Fullscreen HTML5Med Res (??MB)HTML4

The History of Transistors Transistors have reshaped the world since their invention 75 years ago. Computers, mobile phones, tablets, the internet, high definition TVs... none of this would be possible without transistors. While the history of the transistor could fill a book (and properties of transistors several more), this short series of articles covers the most interesting bits. Part 1: by Ian Batty 12 Silicon Chip Lead image: John Bardeen (left) and Walter Brattain (right) explain their invention to William Shockley (centre) Australia's electronics magazine siliconchip.com.au , to most electronic devices, you’ll design repair or understand need to understand how various types of transistors work. Since their first commercialisation, transistors have gone through ten distinct manufacturing methods (and hence transistor technologies). It has been difficult to find, in a single source, straightforward descriptions of transistor construction and operation. These articles are intended for casual reading, as a guide to operation and repairs and as a compact reference work. We’ll start by describing the invention of transistors and the major developments that followed. After that, we’ll have some details of more modern manufacturing techniques, semiconductor physics, doping, and diode and transistor behaviour. Thermionic valve (tube) limitations Valve technology underwent explosive development between Fleming’s patent for the thermionic diode in November 1904 and Bernard Tellegen’s patent application for the pentode in 1926. Receiving valve technology matured in the 1960s with miniature ceramic devices. But three problems inherent in thermionic valves persisted over that time, none of which was ever fully solved. Heater/filament power consumption Leaving aside some special types, amplifying, rectifying and oscillating valves use thermionic emission from a heated filament or cathode. The tiny DL66 hearing-aid output valve delivers only 0.95mW of output, yet needs 12.5mW of filament power – almost thirteen times its useful output. The 6SA7 converter delivers a voltage signal of virtually zero power but demands about 1.9W of heater power! A fair comparison would be between a 12CX6 ‘hybrid’ valve, using both 12V heater and HT supplies, and a relatively early (1965) AF121 diffusion-­ alloy germanium transistor, in both cases having a 12V DC supply. The 12CX6 will draw ~4.4mA of anode/screen current from the 12V HT supply, for a total HT power of about 53mW. The AF121 transistor will draw ~3mA from its 12V supply, giving a comparable figure of 36mW. But we need to add in the valve’s heater current of 150mA. This adds 1.8W, for a total of 1.85W; over 50 times the power for a transistor doing the same job. Limits to miniaturisation Thermionic valves rely on a vacuum to separate their electrodes. Conventional valves use a concentric structure, with the grid or grids and anode surrounding the filament or cathode. This demands great manufacturing precision and limits the minimum size. A planar (layered) structure can be made to much higher precision and much smaller. This approach delivered the 7077 ceramic triode – just 11.3mm tall and 12.2mm in diameter. Impressive as this is, the 7077 is hardly smaller than most transistors of 1958, which points to the limits of thermionic valve miniaturisation. Fig.1 shows a 7077 beside an OC76, one of the second generation of junction transistors, reproduced at an enlarged magnification. Transistors are now manufactured with dimensions measured in nanometres, a degree of miniaturisation impossible for thermionic valves. Frequency limitations Electrons in a thermionic valve pass from the cathode to anode across an evacuated space. At very high signal frequencies, transit time effects (the time taken for electrons to travel that distance) set absolute limits to triode valve operating frequency. The problem is most easily understood by considering the electrons just approaching the grid being out of phase with those just leaving. More are arriving than leaving, or more leaving than arriving. Now that these numbers no longer balance, the grid no longer appears as a small capacitance but instead as a low impedance. This grid loading demands power from the driving stage, even in voltage amplifiers. As a result, the 6BL8 converter has an input impedance of only a few kilohms at around 100MHz, limiting the gain available in an FM radio. Grid loading and other more complex effects set an upper limit of around 2.5GHz for thermionic triode amplifiers such as the 7077, with a few types extending to some 7.5GHz. Three main ‘non-triode’ types of thermionic valve were developed: magnetrons, klystrons and travelling-wave tubes (TWTs). Although these can operate at frequencies approaching 100GHz, only the klystron and TWT can amplify. Klystrons and TWTs are pretty noisy, with the better-performing TWTs having noise figures of about 7dB, making them unsuitable for weak-signal amplification. These three thermionic types are physically large, and the amplifying versions consume many watts of power. One variant of the TWT, the Backward-Wave Oscillator (Carcinotron), can work as an oscillator up to 1THz (1000GHz). Current transistor developments (as of 2022) are yielding low-noise amplifiers with operating frequencies exceeding 500GHz and gains of 20dB. In summary, thermionic valves for general-purpose amplification had reached their limits of development by the early 1960s. The 7077 ceramic triode is a fine example of how far valve development had come and the limits to further practical development. 