Provenance Studies of Theran Lead

There are indications supporting the idea of local lead production, some of which are: (a) the considerably high abundance of lead objects found in the excavations in comparison with the number of bronze objects; (b) the existence of pieces of lead ore in the excavation area, which are reported for the first time; (c) the low Pb-Fe-Cu mineralization that occurs in the underlying metamorphic rocks of Thera.

However, the fact that during the 3rd and 2nd millennia BC Siphnos and Lavrion were the production centres for Pb and Ag makes the above hypothesis debatable. In addition, no metallurgical workshops or slags have yet been found at Akrotiri. Microscopic studies on polished and thin sections, and also Neutron Activation Analysis were carried out on Pb ores from the prehistoric mines of Lavrion and Siphnos. Also, samples were taken of the Pb-Cu mineralization from Athinios on Thera, the site of the modern, now defunct ore mine, along with pieces of lead ore and metallic Pb from the excavations at Akrotiri, and examined by the same methods.

The combination of the trace element and microscopic results constitutes a new attempt to investigate whether there was metal production on LBA Thera, or whether metal objects were imported from other Aegean areas.

INTRODUCTION

The majority of the metal objects (weights, plates, ingots etc.) found in the Late Bronze Age settlement of Akrotiri were lead. Furthermore, the number of lead finds is greater than that of similar finds in Crete or the other Cycladic islands (Petruso 1978; Michailidou 1989). This fact indicates that the inhabitants of Akrotiri were familiar with the use and applications of lead for their needs, and also that they had access to raw material.

During the 2nd and 3rd millennia BC the known production centres of lead in the Aegean region were Siphnos (Wagner and Weisgerber 1985) and Thorikos, near Lavrion (Spitaels 1984). In the Akrotiri excavation site, however, individual pieces of lead ore have been recently identified which, together with the existence of a dim Pb-Cu mineralization in the island, might suggest a possible exploitation of lead during the Bronze Age. These deposits have been mined during the last hundred years. On the other hand, no evidence of metallurgical activity (slags, workshops etc.) has been found in the excavated part of the settlement.

In this paper the provenance of a number of lead objects found in Akrotiri is examined by means of mineralogical and trace element analysis.

GEOLOGICAL RECONNAISSANCE AND SAMPLING

       Thera:      The underlying semi-metamorphosed basement is exposed in the southern part of Thera. It is composed of tertiary phyllitic schists (Pichler and Kussmaul 1972) which have hydrothermally altered, locally and impregnated by ore bearing solutions resulting in conspicuous accumulations of lead, zinc and copper sulphides (Murad and Hubberten 1975). One dim net of quartz veins accompanies the mineralization zones. One part of the mineralization is closely associated with the skarn-forming stage of a granitic intrusion of Late Miocene Age, exposed near Athinios (Skarpelis and Liati 1990). The mineralized zone is located in the Athinios area where an underground modern mine operated during the last century, mined for primary and secondary Pb ores. The galleries which were visited lie between 30 and 60 m above sea level. They follow the direction of the metalliferous veins (strike 120°-150° east, dip 20°-30°south-west) and are orientated in discordance to the phyllitic schist. The mineralization also follows some feather joints (strike 80° west, dip 40°-60° east) developed in the country rock, and they have also been worked. The dimensions and the shapes of the galleries indicated modern mining techniques: height 1.5-2.5 m, width 0.8-1.3 m, shafts of depth up to 10 m and width up to 2.5 m and rooms 3 x 3 x 2 m. The galleries where reconnaissance was possible were inspected in order to find any possible ancient works and sample lead ores. Unfortunately no ancient works or remnants of ancient mining activity were found. Samples from the primary and secondary mineralizations were taken. A description and the mineralogical composition of all the samples is given in Table 1.

Three lead ore pieces were sampled from the excavation area in Akrotiri. Two of them (SN6 and SN7) were hand specimens and one (SN5) was a block of 60 x 50 x 60 cm and more than 150 kg in weight. Additionally, samples were taken from four lead artefacts, also included in Table I.

