Blue Chip

I press a button on the pocket light-emitting diode tester and three specks of plastic and semiconductor shoot out blue and green rays intense enough to hurt my eyes. The two blue devices emit a furious cerulean with just the slightest hint of violet. The green is sharp and richnot that ghastly yellowish hue that had to do if you wanted a "green" LED until recently.

Until, that is, the man who is grinning at me, Shuji Nakamura, got some very bright ideas.

Nakamura, the newest addition to the engineering faculty at the University of California at Santa Barbara, stunned the semiconductor world late in 1999 when he revealed that was leaving Nichia Corporation, a once small and obscure Japanese maker of phosphors for cathode-ray tubes and fluorescent lights. Thanks to Nakamura, Nichia now fabricates the worlds best blue LEDs, the best green LEDs, and the worlds only commercially available blue-violet semiconductor lasersdistinctions that have pushed Nichia to the verge of being a half-billion-dollar-a-year company with sales all over the world. Most remarkable of all, Nakamura, working alone and with a tight research budget, managed to open up a lead measured in years over some of the titans of U.S., Japanese, and European industrial research. Even more incredible, the companies, including Hewlett-Packard, Xerox, Stanley Electric, Sharp, Sanyo, Sumitomo, Toshiba, Toyoda Gosei, NEC, Sony and Philips, still have not caught up. The situation is without precedent in the half-century history of semiconductor research.

Analysts estimate that those companies, along with a couple dozen universities, spent roughly $1 billion in pursuit of blue-light devices since the 1960s. And it is easy to understand why. For more than 25 years, LEDsthe most efficient lights yet producedwere like a third of a rainbow. Red, orange, yellow and that yellowish-green were all you could get. Engineers wanted blue and true green because with those colors, along with the red they already had, they could build fabulous things, such as a a white-light-emitting device as much as 12 times more efficient and 12 times longer-lasting than an ordinary lightbulb. Small wonder, then, that analysts say LEDs are poised to revolutionize the lighting industry and move beyond their familiar role as mere indicator lights on electronic equipment. In the meantime, colored LEDs are being deployed as traffic lights and in displays, the biggest being the eight-story-tall Nasdaq display in New York Citys Times Square.

The potential bonanza does not end there. The blue-light semiconductor laser, an offshoot of the LED, also has tremendous commercial possibilities. These are linked to the fact that the wavelength of blue light is about half that of the infrared semiconductor lasers typically found in CD players and laser printers. A wavelength half as long translates into a cost-free quadrupling of the amount of data that can be put on a CD or in the resolution of a laser printer.

Most of the milestones on the way to these optoelectronics triumphs took place, oddly enough, on the island of Shikoku, something of a backwater in the Japanese chain. There Nakamura was born, raised and educated at the University of Tokushima. He graduated from the "medium-class" institution, as he describes it, with a masters degree in electrical engineering in 1979 and aspired to a career in R&D at a company like Sony or Toshiba. But a professor discouraged him, noting that if he went off to a big city, he would be just another salaryman, unable to afford a house. With a wife and a baby on the way, he reluctantly decided to stay on Shikoku. He got an R&D job at Nichia, which then had about 200 employees, was doing about $30 million a year selling phosphors and was looking to get into the market for LEDs. After a researcher changed jobs, the companys entire R&D department consisted of Nakamura, his boss and one other employee. He was the only person with a masters degree and the only electrical engineer in the company.

"If I had gone to a big company, it would have been very difficult for me to do research freely," he says in retrospect. "At a big company, say Sony, there are very, very good researchers. So I would have to ask them what I could do."

As it turned out, there wasnt much freedom at Nichia, eitherat least not initially. Just after Nakamura joined the company, a sales manager asked Nakamuras boss to direct the young researcher to develop gallium phosphide crystals for use in red and yellowish-green LEDs. And because Nichia was small, Nakamura could not buy the necessary equipment; he had to make it himself.

