Monte Carlo bounds on Hg-rich amalgam gilding in high-resolution neutron tomography of medieval filigree

High-resolution neutron computed tomography (nCT) enables non-invasive investigation of dense metallic cultural-heritage objects, but quantitative interpretation of reconstructed attenuation can be limited by partial-volume effects and region-of-interest (ROI) averaging when thin coatings are compositionally heterogeneous. Here we combine scanning electron microscopy paired with energy dispersive spectroscopy (SEM–EDS) microanalysis, nCT-based segmentation, and a validated Geant4 model of the Dingo thermal neutron imaging beamline (ANSTO) to simulate neutron transmission through a filigree fragment from the 13th-century Chiaravalle Cross (Milan). Surface meshes of the twisted wire, solder joint, and residual gilding were exported from the reconstructed nCT volume and simulated under the Dingo spectrum. We report spectrum-averaged effective attenuation coefficients, Σ eff ( Σ MC from simulation, Σ nCT from reconstruction), extracted using ROI-specific path-length characterisation. For the internal consistency check of the wire and solder joint, simulated values agree with the nCT-derived ranges (wire: Σ MC=2.4~ cm⁻¹ , Σ nCT=2.3 - 2.5~ cm⁻¹ ; solder joint: Σ MC=1.9~ cm⁻¹ , Σ nCT=1.9 - 2.1~ cm⁻¹ ); a two-factor decomposition attributes the wire–solder joint contrast almost entirely to geometry rather than composition. For the residual gilding, SEM–EDS indicates strong local variability, including Hg-rich micro-compositions typical of amalgam gilding. A Monte Carlo sensitivity study across candidate gilding compositions predicts Σ MC spanning approximately 2.6-5.5~ cm⁻¹ over the CT-derived ROI path-length distribution, whereas the nCT-derived gilding values cluster in a narrower interval, consistent with voxel-scale averaging in thin, discontinuous coatings. A patch-mixture forward model converts this into a quantitative bound on the Hg-rich area fraction within the analysed segmented surface-adjacent gilding ROI. Against the empirical upper edge of the nCT interval the model crosses at $$13.8%$$ ( $$95%$$ Monte Carlo endpoint-resampling interval $$11.7-15.8%$$ ); expanding the comparison level by one ROI standard deviation as a conservative sensitivity test gives a model-based bound of $$24.3%$$ for the selected gilding (3)/gilding (4) endmember scenario. Hg-rich residues occupying up to roughly a quarter of the ROI would therefore remain compatible with the observed nCT interval. This is not a formal confidence limit on the Hg-rich fraction: substantial local enrichment cannot be excluded, the amalgam evidence derives from the related SEM cross-section rather than the imaged fragment, and the bound is conditional on the endmember pair. The workflow integrates local SEM–EDS compositional priors with CT-derived geometry and spectrum-resolved neutron transport to test which sub-resolution Hg-bearing scenarios are compatible with ROI-averaged nCT attenuation.