Archaeometallurgy in Europe IV [1 ed.] 8400102878, 9788400102876

Citation preview

Archaeometallurgy in Europe IV

IGNACIO MONTERO RUIZ ALICIA PEREA

ARCHAEOMETALLURGY IN EUROPE IV

BIBLIOTHECA PRAEHISTORICA HISPANA, XXXIII Directora Alicia Perea Caveda, Consejo Superior de Investigaciones Científicas Secretaria Inés Sastre Prats, Consejo Superior de Investigaciones Científicas Comité Editorial Xosé Lois Armada Pita, University College London Primitiva Bueno Ramírez, Universidad de Alcalá de Henares Carmen Cacho Quesada, Museo Arqueológico Nacional Pedro Díaz del Río Español, Consejo Superior de Investigaciones Científicas Teresa Chapa Brunet, Universidad Complutense Leonor Peña Chocarro, Consejo Superior de Investigaciones Científicas Juan Pereira Sieso, Universidad Castilla-La Mancha Consejo Asesor Concepción Blasco Bosqued, Universidad Autónoma de Madrid Francisco Burillo Mozota, Universidad de Zaragoza Felipe Criado Boado, Consejo Superior de Investigaciones Científicas Nuno Ferreira Bicho, Universidade do Algarve Antonio Gilman Guillén, Califormia State University-Northridge Susana González Reyero, Consejo Superior de Investigaciones Científicas Victorino Mayoral Herrera, Consejo Superior de Investigaciones Científicas Ignacio Montero Ruiz, Consejo Superior de Investigaciones Científicas Lourdes Prados Torreira, Universidad Autónoma de Madrid Federico Bernaldo de Quirós Guidotti, Universidad de León Gonzalo Ruiz Zapatero, Universidad Complutense Margarita Sánchez Romero, Universidad de Granada Assumpciò Vila Mitjà, Consejo Superior de Investigaciones Científicas

IGNACIO MONTERO RUIZ ALICIA PEREA (eds.)

ARCHAEOMETALLURGY IN EUROPE IV

CONSEJO SUPERIOR DE INVESTIGACIONES CIENTÍFICAS INSTITUTO DE HISTORIA Madrid, 2017

Reservados todos los derechos por la legislación en materia de Propiedad Intelectual. Ni la totalidad ni parte de este libro, incluido el diseño de la cubierta, puede reproducirse, almacenarse o transmitirse en manera alguna por medio ya sea electrónico, químico, óptico, informático, de grabación o de fotocopia, sin permiso previo por escrito de la editorial. Las noticias, los asertos y las opiniones contenidos en esta obra son de la exclusiva responsabilidad del autor o autores. La editorial, por su parte, solo se hace responsable del interés científico de sus publicaciones. Este volumen ha contado con la aportación económica de la organización del Congreso Archaeometallurgy in Europe IV, celebrado en Madrid entre los días 3-6 de junio de 2015 y organizado por el Instituto de Historia del CSIC en colaboración con el Instituto Arqueológico Alemán de Madrid, el CENIM-CSIC y el Museo Arqueológico Nacional.

Catálogo general de publicaciones oficiales: http://publicacionesoficiales.boe.es Editorial CSIC: http://editorial.csic.es (correo: [email protected])

© CSIC © Ignacio Montero Ruiz y Alicia Perea (eds.), y de cada texto, su autor © De las imágenes, los autores de los capítulos ISBN: 978-84-00-10287-6 e-ISBN: 978-84-00-10288-3 NIPO: 059-17-187-8 e-NIPO: 059-17-188-3 Depósito Legal. M-32.064-2017 Maquetación, impresión y encuadernación: Solana e hijos, A.G., S.A.U. Impreso en España. Printed in Spain En esta edición se ha utilizado papel ecológico sometido a un proceso de blanqueado ECF, cuya fibra procede de bosques gestionados de forma sostenible.

ÍNDICE ELEMENTAL AND PB ISOTOPIC ANALYSIS OF ARCHAEOLOGICAL METALS BY LASER ABLATION-Q/MC-ICP-MS: METHODS RESTRICTIONS AND APPLICATION EXAMPLES, Sonia García de Madinabeitia, María Eugenia Sánchez-Lorda, José Ignacio Gil Ibarguchi, José María Badillo Larrieta..................................................................................................................... 7 ARE PLATINUM AND PALLADIUM RELEVANT TRACERS FOR ANCIENT GOLD COINS? ARCHAEOMETALLURGICAL AND ARCHAEOMETRIC DATA TO STUDY AN ANTIQUE NUMISMATIC PROBLEM, Maryse Blet-Lemarquand, Sylvia Nieto-Pelletier, Florian Téreygeol, Arnaud Suspène...................................................................................................................................... 19 THE GREAT ORME BRONZE AGE COPPER MINE: LINKING ORES TO METALS BY DEVELOPING A GEOCHEMICALLY AND ISOTOPICALLY DEFINED MINE-BASED METAL GROUP METHODOLOGY, Robert Alan Williams.................................................................. 29 BRONZE AGE MINING IN SOUTHEAST SPAIN. NEW COPPER MINES FROM THE JANDULA AND YEGUAS VALLEYS, SIERRA MORENA, Luis Arboledas-Martínez, Charles Bashore, Eva Alarcón-García, Francisco Contreras-Cortés, Auxilio Moreno-Onorato, Juan Jesús Padilla-Fernández................................................................................................................................... 49 NEW DATA ON SCALE PRODUCTION OF COPPER IN THE CULTURE OF THE EL ARGAR. THE DUMPING GROUND FOR PEÑALOSA (BAÑOS DE LA ENCINA, JAÉN), Auxilio Moreno-Onorato, Salvador Rovira-Llorens, Francisco Contreras Cortés, Martina Renzi, Luis Arboledas-Martínez, Eva Alarcón-García, Adrián Mora-González, Alejandra García-García........ 65 SMALL SCALE BRONZE AGE METALLURGY: NEW DATA FROM SANTA LLÚCIA (ALCOSSEBRE, CASTELLÓN, SPAIN), Ignacio Montero-Ruiz, Mercedes Murillo-Barroso, Gustau Aguilella, Salvador Rovira....................................................................................................................................... 79 TIN PRODUCTION IN BRITTANY (FRANCE): A RICH AREA EXPLOITED SINCE BRONZE AGE, Cécile Le Carlier de Veslud, Céline Siepi, Christian Le Carlier de Veslud............................................ 91 ARCHAEOMETALLURGY OF TIN BRONZE IN THE DEH DUMEN BRONZE AGE GRAVEYARD, SOUTHWEST OF IRAN, Omid Oudbashi, Reza Naseri....................................................................... 105 COPPER PRODUCTION IN EARLY BRONZE AGE THASSOS: NEW FINDS AND EXPERIMENTAL SIMULATION IN THE CONTEXT OF CONTEMPORANEOUS AEGEAN METALLURGICAL PRACTICES, Nerantzis Nerantzis, Yannis Bassiakos, Myrto Georgakopoulou, Elena Filippaki, Georgios Mastrotheodoros...................................................................................................................... 115 COMPARATIVE STUDY OF SLAGS FROM TWO DIFFERENT COPPER SMELTING SITES IN THE SOUTHERN ‘ARABA VALLEY, ISRAEL, Shana Shilstein, Uzi Avner, Tal Kan-Cipor Meron, Sariel Shalev............................................................................................................................... 127 INVESTIGATIONS OF A SLAG FROM COPPER SMELTING DISCOVERED AT THE BRONZE AGE SITE PREIN VII/CU IN LOWER AUSTRIA, Roland Haubner, Susanne Strobl, Susanne Klemm....... 135

THE SMELTING FURNACES OF AYN SOUKHNA: THE EXCAVATIONS OF 2013, 2014 AND 2015, Georges Verly.......................................................................................................................................... 143 COMPOSITION OF BRONZE AGE GOLD BRACELETS FROM THE PORTUGUESE AREA, Isabel Tissot, Maria Filomena Guerra ................................................................................................... 159 METALLURGICAL AND TECHNOLOGICAL ASPECTS OF EARLY IRON AGE GOLD, Barbara Armbruster, Maryse Blet-Lemarquand, Bernard Gratuze, Verena Leusch, Ernst Pernicka, Birgit Schorer, Roland Schwab.......................................................................................................................... 169 ESTIMATING THE ECONOMIC AND ECOLOGICAL IMPLICATIONS OF AN IRON SMELTING SITE IN BEREKET, SW-TURKEY, Eekelers Kim, Scott Rebecca, Hodgins Gregory, Muchez Philippe, Poblome Jeroen, Degryse Patrick............................................................................................ 179 ARCHAEOMETALLURGICAL STUDIES OF IRON AGE WEAPONS FROM THE IBERIAN PENINSULA. A VISION IN PERSPECTIVE, Marc Gener Moret......................................................... 195 PUTTING SOME IRON BACK IN THE IRON AGE: A CASE STUDY FROM THE UK, Peter Halkon........................................................................................................................................... 205 THE EARLIEST IRON BLOOMERY IN SOUTHEASTERN NORWAY – TECHNOLOGICAL CONFORMITY AND VARIATION, Bernt Rundberget, Jan Henning Larsen..................................... 217 SUPPLY AND DEMAND: METAL RECYCLING IN SOUTHERN GERMANY AT THE END OF THE LATE IRON AGE, Roland Schwab.......................................................................................... 227 HANDHELD XRF MAPPING OF ELEMENTAL COMPOSITION OF ROMAN SILVER ARTEFACTS: PRELIMINARY RESULTS, Viktória Mozgai, Bernadett Bajnóczi, István Fórizs, Zoltán May, István G. Hatvani, Marianna Dági, Zsolt Mráv, Mária Tóth............................................................................. 237 TECHNOLOGICAL CHANGES IN MINING AND METALLURGY FROM ROMAN TO MEDIEVAL TIMES: EVIDENCE FROM A PB-AG (-CU) ORE DISTRICT IN CENTRAL KOSOVO, Katrin J. Westner, Guntram Gassmann, Sabine Klein, Gabriele Körlin............................................................ 249 ARCHAEOMETALLURGICAL EXAMINATION OF FINDS FROM MEDIEVAL BELL CASTING FOUNDRIES IN HUNGARY, Adrián Berta, Béla Török, Mária Tóth, Péter Barkóczy, Árpád Kovács, Krisztián Fintor.......................................................................................................................... 259 LITHARGE CAKES FROM CASTEL-MINIER (ARIÈGE, FRANCE): UNDERSTANDING STRATEGIES OF THE CUPELLATION IN A MULTI-METALS WORKSHOP FROM THE 14TH CENTURY, Julien Flament, Guillaume Sarah, Florian Téreygeol........................................................ 269 LOST-WAX CASTING PROCESS ANALYSIS OF THE LOST MUISCA SIECHA RAFT, Natalia Rueda Guerrero, Jairo Escobar Gutiérrez.......................................................................................................... 283

ELEMENTAL AND PB ISOTOPIC ANALYSIS OF ARCHAEOLOGICAL METALS BY LASER ABLATION-Q/MC-ICP-MS: METHODS RESTRICTIONS AND APPLICATION EXAMPLES Sonia García de Madinabeitia*, María Eugenia Sánchez-Lorda**, José Ignacio Gil Ibarguchi*,**, José María Badillo Larrieta***

Key words: Laser ablation, LA-Q-ICP-MS, LA-MC-ICP-MS, elemental analysis, Pb isotopes, wet ablation, Hg interference correction.

Abstract The use of laser-ablation (LA) systems coupled to ICP source instruments is becoming progressively common in archaeological studies as it allows for detailed minimally invasive analyses. We have set up improved microanalytical procedures using those techniques for the determination of element contents and Pb isotopic ratios in archaeological metal artefacts, including natural raw materials. A full range of trace elements in archaeological metal objects and ancient coins were analysed with good precision and accuracy using a «wet ablation» technique and a quadrupole ICP-MS instrument, allowing for reliable chemical discrimination among samples. Pb isotopic ratios were obtained on different types of metallic samples using a multicollector ICP-MS instrument. The isotopic results have also good precision and accuracy being reliable and meaningful in terms of geochemical and archaeological significance. Here we present the details of the implemented analytical methods that allow for element and Pb isotope chemical analyses of often highly valuable samples with minimum sample destruction.

Resumen El uso de sistemas de ablación láser (LA) acoplados a equipos con fuente de plasma (ICP) se está generalizando en los estudios arqueológicos ya que permite realizar análisis con un deterioro mínimo de la muestra. Utilizando dichas técnicas, hemos puesto a punto los procedimientos microanalíticos para la determinación de la concentración elemental y las relaciones isotópicas de Pb en muestras arqueológicas metálicas, incluyendo materias primas naturales. Así, utilizando la técnica de «ablación húmeda» y un ICP-MS cuadrupolar, se ha analizado con buena precisión y exactitud un gran número de elementos traza en objetos metálicos arqueológicos y monedas antiguas lo que permite su caracterización química detallada. Las relaciones isotópicas de Pb se han medido con un ICP-MS multicollector, obteniéndose también resultados con buena precisión y exactitud, coherentes con la información geológica y arqueológica. Se presentan los detalles de los métodos desarrollados para un estudio fiable de la composición elemental y de isótopos de Pb con una mínima destrucción en muestras que pueden ser de gran valor.

  *  Geochronology and Isotope Geochemistry FacilitySGIker.   ** Department of Mineralogy and Petrology. ***  Department of Geodynamics Faculty of Science and Technology, University of the Basque Country UPV/EHU.

Palabras clave: Ablación láser, LA-Q-ICP-MS, LA-MC-ICP-MS, análisis elemental, isótopos de Pb, ablación húmeda, corrección de la interferencia de Hg. 7

1. INTRODUCTION

mental and Pb isotopic analysis of archaeological metallic samples, including the determination of ratios involving 204Pb in samples with moderate Hg content.

From very early in the study of ancient metallic objects, one important goal was to establish the geological origin of the metal or metals used to make particular metal/alloy artefacts. This hopefully allowed for directly addressing issues of production, trade relationships and movement of objects in the past. Initially, the approach taken was through elemental chemical analysis of aliquots of the samples of interest. Yet, early elemental analyses failed often as an approach to determine the geological origin of the ores used to manufacture the metallic objects. Two research groups independently suggested that this objective might be better achieved instead by the comparison of lead isotope data from artefacts and metal ores (Brill and Wampler, 1965; Grögler et al., 1966). Systematic analysis and application of lead isotope data to provenance studies began in the 1970s. These analyses, usually as well as those of elemental concentration, were done through dissolution of the sample, which involved sample loss in some measure and, therefore, prevented the application of the techniques to particularly precious archaeological metallic remnants. Besides that, the Pb isotope analysis of dissolved samples, either by thermal ionization or plasma source mass spectrometry, is a highly time consuming technique that requires a careful isolation of Pb from the matrix by chromatographic procedures previous to the instrumental analysis (e.g., Desaulty et al., 2011; Ling et al, 2014), hence moreover restricting its generalized application. Recently, laser ablation (LA) techniques coupled to inductively coupled plasma mass spectrometry (ICP-MS) systems have proven to be particularly advantageous for this kind of fingerprinting or provenancing studies (e.g., Giussani et al., 2009; Resano et al., 2010; Nocete et al. 2014). In effect, LA-ICP-MS provides the detection power, wide linear dynamic range and capability for the analysis of small parts of the sample demanded by this type of application, while offering multi-element information with minimum or no sample preparation, minimal sample damage and a high sample throughput. In this work, the methods implemented at the Geochronology and Isotope GeochemistrySGIker facility of the University of the Basque Country UPV/EHU (Spain) for he analysis of major and trace elements as well as Pb isotopes in archaeological metallic objects using quadrupole (Q) and high-resolution multicollector (MC) LAICP-MS techniques, respectively, are presented. The optimization of techniques previously set up at our facility, in particular those for the trace and U-Pb isotope analysis of rock forming minerals, has allowed us to reliable perform the ele-

2. METHODS 1. Elemental analysis by LA-Q-ICP-MS Elemental analyses are done at the SGIker facility by coupling a laser ablation system to a quadrupole-based ICP-MS instrument (LAQ-ICP-MS). The method employed has been adapted from that proposed by Kovacs et al. (2009) to the equipment available: a New Wave UP213 Nd:YAG solid state laser ablation system and a Thermo Fisher Scientific XSeries-2 Q-ICP-MS instrument with enhanced sensitivity through a dual pumping system. The analyses are done in «wet ablation» mode and the instrument operating parameters and conditions of the laser ablation used are listed in Table 1. Previous to the collection of ablation data a preablation is conducted to remove surface contamination. Typical spots of 100 µm (nominal size), with repetition rates of 10 Hz and laser fluence of ca. 5 J/cm2 allow to quantify the major and trace elements. The ablated material is carried into helium and then mixed with argon and, before injection into the plasma source, it is blended with nebulizer-generated aerosol in a Peltier refrigerated spray chamber with dual inlet of ESI (Elemental Scientific, USA). This procedure, known as «wet ablation» allows the introduction of calibration solutions of known concentration for the quantification of the elements of interest. The tuning and mass calibration of the instrument are performed using the NIST SRM 612 reference glass and, before every analytical session, the signal of 238U is checked and the ThO+/ Th+ ratio is minimized to ca. 1.5 %. Using a spot size of 100 mm, the mean sensitivity on 238U at the instrumental conditions of Table 1 is about 1.400.000 cps/mg g-1. Each single analysis is done in Time Resolved Acquisition (TRA) mode by acquiring signals during 90 s, the first 30 s with the laser shutter closed in order to measure background contributions of the gases and the following 60 s with the laser firing to collect the signals corresponding to the sample. A 30 s delay between samples is used to wash out the system and prevent cross contaminations (ca. 10 s for 6 orders of magnitude). During the whole run a 2 % hydrochloric acid blank is introduced into the plasma to maintain stable conditions. The measured isotopes were: 24Mg, 47Ti, 51V, 52Cr, 55 Mn, 56Fe, 59Co, 60Ni, 65Cu, 66Zn, 75As, 82Se, 105Pd, 107 Ag, 111Cd, 118Sn, 121Sb, 125Te, 195Pt, 197Au, 208Pb and 209Bi, that is, including all the usual elements 8

Table 1. Operating conditions and data acquisition parameters used in the elemental analyses by LA-Q-ICP-MS

In every analytical method, to quantify lead isotope ratios by mass spectrometry there are two key issues that must be solved adequately: (i) the isotopic fractionation during the analysis, which refers to those processes that affect the relative abundance of isotopes and result in measured ratios different from the true ones in the samples analysed; and (ii) the interference of 204Hg on 204Pb, that concerns particularly the analyses by laser ablation methods as explained below. These two issues, if not properly solved, can make impossible the reliable determination of 206Pb/204Pb, 207Pb/204Pb and 208Pb/204Pb ratios. Two main analytical protocols are usually employed to correct for the isotopic fractionation in the analysis of Pb isotopes by LA-MC-ICP-MS: (i) through the mixing of ablated sample with Tl standard in liquid solution (e.g., Walder et al. 1993, Ponting et al., 2003, Resano et al., 2010); and (ii) by the use of the sample-standard bracketing method (e.g., Paul et al., 2005, Standish et al., 2013). These two methods have been tested in our study in order to evaluate their performance and select the best one for obtaining reliable results and long-term stability in routine analysis. The configuration of the MC-ICP-MS instrument is the same for both methods, with the only difference in the sample introduction system since, in the case of measurements with addition

in recent archaeological publications. Data treatment was done by means of the Plasmalab software of the XSeries-2 ICP-MS assuming that the analysis included all the elements present in the sample. The data treatment procedure basically involves: (i) the selection of integration periods for background and sample signal to discard blank contribution; (ii) the establishment of calibration curves by relating blank-corrected signals and known concentrations of the calibration standards; (iii) the assignment of a concentration to each blank-corrected sample signal; and (iv) the adjustment of concentrations to a total recovery condition. This method has been successfully applied to the analysis of archaeological gold samples in a previous work (Nocete et al., 2014) and to other sample types including archaeological bronze and silver (see below), and metal ores. 2. Pb Isotope analysis by MC-ICP-MS Pb isotopic analyses at the SGIker facility are done by coupling the same laser system (New Wave UP213 Nd:YAG) to a magnetic sector Thermo Fisher Scientific Neptune multicollector ICP-MS equipped with 9 Faraday detectors (LA-MC-ICP-MS). 9

Table 2. Operating conditions and data acquisition parameters for the isotopic analyses by LA–MC-ICP-MS

ured 205Tl/203Tl ratio is compared to the certified value of 2.38714 ± 0.001 (May and Waters, 2004). The standard solution is continuously introduced into the plasma through a desolvating Apex IR inlet system (Elemental Scientific, USA) with a 50 µL min-1 PFA micronebulizer, a silica glass spray chamber heated by infra-red radiation and a Peltier effect-cooled condenser. The aerosol formed is mixed with the ablated sample material by means of a Y-connector before the ICP torch. The results are calculated using the software of the ICP for interference and isotopic fractionation corrections. For the second method, where standard bracketing is applied to correct the fractionation, the ablated sample is mixed with Ar and N2 to stabilise the plasma and increase the sensitivity, and the NIST SRM610 glass standard (Jochum et

of Tl standard, it is necessary to inject liquid to the sample, hence «wet ablation» is done, while for the sample-standard bracketing method only dry ablated material arrives to the plasma. The MC-ICP-MS is set up to acquire simultaneously the signals of masses 202Hg, 203Tl, 204 (Pb+Hg), 205 Tl, 206Pb, 207Pb and 208Pb using the operating parameters summarized in Table 2. Tuning of the instrument is performed before every analytical session using the NIST SRM 612 reference glass, the ion lenses are tuned for maximum sensitivity and optimal peak shape at each mass of interest. For the method of Tl standard addition to correct for the isotopic fractionation in all Pb ratios (cf. Walder et al., 1993), a 200 ng mL -1 solution of thalium isotopic standard NIST SRM 997 is introduced in liquid form and the meas10

al., 2011) is analysed periodically between unknown samples to allow for the application of a sample-standard bracketing correction. The data obtained are reduced using the Iolite 3.0 software (Paton et al., 2011) and a specific Data Reduction Scheme (DRS) designed to correct for the fractionation after correction of the interference of 204Hg on 204Pb and the background contribution. The lack of certified reference materials for Pb isotopic analysis of archaeological metallic samples does not allow us to test these two methods on the same matrix than the samples of interest, neither, obviously to perform matrix-matching experiments. Hence, both methods have been applied to the analysis of NIST SRM612 (silica-rich glass) and USGS BCR-2G (silica-poor basaltic glass) reference materials as unknowns for evaluation purposes. The results obtained are represented in the Figure 1. The Pb isotopic ratios obtained by the two methods are in good agreement with those published for the analysed materials by different methods. As an example, in the Figure 1 are represented 20 spot analyses of BCR-2G and 40 of SRM612, measured with and without Tl standard addition, respectively. The mean results for BCR-2G measured with Tl standard added in solution are: 18.81 ± 0.13 for 206 Pb/204Pb; 0.83309 ± 0.00061 for 207Pb/206Pb and 2.061 ± 0.0017 for 208Pb/206Pb. That is, in the range of those published for this reference material: 18.765 ± 0.007 for 206Pb/204Pb; 0.833 for 207Pb/206Pb and 2.066 for 208Pb/206Pb (Georem Preferred Values). Also the results obtained via the sample-standard bracketing technique are acceptable, with mean values for SRM612 of: 17.08 ± 0.02 for 206Pb/204Pb; 0.90738 ± 0.00009 for 207Pb/206Pb and 2.1648 ± 0.0002 for 208 Pb/206Pb, again very similar to those obtained by other authors and methods as compiled by Georem: 17.095 ± 0.0002 for 206Pb/204Pb; 0.9073 ± 0.00029 for 207Pb/206Pb and 2.1647 ± 0.0008 for 208Pb/206Pb. There are, however, significant differences in the precision of the results obtained by each method (Fig. 1). This is easily visualized by the error bars for each point depicted in Figure 1, which are distinctly longer for the measurements with Tl standard addition than those obtained by means of the sample-standard bracketing technique. Overall, while the maximum value of standard error is near 1 % for an individual analysis using sample-standard bracketing, in the case of measurements by the Tl standard addition procedure the error for each spot is usually near 3 %. Thus, taking into account the results obtained on reference materials by these two methods, we advocate to discard the analytical routine including Tl standard addition for LA experiments.

