Introduction

Recent years have witnessed a substantial increase in our understanding of glassmaking in medieval Islamic Iberia (al-Andalus), partially redressing the imbalance between the number of glass samples subjected to quantitative chemical analysis in Islamic Spain and Portugal and those from other contemporary regions under Islamic rule. This is important due to the dynamic, and rather unique political position of the various regions of al-Andalus compared to other Islamic polities in north Africa and the so-called “Islamic heartlands” (Milwright 2010), but also with the Christian kingdoms to the north, to the Pyrenees and beyond. Important assemblages have been published in recent years, from 9th-12th-century Córdoba (Duckworth et al 2015; De Juan et al 2021), Vascos (coordinates: 39° 45′ 21.5634"; -5° 5′ 15.2154") (De Juan and Schibille 2017) and Murcia (Carmona et al 2009). The dataset reported here will add to these few but important contributions to our previously sparse compositional evidence. New patterns are still emerging, as we move towards a more chronologically and spatially refined understanding of glass production, circulation, and consumption over time in al-Andalus.

The assemblage presented here is particularly relevant since Murcia is known to have been an important hub for glass production during the 12th and 13th centuries. Archaeology has attested to the existence of several glass workshops in the city and its hinterland during this period (Baños and Sánchez 2005; Jiménez et al 1998), and the excavation of San Esteban may have added to this evidence (Velo et al 2024). Evidence for this is also found in the surviving written records; in a comment copied by al-Maqqarī, Ibn Sa’īd al-Maghribī highlights glass craftsmanship in several cities in al-Andalus during the mid-thirteenth century, including Murcia (Gayangos 1840–43: 51; 93) (Fig. 1). Chemical data on glass from Murcia was published by Carmona et al. (2009), but this is the first time that glass excavated here has been subjected to trace element analysis, which is proving to be a very powerful tool for their characterisation.

Fig. 1
figure 1

Map with the location of the main sites mentioned in the text

Full size image

Archaeological context and materials

The extensive (c. 30 ha) arrabal of Arrixaca (al-Rišaqa) (coordinates: 37° 59′ 19.8744"; -1° 8′ 0.276") formed in the 11th century outside the northern and western fronts of the city walls of the medina of Mursiya, before undergoing substantial growth in the Almoravid period (mid-11th to mid-12th century). Nonetheless, even by the time of the Christian conquest in 1243, the arrabal included still some unbuilt areas within it. These were occupied by orchards and vegetable gardens, as confirmed by both archaeology (e.g. Navarro and Robles 1999) and the written record. The latter includes the repartimiento de Murcia, the detailed inventory of buildings and plots of land carried out by Castilian officials to organise the allocation of property with which the Crown rewarded members of the conquering host (nobles, military orders, urban militias from Castilian cities, etc.) (Torres 1960: 230).

Rescue excavations in the modern area of San Esteban, which covers a significant portion of the arrabal of Arrixaca, began in 2009. It soon became apparent that the site was among the most representative archaeological spaces of Islamic Murcia, due to its size and the diversity of archaeological features identified within it (houses, commercial buildings, cemeteries). It is currently under systematic investigation by the University of Murcia, and excavations are ongoing (Fig. 2) (Eiroa et al 2021).

Fig. 2
figure 2

Top view of the excavations of the arrabal of San Esteban

Full size image

Glass and vitreous remains from three areas of the site were sampled for this study:

  1. (1)

    “Recinto I” a large building complex, identified as a funduq, a public building to house foreign merchants and their wares, active in the 12th and 13th centuries;

  2. (2)

    “Recinto II”, a large residence contemporary with the funduq;

  3. (3)

    A maqbara, which has yielded over 30 individual burials.

Two samples (SEB22 and SEB25) were taken from items found in the foundation layers of a Late Medieval Christian chapel and could potentially post-date the 13th-century conquest (Fig. 3).

Fig. 3
figure 3

Plan of the excavations of the arrabal of San Esteban-Arrixaca, with the different excavation sectors

Full size image

In addition to the potential remains of glass-working at the site, discussed below, excavations uncovered abundant dumps of ceramic wasters, including a significant number of crucible fragments used to melt lead glazes. The western edge of the arrabal of Arrixaca is known to have housed extensive pottery-production activity (González Ballesteros and Muñoz Espinosa 2024).

The materials examined in this paper include 36 glass samples taken from 35 archaeological objects excavated during the 2018 and 2019 seasons. They include vessel glass, possible glass production waste, and raw glass (Fig. 4). Some objects could not be typologically identified owing to their small size, but were nevertheless sampled because of their colour/opacity, which set them aside from the bulk of glass items found at the site, or their archaeological context.

Fig. 4
figure 4

Archaeological items from which the analysed samples were taken: small vessel glass fragments (top), potential glass-working remains (middle), incomplete shapes (bottom; see Table 1)

Full size image

Others could be confidently linked to various types of glassware, including bottles, jugs, flasks, lamps, and drinking cups. Production remains include drop-like glass lumps, chunks of raw glass (a glass fragment that is the result of melting primary glass from the raw materials or remelting cullet, rather than a piece of glass that has been manipulated into a defined shape, like a flask), and a moil (the glass crown that remains attached to the blowing iron after the vessel is cracked off to work the rim) (see Table 1). The full glass assemblage for the excavation seasons of 2018 and 2019 includes a total of 294 glass fragments, so our samples account for just over 10% of the total. Material from secure archaeological contexts was prioritised, and almost all typologically identifiable fragments from the medieval phases were sampled, so the assemblage considered here is regarded as a broadly representative sample of glass circulation in 12th-13th-century Murcia, at least in those areas excavated to date.