In the horizontal Bridgeman method then used to produce such crystals, the source materialsin this case, gallium metal and phosphorouswere melted in a 1,500 Celsius, hydrogen-oxygen-fueled furnace and reacted on quartz in a vacuum chamber. Nakamura had to learn not only how to make that kind of furnace but also how to weld quartz, a tricky and esoteric task.

The real challenge came after he built the equipment. At 8:00 each morning, he began preparing the quartz and the materials. The reactants went into the furnace, behind a protective shield in his laboratory-office, between 10:00 and 11:00. Then he raised the temperature until 2:00 or 3:00, starting the reaction after the temperature reached 1,100 degrees. Every day, he burned through two tall cylinders each of hydrogen and oxygen.

About three times a month the quartz would crack, letting in oxygen that would react with the phosphorous and cause an explosion. It always occurred around 5:00, when his co-workers were going to their cars in the parking lot, just 150 meters from his laboratory. Dense white smoke would billow out of his windows. Laughing heartily, he recalls that "people would come in asking, 'What happened? Nakamura, you still alive?' By the fifth or sixth time, they didnt bother to come in."

Every time there was an explosion, he had to spray water inside the furnace to keep the remaining phosphorous from igniting. And after the smoke cleared, he had to write a report. His bosses were not amused. "I thought my life will be like this forever," he says.

After three years, he finally got gallium phosphide crystals as good as anything Nichias larger rivals were putting out. Unfortunately, all things being equal, most customers bought from the much larger, well-known companies, leaving Nichia only a sliver of the pie: about $10,000 a month in gallium phosphide sales.

Disappointed, the sales manager made it known he wanted Nakamura to try again with gallium arsenide crystals, also used in LEDs. So in 1983, Nakamura started working with gallium arsenide. His experience was much the same, except this time, when the furnace blew up, it released lethally toxic arsenic oxide. Nakamura had wisely relocated the furnace to a room separate from his laboratory-office. He would wait until the smoke settled and then go in to clean up the mess wearing a full-body covering and breathing from a respirator. "I was very lucky because I was never injured," he explains, laughing at the memory of himself in his "space suit."

After all that, the commercial reception to the gallium arsenide crystals was much the same as it had been with gallium phosphide. Nichias larger rivals again dominated. Worse, the sales staff was starting to blame Nakamura for the disappointing sales, even though his crystals were as good as anything on the market.

Around 1985 the sales manager had another idea: instead of making the crystals for LEDs, Nichia would make the complete LEDs. For this, Nakamura had to teach himself how to do liquid-phase epitaxy, the technology commonly used to grow the light-emitting semiconductor layer on a substrate. He pored over U.S. and Japanese patents and articles in periodicals like Applied Physics Letters and Journal of Applied Physics. By 1988 he had developed excellent red and infrared LEDs by growing gallium aluminum arsenide on a gallium arsenide substrate. But so had Sanyo, Sharp, Stanley, Rohm, Panasonic, Toshiba and many others.

"Gradually, my company became mad at me," he says. Resentful of even his relatively meager R&D expenditures in the face of disappointing sales, some of his more senior co-workers wanted him to resign. Nakamura, for his part, became "very angry." He had also had an insight. "A small company like Nichia should do niche products," he says. "After 10 years, I could understand those things." From his voracious reading, he knew that a blue-light-emitting semiconductor device was the holy grail of optoelectronics, and he resolved to get into the fray.

He remembers his boss, the R&D manager, responding along these lines: "You are crazy. All the big companies and universities havent been able to do that. Why do you think you can do it at a small company?"

So in January 1988 he bypassed his boss and marched into the office of Nichias CEO, Nobuo Ogawa, with a list of demands. He wanted about $3.3 million in research funding to work on blue-light devices and also a year off to study metallorganic chemical vapor deposition, or MOCVD, at the University of Florida. MOCVD was then emerging as the technology of choice for producing exotic semiconductors, such as the ones capable of emitting blue light.

Nakamuras move would probably raise a few eyebrows at most in a small American start-up company, but it was absolutely outrageous in the feudal, seniority-based Japanese system. "I was very mad," he explains, when asked what prompted his ultimatum. "I wanted to quit Nichia. I didnt care about anything. It was OK for them to fire me. I was not afraid of anything."

Much to his amazement, Ogawa simply agreed to all his demands. Nakamura later heard that the CEO told confidants, "Nakamura is a big liar. But he is the best researcher." The liar part apparently referred to what was perceived as Nakamuras failure to generate sizeable revenues. But the compliment about his research skills suggested an underlying empathy between Ogawa and Nakamura. As a maverick chemical engineering researcher in the early 1950s, Ogawa managed to make better phosphors than his competitors. And other than Ogawa and Nakamura, no one else at Nichia had created a new revenue-generating product, Nakamura says. "He had developed phosphors himself, so he knew how difficult it was to make a product just by reading papers and patents," Nakamura explains.

Three months later, in March 1988, Nakamura left for Florida. His co-workers werent too sorry to see him go: though he got what he wanted in his meeting with Ogawa, news of his brash maneuver did not sit well with some other employees. "When they met me, every time, they would ask, 'Why are you still here?'" he recalls. In Florida, some more rude surprises were in store. First, the faculty member who had attracted him to the university with promises of access to MOCVD systems disclosed that he had just lost control of three machines to another faculty member in a turf battle. All the professor had were parts that could be assembled into an MOCVD system. And it turned out that Nakamura was just the guy to do it: lacking a doctorate and a list of published papers, he was "treated like an engineer, not a researcher."

So he put together an MOCVD machine. He did it in 10 months, working 7 days a week, 16 hours a dayabout four hours a day more than his usual work regimen in Japan. He also had a significant career-related insight: "The most important thing I learned at the University of Florida is that a Ph.D. and writing papers is very important in the United States."

Returning to Nichia in March 1989 to begin work on blue-light devices, he had to choose between the two main semiconductors then being developed: zinc selenide and gallium nitride. No contest: he picked gallium nitride, mainly because all of the giants of industry and academia were pursuing the former. "I didnt like competition with big companies," he says. "Even if I succeeded with zinc selenide, I wouldnt be able to compete with big companies."

The material he chose to work with, gallium nitride, had a well-deserved reputation as one of the most difficult of semiconductor materials. "At that time, it was crazy for people to try gallium nitride," he acknowledges. He didnt really think he was going to hit a grand slamproducing commercial blue LEDanyway. He was just hoping to bunt his way to first base: "I wanted to write a paper. I could expect that I could write a paper if I got even a basic result of some kind."

The big players were mostly ignoring gallium nitride for a host of what seemed to be good reasons. The density of defects in the best gallium nitride crystals, at 1.0¹⁰ dislocations per square centimeter, was ten million times higher than what was believed necessary for commercial success, especially in a laser. Even more inauspicious, there was no practical way to make p-type gallium nitride, which has an excess of electron deficiencies known as "holes." Any semiconductor device requires both p-type and n-type materials.

Recalling his big lesson at the University of Florida, Nakamura resolved to write as many research papers as possible. The strategy was perilous, because CEO Ogawa, fearing disclosure of technical secrets, strictly forbid Nichias employees from writing papers or speaking at conferences.

Over the next 10 years, as he coaxed more and more light out of gallium nitride and shot far ahead of his competitors, Nakamura put together a string of achievements that for genius and sheer improbability is as impressive as any other accomplishment in the history of semiconductor research. And it is all documented in a trail of literatureessentially all of it written in secretalmost as stunning. Between 1991 and 1999, he authored or co-authored 146 technical papers, 6 books and 10 book chapters on gallium nitride semiconductors. Nakamuras publishing exploits largely escaped the companys notice because he published his research in fairly obscure journals. He was nevertheless caught once or twice and agreed to abide by the no-publishing rule but then continued to quietly defy it. Nichia was no victim, though: because of Nakamuras work, the company was awarded 68 patents in Japan and 13 in the US, with many others submitted and yet to be decided. By 1994 Nakamuras output had been so prodigious that the University of Tokushima awarded him the thing he seemed to covet most: a doctor of engineering degree.

