Geochemistry of Byzantine and Early Islamic glass from Jerash, Jordan: Typology, recycling, and provenance

Twenty‐two objects of glass from the Decapolis city of Gerasa, N. Jordan, with characteristic vessel forms ranging from Hellenistic to Early Islamic (2nd century BCE to 8th century CE) were analyzed for major and trace elements, and 16 samples for Sr‐isotopes. The majority were produced in the vicinity of Apollonia on the Palestine coast in the 6th–7th centuries CE, and strong inter‐element correlations for Fe, Ti, Mn, Mg, Nb reflect local variations in the accessory minerals in the Apollonia glassmaking sand. The ubiquity of recycling is reflected in elevated concentrations and high coefficients of variation of colorant‐related elements as well as a strong positive correlation between K and P. The high level of K contamination is attributed to the use of pomace (olive processing residue) as fuel, and a negative correlation with Cl, due to volatilization as the glass was reheated. This points to an efficient system for the collection of glass for recycling in Jerash during the latter part of the first millennium CE. Differences in elemental behavior at different sites in the Levant may reflect the context of the recycling system, for example, glass from secular contexts may contain less colorants derived from mosaics than glass associated with churches.


INTRODUCTION
It is now generally accepted that from the late first millennium BCE until the late first millennium CE, the ancient glass industry was centralized. Large-scale natron glass production supplying the entire Eastern Mediterranean region was centered in only a few locations along the Palestine coast and in Egypt. Each production center produced unique glasses due to minor differences in recipe and the local raw materials, but common for the natron glass types is that they were made by mixing calcium carbonate-bearing sand with natron (soda) from salt lakes at Wadi el-Natrun or the Nile Delta (e.g., Brill, 1988;Degryse, 2014;Degryse & Schneider, 2008;Freestone, Gorin-Rosen, & Hughes, 2000;Freestone, Leslie, Thirlwall, & Gorin-Rosen, 2003;Nenna, Vichy, & Picon, 1997). These primary glassmaking centers exported the raw glass to population centers across the ancient world where secondary glass workshops remelted and shaped the raw material into vessels, windows, and jewelry. Whereas the general outline F I G U R E 1 Regional map of Syria-Palestine with location of the study site of Jerash as well as surrounding contemporary cities of Petra and Umm el-Jimal. Also shown are glass production sites along the Levantine coast at Apollonia, Jalame, and Bet Eli'ezer [Color figure can be viewed at wileyonlinelibrary.com] highest area within the walled city (Lichtenberger & Raja, 2015, 2017 ( Figure 2). The project explores mainly domestic complexes of this quarter of the city. Most of the excavated structures stem from the Late Roman to Early Islamic periods. During the excavations, evidences were excavated containing the inventory of houses and among the finds were glass vessels, most of them fragmented. Here, we present major and trace elements for 22 as well as Sr isotopic compositions for 16 glass artefacts excavated during the 2013 campaign of this project, selected to represent the range of glass forms encountered.
The objectives of this study are twofold. The first is to determine the main glass types that reached Gerasa and how this reflects the supplies into the city and thus regional trade networks. The second objective is a detailed characterization of the contaminants and postproduction chemical signatures that became incorporated into the glasses during remelting in secondary glass workshops. The signatures provide clues about the local remelting techniques, furnaces, fuel sources, glass types mixed during melting and/or added colorants which, ultimately, reflect the inner workings and infra-structure of Gerasa within a local and regional context. Islamic periods (5th-8th centuries CE) but we have included earlier representative forms in the analytical sample set.

SAMPLE MATERIAL
Early period material is rarely encountered among the glass finds. Only a few sherds can be assigned to the Hellenistic period (336 BC-30 BC). They belong to cast grooved bowls showing a TA B L E 1 Composition of Corning B glass standard by electron microprobe in this study compared to recommended composition

RESULTS
All samples classify as low-magnesium, low-potassium (<1.5 wt% each of MgO and K 2 O) natron glasses (Lilyquist, Brill, & Wypyski, 1993;Sayre & Smith, 1961). Figure Figure 5). This is important for distinguishing between the different glass groups observed in our study.
