Natron glass has been an important part of Mediterranean and European material culture for nearly two millennia, yet natron glass found in other parts of Eurasia has not been fully studied despite its influence on ancient Asian glass. In this paper, we offer a new interpretation of natron glass found in the East and West. After using trace elements to determine the compositional type and technological sequence of Mediterranean natron glass (8th–12th centuries BCE), we present an analysis of a mid-first millennium BCE glass bead found in Xinjiang, China, which was likely made from Levantine rough glass, and, based on compositional and typological comparisons, identify a common type of plate eye bead across Eurasia. Taking these results together, we suggest that large quantities of Mediterranean soda glassware had reached East Asia by at least the fifth century BCE, which may have been a contributing factor to the development of local glass production in China. The rapid spread of soda glass across Eurasia in the first millennium BCE was likely achieved through a three-stage process involving maritime and overland networks and multiple forms of trade and exchange, suggesting the existence of a highly adaptive and increasingly efficient transcontinental connection along the ‘proto-Silk Road’.
Past interregional interactions in Eurasia have shaped our world in many ways. It is critical to understand how these interactions evolved over time, spread over larger areas, and increasingly influenced culture. Recent studies have revealed prehistoric connections across Eurasia based on evidence of human migrations1,2,3, the spread of domesticated crops4,5, and the spread of pottery6 and metallurgy7. However, before and throughout most of the Bronze Age, the dissemination of people, materials, and innovations generally took a long time, suggesting that the intensity of communication over long distances was limited. For example, wheat did not reach the lower Yellow River basin in China until 6,000 years after its domestication in West Asia8,9. This changed in the second century BCE. BC, when silks, mirrors, and lacquerware from China’s Warring States period were found in tombs from the same period in Siberia and Inner Asia,10 and distant regions were mentioned in Roman and Chinese texts. At the beginning of the surge in intercontinental trade, in the late second century BC, the Han Dynasty envoy Zhang Qian traveled to the heart of Asia. This event is generally considered to mark the beginning of the historic Silk Road. Thus, the period preceding this, the first millennium BC, was crucial in dramatically accelerating and integrating long-distance interactions across Eurasia.
Artificial glass is one of those truly transnational products of civilization that can be used to study interactions over long distances. The history of the origin and spread of glass is varied. Soda glass was first produced in the Middle East and then spread to Europe, Central Asia, and South Asia. Potash glass is found in South, Southeast, and East Asia, as well as Europe. Lead-barium glass first appeared in China and has also been found in other parts of East Asia.11,12 The origin of glassmaking technology in East Asia in the first millennium BCE is uncertain. Archaeological studies of early Chinese glass have revealed the existence of these three main groups,13,14,15 suggesting that this was a period of active cultural interaction and technological exploration. Confirming the authenticity of these early artifacts is critical to determining the origins of Chinese and Asian glass.
Soda glass is a sodium glass made using the mineral natron as a flux, which is characterized by low magnesium and potassium contents. Soda glass appeared in the early first millennium BC and was popular west of the Euphrates until about 800 AD. Soda glass was produced only in certain areas of the Mediterranean, making it ideal for studying intercontinental contacts in the first millennium BC. Our knowledge of soda glass has continued to expand in recent decades, particularly during the Roman Empire and beyond. However, questions remain regarding the raw materials, production processes, and distribution routes of soda glass from the first millennium BC, such as early Iron Age glass, Phoenician glass, and Hellenistic glass. In addition, there are few reports of soda glass items dating from before the Han Dynasty (206 BC–220 AD) in eastern Eurasia. However, if the earliest Chinese glass was imported from the West, most of it would have been soda glass, which was dominant in the West at the time. In fact, the history of soda glass’s spread to the East is poorly understood: this article will attempt to correct this imbalance.
In this paper, we focus on soda glasses from Eastern and Western Eurasia dating to the 8th–2nd millennia BCE, and especially the mid-1st millennium BCE. By examining Mediterranean soda glasses at the macroscopic level, we propose a comprehensive microelement-based framework for grouping soda glasses from this period. In this technological context, we report the discovery of soda glass beads at the Ubao cemetery, which are among the earliest soda glass artifacts found to date in East and Central Asia. Combining this analysis with the composition and type of specific soda glass eyes, we propose a model for the production and distribution of dispersed soda glass wares in this period. Based on these results, we present, to our knowledge, the first comprehensive discussion of the spread of soda glass across Eurasia in the 1st millennium BCE.
Natron and trona (hereafter referred to as natron) are evaporite deposits consisting primarily of sodium carbonate. Most of the available sodium oxide probably came from Egypt, with the most notable source being Wadi el-Natrun17, about 100 km northwest of Cairo. Archaeological evidence, mostly from the Roman period, suggests that natron glass production involved a few large glassmaking centres that produced raw glass from the raw material, and many smaller workshops that melted down the raw glass ingots to produce the final product18,19. This must also date to the first millennium BC.
The history of soda glass began with a phase of experimentation and testing in the early 1st millennium BCE, which resulted in the earliest varieties of soda glass, such as the unstable low-calcium type, the cobalt blue type with a high aluminium content, and the black type with a high iron content20,21,22,23,24,25,26, most of which had a low calcium content. Several centuries later, chemically stable soda glass containing high amounts of calcium (around 5–8%) emerged, and the basic composition of soda glass has remained virtually unchanged since then. This “classic” type of soda glass has been found in the Mediterranean region since the 1st millennium BCE. , including in Italy25,26,27,28,29,30, Greece31,32,33,34 and Spain35, as well as Bulgaria36,37 and Georgia38,39 in the Black Sea region (summary in Supplementary Information (SI)). In addition, soda glass has been found outside the Mediterranean region, for example in Germany40,41, Poland41,42 and as far away as Mali.43 The intensification of glass production in the late Hellenistic period19,44 ultimately laid the foundation for Roman glass, when soda glass flourished as a centre of material culture.
