Regular Article
Production of Silver across the Ancient World
2014 Volume 54 Issue 5 Pages 1085-1092
Details
2014 Volume 54 Issue 5 Pages 1085-1092
The quest for silver through antiquity encouraged a succession of major developments across a variety of technologies. Silver and its minerals occur in a variety of ores, but rarely in more than trace amounts such that in order to discover and extract them various special technologies had to be developed, resulting in the first separation of small quantities from the fourth millennium BC. Through the first millennium BC there was a steady increase in the demand for silver, much accelerated by the introduction of coinage from the Mediterranean to South Asia that led to the major developments in all aspects of mining technology. The continuing demand led to new technologies to extract silver from copper and to recover metal from smelting debris.
Silver occurs as a native metal in limited amounts and also as mineral ores such as the sulphide, argentite, Ag2S and as the chloride, ceragyrite, AgCl, but once again in only very limited quantities.1) Silver is far more prevalent as a minor component in other metal ores, gold and copper, but particularly in those of lead. The presence of minor quantities of silver in lead ore in the Middle East seems to have been known, and extracted from the fourth millennium BC. This was man’s first attempts to separate trace amounts of one metal from another. This was done by the process of cupellation, by which the argentiferous lead was subjected to an oxidising blast at around 1000°C, oxidising the lead to litharge, PbO but leaving the silver as a separate metal phase2,p.221-8,3) The evidence for this was the presence of litharge, which does not occur as a natural mineral, at smelting sites in northern Syria and in eastern Anatolia.4,5) Pieter Meyers6) has studied the composition of early silver artefacts in the Eastern Mediterranean for indications of the likely ores and processes used. Silver extracted by cupellation should contain traces of lead (at least 0.05%, but typically at least an order of magnitude higher), but only negligible amounts of volatile elements such as zinc. For example, the silver artefacts of the Early Bronze Age Argaric culture in Spain of the second millennium BC have very low lead contents and are thus likely to have been smelted directly from silver ores.7,8) The silver artefacts from Mahmatlar in eastern Anatolia, dated to the third millennium BC, have substantial traces of zinc and thus cannot have been cupelled.9)
Thus it is possible to begin to construct a scenario for early silver production. At first the very limited amount of native silver could be utilised, and some would inevitably have been associated with argentite and ceragyrite, which although not especially colourful or shiny compared to copper or even lead ores, are dense and thus of potential interest, and would have been easily smelted. The discovery of cupellation is a little more problematic, but the silver minerals are often associated with lead ores and the resulting mixed metal after smelting would have been decidedly disappointing in appearance. However carrying out the usual metal refining technique of melting in an open crucible would have progressively oxidised the lead, enabling it to be skimmed off. From this it would soon become apparent that small quantities of silver were sometimes to be found in lead ores, but that both high temperatures and an air blast were necessary if the lead was to be oxidised in a reasonable time (They also probably soon noticed that the practitioners of this process often became ill, but this never seems to have concerned anyone right up to the 20th century!).
As noted above, native silver or minerals concentrated enough to be smelted directly only occur in very limited quantities, and already by the first millennium BC virtually all silver must have come from ores containing only a few thousand parts per million at most. The lead ores could either be the primary sulphide, galena, PbS or the secondary carbonate, cerussite, PbCO3, formed by the weathering of the sulphides, and thus tending to occur near the surface. Meyers6) noted that the gold content of the oxidised ores is very much higher than in the primary ores. This is because in the weathering process some of the lighter more electro negative metal ions tend to dissolve and drain down into the deposit but the gold is unaffected and left behind, and thus becomes concentrated.
The near surface deposits would be expected to be the first exploited. At the great Mauryan-period silver mines in the Aravalli Hills (see below), the heavily oxidised, known as gossanised, surface deposits suggest that in the upper levels of the mines the ore would have been principally cerussite10) (although the primary argentiferous galena was certainly also worked from the lower levels). The few contemporary early punch-marked silver coins that have been analysed have gold contents between 0.7 and 1.3%, suggestive of oxidised ores.11) A somewhat similar situation is found at Laurion in Greece where Conophagos12) believed that the upper levels would have been mainly of cerussite. However, if so then these must have been worked out early in the mine’s long history, probably already in the Bronze Age. Analysis of the blebs of metal in some of the Classical period slags led Photos-Jones and Jones13) to believe that cerussite was still the main ore used, but more detailed study showed that the original mineral in the slag had been galena, and which had subsequently oxidised.14) This conclusion is supported by the relatively low gold content of the contemporary Athenian drachma coins.6)
Meyers further postulated that although most Greek silver must have come from primary galena ores, elsewhere, through the Middle East the high gold content of the silver suggest that cerussite was the more common ore. However, there is another possible source- the legendary silver from Tartessos. From the beginning of the first millennium BC the Phoenicians crossed the Mediterranean and began to exploit the resources, particularly in Sardinia and the Iberia Peninsula. In their search for metals they found the vast mineral wealth of the Iberian Pyrites Belt that stretches for hundreds of km across Andalucia and into Portugal.15) In particular they discovered the jarosite ores, rich in silver. These form at the base of the weathered horizon at its junction of with the primary deposit where the more soluble metal salts that percolated from above will have precipitated and where early mining concentrated (Fig. 1).16)
Rio Tinto. East end of Corta Lago: Junction of the primary sulphidic pyrites (white) with the oxidised gossan (dark) above. Note the ancient gallery (circled) exposed in the enriched material just below the contact. (P. T. Craddock 1977).
