Towards an Integrated Approach to Studying the Strati ﬁ ed Ceramics from Dandanakan/Da ş Rabat, Turkmenistan (9th – 12th Centuries A . D .)

This study draws on archaeological, stylistic, and technological evidence to explore ceramic and brick production of the medieval Islamic period in the southern Karakum region in Turkmenistan, home to many urban sites along the so-called Silk Roads. We focus on a 9th – 12th centuries A . D . assemblage recovered from the site of Dandanakan/Da ş Rabat during the ﬁ rst season of ToKa (Town of Karakum project) in 2019. Special emphasis is paid to characterizing the local ceramic fabrics and ceramic technologies through macroscopic examination and petrography, SEM-EDS, and FTIR analyses. Our results show that unglazed and glazed earthenware were manufactured using two local or regional clay outcrops, also employed in the brick kilns detected outside of Dandanakan ’ s city walls. A di ﬀ erent clay was used for the slip of the glazed earthenware. These all had high lead-silica glazes, except for the turquoise glazes detected on both earthen-and siliceous wares.


Introduction
In the early and mid-Islamic periods, the regional metropolis of Merv and its satellite towns saw a renewed political, intellectual, and economic relevance.The ceramic manufacturing to serve such large economies must have been vast, but research on it remains scarce and disconnected, with notice only of kilns at Merv (Zaurova 1958;Herrmann, Kurbansakhatov, and Simpson 1999, 13-15) and in smaller settlements such as Kushmeihan (YUTAKE 1966).Owing to a western focus on the Iranian world and the attention paid to the so-called Silk Roads, scholars have tended to overplay the role of the transregional trade of ceramics, especially as kilns were found further west in Khurasan at Nishapur (Wilkinson 1973) and east in Mawarranahr at Afrasiyab (Shishkina and Pavchinskaya 1992, 35-45), Termez (Fusaro et al. 2022), and Chach/Tashkent (Burjakov and Filanovich 1999).Meanwhile, siliceous wares (stonepaste) are assumed to have been traded from Iran because of extensive evidence of production there and unclear or unpublished data from Turkmenistan, Uzbekistan, and Afghanistan (Lunina 1962, 255;Nemtseva 1969, 194-203;Siméon 2017, 24-25).The recent surge in technological studies of ceramics of the medieval Islamic period has included research on Central Asia but almost exclusively on Mawarrannahr and Semirechye (on glazed wares: Henshaw et al. 2006;Henshaw 2010;Matin, Tite, and Watson 2018;Klesner, Akymbek, and Vandiver 2021;Klesner et al. 2021;also unglazed: Klesner et al. 2019;Martínez Ferraras et al. 2020;Molera et al. 2020;Fusaro et al. 2022).As for the region of the Murghab and Tejen River basins, technological aspects of local medieval ceramic production are still only partially understood.
In this paper, we present an attempt to develop an integrated approach to examining the ceramic assemblage excavated at Daş Rabat, the modern toponym for medieval Dandanaqan (henceforth "Dandanakan").Survey and excavation of the site were carried out in 2019 within the framework of the Towns of Karakum Archaeological Project (ToKa), an international collaboration with the goal of investigating the role played by urban stations in the routes that traversed Khurasan during the 9th-12th centuries A.D. (Wordsworth et al. 2022).The assemblage under study consists of unglazed and glazed wares and fired bricks.The aims of this pilot study are to identify the ceramic fabrics that are representative of the Islamic period assemblage in Dandanakan from a macro-and micro-perspective, which will then allow us to explore local technical traditions.Looking into the future, we will use this data to contextualize Dandanakan ceramic production within the broader technological developments in Khurasan and Mawarrannahr during the 9th-12th centuries A.D.

