Background Mitigation in Dual Phase Xenon Time Projection Chambers

The focus of this thesis is on the identification and removal of background events in the dual-phase xenon TPC experiments of LUX and LZ. In particular, this effort focuses on a few hard to classify background types that appear in the traditional WIMP search and in the EFT search. The first half of the original research presented in this thesis pertains to backgrounds expected to be present in LZ. This includes multiple scatter events which only deposit a proportion of their energy in the LXe, and multiple scatter events which only produce a single ionisation signal. The thesis then goes on to examine how backgrounds such as these, and other backgrounds occurring close to the edge of the LXe can be reduced with improved modelling of ionisation electron paths in the LXe, and describes how this modelling was used by the LZ collaboration. The second half of this thesis focuses exclusively on these multiple scatter events which only produce a single ionisation signal but in the context of LUX data. It begins by examining how previous studies have identified and removed them; then presents a novel method for the identification and removal of these events using a machine learning algorithm called a BDT. Finally, the thesis concludes after using data and simulations to compare this new BDT cut against previously developed cuts for this class of events, and shows this new BDT cut to be the best option for the ongoing EFT analysis of LUX.