1The medieval (ante 14^th c.) stratigraphic layers of the excavations carried out in the ‘80s at the Palazzo Ducale of Genoa (Liguria, NW Italy) are rich in ceramic wares coming from a large part of the Mediterranean (Cabona et al., 1986; Rabino, 2006-2007). However, the high degree of fragmentation of the pottery is an obstacle to a classification only based on typological characteristics, in particular as for the coarse and the glazed monochrome wares, which are very poor in discriminant features. On the other hand, little is known about production centres and diffusion of these classes also at a Mediterranean scale, and few archaeological/typological published data are available for comparative studies (for instance: Grassi, 1999; Waksman, 2002).

2Previous investigations on some medieval glazed monochrome types found in Ligurian and Provençal contexts (Capelli et al., 2002; Capelli et al., 2006) demonstrated the utility of thin section analyses in order to identify different productions and to localise the areas of provenance. On these bases, a interdisciplinary study was carried out on 12th-13th c. glazed cooking wares found at Palazzo Ducale (Capelli et al., 2007a). The integrated petrographic and archaeological study allowed the identification of seven groups (1-7), three of which (1, 4, 5) dominate. Moreover, provenance information was obtained by the comparative study of geological data and the thin section database of Mediterranean ceramics available at the DipTeRis of Genoa. Finally, diffusion data were also attained for a few groups by integrated archaeometric and archaeological analyses of other north Tyrrhenian coeval contexts.

3In this paper we will present the results of the thin section study, as well as those of new XRPD and SEM-EDS analyses on selected samples, mainly aimed at better characterising the seven different productions from a technical point of view, finding out further distinctive elements and evaluating preliminarily if more accurate analytical methods (such as XRF applied on bodies and microprobe, laser ablation, PIXE on glazes) could significantly improve the results.

4In Table 1 are listed the analysed samples of Palazzo Ducale, with a few archaeological data. In addition, a reference sample (7530, Group 7*) from the production of Uzège, southern France (Leenhardt, 1995) was analysed for comparison with the Group 7 sample. In Figure 1 are shown some diagnostic sherds representative of the groups identified by the analyses.

1. 1. Analytical methods

5After the observation of all the recovered sherds (more than 650) of that class by a 10x-magnifying lens, 29 representative samples (all the typologically diagnostic sherds, chosen in collaboration with the archaeologists) were submitted for thin section analysis.

6Among them, 16 samples were selected for SEM-EDS analysis. All samples showing strong evidence of weathering, affecting body and/or glaze, have not been taken into consideration.

Finally, seven representative samples were also analysed by XRD (Tab. 1).

7XRD analyses were carried out using a Philips PW3710 diffractometer. Powdered samples were run between 2.5° and 70°2, with a generator potential of 30 kV, a generator current of 22 mA (using a CuK radiation), a Ni filter, and a scan speed of 1°/min. The software used for XRD data reduction was Philips PC-APD Diffraction Software and MacDiff 3.0.6c.

8SEM-EDS images and analyses were carried out on polished thin sections with a Philips SEM 515 equipped with an energy-dispersive spectrometer (EDAX PV9100). Quantitative analysis calibration was accomplished with synthetic (glass) and natural standards. The results of SEM-EDS analysis of glazes and pastes (major and a few minor elements) have been normalised to 100wt%. Microanalyses of glazes were carried out on selected points (spot mode) avoiding alterations, reaction zones and mineral inclusions; the reported values are the mean of at least three analyses. For paste analysis, the window mode (150x corresponding to 0.5 mm²) was used, avoiding the outer parts of the body, which are influenced by the glaze, and the coarsest inclusions; the reported values are the mean of two or more analyses. SEM-EDS analyses of pastes (bearing in mind the limits due to the small number of samples – in particular as for some groups consisting of one or very few available samples – and the low accuracy of that method, especially for heterogeneous materials) were carried out in order to obtain preliminary data about the chemical composition of the seven groups and to search for possible relationships with petrographic features.

