Particle Detectors II: Cloud and Bubble Chambers

These two instruments record everything and decide nothing. Every charged track in three dimensions, on film — and most of the millions of photographs show nothing worth keeping.

🎯 Why this matters

Once electronics could decide fast enough, the trigger became the most consequential thing an experiment designs, because what it discards is gone for good. The chambers never had that problem, and never had any way to keep up either.

Every detector in §1.13a returns a number — a pulse height, a time, a ring angle. The two on this page return a picture: the complete topology of an event, every charged track, in three dimensions, with its curvature and its ionisation density along the way. For forty years that was the only way to discover a particle you were not looking for, because you cannot trigger on a signature you have not yet imagined.

Both work the same trick. Put matter into a metastable state, one that is desperate to change phase but has no place to start, and let the ion trail left by a passing charge be the place it starts.

📐 Physics you need first — metastable phases and nucleation

Cool a vapour below its condensation point and it should turn into liquid; heat a liquid above its boiling point and it should turn into gas. Neither happens immediately, because a new phase has to begin as a tiny droplet or bubble, and a tiny one is expensive: its surface energy $\sigma \cdot 4\pi r^2$ grows faster than the bulk energy it saves, $\propto r^3$ . Below a critical radius the thing you just nucleated shrinks back. So the vapour sits supersaturated and the liquid sits superheated, both waiting.

An ion changes the arithmetic. Its electric field polarizes the surrounding molecules, which lowers the effective surface energy, which lowers the critical radius — so a droplet nucleates there and nowhere else. In a superheated liquid, the local energy deposit of the ionisation does the same job thermally.

That is the whole physics of both instruments: a metastable medium plus a seed that is a charged particle’s own wake. Everything else — cameras, magnets, timing — is engineering.

The cloud chamber: an instrument that arrived by accident

C. T. R. Wilson was a meteorologist. In 1895, taken with the glories and coronae he had watched from the observatory on Ben Nevis, he built a glass- windowed vessel of air and saturated water vapour that could be expanded suddenly, to study how clouds form. He very soon found that dust was not the only thing droplets condensed on, guessed the extra nuclei were ions, and confirmed it by irradiating the volume with the newly discovered X-rays.

By the end of 1911 the device was good enough to photograph the first tracks of α and β particles (Wilson 1912); by 1923 the cloud chamber was fully developed (Wilson 1933). A particle crossing it leaves a trail of ions, the expansion turns each into a droplet, a flash lamp fires and a camera records the track. Immersed in a magnetic field $\mathbf B$ the tracks curve, and the curvature radius $R$ gives the momentum through Eq. (1.110), $p\,[\text{GeV}] = 0.3\,B\,[\text{T}]\,R\,[\text{m}]$ .

🛠️ Cloud chamber, with the trigger that made it useful

Click a numbered marker for what that piece does.

Redrawn from the description on p. 51. Nothing here is exotic: a piston, a lamp, a camera, two counters and an AND gate.

⚙️ Engineer’s bridge — the cloud chamber is interrupt-driven

Read the picture above as a data-acquisition system and every part is familiar.

The ion trail is an analogue sample held in an RC-like decay: the particle writes it, and it persists for ~0.1 s. The Geiger counters are level-sensitive inputs. The coincidence circuit is a combinational AND. The expansion is a capture strobe issued after the event of interest, and it works because the hold time exceeds the trigger latency.

That is the ordinary architecture — detect, decide, then capture. It is worth naming it explicitly here, because the next instrument breaks it.

Where it breaks: interrupt-driven capture needs the sensor to hold state until the decision arrives. A cloud chamber can because its ions persist for milliseconds — a memory the designer did not have to build. Take that away and the architecture collapses, which is exactly what happens at a modern collider: nothing in the detector remembers anything for 25 ns, so the data must be clocked into a pipeline and held there while the trigger thinks. And this chamber’s “interrupt” costs about a second of dead time while the gas re-compresses and settles, so the duty cycle is set by the actuator rather than by the decision logic — a constraint no electronic trigger has.

The bubble chamber: the same idea, run backwards in time

D. Glaser invented the bubble chamber in 1952; L. Alvarez turned it into an industry. The working fluid is now a liquid, superheated by the expansion instead of a vapour supersaturated by it, and the track appears as a trail of gas bubbles.

