Artificial Intelligence (AI) could revolutionize nuclear physics, according to a recent study from the Institute of Nuclear Physics of the Polish Academy of Sciences (IFJ PAN) in Krakow.
The researchers suggest that AI might soon replace current methods for reconstructing the trajectories of secondary particles created during particle collisions in accelerators like the Large Hadron Collider (LHC).
The principle of reconstructing secondary particles' trajectories based on the impacts recorded during collisions inside the MUonE detector. The successive targets are marked in yellow, and the silicon detector layers are in blue.
Credit: IFJ PAN
In high-energy physics experiments, the process of reconstructing secondary particles' trajectories is complex, especially due to the magnetic field in the detectors causing charged particles to deviate. Classical methods struggle to keep up, particularly with the anticipated increase in collision energy and the number of generated secondary particles. AI, with its ability to quickly recognize universal patterns, could provide an effective solution.
The IFJ PAN team developed an AI in the form of a deep neural network, comprised of five layers and two million configurable parameters, trained with 40,000 simulated particle collisions. This AI has learned to accurately reconstruct particle trajectories, a major advancement for detection techniques.
The MUonE experiment, scheduled to begin at CERN, will represent an opportunity for this AI to prove its effectiveness in a real context. MUonE focuses on an anomaly observed in the magnetic moment of muons, which could indicate the existence of hitherto unknown physical phenomena. The use of AI in this experiment could increase the accuracy of the predictions of the Standard Model, thus potentially contributing to a significant discovery in particle physics.
Thus, AI could mark the beginning of a new era in particle detection techniques, with its deployment expected within the MUonE experiment starting next year, and a targeted phase planned for 2027.