Mass evolution and collapse of protoplanets in massively parallel radiation-hydro simulations of disk instability with protoplanet feedback.

We propose a simulation campaign that considerably extends our ongoing activity to investigate the disk instability model of planet formation. We will perform both ultra-hi res global disk simulations and local simulations of individual protoplanetary clump collapse, following up on Galvagni et al. (2012), including previously missing physical processes such as radiative feedback from collapsing protoplanets (protoplanet feedback), analogous to radiation feedback in protostellar collapse, and opacity evolution du to dust grain settling and coagulation within clumps, which has been preliminarly studied with 1D models (Helled & Bodenehimer 2011). By combining global and local simulations we will complement realistic dynamics and hydrodynamics for clump evolution in disks while allowing to explore the parameter space of detailed physical effects important for clump collapse.
We will address in a novel way the most crucial, long-standing open issue in disk instability, namely whether or not the fnal masses of protoplanets remain below the deuterium burning limit even they survive and evolve over long timescales approaching 100.000 yr.
The massively parallel 3D TreeSPH code ChaNGa, which is written in the Charm ++ environment and achieves dynamic load balancing scaling to many thousand CPU and GPU cores on the hybrid architecture of PizDaint, will be our main tool. ChaNGa will be aided by the recent development of the STARRAD ray-tracing code within from the PASC DIAPHANE project
(Dykes et al., in prep.) combined with two different implementations of Flux-Limiited Diffusion (FLD). STARRAD will be heavily used in the global simulations to model both stellar irradiation and protoplanet feedback . It will be attached to sink particles to model the rapidly collapsing , not fully resolved core region of the protoplanets.
We will run 4 disk simulations with 40 million particles each, enough to resolve the detailed structure of the clumps with a dense core and a rotating circumplanetary disk (Szulagiy, Mayer & Quinn 2016), and 12 local collapse simulations exploring the large parameter space of opacity effects.
Based on our new benchmarks and taking into account the forthcoming node upgrade of PizDaint we request an allocation of 975000 node hrs for one year to complete the planned runs.
We expect the outcome of this project will foster great progress in our understanding of the population of planets potentially formed by disk instability, bringing an important contribution to the NCCR PlanetS of which the PI and part of his group are active members.