Non-Abelian Screening and Asymptotic Freedom

Non-Abelian gauge fluctuations can make a four-dimensional gauge theory weaker at short distance. With $\beta(g)=\mu\,\mathrm dg/\mathrm d\mu$, the one-loop result is $\beta(g)=-b_0g^3/(16\pi^2)+O(g^5)$. Gauge and ghost fields contribute positively to $b_0$, while fermions and scalars screen and reduce it; asymptotic freedom occurs when the complete coefficient is positive.

Required background. Yang–Mills color algebra and perturbative vertices supplies the color factors and ghost vertex. Beta functions, running masses, and field anomalous dimensions supplies the general RG definition and scheme language.

Helpful background. Ultraviolet and infrared fixed points explains what additional evidence is needed when the leading coefficient vanishes or the flow is followed into strong coupling.

The universal one-loop coefficient

For a compact simple gauge factor with Dirac fermions in representations $R_f$ and complex scalars in representations $R_s$,

$$\beta(g) =-\frac{g^3}{16\pi^2}b_0+O(g^5),$$

where

$$b_0 =\frac{11}{3}C_A -\frac{4}{3}\sum_{\text{Dirac }f}T(R_f) -\frac{1}{3}\sum_{\text{complex }s}T(R_s).$$

Every flavor, multiplet copy, and representation multiplicity belongs in the sums. A Weyl or Majorana fermion contributes half of a Dirac fermion in the same representation; a real scalar contributes half of a complex scalar. For product groups this formula is applied separately to each simple factor, using the multiplicities supplied by the other factors.

The leading coefficient is independent of the covariant gauge parameter and of analytic mass-independent scheme changes. Indeed, if $g'=g+a g^3+O(g^5)$, then $\beta'(g')=(\mathrm dg'/\mathrm dg)\beta(g)$ has the same $g'^3$ coefficient. Higher coefficients require more care, and thresholds require matching when a particle mass is crossed.

Where the sign comes from

The divergent one-loop two-point function is transverse after all required diagrams are summed:

$$q^\mu\Pi_{\mu\nu}^{ab}(q)=0, \qquad \Pi_{\mu\nu}^{ab}(q) \propto\delta^{ab}(q^2\eta_{\mu\nu}-q_\mu q_\nu).$$

In background Feynman gauge, a useful decomposition of the numerator $3b_0$ is

One-loop sector	Contribution to $3b_0$
Adjoint ghost determinant	$+C_A$
Covariant kinetic part of the real vector fluctuation	$-2C_A$
Vector spin coupling to the background curvature	$+12C_A$
Each Dirac representation $R_f$	$-4T(R_f)$
Each complex scalar representation $R_s$	$-T(R_s)$
Total	$11C_A-4\sum_fT(R_f)-\sum_sT(R_s)$

The first three rows are a calculation-specific split, not three separately gauge-invariant observables. Their sum $11C_A$ is the gauge-sector result. Omitting the closed ghost loop changes that sum and violates the background Ward or Slavnov–Taylor identity. The determinant derivation and the matter weights are worked out in Srednicki 2007, § 78, pp. 465–471; the ordinary covariant-gauge diagram sum and renormalization check appear in Schwartz 2014, §§ 26.4–26.6, pp. 517–528.

Matter polarization resembles electric-charge screening: more charged species push $b_0$ downward. The non-Abelian gauge sector has the opposite net sign because the self-interacting spin-one field responds to the background curvature. “Antiscreening” is a useful summary of the total gauge-sector effect, not a gauge-invariant assignment to one diagram or a classical medium picture.

Gross and Wilczek and, independently, Politzer identified the negative beta function and its short-distance consequence in Gross and Wilczek 1973, pp. 1343–1346 and Politzer 1973, pp. 1346–1349.

