# News this Week

Science  31 Jan 1997:
Vol. 275, Issue 5300, pp. 613
1. Geophysics

# Deep-Sinking Slabs Stir the Mantle

1. Richard A. Kerr

New images of Earth's interior may end a long-running debate by showing that cast-off slabs of surface rock sink to the very outskirts of the core, mixing the mantle from top to bottom

SAN FRANCISCO—Over the eons, pieces of Earth's ocean floor have vanished like the lost island of Atlantis, slowly sinking below the surface in the stately cycle of plate tectonics. But where do these great slabs of oceanic plate go, and what happens to them inside Earth? Geophysicists have long pondered the issue, for it bears on the inner workings of the planet and on a great debate about the mantle, the inaccessible realm below the crust. Some have suggested that the mantle is divided neatly into two layers, with the slabs confined to the shallow top layer. But many seismologists, painting increasingly detailed mantle images with deep-probing seismic waves, have argued with growing confidence that the slabs dive to the very bottom of the 2900-kilometer-deep mantle, stirring it from top to bottom.

Now, in the view of many seismologists, two detailed seismic images of Earth's interior clearly show that the burial ground of these “lost slabs” is indeed deep within the bowels of the Earth. The discovery suggests that the lower mantle is an integral part of the engine of plate tectonics, which recycles old, cold slabs and sends hot rock back toward the surface.

The new studies, done by independent teams using different kinds of data, have so impressed geophysicists that the great mantle debate may be over. “There's no doubt—you see the slabs clearly penetrating all the way from the surface to the core-mantle boundary,” says seismologist Michael Wysession of Washington University in St. Louis. “In my opinion, this is the final argument. I'm convinced it's some form of whole-mantle circulation.” Although some seismologists are less adamant than Wysession, most agree that whole-mantle mixing is the most reasonable interpretation of the new images, presented here at the December meeting of the American Geophysical Union (AGU). “There's now enough evidence to see that, indeed, in some areas, the slabs do penetrate into the lower mantle,” says seismologist Barbara Romanowicz of the University of California (UC), Berkeley. “It's hard to see what else it could be.”

Two layers or one?

Since the 1960s, geophysicists have known that Earth's internal fires—stoked by heat leaking from the core and by radioactive decay in mantle rock—drive slow convection in the mantle, as a stove burner roils a pot of water. This heat-driven engine reshapes Earth's surface, as tectonic plates move across the surface and then plunge into the interior at so-called subduction zones. But the nature of this convection has long been debated. Laboratory experiments on rocks thought to match the composition of the mantle, plus seismic and other data, suggested that, below a depth of 660 kilometers, the mantle becomes denser and more viscous, like molasses layered beneath water. So, some researchers suggested that each layer churns separately, with no mixing between them.

Some seismologists were slowly putting together a different picture, however. By the early 1980s, they were seeing a glimmer of recognizable slabs in their images of the mantle, assembled from data on the speeds of seismic waves traveling from earthquakes to receivers along many different paths through Earth (Science, 17 August 1984, p. 702). A wave's speed depends on the rock's temperature—the hotter the rock, the slower the speed—and composition; as a result, cold, dense slabs speed up the waves passing through them. To form an image of slabs in the mantle, seismologists need the seismic velocity at each spot along a wave's path and so must mathematically combine tens of thousands of paths. A favorite technique is akin to the computerized axial tomography, or CAT, scan that can turn x-rays crisscrossing your head into a three-dimensional image of your brain.

The early mantle scans were pretty blurry, but some by Thomas Jordan of the Massachusetts Institute of Technology (MIT) and Kenneth Creager of the University of Washington zeroed in on small regions and seemed to reveal cold slabs of oceanic plate diving through the upper mantle—and through the putative barrier at 660 kilometers (Science, 7 February 1986, p. 548). But later imaging showed that the barrier does in places deflect descending slabs away from the deep mantle.

Arguments and counterarguments flew as each side continued to amass new evidence. In recent years, for example, regional seismic imaging around the Pacific Ocean at places such as the Sea of Okhotsk, the Marianas Islands, and central Java “can most easily be interpreted as slabs going through and penetrating to depths of at least 1000 kilometers,” says Thorne Lay of UC Santa Cruz. But just as the barrier to the lower mantle seemed to be smashed, mineral physicists investigating the properties of presumed mantle minerals in the lab began suggesting that the barrier itself is a wide zone, rather than a wall, that stretches down to 1000 kilometers.

With the presentation of the new global imaging studies at the AGU meeting, that debate has now swung heavily in favor of deep slab penetration and whole-mantle mixing. The two groups, led by Stephen Grand of the University of Texas, Austin, and by Rob van der Hilst of MIT, presented the clearest images yet of the entire mantle. Improvements included several kinds of corrections to the seismic wave paths reported by the International Seismological Center, such as precisely relocating source earthquakes, and the use of shorter wavelength waves that can create sharper pictures. The images show slabs extending at least 1000 kilometers below the 660-kilometer barrier at many places around the Pacific where slabs are subducting into the mantle today. “There is no evidence for a barrier to mantle flow at 1000 kilometers,” says van der Hilst.

The slab graveyard

In regions of past subduction, where thousands of kilometers of ocean crust vanished long ago into the planet's interior, the slabs have plunged even farther into the lower mantle. Both studies, for example, reveal great slablike features 500 kilometers wide and many thousands of kilometers long hanging in the lower mantle below the now-vanished Tethys Ocean from the central Mediterranean to Indonesia and under the western Americas from Siberia to South America. And in two other places where the images are particularly sharp due to numerous seismic ray paths—beneath the Caribbean and Central Japan—both studies show slablike features that extend from the top to the bottom of the mantle, just above the molten iron core. “It really looks like slabs slide down to [the mantle bottom],” says Grand.

When they get there, he says, they appear to spread out. On the basis of earlier, fuzzier images, some researchers had suggested that the last 300 kilometers of mantle above the core, dubbed the D″ layer, is a graveyard of old slabs that have come to rest on the bottom. Data from core-skimming seismic waves, compiled last year by Washington University's Wysession, supported that view. And now the broad pattern of fast and slow seismic velocities in the D″ layer of both new images confirms the pattern, says Grand. As glimpsed before, the seismically fast zones of possible sunken slabs appear beneath zones of subduction over the past 150 million years, while the seismically slow, presumably hotter parts of D″ fall beneath the southwest Pacific and Africa, where no subduction has occurred in the past 200 million years.