Table 1 Archaeological items sampled for this study, including detailed typological characterisation and archaeological context
Full size table

Most objects appear on sight to be ‘naturally coloured’, that is, that their colour is the result of the impurities added alongside the silica source (see below), predominantly pale yellow and bluish green, although there are also a number of green and colourless specimens. In addition, there are three specimens of opaque glass (one white and two pale blue), a not uncommon feature of medieval Islamic glasses, including those excavated at the workshop of Puxmarina, barely a few hundred meters from San Esteban (Jiménez et al 1998).

All fragments excluding those thought to be production remains were of blown vessel glass, including some mould-blown examples. Mould-blowing was a common decorative technique in medieval Islamic glasses (Whitehouse 1993; 2001) and has been frequently attested in al-Andalus, suggesting local production from as early as the 9th-10th century (Rontomé 2000). In addition, one fragment, identified as a handle, was stamped with a concentric circle pattern when the glass was still hot.

Methods

The samples were taken manually by Almudena Velo at Laboratorio de Arqueología, Universidad de Murcia in 2019 and mounted at the Tephrachronology Laboratory, Research Laboratory for Archaeology and the History of Art, University of Oxford, by Victoria Sainsbury. They are approximately 4 mm wide and, in the case of vessel glass, as thick as the original item from which they were taken. Samples were mounted in epoxy resin, ground, polished (< 1 μm diamond paste), and analysed by electron microprobe analysis (EMPA) and laser-ablated inductively coupled plasma mass spectrometry (LA-ICP-MS).

For the determination of major and minor elements present in the glass (reported as weight percentage oxides), EMPA analysis was conducted in collaboration with Victoria Smith at the Tephrachronology Laboratory, Research Laboratory for Archaeology and the History of Art, University of Oxford. The instrument used was a JEOL-8600 wavelength dispersive electron microprobe with 15 kV accelerating voltage, 7 nA current and 10 μm diameter beam. The microprobe was calibrated using mineral standards and quantified using the PAP absorption correction method. Corning reference glasses (A, B, C, and D) were analysed, and the results compared with published compositions (Vicenzi et al. 2002; Wagner et al 2012) to verify the accuracy of the instrumental setup. Samples were carbon coated prior to analysis to prevent charging. Peak counting times were 20 s for calcium and potassium; 30 s for silicon, aluminium, and magnesium; 40 s for iron; 50 s for manganese and lead; 60 s for tin and antimony; and 80 s for phosphorus and copper. The results presented are the mean of three (in some samples four, see Online Resource 1-Tab 2) separate readings taken in different positions of each sample.

In order to detect trace elements (reported as parts per million), LA-ICP-MS analysis was conducted by Simon Chenery at the Centre for Environmental Geochemistry, British Geological Survey (Keyworth). Owing to logistical issues, only 24 samples could be analyzed for trace elements. The carbon coating was removed from the sample blocks using alcohol wipes. A NewWave FX 193 nm excimer laser with integral microscope and ablation cell was coupled to an Agilent 7500c series ICP-MS using a helium gas flow. It was calibrated using SRM610 glass (NIST, USA), and quality assessed using SRM612 glass. Internal calibration of the ICP-MS data was realized using the silica (SiO2) results obtained by EMPA. Each sample was ablated three times, and the results presented are the average of these.

Average EMPA and LA-ICP-MS data compared with published values for Corning A, B, C, D and NIST SRM 612 glass standards can be found in Table 2. In some elements (aluminium, lead, soda, iron, magnesia, and titania), the measurement of some reference materials yielded an error of over ± 10% relative with regard to published and certified results. In order to test the reliability of our measurements, we compared our EMPA and LA-ICP-MS (wherever available) results (see Online Resource 1 – Tab 4). With only a few exceptions, the measurements undertaken with both techniques were fairly close, suggesting that our measurements were indeed fit for purpose and that, with due caution, they could be used as a reliable basis for interpretation.

Table 2 Average EMPA and LA-ICP-MS data compared with published values for Corning A, B, C, D and NIST SRM 612 glass standards. Corning A, B, C, D data after Vicenzi et al. (2002) and Wagner et al. (2012) and NIST SRM 612 data after Jochum et al. (2011)
Full size table

Before presenting the results, it must first be noted that none of the glasses analysed correspond to the lead-soda-silica glass reported over the last decade at a number of Andalusi sites, primarily in Córdoba in 10th-11th-century contexts (Duckworth et al 2015; De Juan and Schibille 2017; De Juan et al 2021).Footnote 1 When these were first reported, the possibility that they reflected the addition of metallic lead in the form of calcined litharge to a preformed glass was pointed out (Duckworth et al 2015: 38), a practice known in other Islamic regions (Sayre and Smith 1967: 303). To explore and account for the likely possibility that lead was added to a pre-formed ‘base’ glass rather than included as a raw ingredient in primary glassmaking, unless otherwise stated, the compositions reported are the result of omitting the lead oxide component and the normalisation of the contents for all other elements to 100%.

Results

The results of our EMPA and LA-ICP-MS measurements and the means used for interpretation can be viewed in Online Resource 1 – Tab 2.

All the samples correspond to soda-lime-silica glass. With the exception of SEB20, which will be discussed separately, soda is the main fluxing agent present in all samples, with results ranging 20.35%-12.97% Na2O in non-high-lead glasses, and 14.09%-11.41% Na2O in high-lead glasses. The main network former, silica, ranges from 67.93% to 55.34% SiO2 in non-high-lead glasses and 53.23%-42.39% SiO2 in high-lead glasses. Finally, the chemistry of the glass is stabilised with lime, which in our samples ranges 12.09%-5.06% CaO in non-high-lead glasses and 5.27%-3.16% CaO in high-lead-glasses. This interpretation of the data assumes that lead was not in this instance deliberately used as a flux, a practice attested elsewhere in al-Andalus (Schibille et al 2020). All the concentrations reported in this paragraph are the un-normalised concentrations prior to the removal of lead contents.

