Scene 01 — the starting gun

From a proton pulse
to a published paper.

The ISIS Neutron and Muon Source is a giant accelerator in Oxfordshire: it fires neutrons and muons at materials to reveal where the atoms sit and what they’re doing. This is the story of how that happens — told for the people who keep it running but never studied physics. No equations you can’t skip. Lots of moving pictures.

50×tight bursts (‘pulses’) of protons fired every second

Scene 02 — the machine

First, make protons. Fast ones.

It starts as hydrogen gas, turned into H⁻ ions — protons carrying two spare electrons, so the whole thing is negatively charged. They’re accelerated down a linac (a straight-line accelerator) and into a synchrotron ring 163 m in circumference, where a very thin carbon foil strips both electrons off each ion — leaving bare protons that gain energy each lap until they’re kicked out toward the metal targets where the neutrons and muons are made (next scene).

800 MeVmega-electronvolts of proton energy — 84% of the speed of light, one lap every 0.65 µs

from linac (70 MeV) targets → 163 m around 70 MeV ≈400 MeV 800 MeV
schematic loop — the real bunch makes ~10,000 laps in ~10 ms before extraction

Scene 03 — two probes for the price of one

The beam makes muons, then neutrons.

On the way to Target Station 1 the protons first pass through a 1 cm slice of graphite, where collisions make short-lived particles called pions, which fall apart (‘decay’) within metres into muons — heavy, unstable cousins of the electron. Graphite is thin and light, so most protons punch straight through; the surviving beam then slams into a thick tungsten target, chipping neutrons off its heavy, neutron-packed nuclei. That process is called spallation, and the neutrons spray out in every direction — which is why the moderators in the next scene sit above and below the target, not just downstream of it. (From here we follow the neutrons — the muons’ story continues on the probes page.)

~10–15neutrons knocked out per incoming proton

graphite, 1 cm tungsten muons (μ) → muon instruments neutrons (n)
on loop: the pulse sheds muons at the graphite, then makes neutrons in the tungsten

Scene 04 — slow down

Fresh neutrons are uselessly fast.

Straight off the target they move at tens of thousands of km/s — far too fast to probe atoms. So they pass through a moderator: a small tank of water, liquid methane or liquid hydrogen — all of them packed with light nuclei. Every bounce is an elastic collision, billiard balls rather than absorption, and a light nucleus takes away far more of the neutron’s speed than a heavy one would. Quantum mechanics’ strange gift: every particle of matter also behaves as a wave, and for a neutron, the slower it moves the longer its wavelength. About 14 collisions later they emerge thousands of times slower, with wavelengths that match the spacing between atoms — and energies that match how those atoms are moving.

≈ 1/10,000thof their speed left after moderation — and that’s the point

moderator (H₂O / CH₄ / H₂) cold, slow, useful
hot in, cold out — ‘cold’ just means slow — repeated 40 times a second at Target Station 1

Scene 05 — the race that measures itself

Every pulse is a race down the beamline.

The neutrons in a pulse start almost together: the protons arrive as two bunches 0.32 µs apart, and the moderator holds each neutron for a slightly different time on the way out. Remember: slower = longer wave. So the short-wavelength ones are fast; the long-wavelength ones lag. By the time they reach the sample (the sliver of material a researcher has parked in the beam to study), arrival time tells you the wavelength — the facility’s clock is its most important instrument, and that small smear in the start time is exactly what limits how sharply it can read.

moderator sample · 10 m t = 5.1 ms t = 7.6 ms t = 10.1 ms short wavelength (λ) — arrives first medium λ long λ — arrives last

λ ≈ 3956 · t / L — wavelength in ångströms (Å: a ten-billionth of a metre, about the width of an atom), from arrival time t (s) over flight path L (m). The 3956 isn’t magic — a neutron’s speed × wavelength always equals 3956 m·Å/s. Try it yourself →

on loop: all three leave within microseconds of each other — the clock is what tells them apart

Scene 06 — the data is born

Detectors don’t see pictures. They see events.

Neutrons scatter off the sample and hit banks of detectors. Every hit is just (detector ID, time of flight) — no images, no spectra, only counts and clocks. On most ISIS instruments the electronics don’t keep each hit as its own row: they add 1 to the bucket for that detector and that slice of time, so what lands in the file is already a tally. (Instruments that do save every hit separately are running in ‘event mode’ — both shapes, and why it matters, on the data page.)

10⁶–10⁹neutrons counted in one ‘run’ (a single measurement, minutes to hours)

sample example hits: (det 02, 2.0 ms) (det 07, 3.7 ms) (det 04, 6.5 ms) (det 09, 10.0 ms) counts vs time of flight, building up
on loop: each pulse builds the histogram hit by hit — the list shows a sample of hits

Interlude — scenes 02–06 in one picture

One pulse, end to end.

Forty times a second at Target Station 1: protons accelerate, shed muons at the graphite, make neutrons in the tungsten, slow down in the moderator, then race down the beamline — and every detector hit becomes one more count.

linac synchrotron graphite tungsten moderator flight path · 10 m sample detectors → muon instruments one neutron instrument (of 30+)
on loop — the whole chain, schematic and not to scale

Scene 07 — from log file to physics

Reduction turns events into answers.

Software — at ISIS, mainly Mantid — takes those tallies and converts time of flight into physical units, using the fact that it knows where every detector sits. Then it strips out the machine’s own fingerprint with the help of extra measurements — a standard sample, an empty sample container, a monitor count — some divided out, some subtracted, until what’s left is the sample’s own signal and a model can be fitted to it. The one thing reduction can’t remove is that start-time smear from Scene 05, so it gets built into the fit instead. Out the other end: numbers with error bars that go straight into a paper.

Run the pipeline yourself ▸

Load ConvertUnits Rebin Normalize Fit raw instrument units — raw events physical units — reduced + fitted
on loop — the same data, step by step through the chain

Scene 08 — the punchline

All of this, for ~1,200 experiments a year.

Thousands of visiting researchers — plus the ISIS staff who run the accelerator, the instruments, the sample environments and the software that turns pulses into papers.

3,000researchers supported each year

30+neutron & muon instruments

1984delivering beam since