12.2mm 5mm Fig.1: the 7077 ceramic triode (12.2mm diameter), along with an early germanium alloyed-junction transistor (OC76, 5mm diameter), both shown larger than life. While they are similar in size at this early stage of transistor development, it didn’t take long for transistors to shrink further. siliconchip.com.au Australia's electronics magazine March 2022  13 Early transistor attempts Brought into public consciousness by Michael Faraday’s popularisation in the early 1800s, electrical science seemed to produce a new miracle in every decade of the 19th century. Willoughby Smith’s 1873 work with selenium rods demonstrated an oddity: selenium’s resistance was affected by incident light. Clearly, there was more to electricity than Ohm’s Law of conduction in metals. Smith’s results foreshadowed the semiconductor revolution. Semiconductor diode action was demonstrated by Karl Braun just a year later in 1874, followed by Bose and Pickard’s 1904 practical application of these detectors to radio reception. Many military receivers used in World War I were based on solid-state ‘crystal’ detectors. Once Lee de Forest had demonstrated his triode Audion valves, some must have wondered whether that same principle could be applied to triode amplifiers. In 1925, Julius Lilienfeld formerly of Leipzig University filed a Canadian patent for a solid-state device that foreshadowed the modern field-effect transistor, similar in operation to thermionic triodes. The device passed current through a thin sheet, but subjected that current to a controlling electric field. His US patent was granted in 1930 (see https://patents.google.com/ patent/US1745175). Lilienfeld did not publish research papers, so his device was not taken up by industry. A replica was finally built and tested in the 1990s and proved effective as an amplifier (see https://w. wiki/4YLL). It’s less well known that Lilienfeld also lodged the first known patent for the junction transistor design in 1928 (https://patents.google.com/ patent/US1877140A). This was around 20 years before William Shockley’s 1948 patent (which we now know as the transistor). Shockley’s design was fundamentally identical to Lilienfeld’s (https://patents.google.com/patent/ US2569347A). The two devices are identical in current flow (emitter-through-baseto-collector) and physical design (two back-to-back diodes with one reverse-biased). Fig.2 from Lilienfeld’s 1928 US patent 1877140 shows the circuit for a solid-state amplifier; the current path and structure of the device clearly 14 Silicon Chip Fig.2: Lilienfeld’s junction transistor as part of an amplifier circuit, from US patent 1,877,140. anticipate the junction transistor. The operation of Lilienfeld’s design is described briefly in his patent application. human skill to fabricate with the fineness necessary to produce amplification.” Bardeen and Brattain’s patent (1950) lists the first known transistor specifiFast-forward to 1947 cations for their point-contact design, On receiving his doctorate in 1936, including power gains of just 16-19dB William Shockley was recruited to Bell and a current gain of merely 1.3 times. Laboratories and joined a team of physWalter Brattain, John Bardeen and icists researching solid-state electron- Robert Gibney, working at Bell Labs’ ics.With the outbreak of World War Solid State Physics Group (led by WilII, Shockley began working on radar, liam Shockley), made several attempts joining Columbia University’s Anti-­ at the “solid-state triode”, but many Submarine Warfare Group in 1942. were found to infringe Lilienfeld’s Since thermionic diode mixers had existing patents. proven inadequate at the ultra-high Their first material of choice was silfrequencies used in radar systems, icon, but the high temperatures needed Shockley developed high-­performance (melting point 1414°C) proved diffiultra-high-frequency silicon diode cult, so they switched to germanium mixers. (melting point 938°C). His success led him to consider Shockley started wanting to repwhether his diode design might be licate thermionic triode operation transformed into a triode structure, – electron flow controlled by a non-­ thus allowing amplification. conducting electrode. This looked Papers show that Shockley and his forward to the modern family of field-­ colleague Gerald Pearson had actually effect transistors (including those used built ‘Lilienfeld’ devices but didn’t in CMOS chips such as microprocesrefer to that in their published papers. sors). But Shockley was unable to It is notable that the successful Bar- demonstrate any useful effect. deen and Brattain point-contact tranHe eventually developed a transissistor patent (https://patents.google. tor explicitly using two types of curcom/patent/US2524035A) describes rent carriers: electrons and holes – the Lilienfeld’s 1925 (Canadian patent) junction transistor. We’ll return to that mechanism – which would become a little later. today’s overwhelmingly-used field-­ Bardeen and Brattain, working effect technology – as “…beyond without Shockley due to his abrasive Fig.3: the point-contact transistor, from Bardeen & Brattain’s US patent 2,524,035. These performed reasonably well, but they were tricky and expensive to manufacture. Australia's electronics magazine siliconchip.com.au Fig.4: Shockley’s junction transistor (this drawing from US Patent 2,569,347) was a significant step from the point-contact transistor. It was much easier to manufacture in bulk and less fragile too. Fig.5: Shockley’s patent also included this five-layer compound transistor that was intended to operate similarly to a pentagrid mixer valve like the 6L7. management style, eventually demonstrated the device we know as the point-contact transistor in December 1947 and filed for the patent in 1948 (Fig.3). A press conference in June 1948 showcased the new device. Demonstrations included an amplifier and a radio receiver. Regrettably, no details of that radio are available. John Bardeen, Walter Brattain and William Shockley were jointly awarded the Nobel Prize for Physics in 1956 “for their researches on semiconductors and their discovery of the transistor effect”. Bardeen went on to win a second Nobel Prize in 1972 for his theory of superconductivity. of Shockley (who was unconvinced about the need for ultra-pure stock) and on his own initiative. Teal said, having previously been involved in germanium processing and diverted to other projects, only to return: “If I ever had another idea I considered a world-beater, I’d work on it even if nobody gave me any help.” The program also established processes such as the production of P-N junctions and attachment of leads to devices. Morton made an important strategic decision to share transistor technology with other researchers so that Bell Labs and parent AT&T could benefit from a cooperative approach. Bell Labs held three famous seminars where scientists and engineers visited Bell Labs to learn the new semiconductor technology first-hand. The first meeting, in