Bulletin of the American Physical Society
71st Annual Gaseous Electronics Conference
Volume 63, Number 10
Monday–Friday, November 5–9, 2018; Portland, Oregon
Session BM4: High Performance Computing for Plasma Applications Workshop |
Hide Abstracts |
Chair: Igor Kaganovich, Princeton Plasma Physics Laboratory Room: Oregon Convention Center A105 |
Monday, November 5, 2018 9:30AM - 10:00AM |
BM4.00001: Future of High Performance Particle-in-cell codes computing Johan Carlsson For electrostatic and implicit electromagnetic Particle-in-cell (PIC) codes the global field solve of the Poisson, or implicitly time-discretized Maxwell equations, requires massive communication that limits scalability, especially for three-dimensional simulations. We will discuss best practices, including algorithms and solver libraries. Results from scalability studies will be presented, with an emphasis on electrostatic PIC using the multi-grid method to solve the Poisson equation. Example applications using large-scale simulationswill be presented. An extrapolation to future high-performance computers will be attempted. [Preview Abstract] |
Monday, November 5, 2018 10:00AM - 10:30AM |
BM4.00002: Efficient use of GPUs for Particle-in-Cell codes Peter Messmer The physical accuracy of Particle-in-Cell codes is strongly related to the available computing power. It is therefore no surprise that PIC codes are often among the first to embrace novel hardware architectures like GPU accelerated hybrid systems. With the recently introduced Summit supercomputer at Oak Ridge National Laboratory, largest-scale GPU-accelerated systems have been taken to the next level. And while the fundamental architecture is still unchanged, GPUs have seen tremendous advances over the past decade, both in terms of performance and usability, affecting hardware, software and interconnect. In this presentation, I will review some of the latest architectural changes in the light of core algorithms of PIC codes: How can we optimally deposit charge on latest generation GPUs? How to use multiple GPUs for Poisson solvers? What programming models to use for what parts of the code? etc. In addition, we will also look at hardware features targeting domains like machine learning or computer graphics and how they could be used for kinetic plasma simulations. [Preview Abstract] |
Monday, November 5, 2018 10:30AM - 11:00AM |
BM4.00003: Multi-dimensional fluid simulations of gas discharges Gerjan Hagelaar Numerical simulations based on fluid models are widely used in low-temperature plasma research to gain insight in physical mechanisms and aid the development of gas discharge devices. The computational methodology of these simulations is at present quite mature, and many tools and software products are available to carry them out in 2D or even 3D. However, fluid simulations involve numerous physical approximations which are valid only under certain conditions and limit their accuracy and predictive capabilities. Many of these approximations are specific to low-temperature plasmas, or even to certain discharge types or conditions, and there exists a whole variety of fluid approaches, addressing a variety of discharge conditions. It takes specific expertise to pick the optimal approach and apply it in a meaningful way. In this talk, we present a brief overview of the main fluid modeling approaches with their capabilities and limitations, and we discuss some key approximations (e.g. regarding transport coefficients or plasma sheaths). We also attempt to identify areas where fluid simulations are problematic or not yet mature, so that not even experts know exactly which equations to use, how to efficiently solve them and for what benefit (e.g. magnetized plasma discharges). [Preview Abstract] |
Monday, November 5, 2018 11:00AM - 11:30AM |
BM4.00004: Advanced simulation of arc discharges and their electrodes Mikhail Benilov Plasmas in the bulk of many arc discharges are close to local thermodynamic equilibrium (LTE) and most authors have described arc discharges by means of one-fluid MHD models. While being useful, such models lose their validity in near-electrodes regions, where LTE breaks down. Since appropriate arrangement of current transfer to electrodes is of critical importance, significant efforts have been invested by many workers in order to go beyond LTE MGD models and important advances have been achieved in the course of the last decade. In particular, fully non-equilibrium models have been developed for model 1D problems and 2D low-current arcs. Methods have been developed that employ models of different levels of complexity (assuming quasi-neutrality, or quasi-neutrality and ionization equilibrium, or full LTE) with appropriate boundary conditions at the plasma-electrode interfaces (describing, respectively, the space-charge sheath, or the sheath and the ionization layer, or the sheath, the ionization layer, and the layer of thermal non-equilibrium). Methods of simulation of spots and spot patterns on electrodes have been developed, as well as first-principle models of erosion of cathodes of vacuum and low-pressure arcs. These and other advances are discussed in this talk. [Preview Abstract] |
