Extreme ultraviolet lithography

Based on Wikipedia: Extreme ultraviolet lithography

The Machine That Was Supposed to Be Impossible

There is a machine so complex, so demanding, and so expensive that for decades the semiconductor industry believed it could never be built. It weighs nearly 200 tons. It costs around 180 million dollars. It consumes enough electricity to power a small town. And as of 2025, only one company on Earth knows how to make it.

This is the story of extreme ultraviolet lithography—a technology that experts said would never work, that nearly killed Moore's Law, and that has become the most critical chokepoint in the global technology supply chain.

What Lithography Actually Does

To understand why this machine matters, you need to understand how computer chips are made. The process is surprisingly similar to developing photographs in a darkroom—if that darkroom existed at the atomic scale and required the most precise engineering humans have ever achieved.

A computer chip is essentially a sculpture carved from silicon. The "chisel" is light. You shine light through a stencil called a photomask, and that light creates a pattern on the silicon wafer below. Where light hits, the material changes. Then you wash away either the changed or unchanged material, leaving behind the intricate circuits that make modern computing possible.

The catch is that light has a fundamental limit: it cannot create features smaller than its own wavelength. It's like trying to paint fine details with a brush that's too thick. The brush just smears over the small stuff.

For decades, chipmakers played a cat-and-mouse game with this limit. In the 1960s, they used visible light at 435 nanometers—about the wavelength of deep blue light you can see. Then they moved to ultraviolet at 365 nanometers. Then "deep" ultraviolet at 248 nanometers, followed by 193 nanometers.

Each jump required reinventing the entire manufacturing process. New light sources. New optics. New materials. New everything.

By the early 2000s, the industry hit a wall. They had squeezed everything possible out of 193-nanometer light. The next step would require "extreme" ultraviolet light at just 13.5 nanometers—a wavelength so short it borders on what physicists call soft X-rays.

And extreme ultraviolet light has a rather inconvenient property.

It is absorbed by literally everything.

When Light Itself Becomes the Enemy

Normal light passes through glass. That's why we can make lenses. That's why we can make windows. That's why cameras and microscopes and all of optical engineering works.

Extreme ultraviolet light does not pass through glass. It does not pass through air. It does not pass through anything. Every material on Earth drinks it up like a sponge drinks water.

This creates an engineering nightmare of staggering proportions.

You cannot use lenses to focus extreme ultraviolet light. Lenses are made of glass. Glass absorbs the light. Your lens becomes an opaque wall.

You cannot use air in your system. Air absorbs the light. The photons vanish before they reach the silicon.

You cannot even use normal mirrors. The reflective coating would absorb the light instead of bouncing it.

The only solution is to work in a vacuum, using specially engineered mirrors that exploit a quantum phenomenon called Bragg diffraction. These mirrors are made of 40 to 50 alternating layers of molybdenum and silicon, each layer just a few atoms thick. When extreme ultraviolet light hits this multilayer stack, tiny reflections from each interface add up constructively—like sound waves combining in a concert hall—to bounce back about 70 percent of the light.

Seventy percent sounds good until you realize the machine needs eleven of these mirrors in sequence.

Do the math: 0.7 multiplied by itself eleven times equals about 0.02. Only two percent of the light that enters the system reaches the silicon wafer. The other 98 percent is absorbed by the mirrors themselves, turning into heat that must be precisely managed.

Creating Light from Molten Metal

Here is how you generate extreme ultraviolet light: You shoot tiny droplets of molten tin with a carbon dioxide laser.

That is not a simplification. That is literally what happens.

The machine generates 50,000 tin droplets per second, each about 25 micrometers across—roughly the diameter of a thin human hair. A laser blasts each droplet twice. The first pulse flattens the droplet into a pancake shape. The second, more powerful pulse vaporizes it into a plasma hotter than the surface of the sun.

In this plasma, tin atoms are stripped of many of their electrons, creating highly charged ions. As these ions recapture electrons and drop to lower energy states, they emit photons at precisely 13.5 nanometers—the wavelength the system needs.

The physics here is beautiful but unforgiving. Only tin ions in very specific charge states—from nine-times ionized to fourteen-times ionized—emit light at the right wavelength. Ions in other charge states produce unusable light at other wavelengths. And the same plasma that creates the light also absorbs it, because the ions gobble up the photons they just emitted.

