Three-Dimensional Yang–Mills–Higgs Scaling-Limit Theorems

Rigorous gauge-field scaling limits exist only in sharply specified regimes. A recent theorem constructs weak-coupling, fixed-length U(1) and SU(2) lattice Yang–Mills–Higgs limits in every $d\ge2$; in three dimensions the limiting projected gauge field is a massive Gaussian Euclidean Proca field. It is not a construction of confining four-dimensional Yang–Mills theory or of a non-Gaussian gauge continuum.

Required background. The constructive program and cutoff removal supplies the limit grammar; thermodynamic limits, correlation decay, and phase control separates volume from spacing limits; stochastic quantization: invariant measures and convergence theorems supplies distributional convergence ideas.

Helpful background. Fermionic, gauge, and lattice reflection positivity explains the positivity issue; Gauss-law infrasectors and asymptotic charge classes explains why gauge charges need not be compactly localized.

The lattice theorem and its scaling domain

Take a cubic lattice of spacing $\varepsilon$ in $d=3$, gauge group $G=U(1)$ or $SU(2)$, Wilson plaquette action, and a Higgs field in the fundamental representation constrained to fixed length. The finite-volume probability law is gauge invariant. Unitary gauge sends the Higgs orientation to a fixed vector, leaving a massive gauge-link action. Let $g\to0$ be the gauge coupling and $\alpha\to\infty$ the Higgs length with

$$\alpha g=c\varepsilon, \qquad g=O(\varepsilon^{50d}),$$

for fixed $c>0$. After stereographic projection of links near the identity and the theorem’s rescaling, the resulting random distributional one-form converges in law against compactly supported smooth test one-forms to a Euclidean Proca field of positive mass.

This is Chatterjee 2026, Theorems 3.1–3.2, pp. 10–13: U(1) and SU(2) are separate statements, valid for $d\ge2$. The proof fixes unitary gauge, derives the projected-link density, proves that links stay in a small neighborhood with overwhelming probability, compares local densities to a discrete Proca Gaussian, and then controls the accumulation of errors. The key density estimates and final convergence proofs occupy Chatterjee 2026, §§5.3–5.8, pp. 37–49.

The limiting object is a random distributional one-form. Convergence is not pointwise convergence of link variables, convergence of every Wilson loop, or convergence of a Hilbert-space Hamiltonian. Moreover, the coupling is driven extremely rapidly to zero; the Gaussian limit is consistent with that weak-coupling scaling.

A gauge-invariant worked observable

In the U(1) branch, the curvature $F=dA$ of the limiting Proca one-form is gauge invariant. For a compactly supported smooth test two-form $h$, integration by parts gives

$$F(h)=A(d^*h).$$

Because $d^*h$ is a smooth test one-form, convergence of the projected lattice field $A_\varepsilon$ implies convergence of the smeared linearized lattice curvature $F_\varepsilon(h)=A_\varepsilon(d^*h)$ by the continuous-mapping theorem. Its two-point function tends to

$$\mathbb E[F(h)F(k)]=C_{\mathrm{Proca}}(d^*h,d^*k).$$

This supplies a concrete gauge-invariant correlation observable in the Abelian theorem’s linearized scaling regime. It does not establish convergence of unsmeared curvature, topological flux sectors, or finite macroscopic Wilson loops. In SU(2), the theorem itself is formulated after unitary gauge and stereographic projection; a comparable claim for nonlinear gauge-invariant Wilson observables requires a separate result.

The physical interpretation belongs with Coulomb, Higgs, and confining regimes. The rigorous conclusion is Higgs-regime mass generation in this scaling, not a proof that all three regimes are analytically connected or separated.

An independent check is dimensional and Gaussian. The continuum covariance must be positive on test one-forms and have a massive denominator; taking the exterior derivative annihilates pure-gradient test components. The two-point curvature covariance derived above must agree whether one differentiates the limiting covariance or first differentiates the lattice field and then takes the limit.

Adversarial overclaim test

Demand instead four-dimensional pure non-Abelian Yang–Mills, a fixed nonzero renormalized coupling, an area law for Wilson loops, and a reconstructed mass-gap spectrum. Every decisive hypothesis has changed: there is no fixed-length Higgs field to impose unitary gauge and mass, the coupling scaling is absent, the observable is nonlinear and extended, and no OS-to-Wightman construction is provided. The theorem licenses none of those conclusions.

Even in $d=3$, it would be wrong to call the result non-Gaussian: the main theorem explicitly identifies a massive Gaussian limit and lists construction of a non-Gaussian scaling limit as open; Chatterjee 2026, §3.5, pp. 16–17.

Renormalized stochastic Yang–Mills–Higgs dynamics in three dimensions supplies another important but different result: a canonical Markov process on gauge-orbit space up to possible explosion. It does not by itself provide the invariant Euclidean measure or global constructive QFT claimed here.

Exercises

1. Continuous mapping. Why does $A_\varepsilon\Rightarrow A$ as random distributions imply $dA_\varepsilon\Rightarrow dA$?

Solution

Distributional differentiation is a continuous linear map: $(dA)(h)=A(d^*h)$. The continuous-mapping theorem therefore applies to every finite family of test two-forms.

2. Wilson loops. Explain why convergence against smooth test one-forms does not immediately give convergence of $\exp(i\oint_\gamma A)$.

Solution

The current supported on a curve is singular, not a smooth test one-form. Evaluating a distribution on it and exponentiating require additional regularity or renormalization estimates not contained in the stated topology.

References

• Chandra, Ajay, Ilya Chevyrev, Martin Hairer, and Hao Shen. “Stochastic Quantisation of Yang–Mills–Higgs in Three Dimensions.” arXiv:2201.03487, 2022. Open PDF.
• Chatterjee, Sourav. “A Scaling Limit of SU(2) Lattice Yang–Mills–Higgs Theory.” Probability and Mathematical Physics 7 (2026): 339–381. DOI. Open PDF.