Implementation of Machine Learning Techniques in the Search for Emerging Jets Using the ATLAS Run II Dataset

A search for the novel experimental signature known as ’emerging jets’ in the ATLAS Run II dataset is presented. The emerging jets model assumes a QCD-like dark sector in which jets of dark particles can shower and hadronize. These dark hadrons eventually decay back to Standard Model particles and produce jets at various displacements from the interaction point. These jets ’emerge’ into the detectors, producing a unique signature at collider experiments. Introducing a boosted decision tree into this analysis enables sensitivity to several emerging jets models which cover a wide parameter space, allowing for the possibility of discovery, or alternatively excluding the current theoretical cross-sections. This is shown via a complete Monte-Carlo study, including a thorough evaluation of systematic uncertainties, which corresponds to proton-proton collisions at $\sqrt{s}$ = 13 TeV, and the full ATLAS Run II integrated luminosity of 139 fb$^{−1}$. Although the analysis is currently still blinded to data in the search region, it is validated using a selection of background-only Data which can be compared to the corresponding background Monte-Carlo.