Roof Tiles and Bricks of the Etruscan Domus dei Dolia (Vetulonia, Italy): An Archaeological and Archaeometric Study of Construction Materials

2.1 The Ancient City of Vetulonia and the Identification of the Domus dei Dolia

The city of Vetulonia is in central Italy, in the western side of the Italian peninsula in Southern Tuscany (Figure 1). During ancient times, the city overlooked a big navigable lagoon to the west (Colombi, 2018) and also controlled a vast territory to the north, including the so-called area of the Colline Metallifere, an important Italian mining district particularly abundant in ore natural resources (Bazzoni, Betti, Casini, Costantini, & Pagani, 2015). The most ancient Etruscan archaeological evidence goes back to the ninth–eleventh century BC, but it was between the eighth and the sixth century that Vetulonia gained importance, and it became one of the few important Etruscan city states that survived to Rome expansion until the second century BC. Nevertheless, during the first century BC, the city was involved in the Roman civil war, leading to its inexorable decline and destruction (Gregori, 1991; Semplici, 2015). The most ancient archaeological excavation at Vetulonia was carried out by Isidoro Falchi during the late nineteenth century, discovering the funerary area of the city (i.e. Villanovia phase of the Etruscan Culture, ninth–eighth century BC) in addition to the majestic tumulus tombs (Etruscan Orientalizing Period, eighth to beginning of the sixth century BC), and the Etrusco-Roman neighbourhood (Hellenistic – Late Republican periods, third–first centuries BC) of the ancient city in the area of Poggiarello Renzetti (Cygeilman, 2002).

Figure 1

Identification of the Domus dei Dolia within the Etruscan-Roman neighbourhood of Poggiarello Renzetti, Vetulonia, Italy.

Afterwards, since the middle of the twentieth century, new excavations were carried out by different scholars, bringing to light more evidence of the ancient Etruscan city of Vetulonia included in the Hellenistic (i.e. fourth–third centuries BC) and Late Republican (i.e. second–first centuries BC) periods (Cygielman, 2015; Rafanelli, 2015; Talocchini, 1981; da Vela, 2015). Just in 2009, in the Etrusco-Roman district discovered by Isidoro Falchi at Poggiarello Renzetti, the Domus dei Dolia was discovered (Agricoli, Rafanelli, & Carnevali, 2016).

The Domus is close to other residential Etruscan houses of the Hellenistic period such as the Domus n. 19, the Domus of Medea, and another Domus characterized by the presence of a vast atrium with impluvium and a large cistern (Figure 1). Consequently, the Hellenistic residential area of Poggiarello Renzetti represents the most important archaeological evidence of ancient Vetulonia, representing a period of prosperity and economic renaissance.

The results of the archaeological excavation evidenced that the Domus was destroyed by a fire, which probably also destroyed the whole neighbourhood (Cygeilman, 2016) during the Roman civil war of the first century BC. Besides, the preliminary study of pottery (i.e. black painted ware and Greek-Italic amphorae) supports this hypothesis, suggesting a chronological occupation of the house in a timeframe comprised between the third and the first century BC (Grassigli & Rafanelli, 2019; Rafanelli & Spiganti, 2019).

From the architectural point of view, the uncovered reality represents an important discovery in the field of Etruscan archaeology, as the archaeological structures are quite intact (Figure 2), and well-preserved dwellings with a high rise (i.e. 1.60 m high and 0.6 m thick) are still in place. Based on field observation, three different construction phases have been proposed by archaeologists and, until now (the excavation is still on-going), 12 different divisions of the Domus with a different function have been documented. The complete description of the Domus can be found elsewhere (Coradeschi et al., 2021; Grassigli & Rafanelli, 2019; Rafanelli & Spiganti, 2019).

Figure 2

Planimetry of the Domus dei Dolia, and the divisions brought to light between 2009 and 2015.

Different construction materials were employed to build the Domus dei Dolia. Perimetral and dividing walls were mostly constructed using locally available sandstone. Nevertheless, walls could also be constructed using stones (i.e. lower part of the wall) plus bricks or wattle and daub (i.e. higher part of the wall). These technological options have been widely attested in the Etruscan world (Amicone et al., 2020; Ceccarelli et al., 2020; Miller, 2015; Pizzirani, 2019), but they were prohibited and substituted by fired bricks in Roman territories after the first century BC (Spiganti, Trippetti, Spiganti, & Zoccoli, 2016). Wood was utilized for the construction of the room’s framing elements, for the doors, and may be for the internal support of wattle and daub (Coradeschi et al., 2021). Finally, roof tiles were used in the roof.

