QCD vertices & Feynman rules

Quantum electrodynamics is built from a single three-line vertex: one photon, two fermions, the factor $-iQe\gamma^\mu$ . Quantum chromodynamics inherits that template and then breaks its most comfortable assumption. The quark–gluon vertex looks almost identical — $-\tfrac12 ig_S\lambda^a_{ji}\gamma^\mu$ — differing only by a Gell-Mann matrix element $\lambda^a_{ji}$ threaded through the coupling, a pure number recording which colour the quark carried in and which it carries out. But because the eight gluons themselves carry colour, the interaction cannot stop there: gluons couple to gluons. QCD therefore possesses two vertices with no analogue anywhere in QED — a three-gluon vertex proportional to the structure constants $g_S f^{abc}$ , and a four-gluon vertex proportional to $g_S^2$ . That gluon self-coupling is the whole personality of the strong force: it is why the colour field between two quarks collapses into a narrow flux tube instead of spreading like a Coulomb field, why quarks are confined, and why the coupling weakens at short distance — asymptotic freedom, the property that lets perturbation theory work at all inside a proton. This article is the assembly manual for that machinery. It takes the colour algebra from aq-pp-colour-qcd, the eight gluons from aq-pp-gluons, the QED template from aq-pp-feynman-rules-qed, and the Dirac spinors from aq-pp-dirac-solutions, and forges the full set of QCD Feynman rules — vertices, propagator, and colour factors — behind every strong-interaction cross-section.

Key takeaways

The quark–gluon vertex is $-\tfrac12 ig_S\lambda^a_{ji}\gamma^\mu = -ig_S\,(T^a)_{ji}\,\gamma^\mu$ : the QED vertex with the charge replaced by the strong coupling $g_S$ and dressed by a colour matrix $T^a=\lambda^a/2$ .
The vertex factorises: a colour part $(T^a)_{ji}$ acting on the 3-component colour state, times the ordinary Dirac current $\bar u\gamma^\mu u$ acting on the spinor — the two live in different spaces.
QCD adds two vertices absent from QED — a three-gluon vertex $\propto g_S f^{abc}$ and a four-gluon vertex $\propto g_S^2$ — because gluons carry colour and couple to themselves.
The gluon propagator is $-ig_{\mu\nu}\delta_{ab}/q^2$ ; the $\delta_{ab}$ conserves the gluon type along the line, while the off-diagonal $\lambda^a_{ji}$ at each vertex is what changes a quark's colour.
Colour sums collapse to the SU(3) Casimirs $C_F=4/3$ (quark line), $C_A=3$ (gluon line), $T_F=1/2$ (trace); gluon self-coupling then makes $\alpha_S$ run down at high energy — $\alpha_S(M_Z)=0.1179$ , versus $\alpha_S\sim1$ at the proton scale — which is asymptotic freedom.
Key idea: the QCD vertex is the QED vertex times a colour matrix, but because gluons carry colour they couple to themselves — the three- and four-gluon vertices are the entire origin of confinement and asymptotic freedom.

The core insight

The quark–gluon interaction separates into a colour part (a Gell-Mann matrix element $\lambda^a_{ji}$ ) and a Dirac part ( $\gamma^\mu$ ), so the vertex is the QED factor times a colour number; but because the gluons themselves carry colour, QCD adds three- and four-gluon self-interaction vertices with no QED analogue — the source of confinement and asymptotic freedom.

Where this lives today

The QCD vertices are executed billions of times a second in the analysis of every Large Hadron Collider event. Automated tools — MadGraph, FeynArts, and the loop libraries behind NNLO calculations — implement the quark–gluon, three-gluon, and four-gluon vertices verbatim, while parton showers such as Pythia, Herwig, and Sherpa iterate the $q\to qg$ and $g\to gg$ splittings (governed by $C_F$ and $C_A$ ) to dress bare partons into the jets a detector sees. The self-coupling vertices are not abstractions: the three-gluon vertex was confirmed in the angular structure of four-jet events at LEP, distinguishing QCD from an abelian "QED of colour" that lacks it, and $C_A/C_F=9/4$ has been measured from the relative widths of quark and gluon jets. Lattice QCD, meanwhile, computes the low-energy regime where $\alpha_S\sim1$ and the vertices are no longer perturbative, predicting hadron masses to percent accuracy.

