Chaotic planet formation

Article metrics

Our understanding of how planets formed around the Sun is far from complete, and is increasingly challenged by new discoveries of unusual planetary-mass companions around other stars. The idea that chaotic dynamics — or extreme sensitivity to initial conditions — plays a large role in the formation, evolution and survival of planetary systems has gained popularity in recent years. This is in contrast to earlier ideas that solar systems similar to our own are an inevitable by-product of the formation of Sun-like stars.

On page 635 of this issue, Thommes et al.1 propose that all of the giant planets in our Solar System formed in a narrow region of the disk-like solar nebula that surrounded the young Sun: they ended up in their present widely spaced orbits as a result of violent and chaotic scattering. Also, on page 633, Armitage and Hansen2 propose that the early formation of a Jupiter-mass planet in a nebular disk could ‘trigger’ the formation of other giant planets. Such protoplanets would also form in a narrow region, and would be expected to evolve chaotically into systems that might resemble some of the putative extrasolar planetary systems.

Both groups have used computer simulations to advance these ideas. In our Solar System there are two ‘gas giants’ (Jupiter and Saturn), having a small rocky core surrounded by a large hydrogen and helium atmosphere, and two ‘ice giants’ (Uranus and Neptune), which have icy mantles around their cores and only a thin atmosphere. Theorists have struggled to explain how Uranus and Neptune formed at their present locations in the Solar System, because at such distances from the Sun the nebular disk density would have been low, and so the time required for these planets to grow is thought to be longer than the age of the Solar System.

The Thommes et al.1 model postulates that the four giant planets started out as rocky cores in the Jupiter–Saturn region, and that the cores of Uranus and Neptune were violently tossed out by gravitational scattering from Jupiter and Saturn, which had already accreted large amounts of gas. In the simulations, the ejected planets remain in highly chaotic orbits for a short period of time (a few hundred thousand years), after which they settle down (by means of friction with the nebular disk) and gradually migrate out to their present, nearly circular orbits. The authors start their simulations with four cores embedded in a narrow zone of the solar nebula, and find that in 50% of their trials the giant planets end up more or less where they are now.

The model is unusual in suggesting that all four giant planets originated in the same region of the solar nebula — within a ring 5–10 astronomical units from the Sun (1 au 4 Sun–Earth distance). This is substantially narrower than previous estimates of 10–20 au for the birthplace of Uranus and Neptune3, and far inside their current orbits at 19 and 30 au, respectively. Because the nebular disk density and orbital rotation is much higher at 5 au than at 30 au (ref. 4), the timescale for growing planetary cores is much reduced in this model, and so provides a solution to part of the mystery of how Uranus and Neptune formed. The solution is only partial because one still needs to explain how the Jupiter and Saturn cores accreted large amounts of gas whereas Uranus and Neptune did not.

The biggest remaining question is how realistic is the initial state. For example, it is likely that the violent character of the migration would be damped by gas-drag effects and the self-gravity of the nebular disk, both of which have been neglected in the numerical model. I find it conceivable that a process leading to the formation of a system of planetary cores embedded in a protoplanetary disk would self-regulate to remain ‘marginally stable’ throughout its evolution as the disk mass is gradually exhausted. This means the planets would separate relatively slowly rather than through violent dynamical instabilities as proposed by Thommes et al.1. Further advances in numerical and analytical modelling of the evolution of protoplanetary disks will be needed to address this uncertainty.

Armitage and Hansen2 consider the evolution of giant protoplanetary disks around other stars — disks much more massive than our solar nebula is thought to have been — in which a giant planet forms quickly, possibly by means of a gravitational instability in the disk5. Using computer simulations of disk–planet interactions, the authors find that starting with a marginally stable disk, the embedded giant planet quickly clears a partial gap and also creates a spiral density pattern in the surrounding disk material. The planet accretes mass rapidly through the spiral arms, and when the planetary mass reaches four-to-five times Jupiter's mass the disk rapidly fragments around sites of orbital resonances (these are especially unstable orbits in the disk due to the planet's perturbation). Such ‘triggering’ of planet formation by Jupiter at its orbital resonances has been suggested previously for our Solar System6; the work by Armitage and Hansen represents the first demonstration of this process in a self-consistent computer simulation.

In this simulation, the disk fragments also accrete mass rapidly to become giant planets, with final masses greater than Jupiter. This process yields several giant planets in close proximity to each other, starting out in nearly circular orbits at large distances from the star (several astronomical units). The authors suggest that such a system would subsequently evolve by way of mutual gravitational scattering and eventually result in a planetary system consisting of one or more massive planets in eccentric orbits (although this latter process is not modelled in their simulation). Examples of such systems have been found in recent extrasolar planet searches7,8.

As with the Thommes et al. model, the overriding uncertainty in these computer simulations is the plausibility of the initial state, in this case a single giant planet embedded in a massive protoplanetary disk. Are there reasonable evolutionary paths during star formation that lead to such systems? We also don't yet know whether the disk fragments produced by the simulations actually grow into planets rather than breaking up again; the simulations need to be carried out for longer (with all the necessary physics included) to address this question. Nonetheless, our understanding of planet formation is so limited that plausible new ideas are always welcome.


  1. 1

    Thommes, E. W., Duncan, M. J. & Levison, H. F. Nature 402, 635–638 (1999).

  2. 2

    Armitage, P. J. & Hansen, B. M. S. Nature 402, 633–635 (1999).

  3. 3

    Hahn, J. & Malhotra, R. Astron. J. 117, 3041–3053 (1999).

  4. 4

    Lissauer, J. J. Icarus 69, 249–265 (1987).

  5. 5

    Boss, A. P. Astron. J. 503, 923–937 (1998).

  6. 6

    Patterson, C. W. Icarus 70, 319–333 (1987).

  7. 7

    Marcy, G. W. & Butler, R. P. Annu. Rev. Astron. Astrophys. 36, 57–98 (1998).

  8. 8

Download references

Author information

Correspondence to Renu Malhotra.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Malhotra, R. Chaotic planet formation. Nature 402, 599–600 (1999) doi:10.1038/45123

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.