Integrable Deformations and Renormalization-Group Flows

A relevant deformation of a conformal field theory preserves integrability only when an infinite family of higher-spin conservation laws survives the perturbation. Relevance alone does not ensure this. When the survival test succeeds, ultraviolet operator data, a mass–coupling relation, factorized scattering, and the thermodynamic Bethe-ansatz scaling function can describe one trajectory. Matching only the ultraviolet and infrared endpoints cannot prove that the proposed exact data describe the intervening flow.

Required background. Thermodynamic Bethe ansatz and finite-size ground-state energy supplies the scaling function and effective-central-charge checks. Helpful background. Relevant, marginal, and irrelevant directions supplies the linearized RG classification of perturbations.

Perturbing a conformal fixed point

Start from a two-dimensional CFT and perturb by a spinless operator $\Phi$:

$$\mathcal A = \mathcal A_{\rm CFT} +g\int\mathrm d^2x\,\Phi(x).$$

If $\Phi$ has scaling dimension $x_\Phi<2$, then

$$[g]=2-x_\Phi.$$

Dimensional analysis predicts a mass scale

$$m=\kappa_\Phi\, \lvert g\rvert^{1/(2-x_\Phi)},$$

provided the flow is massive and no other scale intervenes. The exponent is fixed by the CFT dimension; the coefficient $\kappa_\Phi$ depends on the normalization of $\Phi$, the coupling convention, and the model. An exact mass–coupling relation must state all three.

Suppose the CFT contains a holomorphic higher-spin current $J_s(z)$. To first order in $g$,

$$\bar\partial J_s =g\,\mathcal A_s+O(g^2),$$

where $\mathcal A_s$ is proportional, in a convention-dependent normalization, to the simple-pole residue in the OPE $J_s(z)\Phi(w,\bar w)$. The current remains conserved at this order if

$$\mathcal A_s=\partial\Theta_{s-2}$$

for a local field $\Theta_{s-2}$, because then

$$\bar\partial J_s-\partial(g\Theta_{s-2})=O(g^2).$$

The counting compares possible OPE residues with possible total derivatives. It is a selection rule, not an all-orders theorem by itself. Operator mixing, resonances, higher orders, and nonperturbative definition of the charges still have to be controlled. Zamolodchikov develops this conserved-current route from conformal perturbation theory to massive integrable models in Zamolodchikov 1989a, §§ 2–6, pp. 645–673.

The two integrable Ising axes

The critical Ising CFT has energy operator $\varepsilon$ of scaling dimension $1$ and spin operator $\sigma$ of scaling dimension $1/8$. Write

$$\mathcal A = \mathcal A_{\rm Ising} +\tau\int\mathrm d^2x\,\varepsilon(x) +h\int\mathrm d^2x\,\sigma(x).$$

The two coordinate axes illustrate distinct integrable deformations.

Thermal axis

For $h=0$, the deformation is a free massive Majorana theory. The mass is linear in $\lvert\tau\rvert$ after the operator normalization is fixed, the elastic amplitude can be represented as $S=-1$, and the TBA kernel vanishes. The two signs of $\tau$ describe the ordered and disordered massive phases; the particle and order/disorder operator descriptions differ even though the on-shell free-fermion structure is simple.

Magnetic axis

For $\tau=0$ and $h\ne0$, the mass scale behaves as

$$m_1\propto\lvert h\rvert^{8/15}.$$

The surviving higher-spin integrals of motion lead to the celebrated eight-particle $E_8$ mass pattern and a factorized scattering proposal. Zamolodchikov’s original analysis presents the exact mass spectrum and S matrix as a conjecture supported by integrability and bootstrap consistency Zamolodchikov 1989b, pp. 4235–4248. Later exact-data calculations and finite-size comparisons test that identification, but the historical claim should not be silently promoted into a general constructive theorem.

Away from the axes

For generic $\tau h\ne0$, the higher-spin conservation laws compatible with one axis are broken by the other perturbation. Particle production and unstable excitations are then allowed, and the factorized bootstrap no longer describes the generic theory. The fact that both axes leave the same ultraviolet CFT does not make their massive dynamics interchangeable.

Exact-data bridge for a proposed flow

A candidate integrable flow should connect the following data without changing conventions midstream.

