Introduction

1The richness of gold in the Caucasus region is legendary. Since the second half of the 3rd millennium, extensive use of gold is evident through finds in kurgans of the Martkopi-Bedeni stage of the late Kura-Araxes period. A later highlight is the prehistoric gold work of the Trialeti culture in the first half of the 2nd millennium BC demonstrating filigree and granulation techniques. Small beads of the Leilatepe culture from Soyuq Bulaq in Azerbaijan dating to the 4th millennium BC are unique and exceptional examples from Chalcolithic times. A major question arises as to the provenance of the raw materials. Were the rich gold occurrences of the Lesser Caucasus and the southern Greater Caucasus already exploited for local supply at that time, or were gold or finished artefacts imported from other regions? Field work and analytical projects were conducted in recent years in order to characterize the geochemical fingerprint of natural and artefact gold in Transcaucasia (Hauptmann, Klein 2009; Hauptmann et al. 2010; Jansen, Stöllner, Courcier 2014), and for Armenian gold occurrences (Wolf et al. 2011, 2013; Wolf, Kunze 2014).

2The analytical investigation of archaeological gold artefacts is a complicated task given their great cultural value and careful curation. Provenance studies of gold artefacts aiming to trace their raw material sources through chemical analyses are an intensively debated topic in archaeometallurgy (Guerra, Calligaro 2003; Watling et al. 1994; Tamas, Baron, Cauuet 2009; Pernicka 2014). Famous investigations to characterize ancient gold artefacts all over Europe and beyond were performed by Hartmann (1970, 1982) who analyzed large series of around 5,000 objects. Different types of gold were classified according to five major and minor elements (silver, copper, tin, nickel, platinum), and the most suitable tracers to identify the use of alluvial gold for artefacts were identified in the trace elements tin and platinum. But as these analyses were solely focused on artefacts themselves and excluded a characterization of natural gold occurrences, no direct links between the gold types and a provenance of the gold artefacts could be established. Since then, more sensitive and non-destructive micro-invasive techniques were introduced to analytical chemistry, including particle induced X‑ray emission (PIXE) and laser ablation inductively coupled plasma mass spectrometry (LA‑ICP‑MS). In recent years artefacts have been increasingly investigated with trace element measurements by LA‑ICP‑MS, and were also compared with potential natural gold resources (Schmiderer 2008; Hauptmann, Klein 2009; Kovacs et al. 2009; Schlosser et al. 2009; Ehser, Borg, Pernicka 2011; Leusch, Brauns, Pernicka 2016). Lead isotope analysis applied to gold artefacts (Bendall et al. 2009; Hauptmann, Klein 2009; Standish et al. 2013) is rather rare because of analytical issues since typically low Pb concentrations and potential Hg contamination of the gold prevent accurate and precise determination of ²⁰⁴Pb. The investigation of inclusions of natural alloys made of platinum group elements (PGE) for either their chemical composition (Young 1972; Ogden 1976; Meeks, Tite 1980; Zwicker 1998; Miniaci et al. 2013; Troalen, Tate, Guerra 2014; Lemasson et al. 2015; Zaykov et al. 2017) or osmium isotope ratios (Junk, Pernicka 2003; Jansen et al. 2016) is an alternate approach but limited to artefacts where those are present.

3In the current study, results from a geochemical characterization of gold artefacts from Georgia and Azerbaijan, natural gold from the prehistoric mine in Sakdrisi (Georgia), and nearby placer deposits are presented. Ore samples from the Kura-Araxes galleries of Sakdrisi were microscopically investigated and analyzed for their trace elements by sector field inductively coupled plasma mass spectrometry (SF‑ICP‑MS) in order to characterize the ore mineralization. Concentrates of natural gold as well as artifact gold were analyzed by electron probe micro analysis (EPMA) and portable X‑ray fluorescence (pXRF) for their alloy composition, laser ablation inductively coupled plasma mass spectrometry (LA‑ICP‑MS) for their trace element contents. This groundwork will contribute to future provenance studies in a supra-regional context.

