The Origins of Diffused-Silicon Technology at Bell Labs, 1954-55*
Total Page:16
File Type:pdf, Size:1020Kb
Discovery of the Silicon The Origins of Diffused-Silicon Technology Dioxide Layer at Bell Labs, 1954-55* Many of the diffused-silicon wafers by Nick Holonyak, Jr. for our p-n-p transistors (and p-n-p- n switches) were prepared by Frosch ilicon-based transistor and germanium, particularly for switching At the time we began to construct using a phosphorus “pre-deposition” integrated circuit technology has functions. The energy band gap of diffused-impurity silicon switching diffusion followed by a higher- grown so large and become so germanium was too small (0.65 eV) devices, diffused-silicon solar cells temperature “drive-in” diffusion. S 4 When we asked him to reduce the important that it is difficult to recognize compared to silicon (1.11 eV), and thus had been demonstrated, and Prince how it all began. In spite of the obvious yielded an off-state switch—a reverse- and colleagues were then developing pre-deposition time and temperature differences between germanium and biased junction—that was too leaky. diffused-silicon rectifiers.5 In addition, and to increase the drive-in time and silicon, it was not evident during the We knew, as Moll had argued, that Lee and co-workers were attempting temperature, we either lost the n-type early 1950s in what form and substance we had to work with silicon to build to make a germanium transistor with layer completely, or the wafer was the transistor would prevail. What was switching transistors, especially the a diffused base and an evaporated pitted, eroded or destroyed during the need for silicon—at the time such an p-n-p-n switch. This four-layer diode (alloyed) emitter.6 Thus there was reason the dry-gas process. This was a serious intractable, peculiar new technology? was of substantial interest then because to think that diffused-silicon devices problem in late 1954 and early 1955, but In 1954-55, however, the requirement it appeared capable of competing with could indeed be made, particularly if Frosch solved it by a major discovery for low-leakage switching devices at Bell a gas-tube cross-point switch that had impurity diffusion could be developed he made a few months later. Telephone Laboratories led directly to the attracted interest within Bell Labs.2,3 further and sophisticated contacting In the meantime, we were able exploration of impurity-diffusion and At first I proposed to build a germani- schemes be devised. In the latter to devise a practical solution to our problem of constructing a diffused- aluminum-metallization technology to um p-n-p-n switch, but Moll objected case, metal evaporation appeared most make silicon transistors and p-n-p-n on the grounds of the leakiness of this desirable because either a contact or a base alloyed-emitter silicon p-n- switches. This technology, a more or element due to its small energy gap. shallow p-n junction (using thin-layer p device using a trick to cope with FIG. 2. Diffused-base p-n-p silicon mesa transistor made in early 1955 using evaporated less ideal thin-layer technology that can An alternative was to fabricate quickly precision alloying) could be realized. the high phosphorus concentration gold-plus-antimony ring base contact and aluminum-evaporated and alloyed emitter. that Frosch was introducing. But our be referenced from a single surface, led a silicon p-n-p-n switch by introducing Most important, metal evaporation to the 1955 discovery of the protective a point contact on the collector body, could be done in a uniform patterned solution to the problems of diffusion Using the evaporator in Moll’s group, spots on the wafer to form (after etching) silicon-dioxide layer, oxide masking and near the collector junction, of a grown- format. in silicon as well as Tanenbaum’s Goldey began trying to determine separate p-n-p mesa transistors. transistors10 fabricated from patterning, and ultimately to the silicon junction n-p-n transistor. This approach The kind of device technology n-p-n how to evaporate gold-antimony Made before Christmas of 1954, our integrated circuit. would work in principle, Moll agreed, outlined here, which concerned us in Fuller’s double-diffused wafers did not on silicon in an attempt to realize first diffused-base silicon transistors In this article, I recount some of the but he noted the inherent problem 1954, has the important advantage, matter much in the long run, except to n-type contacts or shallow (n on p) n-p had ratios of collector current to exploratory diffused-impurity silicon of poor reproducibility because of particularly for low-power devices, that demonstrate the feasibility of diffused- junctions. In the research department, emitter current of only 0.1–0.2. Small device development at Bell Labs that the point contact. He knew we had all of the vital junction construction silicon devices. A far more important Morris Tanenbaum