2005 Computerworld Honors Program

	Simulation to Predict Comet-Planet Collision Pittsburgh Supercomputing Center / University of Chicago Pittsburgh, PA USA Year: 1995 Status: Finalist Category: Science Nominating Company: Cray Research, Inc.
	Advance simulation of the impact of a comet on Jupiter guided scientists in planning their observations and is aiding them in the analysis of data about the event.
	Fireworks on Jupiter All telescopes pointed to Jupiter this summer as Comet Shoemaker-Levy 9 crashed into the giant planet. As people day-by-day followed the news of Shoemaker-Levy 9's fiery death, a basic scientific question entered public consciousness: Could a similar large comet or asteroid someday come hurtling toward Earth? The short answer is yes. It happened 65 million years ago, with disastrous consequences for the dinosaurs, and the odds are, say astronomers, it will happen again within the next 100 million years. What are the chances human life could survive such a calamity? This summer's event gave scientists an unprecedented chance to observe doomsday from a safe distance and learn from it. "It was a unique event," says astrophysicist Mordecai-Mark Mac Low. "It's the first time we've been able to predict a large planetary impact and then observe it, and it will probably be the only time in our lifetimes that an impact this large occurs." In preparation for the comet crash, Mac Low, a postdoctoral researcher at the University of Chicago, used the CRAY C90 at Pittsburgh to forecast what would happen. "Basically, I was trying to predict the results of the impact so that observers could plan for the event. I looked at things like how bright the flash of the explosion would be and how much material from beneath the Jovian clouds would be lifted above them where it can be observed." Observers worldwide relied on Mac Low's predictions to help plan for the event, and the results suggest that Mac Low's simulations provide a complete description of what actually happened on Jupiter. Using the C90, Mac Low ran simulations that showed the comet would penetrate less deeply and explode more violently than other models predicted. Perhaps the most important result, at least in terms of forecasting the effect of Earth impacts, is that the simulations agree with a mathematical theory called "the pancake model." Forecasting a Big Splash In collaboration with Kevin Zahnle of NASA Ames Research Center, Mac Low ran three types of simulations: (1) a comet fragment entering Jupiter's atmosphere until it exploded, (2) the initial fireball from the explosion, and (3) what happened in the atmosphere after the initial fireball. The researchers used ZEUS, a program developed at the Laboratory for Computational Astrophysics of the National Center for Supercomputing Applications for modeling the gas dynamics of astrophysical phenomena such as the violent shock waves from supernovae. For the last six years, Mac Low has used ZEUS to study interstellar gas dynamics, and he realized he could apply the same method with relatively minor changes to simulate a comet crashing into Jupiter's atmosphere -- basically by shifting the scale from light years to kilometers. "The physics," says Mac Low, "is virtually identical. It's only the details of the composition of the atmosphere that change, and of course the length scales, time scales and density changed -- by 20 orders of magnitude in some cases." Their simulations predicted that the flash from the explosion would last about a minute, with about as much brightness as the sunlit side of Jupiter. Unfortunately for Earth observers, the comet crashed into the back side of Jupiter. Mac Low's calculations suggested, nevertheless, that the fireball would be bright enough to be seen from the NASA spacecraft Galileo or with Earth telescopes as a reflection off one of Jupiter's moons. The strongest prediction from the simulations, the one Mac Low was most confident of, had to do with how deep the comet would dive into Jupiter's atmosphere before exploding. Other models showed it going in hundreds of kilometers, so that its energy is absorbed relatively slowly -- what one researcher called a "soft catch." Mac Low's results showed, however, that impact with Jupiter's atmosphere would rip the comet apart more quickly and violently, with a fierce explosion, after penetrating only about 110 kilometers below the cloud tops. The resulting plume of superheated debris, according to Mac Low's study, would shoot hundreds of kilometers above Jupiter's layered clouds, giving astronomers a good chance to observe the after-effects and, in the process, learning more than we know now about the composition of Jupiter's atmosphere. Inferences from observational data indicate that this prediction was essentially accurate. Mac Low's simulations used a computational grid finer than the other models, suggesting that his results more closely approximated reality. As a check on this surmise, Mac Low ran his code at much lower resolution and got a result similar to the other models. "At low resolution we got one result," says Mac Low, "and at high resolution we got another, and as we continued increasing resolution the result stayed the same." The high degree of detail in Mac Low's study -- made possible by the CRAY C90 -- gave a reasonable basis for astronomers to be optimistic that they would have a good show to watch in July. The Pancake Model The most important result of Mac Low's study is his finding that the numerical simulations agree well with an analytical model called the pancake model. This model assumes that once the aerodynamic force from the comet's impact into the atmosphere overcomes its material strength the comet flattens like a pancake, which greatly increases drag -- essentially stopping the comet in its tracks. If these results prove to be realistic -- and this summer's event could help determine that -- the pancake model can be used to predict what will happen from comet and asteroid impacts on Earth and other planets. "One of the scientific issues we're hoping to get a handle on in terms of Earth impactors," says Mac Low, "is how big a rock do we need to worry about? One of the motives for modeling this impact is to see if we can do reasonably accurate predictions. If we can, we can start talking about how well the Earth's atmosphere protects us."