The three glasses which are typologically Roman are natron-based with CaO and Al 2 O 3 contents below 8 wt% and 2.8 wt%, respectively, and Na 2 O concentrations above 17 wt% (Table 2), consistent with Roman glass produced in the first centuries CE (Jackson, 2005).
Three distinct types of Roman glass are recognized primarily on the basis of their contents of the decolorizers Mn and Sb (Table 4). Values are the means of six separate spot analyses.
F I G U R E 6 Oxide ratio variation diagram of CaO/Al 2 O 3 versus Na 2 O/SiO 2 for Jerash glass groups, which discriminates between Levantine primary productions of glass produced at: (1) Jalame (4th century; data of Brill, 1988, supplemented with 3rd century Rom-Mn glass from Silvestri, 2008, (2) Apollonia (6-7th century, Freestone et al., 2000, Tal et al., 2004, supplemented with vessels from Phelps et al., 2016), and (3) Bet Eli'ezer (7-8th century, Freestone et al., 2000 and unpublished). Original plot lay-out from Phelps et al. (2016). All oxides are in wt% (  Brill, 1984). On the other hand, cat. no. 1 has only background Mn at 146 ppm and is amber which is generally attributed to the presence of a ferri-sulfide complex (Schreurs & Brill, 1984). It may be pertinent that this glass has the highest sulfur content of those analyzed (0.3 wt%) and the absence of added manganese (Tables 2 and 4) is a typical feature of amber glass (Freestone & Stapleton, 2015;Sayre, 1963), as it is an oxidizing agent and the generation of amber requires reducing conditions.
Glass characteristic of Late Roman-Byzantine Palestine was originally defined as Levantine by Freestone et al. (2000) and this term has been heavily used in the literature. It refers to lime and alumina contents in excess of 8 wt% CaO and 2.8 wt% Al 2 O 3 combined with relatively low Na 2 O (<17 wt%) making them distinct from older 1st-3rd century CE Roman glass ( Figure 4; Table 2; Schibille et al., 2017).
Levantine glass has been divided into Levantine I type characterized by 68-71 wt% SiO 2 and 14-16 wt% Na 2 O and Levantine II type with relatively higher 73-76 wt% SiO 2 and lower 11-13 wt% Na 2 O contents (Freestone et al., 2000). More recently, it has been recognized that the original grouping of Levantine I incorporated the products of at least two different primary productions (Al-Bashaireh et al., 2016;Phelps et al., 2016;Schibille et al., 2017) and that these can be more-orless separated on the basis of composition: 6th-7th century CE glass from Apollonia Tal, Jackson-Tal, & Freestone, 2004) and 4th century CE glass from Jalame (Brill, 1988). Figure 6 presents our data in terms of CaO/Al 2 O 3 and Na 2 O/SiO 2 ratios, which pull apart glass from the two production sites (Phelps et al., 2016) and also separate Apollonia glass from natron glass (previously Levantine II) from the Umayyad production site at Bet Eli'ezer, Hadera (Al-Bashaireh et al., 2016). The reference data for Roman period Jalame glass are supplemented with data for Rom-Mn glass from the 3rd century CE Iulia Felix wreck (Silvestri, 2008;Silvestri, Molin, & Salviulo, 2008) and for the Byzantine Apollonia-type with the analysis of vessels from Palestine (Phelps et al., 2016).
Whereas the typologically Roman and Hellenistic glasses analyzed plot on the right-hand side of Figure 6, similar to 4th century Jalame, the 14 Late Roman-Byzantine (4th century CE or later, Table 2) samples which have natural levels of Mn or low Mn levels up to 500 ppm lie to the left, plot in the Apollonia field or have compositions which straddle the boundary between Jalame and Apollonia glasses. In addition to their dating, the low manganese contents of these glasses are consistent with their assignation to Apollonia rather than Jalame since deliberately added Mn is present in about half of the glass analyzed from Jalame (Brill, 1988) whereas this has not been observed in primary glass from Apollonia tank furnaces.
