General characteristics
The results of the electron microprobe (SEM-WDS) compositional analyses of 59 samples of transparent glass from the Jebel Khalid Acropolis are reported as oxides in Table 1. Ancient glass was typically made of silica from sand or quartz pebbles, combined with a soda flux derived from either plant ash or mineral soda (natron), and lime as a stabiliser [3, 22, 51, 54, 61]. Lime was probably introduced to the mineral soda glass batch in combination with the silica sand source (e.g. [22: p. 277, 61]).
After the second millennium BCE, west of the Euphrates glass was predominantly fluxed with mineral soda, or natron, until the 9th century in the Islamic period, when plant ash was used again [22: pp. 271–276, 51: p. 1825, 55]. Ash had continued to be used east of the Euphrates River in Mesopotamia and Iran throughout this time [24: p. 204, 58: pp. 1284–1285].
To this basic silica-soda-lime mix could be added modifying agents that produced colour, removed colour to form colourless glass, or opacified glass. Comparative analysis and interpretation of compositional glass data is initially based on the essential raw materials, the so-called reduced glass composition that excludes intentionally added modifiers such as colourants and opacifiers, and includes the basic glassmaking oxides and their associated ‘contaminant’ oxides. Besides silica, soda and lime which are the major components, magnesia, potash and phosphorus are typically minor impurities associated with a plant ash soda flux, while iron, titania and alumina may enter the glass as natural sedimentary contamination of the silica sand source [8, 11, 16, 25, 36, 37, 61: p. 162T]. Alumina may also be derived from the refractory containers used in glass production [47, 61: p. 175T].
From data obtained for the 13 analysed oxides (Table 1), it was found that the silica content of the glass ranges from approximately 64–74 % (average 68.22 %), while soda is present at levels of approximately 15–19 % (average 17.35 %). Lime levels range widely between approximately 5.6 and 10 %, with the majority falling between 8 and 10 %. The average lime content overall is 8.31 %. All but two Jebel Khalid samples therefore have a typical soda-silica-lime composition, which can be characterised as low-magnesia, low-potash (LMLK), with values below 1.4 % for both oxides. Combined with low phosphorus levels of less than 0.19 % (average 0.10 %), this indicates the use of a natron or mineral soda flux [54].
The presence of elevated levels of potash, magnesia and phosphorus would indicate, on the other hand, that glasses were fluxed with plant ash rather than with mineral soda [3, 8; p. 277, 58]. Glasses fluxed with mineral soda typically contain potash and magnesia at concentrations less than 1.5 % [17]. It is of interest that two samples, JK 20, a greenish-coloured bowl, and JK 52, a colourless bowl, possibly late in the series, exhibit lower lime concentrations at around 4.3 %, while their phosphorus contents are elevated to 0.45 and 0.24 % respectively. Combined with relatively higher levels of magnesia (1.35 and 1.10 %) and potash (1.37 and 1.10 %), this could suggest a different source of flux for these two samples, possibly including the addition of plant ash (e.g. [1; p. 241]).
The alumina values for the Jebel Khalid glasses are typical of sand-sourced silica glass, falling between 1.76 and 2.58 % (average 2.23 %), excluding the same two outliers which have lower alumina concentrations, JK20 (1.08 %), and JK52 (1.24 %) (Fig. 2). The iron oxide content of all glasses ranges from 0.23 to 0.71 %, excluding the higher concentration outlier, JK37 (1.15 %). Iron oxide concentrations are generally below 0.60 % (average 0.39 %). Titania is present in all samples between 0.04 and 0.10 %, and appears to be weakly correlated with iron (R2 = 0.4) as a contaminant of the sand (Fig. 3). The levels of iron oxide, alumina and titania in the Jebel Khalid Hellenistic glasses are consistent with the use of impure, sand-sourced silica.
The two main compositional outliers (JK20 and JK52) were identified because of their inconsistent levels of several key oxides, including lime, alumina, titania, potash, magnesia and phosphorus. Other samples of note are JK46, a greenish-coloured sagged bowl with relatively high titania (0.10 %), and JK21, a green, footed bowl base, with slightly high magnesia content (0.90 %). The majority of the glasses analysed form a tight group when oxides are plotted, regardless of colour, vessel form or type.
Colourants and decolourants
Before analysis, samples were assigned to visual colour groups distinguished as clearly as possible by eye to record the variability of colours produced from much less varied chemical compositions (Table 1). The production of colour is complex and is reliant on more than just composition (e.g. [2, 66]).
Copper and cobalt
The presence of cobalt oxide in glass produces a deep blue, and copper oxide may colour glass red, blue or green [2, 66]. Three of the blue and green glasses from Jebel Khalid owe their colours to the presence of these colourant oxides. The cobalt oxide concentration (0.08 %) for the blue sample from the core-formed alabastron, JK26, provides the source of the blue colour for this glass. Copper oxide was also detected in this sample at 0.12 %. It is well-known that cobalt can impart a deep blue colour to glass even at a concentration of a few hundred ppm (see for example [1: p. 244, 61: p. 163T, 66], and as little as 0.02 % [24: p. 69]). Kaczmarczyk and Hedges [33: p. 151] (see also [57: p. 289]) define a cobalt blue glass as one containing at least 0.05 % cobalt oxide.
In the Late Bronze and early Iron Ages cobalt colourant was predominantly sourced from the Western Desert oases of Egypt in the form of cobaltiferous alum. Glasses coloured with cobalt derived from alum are typified by raised concentrations of alumina, magnesia, iron, nickel, zinc and manganese [23, 32, 33: pp. 41–55, 35, 39, 44, 48, 50: pp. 267–268, see 55: p. 958, 56: pp. 145–146, 57]. These cobalt glasses generally contain lower lime than non-cobalt glasses [18; pp. 272–273, 54; p. 157]. After approximately the 7th century BCE the source of cobalt changed, and the concentrations of the impurities associated with cobaltiferous alum returned to average levels [32: pp. 373–374, 33: pp. 45–47, 53, 52: pp. 51, 54]. Various cobalt ores have been identified in the region, one being an arsenic-nickel cobalt ore known to have been used in glassmaking [19, 24: pp. 69–74].
The levels of iron, alumina, manganese and magnesia in sample JK26 are similar to those in the non-cobalt coloured Jebel Khalid glasses suggesting that the source of colourant was not cobaltiferous alum. LA ICP-MS analyses of the Jebel Khalid glasses confirm that the concentrations of nickel, zinc and arsenic were not elevated either, leaving the source of cobalt as yet unknown (H. Rutlidge, UNSW, unpublished data). Tite and Shortland [57: p. 289] observed that second millennium Egyptian glass coloured with cobalt may also have copper oxide present in the range of 0.02–1.3 %. It is interesting that sample JK26 contains 0.12 % copper oxide, within the range defined by Tite and Shortland that qualifies it as a co-colourant. Elevated cobalt and copper oxides occur also in blue glasses from the Hellenistic period from Gordion in Anatolia [46: p. 68], and from Pichvnari and Tsikhisdziri in Georgia [55: p. 958].
An elevated level of copper oxide is responsible for the blue colour of JK04 (1.96 %) and the green colour of JK01 (1.40 %), both mosaic cane bowls. Other green to pale greenish or pale bluish glasses do not contain a significant quantity of this colourant, and derive their colour from iron, introduced as an impurity in the glass raw materials in similar concentrations in all samples [2: pp. 35, 48, 143, 61: p. 180T, 66: pp. 91, 163–164].
Iron
The majority of sands contain iron as an impurity at usually less than 1 % [33: p. 36–41]. In natron glasses the inclusion of iron in the glass batch as an impurity in silica sand can be responsible for colouring raw glass at levels below 0.5 % (e.g. [29: p. 151, 53]). The colouring effects of iron are complex, depending on its oxidation–reduction equilibrium. Colours produced include blue, green, yellow and brown [2: pp. 143, 155, 66: pp. 89ff, 119–120].
