Chemical Heritage Foundation: Chemical Heritage Magazine

A Most Successful Failure

By David C. Brock

In the annals of the history of science, technology, and business, stories of successes—of those events, individuals, and groups that came to shape history in both intended and unforeseen ways—abound. Stories of failure are less frequently told, but the past is replete with examples of “successful failures”: developments that, despite falling short of their immediate goal, had profound effects.

By most benchmarks the Shockley Semiconductor Laboratory was a business failure. William Shockley (1910–1989), a pioneer of solid-state physics and corecipient of the 1956 Nobel Prize in physics for the invention of the transistor, established the firm in 1955 as a subsidiary of Beckman Instruments after convincing Arnold O. Beckman that a substantial technological and economic opportunity lay in silicon semiconductor electronics. The Palo Alto–based firm met the same fate as many high-technology start-ups that followed: early years of R&D struggle, limited commercial success, and gradual disappearance as a recognizable entity after a series of acquisitions. In the late 1950s the company introduced a line of four-layer silicon diodes to the market; in 1961 Beckman sold the operation to the Clevite Corporation; in 1965 Clevite sold its semiconductor operations to ITT; and by the end of the 1960s the Shockley Semiconductor facility was shuttered.

Although Shockley Semiconductor failed to capitalize on the opportunities in silicon electronics foreseen by Shockley and Beckman, its story is crucial for the development of the semiconductor industry. Shockley assembled a remarkable collection of talented individuals, many of whom subsequently became industrial and technological leaders in the region that would become known as “Silicon Valley.” In 1957 eight of Shockley’s employees left to establish Fairchild Semiconductor, a firm that became the mother organization for hundreds of semiconductor manufacturers and their suppliers. The cofounders’ time at Shockley Semiconductor served as a crash course in entrepreneurship from which their subsequent success at Fairchild was forged.

R. Victor Jones was among the first wave of Shockley’s recruits. Jones’s addition to the laboratory in 1956 was, in Shockley’s estimation, something of a coup. Jones was finishing his doctorate in physics at the University of California, Berkeley, and had accepted a job offer from Bell Telephone Laboratories—the nation’s premier industrial research laboratory. Shockley eventually convinced Jones to refuse the Bell Labs position and to join the handful of Shockley Semiconductor staff occupying a rented storefront in Palo Alto. Shockley was convinced that the future of semiconductor devices lay in two key technological approaches: silicon and diffusion. Scarcely a year earlier the Bell Labs chemist Morris Tanenbaum had used these approaches to create the first diffused-base silicon transistor (see CH, Winter 2004/5, p. 24). (At the time most transistors were made of germanium.) Shockley’s vision was predicated on access to an adequate supply of single crystals of silicon with exacting degrees of chemical purity and crystalline perfection, manufactured using a process pioneered at Bell Labs and later Texas Instruments by the chemist Gordon Teal (see CH, Spring 2006, pp. 33–35).

Not long after joining Shockley Semiconductor Jones accompanied Shockley on a cross-country flight. On the long flight to the east coast Shockley and Jones discussed how the firm could grow purer, more perfect silicon crystals. As the landscape passed beneath them, Jones and Shockley sketched out an answer. In Teal’s conventional crystal-production technique, single crystals were pulled from a melt of silicon held by a graphite crucible heated by an induction coil. The combination of high heat, silicon’s reactivity, and direct contact with graphite made this technique susceptible to contamination. Jones and Shockley envisioned pulling a crystal from a pool of melted silicon atop a large block of ultrapure silicon. The underlying silicon block would rest on a graphite support, but the arrangement would prevent contaminants from the graphite from easily migrating to the surface puddle. The scheme, however, required drastic changes from the conventional approach. Jones and Shockley would need two independent heating systems: an “oven” of sorts to bring the entire silicon block to just below its melting point and an additional “surface heater” to form the puddle on the top. Resistance heating—not induction heating—would be required. They envisioned an electro-optical feedback system to control the temperature of the surface puddle of silicon. Jones and Shockley contemplated yet another leap from existing practice to prevent contamination: instead of growing their crystals in an atmosphere of inert gas, they would grow them in a vacuum.

Upon their return to California Jones made a partial proof-of-concept demonstration for their new crystal-growing approach. He fashioned a benchtop model in which he formed a pool of Wood’s metal (an alloy of bismuth, lead, tin, and cadmium) atop a block of the material, using an oven arrangement to heat the metal block and a surface heater to form the pool. From this pool Jones drew a rough crystal. This demonstration was enough to convince Shockley to launch a major effort to build a full-scale, production-capable silicon crystal puller that embodied all of these innovative ideas. No additional developmental experiments would be required. Shockley believed that they had devised the principles behind what could become the best—and most complex—crystal grower ever.

