---
title: Asymptotic Freedom and Confinement
module: Quantum Chromodynamics
moduleNumber: 8
lessonNumber: 2
order: 802
summary: >
  The QCD beta function is negative: gluon self-interaction antiscreens color, so the
  coupling weakens at short distance (asymptotic freedom) and strengthens at long
  distance (confinement). This lesson computes the one-loop beta coefficient, solves
  for the running of alpha_s and the emergent scale Lambda_QCD, and reads the
  strong-coupling regime as the linear quark-antiquark potential of a color flux tube
  that breaks by pair creation.
topics: [Quantum Chromodynamics]
draft: false
sources:
  - book: Griffiths
    ref: "Ch. 9 — Quantum Chromodynamics, §9.3–9.4"
  - book: Halzen & Martin
    ref: "Ch. 15 (asymptotic freedom, the running coupling, and confinement)"
  - book: Thomson
    ref: "Ch. 10 §10.5 (the running of alpha_s)"
  - book: Tong
    ref: "The Standard Model (Cambridge Part III), §3.1 (asymptotic freedom, the QCD scale, the flux tube)"
---

The QED coupling grows with energy: a bare charge is screened by virtual
electron-positron pairs, so probing at shorter distance penetrates the screening
cloud and sees a larger effective charge, as derived for the
[running of $\alpha$](/particle-physics/qed/renormalization-running-coupling). QCD
does the opposite. The [gluon self-coupling](/particle-physics/qcd/color-su3-gluons)
adds a contribution of the opposite sign that overwhelms quark screening, and the
effective color charge falls at short distance. This lesson establishes that sign
through the one-loop beta function, follows the coupling down to the scale where it
diverges, and interprets the strong-coupling end as confinement.

Throughout, $\hbar = c = 1$, $\alpha_s = g_s^2/4\pi$, and $N_c = 3$ colors with $n_f$
active quark flavors.

## The one-loop beta function

The coupling depends on the scale $\mu$ at which the theory is probed. The dependence
is governed by the beta function, defined as the logarithmic derivative of the
coupling,

$$
\beta(g_s) \equiv \mu\,\frac{\d g_s}{\d\mu}
  = -\frac{b_0}{(4\pi)^2}\,g_s^3 + \mathcal O(g_s^5),
$$

with the one-loop coefficient[^tong-beta]

$$
b_0 = \frac{11}{3}\,C_A - \frac{4}{3}\,T_F\,n_f
    = 11 - \frac{2}{3}\,n_f
\qquad (C_A = 3,\ T_F = \tfrac12).
$$

The two terms have opposite physical origins.

- The $-\tfrac43 T_F n_f$ term comes from **quark loops** on the gluon propagator.
  Like the electron loop in QED, a virtual $q\bar q$ pair screens the color charge and
  drives the coupling up at short distance. Alone, it would give a positive beta
  function, as in QED.
- The $+\tfrac{11}{3}C_A$ term comes from **gluon loops**, available only because the
  gluon carries color. It antiscreens: the cloud of virtual gluons spreads color
  charge outward rather than screening it, so a short-distance probe sees less charge.
  This term has no QED counterpart and it dominates.

For QCD with $n_f \le 6$, the gluon term wins and $b_0 > 0$, so the beta function is
negative. The condition for the antiscreening to survive is

$$
n_f < \frac{11}{2}\,N_c = \frac{33}{2} = 16.5,
$$

comfortably satisfied by the six quark flavors of nature.