The natural gold from the Sakdrisi-Bolnisi district

4The gold-rich mineralization of Sakdrisi is characterized by subvertical stockwork-style lodes in a volcanic tuff of the Cretaceous (Gugushvili et al. 2002; Hauptmann et al. 2010). Samples from the ore lodes in the prehistoric galleries were taken in the last few years for detailed laboratory investigations (Jansen, Stöllner, Courcier 2014; Stöllner 2014). Five chunks of ore were prepared as polished sections and microscopically analyzed for their texture. The gangue mainly consists of quartz, hematite and barite. At the microscopic scale, the gold itself is visible as free-gold particles in limonitic boxwork structures of weathered sulfide minerals (fig. 1). Residues of hypogene sulfidic ore minerals like chalcopyrite and pyrite are present, e.g. as inclusions in quartz which protected the sulfides against oxidation. Supergene minerals like enriched copper sulfides, lead carbonate, silver bromide and bismuth oxide occasionally are visible in the boxwork structures as well.

Fig. 1a – Particles of free gold in boxwork structures in ores from Sakdrisi, boxwork of a dissolved sulfide mineral surrounded by quartz with bright free-gold particles, optical microscopy picture (XPL) (M. Jansen).

Fig. 1b – Particles of free gold in boxwork structures in ores from Sakdrisi, bright free-gold particles in limonitic boxwork (grey), surrounded by quartz (dark grey) and porosity (black), scanning electron microscopy image (BSD) (M. Jansen).

5Bulk analyses of 17 ore samples directly taken from the ore lode in the ancient mine of Sakdrisi were conducted to investigate the trace elements of the ores with SF-ICP-MS (Jansen, Stöllner, Courcier 2014). High gold concentrations up to 880 ppm were observed in these samples (tab. 1). Next to gold and silver, these also contain copper and lead in trace element concentrations (fig. 2). For all samples, the trace elements arsenic, zinc, tin and bismuth were detected in concentrations below 100 ppm and cobalt, nickel, selenium, antimony and tellurium below 10 ppm. The PGE ruthenium, rhodium, palladium, osmium and iridium were found below the detection limit of 0.1‑0.2 ppm.

Fig. 2 – Histograms of trace elements of the bulk analysis of 17 ore specimens from the prehistoric galleries of Sakdrisi; the contents of lead and copper are too low for a prehistoric exploitation of these metals; measured by SF‑ICP‑MS (M. Jansen).

6The gold and silver traces of the ore originate from the free-gold particles (see below), but additional silver could be traced back to secondary phases like silver bromide and sulfide. The copper content is due to minor amounts of copper minerals, and lead traces are due to minor amounts of lead minerals, both observed microscopically (see above). In conclusion, the copper, lead and silver contents of the ore are too low for a prehistoric extraction of these metals, but gold could have been enriched through pounding, grinding and panning of the ore.

7In order to characterize the geochemistry of the free-gold particles themselves, more than 150 kg of ore and 600 kg of backfilling of the ancient galleries from Sakdrisi were crushed down and panned to concentrate the heavy mineral fraction including gold (fig. 3a). Six samples of those concentrates from ground ore, from backfilling of different levels of the galleries, and from a fire setting experiment (Stöllner et al. 2012) contained enough gold for further investigation with the micro analytical methods EPMA and LA-ICP-MS, requiring mounted and polished sections. In addition, a small sample of a gold-rich heavy mineral concentrate from Sakdrisi was melted in a porcelain crucible in normal atmosphere within a muffle kiln to investigate the alteration of the composition due to metallurgical processing (fig. 3b). The free-gold particles formed a prill due to the surface tension of the liquid gold. Other heavy minerals stayed nearly unaltered as friable conglomerate at the bottom of the crucible. The inclusion-free prill was investigated in the same way as the free-gold particles.

