What is mos technology?

MOS manufacturing technology can make tens of thousands of electronic components on a chip with only a few square millimeters. This ic circuit has been widely used in sleeve diagnostic calculator.

It has been 25 years since the invention of the transistor. Because of this epoch-making contribution, electronic products have penetrated into the whole human life and become a very popular thing. Last year1February, the American electronics industry also held a 25th anniversary meeting to celebrate this great contribution of replacing vacuum tubes with solid substances. Looking back at the development history of semiconductor electronic parts, we find that until 1960, electronic instruments were still connected one by one with transistors the size of an eraser on a pencil (or larger), and the average price of each transistor was as high as one dollar. In the late 1960s, scientists began to design various methods to make part or a whole set of electronic circuits on a single silicon wafer, which is the so-called integrated circuit universal integrated circuit. Early integrated circuits can only contain about a dozen electronic components on a chip with a few square millimeters, but today's mass-produced integrated circuits already contain about 3000 electronic components, most of which are transistors. At present, there are as many as ten thousand transistors in some advanced integrated circuits. It seems that we can expect that integrated circuits containing millions of transistor elements will appear in the 1980s [Note 1].

Traditional transistors are called bipolar transistors. Due to the inherent limitations of this transistor production process, it is difficult for us to produce integrated circuits with high component density on one chip. Therefore, the so-called LSI (Large Scale Integrated Circuit) is currently manufactured by the MOS method. The so-called MOS is the abbreviation of metal oxide semiconductor (see Figure 4). With this technology, integrated circuits can be made smaller and contain more components. Moreover, the manufacturing process of MOS is simpler than that of traditional transistors. As we all know, if a product can compete in the market, it is nothing more than high quality and low price, and the production process of MOS is simple, so the cost is low. The average price of each transistor on an LSI with 200 transistors is only one dollar, but it is generally believed that the average price of each transistor can be reduced by 30 times in ten years, and then the price of each transistor will be as cheap as the words printed on paper (see Figure 3). In addition, the improvement of reliability and the reduction of volume and weight of integrated circuits are also one of the reasons why integrated circuits are generally valued and liked. Of course, these reasons are insignificant compared with the cost reduction.

To reduce the cost, only mass production, that is, the manufacture of integrated circuits adopts mass production. The general method is to make many chips go through a series of chemical and metallurgical treatments at the same time, and then through photolithography, diffusion and other procedures, often hundreds of integrated circuits can be produced on each chip. However, although scientists try their best to keep the properties of each wafer the same, and even pay attention to this problem in every weekly procedure, the properties of each wafer can never keep a certain specification, and there are often many defects and dislocations on each wafer, or some unnecessary substances are attached to the surface of the wafer; Moreover, because the precision in integrated circuits is in microns (10-4cm), a defect invisible to the naked eye will often destroy the characteristics of the whole integrated circuit, so some mass-produced ICs mentioned above are often eliminated after quality inspection, so the so-called "yield" problem will appear in IC manufacturing. The success rate of an advanced integrated circuit is often very low in the initial stage of production, but it can be quickly improved from the experience of groping experiments. In recent years, due to the improvement of precision measurement and control instruments, the manufacturing process of integrated circuits has been greatly improved. Therefore, scientists can make more complex integrated circuits. Of course, the more complex and dense the integrated circuits, the lower the success rate, so unless there is a big breakthrough in the manufacturing process, the cost reduction will always reach a certain limit.

The origin of transistor

Transistors on MOS integrated circuits are field effect transistors, commonly called FET (field effect transistor). Its working principle is to apply an electric field in the direction perpendicular to the chip surface to control the conductance between the source and the drain. In fact, this effect was discovered by Edgar Schmidt as early as 1930 (he obtained the patent of field effect device in 1935), but at that time, due to the lack of physical knowledge of crystal surface and thin film, field effect device could not be made, and at that time, because most scientists were committed to the research and development of vacuum tubes, the theory of field effect was also lacking.

At the end of 1930, William B. shockley, a young physicist working in Bell Laboratories, became interested in the possibility of manufacturing electronic components with solid materials. At first, he devoted himself to developing a solid electronic component as a telephone-to-telephone switching system to replace the traditional electromechanical switch. Shackley and some advanced people are convinced that there will be a great demand for telephone exchanges in the near future. If you still use vacuum tubes, it will be uneconomical, and the reliability of vacuum tubes is very low. In Walter Schottky's article on the phenomenon of rectifying the interface between metal and semiconductor (alternating current to direct current), Shackley found that we can use the change of the width of space charge region in semiconductor to amplify the signal (see Figure 2). He is convinced that this space charge region can be used to control the conductance in the semiconductor like a switch valve, so as to achieve the effect of controlling the current between the two electrodes, which is very similar to the principle that a vacuum tube uses the gate voltage to control the current between the two electrodes. 1939, Xue Keli tried to trial-produce this electronic component with copper and copper oxide, but unfortunately it was unsuccessful.

