* * *
I. For example, the engineers and theorists discovered that silicon (which has four electrons in its outer orbit) that was doped with phosphorus or arsenic (which have five electrons in their outer orbits) had spare electrons and thus was a negative-charged carrier. The result was called an n-type semiconductor. Silicon that was doped with boron (with three electrons in its outer orbit) had a deficit of electrons—there were “holes” where some electrons would normally be—and thus was positively charged, making it known as a p-type semiconductor.
II. His son Fred Terman later became the famous dean and provost at Stanford.
III. For a short video of Shannon and his machines juggling, see https://www2.bc.edu/~lewbel/shortsha.mov.
Jack Kilby (1923–2005) at Texas Instruments in 1965.
Kilby’s microchip.
Arthur Rock (1926– ) in 1997.
Andy Grove (1936– ) with Noyce and Moore at Intel in 1978.
CHAPTER FIVE
* * *
THE MICROCHIP
In a paper written to celebrate the tenth anniversary of the transistor, published in 1957 just when Fairchild Semiconductor was formed and Sputnik launched, a Bell Labs executive identified a problem that he dubbed “the tyranny of numbers.” As the number of components in a circuit increased, the number of connections increased way faster. If a system had, for example, ten thousand components, that might require 100,000 or more little wire links on the circuit boards, most often soldered by hand. This was not a recipe for reliability.
It was, instead, part of a recipe for an innovation. The need to solve this growing problem coincided with hundreds of small advances in ways to manufacture semiconductors. This combination produced an invention that occurred independently in two different places, Texas Instruments and Fairchild Semiconductor. The result was an integrated circuit, also known as a microchip.
JACK KILBY
Jack Kilby was another of those boys from the rural Midwest who tinkered in the workshop with his dad and built ham radios.1 “I grew up among the industrious descendants of the western settlers of the American Great Plains,” he declared when he won a Nobel Prize.2 He was raised in Great Bend, in the middle of Kansas, where his father ran a local utility company. In the summer they would drive in the family Buick to far-flung generating plants and, when something had gone wrong, crawl through them together looking for the problem. During one bad blizzard they used a ham radio to keep in touch with areas where customers had lost phone service, and young Kilby became fascinated by the importance of such technologies. “It was during an ice storm in my teens,” he told the Washington Post’s T. R. Reid, “that I first saw how radio and, by extension, electronics, could really impact people’s lives by keeping them informed and connected, and giving them hope.”3 He studied to get a ham operator’s license and kept upgrading his radio using parts that he scrounged.
After being turned down by MIT, he went to the University of Illinois, interrupting his studies after Pearl Harbor to join the Navy. Deployed to a radio repair facility in India, he made runs to Calcutta to buy parts on the black market, using them to build better receivers and transmitters in a pup-tent lab. He was a gentle guy with a wide smile and an easygoing, taciturn manner. What made him special was his insatiable curiosity about inventions. He began to read every new patent issued. “You read everything—that’s part of the job,” he said. “You accumulate all this trivia, and you hope that someday maybe a millionth of it will be useful.”4
His first job was at Centralab, a Milwaukee firm that made electronic parts. It experimented with ways of combining the components used to make hearing aids onto a single ceramic base, a rough precursor of the idea for a microchip. In 1952 Centralab was one of the companies that paid $25,000 for a license to make transistors, and it was the beneficiary of Bell’s willingness to share its knowledge. Kilby attended a two-week Bell Labs seminar—staying with dozens of others at a Manhattan hotel and being loaded every morning onto a bus for Murray Hill—that included in-depth sessions on transistor design, hands-on experience in the labs, and visits to a manufacturing plant. Bell sent all attendees three volumes of technical papers. With its extraordinary willingness to license its patents cheaply and share its knowledge, Bell Labs laid the foundations for the Digital Revolution, even though it didn’t fully capitalize on it.
In order to be at the forefront of transistor development, Kilby realized that he needed to work at a bigger company. Weighing a variety of offers, he decided in the summer of 1958 to join Texas Instruments, where he would get to work with Pat Haggerty and his brilliant transistor research team led by Willis Adcock.
The policy at Texas Instruments was for everyone to take off the same two weeks in July. So when Kilby arrived in Dallas with no accrued vacation time, he was one of the very few people in the semiconductor lab. This gave him time to think about what could be done with silicon other than fabricate it into transistors.
He knew that if you created a bit of silicon without any impurities, it would act as a simple resistor. There was also a way, he realized, to make a p-n junction in a piece of silicon act as a capacitor, meaning it could store a small electrical charge. In fact, you could make any electronic component out of differently treated silicon. From that he came up with what became known as the “monolithic idea”: you could make all of these components in one monolithic piece of silicon, thus eliminating the need to solder together different components on a circuit board. In July 1958, six months before Noyce wrote down a similar idea, Kilby described it in his lab notebook in a sentence that would later be quoted in his Nobel Prize citation: “The following circuit elements could be made on a single slice: resistors, capacitor, distributed capacitor, transistor.” Then he drew a few crude sketches of how to construct these components by configuring sections of silicon that had been doped with impurities to have different properties on a single slab.