$$
% caption: The two contributions to the QCD beta function. A virtual quark loop on the
% gluon line screens color, pushing the coupling up at short distance, exactly as in
% QED. The virtual gluon loop, absent in QED, antiscreens and pushes the coupling
% down. The gluon term is larger, so the net coupling falls with energy.
\begin{tikzpicture}[>=stealth, font=\footnotesize, scale=1.0]
  \definecolor{acc}{HTML}{4A6FA5}
  % --- quark loop (left) ---
  \begin{scope}[shift={(-3.6,0)}]
    \draw[thick] (-1.7,0) sin (-1.5,0.12) cos (-1.3,0) sin (-1.1,-0.12) cos (-0.9,0);
    \draw[thick] (0.9,0) sin (1.1,0.12) cos (1.3,0) sin (1.5,-0.12) cos (1.7,0);
    \draw[thick] (-0.9,0) arc (180:0:0.9 and 0.6);
    \draw[thick] (-0.9,0) arc (180:360:0.9 and 0.6);
    \draw[->] (-0.2,0.6) -- (0.2,0.6);
    \draw[->] (0.2,-0.6) -- (-0.2,-0.6);
    \node[black, font=\scriptsize] at (0,-1.3) {quark loop: screening};
    \node[black, font=\scriptsize] at (0,-1.8) {coupling up (as in QED)};
  \end{scope}
  % --- gluon loop (right) ---
  \begin{scope}[shift={(3.6,0)}]
    \draw[acc, thick] (-1.7,0) sin (-1.5,0.12) cos (-1.3,0) sin (-1.1,-0.12) cos (-0.9,0);
    \draw[acc, thick] (0.9,0) sin (1.1,0.12) cos (1.3,0) sin (1.5,-0.12) cos (1.7,0);
    \draw[acc, thick] (-0.9,0)
      sin (-0.75,0.25) cos (-0.55,0.5) sin (-0.3,0.62) cos (0,0.6)
      sin (0.3,0.62) cos (0.55,0.5) sin (0.75,0.25) cos (0.9,0);
    \draw[acc, thick] (-0.9,0)
      sin (-0.75,-0.25) cos (-0.55,-0.5) sin (-0.3,-0.62) cos (0,-0.6)
      sin (0.3,-0.62) cos (0.55,-0.5) sin (0.75,-0.25) cos (0.9,0);
    \node[black, font=\scriptsize] at (0,-1.3) {gluon loop: antiscreening};
    \node[black, font=\scriptsize] at (0,-1.8) {coupling down (QCD only)};
  \end{scope}
\end{tikzpicture}
$$

## The running coupling and the QCD scale

Solving $\beta(g_s) = -b_0 g_s^3/(4\pi)^2$ and rewriting in terms of $\alpha_s$ gives
the one-loop running

$$
\alpha_s(Q^2) = \frac{\alpha_s(\mu^2)}
  {1 + \dfrac{b_0}{4\pi}\,\alpha_s(\mu^2)\,\ln\!\dfrac{Q^2}{\mu^2}} .
$$

The coupling decreases logarithmically as $Q^2$ increases — asymptotic freedom made
quantitative. It is convenient to absorb the reference scale into a single dimensionful
constant $\Lambda_{\text{QCD}}$, defined as the scale at which the extrapolated
one-loop coupling diverges. In terms of it,

$$
\alpha_s(Q^2) = \frac{4\pi}{b_0\,\ln\!\left(Q^2/\Lambda_{\text{QCD}}^2\right)}
  = \frac{12\pi}{(33 - 2 n_f)\,\ln\!\left(Q^2/\Lambda_{\text{QCD}}^2\right)} .
$$

A classical Lagrangian with massless quarks contains no scale at all; the quantum
theory manufactures one, $\Lambda_{\text{QCD}} \approx 200\text{–}300$ MeV, through
the running of the coupling. This is dimensional transmutation: the dimensionless bare
coupling is traded for a dimensionful scale.[^tong-lambda] For $Q \gg
\Lambda_{\text{QCD}}$ the coupling is small and perturbation theory in $\alpha_s$
converges; for $Q \sim \Lambda_{\text{QCD}}$ it is of order one and perturbation theory
fails, which is precisely the hadronic scale.

The measured coupling runs from $\alpha_s \approx 0.12$ at the $Z$ mass down toward
$\alpha_s \approx 0.09$ at the TeV scale, and the world data over three orders of
magnitude in $Q$ lie on the single predicted curve — one of the most direct
confirmations of the theory.[^pdg-run]

$$
% caption: The measured strong coupling against momentum transfer, on a logarithmic
% energy axis. Points from tau decays, the Z width, and jet rates fall on the single
% asymptotic-freedom curve; the coupling drops from about 0.35 near 1 GeV to 0.118 at
% the Z mass. The extrapolated curve diverges at the QCD scale of a few hundred MeV.
\begin{tikzpicture}[font=\footnotesize, scale=1.0]
  \definecolor{acc}{HTML}{4A6FA5}
  \draw[->, black] (0,0) -- (10.4,0) node[right, font=\scriptsize] {energy (log)};
  \draw[->, black] (0,0) -- (0,4.4) node[above, font=\scriptsize] {$\alpha_s$};
  % y ticks
  \foreach \y in {1,2,3,4}
    \draw[black] (-0.1,\y) -- (0.1,\y);
  \foreach \y/\lab in {1/0.1, 2/0.2, 3/0.3, 4/0.4}
    \node[left, black, font=\scriptsize] at (-0.12,\y) {\lab};
  % running curve: alpha_s falls with log-energy. Approximate smooth decreasing curve.
  \draw[acc, very thick]
    (0.7,3.6) .. controls (2.2,2.3) and (3.4,1.7) .. (5.0,1.35)
    .. controls (6.6,1.05) and (8.4,0.85) .. (9.9,0.75);
  % marker points on the curve
  \fill[black] (1.3,2.75) circle (1.8pt);
  \fill[black] (2.6,1.95) circle (1.8pt);
  \fill[black] (5.0,1.35) circle (1.8pt);
  \fill[black] (7.4,0.98) circle (1.8pt);
  % x labels
  \node[black, font=\scriptsize] at (1.3,-0.4) {1 GeV};
  \node[black, font=\scriptsize] at (5.0,-0.4) {$M_Z$};
  \node[black, font=\scriptsize] at (9.0,-0.4) {1 TeV};
  % annotation
  \node[acc, font=\scriptsize, align=left] at (7.0,2.4) {weaker at\\ higher energy};
\end{tikzpicture}
$$

