WHEN Tom Henyey lowered a set of electronic thermometers into a California drill hole, he wasn鈥檛 expecting a mystery. It was 1965, and Henyey, then a graduate student at the California Institute of Technology in Pasadena, was simply looking for easy data for his thesis. The hole was only a few kilometres away from the San Andreas Fault, where two of the Earth鈥檚 giant tectonic plates have been grinding past each other for several million years. Geologists were sure that the heat generated as rock scraped against rock would be flowing a few hundred metres underground, and Henyey just wanted to measure it. There was only one problem 鈥� the heat wasn鈥檛 there.
Almost three decades and more than a hundred experiments later, geologists are still baffled. Everything they know about rock predicts that the San Andreas should be as rough as sandpaper. Yet it seems to be lubricated, slipping along at least ten times more easily than anyone imagined. 鈥淲e know we鈥檙e missing something important,鈥� says Steve Hickman of the US Geological Survey (USGS). 鈥淚t鈥檚 like being an astrophysicist without knowing quantum mechanics or a biologist before the cell鈥檚 nucleus was discovered.鈥�
The answer to this conundrum is not just of academic interest. Hickman says that the explanation for the coolness of the San Andreas, when it is found, could either open up a new science of earthquake prediction 鈥� or put an end to all hope of forecasting quakes.
Advertisement
The San Andreas Fault spawned two of the costliest earthquakes the US has ever experienced: the San Francisco quake of 1906 and the Loma Prieta quake in 1989, which caused damage worth billions of dollars. No one doubts that there will be another Big One sometime in the future. Moreover, it is not only the San Andreas that slides too easily: the same seems to be true of any of the world鈥檚 major faults, such as the Alpine fault in New Zealand and even of quite different geologic animals 鈥� subduction zones, for instance, where one tectonic plate is pushed under another.
Satellite photos show the San Andreas Fault running like a scar along almost the whole length of California and into Mexico. Over the past 5 million years, the North American tectonic plate has been ambling southeastwards on the east side of the divide, while on the opposite side the Pacific plate moved northwestwards. In that time, the two sides have moved several hundred kilometres relative to each other. In some parts of the fault, stress builds up where a snag holds the two sides of the fault still for decades or centuries. Then something gives, releasing the stress and causing an earthquake, a type of motion called strike-slip.
The wrong rocks
Whatever the secret of the San Andreas, it must involve the physics of friction. One important element is the intrinsic roughness of the rock itself. Laboratory measurements on rocks from the fault suggest they are what friction specialists call 鈥渟trong鈥�: as they slide against each other they should provide a strong braking force holding back the tectonic plates.
But when the heat measurements came to light, many scientists wondered if they had simply been looking at the wrong rocks. The friction values that geologists relied on in the 1960s were based on run-of-the mill varieties of rocks such as limestone, sandstone, quartz and granite. But weaker materials -types of clay 鈥� had already been seen in the fault zone and in samples retrieved by drilling deep into the plates. On the surface, these rocks have far less friction than more typical varieties, so maybe they were lubricating the fault. But then this theory hit a snag. In 1992, Jim Byerlee, a specialist on the friction of rocks at the USGS in Menlo Park, decided to test whether these materials would still be slippery 10 or 15 kilometres beneath the surface, which is where earthquakes start. Byerlee squeezed samples of the clays in laboratory presses to the thousands of atmospheres of pressure they would encounter deep inside the Earth. 鈥淓ven the weakest of these materials transforms at those pressures,鈥� he says. 鈥淭he friction becomes as high as ordinary rock.鈥�
This does not entirely rule out the clay hypothesis 鈥� Byerlee admits that some strange chemistry in the fault, different from what happens in the laboratory, might yet allow the clay to stay greasy. But many scientists turned instead to other explanations. In particular, they began to wonder whether the answer lies not in the type of rock, but in the conditions the rocks find themselves in. After all, the roughness of the rocks is not the only thing responsible for the frictional heating. The pressure pushing the plates together is also a factor: the more firmly they are pushed against each other, the hotter they will get.
