Unsupervised 3D segmentation and statistical corrosion-layer stratigraphy of long-term corrosion in carbon steel archaeological analogs using co-registered bimodal tomograms

Abstract Accurate prediction of long-term carbon-steel corrosion remains a major challenge in safety assessments for deep geological disposal of high-level radioactive waste. Because laboratory and in situ experiments span comparatively short periods, archeological iron offers access to corrosion states developed over centuries to millennia under natural burial conditions. Here, we present a quantitative three-dimensional analysis of corrosion stratigraphy in 10 Roman iron nails, buried for approximately 1700–2000 years, using co-registered neutron and X-ray computed tomography combined with unsupervised clustering in joint attenuation space. Voxel-wise neutron-X-ray attenuation pairs were segmented by K-means clustering at ( k = 4), (6), and (8) to test the reproducibility and useful granularity of attenuation-defined corrosion domains across specimens. At ( k = 4), all specimens exhibited a consistent first-order architecture comprising residual metal, dense product layer, and transformed medium, with preserved spatial ordering around the metallic core. An operational inner-envelope proxy indicated apparent metal loss of 43–65%, with a cohort mean of approximately 53%, corresponding to millimeter-scale penetration and low long-term average corrosion rates. At ( k = 6), reproducible sub-domains emerged within the dense product layer and transformed medium in both attenuation and tomogram space. Targeted Raman measurements showed recurring mixed mineral assemblages rather than phase-pure layers. At ( k = 8), finer subdivision remained possible, but cross-specimen correspondence decreased, particularly within the dense product layer, indicating limited robustness of the finest partition. This framework enables non-destructive, three-dimensional comparison of corrosion stratigraphy and establishes archeological iron as a constrained analog for long-term corrosion behavior in repository-relevant systems across natural burial timescales that are inaccessible to conventional experiments.

Authors

Publication Details

Journal
npj Materials Degradation
Published
2026-09-14
DOI
https://doi.org/10.1038/s41529-026-00882-w
Primary Topic
Corrosion Behavior and Inhibition
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
article

Unsupervised 3D segmentation and statistical corrosion-layer stratigraphy of long-term corrosion in carbon steel archaeological analogs using co-registered bimodal tomograms

Elodie Granget, Laura Brambilla, Markus Ströbl, Nikitas Diomidis et al.
npj Materials Degradation
Corrosion Behavior and Inhibition
article

Unsupervised 3D segmentation and statistical corrosion-layer stratigraphy of long-term corrosion in carbon steel archaeological analogs using co-registered bimodal tomograms

Elodie Granget, Laura Brambilla, Markus Ströbl, Nikitas Diomidis, Laura Cristina, Seren Azad, David Christian Mannes, Fazel Mirzaei, Lara El Arab, Anders Kaestner
article en

Abstract

Abstract Accurate prediction of long-term carbon-steel corrosion remains a major challenge in safety assessments for deep geological disposal of high-level radioactive waste. Because laboratory and in situ experiments span comparatively short periods, archeological iron offers access to corrosion states developed over centuries to millennia under natural burial conditions. Here, we present a quantitative three-dimensional analysis of corrosion stratigraphy in 10 Roman iron nails, buried for approximately 1700–2000 years, using co-registered neutron and X-ray computed tomography combined with unsupervised clustering in joint attenuation space. Voxel-wise neutron-X-ray attenuation pairs were segmented by K-means clustering at ( k = 4), (6), and (8) to test the reproducibility and useful granularity of attenuation-defined corrosion domains across specimens. At ( k = 4), all specimens exhibited a consistent first-order architecture comprising residual metal, dense product layer, and transformed medium, with preserved spatial ordering around the metallic core. An operational inner-envelope proxy indicated apparent metal loss of 43–65%, with a cohort mean of approximately 53%, corresponding to millimeter-scale penetration and low long-term average corrosion rates. At ( k = 6), reproducible sub-domains emerged within the dense product layer and transformed medium in both attenuation and tomogram space. Targeted Raman measurements showed recurring mixed mineral assemblages rather than phase-pure layers. At ( k = 8), finer subdivision remained possible, but cross-specimen correspondence decreased, particularly within the dense product layer, indicating limited robustness of the finest partition. This framework enables non-destructive, three-dimensional comparison of corrosion stratigraphy and establishes archeological iron as a constrained analog for long-term corrosion behavior in repository-relevant systems across natural burial timescales that are inaccessible to conventional experiments.

npj Materials Degradation
Openalex Percentile: Top 24%
Corrosion Behavior and Inhibition
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.