by T. R. Reid
The integrated circuit, however, was so new, and so potentially lucrative, that Mosher, Mims, and Kilby decided to shoot the works—to write a far-reaching patent application that Texas Instruments could wield as an offensive weapon against anyone else who tried to make, use, or sell the device. Accordingly, Mosher and his partners set out to write a document that would leave no doubt about the revolutionary quality of Kilby’s idea. They laid it on thick. Every aspect of the integrated circuit was described as “novel” or “unique,” or both. “Radically departing from the teachings of the art,” the application said, “. . . the present invention has resulted from a new and totally different concept for miniaturization.” As a result “the ultimate in circuit miniaturization is attained.” And for good measure: “...the invention... represents a remarkable improvement over the prior art.”
The document went on to describe two specific circuits that could be fabricated on a semiconductor chip. But Mosher emphasized that Kilby’s invention could be used to build an integrated version of any circuit: “There is no limit upon the complexity or configuration of circuits that can be made in this manner.” To put a point on his sword, Mosher added a thinly veiled warning to anybody who might try to circumvent the patent by making small changes: “Such changes and modifications,” the application said, “are deemed to fall within the purview of this invention.”
For all this rhetorical bravado, the authors of Kilby’s patent application still had to contend with a fairly sticky problem. Since the patent system is designed to let the world know the secrets behind technological advances (and thus, in theory, stimulate new advances), the law requires an inventor to explain in his patent application how the device is made. “The specification,” the statute says, “. . . shall set forth the best mode contemplated by the inventor for carrying out his invention.” Further, the law requires an inventor to provide drawings showing precisely what his new device looks like. This was a problem because nobody at Texas Instruments, not even Jack Kilby, had figured out the best mode for carrying out the invention. And nobody knew what a production-model integrated circuit would look like.
In a way, Kilby’s invention was almost too ingenious, too violent a break with the way things had always been done before. By the late 1950s, engineers had been building electric circuits for almost 100 years, and they had always done it the same way: taking individual resistors, capacitors, and other components and wiring them together. The invention of the transistor had led to dramatic changes in the size, cost, reliability, and efficiency of circuits, but it had not changed the basic structure of discrete components wired together. That was one reason the tyranny of numbers had proven such a baffling problem; to the engineers, a large number of discrete parts were what circuits were all about.
The monolithic idea—the idea that an entire circuit could be a single part, with resistors, capacitors, diodes, and transistors built right in—involved two fundamentally new concepts, which Kilby later defined as integration and interconnection. Integration was the central idea: the realization that all the parts of a circuit could be made from the same material—silicon—and thus a whole circuit could be integrated in a single silicon chip. Interconnection was the recognition that, once all the parts of a circuit were made on a single chip, the connections between the different parts— the wires—could be printed onto the chip as part of the production process. That way, the enormous cost and the inherent unreliability of hand-wiring huge numbers of components to one another would be eliminated.
In the first chips he built, during late summer and fall of 1958, Kilby successfully integrated all the circuit components into a silicon chip. But he didn’t have time to work out the interconnections within the chip. Racing to demonstrate the new idea to his bosses before they lost interest, he had settled for a jury-rigged device in which the separate components on the chip were connected (by hand) with tiny gold wires, giving that first integrated circuit the appearance of an intricate cobweb spun by a golden spider. This crude handmade circuit was a mess—just a sloppy piece of lab equipment, like the dirty glass tube with wires sticking out that was Edison’s first light bulb. The world’s first integrated circuit had been adequate for its purpose—proving that a circuit on a chip would really work—but the job of attaching the wires was much too painstaking and time-consuming for any large-scale production. Before the integrated circuit could become a practical solution to the numbers barrier, somebody would have to deal with the question of interconnection.
By the beginning of 1959, Kilby had somewhat improved his techniques for constructing individual components within a semiconductor chip. With several TI colleagues, he had worked out some reliable designs for different circuits on a chip. But the interconnections business was still up in the air. Kilby had a vague idea, an idea that was similar to the planar process that Noyce’s colleague Jean Hoerni had developed at Fairchild. He thought he could put a layer of silicon oxide on top of the chip, like icing on a cake, and then print fine lines of a conducting metal, such as gold, atop the icing to connect various parts of the circuit. (This was, in fact, the method eventually developed for chip production.) But at the end of January, the only chips in existence still used the handmade cobwebs of wire.
Here was a dilemma: Worried that RCA was going to apply for a patent any minute, the TI people were anxious to file first. But a patent application required a picture. To draw a picture, Mosher’s artist needed a model. But the only model available was Kilby’s crude demonstration chip, with its network of gold wire. After a week or so of deliberation, it was decided to put speed first; the application was filed with a picture of the hand-wired chip:
Even now, more than forty years later, engineers tend to smirk when they see this drawing—it is known in the electronics business as the “flying wire picture”—because it is so far removed from what an integrated circuit is supposed to be.
Jack Kilby, of course, was one of the engineers who knew that the flying wire drawing was fundamentally wrong. Accordingly, the lawyers threw in some language designed to fudge the issue. “Although the invention has been shown and described in terms of specific embodiments,” they wrote, “it will be evident that changes and modifications are possible which do not in fact depart from the inventive concepts taught herein.”
That helped a little, but Kilby still wasn’t satisfied; he had to put in something to show that his revolutionary new circuit would not have to be wired by hand. At the last minute, accordingly, Kilby and Mims added one more paragraph to the application, setting forth the icing-on-the-cake approach to interconnection. “Instead of using the gold wires in making electrical connections,” it said, “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 [oxide] to make the necessary electrical connections.” With this final addition, the application—four pages of pictures and five of text—was delivered to the Patent Office on February 6, 1959.
There, Kilby’s application for a patent on his “Miniaturized Electronic Circuits” was assigned to an examiner who specialized in inventions involving electronic circuits and devices. The Patent Office employs about 2,500 examiners (they are generally lawyers with a background in some technical field) whose job is simple to describe—they have to determine whether an invention is legally entitled to a U.S. patent—but often extremely complicated to perform. The job of patent examiner tends to attract people with a curious, probing intelligence; among the trade’s most illustrious alumni was a hard-working physicist who held the position of “Technical Expert, Third Class” at the Swiss Patent Office in Bern at the turn of the twentieth century—a fellow named Albert Einstein.
The examiner has to determine whether an idea is “new”—a question that requires searching dozens, hundreds, or thousands of earlier patents, reports, and monographs. He has to decide whether it is “useful,” a determination that demands extensive resea
rch in the technical literature. And he has to determine whether the inventor’s application clearly sets forth how the new gadget is to be built. On the average, each examiner is assigned between 75 and 100 patent applications every year. As a result, it regularly takes months, and sometimes years, for an examiner to deliver his initial opinion on an application. If the examiner finds anything amiss—and he almost always does—the Patent Office sends a letter to the inventor explaining the reservations. The inventor gets up to six months to prepare a reply or amend the application; then the examiner may take months more to review the reply. It is not uncommon, particularly for a complex invention in a specialized field, for an inventor to wait years for final action on a patent application.
So it was for Kilby. The process dragged on, with the examiner raising questions and Texas Instruments doing its best to answer them. The first important development came some twenty-six months after the application had been filed. On April 26, 1961, Kilby received a telephone call from the lawyers in Washington, informing him that the first patent ever for an integrated circuit had been granted.
But it had not been granted to Jack Kilby.
Like their counterparts at Texas Instruments, the people at Fairchild Semiconductor took their sweet time at first about developing the monolithic idea into a practical integrated circuit. It was nearly two months after Bob Noyce had set down the basic concept in his notebook before Fairchild started working on the idea; another four months passed before Noyce got around to filing for a patent.
The reason, Noyce explained later, was that, after twenty months of preliminary planning and organizing, Fairchild at the start of 1959 was just beginning to sell its first important product— the double-diffuse transistor. “We were still a brand-new company,” Noyce recalled. “We were worried about basic survival. That meant getting transistors out the door. The integrated circuit seemed interesting, it was something that might make you some money somewhere down the road, but that was not a period when you had a lot of time for it.” During the winter of 1959, the monolithic idea was literally put on a shelf; except for his friend and sounding board, Gordon Moore, Noyce didn’t even mention it to any of his colleagues.
Like their counterparts at Texas Instruments, the people at Fairchild were eventually prodded into action by a rumor— although in this case, the rumor was true. Sometime in late February or early March, word arrived in Silicon Valley that Texas Instruments was about to announce a wholly new kind of circuit that would do away with discrete electric components by integrating all the parts of a circuit into a single silicon chip. For any company that made its money selling discrete components, like Fairchild’s transistors, this was disconcerting news. Their basic product was suddenly faced with obsolescence. Somebody at Fairchild called a meeting to discuss the development. At this session Noyce explained for the first time that he, too, had worked out the basic concept of an integrated circuit. The group decided immediately to get hopping on the new product. To any engineer looking at plans for a possible new product, it was obvious that one of the first steps would be to file for a patent.
Noyce and his patent lawyer, John Ralls, did not have access to Kilby’s patent application; the Patent Office treats pending applications as classified material, and there are no leaks. But by mid-March, when TI publicly announced its new “Solid Circuit,” they knew that the Dallas firm must have filed already. What Fairchild needed, then, was a legal shield—a patent that would differentiate Noyce’s version of the idea from the Kilby invention and thus permit Fairchild to enter the integrated circuit market without fear of legal action by TI.
