The Coulomb gas, screening charges, and the conformal bootstrap

Intuition Beginner

A two-dimensional conformal field theory built from a single free wave field is one of the simplest exactly solvable theories. The wave field is a height that fluctuates over the plane, like the rippling surface of a pond. Out of this one field you can build a whole family of measuring operators by riding the field with different sensitivities.

The clever trick is to put a fixed "background charge" very far away. This far-off charge tilts every measurement by a known amount. Tuning how strongly it tilts is the same as tuning a single number, the central charge, that labels the theory. With the right tilt, this simple free-field model reproduces the critical points of famous lattice models, such as the Ising magnet at its phase transition.

To make all the bookkeeping balance, you sometimes need to add a few neutral "screening" pieces that carry away leftover charge. Counting how many you need, and where their contours can run, is what turns a correlation question into a definite integral you can actually evaluate.

Visual Beginner

Picture the plane with several measuring operators pinned at fixed points. Each carries a small charge label. A large opposite charge sits at the far edge to keep the books balanced. A few extra wandering markers, the screening pieces, slide along loops that thread between the pinned points.

The picture looks like charges in a flat landscape, which is why the method is named after a gas of charges. But these charges are labels on field operators, not real particles sitting in a real electric field. The loops are integration paths, and sliding a marker around a pinned point can multiply the answer by a phase. Keeping track of those phases is the whole game.

Worked example Beginner

Take the model whose central charge is one half. This is the value that describes the two-dimensional Ising magnet exactly at its critical temperature. The theory has three basic operators, with scaling weights equal to $0$, $1/16$, and $1/2$.

The weight $0$ operator is the identity, the "do nothing" measurement. The weight $1/2$ operator tracks the local magnetisation, the tendency of a spin to point up or down. The weight $1/16$ operator tracks the boundary between an up region and a down region, called the disorder or spin-flip operator.

Now ask a concrete question. If you compute how two magnetisation operators a distance $r$ apart correlate, the free-field rule says the answer falls off like $1$ divided by $r$ to the power $4$ times $1/2$, that is like $1$ over $r$ squared. The single number $1/2$ fixes the whole power law. The same machine, with a different tilt, fixes every other critical model in the same family.

Check your understanding Beginner

Formal definition Intermediate+

Work with a single chiral free boson $\phi (z)$ with the propagator normalisation $⟨\phi (z)\phi (w)⟩=-\log (z-w)$. The plain free boson has central charge $1$. The Coulomb gas modifies the stress tensor by a background charge ${\alpha }_0$:

$$T(z)=-\frac{1}{2}(\partial \phi {)}^2+i\sqrt{2}{\alpha }_0{\partial }^2\phi .$$

The added improvement term shifts the central charge to

$$c=1-24{\alpha }_0^2,$$

equivalently $c=1-6Q^2/(\text{unit})$ after setting the background charge $Q=2{\alpha }_0$ at infinity. The vertex operators are the normal-ordered exponentials

$$V_{\alpha }(z)=:e^{i\sqrt{2}\alpha \phi (z)}:.$$

Computing the OPE of $T$ with $V_{\alpha }$ shows that $V_{\alpha }$ is a Virasoro primary of conformal weight

$$h_{\alpha }=\alpha (\alpha -2{\alpha }_0)=\alpha (\alpha -Q).$$

This weight is invariant under the reflection $\alpha \mapsto Q-\alpha =2{\alpha }_0-\alpha$, so two distinct charges represent the same primary weight; this is the charge-conjugation symmetry of the free-field representation.

The two screening charges are the contour integrals of the weight-one vertex operators ${\alpha }_{\pm }$ solving $h_{{\alpha }_{\pm }}=1$:

$${\alpha }_{\pm }={\alpha }_0\pm \sqrt{{\alpha }_0^2+1},Q_{\pm }=\oint dz{V}_{{\alpha }_{\pm }}(z).$$

Because each $V_{{\alpha }_{\pm }}$ has weight one, $Q_{\pm }$ commutes with the Virasoro algebra and can be inserted into a correlator without changing its conformal covariance. The degenerate primaries of the Kac table sit at the lattice of charges

$${\alpha }_{r,s}=\frac{1}{2}(1-r){\alpha }_{+}+\frac{1}{2}(1-s){\alpha }_{-},r,s\ge 1.$$

Convention. This is the Dotsenko-Fateev normalisation, $c=1-24{\alpha }_0^2$; some references absorb factors into $\phi$ and write $c=1-24{\alpha }_0^2$ with a rescaled ${\alpha }_0$. The reflection-symmetric pairing $h_{\alpha }=h_{Q-\alpha }$ is convention-independent.

