SLAC's LCLS-II is rapidly advancing towards MHz repetition rate attosecond X-ray pulses, opening new opportunities to leverage the abundance of data in combination with advances in machine learning (ML) to better align the x-ray source with specific experimental goals. We approach the challenge from both ends of the facility. Starting at the X-ray output, we showcase our low latency, high throughput ML algorithms implemented at-the-edge for X-ray detection and reconstruction in the Multi-Resolution 'Cookiebox' (MRCO) angle resolved electron spectrometer with its 16 electron time-of-flight detectors. MRCO performs spectro-temporal characterization of X-ray profiles with a resolution that allows single shot identification of well-seeded shots versus SASE background at MHz rate. MRCO enables fast feedback, so we also tackle the problem as a control issue, focusing on programmable photoinjector laser shaping to adjust the initial electron bunch. Towards this end of using advances in ML to explore the parameter space for optimizing X-ray production, we present our progress towards a digital twin linking the photoinjector laser all the way through MRCO in the endstation diagnostics.

Linear particle accelerators are elaborate machines that demand a thorough comprehension of their beam physics interactions to enhance performance. Traditionally, physics simulations model the physics interactions inside a machine but they are computationally intensive. A novel solution to the long runtimes of physics simulations is replacing the intensive computations with a machine learning model that predicts the results instead of simulating them. Simple neural networks take milliseconds to compute the results. The ability to make physics predictions in almost real time opens a world of online models that can predict diagnostics which typically are destructive to the beam when measured. This research entailed the incorporation of an innovative simulation infrastructure for the SLAC FACET-II group, aimed at optimizing existing physics simulations through advanced algorithms. The new infrastructure saves the simulation data at each step in optimization and then improves the input parameters to achieve a more desired result. The data generated by the simulation was then used to create a machine learning model to predict the parameters generated in the simulation. The machine learning model was a simple feedforward neural network and showed success in accurately predicting parameters such as beam emittance and bunch length from varied inputs.

Coherent synchrotron radiation (CSR) in linear accelerators (linacs) is detrimental to applications that require highly compressed beams, such as FELs and wakefield accelerators. However, traditional measurement techniques lack the precision to fully comprehend the intricate multi-dimensional aspects of CSR, particularly the varying rotation of transverse phase space slices along the longitudinal coordinate of the bunch. This study explores the effectiveness of our generative-model-based high-dimensional phase space reconstruction method in characterizing CSR effects at the Argonne Wakefield Accelerator Facility (AWA). We demonstrate that the reconstruction algorithm can successfully reconstruct beams that are affected by CSR.

SLAC's LCLS-II is rapidly advancing towards MHz repetition rate attosecond X-ray pulses, opening new opportunities to leverage the abundance of data in combination with advances in machine learning (ML) to better align the x-ray source with specific experimental goals. We approach the challenge from both ends of the facility. Starting at the X-ray output, we showcase our low latency, high throughput ML algorithms implemented at-the-edge for X-ray detection and reconstruction in the Multi-Resolution 'Cookiebox' (MRCO) angle resolved electron spectrometer with its 16 electron time-of-flight detectors. MRCO performs spectro-temporal characterization of X-ray profiles with a resolution that allows single shot identification of well-seeded shots versus SASE background at MHz rate. MRCO enables fast feedback, so we also tackle the problem as a control issue, focusing on programmable photoinjector laser shaping to adjust the initial electron bunch. Towards this end of using advances in ML to explore the parameter space for optimizing X-ray production, we present our progress towards a digital twin linking the photoinjector laser all the way through MRCO in the endstation diagnostics.

Phase space reconstruction using Neural-Networks and differentiable simulation* is a robust beam diagnostic method to obtain complete 4D phase space including the coupling terms such as (x-y’) and (y-x’). In the first experimental demonstration, it was verified that the RMS beam envelope and normalized emittance from the reconstructed phase space are quantitatively similar to those from the conventional beam diagnostics such as quadrupole scan. In addition, here we show the demonstration of the phase space for the i) flat and ii) magnetized beam where the beam has i) very large ratio in between horizontal and vertical emittances (e.g., enx/eny >>1) and ii) transverse coupling induced by non-zero solenoid magnetic field at the cathode (known-as canonical momentum-dominated beam). Through the demonstrations using the experimental data achieved at the Argonne Wakefield Accelerator Facility (AWA), we successfully obtained the information such that the measured flat beam indeed has the emittance ratio larger than 70 with minimized transverse coupling. In addition, we were able to obtain the magnetization from the reconstructed phase space. Moreover, we will compare the beam parameters obtained from the phase space reconstruction and conventional diagnostics and discuss the uncertainty of the parameters.

