Project Title: Ultra-fast ML Algorithms for the ATLAS Global Hardware Trigger

Summary: The next phase of the LHC programme, called High-Luminosity (HL-) LHC and planned for a start at 2030, will generate unprecedented data rates and volumes. In particular, bunches of protons collide every 25ns (40MHz rate) and in every crossing, we expect of order 200 individual proton-proton interactions. To tackle the associated challenges, the ATLAS experiment is carrying out an ambitious programme of upgrades, including to the Trigger and Data AcQuisition (TDAQ) system. The Global Hardware Trigger (GHT) is an entirely new component of the ATLAS TDAQ system. It will receive a data bandwidth of ~50Tbps and has a few microseconds to reject ~97.5% of uninteresting events (1MHz output rate), while keeping the maximum possible acceptance for events in which interesting processes have occurred, for example pair-production of Higgs bosons. In this summer project, you will work as part of the UCL ATLAS GHT team to develop and optimize ML algorithms for the selection of events where one or more energetic electrons or photons were produced, using simulated ATLAS data in HL-LHC conditions. Good knowledge of python and some experience in machine learning (e.g. at the level of the 3rd year course “Practical Machine Learning for Physicists”) and the use of the standard ML libraries (tensorflow, scikitlearn etc) would be essential for doing well in this project.

For more information on all of the above, contact Prof. Nikolaos Konstantinidis