2.2 Geological Setting of the Area

The ancient city of Vetulonia is located on the western side of the Italian Peninsula, in the southern portion of the Tuscany region, roughly 15 km from the seacoast. It is positioned at an altitude of 320 m above sea level with the following coordinates: 42°51′34″N to 10°58′16″E. The geology of the area (Carmignani, Conti, Cornamusini, & Pirro, 2013) is the result of the collision between the African (i.e. Apulia microplate) and European (i.e. Briançonnais microplate) plates during the Tertiary. This event led to the rise of the Northern Apennines (Conti, Cornamusini, & Carmignani, 2020), and in Tuscany, it is possible to observe a complete nappe stack of it, being most geological unit completely exposed. In the area of interest (i.e. Tuscany southwest, Figure 3) the following units can be found: the metamorphic succession of the Tuscan domain, non-metamorphic succession of the Tuscan domain, the Sub-Ligurian domain, the Ligurian domains (i.e. internal and external), in addition to Neogene to Quaternary sedimentary successions. Of all these units derived by the Apulia microplate and the Thetys ocean realm (i.e. Ligurian domains), while the Sub-Ligurian domain represents a transitional domain between the oceanic and the Apulia continental crust. Neogene and quaternary intrusive and effusive acid/felsic volcanic rocks are also present close to the site.

Figure 3

Geological map of the area of Vetulonia where the archaeological site has been discovered. The geological map has been adapted after Carmignani et al. (2013).

3.1 Materials

The archaeological excavation of the Domus dei Dolia began in 2009 (it is still on-going), and the construction materials included in this study (Table 1) were recovered between 2009 and 2015. During this period, only the southern section of the Domus was excavated. Bricks were recovered inside division B, and only some of them were clearly visible during the excavation. In addition, bricks were not fired during the Etruscan period, so humidity, burial conditions, and rain affected their conservation and recovery. Nevertheless, in our case, the Domus was destroyed by a fire, and bricks were consequently partially fired (Spiganti et al., 2016), helping the preservation and recovery of some of them. Archaeological data indicated that they were utilized to complete the construction of a dividing wall made up of sandstones that separated division A from division B (Figure 4a and b). The preliminary macroscopic observation of the bricks suggested that the same raw material was probably exploited to produce them, and two samples from two different bricks were selected for analysis. The dimension of bricks could vary depending on preservation, and the registered dimensions were as follows: 20/26 cm long, 16/21 cm wide, and 11/12.5 cm thick (Pozzi, Giancarlo, Beltrame, & Coradeschi, 2016).

Figure 4

(a) Room B, bricks and roof tiles can be observed lying on the top part of the archaeological deposit, (b) bricks from room B, (c) room C, a picture representing many roof tiles completely fragmented on the archaeological deposits with traces of burning.

Roof tiles were very abundant in all divisions, but few were almost intact. Most roof tiles were extremely fractured because of the roof’s collapse (Figure 4b and c). The best-preserved planar roof tile was 71.5 cm long, 52 cm wide, and 3.5 cm thick. The raised side border was 2–3 cm wide, the internal height was 4.3 cm, and the external height was 7.6 cm. In the case of semi-circular roof tiles, just partial fragments were recovered, and they were a maximum of 19.8 cm long and 17.7 cm wide, and the chord was 10 cm maximum (Pozzi et al., 2016). Unlike bricks, planar and semi-circular roof tiles differed in terms of paste colour, suggesting that more than one raw material was probably exploited in their production. In total, 14 pan and nine (9) cover roof tiles were selected from different divisions of the Domus, being representative of the different pastes observed during sampling.

3.2 Methods

Samples were analysed using optical microscopy (OM), X-ray diffraction (XRD), and X-ray fluorescence (XRF). Principal component analysis (PCA) was used to evaluate XRF results.

4.1 OM

Five different fabrics were identified after thin section analysis (Table 2, Figure 5).

Figure 5

Fabric collected (25× magnification) at crossed polarized light. (a) Fabric A, (b) Fabric B, (c) Fabric C, (d) Fabric D, and (e) Fabric E.

