What Is Etching in Semiconductors? Wet vs Dry Explained

In everyday life

"Etching" was first the name of a printmaking technique. An artist covers a copper plate with a layer that resists chemicals, such as wax. Then they scratch lines through it with a needle to expose the copper. When the plate is dipped in acid, only the exposed lines dissolve into grooves. That gives the printing plate.

Electronics uses the same idea: keep the covered areas and dissolve only the exposed ones. The copper wiring on a printed circuit board is made this way, too, by dissolving the unwanted copper with chemicals ("What Is a Printed Circuit Board?"). Chips do the same thing with far finer patterns.

According to an IEEE Spectrum article (2020), the fine wires in the leading-edge generation of the time are about 36 nm apart. A nanometer (nm) is one millionth of a millimeter. More than 2,000 such lines would fit across the width of a single hair (about 0.09 mm).

Where does etching fit in the process?

A chip's circuits are built by adding a thin film to the wafer and carving that film into the circuit's shape, over and over. Before carving, a pattern is printed onto the film with light. A chemical that changes when light hits it is coated on, and light is shone on it in the shape of the pattern. Then part of the coating is dissolved away. This chemical is called photoresist, or resist for short.

At this point, the pattern exists only in the resist. So the film that the resist does not cover is removed, and the pattern is carved into the film itself. This step is etching.

After etching, the resist has done its job and is removed. In many cases, an oxygen plasma turns the resist into gas, a bit like burning it away. This step is called ashing. The circuit's shape is now left in the film, and work moves on to the next layer.

The steps up to printing the pattern with light are in "How Are Chips Made?" How the light itself works is in "What Is EUV Lithography?" From here on, we look at what happens during etching.

"Wet" dissolves with liquid, "dry" etches with gas

There are two main ways to etch. One is to dip the wafer in a liquid chemical, or spray the chemical onto it, so that the film dissolves. Because it uses liquid, it is called wet etching.

For example, hydrofluoric acid (hydrogen fluoride dissolved in water) readily dissolves films of silicon oxide, the same material as glass. Yet it barely dissolves silicon itself (a University of Texas lab manual). The method is simple, and it is gentle on the wafer.

The other way works inside a chamber with the air pumped out. Electrical energy is fed into a gas there. Electrons are knocked off part of the gas, and it becomes a "plasma" that contains highly reactive particles. No liquid is used, so this is called dry etching. "What Is Plasma?" explains what is inside a plasma.

The big difference between the two is the direction in which they etch. The next section shows why that matters for fine circuits. A side-by-side table is in "Wet vs Dry Etching."

Sideways etching is a problem: isotropic vs anisotropic

A liquid chemical dissolves the film evenly wherever it touches. It eats sideways as fast as it goes down. So it creeps under the edges of the resist, and the cross-section ends up bowl-shaped. Etching at the same speed in every direction is called isotropic etching.

Put simply, while the film is etched 100 nm deep, it also widens by about 100 nm on each side. Say you want to keep a line a few tens of nanometers wide. The film under it would dissolve completely. As patterns got finer, carving them with liquid alone became difficult.

So a way to etch only straight down is needed. A plasma contains particles with a positive electric charge, called ions. Just above the wafer, electric forces pull on the ions, and they rain straight down together. At the bottom of a trench, where ions hit, the reaction runs fast. The sidewalls, which ions rarely hit, are barely etched.

This gives straight, vertical walls. Etching at different speeds in different directions is called anisotropic etching. "What Is Plasma?" explains in detail how the ions come straight down.

This is a simplified picture. Real dry etching adds more tricks, such as coating the sidewalls with a thin protective film while etching. You can also compare the cross-sections in "Isotropic vs Anisotropic Etching."

Etching only the target film: selectivity

Ideally, etching removes only the film you want to remove. In practice, the resist above and the film below also get etched a little. So what matters is how much faster the target film etches than the others. This ratio of etch speeds is called selectivity.

For example, with a selectivity of 20, the film below loses only 1 nm while the target film loses 20 nm. Selectivity changes a lot with conditions such as the type of gas, the pressure and the energy of the ions. So the conditions are chosen separately for each pair of films: the one to etch and the one to keep.

Where to stop? Stop layers and knowing when you're done

The thickness of a film varies slightly across the wafer. When the thinnest spot is fully etched, some film is still left at thicker spots. So etching continues a little longer after the film is gone. This is called overetching.

Etching longer means that, at spots that finished early, the layer below starts to be etched. To prevent this, a film of a different material is sometimes laid under the target film. Under the same conditions, it etches very slowly. This film acts as a "stop layer" and protects the layer beneath. That is why pairs of films with high selectivity matter.

The plasma's glow tells you when etching is done. The color of that glow depends on the kinds of particles in the plasma. When the target film is gone, less material comes off it, and the light in that material's color fades. Spotting this change to know when to stop is called endpoint detection.

When etching deep holes, the resist can wear away before the job is done. In that case, a harder film that etches even more slowly is laid under the resist and used as the pattern's stencil. This film is called a hard mask.

