Kyle Amirie ; Steven Robertson

2020
McGill University Montreal

Abstract: The presence of beam background-related photons in the extra energy distributions of events reconstructed using Belle II data can make it difficult to accurately identify en- ergy missing from the detector that can be attributable to neutrinos or other hypothe- sized weakly interacting particles. By isolating samples of detector information relating to these beam background photons, and samples relating to photons from the intended collision event (signal photons), it is possible to perform a multivariate analysis using a boosted decision tree approach to train a classifier variable for distinguishing between the two types of photons. This classifier can be used to suppress the inclusion of beam background-related photons in extra energy distributions. Six classifier tools were trained on electromagnetic calorimeter information relating to beam background and signal photons from e+e− → γ → μ+μ− and B0 → D∗+l−νl events. These tools were validated as a beam background suppressant when used in reconstructed B0 → D∗+l−νl events to evaluate their performance. Described within is a full description of all six trained classifiers, an examination of the impact training sample characteristics have on their performance, and a number of possible future optimizations to this approach to beam background photon suppression.

Note: Presented on 01 11 2020
Note: MSc