In short: Deep-UV (DUV) lithography prints chip patterns using 248 nm (KrF) and 193 nm (ArF) excimer-laser light because resolution scales with wavelength — shorter light focuses to a smaller spot, so each step down to a shorter wavelength shrinks the smallest feature a stepper can resolve.
Why wavelength decides the node
Photolithography is the step that transfers a circuit pattern from a photomask onto a silicon wafer coated with light-sensitive resist. The hard physical limit on how small a feature you can print comes from diffraction, captured by the Rayleigh criterion:
CD = k1 × λ / NA
Here CD is the critical dimension (smallest printable half-pitch), λ is the exposure wavelength, NA is the numerical aperture of the projection lens, and k1 is a process factor (theoretically floored near 0.25 for single exposure). The equation says it plainly: cut the wavelength and the critical dimension drops in direct proportion. That single relationship is why the industry marched from 436 nm (g-line) and 365 nm (i-line) mercury-lamp sources down into the deep ultraviolet.
Key fact: At a fixed NA and k1, moving from 365 nm i-line to 248 nm KrF improves resolution by about 1.47×; moving on to 193 nm ArF adds roughly another 1.28×.
The excimer-laser sources: KrF and ArF
DUV scanners abandoned the mercury arc lamp for pulsed excimer (excited-dimer) gas lasers, which deliver narrow-band, high-power UV:
- 248 nm — krypton fluoride (KrF): introduced for the 250 nm/180 nm nodes in the late 1990s. KrF enabled the first mass-production sub-quarter-micron logic and DRAM.
- 193 nm — argon fluoride (ArF): the dominant DUV source from the 90 nm node onward. ArF "dry" scanners carried roughly the 90 nm–65 nm range.
- 193 nm immersion (ArFi): a thin film of ultrapure water (refractive index ~1.44) is placed between the final lens and the wafer, raising the effective NA above 1.0 and extending 193 nm light down to the 45 nm node and, with multiple patterning, far below.
Diffraction, in motion
The intuition is optical: light bends around the mask features it passes, and a lens of finite aperture can only capture a cone of that diffracted light. Shorter wavelengths diffract at smaller angles, so more of the diffraction orders fit inside the lens pupil, and the reconstructed image on the wafer stays sharp at a tighter pitch. The animation below shows an exposure wavefront sweeping across a feature and the aerial-image edge tightening as it lands.
Comparing the DUV generations
| Source | Wavelength | Typical nodes | Note |
|---|---|---|---|
| i-line (Hg) | 365 nm | 350–250 nm | Mercury lamp, pre-DUV |
| KrF excimer | 248 nm | 250–130 nm | First true DUV in HVM |
| ArF dry | 193 nm | 90–65 nm | NA < 1.0 |
| ArF immersion | 193 nm (n~1.44) | 45 nm → below | Water raises effective NA > 1.0 |
Key fact: There is no practical excimer line between 193 nm and EUV's 13.5 nm. The gap at 157 nm (F2 laser) was abandoned around 2003 because no durable lens material and resist ecosystem matured, which is why 193 nm immersion plus multiple patterning carried the industry for well over a decade.
Where DUV meets the EDA flow
Shrinking features pushed lithography past simple "shadow printing," and that burden landed squarely on computational lithography and design-for-manufacturing (DFM) tooling. At sub-wavelength pitches the aerial image no longer matches the drawn layout, so the mask itself is modified:
- Optical Proximity Correction (OPC) pre-distorts mask shapes to compensate for diffraction rounding — implemented in tools such as Siemens EDA Calibre nmOPC.
- Resolution Enhancement Techniques (RET) such as sub-resolution assist features and phase-shift masks squeeze k1 lower.
- Multiple patterning decomposition (LELE, SADP) splits one dense layer across two or more DUV exposures; coloring and spacing are verified in Synopsys IC Validator and Cadence Pegasus physical-verification decks.
If you are learning the physical-design and sign-off side, our OPC and RET primer and the predecessor article on g-line and i-line lithography explain how the mask-to-wafer gap is closed. For hands-on tool flows and mentored projects, the CourseTron course catalog walks through a full RTL-to-GDSII lane, and teams needing production tapeout support can look at the Avecas VLSI design services group.
Why shorter light eventually wasn't enough
DUV scaling is a story of diminishing returns. Once 193 nm immersion hit its single-exposure floor around the 38–40 nm half-pitch, each new node required splitting layers into two, three, or four masks — multiplying cost, overlay-error budget, and cycle time. That economic wall is exactly what justified the jump to 13.5 nm extreme ultraviolet (EUV), a 14× wavelength reduction that restores single-exposure printing at the tightest pitches. DUV did not disappear, though: KrF and ArF still expose the majority of layers on virtually every advanced chip, where their lower cost and mature process beat EUV for anything not at minimum pitch.
FAQ
Why 248 nm and 193 nm specifically, not arbitrary values?
These are the natural emission lines of the KrF and ArF excimer gas mixtures. You cannot freely dial an excimer wavelength; it is fixed by the molecular transition of the lasing dimer, so the industry built its optics, resists, and masks around these two stable, high-power lines.
Does immersion change the wavelength of the light?
No. The source stays at 193 nm. The water film has a refractive index of about 1.44, which shrinks the effective wavelength inside the medium and lets the lens collect light at steeper angles, raising numerical aperture above 1.0 and improving resolution without a new laser.
Is DUV obsolete now that EUV exists?
Not at all. Most layers on a modern logic or memory chip are still printed with 248 nm or 193 nm DUV because it is far cheaper and more mature. EUV is reserved for the few critical layers whose pitch is otherwise unreachable without many DUV passes.
What limits how low k1 can go?
Single-exposure k1 is physically floored near 0.25. Below that, image contrast collapses, so designers turn to multiple patterning and strong RET rather than pushing k1 further, which is enforced through DFM decks in Calibre, IC Validator, and Pegasus.
Want to master the lithography-to-layout chain hands-on? Join the CourseTron Nano Fab & Physical Design track and run a real OPC-aware tapeout flow with mentor support.