AI summary by TwinMind

How the EUV Mirrors are Made

Asianometry

19m 16s

EUV Optics Challenge Overview

  • Initial belief: optics would be hardest part of EUV lithography
  • Required perfection: wavefront error must be limited to λ/50
  • EUV wavelength: 13.5 nm
  • Allowable total system error: 260 picometers (13.5 nm ÷ 50)
  • Context: single water molecule = 275 picometers wide
  • System assumption: 6 mirrors total
  • Individual mirror error budget: 106 picometers RMS (260 ÷ √6)
  • Mirror surface sensitivity: 4x more sensitive than lens surfaces
    • Light deformed twice: hitting surface + reflecting off
    • Required surface polish: 53 picometers RMS (106 ÷ 2)
    • Equivalent to hydrogen atom radius
  • EUV system 20x harder than DUV system (6 mirrors vs 60 surfaces)

Bragg Reflector Principles

  • Foundation: 1913 work by Sir Lawrence Bragg and William Bragg
  • Bragg equation: quantified crystalline solid behaviors with x-ray reflection
  • Constructive/destructive interference principles
  • Nobel Prize Physics 1915 (Lawrence Bragg age 25, youngest physics winner record)
  • Bragg reflector: stack of multiple alternating reflective layers
  • Requires precise D-spacing and correct angle for constructive interference
  • Reflections occur at borders between layers

Material Selection for EUV Mirrors

  • Best material pairs: greatest refractive index difference
  • High atomic number materials: molybdenum, tungsten, gold (optically denser)
  • Low atomic number materials: silicon, beryllium, carbon
  • Both materials must not absorb EUV strongly
  • Typical stack: 50-70 layer pairs
  • Each layer pair contributes small reflection → cumulative effect

Layer Specifications and Performance

  • Molybdenum layer: ~3 nm thick
  • Silicon layer: ~4 nm thick
  • D-spacing precision critical: 3% deviation = 36% reflectivity loss
  • Maximum theoretical reflectivity: ~70%
  • Practical performance: lower due to material interactions
  • Molybdenum layers: crystalline structure
  • Silicon layers: amorphous structure
  • Temperature effects: materials diffuse forming silicide compound
  • Silicide absorbs EUV radiation → performance degradation

Substrate Materials and Thermal Management

  • Mirror heating during exposure causes expansion/distortion
  • Early substrates: glass ceramics (Zerodur, Clearceram)
  • Alternative: titanium-doped fused silica (ULE ultra-low expansion glass)
  • Ceramics: inconsistent polishing response
  • ULE: suffers from striations affecting refraction indices

Historical Development

  • 1940s: first nanoscale multi-layers (failed due to diffusion)
  • 1971: Eberhard Spiller at IBM researching LED coatings for IBM 3800 laser printer
  • Computer simulation program for light emission optimization
  • October 1971: fabricated 6-layer mirror (magnesium fluoride + gold)
  • 1973: IBM-German collaboration for x-ray telescopes
  • 1976: 9-layer copper-gold mirror, 3% reflectivity at 20 nm wavelength

Early Manufacturing Methods

  • Electron beam evaporation: vaporizes solid sources into gas
  • Advantages: rotating different sources for alternating layers
  • Challenge: controlling deposition rate consistency with temperature fluctuations
  • Built-in x-ray reflectivity monitoring device

Magnetron Sputtering Breakthrough

  • 1978: Spiller met Troy Barbee at Lawrence Livermore National Lab
  • Barbee pioneered magnetron sputtering method
  • Process: atoms ejected into vacuum after high-energy particle bombardment
  • Setup: substrate on rotating table, two magnetron sputtering sources
  • Control parameters: power to sources, distance, rotation speed
  • Achievement: >60% reflectivity (near theoretical maximum)
  • Challenges: stress on nanometer layers, warping effects

Wavelength Selection Decision

  • 1985: work caught attention of Hiroshi Kinoshita (EUV lithography pioneer)
  • 1986: first reflective EUV system demonstration
  • Wavelength candidates: 13.5 nm, 11.4 nm, 6.6 nm, 4.8 nm
  • 6.6 nm and 4.8 nm: eliminated due to photoresist limitations
  • 11.4 nm: required molybdenum-beryllium (beryllium toxicity concerns)
  • Final choice: 13.5 nm with molybdenum-silicon

Production-Level Manufacturing

  • 1992: Kinoshita at NTT Labs partnered with Tinsley (SVG subsidiary)
  • Aspherical mirrors (non-spherical geometry)
  • Four-step process:
    1. Blanchard rotary grinders for initial machining
    2. Spherical grinding to aspheric contour
    3. CNC machining for special features (holes for light passage)
    4. Traditional optical polishing (lapping)
  • Computer Controlled Optical Surfacing (CCOS) for precision polish
  • Custom interferometer for measurement/calibration
  • Magnetron sputtering for molybdenum-silicon layers
  • Peak reflectivity: 65%

Production Scaling

  • Livermore National Lab + FCO Instruments collaboration
  • Production-level magnetron sputtering tool
  • Capability: coat 4 optics simultaneously
  • Trade-off: faster coating but more defects
  • Critical mirrors (photomask blanks): different method required

Corporate Changes and Ion Beam Figuring

  • 2001: SVG sold to ASML (Dutch company)
  • US regulatory requirement: Tinsley sold to American buyer (SSG)
  • Carl Zeiss became primary optics producer
  • Major challenge: substrate polishing to 50 picometers RMS
  • Solution: Ion Beam Figuring (IBF)
    • Accelerated argon ions toward mirror surface
    • Atom-by-atom polishing precision
    • Continuous metrology throughout process
  • Subnanometer metrology breakthrough enabled precision

Error Classification System

  • Three categories by spatial frequency:
    1. Low spatial frequency (figure error):
      • Ripples >1 mm crest-to-crest spacing
      • Causes astigmatism-like effects
      • Light falls in wrong locations
    2. Mid-spatial frequency roughness:
      • Ripples 1 mm to 1 micrometer spacing
      • Creates flares/halos
      • Light scatters into dark areas
    3. High spatial frequency roughness:
      • Ripples <1 micrometer spacing
      • Reduces reflectivity (1 angstrom roughness = 1% reflectivity loss)
      • Across 6 mirrors = 6% total loss

Performance Evolution Timeline

  • 2005 Alpha demo tool: 0.2 nm (200 picometers) RMS mid-spatial frequency roughness
    • Correlated to ~16% flare
  • 2007 Starlith 3100: <0.15 nm RMS
  • 2011 Starlith 3300: ~0.05 nm RMS
  • 3400: significant reduction
  • 3600: performance data not disclosed (proprietary)

Project Outcome

  • Steady progress without dramatic breakthroughs
  • 2005: confidence Zeiss would achieve targets
  • 2012: optics ready for production
  • Bottleneck shifted: EUV light source became the showstopper
  • Optics problem solved, but production delayed by power source issues

TwinMind

Get summaries like this for your own meetings

TwinMind takes notes in the background during meetings, lectures, and calls, then writes the summary for you.

Free plan to get started

No bot joins your calls

Phone, desktop & Chrome apps

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome