The corresponding author would like to dedicate this work to Uni. Prof. Dr. Reinhard Trettin, Institute for building material chemistry of which he was the director. I was extremely privileged by having such a magnificent mentor in the field of applied mineralogy and archaeometry whose advice and encouragement was never lacking. Sincerely thanks go to Dr. Reaz Motamed, geomorphologist and geologist for useful discussion about the research area. The authors would also like to thank and appreciate anonymous reviewers for their very constructive comments, furthermore for reviewing this article and indicating important highlights to accomplish this paper, which are gratefully acknowledged; any residual errors are our own.
1Ancient ceramics are the main objects obtained from the majority of archaeological excavations and play an essential role in investigating ancient settlements and their technological development. These investigations focus on how people lived and how their technology and ability to use raw materials developed. However, during the time between their initial burial and rediscovery, destruction or changes in the fabrication of ancient ceramics occurred due to in situ mass transposition or the decomposition of the crystalline phase constituents (Maggetti, 1982; Maggetti, 2001).
2In order to investigate how people developed the technology to produce decorative ceramics during the third to second millennium BC archeologists and scientists often considered the city of Haft Tappeh and the Ziggurat of Chogha Zanbil in the Khouzestan province in south-west of Iran (Ghirshman, 1955).
3Haft Tappeh was an ancient city located five kilometers south-east of Susa in the Khouzestan province (Figure 1). Haft Tappeh was a prosperous city in the Elamite Empire, based on the residency of the king Untash Napirisha (Carter & Stolper, 1984; Mofidi Nasrabadi, 2004; Ascalone, 2014). Haft Tappeh and Chogha Zanbil established to the 2nd millennium BC and despite extensive excavations during the modern era, new artefacts continue to be excavated and as such each site has been identified as a potential production center for decorative ceramics (Mofidi-Nasrabadi, 2016).
4The ancient city of Untash (or Dur-Untash in Elamite) with its magnificent Ziggurat of Chogha Zanbil, which was founded by Untash-Napirisha (1275-1240 BC) is located 25 km away from Haft Tappeh (Mofidi Nasrabadi, 2003; Mofidi Nasrabadi, 2004) (Figure 1). Based on their stylistic features, finds in both regions (Haft Tappeh and Chogha Zanbil), can be divided into four sub-periods: (I) Proto-Elamite (4000-2500 BC), (II) Early Elamite (2500-1500 BC), (III) Middle Elamite (1500-1100 BC) and (IV) The recent Empire (750-640 BC) (Potts, 2015). Chogha Zanbil is not only considered as the religious periphery of the Elam dynasty, but also a technical center for manufacturing of materials such as metals, binding material, ceramics and bricks (Thornton, 2010; Weeks, 2016; Amadori et al., 2018).
Figure 1: Location and altitudinal map of Haft Tappeh and Chogha Zanbil in south-east of Iranian plateau / Figure 1 : Localisation et carte altitudinale de Haft Tappeh et Chogha Zanbil au sud-est du plateau iranien
5During the extensive excavations, many ceramic fragments have been unearthed and included in this research and differ in colour due to the differences in firing conditions and raw materials and have some evidence of deterioration effect on their surfaces. Mineralogical and chemical characterization has been carried out on these set of ceramics by the author briefly investigate the development of firing technology in the Elamite period (Emami, 2012; Emami & Trettin, 2012).
6Based on previous petrographic investigations, the ceramics have been shown to have been manufactured from sedimentary and metamorphic based raw materials. Ceramics from Chogha Zanbil and Haft Tappeh are more often classified as CaO-rich calcareous raw material based on the ternary system SiO2 - Al2O3 - (CaO + Na2O + K2O) - (MgO+ƩFexOy). The differences between the mineralogy of ceramics are represented in the occurrence of different plagioclases such as high-temperature plagioclase (anorthite) and wollastonite (Tite, 1999; Velraj et al., 2015). Crystalline phase decompositions, (preliminary minerals as well as high-temperature neo-minerals, measured via QXRPD) have shown that the technology used in their manufacture varied depending on the raw material usage (Emami & Trettin, 2012).
7Statistical modelling could be a useful method for finding the best correlation between each element in an object. For example, heat mapping has been used by many researchers in different areas for clustering the data set from chemical composition. Heat maps are one of the most useful methods for reporting/summarizing data in large datasets. A heat map is a graphical two-dimensional representation of data in which the values of the data are contained in a matrix represented by various colours. Hot colors (such as red, orange etc.) usually indicate relative levels of each element and cold colors (yellow, orange etc.) usually indicate low level of elements. The columns within a heat map might represent different samples while the rows might represent different minerals, chemical compositions etc. The method can show the degree of similarity or a similarity gradient between the geochemical compositions of earthen-ware objects from different sites. Heat maps can also interconnect the chemical composition of materials with a definite mathematical model.
8Linked executive data sets are an emerging tool for studying the similarities between manufacturing techniques versus time in an archaeological site. Previous papers have described methods for concerning chemical data for these purposes (Emami, 2012; Emami & Trettin, 2012), although the results have not explicitly focused on numerical and statistical modelling, nor have they described the degree of associations between ceramics.
9This research focuses on ceramic fragments obtained from periphery of both Haft Tappeh and Chogha Zanbil and to cluster them using time series statistical methods. The statistical model will be established based on the longitudinal dataset according to the chemical data, and mineralogical phase characterization. The application of a proportional hazard Cox model for estimating the probability of similarities between the antique objects from different periods that have been quantitatively measured is a novel application of the method. In addition, innovative approaches for creating longitudinal datasets on ceramics can enhance existing data by providing information on a variety of results and provide flexibility as similarity and chemical composition data from newly unearthed objects becomes available.
