Minisymposium

Future developments towards magnetic confinement fusion energy depend on the understanding of turbulent transport in the edge and scrape-off layer (SOL). Gyrokinetic simulations are among the main tools used to improve our understanding of turbulence in the edge and SOL. Such global, high fidelity simulations provide an accurate description of the relevant physics, however are also highly expensive in terms of computational costs, even when simulating small machines. Approaching reactor relevant, large-scale machines not only increases the computational size drastically, but also imposes new challenges in terms of physics. In this talk we present an overview of the recent progress with the gyrokinetic turbulence code GENE-X, focusing on physical and computational challenges. GENE-X is an Eulerian-type "continuum" code that solves the collisional, full-f, electromagnetic, gyrokinetic Vlasov-Maxwell system on a grid. It is especially targeted towards edge and SOL simulations including the magnetic X-point, using the flux-coordinate independent (FCI) approach. To bridge the gap between simulations of small experiments and future reactors, we pursue two approaches. First, the code is ported to GPUs and, second, we implement an optimized spectral velocity space discretization. These improvements allow the GENE-X code to perform simulations of turbulence in large-scale fusion devices.

We report progress on the development and application of a hybrid gyrokinetic ion–fluid electron model for simulations of cross-separatrix plasma transport in the edge region of a divertor tokamak. The hybrid model includes ion-scale ion temperature gradient (ITG) and resistive drift and ballooning modes, as well as neoclassical ion physics and orbit loss effects. The model is implemented in the finite-volume gyrokinetic code COGENT, which employs a locally field-aligned coordinate system combined with mapped multi-block grid technology to handle strongly anisotropic edge plasma turbulence. Additionally, COGENT utilizes a flexible implicit-explicit (IMEX) time integration framework that can handle the temporal multiscale nature of plasma transport. To that end, the 5D high-dimensional kinetic part of the system that involves ion parallel transients and diamagnetic drift time scales is treated explicitly. The 3D low-dimensional fluid part of the system that contains fast electron physics and Alfven-wave time scales is treated implicitly. The performance of the IMEX scheme is optimized by developing specialized preconditioners for the electrostatic and electromagnetic versions of the hybrid model. Verification results and simulations in realistic single-null geometries are presented.

*Work performed for USDOE, at LLNL under contract DE-AC52-07NA27344.

In this work we describe a fully kinetic PIC MC (Particle in Cell Monte Carlo) code, BIT1, used for the simulation of plasma, neutral, and impurity particle transport in the tokamak edge [2]. BIT1 has several unique features enabling near ab-initio modeling of nonlinear processes in the plasma edge: it can resolve the smallest time and space scales and still simulate large systems, such as the tokamak Scrape-off Layer (SOL) and can track very slow and fast particles with high accuracy, Vmax/Vmin > 1000 and so on. As an example, we present several recent results obtained from BIT1 simulation of the plasma edge of existing and future tokamaks.To address the main drawback of such modeling, which has extremely long simulation times, such as the ITER SOL BIT1 simulation requiring up to 10 million CPU hours, equivalent to half a year of simulation time under realistic conditions, we initiate the development of an exascale version of the BIT1 under the Horizon project Plasma-PEPSC [3]. Furthermore, we outline the BIT1 exascale porting process, which includes algorithmic restructuring, GPU porting, and implementation of highly scalable parallel I/O using openPMD. [1] https://www.iter.org/ [2] D. Tskhakaya, Plasma Phys. Control. Fusion, 59, (2017)[3] https://plasma-pepsc.eu/

The magnetised plasma sheath is a boundary layer, whose thickness is a few ion Larmor radii, appearing next to the solid targets of a fusion device. It is a multi-scale region, including the Debye length and electron gyroradius as additional smaller length scales. It is therefore rich in physics and complex, in particular in the fusion-relevant case of grazing magnetic field incidence at the target. From the point of view of the plasma, the main role of the sheath is to reflect electrons such that a (globally) ambipolar flow of ions and electrons to the solid target is achieved. Therefore, simplified conducting or logical sheath boundary conditions that exploit a constant parallel electron velocity reflection cutoff, related to an unresolved sheath potential drop, have been and are being used in gyrokinetic simulations. Yet, only actual sheath solutions can provide more accurate reflection conditions for electrons, as well as predict the correct plasma-surface interaction by computing the strongly distorted ion distribution function at the target. We present a scheme to iteratively obtain fast numerical solutions of the steady state magnetised sheath for grazing magnetic field incidence. A magnetic moment dependence in the electron reflection cutoff provides more accurate sheath boundary conditions.