报告题目:  Beyondstandard model physics search and the applications of machine learning

报告人: A. Hammad, Departmentof Physics, University of Basel

报告时间:  2021年3月1日下午4点

报告地点: 吉林大学前卫校区物理楼302，在线报告 (online)  https://cern.zoom.us/j/65605627941?pwd=ejh6M3JVWTRHOVNiSmpVelJsVUVEQT09

报告摘要： Since theStandard Model (SM) cannot explain some of the measured phenomena, people startlooking for direct and indirect signatures for physics beyond the SM.

Experimentalsearches for charged lepton flavor violation (CLFV) are the most sensitiveprobes of physics beyond the SM. We discuss the sensitivity of the prospectiveLarge Hadron-electron Collider (LHeC) for CLFV (Indirect) search in aneffective theory approach considering a general effective Lagrangian for theconversion of an electron into a muon or a tau via the effective coupling toneutral gauge bosons. We show that the LHeC could already probe CLFV signatureswith high sensitivity because the converted charged lepton is dominantlyemitted in the backward direction of the detector, enabling an efficientseparation of the signal from the background.

Asan example of the direct search we discuss the signatures of extra scalar Higgsat the future colliders. If the extra Higgs discovered at the LHC or futurecolliders, the question will arise whether CP is violated or conserved in theextended scalar sector. An unambiguous probe of CP violation would be theobservation that one of the extra Higgs particles is an admixture of a CP-evenand a CP-odd state. We discuss the prospects of determining the CP property ofan extra neutral Higgs state H via the angular distribution of final states inthe decay pp → H → τ τ at the high-luminosity (HL) phase of the LHC. Thefact that the new physics cross section has to be small as well the tau pairresonance has poor kinematics, make the normal cut and count analysisinsufficient and has to be extended with Machine Learning (ML) analysis.

Themain goal of Machine Learning Algorithms (MLAs) is to exploit the large datasets in order to reduce the complexity of the data and find features of newphysics hidden in the data. The current most frequently used MLAs are theBoosted Decision Trees (BDT) and the Deep Neural Networks (DNN). Although manydifferent models for machine learning exist, still all of them (as supervisedlearning algorithms) have a common way of dealing with the problem ofextracting new physics signatures from the large background contributions. Oneof the MLAs objectives is its ability to probe the signatures of heavyparticles at the high luminosity LHC with high accuracy. We discuss in detailshow the MLAs (specifically BDT and DNN) can be utilized to improve the collidersearch.

举办单位：36365线路检测中心

　　　　　吉林大学理论物理中心