QCD in e+e- annihilation

Point an electron at a positron, wind the energy up past a few GeV, and let them annihilate. What comes back is not a clean pair of leptons but a spray of hadrons — pions, kaons, protons — organised into tight, collimated jets. This is the cleanest window physics has onto the strong interaction, and the reason is that the entrance to the reaction is pure QED. The pair annihilates into a single virtual photon of known spin and known coupling, exactly as in $e^+e^-\to\mu^+\mu^-$ ; that photon then materialises a quark and an antiquark. Because there are no hadrons in the initial state there are no parton-distribution uncertainties to unfold, and because the final state carries the full centre-of-mass energy the observed jets map directly onto the underlying $q\bar q$ pair. The messy strong dynamics is quarantined to the final state, sitting behind a production vertex we can compute to five decimals. Three measurements make this window famous. First, the ratio of the hadronic rate to the muon-pair rate, $R=\sigma_{\text{had}}/\sigma_{\mu\mu}$ , is a literal counting experiment: it comes out three times larger than a colourless theory predicts, and that factor of three is the number of quark colours. Second, the ratio climbs in steps as the beam energy crosses each new quark threshold — charm near 3.7 GeV, bottom near 10.5 GeV — a staircase that inventories the quarks. Third, the data sit a stubborn ten percent above the naive plateaus, and that excess is not error: it is the first gluon correction, $R_0(1+\alpha_s/\pi)$ , the same gluon that occasionally kicks out a third jet and lets us weigh the strong coupling. QED sets the trap; QCD springs it.

Key takeaways

$e^+e^-\to\gamma^*\to q\bar q$ is a QED-initiated probe of QCD: no parton-distribution uncertainties, and the jets inherit the full $\sqrt s$ .
The ratio $R=\sigma(e^+e^-\to\text{hadrons})/\sigma(e^+e^-\to\mu^+\mu^-)=N_c\sum_q Q_q^2$ — the multiplicity $N_c=3$ is model-free evidence for colour.
$R$ is a staircase: 2 for uds, 10/3 for udsc, 11/3 for udscb, stepping up at each quark threshold $\sqrt s>2m_q$ .
The angular distribution $\mathrm{d}\sigma/\mathrm{d}\Omega\propto(1+\cos^2\theta)$ shows the quark is spin-½, just like the muon.
The measured ~10% excess over each plateau is the QCD correction $R=R_0(1+\alpha_s(s)/\pi+\dots)$ , an inclusive number that measures $\alpha_s$ .
Real gluon emission $e^+e^-\to q\bar q g$ produces three-jet events; their rate fixes $\alpha_s$ and four-jet angles test the non-abelian triple-gluon vertex.
Key idea: because the QED initiation is exactly calculable, the measured hadronic cross section becomes a direct counting experiment for colour (the factor 3 in $R$ ) and a probe of the strong coupling (the $1+\alpha_s/\pi$ correction and the three-jet rate).

Where this lives today

The $R$ ratio has aged into a precision instrument. Its most consequential modern role is the muon anomalous magnetic moment, $(g-2)_\mu$ : the leading hadronic contribution cannot be computed perturbatively, so it comes from a dispersion integral over the measured $R(s)$ at low energies. The uncertainty on that integral — dominated by $e^+e^-\to\text{hadrons}$ data from BES III, CMD-3, BaBar and KLOE — is currently the limiting theoretical error in confronting the Fermilab $(g-2)$ measurement with the Standard Model, and a tension between the data-driven $R$ and lattice QCD is one of the field's live puzzles. Meanwhile the jet observables born here — thrust, jet rates, event shapes — remain the sharpest $e^+e^-$ determinations of $\alpha_s$ , and the four-jet angular correlations measured at LEP confirmed $C_A/C_F=9/4$ , directly verifying the triple-gluon vertex that distinguishes QCD from an abelian theory. Proposed colliders — FCC-ee, CEPC, ILC — would revisit this physics with per-mille precision.

The physical picture

The reaction begins as textbook QED. An electron and positron annihilate into a single timelike virtual photon carrying the entire energy $\sqrt s$ and spin 1. That photon has no memory of what it will decay into; it couples to any charged pair with strength set by the electric charge. Decaying to $\mu^+\mu^-$ gives a lepton pair; decaying to $q\bar q$ gives a quark and antiquark flying apart back-to-back. But a bare colour charge cannot exist alone: as the pair separates, the colour field between them stretches, stores energy, and snaps into new $q\bar q$ pairs that dress into hadrons — two narrow jets pointing along the original quark directions. The first figure is the observable that turns this into a colour census: $R$ versus beam energy.

