APS Logo

GX: a GPU-native gyrokinetic turbulence code for tokamaks and stellarators

POSTER

Abstract

The GX code is a GPU-native radially-local delta-f gyrokinetic turbulence code that uses pseudo-spectral methods in both configuration (Fourier) and velocity (Hermite-Laguerre) space. At high resolution GX is a standard gyrokinetic code, but it can also be successfully run at low resolution (particularly in velocity space) in lieu of uncontrolled approximations, since in the lowest velocity-resolution limit the system corresponds to established gyrofluid models. We demonstrate the robust convergence properties of the model via nonlinear benchmarks against established gyrokinetic codes in both tokamak and stellarator geometries. Coupled with efficient use of GPU computing architectures, GX produces a performance advantage over other gyrokinetic approaches in many cases. GX has been coupled to the Trinity transport solver, which uses a multi-scale approach to solve for time-dependent radial profiles of density, temperature, etc. Turbulent fluxes are obtained from GX calculations, neoclassical fluxes are obtained from a drift kinetic solver, and external sources and edge boundary conditions are supplied by the user. Using the Trinity+GX system, we demonstrate the ability to solve for the time-dependent evolution of core fusion reactor profiles without resorting to reduced models.

Presenters

  • Noah R Mandell

    MIT, Massachusetts Institute of Technology MIT, Massachusetts Institute of Technology

Authors

  • Noah R Mandell

    MIT, Massachusetts Institute of Technology MIT, Massachusetts Institute of Technology

  • William D Dorland

    University of Maryland Department of Physics, University of Maryland, College Park

  • Ian G Abel

    University of Maryland, College Park, IREAP, University of Maryland, College Park

  • Nathaniel Barbour

    University of Maryland, College Park

  • Braden Buck

    Princeton Plasma Physics Laboratory, University of Wisconsin, Madison, Purdue University

  • Rahul Gaur

    University of Maryland, College Park

  • Patrick S Kim

    University of Maryland, College Park

  • Tony Qian

    Princeton Plasma Physics Laboratory, Princeton University