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

Differentiable, CPU-/GPU-Parallelized, Symplectic Spin-Orbit Tracking in SciBmad

THP5302
21 May 2026, 16:00
2h
C.I.D

C.I.D

Deauville, France
Poster Presentation MC5.D02: Nonlinear Single Particle Dynamics Resonances, Tracking, Higher Order, Dynamic Aperture, Code Developments Poster session

Speaker

Joseph Devlin (Cornell University (CLASSE))

Description

SciBmad is a new software library for accelerator simulation and design. Three key pillars which drove the development of SciBmad’s tracking code are full differentiability (forwards-, backwards-, and Taylor-mode), high performance, and strict symplecticity. Differentiable integrators enable the trivial extraction of lattice functions and their higher-order generalizations, facilitate the use of machine learning tools, and make optimizations more efficient. The integrators have also been heavily optimized for performance, with SIMD parallelization on the CPU and GPU parallelization for various hardwares. Finally, SciBmad uses Yoshida’s symplectic schemes to integrate the fully nonlinear Hamiltonian for magnetic multipoles, RF cavities, and combinations thereof, including implicit symplectic integration where required. All elements allow for spin tracking, which is guaranteed to be orthogonal due to the generality of Yoshida’s scheme. Additionally, synchrotron radiation can be turned on in all elements. In this case, the use of a symplectic scheme guarantees that all damping results from the synchrotron radiation and not from numerical errors due to non-symplectic integration.

In which format do you inted to submit your paper? LaTeX

Author

Joseph Devlin (Cornell University (CLASSE))

Co-authors

David Sagan (Cornell University (CLASSE)) Georg Hoffstaetter (Cornell University (CLASSE)) Matthew Signorelli (Cornell University (CLASSE))

Presentation materials

There are no materials yet.