Figure

The R-ratio staircase: R = N_c Σ Q_q² climbs from 2 (uds) to 10/3 (udsc) at the charm threshold near 3.7 GeV to 11/3 (udscb) at the bottom threshold near 10.5 GeV. Narrow resonances (J/ψ, Υ) spike between the steps, and the measured continuum sits about 10% above each plateau — the QCD correction 1 + α_s/π

The staircase settles the colour question, but the excess and the occasional third jet reveal the gluon. Most of the time the quark and antiquark leave two back-to-back jets, but sometimes one radiates a hard gluon before hadronising — QCD bremsstrahlung — and that gluon fragments into a jet of its own, giving three coplanar jets in a Mercedes-star pattern. The rate of these three-jet events, relative to two-jet events, is proportional to $\alpha_s$ : harder, more frequent gluon radiation means a stronger coupling. The second figure contrasts the two topologies and names the quantity, thrust, that separates pencil-like two-jet events from spread-out three-jet ones.

Figure

Two-jet versus three-jet events. Usually q and q̄ leave back-to-back jets with thrust T → 1. When a hard gluon is radiated (q q̄ g) a third jet appears in the plane; a symmetric Mercedes event has thrust T = 2/3. The three-jet fraction is proportional to α_s

The core equations

Everything starts from the QED yardstick — the cross section for annihilating into a point-like spin-½ pair of unit charge. Far above the muon mass and far below the $Z$ pole, where a single photon dominates, the muon-pair cross section is a clean function of $s$ alone.

\sigma(e^+e^-\to\mu^+\mu^-)=\frac{4\pi\alpha^2}{3s} \qquad (1)

where: $\alpha\approx1/137$ — the fine-structure constant; $s$ — the squared centre-of-mass energy; the prefactor evaluates to $86.9\ \text{nb}\,\text{GeV}^2/s$ . Reading: the rate falls as $1/s$ , and this is the denominator every hadronic measurement is divided by. The hadronic cross section has the identical structure — the same photon, the same $1/s$ fall-off — differing only by the quark's fractional charge $Q_q$ and by the fact that each quark comes in $N_c$ colours, all produced. Dividing by equation (1) cancels the kinematics and leaves a pure number.

R\equiv\frac{\sigma(e^+e^-\to\text{hadrons})}{\sigma(e^+e^-\to\mu^+\mu^-)}=N_c\sum_{q:\,2m_q<\sqrt s}Q_q^2 \qquad (2)

where: $N_c=3$ — the number of colours; $Q_q$ — the quark charge in units of $e$ ( $+\tfrac23$ for u,c,t and $-\tfrac13$ for d,s,b); the sum runs over flavours light enough to pair-produce. Reading: $R$ is a dimensionless census of the accessible quarks, and $N_c$ multiplies every plateau — remove colour and the prediction falls threefold below data. The final state also reports the quark's spin through its angular distribution. A spin-½ fermion produced by a transverse virtual photon comes out with the same shape as the muon.

\frac{\mathrm{d}\sigma}{\mathrm{d}\Omega}\propto\bigl(1+\cos^2\theta\bigr) \qquad (3)

where: $\theta$ — the angle between the outgoing quark and the beam axis. Reading: the $1+\cos^2\theta$ shape is the fingerprint of a spin-½ point particle; spin-0 quarks would give $\sin^2\theta$ , which the jet-axis data rule out. Precise data lie above the parton-model plateaus, and the excess is the perturbative QCD correction — a power series in $\alpha_s$ evaluated at the collision scale.

R=R_0\left[1+\frac{\alpha_s(s)}{\pi}+1.41\left(\frac{\alpha_s}{\pi}\right)^2-12.8\left(\frac{\alpha_s}{\pi}\right)^3+\dots\right] \qquad (4)

where: $R_0=N_c\sum Q_q^2$ — the parton-model value; $\alpha_s(s)$ — the running strong coupling at the scale $\sqrt s$ . Reading: the leading $\alpha_s/\pi$ term is a ~5% bump at PETRA energies that shrinks as $\alpha_s$ runs down with energy; because $R$ is inclusive it is one of the cleanest ways to extract $\alpha_s$ . The gluon that causes that correction sometimes appears as a resolved third jet. Its distribution across the Dalitz plane, in the energy fractions $x_i=2E_i/\sqrt s$ , is the order- $\alpha_s$ matrix element.

