<<

to focus the company’ talents and From to Valley: energies on devices that were then at the frontiers of technology, A Saga of Semiconductor Technology Transfer, 1955-61* such as field-effect and the four-layer p-n-p-n , often called the because he held by Michael Riordan a patent on it. At a given potential called the breakdown voltage, current would suddenly begin to flow through espite the fact that the Bell the two-terminal device, switching had it from “off” to “on.” A remarkably Doriginated almost all the silicon simple, compact semiconductor switch, technology eventually used to invent it was the ultimate realization of a the , or microchip, this goal that had been subtly implanted revolutionary breakthrough occurred in Shockley’s mind at Bell Labs two elsewhere—at Instruments, Inc., decades earlier.10 in , and Although a brilliant conception, Corporation in Mountain View, however, the four-layer diode was . In the latter case, the extremely difficult to fabricate with transfer of this technology occurred uniformity and reliability. Workers largely through the offices of a pivotal had to polish a silicon to have but ultimately unsuccessful company, exactingly parallel planes on its Shockley Semiconductor , Fig. 2. Senior staff of Shockley Semiconductor Laboratory toast their illustrious leader two sides and then carefully diffuse which had been formed by (seated at head of table) at a celebratory luncheon on 2 November impurities into the wafer from pioneer William Shockley. After 1956, the day after the announcement of his . Seated at left is both sides simultaneously. Lack of leaving Bell Labs in 1955, he brought and right of him is . stands in the middle and is enough precision meant the diffused together a stellar team of scientists at far right, both lifting glasses. (Courtesy of Corporation) impurities would penetrate to irregular, and engineers later departed (in unpredictable depths, leading to Fig. 1. Shockley Semiconductor Laboratory at 391 San Antonio Road in Mountain View, September 1957) to found Fairchild, would be published in 1958 as the final on this important method. The preprint variations in the breakdown voltage California, as it appeared in 1960 after Clevite Corporation had purchased it and and other electrical characteristics.11 bringing a deep understanding of the volume in the classic series, Transistor also contained crucial new information renamed it Shockley Transistor Company. (Courtesy of Kurt Hübner) Noyce, Moore, and others thought diffused-silicon technologies they had Technology, which was becoming known about how to etch tiny openings in learned while employed by Shockley. throughout the that layer and to use the remaining that this diode was much too difficult One pivotal contribution to silicon a project for the young company to proved satisfying. So by early 1955 of automatic means for production of as “Ma Bell’s cookbook.”5 Shockley oxide as a mask against the diffusion technology, however, originated entirely attempt at such an early stage in its Shockley was seeking to get back into diffused-base transistors.”3 could also call upon his old colleagues of trace impurities into the silicon. at Fairchild—the planar manufacturing evolution. We should concentrate industrial research—this time as the His original plan was to lure away Sparks, Tanenbaum, and Jack Morton, That technique would allow workers to process invented and developed by head of his own company.1 some of his most accomplished Bell head of device development, to visit the establish intricate patterns of n-type the . This crucial In March 1955 his sense of urgency Labs colleagues, such as Tanenbaum new California company and consult and p-type material in the silicon. technique, in which an oxide layer on in these efforts was stimulated by news and Morgan Sparks, who had fabricated with his charges in their converted Shockley circulated this document the silicon surface is used to protect from Bell Labs. Not only had chemist the first junction transistors. But one Quonset hut at 391 San Antonio among his technical staff (Fig. the sensitive p-n junctions beneath it, Henry Theurer finally produced by one they turned him down that fall, Road (Fig. 1). They spent most of that 3), giving them all a crucial early has served ever since 1961 as the basis silicon with impurity levels of less claiming their families were too deeply first year purchasing or building the glimpse of what would become one of microchip fabrication. And Hoerni’s than a part per billion using float-zone rooted in New Jersey and they could necessary equipment, such as silicon of the core enabling technologies conception of this method occurred . But and not bear to leave Bell Labs. Thus he crystal growers and diffusion furnaces, for silicon integrated circuits and substantially earlier than most historical D. E. Thomas had also succeeded in subsequently criss-crossed the United and learning how to use it in making ultimately lead to the emergence of accounts have thus far reported. forming the first diffused-base (and States, trying to recruit the best possible semiconductor devices. .7 By the time Frosch In this article, I try to set the record emitter) silicon transistor using samples scientists and engineers for Shockley On 1 November 1956, marvelous and Derick’s paper was published straight by recounting the important provided by Calvin Fuller (see article Semiconductor Laboratory. And he hired news reached the Mountain View firm in the Journal of The Electrochemical steps and major influences that brought by on p. 30). “Morrie some top-notch ones, too: that Shockley had won the 1956 Nobel Society the following September, silicon technology to California—in Tann has AlSb plus Al bonded,” reads Hoerni from Caltech, Jay Last from Prize in for the invention of the Shockley and his staff were already the process laying the technological a cryptic note in Shockley’s pocket MIT, and Robert Noyce from —as transistor, shared with developing ways to implement this foundations of Silicon Valley. notebook dated 23 March 1955, while well as physical chemist Gordon Moore and Walter Brattain. The next day they disruptive new technology.8 he was still at the Pentagon, referring from the Johns Hopkins Applied Physics stopped working before noon for a Silicon Technology Heads West to micrometers-thin aluminum and Laboratory. Among the eldest of this celebratory luncheon at nearby Rickey’s A Rebellion Erupts impurity layers diffused into group at age 28, Noyce was about the Studio Inn in Palo Alto. A photograph 2 By 1954 William Shockley was the silicon. only one with extensive semiconductor of that gathering (Fig. 2), with Shockley But all was not well at the becoming increasingly frustrated by Over the next few months, Shockley experience, having developed high- at the head of the table being toasted by Mountain View laboratory. By his lack of professional advancement at began talking with executives at RCA, frequency transistors at Noyce, Moore, Last, and others around 1957, members of the technical Bell Labs. Although head of transistor Raytheon, and Philco.4 him, has achieved iconic status in the 6 staff had begun to resent Shockley’s research, he had been passed over for elsewhere about the possibility of Only Noyce among them had lore of Silicon Valley. often heavy-handed management higher positions and seemed mired starting a firm expressly devoted to attended the third Bell Labs symposium Upon returning from style and his quirky selection of at this middle-management level producing diffused-silicon devices. For on transistor technology in January in mid-December, Shockley found a &D projects. After a couple of despite his many publications and most of that summer, he met with 1956, devoted to the advanced package in his from the patent worrisome incidents, he started patents, including the all-important only passing interest until he spoke silicon technologies just then being licensing engineer at investigating the backgrounds ones on the junction transistor. Thus with his friend and Caltech developed—especially diffusion. But Company, AT&T’s manufacturing arm. of a few of his employees. And he began casting around for other alumnus Arnold Beckman, founder and because Beckman had licensed the rights It contained a preprint of an article by he gave one of them a vicious, possible positions, taking a leave of president of Beckman Industries, Inc. In to the transistor patents from Bell’s Carl Frosch and Lincoln Derick titled embarrassing tongue-lashing that absence to serve as a visiting professor early September they met in Newport parent company, AT&T, Shockley and his “Surface Protection and Surface Masking many others overheard. Several of at Caltech and as a military advisor Beach, California, and agreed to found new employees had ready access to the during Diffusion in Silicon.” Frosch the early employees were extremely Fig. 3. Copy of the 14 December 1956 letter to Shockley in the Pentagon’s Weapons Systems had delivered a paper on diffusion in 9 a new division led by Shockley whose information presented there, including dissatisfied and ready to quit. from the Western Electric patent engineer that Evaluation Group. But neither option purposes included “the development early preprints of the papers that the third transistor symposium, but For reasons about which we can accompanied a preprint of Frosch and Derick’s paper on it did not contain much information only guess, Shockley had lost interest oxide masking, plus the initialed routing list of technical about use of a silicon-dioxide layer to in the original goal set for the staff members of Shockley Semiconductor Laboratory * Adapted in part from Crystal Fire: The Birth of the , by Michael Riordan and , W. W. Norton & Co., passivate the silicon surface because who read this paper. ( Archives) (1997), esp. pp. 225-275. See also M. Riordan and L. Hoddeson, “The Moses of Silicon Valley,” Physics Today, pp. 42-47 (December 1997). firm—manufacturing diffused- Bell Labs was still applying for a patent base transistors. He instead began

36 The Interface • Fall 2007 The Electrochemical Society Interface • Fall 2007 37 to focus the company’s talents and From Bell Labs to Silicon Valley: energies on devices that were then at the frontiers of semiconductor technology, A Saga of Semiconductor Technology Transfer, 1955-61* such as field-effect transistors and the four-layer p-n-p-n diode, often called the Shockley diode because he held by Michael Riordan a patent on it. At a given potential called the breakdown voltage, current would suddenly begin to flow through espite the fact that the Bell the two-terminal device, switching Telephone Laboratories had it from “off” to “on.” A remarkably Doriginated almost all the silicon simple, compact semiconductor switch, technology eventually used to invent it was the ultimate realization of a the integrated circuit, or microchip, this goal that had been subtly implanted revolutionary breakthrough occurred in Shockley’s mind at Bell Labs two elsewhere—at Texas Instruments, Inc., decades earlier.10 in Dallas, and Fairchild Semiconductor Although a brilliant conception, Corporation in Mountain View, however, the four-layer diode was California. In the latter case, the extremely difficult to fabricate with transfer of this technology occurred uniformity and reliability. Workers largely through the offices of a pivotal had to polish a silicon wafer to have but ultimately unsuccessful company, exactingly parallel planes on its Shockley Semiconductor Laboratory, Fig. 2. Senior staff of Shockley Semiconductor Laboratory toast their illustrious leader two sides and then carefully diffuse which had been formed by transistor William Shockley (seated at head of table) at a celebratory luncheon on 2 November impurities into the wafer from pioneer William Shockley. After 1956, the day after the announcement of his Nobel Prize. Seated at left is Gordon Moore both sides simultaneously. Lack of leaving Bell Labs in 1955, he brought and right of him is Sheldon Roberts. Robert Noyce stands in the middle and Jay Last is enough precision meant the diffused together a stellar team of scientists at far right, both lifting glasses. (Courtesy of Intel Corporation) impurities would penetrate to irregular, and engineers who later departed (in unpredictable depths, leading to Fig. 1. Shockley Semiconductor Laboratory at 391 San Antonio Road in Mountain View, September 1957) to found Fairchild, would be published in 1958 as the final on this important method. The preprint variations in the breakdown voltage California, as it appeared in 1960 after Clevite Corporation had purchased it and and other electrical characteristics.11 bringing a deep understanding of the volume in the classic series, Transistor also contained crucial new information renamed it Shockley Transistor Company. (Courtesy of Kurt Hübner) Noyce, Moore, and others thought diffused-silicon technologies they had Technology, which was becoming known about how to etch tiny openings in learned while employed by Shockley. throughout the semiconductor industry that layer and to use the remaining that this diode was much too difficult One pivotal contribution to silicon a project for the young company to proved satisfying. So by early 1955 of automatic means for production of as “Ma Bell’s cookbook.”5 Shockley oxide as a mask against the diffusion technology, however, originated entirely attempt at such an early stage in its Shockley was seeking to get back into diffused-base transistors.”3 could also call upon his old colleagues of trace impurities into the silicon. at Fairchild—the planar manufacturing evolution. We should concentrate industrial research—this time as the His original plan was to lure away Sparks, Tanenbaum, and Jack Morton, That technique would allow workers to process invented and developed by head of his own company.1 some of his most accomplished Bell head of device development, to visit the establish intricate patterns of n-type the physicist Jean Hoerni. This crucial In March 1955 his sense of urgency Labs colleagues, such as Tanenbaum new California company and consult and p-type material in the silicon. technique, in which an oxide layer on in these efforts was stimulated by news and Morgan Sparks, who had fabricated with his charges in their converted Shockley circulated this document the silicon surface is used to protect from Bell Labs. Not only had chemist the first junction transistors. But one Quonset hut at 391 San Antonio among his technical staff (Fig. the sensitive p-n junctions beneath it, Henry Theurer finally produced by one they turned him down that fall, Road (Fig. 1). They spent most of that 3), giving them all a crucial early has served ever since 1961 as the basis silicon with impurity levels of less claiming their families were too deeply first year purchasing or building the glimpse of what would become one of microchip fabrication. And Hoerni’s than a part per billion using float-zone rooted in New Jersey and they could necessary equipment, such as silicon of the core enabling technologies conception of this method occurred refining. But Morris Tanenbaum and not bear to leave Bell Labs. Thus he crystal growers and diffusion furnaces, for silicon integrated circuits and substantially earlier than most historical D. E. Thomas had also succeeded in subsequently criss-crossed the United and learning how to use it in making ultimately lead to the emergence of accounts have thus far reported. forming the first diffused-base (and States, trying to recruit the best possible semiconductor devices. Silicon Valley.7 By the time Frosch In this article, I try to set the record emitter) silicon transistor using samples scientists and engineers for Shockley On 1 November 1956, marvelous and Derick’s paper was published straight by recounting the important provided by Calvin Fuller (see article Semiconductor Laboratory. And he hired news reached the Mountain View firm in the Journal of The Electrochemical steps and major influences that brought by Nick Holonyak on p. 30). “Morrie some top-notch ones, too: physicists that Shockley had won the 1956 Nobel Society the following September, silicon technology to California—in Tann has AlSb plus Al bonded,” reads Hoerni from Caltech, Jay Last from Prize in physics for the invention of the Shockley and his