Princeton University professor Hans-Peter Bunge explains that tectonic plates -- 50-mile-thick slabs of rock -- are still sinking into the mantle beneath South America. Throughout geological history, plate movements have built mountains, dragged up volcanoes and caused earthquakes.

Hans-Peter Bunge, Princeton University geology professor, believes we live above an ocean. "The mantle is, if you will, just another ocean inside of the Earth," he says. No, the planet is not full of water, but extreme pressures and temperatures make the rocky middle layer behave like a fluid on geological time-scales.

Bunge submerges into this ocean with TERRA, mantle simulation software from Los Alamos National Laboratory. His purpose is to recreate the flow of rock that sculpts the mountain-studded continents and digs the trench-lined ocean floor. Mantle work continues through the present age, making planet Earth an unfinished story.

Seeking funds for TERRA expeditions, a team including Bunge sent a proposal to NASA's High Performance Computing and Communications (HPCC) Program in 1995. They would enhance TERRA and distribute it via the World Wide Web so others could study the evolving Earth. NASA Headquarters selected the plan, but 1996 negotiations with the HPCC Earth and Space Sciences Project did not go as expected. Associate project manager Robert Ferraro heard talk about tectonic plates building mountains and slipping in earthquakes. To him, that meant TERRA needed to simulate more than a mantle. "Why not seize the opportunity to add the real shapers of the planet surface, a feat no one had attempted?" Ferraro thought. The team agreed to take on this formidable task and include plates in the final software release.

Map by Paul D. Lowman, Jr., and others, NASA/Goddard Space Flight Center

Tectonic plates underlie the continents and ocean floor. This map reflects tectonic activity over the last one million years. 93K size

"We got so excited about the problem because it is at the heart of our science," recollects Bunge, who was then a graduate student at Los Alamos. "When we woke up out of the enthusiasm and realized what a mound of work we had bought ourselves, we all were surprised and shocked for a moment."

Plate work

While he would pursue Ferraro's challenge for the next four years, Bunge first learned about tectonic plates as a physics freshman at the University of Tuebingen in Germany. During that year he went to Italy for a seminar on Mediterranean Sea geology. It showed Bunge that plates underlying the continents and ocean floor closed "a very large ocean that separated Africa from Europe. In the process of that closure, the Alps formed, and a little remnant of that ocean called the Mediterranean Sea was left over."

Hans-Peter Bunge travels back more than 100 million years to recreate the plate motions that formed the Rocky Mountains and other surface geology. 38K size

The freshman was hooked. The seminar's vividness "invoked in me a passion for the geological sciences, in particular for plate tectonics," he says. Adding a second major in geology, Bunge did field work, mapping the Alps his instructors had described.

A few months after his 1988 graduation, Bunge went to South Africa, "a country that represents the geology of the southern continents, which is different from the geology of the northern continents." The field geologist researched gold extraction for a national laboratory.

Besides "seeing the world in all its richness and facets," working in the Alps and South Africa taught Bunge to "pay attention to the data, the actual observations in the field. That is very important for someone who works on theoretical models, because it is quite easy with a model to forget all the data you really should address."

Rocks get stronger

Bunge was unaware of supercomputer models until graduate school. A Fulbright Fellowship funded his study at the University of California, Berkeley, for the 1989-90 academic year. The plan was to return to Germany, but the faculty asked Bunge to stay.

A professor's lecture on geophysical modeling hit Bunge "almost like a revelation." Soon after, the student decided to simulate the mantle for his Ph.D. project. Thesis advisor Mark Richards arranged a research assistantship at Los Alamos, a computing mecca in the New Mexico mountains. There, Bunge met John Baumgardner, creator of the TERRA software that would become key to the NASA research.

Scientists compared results from seismology (a) and simulations (b through d) to test theories for the Rocky Mountains' formation in an uplift called the Laramide Orogeny. Only simulations that include the Farallon Plate scraping beneath the surface for 1,000 miles (c and d) match the current Earth pattern. Colors mark the velocity of rock flow; blues are faster than average, and reds are slower than average. 55K size

Baumgardner asked Bunge to reprogram TERRA for parallel supercomputers. Harnessing several hundred processors would speed the code so it could efficiently cover the eons of geological history. It also would enable carving the virtual mantle into many boxes to simulate small-scale phenomena. The idea proved prescient, for today Bunge can avow, "parallel computers almost overnight have increased our computational ability by a factor of 100 to 1,000."

TERRA reaped the benefits with NASA HPCC support. Since funding started in August 1996, the code reached a high of 121 billion floating-point operations per second. Bunge also modified TERRA to use the Message Passing Interface, which is popular communications software running on everything from Linux clusters of personal computers to the CRAY T3E.

With newfound performance in hand, the team could now apply TERRA to NASA's tectonic plate challenge. The effort hinged on the notion that plates are the cold surface of the mantle. These 50-mile-thick slabs ride atop the rocky ocean like barges, propelled inches per year by the underlying flow. Team members from Harvard to Hawaii wrestled with how the mantle builds plates and carries them along. Bunge and Baumgardner focused their energies on plate rock.

Volcanoes flare and mountains rise where plates sink into the mantle, so the plate rock studies began there. In current geological time, plate sinking limits itself to two snake-like paths framing the Pacific Ocean. Plates are particularly active under Alaska, the Pacific Northwest, western South America and Japan. In TERRA simulations, rock sank at many points around the globe. Bunge consulted gravity data to learn what the simulations were missing.