Fig. 3a – Digital micrographs; gold-rich heavy mineral concentrate from Sakdrisi produced through ore crushing, pounding, grinding and panning (M. Jansen).

Fig. 3b – Digital micrographs; the same concentrate was melted in a porcelain crucible producing a prill next to other barely altered heavy minerals (M. Jansen).

8Next to gold, only silver was detected as main element in the free-gold particles from Sakdrisi (fig. 4). In total 258 single free-gold particles from ore and backfilling material show a composition of around 6±3.5 wt% (mean and standard deviation). The melted gold prill shows a silver content in the higher range of the free-gold particles. Some of the prills from the fire assay experiments (described by Hauptmann 2011; results published by Stöllner 2014) show higher silver contents up to 16 wt% for the bulk of several kg of backfilling material. This could be due to the secondary silver phases detected in the microscopic studies whose silver is collected by the lead in the regulus as well. A maximal range of 2.5 to 16 wt% Ag could be expected for gold exploited from the Sakdrisi galleries, even though leaded smelting to extract gold from the ores, theoretically resulting in a similar silver variation observed for the fire assay experiments, was not applied in the Bronze Age. A range of 2.5 to 9.5 wt% of silver in the gold resulting from the analysis of free-gold particles themselves and the melting experiment is more reasonable for gold extracted in prehistoric times. The latter composition results in a yellowish color of the exploited gold from Sakdrisi. The high silver/gold ratios for the individual ore samples published by Stöllner (2016) do not reflect larger amounts of ore needed for gold exploitation and should not be misinterpreted as bulk analysis of the final product.

Fig. 4 – Diagram showing silver contents (normalized with Au) in different gold samples from the prehistoric mine of Sakdrisi; 258 particles of free gold resulted in a mean of about 6 wt% Ag with a standard deviation about 3.5 wt%; the melted gold shows a value of almost 9.5 wt% which is in the upper range of the single free-gold particles; 21 prills from the fire assay experiments of backfilling show contents up to 16 wt% of silver; measurements with electron probe micro analysis (EPMA) and scanning electron microscopy with energy dispersive X-ray spectroscopy (SEM), showing comparable results by analyzing the prill melted from a heavy mineral concentrate with both methods (M. Jansen).

9Analyses of gold panned from different nearby placer deposits (described by Hauptmann et al. 2010) result in a tendency of higher silver concentrations than the gold from the prehistoric mine of Sakdrisi. In contrast, alluvial gold from two sites in Svaneti shows lower silver contents than observed for the Sakdrisi mine. The gold occurrences from the Sakdrisi-Bolnisi district overlap with occurrences investigated by Wolf et al. (2011, 2013) and Wolf and Kunze (2014). The data sets from the Lesser Caucasus are compiled in fig. 5. The different silver/gold ratios of the various occurrences result in different colors of the exploited gold from these deposits. In addition, one may conclude, the primary occurrences of the alluvial gold in the Sakdrisi-Bolnisi district must have had a different source than that observed in the Sakdrisi galleries due to a tendency of depletion in silver for alluvial gold compared to their primary sources. Therefore, primary sources have a higher silver content than their secondary placer deposits (Dill 2010).

Fig. 5 – Diagram showing silver and copper contents of free-gold particles from the prehistoric gold mine of Sakdrisi in comparison to alluvial gold from the Lesser Caucasus (data from Armenian deposits by Wolf, Kunze 2014); note the logarithmic scale; a tendency of lower silver values for the primary gold from the Sakdrisi galleries in comparison to nearby placers in Sakdrisi-Bolnisi is visible; the results of the Armenian deposits overlap with the composition of the Sakdrisi-Bolnisi district (M. Jansen).

