Semiconductor metrology is becoming more challenging as integrated circuit features shrink and device structures extend into three dimensions. Engineers must characterize not only critical dimensions, but also roughness, layer thickness, interfaces, defects and material properties. The choice of measurement method matters: each instrument observes a different aspect of a structure, over a different area, and with its own sources of uncertainty.
For process engineers, system integrators and manufacturing leaders, the central issue is not simply whether a tool can resolve a small feature. It is whether its measurement model, sampling strategy and error controls provide information that is useful for process development and control.
Why semiconductor metrology is changing
Scaling has historically reduced device dimensions, but continued progress also depends on new architectures. FinFETs give way to gate-all-around designs, while vertical structures and stacked integration add depth, interfaces and additional process steps. Emerging devices may also use materials such as carbon nanotubes or transition-metal chalcogenides, creating measurement needs that differ from those of conventional silicon structures.
As geometry becomes more complex, familiar measurements can change meaning. In a planar transistor, gate width and edge roughness are important. In a FinFET, fin dimensions and their variation also affect performance. In a gate-all-around device, the gate may surround a wire or sheet, while vertical designs can make gate thickness a film measurement rather than a simple linewidth measurement.
Smaller dimensions also make local variation more consequential. Stochastic patterning effects can affect feature fidelity, while missing features and random open or short defects require careful inspection. In dense three-dimensional stacks, a measurement may need to capture buried layers or deep structures without relying on destructive cross-sectioning.
Matching metrology methods to the measurand
No single instrument provides every kind of information needed for advanced integrated circuits. Measurement planning starts by defining the measurand—the specific quantity of interest—and assessing how the instrument interacts with the sample. Relevant considerations include feature geometry, materials, accessibility, sampling area, signal strength and possible measurement-induced damage.
CD-SEM: flexible imaging for process monitoring
Critical-dimension scanning electron microscopy (CD-SEM) is widely used for in-line measurements. A focused electron beam scans the sample, and detected signals are used to assess feature size and shape. Top-down images can support linewidth, contact-hole and edge-roughness measurements. With tilted imaging and model-based analysis, SEM can also provide information about three-dimensional profiles.
CD-SEM combines local imaging with measurements suited to manufacturing environments. Its practical performance, however, is affected by more than beam focus. Drift, vibration, contamination, charging and beam damage can influence results. Reaching smaller dimensions therefore depends on controlling these effects and using models that account for signal generation, sample properties and measurement conditions.
CD-SAXS: ensemble measurements through complex stacks
Critical-dimension small-angle X-ray scattering (CD-SAXS) measures X-rays scattered from periodic structures. Rather than imaging individual features, it fits measured scattering data to shape models to estimate average structural properties, including linewidth, pitch and sidewall angle. Because X-rays can penetrate silicon wafers and optically opaque layers, the method can characterize structures without cross-sectioning them.
Its ensemble-averaging approach can provide information across many repeating devices, while its usefulness depends on the amount of scattering signal. Weakly scattering materials and structures can reduce signal-to-noise ratio and extend measurement times. The brightness of compact X-ray sources is also a constraint on practical throughput.
Scatterometry: fast, model-based measurements
Scatterometry measures changes in light reflected or diffracted from patterned arrays. It can support critical-dimension and overlay measurements, as well as estimates of optical properties. Because it is non-imaging, it can gather measurements quickly and infer dimensions below conventional optical imaging limits by fitting experimental data to simulations.
The fitting process is an inverse problem: different combinations of parameters can produce similar optical signals. Correlation between parameters can therefore increase uncertainty, particularly as structures and material stacks become more complex. Careful experiment design and accurate models are essential. For emerging materials, assumptions about optical properties may also need to be revised.
TEM and AFM: detailed structural and surface information
Transmission electron microscopy (TEM) can provide high-resolution images of cross-sections and interfaces. It is valuable for examining atomic-scale structure, film thickness, strain and material boundaries. However, most samples must be thinned and cross-sectioned, making TEM destructive and limiting how many devices can be assessed. Beam-sensitive, low-contrast materials present additional difficulties.
Atomic force microscopy (AFM) uses a nanoscale tip to sense surface topography and other forces. It can measure vertical topography with sub-nanometre resolution and, depending on the operating mode, probe properties such as electrical fields, work-function differences or magnetic fields. Tip geometry and tip–sample interactions influence interpretation; dense features and confined spaces can be difficult to measure reliably.
Why hybrid metrology matters
Each method has a distinct balance of resolution, coverage, speed and sample impact. Imaging tools can reveal local details, while scattering methods can characterize averages over larger arrays. A model-based optical result may benefit from independent constraints, and high-resolution TEM can help assess or calibrate measurements from other methods—provided its own sample-preparation and measurement uncertainties are considered.
Combining techniques can reduce uncertainty when their measurements contribute complementary information. That requires more than comparing reported dimensions: teams need to understand each instrument’s measurement physics, error sources, sampling approach and model assumptions. Traceability to the SI system of units is also important for establishing confidence in measurement results and calibration.
Practical implications for process and equipment teams
For advanced device development and production, measurement strategy should evolve alongside device architecture and patterning. Teams should define the feature or property that must be controlled, then select methods suited to its geometry, material and location. They should also evaluate whether a result represents an individual feature or an average across an array, and whether measurement conditions can alter a sensitive sample.
As devices become more three-dimensional and material stacks more diverse, metrology will increasingly depend on informed combinations of instruments, models and calibration practices. The practical goal is not to find one universal tool, but to build a measurement approach that delivers relevant, repeatable information for process decisions—while making limitations and uncertainty visible.