The consequence is not a detail. A bubble chamber must be expanded before the particle arrives — the superheated state lasts only milliseconds and the bubbles must be photographed while they are still small. So:

The same physics, opposite causality

↕	Cloud chamber↕	Bubble chamber↕
Medium	gas, supersaturated vapour	liquid, superheated
Expand	after the particle passes
Can it be triggered?	yes — Rossi coincidence	no
Cosmic rays	its natural habitat	useless — arrival times are random
Accelerator	works
Density	gas — interactions are rare	liquid — target AND detector in one volume
Choice of medium	essentially fixed	a design parameter: hydrogen, deuterium, or a heavy liquid

The bubble chamber's great advantage and its fatal limitation are the same sentence. Because the liquid is dense, the chamber is its own target — the interaction vertex is inside the sensitive volume, so nothing is lost in a target wall. Because it must be expanded blind, it can only run where the events come on schedule.

Being your own target means the liquid is a physics choice, not a plumbing one:

liquid hydrogen — the target nucleus is a single free proton, which is the only case where the initial state is unambiguous. Everything in §4 that needs a clean $\pi p$ or $Kp$ initial state comes from hydrogen chambers.
liquid deuterium — a proton and a neutron, so by subtracting hydrogen data you get interactions on neutrons, which cannot be made into a target any other way.
a heavy liquid, typically freon — high $Z$ raises the interaction probability for the processes that need it: photon conversion into $e^+e^-$ pairs (so γ rays become visible), and the tiny cross-sections of neutrinos. Gargamelle’s 1973 discovery of weak neutral currents (§9.7) was freon.

All of them sit in a magnetic field, because otherwise you have topology without momenta.

⚙️ Engineer’s bridge — a pipeline that cannot trigger on its own data

The bubble chamber inverts the architecture above. Capture must be committed before the event exists, so the instrument has no trigger at all — it is a free-running sampler locked to an external clock. Every expansion produces a frame whether or not anything interesting happened.

Two things follow, and both eventually killed it:

The selection moves downstream. With no trigger, every frame must be developed, scanned and measured. The physics rate was never the bottleneck; the human rate was. Berkeley employed rooms of scanners, and the invention of semi-automatic measuring machines was itself a research programme.
The duty cycle is fixed by thermodynamics. Expand, wait for bubbles to grow to a photographable size, flash, re-compress, wait for the liquid to settle. A few frames per second at best.

Compare a modern experiment, which discards 99.999 % of its collisions in hardware within microseconds and never records them at all. The bubble chamber recorded everything and threw most of it away by hand, months later. That is a perfectly reasonable design when your beam delivers a few dozen useful events an hour — and completely hopeless at $10^9$ collisions per second.

Where it breaks: record-everything-and-filter-later is not a naive design, it is the right design in one regime — when events are rare, precious, and cheap to store relative to their scientific value. It fails not because filtering late is wrong but because the arithmetic inverts: at $10^9$ collisions per second the storage cost exceeds the value of the sample by many orders of magnitude, so the irreversible decision has to move upstream into hardware. The lesson is that the architecture is a function of the rate, not a fact about instruments.

Example 1.14 — reading one photograph

🔬 Experiment card — the Alvarez chambers, Berkeley, 1950s

Apparatus

A hydrogen bubble chamber in a uniform 1.5 T field, filled with liquid hydrogen (protons as targets) or deuterium (to reach neutrons), photographed by three cameras at different positions so every event is recorded stereoscopically. The picture reproduced in the book is from the small 10-inch chamber — 25 cm across — exposed to a π⁻ beam from the Bevatron.

What is measured

Nothing but coordinates. The positions of the three vertices and of a series of points along each track, in all three photographs, plus the bubble density along each track by eye. The little spirals curling off the beam tracks are δ-rays — atomic electrons knocked out with enough energy to make their own track, spiralling tighter as they lose it.

What comes out

The complete event — which particles were produced, their masses, their momenta and, from the flight distances, their lifetimes. Two neutral particles that the chamber never saw at all are reconstructed from their charged daughters.