They are thus a decomposition product of very variable composition and appearance, but are mainly the sulphates and oxides of iron, potassium, aluminium of general formula Fe3(OH)6.X(SO4), where X can be a number of metals including arsenic, antimony, copper, lead, bismuth and silver.17,2,p.216-21) As they are from an oxidised ore, the silver typically has a high gold content. Jarosites can be a soft clay or hard rock, with a wide colour range and are not particularly dense such that they do not have the appearance of a promising ore. As such they do not seem to have been exploited by the indigenous Late Bronze Age inhabitants, and the great achievement of the Phoenicians was to discover the jarosite’s potential and to develop ways of successfully smelting them.18) Many of the jarosites have a low lead content and thus lead had to be added to the smelting charge in order to collect the small amount of silver contained in each smelt. Sometimes the lead came from considerable distances as evidenced later in Roman times at Rio Tinto.19) There was a later legend that the Phoenician sailors had so much silver to carry back that they had to throw away their lead anchor stocks and replace them with silver. Could this be a fanciful reference to the Phoenicians bringing in lead from across the seas to smelt the jarosites?
In addition other complex argentiferous lead ores were smelted often adding barites and quartz as fluxes. This produces a mobile slag through which the argentiferous lead could drain easily. This seems straightforward but at many sites some of the slags contain many quite macroscopic fragments of the crushed fluxes that have appearance of having been added very shortly before the slag solidified (Fig. 2), and are accordingly referred to as free silica slags.20) These slags also routinely retain much more argentiferous lead. Thus it would seem that for some reason the slags were solidified by adding the crushed flux which would both stiffen and cool the melt, prior to being removed from the furnace whilst the process was still in progress. The smelting site at Monte Romero, near to Rio Tinto, in Huelva, in the south of Spain, dated to the 7th century BC has been excavated and studied.21,22) There some of the free silica slags were in the form of balls or buns, typically 20–25 cm in diameter, and had been seemingly stored for further processing, although here and at other sites most of the free silica slags were abandoned on the slag heaps with no further processing intended. A possible explanation could be that after the majority of the argentiferous lead had drained through the slags more crushed flux was added to halt the process. Some of the free silica slag could be crushed and assayed to determine the amount of argentiferous lead remaining. Where sufficient was present the slag would be worked into balls and stored for further processing, otherwise the slag was just dumped on the heaps. The resmelting of the free silica slags produced free-flowing tap slags. The silver would have been extracted from the argentiferous lead by cupellation. At Monte Romero a stack of used cupellation vessels, now mainly composed of lead oxide and silicates were found, presumably intended for resmelting to recover the lead.
Section through a typical free silica slag from Cerro de la Tres Aguilas, Spain. Note the large quantity of quite macroscopic quartz fragments. Although they are cracked due to sudden heating, their profiles, especially at the edges are still quite sharp showing that the iron and barium oxides in surrounding molten slag had little or no time to react and begin to dissolve the quartz before the slag had set. (cf the Dariba slags Fig. 13). (P. T. Craddock).
The introduction of coinage from the mid first millennium BC created an enormous demand for silver across the Old World from Spain to India. This resulted in major developments in all aspects of mining technology.23) Indeed many of the most famous mines of antiquity, Rio Tinto, Laurion etc. achieved their maximum production being worked for silver in the later first millennium BC. This is also true in India where the introduction of the silver punch-marked coinage in the mid first millennium BC24,25) created an enormous demand for silver. The sources of the metal have hitherto been obscure with many claiming that silver was not produced on any scale within ancient India.26) However, fieldwork in the Aravalli Hills of Rajasthan in north west India has shown that the Mauryan state developed several truly enormous mining enterprises, both for zinc minerals at Zawar27) and for silver at Dariba and Agucha10,28) (Fig. 3).
