The Transistor Revolution (Part 1)
Dr Bruce Taylor HB9ANY describes the invention of the tiny device that changed the course of radio history.
Dr Bruce Taylor HB9ANY describes the invention of the tiny device that changed the course of radio history.
At Christmas 1938, working in their small rented garage in Palo Alto, California, two enterprising young men called Bill Hewlett and Dave Packard finished designing a novel wide-range Wien bridge VFO. They took pictures of the instrument sitting on the mantelpiece in their house, made 25 sales brochures and sent them to potential customers. Thus began the electronics company that by 1995 employed over 100,000 people worldwide and generated annual sales of $31 billion. The oscillator used five thermionic valves, the active devices that had been the mainstay of wireless communications for over 25 years.
But less than a decade after HP’s first product went on sale, two engineers working on the other side of the continent at Murray Hill, New Jersey, made an invention that was destined to eclipse the valve and change wireless and electronics forever. On December 23rd 1947, John Bardeen and Walter Brattain at Bell Telephone Laboratories (the research arm of AT&T) succeeded in making the device that set in motion a technological revolution beyond their wildest dreams. It consisted of two gold contacts pressed on a pinhead of semi-conductive material on a metallic base.
The regular News of Radio item in the 1948 New York Times was far from being a blockbuster column. Relegated to page 46, a short article in the edition of July 1st reported that CBS would be starting two new shows for the summer season, “Mr Tutt” and “Our Miss Brooks”, and that “Waltz Time” would be broadcast for a full hour on three successive Fridays. But right at the end, after another unexciting story about the broadcasting of road traffic reports, the article mentioned that Bell Labs had demonstrated a small metal cylinder that could “create and send radio waves” but contained “no vacuum, grid, plate, or glass envelope to keep the air away”. It could amplify and oscillate and had been named a “Transistor”.
The name had been chosen by internal ballot among Bell Labs executives and research staff. Semiconductor Triode and Surface States Triode were considered fairly good but unwieldy, and Transistor came out well ahead of Crystal Triode, Solid Triode and Iotatron. Little did the apathetic NYT reporter realise that he had witnessed the first public demonstration of an invention that would spawn a world-changing technology.
Quite independently, Herbert Mataré in June 1948 also invented the point-contact germanium transistor, which he called a Transitron, while working for CFS Westinghouse near Paris. By mid-1949 many of them were in use as amplifiers in the French telephone system. At this time European industry was still recovering from the devastation of war but research at UK companies such as BTH, GEC and STC was not far behind the US and their first products were named Crystal Valve and Germanium Triode as well as Crystal Triode.
The roots of the invention were much older. The rectifying properties of crystals had been discovered by Karl Braun in 1874, before wireless existed, and the cat’s whisker detectors that became popular in the early 1900s were semiconductor diodes in all but the name. Nor was the concept of a three-electrode solid-state amplifier a new one. As far back as 1926, the German-American engineer Julius Lilienfeld, who invented the electrolytic capacitor, filed a patent for an FET-like device that was granted in 1930. Since materials of the purity required weren’t available at the time, it is unlikely that Lilienfeld succeeded in making a working transistor but his claim was strong enough to prevent Bell Labs from patenting the field-effect approach 18 years later.
The theoretical foundations for transistor operation were laid in 1931, when Cambridge University mathematician Alan Wilson formulated the quantum mechanical theory of conduction in semiconductors. He correctly attributed their properties to the presence of impurity atoms in the crystals, opening the way for less empirical work on solid-state devices. During WW2, Wilson worked on radio communications for the secret Special Operations Executive (the ‘Ministry of Ungentlemanly Warfare’) and later on the UK project to develop the atomic bomb. He was knighted in 1961.
With the development of the high-power 10cm cavity magnetron by John Randall and Harry Boot at Birmingham University, high resolution microwave radar became feasible if a reliable detector could be found. After Bell Labs were unsuccessful in developing thermionic valves for this very short wavelength, attention was turned once again to cat’s whisker crystal diodes. In Britain, the Telecommunications Research Establishment (TRE) developed an aluminium-doped silicon cartridge that was manufactured by BTH and GEC. While at TRE, the eccentric Bristol physicist Herbert Skinner improvised a technique for finding a sweet spot for the contact by tapping the crystals with an old pipe-cleaner knife and these detectors famously outperformed those that were not optimised in this way. He refused to allow anyone to clean the dirty knife, which had acquired “exactly the right momentum for the job”!
