The phenomenology of dark matter is complicated if dark matter is a composite particle as a hadron under a dark gauge group. Once a dark parton is produced at a high energy collider, it showers and evolves to a jet-like object, eventually it provides a collider signature depending on interactions with particles of the Standard Model (SM). For example, a finite lifetime of dark hadron would provide a displaced vertex. Thus by considering features in various subdetectors, one can identify a jet from a dark parton ("dark jet") with analysis methods in conventional exotic searches. However if the lifetime of the dark hadron is collider-negligible (too short to manifest a displaced vertex), it would be hard to tag a dark jet over Quantum Chromodynamics (QCD) jets of the SM. Thus conventional analyses with information from various sub-detectors are not enough to probe dark matter physics in general at colliders. We propose an analysis to utilize a combination of jet-substructure variables to identify dark jets over backgrounds. We study features of jet-substructure variables for a dark jet. We identify what parameters in dark jet are relevant to performance of a given jet-substructure variable.
To maximize performance we apply a boost decision tree (BDT) to jet-substructure variables in tagging dark QCD jet over QCD jets. Our result shows that by combining various jet-substructure variables, one could get a good discrimination performance to identify a dark jet over QCD backgrounds.