These deep-diving slabs imply a simple mixing pattern for the mantle, says mantle modeler Michael Gurnis of the California Institute of Technology (Caltech). Both studies suggest that mantle downwelling occurs along only a few lines where cold slabs descend, and hot, buoyant mantle—something seismic imaging has trouble rendering in any detail—rises between these descending arms of mantle circulation (see diagram and sidebar). It seems no more complicated than the broad pattern of plate tectonics seen at the surface. Says Gurnis: “It really solves, I think for good, this issue of whole-mantle versus upper mantle convection. There is whole-mantle convection, and [its form] seems to be rather simple.”

Don L. Anderson of Caltech, a longtime proponent of a layered mantle, admits that the images are impressive. “The amazing thing to me,” he says, “is that their models agree so well even though they use completely different data and [analysis] techniques.” Van der Hilst and colleagues used only P waves—pressure waves akin to sound waves in the air—that pass directly through the mantle or the mantle and the core. Grand, on the other hand, used only shear waves—undulations that resemble ocean waves and that can follow many different contorted paths through the mantle, such as repeatedly bouncing off Earth's surface into deeper rock, then bending to the surface again.

Still, Anderson and other researchers who have advocated a layered mantle are holding out for more evidence. “I guess I'm not convinced,” says mineral physicist Raymond Jeanloz of UC Berkeley. “To what degree is what is inferred from the patterns in the eye of the beholder? I appreciate that the guys who are doing the hands-on data analysis feel the results just leap out at them, that they really see evidence for slab penetration … but how you link up these blobs of high velocity and whether you infer they represent slabs going straight through [the 660-kilometer barrier] is not quite so obvious to me as it is to them.”

Jeanloz and Anderson would both like to know more about the chemistry and physics of the features behind the seismic images. The images could be compared more closely with computer models that simulate how past subduction should have shaped the mantle, Jeanloz suggests. And Anderson would like more comparisons of the seismic data with experimentally determined mineral properties, to prove that the seismically fast features are actually slabs: “Just because you found some [fast] regions, it doesn't mean anything until you know what that means in terms of temperature and chemistry.”

To many seismologists, as Gurnis puts it, these are “details” that “remain to be worked out,” not fatal flaws. And before such studies are finished, even more persuasive images may appear. Grand notes that there are plenty of seismic data in hand that can be analyzed to sharpen the still-fuzzy spots in the images, which reflect gaps in the distribution of earthquakes and seismic stations. Of course, many mysteries remain, such as whether slabs now imaged only into the midmantle can be seen to extend to the bottom and the ultimate fate of slabs that have come to rest at the core-mantle boundary. Slabs may be buried deep, but that doesn't mean that their rest is undisturbed.

2. Geophysics

# 20,000 Leagues Under the Earth

1. Richard A. Kerr

There's no lack of geological processes to study on Earth's surface, but another place on Earth, or rather inside it, may be almost as dynamic. At last month's fall meeting of the American Geophysical Union (AGU) in San Francisco, seismologists sharpened their view of the lowermost mantle, a shadowy land almost 3000 kilometers down. What they see is a realm where “continents” may fall from above, fiery “bogs” stew above the molten iron core, and whole chunks of landscape can loft upward like so many hot-air balloons. Strange as it seems, this topsy-turvy world is just what you would expect to find at the base of the grand circulation pattern that seems to link our own world with the bottom of the mantle (see main text).

Seismologists owe their new window on this realm to seismic waves that pass through the lowermost 20 kilometers of the mantle on their way from earthquakes to distant receivers. Last spring, seismologists Edward Garnero of the University of California (UC), Santa Cruz, and Donald Helmberger of the California Institute of Technology showed that compressional seismic waves (the analog of sound waves in air) traveling at these depths slow down by about 10% when passing beneath the central Pacific. Slower waves usually mean hotter rock, but such a whopping seismic retardation would require either rock that has partially melted into a stiff mush or a dramatically different rock composition.

At the AGU meeting, mineral physicist Quentin Williams and seismologist Justin Revenaugh of UC Santa Cruz and Garnero reported new evidence that this “ultralow-velocity zone,” or ULVZ, consists of partially melted rock. Seismic shear waves, which resemble the undulations of an ocean wave, are particularly sensitive to the stiffness of rock. If the rock in the ULVZ were partially melted, it would slow the shear waves much more than compressional waves. After analyzing waves from 315 earthquakes, the team found that shear waves are indeed slowed three times as much as compressional waves in this region. “The abundance of evidence favors partial melting,” concludes Williams.

The Santa Cruz group also reported that they had broadened their search for the ULVZ beyond the southwest Pacific to include 44% of the globe. They found a strong tendency for the ULVZ to be detectable beneath relatively hot portions of the lower mantle identified by other seismic surveys, and undetectable beneath colder regions. The ULVZ also showed up beneath both the southwest Pacific and Africa, areas where molten rock reaches the surface to form numerous volcanic hot spots, and beneath the hot spots of Hawaii, Pitcairn Island, and Iceland.

To the Santa Cruz group, this link between partially molten deepest mantle, hot areas a few hundred kilometers above, and surface hot spots suggests that plumes of hot rock rise to the surface from the deepest parts of the mantle. Hot plumes have been imaged in the upper mantle, but, for over 25 years, geophysicists have debated how deep their roots go. Now, Williams and his colleagues propose that plumes rise from regions of thick ULVZ, which could cause a plume because the rock's relative fluidity could lead to instability and a plume lift-off. Or the cause and effect might be reversed, Williams says, with the plume causing the thick ULVZ by drawing hot rock upward and thickening an existing but undetectably thin ULVZ.