1. 2. The seven groups. Mineralogical, petrographic and textural analyses of pastes and glazes

9The main features of the different groups resulting from optical and electron microscopy are described below and summarised in Table 2 (pastes) and Table 3 (glazes). Table 2 also shows the mineral phases identified by XRD.

1. 1. Group 1

10Body. Paste inclusions are angular, abundant, medium-sorted and often coarse-grained (Figs. 2-3, n. 1). Most of them are composed of weakly metamorphic quartz+feldspar (granitoid?) rock fragments (up to 1-2 mm in size) and derived minerals (<0.3-0.5 mm): feldspars, sometimes rich in Fe-oxide micro-inclusions (Fig. 4, n. 1), quartz, and rare micas. Fine-grained quartz-micaschist fragments are also present in minor amounts. The clay matrix is generally oxidised and partially vitrified. The macroscopical colour of the body is generally red-orange and homogeneous in the cross section.

11Glaze. Macroscopically, glazes are bright orange in colour. In thin section they are not coloured or, rarely, light yellow. Their thickness is low (<0.10-0.15 mm) but homogeneous (Fig. 3, n. 1). Relic quartz inclusions are rarely observed. In sample 6901, one cluster of cassiterite micrograins was found out in the glaze (Fig. 4, n. 2). The glaze-body interface, with neoformed K-Al-Pb silicates (K-Pb feldspars; Molera et al., 2001) is rather thin (Fig. 4, nos. 1, 2).

Group 2

12It consists of few samples (only one analysed here) slightly different from Group 1 in typological features, fabric texture and composition. It could possibly be considered a sub-group of the former.

13Body. In the coarse fraction (up to 2 mm in size) there are several clasts of fine- or medium-grained quartz +/- feldspar metamorphic rocks, rarely showing possible relic volcanic textures, associated to granitoid (?) rock fragments very similar to those characteristic of Group 1.

Glaze. In thin section, the glaze is light yellow in colour, very thin (ca. 0.06 mm) and regular.

Group 3

14Body. Aplastic inclusions are abundant, poorly sorted and mostly composed of generally fine-grained (<0.3 mm) quartz, feldspar and mica individuals (Figs. 2-3, n. 2). Some coarser fragments (up to 1 mm in size) of quartz-feldspar rocks (meta-granitoids?), as well as fine-grained quartzites and possible acid meta-volcanites, are also present.

15Glaze. The glaze, light yellow in colour under the microscope, is thin (<0.1 mm) but homogeneous (Fig. 3, n. 2). The reactions at the glaze/body interface are little developed (Fig. 4, n. 3).

Group 4

16Body. Paste inclusions are subangular, abundant and medium-sorted. They are mainly composed of fine-grained acid metamorphic rock fragments (quartzite, quartz-schist, mica-schist, and phyllite, up to more than 1 mm in size) and of smaller quartz grains (Figs. 2-3, n. 3; Fig. 4, n. 4). The clay matrix is variably (from poorly- to well-) oxidised, even in a same sherd; the macroscopical colour of the body varies from grey or brownish to red.

The minor presence of calcite, identified by XRD analysis, could be considered of secondary origin.

17Glaze. Glazes are bubble-rich and inhomogeneous in thickness, even if generally thin (<0.1 mm) (Fig. 3, n. 3). Their colour varies from yellow to green depending on the redox conditions. The glaze-body interface is thick and rich in neoformed K-Pb feldspars (Fig. 4, n. 4).

Figure 4: SEM images showing glaze-body interactions.

Group 5

18Body. Paste inclusions are abundant, subangular and medium-sorted (Figs. 2-3, n. 4). They are composed of elements derived from both acid metamorphites and ophiolites: quartzite, quartz-micaschist, micaschist, paragneiss fragments prevailing on serpentinite (generally transformed into Mg-rich olivine during firing) and quartz+albite+epidote metabasite clasts (sometimes >1 mm in size); single mineral grains (<0.3 mm) of quartz, mica (abundant), feldspar, amphibole, epidote and several accessory minerals. The clay matrix is in general partially or totally oxidised and vitrified.

