The dynamical structure of the Solar System can be explained by a period of orbital instability experienced by the giant planets. While a late instability was originally proposed to explain the Late Heavy Bombardment, recent work favors an early instability. We model the early dynamical evolution of the outer Solar System to self-consistently constrain the most likely timing of the instability. We first simulate the dynamical sculpting of the primordial outer planetesimal disk during the accretion of Uranus and Neptune from migrating planetary embryos during the gas disk phase, and determine the separation between Neptune and the inner edge of the planetesimal disk. We performed simulations with a range of migration histories for Jupiter. We find that, unless Jupiter migrated inwards by 10 AU or more, the instability almost certainly happened within 100 Myr of the start of Solar System formation. There are two distinct possible instability triggers. The first is an instability that is triggered by the planets themselves, with no appreciable influence from the planetesimal disk. Of those, the median instability time is $\sim4$Myr. Among self-stable systems -- where the planets are locked in a resonant chain that remains stable in the absence of a planetesimal's disk-- our self-consistently sculpted planetesimal disks nonetheless trigger a giant planet instability with a median instability time of 37-62 Myr for a reasonable range of migration histories of Jupiter. The simulations that give the latest instability times are those that invoked long-range inward migration of Jupiter from 15 AU or beyond; however these simulations over-excited the inclinations of Kuiper belt objects and are inconsistent with the present-day Solar System. We conclude on dynamical grounds that the giant planet instability is likely to have occurred early in Solar System history.

The link between the dynamical evolution of the giant planets and the Kuiper Belt orbital structure can provide clues and insight about the dynamical history of the Solar System. The classical region of the Kuiper Belt has two populations (the cold and hot populations) with completely different physical and dynamical properties. These properties have been explained in the framework of a subset of the simulations of the Nice Model, in which Neptune remained on a low-eccentricity orbit (Neptune's eccentricity is never larger than 0.1) throughout the giant planet instability. However, recent simulations have showed that the remaining Nice model simulations, in which Neptune temporarily acquires a large-eccentricity orbit (larger than 0.1), are also consistent with the preservation of the cold population (inclination smaller than 4 degrees), if the latter formed in situ. However, the resulting a cold population showed in many of the simulations eccentricities larger than those observed for the real population. We focus on a short period of time which is characterized by Neptune's large eccentricity and a slow precession of Neptune's perihelion. We show that if self-gravity is considered in the disk, the precession rate of the particles longitude of perihelion is slowed down. This, combined with the effect of mutual scattering among the bodies, which spreads all orbital elements, allows some objects to return to low eccentricities. However, we show that if the cold population originally had a small total mass, this effect is negligible. Thus, we conclude that the only possibilities to keep at low eccentricity some cold-population objects during a high-eccentricity phase of Neptune are that (i) either Neptune's precession was rapid, as suggested by Batygin et al. (2011) or (ii) Neptune's slow precession phase was long enough to allow some particles to experience a full secular cycle.