High-NA EUV: resolution is not the same as manufacturing yield.

What numerical aperture changes

ASML describes conventional EUV systems with numerical aperture 0.33 and High-NA EUV with 0.55. The increase allows finer optical imaging at the same wavelength. This is a physical capability of the optical system, not an automatic guarantee that every circuit layer should move to the new tool or that a complete chip gains the same amount of density.

Lithography has to transfer a useful pattern repeatedly across real wafers. The printed feature depends on illumination, mask design, resist response, focus, exposure and the subsequent process. A resolution result is therefore one element in a manufacturing chain. The practical comparison should specify the layer, target geometry, acceptable variation and the complete set of steps used to obtain that geometry.

Sources: [1] ASML: lenses, mirrors and numerical aperture

Resolution and depth of focus move differently

A simplified optical relation gives resolution proportional to wavelength divided by numerical aperture, with a process-dependent coefficient. Holding that coefficient fixed, moving from 0.33 to 0.55 gives a ratio of 0.60. This is an illustrative scaling result, not a claim that the minimum usable dimension of every production layer falls by exactly 40%.

Depth of focus scales approximately with wavelength divided by numerical aperture squared, again with process-dependent factors. Under the same simplified assumptions, its ratio becomes (0.33/0.55)^2 = 0.36. The comparison exposes an engineering trade-off: the improved resolving capability comes with a narrower focus allowance. Wafer topography, film uniformity and focus control therefore become part of the benefit, not peripheral concerns.

The stochastic question

At small dimensions, an average printed width is not a sufficient description. Missing features, bridges and local variations can be rare but consequential. A study should distinguish systematic errors from stochastic events and report the area or feature count over which defects were measured. A small, favourable field of view cannot establish the probability of failure across a full die.

We would ask for distributions across wafer position, lots and relevant exposure conditions. Metrology should be linked to electrical tests so that visual conformity is not confused with circuit functionality. The evidence becomes stronger when defect detection, thresholds and sampling are fixed before comparing alternatives, instead of being adjusted after seeing which process appears to perform best.

Compare flows, not individual exposures

A new optical capability may replace some combinations of patterning steps, but the benefit depends on the actual layer and integration strategy. The relevant unit is the complete flow: deposition, lithography, etch, cleaning, inspection and possible rework. A faster isolated step can be offset by additional measurement, a more demanding material stack or a tighter process-control requirement.

ASML's technology explanations help establish the optical direction. For an industrial decision, we would translate that into a layer-specific experiment and avoid treating an equipment roadmap as evidence of a particular customer's yield. An announced capability, a tool installed in a fab and a qualified manufacturing recipe are distinct stages with different acceptance criteria.

Sources: [2] ASML: five technical aspects of High-NA EUV

An illustrative good-die comparison

Consider two hypothetical flows with the same number of candidate dies. Flow A costs 100 arbitrary units per wafer and yields 80% acceptable dies. Flow B costs 120 units and yields 90%. Ignoring every other difference, their relative costs per good die are 100/0.8 = 125 and 120/0.9 = 133.3 units per normalised candidate-die population. Better yield does not automatically make the second flow cheaper.

This is not an estimate of EUV equipment or wafer prices. Real calculations must also account for die area, throughput, tool availability, masks, qualification, inventory and the value of the delivered performance. The example is deliberately simple: it shows why a technically attractive patterning improvement needs a system-level economic comparison rather than a single percentage taken from a presentation.

COTT assessment

High-NA is best understood as an expanded imaging capability coupled to a more demanding control problem. Its value will be demonstrated by robust layers and useful products, not by resolution alone. Our priority would be to connect process-window measurements with electrical behaviour and good-die output under a representative manufacturing schedule.

The same principle applies beyond leading-edge logic. Measurement quality, interfaces between process steps and disciplined uncertainty reporting often decide whether a promising technology can be integrated. This assessment does not imply that COTT owns lithography equipment or operates a semiconductor fabrication line; it identifies the engineering questions we consider important when interpreting public evidence.