Other articles in the series:
- History of the relay
- History of electronic computers
- History of the transistor
- Internet history
For more than a hundred years, the analog dog has been wagging the digital tail. Attempts to expand the capabilities of our senses - sight, hearing, and even, in a sense, touch, led engineers and scientists to search for the best components for telegraph, telephone, radio and radar. It was only by sheer luck that these searches found the way to the creation of new types of digital machines. And I decided to tell the story of this constant , during which telecommunication engineers supplied the raw materials for the first digital computers, and sometimes even designed and built these computers themselves.
But by the 1960s, that fruitful collaboration came to an end, and with it my story. Digital equipment manufacturers no longer needed to look into the world of telegraph, telephone, and radio for new, improved switches, since the transistor itself provided an inexhaustible source of improvement. Year after year, they dug deeper and deeper, always finding ways to exponentially increase speed and reduce cost.
However, none of this would have happened if the invention of the transistor had stopped at .
Slow start
There was little enthusiasm in the popular press for Bell Labs' announcement of the invention of the transistor. On July 1, 1948, The New York Times gave the event three paragraphs at the bottom of the Radio News report. Moreover, this news came after others that were obviously considered more important: for example, the hour-long radio show “Waltz Time”, which was supposed to appear on NBC. In hindsight, we may want to laugh, or even scold unknown authors - how could they not recognize the event that turned the world upside down?
But hindsight distorts perception, amplifying those signals that we know are important, although at the time they were lost in a sea of noise. The transistor of 1948 was very different from the transistors of computers, on one of which you are reading this article (if you did not decide to print it). They differed so much that, despite the same name, and the unbroken line of inheritance connecting them, they must be considered different species, if not different genera. They have different compositions, different structure, different principle of functioning, not to mention the huge difference in size. It was only through constant reinvention that the clumsy device constructed by Bardeen and Brettain was able to transform the world and our lives.
In fact, the single-point germanium transistor did not deserve more attention than it received. It had several defects inherited from the vacuum tube. He, of course, was much smaller than the most compact lamps. The absence of a hot filament meant that it produced less heat, consumed less energy, did not burn out, and did not need to be warmed up before use.
However, the accumulation of dirt on the contact surface led to failures and negated the potential for longer life; it gave a noisier signal; worked only at low power and in a narrow frequency range; failed in the presence of heat, cold or humidity; and it could not be produced uniformly. Several transistors built in the same way by the same people would have defiantly different electrical characteristics. And all this was accompanied by a cost eight times more than a standard lamp.
It wasn't until 1952 that Bell Laboratories (and other patent holders) solved the production problems enough for single-point transistors to become practical devices, and even then they didn't spread much beyond the hearing aid market, where price sensitivity was relatively low. and the benefits in terms of battery life outweighed the disadvantages.
However, the first attempts to turn the transistor into something better and more useful had already begun. They actually started much earlier than the moment when the public found out about its existence.
Ambition Shockley
Toward the end of 1947 Bill Shockley made a trip to Chicago in great excitement. He had vague ideas about how to surpass the transistor recently invented by Bardeen and Brettain, but he has not yet had the chance to develop them. So instead of enjoying a break between stages of work, he spent Christmas and New Year at the hotel, filling out about 20 pages of a notebook with his ideas. Among them was the proposal of a new transistor, consisting of a semiconductor sandwich - a slice of p-type germanium between two pieces of n-type.
Encouraged by having such an ace up his sleeve, Shockley laid claim to Bardeen and Brattain upon their return to Murray Hill, demanding all the credit for inventing the transistor. Wasn't it his field effect idea that got Bardeen and Brettain into the lab? Shouldn't all rights to the patent be transferred to him because of this? However, Shockley's cunning backfired on him: Bell Labs patent lawyers found out that an unknown inventor, , patented the semiconductor field effect amplifier almost 20 years earlier, in 1930. Lilienfeld, of course, never realized his idea, given the state of materials at that time, but the risk of intersection was too great - it was better to completely avoid mentioning the field effect in patent.
So while Bell Labs gave Shockley a generous share of inventor credit, they only mentioned Bardeen and Brattain in the patent. However, what has been done can not be returned: Shockley's ambitions destroyed his relationship with two subordinates. Bardeen stopped work on the transistor and concentrated on superconductivity. He left the laboratories in 1951. Brettain remained there, but refused to work with Shockley again, and insisted on being transferred to another group.
Due to his inability to work with other people, Shockley never got ahead in the labs, so he left as well. In 1956 he returned home to Palo Alto to found his own transistor manufacturing company, Shockley Semiconductor. Before leaving, he separated from his wife Jean when she was recovering from uterine cancer, and got together with Emmy Lanning, whom he soon married. But of the two halves of his California dream—a new company and a new wife—only one came true. In 1957, his best engineers, angered by his management style and the direction in which he was taking the company, left him to found a new firm, Fairchild Semiconductor.
Shockley in 1956
So Shockley dropped the empty shell of his company and took a job in the electrical engineering department at Stanford. There he continued to alienate his colleagues (and his oldest friend, the physicist ) that interested him in the theories of racial degeneration and - Topics unpopular in the US since the end of the last war, especially in academia. He took pleasure in creating controversy, inflaming the media and causing protests. He died in 1989, estranged from his children and colleagues, and visited only by his eternally devoted second wife, Emmy.
Although his pathetic attempts at entrepreneurship failed, Shockley dropped the grain into fertile soil. The San Francisco Bay Area spawned many small electronics firms that were spiced up with funding from the federal government during the war. Fairchild Semiconductor, Shockley's fortuitous offspring, spawned dozens of new firms, a couple of which are known today: Intel and Advanced Micro Devices (AMD). By the early 1970s, the area had earned the derisive nickname "Silicon Valley." But wait, Bardeen and Brettain created the germanium transistor. Where did silicon come from?
This is what an abandoned place in Mountain View looked like in 2009, where Shockley Semiconductor used to be. Today the building has been demolished.
To the silicon crossroads
The fate of a new type of transistor, invented by Shockley in a Chicago hotel, was much happier than that of its inventor. All thanks to the desire of one person to grow a single pure semiconductor crystals. Gordon Teal, a physical chemist from Texas who was studying then-useless germanium for his PhD, took a job at Bell Labs in the 30s. After learning about the transistor, he became convinced that its reliability and power could be significantly improved by creating it from a pure single crystal, and not from the then used polycrystalline mixtures. Shockley dismissed his attempts as a waste of resources.
However, Teal persevered and succeeded, with the help of mechanical engineer John Little, creating an apparatus that extracts a tiny seed of a crystal from molten germanium. Cooling around the nucleus, germanium expanded its crystal structure, creating a continuous and almost pure semiconducting lattice. By the spring of 1949, Teal and Little were able to create crystals to order, and tests showed that they were far ahead of their polycrystalline competitors. In particular, minor carriers added to them could survive within a hundred microseconds or even longer (against no more than ten microseconds in other crystal samples).
Now Teal could afford more resources, and recruited more people to his team, among them was another physical chemist who came to Bell Labs from Texas - Morgan Sparks. They began to change the melt to make p-type or n-type germanium by adding spherules of appropriate impurities. Within a year, they improved the technology to such an extent that they could grow a germanium npn sandwich directly in the melt. And it worked exactly as Shockley predicted: an electrical signal from the p-type material modulated an electric current between two conductors connected to the surrounding n-type pieces.
Morgan Sparks and Gordon Teal at the workbench at Bell Labs
This grown junction transistor outperforms its single point ancestor in almost every way. In particular, it became more reliable and predictable, produced much less noise (and therefore more sensitive), and extremely energy efficient - consuming a million times less energy than a typical vacuum tube. In July 1951, Bell Labs held another press conference to announce a new invention. Before the first transistor ever even hit the market, it was essentially irrelevant.
And yet it was only the beginning. In 1952, General Electric (GE) announced the development of a new process for making junction transistors, the fusion method. In its framework, two balls of indium (p-type donor) were fused on both sides of a thin slice of n-type germanium. This process was simpler and cheaper than growing junctions in an alloy, such a transistor gave less resistance and supported high frequencies.
Grown and alloy transistors
The following year, Gordon Teal decided to return to his home state and took a job at Texas Instruments (TI) in Dallas. The company was founded as Geophysical Services, Inc., and first made equipment for oil exploration, TI opened an electronics division during the war, and now entered the transistor market under license from Western Electric (Bell Labs' manufacturing division).