\frac{1}{\sigma_0}\frac{\mathrm{d}\sigma}{\mathrm{d}x_1\,\mathrm{d}x_2}=\frac{\alpha_s}{2\pi}\,C_F\,\frac{x_1^2+x_2^2}{(1-x_1)(1-x_2)} \qquad (5)

where: $x_1,x_2$ — quark and antiquark energy fractions ( $x_1+x_2+x_3=2$ ); $C_F=\tfrac43$ — the colour factor for gluon emission off a quark; $\sigma_0$ — the lowest-order $q\bar q$ cross section. Reading: the denominators blow up as $x_1\to1$ or $x_2\to1$ , i.e. when the gluon becomes soft or collinear — those poles are why bremsstrahlung is copious and why the correction is a logarithm-enhanced $\alpha_s$ . To separate two-jet from three-jet events without an arbitrary cone, one uses an infrared-safe event shape. Thrust projects all momenta onto the axis that maximises their summed longitudinal component.

T=\max_{\hat n}\frac{\sum_i|\vec p_i\cdot\hat n|}{\sum_i|\vec p_i|},\qquad \tfrac12\le T\le1 \qquad (6)

where: $\hat n$ — the thrust axis; the sum runs over all final hadrons. Reading: a back-to-back two-jet event has $T=1$ (a pencil), a symmetric three-jet event $T=\tfrac23$ , an isotropic spray $T\to\tfrac12$ ; the distribution $\mathrm{d}\sigma/\mathrm{d}T$ is a workhorse for measuring $\alpha_s$ .

Derivation AFrom the muon cross section to the colour count

The goal is to obtain the ratio $R=N_c\sum_q Q_q^2$ from first principles, starting from the QED cross section for $e^+e^-\to\mu^+\mu^-$ and asking what changes when the produced pair is a quark–antiquark pair. The strategy is to isolate the three quantities that differ — the charge at the final vertex, the sum over kinematically open flavours, and the colour multiplicity from summing over the internal colour label — and show that the ratio is therefore a pure number with the colour factor out front, so measuring it counts colours directly.

Step 1. Write the muon cross section as a product of a universal QED core and a final-vertex charge factor. At one photon exchange, $\sigma(e^+e^-\to f\bar f)=Q_f^2\,\dfrac{4\pi\alpha^2}{3s}$ for any point-like spin-½ fermion $f$ of charge $Q_f$ ; the muon has $Q_\mu=1$ , recovering equation (1). Why: the photon couples to the final fermion only through its charge, so replacing the muon by any other charged spin-½ particle changes nothing but the factor $Q_f^2$ and, later, phase-space thresholds.

Step 2. Replace the muon by a quark of flavour $q$ and charge $Q_q$ . The single-flavour, single-colour cross section is $\sigma_q=Q_q^2\,\dfrac{4\pi\alpha^2}{3s}$ . But a quark carries a hidden colour index, and the photon produces the pair in every colour with equal amplitude, so we must sum $|{\cal M}|^2$ over the $N_c$ colour states: $\sigma_q\to N_c\,Q_q^2\,\dfrac{4\pi\alpha^2}{3s}$ . Why: colour is a good internal quantum number the photon does not resolve, so the three colour channels are distinct final states that add incoherently — this is where the factor of three enters, and it enters as a multiplicity, not a coupling.

Step 3. Sum over all flavours the beam energy can produce. A flavour contributes only if the pair can be made on shell, $\sqrt s>2m_q$ , so $\sigma_{\text{had}}=N_c\,\dfrac{4\pi\alpha^2}{3s}\sum_{q:\,2m_q<\sqrt s}Q_q^2$ . Why: below a threshold the corresponding quark simply is not produced, so $R$ is a step function that jumps up each time a new flavour opens — the staircase of the first figure.

