Differentiable simulations for particle tracking in accelerators: analysis, benchmarking and optimization
Francisco Huhn, Francesco M. Velotti · 2025
Read Differentiable simulations for particle tracking in accelerators: analysis, benchmarking and optimization by Francisco Huhn, Francesco M. Velotti (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-08476.txt.
Description
Designing a particle beam line means tuning many settings at once, and slow search methods scale badly as settings multiply. Francisco Huhn and Francesco Velotti explore a smarter route in this 2025 paper: simulations that hand back exact gradients. They built their own such code using automatic differentiation and tested it against the usual trial-and-error gradient estimates.
Two test cases anchor the study, a literature cell with space charge and a realistic future beamline at CERN. The differentiable code matched estimates while running far more cheaply per gradient. Plugged into gradient-based optimizers, it beat gradient-free methods on the tuning tasks. The authors frame the work as a first careful benchmark for this approach in accelerator physics.
Catalog record
| CREATOR | Francisco Huhn, Francesco M. Velotti |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-08476