Elsevier

Icarus

Volume 163, Issue 1, May 2003, Pages 198-231
Icarus

Regular article
Formation of the regular satellites of giant planets in an extended gaseous nebula I: subnebula model and accretion of satellites

https://doi.org/10.1016/S0019-1035(03)00076-9 Get rights and content

Abstract

We model the subnebulae of Jupiter and Saturn wherein satellite accretion took place. We expect each giant planet subnebula to be composed of an optically thick (given gaseous opacity) inner region inside of the planet’s centrifugal radius (where the specific angular momentum of the collapsing giant planet gaseous envelope achieves centrifugal balance, located at rCJ ∼ 15RJ for Jupiter and rCS ∼ 22RS for Saturn) and an optically thin, extended outer disk out to a fraction of the planet’s Roche-lobe (RH), which we choose to be ∼RH/5 (located at ∼150 RJ near the inner irregular satellites for Jupiter, and ∼200RS near Phoebe for Saturn). This places Titan and Ganymede in the inner disk, Callisto and Iapetus in the outer disk, and Hyperion in the transition region. The inner disk is the leftover of the gas accreted by the protoplanet. The outer disk may result from the nebula gas flowing into the protoplanet during the time of giant planet gap-opening (or cessation of gas accretion). For the sake of specificity, we use a solar composition “minimum mass” model to constrain the gas densities of the inner and outer disks of Jupiter and Saturn (and also Uranus). Our model has Ganymede at a subnebula temperature of ∼250 K and Titan at ∼100 K. The outer disks of Jupiter and Saturn have constant temperatures of 130 and 90 K, respectively.

Our model has Callisto forming in a time scale ∼106 years, Iapetus in 106–107 years, Ganymede in 103–104 years, and Titan in 104–105 years. Callisto takes much longer to form than Ganymede because it draws materials from the extended, low density portion of the disk; its accretion time scale is set by the inward drift times of satellitesimals with sizes 300–500 km from distances ∼100RJ. This accretion history may be consistent with a partially differentiated Callisto with a ∼300-km clean ice outer shell overlying a mixed ice and rock-metal interior as suggested by Anderson et al. (2001), which may explain the Ganymede–Callisto dichotomy without resorting to fine-tuning poorly known model parameters. It is also possible that particulate matter coupled to the high specific angular momentum gas flowing through the gap after giant planet gap-opening, capture of heliocentric planetesimals by the extended gas disk, or ablation of planetesimals passing through the disk contributes to the solid content of the disk and lengthens the time scale for Callisto’s formation. Furthermore, this model has Hyperion forming just outside Saturn’s centrifugal radius, captured into resonance by proto-Titan in the presence of a strong gas density gradient as proposed by Lee and Peale (2000). While Titan may have taken significantly longer to form than Ganymede, it still formed fast enough that we would expect it to be fully differentiated. In this sense, it is more like Ganymede than like Callisto (Saturn’s analog of Callisto, we expect, is Iapetus). An alternative starved disk model whose satellite accretion time scale for all the regular satellites is set by the feeding of planetesimals or gas from the planet’s Roche-lobe after gap-opening is likely to imply a long accretion time scale for Titan with small quantities of NH3 present, leading to a partially differentiated (Callisto-like) Titan. The Cassini mission may resolve this issue conclusively. We briefly discuss the retention of elements more volatile than H2O as well as other issues that may help to test our model.

Introduction

The regular satellites of Jupiter and Saturn generally have low inclinations and eccentricities. Perhaps most striking is the progression of satellite density in the Galilean system. Also, the ratios between the satellite systems and the parent bodies of mass and angular momentum are quite similar (Pollack et al., 1991), which suggests a common origin in an accretion disk present about the protoplanets at a late stage of their formation. These properties, taken together with the tantalizing ratio of the largest satellite of each system to its primary Ms/MP ∼ 10−4 (not too dissimilar from the ratio of giant planet to Sun), lead one to think of the Galilean satellite system as a kind of scaled-down solar system.

