Infra-red Spectroscopy of Ceramics from Tell Brak, Syria

Journal of Archaeological Science (2000) 27, 993–1006 doi:10.1006/jasc.1999.0510, available online at http://www.idealibrary.com on

Infra-red Spectroscopy of Ceramics from Tell Brak, Syria Murray L. Eiland and Quentin Williams Department of Earth Sciences, University of California, Santa Cruz, CA 95064, U.S.A. (Received 10 June 1999, revised manuscript accepted 24 October 1999) Infra-red microspectroscopy (IR), in both reflectance and transmission geometries, can be successfully applied to archaeological ceramics. This non-destructive technique provides constraints on the mineralogic make-up of both ceramic bodies and coatings. Using a combination of IR, X-ray diffraction, and inductively coupled plasma analysis, we examine select samples from Tell Brak, Syria, from three periods, the Halaf/Ubaid, Akkadian and 2nd millennium, in terms of their mineralogy and thus firing temperature. Utilising this technique, ‘‘local’’ and ‘‘imported’’ categories of samples are proposed, and developments in firing technology can be identified.  2000 Academic Press Keywords: INFRA-RED SPECTROSCOPY, MESOPOTAMIAN CERAMICS, TELL BRAK, X-RAY DIFFRACTION.

Introduction

few samples may be examined when compared to field observations based on considering body colour. Neither SEM analysis nor colour classification can be considered ideal from a quantitative standpoint, as each are subject to bias. What one researcher observes may not accurately correlate with another. However, both methods—if appropriately applied—can be used to good effect. The former is ideal for field characterisation of a huge number of samples. The latter method, while limited by financial constraints, is suitable for investigations on a limited scale. An analytical method that is both inexpensive and not prone to bias is ideal. Infra-red spectroscopy (IR) is well suited to the analysis of archaeological ceramics, as it offers a fast, inexpensive and quantitative measure of a number of parameters. IR characterisation, which is commonly used on modern ceramics, provides a non-destructive, spatially selective probe of ancient ceramic properties. In essence, the infra-red technique shares a commonality with the system of characterising samples using the Munsell colour chart in that the colour, or spectral signature, of the sample is characterised at wavelengths of 1–200
m. For comparison, visible light lies at 0·4–0·7
m. Because IR involves the complete spectral signature of a sample across this wavelength band, it is considerably easier to quantify than the Munsell method. Here, we characterise a ceramic sequence using IR (Appendix 1). These samples, from Tell Brak in northeast Syria, span from c. 5000 –2nd millennium  (Eiland, in preparation). Replication experiments using local materials (Eiland, in press) have been carried out, and a more detailed analysis of the results will be presented elsewhere. An exhaustive presentation of the

T

here are a number of methods that may be used to constrain the firing environments of archaeological ceramics, but few are commonly used (e.g. Rice, 1987; Orton et al., 1993). Most field classification techniques, which can be conducted with a relatively short ‘‘analysis’’ time, involve considering the colour of the sample. The Munsell colour chart has been used for a number of years (Matson, 1971), particularly for archaeological ceramics from Mesopotamia, modern day Syria and Iraq. This region has a long ceramic history, with many sites yielding literally thousands of sherds. A rapid and inexpensive analytical technique is therefore required. While the Munsell classification scheme is useful, it has a number of drawbacks. Perhaps the most serious is that refiring experiments with local clays are hardly ever performed to check the accuracy of colour/firing temperature correlations. As a result, data relating to the colour of the sherds are difficult to accurately calibrate. In an effort to quantify firing temperature, the scanning electron microscope (SEM) has been used (e.g. Maniatis & Tite, 1981; Tite, 1992). Studies using this method have considered a range of archaeological ceramics, both in an ‘‘as-received’’ state and after refiring to set temperatures in oxidising and reducing atmospheres. Using a visual classification scheme, the microstructures of clay samples are classified into various states of vitrification, and firing at progressively higher temperatures in different environments yields information about ancient firing practices. Unfortunately, this method can be both time consuming and prohibitively expensive. As a result relatively 993

0305–4403/00/110993+14 $35.00/0

 2000 Academic Press

994

M. L. Eiland and Q. Williams

archaeological context of the Tell and its finds is in preparation (Matthews et al., in preparation). These results not only demonstrate the utility of this technique, but also show distinct shifts in firing conditions over time. In some instances, imported wares are also clearly distinguishable from local wares based on the use of distinctive raw materials. In short, we illustrate that IR, particularly when combined with chemical analysis, SEM, and X-ray diffraction, can provide insights into both ancient ceramic technologies and the amount of trade in ceramic materials over time.

Infra-red Spectroscopy Applied to Archaeological Problems IR has been applied to several archaeological problems, although most are ‘‘conservation based’’. The compositions of pigments used in various Medieval manuscripts has been investigated (Orna et al., 1989), and the technique has been successfully applied to identify the pigments used in historical paintings from many periods and places. Glass corrosion has also been the subject of particular study (Sanders et al., 1972; Cooper et al., 1993). Gillard et al. (1994) examined mineralised textile remains using IR spectroscopy to determine a suitable method of preservation, and Weiner et al. (1993) examined archaeological bone to determine the degree of remineralisation. IR has been used to determine the nature of resinous deposits adhering to pottery (Sandermann, 1965; Sauter, 1967). Deposits on the inside of a jar from Godin Tepe, Iran, dating to the mid 4th millennium  (Badler, 1996), and jade from Mesoamerican and Central American sites have also received attention (Curtiss, 1993). The identity of pigments used in rock art from the Pecos region of Texas has also been explored using the related technique of Raman microscopy (Edwards et al., 1988). While the relatively non-destructive nature of IR has been exploited in these studies, the potential for systematically analysing a large number of samples was not emphasised. Amber is the archaeological material that is most closely associated with IR, and it is here that an archaeological methodology using a number of samples spanning a wide temporal range is used. Using IR spectroscopy (Beck et al., 1964, 1965; Beck, 1986; Beck & Shennan, 1991), it was found that the amber trade was much more complicated than previously assumed. The identification of IR with studies of amber in particular, and resinous materials in general, may have discouraged its use for characterising other archaeological materials. Methods used in provenancing archaeological ceramics are often based upon national preferences (Wilson, 1978). As archaeologists usually request that a particular method be used to characterise a sample, it is unlikely that a new technique, with no easy comparisons available from other

sites, would be applied (e.g. Pollard & Heron, 1996). Therefore, infra-red studies of archaeological samples are in their infancy. This is in contrast to the materials science community, which routinely uses spectroscopic techniques to characterise oxide samples.

