Two things happen at very different times: a pair is made, and much later it turns into hadrons. Every calculation in this chapter lives in the gap between them.
🎯 Why this matters
That separation is assumed by every collider prediction ever made. A process nobody can calculate is allowed to sit inside one that can be, provided the two happen on different timescales and neither tells the other anything.Chapter 5 ended with a coupling constant that grows as you look closer. This chapter is about the interaction where it does the opposite — and it starts, as the book’s best chapters do, with an experiment that nobody could explain.
In 1969 the ADONE collider at Frascati measured the rate of hadron production in annihilation and found it several times larger than anyone expected. The measurement was right. The expectation was missing a factor of three, and that factor is the subject of this chapter.
The two-stage picture
Nothing about is mysterious once you split it in two:
The first stage is electromagnetic and was solved in §5.7: a virtual photon materialises as a fermion pair, exactly as it does for muons, with and a distribution. Any quark with can be made.
The second stage is hadronization hadronization the process by which a quark leaving a collision converts the colour-field energy into qq̄ pairs and emerges as a collimated jet of hadrons carrying its original momentum. defined in §4.10 — open in glossary , and it is where the strong interaction takes over. The quark and antiquark fly apart, each radiating gluons, which make more quark pairs, which radiate more gluons — until everything has assembled into colour-neutral hadrons.
💡 What this really says — why hadronization cannot spoil the counting
Hadronization is complicated, non-perturbative, and not calculable from first principles. It ought to ruin any hope of using this process to count quarks. It does not, and the reason is a separation of timescales.
The pair is produced electromagnetically in a time . It hadronizes far later, over — a thousand times longer at PETRA energies. By then the number of pairs made is already fixed. Hadronization decides what comes out; it cannot change how many pairs went in.
This is why the total hadronic cross-section is a clean observable while any individual hadron’s spectrum is not — and it is the same argument that let §5.7 read quark charges off Fig. 5.27 despite never seeing a quark.
Fig. 6.1 — the process, in two stages
Click a vertex or an internal line.
Only the two vertices are calculable. Everything to the right of the second one is hadronization, and the point of the section is that it does not matter.
What a jet is, and why it needs a big collider
The quark’s momentum is and points in a definite direction. The hadrons it turns into inherit that direction — that spray is a jet jet the narrow spray of hadrons that a quark or gluon turns into, carrying essentially the parton's energy and direction; hadronic calorimeters exist to measure it. defined in §1.13d — open in glossary — but each picks up a transverse momentum of 0.5 to 1 GeV — a number set by hadronization, not by the collision. So the spray has an opening angle, Eq. (6.3):
The opening angle of a jet, with √s in GeV. Two quantities of completely different origin divided by each other: a number fixed by the strong interaction, over a number fixed by the machine you built.
Every symbol, one at a time
Hover or tap a symbol above — it lights up in the equation and its meaning, units and type appear here.
with in GeV. The numerator is fixed and the denominator is yours to choose. Build a bigger collider and the jet gets narrower, which is the whole reason jets are a 1979 discovery rather than a 1969 one.
e⁺e⁻ → hadrons, transverse projection
The transverse momentum about the jet axis is fixed at half a GeV or so by hadronization and does not care how hard the collision was — so the cone closes as θ ≈ p_T/(√s/2), Eq. (6.3). Slide down to a few GeV and the two cones swell until they meet: the hadrons are still there and the total cross-section is still measurable, but no jet axis can be assigned. That is exactly why R was measured at ADONE in 1969 and jets had to wait for PETRA in 1979.
The transverse projection, as in the book's Figs. 6.2 and 6.6 — recreated rather than reproduced, so the energy can be changed. At √s = 30 GeV (PETRA, where this was done) a jet is a few degrees wide. Slide down towards the ADONE range of 1.6–3 GeV and the cones swell past 30° and merge: the hadrons are still there, R is still measurable, but there is no axis to fit. Then switch on the third jet.
