Introduction

The potential of lead isotope amount ratios, hereafter referred to as lead isotope ratios, for unravelling the provenance of archaeologic artefacts made from lead was shown by Brill and Wampler already in 1965 (Brill and Wampler 1965). During the following decades, these provenance studies were expanded to silver and bronze artefacts (Wagner et al. 1980, Gale and Stos-Gale 1981, Mabuchi et al. 1985). In parallel, European and Mediterranean mining sites of galena, the major lead ore, but also of ores and minerals being used for copper-based production, have been investigated for their lead isotopic composition (Stos-Gale et al. 1996, 1998). Additionally, the metallurgical processes have been studied to exclude critical isotope fractionation. This was either done via retrospective analysis, which means lead isotope measurements of all materials associated with the production process, such as ores, slags and products (Yener et al. 1991, Gale and Stos-Gale 1996, Georgakopoulou et al. 2011) or via experimental archaeology, which mainly covers the repeat of the metallurgical processes such as cupellation under controlled conditions, while starting material and product were analysed for their lead isotopic composition (Cui and Wu 2011). These activities led to a vast number of published lead isotope data for ores and artefacts with major contributions by Gale, Stos-Gale, Pernicka and Begemann. Although several publications contain a few hundred lead isotope data, there is no complete database freely available. Only very few publications offer a substantial number (> > 1000) of lead isotope data in the form of electronic files or databases, which are freely accessible and easy to use (OXALID 2014, Scaife 1997, Vogl et al. 2018). Such data collections are an indispensable tool for assigning a specific mining site/region to a metal artefact. Even if these data are available, the assignment of the origin or provenance is not straight forward, even if the artefacts are made of pure lead, as is the case for curse tablets (Vogl et al. 2018). Often, two or more mining sites overlap in their lead isotopic composition, due to the geographical proximity of the sites or the geochemical similarity of the ore deposit (Bogdanov et al. 2013). The spread of the lead isotopic composition within one ore deposit can be rather large and, in most cases, it is not analysed in sufficient detail; especially, the ore which has been mined in antiquity is not available anymore. Difficult as well are lead isotopic compositions ranging outside those of any analysed mining site. Either, the mining site is yet unknown or the artefact was made by mixing different sources or by recycling of artefacts of different origin. These constraints not only apply to artefacts made from lead, but as well to those made from silver, which is extracted from lead ores. For other metal artefacts, especially multi-component alloys such as bronze, provenancing is even more difficult, because the lead derives from different constituents or additives and therefore source tracing becomes extremely complex.

Measurement uncertainties, which are not yet available for the majority of the lead isotope data, surely would help in assessing the borders of the mining sites in the lead isotope space and to decide whether or not a specimen can be attributed to a specific source. However, measurement uncertainties cannot solve the problem of overlapping mining sites or missing data in the lead isotope space. These questions can only be solved by taking advantage of other measurement result or other information. This ‘other information’ can be analytical data such as elemental mass fractions or element ratios, iconographic information or historical or archaeologic documents of any kind. Within this study, we applied this approach by combining information from different disciplines, namely isotope analysis (Romer and Born 2009, Vogl et al. 2013), elemental analysis (Riederer 2002) and archaeology (Born et al. 2014), to gain more insight into the provenance of the individual artefacts and whether or not they originate from the same source.

Samples and isotope ratio analysis

Samples

The well-known archaeological precious metal finds of Troy have been discovered by Heinrich Schliemann between 1871 and 1873. The most comprehensive repository, the so-called Treasure of Priam, was saved at the end of May in 1873 (Schliemann 1874). Hubert Schmidt (Schmidt numbers SCH) published this hoard as Treasure A (Schmidt 1902). All 19 or 20 treasure troves have been moved to Moscow and St. Petersburg by Russian troops in 1945 after World War II. Within a restitution between the former Soviet Union and the German Democratic Republic, eight out of 11 silver vessels from Treasure A have been brought to Leipzig in 1979 and after the German reunification, back to the Museum für Vor- und Frühgeschichte, Berlin in 1992. Three great vessels, two bowls, one fragmented beaker and another fragmented jar are now exhibited together with some gold tutuli to the public in the Neue Museum, Berlin (Fig. 1). The authenticity of the vessels has been assessed on the basis of historical photographs, chemical analysis and ancient manufacturing research (Born 1997, Koch and Born 2001, Born and Völling 2006, Born et al. 2014, Völling 2014). The archaeological as well as the finding context could be nearly reconstructed (Völling 2014). One of the results is the cultic use of the great silver vessels with and without pot handles. SCH 5973 (Treasure B) is identical in shape and measure with SCH 5871 and SCH 5872 belonging to Treasure A. During the restoration of the vessels of Treasure A, some milligrams of micro-particles accumulated consisting of metal and corrosion layer. Each of these samples was divided into two subsamples, thus yielding two similar sample sets. A list of the artefacts the samples originate from is given in Table 1.

Fig. 1
figure 1

Photography of the Trojan silver artefacts SCH5868 (small bowl), SCH 5871, SCH 5872 and SCH 5873 (vessels from left to right); photographer: Christa Begal, copyright MVF/SMBK-PK

Full size image
Table 1 Inventory numbers, descriptions and lead isotope ratios (Vogl et al. 2013) of the Trojan silver artefacts, the so-called “Priam’s Treasure”, belonging to the treasure trove A excavated by H. Schlieman; expanded uncertainties U (k = 2) are given in parenthesis and applying to the last two digits
Full size table

Isotope ratio analysis

The samples in our sample set existed in the form of one or more micro-particles derived from the restoration process as mentioned above. All samples were dissolved in PFA beakers using a mixture of concentrated nitric acid and ultrapure water. These samples underwent a lead-matrix separation as described by Vogl et al. (2013) using the Pb·Spec™ resin (Triskem, Bruz, FR). The resulting sample solutions were used to prepare sample filaments, which then were used to measure the lead ion currents on a thermal ionisation mass spectrometer (Sector 54, Micromass, Manchester, UK) in the static multi-collection mode using the automatic measurement mode. The so-determined ion currents at m/z 204, 206, 207 and 208 were calibrated using the primary isotope reference material NIST SRM 981. The achieved standard deviation between individual filaments of the same sample typically is 0.05% or better for the ion current ratio. The relative expanded uncertainty (k = 2) was typically between 0.07 and 0.09% for all isotope amount ratios. The sample preparation and the mass spectrometric measurements have been described in detail and published together with the so-determined lead isotope amount ratios (Vogl et al. 2013).

The second set of samples was analysed by Romer and Born (Romer and Born 2009). A comparison of both data sets is given in Vogl et al. (2013). The data published by Romer and Born do not feature measurement uncertainties, but based on the provided reproducibility a relative expanded uncertainty of ~ 0.1% can be assumed. Considering this, both data sets agree sufficiently well with each other, so that in the course of this interpretation only the data from Vogl et al. (2013) were considered.

Results and discussion

Lead isotope data

Lead has four isotopes and thus the isotopic composition of lead in general is spanned as space in a three-dimensional coordinate system with the coordinate axis 208Pb/204Pb, 207Pb/204Pb and 206Pb/204Pb. Three-dimensional graphs displayed on two-dimensional media can hardly be assessed, especially when applying large datasets. Therefore, the lead isotopic composition most commonly is visualised by the projection of the three-dimensional graph onto the two-dimensional 207Pb/204Pb-206Pb/204Pb and 208Pb/204Pb-206Pb/204Pb diagrams, which hereafter are referred to as 207-206-204 diagram and 208-206-204 diagram in order to improve the readability. Consequently, the lead isotopic composition of the Trojan silver artefacts and the potential mining sites are displayed in the 207-206-204 diagram and the 208-206-204 diagram and are discussed individually. In a second step, the information of both diagrams is combined, because an agreement of two specimen in the lead isotopic composition can only be achieved when there is an agreement in both, the 207-206-204 diagram and the 208-206-204 diagram.

The lead isotope ratios are displayed in Table 1 together with their associated expanded uncertainties. Most of the samples cover the same area in the isotope ratio diagrams, while the lead isotope data of sample SCH 5871 show a disagreement outside the stated uncertainties. Additionally, the samples SCH 5868 and SCH 5870 show a disagreement with the majority of the artefacts in the 207-206-204 diagram, which is outside the stated uncertainties. Therefore, it cannot be completely excluded that samples SCH 5868, SCH 5870 and SCH 5871 were derived from another source or from mixed sources. However, considering that the sample particles contain corrosion or alteration crusts, it is likely that the variation of the lead isotope data of different samples stemming from the same lead source exceeds the expanded uncertainty being calculated for the analytical procedure.

Information on the provenance of these artefacts can only be obtained by comparing the lead isotope data of the artefacts with lead isotope data on silver sources in the ancient world. In general, silver can be found as naturally occurring minerals in the form of solid silver (Ag), argentite (Ag2S), stromeyerite (CuAgS), pyrargyrite (Ag3SbS3) and proustite (Ag3AsS3), which, however, occur very rarely (Holleman and Wiberg 1995). Consequently, these forms did not play an important role for the silver exploitation in the ancient world. The most important source for silver mining is galena (PbS), which contains up to 1% (w/w) silver and which occurs frequently in both ancient Greece and Asia Minor. Secondary sources might be mixed sulphidic deposits. During antiquity, silver extraction was mainly carried out by reduction of galena to lead and subsequent silver separation by cupellation. Silver extracted by cupellation still contain lead with mass fractions in the range from 0.05% (w/w) to 2.5% (w/w) (Gale and Stos-Gale 1981). The elemental composition of the Trojan silver artefacts agrees well with this fact (see Table 2). This strongly supports the assumption that they were produced from Galena by cupellation. Therefore, lead isotope data of silver-bearing ores or minerals, which have been exploited in the Early Bronze Age (EBA) world, are required to trace back the origin of the used ores. Dating of mining sites, however, is difficult to achieve and even if it is successful it does not guarantee that ores and minerals have not been mined before and after this dating (Genter et al. 1980). Thus, no mining sites were excluded here due to unsuitable dating. Lead isotope data of mining sites were used from EBA as well as from antiquity. A very helpful compilation of published lead isotope data from ores and minerals in the Mediterranean area has been provided by Brett Scaife (Scaife 1997). This compilation was used here and has been expanded by additional data such that more than 2700 lead isotopic compositions of ores from Europe and the Mediterranean are covered (Vogl et al. 2018). This compilation was used to compare the lead isotopic composition of the Trojan silver artefacts with those of potential mining sites.

Table 2 Elemental mass fractions (Riederer 2002) of the Trojan silver artefacts excavated by H. Schlieman, belonging to the treasure trove A, the so-called “Priam’s Treasure”
Full size table

Figure 2 shows the 207-206-204 and the 208-206-204 diagrams with the lead isotope data of the Trojan silver artefacts obtained by Vogl et al. 2013. Lead isotope data of the ores of specific geographic regions have been enveloped by elliptical profiles labelled with capital letters A to G. In two cases, Euboea and Almeira, the lead isotope data are distributed over two ellipses, A and C (Euboea) and C and D (Almeira). Lead isotope data of Turkish mining sites are spread all over the displayed area in both diagrams. Therefore, the Turkish data are split. Turkish mining sites which agree well with the Trojan silver artefacts in their lead isotopic composition are labelled individually: The Eastern Troad, consisting of Altınoluk, Balya, Gümüşler, Hahla, Karaydin and Küserlik; the region around Arap Dağ and the Central Taurus region especially around the Bolkardağ Valley, the Niğde massif, and the Aladağ region. Other Turkish mining sites are summarised in the grey shaded area in both diagrams, in the 207-206-204 diagram this is the complete area below 207Pb/204Pb = 15.69 and in the 208-206-204 diagram this is the whole area below a line through origin with an angle of ≈ 30°. The pictograms for the specific Turkish mining sites are explained in the figure caption.

The most important locations in this context are displayed in a geographic map (Fig. 3).

Fig. 2
figure 2

207-206-204 and 208-206-204 diagrams with the lead isotope data of the Trojan silver artefacts and silver-bearing ores from the Mediterranean Area in EBA and antiquity; ♦ = Trojan silver artefacts and selected ore deposits = Taurus, = Arap Dağ, = Altınoluk and Küserlik and Hahlar, Δ = Balya and Gümüşler and Karaydin and Serҫeörenköy; Turkish sources excepting those mentioned above range over large parts in the diagrams which is visualised by the light grey shaded area in both diagrams

Full size image
Fig. 3
figure 3

Map of the Aegean Region with Greece and the western part of Turkey highlighting the locations discussed in the text; B Balya, G Gümüşler, Ka Karaydin, Kü + Ha Küserlik and Hahlar, A Altınoluk (basic map without locations from D-Maps 2017)

Full size image

In the 207-206-204 diagram, the lead isotope data of sample SCH 5871 plots in the same area as group A, showing similar lead isotopic composition as ores from Siphnos (GR), Murcia (ES) and Tuscany (IT). The isotopic compositions of the remaining samples plot in the same area as group C or as the space between A, B and C, showing similar lead isotopic compositions as ores from Antiparos, Euboea, Syros and Thasos (all GR) as well as Tuscany (IT). Additionally, these samples agree well with Turkish ores from Arap Dağ, Hahlar and the Central Taurus, less with ores from Balya, Gümüşler, Karaydin, Küserlik and Serҫeörenköy. The samples SCH 5868 and SCH 5970 show slightly higher 206Pb/204Pb values.

In the 208-206-204 diagram, the lead isotope data of sample SCH 5871 plots in the same area as group A, corresponding with ores from Siphnos (GR), Murcia (ES) and Tuscany (IT), whereas the data points of all other samples plot in the same area as group C and the overlapping area B-C, corresponding with ores from Antiparos, Syros, Thasos (all GR) and Almeira (ES). In the 208-206-204 diagram, the samples SCH 5868 and SCH 5870 fit well with the rest of the group. Except for SCH 5871, all samples correspond well with Turkish ores from Arap Dağ, Hahlar and the Central Taurus, less with ores from Altınoluk, Balya, Gümüşler, Karaydin and Küserlik.

As described above, the information obtained from the 207-206-204 and the 208-206-204 is now combined in order to obtain the agreement/disagreement of the isotopic compositions. Thus, sample SCH 5871 corresponds with ores from Siphnos (GR), Murcia (ES) and Tuscany (IT), while all other samples correspond with ores from Antiparos, Syros and Thasos (all GR) and with ores from Arap Dağ, Halilar and the Central Taurus (TR). Ore deposits at Balya and Serҫeörenköy, which have been discussed as potential sources by Romer and Born (2009), show slightly less agreement with the Trojan silver artefacts; the same accounts for ores from Gümüşler, Karaydin and Küserlik.

The samples SCH 5868 and SCH 5870 are slightly displaced to the other data points in the 207-206-204 diagram but fit well with the group in the 208-206-204 diagram. This might be due to contamination caused by the long underground storage affecting especially surface particles or it simply shows the isotopic spread of the ore deposit. Although a mixing scenario with minute amounts of a second source during any stage of the production process cannot be fully excluded. Following these arguments, these samples can still be assumed to fit with the remaining group. This is also assumed by Romer and Born (2009). Sample SCH 5871, however, falls apart from the group in both plots. A different source than for the other samples is possible, but as well a mixture of different sources.

A statistical evaluation of the lead isotope data for the different ores and minerals is not reasonable because the data have been derived from different sources. This means that different analytical procedures have been applied and thus the obtained results feature a different precision, a different accuracy and potentially a different remaining bias. Additionally, the location and the type of the ore could have been assessed with different accuracy. In most cases, the range of the lead isotopic composition of an ore deposit has not been investigated completely.

From a metrological point of view, measurement uncertainties should be used in any case to decide whether two or more samples agree in their lead isotopic composition or not. After these issues have been clarified, the mapping of provenances can be started. In this case, however, measurement uncertainties are only available for the lead isotopic composition of the Trojan silver artefacts, but not for the lead isotopic compositions of the ores/minerals. Furthermore, the spread of the ore deposits concerning their lead isotopic composition is not completely unravelled. Therefore, all locations overlapping with the lead isotopic composition of the Trojan silver artefacts within the stated uncertainties are equally probable. This accounts for Murcia (ES), Tuscany (IT), Siphnos, Antiparos, Syros, Thasos (all GR) and the Turkish sites Arap Dağ, the Central Taurus, and the Eastern Troad (Balya, Hahlar, Serҫeörenköy, Gümüşler, Karaydin, Küserlik, Altınoluk).

Elemental composition

Further reduction of the potential mining sites could only be applied by using additional information on the samples and the ore deposits. In analytical chemistry, one important parameter is the elemental composition of the ores and artefacts. In this context, however, its use is rather limited because the most important elements such as copper, zinc, arsenic and antimony are heavily depleted during cupellation. Thus, only two quantities remain, which are not affected by mining, cupellation and other metallurgical processes and which can be used for excluding potential mining sites: first the silver mass fraction in the ore and second the silver-gold-elemental-ratio (Au/Ag-ratio) in the ore and in the artefacts.

Gale and Stos-Gale (1981) as well as Wagner et al. (1986) showed that the lower limit at which cupellation was economic and was practiced in EBA ranges around 800 mg silver per kilogramme bullion lead corresponding to approximately 700 mg silver per kilogramme galena. Based on this, ore deposits with silver mass fractions considerably below this threshold (700 mg/kg) very unlikely are the sources for silver production by cupellation in EBA. According to Gale and Stos-Gale (1981), ores from Antiparos, Siphnos and Syros “prove to have significant silver contents” (Gale and Stos-Gale 1981) and are therefore still potential sources. The same applies to ores from Thasos which offer silver mass fractions up to 1000 mg/kg (Pernicka and Wagner 1985). Ores from Gümüşler and Altınoluk can be excluded as potential sources for the Trojan artefacts because they show silver mass fractions below 260 and 400 mg/kg, respectively (Wagner et al. 1984, 1986). Ores from Hahlar, which are epithermal-type deposits in different settings (multi-phase breccia zones, quartz veins, stockwork veinlets) show silver mass fractions around 650 mg/kg and below (Wagner et al. 1984, 1986) and therefore very unlikely are the source for the Trojan silver but cannot be excluded completely as provenance. Ores from Balya, Serҫeörenköy and the Taurus show high silver mass fractions, even above 1000 mg/kg (Wagner et al. 1984, 1986, Yener et al. 1991).

In Balya, hydrothermal stratabound or hydrothermal-metasomatic (Skarn-type) lead-zinc deposits with copper-gold mineralisation can be found (Wagner et al. 1984). In the Central Taurus, especially the Bolkardağ Valley, lead-zinc-copper-tin ores occur, but also rich deposits of silver ores: pyrargyrite and argentite. Bolkardağ Eskişehir seems to be a placer ore deposit as well with gold and silver enrichments in small placer reservoirs, 1–100 mg/kg gold and up to 6000 mg/kg silver (Hauptmann 2008). For ores from Arap Dağ, Karaydin and Küserlik, no silver mass fractions have been reported, for which reason these ores cannot be excluded. This leaves Murcia (ES), Tuscany (IT), Siphnos, Antiparos, Syros, Thasos (all GR) and the Turkish sites Arap Dağ, the Taurus and the Eastern Troad (Balya, Serҫeörenköy, Karaydin, Küserlik) as potential silver sources.

The Au/Ag-ratio is another important parameter because it does not change during cupellation and therefore the Au/Ag-ratio in the artefact should reflect the Au/Ag-ratio in the corresponding ore (Gale and Stos-Gale 1981). The Au/Ag ratios of potential silver sources (ore deposits) with silver mass fractions above 100 mg/kg are visualised in Fig. 4 and are compared with the Au/Ag ratios of the Trojan silver artefacts. Ore deposits whose Au/Ag ratios overlap with those of the Trojan silver artefacts (red colour) are coloured green; ore deposits whose Au/Ag ratios do not overlap are coloured blue. To avoid any confusion in black-and-white prints, the context is explained in detail in the following text.