was attempting to improvements that came soon thereafter helped to establish the fundamental to employ silicon and, if necessary, could be referenced and realized development was about to emerge evaporate and alloy aluminum on “hot” were needed to make these transistors basis for microelectronics. It draws upon generate all the requisite diffusion and (precisely!) from one side of the crystal that would fundamentally change (over 660˚C) silicon, but he had failed to fully functional (See Fig. 2). Double- my direct experiences in the group in metallization technology to realize wafer.7 This indicated that precise semiconductor technology, making realize continuous metallization across diffused silicon n-p-n transistors, all existing diffusion procedures the Device Development Department sophisticated n-p-n, p-n-p and p-n-p-n dimensions could be achieved, which the silicon surface. After Moll obtained fabricated by Tanenbaum and D. E. supervised by John L. Moll, which was device structures. Small silicon crystals is not possible in the case of an alloy obsolete. Tanenbaum’s permission for us to use Thomas using wafers that Fuller had concerned with switching devices. Silicon were then available, typically almost transistor with junctions constructed, Frosch, the man chiefly responsible his evaporator, we quickly solved the prepared by simultaneously diffusing diffused-impurity devices made there in two inches in diameter, but a process more or less imprecisely, on both sides for this advance, was a consummate problem of evaporating aluminum on aluminum and antimony into them, 1954–55 are described, including work technology yielding the switching of a relatively thick crystalline wafer. process chemist familiar with silicon, either hot or cold. We also achieved higher ratios earlier,10 but they and data not broadly known. Much of devices we wanted did not yet exist. many types of procedures and had this technology was soon carried across made precision, shallow alloyed p- required that the p-type base contact be been working for several years on the country to what became known as type contacts or shallow (p on n) p-n alloyed through the n-type emitter, thus impurity diffusion in silicon with his Silicon Valley. junctions. By late 1954 Goldey and resulting in some compromise. If the technician Lincoln Derick (See Fig. 3). I had solved the problem of making evaporated aluminum contact on the In spite of his considerable experience metal contacts on silicon and forming double-diffused n-p-n device was not with dry-gas diffusion procedures, Silicon Technology and uniform, shallow p-n (or n-p) junctions alloyed deep enough and thus formed however, Frosch regularly reduced Switching Devices or shallow ohmic contacts. I wrote a Bell a shallow p-n junction on the top n- many of our silicon wafers to cinders, Labs memorandum with Tanenbaum type layer, an accidental p-n-p-n switch particularly at high temperatures near In the early 1950s, laboratories were on this aluminum metallization and resulted. or above 1100°C. hiring new PhD scientists and engineers shallow junction formation on silicon;8 to work on the transistor and, finding Goldey included this material and many of the new recruits lacking in further results in a later report.9 semiconductor background, began The prototype devices we made in late short courses and training programs 1954 and early 1955 included a diffused- in semiconductor studies. In the base, alloyed-emitter transistor. In fall of 1954, James M. Goldey and I March 1955, these achievements—along joined Moll’s switching-device group with those of our colleagues Calvin and, because of prior semiconductor Fuller, Carl Frosch and Tanenbaum device experience, were ready to start —helped convince transistor- building and learning about silicon development head Jack Morton (after switching devices. Moll, of course, had strenuous argument) to redirect all work already written (with J. J. Ebers) the towards diffused-base silicon transistor Ebers-Moll equations describing the technology. Figure 1 is a photograph 1 of our rudimentary silicon p-n-p switching operation of transistors and FIG. 1. Aluminum evaporated and alloyed on a p-type silicon wafer that had first been had given much thought to why silicon diffused and made n-type with phosphorus. The etched mesas are low-gain p-n-p diffused-base alloyed-emitter transistor; devices must be developed to replace transistors fabricated in late 1954. the emitter is aluminum evaporated and shallow-alloyed into the silicon. In creating this device, phosphorus *Adapted from N. Holonyak, Jr., “Diffused Silicon Transistors and Switches (1954–55): The Beginning of Integrated Circuit Technology,” was first diffused into a p-type silicon in David G. Seiler et al., Editors, Characterization and Metrology for ULSI Technology: 2003 International Conference on Characterization and Metrology for ULSI, AIP Conference Proceedings Vol.