Semiconductor Metrology in 2026: In-Line Measurement, EUV Critical Dimensions and X-Ray CT
Updated: 13 September 2026
Semiconductor manufacturing depends on measurements that most factory visitors never see.
A finished chip may contain structures far smaller than a human hair, repeated across wafers and connected through multiple material layers. Advanced packages can contain several dies, interconnects and internal structures that are difficult or impossible to inspect visually after assembly.
Manufacturing these systems consistently requires measurement at almost every stage.
The problem is that the measurement task becomes harder as the technology improves.
Smaller features push conventional optical methods toward their limits. Three-dimensional structures create parameters that cannot be described by a simple width alone. Advanced packaging moves important interfaces inside the assembly, where line-of-sight inspection is no longer enough.
This is why semiconductor metrology is increasingly becoming part of process control rather than only final inspection.
Why semiconductor metrology is a stronger topic in 2026
A June 2026 NIST-linked review of advanced semiconductor manufacturing discusses in-line metrology for process control alongside semiconductor equipment, manufacturing processes and data analytics.
NIST's CHIPS Metrology Program also identifies measurement as a critical part of future semiconductor manufacturing because devices are becoming smaller, more complex and more layered.
The programme focuses on measurements that are accurate, precise and fit for purpose for microelectronic materials, devices, circuits and systems.
Current measurement challenges include:
- critical dimensions of nanoscale structures;
- three-dimensional device geometry;
- material composition;
- surface and interface properties;
- defect detection;
- advanced packaging;
- thermal behaviour;
- process-induced stress and warpage;
- data integration for process control.
These are not separate from manufacturing yield. A measurement that arrives too late, is too uncertain or cannot distinguish the required structure may not be useful for controlling the process.
What is semiconductor metrology?
Semiconductor metrology is the measurement science used to characterize materials, structures, dimensions, defects and process behaviour during semiconductor design and manufacturing.
Depending on the process step, the measurand could be:
- line width;
- film thickness;
- sidewall angle;
- pitch;
- overlay;
- surface roughness;
- composition;
- stress;
- warpage;
- internal void size;
- defect location;
- thermal properties.
The appropriate method depends on what must be measured and how quickly the result is needed.
Why in-line metrology matters
Semiconductor fabrication consists of repeated process steps. If a process begins drifting, waiting until the completed device fails can be extremely expensive.
In-line metrology places measurement closer to the manufacturing process so the information can support process control.
A useful in-line measurement method should balance several competing needs:
- measurement accuracy;
- repeatability;
- speed;
- small measurement area where necessary;
- minimal damage to the wafer or device;
- sufficient sensitivity to the parameter being controlled;
- integration with manufacturing data systems.
A laboratory method may be very accurate but too slow for production. A fast production method may provide a useful proxy but need regular comparison with a higher-order reference method.
In-line, near-line and off-line metrology
| Measurement location | Typical purpose | Main trade-off |
|---|---|---|
| In-line | Fast process control during normal production flow | Must minimize delay and may use indirect measurement methods |
| Near-line | More detailed measurement close to production | May require additional handling or longer measurement time |
| Off-line / reference | High-detail characterization, validation or root-cause work | Usually slower and may not support every-wafer control |
Strong semiconductor process control often depends on correlation between these levels rather than one instrument replacing all others.
What is critical-dimension metrology?
Critical dimension, or CD, refers to a feature dimension that strongly affects device performance or manufacturing success.
Historically, line width is a common example. Modern structures are more complex, so CD metrology may need to determine multiple geometric parameters instead of one width.
Measurements can involve:
- line width at different heights;
- sidewall profile;
- feature height;
- pitch;
- three-dimensional shape;
- buried structure.
As structures become smaller and more three-dimensional, established optical methods can face increasing sensitivity and parameter-correlation challenges.
Why shorter wavelengths are attractive
Conventional optical critical-dimension methods use visible or ultraviolet wavelengths.
A March 2026 NIST publication explains that shrinking feature sizes, increasing structural complexity and difficult parameter correlations are motivating metrology approaches at vacuum-ultraviolet and extreme-ultraviolet wavelengths.
Shorter wavelengths can interact with nanoscale structures differently and may provide additional sensitivity to features that become difficult to resolve or distinguish using longer-wavelength methods.
