(FORTUNE Magazine) – Never has industry had a greater stake in the process of inventing and producing materials that are the flesh of new technology. Stuff like semiconductors, optical fibers, metallic alloys, and polymers that make up plastics and synthetic textiles, as well as every society-changing blockbuster substance of the future, can only come from the mother of all pantries: the periodic table, with its 90 or so useful elements occurring in nature. It's the basis of every industrial material that ever was, is, or can be. But until now there has been no sure-fire way to tap the pantry's riches.

Rarely, for example, have scientists devised new ways to manipulate the molecular structure of a single element such as carbon, which has long taken such familiar forms as graphite and diamonds. As for blending different atoms, the astronomical number of possible combinations has made it maddeningly hard to find good recipes for creating specific materials. The difficulties were summed up by the late Robert Laudise, former director of semiconductor research at Bell Laboratories (now Lucent Technologies), when he declared, "New materials have always been developed by guess, by God, and by good luck."

These days, says Roger French of Du Pont's Experimental Station in Wilmington, Del., "that's too slow." Researchers are at last finding quicker ways to come up with new stuff:

--At Du Pont, French and his colleagues are using sophisticated trial and error--a highly informed version of the traditional guessing method--to develop polymers that are pivotal to making next-generation computer chips.

--Another company is gunning trial and error to awesome speeds with a technique called combinatorial materials research. Using ingenious techniques, researchers mix, match, react, shake, bake, and otherwise combine atomic ingredients into thousands of recipes for new materials at fantastic tempos, then screen them just as fast for technological promise. The combinatorial approach has already yielded new crystals for digital X-ray systems slated for sale next year, as well as more efficient catalysts for polymer production.

--A third promising method takes the opposite tack. Instead of checking out myriad recipes, researchers use theory and computers to design virtual materials with desired properties before actually making any. The computational approach, as it's called, has led to architectures for new semiconductor structures as well as tougher steel alloys that may wind up in jet engines. Computation's partisans believe that every kind of future material, even something as mundane as next-generation gypsum wallboard less prone to cracking or buckling, could start out as a ghost in a computer.

Du Pont is using long-accumulated, specialized knowledge to add pizzazz to classic trial and error. The chance to capture several hundred million dollars' worth of business in a little-known corner of the chipmaking market provides plenty of motivation for a small team of polymer experts at the Experimental Station in Delaware, just uphill from where Wallace Carothers invented nylon in 1936.

Research manager Curtis Fincher says the ongoing microelectronics revolution is propelling several of Du Pont's latest R&D projects. "Everybody wants more computing power in less space for less dollars while consuming less energy," Fincher explains. "And that depends on impressing more circuitry onto the same space of silicon." The work of competitors like 3M and others at universities and in Japan lends extra urgency.

New materials would make denser circuits possible, says veteran chemist Andrew Feiring, 52, whose expertise centers on the science of fluoropolymers. It's the basis of one of Du Pont's greatest hits: poly- tetrafluoroethylene (PTFE), known to most people as Teflon, which has rung up billions of dollars in revenues over the past 40 years.

For nearly two years, under the guru guidance of Feiring and others, a team of Du Pont researchers has been looking for a fluoropolymer that could be used as a "157-nanometer resist" in chipmaking. The team has been making and testing several new fluoropolymers a week--more than 300 so far. They started with different monomers, the small molecular building blocks that link into giant polymer molecules like rubber, polyethylene, and PTFE. Mixing them with catalysts inside flasks, they produce spoonfuls of novel fluoropolymers with different sequences and arrangements of monomers. Then they screen each polymer in the hope of finding the right set of properties.

Resists are to chipmaking what film has been to traditional photography. They start life as liquids that spread into thin, uniform films when squirted onto spinning silicon wafers, and later dry. When ultraviolet light reaches the resist after passing through a quartz mask embossed with an opaque superfine pattern of chromium corresponding to circuitry, the exposed portions of the resist become soluble and can be washed away.

The nonexposed portions resist this process and become a stencil of the circuitry that will be microfabricated into the silicon wafer's surface. With portions of the resist gone, the exposed areas of the wafer can then be etched, converted into insulating silicon dioxide, overlaid with conductive aluminum, or spiced with electrically charged ions to fine-tune the conductivity of just that tiny exposed area of the chip. Microprocessors, such as Pentium IIIs with nearly ten million transistors, are built up using more than 20 mask-resist-and-processing cycles in clever sequences.

With each further step of miniaturization, materials researchers have had to come up with new resists that work with the shorter, finer ultraviolet wavelengths of light that make it possible to print tinier circuitry in the photolithographic process. The resists Du Pont is pursuing are intended to work with the 157-nanometer wavelength of light that chipmakers hope to use in next-generation processes. Off-the-shelf resists do not respond to this wavelength well enough for patterning the details of tomorrow's chips.

Hence, Du Pont is trying to put something new on the shelf. "The electronics industry," Feiring says, "wants to change as little as it has to and still get to the next generation of chips. So we seek to provide drop-in products that work at 157 nanometers." The finest-grained production processes today use 248-nanometer resists to create circuitry features as small as 150 nanometers. With 157-nanometer resists, it ought to be possible to create features as small as 70 nanometers, or about the width of a cold virus.

The R&D task is akin to finding an all-star football player with straight A's who performs classical ballet, has founded a great Internet startup, and also helps little old ladies cross the street. A 157-nanometer resist must be soluble so that it can be spun onto wafers in liquid form. It has to remain stress-free as it solidifies so that no chip-killing irregularities will form. It has to ignore most light yet interact in just the right ways with the 157-nanometer ultraviolet light coming from the source. What's more, the new resist has to be affordable and easy to make in adequate amounts to meet industry demand.

Feiring admits that no one yet knows how to put all of these qualities together in one material. There is no grand theory pointing to the one recipe, out of the infinitude of possible ones, that will yield the right stuff. That's why Du Pont is banking on trial and error. So far, Feiring says, the researchers have found three promising candidates.

A lot is riding on the effort. If Du Pont and the others in the race fail to come up with a good 157-nanometer resist, a completely different and less time-tested approach will be needed, sooner than expected, to keep the seemingly unstoppable trend of electronic miniaturization from coming to a halt.

The Du Pont team has another reason for sticking with traditional trial and error. Its tests typically require a whole teaspoon of polymer. But if tests require only a speck of stuff, new techniques make it possible to screen thousands of candidates at blinding speed. Symyx Technologies, a fast-growing, 180-employee materials innovation company in Santa Clara, Calif., is betting $90 million from a recent initial public offering that its souped-up trial-and-error approach will kick out a constant flow of new, profitable, high-tech materials. Symyx researchers have been developing tools and methods for doing as many trial-and-error experiments in a single day as used to be done in months or years. Mathematics alone, the company says, suggests that the rate of discovery ought to go up just as dramatically.

Two men are chiefly responsible for the creation of Symyx. About seven years ago Peter Schultz, now the director of the Genomics Institute of the Novartis Research Foundation in La Jolla, Calif., had an epiphany involving trial and error, materials innovation, and the immune system. He knew that to create at least one antibody that will attack a particular kind of bacterium, the human body makes zillions of slightly varying antibodies. Almost always, a few of these variants bind to some spot on the invader and from there mobilize a full-blown immunological defense. It dawned on Schultz that chemists could emulate this approach, making small amounts of zillions of candidate materials and then quickly screening them for interesting properties.

Word of Schultz's vision got to Alejandro Zaffaroni, Silicon Valley's premier maven for molecular technology, whose string of successes includes the founding of Alza, a maker of transdermal and other drug delivery systems, and Affymetrix, a maker of chips for identifying genes. Seeing big possibilities, Zaffaroni in 1994 bankrolled the founding of Symyx, whose first few employees set up shop in extra space at Affymetrix. Two years ago the growing company moved into its own building down the road, and recently it spilled over into a second building.

W. Henry Weinberg, Symyx's chief technology officer, bristles at skeptics who call his company's approach little more than an accelerated random search. Says Weinberg: "We are not chimpanzees sitting at a typewriter waiting for War and Peace to self-assemble." The periodic table harbors too many possible combinations, he emphasizes, for a totally blind approach to work: "There is not enough mass in the universe to make one microscopic speck of each of the infinite number of materials that it is possible to make." Accordingly, Weinberg has hired scores of topnotch people with the expertise and intuition to narrow the quest to promising categories of new materials. Collaborators, including Bayer, B.F. Goodrich, Dow Chemical, Celanese, Agfa-Gevaert, and PE Biosystems as well as the Department of Defense, have committed at least $85 million to targeting materials ranging from X-ray phosphors for radiography to specialty and commodity polymers.

