In particle physics experiments, pileup occurs when two particles hit the same detector (say a calorimeter) at roughly the same time, resulting in what looks like a single event with higher energy than either one. Naturally, this can be a pain when analyzing data, and I've heard bits and pieces about methods for subtracting off the pileup events in the final analysis. Presumably this entails identifying events that seem to have a two-peak structure and either dropping the event from the analysis (eek, systematic error) or finding a way to separate them into the two events they are. All the analysis comes after the experiment, in which pileup can be reduced by segmentation of detectors (identify separate events because they are recorded on different parts of the detector) and faster data-taking (making it easier to identify that double-peak structure).
But I can't seem to find any accessible information on pileup (the only article Google seems to turn up is a fairly advanced one from the LHC, which I can't quite follow) or its reduction, and would really appreciate a brief summary of pileup subtraction techniques or a pointer to the right place to look.
Best Answer
That's a big topic with a lot of pieces in the answer.
The simplest one is that the interaction region has non-trivial extant along the beamline1 and many particles can be grouped into distinct vertex position along the beamline. Particles that come from the same place may belong together and particle that come from different places belong to different interactions.
The vertex position detection capability of the LHC experiments is very good, considerably better than $1\ \mathrm{mm}$ along the beam axis. I've seen event displays with 20+ interaction where each one is clearly separated by vertex position; it is truly astounding how well they can do with just this most basic separating metric.
When groups of tracks that can not be distinguished by vertex position don't seem to belong to a single interaction then you have to get to be smart.2 But as I have never been involve in these analyses I'm not going to guess.
1 The beam focussing for the LHC's interaction regions was designed with this in mind.
2 Lots of combinitorics and matching against well know interaction topologies.