But shorter wavelengths also create new technical challenges, including strong absorption and more demanding optical systems.
What is EUV critical-dimension metrology?
NIST's 2026 work describes approaches that use extreme-ultraviolet wavelengths for model-based critical-dimension measurements.
The work includes:
- EUV diffractometry;
- EUV scatterometry;
- vacuum-ultraviolet and EUV spectroscopic ellipsometry;
- model-based interpretation of diffraction or reflection signals.
The goal is not to take a conventional microscope image with shorter-wavelength light.
In many cases, the measurement uses how light scatters or diffracts from a periodic structure. A physical model is then used to estimate the structure that most likely produced the measured signal.
Why EUV metrology is a model-based measurement problem
Consider a nanoscale line pattern.
The measurement system records an optical response. Software compares that response with predictions from electromagnetic models. The model parameters may include dimensions and material properties.
The final dimensional result therefore depends on more than detector performance.
It can depend on:
- the optical model;
- assumed material properties;
- feature-shape parameterization;
- instrument calibration;
- measurement noise;
- parameter correlations;
- reference measurements used for validation.
This is why model validation is part of metrology.
EUV metrology is not the same as EUV lithography
EUV is widely associated with semiconductor lithography, where extreme-ultraviolet radiation is used to create very small device patterns.
EUV metrology uses EUV radiation for measurement.
The technologies can be related through wavelength scale and semiconductor applications, but their purposes are different.
- EUV lithography: creates patterns.
- EUV metrology: measures or characterizes structures.
Why optical critical-dimension metrology faces limits
Optical scatterometry has been valuable because it is fast and non-destructive.
NIST's work on next-generation in-line metrology notes that semiconductor structures such as gate-all-around devices, 3D NAND and 3D DRAM create new dimensional measurement demands.
As feature sizes decrease and structures become more three-dimensional, traditional optical approaches can face fundamental sensitivity limits.
This is one reason NIST is also developing X-ray approaches such as critical-dimension small-angle X-ray scattering, or CD-SAXS.
Where CD-SAXS fits
CD-SAXS uses X-ray scattering to infer the dimensions and shape of nanoscale periodic structures.
NIST describes it as a candidate method for next-generation in-line dimensional metrology.
X-rays provide different interaction mechanisms and can potentially characterize three-dimensional structures that are difficult for conventional optical methods.
The manufacturing challenge is not only achieving accurate measurements. It is also achieving sufficient throughput for process control.
Why semiconductor packaging creates a different metrology problem
Front-end wafer metrology often deals with small surface or near-surface structures.
Advanced packaging creates another challenge: important structures are buried inside an assembled package.
Modern packages may include:
- multiple dies;
- chiplets;
- fine-pitch interconnects;
- through-silicon vias;
- underfill;
- bonded interfaces;
- multi-material stacks.
Traditional visual inspection cannot directly access many of these internal features.
This increases interest in non-destructive three-dimensional measurement methods such as X-ray computed tomography.
What is X-ray CT metrology for semiconductors?
X-ray computed tomography uses many X-ray projections to reconstruct a three-dimensional representation of an object.
In semiconductor and advanced-packaging applications, XCT can potentially help inspect internal features without physically sectioning the package.
Potential targets include:
- voids;
- internal cracks;
- bond defects;
- interconnect geometry;
- buried structural features;
- assembly defects.
But seeing an object in a reconstructed image does not automatically prove measurement accuracy or defect-detection capability.
Why XCT needs reference artifacts
In June 2026, NIST published a major review of fabricated artifacts, also called phantoms, for X-ray computed tomography.
The report explains that XCT is increasingly being used for high-precision industrial metrology, including semiconductor manufacturing, and that confidence in results requires well-characterized references.
A reference artifact can contain known structures or controlled flaws.
The XCT system is then challenged to detect or measure those known features.
This allows engineers to ask:
- What spatial resolution can the system actually achieve?
- What defect sizes can it detect?
- How does material combination affect detectability?
- How repeatable are the results?
- How does one XCT platform compare with another?
What is a semiconductor XCT phantom?
A phantom is a test artifact designed with known or well-characterized features for evaluating an imaging or measurement system.