Fabric A (Figure 5a). This fabric includes samples VSC-1/2/6/8/9/16/19/20/21, and they are pan roof tiles. In all cases, the ceramic paste is moderately homogeneous and buffy in colour. The colour of ceramic pastes suggests a high calcium content in the raw clay material employed to produce them. Nevertheless, this consideration will be tested/verified by mineralogical and chemical analysis in Sections 4.2 and 4.3. Porosity is abundant, and it is mainly composed of micro to macrochannels and vughs. Channels are generally weakly oriented to the outer surfaces of the fragments in thin sections. Temper (5 to 10%) is very poorly sorted in all cases, and elongated crystals are more abundant than equant ones. The degree of roundness varies from angular to sub-rounded. Grain size distribution is bimodal. Besides, two distinct grain size fractions could be clearly observed. The smaller fraction is mainly composed of sub-rounded/angular grains of coarse silt, while the coarse fraction is composed of very coarse sand to granule angular grains. From the mineralogical point of view, plagioclase feldspars and pyroxenes were dominant, with lower amounts of quartz, amphiboles (few), potassium-rich feldspar (rarely), and opaque minerals. Gabbroic rock fragments were very common. Shale, siltstone, and sandstone could also be rarely observed, in addition to some fragments of volcanic glass.

Fabric B (Figure 5b). This fabric includes samples VSC-3/4/5/7/10/14/17/18/22, and they are four pans and five cover roof tiles. In all cases, the ceramic paste is slightly homogeneous and red/brown in colour. The colour of ceramic pastes suggests high iron content in the raw clay material employed to produce them. Porosity is not very abundant, and it is mainly composed of micro to meso vesicles and vughs. Temper (5 to 20%) is very poorly sorted in all cases, with a similar concentration of elongated and equant crystals. Roundness varies from angular to sub-rounded. Grain size distribution is mostly bimodal. Besides, two distinct grain size fractions could be clearly observed with a different degree of roundness. The smaller fraction (i.e. more rounded) varies from coarse silt to very fine sand, while the coarse fraction varies from coarse sand to granule. From the mineralogical point of view, quartz, plagioclase, and pyroxene were dominant mineralogical phases compared to muscovite, potassium-rich feldspar, and amphibole minerals. Opaque minerals were also identified in thin sections. Unlike fabric A, shale, siltstone, and sandstone rock fragments were common in all cases, and gabbroic rock fragments were less represented.

Fabric C (Figure 5c). This fabric includes a sample VSC-15, and it is a pan roof tile. The ceramic paste is moderately homogeneous and brown in colour. Thus, similar to fabric B, the ceramic paste colour suggests high iron content in the raw clay material employed to produce it. Porosity is not very abundant, being mainly composed of micro to meso vesicles and vughs. Temper (20%) is very poorly sorted, predominating equant crystals. Roundness varies from angular to sub-rounded. Grain size distribution is bimodal, and two distinct grain size fractions could be clearly observed. The smaller fraction is composed of sub-rounded/angular grains of coarse silt, while the coarse fraction is composed of angular grains of coarse sand. From the mineralogical point of view, quartz, big crystals potassium-rich feldspar, and biotite mineralogical phases are dominant. Plagioclase feldspars (few), zircon (rare), and opaque minerals could also be observed. Unlike fabrics A and B, granitic rock fragments were common, but a few sandstone fragments were also observed.

Fabric D (Figure 5d). This fabric includes samples VSC-23/24, and they are two bricks. The ceramic paste is moderately homogeneous and red in colour. Thus, similar to fabrics B and C, the ceramic paste colour suggests high iron content in the raw clay material employed to produce them. Porosity is mainly composed of randomly oriented micro to macro vesicles, vughs, and channels. Temper (40–50%) is moderately sorted, with a clear predominance of equant crystals. Roundness varies from sub-rounded to rounded. Grain size distribution is unimodal in all cases, with the predominance of fine sand grains. From the mineralogical point of view, quartz, potassium-rich feldspar, and muscovite are the dominant mineralogical phases, in addition to plagioclase feldspars (few) and opaque minerals. Amongst rock fragments, shale, and sandstone fragments were very abundant.

