20th Advanced Accelerator Concepts Workshop

6 Nov 2022, 15:00 → 11 Nov 2022, 17:10 America/New_York

Hyatt Regency Long Island

Mark Palmer (Brookhaven National Laboratory) , Navid Vafaei-Najafabadi (Stony Brook University)

The AAC’22 workshop is the 20th in a series of by-invitation biennial fora for intensive discussions on long-term research in advanced accelerator physics and technology. This research supports the development of capabilities for the basic sciences, from photon science to high energy physics, as well as the development of compact accelerators for industrial, medical and security applications.

AAC'22 will be organized into eight working groups covering the following topical areas:

1. Laser-Plasma Wakefield Acceleration 
2. Computation for Accelerator Physics
3. Laser and High-Gradient Structure-Based Acceleration
4. Beam-Driven Acceleration
5. Beam Sources, Monitoring, and Control
6. Laser-Plasma Acceleration of Ions
7. Radiation Generation and Advanced Concepts
8. Advanced Laser and Beam Technology and Facilities

Conference Home:  https://www.aac2022.org/

 

Sunday, 6 November

18:00 → 19:30 Welcome Reception

Monday, 7 November

10:20 → 10:40 Coffee Break

12:10 → 13:30 Lunch

15:00 → 15:30 Coffee Break

Tuesday, 8 November

10:00 → 10:30 Coffee Break

12:00 → 13:20 Lunch

15:00 → 15:30 Coffee Break/Exhibits

Wednesday, 9 November

10:00 → 10:30 Coffee Break/Exhibits

12:00 → 13:20 Lunch

13:30 → 15:00 WGs 2+5 Joint Session: Session 1 of 1

Joint session between working groups 2 & 5:
WG2 - Computation for Accelerator Physics
WG5 - Beam Sources, Monitoring and Control

Conveners: Alexey Arefiev (UC San Diego) , David Bruhwiler (RadiaSoft LLC) , Dr Samuel Barber (Lawrence Berkeley National Laboratory) , Yine Sun (Argonne National Laboratory)

13:30 A Non-Invasive, Single-Shot Emittance Diagnostic, Using Edge-Radiation and a CNN Based Inference Model

Accelerators are moving towards higher repetition rates with extremely high current and brightness beams.
Advanced control techniques using machine learning are required for the optimisation and operation of such accelerators.
These techniques greatly benefit from having single-shot beam measurements.
However, high intensity beams present an issue for conventional diagnostics as they will destroy any material placed in the beamline.
Therefore, single-shot, non-invasive diagnostics are highly desirable.

Edge-radiation is produced in the fringe fields of bending magnets and is common to both linear and circular accelerators.
The radiation is both sensitive to the beam parameters and non-intercepting, making it an ideal candidate for future diagnostics.
Here, we will present experimental results of an edge-radiation based diagnostic at FACET-II.
This diagnostic collects edge-radiation and uses machine learning to characterize the beam.
We will discuss the image processing techniques used and demonstrate how a convolutional neural network can predicts the emittance in a single-shot.

Speaker: Robbie Watt (SLAC)

13:45 Experimental results for machine learning based diagnostics and optimization at FACET-II

This contribution addresses recent progress in machine learning based diagnostics and optimization at the FACET-II facility at SLAC National Accelerator Laboratory. We focus the discussion around three examples: longitudinal phase space diagnostics [1-2], new algorithms for 20x speedup in optimization of beam emittance, and automated sextupole tuning to reduce minimum spot sizes in the FACET experimental area. We also present results describing the application of machine learning based characterization methods to efficiently explore high-dimensional parameter spaces [3], as compared to traditional parameter scans, in the context of emittance optimization of the FACET-II photoinjector. We will show results obtained during the last experimental run and discuss plans for deployment of these tools in future runs, along with their potential to aid in beam setup for experimental configurations, reduce tuning time, and improve beam stability. While we focus on applications at FACET-II, most of these tools can be readily used at other facilities through our open-source software packages (e.g. see [4,5]); we will briefly highlight the current capabilities of these software packages.

[1] C. Emma and A. Edelen, M. J. Hogan, B. O’Shea, G. White, and V. Yakimenko , Phys. Rev. Accel. Beams 21, 112802 (2018)
[2] C. Emma, A. Edelen, A. Hanuka, B. O’Shea, A. Scheinker, Information 2021, 12(2), 61; https://doi.org/10.3390/info12020061
[3] Ryan Roussel, Juan Pablo Gonzalez-Aguilera, Young-Kee Kim, Eric Wisniewski, Wanming Liu, Philippe Piot, John Power, Adi Hanuka & Auralee Edelen Nature Communications volume 12, Article number: 5612 (2021)
[4] C. Mayes, R. Roussel, H. Slepicka, Xopt. 10.5281/zenodo.6991160
[5] C. E. Mayes and P. H. Fuoss and J. R. Garrahan and H. Slepicka and A. Halavanau and J. Krzywinski and A. L. Edelen and F. Ji, W. Lou and N. R. Neveu and A. Huebl and R. Lehe and L. Gupta and C. M. Gulliford and D. C. Sagan and J. C. E and C. Fortmann-Grote, IPAC’21, THPAB217.

