Strong-Coupling Phases and Cross-Method Evidence

A strong-coupling phase claim is reliable when the observable, regulator, infinite-volume limit, continuum limit, and approximation are all explicit and when methods with genuinely different systematics agree in an overlap. For the two-dimensional $O(N>2)$ model, perturbative running locates the generated scale, large $N$ constructs a massive saddle, exact scattering constrains the continuum spectrum, and lattice step scaling tests the continuum approach. None of these statements should be replaced by the weaker fact that continuous symmetry does not break.

Required background. Gap equations, dimensional transmutation, and physical mass distinguishes a saddle scale from a measured correlation length. Helpful background. Symmetry realization and order parameters fixes what “unbroken” means, while nonperturbative regimes, observables, and control supplies the regime-by-regime comparison.

The O(N > 2) vector-mass claim

Use the concrete claim:

In the zero-temperature, infinite-volume continuum $O(N)$ model in $1+1$ dimensions, for fixed integer $N>2$, the lightest state in the vector channel has a nonzero mass $M$, and connected vector correlators decay exponentially at long distance.

This statement fixes the dimension, the range of $N$, the channel, and the observable. Its limit order must still be operational: a regulator-level definition can take $a/\xi\to0$ at fixed physical $ML$ and then $ML\to\infty$, or use a demonstrated path-independent joint limit. Merely saying “infinite-volume continuum” does not specify that order. The claim does not assert spontaneous symmetry breaking, a particular bare-cutoff formula for $M$, or a four-dimensional confinement mechanism.

The relevant continuum correlator is

$$C^{ab}(x) =\langle n^a(x)n^b(0)\rangle_c =\delta^{ab}C(x),$$

with

$$C(x)\sim Z\,\frac{e^{-M\lvert x\rvert}}{\sqrt{M\lvert x\rvert}}$$

for an isolated one-particle contribution in two Euclidean dimensions, up to heavier states and normalization conventions. A finite correlation length can also be defined by the second moment; it need not equal $1/M$ at finite lattice spacing.

Evidence matrix for the O(N > 2) mass statement

What each method contributes to the two-dimensional O(N > 2) vector-channel mass claim.

Method	Computed object	Control and assumptions	Useful cross-check	What it does not establish alone
Weak-coupling RG	Beta function and RG-invariant scale	Short distances, small renormalized coupling, fixed normalization	Universal leading coefficients and ultraviolet scaling of dimensionless observables	An infrared pole or exponential clustering
Large N	Constraint saddle, vector propagator, and systematic 1/N corrections	N tends to infinity at fixed ’t Hooft coupling; regulator and order of limits declared	Same transmuted exponential as RG and compatible vector-channel analytic structure	The exact spectrum at a chosen finite N
Factorized scattering	Continuum vector multiplet, two-body amplitudes, and absence of physical-strip bound-state poles in the minimal solution	Quantum integrability, unitarity, crossing, Yang–Baxter consistency, and bootstrap completeness	Thermodynamic Bethe ansatz or form-factor ultraviolet behavior	That a classical Lax pair automatically survives quantization
Continuum lattice extrapolation	Finite-volume coupling, correlation lengths, step scaling, and dimensionless mass ratios	Several lattice spacings and volumes; controlled a/ξ and L/ξ extrapolations	Universal continuum step-scaling curve and matching to perturbation theory at small volume	Exact continuum behavior from one finite lattice or one fit ansatz
Strong-bare-coupling expansion	Absolutely convergent polymer/graph series in its high-temperature lattice domain	Small hopping parameter or high-temperature domain where activities are summable	Direct exponential bounds and agreement with numerical data inside the convergence domain	The asymptotically free continuum limit where ξ/a tends to infinity

The strong-coupling laboratory map places these controls beside those of the other models, and the regime comparison states the corresponding forbidden transfers.

How the methods meet

Perturbative RG predicts the ultraviolet dependence of a renormalized coupling and constructs a scheme-dependent $\Lambda$ parameter Polyakov 1975, pp. 79–81. Large $N$ gives

$$M \sim \mu\exp\!\left[-\frac{2\pi}{t_R(\mu)}\right]$$

at leading order, matching the RG exponent in the same normalization. Exact scattering then expresses physical masses and amplitudes in terms of one overall scale. The separate ultraviolet matching calculation yields the finite-$N$ relation

$$\frac{M}{\Lambda_{\overline{\mathrm{MS}}}} =\frac{(8/e)^{1/(N-2)}} {\Gamma\!\left(1+\frac{1}{N-2}\right)}, \qquad N\ge3,$$

with $M/\Lambda_{\overline{\mathrm{MS}}}=8/e$ at $N=3$ Hasenfratz and Niedermayer 1990, pp. 529–532. The agreement is nontrivial because it connects infrared spectral data with ultraviolet renormalization, but it is not independent of the perturbative matching used to define $\Lambda_{\overline{\mathrm{MS}}}$. The numerical limiting cases are checked in the chapter’s benchmark.

