Lattice gauge theory and continuum limits

Lattice gauge theory gives us both a mathematically defined regulated model and a practical route to nonperturbative calculations. It preserves gauge invariance while replacing a formal integral over continuum fields by finite-dimensional group integrals. Its success also makes the remaining challenge unusually concrete: what can be controlled when the cutoff disappears?

A finite definition

Prerequisites: gauge transformations and basic path integrals from graduate QFT. The Haar-measure bridge explains the new integration measure.

Outcome: distinguish a finite lattice, physical time evolution, a numerical extrapolation and a continuum existence argument, with the units and assumptions of each kept visible.

A link variable $U_\ell\in SU(N)$ represents parallel transport. Closed products around plaquettes enter the Wilson action,

$$S_W=\beta\sum_p \left(1-\frac1N\operatorname{Re}\operatorname{tr}U_p\right), \qquad \beta=\frac{2N}{g_0^2}.$$

The convention here counts each elementary plaquette once. Other papers may normalize their coupling differently, so numerical values of “$\beta$” cannot be compared before translating conventions.

With normalized Haar measure on each link, a finite lattice has a finite positive partition function. Wilson’s original paper introduced this nonperturbative formulation. Our lattice-definition lesson works through the transformations and classical expansion.

There are two different distances

The lattice spacing $a$ sets the resolution. The physical box length $L=N_sa$ sets the available volume. Decreasing $a$ at fixed $N_s$ shrinks the box, so it does not accomplish both desired limits at once.

If a physical mass remains fixed, its lattice-unit value satisfies

$$m_{\mathrm{latt}}=a\,m_{\mathrm{phys}}\to0, \qquad \xi_{\mathrm{latt}}=\frac1{a\,m_{\mathrm{phys}}}\to\infty.$$

The growing correlation length in lattice sites is compatible with a finite physical correlation length. A credible continuum argument must specify a trajectory with a physical reference scale held under control.

Explore finite volume, continuum limits, and the units lab. These distinctions are essential for interpreting both theorems and simulations.

What a calculation measures

To estimate a glueball mass, choose gauge-invariant operators, calculate their connected Euclidean correlators, and examine their decay. Multiple operators help separate states with different overlaps. Reliable analysis also addresses correlations between measurements, fitting choices, lattice artifacts, and physical volume.

For example,

$$m_{\mathrm{eff}}(t) =\frac1{\delta t}\log\frac{C(t)}{C(t+\delta t)}$$

approaches the lowest visible energy in an appropriate long-time regime. The meaning of $\delta t$ depends on whether time is measured physically or in lattice steps.

The spectroscopy lesson gives the workflow. The glueball benchmark dossier follows an actual continuum-extrapolated calculation, keeping numerical evidence distinct from a constructive proof.

What can be proved at the cutoff?

Strong-coupling expansions and correlation methods can establish genuine statements such as area-law bounds in specified parameter regimes. The 2025 lattice area-law dossier is an example with explicit scope.

Positivity bootstrap methods use exact loop relations and inequalities to constrain observables. Their numerical implementation introduces a further question: how is a displayed bound certified, and what happens as more constraints are included? Read the SU(3) bootstrap dossier and the loop-equation update.

Follow the evidence into the target

Task	Core concept page
Convert transfer eigenvalues to physical energies and distinguish sampling time	Transfer matrix
Calculate a plaquette term and separate a formal series from a uniform bound	Strong-coupling expansion
Hold a reference scale fixed while changing the cutoff and volume	Continuum scaling
Understand an operator basis, generalized eigenvalues and omitted states	Spectral extraction

A useful lattice result should always identify its spacing, volume, coupling normalization, observables, and limit. The Research Atlas keeps those fields visible. The Proof Map shows which further step would connect each result to a physical continuum theory with a mass gap.