Infra-red Spectroscopy: the Technique Infra-red spectroscopy, and the complementary technique of Raman spectroscopy, have been extensively utilised to characterise modern industrial ceramic materials, as well as a broad range of inorganic, organic, and mineralogic specimens. A large number of reviews of the applications and theoretical aspects of infra-red spectroscopy exist (e.g. Herzberg, 1945; Farmer, 1974; Velde, 1977; Nakamoto, 1978; McMillan & Hofmeister, 1988; Mitra, 1989; Williams, 1995; McMillan et al., 1996), and we provide only an abbreviated description of the technique here. The usefulness of infra-red spectroscopy in characterising archeological ceramics lies primarily in its ability to provide a ‘‘fingerprint’’ spectral pattern that can be readily associated with the mineralogic constituents of samples. Essentially every mineral of importance to ceramics has had its infra-red spectra characterised (e.g. Farmer, 1974; Salisbury et al., 1991), and these spectra fall into different groupings depending on the type of bonding within the different constituents of the ceramic. For example, carbonates are readily distinguishable from framework silicates (such as quartz and feldspars) which in turn are easily distinguished from chain silicates (amphiboles and pyroxenes). The overall mineralogic make-up of samples can be used to provide constraints on the firing temperature of the ceramics, an effect noted by Maggetti (1982). Additionally, infra-red spectroscopy can be conducted in a relatively non-destructive, spatially localised manner, and inclusions, coatings and glazes can be characterised. Therefore, infra-red spectroscopy can be rapidly used to determine if ceramics (or coatings) have similar initial compositions and firing histories. In this sense, infra-red spectroscopy can be used as a tool to demonstrate the homogeneity or the degree of heterogeneity of a suite of coeval archaeological ceramics, and thus the relative abundance of imported wares. The simplest underlying principle of infra-red spectroscopy has long been used in the archaeological sciences; that is, characterising the colour of a sample as a classification tool (Matson, 1971). Infra-red spectroscopy involves characterising the colour of samples in the wavelength range of 1–200
m, or equivalently 10,000–50 cm
1 in units of frequency. For ceramic characterisation, we focus primarily on the frequency range between 500–2000 cm
1 (20 to 5
m), as this spectral region proves to be the most diagnostic of different ceramic types. This range of infra-red light is at comparable energies to those of inter-atomic vibrations of solids, liquids, and gasses.

Infra-red Spectroscopy of Ceramics from Tell Brak, Syria 995

The energies of these vibrations are determined by the strengths of bonds within a sample, and the masses of the atoms. Stronger bonds produce higher frequency vibrational bands, as do less massive atoms. The local symmetry surrounding the atoms (particularly in crystal structures) also plays a major role in determining which types of vibrations are infra-red-active within a crystal. For example, calcite (CaCO3) has eight symmetry-allowed infra-red-active vibrations (e.g. Williams, 1995), with three of these lying above 600 cm
1 (typically, spectra of mixed phase materials such as ceramics are not notably diagnostic below 600 cm
1). Of these three, two (near 1400 cm
1 and 880 cm
1) are strong and completely diagnostic of the presence of calcite within the sample (Onomichi et al., 1971; Williams, 1995). These two vibrational bands, the asymmetric stretch and out-of-plane bend of the carbonate (CO2
3 ) unit, respectively, appear in many of the spectra shown in this investigation. These bands are completely distinct from vibrations of silicate phases, and therefore are unambiguously attributable to the presence of a carbonate in the sample. In practice, infra-red spectra can be collected through either absorption or reflection of infra-red light from a ceramic sample. Infra-red absorption measurements of ceramics are often conducted by dispersing powdered ceramic material into a pressed salt matrix (usually potassium bromide) at a level of 1 part in 1000 to 1 part in 100 (Russell, 1974), with typical powder sample sizes of 0·01 mg. For comparison, reflectance measurements can be conducted on a bulk sample, such as a sherd. A flat, polished surface is optimal for such measurements (e.g. Bliss et al., 1990). We found that reasonably high quality and completely reproducible data can be produced from flat regions of sherds (or sherds oriented so that their curvature is small in the region being sampled) in areas as small as 50
m in diameter. In essence, a sherd represents a polygranular compact of crystalline (and often some vitreous) material, and the spectra measured are thus derived from a diffuse reflectance regime (Aronson et al., 1966; Hopfe et al., 1993). It is of considerable importance that the characteristic grain sizes in the bulk (inclusion-free portions) of archaeological ceramics are generally smaller than the wavelength of the incident infra-red light, as shown by SEM observations, described in Appendix 2. Large grains of material can bias the observed reflectance spectra (Bliss et al., 1990), and we occasionally observe such effects when, for example, reflectance spectra are taken near macroscopic inclusions. Reflection experiments sample only the outermost 10
m or so of a sample, while absorption experiments sample the bulk vibrational properties of the powdered material. Accordingly, clean surfaces are crucial in conducting reflectance experiments. In many instances, we found that small amounts of scraping of the surface were necessary to remove dirt or surface alteration products from the sample. In the majority of cases the

uppermost 100–500
m were removed. Because of the spatial selectivity of the infra-red reflectance technique, such scraped areas could be confined to spots less than 1 mm2 in area. The subsidiary advantage of the surface sampling of the infra-red reflectance technique lies in the ability to characterise glazes and washes on sherds comparatively non-destructively, and distinctly from the underlying body of the ceramic. The increased utility and spatial selectivity of the infra-red reflectance technique can largely be traced to technologic advancements in infra-red spectroscopic techniques over the last two decades. In detail, polycrystalline ceramics are not particularly reflective in the infra-red region. Reflectivities at even reflectance peaks are typically only about 1% that of a metallic mirror. These low reflectivities are consequences of both the material properties of oxides, as well as the polycrystalline (and unpolished) surfaces of the ceramics. Accordingly, to conduct spatially selective measurements, in which reflected light is collected from areas with lateral dimensions as small as 50
m, requires that infra-red light be extremely efficiently collected and detected. We utilise a Fourier Transform infrared (FTIR) spectrometer coupled with an infra-red microscope to conduct all of our measurements. The infra-red microscope is equipped with Cassegrain all-reflecting optics (reflecting optics are required for IR microscopes, as conventional lenses absorb infrared light). Specifically, we utilise a Bruker Industries IFS-66v evacuated FTIR spectrometer, equipped with a KBr beamsplitter and glowbar source to conduct all of our measurements. The reflectance measurements are conducted on an infra-red microscope attachment (also Bruker Industries) equipped with a liquidnitrogen cooled intermediate range mercury cadmium telluride (MCT) detector. All spectra were run with 4 cm
1 resolution, with typical collection times of approximately 1–3 min for each reflectance spectrum. The spectra were referenced against a front-silvered aluminum mirror (which has a flat spectral response across the frequency range of interest) to remove instrument response from the spectra. In most cases, multiple spectra were collected from each sherd to both ensure reproducibility and to verify that the spectra were not aliased by a shallow subsurface inclusion.

