000003902 001__ 3902
000003902 005__ 20231005131836.0
000003902 037__ $$aBELLE2-MTHESIS-2023-037
000003902 041__ $$aeng
000003902 100__ $$aYuxin Liu
000003902 245__ $$aStudy of Sudden Beam Losses of SuperKEKB and Development of 3D Track Hardware Trigger using Machine Learning at Belle II
000003902 260__ $$aTsukuba$$bSOKENDAI$$c2023
000003902 300__ $$amult. p
000003902 500__ $$aPresented on 25 07 2023
000003902 502__ $$aMSc$$bTsukuba, SOKENDAI$$c2023
000003902 520__ $$aThe Belle II Experiment, located at the SuperKEKB asymmetric electron-positron collider in Japan, is at the next generation of B-factories, aiming to explore new physics (NP) in the flavor sector and enhance the precision of Standard Model (SM) measurements. SuperKEKB is expected to achieve the luminosity of $6 \times 10^{35} , \text{cm}^{-2} \text{s}^{-1}$ , enabling unprecedented NP searches and measurements of the CKM matrix. However, the increase of higher luminosity faces challenges of sudden beam loss events and increasing level-1 trigger rates. Sudden beam loss events, characterized by rapid beam loss within a few turns, pose risks to the SuperKEKB and Belle II components, with the underlying causes still unknown. To address this, beam loss monitor with fine timing resolution have been installed to pinpoint the location of initial beam loss. Timing analyses have identified the LER D06 section as the region where the earliest loss occurs, suggesting the occurrence of initial beam instability. Based on the analysis results, countermeasures, such as fast beam abort and additional sensors at the D06 section, are planned to protect detectors and collimators from sudden beam loss. The 3D Track Hardware Trigger, responsible for triggering physics events, faces limitations of trigger rate. With increasing luminosity, the large increased background trigger rate nears the maximum limit. Based on the newly deployed fourth generation universal trigger board (UT4, which have general 4 times logic gates than previous board, new neural-network 3D track trigger architectures have been developed with software simulations. These architectures achieved an 50 \% reeducation of total CDC trigger background. Further work will focus on simplifying the architecture and implementing it in UT4 modules.
000003902 700__ $$aAkimasa Ishikawa$$edir.
000003902 700__ $$aTaichiro Koga$$edir.
000003902 8560_ $$fliuyux@post.kek.jp
000003902 8564_ $$uhttps://docs.belle2.org/record/3902/files/BELLE2-MTHESIS-2023-037.pdf
000003902 980__ $$aTHESIS