ScienceNot yet confirmed elsewhere1 publisher3 min readPublished
Coating thickness changes what surface roughness means for pitting on steam-coated aluminum
Shibaura Institute of Technology researchers modeled 90 steam-coated aluminum specimens and found roughness's link to thickness flips sign near 2,200 nm. Thickness and roughness need judging together, though the flip appears only in rougher films and its cause is untested.
The Scientist · Science desk
What happened
- The team fed four measured features (film thickness, RMS roughness, crystallite size and a substrate dislocation index) into a random forest, then read it with SHAP and accumulated local effects.
- Pitting resistance was measured as pitting potential, the voltage at which localized corrosion starts, in 5 wt.% sodium chloride solution at room temperature.
- A Monte Carlo check put measurement uncertainty at about 0.080 V of prediction variability, against an overall model error of 0.292 V.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- decision Anyone tuning steam-coating conditions has to set thickness and roughness as a pair, because the same roughness value is modeled as interacting with thickness in opposite directions on either side of about 2,200 nm.
- constraint The 2,200 nm figure comes from a forest fitted to 90 specimens, and its physical explanation is still a hypothesis, so it is not yet something to write into a coating specification.
- constraint With measurement noise only about a quarter of the prediction error, a quality check built on these four measurements would still leave most of the scatter in pitting potential unexplained.
- capability If validation holds, coating lines could judge films by measured structure from diffraction and microscopy, instead of trusting temperature and time settings alone.
Pitting can bring down an aluminum part suddenly even when its overall corrosion rate is low [13]. Steam coating grows a protective boehmite film from water vapor, but changing the process moves film thickness, surface morphology, crystallinity and substrate defects all at once, so it has been hard to say which of them governs corrosion resistance [1].
The Shibaura team, led by Takahiro Ishizaki and master's student Kei Masuhara, worked with that coupling instead of around it [2]. They measured all four features on 90 A6061-T6 specimens, fitted a random forest, and then used Shapley additive explanations and accumulated local effects to read out what the forest had learned [3][4]. The fitted model predicted pitting resistance more accurately than one that knew only coating temperature and treatment time [10].
Roughness and thickness came out as the two leading descriptors [6]. In the rougher specimens, with RMS surface heights of about 600 to 1,080 nm, the modeled interaction between them turned from positive to negative near 2,200 nm of film [7]. The authors are careful about scope. The flip describes how the two variables combine in the model, not the pitting resistance of every coating above or below that thickness, and the smoother specimens showed no reversal [8].
"The key insight is that higher roughness should not automatically be considered beneficial or harmful," Ishizaki said. "Its meaning depends on the film-growth regime." [16]
Read strictly, the study ranks thickness alongside roughness, not ahead of it, so the case it makes is for setting the pair together [6]. Crystallite size behaves the same way. Its interaction with roughness also changes by region, and the authors warn against a simple "larger is better" rule there [12].
The model's root-mean-square error is 0.292 V, and a Monte Carlo analysis put the contribution from descriptor measurement uncertainty at about 0.080 V [9]. Measurement noise is therefore roughly 27% of the total error [14], and the authors conclude it cannot account for the rest [9]. The published summary does not report the spread of pitting potentials across the 90 specimens, so it is hard to judge how large 0.292 V is against the effect being modeled [3][9].
The thing this doesn't tell you is why. SHAP and accumulated local effects describe how the fitted forest responds to its inputs; the coating's own physics is a separate question [4]. The team proposes that in thin films, roughness may track protective coverage that is still forming. In thick films it may instead mark structural irregularities that give corrosive species a path in. The authors label both as hypotheses that need experimental testing [15].
The descriptors came from X-ray diffraction, confocal laser scanning microscopy and cross-sectional electron microscopy, and the authors see such measurements supporting data-driven coating assessment and quality control once validated [11]. The paper appeared online in npj Materials Degradation on Aug. 18, 2026 [2].
What to watch
- A designed experiment that grows films at matched roughness on either side of 2,200 nm and checks whether pitting potential follows the model's sign change.
- Cross-section imaging of thick, rough films that finds or fails to find the irregularities the authors propose as corrosion pathways.
- Tests of the model on specimens from another batch or another alloy, outside the 90 A6061-T6 samples it was fitted on.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+20
- Incentives
- Insufficient
- Confidence40
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Steam coating offers a water-vapor-based route to protective boehmite films, but processing simultaneously changes film thickness, surface morphology, crystallinity and substrate defects; these intertwined changes make it difficult to determine which physical features govern corrosion resistance.
- [2]
The research team was led by professor Takahiro Ishizaki and master's student Kei Masuhara of Shibaura Institute of Technology; the findings were published online in npj Materials Degradation on Aug. 18, 2026.
- [3]
The team used an interpretable machine-learning framework to examine 90 steam-coated A6061-T6 aluminum specimens.
- [4]
Four descriptors (film thickness, surface morphology quantified as root-mean-square surface height SQ, crystallite size, and a substrate dislocation density index) were analyzed using random forest, Shapley additive explanations and accumulated local effects.
- [5]
Pitting resistance was assessed by measuring the pitting potential, the voltage at which localized corrosion begins, in a 5 wt.% sodium chloride solution at room temperature; higher values indicate greater resistance.
- [6]
The analysis identified surface roughness and film thickness as the leading descriptors in the model.
- [7]
For specimens with SQ values of approximately 600 to 1,080 nm, the modeled interaction between roughness and thickness changed from positive to negative near a film thickness of 2,200 nm.
- [8]
The reversal concerns the interaction between the two descriptors, not the overall corrosion resistance of every coating above or below that thickness; the same reversal was not observed in the lower-roughness range.
- [9]
A Monte Carlo analysis estimated that uncertainty in descriptor measurements contributed approximately 0.080 V to prediction variability, compared with an overall model RMSE of 0.292 V, suggesting measurement uncertainty alone cannot explain the remaining prediction error.
- [10]
The machine-learning model predicted corrosion resistance more accurately than a model based only on coating temperature and treatment time.
- [11]
The descriptors were obtained using X-ray diffraction, confocal laser scanning microscopy and cross-sectional electron microscopy; with further validation, such measurements could support data-driven coating assessment and quality control.
- [12]
The analysis also identified a region-dependent interaction between surface roughness and crystallite size, suggesting their combination matters more than a simple 'larger is better' rule.
- [13]
Localized pitting corrosion can cause sudden failure in aluminum components even when overall corrosion rates are low.
- [14]
Descriptor measurement uncertainty (about 0.080 V) is roughly 27% of the model's overall RMSE (0.292 V).
- [15]
The researchers interpret the pattern as a possible change in what roughness represents: in thinner coatings it may reflect development of protective coverage, in thicker coatings it may be associated with structural irregularities that provide pathways for corrosive species. These interpretations remain hypotheses requiring further experimental testing.
- [16]
"The key insight is that higher roughness should not automatically be considered beneficial or harmful," says Ishizaki. "Its meaning depends on the film-growth regime."
Sources
1 independent publisher whose own reporting we read for this story.
- phys.orgAI reveals how coating thickness changes the link between surface roughness and pitting resistance
1 article · October 9, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.