19Glaze. The glaze colour is typically brown macroscopically and yellow in thin section. The glaze thickness is variable (<0.1-0.2 mm) (Fig. 3, n. 4). The glaze-body contact is irregular and the interface, rich in neoformed K-Pb feldspars growing on mica and feldspar grains of temper, is rather thick (Fig. 5, nos. 1-2).

20Many micro-crystals of an undetermined Fe-rich neoformed lamellar phase are often included in the glaze (Fig. 5, n. 1), enhancing its brown-red macroscopical colour. Several bubbles are also present and, only in 6897, a few cassiterite clusters are scattered in the glaze (Fig. 5, n. 1). Occasionally, serpentinite fragments (transformed into Mg-rich olivine by firing) from the outer part of the body are embedded by the glaze. In that case, these fragments develop thick reaction rims, which are characterised by a spongy glassy texture including scattered Mg-rich phases indeterminable by SEM-EDS (Fig. 5, n. 2).

Group 6

21Body. Paste inclusions are abundant (Figs. 2-3, n. 5). They are composed of fine-grained quartz, mica, feldspar, amphibole and several heavy mineral individuals (<0.2 mm), rare larger gneiss fragments (up to 0.8 mm in size) and siliceous microfossils. In the oxidised clay matrix, micro grains of carbonates are associated with dominant Fe-oxides. Macroscopically, the body is rather homogeneously red-orange in colour.

22Glaze. The glaze colour is generally brown, rarely green-brown macroscopically, and yellow in thin section. Its thickness is about 0.1 mm in one sample and 0.2 mm in the other sample (Fig. 3, n. 5). The glaze-body contact is irregular and the interface is evidenced by the growth of abundant K-Pb feldspars, while many micro-crystals of Ca-rich neoformed phases (diopside and rarer wollastonite) are scattered in the glaze (Fig. 5, n. 3).

Group 7

23This group consists of very rare sherds, only one of which showing discriminant typological features. The reference sample 7530 (Group 7*) from Uzège is very similar to Group 7 in both fabric and glaze features.

24Body. Paste inclusions are almost exclusively composed of quartz grains. The largest ones (up to 1 mm in size) are rounded in shape and relatively scarce, while in the groundmass (<0.1 mm) they are subrounded and more abundant (Figs. 2-3, n. 6). The clay matrix is kaolinite-rich, as suggested by the peculiar brown-grey colour and the Al-rich composition obtained by SEM-EDS analysis.

25Glaze. The glaze is very thin (<0.05 mm) and irregular (Fig. 3, n. 6). The glaze-body interface is well developed. Differently from the other groups, Al-rich silicates (possibly Pb-feldspars) grew instead of K-Pb feldspars at the glaze-body interface (Fig. 5, n. 4) because of the K-poor (kaolinite-rich) body composition.

1. 1. Summary of the discriminant features

26In conclusion, simple macroscopical and microscopical features allow the distinction of the seven groups. The main of them are listed below.

27Group 1 to 3: bright orange (light yellow in thin section) glaze, body with granitoid inclusions (accessory volcanites in Group 2 and 3; finer grain size in Group 3; thin glaze in Group 2). Group 4: low-quality green to yellow glaze (bubble-rich, irregular thickness), body with fine-grained acid metamorphic rock inclusions. Group 5: brown (light yellow in thin section) glaze with abundant Fe-rich neoformed microinclusions, body with acid metamorphic and ophiolitic inclusions. Group 6: brown (yellow in thin section) glaze, body with acid metamorphic inclusions and siliceous microfossils. Group 7: very thin and irregular greyish (colourless in thin section) glaze, body with kaolinitic clay matrix and rounded quartz grains.

