Scientists drilling more than 2 miles deep into the San Andreas Fault have discovered underground patches of talc, nature's softest known mineral, that could help explain the absence of sharp earthquakes where the fault is "creeping."

The surprising result of what one expert calls "a literally groundbreaking project" adds a new understanding to the strange behavior of the infamous fault's 90-mile segment between Parkfield in southern Monterey County and San Juan Bautista in San Benito County.

The finding also marks a significant success for a drilling effort that was once only a dream of earthquake scientists but is now a major international seismic research undertaking known as SAFOD, the San Andreas Fault Observatory at Depth.

The team of geophysicists and seismologists running the project reported Wednesday that their difficult core drilling effort is now within 15 feet of crossing one of the most active traces of the fault for the second time.

Once there, the scientists will be able to measure and analyze clusters of tiny, deep quakes, using seismic instruments that the drillers are delicately maneuvering into a narrow, sharply slanted hole now penetrating the target region.

Already the drillers are retrieving core samples of rock, 4 inches in diameter and some as long as 25 feet, from the fault's deep region that the scientists expect will give them a new understanding of how the Earth's crust is behaving in the San Andreas zone.

"It's an extremely tough, challenging job, and we're all excited right now because the cores we're bringing up we can study just like precious moon rocks," said William Ellsworth of the U.S. Geological Survey's earthquake hazards team in a telephone interview from the drill site some 9 miles northwest of Parkfield.

Diane E. Moore and Michael J. Rymer, two Geological Survey scientists unexpectedly found the slippery, watery, reddish-brown talc embedded as large veins in chunks of a mineral called serpentenite that the drill team brought up to the surface from an earlier crossing of the fault's seismically active trace.

According to Moore and Rymer, aerial magnetic surveys of the Parkfield region have revealed that a flat slab of the talc-containing rock a mile wide and more than 30 miles long lies some 2 miles deep on the fault's northeast side, where the temperature is roughly 233 degrees Fahrenheit.

And because the talc is so soft - the mineral is used in some baby powders - Moore and Rymer wondered if it could account for the phenomenon of fault creep that has caused the sides of the San Andreas in that region to slide past each other at an inch a year for millions of years rather than triggering abrupt earthquakes.

At the Geological Survey's Rock Mechanics Laboratory in Menlo Park, Moore and Rymer analyzed the talc using a variety of instruments, including a scanning electron microscope, and found that at various temperatures, rock faces containing the talc slid easily past each other rather than fracturing - a striking example of fault creep right there in the lab.

The result of their work is being reported today in the journal Nature.

In an accompanying commentary, Christopher Wibberley, a geologist at the University of Nice in France, said he found their report an intriguing explanation for the apparent weakness of the San Andreas Fault where it creeps gently in the Parkfield area rather than sticking until increased strain launches strong earthquakes.

Ellsworth and Stephen Hickman of the Geological Survey and Mark D. Zoback of Stanford, who are directing the SAFOD drilling project at Parkfield, reported that the fault's creeping motion is so powerful that it has already deformed the drill hole's steel casing deep underground and caused it to twist and bend by a few inches - making it extremely difficult to continue drilling down the hole's slanted course.

"Last month, the San Andreas was fighting back," Zoback said, but the diamond-studded drill bit is now moving steadily downward.

All along this segment of the fault, but deep down at the slanted end of the SAFOD hole, small microquakes are being recorded even while the creeping motion continues, for the talc slabs are probably not continuous.

Zoback drew an analogy: "It's like pushing a box across an oily floor on which sand grains are scattered here and there," he said. The box slides easily in spots, then sticks until the pressure of the push unsticks it and it lurches strongly for a moment - that's the quake - before sliding again, he explained.

The network of seismometers throughout the Parkfield region records occasional small quakes with magnitudes around 2 along that 90-mile portion of the San Andreas, but the creeping motion dominates the section's long history of slow movement between Parkfield and San Juan Bautista, Zoback said.

South of Parkfield, however, where talc may not exist underground and seismic stress builds up ceaselessly, the history of the San Andreas is far different: Sharp earthquakes with magnitudes around 6 have struck there at almost regular intervals, roughly every 22 years - in 1857, 1881, 1901, 1922, 1934 and 1966.

Because of that long series, scientists have monitored the fault's behavior with a large array of instruments, and in 1985 they predicted that another magnitude-6 quake might strike the area before 1996. It didn't happen, but in 2004 a sharp magnitude-6 quake did hit about 7 miles southeast, while a magnitude-5 aftershock occurred four minutes later.

All has been quiet on the Parkfield front since then - and while the drillers keep drilling, the fault keeps creeping and the scientists keep studying their drill cores to learn just what is going on far underground that will help them understand how and why earthquakes behave the way they do.