First-principles-based whole device modeling (WDM) of fusion devices is one of the great frontiers of computational plasma physics. Ultimately, success will enable the confident exploration of new ITER scenarios and new reactor concepts in silico, at greatly reduced cost and over a much wider range of possibilities. The challenges associated with connecting simulations of physical processes occurring at vastly different time and space scales to one another are immense and will require significant investments. In particular, success requires further intensive development of the most complex physics modules: the gyrokinetic turbulence kernels. Due in part to significant past SciDAC support, core tokamak ‘anomalous transport’ is no longer an intractable mystery. A validated predictive capability now exists for a range of conditions. However, important problems remain unsolved.

For example: The narrow transport barrier of the H-mode pedestal doubles tokamak energy confinement, but we cannot predict the conditions required to achieve it, its energy losses, or its scope for optimization. Internal transport barriers have proven to be equally challenging. Fluctuations coexisting on ion and electron gyroradius scales yield to heroic simulations today, but remain inaccessible on transport time scales and have not been explored in important parameter regimes. Electromagnetic fluctuations (including microtearing) can be simulated today, but not in the WDM context. Notably, each of these multiscale challenges is associated with high performance operating regimes that magnetic confinement fusion research seeks to exploit in the quest for fusion energy.


In this project we will take on these challenges by applying high-fidelity gyrokinetic simulations to these ‘frontier’ transport problems. In the process, we will achieve profound scientific breakthroughs while also bringing these challenging problems within the scope of WDM. Ultimately our team will deliver to the broader WDM effort advanced 5D gyrokinetic modules for HPC platforms, designed from the outset to meet three goals. First, these inherently multi-scale/multifidelity turbulence modules will be validated against experimental data in a continuing program of focused experimental research. Second, our modules and algorithms will incorporate recent and inspire new advances from the applied mathematics and computer science communities, to guarantee the resilience, extreme scalability, and efficiency required to fit the stringent requirements of WDM. The third goal is essential: we will confront ‘non-asymptotic’ non-local problems that often lie outside the conventional local approximation—and thus necessarily, we will experiment with and develop high-risk, multiscale, multifidelity coupling algorithms that should not be the responsibility of the main WDM framework team.



The Partnership for Multiscale Gyrokinetic Turbulence 

  • David Hatch (Overall Principal Investigator), Frank Jenko, (Max Planck Institute for Plasma Physics & University of Texas-Austin), Mike Kotschenreuther, Craig Michoski (University of Texas-Austin)
  • Greg W. Hammett (Principal Investigator, Princeton Plasma Physics Laboratory), Ammar Hakim (Princeton Plasma Physics Laboratory)
  • Bill Dorland (Principal Investigator, University of Maryland)
  • Jeff Parker (Principal Investigator), Lynda LoDestro (Lawrence Livermore National Laboratory)
  • Darin R. Ernst (Principal Investigator, Massachusetts Institute of Technology), Qingjiang Pan, Manaure Francisquez (Massachusetts Institute of Technology)
  • Antoine Cerfon (NYU Courant Institute)
  • Dan Reynolds (SMU—FASTMATH)
  • Carol Woodward (SMU—FASTMATH)
  • Shan Hongzhang (LBNL—RAPIDS)
  • Lenny Oliker (LBNL—RAPIDS)