Silica and other elements introduced in the glass with the silica source, especially alumina, iron, and titania (which are indicative of the level of purity/maturity of the silica sources), are often of great use to distinguish glass provenance, as they reflect the geological makeup of the silica source employed as raw material for the glass. All of the glasses from San Esteban likely used sand as a source of silica, based on their relatively high Sr content (> 200 ppm Sr), which seems to discard the use of crushed quartz (Brems et al 2014: 53). It must, however, be noted that strontium typically pairs with lime, so in plant-ash glasses such as those found in Iberian Islamic contexts it can also be introduced in the glass with the fluxes, and some samples present Sr contents barely above this cut-off mark – SEB10 (215 ppm Sr) and SEB11 (214 Sr), so the possibility that these used crushed quartz as a source of silica must remain open.

Both alumina and titanium are highly variable in the assemblage (ranging 0.80%-6.65% Al2O3 and 0.06%-0.47% TiO2). The presence of iron is also variable (0.21%-2.16% Fe2O3), although it is fairly low in most samples (< 1% Fe2O3). Generally, these three elements are strongly positively correlated (Al2O3:Fe2O3 = corr. coeff. 0.76; Al2O3:TiO2 = 0.93; TiO2:Fe2O3 = 0.69), suggesting that they can be used as a reliable marker of provenance. It is possible to broadly distinguish two groups in the assemblage, one with relatively low alumina (< 2.2% Al2O3) and titania (< 0.15% TiO2), and another with markedly elevated levels – > 3% Al2O3 and > 0.15% TiO2 – of these elements compared with other early Islamic assemblages, specifically from Tyre (Phelps 2018), Mesopotamia (Schibille et al. 2018), and Egypt (Schibille et al. 2019). These values are, however, common in Andalusi assemblages, as illustrated in Fig. 5. For the relationship of these two broad groups with published Iberian compositional groups, see below.

Fig. 5
figure 5

Elements brought into glass with the silica source can help to distinguish between glasses with different provenances. These scatter graphs compare the alumina and titania (a) and alumina and iron (b) contents of the glasses in the San Esteban assemblage with other published assemblages from the early Islamic Levant (Phelps 2018), Egypt (Schibille et al. 2019), and Mesopotamia (Schibille et al. 2018), 11th-century Vascos (De Juan and Schibille 2017), 9th-12th century Cordoba (Duckworth et al. 2015), and the 12th-century glass workshop of Puxmarina (Carmona et al. 2009). V1-5 refer to de Juan et al.’s (2017) Iberian groups 1–5

Full size image

Another key set of variables are those that reflect the fluxes used in the batch. It is by now widely accepted that, by approximately the turn of the 9th century, mineral fluxes, which had been used as primary flux for glass throughout the Roman period, began to be replaced with plant ash-based fluxes in the Mediterranean basin and continental Europe (Shortland et al. 2006). Typically, the use of plant ash-based fluxes is identified through the presence of potash and magnesia (which appear only in small concentrations in mineral fluxes) in quantities upwards of ≈1.5% in the final glass. The introduction of these ashes resulted in a wider range of flux-related elements in glass production than was typically the case with Roman glass, as the soda-based mineral fluxes used until then were more predictable in terms of chemical composition, as well as displaying less impurities such as magnesia and potash. The environmental factors that can cause variations in the composition of plant ashes are multiple, from the underlying geology to rainfall patterns (Larcher 2003), and determining the specific species used can thus be challenging.

All the samples from San Esteban, except for SEB20 (which, as noted, is examined separately) and SEB11 (see below), display magnesia and potash concentrations that suggest the use of plant ash as flux (ranging from 2.40%-6.16% MgO to 1.46%-3.22% K2O respectively). A significant number of them present fairly high soda contents compared to other early Islamic assemblages – overall, the assemblage presents 17.48% Na2O on average, compared to 12.15% Na2O in Tyre, 14.56% Na2O in Samarra (excluding natron glasses, see below), and 13.02% Na2O in Egypt (again, excluding natron glasses) – and comparable potash contents, leading to a high soda to potash ratio, another feature commonly found in Iberian Islamic contexts, as illustrated in Fig. 6. SEB11, on the other hand, presents much lower magnesia (0.58% MgO) and potash (0.31% K2O), as well as much lower phosphorus (0.07% P2O5) than are typically found in plant ash glasses, so it is concluded that this was a natron glass, a type that is still found in early Islamic assemblages in Mesopotamia and Egypt. All the concentrations reported in this paragraph are the un-normalised concentrations prior to the removal of lead contents.

Fig. 6
figure 6

Soda and potash contents in plant ash-fluxed glasses are indicative of the plant species uses. In this scatter graph, the glasses of San Esteban are compared with other published assemblages from the early Islamic Levant (Phelps 2018), Egypt (Schibille et al. 2019), and Mesopotamia (Schibille et al. 2018), 11th-century Vascos (De Juan and Schibille 2017), 9th-12th century Cordoba (Duckworth et al. 2015), and the 12th-century glass workshop of Puxmarina (Carmona et al. 2009). V1-5 refer to de Juan et al.’s (2017) Iberian groups 1–5

Full size image

Concerning lead and other elements used historically as opacifiers and colouring and decolouring agents in glass (basically, cobalt, copper, tin, and antimony, in addition to lead), several of the glasses from San Esteban were found to have fairly high lead contents – SEB2 (17.10% PbO); SEB10 (14.54% PbO); SEB18 (15.66% PbO) – though still too low to fall into the prominent group of lead-soda-silica glass found in a number of Andalusi contexts. Many other samples present significant minor quantities of lead, always below 0.5% PbO, except for SEB5 (1.77%). The high-lead glasses include one white (SEB2) and two light blue (SEB10 and SEB18) opaque glasses. The white colour of SEB2 is likely the product of the addition of tin (9.97% SnO2), while SEB10 and SEB18 are rendered opaque pale blue by the addition of copper (1.43% and 1.00% CuO respectively) and tin (9.08% and 8.84% SnO2 respectively), as they present almost undetectable quantities of cobalt (3 and 2 ppm Co respectively), another usual ingredient in blue glasses.

