Master thesis defense by Lasse Skaaning Husbjerg
Title: Improvements of Trigger Efficiency in T0+ on Ultra Peripheral Pb-Pb Events at sNN = 2.76 TeV using Machine Learning
Abstract: The classification problem of Quantum Electrodynamic (QED) background versus ultra peripheral Pb-Pb events at sNN = 2.76 TeV using the upgraded T0+ detector system is investigated.
This problem is particularly difficult due to the massive difference in cross section and interaction frequency between the background and the hadronic events, with 79 kbarn for the background versus 8 barn for the Pb-Pb events.
To study this classification problem, a simulation consisting of 20 microchannel plate photomultiplier (MCP-PMT) detectors on the A side and 28 MCP-PMT detectors on the C side of the interaction point in ALICE is developed.
Before applying the simulation on data generated with HIJING, the simulation is thoroughly investigated through various tests, such as measuring the multiplicity resolution. Ultra peripheral events with centrality 80-90% at sNN=2.76 TeV and QED background are then generated using HIJING and several different analysis methods are applied on the background classification.
Using threshold analysis on the number of detected Cerenkov photons results in a trigger efficiency of 93.2% on the ultra peripheral Pb-Pb events, with a background trigger rate of less than 81 Hz.
Applying a Support Vector Machine (SVM) instead of a simple threshold increases the trigger efficiency of T0+ on the ultra peripheral events to 95.9% with less than 83 Hz, corresponding to an increase in trigger efficiency of 2.7 percent.