10Ceramics were collected from the previously excavation at ancient sites of Chogha Zanbil and Haft Tappeh. Forty-nine samples from different periods were selected from Haft Tappeh (n=21) and Chogha Zanbil (n=28) (Table 1). Samples which dated from the 14th century B.C. to the 7th century A.D. show a reddish-beige colour with a smooth buffed surface and visible mineral additives. Samples from 2nd century BC to the 2nd century AD showed acharacteristically dense body with green-grey glaze on their external and internal surfaces. The colour of the body indicates the use of different clay, which was fired under reducing conditions (Pollard & Hatche, 1986; Kerr et al., 2004).
11Quantitative XRD analysis was used for determining the crystalline phases by a PANanalytical instrument. The samples were prepared as powder and measured at the X-ray station. Areas of samples which showed visible damage were excluded from the present study. Analyses were carried out using an X-ray generator operated with a copper target (Kα=1.5406 and Kα2 / Kα1=0.5) at 45 kV and 40 mA. The powder patterns of ceramics were measured between 2θ=5° and 2θ=60°, with a scanning rate of 2θ=0.01° per minute. This large scanning area was required in order to gain a complete diffraction pattern for minerals with low symmetry (Emami et al., 2016). The measured peaks are refined with Rietveld method with X’Pert HighScore Plus software package version 2.2c(2.2.3) for getting the best preferred orientation of the phases (based on Reference Intensity Ratio [RIR]). Clustering of analysis was carried out for distinguishing the ceramics according to the crystalline phase constituents (Emami & Trettin, 2010; Emami, 2012).
12300 mg of sample were powdered and mixed with 0.5% LiBr lithium borate which was used in the same ratio to the sample and pressed into pellets (20 mm diameter, 1 mm thickness) using a press machine. In these small holders the powder was slightly compressed and flattened manually for reducing porosity effects. Prepared samples were subjected to wavelength-dispersive XRF (WD-XRF) analysis on a Bruker spectra plus 2008. The spectrometer is equipped with an X-ray tube with a Pd-anode (50 kV, 200 W), and three position crystal changer where LiF(200) and PET are mounted as standard crystals. Indeed, recent generations of WD-XRF give the probable to decrease the quantity of samples by cautiously developing calibrations on sample’s amount as small as a few grams (De Vleeschouwer et al., 2011). This configuration provides exceptionally high resolution for elemental peaks from oxygen through magnesium (Klein et al., 2004). Experiments were indicated that the use of flux did not affect the calibration and this is additionally clear on the basis of the excellent coefficients of determination noted in the calibration lines of most elements (Georgakopoulou et al., 2017). The equipment was regularly calibrated with 45 commercially certified reference materials, which have been established for the analysis of ceramics and clayey based materials (e.g. NIST97b, NIST98b, NIST679). Quantitative calibration curve has been recognized for each element by using a linear regression between the corrected intensities and the elements concentrations within the samples.
13These samples were also examined by polarizing microscopy on an Olympus XB 51 with AnalySIS Software. Samples were cut into thin sections approximately 30 µm thick of for petrological as well as petrographical analysis.
14Statistical analyses was performed using R software (v3.3.2; lme4, Cox proportional hazards model (PHZ). The duration of stabilizing the geochemical dynamics over time [time-to-becoming dispersed] was analyzed using the Cox-PHZ; with time of follow-up ceramic samples used as the main covariate for evaluating the proportion of geochemical similarity of samples (in two historical locations) as the outcome, with each chronological group (2 in HT and 3 in CG) incorporated as a random-effect [frailty function].
15The Cox proportional hazards model is called a semi-parametric model because there are no assumptions about the shape of the baseline hazard function. There are, however, other assumption such as a linear association between the natural logarithm of the relative hazard and the predictors. Data that have a multilevel structure frequently occur across a range of disciplines, including epidemiology, health services research, public health, and sociology. Cox-PHZ model is prepared for the analysis of multilevel archaeometric data in this study. Cox proportional hazards model with mixed effects incorporate cluster-specific random effects that modify the baseline hazard function. Random effects can be incorporated to account for within-cluster homogeneity in outcomes. The application of these methods can be illustrated using data consisting of ceramic samples and statistical programming languages (R). In all GLMM models, backward elimination was used for sequential removal of non-significant variables, to obtain the minimal statistically-significant model (Huisman & Snijders, 2003). Kinetic modelling of geo-chemical samples were explored and visualized using heat-map CummeRbund v.2.0.0. [http://compbio.mit.edu/cummeRbund/] and R scripts (RStudio software v.3.1.0 [http://www.r-project.org/]).
16Samples were quantitatively measured via QXRPD with Rietveld refinement methods to determine the crystalline phase constituents (Table 1, 2 and 3). The refinement is carried out with a good of fitness (GOF) about “2” for each diffractogram of ceramics, which is very reliable for clustering the data. Based on the phase constituents of the investigated ceramics, quartz and calcite were the most important mineralogical factor for controlling the mineralogical phase decomposition during firing. The crystalline phase analysis suggests that there are two different groups of ceramics based on the raw material processing and firing technology (Figure 2a and 2b). The first group can be classified by the presence of quartz and Ca-rich minerals (such as calcite or dolomite) as additives or primary minerals. The presence of albite, muscovite and sanidine in the first group suggests the use of lithic (granitic) rock fragments. The second group consists of high-temperature phases such as augite, diopside, anorthite, gehlenite and enstatite, which have affirmed the firing condition of ceramics (Cultrone et al., 2001; Rathossi et al., 2010). All these phases were classified as secondary high-temperature minerals from firing (Ricarrdi, 1999). The ceramics were further discriminated based on different raw materials as well as firing conditions (> 970 °C). The estimated firing temperature for both regions is given by Emami and Emami et al. and estimated as being between 1000 and 1050 ºC (Emami, 2012; Emami & Trettin, 2012).