To understand the natron-based glass production technology of this time, we studied the chemical composition of natron glass from the initial phase of the investigations until the late Hellenistic period (c. 8th to 2nd centuries BC). The macroanalysis is based on the compositional data of 145 natron glass samples from 12 sites (see Supplementary Text and Files). Based on the principal component analysis (PCA) of 18 major, minor and trace elements, we decided to mainly use the trace elements Ba, Zr, Ti, Sr, Nd, Th and La for our study, which can be divided into three geochemical groups: Sr and Ba; Zr and Ti; Nd, La and Th. Only one major element, Al, showed the potential to effectively distinguish between the different sample clusters. The PCA processing is described in the Methods Appendix and Supplementary Figures S1 and S2. The PCA results and 2D plots showed that the 145 samples could be consistently divided into three major component types (designated I, II, and III) and one minor type (designated I0) (Fig. 1; see also Supplementary Figs. S2, S3, and S4). Because some glasses are rich in lead and iron, all elements (except Fe and Pb) were compositionally normalized to the iron and lead oxides removed (i.e., A* = A/(1-PbO-Fe2O3), where A is the original element concentration and A* is the normalized concentration). When different glass types were available at a site, we divided the samples into numbered subgroups. The four compositional types sometimes overlap slightly, possibly due to glass recycling or mixing of materials. Other soda glass assemblages dating to this period may also be classified into these types27,28,33,34,37,38,42; details are given in the SI. Supplementary Table S2 provides a statistical summary of the concentrations of each element type.
The chemical composition shows that there are three major types (I, II, and III) and one minor type (I0) of soda glass dating to the 8th–2nd centuries BC. (a) TiO2 and Al2O3. (b) Nd and Sr. (c) Th and Zr. (d) Th and Ba. The specimens come from: Early ‘classical’ soda glass from southern Italy (Subgroup 1: 8th–6th centuries; Subgroup 2: 8th–7th centuries)25,26; Archaic Rhodes, Greece (640–600, two subgroups)31; MG-II specimens from Satricum, west-central Italy (4th–3rd centuries, groups A and B)30; Adriatic, northern Italy (Mediterranean groups-I, -II, -III: 5th, 3rd, 2nd centuries respectively)29; Pydna, Greece (6th–4th centuries)32; Methoni, Greece (6th–4th centuries, two subgroups)32; Son Mas, Majorca, Spain (Subgroup 1: probably 4th–3rd centuries; Subgroup 2: probably 3rd century)35; Dron-Delion, Bulgaria (late 6th century – first half of 4th century)36; and Wubao, China (760–481, discussed in the next section). All dates are BC. All elemental data are normalized (marked with asterisks). Not all samples contain all elements.
A quick inspection of the data (Table S2) provides an idea of the main properties of each silicene source type. From the Mg, K, and Al contents (Fig. 1a), the content of silicate traces such as feldspar, mica, pyroxene, and amphibole can be estimated. Types II and III are rich in Ca, and the Al and K contents are significantly higher than those of the other types. Since quartz sand is typically rich in calcium carbonate and contains aluminous clays and feldspars, it was likely used to make the original Type II and Type III glasses. Type I0 is low in Al, K, Fe, and almost all trace elements investigated here, and has a higher Si content than the other types, so a pure silica source26 such as crushed and ground pebbles or vein quartz is used. Type I is characterized by high and variable Ca and low and less variable Al, K and Fe contents, indicating the use of lime-bearing clasts with minor impurities (e.g. feldspar) as a source of silicon. Types I and I0 are similar in composition, except that the latter contains fewer impurities. Most type I and I0 samples are reported to be relatively rich in Mn.
The trace elements Sr and Ba are associated with lime, mica and feldspar. Type II belongs to the high Sr and Ba range, which is the opposite of types I0, I and III (Fig. 1b, d). Sr substitutes for Ca atoms in different proportions in different minerals. High Sr contents (or, preferably, Sr/CaO ratio) in glass usually indicate the presence of aragonite (as shells) in glassy sands from coastal environments, whereas low Sr contents indicate the predominant use of calcite (as limestone fragments) in the vitrified material45,46. This may be the case for type II and other types, respectively. Ba may substitute for Sr atoms and, to a lesser extent, for smaller Ca atoms. For type II, Ba mainly comes from Sr-rich shells, which explains its overall higher Ba content. Limestone provides part of the barium for types I and III. In addition, potassium feldspar and mica contribute additional Ba to type III (see SI for details).
The heavy mineral composition reveals important information about the origin. Ti is commonly found in rutile, ilmenite and sphene. Zr is closely associated with the presence of zircon, which usually first appears in weathered felsic igneous rocks such as granite in inland areas. Among all the types, the Ti and Zr contents in type II are the lowest (except for type I0) (Fig. 1a, c). The high Zr contents in types I and III indicate that the siliceous source comes from zirconium-rich deposits. The differences in the Zr and Ti ranges among types I, I0 and III may be caused by the heterogeneous distribution of heavy mineral crystals, suggesting that their source is coarse-grained, poorly sorted continental rocks. At the same time, the compact range of Zr and Ti in type II suggests that its original glass was a stable, well-sorted beach sand reserve with fine-grained, homogeneous particles.