🪜 Reading a bubble-chamber photograph (Example 1.14)

Step 1 of 7 — What the film shows

Why you may do this: Seven π⁻ tracks from the Bevatron enter from the left. They are nearly parallel and nearly straight, because the beam is monochromatic and the field bends a 1 GeV track only slightly over this distance.

The small curls coming off some tracks are δ-rays: atomic electrons knocked out with enough energy to make their own short, tightly curling track. They are a nuisance, and also a check that the field direction is what you think.

Redrawn from the book's Fig. 1.21 — a 10-inch hydrogen chamber exposed to a π⁻ beam at the Bevatron. Nothing in this analysis needed electronics: three cameras, a magnetic field and a ruler.

The two V⁰ s and the primary vertex give the reaction

\pi^- + p \to K^0 + \Lambda^0 \tag{1.117}

followed by the two decays

K^0 \to \pi^+ + \pi^-, \tag{1.118}

\Lambda^0 \to \pi^- + p . \tag{1.119}

How much can you not see and still solve the event?

This is the part of Example 1.14 worth doing carefully, because it is a counting argument and it generalises to every experiment in the rest of the book.

At any vertex, energy–momentum conservation is four equations. An unseen neutral particle whose mass you have hypothesised costs three unknowns — its momentum vector; the energy then follows from $E^2 = p^2 + m^2$ . So:

Counting at the primary vertex (beam momentum measured, target proton at rest)

Final state↕	Unknowns↕	Equations↕	Verdict↕
K⁰ Λ⁰ — both reconstructed from their Vs	0	4	over-determined by 4: the hypothesis is TESTABLE
K⁰ Λ⁰ π⁰ — the π⁰ never seen	3	4
two unseen neutrals	6	4	under-determined — the event is unreconstructable

The threshold sits between one unseen neutral and two. That is why final states with several π⁰s were the hard ones for a generation of experiments, and why missing-mass techniques (§4.5) had to be invented — they are what you do when you are on the wrong side of this table.

💡 What this really says — it is a least-squares fit, and you already know how to read it

“The problem is overdetermined” is the physicist’s way of saying you have more equations than unknowns, so you cannot merely solve — you must fit, and the residual is information. Assume every track is a pion, impose conservation at all three vertices, and see how badly the equations disagree relative to the measurement errors. Then assume one track is a proton and do it again. The hypothesis with the acceptable $\chi^2$ is the event.

The data choose, not the physicist. That is the entire value of being over-determined, and it is the same reason you design a sensor system with redundant measurements rather than exactly enough.

The numbers behind the picture

Two quantitative facts are doing silent work in that photograph, and the book states neither.

Why the Vs are inside the chamber at all. A $\Lambda$ travels a mean $\beta\gamma c\tau$ before decaying, with $c\tau = 7.89$ cm; for the K⁰ — decaying here as the short-lived $K^0_S$ , §8.3 — it is 2.68 cm. At Bevatron momenta these come out at a few centimetres — comfortably inside a 25 cm chamber, and comfortably outside the resolution. The chamber size is not arbitrary: it is matched to $c\tau$ .

Why the proton is obvious. The book says the track is identified as a proton “because it is positive and has a large bubble density, corresponding to a large $dE/dx$ ”. How large? At the same momentum, Bethe–Bloch (§1.11) gives a proton four times the ionisation of a pion at 400 MeV/c, because $\beta\gamma$ is 0.43 for the proton and 2.9 for the pion — the pion is near its minimum and the proton is far up the $1/\beta^2$ rise. A factor of four in bubble count per centimetre is not subtle; you can see it.

🔢 Worked example — is a 10-inch chamber big enough?

Take a π⁻ beam at $p = 1.2$ GeV/c on a proton at rest.

Step 1 — is the reaction even open? The threshold is where $\sqrt s = m_K + m_\Lambda$ :

E_\pi^{\text{thr}} = \frac{(m_K+m_\Lambda)^2 - m_\pi^2 - m_p^2}{2 m_p} = 907.5\ \text{MeV},

so $p_\pi^{\text{thr}} = 896.7$ MeV/c. At 1.2 GeV/c we are above it, but not by much — this is a threshold-region experiment.

