The World's Most Important Machine

By around 2015, the semiconductor industry hit a wall with 193nm deep UV lithography — features could no longer be shrunk further with existing technology (2:35). This threatened the decades-long trend of transistor doubling every two years, which had been the primary engine of computing progress since 1965.

Hiroo Kinoshita proposed in the 1980s using ~10-11 nanometer x-ray wavelengths to print much smaller chip features (6:25). The idea was dismissed as impossible because these wavelengths are absorbed by air and lenses, requiring vacuum operation and special reflective optics instead of traditional refractive ones.

Underwood and Barbee pioneered mirrors using alternating ultra-thin layers of materials (e.g., tungsten/carbon, later silicon/molybdenum) that reflect x-rays through constructive interference (7:15). Each layer pair reflects a tiny fraction, but stacking 40-80+ layers achieves ~70% reflectivity — the theoretical maximum for silicon/molybdenum at 13.5nm.

EUV mirrors must be atomically smooth — average bumps only ~2.3 silicon atoms thick (10:08). Scaled to Earth-size, the largest bump on a high NA mirror would be no thicker than a playing card. Zeiss achieved this using ion-beam polishing that nudges atoms into place.