       Lavrion:      The ore mineralization of Lavrion (Fe-Mn carbonates and mixed sulphides of Zn, Fe, Pb) is mainly developed in an autochthon system consisting of marbles, dolomites and schists (Marinos and Petrascheck 1956). This system is generally subdivided into three horizons: the lower marble, the so called Kaisariani-mica schist and the upper marble. Above the autochthon system the series of an overthrusted nape lies in structural unconformity. The nape consists of phyllites and includes scales of limestone, marbles, sericitic schists, quartzite, green stones and serpentine. One granodioritic body, traversed by many small applitic dykes and by hydrothermal quartz, has inthrusted into the autochthon system. Many dykes or sills composed of granodioritic porphyry are associated with the granodioritic intrusion and they traverse the autochthon system reaching up to the overthrusted phyllitic nape. These dykes of a Middle or Late Miocene Age are known as eurites and they are in generally associated with ore mineralization. The sequence of the strata and the position of the thrust plane form a series of contacts where the ores occur. Among the metalliferous contacts the richest in ores are those where the schists form the hanging wall. During antiquity all the metalliferous contacts, where rich Pb-Ag mineralization exists, were mined (Konophagos 1980). However, during the 2nd millennium there is only indirect evidence for a Pb-Ag exploitation (Spitaels 1984). In Thorikos the first metalliferous contact occurs within the upper marble, below the base of the phyllitic nape. In this position the exploitation started at least as early as the 2nd millennium BC. Even today, the ore is exposed and its underground extension is followed by a system of galleries more than 500 m in length. The dim mineralization generally consists of Fe-Mn oxides and secondary Pb-minerals. It is developed in a carbonate matrix (upper marble). This mine was visited and four representative ore samples were collected as well as one sample of metallurgical slag from a nearby Roman furnace. Two other samples, from 3rd metalliferous contact, which was richer in Pb, were collected from the wider Lavrion area (Kamareza) along with one slag sample taken from the same area (Table 1). The last three samples (two ores, one slag) were collected for comparative reasons, in the knowledge that the Kamareza area and the third metalliferous contact were never reported as prehistorically exploited.

       Siphnos:      The island of Siphnos belongs geologically to the western part of the Attic-Cycladic crystalline complex. Its metamorphic series consists of a sequence of alternating thick-bedded marbles and heterogeneous schists and gneisses (Davis 1966). The middle and southern part is built predominantly of marbles, albite-rich green schists and mica schists, while the stratigraphically lower exposed formation is composed of a muscovitic-chloritic-schist, traversed by quartz veins. Ore mineralization occurs in the above mentioned lower schist, in the directly overlying marble, and also in the stratigraphically upper marble. The type of the ore mineralization differs between these two units. The ore mineralization of the upper marble consists of Fe-Mn oxides, appearing in the form of stocks or lenses, streaky or irregularly shaped. These are bodies mainly composed of goethite, hematite, manganese oxides and smithsonite. Secondary Pb and Cu-minerals also appear in lower concentrations, as well as galena and pyrite. The ore mineralization of the lower schist and overlying marble is composed mainly of pyrite, goethite, hematite, Cu-sulphides and secondary Cu-minerals. It is generally accepted that the character of the upper marble mineralization is associated with fillings of Karst phenomena and syndiagenetical procedures (Vavelidis et al. 1985). Only one part of the lower schist ore mineralization has a possible hydrothermal origin. The age of this ore formation is between 110-160 million years (Chalkias et al. 1988). Both metalliferous strata were sampled, the upper marble in Ayios Silivestros and the lower schist in Apokofto. The ore occurrence in Ayios Silivestros was mined in prehistorical times, although no strong evidence exists for exploitation during the 2nd millennium (Wagner and Weisgerber 1985). On the other hand the gold-bearing mineralization of the lower schist is said to have been exploited only during the Archaic and Classical periods (Pernicka and Wagner 1985). This site was sampled for geochemical comparison only.

Lead ore samples or lead pieces of miscellaneous origin (North Europe, Canada, North Greece etc.) were also sampled and used for reference.

It is known that in polymetallic ores any serious variation in the composition of the different metals can also affect the co-existing minor and trace elements composition as well as the mineralogical composition of the ore (Ginzburg 1960). The same phenomenon is also observed in the metallurgical slags, not always expected to be geochemically comparable with the mother ores. For these reasons, the majority of the ore-samples analysed correspond either to typical Pb-bearing occurrences or to sites already exploited during antiquity, such as Ayios Silivestros (Siphnos) and Thorikos (reported as mined during the 2nd millennium BC) where the samples were taken from unexploited supporting columns of ore or very close to existing tool marks. However, a few samples not belonging to these two categories were analysed for comparison. These were SF1, from the Au-bearing mineralization of Siphnos, SN3, SN4, mainly Fe-bearing from Athinios mine, KM3, Fe-Cu-bearing from Kamareza and TR4, KM1 metallurgical slags from Lavrion.

A description of the sampled material is given in Table 1, together with the results of the mineralogical study. XRD analysis confirmed and assisted the results of the ordinary microscopical identification carried out on thin and polished sections.

The comparison of the ore-paragenesis characters of the three metalliferous areas shows that although Lavrion, Thera and Siphnos belong to the Attic-Cycladic crystalline complex, their ore mineralizations differ in a number of properties. Table 2 contains some of these properties which are comprehensively given based on the relationship between the ores and the existing granite intrusions as well as on the age and the type of the ore. It can be concluded that the Thera ore mineralization seems to be similar to the Lavrion one while Siphnos differs subsequently. Based on features of the relationship of the ores to known magmatic centres and on local structural conditions, Petrascheck (1955) concluded that Lavrion may be incorporated within the Aegean Taurus metallogenetic sub-province. Also Launay (1913) stated that Lavrion seems similar with the metalliferous Bokardagh of Taurus in Asia Minor. These aspects in particular have been confirmed by lead isotopic analysis (Wagner et al. 1986).