This construction must be distinguished from the electrostatic Coulomb gas of the Kosterlitz-Thouless transition 08.15.01: there the charges are integer vortex numbers of the XY model interacting through a genuine logarithmic potential at finite temperature, with a fugacity-driven unbinding transition. Here the charges are formal $U(1)$ momenta $\alpha$ of holomorphic vertex operators in a chiral CFT, the neutrality is the sphere sum rule, and there is no temperature.

Key theorem with proof Intermediate+

Theorem (charge neutrality and the screened correlator). On the sphere, a product of vertex operators ${\prod }_i{V}_{{\alpha }_i}(z_i)$ has a nonzero expectation value only if the total charge equals the background charge,

$$\sum _i{\alpha }_i=2{\alpha }_0=Q.$$

When the bare charges fail neutrality, inserting $n_{+}$ screening charges $Q_{+}$ and $n_{-}$ screening charges $Q_{-}$ restores it provided

$$\sum _i{\alpha }_i+n_{+}{\alpha }_{+}+n_{-}{\alpha }_{-}=2{\alpha }_0,$$

and the resulting correlator is the Dotsenko-Fateev integral

$$⟨\prod _i{V}_{{\alpha }_i}(z_i)⟩=\oint \prod _{a=1}^{n_{+}}d{u}_a\oint \prod _{b=1}^{n_{-}}d{v}_b⟨\prod _i{V}_{{\alpha }_i}(z_i)\prod _a{V}_{{\alpha }_{+}}(u_a)\prod _b{V}_{{\alpha }_{-}}(v_b){⟩}_0.$$

Proof. The background charge at infinity assigns the vacuum a charge $-2{\alpha }_0$ when read against the $\phi$ zero mode, so the path integral over the zero mode is a delta function setting the total charge of all insertions to $2{\alpha }_0$. With neutrality met, the Gaussian (Wick) contraction of the exponentials gives

$$⟨\prod _k{V}_{{\beta }_k}(w_k){⟩}_0=\prod _{k<l}(w_k-w_l{)}^{2{\beta }_k{\beta }_l},$$

a product of power laws in the separations, where the ${\beta }_k$ run over both the external charges ${\alpha }_i$ and the screening charges ${\alpha }_{\pm }$ at the contour points $u_a,v_b$. Since each $V_{{\alpha }_{\pm }}$ has weight one, integrating its position over a closed contour leaves the correlator a conformally covariant function of the external points $z_i$. Each independent way of distributing the screening contours produces one conformal block; the number of blocks matches the fusion multiplicity of the degenerate external fields. $\square$

Bridge. This screened integral builds toward the conformal bootstrap: the conformal blocks it produces are exactly the holomorphic objects whose crossing matrix the bootstrap constrains, and the neutrality sum rule appears again in 03.10.04 as the truncation of the Kac table that gives the finite minimal-model fusion rules. The foundational reason the free boson can encode an interacting minimal model is that the background charge makes the naive Gaussian theory non-unitary in a controlled way; this is exactly the Feigin-Fuks resolution of the Virasoro Verma module by free-field Fock spaces. The screening charges are dual to the singular vectors of 03.10.04: putting these together, a singular-vector decoupling on the algebraic side is the statement that a screening contour can be deformed off the correlator on the analytic side, and that duality generalises the Kac determinant from a vanishing condition into an explicit integral formula for every degenerate correlator.

Exercises Intermediate+

Lean formalization Intermediate+

lean_status: none — Mathlib can express Gaussian integrals and some special functions, but not vertex operators, screening contours, or conformal blocks.

-- Pseudocode only: no CFT / vertex-operator structure exists in Mathlib.
axiom FreeBoson : Type
axiom backgroundCharge : ℝ → FreeBoson → FreeBoson

-- Vertex operator weight h_α = α (α - 2 α₀)
def vertexWeight (α α₀ : ℝ) : ℝ := α * (α - 2 * α₀)

-- Screening condition: weight one
axiom screening_weight_one
    (α₀ : ℝ) :
    vertexWeight (α₀ + Real.sqrt (α₀^2 + 1)) α₀ = 1

A genuine formalization needs a vertex operator algebra with a deformed conformal vector, the Feigin-Fuks Fock-space resolution of Virasoro Verma modules, and the Riemann-Hilbert monodromy of the resulting Fuchsian system. None of these is presently in Mathlib.

