1This paper focuses on the archaeo-metallurgical debris from Corta Lago, Riotinto, South-West Spain.
The site is located within a mining district known and exploited since the Late Bronze Age. This ore-rich district is the result of a regional tectonic event that led to the formation of the Iberian Pyrite Belt and characterizes the whole western part of the Sierra Morena to the border with Portugal and part of Portugal itself.
2The deposit of the pyrite belt is characterized by sulphides ( dominant pyrite, sphalerite, galena and chalcopyrite, generally with accessory tetrahedrite-tennantite, cassiterite and pyrrothite, and numerous trace minerals including electrum – Leistel et al., 1998, 13). The deposit comprises different layers which from the bottom to the top are: sulphide stockworks and massive sulphides, showing the same parageneses mentioned above, and gossan. These three geological layers have been formed at the same time so they show the same isotopic signature (Marcoux & Saez, 1994; Leistel et al., 1998; Marcoux, 1998). Between the massive sulphides and the gossan, the phenomena of oxidation and water infiltration formed a secondary enrichment layer (Davies, 1935, Rothenberg and Blanco-Freijero, 1981; Salkield, 1987; Barriga, 1990; Kassianidou, 1992; Hunt Ortiz, 2003). This enrichment layer is the one mainly exploited in antiquity for the production of silver and copper. The layer is characterized by a complex hydro-sulphate called jarosite or more accurately jarosites: XFe3(SO4)2(OH)6 where X can be K, Na, NH4, Ag and Pb. Depending on the main cation accompanying the iron, the jarosite has a different dominant colour, useful in Antiquity to recognise the vein to be exploited.
3Modern quarrying in the Riotinto area during the last century led to the discovery of mines and sections of ancient slag heaps. More than thirty years ago, the Huelva Archaeo-Metallurgical Project, coordinated by Prof. Beno Rothenberg, carried out a survey in the area and discovered different sites from different historical periods. The site of Corta Lago is, stratigraphically, one of the most complete, covering from the LBA to the 2nd century AD. Several seasons cleaned and uncovered a section of metallurgical debris several metres deep, truncated at the top and covered by a 2nd century Roman wall, so we do not have evidence of later activity (Fig. 1).
Figure 1: layers of tapping separated by a thin magnetite layer.
Figure 1 : Niveaux of coulées séparées l’un de l’autre avec un fin niveau de magnétite.
4Two layers of the section were radiocarbon dated; layer 50 dated to the 1st century BC and the layer 92 dated to the 1st century AD. Stylistic analyses of the ceramic allow interpolation of these two dates thus dating the whole section:
layers 91 to 99 Imperial;
layers 67 to 92 Republican tapped slags period II;
layers 50 to 66 Republican plate slags period;
layers 114 to 119 and 1 to 50 Republican tapped slags period I;
layers 108 to 113 Iberians;
layers 100 to 107 Phoenician.
5It should be noted that layer 101 to 119 are physically below layer 1, while layers 1 to 100 follow upon each other.
6The production of silver from jarosite involves a completely different chemical system compared to those we are used to when silver is associated to primary or secondary lead minerals such as galena or cerussite (Laurion – Conophagos, 1980, Rehren and Prange, 1998).
7In this case the main cation is iron and the system is iron-silicate dominated, while in other cases the iron compounds if present, may be associated to lead silicate.
8It is common knowledge that lead is a silver collector, and from a bullion of silver rich lead, through the cupellation we obtain silver.
9In the iron-dominated jarosite system however there is not sufficient lead to create a lead (silver) bullion during the smelting, hence and excess of lead must be added in order to ensure the complete collection of the silver.
10One possible source of lead is the plumbojarosite, part of the jarositic earths beneficiated for the production of silver, so the primary mineral. However the richness of plumbojarosite in the veins varies widely between 3 and 32% (Hunt Ortiz, 2003). This high variance of lead concentration in the mineralization could explain the discovery of archaeological silver free lead ingots in the area. The question of the origin of this extra lead is a key part of our understanding of Roman period silver extraction and some possible interpretations will be discussed later on in this paper.
11This study is focused on the reconstruction of the processes and the comparison between the four different technological periods recognised in the section: pre-Roman, Republican tapped slag (I and II), Republican plate slag, and Imperial tapped slag. The reconstruction of the processes is through chemical-mineralogical analyses of the slags. Bulk analyses is used to reconstruct the charge and the temperature of the process and analyses of the new-formed phases to used to reconstruct other parameters of the process, such as fugacity of oxygen, liquidity of the system, speed and method of cooling.
