Accelerators, Colliders, and Luminosity

An accelerator delivers the energy that a reaction needs and the intensity that a measurement needs. The energy sets which final states are kinematically open; the intensity sets how many events accumulate before statistics limit the answer. The first quantity is fixed by relativistic kinematics and the choice between a fixed target and colliding beams; the second is fixed by the luminosity. This lesson develops both, then applies them to the machines that define the energy frontier.

Natural units and ${\eta }^{\mu \nu }=\mathrm{diag}(+,-,-,-)$ throughout, with the invariant mass of a system written $\sqrt{s}$ where $s=({\sum }_i{p}_i{)}^2$.

Available energy: fixed target versus collider

The quantity that decides which particles a collision can create is the invariant mass of the initial state, $\sqrt{s}$, because $s$ is frame-independent and equals the total energy in the center-of-momentum frame. A final state of total rest mass $M$ is accessible only when $\sqrt{s}\ge M$.

For a beam of energy $E_{\text{beam}}$ and mass $m$ striking a stationary target of mass $M_t$, the two four-momenta are $p_1=(E_{\text{beam}},\vec{p})$ and $p_2=(M_t,\vec{0})$, so

At high energy $E_{\text{beam}}\gg m,M_t$ the first two terms are negligible and

The center-of-mass energy grows only as the square root of the beam energy. The reason is visible in the four-momentum: the produced system inherits the beam's momentum $\vec{p}$ and therefore carries a large kinetic energy of overall motion that cannot be spent on making new particles. Most of the beam energy ends up as the recoil of the center of mass.

A collider removes that waste. Two beams of energy $E$ and opposite momenta meet head-on, so the total momentum is zero and the lab frame is the center-of-mass frame. For equal beams of the same mass,

Every unit of beam energy appears in the collision. The contrast is dramatic at high energy: to reach a given $\sqrt{s}$, a fixed-target machine needs a beam energy that scales as $s$, whereas a collider needs a beam energy that scales as $\sqrt{s}$.

The penalty is the whole reason the energy frontier is a collider frontier. Fixed targets survive where their advantages matter: a stationary target can be dense and thick, giving high interaction rates, and it permits a secondary beam of unstable particles (pions, kaons, neutrinos, muons) to be produced and steered. Colliders own the high-$\sqrt{s}$ regime; fixed-target beams own rate and beam-species flexibility.

Linear and circular machines

A charged particle gains energy only from an electric field along its motion. All accelerators therefore reduce to radio-frequency (RF) cavities that present an oscillating longitudinal field, timed so that a particle crossing the gap always sees an accelerating phase. The two geometries differ in how often a bunch reuses the same cavities.

• Linear accelerator (linac). Cavities are laid in a straight line; each bunch passes once. The final energy is the sum of the gap voltages, so reaching high energy means building a long machine. There is no synchrotron loss and no bending, and the beam is used once per pass, which suits it to a collider only if the spent beams are dumped rather than stored.
• Circular accelerator (synchrotron). Bending magnets return the bunch to the same RF cavities on every turn, so a modest voltage compounds over millions of revolutions. The magnetic field is ramped in step with the rising momentum to hold the orbit radius fixed — the feature that names the machine. A stored beam can circulate for hours, and two counter-rotating beams in the same ring collide at fixed points every turn.

The bend that makes a synchrotron efficient also forces the beam to radiate.

Synchrotron radiation and the electron limit

A relativistic charge on a circular orbit is transversely accelerated and radiates. For a particle of energy $E$, mass $m$, and orbit radius $R$, the energy lost to synchrotron radiation per revolution is

The ${\gamma }^4=(E/m{)}^4$ dependence is decisive. Because the loss scales as the inverse fourth power of the mass, an electron ($m_e=0.511\text{MeV}$) radiates $(m_p/m_e{)}^4\approx 1.1\times 10^{13}$ times more than a proton of the same energy on the same orbit. For electrons a convenient practical form is

At the LEP collider ($R\approx 3.1\text{km}$, $E=100\text{GeV}$ per beam) this is about $2.9\text{GeV}$ radiated per turn — a substantial fraction of the beam energy that the RF system had to replace every revolution. LEP was the practical ceiling for a circular electron collider: the radiated power scales as $E^4/R$, so pushing the energy higher would have demanded an impractically large ring or an impractically large RF plant. Higher-energy electron colliders are therefore linear by design, where the beam never bends and never radiates.

Protons escape this limit by their mass. At the LHC ($E=7\text{TeV}$, $R\approx 4.3\text{km}$) the synchrotron loss is only a few keV per turn. The proton machine is limited instead by the bending field: the momentum a ring can hold is

so a $7\text{TeV}$ beam in the $27\text{km}$ LHC tunnel requires superconducting dipoles of about $8.3\text{T}$. The electron frontier is set by radiation; the proton frontier is set by magnet technology and tunnel size.

Luminosity and event rate

Energy decides what a collision can make; luminosity decides how often it does. For a process of cross section $\sigma$, the event rate is

which defines the instantaneous luminosity $\mathcal{L}$, a property of the machine alone with units of inverse area per unit time (conventionally ${\text{cm}}^{-2}{\text{s}}^{-1}$). The luminosity is set by how much beam is packed into how small a spot, how often. For two Gaussian bunches of $N_1$ and $N_2$ particles colliding head-on at frequency $f$, transverse sizes ${\sigma }_x$ and ${\sigma }_y$,

High luminosity wants many particles per bunch, a high crossing frequency, and tight focusing at the interaction point (small ${\sigma }_x{\sigma }_y$). The final focusing is the job of strong quadrupole magnets that squeeze the beams just before they meet. The LHC reaches $\mathcal{L}\sim 2\times 10^{34}{\text{cm}}^{-2}{\text{s}}^{-1}$, packing $\sim 10^{11}$ protons into each of thousands of bunches and focusing them to a transverse size of order tens of micrometers.