Concerning non-high-lead glasses, cobalt is virtually absent from the assemblage and copper is present in significant but still very small quantities in most samples, although the concentration present in SEB13 (0.83% CuO), could be argued to have been added deliberately as a colourant. This is, however, a fragment of production waste, so special caution is required. Tin is only present in significant quantities (above 100 ppm Sn) in three non-high-lead samples (SEB4a, SEB4b, and SEB17). SEB5, in addition to high relative lead and copper (0.11% CuO), presents significant traces of antimony (284 ppm Sb), which is otherwise only present in negligible quantities in the assemblage.

Discussion. Provenance and technological inferences

For the interpretation of this assemblage, we shall follow the ‘compositional group’ paradigm that has dominated the chemical approach to the study of Roman and late Roman glass, although we have some reservations about this approach being equally valid for medieval glass for a number of reasons. The tight compositional groups that characterise Roman and late Roman glass were the result of the narrow range of raw materials used, notably Egyptian natron and a relatively limited selection of sands in Egypt and the Levant. However, the use of plant ash-based fluxes, different varieties of which can be found pretty much everywhere, multiplied the options for primary glassmaking, theretofore limited to areas near the sources of natron, which added new silica sources into the mix, not to mention the use of much more variable fluxes (see above). Recent work on western Mediterranean beach sands has established that most sands are not suitable for glassmaking in the Roman tradition, among other reasons because of the low lime content of many sands (Brems et al. 2012, 35–36). This, however, could have been compensated by the addition of calcium carbonate in the form of crushed seashells, and it must also be taken into account that the plant ashes used as a flux in medieval Mediterranean glassmaking are lime-rich, also compensating for low calcium levels in the silica source and thus expanding the range of sands suitable for making glass.

Despite this caveat, San Esteban samples SEB5, SEB12, SEB14, SEB15, SEB19, SEB21, SEB22, SEB25, SEB26, SEB28, SEB32, SEB33, SEB34 and SEB35 can be argued to be Iberian in origin on account of the high concentrations of impurities in their silica source (especially alumina), as well as by other silica-related variables, such as the Th/Zr:La/TiO2 ratio (Fig. 7a), which clearly single them out from the glass found in other early Islamic contexts further east. These samples encompass a miscellaneous collection of objects, including blown shapes (yellow, green, bluish green, and colourless) and three fragments of production waste. Meanwhile, the high soda to potash ratio of samples SEB1, SEB2, SEB3, SEB4a, SEB4b, SEB6, SEB7, SEB8, SEB9, SEB13, SEB17, SEB24, SEB29, and SEB31 clearly sets them apart from eastern glasses, as confirmed also by their intermediate contents of alkaline earth metals (Fig. 7b) and by high boron and lithium to soda ratios (Fig. 7c). All of these glasses come from pale yellow blown vessels, except for SEB5 and SEB31, which are blue-green, and SEB13 and SEB24, which come from fragments of production waste.

Fig. 7
figure 7

Elements related to the sources of silica and fluxes can be used to distinguish Andalusi glasses from those produced in other early medieval regions under Islam including the Th/Zr:La/TiO2 ratio (a), the lime and magnesia contents (b), and the boron and lithium to soda ratios (c). It is important to note that only some of the San Esteban glasses presented in this paper are expressed in these graphs, as the trace element data was unavailable for a number of samples. V1-5 refer to de Juan et al.’s (2017) Iberian groups 1–5

Full size image

SEB11, on the other hand, presents an intermediate concentrations of alumina (2.82% Al2O3), titania (0.24% TiO2), and iron (0.86% Fe2O3) as well as Th/Zr:La/TiO2 and lithium and boron to soda ratios compatible with some Egyptian glasses reported by Schibille et al. as “plant ash E1” and “plant ash E2” (2019), in addition to being the only sample in the assemblage with magnesia and potash contents consistent with the use of mineral fluxes, a type of glass still relatively abundant in Egypt in the early Islamic period. Samples SEB27 and SEB30, meanwhile, present the low alumina (1.15% and 1.43% Al2O3 respectively), titania (0.06% TiO2 in both instances), and iron (0.58% and 0.21% Fe2O3 respectively) found in some early Islamic Mesopotamian glasses, specifically groups 1 and 2 of the glasses from Samarra reported by Schibille et al (2018), and the former is also compatible with these glasses in terms of potash to soda ratio. The potash and soda contents of SEB30 (2.15% K2O and 13.10% Na2O) are, however, somewhat lower than in other Mesopotamian glasses, although it must be taken into account that this sample corresponds to a gold leaf-decorated glass vessel, so a Mesopotamian provenance, like with the only other gold leaf-decorated sample analysed to date (De Juan et al. 2021, 9–10; for this decorative technique in al-Andalus see Velo et al. 2022), seems likely.

The small number of glass samples that can be identified as imports can be surprising, in light of a recently analysed assemblage of glass from Umayyad Córdoba (also from an area of arrabal comparable to San Esteban) (de Juan et al 2021), which was shown to include a significant proportion of samples whose composition suggested production in various distant regions under Islam, namely Mesopotamia, the coast of the Levant, and Egypt. The issue is of particular interest for Murcia in the 12th-13th century, because Ibn Mardanīš (r. H542-568/AD1147-1172) is known to have drawn close political and commercial links with eastern Mediterranean Islamic polities to find support against his Almohad enemies in the Iberian Peninsula, resulting in increased trade relations with this region (Eiroa and Gómez 2019, 21–5). The small number of imports identified in Murcia could suggest a greater development of the local glass industry by the 12th-13th centuries than during the period of the Cordoba Caliphate, but it seems more likely that the number of imports detected in Córdoba has more to do with the fact that the study specifically targeted decorated glass shapes, which are more likely to be the subject of long-distance trade than plain glass, and one of the seemingly imported fragments in San Esteban is indeed a gold-leaf decorated vessel. There is, however, enough evidence, supported by our results, to indicate that as far as everyday use glass was concerned, 12th- and 13th-century Murcia was, at the very least, self-sufficient.

Concerning the high-lead glasses, it must be recalled that these have long been recognised as a feature of glass production in Islamicate regions during the Middle Ages (Sayre and Smith 1961), and different traditions of soda-ash lead glass with lead contents sometimes in excess of 60% PbO have been identified (Brill 2001, 28–9; 2009). One of these traditions appear to have emerged in caliphate-period Córdoba (10th-11th centuries). None of the glasses from San Esteban match the lead contents of these (roughly 40–45% PbO) or of the oriental glasses reported by Brill. It is to be noted that two very similar high-lead glasses (Table 3), in this instance opaque emerald green, were reported by Carmona et al. (2009) from the glass workshop of Puxmarina, so it can be argued that these productions were not an exceptional feature in Murcian glassmaking during this period. The main differences in the Puxmarina and San Esteban assemblages are elements brought in by the source of silica, which, as noted, are fairly variable in Iberian glasses.

Table 3 Comparison of the averages of the main elements found in the opaque high-lead glasses from the workshop of Puxmarina (Carmona et al. 2009), and those from San Esteban
Full size table

In a previous study, De Juan and Schibille (2017) suggested a number of compositional groups for Iberian Andalusi glasses, based to a large extent on silica-related impurities. Groups 1, 2, and 3 are characterised by fairly low impurities such as alumina and titania (roughly ranging 1.00%-2.00% Al2O3 and 0.05%-0.10% TiO2), in contrast to groups 4 and 5, which present substantially higher impurities. Group 1 can be distinguished from groups 2 and 3 based on the zirconium contents, which are not only significantly higher, but which also present a much wider range (c. 75–175 ppm Zr vs c. 10–50 ppm Zr in groups 2 and 3). Groups 1 and 2 also present significantly higher silica, lime, and manganese and lower soda and magnesia than groups 3, 4 and 5 (≈66–68% vs ≈58–63% SiO2; ≈8–9% vs 6–7% CaO; ≈1% vs ≈0.5% MnO; ≈12–14% vs ≈17–19% Na2O; ≈2–3% vs ≈4–5% MgO). Finally, groups 1 to 3 also present elevated strontium compared to groups 4 and 5 (≈450–550 ppm vs ≈350 ppm Sr), while group 2 presents fairly low lithium compared to the other groups (≈5 ppm vs ≈50 ppm Li) (averages for these groups can be found in Online Resource 1 – Tab 1).

Group 3 (group 2 has not been identified in San Esteban) is particularly interesting, as it not only yielded very low values for sand impurities, suggesting the use of an artificially purified or a fairly mature source of sand, but also for presenting very narrow ranges of these and other silica-related elements (e.g. Th, Zr) (Fig. 8a and b), indicating a careful sourcing/processing of raw materials. Group 3 is represented by samples SEB1, SEB3, SEB4a, SEB4b, SEB6, SEB7, SEB8, SEB9, SEB17, and SEB19. Significantly, the normalised Al2O3, TiO2, Th and Zr content of all the lead-soda-silica glasses (and some of the low-lead glasses too) analysed in Córdoba in previous studies (including seven out of seven samples from the Umayyad palatine city of Madīnat al-Zāhra, the publication of which – under title Neither of the East Nor of the West – is in preparation by Chloe Duckworth) are directly comparable to this group 3, which could have major implications for the interpretation of glass production in 10th- to 12th-century al-Andalus.

Fig. 8
figure 8

Silica-related elements have been used to define several compositional groups of Andalusi glass, two of which (groups 2 and 3) present very narrow ranges in terms of Al2O3 and TiO2 (a) and Th and Zr (b) suggesting careful sourcing/processing of raw materials. In these figures the San Esteban samples are compared with the assemblages from Vascos (De Juan and Schibille 2017), Murcia-Puxmarina (Carmona et al. 2009), and Córdoba low lead and high lead (normalised) (de Juan et al. 2021). It is noted that some of the samples reported by De Juan and Schibille (2017) as group 3 in their additional material may need reaccommodating with group 2. V1-5 refer to de Juan et al.’s (2017) Iberian groups 1–5

Full size image

All of the glasses assigned to group 3 are pale yellow in colour. Although their iron content is relatively low (0.48% to 0.79% Fe2O3), they present a variable but not particularly high concentration of manganese (ranging from 0.29% to 0.80% MnO), so the ‘natural’ colouring effect conferred by the iron impurities is only partially offset by the presence of manganese (other decolourants, such as antimony, are virtually absent) rendering the glass a pale yellow rather than green.