Fabric E (Figure 5e). This fabric includes samples VSC-11/12/13, and they are semi-circular roof tiles. The ceramic paste is slightly homogeneous and red/brown in colour. Thus, similar to fabrics B, C, and D, the ceramic paste colour suggests high iron content in the raw clay material employed to produce them. Porosity is abundant, slightly alienated to samples margins in thin sections, and it is mainly composed of meso to mega vughs, channels, and planar voids. Temper (5–10%) is very poorly sorted, with a predominance of equant crystals. Roundness varies from angular to sub-rounded. Grain size distribution is bimodal, and two distinct grain size fractions could be clearly observed. The smaller fraction is composed of sub-rounded/angular grains of coarse silt, while the coarse fraction is mainly composed of sub-angular and angular grains of coarse sand. From the mineralogical point of view, quartz is the dominant mineralogical phase, with a lower amount of potassium-rich feldspar, plagioclase feldspars, amphiboles (few), pyroxenes, and opaque minerals. Shale, slate, sandstone, and quartzite were very common, and gabbroic rock fragments (rarely) were rarely identified.

Thin section evaluation evidenced a clear difference in roof tiles and bricks manufacturing technology. In the first case, fabrics A, B, C, and E were largely employed to produce roof tiles. Fabrics A and C were generally utilized to produce pan roof tiles, while fabric E was just utilized to produce cover roof tiles. On the contrary, fabric B could be utilized to produce both pan and cover roof tiles. Our observations suggest that in all cases, the raw material was partially treated, and a bigger inclusion could be added in a second moment. Besides, small and big grains do not generally sediment together, and the presence of sub-rounded inclusion in the smaller grain size fraction also points to a partial treatment of the raw material. Temper amount could change (5–20%) in this category of construction materials. Probably, it was not an important variable or, alternatively, a slightly different amount of tempering material could be added depending on the characteristics of the clayey material employed (loss of volume during drying and firing). This hypothesis would suggest that the clayey raw material employed could be different and heterogeneous for different fabrics.

In the case of bricks (Fabric D), the manufacturing technology employed was completely different. It seems that considering the amount of temper (40–50%), the raw material extracted was not treated at all or, alternatively, just the bigger inclusions were removed from the raw material. Inclusion grain size is quite uniform, and it might support this hypothesis. Maybe this was a specific technological option considering the raw material loss of volume and shrinkage during drying (Ceccarelli et al., 2020). Besides, as already stated in the introductory section, during the Etruscan period, bricks were unfired, and a significant amount of temper was probably needed to preserve the original shape of the brick during drying and to avoid shrinkage.

Regarding the origin of the raw material employed to produce roof tiles and bricks, the exploitation of three different raw materials sources is proposed considering sample characteristics and the geological setting of the area (Figure 3). The first raw material source is represented by fabrics A, B, and E. The colour of ceramic pastes (i.e. buffy, red, and red/brown) suggests that the characteristics of the clayey raw material employed could be slightly different between fabrics and that the sedimentary deposit exploited could be different or heterogeneous. Besides, the ceramic paste of CaO-rich raw materials normally became buffy after firing while, on the contrary, when CaO is not so abundant, it does not. This hypothesis will be verified by XRD and XRF analysis in Sections 4.2 and 4.3, respectively. Regarding inclusions (i.e. minerals and rock fragments), the identification of a similar mineralogical association (i.e. in all cases, mafic minerals such as plagioclase feldspars and pyroxenes were clearly identified in a thin section) suggests/indicates that a raw material with similar mafic inclusions was selected and exploited. The main difference between fabrics A, B, and E resides in the relative abundance of these minerals in sample ceramic pastes. On fabric A, mafic minerals are the dominant mineralogical phases. Conversely, from fabrics B to E, the presence of mafic minerals decreases, while the presence of mineral inclusion such as quartz, muscovite, and potassium-rich feldspars increases. Also, the identification of similar rock fragments points to an analogous interpretation. As in the previous case, mafic (i.e. gabbro rock fragments) and sedimentary (i.e. quartzite, shale, slate, and sandstone) rock fragments were observed in all cases. The presence of mafic rock fragments decreases from fabrics A to E, while the abundance of sedimentary rock fragments increases from fabrics A to E. Considering the geological setting of the area close to the archaeological site, optical microscopy results point to a compatibility with the non-metamorphic succession of the Internal Ligurian Unit (Figure 3) that outcrops in a radius of roughly 4–12 km north/north-east from the archaeological site. They are mainly composed of Jurassic ophiolites (Montanini, Travaglioli, Serri, Dostal, & Ricci, 2006) and their cretaceous sedimentary cover (Carmignani et al., 2013; Conti et al., 2020; Nirta, Pandelli, Principi, Bertini, & Cipriani, 2005). In particular, the sedimentary cover is made by heterogeneous shales (i.e. dark/grey shale, siliceous limestone, and fine-grained quartz arenite bed), including ophiolitic masses (i.e. olistoliths and sedimentary breccias) of variable dimension. Mafic minerals and rock fragments can only be found within this geological unit, and it is localized in the territory. Consequently, this material was selected and preferred (i.e. considering the number of samples included in these fabrics) to produce roof tiles, and the heterogeneity observed in fabrics A, B, and E reflects the heterogeneity of the selected deposit/s of the same geological formation.

