17–22 May 2026
C.I.D
Europe/Zurich timezone

Learning beam dynamics in the latent space of beam distributions

THP5347
21 May 2026, 16:00
2h
C.I.D

C.I.D

Deauville, France
Board: Thursday baguette: BF09
Poster Presentation MC5.D08: High Intensity in Linear Accelerators Space Charge, Halos Poster session

Speaker

Ningdong Wang (Cornell University)

Description

Beam dynamics under collective effects such as space charge remains a computationally expensive challenge. We present a latent space surrogate model for collective beam dynamics that significantly accelerates these simulations. The method uses a variational autoencoder to compress 6D particle distributions into a low-dimensional latent space. A learned latent-space dynamical model then predicts beam evolution directly in the latent space, bypassing expensive space charge solvers. Using simulated data from a space charge dominated lattice, this approach reproduces beam envelope evolution with good agreement to particle-in-cell codes while offering substantial speedups. This framework provides a flexible path towards fast beam prediction for online accelerator modeling.

Funding Agency

This work was supported by the National Science Foundation under Grant No. PHY-1549132 and the Department of Energy's grant DE-SC0025351 and DE-SC0024907.

Paper status =Paper accepted (after submitter validation).

Author

Ningdong Wang (Cornell University)

Co-authors

Georg Hoffstaetter (Cornell University (CLASSE)) Ireanne Cao (Cornell University)

Presentation materials