This war effort also gave rise to the next important breakthrough. While working on radar detectors in 1940, Bell Labs chemist Russell Ohl made the fortuitous discovery of a rectifying junction at a defect in a bar of silicon and noted its photovoltaic behaviour. Ohl had been bitten by the radio bug when using spark transmitters on 150m during WW1 and had built a superhet as early as 1921. He coined the terms n-type, for material containing some atoms of phosphorus, antimony or arsenic (in which conduction is by electrons), and p-type, for material containing some atoms of boron, aluminium or gallium (in which conduction is by holes), and the p-n junction was born. Initially this discovery wasn’t disclosed outside Bell Labs, and Ohl was instructed to cut any chance p-n junctions out of silicon that was sent to his British counterparts.
Work on semiconductors at Bell Labs had begun before the war, and energetic boss Mervin Kelly initiated a new unified solid-state programme in June 1945, as soon as the men who had been assigned to military work began to return to the lab’s new ‘Idea Factory’ at Murray Hill. The aim was specifically to devise an alternative to thermionic valve telephone amplifiers and the initial funding of $417,000 was billed to AT&T.
As manager of the programme, Kelly appointed William Shockley, a London-born physicist who had been engaged on anti-submarine warfare research during WW2. When he learned of Ohl’s discovery, Shockley immediately postulated that it might be possible to make a solid-state amplifier by applying an electric field across a p-n junction but, initially, all attempts to vary the conductivity with the control field failed. Without any tools to see what was happening inside the crystals at the subatomic level, progress was dependent on intuition and trial-and-error. By January 1946 the group admitted that they were “groping in the dark”.
However, the failures led Bardeen to postulate a theory of surface states in semiconductors and to continue experimentation. By late 1947, Bardeen and Brattain had switched from silicon to n-type germanium and felt that they were getting close to success. Then Brattain tried a configuration in which a pair of very closely spaced point contacts were created by using a razor blade to cut a minute gap in the gold foil wrapped around the edges of a small triangle of plastic, Fig. 1. The contacts were pressed into the surface of the germanium by a spring fashioned from a paper clip. By December 16th the team had achieved significant power gain and on December 23rd they demonstrated speech amplification to Bell Labs management. An “entirely new thing in the world” had been created.
On June 30th 1948, Research Director Ralph Bown used a giant cutaway model to introduce the first production transistor to the press, Fig. 2.
Everyone at Bell Labs was familiar with the legend that their company originated because Alexander Bell had beaten Elisha Gray in a race to the patent office. So, the transistor invention was classified as Bell Labs Confidential until it was better understood and patent protection had been applied for. But in whose name? Shockley was Bardeen and Brattain’s supervisor, and was disgruntled that his subordinates had made the breakthrough without his active participation, Fig. 3. Bell Labs’ lawyers advised that Shockley’s own work was overshadowed by that of Lilienfeld, and that Bardeen and Brattain were the actual inventors. Consequently, they refused to put his name on the patent for the point-contact transistor.
Professional jealousy and bruised ego spurred Shockley into a frenzy of independent work, which he initially kept secret from the rest of the group, thus starting to alienate them. By the end of January 1948 he had come up with a theoretical transistor design that worked quite differently from that of Bardeen and Brattain, being composed of a sandwich of p-type germanium between two n-type regions. After it was discovered that minority carriers could indeed traverse the bulk semiconductor, the concept appeared feasible and he successfully patented this bipolar ‘junction transistor’ in his own name. It would eventually supersede the point-contact type but at the time no-one knew how such a device could be fabricated and it was dubbed a ‘persistor’ because it seemed that much persistence would be required to make it.