3. Chemical composition of ceramic bodies

28SEM-EDS chemical analyses show that the pastes of all groups are Si, Al-rich and Ca-poor (CaO<2.1 wt%; Table 4). Such compositions reflect the use of Fe-rich alluvial (Groups 1-6) or kaolinitic clays (Group 7/7*) and silicate-rich temper, confirming the thin section observation. In the ternary diagram CaO+MgO-SiO₂-Al₂O₃ (Fig. 6 above), the compositions fall in a relatively restricted field of the Ca-poor clays. In the ternary diagram CaO+MgO-Fe₂O₃-Al₂O₃ (Fig. 6 below), the higher Al₂O₃ / Fe₂O₃ ratios discriminate Group 7/7* from the other Groups.

29The Groups identified by optical microscopy show minor chemical peculiarities (Table 4) related to their mineralogical and petrographic features. While Al, Ti-rich and Fe, alkali-poor compositions clearly discriminate Group 7/7*, the other Groups show only moderate differences in alkali contents:  K values are higher in Groups 4, 5, and 6; Na values are lower in Group 4 and higher in Group 1.

30At present, no statistical treatment is possible because of the limited number of analysed samples. However, the important dispersion of the chemical results obtained by SEM-EDS even inside of most of the Groups suggests that a precise distinction of groups and origins would be difficult even improving the number of samples.

31The use of advanced chemical methods that analyse trace elements as well, such as XRF, could improve the distinction between the productions. Concerning provenance, however, little results could be obtained in this case by chemical analysis compared to optical microscopy: archaeological hypotheses about the origin are scarce and few reference materials of known production centres are available.

4. Chemical composition of glazes

32Representative glaze samples were investigated by SEM-EDS in order to both improve the technical characterisation of the artefacts and search for variations in chemical composition of glazes to be related to the typological/petrographic groups.

33All glazes show Si, Pb-rich, Ca, alkali-poor composition (Table 5) and fall in the field of the high lead glazes (Fig. 7; Tite et al., 1998). Fe, which is the only colouring agent (the greenish or yellow colours depending on reducing or oxidising conditions), shows variable contents. Both intentional opacifiers and relic components of the glaze mixture are generally absent.

34A moderate chemical zoning was generally observed, which is related to glaze-body interactions (Molera et al., 2001): in particular, Si, Al, and, in most cases, K contents increase from the surface to the base of the glaze, while Pb decreases and diffuses into the outer parts of the body enhancing the vitrification of the clay matrix.

35No clear chemical clusters are generally recognisable in the analysed glazes, even if, taking into account the typological-petrographic groups, some differences could be observed (Tab. 5, Fig. 7): in particular, with some exceptions, Groups 1, 2, 3 show higher Si contents, in Groups 5 and 6 Mg contents are higher, while Group 7-7* is characterised by higher contents of Al and lower contents of Fe.

36In conclusion, SEM-EDS analysis of glazes does not seem to be a powerful tool in order to precisely discriminate the various productions identified, even improving the number of samples. The use of more accurate and sensitive microanalytical techniques taking into account also trace elements could provide further markers in some cases, but the fairly homogeneous recipes may reduce the possibility of a clear distinction between all these groups.

5. Discussion

Provenance and diffusion

37Group 1. It can be referred to the so-called “Pseudo-ligure” type, which is diffused in a large part of the Mediterranean: from Spain to Israel, including Provence, Liguria and Tuscany (Capelli et al., 2006; Baldassarri et al., 2007). The former archaeological hypothesis of a Ligurian (Savona) origin is not supported by the analytical results. In fact, inclusions (weakly metamorphic granitoids, without Alpine metamorphic overprint) are not consistent with Savona and regional rocks (Vanossi, 1991; Giammarino et al., 2002). Moreover, fabrics are quite different from the numerous reference samples of the local ceramic production (Capelli & Cabella, 2005; Capelli et al., 2007b), represented by Group 6. Then, the atelier of Group 1 must be located in other Mediterranean productive areas (such as Spain, Provence, Tuscany and the Aegean-Anatolian sector) where acid metamorphic rocks outcrop.

