0
$\begingroup$

Can anyone explain what Pile ups are in a particle accelerator experiments? How can anyone know that an event is a pile-up interaction and how can anyone get rid of these events?

$\endgroup$

2 Answers 2

1
$\begingroup$

One example:

In Rutherford Backscattering Spectrometry (RBS), one sends an ion beam (protons, alphas, ...) onto a sample and looks for backscattered ions at a particular angle. The detector is a surface barrier detector, basically a thick $pn$ junction. When the ion enters the detector it creates a number of electron-hole pairs, causing the detector to emit a current pulse. This pulse goes into a pre-amp, then into an amplifier/filter that puts out a nearly square, flat-topped pulse. The height of the pulse is proportional to the current in the original pulse, which is proportional to the energy of the incident ion. A multi-channel analyzer (MCA) looks at the pulse height, and adds 1 count to the appropriate bin. In this way you build up your backscattered spectrum.

This is all well and good if the time between detected events is long compared with the temporal resolution of the system. Of course, one wants to record spectra as quickly as possible, so one wants lots of counts per unit time. Even with few counts per unit time, an analysis of the (random) backscattered events will show there is some probability of two events occurring in the detector in a given unit of time. Should a second ion hit the detector before the current pulse from the first ion has not ended, the current generated from the second ion comes out overlapping with the first. The pre-amp has no idea that the total current is from two events, and spits out a total signal that is the sum of the two events. The amp/filter happily takes that signal, and spits out a pulse to the MCA that is the sum of the two events, and one count gets recorded at a higher energy than one would expect. This is pileup - 'fake' counts that are the result of multiple real events nearly coinciding in time in the detector.

For RBS in particular it is useful to note that the expected count rate increases as the energy decreases, since the Rutherford cross section increases. Pileup then smears out your element edges, since there are many low energy events that can contribute to pileup at higher energies. So, one takes the hit and dials back you beam current to decrease the count rate to get nice clean edges.

Asides:

One can do a simple convolution of the spectra with itself to estimate what the envelope of the pileup 'signal' might look like, scale appropriately, and subtract off.

When I was doing RBS back in the day, fast digitizers (GHz + with fast acquisition/reset times) were not available. These days one might grab each signal from the detector, and either reject as pileup based on shape (which messes with your statistics mind you), or perhaps even deconvolve the signals to back out the separate events.

$\endgroup$
1
  • $\begingroup$ Thank you for your explanation Jon. $\endgroup$ Mar 29, 2023 at 1:49
0
$\begingroup$

Pile-ups are due to the fact that particle accelerators such as the LHC do not attempt to make single particles collide - that would be next to impossible. Instead they collide bunches containing many particles. In each collision there could be several dozen individual collision events, and the products of these collision events all hit the detectors are more or less the same time. LHC experimenters are usually looking for unusual collision events, and want to weed out the uninteresting events (which are the majority of events). The uninteresting events are called "pile-up" (or, often, "pileup" without the hyphen).

An overview of some of the ways of reducing the effects of pile-ups in the CMS experiment is given here.

$\endgroup$
1
  • $\begingroup$ Thank you for your explanation. $\endgroup$ Mar 29, 2023 at 1:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.