IPAC'25 - the 16th International Particle Accelerator Conferece

Predicting 3D electric potentials and particle tracking with U-nets

WEPM049

4 Jun 2025, 16:00

2h

Exhibiton Hall A _Magpie (TWTC)

Poster Presentation MC5.D02 Nonlinear Single Particle Dynamics Resonances, Tracking, Higher Order, Dynamic Aperture, Code Developments Wednesday Poster Session

Speaker

Isabella Vojskovic (European Spallation Source)

Description

This study investigates a 3D convolutional encoder-decoder architecture with skip connections to predict coupled 3D electric potentials from charged particle distributions. Using automatic differentiation, the corresponding electric fields are then computed from the predicted potentials.
Our findings demonstrate that the network provides a computationally efficient alternative to traditional numerical methods for solving non-linear space-charge problems in beam dynamics simulations, offering significant potential for improving simulation speed and scalability.