Some rare earth minerals can indicate internal trends in sediments. Nd is a light rare earth element (REE) that is commonly associated with monazite and allanite and can be traced back to clastic sediments and ultimately to source rocks. The Nd-Sr trend of type II samples is nearly linear and extends into some type I samples (Fig. 1b). This correlation between these two divalent elements may be due to the presence of allantite, which contains both Nd and Sr. Th is often associated with heavy rare earth elements due to the similar size of their cations in accessory minerals such as zircon and monazite. Type III Th values are anomalously high compared to other types (Fig. 1c, d), implying that the type III barycrystal species are quite different, as are the different members of the monazite group. The Th-Zr correlation for type III glasses is positive (Fig. 1c), which is due to either the presence of Th in the zircon lattice or the coexistence of zircon with monazite/hesperite as accessory minerals. Thus, the vitrified materials of type III glasses are related to the specific geological environment of the provenance area.
Each type of silica has a different origin, depending on its chemical composition. Type II is similar to, but not identical to, Levantine Roman Type I glass.47 Roman-era writings link the Levant with glass production: Strabo mentions Sidon in Lebanon as a Roman center for glass production; Pliny the Elder suggests that the mouth of the Belus River served as a source of sand for glassmaking. However, direct archaeological evidence for the original production of pre-Roman soda glass is very rare (one late Hellenistic monument dating to the first century BCE has been discovered in Beirut.19) Taking into account all the available information, the most likely place of original production of Type II remains the Syrian-Palestinian coast. In fact, it has been suggested that the original glass of Type II specimens came from the Levant.29,32,35,36,37,38,42 Type III is rich in heavy minerals. Egypt is famous for its heavy mineral sands, which were originally mined in the upper Nile region49. Both Nile particles and African dust have high Nd50 contents, consistent with the unique characteristics of Type III sediments. It has recently been suggested that the original glass of some Type III artefacts was indeed of Egyptian origin34,35. Interestingly, the black, iron-rich bubble glass provides further evidence for this view, as its siliceous source may be geographically related to Type III (see SI). For Type I, the opposite trend in terms of low impurity and heavy mineral contents indicates that the quartz crushed stone used to make the glass was not quarried in the same region as Type III. With the exception of a few later examples from Greece, most Type I artefacts have been found in southern Italy or central Europe. These objects were likely made in Italy and traded across the Alps along the Amber Road. Furthermore, the unique type I0 is found only in southern Italy and in a limited number of samples, which may indicate local production of primitive glass in the central Mediterranean (other possibilities are discussed in SI).
Interestingly, although types I and I0 were produced and distributed far from Egypt, the best-known source of soda, their average soda content was higher than that of the other types (Supplementary Table S2 and Fig. S3a). Probably, since the vitrified material contains little Al, Fe or K, more flux was required to lower the silica melting point to an acceptable level and produce a well-vitrified product. For type I0, the lime component may have been added separately. 26 Although type I siliceous sources may contain limestone fragments, the addition of only additional limestone cannot be ruled out, since some samples are quite rich in calcium but generally do not contain any other impurities. The calcium and sodium contents of type I are very high, higher than those of type I0, and may thus reflect the continuity of the technical procedure of type I0. In addition, some black soda glasses from the same period (7th-6th centuries BC) use a unique silica source similar to Type I and, unlike earlier black glasses, may have had iron added separately. These processes may indicate active control of the functional components, i.e. the ability to add and tune individual components in response to the silica source to produce a desired glass composition. This requires a clear understanding of the functions of the various components and materials. Such an understanding did not exist for early black soda glasses or high-aluminum cobalt blue soda glasses, where the Fe in the black sand or the Al in the cobalt alum automatically acted as a network stabilizer without the glassmakers’ knowledge. Thus, around the 8th-7th centuries BC, glassmakers first created “classic” soda glasses by deliberately controlling the amounts of natron and lime added, marking a technological breakthrough in the history of soda glass. Although the later use of sand from the Levantine coast eliminated the need for additional lime and reduced the soda content slightly, from this time onwards glassmakers were able to adhere to the ‘classic’ recipe for producing high-quality soda glass, using the correct amounts of soda and lime, regardless of changes in the raw materials.
Finally, we consider the temporal order of these types to identify common trends. I0 and I are the earliest types. Type I0 dates mainly to the 8th century BC. Production of Type I glass probably began in the 8th century BC, and most artefacts date to before the 6th century BC. Since both Type I and Type II have been found at Rhodes (640–600 BC)31, Methoni (6th–4th centuries BC)32 and Poland (Hallstatt D)42, the transition between the two types probably occurred in the 7th–6th centuries BC. Type II has the largest number of recorded examples of all the types. Production of Type II glass may have begun as early as the 7th century BC. MG-I and some MG-II vessels are made from Type II parent glass. Dating of specimens from Greece and the Black Sea region31,32,33,34,36,37,38 indicates a continuous supply of Type II glass from the Eastern Mediterranean from the mid- to late 1st millennium BC onwards. Type III artefacts have been found in Italy and the Western Mediterranean. Although they are dated to the 4th–2nd centuries BC, this date should be considered specific to this region. Given the availability of raw materials in Egypt and the earlier appearance of black soda glass as a precedent, the production of Type III unworked glass may have begun earlier in Egypt. Type III includes MG-II and MG-III vessels. Since MG-II vessels in Italy could be made of both Type II and Type III glass, the supply of unworked glass to the region probably occurred during the MG-II period (4th–3rd centuries BC). Since the central and western Mediterranean was connected to the eastern Mediterranean and North Africa via Phoenician maritime trade, the transition from Type II to Type III glass may reflect a change in trade routes as the Phoenicians migrated west at the time. The export of Egyptian glass originals to the central and western Mediterranean was probably short-lived, as later Late Hellenistic and Roman glass reverted to Levantine glass originals.
