Jet characterization in Heavy Ion Collisions by QCD-Aware Graph Neural Networks

: The identification of jets and their constituents is one of the critical problems and challenging tasks in heavy-ion experiments such as experiments at RHIC and LHC. The huge background of soft particles poses a curse for jet-finding techniques. The inabilities or lack of efficient techniques to filter out the background lead to a fake or combinatorial jet formation which may have an erroneous interpretation. This article presents the GraphReduction technique (GraphRed), a novel class of physics-aware attention graph neural networks built upon jet physics in heavy-ion collisions. Since, the number of tracks are expected to vary from one event to another, this approach finds the most likely jet constituent particles on an event-by-event basis. This technique demonstrates the robustness and applicability of this method for finding jet constituent particles and shows the applicability of graph architectures on particle-level classification in each heavy-ion event produced at the LHC.