• 1 Schibille 2022; Colangeli unpublished; De Juan Ares – Schibille 2020.
• 2 De Juan Ares et al. 2019.
• 3 Berrica 2023.

2After the Arabs settled in the urban centres, glass production experienced a crisis as imports of raw natron glass from the east to the Iberian Peninsula declined. During this period of transformation, glass recycling increased steadily.1 Previous studies have identified a crisis in glass production from the second half of the 7^th century, due to a decline in imports and thus consumption in rural areas.2 However, ongoing research on the transitional period between the Visigoth and Emiral periods points to a more complex development.3 By the end of the 8^th century, there are signs of new typologies in rural areas (bottles, globular bowls with pouring spouts or molded rims, vases, large or mold-made plates and cups), replacing earlier traditions.

• 4 Berrica 2021.

3Rural areas were of course also affected by the Arab expansion and Islamization of the Iberian Peninsula from the second half of the 8^th century, and this is clearly reflected in the archaeological record that suggests changes in the manufacture of various materials, including the secondary production of glass.4 The systematic study in the mining village of the Dehesa de Navalvillar in the centre of the Iberian Peninsula offered the opportunity to identify some elements of glass processing at the end of the 8^th century and the first half of the 9^th century (fig.1).

• 5 Gratuze 2016.

5The discovery of a sizeable quantity of glass fragments located in two specific areas of the town has led to an archaeometric study of some of these samples to shed light on their compositional features and supply patterns. 15 samples (table 1) were mounted in resin blocks and repeatedly analysed (n=3) by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) at IRAMAT-CEB (Orléans, France). For the analyses, a Thermo Fisher Scientific ELEMENT XR mass spectrometer was used, equipped with a Resonetics M50E eximer laser operating at a wavelength of 193 nm, 5mJ and a repetition rate of 10 Hz. The analytical protocol used is the same as the one described by Gratuze (2016),5 which includes a pre-ablation time of 15 seconds and the acquisition of 9 mass scans over 27 seconds in counts/second for 58 elements. ²⁸Si is used as internal standard. A set of 5 external standard reference materials (NIST610, Corning B, C and D and APL1) were run at the beginning and end of the analytical sequence for the calibration and calculation of the response coefficient (K_y). Accuracy and precision are within 5% relative for most elements and within 10% for some trace elements.

• 6 López Sáez et al. 2015.
• 7 Colmenarejo – Gómez Osuna 2013; Colmenarejo – Gómez Osuna 2015; Colmenarejo – Gómez Osuna 2016;
• 8 Berrica 2021.

6Dehesa de Navalvillar is located in the area of Colmenar Viejo, Sierra de Guadarrama, Madrid, in the Basin of the Manzanares River. It is an open area of 1072 hectares, which is located at the foot of Pico San Pedro (1425 m.). The Arroyo de la Tejada is the most important source of water, as it runs through the entire meadow from north to south and is born from the Pico de San Pedro itself. Although today the Dehesa de Navalvillar is presented as an open space with little vegetation, during the medieval period it was more wooded.6 The investigation concerning the Dehesa de Navalvillar has involved the study of 13 excavation reports from two different areas7, one in the north and the other in the south, separated by the Arroyo de la Tejada; with the new territorial data that allowed us to clearly see some structures between the two excavations, recognizing at least another 43 buildings, it follows that this would be one of the largest villages found so far in the area of the Central Plateau, with at least 343 buildings (fig. 2). In previous work, this site was represented as a single large village, renamed Dehesa de Navalvillar.8

• 9 Berrica 2020.

7The village was involved in  mining and in the processing of specialized metallurgical material that was distributed at a regional level (the materials found so far are iron, copper, lead), which implies numerous specialized workmen, such as miners, transporters, specialized metallurgists, gold producers and merchants.9 So far, seven metallurgical workshops have been identified in Dehesa de Navalvillar, of which two can be linked to glass production: Metallurgical Workshop H-3.8 and Metallurgical Workshop C/E-93.

• 10 The metallurgic workshops of this article have been studied thanks to the help and collaboration w (...)
• 11 Berrica 2023.