When his boss, Willis Adcock, returned from vacation he was not fully persuaded that this would be practical. There were other things for the lab to do that seemed more pressing. But he made Kilby a deal: if he could make a working capacitor and resistor, Adcock would authorize an effort to do a complete circuit on a single chip.
All went as planned, and in September 1958 Kilby prepared a demonstration that was similar in drama to the one Bardeen and Brattain had done for their superiors at Bell Labs eleven years earlier. On a silicon chip the size of a short toothpick, Kilby assembled the components that would, in theory, make an oscillator. Under the gaze of a group of executives, including the chairman of the company, a nervous Kilby hooked up the tiny chip to an oscilloscope. He looked at Adcock, who shrugged as if to say, Here goes nothing. When he pushed a button, the line on the oscilloscope screen undulated in waves, just as it should. “Everybody broke into broad smiles,” Reid reported. “A new era in electronics had begun.”5
It was not the most elegant device. In the models that Kilby built that fall of 1958, there were a lot of tiny gold wires connecting some of the components within the chip. It looked like expensive cobwebs sticking out of a silicon twig. Not only was it ugly; it was also impractical. There would be no way to manufacture it in large quantities. Nevertheless, it was the first microchip.
In March 1959, a few weeks after filing for a patent, Texas Instruments announced its new invention, which it dubbed a “solid circuit.” It also put a few prototypes on display, with much fanfare, at the Institute of Radio Engineers annual conference in New York City. The company’s president declared that the invention would be the most important one since the transistor. It seemed like hyperbole, but it was an understatement.
The Texas Instruments announcement struck like a thunderbolt at Fairchild. Noyce, who had jotted down his own version of the concept two months earlier, was disappointed at being scooped and fearful of the competitive advantage it might give Texas Instruments.
NOYCE’S VERSION
There are often different paths to the sa
me innovation. Noyce and his Fairchild colleagues had been pursuing the possibility of a microchip from another direction. It began when they found themselves hit with a messy problem: their transistors were not working very well. Too many of them failed. A tiny piece of dust or even exposure to some gases could cause them to fizzle. So, too, might a sharp tap or bump.
Jean Hoerni, a Fairchild physicist who was one of the traitorous eight, came up with an ingenious fix. On the surface of a silicon transistor, he would place a thin layer of silicon oxide, like icing atop a layer cake, that would protect the silicon below. “The building up of an oxide layer . . . on the surface of the transistor,” he wrote in his notebook, “will protect the otherwise exposed junctions from contamination.”6
The method was dubbed “the planar process” because of the flat plane of oxide that sat on top of the silicon. In January 1959 (after Kilby had come up with his ideas but before they were patented or announced), Hoerni had another “epiphany” while showering one morning: tiny windows could be engraved in this protective oxide layer to allow impurities to be diffused at precise spots in order to create the desired semiconductor properties. Noyce loved this idea of “building a transistor inside a cocoon,” and he compared it to “setting up your jungle operating room—you put the patient inside a plastic bag and you operate inside of that, and you don’t have all the flies of the jungle sitting on the wound.”7
The role of patent lawyers is to protect good ideas, but sometimes they also stimulate them. The planar process became an example of this. Noyce called in John Ralls, Fairchild’s patent lawyer, to prepare an application. So Ralls began grilling Hoerni, Noyce, and their coworkers: What practical things could be done with this planar process? Ralls was probing to obtain the widest range of possible uses to put in the patent application. Recalled Noyce, “The challenge from Ralls was, ‘What else can we do with these ideas in terms of patent protection?’ ”8
At the time, Hoerni’s idea was merely designed to build a reliable transistor. It had not yet occurred to them that the planar process with its tiny windows could be used to permit many types of transistors and other components to be etched onto a single piece of silicon. But Ralls’s persistent questioning got Noyce thinking, and he spent time that January batting around ideas with Moore, scribbling them on a blackboard and jotting them into his notebook.
Noyce’s first realization was that the planar process could eliminate the tiny wires that stuck out of each layer of the transistor. In their place, little copper lines could be printed on top of the oxide layer. That would make manufacturing the transistors faster and more reliable. This led to Noyce’s next insight: if you used these printed copper lines to connect the regions of a transistor, you could also use them to connect two or more transistors that were on the same piece of silicon. The planar process with its window technique would allow you to diffuse impurities so that multiple transistors could be placed on the same silicon chip, and the printed copper wires could connect them into a circuit. He walked into Moore’s office and drew the idea on the blackboard for him.