The best way for glassmakers to have avoided the colouring effects of iron was to use pure raw materials, because the use of sand in natron glasses commonly incorporated iron. Colourless glass could also be achieved by adding a decolourant [28: p. 764, 66: p. 97].
Iron oxide concentrations of the Jebel Khalid glass range between 0.23 and 1.15 % (average 0.41 %) and would have imparted the translucent greenish hue, or pale aqua blue due to the presence of ferrous ions (Fe2+), while the presence of ferric ions (Fe3+) in combination with sulphur can produce a dark brown or amber colour in a silica-soda-lime melt under strongly reducing conditions. Sulphur, although not measured in these analyses, is introduced intentionally or as contamination in one or more batch materials, including natron, or from the addition of organic materials and smoke from combustion that produces a reducing atmosphere [2: pp. 85ff, 155, 53: pp. 199, 206–207, 66: pp. 119–120, 240].
Of the 22 amber glasses analysed, the iron oxide content of all but one was low, falling between 0.23 and 0.39 %. The majority of the remaining 23 monochrome glasses in various shades of green, blue, purple and colourless had on average slightly higher iron oxide concentrations ranging up to 0.71 %, average 0.44 %. Sample JK37, from a very light green bowl, has an unusually high iron content (1.15 %), but despite a high manganese content of 1.63 %, it was not completely decoloured.
All but one of the glasses have a similar iron content, and this oxide is probably responsible for producing the various shades of green and light blue. In the absence of other identifiable colouring agents, the amber colour is likely due to the formation of the ferri-sulphide complex in a reducing atmosphere [2, 66]. Iron colouration is counteracted in the colourless glasses by manganese.
Antimony and manganese
Antimony was used from the early 1st millennium BCE, and into the Roman period as a decolourant, an opacifier, and as a fining agent to remove small seeds and bubbles in glass production [2: p. 80, 6: p. 116, 50: p. 272, 61: p. 179T, 66: pp. 116, 118, 121].
By the later Hellenistic period of the late 1st millennium BCE, manganese had come into use as a glass decolourant, and this is borne out by the compositional data of the Jebel Khalid glasses (see also [15: p. 95, 24: p. 246]). Manganese reacts with the iron impurities in the glass melt to counteract the iron-induced green tint by oxidation, producing a colourless glass instead [2, 7: p. 277, 28: p. 764, 66: p. 116]. Jackson [28: pp. 765, 771] noted the increasing use of manganese, rather than antimony, to around 1 % towards the end of the Roman period, defining deliberate addition of this oxide at levels exceeding 0.5 %, in agreement with Brill [7]. Under certain conditions manganese will not decolour, but will produce purple glass [2: p. 50, 66: p. 121], and may also act as a fining agent [61: p. 179T, 66: p. 130].
The low antimony oxide values for most of the Jebel Khalid glasses, including the colourless samples, fall below 0.09 %. Greenish to colourless glasses with slightly elevated antimony levels include colourless JK13 (0.11 %) and JK52 (0.13 %), greenish JK20 (0.10 %), and greenish colourless JK35 (0.22 %).
The eight truly colourless (untinted) samples, including JK13 and JK52, appear to be decoloured by elevated concentrations of manganese oxide, between 1.13 % and 1.77 %. Imperfectly decoloured greenish colourless samples JK15, JK46 and JK49 also have elevated levels of manganese between 1.33 and 1.96 %, and some shades of green, amber and light blue have levels of manganese oxide between 0.5 and 1 %. It is evident from these results that manganese oxide was the predominant glass decolourant in Jebel Khalid glasses (Fig. 4). The sagged bowl sample, JK23, has a manganese oxide level of 1.84 %, which has resulted in purple colouration, rather than decolouration. All of the amber coloured samples have uniformly low manganese oxide concentrations (<0.61 %), with the majority containing less than 0.04 % manganese, and little or no antimony.