Comprehensive Evaluation of Simulation Approaches for the CMS Experiment at the LHC, CERN

Norraphat Srimanobhas^1*

¹Physics, Chulalongkorn Universiry, Bangkok, Thailand

* Presenter:Norraphat Srimanobhas, email:Norraphat.Srimanobhas@cern.ch

This comprehensive review delves into three distinct simulation methods employed in the context of the Compact Muon Solenoid (CMS) experiment at CERN's Large Hadron Collider (LHC). The first approach, known as "Full Simulation" (FullSim), is based on the Geant4 detector simulation, renowned for its precision. The second method, "Fast Simulation" (FastSim), capitalizes on parametric approximations to achieve substantial computational efficiencies. The third method introduces a groundbreaking "Flash Simulation" (FlashSim), which incorporates a Machine Learning-based approach to significantly expedite simulations. We provide valuable insights into their applications, performance, and potential developments.

FullSim has been put to use in the framework of Run-3 and Phase-2, demonstrating notable improvements in Geant4 versions, physics lists, and magnetic field tracking. FastSim, which operates at approximately ten times the speed of FullSim, plays a pivotal role in handling the increased luminosity and detector complexity, albeit with a trade-off of reduced accuracy. To address this challenge, a machine learning-based technique has been introduced to refine FastSim results.

Furthermore, we introduce FlashSim as a pioneering approach that employs Normalizing Flows generative models, yielding remarkable speed enhancements while retaining a high degree of physics accuracy. We engage in a thorough discussion of the trade-offs and distinctive features of each method, emphasizing the potential for their coexistence and mutual reinforcement in meeting the evolving simulation requirements of the CMS experiment in Run-3 and Phase-2.

Keywords: Large Hadron Collider, Simulation, Geant4, Machine Learning