But what could be reducing the pressure on the fault? The answer that first jumped to mind was water, which could find its way to the fault from the surface or be squeezed out of the rocks into the fault zone. Pockets of water at a high enough pressure could take some of the load off the rocks, making it easier for them to slide.
In the summer of 1994, geophysicists led by Casey Moore and Andy Fisher of the University of California at Santa Cruz, discovered a striking example of what water can do. They were studying a subduction zone near Barbados where the South American plate pushes under the Caribbean plate. Because the fault lay beneath the ocean, Moore knew it was taking in plenty of fluid. So the researchers drilled into the fault, and capped off the hole with a pressure gauge.
They discovered that the water pressure in the fault was incredibly high. So high, in fact, that it was actually forcing the plates apart, just as the air under pressure in a tyre lifts a car off the ground. 鈥淭he water is literally lifting off the roof,鈥� says Moore. But the tyre is probably leaky, he says. As the fault moves, the pockets of rock holding the water will rupture and release it. Unless new water flowed in to make good the loss from the spillage, the pressure would drop drastically.
This makes it unlikely that water could be lubricating the San Andreas, and other faults on dry land. True, water could drip into the fault from the plate or percolate up from the mantle underneath the plates. But measurements of the chemistry in fault rocks made in 1993 by Fred Chester at Saint Louis University and by Jim Evans at Utah State University in Logan show that the chemical mixture of the fault is very similar to that of the surrounding rock. 鈥淲e seemed to be dealing with a closed system with limited flow,鈥� says Chester. The likeliest source of the water is the dehydration of rocks near the fault 鈥� a strictly limited resource.
There are a few possible escape routes from this impasse. One way to minimise the amount of water needed was suggested back in 1973 by Art Lachenbruch of the USGS at Menlo Park and Rick Sibson of the University of Otago in New Zealand. Water would not escape so easily from the fault zone if it was at a lower pressure, Lachenbruch and Sibson reasoned. But when an earthquake begins, friction from the rocks could heat the water and raise its pressures until it triggered the lubrication mechanism.
Another possibility is that the water is irrevocably trapped in the rock, unable to escape. Experiments in 1993 by Byerlee and theoretical work by Mike Blanpied at USGS and Norm Sleep at Stanford back up this idea that the fluid could act as a self-healing glue. They reason that the heat generated by friction between the plates during a quake could take the pressurised water deep underground to a high enough temperature to dissolve minerals such as quartz, especially if they are crushed during the quake. Then, after the quake, as fluid rushes towards new cracks, the pressure and temperature will fall, causing the quartz to drop out of solution and seal the openings.
Other scientists are sceptical about these scenarios. 鈥淭he fluids need to be at such a high pressure they will crack most rocks,鈥� says Chris Scholz, a geologist at the Lamont-Doherty Earth Observatory in New York. But the model will not work if any fluid is lost when they crack, he says. Even if there is only a tiny loss each time the fault slips, the plate would soon bleed dry, he argues. Then the fault would be subjected to frictions that are not consistent with the missing heat.
If all the theories involving water turn out to be flawed, what then? Enter Jay Melosh from the University of Arizona with an alternative, water-free theory that he outlines in a paper in this week鈥檚 Nature. Paradoxically, Melosh鈥檚 thoughts were not even close to California when his theory was born: he was studying crater formation on the Moon and planets. When the Moon is struck by a meteoroid, the rock leaves a bowl-shaped pockmark on the surface. But within a matter of minutes, the centre of the crater collapses to form a flat floor, leaving terraced walls like a series of steps into the crater.
The processes involved should be much simpler than what goes on during an earthquake. The Moon has no shifting plates, and in the dry lunar vacuum there is no water and no gases to complicate matters. The only force that could be driving the collapse is the Moon鈥檚 own gravity. A handful of calculations based on a few reasonable assumptions were bound to be enough to describe what was going on 鈥� or so Melosh thought. But the calculations didn鈥檛 work out. 鈥淚t was astonishing,鈥� says Melosh. 鈥淎ccording to the numbers the craters shouldn鈥檛 collapse at all.鈥� His calculations indicated that friction between the soil particles should have been more than enough 鈥� by a factor or ten 鈥� to keep the walls of the original craters in place.