17Analcime is an important secondary crystalline phase which has been detected in the second group of samples and furthermore is a highlight for micritic calcite which has been identified in second groups of materials. Analcime is a common hydrated sodium aluminosilicate (NaAlSi2O6.H2O) this is classified as a feldspathoid mineral. This mineral occurs in seams and cracks within the ceramic texture due to the weathering of high Na-feldspar (albite) ceramics (albite in humid condition → analcime + SiO2). The crystalline phase composition is rendered less useful in this way as there is clear definition between the majority of the pieces in each group despite the raw materials being obtained from geologically similar sites.
18The difference in PCA analysis (Figure 2a and 2b) is therefore based on the different portions of quartz and Ca-rich minerals as additives in a calcareous ceramic (Heimann & Franklin, 1979). However, the shape of the crystalline phases and their preferred orientation are the most essential points for discriminating between the samples gathered from both region, as it illustrated also on the diffractograms in Figure 2, PCA analysis has been performed independent from the clustering and uses the correlation matrix. This mathematical method tries to find systematic variances in a large set of observations, the so-called own-values. The result is only influenced by the comparison parameters. The position of each data set based on its first three Eigenvalues is shown by a small sphere. The colour of the spheres indicates the cluster they belong to. The *** after a scan name indicates the most representative scan of a cluster. The “+” after a scan name indicates the two most different scans within one cluster. PCA suggests better clustering of both groups of ceramics from Haft Tappeh (14th century BC, and 2th century BC to 2th century AC) and Chogha Zanbil (7-8th centuries BC, 9-10th centuries BC, 11-12th centuries BC). The related PCA codes in Figure 2a and 2b to the samples are mentioned also in Table 1. Indeed, the variances in the progress of mineral phases are expected in samples with dissimilar chemical composition. But, the main purpose of this research will be focused on the degree of dissimilarities between the samples which are chemically mostly similar. This is arguable due to the diffractograms of all samples in these two regions and in this case the distance of similarity measurements will be considered (Pillay et al., 2000). By this way the similarity distance between all merged objects (in this case crystalline phase constituents) recorded, in order to construct a PCA distribution diagram.
Figure 2: Cluster analysis of the ceramics from both regions Chogha Zanbil (A) and Haft Tappeh (B), which were discriminated the classification of each groups of ceramics with respect to the crystalline phase calculation through Rietveld / Figure 2 : Analyse de grappes des céramiques des deux régions Haft Tappeh (A) et Chogha Zanbil (B), qui ont été discriminées grâce à la classification de chaque groupe de céramiques par calcul de la phase cristalline par la méthode Rietveld
Table 1: Investigated samples with their functionality and archaeological contextualization / Tableau 1 : Échantillons étudiés selon leur fonction et leur contexte archéologique
Functionality
|
Amount
|
|
Invertar No
|
Archaeological Context
|
1. Haft Teppeh
|
|
|
|
|
|
|
PCA Code
|
|
|
1.1. Big Jar
|
8
|
1
|
HT 84
|
14. Century BC
|
|
|
10
|
HT 72
|
14. Century BC
|
|
|
6
|
HT 52
|
14. Century BC
|
|
|
13
|
HT 89
|
14. Century BC
|
|
|
9
|
HT 71
|