The factorized $O(N)$ S-matrix is constrained by global symmetry, unitarity, crossing, and the Yang–Baxter equation Zamolodchikov and Zamolodchikov 1978, §§ 2–4. Its consistency is much stronger than merely finding infinitely many classical currents. It is nevertheless a bootstrap construction with extensive ultraviolet and finite-volume checks, not by itself a constructive existence theorem identifying a unique microscopic continuum model.

Lattice calculations provide a regulator-level definition. A continuum trajectory requires

$$\frac{\xi}{a}\longrightarrow\infty, \qquad \frac{L}{\xi}\longrightarrow\infty$$

for infinite-volume observables, or a fixed $L/\xi$ for finite-volume step scaling, while dimensionless physical quantities approach regulator-independent limits. The finite-volume running-coupling construction is developed in Lüscher, Weisz, and Wolff 1991, pp. 221–243. Several lattice spacings are needed to resolve corrections such as powers of $a/\xi$ and logarithms. A single large but finite $\xi/a$ is evidence, not the limit itself.

A bounded cluster-expansion result

On a lattice at sufficiently small nearest-neighbor coupling, connected correlations can be expanded as sums of connected polymers. A Kotecký–Preiss-type absolute-convergence condition requires a positive size function $a(\gamma)$ such that the incompatibility-weighted activities obey a bound of the schematic form

$$\sum_{\gamma'\not\sim\gamma} \lvert z(\gamma')\rvert e^{a(\gamma')} \le a(\gamma).$$

When a specified regulator and its activities satisfy such a condition with a suitable spatial weight, every polymer connecting regions separated by distance $R$ carries an overall bound of the form

$$\lvert C_{\mathrm{lat}}(R)\rvert \le A e^{-R/\xi_{\mathrm{sc}}}.$$

This is a genuine exponential-clustering result once the named absolute-convergence hypotheses are verified Kotecký and Preiss 1986, Theorem 1. The page does not claim to have mapped and bounded the activities of a particular $O(N)$ regulator. Even when that step is complete, the result remains deliberately limited: the high-temperature or strong-bare-coupling regime typically has $\xi_{\mathrm{sc}}/a=O(1)$. The asymptotically free continuum limit instead requires $\xi/a\to\infty$. A proof in the first region cannot be promoted to the second without a chain of uniform estimates that bridges the regimes.

This example illustrates why “strong coupling” is not one universal limit. It can mean a large renormalized infrared coupling in a continuum theory, a small hopping parameter in a lattice expansion, or a large bare coupling. The controls and observables differ.

Symmetry is a consistency condition, not the mechanism

Coleman’s theorem excludes spontaneous breaking of a continuous internal symmetry in a relativistic $1+1$-dimensional QFT under its standard assumptions Coleman 1973, pp. 259–264. The massive symmetric $O(N>2)$ phase is compatible with this constraint.

The converse is false. The $O(2)$ compact boson can have algebraic correlations and no spontaneous magnetization. Therefore

$$\text{no continuous SSB} \;\not\Rightarrow\; \text{mass gap}.$$

The nonzero $O(N>2)$ mass is a dynamical conclusion supported by the matrix above. Confusing the theorem with the mechanism erases the decisive distinction between $N=2$ and $N>2$.

Correlated systematics

Two nominally different results are not fully independent when they share:

• the same perturbative matching coefficient;
• the same finite-volume ansatz or lattice ensemble;
• the same leading large-$N$ truncation;
• the same assumed particle content in a bootstrap;
• an order of limits that suppresses the same infrared fluctuations.

An effective comparison varies at least one of these. For example, a finite-volume step-scaling curve can be checked against ultraviolet perturbation theory at small $L$ and against a massive infrared description at large $L$, while simulations with distinct lattice actions test universality of the continuum extrapolation.

Common pitfalls

Equating strong bare coupling with continuum strong dynamics. A convergent lattice strong-coupling series may live far from the asymptotically free continuum trajectory. Its mass scale in lattice units need not have a universal continuum interpretation.