The absence of a thick ULVZ beneath colder mantle also suggests the same mantle circulation pattern that other seismic images imply, Williams says—the descent of cold slabs of tectonic plate from the surface all the way to the lower mantle. Indeed, the colder regions of lowermost mantle coincide with places where the great, cold slabs of tectonic plates have fallen from the surface during the past 200 million years. It all adds up to a picture of the lowermost mantle as a place where sunken slabs are piled to form relatively cold “continents,” surrounded not by seas so much as fiery bogs of partially molten rock that can rise into towering plumes toward the surface—our own world in a fun house mirror.

3. AIDS

1. Jon Cohen

Success often comes at a price. Researchers who gathered here last week for the most influential AIDS meeting in the United States heard one speaker after another praise the dramatic advances in drug treatments that recently have captured the media spotlight. But the 2300 attendees also heard scores of reports about the less glamorous task of filling out and qualifying last year's bold claims. “We're not hearing the headline stuff we heard last year. The data at this meeting were much broader and deeper,” said the conference's chair, virologist Douglas Richman of the University of California, San Diego (UCSD). Organizers of the 4th Conference on Retroviruses and Opportunistic Infections* also had to contend with their own success: The gathering has become so popular that they had to set strict attendance limits, angering many AIDS activists and scientists who were locked out.

David Ho, head of the Aaron Diamond AIDS Research Center in New York City, set the tone for the meeting in the opening session, when he cautioned his colleagues to put the recent progress in AIDS treatment in its proper perspective. “We must dutifully avoid unwarranted triumphalism [as well as] the undue pessimism that prevails in some circles,” he said. “The state of HIV treatment is neither black nor white. We must paint the situation in the proper shade of gray.” Indeed, many fundamental questions about the new anti-HIV drugs are still unresolved, including how best to use them and how to assess their limitations. And the treatment picture will take on even subtler shading as drug companies attempt to bring a flood of new drugs to market.

One aspect of AIDS research in 1997 can be rendered in black and white: For the first time since the start of the epidemic, HIV-infected patients and clinicians have at their disposal an arsenal of potent, virus-crippling drugs. A new class of drugs that inhibit HIV's protease enzyme, which the virus uses to assemble new copies of itself, is the arsenal's mainstay. When combined with older drugs like AZT and 3TC that disable the enzyme that copies the virus into the host cell, the protease inhibitors can pound the virus so hard that, in many patients, even the most sensitive tests cannot find it in the blood cells.

Several studies presented at the meeting attempted to assess the impact of the new drug regimens on patients by looking at how frequently they were becoming ill. As AIDS researchers constantly point out these days, reducing the viral load (the total amount of HIV) in a patient's blood does not necessarily mean he or she will suffer fewer AIDS-related diseases. The hope, of course, is that the drugs will help HIV-infected people to live longer. But at present, most drugs win regulatory approval based primarily on viral-load data, and few of the trials measure clinical outcomes.

The largest analysis of links between new drugs and illness was presented by Yves Mouton of Dron Hospital in Tourcoing, France. Mouton and colleagues looked at 7757 patients from 10 AIDS centers in France. In the study period, between the fall of 1995 and 1996, the use of anti-HIV drugs in these patients jumped by 49%. At the same time, AIDS-defining diseases dropped 36%, said Mouton. The number of days patients were hospitalized also plummeted by more than one-third. Mouton attributed these improvements largely to the drugs.

Mary Ann Chiasson, assistant commissioner of the New York City Department of Health, reported similarly upbeat statistics. In New York City, which accounts for 16% of U.S. AIDS cases, AIDS deaths last year dropped by 30%. But health officials did not attribute the drop to increased use of protease inhibitors. According to Chiasson, the AIDS death rate began to fall before the two most powerful drugs reached the market last spring. She suggested that the decline in deaths may instead be linked more closely to an increase in federal funding in 1994 for AIDS patients, which led to better prevention and treatment of opportunistic infections.

Other researchers reported efforts to unravel how much virus remains in the bodies of people who have “undetectable” HIV in their blood. As Northwestern University molecular biologist Steven Wolinsky stressed, “Even though we don't have evidence of virus in blood by currently available tests, it doesn't mean it's not there.”

One provocative study of twins infected by HIV at birth underscored this point. Katherine Luzuriaga from the University of Massachusetts, Worcester, described how she and her co-workers treated a baby boy and girl with three anti-HIV drugs after they showed signs of infection at 10 weeks of age. The drug combination soon drove viral RNA in the blood down to undetectable levels. Their levels of antibodies against the virus also steadily declined, another signal that the drugs had knocked back the virus. “You can initiate therapy early in children and achieve a spectacular effect,” says Wolinsky.

But after 16 months of treatment, one child's blood suddenly tested positive for HIV RNA. Since then, the researchers have pushed the viral load back down with a different drug combination. But the case highlights the fact that the new drugs have yet to cure anyone of HIV infection—and it's still unclear whether they ever will, says Wolinsky. “The study lends credence to modeling studies that say the duration of therapy has to be measured in years,” Wolinsky says, referring to a mathematical model presented at the meeting by mathematician/immunologist Alan Perelson of Los Alamos National Laboratory and Ho that estimates it will take 2.3 to 3.1 years of viral suppression to rid the body of HIV.

And as Winston Cavert of the University of Minnesota, Minneapolis, noted at the meeting, 99% of the HIV in the body ordinarily resides in the lymph nodes. In collaboration with Sven Danner at the Academic Medical Center in Amsterdam, the Netherlands, Cavert, Ashley Haase, and co-workers compared HIV levels in blood and samples of tonsils (a type of lymph node) biopsied from 10 patients who had taken the protease inhibitor ritonavir along with AZT and 3TC for 24 weeks. The researchers saw “a dramatic decline” in HIV in both blood and tonsils, said Cavert. But this good news was tempered by the bad news that the one tonsil they examined that had no apparent HIV RNA still harbored its DNA—the form of the virus that infiltrates a cell's nucleus and can lie dormant for months or even years.

While such studies help answer fundamental questions about how well the drugs work now, a question on everyone's mind is how long the new combinations of anti-HIV drugs will work before resistant strains of the virus crop up, as they quickly did with first-generation treatments such as AZT. One encouraging study suggests that if patients consistently adhere to a regimen of more than a dozen pills a day, resistance can be kept at bay. UCSD's Joe Wong and colleagues reported data from a study sponsored by Merck & Co. in which 18 of 21 patients who have taken the company's protease inhibitor, indinavir, plus AZT and 3TC, have undetectable or extremely low levels of HIV RNA after 68 weeks of treatment.