Advanced results Master

The decoupling of a singular vector turns a correlator into a differential equation. Let ${\varphi }_{1,2}$ be the level-two degenerate field, with weight $h$ obeying the relation that produces the null state $(L_{-2}-\frac{3}{2(2h+1)}{L}_{-1}^2)|h⟩=0$. Inserting ${\varphi }_{1,2}(z)$ into a four-point function and demanding that the null combination annihilate the correlator yields a second-order ordinary differential equation in the cross-ratio. After fixing three of the four points to $0$, $1$, $\infty$, the equation is Fuchsian with three regular singular points, and its solutions are the Gauss hypergeometric functions ${}_2{F}_1$. These two independent solutions are the holomorphic conformal blocks of the four-point function.

The physical correlator is built by combining holomorphic and antiholomorphic blocks into a single-valued, crossing-symmetric sum:

$$G(z,\bar{z})=\sum _{p,\bar{p}}{X}_{p\bar{p}}{\mathcal{F}}_p(z)\overline{{\mathcal{F}}_{\bar{p}}(z)},$$

where ${\mathcal{F}}_p$ are the hypergeometric blocks and $X_{p\bar{p}}$ are constants fixed by single-valuedness around the singular points. The same four-point function admits two natural decompositions: an $s$-channel OPE as $z\to 0$ and a $t$-channel OPE as $z\to 1$. The hypergeometric connection formulae relate the two bases of blocks.

The conformal bootstrap is the requirement that these two decompositions agree:

$$\sum _p{C}_{12p}{C}_{34p}{\mathcal{F}}_p^{(s)}(z)\overline{{\mathcal{F}}_p^{(s)}(z)}=\sum _q{C}_{14q}{C}_{23q}{\mathcal{F}}_q^{(t)}(z)\overline{{\mathcal{F}}_q^{(t)}(z)}.$$

This associativity of the operator algebra is a system of quadratic equations for the OPE structure constants $C_{ijk}$. Dotsenko and Fateev solved it for the minimal models by computing both sides as Coulomb-gas (Selberg-type) integrals; consistency forces the external charges onto the Kac lattice and selects the truncated minimal-model fusion rules ${\varphi }_{r_1,s_1}\times {\varphi }_{r_2,s_2}=\sum {\varphi }_{r,s}$ over the allowed range.

Synthesis. The central insight is that one free field, deformed by a background charge, computes an entire interacting family: the Coulomb-gas integral is exactly the analytic shadow of the algebraic null-vector decoupling, and putting these together identifies the conformal blocks of 03.10.04 with explicit hypergeometric functions. This generalises the BPZ differential-equation method from a single fixed correlator to a closed bootstrap, where crossing symmetry is dual to the associativity of the OPE; the foundational reason the minimal models are exactly solvable is that the bootstrap equations, infinite in principle, collapse to a finite algebraic problem on the Kac table. The bridge runs in both directions: Coulomb-gas integrals supply the structure constants that the bootstrap demands, and the bootstrap in turn certifies that the free-field representation reproduces a unitary, crossing-symmetric theory.

Full proof set Master

Proposition (level-two null vector gives a hypergeometric four-point function). Let ${\varphi }_{1,2}$ be a primary of weight $h_{1,2}$ whose Verma module contains the singular vector

$$(L_{-2}-\frac{3}{2(2h_{1,2}+1)}{L}_{-1}^2)|h_{1,2}⟩=0.$$

Then the four-point function $⟨{\varphi }_{1,2}(z){\varphi }_2(0){\varphi }_3(1){\varphi }_4(\infty )⟩$ satisfies a second-order Fuchsian ODE whose solutions are Gauss hypergeometric functions.

Proof. The mode $L_{-1}$ acts on an insertion as the holomorphic derivative ${\partial }_z$, and $L_{-2}$ acts through the conformal Ward identity as a first-order differential operator in $z$ built from the weights and positions of the other three insertions:

$$L_{-2}{\varphi }_{1,2}(z)⟼\sum _{i=2,3,4}(\frac{h_i}{(z-z_i{)}^2}-\frac{1}{z-z_i}{\partial }_{z_i})⟨\cdots ⟩.$$

Substituting both replacements into the null condition gives a linear relation between ${\partial }_z^2⟨\cdots ⟩$ and $⟨\cdots ⟩$ together with its first $z$-derivative. Global conformal invariance fixes the dependence on $z_2,z_3,z_4$ up to a function of the single cross-ratio $x=\frac{(z-z_2)(z_3-z_4)}{(z-z_4)(z_3-z_2)}$. Setting $(z_2,z_3,z_4)=(0,1,\infty )$, the relation becomes a second-order ODE in $x$ of the form

$$x(1-x)F^{\prime \prime }(x)+(\gamma -(\alpha +\beta +1)x)F^{\prime }(x)-\alpha \beta F(x)=0,$$

the hypergeometric equation, with $\alpha ,\beta ,\gamma$ determined rationally by the four weights. Its regular singular points are $0,1,\infty$, and the local exponents at $0$ reproduce the two allowed fusion weights $h_p$ in the ${\varphi }_{1,2}\times {\varphi }_2$ OPE. The two Frobenius solutions are the conformal blocks. $\square$

Proposition (charge balance forces the Kac lattice). If ${\varphi }_{1,2}$ has a hypergeometric four-point function whose fusion channels are exactly two, then the external charges must lie on the lattice ${\alpha }_{r,s}=\frac{1}{2}(1-r){\alpha }_{+}+\frac{1}{2}(1-s){\alpha }_{-}$.