Noyce was a talkative bundle of energy and Moore was a taciturn yet insightful sounding board, and they played off each other well. The next leap was easy: the same chip could also contain various components, such as resistors and capacitors. Noyce scribbled on Moore’s blackboard to show how a small section of pure silicon could serve as a resistor, and a few days later he sketched out how to make a silicon capacitor. The little metal lines printed on the oxide surface could integrate all of these components into a circuit. “I don’t remember any time when a light bulb went off and the whole thing was there,” conceded Noyce. “It was more like, every day, you would say, ‘Well, if I could do this, then maybe I could do that, and that would let me do this,’ and eventually you had the concept.”9 After this flurry of activity he wrote an entry in his notebook, in January 1959: “It would be desirable to make multiple devices on a single piece of silicon.”10
Noyce had come up with the concept of a microchip independently of (and a few months later than) Kilby, and they had gotten there in different ways. Kilby was trying to solve the problem of how to overcome the tyranny of numbers by creating circuits with many components that did not have to be soldered together. Noyce was mainly motivated by trying to figure out all the neat tricks that could come from Hoerni’s planar process. There was one other, more practical difference: Noyce’s version didn’t have a messy spider’s nest of wires protruding from it.
PROTECTING DISCOVERIES
Patents present an inevitable source of tension in the history of invention, especially so in the digital age. Innovations tend to proceed through collaboration and building on the work of others, so it is difficult to ascribe with precision the ownership of ideas or intellectual property rights. Occasionally this is made gloriously irrelevant when a group of innovators agrees to engage in an open-source process that allows the fruits of their creativity to be in the public domain. More often, however, an innovator wants credit. Sometimes this is for ego reasons, as was the case when Shockley maneuvered to be listed on the patents for the transistor. At other times it is for financial reasons, especially when it involves companies such as Fairchild and Texas Instruments that need to reward investors in order to have the working capital necessary to keep inventing things.
In January 1959 the lawyers and executives at Texas Instruments began scrambling to file a patent application for Kilby’s idea of an integrated circuit—not because they knew what Noyce was jotting in his notebook but because of rumors that RCA had come up with the same idea. They decided to make the application sweeping and broad. That strategy carried a risk because the claims might be easier to dispute, as happened with Mauchly and Eckert’s broad claims for their computer patent. But if granted it would serve as an offensive weapon against anyone who tried to make a product that was similar. Kilby’s invention, the patent application declared, was “a new and totally different concept for miniaturization.” Although the application described only two circuits that Kilby had devised, it asserted, “There is no limit upon the complexity or configuration of circuits that can be made in this manner.”
In the rush, however, there wasn’t time to produce pictures of the various methods that might work for wiring together the components on the proposed microchip. The only example available was Kilby’s spidery demonstration model with a snarl of tiny gold wires threading through it. The Texas Instruments team decided to use this “flying wire picture,” as it was later derisively called, as the depiction. Kilby had already figured out that there could be a simpler version using printed metal connections, so at the last moment he told his lawyers to add a passage to the application claiming rights to that concept as well. “Instead of using the gold wires in making electrical connections,” it noted, “connections may be provided in other ways. For example . . . silicon oxide may be evaporated onto the semiconductor circuit wafer. . . . Material such as gold may then be laid down on the insulating material to make the necessary electrical connections.” It was filed in February 1959.11
When Texas Instruments made its public announcement the next month, Noyce and his team at Fairchild hastened to file a competing patent application. Because they were seeking a shield against Texas Instruments’ sweeping claim, the Fairchild lawyers focused very specifically on what was special about Noyce’s version. They emphasized that the planar process, which Fairchild had already filed to patent, permitted a printed-circuit method “for making electrical connections to the various semiconductor regions” and “to make unitary circuit structures more compact and more easily fabricated.” Unlike circuits in which “electrical connection had to be made by fastening wires,” declared the Fairchild application, Noyce’s method meant that “the leads can be deposited at the same time and in the same manner as the contacts themselves.” Even if Texas Instruments should be awarded a patent for putting multiple components on a single chip, Fairchild hoped to be awarded a patent for making
the connections through printed metal lines instead of wires. Because this would be necessary for mass-producing microchips, Fairchild knew it would give them some parity in patent protection and force Texas Instruments to enter into a cross-licensing deal. The Fairchild application was filed in July 1959.12
As happened with the patent dispute over the computer, the legal system took years grappling with the issue of who deserved what patents on the integrated circuit, and it never quite resolved the question. The rival applications from Texas Instruments and Fairchild were assigned to two different examiners, who each seemed unaware of the other. Although filed second, the Noyce patent application was ruled on first; in April 1961 it was granted. Noyce was declared the inventor of the microchip.
The Texas Instruments lawyers filed a “priority contest,” claiming that Kilby had the idea first. That led to the case of Kilby v. Noyce, run by the Board of Patent Interferences. Part of the case involved looking at the respective notebooks and other testimony to see who had come up with the general concept first; there was broad agreement, even from Noyce, that Kilby’s ideas had come a few months earlier. But there was also a dispute over whether the Kilby application really covered the key technological process of printing metal lines on top of an oxide layer, rather than using a lot of little wires, to make a microchip. This involved many conflicting arguments about the phrase Kilby had inserted at the end of the application, that “material such as gold may then be laid down” on the oxide layer. Was that a specific process he had discovered or merely a casual speculation he had tossed in?13