Constraints of Firing Temperature The precise mineralogy of samples is known to reflect their firing temperature (Maggetti, 1982). In particular, the absence of carbonate is indicative of firing conditions exceeding 880C, as above this temperature calcite converts to CaO (lime), and CO2 (Southard & Royster, 1936; Wyllie & Huang, 1976). This temperature of 880C is a rather firm upper bound of calcite stability. The presence of silica is known to depress the decarbonation temperature. For example, Maggetti

996

M. L. Eiland and Q. Williams

(1982) observed the disappearance of carbonates in fired illitic calcareous clays near 820C. As almost all Mesopotamian clays contain carbonate (Schneider, 1994), most samples that do not contain carbonate have likely been fired to temperatures sufficient to decarbonate them. Additionally, the amount of vitreous material is invariably augmented at higher firing temperatures. Furthermore, the presence or absence of different clays and their resultant high temperature mineral assemblages can be used to infer maximum firing temperatures, such as, for example, the higher temperature thermal transformations of kaolinite (Roy et al., 1955). After examining an extensive range of archaeological ceramic samples, we found that IR may elucidate the composition of the original clay. As a simple example, pure kaolinite [Al2Si2O5(OH)4], a major component of many ceramics, undergoes a sequence of mineralogic transformations as it is treated at progressively higher temperatures. Specifically, kaolinite dehydrates near 550C to an anhydrous phase with a chemical composition of Al2Si2O7 (metakaolin). This phase alters to a mixture of SiO2 (initially amorphous, but which can ultimately crystallise to quartz or cristobalite) and a structurally complex and defect-rich spinel phase with a stoichiometry such as Al21·33O32 at 950–1000C (e.g. Carty & Senapti, 1998). This spinel phase reacts in turn to form mullite (Al6Si2O13) near 1075C (Carty & Senapti, 1998). The temperature of onset of mullite is in complete accord with that observed in natural samples (Maggetti, 1982). The critical point raised by the transformation sequence undergone by this simple clay is that even though a sample containing a kaolinite-rich starting material undergoes no changes in its chemistry (other than the loss of water near 550C) during firing, the resultant mineralogy of the sample shifts markedly as a function of the temperature to which the sample has been exposed. Thus, kaolinite starting materials converge at high temperatures to SiO2-bearing assemblages with aluminumbearing accessory phases, and this progression can be monitored—and firing temperature bounded—through mineralogic characterisation. Clearly, when more chemically complex clays such as smectites and illites are present, the range of minerals that can result during firing is markedly expanded (e.g. Kingery, 1976).

X-ray Diffraction In addition to IR, we conduct powder X-ray diffraction (XRD) on a number of our samples to both provide cross-checks on the sample mineralogy we derived from infra-red spectroscopy, and to provide an independent determination of sample mineralogy in cases that are difficult to characterise using IR alone. These measurements were conducted using a Norelco goniometer and X-ray generator using copper K radiation. The XRD patterns provide a diagnostic

means of determining which minerals are present within a sample. Powder XRD, as used here, provides more diagnostic information on sample mineralogy than infra-red spectroscopy, but has a few disadvantages. It is not sensitive to any amorphous material present within samples, and has difficulty detecting extremely finegrained materials. Mullite in high temperature ceramics represents a traditional example of this phenomenon (Roy et al., 1955; Bimson, 1969). It requires somewhat larger quantities of powder (typically at least tens of milligrams) than powder infra-red, and it can be more time-consuming than infra-red reflection. Typical laboratory based instruments do not have the spatial selectivity associated with infra-red reflection. However, infra-red spectroscopy and X-ray diffraction represent highly complementary techniques in the analysis of ceramic samples. The former can rapidly yield diagnostic ‘‘fingerprints’’ of different mineralogies, while the latter provides critical information on the actual minerals present within the sample.

The Halaf (5000–4500 BC) and Ubaid (4500–3500 BC) Periods The Halaf period is named after the type of site of that name in northern Syria. While these ceramics are not the earliest from Mesopotamia, Halaf wares are known for their fine fabrics. On the basis of several programs of chemical analysis (e.g. Davidson & McKerrell, 1980), it is likely that these ceramics were traded. Problems in distinguishing between a ‘‘home’’ and ‘‘workshop’’ industry directly influence the interpretation of this pottery. On the basis of excavations at Arpachiyah, Mallowan & Rose (1935) proposed finding evidence of a potter’s shop, with shelves, bone tools, and red ochre. Halaf period kilns have been recovered from several sites, including Yarim Tepe (Merpert & Munchaev, 1973) and Tell Hassan (Quarantelli, 1985). The surfaces and probable degree of vitrification of the Tell Brak samples are consistent with kiln firing. To date there have been few excavations to specifically target this period. As a result, there is much uncertainty around how to identify coarser wares, if indeed they make up as large a part of the ceramic assemblage as during later periods. Tell Aqab in northeast Syria records 40% plain wares (Davidson & Watkins, 1981), while Girikihaciyan in southeastern Turkey had 87% plain wares (Watson & Le Blanc, 1990). The Ubaid period has also received specialist attention (e.g. Oates, 1984; Curtois & Velde, 1991) and again much emphasis is placed upon the technical achievements of this early pottery. All the Halaf/Ubaid samples from Tell Brak are clearly of the finer fabric. No coarser wares from this period were identified.

