Metrology’s Expanding Role in Semiconductor Manufacturing
As semiconductor manufacturing advances towards increasingly complex transistor architectures, smaller device features and three-dimensional integration, metrology is becoming ever more important to process control and manufacturing yield. From critical dimension measurement and lithographic overlay to advanced packaging and buried structures, measurement technologies must evolve to characterise features that are increasingly difficult to access, resolve and verify.
Semiconductor manufacturing has always depended on precise control of dimensions, materials and surface characteristics. However, the transition towards gate-all-around transistors, stacked nanosheets, advanced memory structures and chiplet-based architectures is introducing new measurement challenges. Features must not only be manufactured within increasingly demanding tolerances, but also measured reliably at production speeds, with sufficient accuracy to support process optimisation and yield improvement.
The challenge for metrology suppliers is therefore twofold: developing techniques capable of characterising increasingly complex structures while ensuring that the resulting measurements remain reliable, repeatable and suitable for manufacturing decisions.
Measuring Ever-Smaller Features
Critical dimension (CD) metrology is fundamental to semiconductor process control. The widths, heights, profiles and uniformity of patterned features influence transistor performance, electrical characteristics and device reliability. Even relatively small variations can affect manufacturing yield when repeated across billions of structures on a single chip.
Critical dimension scanning electron microscopy (CD-SEM) is widely used to measure patterned features, particularly where optical techniques cannot provide sufficient spatial resolution. By analysing electron microscope images, manufacturers can determine dimensions such as line widths, contact hole diameters and pattern profiles.
However, obtaining a reliable measurement is not simply a matter of resolving a smaller feature. Image contrast, material composition, electron interaction with the sample and the algorithms used to identify feature boundaries can all influence the reported result. At nanometre scales, defining precisely where a feature begins and ends becomes a measurement science problem in its own right.
Optical metrology provides a complementary approach. Techniques including scatterometry and spectroscopic ellipsometry infer dimensional and material properties from the interaction of light with patterned structures and thin films. These methods can measure multiple characteristics rapidly and are particularly useful for monitoring processes across a wafer.
Rather than directly imaging individual features, many optical techniques use mathematical models to interpret measured signals. Their effectiveness consequently depends on the quality of those models, the assumptions made about the structure and the availability of suitable reference measurements.
Atomic force microscopy (AFM) can provide additional information about surface topography and nanoscale features, although measurement speed, access to particular geometries and probe-related effects can limit its suitability for high-volume production.
The different techniques are therefore complementary rather than interchangeable. Semiconductor manufacturers must select measurement methods according to the feature being characterised, the required uncertainty, the production stage and the time available to obtain a result.
Lithographic Overlay and Alignment
Semiconductor devices are constructed through repeated cycles of lithography, deposition, etching and other processing operations. Each stage introduces new patterns or modifies existing structures, making the alignment of successive layers critical.
Overlay metrology measures the relative positioning of patterns formed during different lithographic steps. Misalignment can affect electrical connectivity, device performance and ultimately the proportion of functional chips produced from a wafer.
Optical overlay systems can measure specially designed targets rapidly, enabling manufacturers to monitor alignment across wafers and identify systematic process variations. The resulting data can be used to adjust lithography equipment and compensate for factors such as wafer distortion, stage positioning and process-induced changes.
As device structures become more complex, however, the relationship between an overlay target and the functional structures it represents becomes increasingly important. Measurements taken from accessible targets do not necessarily reveal every alignment issue within a device, particularly where buried features or complex three-dimensional geometries are involved.
This places greater emphasis on combining different sources of metrology data, understanding their limitations and ensuring that process corrections are based on representative measurements.
The Challenge of Three-Dimensional Structures
One of the most significant developments in semiconductor manufacturing is the move towards transistor architectures that extend beyond conventional planar geometries.
Gate-all-around transistors, which surround the conducting channel with a gate structure, are designed to improve electrostatic control as device dimensions shrink. Stacked nanosheet designs introduce additional geometrical complexity, with device performance dependent on the dimensions, spacing and uniformity of multiple closely packed structures.
These architectures create measurement requirements that cannot always be addressed through top-down imaging alone. Characterising sidewall profiles, buried interfaces and cross-sectional dimensions may require a combination of electron microscopy, specialised imaging techniques and computational analysis.
Cross-sectional transmission electron microscopy can reveal extremely small structures and material interfaces, but sample preparation is often destructive and the area examined is relatively limited. Such techniques provide valuable detailed information, although they are not necessarily suitable for routine measurement of every wafer or device.
This creates a need for measurement strategies that combine high-resolution reference techniques with faster methods suitable for production monitoring. Optical measurements, electron microscopy and other analytical approaches can be used together to establish process relationships, validate measurement models and monitor changes during manufacturing.
