Microstructures of ancient and modern cast silver–copper alloys

M A TE RI A L S C HA RACT ER I ZA TI O N 90 ( 20 1 4 ) 1 7 3–1 8 4

Available online at www.sciencedirect.com

ScienceDirect www.elsevier.com/locate/matchar

Microstructures of ancient and modern cast silver–copper alloys S.M. Northovera,⁎, J.P. Northoverb a

Materials Engineering, The Open University, Walton Hall, Milton Keynes MK7 6AA, UK Department of Materials, University of Oxford, Parks Rd, Oxford OX1 3PH,UK

b

AR TIC LE D ATA

ABSTR ACT

Article history:

The microstructures of modern cast Sterling silver and of cast silver objects about 2500 years

Received 3 August 2013

old have been compared using optical microscopy (OM), scanning electron microscopy

Received in revised form

(SEM), transmission electron microscopy (TEM), scanning transmission electron microscopy

13 December 2013

(STEM), energy dispersive X-ray microanalysis (EDX) and electron backscatter diffraction

Accepted 27 January 2014

(EBSD). Microstructures of both ancient and modern alloys were typified by silver-rich dendrites with a few pools of eutectic and occasional cuprite particles with an oxidised rim

Keywords:

on the outer surface. EBSD showed the dendrites to have a complex internal structure, often

Sterling silver

involving extensive twinning. There was copious intragranular precipitation within the

Castings

dendrites, in the form of very fine copper-rich rods which TEM, X-ray diffraction (XRD), SEM

Precipitation

and STEM suggest to be of a metastable face-centred-cubic (FCC) phase with a cube–cube

Twins

orientation relationship to the silver-rich matrix but a higher silver content than the

EBSD

copper-rich β in the eutectic. Samples from ancient objects displayed a wider range of

TEM

microstructures including a fine scale interpenetration of the adjoining grains not seen in the modern material. Although this study found no unambiguous evidence that this resulted from microstructural change produced over archaeological time, the copper supersaturation remaining after intragranular precipitation suggests that such changes, previously proposed for wrought and annealed material, may indeed occur in ancient silver castings. © 2014 Elsevier Inc. All rights reserved.

1. Introduction Cast silver alloyed with copper has been used from at least the third millennium BC, for making both functional and decorative objects, but, in contrast to wrought silver alloys, their microstructures have been very little studied. It is important to our understanding of the history of these objects to know whether twin-like structures and surface oxidation layers, visible optically in some ancient cast silver objects, result from the manufacturing process, or from subsequent exposure to fire. More particularly, we need to know if they also show age-related microstructural changes. Almost all archaeological silver objects other than coins are hypoeutectic, and a

significant proportion is close to the Sterling composition of 7.5% Cu by weight. This study aimed to clarify the morphology and nature of the phases present in cast silver–copper alloys close to the composition of Sterling silver, both soon after casting and after a very long-term exposure at ambient temperatures, and to determine if there are microstructural features of cast silver alloys that can unambiguously be attributed to age. Objects made from silver alloyed with up to 10% Cu are well known to age-harden by the precipitation of copper, because of the difference in its solid solubility between the typical annealing temperatures of up to 700 °C (7%) and room temperature (< 1%). This age-hardening has been well studied

⁎ Corresponding author. Tel.: + 44 1865 486259; fax: +44 1908 653858. E-mail addresses: [email protected] (S.M. Northover), [email protected] (J.P. Northover). 1044-5803/$ – see front matter © 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.matchar.2014.01.028

174

M A TE RI A L S CH A RACT ER IZ A TI O N 90 (2 0 1 4 ) 1 7 3–1 8 4

in wrought, annealed and quenched pure binary Ag–Cu alloys [1–10]. Both continuous precipitation and discontinuous precipitation (DP) (classified on the basis of transformation mechanism not morphology) have been of interest and there have been many studies measuring DP growth rates at different temperatures mostly above 200 °C (e.g. [11]). The DP commonly takes the form of cellular ‘colonies’ growing away from grain boundaries [12–14], observed (by TEM) in homogenised and quenched Sterling silver aged at 200– 500 °C [9] to consist of fine rods of Cu-rich β phase surrounded by Ag-rich α. At lower ageing temperatures the colonies often appeared finely mottled or structureless rather than lamellar [14]. Homogeneous precipitation occurs alongside DP and hardness measurements [15], differential scanning calorimetry [14,16], dilatometry [14], and resistivity measurements [10,16] have indicated the formation of two metastable precipitates, but no continuous β precipitation prior to DP. During ageing at low temperatures (<175 °C), homogeneous precipitation initiates by the formation of Cu-rich zones, which subsequently lose coherency to form intermediate precipitates as growth proceeds [8,9,17]. On the basis of changes in physical properties, it is generally reported, (e.g. [4]), that, at temperatures >175 °C, intermediate precipitates form directly. However, streaking in TEM diffraction patterns, interpretable as due to the presence of zones, has been reported [8] after very short ageing times at 300 °C. Schweizer and Meyers [18] suggested using DP of copper at silver grain boundaries in wrought silver alloys as an indicator of historic age but it was later shown [19] that, with suitable heat treatments, it is possible to simulate some of these morphologies in modern wrought and annealed Britannia silver (Ag–4.16% Cu). More recently Wanhill and co-workers [20,21] have concluded that DP cannot be used to determine the age of silver objects but it may be evidence that age-related changes in microstructure can occur. The idea that DP can be related to historic age has, however, not previously been tested in cast silver alloys. One aim of this study was to find if there is any evidence of DP or of grain boundary migration in either ancient or modern cast silver alloys. During cooling of a Ag–Cu alloy casting, cored Ag-rich α dendrites form from the melt, and at 779.4 °C eutectic solidifies in between. Quenching produces both many more excess vacancies and a higher supersaturation of Cu, so precipitation in cast Ag alloys might be expected to differ significantly from that reported in experiments on wrought Ag alloys. Studies of the cast alloys are hindered by the very fine scale of the precipitation which forms within the grains

over a broad range of cooling rates. This challenges the resolving power of optical methods, while the use of either X-ray methods or TEM poses significant experimental difficulties because of the small samples of ancient material available for examination. The work presented here used a combination of SEM, EBSD, EDX, TEM, STEM and OM to explore the range of microstructures represented in cast silver from archaeological contexts, and compare them with those observed in modern cast Ag alloys, with the aim of both improving our understanding of precipitation in these alloys, and of finding out if this can provide information on the age of the artefact.

