Models of the Collisional Damping Scenario for Ice Giant Planets and Kuiper Belt Formation Harold F. Levison Department of Space Studies, Southwest Research Institute, Boulder, CO, USA 80302 7 0 [email protected] 0 2 n and a J 8 1 Alessandro Morbidelli 1 Observatoire de la Cˆote d’Azur, Nice, France v 4 4 5 Received ; accepted 1 0 7 0 / h p - o r t s a : v i X r a – 2 – ABSTRACT Chiang et al. (2006, hereafter C06) have recently proposed that the observed structure of the Kuiper belt could be the result of a dynamical instability of a system of 5 primordial ice giant planets in the outer Solar System. According ∼ to this scenario, before the instability occurred, these giants were growing in a highly collisionally dampedenvironment according to thearguments inGoldreich et al. (2004a,b, hereafter G04). Here we test this hypothesis with a series of numerical simulations using a new code designed to incorporate the dynamical effects ofcollisions. Wefindthatwe cannot reproduce theobserved SolarSystem. In particular, G04 and C06 argue that during the instability, all but two of the ice giants would be ejected from the Solar System by Jupiter and Saturn, leaving Uranus and Neptune behind. We find that ejections are actually rare and that instead the systems spread outward. This always leads to a configuration with too many planets that are too far from the Sun. Thus, we conclude that both G04’s scheme for the formation of Uranus and Neptune and C06’s Kuiper belt formation scenario are not viable in their current forms. 1. Introduction The investigation of the primordial processes that sculpted the structure of the Kuiper belt is still an active topic of research. Several models have been developed over the last decade, based on the effects of Neptune’s migration on the distant planetesimal disk (Malhotra 1995; Hahn & Malhotra 1999; Gomes 2003; Levison & Morbidelli 2003; see Morbidelli et al. 2003 for a review). However, many aspects of the Kuiper belt have not yet been fully explained. Moreover, a new paradigm about the giant planets orbital evolution has recently been proposed (Tsiganis et al. 2005; Gomes et al. 2005; see Morbidelli 2005 – 3 – or Levison et al. 2006 for reviews), which calls for a global revisiting of the Kuiper belt sculpting problem. In this evolving situation, Chiang et al. (2006, hereafter C06) have recently proposed an novel scenario. The idea is based on a recent pair of papers by Goldreich et al. (2004a,b, hereafter G04 for the pair and G04a and G04b for each individual paper), who, based on analytic arguments, predicted that originally roughly five planets began to grow between 20 and 40AU. However, as these planets grew to masses of 15M their orbits went ⊕ ∼ ∼ ∼ unstable, all but two of them were ejected, leaving Uranus and Neptune in their current orbits. C06 argued that this violent process could explain the structure of the Kuiper belt that we see today. We review the C06 scenario in more detail in section 2. Like G04, the C06 scenario was not tested with numerical simulations, but was solely supported by order-of-magnitude analytic estimates, which were only possible under a number of simplifications and assumptions. Therefore, the goal of this paper is to simulate numerically C06’s scenario, in order to see if the presence of five Neptune mass bodies (or a similar configuration) in a primordial planetesimal disk is indeed consistent with the observed structure of the outer Solar System (the orbital distribution of the Kuiper belt and of the planets). Because G04 and C06 scenarios heavily rely on the presence of a highly collisional planetesimal disk, we need first to develop a new numerical integrator that takes collisions into account as well as their effects on the dynamical evolution. This code is described and tested in Section 3. In section 4 we then describe the results of the simulations that we did of C06’s scenario. The conclusions and the implications are discussed in section 5. – 4 – 2. A Brief Description of the G04/C06 Scenario As we discussed above, C06’s scenario for sculpting the Kuiper belt is built on G04’s scheme for planet formation. In G04, the authors pushed to an extreme the concept of runaway (Ohtsuki & Ida 1990; Kokubo & Ida 1996) and oligarchic growth (Kokubo & Ida 1998; Thommes et al. 2003; Chambers, 2006) of proto-planets in planetesimal disks. Unlike previous works, they assumed that the bulk of the mass of the planetesimal disk is in particles so small (sub-meter to cm in size) that they have very short mean free paths. In this situation the disk is highly collisional, and the collisional damping is so efficient that the orbital excitation passed from the growing planets to the disk is instantaneously dissipated. With this set-up, the extremely cold disk exerts a very effective and time enduring dynamical friction on the growing planetary embryos, whose orbital eccentricities and inclinations remain very small. Consequently, the embryos grow quickly, accreting the neighboring material due to the fact that gravitational focusing is large. The order-of-magnitude analytic estimates that describe this evolution lead to the conclusion that the system reaches a steady state consisting of a chain of planets, separated by 5 Hill radii embedded in a sea of small particles. As the planets grow, their masses increase while their number decreases. This process continues until the surface density of the planetary embryos, Σ, is equal to that of the disk, σ. If the mass of the disk is tuned to obtain planets of Uranus/Neptune mass when Σ σ then the conclusion is that about 5 of ∼ these planets had to form in the range 20–40AU. G04 argue that when Σ σ the dynamical friction exerted on the planets by the disk ∼ is no longer sufficient to stabilize the planetary orbits. Consequently, the planets start to scatter one another onto highly elliptical and inclined orbits. They assume that all but two of the original ice planets (i.e. three of five in the nominal case) are ejected from the Solar System in this scattering process (no attempts were made to model this event). Once their – 5 – companions have disappeared, the two remaining planets feel a much weaker excitation, and therefore their orbits can be damped by the dynamical friction exerted by the remaining disk. This damping phase is then followed by a period of outward migration (Fern´andez & Ip 1986). They therefore become Uranus and Neptune, with quasi-circular co-planar orbits at 20 and 30 AU. ∼ ∼ C06 argues that this basic scenario, with some small modifications, can explain much of the structure currently seen in the Kuiper belt. The Kuiper belt displays a very complex dynamical structure. For our purposes, four characteristics of the Kuiper belt are important: 1) The Kuiper belt apparently ends near 50AU (Trujillo & Brown 2001; Allen et al. 2001, 2002). 2) The Kuiper belt appears to consist of at least two distinct populations with different dynamical and physical properties (Brown 2001; Levison & Stern 2001; Trujillo & Brown 2002; Tegler & Romanishin, 2003). One group is dynamically quiescent and thus we call it the cold population. All the objects in this population are red in color. The other group is dynamically excited, inclinations can be as large as 40◦, and thus we call it the hot population. It, too, contains red objects, but it also contains about as many objects that are gray in color. The largest objects in the Kuiper belt reside in the hot population. 3) Many members of the hot population are trapped in the mean motion resonances (MMRs) with Neptune. The most important of these is the 2:3MMR, which is occupied by Pluto. 4) The Kuiper belt only contains less than roughly 0.1M of material (Jewitt et al. 1996; ⊕ Chiang & Brown 1999; Trujillo et al. 2001; Gladman et al. 2001; Bernstein et al. 2004). This is surprising given that accretion models predict that & 10M must have existed in ⊕ this region in order for the objects that we see to grow (Stern 1996; Stern & Colwell, 1997; Kenyon & Luu, 1998, 1999). C06 suggests the following explanation for the Kuiper belt’s structure. First, in order to make the edge (characteristic 1 above), they assume that the planetesimal disk – 6 – is truncated at 47AU. This assumption is legitimate given the work of Youdin & Shu ∼ (2002) and Youdin & Chiang (2004) on planetesimal formation. They assume –as a variant of the pure G04 scenario – that some coagulation actually occurred in the planetesimal disk while the planets were growing. This coagulation produced a population of objects with a size distribution and a total number comparable to the hot population that we see today. Because this population constitutes only a small fraction of the total disk’s mass, their existence does not change the overall collisional properties of the disk, which are essential for G04’s story. Thus, during the final growth of the ice giants, C06 envisions three distinct populations: the ‘planetary embryos’ (objects that eventually become Neptune-sized), the ‘KBOs’ (macroscopic objects of comparable size to the current Kuiper belt objects, formed by coagulation), and ‘disk particles’ (golf-ball sized planetesimals that constitute the bulk of the disk’s mass and which have a very intense collision rate and damping). The KBOs are not massive enough to be affected by dynamical friction, but are big enough not to be damped by collisions with the disk-particles. As the planets grow to their final sizes, C06 estimate that the KBOs can be scattered by the growing planets to orbits with eccentricities and inclinations of order of 0.2. These, they argue, become the hot population, which has observed eccentricities and inclinations comparable to these values. After all but two of the original planets are removed by the dynamical instability and the ice giants evolve onto their current orbits inside of 30AU, C06 suggest that there is still a population of very small disk particles between 40 and 47AU. With the planets gone, this disk can become dynamically cold enough to allow large objects to grow in it, producing a second generation of KBOs on low-eccentricity and low-inclination orbits. These objects should be identified, in C06’s scenario, with the observed cold population of the Kuiper belt. – 7 – The mass of the disk between 40 and 47AU, however, should retain on order of half of its original mass, namely about 20 M according to the surface density assumed in C06’s ⊕ Equation (13). How this total mass was lost and how the cold population acquired its current, non-negligible eccentricity excitation, are not really explained by C06. The authors limit themselves to a discussion of the radial migration of Neptune, after the circularization of its orbit, to create the resonant populations (Malhotra 1995 Hahn & Malhotra 1999). This, in principal, might excite the cold classical belt, although C06 admit that it is not at all obvious how planet migration would proceed in a highly collisional disk. As for the mass depletion, collisional grinding is the only mechanism that makes sense at this point in time. However, it is not clear (at least to us) why collisional grinding would become so effective at this late stage while it was negligible during the planet formation and removal phases, when the relative velocities were much higher. At this point we want to emphasize that, although the ideas presented in G04 and C06 are new and intriguing, the papers do not present any actual models. Most of the arguments are based on order-of-magnitude equations where factors of 2 and √3 are dropped and approximate time-scales are set equal to one another in order to determine zeroth-order steady state solutions. In addition, simplifications are made to make the problem tractable analytically, like, for example, at any given instant all of the disk particles have the same size. Another example of a simplification is that the rate of change of the velocity dispersion of the disk particles due to collisions is simply set to the particle-in-the-box collision rate — the physics of the collisions are ignored. While making such approximations is reasonable when first exploring a problem and determining whether it could possibly work, numerical experiments are really required in order to determine