Is the mobile phone "high-end"? In fact, the most important part is made of sand→→→

In the last issue, we introduced the functions of chips, the role of semiconductors, and the advantages of silicon. If you want to make chips, you need to make silicon wafers with high enough purity. So today, let’s continue to talk about how to make silicon wafers that are pure enough. This is a long story, which is permeated with human wisdom.

Copyright image, no permission to reprint

01

Raw material purification

Quartz sand → silicon

A good cook cannot cook without rice. To make pure silicon wafers, the first thing you need is pure silicon raw materials. People mix sand with coke, coal or sawdust, and put the mixture into a graphite arc furnace for high-temperature heating. At a temperature above 1900℃, quartz sand is reduced to silicon through various chemical reactions. Among them, the main chemical reactions are the following two:

SiO2+C=Si+CO2 ↑

SiO
2+2 C=Si+2 CO ↑

Quartz sand, copyrighted image from the gallery, unauthorized reproduction

It seems not difficult to produce pure silicon in just a few chemical reactions. But in fact, the purity of silicon can only reach 98% at most, which is far from being the raw material for silicon wafers and needs further purification.

Liquefaction and purification

One of the most common methods in the purification process is liquefaction , because liquid purification is much easier and there are more methods than solid purification. Therefore, the next step is to chlorinate the crude silicon to form chlorides such as silicon tetrachloride (SiCl4) or trichlorosilane (SiHCl3), which happen to be liquids at room temperature.

After multiple distillations and purification of silicon tetrachloride or trichlorosilane with other liquids, ultra-pure chloride solutions can be obtained. Finally, by chemically reducing the high-purity chloride, we can obtain chip-grade polysilicon with a purity of more than 99.9999999%.

SiCl4+2H2→4HCl+Si

High-purity polysilicon, copyrighted image from the gallery, no permission to reprint

Does the process of making silicon raw materials end here? Not really! Although high-purity polycrystalline silicon has been prepared, the silicon used to make chips must be single-crystalline silicon. Although the two are only one letter apart, there is a huge difference in the arrangement of their internal atoms: the crystal framework structure of single-crystalline silicon is uniform, and the silicon atoms are arranged in an orderly manner; the silicon atoms of polycrystalline silicon are arranged in a disordered manner.

Schematic diagram of monocrystalline silicon (left) and polycrystalline silicon (right), drawing by Wang Zhihao

This is like when workers lay asphalt roads, before pouring the top layer of asphalt, they must first use soil to level the ground and compact the foundation. If the foundation is not flat enough, potholes or cracks will appear on the upper asphalt road. The same is true for silicon wafers. If the structure of the silicon wafer is disordered, that is, there are lattice defects, then after doping, the electrical characteristics of different parts will be very different, and the upper logic circuit will also have major defects. Therefore, single crystal silicon must be used to make chips.

02

From polysilicon to monocrystalline silicon

How to turn polycrystalline silicon into single crystal silicon? This requires special technology.

The most commonly used process for converting polycrystalline silicon into single crystal silicon is the CZ method (hereinafter referred to as CZ method). CZ refers to pulling silicon rods directly from "magma". It is the core process step in silicon wafer production and determines the quality and purity of silicon wafers.

Step 1: Melting Ultra-High Purity Polysilicon

The first step of the CZ direct pulling method is to place ultra-high purity polysilicon material in a crucible and heat it to 1420°C in a closed heat field to melt the polysilicon.

Polysilicon melting, Image source: Silicon Wafer Production Simulation animation

Step 2: Add "seed crystals"

The so-called seed crystal refers to a small crystal that is the same as the target crystal, that is, the seed of the grown silicon rod. Here it refers to a small piece of high-purity single crystal silicon. The seed crystal is the "child" of the silicon rod, usually from the part of the silicon rod with good quality. At this point, you may be curious, where did the first seed crystal come from in the world? Which came first, the seed crystal or the silicon rod?

This is like the question of "which came first, the chicken or the egg", but it is easier to answer. In laboratories that do not consider cost, high-purity seed crystals can be easily obtained, and in general laboratories, ultra-high-purity single crystal silicon can be obtained through methods such as chemical vapor phase technology. Therefore, in the question of which came first, the seed crystal or the silicon rod, the "egg" came first.

