PPPL Computational Sciences Department

Predictive models for communities of users, enabling high-consequence decision-making

The computational sciences department supports critical Fusion Energy Sciences mission elements, such as whole-device modeling, device and reactor design, and the experimental validation and testing of theoretical models. It is also the laboratory's home for broader computationally-intensive research, such as exascale computing, applied mathematics, algorithm design, data science and learning, and quantum information science.

The M3D-C1, Trinity3D, and TRANSP teams directly support experimental teams associated with NSTX-U, DIII-D, JET, KSTAR, EAST, and ITER. The SIMSOPT and STELLOPT groups are actively engaged in the search for optimal stellarator concepts, taking into account engineering constraints, stability, and collisional and turbulent transport. PPPL participates in DOE's Exascale Computing Project through the XGC and WDMApp projects. The Gkeyll group supports tokamak modeling, mirror configurations (with ARPA-E support), and heliospheric plasma physics. Multiple teams are developing machine learning techniques to accelerate PDE solvers, to predict and ultimately mitigate against dangerous transient phenomena in tokamaks, and to bring greater capability to real-time control systems for fusion and chemical reactors alike.

Through a strong partnership with Princeton University, PPPL scientists and engineers have access to two petaflop computers, Traverse and Stellar, as well as 6000+ cores onsite. PPPL scientists also obtain time on major computer systems at NERSC, Argonne, Oak Ridge, the University of Texas, and more.

The Stellar computing cluster at Princeton University hosts a unique and uniquely powerful whole-device modeling environment for the FES community of scientists and engineers who are working together to interpret experimental observations, prepare new experimental campaigns, and design next-generation magnetic confinement fusion facilities.