Projects in Scientific Computing

Download the PDF version of this article as it appeared in Projects in Scientific Computing, 2003.


Imagine that you’re nearsighted. You get new glasses and look out your window and, all of a sudden, the backyard is vividly in bloom with crocuses, daffodils and marigolds you couldn’t see the day before.

For astronomers who gaze deeply into the Milky Way, it’s like that. For millennia, as far as we knew, ours was the only solar system, the Sun the only star, among billions visible, that had planets. Now suddenly, within the last eight years, our galaxy is blooming with planets.

Since 1995, astronomers have found more than 100 extrasolar planets, and almost certainly that’s just the beginning. The new pair of glasses is Doppler spectroscopy, a method that infers the presence of planets, which only faintly reflect light, from the wobble their gravitational tug induces in their host star.

Thomas Quinn.

Thomas Quinn, University of Washington

The method has been so successful, especially as employed by the California team of Geoffrey Marcy and Paul Butler, that as Marcy said three years ago, “Planet hunting has morphed from the marvelous to the mundane.” With a data sample that’s suddenly 10 times larger, it’s not surprising that accepted ideas about planets, including how they come into being, are suddenly opened up for rethinking.

Recently, a research team led by University of Washington astrophysicist Thomas Quinn revisited this old question. How do planets form? Using a powerful new tool — LeMieux, PSC’s terascale system — and advanced software called GASOLINE, developed by Quinn’s group, they simulated the process by which “gas giant” planets, like Jupiter and Saturn, coalesce into solid mass from the swirl of gas that surrounds a young star.

Because of LeMieux, they were able to include 10 times more gas particles in their simulation than previous similar work. Their results with this increased resolution — reported in SCIENCE (Nov. 29, 2002) — mount a convincing challenge to accepted thinking. “We used a new model of planet formation,” said Quinn, “that couldn’t adequately be tested without this kind of computing power, and we found that these giant planets can form in hundreds of years, rather than the millions that the standard model predicts.”

Planet Formation Isn’t Special

Protoplanetary disc.
Click Image to Enlarge

This image from simulations by Quinn and colleagues shows how a protoplanetary disk surrounding a young star begins, in a relatively short time, to fragment and form gas giant planets with stable orbits.

For at least 20 years, the textbook view has been that it takes a long time to cook a planet. Little by little — says the old recipe — clumps of solid mass form within the gas, dust and ice that whirl like pizza dough around a young star. Gravity gathers bits of dust, which merge into boulder-sized chunks, which in turn coalesce into bigger rocks.

In about a million years — according to this recipe — these rocks become planets, and over the next few million years, gas from the disk settles around some of these young planets to grow Jupiter-size gas giants, which have most of their mass in a gaseous envelope surrounding the solid core. According to this view, called the “core-accretion model,” it would take as much as ten million years for a gas giant planet to form.

The core-accretion model first came up for rethinking, says Quinn, in the early 90s, when the Hubble Space Telescope was able to see these gaseous protoplanetary disks (or proplyds). Astronomers realized that these disks can’t last for a long time, roughly a million years, because the gas is rapidly cooked away by radiation from nearby stars.

Still, core accretion held up as a theory, because — so the thinking went — massive planets like Jupiter and Saturn are relatively rare. But the Marcy and Butler team carefully surveyed about 1,000 stars and showed that roughly one out of 10 has a planet, probably more, with most of them gas giants, from roughly the size of Jupiter to ten times larger.

“Now,” says Quinn, “we see that planet formation is not particularly special, and this doesn’t jibe with the standard model. If a gas giant planet can’t form quickly, it probably won’t form at all.”

Fast Cooking with Gravitational Instability

Another way for gas giants to form was proposed in 1997 by astrophysicist Alan Boss of the Carnegie Institution. Using computational simulations, Boss found that gravity could act suddenly to form planet-size clumps out of instabilities in the swirling gas disk, and that the clumps formed this way were massive enough to build up a gas envelope. This model, called gravitational-instability, in theory would form a gas giant in much less time.

Protoplanetary discs of different mass.
Click Image to Enlarge

These face-on snapshots from 3D simulations by Quinn and colleagues show density (increasing from dark blue to magenta, red and yellow) of two differently constituted protoplanetary disks, at 160 years (left) and 350 years (right) of evolution. The two upper panels represent a disk containing about 10 percent less gaseous mass than the lower. Both disks revolve around a solar point mass with density shown to a radius of 20 times the Sun to Earth average orbit (about 93 million miles).

"After about 150 years," says Quinn, "the disks develop trailing spiral arms. In the lighter disk, the spiral grow in amplitude up to about 300 years, and then settle into a nearly stationary pattern. In the more massive disk, a two-armed mode grows up to the point, where after about 200 years, fragmentation occurs along the arms, and more than one distinct clump appears."

Until the work by Quinn’s team with LeMieux, however, no one had been able to do the simulations with enough precision and length of time to make the case effectively. “The main criticism people had,” said Quinn, “is that this model wasn’t ready. Nobody was making predictions with it. But that’s because they didn’t have enough computational horsepower.”

The major obstacles in such a calculation have to do with two factors. One is the inherent non-linearity of gravity. “As mass starts to collect, gravity gets stronger and stronger. The other factor is dynamic range. “We have to go from gas densities in the disk of less than a microgram per cubic centimeter,” says Quinn, “to planetary densities of a gram per cc, a million-fold increase in density.”

Quinn’s team had spent about a decade developing software, called GASOLINE, designed to simulate cosmology. Within the last four years, they added the ability to accurately capture the evolution and structure of galactic gas. Quinn has long been interested in planet formation and realized that the gas dynamics capability of GASOLINE would also work, with relatively minor modification, to simulate planet formation.

“Developing parallel software,” notes Quinn, “is a non-trivial task.” In this case, the software was ready at a time when the gravitational-instability model presented itself as a challenge, and the confluence of software and a powerful system, LeMieux, with a problem that needed to be effectively addressed was propitious.

Running on 32 LeMieux processors, over about 30,000 hours of computing, the simulations traced the gas dynamics using a million particles to represent the protoplanetary gas disk, about 10 times more resolution than had previously been attempted. The result was a distribution of masses and orbits comparable to observed extrasolar planets, formed in only hundreds of years, not millions.

Authors of the article in SCIENCE, along with Quinn, are Lucio Mayer, a post-doctoral researcher with Quinn who’s now at the University of Zurich, Joachim Stadel, University of Zurich, and James Wadsley of McMaster University.

“This work shows,” says Quinn, “that gravitational instability can actually form self-gravitating protoplanets and that long-lived systems with masses and orbits consistent with those of extrasolar planets arise.” In future work, Quinn and colleagues plan to carry out more high-resolution simulations and to explore different conditions of the gaseous disks and also to add a more sophisticated approach to modeling the heating and cooling of the gas.

Thomas Quinn, Univeristy of Washington

Terascale Computing System


Related Material on the Web:
N-Body Shop at the Univeristy of Washington.
Planets May Form Faster than Scientists Thought, PSC News Center.

Lucio Mayer, Thomas Quinn, James Wadsley, Joachim Stadel, "Formation of Giant Planets by Fragmentation of Protoplanetary Disks," Science 298, 1756-59 (2002).

Alan P. Boss, "Giant Planet Formation by Gravitational Instability," Science 276, 1836-39 (2002).

Michael Schneider, Pittsburgh Supercomputing Center

Web Design:
Sean Fulton, Pittsburgh Supercomputing Center

Revised: June 3, 2002