v1.0.1 / chapter 10 of 20 / 01 sep 07 / greg goebel / public domain
* Although the development of nuclear power changed the world in a drastic way, quantum physics also led to an understanding of the "solid state" of matter, and revolution in electronics technology whose impact has been in many ways at least as great.
* As discussed in earlier installments, quantum physics was quickly applied to the issues of chemical bonding and molecular formation. Of course, researchers were also interested in the applications of quantum physics to the behavior of materials at larger scales, in the form of orderly crystals or disorderly amorphous materials, such as glasses.
World War II led to the application of quantum physics to nuclear weapons, but the conflict also led to the development of advanced electronic technologies, particularly radar. These devices were built around triode (three-plate) vacuum tube technologies, but they also incorporated a few "solid state" electronic components. Following up research conducted in the 1870s, by the early 1900s solid-state "crystal rectifiers" were in increasing use. These devices consisted of a metal filament or "whisker" pressed up against a crystal of a material such as galena; electric current would flow through in one direction and not another. Such "cat's whisker" rectifiers were used in early radios to extract the sound signal out of a radio transmission. They were gradually eclipsed by the "vacuum diode", a simple two-plate vacuum tube, which was much more reliable.
Radar demanded higher frequencies and greater sensitivity than a simple radio receiver. Vacuum diodes couldn't handle high frequency signals very well and had to be heated to operate, making them electrically noisy, and the faster, cooler and less noisy cat's whisker rectifiers made a comeback. There were major problems with them, however. Such devices were a bit tricky to produce, since given any lot of them their performance varied wildly; they had to be manufactured by moving the cat's whisker around to find a "hot spot" on the crystal; and they were fragile, easily broken by shock or burned out by high currents, which is why they had gone out of style for a time. In addition, nobody had a clear idea of how they worked.
By the time the US entered the war in late 1941, a number of American organizations were performing research on crystal rectifiers, most prominently Bell Laboratories. Some groups focused on theoretical work, while others on practical experiments, evaluating a large number of materials to see how well they worked as solid-state diodes. Silicon emerged as the most promising material, though there were other interesting prospects. Pure crystals were fabricated and then "doped" with various impurities to see how they affected device performance; boron dopant did much to improve the capabilities of the cat's whisker rectifier.
A great deal of money and effort was placed into learning even this much, and it seems like relatively little to show for all the expense. However, it would pay off very well in the postwar period. Much had been learned about characterizing and manipulating silicon and other materials, and there were researchers who thought that much more could be made of solid-state devices. Such avenues of investigation were too speculative to be seriously pursued during the war in the absence of any obvious short-term payoff for doing so, but as the war's end approached, Bell Labs started up a major research effort into advanced solid-state electronics.
The result was the "transistor", a solid-state replacement for the vacuum tube. It was invented by three Bell Labs researchers -- William Shockley (1910:1989), John Bardeen (1908:1991), and Walter Brattain (1902:1987) -- who got their first transistor to work in late 1947. They would share the Nobel prize in 1956 for their work.
Their first transistor was an impractical lab curiosity; it was unreliable, erratic and mysterious in operation, and difficult or impossible to manufacture in volume. By 1951, Shockley had refined the design into a practical device that could in principle be built in quantity, and he formed his own commercial company to capitalize on the technology. Transistor sales remained low for several years since initially they were over ten times more expensive than triodes, but by the middle of the decade, prices had dropped to the level where the transistor was beginning to push vacuum tubes off to one side.
Shockley was brilliant but some found him obnoxious to work for, even crankish -- late in his life he would all but deliberately raise a storm of public controversy by publishing statistics studies to prove that blacks were inferior to whites at some mental tasks -- and so in 1957 eight of his researchers dropped out to form their own company, Fairchild Semiconductor, which would become the leading edge of a solid-state electronic revolution.
In 1958, one of the "Traitorous Eight", as Shockley called them, a Swiss-born physicist named Jean Hoerni (1924:1997), came up with a much better "planar" method for fabricating transistors. The transistor had finally grown up, but the planar method was only a beginning. In 1959 another one of the Traitorous Eight, Robert Noyce (1927:1990), used planar techniques to fabricate several electronic devices on a single silicon crystal.