In a review of the chemical data on Mediterranean soda glasses from the 8th to 2nd centuries BCE, we propose a framework for compositional types and their diagnostic features. Although the provenance of some samples has previously been proposed individually, our new interpretation, based on a large data set, is organized in a data-driven, top-down manner to explore important systematic trends. Using this framework as a guide, we attempt to identify technological relationships among soda glass finds in China, a country distant from the main soda glass circulation sites.
The Wubao site is located in Kezilkek, Wubao Township, Hami, eastern Xinjiang, 70 km from Hami urban area (Figure 2a). The site was inhabited from the Bronze Age to the early Iron Age. Several glass beads were found in the tomb. Unfortunately, we could only find one blue bead for analysis (Figure 2b). The bead belongs to a later period of occupation, dated to 760–481 BC. The archaeological background of Wubao is described in SI.
Information about Upju. (a) Upju is located in eastern Xinjiang, near the Eastern Tianshan Mountains. (b) Beads from Upju. (c) Close-up showing chains of bubbles trapped in glass. The map in (a) was created using QGIS 3.10 (https://qgis.org) based on the World Physical Map basemap (http://goto.arcgisonline.com/maps/World_Physical_Map) and the World Reference Overlay (http://goto.arcgisonline.com/maps/Reference/World_Reference_Overlay).
This spherical bead is dark blue in colour. Its surface is predominantly smooth. Its body contains numerous bubbles, which are formed as a result of the release of gases during the decomposition of the ingredients during the manufacturing process. The bubbles are spherical rather than flattened or deformed, suggesting that the glass was probably not stretched or compressed. In addition, the surface lacks the annular grooves caused by winding. Therefore, the beads were not made by drawing, cutting or winding, but were most likely made in a mould. Some of the bubbles have formed chains (Fig. 2c), probably due to the following mechanism. The resulting bubbles rise to the surface in groups. While the larger bubbles reach the surface and escape the melt, some small bubbles become trapped in the melt, which becomes increasingly viscous as the temperature decreases. As the glass forms into shape, the bubble chains take on the curvature of the bead shape. Confocal Raman spectroscopy showed that the amorphous Si-O network had only two broad envelopes around 500 and 1000 cm-1, indicating that the granules were well vitrified.
The chemical composition of the beads was obtained using LA-ICP-MS microdestructive analysis. The analytical methods followed protocol 51, which did not include internal standards, details are provided in the Methods Appendix. For LA-ICP-MS, moderate levels of surface corrosion are generally not an issue, as only data obtained after the initial ablation of the surface layer are used. For considerations regarding the accuracy and reliability of LA-ICP-MS analysis of archaeological glass, see Lü and Wu (2019)52. Calibration data are listed in Table 1. The overall uncertainty is typically within 10%, with the exception of Cr, Co, Ni, Zn and Sn, for which the uncertainty is estimated at 15% due to their highly variable distribution in the glass. Upp beads are made of natron-based soda-lime glass. It has low magnesium and potassium contents and low phosphorus content, but contains chloride (1.35% by semiquantitative X-ray fluorescence), which is typical of the use of soda as an alkaline flux. 17 The cobalt content of the Upp beads reaches 914 ppm. Co(II) is a strong dye that gives the beads a deep blue color. The antimony content of the beads was 0.6%, which is consistent with the Sb content of soda glass before 200 BC. 16 Since the Upp beads were translucent and no traces of opalescent calcium antimonate were detected in the Raman spectrum, we believe that the Sb dissolved and acted as a brightening agent.
Similar examples from Upps and China are extremely rare, and we rely on comparison with modern Mediterranean natron glass to establish provenance. The Upps bead has been classified as Type II (Fig. 1), indicating that it was made from Levantine proto-glass extracted from coastal sand. It is often difficult to determine the exact location of production of a second or last product based on ingredient information alone. In searching for potential links, we noted that most Type II artefacts were found in Greece, along the Black Sea coast, or in Italy. In the 8th–6th centuries BC, the Greeks began a major colonial expansion in the Mediterranean and Black Sea regions53, and maritime trade within the Greek world was active in the following centuries. Since this is consistent with the geography and age of Type II artefacts, it can be concluded that Type II artefacts may be closely associated with Greek influence. Some Type II artefacts, including the Uppu beads, may have been produced in a Hellenistic context in Italy, at a time when Greek settlements were spreading across Magna Graecia. The large number of Type II artefacts found in Italy and Central Europe from the mid-first millennium BC suggests that Italy may have been a centre of supply for Type II products. The trace elements in the Uppu beads are also similar to some Italian Type II examples. Of course, glass workshops were present in other Greek sites as well. The workshop at the Greek site of Yagorlyk in southern Ukraine (c. sixth century BC) was still producing plant ash glass and probably very limited quantities of primitive natron glass54 (but not of Type II composition). This suggests that the Black Sea region may not have been a major producer of natron glassware. It therefore seems more likely that the Voupu beads were made in the central Mediterranean, but the possibility that they were produced in the east, for example in a workshop in Greece, cannot be completely ruled out.
