Semiconductor Research and Development at General Electric

Copyright Mark P D Burgess 2008 and 2011

Innovation at General Electric in the field of semiconductors in the 1950s occurred at two main sites: Syracuse and Schenectady. It centres on the pioneering work of Harper North, Robert Hall and William Dunlap at Schenectady and that of John Saby at Syracuse in the period 1948-51. By the time the silicon controlled rectifier, the Diac and Triac were developed in the period 1957-1963 General Electric had assembled a significant team of researchers and engineers at Syracuse, Schenectady and the Clyde rectifier plant.

The Schenectady laboratories date back to 1900 when Willis Whitney, a chemist from MIT, was hired to be its foundation director. Its alumni include some famous names in science and technology such as William Coolidge who worked on medical X-ray technologies, the Nobel prize winning chemist Irving Langmuir and Albert Hull who developed the magnetron and other vacuum tubes. In 1950 the laboratory moved to the Schenectady suburb of Niskayuna.

At Syracuse, Electronics Park was the headquarters for electronics research, development and was General Electric’s main electronics manufacturing site (radio, radar, television and similar equipment). Construction began in 1946 on a park like site of 150 acres and phase one was completed early in 1948. The Electronics Laboratory moved there in February 1948.

The Schenectady laboratories were intended to investigate more basic research whereas the Syracuse laboratories were there to support production. But researchers at both sites all had considerable freedom to investigate new technologies and undertook novel developments. The culture of innovation was also strong at Clyde which was a manufacturing site.

Wartime Semiconductor Research

The invention of the cavity magnetron in Britain and the potential for high resolution radar led to a large scale development programme managed by the Radiation Laboratory at MIT. This included the development of new crystal rectifiers: a redeployment of very old technology now revisited because vacuum tubes could not operate at the GHz frequencies being used. But to meet military standards the humble cat’s whisker crystal detector needed to become a reliable detector.

General Electric was a contractor in the programme and represented by Harper North in the more academic “crystal meetings” which were dominated by researchers from the Radiation Laboratory and the University of Pennsylvania. [Henriksen 1987]

The overall purpose of the General Electric contract was to investigate germanium as a possible improvement over silicon which was the dominant crystal in use. [North 1946]

At General Electric North made germanium from its oxide under hydrogen at 650 C and then fused it into a crystalline mass by heating it at 1015 C. The purity of the resulting germanium (as indicated by its resistivity) was crucially dependent on the quality of the oxide, processing conditions and control over the purity of the hydrogen.

Extensive work was undertaken on germanium dopants, antimony being preferred over phosphorous which was also investigated.

To produce detector cartridges a doped ingot was sliced into 0.02 inch wafers, polished and rhodium plated on one side diced into 0.07 inch squares. These were soldered to the cartridge stud and finally highly polished. Heat treatment and chemical etching were found to be undesirable for radar detectors.

North undertook an extensive investigation of welded point-contact diodes where the contact was welded to the germanium by passing a high current briefly through the unit. Due to their negative resistance characteristics these had high gain but were not used because of their high noise. Unlike conventional point-contact diodes, they were, however, excellent rectifiers. Using germanium with high levels of antimony doping, North obtained very low forward resistance at only 0.4 volts. [Torrey 1948]

North’s coaxial cartridge design followed the dimensional outline of the Sylvania silicon 1N26 but typically utilized antimony doped germanium. The assembly was sealed into silver plated steel tube with low-loss glass. Use of antimony doping (n-type germanium) gave reversed polarity compared to conventional silicon diodes. [Drawing from Torrey 1948]

General Electric 1N23A cartridge [Courtesy Jan de Groot]

First Diodes

The first diodes released by General Electric were based on the North wartime research and the subsequent development of Harper North and his team.

The first diodes released commercially by General Electric were the G5 series submitted for RMA registration in June 1948. They were general purpose small signal types and used North’s welded germanium technology. These were the G5A, G5B, G5C and G5D which corresponded to the series 1N49-1N51.

