by Jon Gertner
So what was the Bell System waiting for? Kelly acknowledged that the phone company would capitalize on the transistor long after “other fields of application” such as the home entertainment industries.4 The recent Justice Department antitrust suit, which was now moving forward, was a stark reminder why: The phone company was a regulated monopoly and not a private company; it had no competitors pushing it to move forward faster. What’s more, it was obliged to balance costs against service quality in the most cautious way possible. “Everything that we design must go through the judgment of lots of people as to its ability to replace the old,” Kelly told an audience of phone executives in October 1951. “It must do the job better, or cheaper, or both.”5 Any element within the system was designed (by Bell Labs) and built (by Western Electric) to last thirty or forty years. Junking a functional part before its time had to be economically justifiable. And if it wasn’t justifiable on economic grounds, it had to be justifiable on technological grounds.
The transistor could not be justified on technological grounds—at least not yet. The awesome intricacy of the phone system was not hospitable to sudden changes, even when they allowed for stark improvements. In time, Kelly remarked, the traffic and data needs of the system would require replacing tubes and switches with transistors.6 The Labs’ managers had already begun planning for new transistorized phones and an electronic switching station that, as it turned out, would take nearly twenty years to fully deploy. In the meantime, the system was working reliably and was giving customers a reasonably good product for their money. AT&T shareholders, who now numbered more than a million, making AT&T not only the world’s largest corporation but the most widely owned, also liked its steady profits and sizable dividends. On Wall Street, brokers called the dependable AT&T a “widows-and-orphans” stock; if you couldn’t rely on anyone else, you could rely on Ma Bell. The paradox, of course, was that a parent corporation so dull, so cautious, so predictable was also in custody of a lab so innovative. “Few companies are more conservative,” Time magazine said about AT&T, “none are more creative.”7
PERHAPS, as Kelly realized, one need not rush the phone system’s evolution. But the business of communications was different than the science of communications, and in the science, Kelly’s employees could do whatever they liked to push ahead. By the mid-1950s, one of the essential questions facing researchers was whether the transistor of the future, and therefore the future of electronics, would be fashioned out of germanium or silicon. All of the transistors so far had been made of germanium. But there were a number of reasons to favor silicon.8 Germanium is far rarer than silicon, which can be derived from sand. If the transistor industry were potentially as enormous as Fortune magazine envisioned, germanium’s scarcity (and its high price) could at some point limit the industry’s growth. More fundamentally, germanium had performance issues. For reasons owing to the behavior of its electrons, as germanium transistors got hotter, they became less reliable. In a very warm environment—150 degrees Fahrenheit and over—they were often useless.
In late 1952, a young chemist with a PhD from Princeton named Morris Tanenbaum joined the Bell Labs research department. As was typically the case with new recruits, Tanenbaum was encouraged by his supervisors to look around the Murray Hill complex—literally to drop in on neighboring laboratories for a few days—and see what kind of research interested him before he settled into a particular project. “The transistor was just a few years old then,” Tanenbaum recalls, “and there was still a lot of awful good work going on with germanium crystals. The question was: Are there better semiconductors than germanium?”9 Tanenbaum found this question intriguing. He began his search by obtaining a variety of materials—aluminum, gallium, indium, and so forth—in the purest states he could find. Usually they came from DuPont, and were sold as powders or in small chunks. He or his assistant would melt them into crystals. When Tanenbaum began his tests, he tried a rare element called tellurium; soon after, following some work he had heard about in the Siemens labs in Germany, he began working with a compound of indium and antimony. At that point, Shockley got involved.
“Bill said, ‘Look, germanium has a number of properties that really aren’t very good,’” Tanenbaum recalls, “‘so let’s really look at silicon.’” Several years before, Bardeen and Brattain and Brattain’s lab mate Gerald Pearson had tried to make silicon transistors but had been discouraged by the results. Shockley wanted to try again. “I was asked if I would like to work with him to do that,” Tanenbaum recalls. “I had heard about Shockley’s reputation,” he says, noting that it almost scared him away from the project. “Shockley had an ego substantially larger than a house—and he deserved that. He was a very brilliant guy. But it turned out I never had any problems. Being a physicist with Shockley, well then, you had better be very, very good or you’re going to have a hell of a time.” But being a chemist, with knowledge outside of Shockley’s sphere of expertise, put Tanenbaum beyond the reach of his bullying.
