Lead isotopes link copper artefacts from northwestern Botswana to the Copperbelt of Katanga Province, Congo

Highlights

• Radiogenic lead isotopic ratios show that Tsodilo Hills copper came from the Central African Copperbelt >1000 km away.

• Tsodilo Hills copper is isotopically and chemically similar to ingots from the Upemba Depression cemeteries of southern DRC.

• The very high purity of much Tsodilo and Upemba copper suggests smelting of pure malachite.

Abstract

Copper was highly valued in sub-Saharan Africa for jewellery and as a store of wealth, but was rarely used for tools or weapons. The Central African Copperbelt is one of the world's largest copper deposits, and is known to have been mined since at least 400–600 cal CE, but has seen very little archaeological investigation. We measured lead isotope ratios and trace element concentrations in 20 copper objects, dating between ca. 650 cal CE and ca. 1200 cal CE, from two sites in the Tsodilo Hills in northwestern Botswana. The results show unequivocally that almost all derive from Copperbelt ore deposits in Katanga Province, Democratic Republic of the Congo, at least 1050 km from Tsodilo. Our results are very similar to those recently obtained for a suite of 45 copper ingots, dated between 9th and 18th centuries cal CE, most of which are from cemeteries in the Upemba Depression, about 200 km north of the Copperbelt (Rademakers et al., 2019).

Introduction

The Central African Copperbelt extends for about 550 km along the border between Zambia and Congo and is up to 150 km wide (Fig. 1). It is the largest copper deposit on the African continent, estimated to contain around 140 MT copper and 6 MT cobalt (Cailteux et al., 2005). The earliest historical records of copper production in this region date from the 1760s, when African traders arrived at Portuguese feiras on the Zambezi with copper and ivory to trade (Sutherland-Harris, 1970). The Portuguese learned that these came from the territory of the Lunda paramount chief Kazembe, and sent expeditions to his capital on the Luapula river, south of Lake Mweru, in 1796, 1802 and 1831 (Burton, 1873; Gamitto, 1960). The earliest eye-witness account of copper mining and smelting on the Copperbelt is a brief mention by two Portuguese who between 1802 and 1810 crossed the subcontinent from modern Angola to the Zambezi River in present Mozambique. In 1804 they observed copper mining and smelting, and in 1808 noted that “to the west of the Luapula [river], green stones are found in the ground called Katanga” (Gamitto, 1960, vol 2:80–85). There was no European interest in mining these deposits until the European nations divided Africa among themselves at the Berlin Conference of 1884/1885. In 1895 the South African mining magnate Cecil Rhodes obtained a concession for his British South Africa Company (BSAC) over the territory that is now Zambia. The first BSAC copper mine opened in 1908, but large-scale mining of the Zambian portion of the Copperbelt did not begin until after the British government terminated the BSAC concession in 1924 (Sikamo et al., 2016). Industrial mining of copper in the Belgian portion of the Copperbelt began, with forced labour, in 1911.

Because industrial mining began so late in this region, there are a substantial number of written accounts, oral histories and photographs of African mining and metallurgy on and around the Copperbelt (discussed by Herbert, 1984 and Musambachime, 2016). In Zambia, only the Chibuluma mine deposits had not been mined before Europeans arrived (Sikamo et al., 2016:491). Copper was still mined by Africans in parts of the Copperbelt as late as 1924 (de Hemptinne, 1926) but ceased soon after this when the Belgian and British colonial governments expropriated the ore deposits. This included the famous Shinkolobwe deposit, which supplied much of the uranium ore for the two atomic bombs which were launched at the end of WWII (Fleckner and Avery, 2005). As late as 1971 there were still former copper workers in Zambia able to provide successful demonstrations of indigenous copper smelting technology (Bisson, 1976; Miller, 1994).