Meanwhile, with patent protection secured and after Naval Research Laboratory staff had withdrawn a claim that they had already made the same thing, Bell Labs announced the invention of the point-contact transistor at a press conference on June 30th 1948. The attendees were given headphones to hear the device amplify and oscillate and to listen to a broadcast received on a radio set that used transistors instead of valves. After the unveiling, the public showed only lukewarm interest but Bell Labs was besieged by requests for sample devices from the electronics industry and the armed forces. By the summer of 1949 the lab had fabricated 4000 working germanium transistors.
Also in that year, the US Justice Department filed a new antitrust suit against the Bell System. In view of this, as soon as the military agreed in late 1951 that transistor technology need not be classified, AT&T made manufacturing licences available without restriction to interested companies in NATO countries and Japan for the relatively modest fee of $25,000 (about $260,000 in today’s money). Licensees, government labs and university researchers were given sample transistors. They were also invited to a technology symposium and a two-day transistor manufacturing plant tour at Western Electric that was attended by 100 representatives from 40 companies, including BTH, Ericsson, GEC, Philips, Siemens and Telefunken from Europe. When published in book form, a revised edition of the symposium proceedings called Transistor Technology was soon dubbed Ma Bell’s Cookbook. In 1955 AT&T relinquished its original transistor patents to stave off forced divestiture, although Western Electric hung on to those covering key manufacturing processes.
In the UK, Mullard started production of the point-contact OC50 and OC51 in 1952, before launching a range of junction transistors the following year. Subsequently, over ten British manufacturers produced hundreds of transistor types for different applications, including CV versions for military use. Mullard’s Reference Manual of Transistor Circuits provided a useful starting guide for many amateur designers but high frequency transistors were still rare and the only transmitter described was a 5-transistor unit for the emergency frequency of 500kHz.
While the first point-contact transistors used polycrystalline germanium, the key to the manufacture of junction transistors is the production of extremely pure single crystals that can be doped very precisely with p-type and n-type impurities. Chance often plays a vital role in discovery. In 1915, the Polish chemist Jan Czochralski absent-mindedly dipped his pen into a crucible of molten tin instead of his inkwell. When he lifted the pen, he found that it drew out a filament of tin that proved to be a single crystal.
Bell Labs chemical engineer Gordon Teal advocated developing this technique for growing large single crystals by drawing a seed slowly from a crucible of molten germanium but he received no support from Shockley or management for this work. Working evenings and weekends, he pursued the idea unofficially, shutting his apparatus away in storage during the day since there was no room for it in the metallurgy lab. By late 1949 he was able to make very pure single germanium crystals and the huge improvement compared with polycrystalline material became evident to all. For the production of wafers for modern VLSI microcircuits, the Czochralski/Teal process is still the basis of the principal method of growing silicon boules, which today may be up to 2m long and 450mm in diameter.
At the same time, William Pfann invented the zone-refining method of purifying a crystal ingot by using induction heating coils to displace a small molten region slowly along its length, Fig. 4. This causes the impurities to be swept to the end of the crystal, which is eventually cut off. For silicon crystals, a variation of this technique called float zone-refining was devised in which contamination is avoided by maintaining the molten portion in place by surface tension, without a containment vessel. A similar process was also developed independently by Siemens in West Germany.
To produce the npn or pnp sandwich of Shockley’s junction transistor, Fig. 5, Morgan Sparks invented a double-doping method of changing the relative impurity concentrations while the crystal was being drawn from the melt, producing the first grown-junction germanium transistor in April 1950, Figs. 6 and 7. After further refinement, Bell Labs announced this invention in July 1951. Meanwhile, the trio that had made the initial invention had broken up. Frustrated by Shockley’s management style, Brattain refused to work with him again, while Bardeen left Bell Labs altogether. This was a foretaste of the problems that would later lead to the downfall of Shockley’s business enterprise.
For transistor manufacture, silicon has inherent advantages over germanium because of its greater abundance, lower leakage current and higher operating temperature range. But the material is more difficult to purify and process and, owing to the lower charge carrier mobility, a silicon transistor must have a much narrower base region for high frequency operation. It wasn’t until 1961 that, with gold doping and epitaxy, a silicon transistor (the 2N709) exceeded germanium speed.
