Late Bronze Age Aegean Metallurgy in the Light of the Thera Analyses

As more samples and analyses become available, the study of the metallurgy of the area has developed from artifactually orientated research into consideration of the technology of metal working and possible explanations for the observed changes in metal types. In the nature of the work the main focus has been on the Chalcolithic and Early Bronze Age periods and by taking material from a wide geographical area, a fragmentary but comprehensible picture has emerged. The situation in the later periods is more complex but it is still useful to study a collection of material, particularly if it is well provenanced and dated. The metalwork from Akrotiri offers such an opportunity. Unfortunately, only 15 artifacts could be sampled, and as the samples were of poor quality, only limited conclusions could be reached on their composition. It is to be hoped that a more comprehensive programme can be undertaken. Although there are problems in the interpretation of metal analyses, it would show at least whether a variety of types of copper are represented and whether there was any selection of materials for particular purposes.

AEGEAN METALLURGY

The pattern of metal use in the Aegean parallels that in many areas of Europe and the Near East with a general change from pure coppers to arsenical bronzes and then true tin bronzes (Caley 1949; Renfrew 1967; Branigan 1974). This progression represents an improvement in the properties of the material, specifically in properties - such as ease of casting, lowering of the melting point, increased hardness and ease of working - that would have been important in the Bronze Age (Charles 1967 and bibliography). The improvement in casting is of prime importance as, although it is not absolutely impossible to cast pure copper in a closed two-piece mould, it is far easier and fewer casting defects result if an arsenic or tin bronze is used. There is some dispute as to whether arsenic or tin bronzes have better properties in respect of hardness and ease of working, and thus whether improvement or scarcity of arsenical ores was the impetus behind the development of tin bronzes. Their properties are related to composition (Slater 1972, 32) but are generally comparable for alloys of the optimum composition, with a tin bronze having a greater maximum potential hardness.

However, too much stress may be put on the importance of this, as there is evidence from the British LBA that the potential of a material was not always fully exploited (Allen et al. 1970). There also has been discussion as to whether the generalised changes in metal composition were the result of deliberate selection of materials. Tin bronzes certainly are the product of a deliberate admixture of tin and copper, whether as metals or ores; because of the paucity of finds of tin metal a suggestion has been made of the direct use of the ore (Charles 1975). The position of the arsenic bronzes is more difficult. There are some arsenic-containing copper ores, notably tetrahedrite Cu₅ (Sb, AS)₂ S₇ , enargite Cu₃ AsS₄ and tennantine Cu₃ AsS₄, whilst arsenopyrite Fe As S can also be found associated with copper sulphide ores. The main copper bearing ores that could have been exploited in antiquity are sulphides, often in the form of pyrites with additional iron. On weathering these deposits consist, in general terms, of an oxidised surface cap possibly containing native copper and copper oxide ores, overlying an enriched zone, formed by leaching of the upper levels, and then the pyrites layer. The complex copper-arsenic-antimony ores are typical of the enriched zone. Thus the sequence through a copper deposit can mirror the different types of copper - with the native and oxide sources yielding almost pure copper, the complex ores giving copper with associated impurities of arsenic (with possibly also antimony, nickel and silver), and with the pyrites ores giving purer copper with some iron. Thus the production of arsenic bronzes may have originally been 'accidental' in the sense of unintentional as a deposit was further exploited but once this better quality copper was produced from ores, not only different in colour than those used previously, but also needing to undergo a different smelting process, it is likely the necessary ores were deliberately selected.

