Answers are grounded in this paper and cite their sources.
Let G be a compact connected Lie group with a bi-invariant metric, and let w be a strictly positive C¹ class function on G. Examples are the Wilson weight at every real coupling β, the Villain weight and the heat kernel. We prove that the lattice gauge measure with plaquette weight w on the torus (ℤ/Lℤ)² satisfies a log-Sobolev inequality for the gradient (Langevin) Dirichlet form, with a constant that depends on G, the metric and w but not on L. Piazza and Shen proved this for SU(N) with the Wilson weight, and remark that their strategy also gives SO(N), N ≥ 3. Their argument needs the Ricci curvature of G to be uniformly positive, and they note in their Remark 1.2 that for this reason it does not extend to U(1) without further ideas. The same obstruction affects U(N) and every group whose Lie algebra has a nontrivial centre. We keep their two-scale decomposition and replace the curvature step for the coarse-grained measure by approximate tensorization of entropy in the sense of Caputo, Menz and Tetali. That criterion needs only sup-norm bounds on ratios of conditional densities, which Doeblin's argument supplies for every compact group. We write out the Caputo–Menz–Tetali theorem for bounded densities on arbitrary product probability spaces. The constant we obtain grows very fast with the coupling. By Piazza and Shen's example, some growth in β cannot be avoided in this generality.
No takes yet. Share an insight, caveat, or question.
Yan Ru Pei (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: