Jets & fragmentation

A quark is never seen alone. The instant an electron and a positron annihilate into a virtual photon and that photon converts to a quark–antiquark pair, the two coloured partons start to separate — and because the strong force does not fade with distance, the colour field between them stretches into a taut tube of stored energy. Long before the quarks can travel a nuclear diameter, that tube snaps: it is energetically cheaper to pull new $q\bar q$ pairs out of the vacuum than to keep stretching. What emerges, a fraction of a femtometre and $\sim10^{-23}$ s later, is not two quarks but two collimated sprays of colourless hadrons, each barrelling along the direction of the parton that seeded it. These are jets, and they are the closest thing to a photograph of a quark that confinement permits. This article is about reading those photographs. The messy, non-perturbative business of turning a parton into hadrons — fragmentation, or hadronisation — happens too late and too softly to change how many quark pairs were made or which way they went, so the clean short-distance physics survives in the jet directions and rates. The two-jet axis follows a $1+\cos^2\theta$ distribution, betraying that quarks are spin-½. The overall hadron rate, the ratio $R$ , carries a factor of three that counts colour. And when a hard gluon is radiated the event grows a third jet, at a rate set by the strong coupling $\alpha_s$ — the discovery that made the gluon visible. Four-jet events, and the difference between quark and gluon jets, then test the non-Abelian $SU(3)$ heart of QCD directly.

Key takeaways

Confinement forbids free quarks, so a produced $q\bar q$ pair fragments into two collimated jets of hadrons that inherit the parton directions and energies.
In $e^+e^-\to q\bar q$ the two-jet axis follows $\dfrac{1}{\sigma}\dfrac{d\sigma}{d\cos\theta}=\dfrac{3}{8}(1+\cos^2\theta)$ , the unmistakable signature of spin-½ quarks.
The hadronic rate $R=N_c\sum_f Q_f^2$ carries a factor $N_c=3$ that counts colour; a small QCD correction $(1+\alpha_s/\pi)$ sits on top.
Hard gluon radiation makes three-jet events $e^+e^-\to q\bar q g$ at a rate $\propto\alpha_s$ , revealing the gluon; four-jet events probe the triple-gluon vertex.
Jets are defined operationally by an algorithm and a resolution $y_{\text{cut}}$ ; event shapes like thrust $T$ quantify how jet-like an event is ( $T=1$ pencil, $T=\tfrac12$ spherical).
Key idea: free quarks and gluons are never seen, but they betray their spin, colour, and gauge structure in the number, angular distribution, and rate of the collimated hadron jets they fragment into.

The core insight

Free quarks and gluons are never seen, but they betray their existence by fragmenting into collimated jets of hadrons whose number, angular distributions, and production rates encode the spin, colour, and non-Abelian gauge structure of QCD.

Where this lives today

Jets are the working language of every high-energy collider. At the LHC the anti- $k_T$ algorithm — a descendant of equation (6) tuned for hadron collisions — reconstructs the jets that tag quarks and gluons in essentially every analysis, from precision QCD to searches for new particles; a Higgs boson decaying to $b\bar b$ , or a top quark, is read off jet momenta and jet substructure. That substructure is a field in its own right: when a heavy particle is boosted, its decay products merge into a single fat jet whose internal pattern (the "Lund plane") distinguishes a $W$ or Higgs from ordinary QCD radiation, and machine-learning taggers now classify quark, gluon, and boosted-object jets by exactly the multiplicity and width differences that the $C_A/C_F=\tfrac94$ ratio predicts. At LEP the event-shape distributions — thrust above all — remain among the most precise determinations of $\alpha_s$ and of its running, and the four-jet angular correlations there measured the triple-gluon vertex, confirming the non-Abelian self-coupling directly. In heavy-ion collisions at the LHC and RHIC, jets become probes of the quark–gluon plasma: a jet punching through the hot medium loses energy and is "quenched," and the imbalance between a jet and its recoiling partner images the plasma's opacity. From the spin meter of $1+\cos^2\theta$ to jet quenching in a little Big Bang, the same principle holds — the parton is invisible, but its jet is not.

The physical picture

Picture the moment of annihilation. The electron and positron vanish into a single virtual photon carrying the whole centre-of-mass energy, and that photon materialises a quark and an antiquark flying apart back-to-back. They are colour-charged, and the QCD field lines between them do not spread out the way electric field lines do — the gluon self-interaction squeezes them into a narrow flux tube whose energy grows linearly with separation. By the time the quarks are a femtometre apart the tube stores enough energy to make new hadrons, so it fragments along its length into a chain of colour-neutral mesons and baryons. Each original quark ends up at the head of a spray of hadrons moving in nearly its original direction: a jet. The first figure freezes this — production, the stretching string, and the two emerging jets — and states the angular law the jet axis must obey.

Figure

e⁺e⁻ → γ* → qq̄ hadronises into two back-to-back jets along the original quark directions. The jet axis follows dN/dcosθ ∝ 1 + cos²θ, the fingerprint of spin-½ quarks, and each hadron carries a fraction z = 2E_h/√s of the beam energy

The story does not stop at two jets. Just before the pair hadronises, either the quark or the antiquark can radiate a hard, wide-angle gluon — a genuine short-distance QCD emission, calculable in perturbation theory — and that gluon, being coloured, fragments into a jet of its own. The event now shows three jets splayed in a plane, the direct fingerprint of the gluon. Their relative rate is proportional to $\alpha_s$ , so counting three-jet events measures the strength of the strong force, and the shape of the energy sharing tests the spin-1 nature of the gluon. The second figure is that punchline: the three-jet topology, its energy bookkeeping, and the fact that the gluon jet is the broader one.

Figure

A three-jet event e⁺e⁻ → qq̄g: a hard gluon radiated off the quark or antiquark makes a third jet. The three scaled energies satisfy x₁ + x₂ + x₃ = 2, the rate is set by (α_s/2π)·C_F with C_F = 4/3, and the gluon jet is broader, radiating C_A/C_F = 9/4 times more than a quark jet

The core equations

The first quantitative handle is the two-jet angular distribution. Because the production step $e^+e^-\to q\bar q$ is identical in spin structure to $e^+e^-\to\mu^+\mu^-$ , the jet axis inherits the same law, and fragmentation — happening long afterwards and nearly along the axis — does not distort it.

\frac{1}{\sigma}\frac{d\sigma}{d\cos\theta}=\frac{3}{8}\left(1+\cos^2\theta\right) \qquad (1)

where: $\theta$ — angle of the jet (thrust) axis to the beam; the $\tfrac38$ normalises the distribution to unity over $\cos\theta\in[-1,1]$ . Reading it: the $1+\cos^2\theta$ shape is the spin-½ fingerprint — a spin-0 quark would instead give $\sin^2\theta\propto(1-\cos^2\theta)$ , which the data decisively exclude. The overall production rate is packaged as the ratio $R$ . The numerator and denominator share the same initial state and photon propagator, so everything cancels except the final-state charge, the colour multiplicity, and a small perturbative QCD correction.

R\equiv\frac{\sigma(e^+e^-\to\text{hadrons})}{\sigma(e^+e^-\to\mu^+\mu^-)}=N_c\sum_f Q_f^2\left(1+\frac{\alpha_s}{\pi}+\mathcal{O}(\alpha_s^2)\right) \qquad (2)

where: $N_c=3$ — the number of colours; $Q_f$ — quark charge in units of $e$ ; the sum runs over flavours with $2m_f<\sqrt s$ . Reading it: the leading integer counts colour ( $R=2,\,\tfrac{10}{3},\,\tfrac{11}{3}$ across the $uds$ , $c$ , $b$ thresholds), while the $\alpha_s/\pi$ term is the gluon-radiation bump that fragmentation smears into the three-jet tail. The third jet is governed by the $\mathcal{O}(\alpha_s)$ matrix element for $e^+e^-\to q\bar q g$ , expressed in the scaled energies $x_i=2E_i/\sqrt s$ . This is the equation that turned the gluon from hypothesis into measurement.

\frac{1}{\sigma_0}\frac{d^2\sigma}{dx_1\,dx_2}=\frac{\alpha_s}{2\pi}\,C_F\,\frac{x_1^2+x_2^2}{(1-x_1)(1-x_2)},\qquad C_F=\frac{4}{3} \qquad (3)

where: $\sigma_0=\sigma(e^+e^-\to q\bar q)$ ; $x_1,x_2$ — scaled energies of quark and antiquark, with $x_1+x_2+x_3=2$ ; $C_F=\tfrac43$ — the quark colour charge (Casimir of the fundamental representation). Reading it: the denominator diverges as $x_1\to1$ or $x_2\to1$ (a soft or collinear gluon), which is why "three-jet" only means anything once a resolution cut is imposed; the overall $\alpha_s$ makes the resolved rate a direct measurement of the strong coupling. To decide how jet-like an event is without ever referring to a parton, one uses an event-shape variable. Thrust maximises the momentum flow along a single axis and needs no arbitrary jet count.

T=\max_{\hat{\mathbf n}}\frac{\sum_i|\mathbf p_i\cdot\hat{\mathbf n}|}{\sum_i|\mathbf p_i|},\qquad \tfrac12\le T\le 1 \qquad (4)

where: $\mathbf p_i$ — momenta of all final hadrons; $\hat{\mathbf n}$ — the unit vector (thrust axis) that maximises the projected momentum. Reading it: $T=1$ is a perfect back-to-back two-jet pencil, $T=\tfrac12$ is a fully spherical event, and a symmetric three-jet event sits at $T=\tfrac23$ — so the thrust distribution alone maps the two-jet-to-multijet transition. Fragmentation itself is quantified by the fragmentation function $D_i^h(z)$ : the number density of hadron $h$ carrying momentum fraction $z$ of parton $i$ . It is not calculable from first principles, but it is universal, and it obeys an exact momentum sum rule.

\frac{1}{\sigma}\frac{d\sigma}{dz}=\sum_q\hat e_q^{\,2}\,D_q^h(z),\qquad \sum_h\int_0^1 z\,D_i^h(z)\,dz=1 \qquad (5)

where: $z=2E_h/\sqrt s$ — the hadron energy fraction; $\hat e_q^{\,2}$ — the relative $q\bar q$ production weight. Reading it: the sum rule states that all of a parton's momentum is shared among the hadrons it produces; at leading order $D(z)$ is independent of energy (scaling), and QCD predicts a slow, logarithmic scaling violation as radiation softens the spectrum. Finally, "jet" must be defined by an algorithm. A pair of objects is merged into one jet when its resolution variable falls below a chosen cut, making the count infrared-safe.

y_{ij}=\frac{2\min(E_i^2,E_j^2)\,(1-\cos\theta_{ij})}{s}\;<\;y_{\text{cut}} \qquad (6)

where: $E_i,E_j,\theta_{ij}$ — energies and opening angle of two objects; $y_{\text{cut}}$ — the resolution parameter. Reading it: this is the Durham ( $k_T$ ) measure; the older JADE choice uses $2E_iE_j(1-\cos\theta_{ij})/s\approx M_{ij}^2/s$ . A smaller $y_{\text{cut}}$ resolves more, softer jets — the number of jets is a function of $y_{\text{cut}}$ , not an absolute.

Derivation AThe two-jet angular distribution as a spin meter

The goal is to show that the $1+\cos^2\theta$ angular law of the two-jet axis is a direct readout of the quark's spin, by tracing it back to the same helicity bookkeeping that governs $e^+e^-\to\mu^+\mu^-$ , and by contrasting it with the $\sin^2\theta$ shape that a hypothetical spin-0 quark would produce — so that a single measured curve settles the question of whether quarks are Dirac fermions. The strategy is to argue that the production vertex is pure QED, that only the final-state spin structure sets the angular shape, and that hadronisation preserves the axis.

Step 1. Start from the known result for $e^+e^-\to\mu^+\mu^-$ (recalled from aq-pp-ee-to-mumu): a spin-1 photon coupling to spin-½ currents gives $\overline{|\mathcal M|^2}\propto1+\cos^2\theta$ at high energy. Replacing the muon with a quark of flavour $f$ changes only the vertex charge ( $e\to eQ_f$ ) and adds a colour sum ( $\times N_c$ ), neither of which depends on angle. Why: the angular dependence lives entirely in the spinor structure of the currents, so scaling the coupling and summing colours rescales the rate but cannot reshape the distribution.

Step 2. Read off the physics of the shape. For massless spin-½ fermions, angular momentum conservation forbids the final pair from being emitted along the beam with the wrong helicity correlation; the surviving amplitudes interfere to give a distribution peaked at $\cos\theta=\pm1$ and dipping (but not vanishing) at $90^\circ$ — precisely $1+\cos^2\theta$ . Why: the $\pm1$ peaking is the statement that a transversely polarised photon prefers to make a fermion pair along its polarisation, a purely spin-1-to-spin-½ selection rule.

Step 3. Contrast the spin-0 hypothesis. A scalar quark pair couples to the photon through $(p_+-p_-)^\mu$ , giving $\overline{|\mathcal M|^2}\propto\beta^2\sin^2\theta$ : the rate vanishes along the beam and peaks at $90^\circ$ — the exact opposite of the data. Why: a scalar carries no spin to align, so the only vector available is the momentum, and the resulting $\sin^2\theta$ is the unmistakable signature of the wrong spin.

Step 4. Connect to jets. Fragmentation redistributes energy among many hadrons but sends them nearly along the parent parton direction, so the reconstructed jet (or thrust) axis tracks the original $q\bar q$ line to within the small $\langle p_T\rangle\sim0.3$ GeV of the hadronisation smearing. Measuring $\tfrac1\sigma\,d\sigma/d\cos\theta$ therefore returns $1+\cos^2\theta$ , and the data at PETRA and PEP confirmed it. Why: because the smearing is soft and symmetric, it averages out of the axis, so the jet distribution faithfully reports the parton-level spin law. *Verified: normalisation $\tfrac38\int_{-1}^1(1+\cos^2\theta)\,d\cos\theta=1$ ; the forward ratio $N(|\cos\theta|=0.9)/N(0)=1.81$ for spin-½ versus $0.19$ for spin-0 — a factor $\sim9.5$ apart. Consistent with Halzen & Martin §13.2 and Thomson §10.6.*

Go deeper · Derivation B · 2 minThree-jet events, soft and collinear singularities, and the measurement of α_s

Derivation BThree-jet events, soft and collinear singularities, and the measurement of α_s

The aim is to derive the structure of the $e^+e^-\to q\bar q g$ rate, equation (3), to expose why it diverges when the gluon is soft or collinear, and to explain how imposing a resolution cut turns those divergences into a finite, calculable three-jet fraction that measures the strong coupling — the chain of reasoning behind the 1979 gluon discovery. We build the emission amplitude schematically, locate its singularities, and show what a jet definition does to them.

Step 1. Attach a gluon to the outgoing quark line. The extra quark propagator and quark–gluon vertex bring a factor $g_s\,C_F^{1/2}$ into the amplitude; squaring and averaging gives the colour factor $C_F=\tfrac43$ and the coupling $\alpha_s=g_s^2/4\pi$ , multiplying the Born rate $\sigma_0$ . The kinematics of a three-body final state are captured by the two scaled energies $x_1,x_2$ , with $x_3=2-x_1-x_2$ . Why: one hard emission is first order in $\alpha_s$ , so the three-jet rate is suppressed by exactly one power of the coupling relative to two-jet — which is why its size measures $\alpha_s$ .

Step 2. Assemble the matrix element. The spinor algebra yields $\dfrac{1}{\sigma_0}\dfrac{d^2\sigma}{dx_1dx_2}=\dfrac{\alpha_s}{2\pi}C_F\dfrac{x_1^2+x_2^2}{(1-x_1)(1-x_2)}$ . Note $1-x_1=x_2x_3(1-\cos\theta_{23})/2$ -type factors: the denominator vanishes when $x_1\to1$ or $x_2\to1$ . Why: $x_1\to1$ means the quark takes nearly all the energy, so the gluon is soft; the vanishing denominator is the infrared (soft) and collinear singularity familiar from bremsstrahlung.

Step 3. Interpret the divergences physically. A soft or collinear gluon cannot be distinguished from no gluon at all — its jet merges with the quark's — so the divergence is not a real infinity but the statement that "three jets" is undefined until you say how well you can resolve them. Introduce the cut $y_{ij}>y_{\text{cut}}$ of equation (6): it excises the singular region and leaves a finite three-jet cross section $\sigma_3(y_{\text{cut}})\propto\alpha_s$ . Why: infrared-safe observables must be insensitive to soft or collinear splitting, and the resolution cut is exactly the operation that enforces this, rendering the perturbative prediction physical.

Step 4. Extract the coupling. The resolved three-jet fraction $\sigma_3/\sigma_{\text{tot}}$ (and, more precisely, the thrust and other event-shape distributions) is linear in $\alpha_s$ at leading order, so fitting the measured rate returns $\alpha_s(\sqrt s)$ . At $\sqrt s\approx34$ GeV, $\alpha_s\approx0.14$ , giving $\alpha_s/\pi\approx0.045$ — a few-percent effect on $R$ but a large, clearly visible population of genuine three-jet events. Why: because the leading dependence is a single power of $\alpha_s$ , the three-jet observables are among the cleanest measurements of the strong coupling and of its running with energy. *Verified: the coefficient $\tfrac{\alpha_s}{2\pi}C_F=\tfrac{2\alpha_s}{3\pi}$ ( $=0.212\,\alpha_s$ ); $x_1+x_2+x_3=2$ ; soft limits at $x_{1,2}\to1$ . Consistent with Ellis, Stirling & Webber §3.2 and Thomson §10.6.*

Go deeper · Derivation C · 2 minFragmentation functions, the momentum sum rule, and scaling violation

Derivation CFragmentation functions, the momentum sum rule, and scaling violation

The purpose is to define the fragmentation function $D_i^h(z)$ that quantifies how a parton's momentum is parcelled out among hadrons, to derive its exact momentum sum rule from energy conservation, and to explain why the naive scaling $D(z)=D(z\text{ only})$ is broken logarithmically by gluon radiation — the same DGLAP evolution that governs parton distributions, run in the opposite direction. We isolate the universal long-distance object, constrain it, and let QCD evolve it.

Step 1. Factorise. The inclusive single-hadron spectrum splits into a calculable short-distance production of parton $i$ times a universal, process-independent probability $D_i^h(z)$ that the parton yields hadron $h$ at momentum fraction $z=2E_h/\sqrt s$ . The same $D_i^h$ measured in $e^+e^-$ describes hadrons in deep inelastic scattering and at hadron colliders. Why: hadronisation is a long-distance process that cannot know how the parton was made, so it is separable and universal — the property that makes jets predictive at all.

Step 2. Impose the momentum sum rule. Every parton must hand all of its momentum to its hadrons, so summing $z$ weighted by $D_i^h$ over all hadron species and integrating gives $\sum_h\int_0^1 z\,D_i^h(z)\,dz=1$ . Why: this is nothing but energy–momentum conservation applied to fragmentation; it fixes the overall normalisation that no perturbative calculation can supply.

Step 3. State leading-order scaling. If the parton fragmented independently of its virtuality, $D(z)$ would depend on $z$ alone — Feynman scaling — and the hadron spectrum would look identical at every collision energy. Why: at the crudest level fragmentation is a fixed branching probability, so the spectrum is energy-independent, mirroring Bjorken scaling in the structure functions.

Step 4. Break the scaling. In QCD a parton radiates gluons before fragmenting, and each emission shares its energy, so at higher $\sqrt s$ there is more phase space for radiation: the $D(z)$ spectrum softens (shifts to smaller $z$ ) and the mean hadron multiplicity rises, $\langle n_{\text{ch}}\rangle$ climbing from $\sim13$ at PETRA ( $34$ GeV) to $\sim21$ at LEP ( $91$ GeV). This evolution is governed by the timelike DGLAP equations. Why: gluon radiation is the calculable QCD correction to naive fragmentation, and its logarithmic build-up is exactly the scaling violation measured across collider energies — a direct test of the running theory. *Verified: sum rule $\sum_h\int_0^1 z D_i^h\,dz=1$ from energy conservation; multiplicity rise $13\to21$ over $34\to91$ GeV is the measured trend. Consistent with Ellis, Stirling & Webber §4 and Thomson §10.6.*

Limiting cases

Case	Result	Meaning
Two-jet limit ( $T\to1$ )	back-to-back pencil, axis $\propto1+\cos^2\theta$	leading-order $q\bar q$ ; the spin-½ angular law
Soft or collinear gluon	$(1-x_1)^{-1}$ or $(1-x_2)^{-1}$ diverges	third jet unresolvable; tamed only by $y_{\text{cut}}$
Symmetric three-jet (Mercedes)	$x_1=x_2=x_3=\tfrac23$ , $T=\tfrac23$	planar event, maximally away from two-jet
Spin-0 quark hypothesis	axis $\propto\sin^2\theta$	vanishes along beam — excluded by data
High energy ( $\sqrt s\uparrow$ )	$\langle n_{\text{ch}}\rangle$ rises, jets narrow	$\alpha_s$ runs down; more, softer hadrons
Gluon jet vs quark jet	$\langle n\rangle_g/\langle n\rangle_q\to C_A/C_F=\tfrac94$	gluon radiates more — broader, richer jet

Common trapReading a jet as if it were the quark itself

It is tempting to treat each jet as a stand-in for a specific quark — to say "this jet is the up quark, that one the anti-down" and read off flavour, charge, or which is quark versus antiquark. Hadronisation erases almost all of that. The colour flow that stitches the $q\bar q$ pair together spreads the original quantum numbers across many hadrons and even across the gap between jets (the "string effect"), so a jet does not carry a clean flavour tag, and the cross section must be summed inclusively over flavours and quoted in $|\cos\theta|$ . Three further cautions ride along. A jet's energy is not exactly the parton's energy — it depends on the jet algorithm and $y_{\text{cut}}$ , and out-of-cone radiation and detector effects shift it. The number of jets is likewise algorithm-dependent: the same event is two jets at loose resolution and three or four at tight resolution, because a soft gluon is physically continuous with no gluon. And you cannot, event by event, say which jet was the gluon — only statistically, since gluon jets are broader and higher-multiplicity ( $C_A/C_F=\tfrac94$ ). What is robust is exactly what the derivations use: directions, rates, and inclusive sums, the infrared-safe quantities that survive fragmentation.

Worked examples

1 · A symmetric three-jet event, fully counted (fully worked)

At $\sqrt s=30$ GeV a $q\bar qg$ event emerges as three coplanar jets $120^\circ$ apart with equal energies. Find each jet energy, the scaled energies $x_i$ , the invariant mass of any jet pair, and the thrust $T$ .

Step 1. Energy sharing. Equal jets split the total energy three ways, $E_i=\sqrt s/3=30/3=\mathbf{10\text{ GeV}}$ each. Energy conservation is automatic since $3\times10=30=\sqrt s$ .

Step 2. Scaled energies. By definition $x_i=2E_i/\sqrt s=2(10)/30=\mathbf{2/3}$ for each jet, and the constraint checks out: $x_1+x_2+x_3=3\times\tfrac23=\mathbf{2}$ , exactly as required for a massless three-body final state.

Step 3. Jet-pair invariant mass. Treating the jets as massless with $120^\circ$ between them, $M_{ij}^2=2E_iE_j(1-\cos120^\circ)=2(10)(10)(1+\tfrac12)=300\text{ GeV}^2$ , so $M_{ij}=\sqrt{300}=\mathbf{17.3\text{ GeV}}$ . Each pair carries a large mass — this event sits deep in the resolved three-jet region, far from any $y_{\text{cut}}$ boundary.

Step 4. Thrust. Aligning the thrust axis with one jet, the projections are $|\mathbf p_i\cdot\hat{\mathbf n}|=p,\,p|\cos120^\circ|,\,p|\cos240^\circ|=p,\tfrac{p}{2},\tfrac{p}{2}$ , summing to $2p$ ; the denominator is $3p$ . Hence $T=2p/3p=\mathbf{2/3}\approx0.667$ , the canonical value for a symmetric three-jet event and far below the two-jet value $T=1$ . Each jet: $E=10$ GeV, $x=2/3$ ; $M_{ij}=17.3$ GeV; $T=2/3$ .

2 · The angular distribution as a spin test

Measured two-jet events give an axis distribution consistent with $1+\cos^2\theta$ . Show quantitatively how this rules out spin-0 quarks by comparing the forward region to $90^\circ$ .

Step 1. Form the discriminating ratio $N(|\cos\theta|=0.9)/N(\cos\theta=0)$ . For spin-½ the shape is $1+\cos^2\theta$ , giving $(1+0.81)/(1+0)=1.81/1=\mathbf{1.81}$ — more events toward the beam.

Step 2. For a spin-0 quark the shape is $\sin^2\theta=1-\cos^2\theta$ , giving $(1-0.81)/(1-0)=0.19/1=\mathbf{0.19}$ — a strong depletion toward the beam.

Step 3. Compare. The two hypotheses differ by a factor $1.81/0.19\approx\mathbf{9.5}$ in this single ratio, far larger than experimental uncertainty. The measured excess of forward jets lands squarely on $1.81$ . The data give $\approx1.8$ : quarks are spin-½, not spin-0.

3 · Quark jets versus gluon jets

Why is a gluon jet broader and richer than a quark jet, and by how much asymptotically?

Step 1. Compare colour charges. A quark radiates gluons with strength set by its Casimir $C_F=\tfrac43$ ; a gluon radiates with the adjoint Casimir $C_A=3$ . The ratio of radiated soft-gluon intensity is therefore $C_A/C_F=3/(4/3)=\mathbf{9/4}=2.25$ .

Step 2. Translate to observables. More radiation means more subsequent splittings, so a gluon jet asymptotically has $9/4$ times the particle multiplicity and a wider angular spread than a quark jet of the same energy. The asymptotic multiplicity ratio is $C_A/C_F=9/4$ .

Step 3. Note the reality. At LEP energies the measured ratio is only $\approx1.5$ , not $2.25$ , because the asymptotic limit is approached slowly (energy-dependent corrections). Still, the sign and size of the effect — gluon jets demonstrably broader — is a direct measurement of the non-Abelian ratio $C_A/C_F$ and hence of the $SU(3)$ colour structure. Gluon jets are broader by a factor approaching $9/4$ , confirming $C_A>C_F$ .

Physics of the portrait

Jets earned their central place in QCD the day the gluon appeared as a third one, and the physicist most identified with pulling that signal out of the data spent a career finding jets that changed the Standard Model.

Origins · 2 minSau Lan Wu and how this was discovered

Sau Lan Wu

born 1940 · led the discovery of the gluon in three-jet events at PETRA

Sau Lan Wu, born in Hong Kong in 1940 and long a professor at the University of Wisconsin–Madison, has a personal motto that doubles as a research programme: make discoveries. In the mid-1970s she worked with Samuel Ting on the experiment that found the $J/\psi$ , opening the charm threshold that shows up as a step in the $R$ ratio. But her signature achievement came at DESY's PETRA collider. Theorists — notably Ellis, Gaillard and Ross in 1976 — had predicted that if the gluon were real and radiated like a coloured gauge boson, hard emission would produce planar three-jet events at a rate set by $\alpha_s$ . Wu, on the TASSO experiment, designed the analysis that would recognise them: she developed jet-finding and event-shape methods robust enough to separate a genuine three-pronged topology from a fluctuating two-jet event. In 1979 TASSO, together with MARK-J, PLUTO and JADE, reported the three-jet events, and with them the gluon — the first direct evidence for the carrier of the strong force and a decisive confirmation of QCD as a non-Abelian gauge theory. The tools she and her contemporaries built — infrared-safe jet algorithms, thrust and sphericity, the statistical separation of quark and gluon jets — are exactly the machinery this article rests on. Decades later Wu turned the same discovery instinct on the Higgs boson, contributing to the ATLAS analyses that found it in 2012, so that her name is attached to three of the era's defining discoveries. Her career is a reminder that a jet is not just a spray of hadrons: read correctly, it is a message from a particle that can never be caught alone.

History

1970

— Bjorken and Brodsky predict that qq̄ production yields two collimated jets of hadrons

1975

— Mark I at SPEAR observes the first two-jet structure in e⁺e⁻ annihilation

1977

— Feynman and Field build a quantitative model of quark fragmentation into hadrons

1978

— Ellis, Gaillard and Ross predict planar three-jet events from hard gluon radiation

1979

— TASSO, MARK-J, PLUTO and JADE at PETRA discover the gluon in three-jet events

Going further

Texts. Thomson, Modern Particle Physics, §10.5–10.6 (QCD in $e^+e^-$ annihilation, jets, and $\alpha_s$ ) · Halzen & Martin, Quarks and Leptons, §13.2 (jets and the parton model) · Griffiths, Introduction to Elementary Particles, 2nd ed., §8.1–8.2 · Ellis, Stirling & Webber, QCD and Collider Physics, §3–4 (three-jet cross section, fragmentation, jet algorithms) · Peskin & Schroeder, An Introduction to QFT, §17.4 (annihilation into hadrons).

Provenance. The point cross section $\sigma=4\pi\alpha^2/3s$ , the $1+\cos^2\theta$ law, and $R=N_c\sum Q_f^2$ are recalled from aq-pp-ee-to-mumu and aq-pp-qcd-in-ee-annihilation; the colour factor $C_F=\tfrac43$ and $C_A=3$ from aq-pp-colour-qcd; the running of $\alpha_s$ from aq-pp-running-alpha-s. The DGLAP evolution of $D(z)$ and the parton shower are owned by aq-pp-parton-showers (forward reference, not re-derived). A verifier should confirm: (1) the normalisation $\tfrac38\int_{-1}^1(1+\cos^2\theta)d\cos\theta=1$ and the forward ratios $1.81$ (spin-½) vs $0.19$ (spin-0); (2) the three-jet coefficient $\tfrac{\alpha_s}{2\pi}C_F=\tfrac{2\alpha_s}{3\pi}$ with $x_1+x_2+x_3=2$ ; (3) the symmetric event $E=10$ GeV, $x=2/3$ , $M_{ij}=17.3$ GeV, $T=2/3$ at $\sqrt s=30$ GeV; (4) $C_A/C_F=9/4$ . Quark charges and the $R$ staircase ( $2,\tfrac{10}{3},\tfrac{11}{3}$ ) are standard.

Prerequisites. Colour and QCD · QCD in $e^+e^-$ Annihilation · $e^+e^-\to\mu^+\mu^-$ · Running $\alpha_s$ . Leads to. Parton Showers · Hadron Collisions · Gluons.

