Implementation of Machine Learning to the Muon g-2 Tracking Algorithms

Student: Antony Hibbert
Supervisor: Joe Price

The Muon E989 g-2 experiment seeks to confirm the results previously yielded by the E821 experiment. In this experiment, the anomalous magnetic moment of the muon was determined with a precision of 0.54ppm. A discrepancy between experimental results and the standard model prediction of more than 3s was found, which could be an indication of beyond Standard Model physics. As it stands, however, this discrepancy could be attributed to a statistical fluctuation, having been observed at only one experiment. If this phenomenon is truly an indicator of new physics, the difference should exceed the 5s discovery threshold. In order to confirm or deny the previously determined experimental results, the E989 experiment will improve the precision to 140ppb by observing the muon spin precession with more than 20 times the statistics of the E821 experiment, while controlling the systematics at the 100ppb level.  

One of the improvements made upon the E821 design is a system of straw tracking detectors, built here at Liverpool. These straw trackers act as a non-destructive beam profile measurement for the duration of each muon fill. 

The straw tracker modules are installed at two locations around the vacuum chamber such that decay positrons will pass through multiple straws on their way to the calorimeters. The straws that are hit can be used to reconstruct positron trajectories which can be extrapolated back to the decay vertex in order to build up the beam profiles. Forwards extrapolation is also made possible to aid the calorimeter calibration, reducing systematic uncertainties associated with pileup.  

The intention of this project is to further improve the prospects of the straw tracking detectors by implementing elements of machine learning to allow for real time track reconstruction, and thus real time construction of muon beam profiles. Simulated data will first be used in order to familiarize key concepts of the machine learning algorithm in a generalized form before it will be applied to the specific case of the g-2 experiment using online data. 

It is then the intention for the algorithm to be applied online data such that track reconstruction using the straw trackers can be achieved in real time.