the opening angle, and where jets become possible
import numpy as np
# Eq. (6.3): the hadrons keep a transverse momentum of 0.5-1 GeV about the quark
# direction whatever the collision energy, so the cone closes as 1/sqrt(s).
print(" sqrt(s) jet half-angle what the event looks like")
for rs in (3.0, 7.0, 10.0, 30.0, 91.2, 200.0):
lo = np.degrees(np.arctan(0.5 / (rs / 2)))
hi = np.degrees(np.arctan(1.0 / (rs / 2)))
look = ("two cones fill the event -- no axis" if hi > 25 else
"resolvable" if hi > 5 else "a pencil")
print(f" {rs:6.1f} GeV {lo:5.1f} to {hi:5.1f} deg {look}")
print()
print("ADONE ran at 1.6-3 GeV and PETRA at 30 GeV. That is the whole story of")
print("why R was measured a decade before anyone saw a jet.")sqrt(s) jet half-angle what the event looks like 3.0 GeV 18.4 to 33.7 deg two cones fill the event -- no axis 7.0 GeV 8.1 to 15.9 deg resolvable 10.0 GeV 5.7 to 11.3 deg resolvable 30.0 GeV 1.9 to 3.8 deg a pencil 91.2 GeV 0.6 to 1.3 deg a pencil 200.0 GeV 0.3 to 0.6 deg a pencil ADONE ran at 1.6-3 GeV and PETRA at 30 GeV. That is the whole story of why R was measured a decade before anyone saw a jet.
R R ratio σ(e⁺e⁻→hadrons)/σ(e⁺e⁻→μ⁺μ⁻) = 3Σz_f², the cleanest measurement of the number of colours; 2, 10/3, 11/3 as charm and beauty thresholds open. defined in §6.1 — open in glossary : the observable that counts colours
Divide the hadronic cross-section by the point-like one from §5.7 and everything kinematic cancels, Eq. (6.4):
Bettini pp. 228–229. Every dynamical factor divides out; what is left is a headcount weighted by charge squared.
Every symbol, one at a time
Hover or tap a symbol above — it lights up in the equation and its meaning, units and type appear here.
the staircase, with and without colour
q = {'u': 2/3, 'd': -1/3, 's': -1/3, 'c': 2/3, 'b': -1/3}
def R(flavours, Nc):
return Nc * sum(q[f]**2 for f in flavours)
print(" threshold region 1 colour 3 colours measured")
rows = [('uds', '2 < sqrt(s) < 3 GeV ', '~2.2'),
('udsc', '5 < sqrt(s) < 10 GeV', '~3.6'),
('udscb', 'above 10 GeV ', '~3.9')]
for fl, lab, meas in rows:
r1, r3 = R(fl, 1), R(fl, 3)
print(f" {lab} {r1:.3f} = {round(r1*9):2d}/9 {r3:.3f} = {round(r3*3):2d}/3 {meas}")
print()
print("Every measured value sits on the 3-colour row, three times the 1-colour one.")
print("ADONE in 1969 saw 1 to 3 where 2/3 was expected -- the first sight of colour,")
print("and unrecognised, because quarks were not yet believed to be real.")threshold region 1 colour 3 colours measured 2 < sqrt(s) < 3 GeV 0.667 = 6/9 2.000 = 6/3 ~2.2 5 < sqrt(s) < 10 GeV 1.111 = 10/9 3.333 = 10/3 ~3.6 above 10 GeV 1.222 = 11/9 3.667 = 11/3 ~3.9 Every measured value sits on the 3-colour row, three times the 1-colour one. ADONE in 1969 saw 1 to 3 where 2/3 was expected -- the first sight of colour, and unrecognised, because quarks were not yet believed to be real.
- u d s, 3 colours = 6/3
- u d s c, 3 colours = 10/3
- u d s c b, 3 colours = 11/3
- the same, with 1 colour
- measured (schematic, from Fig. 6.3)
⚙️ Engineer’s bridge — measuring a multiplicity by ratio
R is a degeneracy measurement, and it works the way degeneracy measurements always work: put the thing you cannot compute into both the numerator and the denominator, and read the integer that is left.
You do this whenever you count channels. The absolute throughput of a memory system depends on clock, protocol overhead, refresh, queueing — none of which you can calculate. But run the same benchmark against one bank and against banks, take the ratio, and falls out cleanly while every uncalculable factor cancels. Nobody needs a model of DRAM timing to count banks.
R is that measurement for the vacuum. The 1/s, the coupling, the phase space and the luminosity are common to both processes and divide out; the number of distinguishable ways to make a quark pair does not. The answer is 3, it is an integer, and it was measured before anyone had a theory that required it.
