Abstract Details
Abstracts
Author: Darin R. Ernst
Requested Type: Poster
Submitted: 2022-03-04 16:05:35
Co-authors: M. Francisquez (PPPL), D. R. Reynolds (SMU), C. J. Balos (LLNL), C. S. Woodward (LLNL)
Contact Info:
Massachusetts Institute of Technology
MIT NW16-132, 167 Albany St.
Cambridge, MA 02139
US
Abstract Text:
To study cross-scale interactions in coupled toroidal ITG/ETG turbulence and provide a practical test-bed for multi-rate and multi-scale algorithms to speed up simulations, we formulated and implemented a reduced 2D toroidal multi-scale gyrofluid model, retaining full FLR effects via Bessel functions, using a modified field equation with appropriate electron response at ion and electron scales. These equations are implemented in a pseudo-spectral code with adaptive, additive multi-rate Runge-Kutta (RK) time integration using the ARKODE [1] library, providing more robust time integration than a naive CFL-constrained RK stepper. Toroidal ITG and ETG mode linear growth rates are close to 2D GENE gyrokinetic results for Cyclone-like parameters. Nonlinear simulations using the model closely match GENE toroidal ITG nonlinear heat fluxes over a wide range of temperature gradients, while matching the nonlinear toroidal ITG critical temperature gradient. As expected, results are sensitive to the choice of closures. In the parameter regime where the 2D approximation is justified, the strength of zonal flows relative to turbulence is much greater in 2D than in 3D. Using the reduced model, multi-scale toroidal ITG/ETG simulations were completed in 50 to 100 times less CPU time than 3D gyrokinetic simulations typically require, with resources characteristic of single scale 3D ITG gyrokinetic turbulence simulations. We have formulated a multi-rate approach which could potentially lead to an additional two orders of magnitude speedup, with adaptive time-stepping for specified accuracy. Successful multi-rate methods, after implementation, verification, and testing in 2D will be transferred to GENE.
[1] SUNDIALS Team, https://computing.llnl.gov/projects/sundials, accessed 3/3/2022.
*Work supported by U.S. DOE OFES Subaward No. UTA18-000276 under DE-SC0018429, DE-AC02-09CH11466, DE-SC0021354; and OASCR SciDAC FASTMath Institute at LLNL under contract DE-AC52-07NA27344.
Comments:
I will be attending TTF, so scheduling Monday would guarantee avoiding conflicts.