Figure 1. Comparison of results obtained by the two methods tested for correction of the fractionation: Tl standard addition and sample-standard bracketing. Only one method for each sample is represented for a better view of the results. Error bars are 2SE of each individual analysis. Filled symbols correspond to certified values of reference materials (see text).

Once decided how the fractionation problem will be solved, there is another important question to address in order to obtain adequate results for Pb isotopic ratios, particularly for those that include 204Pb. That is, how to correct for the contribution of 204Hg to the 204Pb measured mass. Usually, Hg is a cause of concern when performing Pb isotope analyses by laser ablation on any type of sample, be it archaeological or not, because mercury is present as a trace impurity in the gases used in ICP-MS. That said, in the case of archaeological metallic samples this problem will be enhanced by the eventual presence of gold in the samples, since Au amalgamates easily with Hg. To establish the limits of a suitable mathematical correction for the 204 Hg interference, we prepared and analysed a solution of NIST SRM 981, pure lead isotope standard reference material, doped with different concentrations of Hg. The results obtained for the 206Pb/204Pb ratio increase with the content of Hg in the dissolution, although this value keeps in good agreement with the certified value of 16.9416 ± 0.0025 (Gao et al., 2010) in solutions with [Hg]:[Pb] ratios up to 1:5 (Fig. 2). This would imply that in samples with mercury contents higher than five times the content of lead the results would be under-corrected. This limit to the obtention of reliable results should be taken into account particularly in the case of analy11

ter the XVI century. Four different ancient silver coins from a private coin collection were thus selected to test our methods. The silver coins correspond to various geographical areas from the Mediterranean s.l. realm and from diverse historical periods (Fig. 3), which might reflect significant differences in their lead isotopic composition. Samples A and B in Figure 3 correspond to Roman coins, the first one previous to 144 B.C., a Republican denarius, and the second one is an Imperial coin of the Vespasian period minted ca. 70 A.D. Sample C in the same Figure 3 corresponds to a silver coin from the Parthian empire dated at ca. 120 A.D. Sample D comes from a more oriental location, the Indo-Scythian kingdom, dated approximately between 35 B.C. and 5 A.D. Although we lack information about the true location of the mints, the selection was done trying to ensure that quite different mining areas would be sampled.

ses of samples with moderate to high contents of Au, which may have even higher contents of Hg than Pb thus definitely preventing the obtention of proper results by this method.

Figure 2. Measured 206Pb/204Pb ratio of NBS SRM 981 solutions doped with different concentrations of Hg.

Results 3. APPLICATION OF THE METHODS TO REAL SAMPLES

Before assessing the results obtained on the analysed samples, it is worth mentioning that the physical effects of the laser ablation system on the coins were minimal, even immediately after the analyses, that is, without any action aimed to reduce the visibility of the sampling points (Fig. 3A and 3D). The holes generated are almost imperceptible to the naked eye having a maximum of ca. 100 µm in diameter and 60 µm in depth, and the minute black spots formed at the coin surface can be subsequently minimized by gently wiping with a soft rubber (Fig. 3B and 3C). On the other hand, and in order to test if the obtained results may be useful in archaeologial studies, we have selected some traditional graphs to represent our data. The experiment by LA-Q-ICPMS was designed to quantify on the same spot the 22 elements overall reported in archaeological metal studies: Mg, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Pd, Ag, Cd, Sn, Sb, Te, Pt, Au, Pb and Bi. However, in fact only 9 of these elements (Fe, Cu, Ag, Sn, Sb, Au, Pb and Bi) have been detected in at least one sample (Table 3), while the remaining elements were in concentrations below the method detection limit (here established as 3 times the standard deviation of all blanks collected for the samples, and then recalculated to solid sample by comparison with analyses of the unknowns). Ten different spots were analysed in each sample in order to evaluate the reproducibility of the method. The standard deviation of these 10 analyses is always below 15 % and, in most cases, it is below 5 % (Table 3).

Because of the lack of appropriate certified reference materials to validate the application of the methods set up, in this work a number of real samples were repeatedly analysed to evaluate a number of facts, like the effects of the laser ablation on the samples, the reliability of the results in terms of archaeological significance and the long term reproducibility of the method. Samples For the Pb isotope analysis, archaeological bronzes should not be a difficult test since they often contain Pb in considerable amounts (up to more than 10 wt % PbO) while the amount of Hg is restricted. Furthermore, Pb isotopic analysis in bronzes using MC-ICP-MS has been recently addressed with good results by Chen and coworkers (Chen et al., 2013), albeit by means of a much more expensive femtosecond laser ablation equipment. Archaeological gold would be on the other hand a particularly difficult, if not impossible, test because of the facility of this element to amalgamate with Hg. Some preliminary experiments revealed that to solve this problem we should develop a specific set up which is out of the scope of the present study. For that reason, we opted for an a priori moderately difficult sample type for the test purposes, in this case represented by ancient silver, that is, silver preceding the customary blending of this metal following its trade from deposits in America af12

Figure 3. Ancient silver coins analysed by the method proposed here. (A) Roman Republic denarius, red circles are sampled areas without any post-laser ablation cleaning; (B) Roman Imperial coin of Vespasian showing laser ablation traces after cleaning with a soft rubber; (C) Parthian coin showing laser ablation spots after cleaning with a soft rubber; (D) Indo/Scythian coin showing original black spots in sampled areas without further post-ablation cleaning.

the chemical composition of the sample after a quick and minimal invasive analysis, and attests to the potential of the method through the use of more specific diagrams like ternary plots, normalized multielemental diagrams or statistical methods. The same silver coins were analysed for Pb isotopic composition by LA-MC-ICP-MS to evaluate the proposed analytical routine in archaeological studies. Pb isotopic ratios obtained are displayed in Figure 5 and presented numerically in Table 4. The Figure 5 shows how isotopic signatures of different coins are neatly distin-

The elemental analyses revealed that, although we had considered that all the samples corresponded to silver coins, the amount of Ag was in fact highly variable, with silver contents ranging between 34.24 and 94.38 %. Besides that, the contents in some trace elements were markedly different among samples allowing to establish groups based on their chemical composition. Simple binary diagrams like those in Figure 4 show obvious differences among the samples studied, not only in the content of Ag but also in impurities like Fe or Au. The results obtained allow therefore the user to distinguish 13

Table 3. Results of elemental analyses of ancient coins obtained by LA-Q-ICP-MS

14

Figure 4. Binary diagrams for the coins analysed by LA-Q-ICP-MS. See text for details.

guished each other and are comparable to those reported for Mediterranean ores and materials (c.f. OXALID database). In Figure 5, 10 analyses of each coin are represented and, as may be observed in the graph, it is not possible to distinguish the corresponding 10 points. This attests to the low deviation among spot results in the

same sample and to the good reproducibility of the analytical method. It may be observed that the data obtained for the 206Pb/204Pb, 207Pb/206Pb and 208Pb/206Pb ratios for the coins analysed are consistent with those of Mediterranean ores in all relations considered, even those that include 204 Pb values. 15

Figure 5. Lead isotope ratios determined by LA-MC-ICP-MS on ancient silver coins compared with data by solution analysis of Mediterranean ores and archaeological artefacts (c.f. OXALID database). Each symbol in the graph corresponds to 10 analysis of the same coin including analytical errors for each data.

Conclusions

63530-P) and the Universidad del País Vasco (Grupo Consolidado project GIU12/05) is acknowledged. Technical and human support provided by SGIker (UPV/EHU, MINECO, GV/EJ, ERDF and ESF) is also gratefully acknowledged.

The determination of the elemental and Pb isotopic composition of archaeological metallic objects can be achieved with reasonably good precision and accuracy using LA-Q-ICP-MS and LA-MC-ICP-MS methods. The small diameter of the laser ablation spots leaves just minor traces on the archaeological targets. The damage being barely visible to the naked eye, the method may be thus considered a minimally invasive one (Fig. 3). The results obtained by the methods set up at the SGIker facility are repetitive and meaningful from a geochemical/archaeological point of view. The elemental and isotopic data obtained by the methods presented are regarded as useful in the investigation of archaeological metallic samples, in particular as regards issues like identification and origin of ores used for the manufacture of objects, routes of trade and particularities of smelting or refining process.

BIBLIOGRAPHY BRILL, R. H. and WAMPLER, J. M. 1965: «Isotope studies of ancient lead». American Journal of Archaeology 69:165-166. CHEN, K.; FAN, C.; YUAN, H.; BAO, Z.; ZONG, C., DAI, M.; LING, X. and YANG. Y. 2013. Analysis of the Lead Isotopic Composition in Copper Using Femtosecond Laser Ablation MC-ICP-MS and the Application in Ancient Coins. Spectroscopy and Spectral Analysis 33(5): 1342-1349.

Acknoledgements

DESAULTY, A. M.; TELOUK, P.; ALBALAT, P. and ALBARÈDE, F. 2011: «Isotopic Ag–Cu–Pb record of silver circulation through 16th–18th century Spain» Proceedings of the National Academy of Sciences or the United States of America 108 (22): 8947-9316.

Financial support by the Spanish Ministerio de Ciencia e Innovación (grant CGL2015-

GAO, Y.; YANG, Z.; HOU, Z.; WEI, R.; MENG, X. and TIAN, S. 2010: «Eocene potassic and ultrapotas-

16

sic volcanism in south Tibet: New constraints on mantle source characteristics and geodynamic processes«. Lithos 117: 20-32.

MAY, W. E. and WATTERS Jr. R. L. 2004: «NIST [2004] Certificate of Analysis SRM997, Isotopic Standard for Thallium» http://www.nist.gov/srm/index3column.cfm

GIUSSANI, B.; MONTICELLI, D. and RAMPAZZI, L. 2009. «Role of laser ablation-inductively coupled plasma-mass spectrometry in cultural heritage research: A review». Analytica Chimica Acta 635: 6-21.

OXALID http://oxalid.arch.ox.ac.uk PAUL, B., WOODHEAD, J. D. and HERGT, J. 2005:» Improved in situ isotope analysis of low-Pb materials usingLA-MC-ICP-MS with parallel ion counter and Faraday detection». Journal of Analytical Atomic Spectrometry 20: 1350-1357

GRÖGLER, N.; GEISS, J.; GRÜNENFELDER, M. and HOUTERMANS, F. G. 1966: «Isotopenuntersuchungen zur Bestimmung der Herkunft römischer Bleirohre und Bleibarren». Zeitschrift für Naturforschung 21a: 1167-1172.

PATON, C.; HELLSTROM, J.; PAUL, B.; WOODHEAD, J. and HERGT, J. 2011: «Iolite: freeware for the visualisation and processing of mass spectrometric data». Journal of Analytical Atomic Spectrometry 26: 2508-2518.

JOCHUM, K. P.; WEIS, U.; STOLL, B.; KUZMIN, D.; YANG, Q.; RACZEK, I.; JABOB, D. E.; STRACKE, A.; BIRBAUM, K.; FRICK, D. A.; GÜNTHER, D. and ENZWEILER, J. 2011: «Determination of reference values for NIST SRM 610-617 glasses following ISO guide- lines». Geostandards Geoanalytical Research 35: 397-429.

PONTING, M.; EVANS J. A. and PASHLEY V. 2003: «Fingerprinting of Roman mints using lead isotope analysis». Archaeometry 45, 4: 591-597.

KOVACS, K.; SCHLOSSER, S.; STAUB, S. P.; SCHMIDERER, A.; PERNICKA, E. and GÜNTHER, D. 2009: «Characterization of calibration materials for trace element analysis and fingerprint studies of gold using LA-ICP-MS». Journal of Analytical Atomic Spectrometry 24: 476-483.

RESANO, M; MARZO, M. P.; ALLOZA, R.; SAÉNZ, C.; VANHAECKE, F.; YANG, L.; WILLIE S. and STURGEON, E. R. 2010: «Laser ablation single-collector inductively coupled plasma mass spectrometry for lead isotopic analysis to investigate evolution of the Bilbilis mint». Analytica Chimica Acta 677: 5563.

LING, J.; STOS-GALE, Z.; GRANDIN, L.; BILLSTRÖM, K.; HJÄRTHNER-HOLDAR, E. and PER-OLOF PERSSON, P-O. 2014: «Moving metals II: provenancing Scandinavian Bronze Age artefacts by lead isotope and elemental analyses». Journal of Archaeological Science 41: 106-132.

STANDISH, C.; DHUIME, B.; CHAPMAN, R.; COATH, C.; HAWKESWORTH, C. and PIKE, A. 2013: «Solution and laser ablation MC-ICP-MS lead isotope analysis of gold». Journal of Analytical Atomic Spectrometry 28: 217-225.

NOCETE, F.; SÁEZ, R.; BAYONA, M. R.; NIETO, J. M.; PERAMO, A.; LÓPEZ, P.; GIL-IBARGUCHI, J. I.; INÁCIO, N.; GARCÍA, S. and J. RODRÍGUEZ. 2014: «Gold in the Southwest of the Iberian Peninsula during the 3rd Millennium BC». Journal of Archaeological Science 41: 691-704.

WALDER, A. J.; PLATZNER, I. and FREEDMAN, P. A. 1993: «Isotope ratio measurement of lead, neodymium and neodymium-samarium mixtures, hafnium and hafnium lutetium mixtures with a double focusing multiple collector inductively coupled plasma mass spectrometer». Journal of Analytical Atomic Spectrometry 8: 19-23.

17

ARE PLATINUM AND PALLADIUM RELEVANT TRACERS FOR ANCIENT GOLD COINS? ARCHAEOMETALLURGICAL AND ARCHAEOMETRIC DATA TO STUDY AN ANTIQUE NUMISMATIC PROBLEM Maryse Blet-Lemarquand*, Sylvia Nieto-Pelletier*, Florian Téreygeol**, Arnaud Suspène*

traza del oro (sobre todo el platino y el paladio) durante los procesos de refundición, copelación y cementación de este mismo metal. Las pruebas indican que el platino y el paladio constituyen unos indicadores seguros para la trazabilidad del oro antiguo. Por lo tanto, resulta pertinente utilizar dichos elementos para examinar una cuestión muy debatida en los estudios numismáticos: ¿refundió Julio César el oro de Galia (conquistada en el año 51 a.C.) para sus propias acuñaciones de oro, como lo sugieren las fuentes literarias? Para establecer un primer enfoque global en torno a este espinoso asunto, hemos analizado monedas célticas tardías y monedas de Julio César gracias al sistema LA-ICP-MS, con en fin de comparar detalladamente sus contenidos en platino y en paladio.

Abstract Archaeometallurgical experiments were carried out at the platform of Melle (France) to study how the trace elements of gold, in particular platinum and palladium behave when gold is melted, cupelled and then cemented. Our tests proved that platinum and palladium are reliable tracers of ancient gold. They can therefore shed light on the provenance of the gold coined by Caesar in Rome around 46-44 BC. Did he indeed melt down the Celtic gold gained from his conquest of Gaul in 51 BC as suggested by textual sources? To give a first insight into this tricky problem LA-ICP-MS was performed on some Late Celtic coins and some of Caesar’s coins in order to compare especially their platinum and palladium contents, and in fact our first results support this hypothesis.

Palabras clave: oro; elementos traza; platino; paladio; cementación; copelación; LA-ICP-MS; monedas célticas tardías; César

Key Words: gold; trace elements; platinum; palladium; cementation; cupellation; LA-ICPMS; Late Celtic coins; Caesar.

INTRODUCTION Platinum and palladium are impurities of ancient gold that are usually drawn on for provenance studies (Blet-Lemarquand et al. 2014a). For instance, they can be used to distinguish between different stocks of precious metals if their contents scaled to the gold concentrations are significantly different. At the same time, they can also help to confirm that gold coins were melted down to manufacture other coins if the contents of both groups of coins are consistent. Their role as gold tracers is based on the assumption that platinum and palladium do not separate from

Resumen Experimentos arqueometalúrgicos realizados en la plataforma de Melle (Francia) han permitido estudiar cómo reaccionan los elementos

   * IRAMAT-CEB, UMR5060, CNRS / Université d’Orléans, France.    ** L APA-IRAMAT, NIMBE, CNRS, Université Paris-Saclay91191 Gif-sur-Yvette France.

19

ARCHAEOMETALLURGICAL EXPERIMENTS AND LABORATORY ANALYSES OF THE EXPERIMENTAL ARTEFACTS

gold in the course of the ancient metallurgical treatments such as smelting, melting, cupellation or cementation, because these elements are highly unreactive and have high melting points. In fact, there has not been any systematic study to determine how the trace elements of gold behave during metallurgical processes but studies are available that reports experiments on the melting of gold ores or metal gold spiked with some trace elements in oxidising or reducing atmospheres (Raub 1995; Hauptmann et al. 1995; see a brief report in Ehser 2011). Nor has there been any study exploring how trace elements behave during cupellation of gold based alloys. The behaviour of certain elements, how ever, may be generalised from the different works carried out on the cupellation of silver alloys (McKerrell and Stevenson 1972; Pernicka and Bachmann 1983; L’Héritier et al. 2015). Unfortunately, no platinum elements were examined in these articles. Cementation experiments typically focus on the efficiency of the process for major elements and neglect the minor and trace elements (Wunderlich et al. 2014; Geçkinli et al. 2000; review in Craddock 2000c). The present paper presents new analytical and experimental data for the role of platinum and palladium as valuable gold tracers. In addition, we explore the question of provenance of the gold minted by Caesar after he conquered Celtic Gaul in the middle of the first century BC.1 Written sources report that he seized large quantities of precious metals during his campaigns. The validity of these sources can be tested by comparing the composition of Caesar’s gold coins with that of the Late Celtic Gaul gold coins. As Late Celtic gold coins are made of gold based ternary alloys, Romans would have had to melt down and purify the Celtic alloys in order to mint their high purity gold coinage. This is why our archaeometallurgical experiments included several steps, especially cupellation and cementation, in order to replicate the chaîne opératoire that the Romans would have probably followed. Some first results of these archaeometallurgical experiments were published in 2014 in a paper which lays the foundations for the present study (Blet-Lemarquand et al. 2014b).

The archaeometallurgical treatments were conceived to resemble the melting down of Late Celtic gold coins and the refining of their gold.2 The first step was to manufacture a gold-silvercopper alloy having the same major element composition as the gold coins struck by the Celtic tribe of the Arverni in the middle of the first century BC (Tab. 4) and that contains substantial amounts of platinum and palladium. The ternary alloy was then cupelled in order to eliminate copper and other base metals from the alloy. The gold-silver alloy obtained after cupellation was laminated and finally cemented to separate gold from silver.3 Cementation alone should be sufficient to part gold from silver and copper. For this, the gold based alloy needs to be beaten into thin foils to expose the maximum surface area. As it is easier to beat a gold-silver alloy than a gold based ternary alloy a two stage process consisting of cupellation and then cementation was preferred. A 19th-century French gold coin4 was selected for the experiments because previous LA-ICPMS analysis showed it contained about 1,400 ppm platinum and 340 ppm palladium (Tab. 1).5 This 6.40 g coin was melted down with 3.12 g of pure silver and 2.07 g of pure copper to manufacture a ternary alloy of about 50 % gold, 30 % silver and 20 % copper (Fig. 1A). The first purification process was cupellation that requires that lead is added in the right proportions to the ternary alloy. A block of lead coming from a medieval site was selected because previous LA-ICP-MS analysis showed that this lead would not contaminate our samples with platinum or palladium. For the gold-silver-copper alloy to be cupelled successfully, it requires about 10 times its weight of lead to be added to the melt as advised in published tables (Hervé, 1839: 26). Four buttons were cast that were then 2   Descriptions of the ancient gold refining methods can be found in Halleux 1985 and Craddock 2000b. 3   The amalgamation process was not taken into account because it is not attested for Antiquity (Craddock 2000a). It was used to separate gold from platinum in Colombia during the 18th century (Morrisson et al. 1999: 125). Hence this method certainly modifies the platinum and palladium fingerprints of gold. 4   Louis XVIII, 20 francs, 1816, bare head. 5   Information about the LA-ICP-MS analysis of ancient gold coins can be found in Dussubieux and van Zelst 2004; for the depth profile mode developed for this method see Gratuze et al. 2004, Blet-Lemarquand et al. 2009 and BletLemarquand et al. forthcoming (for gold coins) and Sarah et al. 2007 (for silver coins).

1   This study lies within the framework of researches devoted to the provenance of Augustean gold (Suspène et al. 2011; Blet-Lemarquand et al. 2015) and of the earliest gold coins struck in Gaul (current researches led by S. NietoPelletier).

20

Figure 1. The different steps of the archaeometallurgical experiments (from Blet-Lemarquand et al. 2014). From left to right, A: manufacture of the gold-silver-copper alloy; B: gold-silver alloy button obtained after cupellation; C: laminated foils made of gold-silver alloy in the cementation pot before cementation takes place; D: foils after cementation.

put into cupels made from bone ashes. The cupellations were carried out in a muffle furnace under a blast of air with temperatures varying between 950 °C and 1000 °C and four white gold buttons were obtained (Fig. 1B). The process was interrupted in the course of one of the operations because the temperature had dropped. Lead was added to the button and cupellation resumed in a new cupel. The cementation experiments were inspired by Theophilus’ recipes dating back to the 12th century AD (Hawthorne and Smith 2014) as well as by the archaeological finds from Sardis. The excavations of the workshop of Sardis provide the first archaeological evidence of the cementation of gold no later than the middle of the 6th century BC (Ramage and Craddock 2000). The four white gold buttons obtained after cupellation were accordingly laminated to a 50 micrometre thickness.6 About 10 cycles of metal rolling and annealing were necessary to achieve this thickness. Common salt was used as active agent in the cement.7 Twice as much ground brick was added to play the role of inert carrier even though the use of brick is not certain in Antiquity.8 The gold foils were put alternatively with the cement in the parting vessel —a cooking pot— and the mixture was dampened with vinegar (Fig. 1C). The pot was covered with a lid, which was in fact a scorifier that had been turn over and drilled to add a temperature sensor. The process took place in an open furnace fed by charcoal that covered the cementation pot entirely and the temperature was regulated

through an adjustable electric blower. In a second cementation experiment the pot was sealed with a lute9 as recommended by Theophilus in his treatise (Hawthorne and Smith 2014: 109) in order to prevent the acid vapours from escaping. In fact, this lute broke in the course of the cementation. In both experiments the temperatures were maintained below the melting point of the gold-silver alloy foils of approximately 1035 °C. The first cementation lasted eight hours with temperatures maintained between 650 °C and 700 °C and the second one was lengthened to 12 hours and the temperatures were raised to 700 °C-800 °C. The cementation experiments kept the foils intact and enhanced their golden colour. The foils obtained from the first cementation had sometimes remains of cement adhering to their surface (Fig. 1D) whereas the ones coming from the second experiment were much cleaner. The gold based alloys were sampled at every step of the purification process in order to study their microstructure by SEM and their composition using LA-ICP-MS and SEM-EDX. The aims were to check that the metallurgical treatments were efficient and to investigate the trace element composition. The LA-ICP-MS analyses were performed using the time resolved mode that enables to reconstruct depth profiles from the surface to the interior of the object, the laser ablation being made at one spot (see for instance Fig. 4B showing a surface depletion in silver thus an enrichment in gold in a cemented foil). The calculation of the concentrations only considers the signals detected when the interior of the experimental sample is reached by the laser. The obtained results are summarized in Tab. 1.