38A similar hypothesis could be made for Group 2 and Group 3, which are partially comparable with Group 1. Therefore, the possibility of three different centres/workshops located in the same geological/technological area cannot be excluded.

39Group 4. Ceramics attributable to Group 4 were found at Hyères, Provence and Andora, Liguria (Capelli et al., 2006; 2007c). A Ligurian production is possible, because the metamorphic inclusions are consistent with some rock formations (Palaeozoic meta-sediments and meta-volcanites of the Briançonnais Domain) outcropping in the Finale-Noli area, western Liguria (Vanossi, 1991; Giammarino et al., 2002). Provence, Tuscany and the Aegean-Anatolian sector are among the alternative provenance areas.

40Group 5. The peculiar association of acid metamorphic and ophiolitic fragments in the ceramic paste limits the provenance investigation to few geological sectors. The integration with archaeological data points to a probable origin of the Group 5 production from the Aegean-Anatolian area. Even if also in western-central Liguria (as well as in Corse and Tuscany) both rock types outcrop, a regional origin could be excluded on chronological bases. In fact, most of the sherds of Group 5 were found in stratigraphic layers dated to the first half of the 12^th c., while Savona production is thought to start in the second half of the 12^th c. (Cabona et al., 1986).

41Group 6. The origin from Savona (western Liguria) workshops is established by the close relationships between the fabrics of Group 6 samples and kiln wasters of other ceramic classes produced in Savona (Capelli & Mannoni, 2001; Capelli et al., 2007a).

42Group 7. The typological and petrographic features of Group 7 exactly match those of the production of Uzège, Languedoc (Leenhardt, 1995), a reference sample of which (Group 7*) was analysed for comparison. The compositional features of pastes are consistent with a provenance from south-eastern France, where kaolinitic raw materials are diffused and commonly used.

Technical features

43Thin section, SEM-EDS, and XRD analyses show that all samples form a group relatively homogeneous in mineralogical and chemical composition, due to the general use of Ca-poor (Fe-rich or kaolinitic) clays with dominant silicate (quartz) inclusions (temper), which give the cooking wares a good resistance to thermal shocks (Picon & Olcese, 1994; Tite et al., 2001).

44A precise evaluation of firing temperature by XRD is difficult due to the lack of Ca-rich phases (Table 2). In most samples, however, the absence of phyllosilicate peaks and the partial or total vitrification of the clay matrix, observed by microscopy, suggest fairly high firing temperature (>900°C; Maniatis & Tite, 1981; Cultrone et al., 2001). One exception is sample 6895 of Group 5, where peaks of white mica are still present (as well as those of tremolitic amphibole, from the ophiolitic temper; however, the coarse grain of inclusions could be responsible for the preservation of mineralogical relics at temperatures above their thermal stability field). The weak hematite peaks observed in Groups 1, 5, 6 samples are related to the transformation under oxidising conditions of Fe-oxides/hydroxides scattered in the clay matrix.

45All the glazes are Pb, Si-rich and Ca, alkali-poor. Similar recipes were used for the various productions. In all cases, the clear divergence between the paste and glaze composition after subtraction of the PbO content and recasting to 100% points to the addition of Si-rich raw materials to Pb compounds (Tite et al., 1998), which are the only or main fluxing agents.

46No intentional opacifier was used. The occasional presence of cassiterite in only two samples might be explained by casual contamination of the glaze mixture.

47The slight variations in some elements showed by the glazes of a few groups could depend either on slightly different recipes for the glaze mixture or contamination processes by different ceramic bodies (Tite et al., 1998). For example, the partial assimilation of ophiolitic inclusions and kaolinitic matrix could be responsible for the higher Mg and Al contents in glazes of Group 5 and 7-7*, respectively.