The Symyx strategy seems simple enough. First, define the traits of a material you'd like to create. Next, synthesize up to tens of thousands of candidate materials on palm-sized wafers in a single day. Then, screen them the same day to see if they measure up. This is not your father's trial and error. "The old way, you can make and test one material a day; the new way, you can do 10,000," says Tom Mallouk, a combinatorial researcher at Pennsylvania State University. "This is a serious leap."

Even so, the sheer physical difficulty of carrying out mass screening looks formidable. How, outsiders wondered when Symyx was starting, could researchers come up with chemistry clever enough for making so many different materials so quickly? How could they screen them for properties as diverse as catalytic efficiency, superconductivity, and light emission when bombarded with electrons? Recalls Weinberg: "Three- and-a-half years ago everyone assured me I should keep my day job, and that I must have had a stroke. 'This is clearly undoable.' That was the feeling."

One of the doubters was Francis DiSalvo, a materials scientist at Cornell University who is looking for new "thermoelectric" materials for refrigeration systems with no moving parts. Now he's a believer who advises Symyx. Says DiSalvo: "It took a lot of engineering and a lot of effort, but I think combinatorial synthesis of materials is a very close reality," by which he means it will become a primary method for discovering new materials. Still, he notes, "they are waiting to hit a home run."

Symyx says it has made at least a base hit with its new X-ray phosphor. The project began about 2 1/2 years ago. Luc Vanmaele, who is in charge of scouting for new technology for Agfa, a Belgian photographic film and imaging unit of the Agfa-Gevaert group, approached Symyx with a conundrum. For 15 years Agfa has been searching for better X-ray phosphors--typically, crystalline materials that capture X-rays and then reemit the energy as light, much as the phosphors on television screens do after being hit by electrons. The phosphors are the key to making improved, higher-resolution digital "film," the basis for so-called X-ray storage screens used in such applications as mammography.

These are a big advance over single-use X-ray film, which must be developed chemically like a traditional photo. "You can use these screens over and over again," says Vanmaele. "And each time you read the image, you can digitally store it. You can send it over the Internet. You can even process the digital data into a classical X-ray image on film."

Agfa has its own phosphor Merlins, the counterparts of Du Pont's fluoropolymer wizards. "But even with our empirical base, we couldn't see how to go," Vanmaele recalls, adding that "there is no way you can use a computer and just design the phosphor you need." To top it off, "if we went the old way, we would need years and years to develop the chemistry and more years to test the materials that came out."

Vanmaele approached Symyx with the idea of collaborating. Agfa would supply its corporate experience and proprietary database on phosphors. Symyx would develop the synthetic techniques for rapidly making and screening potential phosphor compositions. The companies signed a contract in March 1998. Within three months, a team of Symyx chemists, engineers, software and informatics experts, and robotics engineers had developed chemical methods, special ovens for processing, automation tools, and screening protocols to get to work.

One of the synthesis methods relies on a vacuum chamber containing tiny coinlike samples of about 40 elements. Atoms from any of them can be deposited in different sequences onto tiny spots of a silicon wafer using a clever series of masks. The wafers thus become libraries of new materials, each one a little spot about the size of this printed letter "o." Last year the Symyx team made and screened over 50,000 materials in search of new and better X-ray phosphors. A quick-and-dirty screening method was to shine an ultraviolet light on the wafer libraries. Researchers could quickly see which spots might be promising just from whether they glowed or not. Then they could make larger amounts of the promising materials and test for other important properties, such as how long they last before breaking down.

Five products made it through the screens as leads, Vanmaele says, and one proved best. Agfa is now scaling up production of this phosphor and expects to market X-ray storage screens based on it sometime in 2001.

Symyx has promising leads for other new materials, Weinberg claims. Some are catalysts like those used to convert ethylene gas into milk jugs, food wrap, and a thousand other things. Running the search for new catalysts is Howard Turner, whose previous work at Exxon helped transform the way billions of dollars a year in polyethylene resin is made.

Like a kid in his own candy factory, Turner takes a visitor from room to room. One of the coolest research tools he shows off is a tabletop robotic gadget called the Parallel Polymerization Reactor (PPR), which can carry out 96 different reactions at once. A robot with a vacuum-operated pipette gathers and combines reactants from little vials to create solutions of different catalysts concocted by Turner and his chemist colleagues. These are injected into 12 rows of eight finger-sized vessels, into which ethylene gas is then pumped. When the gas molecules and the catalysts come into contact, the molecules start binding to the catalysts and linking into huge polymeric chains.