Infra-red Spectroscopy of Ceramics from Tell Brak, Syria 997

The Akkadian (2350–2100 BC) Period The most striking aspect of ceramic production from this period is the large number of vessels of a similar colour and fabric. Examination of excavated material suggests that most sherds average yellowish-grey in colour. As a rough estimate only about 5% of the assemblage are moderate reddish-orange. The largescale workshop industries that existed during the Akkadian period had a great impact on the nature of the ceramic assemblage. Assemblages with ‘‘workshop’’ ceramics may obscure ‘‘household’’ scale representation, and may signal a change in the way ceramic shapes may be appreciated as ‘‘ethnic indicators’’. Thus there may be blurring of local ceramic traditions into a ‘‘national’’ style. In thin section (and under artificial light) samples with this ‘‘green’’ colour are moderate yellow– brown (10YR 5/4) and light olive brown (5Y 5/6). The ubiquitous green fabric wares of this period differ in many respects from other wares. One of the most notable features is that these wares have a very fine body fabric with a thin body wall. With such a fine fabric, and essentially no organic temper, a thin body wall is required. During firing water can escape in the beginning stages, and other gasses can pass through the thin body at higher temperatures. Vessels with a similar fabric continue into the 7th century , and are referred to as ‘‘Palace Wares’’ (Rawson, 1954). An important consideration for vessels of this period lies in whether they were made on the site. There is good evidence for the large-scale production of ceramics at this time. The site of Umm al-Hafriyat near Nippur offers some of the best evidence. At least 500 kilns (and bread ovens) from the Akkadian through Old Babylonian periods were uncovered. The kilns were located next to the beds of canals, and were associated with pits for levigation (Moorey, 1994). This large site for ceramic manufacture (along with the allied trade of baking) was dependant upon easy access to transport. The large-scale production of Akkadian ‘‘green’’ vessels is also attested from Tell Leilan, Syria. Fused, nested stacks of 50–65 fine ware bowls were common on the surface of the site. Blackman et al. (1993) examined a select group of wasters to characterise the raw materials and methods of manufacture. They found that the clay used to make the wasters was from the same clay source, which showed little chemical variation, which is not surprisingly reflected in the large number of standardised shapes recovered from this site. This kind of workshop production is consistent with the homogenous nature of Akkadian period vessels recovered from Tell Brak, yet there were no wasters recovered from this period. Perhaps the particular site of manufacture of these vessels was not near the region of habitation. Due to the smoke from kiln firings, and possibly the availability of clay as well, the ‘‘potters’

quarter’’ for such large-scale production could be located in a small site on the periphery of the Tell. Wasters dating to the 2nd millennium were recovered from trench HN, indicating ceramic production on the Tell during this period.

2nd Millennium There are trends that quickly separate 2nd millennium samples from Akkadian samples. The most notable feature is the lack of standardised fabrics, such as the ubiquitous green wares from the Akkadian period. As the preceding period is well known for large-scale industrial control, while the 2nd millennium can be (more or less) characterised by the predominance of smaller states—at times influenced by ‘‘nonMesopotamian’’ peoples—this may come as little surprise. In general, however, the ceramics from this period do have distinctive shapes. Perhaps as a result of problems with defining this period archaeologically, there has been little attention paid to the ceramic technology of this period. Calcite is the most common mineral temper during this era. Most samples appear low fired so that the calcite does not influence the matrix, with a few higher fired such that it interacts with the clay. No sample had fine to medium sand size grains of any other minerals used as a temper. This is a significant departure from the general arkose sandstone parent rock used for so long by Tell Brak potters. The carefully self-slipped surfaces characteristic of ‘‘cooking’’ vessels of this period raise a number of interesting points. Cooking vessels from earlier periods usually exhibit a smoothed surface, but there is usually no effort to make such a thick and presumably impermeable layer. The major question at this juncture is why would a vessel be slipped? It is often assumed that vessels would soon be fouled by foods absorbed by their porous ceramic bodies, and that vessels were either disposed of at a regular rate, or that certain vessels were used for specific foods. This raises questions about the eating habits of the 2nd millennium, as perhaps a vessel’s slipped surface would be effective in providing a barrier to oil and other particular foodstuffs. At the same time it is clear that with a decrease in firing temperature and differing clay recipe, 2nd millennium vessels could not offer as effective a barrier to liquids as their Akkadian counterparts.

Results One of the primary advantages of infra-red reflectance spectroscopy lies in its ability to rapidly and comparatively non-destructively characterise genetic similarities and differences between samples. For example, Figure 1 shows representative reflectance spectra from sherds from the Halaf and Ubaid periods. Sherds from these periods are characterised—with the sole

998

M. L. Eiland and Q. Williams

BK 1

Reflectance

BK 2

BK 3

BK 4

BK 5

1400

1200

1000

800

600

Wavenumber (cm–1) Figure 1. Infra-red spectra of Halaf/Ubaid samples. The strong increasing baseline above 1200 cm
1 in BK 1 and BK 2 is likely associated with particle size effects. With the exception of BK 4, all spectra show evidence of carbonates, as shown by the appearance of peaks above 1300 cm
1. Lower frequency features between 800 and 1100 cm
1 are associated with silicates.

exception of BK 4 (which is the second line from the bottom in Figure 1)—by carbonate stretching peaks near 1300–1500 cm
1, and three components between 950 and 1150 cm
1 in the spectral region associated

with silicate (SiO4-group) stretching vibrations. A weaker lower frequency band also is present near 900 cm
1, which overlaps with the carbonate out-ofplane bending vibration near 870 cm
1 in samples BK 1, BK 2, BK 3 and BK 5. This feature near 900 cm
1 is likely associated with silicate stretching vibrations of a glassy portion of the sample. Peaks near this frequency are observed in reflectance spectra of silicate glasses (e.g. Salisbury et al., 1991). The carbonate out-of-plane bend is a comparatively sharp vibrational bond, which silicate stretching bands in this spectral region tend to broaden. For example, the spectrum of BK 4 in Figure 1 shows no evidence for carbonate in either the asymmetric stretching region above 1200 cm
1 or near 870 cm
1, yet a broad silicate stretching feature is visible slightly below 900 cm
1. Within the spectra of BK 1 and BK 2, the carbonate band near 870 cm
1 likely overlaps a silicate stretching band. This combination produces the broad, nearly split feature between 850 and 900 cm
1 in the spectra of BK 1. The variable amplitudes of the stretching vibrations between 950 and 1150 cm
1 are likely produced by moderate shifts in the vitreous fraction of the samples relative to the amount of crystalline quartz present. SiO2-quartz is confirmed to be present in these samples from the X-ray diffraction results shown in Figure 2, which demonstrates that the primary crystalline phases present within BK 5 are quartz, calcite and feldspar, with the broad background between 18 and 34 being produced by the vitreous portion of the sample (e.g. Bish & Reynolds, 1989). This diffraction pattern is discussed in more detail below. The higher frequency components of the silicate peaks (at 1020–1150 cm
1) in the infra-red spectra are associated with SiO2-quartz (e.g. Scott & Porto, 1967; Moenke, 1974; Salisbury et al., 1991; Williams et al., 1993), while the components between 950 and 1030 cm
1 are most probably produced by the vitreous fraction within the ceramic. Comparable peaks are observed in the infrared spectra of aluminosilicate glasses (e.g. Day &

1500

1500 (a)

(b)

1000

BK 5

Counts

Counts

1000 BK 5 500

500

BK 54

BK 54 0

10

20

30 40 50 Diffraction angle (°)

60

0 20

25

30 35 Degrees

40

45

Figure 2. (a) X-ray diffraction patterns of BK 5 (Ubaid) and BK 54 (Akkadian). All prominent peaks can be attributed to quartz, calcite and feldspar, while the broad background between 18 and 34 is produced by a vitreous fraction within the sample. (b) Expanded view of the diffraction patterns of BK 5 and BK 54, emphasising the primary crystalline diffraction peaks. This angular range is presented primarily for comparison with the results of Figure 4.

