High Performance Computing

Building upon decades of computation and plasma simulation experience, including the successful gyrokinetic codes GYRO and CGYRO, which have had demonstrated impact on the field of plama physics through hundreds of scientific publications. CGYRO has formed the foundation of numerous HPC resource awards including INCITE, ALCC, XSEDE, and featured in multiple HPCwire articles. [Should link to these articles; some pubs below]

Presently, the Center is developing a stable, flexible and efficient fluid solver for the exascale era of computing. The Anti-symmetric, Large-Moment, Accelerated (ALMA) approach to plasma simulation bridges the gap between kinetic and MHD approximations of burning plasmas with the potential to capture multiscale dynamics through extreme-scale simulations. The inherent numerical stability and GPU support designed into ALMA make rapid scaling straightforward, allowing one to go from testing on a laptop to heroic simulations on a supercomputer without so much as recompiling.

Projects

ALMA

ALMA is General Atomics' next-generation numerical engine, designed from the ground up to deliver high-fidelity simulations of plasmas and neutral fluids at the largest and fastest supercomputers. We have demonstrated ALMA's scalability in DOE's and NSF's leadership-class clusters (Summit, Perlmutter, Frontera, Cori), including simulations of 500 billion fluid points running over thousands of nodes with nearly perfect strong scaling. ALMA is based on a new approach developed at GA that allows both plasma and neutral fluids to be solved rigorously using the same software infrastructure. This unique technology provides us with the flexibility to apply our simulation tools to many scenarios, including tokamak edge turbulence and fueling, liquid metal MHD flows, magnetic reconnection in laboratory and space plasmas, turbulence in inertially confined plasmas, and airfoil design – all using the same scalable, high-fidelity software.

Plasma Physics and MagnetoHydrodynamics
Liquid Metal Behavior in Presence of Magnetic Field

Liquid metal systems are being evaluated for application to magnetic fusion systems. The environment includes many characteristics which make behavior challenging to simulate effectively. These include the complex interplay of geometric arrangements, varying magnetic fields, and interaction with neutrons. Good understanding of the material behaviors is important for developing a robust design. The outcomes of this project will influence the design selections for future fusion system designs. Application to benchmark problems will establish the suitability of the technique to these types of liquid metal problems.

Figure 1: Magnetic Rayleigh-Taylor instability, without (left) and with (right) an external magnetic field (which stabilizes the instability).

Computational Fluid Dynamics
Active Flow for Aeronautic Control

As a compressible implicit large eddy simulator (ILES), ALMA is well suited for turbulent flows in complex geometries. One example of such a system is a fluidic oscillator, where a well-designed static channel geometry results in an oscillating “spray” outflow. Lining the trailing edge of an airfoil with fluidic oscillators enables one to control the boundary layer and associated lift, improving the handling and maneuverability of a craft at low speeds. Development of a wall-model that is uniquely designed for ALMA’s anti-symmetric treatment of the fluid equations provides additional boundary layer fidelity while reducing grid and overall computational requirements.

Figure 2: Fluidic oscillator in a cross-flow.

Computation Fluid Dynamics for Non-Ideal Fluids
Cavitation Inception in Turbulent Liquid Metal

Liquid metal target systems may be used as a neutron source through bombardment with high energy level proton beam. Such systems deliver very high energy into the target system which includes liquid metal core retained by solid metal walls. The dynamics within the target include pressure waves, tension forces, and cavitation among other impacts. The resulting wall erosion reduces the overall lifetime of the target system. Developing rigorous simulations of the target behavior will result in more robust systems which will last longer and raise system utility. Finite element analysis approaches have focused on the liquid metal and have been validated in certain regimes. ALMA is being leveraged to develop such GPU-enabled system to develop a time-accurate flowfield to further understanding of the mechanics of pressure wave generation and initiation of cavitation when subjected to high energy deposition.