QCD running and asymptotic freedom

For color group $SU(3)$ with $n_f$ active quark flavors, define $\alpha_s=g_s^2/(4\pi)$. We write the one-loop beta function using a derivative with respect to $\log\mu$, with coefficient $\beta_0=11-2n_f/3$. A formula differentiated with respect to $\log\mu^2$ has half this derivative. Both conventions are common; mixing their logarithms without adjusting the coefficient creates a factor-of-two error even when the sign and dimensions look correct.

The positive gauge contribution to $\beta_0$ competes with the quark contribution. For the ordinary flavor counts relevant to QCD, $\beta_0>0$, so the derivative of a small positive $\alpha_s$ is negative. Increasing the scale then decreases the coupling. This is asymptotic freedom. It depends on non-Abelian gauge dynamics rather than simply on naming the color charge, and a theory with sufficiently many matter fields can lose that leading sign.

At one loop, the inverse coupling is linear in the logarithm of scale. Integrate between two scales that are both perturbative and do not cross a heavy-quark matching threshold. The denominator grows for increasing scale. Higher-loop terms and finite scheme definitions change the quantitative evolution, so a one-loop teaching calculation should not be reported as a precision determination of the coupling or as a current world-average value.

A heavy quark is not handled by changing $n_f$ at an arbitrary point without further thought. Effective descriptions with different active-flavor content are matched near the heavy mass using a stated prescription and perturbative order. The matching scale is an intermediate choice; complete physical predictions should depend on it only through higher-order omissions. Physical thresholds and the subtraction scale also need not be identical, even though choosing a scale near the process's hard momentum often avoids large logarithms.

Introducing a scale $\Lambda$ rewrites the solution as $\alpha_s(\mu)=2\pi/[\beta_0\log(\mu/\Lambda)]$ at this order. The formal pole near $\Lambda$ signals that the small-coupling expansion has become unusable. It does not predict an experimentally infinite scattering probability there. The numerical value assigned to a QCD scale parameter depends on the subtraction convention and flavor content, so its symbol alone is insufficient to compare two quoted results.

Long-distance hadron physics requires additional methods, including lattice calculations, symmetry arguments, effective theories, and fitted nonperturbative inputs. Asymptotic freedom supports short-distance expansions but is not by itself a proof of confinement or a complete derivation of the hadron spectrum. A high-energy collider prediction also contains parton distributions and may contain hadronization corrections. A small hard coupling does not turn the entire incoming proton or final-state event into a weakly interacting elementary object.