The quantity that governs a data set is the time-integrated luminosity,

with the same units as an inverse cross section. It is quoted in inverse barns: $1\text{b}=10^{-24}{\text{cm}}^2$, so $1{\text{fb}}^{-1}=10^{39}{\text{cm}}^{-2}$. The expected number of events of a process is

A cross section of $1\text{pb}$ collected against $L_{\text{int}}=100{\text{fb}}^{-1}$ yields $N=(10^{-12}\text{b})(10^5{\text{b}}^{-1})\times 10^{24}\dots$ — kept in consistent units, $\sigma =1\text{pb}=10^{-3}\text{fb}$ times $100{\text{fb}}^{-1}$ gives $N=10^5$ events. Integrated luminosity is the currency in which discovery reach is priced: a rare process with a small cross section becomes observable only once enough inverse femtobarns have been accumulated.

Machine choices: the electron and proton trade-off

The two collider families answer different questions, and the difference traces back to the two limits above.

• Electron-positron colliders ($e^{+}{e}^{-}$). The colliding particles are pointlike, so the full $\sqrt{s}$ is available to a single elementary reaction and the initial state is known exactly. Events are clean: little hadronic debris, calculable QED backgrounds, and a fixed, tunable collision energy that can be parked on a resonance. LEP scanning the $Z$ pole and measuring its line shape is the archetype. The cost is synchrotron radiation, which caps the energy of a circular machine and forces a linear design above the LEP scale.
• Proton-proton (or proton-antiproton) colliders ($pp$, $p\bar{p}$). Protons are composite, so the elementary collision is between partons that each carry an unknown fraction of the proton momentum; the parton-level $\sqrt{\hat{s}}$ is a distribution below the machine $\sqrt{s}$, and the events are messy with beam remnants and QCD radiation. In exchange, protons do not radiate, so a circular machine reaches far higher energy, and the range of parton energies means one machine probes many scales at once. The LHC is the archetype: a discovery machine that trades a clean initial state for reach.

A common division of labor follows. A hadron collider is a discovery machine — broad reach, high energy, the tool that first sees a new state. An electron machine is a precision machine — clean events at a known energy, the tool that measures the new state's properties. The discovery of the $W$ and $Z$ at the $Sp\bar{p}S$ hadron collider followed by their precision study at LEP is the template, and the same pairing motivates proposals for an $e^{+}{e}^{-}$ Higgs factory to follow the LHC's discovery of the Higgs.

The LHC as a case study

The Large Hadron Collider gathers every element of this lesson into one machine. It is a $pp$ synchrotron of circumference $26.7\text{km}$, chosen because protons do not radiate and a large radius is affordable in reused tunnel. Its energy is set by the bending field: superconducting dipoles of about $8.3\text{T}$ hold each beam at $6.8\text{TeV}$, giving $\sqrt{s}=13.6\text{TeV}$ in Run 3 (design $7\text{TeV}$ per beam, $\sqrt{s}=14\text{TeV}$).¹ Because the colliding objects are partons, the actual parton-level energies span a wide range below $\sqrt{s}$, so a single machine probes electroweak-scale physics and multi-TeV physics simultaneously.

Its luminosity is set by tight focusing of dense bunches: several thousand bunches of $\sim 10^{11}$ protons each, crossing every $25\text{ns}$, squeezed to a transverse size of order $10\mu \text{m}$ at the interaction points, reaching $\mathcal{L}\sim 2\times 10^{34}{\text{cm}}^{-2}{\text{s}}^{-1}$ and delivering more than $100{\text{fb}}^{-1}$ per year. That integrated luminosity is what made the Higgs observable: with a production cross section of tens of picobarns and a $\gamma \gamma$ branching ratio near $2\times 10^{-3}$, only a machine delivering tens of inverse femtobarns produces enough events to lift the signal over the background, the subject of the discovery lesson.

Summary

The invariant mass $\sqrt{s}$ sets the accessible final states, and the choice of machine sets how much beam energy $\sqrt{s}$ costs. A fixed target wastes energy in the recoil of the center of mass, so $\sqrt{s}\simeq \sqrt{2M_t{E}_{\text{beam}}}$ grows as the square root of the beam energy; a collider puts the lab frame at rest, $\sqrt{s}=2E$, and spends every unit of beam energy. Linacs accelerate once and never bend; synchrotrons reuse cavities every turn but pay synchrotron radiation $\Delta E\propto (E/m{)}^4/R$, which caps circular electron machines at the LEP scale and leaves the proton frontier limited by bending field, $p=0.3BR$. Luminosity converts a cross section into a rate through $R=\mathcal{L}\sigma$, and its time integral fixes the event count $N=\sigma L_{\text{int}}$. Electron machines buy clean precision at a known energy; proton machines buy reach with a composite, messy initial state. The LHC realizes the proton strategy — a large low-radiation ring, strong dipoles, and enough integrated luminosity to make rare processes visible. What those collisions register in is the subject of the next lesson.