What Is Photolithography? How Chip Patterns Are Printed

In everyday life: yellow rooms and "litho"

You may have seen photos of chip factories with rooms lit in yellow. Those rooms are where photolithography is done. The chemicals used in this step react to blue light and ultraviolet light. So that light is taken out of the room lighting. "What Is a Cleanroom?" explains how the rooms in a chip factory work.

"Photo" means light. "Lithography" was originally the name of a printing method that used stone plates. So the word means "printing with light." People often shorten it to "litho." It is also called "exposure," after the step of shining light on the wafer. It is probably one of the first words you hear when you join a chip company.

How it works 1: Coat, expose, develop

The goal of photolithography is to make a "stencil" on top of a film. With this stencil, the next step can etch or treat only the chosen spots. The stencil is made of a chemical that changes how easily it dissolves in a liquid when light hits it. This chemical is called photoresist, or resist for short.

First, liquid resist is dripped onto the center of a fast-spinning wafer. The resist spreads outward into a very thin, even film. The spin speed sets the thickness. The wafer is then warmed gently. This drives off the parts of the liquid that evaporate easily (the solvent), and the film firms up.

Next, a machine that shines the light, called an exposure tool, reads marks made on the wafer. It then lines the wafer up with the pattern of the layer below. After that, it shines light through a master plate that carries the circuit pattern. Only the parts of the resist hit by light change.

With many resists, the wafer is warmed once more after exposure. A substance made by the light reacts further when warmed. This widens the difference in how easily each area dissolves. Finally, a liquid is poured on, part of the resist dissolves, and the pattern remains. This step is called developing.

Once the resist pattern is ready, the film is cut away wherever there is no resist. This step is etching ("What Is Etching?"). Instead of etching, a chosen element can be shot in through the gaps. This changes how easily electricity flows in the chosen spots, and it is called ion implantation. The resist is removed once its job is done, and the process moves on to the next layer.

How it works 2: Positive and negative resists

Resists come in two types that react to light in opposite ways. In one, the parts hit by light dissolve. In the other, the parts hit by light stay. The first is called positive resist, and the second is called negative resist.

Even with the same master plate, the two give patterns with black and white reversed. With positive resist, the spots where the plate lets light through become grooves. With negative resist, only those spots remain. Compare the two in the figure below.

The fine layers of today's chips mainly use positive resist. It makes it easier to form fine patterns with sharp edges. Negative resist is used for very thick films and for places that need a special cross-section shape. It is also used for some layers of the most advanced chips. You can compare the two in the table "Positive vs Negative Photoresist."

How it works 3: Shrinking the master pattern to one quarter

The master plate that carries the circuit pattern is called a photomask. One made for tools that shrink the pattern is also called a reticle. It is a quartz glass plate about 152 mm square, with the pattern drawn in a film that blocks light. "Reticle vs Photomask" explains how the two words are used.

On today's main exposure tools, the mask pattern is made four times as large as the pattern printed on the wafer. The tool's lenses shrink it to one quarter as it prints (as ASML explains). A larger mask pattern is easier to make fine and accurate. Any tiny error or small speck of dust on the mask also shrinks to one quarter on the wafer.

On ASML's tools, one shot covers at most 26 mm × 33 mm on the wafer (for example, in the NXT:1965Ci specifications). To cover a whole 300 mm wafer, the tool shifts position and prints about 100 times. For example, ASML states how many wafers per hour its NXE:3400B can process, a tool it began shipping in 2017. It gives that figure for 96 shots per wafer.

How it works 4: Coating and printing are done by separate machines

Photolithography mainly uses two kinds of machines. One coats the resist, warms it and develops it. It is called a coater/developer, and the industry also calls it a "track." The other is the exposure tool, which lines up the wafer and shines the light.

The two are usually placed side by side and connected directly. A wafer is coated in the coater/developer and passed to the exposure tool. After printing, it comes back to be developed. The result in the resist also depends on the time and temperature after exposure. Running the machines as one line makes it easier to keep these conditions the same.

Exposure tools print in one of two ways. In the past, the main type printed the pattern for one section at once. It then moved the wafer to print the next section. This type is called a stepper. Nikon says its NSR-1010G, released in 1980, was the first stepper produced in Japan.

Today's main type uses light shaped like a narrow band. The mask and the wafer move at the same time, in opposite directions, under the band, so the pattern is traced out as they pass. This kind of tool is called a scanner. Because the pattern is shrunk to one quarter, the mask moves four times as fast as the wafer. The lens only has to print well inside the band, which makes it easier to cover a wide area.

Speed matters too. In materials for investors from November 2024, ASML says its NXT:2100i tool can process 295 or more wafers an hour. Assuming 96 shots per wafer, this article makes a simple calculation of its own. Multiply 295 wafers by 96 shots, then divide by the 3,600 seconds in an hour, and you get about 7.9. In other words, by this calculation the tool prints about 8 times a second.