Day in and day out, this robotic system rustles up new catalyst formulations according to a script composed by Turner and his colleagues. It then causes them to react with ethylene and other monomers under different pressures and temperatures. All the while, sensors and a computer gather and record data about the reactions and products at a rate many times faster than by any other method. In one way, the PPR is already bringing in money for Symyx. Dow Chemical has bought the first production version of the device for more than $7 million.

Moving to another room with a one-of-a-kind screening instrument, Turner reveals that he's excited by recent measurements on one of the hundreds of thousands of catalysts that Symyx has made over the past few years. This particular new polymerization catalyst could be ten times better than anything else out there, he says. Lots of things could kill it, of course. It might not be stable enough, or it might not lend itself to large-scale production. Says Turner: "Every day's like a lottery around here."

Weinberg thinks that something even more important than any particular material may well emerge from his company. "The technology being developed in this building will change the way materials science laboratories will look ten years from now," he declares. Maybe. More and more companies are establishing their own combinatorial materials capabilities. If they and Symyx prove that rapid trial and error is better than the traditional version, then, as Francis DiSalvo puts it, "everyone and their grandmothers will be doing it."

That thought gets a cool reception from theoretical physicist Alex Zunger, a leading apostle of the diametrically opposed computational approach. Zunger is a 20-year veteran at the U.S. Energy Department's National Renewable Energy Laboratory (NREL) in Golden, Colo. Over the years, NREL has developed, among other things, photovoltaic materials and devices for converting sunlight to electricity at record efficiencies. Even light-speed trial and error, Zunger says, still amounts to groping in the dark. "You declare ignorance and say, 'We will try everything.' "

Zunger thinks he has a better way. Instead of trying everything, researchers can explore what he calls the "quantum architecture" of new materials. "That way," he says, "you can declare the winner ahead of time."

He goes about this by sitting at a computer he has programmed with sophisticated mathematical models of materials' behavior at the atomic level--in his case the behavior of semiconductors for microcircuitry, solid-state lighting, and solar-energy devices. He types in the attributes of the material he seeks, and then, in relatively short order, the computer stipulates the proportions of the virtual material's atomic ingredients as well as the molecular architecture that ought to lead to the desired behavior. Then it's up to Zunger's more experimentally minded colleagues to turn the virtual material into real stuff.

It sounds utopian. But despite the absence so far of blockbuster materials discovered by theory and computer, the approach is so unlike any previous one that practitioners often sound spiritual about the possibilities. Says Art Freeman, a widely respected computational pioneer at Northwestern University: "What is happening with computational materials science is the dawn of a new age."

Zunger has a special affection for semiconductors. Silicon is the superstar of these materials, but its cousins are found in applications all over the place. Semiconducting crystals of gallium arsenide and aluminum gallium arsenide are the stuff of the little solid-state lasers that read compact disks. And those little amber, red, and green light-emitting diodes (LEDs) shining around the house--and increasingly in brake lights, traffic signals, and other places--are made of semiconductors like aluminum indium gallium phosphide and indium gallium nitride.

Perhaps the most tantalizing fact about semiconductors is that the ones used now--for computer chips, solid-state lasers, solar cells, light and radiation detectors, and LEDs--are variations of a mere ten or so basic recipes. Thus, says Zunger, it's quite likely that breakthrough materials for future electronic devices remain undiscovered.

Zunger's way to search for them relies on a combo of fast computers, quantum mechanics, and computational techniques that can make otherwise impossibly huge calculations more digestible for today's machines. Last November he described in the journal Nature his newest computational method for delineating precise semiconductor structures with a specified "band gap," which is a key to each semiconductor's personality. The different colors of LED materials, for example, reflect their varying band gaps. The right band gap can also help a semiconductor survive in a harsh environment. Transistors made of some higher band-gap semiconductors like silicon carbide, for example, can work in hot places like jet engines, where signals from silicon become disoriented and senile.

To design a semiconductor by computer, physicists like Zunger first decide what properties they want the future material to have. They also stipulate the atomic ingredients they want the computer to play with, since giving it freedom to try everything in the periodic table would overwhelm it. It's not especially hard to gather this input. Scientists have long known which elements generally can be combined to make insulators, conductive metals, and semiconductors. They also know a lot about the physics of how band gaps relate to transistor action or light emission. In aiming to uncover new LED materials or new electronic materials, for example, they often can readily decide which band-gap values to plug into the calculation.