2. Materials and Methods 2.1. Modern Sterling Silver Modern Sterling silver was studied in the form of runners and sprue from lost wax castings, made at a sculpture foundry, into plaster moulds (so that the cooling rates were similar to those of ancient objects cast into clay). Both vacuum cast and traditionally cast material were examined, and samples were taken from the cup end, intermediately, and from the casting end, of runners and sprues of different sizes. Both longitudinal and transverse sections were examined.

2.2. Ancient Silver Samples were taken from a range of cast silver artefacts from the eastern provinces of the Achaemenid Empire (Eastern Iran and Western Afghanistan) dating from 6th–4th century BC. Their compositions, as determined by electron probe microanalysis, are shown in Table 1. Images of the ancient objects are available as Supplementary material in the online version of this article.

2.3. Methods Samples were given a conventional metallographic preparation to a 1 μm polish, then etched with ammoniacal hydrogen peroxide (17 ml NH4OH, 3 ml 30% H2O2, 10 ml H2O). Specimens for SEM and EBSD examinations were more lightly etched than those for OM, and were subsequently given a light C coating, before examining in a Zeiss Supra™ 55VP FEGSEM with an EDAX Genesis 4000 Energy Dispersive X-ray (EDX) spectrometry system and an Oxford Instruments NordlysF EBSD camera. The EBSD data acquisition and post-processing were carried out using HKL fast acquisition 5.11 and Channel-5

Table 1 – Analyses of archaeological cast silver objects studied (all concentrations in wt.%). Sample

Fe

Co

Ni

Cu

Zn

As

Sb

Sn

R2339/mean R2970/mean R2967/mean R2968/mean R2969/mean R2971/mean R2972/mean R2978/mean

0 0.01 0.01 0.01 0.01 0.01 0.01 0

0 0.01 0 0 0.01 0.01 0 0.01

0.01 0.02 0.01 0 0.01 0.01 0.01 0

8.49 8.26 21.86 12.48 10.08 7.25 10.35 12.79

1.19 0.01 0.02 0.02 0.02 0.02 0.01 0.01

0 0.01 0.01 0.01 0 0.01 0.01 0.01

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

Ag

Bi

Pb

Au

S

Al

Si

Mn

90.07 91.55 78 87.35 89.73 92.62 89.47 86.97

0.02 0.01 0.01 0.02 0.01 0.01 0.01 0

0.05 0.09 0.05 0.07 0.1 0.04 0.11 0.03

0.04 0.03 0.03 0.02 0.01 0.01 0.01 0.17

0.06 0.01 0 0.01 0 0 0 0

0.03 0 0 0 0.01 0.01 0 0

0.03 0 0 0 0 0 0 0

0.01 0 0 0 0 0 0 0

M A TE RI A L S C HA RACT ER I ZA TI O N 90 ( 20 1 4 ) 1 7 3–1 8 4

175

Fig. 1 – Microstructures of (a,b,c) large (~11 mm diameter) and (d) small (~3 mm diameter) modern cast Sterling silver runners. (a) Low mag. BSE image showing oxidised rim at the outer surface, (b) OM image and (c) and (d) BSE images.

software, respectively. For TEM EDX, specimens were extracted using a FEI Quanta 3D focused ion beam SEM (FIB–SEM) and examined in a JEOL 2100 using an EDAX Genesis XM4 System 60. A cross-section through a modern cast silver runner was also examined by X-ray diffraction (XRD) in a Bruker D5000 diffractometer, using CuKα radiation and calibrated against a quartz standard.

3. Results 3.1. Modern Sterling Silver The microstructures of the modern traditionally cast and vacuum cast Sterling silver were very similar, comprising predominantly

Fig. 2 – EBSD maps from two different areas of a modern cast Sterling silver runner. (a) Band contrast (BC) and orientation map of area 1. (b) BC map of area 1, with twin boundaries — red. (c) Orientation map of area 2. (d) BC map of area 2, with high angle grain boundaries (HAGBs) — black and twin boundaries — red.

