China’s EUV Metrology Ecosystem Takes Shape Around HHG, SSMB and Plasma Sources

Advanced chipmaking is pushing inspection toward three-dimensional structures and ever-smaller defects. EUV lithography uses a 13.5 nm wavelength, while conventional deep-ultraviolet (DUV) metrology commonly operates at wavelengths such as 193 nm and 248 nm. As critical dimensions, mask defects and transistor geometries shrink, inspection tools need higher spatial resolution without sacrificing speed or damaging the sample.5532

The important distinction is that China is not publicly demonstrating a domestically completed, mass-production-qualified EUV lithography scanner. What is emerging is an ecosystem built around EUV metrology: light sources, multilayer mirrors, vacuum systems, detectors, precision motion platforms, computational models and process-control software. Whether that ecosystem becomes an industrial business will depend on integrating those pieces into systems that can run continuously, remain calibrated and connect measurement data with wafer-fab results.

Why GAA transistors need more advanced EUV inspection

The move from FinFETs to gate-all-around (GAA) transistors introduces more pronounced three-dimensional features. Channels, gates and interfaces can be partially buried beneath other materials, making it difficult for surface-only inspection to determine critical dimensions, sidewall profiles, layer-to-layer alignment and embedded defects. Public reporting identifies GAA structures as a major driver of demand for higher-resolution, non-contact inspection.8

The value of 13.5 nm EUV is not simply that its wavelength is shorter. In scattering measurements, both defect size and illumination wavelength strongly affect signal strength. Shorter wavelengths can improve sensitivity to nanoscale defects, but the actual result also depends on source power, optics, detector noise, sample materials and reconstruction algorithms.18

EUV tools can use scatterometry, coherent diffraction imaging (CDI) and reflective imaging to infer structural parameters from diffraction signals rather than physically contacting the wafer. In one NIST experiment, a tabletop HHG source was used for EUV diffractometry of line-space structures with critical dimensions below 50 nm. Another study found that EUV scatterometry could measure out-of-plane features in two-dimensional interconnect structures with single-nanometer sensitivity.3334

Why EUV masks need “same-wavelength” inspection

An EUV mask is not a simple transparent stencil. It contains a complex multilayer reflective structure, and some phase defects may be difficult to judge at wavelengths other than the one used for exposure. That is why actinic inspection generally means inspection at the same 13.5 nm wavelength used by EUV lithography. It provides a more direct assessment of whether a defect is printable on the wafer. Research has highlighted the importance of 13.5 nm inspection for identifying critical phase defects in EUV multilayer masks.48

This is one reason HHG has attracted attention. It can generate spatially and temporally coherent EUV light, making it useful for coherent scatterometry microscopy, diffraction measurements and lensless imaging. Samsung has publicly described an EUV mask defect review system using a standalone HHG source, showing that HHG has at least entered the technology discussion for specialized mask review and development equipment.42

Four EUV source routes—and their trade-offs

LPP: the clearest power advantage, with the highest engineering burden

Laser-produced plasma (LPP) typically generates 13.5 nm EUV by firing a high-power laser at droplets of molten tin. This is the basic approach used in ASML’s commercial EUV lithography systems.55

LPP’s principal advantage is power. It is the route most closely aligned with the demands of lithography and high-throughput inspection. The trade-off is substantial complexity: high-power drive lasers, tin-droplet control, debris mitigation, collector-mirror contamination, thermal management and maintenance. For inspection equipment, LPP could provide more throughput headroom; for smaller suppliers, it also creates higher capital and service requirements.

DPP: a potentially more compact plasma option

Discharge-produced plasma (DPP) generates EUV plasma through electrical discharge. It may offer a more compact architecture than large accelerator facilities and could suit some metrology power requirements. Chinese reports have identified DPP as one of the domestic routes under exploration.1726

However, DPP still has to address electrode erosion, debris, plasma stability, source lifetime and maintenance. Producing 13.5 nm light is only the first step; a production tool must operate reliably over long periods and within a fab’s maintenance schedule.

FEL and synchrotron sources: powerful research platforms, not ordinary fab-floor tools

Free-electron lasers (FELs) and synchrotron sources offer high brightness, coherence and tunability. They are valuable for reference measurements, materials research, mask analysis and process development. But they generally depend on large accelerator facilities whose size, cost and operating requirements make direct deployment on a conventional wafer-fab floor impractical.

Their more likely near-term role is as a research and standards platform, supplying high-quality reference data for compact metrology systems rather than immediately replacing them in production.

SSMB: a long-term route to high average power

Tsinghua University’s steady-state microbunching (SSMB) concept uses laser modulation and an undulator to create microbunches in an electron storage ring. The microbunching is maintained as the electron beam circulates, enabling coherent radiation at high repetition rates. Tsinghua’s proof-of-principle work demonstrated the underlying mechanism, while subsequent designs have proposed 13.5 nm EUV radiation with high average power and narrow bandwidth.124

SSMB is attractive because it seeks to combine the high repetition rate of a storage ring with the brightness and coherence of coherent radiation. Design materials have proposed output above 1 kW per tool, but that is an engineering target—not evidence that a source has already been built, integrated or qualified for semiconductor production.310

The obstacles include electron-beam stability, laser modulation, storage-ring control, beam quality, facility scale and cost. SSMB is better understood as a potentially transformative long-term source architecture than as an immediately deployable tabletop metrology source.