Catalogue numbers 11a, 11b, 16 which have slightly elevated (≈300 ppm) Mn are separated in the analytical tables and in the figures, as they are shifted slightly towards the Jalame field (see above; Figure 6). However, the low Na 2 O contents of these glasses associate them with Apollonia-type production and we consider them as Apollonia-type glass, with admixing of a small amount of older Mndecolorized material. These are henceforth referred to as "low-Mn" as opposed to glasses with "background" levels of Mn. Cat. nos. 6, 7, and 14 have significantly elevated (> 2000 ppm) manganese and could be interpreted as Jalame-type glasses ( Figure 6; None of the glasses analyzed corresponds to the field of the majority of low-soda high-silica products of the 7th-8th century CE furnaces at Bet Eli'ezer ( Figure 6).

Origins of primary glass types in Jerash
The strontium isotope data indicate that most of our samples have 87 Sr/ 86 Sr ratios close to the Holocene seawater (and beach shell) value    (14) Total procedural blank = 0.06 ng Sr a Two sigma analytical precision corresponding to the trailing digits. b Repeat analysis.
groups and for primary glasses from Apollonia (data from Phelps et al., 2016). These fingerprint the Levantine coast as the source for the Apollonia glasses from Jerash, and show the strong similarities between the Rom-Mn, Hellenistic glasses, and primary Apollonia glass. Rom-Mn glass is generally considered to have originated on the coast of the Levant (Nenna et al., 1997), and these data are consistent with that view.
The Hellenistic glasses are similar in major and trace compositions, and are likely to have originated in the same region.
In the 1st century CE, the production of antimony-decolorized glass was established and it appears to have been preferred for more expensive items such as tableware with cut decoration . The Roman Sb and Roman Mn-Sb glasses from Jerash differ significantly from the Byzantine, Rom-Mn, and Hellenistic glasses in terms of Rb, Ba, and LREE (La, Ce) in particular (Figure 8). This supports the view that Rom-Sb glass was not a product of Palestine and more likely originated from Egypt (Degryse, 2014;Schibille et al., 2017).
The results therefore suggest that glass used at Jerash from the   Kamber et al., 2005). Our groups are compared to primary glass from Apollonia (Phelps et al., 2016). Jerash Byzantine glasses include low and high Mn groups. Note the logarithmic scale [Color figure can be viewed at wileyonlinelibrary.com] could be a sampling effect and small quantities of these types might have reached Jerash; this possibility will be further explored in later studies.
Of particular interest is that no glass which might have originated in the tank furnaces at Bet Eli'ezer (Levantine II of Freestone et al., 2000) has been detected. The Bet Eli'ezer furnaces appear to have been in production from about 670 CE (Phelps et al., 2016). The absence of this glass type in Jerash suggests that none of the glasses analyzed date later than the third quarter of the 7th century CE. This is possible, as typologically all of the forms could date to late Byzantine times or earlier. Furthermore, Bet Eli'ezer-type glass has been identified at another site in Jordan, Umm el-Jimal, located away from the coast but some It is pertinent that the glass of the 2nd-4th century CE typically has a dark weathering patina, whereas the 6th-7th century CE fragments weather to an opaque white. The precipitation of manganese oxide in the weathered layers is well-known as the cause of the darkening of medieval European glass (Schalm et al., 2011) and it seems likely that this is also the case for Jerash, as the Roman-period pieces typically have high levels of MnO, whereas the later glasses typically have MnO at background levels. The Roman-period antimony-decolorized glass, cat. no. 5, has low manganese, and also weathers to an opaque white patina, consistent with these observations.