In the case of fabric C, the minerals and rock fragments identified in thin sections (i.e. feldspars, biotite, quartz, zircon, and granite) point to compatibility with the Neogene and Quaternary acid/felsic magmatic outcrops, located at 6–8 km north-east of the archaeological site. In this case, the Gavorrano monzogranites (Musumeci, Mazzarini, Corti, Barsella, & Montanari, 2005) show a similar mineralogical association, suggesting that the alteration products of this geological formation were employed to produce the only fabric C sample. Also, in this case, the outcrop is extremely localized, and a similar mineralogy cannot be found elsewhere close to the archaeological site. Similar outcrops can only be found in the Elba, Giglio, and Montecristo Islands and north of the city of Piombino at roughly 50–60 km from the archaeological site (Carmignani et al., 2013; Conti et al., 2020).

Regarding fabric D (i.e. bricks), they are compatible with the outcrops of the non-metamorphic succession of the Tuscan domain, where the archaeological and the town of Vetulonia is located (Amendola et al., 2016; Cornamusini, Elter, & Sandrelli, 2002). Besides, the mineralogical association (i.e. quartz, feldspars, and muscovite) and the rock fragments (i.e. shale and sandstone) can be clearly correlated with this geological formation, indicating that bricks were probably produced in close proximity to the archaeological site.

Thus, our observations indicate that roof tiles were possibly not produced in place, and two different sets of raw materials were exploited. These raw materials could be found in a radius of roughly 12 km from the archaeological site. So, they were imported into the city. Considering that the Domus is not the only one within the archaeological area of Vetulonia, a massive production of roof tiles was required. Consequently, the selection of the geological deposit was likely based on factors such as distance, abundance, accessibility, space to process the raw material, fuel and water availability, and security. Differently, in the case of bricks, they were probably produced very close to the archaeological sites, using a third and completely different raw material. So, for bricks, the raw material was selected according to accessibility and local availability. The technology applied was also different. In the case of roof tiles, there was no distinction in the technology used to produce them, the typology was not an important variable and similar production criteria were employed in all cases. Differently, in the case of bricks, the amount of temper was an important issue, and it had to be high to avoid bricks shrinkage during drying to preserve the original shape. This technological option in brick manufacture has been very common (Nodarou, Frederick, & Hein, 2008) in the Mediterranean area since pre-history.

4.2 XRD

The mineralogical analysis of the samples verified optical microscopy observations, and it added new data to evaluate samples characteristics and the firing technology applied. Moreover, as already stated in previous sections, the Domus was destroyed by a fire, and roof tiles and bricks were exposed to a high temperature. Roof tiles were generally fired (Ceccarelli et al., 2020), but bricks were not (Pizzirani, 2019). So, it can be possible to understand to which extent the “destruction” event influenced the original mineralogy of the samples. Based on the results obtained, samples were grouped into five different XRD groups (Table 3), which reflect the subdivision proposed in Section 4.1.

XRD group 1. Fabric A samples are included in this group. Quartz, plagioclase feldspar, pyroxene, and akermanite/gehlenite mineralogical phases were always identified. On the contrary, potassium-rich feldspar, hematite, analcime, amphibole, and illite/muscovite mineralogical phases may or may not be present.

XRD group 2. Fabric B samples are included in this group. Quartz, plagioclase feldspar, potassium-rich feldspars, pyroxene, hematite, and illite/muscovite mineralogical phases were identified in all cases. On the contrary, calcite, and amphibole mineralogical phases may or may not be observed.