The Fairchild "integrated circuit (IC)" wasn't entirely news. The basic idea had been floating around for several years, and earlier in 1959 another electronics firm named Texas Instruments (TI) had announced an IC, developed by TI researcher Jack Saint Clair Kilby (1923:2005). However, Kilby's IC had used much more cumbersome fabrication methods, including some unbelievably painstaking hand wiring, and the Fairchild approach was far more practical. In any case, both Fairchild and TI moved ahead rapidly and were producing ICs in quantity by 1962. From that time, the size of devices on an IC repeatedly halved rapidly and the number of devices on a IC doubled accordingly, and now ICs are routinely produced with millions, even tens of millions of devices on one chip.
Bob Noyce and Jack Kilby would share honors as the founding fathers of IC technology, in some places almost revered as demigods. Noyce would help form up a new company, Intel, later in the 1960s. In 2000, Kilby would share the Nobel prize in physics with two other semiconductor researchers, the German-American physicist Herbert Kroemer (born 1928) and the Russian physicist Zhores I. Alferov (born 1930), both of whom had made major contributions to the scientific foundation of solid-state electronics.
* Design of solid-state electronic devices leveraged off theoretical studies of the solid state using the new quantum physics. Solid state physics focused initially on crystalline materials, since they were more orderly and easy to analyze than amorphous materials. A knowledge of chemical bonding rules led to the understanding of the different classes of crystals, such as face-centered cubic, body-centered cubic, and other organizations, but more importantly for present purposes led to an understanding of the electronic properties of materials.
As discussed earlier, in individual atoms electrons can only occupy certain given energy states, with the set of energy states unique to each element. In addition, electrons, being half-spin particles or fermions, obey the Pauli exclusion principle, meaning that no two electrons in the same quantum system can have the same energy state. This has interesting implications when a set of such atoms is brought together in a crystal. Since the crystal forms one big quantum system, electrons that are at the same energy state when they are components of individual atoms can no longer be at the same energy state. The result is that what was a single energy state in individual atoms becomes a broad range or "band" of energy states in the crystal, with electrons shifting up or down in energy to allow them to coexist. The energy bands of the crystal may overlap, or they may be separated by "forbidden regions" or "bandgaps" where there are no allowed energy states. It is the spacing of the energy bands and the bandgaps that help explain the electronic behavior of the solid.
At absolute zero, the electrons in the crystal will all be in their minimum energy states. The maximum occupied energy state in this circumstance is called the "Fermi level" or "Fermi state". If the temperature is raised, electrons will move to higher energy states -- if they are able. If the Fermi level lies in the middle of an energy band, then there is no obstacle to increasing the energy level of the electrons. A material with such an arrangement is highly conductive, in both the thermal and electronic sense. Such a crystal tends to absorb light, since the electrons can be excited by incident photons, and so are generally opaque. Good examples of crystalline "conductors" are metals such as copper, gold, silver, and aluminum.
Incidentally, if two different metals with different Fermi levels are placed in contact, the metal with the higher Fermi level will dump electrons into the empty states of the metal with the lower Fermi level, or in other words there's a "contact potential" or "contact voltage" between the junctions. A familiar example is the unpleasant little electrical charge that flows when a tang of a metal fork hits a metal filling in a tooth. This phenomenon can't be used as a "perpetual motion" class machine, since if the dissimilar metals are connected to an electrical circuit, electrons are displaced through the circuit until the charge imbalance sets up a reverse potential that stops the flow. However, in a storage battery the current can be driven through a conductive "electrolyte", usually a strong acid like sulfuric acid or strong base like potassium hydroxide, resulting in chemical reactions between the metals and the electrolyte that allow current to flow until the reactions go to completion. Dissimilar metals are also used for temperature measurement, since the potential increases with temperature, and a wide variety of "thermocouples" with different properties are available.