CHEMICAL ANALYSIS

All the samples, geological and archaeological, were analysed by neutron activation analysis for 7 elements (Ag, As, Co, Cu, Fe, Sb, Sc). Two portions (about 50 mg each) from every sample were analysed in order to duplicate the results and reduce errors. The samples underwent two separate irradiation and counting procedures. A short irradiation of 5 min in the Demokritos swimming pool reactor at a thermal flux of 2 x 10¹³ n.cm^-2.sec^-1 followed by a 60 sec count, 3 days after the irradiation for the determination of As, Sb and Cu. A longer irradiation (1.5h) was carried out also at the same flux, followed by a 7 day cooling period and 10 min counting for the determination of Sc, Ag, Co and Fe. In addition the samples were analysed for Pb, Al, Zn and Ca using the energy dispersive system EDAX 9900 coupled to a Phillips 515 SEM.

RESULTS AND DISCUSSION

This study was initiated after the identification among the stones in the Akrotiri excavation of one block of secondary Pb-ore (cerrusite) along with two other Pb-ores of the same nature, but smaller in size. These findings provoked two questions: what was the origin of these pieces and whether there was any relation between them and the lead objects found in Akrotiri.

The main goal of the data analysis was to determine similarities or differences within the neutron activation analysis results of trace elements. In principle, ores from different geological formations should have different trace element composition, and objects made from a particular ore should have a similar composition to it. However, the direct comparison of trace element concentrations between the several samples analysed is not realistic because the Pb concentrations vary significantly (1-91%) and the ore samples differ in their nature (some of them are galena, others rich in Fe etc.). To overcome this problem two approaches were used: the comparison between ratios of concentrations and the notion of characteristic vectors derived from the variance-covariance matrix of the data group.

The examination of the chemical data (Table 3) shows that the ore pieces found in Akrotiri cannot be related to the Siphnian ones. Siphnian ores contain Sb in the order of 0.1-17% (see also Vavelidis et al. 1985) while Akrotiri ores contain Sb in the order of hundreds of ppm. The Siphnian sample SF2 (52 ppm Sb) was derived from the lower schist mineralization which is not Pb-bearing (Vavelidis et al. 1985). Thorikos samples (Fe-Zn bearing samples with smithsonite and manganese oxides) are not as well related to the ore pieces found in Akrotiri. The Akrotiri samples contain less than 1.5% Zn and less than 2.5% Fe.

As a first approach the ratios of particular elements (Sb, As, Ag, Co) which are associated with Pb mineralization were used for the statistical analysis (Pernicka 1988). The comparison of ratios instead of absolute concentrations is a form of standardization for the major element variations (Seeliger et al. 1985). The pairs consisted of elements with similar geochemical character. As can be seen from the plots of Fig. 1 and 2 the ratios Ag/Pb plotted against Ag/Co and As/Sb against Sb. The concentrations of As and Sb have been proved useful in studies where Pb objects are involved (Grimanis 1970). Sample SF1 is not included in Fig. 1 because Ag and Co were not detected due to high dead time, as a result of the high Sb concentration in this particular sample.

In the bivariate plots of Fig. 1 and 2 only two variates were accommodated at a time. However, an existing discrimination is a multivariate event because it is based on more than two variates. In order to include more than two elements simultaneously in the evaluation of the results, the data were transformed to the variance-covariance pattern established for the set. With this transformation the simple element axes (vectors) change to the characteristic vectors (eigenvectors) derived from the variance-covariance matrix of the data group. These vectors are linear combinations of the original co-ordinates and they are independent of each other. Then all the original sample measurements were projected on these orthogonal vectors and divided by the square root of the characteristic variance. In this way a new set of standardized co-ordinates was obtained for each data point, from which all correlation has been removed. The new set of characteristic vectors in no way alters the position of the data points in a multivariate space. They are similar to the original ones but their origins have been transferred to the centroid of the data group and have been rotated to direction of maximum and minimum variances within the group (Sayre 1977). There is one limitation in the multivariate calculations: the number of samples must be at least three times as much as the variates (elements). It was found that four elements could be used simultaneously in order to obtain successful results in this study. The first series of calculations included the concentrations of As, Sb, Co and Cu. From the four eigenvectors formed, the projections of all the data points on vectors 2 and 3 were used, to illustrate the separations in the bivariate plot of Fig. 3. Vector 2 was mainly loaded by the concentrations of As and Co and represented 14% of total variation. Vector 3 was loaded by Sb and Co and represented 31% of the total variation within the data. In an attempt to use more elements for the multivariate calculation the concentration of Cu was replaced with Ag and the procedure was repeated. Vectors 2 and 3, which account for 13% and 34% of the total variation respectively, were used again (Fig. 4). Vector 2 was loaded equally by all four elements and Vector 3 by Sb and Ag.

