The World's Most Important Machine

Technological concepts & key product/engineering features

Microchips as nanoscopic “computing cities”

• Transistors store bits as ones/zeros.
• Computation speed improves as transistors shrink, which:
  • reduces signal travel time
  • increases transistor density
• Moore’s Law: transistor count roughly doubles every ~2 years for decades, described as stalling around ~2015.

Why EUV became necessary

• Traditional photolithography using 193 nm deep UV (from the late 1990s onward) hit a physical/engineering limit around ~2015.
• To keep shrinking features, the industry moved to extreme ultraviolet (EUV) lithography.

Photolithography fundamentals (limits from optics)

• Uses masks/reticles and forms patterns via diffraction + interference.
• Resolution depends on optical parameters such as:
  • numerical aperture (NA)
  • wavelength (conceptually described using the Rayleigh equation)
• Since NA can’t increase indefinitely, smaller feature sizes eventually require shorter wavelengths.

Switch from x-rays to EUV via multilayer mirror physics

Shorter-wavelength lithography faced constraints:

• x-rays absorb in materials and air
• operation requires vacuum
• standard lenses don’t work, so mirrors are required

A key enabling idea:

• Multilayer reflective mirrors (with tuned tungsten/carbon layering; later discussed as layered EUV designs)

Mirror performance requirements:

• Mirrors must be extremely smooth at atomic scales to prevent scattering (roughness is referenced in “atoms thick” scale).
• Precision alignment is also critical—control down to pico-radians is mentioned.

EUV lithography machine as a “complete system”

The EUV machine combines multiple subsystems:

• A laser-produced plasma (LPP) EUV light source
• A reflective multilayer collector mirror (high sensitivity to contamination)
• An illuminator + reticle
• Projection optics to image and reduce patterns onto wafers

The “hardest and most crucial” step:

• Forming patterns via photolithography through reticles, where EUV is reflected (not transmitted).
• This is necessary because EUV would be absorbed by standard mask materials.

---

EUV light source: core mechanism and performance targets

Tin droplet LPP method

• EUV is generated by hitting a tin droplet with an intense laser to create plasma.
• Tin is preferred over xenon because it provides higher emission in the desired EUV wavelength region, improving conversion efficiency.
• Subsystem performance targets:
  • ~50,000 droplets/second
  • droplets must be identical and perfectly timed

Droplet creation workaround:

• Molten ultra-pure tin is formed into a jet and modulated so chaotic breakup yields regularly spaced droplets before laser impact.

Droplet timing & laser control

• Laser firing is synchronized to the moving droplets (analogy: hitting a distant hole each time).
• Laser curtains/monitoring detect droplet position using scattered photons so the pulse arrival can be scheduled.

Major breakthrough: the “pancake” target

Instead of a single shot, the system uses multiple pulses:

1. Pre-pulse(s) flatten/rarefy the droplet into a pancake
2. Main pulse fully vaporizes/ionizes it into plasma

Benefit:

• Increases EUV output without proportionally increasing debris and neutral reabsorption.

Power milestones (engineering progress)

Referenced milestones include:

• Early Engineering Test Stand: about 9.8 W at 13.4 nm EUV, printing ~70 nm features (feasibility demonstration)
• ~11 W by 2011 (slow output; about 5 wafers/hour)
• 100 W achieved around 2014 after improvements
• By ~2015, expectations moved toward much higher output and throughput (e.g., ~200 W and ~125 wafers/hour mentioned for viability)
• Shipping sources (most recent): around 500 W
• Pulse rates: ~60,000 pulses/second
• Lab roadmap: demonstration of 100,000 droplets/second

---

Throughput economics and why ASML became the bottleneck supplier

Throughput requirement for cost-effective manufacturing

• Early EUV prototypes could print only ~10 wafers/hour.
• For economics, EUV needed hundreds of wafers/hour, running 24/7.

Optical loss from multiple reflections

• EUV imaging relies on many high-reflectivity mirrors:
  • collector → mirrors → reticle → projection optics
• Even with mirrors at ~70% reflectivity, repeated bounces compound losses so that only a tiny fraction of photons reach the wafer.

Only one company could commercialize

• The text frames EUV tooling as constrained such that only ASML (Netherlands) could build working systems at scale.

---

Contamination control: protecting the collector mirror

Hydrogen gas as in-situ cleaning and protection

• Debris from tin plasma can deposit on the collector mirror.
• Hydrogen is used to:
  • slow/cool tin particles
  • reduce deposition
  • convert contaminants into volatile byproducts (described via stannane formation)

Hydrogen must be carefully tuned:

• too little → dirtier mirrors
• too much → EUV absorption and overheating

Shockwave “mini supernova” model for hydrogen flushing

• Ultrafast imaging and the Taylor–von Neumann–Sedov explosion model were used to understand repeatable hydrogen shockwaves after each plasma event.
• This led to required very fast hydrogen flushing speeds (e.g., ~360 km/h).

Oxygen fix for longer continuous running

• Collector top coating degradation caused frequent cleaning (every ~10 hours).
• Engineers observed that opening the machine lets oxygen in, unexpectedly improving mirror cleanliness.
• Experiments with oxygen levels enabled longer continuous operation, improving commercial viability.

---

Overlay and alignment: key “precision manufacturing” constraint

Overlay accuracy

• Maximum overlay error between layers: ~1 nanometer (about five silicon atoms).
• Alignment must remain correct across layers while heat/power variations could otherwise cause misalignment.

Example of “budgeting”

• Engineering described as splitting a nanometer allowance into smaller portions across subsystems—“fight for the nanometer.”

---

Product/production details and scale of the machine

Cleanroom requirements

• ASML assembly described as extremely clean:
  • on the order of 10 particles per cubic meter above a size threshold
• Air showers and strict particulate limits help prevent defects.

Physical scale and logistics

• Shipping includes thousands of components and large-scale transport (many containers, trucks, and aircraft noted).

---

Main speakers / sources (as referenced in the subtitles)

• Derek (narration/technical explainer)
• Casper (interviewer/another presenter)
• Marc (ASML facility guide/engineer)
• Jos Benschop (ASML early researcher; highlighted as a key person)
• Martin van den Brink (ASML CTO; EUV champion)
• Andrew Hawryluk (Lawrence Livermore / EUV x-ray approach researcher; later EUV work)
• Hiroo Kinoshita (Japanese scientist; proposed earlier x-ray lithography concept)
• Bill Brinkman (Bell Labs executive who encouraged support)
• Zeiss (mirror supplier partner referenced)
• Brilliant (video sponsor mentioned at the end)