BEGIN:VCALENDAR VERSION:2.0 CALSCALE:GREGORIAN PRODID:UW-Madison-Physics-Events BEGIN:VEVENT SEQUENCE:0 UID:UW-Physics-Event-9332 DTSTART:20250725T150000Z DTEND:20250725T170000Z DTSTAMP:20260413T102702Z LAST-MODIFIED:20250714T170956Z LOCATION:https://uwmadison.zoom.us/j/93565126777?pwd=k2hKi3eHH2dZ0QmL9 hxL6fgUW6uHBE.1 SUMMARY:Search for Dark Matter with the ATLAS Detector and Development of a Novel Track Reconstruction Algorithm based on Graph Neural Netwo rks for the ATLAS Inner Tracker\, Thesis Defense\, Tuan Minh Pham\, Ph ysics PhD Graduate Student DESCRIPTION:This thesis is divided into two main parts. The first part presents a summary of dark matter searches performed by the ATLAS exp eriment and a statistical combination of the three most sensitive anal yses. The results are interpreted within the framework of a Two-Higgs- Doublet Model extended by a pseudoscalar mediator (2HDM+a). These anal yses are based on 139 fb-1 of proton-proton collision data collected a t a center-of-mass energy of 13 TeV during Run 2 of the LHC. The combi ned analyses target final states involving large missing transverse en ergy and a visible signature from the decay of a Standard Model Higgs boson or Z boson\, as well as processes involving the production of ch arged Higgs bosons. This work provides the most comprehensive set of c onstraints on the 2HDM+a model published by ATLAS to date.
\n
\n The second part focuses on the reconstruction of charged-particle trac ks in the ATLAS Inner Tracker (ITk)\, which is confronted with the ext reme pile-up conditions expected in the High-Luminosity phase of the L arge Hadron Collider (HL-LHC). Given the anticipated increase in insta ntaneous luminosity and associated event complexity—resulting in up to 200 simultaneous interactions per bunch crossing—traditional reco nstruction algorithms face significant computational challenges. To ad dress this\, a novel track reconstruction algorithm based on Graph Neu ral Networks (GNNs) has been developed and evaluated. Using full detec tor simulation data on realistic ITk geometry\, we demonstrate competi tive physics performance of the GNN-based tracking approach with respe ct to the current tracking algorithm. The computational efficiency is optimized and measured in detail. This approach shows significant pote ntial for efficient pattern recognition in dense detector environments \, leveraging modern hardware accelerators such as GPUs and FPGAs for fast and scalable event reconstruction. URL:https://www.physics.wisc.edu/events/?id=9332 END:VEVENT END:VCALENDAR