Several reports at the conference, however, indicated that in the real world, many patients don't follow drug regimens strictly, and it may only take a few days off treatment for resistant strains to appear. Moreover, as many presenters pointed out, physicians who aren't current with the latest research and prescribe single drugs or weak combinations that do not fully suppress HIV often inadvertently encourage resistant strains. “Great mistakes are being made by people who think they know what they're doing,” said Joep Lange of Amsterdam's Academic Medical Center.

Lange was particularly critical of clinical trials that included as part of their designs “suboptimal therapy,” which he blamed on “regulatory requirements, stupidity, and greed.” He singled out drug companies for testing what he called “incestuous combinations” of compounds that they own, rather than working with other companies to optimize treatments.

Part of the problem, says meeting vice chair Constance Benson of Rush Medical College in Chicago, is simply that changing a clinical trial to reflect new findings requires input from many people. Benson acknowledges that right now, for instance, the AIDS Clinical Trials Group (ACTG), sponsored by the U.S. National Institutes of Health, has “three or four” trials under way that include potentially suboptimal treatments. In each case, the researchers had designed the studies before the potent drug combinations now considered “optimal” were available. Benson, a member of ACTG's executive committee, says the group will meet soon to discuss changing the studies.

But according to several presenters, even people who develop resistance to today's drugs still have reason for hope because of the many new drugs in the pipeline. The hands-on favorite to make it to market next is a protease inhibitor, nelfinavir, which William Powderly of Washington University in St. Louis noted could give patients a big leg up in the resistance battle. HIV strains that are resistant to one protease inhibitor often are resistant to others. But an HIV strain that develops resistance to nelfinavir still is susceptible to other protease inhibitors, says Powderly.

Made by Agouron in La Jolla, California, the drug has been tested in nearly 700 patients. Powderly reported that when the drug was combined with AZT and 3TC, investigators could not detect HIV in the blood of fully 60% of the patients after 6 months. The triple therapy also boosted CD4 cells, the immune-system actors HIV destroys, by an average of 155 to 160 cells per cubic milliliter of blood. This is comparable to the current triple therapies now being used. Agouron has asked the Food and Drug Administration to license the drug and is hoping to receive approval in the next few months.

Further down the pipeline is a new protease inhibitor described at the meeting by scientists from Abbott Laboratories, the makers of ritonavir. Called ABT-378, the drug is 10 times more potent at knocking back HIV than are other protease inhibitors, said company scientist Hing Sham: “Importantly, ABT-378 shows potent inhibition of even highly resistant HIVs.” The drug is currently in small safety trials in uninfected humans.

Still very much up in the air is the fate of the conference itself. The gathering turned away many would-be attendees, including some AIDS activists who threatened to disrupt the conference and even harm organizers. Indeed, safety concerns led the organizers to hire a team of security guards, who shadowed every move the organizers made. Conference chair Richman says the organizers haven't yet decided whether they will allow more people into the meeting next year. “We hear strong opinions in both directions,” he says. They plan to make their decision after reviewing comments from this year's attendees.

• * 4th Conference on Retroviruses and Opportunistic Infections, 22-26 January, Washington, D.C.

4. Physics

# First Atom Laser Shoots Pulses of Coherent Matter

1. Gary Taubes

To the uninitiated, a laser is a pin-thin beam of brightly colored light that you'd be wise not to shine in your eyes. To connoisseurs, it is a coherent beam of photons locked in identical quantum states, meaning they all have exactly the same wavelength and travel precisely in step, crest to crest, trough to trough. Now, the word laser has taken on yet another meaning: a beam of atoms marching in quantum lockstep, like the photons of a light laser.

Such a laser could aid everything from atomic clocks to chipmaking. Two years ago, physicists achieved the crucial starting point when they created an exotic state of matter known as a Bose-Einstein condensate. Now, a group at the Massachusetts Institute of Technology (MIT) led by Wolfgang Ketterle reports on page 637 that they have shaped this novel material into pulses of atoms that have the hallmarks of a laser beam. “The experiments are gorgeous,” says Oxford University physicist Keith Burnett, and the demonstration that this is really a laser is “the most beautiful clear evidence.”

A Bose-Einstein condensate is a dense cloud of atoms cooled in a magnetic trap to within an iota of absolute zero, where their quantum-mechanical waves merge. The formerly disparate atoms take on the characteristics of a single particle, in which the microscopic laws of quantum physics are writ large. Simply making a condensate is “bloody difficult,” says Burnett, let alone turning it into a laser. Since Eric Cornell, Carl Wieman, and their colleagues at the National Institute of Standards and Technology and the University of Colorado made the first one in 1995 (Science, 14 July 1995, p. 198), only Ketterle's group and a team at Rice University led by Randy Hulet have been able to follow suit. Now, the MIT group has found a way to extract pulses of atoms from a condensate and has shown, by allowing two pulses to interfere with each other, that each constitutes the single coherent wave required of a laser.

The first step, creating what laser physicists call an output coupler to extract the atoms from the trap, was relatively easy—“peanuts,” says Ketterle. In a conventional laser, the output coupler simply taps light from the lasing cavity, where it is bouncing back and forth between mirrors. “Laser light is like a big wave,” Ketterle explains, sloshing back and forth between the mirrors. “You want to take a little bit out for the beam, and then the big wave is amplified again and regenerated.” An ordinary laser relies on leaky mirrors to allow perhaps 10% of the light to escape and form a beam.

For the atom laser, says Ketterle, the MIT team opened a leak in the trap confining their sodium atoms. The trap, he says, “can be described loosely as like atoms bouncing back and forth between magnetic walls.” The walls, however, only retain atoms whose spin axis is pointing up. Flip those spins, and “the restoring forces become expelling or repulsive forces.” So the MIT researchers simply apply another magnetic field to the atoms, which tilts their spins to any desired angle. “We varied the angle between 0 and 180 degrees, and at 0 degrees the magnetic mirror was still reflective, so nothing was coupled out; and at 180 degrees, everything was coupled out.” By controlling the angle, the researchers could then “pulse out” portions of the condensate, the way a laser pulses out dollops of coherent light.