staff were already the process laying the technological a cryptic note in Shockley’s pocket MIT, and Robert Noyce from Philco—as transistor, shared with John Bardeen developing ways to implement this foundations of Silicon Valley. notebook dated 23 March 1955, while well as physical chemist Gordon Moore and Walter Brattain. The next day they disruptive new technology.8 he was still at the Pentagon, referring from the Johns Hopkins Applied Physics stopped working before noon for a Silicon Technology Heads West to micrometers-thin aluminum and Laboratory. Among the eldest of this celebratory luncheon at nearby Rickey’s A Rebellion Erupts antimony impurity layers diffused into group at age 28, Noyce was about the Studio Inn in Palo Alto. A photograph 2 By 1954 William Shockley was the silicon. only one with extensive semiconductor of that gathering (Fig. 2), with Shockley But all was not well at the becoming increasingly frustrated by Over the next few months, Shockley experience, having developed high- at the head of the table being toasted by Mountain View laboratory. By his lack of professional advancement at began talking with executives at RCA, frequency germanium transistors at Noyce, Moore, Last, and others around 1957, members of the technical Bell Labs. Although head of transistor Raytheon, Texas Instruments and Philco.4 him, has achieved iconic status in the 6 staff had begun to resent Shockley’s research, he had been passed over for elsewhere about the possibility of Only Noyce among them had lore of Silicon Valley. often heavy-handed management higher positions and seemed mired starting a firm expressly devoted to attended the third Bell Labs symposium Upon returning from Stockholm style and his quirky selection of at this middle-management level producing diffused-silicon devices. For on transistor technology in January in mid-December, Shockley found a R&D projects. After a couple of despite his many publications and most of that summer, he met with 1956, devoted to the advanced package in his mail from the patent worrisome incidents, he started patents, including the all-important only passing interest until he spoke silicon technologies just then being licensing engineer at Western Electric investigating the backgrounds ones on the junction transistor. Thus with his friend and fellow Caltech developed—especially diffusion. But Company, AT&T’s manufacturing arm. of a few of his employees. And he began casting around for other alumnus Arnold Beckman, founder and because Beckman had licensed the rights It contained a preprint of an article by he gave one of them a vicious, possible positions, taking a leave of president of Beckman Industries, Inc. In to the transistor patents from Bell’s Carl Frosch and Lincoln Derick titled embarrassing tongue-lashing that absence to serve as a visiting professor early September they met in Newport parent company, AT&T, Shockley and his “Surface Protection and Surface Masking many others overheard. Several of at Caltech and as a military advisor Beach, California, and agreed to found new employees had ready access to the during Diffusion in Silicon.” Frosch the early employees were extremely Fig. 3. Copy of the 14 December 1956 letter to Shockley in the Pentagon’s Weapons Systems had delivered a paper on diffusion in 9 a new division led by Shockley whose information presented there, including dissatisfied and ready to quit. from the Western Electric patent engineer that Evaluation Group. But neither option purposes included “the development early preprints of the papers that the third transistor symposium, but For reasons about which we can accompanied a preprint of Frosch and Derick’s paper on it did not contain much information only guess, Shockley had lost interest oxide masking, plus the initialed routing list of technical about use of a silicon-dioxide layer to in the original goal set for the staff members of Shockley Semiconductor Laboratory * Adapted in part from Crystal Fire: The Birth of the Information Age, by Michael Riordan and Lillian Hoddeson, W. W. Norton & Co., New York passivate the silicon surface because who read this paper. (Stanford University Archives) (1997), esp. pp. 225-275. See also M. Riordan and L. Hoddeson, “The Moses of Silicon Valley,” Physics Today, pp. 42-47 (December 1997). firm—manufacturing diffused- Bell Labs was still applying for a patent base transistors. He instead began

36 The Electrochemical Society Interface • Fall 2007 The Electrochemical Society Interface • Fall 2007 37 Riordan had learned and developed at Shockley staff members remaining at Shockley (continued from previous page) Semiconductor. Roberts grew silicon Semiconductor Laboratory were still crystals while Moore tackled issues of struggling to four-layer . first on three-layer devices such as the diffusing impurities into them. Last and diffused-base transistor, they argued, and Noyce took on the intricate tasks involved The Planar Manufacturing get a product out the door. After that they in oxide masking and , could begin attempting the more difficult developing a “step-and-repeat” camera to Process problem of fabricating the four-layer aid them in defining tiny, precise n-type But a serious problem emerged with diode. But Shockley was not listening and and p-type areas in the silicon wafers.16 these mesa transistors that threatened began devoting more and more of the By May 1958 the group had the continued success of the fledgling company’s resources to this pet project. successfully fabricated prototype n-p-n company. Some of these devices He eventually set up a separate R&D silicon transistors, based on the “mesa” experienced catastrophic failures that group working mostly in secret on the structure pioneered by Tanenbaum and were traced to bits of foreign material four-layer diode, which further angered Thomas at Bell Labs.17 In such a structure, being dislodged and becoming attached technical staff members left out of the a thin p-type base layer and an n-type to the exposed junctions, attracted loop. emitter layer were diffused into an n-type by strong electric fields there. Merely The conflict reached the boiling point silicon wafer that formed the collector tapping on the metal cans housing in mid-May 1957, when Beckman came of this bipolar junction transistor. Oxide these transistors with a pencil eraser was up for a meeting to discuss the division’s masking and photolithography were used often sufficient to trigger such failures. problems and plans. Taking umbrage at his partner’s proposals and request Fig. 5. Cross-sectional drawings of mesa and planar transistors, copied from Jean Hoerni’s 1961 publication, “Planar Silicon Transistors and to control expenses, however, Shockley Diodes,” Fairchild Semiconductor Corp. Report No. TP-14. (Stanford University Archives) stood up and stomped out of the room, shouting, “If you don’t like what we’re doing here, I can take this group and get While the Fairchild founders were called this “the most the failure-prone mesas (Fig. 5). What’s support any place else!”12 busy setting up offices and equipment important in the history of more, he soon showed they had far But nobody left the room except on Charleston Street, Hoerni had the semiconductor industry.”23 superior electrical characteristics—with Shockley. His outburst triggered a four- already begun to consider a new and When Hoerni conceived this higher gains, and leakage currents month sequence of events that led in different way to fabricate transistors. approach in December 1957, Fairchild often less than a nanoampere. As September to the departure of eight of his An admitted contrarian, he suggested lacked the technical abilities needed Hoerni concluded in a paper that he top lieutenants—including Hoerni, Last, that instead of etching away the to implement it. And the new firm’s presented at an October 1960 meeting Moore, and Noyce. A pivotal moment in oxide layer at the end of processing, talents and energies had to be devoted on devices in , , the history of the semiconductor industry, it instead be left in place to protect the almost entirely to getting its first mesa “the planar design offers considerable it is briefly recorded