Step 4. Divide by the muon cross section (1). The universal core $4\pi\alpha^2/3s$ cancels exactly, leaving $R=N_c\sum_q Q_q^2$ . Evaluate the plateaus: below charm, $R=3(\tfrac49+\tfrac19+\tfrac19)=3\cdot\tfrac23=2$ ; above charm, $R=3\cdot\tfrac{10}{9}=\tfrac{10}{3}$ ; above bottom, $R=3\cdot\tfrac{11}{9}=\tfrac{11}{3}$ . Why: the cancellation is the whole point — all the messy kinematics divides out and what remains is a clean count, so the observed $R\approx2$ below charm can only be reconciled with the quark charges if $N_c=3$ . *Verified: R(uds)=2, R(udsc)=10/3, R(udscb)=11/3 with N_c=3; without colour the predictions would be 2/3, 10/9, 11/9 — a factor 3 below data.*

Go deeper · Derivation B · 2 minWhy real and virtual gluons conspire to give 1 + α_s/π

Derivation BWhy real and virtual gluons conspire to give 1 + α_s/π

The goal is to explain how the finite correction factor $(1+\alpha_s/\pi)$ arises even though each of its two ingredients — real gluon emission $e^+e^-\to q\bar q g$ and the virtual gluon loop correcting $e^+e^-\to q\bar q$ — is separately infrared-divergent. The plan is to identify the soft and collinear singularities in the real-emission rate, note that the virtual correction carries equal and opposite singularities, and invoke the Kinoshita–Lee–Nauenberg theorem: any sufficiently inclusive observable, such as the total hadronic cross section, is finite because those divergences cancel in the sum over degenerate final states.

Step 1. Integrate the real-emission matrix element (5) over the Dalitz plane. The denominators $(1-x_1)(1-x_2)$ produce divergences at the boundaries: a soft singularity where the gluon energy $x_3\to0$ , and collinear singularities where the gluon lines up with the quark or antiquark. Regulated, the real contribution behaves as $\sigma_{\text{real}}\sim\sigma_0\,\dfrac{\alpha_s}{\pi}C_F\big(\text{poles}+\text{finite}\big)$ . Why: a massless gluon costs no energy in the soft limit and no transverse momentum in the collinear limit, so the phase space for emitting one diverges — physically, a colour charge always radiates infinitely many very-soft gluons.

Step 2. Compute the virtual correction: the one-loop vertex graph where a gluon is emitted and reabsorbed by the $q\bar q$ pair. Interfered with the tree amplitude it contributes at the same order $\alpha_s$ , and its integral over the loop momentum carries soft and collinear poles that are equal in magnitude and opposite in sign to those of Step 1. Why: the virtual gluon explores the same soft and collinear regions of momentum space as the real one, and unitarity guarantees the two are two faces of the same physics.

Step 3. Add them. A detector cannot distinguish a $q\bar q$ pair from a $q\bar q$ pair plus an undetectably soft or exactly collinear gluon — they are the same event. Summing over these degenerate final states, the poles cancel by the KLN theorem and the remainder is finite: $\sigma_{\text{had}}=\sigma_0\big(1+\tfrac{\alpha_s}{\pi}\big)$ . Why: the cancellation is not a coincidence but a theorem — any infrared-safe observable, one that does not change under soft emission or collinear splitting, is calculable in perturbative QCD despite the singular pieces.

Step 4. Divide by the muon cross section to update the ratio: $R=R_0\big(1+\alpha_s(s)/\pi+\dots\big)$ , equation (4). At $\sqrt s\approx34$ GeV, $\alpha_s\approx0.14$ , so the correction is $(1+0.14/\pi)\approx1.045$ — the ~5% lift seen above the top plateau. Why: because $\alpha_s$ runs, the size of the bump depends on energy, and matching the measured $R$ to (4) turns the total hadronic rate into a clean determination of the strong coupling. *Verified: individual real and virtual pieces are IR-divergent; their sum gives the finite factor 1 + α_s/π; numerically (11/3)(1+0.14/π) ≈ 3.83.*

Go deeper · Derivation C · 2 minReading α_s off the three-jet rate

Derivation CReading α_s off the three-jet rate

The goal is to show how resolved three-jet events isolate the gluon and provide a direct measurement of $\alpha_s$ , and why the same matrix element that produces the two-jet-dominated sample also has a well-defined hard three-jet region. The approach is to use the Dalitz energy fractions $x_i=2E_i/\sqrt s$ to map the kinematics of $q\bar q g$ , identify where events are two-jet-like versus genuinely three-jet, and define a resolution criterion that makes the three-jet fraction a finite, calculable multiple of $\alpha_s$ .

