A Synchrotron‐Based Study of the Mary Rose Iron Cannonballs

Abstract Post‐excavation iron corrosion may be accelerated by the presence of Cl−, leading to conservation methods designed to remove Cl. This study exploits a unique opportunity to assess 35 years of conservation applied to cast‐iron cannon shot excavated from the Mary Rose. A combination of synchrotron X‐ray powder diffraction (SXPD), absorption spectroscopy (XAS), and fluorescence (XRF) mapping have been used to characterise the impact of conservation on the crystalline corrosion products, chlorine distribution, and speciation. The chlorinated phase akaganeite, β‐FeO(OH,Cl), was found on shot washed in corrosion inhibitor Hostacor IT with or without an additional reduction stage. No chlorinated phases were observed on the surface of shot stored in sodium sesquicarbonate (Na2CO3/NaHCO3); however, hibbingite, β‐Fe2(OH)3Cl, was present in metal pores. It is proposed that surface β‐FeO(OH,Cl) formed in the early stages of active conservation owing to oxidation of β‐Fe2(OH)3Cl at near‐neutral pH.

For over 2000 years,i ron has been used to manufacture weapons,t ools,c eremonial items,a nd more.H owever, surviving artefacts are prone to permanent loss or damage through the action of corrosion. During burial, the metal oxidises by an electrochemical process,w here the cathodic reaction and oxidising agent are dependent on environmental conditions. [1] Commonly,t his involves either H 2 evolution or O 2 reduction, [2] with multiple pathways involved in the reaction, resulting in complex corrosion layers. [3][4][5] Under favourable conditions,t he corrosion rate may be sufficiently low to allow exceptional preservation of artefacts.O ne such case is the shipwreck of King Henry VIIIsflagship,the Mary Rose (1511-1545), which sunk off the coast of Portsmouth on July 19, 1545. Buried in sediment 14 mb elow the surface, [6] the ship was held in an environment with dissolved O 2 concentration of 0mgL À1 and redox potential, E h ,b etween À34 and À110 mV. [6] These conditions allowed the ship and ca. 19 000 artefacts to survive until excavation between 1979 and 1982.
Iron objects from sites with good preservation conditions face ag reater threat from corrosion that occurs after excavation. Exposure to air and water can cause oxidation of Fe 0 to stable Fe II ,F e III ,a nd intermediate Fe II/III compounds, [1] leading to rapid deterioration (Figure 1a). Furthermore,i nt he presence of Cl À ,t he reaction rate increases, [7] resulting in preservation challenges for objects buried in chlorine-rich environments,such as seawater. [7][8][9][10] To mitigate this,several desalination treatments have been proposed that aim to remove as much Cl as possible from the artefact. These techniques can be divided into two categories:r eductionbased, [11][12][13] where Cl À is removed by transformation of chorine-containing Fe III crystals,a nd washing methods [10,14] that remove chlorine by diffusion into aqueous solution. [15] Comparison of different conservation methods [10,[16][17][18] has been limited by the variability of both objects and burial environments.Asaresult, it is often not possible to attribute differences in treatment success to technique used, while studying material that accurately reflects an archaeological artefact. To overcome these issues,t his work focuses on the collection of 1248 cast iron cannon shot from the Mary Rose.