in yet another sensitive junctions. He entered this transistors out the door. But once the improvement and stabilization of the Shockley notebook: “Wed 18 Sep—Group revolutionary new idea on pages 3 and techniques of diffusion, oxide masking parameters most likely to suffer from resigns.”13 4 of his crisp new lab notebook on and photolithography had been surface contamination.”24 1 December 1957, hardly more than mastered in 1958, he returned to his On 14 January 1959, Hoerni had A Successful Semiconductor two months after the firm had been conception as a way to address their had a Fairchild secretary type up a Company founded!21 reliability problems. With the help of formal patent disclosure to give the Titled “Method of protecting Last, who made the extra mask he company’s attorney. Apart from a few Within days the group of eight exposed p-n junctions at the surface needed, Hoerni proved out this idea in typographical changes and improved rebels—which also included the of silicon transistors by oxide masking March 1959, fabricating prototype planar drawings, it is completely identical to metallurgist Sheldon Roberts and techniques,” it was witnessed that transistors that avoided the problems of the 1 December 1957 entry in his lab engineers , , and same day by Noyce. The very first notebook—indicating the great —signed a $1.38 million paragraph is revealing: accuracy of his original theoretical agreement brokered by the investment- The general idea underlying insight.25 Just a few months later banking firm Hayden Stone with Fairchild this invention is the building Moore and Noyce, who had by then Camera and Instrument Corporation of Fig. 4. Official photograph (circa 1960) of the eight Fairchild founders, who later became famous in Silicon Valley lore as the “Fairchild Eight.” Left to right, Gordon Moore, up of an oxide layer prior to emerged as the natural leaders of Syosset, New York, on Long Island (Fig. Sheldon Roberts, Eugene Kleiner, Robert Noyce, Victor Grinich, Julius Blank, Jean diffusion of dopant atoms the new firm, decided to reorient 4). Using these start-up funds, the eight Hoerni, and Jay Last. (Wayne Miller photograph courtesy of Fairchild Semiconductor at the places on the surface Fairchild’s transistor production began setting up a new firm known as Corp. and Magnum Photos) of the transistor at which p- lines to this promising new planar the Fairchild Semiconductor Corporation n junctions are expected to approach.26 just over a mile away on Charleston Street emerge from the body of the in Palo Alto.14 semiconductor. The oxide layer The Monolithic Idea At almost exactly the same moment, to define these regions. But after attaching This was a fundamental design flaw of so obtained is an integrant Frosch and Derick’s revolutionary paper the electrical leads, workers etched away the exposed mesa structure. A possible (sic) part of the device and will Becomes Reality was published in the September 1957 issue any remaining portions of the silicon- solution being considered was to try to protect the otherwise exposed of the Journal of The Electrochemical Society. dioxide sheath, exposing the underlying leave the oxide layer in place over the junctions from contamination Only nine days after Hoerni’s But because they had read the preprint of junctions. This was standard practice in junctions—or deliberately form one there and possible electrical leakage formal patent disclosure, Noyce, this paper while working at Shockley Lab, the semiconductor industry at the time, in order to protect them during the final due to subsequent handling, stimulated in part by the urgings of the Fairchild upstarts had a big head start for the oxide layer was widely viewed as processing steps.19 cleaning, [and] canning of the the patent attorney, began writing in on most of the competition. According to “dirty”—especially at Bell Labs. A far more elegant approach was put device.22 his lab notebook under the heading, Roberts, some of them—including Hoerni, This approach had the further “Method of isolating multiple Fairchild’s transistors were the first forth by Jean Hoerni, who had been 27 Moore, and Noyce—had already begun commercially successful mesa transistors, leading Fairchild’s efforts to fabricate p-n- advantages that all electrical leads devices.” He thought it a terrible experimenting with double diffusion and beating out Texas Instruments, which p transistors to satisfy IBM’s specifications. could be attached and all processing waste that all these transistors on a oxide masking while they were there.15 was also working on similar devices. A theoretical physicist by training, steps performed from the same single wafer had to be apart, have The first commercial product Not only did they meet IBM’s stringent Swiss-born Hoerni had earned advanced side of the silicon wafer—features leads attached to them individually, attempted by Fairchild was a high- specifications, but they also began to degrees from the Universities of Geneva that would eventually prove to and then be soldered back together frequency, diffused-base silicon tran- be used in other avionics applications. and Cambridge, and often worked on be of inestimable value when this with other components to assemble sistor to drive core memory in an onboard Fairchild Semiconductor recorded sales his own. At Shockley Lab he at first planar manufacturing process Fig. 6. Drawings from Robert Noyce’s patent on the complete circuits. What if all the that IBM was making for the was later applied to fabricating integrated circuit, showing how he proposed to use Hoerni’s interconnections could be instead of over $500,000 —almost all of it had an office separate from the Quonset planar processing technique to form the necessary p-n guidance-and-control system of the -70 coming from this product—in 1958, its hut, where he could go off by himself integrated circuits. With only placed directly on the wafer surface junctions in silicon beneath a protective oxide layer. Silicon during the manufacturing process? bomber. The eight partners began first full year of operations, and ended the to calculate diffusion curves and other slight exaggeration, historian of is denoted by 11 here and by 12. applying the silicon technology they year in the black.18 Meanwhile, the loyal useful theoretical quantities.20 technology Christophe Lécuyer has Here again, the silicon-dioxide

38 The Electrochemical Society Interface • Fall 2007 The Electrochemical Society Interface • Fall 2007 39 Riordan had learned and developed at Shockley staff members remaining at Shockley (continued from previous page) Semiconductor. Roberts grew silicon Semiconductor Laboratory were still crystals while Moore tackled issues of struggling to make four-layer diodes. first on three-layer devices such as the diffusing impurities into them. Last and diffused-base transistor, they argued, and Noyce took on the intricate tasks involved The Planar Manufacturing get a product out the door. After that they in oxide masking and photolithography, could begin attempting the more difficult developing a “step-and-repeat” camera to Process problem of fabricating the four-layer aid them in defining tiny, precise n-type But a serious problem emerged with diode. But Shockley was not listening and and p-type areas in the silicon wafers.16 these mesa transistors that threatened began devoting more and more of the By May 1958 the group had the continued success of the fledgling company’s resources to this pet project. successfully fabricated prototype n-p-n company. Some of these devices He eventually set up a separate R&D silicon transistors, based on the “mesa” experienced catastrophic failures that group working mostly in secret on the structure pioneered by Tanenbaum and were traced to bits of foreign material four-layer diode, which further angered Thomas at Bell Labs.17 In such a structure, being dislodged and becoming attached technical staff members left out of the a thin p-type base layer and an n-type to the exposed junctions, attracted loop. emitter layer were diffused into an n-type by strong electric fields there. Merely The conflict reached the boiling point silicon wafer that formed