NIST's 2026 review groups XCT artifacts into areas such as:
- spatial-resolution assessment;
- defect-detection studies.
Examples include line-pair structures, Siemens-star patterns and artifacts containing controlled internal flaws.
Advanced fabrication methods can create very small and controlled features using techniques such as focused ion-beam milling, laser micromachining and deep reactive ion etching.
Why artificial defects are useful
A defect-detection system cannot be validated only by scanning parts where the true defect population is unknown.
Controlled artificial defects provide a form of ground truth.
If an artifact contains a known flaw of known size and location, engineers can evaluate whether the XCT system detects it reliably.
This supports probability-of-detection studies and comparison of measurement configurations.
The difficult part is making an artifact that represents real semiconductor-package geometry and material combinations closely enough to be useful.
Why advanced packaging makes XCT validation harder
Semiconductor packages combine materials with different densities and X-ray attenuation characteristics.
Features can also exist at very different size scales inside the same assembly.
This creates problems such as:
- limited contrast;
- resolution trade-offs;
- reconstruction artifacts;
- occlusion by dense materials;
- complex three-dimensional geometries;
- difficulty creating representative reference samples.
A phantom that works well for a simple material may not adequately represent a heterogeneous advanced package.
Defect detection is not the same as dimensional measurement
XCT can be used for both defect detection and dimensional measurement, but the validation questions differ.
| Application | Main question | Typical evidence needed |
|---|---|---|
| Defect detection | Can the system reliably find a flaw of a given type and size? | Known defects, detection statistics, probability-of-detection analysis |
| Dimensional metrology | How accurately can the system determine a size, location or geometry? | Calibrated artifacts, traceability, uncertainty and geometry-error control |
A system that can see a defect is not automatically capable of measuring its dimensions with the uncertainty required for an engineering decision.
Why semiconductor metrology needs reference methods
Many semiconductor measurements are indirect.
Scatterometry, X-ray methods and tomography use physical signals that must be interpreted using models or reconstruction algorithms.
This makes reference methods important.
A faster in-line method may be compared against:
- electron microscopy;
- atomic-force microscopy;
- cross-sectional methods;
- synchrotron measurements;
- traceable dimensional artifacts;
- other higher-order characterization techniques.
The goal is not necessarily for every production measurement to use the reference method.
The goal is to establish confidence in the faster method used for routine process control.
What does fit-for-purpose mean in semiconductor metrology?
The most accurate instrument is not automatically the best production instrument.
A fit-for-purpose measurement system provides enough information, at sufficient confidence and speed, for the decision that must be made.
For example:
- a research tool may prioritize ultimate accuracy;
- a fab process-control tool may prioritize throughput and stability;
- a failure-analysis tool may prioritize three-dimensional detail;
- a packaging inspection system may prioritize non-destructive defect detection.
Measurement capability should therefore be evaluated against the manufacturing decision, not against one universal definition of "best."
Where data analytics enters semiconductor metrology
Modern fabs create large volumes of measurement and process data.
A single metrology result becomes more valuable when it can be connected with:
- process conditions;
- equipment state;
- wafer history;
- material lot;
- yield outcomes;
- failure analysis.
This is why the 2026 NIST manufacturing review discusses in-line metrology together with process control and data analytics.
But data quantity does not remove measurement uncertainty.
Analytics can find patterns in measurements. It cannot correct an unknown measurement bias unless the measurement system itself is characterized.
AI and semiconductor metrology
AI can help classify defects, reconstruct images, optimize measurement recipes or find correlations in large semiconductor datasets.
The same trust problem remains:
What physical evidence shows that the AI output is correct enough for the intended decision?
AI performance should be evaluated using suitable reference data and realistic manufacturing conditions.
For defect detection, this may require known defect artifacts.
For dimensional measurement, it may require traceable reference structures and uncertainty evaluation.
Why this topic matters to India in September 2026
India's semiconductor manufacturing ecosystem is expanding, making measurement capability increasingly relevant to domestic fabs, packaging operations, equipment suppliers and quality laboratories.
The Government announced that SEMICON India 2026 will be held from 17 to 19 September 2026 in New Delhi under the theme "Silicon to Systems: Building the Ecosystem."
