In the last two years, the entire semiconductor industry has been in the limelight of the confrontation between the United States and China, and the design than the domestic chip is more stuck in the place of manufacturing. Specifically, a series of key equipment and materials in the manufacturing process, including but not limited to photo-lithography, photo-resist, thin film deposition equipment, ion implantation machines, etc. These are basically the core steps monopolized by foreign manufacturers. I will be in the future series, a detailed explanation, so you understand how the chip is made step by step, by the way, will also analyze the road of domestic chips, in the end, what hurdles to overcome.
First of all, a simple comb through the basic process of making a chip with sand: take Intel’s short science video, for example, its own title is called “From Sand to Silicon, the making of a chip”, the reason why the major manufacturers say so is that the mainstream chip is still silicon-based Chip, silicon and precious metals are different, is a rotten element, the content in the earth’s crust, up to 26.3%, second only to oxygen, and silicon and oxygen together, the composition of the sand everywhere in nature, the main component of silicon dioxide.
So the first step in chip making is to reduce the silicon dioxide into silicon ingot. After purification and straight pulling method, a long silicon rod is taken out, and then this rod is cut and polished to produce a disc-like raw wafer. It is then sent to a wafer fab where the physical structure of the transistor is sculpted through photo-lithography and etching, and given electrical characteristics through ion implantation and lamination. Thus hundreds of millions of electronic devices and their corresponding logic circuits are shaped in these constantly repeated processes into hundreds of chips on a single wafer, and finally they are cut and separated and packaged for testing, completing the manufacture of a chip.
Today we will start with sand to see what it takes to get a long, black, straight, large stick from a pile of inconspicuous grains of sand. First of all, the sand used to make the chip, not the kind of river sand on the construction site, but a higher silicon content of silica, the main component and the same as the sand, is silicon dioxide. The chip’s previous life, a total of three lives to experience three lives, melting and lifting. This former former life, respectively, silica stone, to silicon ingots, silicon ingots to frosted models of silicon rods (poly-crystalline silicon), frosted models to mirror models of silicon rods (mono-crystalline silicon).
1 Smelting silicon ingots
Let’s start with the first step, the production of silicon ingots, where a crucible (for gold refining) is used, also called a mineral furnace, or an electric arc furnace, usually 12 meters in diameter and made of graphite, with two three-meter long graphite electrodes in the middle for heating. Graphite, like diamond, is an isomer of carbon with a melting point of 3800 degrees Celsius and a very small coefficient of expansion, which makes the graphite crucible the most ideal smelting equipment. In the crucible, about 60 tons or so of silica are put in, and about 25 tons of coal and wood chips are added. Under the effect of high temperature of 2000 degrees Celsius, the silica gradually melts and produces a magical chemical reaction with carbon – the reduction reaction, the main reaction here, is silica and carbon, producing silica monomers and carbon monoxide. The secondary reaction, which occurs in the lower temperature region of the furnace, produces silicon carbide. If the content of silicon dioxide in the furnace is high enough, silicon carbide can be used as an intermediate product and continue to be converted into silicon and carbon monoxide, and the smelted silicon solution is condensed to give a relatively pure silicon ingot.
This is smelting grade industrial silicon with a purity of about 98%-99%. Although it still contains small amounts of impurities such as iron and aluminum, it is already an important material for the chemical, metallurgical and construction industries.
China’s industrial silicon production is about 65% of the world’s. After all, alchemy is just a basic skill for China, which has a full range of industrial disciplines, and there are no complex technical thresholds or equipment requirements. And 40% of the cost of silicon refining is the electricity consumed by electric heaters. So the relatively cheap price of electricity is also a favorable condition for China to produce silicon. But for the semiconductor industry, 99% purity, is not enough.
2 Refined poly-silicon
So the second step is the purification of industrial silicon, which is the mainstream industry practice of purification with the help of hydrogen chloride gas, also known as Siemens-Verfahren, since it was first developed by Siemens in 1955. In practice, the silicon ingot is crushed into slag, which reacts with hydrogen chloride at 325 degrees to produce hydrogen gas and the target product trichinosis. The main impurities in this step include the gases ferric chloride, aluminum hydrochloride and silicon hydrochloride. Using the different boiling points of these gases, we can separate the lower boiling point trichinosis gas through condensers and distillation columns with temperature control. The next step is to reduce the high purity trichinosis back to solid silicon. This is done by passing hydrogen gas at a high temperature of 1100 degrees Celsius to produce silicon, hydrogen chloride and silicon hydrochloride.