We are fortunate to have these historical records, as almost all traces of the ancient copper mines in Zambia have been destroyed. After 1924 foreign investment poured into the Copperbelt mines, and by the late 1960s Zambia produced 12–13% of the world's copper (Sikamo et al., 2016). By 1970 the only large precolonial mine remaining in Zambia was Kansanshi, which was investigated by Michael Bisson for his doctoral dissertation, under the direction of Brian Fagan (Bisson, 1976, 2000). This is still the only archaeological study of precolonial mining in Zambia or in the Democratic Republic of the Congo. Until recently, industrial exploitation of the very rich copper deposits in the Katanga (Congo) portion of the Copperbelt had been much less intensive than in Zambia, but since 2000 several massive new mines have opened and more are in development. At least one of these mining companies has commissioned a cultural heritage survey (Arazi et al., 2012) but at present there are only three archaeologists in Congo – the largest nation in Africa by area – with training to the PhD level. It is therefore highly unlikely that any archaeological investigations of ancient mines will be undertaken before they are destroyed.

Because of the lack of research and the destruction of ancient mining sites, the prehistory of mining in Copperbelt ore deposits must be inferred indirectly from dated copper artefacts that can be linked by form or by chemistry to Copperbelt mining districts. The earliest published radiocarbon measurement for copper smelting on the Copperbelt is 1555 ± 65 bp (Hv 11402; Anciaux de Faveaux and de Maret, 1984) which gives a calibrated age of 416–648 CE (Calib 7.01, SHCal13, 2 sigma). Excavations during the 1960's and 1970's of cemeteries in the Upemba basin, about 200 km north of the Copperbelt, recovered many hundreds of distinctive HIH, HX, and HH copper ingots (de Maret, 1985, 1992, 1995). Radiocarbon dates from these cemeteries have allowed these shapes to be placed in a chronological sequence from the 7th through the 18th centuries cal CE (de Maret, 1981, 1995; Nikis and Livingstone Smith, 2017). Moulds matching some of these shapes have been found at prehistoric mines on the Copperbelt. Copper ingots in these same shapes have also been found as far away as southern Zimbabwe (around 1100 km from the Copperbelt) and central Malawi (around 950 km from the Copperbelt), but so (in some cases) have moulds (Nikis and Livingstone Smith, 2017; Swan, 2007). This suggests that imported ingots sometimes may have been used as templates for moulds, which were then used to cast copper smelted from local ore deposits. Furthermore, the great majority of copper objects from archaeological excavations in the southern third of Africa are not ingots, and do not have distinctive shapes that could link them to the Copperbelt. Clearly some criteria other than shape must be used to infer the geological provenance of these copper objects.

Lead isotopes and trace elements are the data of choice for inferring the geological provenance of copper objects (Pernicka, 2014). Lead isotope ratios are the first line of attack because they are not altered (fractionated) by smelting, whereas the concentrations and ratios of many trace elements in ores are altered during smelting. However, a restricted set of elements is partly or entirely transferred from the ore into metallic copper. These include gold, silver, nickel, cobalt, bismuth and the platinum group elements (PGE). The absolute abundances and ratios of these elements are often useful in deciding what type of ore was smelted, which may help to constrain geological provenance.

There is only one laboratory for high-precision heavy isotope measurements in Africa (at the University of Johannesburg), so this technique has only recently begun to be applied to southern African archaeology, with measurements made in American and Belgian laboratories (Molofsky et al., 2014; Rademakers et al., 2018, 2019). (For an earlier low-precision study that measured only ²⁰⁷Pb/²⁰⁶Pb ratios, see Miller et al., 2005). The southern third of the African continent should be, in principle, among the most favorable regions in the world for provenance with lead isotopes. This is because its copper ores were formed over an extraordinarily long time range (3300–400 Ma) and therefore should exhibit a much wider range of lead isotope ratios than regions like the Mediterranean or the Andes, where ore formation was concentrated into much shorter intervals of time. To investigate this hypothesis, the first author collated and plotted 559 published lead isotopic measurements on metallic sulfides and carbonates in the geological literature for mining provinces in southern Africa (Stephens, 2016). (Although there should in theory be no fractionation of lead isotope ratios when sulfides oxidize to carbonates, this remains to be checked.) The main conclusions obtained are these:

• There is a much wider range of isotope ratios in southern Africa than in any other region where lead isotope analysis has been used by archaeometallurgists. Fig. 2 plots two lead isotope ratios for these samples against those for non-ferrous ore samples for the circum-Mediterranean, where most copper ore deposits were formed during the Hercynian orogeny (330–270 Ma), and the Andes cordillera, where most metal deposits formed within the last 70 Ma.

