Lithography is the process used to transfer circuit patterns onto silicon wafers during chip manufacturing. It works like a microscopic projector: light shines through a patterned template called a photomask, and a lens system shrinks that pattern down onto the wafer’s surface, where a light-sensitive coating captures it. This single step defines where every transistor, wire, and connection will be on a finished chip, making it the most critical (and expensive) part of semiconductor fabrication.
How the Process Works
The wafer is first coated with a thin layer of photoresist, a polymer that changes its chemical properties when exposed to light. The lithography machine, called a scanner or stepper, projects light through the photomask onto the wafer. Where the light hits, the photoresist either weakens or hardens, depending on the type used. After exposure, a chemical developer washes away the unwanted portions, leaving behind a precise pattern on the wafer surface.
That pattern then serves as a stencil. In subsequent steps, material is either etched away or deposited only in the areas the photoresist left exposed. Once those steps are done, the remaining photoresist is stripped off, and the process repeats for the next layer. A modern processor can require 80 or more lithography passes to build up all its layers.
Positive vs. Negative Photoresist
Two types of photoresist behave in opposite ways. A positive photoresist weakens where light strikes it, so the exposed areas dissolve during development, leaving a hole in the coating. A negative photoresist toughens where light strikes it, so the unexposed areas dissolve instead, leaving a raised feature behind. Positive resists can resolve features down to about 0.5 micrometers, while advanced negative resists can reach as small as 7 nanometers. Positive resists are more expensive but offer better step coverage on uneven surfaces. Negative resists adhere better to silicon and stand up well to wet chemical processing.
What Determines How Small You Can Print
The smallest feature a lithography system can print is governed by physics, specifically a relationship called the Rayleigh criterion. The resolution depends on three things: the wavelength of light used, the numerical aperture (NA) of the lens system, and a process factor that accounts for real-world engineering tricks. Shorter wavelengths and higher numerical apertures both allow finer details.
Numerical aperture describes how much light the optics can gather and focus. A lens with a higher NA bends light at steeper angles, producing a tighter focal point and resolving smaller features. ASML’s most advanced DUV (deep ultraviolet) systems reach an NA of 1.35 using refractive lenses more than 1.2 meters tall, weighing over a metric ton. Their latest EUV (extreme ultraviolet) systems hit an NA of 0.55 using mirrors instead of lenses. Even though 0.55 is numerically lower than 1.35, EUV prints far smaller features because its wavelength is so much shorter.
DUV vs. EUV Lithography
DUV lithography uses ultraviolet light with wavelengths of 248 or 193 nanometers. It has been the industry workhorse for decades, and manufacturers squeezed extraordinary performance out of it using tricks like immersion (putting water between the lens and wafer to increase NA) and multiple patterning (exposing the same layer two or more times with shifted masks). These techniques pushed DUV well beyond its theoretical single-exposure limits, but they add cost and complexity with every extra pass.
EUV lithography uses light at 13.5 nanometers, roughly 14 times shorter than the most common DUV wavelength. That dramatic reduction in wavelength allows single-exposure printing of features that would require multiple DUV passes. The tradeoff is that 13.5 nm light is absorbed by nearly everything, including air and glass. EUV systems operate in a vacuum and use multilayer mirrors instead of lenses. The largest mirrors are about 1 meter across and polished to a smoothness measured in tens of picometers, which is on the scale of individual atoms.
ASML is the only company in the world that manufactures EUV lithography machines. Their newest platform, the TWINSCAN EXE:5000 series, is a “High NA” system with an NA of 0.55 (up from 0.33 in previous EUV systems). It can print with a resolution of just 8 nanometers and is designed to support chip production starting at the 2 nm logic node and extending to multiple future generations.
Why It Matters for Chip Nodes
When you see labels like “3 nm” or “2 nm” describing a chip, those refer to process nodes, which are loosely tied to transistor density rather than any single physical measurement. Lithography is the bottleneck that determines whether a chipmaker can move to the next node, because shrinking transistor features requires printing finer patterns.
TSMC’s N2 process entered volume production in late 2025, and Intel began production on its comparable 18A node the same year. Samsung is targeting mass production of its 2 nm node (SF2) in 2026, with a major new fab in Taylor, Texas. Japan’s Rapidus is aiming for 2 nm class production around 2027. Each of these manufacturers depends on the latest EUV and High NA EUV systems to hit those targets.
Photomasks and Pellicles
The photomask is essentially the master blueprint. It contains the circuit pattern at a larger scale (typically 4x), which the lithography optics then shrink down when projecting onto the wafer. A single advanced mask set for one chip design can cost millions of dollars, and any defect on the mask gets replicated on every wafer it prints.
To prevent dust or particles from landing on the mask and creating defects, a thin transparent membrane called a pellicle is stretched over the mask surface. In DUV systems, pellicles are straightforward. In EUV systems, they’re a serious engineering challenge because so few materials are transparent to 13.5 nm light while also surviving the intense energy levels inside the machine. Developing pellicles that hold up under high-power EUV exposure remains an active area of work across the industry.
Nanoimprint: A Different Approach
Not all next-generation lithography uses light. Canon has developed nanoimprint lithography (NIL), which works by physically stamping the circuit pattern onto the wafer, pressing a patterned template directly into a resin coating. It’s a conceptually simple idea, though the engineering behind it is anything but.
Canon estimates NIL consumes roughly one-tenth the energy of a comparable EUV system. The machines are also far smaller: a cluster of four NIL systems takes up less than half the volume of a single EUV scanner, which is about the size of a double-decker bus at roughly 200 cubic meters. If the technology matures as planned, it could deliver EUV-quality features at significantly lower cost.
The challenges are real, though. In standard lithography, the photoresist is coated evenly across the wafer. NIL can’t do that because excess resin squeezes out from under the stamp during imprinting, creating defects that interfere with the next step. Controlling resin distribution precisely enough for high-volume manufacturing has been one of the core difficulties. Canon shipped its first system to a U.S. defense research consortium, the Texas Institute for Electronics, and the technology is still working toward the defect rates and throughput that high-volume production demands.