Secondary processing phenomena: recycling in the Apollonia-type glasses
The discussion below will focus on the Apollonia-type glasses since these are by far the most dominant group in our sample-set and because they show distinct features that relate back to production and postproduction processes in secondary workshops. The location of the glass workshops which made the vessels in Jerash have yet to be determined by excavation, but it seems very likely that, like other cities in the region in Late Antiquity they were in the immediate vicinity.
A number of compositional effects might be anticipated from the mixing and remelting processes which comprise a glass recycling system: (1) mixing of different primary glass compositions, (2) contamination from the melting furnace/crucibles and iron glass working tools, (3) contamination with colorants and decolorizers from the incidental inclusion of old colored glass in the batch, (4) contamination by components of fuel and fuel ash, and (5) loss of volatile components to the furnace atmosphere. By definition, if a glass object is remelted to make a new one, then it is recycled, and evidence for remelting is therefore evidence for recycling.
In terms of mixing different glass types, we have observed above that the Byzantine glasses with high Mn lie closer to earlier Roman glasses in terms of major components such as Na 2 O ( Figure 6) and that this is the result of mixing Apollonia-type glass with Roman Mndecolorized glass. No other compelling evidence of mixing of primary glasses is recognized here, but this process is implicit in some of the data, for example, the behavior of colorant-related elements, below.
There is a strong correlation between Fe and Mg in the Apolloniatype glasses from Jerash which is also present for other transition metal elements such as Ti, V, and Nb (Figure 9). An enrichment in Fe has been observed in Roman Mn-Sb glass from York, UK and ascribed to contamination from ceramic melting pots or iron blowpipes . If the Fe was derived from an iron tool, departure from the trend with MgO would be expected and this does not occur (Figure 9a). Contamination from the furnace is a possibility, but the high Mg:Fe ratio of 1:1 indicates control from heavy minerals such as amphibole, pyroxene, spinel, and zircon (e.g., Molina, Scarrow, Montero, & Bea, 2009) rather than clays which are generally strongly dominated by Fe relative to Mg (e.g., Kamber, Greig, & Collerson, 2005). Therefore, the covariations in the Apollonia samples are a primary feature controlled by differences in the accessory mineral assemblage of the individual batches of coastal sand used for their production. A similar explanation has also been offered by Schibille et al. (2017) for FeO-MgO covariations in Rom-Sb glasses that they analyzed from Carthage. However, the Rom-Sb glasses reported by Schibille et al. (2017) have an MgO:FeO ratio of 2:1, relative to Mg and Fe covariation of 1:1 in the Jerash glasses ( Figure 9). These very different Mg/Fe ratios support the hypothesis (see above) that the Sb-decolorized glass did not originate in the Levant, but elsewhere, possibly Egypt.
In addition to iron and other transition metal oxides, an increase in alumina concentration might be expected if a glass was significantly contaminated by furnace ceramic during remelting. Figure 6 shows no enrichment in Al 2 O 3 of the Jerash samples relative to Apollonia primary glass and we can therefore assume from this and the iron oxide that contamination of the glass from ceramics (furnace) during any recycling that occurred was minimal. This differs from the conclusions of  and may reflect the arrangement of the furnace. In the Levantine region, there is limited evidence for the use of pots or crucibles in which glass was melted and it appears that even at the secondary stage, the glass was melted in tanks (e.g., Gorin-Rosen, 2000); this was also the case in larger centers in the West, for example Roman London (Wardle, 2015). There is evidence, however, for melting pots at York, UK studied by Jackson and Paynter (op. cit.). Tanks will typically have had a much larger volume to surface area ratio than pots or crucibles and the interaction between the walls of the tank bulk of the glass will have been correspondingly less, explaining the discrepancy.
