Hong Kong， China, November 10, 2026 — Abstract
Short‑pulse accelerated irradiation widely screens Ru‑capped, carbon‑interlayer and B₄C barrier modified Mo/Si multilayers for EUV lithography. Efficient for preliminary screening, pulsed high‑fluence tests differ physically from continuous steady‑state irradiation at production power density.
Extrapolating pulsed performance to steady‑state introduces condition‑dependent bias with no universal scaling factor, governed by power density, total dose, multilayer microstructure and residual‑gas chemistry. Pulsed threshold advantages cannot guarantee long‑term gains under production‑like continuous irradiation.
Without a well‑characterized public steady‑state reference baseline, cross‑laboratory lifetime‑ranking inconsistencies will arise in production lifetime budgeting, triggering technical disputes, redundant prototyping and flawed maintenance estimates. This paper presents falsifiable deductions and does not negate existing multilayer modification schemes.
Keywords: EUV lithography, multilayer mirrors, lifetime prediction, accelerated testing, metrology standards
Acknowledgements
This manuscript is condensed from the Chinese‑language industry column “Carbon‑Silicon Tradition”; the original article is available upon author request. No conflicts of interest are declared.
1. Introduction and Physical‑Dynamics Background
EUV Mo/Si multilayer mirror lifetime assessment guides lithography maintenance budgeting and material screening. Accelerated testing dominates candidate ranking, yet fundamental physical mismatches between test conditions and production environments remain under‑appreciated. This work only discusses evaluation‑methodology risks and does not invalidate existing modified‑multilayer routes.
Mo/Si multilayers degrade via coupled pathways: interdiffusion, vacancy bubble formation, residual‑gas oxidation¹, and thermally aggravated roughening². Damage rates depend strongly on photon profile, thermal history and cavity chemistry.
Short‑pulse high‑fluence accelerated tests deliver high‑throughput screening with low sample consumption. However, pulsed and continuous production exposures differ fundamentally in photon timing, transient thermal response, defect kinetics and residual‑gas interactions³. Some pulsed‑excited failure modes are non‑dominant under operational power levels.
Simulations propose plausible physical scenarios but cannot replace experimental measurements. Pulsed‑to‑steady‑state bias is state‑dependent with no universal conversion coefficient. Accelerated testing remains valid for its original screening purpose.
2. Hierarchy of Engineering Evidence
Production‑grade lifetime conclusions cannot be drawn from low‑confidence evidence. Evidence follows this priority hierarchy:
3. Simulation: Proposes physical possibilities.
4. Short‑pulse accelerated test: Preliminary screening to eliminate inferior candidates.
5. Sub‑system prototype test: Validates reproducibility under near‑tool conditions.
6. Production‑relevant long‑term continuous steady‑state measurement: Supplies datasets for cross‑laboratory public baseline construction.
A public baseline requires process‑stable Si‑capped Mo/Si reference samples irradiated under production‑like continuous EUV to collect multi‑point dose‑reflectivity decay curves. Until established, any “enhanced long‑term lifetime” claim for modified mirrors remains a non‑comparable candidate judgement. Inconsistent prototype hardware cannot substitute for a universal public reference.
3. Cognition‑Paradigm Inertia
Two cognitive biases distort multilayer lifetime evaluation systematically.
First, the law of the instrument bias: easy access to accelerated hardware breeds over‑reliance on pulsed testing, forming flawed implicit logic: “higher pulsed damage threshold equals longer operational lifetime”.
Second, misattribution bias: prototype‑accelerated prediction divergences are commonly attributed to sample scatter or instrument noise, while systematic methodology bias is rarely challenged.
Screening‑oriented accelerated tests are often misused as conclusive performance metrics without steady‑state calibration. This three‑stage risk‑propagation chain emerges:
1. Material screening: Candidates filtered by pulsed thresholds; extrapolation bias hidden pre‑prototype validation.
2. Sub‑system prototype validation: Top accelerated‑screen candidates may show degraded or reversed long‑run prototype performance. Anomalies are dismissed as random fluctuation, propagating risk downstream.
3. Production‑grade lifetime judgement: Uncalibrated accelerated metrics yield incomparable inter‑lab lifetime rankings for identical multilayers.
Discrepancies stem from systematic test‑setup offsets rather than human error. Outcomes include technical deadlocks, redundant R&D, wasted prototype manpower and miscalculated maintenance budgets, all rooted in missing continuous‑irradiation reference baseline.
4. Academic‑Industrial Incentive & Public‑Goods Dilemma
Current R&D focuses on iterative multilayer material improvement, while open long‑term steady‑state baseline datasets remain scarce.