8The metallurgical workshop originally had two previous phases: First it was built with perishable material and the metal furnace (USM 38011) was a pit dug in the ground. Then in the late 8^th century, a workshop (USM 38006) was built on a stone plinth, with adobe walls and a roof made of curved tiles. The first more primitive furnace USM 38011 has an elliptical shape and is probably double-chambered with dimensions of 30×25×10 cm. At the time of the blockage, remnants of a grey ash layer were found inside the furnace. Far more sophisticated is the USM 38006 furnace (fig. 3). It is a circular furnace built with hewn stones arranged vertically to accommodate the second floor and to which two other chambers were connected. Chamber A is where the bellows were attached to the wall to feed the combustion chamber in the lower part of the furnace. This lower chamber has a floor characterized by a platform of insulating clay and the remains of broken tiles nested with mud. The second chamber B is a small stove with a floor of fragmented tiles and an adobe dome above. The second floor is characterized by perforated adobe bricks, on which the crucibles for the production of metal and glass were probably placed. The top of the furnace was made of clay and tiles, we know this from the perpendicular collapse of the building materials found inside the furnace chamber.11

• 12 De Juan Ares et al. 2019.

9The materials directly associated with this furnace and the other from the same phase are 430 g of iron slag, two chunks of blue glass (possibly raw glass), two fragments of glass goblets and a fragment of a glass crucible (sample 7) with a three mm layer of molten glass still attached to its inside (fig. 4). A number of glass fragments were found in the vicinity of this workshop, including cups, bowls and plates with a large diameter. These data are significant because it was during the first half of the 8^th century when we find open forms of large dimensions,12 in fact, the plates of the Dehesa de Navalvillar often exceeded 20 cm in diameter, compared to the earlier period in which the plates ranged between 16 and 18 cm in diameter.

• 13 De Juan Ares et al. 2019.
• 14 Foy 2003.

10The fragments found are not particularly abundant, but the results indicate that the typologies most common in rural areas were tableware. These data are consistent with the type of glass found in other rural settlements from the Early Middle Ages in the centre of the peninsula.13 The most unusual fragment is the bluish plate 3002_13, decorated with spiral waves and made by casting or mould blowing. This is without a doubt the most decorated piece so far, together with the goblet 5204_24 (fig. 5). It is unique in the rural areas of the central Iberian Peninsula at the time. It is a tall-footed goblet with a twisted solid stem. Two glass fragments were found, belonging to Phase 3 and therefore, according to the stratigraphy of the complex, they belong to the second half of the 8th century. One of these is a solid stem of the Foy 27 type,14 where the stem is slightly modelled at the top. This workshop produced bowls and cups made of glass.

• 15 Berrica – Carrera 2020.
• 16 Azkarate Garai – Solaun 2016.

11In building E-93, 4 phases have been identified with a series of rooms that were added successively between the second half of the 8^th and the middle of the 9^th century. This workshop was also built with perishable material in its first phase, to be modified later (Phase 2). In Phase 2, room C was attached to room B (fig. 6). For this, the walls UUSSMM 931201, 931301 and 931401 were built with medium and large stones locked with mud. The room follows the natural slope of this area, and several layers of clay and stone benches were used to level the room. The roof of the building had curved tiles. In this room, a stone forge was built on the east side, with slabs 60 cm long and 20 cm high. The oven consists of two shelves, the upper one and the floor made of broken tiles stuck together with adobe. We found a hole in the wall, UEM 931301 that was used to let air into the furnace. Next to it were the remains of a bellow. In addition, it was possible to delimit the workspace, formed by a stone wall that also delimited the production area. Inside this metallurgical workshop and related to this new blast furnace, semi-finished products (iron ingots), fifteen whole or fragmented knives, iron slag and a porphyry anvil were found (fig. 6).15 Therefore, it is likely that this workshop was dedicated to the manufacture of knives, as has been observed in the Bagoeta deposits (Bizkaia, Basque Country).16

• 17 Lilliquist et al. 1993.
• 18 Foy et al. 2003; Freestone et al. 2018; Freestone 2021; Schibille 2022.

14The results of the LA-ICP-MS analysis (table 2) show that the 15 fragments can be classified as natron-type glass, although some of the samples have K₂O and MgO concentrations of more than 1.5% (fig. 10), which is normally defined as the threshold for distinguishing between natron and soda-rich plant ash glass.17 As far as is currently known, the primary production of natron glass in the first millennium CE took place exclusively in the Levant and Egypt, from where at least ten major primary glass production groups are known to have originated.18 These different glass groups have been characterised based on elements related to the silica source such as aluminium and titanium (fig. 9).

• 19 Foy et al. 2003, Ceglia et al. 2019; Schibille et al. 2016.
• 20 Freeston 2020.
• 21 Freeston 2020.
• 22 Schibille et al. 2016.