Why are deep, narrow holes hard?

The number that tells how many times deeper a hole or trench is than it is wide is called the aspect ratio. If the depth is 50 for a width of 1, the aspect ratio is 50. The bigger the number, the narrower and deeper the hole.

Deep holes matter most for a type of memory used in SSDs and other devices, called 3D NAND. First, hundreds of layers of material films are stacked. Next, a hole is etched through the layers from top to bottom. Then small parts that store data (cells) are formed along the hole. "How Do SSDs Work?" explains the stacking.

This hole is called a channel hole (or memory hole). A Lam Research engineer gives a rough guide for 128 layers: a diameter of about 100 nm and a depth of about 5–6 µm. A micrometer (µm) is one thousandth of a millimeter, so the aspect ratio is 50–60. Scaled up to a hole 1 cm wide, it would be more than 50 cm deep.

As a hole gets deeper, ions and radicals (highly reactive particles) have a harder time reaching the bottom. The gases made by etching also struggle to escape from the narrow hole. So etching slows down as the hole gets deeper. If ions come in at a slight angle and hit the sidewall, the hole can also bulge or bend partway down.

If the hole's shape is uneven through hundreds of layers, the cells turn out differently from layer to layer. So work goes on to find ways to etch fast while keeping the same width from top to bottom. In July 2024, Lam Research announced that it can etch holes more than 50 times deeper than their width. It says the width stays almost the same from top to bottom. In June 2023, Tokyo Electron announced a technique for 3D NAND with more than 400 layers. The company says it can etch a hole 10 µm deep in 33 minutes.

Both companies' techniques are "cryogenic etching," which cools the wafer far below 0 °C while etching. The performance figures in both cases come from the companies' own announcements. No other organization has checked them.

One atomic layer at a time: atomic layer etching

At the other extreme, sometimes you need to remove only a tiny, exact amount of a very thin film. In the most advanced transistors, a film just a few nanometers thick must be etched by a set amount. Ordinary etching runs continuously. So even a slight error in when it stops turns straight into too much or too little etching.

For this, a method that alternates between two steps is used. First, a gas is made to stick to the surface, so only the top layer of atoms becomes easy to remove. Next, low-energy ions hit the surface and remove only that changed layer. Each step nearly stops on its own once one surface layer is done.

Each cycle removes a fixed amount, about one layer of atoms. The number of cycles sets how much is removed, in fine steps. This method is called atomic layer etching (ALE). In 2014, Lam Research announced that it had added an atomic layer etching capability to one of its existing tools. Lam describes the capability as ready for volume production.

This shows a typical sequence, simplified. Because each cycle removes so little, it takes time, so it is used in steps where precision matters most. How it differs from other plasma etching methods is compared in "RIE vs ICP-RIE vs ALE."

Etching equipment and the major makers

A dry etching tool has four main parts. The first is a vacuum chamber that holds the wafer. The others are a system that feeds in the gases, a power supply that creates the plasma, and a stage that holds the wafer at a set temperature. The major makers are Lam Research and Applied Materials in the US, and Tokyo Electron in Japan. All three present etching tools among their products.

For etching 3D NAND holes, Lam Research and Tokyo Electron have announced new techniques, as described above. Applied Materials offers a tool that etches conductive films for memory such as 3D NAND and for logic chips. For the full range of tool types and makers, see "Semiconductor Manufacturing Equipment."

Safety and the environment

Some of the liquids and gases used in etching are hazardous. For example, hydrofluoric acid can enter the body through the skin, and burn symptoms may show up only later (US NIOSH). Carbon tetrafluoride (CF4), used in dry etching, warms the climate about 6,600 times as much as the same mass of carbon dioxide. It also stays in the atmosphere for about 50,000 years (values from the IPCC Fifth Assessment Report; the warming effect is compared over 100 years). So fabs handle these materials inside closed equipment, and they treat the exhaust in an abatement system before releasing it.

Common misconceptions

By the numbers

A depth of 5–6 µm is less than a tenth of the width of a hair (about 0.09 mm). That sounds very shallow. But for a hole 100 nm wide, it is more than 50 times the width. And a single wafer carries countless such holes, each with the same shape from top to bottom.

Learn more

• In this series: "How Are Chips Made?" (the whole process), "What Is Plasma?" (how ions come straight down), "What Is a Silicon Wafer?", "What Is a Cleanroom?", "How Are Chips Tested?" (inspections that measure etched shapes)
• Compare in a table: "CCP vs ICP Plasma" (ways of making plasma), "SC-1 vs SC-2 vs Dilute HF" (cleaning with chemicals and dissolving thin films), "Damascene vs Subtractive Etch Interconnects"
• Go deeper with terms: ARDE (deeper holes etch more slowly), ICP (high-density plasma), perfluorocarbons (greenhouse gases)
• See it in figures: diagrams of the lithography and etching steps, and of how films are added, are in the list of figures.