The Non-Abelian Mass Gap as a Spectral Statement

A gauge-theory mass gap is a statement about the spectrum of the physical, gauge-invariant theory: after choosing a vacuum sector and taking the continuum and infinite-volume limits, every nonvacuum state lies at least a positive energy $\Delta$ above the vacuum. It is not a mass parameter fitted to a gauge-fixed gluon propagator, and it is not synonymous with an area law or the absence of colored asymptotic particles.

Required background. Confinement definitions separates the gap from line and screening criteria. Spectral decomposition of two-point functions supplies the Hilbert-space expansion.

Helpful background. Euclidean correlators and Schwinger functions supplies reflection positivity and reconstruction assumptions.

Shared comparison. The diagnostic map exhibits gapped theories without an area law, and the claim–evidence comparison keeps the physical gap distinct from its proposed mechanism. The later exact and rigorous status comparison places the Yang–Mills statement beside lower-dimensional theorems without upgrading either.

The gap lives in the physical spectrum

Let $\mathcal H_{\mathrm{phys}}$ satisfy the gauge constraints, and choose a translationally invariant vacuum $|\Omega\rangle$ in one superselection sector. Normalize its energy to zero. The spectral gap is

$$\Delta =\inf\left(\operatorname{spec}H\big|_{\mathcal H_{\mathrm{phys}}} \setminus\{0\}\right)>0.$$

If there are several exactly degenerate vacua, remove their finite-dimensional vacuum subspace before taking the infimum. If a continuous symmetry is spontaneously broken, its Goldstone states force $\Delta=0$. In infinite volume, the infimum can be the threshold of a continuum; a stable one-particle pole is not required by the definition.

The order of limits is essential. A finite spatial box often has a discrete positive first level even when the infinite-volume theory is gapless. A lattice regulator may also possess a gap in lattice units that vanishes after the spacing $a\to0$. Writing $L$ for the physical box size, not the number of lattice sites, the continuum claim is therefore

$$\Delta_{\mathrm{phys}} =\lim_{L\to\infty}\lim_{a\to0} \frac{(a\Delta)(a,L)}{a}>0,$$

with bare parameters tuned to a renormalized continuum trajectory and finite-volume effects controlled. The notation emphasizes the logical tests; a practical calculation may coordinate the limits differently while demonstrating the same regulator-independent result.

Gauge-invariant correlators expose spectral support

For a renormalized gauge-invariant local operator $\mathcal O$ with $\langle\mathcal O\rangle=0$, define the zero-momentum Euclidean correlator

$$C_{\mathcal O}(\tau) =\int\mathrm d^{d-1}x\, \langle\mathcal O(\tau,\mathbf x)\mathcal O(0)\rangle_c.$$

Reflection positivity and a transfer-matrix reconstruction give a positive spectral representation,

$$C_{\mathcal O}(\tau) =\int_{0}^{\infty}\mathrm dE\, \rho_{\mathcal O}(E)e^{-E\tau}, \qquad \rho_{\mathcal O}(E)\ge0.$$

If the lowest support in that channel is $m_{\mathcal O}>0$ and has nonzero overlap, then

$$-\frac{\mathrm d}{\mathrm d\tau}\log C_{\mathcal O}(\tau) \longrightarrow m_{\mathcal O} \qquad(\tau\to\infty).$$

This determines the lightest state coupled to $\mathcal O$, not automatically the global gap. To bound $\Delta$, the operator set must cover all physical symmetry sectors that could contain a lighter state. Exponential decay of one glueball channel alone is insufficient.

For pure four-dimensional Yang–Mills, representative interpolators include

$$\mathcal O_{0^{++}}=\operatorname{tr}F_{\mu\nu}F_{\mu\nu}, \qquad \mathcal O_{0^{-+}}=\operatorname{tr}F_{\mu\nu}\widetilde F_{\mu\nu}.$$

Their Euclidean correlation matrices can be projected into irreducible spin, parity, and charge-conjugation channels. A variational basis improves overlap and separates excited states. Renormalization, finite volume, discretization, scale setting, and continuum extrapolation belong to the regulated measurement workflow; the spectral definition does not disappear when those systematics become difficult.

Propagator scales are not the spectral gap

A gauge-fixed gluon two-point function can show a finite infrared scale, complex poles, or positivity violation. These features may constrain a mechanism, but the field $A_\mu^a$ is not a gauge-invariant operator on $\mathcal H_{\mathrm{phys}}$. Its pole structure can depend on the gauge, gauge-fixing prescription, and treatment of Gribov copies.

Likewise, at nonzero temperature a screening mass extracted from spatial decay is an eigenvalue of a spatial transfer problem. It need not equal the real-time Hamiltonian gap. A Debye mass, a dual-photon mass in a compactified effective theory, and a four-dimensional glueball mass are each meaningful within their definitions, but they are not interchangeable.

The invariant test is simple: identify a gauge-invariant operator or physical sector, state the geometry and state, and show positive spectral support bounded away from zero after the relevant limits.

Mass gap and confinement remain independent

Three comparisons prevent the common conflation.

• A fundamental Higgs regime can be fully gapped while its fundamental Wilson line is screened.
• The massless Schwinger model has a generated gauge-invariant boson mass and screens external charges rather than supporting an indefinitely rising static energy.
• A Coulomb phase has a massless photon and no area law, so both claims fail together—but this single example does not make them equivalent.

Conversely, a line-operator area law concerns a nonlocal probe and does not by itself construct all gauge-invariant correlators or prove a positive lower bound for the full spectrum. Establishing both properties requires two arguments.

The four-dimensional Yang–Mills problem

The Clay problem asks for a nontrivial quantum Yang–Mills theory on $\mathbb R^4$ for every compact simple gauge group, satisfying sufficiently strong axiomatic properties, together with a mass gap $\Delta>0$ Jaffe and Witten 2006, pp. 129–152. It joins existence and gap; a numerical glueball spectrum in a regulated theory does not by itself supply the requested construction.

As checked on 2026-08-09, the Clay Mathematics Institute continues to state that no proof is known Clay Mathematics Institute, “Yang–Mills & the Mass Gap”. This official status concerns the existence-and-gap theorem. Physical evidence for four-dimensional confinement is a separate, observable-dependent assessment; neither should be used to claim the other has been proved.

Exercises

1. Channel mass versus global gap. A correlator of $\operatorname{tr}F^2$ decays as $e^{-m\tau}$. What has been shown, and what remains to establish the theory’s spectral gap?

Solution

The result shows a lowest spectral support $m$ in the scalar channel, assuming reflection positivity, continuum control, and nonzero overlap. A lighter state could exist in another physical symmetry sector. Establishing the global gap requires covering all sectors or proving a lower bound on the full physical spectrum.

2. Finite-volume trap. A massless particle in a periodic box of side $L$ has lowest nonzero momentum $2\pi/L$. Why is this not a mass gap?

Solution

The finite-size level spacing vanishes as $L\to\infty$. A physical mass gap must approach a strictly positive value in that limit, rather than being set by the infrared regulator.

References

• Clay Mathematics Institute. “Yang–Mills & the Mass Gap.” Millennium Prize Problems. Accessed August 9, 2026. Official problem page.
• Jaffe, Arthur, and Edward Witten. “Quantum Yang–Mills Theory.” In The Millennium Prize Problems, edited by James Carlson, Arthur Jaffe, and Andrew Wiles, 129–152. Providence, RI: American Mathematical Society and Clay Mathematics Institute, 2006. Official PDF.