Running and dimensional transmutation

At one loop,

$$\frac{\mathrm d}{\mathrm d\ln\mu}\frac1{g^2} =\frac{b_0}{8\pi^2},$$

so

$$\frac1{g^2(\mu)} =\frac1{g^2(\mu_0)} +\frac{b_0}{8\pi^2}\ln\frac{\mu}{\mu_0}.$$

For $b_0>0$, $g(\mu)\to0$ as $\mu\to\infty$. The same solution can be written

$$g^2(\mu)=\frac{8\pi^2}{b_0\ln(\mu/\Lambda)}, \qquad \Lambda =\mu\exp\left[-\frac{8\pi^2}{b_0g^2(\mu)}\right]$$

at one-loop accuracy. A dimensionless input coupling has been traded for the RG-invariant scale $\Lambda$; this is dimensional transmutation. The formula is reliable only where $g^2/(16\pi^2)\ll1$ and logarithms are controlled. The apparent pole at $\mu=\Lambda$ marks the failure of the one-loop weak-coupling solution, not a perturbative prediction of a physical singularity.

The SU(N) matter boundary

For $n_f$ Dirac fermions in the $SU(N)$ fundamental and no scalars,

$$b_0=\frac{11}{3}N-\frac{2}{3}n_f.$$

Thus one-loop asymptotic freedom requires

$$n_f<\frac{11}{2}N.$$

For $SU(3)$,

$$b_0=11-\frac23n_f, \qquad b_0(n_f=6)=7, \qquad b_0(n_f=17)=-\frac13.$$

The largest integer $n_f$ with positive $b_0$ is therefore $16$. At $b_0=0$, if a matter content permits it, the one-loop term is inconclusive and the higher-order beta function controls the first nonzero flow. A negative $b_0$ means the Gaussian point is not ultraviolet-attractive in this one-loop gauge coupling; it does not by itself rule out another ultraviolet completion.

What the sign does and does not say

A positive $b_0$ establishes a controlled weak-coupling ultraviolet regime: at fixed momentum ratios, sufficiently short-distance quantities admit an expansion in a small running coupling, subject to the usual infrared-safe observable definition. It also predicts logarithmic scaling violations and fixes the leading relation between $g(\mu)$ and $\Lambda$.

It does not establish confinement, a mass gap, chiral symmetry breaking, a spectrum of hadrons, or the infrared endpoint of the flow. Running toward larger $g$ only says that the ultraviolet expansion eventually loses control. Nonperturbative input is required beyond that point. Nor is a gauge-dependent “effective charge” extracted from an arbitrary Green function automatically the physical running coupling; a scheme and observable definition must be supplied.

Independent checks and limits

1. Sign check: differentiating $1/g^2$ must give $+b_0/(8\pi^2)$. Hence $b_0>0$ makes $g$ decrease, not increase, with $\mu$.
2. Group check: pure $SU(N)$ gives $b_0=11N/3$; one fundamental Dirac field reduces it by $2/3$ because $T(F)=1/2$.
3. Ghost check: the gauge-sector sum must be gauge-parameter independent and transverse. A result obtained from gauge loops alone is incomplete.
4. Scale check: differentiating $\Lambda=\mu e^{-8\pi^2/(b_0g^2)}$ with the one-loop beta function gives zero up to higher-order terms.

Common pitfalls

Calling every negative loop contribution screening. The gauge/ghost split depends on gauge and organization. Only the complete $b_0$ has the universal one-loop meaning stated here.

Forgetting representation multiplicities. A field charged under a product group appears multiple times from the viewpoint of one factor. Include the dimension of the spectator representation in its $T(R)$ sum.

Inferring infrared physics from the ultraviolet sign. Asymptotic freedom is a short-distance statement. It is compatible with several infrared behaviors and is not a proof of confinement.

References

• David J. Gross and Frank Wilczek, “Ultraviolet Behavior of Non-Abelian Gauge Theories,” Physical Review Letters 30 (1973), 1343–1346, DOI.
• H. David Politzer, “Reliable Perturbative Results for Strong Interactions?” Physical Review Letters 30 (1973), 1346–1349, DOI.
• Matthew D. Schwartz, Quantum Field Theory and the Standard Model, Cambridge University Press (2014), Chapter 26, DOI.
• Mark Srednicki, Quantum Field Theory, Cambridge University Press (2007), §§ 73 and 78, DOI.