Hitting a barrier deep within the planet.

Earthquake waves can reduce a city to rubble in seconds. But when the same vibrations pass deep through Earth's interior, they lose their destructive power and turn into valuable scientific data. By studying thousands of distant tremors, two seismologists have now uncovered important clues concerning Earth's internal recycling system.

For two decades, geoscientists have debated what happens to the ocean floor when it subducts, or dives down into the Earth's mantle, the region whence it came. While some researchers believe the oceanic plates sink all the way to the top of the iron core, others think the slabs of ocean floor get trapped in the upper mantle. The new data cannot resolve the question of how low slabs go, but they do reveal that the boundary between the upper and lower mantle presents a formidable barrier to sinking slabs.

Peter M. Shearer and T. Guy Masters of the Scripps Institution of Oceanography in La Jolla, Calif., analyzed waves from earthquakes that reflect off the 660-kilometer-deep boundary between the upper and lower mantle. By analyzing 3,000 seismograms from shocks around the world, they produced the first global map of the boundary's topography, which they discuss in the Feb. 27 NATURE.

The mantle, which makes up 84 percent of Earth's volume, extends from a depth of 2,900 kilometers all the way up to within 70 kilometers of the surface. The boundary between the upper and lower mantle is defined by a discontinuity in the speed of seismic waves. As earthquake waves pass down through the mantle, they speed up abruptly at a depth of 660 kilometers. Geoscientists think this acceleration arises mainly from a physical change in the crystal structure of mantle rock. High-pressure experiments indicate that the rock above the boundary has a more open crystal configuration, whereas the crystals in the lower mantle are sequeezed into a denser arrangement. Rocks in the upper mantle might also contain chemicals that differ from those in the lower mantle.

Shearer and Masters report that their topographic map shows broad depressions in the seismic boundary beneath plate subduction zones. The low spots measure about 1,500 kilometers across and extend as much as 30 kilometers below other parts of the boundary. The clearest depression appears northwest of the Kuril-Kamchatka region, where the Pacific plate dives under the Asian plate.

These results suggest that the sinking Pacific slab does not pass cleanly down into the lower mantle, according to the researchers. Instead, they believe, the plate bends at the boundary and moves horizontally across it, forming the depression that appears on their map.

"It's clear that there is resistance to slab penetration at the 660-kilometer boundary," Shearer says.

It is not clear, however, what happens after the slab deflects horizontally and slows its descent. The cold oceanic rock could form a thick blob that continues to sink deep into the lower mantle at a much slower rate, or it might just pile up along the boundary and remain stuck in the upper mantle.

"Its ultimate fate is still in question," says Thorne Lay, a seismologist at the University of California, Santa Cruz.

The answer has important ramifications because it indicates what kind of heat engine drives the motion of the lithospheric plates forming Earth's outer shell. If slabs sink all the way to the core, then the mantle resembles a pot of boiling soup, with hot material rising from bottom to top and cold material sinking from top to bottom. Conversely, if subducting slabs do not descend much below 660 kilometers, then the mantle would resemble a double boiler, with very little mixing across the boundary between lower and upper mantle.

While seismologists in the 1970s generally thought that subducting slabs could not make it into the lower mantle, researchers discovered seismic evidence in the late 1970s and 1980s suggesting that subducting plates in fact penetrated deep into the lower mantlel. In the last few years, seismic studies have grown equivocal: Some show deep slabs, while others show deflected ones.

The data collected by Shearer and Masters provide a new approach to the perennial problem, Lay says.

"Trying to understand the deep structure of the Earth and the processes in the Earth is a formidable and challenging problem. We have better knowledge about the interior of the sun that about the interior of the Earth," he notes.