the collector tapping on the metal cans housing in mid-May 1957, when Beckman came of this bipolar junction transistor. Oxide these transistors with a pencil eraser was up for a meeting to discuss the division’s masking and photolithography were used often sufficient to trigger such failures. problems and plans. Taking umbrage at his partner’s proposals and request Fig. 5. Cross-sectional drawings of mesa and planar transistors, copied from Jean Hoerni’s 1961 publication, “Planar Silicon Transistors and to control expenses, however, Shockley Diodes,” Fairchild Semiconductor Corp. Report No. TP-14. (Stanford University Archives) stood up and stomped out of the room, shouting, “If you don’t like what we’re doing here, I can take this group and get While the Fairchild founders were called this planar process “the most the failure-prone mesas (Fig. 5). What’s support any place else!”12 busy setting up offices and equipment important innovation in the history of more, he soon showed they had far But nobody left the room except on Charleston Street, Hoerni had the semiconductor industry.”23 superior electrical characteristics—with Shockley. His outburst triggered a four- already begun to consider a new and When Hoerni conceived this higher gains, and leakage currents month sequence of events that led in different way to fabricate transistors. approach in December 1957, Fairchild often less than a nanoampere. As September to the departure of eight of his An admitted contrarian, he suggested lacked the technical abilities needed Hoerni concluded in a paper that he top lieutenants—including Hoerni, Last, that instead of etching away the to implement it. And the new firm’s presented at an October 1960 meeting Moore, and Noyce. A pivotal moment in oxide layer at the end of processing, talents and energies had to be devoted on electron devices in Washington, DC, the history of the semiconductor industry, it instead be left in place to protect the almost entirely to getting its first mesa “the planar design offers considerable it is briefly recorded in yet another sensitive junctions. He entered this transistors out the door. But once the improvement and stabilization of the Shockley notebook: “Wed 18 Sep—Group revolutionary new idea on pages 3 and techniques of diffusion, oxide masking parameters most likely to suffer from resigns.”13 4 of his crisp new lab notebook on and photolithography had been surface contamination.”24 1 December 1957, hardly more than mastered in 1958, he returned to his On 14 January 1959, Hoerni had A Successful Semiconductor two months after the firm had been conception as a way to address their had a Fairchild secretary type up a Company founded!21 reliability problems. With the help of formal patent disclosure to give the Titled “Method of protecting Last, who made the extra mask he company’s attorney. Apart from a few Within days the group of eight exposed p-n junctions at the surface needed, Hoerni proved out this idea in typographical changes and improved rebels—which also included the of silicon transistors by oxide masking March 1959, fabricating prototype planar drawings, it is completely identical to metallurgist Sheldon Roberts and techniques,” it was witnessed that transistors that avoided the problems of the 1 December 1957 entry in his lab engineers Julius Blank, Victor Grinich, and same day by Noyce. The very first notebook—indicating the great Eugene Kleiner—signed a $1.38 million paragraph is revealing: accuracy of his original theoretical agreement brokered by the investment- The general idea underlying insight.25 Just a few months later banking firm Hayden Stone with Fairchild this invention is the building Moore and Noyce, who had by then Camera and Instrument Corporation of Fig. 4. Official photograph (circa 1960) of the eight Fairchild founders, who later became famous in Silicon Valley lore as the “Fairchild Eight.” Left to right, Gordon Moore, up of an oxide layer prior to emerged as the natural leaders of Syosset, New York, on Long Island (Fig. Sheldon Roberts, Eugene Kleiner, Robert Noyce, Victor Grinich, Julius Blank, Jean diffusion of dopant atoms the new firm, decided to reorient 4). Using these start-up funds, the eight Hoerni, and Jay Last. (Wayne Miller photograph courtesy of Fairchild Semiconductor at the places on the surface Fairchild’s transistor production began setting up a new firm known as Corp. and Magnum Photos) of the transistor at which p- lines to this promising new planar the Fairchild Semiconductor Corporation n junctions are expected to approach.26 just over a mile away on Charleston Street emerge from the body of the in Palo Alto.14 semiconductor. The oxide layer The Monolithic Idea At almost exactly the same moment, to define these regions. But after attaching This was a fundamental design flaw of so obtained is an integrant Frosch and Derick’s revolutionary paper the electrical leads, workers etched away the exposed mesa structure. A possible (sic) part of the device and will Becomes Reality was published in the September 1957 issue any remaining portions of the silicon- solution being considered was to try to protect the otherwise exposed of the Journal of The Electrochemical Society. dioxide sheath, exposing the underlying leave the oxide layer in place over the junctions from contamination Only nine days after Hoerni’s But because they had read the preprint of junctions. This was standard practice in junctions—or deliberately form one there and possible electrical leakage formal patent disclosure, Noyce, this paper while working at Shockley Lab, the semiconductor industry at the time, in order to protect them during the final due to subsequent handling, stimulated in part by the urgings of the Fairchild upstarts had a big head start for the oxide layer was widely viewed as processing steps.19 cleaning, [and] canning of the the patent attorney, began writing in on most of the competition. According to “dirty”—especially at Bell Labs. A far more elegant approach was put device.22 his lab notebook under the heading, Roberts, some of them—including Hoerni, This approach had the further “Method of isolating multiple Fairchild’s transistors were the first forth by Jean Hoerni, who had been 27 Moore, and Noyce—had already begun commercially successful mesa transistors, leading Fairchild’s efforts to fabricate p-n- advantages that all electrical leads devices.” He thought it a terrible experimenting with double diffusion and beating out Texas Instruments, which p transistors to satisfy IBM’s specifications. could be attached and all processing waste that all these transistors on a oxide masking while they were there.15 was also working on similar devices. A theoretical physicist by training, steps performed from the same single wafer had to be cut apart, have The first commercial product Not only did they meet IBM’s stringent Swiss-born Hoerni had earned advanced side of the silicon wafer—features leads attached to them individually, attempted by Fairchild was a high- specifications, but they also began to degrees from the Universities of Geneva that would eventually prove to and then be soldered back together frequency, diffused-base silicon tran- be used in other avionics applications. and Cambridge, and often worked on be of inestimable value when this with other components to assemble sistor to drive core memory in an onboard Fairchild Semiconductor recorded sales his own. At Shockley Lab he at first planar manufacturing process Fig. 6. Drawings from Robert Noyce’s patent on the complete circuits. What if all the computer that IBM was making for the was later applied to fabricating integrated circuit, showing how he proposed to use Hoerni’s interconnections could be instead of over $500,000 —almost all of it had an office separate from the Quonset planar processing technique to form the necessary p-n guidance-and-control system of the B-70 coming from this product—in 1958, its hut, where he could go off by himself integrated circuits. With only placed directly on the wafer surface junctions in silicon beneath a protective oxide layer. Silicon during the manufacturing process? bomber. The eight partners began first full year of operations, and ended the to calculate diffusion curves and other slight exaggeration, historian of is denoted by 11 here and silicon dioxide by 12. applying the silicon technology they year in the black.18 Meanwhile, the loyal useful theoretical quantities.20 technology Christophe Lécuyer has Here again, the silicon-dioxide

