Modalities
Atomic Force Microscopy
Atomic force microscopy (AFM) measures surfaces through interactions between a sharp probe and a specimen. It can resolve small height differences and, with suitable measurement modes, investigate local mechanical and electrical behaviour.
AFM connects surface structure with material behaviour. Thin-film steps, particles and patterned surfaces offer insight into how a material was made and how it might perform.
Measurement and Interpretation
An AFM image reflects both the specimen and the instrument. Tip shape can broaden features, while feedback and calibration affect the recorded height.
A useful measurement needs a defensible reference. Step height depends on the chosen comparison regions and background correction; roughness also depends on scan area, sampling and processing.
Different signals support different physical claims. Phase contrast alone cannot uniquely identify a material, while estimating elastic modulus from force measurements requires calibrated probe parameters and an appropriate contact model.
AI Agent Potential
AI agents could make AFM analysis more consistent and less repetitive. They can decode native files, compare scans and carry a reproducible analysis across many specimens.
Reliability depends on preserving meaning through those calculations. A misread unit changes the length scale, an overlooked structure leaves the account incomplete, and a missing measurement must remain distinct from zero.
Scientific judgement means recognising what the evidence supports. An agent should distinguish specimen features from probe or processing effects, and identify when calibration or further measurements are needed.
Step-height Example
A useful analysis must decide what can be reported. Here, an agent would estimate the local height of a fictional coating edge and judge whether two scan directions support a single value.
Compare the background corrections, then introduce a scan disagreement. The shaded bands mark exposed substrate (L) and coating (U); the dashed outline shows the fitted region. The surface is generated solely for this demonstration.
SYNTHETIC EXAMPLE
Static view: uncorrected heights from a stable scan pair. Enable JavaScript to compare the other choices.
Raw forward scan · 4 × 3 µm. L and U are comparison bands. No plane is subtracted.
ForwardReverse
Mean along y; both directions expressed in the same x coordinate.
3 Decide what to report
The raw difference includes the sloping background. An agent would fit an appropriate reference region before interpreting it as a step height.
Agreement between scan directions does not remove this shared bias.
How this illustration works
The generated surface has a 16 nm step on a shared planar background, with small ripples. The illustration assumes correct height calibration; it does not model tip geometry or feedback.
Each estimate subtracts the mean height in L from that in U after the selected plane correction. Both bands exclude the edge. The two scans are processed separately, with fixed plot scales.
The substrate fit works here because its slope continues beneath the coating. Curvature or non-parallel terraces would need a different reference treatment.
The direction-dependent pair introduces a smaller step in the reverse scan. This discrepancy illustrates a consistency check, rather than a particular instrument fault; the ripples do not represent measurement uncertainty.
A local height estimate supports a local claim. It does not establish a coating’s composition, mechanical properties or uniformity across a component.
This is purely an example task and is not part of GDPx.