The reason this is worth naming: colour is not directly observable. Every hadron is a colour singlet, so no experiment ever sees a red quark. Yet the count of colours is a number you can read off a cross-section ratio to two significant figures — because a degeneracy shows up in a rate whether or not you can observe the states individually.
Where it breaks: reading a degeneracy off a rate requires everything except the degeneracy to cancel between numerator and denominator. R works because and differ only in charge and colour — same photon, same spin structure, same . Move to where that fails and the count fails with it: near a flavour threshold the phase space is not common to the two, so the steps are smeared rather than sharp, and on a resonance the ratio measures the resonance instead. Counting states by their effect on a rate is only as clean as the thing you divide by.
The residue above the steps, and Question 6.1
Look again at the plot: the data sit a few per cent above the 3-colour lines. That is not experimental error. The quark can radiate a gluon before hadronizing, and that opens extra ways to make a hadronic final state. QCD predicts the correction exactly at first order:
The R ratio with its first-order QCD correction. The 3 counts colours, the sum counts flavours and their charges, and the bracket is the only part of this expression that knows the strong interaction exists.
Every symbol, one at a time
Hover or tap a symbol above — it lights up in the equation and its meaning, units and type appear here.
which the book sets as Question 6.1: read at GeV off the figure.
🔢 Worked example — Question 6.1, and why the answer is nearly useless
At GeV, against the parton-model . So
The accepted value near 40 GeV is about 0.14. The estimate is in the right place and no better than that — and the reason is worth more than the answer.
The QCD correction is . So the whole signal is 4 % of the quantity being measured, and the error on is times the fractional error on : read 1 % wrong and moves by 0.03, a 25 % error. To pin to a per cent this way you would need to 0.1 %, which is far beyond what a total cross-section measurement can do.
That is why §6.5 does not use R. It uses jet rates, event shapes, and scaling violations — observables where enters at leading order rather than as a 4 % correction. The lesson generalises: an observable in which your parameter is a small correction is a bad instrument for that parameter, no matter how precisely you measure it.
Question 6.1, and its error budget
import numpy as np
R_parton = 3 * (2*(2/3)**2 + 3*(1/3)**2) # u c up-type, d s b down-type
print(f"parton model above the b threshold: R = {R_parton:.4f} = 11/3")
for Rm in (3.80, 3.85, 3.90, 3.95):
print(f" reading R = {Rm:.2f} off Fig. 6.3 -> alpha_s = {np.pi*(Rm/R_parton - 1):.3f}")
print(f"\nthe accepted alpha_s near 40 GeV is ~0.14, i.e. R = {R_parton*(1+0.14/np.pi):.3f}")
print(f"the correction alpha_s/pi is only {0.14/np.pi*100:.1f}% of R, so")
print(f"a 1% error on R becomes an error of {np.pi*0.01:.2f} on alpha_s -- about 25%.")
print("Sec. 6.5 measures alpha_s from jet rates instead, where it enters at leading order.")parton model above the b threshold: R = 3.6667 = 11/3 reading R = 3.80 off Fig. 6.3 -> alpha_s = 0.114 reading R = 3.85 off Fig. 6.3 -> alpha_s = 0.157 reading R = 3.90 off Fig. 6.3 -> alpha_s = 0.200 reading R = 3.95 off Fig. 6.3 -> alpha_s = 0.243 the accepted alpha_s near 40 GeV is ~0.14, i.e. R = 3.830 the correction alpha_s/pi is only 4.5% of R, so a 1% error on R becomes an error of 0.03 on alpha_s -- about 25%. Sec. 6.5 measures alpha_s from jet rates instead, where it enters at leading order.
Two more things the jets prove
The quarks have spin 1/2
If the quark is a spin-1/2 point particle, the pair must come out with the same angular distribution as a muon pair, Eq. (6.8):
The angular distribution of a pair of spin-½ point particles produced through one photon — identical for μ⁺μ⁻ and for qq̄ apart from the charge factor. Measuring the shape measures the quark's spin.
Every symbol, one at a time
Hover or tap a symbol above — it lights up in the equation and its meaning, units and type appear here.
The angle is measured to the jet axis, not to a quark, and only is available — the quark jet and the antiquark jet look alike, so which is which cannot be told. The CELLO measurement at 36.8–46.8 GeV follows the curve.
