In 1987 Audrey Engelsberg, a graduate student at Rensselaer Polytechnic Institute in Troy, N.Y., was having a particularly bad day in the lab. "I was doing my doctorate thesis on laser-assisted chemical vapor deposition, looking at trimethylaluminum on silicon dioxide and aluminum oxide surfaces," she recalls. "Those are typical surfaces you'd find in semiconductor manufacturing. What I wanted to do was write metal lines using trimethylaluminum as a precursor material that was able to break apart the metal groups to put aluminum metal down. . . . Instead of making aluminum metal, I took off all the CH3 groups, took off all the carbon contamination on the surface, and just left the aluminum oxygen silicon structure intact." Much like Newton as he sat and rubbed the welt on his head, the pleasure of this discovery did not immediately edify Engelsberg. "It was very disconcerting, because I had just screwed up my thesis. I'm thinking I'm never going to get out of Troy," adds Engelsberg. She then lamented to friend Joseph Dehais, who asked if the failure could be repeated. "So I did the experiment a half dozen times and got the same result." What those repetitions showed was that Engelsberg had stumbled upon a "dry" cleaning method that could remove minute contaminants in an extremely environmentally sensitive manner. Engelsberg -- who did eventually earn her Ph.D. -- and Dehais sought legal counsel and filed for a patent in 1988. Approval came in 1991, followed by the formation of Cauldron LP. The Radiance Process is marketed by Radiance Services Co., a Bethesda, Md., subsidiary of Cauldron LP's corporate general partner Cauldron Co. The technology, trademarked as the Radiance Process, has been demonstrated to clean -- without water or chemicals -- contaminants to at least 0.1 micron from wafers, semiconductors, flat-panel displays, optics, fiber-optic cables, and industrial metals. Engelsberg is now director, vice president, and chief technical officer of Radiance Services Co., and Dehais is chairman. They are joined by Donna Fitzpatrick, CEO and president, formerly an undersecretary in the Dept. of Energy. "We have now been patented in Europe, Canada, Australia, and Taiwan," says Fitzpatrick."We have patents pending in the former USSR, Brazil, Mexico -- 34 jurisdictions in all. We have filed four improvements and extensions of patents in the U.S. on the process: how it works and what it can be used for. We are really internationally patented, probably 80% to 90% of the industrial world." The Radiance Process operates at the substrate surface by breaking down contaminant surface bonds in a multiple photon interaction. It does this by using a photon source (preferably a deep ultraviolet excimer laser), an optical train, a substrate fixture, and inert gas delivery and exhaust systems, carrying nitrogen or in some cases argon. The process breaks the bonds of particulate and thin-film contamination without ablation or melting. The removed nontoxic microcontaminants -- such as dust or fingerprints, etc. -- can be exhausted directly into the environment; toxic particles are trapped for proper disposal. "The light lifts material from the surface, and the gas just sweeps it away so that you don't use any liquids, you don't have any waste materials, you don't have any residue left on the surface," says Fitzpatrick. "The energy fluxes that we use are so low that we're not ablating or melting the surface or causing internal damage -- to quartz, for example. "There are lots of industries . . . that need very clean surfaces, and they want very much to get away from washing. They don't want any more acid, bases, solvents, water. This is all very expensive for them and often doesn't do the job at the level of cleanliness they now need for precision work." As the industrial need to clean ever-smaller particles arises, Fitzpatrick says the process will keep pace. "We claim 0.1 micron because that is what we can document reliably. That's not the limit of the process' effectiveness." The semiconductor industry has begun to pay particular attention to the work done at Radiance Services. Semiconductor manufacturing facilities, called fabs, can't rise fast enough to support the demand for electronic devices, and their operators are spending millions of dollars on ultrapure water and chemical washes. But that industry carries a culture that fears the cutting edge and eyes advances suspiciously. Because of that climate, an agreement between Radiance Services and the Industrial Affiliate Program of Interuniversity Microelectronics Center (IMEC), Leuven, Belgium, carries weight with the semiconductor industry. IMEC develops, characterizes, and evaluates microelectronic technology. Radiance is participating in IMEC's Ultra Clean Processing Program, along with chipmakers such as Intel, Motorola, and