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. A positive coefficient makes the gauge coupling ultraviolet-attractive near the Gaussian point; if other couplings are present, their flow must also remain under control.

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. Majorana fermions and real scalars require a compatible real representation, and a chiral Weyl matter list must define an anomaly-free gauge theory. For product groups this formula is applied separately to each simple factor, using the multiplicities supplied by the other factors. The trace normalization is $\operatorname{tr}_R(T^aT^b)=T(R)\delta^{ab}$ , with $T(\mathrm{adj})=C_A$ .

The remainder $O(g^5)$ denotes the single-gauge-coupling expansion. With independent Yukawa, scalar, or additional gauge couplings, read it as higher-loop terms in the coupled system, not a bound in $g$ alone. The displayed one-loop coefficient is unchanged. In a mass-independent subtraction scheme, the full-theory coefficient remains independent of particle masses even below a threshold. For a low-energy description with heavy species removed, match to the appropriate effective theory; the short-distance active-species interpretation applies above their masses.

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 ghost determinant gives a different coefficient. The diagrammatic weights, including the factor for a real vector and the closed-ghost-loop sign, are displayed in Srednicki 2006, §78, pp.469–471, Eqs.(78.28)–(78.36), author-manuscript PDF. The local calculation below explains the same weights without assigning physical meaning to an individual diagram.

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.

Extracting the local ultraviolet coefficient

For this calculation only, continue to flat Euclidean space and use background Feynman gauge. Take a smooth background with periodic boundary conditions or sufficient falloff to discard integrated total derivatives. We need the small-proper-time coefficient, not the complete determinant or a stable chromomagnetic vacuum; a finite upper proper-time cutoff separates this ultraviolet calculation from zero modes and infrared instabilities. The background-field construction gives the operators

\begin{aligned} \Omega_{\mu\nu}&=[\bar D_\mu,\bar D_\nu] =-ig\bar F^a_{\mu\nu}T^a_{\mathrm{adj}},\\ (P_1)_{\mu\nu}&=-\bar D^2\delta_{\mu\nu}-2\Omega_{\mu\nu}, \qquad P_0=-\bar D^2. \end{aligned}

For a Laplace-type operator $P=-(\bar D^2+E)$ , the four-dimensional curvature coefficient, with its common $(4\pi)^{-2}$ factor suppressed and integrated total derivatives discarded, is

h_4(P)=\operatorname{tr}\left( \frac12E^2+\frac1{12}\Omega_{\mu\nu}\Omega_{\mu\nu} \right).

Here $\operatorname{tr}$ acts on the finite internal indices, and a total derivative has been integrated out. This local expansion is the mathematical input, rather than a claim about the finite nonlocal effective action: Vassilevich 2003, §4.1, pp.39–40, Eqs.(4.25), (4.28), preprint PDF. The $E^2/2$ term also follows by expanding $e^{sE}$ for a constant commuting $E$ .

Set $\mathcal Q=g^2C_A\bar F^a_{\mu\nu}\bar F^a_{\mu\nu}$ . Anti-Hermiticity of the connection gives $\operatorname{tr}_{\mathrm{adj}}\Omega^2=-\mathcal Q$ . The vector connection has four identical components, whereas $(E_1)_{\mu\nu}=2\Omega_{\mu\nu}$ mixes them. Therefore

\begin{aligned} \operatorname{tr}_{1,\mathrm{adj}}\Omega^2&=-4\mathcal Q,\\ \operatorname{tr}_{1,\mathrm{adj}}E_1^2 &=4\sum_{\mu,\nu}\operatorname{tr}_{\mathrm{adj}} \Omega_{\mu\nu}\Omega_{\nu\mu}=4\mathcal Q,\\ h_4(P_1)&=-\frac{\mathcal Q}{3}+2\mathcal Q =\frac53\mathcal Q,\qquad h_4(P_0)=-\frac{\mathcal Q}{12}. \end{aligned}

The positive sign in the second line uses $\Omega_{\nu\mu}=-\Omega_{\mu\nu}$ . The real-vector Gaussian has weight $1/2$ , while the independent ghost–antighost pair has weight $-1$ . Thus

H_4=\frac12h_4(P_1)-h_4(P_0) =\underbrace{-\frac{\mathcal Q}{6}}_{\text{vector kinetic}} +\underbrace{\mathcal Q}_{\text{vector spin}} +\underbrace{\frac{\mathcal Q}{12}}_{\text{ghost}} =\frac{11}{12}\mathcal Q.