It should be mentioned, first, that the number of samples analysed is not sufficient to allow definite conclusion, as it does not satisfy statistical requirements. However, as will be shown, they give some encouraging preliminary results. Looking at all the plots it is clear that the samples from Thorikos (TR1, 2, 3, 5) tend to cluster together or at least to concentrate on the same side of the dividing lines. Sample TR4 is a slag and does not cluster with the rest of the Thorikos samples (Fig. 2, 3, 4) because during the smelting of Pb ores the As and Sb contents decrease (Pernicka and Bachman 1983). The situation with the Kamareza samples is quite different. Sample KM2 (Pb-bearing) lies at the same side of the diagrams as the Thorikos samples, but not always close to them (Fig. 1, 4). This is also the case with KM1, although it is a slag. However, KM3 (Fe-Cu-bearing) is usually isolated, showing no relation with the others (Fig. 2, 3, 4).

Another group of samples that go together in all the plots, are the lead objects from Akrotiri (SN8, 9, 10, 11) along with one Pb-ore sample from Athinios (SN1). Close to them and always at the same side of the diagrams are the second Pb-ore sample from Athinios and the three Pb-ore pieces found in Akrotiri (SN5, 6, 7). The other two samples from Athinios (SN3, 4), lower in Pb than SN1 and SN2, do not behave in the same way. They are found either at the side of Thorikos samples (Fig. 1) or at the side of Theran samples (Fig. 2). However, in Fig. 3 and 4 they cluster together with the Theran ones.

Furthermore the Siphnian samples fall far from both the Thorikos and Theran ones. This is an additional indication that the Akrotiri ores are not geochemically correlated with the Siphnian mineralization. From the study of the plots in Fig. 2, 3, 4 it is also evident that the Siphnian samples differ very significantly from each other. That is to be expected because the sample SF2 derives from the lower schist mineralization which, as already described, is non Pb-bearing. It is a gold-bearing sulphide mineralization which contains a few hundred ppm of Au and corresponds to the TG91-9 sample of a previous work (Vavelidis et al. 1985).

The general impression given from all the plots is a clear tendency for distinction among Thorikos, Theran and Siphnian ore samples. The ore samples found in Akrotiri along with the lead objects tend to fall together with the Athinios samples, indicating common provenance. The lead ore or pure lead reference samples (GL12, GL2, PB2, PB3) usually lie at the periphery of the plots. There is one complication with sample PB1 (lead from Parthenon with possible Lavrion provenance) because it tends to cluster close to the Theran group (Fig. 1). However, in all the other plots this sample is found either closer to the Thorikos samples or between the Theran and Thorikos groups.

The particular complications observed if one was to compare Lavrion and Thera, which do not change the general impression, are results of the similarities between the ore genesis of Lavrion and Thera as already has been described. The geological similarity imposes a factor of additional complexity in provenance studies of lead, where Thera and Lavrion are involved. Gale and Stos-Gale (1981), working with a wide range of metal material (ores and objects) from the Aegean region, came to conclusions that could not support the present results. In the case of Theran lead (26 objects) they suggest a Lavrian provenance, based on lead isotopic analysis. However, the analyses did not include samples of lead ores from the site of Akrotiri, which according to the neutron activation results seems to relate to the Athinios Pb-bearing samples. This shows the necessity for a co-operative investigation, based on a large number of samples (geological and archaeological) and a variety of analytical techniques.

CONCLUSIONS

At the present stage of the work the conclusions should be considered as preliminary and used as a guide for further research. They could be summarized as follows:

1. Siphnos seems not to be a source of Theran lead.
2. According to the material analysed, the Fe-Zn-Pb mineralization of Thorikos, reported to have been mined during the 2nd millennium BC, does not seem to supply lead to Thera.
3. The ore pieces found at Akrotiri appear geochemically similar to the Pb-ores of Athinios mine. They are also similar to the Akrotiri Pb objects.
4. The ore mineralization features of Thera and Lavrion are similar and this makes it difficult to discriminate between these areas geochemically. On the other hand a concrete distinction by means of geological and geochemical features is possible for the Pb-Ag metallogenesis of Siphnos.
5. Future studies on lead provenance should take into account the geological features of the studied areas. In particular, ores similar in nature should be compared (i.e. primary, secondary etc.).

The combination of more than one analytical method could provide more reliable answers to lead provenance problems.

--------------------------------------------------

--------------------------------------------------