Given their similarities in distances, masses, and densities, an issue we wish to focus on is how to view Titan in light of the Galilean satellite system, especially Ganymede and Callisto. Yet, the differences between these three satellites are as intriguing as their similarities. The Galileo mission moment of inertia data are consistent with a fully differentiated Ganymede, but only a partially differentiated Callisto (Anderson et al., 1998). Moreover, Callisto shows no evidence of tectonic activity. Also, the association of craters with the presence of CO2 in Callisto but not Ganymede (Hibbitts et al., 2000) as well as the degradation of craters presumably due to the sublimation of CO2 in Callisto but not Ganymede (Moore et al., 1999), which is consistent with the presence of a CO2 atmosphere in Callisto (Carlson, 1999), seems to require that Callisto be assembled with and retain oxidized ices more volatile than H2O. In the case of Titan, it is probably the presence of methane in the atmosphere that has received the most attention Lunine 1989, Coradini et al 1989, Prinn and Fegley 1989.

Recently, Anderson et al. (2001) have investigated two and three layer models for Callisto’s internal structure assuming hydrostatic equilibrium. For the two layer models these authors find two limiting cases: a relatively pure ice shell about ∼300 km overlying a mixed ice and rock-metal interior, and a thick ≳1000-km ice and rock-metal outer shell overlying a rock-metal core. Since it is difficult to reconcile a metallic core with a partially differentiated state the former solution appears more likely. Given that accreting bodies allocate a fraction of their energy as surface heat Schubert et al 1981, Coradini et al 1982, fast satellite accretion would melt the water ice and lead to rock separation and runaway differentiation (Friedson and Stevenson, 1983). Previous attempts to explain an undifferentiated Callisto have relied on fine-tuning parameters Schubert et al 1981, Coradini et al 1982, Lunine and Stevenson 1982. Although it is possible that nonhydrostatic effects in Callisto’s core could be large enough to allow for complete differentiation of this satellite and still be sufficiently small in Ganymede’s core to have avoided detection, we regard this possibility as unlikely. Instead, we favor a model that makes Callisto slowly.

Other issues also seem difficult to explain. For instance, one might expect the outermost Galilean satellite to have significantly less angular momentum than the preceding satellite. It would seem unlikely that the satellite disk would have enough surface density to make a satellite the size of Callisto at 26RJ, but form no smaller objects outside its orbit. Furthermore, the separation between Ganymede and Callisto (∼10RJ) is so large that one is led to wonder why there are no satellites in between at ∼20RJ (see Mosqueira and Estrada (2003), hereafter Paper II, for a brief discussion of orbital stability). One can always argue serendipity, but the Galilean satellite system is sufficiently regular that we reserve this explanation as a last resort.

A related point can be made concerning Titan and Iapetus. If we form the satellites out of a continuous, smoothly varying accretion disk, it would seem difficult to explain why there are no large satellites between Titan at ∼20RS and Iapetus at ∼60RS (Hyperion does not have enough mass to affect this argument).

Also, one must account for the differences between the satellite systems of Jupiter and Saturn. In the case of Saturn’s satellite system, the concentration of mass in Titan needs to be addressed. But perhaps the most perverse difference between the two satellite systems is the fact that whereas the Galilean satellites get rockier closer to the planet, the inner satellites of Saturn appear to be made mostly of ice! Even so, we attempt a combined model for both Jupiter and Saturn (as well as Uranus).