Step 2 — the products are slow. $\sqrt s = 1779.6$ MeV, giving the two neutrals $p^* = 352$ MeV/c back to back in the CM, which the boost ( $\beta_{\text{cm}} = 0.559$ ) spreads over 0–836 MeV/c for the K⁰ and 364–1214 MeV/c for the Λ⁰ in the lab.

Step 3 — how far do they get? $\ell = \beta\gamma\, c\tau = (p/m)\,c\tau$ :

\ell_{K} \le \frac{836}{498}\times 2.68\ \text{cm} = 4.5\ \text{cm}, \qquad \ell_{\Lambda} \le \frac{1214}{1116}\times 7.89\ \text{cm} = 8.6\ \text{cm}.

Both are a few centimetres in a 25 cm chamber, so both Vs open inside the sensitive volume with room for the daughters to be measured. A chamber ten times smaller would have shown the primary vertex and nothing else; ten times larger and the daughter tracks would leave before you had used the volume. The instrument fits the physics because it was built to.

Reproduce it

import numpy as np

mpi, mp, mK, mL, me = 139.57039, 938.27209, 497.611, 1115.683, 0.51099895
ctK, ctL = 2.6844, 7.89                      # cm, K0_S and Lambda

def pstar(M, m1, m2):                        # CM momentum of a 2-body state
    return np.sqrt((M**2-(m1+m2)**2)*(M**2-(m1-m2)**2))/(2*M)

E_thr = ((mK+mL)**2 - mpi**2 - mp**2)/(2*mp)
print(f"threshold: E_pi = {E_thr:.1f} MeV, p_pi = {np.sqrt(E_thr**2-mpi**2):.1f} MeV/c")

ppi = 1200.0
Epi = np.hypot(ppi, mpi)
rs  = np.sqrt(mpi**2 + mp**2 + 2*Epi*mp)     # sqrt(s)
ps  = pstar(rs, mK, mL)
bcm = ppi/(Epi+mp); gcm = 1/np.sqrt(1-bcm**2)
print(f"p_pi = {ppi:.0f} MeV/c: sqrt(s) = {rs:.1f} MeV, beta_cm = {bcm:.3f}, "
      f"p* = {ps:.1f} MeV/c")
for name, m, ct in (("K0", mK, ctK), ("L0", mL, ctL)):
    pmax = gcm*(ps + bcm*np.hypot(ps, m))    # emitted forward in the CM
    print(f"   {name}: lab p up to {pmax:4.0f} MeV/c -> "
          f"mean flight up to {pmax/m*ct:.1f} cm")
print(f"   chamber is {10*2.54:.1f} cm across, so both Vs open inside it")

print(f"K0 -> pi+ pi- : p* = {pstar(mK, mpi, mpi):.2f} MeV/c")
print(f"L0 -> p pi-   : p* = {pstar(mL, mp, mpi):.2f} MeV/c")

K, ZA, I, rho = 0.307075, 0.99212, 21.8e-6, 0.0708   # liquid H2
def dedx(p, M):                              # Bethe, MeV/cm
    E = np.hypot(p, M); b = p/E; g = E/M; bg = b*g
    Tmax = 2*me*bg**2/(1 + 2*g*me/M + (me/M)**2)
    return K*ZA/b**2*(0.5*np.log(2*me*bg**2*Tmax/I**2) - b**2)*rho
p = 400.0
print(f"dE/dx in liquid H2 at p = {p:.0f} MeV/c: "
      f"proton/pion = {dedx(p, mp)/dedx(p, mpi):.2f}  "
      f"(bg = {p/mp:.2f} vs {p/mpi:.2f})")

print(f"Alvarez chamber, reading the primes as INCHES: "
      f"{72*.0254:.2f} m x {20*.0254:.2f} m x {15*.0254:.2f} m")
print(f"                 reading them as FEET:        {72*.3048:.1f} m long")