10The trace element concentrations of natural gold from different sites were found to be very low in general (fig. 6). Considering the geoscientific definition of trace elements that these do not form their own phases and are to be found in concentrations lower than 1,000 ppm, a majority of LA‑ICP‑MS measurements resulted in mixing analyses due to mineral inclusions and adhesions of the gold particles. This can be clarified by a time resolved evaluation of the LA‑ICP‑MS measurement showing increases of intensities if inclusions are ablated. Mixing analyses were excluded from the data set since they do not reflect the composition of the gold itself. This is also necessary if one compares data sets from the literature including that for Armenian gold occurrences by Wolf and Kunze (2014) where some of the analyses can be identified as mixing analysis.

Fig. 6a – Trace element pattern of natural gold occurrences of the Sakdrisi-Bolnisi mining district; trace element pattern of free-gold particles (mean of 26 particles) from the prehistoric mine of Sakdrisi and the prill (mean of multiple measurements) from a melting experiment; the gold is essentially pure without other trace elements than copper and iron; light changes due to the melting in oxidizing atmosphere were detected; multiple trace elements below detection limit (M. Jansen).

Fig. 6b – Trace element pattern of natural gold occurrences of the Sakdrisi-Bolnisi mining district; trace element pattern of various alluvial gold occurrences (mean of multiple particles) in the Sakdrisi-Bolnisi district; they do not show significant differences inhibiting provenance studies on the base of trace element patterns; multiple trace elements below detection limit; measured by LA‑ICP‑MS (M. Jansen).

11The gold from different sources is essentially pure. Besides Cu and Fe, which were detected in concentrations between 10 to 1,000 ppm, no other trace element exceeded 10 ppm, and for most of the samples these were found to be below the detection limits (fig. 6). Thus, the gold from different occurrences is not distinguishable using trace elements.

12By comparing the prill from the melting experiment with the gold-rich concentrate from Sakdrisi, minor changes in the chemical composition are observed (fig. 6a). Copper was slightly enriched in the prill. This could be traced back to copper sulfides identified in the heavy mineral concentrate. Significant changes can be expected by melting gold in a reducing atmosphere owing to the reduction of metal oxides originating from the different heavy minerals in the concentrate. An experiment melting the same gold-rich heavy mineral concentrate from Sakdrisi in reducing atmosphere will be analyzed in the future. Major changes in the chemical composition might not to be expected for the primary gold from Sakdrisi because of the compound of the heavy minerals mainly consisting of barite, hematite and quartz. Minor changes could be the enrichment of copper, lead and bismuth due to the small amounts of supergene phases and relics of hypogene minerals observed during the microscopic examination (see above).

Bronze Age gold artefacts from Georgia

13In 2010, prehistoric gold artefacts from Georgia were investigated with a handheld portable X-ray fluorescence (HH‑pXRF) device in the Georgian National Museum in Tbilisi. A decrease of the silver content was detected for artefacts from the 1st millennium BC in comparison to Bronze Age artefacts. This could be related to the introduction of the gold-silver parting process at this time (Gambashidze, Hauptmann, Stöllner 2011), by the use of different gold occurrences which are poor in silver like alluvial gold from Svaneti, or by an additional alloying of gold with silver in the production of the earlier Bronze Age artefacts. In addition to the measurements with pXRF, small samples (usually less than 10 mg) were taken from dozens of fragmented gold artefacts for further investigations in the laboratories of the German Mining Museum in Bochum and the Institute for Geosciences of the Goethe University in Frankfurt am Main, and samples from earlier studies (Hauptmann, Klein 2009; Hauptmann et al. 2010) were reanalyzed to ensure comparability, leading to a total of 70 Bronze Age artefacts.

Alloy components

14Besides some outliers, the gold artefacts are characterized by a silver content between 10‑40 wt% and also contain copper between 1‑6 wt% as main alloy components (fig. 7), which is a typical composition for archaeological gold artefacts. The silver contents of the artefacts are within the variation bracket that was also found for natural gold in the Lesser Caucasus. The composition of gold and silver therefore could reflect the use of local gold sources. To investigate the possibility of an intentional alloying of gold with silver to produce white gold artefacts, one could consider the relation between the silver and lead within the gold. Silver was exploited from silver-bearing lead ores like galena and cerussite in the Bronze Age. Thus, an increase of lead for artefacts with higher silver contents could be expected if silver was added to produce an intentional alloy (Hauptmann et al. 2010).