15The 15 samples from Navalvillar have highly variable features, but they can be broadly divided into Levantine and Egyptian glass groups. The samples with higher TiO₂/Al₂O₃ ratios are consistent with Egyptian glasses of the Foy 2.1 or Magby type,19 while the glass with lower TiO₂/Al₂O₃ ratios corresponds to Levantine glass from Apollonia-Arsuf.20 The Levantine glasses match the reference material quite well, but the Egyptian glasses cannot be clearly classified due to contradicting attributes. They appear closely aligned with the so-called Foy 2.1 glass type in terms of their content of alumina, titanium and zirconium (fig. 10; Table 2). Foy 2.1 was produced and circulated widely between the second half of the fifth to the sixth century CE.21 However, the samples from Navalvillar have on average somewhat higher magnesium and phosphorus contents, and to a lesser extent also potassium (fig. 10, 11). In this respect they resemble more the so-called Magby group, also an Egyptian glass that is intimately related to Foy 2.1 but that mostly dates to the late sixth and seventh century CE. On account of its higher magnesium, potassium and phosphorus levels it is assumed that these Magby glasses contain a plant ash component.22

• 23 Degryse 2014.

16The deviation of the composition from the reference data is probably related to the degree of recycling evident in the Egyptian glasses from Navalvillar (fig. 10). Recycling markers are pronounced in all of the Egyptian glasses, reflected in the colouring and opacifying elements in excess of natural impurities in sand raw materials, particularly antimony (> 30 ppm) as well as copper and lead (> 100 ppm).23 The elevated phosphorus concentrations may also have been caused in part by fuel ash vapour. However, the high magnesium clearly suggests the incorporation of an ash component. In contrast, the Levantine glasses do not show any clear signs of recycling in the form of colorant or opacifying related elements with the exception of 2 samples (009, 014). Nevertheless, the phosphorus levels are also somewhat increased in direct comparison with the reference group (fig. 11). It is therefore likely that these Levantine glasses have also been subjected to recycling or at least that the glass was exposed to fuel vapour for a prolonged period of time.

• 24 Foy 2000.
• 25 Saguí – Lepri 2016; Gómez 2012; Sánchez Pardo 2014 Sánchez Pardo 2018.
• 26 Foy 2020; Rodríguez Martorell et al. 2020.
• 27 Ferri 2021.
• 28 Foy 2000.

17Studies of glass furnaces throughout the Mediterranean have shown that there is no single type of furnace,24 but rather various circular25 or quadrangular-rectangular shapes,26 and glass furnaces have even been found positioned attached to the workshop wall.27 Therefore, it is more than likely that the differences in the furnaces are due to the requirements of the glazier and the construction possibilities, in addition to the type of specialization that the workshop had.28

• 29 Del Trente – Orlandi 2016.
• 30 Saguí – Lepri 2016.
• 31 Ferri 2021.
• 32 Munro 2010.

18In this study, however, we must take into account the fact that the furnaces at Dehesa de Navalvillar were not built specifically for glass working. Dehesa de Navalvillar was known in the region for its metallurgical furnaces, which had unprecedented blast furnace technology at a surprisingly early date (mid-8^th century) and for its prolonged craft activity. It is therefore possible that itinerant glass workers flocked to the village and made use of the available facilities. In other areas residues of glass and iron work have been found in the same workshop, although the data is quite sparse and at times inconclusive, such as at Aiano Torraccia, Chiusi;29 Crypta Balbi;30 Comacchio;31 Faragola, Monte Gelato, Pieve del Manerba, San Felice-Gravina.32 The Dehesa de Navalvillar presents a different scenario, because here it is not only about sharing the same workspace, but also about using the same furnace.

• 33 Schibille et al. 2022.

19Evidence of production has been found in the Dehesa de Navalvillar, and while not abundant, it is significant. Two fragments of the same crucible were found in workshop H3.8 with glass still adhering to the ceramic body. The glass layer (007) corresponds to the composition of the Foy 2.1 and/or Magby glass category, but has slightly higher lithium contents compared to the reference glasses (Table 2). This was observed in relation to the glass processing activities at El Tolmo de Minateda33 and appears to be indicative of contamination caused during secondary glass working and/or recycling. The Workshop C-E93 contains fragments of glass cut during the production process, a moil from a blow pipe, and glass kept in the storage areas for later recycling. In addition, they were separated by color (yellow glass and colourless glass).