• Major mining districts are generally well differentiated from each other by lead isotopes, except on the Zimbabwe craton, where most ore deposits are of Precambrian age and probably have similar lead isotope ratios.

• The ore deposits of the Copperbelt are among the youngest deposits, and have the highest (most radiogenic) lead isotope ratios. Copperbelt deposits are easily distinguished from those of Zimbabwe, South Africa and Botswana. At the lower end of this range the ratios for some Copperbelt deposits do overlap those of northern Namibia, since ore deposits in both regions were formed by the same geological event, the Damaran-Lufilian orogeny (ca. 590–512 Ma).

No lead isotope measurements are yet available for some ore deposits with evidence of preindustrial mining, so the first and second authors are currently expanding this database by measuring lead isotope ratios for ore samples from these deposits.

The type of metal analyzed is also an important consideration in a lead isotopic provenance study, as alloy type and composition can affect the interpretation of the results. In southern Africa copper was mostly used unalloyed, and therefore lead isotope signatures should reflect the geological provenance of the copper metal. Before 1000 cal CE copper is relatively scarce in the archaeological record (except in the Upemba basin of Katanga, Fig. 1), and was used exclusively for small items of jewellery (Miller, 1996, 2003). After 1000 cal CE there was a marked increase in the amount of copper deposited in archaeological sites, and copper was used as a form of wealth (as ingots, or, after 1300 cal CE, drawn into wire) as well as in jewellery (Miller, 2010). Copper was very rarely used for tools or weapons; almost all of these were made of iron. Although there are many tin deposits in southern Africa, bronze does not appear in the archaeological record of this region until after 1200 cal CE (on current evidence). Some bronze was certainly made in northern South Africa after 1200 cal CE, and can be distinguished from bronze imported from the Islamic world by having very low concentrations of lead, and often by having anomalously high lead isotope ratios. Molofsky et al. (2014) show that these African bronzes must have been made by adding tin containing tiny amounts of radiogenic lead to copper that was almost lead-free – so in these cases the lead isotope ratios indicate the source of the tin, not the source of the copper. No bronzes at all have been reported from the cemeteries of the Upemba depression in Katanga (Fig. 1), which have been radiocarbon dated to the 7th through the 18th centuries cal CE.

Inferring provenance of brass using lead isotopes can be tricky, as lead is often added during the production of this alloy. The very rare objects of brass dated before 1500 cal CE in the southern African archaeological record are certainly imports from the Islamic world, probably brought by Swahili merchants, as the few analyzed samples of copper alloys in Swahili sites are mostly leaded brasses (Kusimba et al., 1994; Killick and Fenn, unpublished data). There is no evidence that brass was made in southern Africa until the nineteenth century, when imported metallic zinc (from India?) may have been blended with local copper (Miller, 2010). Also in the nineteenth century, African metallurgists near the mouth of the Congo River, some 1100 km NW of the Copperbelt started to produce copper-lead alloys (Rademakers et al., 2018). A 15th-17th cal CE sample from the Niari Basin also has a high concentration of lead (more than 2%), but it is unclear if this is a product of accidental co-smelting or a voluntary mixing of lead and copper (Rademakers et al., 2018). To our present knowledge the nineteenth century marks the first intentional production of lead metal and leaded copper in the southern half of the African continent. Both the copper in Swahili sites (Kusimba et al., 1994; Killick and Fenn, unpublished) and copper from the central European mines of the Fugger family, who were the main suppliers to the Portuguese (Hauptmann et al., 2016), contain several percent lead. The overwhelming majority of analyzed copper objects from archaeological sites in southern Africa are of unalloyed copper. Provenance can be obscured by melting together copper from different sources, but we cannot know how much of a problem this may be until we actually try to fit lead isotope ratios of artefacts to those of potential sources.