During WW2 great strides had been made in the art of purifying the material, and by 1945 silicon of 99.999% purity was available. Further improvements by DuPont allowed the first grown-junction silicon transistor to be fabricated at Bell Labs in January 1954 but in spite of this success the development wasn’t followed up commercially by Western Electric, the manufacturing side of the AT&T group.
Meanwhile Teal had left Bell Labs to join a fledgling start-up company that would become called Texas Instruments. After telling Teal repeatedly “This business is not for you. We don’t think you can do it”, Bell Labs finally agreed to accord a licence to TI in 1952. But instead of making only germanium transistors Teal established a group to work on grown-junction silicon transistors and 150 good ones had been made by the time of the IRE National Conference in May 1954. At the conference, speaker after speaker reported that the development of the silicon transistor would take several years so that Teal created a sensation when he revealed that he had a handful of them in his pocket and that commercial production at TI was beginning immediately. With no effective competition, TI then dominated the market for silicon transistors for several years. The little company was suddenly in the big league.
In the UK, the TRE electronics engineer Geoffrey Dummer, a specialist in radar synthetic trainers, presented a fundamental paper in 1952 describing the concept of the integrated circuit. But British semiconductor manufacturers were unwilling to take the risk of investing in the idea and it would be six years before radio engineer Jack Kilby (W9GTY) of TI was awarded a patent for the first crude IC. He demonstrated that transistors, diodes, resistors and capacitors could be fabricated in a single chip, although his hybrid design required them to be interconnected by flying bonding wires. Bell Labs researchers also missed the integrated circuit and Western Electric had to enter cross-licensing agreements to make them.
In addition to hearing aids, it seemed natural that portable radio receivers would be among the first commercially made consumer products to employ transistors. Demonstration sets were shown at the 1952 Components Exhibition in Paris and at the 1953 Dusseldorf Radio Fair, where the Intermetall Corporation that had been founded by Mataré on his return to Germany demonstrated a prototype receiver that used four hand-assembled point-contact transistors.
But the established radio manufacturers were initially reluctant to change over from thermionic valves, which were cheaper than transistors at the time. So, in October 1954 Texas Instruments and Industrial Development Engineering Associates (IDEA) in Indianapolis teamed up to launch the famous $49.95 Regency TR-1, Fig. 8. This pocket-sized MW AM set used four npn germanium junction transistors in a superhet circuit with one local oscillator/mixer, two 262kHz IF stages and one audio amplifier. (The low IF was chosen to increase the gain). Before teenagers adopted it as a personal music player, the radio was marketed as a ‘security item’ since the USSR was now an atomic power. In spite of its somewhat mediocre performance, the radio was a success and almost 100,000 sets were sold in the first 12 months. But having underestimated the cost of making the radio, TI hardly broke even on the venture and after that the company concentrated on making semiconductors for other manufacturers to incorporate in their products.
One of the enterprises that emulated the TR-1 was a small Japanese company called Tokyo Tsushin Kogyo (Totsuko). With little more information than a Bell Lab’s licence and Ma Bell’s Cookbook, they started manufacturing point-contact and then junction transistors, obtaining a reasonable yield of about 5%. After a false start with a portable radio whose plastic case was found to come apart in the summer heat, they launched their TR-55 model in August 1955. This 5-transistor set was Japan’s first transistor radio and it spearheaded the entire Japanese consumer electronics industry. For the product launch, Totsuko’s owners changed their tongue-twisting brand name to one that could more easily be pronounced in the West. They called it Sony. The name became well known outside Japan after discerning thieves who broke into a New York warehouse ignored other makes to steal only the 4000 Sony radios stored there.
In 1953 Philco developed electrolytically machined surface-barrier germanium transistors that could operate up to 60MHz and manufacturing licences were granted to other companies including Sprague and a division of Plessey. In 1955 these transistors were used in the first commercial transistor car radio and in 1958 improved versions were orbited in the 108MHz beacon transmitter carried by Explorer 1, the first US satellite.
Next month I will conclude by talking about the first uses of transistors in amateur radio and more recent developments in transistor technology.
This article was featured in the October 2018 issue of Practical Wireless