38 The Electrochemical Society Interface • Fall 2007 The Electrochemical Society Interface • Fall 2007 39 Riordan then only to check up on how work was channels with an inert epoxy resin. But at Stanford University. He is author Laboratories Record (June 1958), pp. Exposed p-n Junctions at the Surface (continued from previous page) progressing.30 in the fall of 1960, p-n junction isolation of The Hunting of the Quark (Simon & 202-206. of Silicon Transistors by Oxide Production of planar transistors was proved the superior alternative. The Schuster, 1987) and coauthor of Crystal 12. Shockley quoted by Moore, interview Masking Techniques,” 14 January proceeding apace in 1959, especially group made successful prototype flip- Fire: The Birth of the Information Age with Hoddeson and Riordan. layer played the key role as an enabling 1959, copy provided courtesy of Jay at the urging of Autonetics, a division flop circuits involving four transistors (W. W. Norton, 1997), which won the 13. William Shockley, “Golden West technology. Last. of North American Aviation that was (Fig. 7) and followed them up with other 1999 Sally Hacker Prize of the Society Theme Book” (personal diary, 1956- “In many applications now it would 26. . Lécuyer, Making Silicon Valley: building computer systems for the integrated logic circuits. In March 1961 for the History of Technology. A Fellow 57), Stanford University Archives, be desirable to make multiple devices Innovation and the Growth of High Minuteman missile and had extremely Fairchild announced the “Micrologic” of the American Physical Society and a Shockley papers, accession listing 95- on a single piece of silicon in order to be Tech, 1930-1970, .p. 152, The MIT stringent demands for precision and family of integrated circuits, half a Guggenheim Fellow, he received the 153, box 2B, p. 58. able to make interconnections between Press, Cambridge, MA (2006). reliability. It helped Fairchild to learn year ahead of the Texas Instruments prestigious Andrew W. Gemant Award 14. The most compete reference on the devices as part of the manufacturing 27. Robert N. Noyce, Fairchild Semi- that a Bell Labs group under M. M. competition. The use of Hoerni’s planar of the American Institute of Physics. formation and early years of the process, and thus reduce size, weight, conductor Corp. lab notebook, Atalla had developed “thermal” methods processing technique had made all the He may be reached at mriordan@ucsc. Fairchild Semiconductor Corporation etc., as well as cost per active element,” pp. 70–74 (23 January 1959), copy of growing the oxide, mitigating the difference.33 is Chapter 4 of Christophe Lécuyer’s Noyce began.28 To Hoerni’s planar edu. provided courtesy of Jay Last. problems of dangling bonds and Making Silicon Valley: Innovation and process, he added the provision that 28. Robert N. Noyce, Fairchild Semi- “surface states” at the interface between the Growth of High Tech, 1930-1970, pp. aluminum interconnections between Summary conductor Corp. lab notebook, p. the oxide layer and the silicon.31 Using References 129–167, The MIT Press, Cambridge, the individual devices be placed on top 70 (23 January 1959), copy provided this thermally grown oxide layer, MA (2006). See also Riordan and of the oxide layer, which would insulate Throughout this technological evo- courtesy of Jay Last. Atalla and Dawon Khang subsequently Hoddeson, Crystal Fire, pp. 262-269, these leads from the silicon beneath lution, from diffused-base transistors 1. M. Riordan and L. Hoddeson, 29. Robert N. Noyce, U.S. Patent No. made the first truly successful field- W. W. Norton & Co., New York it. And these circuit components to planar integrated circuits, the Crystal Fire, pp. 225-230, W. W. 2,981,877, “- effect transistors in 1960—what soon (1997); Leslie Berlin, The Man Behind could be electrically isolated from one silicon-dioxide layer discovered in Norton & Co., New York (1997). and-Lead Structure,” filed 30 July became known as the metal-oxide- the Microchip: Robert Noyce and the another by interposing back-to-back p- 1955 by Frosch and Derick remained 2. Riordan and Hoddeson, Crystal 1959, awarded 25 . Here semiconductor, or MOS, transistor (see Invention of Silicon Valley, pp. 82-157, n junctions between them. These were at the center of the action. Without Fire, pp. 230, W. W. Norton & it is crucial to note that article by Ross Bassett on p. 46) .32 , New York the principal new ideas at the heart this supple, flexible interface between Co., New York (1997); Riordan had conceived similar ideas several It took Last’s development team nearly (2005); and Ross Knox Bassett, To of Noyce’s famous integrated-circuit silicon and its environment, the and Hoddeson, Physics Today, p. 43 months earlier at Texas Instruments, two years to produce commercially the Digital Age: Research Labs, Start- patent, titled “Semiconductor Device- microchip as we know it today would (December 1997). See also William and that the two men are generally available integrated circuits using Up Companies, and the Rise of MOS and-Lead Structure,” which he applied be unthinkable. It is the sine qua non of Shockley, “Memorandums,” perceived as the co-inventors planar processing. Higher precision Technology, pp. 45-53, The Johns for later that year (Fig. 6).29 the semiconductor industry, and what (personal diary 1955–56), Stanford of the integrated circuit. Kilby, was needed for oxide masking and Hopkins University Press, Baltimore, The difficult task of actually most distinguishes silicon from all other University Archives, Shockley however, did not have any idea of photolithography—and more steps were MD (2002). fabricating microelectronic circuits material alternatives. papers, accession listing 95-153, the planar technology used in the involved. And a big problem was isolating 15. C. Lécuyer, Making Silicon Valley: according to these prescriptions fell to Although ultimately unsuccessful box 2B. Noyce patent, and proposed that the circuit elements. Workers even Innovation and the Growth of High Jay Last, who formed a development financially, the Shockley Semicon- 3. Riordan and Hoddeson, Crystal individual components on a chip be attempted a physical-isolation approach Tech, 1930-1970, p. 150, The MIT group to realize this goal. In March 1959 ductor Laboratory played an important Fire, pp. 231-234, W. W. Norton connected by what became known suggested by Last that involved etching Press, Cambridge, MA (2006). Noyce took over as Fairchild’s General technology-transfer role in this saga. & Co., New York (1997); Riordan as “flying wires.” See Jack Kilby, channels between the components from 16. C. Lécuyer, Making Silicon Valley: Manager and Moore replaced him as It acted as the crucial incubator where and Hoddeson, Physics Today, p. 44 IEEE Transactions on Electron Devices, the back of the wafer all the way to the Innovation and the Growth of High research director. Noyce rarely showed men came together with ideas on the (December 1997). 23, 648 (July 1976); and Michael S. fragile oxide layer, then filling these Tech, 1930-1970, pp. 140-148, The up in the laboratory after that—and West Coast, and began working together 4. Riordan and Hoddeson, Crystal Fire, Wolff, “The Genesis of the Integrated on the revolutionary implications of the pp. 236-240, W. W. Norton & Co., MIT Press, Cambridge, MA (2006). Circuit,” IEEE Spectrum (August new silicon technologies pioneered in New York (1997). 