AI summary by TwinMind
The Extreme Engineering of ASML’s EUV Light Source
Asianometry
17m 20s
EUV Lithography Introduction
- Extreme ultraviolet lithography: 20+ years development, now commercial reality
- ASML machines: multi-million dollar, use EUV light for impossibly small patterns
- $150 million lithography machine powered by EUV light source
- Focus: lasers firing pulses at tin droplets creating 13.5 nanometer wavelength light
Historical Context & Industry Challenges
- Semiconductor engineers collaborate to decide next-generation chip etching technology
- Previous generation: 193 nanometer wavelength light
- Extended via immersion lithography (shooting light through ultra-pure water)
- Double patterning technique for improved feature density
- Need to move from 193nm to 13.5nm EUV light for advancement
Two Major Technical Requirements
- High conversion efficiency requirement:
- Generate sufficient EUV light at correct 13.5nm wavelength
- Light must emit backwards toward collector optics
- Mirror shaped like headlamp collects/directs light
- Must generate light relatively efficiently
- Debris mitigation requirement:
- Laser firing creates debris contaminating optic system
- Debris dampens performance, affects cost of ownership
Early EUV Development (1986-2002)
- 1986: Japanese researcher Hiro Kunoshita began experimenting with X-ray wavelengths
- Achieved 500 nanometer half-pitch patterns
- 1990: Bell Labs achieved 50 nanometer half-pitch print
- Early experiments: Used synchrotron radiation
- Synchrotron: donut-shaped particle accelerator, continuously circulating beam
- Problem: emitted UV light over too broad field of view
- Insufficient light reached collector mirror
- Goal post movement: Originally scheduled for 100nm node, moved to 70nm, then 50nm
- Cascading effects: Carl Zeiss optics needed more mirrors for higher resolution
- Each mirror reflects maximum ~70% of EUV light
- Required stronger singular plasma light source: minimum 250 watts
Laser Produced Plasma (LPP) Development
- Concept: Point laser at target (gold plate/wire) to create plasma
- Three target materials evaluated: lithium, tin, xenon
- Xenon experiments:
- Used YAG lasers (yttrium aluminum garnet, crystal-powered)
- 1 kilowatt YAG lasers at xenon jet, 10,000 cycles/second
- Sandia Laboratory suppressed xenon ice particles with gas jet
- Problem: xenon ice plasma emitted high energy ions degrading optics
- Tin experiments:
- Tin plasma emits lots of UV light at required wavelength
- Initial conversion efficiency terrible with tin plate
- Tin debris issue: 1.2nm thick tin layer causes 20% collector efficiency decline, 10% total tool output decline
2002 Dallas Industry Decision
- Industry leaders debated xenon vs tin options
- Xenon rejection: fundamental roadblock, conversion efficiency <1%, cannot generate more than several tens of watts
- Tin selection: debris nightmare but no fundamental physics issues
- Industry made massive turn toward tin plasma
Tin Target Optimization
- Original tin plate problems:
- Low spectral efficiency (~1%)
- High opacity plasma blocked 13.5nm light from reaching mirror
- Solution: Concave shape instead of flat
- Creates 13.5nm light with far more intensity
- Improves overall conversion efficiency to acceptable level
- Debris minimization: Smallest possible target = droplets
Dual-Laser Droplet System
- Challenge: Small as possible + concave shaped target
- Solution: Two laser shots hitting falling droplet twice
- 100,000 times per second, "lasers akimbo"
- Pre-pulse (first shot):
- Low energy laser pulse
- Creates pressure waves reshaping tin droplet from sphere to concave sheet
- Must hit exact right location or sheet tilts wrong direction
- Wrong angle reflects second shot back, damaging laser machine
- Main pulse (second shot):
- Amplified carbon dioxide laser, 10 micrometer wavelength
- Remakes droplet into acorn shape, vaporizes creating dense plasma
- Plasma emits 13.5nm light
- Plasma hotter/denser than lightning, but less than sun's core
System Performance Improvements
- 2010: Systems struggled past 10 watt UV power level
- 2016: Pre-pulse addition + more powerful CO2 lasers
- Scaled from 10 watts to 200+ watts
- Final conversion efficiency: 6% (massive leap from initial 1%)
- Target performance: Stable 250 watt power level, 125 wafers/hour operation
- Future pathway: 450 watts possible
Machine Components & Operation
- High powered CO2 laser: includes master oscillator + power amplifier (MOPA)
- Beam transport system: handles laser focusing/positioning
- Vacuum vessel: low pressure hydrogen gas environment
- Houses droplet generator
- Metrology modules for performance monitoring
- EUV collector mirror with hole for laser
- CO2 laser specifications: >20 kilowatts power, requires significant amplification
- Installation: Typically under fab clean room floor due to size
- Droplet specifications:
- 50 kilohertz rate (50,000 cycles/second)
- 30 micrometer wide droplets
- 80 meters/second speed (288 km/hour)
- 1.6mm separation between droplets
- Prevents plasma interference between droplets
Debris Mitigation Strategy
- Collector mirror survival: Must withstand 100+ billion pulses over 30,000 hours
- Threats: Tin debris particles, high energy ions, neutrals from plasma
- Solution: Hydrogen buffer gas pumped into vacuum chamber
- Cools plasma
- Blocks percentage of tin ions
- Etches tin layers from mirror surface
Development Philosophy Comparison
- Previous laser development (193nm): Trial and error approach, basic technologies existed
- EUV development: Required deep physics understanding
- Calculated exact opacity, temperature, droplet size, ion density, emission duration
- Built machine to achieve calculated parameters
- Analogy: Previous = Edison's empirical method, EUV = Tesla's theoretical approach
- Modern day moon landing level achievement
