point kinematics alone would permit a scalar coefficient. The distinction between fusion multiplicities and normalized structure constants is emphasized in Ginsparg 1990, §§4–5.

Exact constructions need more than a central charge

The figure places the minimal model beside four other compact constructions. Inspect the arrows into the center: each route must supply a chiral algebra, a spectrum, a rule selecting or projecting states, and a modular completion.

Schematic comparison of rational and compact constructions. Matching central charges is never sufficient: null quotients, momentum–winding lattices, affine integrability, branching, projections, twisted sectors, and left–right modular completion determine different theories.

An equivalent construction checklist is:

Construction	Chiral algebra	Spectrum and selection	Modular-completion obligation	Normalization anchor
Virasoro minimal model	Irreducible Virasoro modules	Kac table modulo $(r,s)\sim(p-r,p'-s)$ and null submodules	Choose a nonnegative integral left–right pairing and check sewing	Unit two-point functions and a declared sign for $C_{\sigma\sigma\epsilon}$
Compact boson	$U(1)$ current algebra, enlarged at special radii	Even momentum–winding lattice with locality cocycles	Sum the full lattice and all sectors required by any quotient	$X_LX_L\sim-\log z$ and a declared radius convention
WZW model	Affine $\widehat{\mathfrak g}_k$	Integrable highest weights at level $k$	Pair affine characters and satisfy fusion/factorization	Long-root length and current-OPE metric
Coset	Commutant of $\widehat{\mathfrak h}\subset\widehat{\mathfrak g}$	Branching selection, field identifications, fixed-point resolution	Include every branching sector in a modular-closed pairing	Embedding index and inherited levels
Orbifold	Invariant algebra plus twisted modules	Projection in every twisted sector	Sum over commuting twists, including fixed-point multiplicities	Group action, cocycle phases, and $1/\lvert G\rvert$ factor

Comparing low-dimensional model evidence

The following shared comparison keeps spectrum type, diagonalizability, positivity, and the meaning of correlators in separate fields. Rows outside the rational unitary core are included only to mark where the assumptions change; their constructions belong to the next chapter.

Benchmark	Spectrum and measure	Dilatation action	Inner product	Correlator status	Fusion or OPE	Modular behavior	Exact evidence and normalization
Ising $M(3,4)$	Three discrete Virasoro sectors; counting measure	Diagonal	Unitary	Ordinary single-valued bulk correlators after chiral pairing	Finite semisimple fusion	Three characters transform by finite $S,T$; diagonal $Z$	Null equations, exact four-spin blocks, characters; $\langle\phi\phi\rangle=1$
Yang–Lee $M(2,5)$	Two discrete Virasoro sectors; counting measure	Diagonal	Nonunitary, indefinite	Ordinary bulk correlators after pairing	Finite semisimple fusion	Two-character finite modular representation	BPZ equations and exact characters; negative primary weight records nonunitarity
Compact boson	Countable discrete momentum–winding lattice	Diagonal	Unitary at positive radius	Ordinary after zero-mode neutrality and cocycle completion	Lattice addition with locality phases	Theta-function lattice sum is modular invariant	Gaussian correlators; $X_LX_L\sim-\log z$ fixes dimensions
Noncompact free boson	Continuous momentum with Lebesgue measure	Diagonal	Delta-normalized positive generalized states	Distributional in momentum labels; ordinary for suitable wave packets	Continuous charge addition	Integral kernel and target-volume divergence	Gaussian path integral; zero-mode measure must be declared
Logarithmic $c=-2$ example	Discrete generalized modules	Jordan blocks occur	Nonunitary	Ordinary position-space logarithms, with generalized-state pairings	Nonsemisimple fusion	Characters alone may not close; generalized traces can be needed	Exact hypergeometric degeneracy and logarithmic two-point forms
Liouville theory	Continuous spectrum with a nontrivial spectral measure	Diagonal on the principal spectrum	Unitary in its standard $c>25$ regime	Distributional in continuum labels; smeared observables are ordinary	Continuous fusion integral	Modular transformations act by an integral kernel	Exact structure constants and conformal blocks require fixed delta normalization

Reproducible checks

A reproducible calculation should test the Kac identifications, null levels, normalized BPZ blocks on a fixed complex domain, fusion closure, character coefficients, and exact $S,T$ identities with declared tolerances.

Common pitfalls

Confusing a chiral solution with a local correlator. A BPZ solution is generally multivalued. A full correlator requires a left–right pairing whose monodromies cancel and whose channel limits have admissible OPE coefficients.

Reading OPE coefficients from fusion multiplicities. Fusion says which representations may appear and with what multiplicity. Three-point normalization supplies additional continuous or discrete data.

Calling every minimal model unitary. Only the adjacent series $p'=p+1$ is unitary. Finite Kac data and exact solvability do not imply positivity.

Exercises

Derive $\sigma\times\sigma$ from the general fusion bounds.

Solution

For $\sigma=(1,2)$ in $M(3,4)$, the $r$ sum begins and ends at $1$. The $s$ sum runs in steps of two from $|2-2|+1=1$ to

$$\min(2+2-1,\,2\cdot4-2-2-1)=3.$$

Thus $(1,2)\times(1,2)=(1,1)+(1,3)=\mathbf1+\epsilon$.

References

• Belavin, Alexander A., Alexander M. Polyakov, and Alexander B. Zamolodchikov. “Infinite Conformal Symmetry in Two-Dimensional Quantum Field Theory.” Nuclear Physics B 241, no. 2 (1984): 333–380. DOI.
• Di Francesco, Philippe, Pierre Mathieu, and David Sénéchal. Conformal Field Theory. Graduate Texts in Contemporary Physics. New York: Springer, 1997. DOI.
• Ginsparg, Paul. “Applied Conformal Field Theory.” In Fields, Strings and Critical Phenomena, Les Houches Session XLIX, edited by Édouard Brézin and Jean Zinn-Justin, 1–168. Amsterdam: North-Holland, 1990. arXiv.