12The Bronze Age layers are characterized by so-called free-silica or ball slags, which are found also in other sites in the same region, for instance Monte Romero (Kassianidou, 1992). These slags are characterized by the presence of residual crystals of quartz and barite, already visible at the macroscopic scale. The main mineral is a feldspar-like mineral with the approximate formula (K, Ba, Fe, Ca, Pb)Al2Si2O8, found associated with pyroxene (Ca(Mg,Fe)Si2O6), cerussite and lead oxide. The crystals are fine in size which suggests a fast cooling, but the structure of the crystals (from hopper to polyhedral) suggest a slow cooling (Donaldson, 1976, Ettler, 2009). The explanation is possibly a low quantity of liquid in the cavities between the rigid structure of the solid crystals at the moment of crystallisation. The temperature of the liquidus can be calculated at 1100°C. Quartz and barite residual crystal show low indication of chemical reaction with the melt, and the author agrees with Tylecote hypothesis that gangue “sand” was added to thicken the slag “like a sponge with water” in order to help its removal (Tylecote 1987, 306-307).
13The Phoenician layers are characterised by crushed slags. These slags are mainly constituted by olivine, with H-chain morphology, which suggests a fast cooling. Crystallized magnetite is also present, disseminated in the slag and with a skeletal structure, indicating crystallization in the oxidizing condition when the amount of residual liquid was very low, since the length of the olivine suggests a begging of crystallisation already inside the furnace. The glass phase corresponds to a pyroxene-like composition. Residual aggregates of quartz were detected. Poly-metallic prills were detected (Pb, Sn, Cu, Zn, Ag, Sb) suggesting the use of jarosite mixed with poly-metallic sulphides as raw material. The system was not completely liquid, as indicated by the residual (non-reacted) quartz. The association of a glass phase with pyroxenitic composition, the skeletal (H-chain morphology) of the olivine and the residual quartz seem to suggest a rapid cooling. The temperature of the smelting can be estimated at 1100°C.
14The Iberians layers are predominantly characterised by crushed slags. The mineralogical association is dominated by fayalite with very fine H-shaped habit, indicating fast crystallisation and corroborating the indication given by chemical uniformity resulting from area analyses, even if residual barite is detected, suggesting then an excess of this component. The bulk composition suggests a smelting temperature of 1100°C.
Figure 2: Layers of olivines crystallized from two different smelting events.
Figure 2 : Niveaux d’olivines cristallisées de deux différents événements de fusion.
15The Republican period can be further divided in two different phases: the tapped slag and the plate slag phase.
The first phase is the tapped slag: the samples are characterized by the typical lava-like flow slags, and the tapping hole size can be estimated at c. 1 cm diameter. The slags were tapped one on top of the other and the different layers are recognisable because of an oxidation layer of magnetite, indicating a short length of time between the different tapping (Fig. 2).
16Olivine is the main mineral and shows a skeletal structure and a spinifex texture, representing a very fast cooling. The system was completely liquid with no excess of any of the components, and the temperature of the smelting can be estimated at 1100°C.
17The second phase is characterized by plate slags. These slags derive from a process of tapping in a fore-hearth and then “peeling” away the slag layer by layer from there. Olivine is still the main mineral, associated in this case with a feldspathoid-like mineral, with the formula (K, Fe, Ba)AlSi2O6. The olivines are skeletal, but their size is increased compared to the first phase and there is no spinifex texture visible, so the cooling rate has decreased relative to the first phase. The top and bottom surfaces are characterised by dendritic magnetite indicating the formation of top and bottom “crusts” within which densely liquid material was more slowly crystallising. The system was completely liquid, no residual components are detected and the temperature of smelting is calculated at 1100°C. The samples from this phase are the ones with lowest concentration of silver.
18The Imperial phase is also a tapping phase, but in this case we observe an increase in the lava-flow like structures typical of the tapping. This can be linked with the size of the tapping hole diameter which can be estimated at 2 cm (while it could be estimated around 1 cm in the previous phases). The main mineral is olivine but in this case two families of olivines are detected, a first crystallised phase has hopper structure indicating a slower cooling rate (Donaldson, 1976), starting in the furnace, and a second family with skeletal small size habit which crystallised within the residual liquid at the moment of the tapping. This observation is parallel to the observation of high variance in area analyses, since the first crystallisation would impede the movement of the chemicals in such a semi-rigid system. More layers of slag were tapped one on top of the other, indicated by the fact that the spinifex structure of the olivine at the surface of every tapping event can be seen inside the analysed samples. In this case there is no presence of oxidation layers between the different tapping events (Fig. 3) indicating a rapid succession. This too is a completely liquid system with no component in excess and with a smelting temperature of around 1100°C.