Inserting the seed crystal, Image source: Silicon Wafer Production Simulation animation

Step 3: Pull out and rotate

Back to the third step of the CZ method, the seed crystal is slowly pulled vertically out of the "magma" and rotated. The crystal will grow at the bottom of the seed crystal and gradually grow as the seed crystal is pulled to form a crystal rod. The properties of the grown crystal and the seed crystal are the same, both are single crystal silicon.

Vertical stretching to form a crystal rod. Image source: Silicon Wafer Production Simulation animation

This method may sound simple, but it is actually much more difficult than expected. In order to prepare highly uniform silicon rods, this large pot of silicon "paste" that is like magma needs to be kept at a stable temperature. At the same time, the pulling and rotating speed of the silicon rods also need to be extremely stable. In addition, the entire crystal pulling process needs to be carried out in a high temperature and negative pressure environment.

Nowadays, the diameter of wafers is getting bigger and bigger, from 4 inches (1 inch = 2.54 cm) to 12 inches and even 18 inches in the future. People are pursuing larger diameters. This is because the larger the diameter of the silicon wafer, the more chips can be made from the same silicon wafer, which reduces the cost accordingly.

However, the increase in silicon wafer diameter represents an exponential increase in manufacturing difficulty. First, the diameter of the crystal rod corresponding to the silicon wafer is required to be thicker, so the size of the thermal field used for heating must also be increased accordingly, and the convection of the magma will become more complicated. At the same time, the temperature gradient of the solid-liquid interface and the oxygen concentration distribution become difficult to control, which means that the control requirements for crystal pulling are also more complicated.

Copyright image, no permission to reprint

However, smart people have perfectly solved these problems by adding a magnetic field to the traditional CZ device system. Since molten silicon can conduct electricity, it will be affected by the force generated by the interaction between the magnetic field and the flow, which can change the convection of the "magma". In addition, under the appropriate magnetic field distribution, the crystal growth process can also reduce the impurities such as oxygen, boron, and aluminum entering the silicon melt through the crucible, thereby preparing high-resistivity silicon rods with controllable oxygen content and better uniformity.

This processing method that adds a magnetic field device to the traditional CZ method is called the magnetic controlled CZ method (hereinafter referred to as the MCZ method). These customization advantages also make it the mainstream process technology at present. The MCZ method can be divided into the longitudinal magnetic field method, the transverse magnetic field method and the cusp magnetic field method according to the different applied magnetic fields. As the name implies, the directions of the applied magnetic fields are different, and they can achieve different functions and have different characteristics.

Schematic diagram of longitudinal magnetic field method, transverse magnetic field method and cusp magnetic field method

(Image source: Global Wafers Japan)

The process of drawing a crystal rod is a complex systematic control process with high technical difficulty, which requires a long period of experience accumulation and optimization. At present, in addition to the CZ method, there is also a floating zone melting method (hereinafter referred to as the FZ method) for the preparation of single crystal silicon. The floating zone melting method uses heat energy to generate a molten zone at one end of the rod and then melt the seed crystal. By adjusting the temperature, the molten zone slowly moves to the other end of the rod, passing through the entire rod, and growing into a single crystal in the same direction as the seed crystal.

The CZ method and the FZ method each have their own advantages and disadvantages:

The advantages of the Czochralski method are that the silicon produced has a higher oxygen content, greater mechanical strength , and is easier to make large-sized silicon rods. At the same time, the Czochralski method has lower costs and faster crystal growth. Therefore, about 85% of single crystal silicon wafers are now produced using the Czochralski method.

However, the FZ method also has its own advantages. For example, the resistivity of single crystal silicon produced by the FZ method is very high , which is particularly suitable for high-power devices such as detectors and rectifiers. In addition, since the FZ floating zone melting method avoids contamination caused by the crucible, the purity of single crystal silicon can be higher . However, its disadvantage is that the silicon rods produced are small in size, with a maximum of only 8 inches, and it is difficult to make them larger.