This is why extreme ultraviolet sources are so inefficient. The wall-plug efficiency—the ratio of usable light output to electrical power input—is around 0.02 percent. To generate 200 watts of extreme ultraviolet light at the point where it enters the optical system, you need about one megawatt of electrical input power.

A conventional lithography tool uses 165 kilowatts.

The Debris Problem

Vaporizing 50,000 tin droplets per second creates another problem: what happens to all that tin?

Some of it shoots outward at high velocity. Some condenses into microscopic droplets. Some deposits onto the collector mirror—the first mirror in the optical chain, which is positioned terrifyingly close to the plasma to capture as much light as possible.

If tin accumulates on the collector mirror, it absorbs light instead of reflecting it. The mirror becomes less efficient. Eventually it fails entirely.

The solution involves hydrogen gas. The machine fills its source chamber with hydrogen, which does several things simultaneously. The gas slows down tin ions flying toward the mirror. It pushes back some of the debris. And most cleverly, it enables a chemical reaction: solid tin on the mirror surface combines with hydrogen atoms to form stannane—tin hydride gas—which floats away and can be pumped out of the system.

Even with this cleaning system, collector mirrors degrade about 0.1 to 0.3 percent per billion laser pulses. At 50,000 pulses per second, a billion pulses takes less than six hours. The mirrors lose roughly ten percent of their reflectivity every two weeks.

The collector mirror must be replaced about once a year—an expensive and time-consuming operation for a machine that already costs more than a Boeing 787 Dreamliner.

The Mask That Reflects Instead of Blocks

In conventional lithography, the photomask works by blocking light. A chromium pattern on a quartz plate casts shadows. Simple.

In extreme ultraviolet lithography, light cannot pass through quartz. The mask must work by reflection instead. This inverts the entire concept.

An extreme ultraviolet mask is itself a multilayer mirror—those same 40 to 50 alternating layers of molybdenum and silicon—with a tantalum-based absorbing layer on top that defines the pattern. Where you want light, the multilayer reflects it. Where you want darkness, the absorber swallows it.

But this creates a geometric problem that conventional masks don't have. Because the light comes in at an angle and reflects out at an angle, the features cast shadows. These shadows are asymmetric, distorting the pattern in ways that must be carefully compensated.

Making these masks is extraordinarily difficult. Only two companies in the world—AGC Inc. and Hoya Corporation, both Japanese—produce the blank substrates. The multilayer must be deposited with atomic precision, typically using ion-beam equipment from a company called Veeco. The pattern is written using electron-beam lithography, because even extreme ultraviolet light cannot pattern its own masks—you need something with an even smaller wavelength.

Any defect in the mask will print onto every single chip that mask produces. Finding and repairing defects at this scale requires specialized metrology tools that themselves push the boundaries of measurement science.

The Company That Holds All the Cards

In the early days of lithography, several companies competed to build these machines. Canon and Nikon, the Japanese camera giants, were leaders. An American company called Silicon Valley Group (SVG) was a significant player. And a small Dutch company called ASML was an upstart contender.

When extreme ultraviolet lithography emerged as the next frontier, most experts expected Canon and Nikon to dominate. They had the experience. They had the resources. They had the relationships with chip manufacturers.

But extreme ultraviolet was different.

The fundamental research happened at American national laboratories—Lawrence Livermore, Lawrence Berkeley, and Sandia. These labs were funded by the U.S. government to solve the technical challenges that no private company could tackle alone. And the results of this research were transferred to industry through a carefully structured partnership called the Extreme Ultraviolet Limited Liability Company, or EUV LLC.

Access to this intellectual property required licensing. And here, ASML made a crucial move. In 2001, they acquired Silicon Valley Group, which gave them a direct path to the technology being developed at the national labs.

Canon and Nikon were not granted the same access.

ASML also partnered with ZEISS, the German optics company, which had the expertise to manufacture multilayer mirrors with the required precision. ZEISS calls these "the most precise mirrors in the world"—and they are not exaggerating. The mirrors must be accurate to within a fraction of a nanometer, which means locating individual surface imperfections and knocking off individual molecules using techniques like ion beam figuring.

The first prototype extreme ultraviolet system, completed in 2006, processed one silicon wafer in 23 hours. This was useless for manufacturing—a factory needs to process hundreds of wafers per hour to be economical.

But ASML kept improving. By 2018, they had commercial systems in production. By 2022, their machines could process up to 200 wafers per hour. MIT Technology Review called it "the machine that saved Moore's Law."