Figure 3: Corta Lago Section, drawn by Brenda Craddock.
Figure 3: Section stratigraphique de Corta Lago (Brenda Craddock).
19The chemical composition of the material found in the different periods is fairly homogenous: the major oxides have the average of: SiO2 20-30%Fe2O3 50-60%, CaO 0.1-3%, BaO 1-4%, Al2O3 3-6% corresponding to the field of stability of olivine in the diagram SiO2, Fe2O3, Al2O3, which is the minimum temperature field. The raw material is always the jarosite and its quartz/barite gangue mixed with poly-metallic sulphides (association typical of the enrichment zone of the massive sulphide in the Iberian Pyrite Belt). During the Roman phases it is observed complete understanding of the proportions between these two materials indicated by the complete liquidity of the system, while during the pre-Roman periods the proportion was not completely standardised, as shown by the presence of residual gangue material in the slags. Improvement in the loss of silver between the Bronze Age and the Phoenician and Roman periods is noticeable since it decreases by 1 to 2 orders of magnitude.
20The temperature of the process is around 1100°C (Geller et al., 1934 and 1943).
During the Roman period we observe some technological improvements.
The first macroscopic change is the plate slags period. At this time there is a transition from free tapping to fore-hearth tapping. The capability of recovering silver is improved, and this change may be due to a geological constraint (ore poorer in lead, richer in silver).
21During the two tapping periods we identify two technological changes. The first one is the increase in size of the tapping hole from around 1cm to 2 cm and this is possibly connected with an increase in the furnace size. The second change is the absence of the brief temporal gap between successive tapping during the Imperial period, a gap that is visible during the first Republican period. Both changes indicate an increased rate of production in the later period.
22Three ingots have been analysed during this project, two of them are lead ingots with small calcium impurity, while the third one is a copper iron sulphide ingot (Cu ≈ 40%; Fe ≈ 30%; S ≈ 30%).
Thirty three slag samples, nine semi-reacted ores and 3 ingots have been analysed by MC ICP-MS to trace their isotopic signature in order to give an insight into the origin of the extra lead.
Figure 4: Diagram of the Isotope results for the slags, the ingots and the ores.
Figure 4 : Diagramme des isotopes du plomb dans les scories, lingots et minerai.
23The results show a mixing line of isotopic signature (Fig. 4). Comparing the slag isotopic data with a preliminary selection of ore isotopic data, one extreme of the mixing line coincides with the isotopic signature of the massive sulphide of the pyrite belt (Fig. 4, Gale & Stos-Gale, 1989), while the other extreme (where most of the Phoenician and the Republican plate slags are plotted), looks convincingly similar to the Cartagena isotopic signature. This hypothesis of an added component of Carthagenean lead is also supported by the archaeological evidence of two ingots with the stamp “Carthago” recorded in the literature (Domergue 1987, 1990). The isotopic signature of the two lead ingots confirms the mixing line Cartagena/Rio Tinto mines, and the matte ingot confirms the Cartagena isotopic signature. The two ingots purified in silver through cupellation represent the stock of lead recycled in the system. The matte ingot represents a waste possibly arrived from Cartagena with the load of lead ingots, and recognised and discarded by the smelters in Rio Tinto. This result suggests two possibilities: during these two periods either a vein particularly poor in plumbojarosite was excavated or the ancient smelters were adding extra lead to be on the safe side for a complete collection of silver. In order to distinguish between these possibilities we studied the crystallographic structure of the jarosite. The organisation of the ions Fe, O, S and H is such that big cavities are left to be occupied by the other cations (K, Na, NH4, and Pb, Ag occupies a different lattice position). Considering the ratio Na2O/PbO and K2O/PbO (Fig 5a and 5b), for those slags for which the addition of lead metal is proved by the isotopic signature we observe that the Republican plate slags are relatively more alkali enriched. The relative enrichment in alkali can be read as an effective decrease of plumbojarosite in the vein, there is no evidence for the alkali to be the results of a bigger use of fuel. This lack of plumbojarosite had to be compensated by the addition of lead metal from elsewhere. Such a fine understanding of the ore would indicate a routine assaying of the ore for both lead and silver.
