Accelerators and Detectors
Buried a hundred metres under the French–Swiss border runs the largest and coldest machine humanity
has ever built: the Large Hadron Collider, a ring
27\ \text{km} around. Inside two hair-thin beams of protons scream in
opposite directions at 99.9999991\% of the speed of light, each proton
carrying 6.5\ \text{TeV} of energy, and cross paths eleven thousand times
a second. Where they meet, a fraction of that colossal energy freezes — for an instant — into brand
new particles that have not existed in bulk since the universe was a trillionth of a second old.
Wrapped around each collision point sits a cathedral-sized instrument, ATLAS or CMS, that reconstructs
the debris of every crossing in nanoseconds.
This page rests on a single sentence, worth reading twice: we accelerate particles to enormous
energy, collide them head-on, and reconstruct what happened from a layered detector — with the number
of interesting events set by the product of cross-section and luminosity. Everything below
unpacks that one idea: why we need the energy, how a circular collider delivers it, how the onion-like
detector reads the wreckage, and the bookkeeping — R = L\,\sigma — that
turns a machine into a discovery rate.
Why on earth do we need so much energy?
There are two independent reasons, and a good particle physicist keeps both in mind at once.
Reason one — to create mass. Einstein's
E = mc^2 runs in both directions. Kinetic energy poured into a collision can
crystallise into the rest-mass of particles that were not there before. To make a Higgs boson of mass
m_H \approx 125\ \text{GeV}/c^2 you must supply at least
125\ \text{GeV} of energy in the collision; to make a top quark
(\sim 173\ \text{GeV}/c^2) you need more still. Heavier particles simply cost
more energy. There is no other way to summon them into existence.
Reason two — to see small things. Every particle has a quantum wavelength set by its
momentum through the de Broglie relation,
\lambda = \frac{h}{p}.
A microscope cannot resolve detail finer than the wavelength it probes with, and the same is true here:
to look inside a proton — to resolve structure at
10^{-18}\ \text{m} — you need a probe with a tiny wavelength, which by
\lambda = h/p means a huge momentum, hence huge energy. High energy is a
short ruler. Together these two reasons explain the whole arms race of ever-bigger machines: more
energy means both heavier particles and finer resolution.
The circular collider
How do you give a proton 6.5\ \text{TeV}? Not in one push. You bend the
beam into a ring and let it come round again and again, adding a little energy each lap. Two families
of hardware do the work:
-
Bending magnets — superconducting dipoles (at the LHC,
8.3\ \text{T}, cooled to 1.9\ \text{K}, colder
than outer space) whose magnetic field curves the charged protons around the
27\ \text{km} ring. Faster protons are stiffer and need a stronger field
to hold the same radius, which is why the top energy of a ring is set by how strong you can make its
magnets.
-
Radio-frequency (RF) cavities — resonant chambers holding an oscillating electric
field, timed so that each bunch of protons surfs a forward-pushing wave every time it passes. Lap
after lap the RF cavities do the actual accelerating, nudging the energy up until the beam is at full
strength.
The magnets steer; the cavities accelerate. Round and round for twenty minutes until the beams are
ready, then they are focused to a width thinner than a human hair and steered into each other at the
heart of each detector.
Because the dipole magnets bend it. A charged particle in a magnetic field feels a force
perpendicular to its motion, so it travels a circle of radius
r = p/(qB). Rearranged into the units particle physicists actually use,
this is the wonderfully handy rule
p\,[\text{GeV}/c] \approx 0.3\,B\,[\text{T}]\times r\,[\text{m}].
It says a stronger field or a bigger ring holds a higher-momentum beam. The very same formula, read
backwards, is how the detector measures momentum: watch how tightly a track curves in a
known field and you read off p. One equation, two jobs — steering the
beam and weighing the debris.
Centre-of-mass energy: why colliders beat fixed targets
You could imagine a cheaper machine: accelerate one beam and fire it at a block of metal sitting still.
It was done for decades, and for making new heavy particles it is a terrible bargain. The reason is the
centre-of-mass energy \sqrt{s} — the energy actually
available to create new stuff, as opposed to energy locked up in the debris flying forward to conserve
momentum.
For two beams of energy E colliding head-on (equal and
opposite momenta, so the total momentum is zero), all of the energy is available:
\sqrt{s} = 2E \qquad\text{(symmetric collider).}
For a beam of energy E hitting a fixed target of mass
m, most of the energy is wasted carrying the recoil forward, and the useful
energy only grows as the square root of the beam energy:
\sqrt{s} \approx \sqrt{2\,E\,m\,c^2} \qquad\text{(fixed target).}
The difference is dramatic. Double a collider's beam energy and you double
\sqrt{s}. Double a fixed-target beam and you gain only a factor
\sqrt{2}\approx 1.4. That linear-versus-square-root scaling is exactly why
the LHC — two 6.5\ \text{TeV} beams meeting head-on for
\sqrt{s}=13\ \text{TeV} — collides beams instead of hitting a wall.
The layered detector: an onion around the collision
No single instrument can measure everything, so a modern detector is built in
concentric layers, each tuned to a different job. A particle born at the centre
punches outward through them in turn, and by comparing what each layer saw, physicists reconstruct its
identity, direction, momentum and energy. Reveal the figure ring by ring.
Read the layers as a checklist of clues:
-
Inner tracker. Sitting in a strong magnetic field, it records the curved paths of
charged particles. Tight curve = low momentum, gentle curve = high momentum, by the very
p \approx 0.3\,B\,r rule from the magnets. Neutral particles leave no
track here at all.
-
Electromagnetic calorimeter. Electrons and photons dump all their energy here in a
shower, which the calorimeter measures by absorption. Anything that stops in this layer with
no track before it was a photon; with a track, an electron.