If the Fermi level is at the very top of an energy band, the electrons in this "valence band", so named because it is the outermost or valence electrons that occupy this range of states, have to be given enough energy to jump up over the bandgap into the neighboring higher "conduction band" in order for them to become mobile, or in other words act as "charge carriers". If the bandgap is large, electrons tend to stay in the valence band, and the crystal has poor thermal and electronic conductivity. It also tends to not absorb light and be transparent. Good examples of crystalline "insulators" are quartz and diamond. However, if the bandgap is relatively small, the material has properties intermediate between those of a conductor and an insulator. Such materials are "semiconductors", with the best-known example of course being silicon. The crystals used in the old cat's whisker rectifiers were all semiconductors.
* While quantum physics often seems mystical and impractical, the quantum physics of semiconductors is of great practical importance, and in fact it could be thought of as the most widespread application of quantum physics. Since silicon is the most familiar semiconductor material, this discussion will focus on silicon and discuss alternate materials later.
Pure or "intrinsic" silicon isn't all that useful in itself. Heat and light will cause electrons to jump from the valence band to the conduction band, and so intrinsic silicon can be used for temperature or light sensors. Its behavior becomes more interesting if impurities are added to the crystal structure of silicon.
Silicon has a valence of four, or in other words it typically forms four bonds. In a silicon crystal, these are purely covalent bonds, with each silicon atom linked to four neighbors in a tetrahedral structure, though the structure can be modeled more simply if less accurately in a two-dimensional structure that consists of a nice rectangular grid of silicon atoms.
Now suppose that a trace of phosphorus atoms is introduced or "doped" into this nice neat crystal matrix. Phosphorus has a valence of five, so a phosphorus atom introduce a "dangling" bond into the matrix. The end result is that this frees up electrons to act as charge carriers, and the conductivity of the silicon is greatly increased. From the more abstract point of view of band theory, the presence of the phosphorus introduces a new set of states in the bandgap, just below the conduction band.
Next, suppose a trace of boron atoms are introduced or doped into the silicon matrix instead of phosphorus atoms. Boron has a valence of three, and so its introduction into the matrix leaves silicon bonds dangling. This also increases the conductivity of the silicon. From the point of view of band theory, the presence of the boron introduces a new set of states in the bandgap, just above the valence band.
On a per-atom basis, boron doesn't contribute as much to conductivity as much as phosphorus does. Imagine as a very rough analogy the intrinsic silicon matrix as a flat box loaded with billiard balls that are packed so tightly that they can't move around. Adding phosphorus to the matrix is like throwing more balls on top of the box, and these new balls can move around very easily.
Adding boron to the matrix is, in contrast, like removing some of the balls from the box; the remaining balls can shift around, but not as efficiently. Interestingly, what appears to be moving around in this case is not the balls themselves but the "holes" left when the balls were removed, and so the charge carriers in silicon doped with boron are called "holes". This is obviously less efficient and so the "charge carrier mobility" is lower. This analogy can't be pushed too far: it's strictly mechanical, while in the case of doped silicon what's being shuffled around are electrons and available energy states.
In general, valence-five atoms like phosphorus are called "donor" impurities, since they "donate" electrons, and the resulting altered silicon matrix is known as "N-type" silicon. Valence-three atoms like boron are called "acceptor" impurities, since they "accept" electrons (creating holes), and the resulting silicon matrix is known as "p-type" silicon. It must be emphasized that in either case, the matrix is electrically neutral: the number of protons equals the number of electrons. The only thing that has been done is to tinker with the crystal structure to improve the mobility of charge carriers.
This sort of tinkering with charge carriers can be used to build an electronic switch. Dopants can be introduced into a silicon crystal by exposing the crystal to a hot gas containing the dopant; a "mask" of silicon oxide or "oxide", a form of insulating glass laid down on top of the crystal, can be patterned and chemically etched in a process conceptually similar to silkscreening to ensure that the doping is only introduced to selective exposed regions of the crystal. The length of time of the exposure determines the level of concentration of dopants and their depth of penetration. The oxide layer can be stripped off, redeposited, masked, and etched to create several different patterns on the crystal. This is essentially the planar fabrication method invented by Jean Hoerni.