If we take the satellite systems of Jupiter and Saturn and add the amount of gas necessary to create a solar composition mixture the resulting disks have a total angular momentum comparable to the spin angular momentum of the parent planet (Stevenson et al., 1986). The issue arises whether or not one would expect the circumplanetary disk to exhibit a solar mixture of elemental abundances of water and ice bearing materials. One can think of several processes that modified the abundances of rock and ice from their solar abundances. Yet, the fact that the similarly sized Ganymede, Callisto, and Titan all deviate from solar mixture by the same proportion (∼60% rock, ∼40% ice by mass) seems to indicate that one should be guided by solar mixtures and investigate mechanisms for deviation from them, such as size-dependent water vaporization on one end and water enrichment by composition selective mechanisms on the other. If so, one might calculate models with “minimum mass” by augmenting the mass of the satellites by some factor (typically ∼100; Pollack et al., 1994), corresponding to the mass ratio of gas to rock-metal/ice in the solar nebula. This factor might be decreased somewhat in view of the heavy-element enrichment of the giant planets or increased in view of the possible loss of some of the accreting materials as a result of the specifics of the process used to make the planet and satellites.

In order to arrive at a specific model for the formation of regular satellites in a gaseous medium we need to characterize the subnebular viscosity. It has been suggested that because of the stabilizing influence of a positive radial gradient in specific angular momentum, turbulence in a Keplerian disk is not self-sustaining unless a source of “stirring” is found Ryu and Goodman 1992, Balbus et al 1996, Stone and Balbus 1996. As a result, one needs to identify a specific mechanism that can maintain turbulence in the dense, high orbital frequency subnebula. One such suggestion is that convection drives turbulence Cameron 1978a, Lin and Papaloizou 1980, Ruden and Lin 1986; however, eventually particle growth may stop convection by diminishing the Rosseland mean opacity and weakening its temperature dependence (Weidenschilling and Cuzzi 1993). Given the fast dynamical time scale and the high particle density of the subnebula disk, coagulation and settling for sticky particles may take place on a time scale faster than disk evolution. Furthermore, if convection drives turbulence then angular momentum transport may be weak (α ∼ 10−5; Stone and Balbus, 1996) and directed inwards Ryu and Goodman 1992, Kley et al 1993, Stone and Balbus 1996, Cabot 1996, which would essentially terminate gas accretion onto the primary. Another possibility is that turbulence is driven by a magnetohydrodynamic (MHD) instability (Balbus and Hawley, 1991). But this is also unlikely to apply (Gammie, 1996) in the dense and relatively cool subnebula disk. Alternatively, there are a variety of ways that accretion itself, or the gravitational energy released by it, can provide the source of free energy that can drive turbulence. It has been pointed out (but not quantitatively explored) that a turbulent shear layer, where the angular momentum of the infalling gas is adjusted to the angular momentum of the Keplerian disk flow, exists below an accretion shock and may provide a localized viscosity Cassen and Moosman 1981, Cassen and Summers 1983. More recently it has been shown that a bump in the temperature profile of the disk, as may result from accretion, that leads to a strong radial entropy gradient generates Rossby waves and localized turbulence Lovelace et al 1999, Li et al 2000. Similarly, but more generally, Klahr and Bodenheimer (2001) study a global baroclinic instability as a source of turbulence and outward angular momentum transport in Keplerian accretion disks characterized by a negative radial entropy gradient.

To create a coherent scenario of satellite formation, the source of the solids that go into the satellite systems must be considered. The concentration of rock/ice to gas in the subnebula may depend on the ability of the protoplanet to disturb the orbits of planetesimals situated within a few AU of its orbit into ones that crossed its orbit. One would expect that in a time scale much shorter than the lifetime of the solar system virtually all the planetesimals located in the outer solar system would have their orbits perturbed into giant planet crossing orbits (Gladman and Duncan, 1990). What happens to such a planetesimal depends on the size of the planet at the time of crossing. If the giant planet’s envelope filled a fair fraction of its Hill radius, as it probably did during most or all of its gas accretion phase (unless significant amounts of gas accreted through the gap after gap-opening), then the distended atmosphere would have greatly increased the planet’s cross-section Bodenheimer and Pollack 1986, Pollack et al 1996. Early arriving (before the completion of planetary accretion) icy planetesimals of size <10 km (Zahnle, private communication) may break up in the contracting envelope of the giant planet, and their condensable content may then be entrapped in the gas and left behind in the form of a circumplanetary disk. On the other hand, most late arriving planetesimals may have been scattered to further regions of the solar system with some sent to the Oort cloud and some lost altogether. Thus, our model relies on early arriving planetesimals that break up or dissolve in the extended giant-planet envelopes to provide the bulk of the material that will eventually make the satellite systems, delivered to the satellite disk in a time scale given by the envelope collapse time.