At this temperature, only silicon is solid, so you will see in the reactor that black silicon is slowly growing out. In about a week’s time, about 250 kilograms of silicon rods can be accumulated, and the silicon content can be as high as 99.999999999% (a total of 11 9s). The silicon rods at this point, however, are still frosted models with a pockmarked surface, which is because the silicon manufactured at this step is polycrystalline, meaning that its crystal frame structure is not uniform and the whole is made up of numerous small irregular crystals. The polysilicon produced at this stage can be used in the photovoltaic industry, such as making solar modules and panels.
In 2019, China’s polysilicon production capacity, accounting for as much as 69.2% of the world’s first. Domestic leading manufacturers, such as GCL-Poly, Xinte Energy, Tongwei, Da all still continue to expand production. Germany’s WACKER and other old overseas factories to lose their armor. In short, to the polysilicon this step, we are not only not stuck neck, but also a great advantage. But the chip needs more is a uniform and continuous lattice, the electrical properties of stable monocrystalline silicon.
3 Manufacture of poly-silicon
So the third step is to turn the frosted polycrystalline silicon rods into mirror-like monocrystalline silicon. In this step, the industry’s mainstream production method is the Czochralski Process. Rods. In this process, the high-purity polysilicon obtained in the previous step is heated and melted in a quartz crucible. The melting point of quartz is about 1700 degrees Celsius and the melting point of silicon is about 1400 degrees Celsius. The temperature here is generally controlled to be slightly higher than the melting point of silicon, and then a small strip of crystal seeds, i.e. a tiny single crystal of silicon, is dipped into the silicon melt, and then slowly lifted upward, rotating and pulling. comes out.
And the thickness and quality of the pulled out big bar depends on the working temperature, rotation speed and lifting speed. Here is a reminder of a detail that in the traditional process, there is often a neck of a few millimeters in diameter at the top of the large bar. This is because when the crystal seeds are first in contact with the solution, high frequency slip dislocations occur due to thermal shock, resulting in a large number of crystal defects in the initial section, so at the beginning, a neck of about 10 cm is pulled at high speed (6mm/min) to slow down the dislocation phenomenon until it disappears completely, then the speed is reduced and the large diameter rod is pulled. At this point, the solidified silicon rod, like the crystal seed, is a smooth single crystal silicon. However, the presence of a thin neck limits the weight of a single silicon rod. If you pull half broken, it would be embarrassing. So the diameter of 8 inches of silicon rods, generally pull 6 meters long, 12 inches of large rods, generally pull 1 m 5. In addition to the pulling method.
Other methods of making monocrystalline silicon include the Floating-Zone process, also known as zone melting. This method is equivalent to melting and solidifying silicon rods in sections, using the difference in impurities between the solid and liquid phases to refine and purify monocrystalline silicon. By not using a crucible, this approach avoids the impurity component of the pulling method that is carried by the vessel. However, the pulling method also has its own benefits, such as the possibility of doping directly in the melt of the crucible, that is, adding materials such as boron and phosphorus to produce impurity semiconductors with different electrical properties than pure silicon. Doping is an important process in chip production, which will be explained in detail later. Now we have completed the third step, from polycrystalline silicon rods refined into monocrystalline silicon rods, which will be cut into thin slices, is the raw material for making chips, the production of wafers of silicon.
So the silicon rod pulled in this step will determine the size of the wafer. If you want to make an 8-inch wafer, you have to pull it thinner; if you want a 12-inch wafer, you have to pull it thicker. The relationship between the diameter of the silicon rod and the maximum lifting speed can be estimated using the heat balance between the melt, the silicon rod and the radiation heat. In 12-inch wafer size, for example, 1 m 5 silicon rod, need to pull about 8 hours, theoretically the diameter of course, the thicker the better, because the larger the individual wafer area, the more chips made, the lower the cost of sharing down. But at the same time the process difficulty and equipment costs, will also skyrocket. For example, an 8-inch wafer fab, it will cost $1.6 billion, while the 12-inch plant, it will take $3-4 billion. Have to sigh, do the chip really burn money.
Summary.
To summarize, now that we have finished taking a silicon rod out of the sand, the next step is to cut it into silicon wafers, ready to start the front-end production of chips. Of course slicing itself is not so simple, to go through rolling grinding, chamfering, fine research, back roughness, chemical mechanical polishing and so on a series of operations, the technical threshold of these steps, gradually rising, the requirements for fine processing gradually perverted, it is from this stage, China’s production capacity share and the degree of autonomy decline, the road to domestic production of chips, will also usher in the first necked link.
At present, the silicon wafer market is mainly controlled in the hands of five companies, respectively, Japan’s Shin-Etsu Chemical and Seiko Group, followed by Taiwan’s Global Wafer, Germany’s Siltronic (Siltronic), and South Korea’s SK Siltron, they add up to control 90% of the global silicon wafer supply, specific market analysis and manufacturing details, will continue in the next section.
The above information comes from the talk three circles bilibili video, thanks to the author’s detailed explanation