Speaker: Claudio Emma

14:00 Detailed Phase Space Reconstructions from Accelerator Beam Measurements Using Differentiable Simulations

Characterizing the phase space distribution of particle beams in accelerators is a central part of accelerator understanding and performance optimization. However, conventional reconstruction-based techniques either use simplifying assumptions or require specialized diagnostics to infer high-dimensional (> 2D) beam properties. In this work, we introduce a general-purpose algorithm that combines neural networks with differentiable particle tracking to efficiently reconstruct high-dimensional phase space distributions without using specialized beam diagnostics or beam manipulations. We demonstrate that our algorithm reconstructs detailed 4D phase space distributions with corresponding confidence intervals in both simulation and experiment using a single quadrupole and diagnostic screen. This technique allows for the measurement of multiple correlated phase spaces simultaneously, enabling simplified 6D phase space reconstruction diagnostics in the future.

Speaker: Ryan Roussel (SLAC National Accelerator Laboratory)

phase_space_reconstruction_AAC22.pdf

14:15 ADJOINT OPTIMIZATION OF CIRCULAR LATTICES

Design of circular lattices involves optimizing figures of merit (FoMs) characterizing the beam properties subject to the constraint that the beam distribution function be approximately periodic in trips around the lattice. We are developing an algorithm that accomplishes this with minimal computational effort. The algorithm takes advantage of recent developments in adjoint techniques that allow the derivatives of the FoM with respect to the many parameters describing the lattice to be evaluated. The present description of the accelerator is based on the 10 second moments of the beam distribution function in the transverse phase space. However, extensions to kinetic descriptions will be discussed. Our algorithm, which we name “Adjoint with a Chaser”, works as three separate minimizations run concurrently. These three working together force the beam into a periodic state, while varying parameters to minimize an FoM. Examples relevant to the Maryland lattice UMER will be presented.

Speaker: Thomas Antonsen

Antonsen_AAC_V1b.pdf

14:30 Simulating electron beams in RF cavities with beam loading

High-brightness electron photoinjectors and electron linacs are fundamental to many advanced accelerator concepts and associated applications (see e.g., Ref. [1] and references therein). The industrial, medical and homeland security markets for low-to-moderate energy electron linacs are growing rapidly. To meet the design challenges for these divergent applications, with modest software development resources, a simulation code must meet the following requirements: a reduced-model algorithm that includes beam loading in rf cavities; phase space conserving algorithms; a single-source implementation that executes efficiently on many CPUs, on one or more GPUs, and on heterogeneous supercomputing architectures; as well as easy benchmarking with high-fidelity community codes.
We present recent work with the open source Hellweg code [2-4], which is routinely used to design TW electron linacs, showing 1000x speedup as compared to CST Particle Studio. We plan to refactor Hellweg’s C++ source code to make effective use of the AMReX framework [5,6], joining an ecosystem of massively parallel accelerator physics codes under development at Berkeley Lab. The reduced model algorithms in Hellweg will play an important role, in concert with other more high-fidelity PIC codes.
We will describe the underlying algorithms in Hellweg, as well as recent and ongoing generalizations. Recent work on traveling wave linac simulation and design will be presented. Time will also be devoted to a discussion of future plans, which include treatment of Touschek scattering, thermionic and photocathode electron guns, and a modified algorithm to conserve phase space.

[1] F. Stephan et al. “High Brightness Photo Injectors for Brilliant Light Sources,” Synchrotron Light Sources and Free-Electron Lasers (2020). Ed. by E. Jaeschke et al.
[2] S. V. Kutsaev et al. “Generalized 3D beam dynamics model for industrial traveling wave linacs design and simulations,” NIM A 906 (2018), p. 127.
[3] Y. Eidelman et al. “Ellipsoid space charge model for electron beam dynamics simulations,” Phys. Part. Nucl. 52 (2021), p. 477.
[4] The open source Hellweg repository, https://github.com/radiasoft/rslinac
[5] W. Zhang et al. “AMReX: a framework for block-structured adaptive mesh refinement,” Journal of Open Source Software 4 (2019), p. 1370.
[6] The open source AMReX repository, https://github.com/AMReX-Codes/amrex

Speaker: Dr David Bruhwiler (RadiaSoft LLC)

20221109_Bruhwiler_Hellweg_AAC_v2.pdf

14:45 Discussion

Thursday, 10 November

10:00 → 10:30 Coffee Break/Exhibits

12:00 → 13:20 Lunch

15:00 → 18:00 Afternoon at Leisure

Friday, 11 November

10:00 → 10:30 Coffee Break

12:00 → 13:00 Lunch

14:40 → 15:00 Coffee Break