Assessment of a new sub-grid model for magneto-hydrodynamical turbulence. I. Magnetorotational instability

TL;DR

This paper introduces a new physically motivated sub-grid model, MInIT, for magneto-hydrodynamical turbulence that effectively captures small-scale stresses in MRI simulations, reducing computational costs and improving accuracy.

Contribution

The paper presents the MInIT model, a novel mean-field approach based on stress tensor evolution, validated against DNS data, and showing robustness across filter scales.

Findings

MInIT achieves less than one order-of-magnitude difference with DNS data.

The model's performance is independent of filter size, unlike gradient models.

MInIT has potential applications in simulating highly-magnetized astrophysical objects.

Abstract

Insufficient numerical resolution of grid-based, direct numerical simulations (DNS) hampers the development of instabilitydriven turbulence at small (unresolved) scales. As an alternative to DNS, sub-grid models can potentially reproduce the effects of turbulence at small scales in terms of the resolved scales, and hence can capture physical effects with less computational resources. We present a new sub-grid model, the MHD-instability-induced-turbulence (MInIT) mean-field model. MInIT is a physically motivated model based on the evolution of the turbulent (Maxwell, Reynolds, and Faraday) stress tensors and their relation with the turbulent energy densities of the magneto-rotational (MRI) and parasitic instabilities, modeled with two partial differential evolution equations with stiff source terms. Their solution allows obtaining the turbulent stress tensors through the constant coefficients that link them to the energy densities. The model is assessed using data from MRI in-box DNS and applying a filtering operation to compare the filtered data with that from the model. Using the $L_2$-norm as the metric for the comparison, we find less than one order-of-magnitude difference between the two sets of data. No dependence on filter size or length scale of unresolved scales is found, as opposed to results using the gradient model (which we also use to contrast our model) in which the $L_2$-norm of some of the stresses increases with filter size. We conclude that MInIT can help DNS by properly capturing small-scale turbulent stresses which has potential implications on the dynamics of highly-magnetized rotating compact objects, such as those formed during binary neutron star mergers.