Infra-red Spectroscopy of Ceramics from Tell Brak, Syria 999 Table 1. The Inductively Coupled Plasma (ICP) analysis was performed at Royal Holloway, University of London, Geology Department. The particulars of this method will be found in many introductions to archaeological science (Pollard & Heron, 1996: 31–36).The results for Al2O3, Fe2O3, MgO, CaO, Na2O, K2O, TiO2, P2O5 and MnO are quoted as weight per cent oxides. One of the limitations of the ICP technique is that silica (SiO2) is not typically measured. It is usually calculated by adding together all analysed elements and subtracting that value from 100%, as has been done here

BK1 BK2 BK3 BK4 BK5 BK14 BK15 BK21 BK53 BK54 BK 55 BK59 BK 62 BK64

Al2O3

Fe2O3

MgO

CaO

Na2O

K2O

TiO2

P2O5

MnO

14·27 12·39 12·43 12·49 11·35 6·65 11·83 15·6 10·78 13·06 10·95 8·12 8·52 11·32

8·02 6·76 7·02 6·75 6·14 3·39 7·83 9·16 5·8 9·97 5·5 4·27 4·47 5·72

5·49 3·95 3·8 4·32 5·8 2·34 4·02 5·77 3·88 6·08 4·52 2·72 4·55 4·51

7·97 13·34 14·62 15·34 15·27 28·60 17·06 4·39 14·83 11·19 18·68 26·57 21·5 20·22

1·03 1·1 0·78 1·01 0·78 0·16 0·55 1·62 0·65 1·58 1·18 0·37 0·38 0·6

2·62 2·32 2·13 2·35 2·32 1·6 2 2·24 2·63 1·83 1·41 2·13 2·18 2·01

0·73 0·76 0·77 0·79 0·68 0·4 1·1 1 0·67 1·32 0·64 0·48 0·5 0·66

0·57 0·26 0·38 0·31 0·3 0·13 0·34 0·27 0·50 0·36 0·33 0·30 0·22 0·29

0·13 0·11 0·1 0·11 0·14 0·05 0·12 0·15 0·1 0·14 0·09 0·07 0·07 0·11

Rindone, 1962; Sweet & White, 1969). The variations in amplitude of these four silicate stretching peaks between the different spectra of Figure 1 likely reflect relatively minor variations in starting material(s) and/or firing conditions among these sherds. It is clear that in this case the IR measurements are sensitive to a different set of variables than either chemical or particularly petrographic analysis. The IR measurements are sensitive to the amount of vitreous material relative to SiO2. In contrast, petrographic analysis suggests that while minor differences in bulk chemistry may exist, these samples can be further divided on the basis of visible inclusions. Despite this, the clays used for all samples are broadly similar, which is confirmed by the chemical analyses of Table 1. Samples BK 2, BK 3, BK 4 and BK 5 are essentially chemically identical, while BK 1 is depleted in calcium, but enriched in iron, aluminum and magnesium relative to the other samples. By difference, all samples have between 56·9% (BK 4) and 59·5% weight per cent silica (BK 1). This difference implicitly assumes a minor weight percentage of carbon in these samples, as carbon is not quantitatively analysed. Given this near constancy in silica content, the similarity in the infra-red spectra of the silicate stretching vibrations of these sherds is not surprising. With respect to firing temperature, it is notable that all samples from this period appear to have large degrees of vitrification. Only BK 4 apparently has no carbonate present within it, as shown by both the lack of an 880 cm
1 peak and no significant amplitude in the 1300–1500 cm
1 range in the spectrum of this sample. The spectra of BK 1 and BK 2 also have a steadily increasing baseline above 1200 cm
1. This increase is associated with particle scattering effects, and is commonly observed in fine grained aggregates (van de Hulst, 1965). This effect is simply generated by the wavelength of incident

infra-red light being comparable to the grain size of the sample, and is similar to the effect that produces the vivid colour of oil films on water. The presence of such scattering in both BK 1 and BK 2 implies that these samples have similar grain sizes. According to SEM images, samples BK 3, BK 4 and BK 5 are characterised by coarser grain sizes, and show no such scattering of infra-red light. The spectra of Akkadian-era sherds are qualitatively different in their variability from those of the earlier samples (Figure 3). This figure shows representative spectra of different families of sherds. The spectra of samples BK 14 (cooking pot), BK 54 (basalt tempered vessel) and BK 55 (‘‘Akkadian green’’) are dominated by the presence of large quantities of carbonates, indicating that these ceramics are highly carbonateenriched relative to those of the earlier period (Figure 1). BK 53 (‘‘Sumerian’’ sherd) perhaps represents the most similar sample to those of the previous era, with some similarities between the spectrum of this sample and that of BK 1. This sample also has an estimated silica content that is similar to that of BK 1 (60·2% versus 59·5%), although we do not suggest that these two samples must necessarily originate from a similar region. X-ray diffraction confirms that the Akkadian basalt tempered vessel BK 54 (Figure 2(a), (b)) differs markedly in its mineralogic make-up from ceramics such as BK 14 and BK 59 (Figure 4). While quartz and particularly calcite are present in abundance in BK 54, the X-ray diffraction pattern is considerably more complex, with lines likely produced by a plagioclase-fraction within the sample, as demonstrated by the strong diffraction lines near 28 and additional peaks between 20 and 25. Notably, the infra-red spectral peaks of feldspars and quartz overlap in the silicate stretching region (Williams et al., 1993; Williams, 1998). Submicron grain sizes can produce

1000 M. L. Eiland and Q. Williams

BK 14

BK 15

Reflectance

BK 21

BK 53

BK 54

BK 55

1300

1100

900

700

Wavenumber (cm–1) Figure 3. Infra-red spectra of Akkadian era ceramics.

