Access Official Technical PDF

Share Summary:LinkedInEmail

EUV lithography reached an important manufacturing milestone in 2018, when exposure systems were reported ready for high-volume production of advanced semiconductor devices. For semiconductor process and manufacturing teams, the significance goes beyond a shorter exposure wavelength: EUV offers a way to simplify patterning steps while meeting demanding requirements for critical dimension (CD) uniformity and overlay. Its production readiness depends on coordinated progress in the scanner, light source, mask, resist and defect-control processes.

Why EUV lithography changes the patterning equation

Optical lithography resolution depends on exposure wavelength, numerical aperture (NA) and a process factor called k1. As device features continued to shrink, 193-nm immersion lithography approached its practical resolution limits. Multiple patterning extended that capability, but forming a single layer could require several exposures. Each additional patterning step adds process complexity and makes CD control, alignment between exposures and defect management more challenging.

EUV uses light at a wavelength of 13.5 nm. At the time described in the 2019 technical report, a single EUV exposure could produce patterns that otherwise required multiple 193-nm immersion exposures. This can reduce the number of patterning steps and associated process time and fab tool footprint. It also avoids some of the pattern-fidelity, CD-uniformity and overlay challenges that can accumulate across multiple exposures. The practical benefit depends on the layer and process: EUV does not remove the need for careful integration, but it can make selected patterning flows simpler.

How an EUV lithography scanner works

EUV light is absorbed by materials, including air, so the exposure path cannot use conventional refractive lenses. Instead, an EUV scanner operates in vacuum and relies on reflective optics. The mask, or reticle, and the illumination and projection optics use mirrors with molybdenum-silicon multilayer coatings to reflect the 13.5-nm light. Because mirror contamination can reduce reflectivity, maintaining a clean vacuum environment is essential; hydrocarbon and hydrogen-oxide contamination must be controlled.

Generating and shaping the light

The source described in the report uses laser-produced plasma. Tin droplets are struck by a high-power, high-frequency CO2 laser, creating plasma that emits EUV light. A collector mirror gathers and focuses that light toward an intermediate focus, from which it enters the exposure system. Inside the scanner, illumination optics shape the light, while projection mirrors image the mask pattern onto the wafer.

Short wavelength alone does not guarantee high-quality imaging. EUV mirror surfaces require tight control of profile and roughness to limit aberration and flare. The mask also presents specific imaging challenges. The angled chief ray creates shadowing effects that can cause differences between features with different orientations, so mask patterns may need pre-correction. The absorber’s thickness relative to the EUV wavelength also makes three-dimensional mask effects important.

Balancing imaging quality, overlay and stochastic effects

Production depends on control across the wafer and between exposures. The report described NXE:3400B performance with full-wafer CD uniformity below 0.5 nm. Matched-machine overlay was reported at 1.1 nm on some systems, while a 2-nm specification was consistently met. These results were presented as meeting the requirements for 5-nm-generation logic and 16-nm-generation DRAM devices. Improvements to reticle and wafer stage clamp flatness, along with calibration, contributed to the reported overlay progress.

Another consideration is local CD uniformity (LCDU), closely associated with line-edge roughness and stochastic variation. EUV photons have higher energy than 193-nm DUV photons, and the report notes greater stochastic noise at the same dose. The resulting pattern variability is influenced by factors including resist behavior, image quality, dose and normalized image log slope. Improving resist performance, optimizing mask bias and improving the image are therefore important to controlling LCDU and total CD uniformity.

Productivity and defect control for production

Source power is a major contributor to scanner productivity. The 2019 report described EUV source power above 250 W and demonstrated throughput exceeding 140 wafers per hour at a dose of 20 mJ/cm2 on the NXE:3400B configuration. Source development included improvements to tin target formation, conversion efficiency and CO2 laser output. The report also noted that transmission and exposure overhead affect throughput, and that excess source power could support higher throughput or help extend collector lifetime.

Productivity gains must be balanced against reticle protection and defectivity. The report described a reduction in reticle front-side particle adders from roughly 100 per 10,000 exposed wafers to fewer than 10 in the latest configurations discussed. A roadmap aimed at fewer than one adder per 10,000 wafers addressed particle generation, release and transport to the reticle.

EUV pellicles are another part of this control strategy. A pellicle protects the mask from particles, but its membrane must combine high, uniform EUV transmittance with heat resistance, mechanical strength and durability under illumination. The work described focused on polycrystalline silicon-based membranes, with further development needed in cleanliness, transmittance uniformity and power capability. Because available inspection tools could not inspect through the pellicle material, a detachable frame concept allowed pellicle removal for mask inspection.

High-NA EUV and the next scaling challenge

At the time of the report, development had begun on a higher-NA EUV system at NA 0.55, compared with 0.33 for the systems discussed. The aim was to support continued scaling and EUV single-patterning at smaller pitches. For some structures, a higher-NA system could enable a single exposure where a lower-NA approach would need multiple exposures, improving effective throughput for that patterning task.

Moving into lower-k1 imaging remains dependent on more than scanner specifications. The report identifies resist and mask-material improvements as necessary to extend EUV single-patterning capability. Process engineers must consider these materials alongside optics, dose, mask effects and the patterning requirements of each layer.

Conclusion

EUV lithography’s transition toward high-volume production was built on progress across the complete exposure ecosystem: high-power sources, reflective optics, precise stages, improved overlay, reticle defect control and pellicle development. For manufacturing decision-makers, the practical case is the potential to replace selected multi-exposure flows with simpler single-exposure patterning while preserving demanding imaging performance. Continued scaling, however, requires joint attention to scanner productivity and to resist, mask and defect-control challenges.

Access Official Technical PDF

About The Author