That was the peanuts part, which Ketterle's group reported at a conference in Sydney, Australia, last July and in the 27 January Physical Review Letters. What's reported in this issue of Science is the challenging part: showing that these dollops of condensate are coherent, which means the quantum-mechanical wave functions of the particles are all oscillating up and down in phase. To show that, says Ketterle, “you have to overlap matter waves from two different sources, [just as] you can prove light is a wave by passing it through two slits and looking at the interference pattern.”

It took months of hard work to do this. “It only looked easy after it was finished,” says Ketterle. First, he and his colleagues created two condensates by beaming a laser up through the middle of their magnetic trap. The laser light repelled the atoms and split the condensate into two distinct halves. For this test, there was no need to pulse the condensates out of the trap; instead, the group just turned off the trap and let them free fall. As the condensates fell, they expanded into the surrounding vacuum until they overlapped and interfered, demonstrating the atomic version of the bright and dark fringes in an interference pattern.

“The density of the overlapping region is modulated,” says Ketterle. “Every 15 microns, we have matter, no matter, matter, no matter. Now, we just shine some light onto the pattern and see this shadow with black-and-white stripes.” Says Burnett, “It's not just a little crappy demonstration but a big, juicy interference pattern.”

Having proved the condensate is coherent, Ketterle and his colleagues can use the output coupler to extract the condensate in pulses, which makes the setup effectively the first primitive atom laser and raises the question of where they go next. So far, they have been able to get eight pulses out of a condensate before they have to reload, which takes 20 to 30 seconds. One of their first goals is to figure out a way to restock the condensate as they go along to create the atomic version of a continuous wave laser. “Remember, these things are a few weeks old,” Ketterle says, “and we need a major improvement in output power, a major reduction in complexity, and also improvement in shaping the pulses.”

At that point, any field that relies on beams of atoms might benefit from the brighter and better controlled beams of an atom laser. Atomic clocks, which are based on the vibrations of atoms drifting through a cavity, are one candidate. Another is nanolithography, the technique by which circuit designers lay out minuscule features. It now depends on a mask or stencil to control where atoms or light land on a surface, but an atom laser—which could be focused and directed like a light laser—might provide a way of writing the patterns directly, says University of Texas physicist Dan Heinzen.

The technology does seem to come with a handicap: Unlike light, an atomic laser beam can't propagate freely through the atmosphere. But Burnett says it's too early to focus on limits. After all, at the birth of the light laser, “people talking about applications really didn't imagine them being in every supermarket check-out counter.”

5. Physics

# A New Recipe for Atom Condensates

1. Gary Taubes

While researchers at the Massachusetts Institute of Technology have concentrated on turning a Bose-Einstein condensate into an atom laser (see main text), a group at the University of Colorado and the National Institute of Standards and Technology (NIST) in Boulder has found a new recipe for these exotic assemblages of atoms. They serendipitously created two of them coexisting in the same magnetic trap, “a little bit like a big blob of oil next to a big blob of vinegar,” says NIST's Eric Cornell.

In a paper in the 27 January Physical Review Letters, Cornell, Carl Wieman, and their colleagues—who created the first Bose-Einstein condensate in 1995—report that they cooled a single cloud containing rubidium atoms in two subtly different quantum states. The states are distinguished by whether the electrons and nucleus of each atom have spins that are oriented in the same or opposite directions. Because the mechanism used to cool the atoms into a Bose-Einstein condensate works on only a single state, the researchers normally “take care to put all the atoms in the same internal state,” says Cornell. “But on this particular day, the apparatus was not working very well, and almost by accident we got two internal states” in the trap. The system promptly cooled the atoms in one state, but the other atoms cooled “sympathetically,” says Cornell, by losing heat to the adjacent, already-cooled cloud.

The end result was two distinct clouds, says Cornell: “They're all exactly the same isotope of rubidium, but you have one rather distinct cloud in one internal state and another distinct cloud in another internal state. They do overlap a little bit, but they find each other repulsive.”

To Wieman, “That is the most remarkable thing about the experiment.” He adds that the interaction holds “a lot of interesting physics.” The sympathetic cooling process should also expand the repertory of condensates, he says. “There are all kinds of other atoms one can now stick in the trap and turn into Bose-Einstein condensate by sympathetic cooling. It's like a refrigerator where you only know how to cool Freon directly, but you can get anything cold by sticking it in thermal contact with the Freon.” The condensate chefs now have a new tool.

6. Developmental Biology

# A “Master Control” Gene for Fly Eyes Shares Its Power

In a startling experiment reported 2 years ago, Swiss biologists caused surplus eyes to sprout on fruit flies' wings, legs, and antennae—all by manipulating a single gene called eyeless (ey). Grotesque as this spectacle was, researchers hailed it at the time: Besides shedding light on eye development, it also supported the seductive idea of “master control genes” that can single-handedly order up complex organs by turning on other genes. But now it seems that ey has a partner—perhaps even two—and the all-powerful master controller may be merely a member of a committee instead.

New work reported in the January issue of the journal Development shows that a fly gene called dachshund (dac) can, like ey, give rise to ersatz eyes when turned on in out-of-the-way places such as a developing leg or antenna. And the researchers also discovered that ey can't build these so-called ectopic eyes in flies missing dac—an indication that the two genes normally work together. “It's really an oversimplification to say that any one gene is the master-control gene for eye development,” concludes developmental geneticist Graeme Mardon of Baylor College of Medicine in Houston, who authored the study with technician Weiping Shen.

The Baylor team's result “is a very interesting discovery,” agrees Nancy Bonini, a Drosophila geneticist at the University of Pennsylvania. “If dac had been found before ey, you might say that dac is ‘the’ master regulatory gene in eye development. So, maybe we should think differently about these terms.” But Walter Gehring, the Swiss geneticist who led the original dramatic ey study—and whose lab recently discovered yet another eye-forming fly gene, christened twin-of-eyeless (toy)—says eyeless is still the master switch. “I don't think this [label] has to be revised,” he says.