Working with silicon, as Tanenbaum soon discovered, was far more frustrating than working with Shockley. Still, the rewards of success seemed obvious to most people in the solid-state area. “If anyone could actually build a silicon transistor,” Tanenbaum remarks, “then we knew it would even work in boiling water.” DuPont was already selling what it called “pure silicon” for semiconductor devices; according to the Wall Street Journal, the company was then charging $430 a pound.10 The product, which arrived at Bell Labs as a powder, was a useful starting point. But it was only a starting point. Silicon has an extraordinarily high melting point, about 2,500 degrees Fahrenheit, and is easily tainted in the “melt” by all sorts of other unwelcome elements. The residue from the crucible that it has been melted in, for instance, can easily ruin its potential as an electronic device. Indeed, while it was true that the metallurgists at Bell Labs wanted their silicon to have small amounts of impurities, they only wanted certain kinds of impurities, so as to affect the conduction in useful ways. They wanted a few atoms of one specific type of element to transform the silicon into the negative n-type and a few atoms of another specific type of element to transform the silicon into the positive p-type. Then they wanted to join the two. The junction between the p-type and n-type was where the movements of electrons and holes resulted in transistor action.
Success, to a certain degree, came quickly. After several months of work, Tanenbaum, with the help of his lab technician, Ernie Buehler, “grew” a long crystal of silicon through a complex process that involved varying the rate at which the crystal was being “pulled” up from the molten silicon. By varying the rate, the men could alternate the amount of n-type and p-type impurities that were incorporated into the crystal. When they were done, this long crystal—about four and a half inches long and three-quarters of an inch wide—had dozens of tiny n-p-n sandwiches all stacked up, giving it the appearance of a thin rod made up of tiny gray slices piled on top of one another. After slicing one of the n-p-n portions from the tiny stack, the men fashioned, in January 1954, the world’s first working silicon transistor. A few months later, Gordon Teal, a former Bell Labs metallurgist who had pioneered techniques for making germanium crystals before joining a tiny semiconductor company called Texas Instruments, unveiled his own silicon transistor. But neither of these developments were cause for celebration. One of the drawbacks of Buehler and Tanenbaum’s silicon transistor was their complicated fabrication method, which seemed unsuitable for mass production. To actually change the course of an entire industry, a silicon transistor would have to be reliable and easy to make.
For nearly a year afterward, Tanenbaum kept working with silicon. He was in Building 2 at Murray Hill, with Shockley’s research team; just down the hall was Jack Morton’s transistor development group. In Building 1, across a courtyard, Tanenbaum had a chemistry colleague named Cal Fuller. Like many of his fellow researchers at the Labs, Fuller had a background that should never have led him into a distinguished life in science. A slim and
scholarly-looking man, the son of a bookkeeper who was raised in a poor family in Chicago, Fuller as a teenager had experimented diligently with his wireless radio and home chemistry set. He never imagined he would go to college. But then his high school physics teacher thought otherwise. She was Mabel Walbridge, “a very demanding teacher,” Fuller recalled, “a woman in her fifties or early sixties” who had taken courses earlier in her life from Robert Millikan. She knew that the University of Chicago offered exams to high school students in science and math. She also knew that for those who passed, the university “provided full tuition for the first year and, if you were among the top twenty-five students in your class at the university, for the following three years.” Walbridge demanded that Fuller pursue the scholarship. “She tutored me nights on her own time,” Fuller said, still astounded by the fact sixty years later, “so that when I took the exam I was prepared for almost every question.” Fuller went on to an undergraduate degree at Chicago and a PhD in chemistry. He put himself through graduate school by working the four-o’clock-to-midnight shift at the Chicago Tribune.11
His early work at Bell Labs had been on plastics and rubber. After Mervin Kelly reorganized the Labs research department after the war, however, Fuller begun working on semiconductors such as germanium and silicon, with a special interest in how infinitesimal impurities affected them. He had noticed that a germanium crystal could be rendered impure if someone touched it after handling a brass doorknob,12 and he wondered if there was a way to take advantage of the remarkable sensitivity of these crystals to impurities. By the time Morry Tanenbaum got to know him, Fuller was using a technique called diffusion that suggested a way to manipulate the concentrations of impurities in silicon with remarkable precision.
