Explore / Foundry Processes / Lithography: the Resolution Limit and Multi-Patterning

Lithography: the Resolution Limit and Multi-Patterning

Foundry Processes foundational the industry; ASML (Advanced Semiconductor Materials Lithography) why pitches are what they are

The idea

Every layer's pattern is projected from a mask onto photoresist on the wafer. Optics limit how fine a pitch one exposure can resolve:

minimum half-pitch = k1 × λ / NA

λ the light's wavelength: 193 nm for the deep-ultraviolet scanners that carried the industry from 90 nm to 7 nm, 13.5 nm for EUV (Extreme Ultraviolet)
NA numerical aperture of the projection lens; immersion (water between lens and wafer) raised it to about 1.35 at 193 nm
k1 a process factor with a physical floor of 0.25 for a single exposure; resolution enhancement (phase-shift masks, optical proximity correction, off-axis illumination) pushes k1 toward that floor

At 193 nm immersion the floor gives a minimum pitch near 76 to 80 nm. The 20 nm generation needed finer metal than that, so the industry turned to multiple patterning: splitting one layer across two or more exposures (LELE (Litho-Etch-Litho-Etch)) or defining the pitch with self-aligned spacers (SADP (Self-Aligned Double Patterning)). Each carries a design constraint, coloring for the former, a fixed line grid for the latter, that the router must model; the routing section's multi-patterning page works through both.

EUV (Extreme Ultraviolet)

EUV (Extreme Ultraviolet) lithography at 13.5 nm resolves in one exposure what needed three or four 193 nm exposures. It entered high volume at the 7/5 nm generations (2019 onward) and removed multi-patterning from many critical layers, simplifying the rules the router sees. It brings its own limits: photon shot noise (stochastic defects) at low dose, mask defects, throughput, and cost per layer, so foundries still reserve it for the layers that need it and keep 193 nm for the rest. High-NA (numerical aperture 0.55) EUV (Extreme Ultraviolet) tools, arriving in the mid-2020s, extend single-exposure resolution once more.

What the physical-design engineer inherits

Which layers are multi-patterned decides whether the router faces coloring constraints or a fixed line grid on those layers.
Rule complexity follows the patterning: forbidden pitches, restricted jogs, and cut-mask spacing rules trace directly to how the layer is printed.
OPC (Optical Proximity Correction) happens after tape-out, when the mask shop distorts the drawn shapes so the printed ones match; the design sees it only as the DFM (Design For Manufacturability) rules and recommended patterns that keep OPC (Optical Proximity Correction) solvable.

Pros, cons, and the trade-off

Strengths

Multi-patterning extended 193 nm optics through four generations no one expected them to reach.
EUV (Extreme Ultraviolet) restored single exposure on critical layers and simplified their design rules.

Weaknesses

Multi-patterning multiplies mask cost, process steps, and overlay error, and pushes constraints into design.
EUV (Extreme Ultraviolet) tools cost hundreds of millions each, and stochastic defects set a dose-versus-throughput trade every layer negotiates.

The trade-off

Resolution has been bought twice: first with design constraints (multi-patterning), then with capital (EUV (Extreme Ultraviolet)). Either way the router pays in rules, and knowing how a layer is printed is knowing why its rules look the way they do.

Where it stands today

193 nm immersion prints the majority of layers at every node; EUV (Extreme Ultraviolet) prints the critical layers at 7 nm and below at the leading foundries.
Multi-patterning persists on layers where EUV (Extreme Ultraviolet) is not used, and returns even for EUV (Extreme Ultraviolet) at the finest pitches of the newest nodes.
Open PDKs (Process Design Kits) are single-patterned; the ASAP7 (Arizona State University 7 nm predictive PDK (Process Design Kit)) kit models EUV (Extreme Ultraviolet) (Extreme Ultraviolet) single exposure on its critical layers by design.