Developing Machine Learning methods to constrain the properties of the Quark-Gluon Plasma

Supervisors: Jaime Norman, Roy Lemmon, Marielle Chartier (UoL)
Institution: University of Liverpool

At sufficiently high energy densities QCD predicts that hadronic matter exists in the form of a deconfined state of quarks and gluons, the Quark-Gluon Plasma (QGP), and not in the form of confined, colour-singlet, constituent hadrons. This state of matter filled the universe up to a few microseconds after the Big Bang and is produced in the lab using ultra-relativistic heavy-ion collisions. The study of properties of the Quark-Gluon Plasma, which include specific viscosity, energy transport, and microscopic structure, aims to tell us about the properties of the universe at its early stages, and in general the properties of QCD matter in its most extreme phase.

Measurements of many different observables have been made in heavy-ion collisions at the LHC (CERN, Geneva) and RHIC (USA), all of which must be sensitive to the same underlying physics, and detailed comparison of measurements to theoretical models are required to extract quantitative information from these collisions.

This PhD project will focus on developing novel algorithms and machine learning-based approaches to measurements in heavy-ion collisions, and phenomenological approaches to rigorously compare measurements to models. This project will involve two main research directions:

  • The student will work on the ALICE experiment – the LHC experiment dedicated to studying heavy-ion collisions. In particular, they would focus on the measurement of ‘jets’ – collimated sprays of hadrons produced in high-energy particle collisions which provide a unique probe of the QGP as they interact with the QGP at all stages of its evolution. The student will explore the use of Machine Learning as a tool to improve the measurement of jets.
  • The student will work on developing and optimising techniques to improve data-model comparisons of heavy-ion data, as part of the JETSCAPE collaboration. Bayesian parameter estimation provides the most natural method for rigorous, multi-observable data-model comparisons, and they will explore and develop new, more efficient approaches for Bayesian analyses, in particular focusing on approaches to optimise posterior sampling.

The student will be provided with comprehensive training in data science through LIV.INNO’s structured training program, as well as courses on theoretical and experimental particle physics. A 6-month industry placement will complement your training. This project will be carried out over 48 months based at the University of Liverpool but you have the opportunity to spend a period of time at CERN.

https://www.liverpool.ac.uk/study/postgraduate-research/how-to-apply/

Please ensure you state Dr Norman, Prof Chartier and Dr Lemmon as the proposed supervisors on your application form and quote studentship reference: PPPR056.

View on Find a PhD