Schematic diagram of FZ method, image source: Global Wafers Japan

After completing the above production process, we finally obtained almost pure silicon rods. Then we enter the next processing stage.

03

Silicon rod cutting and grinding

The silicon rods will then be cut off from the head and tail, and the good quality silicon rods will be cut into "seed crystals" for the next growth. Since the straight-pulled silicon rods are not perfect cylinders, the remaining silicon rods will be cut into appropriate sizes and put into the machine to slowly roll and grind the sides to form the required radius and shape.

Silicon rod cutting, copyrighted image, unauthorized reproduction

Next, the ground silicon rods are cut into slices . In the past, cutting silicon rods was like cutting lamb slices at home, one slice at a time. Although the cut surface was smooth, the efficiency was too low. Nowadays, people are more likely to use multi-wire cutting machines with diamond wires. The number of slices cut each time is linked to the number of diamond wires. Although the cut surface is not as smooth as the previous internal circle cutting machine, it is more efficient.

Multi-wire cutting, Image source: Global Wafers Japan

The cut silicon wafer will be mechanically polished to make its surface smoother. Some silicon wafers also need to have a rough backside to artificially create defects so that impurities added in subsequent processes can be trapped on the backside to protect the device. In addition, the edge of the silicon wafer needs to be polished into an arc shape to prevent edge cracking and facilitate subsequent photolithography.

After polishing, it is placed in nitric acid or hydrofluoric acid for chemical etching to remove the mechanical damage accumulated on the silicon wafer during the previous polishing process and the abrasive mixed into the surface of the silicon wafer.

After a series of processes such as polishing and etching, the surface of the silicon wafer is as smooth as a mirror, but it is still not enough for chip manufacturing.

Grinding and etching, Image source: Global Wafers Japan

04

Silicon wafer polishing and cleaning

Although the surface of the silicon wafer is as smooth as a lens at this point, it still needs to be polished chemically and mechanically , a step that combines physical and chemical polishing methods. The silicon wafer is first mounted on a rotating polishing machine, where a thin layer of its surface is first chemically oxidized by the abrasive fluid and then physically polished by the polishing pad until the silicon wafer is polished into a nearly perfect mirror surface.

Chemical mechanical polishing and cleaning, Image source: Global Wafers Japan

After this step, the flatness of the silicon wafer will reach a very high level. The flatness of a 12-inch silicon wafer is required to be controlled within 51 nanometers. Most people may not understand this flatness, but if we magnify it millions of times, it is equivalent to the maximum fluctuation of no more than 25 centimeters in a circle with the distance from Beijing to Shanghai as the diameter .

After that, the silicon wafer needs to be cleaned with deionized water and various chemical solvents to remove various dust and impurities adhering to the surface of the silicon wafer during the manufacturing process. These particles will affect the chip manufacturing process and easily cause short circuits or open circuits in the device.

Copyright image, no permission to reprint

Finally, the silicon wafers are inspected to ensure critical wafer flatness and surface cleanliness (no particles), and to ensure that indicators such as warpage, oxygen content, and metal residue meet the standards to ensure wafer quality. After various tests such as electron microscopy and optical scattering meet the standards, the silicon wafers are placed in clean shipping boxes, sealed in special moisture-proof bags, and safely sent to the next factory for other processes.

Seeing this, you may breathe a sigh of relief. The silicon wafer preparation process is finally over. But unfortunately, this is just a blank sheet of paper. The silicon wafer still needs to go through a series of operations such as lithography, epitaxy, and etching before it can become a wafer containing hundreds of chips. Then it needs to be cut and packaged before it can become individual chips and enter the market.

You may be shocked that a seemingly insignificant silicon wafer actually has such a complex manufacturing process. But at the same time, you may also be curious about the development of my country's silicon wafer manufacturing technology and the proportion of my country's silicon wafer industry. Please listen to the next analysis.

Produced by | Science Popularization China

Author: Wang Zhihao (Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Sciences)

Producer｜China Science Expo

Submitted by: Computer Information Network Center, Chinese Academy of Sciences

The cover image and the images in this article are from the copyright library

Reproduction of image content is not authorized