Three-dimensional gauge theory as a testing ground

Changing the spacetime dimension changes both dimensional analysis and the analytical difficulty of constructing gauge fields. Three-dimensional results can isolate an important obstacle while leaving the four-dimensional problem open.

Prerequisites: dimensional transmutation and semigroups.
Outcomes: compute the coupling’s dimension; distinguish local stochastic existence from equilibrium construction; retain the model and lifetime of a theorem.

Work out what the coupling can tell you

Use the site’s convention that the coupling is outside the curvature. Since $D=\partial-iA$, the field $A$ has mass dimension 1 and $F$ has dimension 2. A dimensionless action

$$S_E=\frac1{2g_d^2}\int d^dx\,\operatorname{tr}F_{\mu\nu}F_{\mu\nu}$$

requires

$$[g_d^2]=\text{mass}^{\,4-d}.$$

In three Euclidean dimensions $g_3^2$ is a mass scale. In four it is dimensionless at the classical level. In two it has dimension mass squared.

If an uncompactified pure three-dimensional theory has no other scale and a well-defined mass $M$, dimensional analysis suggests a form $M=c_Gg_3^2$, where $c_G$ is dimensionless and depends on normalization and group. This is a conditional dimensional statement: it neither proves that $M$ exists nor that $c_G$ is positive. A finite torus adds its circumference; a Higgs sector adds masses and couplings; neither setting has only one scale.

Three Euclidean dimensions do not mean three spatial dimensions

A Euclidean three-dimensional theory would reconstruct, when appropriate hypotheses hold, a theory with two spatial directions and one physical time. In stochastic quantization, an additional evolution parameter $s$ drives a random process on three-dimensional configurations. That parameter is auxiliary; it is not a fourth physical coordinate.

A schematic pure-gauge stochastic equation has a dissipative drift, a gauge-fixing term, renormalization terms, and noise. Its solutions describe how a random gauge configuration changes with $s$. A theorem controlling such solutions can address singular products and gauge covariance without yet producing an invariant probability measure.

Two source statements, with their boundaries intact

Chandra, Chevyrev, Hairer, and Shen arXiv:2201.03487v2, §1, Theorems 1.7 and 1.9 construct renormalized local stochastic Yang–Mills–Higgs dynamics for a compact Lie group on trivial bundles over $\mathbb T^3$ and a gauge-covariant process on gauge orbits, allowing possible finite-time blow-up. The paper develops a distributional state space. The Higgs field takes values in a finite-dimensional real inner-product space carrying an orthogonal representation of the group; that representation is part of the model. The pure-gauge case is a specialization. This is not an infinite-volume physical-gap theorem.

Chevyrev and Shen arXiv:2503.03060v2, §1.1 and Theorem 1.6 in §1.2 strengthen the uniqueness of gauge-covariant renormalization, detecting an incorrect finite gauge-field mass shift through suitably regularized Wilson loops. They work with a connected compact Lie group and allow the pure case $V=\{0\}$. Their introduction explicitly separates this dynamical result from the still-open invariant-measure construction on $\mathbb T^3$. Their Definition 1.2 retains paths with finite-time blow-up.

These are selected-statement and scope readings of the cited versions, checked 24 October 2026, not independent proof verification.

The missing steps are logically separate

Suppose an approximation converges on every sufficiently short stochastic-time interval up to its lifetime. The following conclusions still require arguments:

• The process is defined for all stochastic time, if global existence is claimed.
• An invariant probability measure exists and is identified.
• The selected approximation converges to that equilibrium law.
• The law has the relevant physical positivity and reconstruction properties.
• Infinite-volume limits and a physical spectral gap hold.

These are not interchangeable. Even a rapidly mixing globally defined process might have been designed for an equilibrium measure whose physical correlations are gapless. Rescaling stochastic time changes a relaxation rate without changing the stationary law.

The uniqueness question is valuable because it can help identify possible limiting dynamics. It does not, on its own, prove that every proposed lattice approximation has a limit. The paper’s §1 discussion explicitly leaves the three-dimensional lattice-dynamics scaling limit as another task.

Exercise: audit a one-sentence claim

A report says: “There is a unique gauge-covariant renormalization of the three-dimensional stochastic equation, therefore four-dimensional Yang–Mills has a positive physical mass gap.” Identify three unsupported transitions.

Solution. Uniqueness of a permitted renormalization does not supply global dynamics or equilibrium existence. The stochastic generator is not the physical Hamiltonian. A theory on $\mathbb T^3$, potentially with matter, is not pure Yang–Mills on $\mathbb R^4$. Any transfer needs explicit new results rather than a change of terminology.

For the research context, use the stochastic dossier and stochastic quantization. Return to weak convergence when asking how several possible approximation limits can be identified.