Archaeological and Geological Contexts
In the 19th century A.D., the local toponym Daş Rabat (transcribed as "Ташъ Рабадъ") was marked on Russian Imperial maps on the desert route which crosses from the Tejen River to the Murghab River (Il'ina 1891).Its position, to the south of the first east-west Transcaspian Railroad (completed in A.D. 1886), demonstrates the shifting axis of travel in the late 19th century A.D. away from connections with the Qajar empire and the prioritization of routes to Ashgabat and the Caspian (Figure 1).Owing to its remote location, surrounded by desert, its significance only came to light during the Great Patriotic War (World War II) when archaeologists were sent to investigate the origins of elaborate fragments of stucco that had been retrieved by a local military division (Ershov 1947).Triangulating the position of the site with medieval Arabic geographies, Ershov and Maruschenko were able to identify the ruins as those of Dandanakan, a small trading settlement on the high road to Khurasan.Noting the importance of the site and the stucco, the Soviet team undertook rapid rescue excavations in 1942 to examine the main mosque in the city, uncovering remains dating from approximately the 11th-12th century A.D. (Pribytkova 1964).
The extant remains of the site are a low earth mound of occupation and building debris, in the form of an almost exact square of approximately 215 × 215 m.Additional remains of extramural buildings and probable industrial sectors can be seen on satellite imagery as mounds beyond this core area.A team of Turkmen archaeologists began to reinvestigate the site in 2017 with a series of excavations to ascertain the preservation and layout of the settlement.In 2019, this was expanded to form ToKa, a collaborative project with the Metropolitan Museum of Art, New York, to explore the stratigraphy of the site and its parallels at other desert cities.One aspect of this initiative was to excavate a new trench, 6 × 8 m (DDK1), through an area interpreted as domestic remains to assess the occupation stratigraphy of the site and how it relates to the traces of houses and streets visible from aerial imagery (Wordsworth et al. 2022).One short season of excavation was completed in 2019, revealing the remains of a collapsed building and adjacent street oriented on the same alignment as the city.Postoccupation digging at the site to recover bricks had damaged some of the uppermost building remains, but aside from these later interventions, several layers of in situ collapse were excavated, as well as the upper occupation layers of the street.The material from these strata collectively represents the final stage of medieval habitation of the settlement of Dandanakan, after which it seems to have been completely abandoned.Following this initial season of work, excavation in this trench was paused owing to the global pandemic, and the corpus presented below represents the sum total of ceramics retrieved from this first season.
In addition to trench DDK1, an initial walkover survey of the site identified a number of brick kilns that had been exposed by modern agricultural activity, characterized by visible burnt material, vitrified earth, and misfired bricks.A stratigraphic section through one of these kilns was cleaned (DDK2) in order to obtain secure dating material from two layers of in situ burnt remains, processed using flotation.Bricks from the kilns were also sampled as petrographic examples of local manufacture and incorporated into the analysis below.
Dandanakan is located between the alluvial fans of the Murghab and Tejen Rivers, which flow down from the Hindu Kush range in eastern Afghanistan to the region of the city of Mary in southeastern Turkmenistan.This region is underlain by the Amu Darya (Karakum) basin.The bedrock of the Amu Darya basin has alternating layers of continental and marine rock fragments, interbedded with carbonates, salt, and humic organic matter (Brunet et al. 2017;Ghassemi and Garzanti 2019).Covering the bedrock is Neogene and Quaternary sedimentary, metamorphic, and volcanic rock fragments brought by wind and by the rivers that flow through the region (Marcolongo and Mozzi 1998; Garzanti et al. 2018); the frequency of metamorphic rock fragments increases from the east to the west of the basin.The geology of the Karakum region is different from southwestern and northwestern Turkmenistan.The southwestern part is dominated by the Köpetdag, marked by the presence of limestone and derivative minerals in the west and center and volcanic ash and sandstone in the east (Coolidge 2005).The northwestern part of Turkmenistan is underlain by shallow marine sediments and Palaeozoic to Middle Triassic igneous and metamorphic rocks, including gabbro, granite-gneiss, and porphyroid (Lyberis and Manby 1999).