September 1951, specifically addressed military uses and applications. Proposals to classify the transistor (and thus make it unavailable to the civilian world) were, thankfully, not pursued. In April 1952, over 100 representatives from 40 companies that had paid a US$25,000 patent-licensing fee came for a nine-day Transistor Technology Symposium, including a visit to Western Electric’s ultramodern transistor manufacturing plant in Allentown, PA. There were participants from such Out of the lab and into the fab Bell Telephone Laboratories realised that the transistor was a revolutionary device with the potential to transform electronics. Bell Labs pursued a vigorous program of “fundamental development” in the late 1940s and early 1950s, promoting rapid improvements in transistors and other solid-state devices. Electrical engineer Jack Morton led this program, developing processes such as zone refining – critical to the high purity of materials needed – and growing single crystals of germanium and silicon. Gordon Teal perfected his refining processes against the advice Fig.6: while it’s reasonably certain that Lilienfeld transistors were built, there isn’t much information left outside his patent on just how they worked. Their structure is quite different from modern transistors and, as shown here, they were made from metals and a semiconductor (silver sulfide). siliconchip.com.au Australia's electronics magazine electronics titans as GE and RCA, as well as from then-small firms including Texas Instruments. The Bell Telephone Company was established by Alexander Graham Bell, who had started as a teacher of the deaf and who spent much of his career in service to the hearing-impaired. So, commendably, Bell Labs waived all patent royalties for the very first transistorised consumer product in 1953 – a hearing aid. The one-month wonder The personal story of the people behind the junction transistor is as interesting as the story of the technology itself. In brief, Bardeen and Brattain were pivotal in developing the working point-contact transistor, and Shockley felt that he had been excluded from the project; indeed, the patent was issued only to Bardeen and Brattain. So, working by himself, Shockley designed the junction transistor in one month, claiming an entirely new approach to solid-state electronics (https://patents.google.com/patent/ US2569347A), and one which would become the basis for all subsequent development. Fig.4 (from the patent) is the basic ‘triode’ device. Aside from Shockley’s design using homogeneous material (all germanium), it is remarkably similar to Lilienfeld’s 1928 patent, shown in Fig.2. The patent shows that Shockley also envisaged complex multilayer devices. One five-layer proposal (Fig.8 in the original Shockley patent, or Fig.5 here) would operate similarly to the 6L7, a pentagrid mixer valve. Shockley defined two active layers (92 and 94) for signal injection and local oscillator injection. Early junction transistor designs Shockley’s design differs from Lilienfeld’s earlier junction transistor (Fig.6) in several important ways. As with the vacuum tube triode, both Lilienfeld and Shockley aimed to produce a space charge within the device that could then be controlled by an electrode intermediate between the ‘emitting electrode’ and ‘collecting electrode’ (cathode and anode in valve terminology). In this most fundamental way, junction transistors are similar to vacuum triodes. March 2022  15 Regrettably, no detailed description of Lilienfeld’s device exists. What follows is based on the Lilienfeld Patent. Lilienfeld’s device used a thin, deposited intermediate base layer overlaid on a substrate. The base was then deliberately micro-fractured to present a fine mesh-like surface. The collector deposition covered the base surface, penetrating gaps in the base ‘mesh’ layer to give electrical contact with the emitter layer. In operation, the forward-biased base-emitter junction created a space charge in the interface between base and emitter, presumably of electrons. The space charge existed on the underside of the base, just as with any planar diode such as a copper-­ oxide rectifier. But the space charge also existed in the minute crevices in the base layer, so it was subject to the attracting field from the collector. Thus, the base-emitter forward bias would establish a space charge that could be drawn through the base region to become collector current. By contrast, Shockley’s design created a voluminous space charge entirely within the homogeneous and continuous base layer, allowing the space charge to diffuse in all directions throughout the base, most notably towards the base-collector junction. This is described in detail in his patent, beginning on page 29. To reinforce the distinction, Lilienfeld’s design created a useful space charge at the interface between base and emitter, where Shockley’s created it entirely within the base. Lilienfeld’s collector-base junction is – like Shockley’s – reverse-biased. For both designs, zero bias means zero collector current. They both operate in contrast to a vacuum triode, where zero grid bias means maximum anode current. It’s not known how well Lilienfeld’s device worked. Shockley’s design intentionally created a large surplus of charge carriers in the base region due to its low doping concentration. Lilienfeld does not address this matter, and it is unclear how such a surplus could have been established, and consequently, whether his device could have worked as he claimed. Lilienfeld states a base thickness of 200µm, comparable to first-generation grown-junction transistors. He also mentions the need for overall small size to lessen capacitive effects. 