Identifying quenching effect in heavy-ion collisions with machine learning

Quantum chromodynamics (QCD), the theory of the strong interactions, predicts a deconfined state of quarks and gluons, known as quark gluon plasma (QGP) at high temperature and/or density. This extremely hot and dense matter, believed to exist in the early universe, can be recreated in heavy ion collisions at particle accelerators such as RHIC and LHC. One of the key signatures of QGP formation is the quenching of jets, which are collimated sprays of hadrons, due to their interaction with QGP leading to parton energy loss and jet substructure modifications. While these modifications have been observed experimentally with a variety of methods averaged over many collision events, the jet-by-jet identification of quenched jets remains difficult.

We designed a machine learning approach to identify quenched jets based on their substructure. The jet showering processes are simulated with a jet quenching model Jewel and a non-quenching model Pythia 8. Sequential substructure variables are extracted from the jet clustering history following an angular-ordered sequence and are used in the training of a neural network built on top of a long short-term memory network. We show that this approach successfully identifies the quenching effect in the presence of the large uncorrelated background of soft particles created in heavy ion collisions.