Access Official Technical PDF

Share Summary:LinkedInEmail

As semiconductor structures become more three-dimensional, dimensional control increasingly depends on measurements that capture more than top-surface height. 3D-AFM metrology extends atomic force microscopy (AFM) to features such as steep or reentrant sidewalls, providing profile information relevant to linewidth, height, pitch and sidewall angle. Reliable results, however, depend on more than the instrument’s ability to reach a surface: tip geometry, scanner calibration and measurement uncertainty all matter.

What makes an AFM measurement three-dimensional?

Conventional AFMs can produce images with x, y and height data, but their tip-to-sample position control is generally applied along the vertical axis. A tapered probe may also be unable to reach a near-vertical wall because the tip itself is geometrically blocked. The resulting image can therefore omit important parts of a patterned feature or include artifacts.

In this context, “3D-AFM” describes a class of advanced AFM methods designed to acquire sidewall information. It does not mean that current instruments have equivalent force sensing and position control in all three axes. Different implementations combine specialized probes, alternative scanning and data-acquisition strategies, or additional sensing and control to access steep surfaces.

How 3D-AFM measures semiconductor sidewalls

Critical dimension AFM (CD-AFM) is a widely used implementation. It combines a flared probe, vertical cantilever oscillation, lateral tip dithering and bidirectional feedback. When the system detects a change in surface slope, it can adjust the servo direction to follow the sidewall. Adaptive data spacing allows measurements to be concentrated according to the surface profile, rather than relying only on fixed lateral intervals.

Other approaches use different strategies. Step-in scanning retracts and advances the probe rather than maintaining continuous contact, helping mitigate edge artifacts associated with tip bending and vertical feedback. Controlled tilting of the AFM head can provide another route to sidewall access, although results depend on factors including axis decoupling and the accuracy of stitching data collected at different tilts. These methods have distinct operating principles; none removes the need to understand the instrument’s limitations.

For semiconductor process control, the value of sidewall information is clear in structures such as fin-based field-effect transistors. Relevant dimensions can include fin height, widths at different heights, pitch and sidewall angle. A profile that captures multiple dimensions can help characterize a feature more completely than a single top-down width measurement. Whether a particular feature can be measured still depends on probe width, working length and physical access.

Calibration must include the probe and the scanner

Scanner calibration establishes how commanded movement relates to measured displacement. Vertical scales are commonly calibrated with height standards and lateral scales with pitch standards. Calibration should also consider linearity across the measurement range and local deviations: a scanner’s response may vary depending on where a feature lies within that range. Using standards with dimensions close to the features of interest can help account for these local effects.

For sidewall and linewidth measurements, scale calibration alone is not enough. The finite probe size changes the apparent profile: the recorded path reflects the interaction between the tip and sample, rather than the sample shape alone. Tip width and, where needed, tip shape must therefore be characterized so that probe effects can be accounted for in the measurement.

One route to calibrating probe width uses reference features whose dimensions are characterized by transmission electron microscopy (TEM). In lattice-based approaches, crystalline spacing provides a dimensional reference. Because edge locations in TEM images can be difficult to resolve, the characterization requires careful evaluation and an uncertainty assessment. Reference samples also need periodic checks for damage or change.

Angle verification is a separate but related task. A misalignment between the instrument head and sample axes can affect reported sidewall angles. Image reversal—measuring the same feature in different scan directions or sample orientations—can reveal such misalignment. Comparisons with cross-section TEM or scanning electron microscopy can support verification over the angle range used in routine measurements.

Make uncertainty specific to the measurement

Measurement uncertainty is not a single, permanent rating for an AFM. It applies to a particular measurand and set of measurement conditions. For linewidth, for example, relevant contributions may include repeatability, sample uniformity, tip size, tip bending, scale factor and scanner nonlinearity. The importance of each source depends on the measurement model.

Two complementary approaches are described for evaluating performance. A linearity comparison against a reference measurement system can help assess residual error and compare instruments, but does not by itself identify every underlying error source. A component-based uncertainty budget instead estimates individual contributions, including both statistically evaluated effects and those assessed through models or other evidence. It can make dominant contributors easier to identify, while requiring the analyst to consider whether error sources are correlated.

The practical objective is to match the rigor of the analysis to the application, whether that is routine process monitoring, tool matching or reference measurement. Stating the uncertainty alongside the result helps decision-makers judge whether a measurement is fit for its intended tolerance.

Roles in reference and contour metrology

A calibrated and characterized 3D-AFM can serve as a reference measurement system (RMS) for faster inline tools. In this arrangement, the AFM measures reference wafers or artifacts, which are then used to evaluate production instruments. Where the calibration chain is traceable to the SI metre, this approach can transfer dimensional consistency to measurements made by those tools. It is especially relevant when inline results may be influenced by material properties as well as feature geometry.

3D-AFM has also been proposed for contour metrology, which checks whether lithographically printed features match intended layouts. Accurate width information can support contour extraction, while a full profile may provide additional dimensional context. Extracting planar contours from CD-AFM data can require measurements with orthogonal scan directions because sidewall information is acquired along a designated lateral axis.

Conclusion

3D-AFM gives semiconductor metrology teams a way to access sidewalls and profile details that conventional AFM may not capture. Its practical value depends on a complete measurement chain: calibrated scanner scales, characterized probe geometry, verified alignment and an uncertainty evaluation suited to the measurand. For engineers and process leaders, these steps are essential to interpreting profiles consistently and using AFM data effectively in process control, tool evaluation and reference metrology.

Access Official Technical PDF

About The Author