Predicting Leading Vertex Multiplicity using Graph Neural Networks in High Pileup Environments + Electron Ion-Collisions and their Soft Photons

by Rubi Hason + Eden Mautner

at Particles and Fields Seminar

Mon, 15 Jul 2024, 14:00
Sacta-Rashi Building for Physics (54), room 207


[This week we again have two half hour talks]
[1st talk] At the Large Hadron Collider (LHC), proton collisions produce two main types of interactions: hard scatter and soft scatter (pileup). The average number of pileups per event has increased significantly over the years, from 14 in 2015 to 40 in 2018 during Run 2. The pileups are projected to exceed an average of 200 per event by 2026 when transitioning to the High-Luminosity LHC.
A high number of pileup events alongside the hard scatter process introduces difficulties, as tracks originating from pileups can be mistakenly classified as originating from the leading vertex. This misclassification affects the multiplicity of the leading vertex, which is defined as the number of charged particles associated with it.
In this study, we use the ACTS framework which starts by simulating collisions using Pythia8, which is then propagated through the detector simulation and then undergoes reconstruction for the tracks and vertices. We use the output data to train a machine learning model (GNN) which predicts the multiplicity of the leading vertex for each event.

[2nd talk] The Electron-Ion Collider (EIC) is a particle collider planned for construction at Brookhaven National Laboratory, aiming to be operational in the early 2030s. Investigating collisions of high-energy electrons with high-energy ion beams at the EIC may reveal fundamental properties of nucleons and nuclei, touching on the origins of nuclear mass and the distribution of matter at the nuclear level. One of many new and exciting measurements that could be performed at the EIC is the measurement of exclusive vector-mesons production, such as J/psi, with the ability to differentiate, for the first time, between coherent and incoherent production.
In this research I examine incoherent, quasi-coherent, and coherent vector-meson production in ePb collisions, produced by using advanced Monte Carlo event generators such as BeAGLE and eStarLight. The events are then simulated through a detailed virtual representation of the ePIC@EIC detectors so we could learn more about both the detectors measurement sensitivity and about the physics behind these events. One of the main goals in this study is to help us design the new B0 detector, maximizing it’s measuring capabilities, while taking into account the complex geometrical constraints that we encounter.
In this talk my goal is to inform about the EIC project, specifically about the progress that we have made towards building the B0 detectors, and talk about the importance of including the new B0 detector at the EIC when aiming to measure processes that have never been measured.

Created on 10-07-2024 by Citron, Zvi (zhcitron)
Updaded on 10-07-2024 by Citron, Zvi (zhcitron)