XRD group 3. Fabric C samples are included in this group. Quartz, plagioclase feldspars, potassium-rich feldspars, hematite, biotite, and illite/muscovite minerals were identified.

XRD group 4. Fabric D samples are included in this group. Quartz, plagioclase feldspar, potassium-rich feldspars, and illite/muscovite mineralogical phases were identified in all samples. Smectite and chlorite clay minerals were also identified.

XRD group 5. Fabric E samples are included in this group. Quartz, plagioclase feldspar, potassium-rich feldspars, pyroxene, and illite/muscovite mineralogical phases were identified in all cases, while hematite, calcite, and amphibole may or may not be observed.

As observed during optical microscopy analysis, XRD groups 1, 2, and 5 show a slightly different mineralogy. In XRD group 1, plagioclase feldspars and pyroxenes were observed in thin sections, but akermanite/gehlenite (i.e. sorosilicate) minerals were not. Considering ceramic pastes colour (i.e. buffy), the identification of sorosilicate minerals on XRD patterns strongly suggests that plagioclase feldspars (i.e. tectosilicate) and pyroxenes (i.e. inosilicates) also formed during burning within the ceramic paste. Besides, when the carbonate component of the original raw material is significant, high-temperature calcium/magnesium-rich mineralogical phases nucleate at 800°C (Beltrame et al., 2021; Heimann & Maggetti, 2019; Trinidade, Dias, Coroado, & Rocha, 2009). So, plagioclase feldspars and pyroxenes were included in the tempering material, but they also formed in the ceramic paste in addition to akermanite/gehlenite. The same reasoning is not valid in the case of XRD groups 2 and 5, as akermanite/gehlenite was not identified in XRD patterns, and a difference in the carbonate component within the raw material exploited can be assumed.

Moreover, considering the same XRD groups, the semi-quantification of quartz, plagioclase feldspar, and pyroxene mineralogical phases clearly change. Quartz abundance increases in XRD groups 2 and 5, while the plagioclase and pyroxene representativity decrease accordingly. These results support the observation developed in the case of XRD group 1, and the interpretation proposed in the optical microscopy section. These observations indicate that the original raw materials employed to produce XRD groups 1, 2, and 5 contained similar mineralogical species (i.e. as inclusion), but on the clayey fraction, the carbonate component could vary. Consequently, XRD patterns reflect the heterogeneity of the original geological formation exploited in their production.

For XRD groups 3 and 4, results corroborate optical microscopy observation.

Regarding firing technology, bricks and roof tile are two different classes of construction materials, and they were produced using a different technology (Bonetto, 2019; Ceccarelli et al., 2020; Pizzirani, 2019). So, different considerations can be made after evaluating XRD results to determine sample's thermal history. Besides, specific mineralogical phases might disappear or nucleate when a clayey raw material is fired. Phyllosilicates and carbonate disappear during firing depending on the maximum firing temperature attained. Illite/muscovite disappear above 950/1,000°C, smectites at 500°C, while chlorite dehydroxylation occurs around 650°C (Beltrame et al., 2020; Maritan, Nodari, Mazzoli, Milano, & Russo, 2006; Trinidade et al., 2009). Between 700 and 800°C, CO₂ is removed from carbonates, and free lime can combine with melted phyllosilicates in the formation of new calcium-rich mineralogical phases (Fabbri, Gualtieri, & Shoval, 2014; Trinidade et al., 2009), including plagioclase feldspars, pyroxenes, and akermanite/gehlenite. This reaction starts at 800°C (Heimann & Maggetti, 2014, 2019; Noll & Heimann, 2016; Trinidade et al., 2009), and it can continue until a higher temperature (i.e. more than 1,000°C). Hematite nucleates at 750°C (Beltrame et al., 2020).

Thus, the identification of clay minerals on bricks (i.e. XRD group 4) indicates that the fire that destroyed the Domus was not sufficiently strong for the clay minerals to collapse and disappear from XRD patterns. Besides, bricks were generally unfired, and the identification of clay minerals points to this interpretation. In the case of roof tiles, considering XRD results, the fire did not influence the original mineralogy because the fire temperature was too low. Consequently, the firing technology applied can be established in the case of roof tiles. Regarding XRD group 1, the firing temperature was higher than 800°C and up to 1,000°C, considering that illite/muscovite was not detected in two different samples. Differently, XRD groups 2, 3, and 5 samples were probably fired between 750 and 950°C. Besides, hematite and illite/muscovite were always detected in the analysed samples.