How it works 5: How fine a pattern can be printed

Light spreads out when it passes through a narrow gap. The finer the mask pattern, the more the light spreads after passing through, and the blurrier the image. If the pattern is too fine, neighboring lines merge and can no longer be told apart.

The shorter the wavelength of the light, the less it spreads. The wavelength is the distance from one crest of a light wave to the next. Also, the wider the angle from which a lens can collect light, the finer it can print. The number that describes this angle is called the numerical aperture (NA).

In short, the finest printable feature is roughly "proportional to the wavelength and inversely proportional to NA." In optics this is called the Rayleigh equation, after the physicist Rayleigh. In practice, it also includes a factor set by tricks in the lighting and the resist. "What Is EUV Lithography?" explains the formula. You can try different numbers in the lithography resolution calculator.

Factories choose the light to match how fine each layer must be. A nanometer (nm) is one millionth of a millimeter. Ultraviolet light with wavelengths of 365 nm, 248 nm and 193 nm is mainly used. The finest layers use extreme ultraviolet (EUV) light, with a wavelength of 13.5 nm. EUV tools are very expensive, so they are not used for coarse layers. "What Is EUV Lithography?" covers how EUV works.

How it works 6: Stacking layers within a few nm

A chip is built by stacking dozens of layers. The pattern of each upper layer must line up exactly with the pattern below. For example, a thin pillar that connects the wires above and below must sit right on top of the lower wire. How accurately a layer is placed relative to the one before is called overlay.

The error allowed is much smaller than the wire spacing. Around 2020, leading-edge chips had fine wires lined up at a spacing of about 36 nm (IEEE Spectrum). So the error must be kept to a few nm. In the same November 2024 materials, ASML says the error between layers printed with two NXT:2100i tools is 1.3 nm or less.

To achieve this, the exposure tool uses light to read marks at many points on the wafer. It works out where the layer below is, and only then prints. The temperature is also kept the same. Silicon expands slightly when it warms, and that alone can cause errors on the nm scale. "What Is a Cleanroom?" also touches on this.

When one exposure is not enough: multiple patterning

There is a limit to how fine an exposure tool can print in one go. Finer patterns are made in several rounds. This method is called multiple patterning.

One way is to split the pattern across two masks, and repeat exposure and etching twice. Every other line goes on the second mask, so the line spacing can be half that of a single exposure. But the second round must line up exactly with the first, which makes the overlay requirement even stricter.

Another way starts from lines printed in one exposure. A film is added on both sides of each line, and that film is used as the new lines. Each line gives two lines, and their positions are set by the film thickness. Both ways add steps, so they cost time and money. "SADP vs SAQP vs LELE" compares the methods.

Why it is such a costly step

Photolithography takes up a large share of the cost of making chips. The International Technology Roadmap for Semiconductors (ITRS) gathered the industry's outlook for chip technology. Its 2007 edition includes an estimate. Counting masks, resist and measurement, lithography made up 30–40% of total manufacturing cost. But the report also says this varies widely with the type and number of products made and the age of the equipment. It is an old estimate, so today's share may be different.

There are three main reasons. First, the step is repeated for every layer. ASML explains that patterns are printed more than 100 times before a chip is finished. Second, the exposure tools themselves are expensive. For the most advanced EUV exposure tools, ASML describes itself in its 2025 annual report as the world's only maker. You can find rough tool prices in "Semiconductor Manufacturing Equipment."

Third, a mask is needed for each type of chip and for each layer. One chip can only be made once dozens of masks are all ready. The more chips are made from the same masks, the lower the cost of each chip.

Who makes them?

Many of the tools and materials for photolithography come from Japanese companies. The table lists what each company says about itself, with the date.

Printing by pressing a mold is called nanoimprint lithography. "Semiconductor Manufacturing Equipment" gives an overview of the tool makers.

Common myths

By the numbers

Lined up, the numbers show how hard this step is. Patterns are printed about 100 times on each wafer, at a pace of nearly 300 wafers an hour. Each time, the error against the layer below is kept to a few nm or less.

Learn more

• Related articles: "What Is EUV Lithography?" (the most advanced light and tools), "What Is Etching?" (the step that uses the stencil to cut), "What Is Thin-Film Deposition?" (the step that adds the film before etching), "What Is a Silicon Wafer?" (why a flat plate is needed), "What Does 2nm or 3nm Mean?" (generation names and real sizes)
• Compare in a table: "EUV vs DUV Lithography," "OPC vs Phase-Shift Mask vs Off-Axis Illumination" (tricks that push toward the limit)
• Try the numbers: in the lithography resolution calculator, change the wavelength and NA to see how the finest printable feature changes.
• See it in figures: figures of the lithography and etching steps are in the list of figures.