The solar system’s outer planets (Jupiter, Saturn, Uranus, and Neptune) formed in just a few million years, while gas was still present in the Sun’s primordial protoplanetary disk. Although the evolution of these outer planets is well studied, the leading models seem to be incompatible with the solar system’s terrestrial system (Mercury, Venus, Earth, and Mars). By performing thousands of N-body computer simulations of an orbital instability in the outer solar system occurring in conjunction with terrestrial planet formation, my work has helped develop a simple and elegant explanation for Mars’ small size and rapid growth. My instability simulations consistently outperform control runs when measured against a variety of success criteria. These include the orbits, masses, and geological formation timescales of the planets; the size and orbital structure of the asteroid belt; and the water content of Earth. When the instability occurs approximately 1–10 million years after gas disk dispersal, “Mars” is just one of several Mars-sized objects with similar orbits (Fig. 1). The frequent perturbations from the increasingly eccentric orbits of Jupiter and Saturn quickly cause these objects (and similar bodies in the asteroid belt) to either be ejected from the system or scatter inward towards the proto-Earth (sometimes to deliver water). In successful simulations, Mars undergoes no further accretion events after the instability, while Earth and Venus continue to grow (thus matching their relative geological formation times).

However, accurately modeling the late stages of planet accretion is subject to numerical limitations and simplifications. In particular, to keep the calculation tractable, most authors employ integration schemes that neglect collisional fragmentation. The initial planetforming disk, which in reality contained millions of solid objects with a range of masses, must also be approximated with just over a thousand bodies (the majority of which are assumed not to interact gravitationally with one another).  Recently, I have been using graphical processing units (gpus; which greatly speed up simulations by performing calculations in parallel) to continue my study of terrestrial planet formation, thoroughly investigate the complex orbital dynamics within the asteroid belt, probe observational constraints on theoretical models by studying the tail-end phase of bombardment and clearing in the young solar system, and reevaluate the common initial conditions used when studying terrestrial planet formation.