This formation model is consistent both with a model that captures irregular satellites at a time when the proto-planetary envelope was collapsing rapidly and extended several hundred planetary radii (Pollack et al., 1979) and with a model that captures irregular satellites using a long-lived circumplanetary gas disk (Cuk and Burns, 2002). Here late arriving interplanetary debris plays a role in that it can threaten the survival of regular satellites close to their primary. Hence, the large disparity in masses between Titan and all other moons of Saturn may in part be the result of the break-up of satellites by high-velocity impacts (e.g., Lissauer, 1995; but note that gas would still be needed to clear up the collisional debris and prevent re-accretion). In contrast, a starved disk model (Stevenson, 2001) relies on the late arriving planetesimals or flow through the gap to form a disk around the planet out of which all the regular satellites will eventually accrete. One should keep in mind, however, that most planetesimals were probably scattered or the giant planets would have ended up with too much high-Z mass (Podolak et al., 1993) and that most of the mass in the nebula disk at late times is in the form of planetesimals Mizuno et al 1978, Weidenschilling 1997. Furthermore, the high specific angular momentum of gas arriving at late times may place it in orbit well outside the region where most of the satellite mass is found.

In Section 2 we organize the satellite systems of the giant planets according to the Hill radius of the primary. In Section 3 we characterize the subnebulae of giant planets, especially that of Jupiter. In Section 4 we discuss the accretion of the Galilean satellites, reserving discussion of Callisto for Section 5. In Section 6 we turn to Saturn’s satellite system. In Section 7 we discuss the satellite system of Uranus. In Section 8 we make some comments on an alternative satellite accretion model that leads to a long accretion time scale for every satellite. In Section 9 we present our conclusions and discussion. In Paper II we turn to the migration and survival of full-sized satellites.

Section snippets

Regular satellites of giant planets

We begin with a brief comparative discussion of the satellite systems of the giant planets. We compare satellite positions mainly in terms of the Hill radius RH = a(MP/3M)1/3 of the planet (and the concomitant centrifugal radius rcRH/48; see Sect. 3). In Fig. 1, we plot the locations of the regular satellites (solid circles) and the innermost irregular satellites (open circles) in units of the Hill radius of the giant planet. The bold dashed line describes the position of the centrifugal

The giant planet subnebula

The “minimum” mass subnebula we use here is one of solar nebula composition that provides just enough mass to form the observed satellite systems with the observed rock/ice mass ratio. Given Jupiter’s relative enrichment in heavy elements with respect to the solar nebula, the minimum mass subnebula is not a firm lower bound. On the other hand, inefficiencies in the satellite formation process and depletion of solids due to planetesimal formation mean that it is not a firm upper bound either.

Galilean satellite accretion and evolution

In analogy to gas-free planetary accretion, we begin the problem of satellite accretion by calculating characteristic sizes of satellitesimals and satellite embryos for our disk parameters assuming a satellitesimal density of ρs = 1.5 g cm−3. Though our problem differs markedly from one in which the satellites are accreted in the absence of gas, we will show later that the characteristic sizes one obtains in the presence of gas are roughly consistent with the ones we give below, which are meant

Slow formation of callisto

Our model has Callisto forming from an extended, low optical depth gas disk. We expect this gas disk to be largely quiescent with very low gas viscosity. This means that the dust and rubble layer will quickly settle down to the midplane within a scale-height much smaller than the gas scale-height. The size of the dust and rubble layer is determined by shear turbulence close to the midplane (Cuzzi et al., 1993).

