K1 Factor

The k1 factor is the single number in the Rayleigh resolution equation that measures how hard a lithography process is being pushed relative to the physical limit of its own optics — the term that turns "we're imaging below the wavelength of light" from a vague boast into a precise, auditable difficulty score:

$$\text{CD} = k_1 \cdot \frac{\lambda}{\text{NA}}$$

where $\lambda$ is the exposure wavelength, $\text{NA}$ is the numerical aperture of the projection lens, and $\text{CD}$ is the smallest feature the process can reliably print. Everything else in this equation — EUV's 13.5nm wavelength, High-NA's 0.55 aperture — is a property of the scanner. $k_1$ alone is a property of *how aggressively the process is run* against the physics those scanners provide, and its theoretical floor is fixed by diffraction itself:

$$k_1 \geq 0.25 \quad \text{(the absolute physical limit for a single exposure, no tricks)}$$
k1 Factor — How Far Below the Comfortable Limit a Node Is Running same wavelength and lens, falling k1 buys smaller features at the cost of process window k1 factor → process window / manufacturability → window collapses as k1 → 0.25 k1 ≈ 0.6 comfortable, single-exposure k1 ≈ 0.4 needs OPC + SRAF k1 ≈ 0.28 needs multi-patterning or High-NA a new scanner resets λ or NA; it does not raise k1 back up — that has to be fought for on every layer, every node

Why this one dimensionless number is the real scoreboard behind every "how did they print that" claim in semiconductor lithography:

  • It is the actual metric of process difficulty, not resolution itself. Two fabs printing the identical 20nm feature on the identical scanner can be running very different $k_1$ values depending on how much resolution-enhancement effort (optical proximity correction, sub-resolution assist features, source-mask optimization) they throw at the problem — a lower $k_1$ means the same hardware is being pushed harder, with a correspondingly smaller [process window](process-window) and higher sensitivity to every other variable in the fab.
  • It explains why multi-patterning exists at all. Single-exposure $k_1$ cannot physically go below roughly 0.25 — below that, diffraction simply doesn't deliver enough contrast to resolve the pattern, no amount of software correction fixes it. Splitting one layer's pattern across two or more exposures (LELE, SAQP) is the industry's way of printing features that would require an impossible sub-0.25 $k_1$ on a single exposure, by effectively doubling or quadrupling the achievable pitch density per unit $k_1$.
  • It is the lever behind EUV and High-NA's entire value proposition. Moving from 193nm immersion (λ=193nm) to EUV (λ=13.5nm) isn't just "a shorter wavelength" — it is a way to hit the same CD at a *larger*, safer $k_1$, buying back process window that had been spent fighting multi-patterning complexity at the previous node. High-NA repeats the same trade at the next node by raising NA instead of shrinking λ.
  • Every resolution-enhancement technique on a mask exists to let a fab run a lower $k_1$ without losing yield. OPC, SRAF, and inverse lithography don't change the optics at all — they reshape the mask pattern so that the diffraction-limited image still lands close enough to the target shape even as $k_1$ is pushed toward its 0.25 floor, which is why these techniques become mandatory, not optional, exactly at the nodes where $k_1$ gets tight.
$$\Delta(\text{process window}) \propto -\,\Delta k_1$$

is the practical summary: every tenth of a point shaved off $k_1$ to hit a smaller [chip](chip) feature is paid for somewhere else in the fab — in a narrower [process window](process-window), a higher mask error enhancement factor, or a bigger [WFE](wfe) line item for the next scanner generation that can restore the number back to a comfortable range.

Take k1 factor further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app