Skip to content

Silicon Wafers: From Sand to Wafer ​

After reading this section, you will be able to describe the four stages a wafer goes through, and use the calculator below to explain why large chips are expensive and why wafers keep getting bigger.

From Quartz to "Nine Nines" ​

The raw material of chips is silicon, and silicon comes from quartz. Getting from quartz to chip-grade silicon takes several rounds of purification:

  1. Metallurgical-grade silicon: smelted from quartz, with a purity of about 98%.
  2. Polysilicon: using the "Siemens process" invented by Siemens in 1954, metallurgical-grade silicon is converted into trichlorosilane, then reduced back into high-purity silicon in a reactor, growing into polysilicon rods 2 to 3 meters long. Electronic-grade polysilicon has impurities below one part per billion. As of 2021, nearly all polysilicon was still produced this way.
  3. Single-crystal silicon ingots: the grain orientations in polysilicon are disordered, while chips need single crystals with atoms arranged in perfect order. The next step is "pulling a single crystal."
  4. Wafers: the silicon ingot is sliced into thin discs and polished. The finished product reaches a purity of 99.9999999% (commonly called "nine nines") or higher.

Wikipedia also mentions a figure: Chinese companies account for about 90% of global polysilicon capacity, and seven of the top ten producers are Chinese companies.

The Czochralski Method: An Accidental Invention ​

Today, about 90% of semiconductor devices use silicon made by the "Czochralski method" (CZ method for short). Its invention was an accident: while writing, the Polish chemist Czochralski dipped his pen not into an inkwell but into molten tin, and when he pulled it out, it drew out a thin metal filament.

Line them upPull a single-crystal silicon ingot with the Czochralski method; click in order
  1. 1?
  2. 2?
  3. 3?
  4. 4?
  5. 5?

Click them one by one, in order.

Slicing and Polishing: Why Wafers Are Round ​

The silicon ingot is a cylinder, so the wafers cut from it are round. The sequence is: slice with a wire saw → machine it flatter → chemically etch away the lattice damage caused by cutting → polish. A 300 mm wafer is about 0.75 mm thick.

Wafers of 200 mm and above have a small notch on the edge, used to mark the lattice orientation.

Wafers Kept Getting Bigger, Then Stopped at 300 mm ​

Wafer diameters kept growing: 1960 saw 25 mm (1 inch), 1969 saw 50 mm, 1976 saw 100 mm... Starting in 2000, the industry moved from 200 mm to 300 mm, which Wikipedia records as cutting the price per chip by about 30% to 40%.

Why does bigger mean cheaper? Because the cost of many process steps (such as one etch) depends on how many wafers were processed, not much on wafer area. With a larger wafer, the same step yields more chips.

The next step should have been 450 mm, but progress stopped around 2016. Wikipedia gives these reasons: 450 mm wafers were projected to cost 4 times more, with equipment costing 20% to 50% more, so the return didn't pencil out. Another intuitive difficulty: a cassette of 25 300 mm wafers weighs about 7.5 kg; switched to 450 mm, it would be 45 kg and would need a crane to move.

Today 300 mm is the standard for the most advanced fabs; many products that don't need the most advanced processes, such as analog, RF, and power chips, are still made on 200 mm wafers.

Try It Yourself: How Many Chips Can One Wafer Yield ​

Drag it yourselfHow many chips can be cut from one waferGeometric layout + Wikipedia estimation formula
Complete chips (by count)
621
Formula-based estimate
640
Usable chips
496

The larger the chip, the fewer fit on a wafer and the higher the proportion wasted at the edges; when the same defect lands on a larger chip, the damaged area is also bigger. These are the two reasons big chips are expensive.

Give it a try:

  • Drag the chip area from 100 to 600 square millimeters: the number of complete chips drops by more than a factor of 6. When big squares are packed into a circle, the wasted fraction at the edge is larger.
  • Switch to a 200 mm wafer, and the count for the same chip drops by more than half. That's why moving to bigger wafers saves money.
  • Yield is an assumption you set yourself. With the same number of defects, more of them land on large chips and ruin them.

The "formula estimate" next to the calculator uses the approximation from Wikipedia's "Wafer (electronics)" article: chip count ≈ wafer area ÷ chip area − wafer perimeter ÷ chip diagonal.

Who Makes Silicon Wafers ​

According to Wikipedia, Japan's Shin-Etsu Chemical is the company with the largest share of the global semiconductor silicon wafer market. The sources don't give shares for other wafer makers, so they aren't listed here.

Quiz timeWhy does moving from 200 mm to 300 mm wafers lower the cost per chip?

Check Yourself ​

  • I can describe the four stages: quartz → polysilicon → single-crystal ingot → wafer
  • I can explain what the Czochralski method does and why wafers are round
  • I used the calculator to see the relationship between chip area and chip count

Next section: Cycles Inside the Fab.

gaodong.aiAboutPrivacyContact

Independent tutorials, not affiliated with any product mentioned. Product names and trademarks belong to their owners. Check official pages for prices and policies.