1. UV operator: identify the CFT, the perturbing operator, its normalization, and the sign and units of its coupling.
2. Conserved charges: show which higher-spin currents survive and what quantum anomaly or mixing analysis is required.
3. Mass scale and spectrum: state the mass–coupling relation and stable particle multiplets.
4. Two-body amplitudes: impose factorization, Yang–Baxter where needed, unitarity, crossing, pole residues, fusion, and a named CDD choice.
5. Finite-size function: solve the TBA system with its kernel and statistics convention over intermediate as well as asymptotic $mR$.
6. Observable checks: compare finite-volume levels, form factors, sum rules, or independent regulated calculations without reusing fitted data as a test.

The integrability exact-data chain organizes these steps and shows where each extra assumption enters. The exact and rigorous status comparison distinguishes an exact description within the accepted integrable model from a rigorous construction of that flow.

Why endpoint matching is insufficient

For a massive flow from a unitary CFT, one expects after bulk subtraction

$$c_{\rm eff}(R) \longrightarrow \begin{cases} c_{\rm UV}, & R\to0,\\ 0, & R\to\infty \end{cases}$$

when the infrared has a unique gapped vacuum in the chosen sector. These are necessary checks, not a proof of the entire trajectory.

Different scattering kernels can share the same ultraviolet effective central charge and the same trivial massive infrared limit. A CDD deformation can preserve unitarity and crossing while changing phase shifts and intermediate-size energies. To identify a flow, test the full scaling function, mass ratios, mass–coupling normalization, pole structure, and independent observables. Degenerate vacua, nonunitary endpoints, or massless flows also change the naïve endpoint formula and must be declared.

Model-specific limits

Integrability statements do not automatically survive:

• a second relevant perturbation;
• a boundary condition that lacks compatible reflection charges;
• an irrelevant deformation with uncontrolled ultraviolet behavior;
• a massless limit in which left- and right-moving completeness changes; or
• a CDD deformation whose ultraviolet completion is not established.

The Ising axes give a counterexample to universality from common UV data: thermal and magnetic perturbations begin at the same CFT but have different spectra, scattering, operator maps, and finite-size functions.

Common pitfalls

Calling every relevant perturbation integrable. Relevance predicts how a coupling grows. Integrability requires surviving higher-spin conservation laws and their quantum consistency.

Inferring the trajectory from two limits. UV and IR endpoints constrain a scaling function but do not determine it. Intermediate finite-size data are part of the identification.

Suppressing operator normalization in a mass–coupling relation. Rescaling $\Phi$ inversely rescales $g$ and changes $\kappa_\Phi$. Quote the convention before comparing coefficients.

Exercises

1. Derive the mass exponent for perturbation by an operator of dimension $x_\Phi$ in two dimensions, and apply it to the Ising magnetic field.

Solution

The action is dimensionless, so $[g]=2-x_\Phi$. If $g$ is the only dimensionful input, $m\propto\lvert g\rvert^{1/(2-x_\Phi)}$. For $x_\sigma=1/8$,

$$\frac{1}{2-x_\sigma} =\frac{1}{2-1/8} =\frac{8}{15},$$

hence $m_1\propto\lvert h\rvert^{8/15}$.

2. A candidate TBA has the correct limits $c_{\rm eff}(0)=1/2$ and $c_{\rm eff}(\infty)=0$. List three further checks needed before identifying it with the magnetic Ising trajectory.

Solution

Suitable independent checks include the $E_8$ mass ratios and pole residues, the correctly normalized $m_1$–$h$ relation, the full intermediate-$mR$ scaling function or finite-volume levels, and operator form factors or sum rules. Any three address information that the two endpoint values do not contain.

References

• Zamolodchikov, Alexander B. “Integrable Field Theory from Conformal Field Theory.” In Integrable Systems in Quantum Field Theory and Statistical Mechanics, Advanced Studies in Pure Mathematics 19 (1989): 641–674. DOI.
• Zamolodchikov, Alexander B. “Integrals of Motion and S-Matrix of the (Scaled) $T=T_c$ Ising Model with Magnetic Field.” International Journal of Modern Physics A 4 (1989): 4235–4248. DOI.