Having been produced in bulk, [19] the shot were buried together and their relative uniformity maintained until excavation, when they were treated by an umber of conservation methods and exposed to varying environmental conditions. [20] Immediately after excavation, all of the shot were immersed in ahigh pH solution (either NH 3 ,NaOH, or an equimolar mixture of Na 2 CO 3 /NaHCO 3 )until undergoing active conservation, though some shot have remained in passive storage to the present day.F or this study,t hree differently treated categories of shot have been investigated: 1) SS:p assive storage in 0.15 m (pH 10) sodium sesquicarbonate,N a 2 CO 3 /NaHCO 3 ,s olution 2) HW:s ame as (1) until 2010-12, when rinsed in tap water, washed in 3o r4consecutive baths of 1% v:v corrosion inhibitor Hostacor IT:water, [21] (Scheme 1), pH 4.5-8.5, and dried in 2-stage acetone:water (1:1 and 1:0) series. Now stored in acontrolled environment (20 8 8C, 20 %RH) 3) HWAS:s ame as (2) but underwent additional alkaline sulfide,N aOH/Na 2 SO 4 ,p H12-13 reduction treatment. [13] Now stored in controlled (20 8 8C, 20 %R H) environment Tw ot ypes of samples were collected:b ulk corrosion powders from the object surface (-C samples) and cut crosssections mounted in polyester resin (-S samples;T able 1). Crystalline phases were studied by SXPD of -C samples,while the chlorine distribution and speciation of -S samples were probed by synchrotron XRF mapping and Cl k-edge X-ray absorption near-edge spectroscopy (XANES), respectively. Details of the method may be found in the Supporting Information.
During cutting, cracks were observed in HW and HWAS shot, presenting as tructure (Figure 1b,c) that consists of at hin (< 0.5 mm) surface corrosion layer over the original Tu dor metal. Corrosion around large cracks and casting voids is associated with severe degradation and mechanical failure. In some cases,t his has led to fracturing, indicating that corrosion is occurring inside the artefact, rather than from the outside in.
Thec hlorinated phase most often associated with archaeological iron corrosion in chlorinated media is akaganeite, [22][23][24] commonly referred to as b-FeO(OH), though more closely defined as FeO 0.833 (OH) 1.167 Cl 0.167 . [22] Akaganeite has amonoclinic crystal structure (space group: I2/m), built from edge-and corner-sharing Fe(O,OH) 6 octahedra that form atunnel structure similar to the hollandite crystal. [22,25,26] Tw o Cl sites may be distinguished: [23] within the tunnel structure, Cl str ,a nd adsorbed to the surface, Cl sur of the crystal, Figure 4b.The size of individual b-FeO(OH) crystals depends on the growth conditions,but is typically small, around 0.15 0.03 mm, resulting in al arge proportion of Cl sur . [22] During aw ashing-based conservation treatment, mobile Cl sur is removed, while Cl str is thought to be unaffected. [14,22,23,27] Xray diffraction of akaganeite gives two low-angle peaks,which at the wavelength used in this study (l = 0.82578 )appear at 2q values 6.358 8 and 8.958 8,c orresponding to the (101) and (200) planes,r espectively.T he insets in Figure 2a highlight the location of these peaks (dotted grey lines) and show that, somewhat surprisingly,a kaganeite is not observed in the surface corrosion products of any of the SS shot, while it is present on the surface of all HW and HWAS shot. Phase identification of the powder diffraction profiles,F igure 2b, instead shows ac ombination of phases from the burial environment:c alcite (CaCO 3 ), quartz (SiO 2 ), and aragonite (CaCO 3 ); related to the microstructure of the metal:graphite and cementite (Fe 3 C);a nd commonly reported [3,14] marine iron corrosion products:goethite a-FeOOH, magnetite Fe 3 O 4 and lepidocrocite g-FeOOH.
