AI summary by TwinMind
A Deep Dive Into Canon’s Nanoimprint Lithography
Asianometry
20m 7s
Canon's NIL Machine Announcement
- Mid-October 2023: Canon announced FPA 1200 NZ2C nano imprint lithography (NIL) machine
- Technology works like stamp mechanism
- Canon claims capability to print features suitable for 5nm class chips
Optical Lithography Background
- Uses high energy light, sophisticated optic system, light-sensitive resist chemical
- Stencil pattern onto wafer, then etch to make permanent
- Frequently seen as having most economic value - establishes design performance limits
- Key advantage: throughput - ASML machines process 200-300 wafers per hour
- Machines constantly depreciating, costing fab owners money every second
NIL Technology Origins & Development
- 1995: First proposed by Professor Steven Y. Cho at University of Minnesota (now Princeton)
- Original setup: 55nm thick plastic layer on substrate as resist
- Mask: silicon dioxide on silicon substrate, patterned via electron beam lithography
- Process: heat both mask/resist beyond glass transition temperature, press at 1,900 lb/sq inch, cool, separate
- 1999: Cho founded startup Nanex around technology
- Inspiration from compression masking for thermoplastic polymers (used for CD production)
- Related to Liga microfabrication technique from early 1980s Karlsruhe Institute
NIL Variants
Hot Embossing (Original 1995 Method)
- Heat mask and resist beyond glass transition temperature
- High pressure stamping process
UV NIL (1996)
- November 1996: First presented by Phillips team led by Yan Heisa
- Transparent mask with liquid photosensitive resin resist
- UV light curing instead of heat/pressure
- Room temperature process, no heating required
- Faster overall - skip cooling process
- Better integration with existing semiconductor technologies
Step and Flash NIL
- Late 1990s/early 2000s: Invented at Texas Materials Institute, UT Austin
- Semiconductor Research Corporation provided early funding
- Transparent mask positioned above resist with small gap
- Inject liquid resist, pulled by capillary effect
- UV flash curing, then mask removal
Jet and Flash (J-FIL)
- Later development: UT Austin team rework of Step and Flash
- Inkjet device deposits picoliter droplets of photosensitive resist
- Droplets arranged to accommodate chip design patterns
- Air pressure bends mask for contact from middle outwards
- Careful process to avoid bubbles while maintaining speed
Commercialization Efforts
- 2001: UT Austin professors SV Serena Vassen and C. Grant Wilson founded Molecular Imprints Inc
- 2003: MIT Technology Review named NIL one of 10 emerging technologies to change world
- Appeared on International Roadmap for Semiconductors as Next Generation lithography candidate for 22nm node
- Other candidates: 157nm lithography, EUV, electron projection, x-ray projection
- Multiple startups joined race: Nanex, Molecular Imprints, Obate, EV Group
- Industry ultimately chose immersion lithography for 22nm node, later consolidated on EUV
NIL Technical Challenges
Mask Quality Issues
- Must be 1:1 perfect replica of chip design (no optical shrinking like photolithography)
- Need near-perfect masks and inspection methods for 10-20nm features
- EUV struggled with actinic inspection tools using production-grade UV light
- Leading edge semiconductor mask sets cost tens of millions of dollars
Mask Damage Problems
- Similar to 1960s contact lithography issues
- Hard particles on wafer can permanently damage expensive masks
- Damaged masks create "repeater" defects on wafers
- Molecular Imprints solution: template replication - single master mask creates ~100 cheaper clones
Image Transfer Defects
- Non-fill defects: Resist doesn't properly fill mask grooves/trenches, creates bubbles
- Separation defects: Shear forces during mask removal damage transferred pattern
- Particularly problematic with large, irregular features
Overlay Challenges
- Must position stamp with nanometer accuracy at high speed
- Overlay errors require wafer removal and resist stripping (unlike photolithography re-flash capability)
- Significantly raises cost of errors
- Whole-wafer stamping had serious defect/quality control issues
Throughput Limitations
- ASML fastest machines: 200-300 wafers/hour (5 wafers/minute at high end)
- NIL liquid/semi-liquid handling makes high speed challenging
- Multiple parallel stations difficult to implement
- Molecular Imprints first tool (Impro 500 J-FIL): ~20 wafers/hour throughput
Resource Constraints
- NIL backed by startups vs. EUV/electron projection backed by major corporations
- Molecular Imprints raised $60-70M from VC/industry/government
- Insufficient resources compared to competition with millions of dollars and hundreds of engineers
Alternative Applications
Magnetic Storage Industry (2010s)
- Patterned media/bit patterned media for hard drives
- Magnetic islands dozen nanometers large, each containing one bit
- Industry ultimately used perpendicular magnetic recording and heat-assisted magnetic recording (HAMR)
- Pattern media remains on future roadmaps
Other Applications
- Optical devices, biotech, MEMS applications
Canon Acquisition & Development
- 2014: Canon acquired Molecular Imprints, formed Canon Nanotechnology division
- 2017: Reported overlay tools progress, throughput approaching 80 wafers/hour
- Template replication tool: FPA 1100 NR2 as key enabling technology
- Goal: Each mask clone lasts 1,000+ wafer runs
- 2019: Toshiba Memory customer achieved ~340 runs
- Alignment system: Peer through transparent mask, align using wafer marks to 1nm precision (10 atoms wide)
- Defect mitigation: High performance filters, air curtain to direct particles away
Current Strategy & Future Outlook
- Canon recognizes immediate challenges, particularly defect issues
- Initial target: Semiconductor memory industry (flash memory)
- Memory advantages: 15nm or smaller features needed, more redundant structures easier to imprint, more defect-forgiving
- Logic strategy: Hybrid approach integrating NIL into process flows alongside photolithography
- Potential for special layers or chip types in logic fabs
- Emphasis on industry-wide standards and integration compatibility
