Eliminating turbulent self-interaction through the parallel boundary condition in local gyrokinetic simulations

TL;DR

This paper identifies and addresses the issue of turbulent self-interaction in local gyrokinetic simulations caused by the parallel boundary condition, demonstrating how domain extension can eliminate this artifact and improve accuracy.

Contribution

The study reveals the impact of turbulent self-interaction on simulation accuracy and proposes extending the simulation domain to eliminate this effect, enhancing the fidelity of gyrokinetic models.

Findings

Self-interaction modifies heat flux by about 40% in typical cases.

Increasing parallel length and binormal width reduces self-interaction effects.

Effect is most significant with kinetic electrons, low magnetic shear, and steep gradients.

Abstract

In this work, we highlight an issue that may reduce the accuracy of many local nonlinear gyrokinetic simulations - turbulent self-interaction through the parallel boundary condition. Given a sufficiently long parallel correlation length, individual turbulent eddies can span the full domain and "bite their own tails," thereby altering their statistical properties. Such self-interaction is only modeled accurately when the simulation domain corresponds to a full flux surface, otherwise it is artificially strong. For Cyclone Base Case parameters and typical domain sizes, we find that this mechanism modifies the heat flux by roughly 40% and it can be even more important. The effect is largest when using kinetic electrons, low magnetic shear, and strong turbulence drive (i.e. steep background gradients). It is found that parallel self-interaction can be eliminated by increasing the parallel length and/or the binormal width of the simulation domain until convergence is achieved.