14. Century BC
|
|
|
8
|
HT 64
|
14. Century BC
|
|
|
5
|
HT 46
|
14. Century BC
|
|
|
7
|
HT 59
|
14. Century BC
|
|
|
|
|
|
1.2. Dish
|
1
|
12
|
HT 88
|
14. Century BC
|
|
|
|
|
|
1.3. Big open bowl
|
7
|
2
|
HT 14
|
14. Century BC
|
|
|
3
|
HT 23
|
14. Century BC
|
|
|
4
|
HT 31
|
14. Century BC
|
|
|
11
|
HT 74
|
14. Century BC
|
|
|
21
|
HT 56
|
2. Century BC till 2. Century AD
|
|
|
17
|
HT 42
|
2. Century BC till 2. Century AD
|
|
|
20
|
HT 54
|
2. Century BC till 2. Century AD
|
|
|
|
|
|
1.4. Storage Vessel
|
5
|
16
|
HT 32
|
2. Century BC till 2. Century AD
|
|
|
15
|
HT 30
|
2. Century BC till 2. Century AD
|
|
|
18
|
HT 44
|
2. Century BC till 2. Century AD
|
|
|
19
|
HT 45
|
2. Century BC till 2. Century AD
|
|
|
14
|
HT 6
|
2. Century BC till 2. Century AD
|
|
|
|
|
|
|
|
|
|
|
2.1. Bowl
|
4
|
1
|
CZ 16
|
11. - 12. Century BC
|
|
|
14
|
CZ 26
|
11. - 12. Century BC
|
|
|
15
|
CZ 27
|
9. - 10. Century BC
|
|
|
8
|
CZ 15
|
9. - 10. Century BC
|
|
|
|
|
|
2.2. Big bottle
|
1
|
12
|
CZ 24
|
9. - 10. Century BC
|
|
|
|
|
|
2.3. Big open container
|
10
|
17
|
CZ 33
|
11. - 12. Century BC
|
|
|
18
|
CZ 34
|
11. - 12. Century BC
|
|
|
24
|
CZ 57
|
9. - 10. Century BC
|
|
|
7
|
CZ 13
|
9. - 10. Century BC
|
|
|
21
|
CZ 39
|
9. - 10. Century BC
|
|
|
16
|
CZ 28
|
9. - 10. Century BC
|
|
|
3
|
CZ 4
|
9. - 10. Century BC
|
|
|
28
|
CZ 11
|
9. - 10. Century BC
|
|
|
2
|
CZ 2
|
9. - 10. Century BC
|
|
|
6
|
CZ 9
|
7. - 8. Century BC
|
|
|
|
|
|
2.4. Storage Vessel
|
10
|
20
|
CZ 36
|
11. - 12. Century BC
|
|
|
9
|
CZ 17
|
11. - 12. Century BC
|
|
|
13
|
CZ 25
|
9. - 10. Century BC
|
|
|
4
|
CZ 5
|
9. - 10. Century BC
|
|
|
26
|
CZ 68
|
9. - 10. Century BC
|
|
|
27
|
CZ 90
|
9. - 10. Century BC
|
|
|
5
|
CZ 7
|
9. - 10. Century BC
|
|
|
23
|
CZ 49
|
9. - 10. Century BC
|
|
|
10
|
CZ 19
|
7. - 8. Century BC
|
|
|
11
|
CZ 22
|
7. - 8. Century BC
|
|
|
|
|
|
|
|
|
|
|
2.6. Dish
|
2
|
22
|
CZ 48
|
7. - 8. Century BC
|
|
|
25
|
CZ 65
|
7. - 8. Century BC
|
|
|
|
|
|
2.7. Jar
|
1
|
19
|
CZ 35
|
7. - 8. Century BC
|
Table 2: Quantitative phase analysis by Rietveld refinement methods of Haft Tappeh in different periods / Tableau 2 : Analyse de phase quantitative par la méthode de raffinement Rietveld des céramiques Haft Tappeh à différentes periods
|
Qz
|
Cc
|
An
|
Or
|
Ab
|
Do
|
Di
|
Ag
|
En
|
Ms
|
Ill
|
Gh
|
Anl
|
HT84
|
9
|
7,9
|
4,8
|
1,2
|
4
|
0
|
13,8
|
45,7
|
0
|
0
|
1,8
|
6,1
|
5,7
|
HT72
|
28,4
|
33,3
|
7,6
|
0
|
8,1
|
0,9
|
7,3
|
0
|
0,5
|
5,6
|
4,9
|
2,4
|
0,9
|
HT52
|
10,5
|
30,5
|
4,2
|
2
|
3,8
|
0,3
|
17,3
|
17,2
|
0
|
1,4
|
1,9
|
7,4
|
4,2
|
HT89
|
27
|
36,7
|
0
|
6,5
|
0
|
0,4
|
17,4
|
8,7
|
0
|
1,6
|
0,2
|
1,5
|
0
|
HT71
|
15,7
|
9,6
|
5
|
0,2
|
5,7
|
0,1
|
12,2
|
41,3
|
0
|
0
|
0
|
6,4
|
3,9
|
HT64
|
36,5
|
28,3
|
6
|
3,2
|
3,7
|
0
|
6,8
|
8,1
|
2,1
|
2,4
|
2,2
|
0,1
|
0,7
|
HT46
|
6,7
|
7
|
5,8
|
5,2
|
1,5
|
0
|
3
|
62
|
1,3
|
0,5
|
0
|
3,3
|
3,6
|
HT88
|
48,7
|
8,9
|
0
|
0,3
|
4,7
|
0,6
|
18,4
|
0,9
|
0,6
|
1,1
|
2,3
|
13,3
|
0,2
|
HT14
|
24,2
|
14,3
|
6,4
|
3,5
|
1,2
|
0,8
|
11,7
|
12,3
|
3,6
|
4,2
|
1,6
|
15,5
|
0,6
|
HT23
|
28,1
|
23,6
|
6,5
|
7,2
|
3,8
|
0
|
7,7
|
21,6
|
0
|
1,6
|
0
|
0
|
0
|
HT31
|
11,7
|
10,8
|
0,2
|
1,7
|
6,7
|
0,5
|
30,5
|
22,6
|
2,4
|
1,1
|
0,1
|
6,1
|
5,6
|
HT74
|
24
|
13,6
|
4,5
|
0
|
7,1
|
0
|
16,6
|
14,3
|
3,3
|
0
|
3,6
|
0,1
|
0,5
|
HT59
|
17,1
|
6,3
|
8,5
|
1,6
|
12,7
|
0
|
3,3
|
41,4
|
0
|
0
|
0
|
4
|
5
|
|
Haft Tappeh 2th century BC till 2th century AD
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
HT56
|
22,3
|
16,4
|
8,3
|
0,6
|
0
|
0,6
|
14,1
|
17,6
|
2,3
|
3,3
|
3
|
7
|
4,6
|
HT42
|
22,8
|
24,6
|
0,8
|
4,1
|
1,8
|
0
|
6,6
|
26,7
|
4,2
|
0
|
0,1
|
8,3
|
0
|
HT54
|
10,1
|
14
|
15,9
|
0,9
|
0,8
|
0,6
|
27,9
|
13,4
|
0,2
|
1,4
|
0
|
8,3
|
4,9
|
HT32
|
15,7
|
29,4
|
0
|
3,4
|
1,9
|
0,6
|
11,1
|
7,8
|
5,6
|
0,9
|
5,1
|
17,3
|
1,1
|
HT30
|
19,5
|
29,9
|
3,1
|
0,1
|
1,3
|
0
|
5,1
|
3,8
|
3
|
0
|
8,5
|
21,6
|
0
|
HT44
|
24,3
|
28,4
|
0
|
0,8
|
5,7
|
0
|
17,9
|
5,5
|
1,2
|
0,4
|
4,5
|
11,3
|
0
|
HT45
|
16
|
11,8
|
5,5
|
0,3
|
0,8
|
0,6
|
19,8
|
30,2
|
2,5
|
1,5
|
0
|
7,6
|
3,4
|
HT6
|
27,1
|
33,3
|
7,2
|
1,6
|
0,3
|
0,1
|
5,7
|
6
|
2,9
|
1,8
|
3,6
|
10,5
|
0
|
Qz: Quartz (ICSD 79634), Cc: Calcite (ICSD 40544), An: Anorthite (ICSD 34942), Or: Orthoclase (ICSD 34742), Ab: Albite (ICSD 87655), Do: Dolomite (ICSD 40971), Di: Diopside (ICSD 64977), Ag: Augite (ICSD 56921), En: Enstatite (ICSD 24464), Ms: Muscovite (ICSD 202263), Ill: Illite (ICSD 55333), Gh: Gehlenite (ICSD 20392), Anl: Analcime (ICSD 87555)
Table 3: Quantitative phase analysis by Rietveld refinement methods of Chogha Zanbil in W% / Tableau 3 : Analyse de phase quantitative par la méthode de raffinement Rietveld des céramiques Chogha Zanbil à différentes periods
Chogha Zanbil 7-8th century BC
|
|
Qz
|
Cc
|
An
|
Or
|
Ab
|
Mc
|
Sa
|
Do
|
Di
|
Ag
|
En
|
Ms
|
Ill
|
Gh
|
Anl
|
CZ9
|
33,1
|
5,1
|
16,3
|
13,3
|
0
|
1,2
|
3
|
0,1
|
21
|
3,8
|
1,9
|
0
|
0
|
0,6
|
0,5
|
CZ19
|
11,2
|
30,3
|
0
|
6,9
|
8,4
|
0
|
4,5
|
0
|
0
|
38,7
|
0
|
0
|
0
|
0
|
0
|
CZ22
|
9
|
1,7
|
11,2
|
11
|
0
|
9,7
|
5
|
0,7
|
19
|
24,2
|
5,1
|
0,4
|
0
|
0
|
3,1
|
CZ48
|
0,3
|
1,4
|
23
|
1,2
|
0
|
0
|
2
|
0
|
30,4
|
24,9
|
2,9
|
1,2
|
2,3
|
0
|
10,4
|
CZ65
|
0,7
|
0,4
|
23,6
|
0
|
0
|
1,2
|
0,1
|
0
|
60,3
|
3,6
|
0
|
0
|
0
|
0
|
10,2
|
CZ35
|
23,3
|
18
|
17
|
9,6
|
0,3
|
2,9
|
2,9
|
3,7
|
10,1
|
6,7
|
4,4
|
0,9
|
0
|
0
|
0,1
|
Chogha Zanbil 9-10th century BC
|
CZ27
|
21,7
|
45
|
6,4
|
2,6
|
0
|
3,8
|
5,2
|
0
|
4
|
10,3
|
0
|
1
|
0
|
0
|
0
|
CZ15
|
26,1
|
22,9
|
11,1
|
0
|
3,7
|
0
|
0
|
0
|
0
|
26,6
|
0
|
9,6
|
0
|
0
|
0
|
CZ24
|
12,4
|
5,2
|
17,8
|
4,8
|
0
|
1
|
6,1
|
0
|
21,8
|
17,5
|
1,8
|
0,6
|
0
|
7,5
|
3,7
|
CZ57
|
29,2
|
30,6
|
6,6
|
2,6
|
0,5
|
0
|
3,5
|
0
|
0
|
3
|
0
|
17,3
|
5,8
|
0
|
1,1
|
CZ13
|
33,4
|
21,4
|
7,8
|
0,9
|
0
|
0
|
0
|
0
|
18,8
|
6,6
|
0
|
11,1
|
0
|
0
|
0
|
CZ39
|
20,9
|
5
|
0,3
|
7,1
|
0
|
11,1
|
0,5
|
0,8
|
16,7
|
30,8
|
3,8
|
0,3
|
0,4
|
2,1
|
0,3
|
CZ28
|
34,8
|
22,1
|
9,5
|
0,2
|
4,5
|
1,5
|
6,8
|
0
|
6,2
|
5,2
|
0
|
9,2
|
0
|
0
|
0
|
CZ4
|
32,4
|
12,1
|
10
|
7
|
0
|
0
|
4,8
|
0,6
|
9,7
|
11,3
|
3,9
|
4
|
0
|
4,1
|
0
|
CZ11
|
22
|
8,5
|
0
|
0,5
|
16,9
|
0
|
0
|
0
|
0
|
52,1
|
0
|
0
|
0
|
0
|
0
|
CZ2
|
19
|
13,2
|
14
|
5,1
|
1,8
|
0
|
0
|
0
|
43,4
|
0
|
0
|
0
|
0
|
0
|
0
|
CZ25
|
24,4
|
12,5
|
10,7
|
4,1
|
0
|
0
|
2,6
|
0
|
0
|
38,3
|
5,4
|
2,1
|
0
|
0
|
0
|
CZ5
|
34,2
|
13,3
|
13
|
0
|
0,8
|
11
|
0,4
|
0,3
|
5,3
|
6,3
|
4,8
|
8,4
|
0
|
1,8
|
0,3
|
CZ68
|
31,4
|
5,6
|
20,5
|
13,9
|
0
|
4,1
|
2
|
0
|
16,4
|
2,1
|
4,1
|
0
|
0
|
0
|
0
|
CZ90
|
23,3
|
0
|
11,4
|
0,4
|
0
|
0
|
2,6
|
0
|
55,9
|
4,3
|
0
|
0
|
0
|
0
|
0
|
CZ7
|
26,6
|
28,3
|
13
|
2,2
|
0
|
0
|
0
|
0
|
0
|
24,6
|
2,3
|
3,1
|
0
|
0
|
0
|
CZ49
|
21,8
|
24,6
|
11,3
|
0
|
0
|
0
|
6,4
|
0
|
10,4
|
7,3
|
2,7
|
3,6
|
0
|
11,2
|
0,8
|
Chogha Zanbil 11-12th century BC
|
CZ16
|
12,1
|
5,3
|
15
|
0
|
0
|
0
|
0
|
0
|
22,4
|
45,2
|
0
|
0
|
0
|
0
|
0
|
CZ26
|
23,4
|
37,7
|
8,3
|
3,8
|
0
|
4,9
|
1,2
|
0
|
7,2
|
9,4
|
0
|
4,1
|
0
|
0
|
0
|
CZ33
|
27,1
|
30
|
9
|
0,6
|
0
|
0
|
4,3
|
0
|
1,2
|
1,6
|
2,1
|
22,2
|
1,9
|
0
|
0
|
CZ34
|
23,1
|
37,8
|
11,6
|
0
|
0
|
0,2
|
3,3
|
1,1
|
6,2
|
6,6
|
6
|
0,3
|
0
|
3,2
|
0,5
|
CZ36
|
8,2