These traces agree with Vassilevich 2003, §4.2.1, pp.40–41, Eqs.(4.30)–(4.34), preprint PDF; his total heat coefficient is twice this Gaussian-weighted one. The sign of the effective-action pole follows from $\operatorname{Tr}\ln P=-\int ds\,\operatorname{Tr}e^{-sP}/s$ . In $d=4-2\epsilon$ , its curvature term and the counterterm that cancels it are

\begin{aligned} \Gamma^{(1)}_{\mathrm{div}} &=-\frac1{16\pi^2\epsilon}\int d^dx\,H_4,\\ \delta\mathcal L_{\mathrm E} &=\frac14\delta Z_{\bar A}\bar F^a_{\mu\nu}\bar F^a_{\mu\nu}, \qquad \delta Z_{\bar A}=\frac{11C_Ag^2}{48\pi^2\epsilon}. \end{aligned}

Only the pole is displayed: the dimensionally continued coefficient and scale factors change finite terms, which are not needed here. Background covariance fixes $Z_gZ_{\bar A}^{1/2}=1$ , so $Z_g=1-b_0g^2/(32\pi^2\epsilon)+\cdots$ with $b_0=11C_A/3$ . Differentiating $g_0=\mu^\epsilon Z_gg$ at fixed bare coupling, including $\beta_d=-\epsilon g+\cdots$ , gives $\beta(g)=-b_0g^3/(16\pi^2)$ at one loop. The kinetic factor $1/4$ is what converts $11/12$ into $11/3$ .

The same coefficient explains the matter weights. For one complex scalar, $E=0$ , $\operatorname{tr}_R\Omega^2=-\mathcal Q_R$ , and the determinant weight is $+1$ , where $\mathcal Q_R=g^2T(R)\bar F^2$ . For a Dirac field, square its Euclidean Dirac operator: $E=\gamma^{\mu\nu}\Omega_{\mu\nu}/2$ , with $\gamma^{\mu\nu}=[\gamma^\mu,\gamma^\nu]/2$ . The spin trace gives $\operatorname{tr}\Omega^2=-4\mathcal Q_R$ and $\operatorname{tr}E^2=2\mathcal Q_R$ . Its Grassmann determinant has weight $-1/2$ after squaring. Consequently

H_4^{\mathrm{complex\ scalar}}=-\frac{\mathcal Q_R}{12}, \qquad H_4^{\mathrm{Dirac}}=-\frac12 \left(-\frac{\mathcal Q_R}{3}+\mathcal Q_R\right) =-\frac{\mathcal Q_R}{3}.

Multiplication by four gives $-T(R)/3$ and $-4T(R)/3$ in $b_0$ . Taking half as many real or spinor degrees of freedom gives the stated real-scalar and Weyl weights. These are ultraviolet coefficients, so masses do not change them in the mass-independent scheme; decoupling is a separate matching step. The QFT II calculation develops the proper-time normalization and additional controls.

Running and dimensional transmutation

At one loop,

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

\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. At large rank or matter multiplicity, the corresponding group- and multiplicity-enhanced loop parameters must also be small. 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 ultraviolet attraction in the gauge-coupling direction. In pure Yang–Mills theory, or on a trajectory where all other couplings also remain perturbatively controlled, sufficiently short-distance quantities at fixed momentum ratios admit a weak-coupling expansion, subject to an infrared-safe observable definition. With Yukawa or scalar self-couplings, the sign of $b_0$ alone does not establish a complete ultraviolet trajectory or stability of the scalar potential. The gauge running still predicts its leading logarithmic scaling violations and the one-loop 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

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

Exercise: spectator multiplicities

Consider an $SU(3)\times SU(2)$ theory with one Dirac multiplet in $(\mathbf3,\mathbf2)$ and one complex scalar in $(\mathbf3,\mathbf1)$ . Use $T(\mathbf N)=1/2$ for the fundamental of $SU(N)$ . Find the one-loop coefficient for each gauge factor. Does positivity of both coefficients prove that every coupling is asymptotically free?

Solution

From the viewpoint of $SU(3)$ , the Dirac multiplet supplies two fundamental copies and the scalar supplies one. From the viewpoint of $SU(2)$ , there are three fundamental Dirac copies and the scalar is neutral. Thus

\begin{aligned} b_0^{SU(3)} &=11-\frac43\left(2\cdot\frac12\right) -\frac13\left(\frac12\right)=\frac{19}{2},\\ b_0^{SU(2)} &=\frac{22}{3}-\frac43\left(3\cdot\frac12\right) =\frac{16}{3}. \end{aligned}

Both gauge directions are ultraviolet-attractive at this order. The calculation says nothing sufficient about the scalar self-couplings or any allowed Yukawa couplings; those require their coupled beta functions and a stable scalar potential. Treating the Dirac multiplet as one copy for each factor would miss the spectator dimensions.

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.
Mark Srednicki, Quantum Field Theory, prepublication author manuscript (2006), §78. Author’s draft and published-edition errata. Open PDF. Page locators above refer to this manuscript.
Dmitri V. Vassilevich, “Heat kernel expansion: user’s manual,” Physics Reports 388 (2003), 279–360. DOI. Open PDF, arXiv:hep-th/0306138v3. Page locators above use the preprint’s printed pagination.