In today's article, we will tell you the whole process from a pile of sand to a powerful integrated circuit chip step by step. Basic raw materials for making chips
If you ask what is the raw material of the chip, everyone can easily give the answer-silicon. That's true, but where does silicon come from? In fact, it is the most inconspicuous sand. It's hard to imagine that the chip with expensive price, complex structure, powerful function and full of mystery actually comes from worthless sand. Of course, there must be a complicated manufacturing process. But you can't just grab a handful of sand as raw material. You must choose carefully and extract the purest silicon material from it. Imagine if the chip is made of the cheapest and richest raw materials, what will the quality of the finished product be like? Can such a high-performance processor still be used?
Besides silicon, an important material for making chips is metal. Today, aluminum has become the main metal material for making internal fittings of processors, while copper has been gradually eliminated. There is a reason for this. At the current chip operating voltage, the electromigration characteristics of aluminum are obviously better than that of copper. The so-called electromigration problem refers to that when a large number of electrons flow through a section of conductor, the atoms of the conductor material leave their original positions due to the impact of electrons, leaving vacancies. Too many vacancies will lead to the disconnection of wires, while atoms leaving the original position will stay in other positions, causing short circuits in other places, thus affecting the logic function of the chip and further causing the chip to be unusable. This is why many Northwood Pentium 4 have been replaced by SNDS (Northwood Outbreak Syndrome). When enthusiasts overclocked Northwood Pentium 4 for the first time, they were eager for success and greatly increased the chip voltage. Serious electromigration problems caused the chip to be paralyzed. This is Intel's first attempt at copper interconnection technology, and obviously some improvements are needed. On the other hand, the application of copper interconnection technology can reduce the chip area, and at the same time, because the resistance of copper conductor is lower, the current passing through it is faster.
In addition to these two main materials, some kinds of chemical raw materials are needed in the design process of the chip, which play different roles, so I won't go into details here. In the preparation stage of chip manufacturing, after the collection of basic raw materials is completed, some of these raw materials need to be pretreated. As the most important raw material, the treatment of silicon is very important. First of all, silicon raw materials should be chemically purified to reach the level of raw materials used in semiconductor industry. In order to make these silicon raw materials meet the processing needs of integrated circuit manufacturing, they must be shaped. This step is completed by melting silicon raw materials and then injecting liquid silicon into a large high-temperature time-sensitive container.
Then, the raw materials are melted at high temperature. We learned in middle school chemistry class that atoms in many solids are crystal structures, and so is silicon. In order to meet the requirements of high-performance processors, the whole silicon raw material must be high-purity monocrystalline silicon. Then the silicon raw material is taken out from the high-temperature container by rotation and stretching, and a cylindrical silicon ingot is made at this time. According to the technology currently used, the diameter of the circular section of the silicon ingot is 200mm. But now Intel and other companies have begun to use 300 mm diameter silicon ingots. It is quite difficult to increase the cross-sectional area without changing the characteristics of silicon ingots, but it can be achieved as long as enterprises are willing to invest a lot of money in research. The factory built by Intel to develop and produce 300mm silicon ingots cost about $3.5 billion. The success of new technology enables Intel to manufacture more complex and powerful integrated circuit chips. The 200 mm silicon ingot factory also spent1500 million dollars. The manufacturing process of the chip is introduced from the slicing of silicon ingot.
After the single crystal silicon ingot is made into a silicon ingot and guaranteed to be an absolute cylinder, the next step is to slice the cylindrical silicon ingot. The thinner the slice, the less material is used, and naturally more processor chips can be produced. Slices should also be mirrored to ensure that the surface is absolutely smooth, and then check for distortion or other problems. The quality inspection in this step is particularly important, which directly determines the quality of the finished chip.
The new slice of monocrystalline silicon ingot should be doped with some substances to make it a real semiconductor material, and then transistor circuits representing various logic functions should be engraved on it. Doped substance atoms enter the gaps between silicon atoms, and atomic forces interact, thus making silicon raw materials have the characteristics of semiconductors. Nowadays, multi-choice CMOS process (complementary metal oxide semiconductor) is used in semiconductor manufacturing. The word complementary represents the interaction between N-type MOS transistor and P-type MOS transistor in semiconductor. N and p represent the negative electrode and the positive electrode in electronic technology respectively. In most cases, the P-type substrate is formed by slicing and doping chemicals, and the logic circuit engraved on it should be designed according to the characteristics of nMOS circuit. This kind of transistor has higher space utilization rate and more energy saving. At the same time, in most cases, it is necessary to limit the appearance of pMOS transistors as much as possible, because in the later stage of the manufacturing process, N-type materials need to be implanted into P-type substrates, and this process will lead to the formation of pMOS transistors.