Melosh thinks he has a solution that might also explain the missing heat of the San Andreas. Obviously enough, there will be friction between the soil particles of a Moon crater only while they are in contact with each other. If a meteor strikes the surface of the Moon, the particles start bouncing about -and the friction disappears. The intrinsic stickiness of all the surfaces is the same, but there are moments when each particle is detached from the others and has the equivalent of zero friction. It could be a similar story for the San Andreas, says Melosh. Vibrations generated in the first few moments of an earthquake could separate the rocks, making them behave like a slick liquid.
Jim Rice, an earthquake theorist at Harvard isn鈥檛 persuaded. He says he can鈥檛 rule out Melosh鈥檚 theories in every case, but he doesn鈥檛 think they describe the typical earthquake. 鈥淔or his model to work, these vibrations have to be very intense and stay at the fault rather than disperse,鈥� says Rice. But he points out that energy from the quake tends to radiate quickly and efficiently away from the source, which is precisely why we feel earthquakes at the Earth鈥檚 surface.
Melosh agrees that some focusing mechanism is required, but says the structure of the fault itself would help keep enough energy trapped. He cites investigations in 1993 led by Chester which examined the section of the San Andreas that passes through the San Gabriel Mountains. On this ancient part of the fault, the strata that millions of years ago were 3 to 5 kilometres down are now near the surface. Chester found that the plate border consists of a very finely ground rock layer 1 to 10 centimetres thick, sandwiched between 5-metre layers of highly fragmented rock 鈥� a structure thought to be typical of an old strike-slip fault. Melosh says his computer model shows that the inhomogeneity caused by the cracks will bounce vibrations from the quake back to the thin layer of pulverised rock.
Seismic setback
But there is another problem, and it is shared by Lachenbruch and Sibson鈥檚 dynamic water model or indeed any model that assumes that the friction of the rocks remains high until the quake begins. For these models to work, enough strain must be built up before the earthquake to overcome the initial high friction in rocks all along the eventual rupture zone. This strain is then released as measurable seismic energy when the fault ruptures. But even the largest quakes seen so far on the San Andreas have released too little seismic energy for these models to fit.
Joe Andrews, also from the USGS at Menlo Park, believes there鈥檚 a loophole that can rescue dynamic theories, one which has important implications for earthquake prediction. Perhaps strain builds up until it can overcome the initial high friction in one small 鈥渋nitiation point鈥�. This region 鈥� around ten per cent of the rupture zone 鈥� then begins to move. As it does so, it triggers a drop in the initial friction of the rest of the rocks (because the water heats up, or the rocks start vibrating). This means that much less strain energy is needed to shift the whole rupture zone, and there would be no large release of either heat or seismic energy during the quake.
If there is just one initiation point for a given portion of the fault, it should be possible to predict local quakes reliably by monitoring each of these points closely, Andrews says. Any activity at the initiation point would warn of problems to come. However, Andrews鈥檚 suggestion, if true, could just as easily have gloomy consequences: if such nucleating centres crop up chaotically and unpredictably along the fault, then very extensive 鈥� and unreasonably expensive 鈥� measurements would be needed to detect a coming quake. The prospects for detecting quakes would have taken a giant step backwards.
Hunt the hypothesis
With so many competing ideas about exactly what happens during a quake, a practical earthquake prediction programme is still a distant dream. 鈥淲e need to narrow down this growing plethora of hypotheses,鈥� says Mark Zobach of Stanford University. To do that he wants to drill a hole kilometres into the heart of the San Andreas Fault, a multimillion dollar project that he wants to complete by the turn of the century. The data that they will retrieve will make the deep bore worth the cost. If unknown types of clay are hidden in the darkness of that hole, the researchers will discover them. If the fluid pressure is fantastically high, they will measure it. And if Melosh-type vibrations are running through the fault, they鈥檒l detect them.
And what if they find them all? Rice acknowledges the possibility of this unsettling outcome. 鈥淲e all hope for the simple answer, but we have to admit that there may not be one. All these factors may come into play,鈥� he says. 鈥淎nd maybe one or two other things we haven鈥檛 even thought of.鈥�
