Regional Differences in the Chemical Composition of Cuneiform Clay Tablets

To elucidate regional differences in the chemical composition of cuneiform clay tablets originating from Iraq and Turkey, chemical analysis was conducted using portable X-ray fluorescence analyzers. The analysis included clay tablets from the 21 areas of Ur, Larsa, Lagash, Uruk, Umma, Adab, Drehem, Nippur, Kish, Borsippa, Dilbat, Babylon, Sippar, Nusi, Nimrud, Nineveh, Tell Brak, Tell Halaf, Boghazkoy, Kultepe, and Alalakh, currently stored in the Yale Babylonian Collection at Yale University and the British Museum. Multivariate statistics such as principal component analysis, discriminant analysis, and cluster analysis were applied to the chemical analysis results. Based on the chemical compositions for Ca, K, and Fe, the clay tablets were classified into four groups corresponding to the upper stream area of the Tigris and Euphrates River, the lower stream area of the Tigris and Euphrates River, the northern and central areas in Turkey, and the southern area in Turkey. This grouping was determined mainly by a difference in Ca content dictated by the local geology.


Introduction
To obtain chemical compositional information to elucidate the provenance of cuneiform clay tablets, we conducted non-destructive chemical analysis of clay tablets originating from Iraq and Turkey.Chemical compositional data were obtained using portable X-ray fluorescence analyzers and multivariate analysis was applied for the obtained data.Chemical analysis data were collected for 540 clay tablets stored in the Yale Babylonian Collection of Yale University (in 2009, 2011, and 2012), and 94 clay tablets including bullae and seals stored in the British Museum (in 2013).
Cuneiform clay tablets made during the Mesopotamian period were used as writing media for cuneiform characters.The use of clay tablets dates back to around 3300 BC.A large number of clay tablets were made in the Ur III period (2112BC to 2004 BC).Economic, administrative, or agricultural information was recorded on the clay tablets.In this research, chemical analysis was conducted on the economic and administrative clay tablets whose information on the provenance was available.
Neutron activation analysis has been previously used to analyze clay tablets (Artzy et al., 1976;Dobel et al., 1977;Goren et al., 2009).However, this method is not favorable because the method is destructive.In contrast, a recently developed portable X-ray fluorescence analyzer (pXRF) is non-destructive and quick, with good sensitivity.In-situ analysis using pXRF has been applied to the various fields including archaeology and artifacts (e.g., Williams-Thorpe et al., 2008;Cesareo et al., 2008).Chemical compositional analysis of clay tablets using pXRF has been previously performed by Goren et al. (2010), mainly on clay tablets from Egypt and Boghazkoy (Hattusa).In this study, we carried out the chemical analysis of clay tablets from Iraq and Turkey.Clay tablets from Egypt were not analyzed in this research.

Chemical Analysis
The chemical compositions of Ca, K, Fe, Ti, Rb, Sr, Mn, and Zn, which were analyzed with good precision by pXRFs for all the clay tablets from each area (listed in  Grouping on the basis of the principal component analysis and the cluster analysis. Judging from the detection limit and precision for the pXRF calibrations, the chemical analysis data of Ca, K, Fe, Ti, Rb, Sr, Mn, and Zn were selected as candidates for multivariate analysis.However, as Mn and Zn frequently showed unusual values (Table 2), the multivariate analysis was conducted excluding these elements.Manganese concentration has been frequently observed on the surface of clay tablets, owing to Mn precipitation by a manganese oxidizing microbe (Laurito et al., 2005;Gütschow, 2012;Uchida & Watanabe, 2014).This may be the cause of the compositional variation in Mn on the surface of the clay tablets.

Multivariate Analysis
Principal component analysis, cluster analysis using Ward's method, and discriminant analysis were performed on the data shown in Table 2. Excel "Tokei" for Windows (Social Survey Research Information Co., Ltd., Tokyo, Japan) was used for the multivariate analysis.
Based on results of the principal component analysis using the averaged chemical analysis data in Table 2 for the above-mentioned six elements (Ca, K, Fe, Ti, Rb, Sr, Mn, and Zn), the areas where the clay tablets were made were classified into four groups (Figure 2(a)).Group 1 corresponds to the lower stream area of the Tigris and Euphrates River, Group 2 corresponds to the upper stream area of the Tigris and Euphrates River, Group 3 corresponds to the northern and central areas (Boghazkyoy and Kultepe) in Turkey, and Group 4 corresponds to the southern area (Alalakh) in Turkey.Because Group 1 slightly overlaps with Group 2 in Figure 2(a), we tried to eliminate some elements from the six elements to find the best combination of elements.As a result, we found that the combination of Ca, K, and Fe was best for the grouping.The result is shown in Figure 2(b).
As it was revealed by the principal component analysis that Ca, K, and Fe are the useful components for the discrimination of the clay tablets, the Ca-K-Fe triangular diagram was drawn for all the clay tablets using the chemical analysis data in Appendix based on the above-mentioned four groups (Figure 3).The four groups are well distinguished with some overlap in Figure 3.
Cluster analysis using Ward's method for Ca, K, and Fe of the clay tablets in Table 2 were carried out and the result is shown in Figure 4 as a dendrogram.The cluster analysis was conducted for normalized values obtained by dividing the concentration for each element by the average value.At a distance of 1 in the dendrogram (Figure 4), the clay tablets could be classified into the same four groups obtained by the principal component analysis (Figure 2).
Results of the discriminant analysis using the averaged chemical analysis data for Ca, K, and Fe of the clay tablets in Table 2 and the four groups obtained from the principal component and cluster analyses are shown in      2 and the four groups obtained from the principal component analysis (Figure 2) and the cluster analysis (Figure 4).The four groups by the principal component analysis and the cluster analysis are well separated.The symbols are the same as those in Figure 2.

