2 days ago
Want even tinier chips? Use a particle accelerator
Semiconductor chips are among the smallest and most detailed objects humans can manufacture. Shrinking the scale and upping the complexity is a fight against the limits of physics, and optical lithography—etching nanometre-scale patterns onto silicon with short-wavelength light—is its most extreme frontier. ASML , a Dutch firm that builds such lithography tools, takes an almost sci-fi approach by blasting molten tin droplets with lasers in a vacuum to produce extreme ultraviolet (EUV) light with a wavelength of just 13.5nm. Now, some researchers hope to generate more powerful EUV beams with a particle accelerator that propels electrons to nearly the speed of light.
The need for this radical proposal stems from a fundamental limitation of current EUV sources: they struggle to generate enough power to reliably etch circuits onto silicon. In a lithography tool such as ASML's, the EUV beam bounces off nearly a dozen mirrors before it hits the silicon. EUV light is so easily absorbed, though, that even in a vacuum-sealed chamber with ultra-specialised mirrors, each reflection saps 30% of the light's energy. By the time the photons reach the wafer, less than 2% of the original EUV energy remains. Without enough power, reliability and precision plummet.
One way to boost energy is to bombard the wafer with multiple doses of EUV light, a trick that slows down the chip-manufacturing process. The other approach is to increase the power of the photons. ASML's latest rig uses a light source that operates at 500 watts, nearly twice the power of its previous machines. To speed up production or to shrink feature sizes even further, the light source must get stronger. ASML currently has a road map to develop a one kilowatt light source.
A more radical solution is to use a free-electron laser (FEL), where electrons travelling near the speed of light are manipulated to emit EUV radiation. The FEL process begins with a powerful electron gun that injects a beam of the particles into a miniature racetrack. The electrons then pass through a linear accelerator, which propels them to nearly the speed of light. Once accelerated, they enter a roughly 200-metre-long structure called an undulator, where a series of magnets generate a field whose polarity flips periodically. This wiggles the electrons, causing them to emit a beam of EUV photons with a specific wavelength.
Nicholas Kelez, the boss of xLight, a Silicon Valley startup developing FEL-based lithography, described the technology as a more powerful and tuneable 'new light bulb" that he believes can be swapped into existing optical lithography machines. xLight expects to deliver the first commercial system within four years.
Another research group, at the High Energy Accelerator Research Organisation (KEK) in Japan, has already demonstrated the ability to generate light at 20 micrometres (millionths of a metre)—far longer than the 13.5nm wavelengths ASML is capable of but a step towards refining the process. Chinese researchers are also exploring FEL technology in their quest to develop an independent EUV machine.
Generating light using a FEL has some advantages over using lasers. The first is power: a lithography machine based on a FEL-based light source can be around six times more energy-efficient than a laser-plasma tool. Dispensing with molten-tin droplets also reduces the risk of contamination. Tuning such a machine for smaller wavelengths is also, at least theoretically, much easier: all that needs doing is tweaking the settings on the electron gun and the undulator. It would also be cheaper. A single FEL system can be repurposed to provide light for multiple lithography machines, allowing its operator to distribute the fixed costs across multiple chip-etching tools. Nakamura Norio from KEK estimates that the construction cost is around half that of a laser-based EUV tool and the running costs are around a fifteenth.
For now, all this is theoretical. Whereas ASML's EUV machines are proving themselves in high-volume manufacturing, FEL-based lithography is still in the experimental phase. But in the high-stakes world of chipmaking, any edge is worth chasing.
Curious about the world? To enjoy our mind-expanding science coverage, sign up to Simply Science, our weekly subscriber-only newsletter.