6   The gold foils excavated in Sardis which were subjected to cementation were obviously hammered and not laminated but for the sake of convenience we preferred to laminate our samples. 7   A review of the different recipes for cements was established by Halleux (1983). 8   Craddock 2000d: 204.

  The lute was made of a mixture of clay and horse dung. 9

21

Table 1. Elemental composition of the coin and of the alloys obtained by archaeometallurgical experiments (LA-ICP-MS analysis). Each content is the average of the concentrations calculated for 3 to 6 micro sampling. Contents in weight % (Au, Ag, Cu) or weight ppm (other elements). AVG: average, SD: standard deviation

The cupellation proved to be efficient because copper was to a great extend eliminated from the alloy: its content was reduced from 19.4 % to about 1.6 % or even lowered to 0.1 % (Tab. 1) in the sample which was subjected to a two stage cupellation (see above).

(Zn, As, Sb, Sn, Pb, Fe) when the gold was melted down (comparison between gold coin and AuAg-Cu alloy on Fig. 2A) and these parameters were still lowered during the cupellation (comparison between Au-Ag-Cu alloy and cemented alloy on Fig. 2). In contrast, the Pt/Au and Pd/ Au ratios remained unchanged between the gold coin and the cupelled alloy (Fig. 2B). It can be noticed that the cupellation left some lead (0.2 to 0.5 %) in the alloy (Tab. 1). It can be concluded that most of the trace elements were separated from the gold during melting and cupellation whereas platinum and palladium remained with adhered to gold in the course of both operations. The cementation led to the partial elimination of silver as can be seen in several ways. The foils look more golden once they were cemented. At high magnification using a scanning electron microscope their surfaces reveal characteristics of cementation: porosities which result from the parting process (they extend along the grain boundaries and the triple points), and deposits of silver chloride formed from the attack of the silver (Fig. 3).10 The surface of the cemented samples is also depleted in silver compared to the foils that have not been cemented: a minimum of 8-10 % silver was obtained for the foils from the first cementation and the content of sil-

Figure 2. Behaviour of the trace elements during melting and cupellation.

The contents of the trace elements were divided by the gold concentration for each sample in order to determine how they behaved relative to gold (Fig. 2). The ratios of most of the trace elements were reduced by a factor of 10 to 100

  See Craddock 2000c: 181 for explanations on the chemistry of the salt cementation process. 10

22

other element are of less reliable when trying to trace the life cycle of gold coins.

ver is around 5 % for the sample from the second experiment (SEM-EDX analysis). However the cross-section of a foil led us to think that the first cementation did not consistently reach the core of the foil (Fig. 4A). In fact, the depth profile analysis of the cemented foils clearly established that the surface is enriched in gold while the composition of the core remained unchanged (Fig. 4B and Tab. 1). The second cementation was more efficient than the first in that a depletion of silver was achieved also inside the foil (decrease of 5 % see Tab. 1). It seems that the cementation had an effect on the composition of some minor elements in the interior of foils. Lead contents were lowered by a factor of ten and at the same time the binary alloys were enriched in some trace elements (Sb, As, Zn, Sn, Bi), probably as a result of contaminations from the cement (Tab. 1). The Pt/Au and Pd/Au ratios remained the same after the two cementation experiments (Tab. 1). However, we cannot strictly deduce from these analyses carried out inside the foils that cementation has no effect on these parameters because the enrichment of gold inside the foils was at best slight. But the depth profile analyses undemiably showed that the Pt/Au and Pd/Au ratios are constant from the cemented surface to the interior of the foil (Fig. 4b). Thus, cementation has no effect to these ratios.

PROVENANCE OF THE GOLD COINS MINTED BY CAESAR BETWEEN 46 AND 44 BC IN ROME During the Roman Republic, gold had only been coined in times of emergency (like the Hannibalic War or the First Civil War, Burnett 2004: 49). Caesar is actually the first Roman minting authority who regularly produced gold coins from 46-44 BC at his own mint located somewhere in Rome (Woytek 2003: 263-271). It is thought that Caesar melted down the huge quantities of gold that he had seized during his campaigns in Gaul (see Suetonius)11 and the precious metal he took from the state’s treasury in 49 BC. Pliny the Elder reported that Caesar looted in the aerarium not only 30 million sestertii or coins but also 15,000 ingots of gold and 30,000 ingots of silver (see Pliny).12 Several scholars have already discussed the impact of the arrival of great quantities of precious metals on the Roman economy (Castelin 1977, Nash 1987: 34 or van Heesch 2005). This paper contributes new information to this discussion in the form of analytical data of Ceasar’s gold coins in comparison with Late Celtic coins. In so doing, we do not claim to solve this mystery once and for all, but rather offer an important piece of new evidence about the economic ramifications of Caesar’s conquests. For this purpose, 15 gold coins (aurei) minted by Caesar and believed to be representative of Caesarian coins (particularly the Hirtius issue) were selected for analyses (Tab. 2 and Fig. 5). The composition of the gold circulating in Gaul at the time of Caesar’s campaign can be gleaned on the one hand from the composition of Late Celtic coins and on the other hand from Celtic gold objects. Unfortunately, the trace element compositions of the latter are still unknown. The advantage of coins over object is that they were manufactured in large quantities and that consequently the composition of one single specimen is expected to represent the mean composition of a set of them belonging to the same series. The Celtic coins selected for our study possibly date back to the 1st century BC (Aubin et al. 2011; Barrandon et al. 1994; Nieto-Pelletier et al. 2011) and are attributed to different tribes:

Figure 3. Cemented foil after the second experiment. SEM image obtained with back scattering electron detector. Surface of the Au-Ag alloy with remains of silver chloride.

Our archaeometallurgical experiments firmly established that platinum and palladium can be relevant tracers to track a gold bullion. Consequently, platinum and palladium contents can confirm whether or not ancient gold coins were melted down and used as bullions to strike new specimen. It should be noted that most of the

  Suetonius, Vie de Jules César, LIV, 2.   Pliny, Naturalis Historia, XXXIII, 17. Unfortunately, as pointed out by Crawford, «We have no idea how large theses bars of gold and silver taken from the aerarium were» (Crawford 1974: 639). 11 12

23

Figure 4. Cemented foils after the first experiment, left to right. 4A: Cross-section of 6.2.a observed using a binocular lens; 4B: Depth profile analysis of 6.1.b (LA-ICP-MS analysis).

Cenomani, Arverni and Osismii (Tab. 3 and Fig. 5). The concentrations of more than 50 coins analysed using LA-ICP-MS were gathered for our study. The protocol of analysis of the coins resembles that applied to the gold foils and uses the depth profile mode to determine the composition beyond the surface that may be enriched in gold as a result of silver and copper depletion13 and that may be «contaminated» in other elements. The laser diameter was lowered to 80 micrometres so as the craters left on the coin can not be seen with the naked eyes. Three different micro-samplings were carried out on each coin and the calculated contents are the results of the means.

contrast, the Caesar’s aurei were struck from purified gold that contains about 99.4 % gold (Tab. 4). If Caesar has indeed reused the highly debased Late Celtic coins to mint his coinage, purification processes must have necessarily been carried out to eliminate copper and silver probably by cupellation and cementation. One could wonder if it was worthwhile to purify such highly debased gold. The parting vessels discovered in Roman archaeological contexts in Britain give weight to the hypothesis of the recycling of the Iron Age precious metal by the Romans (Bayley 2009: 430).14 As we have demonstrated, it is relevant to compare Pt/Au and Pd/Au ratios to discuss the provenance of gold (see above). The contents of platinum and palladium are normalized to gold and plotted on a binary graph for the Late Celtic and the Caesar’s coins (Fig. 6). Most of the Late Celtic coins fall into an ellipse characterised by Pt/Au ratios ranging from 20 and 100, Pd/Au ratios between 8 and 40 and Pt/Pd ratios around 2.5. The two Arvernian gold coins have higher Pt and Pd contents in relation to Au but their Pt/Pd ratios are close to those of the other Late Celtic coins. It would be necessary to increase the analytical data on the Late Celtic coins especially the Arverni coins to confirm these trends.

Figure 5. Pictures of some of the coins analysed for this study. 1. Osismii; Caesar’s aurei: 2. RRC 466/1 (A. Hirtius); 3. RRC 475/1b (L. Plancus); 4 RRC 481/1 (COS QVINC). Sources: IRAMAT-CEB A. Arles and G. Sarah (Celtic coin) and gallica.bnf.fr (Roman coins).

14   For instance, J. Bayley identified parting crucibles among archaeological remains found in Chichester (Britain), a Roman settlement from immediately after the Claudian conquest of Britain in AD 43 (Bayley 2009). This site which hosted the 2nd Legion for up to two years is located near Selsey, an oppidum of the Atrebatic kingdom, where large quantities of coins and fragments of gold have been found (Bayley 2009: 430). Previous analysis established that these late Iron Age gold coins were made from ternary alloys containing around 40-50 % gold. Bayley concludes that «the late Iron Age precious metal was requisitioned in the period following the Claudian conquest and [...] (purified) so it could be subsumed into the precious metal pool the Romans had in circulation.» We are indebted to J. Bayley for directing our attention to these researches.

The three series of Late Celtic coins are made of different gold-silver-copper alloys with gold contents ranging from about 50 % to 20 % depending on the issuing authorities (Tab. 4). In 13   Three patterns of depth heterogeneities were obtained for the Celtic coins from the hoard of Laniscat, see Blet-Lemarquand et al. forthcoming.

24

Table 2. Types of the Caesar’s gold coins selected for LA-ICP-MS analysis. All these coins are kept at the Bibliothèque nationale de France (BnF), département des Monnaies, médailles et antiques. RRC = Crawford 1974. The dating of coins, the number of specimen known and of dies were quoted from Roman Republican Coinage (Crawford 1974: 93-4; 888)15

Table 3. Late Celtic gold coins selected for LA-ICP-MS analysis. These coins are kept at the BnF or at the Musée archéologique du Mans (hoard of Les Sablons) or at the Direction Régionale des Affaires Culturelles de Bretagne, Service Régional de l’Archéologie (hoard of Laniscat)

Table 4. Contents of gold, silver, and copper determined for the Late Celtic coins and for the Caesar’s coins. Average (AVG) and standard deviation (SD). For the results of late Celtic coins, see Aubin et al. 2011 (Cenomani); Nieto-Pelletier 2013 (Arverni) and Nieto-Pelletier et al. 2011 (Osismii). Caesar’s gold: unpublished results

15

used – from 45 onwards – other supplies than the gold recovered from the Celtic coins.16 In summary, the Hirtius issue, which represents the bulk of the gold minted by Caesar in Rome, appears to be consistent with the Late Celtic gold coins when the Pt and Pd contents are compared. At first sight, the hypothesis of recycling of the Late Celtic gold coinage by Caesar to mint his Roman coins cannot be ruled out however, as expected, other sources of gold were also used. The study is still ongoing and further analyses are planned in the framework of a next research program to complete a diachronic overview of the gold coined in the Western part of

The 15 Roman coins show various Pt and Pd fingerprints stretching from low values which match the Cenomani and Osismii gold compositions to higher contents which are more consistent with the trend of the two Arvernian coins. However, three Caesar’s coins seem to deviate from the Late Celtic gold because they have a lower Pt/Pd ratio of 1.3 (grey dotted line): they belong to the Plancus issue (RRC 475) or have COS QVINC on their reverse (RRC 481). This progression in the composition of the gold minted by Caesar is consistent with the dating of coins and could let us to think that Caesar also 15   The actual numbers of coins of these types are most likely more numerous than the numbers quoted in Crawford 1974 because other specimens were found since. One should also keep in mind that the assessment of the numbers of dies are very arbitrary and only indicate that the Hirtius issue was really very common whereas the Plancus and especially the Caesar issues were much less widespread.

  It is known from the literary sources that Caesar stole precious metal from the state’s treasury in 49 BC (see Pliny the Elder). But we do not know where this gold accumulated in the aerarium came from. Gold previously sent by Caesar to Rome during 58 and 49? More ancient Celtic gold? Gold coming from outside? This question remains open. 16

25

Late Celtic coins and at first sight their platinum and palladium chemical fingerprints partially match each other. The analytical results are consistent with the written sources. These are however preliminary results which need to be confirmed and qualified by further analyses especially on Late Celtic coins and on Celtic jewels within the framework of future research programs.

the Mediterranean basin between the end of the 4th century BC and the 1st century AD17. This data will give us arguments to hypothesize which other gold supplies could be available for Caesar. CONCLUSION Archaeometallurgical experiments provided firm evidence that the Pt/Au and Pd/Au ratios remain unchanged when the gold based alloy is melted down, cupelled and then cemented. We were not able to demonstrate that cementation keeps these ratios consistent, but given the moderate temperature of the cementation process we expect this to be the case as platinum and palladium do not react with the cement at such low temperatures. Hence, platinum and palladium can serve as relevant tracers to study the provenance of gold. All other trace elements of gold, in contrast, do not survive the melting and refining of gold. This property of platinum and palladium was applied to explore the provenance of the numerous gold coins minted by Caesar in Rome around 46-44 BC. The elemental composition of Caesar’s gold coins were accordingly compared to some

ACKNOWLEDGEMENTS All the archaeometallurgical experiments were conducted on the national experimental archaeology platform located in Melle (France) in the framework of a PCR «Paléométallurgies et experimentations. Recherches sur les chaines de production des métaux aux périodes anciennes» 2013-2015 directed by F. Téreygeol. We are grateful to F. Duyrat and to D. Hollard (Bibliothèque nationale de France, département des Monnaies, médailles et antiques), to Y. Menez (SRA Bretagne), to B. Mandy (SRA Pays de la Loire) and to M. Thauré (Musée archéologique du Mans), for allowing us to carry out analysis on their coinages. We thank Francesca Silenzi who made the second experiment of cementation.

Figure 6. Scatterplot of the ratios of palladium and platinum to gold for the Late Celtic coins compared with the Caesar’s coins (LA-ICP-MS analysis). Cenomani’s results: Aubin et al. 2011; Arverni’s and Osismii’s gold: unpublished results. Caesar’s gold: unpublished results18.

micro-sampling. In fact the relative standard deviation is around 30 % for the three ablations carried out on the Caesar’s coins. Despite these high fluctuations the Pt/Pd ratio remains steady for each coin. This has already been noticed for other antique gold coins (Blet-Lemarquand et al. forthcoming).

  APR IA 2016-7 «AUREUS. À la naissance du monnayage d’or romain : étude et caractérisation de l’or monnayé en Occident de la fin de la période hellénistique au premier siècle de notre ère» (directed by A. Suspène) founded by the Région Centre-Val de Loire. 18   The uncertainties on these ratios are mainly due to variations in the Pt and Pd contents at the scale of the 17

26

BIBLIOGRAPHY

CASTELIN, K. 1977: «L’or de la Gaule et César». Cahiers numismatiques 53: 62-68.

AUBIN, G.; BARRANDON, J.-N. and LAMBERT, C. 2011: «Le dépôt monétaire des Sablons, Le Mans (Sarthe) : 152 statères gaulois en or allié». In M. Amandry (ed.): Trésors d’or. Les Sablons (Le Mans), Lava (Corse), Partinico, Martigné-sur-Mayenne. Trésors monétaires XXIV. Paris: 1-90, pl. 1-9, A-B.

CRADDOCK, P. T. 2000a: «Appendix 3. Early History of the Amalgamation Process». In A. Ramage and P.T. Craddock (eds.): King Croesus’ gold: Excavations at Sardis and the History of Gold Refining. British Museum Press, Cambridge: 233-237.

BARRANDON, J.-N.; AUBIN, G.; BENUSIGLIO, J.; HIERNARD, J.; NONY, D. and SCHEERS, S. 1994: L’or gaulois. Le trésor de Chevanceaux et les monnayages de la façade atlantique, Cahiers Ernest-Babelon 6. CNRS Éditions. Paris.

— 2000b: «Chapter 2. Historical Survey of Gold Refining. 1 Surface Treatments and Refining worldwide, and in Europe prior to AD 1500». In A. Ramage and P.T. Craddock (eds.): King Croesus’ gold: Excavations at Sardis and the History of Gold Refining. British Museum Press, Cambridge: 27-53.

BAYLEY, J. 2009: «The Discovery of Precious Metal Refining in Roman Chichester». In J.-F. Moreau, R. Auger, J. Chabot and A. Herzog (eds.): Proceedings Actes ISA 2006: 36th International Symposium on Archaeometry, 36e Symposium International d’Archéométrie, 2-6 May 2006, Quebec City, Canada. Québec: 425-432.

— 2000c: «Chapter 8. Replication Experiments and the Chemistry of Gold Refining». In A. Ramage and P.T. Craddock (eds.): King Croesus’ gold: Excavations at Sardis and the History of Gold Refining. British Museum Press, Cambridge: 175-183. — 2000d: «Chapter 10. Reconstruction of the Salt Cementation Process at the Sardis Refinery». In A. Ramage and P.T. Craddock (eds.): King Croesus’ gold: Excavations at Sardis and the History of Gold Refining. British Museum Press, Cambridge: 200211.

BLET-LEMARQUAND, M.; NIETO-PELLETIER, S. and GRATUZE, B. Forthcoming: «Depth profile LAICP-MS analysis of antique gold coins». In K. Sheedy et al. (eds): Mines, Metals and Money in Attica and the Ancient World. Proceedings of the International conference at the Epigraphic and Numismatic Museum, Athens, April 20-22 2015.

— 2010: Early Metal Mining and Production. London. CRAWFORD, M. 1974: Roman Republican Coinage, Vol. I & II. London.

BLET-LEMARQUAND, M.; NIETO-PELLETIER, S. and SARAH, G. 2014a: «L’or et l’argent monnayés». In Ph. Dillmann and L. Bellot-Gurlet (eds.): Circulation et provenance des matériaux dans les sociétés anciennes. éditions des archives contemporaines. Paris: 133-159.

DUSSUBIEUX, L. and VAN ZELST, L., 2004: «LA-ICPMS analysis of platinum-group elements and other elements of interest in ancient gold». Applied Physics A 79: 353-356. EHSER, A.; BORG, G. and PERNICKA, E. 2011. «Provenance of the gold of the Early Bronze Age Nebra Sky Disk, central Germany: geochemical characterization of natural gold from Cornwall». European Journal of Mineralogy 23-6: 895-910.

BLET-LEMARQUAND, M.; NIETO-PELLETIER, S. and TÉREYGEOL, F. 2014b: «Tracer» l’or monnayé : le comportement des éléments traces de l’or au cours des opérations de refonte et d’affinage. Application à la numismatique antique», Bulletin de la Société Française de Numismatique 4: 90-95.

GEÇKINLI, A. E.; ÖZBAL, H.; CRADDOCK, P. T. and MEEKS, N. D. 2000: «Chapter 9. Examination of the Sardis Gold and the Replication Experiments». In A. Ramage and P.T. Craddock (eds.): King Croesus’ gold: Excavations at Sardis and the History of Gold Refining. British Museum Press, Cambridge: 184-199.

BLET-LEMARQUAND, M.; SARAH, G.; GRATUZE, B. and BARRANDON, J. N. 2009: «Nuclear methods and Laser Ablation Inductively Coupled Plasma Mass Spectrometry: how can these methods contribute to the study of ancient coinage?». Cercetări numismatice XV: 43-56.

GRATUZE, B., BLET-LEMARQUAND, M. & BARRANDON, J. N. 2004: «Caractérisation des alliages monétaires à base d’or». Bull. de la Société Française de Numismatique 6: 163-169.

BLET-LEMARQUAND, M.; SUSPÈNE, A., and AMANDRY, M. 2015: «Augustus’ gold coinage: investigating mints and provenance through trace element concentrations». In A. Hauptmann and D. ModaressiTehrani (eds): Archaeometallurgy in Europe III, Proceedings of the 3rd International Conference Deutsches Bergbau-Museum Bochum, June 29 - July 1 2011: 107-113. Bochum.

HALLEUX, R., 1985: «Méthodes d’essai et d’affinage des alliages aurifères dans l’Antiquité et au Moyen Age». In C. Morrisson, B. Brenot, J.-P. Callu, J.-N. Barrandon, J. Poirier and R. Halleux (eds.): L’or monnayé I: purification et altérations de Rome à Byzance, Cahiers Ernest-Babelon 2. CNRS Éditions. Paris: 39-77.

BURNETT, A. 2004: Coinage in the Roman world. First published 1987. London.

27

PERNICKA, E. and BACHMANN, H. G. 1983: «Archäometallurgische Untersuchungen zur antiken Silbergewinnung in Laurion. III. Das Verhalten einiger Spurenelemente beim Abtreiten des Bleis». Erzmetall 36 (12): 592-597.

HAUPTMANN, A.; REHREN, T. and PERNICKA E. 1995: «The composition of gold from the ancient mining district of Verespatak/Rosia Montana, Romania» In G. Morteani and J.P. Northover (eds.): Prehistoric gold in Europe: mines, metallurgy and manufacture. Proceedings of the NATO Advanced Research. Workshop on Prehistoric Gold in Europe, held at Seeon, Germany, from September 26 to October 1, 1993: 369384. Dordrecht.

RAMAGE, A., and CRADDOCK, P. T. 2000: King Croesus’ gold: Excavations at Sardis and the History of Gold Refining: Archaeological Exploration of Sardis. British Museum Press, Monograph 11. Cambridge.

HAWTHORNE, J. G. and SMITH, C. S. (trans. and ed.) 2014: Theophilus. On Divers Arts, New York.

RAUB, C. 1995: «The metallurgy of gold and silver in prehistoric times». In G. Morteani and J.P. Northover (eds.): Prehistoric gold in Europe: mines, metallurgy and manufacture. Proceedings of the NATO Advanced Research. Workshop on Prehistoric Gold in Europe, held at Seeon, Germany, from September 26 to October 1, 1993: Dordrecht 243-259.

HERVÉ, A. C., 1839: Nouveau manuel complet des alliages métalliques: Contenant la préparation de ces alliages : leurs principales propriétés ; leur emploi ; leur existence dans la nature ; leur analyse, etc., Paris. L’HÉRITIER, M.; BARON, S.; CASSAYRE, L. and TÉREYGEOL, F. 2015: «Bismuth behaviour during ancient processes of silver–lead production». Journal of Archaeological Science 57: 56-68.

SARAH, G.; GRATUZE, B. and BARRANDON, J. N. 2007: «Application of laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) for the investigation of ancient silver coins». Journal of Analytical Atomic Spectrometry 22(9): 1163-1167.

MCKERRELL, H. and STEVENSON, R. B. K. 1972: «Some analyses of Anglo-Saxon and associated Oriental coinage». In E.T. Hall and D.M. Metcalf (eds.): Methods of Chemical and Metallurgical Investigation of Ancient Coinage. Royal Numismatic Society, London: 195-210.

SUSPÈNE, A.; BLET-LEMARQUAND, M. and AMANDRY, M. 2011: «Les monnaies d’or d’Auguste  : l’apport des analyses élémentaires et le problème de l’atelier de Nîmes». In N. Holmes (ed): Proceedings of the XIV International Numismatic Congress (Glasgow 2009). London: 1073-1081.

MORRISSON, C.; BARRANDON J. N. and MORRISSON, C. 1999: Or du Brésil, monnaie et croissance en France au XVIIIe siècle, Cahiers Ernest-Babelon 7. CNRS Éditions. Paris.

VAN HEESCH, J. 2005: «Les Romains et la monnaie gauloise: laissez-faire, laisser-aller ?». In J. Metzler and D. Wigg-Wolf (eds.): Die Kelten und Rom: neue numismatische Forschungen. Studien zu Fundmünzen der Antike, Band 19, Philipp von Zabern. Mainz am Rhein: 229-245.

NASH, D. 1987: Coinage in the Celtic world. London. NIETO-PELLETIER, S. 2013: Catalogue des monnaies celtiques, CMC 1-Les Arvernes. BnF/MAN. Paris.

WOYTEK, B. 2003: Arma et Nummi. Forschungen zur römischen Finanzgeschichte und Münzprägung der Jahre 49 bis 42 v. Chr. Österreichischen Akademie der Wissenschaften. Vienne.

NIETO-PELLETIER, S.; GRATUZE, B. and AUBIN, G. 2011: «Le dépôt monétaire gaulois de Laniscat (Côtes-d’Armor) : 547 monnaies de bas titre. Étude préliminaire». In N. Holmes (ed): Proceedings of the XIV International Numismatic Congress (Glasgow 2009). London: 1217-1225.

28

THE GREAT ORME BRONZE AGE COPPER MINE: LINKING ORES TO METALS BY DEVELOPING A GEOCHEMICALLY AND ISOTOPICALLY DEFINED MINE-BASED METAL GROUP METHODOLOGY Robert Alan Williams*

Abstract

Resumen

Understanding the connectivity between Bronze Age communities involving the movement of people, materials and technological knowledge can be assisted by re-invigorating metal provenance studies. Completely rethinking our methodology by switching around from an artifact-based metal group approach to a new mine-based metal group approach based on the discovery of major Bronze Age mine sites in recent decades. This approach relies on using expertise from geological disciplines combined with the latest analytical techniques and are here applied to the Great Orme mine in north Wales which is one of the largest surviving Bronze Age copper mines in Europe. The results show this mine to be a major source of arsenic–nickel metal contrary to low impurity claims of the past literature. The evidence suggests that the ‘golden age’ of production at the Great Orme was around 1500 to 1400 BC in the early Middle Bronze Age (Acton Park), when it probably dominated the metal supply in Wales and Lowland Britain with some exchange to the near continent.

La comprensión de la conectividad de las comunidades de la Edad del Bronce implica el movimiento de personas, materiales y conocimiento tecnológico y en este sentido los renovados estudios sobre procedencia del metal pueden ser de gran ayuda. Es necesario repensar nuestra metodología cambiando el enfoque desde el grupo metálico basado en el objeto hacia un nuevo enfoque de grupo metálico basado en las minas a partir de la identificación de las principales minas de la Edad del Bronce realizado en las últimas décadas. Este enfoque conecta los conocimientos de disciplinas geológicas con las últimas técnicas analíticas y aquí se aplica al caso de la mina de Great Orme en el norte de Gales, que es una de las minas de cobre de la Edad del Bronce más grandes que sobreviven en Europa. Los resultados muestran que esta mina fue un recurso importante de metal con arsénico y níquel a pesar de que anteriormente se le ha considerado con nivel bajo de impurezas. Los datos sugieren que la «edad de oro» de la producción en el Great Orme Fue alrededor de 1500 a 1400 a.C. en la Edad del Bronce Medio temprano (Acton Park), cuando es probable que dominó el suministro de metal en el País de Gales y de las tierras bajas de Inglaterra con algunos intercambio con el área continental próxima.

Key words: Bronze Age, Britain, Wales, Great Orme, metal, copper, mines, provenance, minebased metal group, arsenic, nickel, Acton Park, exchange, connectivity, geology, geometallurgy.

Palabras clave: Edad del Bronce, Inglaterra, Gales, Great Orme, metal, cobre, minas, procedencia, grupo metálicos basado en minas, arsénico, níquel, Acton Park, intercambio, conectividad, geología, geometalurgia.

*  Department of Archaeology, Classics and Egyptology, University of Liverpool, 12-14 Abercromby Square, Liverpool L69 7WZ, United Kingdom.

29

METAL PROVENANCE STUDIES AND ARTEFACT-BASED METAL GROUPS

lished a classification independent of typology with a view to identifying ‘workshops’ (Pernicka 2014,242; Roberts 2014,424). They set up 12 and later 29 artefact-based metal groups using the principal impurities found in copper, namely arsenic, nickel, antimony, silver, bismuth, lead and cobalt. These groups were statistically based without proper geological and archeological foundations. The SAM programme and similar ones in other countries have been widely criticized (Tylecote 1970, see Pernicka 2014,244-247 for a review) but form a useful legacy database provided the accuracy limitations of the analytical method used are borne in mind. In the late 1970s in Britain, Peter Northover’s (1980) pioneering work involved analyzing numerous artefacts from the Bronze Age collection in the National Museum of Wales using electron microprobe analyses. He defined 14 artefactbased metal groups based on trace/minor element chemistry, which were later extended to the whole of the Britain (Northover 1991). Later Rohl and Needham (1998) defined 23 British artefact-based groups incorporating trace elements, lead isotopes and a typological/chronological dimension. The common thread of the various research groups was the setting up of artefact-based metal groups but there was a gradual realization that there were complications in the trace element chemistry approach alone. Firstly, there were some overlaps in the geochemistry between some mines and mining regions. Secondly, there was a split of the trace/minor elements in the ores between the copper metal and slag during smelting. These might vary depending the smelting conditions such as the redox and the temperature although a better understanding was gradually gained (Tylecote et al 1977, Pollard et al. 1991, Pernicka 2014, 252-253, Hauptmann 2007,27,204-207). Given these potential weaknesses from using a trace/minor element ‘signatures’ there was renewed optimism with the introduction of lead isotope measurements which came to prominence in the 1980s and 1990s (Pernicka 2014, 247). This offered, in principle, an isotopic ‘signature’ or ‘fingerprint’ that would, in theory, be specific to each ore deposit and would apparently be fully preserved during smelting (Pernicka 2014, 248, see also Pollard and Bray 2014,232). This proved very valuable in excluding deposits as sources of particular copper artefacts (e.g. Rudna Glava mine, see Pernicka 2014, 250256) even when the actual sources sometimes remained unknown. The whole subject ran into a confidence damaging controversy in the early 1990’s over various issues relating to possible fractionation, precision, overlapping ore deposits and the mixing of metal sources (see sum-

Increasing our understanding of the connectivity between Bronze Age communities involving the movement of people, materials and technological knowledge can potentially be assisted by re-invigorating metal provenance studies. A re-thinking of our methodology needs to be based on bringing in practical expertise from other disciplines. Part of this involves systematic ore sampling and using the latest and most appropriate scientific analytical techniques to provide representative and accurate results. Many archaeologists in the past have eagerly pursued the hope of being able to analyze Bronze Age metal artefacts and trace them to a mine or mining area. Pernicka (2014) and Pollard and Heron (1996) reviewed developments starting with the aspirations of researchers in Europe in the nineteenth century. In Britain, William Gowland (1906 and 1912), professor of metallurgy, at the Royal School of Mines sought origins of the metals and noted the presence of arsenic and silver in analyses of Irish copper axes. Advances in analytical techniques allowed a larger number of samples to be analyzed with pioneering work by Otto and Witter (1952) in Halle who analyzed 1300 mainly German artefacts but few ores. Also Preuschen and Pittioni (1937) in Vienna who analysed over 6000 artefacts and 2000 ores from Central Europe, unfortunately only recording the presence or absence of elements rather than quantification (Pernicka 2014,240-241). Both these projects proposed their own artifact-based groups with some regard to general ore types. The scientific approach of the Austrian work influenced British researchers (Davies 1935,v, Coghlan 1958,57). Research resumed after the second world war and in 1945 the Ancient Mining and Metallurgy Committee was set up in Britain chaired by Herbert Coghlan and involved many of the most prominent Bronze Age scholars of the period including Gordon Childe and Stuart Piggott. Eventually this group had, what has subsequently proved to be, a remarkable success in tentatively identifying the unusual fahlore copper ores from «the Cork-Kerry region» in SW Ireland as the probable source of the earliest Irish and British copper (Coghlan 1958, Coghlan and Case 1957, Coghlan et al. 1963,15). They also proposed three artefact-based metal groups. In the 1960s and the 1970’s there was the huge SAM project of Siegried Junghans and Edward Sangmeister based in Stuttgart and Freiburg using optical emission spectroscopy, which eventually analyzed about 22,000 metal artefacts across Europe (Junghans et al. 1968 &1974). They were unable to match artefacts to mines or mining areas and so instead estab30

mary by Tite 1996, Pollard and Bray 2014,229). However, most of these criticisms were addressed successfully and the technique has re-emerged with more accurate measurements and a clearer acknowledgement of the technique’s limitations (Pernicka 2014,263). As with the trace/minor element studies, as more and more ore deposits were analyzed, occasional overlaps between mines or mining areas were discovered (Pollard and Bray 2014,231) and also, where significant uranium or thorium was present in the deposit, some very broad isotopic ‘signatures’ emerged (Pernicka 2014,249). While both chemical and isotope methods have some drawbacks, when they are used in combination they can be provide compelling evidence because they are both completely independent methods. Hence, the best practice is to use both methods in provenance studies (Pernicka 2014,263) wherever possible. In addition, information on the date of mining activity in a particular area can further strengthen the evidence linking mines to objects of the same age (Pernicka et al. 1997,143). Roberts (2014,433) pointed out another requirement to increase the probability of success of such studies, is that the ore source should have been the dominant source of metal in an area for substantial period of time in

order to make an impact on the regional artefacts as the dominant metal composition. Recycling, to some degree, is very likely to have occurred since earliest times. There is some debate about the extent of recycling in the Bronze Age and the degree of volatile loss on remelting (Pernicka 2014,254-258, Bray and Pollard 2012,854,865). Some recycling would have been of metal objects from one dominant primary source and so the isotopic and chemical signature would be unchanged apart from some possible volatile loss (see below). Even when metals from two different sources were used the resulting metal will lie between the primary sources in terms of trace/minor elements and lead isotopes (mixing lines) but obviously these signatures will become of little use with multiple metal sources (Pernicka 2014,258). Also re-melting metal may reduce the content of volatile elements, particularly of arsenic and antimony (Bray and Pollard 2012,854) under certain redox/remelting conditions and consequently slightly increase the level of non-volatile elements. However, the presence of nickel may inhibit arsenic loss (Sabatini 2015) and also losses may not be significant at low concentrations and with particular redox/remelting conditions. Sudden and dramatic changes in the dominant metal composition (both chemi-

Figure 1. Dated Bronze Age copper mines and trials in Britain and Ireland with main mines named.

31

Figure 2. Seven broadly dominant artefact-based metal compositions in the Copper/Early Bronze Age and in the metalwork phases of the Middle Bronze Age and Late Bronze Age. Note the elemental proportions shown are not to scale and are symbolic only. Approximate correspondence to Northover groups (A, C, M, N, P & S) is indicated. Chemistry can be used to distinguish most groups but those with a thick black border (Groups 4, 5 & 6) require lead isotope to separate them.

cally and isotopically) are sometimes clearly visible in the archaeological record. For instance, in Britain from the Early to Middle Bronze Age (Arreton to Acton Park) and later in the Middle Bronze Age (Penard to Wilburton). This suggests that the effects of recycling, at least in some periods and in particular regions, can be muted or diluted to low levels when a major flow of new primary metal becomes established. This type of sudden change may correlate with a major change in the dominant exchange network due to broader societal changes and/or the exhaustion/discovery of particular ore deposits. The discovery in Ireland and Britain of numerous Bronze Age mines and trials since the 1980s (Fig. 1) has transformed our understanding of potential copper sources (Timberlake 2009, O’Brien 2004) but probably not many of them made a significant contribution to the metal supply. At the Ross Island mine in SW Ireland, the application of lead isotopes combined with the very distinctive chemistry of the grey fahlore copper ore (As, Sb and Ag in tennantitetetrahedrite) matched the artefact-based metal group ‘A’ which dominated the Irish and British Copper Age and Early Bronze Age (Fig. 2).

Smelting experiments with samples of probable ore proved difficult but eventually three copper prills were micro-analysed and were found to be consistent with ‘A’ metal (O’Brien 2004, 532). However, since the Ross Island work, the matching any of the other dominant metal groups during the rest of the British Bronze Age to British, Irish or continental mines has stalled. This has been partly due to the papers in the late 1990s onwards, which claimed that most British (and some Irish) Bronze Age mines could only produce copper with low impurity levels (Ixer and Budd 1998,26). However, this conclusion was based only on mineralogical studies without any geochemical analyses. This low impurity claim only matched a minor metal group, mostly in the Early Bronze Age, while the Middle Bronze Age was dominated by arsenic-nickel compositions (Northover 1991,65) (Fig. 2). In addition to this difficulty, the wide range of lead isotope signatures in British mines and artefacts, revealed by pioneering work of Rohl and Needham (1998), seemed to make the whole business of unraveling which artefacts were from British mines too difficult and so little further work has been done. There has been some debate about 32

duced numerous macro or bulk homogenized crushed samples for analysis. The major weakness of some previous studies is that selected micro-sized grains of a copper mineral in the ore was used rather than the whole suite of minerals in the ore. The mine-based metal group is defined in two independent ways. Firstly, by establishing in detail the broad natural geochemical range (‘signature’) for all the key impurities and secondly, by defining the full lead isotopic range (‘signature’). A crucial aspect is deciding upon the correct ores to sample, which depends on a detailed mineralogical understanding of the ores (Baron et al 2013,2 , Killick 2014,11) in parallel to an archaeological/geological understanding of the mining remains. For instance, the presence or absence of separate or contemporary lead mineralizing events with the copper mineralization can be crucial in lead isotope interpretation. Overall, this approach is much more likely to reveal the probable metal composition range and isotopic range that the ore deposit could have produced. In the past, the inherent assumption has often been that a mine’s ores would produce a narrow metal composition whereas in reality a range of compositions would usually be produced given the natural variation of ore deposits. Combining this approach with smelting experiments, using the actual ores from the Bronze Age workings of a particular mine,

a possible minor Copper Age copper source(s) in SW England (Budd et al 2000, Bray 2012, 60) based on rare highly radiogenic artefacts but the location of the source mine(s) remains elusive. MINE-BASED METAL GROUPS – A NEW APPROACH A new approach is required to give fresh impetus to the stalled process of matching Bronze Age artefacts to the copper produced by the principal British Bronze Age copper mines discovered over the last 30 years. The new approach involves establishing the concept of mine-based metal groups by harnessing knowledge from other disciplines, namely ore geology, ore mineralogy, geochemistry, geometallurgy and pyrotechnology. Now that numerous Bronze Age copper mining sites have come to light there is the opportunity of turning the usual provenance question around. Rather than seeking to assign the relatively artificial artefact-based metal groups to mines/mining regions we can now define mine-based metal groups, each of which might partially or fully overlap with several existing artefact-based groups. A mine-based metal group is based on extensive and systematic ore sampling from actual Bronze Age mine workings, from which are pro-

Figure 3. General view of the Great Orme mines.

33

Figure 4. Bronze Age chronology of main mines and approximate radiocarbon. Date ranges at the 2 sigma level and associated metalwork assemblages. Radiocarbon data from Timberlake and Marshall (2013).

THE GREAT ORME BRONZE AGE COPPER MINE

can lend further strength to the defining the limits of a mine-based group and to allow for an understanding of the split of trace/minor elements between copper metal and slag. The research team at Ross Island (O’Brien 2004) went in the direction of the new approach being proposed, however, they were unable to sample primary and probable secondary ores from the extensive flooded Bronze Age workings in order to establish the full range of ore compositions in the mine. In addition, with the ore fragments that were selected, they relied on micro-analysis (electron micro-probe) of tiny individual ore mineral phases within naturally heterogeneous ore samples rather than on crushed and homogenized macro ore sampling, encompassing all phases present for geochemical analysis. This prevented the full range of practical ‘run of mine’ ore compositions from being established. Hence, there was a heavy reliance on the artefact-based metal group (‘A’ metal) to define the full range of metal compositions that the mine probably produced rather than being defined independently by the mine’s ores.

The ideal candidate to test the mine-based metal group approach is the Great Orme Bronze Age copper mine on the coast of North Wales (Figs. 1 & 3), which is generally agreed to be one of the largest Bronze Age copper mines in Europe (O’Brien 2015,146). There is access to around 6km of Bronze Age workings with many areas having residual ore in situ. In addition, there are two additional sources of material to test the whether the mine-based group established from the ores is likely to be correct. Firstly, there are the bronze fragments excavated in the mine, which are probably from metal mining tools made with the local ore, and secondly, the copper prills from the near-by fragmentary smelting site. The Great Orme Bronze Age copper mine is located on a Carboniferous Limestone headland above the seaside town of Llandudno and the workings include a large opencast and tunnels up to 70m metres deep. The extensive ancient workings were found under the nineteenth century mine tips in 1987. Excavations have found over 2400 stone hammers and around 30,000 bone fragments, often used as mining tools (Jen34

kins and Lewis 1991, Dutton et al 1994, Lewis 1996, Wager 2001, James, 2011). A mining museum was opened in 1992. At this point it is worth noting a recent review of the chronology of known British Bronze Age copper mines (Fig. 4) based a Bayesian analysis of radiocarbon dates (Timberlake and Marshall 2014). This has suggested that after earliest metallurgy from the Irish Ross Island mine from around 2400 BC (until around 1900/1800 BC) there was a wave of exploration in western Britain, possibly from Ireland, in Mid-Wales and north Wales from around 2100 BC. This appears to have lead to many trials and the opening of mines at Parys Mountain and Copa Hill mines (Fig. 4) plus, slightly later, some workings in central north-west England (Alderley Edge and Eaton) but all had closed by around 1600 BC. The only exception is the Great Orme mine whose radiocarbon dates stretch from around cal. 1884 to 933 BC (2 sigma range, James 2011, 45) indicating that it continued after all the other known British mines had closed. The closures may have been related to the Great Orme’s rich and easily worked oxide ores, which contrasted with poorer less oxidised deposits in much hard rocks at the other mines. What is not highlighted in the literature is that the initial British exploration activity, around 2100 BC, corresponded with the relatively sudden change to full tin bronze in Britain and may suggest the wave of exploration also revealed the cassiterite deposits of Cornwall and/or Devon in SW England. Gold has also being recently linked more to this area (Standish et al 2015,18) than the traditionally assumed sources in Ireland. The Great Orme mine’s radiocarbon dates go back to the late EBA although most dates lie in the MBA. Northover (1991,65) noted the sudden change in the dominant metal composition between the CA/EBA (As-Sb-Ag ‘A’ metal) to the MBA (nickel-arsenic metal). He suspected a source for the nickel-arsenic metal in north Wales (possibly Snowdonia) or Shropshire. Given the discovery of the very extensive Great Orme Bronze Age mine in 1987 with its many MBA radiocarbon dates, this would seem to be the obvious source. However, during the late 1990s a series of influential papers stated this was definitely not the case (Ixer and Davies 1996, Ixer and Budd 1998 and Ixer 2001). They stated that «… the mineralogy of the ores shows that they could only produce trace element poor copper metal like most Bronze Age ores from the British Isles … and that their usefulness in provenancing Bronze Age metal work based upon distinctive trace element signatures is very limited» (Ixer & Budd 2001,218). This conclusion was reflected in a number of other papers including Budd (2000) «…the Great Orme ores contain no ar-

senic whatsoever» , Craddock (1994,76) «…copper from the Great Orme cannot be identified in the contemporary bronzes…the ore is characterized by very low nickel contents» and Northover (1999,223) «…..it is incompatible with the vast bulk of Middle Bronze Age metal». This view has persisted and is stated in a recent book on prehistoric copper mining in Europe, « The copper produced was high purity, making it difficult to follow its circulation in the wider pool of metal in that period» (O’Brien 2015,150). Hence, the consequence of this low impurity conclusion was that the Great Orme mine was not considered as important in the Bronze Age as the size of the workings would suggest. However, Ixer’s conclusions were based on mineralogical observations only (noting the lack of specific arsenic or nickel minerals) but without geochemical analyses, which might have detected high levels of impurities within the structure of the various copper minerals and other major gangue (waste) minerals (particularly iron oxides) present in the ore. Some scholars such as Rohl & Needham (1998,111,181) using some lead isotope data points and others (Northover in Lynch et. al., 2000,99, Timberlake 2009,115, Bray 2012,60) expressed the view that the mine may have been an important source of MBA metal (and some EBA metal) but offered no explanation to resolve the conflicting mineralogical, geochemical, smelting, isotopic and artefactual analysis evidence. Using the mine-based metal group methodology at the Great Orme should not only demonstrate the methodology in practice but also unravel the various claims made about this mine in the past. MATERIALS ANALYZED (ORES, BRONZE PARTICLES AND COPPER SMELTING PRILLS) Three groups of materials have been studied and analyzed in detail, copper ores, bronze particles and copper smelting prills. Firstly, the copper ores from the several kilometers of Bronze Age workings at the Great Orme mine were sampled using specialist knowledge of the ore geology to guide the sampling process. This work established for the first time that the main ore type was a dark malachite-goethite ore formed by in situ supergene oxidation of the primary chalcopyrite (copper-iron sulphide) in north-south trending anastomosing veins which in places merge to form larger ore bodies mined out in large chambers. Associated with the goethite (α-FeO(OH)) are amorphous iron oxides and the general overall term ‘limonite’ has been used by Ixer (2001,215) to encompass all the iron oxides present. Traces of unconverted sulphides are 35

present in some ores. There are minor amounts of green malachite-only and azurite-only ores, usually heavily diluted by gangue (waste) minerals, where the copper has moved away from the main primary veins in solution and precipitated in other locations in the surrounding area (e.g. within mudstone layers). An important observation is that there is at least one vein of lead ore (galena) that crosses and predates the copper deposit (Ixer 2001,217). This is important when considering the lead isotope results and the accidental or deliberate alloying of lead with the copper metal produced. A key feature of the malachite-goethite veins is that they were usually easy to extract with bone tools because the dolomite each side was soft and friable (Lewis 1996,78). Ixer (2001,216) attributed the phenomena to the supergene weathering causing local dedolomitization and partial dissolution of the dolomite host rock locally adjacent to the vein. This made an enormous difference to the ease of working and probably explains why firesetting remains are not very common. The second group of materials analyzed are the copper prills from the nearby small truncated Late Bronze Age (around cal. 900 BC) smelting site at Pengwern about 1.2km from the mine excavated in 1998 and 2011. An archaeometallurgical study of the prills and slags was recently published and revealed the simple smelting of oxidised ores producing copper prills without a full molten slag and showing geochemical links to the Great Orme ores (Williams 2014,104,108). The third group of materials analyzed are the bronze particles excavated in the mine in the 1990s plus a bronze tip that was found in 1831 by the miners in an ancient working (Lewis 1996,131 and Williams, forthcoming). These fragments could be from bronze picks like those found in the Hallstatt salt mines (Kern et al 2009) and in the Mitterberg copper mines (O’Brien 2015,211).

more recently cconfirmed by microwave plasma atomic emission spectroscopy (MP-AES) using an Agilent 4200, both in the Department of Archaeology, Classics and Egyptology at the University of Liverpool. The results obtained were consistent with the results from other techniques used but not presented here (XRF-WD, pXRF, SEM-EDS and LA-ICP-MS). For the lead isotope work, data was obtained on ore and metal samples supplied to the NERC Isotope Geosciences Laboratory at Keyworth who used a Thermo Fisher Neptune Plus MC-ICP-MS. Certified and standard reference materials were used with all the analyses. Full details of all these techniques will be included in forthcoming papers and a PhD thesis. For chemical analysis, twenty-eight ore samples were analyzed incorporating intimately associated gangue (waste) minerals in case they were a source of trace elements. Given their inhomogeneity, at least 5 grams (wherever possible) of a hand-picked ore concentrate was crushed and from which a representative 1 gram sample taken. This was dissolved in aqua regia, filtered and then diluted for analysis. The AAS results show very good correlation with key trace elements in the certified standards (As and Ni within 0.6% relative to the certified value). The metal samples (bronze particles from the mine and copper prills from the smelting site) were usually 10 to 20 mg in weight and dissolved in aqua regia according to a well established technique (Hughes et al 1976) and diluted for analysis. MP-AES was used when this new equipment became available during this project and which has a greater sensitivity than AAS. Pre-existing metal analyses have been included from the literature (Lewis 1996,131) and the OXSAM database based on AAS and electron-microprobe data. Twenty new lead isotope analyses were obtained using either 50mg from each crushed ore concentrate or 10mg from the metals (which are more homogenous than ores). The lead isotope data was produced at the NERC Isotope Geosciences Laboratory at Keyworth using a Thermo Fisher Neptune Plus MC-ICP-MS. Pre-existing isotope data has also been included (Joel et al 1995, Rohl & Needham 1998, Northover 1982a)

ANALYTICAL TECHNIQUES Several analytical techniques have been used to ensure the data obtained is robust and not affected by the peculiarities of one particular technique. The accurate quantification of the minor and trace elements in rich copper ores is a significant analytical challenge because the very high levels of copper and iron in the ores. This results in a wide range of emission or absorption peaks that sometimes interfere with those from other elements of interest and so development work has been required to ensure good quality data. The chemical data presented in this short paper was obtained by using atomic absorption spectroscopy (AAS) using a Perkin Elmer 3110 and

RESULTS - CHEMICAL ANALYSES Table 1 shows chemical analysis data on ores for the key elements that are normally used in characterizing copper alloys and were mainly obtained using AAS. The initial data from the other techniques, particularly MP-AES, described above gave generally similar results and included additional major, minor and trace elements and will be reported in a future paper. However, 36

Table 1. Minor and trace element macro analyses of Great Orme ores using AAS except where indicated

the AAS results contain the key data required to initially define the main chemical analysis part of a mine-based metal group. Two key elements to consider are arsenic and nickel and a simple plot is shown in Fig. 6. If the papers by Ixer and co-authors (1996,1998, 2001) claiming only low impurity levels in the Great Orme ores had been correct we would expect the copper ores analyses to lie in the shaded square shown in Fig. 5, in the bottom left hand corner (all below 0.1%). However, plotting the ore results normalized to 100% copper (see below) dramatically shows that there are substan-

tial amounts of arsenic and nickel in all the malachite-goethite ores. Laser ablation ICP-MS work has demonstrated that while the malachite can be low in impurities, the intergrown goethite with its absorbent structure acts as sponge for high levels of trace elements (Manceau et al. 2000). The large square in Fig. 5 is the average of all the ore results, which is useful to consider as naturally occurring ores are usually much more heterogeneous than metals (which are mixed during smelting and refining). The much rarer malachite-only ores are lower in impurities and are shown as small triangles. 37

Figure 5. Arsenic and nickel plot of Great Orme ores (average is large square), bronze particles from the mine, copper prills from Pentrwyn smelting site and the typical range of British Bronze Age metalwork data. Shaded square indicates where low impurity metal would plot (less than 0.1%).(Data: Table 1, Lewis 1996, Williams 2014, and Williams forthcoming).

