The World's Most Important Machine

Moore’s Law didn’t stall because chipmakers ran out of ideas—it stalled because photolithography hit physical limits, and the industry needed an “impossible” machine to keep shrinking transistors. Extreme ultraviolet lithography (EUV) became that lifeline: a system that can print features at nanometer scales by generating 13.4 nm EUV light, reflecting it off ultra-smooth multilayer mirrors, and aligning multiple chip layers with about one nanometer of overlay error—roughly five silicon atoms.

The bottleneck traces back to how chip patterns get transferred. Traditional photolithography shines light through a mask, but as feature sizes approach the light’s wavelength, diffraction and interference distort the image. Designers can fight this with larger numerical aperture optics, but numerical aperture has a ceiling. The industry therefore kept moving to shorter wavelengths—eventually settling on 193 nm deep UV—until around 2015 when further shrinking became untenable. EUV offered a new route: use much shorter wavelengths (around 13 nm) so diffraction blurs less, but that introduces a cascade of engineering problems.

Early x-ray lithography work hinted at the path but faced skepticism and practical barriers. Hiroo Kinoshita proposed using ~10 nm x-rays, yet x-rays are absorbed by air and require vacuum systems and mirrors smooth on atomic scales. Curved multilayer mirrors—built from alternating tungsten and carbon layers—proved x-rays could be reflected, but the reflectivity and infrastructure demands made the approach feel unrealistic. Similar efforts at Lawrence Livermore National Laboratory and later Bell Labs partnerships helped keep x-ray lithography alive long enough for the field to pivot toward EUV.

EUV’s core challenge is efficiency. EUV light at 13.4 nm must bounce off multiple mirrors (including a reflective reticle), and each reflection loses power. Even with ~70% reflectivity per mirror, repeated bounces drain photons quickly, forcing the system to generate far more EUV power than early prototypes could deliver. The solution required a scalable EUV light source and a way to protect the collector mirror from debris.

ASML’s approach uses laser-produced plasma. A high-power laser hits a stream of tin droplets, heating them to extreme temperatures and producing EUV photons. Tin’s emission spectrum around 13.5 nm improves conversion efficiency compared with xenon, but the droplets also create debris that can foul the collector. The machine therefore runs with low-pressure hydrogen to slow and cool tin particles; any tin that reaches the collector is converted into stannane gas and removed while the system operates. Heat management became its own science problem—shockwaves in the hydrogen were modeled with the Taylor–von Neumann–Sedov framework, guiding the required hydrogen flushing speeds.

Power targets kept moving. Early EUV prototypes could print only about 10 wafers per hour, while manufacturing viability demanded hundreds of wafers per hour. ASML and its partners (notably Zeiss for optics) pushed source power from tens of watts toward 100 watts and beyond, using multi-pulse laser schemes that first “pancake” the droplet and then ionize it more effectively. They also had to solve timing: droplets travel through a turbulent hydrogen flow, so laser pulses must hit each droplet with golf-ball precision through a tornado.

By 2016, orders began flowing, and the latest systems expanded optical performance with higher numerical aperture (0.55) to print smaller features. The result is a machine so complex it’s shipped in thousands of parts and tens of trucks—yet it’s now embedded in the supply chain behind today’s most advanced chips. The central takeaway is that EUV succeeded only after decades of iterative breakthroughs across optics, plasma physics, contamination control, and precision metrology—turning a “brick wall” into a manufacturing reality.