Fig. 4
figure 4

Au/Ag ratios of silver-bearing ores from the Mediterranean area with silver mass fractions above 100 mg/kg compared with Au/Ag ratios of the Trojan silver artefacts

Full size image

The Trojan silver artefacts show Au/Ag-ratios ranging from 3.2·10−4 to 2.0·10−3 (Table 2). Only one of two subsamples of SCH5873 shows a value of < 1.1·10−4, whereas the other subsample shows a value of 3.2·10−4. The Au/Ag-ratios for galena ores from Antiparos are approximately 2.8·10−4 and for galena ores from Syros they range between 1.4·10−6 and 1.4·10−4, whereas galena ores from Siphnos show Au/Ag-ratios from 8.0·10−6 up to 8.2·10−3 (Gale and Stos-Gale 1981). The Au/Ag ratio for galena from Thasos range between 10−6 and 10−4 (Pernicka and Wagner 1985). Due to these data, ores from Thasos can be excluded with high probability. The Au/Ag-ratios of ores from Syros and Antiparos are at or slightly below the lower edge of the Au/Ag-ratios of the Trojan silver artefacts. Based on these data, it is unlikely that ores from Syros and Antiparos are the source, but they cannot be excluded completely. The Au/Ag-ratios of galena from Siphnos cover those of the Trojan silver artefacts. Additionally, ceramic sherds from mining sites at Siphnos have been dated by thermoluminescence (Pernicka et al. 1985) and 14C techniques to the first half of the third millennium BC (Gale and Stos-Gale 1981) supporting galena mining and silver exploitation during this time. This still keeps ores from Siphnos as a potential silver source. The ores from Gümüşler show Au/Ag-ratios of less than 3.1·10−4, whereas the ores from Serҫeörenköy range between 1.7·10−3 and 2.2·10−2 (Wagner et al. 1984, 1986). Both data do hardly agree with the values found for the Trojan artefacts. Au/Ag-ratios for several ore samples from Balya can be calculated from data published by Wagner et al. (1984) and result in values ranging from 6.7·10−5 to 1.4·10−3. Two out of four Au/Ag-ratios agree with those calculated for the Trojan artefacts and two Au/Ag-ratios are below. Due to this, Balya still has to be regarded as potential source. The Au/Ag- ratios obtained for the ores from Altınoluk (5.2·10−4 to 1.0·10−3) and Hahlar (3.9·10−4 to 4.2·10−4) agree with those from the Trojan artefacts, the same accounts for Central Taurus ores (9.0·10−4 to 5.2·10−2). For ores from Arap Dağ, Karaydin and Küserlik, no data are available.

According to the Au/Ag ratios, galena from Siphnos, Balya, Altınoluk, Hahlar and the Central Taurus with high probability are potential silver sources; ores from Syros, Antiparos, Serҫeörenköy and Gümüşler show low probability. Arap Dağ, Karaydin, Küserlik, Murcia and Tuscany cannot be assessed due to a lack of data.

Yener et al. (1991) analysed seven specimens of silver jewellery being excavated by Schliemann and now treasured in the Istanbul Archaeological Museum. They assigned four silver artefacts with high probability to Taurus ores and the remaining three jewelleries more probably to Eastern Troad ores. The closest agreement of the lead isotope data has been obtained for the Trojan silver artefacts and the galena ores from Taurus, which agrees well with the findings of Yener et al. (1991).

Summarising the findings in this chapter, we can reduce the possible silver sources with a high probability to Balya, the Central Taurus and Siphnos. Silver sources with low probability are Antiparos, Syros, Serҫeörenköy and Hahlar. Due to a lack of data, Arap Dağ, Karaydin, Küserlik, Murcia and Tuscany still have to be considered as potential silver sources.

Archaeological information

Information on the cultures of the EBA and their trade relations can also help to reduce the number of potential silver sources. According to A. F. Harding (2000), it is most likely that open sea crossings have been avoided due to the used ship types and the lack of navigation. Therefore, coast hopping was the type of movement. Also, Maran (1998) excludes navigation by Cycladic longboats before 2200 BC.

EBA cultures in Italy and Spain date from 2550 and 2300 BC onwards and no silver findings are reported for Italy and Spain around this time, while the Trojan silver artefacts are dated to 2300 BC (Bachhuber 2009). Consequently, Murcia and Tuscany are very unlikely the geographic origin of the silver because the ore deposits in Western Anatolia could be reached by the caravan route.

