Skip to content

description.md

Latest commit

 

History

History
20 lines (14 loc) · 3.77 KB

description.md

File metadata and controls

20 lines (14 loc) · 3.77 KB
Raw
widget blank
weight 20
headless true
design
columns
1

The Gysela-X project consists in the development of a series of gyrokinetic codes with the main aim to allow exascale simulations of tokamak plasma turbulence and transport including a X-point geometry.

The legacy Fortran code GYSELA (for GYrokinetic SEmi-LAgrangian)1 has been developed at CEA/IRFM for over 20 years through national and international collaborations with a strong interaction between physicists, mathematicians and computer scientists. It is a global full-f nonlinear gyrokinetic code that simulates electrostatic plasma turbulence and transport in the core of Tokamak devices. It evolves the complete 5-dimensional (3 space coordinates, 2 velocity coordinates) guiding-centre distribution function of the plasma's ion and electron species in the electrostatic limit, over a full torus. The time-evolution of this distribution function $f(r,\theta,\phi,v_\parallel,\mu)$ is governed by the gyro-averaged Vlasov equation, the so-called gyrokinetic equation, which is self-consistently coupled to the 3D quasi-neutrality equation (electrostatic and long wavelength limit of Maxwell's equations). Versatile sources of heat, momentum, particles and vorticity are also commonly used to achieve a steady-state in long simulation runs.

However, upgrading such a long-lived code to run efficiently on massive nmubers of GPUs while also extending the code's physics capacities has proven to be an impossible task. Therefore, we ahve been extracting key kernels from the Fortran code and developing new ones from scratch in ordr to create an open-source library written in modern C++ and using modern libraries including Kokkos and PDI. The development of this library, called Gyselalib++, started during EoCoE-III. It contains the elementary kernels required for (gyro-)kinetic codes based on a semi-Lagrangian algorithm to evolve the distribution functions. This shared library reveals extremely efficient by easing the implementation in the two flagship codes of numerical optimizations tested in simplified configurations. These two flagship codes are Gysela-X++ to simulate tokamak plasma turbulence and transport in the presence of a X-point, and Gysela-stellarator to account for 3D magnetic equilibria (stellarators, tokamaks with Resonant Magnetic Perturbations or ripple).

The main goal of Gysela-X++ is to simulate an ITER plasma including the relevant physics, namely X-point, neutrals, realistic sources and electromagnetic effects. The operational regime of ITER plasmas is expected to depart from the one of current devices, especially regarding their distance to marginality. There, recent publications suggest that flux-driven simulations could be mandatory to predict the confinement time reliably. However, this challenging task remains so far out of reach of current first-principles simulation tools because of the extreme resolution required. The usage of exascale resources will allow Gysela-X++ to generate high fidelity predictions for ITER. In addition, a database generated from its outputs to allow the development of lower fidelity tools in preparation of ITER exploitation.

So far, flux-driven gyrokinetic simulations in 3D magnetic equilibria (stellarator, tokamak with RMP or ripple) do not exist. Within EoCoE-IV, we will start the development of the Gysela-stellarator code which aims at simulating plasma transport in such equilibria.

Footnotes

  1. [Virginie Grandgirard, Jérémie Abiteboul, Julien Bigot, Thomas Cartier-Michaud, Nicolas Crouseilles, et al.. A 5D gyrokinetic full-f global semi-lagrangian code for flux-driven ion turbulence simulations. 2015.]({{< ref "/publication/grandgirard-cpc-2016/index.md" >}}) ↩