	These computations provided an immediate scientific benefit by helping astronomers to plan for the summer 1994 comet crash. Mac Low's computer simulations were more detailed than others done prior to the event, and they indicated that, even though the comet was aimed at the far side of Jupiter, the resulting explosions would be likely to produce effects observable from the Earth. They also indicated it would be visible from the Galileo spacecraft. Furthermore, they gave observers a coherent picture of the expected effects, allowing them to plan appropriate series of observations for each impact. Now that the impact has occurred, Mac Low's simulations are proving to be the most complete description of the explosion. They are able to explain the major observational results in a coherent story. These results include the presence of elemental sulfur above Jupiter's visible clouds after the explosions. The models also appear to yield the size and mass of the comet fragments, giving unique insight into the structure and composition of comets. Mac Low's work is playing a central role in gleaning knowledge of both Jupiter and comets from this spectacular event.
	The results of this research could not have been obtained without using the techniques of computational gas dynamics running on powerful supercomputers. Mac Low's simulations gave information that couldn't be obtained through laboratory techniques. His work exemplifies the value of computational science, representing a third kind of scientific model alongside the traditional analytical and experimental approaches.
	Mac Low's modeling used a scientific program, ZEUS, developed at the Laboratory for Computational Astrophysics of the National Center of Supercomputing Applications, where Mac Low has been a collaborator for the past seven years. He has contributed to the development of this program, which has proven itself as a tool in modeling astrophysical gas dynamics problems such as the violent shock wave produced by a supernova. ZEUS is freely available to researchers, and is generally recognized as one of the most versatile programs available in astrophysical gas dynamics. Mac Low's insight, once he learned of the predicted comet crash, was to realize that the problems he had been computing with ZEUS could be adapted with relatively minor changes to simulate a comet crashing into Jupiter's atmosphere. Although a supernova is an immensely larger and more violent astrophysical event than a comet crash, Mac Low understood that the physics were similar and that only the scales of time, length and density needed to be changed, along with adding the details of the Jovian atmosphere. His resourcefulness in seizing this opportunity has proven to be of great benefit to astronomy in planning for and interpreting this unique event.
	xe 4s computational results appear to give a complete description of the explosions resulting from the comet fragments entering Jupiter's atmosphere. This suggests that the analytical model underlying his approach -- called "the pancake model" -- can be relied on to predict other similar events. Even more importantly, it can be used in research to predict the effect of asteroid and comet impacts with Earth's atmosphere. Partly as a result of this summer's comet crash, planning is now underway -- with discussion in Congress -- for NASA or another agency to develop a predictive capability that could provide a measure of protection from catastrophic events on Earth such as the one that wiped out the dinosaurs 65 million years ago. Mac Low's work, in collaboration with Kevin Zahnle, will help to determine the feasibility of such a program.
	Mac Low's first obstacle was a technical issue: how to properly model the entry and explosion of comet fragments. Other attempts to model this problem suffered badly from underresolution, giving incorrect results. Because Mac Low had experience working on physically similar problems in interstellar gas dynamics, using a well developed code (ZEUS) and high- performance computing (the CRAY C90), he was able to quickly apply higher levels of resolution to the problem and get answers that agree with analytical theory and, apparently, with observation. The second obstacle was time. Results needed to be generated quickly in order to use them in planning observations of the comet impacts. Major ground-based and space-based observatories normally develop their schedules months in advance, which meant that plans for observing the impact had to be finalized during winter 1993-94, only six to eight months after Comet Shoemaker-Levy 9 was discovered. This time pressure meant that normal channels for funding research had to be shortcut. The availability of supercomputer resources (at Pittsburgh Supercomputing Center) and funding nominally devoted to other projects proved vital to producing results in the required time frame.