Even so, Zunger says the number of possible configurations his computers would have to consider could easily reach into the hundreds of trillions, which is far too many to be practical. One way he avoids this problem is by building into his programs efficient ways to sample that huge number of possibilities and detect when a sampled configuration is getting close to having the desired attributes. In his Nature article, Zunger told of using just such a technique to predict precise arrangements of aluminum, gallium, and arsenic atoms that have the largest possible band gap. Materials with larger band gaps, Zunger says, could help energy researchers develop more efficient photovoltaic cells as well as blue solid-state lasers useful in color printers and data-storage systems, and in improved night-imaging and vision devices.

The material structures he came up with are not simple. They're a special class of crystals known as superlattices. One described in his Nature paper has two atomic layers of gallium arsenide, topped by a layer of aluminum arsenide, topped by four layers of gallium arsenide, topped finally by a layer of aluminum arsenide. Superlattices can be grown, atomic layer by atomic layer, with techniques such as molecular beam epitaxy (MBE), which is like spray-painting with atoms. Zunger hopes his paper will inspire an MBE expert to make real versions of his virtual superlattice.

It's too early to say whether Zunger's superlattice will ever join the rarefied pantheon of important semiconductors. But he thinks his quantum architecture approach not only will lead to specific new materials but also could complement combinatorial research by helping to narrow dramatically the vast number of possibilities.

New semiconductors are not the only materials that are making their first appearance on computer screens. Another computational crowd is using theories and computers to devise more complicated stuff like metallic alloys and gypsum wallboard. Unlike semiconductors, whose properties are usually determined by the regimented structures of single crystals, these materials have a hierarchy of structures. The hierarchy begins with atoms and molecules and ascends through various levels--including fibers and tiny crystalline or amorphous grains--to fabricated shapes such as a gear or a sheet of wallboard.

One of the foremost players here is materials researcher Greg Olson, 52, of Northwestern University. In 1996 he and three colleagues decided that their theoretical grasp of how materials are formed and how the resulting structures affect their characteristics were far enough along to turn what Olson calls "materials by design" into a business. They founded QuesTek Innovations, a small but growing 18-person operation in Evanston, Ill.

The materials-by-design doctrine, as proclaimed by Olson, sounds like a neat idea--if it works. He says: "Instead of designing things with constraints due to available materials, invent new materials that will make your designs work." Easier said than done, of course. It has taken years for Olson and his colleagues to develop theoretical descriptions and computer models of the complex ways the normally invisible interiors of materials like steel alloys evolve as they are being made and, in the case of metals, as they solidify.

At times the researchers use theories like Zunger's, which focus on the quantum mechanical level, or the most basic behavior, of atoms and electrons. That's appropriate when the quest is for materials with specific electronic or light-handling properties, such as the speed at which electrons move in a semiconductor crystal or the specific wavelengths a material absorbs or emits. But quantum mechanics can even help alloy designers. Says Olson: "We have used quantum-level theories to predict all of the alloying elements you might want to use to enhance cohesion between an alloy's grains." That's one way to make alloys stronger, and Olson expects to secure patents based on these discoveries.

Yet quantum mechanics is just part of it. A more central and versatile element is thermodynamics: the play of heat and chemistry on the slam dance of iron, cobalt, nickel, chromium, molybdenum, and other alloying ingredients as they combine, react, redistribute, and settle into the internal and surface microstructures of alloys. Working mostly at this level of theory, QuesTek has developed its first three products, all specialized steel alloys.

One is GearMet C69, an alloy with seven elements. It's a member of a new class of gear and bearing alloys that combine hardened, damage-resistant surfaces with a tough, ductile core that can withstand service in demanding places like jet engines. The original impetus was to come up with an alloy that could shave a ton off an average airliner's weight. Reducing the size of gears in a jet engine would do the trick, but that requires tougher alloys to handle the added stresses that come with smaller contact areas in a gear train. QuesTek is now working to design such alloys under a contract with jet-engine maker Pratt & Whitney, which will evaluate their suitability for future engines.

GearMet C69 is prepared in a series of melting, cooling, and reheating steps, which cause metal-carbide particles to form and disperse throughout the alloy when it is subsequently molded to make a gear. These particles strengthen the metal by blocking tear- and crack-creating motions between its grains. Yet the alloy remains ductile and machinable enough to form into complex shapes. Its surfaces are superhard too.

A related alloy, GearMet C61, is a bit softer. Made into transmission gears, it's undergoing a second round of tests for the Newman-Haas car-racing team, formed in the 1980s by actor Paul Newman and veteran car racer Carl Haas. "The goal is to make unstrippable gears," says Olson. The first round of this project ended in frustration because the supplier delivered an alloy with metal-weakening contaminants.