The question of the production of arsenic bronzes - whether as the result of the sequence outlined above or by the addition of arsenic-rich material to copper (McKerrell 1976; Eaton & McKerrell 1976) - is important for the study of the Late Bronze Age, when arsenic bronzes had predominately been superceded by tin bronzes, because if it is simply a question of different mixtures at different periods the same basic copper might have been used throughout. There are two main types of arsenic bronzes, those with appreciable amounts of antimony nickel or silver, and those relatively pure in these other elements. Some of the most outstanding examples of copper-arsenic alloys come from the Caucasus, where it is suggested there were additions of arsenic rich materials to produce arsenic contents up to 30% (Chernykh 1965). but it is not necessary to have the same method for the production of alloys of lower arsenic content. In central Europe these alloys are thought to have been produced by the use of tetrahedrite ores, their use charted by the associated antimony content. In the Near East/Aegean area enargite and tennantite are more common than tetrahedrite and thus the source of arsenic is more difficult to determine and the alloy compositions are more varied. In this region the idea of deliberate alloying is based partially on analyses from Horoztepe and Kultepe with a base copper postulated with no antimony or arsenic with nickel being introduced with the arsenic (McKerrell 1976). This idea has certain attractions but the nickel/arsenic relationship is not as direct as suggested and the situation may be more complex. It is clear from the Aegean that, as paralleled elsewhere, there is a decrease in arsenic content of the metal with the introduction of tin bronzes, with a similar but less clearly defined, relationship between the antimony and tin contents (figs. 1 & 2). It is just as likely that this change in composition of the base copper arose from the exploitation of pyrites ores (either because of exhaustion of the enriched zone or a change of ore body) as from a substitution of tin for arsenic as a deliberate alloying agent. This is supported by a variable but generally increased iron content.

Unfortunately, further discussion about sources and trade is limited without an understanding of the basic processes. As Muhley has recently said "We thus come to the conclusion that the analytical evidence available at present tells us very little about the copper trade in the Mediterranean during the Late Bronze Age... Before we can say anything about trade we must first master the basic technology of Bronze Age copper metallurgy"  (1977).

POTENTIAL SOURCES OF COPPER AND TIN

The sources of copper and tin available to Aegean metalworkers have recently been documented (Muhley 1973) and there has been a lively discussion on the tin sources (eg. Dayton 1971; Muhley & Wertime 1973) and the arguments will not be rehearsed here.

It has been suggested that in the Aegean EBA there may have been local developments in metalworking based on local ore sources, the evidence coming from typology, artifact distribution, evidence from the previous neolithic period and finds related to technology such as moulds (Renfrew 1972, 308 ff; Branigan 1974, ch II & III). The sources of the metal cannot, at present, be determined by metallurgical analysis particularly because of the widespread use of arsenical coppers and their variable composition (see a recent survey by Tylecote 1976). Artifact typology and chronology of changes in metal use may reflect similarities and differences between areas, but this does not necessarily imply contact and certainly not trade in metal. A new slant has been put on EBA metallurgy with the study of extensive mines dated to these periods in their respective areas with variations in the quality of the dating evidence (Rothenberg 1972; Jovanovic 1971; Giles et al. 1974).

The main evidence for trade in the Late Bronze Age comes from the ingots (Buchholz 1959; Muhley et al. 1977; Mosso 1910), whose distribution in the East Mediterranean reflects the importance of Cyprus and Crete in metalworking at this time. The analyses of the ingots reflect the problems in assigning provenance. The source of one ingot had previously been ascribed to the mines at Ergani Maden because of the cobalt content (Maddin & Muhley, 1974). However, as further studies showed that the ingots were produced from weathered and unweathered sulphide ores, and the unweathered ores from Cyprus do contain cobalt, no direct assignment can be made. Additionally zinc is regarded as an important impurity in Cypriot ores, but zinc was not detected even in the ingots from Cyprus. This in no way diminishes the importance of Cyprus as a source of metals in the LBA, which is supported by strong evidence, but highlights the problems of source attribution on no evidence but that of analysis as discussed below.

THE USE OF METAL ANALYSES

In addition to providing a further means of describing an object and to chart changes in composition with time, analyses of prehistoric metal objects are often used in attempts to identify objects of similar composition and thus, implicitly, those produced from the same source material. The ultimate goal may be the geographic source of the ore if indeed, all potential sources can be located. There are however, problems in the interpretation of analytical data. Apart from the obvious errors in the analytical method, segregation may produce variations in composition within an object (Coghlan 1953; Charles 1973). Such variations may be smaller than the observed analytical error (Craddock 1976) but they may affect specific uses of the data (Slater & Charles 1970).