Baseline measurement is an engineering public good: it demands high‑end hardware, lengthy cycles, carries high failure risk, and yields limited conventional high‑impact publications. Individual organisations bear heavy direct costs while benefits spread to the whole EUV optics community. Market‑driven spontaneous baseline generation is insufficient, leaving fragmented, mutually incomparable published conclusions.
5. Deferred‑Systemic‑Risk Propagation
Upstream material‑screening risks propagate sequentially through engineering workflow.
Laboratory ranking bias → wasted prototype resources → system‑level lifetime budget deviation → inconsistent supply‑chain data and schedule delays
This constitutes deferred systemic risk. Hazards are concealed within early laboratory experiments; financial penalties and schedule slippage only surface in production, at sharply higher remediation cost.
6. Falsifiable Verification Framework
Immediate large‑scale experimental campaigns are not mandated. This framework resolves future lifetime‑ranking conflicts:
7. Deploy reproducible, process‑controlled Si‑capped Mo/Si reference specimens for multi‑point dose‑reflectivity decay measurement under production‑relevant continuous EUV irradiation.
8. Quantify pulsed‑to‑steady‑state extrapolation bias against power density, total dose, multilayer microstructure and residual‑gas chemistry. Build condition‑mapping tables and mark non‑extrapolable regimes explicitly. Prioritise steady‑state measurement without dual‑condition hardware.
9. Re‑calibrate modified mirror performance to the steady‑state baseline and inspect lifetime‑ranking inconsistency.
10. Define safe accelerated‑test extrapolation boundaries and prohibited extrapolation regimes.
11. Set minimum completeness criteria for baseline datasets covering sampling strategy, metadata and sample‑control specifications under practical constraints.
Core objective: convert hidden systematic bias into quantifiable engineering variables.
7. Extended Open‑Ended Engineering‑Management Discussion
These open questions treat baseline metrology as a shared engineering public good and do not endorse a single technical path:
8. With high hardware costs, how can industrial consortia or public platforms share experimental burdens and avoid duplicate baseline work?
9. How can journals recognise high‑value baseline work with limited conventional publishable output?
10. Can a meta‑evaluation framework support bounded conditional extrapolation from accelerated datasets, instead of blanket approval or rejection?
11. What traceability, metadata and long‑term archiving standards sustain cross‑laboratory baseline reproducibility?
12. With incomplete baseline datasets, how should accelerated‑test extrapolation uncertainty feed formally into engineering risk assessment?
13. Conclusion
This article proposes open‑ended engineering propositions covering physics, experimental methodology, cognitive paradigm, industrial incentives and project management. Its explanatory value unfolds alongside EUV optical engineering advances; all deductions await future experimental verification.
Cross‑laboratory lifetime‑ranking divergence during production‑grade mirror qualification will empirically confirm the cognition‑paradigm inertia and deferred‑risk propagation described herein. Adoption of this verification framework can mitigate cross‑laboratory inconsistencies in multilayer‑mirror lifetime budgeting.
References
[1] Hill, S. B., Grantham, S., Tarrio, C., et al. “Critical parameters influencing the EUV‑induced damage of Ru‑capped multilayer mirrors”, Proc. SPIE 6517, Emerging Lithographic Technologies XI, 65171O (2007). https://doi.org/10.1117/12.710432¹
[2] Hill, S. B., Ermanoski, I., Tarrio, C., et al. “EUV testing of multilayer mirrors: critical issues”, Proc. SPIE 6151, Emerging Lithographic Technologies X, 61510Q (2006). https://doi.org/10.1117/12.656207²
[3] Kakutani, S., Niibe, M., Gomei, Y., et al. “Effect of residual gas atmosphere on lifetime of Ru‑capped EUVL projection optics mirror”, Proc. SPIE 6151, Emerging Lithographic Technologies X, 61510R (2006). https://doi.org/10.1117/12.656221³
[4] Bender, H., Wu, C.‑C., Jonckheere, R., et al. “Lifetime test on EUV photomask with EBL2”, Proc. SPIE 11178, Extreme Ultraviolet Lithography X, 111780E (2019). https://doi.org/10.1117/12.2529741
[5] Hayase, N., Nakazaki, M., Yoneda, H., et al. “Impact of Ru capping on blistering in EUV masks”, Proc. SPIE 14273, Photomask Technology 2026, 142730D (2026). https://doi.org/10.1117/12.3116493
Qingjia Huang
Contact Info:
Name: QingJia Huang
Email: Send Email
Organization: Hong Kong 5230 Unlimited Company
Website: https://mbd.baidu.com/ma/s/Ww8A42rG
Release ID: 89208900
If you detect any issues, problems, or errors in this press release content, kindly contact error@releasecontact.com to notify us (it is important to note that this email is the authorized channel for such matters, sending multiple emails to multiple addresses does not necessarily help expedite your request). We will respond and rectify the situation in the next 8 hours.
Source: MarketersMedia