AFM
Nanoscale topography, quantified.
AFM drags or taps a nanometer-sharp tip across a surface and builds a true height map, roughness with real units, line profiles across features, and, in force mode, the mechanical properties of the surface itself.
What it measures
Unlike electron microscopy’s intensity images, AFM output is quantitative height, z(x, y), which makes surface metrics computable rather than impressionistic:
- Roughness: Ra = (1/n)·Σ|zᵢ − z̄| (mean absolute deviation), Rq/RMS = √[(1/n)·Σ(zᵢ − z̄)²], and Rmax, the worst peak-to-valley excursion. Ra and Rq summarize texture; Rmax flags the single defect that punctures a separator.
- Line profiles and features: step heights, grain and particle sizes with full distributions.
- Force-distance curves: pressing the tip into the surface and retracting gives adhesion force from the pull-off event and elastic modulus from fitting the contact region with Hertz or DMT mechanics.
- Phase imaging: in tapping mode, the phase lag maps stiffness and adhesion contrast, separating material domains that look identical in height.
How to read the output
Level first, read second. Raw AFM data sits on sample tilt and scanner bow; plane-correction and line-by-line leveling must come before any roughness number, and aggressive leveling can also erase real long-wavelength waviness, so check what the flattening removed. Then read the metrics together: two surfaces with the same Ra can differ wildly in Rmax, and it is usually the extreme event that matters for adhesion or puncture. For force curves, the approach-retract hysteresis is the signal: pull-off depth reads adhesion, and the contact-region slope reads stiffness, valid only while the tip stayed clean and the model assumptions (tip shape, indentation depth) hold.
A real use case
An anode line sees intermittent coating adhesion failures, peel strength varies lot to lot with identical slurry. AFM on the incoming copper foil puts numbers on a suspicion: the failing lots’ foil runs Ra well below the good lots’, smoother than the process wants, because the anchor profile that the coating keys into is missing, while Rmax stays comparable. Line profiles confirm the texture difference is broadband, not a few scratches. A minimum-Ra window goes into the foil supplier spec, and the adhesion excursions stop arriving with particular deliveries.
Common mistakes
- Quoting roughness from unleveled (or over-leveled) data, the leveling choice can move Ra more than the sample does.
- Comparing Ra across different scan sizes. Roughness is scale-dependent; a 1 µm scan and a 50 µm scan of the same surface give different numbers by construction.
- Ignoring tip artifacts. A blunted or contaminated tip widens every feature and flattens every modulus; sharp features that all share the same shape are the tip imaging itself.
- Fitting force curves with Hertz when adhesion is significant, that is DMT territory, and the model choice changes the modulus.
- Treating one 10 µm field as the surface. Sample several locations before turning a map into a spec decision.
Leveled maps, honest metrics, fitted force curves
Niobia takes AFM data through the corrections first, plane-correction and line-by-line leveling guidance, then computes the roughness set (Ra, Rq/RMS, Rmax), extracts line profiles, and runs grain and particle analysis with size distributions. Phase images get histogram analysis to separate material domains, and force-distance curves are interpreted for adhesion force and elastic modulus with the contact mechanics stated, Hertz or DMT, named explicitly because the choice moves the answer. The foil-QC comparison above is a multi-sample roughness readout, not a gallery of pretty height maps.
Frequently asked
What is the difference between Ra and Rq, and when does Rmax matter?
Ra is the mean absolute height deviation; Rq is the RMS version, which weights large excursions more. They track texture. Rmax is the single worst peak-to-valley event in the scan, the number that matters when one tall defect can puncture a separator or short a cell.
Why do my roughness numbers differ from the supplier's datasheet?
Usually scan size and leveling. Roughness is scale-dependent and leveling-dependent, so the comparison is only valid at matched scan area, instrument mode, and flattening procedure, worth specifying all three in any incoming-QC spec.
Can AFM measure mechanical properties of battery materials?
Yes, force-distance curves give local adhesion and elastic modulus, useful for binder distribution, coating stiffness, and SEI mechanics. The caveats are tip calibration and model validity; modulus values are comparative gold but absolute silver.