EUV Lithography, Explained Without the Hype

Extreme ultraviolet lithography gets more breathless coverage than almost any other piece of fab equipment, so it’s worth saying plainly what it actually does: it projects a mask pattern onto a photoresist-coated wafer, at a wavelength short enough to resolve features a few nanometres across. That’s it. The complexity is entirely in how hard 13.5nm light is to generate, focus, and not have absorbed by literally everything, including air.

Because EUV light is absorbed by almost every material, the whole optical path — source to mask to wafer — has to run in vacuum, using mirrors instead of lenses (glass absorbs EUV completely). Generating the light in the first place means vaporising tin droplets with a pulsed laser twice, tens of thousands of times a second, and collecting the resulting plasma’s emission with mirrors that have to be replaced on a maintenance schedule because the tin residue degrades them.

What this buys a chip design is fewer lithography passes for the same feature size. Older 193nm immersion tools can print small features too, but only by exposing the same layer multiple times with different masks (multi-patterning) — more steps, more alignment error, more chances to introduce a defect. One EUV pass often replaces three or four multi-patterned ArF passes.

The tradeoff nobody puts on a spec sheet: EUV tools are the single most expensive piece of capital equipment in a modern fab, there are effectively two vendors on Earth who make them, and the machine’s uptime is a genuine node-level risk, not a rounding error.