Ask the tutor: Ask me to walk the $q\bar qg$ Dalitz plot and show where the soft and collinear gluon singularities sit, or to trace how the thrust distribution turns a spray of hadrons into a measurement of $\alpha_s$ .

Why do quarks make jets at all instead of flying off freely?

Because the strong force does not weaken with distance the way electromagnetism does. As a quark and antiquark separate, the gluon self-interaction confines the colour field into a narrow flux tube whose energy grows roughly linearly with length, about 1 GeV per femtometre. Long before the quarks travel a nuclear diameter, the tube holds enough energy to create new $q\bar q$ pairs from the vacuum, and it does — repeatedly, snapping into a chain of colour-neutral hadrons. The debris travels along the original quark direction because the quark was moving fast when the tube broke, so what began as one energetic quark ends as a collimated jet. Free quarks are never produced; jets are confinement's way of conserving momentum while hiding colour.

How can jets be calculable if hadronisation is non-perturbative?

Through a separation of scales. The production of the partons and any hard gluon emission happens at short distances where $\alpha_s$ is small and perturbation theory works; hadronisation happens later, at long distances, where it is not calculable. The key fact is that the long-distance step neither changes the number of partons produced nor deflects them much — it only dresses each one into hadrons moving nearly along its direction. So infrared-safe observables — jet directions, rates, inclusive sums, event shapes — are set by the calculable short-distance physics, while the incalculable fragmentation cancels or averages out. This is why you can predict a three-jet rate to percent accuracy without ever computing how a single pion forms.

What actually defines a jet — where does one jet end and another begin?

A jet is defined by an algorithm plus a resolution parameter, not by nature drawing a line. A sequential algorithm computes a distance $y_{ij}$ between every pair of objects (equation 6) and repeatedly merges the closest pair until all remaining pairs are farther apart than a chosen cut $y_{\text{cut}}$ ; whatever clusters remain are the jets. Change $y_{\text{cut}}$ and you change the count: loose resolution sees two jets where tight resolution sees three or four, because a soft or collinear gluon is physically continuous with no gluon at all. What makes this respectable rather than arbitrary is infrared safety — a good algorithm gives the same answer whether or not a very soft or collinear particle is present, so the theory can predict the jets it defines.

Can you tell which jet came from the quark, the antiquark, or the gluon?

Not event by event, and only statistically overall. Hadronisation smears the original flavour and colour across many hadrons and even between jets, so no jet carries a clean "I was the up quark" label — which is exactly why cross sections are summed inclusively over flavour and quoted in $|\cos\theta|$ . The gluon jet is a partial exception, but only in aggregate: because the gluon's colour charge $C_A=3$ exceeds the quark's $C_F=\tfrac43$ , gluon jets are on average broader and contain more particles, approaching a ratio $C_A/C_F=\tfrac94$ . Techniques like tagging the two most energetic jets as the quark pair (the gluon is usually softest) work statistically, and modern taggers exploit substructure, but any single jet's parentage is a probability, not a fact.

Why does a hard gluon give a third jet rather than just widening the two?

It depends on how hard and how wide-angle the gluon is. A soft gluon, or one emitted almost parallel to its parent quark, simply merges into that quark's jet and slightly broadens it — this is the collinear and infrared region where the three-jet rate formally diverges and where "three jets" is not a meaningful description. But a gluon carrying a substantial fraction of the energy at a large angle to both quarks forms its own resolved cluster of hadrons, well separated from the other two, and the event genuinely has three jets. The boundary between "widens a jet" and "makes a new jet" is precisely the resolution cut $y_{\text{cut}}$ , which is why the gluon was discovered as a population of clearly planar, three-pronged events rather than as a subtle broadening.

Do jets at the LHC mean the same thing as jets in e⁺e⁻?

The physics is the same — a parton fragmenting into collimated hadrons — but the environment is messier, so the tools differ. In $e^+e^-$ the whole event comes from a clean $q\bar q$ (or $q\bar qg$ ) system, so global event shapes like thrust work beautifully. At a hadron collider the colliding protons leave beam remnants and there are many simultaneous interactions, so one uses jet algorithms tailored to that clutter — chiefly anti- $k_T$ , which builds circular, pile-up-resistant jets around the hardest particles. The underlying quantities are identical: jet momenta track partons, gluon jets are broader than quark jets by the same colour ratio, and jet rates measure $\alpha_s$ . What changes is the bookkeeping needed to subtract the underlying event, not the meaning of a jet.