8000

Counts

6000 4000 2000

BK 14 BK 59

Quartz Calcite

0 20

25

30 35 Diffraction angle (°)

40

45

Figure 4. X-ray diffraction patterns of BK14 (Akkadian era) and BK59 (2nd millennium), compared with the diffraction patterns of mineral samples of pure quartz and calcite.

notable broadening of the infra-red spectral bands (Hadni, 1974; Salisbury et al., 1991). Accordingly, distinguishing between these two framework silicates in fine-grain-sised, multiphase ceramics is difficult based on IR results alone. In particular, the locations of the infra-red bands generated by silicate tetrahedra in quartz and plagioclase are grossly similar, and fine

particle sizes tend to broaden these features. The X-ray diffraction pattern of the Akkadian cooking ware, BK 14 (Figure 4), is far simpler, with abundant calcite being juxtaposed with a smaller quantity of quartz as the primary (and sole) crystalline mineral phases. This pattern shows essentially no broad background between 18 and 30, such as is generated by vitreous material (Bish & Reynolds, 1989). The net conclusion from the amplitude in the 18–34 region of the X-ray diffraction patterns is that BK 5 has a larger percentage of vitreous material than BK 54, and each of these have a substantially greater amount of vitreous material than BK 14 and BK 59 (2nd millennium), which are comparable in their relatively small amplitude background between diffraction angles 18–34. Moreover, BK 5 has a smaller concentration of feldspar, and higher concentration of quartz relative to BK 54. This is demonstrated by the smaller relative amplitude of the feldspar peaks near 29 in BK 5 relative to the same peak in the diffraction pattern of BK 54 (Figure 2(a), (b)). Accordingly, variations in feldspar and carbonate content appear to be the primary mineralogic discriminants between these different ceramics: feldspar content is largest in the basalt tempered ware (BK 54: Figure 2) of the Akkadian era, is lower in the Halaf/ Ubaid samples, and is vanishingly small in the calciterich cooking pot (BK 14: Figure 4). For comparison, pottery of the 2nd millennium (BK 59) has a feldspar content which approaches that of BK 14, and is similarly dominated by quartz and calcite as virtually the sole crystalline phases. The overall interpretation of the highly variable results from this suite of samples is that pottery of the Akkadian period from the Tell Brak assemblage can be characterised by the usage of different starting materials and recipes, unlike the relatively homogenous starting materials encountered during the Halaf/ Ubaid periods. Firing temperatures were likely lower for the majority of ceramics produced during the Akkadian period—although many of the ‘‘mass produced’’ wares show evidence of extensive vitrification—but many samples are carbonate-rich. Some of these shifts may be produced by differences in local manufacturing techniques; however, the magnitude of the variability strongly suggests that a larger component of imported pottery is present among these sherds. BK 21 (fine red bowl) was the only sherd of this ware recovered, with little doubt that this sample was imported. In thin section it was almost pure clay, and chemically it appears the same. BK 15 (painted ware) is also distinctive from what one can almost certainly assume to be the locally made BK 55 (‘‘Akkadian green’’). In contrast to the oxidised-fired wares, with a red colour, the majority of Akkadian period vessels were fired in a reducing atmosphere. Reduction firing in a kiln requires more fuel, as combustion is incomplete, but vitrification occurs at a lower temperature. Petrographic characterisation indicates that finely ground

Infra-red Spectroscopy of Ceramics from Tell Brak, Syria 1001

Reflectance

BK 59

BK 62

BK 64

1300

1100

900

700

Wavenumber (cm–1) Figure 5. Infra-red spectra of 2nd millennium ceramics.

calcite was added to the clay recipe, which is attested by several low fired samples where the calcite has not decarbonated. Adding calcite lowered the temperature of vitrification further, as well as changed the body colour from red to green. One can conclude that during this period—at the latest—Tell Brak potters understood the effect of fine grained calcite on matrix colour. Another advantage of firing in a reducing atmosphere is that vitrification is better controlled over a wider temperature range, from c. 850–1050C (Maniatis & Tite, 1981; Schneider, 1989). Without accurate means of determining temperature, this wide temperature range is significant. Most Akkadian vessels were fired to a high enough temperature to reduce porosity. At this critical point, it would be easy to over-fire a batch of vessels if there were not considerable latitude in firing temperature. Finally, Figure 5 shows representative spectra of samples from the 2nd millennium. These are generally quite similar to one another, with a broad, dominant silicate stretching band centred near 1020–1050 cm
1 in BK 59 and BK 62. It is likely that this peak is generated by extremely fine-grained SiO2-quartz. Finegrained materials typically have broadened spectral features relative to coarser-grained ceramics (e.g. Salisbury et al., 1991). These samples also appear to

have an intermediate carbonate/silicate ratio relative to the comparatively carbonate-poor Halaf/Ubaid samples (Figure 1) and some of the carbonate-rich Akkadian-era samples (Figure 3). The presence of carbonate is indicated by the absorption above 1300 cm
1 and strong absorption near 880 cm
1 in these samples. The X-ray diffraction results (Figure 4) on sample BK 59 are grossly similar in terms of the crystalline phases present to that of samples likely to have been produced locally. Quartz and calcite dominate the sample, with a minor feldspar component which produces the weak diffraction peaks between 27 and 29. Therefore, the 2nd millennium samples generally appear to be fired at lower temperatures than the Halaf/Ubaid samples, but their overall mineralogy (and chemistry: Table 1) is similar to those of the earlier samples. The comparison with the Akkadian-era samples is more difficult. Sample BK14 (cooking ware) alone appears to represent the closest proxy to these later samples. This may suggest that there was a shift in locally obtained clays for the ‘‘common’’ wares produced during the Akkadian and 2nd millennium. The latter samples show great uniformity in their chemical composition—unlike the samples from the Akkadian period—but in general have low Al fabrics that are consistent with natural clay (such as BK 59) or perhaps natural clay that has been slightly levigated (BK 64). This significant shift in composition from the Akkadian period probably does not reflect a major difference in raw materials, as would be the case if there were large scale imports from a distinctive source region. BK 59 (ring base bowl) would at first appear to be distinct. Despite the very high Ca content and the number of notably low ‘‘minor’’ element values, this sample does not differ greatly from the group as a whole. The net results of the infra-red and X-ray diffraction studies are rather simply summarised. Reasonable uniformity of ceramic materials existed in both the Halaf/Ubaid and 2nd millennium. This trend is manifested in the broad similarity between the spectra within Figure 1 and within Figure 5. These ceramics were likely local and are mineralogically dominated by quartz and calcite, with degree of vitrification appearing to have been higher in the late Neolithic. During the Akkadian period, some ceramics of affinity with both later and earlier samples are found [for example, BK 14 with 2nd millennium samples such as BK 59 (Figures 3, 4 & 5), and BK 53 with late Halaf/Ubaid samples such as BK 1: Figure 1 & 3], but a broader variability in both spectra and mineralogy is observed during this era, thus implying a greater abundance of imported wares from distinctive source regions.