Materials and Methods
Our study focused mainly on the materials recovered from the collapse and abandonment of the excavated building in DDK1, as the final occupation layers of the structure have yielded few suitable fragments at this stage.The ceramics from the collapse, as well as the 14 C dates from the final occupation layer, suggest a date of abandonment at some point from the late 11th-mid-12th century A.D. ( 14 C analysis for the final occupational layer of the street suggests a date of A.D.  1027-1156 with 95.5% confidence).From DDK1, we sampled 27 ceramics and one brick, and from the brick kiln at DDK2, we sampled two bricks.The material from the kiln is crucial to set up a comparison for locally-made ceramics. 14C of shortlived plant remains (twigs and a wheat grain) from burning layers within the kiln imply that it was in use at some point between A.D. 1040 and 1163 (with 95.5% certainty).
The 30 ceramic and brick samples are representative of the range of the whole excavated assemblage from 2019, which comprised (excluding the topsoil), 1322 ceramic fragments, of which 72% (n = 950) were unglazed, of mostly closed forms, and 28% (n = 372) glazed, mostly open forms of a variety of monochrome glazes (green, yellow, brown, and turquoise), either slipped or un-slipped, mottled, splashed, painted (brown on white, brown/green/red, red on white, and "buff ware"), sgraffito, and mottled sgraffito productions. 1 All ceramic fragments were subjected to macroscopic examination of fabric on fresh breaks, surface treatment, and form, and were photographed; rims were measured; and, new shapes were drawn.To record the macroscopic characteristics of the fabrics, we used a 10x magnifying hand lens, which allowed us to classify them into macroscopic fabric groups based on a set of standardized criteria, such as the frequency and size of voids and inclusions, color and shape of inclusions, and color, texture, density, and hardness of the fabric.We assigned different fabric groups to unglazed and glazed ceramics, even when their macroscopic characteristics were similar: our rationale being that their production required different technologies.The macroscopic fabric groups constituted the basis for selecting the 30 samples for technological analyses (Figures 2-5).
The technological analyses were conducted at the Pitt-Rivers Laboratory for Archaeological Science, University of Cambridge (Table 1).Thin-section petrography and scanning electron microscopy energy dispersive spectrometry (SEM-EDS) followed the same protocol described by Ting and colleagues (2021).We carried out petrographic analysis on all selected samples to characterize the microscopic features of the fabrics and evaluate their correlation with the macroscopic counterparts.The microscopic features we recorded included the mineralogical composition and textural features such as the size, shape, frequency and orientation of inclusions and voids, and the color and optical activity of the clay matrix, loosely following Whitbread's (1995) descriptive system.The estimation of the abundance of inclusions was made using the percentage charts developed by Matthews, Woods, and Oliver (1991).We further used these microscopic features as a proxy to determine the potential provenance of the samples by comparing them with local geological maps and information (Quinn 2022).
We carried out SEM-EDS analysis on 12 samples, including all but one poorly preserved glazed ware and one unglazed ware with a whitish surface (DK11).These samples were selected due to the presence of surface decoration such as glaze and possibly slip, which warrant a higher-resolution technological analysis by SEM-EDS.Three glass standards were also analyzed, showing that the data generated by our SEM-EDS has good accuracy and precision (Supplemental Material 1).SEM-EDS analysis can focus on a specific area of the sample at high magnification, making it suitable for characterizing the composition and microstructure of the ceramic body, slip, and glaze, where applicable (Pradell and Molera 2020).
Based on the macroscopic and microscopic observations, we submitted a subset of eight samples-consisting of two bricks and six glazed and unglazed ceramic samples-for refiring experiments.Noteworthy is that there may seem to be little overlap between the samples submitted for SEM-EDS and the refiring experiments.This is due to the small number of samples we were allowed to take from each sherd as per the local regulations of exporting samples.That being said, the samples we selected for the refiring experiments should reflect the range of ware types and fabric groups (both macroscopic and microscopic) that are present within the assemblage.The bricks were refired between 500 and 900°C, whereas the ceramic samples were refired between 600 and 900°C, all at an interval of 100°C.The samples were then analyzed using Fourier-transformed infrared spectroscopy (FT-IR) to estimate the firing temperature based on the changes in mineral phases that occurred during refiring (Shoval 2016).A Thermo iS5 FTIR spectrometer and the KBr method were used, in which the samples were crushed into powder and mixed with KBr powder to form pellets for analysis.Spectra were collected between 4000 and 400 cm -1 at 4 cm -1 resolution.

Macroscopic examination
We identified 20 macroscopic fabrics, including 16 earthenware fabrics and four siliceous ones (Supplemental Material 2)."Earthenware" here simply refers to ceramics made of a clay paste as opposed to a siliceous paste; this term is commonly used to describe clay-paste ceramics across Islamicate lands.Among the earthenware fabrics, 11 are associated with unglazed ceramics and five with glazed ceramics.The earthenware fabrics E1-E8 and E20-E24 show an overall low variation in their macroscopic characteristics, as they all have a fine and consistent granulometry, fine-grained inclusions, and no obviously-added temper.These fabrics do, however, vary in color and hardness, which we posited may be caused by a slight difference in composition and/or firing conditions, a point we will address below.One fabric (E7) stands out from other earthenware fabrics for its coarser-grained granulometry, laminated texture, and possible occurrence of vegetal inclusions.Fabrics that have been tentatively described as "sandy" are predominant, and four of these (E1, E2, E20, and E21, plus the variant E1b) account for almost 84% of the DDK1 assemblage (SM Figure 1).Within the "sandy" fabrics, we noted the similarity between unglazed fabric E1 and glazed E21 and between unglazed E2 and glazed E20 and E22."Non-sandy" fabrics only comprise E4 and E24 (0.6% of the assemblage).
Most of the unglazed ceramic fragments could not be assigned to specific forms and therefore did not allow a thorough understanding of wares, but some correlations were noted among macroscopic fabrics and vessel types.Jugs (i.e.closed forms with a neck and handle) were predominantly linked to fabric E2 (16 fragments; one jug fragment was E1).Fragments of tandir/oven covers or lids (thick flat objects with occasional traces of burning) belonged to our coarser-grained fabric E7 (six fragments).Our only example of a sphero-conical vessel was associated with a fabric (E8) that is notably different from other earthenware fabrics.Less clear is the correlation of closed forms with relief ware (i.e.mold-made) and the associated fabrics E1 and E3, as their occurrence was rare (only two and one fragments, respectively).
For the glazed ceramics, we noticed varying degrees of correlation between the macroscopic fabric and the vessel form and surface treatment (SM Figure 2).Almost all glazed earthenware (83%) recovered from Dandanakan were associated with three larger fabric groups (E20, E20b, and E21), of which E20 is by far the largest attested (63% all of glazed).The surface treatments associated with E20, E20b, and E21 included splashed, sgraffito, monochrome glazed, and painted decoration, but notably did not include monochrome mottled glazes (see below).Painted glazed ceramics have the highest frequency of occurrence overall and likely comprise a variety of productions; of these, three broadly defined decorative styles were identified.The most common one is "Painted Brown on White:" sparsely painted with recurrent epigraphic motifs, it shows a rather diluted brown/black paint over a thin white slip under a colorless transparent glaze.The second most common, "Painted in Brown/Green/Red," shows the painted decoration over a thicker layer of white slip; incised decoration is present in some cases.There are also a small number of buff-colored bowls that have green, yellow, and brown/black painted decoration under a transparent colorless glaze ("Buff ware").Another clear correlation observed was that monochrome turquoise glazes were exclusively paired with fabrics E21 and E20b, regardless of their variation in surface treatment and decoration ("turquoise:" plain, incised, transparent turquoise or opacified turquoise and yellow glaze).The most outstanding distinction was that the monochrome mottled glazed ceramics, with or without sgraffito, were strongly associated with fabric E22.This fabric is also associated with two specific vessel shapes: medium-deep bowls with a rounded profile and incised lines below the exterior rim and a smaller version of the same bowl form but with no incised lines.
The macroscopic subdivision of the siliceous fabrics was not as straightforward, with only slight variation in hardness and texture between our samples, which we have divided into four fabric groups.

