The analyses of sunspot observations revealed a fundamental magnetic property of the umbral boundary, the invariance of the vertical component of the magnetic field. We aim to analyse the magnetic properties of the umbra-penumbra boundary in simulated sunspots and thus assess their similarity to observed sunspots. Also, we aim to investigate the role of plasma $beta$ and the ratio of kinetic to magnetic energy in simulated sunspots on the convective motions. We use a set of non-grey simulation runs of sunspots with the MURaM code. These data are used to synthesise the Stokes profiles that are then degraded to the Hinode spectropolarimeter-like observations. Then, the data are treated like real Hinode observations of a sunspot and magnetic properties at the umbral boundaries are determined. Simulations with potential field extrapolation produce a realistic magnetic field configuration on their umbral boundaries. Two simulations with potential field upper boundary, but different subsurface magnetic field structures, differ significantly in the extent of their penumbrae. Increasing the penumbra width by forcing more horizontal magnetic fields at the upper boundary results in magnetic properties that are not consistent with observations. This implies that the size of the penumbra is given by the subsurface structure of the magnetic field. None of the sunspot simulations is consistent with observed properties of the magnetic field and direction of the Evershed flow at the same time. Strong outward directed Evershed flows are only found in setups with artificially enhanced horizontal component of the magnetic field at the top boundary that are not consistent with the observed magnetic field properties at the UP boundary. We want to stress out that the `photospheric boundary of simulated sunspots is defined by a magnetic field strength of equipartition field value.
تحميل البحث