16 Silicon Chip Lilienfeld’s 1925/1930 field-effect patents did specify “a film of copper sulphur [sulfide] compound”, but only to provide the extremely thin, high-­ resistance film needed for his design. Modern field-effect devices use a single semiconductor (silicon) for all of the device’s elements. Oskar Heil filed for a UK patent on a field-effect device in 1935 (patents. google.com/patent/GB439457A). His device specifically described the use of semiconductors and a thin insulating layer between the control electrode and the conduction part. It’s essentially the insulated-gate construction of virtually all metal oxide semiconductor (MOS) devices, from memory chips to microprocessors. Shockley Semiconductor Lab Arnold Beckman had built a substantial instrument company by the 1950s, beginning with a successful pH meter. He and Shockley had been friends for some years. Shockley left Bell Labs in 1955 and negotiated with Beckman to form his own Shockley Semiconductor Laboratory in mid-February 1956 (Fig.7). Gathering a stellar team of physicists and engineers and intending to develop and market junction-technology transistors, this ought to have been a very successful industry startup. Amusingly, Beckman had already paid Bell Labs the $25,000 licence fee for patent rights to transistors. Shockley Semiconductors was even forced to send two of its employees to the final Bell Labs Seminars on diffusion so that Shockley’s new company could be updated on the latest transistor theories. After a year’s intensive effort, Shockley’s company had failed to sell a single device, and Shockley had proven a poor leader. Rather than directing efforts towards perfecting his own patent of the junction transistor, he proposed distracting projects, including the development of his Shockley Diode. It was a four-layer PNPN device that would develop into the SCR/thyristor, today widely used in power control and finally developed as the Triac family of devices. This focus seems puzzling at a time when analog signal processing and amplification dominated domestic, communications, telephone and military electronics. But Bell’s gargantuan Australia's electronics magazine telephone network was switched by noisy, power-hungry electromechanical relays and switches needing constant maintenance. Shockley’s device would have revolutionised telephone exchange technology. But internal friction, fuelled by Shockley’s domineering management style, led to the exodus of Julius Blank, Victor Grinich, Jean Hoerni, Eugene Kleiner, Jay Last, Gordon Moore, Robert Noyce and Sheldon Roberts. This group are known as the “Fairchild Eight”. Shockley Semiconductor was sold to Clevite in 1960, having produced no commercial product. The “eight” were snapped up by the Fairchild Camera and Instrument Company, spinning off to become Fairchild Semiconductors. Work there ultimately led to the planar design, the basis of all modern silicon devices, from single transistors to microprocessors and memory chips with billions of individual components per chip. Fairchild Semiconductor remained a commercial success until September Fig.7: public artworks in Mountain View, California commemorate the site of Shockley Labs. Source: Wikimedia user Baltakatei (CCAShare Alike 4.0 International) siliconchip.com.au 2016, when the company was acquired by ON Semiconductor (previously Motorola’s semiconductor division). Ironically, Robert Noyce’s management style led the inventors of the integrated-­ circuit op amp to desert Fairchild and join National Semiconductor (which merged with Texas Instruments in September 2011), taking their extensive analog design expertise with them. Technologies in more detail Having gone over the basic history of transistors, let us take a more detailed look at the different processes used to fabricate those early transistor types. Point contact transistors The path to Bell Labs’ most famous patent was somewhat torturous. Bell Laboratories was formed in 1925 as an amalgamation of the research arms of Western Electric and American Telephone & Telegraph. Aside from their principal work on telephone systems, Bell Labs contracted to the US Government and, focusing on basic research, produced several Nobel Prize winners. The Bell System Technical Journals (https://archive.org/details/bstj-­ archives) detail Bell’s work from 1922 to 1983, which includes some of the foundations of today’s electronic and communications technologies. Shockley had begun from a ‘thermionic triode’ perspective, intending to pass current through a single piece of semiconductor. He would add an insulated metal contact with an applied potential on one side and use that contact’s electric field to control the current in the main channel. Fig.8 shows his intended device, which today we would call a depletion-­mode Mosfet. Over some two years of frustration, Shockley attempted to demonstrate his expected effect and failed each time. At this early stage of research, no one had anticipated two requirements: near-absolute purity of the semiconductor material and crystal regularity approaching perfection, especially at the surface. Fig.8: this is the device that Shockley was trying to build – essentially a semiconductor analog of the vacuum tube triode. Such devices were eventually built and are known as depletion-mode Mosfets (they’re similar to JFETs but have an insulating layer between the gate and channel) like the BSS139. Fig.9: the operation of point-contact transistors is still not fully understood, and it probably never will be as they are obsolete devices and there is no longer any active research. This is our best guess as to how they work. siliconchip.com.au Australia's electronics magazine Later research proved that the chaotic ‘tangled’ surface states which diffused and opposed any external field’s influence were the principal cause of Shockley’s failures. Gordon Teal’s advice regarding feedstock purity (noted earlier) and crystal regularity may well have delivered Shockley the device he had envisaged, had Shockley heeded it. In desperation, Bardeen and Brattain flipped the device: current would enter via a surface emitter contact, flow through the base material, then exit via a second collector contact. The strangest (but most successful!) results were obtained by adding a small drop of liquid – some electrolyte from a butchered electrolytic capacitor – to improve conductivity between the applied electrodes and the germanium surface. Finding an effect only at very low frequencies, they reasoned that a point contact (of the smallest possible diameter) would establish an intense electric field at the surface and perhaps give a higher operating frequency. They calculated they would need a separation between the points of about 0.002in (close to 50μm). Bardeen and Brattain then took a shortcut. Rather than waste time manipulating fine-pointed wires, Brattain had an assistant attach a strip