Since natron was not mined for glassmaking in East and Central Asia, and the raw material for natron glass was not found, early natron glass from China was imported as a finished product. Wubao beads are among the earliest soda glass objects discovered in East and Central Asia, and possibly the earliest. Its analysis also provided the first high-quality data on soda glass from the region. Prior to this study, eyeballs from Juxian55 and Xujialing56 in Henan Province and Zenghoui57 in Suizhou, Hubei Province were thought to be made of soda glass, and these beads were dated to the late Spring and Autumn period (771–476 BC) and early Warring States period (475–221 BC). In particular, 173 beads were found at Zenghoui. Recently, natron eyeballs dating to the late Spring and Autumn period were discovered at Dongdazhangzi, Jianchang, Liaoning Province58. Moreover, some artifacts previously described as soda-lime glass are likely natron-based. Based on published data, we suggest that soda glass has been found at Hougudui (late Y&A)59 in Gushi, Henan, central China, Leigudun Tomb No. 2 (WS)60,61 in Suizhou, Hubei, central China, and Majiayuan (late WS)62 in Tianshui, Gansu, central-west China. All of these artifacts are eyeballs. It is important to note that high-quality data on the original, uncorroded glass are needed for detailed provenance studies. The availability and quality of data across these analyses varied. For many elements, quantitative information is either heavily weathered or lacking due to limitations of analytical methods. We empirically identified potential soda glasses based on the available data and concluded that these beads exhibit major element levels characteristic of “classic” soda glasses (see SI). Figure 3 shows the locations in China where soda glasses may have been found. With the exception of the discoveries at Dongdachangzi and Majiayuan, the others were found in areas that belonged to or were heavily influenced by the state of Chu.
Due to limitations in the available data, we adopted a mixed typology approach to further examine natron-stratified eyeballs from China. Eyeballs are so called because smaller patches of one color are superimposed on larger patches of another color, resembling an eye. On the bead surface, these patches appear as mosaic-like areas containing concentric rings surrounding the pupil. Previous reviews of Chinese eye beads have considered bead shape and eye shape63 but have not focused on color. The colors of the different compositions indicate the glass materials available to the beadmaker and, by extension, the coloring methods used in a given workshop. Among spherical layered eyeballs found across Eurasia and dated to the mid-first millennium BCE, we distinguish four types, designated EB-a through EB-d, based on published images, primarily based on the color of the components. Each eyeball type has a unique combination of body, pupil, ring, and inlay area colors (Table 2), and the placement of the eyes is different.
Compositional data for some beads have been published. The Altdorf bead 41 (EB-b) is made of soda glass, and beads from Zenghoui 57 (EB-a, -b, -c), Xujialing 56 (EB-c), Hougudui 15 (EB-c), and Shanghai Glass Museum 64 (EB-a, -b, -d) may also be natron-based. In addition, the Pichvnari beads, which may be types EB-b and EB-d, are considered to be soda glass according to text descriptions38. Notably, beads from Dren-Delyan36 (EB-a, -b), Owidz42 (EB-b), and Son Mas35 (EB-d) have trace element data, all of which are type II. Beads of the same type may have the same or very similar composition. Therefore, we believe that all four types of eyeballs are made of soda glass, and at least EB-a, EB-b and EB-d are of type II composition, which is the same type as the Uppa beads. It is therefore reasonable to assume that these eyeballs are comparable in composition to the Greco-Italian soda glass.
The coloration of the base glass indicates a common technological basis. In these eyeballs, glasses of the same color contain the same colorants and opacifiers: white glass is rich in Sb and Ca, indicating the presence of calcium antimonate35,36,41,42,56,57,64; blue or blue-green glass is colored with Cu36,56,57,64; yellow glass contains large amounts of Sb and Pb and is colored with lead antimonate36,41,42,57; dark blue glass is colored with Co, and most of them also contain some Cu35,36,41,42,56,57,64, like the Upp beads.
Similarities in type and composition strongly suggest a common origin. From these common features, we infer that around the middle of the first millennium BC, most eyeballs were produced by a few specialist artisans using a limited range of standardized base glasses supplied by large workshops throughout the Mediterranean via a well-developed network. Since all four types of glass have been found on the Black Sea coast and in China in the 5th to 4th centuries BC, we suggest that glassware entering China was largely dependent on glassware available in the Black Sea region, and that large quantities of Mediterranean natron glass beads may have reached central China by the early Warring States period.
For soda glassware to have reached Xinjiang and central China, an extensive trade or exchange network must have existed. The Eurasian steppe is a vast, continuous ecological region that extends from Eastern Europe to Siberia and was once considered a corridor connecting the major civilizations of Eurasia.10,67,68 The steppes may have been the routes for the dissemination of Mediterranean glass. Before the historical Silk Road, the route that connected western and eastern Eurasia may be called the “Original Silk Road.” Figure 3 is a schematic diagram of the “life cycle” of soda glassware in the context of the Original Silk Road. Figure 3 highlights the locations where soda glass has been found in grasslands. Although rare, these finds suggest that the Mediterranean may have been connected to China via the steppes.