First we calculate characteristic masses and lengths in analogy to the gas free

Saturn’s regular satellite system

In order to apply our model to Saturn we first need to constrain the nebula parameters for Saturn as we did for Jupiter. First, we note that the ratio of the reconstituted Galilean satellite masses to the saturnian satellite masses is ≈3.7. On the other hand, the ratio of the atmospheric envelopes of Jupiter to Saturn is ≈3.7 for giant planet core masses of ≈12 Earth masses, consistent with nominal values.

In Fig. 2, we plot two models. The first model simply assumes the same mass ratio (∼3.7)

Uranian satellite system

As we did in the case of Saturn, we begin our discussion of the satellite system of Uranus by comparing the ratio of masses of the atmospheric envelopes to the ratio of masses of the satellite systems. Assuming Saturn to have a ∼15 Earth mass core and Uranus a smaller ∼10 Earth mass core, the mass ratio of the envelopes for the two planets is ∼18. While substantially uncertain, this value compares favorably with the mass ratio of the satellite systems of the two planets ∼15.

All the regular

Starved disk model

A scenario in which the giant planet satellites accrete from a disk produced by the direct infall of gas and solids from heliocentric orbit, leading to long satellite assembly times ∼106 years despite the short disk accumulation times ∼102 years (Stevenson, personal communication), has numerous issues to overcome. Here we mention some of the outstanding ones and leave development of such a model (if viable) for later work.

First is the issue of satellite survival. Recent numerical simulations

Discussion and conclusions

We have used a consistent model for the accretion of regular satellites of Jupiter, Saturn, and Uranus. We do exclude coorbitals, small regular satellites found close to the giant planet, and the satellites of Neptune from consideration. The coorbitals will require a separate treatment, which we do not attempt here. Small satellites close to the planet are likely to have undergone significant collisional or tidal evolution after their formation, so they may not provide useful constraints on

Acknowledgements

We thank Jeffrey Cuzzi, Kevin Zahnle, Doug Lin, Peter Bodenheimer, Sandy Davis, Dale Cruikshank, and Jeff Moore for discussions. One of us (I.M.) had numerous, very helpful discussions with Dave Stevenson. We also thank Jeffrey Cuzzi, Kevin Zahnle, and Jack Lissauer for reading the manuscript and suggesting improvements, and Steve Squyres for discussions, comments on the manuscript, and generous support of this work. This research was supported by a grant from the Planetary Geology and

References (127)

  • D.R. Davis et al.

    The missing Psyche familycollisionally eroded or never formed?

    Icarus

    (1999)
  • S.F. Dermott et al.

    The determination of the mass and mean density of Enceladus from its observed shape

    Icarus

    (1994)
  • B. Dubrulle

    Differential rotation as a source of angular momentum transport in the solar nebula

    Icarus

    (1993)
  • A.J. Friedson et al.

    Viscosity of rock-ice mixtures and applications to the evolution of icy satellites

    Icarus

    (1983)
  • A. Fujiwara et al.

    Destruction of basalt bodies by high-velocity impact

    Icarus

    (1977)
  • B. Gladman et al.

    The discovery of Uranus XIX, XX, and XXI

    Icarus

    (2000)
  • P. Goldreich et al.

    The formation of the Cassini division of Saturn’s rings

    Icarus

    (1978)
  • J. Goodman et al.

    Secular instability and planetesimal formation in the dust layer

    Icarus

    (2000)
  • R. Greenberg et al.

    Size distribution of particles in planetary rings

    Icarus

    (1977)
  • J.M. Hahn et al.

    Resonant trapping in a self-gravitating planetesimal disk

    Icarus

    (1995)
  • K. Hourigan et al.