The eye-popping powers of dac were discovered by accident. The gene got its name several years ago, when Yale University biologist Iain Dawson came across a mutation in the fruit fly Drosophila melanogaster that resulted in short, stubby legs—and also affected the arrangement of the 800-some individual eyes (called ommatidia) in each of the flies' compound eyes. Mardon, then a postdoc in the lab of geneticist Gerald Rubin at the University of California, Berkeley, found the gene independently. He went on to clone it and discovered that the protein it encodes resides in the cell nucleus, suggesting that dac helps regulate the expression of other genes.

But Mardon couldn't find which genes those might be. Then, in 1995, Gehring and colleagues Georg Halder and Patrick Callaerts at the University of Basel in Switzerland published their study on ey. To trick the gene into becoming active where it should be dormant, they used genetically engineered fly larvae that produced a gene-activating protein called GAL4 in many different body parts, such as wings, legs, and antennae. Then, they mated these flies to others in which ey was connected to a control switch activated by GAL4. The result was a brood of flies with eyes in unorthodox places (Science, 24 March 1995, pp. 1766 and 1788).

Mardon, eager “to see what dac might be doing,” borrowed the technique, linking not ey but dac to the GAL4-activated control switch. He and Shen found that 20% of the resulting flies developed clusters of fully formed ommatidia in odd locations. That's a much lower fraction than the Gehring team's 100%—perhaps, Mardon speculates, because making eyes in certain places in the body would require genes that dac does not activate, but ey does. Intriguingly, the Baylor team also found that ectopic expression of ey induces dac expression in the same places, and that dac can also turn on ey in a subset of these cells. And when Gehring's experiment is repeated in flies lacking dac, no ectopic eyes form. All this suggests to Mardon that the two genes evolved as partners, reinforcing each other's eye-building signals in a positive feedback loop.

So, which is the true master-control gene for the fly eye? Neither, says Mardon. Both, suggests Ulrike Heberlein, a Drosophila geneticist at the University of California, San Francisco. “Maybe we need to talk about a hierarchy of master regulators,” she says. But Gehring maintains that between ey and dac, ey is still the master. In his view, the term means that “if you make a gain-of-function mutation or switch the gene on ectopically, you get a complete wing or leg or eye or body segment,” he says. dac fits this definition, but he thinks it doesn't quite qualify as a master gene, because ey always induces dac, but dac can't always induce ey. To him, this suggests that ey is higher in the regulatory hierarchy. Or perhaps, suggests University of Southern California geneticist Kevin Moses, it all boils down to semantics: “Any genetic element that can become critical can be seen as a ‘master regulator.’”

The “big game” now, says Gehring, is to map out the eye-development pathway in detail, assigning places to ey, dac, and a handful of other genes known to be involved—including the recently discovered toy, a possible “co-master” regulator that seems to help activate ey, as Gehring reported at a conference in Tennessee last June. The gene toy is even more closely related than ey to pax-6, a mammalian gene involved in eye development, and may be the ancestral fly-eye gene, with ey an accidental duplicate that later took over most of toy's job, Gehring speculates. Whichever gene comes out on top, there will be plenty of depth left to plumb below.

7. Astronomy

# Gas Clouds May Be Relics of Creation

1. James Glanz

TORONTO—Astronomers who study a mysterious set of gas clouds speeding through the Milky Way can appreciate the plight of the very nearsighted tourist in Africa. A gnat seems to be crawling across her glasses, but when she removes them, the gnat is still there; finally she realizes that a rhino is charging over a ridge. Astronomers haven't had a similar flash of recognition yet. But some have proposed that what they thought were gnats—the spindrift of supernovas exploding in our galaxy—might be something much grander: huge, distant remnants of the galaxy's formation that extend well beyond the Milky Way and could fuel the formation of new stars for billions of years into the future.

The proposal, presented during an American Astronomical Society meeting held here from 12 to 16 January, relies on computer simulations of gas left over from the formation of the first galaxies and clusters of galaxies. The simulations show that these leftovers could survive until the present as clouds of gas roiling in the gravity of our Local Group of galaxies. “We start these [models] at about 1 billion years after the big bang and just let them evolve,” says Leo Blitz of the University of California, Berkeley. Fast-forwarding to the present, “we get what we see”—assuming the observed high-velocity clouds are lumbering masses in deep intergalactic space.

Few astronomers believe that the evidence is strong enough yet to prove these cosmic claims. But by pointing to the kinds of observations that could finally pin down the nature of the clouds, says Joel Bregman of the University of Michigan, “this breathes new life into the problem.”

The mystery dates back to the 1960s, when observations of radio emissions from hydrogen atoms in interstellar space showed that some of them belonged to clouds stampeding in all directions at hundreds of kilometers per second relative to Earth. The most complete catalog to date, compiled by Bart Wakker of the University of Wisconsin, lists about 550 of these rogue clouds. The biggest obstacle to understanding them is astronomers' ignorance of their distance and thus their actual size. “Distance is the most critical, but the most difficult,” says Wakker.

Still, Wakker thinks the clouds are closely associated with our galaxy. One possibility is that they are the handiwork of supernovas. By driving gas out of the plane of the galaxy in “fountains” that would tumble back, supernovas could stir up clumps of interstellar gas. “There are supernovas in the [galactic] plane; they do explode, so where does the gas go?” asks Wakker. “The galactic fountain seems reasonable.” Wakker has shown that this picture can account for most of the observations, although he and Bregman concede that it has a hard time explaining the very fastest clouds.

Blitz, along with David Spergel of Princeton University, Dap Hartmann of the Harvard-Smithsonian Center for Astrophysics in Cambridge, Massachusetts, W. Butler Burton of the University of Leiden in the Netherlands, and Peter Teuben of the University of Maryland, College Park, decided to try out a grander picture. They suggest that, rather than lying 10,000 or 15,000 light-years away, the high-velocity clouds are scattered on scales of more than a million light-years, stretching well beyond our galaxy toward its neighbors. And instead of being run-of-the-mill interstellar gas, they are relics of the great filament of primordial gas that coalesced to form the entire Local Group of galaxies. If so, the gravity of the Andromeda galaxy and the Milky Way could be accelerating the clouds to the high velocities that have puzzled observers.