A distinctive group of elements, including Cu, Sn, Pb, Co and Sb, shows different behavior. Figure 10 shows that where crustal values are available (Kamber et al., 2005), these elements show a substantially higher level of enrichment in our glasses than those elements associated with accessory minerals. They are the elements associated with glass coloration and, following earlier studies (Freestone, Ponting, & Hughes, 2002;Jackson, 1996;Mirti, Lepora, & Saguì, 2000), it is considered that a significant component originates in the incidental incorporation of small amounts of earlier colored glasses in recycling processes. This effect of recycling on the distributions of these elements is conveniently illustrated in terms of the coefficients of variation (relative standard deviations) for the individual elements ( Figure 11). The colorant elements have very high CVs due to the imperfect nature of the recycling process and the failure to completely mix and homogenize separate glass batches. Furthermore, several element pairs show very strong correlations, such as Cu-Sn, and Pb-Sb F I G U R E 1 0 Trace element concentrations (ppm) related to colorants addition to glasses normalized to weathered continental crust (MUQ of Kamber et al., 2005). Our groups are compared to primary glass composition from Apollonia (Phelps et al., 2016). Jerash Byzantine glasses include low and high Mn groups. Note the logarithmic scale [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 1 1 Coefficients of variation (= relative standard deviations) for trace elements in for all Apollonia-type glasses with background levels of Mn (Table 2). Sand-related elements typically have low CVs whereas those associated with colorants are high. Elements associated with alkali and ash-U, Rb, and B are intermediate (R 2 of 0.75 and 0.81, respectively), and these appear to reflect specific coloring agents such as bronze scale and lead antimonate. The implication is that, whereas the Apollonia-type glasses from Jerash show features fully consistent with a single primary production, there has been significant recycling and this is reflected in the colorants. Furthermore, analysis of glass from tank furnaces on the Levantine coast indicates Pb values typically less than 10 ppm, and Cu values less than 5 ppm (Brems et al., in press;Phelps et al., 2016), whereas with only one exception, our Apollonia-type glasses contain higher levels (Table 4) suggesting that the great majority of the Apollonia-type glass analyzed here contains some recycled material.

Influence from fuel and furnaces during recycling in Jerash
Evidence that the Apollonia-type glasses had been through one or more episodes of recycling has been inferred above from the colorant element concentrations. The effects of workshop practices on glass composition have been explored experimentally by Paynter (2008) F I G U R E 1 2 Oxide variation diagram of P 2 O 5 versus K 2 O in wt% (representing contamination from fuel ash) for Jerash Byzantine glass groups with background and low Mn compared to Byzantine glass compositions observed at other Levant cities. Data for Petra, Deir Ain Abata (Rehren et al., 2010), Ramla, Israel , and Umm el-Jimal, Jordan (Al-Bashaireh et al., 2016). Primary glass from Apollonia from Phelps et al. (2016). R 2 value is for fitted regression line through Jerash glass group [Color figure can be viewed at wileyonlinelibrary.com] who showed that in addition to accumulation of Al and Fe from the melting pot, remelting of Roman-type soda lime glasses in reconstructed Roman glass furnaces subjects the glass to contamination by K from fuel ashes and/or vapors. While we can expect the contaminants to have been strongly controlled by the type of glass, the fuel, the firing temperatures and the type of clays used to make the furnace, Paynter's study provide some important clues about potential influences from furnace and fuel.
Concentrations of K 2 O and P 2 O 5 in Levantine I glasses from Jerash are high, up to 1.33 and 0.21%, respectively. Not only are these values twice as high as in glass from the primary furnaces at Apollonia (Freestone et al., 2000;Tal et al., 2004), but these two components are strongly correlated (R 2 of 0.88 in Figure 12). The K 2 O and P 2 O 5 correlation observed for the Jerash glasses is most likely the result of interaction with the fuel ash and fuel ash vapors during remelting and/or working as has been observed for Apollonia-type glass at other contemporary sites such as Petra, Jordan (Rehren, Marii, Schibille, Stanford, & Swan, 2010), Ramla, Israel  and  This may be due to the configuration of the Jerash furnace(s), so that the glass was protected from contamination by solid ash, and the contamination was largely from the vapor, but it could also be due to the type of fuel used.