These were followed by the G5E (1N63) that December, G5G (1N65) and G5F (1N64) in 1950 [RMA Release 670]

In 1950 the package outlined changed to the newer format illustrated below. In addition the G5 series was extended to include hermetically sealed ceramic cases as the G5K, G5L and G5P (1N69, 1N70 and 1N81) from 1950. [RMA release 891B, 1951, RMA release 1016 1951]

The G range of welded germanium diodes was extended with the G6 and G7 series. The G6 diode was intended as a VHF detector and meter rectifier and the G7 series for UHF applications. A summary of the range is shown in the following data sheets (click on these thumbnails for full resolution). For the story of their manufacturing process see "Welded Germanium Diodes."

The G8 matched pairs and G9 bridge rectifiers completed the commercial G series. By now alloy junction rectifiers known then as “diffused junction rectifiers” were in development. The G10 rectifier made a brief appearance as noted below.

Germanium Whisker Transistors

Robert Hall returned to General Electric Research Laboratories Schenectady just as Bell announced its point-contact transistor. Hall had just graduated PhD from Caltech and had previously worked at the laboratories on a range of projects including the Harper North Wartime radar diode project. Dr Albert Hull was Assistant Director of the Laboratory. He was known for his collegial management style with a relaxed approach to the research agenda of his staff members. Hall recalls that Hull came in with the reprint of the Physical Review letters that announced the transistor [Bardeen 1948] and said “Robert here’s an interesting development from Bell Laboratories. It looks like something pretty new and exciting. Would you like to look into it and see if there’s anything interesting there.” [Choi 2004]

General Electric had all the knowhow it needed to duplicate the Bell design quickly. North’s diode programme had diodes that could handle 100 volts back voltage which was state of the art at that time. (Purdue University produced the high back voltage germanium for the Bell point contact transistor.) North’s diode programme and related research ensured it had a manufacturing understanding of the key technologies:

High back voltage poly-crystalline germanium;

Doping;

Point contact design;

Welding or forming; and

Assembly and encapsulation.

Its first designs were crude with two pins for the collector and emitter with the base connection through the case in the manner of the Bell Type A. No socket was available and users were recommended to use a 5 pin subminiature tube socket using positions 2 and 3 for the collector and emitter and to create a base connection by inserting phosphor-bronze strips in positions 1 and 4 and bending them so they contacted the transistor case.

The transistors were known as germanium triodes or germanium whisker transistors. Two types were produced evolving from prototype coding through the familiar “G” designation in use for its point-contact diodes and finally adopting RMA registration:

The prototype numbers appear in early data sheets and in a General Electric price list dated June 1^st 1951 in which the new SX-4A and Z2 transistors were priced at a massive $29 each.

The two transistors had the same mechanical and electrical characteristics but the switching transistor was tested for “trigger action” or negative resistance. [General Electric 1950 courtesy Jack Ward]

From 1952 the transistors had 3 pins with the base connection being soldered to the outside of the transistor’s case.

See data sheet ECG-3B.

Early General Electric G11 and G11A point-contact transistors.[Courtesy Jack Ward]

Production General Electric G11 point-contact transistor [Courtesy Jan de Groot]

Hall PN Junctions and Diffused Rectifier Research

Hall had the opportunity to take a more fundamental approach to support semiconductor development.

Pilot scale quantities of the G-10 Rectifier were offered in Tele-Tech January 1952

Full Data Here

By November 1952 the "New Style" Rectifier was Advertised in Tele-Tech

Recombination Theory

Hall carried out extensive characterization of the rectifiers he made from high purity germanium wafers doped on opposite faces with acceptor and donor impurities. These were known as PIN types implying intrinsic semiconductor at the barrier. Current flow in the high purity region is determined by the generation and recombination of holes and electrons and these structures enabled the study of these phenomena. Recombination was expected to be proportional to the square of the carrier concentration at high levels of injection but Hall found a linear dependence. His observations could be accounted for by assuming that recombination takes place largely through recombination centers distributed through the germanium [Hall 1952].