Petrographic analysis
We switch to referring the samples by their laboratory number (lab no.) here and below.Petrographic analysis of 30 samples reveals the presence of three petrofabric groups and three outliers (see Supplemental Material 3 for detailed description).The petrofabric groups are the Quartz Opaque Group, Quartz Biotite Group, and Quartz Group.The majority of the samples are placed in the Quartz Opaque Group (n = 13) and Quartz Biotite Group (n = 10); these samples consist of unglazed and glazed ceramics and fired bricks.The unglazed and glazed ceramic samples of the Quartz Opaque Group and Quartz Biotite Group have a fine-grained fabric, exhibiting a similar suite of mineral inclusions that comprise quartz, opaque (nodules that are rich in iron oxide), biotite, plagioclase feldspar, quartzite, and amphibole.The relative abundance of inclusions varies between these two petrofabric groups: quartz and opaque are the dominant type of inclusions of the Quartz Opaque Group (Figure 6A-B), and quartz and biotite in the Quartz and Biotite Group (Figure 6D-E).In both cases, the petrofabric group is split into two subgroups: Subgroup A has a less overall abundance of inclusions that are finer-grained than Subgroup B. All unglazed and glazed ceramic samples belonging to the Quartz Opaque Group and Quartz Biotite Group display similar textural characteristics, featuring inclusions that are sub-angular and sub-rounded in shape.The inclusions and voids align parallel to the margin of the thin sections.The clay matrix is optically inactive; that is, the color of the matrix does not change when rotating the sample in cross polarization (XP) in microscopy.We also assigned the fired bricks as an associated member of these two petrofabric groups, as they share the same mineralogical composition profile with the ceramic samples, even though the fired bricks have different textural features (Figure 6C, 6F).
Four glazed ceramic samples belong to the Quartz Group (Figure 7A).This petrofabric stands out from the Quartz Opaque Group and Quartz Biotite Group: quartz grains are the predominant type of inclusion, often appearing to be uniform in grain size and shape.These quartz grains are surrounded by a small amount of clay, which sometimes contains minerals such as biotite and pyroxene.We also identified three outliers-all unglazed ceramic samplesthat do not match with the description of the petrofabric groups above.DK09 and DK13 have a notably coarsegrained fabric (Figure 7B-C).DK09 has quartz and quartzite as the principal type of inclusion, whereas DK13 has a high frequency of amphibolite, serpentinite, and quartz.Another outlier is DK14, which has a distinct fabric full of micritic calcite (Figure 7D).

Ceramic body
Focusing on the 12 samples submitted for SEM-EDS analysis, a distinction can be easily drawn between the samples with exceptionally high silica (SiO 2 ) content (DK23, DK24, DK25, and DK26) and those with silica content that is more typical of most earthenware ceramic composition (DK11, DK15, DK16, DK17, DK18, DK20, DK21, and DK22) (Supplemental Material 4).The samples with a lower silica content are characterized by a higher lime content, pointing to the use of calcareous clays.DK11, DK15, DK16, DK17, DK18, and DK22-all belonging to the Quartz Biotite Group-have a slightly higher magnesia, alumina, and iron oxide content, which are characteristic of the composition of biotite, a major mineral component of these samples.The silica-rich samples correspond with the Quartz Group; their chemical and mineralogical features are consistent with what is known as a stonepaste or siliceous body (Allan 1973;Mason 2003Mason , 2004;;Mason and Golombek 2003;Rugiadi 2011).In our samples, no glass fragments in their original shape are present (Mason and Tite 1994;Tite, Wolf, and Mason 2011), but areas with crystalline phases can be found in some spaces between the quartz grains.Analysis of these crystalline phases shows that they are rich in soda (Na 2 O), alumina (Al 2 O 3 ), magnesia (MgO), and lime (CaO), in addition to their high silica content; the lead oxide (PbO) content is below the limits of detection.These observations suggest that the glass fragments added to the fabric of the Quartz Group were alkaline-based, which only survive as extended areas of interparticle glass (Tite, Wolf, and Mason 2011).