17. M. Tanenbaum and D. Thomas, 1976), pp. 38–45. the mid- by Bell Labs. They were 5. F. J. Biondi, et al., Editors, Transistor “Diffused Emitter and Base Silicon 30. Jay Last, private communications. so revolutionary, in fact, that it took Technology, 3 Vols., D. Van Nostrand, Transistors,” Technical Lécuyer observes, “Interestingly, a small rebellion, much celebrated in New York (1958); Vol. 3 is devoted Journal, Vol. 35 (January 1956), pp. Noyce played no role in this effort.” Silicon Valley lore, to realize them. to diffusion technologies, including 1–22. See C. Lécuyer, Making Silicon Valley: The development and implementation contributions by Frosch, Fuller, 18. C. Lécuyer, Making Silicon Valley: Innovation and the Growth of High of planar processing at Fairchild and Tanenbaum. Noyce attended Innovation and the Growth of High Tech, 1930-1970, p. 157, The MIT Semiconductor Corporation was what the third symposium as a Philco Tech, 1930-1970, p. 149, The MIT Press, Cambridge, MA (2006). really threw the floodgates open to employee. Press, Cambridge, MA (2006). 31. M. M. Atalla, et al., “Stabilization of extremely reliable silicon transistors 6. Riordan and Hoddeson, Crystal Fire, 19. C. Lécuyer, Making Silicon Valley: Silicon Surfaces by Thermally Grown and microchips of endlessly growing pp. 241-243, W. W. Norton & Co., Innovation and the Growth of High Oxides,” Bell System Technical Journal, complexity. New York (1997). Tech, 1930-1970, pp. 149-150, The Vol. 38 (1959), pp. 749–783. The junction transistor’s inventor, 7. Riordan and Hoddeson, Physics MIT Press, Cambridge, MA (2006). 32. On the origins of the MOS transistor, William Shockley, dimly perceived Today, p. 45 (December 1997). 20. C. Lécuyer, Making Silicon Valley: see Ross Knox Bassett, To the Digital this bright future, but he lacked the 8. C. J. Frosch and L. Derick, Journal of Innovation and the Growth of High Tech, Age: Research Labs, Start-Up Companies, management skills to bring it off under The Electrochemical Society, 104, 547 1930-1970, pp. 150-152, The MIT and the Rise of MOS Technology, pp. his direction. In the early , he (1957). Press, Cambridge, MA (2006); also 22–33, The began teaching at Stanford University, 9. Gordon Moore, interview with Jay Last, private communication. Press, Baltimore, MD (2002); and first as a lecturer and then as a L. Hoddeson and M. Riordan, 11 21. Jean A. Hoerni, Fairchild Semi- , IEEE Transactions on professor of engineering and applied January 1996. conductor Corp. lab notebook, pp. 3- Electron Devices, 23, 655 (July 1976). science, while his company slowly 10. Soon after Bell Labs research 4 (1 December 1957), copy provided 33. C. Lécuyer, Making Silicon Valley: failed and was eventually dissolved. director hired courtesy of Jay Last. Innovation and the Growth of High Yet without Shockley’s contributions Shockley in 1936, he impressed on 22. Jean A. Hoerni, Fairchild Semi- Tech, 1930-1970, pp. 155-59, The MIT in bringing together this stellar group his new employee the needs of the conductor Corp. lab notebook, p. 3 Press, Cambridge, MA (2006); and of researchers and introducing them to Bell System to replace its unreliable (1 December 1957), copy provided Jay Last, private communication. silicon technology, there would be no electromechanical switches with courtesy of Jay Last. Silicon Valley—at least not in Northern a better, faster -state version. 23. C. Lécuyer, Making Silicon Valley: California. Kelly’s urgings helped stimulate Innovation and the Growth of High Shockley to invent the junction Tech, 1930-1970, p. 150, The MIT transistor and probably promoted Press, Cambridge, MA (2006). About the Author his interest in the four-layer diode. 24. Jean A. Hoerni, “Planar Silicon Fig. 7. One of the earliest prototype integrated circuits fabricated by Jay See Riordan and Hoddeson, Crystal Transistors and Diode,” Fairchild Last’s development group at Fairchild using the planar processing technique. Michael Riordan earned his PhD in Fire, pp. 81-82, W. W. Norton & Co., Semiconductor Report No. TP-14, This “flip-flop” logic circuit employed four transistors. (Courtesy of Fairchild physics from MIT. He is an adjunct New York (1997). p. 9, Fairchild Collection, Stanford Semiconductor Corp.) professor of physics at the University of 11. On the difficulties of fabricating University Archives. California, Santa Cruz, and a Lecturer uniform, reliable four-layer diodes, 25. Jean A. Hoerni, Fairchild patent see W. J. Pietenpol, “Transistor disclosure, “Method of Protecting Designs: The First Decade,” Bell 40 The Electrochemical Society Interface • Fall 2007 The Electrochemical Society Interface • Fall 2007 41 Riordan then only to check up on how work was channels with an inert epoxy resin. But at Stanford University. He is author Laboratories Record (June 1958), pp. Exposed p-n Junctions at the Surface (continued from previous page) progressing.30 in the fall of 1960, p-n junction isolation of The Hunting of the Quark (Simon & 202-206. of Silicon Transistors by Oxide Production of planar transistors was proved the superior alternative. The Schuster, 1987) and coauthor of Crystal 12. Shockley quoted by Moore, interview Masking Techniques,” 14 January proceeding apace in 1959, especially group made successful prototype flip- Fire: The Birth of the Information Age with Hoddeson and Riordan. layer played the key role as an enabling 1959, copy provided courtesy of Jay at the urging of Autonetics, a division flop circuits involving four transistors (W. W. Norton, 1997), which won the 13. William Shockley, “Golden West technology. Last. of North American Aviation that was (Fig. 7) and followed them up with other 1999 Sally Hacker Prize of the Society Theme Book” (personal diary, 1956- “In many applications now it would 26. C. Lécuyer, Making Silicon Valley: building computer systems for the integrated logic circuits. In March 1961 for the History of Technology. A Fellow 57), Stanford University Archives, be desirable to make multiple devices Innovation and the Growth of High Minuteman missile and had extremely Fairchild announced the “Micrologic” of the American Physical Society and a Shockley papers, accession listing 95- on a single piece of silicon in order to be Tech, 1930-1970, .p. 152, The MIT stringent demands for precision and family of integrated circuits, half a Guggenheim Fellow, he received the 153, box 2B, p. 58. able to make interconnections between Press, Cambridge, MA (2006). reliability. It helped Fairchild to learn year ahead of the Texas Instruments prestigious Andrew W. Gemant Award 14. The most compete reference on the devices as part of the manufacturing 27. Robert N. Noyce, Fairchild Semi- that a Bell Labs group under M. M. competition. The use of Hoerni’s planar of the American Institute of Physics. formation and early years of the process, and thus reduce size, weight, conductor Corp. lab notebook, Atalla had developed “thermal” methods processing technique had made all the He may be reached at mriordan@ucsc. Fairchild Semiconductor Corporation etc., as well as cost per active element,” pp. 70–74 (23 January 1959), copy of growing the oxide, mitigating the difference.33 is Chapter 4 of Christophe Lécuyer’s Noyce began.28 To Hoerni’s planar edu. provided courtesy of Jay Last. problems of dangling bonds and Making Silicon Valley: Innovation and process, he added the provision that 28. Robert N. Noyce, Fairchild Semi- “surface states” at the interface between the Growth of High Tech, 1930-1970, pp. aluminum interconnections between Summary conductor Corp. lab notebook, p. the oxide layer and the silicon.31 Using References 129–167, The MIT Press, Cambridge, the individual devices be placed on top 70 (23 January 1959), copy provided this thermally grown oxide layer, MA (2006). See also Riordan and of the oxide layer, which would insulate Throughout this technological evo- courtesy of Jay Last. Atalla and Dawon Khang subsequently Hoddeson, Crystal Fire, pp. 262-269, these leads from the silicon beneath lution, from diffused-base transistors 1. M. Riordan and L. Hoddeson, 29. Robert N. Noyce, U.S. Patent No. made the first truly successful field- W. W. Norton & Co., New York it. And these circuit components to planar integrated circuits, the Crystal Fire, pp. 225-230, W. W. 2,981,877, “Semiconductor Device- effect transistors in 1960—what soon (1997); Leslie Berlin, The Man Behind could be electrically isolated from one silicon-dioxide layer discovered in Norton & Co., New York (1997). and-Lead Structure,” filed 30 July became known as the metal-oxide- the Microchip: Robert Noyce and the another