ASML and ZEISS outline Hyper-NA EUV system for 5-nanometer patterning

The story

Engineers at ZEISS Semiconductor Manufacturing Technology and ASML have published a peer-reviewed blueprint for a possible successor to today's most advanced chipmaking systems. Their paper describes "Hyper-NA" extreme-ultraviolet lithography with a numerical aperture of at least 0.75, compared with 0.55 for the current High-NA generation. Reuters reported on October 8 that a commercial system could be about a decade away. ASML has started development, but has not committed to producing the machine.

Numerical aperture measures how much light an optical system can collect and focus. Raising it improves resolution, allowing a lithography tool to project smaller patterns onto a light-sensitive layer on a silicon wafer. The authors, Michael Patra, Heiko Feldmann, Jens Timo Neumann and Gerardo Bottiglieri, examine an optical architecture intended to print features near 5 nanometers. That is more than one-third smaller than the roughly 8-nanometer resolution ASML publishes for its 0.55-NA EXE platform.

The 5-nanometer figure describes an individual printable feature, not a commercial process-node name and not the size of an entire transistor. Modern node labels are branding conventions that combine many dimensions and manufacturing choices. The practical promise is narrower but still consequential: finer lithographic resolution can help chipmakers pack circuitry more densely and improve performance or energy efficiency, even as three-dimensional transistor and packaging designs carry more of the scaling burden.

The proposal is evolutionary in several important respects. Reuters reported that the paper says ASML's existing 13.5-nanometer EUV light source could be reused and that ZEISS can already manufacture mirrors with sufficient precision. The machine would be only modestly larger than current High-NA systems, which are themselves approximately the size of a double-decker bus. Reuse does not make the project simple, but it could preserve substantial parts of the supplier base, cleanroom infrastructure and technical knowledge assembled around EUV.

The hard problems move into the optical and process details. SemiEngineering's summary of the paper identifies reticle three-dimensional effects, polarization, projection optics and the wider EUV ecosystem as core design questions. At higher aperture, light reaches the mask and mirrors at more demanding angles. That can distort how a pattern is transferred, reduce contrast between differently polarized components and tighten tolerances for masks, resists, overlay and contamination control. A theoretical resolution advantage is useful only if a production tool can repeat it across a full wafer at acceptable speed and yield.

Hyper-NA also arrives before High-NA has finished its own industrial ramp. ASML's EXE:5200B uses 0.55-NA optics jointly developed with ZEISS. Reuters says Intel has begun using High-NA in production, while Samsung Electronics and SK Hynix plan to follow in 2028 and TSMC in 2030. Those deployments will generate evidence about throughput, maintenance, process complexity and cost that should shape whether customers want another major aperture transition.

The timing matters because lithography is both a technical bottleneck and an industrial concentration point. ASML is the only commercial supplier of EUV scanners, and ZEISS supplies the precision projection optics. Extending the same platform would reinforce that partnership's influence over the semiconductor roadmap. It would also force mask makers, resist suppliers, metrology companies and chip manufacturers to invest well before a production machine ships. Conversely, failure to make Hyper-NA economical could strengthen alternatives such as additional patterning steps, directed self-assembly or more aggressive three-dimensional integration.

The paper draws a boundary beyond the proposed system. Further resolution gains after Hyper-NA would probably require a shorter light wavelength rather than another increase in aperture, the authors say. That would be a deeper reset: new sources, mirrors, materials and process controls would be needed. In that sense, Hyper-NA may represent an effort to extract one more major generation from 13.5-nanometer EUV before the industry faces a more disruptive change.

INNOVOX analysis: the important development is not a finished machine or a guaranteed product date. It is a technically detailed path that makes the post-High-NA roadmap more concrete. The concept reduces risk by retaining the light source and drawing on mirror capabilities ZEISS says are already within reach. The remaining risks are systemic: mask imaging, polarization, wafer-level process windows, throughput, reliability and capital cost must work together. A scanner that resolves 5-nanometer features but lowers factory productivity would not automatically improve chip economics.

What to watch next is the transition from paper architecture to measured hardware. A prototype optical column, full-field exposure data and demonstrations using production-relevant masks and resists would materially strengthen the case. Customer commitments would show whether leading foundries see value beyond 0.55 NA. Until those milestones appear, Hyper-NA should be treated as a credible research and development direction from the industry's central lithography partnership, not as a scheduled commercial machine.

INNOVOX analysis

Hyper-NA is an attempt to extend optical scaling without replacing the entire EUV ecosystem. Reusing the light source and building on ZEISS mirror capability could reduce one class of risk, but mask behavior, polarization, process control and economics will decide whether the architecture becomes a manufacturable tool rather than an elegant optical design.

What to watch

Watch for a formal ASML product commitment, a prototype optical column, full-field exposure results, compatible masks and resists, throughput and overlay data, and customer roadmaps beyond today's 0.55-NA systems. Cost per useful wafer layer will matter more than resolution alone.