Fig. 7 – Diagram showing the main alloy components copper and silver content of 70 Bronze Age gold artefacts from Georgia in comparison to natural gold from the Lesser Caucasus (Wolf, Kunze 2014; own data); the artefacts are characterized by a silver content up to 40 wt% in the range of natural gold, and copper typically was found up to 6 wt% for artefacts but not significantly for natural gold (M. Jansen).

15In fig. 8, the silver and lead concentrations of the artefacts are shown for the Bronze Age gold artefacts from Georgia and, for comparison, 3rd millennium BC gold artefacts from Ur. Lines of the least-squares regression are presented for both data sets. For the Georgian artefacts, this line increases with higher silver contents. In other words, there is a relation between silver and the trace element lead for artefacts from Georgia. We cannot exclude that at least some of the Bronze Age artefacts reflect an intentional alloy obtained by adding lead-bearing silver, suggested by the linear regression. The contrary is true for the artefacts from Ur. The lead concentration does not change with an increasing silver content. The lead in the gold artefacts from Ur might originate from lead-rich phases in the gold-bearing heavy mineral concentrate, which apparently were present at the same level for concentrates with different gold/silver ratios.

Fig. 8a‑b – Diagram showing the silver content and the trace element concentrations for the Bronze Age artefacts from Georgia (upper diagram, 8a) and for gold artefacts from Ur for comparison (lower diagram, 8b); lines of the least-squares regression are given for both data sets; for Georgia, this linear regression is increasing (coefficient of determination r²=0.17), while for Ur this is not the case (r²=1.68E‑04); there is a relation between silver and lead for the artefacts from Georgia, but there is no relation for artefacts from Ur; measured by EPMA (Ag) and LA‑ICP‑MS (Pb) (M. Jansen).

16It is evident that gold artefacts contain copper in an order of magnitude which is not observed for natural gold (fig. 7). This composition could be an intentional alloy, but might be a contamination due to co-melting of gold and copper minerals originating from the heavy mineral concentrate produced by panning (Hauptmann, Klein 2009; Hauptmann et al. 2010).

17The silver contents should be identical for both gold artefacts and their raw material sources if specific gold occurrences are exploited to create the artefacts. This is true if no metallurgical operations were conducted after exploitation to change the silver content. The relation of silver and lead for artefacts from the 2nd millennium and some from the 3rd millennium was discussed above: thus, silver could have been added to create an alloy higher in silver for some of the artefacts. However, the parting process was not introduced before the first millennia BC. Gold exploited from Sakdrisi and processed to form artefacts should keep its silver/gold ratio, if no silver was added intentionally. Since only a few artefacts of the 3rd millennium BC show comparable silver contents (fig. 9), the gold of Sakdrisi could be excluded for most of the artefacts, if unalloyed gold from a single source was used.

Fig. 9 – Diagram showing the silver content (normalized with gold) of 3rd millennium BC artefacts from Georgia in comparison to free-gold particles and prills from the fire assay; the variation of silver in gold from Sakdrisi is marked by dashed lines; only a few artefacts from Zitsmatiani, Ananauri and Znori are consistent with the silver content of gold from Sakdrisi; in this diagram, the horizontal axis is used to separate different sites from each other (M. Jansen).

