AI summary by TwinMind
EUV’s Most Difficult Challenge
Asianometry
14m 17s
EUV Reticle Structure and Critical Requirements
- EUV reticle/mask is mirror with protective capping layer and multi-layer Bragg reflector
- 40-50 alternating layer pairs of molybdenum and silicon
- Mirror layer sits on low thermal expansion glass substrate
- Must be extremely clean and stable across wide temperature range
- Chip design pattern applied using 2-3 absorber layers (tantalum, titanium, or nickel compounds)
- EUV light reflects off reticle like patterned mirror onto wafer
- Particle several microns high can negatively affect reticle flatness
- Must achieve zero defect rate - any defects larger than certain size print on wafers
Four Categories of Mask Defects
- Particle defects: Stray particles invading mask during fab process
- Process defects: Flaws occurring in fab process steps
- Absorber defects: Issues with absorber layer (familiar to fabs)
- Multi-layer defects: New category relating to mirror layer EUV light reflection
- Amplitude defects: Particles/pits on/near mirror surface, absorb EUV light causing contrast issues
- Phase defects: Bumps/defects on mirror surface from substrate particles layered over during deposition
- More problematic - very pronounced and printable
- As small as 1 nanometer height/depth can print onto wafer
EUV Mask Manufacturing Process
- Two overarching processes: produce blank reticle, then pattern with chip design
- ~30 named process steps total
- Absorber layer application comes late in sequence
- Cleaning steps alone make up ~20 of whole sequence
- Patterning step uses traditional 193nm photomask techniques (e-beams)
- Zero defect blanks still called "champions" due to rarity
- Accepting only champions would make mask yield rate abysmal, raising EUV costs
Actinic Blank Inspection Development
- Early inspection used DUV light but mirror layer too reflective
- DUV tools cannot detect defects deeper than 2-13 layer pairs within mirror
- Only EUV light itself (actinic light) can penetrate multi-layers sufficiently
- Required actual EUV light source for blank/mask inspection
- LaserTech and IDEC alliance developed inspection tool based on Japan's AIST concept
- ABICS E120 tool: 4+ meters wide, 5+ meters long - like mini EUV lithography machine
- EUV light source reflects off EUV mirrors, illuminates ~1 square micrometer of blank
- Defect-free areas keep reflected light in "bright field"
- Defects scatter light into "dark field" - detection indicates defect presence
- Tool represents 20+ years R&D, received breakthrough awards 2017-2018
- Cannot inspect substrate surface where many phase defects originate
- Critical for driving down mask blank defects over years
Defect Repair Methods
- Amplitude defects: Remove particles to restore mirror reflectivity
- Can remove up to 20 mirror pairs with acceptable reflectivity loss
- Methods: focused ion beam or e-beams
- Phase defects: Far more challenging, no good repair method yet
- 2002 study tried "banging" bumps back down - unsuccessful
- Current industry standard: Carl Zeiss MERIT photomask repair tool
- Uses low voltage electron beam to fix lines
- Pattern shift mitigation: Reposition chip design to place absorber layer over defect
- Requires special EDA software and exact defect knowledge from actinic inspection
EUV Pellicle Challenge and Solutions
- Traditional lithography uses pellicles: thin organic film over metal frame (~2.5mm)
- IBM invented 1977, industry adopted quickly
- Traditional DUV pellicles: thin polymers, 99% transparent to radiation
- EUV pellicle challenges:
- Most materials strongly absorb EUV light
- Light passes through pellicle twice (before hitting mirror, after reflecting)
- Reduces EUV intensity by 81% (90% squared)
- Weaker light = less throughput = higher capital expenditure per wafer
Pellicle Material Research Timeline
- 2006: ASML led widespread search for suitable EUV pellicle
- Silicon approach: First serious candidate
- Silicon compounds have relatively low extinction coefficient
- Fabricated silicon honeycomb mesh membranes ~10nm thick
- Theoretically could reach 90% EUV transmission threshold
- Problem: Silicon membranes become fragile under high-powered EUV exposure, start wrinkling
- Graphene research: Carbon nanomaterial candidate
- 100x stronger than steel, high thermal tolerance, very EUV transparent
- Problem: ASML couldn't grow enough without serious defects
- 2008: No suitable pellicle found for alpha demo tool
- Dual approach: Traditional pellicle system vs "clean system" without pellicle
- Clean system required extensive supply chain particle control
- Additional engineering to prevent EUV plasma particles reaching mask
- More inspections = less wafer production time
Final Pellicle Solutions
- Best early solution: Polysilicon and silicon nitride film (25-55nm thick)
- Partially met ASML needs but suboptimal
- 2016: ASML still encouraging industry partners for better solution
- Carbon nanotubes breakthrough (first investigated 2015):
- Carbon atoms rolled into tubes
- 96.5% transmissible to EUV light
- 10x stronger than traditional fibers
- Multiple nanotube layers mixed into ink, sprayed onto surface
- Highly customizable properties
- Can enhance other materials (e.g., adding to silicon nitride increases strength/thermal conductivity)
- Freestanding pure carbon nanotube solutions possible
Industry Achievement and Transparency
- EUV pushed limits of what many thought possible
- EUV was 15-20 years late but ultimately successful
- ASML provided remarkable transparency throughout development process
- Explained technical challenges, final goals, expected milestones
- No "aha moment" - long slow trudge of incremental progress
- Achievement represents sustained global semiconductor industry effort
