Why are silicon wafers round?

If you have ever seen an etched silicon wafer, the large cylindrical discs that are chopped up into computer chips, one question immediately comes to mind: Why is it round?

Microchips are rectangular or square, so why do we etch them on circular wafers? Around the edges of every wafer are partial chips that will never make it into a server, smartphone, or laptop. Common sense would suggest that a circular wafer wastes space and energy. Wouldn't it be more efficient to start with a square?

The simple answer lies in metallurgy. Specifically, a century-old procedure for growing crystals called the Czochralski method, which continues to be the industry-standard way silicon ingots are made. Semiconductor manufacturing uses crystalline silicon, and producing flawless silicon crystals requires a careful process that yields a large cylindrical ingot. The ingot is then sliced into the thin wafers we know and use.

The more interesting answer, though, goes beyond manufacturing history, including a combination of physics and economics. Round wafers are more than a legacy convention. Their shape helps manufacturers create the remarkably uniform conditions required to build billions of transistors just nanometers apart. According to IBM Research's experts, the advantages of circular wafers ripple through nearly every stage of semiconductor manufacturing.

The Czochralski method

A modern wafer is sliced from a large cylindrical crystal of silicon. Growing a cylinder is a natural outcome of the crystal-pulling process that has been used by the semiconductor industry for decades. Purified silicon, usually in the form of quartz, is heated to 1,420°C in a furnace, and a single ‘seed’ of crystalline silicon is attached to a rod and dipped into the molten mass. The rod is rotated in one direction as it is lifted from the crucible while the vessel rotates the opposite way, ensuring a circular shape as the ingot grows.

Wafers for semiconductor manufacturing are produced today via a scaled-up version of Polish chemist Jan Czochralski’s invention. In 1916 he wrote a paper on a method he had developed for growing large metallic crystals, titled “Ein Neues Verfahren zur Messung des Kristallisationsgeschwindigkeit der Metalle” (“A New Method for the Measurement of the Crystallization Rate of Metals”). He would die before it was ever used for producing silicon wafers.

As with many other world-changing discoveries like penicillin and popcorn, the Czochralski method started with an accident. Roughly 110 years ago, in Czochralski’s laboratory at the German electrical equipment manufacturer Allgemeine Elektricitäts-Gesellschaft, Czochralski was tasked with developing and improving materials for early electrical cables and machinery. In a moment of distracted exhaustion, the story goes, the chemist accidentally dipped his pen into a crucible of cooling molten tin instead of his inkwell. He drew out a strand of tin, which he eventually demonstrated was a single crystal.

He wrote up his findings, which were published two years later in the German chemistry journal Zeitschrift für Physikalische Chemie. Czochralski never found a use for the technique, which he had proposed as a method for measuring rates of metal crystallization, rather than growing crystals at an industrial scale. For the next few decades until his death in 1953 at the age of 67, he continued his material science work which notably yielded a new alloy for railroad carriage bearings — an invention credited with expanding rail transport in Europe and the USA. It wasn’t until 1954, a year after Czochralski’s death, that a team at Bell Labs created the first silicon transistor.

“I don’t think much has changed since silicon wafer fabrication started,” said Griselda Bonilla, senior technical staff member at IBM Research. Part of the reason is that it works exceptionally well for producing defect-free silicon, she said.

A perfect circle

Crystal growth is only the beginning. There are good reasons the semiconductor industry hasn’t been clamoring to move away from circular wafers. Circular geometry offers important physical advantages. First, there’s heat. A circular geometry evenly distributes thermal and mechanical stress from the center to the edge.

“From the extensive work we have done on stress management in rectangular chips, we know that sharp corners are significant stress concentrators,” said Bonilla. “When you’re trying to grow a crystal that’s defect free, a cylindrical geometry is inherently easier to manage because it distributes stress more uniformly.” Crystal growing has gotten sophisticated enough that nowadays we take silicon purity for granted, but processes matter.

The uniform rotational balance of a circle is also important to chip production processes. Those benefits become even more important once a wafer enters the fabrication facility.

Many semiconductor manufacturing processes rely on rotation. During lithography, for example, a light-sensitive material called photoresist is deposited onto the wafer through a process known as spin coating. The wafer spins at high speed, allowing the liquid photoresist to spread evenly across the surface.

“The fact that you have this circular wafer that is spinning, it allows the chemical to spread perfectly evenly across this round surface due to its rotational symmetry,” said Luciana Meli, a patterning expert at IBM Research. With a square wafer, material would accumulate differently at the corners, creating undesirable variations.

The same principle applies to other steps such as chemical-mechanical polishing (CMP), which smooths the wafer's surface between manufacturing stages. A round wafer allows the polishing process to remain more consistent from edge to edge.

Circular wafers also simplify one of the most important challenges in modern semiconductor manufacturing: uniformity. Today's advanced chips may contain dozens, hundreds, or even thousands of individual dies on a single wafer. Each die must be manufactured almost identically. Any variation in structure can reduce yield and increase costs. According to Eric Miller, whose work at IBM Research focuses on plasma etch, the circular shape simplifies many manufacturing problems, making them easier to control.

“Because the wafers are round, most process variation will be symmetrical and can be addressed with radial adjustments in the process chamber” Miller said. This is especially important for plasma-based etching. Round process chambers and round wafers naturally support symmetric flow and temperature distribution.

Smaller transistors, same size wafers

The need for uniformity is only growing as transistors continue to shrink. IBM recently announced a sub 1-nm node process, resulting in the smallest transistor design in the world. Uniformity is the key connection between the shape of wafers and the semiconductor industry's future, according to Meli. As manufacturers push toward ever-smaller features, they have less tolerance for variation. “It's more and more important to be uniform and have pattern fidelity,” she said.

That emphasis on precision is one reason IBM's researchers see technologies such as High-NA EUV lithography as so important. Future generations of patterning and etching will require even tighter control over dimensions, placement accuracy, and process variability.

So why not switch to square wafers and gain more usable area? The industry's consensus is that the tradeoff is not worth it. “There are many processes that really leverage the fact that these are round,” Meli said. Changing the shape of wafers today would require far more than redesigning the wafers themselves. It would mean rebuilding an ecosystem of tools, chambers, handling systems, recipes, and process controls — an ecosystem that has been optimized over decades.

“If you go away from a round chamber, you lose that built-in symmetry,” Miller said. “That would be the biggest challenge in going away from a round wafer.”

The industry has, however, experimented with changing wafer size — perhaps the only thing that has changed about wafers, according to Bonilla. Some of the first wafers were about 1.5 inches in diameter, and over the years they grew to a few inches, then to today's 300-millimeter (approximately 12-inch) standard. There was some industry effort in the past 20 years to grow wafers to 450 millimeters, which would yield more chips per wafer. While larger wafers offered some technical advantages, the required investment across the semiconductor manufacturing ecosystem to make the transition was significant. The current wafer tooling ecosystem is still set up for 300-millimeter wafers — which strike the balance between yield and cost that the semiconductor industry has agreed is acceptable.

In the end, silicon wafers are round not just because crystals grow that way. Circular wafers create the uniform conditions needed for nearly every major fabrication step, from coating and polishing to deposition and plasma etching. The small amount of lost real estate near the edges turns out to be a modest price to pay for the consistency required to manufacture the world's most advanced chips. As semiconductor technology continues to advance, the uniformity that round wafers bring may become more valuable than ever.