Mask Contamination: The Big Operational Headache for EUV Lithography

Pellicles lasting less than a day of exposures, masks getting dirty at unknown times and getting worn out by cleans

No semiconductor process module is more operationally and executionally sensitive to contamination than the EUV lithography machine. Each EUV exposure is extremely expensive. Any particle on the imaged portion of an EUV mask would print a defect on thousands of wafers in a day. These masks that are loaded in from outside, and may become contaminated in the process. As in DUV lithography, pellicles are supposed to protect the mask surfaces which are imaged from being contaminated. But for EUV, pellicles currently cannot be taken for granted.

The Status of EUV Pellicle Lifetime

S&S Tech, the Korean EUV pellicle maker, published lifetime results for metal silicide and CNT pellicles last year. After 4 hours of exposure for a 400 W source, the transmission of the metal silicide pellicle went from 92% to 90%, and rupture occurred after 5 hours [1]. For the CNT pellicle, lifetime was better at 13 hours in an EUV scanner environment, but a special surface treatment and coating were needed. CNT degradation was already observed after 3 hours of hydrogen radical exposure [2]. These findings indicate that pellicles are still far from providing sufficient protection for high volume manufacturing use.

Keeping the Mask Clean: Another Lifetime Limiter

Without pellicles, particle adders are expected to accumulate on the EUV masks. ASML had targeted one adder every 10000 wafers (about a week) [3]. SK hynix had previously reported values as high as one adder per day [4], but most recently Micron had assumed one adder every 20000 wafers [5]. The fact that we should not expect the masks to stay clean indefinitely means two things: (1) periodic wafer inspection is necessary to minimize loss of yield from dirty masks that were not caught, and (2) upon finding a contaminated mask, it must be cleaned. The periodic wafer inspection is a tradeoff between yield risk and inspection cost [3,6]. Inspections might have to be every 100 wafers or so [3]. However, mask cleans have recently been found to be a lifetime limiter as well [5,7,8].

TSMC found that oxygen diffusing through the ruthenium capping layer from the cleaning ambient led to SiO2 growth that led to peeling in the areas not covered by absorber [7]. Samsung had found that standard sulfuric acid peroxide mixture (SPM) could affect absorber CD after as few as 10 cleans [8]. Micron had done the most recent comprehensive study of cleaning effect on mask performance [5]. Even after one clean, low-n absorber CD and sidewall profile were affected. Some vulnerable 2D features suffered significant depth of focus loss and exposure latitude loss after 10 mask cleans.

Assuming two weeks (~20000 wafers) per mask clean, a lifetime of 10 cleans means replacing an EUV mask twice a year. This is unheard of in DUV lithography. Optical masks have an expected lifetime of 5x compared to EUV masks [9]. If the scanner throughput is improved to more than 1500 wafers/day, then in fact, the lifetime is reduced even further. This may raise the urgency for pellicle lifetime improvement. However, we still expect mask cleaning frequency to be determined by pellicle lifetime. What’s worse, Samsung provides a foreboding heads-up: diffusion of hydrogen molecules to remove contaminants on the EUV mask surface can be impeded by the pellicle [8]. This implies that the EUV mask with pellicle might be subject to more frequent cleaning than without pellicle. The frequencies of wafer inspections and mask replacements are part of the annoying considerations of using EUV lithography.

References

[1] M. Choi, C. Park, and J. Hong, “Development and optimization of metal silicide EUV pellicle for 400W EUV lithography,” Nanotechnology 36, 06LT01 (2025).

[2] N. H. Lee et al., “Development of Pellicle Manufacturing Technology for High-power EUV Lithography,” Proc. SPIE 13686, 136860W (2025).

[3] M. van de Kerkhof et al., “Advanced particle contamination control in EUV scanners,” Proc. SPIE 109570U, (2019).

[4] Y. Hyun et al., “EUV mask particle adders during scanner exposure,” Proc. SPIE 9422, 94221U (2015).

[5] V. V. Nair et al., “EUV Mask Lifetime Evaluation using Stochastic Process Window Through Mask Clean Cycles,” Proc. SPIE 13424, 1342406 (2025).

[6] F. Chen, EUV Lithography Without Pellicles: Accounting for Low Yields.

[7] C-H. Wang et al., “Novel capping layer evaluation for EUV mask,” Proc. SPIE 13687, 1368707 (2025).

[8] J. Choi et al., “Film loss-free cleaning chemicals for EUV mask lifetime elongation developed through combinatorial chemical screening,” Proc. SPIE 9635, 96350C (2015).

[9] G. Scheid, Economics of Mask, October 2023.