The physical picture

A QCD diagram is read exactly like a QED one — attach a factor to every line and vertex, multiply, contract — but the alphabet is larger. Every quark line carries, on top of its Dirac spinor, a three-component colour label; every gluon carries one of eight colour–anticolour combinations; and at each quark–gluon dot the colour can change, tracked by which Gell-Mann matrix sits in the vertex. The first figure lays out the complete set of QCD building blocks, so the new pieces — the gluon self-couplings — sit next to the familiar quark vertex.

Figure

The four building blocks of QCD: the quark–gluon vertex −½ig_S λ^a γ^μ where the quark changes colour r → b and the gluon carries r b̄; the three-gluon vertex ∝ g_S f^{abc}; the four-gluon vertex ∝ g_S²; and the gluon propagator −i g_μν δ_ab ∕ q². The last two vertices have no analogue in QED

The self-coupling inverts the sign of the most important number in the theory. In QED, virtual electron–positron pairs screen a charge, so the coupling grows as you probe closer; in QCD the gluon loops — permitted only because gluons couple to gluons — antiscreen colour charge and overwhelm the quark screening, so the coupling shrinks at short distance and grows at long distance. The second figure is the measured consequence: the running coupling $\alpha_S(Q)$ , pinned at $\alpha_S(M_Z)=0.1179$ , the single most important number in perturbative QCD.

Figure

The running strong coupling α_S(Q) from the one-loop QCD prediction, anchored to the world-average α_S(M_Z) = 0.1179 at Q = 91.19 GeV: the coupling falls from α_S ≈ 0.3 near 1 GeV toward α_S ≈ 0.09 at 1 TeV. This decrease — asymptotic freedom — exists only because the gluon self-coupling makes β_0 = 11 − ⅔n_f positive

The core equations

The QCD Feynman rules are the QED rules with a colour structure on every line plus an extra pair of self-coupling vertices. The quark–gluon vertex is the anchor: the QED vertex with charge $Qe$ replaced by $g_S$ and multiplied by a Gell-Mann matrix element.

V_{qqg}=-\tfrac12\,ig_S\,\lambda^a_{ji}\,\gamma^\mu=-ig_S\,(T^a)_{ji}\,\gamma^\mu,\qquad T^a\equiv\tfrac12\lambda^a \qquad (1)

where: $g_S$ — strong coupling, $\alpha_S=g_S^2/4\pi$ ; $\lambda^a$ ( $a=1\dots8$ ) — Gell-Mann matrices; $T^a=\lambda^a/2$ — SU(3) generators; $i,j$ — incoming/outgoing quark colour; $\gamma^\mu$ — Dirac matrix carrying the Lorentz index. Reading: $\gamma^\mu$ acts on the 4-spinor, $(T^a)_{ji}$ on the 3-colour state — different spaces, so the vertex factorises into a Dirac current times a colour number. The gluon propagator carries the photon's Lorentz part and a colour Kronecker delta that keeps the gluon type fixed along the internal line.

D^{ab}_{\mu\nu}(q)=\frac{-ig_{\mu\nu}\,\delta^{ab}}{q^2}\ \ (\text{Feynman gauge}),\qquad a,b=1\dots8 \qquad (2)

where: $q$ — gluon four-momentum fixed by conservation at the vertices; $g_{\mu\nu}$ — Minkowski metric; $\delta^{ab}$ — enforces that the gluon type emitted at one vertex is the same absorbed at the other. Reading: the massless $1/q^2$ pole is shared with the photon, but colour is carried by the vertices, not the propagator. The defining novelty is the three-gluon vertex, forced into existence by the non-abelian field strength. With all three momenta $p_1,p_2,p_3$ flowing into the vertex and labels $(a,\mu),(b,\nu),(c,\rho)$ :

V_{3g}=g_S f^{abc}\big[g^{\mu\nu}(p_1-p_2)^\rho+g^{\nu\rho}(p_2-p_3)^\mu+g^{\rho\mu}(p_3-p_1)^\nu\big] \qquad (3)

where: $f^{abc}$ — the totally antisymmetric SU(3) structure constants, $[T^a,T^b]=if^{abc}T^c$ ; the bracket is the unique momentum structure antisymmetric under interchange of any two gluons. Reading: unlike the constant QED vertex, this one is momentum-dependent — a fingerprint of the derivative in $F^a_{\mu\nu}$ — and vanishes for an abelian group ( $f^{abc}\to0$ ), which is why the photon has no self-coupling. The four-gluon vertex completes the self-interaction set, quadratic in the coupling. With colour–Lorentz labels $(a,\mu),(b,\nu),(c,\rho),(d,\sigma)$ :

V_{4g}=-ig_S^2\big[f^{abe}f^{cde}(g^{\mu\rho}g^{\nu\sigma}-g^{\mu\sigma}g^{\nu\rho})+f^{ace}f^{bde}(g^{\mu\nu}g^{\rho\sigma}-g^{\mu\sigma}g^{\nu\rho})+f^{ade}f^{bce}(g^{\mu\nu}g^{\rho\sigma}-g^{\mu\rho}g^{\nu\sigma})\big] \qquad (4)

where: the sum on $e$ runs over $1\dots8$ ; each term pairs two structure constants with a metric combination. Reading: this vertex is momentum-independent and order $g_S^2$ , one power of $\alpha_S$ deeper than the three-gluon vertex; it is required for gluon–gluon scattering to respect gauge invariance. Every colour sum in a physical rate collapses onto three SU(3) invariants — the Casimirs — the numbers that actually multiply cross-sections.

\sum_a (T^aT^a)_{ij}=C_F\,\delta_{ij},\qquad f^{acd}f^{bcd}=C_A\,\delta^{ab},\qquad \mathrm{Tr}(T^aT^b)=T_F\,\delta^{ab} \qquad (5)

where: for SU(3), $C_F=(N^2-1)/2N=4/3$ , $C_A=N=3$ , $T_F=1/2$ . Reading: $C_F$ appears whenever a gluon attaches to a quark line, $C_A$ to a gluon line, $T_F$ in every closed quark loop — a diagram's entire colour content reduces to products of these three. Finally, the coupling runs, and the sign of its running is dictated by the very self-coupling that equations (3)–(4) encode. The one-loop beta function is:

\mu^2\frac{d\alpha_S}{d\mu^2}=-\frac{\beta_0}{4\pi}\alpha_S^2+\cdots,\qquad \beta_0=\frac{11}{3}C_A-\frac{4}{3}T_F n_f=11-\frac{2}{3}n_f \qquad (6)

where: $n_f$ — active quark flavours; $+\tfrac{11}{3}C_A$ is the gluon (antiscreening) contribution, $-\tfrac43 T_F n_f$ the quark (screening) one. Reading: because $11$ beats $\tfrac23 n_f$ for all $n_f\le16$ , $\beta_0>0$ and the coupling decreases with energy — asymptotic freedom, derived in full in aq-pp-running-alpha-s and aq-pp-asymptotic-freedom.

Derivation AWhere the quark–gluon vertex comes from and why it factorises

The goal is to show that the vertex $-ig_S(T^a)_{ji}\gamma^\mu$ is not an independent postulate but is dictated by promoting the QED gauge principle from the abelian group U(1) to the non-abelian group SU(3), and to make explicit why the colour part factorises cleanly from the Dirac part. We start from the free quark Lagrangian, demand invariance under a local SU(3) colour rotation, introduce the covariant derivative this forces, and read the single cubic interaction term off the result.

Step 1. Begin with the free Dirac Lagrangian for a colour triplet, $\mathcal{L}_0=\bar q_i(i\gamma^\mu\partial_\mu-m)q_i$ , summed over $i=1,2,3$ . It is invariant under a global SU(3) rotation $q\to\exp(i\alpha^a T^a)q$ but not a local one $\alpha^a\to\alpha^a(x)$ , because $\partial_\mu$ then strikes $\alpha^a(x)$ . Why: demanding that the choice of what "red, green, blue" mean be free at every point — the non-abelian gauge principle — forces the gluons into existence, just as locality forced the photon in QED.

Step 2. Restore invariance with the covariant derivative $D_\mu=\partial_\mu-ig_S A^a_\mu T^a$ , where the eight gluon fields $A^a_\mu$ transform so as to cancel the unwanted term. The Lagrangian becomes $\mathcal{L}=\bar q(i\gamma^\mu D_\mu-m)q=\mathcal{L}_0+g_S\bar q\gamma^\mu T^a q\,A^a_\mu$ . Why: the matrix $T^a$ inside the derivative is the entire difference from QED, where the coupling was a mere number $Qe$ ; here it is a $3\times3$ matrix, which is what lets gluons rotate colour.

Step 3. Read off $\mathcal{L}_{\text{int}}=g_S\,\bar q_j\gamma^\mu (T^a)_{ji} q_i\,A^a_\mu$ — two quark fields, one gluon field, a single three-line vertex. The Feynman rule is $i$ times the coefficient with fields stripped: $i\times g_S(T^a)_{ji}\gamma^\mu$ , which in Thomson's convention ( $\lambda^a=2T^a$ , plus a sign choice) is $-\tfrac12 ig_S\lambda^a_{ji}\gamma^\mu$ . Why: the number of fields in the term is the number of lines at the vertex, and $(T^a)_{ji}$ records the colour rotation the gluon performs.

Step 4. Observe the factorisation. Since $T^a$ acts only on the 3-colour index and $\gamma^\mu$ only on the 4-spinor index, the current splits as $[\,\bar c_j(T^a)_{ji}c_i\,]\times[\,\bar u\gamma^\mu u\,]$ — a colour number times the identical Dirac current of QED — and for definite in/out colours the first bracket is just $(T^a)_{ji}$ , one entry of a Gell-Mann matrix. Why: the two structures live in different vector spaces and never mix, so every QCD amplitude is a QED-like kinematic factor times a purely group-theoretic colour factor — the organising principle of the whole subject.

Go deeper · Derivation B · 1 minThe gluon self-coupling from the non-abelian field strength

Derivation BThe gluon self-coupling from the non-abelian field strength

The goal is to trace the three- and four-gluon vertices to their single origin: the extra commutator term in the QCD field strength that has no counterpart in electromagnetism. We write the abelian and non-abelian field strengths side by side, square the non-abelian one to form the kinetic term, and identify the cubic and quartic pieces that become the self-interaction vertices — showing that gluon self-coupling and the gluon's colour charge are one fact stated twice.

Step 1. In QED the field strength $F_{\mu\nu}=\partial_\mu A_\nu-\partial_\nu A_\mu$ is linear in the field, so $-\tfrac14 F_{\mu\nu}F^{\mu\nu}$ is purely quadratic — the photon does not self-interact. Why: a quadratic Lagrangian gives only a propagator and no vertex, which is why light passes through light.

Step 2. In QCD the field strength gains a commutator term, $G^a_{\mu\nu}=\partial_\mu A^a_\nu-\partial_\nu A^a_\mu+g_S f^{abc}A^b_\mu A^c_\nu$ , because the gauge fields do not commute. Why: the $f^{abc}A^bA^c$ piece says the gluon field itself carries colour charge — the structure constants are the colour charges of the gluons, just as $(T^a)_{ji}$ were the quarks'.

Step 3. Form the kinetic term $-\tfrac14 G^a_{\mu\nu}G^{a\,\mu\nu}$ and expand. The cross term between the derivative part and the $g_S f^{abc}A^bA^c$ part is cubic in $A$ — a three-gluon interaction $\propto g_S f^{abc}(\partial A)AA$ , whose Feynman rule is equation (3). Why: a derivative on a gluon becomes a momentum, which is why the three-gluon vertex is momentum-dependent while the QED vertex is not.

Step 4. The square of the $g_S f^{abc}A^bA^c$ term is quartic in $A$ — a four-gluon interaction $\propto g_S^2 ff AAAA$ , the momentum-independent vertex of equation (4). Why: the two self-coupling vertices are the two ways of distributing the new term across the squared field strength; kill $f^{abc}$ and both vanish, recovering QED — their presence is the entire dynamical difference between the two forces.

Go deeper · Derivation C · 1 minThe colour Casimir C_F = 4/3 and the quark–antiquark colour factors

Derivation CThe colour Casimir C_F = 4/3 and the quark–antiquark colour factors

The goal is to compute the colour factor $C_F$ that multiplies every quark–gluon coupling, then use the colour algebra to explain why a quark and antiquark bind in a colour singlet but repel in an octet — the group-theoretic reason mesons exist. We evaluate the fundamental Casimir from the generator normalisation, then decompose the one-gluon-exchange colour operator on the two channels.

Step 1. The generators are normalised by $\mathrm{Tr}(T^aT^b)=\tfrac12\delta^{ab}$ , i.e. $\mathrm{Tr}(\lambda^a\lambda^b)=2\delta^{ab}$ . Summing $a=b$ over all eight gives $\mathrm{Tr}\big(\sum_a\lambda^a\lambda^a\big)=2\times8=16$ . Why: this single trace fixes the scale of the colour charge, and everything downstream is bookkeeping on it.

Step 2. Because $\sum_a\lambda^a\lambda^a$ commutes with every generator, by Schur's lemma it is a multiple of the identity, $\sum_a\lambda^a\lambda^a=c\,\mathbb{1}_3$ . Taking the trace, $3c=16$ , so $c=16/3$ and $\sum_a T^aT^a=\tfrac14 c\,\mathbb{1}=\tfrac43\mathbb{1}$ , giving $C_F=4/3$ . Why: $C_F=4/3$ is the effective "colour charge squared" a quark presents to a gluon — the QCD analogue of $Q^2$ in QED.

Step 3. For one-gluon exchange between a quark ( $\mathbf{T}_1$ ) and antiquark ( $\mathbf{T}_2$ ), the colour factor is $\mathbf{T}_1\!\cdot\!\mathbf{T}_2=\tfrac12\big[(\mathbf{T}_1+\mathbf{T}_2)^2-\mathbf{T}_1^2-\mathbf{T}_2^2\big]$ , with each carrying Casimir $C_F=4/3$ . Why: rewriting the product as a difference of total and individual Casimirs lets the answer be read off from the SU(3) representation of the pair.

Step 4. Decompose $3\otimes\bar3=1\oplus8$ . The singlet has total Casimir $0$ , giving $\mathbf{T}_1\!\cdot\!\mathbf{T}_2=\tfrac12[0-\tfrac83]=-\tfrac43$ — attractive. The octet has total Casimir $C_A=3$ , giving $\tfrac12[3-\tfrac83]=+\tfrac16$ — repulsive. Why: the negative singlet factor binds a colour-neutral $q\bar q$ pair into a meson ( $V=-\tfrac43\alpha_S/r$ ), while the positive octet factor forbids coloured bound states — the colour algebra selects exactly the hadrons nature shows us.

Limiting cases

Case	Result	Meaning
Abelian limit ( $f^{abc}\to0$ )	3g and 4g vertices vanish	QCD collapses to a QED-like theory of a non-self-coupling "photon"; no confinement, no asymptotic freedom
Single quark line	colour factor $C_F=4/3$	a gluon attached to a quark; the QCD analogue of the QED charge-squared
Gluon line	colour factor $C_A=3$	a gluon attached to a gluon; larger than $C_F$ , so gluons radiate $9/4$ as much
Closed quark loop	colour factor $T_F\,n_f=\tfrac12 n_f$	quark-loop screening; opposes the gluon antiscreening in $\beta_0$
High energy ( $Q\to\infty$ )	$\alpha_S(Q)\to0$ logarithmically	asymptotic freedom — quarks act free; perturbation theory becomes exact
Low energy ( $Q\to\Lambda_{\text{QCD}}\!\approx\!0.2$ GeV)	$\alpha_S\to\mathcal{O}(1)$ , diverges	confinement regime — perturbation theory fails, quarks are permanently bound
Colour-singlet $q\bar q$	$\mathbf{T}_1\!\cdot\!\mathbf{T}_2=-\tfrac43$	attractive; the bound meson channel, $V=-\tfrac43\alpha_S/r$

Common trapThe δ_ab in the gluon propagator means gluons cannot change colour

The single most common misreading is to look at the propagator's $\delta^{ab}$ and conclude that gluons conserve colour the way a neutral photon conserves charge. They do not. The $\delta^{ab}$ only says the type of gluon ( $a$ ) emitted at one vertex is the same type absorbed at the other — a statement about the internal line. What changes colour is the vertex: the factor $(T^a)_{ji}=\tfrac12\lambda^a_{ji}$ is off-diagonal for six of the eight gluons, so a red quark emitting a gluon of type $\lambda^1$ or $\lambda^4$ turns into a blue or green quark. The gluon carries away the colour difference — a colour–anticolour object like $r\bar b$ — which is exactly why gluons, unlike the colour-blind photon, source further gluons. Confusing "same gluon type on the line" with "colour unchanged at the vertex" erases the non-abelian content of the theory.

Worked examples

1 · The colour factor for one-gluon exchange (confidence builder)

Compute the colour factor that multiplies the QED-like amplitude when a quark and antiquark exchange a single gluon, first for a colour singlet (a meson) and then for a colour octet, and state the resulting static potential.

Step 1. Strip the kinematics. The amplitude factorises as $\mathcal{M}=[\text{colour factor}]\times[\text{Dirac part identical to QED}]$ , with the colour factor the expectation value of $\mathbf{T}_1\!\cdot\!\mathbf{T}_2\equiv\sum_a T^a_{(1)}T^a_{(2)}$ in the pair's colour state — the only piece new relative to QED.

Step 2. Use $\mathbf{T}_1\!\cdot\!\mathbf{T}_2=\tfrac12[(\mathbf{T}_1+\mathbf{T}_2)^2-\mathbf{T}_1^2-\mathbf{T}_2^2]$ . Each quark and antiquark has $\mathbf{T}^2=C_F=\tfrac43$ , so the individual terms subtract a fixed $\tfrac83$ in every channel.

Step 3. Insert the total Casimir for each channel of $3\otimes\bar3=1\oplus8$ . Singlet (total Casimir $0$ ): $\mathbf{T}_1\!\cdot\!\mathbf{T}_2=\tfrac12[0-\tfrac83]=\mathbf{-\tfrac43}$ . Octet (total Casimir $C_A=3$ ): $\mathbf{T}_1\!\cdot\!\mathbf{T}_2=\tfrac12[3-\tfrac83]=\mathbf{+\tfrac16}$ .

Step 4. Read the physics. A negative colour factor is an attractive potential; combined with Coulomb-like gluon exchange it gives $V(r)=-\tfrac43\dfrac{\alpha_S}{r}$ , the short-distance part of the charmonium and bottomonium spectra. The octet's $+\tfrac16$ is repulsive, so no coloured bound state forms. Singlet colour factor −4/3 (attractive, binds mesons); octet +1/6 (repulsive); potential V = −(4/3)α_S/r.

2 · The strong coupling as a number, and the sign of the running

Given the world-average $\alpha_S(M_Z)=0.1179$ at $M_Z=91.19$ GeV, find the vertex coupling $g_S$ , compare it to the QED vertex strength, and show from $\beta_0$ that the coupling weakens at higher energy.

Step 1. Extract $g_S$ from $\alpha_S=g_S^2/4\pi$ : $g_S=\sqrt{4\pi\alpha_S}=\sqrt{4\pi\times0.1179}=\mathbf{1.22}$ . Compare with QED, where $e=\sqrt{4\pi\alpha}=0.303$ : the strong vertex is $g_S/e\approx4.0$ times stronger even at the $Z$ mass, and far stronger at low energy.

Step 2. Account for colour. The effective strength on a quark line is $g_S^2 C_F$ , on a gluon line $g_S^2 C_A$ , so $C_A/C_F=3/(4/3)=\mathbf{9/4}$ : a gluon radiates gluons $2.25$ times as readily as a quark — measured in the different rates of quark and gluon jets.

Step 3. Evaluate the beta-function coefficient at $n_f=5$ : $\beta_0=11-\tfrac23(5)=11-\tfrac{10}{3}=\mathbf{7.67}$ , positive because the gluon antiscreening term $\tfrac{11}{3}C_A=11$ dominates the quark screening term $\tfrac43 T_F n_f=\tfrac{10}{3}$ .

Step 4. Read off the running. With $\beta_0>0$ , equation (6) gives $d\alpha_S/d\ln\mu^2<0$ : the coupling falls as the scale rises — from $0.118$ at $M_Z$ to about $0.088$ at $1$ TeV, rising to $\approx0.30$ at $2$ GeV. g_S = 1.22 at M_Z (≈4× the QED coupling); β_0 = 7.67 > 0, so α_S decreases with energy — asymptotic freedom, driven by the gluon self-coupling.

Physics of the portrait

The QCD vertex wears one man's mathematics: the matrices $\lambda^a$ in the coupling are Gell-Mann's, the SU(3) they generate is his, and "quark" and "quantum chromodynamics" are both his coinages.

Origins · 1 minMurray Gell-Mann and how this was discovered

Murray Gell-Mann

1929–2019 · SU(3) symmetry, quarks, and the naming of QCD

Murray Gell-Mann was the great classifier of the particle era, and the colour structure of the QCD vertex is built almost entirely from his tools. A prodigy who entered Yale at fifteen, he introduced "strangeness" in 1953, and in 1961 organised the exploding zoo of hadrons into the Eightfold Way — multiplets of the group SU(3), whose eight generators are the Gell-Mann matrices $\lambda^a$ that now sit inside every quark–gluon vertex. That flavour SU(3) predicted a missing particle, the $\Omega^-$ , found in 1964 with exactly the mass he specified. In the same year he proposed (independently of George Zweig) that hadrons are built from three fractionally-charged constituents, which he named quarks after a line in Joyce's Finnegans Wake. When the model demanded three quarks in the same state for the $\Delta^{++}$ and $\Omega^-$ — apparently violating Fermi statistics — the fix was a new threefold charge, colour, gauged by a second SU(3); Gell-Mann, with Harald Fritzsch and Heinrich Leutwyler, assembled this into the theory he christened quantum chromodynamics in 1973, the year asymptotic freedom made it credible. He won the 1969 Nobel Prize for the classification work that made all of it possible — and it is his SU(3) a physicist invokes every time a $\lambda^a$ is written at a gluon vertex.

History

1954

— Yang and Mills construct the first non-abelian gauge theory, the template for gluon self-coupling

1961

— Gell-Mann and Ne'eman organise the hadrons by SU(3), the Eightfold Way

1964

— Gell-Mann and Zweig propose quarks; colour is introduced to rescue the statistics

1971

— 't Hooft and Veltman prove non-abelian gauge theories are renormalisable

1973

— Gross, Wilczek and Politzer discover asymptotic freedom; Gell-Mann, Fritzsch and Leutwyler name QCD

Going further

Texts. Thomson, Modern Particle Physics, §10.2–§10.4 (QCD, colour factors, and running coupling) · Peskin & Schroeder, An Introduction to Quantum Field Theory, §16.1 (non-abelian gauge invariance) and the QCD Feynman rules on the inside cover · Halzen & Martin, Quarks and Leptons, §14.1–§14.4 · Griffiths, Introduction to Elementary Particles, 2nd ed., §8.4–§8.5 · Schwartz, Quantum Field Theory and the Standard Model, Ch. 25–26. Key results to keep in mind: the vertex $-ig_S(T^a)_{ji}\gamma^\mu$ is $i$ times the interaction term of $\mathcal{L}=\bar q(i\gamma^\mu D_\mu-m)q$ with $D_\mu=\partial_\mu-ig_S A^a_\mu T^a$ ; the three- and four-gluon vertices arise from $-\tfrac14 G^a_{\mu\nu}G^{a\mu\nu}$ with the commutator term in $G^a_{\mu\nu}$ ; the SU(3) Casimirs are $C_F=4/3$ , $C_A=3$ , $T_F=1/2$ , with $q\bar q$ colour factors $-4/3$ (singlet) and $+1/6$ (octet); and $g_S=\sqrt{4\pi\alpha_S}=1.22$ at $\alpha_S(M_Z)=0.1179$ , with $\beta_0=11-\tfrac23 n_f>0$ .

Prerequisites. Colour and QCD · Gluons · Feynman Rules of QED · Dirac Spinor Solutions. Leads to. Colour Factors · Running of $\alpha_S$ · Asymptotic Freedom.

Ask the tutor: Name any strong-interaction process — a quark scatter, a gluon splitting, a quarkonium binding — and I will assemble its amplitude and colour factor from the QCD rules step by step.

Why is the QCD vertex written with λ^a in some books and T^a in others?

They are the same thing up to a factor of two: the SU(3) generators are $T^a=\tfrac12\lambda^a$ , with $\lambda^a$ the eight Gell-Mann matrices. Thomson writes the vertex as $-\tfrac12 ig_S\lambda^a_{ji}\gamma^\mu$ , Peskin and Schwartz as $ig_S(T^a)_{ji}\gamma^\mu$ (convention-dependent sign) — identical once you substitute $T^a=\lambda^a/2$ . The generators are the more fundamental object, satisfying the clean algebra $[T^a,T^b]=if^{abc}T^c$ and normalisation $\mathrm{Tr}(T^aT^b)=\tfrac12\delta^{ab}$ , from which the Casimirs $C_F=4/3$ and $C_A=3$ follow. When comparing formulas across textbooks, first check whether an expression uses $\lambda^a$ or $T^a$ ; a stray factor of two in a colour factor almost always traces to this choice.

If the gluon propagator conserves colour type, how does a gluon change a quark's colour?

The colour change happens at the vertex, not on the propagator. The propagator's $\delta^{ab}$ only says the internal line carries one definite gluon type from end to end. But the vertex factor $(T^a)_{ji}=\tfrac12\lambda^a_{ji}$ is off-diagonal for six of the eight gluons: $\lambda^1,\lambda^2$ mix red and blue, $\lambda^4,\lambda^5$ mix red and green, $\lambda^6,\lambda^7$ mix green and blue. So a red quark emitting a $\lambda^4$ -type gluon becomes green, and the gluon carries off the red–antigreen colour it removed; only the diagonal gluons ( $\lambda^3,\lambda^8$ ) leave the colour unchanged. The gluon is thus a colour–anticolour object, which is precisely why — unlike the neutral photon — it radiates further gluons.

Why are there exactly eight gluons and not nine?

Because the gauge group is SU(3), not U(3). U(3) would have all nine independent $3\times3$ Hermitian generators, including the identity — a colour-singlet ninth gluon coupling to total colour like a photon, giving the strong force an unwanted infinite-range, colour-blind component. SU(3) removes that singlet by demanding the generators be traceless, leaving $3^2-1=8$ matrices — the Gell-Mann $\lambda^a$ . A singlet gluon could be emitted by any colour-neutral hadron and would produce long-range forces between distant protons, which are not observed. The eight-ness is thus not an accident but the statement that only the traceless part of colour is gauged.

How can the strong coupling be "strong" if α_S(M_Z) = 0.118 is smaller than one?

It is strong at low energy and only moderate at high energy — that is the whole point of a running coupling. At the $Z$ mass, $\alpha_S=0.118$ is about sixteen times the electromagnetic $\alpha\approx1/137$ , but still small enough for perturbation theory. As you probe softer scales $\alpha_S$ grows: near $1$ GeV it approaches unity, and around $\Lambda_{\text{QCD}}\approx0.2$ GeV the perturbative formula diverges. That non-perturbative regime is where quarks are confined and hadrons form. So "strong force" refers to its low-energy behaviour, where it binds protons and neutrons; at collider energies asymptotic freedom tames it into something perturbation theory can handle.

What are Faddeev–Popov ghosts and do I need them for tree diagrams?

Ghosts are unphysical scalar fields, obeying Fermi statistics despite spin zero, added to QCD in covariant gauges to cancel the unphysical (longitudinal and timelike) polarisations that the gluon propagator $-ig_{\mu\nu}\delta^{ab}/q^2$ propagates. They appear only in loops, circulating with a ghost–gluon vertex $\propto g_S f^{abc}$ , restoring gauge invariance and unitarity. For tree-level diagrams — where this article works — you never need them: no closed loops, no unphysical polarisations to cancel. You first meet ghosts in the gluon self-energy, where they contribute to the beta function alongside the gluon and quark loops; in physical (axial) gauges they vanish entirely, at the cost of a messier propagator.

Do the three- and four-gluon vertices really matter, or are they a technicality?

They are the single most important feature of QCD. Without gluon self-coupling the beta-function coefficient would be $\beta_0=-\tfrac43 T_F n_f<0$ , the coupling would grow at high energy exactly like QED, and there would be no asymptotic freedom, no confinement, and no reason for quarks to bind. The $+\tfrac{11}{3}C_A=11$ antiscreening term in $\beta_0$ comes entirely from the self-interaction and is what flips the sign. Experimentally the vertex is not optional either: four-jet angular distributions at LEP fit only with the SU(3) value, ruling out abelian alternatives. Far from a technicality, these two vertices are why the strong force behaves like nothing else in physics.