Speaker
Description
The scattering of a well-controlled probe particle is a time-honored method to explore the structure and dynamics of matter. In heavy-ion collisions, the interaction of jets with the Quark-Gluon Plasma (“jet quenching”) promises such a probe. However, well-controlled jet reconstruction in the complex environment of nuclear collisions is extraordinarily difficult – the thousands of particles generated in a head-on collision of lead ions at the LHC do not come labeled “jet” or “background.” The HEP jet toolkit for mitigating underlying event and pileup effects in p+p collisions is helpful for very energetic jets in this environment, where the signal/background is large. However, such jet-wise correction approaches fail badly for by far the largest part of jet production phase space, where jet quenching effects are expected to be sizable, and where we need to measure to track the dissolution of jets in the plasma. Machine Learning is useless in this region, because its training requires accurate background modeling (which is what the problem is). The solution, unsurprisingly, is to think about the problem statistically, carrying out data-driven background corrections at the level of ensemble-averaged distributions. This approach has been successfully applied in a wide variety of heavy-ion jet measurements at both the LHC and RHIC, enabling jet quenching measurements over the full jet production phase space at both colliders, and mapping experimentally how moderate-energy jets get blown apart by the QGP and how the QGP responds to that excitation. I will outline the statistical approach to heavy-ion jet measurements, and discuss notable recent results employing it from both the ALICE and STAR experiments.