Advanced Modeling of Electroenergetic Devices

Primary Investigators: J. P. Verboncoeur and A. J. Christlieb
Funding Agency: AFOSR
Funding Amount: $1.05M

Electro-energetic radiation sources from the microwave to terahertz frequencies comprise one of the core technological advantages of the US warfighter, with applications ranging from communications to radar to electronic warfare. As fabrication costs have escalated, the research community increasingly relies on computational tools to develop and design the next generation of sources. These problems require a kinetic model to address with high fidelity. A comprehensive, state-of-the-art, particle-in-cell modeling suite is proposed. Based on the existing AFOSR-funded code, OOPIC, and its one-dimensional counterpart, OOPD1, a three dimensional extension employing object oriented technology will be developed. The object technology allows for encapsulation of models, enabling team and remote development as well as a code lifetime beyond the dwell time of the developers. As with its predecessors, the new code will be distributed at the source level through an open source license consistent with the DOD Open Technology Development Roadmap Plan. The new code will also include algorithm advances and high performance computing capabilities which enable application to large real world problems, with scalability from single processor workstations, to modest cost clusters, to large scale parallel installations. The regime of validity will be expanded with fluid algorithms for high density plasmas, Monte Carlo collision models, surface physics, and flexible boundary conditions including cut-cell conformal boundaries. In addition, the code suite will be enhanced to include dynamic control of the simulation employing the Python language to choose algorithms in real time, as well as create and implement new diagnostics and generate plots at run time. The Python front end will also facilitate the control of massively parallel runs, and optimization and parameter searches employing user-designed algorithmic feedback. New processing capabilities, including GPU coprocessors and Knights Corner, will be used to develop adaptive high performance algorithms capable of portioning the workload dynamically across available hardware components. The code will be benchmarked against existing 3D PIC codes including ICEPIC, Magic-3D, and VORPAL, as well as available experiments in academia and government labs. The result of this research will be an advanced, portable, scalable particle simulation code suite serving the present and future needs of the research community with a free source-level distribution which can be modified and expanded by advanced developers, while remaining accessible to the casual user. A significant side effect of the research program is the training of the next generation of plasma simulation researchers, and advancement of the state of the art through publication.

The work plan includes the following elements:

  • Object-oriented architecture for OOPIC 3D
  • Python front end for dynamic control and specification of models and diagnostics
  • Self-optimizing algorithms for gather/scatter optimized by compute component and memory bandwidth
  • GPU/Knights Corner engine for optimized PIC computation of 3D particle dynamics
  • 3D electrostatic solver and boundary conditions
  • 3D electromagnetic solver and boundary conditions
  • Monte Carlo collision model
  • Cut-cell boundary capability
  • Noise control models
  • Hybrid particle-fluid model with collisions
  • Online distribution