After the dosing work is completed, the standard section is completed. Then each slice is heated in a high temperature furnace, and a silicon dioxide film is formed on the surface of the slice by controlling the heating time. By closely monitoring the temperature, air composition and heating time, the thickness of the silicon dioxide layer can be controlled. In Intel's 90 nm manufacturing process, the width of the gate oxide is as small as 5 atoms. This gate circuit is also a part of transistor gate circuit. The function of transistor gate circuit is to control the electron flow in it. By controlling the gate voltage, the electron flow is strictly controlled regardless of the voltages at the input and output ports. The last step of preparation is to cover the photosensitive layer on the silicon dioxide layer. This material layer is used for other control applications in the same layer. This layer of material has a good photosensitive effect after drying, and can be dissolved and removed by chemical method after the lithography process.
Lithography is a very complicated step in the current chip manufacturing process. Why do you say that? Photolithography process is to use a certain wavelength of light to carve a corresponding notch on the photosensitive layer, thus changing the chemical characteristics of the material there. This technology is very strict with the wavelength of the light used, which requires the use of short-wavelength ultraviolet rays and lenses with large curvature. The etching process is also affected by stains on the wafer. Every step of etching is a complicated and delicate process. The amount of data needed to design each process can be measured by 10GB, and the etching steps required to manufacture each processor are more than 20 steps (etching one layer at a time). Moreover, if the etching map of each layer is magnified many times, it can be compared with the map of new york plus suburbs, or even more complicated. Imagine reducing the map of new york to a chip with an actual area of only 100 square millimeter. You can imagine how complicated the structure of this chip is.
When all these etching jobs are completed, the wafer is turned over. Short wavelength light shines on the photosensitive layer of the wafer through the hollowed-out gap on the timely template, and then the light and the template are removed. The exposed photosensitive layer material is removed by chemical method, and silicon dioxide is generated right below the hollow position.
After the doped material in the remaining photosensitive layer is removed, what remains is a silicon dioxide layer filled with grooves and an exposed silicon layer below this layer. After this step, another silicon dioxide layer is completed. Then, another polysilicon layer with a photosensitive layer is added. Polysilicon is another type of gate circuit. Because metal raw materials (therefore called metal oxide semiconductors) are used here, polysilicon allows the gate circuit to be established before the transistor queue port voltage takes effect. At the same time, the photosensitive layer is etched by short wavelength light passing through the mask. After another etching, all the required gate circuits have been basically formed. Then, the exposed silicon layer is subjected to ion bombardment by chemical means, and the purpose here is to generate an N channel or a P channel. This doping process produces all transistors and their circuit connections. No transistor has an input and an output, and both ends are called ports.
Repeat this process.
From this step, you will continue to add layers, add silicon dioxide layers, and then photoetch once. Repeat these steps, and then a multi-layer three-dimensional architecture will appear, which is the embryonic state of the processor you are currently using. Conductive connection between layers is realized by metal coating technology. Today's P4 processor uses 7 layers of metal connections, while Athlon64 uses 9 layers. The number of layers used depends on the original layout design and does not directly represent the performance difference of the final product.
In the next few weeks, we need to test the wafers one by one, including the electrical characteristics of the wafers to see if there are any logical errors, if so, which floor, and so on. Then, each defective chip unit on the wafer will be tested separately to determine whether the chip has special processing requirements.
Then, the whole wafer is cut into individual processor chip units. In the initial test, those units that failed the test will be abandoned. These cut chip units will be packaged in some way, so that they can be successfully inserted into the motherboard of a certain interface specification. Most Intel and AMD processors are covered with heat sinks. After the processor is completed, it is necessary to carry out all-round chip function test. This company will produce different grades of products. Some chips have relatively high operating frequency, so the names and numbers of high-frequency products are marked, while those chips with relatively low operating frequency are modified and marked with other low-frequency models. This is the market positioning of different processors. Moreover, some processors may have some shortcomings in chip functions. For example, if it has a defect in cache function (this defect is enough to paralyze most chips), then they will be shielded from some cache capacity, which will reduce performance and of course reduce the price of the product. This is the origin of Celeron and Sempron. After the chip packaging process is completed, many products need to be re-tested to ensure that the previous manufacturing process is not omitted, and the products fully meet the specifications and have no deviation.