Trace element fingerprints

18The trace element pattern of the Bronze Age artefacts is very variable. It is obvious that higher concentrations of trace elements can be detected for the artefacts than were observed in natural gold sources. Artefacts contain normally more than 10 ppm and up to 1,000 ppm of different trace elements (fig. 10), while natural gold usually shows the same trace elements in concentrations below 10 ppm (fig. 6). Where do the trace elements of artefacts come from? Trace elements could have been enriched by metallurgical processes and could be dramatically altered by processes of remelting, refining, leaded smelting and mixing with gold from different sources. Thus, we can conclude that it is not possible to trace artefacts back to their deposits taking trace element characterizations from natural free-gold particles, both primary gold and alluvial gold. If silver was added as an alloy component for some of the artefacts, trace elements of the gold could have been blurred by higher contents of the elements deriving from the silver. For non-alloyed gold, co-melting of heavy minerals, either from different ore or gangue minerals deriving from a primary deposit or from heavy minerals of a placer deposit, must be taken in account. This has already been discussed for Georgian artefacts by Hauptmann and Klein (2009). Similar observations were recently made for the gold of Varna (Bulgaria) by Leusch, Brauns, Pernicka (2016).

Fig. 10 – Diagram showing the trace element pattern of 70 Bronze Age gold artefacts from Georgia; the trace elements Sn, Pt, Pb and Zn are enriched for all artefacts; the mean of all other trace elements is higher than 10 ppm; dashed lines indicate maxima and minima for the specific trace elements; measured by LA‑ICP‑MS (M. Jansen).

19The use of multiple gold sources can be assumed if we observe a high variation in the trace element fingerprint of the Bronze Age artefacts, which might be not surprising since the artefacts cover a period of several centuries. However, by analyzing smaller groups of artefacts, e.g. 23 Middle Bronze Age gold artefacts from Tzalka (Trialeti), the variation in the trace element compositions of these is larger than for 43 gold artefacts from the royal tombs of the Early Dynastic Ur, Mesopotamia (fig. 11). This might be due to a different type of exchange occurring at each site. For the Early Dynastic city of Ur, a stable gold supply from a single source can be expected (Jansen, Hauptmann, Klein 2016), while a small scale tribal exchange of the Trialeti culture maybe led to a more complex composition due to the accession of multiple sources by different groups.

Fig. 11 – Diagram showing the trace element pattern (minima and maxima) of Early Dynastic period gold from Ur in comparison to gold artefacts from Tzalka; note the logarithmic scale; the larger variability in the artefacts from Tzalka might be due to the use of several alluvial gold occurrences, while in the Early Dynastic period Ur we expect a stable gold supply by the import of gold from a single occurrence; measured by LA‑ICP‑MS (M. Jansen).

20Even though the artefacts cannot be traced back to individual sources using trace elements, the composition of the gold artefacts bears geochemical information about the deposits they originate from. The importance of the PGE platinum was already discussed by Hartmann (1970, 1982). The six platinum group elements can be divided in platinum‑like PGE (PPGE), including platinum, palladium and rhodium, and iridium‑like PGE (IPGE) including osmium, iridium and ruthenium. These two groups behave geochemically differently, which also has consequences for their behavior during metallurgical operations. While IPGE can be found as inclusions of natural osmium-iridium-ruthenium alloys due to an extremely high melting point, PPGE minerals are soluble in gold and can be found as trace elements. Nearly all of the gold artefacts can be characterized as alluvial gold since they contain both tin (Sn) and platinum (Pt) as trace elements (fig. 12). These originate from tin minerals like cassiterite (SnO₂) and platinum minerals like isoferroplatinum.

Fig. 12 – Diagram of the trace elements tin and platinum showing 70 gold artefacts from Georgia; primary tin deposits occur in acidic rocks, while primary PGE deposits are formed in (ultra-)basic rocks; both Martkopi-Bedeni and Trialeti culture gold can be characterized as placer gold due to the combination of these elements in the trace element pattern; measured by LA‑ICP‑MS (M. Jansen).