4.3 XRF and PCA

The chemical analysis result table is presented in a separate annexed file. The statistical treatment of sample's chemical composition by PCA extracted five different PCs, and they describe 87.43% of the total variance (Table 4). The first two PCs were the most significant, representing 58.57% of the total variance.

PCs 1 and 2 have different loadings (Table 5). In the case of PC1, the most important were represented by K₂O, TiO₂, Rb, Y, and Zr (i.e. positive), and Na₂O, CaO, MgO, and Sr (i.e. negative). They describe the abundance of these oxides/elements inside the raw material exploited. K₂O and Rb can be hosted both by tectosilicates (i.e. feldspars) and by phyllosilicates (i.e. illite/muscovite, biotite). They were both identified during optical microscopy (as lithoclasts or individual minerals) and XRD analysis. Titanium oxide, yttrium, and zircon chemical elements are normally hosted by different accessory minerals. Amongst them, oxides (i.e. ilmenite, rutile, anatase) or nesosilicates (i.e. zircon) can be mentioned. These oxides/elements can be present in the finer grain size fraction of a sediment, sedimentary, and felsic/acid rocks. CaO, MgO, and Sr might represent samples carbonate component, but these oxides (i.e. including Na₂O)/chemical elements can also be hosted by some mafic minerals, such as tectosilicates and inosilicates (i.e. plagioclase feldspars, pyroxenes, and amphiboles). So, PC1 describes the variation of the sedimentary/felsic vs the carbonate/mafic components inside the analysed samples, and these characteristics are widely discussed in Sections 4.1 and 4.2.

Regarding PC2, significant positive loading has been observed in the case of Al₂O₃, P₂O₅, Fe₂O₃, U, and LOI, contrasted by a negative loading of SiO₂ (Table 4). While the variability of uranium and phosphorous is not clear, Al₂O₃, Fe₂O₃, and LOI give important indications. Aluminium and iron oxides are important components in phyllosilicates, while LOI represents hydrous minerals (i.e. mainly phyllosilicates and amphiboles). SiO₂ (i.e. silica) can be included inside different mineralogical phases, and quartz (i.e. tectosilicates) is probably the most significant in this case. So, PC2 describes the amount of hydrous minerals (i.e. mainly phyllosilicates) compared to the amount of quartz (i.e. tectosilicates) inside the analysed samples. Thus, PC2 describes the clay to temper ratio. Nevertheless, in our case, the clay-to-temper ratio was not established, and only an estimation of temper by image analysis was included in the optical microscopy section.

Thus, the biplot (Figure 6) clearly summarizes sample subdivision based on their characteristics, and by combining OM and XRD results, the following observation can be made on fabrics:

1. The chemical composition of fabrics A, B, and E (i.e. XRD groups 1, 2, and 5) samples is explained by the variation of the carbonate component inside the ceramic paste, the variation of mafic and sedimentary minerals/rock fragments, and the clay-to-temper ratio.

2. Fabric A (i.e. XRD group 1) samples are rather uniform in chemical composition, and the carbonate/mafic component was extremely important in this case.

3. Fabric B (i.e. XRD group 2) samples show stronger variability along the PC2 axe. Besides, temper evaluation by 2D image analysis also shows the biggest temper concentration range (i.e. 5–20%). Moreover, the importance of the carbonate/mafic components decreases when compared to fabric A samples.

4. Fabric E (i.e. XRD group 5) samples are rather uniform in chemical composition, and the abundance of clay, sedimentary minerals, and rock fragments were important components in this case.

5. The chemical composition of fabric C (i.e. XRD group 3) samples is explained by the abundance of felsic mineral and rock fragments. Besides, quartz, biotite, and potassium-rich feldspars were important mineralogical phases, as evidenced by optical microscopy and XRD results. Also, the zircon mineral was an important component of the raw material exploited.

6. The chemical composition of fabric D (i.e. XRD group 4) samples is explained by the abundance of quartz and sedimentary rock fragments and minerals. As evidenced during optical microscopy observation, this fabric was extremely abundant in temper.

Figure 6

Bi-plot of PC1 and PC2. Blue axes represent PC1 and PC2 loadings for all variables presented in Table 4, while black axes represent sample scores for PC1 and PC2.