176

M A TE RI A L S CH A RACT ER IZ A TI O N 90 (2 0 1 4 ) 1 7 3–1 8 4

Ag-rich dendrites with the dendrite arms at approx. 90° to their length, with small volumes of eutectic between some of the dendrite arms. The dendrites contained fine precipitates which deep etching showed primarily to be arrays of rods. Optical and backscattered electron (BSE) SEM micrographs are shown in Fig. 1. One small runner had a slightly different morphology with blobs of Cu rather than eutectic pools, as seen in Fig. 1(d). This runner also had areas within some dendrites that appeared to contain no precipitated rods. The compositional contrast of the BSE images, e.g. Fig. 1(c), showed the rods to be Cu-rich and only 40–60 nm across, so they are visible in the optical images only through the wide trench etched at their interfaces with the silver-rich matrix. Within the colonies the rods were approximately 0.8–1.0 μm apart. EDX analysis gave an overall composition of Ag–7% Cu, with the Cu-rich phase of the eutectic analysed as Cu–5% Ag and the Ag-rich matrix of the dendrites as Ag–3% Cu. The eutectic pools appear to consist of alternating Cu-rich, β, and Ag-rich, α, lamellae or β rods separated by α, each about 400 nm wide, with slightly more β at their edges. These pools are surrounded by narrow, rod-free, zones at the dendrite edges as seen in Fig. 1(c). Although there is continuity between the α of the dendrites and of the eutectic, there was no connection between the β phase of the eutectic and the Cu-rich rods. In some areas the rods were straight and in other areas they were bent. The occurrence of these rods was quite general and other runners showed rod spacings down to 0.5 μm, with most 0.6–0.8 μm. Quite near the edge of one sample there were some very long rods (>30 μm) which were mostly straight but with a slight bend at one point. Some areas contained precipitation in a Widmanstätten-like pattern but in others it was more irregular. As seen in Fig. 1(a), close to the outer surface of the runners was a clearly defined oxidation layer containing cuprite (Cu2O) particles of a range of sizes. Oxygen has a low solid solubility in silver and Ag2O is unstable at the alloy liquidus temperature. The solubility of oxygen in molten silver is much higher and even in the vacuum-cast material there was some cuprite (identifiable by its characteristic angular morphology and red colour in dark field OM) but there was more in the traditionally cast material. Within the body of the casting, cuprite is the first material to freeze and the larger particles tend to end up in the last areas to freeze, i.e. between the dendrites. In some regions of the traditionally cast material, cuprite was the major interdendritic phase. As seen in the EBSD maps of Fig. 2, rather than the dendrites being monolithic and of constant orientation, they have a complex internal structure, containing some low angle boundaries but, most markedly, with significant twinning. This is shown by the red lines marking the twin boundaries in Fig. 2(b) and (d). Twins were frequently seen in only one set of the interpenetrating orientations (Figs. 2(b), (d) and 3); usually, but not always, this was the orientation giving higher band contrast. In many cases material of the two different orientations contained different levels of plastic strain from each other, e.g. Fig. 3. Fig. 2(c) and (d) also shows that the extended mutually interpenetrating dendritic volumes were themselves sometimes twin-related. The α and β of the eutectic frequently had a heterotwin relationship as also shown in Fig. 2(b).

Fig. 3 – EBSD maps from an area close to the centre of a transverse section through a modern Sterling silver runner. (a) Orientation map, (b) BC map with twins boundaries — red, and (c) grain average misorientation map.

EBSD orientation maps showed a particularly high level of deformation in some parts of the dendrites close to the centre of a runner. Since this effect was more marked towards the centre of the runner, it is more likely caused by the thermal stresses generated on cooling rather than stresses induced on transformation. Although the heterotwin structure of the eutectic was clearly shown in the orientation maps, the rod-shaped precipitates within the dendrites were invisible in orientation maps, showing them to be cube–cube oriented to the α. However the precipitates were visible in some of the band contrast (BC) maps as lines of reduced BC e.g. the fine darker lines in the areas marked R in Fig. 2(b).

3.1.1. X-ray Diffraction A spectrum obtained from a transverse cross-section through one of the larger modern cast Sterling silver runners is shown in Fig. 4, with the predicted positions and relative intensities of the peaks expected from random polycrystalline pure Ag and pure Cu also shown. All the XRD peaks could be indexed

M A TE RI A L S C HA RACT ER I ZA TI O N 90 ( 20 1 4 ) 1 7 3–1 8 4

177

Fig. 4 – XRD spectrum from a cross-section of a runner of a modern Sterling silver casting (CuKα radiation). as either FCC silver with a lattice parameter slightly smaller than that of pure Ag or FCC copper with a lattice parameter slightly larger than that of pure Cu. The intensity of the copper peaks was < 3% that of the Ag peaks and the shift in the Ag

lattice parameter was greater than that in the Cu. The lattice parameters calculated from this spectrum were a = 0.40707 ± .00007 nm and a = 0.3619 ± .00005 nm for the Ag-rich and Cu-rich phases respectively.

Fig. 5 – Microstructure of a cast silver horse's head. (a) BSE SEM image, (b) EBSD band contrast map showing high angle grain boundaries (black) and twin boundaries (red), and (c) EBSD orientation map over a grain boundary.

178

M A TE RI A L S CH A RACT ER IZ A TI O N 90 (2 0 1 4 ) 1 7 3–1 8 4

The samples taken from ancient objects were too small for XRD analysis.

3.2. Ancient Cast Silver Objects Many of the ancient cast silver objects had an oxidised layer at their outer surfaces similar to that seen in the modern silver (Fig. 1(a)). Fig. 5 shows an SEM image and EBSD maps from a cast vessel in the shape of a horse's head (composition R2970 in Table 1). Fig. 5(a) clearly shows colonies of rods and pools of eutectic, but the cast structure rather obscures the grain boundaries in both optical and SEM images. Fig. 5(b) shows that the as-cast grain structure was similar to that of the modern Ag alloy in that there was twinning. Orientation images, however, as in Fig. 5(c), clearly showed interpenetration of adjoining grains, suggesting cellular growth at boundaries.

Fig. 6 – BSE images of a drinking cup in the shape of a ram's head.

