Numerical Convergence of 2D Solar Convection in Implicit Large-eddy Simulations
Abstract
Large-eddy simulations (LES) and implicit LES (ILES) are wise and affordable alternatives to the unfeasible direct numerical simulations of turbulent flows at high Reynolds (Re) numbers. However, for systems with few observational constraints, it is a formidable challenge to determine if these strategies adequately capture the physics of the system. Here, we address this problem by analyzing numerical convergence of ILES of turbulent convection in 2D, with resolutions between 642 and 20482 grid points, along with the estimation of their effective viscosities, resulting in effective Reynolds numbers between 1 and ~104. The thermodynamic structure of our model resembles the solar interior, including a fraction of the radiative zone and the convection zone. In the convective layer, the ILES solutions converge for the simulations with ≥5122 grid points, as evidenced by the integral properties of the flow and its power spectra. Most importantly, we found that even a resolution of 1282 grid points, $\mathrm{Re}\,\sim \,10$ , is sufficient to capture the dynamics of the large scales accurately. This is a consequence of the ILES method allowing the energy contained in these scales to be the same in simulations with low and high resolution. Special attention is needed in regions with a small density scale height driving the formation of fine structures unresolved by the numerical grid. In the stable layer, we found the excitation of internal gravity waves, yet high resolution is needed to capture their development and interaction.
- Publication:
-
The Astrophysical Journal
- Pub Date:
- April 2022
- DOI:
- 10.3847/1538-4357/ac54b7
- arXiv:
- arXiv:2202.02767
- Bibcode:
- 2022ApJ...928..148N
- Keywords:
-
- Solar convective zone;
- Hydrodynamical simulations;
- Computational methods;
- 1998;
- 767;
- 1965;
- Astrophysics - Solar and Stellar Astrophysics
- E-Print:
- 15 pages, 9 figures, accepted for publication in The Astrophysical Journal