Slip
Slip, here, refers to an extra layer of clay applied to the surface of a clay body; it is usually formed by fluid suspension and can be mixed with other materials to enhance its color or properties (Rice 2005).We identified the presence of slip in five samples, including one from the Quartz Opaque Subgroup B (DK21), one from the Quartz Biotite Subgroup A (DK17), and three from the Quartz Biotite Subgroup B (DK11, DK15, and DK16) (Supplemental Material 5).In terms of the microstructure, the slip of DK11-the only unglazed sample submitted for SEM-EDS-stands out for its thinness and a lack of inclusions (Figure 8A-B).The slip of the remaining samples, all glazed wares, have varying thicknesses, all characterized by the presence of inclusions that are uniform in shape and grain size (Figure 8C).The slip of DK15 and DK16 contains some red inclusions that are visible in optical microscopy (Figure 8D).These red inclusions appear as bright particles in SEM-EDS.
Compositionally, the slip of the unglazed sample (DK11) has a high lime content, suggesting that calcareous clay was used, similar to the ceramic body.This contrasts with the slip of the glazed samples (DK15, DK16, DK17, and DK21), which has a low lime content.These slips also have a higher alumina and lower iron oxide content than the associated ceramic body, implying that different clays were used to make the slip and ceramic body, with the clay for slip being alumina-rich and non-calcareous.The slip of the glazed samples is further characterized by a high lead oxide content.This could be due to the addition of lead oxide to the clay (Molera et al. 2020) and/or a reaction with the lead glaze (see below for glaze composition).The inclusions identified in the slip microstructure are quartz, evidenced by an elevation of the silica content in the area with inclusions as opposed to the area without inclusions.The red inclusions in DK15 and DK16 are rich in iron oxide.

Glaze
Glaze is present in 11 samples, including two from the Quartz Opaque Subgroup B (DK20 and DK21), two from the Quartz Biotite Subgroup A (DK17 and DK22), three from the Quartz Biotite Subgroup B (DK15, DK16, and DK18), and four from the Quartz Group (DK23, DK24, DK25, and DK26).In terms of microstructure, the presence of onion-like layers in most samples suggests that the glaze was subjected to post-depositional corrosion.A clear distinction is noted between the slipped and unslipped samples of the Quartz Opaque and Quartz Biotite Groups.The glaze of DK15, DK16, DK17, and DK21-all slipped-has no particle or microcrystalline throughout; there is a clear interface between the glaze and associated ceramic body (Figure 9A-B).The glaze-ceramic body interface is less well-defined for the unslipped samples (DK18, DK20, and DK22).DK18 and DK20 have abundant needle-like and rhombic microcrystallites along the interface (Figure 9C-D), whereas these microcrystallites disperse throughout the glaze of DK22.DK20 is further characterized by the presence of bright microcrystallites throughout the glaze (Figure 9E-F).The glaze of all Quartz Group samples has a congregation of angular quartz grains along the ceramic body (Figure 10A-B), while the rest of the glaze has no particles or microcrystallites.DK24 is an exception, as its glaze has bright microcrystallites scattered throughout, similar to DK20 (Figure 10C-D).
The Quartz Opaque Group and Quartz Biotite Group have a different glaze composition from the Quartz Group (Supplemental Material 6).All but one sample (DK20) of the Quartz Opaque Group and Quartz Biotite Group have high lead oxide content in the glaze (≥ 50 wt%).These samples also have low alkali and alkaline-earth oxides, with less than 3 wt% of soda, magnesia, and potash in sum.DK20 has a lower lead oxide (≤ 40 wt%) and higher alkali and alkaline-earth oxides concentration (> 6 wt%).Analysis of the needle-like and rhombic microcrystallites in the glaze of the unslipped samples (DK18, DK20, and DK22) shows that they are lead-feldspar and wollastonite, respectively.
Judging from their small grain size, the microcrystallites are likely newly-formed as a result of the reaction between the glaze and associated calcareous ceramic bodies that had not undergone a first firing, or "biscuit-firing."This may also explain an elevation in the lime content of the glaze of the unslipped samples compared to the slipped ones.
The glaze of the Quartz Group samples has high alkali and alkaline-earth oxides, with the sum of soda, magnesia, and potash ranging from 12.9-19.6wt%.A lower sum is recorded in DK24, probably due to a proportionally higher lead oxide content in this sample.Lead oxide is also found to be present in DK26, but its content is lower than DK24; in both cases, the lead oxide content is still lower than the samples of the Quartz Opaque Group and Quartz Biotite Group.Lead oxide is not present or below the limits of detection in DK23 and DK25.A higher lime content is also detected in DK24 and DK26 than the other two samples of the Quartz Group.
Regardless of the petrofabric group, copper oxide (CuO) was measured in the samples with green (DK15, DK17, DK18, and DK21) or turquoise (DK20, DK23, DK24, and DK25) glazes (Tite 2011).An elevated iron oxide content is associated with the samples with yellow glaze (DK15 and DK22).Cobalt (CoO) and manganese (MnO) oxides, together with a higher iron oxide and the presence of arsenic trioxide (As