When pursuing his own Ph.D., Berkeley's Mark Richards discovered gravity to be a good indicator for rock strength inside the Earth. Satellite measurements of gravity "uniformly suggest that the deeper mantle is stronger than the upper mantle," Bunge says. The mantle depths' higher pressures seem to strengthen rock. Arduous months spent writing and testing code to mimic this behavior paid off: TERRA reproduced the plate sinking pattern under today's Earth.

Image by Hans-Peter Bunge, Princeton University

At least 64 million years ago, the Farallon and Kula Plates spread apart beneath the Pacific Ocean as they began to sink beneath North America. The shaded area indicates uncertainty of the spreading location. 29K size

Programming mathematical rocks that could gather strength opened the door to modeling plates' dual identity. Especially at the mantle surface, a plate is at once strong and weak, depending on where one looks. "Plates are very rigid inside yet separated by faults that accommodate their relative motion," Bunge explains. Having captured plates' "extreme variation in strengths," Baumgardner is testing the feasibility of earthquake prediction with another software code. [See Drilling through the Earth, INSIGHTS, August 1999.]

TERRA was growing into a capable tool for studying plate tectonics. The simulated rocks varied in strength at the right places. Thanks to colleagues' creativity, the TERRA mantle spawned and moved plates. The time had come to apply the software to a big problem -- rebuilding the globe from the inside out.

In geological time, the world overturns itself every 150 million years or so. That is how long it takes for rock from the bottom of the mantle to reach the top and vice versa. Bunge contacted University of Texas seismologist Stephen Grand for clues about the recent past.

Seismology records sound waves as they travel through the planet after earthquakes. Over a decade, Grand had compiled the speeds of 20,000 sound waves, which amount to an X-ray movie of the up-and-down rock flows. "Peter is trying to simulate those movements in the Earth," relates Grand, a professor of geological sciences at the Austin campus."Seismology says whether his simulations are right or not."

Worldwide, TERRA got things right. For instance, plate sinking under the Pacific matched seismological data. [See INSIGHTS, August 1999.] Next up was a region that has been a thorn in the side of plate tectonics theory.

Image by Hans-Peter Bunge, Princeton University

Circles on this topographic map estimate the eastern extent of regions affected by the scraping of the Farallon Plate under North America. 41K size

Scaling the Rockies

The new plate riddle arose from Bunge's fall 1998 move to Princeton as assistant professor of geosciences. As Bunge daily strolled through gothic Guyot Hall, a giant globe on the first floor bothered him. It was the Rocky Mountains. Mountain ranges spring from where plates sink or collide, but the Rockies are 1,000 miles from the nearest plate boundary.

"Other mountains reflect the standard view of plate tectonics," Bunge says. "The Rockies defy that notion. There is no reason for them to be where they are if tectonics theory is right."

Sometimes theory must bend. A plate normally sinks at a 45-degree angle. The Farallon Plate that built the Rockies started out that way. It plunged beneath the North American Plate, dragging up volcanoes to form the Sierra Nevada in what is now California. Then nothing. No surface activity until the Rockies pop up far, far to the east, which struck Bunge as "a very funny interlude."

At an overlook in the southern Rocky Mountains, John Baumgardner describes the TERRA software used to simulate the Farallon Plate that built the mountains. 55K size

Reconstructing this lull is not easy, for the Farallon Plate is gone. It disappeared into the mantle long ago. Amazingly, Stephen Grand's seismic compilation recovered Farallon. The plate lies under the East Coast of the United States. "I consider it one of the major triumphs in seismology to have shown in such great detail an oceanic plate that is entirely now inside of the Earth's interior," Bunge revels, "showing it to us all the way from the bottom of the mantle, 5,000 miles below our feet."

Bunge mounted a simulation to test Farallon sinking scenarios. To follow one plate, one must compute the whole world. This was a task for NASA HPCC's CRAY T3E at Goddard Space Flight Center. With so few powerful machines available, "being able to use one of these computers makes a huge difference," Bunge stresses. Solving equations in 10 million computational boxes, the simulation consumed 128 T3E processors for four days.

Of all the simulated cases, only a remarkable process gets the Farallon Plate beneath the East Coast. Somewhere between 40 and 80 million years ago, it wandered inland from the Sierra Nevada. Farallon scraped under the bottom of North America for 1,000 miles before -- at last -- diving into the mantle. The Rockies rose in its wake.

A 1970s era theory for the Rockies' unusual location seemed plausible, even likely. The journal Nature published the findings in its May 18 issue. In Bunge's view, "we're trying to put the pieces of the puzzle dynamically together. The seismologists are for the first time really showing them to us. In a computer model we can find out why they are there."

Such collaborative strides prompt him to be optimistic about new discoveries in the Earth sciences. And NASA helped sound the call with its request to add plates to TERRA. "Without that extra push, we would not have gotten as far as we have," Bunge says.

Credits for Insights Magazine go to the following people along with the writers and photographers who contribute to each issue and the researchers and specialists whose material is highlighted:

Program manager: Dr. Eugene L. Tu
Deputy program manager: Dr. William R. Van Dalsem
Editor, writer, photographer and designer: Judy Conlon
Web designer and programmer: Sue Cox

Insights was published by the HPCC Program office and produced by Raytheon contractor staff at NASA Ames Research Center.