Constrained Hyperbolic Divergence Cleaning for Smoothed Particle Magnetohydrodynamics

TL;DR

This paper introduces a constrained hyperbolic divergence cleaning scheme for Smoothed Particle Magnetohydrodynamics that guarantees energy conservation or dissipation, improving stability and accuracy over previous methods.

Contribution

It develops a novel constrained formulation using conjugate operators, enhancing divergence cleaning stability and reducing magnetic energy errors in SPMHD simulations.

Findings

Maintains divergence errors at 0.1-1% levels, significantly better than previous methods.

Improves momentum conservation by up to two orders of magnitude in 3D simulations.

Enhances numerical stability without additional computational cost.

Abstract

We present a constrained formulation of Dedner et al's hyperbolic/parabolic divergence cleaning scheme for enforcing the \nabla\dot B = 0 constraint in Smoothed Particle Magnetohydrodynamics (SPMHD) simulations. The constraint we impose is that energy removed must either be conserved or dissipated, such that the scheme is guaranteed to decrease the overall magnetic energy. This is shown to require use of conjugate numerical operators for evaluating \nabla\dot B and \nabla{\psi} in the SPMHD cleaning equations. The resulting scheme is shown to be stable at density jumps and free boundaries, in contrast to an earlier implementation by Price & Monaghan (2005). Optimal values of the damping parameter are found to be {\sigma} = 0.2-0.3 in 2D and {\sigma} = 0.8-1.2 in 3D. With these parameters, our constrained Hamiltonian formulation is found to provide an effective means of enforcing the divergence constraint in SPMHD, typically maintaining average values of h |\nabla\dot B| / |B| to 0.1-1%, up to an order of magnitude better than artificial resistivity without the associated dissipation in the physical field. Furthermore, when applied to realistic, 3D simulations we find an improvement of up to two orders of magnitude in momentum conservation with a corresponding improvement in numerical stability at essentially zero additional computational expense.