## Confinement and the color flux tube

Asymptotic freedom is a statement about short distances; confinement is its
long-distance counterpart. As the separation between a quark and an antiquark grows,
$\alpha_s$ grows, and the field between them does not spread out like a Coulomb field.
The gluon self-interaction pulls the chromoelectric field lines into a narrow **flux
tube** of roughly constant cross section, so the field energy grows linearly with
length rather than falling off. The static potential between a heavy quark and
antiquark is

$$
V(r) = -\frac{4}{3}\,\frac{\alpha_s}{r} + \sigma\,r ,
$$

a short-distance Coulomb term with the color factor $C_F = \tfrac43$, plus a linear
term with **string tension** $\sigma \approx 0.9$ GeV/fm $\approx (0.42\ \text{GeV})^2$
on dimensional grounds fixed by $\Lambda_{\text{QCD}}$.[^grif-conf] Separating the pair
costs energy without bound, so an isolated quark cannot be produced: the only
finite-energy states are color singlets. This is the dynamical content behind the
color-singlet rule stated in the [quark model](/particle-physics/quark-model/color-confinement-exotics).

$$
% caption: The contrast between electromagnetic and color fields. Two electric charges
% spread their field lines through all space, and the potential falls as one over r.
% A quark and antiquark instead compress the color field into a tube of fixed width,
% so its energy grows linearly with separation and the pair cannot be pulled apart.
\begin{tikzpicture}[>=stealth, font=\footnotesize, scale=1.0]
  \definecolor{acc}{HTML}{4A6FA5}
  % --- QED spreading (left) ---
  \begin{scope}[shift={(-3.6,0)}]
    \fill[black] (-1.1,0) circle (2.6pt);
    \fill[black] (1.1,0) circle (2.6pt);
    \foreach \a in {25,55,-25,-55}
      \draw[black] (-1.1,0) .. controls (0,{0.9*sin(\a)}) .. (1.1,0);
    \draw[black] (-1.1,0) -- (1.1,0);
    \node[black, font=\scriptsize] at (0,-1.6) {electric field: spreads};
    \node[black, font=\scriptsize] at (0,-2.1) {potential $\sim 1/r$};
  \end{scope}
  % --- QCD flux tube (right) ---
  \begin{scope}[shift={(3.6,0)}]
    \fill[black] (-1.3,0) circle (2.6pt) node[above, font=\scriptsize] {q};
    \fill[black] (1.3,0) circle (2.6pt) node[above, font=\scriptsize] {$\overline{q}$};
    \draw[acc, thick, fill=acc!12] (-1.3,-0.22) rectangle (1.3,0.22);
    \node[black, font=\scriptsize] at (0,-1.6) {color field: flux tube};
    \node[black, font=\scriptsize] at (0,-2.1) {potential $\sim \sigma r$};
  \end{scope}
\end{tikzpicture}
$$

## String breaking

The linear rise does not continue forever. Once the energy stored in the tube exceeds
twice the mass of the lightest quark, $\sigma r > 2 m_q$, it becomes energetically
favorable to create a new $q\bar q$ pair from the vacuum. The pair snaps the string,
and each end caps off into a color-singlet hadron. Pulling a quark out of a proton
does not liberate it; it produces a jet of hadrons, because the energy invested in
separation is converted into new quark-antiquark pairs.

