Cymer fed a laser 50,000 tin droplets a second to make the light that prints 3nm chips

By the 2000s the industry had run out of wavelengths: to keep Moore's Law on schedule, chipmakers needed lithography at 13.5 nm, a light no laser exists to produce and no lens can transmit. The light had to come from a plasma hot enough to strip tin atoms to charge states Sn IX through Sn XIV, whose emission lines cluster at 13.5 nm — created 50,000 times a second, for years, inside a vacuum vessel, without wrecking the optics that collect the glow.

The obvious attempts failed. Burning solid tin targets coated the collector mirror in debris, and discharge designs put eroding electrodes in the blast zone at low power. Cymer, the San Diego company that already made the deep-ultraviolet light sources for most chip scanners, spent years and capital on the droplet-fed answer: a generator firing ~25 micron molten tin drops at 70 m/s, a pre-pulse that expands each drop into a low-density pancake, and a 10.6 micron CO2 main pulse that ionizes it into a plasma emitting at 13.5 nm with 5–6% conversion efficiency — about 250 watts of usable light.

ASML, the only lithography system maker still standing in EUV, decided the light source was too critical to leave in someone else's hands. On October 17, 2012 it agreed to buy Cymer for EUR 1.95 billion in cash and stock — $20.00 plus 1.1502 ASML shares per share, a 61% premium — explicitly 'to accelerate the development of EUV.' By then ASML's pre-production NXE:3100 scanners had already exposed more than 23,000 wafers at customer sites.

The deal closed on May 30, 2013 at about $3.7 billion, richer than announced because ASML's stock had nearly doubled. Cymer kept running as ASML's US-based source division, still serving rival scanner makers at arm's length.

Feed the plasma droplets; let the target renew itself

Tin is the only practical emitter at 13.5 nm, so the real design problem was not the plasma but surviving it — droplets solve that by making the target disposable.

The two-pulse scheme shapes the target before burning it: the pre-pulse pancake lets the CO2 pulse couple efficiently, pushing conversion efficiency to the 5–6% manufacturing needs.

Hydrogen gas converts deposited tin into stannane gas that pumps away, so the collector mirror cleans itself in place instead of shutting the tool down.

ASML bought the source maker rather than betting its EUV program on supplier negotiations — vertical integration removed the riskiest single component from the supply chain.

When a component keeps getting destroyed by its own work, stop toughening it — make it disposable and feed it continuously. A self-renewing consumable beats armor.

Every leading-edge chip fab now runs on this architecture: ASML is the only company selling EUV lithography systems, and its current-generation sources deliver about 250 W of in-band power at intermediate focus, with high-NA systems targeting 500 W. Collector mirrors still erode — about 0.1–0.3% reflectivity loss per billion pulses — so they are swapped roughly annually, and the tin-droplet generator remains the heart of the machine. Chips patterned by 13.5 nm light sit inside essentially every modern smartphone and data-center processor.