Three US companies are developing accelerator-based alternatives to the tin-plasma source at the heart of ASML’s EUV scanners. They promise more usable power, cleaner and more controllable light and ultimately the freedom to move beyond today’s 13.5-nanometer wavelength. The biggest troublemaker in the run-up to EUV industrialization has been tamed. ASML last year demonstrated 1,000 watts of EUV source power and says the laser-produced plasma architecture can scale further, a spokesperson confirms. According to the litho giant, other components, such as the optics or the stages, will ultimately limit the performance of its scanners. And yet, there’s not one, not two, but three companies that believe there’s a better way to make EUV light. All hailing from the US, XLight, Inversion Semiconductor and Tau Systems are developing sources that replace exploding tin droplets with high-energy electrons from a free-electron laser (FEL). Instead of heating tin until it radiates at the desired wavelength, they accelerate electrons to close to the speed of light and manipulate their trajectories to make them emit EUV radiation. Until recently, FEL-based EUV didn’t draw much attention. Then Elon Musk gave the technology an unexpected publicity boost. In August, responding to speculation that the enormous Terafab planned by Tesla and SpaceX might contain a particle accelerator for an FEL-based EUV source, Musk replied simply: “FEL FTW” (free-electron laser for the win). The cryptic endorsement put the alternative to tin-based EUV in the spotlight, although so far, no other chipmakers have publicly backed the technology. XLight generates EUV light in a facility outside the fab. Credit: XLight Of the three contenders, XLight has attracted the most financial firepower. In June, the US government finalized 150 million dollars in Chips Act support for the company to construct and demonstrate its first FEL prototype at the Albany Nanotech Complex in New York. XLight plans large, centralized installations next to fabs that can deliver up to 120,000 watts of EUV source power and distribute it to as many as 16 EUV scanners. Tau Systems also wants to build a FEL but is taking almost the opposite approach. Instead of a conventional radio-frequency accelerator stretching over a large facility, it plans to use laser wakefield acceleration: An intense ultrashort laser pulse creates a plasma wave that accelerates electrons to relativistic speeds over just a few centimeters, after which an undulator converts their energy into EUV light. Tau’s goal is to make the whole system compact enough to retain the current one-source-per-scanner architecture. Inversion Semiconductor occupies an intermediate position. It’s developing compact, multi-turn energy recovery linacs, or ERLs, with one lithography source eventually serving three or four scanners. Before tackling lithography, however, it intends to commercialize a smaller synchrotron source for EUV mask metrology, both as a business in its own right and as a steppingstone toward putting particle accelerators into semiconductor fabs. Multilayer Unsurprisingly, the three companies have similar arguments for abandoning tin. The most obvious is power. Particle accelerators can generate very high average beam powers, and accelerator-based EUV concepts producing several kilowatts have existed on paper for years. More source power could provide more photons at the wafer, allowing higher exposure doses without sacrificing throughput. That may become an important feature as shrinking features make stochastic defects more difficult to control. Stochastic defects are random printing errors caused by the statistical nature of photons and chemical reactions in the photoresist. As features shrink, fewer photons contribute to each feature, increasing the chance that random fluctuations produce missing, malformed or misplaced structures. Increasing power mitigates this problem. It’s an open question how far ASML intends to sca