of gold foil to a plastic wedge. Brattain then slit the foil with a fine knife and used the plastic wedge to press the two gold electrodes against the germanium base substrate. It was a revolutionary transposition. The first crude transistor’s operation came from abandoning expected theory and inventing a wholly new device with no ancestor: Lilienfeld and Heil’s prior devices (the bipolar junction and field-effect forms) contributed nothing to this radical invention. This also demonstrated that a transistor need not be made from only semiconductors: metal-semiconductor interfaces would also work, a fact exploited by later developments of micro-alloy and Schottky devices. The exact physics of the point-­ contact transistor (Fig.9) have never been fully described. Coblenz and Owens, writing in the 1955 book “Transistors: Theory and Applications” state “theories which adequately explain all the known phenomena of point contact operation have not been completed.” March 2022  17 Fig.10: the first point-contact transistor, created by Bardeen & Brattain. Source: Wikimedia user Unitronic (CC BY-SA 3.0) Fig.11: commercial point-contact transistors. They were potted in a plastic compound to protect the physically fragile device and prevent moisture/dust/etc from affecting their operation. Despite being produced commercially, they were still essentially hand-made devices and thus expensive. Image copyright 20012017 by Jack Ward, Transistormuseum.com It appears that much of the action took place under the upper surface of the germanium body. Still, it was the neutralisation of surface states in the collector region that contributed to increased collector current and thus current gain. The simplest complete explanation appears in the book “Fundamentals of Transistors” by L. M. Krugman & John F. Rider (1954) – see archive.org/ details/FundamentalsOfTransistors As well as owing nothing to any previous electronic device, the point-­ contact transistor’s method of operation is unlike any that followed it (including junction and field-effect transistors); its operation was unique. This allowed Bell’s patent attorneys to file with confidence. Most equipment using point-contact transistors has not survived. The majority is preserved in museums and the hands of collectors, with rare examples available online. As shown in Fig.9, the electron flow from the base to the emitter liberates ‘holes’ in the crystal. The liberated holes form a space charge and are attracted to the negative potential of the collector. Arriving at the collector, the holes from the space charge recombine with electrons entering from the collector lead. This recombination provokes additional collector current. Were the collector current only due to the space-charge holes from the emitter-base region, the collector current would be about the same as the emitter current. The transistor would show an emitter-collector current gain of about unity. But the extra collector electron flow to the base means that the collector current is greater than the emitter current. The result is a collector current about 2.5 times the emitter current. The microscopic contacts produce very strong local fields in the substrate, essential for power gain. Even in production, this was a hand-made structure, with the refinement of a ‘flash’ of current to form a more effective collector site. Somewhat reminiscent of Lee de Forest’s difficulties in understanding his Audion, Bardeen and Brattain struggled to describe the device they had invented. There was little ‘transistor action’ deep in the bulk of the crystal – the 18 Silicon Chip current amplifying action was mainly at (and just below) the surface. Yet today, bulk conduction is the sole mechanism used in bipolar transistors. Abandoning the idea of surface-only activity, the principle of bulk conduction was proven by John Shive in 1948 (see siliconchip.com.au/link/ abbe). This paved the way for Shockley’s groundbreaking junction transistor patent. The point-contact transistor’s handmade structure was difficult to manufacture with widely-varying characteristics, and susceptibility to surface moisture. This demanded meticulous and complete protection of the surface, leading to the development of Fig.13: the first European prototype transistor, made by Herbert Mataré in June 1948 by F & S Westinghouse in Paris, France. Source: Deutsches Museum, Munich, Archive, R5432 Australia's electronics magazine siliconchip.com.au applications, saw transistors in limited commercial use in the United States by 1953. RCA released several types and registered them with the then-new industry body, the Joint Electron Devices Engineering Council (JEDEC); the 2N21~26 and 2N50~53 series, 2N32/33 and 2N110 among them. TI also offered their Type 100 and Type 101 devices. Simultaneous discovery Fig.12: the physical structure of the prototype transistor shown in Fig.10. hermetic (airtight/watertight) sealing techniques – see Figs.11 & 14. One manufacturer said that the first transistor off his production line had cost a million dollars; 1954 dollars at that! The point-contact transistor had an appalling noise figure of about 45dB and was unreliable. It also exhibited negative resistance, causing it to oscillate and making it unusable as an amplifier in some configurations. It was, however, the only proven solid-state amplifying device in the early 1950s. Its small size and low power consumption made it a candidate for hearing aids. This, along with telephone repeater (amplifier) The improvement of radar technology was critical to aerial warfare in World War II, with both sides making full use of this technology. Heinrich Welker worked on the production of ultra-pure germanium crystals at the University of Munich during World War II. At around the same time, Herbert Mataré worked on microwave mixer diodes at the Telefunken plant in Silesia (at Bielawa, now part of Poland). Radar receivers must detect very faint signals – any noise generated within the receiver reduces sensitivity and, therefore, the maximum detection range. Local oscillator noise is the limiting factor in a set with a diode detector but no RF amplifier. Mataré discovered that a balanced push-pull detector, with two antiphase local oscillator signals, cancelled some of the local oscillator noise Fig.14: production versions of the European transistor, known as “Transistrons”. Inside each tube is a point-contact transistor. Source: Deutsches Museum, Munich, Archive, R5432 siliconchip.com.au Australia's electronics magazine and gave much-improved sensitivity. Mataré used point-contact diode mixers, the only device that would work at radar frequencies. Experiments in 1944 with two contact wires (for a push-pull circuit) showed that if the wires were very closely spaced, current in one wire would influence that in the other (see siliconchip.com.au/link/abbf). Mataré had discovered, prior to and independent of the work at Bell Labs, the principle of the point-contact transistor. Wartime demands prevented Mataré from pursuing his ‘transistor’ observations. Following the German surrender, Mataré taught physics at a US military academy near Kassel and Aachen university. During one briefing session, he was invited to move to Paris and set up a semiconductor plant for F.V. Westinghouse. Mataré and Welker’s research led to the production of diodes in 1946. Taking up his ‘double diode’ design, Mataré was granted US patent 2,552,052, lodged April 21st, 1948. More importantly, Mataré was able to demonstrate amplification in that year, 1948 – see Figs.13 & 14. His development program differed from that of Bardeen, Brattain and Shockley, as shown by Mataré’s different approach to surface preparation (see the PDFs at siliconchip. com.au/link/abbg and siliconchip. com.au/link/abbh). Like the Bell Labs team, Mataré and Welker struggled to unravel and understand the complex mixture of bulk and surface effects. Their first confirmed device was demonstrated in July 1948. Bell Labs’ release of their design prompted Mataré and Welker to rush a patent application to the French office. Their company, F.V. Westinghouse, applied for a French patent on August 13th, 1948, granted on March 26th, 1952. Stuck for a name, the French device became the “Transistron” to differentiate it from Bell Labs’ transistors. Transitrons were successfully used as early as May 1949 in telephone repeaters and were widely used by 1950. Despite the French devices being reported as superior to those from Bell, in the words of Michael Riordan, “Europe missed the transistor”. The French government, distracted by the threat of nuclear warfare with the Soviet Union, failed to support March 2022  19 Fig.15: probably the first public demonstration of a transistor radio at the 1953 Düsseldorf Radio Fair in Germany. Fig.16: a closer view of the radio shown in Fig.15. semiconductor manufacturing. Mataré left France for Germany and founded Intermetall (“Semiconductor”) in Düsseldorf, Germany. At the 1953 Düsseldorf Radio Fair, “a young lady wearing a black sweater and a multicoloured flowery skirt demonstrated to the public a tiny battery-operated transistor radio” – shown in Figs.15 & 16. The revolutionary work of Bardeen, Brattain, Mataré and Welker resulted in the creation of a solid-state amplifier that owed nothing to any ‘prior art’. However, the point-contact transistor was a dead end; poor performance, reliability and economics of manufacture condemned it to the dustbin of history. No complete functional and mathematical description of the device is ever likely to be written. with point-contact technology. His patent (https://patents.google.com/ patent/US2763832) gives an excellent description of the grown-­junction process. Source material of exceptionally high purity (highly regular germanium with no crystalline faults) was critical to reliable transistor production. Among other requirements, exceptional purity meant that electrical conductivity would be due only to carefully-measured doping chemicals, resulting in devices with predictable characteristics. Ordinary chemical methods were unable to produce highly-purified, regular crystalline stock. Zone refining passes ingots through a coil that heats the stock to its melting point. As the ingot passes through, it solidifies in cooler parts of the furnace. Impurities remain in solution and are ‘swept’ backwards relative to the ingot’s motion. In practice, furnaces used several heating coils, producing multiple refining zones in a single pass (see Fig.17). Germanium’s relatively low melting temperature allowed it to be conveyed in graphite ‘boats’. While this method gave much higher purity than simple chemical methods, it could not produce the ultra-high purity needed for transistor manufacturing. What about Doctor Adams? There are online claims that New Zealander Robert George Adams made transistor devices in the 1930s. For example, see http://blog.makezine. com/2009/04/02/the-lost-transistor/ You will find many references to candidates for ‘the inventor’ of the transistor. Some of these appear credible, others simply argumentative. I have focused on designs that were patented, and – more importantly – were either the direct antecedents of commercial devices or commercial devices themselves. Junction technology Taking up Shive’s work on bulk conduction (which had led to Shockley’s Junction Transistor patent), Gordon Teal’s patent for grown-junction devices revolutionised transistor manufacturing, making a complete break 20 Silicon Chip Fig.17: zone refining was one early method of purifying germanium feedstock. By passing an ingot through multiple induction heating coils, impurities could be ‘swept along’ the rod and ultimately removed. Australia's electronics magazine siliconchip.com.au Silicon’s much higher melting point necessitated running the ingot vertically without any form of container or support, relying on molten silicon’s natural cohesion to restrain the molten zone and not let the ingot collapse. This method needs no mechanical support. It also gave very high purity, so it was adopted for germanium. For germanium, Teal’s method was to melt well-purified germanium at about 940°C, then dip a seed crystal into the liquid, slowly rotating and withdrawing the seed vertically (at about 60 rpm and 0.8mm/second), as shown in Fig.18. The ‘pulled’ melt solidified into a highly-purified cylindrical crystal with a regular structure. The pulling furnace used a dry hydrogen atmosphere as air would affect the nature of the pulled crystal. This method worked equally well for germanium or silicon. Critically, it opened the door to the