Note that all these results have been normalized to 100% copper from variable ores that contain 16% to 38% copper using the convention used by Pernicka (2004,315). For simplicity, this is based on the ideal smelting situation where there would be a 100% transfer of nickel and arsenic into to the copper metal produced rather than into the slag phase. There is evidence in the literature that high levels of transfer are achieved in reducing atmospheres as these two elements preferentially move to the copper metal (Tylecote et al 1977,19, Hauptmann 2007, 27,204-207). The other key thing to consider is that smelting oxide ores (albeit with slight traces of sulphides in some ores) do not require the oxidizing roasting step that sulphide copper ores require before smelting to remove the sulphur and which would cause some arsenic loss. Therefore, going straight to a reducing smelting stage with oxide ores means a large proportion of arsenic in the ores can be retained. Hence, the level of arsenic required in an oxide ore to give a metal with a particular arsenic content in the copper metal is lower than with a sulphide ore. To test the claims in the literature of high arsenic and nickel retention during smelting oxide ores, two smelting experiments were under-

taken separately with two people experienced in experimental smelting. Firstly with David Chapman from Ancient Arts and later another experiment with Simon Timberlake, both using pre-analyzed Great Orme ores. These field experiments using simple bellows-powered pit boles and ores introduced as coarse milled powders, will be reported in detail in a future paper. However, the initial analyses of the copper prills are indicating high arsenic and nickel retentions sometimes exceeding 80% even without the probable accumulated generational skills of the ancient smelters in achieving the best reducing conditions and temperatures for optimum smelting results, probably achieved by attention to colour and smell, etc. Returning to the arsenic–nickel graph of the ores, which is starting to define range of compositions for these two elements. We can test this range by plotting the results from the bronze particles from the mine and we see a strong coincidence with the average ore analysis. If we plot the copper from the Pentrwyn smelting prills they have lower levels of impurities but still within the range of the ores. Hence, the geochemical signature of the ores is emerging with the broad ellipse encompassing all points but with an inner 38

Figure 6. Arsenic – Nickel plot of the Acton Park metalwork assemblage (c 1500 – 1400 BC) and similar metalwork from the Voorhout hoard in Holland and from Tréboul in Brittany (star symbols are mean values). The ellipses define the inner and outer range of ore samples. (Data: Rohl and Needham 1998, OXSAM database, Briard et al 1984, Northover 1982).

ellipse defining a core area containing most of the metal compositions produced (Fig. 6 and 7). This defined geochemical range can now be compared against the Acton Park assemblage from the early Middle Bronze Age whose type locality is near Wrexham in North Wales. This data (Rohl and Needham 1998 and OXSAM online database) defines an area almost identical to that of the ores (Fig. 6). In addition, the Voorhout hoard in Holland has been identified as belonging to the Acton Park assemblages (Butler 1963, Northover 1982b,54 and 1989,220 ), also plot close to the centre of the area defined by the average ore. A similar correlation is seen for the Tréboul hoard from Brittany. So the geochemistry part of the mine-based group is starting to be defined, essentially ranging from low to high arsenic and nickel. Similar work on other elements shows that the Great Orme ores have characteristically both low antimony and fairly low silver levels with variable cobalt. The lead (Pb) can vary from very low levels to several percent due the lead vein that crosses the Great Orme site and so some addition could have been unintentional. However, the very high levels in a few artefacts mean deliberate lead alloying cannot be excluded as a possibility. The

mine-based metal group that emerges coincides with Northover’s artefact based-metal groups M1, M2 and parts (but not all) of his groups O, N and P (Northover 1980,237). As previously mentioned, Northover defined 14 artifact-based metal groups in the British Bronze age and Rohl and Needham defined 23 groups incorporating lead isotopes. However, for simplicity it is useful to focus discussion on the dominant artefact-based metal groups during the Early Bronze Age and the various periods in the Middle/Late Bronze Age. Seven dominant groups have been very broadly defined in Fig. 2 and those roughly equivalent Northover’s artefact-based groups (A, C, P and S) are indicated. The Group 4 range encompasses Northover’s M1 and M2 metal plus part of his O metal. Group 5 includes his N2 metal. Group 4 and 5 can be divided broadly on As>Ni and As0 - 0.249) size fractions (Orton et al. 1997; Gámiz Caro et al. 2013). Many of them have pouring spout. The ceramic is highly vitrified in surface by thermal effect, suggesting working temperatures well above 1,100º C. The fabric of the deep crucibles (Fig. 5: C and D) has as particular feature the intentional addition of vegetal temper. With respect to other fabric matrices, this is a somewhat finer, loamy, very porous and much more lean, with a more homogeneous distribution of tempers. Mineral tempers (quartz, quartzite, feldspar, mica schist and mica) are medium or large-sized (0.5 a 1). Vegetal temper is predominant. One aspect to highlight of these vessels is that they often present a flat rim with a series of (2-4) semi-circular or oval notches scattered on it. This

does not occur in domestic pottery. Sometimes the rim may be bevelled to the inside or, in the case of vessels with less thick walls, to present a sort of central groove. In all three cases these traits could be used to fit a cover. There are also some specimens with rounded rim, regardless of the wall thickness, although they are less numerous. The deep crucibles can also have a pouring spout. The slaggy layer of these vessels is highly significant. In most of them it tends to be thin, whitish-yellowish-green in colour. In others cases, as in most bowl-shaped crucibles, the slag show a dark colour and a considerable thickness that may even exceed the rims. Finally, eight fragments belonging to ingot mould have been collected. Their fabric is rude, similar to the one of the bowl-shaped crucibles. QUANTIFICATION OF THE DUMP CONTENT (Fig. 6) The excavation of this small space (3.95 m3) has allowed the recovery of a large number of metallurgical wastes that can be summarised as follows: 69

Figure 6. Quantification of metallurgical ceramic waste in the dump and inside the village.

– 16.2 kg of slag – 4 kg minerals, mainly green copper ores but also galena and iron oxide (Fig. 6). –  1.771 fragments of deep crucibles (48.6 kg). – 98 fragments of bowl-shaped crucibles (1.6 kg). – 225 fragments of indeterminate crucibles (1.6 kg). – Fragments of 8 moulds (4.1 kg) – 34.7 kg of lining fragments (Fig. 6).

Table 1. Weights of metallurgical debris of Peñalosa

In the preliminary conclusions on the Peñalosa metallurgy was proposed that the deep vessels would be used mainly for melting tasks while the bowl-sized ones would be ore smelting pots (Moreno Onorato et al. 2010: 320). Now, with a higher number of samples analyzed it appears that both types of crucibles were indistinctly used to smelt or to melt, being smelting the prevailing function (Tab. 2). This is quite consistent as more crucibles are needed to produce the metal than to melt it.

COMPARISON OF THE METALLURGICAL DEBRIS OF THE DUMP AND THE VILLAGE (Figs. 7 and 8) The table 1 shows the weights of the metallurgical residues found in the village (Moreno Onorato 2000) and the ones corresponding to the dump. Considering the global numbers we can conclude that the amount of metallurgical ceramics is similar in both areas while the remains of copper ore are more than twice in the village than in the dump. On the other hand, the amount of slag in the dump is much higher, as expected. The functional classification of metallurgical ceramics for smelting or melting operations is based on the characteristics of the slag adhering to the wall of the vessel determined in the laboratory. The former usually retain relicts of the original ore, and other minerals such as magnetite and leaded compounds have been formed that cannot be properly explained by the composition of the clay (see below).

THE CERAMIC ANALYSIS: RESULTS The analytical data of the ceramic matrices allow us to determine its great technological, stylistic and functional similarity. They are ceramic manufactured exclusively for the development of metallurgical tasks. The material used to make the crucibles is common clay whose composition is shown in the table 3. As can be seen, it is poor in calcium and with an iron content ranging from 2.2 to 8.0%. Some specimens are contaminated by copper salts after their metallurgical usage. 70

Table 2. Function of the metallurgical ceramic analysed. Percentages calculated within each of the spaces under consideration. It is assumed that each sample belongs to a different crucible

Table 3. Analysis of metallurgical ceramics (SEM microanalysis; wt. %; nd: not detected)

Figure 7. Quantification of metallurgical ceramics in the dump and inside the village.

71

Figure 8. Estimated number of items in the village and in the dump.

Figure 10. Section of a smelting crucible. Ceramic body with heterometric quartz grains and high porosity. SEM image, backscattered electron.

In many cases the outer surfaces of the crucibles readily disintegrate, which is due to a poorly cohesive clay structure suggesting low temperature firing or that the vessel was fired mainly from the inside during the metallurgical process causing an uneven heating of the ceramic body. This will result in weak and easily broken containers (Fig. 9). In other cases the ceramic body matrix shows better quality. The figure 10 is an example. No reaction between temper and clay is observed, but the high porosity suggests gasification of clay compounds that use to occur at temperatures above 1,100 ºC. This is probably due to its metallurgical use because the common domestic pottery of the Bronze Age was not usually fired to such a high temperature.

Figure 11. Section of a smelting crucible. Contact zone between the slag (left) and the clay (right). SEM image, backscattered electrons.

A thick layer of thermal alteration, usually more than one millimetre thick, with large vacuoles, is formed on the inner surface of the smelting crucibles (Fig.11). Instead, the altered layer is much thinner in the melting pots. The reaction with the ceramic materials is less aggressive (Fig. 12). Fig. 9. Section of a smelting crucible. Observe the poorly cohesive clay structure on top of the thick slag layer. The temper is coarse quartz and mica-schist grains. SEM image, backscattered electrons.

72

more from the quantitative than the qualitative point of view. So far we only had materials removed from the workshops in and outside the huts. Regarding minerals, previous studies determined they come from two different sources, one of copper oxides and carbonates and another of a more complex mineralization in which copper and lead are associated (Moreno Onorato et al. 2010). LIA analyses had identified two mines (Hunt Ortiz et al. 2011) one of which, Mina del Polígono, is rich in this combination of copper-lead ore (Arboledas Matínez and Contreras Cortés 2010). A wide series of new analysis of materials from the dump have been performed using a pXRFED Innov-X Alpha spectrometer, of National Archaeological Museum, operated by Ignacio Montero, and an environmental SEM Fei Inspect with detectors of secondary and backscattered electrons and a built Oxford Instruments Analytical analysis system-Inca, of the National Museum of Natural Sciences in Madrid (CSIC), operated by the microscopists L. Tormo and M.M. Furió. These new analyses are in addition to those already previously obtained of samples from the village, which were already disclosed in previous works (Moreno Onorato et al. 2010; Rovira Llorens et al. 2015).

Figure 12. Section of a melting crucible. Note the thin layer of slag (left) on the clay surface. SEM image, backscattered electrons.

ARCHAEOMETALLURGICAL SURVEY The excavation of the dump has significantly broadened the spectrum of metallurgical debris,

Table 4. p XRF-ED bulk chemical composition of copper ores from the dump (wt. %; surface analysis; nd.: not detected; LE: light elements)

73

low melting point slag but a conglomerate containing solid and partially fused materials. Another frequent mineral is magnetite, which together with delafossite are good indicators of an environment into the hearth whose oxygen fugacity is variable causing that chemical reactions took place sometimes under reducing conditions and sometimes under oxidizing ones. A good example of this is documented in figure 14, where fayalite and magnetite coexist in the same field of a slag.

The table 4 shows the analysis of a selection of minerals from the dump. As can be seen, most are ores with little gangue (small amount of light elements), suggesting that the ore was carefully selected in the mine to prevent the transport of unwanted material. This is also suggested by the absence of residues of ore dressing, both in the villages and in the dump. On the other hand, the fact that among the minerals in the dump predominate the copper-lead associations is telling us that perhaps the dump accumulates metallurgical remains of a period in which they were preferably exploited, probably from Mina del Polígono. This idea is also confirmed by the analysis of slags, being predominant the copperlead ones, as we will show below. Comparing the bulk compositions of ores from the huts (Rovira Llorens et al. 2015: 357) and the dump we observe that leaded minerals are predominant in the latter (Fig. 13), as it has been said before.

Figure 14. Fayalite skeletons and magnetite crystals in a slag. Dark grey sticks growing in the matrix are hedenbergite. SEM image, backscattered electrons.

These features are found in immature slags well known in Spain and other regions of the Old World since the Chalcolithic period (Rovira Llorens and Renzi 2010; Haupmann 2007), and they are referring to a primitive metallurgy to obtain raw copper using open fires and ceramic vessels as smelting reactors or containers. In these circumstances, the redox conditions of the system are very variable, sometimes oxidizing condition (which encourages the formation of magnetite and delafossita, for example), sometimes achieving reducing conditions to get metal and other compounds, depending on the composition of the charge. Semiquantitative analysis of bulk composition of slags from the dump and the huts does not show significant differences (Fig. 15), although leaded slags are more frequent in the dump, as could be expected after the study of the minerals. What really makes original and surprising the Peñalosa metallurgy is the exploitation of copper-lead minerals, undoubtedly abundant in Sierra Morena. The copper-lead slags are

Fiigure 13. Elemental bulk composition of copper ores from Peñalosa.

Previous analyses allowed defining the characteristics of the slags from Peñalosa (Rovira Llorens et al. 2015: 357-359). They can be summarised as follows: immature slags containing plenty of free silica, corresponding to physical and chemical not-in-equilibrium systems resulting of direct reduction of the ore without adding slagging and fluxing minerals. Molten material forms a silicate matrix in which metal prills are embedded. It is common to find remnants of non-reduced original mineral charge and/or its transformation to cuprite, delafossite (if iron oxide is in the gangue) and crystal formations of hedenbergite, melilite, akermanite, wollastonite and other silicates in the series. Fayalite is found in iron-rich slags but the whole product is not a 74

used. To decide whether the slag on a sherd has been formed in a process of smelting or melting is simple when relicts of primary ores are identified by the SEM analysis: primary ores are not expected to be found in a crucible used to melt metal. But if that is not the case the decision is not so simple because some of the minerals identified in the slag could also be formed by reaction of the metal impurities with the crucible walls and ashes. There is, however, a subjective argument that can be applied in these cases. While a reduction process is time consuming at high temperature, melting a similar mass of metal requires a much shorter time if the thermal conditions are suitable. The reaction time is an important factor that makes that in the first case the slag layer formed on the ceramic surface is thicker than in a melting crucible. At least that is what is experimentally deduced. In fact in many melting experiments there is no time enough to form slag but a thin glaze layer (Rovira 2012). The criterion applied for classifying the metallurgical ceramics of Peñalosa is based, therefore, on the results of the analysis and, if the results have not been sufficiently conclusive, on the formal aspects of the slaggy layer. The new analytical data obtained from the dump have largely confirmed the metallurgical aspects defined in previous works (Rovira Llorens et al. 2015), so we see no need to expand with more data. There is, however, an intriguing issue on the Peñalosa metallurgy: what kind of pyrometallurgical structures were used to obtain copper? The excavations of the settlement have not provided any light on the matter. Were they simple fireplaces similar to the domestic hearths? It is likely that the crucibles were stationed in places without special preparation, conveniently surrounded and covered by burning coal. The remains would leave an installation of this type would be an ashy soil with charcoal and perhaps some fragment of ore. However, among the remains recovered in the dump are small pieces of clay that appear to have been part of the lining of a structure dug into the ground. The prepared surface is smoothed and frequently altered by a slag layer whose composition is similar to the slag on smelting crucibles (Tab. 5). If our interpretation is correct as can be deduced from the analysis, in addition to smelting crucibles simple pits small in size were also used for metal obtaining.

characterized by a complex silicate containing lead as melted material forming the matrix. In some cases almost all the lead goes to the glassy matrix and the metal droplets embedded in de slag contain only small amount of lead id any (Fig. 16). In other samples, however, copperlead droplets and even metallic lead inclusions are found. When refining the raw copper most of lead reacts with the crucible clay forming a leaded glassy layer (Rovira Llorens et al. 2015: 360-361).

Figure 15. Elemental bulk composition of slags from Peñalosa.

Figure 16. Copper-lead slag. Light grey leaded glassy matrix, fayalite (grey) and hedenbergite (black). Crystals and dentrites of magnetite are also visible. White spots of metal. SEM image, backscattered electron.

As we have said, in many of the metallurgical tasks performed at the site ceramic vessels were 75

Table 5. pXRF-ED bulk chemical composition of the slag layer on clay lining samples from the dump (wt. %; surface analysis; nd.: not detected; LE: light elements)

CONCLUSIONS

in the basin of the Rumblar river (Contreras Cortés et al. 2005; Arboledas Martínez and Contreras Cortés 2010; Arboledas Martinez et al. 2015), one in Linares and 18 in the Jándula-Yeguas valleys, which are discussed in another work of this monograph (Arboledas et al. in press). These mines provide remains of material culture of the Bronze Age, mainly ceramics and stone hammers similar to those found in Peñalosa. In one of them, the mine Jose Martin Palacios (Baños de la Encina) have been documented metallurgical remains (two fragments of smelting crucibles) dated in the Bronze Age (Arboledas Martínez et al. 2015). In addition, LIA analysis performed on archaeometallurgical material from Peñalosa and several of the mines located in the Rumblar basin relate mines in this valley with the Peñalosa site and even with some copper objects of the El Argar Culture in the Almeria province (StosGale et al. 1999). If the circulation of metal within El Argar territory is still confirmed, the role of the Rumblar basin and Peñalosa could be interpreted in two ways: either we are talking of the most important metallurgical focus of the El Argar State (Lull Santiago et al. 2010b) or we are describing a highly specialized area of this industry, controlled by a few elites established there to control the flow of metal beyond the Upper Guadalquivir region (Moreno Onorato and Contreras Cortés 2010). Such regional scale production is backed by a singular fact: the manufacture of standardized copper ingots, which as a currency would be distributed by wide territories. In Peñalosa have appeared not only copper ingots (Moreno Onorato 2000) but numerous sandstone moulds to produce them (Contreras Cortés et al. 2010). In the archaeological record obtained previously, the relatively small amount of archaeometallurgical remains found in the huts and elsewhere in the village could give the impression of a modest, domestic activity. The content of the dump, however, show the great importance of this activity and gives us a more detailed picture of the ore processing in the site. It would be important in the future to determine the entire space occupied by the dump in order to make a more precise quantification this aspect.

The metallurgical debris recovered in the dump from Peñalosa reinforce previous studies that illustrate the primitive nature of mineral processing based on the use of smelting crucibles, a technology that generates little slag after selecting rich copper ores. The recovery of more than 37 kg of lining fragments stresses the importance of using hearts dug in the ground for mineral reduction operations, a method that had not been identified in previous studies of the metallurgical remains of the village. Two kinds of ores were worked: a) copper oxides and carbonates, and b) copper-lead oxides, carbonates and sulphides. These minerals produce two types of dross which compositions are markedly different, suggesting that probably they were not exploited simultaneously. Two mines that provide these ore types have been identified by LIA analysis (Arboledas Martínez and Contreras Cortés 2010; Hunt Ortiz et al. 2011). The metallurgical ceramics in the dump added to those found in residential areas allow the reconstruction of basic, standardized types, which are designed and made with special technology (types of temper used) exclusively for metallurgical tasks. The analysis of copper metallurgy in Peñalosa is giving us a series of keys on the importance of this activity in the Bronze Age of the Upper Guadalquivir region and how a large community, both in population and settlements, could support themselves in this space based fundamentally on the metal production. There are a number of elements that support a large-scale copper production and distribution control. A first indication is the existence of a large settlement in areas like the Rumblar river basin where numerous villages and small pillboxes have been documented (Contreras Cortés 2000; Contreras Cortés and Cámara Serrano 2002). It is a hierarchical settlement, with villages of various sizes that seems to have been established specifically for the metallurgical activity, with mining towns near copper outcrops and metallurgical villages specialized in mineral processing. A second indication of this large-scale production is the location of 21 prehistoric mines, two 76

Peñalosa becomes a unique site to investigate the technological characteristics of the production of copper in the Middle Bronze Age, to assess the role of the metal production and its scale within El Argar Culture, and provides suggestions to reconstruct the social and economic relations of a community located in the south of the Iberian Peninsula.

Contreras Cortés, F. and Cámara Serrano, J. A. 2002: La jerarquización social en la Edad del Bronce del Alto Guadalquivir (España). El poblado de Peñalosa (Baños de la Encina, Jaén). British Archaeological Reports, International Series 1025, Archaeopress. Oxford. Contreras Cortés, F.; Dueñas Molina, J.; Jaramillo Justinico, A.; Moreno Onorato, A.; Arboledas Martínez, L.; Campos López, D.; García Solano, J. A. and Pérez, A. A. 2005: «Prospección arqueometalúrgica en la cuenca alta del río Rumblar», Anuario Arqueológico de Andalucía 2002, II. Actividades Sistemáticas: 22-36.

ACKNOWLEDGMENTS This research has been carried out within de framework of two projects: Peñalosa Project (Consejería de Cultura, Junta de Andalucia) and «La minería en el Alto Guadalquivir. Formas de construcción histórica en la antigüedad a partir de la producción, consumo y distribución de los metales» (HAR2011- 30131-C02-01, Spanish Ministry of Economy and Competitiveness).

