Discriminating Cosmics from Collision Muons in the ATLAS Detector

In this report the topological and timing differences between muons arising from cosmic rays and muons from proton-proton collisions at the LHC are explored. Topological information from muon segments and jets reconstructed by the ATLAS detector is studied, using data from Run-2 of the LHC, which lasted between June 2015 to November 2018. By reducing the detector’s data down to a simpler format containing just the important observables, multivariate analysis was carried out successfully by implementing the TMVA tool native to ROOT. The gradient boosted decision tree machine learning algorithm was utilised for muon segment observables in order to generate a composite variable of 25.50 maximum discriminating significance at an optimal cut of -0.23. In addition, it was discovered that the properties of jet width and timer provided a precise discrimination between signal and background. The composite variable had a maximum significance of 31.23 at a cut of -0.07, making jet properties a more accurate discrimination method than segments. Matching the segments to the jets resulted in a maximum significance of 31.38 at a cut of -0.07, allowing for an accurate preliminary study on cosmics using boosted decision trees.