Teal brought with him new skills learned in the labs: the ability to grow and silicon single crystals. The most obvious weakness of germanium was its sensitivity to temperature. When exposed to heat, the germanium atoms in the crystal quickly shed their free electrons, and it increasingly became a conductor. At a temperature of 77 ° C, it stopped working altogether, like a transistor. The main target of transistor sales was the military - a potential consumer with low price sensitivity and a huge need for stable, reliable and compact electronic components. However, temperature-sensitive germanium would not be useful in many military applications, especially in aerospace.
Silicon was much more stable, but at the cost of a much higher melting point, comparable to that of steel. This caused enormous difficulties, given that very pure crystals were required to create high-quality transistors. The hot molten silicon would soak up contaminants from whatever crucible it was in. Teal and his team at TI overcame these challenges with DuPont's ultra-pure silicon samples. In May 1954, at a conference of the Institute of Radio Engineers in Dayton, Ohio, Teal demonstrated that the new silicon devices produced in his laboratory continued to work even when immersed in hot oil.
Successful Upstarts
Finally, about seven years after the first invention of the transistor, it could be made from the material with which it had become synonymous. And about the same time will pass before the appearance of transistors, roughly resembling the form that is used in our microprocessors and memory chips.
In 1955, scientists at Bell Labs successfully learned how to make silicon transistors with a new doping technology - instead of adding solid balls of impurities to a liquid melt, they introduced gaseous additives into the solid surface of a semiconductor (). By carefully controlling the temperature, pressure and duration of the procedure, they achieved exactly the required depth and degree of doping. Increased control over the manufacturing process gave greater control over the electrical properties of the final product. More importantly, thermal diffusion made it possible to produce a product in batches - it was possible to dope a large silicon slab, and then cut it into transistors. The military provided funding for Bell Labs, since the organization of production required high upfront costs. They needed a new product for an ultra-high-frequency line for early radar detection (""), a string of arctic radar stations designed to detect Soviet bombers flying in from the North Pole, and they were willing to shell out $100 for a transistor (those were the days when a new car could be bought for $2000).
alloying together with , which controlled the arrangement of impurities, opened up the possibility of etching the entire circuit entirely on a single semiconductor substrate - this was thought up at the same time at Fairchild Semiconductor and Texas Instruments in 1959. "» from Fairchild used chemical deposition of metal films connecting the electrical contacts of the transistor. It eliminated the need for manual wiring, reduced production costs, and increased reliability.
Finally, in 1960, two engineers from Bell Labs (John Atalla and Daewon Kahn) implemented the original concept of Shockley field effect transistor. A thin layer of oxide on the surface of the semiconductor was able to effectively suppress surface states, as a result of which the electric field from the aluminum gate penetrated into the silicon. Thus was born MOSFET [metal-oxide semiconductor field-effect transistor] (or MOS structure, from metal-oxide-semiconductor), which turned out to be so easy to miniaturize, and which is still used in almost all modern computers (interestingly, Atalla was originally from Egypt, and Kang from South Korea, and practically only these two engineers from our entire history do not have European roots).
Finally, thirteen years after the invention of the first transistor, something appeared that resembled the transistor in your computer. It was easier to manufacture, it used less energy than a junction transistor, but it was rather slow in responding to signals. It wasn't until the proliferation of large integrated circuits with hundreds or thousands of components on a single chip that the advantages of FETs came to the fore.
Illustration from a patent for a field-effect transistor
The field effect was Bell Labs' last major contribution to the development of the transistor. Major electronics manufacturers such as Bell Laboratories (with their Western Electric), General Electric, Sylvania, and Westinghouse have built up an impressive body of semiconductor research. From 1952 to 1965, Bell Laboratories alone filed over two hundred patents on the subject. Yet the commercial market was quickly taken over by new players such as Texas Instruments, Transitron, and Fairchild.
The early transistor market was too small for the big players to notice: about $18 million a year in the mid-1950s, compared to a total electronics market of $2 billion. However, the research labs of these giants served as unintentional training camps where young scientists could absorb the knowledge about semiconductors and then move on to selling their services to smaller firms. When the vacuum tube market began to shrink seriously in the mid-1960s, it was too late for Bell Labs, Westinghouse and others to compete with the upstarts.
The transition of computers to transistors
In the 1950s, transistors invaded the world of electronics in four major areas. The first two were hearing aids and portable radios, where low power consumption and consequent long battery life overrode other considerations. The third was military use. The US Army had high hopes for transistors as rugged, compact components that could be used in everything from field radios to ballistic missiles. However, at first, their spending on transistors was more like a bet on the future of technology than a confirmation of their then value. And finally, there were digital calculations.