• 34 Gómez 2017.
• 35 De Juan Ares et al. 2019.

21This type of moulded decoration on rural tableware is unexpected and has not been recorded outside urban centres. It has been found at Recópolis in the 6th-century stratigraphy in both closed forms such as cups, as well as open forms (e.g. plates).34 In rural areas, however, this is a novelty, because the tableware recovered in other villages in the centre of the peninsula does usually not have any decoration.35 It is therefore likely that these fragments came from an urban centre such as Recópolis or similar.

• 36 Sánchez de Pardo 2018.
• 37 De Juan Ares et al. 2019.
• 38 Freestone 2020.

22In residential building 2, only two glass fragments have been found, which can be attributed to the second half of the 8^th century. One of them is a solid stem of the Foy 27 type, where the stem is slightly modelled in the upper part. There are many examples of this type of goblet in the Iberian Peninsula during the early medieval period,36 and they are also found in rural areas in the centre of the peninsular.37 The piece has a translucent aquamarine blue-green color that may indicate a Levantine origin of the base glass.38

• 39 De Juan Ares et al. 2019.
• 40 Sánchez de Pardo 2018.
• 41 Foy 2020.

23Goblet 5204_24 is currently a unicum in the rural areas in the centre of the peninsula. It is a high-footed cup with a solid and twisted stem. The only roughly similar objects are known from Gózquez, where a slightly modelled stem has been found,39 while an example of the twisted decoration has been documented in the city of Zaragoza40 the chronology of which coincides with that of the Dehesa de Navalvillar (first half of the 8^th century). Outside the Iberian Peninsula, this decoration has come to light at Sabra al-Mansuriya in the form of stems of hanging glass lamps.41 On the whole, it cannot be considered a very common type of decoration. A dark blue thread can be seen embedded in the vitreous matrix of the goblet, but we do not know if this was a firing defect or a decoration indented by the glassmaker, as the piece is too fragmented.

• 42 Sánchez de Pardo 2018.
• 43 Gómez 2012.
• 44 Berrica 2022.
• 45 Berrica 2022.
• 46 Keller 2004.

24Of particular interest are the glass fragments that were stored in the warehouse ready to be recycled. Remains of broken glass have been found in other contexts of the Iberian Peninsula, which were preserved because they were pieces that had already fallen into disuse and were broken.42 Similar finds are known, for example, from Recópolis, where a heap of broken glass was recovered from next to the second workshop (late 7^th-early 8^th century),43 and a unique case is that of Monte de la Villa (Villaviciosa de Odón), where two handfuls of glass were kept in a cabin in a small village where no craft activity was recorded.44 This could be an important piece of information for the collection and subsequent sale of cullet to itinerant glass workers who moved from village to village to collect material for recycling. This then would seem to reflect a local or regional exchange and closed circuit of broken glass. The village of Monte de la Villa is contemporaneous with Dehesa de Navalvillar, and both villages were abandoned in the middle of the 9^th century. It is therefore a legitimate hypothesis that part of the work of itinerant glass workers was to buy up glass cullet in nearby villages and bring them to Dehesa de Navalvillar to be processed and transformed into new vitreous objects.45 Not only were these practices widespread at the local or regional levels, but the circulation of broken glass for recycling also took place on a large scale, as evidenced by the cargo of copious quantities of broken glass from sunken ships.46

• 47 Ceglia et al. 2019.

25If we assume that the glass workers sourced their material from a wide variety of contexts, we can more easily explain the different recycling patterns that we observed in the Levantine and Egyptian base glasses. This could mean that some of the recycling effects we see, such as the enhanced colouring and opacifying elements in the Egyptian glass, derive from earlier times and previous recycling processes. This also makes sense with regard to the relative availability of coloured glass, which could have ended up in the recycling pot. Foy 2.1 that has systematically higher recycling markers is a slightly older glass group,47 when coloured (Cu, Pb) and decolourised (Sb) glass may still have been more widely available. There is no increase in colourants, decolourants or opacifiers in the Levantine glass. The exception are two samples that seem to correspond to earlier Roman glass, more precisely a mixture of Roman Mn and Roman Sb glass (table 2). The Levantine samples of the Apollonia-type were not adulterated by the accidental admixture of elements related to colours or opacifiers (fig. 10a). They represent the last glass imports from the eastern Mediterranean in the sixth or perhaps seventh century CE.