Three-stage model of the “life cycle” of soda glass. Production (yellow line): Unworked glass was transported from the Levant to workshops throughout the Mediterranean for secondary production. The solid line shows a possible route for Type II glass. Egypt also supplied unworked glass, but it is unlikely to have provided the raw material for the products shipped to eastern Eurasia. Distribution (purple line): As Greek colonization expanded, finished glass was traded throughout the Mediterranean and the Black Sea. Dissemination (red line): Nomadic peoples carried small glass ornaments with them as they traveled across the Eurasian steppes, effectively helping to move soda glass eastward. The routes shown are not entirely accurate. Sites east of the Black Sea where soda glass dating to the mid-first millennium BC has been found are marked with orange dots. Locations in China where early soda glass has been discovered: Central (State of Chu): 1 Xujialing, 2 Jiuxian, 3 Hougoudui, and 4 Zenghouyi/Leigudun; North: 5 Jianchang; West: Wupu, 6 Majiayuan. This map was created using QGIS 3.10 (https://qgis.org) using the Physical Map of the World (http://goto.arcgisonline.com/maps/World_Physical_Map) as the basemap.
The first phase of the original Silk Road involved links between production sites, probably involving highly active seafaring groups such as the Phoenicians. Unprocessed glass from Syria-Palestine, Egypt and elsewhere was shipped to secondary facilities across the Mediterranean, particularly Italy in this study, as shown in the ‘production’ line in Figure 3. The production of soda glass required not only the pooling of resources from different locations (trade in goods), but also the development of local expertise in glass processing at different stages (division of labour). This inter-regional cooperation can be seen as a prototype of the ‘supply chain’ that justified the existence of the early glass industry.
The finished glass products were supplied to the Greek market as well as to other Mediterranean markets. Trade between workshops and consumers resembles the multi-channel sale of goods that is represented as part of the “distribution” in Figure 3. Greek expansion also contributed to the spread of Mediterranean material culture in the Black Sea, bringing Mediterranean goods into contact with local groups.53 Blue-and-white eyeballs are often found in Scythian tombs near Greek colonies on the Black Sea coast.69 Museums in Georgia hold a large collection of blue beads and blue-and-white eyeballs from Colchis and Greek settlements (see for yourself).
From the Black Sea, nomadic peoples roamed the steppes eastward, probably carrying trinkets with them. Natron glass beads from the early Sarmatian period (4th–2nd centuries BC) have been found at Pokrovka, Orenburg, Russia, and are similar in style to beads from Scythian sites in the Crimea and Donbass.69,70 The Pokrovka beads include solid blue beads, blue and white eye beads, and black glass beads, representing several types popular in the Mediterranean. Eyeballs and core glass vessels from the Filippovka I66 cemetery also provide evidence of the role of the southern Urals in Greek and Mediterranean glass flows. In addition, glass with high alumina and sodium content has been found at Pokrovka and Pichvnary from the first millennium BCE, and potassium glass, which may have been produced in South Asia, has been found at northwestern Xinjiang and Pokrovka38,69,71. These results demonstrate the role of grasslands in connecting different sources and facilitating some degree of diffusion, which could occur in both directions. The long-distance dissemination of glass ornaments is unlikely to have occurred overnight. It has been suggested that pastoralists may have moved goods seasonally, essentially acting as intermediaries.69 It has been suggested that nomadic migration routes were precursors to the Silk Roads in the mountains of Inner Asia,72 and similar dynamics may exist for long-distance links across the steppes. Internal population movements between eastern and western Scythian groups in the first millennium BCE may also have facilitated the dissemination of material culture.73 Importantly, despite the lack of the involvement of intermediaries such as diplomats and professional traders who later travelled along the Silk Roads, the movement of early soda glass along the Proto-Silk Roads was highly directional and relatively rapid. This suggests that some form of integration of intergroup communication networks may have occurred by the mid-first millennium BCE. Glass ornaments were part of luxury adornments, often displayed as a symbol of social or political status, and demand for these items may have stimulated exchanges between groups. It has been suggested that the interconnected network of a small, well-known elite group of people may have been sufficient to disseminate exotic and valuable items over long distances.10 Although interactions between groups may initially begin with trade or the exchange of small quantities of valuable goods, the process gradually strengthens ties, increases the efficiency of interregional communication, and eventually leads to an expansion of economically significant trading activities and interactions. The route across the grasslands is represented by the ‘diffusion’ line in Figure 3.
Since the Bronze Age, eastern Xinjiang, where Wubao is located, has been a hub connecting the eastern Hexi Corridor and the northern steppes. Sites in the northern Tian Shan foothills of the Hami region date back to the second millennium BCE, and their pottery and bronze artifacts are similar to those of the Xiba culture of the Hexi Corridor.74,75 There were also routes into the steppes of eastern Xinjiang, and northwestern Xinjiang may have facilitated this connection. Imported painted ceramic beads with high soda content have been found at the Yar 76 cemetery near Wuburg. Some bronze artifacts found in the Hami region may have been made from ore from Ili in northwestern Xinjiang.77 Mixed alkali-stained ceramic beads were reported to have been found at the Tianshan Northern Road,78 while similar mixed alkali-stained ceramic beads were found at the Andronovo-influenced site of Wenquan in western Xinjiang, dated to the 19th–15th centuries BC and thought to be of European origin.79 It is also worth noting that Xinjiang can also be accessed via the Central Asian valleys. Increased rainfall created favorable conditions for social development in Central Asia during the 2nd millennium BC.80 Agricultural intensification and sociopolitical changes in Central Asia81 and the development of city-states in the Tarim Basin82 also date back to the 1st millennium BC. However, material cultural evidence points to early interaction with the steppes of eastern Xinjiang. Since few soda glass artifacts have been found in Central Asia, and the soda glass found in oases is dated to a later period than that found in eastern Xinjiang and central China83,84,85,86, it is currently impossible to link the spread of early soda glass to this alternative route, but future archaeological discoveries may update this understanding.