Step 1. Parameterise the three-body final state by the energy fractions $x_1,x_2,x_3$ (quark, antiquark, gluon), constrained by energy conservation to $x_1+x_2+x_3=2$ and each $0\le x_i\le1$ . The allowed region is a triangle; the physical Dalitz region for massless partons fills it. Why: two independent fractions fix the event shape entirely, so the single matrix element (5) contains all three-parton configurations from nearly-two-jet to symmetric Mercedes.

Step 2. Locate the singular edges. The matrix element (5) peaks where $x_1\to1$ or $x_2\to1$ — the corners and edges of the triangle — which correspond to a soft gluon ( $x_3\to0$ ) or a gluon collinear with a quark. These are precisely the two-jet-like configurations. Why: the bulk of the cross section piles up against the two-jet limit, which is why most events look like two jets even though a gluon is almost always present at some softness.

Step 3. Impose a jet-resolution cut to define "three-jet". Require the invariant mass between every pair of partons to exceed a fraction $y_{\text{cut}}$ of $s$ , i.e. $y_{ij}=m_{ij}^2/s>y_{\text{cut}}$ ; this excludes the soft and collinear corners and leaves a finite integral. The surviving fraction is $f_3=\dfrac{\alpha_s}{2\pi}C_F\,g(y_{\text{cut}})$ for a calculable function $g$ . Why: an infrared-safe definition is mandatory — without $y_{\text{cut}}$ the three-jet rate would diverge, but with it the rate is a finite pure multiple of $\alpha_s$ , so counting three-jet events measures the coupling.

Step 4. Invert to get the coupling: $\alpha_s\propto f_3/f_2$ at fixed $y_{\text{cut}}$ , and equivalently fit the thrust or jet-mass distributions. The PETRA and LEP experiments did exactly this, and the values agree with those from $R$ and from the $Z$ lineshape. Why: multiple independent observables built on the same $O(\alpha_s)$ physics all return one $\alpha_s(s)$ , and their consistency — plus its decrease with energy — is a direct confirmation of asymptotic freedom. *Verified: x₁+x₂+x₃=2; matrix element (5) singular as x₁,x₂→1 (soft/collinear two-jet limit); three-jet fraction finite and ∝ α_s once y_cut regulates the edges.*

Limiting cases

Case	Result	Meaning
Below charm threshold	$R=2$	only u,d,s open: $3(\tfrac49+\tfrac19+\tfrac19)=2$ — the cleanest colour count
Above charm ( $\sqrt s\gtrsim4$ GeV)	$R=\tfrac{10}{3}$	c added: $R$ jumps by $3\cdot\tfrac49=\tfrac43$ at the charm step
Above bottom ( $\sqrt s\gtrsim11$ GeV)	$R=\tfrac{11}{3}$	b added: jumps by $3\cdot\tfrac19=\tfrac13$ ; the continuum plateau at PETRA/PEP
On a narrow resonance	$R$ spikes far above plateau	$J/\psi$ , $\Upsilon$ : bound $c\bar c$ , $b\bar b$ states, not the continuum — excluded from the parton count
High energy, one photon	$R\to R_0(1+\alpha_s(s)/\pi)$ , $\alpha_s\to0$	asymptotic freedom: the correction shrinks logarithmically as $\sqrt s\to\infty$
Near the $Z$ pole (91 GeV)	photon formula fails	$Z$ exchange and $\gamma$ – $Z$ interference dominate; use the full electroweak $R_Z$

Common trapThe 10% excess over each plateau is experimental error

It is not error — it is the signal. The parton-model prediction $R=N_c\sum Q_q^2$ gives the plateaus 2, 10/3, 11/3, but the continuum data sit reproducibly a few percent above them, and that offset is the genuine perturbative-QCD correction $R_0(1+\alpha_s(s)/\pi+\dots)$ from gluon radiation. Two related traps follow. First, $R$ is not a fixed number: it is a staircase that steps up at each flavour threshold, so quoting "R ≈ 3.9" without an energy is meaningless. Second, the narrow resonance spikes ( $J/\psi$ at 3.10 GeV, the $\Upsilon$ family near 9.5–10.6 GeV) are not part of the smooth count — they are bound quarkonium states on top of the continuum, and you must sit between them to read the plateau. Treat the excess as physics and the resonances as separate, and $R$ becomes both a colour census and an $\alpha_s$ -meter.

