A Data-Driven Magnetic Field Map for Muon g-2 Using Physics-Informed Neural Networks

In 2023, the Fermilab Muon g-2 collaboration released the Run-2/3 results of the muon's magnetic moment anomaly with a uncertainty of 215 ppb. Part of the experiment involves a precision measurement of the magnetic field in the muon storage ring. In both the Run-1 and Run-2/3 published values, the magnetic field map is determined using a model of multipole moments. This model was converged upon by two independent teams during the Run-1 analysis. In this talk, we present a new, fundamentally different method of generating the magnetic field map from the field measurements using machine learning and Physics-Informed Neural Networks (PINNs). This experiment presents an ideal opportunity to test Implicit Neural Representations (INRs) for field mapping for two key reasons. First, the Muon g-2 field measurement has undergone rigorous verifications with high levels of confidence. This provides a peer-reviewed method to compare to the results from the new neural network. Second, there are six runs worth of magnetic field data that can be used for training, validation, and testing, alleviating the data-hungry nature of neural networks. This talk will include a brief introduction to INRs and PINNs, an overview of the Muon g-2 field measurements, and a comparison with results from the multipole moment analysis.Mu