Calling a fit a controlled extrapolation. Several fit forms can describe a narrow range of lattice spacings. Stability under removing coarse points, changing the regulator, and using theoretically motivated correction terms is part of the evidence.

Counting shared assumptions twice. RG matching and an exact mass–scale comparison are powerful, but not independent if both use the same perturbative coefficient without an additional infrared check.

Exercises

1. Suppose a lattice calculation gives correlation lengths $\xi/a=8,16,32$ and a ratio

$$R(a)=\frac{M_2(a)}{M_1(a)} =R_0+c\left(\frac{a}{\xi}\right)^2.$$

Describe a minimal continuum test.

Solution

Fit the three or more spacings to the predicted correction form, repeat after removing the coarsest point, and compare with a fit that includes the leading allowed logarithmic or higher-order correction. Check that $L/\xi$ is held fixed or large enough that finite-volume effects are smaller than the extrapolation error. Agreement of $R_0$ across a second lattice action would test regulator universality.

2. Explain why exponential clustering from a high-temperature cluster expansion does not, by itself, prove the continuum $O(3)$ mass gap.

Solution

The cluster expansion controls a region where polymer activities are small and the correlation length in lattice units is typically finite. The continuum $O(3)$ theory is reached by tuning the bare coupling so that $\xi/a\to\infty$. Unless the bounds remain uniform along that trajectory and identify the continuum channel, the convergence domain does not reach the required limit. The result is rigorous within one regime but not a bridge between regimes.

3. Evaluate the exact mass normalization at $N=3$ and derive its first large-$N$ correction.

Solution

At $N=3$, $\Gamma(2)=1$, so

$$\frac{M}{\Lambda_{\overline{\mathrm{MS}}}}=\frac8e.$$

Let $\varepsilon=1/(N-2)$. Using $\log\Gamma(1+\varepsilon)=-\gamma_E\varepsilon+O(\varepsilon^2)$,

$$\log\frac{M}{\Lambda_{\overline{\mathrm{MS}}}} =\varepsilon\log(8/e)-\log\Gamma(1+\varepsilon) =\frac{\log(8/e)+\gamma_E}{N-2} +O\!\left((N-2)^{-2}\right).$$

Thus the ratio approaches one. This tests the large-$N$ limit while retaining the fact that the exact finite-$N$ normalization required separate ultraviolet matching.

4. Two papers quote the same $M/\Lambda_{\overline{\mathrm{MS}}}$ value, one from exact thermodynamics and one from a lattice continuum extrapolation calibrated with the same perturbative matching coefficient. Are they fully independent? Propose a stronger cross-check.

Solution

They are not fully independent in their ultraviolet normalization because both inherit the same perturbative conversion to $\Lambda_{\overline{\mathrm{MS}}}$. The lattice regulator, continuum extrapolation, and infrared measurement still provide distinct systematics. A stronger comparison can add a finite-volume step-scaling curve across several lattice actions, an exact-scattering prediction for a dimensionless finite-volume observable, or a form-factor correlator whose normalization does not reuse the same matching coefficient. The shared assumption should be recorded rather than counted twice.

Continue

The Principal Chiral Model and the Integrability Bridge isolates the extra quantum tests behind exact scattering. The Gross–Neveu Model and Dynamical Mass Generation gives a second large-$N$ gap calculation whose symmetry constraints differ from the $O(N)$ model.

References

• Coleman, Sidney. “There Are No Goldstone Bosons in Two Dimensions.” Communications in Mathematical Physics 31 (1973): 259–264. DOI.
• Hasenfratz, Peter, and Ferenc Niedermayer. “The Exact Mass Gap of the O(N) σ-Model for Arbitrary N ≥ 3 in d = 2.” Physics Letters B 245 (1990): 529–532. DOI.
• Kotecký, Roman, and David Preiss. “Cluster Expansion for Abstract Polymer Models.” Communications in Mathematical Physics 103 (1986): 491–498. DOI.
• Lüscher, Martin, Peter Weisz, and Ulli Wolff. “A Numerical Method to Compute the Running Coupling in Asymptotically Free Theories.” Nuclear Physics B 359 (1991): 221–243. DOI.
• Polyakov, A. M. “Interaction of Goldstone Particles in Two Dimensions. Applications to Ferromagnets and Massive Yang–Mills Fields.” Physics Letters B 59 (1975): 79–81. DOI.
• Zamolodchikov, Alexander B., and Alexey B. Zamolodchikov. “Relativistic Factorized S Matrix in Two-Dimensions Having O(N) Isotopic Symmetry.” Nuclear Physics B 133 (1978): 525–535. DOI.