TSMC's Burn Lin swapped a new laser for a puddle of water to keep shrinking chips

Chips are printed with light, and by 2002 the 193nm argon-fluoride laser had reached the edge of what dry optics could resolve. The roadmap's answer was the next wavelength: 157nm fluorine lasers with entirely new optics. As Burn-Jeng Lin's 2004 review laid out, that path was failing on contact with reality — thin-film pellicles could not survive more than about ten exposures, no manufacture-worthy 157nm resists existed, and calcium-fluoride lens material still fell short on quality and quantity.

Lin had proposed immersion at IBM decades earlier and joined TSMC in 2000. His 2002 letter — the 'k3 coefficient' analysis — showed the gap between a scanner's final lens and the wafer was an unused variable in the resolution equation: fill it with water, index 1.44 and nearly transparent at 193nm, and 193nm effectively becomes 135nm with better depth of focus, using the same lenses, lasers and resists.

His 2004 review then simulated the 65, 45 and 32nm nodes across dry and immersed systems, making the case end to end. MIT's Switkes and Rothschild had already shown immersion's ~40 percent resolution gain at 157nm using perfluoropolyethers; water at 193nm did the same trick with a fluid every fab had in the building.

The industry pivoted fast. In January 2004 IBM's director of silicon technology announced it would commercialize lithography run through water, and ASML and Nikon turned immersion into a product line; the 157nm programs wound down without ever shipping a production tool. Immersion scanners became the leading edge's workhorse — ASML's TWINSCAN NXT platforms hit 275 wafers per hour by 2015 — and when a 1.35-aperture tool bottomed out at 36nm, multiple patterning stretched the same technology to 10nm-class chips, like Samsung's 2016 triple-patterned SoC.

Put water between lens and wafer: 193nm acts like 135nm

Water transmits 193nm light well and raises the effective wavelength conversion, so both resolution and depth of focus improve with the same lenses and lasers.

The 157nm path needed pellicles, resists and calcium-fluoride optics that did not exist at manufacturing quality; immersion needed a liquid every fab already had.

The gain compounds with everything already installed: foundries multiplied the value of their ArF fleets instead of waiting on an entirely new supply chain.

It was analysis first, hardware second — Lin's scaling equations let toolmakers and chemists know exactly what to build before committing capital.

When the roadmap says rebuild everything, look for the unused variable in the physics you already own. Lin found one in the resolution equation and spared the industry new lasers, lenses and resists.

Immersion DUV carried every leading-edge node from 65nm down through 45, 32 and into the 10nm class via multi-patterning, and it still prints the majority of layers in advanced chips today. Lin was awarded the IEEE Jun-ichi Nishizawa Medal in 2013 and the SPIE Mozi Award in 2023, and was widely dubbed the father of immersion lithography. The episode became the standard example of beating a wavelength race with a glass of water: when the roadmap looked like a cliff, the cheapest variable in the equation turned out to be enough.