by interposing back-to-back p- 1955 by Frosch and Derick remained 2. Riordan and Hoddeson, Crystal 1959, awarded 25 April 1961. Here semiconductor, or MOS, transistor (see Invention of Silicon Valley, pp. 82-157, n junctions between them. These were at the center of the action. Without Fire, pp. 230, W. W. Norton & it is crucial to note that Jack Kilby article by Ross Bassett on p. 46) .32 Oxford University Press, New York the principal new ideas at the heart this supple, flexible interface between Co., New York (1997); Riordan had conceived similar ideas several It took Last’s development team nearly (2005); and Ross Knox Bassett, To of Noyce’s famous integrated-circuit silicon and its environment, the and Hoddeson, Physics Today, p. 43 months earlier at Texas Instruments, two years to produce commercially the Digital Age: Research Labs, Start- patent, titled “Semiconductor Device- microchip as we know it today would (December 1997). See also William and that the two men are generally available integrated circuits using Up Companies, and the Rise of MOS and-Lead Structure,” which he applied be unthinkable. It is the sine qua non of Shockley, “Memorandums,” perceived as the co-inventors planar processing. Higher precision Technology, pp. 45-53, The Johns for later that year (Fig. 6).29 the semiconductor industry, and what (personal diary 1955–56), Stanford of the integrated circuit. Kilby, was needed for oxide masking and Hopkins University Press, Baltimore, The difficult task of actually most distinguishes silicon from all other University Archives, Shockley however, did not have any idea of photolithography—and more steps were MD (2002). fabricating microelectronic circuits material alternatives. papers, accession listing 95-153, the planar technology used in the involved. And a big problem was isolating 15. C. Lécuyer, Making Silicon Valley: according to these prescriptions fell to Although ultimately unsuccessful box 2B. Noyce patent, and proposed that the circuit elements. Workers even Innovation and the Growth of High Jay Last, who formed a development financially, the Shockley Semicon- 3. Riordan and Hoddeson, Crystal individual components on a chip be attempted a physical-isolation approach Tech, 1930-1970, p. 150, The MIT group to realize this goal. In March 1959 ductor Laboratory played an important Fire, pp. 231-234, W. W. Norton connected by what became known suggested by Last that involved etching Press, Cambridge, MA (2006). Noyce took over as Fairchild’s General technology-transfer role in this saga. & Co., New York (1997); Riordan as “flying wires.” See Jack Kilby, channels between the components from 16. C. Lécuyer, Making Silicon Valley: Manager and Moore replaced him as It acted as the crucial incubator where and Hoddeson, Physics Today, p. 44 IEEE Transactions on Electron Devices, the back of the wafer all the way to the Innovation and the Growth of High research director. Noyce rarely showed men came together with ideas on the (December 1997). 23, 648 (July 1976); and Michael S. fragile oxide layer, then filling these Tech, 1930-1970, pp. 140-148, The up in the laboratory after that—and West Coast, and began working together 4. Riordan and Hoddeson, Crystal Fire, Wolff, “The Genesis of the Integrated on the revolutionary implications of the pp. 236-240, W. W. Norton & Co., MIT Press, Cambridge, MA (2006). Circuit,” IEEE Spectrum (August new silicon technologies pioneered in New York (1997). 17. M. Tanenbaum and D. Thomas, 1976), pp. 38–45. the mid-1950s by Bell Labs. They were 5. F. J. Biondi, et al., Editors, Transistor “Diffused Emitter and Base Silicon 30. Jay Last, private communications. so revolutionary, in fact, that it took Technology, 3 Vols., D. Van Nostrand, Transistors,” Bell System Technical Lécuyer observes, “Interestingly, a small rebellion, much celebrated in New York (1958); Vol. 3 is devoted Journal, Vol. 35 (January 1956), pp. Noyce played no role in this effort.” Silicon Valley lore, to realize them. to diffusion technologies, including 1–22. See C. Lécuyer, Making Silicon Valley: The development and implementation contributions by Frosch, Fuller, 18. C. Lécuyer, Making Silicon Valley: Innovation and the Growth of High of planar processing at Fairchild and Tanenbaum. Noyce attended Innovation and the Growth of High Tech, 1930-1970, p. 157, The MIT Semiconductor Corporation was what the third symposium as a Philco Tech, 1930-1970, p. 149, The MIT Press, Cambridge, MA (2006). really threw the floodgates open to employee. Press, Cambridge, MA (2006). 31. M. M. Atalla, et al., “Stabilization of extremely reliable silicon transistors 6. Riordan and Hoddeson, Crystal Fire, 19. C. Lécuyer, Making Silicon Valley: Silicon Surfaces by Thermally Grown and microchips of endlessly growing pp. 241-243, W. W. Norton & Co., Innovation and the Growth of High Oxides,” Bell System Technical Journal, complexity. New York (1997). Tech, 1930-1970, pp. 149-150, The Vol. 38 (1959), pp. 749–783. The junction transistor’s inventor, 7. Riordan and Hoddeson, Physics MIT Press, Cambridge, MA (2006). 32. On the origins of the MOS transistor, William Shockley, dimly perceived Today, p. 45 (December 1997). 20. C. Lécuyer, Making Silicon Valley: see Ross Knox Bassett, To the Digital this bright future, but he lacked the 8. C. J. Frosch and L. Derick, Journal of Innovation and the Growth of High Tech, Age: Research Labs, Start-Up Companies, management skills to bring it off under The Electrochemical Society, 104, 547 1930-1970, pp. 150-152, The MIT and the Rise of MOS Technology, pp. his direction. In the early 1960s, he (1957). Press, Cambridge, MA (2006); also 22–33, The Johns Hopkins University began teaching at Stanford University, 9. Gordon Moore, interview with Jay Last, private communication. Press, Baltimore, MD (2002); and first as a lecturer and then as a L. Hoddeson and M. Riordan, 11 21. Jean A. Hoerni, Fairchild Semi- Dawon Kahng, IEEE Transactions on professor of engineering and applied January 1996. conductor Corp. lab notebook, pp. 3- Electron Devices, 23, 655 (July 1976). science, while his company slowly 10. Soon after Bell Labs research 4 (1 December 1957), copy provided 33. C. Lécuyer, Making Silicon Valley: failed and was eventually dissolved. director Mervin Kelly hired courtesy of Jay Last. Innovation and the Growth of High Yet without Shockley’s contributions Shockley in 1936, he impressed on 22. Jean A. Hoerni, Fairchild Semi- Tech, 1930-1970, pp. 155-59, The MIT in bringing together this stellar group his new employee the needs of the conductor Corp. lab notebook, p. 3 Press, Cambridge, MA (2006); and of researchers and introducing them to Bell System to replace its unreliable (1 December 1957), copy provided Jay Last, private communication. silicon technology, there would be no electromechanical switches with courtesy of Jay Last. Silicon Valley—at least not in Northern a better, faster solid-state version. 23. C. Lécuyer, Making Silicon Valley: California. Kelly’s urgings helped stimulate Innovation and the Growth of High Shockley to invent the junction Tech, 1930-1970, p. 150, The MIT transistor and probably promoted Press, Cambridge, MA (2006). About the Author his interest in the four-layer diode. 24. Jean A. Hoerni, “Planar Silicon Fig. 7. One of the earliest prototype integrated circuits fabricated by Jay See Riordan and Hoddeson, Crystal Transistors and Diode,” Fairchild Last’s development group at Fairchild using the planar processing technique. Michael Riordan earned his PhD in Fire, pp. 81-82, W. W. Norton & Co., Semiconductor Report No. TP-14, This “flip-flop” logic circuit employed four transistors. (Courtesy of Fairchild physics from MIT. He is an adjunct New York (1997). p. 9, Fairchild Collection, Stanford Semiconductor Corp.) professor of physics at the University of 11. On the difficulties of fabricating University Archives. California, Santa Cruz, and a Lecturer uniform, reliable four-layer diodes, 25. Jean A. Hoerni, Fairchild patent see W. J. Pietenpol, “Transistor disclosure, “Method of Protecting Designs: The First Decade,” Bell 40 The Electrochemical Society Interface • Fall 2007 The Electrochemical Society Interface • Fall 2007 41