The difficulty is ensuring that results from different techniques remain comparable. Differences in measurement principles, sampling locations and data interpretation can introduce discrepancies that must be understood before measurements can be used confidently for process control.
Advanced Packaging Creates New Requirements
Metrology challenges are also expanding beyond the fabrication of individual transistors and integrated circuits. Advanced packaging technologies increasingly combine multiple dies or chiplets into integrated systems, placing new demands on dimensional control, alignment and surface quality.
Hybrid bonding, which can create electrical and mechanical connections between closely aligned surfaces, is one example. Successful bonding depends on factors including surface topography, cleanliness, alignment accuracy and the condition of the interfaces being joined.
Small variations in surface flatness, local defects or relative positioning can affect bonding quality and electrical connectivity. Metrology must therefore support both the characterisation of individual components and the verification of their relative positions before integration.
As manufacturers adopt more complex packaging architectures, the measurement task becomes increasingly multidimensional. Surface topography, local height variation, wafer distortion, alignment and defects may all need to be evaluated as part of the same production process.
This creates opportunities for optical surface measurement, interferometry, high-resolution imaging and automated inspection systems. It also increases the importance of integrating metrology at different stages of production, allowing potential problems to be identified before expensive downstream processing or assembly operations are completed.
From Measurement to Process Control
The commercial value of semiconductor metrology extends beyond the ability to measure increasingly small features. Measurements must also be obtained quickly enough to support manufacturing decisions.
In high-volume semiconductor production, inspection and measurement systems generate large quantities of data across multiple process steps. The challenge is to identify meaningful variations, distinguish normal process behaviour from emerging problems and translate results into appropriate corrective action.
Statistical process control provides an established framework for monitoring variation and identifying changes in manufacturing performance. Increasingly, automated data analysis is also being used to identify patterns across large datasets, correlate measurements from different process stages and help engineers investigate potential causes of defects or yield loss.
Artificial intelligence may extend these capabilities through automated image analysis, anomaly detection and the interpretation of complex measurement data. However, AI-assisted metrology introduces additional questions concerning model validation, repeatability and the reliability of measurement results under changing process conditions.
A system that consistently identifies a feature in an image does not necessarily guarantee that its reported dimension is accurate. Measurement systems must still be evaluated against suitable references, with uncertainty and limitations understood before results are used to control manufacturing processes.
The integration of metrology data with manufacturing equipment and process-control systems is therefore as important as improving the resolution of individual instruments. The objective is to identify variations early, determine their significance and reduce the time between detecting a problem and implementing a corrective response.
Measurement Reliability Remains Fundamental
As semiconductor geometries become smaller and more complex, measurement uncertainty and traceability remain central concerns.
Different measurement techniques may produce different results when examining the same feature, particularly when they rely on different physical principles or definitions of a feature boundary. These differences must be understood rather than automatically interpreted as evidence that one instrument is incorrect.
Calibration, reference materials, validated measurement procedures and appropriate uncertainty evaluations provide the foundation for reliable dimensional data. For advanced semiconductor applications, this can require specialised reference structures and measurement models that account for the interaction between the instrument and the material being examined.
Sampling is another important consideration. High-resolution measurements may provide detailed information about a small number of features, while production decisions often depend on understanding variations across entire wafers, multiple process steps and large manufacturing populations.
Manufacturers must therefore balance measurement resolution, sampling strategy, throughput and cost. A highly detailed measurement is not necessarily the most useful production measurement if it cannot be obtained frequently enough to identify changes in process behaviour.
The challenge is to develop measurement strategies that combine detailed characterisation with representative, repeatable monitoring at production scale.
Metrology’s Expanding Role in Semiconductor Manufacturing
The evolution of semiconductor manufacturing is changing both what must be measured and how measurement data is used. Smaller features remain important, but increasingly complex three-dimensional structures, advanced packaging and tighter integration between manufacturing processes are creating new requirements for dimensional and materials characterisation.
Optical metrology, electron microscopy, atomic force microscopy and computational methods each provide capabilities that address different aspects of the problem. Their effective integration will be important to maintaining reliable process control as device architectures continue to evolve.
At the same time, automated analysis and AI-assisted interpretation offer opportunities to extract more value from the growing volume of metrology data. Their contribution, however, will depend on maintaining confidence in measurement accuracy, repeatability and traceability.
For semiconductor manufacturers, metrology is no longer simply a means of verifying that a component meets its dimensional requirements. It is an essential part of understanding complex manufacturing processes, controlling variation and improving yield.
As semiconductor architectures become more demanding, the ability to measure reliably, interpret results correctly and feed that information back into production will remain a critical factor in the development and manufacture of next-generation devices.
Contributing Author: Peter Lawson