whether the process does indeed act as the analytic expressions predict. Finally, many of the steps in these scenarios are not justified. Of particular interest – 8 – to us is the stage when, according to G04, all but two (i.e. three of five in the nominal case) of the original ice giants are removed from the system via a gravitational instability. The papers present order-of-magnitude equations that argue that such an instability would occur, but the authors are forced to speculate about the outcome of this event. Indeed, we suspect, based on our experiences, that G04’s expectations about the removal of the ice giants are naive. In particular, Levison et al. (1998) followed the dynamical evolution of a series of fictitious giant planet systems during a global instability. They found that during the phase when planets are scattering off of one another, the planetary system spreads to large heliocentric distances, and, while planets can be removed by encounters, the outermost planet is the most like to survive. Similarly, Morbidelli et al. (2002) studied systems of planetary embryos of various masses originally in the Kuiper belt and found that in all cases the embryos spread and some survived at large heliocentric distances. From these works we might expect that G04’s instability would lead to an ice giant at large heliocentric distances (but still within the observation limits), rather than having a planetary system that ends at 30AU with a disk of small particles beyond. Granted, the simulations in both Levison et al. (1998) and Morbidelli et al. (2002) did not include a disk of highly-damped particles that can significantly affect the evolution of the planets, so new simulations are needed to confirm or dismiss the G04/C06 scenario. In this paper we perform such simulations. We are required to develop a new numerical integration scheme to account for the collisional damping of the particle disk. This scheme is detailed and tested in the next section. – 9 – 3. The Code In this section we describe, in detail, the code that we constructed to test the C06 scenario. Before we can proceed, however, we must first discuss what physics we need to include in the models. As we described above, our motivation is to determine how a system containing a number of ice-giant planets embedded in a disk of collisionally damped particles dynamically evolves with time. Our plan is to reproduce the systems envisioned by G04 and C06 as closely as possible rather than create the most realistic models that we can. Thus, we purposely adopt some of the same assumptions employed by G04. For example, although G04 invoke a collisional cascade to set up the systems that they study, their formalism assumes that the disk particles all have the same size and ignore the effects of fragmentation and coagulation. We make the same assumption. In addition, although G04 invokes a collisional cascade to grind kilometer-sized planetesimals to submeter-sized disk particles, they implicitly assume that the timescale to change particle size is short compared to any of the dynamical timescales in the problem. Thus, their analytic representation assumes that the radius of the disk particles, s, is fixed. They determine which s to use by arguing that disk particles will grind themselves down until the timescale for the embryos to excite their orbits is equal to the collisional damping time (which is a function of s). Then s is held constant. We, therefore, hold s constant as well. In addition, G04 does not include the effects of gas drag in their main derivations, we again follow their lead in this regard. Our code is based on SyMBA (Duncan et al. 1998, Levison & Duncan 2000). SyMBA is a symplectic algorithm that has the desirable properties of the sophisticated and highly efficient numerical algorithm known as Wisdom-Holman Map (WHM, Wisdom & Holman 1991) and that, in addition, can handle close encounters (Duncan et al. 1998). This technique is based on a variant of the standard WHM, but it handles close encounters by – 10 – employing a multiple time step technique introduced by Skeel & Biesiadecki (1994). When bodies are well separated, the algorithm has the speed of the WHM method, and whenever two bodies suffer a mutual encounter, the time step for the relevant bodies is recursively subdivided. Although SyMBA represented a significant advancement to the state-of-art of integrating orbits, it suffers from a basic and serious limitation. At each time step of the integration, it is necessary to calculate the mutual gravitational forces between all bodies in the simulation. If there are N bodies, one therefore requires N2 force calculations per time step, because every object needs to react to the gravitational force of every other body. Thus, even with fast clusters of workstations, we are computationally limited to integrating systems where the total number of bodies of the order of a few thousand. Yet, in order to follow both the dynamical and collisional evolution of the numerous small bodies present during the G04’s scenario, we need to implement a way to follow the behavior of roughly 1026−29 particles. This clearly is beyond the capabilities of direct orbit integrators. Only statistical methods can handle this number of objects. In the following, we describe our approach to this problem. As described above, the systems that C06 envisions have three classes of particles: the planetary embryos, the KBOs, and the disk particles. Each class has its unique dynamical characteristics. The embryos are few in number and their dynamics are not directly effected by collisional damping. Thus, in our new code, which we call SyMBA COL, they can be followed directly in the standard N-body part of SyMBA. The KBOs are not dynamically important to the system from either a dynamical or collisional point of view. Since we are more concerned here with the final location of the ice giants than the dynamical state of the Kuiper belt, we ignore this population. Finally, we need to include a very large population of submeter-sized particles that both dynamically interact with the rest of the system and