Nevertheless, Shockley hedged his bets on the advanced crystal puller by directing some of the firm’s top engineering talent to work simultaneously on conventional crystal pullers. Shockley had recruited the engineers Dean Knapic, Eugene Kleiner, and Julius Blank from a major Western Electric plant in New Jersey. Their primary responsibility was working with Jones on the advanced crystal puller, but they were also tasked with building conventional pullers. Even so, by 1956 the effort to build Jones and Shockley’s crystal puller consumed a great proportion of the firm’s financial resources and staff time. Although the prototype for the advanced crystal grower soon rose to the roof and took up a large space at the back of the production facility, Jones and the engineers worked in relative isolation. No members of the staff beyond the Jones-Shockley-engineers circle were involved in or even briefed on the project. Members of the technical staff reported directly to Shockley, who in turn set assignments and goals and coordinated staff efforts. One striking result of this “hub and spokes” managerial approach was that MIT-trained metallurgist C. Sheldon Roberts, the single member of the staff who had experience with crystals and crystal growing, was not involved in the advanced crystal- puller project.

In early 1957 the Jones-Shockley crystal grower was ready for a preliminary test. The result was a spectacular and distressing failure: the sapphire rods used to support the oven’s molybdenum resistance windings weakened during heating, causing an electrical short. A tremendous bang thundered through the laboratory, and the circuit breaker panel all but flew off the wall. The puller sustained considerable damage and Shockley’s enthusiasm for the project evaporated just as quickly as it had earlier condensed. He ordered the machine to be repaired and redesigned, but he also shifted emphasis to the staff’s efforts to build conventional crystal pullers.

Within a few months Knapic, Blank, and Kleiner had succeeded in building conventional crystal pullers and had produced single-crystal silicon for the firm. Prompted by Robert Noyce, an MIT-trained physicist, Blank even began working on a resistance heating system for a conventional crystal puller—an incremental improvement on the conventional approach. But before Blank could complete this developmental work he joined a group of his coworkers (Vic Grinich, Jean Hoerni, Kleiner, Jay Last, Gordon Moore, Noyce, and Roberts) who resigned en masse to form Fairchild Semiconductor.

The sources of the Fairchild Semiconductor cofounders’ dissatisfaction with Shockley Semiconductor were varied but tended to center on organizational focus and management practices. The advanced crystal-puller project epitomized many of their concerns. It belied an emphasis on “blue sky” developments—pushing out the research frontier—rather than on moving quickly and directly to manufacturing devices. Sometimes Shockley seemed focused on the production of research and academic presentations; other times he strongly expressed the view that getting products to market was paramount. With these shifting expectations the identity of the firm was rendered uncertain, leading, perhaps inevitably, to frustration.

The departing group turned their hard-earned lessons into prompt success at Fairchild. Roberts, the chemist responsible for providing the fledgling company with silicon crystals, thought that the advantages of resistance-heated pullers lay in their low cost and reliability. As a result Fairchild’s crystal pullers were less showcases of engineering finesse than production tools that reflected the new company’s unwavering focus on manufacturing—a reaction to the founders’ experience at Shockley. Blank, in his words, “overdesigned” the crystal pullers, provisioning them with seemingly overpowered motors and overrated electrical components, ensuring that the machines ran continuously, required minimal maintenance, and withstood whatever abuse might befall them at the hands of their operators.

Abundant single crystals of silicon enabled Fairchild Semiconductor to bring the first diffused silicon transistor to the open market within a year of its formation. By January 1959 the company had achieved two central innovations in the history of semiconductor electronics: Hoerni’s creation of the planar process and Noyce’s conception of the planar integrated circuit. (For more on the integrated circuit at Fairchild see CH, Spring 2005, pp. 31–32.) The crux of these inventions—the foundation of the microelectronics revolution of the past 40 years—is the controlled use of oxide layers that form on the surface of wafers cut from silicon single crystals. The invention and manufacture of the planar integrated circuit at Fairchild was only possible with an abundance of adequate-quality silicon single crystals, and this was only possible given the most “successful failure” of Shockley Semiconductor. As Gordon Moore has quipped, “Shockley brought the silicon to Silicon Valley.”

This article is based in part on oral history interviews conducted with R. Victor Jones, Julius Blank, Gordon Moore, and Jay Last as part of CHF's Chemical History of Electronics program.

For Further Reading

Brock, David, ed. Understanding Moore’s Law: Four Decades of Innovation. Philadelphia: Chemical Heritage Foundation, 2006.

Lécuyer, Christophe. Making Silicon Valley: Innovation and the Growth of High Tech, 1930–1970. Cambridge, MA: MIT Press, 2005.

Shockley, William. “Transistor Electronics: Imperfections, Unipolar and Analog Transistors,” Proceedings of the IRE 40:11 (1952), 1289–1313.

Thackray, Arnold; Minor Myers, Jr. Arnold O. Beckman: One Hundred Years of Excellence. Philadelphia: Chemical Heritage Foundation, 2000.

 

David C. Brock is a senior research fellow with CHF’s Center for Contemporary History and Policy and the editor of Understanding Moore’s Law: Four Decades of Innovation (CHF, 2006).


Lawrence M. Principe is the Drew Professor of the Humanities in the Department of the History of Science and Technology and the Department of Chemistry at Johns Hopkins University. His many publications include Alchemy Tried in the Fire: Starkey, Boyle, and the Fate of Helmontian Chymistry, with William R. Newman.

Marjorie Gapp is curator of art and images at CHF.

©2007 Chemical Heritage Foundation