University of California, Irvine, physicist Zhihong Lin is at the forefront of fusion simulation, using Oak Ridge National Laboratory's (ORNL's) “Jaguar” Cray XT3 system to simulate microturbulence in magnetically confined thermonuclear plasmas.
Lin is a member of the “Center for Gyrokinetic Particle Simulation of Turbulent Transport in Burning Plasmas” (GPSC), a fusion energy SciDAC (Scientific Discovery through Advanced Computing) project led by W. W. Lee of the Princeton Plasma Physics Laboratory (PPPL). Lin is the original developer of the Gyrokinetic Toroidal Code (GTC) [See “ORNL's Jaguar Fastest for Fusion Turbulence Simulations,” http://www.hpcwire.com/hpc/943289.html], which is the flagship code at UC–Irvine and PPPL for studying plasma microturbulence in advanced fusion systems.
He was recently named a fellow of the American Physical Society (APS), which recognized him “for fundamental contributions to the understanding of zonal flows and turbulence spreading and to the pioneering development of massively parallel gyrokinetic particle simulations on modern leadership class supercomputers.”
In this interview, Lin talks about his work, the importance of simulation to fusion research, and the people who influenced his career.
HPCwire: Congratulations on being named an APS fellow. Please describe briefly the work that led to this honor.
Lin: Thank you. My primary area of research interest is studying plasma turbulence and transport via large-scale particle simulations using massively parallel supercomputers. Turbulent transport is one of the most important scientific challenges in burning plasma experiments such as the International Thermonuclear Experimental Reactor (ITER), which is the crucial next step in the quest for fusion energy. Research advances in this application area can also be expected to help accelerate progress in understanding fundamental problems in space and astrophysical plasmas.
HPCwire: What is your relationship with other members of the ORNL and PPPL team?
Lin: The success of the GTC project is a result of the productive collaborations between many talented SciDAC GPSC researchers. I have close collaborations with GPS team members such as Dr. Scott Klasky at ORNL and Dr. Stephane Ethier at PPPL. I am currently the GTC team leader for the SciDAC GPSC project.
HPCwire: How did you get started using supercomputers in your research?
Lin: My interest in parallel computing started in graduate school and was inspired by the vision of my mentor at Princeton University, Prof. Bill Tang. Near the end of my PhD thesis project, I spent 2 months during the summer of 1994 at the Advanced Computing Laboratory at LANL [Los Alamos National Laboratory] to learn how to use a parallel computer — CM-5, made by Thinking Machines Corporation — with a language called CMFortran. During the last week of my stay at LANL, Thinking Machines Corporation went bankrupt. Nonetheless, I did learn the basics of parallel programming.
HPCwire: How has the evolution of supercomputers affected fusion research over the years?
Lin: Addressing the dynamical processes in fusion plasmas characterized by a large number of degrees of freedom and disparate spatial-temporal scales, fusion simulation is a grand-challenge application that pushes the frontier of high-performance computing. The 3D particle-in-cell code for fusion simulations has necessarily evolved in a continuous way in order to efficiently utilize the most powerful computers available. Along with the evolution of computing hardware, parallelization, optimization, visualization, statistical analysis, and data management have become more sophisticated and specialized. Consequently, cross-disciplinary, multi-institutional collaborations have become a clearly visible trend in fusion simulations, which often involve computational and analytical plasma physicists in alliance with applied mathematicians and computational and computer scientists. Therefore, it is crucial to carefully plan the coordination and collaboration across the boundaries of disciplines and institutions.
Other fusion SciDAC projects have also benefited significantly from the interdisciplinary alliances enabled by the SciDAC program and by access to increasingly powerful computational platforms. Examples of such projects include the “Simulation of Wave Interactions with Magnetohydrodynamics” led by Dr. Don Batchelor of ORNL and the “Center for Extended Magnetohydrodynamic Modeling” led by Dr. Steve Jardin of PPPL. However, I am not familiar with the associated codes.
HPCwire: How has it affected your own career?
Lin: My own career certainly benefits from the opportunities offered by advanced simulations. Collaborations within and outside the SciDAC framework provide complementary expertise in computational and computer sciences. Access to the state-of-the-art computational resources enables simulations of new parameter regimes of interest to the theoretical and experimental communities. As a new tool for discovery, advanced simulation opens up an exciting frontier of research for new players, which complements the traditional analytic theory populated by well-established scientists. Finally, the training in computational physics is of increasingly high demand in research and academic institutions.
HPCwire: Who are your role models in physics?
Lin: I have personally benefited from the wisdom of the late Prof. Marshall Rosenbluth, who was a prominent pioneer of plasma physics and fusion energy science. In an interdisciplinary sense, Prof. Rosenbluth was perhaps best known for inventing the Monte-Carlo method, with his seminal 1953 paper having been cited nearly 10,000 times.
HPCwire: You are one of the biggest users of the Cray XT3 system at ORNL's National Center for Computational Sciences (NCCS). Could you please describe your experience there?
Lin: My experience with NCCS has been very positive. After some initial hiccups following a steep curve of hardware upgrades on the Cray XT3, a number of GTC simulations have now been able to run smoothly on 6,400 cores for several days, thanks to support by the account and operation staff, who are very accessible and responsive. The mechanism of a project contact person (Scott Klasky for the GPSC SciDAC project) provides an effective channel for technical communication between the science application team and the computational support team. Finally, our project benefits greatly from the expertise of the computational science team at NCCS in extracting physics out of a large amount of simulation data, in particular, through collaborations with Dr. Klasky on visualization, data management, and work flow, and with Dr. George Ostrouchov on advanced statistical analysis.
HPCwire: What new science and breakthroughs can you point to from the use of simulation in fusion science?
Lin: Every order of magnitude increase in computing power, along with progress in physics models and numerical algorithms, enables qualitatively new science to be addressed via advanced simulations. Specific examples since the mid-'90s include 3D global simulation, the simulation of zonal flow physics, transport-size scaling, and the simulation of electron turbulence.
HPCwire: Your work is geared toward support of the ITER experimental fusion reactor. How has this work helped ITER?
Lin: ITER is the second largest international science collaboration, with a budget of $13 billion funded by the United States, European Union, Japan, China, India, Korea, and Russia. ORNL currently houses the U.S. ITER Project Office. Achieving the design goal of ignition in ITER depends on a good confinement of the burning plasmas. Current studies in plasma turbulence aim at significantly improved understanding of the confinement properties in parameter regimes relevant to ITER. The fusion SciDAC projects will also lay the groundwork for an integrated fusion simulation project targeting predictive simulations of ITER experiments when ITER becomes operational in the middle of the next decade.
HPCwire: How will supercomputers fit in with ITER in the coming years?
Lin: Theory and simulation can be expected to play a greater role in the ITER project as we explore the new regime of burning plasma physics. First-principles simulations could directly address the parameter regimes inaccessible by conventional experimental and analytic theory techniques. A specific example is the turbulence driven by the fusion products (a-particles), which are expected to provide the main heating source in ITER. Such turbulence can lead to the loss of these energetic particles before the energy is deposited to the plasmas. Physics-based predictive simulation tools can enhance the efficiency of the utilization and optimize operation schemes for ITER. ITER simulations focus on the extension of current simulation capabilities into the burning plasma regime and on the development of integrated simulation capability for addressing simultaneously multiple scales and multiple processes in fusion plasmas. A central theme is predictive simulation integrating both modeling and physics simulation codes for the purpose of benchmark validations on existing tokamak experiments. The ultimate goal here is to provide a reliable, comprehensive simulation package for ITER plasmas. These new capabilities will certainly require much more powerful computing capabilities than are currently available.