A plausible fuel for Jerash is olive pits, given the finds of olive crushing mills and olive pits in many layers in Jerash. There is little doubt that olives have played a role regionally and oil production in general was significant (e.g., Ali, 2014). Rowan (2015) has drawn attention to the extensive evidence for the use of pomace, olive-pressing waste, as a fuel in antiquity. Olive pits as fuel for glass production are particularly suitable, since their fire burns hotter than wood and therefore they have excellent qualities for glass melting. Large amounts of charred olive pits were found close to glass furnaces in Beth Shean (Gorin-Rosen, 2000) and Sepphoris (Fischer & McGray, 1999), but until now evidence for the actual use of these for firing has not been drawn from the chemistry of the glass samples. Data on the chemistry of olive residues is available due to modern interest in their potential as a biofuel. CaO for the ash of "olive residue" (Gogebakan & Selçuk, 2009). It is clear that the potash to lime ratio of olive pit/residue ash is significantly higher than those of most hard and soft wood ashes, in which lime is generally in excess of potash (e.g., Misra, Ragland, & Baker, 1993).
Therefore, furnaces operating with a high proportion of olive pits in the fuel would produce ash with substantially more K 2 O than those firing mainly wood. The high level of enrichment of potash observed in this study and in other glasses from Jordan strongly suggests that olive pits were a significant component of the fuel used, consistent with the archaeological evidence from the region. Miranda et al. (2008) report that their olive pit ash also contained 3.43% P 2 O 5 , which would volatilize and explain the correlation observed between phosphate and potash.
We observe a negative correlation between potash and chlorine, which has previously been observed in glasses from Umm el-Jimal originates from the natron and would normally be expected to show a positive correlation with soda (Na), also coming from the natron, and be stabilized in the melt due to sodium-chloride (Dalou, Le Losq, Mysen, & Cody, 2015). However, given the volatile nature of chlorine, as well as the alkalis, repeated melting, particularly at high temperature, inevitably leads to Cl (and to a lesser degree alkali) loss (Freestone & Stapleton, 2015). This does not explain the antithetical relationship seen for Cl and K in Jerash Byzantine glass ( Figure 14). As for Umm el-Jimal and Petra, we ascribe this correlation to a combination of recycling (leading to chlorine loss) and contamination by fuel ash (leading to increased potassium). Moreover, the strong negative K-Cl correlation (R 2 = 0.63) compared to other sites in the region (Umm el-Jimal at 0.25 and Petra at 0.24), in addition to even stronger positive K-P correlation (R 2 = 0.88; Figure 12), suggests that glass recycling was more intensive at Jerash. Jackson, Paynter, Nenna, and Degryse (2016) (Carroll, 2005;Metrich & Rutherford, 1992;Veksler et al., 2012). This expectation is realized for medieval and postmedieval glasses where Na and Cl are positively correlated (Schalm, Janssens, Wouters, & Caluwé, 2007;Wedepohl, 2003) and supported by evidence of immiscible droplets of sodium chloride in ancient Cl-rich glass soda lime-silica glasses (Barber & Freestone, 1990; Barber, Freestone, & Moulding, 2009). Finally, the importance of alkali-Cl complexes in the melt and inevitability of Cl loss during fusion are corroborated by the relatively high chlorine contents of Roman amber glass which have been suggested to be the result of relatively short melting durations used to preserve the color (Freestone & Stapleton, 2015). In conclusion, our observation that chlorine abundance is antithetical to potash for Jerash Byzantine glass and the lack of demonstrable correlations with soda and lime is consistent with recycling, and thus not a feature of primary glass production. We are not proposing that Cl abundance is a universal tracer of recycling, melting duration or melting temperature, but rather, when one considers glasses of similar major element composition from similar technological context, chlorine content coupled with correlations (or not) with other glass constituents is a useful indicator of recycling.