|
5,5
|
7
|
2,1
|
0
|
0
|
2,6
|
0
|
64,5
|
1,4
|
1,2
|
0
|
0
|
3,5
|
4
|
CZ17
|
4,9
|
4,8
|
13,4
|
0,8
|
0
|
4,2
|
0,2
|
0,8
|
31,9
|
25,1
|
2,6
|
0,9
|
2,1
|
0
|
8,2
|
Qz: Quartz (ICSD 40009), Cc: Calcite (ICSD 16710), An: Anorthite (ICSD 22022), Or: Orthoclase (ICSD 34742), Ab: Albite (ICSD 87654), Do: Dolomite (ICSD 40971), Di: Diopside (ICSD 69701), Mc: Microcline (ICSD 100495), Sa: sanidine (ICSD 81384), Ag: Augite (ICSD 85158), En: Enstatite (ICSD 24464), Ms: Muscovite (ICSD 60569), Ill: Illite (ICSD 55333), Gh: Gehlenite (ICSD 20392), Anl: Analcime (ICSD 87555)
19Petrography of archaeological ceramics from Haft Tappeh and Chogha Zanbil was performed in order to examine the composition of body of the pieces and their mineralogical petrological interpretation. The petrography of the ceramic body can provide information on raw material reservoirs as well as manufacturing processes by means of composition, color and heterogeneity (texture) of the matrix (Whitbread, 1986).
20The b-fabric (birefringent fabric) of the ceramics suggest that the raw materials were collected from a residual clayey reservoir (naturally heterogeneity), while they contain extreme altered minerals within their fabrics, such as calcareous clay matrix (Figure 3a) from Haft Tappeh (14th century BC). In Figure 3b the other characteristics of the clayey are visible due to the association of bright admixtures of fine quartz (<7 µm) aggregates inclusive calcite (or calcareous rich fabrics) (<5 µm) inclusions. The b-fabric of the ceramic in Figure 3b from 2th BC-2th AD changes from the upper right to the lower left side of the Figure depending on the secondary clay deposit, which has been used within it (Quinn, 2013). The observation of these ceramics gives the conclusion that the fabric of sample in Figure 3a is more compact when compared to sample in Figure 3b. The admixtures originated from igneous host rocks due to the coexistence of albite and microcline as typical lithic (granitic) inclusion. Conversely, Figure 3b shows decalcification deterioration effects and a firing temperature more than 870 ºC (Cultrone et al., 2001; Rathossi, 2010). Clay raw materials differ between Chogha Zanbil and Haft Tappeh with respect to the b-fabrics and mineral inclusions. Figure 3C shows the orthoclase and 3D topical large crystal of biotite with its cleavage parallel to the (001) crystallographic axes as typical additives from the granitic rocks. Both ceramics were fired at high temperatures (>950 °C). This will be achieved by means of yellowish birefringent of its interference color due to the loss of Fe+3 from its crustal structure via high temperature reaction (Emami & Trettin, 2010; Deer et al., 2011). The firing temperature seems to have the similar gradient in both regions. Calc (CaO) aggregates in Figure 3e (bright brown aggregate in left side) showed the reaction of calcite above 870 ºC, which were followed by the formation of gehlenite, and clynopyroxene (bluish rims around the aggregate) due to humidity (Böttger et al., 2002). The presence of tiny micrite within the matrix of ceramic improved the residual clay deposit. Calcite appears mostly as secondary new-crystallization based on the humidity transfer in the ceramic’s fabric with very high birefringent and due to the conoscopic illuminations they are appearing optically as uniaxial crystals (Figure 3f). This means that the reaction would have occurred over a long time or long soaking time, therefore, the crystals had enough time to growing in the saturated Ca, and dynamic electrolytes from the burial conditions.
Figure 3: A) Haft Tappeh ceramic with calcareous fabrics and plagioclase with coexistence of alkalifeldspar. B) Residual clay with dissolved aggregates of calcite and the accumulation of mg rich rims as reason of using dolomite. C) Plagioclase within the matrix of a Chogha Zanbil sherd. D) Biotite with bright yellow interference color as a function of high temperature. E) residual clayey fabric with calcite aggregates which has been reacted to different silicate. F) secondary uniaxial crystals of calcite within the ceramic matrix / Figure 3 : A) céramique Haft Tappeh avec matières calcaires et plagioclases, et coexistence de feldspaths alcalins. B) Argile résiduelle avec des agrégats dissous de calcite et accumulation de pourtours riches en mg résultant de l’usage de la dolomite. C) Plagioclase dans la matrice d’un tesson de Chogha Zanbil. D) Biotite avec une couleur d’interférence jaune vif liée à la température élevée. E) matière argileux résiduel avec des agrégats de calcite ayant réagi à différents silicates. F) cristaux secondaires uniaxiaux de calcite dans la matrice céramique
21Ceramic production in antiquity proved to have maximum control of fabrications and choosing of aggregates (additives). However, this kind of pyro-technology is restrictive due to the firing conditions (Tite, 1999). Standardization of the products is often demanded by the potters to increase the overall quality of the finished product.
22Chemically, ceramics from both regions are divided in two groups based on Noll diagram (SiO2-Al2O3-CaO+MgO system), which is illustrated in Figure 4 (based on Table 4). The ceramics were divided into two subgroups; namely high calc and very high calc ceramics due to the anorthite – diopside line in the Noll ternary diagram (Emami et al., 2008). Both groups of ceramics from Haft Tappeh (Figure 4a) and Chogha Zanbil (Figure 4b) were scattered in a triangle between high calc and very high calc, whereas the ceramics from 7th-8th century showed an increase in the level of understanding of ceramic manufacture. Figure 4 shows that Chogha Zanbil samples vary drastically from calc rich toward very calc rich from 12th to the 7th century BC whereas samples from Haft Tappeh are far more clustered.
Figure 4: Distribution of the samples. A) Haft Tappeh, and B) Chogha Zanbil, in system SiO2 – CaO + mgO – Al2O3 after Noll 1999. The samples are divided in two groups of calc rich and very calc rich based on the points around diopside–anorthite knode / Figure 4 : Distribution des échantillons dans A) Haft Tappeh, et B) Chogha Zanbil, le système SiO2 - CaO + mgO - Al2O3 d’après Noll 1999. Les échantillons sont divisés en deux groupes riches en calcaire et très riches en calcaire sur la base des points autour de la distinction diopside-anorthite
Table 4: Bulk chemical composition of all ceramic finds from both places in different periods in W% / Tableau 4 : Composition chimique (en W%) de toutes les céramiques des deux endroits à différentes périodes
23Heat-maps evaluated the different concentration of elements in the ceramic body from all periods via XRF in order to get the best relationship between the presences of chemical compositions of materials at both sites (Table 4).
24These simple heat maps provide an immediate visual summary of total elements/oxides densities for all samples at the two separate locations. All samples were combined and only focused on element density variations in relation to historical site. However, the navy colour in both heatmaps represented the undetermined values which are programmed in R statistical software as “NAs” which means indeterminate values (Figure 5).
25The relationship between the measured concentrations of oxides in this diagram is then interpreted by means of the value of the concentration of each element represented by the colour (Huisman & Snijders, 2003). In this case, the dark reddish colour will describe the more concentration which has been reduced to zero by means of blue colour. Each horizontal line (each scan of x-aches) belong to the definite oxides (in W%) for each group of samples from each site.
Figure 5: Kinetic modelling of composted chemical data from all samples generated over time at two historical locations (Chogha Zanbil vs Haft Teppeh). These heatmaps represented the total density distribution of elements/oxides, which were collected at two historical locations, separately. A) Chogha Zanbil and B) Haft Teppeh. The red color shows the high density of oxides and yellow represents the low density of them / Figure 5 : Modélisation cinétique des données chimiques compostées au fil du temps (Chogha Zanbil contre Haft Teppeh). Distribution totale des oxydes de tous les échantillons qui ont été recueillis à deux endroits historiques, A) Chogha Zanbil et B) Haft Teppeh. La couleur rouge montre la haute densité d’oxydes et le jaune représente la faible densité d’entre eux
26The quantitative similarities between the components via time is discriminated in Figure 6 due to the probability of the samples during 14th century BC till 2th century AC in Haft Tappeh and 12th till 7th century BC in Chogha Zanbil. Figure 6 shows predicted proportion of geo-chemical similarity in all samples collected from two historical locations, A) Chogha Zanbil and B) Haft Teppeh, over time.
27Time series analyses, including survival analysis, refer to a statistical analysis that delves into the assessment of the time duration interval until one or more events happen. In business, finance or economic analysis, survival analysis is also referred to as duration modelling, most often dealing with answering financial or commercial questions, for example how many employees will stay in a given industry during a serious recession causing massive layoffs. Survival analysis in these fields can also assess how certain factors play to increase or decrease the rate of stability (e.g. survival) or persistence of certain economic or financial events. Under the survival analysis branch of statistics, the time duration until a certain event occurs is normally the outcome variable (Tableman & Kim, 2003). The Cox proportional hazards regression model can be written as follows:
28h(t) = h0(t) exp (b1X1 + b2X2 + … + bpXp)
29Where h(t) is the expected hazard at time t (in our case the proportion of dispersion in geo-chemicals), h0(t) is the baseline hazard and represents the hazard when all of the predictors (or independent variables) X1, X2, Xp are equal to zero. Notice that the predicted hazard (i.e., h(t)), or the rate of suffering the event of interest in the next instant, is the product of the baseline hazard (h0(t)) and the exponential function of the linear combination of the predictors. Thus, the predictors have a multiplicative or proportional effect on the predicted hazard.