$$
% caption: String breaking. As the quark and antiquark separate, the flux tube stores
% energy in proportion to its length. When that energy reaches twice the light-quark
% mass, the tube snaps by pulling a new pair from the vacuum, and two color-singlet
% mesons result instead of a free quark. This is why quarks are never seen in isolation.
\begin{tikzpicture}[>=stealth, font=\footnotesize, scale=1.0]
  \definecolor{acc}{HTML}{4A6FA5}
  % before
  \begin{scope}[shift={(0,1.3)}]
    \fill[black] (-2.3,0) circle (2.4pt) node[above, font=\scriptsize] {q};
    \fill[black] (2.3,0) circle (2.4pt) node[above, font=\scriptsize] {$\overline{q}$};
    \draw[thick, fill=black!8] (-2.3,-0.18) rectangle (2.3,0.18);
    \node[black, font=\scriptsize, right] at (2.7,0) {stretch};
  \end{scope}
  % after
  \begin{scope}[shift={(0,-0.9)}]
    \fill[black] (-2.3,0) circle (2.4pt) node[above, font=\scriptsize] {q};
    \fill[acc] (-0.5,0) circle (2.4pt) node[above, acc, font=\scriptsize] {$\overline{q}$};
    \fill[acc] (0.5,0) circle (2.4pt) node[above, acc, font=\scriptsize] {q};
    \fill[black] (2.3,0) circle (2.4pt) node[above, font=\scriptsize] {$\overline{q}$};
    \draw[thick, fill=black!8] (-2.3,-0.18) rectangle (-0.5,0.18);
    \draw[thick, fill=black!8] (0.5,-0.18) rectangle (2.3,0.18);
    \node[black, font=\scriptsize, right] at (2.7,0) {two mesons};
    \node[acc, font=\scriptsize] at (0,-0.55) {new pair};
  \end{scope}
\end{tikzpicture}
$$

## Two regimes, one theory

The negative beta function ties the two faces of QCD together. At high energy the
coupling is small: quarks and gluons are nearly free, and cross sections are computed
in perturbation theory, the domain of [deep inelastic
scattering](/particle-physics/qcd/deep-inelastic-scattering-partons) and
[jets](/particle-physics/qcd/jets-hadronization). At low energy the coupling is large:
color is confined, the spectrum is hadrons, and quantitative predictions require
lattice methods rather than diagrams. The crossover is the single scale
$\Lambda_{\text{QCD}}$, generated from a classically scale-free theory by the running
of the coupling.

The following table contrasts the two limits.

| Regime | Distance | $\alpha_s$ | Degrees of freedom | Method |
| --- | --- | --- | --- | --- |
| Asymptotic freedom | $r \ll \Lambda_{\text{QCD}}^{-1}$ | small | quarks, gluons | perturbation theory |
| Confinement | $r \gg \Lambda_{\text{QCD}}^{-1}$ | large | hadrons | lattice / models |

[^tong-beta]: The one-loop beta function $\beta(g) = -b_0 g^3/(4\pi)^2$ with $b_0 = \tfrac{11}{3}N_c - \tfrac23 n_f$, and the asymptotic-freedom bound $n_f < \tfrac{11}{2}N_c$, are given in Tong, _The Standard Model_ (Cambridge Part III), §3.1.1, [damtp.cam.ac.uk/user/tong/standardmodel.html](http://www.damtp.cam.ac.uk/user/tong/standardmodel.html); see also Griffiths, §9.3, and Halzen & Martin, Ch. 15. The 2004 Nobel Prize recognized Gross, Wilczek, and Politzer for the 1973 discovery of asymptotic freedom.

[^tong-lambda]: The definition of $\Lambda_{\text{QCD}}$ through the divergence of the one-loop coupling, and the interpretation as dimensional transmutation, follow Tong, §3.1.1. The value $\Lambda_{\text{QCD}} \approx 200$–$300$ MeV (scheme- and $n_f$-dependent) is from the Particle Data Group, [pdg.lbl.gov](https://pdg.lbl.gov).

[^grif-conf]: The Cornell potential $V(r) = -\tfrac43\alpha_s/r + \sigma r$ and the string tension $\sigma \approx 0.9$ GeV/fm are discussed in Griffiths, §9.4, and Halzen & Martin, Ch. 15. Numerical values from the Particle Data Group, [pdg.lbl.gov](https://pdg.lbl.gov).

[^pdg-run]: The compilation of $\alpha_s(Q)$ measurements across energy scales and the world average $\alpha_s(M_Z^2) = 0.1180 \pm 0.0009$ are from the Particle Data Group review of QCD, [pdg.lbl.gov](https://pdg.lbl.gov).