The Tsodilo Hills are a group of four inselbergs in the northwestern corner of Botswana, at the northern edge of the Kalahari Desert (Fig. 1). Excavations in rock shelters within these hills show that some were occupied from the Middle Stone Age (beyond the range of radiocarbon dating). In 2001 the Tsodilo Hills were designated a UNESCO World Heritage Site to protect their extraordinary wealth of rock art (Campbell et al., 2010). There were only hunter-gatherers in this region until around 2000 bp, when the appearance of cattle, sheep and the first pottery in archaeological sites signals the arrival of pastoralists from the north (Robbins et al., 2008).

After about 650 cal CE, new populations arrived and occupied the only two flat plateaux in the Tsodilo Hills. These occupations produced the archaeological sites of Divuyu (Denbow, 2011) and Nqoma (Wilmsen, 2011). These peoples made pottery, cultivated sorghum, herded cattle, sheep and goats, and discarded small items of iron and copper jewellery (Miller, 1996). Divuyu pottery styles are limited in variety and closely resemble wares in central Angola and the lower Congo region (Denbow, 2014:166–168). The Divuyu plateau was abandoned ca. cal 800 cal CE and never reoccupied. Nqoma presents a more complex picture. In the earliest levels – prior to about 900 cal CE – sherds are in part similar to those of Divuyu (which continued to be deposited at Nqoma even after the abandonment of Divuyu); but there are other distinct sets of sherds that are earlier or perhaps contemporary with Divuyu, and which may have been made by pastoralists. Many of these are thin-walled with a hard paste and a variety of design techniques employing surface patterns, including what may be cord-rolled texturing overdrawn with finely-incised curvilinear motifs; many of these vessels had a red slip. From about 900 to 1200 cal CE at Nqoma, collared, thick-rimmed jars and small elaborately decorated serving bowls associated with Zambezi traditions around Victoria Falls and northern Zambezi are the most common in all parts of the site. This must reflect a significant influx of peoples from what is now southern Zambia (Denbow, 2014:168–172). Our analyses by optical petrography of sherds from Nqoma and Chobe-Zambezi sites, along with clays from a large number of parent rock exposures, have confirmed that some pots from Nqoma were made in the Zambezi valley, some 400 km to the northeast (Wilmsen et al., 2009).

The major attraction of the Tsodilo Hills appears to have been thin seams of specular hematite (specularite) within the quartzite and mica schists that comprise the hills. Specularite (Fe₂O₃) is a glittering blue-black mineral that was a valued cosmetic in southern Africa from very early times to well into the twentieth century. At Tsodilo hard rock mining for specularite was carried out on a large scale from an as-yet undetermined date; 20 mines have been located (Campbell et al., 2010), with 500–1000 tons of rock excavated from some of the larger examples. A series of radiocarbon ages indicates that mining, complete with tunnels and inclined shafts up to 43 m long, was most intensive during the Divuyu/Nqoma period, cal. 750–1025 cal CE (Wilmsen et al., 2013). The current name Tsodilo almost certainly stems from the Tswana word for specularite, sebilò, thus forming a symbiotic association of the place with its highly valued product.

Shortly after mid-8th century, glass beads of the Chibuene-Zhizo series – manufactured from glass probably made in northeastern Iran – arrived at Nqoma along with marine cowrie and conus shells, marking the start of indirect connections between the center of the continent and the Indian Ocean littoral (Denbow et al., 2015:2; Wilmsen, 2017). The Tsodilo Hills were clearly a magnet for diverse peoples for several millennia, but this came to an end about 1200 cal CE when these plateaux were abandoned. This decline is noted at other sites in Botswana, and possibly reflects the growth of communities along the Limpopo valley, between present South Africa and Zimbabwe, which became the endpoints for Swahili travelers (Denbow et al., 2015).