By 1992 Nakamura had developed a heat-based process to produce p-type gallium nitride in commercial quantities; all commercial p-type gallium nitride is now produced with his method. Around the same time, in the early 1990s, he was striving to improve the quality of the indium gallium nitride film, grown on gallium nitride, in his device. This film was the light-emitting "active" layer, where electrons and holes combined and released photons.

The foundation of Nakamuras success was a deep understanding not only of semiconductor crystal growth but of the machines that accomplished it. He recognized that commercially available MOCVD machines could not grow indium gallium nitride films good enough to emit light brightly, so he set about modifying his setup. From his years of building reactors and furnaces, he knew how to weld quartzwhich enabled him to quickly modify the conduits that conveyed the reactants in an MOCVD machine. Also, having assembled an MOCVD system, he was intimately familiar with how it worked.

No one else doing MOCVD research then (or now) had Nakamuras range of skills. Thus, he could perform modifications himself that for another researcher would probably mean writing a specification, completing a purchase order and bringing in a vendor or technical specialist, in other words, a delay of a couple of months rather than a couple of hours.

Every morning, Nakamura modified the reactor. Every afternoon, he grew four or five samples. After about two years of seven-day weeks, he hit on the configuration that would ultimately put him ahead of the pack. He called it "two-flow MOCVD." In a conventional MOCVD system, semiconductors are created as reactant gases flow over a substrate, parallel to its surface. In Nakamuras system, one gas flows parallel and the other flows perpendicular to the surface. The configuration suppresses thermal convection currents on the substrate and cools the reactant gases just before they react. This lower temperature, in turn, leads to more stable reactions and higher-quality films.

In mid 1991 he produced a film in which the electrons were unusually mobileabout 43 percent more mobile than anything previously seen in the semiconductor. "It was the most exciting day of my life," he recalls. "I had never been number one in the world. Now I was number one." Nine years later, hes still there.

Tweaking the two-flow MOCVD system sufficed to produce first blue, and then green, LEDs of higher and higher brightness. But to get a dependable laser, he had to tackle the defect-density problem. The breakthrough here was inspired by a talk by NEC researchers, in the spring of 1997, at a panel session in which Nakamura was the chairman. After making some major inferences from the researchers purposefully oblique statements, Nakamura began exploiting a technique known as lateral overgrowth. By growing a layer of silicon dioxide strategically, he managed to block some of the defects in the gallium nitride crystal. The method could reduce the defect density only in a minuscule volume of the crystal, but the low-defect space was large enough to contain the indium gallium nitride active region. By using lateral overgrowth, within six months, he increased the lifetime of his blue semiconductor lasers from about 300 hours to several thousand hours. By the end of the year, he says, he achieved the 10,000 hours needed for a commercial product.

In 1999 Nichia began selling five-milliwatt blue semiconductor lasers. In the laboratory, meanwhile, Nakamura and his colleagues kept improving the lasers, getting higher power levels and frequencies. In October 1999, Nichia began selling a five-milliwatt violet laser with a wavelength of 405 nanometers, the shortest ever for a semiconductor laser. As of this writing, no one else has succeeded in making such a laser. In the laboratory, Nakamura had blue lasers with power levels above 30 milliwatts. He declines to give a precise figure, but the power levels necessary for laser printers are around 50 to 60 milliwatts.

Last October, having done everything he wanted to with gallium nitride and weary of a Japanese industrial R&D system that he characterizes as "communist," Nakamura decided to leave Nichia. Though his inventions had swelled Nichias profits from under $100 million to over $400 million, Nakamura was being paid only $100,000 a year, he says. In 1999, Nichias products accounted for 40 percent of the $300-million-a-year market for nitride LEDs, according to Strategies Unlimited of Mountain View, Calif. That market is growing at a rate of 40 to 50 percent a year, says the company.