Reflecting the SXPD results,w here no chlorinated corrosion products were observed on the surface of SS shot, Cl elemental maps of -S samples (Figure 3and the Supporting Information, Figure S7) show no chlorine at the surface of SS shot, while alocalised layer of Cl is present on the outer edges and around voids of HW and HWAS shot. However,inone SS shot, the example shown in Figure 3, an area of Cl in the inner region of the sample can be seen. Comparing the Cl k-edge XANES in this inner area to the chlorine species observed on aHWtreated shot (Figure 4) it can be seen that the Cl species is different. To identify the chlorine species present in the Mary Rose samples,alibrary of Cl standards was prepared by ac ombination of chemical purchases and synthesis (see the Supporting Information). Cl XANES from the standards were used to fingerprint spectral features in the sample datasets.T he library of standards included al ab-synthesised sample of hibbingite, b-Fe 2 (OH) 3 Cl, aprecursor to akageneite that has been observed on archaeological iron [3,28,29] but rapidly oxidises [30,31] to akaganeite in storage [32] or during  conservation [14] and al ab-synthesised standard of b-FeO-(OH), prepared by hydrolysis of an Fe 3+ chloride solution. To simulate the effect of aw ashing treatment, the akaganeite standard was immersed in 500 mL distilled H 2 Of or 1month at ambient conditions, with solution changes every 48-72 hours, giving as eries of incrementally washed standards AKA-1 (0 washes) to AKA-8 (7 washes). Comparing the spectra from the HW and SS shot, it can be seen that the HW spectrum is consistent with b-FeO-(OH), Figure 4c,while the spectrum from beneath the surface of the SS shot, Figure 4d,i sc onsistent with the precursor Fe II chloride, b-Fe 2 (OH) 3 Cl. Looking at the pre-edge region (Figure 4a), as mall feature is visible at 2819-2822 eV for the unwashed akaganeite that is decreased in amplitude for the washed standard. TheX ANES data from the series of washed standards, AKA1-8, were fitted with three peaks to compare the relative intensities of the observed features ( Figure 5). While the contribution from the second and third peaks remains constant throughout the series (Figure 5c, squares), the contribution from the first peak, the pre-edge,d ecreases with an increasing number of washes (circles and stars). This feature arises from electronic transitions in partially-bound Cl sur , [33] demonstrating that chlorine is lost from this site during washing. Fort he HW sample,areduced amplitude is observed, indicating that the phase has gone through several washing  [35] software and semiquantitative phase proportions, based on the reference intensity ratio method, [36] I/I c (c = corundum)accuracy AE 5wt%. stages to remove Cl sur . From this,i tm ay be inferred that b-FeO(OH) formed prior to,o ri nt he early stages of the washing treatment. Immediately before the HW shot underwent active conservation, it was stored at pH 10 and was analogous to SS shot, that is,nosurface b-FeO(OH), but subsurface chlorine present in pores as b-Fe 2 (OH) 3 Cl. Studies have shown that oxidation of hibbingite to akaganeite in solution is only thermodynamically feasible in the pH range 4-6, [30] and that complete transformation can occur in about 7hours. [30] As aresult, it is proposed that the sudden change in pH from 10 to near-neutral during active conservation, coupled with removal in tap water of adhering corrosion layers,l ed to exposure of previously-blocked pores and cracks,e nabling mobilisation of sub-surface chlorine and subsequent oxidation to akaganeite at the object surface.This transformation would have occurred while the artefact was in the first washing bath, resulting in loss of Cl sur in successive washes.T his conclusion is supported by ar ecent in situ study [34] of hibbingite on archaeological iron, where akaganeite was not observed to form in NaOH, but was observed after drying at the end of the treatment. However,while this mechanism can be used to explain akaganeite formation on this sample,i to nly represents as ingle location on as ingle artefact. In cases where fracturing has occurred to an object and an ew surface is exposed to the external environment, corrosion products can be observed [24] to form while in storage or on display.R ather, this investigation has shown that, alongside its formation after conservation, there are additional opportunities for akaganeite to form on archaeological iron, such as during or in-between treatment stages.
In conclusion, this work has used ac ombination of synchrotron techniques to gain an unprecedented insight into the effect of conservation choices on iron corrosion. It has been shown that during multi-decade immersion in sodium sesquicarbonate solution, chlorine is removed from the outer surface of artefacts;however, Cl can remain trapped in pores within the metal, in the form of hibbingite, b-Fe 2 (OH) 3 Cl. On exposure to oxygen at near-neutral pH, this phase rapidly oxidises to b-FeO(OH). In the case of the Mary Rose shot, it is proposed that this transformation occurred during treatment, owing to ac ombination of pH change and surface removal. Using chlorine XANES spectra, it has been shown that, years after conservation, it is possible to differentiate iron corrosion products formed during treatment. This reveals that post-conservation studies of artefacts,e ven decades after treatment, have an important role to play in the future development of conservation.