In diffusion, a long silicon crystal (it looked about the size of a pretzel rod, says Tanenbaum) is cut into thin, round slices; the slices are then placed in a furnace. In the furnace, the silicon slices are exposed to a gas containing an impurity, such as aluminum. “At high temperatures, these impurity atoms bombard the crystal surface and slowly force their way into the interior,” the Bell Laboratories Record explained. Or to put it another way, the impurities in the furnace atmosphere, depending on the type, could create, on top of, say, an n-type silicon slice, exceedingly thin layers of p-type and n-type silicon stacked on top of one another. One might envision a crystal disc the size of a dime that emerges from a furnace with two thin coatings on top, each less than a thousandth of an inch thick. These are, respectively, the p-type and n-type layers.
Now that he had the right materials, Tanenbaum faced the challenge of actually making an electrical contact with the thin middle p-layer within the diffused silicon disc. That middle layer was far finer than a human hair. He spent weeks on the problem, trying to grind the dime-sized crystal on an angle to attach a wire, and attempting all sorts of other tricks. “Finally, one night, I went back into the lab because my wife was having a bridge game,” he recalls, and by trying a blunt method—in his lab notebook he wrote, will try direct approach—he melted an aluminum wire “through” the thin top layer. He made a good contact. It was late on the evening of March 17, 1955. When he took some instrument readings, he was shocked to see that the device performed better than any germanium transistor then in existence. In his notebook he wrote, This looks like the transistor we’ve been waiting for. It should be a cinch to make. “Right away,” he recalls, “I knew that this would be very manufacturable.” He drove home like a demon to tell his wife. He could barely sleep, wondering if he had imagined the whole thing, and rushed back to the Labs in the morning to test it. Almost immediately the supervisors were called in, including Jim Fisk, who had returned from the Atomic Energy Commission and was now Bell Labs’ chief of research.
Jack Morton was in Europe, Tanenbaum recalls, but cut his trip short and flew home when he was told the news. Right away Morton—whose opinions on transistor innovation were akin to a final judgment at Bell Labs—understood the potential value. Even Kelly, too busy with management to properly assess the technical details that entered into Morton’s calculus, would defer to him on such matters. If Morton was on board with the diffused silicon transistor, Kelly was, too. That meant the future would be silicon.
DIFFUSED SILICON had another use, too.
Fifteen years had passed since the day Walter Brattain had been ushered into Mervin Kelly’s office to regard a strange piece of silicon that had been discovered down in Holmdel, New Jersey. The men had shone a light on the blackened chunk and the resulting electric charge had stunned them. In later years it came to be understood that this chunk of silicon contained a naturally occurring p-n junction where two types of silicon met. The junction is extremely photosensitive. In very general terms, the photons in light are hitting the semiconductor crystal and “splitting off” electrons from their normal location in the crystal; the process, if properly captured, can create a flow of electrons, that is, a flow of electricity.13 Kelly and Brattain and Ohl didn’t know it at the time, but in Kelly’s office the men had been looking at the world’s first crude silicon solar cell.