Significantly, the chronological range of the lead-soda-silica glasses (the 10th and 11th centuries) is, based on the evidence, shorter than that of the base glass to which lead was likely added and seems to closely coincide with the heyday of the Umayyad caliphate in Córdoba. It would be neglectful not to examine the possibility that these lead-soda-silica glasses were in some way connected with the state-controlled industrial facilities (ṭirāz) set up by the Umayyad caliphate in Córdoba in the 10th and 11th centuries (López-Marigorta 2023a). These royal workshops are known to have engaged in pyrotechnological activity, specifically with the production of ‘green and manganese’ tin-glaze ceramics for the royal house and its retainers (Barceló 1993), and the possibility that they also produced glass must not be ruled out; after all, these glasses appear to be restricted to the 10th and 11th century and to be proportionally more abundant, by a long way, in Madīnat al-Zahrā (37° 53′ 10.4274"; -4° 52′ 3.864"), the caliph’s palatial city near Cordoba (more in Neither of the East Nor of the West, see above) and Córdoba itself (Duckworth et al. 2015; De Juan et al. 2021) than anywhere else. Significantly, these glasses consistently present higher soda/total flux ratios than other Andalusi glasses, also suggesting the use of purer/purified ashes, again increasing the impression that these were high-end glasses (for the purification of plant ash see Cagno et al. 2008; 2012a; b). The ṭirāz system was characterised by the manufacture of products with a relatively wide, but invariably high, range of qualities (López-Marigorta 2023b). Let us also recall that Cordoba and Madīnat al-Zahrā housed mints that struck millions of silver dirhems per year, and that the cupellation of galena must have been carried out there in an industrial scale (Grañeda 1998; 1999), so there must have been abundant lead available to add to the glass; recent isotopic analyses have shown that the glass added to earlier Cordoban glasses (8th-9th centuries) came from local ore sources (Schibille et al. 2020, 16246).

At any rate, the use of some of the ingredients employed in these glasses appears to have survived beyond the ‘life-span’ of the lead-soda-silica glasses in so-called Iberian group 3, as attested by the San Esteban assemblage, although in this instance without the lead and (in most samples) with less pure plant ashes (although the relatively narrow range of ash-borne impurities, such as magnesia and potash, perhaps indicates a carefully controlled ashing process all the same). As has been shown with other crafts (López-Marigorta 2023b), the end of the Caliphate did not mean the end of ṭirāz crafts, simply their dispersion in the form of smaller workshops that endeavoured to follow the ṭirāz’s production traditions, although generally the quality of their output suffered from the dissolution of the sophisticated supply systems that they had enjoyed previously.

As noted, groups 4 (SEB12, SEB14, SEB15, and SEB21) and 5 (SEB5, SEB13, and SEB22), especially the former, are characterised by significantly higher impurities, and the suggested compositional ranges are considerably wider than in the previous groups. This dispersion is likely indicative of less controlled raw material selection/processing, mixing of different glasses, or both. They also present a wider range of colours (see Table 1), and also of elements that can have a colouring effect on the glass, such as copper, antimony, etc., again suggesting less careful sourcing of raw materials and, perhaps, working practices, as variables such as furnace atmosphere, melting time, etc. can also affect the colour of the glass (Green and Hart 1987, 276–7).

Samples SEB24, SEB29, and SEB31 present fairly low silica-related impurities, but in the absence of trace element data it is impossible to tell whether they belong to groups 1, 2, or 3, while SEB35, SEB28, SEB32, SEB33, and SEB34, on the one hand, and SEB25, on the other, display the high alumina and titania that characterise groups 4 and 5, respectively, although this conclusion must remain tentative as no trace element data is available for these samples.

Finally, the flux-related indicators (see above) strongly suggest the use of plant species that yield a high soda to potash ratio and high normalised lime-plus-magnesia ratio, for instance those elaborated with Salsola soda (Tite et al. 2006, 1290; see also Ashtor and Cevidalli 1983, 496) (see Table 4) which are found in abundance in coastal marshlands in the Mediterranean coastline of Iberia and its hinterland. In fact, the south-eastern coast of the Iberian Peninsula is known to have exported the ashes of barrilla (the vernacular name for a number of related taxa, e.g. S. kali, S. soda, S. sativa) from at least the 15th to the 17th century (Girón-Pascual 2018), and barrilla ashes from the salt pans of San Pedro del Pinatar (coordinates: 37° 49′ 16.3914"; 0° 46′ 5.772"), near Murcia, collected by two of the authors of this paper, were successfully used to experimentally make glass in combination with other local raw materials (Pearson et al. 2021; Govantes-Edwards and Pearson 2023). The data from San Esteban, therefore, uphold the idea that the Andalusi glass industry was from relatively early on, in the 10th century if not earlier, overwhelmingly dominated by recipes that used soda-rich plant ashes as flux. The use of these ashes is also mentioned in several medieval Iberian technical recipes (Govantes-Edwards et al. 2016; 2020a), and, in his travels to the Iberian Peninsula in the late 15th century, Hyeronimus Müntzer wrote about the glass-making properties of the barrillas found in the vicinity of Alhama de Murcia (coordinates: 37° 51′ 4.5"; -1° 25′ 27.1194"), near the city of Murcia (Puyol 1924). However, this does not necessarily mean that other types of glass were not in circulation. We know that in some areas, such as late medieval Malaga, the stock of circulating glass included a good deal of recycled natron Roman glass (Duckworth and Govantes-Edwards 2022). Also, a workshop in Jaén, the activity of which straddled the Christian conquest of the city in 1246, appears to have used mixed-alkali plant ash fluxes with high soda (≈10%-18% NaO2) and potash (≈5%-12% K2O) well above that found in barrilla ash glass (more in Neither of the East Nor of the West, see above). It is plausible to think that these glasses were made with ashes of mastic tree (Pistacia lentiscus). This is a common wild plant species in the south of the Iberian Peninsula that was, like barrilla, traditionally used to make soap in Spain (Gras et al. 2014). None of these mixed-alkali glasses have been found in either Puxmarina or San Esteban, but several samples (reported as “outliers” by De Juan and Schibille 2017 and 2021) in Vascos and Córdoba are compositionally very similar, perhaps suggesting that they were characteristic of inland regions where barrillas are harder to come by than near the coast.

