Advanced Modeling of Particle Accelerators Using SciDAC Beam Dynamics Codes

PI: R.D. Ryne³, K. Ko⁵, PI (Advanced accelerators): W.B.Mori¹ co-PIs: D.Bruhwiler², E.Esarey³, T.Katsouleas , Compuational science support: V. Decyk¹, Researchers: F.S.Tsung¹, Students: S.Deng⁴, C.K.Huang¹, G. Fubiani³, Affiliates: J.Cary⁶, D.Dimitrov², B.Shadwick³, ISIC/SAPP/other affiliates: P. Colella³, P. McCorquodale³, D. Serafini³ (APDEC)

¹University of California at Los Angeles, ²Tech-X, Boulder Co., ³Lawrence Berkeley National Lab, ⁴University of Southern California, ⁵Stanford Linear Accelerator Center, ⁶University of Colorado at Boulder

Summary

The goal of the SciDAC accelerator modeling project is to develop a new generation of terascale accelerator simulation codes that will allow the accelerator community to address the most challenging problems in 21st century accelerator design and optimization. This paper describes the advanced accelerator piece of the project. This effort involves the development of highly optimized particle-based software modules for modeling plasma-based acceleration. Emphasis is placed on single processor efficiency, parallel scalability, and adding realism to accurately model experiments. The goal is to develop a virtual accelerator capable of accurately modeling 100GeV or greater plasma-based accelerator stages on “terascale” computers.

The future of high-energy physics depends on the development new high-gradient accelerator technology which could reduce the size and cost of future accelerators. Advanced accelerator R and D, relies on experiments which are becoming larger and more complex. SciDAC’s accelerator modeling project is providing advanced accelerator scientists with new simulation tools for terascale computers that are enabling detailed modeling of plasma accelerator experiments in their full-scale for the first time. Furthermore, they are building the foundation for codes that will be able to design and test future experiments in high fidelity in advance of their construction.

The advanced accelerator effort has achieved several milestones. Much of this productivity is due to the SciDAC team approach. Highlights that the SciDAC project has contributed to include:

• A suite of parallelized, full electromagnetic particle-in-cell (PIC) modules and algorithms. These include the codes OSIRIS, OOPIC, Vorpal, and quickPIC. This multi-code effort and the SciDAC team approach make progress much more rapid and reliable. For example the 2D code OOPIC uses very mature ionization algorithms while OSIRIS is a highly optimized 2D and 3D code. As part of the SciDAC project, ionization modules have been added to OSIRIS. By comparing the 2D results against those in OOPIC the routines were debugged significantly faster. Therefore, we will very soon have the capability to more realistically model plasma experiments in 3D with algorithms which have shown 90% parallel efficiency on over 1000 processors at NERSC.
• A modern parallel PIC Framework called UPIC. This is a unified environment containing all the components needed for writing parallel PIC codes. It supports multiple numerical methods and physics approximations and it is designed with many error checks and debugging tools. The 3D code has scaled well on up to 2048 processors using 12 billion particles. Pieces and/or concepts from UPIC are used in the beam dynamics codes, the plasma code quickPIC, and PIC codes in other SciDAC projects.
• The development of the new codes quickPIC and Vorpal. The development of quickPIC is an excellent example of the SciDAC team approach. Based on some reduced physics models it was recognized that a new reduced description PIC code could be constructed by imbedding a 2D electrostatic parallel PIC code inside another 3D parallel PIC code. This was then done rapidly using the UPIC Framework. Furthermore, by benchmarking quickPIC against a non-reduced plasma code (OSIRIS) it was recognized that more physics needed to be added for certain parameters. This work is ongoing. The team approach also benefited Vorpal which developed rapidly using concepts from OOPIC and it includes the ability to replace the particles with fluid elements.
• The suite of codes has been systematically used to model several plasma-based accelerator experiments. Emphasis was given to the E-162 plasma wakefield experiments done at SLAC and the laser wakefield accelerator experiments done at LBNL. The simulation tools were used by a team of scientists on the SciDAC project and by scientists who are not on the project but are involved in the experiments.
• The application of the plasma codes to the E-cloud problem. The need to understand how intense positively charged beams interact with low density electron clouds is one of the most important challenges for existing and future circular machines. The plasma modeling tool quickPIC, that was partially developed by the project,was used by accelerator physicists to model the E-cloud problem. This tool was a tremendous extension over the existing tools. It is completely parallelized, it includes image charges in the conducting walls and it accurately includes the continuous effect of the cloud per turn. Members of the E-cloud community and plasma physicists added the single particle dynamics from the external magnets into quickPIC. These scientists used quickPIC to model over 200 turns (>1000 kms) of the SPS proton ring at CERN. The results are dramatically different from those given by prior models and this new tool offers hope for resolving the discrepancy and experimental results.
• Collaboration with the APDEC ISIC to implement a parallelized Poisson solver for complicated conducting boundaries (P. Collela et al, APDEC). Our E-cloud model uses a square shaped conductor. However, in the FNAL booster the pipe is round. So the Poisson solvers from the APDEC are being added into quickPIC for an elliptical shaped conductor.
• Modeling planned experiments. Using OOPIC it was recognized that for the short-electron bunches to be used in the future E-164 experiment at SLAC that the plasma may be created by the tunnel ionization produced by the self-fields of the beam. This could greatly simplify this and follow on experiments. The codes OSIRIS, quickPIC, and Vorpal are also being used by a host of experimentalists to help design the new experiments.

Our goals for the next year are: to finish developing a 3D ionization capability,to maintain parallel scalability to 3000 processors, to use the codes to model 10 to 100 GeV stages, to model ongoing experiments, to add new Poisson solves into quickPIC, and to improve the iteration loop in the reduced physics model in quickPIC. Our project allocation requests are 2M hrs in FY04 and 5M hrs in FY05.

For further information contact:
Prof. Warren B. Mori, co-Principal Investigator
Phone: 310-206-0372, Mori@physics.ucla.edu
Web site: http://scidac.nersc.gov/accelerator/

back to project page