In preparing his application, Noyce had one significant advantage: Kilby had been forced to file for a patent before he had worked out the problem of interconnections within the chip; Noyce’s formulation of the idea covered both integration and interconnection. This was a result of the different routes the two inventors had taken to arrive at the integrated circuit. Kilby had first hit upon the concept of integration—of building all the parts of a circuit in a monolithic chip of silicon—and had moved from there to consideration of interconnections. Noyce, in contrast, had first recognized the possibility of printing connecting strips of metal on a chip—something made possible by Jean Hoerni’s invention of the planar process—and the notion of interconnection had led him to the idea of integration. By the spring of 1959, Fairchild was busily engaged in working out the details of Hoerni’s planar process, and thus Noyce had no difficulty providing a description and a drawing of a chip with interconnections built right in. Accordingly, Noyce and Ralls titled their application “Semiconductor Device-and-Lead Structure” (“lead” is the electrician’s term for a connecting wire in a circuit), and they strongly emphasized the interconnections aspect of the Noyce circuit. The application listed three “principal objects” of the invention. The first one was interconnections: “to provide improved device-and-lead structures for making electrical connections to the various semiconductor regions.”
Noyce’s application went on to describe a “unitary circuit structure” that would permit integrating “more than one circuit device, into a single body of semiconductor.”
“According to prior practice,” it continued, “electrical connection . . . had to be made by fastening wires directly to the [components]. . . . By means of the present invention, the leads can be deposited at the same time and in the same manner as the [components] themselves.”
To the four-page written description of Noyce’s idea, Ralls added three pages of pictures of typical circuits that could be integrated onto a chip. There were no flying wires. There were no wires at all. The drawings show a structure that is essentially the same as the integrated circuits being produced today.
It was midsummer before Ralls and Noyce were satisfied that everything was in order; the patent application was finally filed on July 30, 1959. It was assigned, evidently, to an examiner who was not aware of the earlier application from Texas Instruments, and it moved ahead at what is, for the Patent Office, lightning speed. Twenty-one months after it was filed, Noyce’s application was granted: U.S. Patent No. 2,981,877. By formal decree of the United States of America, Robert N. Noyce—the second person to come up with the idea—had been officially declared the inventor of the integrated circuit.
The award of an integrated circuit patent to Noyce evoked consternation, but not outright panic, at Texas Instruments. Kilby and his lawyers, after all, were veterans of the patent game; they knew that some applications move through the Patent Office faster than others, and that it is not particularly unusual for the second version of an invention to be the first patented. This happens so often, in fact, that the government has a special procedure— called an interference proceeding—and a special board—the Board of Patent Interferences—to consider the claims of inventors who find themselves in Kilby’s position. The basic rule governing an interference is that priority prevails—that is, whichever inventor can prove to have had the idea first gets the patent.
Mosher filed the necessary papers, and in May 1962 both Noyce and Kilby received a copy of Commerce Department Form POL-102, declaring that the Board of Patent Interferences had convened Interference No. 92,842, “Kilby v. Noyce,” to determine who had really been the first to invent the microchip. The board enclosed a short form asking each man to list the earliest date on which he could prove he had had the idea. Since both Kilby and Noyce had maintained lab notebooks precisely for this purpose, both were able to provide an exact answer: July 1958 for Kilby, January 1959 for Noyce. With these preliminaries concluded—they consumed about ten months—the stage was set for a final determination.
Actually, it was not quite set. Before the central legal battle could get under way, the lawyers fought out a series of preliminary skirmishes:
“Motion to Dissolve Under Rule 232 (a) (2)”
“Opposition to Motion to Dissolve Under Rule 232 (a) (2)”
“Motion to Dissolve Under Rule 232 (a) (3)”
“Opposition to Motion to Dissolve Under Rule 232 (a) (3)”
“Request to File Affidavits”
“Opposition to Motion to File Affidavits”
While this was going on, the Patent Office concluded that Jack Kilby’s initial application was satisfactory after all. In June 1964, Kilby was granted a patent—No. 3,138,743—for the integrated circuit. This made the interference proceeding even more crucial, and the lawyers went back to work:
“Motion for Extension of Time”
“Opposition to Motion for Extension of Time”
Each of these preliminary disputes took a few months to resolve. And so it was not until July 28, 1964—more than two years after the interference was begun—that the inventors and their lawyers gathered in Mosher’s Washington office to hear the first piece of evidence in the case.
The session was a brief one, directed strictly to prove that Kilby had made the invention first. Kilby explained that he had gotten the idea in July 1958, and one of Mosher’s associates related how he had filed Kilby’s patent application in February 1959. There was a short cross-examination, and everybody went home.