Contreras Cortés, F.; Moreno Onorato, A.; Arboledas Martínez, L.; Alarcón García, E.; Mora González, A.; Padilla Fernández, J. J. and García García, A. 2014: Un poblado de la edad del bronce que tiene mucho que decir, Peñalosa: últimas novedades, Cuadernos Prehistoria y Arqueología de la Universidad de Granada 24: 111-145. CONTRERAS CORTÉS, F.; MORENO ONORATO, A, and CÁMARA SERRANO, J. A. 2010: «Los inicios de la minería. La explotación del mineral de cobre». In F. Contreras y J. Dueñas (Dirs.): La minería y la metalurgia en el Alto Guadalquivir: desde sus orígenes hasta nuestros días. Instituto de Estudios Giennenses, Diputación Provincial de Jaén. Jaén: 43-122.

REFERENCES ARBOLEDAS MARTÍNEZ, L.; ALARCÓN GARCÍA, E.; CONTRERAS CORTÉS, F.; MORENO ONORATO, A.; PADILLA FERNÁNDEZ, J. J. and BASHORE, CH. forthcoming: «Prospección arqueominera selectiva e intensiva en la cuenca media/alta del río Jándula (Jaén)». Anuario Arqueológico de Andalucía 2014, Jaén. Sevilla.

CORTÉS SANTIAGO, H. 2007: «El papel de los elementos cerámicos en los procesos metalúrgicos. El caso de Peñalosa, grupo estructural VI». Revista Arqueología y Territorio 4, Universidad de Granada. Granada: 47-69.

ARBOLEDAS MARTÍNEZ, L.; ALARCÓN GARCÍA, E.; CONTRERAS CORTÉS, F.; MORENO ONORTO, A.; PADILLA FERNÁNDEZ, J.J. and MORA GÓNZALEZ, A. 2015: «La mina de José Martín Palacios-Doña Eva (Baños de la Encina, Jaén): la primera explotación minera de la Edad del Bronce documentada en el sureste de peninsular». Trabajos de Prehistoria 72: 145-162.

GÁMIZ CARO, J.; DORADO ALEJOS, A. and CABADAS BÁEZ, H. V. 2013: «Análisis de cerámica prehistórica con estereomicroscopía: una guía revisada sobre la descripción de las fases de producción», Cuadernos de Prehistoria y Arqueología de la Universidad de Granada 23: 365-385. HAUPTMANN, A. 2007: The Archaeometallurgy of Copper. Evidence from Faynan, Jordan. Springer. Berlin and New York.

Arboledas Martínez, L. and Contreras Contreras, F. 2010: «La mina del Polígono o Contraminas (Baños de la Encina, Jaén). Evidencias de la explotación de mineral de cobre en la antigüedad». Cuadernos de Prehistoria y Arqueología de la Universidad de Granada 20: 355-379.

HUNT ORTIZ, M. A.; CONTRERAS CORTÉS, F. and ARBOLEDAS MARTÍNEZ, L. 2011: «La procedencia de los recursos minerales metálicos en el poblado de la Edad de Bronce de Peñalosa (Baños de la Encina, Jaén). Resultados de análisis de isótopos de plomo». V Simposio Internacional sobre Minería y Metalurgia Históricas en el Suroeste Europeo, Homenaje a Claude Domergue (León 2008): 197-208.

Contreras Cortés, F. 2000: Proyecto Peñalosa. Análisis histórico de las comunidades de la Edad del Bronce del Piedemonte Meridional de Sierra Morena y Depresión Linares-Bailen. Arqueología Monográficas 10, Consejería de Cultura, Junta de Andalucía. Sevilla.

LULL SANTIAGO, V.; MICÓ, R.; RIHUETE HERRADA, C. and RISCH, R. 2010a: «Las relaciones políticas y económicas de El Argar». Revista Menga 1: 11-35.

Contreras Cortés, F. and Cámara Serrano, J. A. 2000: II. El poblado de la Edad del Bronce de Peñalosa (Baños de la Encina, Jaén). 4. La cerámica. In F. Contreras (coord.): Proyecto Peñalosa. Análisis Histórico de las Comunidades de la Edad del Bronce del piedemonte meridional de Sierra Morena y Depresión Linares-Bailen, Consejería de Cultura, Junta de Andalucía. Sevilla: 77-128.

LULL SANTIAGO, V.; MICÓ, R.; RIHUETE HERRADA, C. and RISCH, R. 2010b: «Metal y relaciones sociales de producción durante el III y II milenio ANE en el sudeste de la Península Ibérica». Trabajos de Prehistoria 67 (2): 323-347.

77

Arqueología de la U.A.M.». Cuadernos de Prehistoria y Arqueología de la Universidad Autónoma de Madrid 37-38: 105-120.

MONTERO, I. and MURILLO, M. 2010: «La producción metalúrgica en las sociedades argáricas y sus implicaciones sociales: una propuesta de investigación». Revista Menga 1: 37-51.

ROVIRA LLORENS, S. and RENZI, M. 2010: «Las operaciones pirometalúrgicas y sus subproductos». In I. Montero Ruiz (coord.): Manual de Arqueometalurgia. Museo Arqueológico Regional, Sección de Arqueología del Colegio de Doctores y Licenciados en Filosofía y Letras y Ciencias. Madrid: 87-122.

MORENO ONORATO, A. 2000: «La metalurgia de Peñalosa». In F. Contreras Cortes (Coord.): Análisis histórico de las comunidades de la Edad del Bronce del Piedemonte meridional de Sierra Morena y Depresión Linares-Bailén. Proyecto Peñalosa, Arqueología. Monografías 10, Consejería de Cultura. Sevilla: 167-222

ROVIRA LLORENS, S.; RENZI, M.; MORENO ONORATO, A. and CONTRERAS CORTÉS, F. 2015: «Copper slags and crucibles of copper metallurgy in the Middle Bronze Age site (El Argar Culture) of Peñalosa (Baños de la Encina, Jaen, Spain)». In A. Haptmann and D. Mondarressi-Tehrani (eds.): Archaeometallurgy in Europe III. Proceedings of the 3rd International Conference. Deutsches Bergbau-Museum Bochum. June 29-Luly 1, 2011. Der Anchnitt. Bochum: 355-362.

MORENO ONORATO, A. and CONTRERAS CORTÉS, F. 2010: «La organización social de la producción metalúrgica en las sociedades argáricas: el poblado de Peñalosa». Revista Menga 1: 53-76. MORENO ONORATO, A.; CONTRERAS CORTÉS, F.; RENZI, M., ROVIRA LLORENS, S. and CORTÉS SANTIAGO, H. 2010: «Estudio preliminar de las escorias y escorificaciones del yacimiento metalúrgico de la Edad del Bronce de Peñalosa (Baños de la Encina, Jaén)». Trabajos de Prehistoria 67 (2): 305-322.

STOS-GALE, S.; HUNT-ORTIZ, M. and GALE, N. (1999): «Análisis elemental y de isótopos de plomo de objetos metálicos de los sondeos de Gatas». In P. Castro, R. Chapman, S. Gili, V. Lull, R. Micó, C. Rihuete, R. Risch and M.E. Sanahuja: Proyecto Gatas 2. La dinámica arqueoecológica de la ocupación prehistórica. Junta de Andalucía. Sevilla: 347-361.

ORTON, C.; TYERS, P. and VINCE, A 1997: La Cerámica en Arqueología. Crítica. Barcelona. ROVIRA LLORENS, S. 2012: «Arqueometalurgia experimental en el departamento de Prehistoria y

78

SMALL SCALE BRONZE AGE METALLURGY: NEW DATA FROM SANTA LLÚCIA (ALCOSSEBRE, CASTELLÓN, SPAIN) Ignacio Montero-Ruiz*, Mercedes Murillo-Barroso**, Gustau Aguilella***, Salvador Rovira****

Key words: technology, Bronze Age, metallography, lead isotopes, elemental analysis, recycling.

Abstract Ancient technology is usually described and defined by that element that remains in the archaeological record. So, it is frequent that the main information comes from big sites or from areas where the economic activity is well defined. However, less common activities or even small scale activities are not always well identified due to the small quantities of residues that could be generated. This is true for ancient mining works, but also for metallurgical activities in small settlements. We present the case of Santa Llucia, an Early Bronze Age settlement, dated around 2200-1800 cal BC, where an exceptional find showed an unusual perspective of metal consumption. Seven small metal items (smelting prills, a possible baringot, a rivet and a sheet fragment) with a total weight of 4.1 g were found inside of a small ceramic vessel and close to it a very small riveted dagger (5.75 gr) was found. Elemental analysis (XRF), metallography and lead isotope analysis were used to define the technology and provenance of these metals. We try to understand how the final composition and the isotopic signature could affect by a recycling activity with this small hoard, taking into account that new and scrap metal has been identified in the same context.

Resumen El estudio de la tecnología antigua viene condicionado por los elementos del registro arqueológico conservados. De este modo es frecuente que la información proceda principalmente de sitios importantes o con áreas de actividad bien definidas. Sin embargo, actividades de pequeña escala o menos especializadas no están bien identificadas por el escaso registro que pueden dejar en los asentamientos. Esto es especialmente cierto en la minería pero también en la actividad metalúrgica en yacimientos pequeños. Presentamos el caso de Santa Llúcia, un poblado de la Edad del Bronce fechado alrededor del 2200-1800 cal a.C., en el que un excepcional hallazgo permite identificar practicas poco frecuentes de consumo de metal. Siete pequeños fragmentos de metal de apenas 4,1 g (gotas de reducción, un lingote-barra, un remache y una lámina doblada) fueron recuperados dentro de una pequeña vasija cerámica junto a un pequeño puñal de remaches (5.75 g). Se ha realizado el análisis elemental, la metalografía y los análisis de isótopos de plomo, cuyos resultados nos permiten describir la tecnología y la procedencia de estos metales. Pretendemos comprender como la composición final y la signatura isotópica puede estar afectada por la actividad de reciclado, teniendo en cuenta que tanto metal primario como reciclado aparece en el mismo depósito.

**** Instituto de Historia-CSIC, Spain. **** University of Granada. **** Diputación de Castellón. **** Museo Arqueológico Nacional, Madrid.

79

Palabras clave: Tecnología, Edad del Bronce, Maetalografía, Isótopos de plomo, Análisis Elemental, Reciclado.

pel and a hospice built during the XVII century A.D. The whole area is currently declared as BIC (Spanish acronym for Asset of Cultural Interest). The reuse of this space has conditioned the conservation of the Pre and Protohistoric remains, where some areas were not affected by the subsequent constructions. In such a way, the area subject of archaeological excavations is located south of the plaza that precedes the chapel of Santa Llúcia, where in 2012 a series of remains and some traces of ancient superficial structures were identified (Fig. 2).

1. INTRODUCTION One of the most noted characteristics of metals in relation to other materials such as stone, ceramics or bone, is the possibility of it being recycled. From an archaeological point of view, it is difficult to assess the impact of this activity in the earliest prehistoric metallurgy, being more visible during historic periods where we can count on textual evidence that confirms the practice of recycling ancient metal (Bray et al. 2015). The recycling process eliminates its own trace and can only be documented in unique contexts and findings, such as Late Bronze Age deposits with hoarded or broken objects1, the identification of certain characteristics in the composition of metallic objects that can suggest the use of recycled metal, or the identification of mixture lines through lead isotope analysis (Pernicka 2014). These three arguments are the ones we have tried to follow in this paper regarding Early Bronze Age Metallurgy in the Iberian Peninsula. For some authors (Lull et al. 2010) recycling is an extended practice throughout the argaric Bronze Age in Iberia, fact that would justify the relatively low number of metallic objects that have been discovered. Nevertheless, as we have already stated in other papers (Montero y Murillo-Barroso 2010), neither the composition of objects nor lead isotope analysis support at the moment the existence of intense metallic recycling. The discovery of the metallic deposit of Santa Lucia that we present in the paper will allow us to reflect on the impact of metal recycling and its relation to the volume of metallurgical production.

Figure 1. General location of the Santa Llúcia site.

Until now a total of 4 short two week campaigns have been carried out during the summers of 2012, 2013, 2014 and 2015, with the objective of evaluation the state of preservation of the remains as well as the chronological and cultural sequence of the site. Upon the bedrock, a first occupation belonging to the Early Bronze Age has been identified, having been dated by 14C,2 dating from the last centuries of the III Millennium BC to the first few of the II millennium, phase in which the metallic remains studied in this paper were located. This Bronze Age level is poorly preserved in some areas, having been affected by subsequent phases which hamper the interpretation of the site since it cannot be understood in extension. Thus, some poorly preserved walls have been identified, as well as some supporting structures and small combustion places. In fact, we can see how the best preserved records belonging to this phase are located where the slope and the irregularities and depression of the terrain have allowed its conservation.

2. THE SITE: SANTA LLÚCIA (ALCALÀ DE XIVERT – ALCOCEBRE, CASTELLÓN) The site of Santa Llúcia is located upon a hilltop on the southern extreme of the Irta costal mountain range, at 315 a.m.s.l. The location and height gives this site a privileged visibility of a large portion of the southern coast lineal of Castellon and the sea (Fig. 1). The archaeological site is currently under historical constructions, specifically the remains of a medieval fortress and tower, as well as a cha1   Pernicka (2014: 257) also mentioned bun-shaped ingots with semi-molten pieces of identifiable metal objects from the Late Bronze Age in western Switzerland.

2  The results are: 2140-1950 Cal. B.C. (Beta-336275) from a carbon sample, and 1930-1750 Cal B.C. (Beta4147170) from a bone sample.

80

Figure 2. General view of the archaeological site and the area of works in Santa Llúcia.

Figure 3. Bronze Age small pottery vessel where the metallic items were found inside.

81

Figure 4. Copper items found inside the pottery vessel.

Figure 5. Copper dagger with 3 rivets found close to the pottery vessel.

3.  THE FINDING

of these riveted daggers, despite the typological variability, is usually situated between 60-100g and 15-20 cm in maximum length (Montero 1994; Brandherm 2003). In the neighboring site of Tossal del Mortorum another similar worn down dagger was also recovered. We must highlight the fact that the three rivets are significantly different form one another. Two of them present deformed ends, one has a rectangular section, another one a square section and the third is more similar to a small sheet. Their dimensions are also very small (3.5% As). Prill B on the other hand, with a slower cooling process, could also be a product of primary metal production, though its levels of arsenic are considerably lower (1.42%) If we consider the 3 prills as the reference for the provenance of primary metals, we can observe that they are located at the extremes of the distribution of the objects from both Tossal de Mortorum and Santa Llucia (Fig. 8). We can also add the bar-ingot with high arsenic percentage (4%) to this group of raw material identified with the “O” symbol in the figures The remaining metallic objects from both sites conserve an isotopic signature compatible to its production from the same primary metals, with the exception of the halberd from Tossal del Mortorum located in an outer position(Fig. 9). In other words, all the small objects (awls and

rivets) and the daggers could have been manufactured from the same raw material. The metals with lower amounts of arsenic could reflect the tendency to the progressive loss of arsenic after each melting process if there is no addition of primary metal. In this case the isotopic signature would reveal both the primary metal and the recycled metal without any substantial changes in its value due to similar ratios of its origin. To test if local or imported metal were in use, lead isotope data of copper minerals are available from Castellon province (Montero Ruiz et al. 2014 and unpublished data) and from the neighboring area of the Priorato (Tarragona) where the two only mines that present evidence of prehistoric exploitation in the region are located: La Solana del Bepo and Mina Turquesa (Rafel et al 2014). La Solana del Bepo is a copper-iron mineralization without arsenic, while Mina Turquesa is the only mine in the Montsant area that could produce metals with arsenic. The Linda Mariquita mine, in the district of MBF, also contains copper minerals with high levels of arsenic, antimony, silver and lead (Montero Ruiz et al. 2012: 170. tab 1). The copper minerals from Castellon present variable compositions, from Falhore-type minerals with arsenic, antimony, and lead, and primary copper-iron sulfates. In the case of the Solaig mine the mineralization of Cu-As is the only known one in the region that could produce primary metal similar to the one studied at Santa Llucia and Tossal del Mortorum. But neither Mina Turquesa nor the Solaig mines have similar isotopic rations to any metals from the Bronze Age sites under study, and therefore must be discarded as the origin of them (Fig 10). The closest mine to Santa Llusia and Tossal del Mortorum with isotopic data clearly com-

Figure 9. Lead isotope data for Santa Llúsia and Bronze Age items from Tossal del Mortorum comparing other lead isotope ratios.

Figure 10. Lead isotope data for Santa Llúsia and Tossal del Mortorum in comparison with Turquesa (Tarragona) and Solaig (Castellón) mines.

7.  INTERPRETATIVE MODEL

86

Figure 11. A and B: Comparative lead isotope data between archaeological items and local ores from Tarragona and Castellón provinces.

tern in metals does apparently not match with these mines. If ores from Linda Mariquita were smelted it could be expected to find some antimony and silver in the metal, apart from the arsenic. However the high detection limit for both elements (0.15 %) in this analysis could be hiding this relationship. If we observe figures 8 and 9 we can see that the archaeological materials change their relative position if different ratios are combined. This effect could suggest metal from different resources, but with close isotopic signatures. Prill B has a lower amount of arsenic and is placed in a differentiated area regarding the alignment of the other metals and mines that have been compared, and the position of rivet E with a low amount of arsenic (0.54%) is also apart from the other metals in some figures. This isotopic distribution would restrict the local supply of copper

patible is the lead mine of Miravet (Cabanes) in the same municipality as the site of Tossal del Mortorum (Montero Ruiz et al. 2014). We do not have any record on the existence of copper mineralization in its vicinity, and since it is galena, it cannot be linked to Bronze Age materials. Nevertheless, the Mines del Misteri in Castellon, where we have been able to document both galena and chalcopyrite (Montero Ruiz et al. 2014), and the Mine of Linda Mariquita with some copper arsenide minerals (Montero Ruiz et al. 2012), both present similar isotopic ratios to some of the ones obtained from the archaeological material. The isotopic field of these mines, especially Linda Mariquita, partially overlapped the distribution range of the archaeological items (Fig. 11). This could be a starting point for considering the option of the use of local and regional minerals, although the trace element pat87

to the upper part of the figures 11 A and B and would limit the Misteri mines as possible source of copper. Finally, must be highlighted that the halberd is located in a separate position in all the figures and probably the metal comes from a different resource than the rest of objects and without matching local ores.

en Halbinsel. Prähistorische Bronzefunde. Abteilung VI. Stuttgart: Franz Steiner. BRAY, P.; CUÉNOD, A.; GOSDEN, C.; HOMMEL, P.; LIU, R. and POLLARD, A.M. 2015: «Form and flow: the ‘karmic cycle’ of copper». Journal of Archaeological Science 56: 202-209. CHERNYSHEV, I. V.; CHUGAEV. A. V. and SHATAGIN, K. N. 2007. «High-precision Pb isotope analysis by multicollector-ICP-mass-spectrometry using 205 Tl/203Tl normalization: optimization and calibration of the method for the studies of Pb isotope variations». Geochemistry International 45: 1065-1076.

8. CONCLUSIONS This study could reveal a dual pattern in metal production and use; the supplying and exploitation of local and regional resources are combined with foreign metal, as could be the case of the halberd. The Santa Llusia´s hoard suggest a panorama of small scale production that includes primary metal that may circulate at a local scale, either in the form of prills or bar-ingots. The volume of the circulation of metal, given the form and size it adopts, does not seem to reach great proportions. There was a certain contribution of recycled metal used for small objects that would not alter the isotopic signature, although the arsenic proportion in metal diminished. A good example could be the prill C and an awl from Tossal del Mortorum (n.º 42408) both with similar isotopic rations, but with very different amount of arsenic, which is practically absent in the awl. Recycling seems to be the last option in the metal use, as we observe in the constant resharpening of the daggers from both sites with a very small size of the blade. Beside this fact, we can also detect trade or circulation of objects with a higher social relevance, as is the case of the halberd, maybe at a supra-regional scale. This frame suggests the coexistence of two different economies during the Bronze Age: a small scale domestic production restricted to the creation of small tools (awls or rivets) that could be developed in small settlements, but that is dependent on trade for obtaining larger elements or objects with a higher social relevance, which leads to the necessary existence of supra-regional relationships for obtaining objects that require a larger volume of metal. It would be interesting to observe the pattern of halberd distribution within this frame, and extend our sampling to other areas of the Iberian Peninsula that could infer possible trade at a macro-regional scale.

GUSI, F. and OLARÍA, C. 2014: Un poblado fortificado del Bronce medio y Bronce final en el litoral Mediterráneo: Orpesa la Vella (Orpesa la Vella. Castellón. España). Monografies de Prehistòria i Arqueologia Castellonenses 10. Diputació de Castelló. Castelló. LULL. V.; MICÓ. R., RIHUETE. C. and RISCH, R. 2010. «Metal y relaciones sociales de producción durante el III y II milenio ANE en el sudeste de la Península Ibérica». Trabajos de Prehistoria 67(2): 323-347. MONTERO RUIZ, I. (1994): El origen de la metalurgia en el sudeste de la Península Ibérica. Instituto de Estudios Almerienses. Colección de Investigación. n.º 19, Almería. — (2017): «Metales y metalurgia en el yacimiento del Tossal del Mortòrum (Cabanes, Castellón)». En Aguilella Arzo, G. (coord.): Tossal del Mortòrum un assentament de l´Edat del Bronze i del Ferro Antic a la ribera de Cabanes (Castelló). Monografies de Prehistoria i arqueologia Castellonenques, 12: 97-106. Servei d´investigacions Arqueologiques i Prehistoriques, Castello. MONTERO RUIZ, I. and MURILLO-BARROSO, M. 2010: «La producción metalúrgica en las sociedades argáricas y sus implicaciones sociales: una propuesta de investigación». Menga 1: 37-51. MONTERO-RUIZ, I.; RAFEL, N.; ROVIRA, M. C.; ARMADA, X.-L.; GRAELLS, R.; HUNT, M.; MURILLO-BARROSO, M.; RENZI, M. and SANTOS, M. (2012): «El cobre de Linares (Jaén) como elemento vinculado al comercio fenicio en El Calvari de El Molar (Tarragona)». Menga 3: 167-186.

REFERENCES

MONTERO RUIZ, I., AGUILELLA, G. and ROVIRAHORTALÁ, M. C. (2014): «Plomo metálico en yacimientos de la I Edad del Hierro en la Provincia de Castellón: Explotación de recursos mineros y circulación del metal». Actas del X Congreso Ibérico de Arqueometría (Castellón, 16-18 octubre 2013). Castellón: 200-214.

BRANDHERM, D. 2003: Die Dolche und Stabdolche der Steinkupfer- und der älteren bronzezeit auf der Iberisch-

PERNICKA, E. 2014: «Provenance determination of archaeological metal objects». In Roberts. B.W.;

88

Thornton. C.P. (Eds.): Archaeometallurgy in Global Perspective: Methods and Syntheses. Springer. New York: 239-268.

dernos de Prehistoria y Arqueología de la Universidad de Granada 24: 147-166. ROVIRA. S. and GÓMEZ RAMOS, P. 2003: «Las Primeras Etapas Metalúrgicas en la Península Ibérica». III. Estudios metalográficos. Madrid.