Linear particle accelerators are elaborate machines that demand a thorough comprehension of their beam physics interactions to enhance performance. Traditionally, physics simulations model the physics interactions inside a machine but they are computationally intensive. A novel solution to the long runtimes of physics simulations is replacing the intensive computations with a machine learning model that predicts the results instead of simulating them. Simple neural networks take milliseconds to compute the results. The ability to make physics predictions in almost real time opens a world of online models that can predict diagnostics which typically are destructive to the beam when measured. This research entailed the incorporation of an innovative simulation infrastructure for the SLAC FACET-II group, aimed at optimizing existing physics simulations through advanced algorithms. The new infrastructure saves the simulation data at each step in optimization and then improves the input parameters to achieve a more desired result. The data generated by the simulation was then used to create a machine learning model to predict the parameters generated in the simulation. The machine learning model was a simple feedforward neural network and showed success in accurately predicting parameters such as beam emittance and bunch length from varied inputs.

The RHIC heavy ion program relies on a series of RF bunch merge gymnastics to combine individual source pulses into bunches of suitable intensity. Intensity and emittance preservation during these gymnastics require careful setup of the voltages and phases of RF cavities operating at several different harmonic numbers. The optimum setting tends to drift over time, degrading performance and requiring operator attention to correct. We describe a reinforcement learning approach to learning and maintaining an optimum configuration, accounting for the relevant RF parameters and external perturbations (e.g., a changing main dipole field) using a physics-based simulator at Brookhaven Alternating Gradient Synchrotron (AGS).

Advances in high-performance computing have enabled detailed physics simulations, including those with nonlinear collective effects such as space charge, to be deployed online in a control room setting to aid operator intuition and be used directly in automatic tuning. Simultaneously, machine learning (ML) has enabled deployment of detailed models online with sub-second execution time, opened up new avenues for adapting simulation models to more closely match real accelerator behavior, and enabled novel ways to combine detailed physics simulations and ML-based tuning. This contribution will provide an overview of how these tools are being developed and successfully applied at SLAC, with an emphasis on experimental demonstrations. This includes improvements in adaptive calibration methods, novel approaches to simulation (e.g. differentiable physics combined with ML), and the use of system models in ML-based tuning (e.g. Bayesian optimization with system model priors, iterative simulation and ML tuning to aid LCLS-II injector commissioning). Discussion of the software infrastructure required to achieve this and deploy these solutions into regular operation will also be discussed.

Next-generation accelerator concepts, which hinge on the precise shaping of beam distributions, demand precise diagnostic methods capable of reconstructing beam distributions with 6-D phase spaces. However, the characterization of 6-D beam distributions using conventional techniques necessitates hundreds of measurements, using hours of valuable beam time. Novel diagnostic techniques are needed to reduce the number of measurements required to reconstruct detailed, high dimensional beam features for precision beam shaping applications. In this study, we present a novel approach to analyzing experimental measurements using generative machine learning models of 6-D beam distributions and differentiable beam dynamics simulations. We demonstrate in simulation that using our analysis technique, conventional beam manipulations and diagnostics can be used to reconstruct detailed 6-D phase spaces using as few as 20 beam measurements with no prior training or data collection. These developments enable detailed, high dimensional phase space information to be obtained for precision control and improved understanding of complex accelerator beam dynamics.

Coherent synchrotron radiation (CSR) in linear accelerators (linacs) is detrimental to applications that require highly compressed beams, such as FELs and wakefield accelerators. However, traditional measurement techniques lack the precision to fully comprehend the intricate multi-dimensional aspects of CSR, particularly the varying rotation of transverse phase space slices along the longitudinal coordinate of the bunch. This study explores the effectiveness of our generative-model-based high-dimensional phase space reconstruction method in characterizing CSR effects at the Argonne Wakefield Accelerator Facility (AWA). We demonstrate that the reconstruction algorithm can successfully reconstruct beams that are affected by CSR.

The recent development of advanced black box optimization algorithms has promised order of magnitude improvements in optimization speed when solving accelerator physics problems. These algorithms have been implemented in the python package Xopt, which has been used to solve online and offline accelerator optimization problems at a wide number of facilities, including at SLAC, Argonne, BNL, DESY, ESRF, and others. In this work, we describe updates to the Xopt framework that expand its capabilities and improves optimization performance in solving online optimization problems. We also discuss how Xopt has been incorporated into the Badger graphical user interface that allows easy access to these advanced control algorithms in the accelerator control room. Finally, we describe how to integrate machine learning based surrogate models provided by the LUME-model package into online optimization via Xopt.

The LCLS-II is a high repetition rate upgrade to the Linac Coherent Light Source (LCLS). The emittance and dark current are both critical parameters to optimize for ideal system performance. Here we summarize the role these tools played in the commissioning period and are playing in the current operational stage of the LCLS-II injector, which provides an example of how other accelerator facilities may benefit from combining online modeling and optimization infrastructure. We also describe current progress on creating a fully deployed digital twin of the LCLS-II injector based on a combination of ML modeling and physics modeling, using the LUME software suite and various ML-based characterization tools. Finally, we will describe current efforts and plans to leverage the online LCLS-II injector model in fast optimization and control schemes.