Mesopotamian written records provide some idea on the organisation and production capacity of metal workshops in centralised economies (Reiter 1997, Günbattı 1998, Yakar 2008) in the Late EBA around 2300 BC (Bachhuber 2009, Saszı and Korfmann 2000). The use of metals as raw material and finished product (Archi 1988, Pomponio and Xella 1997, Pettinato 1977) points to the fact that production and trade were under control of a central authority. Although written records have not yet been discovered in Troy, all treasure troves are the best demonstration for the existence of an authority within a local workshop where even prestige goods and symbols of power could have been produced in metal (Völling et al. 2012). Efe (2002) showed convincingly the interaction between cultural/political entities and metal working in Western Anatolia during the Late EBA II and EBA III periods: caravan routes from the Taurus region lead across the plains of Afyon, the valley of upper Sakarya, Eskişehir and the plains of Iznik and Inegöl; many EBA III sites were situated on the way into the Troad (Völling 2014). In the Central Taurus region, another important mine is located providing tin in the EBA: Kestel/Göltepe (Earl and Özbal 1996). The exploitation of tin from the Kestel mine is controversially discussed, but the mineralisation also contains gold as Hauptmann (2008) noted. Regardless of the tin exploitation, the different metal deposits at Kestel/Göltepe provided good reasons for trading between Syro-Cilicia and the north Aegean along EBA III sites (Efe 2002). The south-western line headed into the Izmir region, in which the Arap Dağ ore deposit is located nearby EBA I-III sites with intensive metal production. The Izmir region provides a natural bridging function between the Anatolian mainland and the western Aegean (Şahoğlu 2009). Liman Tepe, a harbour town that was continuously inhabited from the Chalcolithic to the end of the Late Bronze Age, played an important role in the Aegean and Anatolian trade network during the later part of EBA II and earlier EBA III. Bakla Tepe reflects extensive metallurgical activities from the Late Chalcolithic to the Bronze Age. Grave furnishings include rich metal objects of silver and gold as well as a rich variety of pottery at the end of EBA II. Similar to Late Troy II, there were contemporaneous settlements and metal working activities in Late EBA II-III near Arap Dağ (Şahoğlu 2005; Şahoğlu 2009).

Although the Eastern Troad should be the closest ore deposit to Troy for gold (Astyra) and silver (Balya), Taurus and Arap Dağ are more likely the source areas for Trojan silver. Genter et al. (1980) stated Balya (TG 18) as a deposit mined in Hellenistic-Roman times. Wagner et al. (1984) described Balya as ‘prehistoric’ without further specification and left unclear whether it means in the EBA or not; furthermore, they stated ‘On the base of these data it appears that the analysed EBA metal artefacts of the Troad were predominantly produced from ores of other regions’ (Wagner et al. 1984). The period of mining in Gümüşler is unknown, Karaydin is supposed to be Roman and Küserlik was not mentioned at all. Wagner et al. (1984) as well as Wagner and Öztunalı (2000) emphasised that Balya Maden is one of Turkey’s major lead-silver deposits and was the traditional mining centre of Northwest-Anatolia between 1880 and 1935, but “’or prehistoric periods the situation might have been quite different’ (Wagner and Öztunalı 2000). Romer and Born (2009) supposed also Balya, Serҫeörenköy or Thasos as possible silver sources, but they do not provide any details of lead-silver mining in the EBA. Cilicia is well documented in the EBA by Mesopotamian written records. Especially, the Taurus region was well-known for mining of tin—a high coveted metal—silver, copper and perhaps gold, which could be traded from the Central Taurus to Troy. Basically, the corpus of archaeological finds in Tarsus and Mersin (Mellink 1986) synchronise with Troy and is an inherent part within the Aegean-Anatolian distribution in EBA II-III (Çalış-Sazcı 2006).