prints

threshold: E_pi = 907.5 MeV, p_pi = 896.7 MeV/c
p_pi = 1200 MeV/c: sqrt(s) = 1779.6 MeV, beta_cm = 0.559, p* = 352.2 MeV/c
 K0: lab p up to  836 MeV/c -> mean flight up to 4.5 cm
 L0: lab p up to 1214 MeV/c -> mean flight up to 8.6 cm
 chamber is 25.4 cm across, so both Vs open inside it
K0 -> pi+ pi- : p* = 205.97 MeV/c
L0 -> p pi-   : p* = 100.58 MeV/c
dE/dx in liquid H2 at p = 400 MeV/c: proton/pion = 4.31  (bg = 0.43 vs 2.87)
Alvarez chamber, reading the primes as INCHES: 1.83 m x 0.51 m x 0.38 m
               reading them as FEET:        21.9 m long

Erratum — the book’s 72-foot bubble chamber

p. 52 describes Alvarez’s principal device as “a large hydrogen bubble chamber 72′ long, 20′ wide and 15′ deep”. The prime mark means feet; that chamber would be 21.9 m long — a 22-metre vessel of superheated liquid hydrogen, which nobody has ever built and nobody would want to.

It is the famous 72-inch chamber: 1.83 m × 0.51 m × 0.38 m, which is a large but entirely sane object. The double-prime is used correctly four lines later for the 10″ chamber of Fig. 1.21, so this is a typography slip rather than a misconception — but it is a factor of twelve, and the phrase “large hydrogen bubble chamber” gives a reader no reason to doubt it.

The habit worth taking from this: primes and double-primes are one glyph apart and mean units a factor of 12 apart. Convert to SI the moment you read them.

What you can actually measure on a photograph

The coordinate resolution is about one tenth of a bubble radius, and the bubble radius is set by the liquid and by how soon after the expansion you fire the flash:

Bubble radius, and what it buys

Chamber↕	Bubble radius↕	Coordinate resolution↕	What that is good for↕
heavy liquid (freon)	≈ 1 mm	≈ 100 μm	γ conversions and neutrino interactions — you accepted the resolution to get the cross-section
liquid hydrogen	≈ 0.1 mm	≈ 10 μm	the clean initial state; the workhorse of hadron spectroscopy
LEBC (rapid cycling H₂)	≈ 10 μm	≈ 1 μm

Compare the emulsion of §1.13a, at sub-micrometre resolution but with no time information and no trigger. Every detector in this chapter is a trade between where, when and how often — none of them wins all three.

Bubble chambers repaid the effort: the discovery of most of the unstable hadrons of Chapters 3 and 4, the evidence that assembled the quark model, the first neutrino physics, weak neutral currents, and the early study of nucleon structure. Then the rate went up and the film ran out.

🔑 If you remember only three things

An instrument that cannot be triggered cannot be scaled. The rate is set by how fast you can cycle and photograph, not by how fast the events arrive.
What you cannot see is recovered by conservation, not by looking. The neutrals leave no track, and their momenta come out of the imbalance in everything that did.
The record outlives the analysis. A photograph can be re-measured decades later, which is why these chambers yielded results their builders never set out to find.

Where this goes next

§1.13c is Example 1.15 — how you actually get $R$ , and hence $p$ , from three measured points on a track — and then the electronic descendants of the chamber: Geiger, proportional, MWPC, drift, TPC.
§2.2 is what the V⁰s turned out to mean.
§8.3 is why the K⁰ has two lifetimes, one of which is the 2.68 cm used above.

✅ Check yourself — the imaging detectors

0/6 answered · 0 correct

1.A cloud chamber can be triggered and a bubble chamber cannot. What is the physical reason?
2.Bubble chambers were useless for cosmic rays but ideal at an accelerator. Which statement captures why?
3.In the Alvarez event, the primary vertex produced two neutrals and both were reconstructed. Suppose the reaction had instead been π⁻p → K⁰ Λ⁰ π⁰, with the π⁰ never seen. What happens?
4.One track from the lower V is identified as a proton rather than a pion, by eye, from bubble density alone. Roughly how big is the effect at p = 400 MeV/c in liquid hydrogen?
5.Why is a 10-inch (25 cm) chamber the right size for seeing both decays of π⁻p → K⁰Λ⁰ at Bevatron energies?
6.Bubble chambers were abandoned even though they still give the most complete picture of an event. What ended them?