First effort to couple nonlinear gyrokinetic codes advances fusion whole-device simulations

Building on the increasing importance of first principles–based simulations for understanding turbulent transport of heat, particles, and momentum in fusion devices such as ITER, researchers funded by the Exascale Computing Project have coupled the GENE Eulerian code, which simulates microturbulence at the plasma core, with the PIC code XGC, which simulates large fluctuations at the edge and open field line regions, to deliver the best approach to simulating the processes at play in the whole fusion device. The team also has performed the first simulations of ion temperature gradient–driven and trapped electron–driven turbulence, principal instabilities causing turbulence in toroidal fusion plasmas, using their method. Their ion temperature gradient–driven turbulence work was published in the January 2021 issue of Physics of Plasma.

The researchers aim to develop a high-fidelity Whole Device Model (i.e., WDMApp) of magnetically confined fusion plasmas, a necessary tool to make predictions for planned experiments on ITER and to optimize the design of future next-step fusion facilities. Because these devices will operate in high-fusion-gain physics regimes never achieved by any current or past experiments, successful development of predictive simulations will enable the reliable exploration of new scenarios and new reactor concepts at a significantly reduced cost and over a much broader range of possibilities.

This work extends the capabilities of exascale computing to fusion research and establishes the validity and scalability of the code-coupling approach for whole-device simulations. The team’s approach addresses the risks and challenges associated with coupling codes by ensuring the time-dependent macroscopic paths of the codes’ component equations are consistent with turbulence, using a unified solver with a unified boundary condition, and utilizing ADIOS2, an open-source framework that addresses scientific data management challenges, to control data movement between the GENE Eulerian and XGC codes and minimize data movement costs versus code execution costs.


Merlo, G., S. Janhunen, F. Jenko, A. Bhatterjee, C.S. Chang, J. Cheng, P. Davis, J. Dominski, K. Germaschewski, R. Hager, S. Klasky, S. Parker, and E. Suchyta. “First Coupled GENE-XGC Microturbulence Simulations.” Physics of Plasma 28 (January 2021): 012303.