Friends familiar with the semiconductor manufacturing process know that before cutting and packaging, all manufacturing processes of the chip are performed on a wafer. However, the chips we see are all square-shaped, and when manufacturing chips on a circular wafer, there will always be some areas that are not utilized. So, why not use square wafers to increase the utilization rate?
Actually, this question is quite easy to answer. The wafer (originally a silicon wafer) is cut from a cylindrical silicon ingot, which is why its cross-section is circular. Now, another question arises: why is the silicon ingot in the shape of a cylinder? In this article, I will introduce you to the upstream industries of semiconductor manufacturing, the process of making silicon wafers and wafers.
Wafers are more suitable for making chips than “square wafers”
We often joke that chips are essentially made of a bunch of sand, and this statement is not wrong. The foundation of chip manufacturing – single-crystal silicon – is derived from quartz sand.
The key to transforming sand into chips lies in the purification of silicon and the development of the single-crystal silicon preparation process. In 1916, the Polish chemist Jan Czochralski accidentally dipped a pen into a crucible of molten tin instead of an ink bottle. When he pulled out the pen, he found a solidified metal wire attached to the nib. Czochralski verified through experiments that this metal wire was composed of a single crystal of metal with a diameter reaching millimeters. Subsequently, this method underwent continuous iterations by chemists and eventually produced single-crystal silicon. This method is also known as the Czochralski method or the Czochralski pulling method.
In contrast to single crystals, polycrystalline silicon is composed of many small crystalline grains with no regular arrangement between them. Single crystals themselves are completely large crystals with ordered arrangements of atoms or ions. Since polycrystalline silicon does not have the repetitive single-crystal structure that provides stable electrical and mechanical properties like single-crystal silicon, only single-crystal silicon is used to make the silicon wafers for chip production.
The process of the Czochralski pulling method involves first heating high-purity silicon in a crucible until it becomes molten. Then, a seed crystal (crucible) is placed at the end of a precisely oriented rod and immersed into the molten silicon. The rod is then slowly pulled upwards while rotating. By precisely controlling the pulling and rotation speed as well as the temperature, a larger cylindrical single-crystal silicon ingot can be obtained at the end of the rod. Subsequently, the silicon ingot goes through processes such as grinding, polishing, and cutting to obtain a usable round silicon wafer.
Therefore, it can be said that the circular shape of the wafer is due to the “round” shape of the silicon ingot. However, to be accurate, the wafer is not perfectly circular. Typically, a notch is ground around the wafer after it is processed into a wafer (a flat cut below 200mm or a small cut above 200mm) to indicate the crystal’s growth direction and facilitate the positioning of the wafer during subsequent photolithography and etching steps.
In fact, the silicon ingot can be cut into rectangular blocks before slicing, which would result in “square wafers” directly during subsequent slicing. However, this method is not used for producing silicon wafers for chips, and there are several reasons for this:
Firstly, the circular shape is more suitable for photolithography. Before photolithography, the wafer needs to be uniformly coated with a layer of photoresist on its surface, and the uniformity of the coating directly affects the chip’s yield. The commonly used method is to apply the resist at the center of the wafer and then spin it to spread the photoresist across the entire wafer. Due to the liquid’s viscosity, surface tension, and air resistance, when using square wafers, photoresist tends to accumulate at the four corners after spinning, leading to poor photolithography results, decreased yield, and more wastage.
Furthermore, the circular shape provides higher structural strength due to the presence of edge stress. Before becoming a wafer, the silicon goes through multiple processes such as photolithography, etching, and chemical polishing, leading to the accumulation of stress at the outer edge of the wafer. Therefore, the sharp corners of square wafers would lead to stress concentration, making them more susceptible to breakage during the production process and adversely affecting overall yield.
Keen observers may notice that some wafers have no chips on their outer edges, while others have complete circuit patterns covering the entire wafer, including the outer edges.
This is actually related to the size of the photomask used in photolithography. Photolithography essentially involves exposing a silicon wafer coated with a photoresist to specific light (such as extreme ultraviolet, EUV) through a photomask, which leaves the required circuit patterns on the wafer’s surface. The photomask itself is square-shaped and consists of many grids, with each grid called a “Shot,” which is the smallest unit of exposure. Each Shot includes one or more Dies with peripheral test circuits. Since the Shot is square, the Dies are also square. The photomask’s size typically covers the entire area of the wafer, resulting in incomplete small squares at the wafer’s edges.
Furthermore, due to the edge effect, not creating peripheral circuits after photolithography can affect the material density inside and outside the wafer, thereby impacting overall yield. Industry experts explain this phenomenon: “In the chip manufacturing process, the wafer is continuously thickened, especially in the later stages during metal and via fabrication, where multiple CMP (Chemical Mechanical Polishing) processes are used. If there are no patterns at the wafer’s edge, the edge polishing rate will be slow, leading to height differences between the edge and center, which will, in turn, affect adjacent complete chips during subsequent polishing processes. Therefore, even as a dummy pattern, the incomplete shots at the wafer’s edge need to be exposed normally.”
Circular chips are, in fact, more challenging to manufacture
After going through steps such as coating, photolithography, etching, and ion implantation, individual chips are manufactured on the silicon wafer. However, at this stage, the chips are still “attached” to the wafer and need to be cut to become separate individual chips.
