A differentiable simulation package for performing inference of synchrotron-radiation-based diagnostics.

Area of Science:

• Particle Accelerator Physics
• Computational Physics
• Optics

Background:

• Particle accelerator development demands higher beam quality, currents, and repetition rates.
• Traditional beam diagnostics face destruction challenges with intense beams.
• Non-invasive diagnostics using synchrotron radiation are complex and computationally intensive.

Purpose of the Study:

• Introduce SYRIPY, a Python software package for synchrotron radiation-based diagnostics.
• Enable efficient and accurate inference of particle beam parameters.
• Provide a differentiable tool leveraging GPU acceleration for statistical inference.

Main Methods:

• Developed SYRIPY using PyTorch for differentiability and GPU acceleration.
• Integrated modules for particle tracking, Lienard-Wiechert field calculation, and Fourier optics propagation.
• Performed start-to-end simulations of synchrotron radiation detection.

Main Results:

• SYRIPY demonstrated good agreement with the established SRW package.
• Achieved up to a 50x speed improvement compared to existing methods.
• Successfully applied SYRIPY for Bayesian inference of beam parameters using stochastic variational inference.

Conclusions:

• SYRIPY offers a powerful and efficient tool for synchrotron radiation diagnostics.
• The software facilitates advanced beam parameter inference in particle accelerators.
• SYRIPY's performance and features advance non-invasive diagnostic capabilities.