Conclusions IR offers considerable advantages over other techniques. Potentially the most significant is the rapid

1002 M. L. Eiland and Q. Williams

assessment of a number of samples in a short period of time. Essentially no preparation is required for the majority of archaeological ceramic samples, and analysis time of individual samples is usually less than 2 min. Unlike other methods, such as ICP, IR can give information about the raw materials and firing history of ceramic samples, and the spectra can be quickly matched with similar spectra, making rapid assessment of numerous samples possible. Calcite is particularly reflective, and can serve as a useful diagnostic mineral to estimate maximum firing temperature. There is also an increasing demand for non-destructive methods of analysis. Museum curators are becoming less enthusiastic about releasing shards for destructive tests, while complete vessels have almost never been sampled. Infra-red spectroscopy, when used in reflectance mode, is non-destructive, and if a sample of powder is required, it can often be obtained from a broken surface of a sample. With further work to define the dependence of the method on grain size, starting materials, and firing conditions, this technique can be quantitatively applied to a wide range of materials from a number of sources. Our results demonstrate general mineralogic differences and affinities between ceramics of various cultures. Accordingly, dramatic shifts in firing technology and/or in preparation of raw materials as well as presence or absence of imported wares can be identified within the archaeological record. It is perhaps the change in raw materials over time, from source materials that in the main will have been locally collected, where the greatest interest lies. While many analytical programmes identify imports on the basis of similarity or difference with a control group, it is clear that particular attention must be paid to both the time period in question and the type of vessel. In the case of samples from the Ubaid period in particular, it has been suggested that unusually effective kiln design was a by-product of the development of copper metallurgy (Matson, 1955). While this may be true of the luxury wares so far identified, much more remains to be done with plain wares that may have made up the bulk of the ceramic assemblage from this period. But imported wares also have an important role in reconstructing ancient technology, as our results demonstrate that a larger suite of ceramic types was utilised in the Akkadian era than in either the preceding Halaf/Ubaid periods or the subsequent 2nd millennium. Moreover, samples of Akkadian ceramics appear to be characterised by large variations in degree of vitrification, as documented by the SEM results described in Appendix 2. This trend reflects both the mass produced tradition of centralised ceramic production, as well as that more limited production, perhaps from distant areas, was also a factor in trade. Several samples from the 2nd millennium show surprisingly little variability, suggesting that by this point, extensive trade may not have been a large factor. While the large

Figure 6. 102 BK 2 Halaf sample. This sample shows extensive vitrification, with regions with extensive glass formation that have completely obscured the matrix texture. There is slight evidence for bloating pores. The clay used to prepare this sample appears to be of a finer texture than that used for the majority of the Tell Brak samples, and is noticeably finer than the Ubaid period samples.

scale production of pottery continued in much the same way as before, the firing temperatures attained during the later periods appear to be lower.

Acknowledgements The Mineral Physics Laboratory at UCSC is supported by the NSF and the W. M. Keck Foundation. The authors would also like to thank the Tell Brak research team for providing samples from the 1994–1996 seasons for analysis. Thanks are also due to Chris Bendall and Helen McDonald for reading drafts of this paper.

Appendix 1 Catalogue For a more complete description of the wares see the Tell Brak Final Report (Matthews et al., in preparation). The first number recorded (abbreviated with the letters BK) refers to the chemical analysis. The numbers in parenthesis are the site numbers. Here, particular attention is paid to differences that were elucidated by thin section studies (abbreviated as TT). BK 1 (227.1). Halaf body sherd. Red surface is very rough, with irregular undulations from coarse inclusions, although no mineral temper is visible. TT: this sample is more in keeping with other wares recovered from the site. The dominant inclusion is fine sand sised quartz (7%). There is a small amount of calcite (3%), and a pellet of clay 2 mm long. BK 2 (242.1) (Figure 6). Halaf rim and body sherd. By eye the fabric is similar to the ‘‘Akkadian green’’ wares, with no visible inclusions. The reddish brown painted decoration on the outer surface is reminiscent of basketry. TT: this sample does not resemble the

Infra-red Spectroscopy of Ceramics from Tell Brak, Syria 1003

Figure 7. 94 BK 14. The Akkadian period cooking pot shows slight partial vitrification. The original matrix texture is largely intact. In thin section this sample also has a large quantity of calcite grains which show only slight evidence of heating. These large grains of calcite had essentially no effect on the vitrification structure, as would a similar quantity of fine grained calcite. The required temperature for decarbonation was clearly not reached.

Ubaid sample (BK 5) and has little affinity with the other Halaf sherd (BK 1). Coarse silt sised quartz (10%) predominates, with few grains of plagioclase. The amount of muscovite (5%) is the largest amount of any sample from the site. Voids (5%) in the fine sand range indicate calcite. There was grog in the fine sand size range (3%). The angular nature of the mineral inclusions indicates that the temper was a ground rock, but of a different origin from those used in the majority of other wares from Tell Brak. One may confidently consider this an import. BK 3 (750.1). Halaf rim/body sherd. No TT. The reddish fabric of this sherd has been coated with a bright red slip and burnished. There is a band of black wash on either side of the lip extending down 1 mm. By eye small fragments of calcite are visible. BK 4 (750.1). Ubaid period body sherd. No TT. Buff fabric of this sample is much like MU 4. The painted decoration is light, and the surface of the sherd is pitted, consistent with a long period of circulation. No inclusions are visible in the reddish/buff fabric except for calcite. BK 5 (757.1). Ubaid rim/body sherd. Buff self slipped fabric with distinct calcite inclusions. TT: 10% inclusions of fine sand size. The majority is calcite with about 2% quartz. There are several grains of shell, and one grain of amphibole (oxy-hornblende). The transformation of hornblende to oxy-hornblende takes place at about 800C when a sample is heated in air (Deer et al., 1992). BK 14 (1017.1) (Figure 7). Akkadian period rim/ handle sherd of a bulbous handled calcite tempered cooking pot. Outer surfaces self slipped. TT: calcite 15% average coarse sand size. This very distinctive

Figure 8. 96 BK 21. Fine red bowl from the Akkadian period. This sample has a distinctive ‘‘ring’’ when tapped against a hard surface. This image shows extensive vitrification. There is a fine network of bloating pores. This mixture had a relatively small amount of CaO (4·39%). Maniatis & Tite (1981) found that non-calcareous clays were, when compared to calcareous clays, unstable over a short temperature range. Their experiments demonstrated that, in a reducing atmosphere (as in this case), fine bloating pores were formed at 800–900C, while medium bloating pores were formed from 900–1000C. This suggests that while this vessel may have a structure that appears to be more affected by heat that the calcareous clays, it was probably fired within the same temperature range.