In the early 1950s, Cal Fuller, who had made the diffused silicon that Morris Tanenbaum used to fashion his silicon transistor, was also working at Murray Hill with the experimental physicist Gerald Pearson. A genial presence at the Labs, slim and handsome and neatly kept, with his dark hair always combed straight back from his forehead, Pearson was Walter Brattain’s old laboratory mate, the same G. L. Pearson who had signed Brattain’s lab notebook on that fateful Christmas Eve in 1947 after Brattain and Bardeen had demonstrated the transistor for the Bell Labs brass. Now Fuller and Pearson were trying to build something with diffused silicon called a silicon power rectifier. In the course of the work on the device, Pearson noticed it was highly sensitive to light.14 Pearson had an old college friend at the Labs named Daryl Chapin, who he knew was trying to develop power sources for remote telephone installations. Often these remote installations—places where phone repeaters might be located, for instance—used diesel generators or dry cell batteries. The batteries had problems in humid weather. Pearson wondered if Chapin might be able to use the power of the sun.
With Pearson as the go-between, the three men—Fuller, Pearson, and Chapin—created over the course of a few months what the Labs eventually called a silicon solar battery. These were thin strips of specially diffused silicon, connected to a circuit, that in sunlight could generate a steady electric voltage. The “battery” was not the first solar cell; functional ones had been made before from the element selenium, for instance. But this was, by Bell Labs’ calculations, “at least fifteen times more efficient than the best previous solar energy converter.”15 That made it the first truly usable solar power device. The batteries promised to last forever, since they had no moving parts. What was striking but almost always overlooked about its invention, Fuller later recalled, was that all three inventors of the device were working in different buildings. “The solar cell just sort of happened,” he said. It was not “team research” in the traditional sense, but it was made possible “because the Labs policy did not require us to get the permission of our bosses to cooperate—at the Laboratories one could go directly to the person who could help.”16
The silicon solar cell generated a hurricane of publicity when it was unveiled. “The subsequent attention,” Fuller recalled, “which exceeded that of the announcement of the transistor, was unbelievable.”17 By the front-page newspaper headlines,18 one might easily imagine that the cells’ ability to effectively harness the sun meant modern society had reached a pivotal juncture, and that soon enough the world’s energy supplies would be clean and inexhaustible. For those who knew anything about transistors, which had extremely low power requirements and could thus be a perfect match for the new solar technology, the news seemed yet another fantastical bit of Bell Labs augury. On October 4, 1955, a test project for the solar cells was set up by the Labs’ engineers at a remote rural phone installation in Amer
icus, Georgia, 135 miles south of Atlanta. For six months, they powered equipment at the installation, and hinted at the remarkable future where power could be generated anywhere the sun shone.
And then the great excitement of the solar breakthrough dimmed. As Pearson would later recall, the installation was “a huge technical success, but a financial failure.” The solar battery could power the remote telephone equipment with ease. But for the power they generated, the solar cells, at several hundred dollars per watt, simply cost too much.19 In 1956, Daryl Chapin figured that it would cost the average homeowner nearly $1.5 million to buy enough Bell solar cells to power his house.20 By one of Kelly’s fundamental dictums of innovation—something that could do a job “better, or cheaper, or both”—the cost of the cells and the results in Georgia suggested solar power was not going to be a marketable innovation anytime soon. Sometimes, in describing a new invention that seemed technically brilliant but impractical, industrial scientists would quip that they had found “a solution looking for a problem.” The silicon solar cell needed a problem, as yet unimagined, to appear.
THE PUBLICITY AROUND inventions like the solar cell tended to distort public perceptions about the actual work being done inside the Labs. Kelly would often point out that the Labs workforce—including PhDs, lab technicians, and clerical staff—by the early 1950s totaled around nine thousand.21 Only 20 percent of those nine thousand worked in basic and applied research, however. Another 20 percent worked on military matters. Meanwhile, the rest of the Labs’ scientists and engineers—the majority—toiled on the never-ending job of planning and developing the system. Their work was arguably less glamorous. The research scientists at the Labs were thinking ahead to a glorious future that was ten or even twenty years away. The development and systems engineers were thinking about what they could do in the next year or two or three. And yet the projects undertaken by the latter group during Kelly’s presidency were in many ways just as ambitious as those done in research; one might see that they were logistically more difficult. In development, mistakes were not excusable. Building a new product or invention, and then putting it into the working telephone system, demanded perfection.