Although metal was first introduced to the Tsodilo Hills by the people who settled at Divuyu, it is relatively uncommon at that site; iron and copper together make up just 1% of the site inventory. There were 196 pieces of iron, 11 of copper, and two bimetallic (Miller, 1996, Table 2). At Nqoma the picture is diametrically opposite: for a brief time, from ca. mid-8th century to ca. mid-11th century, Nqoma was the richest southern African site yet known in iron and copper: 2673 iron, 191 copper, and two brass pieces were found (Wilmsen, 2011, Table 3 – these numbers supersede those in Miller, 1996, Table 3). About 25% of the copper at Nqoma is from the lower levels and associated with Divuyu pottery; about 75% is from the upper levels representing the arrival of new populations between ca. 900 and ca. 1200 cal CE. The two brass objects were both near-surface, and are a boot eyelet and a cartridge from a firearm. The most striking finding is that in both sites the metal objects in both iron and copper were overwhelmingly small fragments of jewellery (beads, twisted helices, flat strip, chain, wire, etc. – see Miller, 1996: Table 2, Table 3). The rarity of iron tools at Divuyu and Nqoma – a few arrowheads, chisels, axes and pointed rods – is interpreted to mean that iron tools were too useful to be casually discarded or placed in burials (Miller, 1996:92–96).

Miller published metallographic studies of 14 copper artefacts (four from Divuyu, ten from Nqoma) (Miller, 1996). He measured bulk chemical compositions and the compositions of non-metallic inclusions in seven samples by energy-dispersive XRF on a scanning electron microscope (SEM-EDAX). No tin, zinc, arsenic or lead was detected in any of these (detection limit ca. 0.1 wt %). Two of the objects from Divuyu contained detectable Fe (1.2 wt% and 7.0 wt%), and one from Nqoma had 1.5 wt%; the other four had <0.1%. Non-metallic inclusions in most of the specimens were copper sulphides with up to 12 wt% Fe; some inclusions had detectable silver (0.1–2.5 wt%). One specimen from Divuyu had oxide inclusions containing Fe, Cu and Ni; and one from Nqoma, with 4.5 wt% P, had iron phosphide inclusions (Miller, 1996: Tables 5 and 8).

The copper artefacts are of particular interest because there are no copper deposits near the Tsodilo Hills that could have been exploited a thousand years ago. The Central Kalahari Copper Belt (Borg and Gauert, 2018) was not accessible because it is buried under 100–200 m of aeolian Kalahari Sands. The closest ore deposits with surface exposures (Fig. 3) are at Tsumeb in Namibia (425 km west), and within Botswana in the Precambrian greenstone belts of the Bushman Shear Zone (>545 km east) and the Matsitama Schist Belt (>585 km east). Salvage excavations have been done on 5 of 59 known precolonial copper mines in Botswana (van Waarden, 2016) but only a few radiocarbon dates are available, all of them after about 1500 cal CE. This does not exclude the possibility of copper production before 1500 cal CE, and more work is needed to clarify the start of copper mining in Botswana. No archaeological investigations have been done yet around Tsumeb in Namibia.

For this study 20 copper artefacts were selected by Wilmsen for analysis. Eight from Divuyu represent over half of the total copper objects excavated from the site; the 12 from Nqoma are only 6% of the 191 total copper objects excavated (Wilmsen, 2011: 100). The sample includes a wide variety of object types, and represents all of the recovery contexts at Divuyu and seven of the ten recovery contexts for copper objects at Nqoma (Denbow, 2011: 78; Wilmsen, 2011: 100). They were all very small, with masses between 0.06 g and 0.25 g (Table 1). All sampled objects are associated with dates from 650 to 1190 cal CE and we believe these to be representative for the dates of the sampled copper objects, as evidence for bioturbation was minimal at both Divuyu and Nqoma. The samples from Nqoma span the whole period of occupation. Samples N1143, N1429, N1625, N2340 and N247 were associated with Divuyu-style pottery and thus dated roughly between 650 cal CE and 880 cal CE (Table 1). Samples N1075, N1150, N1680 and N1682 were associated with only Nqoma-style pottery, and thus date in the range 900–1200 cal CE. Three samples – N40, N2500 and N1221 – were not securely associated with any particular style of pottery but either have radiocarbon dates which calibrate to the Nqoma period of occupation or are stratigraphically associated with radiocarbon samples that date to the Nqoma period of occupation (Table 1).