Suppose two isolated regions in a p-type crystal are doped to be highly conductive using an N-type dopant like phosphorus, with the two regions set closely apart, the gap between them being known as the "channel". The source and drain are mated to metal connections that are part of a electric output circuit, while the channel is covered with a layer of oxide covered in turn with a tiny plate of metal, with this plate connected to an input electric circuit. If a negative voltage is placed on this plate, known as "gate", it will impede the flow of electrons through the channel, shutting off the flow of current between source and drain. If a positive voltage is placed on the gate, it will allow electrons to flow freely between the source and drain.
This electronic switch is a "metal-oxide-semiconductor field effect transistor (MOSFET)", and most integrated circuits contain thousands, even millions of them. It is a conceptually simple device, based on the manipulation of the conductivity of silicon through selective doping. Its operational characteristics are surprisingly similar to those of the old vacuum triodes.
Incidentally, P-type doping could have been used instead of N-type doping, though in that case a negative voltage would turn the switch off and a positive voltage would turn the switch on. Although the charge carrier mobility of P-type devices is lower than that of N-type device and so a P-type MOSFET is slower than an N-type MOSFET, building a circuit with both types of MOSFET, one turning off while the other turns on, has advantages in reducing the power draw of the switch, and such "complementary MOS (CMOS)" devices are an effective standard for digital electronic systems at present.
* Although CMOS circuits have made modern computers possible, from the point of view of physics, more interesting things can be done by mating two pieces of doped silicon. Suppose half of a single crystal of silicon is doped with boron, making it P-type, and the other half is doped with phosphorus, making it N-type. The result is that a "PN junction" is formed between the two halves. The N-type silicon has an excess of electrons, while the P-type silicon has an excess of holes. The electrons on the N-type silicon side of the crystal migrate into the P-type silicon side of the crystal, depleting it of electrons or in effect leaving behind holes. Once this happens, the crystal is no longer electrically neutral, or at least the regions around the PN junction aren't: it's got a net surplus of electrons on the P side, making that region negatively charged, and a net deficit of electrons (or a surplus of holes) on the N side.
This creates a "junction potential" Vj of about 0.7 volts in silicon. Just to make things even more confusing, the migration of charge carriers across both sides of the PN junction depletes the region around the junction of free charge carriers, and so it is known as the "depletion region".
Now imagine what happens if a negative voltage is applied to the N-type region and a positive voltage is applied to the P-type region. This will drive more charge carriers across the PN junction, which will result in an even bigger depletion region and a greater potential that will block current flow. However, if the polarity of the voltage is reversed, with a positive voltage applied to the N-type region and a negative voltage applied to the P-type region, charge carriers will flow out of the silicon crystal and a current will flow. The PN junction acts as a one-way valve for electric current, or in other words a "rectifier" or "diode".
If the reverse voltage is increased enough, the electrical potential will be great enough to tear loose charge carriers even through the junction potential, resulting in "breakdown" operation. This can actually be useful, since the voltage across the junction remains much the same over a wide range of current, and a diode can be designed to breakdown at a specific voltage, Vbr, to act as a voltage reference. Such devices are known as "Zener diodes".
It is possible to build a transistor using a crystal with three regions, either N-P-N or P-N-P, and two junctions, and in fact such "bipolar junction transistors (BJTs)" well preceded MOSFETs. However, explaining the operation of a BJT is more complicated than explaining the operation of a MOSFET, and really more appropriate for a document on solid-state electronics. The BJT is only mentioned here as a lead for further study elsewhere.
Although this discussion specifically focuses on the physics of silicon, it also applies to other semiconductors used in electronics. Germanium, another valence-four atom, was once used in power devices because it had some features that allowed it to handle high power levels. Gallium arsenide, a compound of gallium and arsenic, is often used in very fast circuits because of its higher charge carrier mobility. Gallium arsenide is known as a "III-V" compound, since gallium is a valence-three atom while arsenic is a valence-five atom; the even combination of the two results in a crystal matrix with an overall valence of four. It also should be noted that N-type valence-five dopants include not only phosphorus, but also antimony and of course arsenic, while P-type dopants include not only boron, but also aluminum, indium, and of course gallium. Various mixtures of III-V compounds often show up in optoelectronic devices.
