A new laser-based EUV source can help crack tough material problems that bedevil 7-nm and under photoresists.
Extreme ultraviolet photolithography, a potential savior of Moore’s Law, has been a long time coming. A dozen years ago, roadmaps called for EUV to arrive in 2011. Not until last year did it finally get up and running.
EUV sources have reached the 200-watt level needed for semiconductor fabrication. Yet, defects in exposed photoresists are limiting yields at the current 7-nanometer node, and bigger problems are looming for future 5- and 3-nm nodes. Now a new laboratory EUV source based on state-of-the-art lasers is giving developers the higher resolution in space and time that they need to understand and solve those problems.
Moving lithography to the EUV band meant big changes in both materials and light sources. The new 13.5-nanometer EUV plasma sources replaced ultraviolet lasers emitting at 193 nm. Photon energy increases as wavelength decreases, so each photon from the new laser-driven plasma EUV source carries 14 times the energy than photons from the old laser source. Higher-energy photons require new photoresist materials—a challenging chemistry problem that remains a work in progress. Newly developed photoresists suffer from seemingly random defects called “stochastic print failures.” This issue has become a top problem in EUV lithography, says John Petersen, principal scientist for exploratory materials at IMEC in Belgium.
“We need to understand the real chemistry of what’s going on,” says Paul van der Heide, director of IMEC’s materials and analysis group. To do that, the company teamed up with KMLabs of Boulder, Colo., to build a high-resolution EUV imaging and ultrashort pulse laboratory in Belgium. Petersen and others described the facility at the SPIE Advanced Lithography conference that took place from 25–28 Feb in San Jose.
The system built by KMLabs produces EUV pulses by focusing high power pulses from an infrared laser into a gas to generate the high harmonics of the laser light. The process generates pulses lasting from picoseconds (10-12) to attoseconds (10-18) at wavelengths that can be tuned between 6.5 and 47 nanometers. The tunable wavelength and adjustable pulse lengths make the high-harmonic source much better for measurement than the brighter plasma source for exposing photoresists. Harmonic generation also produces laser-like EUV light, which gives very high resolution and allows imaging without lenses—a big advantage because solid lenses cannot focus EUV light. The result is a powerful measurement tool that also can perform other measurements that now require massive sources like synchrotrons.
The high harmonic output is bright enough to do high-resolution interferometric imaging on micrometer-scale areas and features down to 8 nm. It also can observe extremely fast molecular dynamics and ionization processes in materials, crucial for understanding chemical processes. Many material suppliers are testing the deposition of thin layers underneath a resist to improve its performance, but they lack a way to probe what happens when the layer is exposed. “We can probe that with this lab,” says Petersen.
The facility can also probe the troublesome stochastic resist failures, providing data that may help researchers prevent them. The next step after that will be identifying and trying to fix other troublesome effects that produce noise above the inevitable photon shot noise caused by the high energy of individual EUV photons. That shot noise floor may threaten scaling to geometries smaller than the 3 nm node.
However, the new EUV system has its own limits. “We are a measurement source, not a source for fab photolithography,” says KMLabs CEO Kevin Fahey. The beam is much too weak for chip fabrication, but it can be focused to illuminate a micrometer-scale area brightly enough for high-resolution interferometric imaging with sub-wavelength resolution to test resists.
“Using high-harmonic generation for photolithography is not outside the realm of possibility,” says KMLabs co-founder Henry Kapteyn, but it would require major new developments and could be decades away.