-----------------------------------------------------------
** Date: January 22 (W) 4:30 pm - 5:30 pm** Place: B109** Speaker: Dr. Kihong Park (IBS-CUP)** TITLE : “Study of dark photons at electron-positron colliders based on high performance computing and machine learning"** Abstracts:
Approximately 380,000 years after the Big Bang, analysis of the Cosmic Microwave Background (CMB) radiation revealed that the universe contains more than five times as much dark matter as visible matter. Dark matter, which has existed in the universe since its early stages, is believed to play a crucial role in forming and maintaining major structures in the universe, such as galaxies and galaxy clusters, through gravitational interactions. Therefore, dark matter is not only the dominant matter of the universe but also serves as a key to understanding the evolution and astronomical phenomena of the universe. However, dark matter interacts with ordinary matter primarily through gravity. In addition, the cross-section of dark matter is more than 1000 times smaller than that of the Standard Model particles. Due to these challenges, dark matter is not directly detected even though intensive efforts are made to detect dark matter experimentally based on various theoretical models.
Because there is no suitable dark matter candidate in the Standard Model, dark sector composed of unknown particles can be introduced. Similar to how photons mediate electromagnetic interactions in the Standard Model, the dark sector may contain dark photons. These hypothetical particles can collapse into the Standard Model, allowing the analysis of data left by Standard Model particles in detectors to search for dark photon signals. However, effective removal of background events and massive data production are required to search for dark photon signals. In this regard, researches based on high performance computing and machine learning techniques are expected to be promising.
Thus, we studied dark photons at electron-positron colliders based on high performance computing and machine learning. The experiments included the current Belle II experiment and future experiments of the Circular Electron-Positron Collider (CEPC), Future Circular Collider (FCC)-ee, and International Linear Collider (ILC). Specifically, the Belle II experiment at KEK (High Energy Accelerator Research Organization) in Japan is an international collaboration. Korea Institute of Science and Technology Information (KISTI) has played role in data handling system of the Belle II experiment. During the COVID-19 pandemic, we have developed the remote control room and have operated control room shift remotely. Therefore, we have taken data acquisition experiment even if we do not visit the KEK, Japan.
We investigated the effects of variables of center-of-mass energy, dark photon mass, and coupling constant on the cross-section of signal event modes, including single dark photon mode (e+e- -> mu+mu-A’) and double dark photon modes (e+e- -> A’A’ and e+e- -> A’A'gamma).
Using the KISTI-5 supercomputer, we efficiently generated signal and background simulation data. One million events were produced for each center-of-mass energy of 91, 160, 240, 250, 350, 500, and 1000 GeV. A simplified model is used for the signal events and the Standard Model is used for background events. Delphes simulation was performed to simulate detector responses for the CEPC (91, 160, and 240 GeV), FCC-ee (91, 160, 250, and 350 GeV), and ILC (250, 500, and 1000 GeV). We then reconstructed dark photons from detector data. To reduce the background events, we used Boosted Decision Trees as a machine learning method. As a result, for each experiment, background events were effectively reduced, and high-purity signal events were obtained. The predicted numbers of signal events and detector efficiencies were calculated.
The results of this study will serve as a reference of searching for dark photons in future experiments. In addition, the methodologies used in this study will help searching for dark matter at current experiments of Belle II and BESIII.
—————————————————————————————