After World War II, Shackley returned to work in Bell Laboratories. He, and Walter H. Blatin) [30] began to study the field effect amplification in germanium (Ge) semiconductor (because the physical properties of germanium were much better than copper oxide at that time). Their research on semiconductor surface contact and space charged area finally invented the "point contact transistor" in 1947. Although this kind of point contact transistor can't be mass-produced, they have confirmed the idea of making electronic components with semiconductors anyway, and the rest seems to be only technical problems. Sure enough, the junction transistor was made in 1948. A junction transistor or bipolar transistor has two junctions; These two junctions divide the semiconductor into three regions, namely emitter, base and collector. The current flowing from the emitter to the collector can be controlled by the tiny signal of the base, so it has the function of signal amplification.

Although the invention of the transistor excited the scientific community for some time, under the leadership of Shackley, the interest of Bell Laboratories scientists in field effect has not decreased at all. 1948 Gerald L. Pearson and Shackley discovered the field effect phenomenon in the pn junction of silicon wafers (Note 4), 1952 Shackley published the field effect transistor theory. In the second year (1953), george C. Dacey and M. Ross designed the field effect transistor, but the field effect transistor at that time used electric field to control the conduction phenomenon in Ge. This kind of field effect transistor is only used in some special occasions because of its high price and limited advantages over ordinary transistors.

Scientists have found that silicon has high stability to temperature and is easy to control in manufacturing, so the cost is low. After 1950, Si gradually replaced Ge as the material of transistor. Scientists have made rapid progress in the study of silicon crystal surface, and the manufacturing technology of components is also changing with each passing day; Therefore, the interface phenomenon between silicon and silicon dioxide is gradually understood and can be controlled, and the stability of manufactured electronic components is getting higher and higher. Dawon Kahung and John Atalla of Bell Laboratories use an insulated electrode (they call it a gate) to induce a conductive channel between the P-N junctions in 1960 to control the conductive state in the crystal. According to this idea, two years later, Stephen Hough R. Hofstein and Frederick P. Heiman of American Radio Company finally designed the field effect transistor (FET). Its structure is as follows: N-type or P-type impurities are introduced into two different places on the silicon wafer as the source and drain, a layer of silicon dioxide insulator is grown on the silicon wafer between the two electrodes, and then a layer of metal is plated on the silicon dioxide as the gate. From the longitudinal section, its structure is metal oxide semiconductor, so it is called MOS transistor.

We take N-type semiconductor as an example to illustrate the working principle of MOS. When a voltage is applied between the source and the drain, whether the conduction between them is good or not can be determined by the amount of charge in the channel, and the charge in the channel can be induced by the voltage of the gate. From the knowledge of electromagnetism, we all know that if the grid is charged, the semiconductor under the grid will cause some charges with opposite signs, which can form a so-called channel. The width of this channel (that is, the amount of charge caused) is directly proportional to the voltage of the gate, so we can use the voltage of the gate to control the current between the source and the drain. In fact, if the voltage applied to the gate does not exceed the so-called threshold voltage, the conductance between the source and the drain is still very small, but once the threshold voltage is exceeded, the conductance increases sharply, so the current between them increases sharply. The voltage on the gate of the N-type semiconductor is negative, so the induced charge is positive [Note 5]. This channel is called a P-channel enhancement transistor. If the semiconductor is p-type and its source and drain are n-type, then the voltage on the gate should be positive and the induced charge should be negative. The transistor at this time is called N-channel enhancement transistor. There is also a FET with the same structure as above, except that there is a charged channel between the source and drain when the gate voltage is zero (the charge of this channel is the same as that of the source and drain). When the gate voltage is applied, the charge in the channel decreases (for example, it was originally an N-channel, and the number of electrons in the channel decreases due to the electric field after the negative voltage is applied), so the current between the two electrodes is the largest when the gate voltage is zero, and the boosting current decreases. According to the different channel charges, this kind of transistor is called N-channel depletion transistor and P-channel depletion transistor. However, in practical application, the enhanced field effect transistor has great plasticity, so it is mostly used in circuits.

metal oxide semiconductor transistor

As we said before, the manufacturing process of MOS transistors is much simpler than that of traditional transistors. Therefore, manufacturing MOS integrated circuits is much simpler and easier than using old transistor integrated circuits. Take an ordinary inverter as an example. If a junction transistor is used, it needs four different diffusion steps and six sets of shields. (Note: Please refer to the article "Ion implantation technology" in Volume IV of York for the function of screen. However, if MOS transistors are used, only one diffusion step and five sets of screens are needed. It is precisely because of the above advantages and low cost that the research of MOS has been widely concerned since 1960. Scientists have spent several years studying and solving the instability of the interface between silicon wafer and silicon oxide and the characteristics of silicon oxide itself. In the past six years, MOS integrated circuits have grown from scratch to 48 million this year, with a total value of $250 million. It is estimated that there will be as many as 400 million bipolar transistor integrated circuits (with a total value of 720 million US dollars) this year. Readers can find from the above figures that the growth rate of MOS integrated circuits is quite amazing.

