Running and Dynamical Scales in Gauge Theories

In four-dimensional gauge theory the classical coupling $g$ is dimensionless, but quantum fluctuations make it scale dependent. When the one-loop coefficient $b_0$ is positive, integrating the beta function replaces the boundary value $g(\mu_0)$ by a dimensionful parameter $\Lambda$. This dimensional transmutation is an exact change of RG coordinates at the stated perturbative order; it signals where weak coupling fails, but it does not by itself prove confinement, a mass gap, or any particular infrared phase.

Required background. Running couplings and dimensional transmutation supplies the renormalization-group equation and its interpretation.

Helpful background. Scheme transformations and RG invariants supplies the distinction between an invariant trajectory and a scheme-independent numerical parameter.

One-loop gauge running

Define

$$\beta(g)\equiv\mu\frac{dg}{d\mu} =-\frac{b_0}{16\pi^2}g^3+O(g^5).$$

For a compact simple gauge group with Dirac fermions and complex scalars,

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

A real scalar contributes half the complex-scalar term. The group factors obey $f^{acd}f^{bcd}=C_A\delta^{ab}$ and $\operatorname{tr}_{R}(T^aT^b)=T(R)\delta^{ab}$.

The coefficient can be checked against its physical and diagrammatic sources.

Fluctuation	Contribution to $b_0$	Effect
Gauge and ghost fields	$+\tfrac{11}{3}C_A$	Non-Abelian antiscreening
Dirac fermion in $R_f$	$-\tfrac{4}{3}T(R_f)$	Screening
Complex scalar in $R_s$	$-\tfrac{1}{3}T(R_s)$	Screening

Background-field gauge makes the comparison especially direct: gauge invariance forces the vacuum-polarization tensor to be transverse, and the coefficient of its logarithmic divergence determines the coupling counterterm. Gauge, ghost, fermion, and scalar diagrams separately depend on the chosen organization, but their sum gives the gauge-independent $b_0$. The pure-gauge and fermion calculation is developed in Schwartz 2014, §§26.4–26.6, pp. 517–528.

For $SU(N)$ with $n_f$ Dirac fundamentals and no scalars, $C_A=N$ and $T(F)=\tfrac12$, so

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

Thus $n_f<11N/2$ gives one-loop asymptotic freedom. At equality the one-loop term vanishes and the next coefficients decide the local flow; setting $b_0=0$ is not evidence for an exactly marginal coupling.

Integrating the flow and generating Λ

At one loop,

$$\frac{d}{d\ln\mu}\frac1{g^2} =-\frac{2}{g^3}\beta(g) =\frac{b_0}{8\pi^2}.$$

Integration between $\mu_0$ and $\mu$ gives

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

For $b_0>0$, define

$$\Lambda_{\rm scheme} \equiv \mu\exp\!\left[-\frac{8\pi^2}{b_0g^2(\mu)}\right].$$

Solving for the coupling yields

$$g^2(\mu)= \frac{8\pi^2}{b_0\ln(\mu/\Lambda_{\rm scheme})},$$

which is reliable only while the omitted higher-order terms are small. The promised RG check is explicit:

$$\frac{d\ln\Lambda_{\rm scheme}}{d\ln\mu} =1+\frac{16\pi^2}{b_0g^3}\beta(g) =0+O(g^2).$$

The arbitrary subtraction scale $\mu$ has disappeared in favor of $\Lambda$. A classically dimensionless parameter has been traded for a dimensionful integration constant. In a massless theory for which $\Lambda$ is the only generated scale, physical masses, if generated, take the form

$$M_i=c_i^{\rm scheme}\Lambda_{\rm scheme},$$

where the nonperturbative coefficient compensates the scheme dependence of $\Lambda$.

As a concrete group-theory check, six-flavor $SU(3)$ has

$$b_0=11-\frac{2}{3}(6)=7, \qquad \Lambda=mu_0\exp\!\left[-\frac{8\pi^2}{7g^2(\mu_0)}\right]$$

at one loop. The exponential sensitivity is why a modest ultraviolet coupling can generate a much smaller infrared scale.