fabric contains the most calcite of any sample examined from Tell Brak. There is a small population (5%) of organic voids. Rounded quartz in the fine sand size range (1%) was apparently not added intentionally as temper. ICP analysis shows that this sample has the least ‘‘clay’’ and the most Ca out of all the samples examined from the site. This vessel was almost certainly imported. BK 15 (1098.1). Akkadian period body sherd— clearly wheel formed—with painted design. The painted surface was made by incising (using a tool 2 mm across) the red/brown coating away from select areas. TT: bimodal population of grains. 10% calcite is in the fine sand range, with fine sand sised voids (7%) probably from calcite. There are few grains of basalt and disaggregated plagioclase and pyroxene in the fine sand to medium sand range. Calcite has been crushed finer than the significantly harder basalt. ICP analysis indicates that the clay used to make this sample is consistent with local materials. BK 21 (1195.1) (Figure 8). Akkadian period rim/body sherd of a wheel formed red bowl. Outer surface is decorated with a fine brown/black paint or wash 21 mm down the outer surface of the vessel and 3 mm down the inside. The remainder has been carefully self slipped and burnished. By eye the fabric is very fine with no visible mineral inclusions. TT: very distinctive red fabric with 7% mineral inclusions, generally well sorted in the very fine sand range. At this size range the majority appear as ‘‘quartz’’, but 3% that

1004 M. L. Eiland and Q. Williams

Figure 9. 103 BK 55. This ‘‘typical’’ Akkadian fabric shows extensive vitrification throughout. It appears that a fine particle size assisted the regular sintering of the grains. Thin section studies also showed that this sample, in keeping with others of the type, contained a sizable fraction of burned out organic material. This likely promoted the release of gas from the fabric as it was formed during firing, and inhibited the formation of bloating pores.

are clearly plagioclase feldspar, and 2% biotite. Biotite is scarce in samples from Tell Brak. ICP analysis shows that this ‘‘high clay’’ sample is atypical of the Akkadian assemblage and represents a completely different set of materials than the other samples from this period. BK 53 (1074.1). Akkadian period rim/body sherd. Shape and decoration are consistent with a southern (or ‘‘Sumerian’’) origin. Decoration required two different implements for wavy lines (four tined implement), and straight lines (seven toothed tool). By eye the fabric is lighter (a light buff) than most of other fabrics recovered from this period. TT: reddish/green fabric is atypical, and there is a large amount (3%) of very fine sised hematised grains. There is fine sand sised quartz (7%). Calcite (5%) survives in the middle of the sample. Isolated regions that have experienced less heat show a large number of fine sand sised calcite grains. There is 10% organic voids. BK 54 (1015.1). Akkadian period rim/body sherd of a basaltic tempered vessel, probably paddle formed. Inside of the sherd bears carbon deposits. TT: profusion of basaltic grains. This is unlike AK 8, which only has small amounts of basalt. The fragments (15%) are poorly sorted, average fine sand sised grains up to isolated examples 2 mm. Long Alkali basalt consists of laths of plagioclase feldspar, titanium aurgite, and olivine. There is no evidence that grains of a particular size were desired. There is no trace of calcite added intentionally, but there is a small amount of fine sand to silt sised grains associated with the clay. The chemical analysis also indicates that this vessel is imported. BK 55 (1070.1) (Figure 9). Akkadian period sherd of a fragmentary wheel formed base, 5 cm in diameter. TT: green fabric with large number of voids (10%), at

Figure 10. 99 BK 64 2nd millennium wheel formed bowl. The image shows partial vitrification. While the matrix is largely intact, the finer population of grains have clearly been sintered. While sand sised calcite grains have undergone heat damage, much calcite still remains (as is confirmed by thin section studies). It seems that, unlike the Akkadian period samples, where silt sised calcite was used to lower the sintering temperature, potters of this period made lower fired ceramics with different properties.

least 3% are from calcite. There are silt sised grains of quartz associated with the clay (1%). The margins of the sample are vitrified so that they do not transmit light, while the core is not so well fired. The core contains silt sised grains of calcite. There is good evidence from other 2nd millennium samples that the majority of calcite was added in the silt size range, and this evidence disappears when the vessel is fired to a high temperature. BK 59 (254.1). Second millennium rim/body sherd of a wheel formed ring base bowl, outer surface smoothed, washed, and burnished. It is made of typical ‘‘cooking ware’’ fabric. TT: calcite predominates (15%) poorly with a size range from fine sand to grains 1 mm long. All grains show little heat damage. The clay is associated with a sizable amount of silt that indicates a low firing temperature. This sample has effectively no porosity from organic material or forming. BK 62 (1.1). Second millennium rim/body sherd of plate with a simple rim and a semi-circular lip of ‘‘Mitanni’’ type. Reddish wheel thrown fabric, inner surface has been smoothed and self slipped. TT: calcite (10%) in the fine sand size range, and shell (3%) averaging 1 mm long. The matrix contains a large fraction of silt. The ‘‘organic temper’’, visible by eye appears to be shell that has been degraded by heat, although there could be also be a small fraction of vegetal material. BK 64 (142.5) (Figure 10). Second millennium rim/body sherd of a light buff wheel thrown bowl with an angular rim, light well fired fabric, and burnished surface. TT: similar to the other two second millennium samples except that calcite makes up about 13% of slide.

Infra-red Spectroscopy of Ceramics from Tell Brak, Syria 1005

Appendix 2 SEM images SEM micrographs, taken at low magnification, give an indication of firing temperature in antiquity, and provide an important cross-check to the IR spectra. The long depth of focus on these freshly fractured surfaces demonstrates the amount of glass (vitrification) formed during firing. Previous studies of SEM images of archaeological ceramics (e.g. Maniatis & Tite, 1981; Tite, 1992) have divided samples into discrete groups based on visual characteristics of the matrix, and recorded the temperatures at which vitrification structures are attained. These authors paid particular attention to calcite which, if dispersed throughout the sample in fine grains, led to a stable vitrification structure (in calcareous clays) when fired from 850–1050C. Both oxidising and reducing atmospheres led to a similar temperature range (and therefore similar vitrification structures). On the basis of examination using a petrographic microscope, it is clear that a number of voids in the samples are from burned out organic material, and not necessarily from calcite. As most ceramics from Tell Brak were fired to near 900C, there is essentially no organic material remaining that could be used for accurate identification of this material (Johnson et al., 1988).

Terminology and temperatures (for calcareous wares) No vitrification: individual grains with no melt phase,