Imagine that square chips can be easily cut out with just a few cuts. If the chips were circular, it would likely take several times longer to cut them. From a packaging perspective, square chips are also more conducive to wire bonding operations, and even in flip-chip packaging, the square shape is more convenient for aligning the I/O interfaces with the solder pads using machines.
Most importantly, circular chips do not solve the problem of silicon wafer area waste. Cutting many square regions from a wafer leaves no gaps in between; only the edges of the wafer are left unused. However, if many circular regions are cut from a plane, there will inevitably be wasted areas in between, and it still cannot avoid waste at the wafer’s outer periphery.
Indeed, saving the wafer area is always an important topic. The more chips that can be produced on a wafer, the higher the production efficiency, and the lower the cost per individual chip. Currently, the most effective way to improve production efficiency is by increasing the wafer’s surface area, a concept familiar to us as calculus.
From the image, it is evident that when the chip area is fixed, using larger wafers can effectively increase wafer utilization.
When producing 100mm² chips on a 12-inch wafer, approximately 660 chips can be manufactured, while using 8-inch wafer results in only 180 chips. The wafer area decreases by 50%, but the chip quantity reduces by 72%. Consequently, 12-inch wafers have become the primary battleground for larger IDM and foundry manufacturers worldwide. In China, only a few companies currently possess the technology for manufacturing 12-inch semiconductor wafers, and domestic enterprises are accelerating their efforts to catch up with global leaders.
Regarding wafer utilization, circular chips are not feasible at present, but there indeed exist square-shaped “wafers,” and they are not only present but also quite common.
Square photovoltaic silicon wafer
In addition to chip manufacturing, silicon wafers are also an extremely important component in the field of photovoltaics.
Photovoltaic (PV) power generation utilizes the photovoltaic effect of silicon wafers to directly convert solar radiation into electrical energy. The photovoltaic effect of crystalline silicon is applicable to both polycrystalline and monocrystalline silicon, but monocrystalline silicon has a more complete crystal structure with uniform optical and electrical properties, higher mechanical strength, and more efficient photovoltaic conversion. Therefore, monocrystalline solar cells can achieve 2-3 percentage points higher conversion efficiency compared to polycrystalline solar cells.
The initial stages of the manufacturing process for monocrystalline silicon used in photovoltaics are the same as those for monocrystalline silicon used in chips. Both processes involve heating high-purity silicon to a molten state and then pulling a monocrystalline silicon ingot from it. Before slicing, the photovoltaic silicon ingot is cut into rectangular blocks, resulting in a square cross-section for the silicon wafers. The reason for using a square shape is also straightforward – if photovoltaic cells were circular, there would be gaps between multiple cells arranged in a solar panel, reducing the overall conversion efficiency.
Compared to chip manufacturing, the requirements for silicon purity in photovoltaic panel production are slightly lower. The purity standard is typically 99.9999%, which does not reach the 99.999999999% purity required for chip fabrication.
| Semiconductor Silicon Wafer | Photovoltaic Silicon Wafer | |
| Purity Rubric | 99.999999999% | 99.9999% |
| Types of silicon raw materials | Monocrystalline Silicon | Monocrystalline silicon, polycrystalline silicon. Polysilicon accounted for 60% |
| Surface standard | The flatness and smoothness are controlled within 1nm | Standards are lower than in semiconductors |
| Shape and Dimension | Round, diameter 150mm, 200mm, 300mm | Square, side length 125mm, 150mm |
| Application | Mainly used in the production of chips, is widely used in communications, consumer electronics, automotive, industrial and other fields | Mainly used in the production of chips, is widely used in communications, consumer electronics, automotive, industrial, and other fields |
The difference between semiconductor silicon and photovoltaic silicon
Conclusion
The answer to the question posed in the title is that chips are square because it is easier to cut them, and they are more convenient to control during the subsequent packaging stages. Moreover, square chips address the issue of wafer area waste more effectively. On the other hand, wafers are circular primarily due to mechanical factors, which make the production process more convenient and yield higher. The silicon ingot used in the manufacturing process is naturally in cylindrical shape, which leads to the circular shape of wafers. However, in the photovoltaic field, square silicon wafers do not waste space during cell packaging, so square-shaped wafers are used for photovoltaic cells.
Related:
- Silicon Etching: Gas Choices for Precision and Performance
- Understanding Start Oxide in Semiconductor Making
- Discover 3 TGV Hole Shapes and Their Unique Applications
- Here’s Why Wafer Carriers Always Hold 25 Wafers Only
- Gold in Chips: Costly Yet Crucial to Semiconductors
- IC Quality Control Explained: CP, FT and WAT Testing
- Dry Etching Technology: Principles, Features and Uses
- Everything You Need to Know About Silicon CPO Technology
- Rigid and Flexible Semiconductors: The Future of Electronics
- Silicon Motion Unveils 128TB PCIe 5.0 SSD with SM8366
- PCIe 6.0 SSD Controllers in Development by Silicon Motion
- TSMC 2nm Plant: Fatal Accident Leads to Immediate Closure
- IGBT Power Semiconductors: definition and Applications
- Intel Delays Germany Plant Due to Subsidy Issues
- SEMI: Global Wafer Capacity Growth in 2024 and 2025
- PSMC Partners with Tata for Semiconductor Plant in India
- Understanding How CPU Works Without a Start Location
1. This article is compiled from online sources. If there is any infringement, please contact us for removal.
2. The published content represents the views of the author and not the position of Yunze.