Additionally hot working of alloys of arsenic and antimony results in a reduction in the concentration of those elements (McKerrell & Tylecote 1972). There are further problems if an object is to be related to a source or objects thought to be from the same source. There may have been reuse of scrap or trace elements introduced by alloying.

The more volatile elements such as zinc and arsenic may be lost if the smelting process involves roasting, as is the case for sulphide ores or the smelting process may even introduce additions with the flux. Ore bodies may not have a uniform composition, in either major or trace elements (Chernykh 1966; Griffiths et al. 1972). It is immaterial whether ancient mines were deep or shallow as these variations affect the whole deposit, and not just the surface layers, and would not have been averaged in the small amounts of ore used for each smelt. A further set of problems have been highlighted by variations in analytical results produced from analyses of samples of the same metal by 21 laboratories (Chase in Beck 1975). As a result of these and other considerations Craddock (1976) has rightly issued a caveat on the comparison of analyses from different programmes.

As some of these problems cannot be resolved we should accept that the study of metals has its own limitations and may not produce results comparable to those, say, from obsidian. However, on the positive side, general trends can be recognised and on occasion composition may be sufficiently diagnostic for specific interpretations to be made.

THERAN SAMPLES

The Theran samples raise an additional problem as they are of poor quality. Only broken objects could be sampled and they all showed varying amounts of corrosion (sample 14 being discarded as it was completely corroded), although the most obvious corrosion products were discarded before analysis. There has been discussion as to the relationship of the quality of the sample and the value of the analytical results (eg. Jedrzejewska 1962; Organ 1962) but it is generally accepted that a sample should be taken from the bulk uncorroded metal away from surface layers which, whilst not necessarily showing signs of corrosion, may have been affected by surface leaching, preferential corrosion, cleaning surface enrichment etc. In some of the Theran samples mineralisation had extended through the object. As it was recognised that this would probably have affected the concentrations of the major elements the samples were analysed by neutron activation analysis to determine the current composition and whether any trace elements were present at diagnostic values. Samples 1 - 11 were kindly analysed by Dr. A. MacKenzie of the Scottish Universities Research and Reactor Centre, East Kilbride and Dr. MacKenzie describes the analytical method at the end of this paper. Samples 12, 13 & 15 were analysed by his predecessor, Dr. R. Wellum, using a similar method. The lead and nickel contents are being measured by X-ray fluorescence at the Kelvin Laboratory, University of Glasgow by courtesy of Dr. K. Ledingham.

The analytical results are presented in Tables 1 and 2, with a measure of the method error. The copper content was determined to give some idea of the degree of mineralisation and the results are presented in terms of absolute values and in relation to the copper content. Whilst the results are given for 12 elements, selected as potentially the most useful, the samples were also scanned to see if other elements were showing systematic variations. Fig. 3 gives the sample numbers related to the find spots of the objects. 

DISCUSSION OF RESULTS

There are over 50 copper-based objects in the Apothiki on Thera, plus the objects in the National Museum in Athens. Therefore, the 14 analyses represent a small sample, and they cannot, a priori, be recognised as a representative sample as they come from broken objects - 12 from a tripod, 2 from a pin, 5 from a hook, 10 from bronze sheet and 7 from a 'bronze plate'.

The analyses are presented graphically in figures 4 - 8, with the values as given in Table 1. The comparative analyses are those given by Bittel (1959), Bossert (1967) Craddock (1976), Goldman (1931), Junghans et al. (1968), Lamb (1936) Schmidt (1902) Tsountas (1908) and Vollgraff (1906). The use of these analyses is obviously contrary to the comments by Craddock (1976) and the problems discussed above, but they are used only to display general trends, not in terms of their detail. Detailed discussion would also be precluded by the quality of the Thera analyses.

The values in the final column of Table I give an estimate of the corrosion of the samples (although the lead contents, not yet available, may also make an additional major contribution). This is also emphasised by comparison of Tables I and II. The corrosion of copper-based alloys is complex with certain elements tending to be concentrated in the corrosion layers (Catling & Jones 1977), and the values in Table II would only be 'better' if all elements corroded at the same rate as copper. It has been argued elsewhere that ratios of element concentration might give a better guide to the composition of the source material in the case of elements like arsenic and antimony which might be affected in a similar way by the smelting process (Slater 1972) and a comparable argument could be advanced for corroded objects, with the values for elements more resistant to corrosion, like gold and silver, given as absolute values.

