COLLOQUIUM by PhD candidates

Friday 3 May 2024, 11:00 → 12:00 Europe/Amsterdam

Colloquium Room (Nikhef Vertex)

"Resolving the proton content to percent accuracy" by Giacomo Magni (VU, Nikhef)

Abstract:

Hadronic Colliders, as the LHC, are the main tool used to investigate Standard Model parameters and properties. A faithful interpretation of their measurements requires both a complete knowledge of the detector's behaviour, as well as the computation of precise theoretical predictions of the observed quantities. In the latter, one of the main sources of uncertainty comes from the incomplete knowledge of Parton Distributions (PDFs), i.e. the functions describing how the initial state quarks and gluons are mixed inside the colliding hadrons (typically a proton). PDFs encode the non-pertubative nature of the QCD bound states and their a priori computation is highly non-trivial. However, thanks to their universality, they can be fitted from "simpler" experimental measurements, such as electron-proton scattering, which are compared with accurate theoretical predictions. In this context, I will present the progress to extend PDFs determination to approximate 3-loop accuracy in QCD or account for QED effects, discussing their implications for some cross-sections at the LHC.

 

"Advancing the next generation of gravitational wave searches with machine learning" by Melissa Lopez (UU, Nikhef)

Abstract:

Gravitational waves provide a fresh perspective for exploring our cosmos. They present a unique opportunity to investigate the formation of intermediate mass black holes, which serve as a bridge between stellar mass and supermassive black holes. Until the detection of GW190521, these objects remained elusive, but this discovery has shed light on stellar evolution and galaxy formation. Observing their population would provide valuable insights, but they are difficult to detect due to their similarities with the background of gravitational wave detectors. To enhance the current search for intermediate mass black holes, we propose integrating the state-of-the-art model-based method with a machine learning algorithm. This combination creates a synergistic relationship. In this presentation, we demonstrate the high performance of this method using both simulated and real data, and discuss the potential and challenges of machine learning-based search algorithms.