Abstract
The moons of giant planets are believed to form in situ in circumplanetary discs (CPDs). Here, we present an N-body population synthesis framework for satellite formation around a Jupiter-like planet, in which the dust-to-gas ratio, the accretion rate of solids from the protoplanetary disc, the number, and the initial positions of protosatellites were randomly chosen from realistic distributions. The disc properties were from 3D radiative simulations sampled in 1D and 2D grids and evolved semi-analytically with time. The N-body satellitesimals accreted mass from the solid component of the disc, interacted gravitationally with each other, experienced close-encounters, both scattering and colliding. With this improved modeling, we found that only about 15 per cent of the resulting population is more massive than the Galilean one, causing migration rates to be low and resonant captures to be uncommon. In 10 per cent of the cases, moons are engulfed by the planet, and 1 per cent of the satellite-systems lose at least 1 Earth-mass into the planet, contributing only in a minor part to the giant planet’s envelope’s heavy element content. We examined the differences in outcome between the 1D and 2D disc models and used machine learning techniques (Randomized Dependence Coefficient together with t-SNE) to compare our population with the Galilean system. Detecting our population around known transiting Jupiter-like planets via transits and TTVs would be challenging, but 14 per cent of the moons could be spotted with an instrumental transit sensitivity of 10−5.