This effect was not observed in any of the modern Sterling silver. Fig. 6 shows BSE images of a rhyton (drinking cup) in the shape of a ram's head (composition R2967 in Table 1). As would be expected from its higher Cu content, there was a much larger proportion of eutectic in this microstructure. Although there is no obvious compositional contrast within the α dendrites, they contain linear features and indications of a ‘basket-weave’ structure, aligned with the major growth directions of the dendrites. There also appears to be a rim of irregular thickness around the edges of the dendrites, isolating the basket-weave structure from the eutectic. Many of the ancient objects had microstructures very similar to those seen in the modern cast silver with twinning and with extensive precipitation of rods within the α dendrites. Fig. 7 illustrates these features in optical and BSE SEM images of two beakers (compositions R2969 and R2968 in Table 1). OM and BSE SEM of two other beakers (R2971 and R2972 in Table 1) similarly showed evidence of twinning, a similar regular arrangement and morphology of the eutectic pools and linear and dotted features within the dendrites. Fig. 8 shows the unusual microstructure of a sample taken from the rim of the tube of an eagle rhyton (composition R2339 in Table 1). The remnants of the dendritic coring are visible within the grains, but, unlike the other material studied, there is no twinning within the α, and the colonies of DP at the grain boundaries have a spiky outline. There was no evidence from OM or SEM of intragranular precipitation, but no ion imaging or TEM has been performed to confirm this. A sample of metal

Fig. 7 – Microstructure of (a) beaker with three ibex (OM image) and (b) double ibex beaker (BSE SEM image).

M A TE RI A L S C HA RACT ER I ZA TI O N 90 ( 20 1 4 ) 1 7 3–1 8 4

179

as seen around B in Fig. 9(b). Drilling down into the surface, using a focused Ga ion beam, over the interface between a cellular colony (marked A in the figure) and the ‘untransformed’ dendrite (marked B), showed this ‘basket-weave’ not to be a surface effect but, as shown in area B of the ion image Fig. 9(c), to result from the presence of very fine crystallographically-arranged precipitates within the silver-rich dendrites. Ga ion images and TEM showed the cellular colony to consist of copper-rich rods surrounded by silver-rich regions. Neither of the rods examined from two different colonies had a cube–cube orientation relationship to the adjoining silver and the relationship between the two phases could not be indexed as a simple heterotwin. Fig. 10(a) is a bright field STEM image of an intragranular region away from the discontinuous precipitation; STEM EDX analysis along the line AB, Fig. 10(b), shows the bright areas to be Cu-rich precipitates around 30 nm across frequently arranged in rows. A selected area diffraction pattern from this area, Fig. 10(c), showed a cube–cube orientation between the matrix and the Cu-rich precipitates and, within the accuracy of measurements on the diffraction patterns, the ratios of the lattice spacings agreed with those of Cu-rich and Ag-rich phases of the equilibrium composition. The cube–cube orientation gives rise to very strong double diffraction effects, giving rows of satellite spots beside the strong matrix reflections. None of the diffraction patterns showed streaks, which would have indicated that the precipitates are thin plates, or sidebands, which would have indicated regular compositional fluctuations arising from spinodal decomposition. The dark field TEM image of Fig. 10(d) was obtained using the reflection marked 1 in Fig. 10(c). Closer to the interface with the cellular colony the small Cu-rich precipitates were sparser and not apparently arranged in rows. It was noticeable that there was quite a high dislocation density in the matrix.

4. Discussion

Fig. 8 – Microstructure of the rim of a cast silver rhyton. (a) OM image, (b) secondary electron (SE) SEM image, and (c) EDX composition map.

from the eagle supporting the cup (of the same composition except for not containing any Zn) did not show any intragranular precipitation either, but did show twinning. Areas of DP in this sample had smoothly curved rather than spiky boundaries. Material from a cast silver chicken, R2978 of Table 1, which clearly showed features seen in the other cast material, was selected for further examination, and a TEM sample was extracted using a FIB–SEM. Fig. 9 shows optical, SEM and ion images from the same area. The oxidised outer surface layer, the angular cuprite, the dendritic structure with eutectic pools in between the dendrite arms, and a similar structure within the cellular colonies to that seen in the eagle rhyton, are all clearly visible. Additionally the SEM showed a very fine ‘basket-weave’ etching effect within the ‘untransformed’ parts of the dendrites

4.1. Oxidised Surface Layer The surface layers of even vacuum-cast Sterling silver, cast into plaster moulds, are oxidised with a surface layer similar to that seen in Fig. 1. This shows that the presence of an oxidised surface layer, seen on many of the ancient cast silver artefacts, does not imply any exposure to fire after manufacture but must be due to oxygen from air in the mould matrix.

4.2. Twinning Since Ag has a relatively low stacking fault energy, both growth and deformation twins form very readily and, as seen in Figs. 2, 3, 5 and 7, extensive twinning within the α was typical of both the ancient and modern as-cast microstructures. The interpenetrating volumes related by twinning such as in Fig. 2(c) and (d) might be the result of growth twinning during dendrite formation, with different dendrite arms, one twinned and the other not, freezing together. In other cases, such as Figs. 2(b), 3, 5, and 7(a), twinning most likely, resulted from deformation caused by shrinkage stresses on cooling. Twinning was

180

M A TE RI A L S CH A RACT ER IZ A TI O N 90 (2 0 1 4 ) 1 7 3–1 8 4

Fig. 9 – Microstructure of a cast silver chicken. (a) OM image, (b) SEM image, and (c) Ga ion image of a FIB cut section.

a

c

b

d

Fig. 10 – (a) Bright field STEM image of an intragranular region away from any DP. (b) Composition along line AB (total length 0.476 μm). (c) SADP of an area within (a). (d) Dark field TEM image of copper-rich precipitates using reflection marked 1 in (c).