Worked examples

1 · The colour count and the size of the cross section (fully worked)

At $\sqrt s=34$ GeV (a PETRA continuum energy, above the bottom threshold and below the $Z$ ), predict $R$ , then the actual hadronic cross section in picobarns, and state what the measurement of $R\approx2$ below charm proves.

Step 1. Count the open flavours. At 34 GeV the accessible quarks are u, d, s, c, b (top, at 173 GeV, is far out of reach). Their squared charges: $Q_u^2=Q_c^2=\tfrac49$ , $Q_d^2=Q_s^2=Q_b^2=\tfrac19$ . Sum $=\tfrac49+\tfrac19+\tfrac19+\tfrac49+\tfrac19=\tfrac{11}{9}$ .

Step 2. Apply colour: $R_0=N_c\sum Q_q^2=3\cdot\tfrac{11}{9}=\mathbf{11/3\approx3.67}$ . Add the QCD correction with $\alpha_s(34\text{ GeV})\approx0.14$ : $R=\tfrac{11}{3}(1+0.14/\pi)=3.667\times1.0446=\mathbf{3.83}$ , matching the measured continuum.

Step 3. Turn $R$ into a cross section. The muon yardstick is $\sigma_{\mu\mu}=86.9\ \text{nb}\,\text{GeV}^2/s=86.9/34^2=86.9/1156=0.0752\ \text{nb}=\mathbf{75\ pb}$ . Hence $\sigma_{\text{had}}=R\,\sigma_{\mu\mu}=3.83\times75=\mathbf{288\ pb}$ .

Step 4. State the colour proof. Below charm the same arithmetic gives $R_0=3\cdot\tfrac23=2$ ; a colourless theory ( $N_c=1$ ) predicts $R=\tfrac23\approx0.67$ . The measured $R\approx2$ is three times larger, with no free parameter to absorb the gap — the quark charges are fixed by the hadron spectrum. The factor 3 is colour: $N_c=3$ , read directly off $R$ .

2 · Extracting the strong coupling from R

A continuum measurement returns $R=3.84\pm0.05$ at $\sqrt s=34$ GeV. Extract $\alpha_s$ using the leading QCD correction, and comment on the precision.

Step 1. Use $R=R_0(1+\alpha_s/\pi)$ with $R_0=11/3=3.667$ . Solve: $\alpha_s/\pi=R/R_0-1=3.84/3.667-1=1.0472-1=0.0472$ .

Step 2. Multiply by $\pi$ : $\alpha_s=0.0472\times3.1416=\mathbf{0.148}$ . This is a sensible value at 34 GeV, consistent with the running from the modern $\alpha_s(M_Z)=0.118$ .

Step 3. Comment on precision. The extraction hinges on the difference between $R$ and $R_0$ , so a 1.3% error on $R$ becomes a ~30% error on $\alpha_s$ : $\alpha_s=0.148\pm0.043$ . Inclusive $R$ gives a clean but statistically blunt $\alpha_s$ ; three-jet rates and event shapes, where the QCD effect is $O(1)$ not $O(5\%)$ , are far more sensitive — which is why the sharpest $e^+e^-$ determinations of $\alpha_s$ come from jets, not from $R$ .

3 · Thrust of a symmetric three-jet event

A three-jet event has its three jets at 120° in a plane with equal energies $E_1=E_2=E_3=\sqrt s/3$ . Show that its thrust is $T=2/3$ .

Step 2. Project. Jet 1 contributes $|\vec p_1\cdot\hat n|=\sqrt s/3$ . Jets 2 and 3 are at 120° to $\hat n$ , so each projects $|\cos120°|=\tfrac12$ : contribution $\tfrac{\sqrt s}{3}\cdot\tfrac12=\tfrac{\sqrt s}{6}$ each.

