M. Járó - L. Költő szerk.: Archaeometrical research in Hungary (Budapest, 1988)
Analysis - BALLA Márta, BÉRCZI János, KEÖMLEY Gábor, ROSNER Gyula, GABLER Dénes: Provenance studies of ceramics by neutron actiwtion analysis
to determine as many isotopes as possible. For gamma-ray spectrometry measurements Ge/Li semiconductor detectors (ORTEC)are used with an energy resolution of 2 keV for the 1333 keV peak of Co— 60 at a relative efficiency of 12.6%; these detectors are connected to a multichannel analyser (CANBERRA-80) through amplifier and signal transformers. Evaluation of the gamma spectra, i. e. peak search, determination of the energy which corresponds to the gamma-lines, and the calculation of the peak areas and their error are carried out by a PDP 11/23 computer, using the program system SPECTRAN-F. The following elements were determined in the different sherds: Sc, Cr, Fe, Co, Rb, Cs, La, Ce, Eu, Yb, Lu, Hf, Th,U. For selecting the elements to be determined the following further points of view were considered (bearing in mind the measuring technique ): — To ensure the lowest detection hmit and relative error that radioisotope was chosen whose gamma line yielded the greatest peak/background ratio free of spectral disturbance. These two conditions were significantly influenced by the choice of cooling time . - Such radioisotopes were chosen which have a gamma-line (or lines) free of spectral disturbance. — Cooling times were selected in order to measure most of the isotopes in both spectra. More lines of an isotope were measured in some cases thereby providing more concentration data for a given element. In these cases the mean value of more measurements or the result with a smaller measuring error was accepted. — Only those elements were utilized for statistical evaluation for which it was unambiguous that analytical uncertainty could not affect the grouping. - A great number of further elements can be determined with the applied measuring conditions. Some of these do not furnish reliable data because of their partly or entirely disappearing during baking (Br, As e. g.), the others can only be measured with great error and poor reproducibility because of their adverse nuclear parameters (e.g.Zr). Processing of the measured data A considerable amount of analytical information can be obtained from the concentration of elements and the errors but their successful review, systematization and the grouping of samples that are similar can be accomplished only by applying computerized mathematical statistical methods. The primary point of view when choosing the methods was that data processing should modell the questions arising from archaeological investigation and the results should directly answer the questions. Two initial situations were modelled: - We do not have "a priori" information on the origin of the samples. A typical example of this approach could be the analysis of finds of an excavation. The ceramics found are grouped according to the distribution of their trace elements, without knowing the workshops where they were produced. A suitable mathematical method for this approach is cluster analysis. The basis of the method is the following. In the first step the similarity index is to be determined i. e. the variable expressing the similarity of two samples. A great number of algorithms are used to define the similarity index. Of these algorithms, the two that are most widely used were chosen, based on the calculation of distance and correlation coefficient, respectively. The distance coefficient (dy) is the distance of two samples measured in the n-dimensional space where n is the number of chemical elements.