Table 4 The contents of soda, potash, lime, and magnesia found in glass samples can reflect the composition of the plant ashes used as a flux in glassmaking, which present different soda to potash ratios and normalised lime plus magnesia ratios (CaO + MgO/Na2O + K2O). The compositions of the plant species listed in the second block of the table (*) can be found in Tite et al. 2006, and those in the third block (**) in Ashtor and Cevidalli 1983)
Full size table

Importantly, a not insignificant number of samples from San Esteban can be confidently regarded as glass production waste. Although no glass-production facilities have been identified to date in San Esteban, these remains are suggestive of glass-working activity; the fact that the arrabal was not fully built at the time – see Introduction – makes this particularly plausible, as a significant proportion of glass production facilities found to date in al-Andalus were built in relatively sparsely populated areas (Duckworth and Govantes-Edwards 2015; 2020). This is compounded by the nearby presence of pottery production remains, including crucibles used to make glazes, and the archaeological and historical data that point to Murcia as an important glass production hub during the 12th and 13th centuries (see above). A chunk of raw glass (SEB27) has tentatively been identified as a Mesopotamian import. This, however, does not necessarily imply (apart from the extra caution required with samples for which no trace element data is available) that Mesopotamian glass was being imported raw to Murcia. Although there is evidence for trade in raw glass in the medieval Mediterranean (Bass et al. 2009), the ability to melt glass from batch (including high-quality glasses) seems too widespread in al-Andalus for these sort of imports to have been practical in any way. More likely, this suggests the remelting of Mesopotamian glass cullet for reuse (possibly broken imported finished products).

Averages of the various Iberian glass groups, as represented in San Esteban, can be found in Online Resource 1 – Tab 1.

Discussion. Recycling

In relation to recycling, a word of caution is perhaps in order. Work undertaken in recent years has led to a much better understanding of chemical markers that flag potential recycling practices, such as the presence of trace but significant levels of opacifiers and chromophores in ‘naturally coloured’ non-opaque glasses, thought to be the result of the addition (accidental or otherwise) of coloured or opaque glasses to a glass batch (Freestone 2015, 35). It is important to note, however, that these markers have largely been identified in the context of the study of Roman, late Roman, and late antique natron glass production, in which the variables of glass chemistry, from the silica source to the choice of flux, are much narrower than those in glasses postdating the 10th century, by which time the use of plant ash-fluxes was widespread.

This is compounded by two additional factors. First, rather than defining a technical process that follows a relatively straightforward chaîne opératoire, the term “recycling” covers a broad set of practices mobilised in very different settings, from the opportunistic scavenging of materials in conditions of raw material scarcity (e.g. in glass, this is clearly reflected in the dismantling of church windows for remelting described by Gregory of Tours’s Liber in gloria martyrum, see Van Dam 2004, 58); to fully organised economic operations to ensure the reclamation of readily accessible ‘rubbish’ as an optimising economic mechanism (e.g. Martial’s Epigrams seem to describe such a system around glass in early imperial Rome, see Harrison 1987, 203–7); and highly selective operations aimed at producing prestige items (as has been suggested for glass in Visigothic Spain by Govantes-Edwards et al. 2020b). Therefore, a wide range of possible practices and ways of material “re-categorisation” (Duckworth et al. 2020, 450), which requires complex explanations beyond the ‘stock’ answer of recycling as a response to supply shortages. Second, there is the issue of the visibility of recycling, recently summarised by Duckworth (2020). Like any other technological practice, recycling can be undertaken with varying levels of skill and resources; for instance, the result of combining well-sorted glasses (like-with-like, based on external appearance), which probably present very similar compositions, will deviate chemically but little, if at all, from the original glasses, obscuring the practice to the analyst. Therefore, all conclusions regarding recycling in a given glass assemblage must always be suspect of understating the scale of actual recycling practices, as only less careful recycling is likely to have a visible compositional effect. Also, since compositional groups, even assuming that this is a valid paradigm for plant ash-based glasses, in medieval Iberia are still being defined, ‘intermediate composition’ criteria (a glass which, as a result of the mixing of two groups, presents a chemistry somewhere between these) (Paynter and Jackson 2016) lose a good deal of their value.

In San Esteban, some of the classic markers of recycling are virtually absent from the assemblage (e.g. cobalt and antimony), while lead was purposefully added to some glasses. However, many of the non-opaque, naturally coloured Iberian glasses in the assemblage display small but significant concentrations not only of lead and copper (see above), but also of tin, and, significantly, manganese. Most of these elements are present in samples from all compositional groups defined to date in widely varying quantities, in relative terms. It is assumed that these elements were not deliberately added to the glass, as seems to be suggested by the small quantities involved (see Fig. 9). In the figure, this is accounted for by excluding all samples in which the element considered is present above an arbitrary 1% cut-off mark. Based on an understanding grounded on the study of Roman and late Roman glass, this should be considered a reflection of recycling practices. In any case, although the number of Andalusi samples analysed to date is still too small to take conclusions too far, systematic recycling practices with a recognisable signature such as those found in some late antique glasses are yet to be identified in Andalusi glass: e.g. in 6th-7th-century natron glasses, the majority of Egyptian Foy 2.1 glasses present significant quantities of antimony, thought to be result of the addition of earlier Sb-decolourised glasses to the batches (Ceglia et al. 2017), while their coeval Levantine glasses are in the vast majority of cases free from recycling markers.