FTIR
We conducted FTIR analysis on eight samples, all with earthenware fabrics.For the Quartz Opaque Group, DK02, an unglazed ceramic of Subgroup A, displays a similar spectrum when refired at 600 and 700°C, with a CO 3 band of calcite at 1420 cm -1 .This CO 3 band became weaker when the sample was refired at 800°C and almost disappeared at 900°C, while a weak SiO band developed at 920 cm -1 .This SiO band could be indicative of the presence of gehlenite, a firing silicate that is formed when the clay is fired between 800 and 900°C.These observations help place the firing temperature of DK02 between 700 and 800°C.Very little change occurred to the spectra of DK05 and DK19-an unglazed ceramic of Subgroup A and a glazed ceramic of Subgroup B, respectively-when the samples were refired at 600, 700, and 800°C, all marked by an absence of the CO 3 band of calcite at 1420-1450 cm -1 .Calcite begins to disintegrate when fired above 700°C (Maggetti 1982;Fabbri, Gualtieri, and Shoval 2014).This, coupled with the presence of a weak SiO band of firing silicate at 920 cm -1 (gehlenite) when the samples were refired at 900°C, suggests that DK05 and DK19 were fired well above 700°C and possibly between 800 and 900°C.The spectra of DK28, the fired brick affiliated with the Quartz Opaque Group, show a gradual shift of the main SiO band from 1039 at 500°C to 1078 at 900°C and a gradual lowering of the CO 3 band of calcite until it almost disappeared at 900°C.The peak at 2950-2850 cm -1 , which reflects the presence of organic matter, disappeared after the sample was refired at 600°C.Judging from continuous reaction during refiring, it is likely that the brick was fired below 500°C, but the temperature range cannot be determined, as the sample was not refired at a lower temperature.
Only the samples of Subgroup B of the Quartz Biotite Group were analyzed.The CO 3 band of calcite at 1420 cm - 1 in the spectra of DK11 and DK15-an unglazed and a glazed sample, respectively-disappeared when DK11 was refired at 800°C and DK15 at 900°C.No calcite band is observed in the spectra of DK18, a glazed sample, from the onset of the refiring experiment.The SiO band of firing silicate appeared at 920 cm -1 (gehlenite) when DK15 and DK18 were refired at 900°C.These observations help place the firing temperature of DK11 to between 700 and 800°C and DK15 and DK18 between 800 and 900°C.The difference in the CO 3 band of calcite between DK15 and DK18, both glazed samples, may reflect DK18 being fired at a temperature closer to 900°C than DK15.DK27, the brick affiliated with the Quartz Biotite Group, shows a similar trend in its spectra to DK28; it is postulated that the brick was fired at temperatures below 500°C.

Establishing the potential provenances of our samples
Inclusions such as quartz, opaque, biotite, amphibole, pyroxene, and plagioclase feldspar that are characteristic of the Quartz Opaque and Quartz Biotite Groups-identified by petrographic analysis-are common constituents of the sediments underlying the southern Karakum region, where Dandanakan is located.Such correspondence suggests that the vessel and brick fragments assigned to the Quartz Opaque and Quartz Biotite Groups were produced using regional raw materials.The identification of the use of both of these two petrofabrics in making bricks at Dandanakan suggests that both subgroups may represent local geologies, assuming that the raw clay for their production was not carried over significant distances.
It is challenging to determine the origins of the Quartz Group (stonepaste), even though Mason (2004) proposed some potential criteria for the petrographic characterization of stonepaste from different regions.Quartz-the principal ingredient to make this petrofabric-is found across a wide variety of geologies, making it an ineffective marker in establishing ceramic provenance.The little clay that is used to make the paste malleable is usually highly refined and homogenous and displays few additional characteristic minerals.The glaze of the Quartz Group may possibly provide us with some indications of where these vessels might have come from.The high soda, magnesia, and potash concentration in the glaze suggests that plant ash was the source of the alkaline flux used.Comparing our samples and published examples of coeval alkaline glazes in Central and western Asian regions (Mason and Tite 1997;Henshaw 2010;Rante and Collinet 2013;Matin 2018;Matin, Tite, and Watson 2018;Molera et al. 2020;Klesner et al. 2021), the alkaline flux of DK23, DK24, DK25, and DK26 is similar to the reported contents of the turquoise glazed stonepaste found in Nishapur (Rante and Collinet 2013).The cobalt colorant added to the glaze of DK26 contains manganese, iron, and arsenic oxides, a formula consistent with the cobalt sources originating from Iran (Matin and Pollard 2017).While it is tempting to agree with the dominant narrative in assigning Dandanakan stonepaste to an Iranian production source, the absence of comparative studies on the Merv stonepaste, combined with the possibility that glaze raw materials were transported, makes any firm conclusion at this point impossible.
Three earthenware outliers (DK09, DK13, and DK14) all have significantly different mineralogical and textural characteristics from the Quartz Opaque and Quartz Biotite Groups, implying that these vessels are not local to the Mary region.DK13 is distinguishable for the presence of metamorphic rocks such as quartzite and amphibolite, similar to the Group B petrofabric identified by Rante and Collinet (2013) in their study of the glazed and unglazed pottery from Nishapur.Rante and Collinet's Group B petrofabric contains quartz, plagioclase feldspar, mica, amphibole, and sedimentary rock fragments, likely coming from the soil of the first slopes of the Binalud Mountains that border Nishapur (Rante and Collinet 2013, 86, 128-131).A Nishapur origin is, therefore, postulated for DK13.DK09 and DK14 are not similar to the petrofabrics characteristic of Nishapur and other Iranian productions (Mason 2004) nor to the limited studies of petrofabrics from Merv (Casellato et al. 2007) and the Köpetdag region (Coolidge 2005).The presence of quartzite and limestone in DK09 and the micritic nature of DK14 nonetheless seem to be consistent with the geology of the western and central part of Köpetdag, respectively, but there is not enough evidence to determine the provenance of these two samples.