To test this idea, says Blitz, the team used “an extremely simplified model of our local region of the universe.” In the model, the newborn Milky Way and Andromeda galaxies first draw apart with the general expansion of the universe, then move closer again because of their mutual attraction. Based on the changing gravitational field created by the two galaxies, the model calculates how nearby gas clouds should move, how much of the gas should get swallowed up by the galaxies, and how much should survive to the present epoch.

Hartmann says the model's predictions of where the clouds should tend to congregate in the sky and how fast they should move just about match his own detailed radio observations. The model predicts, for example, that clouds should be concentrated along a line connecting the Milky Way and Andromeda—the orientation of the original gas filament that formed the Local Group. The clouds do seem to cluster along that line, which impresses David Weinberg, a specialist in cosmic structure formation at Ohio State University. Throughout space, galaxies form patterns “like beads on a string,” says Weinberg. “If this is correct, then in addition to seeing the beads, you can still see the string.”

These leftovers, if that's what they are, should amount to roughly 100 billion solar masses of material—enough to nourish star formation in the Milky Way for billions of years. That would brighten the galaxy's future, says Blitz, who notes that observers have had a hard time finding enough fresh gas to sustain the present rate of star formation for much longer. Before astronomers draw too many conclusions, though, they want definitive evidence for or against the new cloud theory. So far, one observation has given it a boost by showing that the composition of one high-speed cloud could be extragalactic. But another has raised doubts by showing that a different cloud lies relatively nearby—too close for comfort in the new scenario. The cloud watchers are still waiting for that shock of recognition.

# Japanese Mission Stretches Limits of Interferometry

1. Dennis Normile

TOKYO—Japanese engineers are readying for launch on 7 February a satellite that should give radio astronomers a much better look at black holes and other extremely energetic objects in the universe. The satellite—if it's lofted successfully—is also expected to boost the fortunes of one of the country's premier research institutes as it seeks a larger role in Japan's exploration of space.

The $90 million MUSES-B satellite, built by the Institute for Space and Astronautical Science (ISAS), will be the first space-based antenna dedicated to very long baseline interferometry (VLBI). The technique allows astronomers to combine signals from widely spaced antennas, generating images as though they were produced by a single instrument with a huge collecting area. MUSES-B, which has an orbit that fluctuates between 1000 and 20,000 kilometers above Earth's surface, will take this technique to new heights, working in tandem with ground-based telescopes to generate images with unprecedented resolution. Says Hisashi Hirabayashi, project scientist for ISAS: “This will be the first mission ever in the field of radio astronomy to [have] a synthetic-aperture radio telescope bigger than the Earth.” Hirabayashi and his colleagues will be watching anxiously to see this novel vision take shape next week. Their anxiety will be sharpened by the fact that the 830-kilogram MUSES-B will be entrusted to the first launch of ISAS's new M-V rocket, which is capable of lifting more than twice the payload of previous ISAS launchers. It will also provide the first test of an innovative tension-truss design, with a half-dozen booms and cables providing structural support for the 8-meter antenna after it is unfurled in orbit. And if that weren't enough, the satellite's observations, at wavelengths of 18, 6, and 1.3 centimeters, will require exacting choreography among up to 10 telescopes scattered around the globe and five tracking and data-relay stations—all part of a 25-telescope network called the VLBI Space Observatory Program (VSOP). The end product will be 3 to 5 years of data that should sharpen astronomers' understanding of the environment surrounding black holes, the characteristics of star-forming regions, and radio sources throughout the universe. “It's a major step forward for VLBI,” says Roy Booth, director of the Onsala Space Observatory in Sweden, which is participating in VSOP. In particular, astronomers hope to use VSOP's sharp vision to peer into the nuclei of active galaxies, regions smaller than the solar system that pack the energy and output of an entire galaxy of stars. The nuclei—“the most powerful objects in the universe,” notes David Meier, an astrophysicist at NASA's Jet Propulsion Laboratory (JPL) in California—are believed to be comprised of black holes surrounded by rotating rings of gas, or accretion disks. If so, astronomers hope that their glimpse of the accretion disk should shed light on how these systems generate their energy, says radio astronomer Makoto Inoue of Japan's National Astronomical Observatory (NAO). Astrophysicists also plan to use VSOP to probe the origins of enormous jets of gases that appear to spurt from active galactic nuclei at velocities very near the speed of light. “This phenomenon is physically very interesting, but it is not understood at all,” says Hirabayashi. Other enigmatic phenomena high on the list of VSOP's priorities are maser spots, or point sources of intense microwave radiation. Masers, whose origins are poorly understood, often are found in regions of star formation. They can be used as celestial markers for such regions because their position, direction, and velocity can be determined from their frequency and movement. That information helps scientists determine the distance of these regions from Earth, and their movement can reveal whether the gaseous clouds that envelop the regions are turbulent or flowing linearly. “We'd like to study the maser process for its own sake and also for what we can learn about star formation,” says Onsala's Booth. VSOP may also turn its attention to supernovae. “We hope to observe how their shapes evolve after the explosion,” says Hirabayashi, at a level of detail not now possible. And general surveys to refine details of known radio sources could always turn up a surprise. “I think a good mission not only solves some problems, but also finds new phenomena [for] the next mission,” Hirabayashi adds. To coordinate these observations, ISAS and NAO scientists will work with colleagues around the world. An international panel will sort through requests for viewing time. ISAS will operate one of the five stations, with three to be run by JPL and the fifth by the U.S. National Radio Astronomy Observatory, which will also help process and analyze the data from the satellite and ground-based antennas. JPL has set up its own 5-year,$80 million program to work with both VSOP and a Russian space VLBI project, called RadioAstron, that is still under development.

The project is a major step for ISAS and NAO, which teamed up with ISAS because it had no independent launch capability. ISAS—whose scientists have done ground-breaking work in x-ray astronomy and whose YOHKOH solar satellite returned some of the most detailed x-ray images of solar flares ever recorded—is trying to squeeze maximum scientific benefit out of a budget of only \$190 million. NAO also has a distinguished track record in solar observations and radio astronomy.