Monday, November 5, 2018 11:30AM - 12:00PM |
BM4.00005: Atomistic simulations of plasma-surface interaction for ALD and ALE processes Satoshi Hamaguchi, Yuki Okada, Michiro Isobe, Tomoko Ito, Kazuhiro Karahashi Molecular dynamics (MD) simulation and quantum mechanical (QM) first-principles simulation are powerful tools to analyze complex surface reaction mechanisms of atomic layer deposition (ALD) and atomic layer etching (ALE) processes. For example, in ALE of SiO2 films, deposition of a few-angstrom deep fluorocarbon (FC) layer on a SiO2 film and a subsequent application of low-energy Ar$+$ ions to the fluorocarbon-deposited SiO2 film is known to cause sub-mono-layer etching of the SiO2 surface. MD simulation of such a process has shown that low-energy Ar$+$ ion bombardment causes a mixing of the FC layer with the underneath SiO2 surface as well as preferential sputtering of O atoms, resulting in the formation of a relatively Si-rich thin layer incorporating F and C atoms. In other words, Ar$+$ ion bombardment promotes two competing surface reactions: one is the formation of volatile SiFx moieties, which may lead to the desorption of surface Si atoms, and the other is the formation of a SiC network, which can hinder such desorption. For ALE of metal surfaces due to the formation of metal organic complexes, QM simulation can reveal energetically preferred surface reactions. Comparison of the simulation results with experimental observations for such processes will be also discussed. [Preview Abstract] |
Monday, November 5, 2018 12:00PM - 12:30PM |
BM4.00006: Electron-molecule collision calculations for plasma physics applications Jonathan Tennyson Plasma models require data on electron driven process for a wide variety of species. Cool plasmas such as those that occur naturally in the Earth's ionosphere and the interstellar medium or those harnessed for a wide variety of technological processes are largely molecular. These molecular plasma include many transient species whose behavior when colliding with electrons is poorly known and hard to characterize experimentally. My group performs calculations which use the fully quantum-mechanical R-matrix method to perform calculations on a variety of plasma problems including rotational excitation of interstellar molecular ions, excitation of fusion plasma edge species and collisions with molecules important for plasma etching. My talk will feature illustrative examples of such applications including (a) Rotational excitation and dissociative recombination of interstellar molecular ions; (b) Development of a radiative-collisional model for BeH/BeD/BeT, a species whose emission spectra is being actively monitored in fusion plasmas; (c) An electron chemistry of the NFx system which is important in remote plasma sources which are being developed for isotropic etching and thin film deposition in microelectronics fabrication. [Preview Abstract] |
Monday, November 5, 2018 12:30PM - 2:00PM |
BM4.00007: LUNCH BREAK (12:30PM-2:00PM) |
Monday, November 5, 2018 2:00PM - 2:30PM |
BM4.00008: Mullti-scale Methods for Plasma Chemistry Davide Curreli One of the key steps in modeling chemically-reactive plasmas is the construction of a reaction network and the determination of the reaction rates for each reaction participating to the network. This piece of information is in general difficult to construct or obtain, but is a fundamental link between the microscopic physics of atomic/molecular collisions, and the macroscopic transport of plasma species. In this talk we review some of the fundamental computational approaches for the construction of a reaction network and the determination of the corresponding reaction rates. We review different approaches and software for the determination of the thermo-chemistry and plasma-chemistry branches, including both commercial packages at a mature stage of development (eg: ChemKin, Quantemol, etc.) and open-source tools widely available to the community at several stages of development (eg: ZDPlasKin, Crane, etc.). We include in the talk mini-tutorials which can serve as training exercises. Examples include not only well-known systems such as noble gases and O2/N2, but also more complex reaction networks recently tackled, such as the uranium/oxygen system, and their inclusion into macroscopic transport solvers, such as COMSOL and the new MOOSE-based plasma application ZAPDOS-CRANE. [Preview Abstract] |
Monday, November 5, 2018 2:30PM - 3:00PM |