Fig. 9
figure 9

The presence of chromophores, decolourants, and opacifiers in naturally-coloured, non-opaque glasses is typically regarded as a sign of recycling. A significant number of samples from the five Andalusi compositional groups discussed in this paper present small but significant, and widely-varying, concentrations of some of these elements, including lead (in wt.%) (a), copper (in ppm) (b), manganese (in wt.%) (c), and tin (in ppm) (d). All samples with content above 1% of any of these oxides have been removed to account for the possibility of their deliberate addition. V1-5 refer to de Juan et al.’s (2017) Iberian groups 1–5

Full size image

In any case, although the need to wait for more data to refine our understanding of this complex practice cannot be overemphasised, the available evidence indicates that Iberian group 2, which presents the smaller contents and the narrower ranges in recycling markers, was subject to recycling either less often or more carefully (e.g. sorting like-with-like) than other groups. At any rate, the presence of these chromophores and opacifiers in many ‘naturally-coloured’ glasses seems to suggest that glass recycling was widespread in Andalusi Murcia, in line with recent arguments that suggest that recycling was an everyday practice in glassmaking in Antiquity and the Middle Ages (Duckworth 2020).

This said, the discussion of glassmaking in al-Andalus must take into consideration a factor that plays a much more minor role in the Roman and late Roman periods, such as the possible interaction, even cross-breeding, between glass- and glaze-making. This has been the subject of some specific research in al-Andalus (Salinas et al. 2021), but the extent and scope of this interaction was much greater than is generally understood (Govantes-Edwards 2025). It is interesting to note that 10th-to-12th century Andalusi ceramics, for instance the so-called “green and manganese” wares and other lead and tin glazes (Salinas and Pradell 2020), made abundant use of some of the elements taken as markers of recycling. It is not unthinkable that glass and glazes were melted in the same furnaces and, even if different crucibles were used (which is by no means certain), some cross-contamination is to be expected. Portable XRF analyses undertaken in situ in late medieval and early modern furnaces known to have been used to make glazed pottery in the Alhambra (Granada) (coordinates: 37° 10′ 34.392"; -3° 35′ 17.4474"), showed significant traces of lead, copper, tin, and antimony on the furnace walls (Welham et al. 2022, 61–7; García Porras et al. 2022, 127), and one area tentatively identified as the potential location of the workshop of the pyrotechnological ṭirāz in Madīnat al-Zahrā has been characterised based on the presence of high soil pollution of lead, copper, and manganese (Govantes-Edwards and Duckworth 2020). As such, the possibility that some of the presence of opacifiers, colourants, and decolourants in Andalusi glasses is the result of these cross-craft practices should also be considered and taken into account in future research.

Finally, SEB20 is interesting insofar as, while presenting very high lead (9.46% PbO) and low silica (42.01% SiO2), like the other high-lead glasses, it has extremely low soda (3.93% Na2O) and high alumina (10.87% Al2O3) and iron (4.15% Fe2O3) as well as exceptionally high lime (21.10% CaO) and high concentrations of some elements that are generally regarded as indicative of recycling, such as copper (229 ppm Cu) and antimony (642 ppm Sb). Similar levels of alumina, iron, and silica are typical of contemporary Andalusi bangles that are held to be the tail-end products of repeated iterations of glass recycling (Govantes-Edwards and Duckworth 2024), once the glass was no longer deemed suitable for any shaping technique other than cold-working (see a similar phenomenon in North Africa in Duckworth et al. 2016). The high alumina and particularly lime of the sample strongly suggests that this chunk is the end result of a similar process of repeated recycling.

Conclusion

The archaeometric analysis of 36 samples of glass found during the 2018–2019 excavation campaigns in the urban site of San Esteban has yielded highly interesting data concerning the production, circulation, and consumption of glass in the arrabal of Arrixaca. The items sampled were mostly found in archaeological contexts dated to the 12th and early 13th century, and largely comprised tableware, containers, and lamps, as well as some glass production waste.

Chemical analysis shows that nearly all sampled items were lime-soda-silica glass, using a plant ash-based flux. The comparison with other Andalusi glasses and the Islamicate world more broadly reveals that most samples (29) can be more or less comfortably linked to published Iberian compositional groups, while a small group (three) are likely imports from other Islamicate areas.

Concerning Iberian glasses, it has been argued that one group present in San Esteban (group 3), could represent the late iteration of earlier production traditions related to the Umayyad ṭirāz, based on the narrow compositional range in both silica- and flux-related elements, which suggests a careful sourcing and processing of raw materials, and their similarity with glasses found in Caliphate of Córdoba contexts. Other groups (4 and 5), as currently outlined, present much wider compositional ranges, and it should not be surprising if, as our maps and chronological lines become populated by more data points, their features become clearer and they are subdivided into further groups (again, if the ‘compositional group’ paradigm that we have been using is justified for Andalusi glass in a broad sense).

The Iberian samples characterised include tableware, containers, and lamps, and no relation has been identified so far between compositional group and typology. Most of the objects were made in naturally-coloured yellowish glasses, although bluish green and colourless glasses were also attested, as well as several fragments of tin-and-lead-opacified glass (white and pale blue). Opaque glasses, while not overabundant, have now been identified in a handful of Andalusi sites, including the site of Siyāsa, Murcia (coordinates: 38° 13′ 38.6754"; -1° 25′ 21.0354") (Navarro and Jiménez 2007) and the workshop of Puxmarina (Carmona et al. 2009), so the possibility that these were common Murcian productions appears plausible.

The data also suggests that a significant proportion of these glasses may have been subject to recycling practices. However, the possibility that some chemical markers which have been traditionally used to identify recycling may in Andalusi glass be indicating cross-craft interaction with the glazed ceramics industry, which was an important economic sector in al-Andalus in general and 12th-13th-century Murcia in particular (Navarro 1984), must be considered.