“So I was pretty well convinced that I had a new model of recombination which I discussed at a meeting of the Physical Society. Bill Shockley was there. After the meeting he quizzed me at length about this, and apparently was convinced the mechanism was quite important.“ [Choi 2004]

In 1952 Shockely and Read carried out an extensive analysis based on Hall’s hypothesis. They named the recombination sites “traps” that sat at energy levels mid-way between the conduction and valance bands. They agreed with Hall: “It should be noted that at high carrier densities the rate of recombination through traps is linear in the carrier density whereas any direct recombination would be quadratic. This is in agreement with the findings of R N Hall for P-N junctions operating with high injected densities in the region of recombination.” [Shockley 1952]

This form of recombination is now known as Shockley Read Hall recombination. The ideal trap energy level is half way between the valance and conduction bands. Hall found a value of 0.22 ev for the position of the traps above the valence band or below the conduction band. [Hall 1952]

Reflecting on Hall’s achievements (including his later development of the first semiconductor laser) Holonyak muses: “If I had been giving Nobel Prizes for semiconductors, after Bardeen, Brattain, and Shockley, the next guy who would have gotten the Nobel Prize would have been Bob Hall, my colleague, later, in Schenectady.” [Ashrafi 2005]

Rectifier Production at Syracuse

In 1949 Hall’s work on the rectifier was transferred to the Electronics Laboratory at Syracuse for the development of production processes.

General Electric had a diode programme since 1942 when Harper North began work on an improved radar diode. Both silicon and germanium had been considered but germanium proved best. This work was not unique to General Electric: Sylvania, Western Electric and Raytheon had also developed radar diodes. Following the war General Electric had a strong diode business.

Paul Jordan hired Addison Sheckler in March 1948 to work on nuclear instruments at Syracuse. In late 1948 production asked for help on purifying the germanium they were using. At the time Eagle Picher Co was their sole supplier and quality was variable. A chemist at Syracuse, Fred Pingert suggested this could be done by the Bridgeman method (progressive feezing). Sheckler developed the method using high purity General Electric carbon boats. Crystallisation gave excellent purity gains at a yield of 80% (20% had the impurities).

Sheckler writes “About the end of 1949 R.N. Hall and W. Crawford Dunlap of GE’s Research Laboratory came up with the alloy/diffused junction using antimony and indium for contacts. (The Hall-Dunlap patent). We at Electronics Laboratory picked this up immediately and started to make diodes. It very quickly became apparent that we were going to need single crystal germanium for a successful product.” [Sheckler 2004].

Thus the Laboratory assigned Sheckler to build a crystal pulling apparatus which Sheckler says he had operational in only five weeks. He notes that subsequently he discovered that Hall already had single crystal pulling working at Schenectady. Apparently the relations between the two laboratories were somewhat competitive.

Sheckler recalls that about this time the “Laboratory Management” told Jordan to get his team back to some “real work” or be fired. This meant tube electronics. Jordan went to the head of the Electronics Division, Dr W R G Baker and pitched the future of semiconductors and Baker bought it with sanction for $10,000 in continuing development. The $10,000 became $30,000 and secured the place of General Electric in semiconductors.

In 1950 everything to do with the new rectifier needed to be developed. There were no useful pre-existing processes. The interim can consisted of two copper cups one fitting inside the other insulated by a paper washer, sealed with an O-ring and crimped. 50,000 diodes m

Junction Transistor Development

Work on junction transistors was carried out by Hall at Schenectady (new forms of grown junction transistors) and Saby at Syracuse (the alloy junction transistor). Later Hall worked on power alloy junction transistors.

Commercially the alloy junction transistor became the workhorse device through the 1950s and were made by many semiconductor companies.

These developments reflect the cultures of the two sites. Hall’s grown junction work depended on an exquisite understanding of the kinetics of segregation of impurities between the semiconductor melt and the solid phase. His rate growing and meltback junction transistors could not be “discovered;” they were based on the segregation studies.

