Searching for Dark Matter at the LHC: Evaluating and Optimising Performance of a Higgs Event Classifier

Abstract

Dark Matter comprises more than 25% of all mass-energy in the Universe. Very little is known about it, though it is unlikely to be a known Standard Model particle. Since the discovery of the Higgs boson in 2012, new direct production Dark Matter searches at the LHC have become possible through the examination of events in which the Higgs recoils off undetected particles. This ‘missing momentum’ signature can be identified in the CMS experiment as evidence for Dark Matter. In order to study such events, it is first necessary to identify, or ‘tag’ objects that are consistent with having come from the decay of a Higgs boson. We tag these ‘jets’, which are narrow cones of particles formed by hadronisation of a quark or gluon, as likely Higgs candidates by searching in the most prevalent decay channel, \(H \rightarrow b\bar{b}\). Specifically, we look for events that produce a high-momentum, ‘boosted’ bb jet, since Dark Matter events are far easier to identify against backgrounds when they occur with large momentum transverse to the beam line. Using a Boosted Decision Tree algorithm, we evaluate the Higgs tagger currently in use at CMS and attempt to optimise it by increasing signal efficiency while simultaneously decreasing background efficiency.