Whenever the SLAC Public Gas Pedal Research Facility’s direct gas pedal is on, bunches of around a billion electrons each move together at almost the speed of light through metal funneling. The accelerator’s particle beam, made up of these bunches of electrons, is used to study the atomic behavior of molecules, novel materials, and many other topics.
However, it is difficult to predict how a particle beam will behave during an experiment, leaving scientists with only a rough idea of what it will look like as it moves through an accelerator.
An algorithm that better predicts a beam’s distribution of particle positions and velocities as it zips through an accelerator has now been developed by researchers at the University of Chicago, the Department of Energy’s Argonne National Laboratory, and the SLAC National Accelerator Laboratory.
“Our algorithm uses information about a beam that is typically ignored in order to create a more accurate picture of the beam. Although we have many different ways to manipulate particle beams inside of accelerators, we don’t have a very precise way to describe a beam’s shape and momentum.”
SLAC accelerator scientist and lead author Ryan Roussel .
As accelerator facilities operate at ever-increasing energies and generate increasingly complex beam profiles, this in-depth beam information will help scientists carry out their experiments more reliably. In Physical Review Letters, the researchers described their algorithm and method in detail.
“We have a variety of ways of controlling molecules radiating within gas pedals; however, we don’t have a truly exact method for depicting a pillar’s shape and energy,” SLAC gas pedal researcher and lead creator Ryan Roussel said. “Our calculation considers data about a bar that is typically disposed of and utilizes that data to illustrate the pillar.”
Typically, researchers describe the positions and speeds of particles in a beam using a few summary statistics that give an idea of the beam’s overall shape. However, this method leaves out a lot of information that could be useful. On the other hand, bar researchers can take numerous estimations of the actual pillar and attempt to reproduce, in some cases utilizing AI, what the shaft would resemble under various exploratory conditions; however, those techniques require a ton of information and a great deal of computational power.
The team tried a new approach for this study: they developed a machine learning model that makes use of our comprehension of beam dynamics to predict the phase space distribution, or the distribution of positions and speeds of particles within the beam.
The team used their model to interpret experimental data from the Argonne Wakefield Accelerator at the DOE’s Argonne National Laboratory in order to put their theories to the test. Counting the physical science of molecule pillar elements with the trial information permitted the specialists to precisely recreate the fine subtleties of the bar utilizing just 10 data points of interest — an errand that could take up to 10,000 data points of interest for some AI models that do exclude a model of shaft physical science.
According to SLAC accelerator scientist and co-author Auralee Edelen, “most machine learning models don’t directly include any notion of particle beam dynamics to speed up learning and reduce the amount of data required. We have demonstrated that even astonishingly small amounts of data can be used to infer extremely intricate high-dimensional beam shapes.”
Currently, the algorithm is able to reconstruct a beam model along its left-right and up-down axes, as if the group of particles were a pancake moving down the accelerator path. 4D beam phase space is the name given to this kind of reconstruction. The next step is for the researchers to try out the algorithm in action by reconstructing complete 6D phase space distributions, including the positions and speeds of particles along the beam’s path.
“In general, the calculation is a significant change in outlook in the manner in which we break down exploratory gas pedal information at offices today,” Roussel said.
He stated, “We can now use particle beam data in a more comprehensive and potent way to improve our scientific goals at accelerators everywhere.”
More information: R. Roussel et al, Phase Space Reconstruction from Accelerator Beam Measurements Using Neural Networks and Differentiable Simulations, Physical Review Letters (2023). DOI: 10.1103/PhysRevLett.130.145001