M A TE RI A L S C HA RACT ER I ZA TI O N 90 ( 20 1 4 ) 1 7 3–1 8 4

particularly marked close to the centre of the modern runners, a region where a high level of deformation was apparent from the orientation maps. It was notable that there was no twinning in a sample taken from the rim of the thin-walled cup of the eagle rhyton but there was twinning in a sample taken from the eagle supporting it where the section was thicker. Where the precipitation was visible in the BC maps there was no apparent relationship between the precipitates and the twins. Grain average misorientation maps, such as Fig. 3(c), showed that the marked difference in band contrast, often seen between the different areas of different orientations, was due to differences in the levels of plastic strain in the material of the two different orientations, rather than from dynamical diffraction effects (which can also cause differences in the sharpness of patterns from differently oriented perfect crystals). This difference in plastic strain probably arises from the different orientations of the different crystallites with respect to the contraction stresses produced by solidification and cooling. Differently oriented regions will experience different resolved shear stresses on their favoured slip systems, and so deform plastically to differing extents. Regions unfavourably oriented for slip may also deform by twinning. When Cu precipitates from the supersaturated solid solution of the silver alloy, the transformation produces a large volume change but the areas of reduced BC in the EBSD maps are strongly localised, delineating the edges of the precipitates, as seen in Fig. 2(b), and so are unlikely to initiate the larger scale twinning seen so extensively. Since the rod-like precipitates were seen throughout both the twinned and un-twinned areas of the cast structure the twinning is unlikely, in any way, to relate to strains produced on transformation.

4.3. Orientation Relationships EBSD observations frequently showed a hetero-twin relationship between the Ag-rich and Cu-rich phases in the eutectic. However neither of the two Cu-rich rods, from different cellular regions, observed by TEM were in either a heterotwin or a cube– cube relationship to the adjacent Ag-rich phase. The invisibility of the rod-shaped precipitates in the EBSD orientation maps and the orientation relationships seen in the TEM diffraction patterns confirm that the small Cu-rich precipitates have a cube–cube crystallographic relationship to the parent α dendrite, as previously reported for very fine continuous precipitates in quenched Ag–7.3wt.% Cu aged at 300 °C [8]. The dislocations in the plastically deformed matrix would provide both heterogeneous nucleation sites for the intragranular precipitates and rapid diffusion paths for solute supply during their growth. There is a large lattice mismatch between Cu and Ag, and the visibility of the Cu-rich rods in the BC images probably results both from the strain field of the array of interfacial dislocations which accommodates this mismatch and directly from the local coherency strains at the interface.

4.4. Precipitate Morphologies and Transformation Mechanisms The small eutectic pools between the dendrite arms in the modern Sterling silver appeared either as intimate mixtures of α and β or as globular copper. The globular form was observed in the smallest sprue and this is consistent with the

181

increased tendency of alloys, far from the eutectic composition, to form divorced eutectics at higher cooling rates [22]. The as-cast microstructure of alloys of the eutectic composition, although frequently referred to as lamellar, has been seen to consist of colonies of copper rods embedded in silver [23,24]. Barrett et al. [25] quote Hansen [26] as finding slow cooling of silver-rich Ag–Cu occasionally to produce structures showing ‘needles similarly oriented in a rough way’. TEM observations [9] have shown that DP of Cu produced in Sterling silver by quenching and ageing consisted of fine rods, and this was confirmed, by the TEM of the present study, also to be the case in DP in ancient cast silver alloys. Although on a much coarser scale than the Cu-rich precipitates generally found within the α of both the ancient and modern Ag alloys in this study, these observations all confirm that a needle-like morphology is energetically favoured when the Cu-rich phase precipitates from the Ag-rich one. Careful examination of micrographs of Ag–15% Cu etched in dilute nitric acid published in 1897 [27] and reproduced in 1919 [24] shows colonies of rods or lamellae within the Ag dendrites very similar to those seen in the present study but unremarked by previous authors. EBSD showed the presence of some low angle boundaries within the dendrites, and it is likely that the slight bend observed in the particularly long rods seen in one sample is due to their crossing such a boundary. The differences between the areas where the precipitates in the α dendrites formed a Widmanstätten-like pattern and those where they did not are probably due to differences in their cooling rates. It is known that in Al–Cu, precipitates formed at higher temperatures by slow initial cooling are more likely to form Widmanstätten patterns than faster cooled specimens where precipitate growth takes place mostly at lower temperatures. Barrett et al. [25] tried many different heat treatments on Ag alloys with 3–12% Cu, attempting to produce a Widmanstätten structure and achieved it only in an Ag–4% Cu alloy which was cooled extremely slowly over 110 K from the solidus, then air cooled to room temperature. It is possible that the areas where the rods were arranged into more geometrical patterns cooled more slowly, and that those regions where the rods were simply aligned were those which cooled faster. In the regions which cooled faster, the relationship between crystallographic directions allowing particularly fast growth and the local direction of maximum temperature gradient was more likely to be the dominant influence. < 100 > are frequently directions of fast growth in FCC crystals and the disposition of the dendrite arms in the modern cast Sterling silver suggested this to be the case here. EBSD of regions where the rods were approximately perpendicular to the specimen surface also showed the surface normal to be close to [100] there. The morphology of the precipitation within the dendrites of the modern cast Sterling silver is very similar to that of the large cellular colonies in some of the dendrites of the ancient cast silver chicken, suggesting that the structure seen in the modern silver arises by discontinuous precipitation. This has subsequently been confirmed by identifying characteristic boundaries within the dendrites in both cases [28]. Under isothermal conditions, decomposition by discontinuous precipitation, to form colonies of rods, is favoured over homogeneous precipitation under conditions of high supersaturation and low temperatures [6,7]. The cooling rate of the cast silver