Proof. Charge neutrality on the sphere requires ${\sum }_i{\alpha }_i+n_{+}{\alpha }_{+}+n_{-}{\alpha }_{-}=2{\alpha }_0$ for nonnegative integers $n_{\pm }$. A two-channel fusion corresponds to exactly two admissible screening counts, which is possible only when each external charge differs from ${\alpha }_0$ by a half-integer combination of ${\alpha }_{+}$ and ${\alpha }_{-}$. Writing $\alpha ={\alpha }_0-\frac{1}{2}r{\alpha }_{+}-\frac{1}{2}s{\alpha }_{-}$ and using ${\alpha }_{+}+{\alpha }_{-}=2{\alpha }_0$ recovers ${\alpha }_{r,s}$, with $r,s$ positive integers labelling the Kac table. The two screening routings are the two fusion channels, matching the two Frobenius exponents of the hypergeometric equation. $\square$

Connections Master

• Minimal models, the Kac formula, and null vectors 03.10.04 — this unit is the analytic companion to the algebraic minimal-model construction. The degenerate primaries and singular vectors defined there are realised here as charges on the Kac lattice whose null-vector decoupling produces the hypergeometric blocks; the Coulomb-gas integrals compute the very structure constants whose existence the Kac determinant only certifies.

• CFT basics 03.10.02 — the vertex operators, stress tensor, OPE, and central charge used throughout are the general apparatus introduced there. The background-charge deformation specialises the free-boson stress tensor of that unit to the $c<1$ range.

• Gaussian free field / free boson 08.06.01 — the entire construction is a controlled deformation of the Gaussian free field. The Wick-contraction formula for vertex-operator correlators is the free-field two-point function exponentiated, and the screening contours integrate that Gaussian data.

• The Kosterlitz-Thouless transition / 2D Coulomb gas 08.15.01 — a deliberate contrast, not a dependency. That unit's Coulomb gas is a finite-temperature electrostatic gas of integer vortex charges with a real logarithmic potential and an unbinding transition; the present construction's charges are formal $U(1)$ momenta of holomorphic vertex operators with no temperature. The shared name reflects only the common Gaussian / logarithmic-correlator origin.

Historical & philosophical context Master

Feigin and Fuks gave the free-field resolution of Virasoro highest-weight modules in 1982, exhibiting the singular-vector structure of degenerate representations through Fock spaces of a single boson with a background charge ^{[Feigin-Fuks 1982]}. Belavin, Polyakov, and Zamolodchikov had, in the same period, identified the degenerate fields whose null vectors give differential equations for correlators ^{[Belavin-Polyakov-Zamolodchikov 1984]}.

Dotsenko and Fateev turned this into a computational engine in two papers of 1984 and 1985, representing minimal-model correlators as multiple contour integrals of screened vertex operators and evaluating the four-point structure constants explicitly ^{[Dotsenko-Fateev 1984]}. Their evaluation of the crossing equations as Selberg-type integrals supplied the first complete solution of the conformal bootstrap for an interacting unitary family ^{[Dotsenko-Fateev 1985]}.

Bibliography Master

• Feigin, B. L. & Fuks, D. B., "Invariant skew-symmetric differential operators on the line and Verma modules over the Virasoro algebra", Functional Analysis and Its Applications 16 (1982), 114–126.
• Belavin, A. A., Polyakov, A. M. & Zamolodchikov, A. B., "Infinite conformal symmetry in two-dimensional quantum field theory", Nuclear Physics B 241 (1984), 333–380.
• Dotsenko, Vl. S. & Fateev, V. A., "Conformal algebra and multipoint correlation functions in 2D statistical models", Nuclear Physics B 240 (1984), 312–348.
• Dotsenko, Vl. S. & Fateev, V. A., "Four-point correlation functions and the operator algebra in 2D conformal invariant theories with central charge $c\le 1$", Nuclear Physics B 251 (1985), 691–734.
• Di Francesco, P., Mathieu, P. & Sénéchal, D., Conformal Field Theory, Springer, 1997. Chapters 8–9.
• Ginsparg, P., "Applied Conformal Field Theory", Les Houches Lectures (1988), arXiv/9108028.