30The Cox model approach is a case (the chemical composition of the ceramic sample)-crossover design for these archaeometric data. It is more efficient than a semisymmetric bidirectional case-crossover design only because more referent time points are used. The case-crossover design requires a choice of referent method for choosing control time periods (referent windows). With a valid referent method a localizable design and a conditional likelihood are constructed by conditioning on the proportion of events (geochemical similarity) experienced by each sample (ceramic) over the study period.
31The solid lines demonstrate the variation of the samples from Chogha Zanbil, and each step is the periods within the samples are excavated (Figure 6). The doted lines demonstrate the chemical change in Haft Tappeh area.
32The proportion of geochemical similarity in two historical locations over time was analyzed using Cox-PHZ, which showed that the number of millenniums for stabilizing the ceramic dynamics in Haft Teppeh (Location B) are significantly higher than those samples estimated for Choga Zanbil (Location A) [p < 0.001].
33The shaded area shows 95% confidence interval (CI). The solid black line shows the model predicted the probability of geochemical similarity over time in Chogha Zanbil. The dotted line represents the model (Cox-HPZ) predicted the probability of geochemical similarity in Haft-Teppeh samples being detected (n=693; χ21=480.11; p< 0.001). The Cox-HPZ model has been demonstrated the highly significant difference between samples collected from these two areas based on the time period. The statistical variation estimated based on the number of samples and their oxides and are discriminated of about 695 with the given mathematical algorithm. The highly significant P-value represents the greater dissimilarities between these two regions. The variation is measured due to the sample variance and population (Sample Error). In this case, the graph shows the faster duration of geo-chemical dissimilarities for Chogha Zanbil samples rather than Haft Tappeh. This means that the chemical composition of the ceramics from different period in Chogha Zanbil has more variation than in Haft Tappeh.
Figure 6: Dynamics of geo-chemical similarity in historical samples versus time / Figure 6 : Dynamique de la similarité géochimique des échantillons historiques en fonction du temps
34Ceramics are amongst the most significant materials that are excavated and studies from archaeological excavations. Not only does the production of decorative ceramics represent changes in the processing of raw materials it also represents the evolution of manufacturing technologies (Tite, 1999).
35Chemically, the samples from Chogha Zanbil and Haft Tappeh are characterized by their calc constituents. However, the clayey raw materials remained constant due to the presence of Al2O3 in the samples. Augit, diopside, quartz and calcite are the dominant crystalline phase occurrences in the Haft Tappeh. Ceramics Ceramics from Chogha Zanbil are characterized by the presence of more quartz and lower quantities of augite and diopside calcites. Sanidine and microcline are more prevalent in Chogha Zanbil ceramics.
36The ceramics studied in the current paper were collected from two important regions, Chogha Zanbil and Haft Tappeh, within the same cultural period. On cursory analysis, it seems that the pieces are similar to each other but based on the mineralogical chemical interpretations, they might be contrary to each other. The crystalline phase composition, as well as petrological studies, discriminates the samples via different phase compositions through XRD and cluster analysis (via qualitative and quantitative Rietveld Methods). However, the use of statistical calculations, dynamic of geo-chemicals and longitudinal modelling further diversifies the objects through time with regard to their chemical compositions. In this case, Cox-PHZ modelling has predicted the probability of geochemical similarities over time in both areas.
37According to the calculated heat-map represented in Figure 5a and 5b, the ceramics from both regions do not have the similar variety. In Haft Tappeh the colours were mostly focused on the cluster of ceramics which have been enriched from CaO, SrO, Fe2O3, Cr2O3, MnO, Na2O, SO3 and P2O5 with no attention to the periods in which the ceramics were manufactured. On the opposite, the concentration of CaO, SrO, Fe2O3 Cr2O3, MnO and Na2O in Chogha Zanbil ceramics is generally lower than Haft Tappeh ceramics. The heat-map demonstrates a dialogue between the visual concentrations of oxides from both regions.
38The Cox proportional hazards model is called a semi-parametric model because there are no assumptions about the shape of the baseline hazard function. However, there are other assumptions such as a linear association between the natural logarithm of the relative hazard and the predictors. Cox proportional hazards models with mixed effects incorporate cluster-specific random effects that modify the baseline hazard function. Random effects can be incorporated to account for within-cluster homogeneity in outcomes.
39The chemical composition in Chogha Zanbil changed periodically over time from 7-8th century BC to 9-10th century BC toward 10-11th century BC. However, the changes in chemical characteristics of materials from Haft Tappeh occurred over a prolonged period. This suggests that the recipe for ceramics in Chogha Zanbil was significantly altered, whereas in Haft Tappeh it remained mostly unchanged.
40This study aimed to use generalized linear mixed models (GLMM) to describe the dynamics of mineralogical character in the ceramic matrix, including chemical parameters and time. With these investigations, it is possible to develop a model of technological development based on the statistical measurements and mineralogical- chemical characteristics.
41From an archaeometrical point of view, the studied objects have a unique structure and exhibit very noticeable differences in firing processes (particularly temperature) and mineralogical generating process during the time.
42Statistical models provide mathematical comparisons of information extracted from the geochemical data. As a result, statistical modelling techniques can provide passable models to discriminate between ancient ceramic matrices which have been separated chronologically and geographically. In addition, this method could potentially provide better results for distinguishing archaeological objects, based upon their raw material composition. Based on the current dataset, it is possible to determine the level of similarities in different samples due to technological advances in a defined time period, and potentially obtain a quantitative adjustment of comparisons in a group of materials with regards to their manufacturing processes.
43The dynamic statistical model of the samples chemical composition shows dissimilarities between the ceramics with respect to their chemical composition and phase constituents. The perspectives and the methodology of the implementation of the Cox-HPZ model presented in this paper are generic and could be directly adapted to other systems and antique samples. It also might be developed to give a more detailed description of ancient ceramics.