BM4.00009: Update on Code Validation and Verification Miles Turner If computer simulation is an important tool for low-temperature plasma physicists, as most believe, then the correctness of the models and codes that are used must be a matter of concern. Correctness in this context is something in need of demonstration. Modern methodologies for demonstrating code and model correctness are usually called ''Verification and Validation'' or V\&V. Verification refers to the process of testing a code, in other words showing that the code accurately solves the mathematical model that it purports to embody. Validation tests the mathematical model against reality, by comparison with experiments. Clearly, verification must precede validation, and both are (ideally) required before a code can be accepted as an accurate tool. This paper reports on progress in applying the techniques of V\&V in low-temperature plasma physics. The main tools of V\&V will be discussed, together with the outstanding difficulties that remain in applying these methods to the problems of low-temperature plasma physics. [Preview Abstract] |
Monday, November 5, 2018 3:00PM - 3:30PM |
BM4.00010: High-Performance Computer modeling of laser-plasma interactions. Jean-Luc Vay The workhorse algorithm for the modeling of laser-plasma interactions is the Particle-In-Cell (PIC) methodology, where particle beams and plasmas follow a Lagrangian representation with electrically charged macroparticles while electromagnetic fields follow a Eulerian representation on (usually Cartesian) grids. The complexity of the phenomena that often involve large ranges of space and time scales have driven the development of codes that run on the largest available supercomputers, and has also driven the development of novel algorithms. We will review the latest advances with examples of applications and discuss the challenges and efforts in preparation of upcoming exascale supercomputers. [Preview Abstract] |
Monday, November 5, 2018 3:30PM - 4:00PM |
BM4.00011: PIC Simulation of Magnetic Fusion Plasmas on High Performance Computers C-S Chang, W. Lee, Z. Lin, Scott Parker, W. Tang, w. Wang This talk will overview the state-of-the-art HPC-oriented PIC techniques, which are used in the magnetic fusion plasma research and which could be useful in the application plasma research. Magnetic fusion plasma is governed by multi-scale multi-physics phenomena that span millions of order in configuration space and billions of order in time-space, with many of the multi-physics being non-separable in space and time scales. The discussion topics will include the concept of the magnetic field-line following meshing, the turbulent and non-turbulent physics, the delta-f and total-f techniques, the whole-device modeling, the multiscale self-organization phenomenon, the quite start, the collision operators, the 5-D gyrokinetic and 6D full kinetic simulations, the addition of velocity-space grid to PIC mesh, the neutral particle recycling and Monte Carlo transport, the large scale parallelization and programing models, the optimization on CPU and GPU type accelerators, the load balancing, the in-memory in-transit data managements, the employment of machine learning technologies, and others. [Preview Abstract] |
Monday, November 5, 2018 4:00PM - 4:30PM |
BM4.00012: Bridging the scales of simulations for plasma material interaction Longtao Han, Igor Kaganovich, Predrag Krstic It is important to understand how plasma properties influence the surface processes of materials in different levels, which may explain the synthesis, phase change or damage of materials. This talk will give an overview on simulation approaches that are being used in the modeling of plasma material interaction. Density functional theory methods (molecular DFT, plane wave DFT), quantum classical molecular dynamics methods (e.g., DFTB, DC-DFTB), classical molecular dynamics (e.g., LAMMPS) and kinetic Monte Carlo methods will be covered with examples in recent studies. [Preview Abstract] |
Monday, November 5, 2018 4:30PM - 5:00PM |
BM4.00013: Discrete Velocity Methods and Adaptive Kinetic-Fluid Models Vladimir Kolobov The particle transport in plasma can be described by either kinetic or hydrodynamic (fluid) models. Kinetic models provide detailed description in terms of particle velocity distributions, which obey the Boltzmann, Vlasov, and Fokker-Planck kinetic equations. Direct numerical solution of the kinetic equations can be obtained using discrete velocity methods. The fluid models describe plasma on a macroscopic level and provide coupling between the particle transport and electromagnetics. In this tutorial, we will describe the kinetic solvers based on discrete velocity methods and hybrid kinetic-fluid models combining the accuracy of kinetic solvers with the efficiency of fluid models. Adaptive kinetic-fluid models are particularly important for plasma, which is characterized by a wide range of temporal and spatial scales due to the disparity of electron mass and the masses of heavy species (ions, atoms). The research challenges in this field are associated with identifying correct criteria for selecting appropriate models (which are different for electrons, ions, atoms, and photons), closure of fluid models for collisional and collisionless plasmas, coupling kinetic and fluid solvers at interfaces (which can dynamically evolve), and implementing these hybrid algorithms into smart software for practical plasma engineering on modern computing systems. [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