Why HHG is a leading candidate for next-generation metrology

High-harmonic generation (HHG) focuses an ultrafast laser into a gas or another medium to produce harmonics extending into the EUV and soft-X-ray ranges. Compared with tin-plasma sources, HHG offers several features that are particularly useful for metrology:

• Tabletop potential: It can operate in a laboratory-scale system rather than requiring a large synchrotron facility.
• High coherence: Spatial and temporal coherence support CDI, scatterometry, ptychography and mask review.
• Wavelength flexibility: Different wavelengths can be selected for different materials, structures and measurement models.
• Lower debris risk: HHG does not rely on tin-droplet plasma, reducing one source of contamination-control pressure.
• Ultrafast operation: Its short pulses are useful for time-resolved measurements and materials research.3544

HHG is particularly well suited to applications that require high coherence but not lithography-grade power. NIST has used tabletop HHG for EUV critical-dimension diffractometry, and research literature continues to identify HHG as an important source for EUV metrology and imaging.3335

Its limitations are equally clear. Conversion efficiency and delivered EUV power remain low, while beam stability, long-duration operation, repetition rate, optical losses and large-area scanning throughput all require improvement. An engineering analysis of full-size EUV mask inspection estimated that some applications could require roughly 1–100 mW of source power. The actual requirement varies with defect sensitivity, scan area, signal-to-noise ratio and throughput, so no single power figure should be treated as a universal threshold.51

Public reports say that ASML, Intel, Samsung and TSMC have used or evaluated HHG-related measurement and inspection technologies.1824 This is best interpreted as evidence that industry is testing HHG’s potential—not as proof that HHG has broadly replaced plasma sources or become the standard source for every high-volume inspection tool.

China’s domestic ecosystem: from sources to system integration

Public reporting identifies Chinese companies including Huaxin Aurora, Langdao Technology, Yunqi Jiyao, Huari Laser and Haoyu Xinguang. The companies are described as operating across source development, lasers, equipment and system integration, with Haoyu Xinguang focusing particularly on the HHG route.17

Among universities and research institutions, Tsinghua’s SSMB program is one of the clearest publicly documented accelerator-source initiatives. Its proof-of-principle experiment was completed with Helmholtz-Zentrum Berlin and Germany’s Physikalisch-Technische Bundesanstalt, underscoring the depth of accelerator and metrology expertise still required in this field.214

A durable domestic ecosystem must also fill the gaps beyond the source itself:

• EUV multilayer mirrors, vacuum chambers and contamination control;
• detectors, precision stages, alignment, calibration and radiometry;
• scattering models, image reconstruction, defect classification and process-control software;
• mask blanks, pellicles, photoresists, films and wafer-process test structures;
• close cooperation with fabs to correlate measurements with actual printing, device electrical results and yield.

In other words, localization cannot be measured simply by counting companies developing EUV sources. The more meaningful questions are whether suppliers can deliver complete tools, maintain a stable calibration chain and establish a feedback loop using real production data.

Five hurdles between a laboratory prototype and a production tool

1. Effective optical power and throughput

Power generated by a source is not the same as useful power reaching the sample, passing through the optics and becoming a measurable signal. A tool must balance defect sensitivity, scan area, exposure time and noise while maintaining an economically meaningful inspection rate.

2. Long-term stability

Fabs need equipment that operates consistently across shifts and months, not an experimental setup that performs well for a short demonstration. Wavelength, dose, pointing, coherence and detector response must all be monitored, calibrated and maintained.

3. Source and system lifetime

Plasma components, lasers, vacuum systems, mirrors and detectors all affect the maintenance interval. Debris, optical degradation and replacement costs ultimately determine the tool’s total cost of ownership.

4. Correlation with wafer results

A high-resolution image is not the end goal. The tool must show that measured critical dimensions, overlay errors or defect signatures predict actual wafer printability, device performance and yield. Without that correlation, measurements are unlikely to become part of production control.

5. Process and software integration

EUV inspection must work with masks, photoresists, films, wafer topography, automated handling, data interfaces and factory process-control systems. Long customer qualification cycles, comparative testing and field data often matter more to commercialization than a single laboratory demonstration.3050

How to judge whether China’s EUV metrology is ready

The most important indicators will not be claims that a team has produced 13.5 nm light. They will be verifiable production metrics: continuous-operation data, delivered sample power, scan throughput, calibrated defect sensitivity, correlations with electron-beam or synchrotron measurements, and links to wafer electrical results. Long-term fab trials and customer qualification would be stronger evidence still.

Based on the public information available, a plausible path is staged. HHG may enter research, mask review and specialized metrology first. DPP and other compact plasma sources could target selected in-line applications. Synchrotron facilities and SSMB may continue to support advanced source research and metrology science, while LPP remains important wherever the highest power and throughput are required.

China’s EUV metrology ecosystem is therefore beginning to take shape, but an emerging ecosystem is not the same as mass-production readiness. The decisive challenge is not merely generating 13.5 nm photons. It is integrating the source, optics, detector, algorithms, materials knowledge and fab validation into a reliable tool that can run continuously and deliver data useful for production decisions.