The same August 2026 Government announcement states that twelve projects had been approved under Semicon 1.0.
As manufacturing capacity grows, the requirement is not only for fabrication and packaging equipment.
The ecosystem also needs:
- process-control metrology;
- reference materials and artifacts;
- inspection and defect-detection capability;
- measurement software;
- calibration and traceability infrastructure;
- trained measurement specialists.
A semiconductor ecosystem cannot maintain yield and quality without a measurement ecosystem.
What should an engineering team ask before adopting a semiconductor metrology method?
- What exact physical quantity or defect is being measured?
- What process decision will use the result?
- What resolution is required?
- What measurement uncertainty is acceptable?
- What reference method will validate the result?
- Does the method need to be non-destructive?
- How much measurement time can the production flow tolerate?
- How are model assumptions controlled?
- How is instrument drift monitored?
- How are software and recipe changes validated?
- Can the data be connected reliably with process history?
What semiconductor metrology should not be allowed to hide
- poorly characterized reference artifacts;
- unvalidated reconstruction algorithms;
- model assumptions with strong parameter correlation;
- unknown instrument drift;
- insufficient spatial resolution;
- defects below the demonstrated probability of detection;
- material combinations that change imaging performance;
- software updates that alter measurement results;
- measurement uncertainty that is too large for the process decision.
Three metrology directions to watch
1. Shorter-wavelength critical-dimension measurement
EUV and X-ray approaches are becoming more important as optical methods face increasing challenges with smaller and more complex structures.
2. Reference artifacts for three-dimensional inspection
XCT needs better phantoms and controlled defect artifacts so performance can be quantified rather than assumed.
3. Faster measurement connected directly to process control
Measurement methods will increasingly be judged by whether they can support manufacturing decisions quickly enough while preserving confidence in the result.
The central challenge is confidence, not only resolution
Semiconductor metrology is often described as a race toward smaller measurable features.
Resolution matters, but it is not the only issue.
A useful manufacturing measurement must also have:
- known meaning;
- adequate uncertainty;
- validated models;
- stable instrument performance;
- reference evidence;
- sufficient speed;
- integration with the process decision.
As semiconductor manufacturing moves deeper into nanoscale and three-dimensional structures, measurement confidence becomes as important as raw instrument capability.
Frequently asked questions
What is semiconductor metrology?
Semiconductor metrology is the measurement science used to characterize dimensions, materials, structures, defects and process behaviour during semiconductor manufacturing.
What is in-line metrology in semiconductor manufacturing?
In-line metrology provides measurements close to or within the normal production flow so results can support process control without excessive manufacturing delay.
What is EUV critical-dimension metrology?
EUV critical-dimension metrology uses extreme-ultraviolet wavelengths and model-based optical methods such as diffractometry or scatterometry to characterize nanoscale semiconductor structures.
Why is X-ray CT useful for semiconductor packaging?
X-ray computed tomography can inspect internal three-dimensional structures non-destructively, making it useful for defects and features hidden inside advanced semiconductor packages.
What is an XCT reference artifact or phantom?
It is a test artifact containing known or well-characterized structures or defects used to evaluate X-ray CT spatial resolution, dimensional capability or defect-detection performance.
Why are reference measurements important in semiconductor metrology?
Many semiconductor measurements are indirect and model-based. Reference measurements help validate the faster methods used for production and establish confidence in their accuracy and uncertainty.
Sources checked
- NIST: Innovations in advanced processes and systems for semiconductor manufacturing , published 30 June 2026.
- NIST: Harnessing the extreme ultraviolet for critical dimension metrology , published 17 March 2026.
- NIST: Literature Review of Fabricated Artifacts for X-ray Computed Tomography , published 22 June 2026.
- NIST CHIPS for America: CHIPS Metrology Program .
- Government of India, Ministry of Electronics and Information Technology: SEMICON India 2026, Silicon to Systems: Building the Ecosystem , posted 18 August 2026.
Editorial note: Semiconductor measurement technologies evolve quickly. This article explains public measurement-science and government information available as of 13 September 2026. Specific semiconductor processes require method-specific validation, reference measurements and uncertainty evaluation before metrology results are used for manufacturing acceptance or automatic process control.
Comments
Post a Comment