Exploring the relationships between the macroscopic fabrics and petrofabrics
By cross-referencing the petrofabric groups with macroscopic fabrics, we notice that the main distinctions evidenced by the two methods align (Table 2).The samples of fabrics E1 and its variant E1b, its coarser equivalent E7, and its glazed equivalent E21-which were macroscopically recognized as similar-all fall within the Quartz Opaque Group.Similarly, two-thirds of the samples of fabrics E2, its variant E2b, and its glazed equivalent E20-macroscopically recognized as similar-all align with the Quartz Biotite Group.On the other hand, we cannot see any clear correlation between macroscopic fabrics and the petrofabric subgroupings.
Considering the whole ceramic assemblage of the site, it is clear that the two main petrofabric groups, Quartz Opaque and Quartz Biotite, considered local to the southern Karakum region and available close to Dandandakan, are representative of all the glazed and most of the unglazed ceramic types (macroscopic fabrics unglazed E1, E1b, E2, E2b, E3, E4, and E7, of different closed forms; glazed E20, E20b, E21, E22, and E24 of all open forms, monochromes, splashed, painted, sgraffito, mottled, and mottled sgraffito), suggesting that these types were all made in the region or nearby (the exception being a few outliers/imports mentioned above; see Table 2, SM Figure 1).The FTIR data further suggest that the variations that we observed visually in the color and hardness of these fabrics are not to be ascribed to firing temperatures, as we tentatively postulated, but were due to the exploitation of slightly different outcrops of two main raw sources (Quartz Opaque and Quartz Biotite).
Firing temperatures employed align closely with three broad groups of artifacts-unglazed vessels, glazed vessels, and bricks.The firing temperature of the unglazed vessels analyzed ranged between 700 and 800°C (DK02 and DK11), with one exception possibly fired above 800°C (DK05), while the firing temperature of the glazed vessels seems to have been slightly higher, in the range of 800-900°C (DK15, DK18, and DK19).Bricks were fired at much lower temperatures, below 500°C (DK28 and DK27).Slips found on glazed earthenwares were made by mixing an alumina-rich non-calcareous clay with quartz grains (DK17 and DK21), with the occasional presence of iron oxide-high red inclusions (DK15 and DK16), while the one slip analyzed of an unglazed fragment used a calcareous clay similar to the ceramic body (DK11).The glazes on earthenwares are all of the high lead-silica type (DK15, DK16, DK17, DK18, DK21, and DK22).The one exception is the turquoise glazed fragment that has a lead oxide and plant ash glaze with lead-tin microcrystallites as an opacifying agent (DK20), the latter being a feature not found in similar turquoise glazed ceramics from Akhskiket, Termez, and Nishapur (Henshaw 2010;Rante and Collinet 2013;Molera et al. 2020).Glazes on stonepaste (Quartz Group petrofabric) are alkali and alkaline-lime, although in two cases, low percentages of lead oxides are also present (DK23 and DK25; DK24 and DK26).
Three of the macroscopic fabrics that were rarely attested did not fall into the regional petrographic groups, resulting instead in outliers, and are thus likely imports.These include the one fragment of a spheroconical vessel (E8: one fragment), a large storage jar (E6: nine fragments), and an unidentified closed form with comb incised decoration (E5: 10 fragments).