- spin 1/2: 1 + cos²θ
- spin 0 would give sin²θ = 1 − cos²θ
- CELLO (schematic, from Fig. 6.4)
The gluon exists, and it is a vector
About 10 % of events at –100 GeV show a third jet: a quark radiated a hard gluon gluon the eight massless vector mediators of QCD, each carrying a colour and an anticolour. There is no ninth: the singlet combination is colourless and would not couple. defined in §6.1 — open in glossary , which hadronized like everything else. That is how the gluon was discovered, at PETRA in 1979 — switch the toggle on the widget above.
Establishing its spin takes one more step, because a gluon jet and a quark jet look alike. The trick: order the three jets by energy, call the softest one the gluon, boost into the frame of the other two, and measure the angle between the gluon and the common axis of the pair. The distribution of differs sharply for a scalar and a vector, and TASSO’s data pick the vector.
- vector gluon, J^P = 1⁻
- scalar gluon, J^P = 0⁺ — excluded
- TASSO (schematic, from Fig. 6.7)
Erratum — a mislabelled line in Fig. 6.3
The two right-hand labels of Fig. 6.3 both read “(1 colour)”:
- “u d s c b (1 colour) = 11/3”, drawn at R ≈ 3.67
- “u d s c b (1 colour) = 11/9”, drawn at R ≈ 1.22
The upper one should read “(3 colours)”. 11/3 is precisely three times 11/9, and every other pair of labels on the same figure is tagged correctly — “u d s (3 colours) = 6/3” against “u d s (1 colour) = 6/9”, and “u d s c (3 colours) = 10/3” against “u d s c (1 colour) = 10/9”.
It matters because the mislabelled line is the one the data lie on. Read as printed, the figure says the measurements agree with the colourless prediction above the beauty threshold — the exact opposite of what the section spends four pages establishing.
What this section established
| result | how | what it would take to doubt it |
|---|---|---|
| There are three colours | R is three times Σz², at every energy and across three flavour thresholds | A factor of 3 hiding in the hadronization — but hadronization happens long after the pair count is fixed |
| Quarks are spin-1/2 point particles | the jet axis follows 1 + cos²θ, and the 1/s law holds between thresholds | Substructure would show as a form factor, i.e. a departure from 1/s. None is seen |
| Gluons exist | a third jet in ~10 % of events above 30 GeV | Nothing else makes a hard, isolated, colour-carrying spray |
| Gluons are vectors | the cos φ distribution of the softest jet excludes J^P = 0⁺ | The two hypotheses differ most where statistics are thinnest — this is the weakest of the four |
| …and gluons are coloured | the few per cent excess of R above 3Σz², the (1 + α_s/π) factor | It is only a 4 % effect, which is why §6.5 measures α_s another way |
Notice what is <em>not</em> here: no Lagrangian, no group theory, no SU(3). Everything above was measured before §6.3 writes down a theory — which is the order the book chose and the order this site keeps.
🔑 If you remember only three things
-
A jet is an observable the machine creates. Below a certain energy the hadrons never collimate, so the thing being counted did not exist at earlier accelerators.
-
The jets prove more than the counting does. Their angular distribution gives the quark’s spin and their occasional third member gives the gluon.
-
The interesting part is the residue. R sits slightly above the parton-model steps, and that small excess is where the strong coupling first shows itself.
Where this goes next
- §6.2 is the other half of the evidence: deep inelastic scattering, which shows the same quarks from the inside of a proton rather than from the outside of a collision.
- §6.3 finally writes the theory — SU(3), and the fact that gluons carry the charge they mediate.
- §6.5 measures properly and finds it falling with energy, the opposite of §5.8’s α. The same widget draws both.
- §5.7 is the denominator of R, and worth rereading now that the numerator has a name.
✅ Check yourself — jets, R, and the discovery of colour
0/5 answered · 0 correct
1.Hadronization is non-perturbative and not calculable. Why does that not ruin R as a measurement of the number of colours?
2.Slide the jet widget down towards √s = 3 GeV. Why do the jets disappear while R remains perfectly measurable?
3.R is measured as a ratio rather than as an absolute cross-section. What does that buy?
4.Question 6.1 extracts α_s ≈ 0.2 from R at 40 GeV, against an accepted 0.14. Why is the method so blunt?
5.Of the four results this section establishes, which rests on the weakest evidence, and why?