21In primary deposits, tin and PGE minerals (PGM) are separated from each other, since they originate from different types of host rocks. PGM are formed in ultrabasic or basic rocks, while primary tin minerals are formed in acidic rocks. Thus, the milieus of the primary deposits exclude each other. Instead, several primary deposits of gold, tin and PGE were eroded and accumulated together in placer deposits.

22The characterization as alluvial gold is also important if lead isotope analysis (LIA) is applied to archaeological gold artefacts (e.g. Bendall et al. 2009; Hauptmann et al. 2010; Standish et al. 2013). LIA is applied successfully e.g. for silver and copper-based artefacts to trace back to the metal sources since the ores contain lead, and since the lead isotope signature from the deposits is invariable and not altered through smelting and processing. As we observed, lead was found in higher concentrations in the artefacts than in natural free-gold particles from a primary deposit and from micro nuggets of alluvial gold. Consequently, the lead from gold artefacts does not originate from the natural gold itself, but eventually from accessory lead-rich phases of the ore mineralization if a primary gold deposit like Sakdrisi was used. Since we were able to characterize the Bronze Age gold as coming from placer deposits, one should think about the lead in such deposits. Lead compounds of different sources can be discharged in the placer deposit, which do not have to reflect the lead isotope composition of the primary gold source. Thus, we can conclude that lead isotope ratios determined for gold artefacts do not trace back to the primary gold sources, but that the lead originates from lead-rich components in heavy mineral concentrates which were co-melted with the gold and so that subsequently the lead was incorporated in the gold. As a consequence, the lead isotope ratios of gold artefacts from alluvial deposits reflect several ore sources from a region, and possibly not single sources. Given this, even for unalloyed artefacts we are limited to discussing regional patterns rather than localized variation. In addition, we discussed the possibility of an intentional alloying with lead-bearing silver. LIA for those artefacts would trace back to the source of the alloy component silver and not to the gold source. Our lead isotope analyses for occurrences and artefacts will be discussed elsewhere, but these strict limitations should be emphasized here.

Bronze Age gold artefacts from Azerbaijan

23A group of 13 Bronze Age and two Chalcolithic gold artefacts from Azerbaijan was brought to the laboratories of the German Mining Museum and the Institute for Geosciences at the Goethe University in 2014. The characterization of most of the artefacts provides similar results to those from Georgia. Eleven Bronze Age artefacts can be assigned to placer gold deposits since they contain tin and platinum like the Bronze Age artefacts from Georgia (fig. 13). In several artefacts, PGE inclusions of the composition Os‑Ir‑Ru were identified. These include gold artefacts from Mentesh Tepe and Soyuq Bulaq (fig. 14). The PGE inclusions are tracers for alluvial gold deposits as well, since they are formed within a mineral named chromite in ophiolite complexes which do not contain syngenetic gold mineralization. The closest ophiolite complexes are those of the Sevan-Akera suture zone near Sotk, the largest gold deposit of the Lesser Causasus.

Fig. 13 – Diagram of the trace elements tin and platinum showing 15 gold artefacts from Azerbaijan; most of the artefacts show identical values to those from Georgia; the gold of those can be characterized as placer gold due to the combination of the elements Pt and Sn in the trace element pattern; two spiral rings show extremely low tin and platinum contents; also, the Chalcolithic gold beads from Soyuq Bulaq are lower in the trace element tin; different sources can be assumed; measured by LA‑ICP‑MS (M. Jansen).

Fig. 14a – Digital micrograph of PGE inclusions consisting of osmium, iridium and ruthenium identified in gold artefacts from Chalcolithic Soyuq Bulaq (M. Jansen).

Fig. 14b – Digital micrograph of PGE inclusions consisting of osmium, iridium and ruthenium identified in gold artefacts from Bronze Age Mentesh Tepe (M. Jansen).

