Jefferson Lab's small business partner, RadiaSoft LLC, has been working collaboratively with the lab since it was founded in 2013 on a number of advanced accelerator projects, including their most recent work with the lab on machine learning.
Partnership with small business yields big advances for accelerators
NEWPORT NEWS, VA — In accelerator physics, simulating physical phenomena with specialized computer code is the norm. Those codes, however, are not always easily accessible nor do they reach their full potential. RadiaSoft LLC, one of Jefferson Lab’s small business partners, is working with the lab to change that.
Founded in 2013, RadiaSoft LLC is a Colorado-based science consulting company. While it has been informally collaborating with Jefferson Lab since its incorporation, in 2016 a formal partnership was established to explore the topic of magnetized dynamic friction for relativistic colliders. This work led to fundamentally new results for the understanding and quantitative calculation of the magnetized dynamic friction force for the electron-ion collider parameter regime, with RadiaSoft scientists showing significant qualitative and quantitative differences with previously published results. Using the open-source JLab Simulation Package for Electron Cooling (JSPEC) code developed at JLab, RadiaSoft ultimately developed the Sirepo/JSPEC graphical user interface for numerical simulations on the electron cooling process.
In 2017, RadiaSoft LLC started work on a project using ray-tracing simulation code Zgoubi on spin dynamics for polarized electrons and ions in collider designs. A widely used collider-simulation code, Zgoubi has broad capabilities, but is difficult to learn. RadiaSoft developed a graphical interface to improve ease of use and hosted an educational Zgoubi workshop in 2019. JLab’s Fanglei Lin and Vasily Morozov were both involved in developing a tutorial on electron-ion spin dynamics that studied designs under consideration for a collider.
Jefferson Lab continues to work with RadiaSoft LLC as a Small Business Innovation Research (SBIR) partner on new and exciting projects like these. The relationship between the lab using these accelerator codes and the business outfitting them with graphical interfaces that provide interactive visual components for computer software is mutually beneficial. During the partnership, RadiaSoft significantly improved its technology for containerizing accelerator simulation codes with Docker, specifically with regard to reproducible computing. This work will enable particle accelerator design teams to archive simulation results for advanced accelerators in order to reproducibly repeat and extend simulations during future project commissioning or upgrades.
"Accelerators like Jefferson Lab’s CEBAF and the future electron-ion collider provide a powerful engine for scientific discovery, and a driver for advanced manufacturing and new applications in the United States. We are proud to be supporting the development of relevant accelerator science and technology through the DOE Office of Nuclear Physics SBIR program,” says RadiaSoft LLC President and CEO David Bruhwiler.
Most recently, the lab and RadiaSoft are collaborating on developing advanced machine learning algorithms for particle accelerator control systems. When completed, these algorithms will power a software toolbox that allows users to rapidly build web-based control interfaces for data analysis, automated tuning, and advanced fault detection in particle accelerators. “This type of collaborative work with small business,” says Chief Technology Officer Drew Weisenberger, “is a key to technology development that benefits our science programs and beyond.”
Contact: Deborah Dowd, Jefferson Lab Communications Office, 757-269-7180, dowd@jlab.org