    Radial migration of preplanetary materialimplications for the accretion time scale problem

    Icarus

    (1984)
  • D.M. Kary et al.

    Nebular gas drag and planetary accretion

    Icarus

    (1993)
  • D.G. Korycansky et al.

    Numerical models of giant planet formation with rotation

    Icarus

    (1991)
  • J.S. Lewis

    Low temperature condensation from the solar nebula

    Icarus

    (1972)
  • J.J. Lissauer

    Urey prize lectureon the diversity of plausible planetary systems

    Icarus

    (1995)
  • J.I. Lunine et al.

    Formation of the Galilean satellites in a gaseous nebula

    Icarus

    (1982)
  • R. Malhotra

    Orbital resonances in the solar nebulastrengths and weaknesses

    Icarus

    (1993)
  • R. Malhotra et al.

    The role of secondary resonances in the orbital history of Miranda

    Icarus

    (1990)
  • J.M. Moore et al.

    Mass movement and landform degradation on the icy Galilean satellitesresults of the Galileo nominal mission

    Icarus

    (1999)
  • O. Mousis et al.

    An evolutionary turbulent model of Saturn’s subnebulaimplications for the origin of the atmosphere of Titan

    Icarus

    (2002)
  • T.C. Owen et al.

    Decoding the dominothe dark side of Iapetus

    Icarus

    (2001)
  • J.B. Pollack et al.

    The formation of Saturn’s satellites and rings as influenced by Saturn’s contraction history

    Icarus

    (1976)
  • J.B. Pollack et al.

    A calculation of Saturn’s gravitational contraction history

    Icarus

    (1977)
  • J.B. Pollack et al.

    Gas drag in primordial circumplanetary envelopesa mechanism for satellite capture

    Icarus

    (1979)
  • J.B. Pollack et al.

    Formation of the giant planets by concurrent accretion of solids and gas

    Icarus

    (1996)
  • I. Adachi et al.

    The gas drag effect on the elliptic motion of a solid body in the primordial solar nebula

    Prog. Theor. Phys.

    (1976)
  • T.J. Ahrens et al.

    Strength versus gravity dominance in catastrophic impacts

    Lunar. Planet Sci.

    (1996)
  • J.D. Anderson et al.

    Distribution of rock, metals, and ices in Callisto

    Science

    (1998)
  • P. Artymowicz et al.

    Mass flow through gaps in circumbinary disks

    Astrophys. J.

    (1996)
  • S.A. Balbus et al.

    A powerful local shear instability in weakly magnetized disks. II. Nonlinear evolution

    Astrophys. J.

    (1991)
  • S.A. Balbus et al.

    Nonlinear stability, hydrodynamical turbulence, and transport in disks

    Astrophys. J.

    (1996)
  • Beatty, J.K., ed., 1999. The New Solar System. Cambridge Univ. Press,...
  • S.F. Bell

    CallistoJupiter’s Iapetus?

    Lunar and Planet Sci.

    (1984)
  • M.E. Brown et al.

    Detection of water ice on Nereid

    Astrophys. J.

    (1998)
  • G. Bryden et al.

    Tidally induced gap formation in protostellar disksgap clearing and suppression of protoplanetary growth

    Astrophys. J.

    (1999)
  • G. Bryden et al.

    On the interaction between protoplanets and protostellar disks

    Astrophys. J.

    (2000)
  • W. Cabot

    Numerical simulations of circumstellar disk convection

    Astrophys. J.

    (1996)
  • A.G.W. Cameron

    Physics of the primitive solar accretion disk

    Moon and Planets

    (1978)
  • A.G.W. Cameron

    Physics of the primitive solar nebula and of giant gaseous protoplanets

  • Canup, R.M., Ward, W.R., 2000. A possible impact origin of the uranian satellite system. Proceedings of the 32nd DPS...
  • Cited by (218)

    View all citing articles on Scopus
    View full text