ASML's revenue in 2021 reached 27.4 billion euros—dwarfing Canon and Nikon's lithography businesses. They became the world leader in photolithography equipment. More importantly, they became a monopolist in the most advanced technology.

No other company on Earth can build these machines.

The Geopolitical Chokepoint

When a technology becomes critical to national security and only one company can provide it, geopolitics inevitably follows.

Modern semiconductors are essential for everything from smartphones to artificial intelligence to advanced weapons systems. The most advanced chips—those made with extreme ultraviolet lithography—are produced by a handful of companies in Taiwan, South Korea, and the United States. None are produced in China.

The United States has pressured the Dutch government to restrict ASML's sales to China. ASML has followed these export controls and, as of 2025, cannot ship extreme ultraviolet systems to Chinese customers.

China has not accepted this situation quietly. Chinese companies like Huawei and the Shanghai Micro Electronics Equipment company (SMEE) have filed patents related to alternative extreme ultraviolet approaches. The Chinese government has poured resources into domestic development.

In December 2025, Reuters reported that China had secretly completed a prototype extreme ultraviolet machine in Shenzhen. If true, this would be a remarkable achievement—but prototypes are very different from production systems. The Reuters report suggested working chips would not be produced until 2028 to 2030.

The technology gap remains enormous. ASML has been developing extreme ultraviolet lithography for decades, with contributions from national laboratories, university research programs, and hundreds of specialized suppliers. Replicating this ecosystem from scratch is a challenge that may take China many years to overcome.

The Man Who Started It All

The concept of extreme ultraviolet lithography originated with a Japanese engineer named Hiroo Kinoshita, working at Nippon Telegraph and Telephone in the mid-1980s.

Kinoshita first proposed the idea when others considered it absurd. Light at 13.5 nanometers? Mirrors instead of lenses? Vacuum instead of air? The challenges seemed insurmountable.

But Kinoshita persisted. He demonstrated the first extreme ultraviolet images at a 1986 meeting of the Japan Society of Applied Physics. Despite skepticism from colleagues, he continued research and organized joint programs with American researchers in the early 1990s.

Meanwhile, scientists at Bell Labs published a 1991 paper demonstrating "soft X-ray projection lithography" at 13.8 nanometers—confirming that the physics was sound even if the engineering remained daunting.

Kinoshita's persistence, combined with massive investments from the U.S. national laboratories and private industry, eventually made the impossible possible. But it took nearly four decades from concept to commercial production.

What Comes Next

Extreme ultraviolet lithography has enabled chip manufacturers to reach 5-nanometer and 3-nanometer process nodes—transistors so small that a single silicon atom is several percent of their total size. Further shrinking approaches fundamental physical limits.

But the industry is not standing still. High-NA extreme ultraviolet—machines with improved numerical aperture that can print even smaller features—are already in development. These systems will require even more powerful light sources, pushing beyond 500 watts. They will demand even more precise optics.

And they will cost even more than today's machines.

The extreme ultraviolet lithography market is projected to grow from about 9 billion dollars in 2024 to over 17 billion dollars by 2030. This growth reflects insatiable demand for smaller, more powerful, more efficient chips—for smartphones, for artificial intelligence, for autonomous vehicles, for applications we haven't yet imagined.

The machine that was supposed to be impossible has become the foundation of modern technology. And the company that builds it has become one of the most strategically important enterprises in the world.

The Numbers That Define Impossibility

Sometimes the best way to appreciate an engineering achievement is through its specifications. Here are some of the numbers that define extreme ultraviolet lithography:

• Wavelength of light: 13.5 nanometers, about 14 times shorter than deep ultraviolet
• Tin droplet production: 50,000 droplets per second
• Laser intensity: 10 to the 11th watts per square centimeter—a hundred billion times more intense than a conventional lithography laser
• Mirror reflectivity: about 70 percent per mirror
• Total light reaching wafer: about 2 percent of source output
• Wall plug efficiency: 0.02 percent
• Power consumption: more than 1 megawatt
• Machine weight: nearly 200 tons
• Machine cost: approximately 180 million dollars
• Wafer throughput: up to 200 wafers per hour
• Companies that can build it: exactly one

Each of these numbers represents a triumph over physics, chemistry, and engineering. Each was once considered an insurmountable obstacle. Together, they represent perhaps the most sophisticated machine humanity has ever created.

And every advanced computer chip in your pocket was made by a machine just like it.