Challenges in Self-Consistent Full-Wave Simulations of Lower Hybrid Waves

• Published in: IEEE transactions on plasma science 2010-09, Vol.38 (9), p.2136-2143
• Main Authors: Wright, John C, Jungpyo Lee, Valeo, Ernest, Bonoli, Paul, Phillips, Cynthia K, Jaeger, E F, Harvey, Robert W
• Format: Article
• Language: English
• Subjects: Algorithms; Approximation; Computation; Computational modeling; Computer architecture; Computer simulation; Conferences; Diffraction; Distributed computing; Electron distribution; Electrons; Frequencies; Mathematical analysis; Plasma; Plasma simulation; Plasma waves; Propagation; Radio frequency; Ray tracing; Simulation; vector wave equation; Wave propagation; waves; X-ray measurements
• ISSN: 0093-3813; 1939-9375
• DOI: 10.1109/TPS.2010.2055167

Analysis of wave propagation in the lower hybrid range of frequencies (LHRF) in the past was done using ray tracing and the Wentzel-Kramers-Brillouin approximation taking advantage of the very small scale of those waves. To include the effects of wave diffraction and focusing in this regime, full-wave simulation is necessary but requires significantly more computational power. In both ray tracing and full-wave simulations in the LHRF, it is also essential to include the self-consistent evolution of the electron distribution in response to the waves. This adds a considerable computational burden in constructing the stiffness matrix for the system [Valeo , "Full-wave Simulations of LH wave propagation in toroidal plasma with non-Maxwellian electron distributions," 18th Topical Conference on Radio Frequency Power in Plasmas, AIP Conference Proceedings (2007)]. Advances in algorithms and the availability of massively parallel computer architectures have permitted the solving of the Maxwell-Vlasov system for wave propagation directly [Wright , Phys. Plasmas (2009), 16, July]. We will discuss the various modeling advances that have led to this capability, including various memory-management approaches, physics-motivated algorithm adaptions appropriate to the LHRF, and improvements in the matrix solver to minimize communication overhead when using thousands of cores on leadership-class computer platforms. Of particular importance have been the use of verification and validation techniques and the analytic approximations to the imaginary (pole residue) contribution to the plasma dielectric response.