Compositional dependence upon the context of the glass recycling economy
The Jerash data emphasize the complexity of the glass recycling process and the dependence of the composition of the recycled glass upon the local social context. It has been observed that the characteristics of recycling differ from those in some western contexts, such as York, as contamination from container ceramics is not apparent. Furthermore, the elevated values and strong correlations for potash and phosphorus observed here are not as apparent in western Roman glasses which are believed to have been recycled (e.g., Freestone, 2015;Silvestri, 2008) or even in Apollonia-type glass from Israel (Phelps et al., 2016) and this may be related to the fuel used.
It is also noted that in Umm el-Jimal, in northern Jordan, Apolloniatype glasses show a greater overall enrichment in trace metals generally added as colorants, where Cu and Pb enrichments are detectable using EPMA rather than the trace levels observed here (Al-Bashaireh et al., 2016). This is likely to stem from the nature of the reservoir of glass undergoing recycling. The Umm el-Jimal glass was recovered from churches where storage of colored glasses from mosaics for recycling might be expected, as has been observed at Petra (Marii & Rehren, 2009). We speculate that this led to relatively high contents of glass colorants in the glass from Umm el-Jimal. The glass from Jerash analyzed here originates from domestic houses, shows relatively weak enrichment in colorant elements but strong evidence of recycling in the fuel-related components.
There is substantial archaeological evidence from Jerash which attests to the collection of glass possibly for recycling. Such glass heaps stem from the churches and especially from the passage north of St.
Theodore and from a room under the north stairs from the Fountain Court (Baur, 1938 in: Kraeling 514-515). Also, evidence for already recycled material in the form of glass cakes probably prepared for the production of glass tesserae has been found in Jerash in the so-called Glass Court (Baur, 1938 in: Kraeling 517-518). Future work, comparing glass associated with the churches with that from more secular, domestic contexts, might cast light on the organization of the glass industry in the Levant at this time.

CONCLUSIONS
The excavated glass from Jerash, Jordan, dating to between the Hellenistic and the Late Byzantine periods, derives mainly from the Levantine coast with some, possibly Egyptian, antimony-decolorized glass in the Roman period. The Byzantine glass, which dominates the assemblage, derives mainly from the tank furnaces located in or around Apollonia. A consideration of the manganese contents of the Apollonia-type glass indicates that it is generally present at background levels, and where present is the result of remelting and mixing of Roman glass during recycling.
Significant evidence for recycling is observed in the form of elevated potash and phosphate contamination from the fuel, as well as elevated transition metals. Concomitantly, there was a depletion in chlorine, due to volatilization at high temperature. For the first time, we draw attention to the effect of recycling on the coefficients of variation of trace elements in the glass. These types of indicator can provide clues as to the relative intensity of the recycling process, which elemental concentrations alone do not.
Despite the apparent proximity of Jerash to primary glass production sites near the Levantine coast, an efficient system for recycling of old glass must have been in place. The implication of a well-organized recycling system in Jerash suggests limited glass import from the Levantine coast and elsewhere, which is supported by the finds of only few Roman glasses and a lack of Egyptian-type glasses. The localized nature of recycling in Jerash displays important regional differences, which we relate to differences in interaction zones and proximity to the production sites at the Mediterranean coast.
The characteristic K-enrichment observed at Jerash and other Levantine locations has implications for the type of fuel used and is likely to indicate a significant component of olive-pressing residue. Differences between Jerash and other sites in the region such as Umm el-Jimal suggest that the nature and degree of over-printing of primary compositions by secondary recycling processes are specific to the context within which the recycling took place. Technological factors relating to local practice, as well as the types and quantities of glass available for recycling, may provide a fingerprint of the secondary workshop. In favorable circumstances, this may allow the attribution of glass vessels to secondary workshops through elemental analysis.
The phenomenon of recycling fits well to the overall economic situation of the cities in the region of the 7th-8th centuries CE. The towns underwent a considerable process of on the one hand urban industrialization and on the other localization of trade networks (see e.g., Avni, 2014, 290-294;Walmsley, 2000, 305-309;321-329;335-337).
Both processes are apparent in recycling which attests to local production within the city and limited supply of (or high demand for) raw materials.