Scheme changes and thresholds

A finite coupling redefinition

$$g'=g+a g^3+O(g^5)$$

leaves $b_0$ unchanged but gives

$$\frac1{g'^2}=\frac1{g^2}-2a+O(g^2), \qquad \Lambda'=\Lambda\exp\!\left(\frac{16\pi^2a}{b_0}\right).$$

Therefore “$\Lambda$ is RG invariant” means it is independent of the sliding scale within a declared scheme. Its numerical value is not scheme independent. Ratios of physical observables are scheme independent after the matching coefficients are transformed consistently.

Massive particles add a second qualification. A mass-independent subtraction scheme does not automatically remove a field when $\mu$ drops below its mass $M$. One constructs a low-energy effective theory, matches its gauge coupling to the high-energy coupling near $M$, and runs with the new coefficient. At leading matching order,

$$g_{\rm low}(M)=g_{\rm high}(M),$$

so

$$b_{\rm high}\ln\frac{M}{\Lambda_{\rm high}} =b_{\rm low}\ln\frac{M}{\Lambda_{\rm low}}.$$

Finite matching corrections modify this relation at higher orders. Running straight through a heavy threshold with one fixed $b_0$ creates spurious logarithms and the wrong transmuted scale.

Question	Invariant statement	Necessary qualification
Does $g$ decrease toward the ultraviolet?	Yes at sufficiently weak coupling if $b_0>0$	Higher loops and thresholds alter the quantitative trajectory
Is $\Lambda$ independent of $\mu$?	Yes to the computed RG order	Its normalization depends on scheme
Is a mass scale generated?	The RG trajectory contains a dimensionful invariant	A physical mass requires nonperturbative or other infrared dynamics
Does the coupling grow near $\Lambda$?	The one-loop expression becomes strong	The pole is outside perturbative control, not a literal observable singularity
Does asymptotic freedom imply confinement?	No	Infrared phase diagnostics must be supplied independently

What running can and cannot establish

For $b_0>0$, the ultraviolet statement is controlled: at sufficiently large $\mu$, $g(\mu)$ is small and the theory is asymptotically free. Evolving downward, the one-loop formula identifies the scale at which its own expansion fails. It does not determine whether the infrared theory confines, flows to an interacting fixed point, develops a Higgs regime, breaks chiral symmetry, or remains gapless.

Indeed, changing matter content can produce a perturbative infrared fixed point when the one-loop and two-loop terms balance. The relevant continuation is ultraviolet and infrared fixed points in QFT. Conversely, evidence for a mass gap or string tension must come from gauge-invariant spectral, line, semiclassical, or lattice observables.

Independent checks and failure diagnoses

• Sign check: for pure Yang–Mills, $b_0=11C_A/3>0$, so $g$ must decrease as $\mu$ increases.
• Group-factor check: each Dirac fundamental of $SU(N)$ lowers $b_0$ by $2/3$, independent of $N$ with the stated trace normalization.
• Invariant check: differentiate the proposed $\Lambda$ using the same beta-function truncation; residual terms should begin beyond that order.
• Scheme check: apply a finite $g\mapsto g'$ transformation and verify that a physical prediction is unchanged after matching.
• Threshold check: compute the running piecewise and vary the matching scale near $M$; the residual variation estimates missing higher orders.
• Domain check: require $b_0g^2/(16\pi^2)\ll1$ before trusting a one-loop numerical prediction.

Common pitfalls

Reversing the logarithm. With $b_0>0$, $1/g^2$ grows with $\ln\mu$. A formula predicting a larger coupling in the ultraviolet has a sign error.

Calling the one-loop pole a physical particle or transition. It marks the breakdown of the approximation. Infrared observables require other methods.

Treating $\Lambda$ as scheme independent. Its RG invariance and its scheme dependence are compatible. Only matched physical quantities are scheme independent.

Ignoring decoupling. Heavy fields still appear in mass-independent beta functions until an effective theory is matched. Use the coefficient appropriate to each energy interval.

Continue to theta dependence in Yang–Mills and QCD for the independent topological parameter. Return to line-operator diagnostics before turning strong running into a claim about confinement.

References

• Schwartz, Matthew D. Quantum Field Theory and the Standard Model. Cambridge University Press, 2014, §§26.4–26.6, pp. 517–528. DOI.