24Two gold spiral rings of Hasansu have an atypical composition and were attested to derive from a primary gold source due to its lack of the trace elements platinum and tin (Jansen, Stöllner, Courcier 2014; Courcier et al. 2016), and they are identical with the silver content observed for the free-gold particles of Sakdrisi. The gold from the Late Chalcolithic beads of Soyuq Bulaq seems to originate from a different source than the Bronze Age gold. This is most evident through low tin concentrations around 1 ppm, while the Bronze Age gold is characterized by tin contents above 10 ppm. The identification of a few PGE inclusions in the gold from Soyuq Bulaq (see above) lead to the interpretation, that also for these artefacts gold placer deposits were used. Identical biconical gold beads of different sizes, like the two examples analyzed in this study, were found at Tell Brak (Syria). They belong to a cache of more than 350 beads made of various materials that was found beneath a residential building of the mid 4th millennium BC (Emberling, McDonald 2002, fig. 3). The similarity between those might underline the contact between the Caucasus and northern Mesopotamia in the Chalcolithic period, which is evident through the ceramics inventory (Lyonnet et al. 2008).

Conclusions

25Natural and artefact gold from the Lesser Caucasus was analyzed for its geochemical fingerprint. Natural gold from primary and alluvial occurrences does not contain significant copper and trace element concentrations, unlike what was observed for gold from the artefacts. The increase of trace elements in artefacts might be due to a co-melting of gold bearing heavy mineral concentrates produced by panning or, alternatively, due to processing of gold ores. In addition, silver with its own trace element fingerprint potentially might have been added to produce white gold for some of the artefacts. This means that trace elements cannot be reliably used to connect gold artefacts with specific gold occurrences since, together with alloy components, heavy mineral concentrations have to be taken in account. Thus, the geochemical signature of unalloyed gold artefacts is to be explained by a mix of different heavy minerals forming a placer deposit, or the compounds of a heavy minerals concentrate obtained by crushing, grinding and panning of ore from a primary deposit like Sakdrisi. The combination of tin and platinum in the trace element pattern of all Bronze Age artefacts from Georgia and most of the artefacts from Azerbaijan can be used as an indicator that alluvial gold from placer deposits was used.

26The earliest gold finds in the Lesser Caucasus, the Chalcolithic beads from Soyuq Bulaq (Azerbaijan), are different from the Bronze Age gold artefacts since the former reveal only a small amount of tin. A few inclusions of PGE alloys characterize the gold of Soyuq Bulaq as placer gold as well, but the gold source seems to be different from those used for the Bronze Age artefacts. The trace element fingerprint of the Bronze Age gold artefacts is typical of alluvial deposits, with one exception: two spiral rings from Hasansu (Azerbaijan) that are characterized by low trace element concentrations including tin and platinum. This might be due to the use of a primary gold deposit like Sakdrisi.

27We demonstrated that no other metal could have been exploited in the prehistoric galleries of Sakdrisi, and besides the use of hematite as a pigment, which might have played a role at least as a by-product, only gold can be found in exploitable concentrations in the ores. So where is the gold of Sakdrisi? First, we must consider that the artefacts presented in this study are mostly associated with the Trialeti culture and are younger than the mining activities in Sakdrisi, and that only a few of the Martkopi-Bedeni stage artefacts were sampled, which slightly overlaps with the younger radiocarbon dating of Sakdrisi (dates published most recently in Stöllner 2014). In general, the mining operations in Sakdrisi are earlier in date than the first extensive use of gold in Georgia. Secondly, we have to keep in mind that smaller portions of gold from Sakdrisi could have been mixed with gold deriving from other sources. The signature of gold from Sakdrisi which is low in trace elements could have been blurred by alluvial gold characterized by higher trace elements like tin and platinum, and these would be visible in the geochemical fingerprint of the mixed gold leading to a domination of an alluvial character. In this context, we must emphasize the fact that we would not have any suitable tracers available to identify gold from Sakdrisi at all. Finally, long-distance trade should be taken into account. Gold finds from sites like Arslantepe, Basur Höyük or even the Majkop culture are contemporaneous and should be considered for future research.