Not a duel, but a stampede of bunches
From the accelerator rung you already know that a beam is not a smooth stream of protons but a train of tightly packed clumps called bunches. At the LHC each bunch holds something like a hundred billion protons, squeezed into a sliver thinner than a human hair, and two such trains race in opposite directions around the ring. The interesting physics only happens where the two trains are steered to pass through each other — the interaction point, the dead centre of a detector. Picture two swarms of bees flown head-on through one another rather than two single bullets aimed to meet.
Here is the first thing that surprises newcomers: when two bunches of a hundred billion protons each fly through one another, almost nothing happens. Protons are mostly empty space, the beam is mostly empty space, and a head-on hard collision between two of them is a rare accident. The overwhelming majority of protons in both bunches sail straight past everything and continue on around the ring untouched. Out of those two hundred billion protons, only a handful actually strike hard enough to do anything we care about. The art of collider physics begins with accepting that you are fishing for rare events in an ocean of near-misses.
What actually collides is not the proton
When one proton does strike another hard, the proton itself is not really the projectile. As you learned in the hadron rung, a proton is a seething bag of quarks and gluons — its three valence quarks plus a churning sea of gluons and quark-antiquark pairs, all sharing the proton's momentum. In a violent collision it is one of those internal constituents, a single quark or gluon called a parton, that does the actual smashing against a parton from the other proton. The two partons each carry only a fraction of their parent proton's energy, and that fraction is different every single time.
This has a consequence you must internalize early. Even though every proton in the beam carries the same fixed energy, the energy available in the actual hard collision varies wildly from crossing to crossing, because it depends on which constituents happened to meet and what fraction each was carrying. The probability of finding a parton with a given momentum fraction is described by the parton distribution function, measured in earlier experiments. So a proton collider is secretly a broadband machine: a single beam energy delivers a whole spectrum of effective collision energies, which is part of what makes it such a good discovery tool — you scan a range without retuning the magnets.
Because the colliding partons share their momentum unequally and at random, the products of the hard collision are usually thrown forward or backward along the beam, not symmetrically. This is why analysts care so much about momentum measured sideways to the beams — the part of the picture that is on average balanced. It also explains why the debris of a collision never looks like a neat fireworks burst centred on the crossing point; it is lopsided, smeared along the beam direction, and you have to read it knowing the boost was unknown from the start.
Pile-up: many collisions stacked into one snapshot
Now the complication that defines modern collider life. We said most protons miss — but with a hundred billion protons in each bunch, even a small per-collision rarity adds up. At every bunch crossing it is not one pair of protons that collides but typically dozens of pairs, all at the same instant, all within a region a fraction of a millimetre across. The detector cannot tell these apart in time: they are simultaneous. So a single recorded snapshot contains the debris of many independent proton-proton collisions superimposed. This stacking is called pile-up, and it is one of the central headaches of analysis.
Imagine trying to photograph one specific firework while forty other fireworks go off in the same frame at the same moment. Forty separate, boring collisions might bury the one rare and interesting collision you actually care about, dumping hundreds of extra particles into the detector and confusing every measurement. The chief defence is geometry: each proton-proton collision happens at a slightly different point along the few-centimetre length of the bunch, so the tracker can trace particles back to distinct vertices strung along the beam line and assign each track to the collision it came from. Reconstructing and separating these vertices is the first real job of the software the instant an event is saved.
Pile-up is not a flaw to be eliminated — it is the price of high luminosity, and luminosity is what finds rare physics. The deal physicists strike is to accept a crowded snapshot in exchange for vastly more chances at the rare process, then spend enormous effort cleaning up afterward. As beams are squeezed ever tighter to deliver more collisions, pile-up grows, and future upgrades will pack well over a hundred simultaneous collisions into every crossing. Much of detector design — fine spatial segmentation, precise timing down to tens of trillionths of a second — exists for the single purpose of pulling one signal collision out of that crowd.
What one recorded event actually contains
So what does the detector save when it decides to keep a crossing? Not a movie, not even a picture in the everyday sense. An event is the complete set of electronic readings from every channel of the general-purpose detector at that one bunch crossing — millions of numbers: which silicon strips lit up, how much energy landed in each calorimeter cell, which outer muon chambers fired. It is a frozen, three-dimensional pattern of hits and energy deposits, the raw fingerprints we met in the detector rung, captured for one instant and written to disk. Everything else — particles, momenta, masses — is something we infer from these numbers afterward.
From that raw pattern, reconstruction software builds a list of physics objects: charged-particle tracks bent in the magnetic field, clusters of energy in the calorimeters, jets of hadrons gathered up from the hadronization of quarks and gluons, identified electrons and muons, and the ledger of momentum balance sideways to the beam. When that ledger does not add up — when momentum is missing — the gap is recorded as missing transverse energy, the calling card of neutrinos or of anything else that escaped unseen. A reconstructed event is therefore a curated summary: from millions of raw hits down to perhaps a dozen objects with measured momenta, energies, charges, and identities.
From a saved event to a measurement
One number makes the scale of the problem vivid. The LHC's bunches cross about forty million times every second, but the detector can only afford to permanently store a few hundred to a couple of thousand events per second — the rest are simply too much data to keep. So the very first thing that happens to a crossing is a brutal, near-instant decision about whether it is even worth recording. That split-second filter is the trigger, and it throws away more than 99.99% of all crossings forever, keeping only those whose raw pattern hints at something energetic or unusual. The next guide is devoted to it; for now, just hold the staggering ratio in mind.
- Two bunches cross at the interaction point; tens of proton-proton collisions happen at once, all piled together — this is the pile-up you must later untangle.
- The detector's millions of channels record a raw pattern of hits and energy deposits — the whole snapshot is what we call one event.
- The trigger decides in a flash whether the pattern is interesting enough to keep, discarding the vast majority of crossings on the spot.
- Surviving events are reconstructed into tracks, jets, and identified particles, then poured into a vast pile where the statistics of millions of events — not any single one — eventually reveal a signal.
That is the whole arc of this rung in miniature. A bunch crossing produces a stampede of simultaneous collisions; one frozen snapshot of all of them is an event; a trigger keeps the rare promising ones; reconstruction turns raw hits into physics objects; and only the accumulated statistics of a mountain of events, sifted for signal against background and held to a punishing standard of significance, can become a measurement or a discovery. Everything that follows — triggering, signal versus background, the five-sigma rule, and the famous discoveries — is the patient art of digging that result out of the pile we have just learned how to build.