Early Stages of Metal Corrosion in Coastal Archaeological Sites: Effects of Chemical Composition in Silver and Copper Alloys.

In this study, metal disks with different chemical composition (two Ag-based alloys and three Cu-based alloys) were buried in the soil of coastal archaeological sites for a period of 15 years. The aim was to naturally induce the growth of corrosion patinas to obtain a deeper insight into the role of alloying elements in the formation of the patinas and into the degradation mechanisms occurring in the very early stages of burial. To reach the aim, the morphological, compositional and structural features of the patinas grown over 15 years were extensively characterized by optical microscopy, field emission scanning electron microscopy coupled with energy dispersive spectrometry, X-ray diffraction and micro-Raman spectroscopy. Results showed that the Cu amount in Ag-based alloys strongly affected the final appearance, as well as the composition and structure of the patinas. Corrosion mechanisms typical of archaeological finds, such as the selective dissolution of Cu, Pb and Zn and internal oxidation of Sn, occurred in the Cu-based alloys, even if areas enriched in Zn and Pb compounds were also detected and attributed to an early stage of degradation. In addition, some unusual and rare compounds were detected in the patinas developed on the Cu-based disks.

Reproducing bronze archaeological patinas through intentional burial: A comparison between short- and long-term interactions with soil.

The reproduction of archaeological corrosion patinas is a key issue for the reliable validation of conservation materials before their use on cultural objects. In this study, bronze disks were intentionally buried for 15 years in the soil of the archaeological site of Tharros, both in laboratory and in situ, with the aim of reproducing corrosion patinas typical of archaeological artifacts to be used as representative surfaces for testing novel cleaning gels. The microstructural, microchemical and mineralogical features of the patinas were analyzed by a multianalytical approach, based on optical microscopy (OM), field emission scanning electron microscopy coupled with energy dispersive spectrometry (FE-SEM-EDS) and X-ray diffraction (XRD). The patinas developed in 15 years were compared with an archaeological bronze recovered from the same site after about two thousand years of burial (referred as short-term and long-term interaction, respectively). Results revealed a similar corrosion behavior, especially in terms of chemical composition and corrosion mechanisms. XRD detected the ubiquitous presence of cuprite, copper hydroxychlorides and terrigenous minerals, while OM and FE-SEM-EDS analyses of cross-sections evidenced similar patinas' stratigraphy, identifying decuprification as driving corrosion mechanism. However, some differences related to the type of local environment and to the time spent in soil were evidenced. In particular, patinas developed in situ are more heterogeneous and rougher, while the archaeological one is thicker and presents a major amount of cuprite, terrigenous deposits and uncommon corrosion compounds. Based on our findings, the disks buried in situ were selected and used as disposable substrates to study the cleaning effect of a novel polyvinyl alcohol (PVA)-based gel loaded with a chelating agent (Na2EDTA · 2H2O). Results show that the gel is effective in removing disfiguring degradation compounds and preserving the stable and protective patina. Based on the conservation needs, the time of application can be properly tuned. It is worth noticing that after a few minutes the green corrosion products can be selectively removed. The EDS analysis performed on the gels after cleaning reveals that they are highly selective for the removal of copper(II) compounds rather than Cu(I) oxide or Cu(0) from bronze substrates.

Chemical and structural variability in cubic spinel oxides.

The empirical relations between cubic spinel oxides of different compositions were investigated using data from 349 refined crystal structures. The results show that the spinel structure is able to tolerate many constituents (at least 36) by enlarging and decreasing the tetrahedra and octahedra. This is reflected in a large variation in tetrahedral and octahedral bond distances. The oxygen positional parameter (u) may be regarded as a measure of the distortion of the spinel structure from cubic close packing or of the angular distortion of the octahedron. The distortion can best be explained in terms of ionic potential (IP), which merges the size and charge properties of an ion. Sterically induced distortion depends on ion size, whereas electrostatically induced distortion is caused by cation-cation repulsion across faces of tetrahedra and shared edges of octahedra. The strong correlations between the u parameter and the IP at the T and M sites are consistent with the main role played by the both charge and size. Large distortions (u ≫ 0.27) result in oxygen-oxygen distances of the octahedron shorter than 2.50 Å, which would lead to structural instability because of increased non-bonded repulsion forces between the oxygen atoms.

Bond valence at mixed occupancy sites. I. Regular polyhedra.

Bond valence sum calculations at mixed occupancy sites show the occurrence of systematic errors leading to apparent violations of the Valence Sum Rule (bond valence theory) in regular and unstrained bonding environments. The systematic deviation of the bond valence from the expected value is observed in the long-range structure, and is discussed from geometric and algebraic viewpoints. In the valence-length diagram, such a deviation arises from discrepancies between the intersection points of the long-range bond valences and the theoretical bond valences with the valence-length curves of involved cations. Three factors cause systematic errors in the bond valences: difference in atomic valences, bond valence parameters Ri (the length of a bond of unit valence) and bond valence parameters bi (the bond softness) between the involved cations over the same crystallographic site. One important consequence strictly related to the systematic errors is that they lead to erroneous bond strain values for mixed occupancy sites indicating underbonding or overbonding that actually does not exist.

Mean bond-length variation in crystal structures: a bond-valence approach.

The distortion theorem of the bond-valence theory predicts that the mean bond length 〈D〉 increases with increasing deviation of the individual bond lengths from their mean value according to the equation 〈D〉 = (D' + ΔD), where D' is the length found in a polyhedron having equivalent bonds and ΔD is the bond distortion. For a given atom, D' is expected to be similar from one structure to another, whereas 〈D〉 should vary as a function of ΔD. However, in several crystal structures 〈D〉 significantly varies without any relevant contribution from ΔD. In accordance with bond-valence theory, 〈D〉 variation is described here by a new equation: 〈D〉 = (DRU + ΔDtop + ΔDiso + ΔDaniso + ΔDelec), where DRU is a constant related to the type of cation and coordination environment, ΔDtop is the topological distortion related to the way the atoms are linked, ΔDiso is an isotropic effect of compression (or stretching) in the bonds produced by steric strain and represents the same increase (or decrease) in all the bond lengths in the coordination sphere, ΔDaniso is the distortion produced by compression and stretching of bonds in the same coordination sphere, ΔDelec is the distortion produced by electronic effects. If present, ΔDelec can be combined with ΔDaniso because they lead to the same kind of distortions in line with the distortion theorem. Each D-index, in the new equation, corresponds to an algebraic expression containing experimental and theoretical bond valences. On the basis of this study, the ΔD index defined in bond valence theory is a result of both the bond topology and the distortion theorem (ΔD = ΔDtop + ΔDaniso + ΔDelec), and D' is a result of the compression, or stretching, of bonds (D' = DRU + ΔDiso). The deficiencies present in the bond-valence theory in explaining mean bond-length variations can therefore be overcome, and the observed variations of 〈D〉 in crystal structures can be described by a self-consistent model.