-
Hadronic calorimeter. Heavier, strongly-interacting particles — protons, pions,
neutrons — punch through the EM layer and deposit their energy here. A neutron shows up as energy
here with no track anywhere before it.
-
Muon chambers. Muons are penetrating: they ignore both calorimeters and sail out to
the very edge, where dedicated chambers catch them. A track that reaches the outside is a muon's
signature.
-
What you never see. Neutrinos interact so feebly they pass through the whole
apparatus undetected. You infer them from a momentum imbalance: add up the
transverse momenta of everything you did see, and if the books don't balance, the shortfall
— the missing transverse energy — carried off invisibly, is your neutrino.
Cross-section and luminosity: turning a machine into a rate
Colliding beams is not enough; you need to know how often an interesting collision happens.
Two numbers control that, and their product is the whole game.
The cross-section \sigma is the effective target
area a given process presents — how "likely" that reaction is, expressed as an area. It is measured in
barns, where 1\ \text{barn} = 10^{-24}\ \text{cm}^2 (the
whimsical name comes from being, for a nucleus, "as big as a barn door"). A common process has a large
cross-section; a rare one like Higgs production has a tiny cross-section, measured in picobarns
(10^{-12}\ \text{barn}). Crucially, \sigma is a
probability per unit incoming flux — it is not the physical size of any particle.
The luminosity L describes how hard the machine is pushing:
how many particles it crams through unit area per second. It has units of
\text{cm}^{-2}\,\text{s}^{-1}, and the LHC reaches around
L \approx 10^{34}\ \text{cm}^{-2}\text{s}^{-1}. Higher luminosity means
tighter, denser bunches crossing more often.
-
Instantaneous rate. The number of events per second of a process with
cross-section \sigma at luminosity L is
R = L\,\sigma.
-
Total harvest. Over a run, the total number of events is the cross-section times
the integrated luminosity
\mathcal{L} = \int L\,dt:
N = \sigma \int L\,dt = \sigma\,\mathcal{L}.
-
Units. Integrated luminosity is quoted in inverse barns — an LHC run delivers
of order 100\ \text{fb}^{-1} (inverse femtobarns). Multiply by a
cross-section in femtobarns and the barns cancel, leaving a pure count of events.
This is why discovery is a race on two fronts: build the energy high enough to make the
particle (that sets \sigma), then pile up enough luminosity to make
enough of them to see above the background.
Worked examples
Example 1 — momentum from a curving track. A charged particle leaves a track of
radius r = 5\ \text{m} in the tracker's
B = 2\ \text{T} field. Using the handy rule,
p \approx 0.3\,B\,r = 0.3 \times 2 \times 5 = 3\ \text{GeV}/c.
A gently curving (large-radius) track means a high momentum; a tightly wound one means a slow, soft
particle.
Example 2 — centre-of-mass energy. The LHC collides two proton beams of
E = 6.5\ \text{TeV} head-on. Since the collision is symmetric,
\sqrt{s} = 2E = 2 \times 6.5 = 13\ \text{TeV}.
A fixed target of protons (mc^2 \approx 0.94\ \text{GeV}) hit by that same
beam would give only \sqrt{s} \approx \sqrt{2 \times 6500 \times 0.94}\approx 111\
\text{GeV} — a hundred times less useful energy from the identical beam.
Example 3 — event rate. A process has cross-section
\sigma = 1\ \text{nb} = 10^{-33}\ \text{cm}^2 and the machine runs at
L = 10^{34}\ \text{cm}^{-2}\text{s}^{-1}. The events pour out at
R = L\,\sigma = 10^{34} \times 10^{-33} = 10\ \text{events per second.}
Example 4 — the total harvest. A rare process has
\sigma = 1\ \text{pb} = 10^{-36}\ \text{cm}^2. Over a run the experiment
accumulates an integrated luminosity \int L\,dt = 100\ \text{fb}^{-1} =
10^{41}\ \text{cm}^{-2}. The total number of events collected is
N = \sigma \int L\,dt = 10^{-36} \times 10^{41} = 10^{5}\ \text{events.}
A hundred thousand events of a process a billion times rarer than the common ones — pulled out of the
pile because the luminosity was high enough.
Misconception: "the collider smashes protons open to see the little bits inside." This
is the picture most people carry, and it is wrong. You are not cracking a walnut to inspect the kernel.
The reason the collisions must be so energetic is not to overcome some sticky glue holding the pieces
together — it is that the kinetic energy of the collision is converted into new mass
via E = mc^2. The Higgs boson, the top quark, the W and Z bosons — these are
not shards that were hiding inside the proton. They are created fresh out of pure energy at the
moment of collision, then decay again in a flash. The collider is less a hammer and more a forge: it
turns energy into matter that did not exist a moment before.
A second trap: "a bigger cross-section means a bigger particle," or "luminosity is how bright
the beam glows." Both confuse a name for a picture. A cross-section
\sigma has units of area, yes, but it is a measure of
probability — how likely a reaction is per unit flux — not the physical girth of anything. And
luminosity has nothing to do with light: it counts how many particles cross unit area per second, a
measure of collision intensity. Keep the meanings straight and the master formula
R = L\,\sigma reads plainly: rate equals intensity times probability.
Bunches cross 40 million times a second, and each crossing can spray out
thousands of particles — far too much to record. If ATLAS wrote every collision to disk it would fill
the world's storage in minutes. The answer is the trigger: a lightning-fast,
multi-layered decision system that looks at each event in microseconds and throws away
99.998\% of them, keeping only the rare crossings with the tell-tale
signatures of something interesting — a high-momentum muon, a big lump of missing energy, a pair of
photons. Of forty million crossings a second, roughly a thousand are saved. Discovery physics is,
quietly, as much a problem of deciding what to ignore as of what to measure.