RAFEL, N.; MONTERO. I.; SORIANO, I.; HUNT, M.A. and ARMADA. X-L. 2014: «Nuevos datos sobre la minería Pre y Protohistórica en Cataluña». Cua-

89

TIN PRODUCTION IN BRITTANY (FRANCE): A RICH AREA EXPLOITED SINCE BRONZE AGE Cécile Le Carlier de Veslud*, Céline Siepi*, Christian Le Carlier de Veslud**

Abstract

varias fuentes. En primer lugar se analizaron las prospecciones geológicas exhaustivas realizadas por el BRGM en los años 60-80 que permitieron localizar las áreas mineralizadas, describiendo las configuraciones y las asociaciones de minerales. Se analizaron luego los numerosos informes de prospecciones geológicas y las publicaciones realizadas durante las explotaciones y prospecciones modernas (mediados del siglo XIX y principios del XX) donde se hacen mención de trabajos antiguos de minería. Por último, las prospecciones recientes han permitido descubrir los vestigios de un taller de fusión que fueron fechados de la Edad del Hierro, los resultados analíticos son presentados en este estudio. Teniendo en cuenta todos estos datos, es posible afirmar que la explotación de estaño comenzó en la Edad de Bronce. Esta actividad se continuo probablemente hasta la Edad Media en diferentes lugares del macizo. En este periodo, la explotación del estaño parece haber sido interrumpida, hasta su redescubrimiento a principio del siglo XIX.

This work aims at revisiting the tin archeometallurgical potential in the Armorican Massif. For that purpose, several sources were synthesized. Firstly, extensive geological field surveys performed by the French geological survey in the 60-80’s, which allowed to locate mineralized areas and document their configuration and mineral associations. Secondly, many reports of geological prospecting and publications written during modern tin exploitation (mid-XIXth mid-XXth century), mentioning the presence of ancient mining works were considered. Finally, present-day archeological surveys highlighted the remains of a reduction workshop, dated to Iron Age, and whose results are presented in this study. From all these data, it comes out that the tin exploitation started at the Bronze Age. It continued in different places in the Massif until the Middle Age. At this period, the tin exploitation seems to stop until its rediscovery, in the beginning of the XIXth century.

Palabras clave: Casiterita, minas de estaño, metalurgia del estaño, Bretaña, Edad del Bronce, Edad del Hierro, periodo romano, Edad Media.

Keywords: Cassiterite, tin mines, tin metallurgy, Brittany, Bronze Age, Iron Age, Roman period – Middle Age

Résumé

Resumen

Ce travail se propose de reconsidérer le potentiel archéométallurgique en étain dans le Massif armoricain. Pour cela, plusieurs sources ont été synthétisées. D’une part, les prospections géologiques extensives réalisées par le BRGM dans les années 60-80, qui ont permis de localiser toutes les zones minéralisée en décrivant leur configuration et les associations minéralogiques. D’autre part, les nombreux rapports de prospection géologique et les publications réalisés au

En este trabajo se propone reconsiderar el potencial arqueo-metalúrgico del estaño en el Macizo Armoricano. Para ello se sintetizaron

  *  Laboratoire Archéosciences. Rennes-UMR 6566 CReAAH/CNRS.   ** Laboratoire Géosciences Rennes-UMR 6116 CNRS.

91

instance, in the French Massif Central, lots of ancient gold and tin mines, as well as smelting workshops were uncovered (Cauuet 2006, Abraham and Morasz 1997, Cauuet et al. 2004). The Armorican Massif (hereafter referred as AM) has also a long tradition of tin metallurgy attested by numerous Bronze Age artifacts found in terrestrial hoards (Boulud and Fily 2009). The typology of these metallic artifacts is also in favor of a local production (Briard 1965). Nevertheless, there is no direct evidence that local ores would have been used. This points out to a need for establishing relationships between ore deposits, mines, workshops and finally artifacts. For this reason, since 2010, a more systematic search of tin mines and workshops in the AM has been initiated by the authors, whose first results are given in this paper. The purpose of this study is particularly focused on the search of ancient workshops, in order to reestablish the history of tin metallurgy in the AM, based on a large bibliographic study of geological and archeological data. Thus, the first part of this paper provides a description of the geological context of the AM in relationship with tin deposits. The second part synthesizes available information about ancient tin exploitation obtained during geological prospecting as well as during modern exploitation of tin deposits (mid-XIXth mid-XXth century). The third part presents first results concerning workshop remains obtained during present-day field archeological prospecting. Finally, a synthesis of the history of tin metallurgy in the AM is proposed.

cours des exploitations et prospections modernes (mi-XIXème début XXème siècle) mentionnant des traces de travaux miniers anciens. Enfin, des prospections récentes, qui ont livré les restes d’un atelier de réduction daté de l’âge du Fer dont les résultats analytiques sont présentés dans cette étude. A la lecture de l’ensemble des données disponibles, il est possible d’affirmer que l’exploitation de l’étain a débuté à l’âge du Bronze. Elle s’est probablement prolongée en continu, en différents lieux du massif, jusqu’au Moyen-Age central. A cette date, l’exploitation de l’étain semble s’être arrêtée jusqu’à sa redécouverte au début du XIXème siècle. Mots clés: Cassitérite, Métallurgie de l ‘étain, Bretagne, Age du Bronze, Âge du Fer, Période Romane, Moyen Âge INTRODUCTION The manufacturing of bronze artifacts represents a revolution in the production of metallic objects in ancient societies. However, the attention of researchers remained for a long time focused on the search of the sources of the main constitutive metal, ie copper, with only few works on tin, despite pioneering works such as Davis (1935), Forbes (1964), Penhallurick (1986), Tylecote (1986), Tylecote et al. (1989), Mulhy (1985) for Europe, and Franklin and al. (1977) for the Near and Middle East. Studying ancient tin metallurgy is challenging due to: (1) the destruction of many ancient mining structures by subsequent mining activity, (2) the small amount of slags available. Therefore, in well documented areas, such as Dartmoor (Newman 1996), Cornwall (Gerrard 1996), Bodmin Moor (Herring 1996), NW of the Iberian Peninsula (Meredith 1998) and Turkey (Yener and Ozbal 1987), it was possible to point out a significant ancient mining activity from Bronze Age to more recent periods. By contrast, there are regions with a high archeological potential, such as Erzgebirge, but where more recent mining activity erased potential ancient ones. In order to overcome this loss of information, an archeometric approach is often proposed as a way to characterize tin producing regions and to identify the main trade and exchange routes e.g. using lead isotopes (Molofsky et al. 2014) or tin isotopes (Haustein et al. 2010). Concerning France, in the XIXth century, some local scholars and mining engineers performed field investigations in some mineralized areas, especially in Brittany and located old mines (Pitre de Lisle du Dreneuc 1880; Kerviler 1882; Davy 1897). More systematic archeological surveys began later, in the XXth century. For

1. TIN MINERALIZATIONS IN ARMORICAN MASSIF In Europe, the largest tin-rich provinces are located in the Hercynian belt. The distribution and the size of these deposits are related to the structural organization of this belt. The most important mining district (Cornwall) corresponds to the most external zone of the belt. The more internal Erzgebirge district is clearly poorer. Finally, the most internal parts of the belt, for instance the French Central Massif, the AM and NW Iberian Peninsula, are the less rich. In addition, the distribution and the size of deposits are extremely variable (Chauris 1981). In the AM, tin deposits are frequently spatially linked to hercynian leucogranites, although direct relationships are still difficult to establish (Chauris, 1977). The granites emplacement generally leads to a fracturing of surrounding rocks and to the development of significant veins network in which hydrothermal fluids bearing metallic elements, among which tin, may circulate. Therefore, the main tin-bearing mineral, cassit92

Figure 1. Geological map of the Armorican Massif and related tin rich areas. erite (SnO2) may be found either (1) in impregnation around the leucogranites, or as inclusions in pegmatites, or highly concentrated in greisens; or (2) in vein network, more or less close to the granite, and always in relation with it. This kind of intercrossed vein networks, common in the neighborhood of granites, are called stockwerk and may be as large as 700m. Some of these veins still outcrop today whereas other may have been eroded. In this last case, cassiterite grains may be transported by water and deposited in quiet zones, thus forming secondary deposits, or placer, either fluvial or marine, of variable size. These deposits are generally located at a short distance from the primary deposits from which they originate. Due to their high density, tin minerals are concentrated in the sands or in the alluviums. These kinds of accumulations are extremely numerous in Brittany. Cassiterite displays various shapes and colors (Fig. 2). Crystals are generally octahedral, often blackish in the pegmatites, but reddish in the high-temperature hydrothermal veins. It may also appear as stings or as a characteristic twin (Chauris, 1981). In alluvial deposits, colors are more variable, ranging from brown-black to red-

orange-yellow, and even sometimes colorless. In the veins and in the greisens, the cassiterite is frequently associated with metallic sulphides, such as pyrite, arsenopyrite, chalcopyrite, but also oxides, such as rutile, ilmenite, magnetite, colombo-tantalite. These minerals are included in a gangue consisting essentially of quartz, but also of silicates (muscovite, tourmaline, staurotide…), and phosphates (monazite…). Finally, accessory minerals may be found as inclusions in the cassiterite, such as ilmenite, chromite, columbo-tantalite or monazite. In the placers, most of the minerals forming the gangue are no longer present. By contrast garnet, disthene or andalusite may appear. Heavy minerals, such as zircon, monazite, arsenopyrite, and particularly ilmenite and magnetite are concentrated together with cassiterite. Native metals may also be concentrated in these sands such as gold, bismuth, and even platine. These different kinds of minerals associated with cassiterite in both type of tin orebodies will be the source of chemical elements, typically arsenic, bismuth and even nickel that will be associated with the obtained tin metal after reduction. 93

Figure 2. Different shape and size of cassiterite grains from different origins in Brittany.

2. STATE OF KNOWLEDGE ABOUT MINES AND WORKSHOPS IN THE ARMORICAN MASSIF

activity was never mentioned. Only slags were found close to coastal valleys, but their relationship with tin metallurgy has still to be confirmed as they could also correspond to iron slags. Today, the coast is very urbanized and slags have disappeared. Veins outcropping in the beach don’t seem to have been mined. But their weathering releases cassiterite that is concentrated by the waves in coastal placers. The same is true for the Pénestin beach, whose sand was briefly operated between 1850 and 1913. This location is regularly quoted in the archaeological literature as an ancient mine because of its name meaning «Tin Point». However, no evidence confirms this hypothesis. Another tin-rich beach is located in Betahon (Lulzac 2012) where there is a locality named «Palus Stean» (the Tin Marsh). We can also mention Carantec beach in northern Finistère or Lanildut coast at the west of Saint-Renan granite (Chauris 1967). In summary, tin-rich placers in beaches are present in various locations in the AM at the mouths of rivers, resulting from weathering of mineralized areas. These placers, exploited during modern periods, could also have been exploited in the past without leaving any direct evidence.

The issue with tin metallurgy in the AM is that many ancient mining structures have now disappeared, due to subsequent mining activity or agricultural tillage. However, by collecting data in bibliography, which are abundant, we can evaluate the extent of this activity. Firstly, observations of still existing archaeological structures by XIXth and XXth scholars, mining engineers and local erudites from scientific societies allowed localizing old mines and confirming the existence of workshops through the presence of slags (Dubuisson 1828, Blavier and Lorieux 1836, Durocher 1857, Davy 1897, Kerforne 1903). Secondly, the BRGM (French Geological Survey) realized, from the 60s to the end of the 90s, a large survey of AM mineral resources (Lulzac, 2012). In total, more than 61000 samples were collected on a surface of 50000 Km2 (Fig.1). Some very rich areas have been mined at modern period (La Villeder, Piriac, Nozay-Abbaretz, Saint-Renan, Plougasnou, Montbelleux). Many of them were previously excavated. Other tin poorer areas seem to be mining free, although recent archaeological surveys sometimes indicate the presence of ancient mining works. We will now focus on some particularly specific mining contexts and associated workshops: (1) marine placers, (2) mines in rocks, (3) alluvial placers.

2.2. Rock mines Rock mines are rare in the AM because mineralized veins outcrop rarely. At Montbelleux, close to cassiterite veins, pits have been observed (Kerforne 1926). However, without excavations, it is impossible to determine their origins, metallurgical pits or granitic sand quarries. At La Villeder (Fig. 1), modern mining between 1834 and 1908, operated by digging large trenches, mine shafts and galleries (Lodin 1884). Most of old mining remains have been destroyed at this

2.1.  Marine deposits The sea front of Piriac (Fig. 1) has been mined from 1830 to 1929, during several periods interrupted by breaks (Lulzac 2012). Ancient mining 94

Villeder (Fig. 1), stream works have been identified, exactly where slags have been found by Lodin (1884). Since then, this area has been totally transformed by modern mining. To the east, in the Montaigu massif (Bonnici and Heinry 1969) have found slags in the stream sediments (not preserved). Around Langonnet and CléguerecQuistinic localities (Gautsch and Lulzac 1962), a lot of old works more or less reworked were observed in stream sediments (Kermadec valley and at Kerbic location). Some small pits filled by washed clay and sand were found during modern mining of Pontigou in 1975, and at Hernec and Crémenec, slags were found but not preserved. At Limerzel (Fig. 1), in both Kerdoret and Le Temple valleys, charcoal and slags were found in reworked layers (Lulzac 1970). By 14C, the first site was dated to the very end of the Second Iron Age (92 cal BP - 20 cal AD) whereas the second was dated to Middle Ages (950 - 1290 cal AD) (Giot 1970). In both cases, no evidence was visible on surface, due to a 1m thick clay alluvium. At Le temple site, slags were sampled and analyzed (Mahé-Le Carlier and al. 2001). The size of slags varies from 1-2 mm to 1 cm. They correspond to partially crystallized slags (Nb-Ta-Fe oxides, potassium feldspar and leucite) with a tin-rich vitreous matrix. They contain also desegregated cassiterite and numerous tin prills. At Nozay-Abbaretz, stream sediments close to linear rock mine were mined. Some pieces of wood found in situ allowed dating the site to Early Middle Ages (567–969 cal AD) (Giot 1970). In northern Finistère, the modern mining in Lanmeur area, worked between 1966 and 1974, uncovered old works invisible from the surface: at Kergueff, some pits were filled by washed clay and charcoal, whereas at Mesquéau, slags were found but not preserved (Lulzac 1967). Finally, Saint-Renan area (Fig. 1) contains a huge alluvial potential. At the beginning of the modern mining of the principal stream placer (1960 - 1974), a 30 m wide and 8-9 m deep trench was discovered in the Vouden valley. It

period. Today, only a large trench, deep of 5 to 10 m, known as «Tranchée des Anciens» is preserved. However, before the modern mining, De Limur (1878) observed three 100 m long pits, with large piles of broken rocks as well as broken slags containing metallic tin grains in vicinity (Simonin 1866). Some roman remains were also observed such as fragments of tiles and potteries and a water pipe for washing. Since the closure of the mine, the forest developed so that it is difficult to differentiate old and modern piles and to localize slag piles during surveys. Finally, it is not sure that slag piles still exist, because of the potential reuse of these ones as an ore. The Abbaretz-Nozay mine, the most important mine in the AM, is better documented. During modern mining, field observations were made (Pitre de Lisle du Dreneuc 1880; Kerviler 1882, Leroux 1912) and artifacts were collected (Champaud 1957). All of the ancient works correspond to a rectilinear succession of open-pits (Fig. 3) following mineralized veins along 10 km (Kerviler 1882). In some very tin-rich locations, excavations were very impressive, with 100 m-long, 30-60 m-wide x 2-6 m-deep openings which were separated by less wide trenches. Davy (1897) observed slags close to these mining structures and performed chemical analyses which confirmed a relationship with tin metallurgy. Champaud (1957) observed old mining works and noted that rocks were fractured, suggesting a possible use of wedges and iron hammers, ore fire-setting. Gallo-roman and Merovingian ages may be proposed thanks to coins collected during modern mining (Champaud 1955). Today, almost all these remains have disappeared due to the modern mining activity and agricultural works. 2.3. Alluvial deposits Ancient alluvial stream mining was also mentioned by the BRGM surveys. For instance, at La

Figure 3. Sketch of the Abbaretz-Nozay mining line/beamworks, as observed by Pitre de Lisle du Dreneuc (1880).

95

dated was found, only 14C method allows dating the associated activity. Similarly, it is difficult to date ancient works in hard rock mines. Favorable climatic conditions allow the vegetation to recolonize rapidly mined areas. Today, the hard rock tin mines in AM seem more likely to be dated to the Gallo-Roman period. Further investigations and prospecting are required in order to know if these exploitation could have been initiated earlier, during the Iron Age or before.

was filled by alternating layers of clay and sand containing roughly cut pieces of wood. No sample was collected at that time. Also, on the top of the main placer, 2-3 m wide pits were found and old works (without more explanation in the report) were recognized in, sometimes associated with slags (not preserved). In the Délé valley, a 15m wide and 2 m deep trench, filled by clay containing charcoal and slags, was dated to Middle Bronze Age, (1387 - 1084 cal BC, Giot and Lulzac 1998). 120 m downstream, there is another similar trench existed from which slags were sampled and analyzed (Mahé-Le Carlier and al. 2001). These slags have been probably crushed because their size is very small, around 1-2 mm. Two types of slags can be recognized. The first one corresponds to a tin-rich vitreous slag including small tin prills. The second one contains cassiterite grains in the inner part. The cassiterite crystals may be rounded and non-molten or partially desegregated. In the former case, the outer part of the slag consists in tin poor vitreous matrix whereas in the latter, the matrix is tin rich.

3. RECENT ARCHAEOLOGICAL SURVEYS ON WORKSHOPS IN THE ARMORICAN MASSIF: CHARACTERIZATION OF LIMERZEL-KERDORET SITE SLAGS The purpose of this paragraph is to present the first results obtained by field surveys on several sites. These surveys were particularly focused on searching mines where larges trenches were mentioned, or metal/tin workshops, through the presence of slags. In some areas, such as SaintRenan or Lanmeur, these researches are still unsuccessful, but in the Limerzel area, a workshop was found in the Kerdoret valley, as explained below.

2.4. Surveys discussion From our bibliographical study, it is clear that ancient tin mines are numerous in the AM and may be found in various mineralized areas, obviously in the high grade ones, but also in the less important ones. However, it is difficult to discover remains, the majority of the sites corresponding to stream mining covered by a thick layer of alluvium over the centuries, leaving no visible evidence at the surface. In addition, modern mining exploitation also likely destroyed potential older remains. Similarly, the slag stacks described by scholars from the XIXth century were probably reused as ore during modern periods. Finally, it is almost impossible to prove the exploitation of marine placer, as the potential remains would have been rapidly destroyed by the sea. For instance, the Piriac and Pénestin deposits, frequently cited as «ancient mines», do not display any archeological evidence of an ancient exploitation. Therefore, when archeological surveys only provide limited results, a careful examination of geological surveys may provide very interesting clues, especially concerning undercover alluvial mining exploitations. However, these surveys were focused on economic geology issues rather than on archeological ones, and only few sampling, analysis or dating were performed. Despite these drawbacks, our synthesis shows that alluvial exploitations have existed from Bronze Age to Middle Age. Interestingly, all the remains uncovered display the same prospecting characteristics that are layers of wastes with charcoal and slags. As no other material that could be

3.1. The Questembert mineralization and site location

Limerzel-Kerdoret site is located within the Questembert leucogranite (Fig. 4). The tin-rich basin of Tévelo in which is located the Limerzel site, extends over 8 Km. It corresponds to the main stanniferous area of Questembert district. Tin deposits correspond to aplite and pegmatite veins emplaced in micaschists host-rocks. These numerous veins are not exploitable but may correspond to the source of tin placerslocated in the numerous rivers incising the bedrock (Fig. 5). In stream placer, most of cassiterite grains have rounded shapes and are free from any silicated gangue. In a lower proportion, grains may be angular, that either (1) may be associated with gangue minerals like feldspar, tourmaline, ilmenite, rutile from foliated granite in the south of the Questembert massif; (2) may have a aplitic or pegmatitic origin and may be associated to quartz, potassium and sodic feldspar, muscovite and more rarely tourmaline (3) may be associated with quartz from sulfide veins injected along the southern boundary of the Questembert massif. In conclusion, Tévelo basin cassiterites may have various origins, with or without their gangue, and accompanied by heavy minerals like ilmenite, rutile, columbotantalite and monazite. 96

Figure 4. Localisation of Kerdoret and Le Temples valleys and location of some tin placers.

Figure 5. Localisation of Limerzel site in relation to the geological context. a variable aspect (Fig. 7). Some slags have a bubbled texture and a dull aspect due to weathering. Other ones have a homogeneous and vitreous texture, and contain mineral inclusions, particularly quartz. Most of the samples are broken but some pieces correspond to small-sized castings and drops. Marks observed on these last samples correspond to contact with charcoal rather than with soil, and so they result from internal melting. Nineteen slag samples were included in resin blocks, polished for a microscope examination followed by EDS-SEM chemical analyses (Tab. 1). Slags have various microscopic textures, mainly vitreous but some samples are partially crystallized. In all cases, tin prills and rare unmolten cassiterite grains and quartz are present. In vitreous slags, flow texture is visible reflecting a high temperature in the furnace, probably around 1100-1200 °C as proposed by some authors (Tylecote and al. 1989). Metallic tin prills

3.2. The Limerzel Kerdoret site Lulzac (1970) described underneath a clay cover, a layer (Fig. 6) with charcoal, gravel and arenite containing a small proportion of brownish slags and small-sized metallic tin globules (Fig. 7). During a field survey in 2012 this layer was found at the river level below 1.2 m of alluvium, in agreement with Lulzac findings. A charcoal fragment was sampled and dated by 14C, yielding an age ranging from 92 cal BC to 20 cal AD. This age indicates the presence of a tin reduction workshop at the end of the second Iron Age or at the beginning of the roman period. 3.3. Slags examination In addition to charcoal residues, the material obtained from the layer displayed fragments of granite and brownish, mm- to cm-sized slags with

Figure 6. Charcoal layer including tin slags from Limerzel Kerdoret valley.

97

Figure 7. Tin slags from Limerzel Kerdoret valley.

Figure 8.: dendritic crystals of colombo-tantalite (in the left).

Figure 9. Stubby crystals of Ti-Sn rich oxides (in the right).

may be numerous in the matrix, but irregular distributed, with zones of higher concentration. SEM-EDS analyses indicate a pure tin metal since no other chemical element was detected. In partially crystallized slags, the crystal shapes are either skeletal or dendritic, and distributed in the whole vitreous matrix. These crystals may be colombo-tantalite (Fig. 8) which incorporate some chemical elements present in molten slag during crystallization like Nb, Ta, W , Ti and Fe (respectively in % element; Ti=56.9%; Nb=16.5%; Ta=13.0%; Fe=4.6%; W=1.7%). There are also stubby Ti-Sn rich oxides (Fig. 9) including Nb, Ta and W (respectively in % element; Ti=22.8%, Sn=43.2%; Nb=14.8%; Ta=12.6%; W=1.3%). In order to better characterise the charge of the furnace and the nature of the gangue, SEMEDS analysis was carried out on 2 mm-sized

spots in the matrix area of the slags. 2 to 3 analyses were performed per sample, and the mean values are given in table 1. These analyses indicate that the slags are relatively homogeneous, for all elements except Sn which vary from 2.9 to 22%. This strong variation is counterbalanced by limited variation of the other elements. The Si, Al, Na and K contents are in agreement with the presence of some gangue minerals such as potassium and sodic feldspar and perhaps tourmaline. The very high content of Ti (1.5 to 11.8%, mean content 7.5%) suggests also the presence of ilmenite and rutile. Finally, the content of Nb is also significant (0.2 to 2.7%, mean content 1.6%) and may be related to the presence of columbotantalite. In summary, slag compositions can be explained by the presence of gangue minerals in the furnace charge.