Syros and Antiparos are at least potential prehistoric Cycladic lead-silver sources for Trojan artefacts. The mine of Agios Georgios Helen with potential ‘prehistoric’ mining in Antiparos is listed (TG 68) by Genter et al. (1980) without mentioning Syros, where lead-silver deposits are available in the southern part (Maran 1998, Ivanova 2008). Gale and Stos-Gale (1981) found EBA pottery, dated as Early Helladic (EH) II, inside and outside the mines of Rosos and Komito in southern Syros, but the sherds have not yet been published. In the fortified settlement Kastri in the north-eastern part of Syros, Bronze artefacts, weapons, tools and slag were found in room 11. Tools for metal production were niched in the wall (Maran 1998), i.e. the same location as for Treasure A in Troy that was niched in the wall of gate FM (Korfmann 2001). The main criterion of a postulated synchronism between Kastri and Troy is the identical lead isotopic composition in bronze artefacts found in Kastri as well as in Troy (Maran 1998). The copper is alloyed with a high percentage of tin (≥ 2% w/w), which is unknown before that time in both Central and the Eastern Aegean, but was only known in the Poliochni phase yellow and Late Troy II (‘treasure horizon’). The lead isotopic compositions of all bronze objects do not coincide with those of Lavrion, Siphnos and Kythnos. The artefacts from Kastri and Troy have to be derived from ore deposits of earlier exploitation time than those in the Aegean and Western Anatolia. Maran (1998) used the term ‘foreign metal’ and suggested that the ‘exotic’ lead isotope data are caused by lead incorporated in tin or copper or both, a phenomenon which occurs not only for Kastri and Troy in Late EBA, but also for other Aegean areas.

The shapes and measures of SCH 5871, SCH 5872 and SCH 5973 are impressively identical and indicate that they were with certainty produced in the same workshop, presumably in Troy with the attestation of an extensive metal production in Anatolia (Müller-Karpe 1994).

Conclusions

Lead isotope data of the Trojan silver artefacts (Vogl et al. 2013) together with elemental data and archaeological data were used to obtain information on the ore provenance of the artefacts. The here presented results are in contrast to the findings published by Romer and Born (2009), but partially agree with their findings published in 2014 (Born et al. 2014). The most important difference of the here presented work is the consideration of the Central Taurus region and its assessment as most probable silver source for the Trojan silver artefacts, which will be summarised in the following.

In EBA, silver has been obtained mainly by cupellation of lead ores such as galena. The remaining lead impurities in the silver can be used to obtain information on the origin of the used ore. Comparing the lead isotope ratios of the Trojan silver artefacts with those of silver containing ores in the Mediterranean Area from the literature shows an overlap for mining sites from Siphnos, for which the earliest proves for metal production in the Aegean exist (GR), Murcia (ES) and Tuscany (IT) for sample SCH5871. The lead isotope ratios of all other samples overlap with those of mining sites from the Taurus, Arap Dağ, the Eastern Troad, Altınoluk (all TR) and Antiparos, Syros, Thasos (all GR) starting with the closest agreement. A more detailed allocation is not possible because the lead isotopic composition of the different locations overlap partially with each other and because the lead isotopic composition of the Trojan silver artefacts cover several locations with their measurement uncertainty.

A further exclusion of mining sites is only possible by using additional information such as the elemental composition of artefacts and ores and archaeological data. The silver mass fraction is an important parameter because galena with silver below 700 mg/kg was very unlikely to be extracted in EBA. Due to this, Gümüşler and Altınoluk can be excluded as ore deposit and Hahlar is less probable. A second important parameter, which is based on the elemental composition, is the Au/Ag-ratio, which stays constant in the cupellation process and thus should provide similar values in the ore as well as in the artefacts. Based on this, Thasos could be excluded as a potential source. Gümüşler, Serҫeörenköy, Syros and Antiparos ranging at or slightly below the Au/Ag ratios of the Trojan silver artefacts turned out to be a less probable silver source.

Archaeological data help to further reduce the possible sources. Open sea crossing is not known for EBA and is being excluded by Harding (2000) and Maran (1998), which makes Murcia and Tuscany a highly implausible silver source. Further archaeological information on culture and trade in the EBA Mediterranean Area favour Arap Dağ and the Taurus region as potential sources.

Summarising all information discussed above, the sources for the Trojan silver artefacts can be listed as follows, starting with the mining site of the highest probability: Central Taurus > Arap Dağ > Eastern Troad. In other words, the Central Taurus is the most probable silver source for the Trojan silver artefacts, while Arap Dağ and the Eastern Troad (especially Balya, Karaydin and Küserlik) cannot be completely excluded. The only exception is artefact SCH5871, which to a higher degree agrees with mining sites from Siphnos.