How a wafer is made

As you well know, chips are made on wafers, also known as wafers. In this article, we are going to talk about this key piece in the semiconductor industry. This way you will know everything you need to know about the wafer , that strange unknown to many…

Keep in mind that the number of chips that can be produced will depend on these wafers , both CPUs, GPUs, SDRAM memory, flash, etc., as well as other types of chips or semiconductor devices.

Index of contents

• What is a chip?
• What is a wafer?
• How a wafer is made
• wafer sizes
• Parts of a wafer
• Wafer types
• Applications

What is a chip?

A chip, also called an integrated circuit , is nothing more than a monolithic semiconductor chip on which a circuit has been etched. For this, numerous techniques of diffusion, ion implantation, epitaxial growth, oxidation, photolithography, deposition, chemical attack, etc. are used.

For more information you should also read our tutorial on how to make a chip .

What is a wafer?

A wafer or wafer is nothing more than a thin sheet of semiconductor in a circular shape. This wafer acts as a semiconductor substrate to create the necessary electronic devices and the successive interconnection layers to create the chip or integrated circuit.

When a wafer is processed in a fab or foundry, what is actually being manufactured is hundreds of independent chips that will later be divided.

Of course, the most frequent material for wafers is silicon (Si wafer) . An abundant semiconductor with very interesting properties for the manufacture of chips. Obviously, not just any silicon is worth it, it must be very pure and with a crystalline structure.

For this reason, MGS (Metallurgical-Grade Silicon) silicon, or metallurgical-grade silicon used in other industries, is not considered pure enough for the semiconductor industry. For this reason, a high purity silicon called EGS or SGS (Electronic-Grade Silicon or Semiconductor-Grade Silicon) must be refined and created . An electronic grade silicon that is suitable for creating chips.

In addition, it cannot be in an amorphous form, or in a polycrystalline form, it must be in a monocrystalline form , that is, with the atoms ordered forming a crystalline structure with the same orientation.

It must also be said that although silicon may seem metallic, it is not entirely a metal. And it is that metals have free electrons that move easily between atoms, making them good electrical conductors. In contrast, pure silicon in crystalline form is almost an insulator, but this property can be modified by a dopant element .

A dopant is nothing more than a small amount of impurities that is introduced into the silicon crystal to change its conductive properties. Depending on the type of doping, the semiconductor may be N-type or P-type . For example, arsenic, antimony, and phosphorous are N-type dopants for silicon, while aluminum or boron are P-type.

Depending on the amount of dopant, the semiconductor can be considered extrinsic, when it is lightly doped, or degenerate when it has a high level of dopant.

You also have to read our article on why wafers are not different?

How a wafer is made

Once the silicon mineral has been obtained, usually from sand, it can go through a series of processes to purify it. To be suitable for making wafers, a purity of approximately 99.9999999% is needed . Once this has been achieved, the pieces of silicon are put into a crucible and melted.

Through a process known as Float-Zone or Czochralski , it is possible to create a monocrystalline silicon ingot or cylinder. To create a single crystal, it is very expensive and time consuming in the laboratory. Therefore, a seed crystal is simply created and through the Czochralski growth process, this seed crystal will be submerged in the crucible with the molten silicon and it will begin to spin, raise the seed crystal, and play with the temperatures.

The result is that a cylindrical ingot of single crystal crystal grows , with the same orientation as the seed crystal. It’s like the process of creating cotton candy at fairs, the cotton sticks to the stick in layers and grows in volume…

Once the ingot has been obtained, it will go through a series of processes :

1. Let it cool.
2. The ends are cut off.
3. Using a saw, the ingot is “sliced”, thus obtaining thin wafers.
4. It will also go through some polishing processes to leave a perfect surface on the wafer that will serve as the basis for manufacturing the chips.
5. And, of course, it will go through processes to remove unwanted particles and leave it clean.

wafer sizes

Throughout history, various wafer sizes have been used , from the smallest to the largest today:

