Applying machine learning to determine impact parameter in nuclear physics experiments

• DOI: 10.48550/arxiv.2107.13985
• Publication Date: 2021-01-01
• Authors: C. Y. Tsang; Yongjia Wang; M. B. Tsang; J. Estee; T. Isobe; M. Kaneko; M. Kurata-Nishimura; J. W. Lee; Fupeng Li; Qingfeng Li; W. G. Lynch; T. Murakami; R. Wang; Dan Cozma; Rohit Kumar; Akira Ono; Yingxun Zhang
• Keywords: Impact parameter; Experimental data

Machine Learning (ML) algorithms have been demonstrated to be capable of predicting impact parameter in heavy-ion collisions from transport model simulation events with perfect detector response. We extend the scope ML application experimental data by incorporating realistic response S$\pi$RIT Time Projection Chamber into generated UrQMD resemble data. At 3 fm, predicted is 2.8 fm if used for training and testing; 2.4 included testing, 5.8 trained applied testing that has The last result not acceptable illustrating importance including developing algorithm. also test dependence applying on simulated four different models as well using input parameters model. Using Sn+Sn at E/A=270 MeV, determined agree experimentally multiplicities, except very central peripheral regions. selects collision better allows determination beyond sharp cutoff limit imposed methods.