North–south movement between the steppes and north-central China is also possible, and material culture provides much evidence for this interaction, such as bronze weapons and horse-drawn chariots.87 This is supported by sodium-rich painted ceramic beads excavated in Shanxi Province in northern China in the early first millennium BCE, which may have been imported from the Near East.88 Although it is unclear how the Chu elite in central China were related to those on the northern steppes, and direct links seem unlikely, Chu, as one of the major powers in China in the mid-first millennium CE, was certainly well positioned to acquire exotic and valuable objects, thereby indirectly facilitating their movement. The fact that the natron glass beads found in central China represent eyeballs may reflect a cultural or aesthetic preference for eyeballs among the Chu people. Early soda glass could have entered central China via two routes: either via Xinjiang and the Hexi Corridor, or directly from the steppes. As for early soda glass, no single-color beads have been found in central China, and no eye-shaped beads have been found in Xinjiang so far. This seems to support the possibility that these two areas had direct contact with the grasslands. It may also be suggested that there was more than one chain of transmission, with the eastward movement of material culture bifurcating in the Altai Mountains.
The Central Plains region acquired decorative glassware from a variety of sources. Sometimes glasses made using different techniques are found in the same tomb, such as soda glass and wood ash glass in the tomb of Marquis Yi of Zeng,57 and soda glass and potash glass in Tomb No. 2 of Leigudong.60 The wide variety of early glass found in China probably reflects its original status as a recipient rather than a producer of glass. The demand for translucent decoration may have prompted China to first engage in glassmaking by imitating foreign glass, especially eyeballs. Eyeballs made of lead-barium glass89,90 and potash glass91 soon began to appear in Warring States tombs. Experimentation was needed before the glass could be successfully made. If imported glass was influential in the invention of Chinese glassmaking, then soda glass may have been used as a technical standard, since early Chinese glassmaking also used fluxes of mineral origin. Previously published evidence includes glasses containing lead-barium and fairly high levels of sodium and sometimes lime15,92, suggesting the possibility of an intermediate stage in the development of glassmaking or processing techniques. However, unlike Western soda glass, there is no variation in Chinese glass that reflects centuries of experimental composition. Chinese glassmakers, with no prior experience in glassmaking, quickly identified suitable conditions for glassmaking using local materials. This seemingly sudden start of glassmaking in China meant that some understanding of the functional components had to be developed. This may be partly due to the examples of primitive porcelain glazes mentioned earlier.93 It is also possible that as glass became more widespread in the Mediterranean, basic knowledge of the technical properties of glass was passed down orally in the constant flow of Mediterranean glassware and quickly found its way into existing Chinese ceramic and possibly bronze making skills.
In the following centuries, as glassmaking technology developed in the Mediterranean, the abundance of large solid beads and eye beads decreased, so that by the Han Dynasty they were rare in central China as well, indicating a transcontinental, interconnected glass market. Meanwhile, trade routes through the oasis states of Xinjiang and the Hexi Corridor became more important during the Han Dynasty. Further imports of soda glass to China were via the modern Silk Road, with much of it coming from the Roman and Sasanian periods.94 Barium lead glass also appears in Xinjiang,86,95 reflecting China’s growing influence.
In this study, we evaluate first-millennium BCE soda glass and examine its early eastward spread across Eurasia. Of the four types of Mediterranean soda glass technology identified here, Type II is arguably the most influential type of soda glass in the East, as demonstrated by the analysis of a single soda glass bead from Xinjiang, and this is true for most, if not all, of the four soda glass beads that arrived in China from the Mediterranean and Black Seas. We propose a three-stage model for the spread of soda glass, involving an ancient “supply chain,” a distribution network, and rapid delivery between populations at each stage. As glass beads spread eastward, they transformed from a common commodity into an exotic status symbol, becoming part of a prestige goods economy that in turn facilitated the integration of intergroup networks.
The picture we present to our readers is far from complete. The next step in our future research will be to integrate isotopic analysis with dyes and opacifiers. Our discussion is necessarily limited by the available samples and dates. The discovery of archaeological evidence of production activities will improve the study of production sites. Because of the large geographical gap between East Asia and the Black Sea, future analysis of glass found in Russia, Kazakhstan and Mongolia will be important in completing the overall picture.
The original Silk Roads, like the historical Silk Roads, were a network of communication channels between East and West. As this study shows, sustained contacts on the steppes before the rise of China and the Roman Empire may have led to the introduction of Mediterranean natron glass in Xinjiang by the mid-first millennium BCE to the early fifth century BCE, no later than its introduction to the Mediterranean. By the first half of the fifth century BCE, increasing numbers of natron glass beads had reached central China, possibly giving rise to indigenous glassmaking in China. The flow of natron glass beads reflects the relatively rapid and unimpeded dissemination of material culture along the original Silk Roads, which was by no means the result of random movement or sporadic exchanges. Our research suggests that transcontinental connections became increasingly efficient and interactive in the first millennium BCE, making possible the coming era of intensive communications and specialized trade.