182

M A TE RI A L S CH A RACT ER IZ A TI O N 90 (2 0 1 4 ) 1 7 3–1 8 4

chicken appears to have been such as to allow both competing processes of decomposition, homogeneous precipitation and cellular precipitation, to occur simultaneously. Previous TEM, carried out on quenched and aged Sterling silver [8,9,29], has shown precipitation during low temperature ageing to initiate as plate-shaped, coherent Cu-rich clusters/ zones on {100} planes, causing streaking in the diffraction pattern normal to <100>, which were still visible after 8 h at 200 °C [8]. At longer ageing times the platelets were replaced by more spherical or oblong precipitates, of an incoherent intermediate phase, which fell out during specimen preparation [9]. Pawlowski [17] suggested that the shape of the initial clusters and transition phases were very sensitive to third element additions and that in the pure binary alloy they were spherical. He reported that the addition of only 0.02% Ti caused a change to cube–cube oriented platelets of an intermediate β′1 phase while with 0.2% Ti additions the β′ phase formed in elliptical particles about 10–20 nm in size with no strong preferred alignment. Microprobe analysis of the cast Ag–Cu alloy investigated here by TEM showed that the only impurities present at a level above 0.01% were 0.17% Au and 0.03% Pb, which would both be expected to remain in solid solution. The shapes of the precipitates seen here by TEM are very similar to those of the intermediate precipitates seen in [8] and [9],and since these precipitates, presumably formed during cooling, probably nucleated and grew at above 200 °C. The absence of any streaking is consistent with the inferences from X-ray measurements, hardness and resistivity changes [4], that, above about 175 °C, precipitation proceeds directly by nucleation and growth of the intermediate phase, without the prior formation of fully coherent clusters.

4.5. Composition of the Transformation Products The Ag–Cu system shows quite large deviations from Vegard's law [30] but the lattice parameters of both Cu-rich and Ag-rich homogeneous annealed and quenched alloys have been measured using XRD [2,31]. The equilibrium lattice parameters of the Ag-rich α differs from that of the solid solution containing 7.5% Cu by only about 1% and from that containing 12.8% Cu by only about 1.5%, while the lattice constant of Cu containing Ag at the concentration of its maximum solubility at the eutectic temperature, is less than 1% greater than that of pure Cu. These differences are within the accuracy of conventional TEM comparisons of lattice parameter so TEM SADPs could not distinguish between precipitates of equilibrium composition and the metastable compositions of the products of cellular decomposition. The room temperature equilibrium solid solubilities are so low that the lattice constants of the equilibrium α and β may be taken as those of pure Ag (0.40857 nm) and pure Cu (0.36146 nm) respectively. The measured lattice parameters of 0.40707 ± .00007 nm and 0.3619 ± .00005 nm show that neither phase has reached the equilibrium composition and that the supersaturation of Cu in the α could provide a chemical driving force for further transformation over extended periods at room temperature. Both the Cu peaks of the XRD spectrum and 1

There are different nomenclatures used in the literature, but here α and β are used to designate Ag-rich and Cu-rich phases respectively.

those from the Ag will include contributions from both the material in the eutectic pools (<5%) and that in the dendrites. The phase fractions within the dendrites cannot be determined optically because of the fine scale of the precipitation but the BSE and ion channelling images show the copper-rich phase to comprise less than 10% of the total. The phase diagram suggests that the copper-rich phase should make up about 1/3 of the eutectic pools but the proportions appeared roughly equal in most cases, demonstrating Cu rejection from the dendrites during solidification. The Ag-rich phase of the eutectic appears continuous with that of the dendrites but the Cu-rich phases in each appear totally isolated from each other. The marked disparity in the relative intensities of the X-ray peaks from those predicted for a random polycrystal reflect the strong crystallographic texture arising from the dendritic growth. The compositions in the eutectic would be expected to be close to those given by the limits of solubility at the eutectic temperature but the compositions within the dendrites might be different for two different reasons. Firstly, the equilibrium solid solubility of Cu in Ag changes with temperature, so the composition of the core of the dendrite, which formed at a higher temperature, will probably be different from that at its edges, which solidified at a lower temperature. Secondly, during growth of a cellular colony the compositions of the two product phases are generally metastable with respect to the equilibrium composition. Gust et al. [15] determined the limit of stability of DP in silver-rich Ag–Cu alloys to lie about 3.6wt.% Cu within the limit of the equilibrium two phase region. As the equilibrium solid solubility of Cu in Ag decreases from over 7% around the eutectic temperature to less that 1% at room temperature, β formed during the early stages of cooling could leave a significant supersaturation of Cu in the remaining α which might provide a driving force for grain boundary migration or further phase transformation in the cold casting over extended periods of time at room temperature. The rim around the dendrites seen in Fig. 6 suggests that during cooling there is competition for solute between the continuous precipitation of β or β′ in the α dendrites and further precipitation on the β present in the eutectic. The waviness of the rim's boundary may be the result simply of a sensitivity of the etch to a particular compositional threshold but, alternatively, may mark the compositional limit of stability of the metastable β′, as determined on the basis of hardness measurements [15].

4.6. Possible Microstructural Changes Over Archaeological Time The limits of the spinodal reaction have been calculated [32] to lie at about 460 °C for Ag–12wt.% Cu and about 140 °C for Ag–6 wt.% Cu. Since diffusion rates are very low at 140 °C, this would appear to preclude any contribution from spinodal decomposition during cooling of cast Sterling silver. It does, however, suggest that, unless the supersaturation of Cu in the α has already been reduced by the nucleation and growth of precipitates during cooling, spinodal decomposition might occur in this alloy over archaeological time. For alloys richer in Cu, such as that of the cast silver chicken, R2978, with 12.8% Cu, the spinodal limit would be above 400 °C so much more of the cooling time would have been spent at temperatures where both decomposition was energetically favoured and diffusion

