Neutron Activation Analysis of Aegean Obsidians

Source determination has been obtained using eight analysed elements; the artefacts have been assigned to one or other of the two sources on Melos and Giali. Implications of their source assignments are discussed.

INTRODUCTION

The success of trace element analysis in characterising obsidian sources is now well established. Renfrew and his co-workers (Cann & Renfrew 1964; Renfrew, Dixon & Cann 1965, 1966, 1968) have shown that the use of optical emission spectroscopy enabled broad divisions to be made amongst obsidian in the Mediterranean, Aegean and Western Asian regions, the divisions being based chiefly on variations in barium and zirconium concentrations. Other elements, notably lithium, iron, yttrium, magnesium, lead and calcium were found to be useful in breaking down the broad divisions into meaningful groups. The range of trace elements available for source discrimination can be extended significantly using the technique of neutron activation analysis (N.A.A.). A considerable body of literature has built up relating to its general application in archaeology and its particular use in obsidian analysis, for example Gordus, Wright & Griffin 1968; Hallam, Warren & Renfrew 1976. It is fortunate that to, an extent, N.A.A. complements the traditional spectrometer method of analysis in that its greatest sensitivity is to those trace elements, such as the rare earths, that had hitherto been assessed only qualitatively. It has also the advantage of being non-destructive of the sample, so that, if necessary, complete small artefacts may be examined.

The first neutron activation analysis of Aegean obsidian was undertaken by Aspinall, Feather & Renfrew (1972). Refined source discrimination was obtained using a linear combination of trace element concentrations. Of particular interest was the ability of the technique to distinguish between the two Melian sources of Sta Nychia (Melos A) and Dhemenegaki (Melos D). Numerous archaeological site specimens from mainland Greece and Aegean islands were unambiguously assigned to sources in this study. The initial work is now being consolidated by examinations of large numbers of source samples and, where appropriate, archaeological specimens. This report is concerned with the analysis of selected samples from Akrotiri and from the collection obtained by Professor J.D. Evans in Crete.

We have compared the results of this analysis with those for our established sources and, in this way, defined the origin of the Theran and Cretan material.

EXPERIMENTAL METHOD

The technique of modern N.A.A. has been adequately described elsewhere. (see for example Tite 1972) and only the relevant details are given here. Fragments of obsidian in the mass range 100 to 2,500 mg were cleaned by immersion in hydrofluoric acid followed by washing in high purity water. After drying and weighing, the samples were wrapped in aluminium foil and batched with powdered samples of a pottery standard based on that produced and circulated by Professor I. Perlman.

Concentrations of many elements in this pottery have been established by N.A.A and other methods, so that it provides a convenient multi-element standard.

The batched samples were irradiated in the Herald nuclear reactor of A.W.R.E., Aldermaston for a 48-hour period at a thermal neutron flux density of 1.8 X 10¹² n cm^-2 sec^-1. On removal from the reactor, the predominant gamma radioactivity was ascribed to ²⁴Na, which has a 15-hour half life for decay. No analysis was possible until this activity had decayed to a measurable level; thus any radio-isotopes with shorter half-lives than that of ²⁴Na could not be assessed. The examination of the specimens twenty days later allowed measurements to be made of long-lived radio-isotopes. All the measurements were taken using large volume Ge-Li detectors couped to multi-channel data accumulation systems; obsidian and standard pottery spectra were compared using computer program SPECT, developed at Bradford, for concentration estimation.

In all, twenty-one elements were identified in obsidian, but quantitative estimates were only possible for fifteen of them owing to the presence of interfering activities in the relevant regions of the spectra. Indeed several of the measured elements were subject to experimental inaccuracies of up to 20 % and subsequent data interpretation was weighed accordingly. The measured rare earths showed the expected correlation with each other but all were retained for possible source discrimination on geological grounds.

Two other features were also evident in the absolute concentration values. They were a) an overall inter-element correlation for individual samples and b) a rather wide concentration scatter for individual elements in a group of source specimens. These effects undoubtedly arose, in part, from non-uniform neutron flux densities through specimen containers during irradiation and partly from the varying "geometry" presented to the Ge-Li detector by obsidian samples of differing size and shape. A considerable improvement in sample grouping was obtained by "normalising" data for each sample to the concentration of a particular element. Scandium was chosen for this normalisation; its activity can be measured with high accuracy in the spectra of both short and long-lived isotopes. 

In our earlier work we reported that, on statistical grounds, the elements Cs, Ta, Rb, Th, La, Ce, and Fe gave high discrimination amongst sources when expressed as concentrations relative to Sc. We grouped the data into the two linear function Fe/Sc and

1/Sc (Cs + Ta + Rb/100 + (Th+La+Ce)/10 )

as giving the best overall two dimensional interpretation. There remained, however, a close grouping between the Giali and Ciftlik clusters that was readily resolved when a third function, Co/Sc, was utilised. In Table 1a we show the range of values for these functions for the small number of source samples then analysed. Of the seven site samples received from Akrotiri, six were broken into two or more fragments before irradiation so as to give a concentration homogeneity check. Seven samples from Knossos were selected by Dr. G. Diamond as possible specimens of differing origin. The results of these analyses are given in Table 1b together with their source assignments.

DISCUSSION OF RESULTS

The analysis of obsidians tfom Akrotiri yielded the expected result that the specimens are Melian in origin. However, two points of interest arise. Firstly, one of the donated specimens clearly originates from the Dhemenegaki sources.

A study of the relative numbers of samples from the two Melian sources therefore becomes feasible. There is no clear distinction in working quality of obsidian from the sources, but any territorial division amongst working groups can now be easily recognised. Secondly, the analysis emphasises once again the homogeneity of usable obsidian. There is a total range for all twelve fragments from the Sta Nychia source of 8 % for the linear discriminating function and of 6 % for the Fe : Sc ratio.

They compare favourably with the experimental accuracy that may be achieved for these elements by the N.A.A. method.

The specimens from Knossos divide unambiguously into three groups with origins on Melos and Giali. It can seen that it is necessary to invoke the use of the Co : Sc ratio to place specimens in the Giali rather than the Ciftlik group. Renfrew et al have reported the presence of Giali obsidian in Crete in Middle or Late Minoan contexts, the material being readily identifiable because of its characteristic white flecking. The specimens analysed by us, however, were not so readily designated and only the detailed studies of Dr. Diamond caused him to suggest further analysis. We have been informed by Professor Evans that they derive from Early Neolithic levels, a fact that gives rise to speculation on early contacts between Knossos and the Eastern Aegean.

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

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