Resolution

Resolution is the smallest feature — a line, a space, or a contact hole — that a lithography system can reliably print to spec across an entire wafer, not just under ideal conditions in a lab. It's governed by the Rayleigh criterion, the same relationship that sets resolving power in any optical system: Resolution = k₁ × λ / NA, where λ is the exposure wavelength, NA is the projection lens's numerical aperture, and k₁ is a process-complexity factor that captures everything about illumination, resist, and mask technique that isn't wavelength or lens.

The same wavelength and lens can still print very different resolutions, because feature geometry itself changes the difficulty. Dense, periodic lines and spaces are the easiest case — their regular pitch produces clean, predictable diffraction orders the lens can capture reliably. Isolated lines are harder, since there's no neighboring pattern contributing extra diffracted light to sharpen the image. Contact holes are the hardest conventional case of all — they're two-dimensional features that need resolution control in both axes simultaneously, not just one. End-of-line features add their own distinct optical challenges on top of that, which is why OPC correction (see our OPC article) treats line ends as a special case requiring their own hammerhead-style compensation.

k₁ is where decades of "resolution enhancement" techniques actually live. The theoretical floor for k₁ is 0.25; practical manufacturing runs closer to 0.28–0.35 depending on feature type, and the entire discipline of computational lithography exists to push k₁ down toward that floor without changing wavelength or lens at all. Phase-shift masks sharpen image contrast through destructive interference (see our photomasks article). Off-axis illumination reshapes which diffraction orders the lens captures in the first place. Multiple patterning techniques — LELE, SADP, SAQP — split one layer's pattern across two to four separate exposures specifically to reach pitches no single exposure could resolve on its own. Inverse lithography technology computes an optimal, often curvilinear mask shape directly from the target print, rather than correcting a conventional rectilinear mask after the fact.

Each full wavelength or NA generation has historically bought the industry roughly a 0.7x resolution improvement, arriving every two to three years — the rhythm that underlies Moore's Law scaling as much as transistor architecture does. g-line and i-line (365–436nm, mercury arc lamp) covered feature sizes down to roughly 250nm; KrF (248nm) and dry ArF (193nm) carried the industry through the 110nm to 65nm range; ArF immersion, by raising NA to 1.35 in water, pushed a single exposure down to roughly 38nm; and EUV's 13.5nm wavelength reaches approximately 13nm at today's 0.33 NA, with High-NA EUV's 0.55 NA targeted at roughly 8nm — the next resolution step the industry is counting on before anyone has to solve what comes after EUV.