For ISAS, which is trying to extend its technological capabilities in anticipation of future missions to the moon and Mars, a lot will be riding on next week's launch. Not only will it be a critical first test for the agency's new rocket, but MUSES-B is also breaking new ground in such areas as orientation control and high-bit-rate signal transmission. Indeed, in many ways, MUSES-B is an engineering test mission that just happens to be carrying a payload of interest to radio astronomers and astrophysicists. “Our mission just fit [ISAS's needs],” says Hirabayashi, a former NAO astrophysicist who joined ISAS to oversee the scientific aspects of the mission. But with the international community hungering for this kind of facility and a long list of interesting phenomena to observe, he says, “this is very good timing.”

9. Genetics Research

# Glaucoma Gene Provides Light at the End of the Tunnel

1. Gretchen Vogel

Glaucoma is an insidious disease. Just ask Kirby Puckett, the former Minnesota Twins outfielder, who was forced to retire from baseball because the disease, which gradually and painlessly destroys peripheral vision, irreversibly damaged his right eye before it was detected. Glaucoma, which blinds almost 12,000 people in the United States each year, has been just as elusive in the laboratory. Although it's apparently caused when pressure builds up in the eye and damages the optic nerve, exactly what causes the increased pressure and how it kills nerve cells is unclear in most cases. Now, new results may lead not only to a better understanding of the biochemical causes of the disease, but to better tools for early diagnosis as well.

On page 668, molecular geneticists Edwin Stone and Val Sheffield from the University of Iowa College of Medicine in Iowa City, with colleagues from six other institutions, report that they have identified the gene at fault in juvenile glaucoma, an aggressive hereditary version of the disease that strikes as early as the teenage years. “It's a very exciting find,” says Ellen Liberman, a glaucoma expert at the National Eye Institute (NEI). This “is the first time anybody has ever identified something specific … that indicates what might be going on.” The discovery of the gene has researchers scrambling to understand how glaucoma could result from defects in the gene's protein product, which is called TIGR and is made by cells that help control eye pressure.

But perhaps even more important, eye experts say, identification of the gene may aid in glaucoma diagnosis, which is difficult in the early stages of the disease. Although juvenile glaucoma accounts for fewer than 1% of all cases, early indications are that mutations in the TIGR gene cause at least 3%. This means that “the gene could allow us to identify up to 100,000 people in the United States who have this disease and otherwise wouldn't know about it and would risk losing their vision,” says Sheffield. These people could then be treated with drugs that, by lowering the pressure inside the eye, can prevent loss of sight if the disease is caught early enough.

To find the gene, Sheffield and his colleagues at Iowa used standard genetic linkage studies to identify DNA markers that are consistently inherited with the disease. By 1993, they had homed in on a region on chromosome 1 that carried several such markers, an indication that it also carried the glaucoma-causing gene. Researchers around the world found more linkages to the region, and a race was on to find the specific gene.

Using genetic data from more than 100 affected people in eight families, the Iowa researchers narrowed their search to a region of about 1 million base pairs. Combing the Human Genome Project's gene catalog, they found three genes within their suspect interval. Closer examination revealed that one, which encodes TIGR, contained mutations in five of the eight families initially tested, but not in any of 100 healthy controls—an indication that the gene is the one at fault.

The gene's effects may not be limited to the juvenile-onset families, however. The Iowa team found that it was also mutated in a family with 15 adult-onset glaucoma victims and in three of 100 randomly chosen adult-onset patients. In all, the researchers screened 330 unrelated glaucoma patients to come up with their estimate of a 3% frequency for TIGR mutations in all glaucoma cases.

That leaves other groups racing to find what may be a multitude of genes responsible for the glaucoma patients who do not seem to carry TIGR mutations. In fact, in the February American Journal of Human Genetics, geneticist Mary Wirtz of Oregon Health Sciences University and her colleagues report that they have linked a suspect region on chromosome 3 to adult-onset glaucoma.

But until other genes are found, the TIGR gene provides the only clues to the disease mechanism. For almost 10 years, Jon Polansky, a medical researcher who studies cell biology and hormone action, and molecular biologist Thai Nguyen, both at the University of California, San Francisco (UCSF), have been studying TIGR's connection to a different type of glaucoma. Doctors have known for decades that inflammation-reducing glucosteroids can cause a rise in eye pressure, especially in glaucoma patients and their relatives. Polansky and Nguyen identified the protein while studying the effects of steroids on the trabecular meshwork cells, which Polansky had induced to grow in lab cultures. In the eye, the trabecular meshwork cells help regulate eye pressure by controlling the drainage of fluid from the eye as new fluid is produced.

The cultured cells, when treated with steroids, secreted a protein which Nguyen, who had cloned the corresponding DNA, called TIGR (for trabecular meshwork inducible glucocorticoid response protein). Polansky says more recent experiments have suggested that high levels of the TIGR protein make the meshwork of cells less permeable. He suggests that an excess of TIGR may gum up the space between the meshwork cells and block the normal outflow of fluid from the eye. Whether the mutations in the TIGR gene have a similar effect or lead to glaucoma in some other way remains to be established, however. “We have a lot of work to do to figure out what this protein actually does,” NEI's Liberman says.

The Iowa researchers are already on their way. To study the protein's effects in a living system, they are trying to develop genetically engineered mice that produce either the mutated form of the TIGR protein or no TIGR at all. But mouse eyes and human eyes have significant differences, and Iowa's Stone says he and his colleagues also plan “to find as many actual human beings who are running around with this mutation as possible.” A larger sample will allow the researchers to get a better fix on how widespread TIGR mutations are and to uncover any correlation between the gene and specific symptoms or response to treatments.

“Until we figure out what TIGR is doing, anything is an open possibility,” says Liberman. Julia Richards, a molecular geneticist at the University of Michigan, who has also been searching for the juvenile glaucoma gene, agrees. The discovery, she says, will encourage researchers who have been hunting the gene to “get on with the business of determining the underlying mechanism … which is why we're really in the field in the first place.”