Saby worked in a production laboratory and adopted alloy diffusion processes that had been invented by Hall and Dunlap and developed for production of rectifiers at Syracuse and doped single crystal germanium wafers also used in rectifiers. This was a pragmatic approach which led him to useable transistors quite quickly.

Grown Junction Transistor Development

Grown junction transistors are discussed in the companion article on this site Early Transistor Technologies.

Hall used the Czochralski method to grow single germanium crystals and used this to make grown junctions using the double doping method pioneered by Gordon Teal and Morgan Sparks at Bell [Shockley 1976]. In this method an NPN grown junction, for example, is made by commencing with an N-type semiconductor from which a single crystal is started. Then a P-type impurity is added to convert the melt to P-type and produce a P region in the crystal. Finally the melt is restored to N-type by adding a suitable quantity of N-type impurities.

As part of his studies on the purification of germanium, Hall measured the segregation constants of a range of impurities. P-type impurities have high segregation constants. Hall had discovered this early on when he tried to make N-type germanium by doping with arsenic: his germanium was contaminated with boron and this co-crystallized first creating P-type germanium. The arsenic followed creating N-type germanium forming Hall’s “barrier-less” rectifier. Other P-type impurities such as gallium and indium behave similarly and their segregation constants are relatively independent of crystallization rates. Incorporation of antimony, however, is strongly dependent on crystallization rate: more is incorporated at high growth rates. Thus Hall found he could pull a single crystal from a melt containing antimony and indium that was P-type when grown slowly and N-type when grown rapidly. By repeated cycling of the growth rates he could produce “more than a hundred uniformly spaced PN junctions.” [Hall 1952]

Double doping transistors had poor high frequency performance due to their thick base layer. The rate grown process gave greatly improved high frequency performance.

Rate-grown transistors were manufactured by General Electric at Syracuse using this process. Early examples were the 2N78 and 2N167. [Gottlieb 1959]

In September 1954, General Electric announced a new production facility for its rate grown transistors would soon be commissioned noting that “at present, GE is making a few hundred transistors a week on its pilot line in Syracuse. A rate grown ingot is being produced every ten days, which is currently enough to keep the line supplied for about ten days. The firm is presently mechanizing the fabrication process of attaching the leads to the transistor bars.” See Making a 2N78 Transistor

“The rate grown process, first announced a year ago, involves varying the heat and withdrawal rate controls cyclically during the crystal-growing process. As a result, as many as a hundred usable wafer-thin slices of germanium can be cut from a six-inch ingot. With other processes only one useful slice is formed per crystal. The disc shaped slices are then diced into thousandths of an inch long, with a junction layer through the center of each. The more cycles there are in the crystal-growing process, the more layers there are for slicing and dicing.” [Electronics 1954]

By April 1955 the 2N78 transistor was available in volume quantities:

Advertising for Mass Production of the 2N78 in Electronics April 1955

“The General Electric N_P_N transistor type 2N78 is designed for RF and IF amplification in broadcast receivers. It employs a new and revolutionary manufacturing technique – the exclusive GE rate-growing process. This process makes possible stable operation at junction temperatures up to 1—C. The 2N78 is hermetically sealed and rated for a collector dissipation of 50 mw in 30C free air.” [General Electric datasheet ECG-78]

This set of 2N78 are from mid 1955 representing some of the first from the new production line:

Early 2N78 and a 2N167

The melt-back process invented by Hall [Ward 2001] also exploited the stronger incorporation of P-type impurities in germanium at slow crystallization rates and can be used to make NPN transistors. In this case small bars of single crystal N-type germanium with a minority amount of P-type impurities are prepared. One end of the bar is melted and on solidification P-type germanium forms first followed by N-type thus converting a uniformly N-type bar into a NPN structure. The General Electric 2N1289 is an example of a melt-back transistor.

Relative high frequency performance of these types is shown in the table.