High-Performance Computing High-performance computational models of stellarators, left and right, with an engineering image in between Pause video ￭ Mission Statement The Computational Sciences Department supports world-class plasma physics research and energy applications and develops novel computational methods with applications across the whole of science and engineering today. We pursue Fusion Energy Science's goal of establishing the scientific knowledge to develop a source of safe, clean and abundant fusion energy and advancing our understanding of matter at very high temperatures and densities.We also support broad mission elements of the U.S. Department of Energy (DOE), ranging from modeling and predicting complex processes important to the department’s Advanced Scientific Computing Research (ASCR) program to advancing entirely new ways to generate energy through high-impact technology Advanced Research Projects Agency-Energy (ARPA-E) programs that have the potential to radically improve U.S. economic prosperity.Computational simulations facilitate the design of complex systems. Employing powerful supercomputers PPPL’s computational physics staff makes use of powerful supercomputers on the Princeton University campus and the extraordinary powers of national supercomputers for complex, mission-critical computations.Stellar, a computing cluster at the University’s High-Performance Computing Research Center, allows our computational physicists to test and fine-tune computer codes before running them on national supercomputers, where time is limited. These codes are used to model experiments aimed at developing fusion as a safe, clean and plentiful method of generating electricity and predicting disruptions that can halt fusion reactions and damage fusion devices. PPPL codes are used throughout the fusion world.Another computing cluster at Princeton University, called Traverse, is equipped with graphics processing units (GPUs) that facilitate artificial intelligence for codes that use GPUs. The computational physics community at PPPL thus uses the most powerful and sophisticated supercomputers to enable our mission of creating energy from fusion for a virtually inexhaustible power supply for all humankind. Cutting the ribbon for the Traverse supercomputer, from left: Craig Ferguson, PPPL deputy director for operations and chief operating officer; Steve Cowley, PPPL director; David McComas, Princeton University’s vice president for PPPL; Chelle Reno, Princeton University’s assistant vice president for operations for PPPL; and Jay Dominick, Princeton University’s vice president for information technology and chief information officer. (Photo by Denise Applewhite/Princeton Office of Communications) Research We support critical fusion computational developments ranging from the modeling of whole-device fusion plasmas to the experimental testing and validation of theoretical models and fusion reactor design. We are also the PPPL home for broader and intensive computational research, including exascale computing, applied mathematics, algorithm design and quantum materials and devices.High-performance PPPL codes support experiments on the National Spherical Torus Experiment-Upgrade (NSTX-U), the Laboratory’s fusion facility, and tokamaks worldwide. The team operating a widely used stellarator code is actively engaged in the search for optimal stellarator concepts, while other teams are developing machine-learning techniques to predict and mitigate dangerous tokamak disruptions and applying codes to the study of plasma from the microscopic to the astrophysical scale. Overall, advanced computation is at the frontier of fusion energy development and plasma science. Whole-device modeling Computational scientists aim to create highly complex computer simulations of the plasma that fuels fusion reactions in doughnut-shaped fusion facilities called tokamaks and twisty fusion facilities called stellarators. Unlike current simulations, which model only part of the hot, charged plasma gas, the new simulations will display the physics of entire plasmas all at once. The modeling will enable physicists to understand a plasma fully, allowing them to predict its behavior and design more efficient fusion reactors.To complete this task, the scientists will rely on exascale computers, now coming online, that can perform a billion billion (1018) operations per second. Codes designed to run on these computers will be used to understand everything from the shape of an entire plasma to how the gas absorbs and releases energy. Related links PPPL Computational Sciences Department website High Performance Computing