Conclusion
The results presented above allow us to distinguish the fabrics of the ceramics that were made with regional raw materials, perhaps local to Dandanakan or slightly further afield, from those which were likely imported.The data generated by the petrographic and FTIR analyses seem to have suggested that the variations observed macroscopically were due to the use of different outcrops of raw materials rather than different firing condition.
Our analyses also illustrate that fabric-whether it be observed macroscopically or petrographically-crosscut ceramic ware types, implying that the same regional raw clays were used by workshops to produce a wide range of unglazed and glazed vessels.Combined with evidence for consistent slip and glaze recipes and firing temperatures, this might suggest that regional ceramic workshops in southern Karakum or at Dandankan itself had the same access to raw materials and possibly shared knowledge of production techniques.This observation also holds true across styles that could be chronologically diverse, suggesting comparatively stable traditions of pottery manufacture.We also identified the presence at Dandanakan of imported unglazed pottery (one spheroconical vessel, a large storage jar, and an unidentified closed form) and of a glazed ceramic with a siliceous body.
In sum, in the late 11th and 12th centuries A.D., the production centers of the southern Karakum that provided earthenware ceramics to be consumed in Dandanakan exploited at least two sources of raw clay; potters seemingly used the same clays to make the body of and to slip unglazed vessels, while they slipped glazed ceramics with a mixture of alumina-rich, non-calcareous clay with quartz, and at times red oxides (DK15 and DK16).Meanwhile, they used high lead-silica glazes for painted and sgraffito ceramics, consistent with similar production elsewhere in Iran, Khurasan, and Mawarrannahr, but used a different glaze for turquoise glazed ceramics (plant ash glaze opacified with lead-tin microcrystallites) that is not paralleled in Iran nor Mawarrannahr.Finally, these regional production centers adapted their firing temperatures to the technology employed (unglazed, glazed, or bricks).Stonepaste ceramics available at Dandanakan in the same period, which may well have been imported, used alkali and alkaline-lime glazes, different from those of locally-made glazed earthenwares.Despite the pilot nature of our study, the results allowed us to reassess our research questions about local production; also, they tentatively suggest the need to re-evaluate the role of trans-regional trade of ceramics in the medieval period.Far from being finalized, this work needs to be refined by deeper stratigraphic investigations and a wider sampling for analyses, which will help clarify whether the results presented here are part of a widespread pattern or an exception.

Endnotes
1.The percentages used throughout the text are percentage sherd counts of the total assemblage from this specific excavation.These are only indicative figures for comparison within the assemblage and are not intended as an independent measure of quantification.Estimated Vessel Equivalents (EVEs) based on the percentage of rim sherds preserved have been recorded for the assemblage, but owing to the small size of this initial study and the high level of fragmentation, they are not used here.Following the continuation of the excavation in 2023 and the greater number of sherds available, percentages based on EVEs will also be included in future studies.

Figure 1 .
Figure 1.Map showing the location of Dandanakan and the contexts where the ceramic and fired bricks samples came from.

Figure 2 .
Figure 2. The earthenware ceramic samples with E1-E2 fabrics included in this study.

Figure 3 .
Figure 3.The earthenware ceramic samples with E3-E8 fabrics included in this study.

Figure 4 .
Figure 4.The glazed ware samples with E20-E22 fabrics included in this study.

Figure 5 .
Figure 5.The glazed ware sample with E24 fabric, earthenware samples with E7 fabric, and the ceramic samples with special fabrics included in this study.
2 O 3 ), are found in the blue glaze of DK26.Analysis of the bright microcrystallites dispersed throughout the glaze of DK20 and DK24 shows that they are rich in lead and tin oxides, acting as an opacifying agent (Mason and Tite 1997; Tite et al. 2015; Matin, Tite, and Watson 2018).

Figure 8 .
Figure 8. Images showing the slip of A-B) an unglazed ware (DK11) and C-D) a sgraffiato splash glazed ware (DK15).Images in the left column are backscattered electron images (BSE) taken by SEM-EDS and those in the right are taken with a digital microscope.All images are taken at 200x magnification.

Figure 9 .
Figure 9. Images showing A-B) a clear glaze-ceramic interface of DK16 (Quartz Biotite Subgroup B), C-D) the growth of newly-formed microcrystallites along the interface of DK18 (Quartz Biotite Subgroup B), and E-F) the presence of lead-tin microcrystallites in the glaze of DK20 (Quartz Opaque Subgroup B).Images in the left column are BSE taken by SEM-EDS, and those in the right are taken with a digital microscope.Images are at different magnifications.

Figure 10 .
Figure 10.Images showing A-B) the presence of two glaze layers of DK23 (Quartz Group) and C-D) the presence of lead-tin microcrystallites in the upper glaze of DK24 (Quartz Group).Images in the left column are BSE taken by SEM-EDS, and those in the right are taken with a digital microscope.Images are at different magnifications.

Table 1 .
Description of the ceramic and fired brick samples included in this study and the analyses carried out on them.

Table 2 .
Summary of the macroscopic, petrographic, and technological analyses.