In the computer field, the shortcomings of vacuum tube switches were well known, with some skeptics before the war even believing that the electronic computer could not be made a practical device. When thousands of lamps were assembled in one device, they devoured electricity, giving out a huge amount of heat, and in terms of reliability, one could only rely on their regular burnout. Therefore, the low-power, cold, and filamentless transistor became the savior of computer manufacturers. Its shortcomings as an amplifier (such as a noisier output) were not such a problem when used as a switch. The only obstacle was the cost, and in due time it will begin to fall sharply.
All of America's early experimentation with transistorized computers occurred at the intersection of the military's desire to explore the potential of a promising new technology, and the engineers' desire to upgrade to better switches.
Bell Labs built the TRADIC for the US Air Force in 1954 to see if transistors would make it possible to install a digital computer on board a bomber, replacing analog navigation and targeting assistance. MIT's Lincoln Lab developed the TX-0 computer as part of an extensive air defense project in 1956. The machine used another version of the transistor, the surface barrier, well suited for high-speed computing. Philco built its SOLO computer under a contract from the Navy (but actually at the request of the NSA), finishing it in 1958 (using another version of the surface barrier transistor).
In Western Europe, under-resourced during the Cold War, the story was very different. Machines such as the Manchester Transistor Computer, (another name inspired by the ENIAC project and spelled backwards), and the Austrian were side projects that used resources that their creators could scrape together - including first-generation single-point transistors.
There is a lot of controversy over the title of the first computer to use transistors. Everything, of course, rests on the choice of the correct definitions of such words as "first", "transistor" and "computer". In any case, we know where the story ends. The commercialization of transistorized computers began almost immediately. Year after year, computers for the same price became more powerful, and computers of the same power became cheaper, and this process seemed so inexorable that it was elevated to the rank of law, next to gravity and the conservation of energy. Do we need to argue about which pebble was the first in the collapse?
Where did Moore's law come from?
As we near the end of the history of the switch, it is worth asking the question: what caused this crash to occur? Why does Moore's law exist (or existed - we will argue about this another time)? There is no Moore's law for airplanes or vacuum cleaners, just as there is none for vacuum tubes or relays.
The answer has two parts:
- Boolean properties of a switch as an artifact category.
- Ability to use purely chemical processes for the manufacture of transistors.
First, about the essence of the switch. The properties of most artifacts must satisfy a wide range of inexorable physical constraints. A passenger plane must support the combined weight of many people. The vacuum cleaner must be able to suck in a certain amount of dirt in a certain time from a certain physical area. Airplanes and vacuum cleaners would be useless if reduced to the nanoscale.
A switch, on the other hand, an automatic switch that has never been touched by a human hand, has far fewer physical limitations. It must have two different states, and it must be able to tell other similar switches to change their states. That is, all he has to be able to do is turn on and off. What is so special about transistors? Why haven't other kinds of digital switches experienced such exponential improvements?
Here we come to the second fact. Transistors can be manufactured by chemical processes without mechanical intervention. From the very beginning, the use of chemical impurities has been a key element in the production of transistors. Then came the planar process, which eliminated the last mechanical step from production—connecting wires. As a result, he got rid of the last physical limitation on miniaturization. Transistors no longer needed to be large enough for human fingers - or for any mechanical device. It was all done by simple chemistry, on an unimaginably small scale: acid to etch, light to control which parts of the surface would resist etching, and vapors to inject impurities and metal films into the etched tracks.
Why is miniaturization needed at all? Reducing the size gave a whole galaxy of pleasant side effects: increased switching speed, reduced power consumption and cost of individual instances. These powerful incentives have spurred everyone on to look for ways to further reduce switches. And the semiconductor industry has gone from making fingernail-sized switches to packing tens of millions of switches per square millimeter in the lifetime of one man. From asking eight dollars for one switch to offering twenty million switches for a dollar.
Intel 1103 memory chip from 1971. Individual transistors, only tens of micrometers in size, are already indistinguishable by the eye. And since then they have decreased by a thousand times.
What else to read:
- Ernest Bruan and Stuart MacDonald, Revolution in Miniature (1978)
- Michael Riordan and Lillian Hoddeson, Crystal Fire (1997)
- Joel Shurkin, Broken Genius (1997)
Source: habr.com