48The relatively important presence of Fe (FeOwt% always >1) might be due either to impure silica sources or, less probably, to mobilisation processes from the clay body (Fe contents do not increase from the glaze rim to the glaze-body interface). The hypothesis of an intentional use of Fe as colouring agent can be excluded, as the main function of glaze in that rough cooking pottery is to impermeabilize the vessel, while appearance is less important.

49A greater degree of variability must be supposed for the atmosphere of kilns and the number of firings. The analyses of the macroscopical features (homogeneity, colour, oxidation) of glaze and body and the small-scale glaze-body interactions (extent of mobilisation of chemical elements and growth of neoformed phases) point to a single firing for Group 4 and, possibly, 7-7*, and to a double firing for Groups 1-3 (Tite et al., 1998). More doubts exist as regards Groups 5 and 6, characterised by a thick glaze-body interface, which is generally referred to single-firing processes (Molera et al., 2001). However, in this case the homogeneous oxidation of both body and glaze may suggest double firing and the presence of abundant reaction phases might be explained by slow cooling rates at high temperatures (Parmelee, 1948; Tite et al., 1998).

Conclusions

50The archaeometric study of the 12^th-13^th c. pottery assemblage of Palazzo Ducale allowed a precise typological classification, never attained before, and provided new data about provenance, diffusion and technique of Mediterranean glazed cooking wares, contributing to improved knowledge of the routes of trade and technology in a critical period of the Middle Ages.

51Because of the technical and compositional homogeneity as well as the scarceness of reference materials, XRD, SEM-EDS and other chemical analyses demonstrated in this case to be less useful than optical microscopy for the distinction of the different productions of glazed cooking wares as well as for provenance and diffusion studies.

52Fabric analyses led to the distinction of seven groups (1-7), which are different from each other in compositional, technical and typological characteristics. They can be referred to at least five workshops or production centres, three of which (Groups 1-3, 4, and 5) are the most represented in terms of sherds abundance.

53The comparative study of fabrics with both geological data and our pottery thin section database points to the import to Genoa, during the 12^th-13^th c., of cooking wares produced in the northern Mediterranean area included between Provence and the Aegean-Anatolian sector. In contrast with former archaeological hypotheses, only a minor amount of regional products was found. On the other hand, an early Savona production of glazed cooking ware (Group 6) was recognised. It came before the “Graffita arcaica tirrenica”, which started in the late 12^th c. (Varaldo, 1997). Moreover, the analyses do not exclude the possibility of another regional production (from western Liguria), started at the beginning of the 12^th c. (Group 4).

54It is noteworthy that no imports were found at Palazzo Ducale of glazed cooking wares coming from other important production areas, such as North Africa, western Sicily and southern Spain, even if numerous glazed tablewares came to Genoa from these regions in the same time span (Cabona et al., 1986).

55The results of this study may point to the existence, in most cases, of specialised productions of glazed cooking ware, also located in different workshops or production centres from those of tablewares (because of the necessity of peculiar raw materials?). In spite of the plurality of workshops, a fairly homogeneous technical knowledge in the production of glazed cooking wares was diffused in the Mediterranean during the 12^th-13^th centuries AD. Similar quartz-rich, Ca-poor raw materials were generally chosen for ceramic bodies and similar recipes (impure quartz-rich sands mixed with Pb compounds as fluxes) were used for the glazes.

This work is dedicated to Tiziano Mannoni, deceased on october 2010. He also contributed to our research with useful discussions about typological and archaeological problems. We are grateful to: A. Rabino for the help in the choice of the samples; L. Vallauri, who helped us to identify the origin of Group 7 and provided the reference sample; A. Cucchiara and L. Negretti, who performed XRD and SEM-EDS analyses; two anonymous referees for their helpful suggestions and criticism. This work has been carried out in the framework of the Research Cooperation Program for archaeometric analyses between DipTeRis and Soprintendenza per i Beni Archeologici della Liguria.