Consideration
The clay tablets were classified into four groups, as presented above, based on the multivariate analyses including the principal component analysis, cluster analysis (Ward's method), and discriminant analysis.This grouping was determined mainly by the contents of Ca, K, and Fe.In particular the Ca content was essential for the grouping.The Ca content in the clay tablets decreases remarkably in the groups in the following order (Figure 3): the southern area (Alalakh) in Turkey (Group 4: 22.0% in average), the upper stream area of the Tigris and Euphrates River (Group 2: 13.3%), the lower stream area of the Tigris and Euphrates River (Group 1: 11.0%), and the northern and central areas (Boghazkyoy and Kultepe) in Turkey (Group 3: 4.1%).As the raw materials of clay tablets comprised sediments from rivers or canals, their chemical content is likely controlled by the surrounding geology, especially the distribution of limestone.Goren et al. (2010) conducted non-destructive chemical analysis using pXRF and multivariate analysis for clay tablets mainly from Ugarit (corresponding to Alalakh/Group 4 in this study), Mitanni and Assyria (corresponding to the upper stream area of the Tigris and Euphrates River/Group 2 in this study), Babylonia (corresponding to the lower stream area of the Tigris and Euphrates River/Group 1 in this study), and Hattusa (corresponding to Boghazkyoy and Kultepe/Group 3 in this study).Clay tablets from Egypt were also included in that study.On the basis of multivariate analysis, Goren et al. (2010) classified these clay tablets into five groups: Ugarit, Mitanni, Babylonia, Hattusa, and Egypt.They suggested that the above grouping might appear in a Ti-K diagram.In this context, we drew a Ti-K diagram for all the clay tablets investigated in this study (Figure 6), and found that the four groups were distinguished well with some overlap.Comparison of the compositional data for Ti and K in this study with those from Goren et al. (2010) shows a systematic difference.The concentration levels of K and Ti in the Goren study are significantly lower than those in this study.The difference is especially large for K; the K content found by Goren et al. (2010) is around half of that in this study.Calibration of the pXRF instruments for our study was conducted using reference rocks as mentioned above, whereas Goren et al. (2010) did not calibrate their pXRF.Calibration of pXRF using reference materials is essential (Ross et al., 2014), and thus this difference is the likely reason for the discrepancy in chemical compositions between the two studies.Discrepancies in the results were also evident for Mn and Rb.Uchida & Watanabe (2014) conducted chemical analysis of clay tablets from the lower stream area of the Tigris and Euphrates River using both an electron probe X-ray microanalyzer (EPMA) and pXRF.The results obtained by EPMA coincide with those obtained by pXRF in this study.Therefore, it is concluded that the results obtained in Goren et al. (2010) are not correct because the pXRF was not calibrated using standard materials.

Summary
In summary, the principle component and cluster analyses using the average chemical compositions for K, Fe, and Ca found in clay tablets resulted in their classification into four groups, corresponding to the upper stream area of the Tigris and Euphrates River, the lower stream area of the Tigris and Euphrates River, the northern and central areas in Turkey, and the southern area in Turkey.This grouping was determined mainly by a difference in Ca content dictated by the local geology.

Figure 1 .
Figure 1.Location map of the areas where the clay tablets investigated in this research originated.Position of the ancient Tigris and Euphrates rivers is taken from Postgate (1992).

Figure 2 .
Figure 2. Results of the principal component analysis of the clay tablets.(a) Results using the averaged data for Ca, K, Fe, Ti, Rb, and Sr in Table 2; (b) results using the averaged data for Ca, K, and Fe, in Table2.Group 1 corresponds to the lower stream area of the Tigris and Euphrates River, Group 2 corresponds to the upper stream area of the Tigris and Euphrates River, Group 3 corresponds to the northern and central areas (Boghazkyoy and Kultepe) in Turkey, and Group 4 corresponds to the southern area(Alalakh)  in Turkey.

Figure 3 .
Figure 3. Ca-K-Fe triangular diagram for all the clay tablet using the data for Ca, K, and Fe in Appendix.

Figure 4 .
Figure 4. Result of the cluster analysis (Ward's method) using the averaged data for Ca, K, and Fe of the clay tablets in Table2.The same grouping as in Figure2was obtained in the cluster analysis.

Figure 5 .
Figure 5.The discriminant functions W1 and W2 describe 87.63% and 11.68%, respectively.The four groups corresponding to those obtained by the principal component analysis and the cluster analysis are well separated in Figure 5.

Figure 5 .
Figure 5. Results of the discriminant analysis using the averaged data for Ca, K, and Fe of the clay tablets in Table2and the four groups obtained from the principal component analysis (Figure2) and the cluster analysis (Figure4).The four groups by the principal component analysis and the cluster analysis are well separated.The symbols are the same as those in Figure2.

Figure 6 .
Figure 6.Ti-K diagram for all the clay tablets based on the grouping obtained by the principal component analysis (Figure 2) and the cluster analysis (Figure 4) using the data in Appendix.The four groups were distinguished with some overlap.

Table 1 .
List of clay tablets studied in this research.

Table 2 .
Averaged chemical compositions and standard deviations (1σ) of the clay tablets determined by the portable XRFs. *