Step 3. Sum and divide: numerator $=\tfrac{\sqrt s}{3}+2\cdot\tfrac{\sqrt s}{6}=\tfrac{\sqrt s}{3}+\tfrac{\sqrt s}{3}=\tfrac{2\sqrt s}{3}$ , so $T=\dfrac{2\sqrt s/3}{\sqrt s}=\mathbf{2/3}$ . A symmetric Mercedes event sits at the low-thrust edge $T=2/3$ , maximally far from the two-jet value $T=1$ — which is why the thrust distribution's approach to $T=1$ measures the two-jet-to-three-jet transition, and hence $\alpha_s$ .

Physics of the portrait

The ratio $R$ answers a question of principle with a single plotted number. It turned colour from a bookkeeping device — invented to save the Pauli principle in the $\Delta^{++}$ — into a directly counted quantum number, and the same plot weighs the strong coupling.

Origins · 1 minthe R ratio and how this was discovered

the R ratio

R = σ(hadrons)/σ(μμ) = N_c Σ Q_q², the colour-counting observable · QCD in electron–positron annihilation

Colour was proposed in the mid-1960s to rescue quark states such as the spin-³⁄₂ $\Delta^{++}=uuu$ , whose symmetric wavefunction violated Fermi statistics unless the three quarks carried a hidden three-valued label — a step taken by Oscar Greenberg (1964) with parastatistics and by Moo-Young Han and Yoichiro Nambu (1965), who gauged an SU(3) of colour. For years this was a theorist's fix with no direct measurement. The ratio $R$ changed that. Because the QED vertex is exactly known, $R=N_c\sum Q_q^2$ isolates the colour multiplicity as an overall factor, and the early SPEAR and DORIS data below charm sat at $R\approx2$ , not $\tfrac23$ — colour, counted. The same experiments mapped the staircase: the charm step announced by the $J/\psi$ (the November Revolution of 1974), the bottom step by the $\Upsilon$ (1977). Then the continuum refused to sit on the plateaus, running ~10% high, and three-jet events appeared at PETRA in 1979 — the gluon, radiated as QCD bremsstrahlung, made visible. In one observable, measured across two decades of energy, $R$ delivered the colour factor, the quark charges, the quark inventory, the strong coupling, and the gluon. It remains a precision input today, feeding the hadronic vacuum polarization that limits the prediction of the muon's anomalous magnetic moment.

History

1964

— Greenberg introduces parastatistics; Gell-Mann and Zweig propose quarks

1965

— Han and Nambu gauge an SU(3) colour symmetry

1970

— early SPEAR-era measurements of R hint at the factor of three

1974

— the J/ψ discovered at SLAC and Brookhaven opens the charm step in R

1975

— Mark I finds the τ lepton and confirms R ≈ 2 below charm

1979

— TASSO and the other PETRA experiments discover three-jet events, the gluon

1989

— LEP and SLC turn on at the Z, measuring R to sub-percent precision

Going further

Texts. Thomson, Modern Particle Physics, §10.6 (QCD in $e^+e^-$ annihilation) and §6 (the $e^+e^-\to\mu^+\mu^-$ cross section, Eqs. 10.18–10.22) · Halzen & Martin, Quarks and Leptons, §11–12 · Peskin & Schroeder, An Introduction to QFT, §5.1 and §17.2 (the $R$ ratio and its QCD correction) · Griffiths, Introduction to Elementary Particles, §8–9 · Ellis, Stirling & Webber, QCD and Collider Physics, Ch. 3.

Provenance. The QED core $\sigma_{\mu\mu}=4\pi\alpha^2/3s$ is recalled from aq-pp-ee-to-mumu; the colour factor $N_c=3$ and $C_F=4/3$ from aq-pp-colour-qcd; the running $\alpha_s(s)$ from aq-pp-running-alpha-s; the $1+\cos^2\theta$ shape from aq-pp-differential-cross-sections. A verifier should confirm: $R=2,10/3,11/3$ with $N_c=3$ ; $\sigma_{\mu\mu}=86.9\ \text{nb}\,\text{GeV}^2/s=75$ pb at 34 GeV; $(11/3)(1+\alpha_s/\pi)=3.83$ at $\alpha_s=0.14$ ; $\alpha_s=0.148$ from $R=3.84$ ; thrust $T=2/3$ for the symmetric three-jet event; $C_A/C_F=9/4$ . Quark charges and masses are standard PDG values.

