Bulletin of the American Physical Society
59th Annual Meeting of the APS Division of Plasma Physics
Volume 62, Number 12
Monday–Friday, October 23–27, 2017; Milwaukee, Wisconsin
Session PO6: Computation |
Hide Abstracts |
Chair: Frank Tsung, University of California, Los Angeles Room: 202C |
Wednesday, October 25, 2017 2:00PM - 2:12PM |
PO6.00001: Development of high performance particle in cell code for the exascale age Giovanni Lapenta, Jorge Amaya, Diego Gonzalez Magnetized plasmas are most effectively described by magneto-hydrodynamics, MHD, a fluid theory based on describing some fields defined in space: electromagnetic fields, density, velocity and temperature of the plasma. However, microphysics processes need kinetic theory, where statistical distributions of particles are governed by the Boltzmann equation. While fluid models are based on the ordinary space and time, kinetic models require a six dimensional space, called phase space, besides time. The two methods are not separated but rather interact to determine the system evolution. Arriving at a single self-consistent model is the goal of our research. We present a new approach developed with the goal of extending the reach of kinetic models to the fluid scales. Kinetic models are a higher order description and all fluid effects are included in them. However, the cost in terms of computing power is much higher and it has been so far prohibitively expensive to treat space weather events fully kinetically. We have now designed a new method capable of reducing that cost by several orders of magnitude making it possible for kinetic models to study macroscopic systems. [1] Lapenta et al, JCP, 334 (2017): 349-366. JPP, 83.2 (2017). [Preview Abstract] |
Wednesday, October 25, 2017 2:12PM - 2:24PM |
PO6.00002: Algorithm implementation and testing to ensure consistency of Gauss’s law in OSIRIS Kyle Miller, Paul Elias, Ricardo Fonseca, Benjamin Winjum, Frank Tsung, Viktor Decyk, Warren Mori Electromagnetic particle-in-cell (PIC) simulations compute the trajectories of particles as they interact via fields calculated by numerically solving Maxwell's equations on a grid using currents (and charge densities) from the particles. Within PICKSC, UCLA maintains a variety of open-source and open-access codes. These include OSIRIS---developed in partnership with IST---and UPIC-EMMA. Standard OSIRIS uses a rigorous charge-conserving current deposit to ensure the consistency of Gauss's law together with a finite-difference (FD) solution to Maxwell's equations. It also contains options for spectral (FFT) and hybrid (FFT and FD) field solvers, as well as a customized, higher-order FD field solver to help mitigate the numerical Cerenkov instability. The standard charge conserving current deposit is only valid for second-order accurate FD solvers. Another option for maintaining the consistency of Gauss's law is the Boris correction, where a ``direct'' current deposit is used and the electric field is corrected through the use of a Poisson solve. The Boris correction---with both exact and iterative multigrid Poisson solves---has been implemented into OSIRIS. Preliminary analyses of timing and fluctuations levels will be presented, including the effects of different particle orders. [Preview Abstract] |
Wednesday, October 25, 2017 2:24PM - 2:36PM |
PO6.00003: Implementation of collisions on GPU architecture in the Vorpal code Jarrod Leddy, Sergey Averkin, Ben Cowan, Scott Sides, Greg Werner, John Cary The Vorpal code contains a variety of collision operators allowing for the simulation of plasmas containing multiple charge species interacting with neutrals, background gas, and EM fields. These existing algorithms have been improved and reimplemented to take advantage of the massive parallelization allowed by GPU architecture. The use of GPUs is most effective when algorithms are single-instruction multiple-data, so particle collisions are an ideal candidate for this parallelization technique due to their nature as a series of independent processes with the same underlying operation. This refactoring required data memory reorganization and careful consideration of device/host data allocation to minimize memory access and data communication per operation. Successful implementation has resulted in an order of magnitude increase in simulation speed for a test-case involving multiple binary collisions using the null collision method. [Preview Abstract] |
Wednesday, October 25, 2017 2:36PM - 2:48PM |
PO6.00004: The utility of continuum simulations for direct current and microwave microplasmas Venkattraman Ayyaswamy, Arghavan Alamatsaz, Abhishek Kumer Verma State-of-the-art microplasma devices have contributed to several challenges that require a fundamental understanding of the various mechanisms involved in order to achieve optimal operation for a given application. In this context, the role of computations cannot be stressed enough. Historically, the computational techniques used for simulating plasmas belong to two categories – continuum/fluid and kinetic methods. The primary goal of the current work is to perform an exhaustive comparison of continuum and kinetic simulations for a range of operating conditions. Kinetic simulations using the particle-in-cell with Monte Carlo collisions (PIC-MCC) method and continuum simulations using the full-momentum equation are performed at various operating conditions. It is shown that using the electron energy distribution function (EEDF) predicted by BOLSIG+ in continuum simulations of direct current microplasmas leads to a significant under-prediction of plasma densities. The discrepancy between kinetic and continuum simulations is attributed to the presence of hot electrons created as a result of secondary emission. On the other hand, continuum simulations performed for a microwave microplasma operating at 0.5 GHz showed excellent agreement with kinetic simulations. [Preview Abstract] |