Another of QuesTek's maiden products, Ferrium CS62, resulted from an attempt to design an alloy with two properties that metallurgists normally can't optimize simultaneously. Says Olson: "When you push mechanical properties like hardness, you usually can't go further with the stainless, corrosion-resistance property." In principle, parts made of the new alloy would last longer in outboard motors and in machinery on aircraft carriers.

Despite engineering tests confirming that the materials-by-design approach can zero in on superior metallic recipes from the infinitude of possibilities, the industrial world isn't beating a path to QuesTek's doorstep. "No one is buying production quantities of our materials yet," says co-founder Charles Kuehmann, 34. Government and industrial contracts keep the company going.

Olson says he's excited by a new research grant from the Department of Energy's Strategic Environmental Research and Development program. The task, says Olson, is "to make a big problem go away." The problem is the toxic cadmium anti-corrosion coatings on aircraft landing gear, which wear away and must be repeatedly applied. QuesTek's challenge, in Olson's words: "Can we design a steel alloy that gives the same mechanical performance as current landing gear steels, but also gives you the stainless property without cadmium coatings? The aircraft industry has spent tens of millions of dollars trying to find [such] an alloy, without success."

The QuesTek approach has attracted materials techies from some unlikely places. One is Brian Burrows, R&D vice president at USG, a leading maker of wallboard. "We have had a vision for about 12 years of going high tech," says Burrows. QuesTek is now part of his vision.

"We had lots of empirical data," Burrows explains. "We had rules of thumb, experience, and our base knowledge." But no one really understood the detailed mechanics by which wallboard cracks and fails. Gypsum, or calcium sulfate, reacts with water and additives to become the final wallboard material: mainly crystals of calcium sulfate dihydrate that organize into a porous labyrinth of plasterlike stuff. Burrows suspected that a solution to the cracking mystery could lead to wallboard in larger sizes that would make construction more efficient, to lower rates of product failure, and even to faster processes that could make huge capital expenditures for new plants unnecessary.

QuesTek came into the picture two years ago, when Burrows met Greg Olson during an engineering competition the two were judging at Northwestern. The first step in their collaborative effort was Olson's weeklong visit to Burrows and his staff at USG's research center in Libertyville, Ill. "He quizzed us about what we understand about gypsum and its properties, and how we manufacture it," Burrows says. "Then he went to QuesTek and drew up a model with all the elements that go into gypsum wallboard." One goal is to find new overall internal structures of the meat of wallboard--calcium sulfate dihydrate--that would result in a stronger product.

"We are one mile down a 100-mile track," Burrows admits. The challenge comes from the complexity of gypsum. It's not simple like a crystal; it's porous, varies in density, and has many imperfections. With QuesTek on the case, however, Burrows hopes that more details about how gypsum powder, additives, water, and heat become wallboard will help his company stay in the forefront of its industry.

While researchers pursue ever more complicated atomic combinations, an occasional surprise occurs on another front: altering the molecular structure of a single element. Last summer Richard Smalley of Rice University, co-winner of the 1996 Nobel Prize in chemistry, regaled the Senate subcommittee on science and technology with descriptions of an amazing new class of ultra-tiny structures known as carbon nanotubes. Like diamond and graphite, they're nothing but carbon atoms. In size, they're to a human hair what a human hair is to a sequoia.

Computer models of these carbon molecules, whose first physical samples were made in 1991 by researchers at Japan's NEC, look like chicken wire rolled into cylinders. "Nanotubes are incredible," Smalley told the Senators. "They are expected to produce fibers 100 times stronger than steel with only one-sixth the weight--almost certainly the strongest fibers that will ever be made out of anything." The catch: No one yet knows how to make them cheaply. One enthusiast is NASA researcher Bradley Files, a former student of Olson's. He and his colleagues zap finger-thick graphite bars with lasers to produce about a gram of carbon nanotubes a day. On the market, that smidgen would go for about $1,000.

At lower costs, Smalley has speculated, the possibilities are literally out of sight. Carbon nanotubes, he says, could be the right stuff for building elevators to space. He may be half kidding, but it's fun to dream. Just press the up button, and a load of pretzels and Pokemon videos could ascend along a nanotube cable 22,000 miles straight up to a receiving bay on a geostationary space station. All of this from merely rejiggering one of the periodic table's most common atomic ingredients.

Stories from FORTUNE's Industrial Management & Technology section can be found at www.fortune.com/imt. feedback: iamato@fortunemail.com