first truly successful transistor construction: grown junction. Semiconductor doping Practical semiconductors use highly-purified feedstock with tiny amounts of purposefully-added elements other than germanium or silicon. These ‘doping’ elements greatly improve conductivity (pure germanium and silicon are both very poor conductors). Doping creates the P- or N-type materials needed to make diodes, transistors and integrated circuits. Just one doping atom for about every ten million germanium atoms will give the conductivity needed for semiconductor action. A pentavalent element such as phosphorus donates electrons, so it is a donor impurity, making an N-type semiconductor. This is different from a common metallic conductor, which has a population of free electrons; the excesses in P- and N-type semiconductors are permanent, not like the mobile ‘electron clouds’ in metals which are, overall, electrically neutral. A trivalent acceptor impurity (such as aluminium) ‘steals’ an electron, leaving a positively-charged hole in the germanium, making it P-type. This means that the semiconductor has a permanent positive charge. Holes can be made to move in P-type material by an electric field, just as electrons can be made to move in N-type material. An excess of electrons (N-type) or holes (P-type) means that a doped semiconductor is a good conductor. It’s the ability to create different conductors with different current carriers that makes semiconductor devices possible. This is why the purity of the raw stock is critical. Precise electrical characteristics can only be guaranteed by starting from raw material of virtually absolute purity and adding precisely-­ controlled amounts of impurities. We’ll have more details on the effect of doping in a later article of this series. Teal’s development on the basic refining process was to add minute concentrations of doping gases to the furnace atmosphere. With an arsenic-­ containing atmosphere, P-type germanium was pulled. For N-type, phosphorus could be used. Fig.19 shows the process, with a doping ‘pill’ (rather than a gaseous doping atmosphere) controlling semiconductor polarity. But if the atmosphere were changed from, say, arsenic-rich to phosphorus-­ rich during a pull, the drawn crystal would begin as P-type, then transition to N-type. On completion of the pull, the crystal cylinder could be sliced, discarding most of the ends and leaving a disc containing the P-N junction, then cut across the disc to separate out numbers of individual square junctions. Voila! Diodes. Grown junctions If the pull was conducted slowly, and the melted pool sequenced from arsenic-rich to phosphorus-rich then back to a final arsenic-rich composition, the pull would contain three regions: P-type, N-type and P-type in a ‘sandwich’ (https://patents.google. com/patent/US2631356). Fig.18: one of the biggest breakthroughs in semiconductor manufacturing (which is in use to this day) was the pulling furnace process for generating ultra-pure giant crystals of germanium or silicon. These days, silicon crystals up to 400mm in diameter are made, although 300mm is a more typical size. Fig.19: it is possible to dope the molten germanium during the crystal pull. This results in graduated doping across the length of the crystal, or possibly even different doping zones within the crystal. siliconchip.com.au Australia's electronics magazine March 2022  21 This construction resulted in a large, single ‘transistor’. Fig.20 shows how careful slicing and dicing yields numerous individual transistors. This was William Shockley’s original Bell Labs patent. The world’s first transistor radio (the 1954 Regency TR-1) used grown-junction transistors (types X1 to X4) from the newly-formed Texas Instruments. Many types were given ‘in-house’ numbers, and grown-junction technology was being phased out as the 2Nxxx JEDEC nomenclature became established. NPN types 2N27~29 are among the registered grown-junction devices. The grown-junction process favours NPN construction. Many early transistors are NPN, including those in the Regency TR-1. NPN types also appear in the Regency’s TR-4/TR-5 and the Zenith Royal 500, implying that grown-junction technology was used at least until the issue of the Royal 500’s IF devices, type 2N216. TI released their germanium type 200 and type 201 in 1953 and returned to the technology with their silicon 2N389, as one JEDEC-registered example. Being a single, solid crystal, the grown junction was much more reliable and stable than the point-­contact construction. Since the regions – and their junctions – had been doped during the pull, no ‘forming’ was needed, as was necessary for the point-contact types. The characteristics were essentially stable from the moment of solidification until the end of life. Each sawn sliver needed to be mounted in a case and connections made to it, with the principal difficulty being the base’s location between the outer emitter and collector regions. There were also practical limits to base thinness – thinner bases give better gain and higher operating frequency, so this manufacturing technique limited the achievable performance. Fig.21 shows the long sliver sitting horizontally, soldered at each end to the emitter and collector lead-out wires. millions or even hundreds of kilometres per hour. Instead, they diffuse, like a swarm of bees buzzing about. This means that the current carrier (hole or electron) lifetime is critical – they must exist long enough in the base to complete their slow journey across it. It’s this diffusion process that held the key to transistor operation. Lee de Forest, believing that current flow in his “Audion” was solely dependent on gas ions, did not fully understand valve operation and could not capitalise on his invention. It was Thinner bases Irving Langmuir who discovered the For VHF and UHF operation, triode vital need for near-perfect evacuation valves become smaller and smaller, of valve envelopes. with anode-cathode spacings meaLikewise, transistor development sured in tenths of a millimetre or less. did not truly take off until the nature of Audio transistors need base thick- base diffusion was understood. Once nesses of micrometres, some one-­ it was, the principal effort was aimed thousandth of their valve equivalents. at reducing the