(Author Not Attending)
|
PO6.00005: A Lagrangian code for simulating compressible large deformation multi-material MHD processes Bo Xiao, Hai-bo Zhao, Jin-song Bai, Ganghua Wang A 2-dimensional Lagrangian code, named TriAngels-MHD, is made for the simulation of compressible large deformation multi-material MHD processes. The Lagrangian scheme is built on pure triangular mesh. The MHD simulation is split into a hydrodynamic step (ideal MHD step) and a magnetic diffusion step. For the hydrodynamic step, to conquer the mesh distortion problem in a Lagrangian scheme, a dynamic local remeshing algorithm is designed. And to mitigate the checkerboard oscillation problem that is typical for a triangular mesh based Lagrangian simulation, a matter flow term is introduced for each grid edge, which compensates for the non-bending of a grid edge. For the magnetic diffusion step, the Joule heat is calculated based on a formula of $\frac{\partial e_J}{\partial t}=\nabla\cdot(\frac{\eta}{\mu_0}\vec{B}\times(\nabla\times\vec{B}))-\frac{\partial}{\partial t}(\frac{1}{2\mu_0}B^2)$. This scheme insures the equality of the total Joule heat production and the total electromagnetic energy loss in the system. Typical simulations are carried out to test the performance of the Lagrangian code, including the pure hydrodynamic processes of Noh explosion test problem and triple-point problem and the MHD process of magneto-Rayleigh-Taylor instability evolution. [Preview Abstract] |
Wednesday, October 25, 2017 3:00PM - 3:12PM |
PO6.00006: Block Preconditioning to Enable Physics-Compatible Implicit Multifluid Plasma Simulations Edward Phillips, John Shadid, Eric Cyr, Sean Miller Multifluid plasma simulations involve large systems of partial differential equations in which many time-scales ranging over many orders of magnitude arise. Since the fastest of these time-scales may set a restrictively small time-step limit for explicit methods, the use of implicit or implicit-explicit time integrators can be more tractable for obtaining dynamics at time-scales of interest. Furthermore, to enforce properties such as charge conservation and divergence-free magnetic field, mixed discretizations~using volume,~nodal, edge-based, and face-based degrees of freedom are often employed~in some form. Together with the presence of stiff modes due to integrating over fast time-scales, the mixed discretization makes the required linear solves for implicit methods particularly difficult for black box and monolithic solvers. This work presents a block preconditioning strategy for multifluid plasma systems that segregates the linear system based on discretization type and approximates off-diagonal coupling in block diagonal Schur complement operators. By employing multilevel methods for the block diagonal subsolves, this strategy yields algorithmic and parallel scalability which we demonstrate on a range of problems. [Preview Abstract] |
Wednesday, October 25, 2017 3:12PM - 3:24PM |
PO6.00007: Second Energy Variation for Heterogeneous Systems with Electrostatic and Magnetostatic Interaction Michael Greenfield, Pavel Greenfield Systems carrying electric charges and magnetic dipoles are widespread in nature and applications. Variational principles and methods play key role in modelling these systems, and they have been developed for a couple of centuries (see, Landau, L.D. and Lifshitz, E.M. Electrodynamics of Continuous Media, Pergamon, 1963). Nonetheless, even a quick glance at the literature shows that the variational approach remains unfinished. In particular, in terms of calculus of variations, all the studies are based on the analysis of first energy variations (i.e., analysis of ponderomotive forces and conditions of equilibrium.) The analysis of the second variation, the cornerstone of the stability analysis, is not even touched in a systematic manner. We partially fixed this gap in (Grinfeld, M., Grinfeld, P., A Variational Approach to Electrostatics of Polarizable Heterogeneous Substances, Advances in Mathematical Physics, Article ID 659127, 2015). In this paper, we present corresponding results relating to the problems of plasma physics. [Preview Abstract] |
Wednesday, October 25, 2017 3:24PM - 3:36PM |
PO6.00008: Bayesian equilibrium inference in the Minerva framework Jakob Svensson, Oliver Ford, Sehyun Kwak, Lynton Appel, Kian Rahbarnia, Joachim Geiger, Jonathan Schilling The Minerva framework is a scientific modelling system based on Bayesian forward modelling and is used at a number of experiments. The structure of the framework makes it possible to combine flux function based, axisymmetric or full 3D models. A general modularity approach makes it easy to replace underlying physics models, such as the model for force balance and corresponding current distribution. We will give an overview of the different models within Minerva for inference of equilibrium field and flux surfaces, for both tokamaks and stellarators. For axisymmetric devices, three methods of increasing complexity, Gaussian process based Current Tomography (CT), an iterative Grad-Shafranov solver, and a full nonlinear Grad-Shafranov based model, will be demonstrated for the JET device. The novel nonlinear Grad-Sharanov model defines a proper posterior distribution for the equilibrium problem thus defines the space of possible equilibrium solutions, and has the capacity to include any nonlinear constraints (e.g. from models of profile diagnostics). The Bayesian approach further allows uncertainties on the equilibrium parameters to be calculated. For the W7-X stellarator, two models based on the VMEC 3D solver and a fast function parameterization approximation will be demonstrated. [Preview Abstract] |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700