Ancient mines in the Aravalli Hills of North West India, A Agucha, D Dariba and Z Zawar, together with B, the ancient port of Bharuch or Broach. (T. Simpson).
The Aravalli Hills are formed of Precambrian rocks tilted almost vertically. At Dariba and Agucha the metalliferous ore is contained in the hard calc silicate rocks and the adjacent and much softer graphite mica schists. The latter were much easier to mine but their complex mineralogy with mica, sillimanite, potassium feldspars etc. made them difficult to smelt (see below). The primary ore minerals are of mixed iron, zinc and lead sulphides, of which the lead is argentiferous at Dariba and Agucha.29) The near-surface deposits at both mines are extensively gossanised and it is likely the argentiferous lead ore would have been principally the carbonate, cerussite, PbCO3 and the sulphate, anglesite, PbSO4. At Agucha the complex antomonide ore, freibergite, ([Cu, Ag, Fe]12 [Sb, As]4 S13) was also a significant source of silver.
At Dariba a series of small near-surface mines survive, principally in the graphite mica schist (Figs. 4, 5 & 6) dug both for their silver and to access the ore in the calc silicate deposits below, and there is evidence that similar workings were also once present at Agucha. These are likely to have been the earliest workings and mining continued at depth following the ore bodies down. This created a series of massive near vertical workings, known as stopes, penetrating down for hundreds of metres, well below the water table (Fig. 7). The graphite schist and surrounding country rocks are pervious which created a major drainage problem. As the workings at both mines are well below the surrounding plain there was no possibility to drive a drainage adit from the workings out to an adjoining river valley as the Romans had dug at Rio Tinto. Instead the water had to be raised to the surface and a series of bailing ponds were encountered at the top of one of the stopes at Dariba (Fig. 8), rather similar to systems recorded in Japanese mines in the 19th century.30)
Plan and section of Dariba. (HZL).
Looking north along North Lode (photo position at arrow A on Fig. 4). Note the quartzite chert ridge with workings in the graphite mica schist to either side (see Figs. 6 & 7). In the foreground is a later opencast in the calc-silicate which has gone through earlier galleries now exposed on the left hand side. (P. T. Craddock).
Mining in the graphite mica schist alongside the upstanding quartz dykes at North Lode. (P. T. Craddock).
Dariba South Lode: Near vertical stope typical of the large scale workings in these mines. The arrow indicates the bailing pool (Fig. 8). (L. Willies).
Small bailing pond and timber dam at the top of the large stope (Fig. 7). The wooden ladder on the left is original and leads up to the next pool. Note the large beam across the stope, probably used for hauling water up to this level. (L. Willies).
At Agucha the recent open cast mining of the Mauryan mine has exposed a complex series of workings beneath the ancient opencast (Fig. 9), the sophisticated layout of shafts and cross cuts being partly determined by considerations of ventilation and drainage of the workings at depth as well as access to the ore (Fig. 10).
The ancient open cast mine at Agucha before modern mining commenced. A complex system of deep mines lay beneath (Fig. 10). (P. T. Craddock).
Agucha: Conjectured arrangement of the west end of the deep workings, based on drilling and development at the modern mine (the squares are at 10 m vertical and 100 m horizontal intervals). Note the paired shafts, probably for ventilation purposes. (HZL).
When the richest ore had been extracted by deep mining major opencast mining was undertaken at both mines. The opencast pit over the East Lode at Dariba must surely be the largest such mine working to survive from antiquity (Figs. 4 & 11). At this mine there was an additional problem, because of the steep hillside on the west side, all the waste had to be dumped on the other side which was of unstable alluvium. This had to be supported and thus a vast timber complex of benches was constructed (Fig. 12). The benches run along one side for several hundred metres and the benches have been identified descending at four levels and further levels may continue down beneath the present fill of the opencast.
East Lode opencast seen from the top of South Lode (photo position at arrow B on Fig. 4). with the exposed section of the timber revetment (Fig. 12) circled. This is certainly the largest opencast metal mine of antiquity. Note the enormous quantities of waste dumped on the alluvium on the far side that had to be supported by the revetment. (L. Willies).
(a) Excavated section of the East Lode timber revetment showing part of one of the lifts. Note the old ladder stiles reused as backing behind the vertical timbers. (P. T. Craddock). (b) Isometric drawing of the excavated section of the revetment. (B. R. Craddock).
The mined ore would have been beneficiated by crushing, hand picking and washing. At Dariba a series of mortars survive cut into the hard calc silicate rock surrounded by enormous heaps of bean-sized waste, and similar heaps exist at Agucha.