Figure 5a: Diagram PbO/Na2O for the slag samples.
Figure 5a : Diagramme PbO/ Na2O dans les scories.
24We do not observe any difference in the ratio alkali/PbO for the Phoenician slags. The homogeneity of the Phoenician slags with the other slags in this respect but the difference in their isotopic signature can be explained by considering that during the Phoenician period extra lead was possibly added despite a sufficient richness in plumbojarosite of the vein. Unfortunately the analysis of the ratio Ag/Pb does not give information about the richness in silver of those veins poorer in plumbojarosite.
Figure 5b: Diagram PbO/K2O of the slag samples.
Figure 5b :Diagramme PbO/ Na2O dans les scories.
25The metallurgical processes during the periods considered have the same raw material (jarosite, poly-metallic sulphides, quartz and barite) plus a varying amount of extra lead metal. The temperature is estimated to around 1100°C.
26The main technological change reconstructed between the pre-Roman and Roman periods is the standardisation of the process and the understanding of the correct proportion ore/gangue and the firing time. The complete liquidity of the system achieved in this way resulted in a decrease in the loss of silver. The main difference between the Phoenician and the Roman periods is an increased understanding of the ore, possibly by regular assay, and the import of lead metal only when the plumbojarosite in the ore was not sufficient.
27Within the Roman period the production increased in scale and the diameter of the tapping hole doubled reflecting a probable increase of the size of the furnace. There was also continuity of production with no detectable temporal gap between successive tappings in the later period.
28The preliminary isotope data do not appear to give a decisive distinction in the use of extra lead in different periods, even though it appears that during different periods the source of extra lead was always the same one, possibly Cartagena. The main difference (not completely coincident with the different phases) is the amount of extra lead metal used, due possibly to the varying richness in plumbojarosite of the exploited vein. Considering that most of the data coming from the Phoenician and the Republican plate slag period are at the far end of the mixing line, we could argue that in these two periods more extra lead metal was used. If we consider the Phoenician slag samples we do not notice any relative enrichment in alkali that would suggest a deficit of plumbojarosite in the vein. From this we can deduce that Phoenician smelters were adding extra/imported lead metal even when it was not required.
29Considering the skills and the metallurgical knowledge identified in the first Republican period, the lead isotope signature found for the Republican plate slags may be related to veins poorer in plumbojarosite and needing a larger addition of lead metal from elsewhere; an hypothesis confirmed by the relative enrichment in alkali of these slags.
30The puzzling period is what we called the Republican plate slags period. What was the need for this change in tapping process? Did the production slow down? Several reasons can be speculated: a more difficult mining situation that would request a larger capital investment; the need for more material coming from another region as indicated by the isotope results; a lack of capital to invest due to a political or socio-economical situation. The 1st century BC coincides with the political instability of the Roman Republic, followed by the stability brought by the settlement of the Empire. This period is also characterised by the presence of the procurator metallorum and a panel pf private investors (Avery, 1974), sharing the ownership and the exploitation of the mines. This situation changes during the Empire when the State intervenes in the ownership and the amount of capital increases.
Paul Craddock, Brenda Craddock, Phil Andrews and John Merkel helped a lot with their memories and documentation. Richard Harrison work on the archaeological stratigraphy and summary of all the excavation notes was immensely important for my project.
We are very grateful to Prof. Aurelio Perez Macias and Prof. Reinaldo Saez (University of Huelva) for the support, the company and the communications during my field trip to the Huelva Province, and to all the staff of the Museo Minero de Rio Tinto and in particular to Aquilino Delgado Dominguez for the help to select the material in the museum storage and the visits to the mine and the joyful company during my stay in Rio Tinto.
We thank Gilberto Artioli for the use of XRD at the Department of Mineralogy at the University of Milano and Monica Dapiaggi for all the advices for the use of the machine. The XRF, SEM and OM analyses were done at the Wolfson Archaeological Science Laboratories at the Institute of Archaeology UCL, with the important co-operation of Simon Groom (XRF laboratory supervisor) and Kevin Reeves (SEM laboratory supervisor). The MC-ICP-MS analyses were done at the Geology Department, Royal Holloway, University of London.