Planetary formation
Earth’s core accounts for ~33% of its total mass, consistently with what expected from the Sun's abundance of rock-forming elements. Mars’ has a smaller core-mass fraction than Earth’s according to data from NASA Insight. Mercury’s, some super-Earths like K2-229b, and perhaps some asteroids like 16 Psyche have an anomalously large metal content. How does this planetary diversity arise? My research aims to test the hypothesis that this planetary diversity is the result of giant impacts, which are energetic collisions between proto-planets thought to occur near the end of planetary formation. To test this hypothesis, I develop novel theories and machine-learning models aimed at improving the realism of giant impact in planet formation studies.
The role of giant impacts in planetary formation
Gabriel & Cambioni 2023, AREPS
Planets are thougth to conclude their growth through a series of energetic, global impacts whose outcomes strongly depend on the impact conditions. How do these giant impacts affect planetary composition and evolution? Will the projectile survive as a distinct object or being buried forever in the target planet? Will a moon form from the debris disk? In this review, we discuss how continued improvement in computer models and theory have helped addressing these questions and improved our ability to deduce probable collision conditions from observations of collision remnants. Despite these advancements, many open questions remain, as even the type of giant impact that formed Earth's Moon remains debated. Among many findings, we encourage the use of probability theory to evaluate distinct formation hypothesis among different sequences of giant impacts that can produce similar planets (e.g., side figure).
Videos of collision simulations (Supplementary materials in Gabriel & Cambioni 2023, AREPS)
Left axis: Diversity of metal contents for bodies in a population (measured as the number of different hit-and-run species—that is, the subset of the population whose bodies have survived the same number of hit-and-run collisions) as a function of total number of giant impacts. For each collision, the outcome (accretion or hit-and-run) is decided via coin flip (50-50% probability); metal content increases only in case of hit-and-run. Right axis: the fraction of bodies with same metal content to those of the initial population, whose bodies are initially identical to one another. This statistical exercise shows that the more collisionally evolved a population is, the higher its expected compositional diversity (or Shannon's information entropy), the higher the "information loss" about its initial metal content. From Figure 6 from Gabriel & Cambioni 2023.
The effect of inefficient accretion on planetary differentiation
Cambioni et al. 2021, PSJ
In this work, my colleagues and I studied the effect of collisional fragmentation of planets following giant impacts on their core formation following magma-ocean formation using the machine-learning models of giant impacts developed in previous studies (Cambioni et al. (2019), ApJ; Emsenhuber et al. 2020, ApJ, described below). We find that the model of perfect merging and the machine-learning treatment provide similar predictions for the mass and core mass fraction of planets more massive than 0.1 Earth’s masses (~1 Mars’ mass). At smaller scales, however, the inefficient accretion model predicts a higher degree of planetary diversity in terms of core mass fraction due to a predominance of hit-and-run collisions. This suggests that giant impacts played a key role in shaping the diversity of the core-mass fractions of the terrestrial planets in the solar system.
See my presentation of this work here: https://vimeo.com/457763367
The effect of inefficient accretion on planetary formation
In this work, we explore the effect of collisional fragmentation on the masses and composition terrestrial planets. We run N-body simulations of terrestrial planet formation by modelling collisions as perfectly inelastic and compare the results to simulations in which collision outcomes are predicted using using a machine-learning model trained to mimic high-resolution simulations of giant impacts. The solar system architectures obtained with the latter approach feature a much wider range of terrestrial planet masses and enhanced compositional diversity than what obtained using perfect merging of the colliding bodies, highlighting the importance of realistically modelling collisions in planet formation studies.
Code here: https://github.com/aemsenhuber/collresolve
Using machine learning to realistically model giant impacts in planet formation studies
The outcomes of giant impacts vary significantly depending on the impact conditions, ranging from accretion to the projectile by the target planet, to disruption of both bodies, to hit-and-run collisions, in which the projectile grazes the target and escapes accretion. Despite the importance of giant impacts in shaping the mass, composition, and orbits of planets, their modeling in planet formation and evolution studies has been traditionally simplified assuming inelastic collisions. This oversimplification was due to the high computational cost of running “on-the-fly” impact simulations. In this paper, my colleagues and I put forward a new method to solve this bottleneck by training machine-learning algorithms to mimic the outcome of expensive, high-resolution giant impact simulations. The simulations were run with the Smoothed Particle Hydrodynamics approach.