Texas Instruments, says Engelsberg. Radiance Services installed its technology at IMEC this summer and the R&D house began testing in September. "We have become an industrial affiliate there and worked out with them a research agenda," says Fitzpatrick. "First, they're going to validate some of the things that we've already done -- you won't have to take our word for it -- then they will do some more advanced work that will take advantage of their very large array of diagnostic tools. Because they are a working semiconductor fab, they will be able to test the process at every stage of the chip-manufacturing process and compare it to wet chemical cleaning for efficiency and cost, for example." One current standard in chip cleaning is a wet method known as RCA Standard Clean, and Radiance estimates that 40-year-old process to be at least three to seven times more expensive than the Radiance Process. At an equal-yield performance, Radiance projects its process would cut capital fab costs by 10% and operating costs by 20% per year. Even with such a promising cost model, Engelsberg does not expect rapid conversion by the semiconductor industry. "I think adoption is paradigm-shifting. You're asking people to work in a field that they're not used to, they didn't know really exists. When you're used to working in thermodynamics, it's pressure, temperature, force. . . . There is a groundwork of information you can always fall back on. We say to someone, 'You have to look at this in terms of quantum mechanics and bonds, or energy.' Their eyes start to glaze over." Although the cultural shift may be time-consuming, Fitzpatrick says the physical adoption of the process is simple. "Everything is commercially and readily available off the shelf, and we don't use any strange bits of hardware, anything exotic," she says. "It's just knowing how the light should be applied and how the gas should be flowed over the surface that's important . . . . It doesn't require any infrastructure support beyond a gas line and a small exhaust fan and an electric cord. "The process now for cleaning semiconductor wafers uses a tremendous amount of water and chemicals, such as hydrochloric acid, hydrogen peroxide, sodium hydroxide, isopropyl alcohol, and lots and lots of high-purity deionized water. All of this involves pipes, tanks, pumps, handling and mixing apparatus, monitors, heaters, coolers, drains -- the whole works. A typical semiconductor plant will use 3 million to 6 million gallons of water per day, most of it for cleaning. "A typical plant will have its own deionized water plant, which costs about $60 million. In fact, about 10% of the whole capital cost of the plant-a plant costs $1 billion dollars and up today-is just the equipment and infrastructure to support cleaning." "What they claim -- and I think their claim should be taken very seriously -- is very promising, both for its environmental impact and also for its technology power of being able to remove small particulates in very, very small and complex geometries," says Nader Pakdaman, senior analyst-semiconductor materials and manufacturing, Dataquest Inc., San Jose. John Dennis, president of Mr. Clean Consulting Inc. and Mr. Clean Conferences Inc., a Black Forest, Colo., company that runs conferences aimed at cleaning-contamination control, says he has been aware of the Radiance Process for about two years. "It certainly looks like a very viable new technology that has some great potential," says Dennis. "One, it seems to be doing a very, very good job. Second, you've essentially got nothing going down the drain, which environmentally is extremely good." The testing underway at IMEC may yield a report as soon as early next year, says Fitzpatrick, but both she and Engelsberg say Radiance is proceeding concurrently with other commercialization efforts. Two equipment manufacturers have signed vendor licenses for the technology. IMEC is "not holding us up, both because there are other people that are convinced that we can clean at least bare silicon wafers, and also there are these other industries that won't be affected by the IMEC results," says Fitzpatrick. "Our licensed equipment vendors are doing design work for a wafer cleaner for the Dept. of Defense, a system for cleaning photostencils and lithography stencils, a system for tire molds, a system for cleaning metal foils." The process also is being used to clean ancient metal artwork, and could also be upscaled to clean auto and aircraft structures, thus improving the effectiveness of glues and reducing the reliance on welds in those industries. "Frankly, we have not had the imagination to figure out for ourselves what this thing is good for," says Fitzpatrick. "People are coming to us and saying, 'Here's my problem.' . . . It's working out to be better than we had dreamed."