width of the base juncWhy is this? tion. By the necessity of its microthin Electron flow in valves is driven by base, every transistor is going to be the anode-cathode voltage. As soon as a tiny device compared to its valve an electron escapes the space-charge cousins. cloud around the cathode, that electron is powerfully accelerated by the Conclusion anode-cathode field. All of the manufacturing methods A speed of 300 million kilometres described above are now obsolete. per hour (!) is common, and you may The second article in this series, to see perfectly good receiving valves be published next month, describes with a faint blue glow on the inside of improvements upon these techniques the glass envelope. This is caused by which included alloyed-junction electrons that miss the anode hitting transistors, diffused construction, the envelope so powerfully that they graded doping, base-substrate etchcause the glass to fluoresce. ing, micro-alloy diffusion and all-­ Electrons and holes in the transistor diffusion techniques. do not experience such an accelerating Having explained those, we’ll then field in the base. The base is essentially cover in detail the two transistor manat a constant potential across most of ufacturing methods still in use: mesa its width – there is no powerful field and epitaxial planar, both of which rely SC to accelerate electrons or holes to on photolithography. Base Collector Base region Emitter Fig.20: many grown-junction transistors are made in a single ‘pull’. After the billet is complete (with a thin P-doped layer in the middle), it is sliced into hundreds or thousands of slivers to form individual transistors. After having leads attached, they are encapsulated. Fig.21: a photo of a grown-junction transistor. The base connection wire is very thin since it must connect to the narrow base region in the middle of the slice. Source: David Forbes [CC BY-SA 3.0] Australia's electronics magazine siliconchip.com.au 22 Silicon Chip A Timeline of the Transistor 1873 Willoughby Smith 1948 Mataré & Welker 1956 Abramson & Danko Photoelectric Effect Point-Contact Photolithography He discovered that the electrical resistance of selenium varies with the amount of light falling on it. They independently developed a pointcontact transistor called the “transistron” that was used in France’s telephone network. This early technique was the start of mass PCB fabrication, and involved board lamination and etching. 1874 Karl Braun 1948 John Shive 1957 J. R. A. Beale Diode Detection Bulk Conduction Alloy-Diffused Transistor Braun noted that, when probing a galena crystal with a metal wire, current only flowed freely in one direction. Shive proved that conduction could occur through the bulk of a crystal, paving the way for Shockley’s junction transitor. See video: https://youtu.be/s2H3u-OPSIE Beale reported experimental production with operating frequencies up to 200MHz. 1904 Bose & Pickard 1950 Morgan Sparks 1958 Fred B. Maynard Practical Detectors Grown-Junction Micro-Alloy Diffusion The cat whisker detector was one of the most common early type of semiconductor diode, frequently used in crystal radios. Sparks helped develop the microwatt bipolar junction transistor. A grown-junction transistor can be seen at https://w.wiki/4Yv2 A transistor which employs a base layer with a graded impurity concentration, which is then etched to produce a thin active section. 1925 Julius Lilienfeld 1951 Christensen & Teal 1958 Arthur Varela Field-Effect Principle Epitaxial Fabrication Surface Barrier Lilienfeld filed a patent describing a thin-film device that is now recognised as a precursor to the FET (field-effect transistor). Also called epitaxial deposition, this technique increased both the transistor’s breakdown voltage and switching speed. Varela used chemical etching to create very a thin base structure, with the emitter and collector “plated” into the base wells. 1928 Julius Lilienfeld 1952 William Pfann 1959 Jack Kilby Junction Transistor Zone Refining Integrated Circuit (fabricated) Lilienfeld filed a patent describing a 3-layer device whose structure would be developed by William Shockley as the junction transistor. Also called zone melting, this is a technique used to purify materials and was first used for germanium transistors. Kilby created the first prototype IC, which was a hybrid, not monolithic. A photo of him can be found at: https://w.wiki/4Yvw 1935 Oskar Heil 1952 Pankove & Saby 1959 J. F. Aschner Field-Effect Alloyed-Junction Mesa Transistors Heil discovered the possibility of controlling the resistance of a semiconducting material with an electric field (as in a MOSFET). Alloy-junction transistors were well-suited for mass production, but suffered from poor RF performance. One of these transistors can be seen at https://w.wiki/4YvL Produced by Fairchild Semiconductor, but developed at Bell Labs in 1955. Both base and emitter were diffused, but they still suffered from leakage. 1943 Paul Eisler 1953 Herbert Kroemer 1959 Atalla & Khang Printed Circuitry Drift-Field Transistors The MOSFET Eisler designed a radio in 1942, the first to use a PCB. He was granted a patent for it in 1943. High-speed bipolar junction transistor using graded doping. At Bell Labs, Atalla’s work on oxidising silicon surfaces led (with Khang) to the MOSFET, and to planar transistors and the monolithic IC. 1944 Herbert Mataré 1953 Dacey & Ross 1962 Jean Hoerni Point-Contact Effect Field-Effect Transistor Epitaxial Planar Mataré noticed this effect while developing crystal rectifiers from silicon and germanium during WW2. A working JFET was built by George Dacey and Ian Ross. A photo of them can be found at siliconchip.com.au/link/abcb An oxide layer is left in place on the silicon wafer, reducing leakage. 1947 Bardeen & Brattain 1953 Harwick Johnson 1963 Sah & Wanlass Point-Contact Transistor Monolithic Integrated Circuit CMOS At Bell Labs, these two, led by Shockley, created the first point-contact transistor from germanium. A patent for a phase-shift oscillator fabricated in a single “slice” of semiconductor, which needed no interconnecting wires. CMOS (complementary MOSFET) technology was developed at Fairchild Semiconductor, paving the way for the computer revolution. siliconchip.com.au Australia's electronics magazine March 2022  23