M A TE RI A L S C HA RACT ER I ZA TI O N 90 ( 20 1 4 ) 1 7 3–1 8 4

rates were still high enough to allow transformation to occur. However, in the present study there was no diffraction evidence (in the form of sidebands around principal TEM diffraction spots) for spinodal decomposition. The ancient cast Ag–Cu alloy objects displayed a wide range of microstructures. Some, such as the beaker shown in Fig. 7, were very similar to those of modern cast silver but others, such as that of the rim of a cast silver rhyton and the cast silver horse's head (Figs. 9 and 5), are quite different. The rim of the rhyton has a very thin section, so it may both have cooled faster than most of the other material studied, suppressing precipitation, and have generated lower internal stresses during cooling than a thicker sheet would have done. Both these effects would have reduced the tendency to twinning and might explain why the sample from this object did not show any twins. In addition to the discontinuous precipitation as seen in the eagle rhyton and the chicken, the fine-scale interpenetrations of adjoining grains shown in the EBSD maps of Fig. 5 from the cast horse's head clearly suggest secondary cellular growth. No similar boundary modification was observed in any of the modern Sterling silver. This may be due to differences in their cooling histories, or might suggest that, over archaeological time, it is possible for boundary modification, similar to that thought to occur in some wrought and annealed structures, to take place in cast material. This would only be possible if, after cooling, there remained sufficient supersaturation of Cu in the α, to drive grain boundary migration and further transformation towards the room temperature equilibrium compositions of α and β. The XRD measurements on the modern cast Sterling silver suggest that this is, indeed, the case. Unlike the case of wrought and annealed material in which boundary modifications produced by short term high temperature ageing cannot be distinguished from those produced at ambient temperature over archaeological time, the simultaneous effects of any heat treatment on the intragranular precipitates present in these cast microstructures might allow such a distinction.

4.7. Confirmation of Age-related Boundary Changes in Cast Structures Metallographic methods that rely on surface relief or colour contrast often do not reveal the as-cast grain boundaries in cast Ag–Cu; this effect is compounded by the obscuring effect of general discontinuous precipitation within the grains. This limits the usefulness of optical and both SE imaging and BSE imaging in the SEM in this context. Although TEM is useful in determining the crystallography of the fine precipitation, it cannot be used to locate and examine a sufficient length of as-cast grain boundaries to identify and characterise any modifications. EBSD is the only technique that can give the necessary crystallographic information over a wide enough area. Experience with wrought, annealed and quenched silver alloys from archaeological contexts, has shown there to be a range of Cu contents, and hence supersaturations, where boundary changes identified as age-related [20] are most clearly defined. This range is approximately 3–6% Cu. This suggests that there may also be a critical range of Cu contents where age-related boundary changes may be visible in cast

183

Ag alloy structures. Because the extensive discontinuous precipitation during cooling has removed a certain amount of Cu from solid solution the copper contents will be higher than that in wrought Ag–Cu alloys. In the cast silver horse's head where the current work proposes such changes have occurred (Fig. 5) the copper content was 8.3% which gives a measure of the likely range required. To establish whether such modifications are typical in archaeological samples will require a systematic survey of ancient samples of a range of Cu contents of 0–20% from objects of similar mass and cross-section, and of similar age. The similarity of mass and cross-section will ensure, as far as possible, a similar thermal history. The proportion of Cu removed from the solid solution by discontinuous precipitation during cooling of the casting is, as yet, not well-defined, but using pieces expected to have had a similar cooling rate will, as far as possible, remove one variable. If this work does indicate a definite composition range over which these boundary changes are observed, it is then necessary to investigate if similar changes can be induced in modern cast Ag–Cu alloys of a similar composition to decide if they can unambiguously be identified as age-related. By analogy with work on wrought and annealed Britannia silver [19] such changes might be expected to occur at low supersaturations and low temperatures. If the survey of different Cu contents shows Sterling silver to lie within the range of Cu contents where such changes occur, a simple first experiment would be to heat treat Sterling silver sprue at temperatures between 150 and 350 °C for times ranging from 0.5 to 150 h.

5. Summary/Conclusions The microstructures of modern cast Sterling silver and a variety of ancient cast Ag–Cu alloy objects of roughly similar Cu contents have been examined by a variety of metallographic techniques. The principal conclusions are: • The twinned structures and oxidised surface layer of ancient as-cast Ag–Cu alloys are exactly paralleled in modern as-cast Sterling silver. • A dendritic structure with intragranular precipitation, in the form of fine Cu-rich rods, whose alignment depends on both crystallography and heat flow, is characteristic of both ancient and modern as-cast Ag–Cu alloys. • The Cu-rich intragranular precipitates have a cube–cube orientation to the silver matrix. • A heterotwin relationship is commonly found between the phases in the eutectic pools. • The fine rods appear to have formed by discontinuous precipitation. • The Ag-rich α phase in castings of Sterling silver remains supersaturated with Cu. • There was no evidence of spinodal decomposition in either the ancient Ag–Cu alloys or the modern cast Sterling silver. • Some ancient as-cast Ag–Cu alloys show secondary cellular precipitation and other possibly age-related grain boundary modifications not seen in the modern material.

184

M A TE RI A L S CH A RACT ER IZ A TI O N 90 (2 0 1 4 ) 1 7 3–1 8 4

• An EBSD survey of ancient cast Ag–Cu alloy objects of a range of Cu contents but similar age and thermal history is required to establish under what conditions such modifications may be typical in archaeological samples. • Experiments on modern cast Ag–Cu alloys are needed to determine if such changes can be induced by heat treatment.

Acknowledgements The authors are grateful to Mr Gordon Imlach of Materials Engineering, The Open University, Mrs Heather Davies of The Open University Interfaculty Electron Microscope Unit, Dr Diane Johnson of The Open University Centre for Earth, Planetary and Space Research and Dr Phil Holdway of Oxford Materials Characterisation Services for their generous assistance in SEM imaging, STEM, TEM specimen preparation using the FIBSEM, and XRD respectively. Thanks are also due to Mlle Olivia Masson for making the ancient objects available for study, to Mr Rungwe Kingdom of Pangolin Editions for providing the modern Sterling silver sprue, and to Ms Alison Wilson for perfecting the etch.

Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.matchar.2014.01.028.

REFERENCES

[1] Norbury AL. The effect of quenching and tempering on the mechanical properties of standard silver. J Inst Met 1928;39:145–61 [and discussion 162–72]. [2] Ageew N, Hansen M, Sach G. Entmischung und Eigenschaftanderungen ubersattigter Silber-Kupferlegierungen. Z Phys A 1930;66(5–6):350–76. [3] Wiest P. Age-hardening process in silver–copper single crystals. Z Metallkd 1933;25:238–40. [4] Cohen M. Aging phenomena in a silver-rich copper alloy. Trans Am Inst Min Metall Eng 1937;124:138–57. [5] Cox WF, Sykes C. Precipitation in the alloys of copper and silver during age-hardening. J Inst Met 1940;66:381–7. [6] Jones FW, Leech P, Sykes C. Precipitation in single crystal of silver-rich and copper-rich alloys of the silver–copper system. Proc Roy Soc A 1942;181(985):154–68. [7] Gayler MLV, Carrington WE. Metallographic study of the precipitation of copper from a silver-rich silver–copper alloy. J Inst Met 1947;73:625–39. [8] Leo W. An electron microscopical study of precipitation in hardenable silver–copper alloys. Z Metallkd/Mater Res Adv Tech 1967;58(7):456–61. [9] Scharfenberger W, Schmitt G, Borchers H. Kinetics of discontinuous precipitation in silver alloys with 7.5 wt.% Cu. Z Metallkd/Mater Res Adv Tech 1972;63(9):553–60. [10] Youssef SB. Resistometric study of Ag–8 at% Cu alloy aged in the temperature range 0.4–0.65 Tm. Phys B Condens Matter 1996;228(3–4):337–41.

[11] Gust W, Predel B, Diekstall K. Discontinuous precipitation in silver–6.2 at% copper tri-crystalline specimens — 2. Experimental results for fine lamellar discontinuous precipitation at three grain boundaries. Z Metallkd/Mater Res Adv Tech 1978;69(2):75–80. [12] Rose R. The precipitation of copper from a silver–5.5% copper solid solution at 220 °C. Acta Metall 1957;5(7):404–5. [13] Predel B, Ruge H. Untersuchung der Kinetik der diskontinuierlichen Entmischung Übersättigter Silber-Mischkristalle. Z Metallkd 1968;59:777–81. [14] Hamana D, Boumaza L. Precipitation mechanism in Ag–8 wt.% Cu alloy. J Alloys Compd 2009;477(1–2):217–23. [15] Gust W, Predel B, Diekstall K. Experimentelle Ermittlung von Drei Metastabilen Löslichkeitslinien auf der Ag-seite des Zustandsdiagramms Ag–Cu. Acta Metall 1978;26(2):241–5. [16] Colombo S, Battaini P, Airoldi G. Precipitation kinetics in Ag–7.5 wt.% Cu alloy studied by isothermal DSC and electrical-resistance measurements. J Alloys Compd 2007;437(1–2):107–12. [17] Pawlowski A. Mechanism of precipitation in aged silver–copper alloys with small additions of titanium or magnesium. Arch Hut 1981;26(4):601–15. [18] Schweizer F, Meyers P. Authenticity of ancient silver objects: a new approach. Masca J 1978;1(1):9–10. [19] Dye A. Precipitation phenomena in silver–copper alloys. MChem Thesis. Oxford University; 1998. [20] Wanhill RJH, Hattenberg T, Northover JP. EBSD of corrosion, deformation and precipitation in the Gundestrup cauldron. In: Ferreiro da Silva AC, Menino Homem P, editors. Ligas Metálicas: Investigação e Conservação. Porto: Faculdade de Letras da Univsidade do Porto; 2008. p. 47–61. [21] Wanhill RJH. Significance of discontinuous precipitation of copper in ancient silver. Metallogr Microst Anal 2012;1:261–8. [22] Dantzig JA, Rappaz M. Solidification. Lausanne: EPFL Press; 2009 367. [23] Osmond F, Stead JE. Microscopic analysis of metals. London: Charles Griffin & Company, Limited; 1904. [24] Smith EA, Turner H. The properties of standard or Sterling silver, with notes on its manufacture. J Inst Met 1919;22:149–92. [25] Barrett CS, Kaiser HF, Mehl RF. Studies upon the Widmanstätten structure, VII — the copper–silver system. Trans Am Inst Min Metall Eng 1935;117:39–60. [26] Hansen M. Die Härte silberreicher Kupfer–Silberlegierungen. Bestimmung der Löslichkeit von Kupfer in Silber mit Hilfe von Härtemessungen. Z Anorg Allg Chem 1929;186(1):41–8. [27] Osmond F. Microscopic examination of alloys of the silver–copper group. Bulletin de la société d'encouragement pour l'industrie nationale, 2; 1897 837 [5th series]. [28] Northover SM, Northover JP, Wilson A. Discontinuous precipitation in silver–copper alloys. Euromat 2013. Spain: Seville; September 2013 8–13. [29] Pawlowski A, Dutkiewicz J, Litynska L, Ciach R, Salawa J. Phase transformations in silver–copper alloys containing magnesium or titanium additives. Arch Hut 1981;26(4):601–15. [30] Lubarda VA. On the effective lattice parameter of binary alloys. Mech Mater 2003;35(1–2):53–68. [31] Ageew N, Sachs G. Z Phys 1930;3:293–303. [32] Murray JL. Calculations of stable and metastable equilibrium diagrams of the Ag–Cu and Cd–Zn systems. Metall Mater Trans A 1984;15(A(2)):261–8.