Prerequisites. e+e- → μ+μ- · Colour in QCD · Running of α_s · Differential cross sections. Leads to. Jets and Fragmentation · Asymptotic Freedom · Quarkonium · The Z Resonance at LEP.

Ask the tutor: Ask me why the factor of three in R is impossible to fake, or to derive the soft-gluon poles in the three-jet matrix element step by step.

Why is e+e- annihilation called a "clean" probe of QCD when the final state is a mess of hadrons?

"Clean" refers to the entrance, not the exit. The initial state is two point-like leptons, so the annihilation into a virtual photon is pure QED — no parton distribution functions to unfold as in proton collisions, and no spectator remnants down the beampipe. The full energy goes into the $q\bar q$ pair, so the jets carry the hard-interaction kinematics directly. The hadronisation that turns quarks into pions is complicated but soft and universal, and — for inclusive quantities like $R$ or infrared-safe event shapes — either divides out or is calculable. The strong physics is isolated behind a known door.

If quarks are never seen free, how can R count them?

Because $R$ is an inclusive rate, not a flavour-tagged one. You need not identify which quark was produced event-by-event; you only add up all hadronic final states and divide by the muon rate. The probability to produce any hadrons is, at the moment of annihilation, the probability to produce a $q\bar q$ pair — hadronisation happens later with unit probability and does not change the count. This is quark–hadron duality: summed over enough final states and away from resonances, the hadronic cross section equals the partonic one. So $R$ counts the quarks that could be produced, weighted by charge and colour, even though each immediately dresses into a jet.

What exactly is the difference between R jumping at a threshold and a resonance spike?

A threshold step is a permanent increase in the continuum level: once $\sqrt s$ exceeds $2m_q$ , that flavour is produced for all higher energies, so $R$ rises to a new plateau and stays. A resonance spike is a narrow, temporary enhancement at one energy where a $q\bar q$ bound state — the $J/\psi$ ( $c\bar c$ ) or the $\Upsilon$ ( $b\bar b$ ) — forms directly; it sits on the continuum but between resonances $R$ returns to its plateau. When you quote the parton-model $R$ you sit in the continuum, away from the spikes, which are studied separately as quarkonium spectroscopy.

Where does the number 1 + α_s/π actually come from?

It is the sum of two order- $\alpha_s$ diagrams that are each infinite alone. The first is real gluon emission, $e^+e^-\to q\bar q g$ , whose rate diverges when the gluon is soft or collinear with a quark. The second is the virtual correction, a gluon emitted and reabsorbed within the $q\bar q$ pair, carrying exactly the opposite infinities. Because a detector cannot tell a $q\bar q$ event from one with an unmeasurably soft gluon, the physical quantity is their sum, and there the infinities cancel (the KLN theorem). What survives is a finite $\alpha_s/\pi$ multiplying the lowest-order rate — small, a few percent, because $\alpha_s$ is a few tenths and $\pi\approx3$ .

Does R ever include the top quark, and why does it stop at 11/3?

Not at any energy $e^+e^-$ colliders have reached. The top mass is about 173 GeV, so a $t\bar t$ pair needs $\sqrt s>346$ GeV — beyond LEP2's 209 GeV. Below that the heaviest accessible flavour is bottom, so the count tops out at $R=3(\tfrac49+\tfrac19+\tfrac19+\tfrac49+\tfrac19)=11/3$ . A future collider crossing the $t\bar t$ threshold would step $R$ up by $3\cdot(\tfrac23)^2=\tfrac43$ to $5$ . Note also that near the $Z$ pole the photon formula breaks down — $Z$ exchange dominates — so the "11/3" plateau refers to the photon-dominated continuum below ~60 GeV.

How do four-jet events test QCD beyond just measuring α_s?

Four-jet events arise two ways: a quark radiates two gluons, or a gluon splits ( $g\to gg$ or $g\to q\bar q$ ). Crucially, gluon splitting into two gluons exists only because QCD is non-abelian, with a triple-gluon vertex that photons lack. The angular correlations between the four jets (the Bengtsson–Zerwas and Nachtmann–Reiter angles) depend on the colour factors $C_A$ , $C_F$ , $T_F$ . Measuring them at LEP returned $C_A/C_F=2.25$ , exactly the SU(3) value $9/4$ , ruling out abelian alternatives. So four-jet events confirm the gauge group of QCD.