The excavations at both mines failed to discover any remains of intact smelting units, but from the very many fragments of refractory ceramics in the slag heaps it has been possible to conjecture their form. There were many thick, crude curved pieces that could have come from hemispherical shapes of approximately 30 cm diameter. There were vitrified and slagged on their concave, i.e. their inner surfaces. These could either have come from a bowl furnace or a smelting hearth. On some fragments which included the edge, it was clear that the slag flow was away from the edge, suggesting that the ceramic had been set in ground and thus was a hearth lining, similar to those used in Japan until the 19th century.30)
The smelting slags at both mines were very heterogeneous, containing many fragments of unreacted gangue minerals such as mica, and clearly must have been very viscous (Fig. 13), unlike the contemporary copper and iron smelting slags at Dariba which were homogenous and clearly had been fully mobile. The viscosity and high melting point is due to the aluminium content of the gangue materials in the graphite mica schists (see above and Fig. 14). However the survival of the gangue fragments suggests either that they were added late in the process, recalling the free silica slags from southern Spain, or alternatively a relatively short reaction time at high temperature. The latter is reinforced by the studies on the vitrification of the smelting hearth fragments which suggest temperatures in the region of about 1150°C maintained for only a little over an hour, a very short smelting time. Despite this the slags contain relatively few blebs of argentiferous lead, certainly when compared to the Phoenician free silica slags discussed above, indicating a good separation of slag and metal, with very little silver being lost to the slags. It is possible that the slags were mechanically worked whilst still semi-molten to squeeze out much of the lead.
SEM photomicrograph of slag 32274 from Dariba Site 7, showing large relict potassium feldspar (dark) with Zn–Fe–S and PbS inclusions (bright spots) surrounded by dendritic olivine and hyalophane (grey tones) (cf the free silica slag Fig. 2).
Plot of Al2O3 against FeO for the Dariba and Agucha slags. The Dariba copper slags are post-Medieval and their aluminium contents are lower than those in the earlier silver lead slags as they tend to have come from the calc silicates rather than the graphite schists.
The next stage was the extraction of the silver from the lead by the process of cupellation. At Agucha some pits, also of the third century BC, were excavated that contained a great deal of cupellation debris. This included many very small cupels that must have been used to assay each batch of ore to discover the silver content. There were also fragments of furnace lining that had been attacked by lead oxide causing very extensive glazing together with long runs and drips of lead silicate that enabled the orientation of each fragment to be ascertained. Studies on these refractories showed that they had been at temperatures of about 1100°C for many hours suggesting an industrial-scale continuous process under ceramic hoods, similar to those envisaged by Conophagos3,12) at the contemporary silver mines at Laurion.
Argentiferous copper was also a significant source of silver in the past. The silver could be extracted by the process of liquation, in which the molten copper was mixed with lead, whereupon the silver transferred to the lead from which it could be recovered by cupellation.2,p.232) Liquation was used in Medieval Europe and in the Far East, but the origins of the process are uncertain. Rovira and Renzi31) believe that the process was used by the Tartessians early in the first millennium BC, and it has been suggested that the process could have been used somewhat earlier in the Late Bronze Age on Sardinia.32) The Romans certainly utilised liquation,2,p.233) but it does not seem to have been used in the Far east until after European contact.33) The Romans were also developing methods to recover silver lost in earlier processes, for example it is likely that the old Greek slags and tailings were reworked at Laurion.14) At Rio Tinto there is evidence for the processing of other smelting debris to recover the silver. When smelting ores that are rich in arsenic and antimony there is a strong possibility that a viscous layer of iron arsenides and antimonides will form in the furnace, known as speiss. This is a problem in itself but unfortunately speiss absorbs the forming silver and thus there is a considerable potential loss. Fragments of speiss occur frequently in the slag heaps at Rio Tinto and usually contain several thousand ppm of silver, but in one area well away from the main heaps, large amounts of speiss are found which is almost silver free, together with slags that are very rich in arsenic. From another part of the site a large partially vitrified crucible was found which had a few percent of lead and large quantities of arsenic. This was reported as evidence for the treatment of the speiss to recover the silver.34) Subsequently it was suggested that the crucible was in fact a cupel,35) but reanalysis of the original sample confirms the relatively low lead content and the high arsenic, incompatible with use as a cupel.15)
Man’s quest for silver from the earliest times has involved new technologies and great endeavours. From the development of sophisticated chemical treatments to the establishment of huge mining enterprises in remote locations all demonstrate the determination to produce the maximum amount of silver possible.