* Incidentally, a rectifying junction can be formed between a metal and an N-type semiconductor; electrons can migrate from the metal into the semiconductor easily, but can't move back so easily. Such a device is known as a "Schottky diode". It's leakier but faster than a conventional silicon diode, since the junction stores less charge and so takes less time to "drain". Care has to be exercised in the design of integrated circuits to make sure that the final layer of metal contacts laid down on the semiconductor chip don't form unintended and unwanted rectifying junctions.
Of course, the old cat's whisker rectifiers were metal-semiconductor devices. While it seems possible to have built such a thing by trial-and-error, it is still hard to believe that such a thing could be reliably manufactured in volume when nobody had a clear idea of how it worked, and it is likely the scrap rate off the end of the production line was high.
* The PN junction has a number of other interesting properties. For example, if a wide flat semiconductor crystal is doped to form a PN junction just under the surface and covered with a thin metallic connection layer on top, then if the crystal is exposed to light, the photons in the light will kick electrons up across the band gap and cause an electric current to flow. This is the basic principle of the "photovoltaic (PV)" cell or "solar power cell". The size of the bandgap is related to the wavelengths of photons a PV cell can absorb, since the photon must have enough energy to kick an electron up across the bandgap, and photon energy increases as wavelengths decrease. Any excess energy is wasted. For this reason, some high-efficiency PV cells have been built that have multiple PN junctions, stacked on top of each other, with the bandgap decreasing with lower junctions to soak up as wide a band of light as possible.
Similarly, if a current is passed through a diode, then if this diode is properly designed the recombination of charge carriers as they pass through the PN junction will release light. This is the basis of the "light emitting diode (LED)". Silicon can't be used for LEDs, since its bandgap is too small to generate photons of useful wavelengths, though solid-state researchers are working on tricks to see if silicon can be coaxed into acting as a practical light-emitter.
Gallium arsenide, often with traces of other III-V compounds, is a common material for building red LEDs, with careful manipulation of types and quantities of dopants giving different colors of light. From the early 1990s, gallium nitride has been used to build blue-light LEDs. Work is currently underway towards zinc-oxide LEDs that operate in the ultraviolet. A UV LED is regarded as ideal for LED applications in lighting, Since UV is more energetic than visible light, it can be used to excite red, green, and blue light-emitting phosphor materials, with the proportion of phosphors adjusted to give any desired color.
* This discussion has assumed semiconductors in the form of orderly crystals, but it is also possible to build semiconductor devices from amorphous semiconductors. Amorphous semiconductors do possess a band structure along the lines of those of their crystalline counterparts, but their random nature means less predictable properties and in particular results in low charge-carrier mobility, a thousandth of that of crystalline semiconductors, and so low speeds, in much the same way that it is easier to walk through the neat rows of a tree farm than struggle through the brush in a forest. However, amorphous devices are much easier and cheaper to fabricate than crystalline devices, and low-cost amorphous silicon PV panels are now sold in developing countries to provide household power.
There has also been work on developing synthetic polymers with semiconductor properties by paying close attention to the polymer structure. Performance of such materials is also low, but the idea of building "plastic transistors" and other electronic components is very appealing, since simple electronic systems using plastic components would be very cheap, even disposable, and could be laid down on almost any substrate, such as the wall of a cardboard box.
* In the decades following World War II, research into the solid state expanded in range, covering not merely crystalline substances but also amorphous materials such as glasses, as well as liquids. Obviously, the term "solid state physics" was hard to apply to liquids, and so a new umbrella term, "condensed matter physics", was invented.
Along with the studies of semiconductor physics, condensed matter physics also covers studies of materials known as "superconductors" that lose their electrical resistance when cooled below a certain temperature, as well as "masers" that produce a tightly focused microwave beam and similar "lasers" that produce a tightly focused light beam. These topics are discussed in the following chapters.