To determine whether the samples could be successfully clustered, we processed the compositional data using principal component analysis (PCA) and considered the contribution of different elements. We examined data for the following 18 elements: Si, Na, Mg, Al, K, Ca, Fe, Pb, Ti, Ba, Zr, Sr, Nd, Mn, Cr, V, Th, La. The resulting first and second principal components (PCs) explained a total of 45.2% of the variance (23.6% and 21.6%, respectively). Individual data points could be successfully grouped into four clusters: I, II, III, and I0. The elements that contributed the most to PC were Sr, Ba, Th, Al, Nd, and Ti, followed by Zr and La. The elements most closely associated with PC1 were Th and Nd, and those most closely associated with PC2 were Sr and Ba. The results are generally consistent with the mineralogical understanding of the roles of different elements in glass composition: trace elements are the most promising indicators of the original minerals in the raw material; for most major elements, comparative trends are not easy to discern, since overall variations in mineral abundances can lead to variations in major element concentrations; most REEs are considered to show roughly uniform patterns that are not indicative of a source18,47, since the origin of these REEs is obscured by numerous associated minerals; However, some LREEs are notable exceptions and may indicate the presence of accessory minerals. Based on the PCA results, we decided to use the trace elements Ba, Zr, Ti, Sr, Nd, Th and La, as well as one major element Al, for our analysis. Supplementary Figure S1 shows the correlations between elements and PC, indicating that these eight elements are best represented in PC and are therefore suitable for distinguishing between the different sample clusters. All of these elements are associated with siliceous sources, so their bivariate plots provide important information on the primary production. From a mineralogical point of view, the trace elements can be divided into three geochemical groups: Ba and Sr, which are usually associated with lime, mica or feldspar; Zr and Ti, both associated with heavy minerals; Nd and La, which are light rare earth elements, are also associated with some heavy minerals. Th often coexists with rare earth elements and heavy minerals and has characteristics of the last two groups. The major element Al is associated with silicate traces such as feldspar. After reducing the data set to the selected elements, we ran the data again with PCA to ensure consistent clustering of the samples. The first two principal components now explain 83% of the total variance. Supplementary Figure 2 shows the four groups identified by PCA. In order to demonstrate the grouping results with mineralogical and archaeological significance, we chose to plot the following two-dimensional relationship diagrams to show the four compositional types sequentially: Fig. 1 (a) Ti vs. Al; (b) Nd vs. Sr; (c) Th vs. Zr; (d) Th vs. Ba; Supplementary Fig. S3 (a) Al vs. Na; (b) Sr vs. Ca; (c) Sr vs. Ba; (d) Ti vs. Ba; (e) K vs. Ba; (f) Zr vs. Ba; Supplementary Fig. S4 (a) 1/Zr vs. 1/Ti; (b) Ti vs. Nd; (c) Sr vs. Ti; (d) Zr vs. Nd; (e) Th vs. Nd; (f) Th vs. Sr; (g) La vs. Zr; (h) Th vs. La. To better illustrate the different clusters, we generally use elements from different geochemical groups to represent the two-dimensional relationships, although some element combinations were chosen to stimulate discussion in the text. As a side note, we have found that if most trace element data are not available, compositional type can usually be inferred from Al and Ti contents, although this should be taken with a grain of salt if trace element data are not available.
The LA-ICP-MS analyses were carried out at the Key Laboratory of Crust-Mantle Materials and Environment, Chinese Academy of Sciences, University of Science and Technology of China96. The samples were first cleaned in an ultrasonic bath to remove surface contaminants. The signal intensity was recorded using an Agilent 7700e ICP-MS instrument coupled with a GeolasPro ArF excimer laser sampling system (193 nm). The ablation spot aperture size was 44 μm and the repetition rate was 10 Hz. Helium was used as the carrier gas at a flow rate of 900 mL/min, and argon was used as the make-up gas, which was mixed with helium before entering the ICP. For each analysis, after approximately 20 seconds of background data collection, the sample was laser ablated for 40 seconds for data collection, followed by a gas flow purge for approximately 35 seconds. We performed the calibration using the procedure described by Liu et al. 51 using multiple external standards and no internal standard. This calibration protocol calculates the ablation yield correlation coefficient (AYCF) by normalizing the total metal oxide abundance to 100%, eliminating the need for internal standards. The sequence begins and ends with the analysis of five reference materials (NIST SRM 610, NIST SRM 612, BHVO-2G, BCR-2G, and BIR-1G). NIST SRM 610 is analyzed regularly to correct for time drift. The data were processed using ICPMSDataCal software 51. A total of five surface points were analyzed and the calibration data were averaged to obtain the final value. The uncertainty in the concentration of each element is a combination of the systematic error of each ablation data point and the standard deviation of multiple images.
All data used to inform the analyses of this study were presented in this manuscript or published in the cited references. A compilation of these data is also included in the supplementary spreadsheet file.
Haak, V. et al. Large-scale steppe migrations are one of the sources of the spread of Indo-European languages in Europe. Nature 522, 207–211 (2015).
De Barros Damgaard, P. et al. 137 ancient human genomes from the Eurasian steppe. Nature 557, 369–374 (2018).
Spengler, R., et al. “Early agriculture and crop dispersal among mobile Bronze Age pastoralists in Central Eurasia.” Proc. R. Soc. B Biol. Sci. 281, 20133382 (2014).
Zhou, S., et al. 5,200-year-old cereals from the Eastern Altai provide new insights into the dating of trans-Eurasian crop exchange. Nature Plants 6, 78–87 (2020).
Jordan, P. et al. Modelling the spread of pottery technology in Afro-Eurasia: new ideas and future research. Antiquity 90, 590–603 (2016).
Killick, D. and Fenn, T. Archaeometallurgy: the study of pre-industrial mining and metallurgy. Rev. Anthropol. 41, 559–575 (2012).
Long, T. et al. Early history of wheat in China from 14C dating and Bayesian dating modeling. Plant Nature 4, 272–279 (2018).
Post time: Mar-27-2025