Figures 4 - 6 show that the Thera analyses are not outside the general picture for the Aegean LBA (figure 6 is an enlargement of part of 5) with respect of the tin and arsenic contents. Figure 7 shows a slight clustering of the arsenic/antimony ratios about 7 and it would be tempting to consider a single source but even allowing for errors, there seems to be no correlation with the silver, gold or other trace element concentrations (figs 7 & 8). It may be easier to show similarities in composition amongst objects from one area in the EBA, eg. Renfrew's analyses 40 & 56; 58, 59 and 63 of material from Amorgos, when there is less danger of mixing of materials and additions from alloying elements. All the objects contain cobalt and all but three zinc, albeit in trace quantities - the elements suggested as being diagnostic of the Ergani and Cypriot mines. In this context it is interesting that of the 112 analyses for Late Bronze Age material published by Craddock (1976) only 6 had appreciable zinc contents of which two (190 and 191) came from Cyprus. The concentrations of some trace elements were measured in the hope, not fulfilled, that they might be diagnostic.

All the objects contained iron, which can arise through the incomplete slagging of pyrite ores, and, in general, the iron content of Aegean copper based objects increases from the EBA to the LBA.

In conclusion, it would unfortunately be dangerous to say anything with certainty about these 14 objects on the basis of their current composition other than the fact that 10 were deliberate tin bronzes, and with the hope that the study can be expanded.

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

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

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

ANALYSIS OF BRONZE SAMPLES

FROM THERA - ANALYTICAL METHODS

SAMPLE PRE-TREATMENT AND STANDARDS

Eleven samples of bronze from Santorini were analysed for 12 elements by instrumental activation analysis. The samples, with the exception of number 2, were in a fragile, corroded condition and were powdered using an agate mortar and pestle. About 0.1 g of each sample was weighed into a cleaned polythene vial to give uniform sample geometry for irradiation and counting. Sample 2 could not be readily powdered and, in this case, a small piece of the solid sample was used, having a similar geometry to the powedered samples.

Copper standards for irradiation with the samples were prepared from finely cut copper wire, in each case about 0.1 g of copper being used in the same type of vial as used for the samples. Multi-element solution standards were prepared from Specpure materials for the remaining elements of interest, the volume of standard used again being matched with the sample volume.

The elements for which analysis was performed are shown in Figure A along with isotope and decay data used (Weast 1974 - 75).

IRRADIATIONS AND COUNTING

Neutron irradiations were performed at a flux of 3.8 X 10¹² n cm^-2 sec^-1 in the UTR 300 research reactor at East Kilbride. In an initial 20 second irradiation for copper determination, samples plus standards were transferred to and from the reactor via a pneumatic "rabbit" system. After irradiation, the vials were wiped clean and counted the same day for ⁶⁴Cu. The main irradiation of the samples was for 4 hours in the central vertical stringer of the reactor, after which a 7 day decay period was allowed before gamma spectra were recorded for the determination of As, Au, Sb and Sc. After a further decay period of 20 days, the samples were re-counted for Ag, Co and Sn determinations and after a further 30 days, the analytical sequence was completed by determination of Cr, Fe, Zn and Cs concentrations.

Gamma spectra were recorded using an 80cc Ortec GeLi detector, having a resolution (FWHM) of 2keV. A low background is achieved by the use of 4 inch lead shielding round the detector. A Data General 'Nova 2/4' computer (16 bit word; 8K memory) is used as a multi-channel analyser to record spectra using a program (MCAN 3) developed at East Kilbride by Dr R. Wellum. Spectra thus obtained are transferred to floppy disc for storage prior to analysis. Data interpretation is achieved by transfer of the spectrum to a hard-disc based Data General 'Nova 3/12' computer (16 bit word, 32K memory). The spectrum analysis program PEAKB, also developed at East Kilbride, locates peak positions, calculates their energy from a given calibration and carries out a smoothed Covell calculation of the peak area.

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

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