WAFER SIZE (IN DIAMETER MM AND APPROX. INCHES)	Thickness	Year	Weight	Die/wafer (100mm2)
25 mm (1”)		1960
51 mm (2”)	275 μm	1969		9
76 mm (3”)	375 μm	1972		29
100 mm (4”)	525 μm	1976	10 g	56
125 mm (4.9”)	625 μm	1981		95
150 mm (6″)	675 μm	1983		144
200 mm (8″)	725 μm	1992	53 g	269
300 mm (12″)	775 μm	1999	125 g	640
450mm (17.7”) (proposed)	925 μm	–	342 g	1490
675 mm (26.6”) (theoretical)	A stranger	–	A stranger	3427

Currently, cutting-edge fabs use 300mm diameter wafers , 11.8″, although it is usually rounded and called 12-inch. Also years ago the use of 450mm was proposed, but this diameter has not been chosen because the advantages are not great enough with respect to 300mm to justify the expense to adapt all the machinery to this diameter.

In the last column you have the approximate number of chips that can be produced in a wafer of that size if we take into account that the chip is 100 square mm. However, 100 is a small value for the most advanced chips, such as the CPU or GPU, since, for example, the NVIDIA GeForce RTX 4090 has a size of 608 mm², and a 13th Gen Intel Core reaches 257 mm². Therefore, for CPUs an amount of 300 chips per wafer would be estimated, while for GPUs considering this size, it would be around 100 units.

Parts of a wafer

Next we are going to see some of the parts and functions of a wafer:

• Edge die: these chips are not complete, they are the ones on the edge of the wafer, and are considered a loss of production. The larger the wafer, the less loss they have from this problem.
• Scribe lines: they are lines of drawing, the lines that there are of space between the different chips. These parts are used to pass the saw or chip separation method and do not damage functional parts.
• Chip or die– May also be called a dice, die, etc. This is the integrated circuit itself.
• Chamfer or flat zone: some wafers have a chamfered edge, that is, flat. This cut is made to determine the orientation of the crystalline structure of the silicon from which the wafer is made. However, this is no longer commonly used today.
• TEG (Test Element Group): this is an integrated circuit different from the one that is intended to be manufactured on the wafer, and which is usually in several units per wafer, distributed along the surface. These chips are used to perform tests and verify that everything works properly, although they will then be discarded and only the rest of the functional chips will be left.

There are wafers called MPW (Multi-Project Wafer), and as the name suggests, they are not used to make a single identical chip, but a multitude of different chips are made from one another. In this way, the wafer is used to produce chips for institutions that do not want to mass-produce, for universities, for those who do not have a foundry and need to implement a physical chip, etc.

Wafer types

There are several types of silicon wafers, or wafers , with different purposes. To get an idea of ​​what these types are, we mainly have:

• Undoped: These are wafers made with processes such as FZ or Float-Zone, and that do not contain dopants, known as intrinsic for the type of semiconductor.
• Doped: they are wafers in which some impurity or dopant has been introduced. The dopant is usually added to the mixture. In this way, wafers with P or N substrate are obtained. P-type wafers have excess positively charged holes, and N-type wafers have excess negatively charged electrons. And, as I mentioned before, depending on the amount of dopants, it can be degenerate or extrinsic.
• Other types: there are also special cases of SOI (Silicon On Insulator) wafers such as PD-SOI and FD-SOI, SON (Silicon On Nothing), etc.

Here we are talking about silicon wafers, but they can be many other semiconductors, such as germanium, GaAs (Gallium Arsenide), InSe (Indium Selenium), InGaN (Gallium-I ndium Nitride) , etc.

Generally, the amount of dopant that is added is usually between 10 ¹³ and 10 ¹⁶ atoms of dopant for each cm ³ of the material to be doped. For example, when it is lightly or lightly doped, such as an N-type or a P-type, the dopant is usually in concentrations of 1 dopant atom per 100,000,000 silicon atoms. When it is high or heavy doping, such as N+ or P+ types, then the amount is one dopant atom per 10,000 silicon atoms.

Applications

Finally, we are going to see the utility of a wafer or wafer , since it is not only for what you imagine:

• Manufacture of semiconductor devices such as transistors, diodes, sensors, etc.
• Manufacture of chips for electronics, computing, etc.
• Manufacture of optical or optoelectronic elements, such as lasers, LEDs, IR, CIS image sensors, etc.
• Manufacture of MEMS mechanisms.
• Manufacture of solar cells for the panels, whether made of amorphous, monocrystalline silicon, cadmium telluride, perovskite, etc.
• Manufacture of certain special coating materials for the aerospace industry.