Within a month, as word got out of his decision to leave Nichia, Nakamura was offered professorships at 10 U.S. universities and two European ones, and at five U.S. companies. One of the companies (he declines to identify it) offered him a salary of $500,000 a year and stock options worth $10 million. "It was unbelievable to me," he says. He was set to sign with the company, but a professor at one of the universities that was courting him advised that if he took an industrial job, Nichiawhich held the patents on all his gallium nitride breakthroughswould sue him the moment he did anything even remotely related. "I was afraid of those things," he explains. In effect, he was a prisoner of his own painstakingly written patents. Ironically, patenting all of the advances was his idea, not his superiors, he claims.

After mulling things over, Nakamura accepted an offer from the University of California at Santa Barbara, which already had an impressive gallium nitride effort and promised him an MOCVD laboratory of his own (which would supplement another one the university already had). When he left Japan, five television stations showed up at the airport to videotape his departure. When he got off the plane in Santa Barbara, another Japanese station was there to record his arrival, and yet another was at U.C.S.B. to document his first day at his new job.

A 46-year-old who has already revolutionized optoelectronics and earned a spot in the semiconductor pantheon, Nakamura is as restless and driven as ever. Asked what he wants to do now, he is blunt, if a little nonspecific: "I want to achieve the American dream," he exclaims, laughing. "Thats why I came here. I couldnt achieve the American dream in Japan. Here, I can start a venture companyin five or 10 years, if I could invent a new device. To get funds, Ill have to keep working on nitride devices, because I am famous for nitride devices."

Thats not all hell be doing; for the first time in his life hell have to teach graduate students and write proposals for research grants. With characteristic candor, he admits hes not looking forward to either. "I have to read these books," he says, waving a hand toward a wall of science and technology texts in his small, neat office at U.C.S.B. For him, reading textbooks is like rehashing what is already known. "I dont like to imitate technology developed by other groups," he adds. "I like to do new things."

In the months and years to come, the world will be seeing more and more of some of Nakamuras "old" things. His green LEDs have begun replacing the incandescent lightbulb and green glass in traffic signals; the semiconductors last much longer and use about a tenth of the electricity, saving about $14 per year per traffic light. Dozens of huge full-color LED displays have also been built, many by Sacco Smartvision, a Montreal-based company. The companys flagship is a 27-meter by 37-meter monitor in New York Citys Times Square. Built for the stock exchange Nasdaq, the display has over 18 million LEDs, of which 4.6 million are blue and 6.9 million are green (the rest are red).

Meanwhile General Electric, Philips and Siemens are all trying, along with R&D partners, to build solid-state lighting based on gallium nitride LEDs. Today, white-light-emitting LEDs are expensive to produce and their luminous efficiency is generally no higher than that of ordinary incandescent bulbs. So for the moment they are limited to a few very small nichesfor example, in automobiles and portable lighting, where their ruggedness and small size are more important than their high cost. Researchers are confident, though, that over the next several years the efficiency of white LEDs will go up and the costs will come down, pitting them more squarely against incandescents.

Products based on the blue and violet lasers are also in the works. Sony and other companies are working on DVD and CD players based on blue-violet lasers, which should start appearing on the market next year or in 2002. As for laser printers, Xerox is known to be working on one. But it is based on the companys own blue-light semiconductor laser, which is believed to be inferior to Nichias.

As products based on Nakamuras work start pervading the market, his legacy in semiconductors may outshine his other bequest. At a time when invention tends to be dominated by faceless teams at huge multinational corporations, he showed it is still possible, even in semiconductor electronics, for an inventor with enough talent and determination to triumph despite daunting disadvantages in resources. As he strives to achieve his own version of the American dream, he has lifted the spirits of many others struggling toward theirs.