Like a vacuum tube, MOS uses voltage to control current, with high input impedance and quite linear ratio of output to input. The junction transistor is controlled by current, so its characteristics are not as linear as MOS, and its input impedance is much smaller than MOS. Secondly, under both conductive and non-conductive conditions, MOS consumes much less energy than junction transistors. But so far, the working speed of the MOS transistor we made is not as fast as that of the general transistor. However, this speed difference is mainly due to the immature manufacturing process of MOS, not the theoretical limitation of MOS itself. From the current situation, because of their respective advantages and disadvantages, engineers who design instruments often hesitate between the two, but I personally think that MOS will play a more important role in digital electronic circuits in the late 1970s.

At present, hundreds of MOS integrated circuits are used in desktop calculators and various electronic devices, including the simplest logic circuits to integrated circuits containing memory cells and logic. In addition to the need for high-speed electronic computers, almost all new electronic devices have some MOS lines in them.

MOS calculator

Probably the biggest commercial application of MOS is desktop calculator and pocket calculator. Before MOS application, desktop calculators were mostly designed with electromechanical parts, so the cost of each calculator was about 500 to 1000 USD. Later, after the integrated circuit of bipolar transistor was applied to the world, the quality certainly improved a lot, but in terms of cost, the improvement was not great. But by 1969, we have been able to design all the calculation units in the calculator on several integrated circuits. Only three years later, we can now design the whole complex calculator circuit on a MOS integrated circuit (see Figure 4). Using this MOS integrated circuit greatly reduces the cost of the calculator. Now you can buy an efficient calculator for 50 ~ 200 yuan dollars. I believe that the price of this calculator will be cheaper and the quality will be better in the near future.

Although MOS is not fast enough to be used in the central processing system of large computers, the price of MOS integrated circuits is getting lower and lower, and it can compete with magnetic ring memory at present. It is believed that all memory cells in future computers will be replaced by MOS. At present, the price of each bit in MOS is about 0.8 cents. Recently, a random access memory (RAM) is made of MOS, the price of which is comparable to that of magnetic ring memory. Its advantages are less power required and less heat generated, so the density of storage cells in the memory can be designed to be very high when designing a computer. In addition, when a magnetic coil is used as a memory, high-quality wires are required. For the sake of economy, this kind of high-quality wire is often shared by all magnetic coils, which virtually limits the function of the computer. However, when using MOS memory, because its data can be replaced by integrated circuits, computer designers can freely arrange their memories, so that the whole computer has better efficiency without worrying about the cost. Although manufacturers of magnetic ring memory are striving to compete with MOS memory, I firmly believe that it is only a matter of time before MOS replaces magnetic ring memory.

What are PMOS, NMOS and CMOS?

Looking back on the development history of semiconductor technology, we can see that the whole semiconductor technology has been progressing due to the high research and development of semiconductor materials, structures and circuits. In terms of MOS, its application has been quite extensive, but it is still expanding. The earliest MOS integrated circuit on the market is P-channel enhancement (PMOS). At present, this kind of MOS accounts for about 80% of all MOS integrated circuits, which is probably the reason why the production process of PMOS is easier to control! However, the current technology can already manufacture other types of MOS, such as NMOS (N-channel enhanced MOS) and CMOS (complementary MOS), in which NMOS and PMOS are combined. Because electrons move more easily than holes, the operation speed of NMOS is about 2 ~ 3 times faster than that of PMOS, so NMOS is used in some places where the speed factor is more important to obtain the best effect of the whole integrated circuit.

At present, CMOS is receiving extensive attention, and it is likely to become the most important of all components. The combination of N-channel and P-channel may be the best of all integrated circuits at present. The simplest CMOS circuit is an inverter (see Figure 5), which consists of PMOS and NMOS in series. At present, this circuit consumes the least power among all semiconductor elements. By combining these inverter circuits properly, we can design many useful circuits with little power consumption. For example, a 14-level binary counter (14-level binary counter), which is commonly used for timing, consumes only 2.5 microwatts (10-6 watts) of energy at 5 volts, which is only116 watts when using PMOS or bipolar transistor integrated circuits. This is really important for some instruments with limited power supply, and any battery-powered equipment should consider using CMOS.

PMOS and NMOS can also be connected in parallel to form a transmission switch, which can transmit digital signals or analog signals in both directions. Theoretically, this circuit can also be obtained by combining NPN and PNP transistors, but this circuit is very uneconomical, and there is another advantage of using low-cost CMOS, so CMOS should be used where the noise signal is strong. The circuit designer found that we can get any logic circuit and switch circuit we need by properly combining inverter circuit and switch circuit.

A great commercial application of integrated circuits, especially CMOS, is to make electronic watches or clocks, the accuracy of which is beyond the reach of any mechanical clock. It uses electronic counting circuit to divide the natural vibration frequency into several kinds of electrical signals to drive the hands on clocks and watches, and even directly connects these signals to electro-optical digital devices such as liquid crystal (liqguid crystal) and light emitting diode. In this way, we can know the time directly from the indicated figures. It seems that this kind of cheap electronic watch is bound to change the whole watch industry.

Theoretically, the working speed of MOS should only be related to the mobility of charge carriers and the distance traveled by carriers, so its working speed should be similar to that of the fastest transistor. But at present, the operation speed of MOS we make is much slower than that of bipolar transistor. What is the reason? Theoretically, since there is no limit, it must be a structural problem. In the past, when we diffused the source and drain, we often formed a considerable capacitance between the source, the drain and the silicon substrate. Due to these capacitors, the operation speed of the whole MOS is slowed down. Now scientists are using various methods to reduce these capacitors to improve the speed. I believe that the operation speed of MOS integrated circuits will be greatly improved in the future.

What is SOS?

In the manufacturing process and working principle of MOS (refer to Figure 2 and Figure 6), we can find that the silicon wafer actually used is only the surface layer, which is really not that thick, but the silicon wafer that is too thin is too broken to operate, so scientists thought of another method, that is, trying to plate a silicon single crystal film (about 10-4cm thick) on artificial sapphire, and then making MOS structure on this film. It is found that with this structure, the source-drain voltage is about 20 times lower than that of silicon wafer. Moreover, we can chemically etch away the silicon single crystal film between the transistors to achieve the effect of isolation, and then we evaporate the metal, so that the transistors can be connected with each other to form the circuit we need. It should be pointed out here that most metals are plated on sapphire, unlike previous MOS, which are all plated on silicon, so there will be no extra capacitance. This component made of sapphire coated with a silicon single crystal film is called SOS, which is the abbreviation of the English letter Silicon Sapphire. At present, this kind of SOS integrated circuit is only used in some special occasions because of its immature technology and relatively expensive price.

label

MOS can be successfully used as a field effect transistor, and we can also use silicon dioxide as a capacitor between the gate and the silicon substrate. Capacitors can store charge. If we arrange these MOS capacitors properly, we can use the clock pulse signal to control the charge transfer from one capacitor to another. Based on this principle, we can use MOS as a shift register in data processing system. In addition, MOS capacitors can also be used as photosensitive elements. When light irradiates the element, electric carriers will be generated and stored in the MOS capacitor. Later, when a series of clock pulse signals are input, we can read the signals generated by light. At present, a TV camera has been made, which is only as big as the palm of your hand and weighs less than one pound. It is made of this component. This MOS photosensitive element can also be applied to slow scan television, high-definition fax and other instruments requiring high resolution. We can imagine the application prospect of this component in industry or other entertainment consumption in the future.

Looking back on the development history of MOS, its theory has been interpreted by scientists for a long time, but it is only in recent years that real MOS devices have been widely used in the market. It can be seen that an idea that sounds reasonable is often realized by technology. Can you stop burying your head in scientific research in order to catch up with others? Encourage future young friends.

The original text was translated from Scientific American.

August 1973

Note 1: With the development of ion implantation technology and the improvement of crystal quality, this kind of integrated circuit seems to be just around the corner. (Please refer to Volume 4 of Science Monthly 10)

Note 2: We use N-type silicon crystal to illustrate this phenomenon. When a metal contacts a semiconductor, electrons in the N-type crystal near the interface will be repelled, so there will be a positively charged ion region near the interface, which we call the space charge region.

Note 3: Shackley, Badin and Braden won the 1956 Nobel Prize in Physics for inventing transistors. Among them, Braden visited China last September.

Note 4: The junction formed by bonding N-type crystal and P-type crystal is called PN junction, but in actual manufacturing, trivalent (or pentavalent) atoms are infiltrated into N-type (or P-type) primary crystal by diffusion or ion implantation technology to form this junction.

Note 5: In semiconductor science, this kind of positive charge is called "hole" because it is actually formed due to the lack of an electron on the bond of the crystal structure, and this kind of hole can easily snatch electrons from other bonds, resulting in the flow of electrons, which can be regarded as the flow of holes, but in the opposite direction to the flow of electrons. The reader should note that the positive charge between this positively charged hole and the previous space charge is completely different. The positive charge in the space charge region is generated by ions and is fixed, but due to the application of electric field, holes can flow to generate current.