Table 1. Tin slags EDS chemical analyses from Limerzel Kerdoret valley

98

4. GENERAL DISCUSSION: A FIRST APPROACH OF TIN METALLURGY HISTORY IN ARMORICAN MASSIF

3.4. Limerzel Kerdoret site discussion Part of the small-sized Limerzel Kerdoret slags display a droplet morphology. Their round shape and the marks on their surfaces suggest that flow may have occurred in the furnace between the charcoal pieces. The small size of the droplets implies that furnaces had also a small size. Unfortunately, no furnace was discovered in the AM which could confirm this hypothesis. Moreover, the presence of cassiterite grains in slags implies that cassiterite was not crushed before being charged in the furnace. The same observation was also done on another Armorican tin slags ie. Bronze Age site of Saint Renan and Middle Ages site of Le Temple Limerzel (MahéLe Carlier et al. 2001). Most of the other slags were partially crushed. Nevertheless, no large tin prills were observed in these slags. Slag crushing seems to be common in tin workshops as observed by Mahé-Le Carlier and al. (2001) for the Bronze Age site of Saint Renan and the Middle Age site of Limerzel Le Temple from the AM. Scholars who surveyed the area during the XIXth century, noted that crushed slags with tin prills were present in stacks close to mines of La Villeder (Simonin 1866) and Abbaretz-Nozay (Davy 1897). In South Africa, slags analyzed from the Rooiberg valley were also crushed (Chirikure et al. 2010). The same is true for Cornwall and Devon protohistoric as well as medieval workshop sites (Malham 2010). It may be suggested that crushing the slags helped to recover metallic tin from prills, but it is likely that the separation of the metal prills from the slag was neither easy nor complete. The high content of Ti in slags implies the presence of numerous heavy minerals like ilmenite and rutile in the furnace charge. Only the use of stream cassiterite can lead to such concentration, as proposed by Chirikure and al. (2010) in the Rooiberg valley (South Africa). However, we note also high content of Si and Al in slags, which can be explained by the presence of silicate gangue minerals, probably attached to cassiterite grains. This suggests that tin sources are close to the placer as mentioned by geologists (Lulzac 1970). So, the furnace charge was not composed exclusively of cassiterite grains but also of heavy minerals characteristic of placer and of silicate gangue. Tylecote et al. (1989) propose that charge of protohistoric and medieval furnaces at Caerloggas (south west of England) was composed of 65% of cassiterite. This value seems to be relevant for the Limerzel Kerdoret site. Indeed, Caerloggas slag compositions (Tylecote et al. 1989) are very close to slags analyzed in our study. The rest of the furnace charge was composed by minerals present in placer. No flux was added to improve the process.

In Western Europe, first low tin content bronze artifacts appear at the beginning of the 3rd millennium BC in Montenegro, in the center of Germany, in the north of Italy and in the north-east of Spain (Fernandez Miranda and al. 1995, Primas 2002). In England and Ireland, the transition between pure copper/arsenical copper and bronze is at the end of 3rd millennium BC (Needham 1996, O’Brien 2004). In France, the first tin bronze artifacts are associated with the Bell-Beaker Culture (Gandois 2009), followed by a larger development in the whole country during the A1 Bronze Age (2300-2000 BC, Gandois 2009), as in England and Ireland. Nevertheless, a higher concentration of bronze artifacts is observed in the Western part of the country, for instance in the numerous tumuli preferentially found in the north of Finistère (Briard 1965). Is it possible to explain this richness by the proximity of the large tin deposits of Saint Renan, or alternatively by a cultural proximity with England? Anyway, no evidence of such early tin exploitation in Brittany has yet been documented. During the Bronze Age period, bronze production in north-west France was extremely significant, as attested by the numerous metallic hoards found (Boulud and Fily 2009). Was tin produced locally or imported from neighboring countries, particularly from the British Islands? Large scale tin exploitation is attested in Cornwall (Tylecote 1986) and in Iberian Peninsula (Meredith 1998). These regions are classically considered as the proximal sources of tin for Western Europe during Bronze Age (Pliny the Elder, Strabo). Moreover, the discovery of a shipwreck containing copper and tin ingots on the Devon coast (Roberts and Veysey 2011) clearly shows that metal crossed the Channel. However, the tin deposits of Brittany have also been used, with for instance the Saint Renan workshop (Middle to Late Bronze Age, Giot and Lulzac 1998). Nevertheless, it is not yet possible to determine if the tin production in the AM was sufficient to supply the local bronze production during this period. Lulzac (2012) estimates that about 15 t of cassiterite was produced in the Saint Renan district. But as artifacts dated to Iron Age and Gallo-roman period were also collected on this site, (Giot et Lulzac 1998) suggest that this site could also have been used after the Bronze Age. The chronology of alluvial sites being difficult to establish, it is unlikely to answer these question from field surveys only. At the end of the Bronze Age, lead starts to be added in the alloys. This process is increasingly used during the first Iron Age, characterized by 99

the huge production of famous Armorican type socketed axes (Briard 1965) dated to Hallstatt D1-D2 (Aranda et al. 2013). Briard (1965) suggested that this lead add could reflect a replacement of tin, whose resources in Brittany would have been almost depleted by the extensive Bronze Age exploitation. However, as demonstrated by our bibliographic synthesis, it is unlikely that the AM could have been short of tin for bronze production. In addition, metal analyses of Armorican type socketed axes (Aranda and al. 2013) indicate that lead was not added as a replacement of tin, but rather to counterbalance a lack of copper, as the AM is poor in copper resources. Tin and copper resources being unevenly distributed in Europe, long distance trade routes for ore, ingots and artifacts have been established since the Bronze Age from producing regions to consuming regions. This is attested by antic writers such as Pliny the Elder or Strabo, who suggest that Phoenicians could have reached Atlantic coasts of France and England. This claim is disputed by many modern historians (argument echoed by Routhier 1999), who think that Phoenicians have not passed Spain and that it is local populations from the NW of Europe which came in contact with the Phoenicians. During the Iron Age, the Veneti, a Gallic tribe settled along the southern coast of Brittany, controlled the maritime trade between Great Britain and south Brittany (Mairecolas and Pailler 2010). Ancient authors cite namely SW of England and NE of Spain as tin producing regions. Tin from England took several possible routes toward the Mediterranean area. Ships may follow the coast of Brittany and then Spain, or disembark the cargo at the mouth of Loire or Garonne rivers, to finally use these rivers to cross France. Alternatively, ships may cross the channel and unload their cargo on the northern coast of Brittany. Then, the cargo was transported across the Brittany peninsula to the Loire river (Mairecolas and Pailler 2010). Anyways, it is clear that the Gauls of Armorica were perfectly aware of tin and its production. Why would they have not produced it from the Armorican deposits? If no ancient text explicitly mentions Armorica as a producing region, archeological surveys today point out remains, such as those of the Limerzel-Kerdoret workshop, dated to the end of the Iron Age. Concerning the Gallo-Roman period, there is today no evidence of exploitation of alluvial deposits, although Gallo-Roman artifacts may have been found in their vicinity. Most of the mining activity was focused on open-pit mines. This is the case for instance, of the large, 10-km long Abbaretz–Nozay mine, or of the Villeder mine. The Abbaretz–Nozay open-pit mine was dated

thanks to coins found during modern mining works, with age ranging from 14 BC (bronze Augustus as) to 263 AD (Postumus billon, Champaud 1957). In this area, tin-bearing rocks are particularly fractured. Therefore, the tin exploitation was probably relatively easy, thanks to wet wood wedge, iron picks and hammers, as well as fire-setting. Reduction and treatment workshops were likely located in vicinity because stacks of slags have been noted by local erudites during the XIXth century. During this period, as it is the case for the other metals, the estimated tin production from the mines is significantly larger than for the alluvial placers, with about 1200 t of cassiterite produced in Abbaretz-Nozay place (Lulzac 2012). Assuming a 50% cassiterite-to-metal yield in agreement with experimental data (Timberlake 1994, Tylecote et al. 1989), about 600t of metal tin could have been produced in that area. This amount is not exceptional, especially compared to the potential production from SW of England or NW of Spain, however it is not negligible. It is likely that all tin deposits in the Roman Empire were exploited simultaneously, as demonstrated by the ingot typology found in shipwrecks. For instance, handbag-sized ingots found in the PortVendres II shipwreck (Mediterranean sea) probably came from Spain (Colls et al. 1975) whereas the rectangular, plano-convex ones found in the northern coast of Brittany (unpublished data from a shipwreck excavated in august 2015 in Roscoff by DRASSM) may come from SW of England. Why was the Abbaretz-Nozay mine abandoned by the Romans? This is probably in relation with a global unrest of the region at the end of the IIIrd century AD, associated with a decrease of large scale trade exchanges. However, the metallurgical activity has not stopped. For instance, iron reduction workshops are known at this period, but with significantly lower production volumes and probably for a local use. Is the same true for tin? Today, we have no evidence of tin production workshop during the Late Roman Empire. The next period for which a tin production may be found is the VI and VIIth centuries AD. In the vicinity of the Abbaretz-Nozay open pits, wood poles have been uncovered on an alluvial exploitation and have been dated to early medieval age (Giot 1970). No slag was found, but mineralized layers have been reworked. The coins found in that area (2 Merovingian gold coins and a bronze coin of Mauricius Tibere from Carthage in 585, Champaud 1957) indicate that its exploitation lasted for several centuries after the departure of the Roman. However, it is difficult to determine its use, because a generalized use of tin occurred only later, during the Middle

100

Ages. Nevertheless, according to Leniaud and al. (1991), the Nantes cathedral, consecrated in 580, would have had a roof totally covered with tin. As Nantes is at about 50 km from Abbaretz-Nozay, it is possible that this mining district would be a source of this tin. The last known period of activity is the Middle Age, with the Limerzel-Le Temple reduction workshop (950-1290 cal AD). Here, a placer was exploited. The particularity of the analyzed samples comes from the presence of a lead plate in which cassiterite grains are molten (Mahé-Le Carlier and al. 2001). This indicates that (1) on the exploitation site, a lead-tin alloy was produced, and (2) this alloy was made by fusion and direct reduction of cassiterite in the lead. However, the presence of tin reduction slags also indicates that tin metal was also produced. This period was marked by a wide development of tin for different purposes, particularly for the production of tableware for wealthy families. The tin-lead alloy was also used for the welding and brazing of tin artifacts, and for lead sarcophagus. Finally, a common use was the production of banners for pilgrimage. A workshop dated to the XIVth century found in the Mont Saint Michel very well exemplifies the extensive use of both metals (Labaune-Jean 2010). Nevertheless, it cannot be demonstrated that the Armorican exploitations could have supplied this workshop. Indeed, at this period, the tin production from Erzgebirge is highly increasing, for which texts from Agricola in 1556 mention alluvial as well as hard rock mines (France-Lanord 1992). The same is true for Cornwall with alluvial exploitation since the XIIIth century, followed by a development of hard rock mines since the XVIth century (Davis 1935, Tylecote et al. 1989). It seems that the production of these different regions was really significant during the Middle Ages, and that could have led to the establishment of a network of metal distribution in the whole Europe. Finally, during this period, conflicts between lords were frequent in Brittany, and particularly the War of the Breton Succession (1341-1361). This permanent unrest likely prevented the development of tin exploitation, up to its abandonment. It is only during the XVIIIth century that tin was rediscovered in the AM.

results from geological prospecting and drilling, it is clear that many mineralized areas have been exploited. Unfortunately, the limited amount of remains and the lack of exhaustive archaeological prospecting do not allow dating most of these exploitations. Nevertheless, the available ages show that the region has been exploited from the Bronze Age to the end of the Late Middle Ages. From a metallogenic point of view, three kinds of deposits and associated exploitations may be considered in the AM: (1) the numerous fluvial placers that can explain why most of the known exploitations are of alluvial type. (2) lode type deposits, but only two hard rock mines are known, both during the Gallo-roman period; and potentially (3) marine placer, that should be considered only as a possibility as no archaeological evidence of exploitation was found in the field. To complement these results, field prospecting will be continued in order to locate new production workshops. In parallel, a characterization of elementary and lead isotopic signatures of Armorican deposits is currently processed. Comparison between these ore signatures and elementary and isotopic compositions of artifacts and ingots may demonstrate the significance of the AM as a source of tin for several periods.

CONCLUSIONS AND PERSPECTIVES

BOULUD S. and FILY, M. 2009. «Les dépôts métalliques de l’extrême fin du Bronze final en Bretagne: nouvelle évaluation des données à la lumière des découvertes récentes», in  A. Daubigney, P.-Y. Milcent, M. Talon, J. Vital (dir.), «De l’âge du Bronze à l’âge du Fer en France et en Europe occidentale (Xe-VIIe siècle av. J.-C.). La moyenne vallée du Rhône aux âges du Fer», XXX colloque international de l’AFEAF, 284-298.

Although many tin deposits are known in Brittany, this region is generally not considered as an important tin producer, whatever the period. Thanks to the bibliographic synthesis performed in this study, including ancient and recent archeological results, but also invaluable

BIBLIOGRAPHY ABRAHAM P. MORASZ, J. G. 1997. «Occurrence d’une, métallurgie de l’étain en Viadène (Nord Aveyron). Prospections archéologiques et recherche des gisements», Pallas, 46, 219-231. ARANDA, B.; LE CARLIER DE VESLUD C.; MARCIGNY C. and LE BANNIER J. C. 2013: «Le dépôt de haches à douille de type armoricain du Hallstatt D1-D2 de Trelly (Manche). Interprétation à partir des analyses chimiques élémentaires». Bulletin de Société Préhistorique Française, t 110, 105-119. BLAVIER E. and LORIEUX, T. 1836. «Note sur un gîte d’étain nouvellement découvert à la Villeder (Morbihan)». Annales des mines, 3ème série, t VI, p 381. BONNICI, J. P. and HEINRY, C. 1969. «Recherche des gîtes stannifères primaires à partir des indices alluvionnaires trouvés en Mayenne», rapport BRGM 17p.

101

BRIARD, J. 1965: «Les dépôts bretons et l’âge du Bronze atlantique», Travaux du Laboratoire d’Anthropologie préhistorique, Rennes, 353 p.

DUROCHER, J. 1857: «Observations sur les gîtes stannifères de la Bretagne». Comptes Rendus de l’Académie des Sciences. XLV 522-525.

CAUUET, B.; GUILLAUMET, J.-P. and TAMAS C. G. 2004: «Recherches sur les exploitations minières anciennes du Morvan, Bibracte», Centre archéologique européen, 341-357.

FERNÁNDEZ MIRANDA, M.; MONTERO-RUÍZ, I.; and ROVIRA, S. 1995: «Los Primeros Objectos de Bronce en el occidente del Europa». Trabajos de Prehistoria 52, 1, 57-69.

CAUUET, B. 2006: «Anciennes mines d’or et d’étain en Limousin (Archives et ancien cadastre au service de l’Archéologie minière)», Archives en Limousin, n°27, 48-54.

FORBES, R.J. 1964: «Studies in Ancient Technology», volume 9, Leiden, éditions E.J. Brill, 288p.

CHAMPAUD, C. 1955: «Notice sur trois types d’outils gallo-romains retrouvés dans l’exploitation minière d’Abbaretz (Loire Inférieure)», Annales de Bretagne. T LXII, 293-299.

FRANKLIN, A. D.; OLIN, J. S. and WERTIME, T.A. 1997: «The search for Ancient Tin», Seminar held at the Smithsonian Inst. and the National Bureau of Standards, Washington DC, Library of Congress.

— 1957: «L’exploitation ancienne de cassitérite d’Abbaretz-Nozay (Loire-Inférieure). Contribution aux problèmes de l’étain antique», Annales de Bretagne et des Pays de l’Ouest, 46-96.

GANDOIS, H. 2009: «L’adoption du bronze à l’étain en France», mémoire Master II, Paris I.

CHAURIS, L. 1967: «Les gisements d’étain du pays du Léon (Finistère)», Pen ar Bed, 49, 59-68. —  1977: «Ceintures granitiques hercyniennes et minéralisations stanno-wolframifères dans le massif Armoricain», Comptes Rendus de l’Académie des Sciences de Paris, 284, serie D, 1009-1012. —  1981: «Une province stannifère: le Massif Armoricain (France)», Chronique de la recherche minière, 462, 5-42. CHIRIKURE, S.; HEIMANN, R. B.; and KILLICK, D. 2010: «The technology of tin smelting in the Rooiberg Valley, Limpopo Province, South Africa, ca. 1650-1850 CE», J. of Arch. Sci., 37, 1656-1669. COLLS, D.; DOMERGUE, C. and LAUBERHEIMER, F.; LIOU, B. 1975: «Les lingots d’étain de l’épave de Port-Vendres II», Gallia, t 33, 61-94. DAVIES, O. 1935: «Roman mines in Europe», Oxford, 291p. DAVY, L. 1897: «Une ancienne mine d’étain entre Abbaretz et Nozay (Loire inférieure)». Bulletin. Société. Sciences. Naturelles de l’Ouest de la France, VII, 281-296. DE LIMUR, A. 1878: «Note sur les gisements de l’étain en Bretagne au point de vue de son commerce à la période dite du Bronze». Bull. de la Société Polymathique du Morbihan, 124-131. DUBUISSON, F. R. A. 1828: «Sur l’existence de l’oxyde d’étain dans le département de la Loire Inférieure et sur le moyen de la reconnaître». Le Lycée Armoricain, 46-48.

FRANCE LANORD, A. 1992: «Agricola 1556. De Re Metallica», ed. Klopp Thionville, 512p.

GAUTSCH, J. P.; and LULZAC, Y. 1962: «Nouveaux travaux sur les indices de Langonnet (Morbihan)». Rapport BRGM R5039. GERRARD, S. 1996: «The early south-western tin industry : an archaeological view», The Bulletin of the Peak District Mines Historical Society, 13, 2, 67-83. GIOT, P. R. 1970: «Chronique des datations radiocarbone armoricaines». Annales de Bretagne, 77, 1, 155-160. GIOT, P. R. and LULZAC, Y. 1998: «Datation à l’Age du Bronze d’une exploitation de cassitérite dans le Finistère». Bulletin Société Préhistorique de. France t 95, 4, 597-600. HAUSTEIN, M.; GILLIS, C. and PERNICKA, E. 2010: «Tin isotopy. A new method for solving old questions», Archaeometry, 52, 5, 816-832. HERRING, P. 1996: «Mapping the Mines and Streamworks of Bodmin Moor», in «The Archaeology of Mining and Metallurgy in South-West Britain», (s.d.) Newman P., Mining History : Bulletin Peak District Mines Historical Society, Volume 13, n° 2, 64-66. KERFORNE, F. 1926: «La région minéralisée de Montbelleux (I. et V.)» Bulletin. Société. Géologique et Minéralogique de Bretagne, 7, 59-67. — 1903: «Découverte d’un gisement d’étain et de wolfram dans l’Ille et Vilaine». Sci. Méd. Ouest. T 12, 448-449. KERVILER, 1882: «La grande ligne des mardelles gauloises de la Loire Inférieure». Bulletin Archéologique de l’Association Bretonne, 39-7. LABAUNE-JEAN, F. 2010: «Une production d’enseignes de pèlerins au Mont-Saint-Michel». Archéopages

102

n°18, .80-81, http://www.inrap.fr/userdata/c_bloc_ file/6/6905/6905_fichier_actualite18-labaunejean.pdf. LENIAUD, J. M.; BIENVENU, G.; CURIE, P.; DABOUST, V.; ERAUD, D.; GROS, C.; JAMES, F. C.; and RIFFET, O. 1991: Nantes, la Cathédrale – Loire Atlantique, 4-5. LEROUX, A. 1912: «Les buttes de Nozay et d’Abbaretz. Nouvelles hypothèses sur leur origine et leur attribution». Bulletin. Société Archéologique de la Loire Inférieure. t 53, 147-167. LODIN, 1884: «Note sur la constitution des gîtes stannifères de la Villeder (Morbihan)». Bulletin Société Géologique de France 3, 12, 645-666. LULZAC, Y. 1967: «Etude et évaluation des flats stannifères de la région de Plougasnou près Lanmeur (Finistère)». Rapport BRGM, DRMM 67 A4. — 1970: «Les formations stannifères primaires du bassin du Trévelo en Limerzel (Morbihan)», rapport BRGM, 34p. — 2012: «L’étain armoricain. Où le trouve-t-on ? Où l’a-t-on exploité ?» Mines et Carrières, hors-série 196, 5-56. MAHÉ-LE CARLIER, C., LULZAC, Y. and GIOT, P. R. 2001: «Etude des déchets de réduction provenant de deux sites d’exploitation d’étain armoricain de l’âge du Bronze et du Moyen Age», Revue Archéologique de l’Ouest, 18, 45-56. MAIRECOLAS, M. and PAILLER, J. M. 2010: «Sur les «voies de l’étain» dans l’ancien Occident. Quelques jalons». Pallas, 82, 139-167. MALHAM, A. 2010 : «The classification and interpretation of tin smelting remains from south west England», Ph D. Thesis, University of Bradford, 686p. MEREDITH, C. 1998: «An archaeometallurgical survey for ancient tin mines and smelting sites in Spain and Portugal, Mid-Central Western Iberian geographical region, 1990-1995», British Archaeological Report International Series, n.° 714, 205p. MOLOFSKY, L. J.; KILLICK, D.; DUCEA, M. N.; MACOVEI, M.; CHESLEY, J. T.; RUIZ, J.; THIBODEAU, A. and POPESCU, G. C. 2014: «A novel approach to lead isotope provenance studies of tin and bronze: applications to South African, Botswanan and Ro-

manian artefacts», J. of Archaeological Science, 50, 440-450. MULHY, J. D. 1985: «Sources of tin and the beginnings of bronze metallurgy», American Journal of Archaeology, 89, 2, 275-291. NEEDHAM, S. 1996: «Chronology and periodisation in the British Bronze Age». Acta Archaeologica 67, 121-140. NEWMAN, P. 1996: «Recording the tinworks of Dartmoor forest», the Bulletin of the Peak District Mines Historical Society, 13, 2, 143-149. O’BRIEN, W. 2004: «Ross Island. Mining, Metal and Society in Early Ireland». Galway Univ. Press, 802p. PITRE DE LISLE DU DRENEUC 1880: «Dictionnaire Archéologique de la Loire Inférieure», Bulletin de la Société Archéologique de Nantes, XIX, 117-182. PENHALLURICK, R. D. 1986: «Tin in Antiquity», Institute of Metals, London, 261p. PRIMAS, M. 2002: «Early Tin Bronze in Central and Southern Europe». In M. Bartelheim, E. Pernicka & R. Krause. The Beginnings of Metallurgy in the Old World. Eds, Rahden/Westf: Marie Leidorf. 303314. ROBERTS, B. VEYSEY, C. 2011: «Trade places». British Museum Magazine, 44-45. ROUTHIER, P. 1999: «Voyage au monde du métal», coll. «Pour la Science», Belin, Paris, 159p. SIMONIN, L. 1866: «Sur l’ancienne exploitation des mines d’étain de la Bretagne», Comptes Rendus de l’Académie des Sciences, Paris, 346-347. TIMBERLAKE, S. 1994: «An experimental tin smelt at Flag Fen», Hist. Metallurgy, 28, 2, 121-128. TYLECOTE, R. F. 1986: «The Prehistory of Metallurgy in the British Isles», Institute of Metals, London. TYLECOTE, R. F., PHOTOS, E. and EARL, B. 1989: «The Composition of Tin Slags from the South-West of England», World Archaeology, 20 (3), 434-445. YENER, K. A. and OZBAL, H. 1987: «Tin in the Turkish Taurus mountains : the Bolkardag mining district», Antiquity, 61, 220-226.

103

ARCHAEOMETALLURGY OF TIN BRONZE IN THE DEH DUMEN BRONZE AGE GRAVEYARD, SOUTHWEST OF IRAN Omid Oudbashi*, Reza Naseri**

Abstract The recent archaeological excavation in the Deh Dumen Bronze Age graveyard in the Zagros Mountain, southwest of Iran, has led to find various objects including metallic artefacts. Based upon archaeological studies, this graveyard belongs to the Early/Middle Bronze Age (e.g. 3rd Millennium BC). To identify manufacturing process in archaeological metal finds of Deh Dumen, 10 samples from 8 broken vessels were studied and analysed by Scanning Electron Microscopy coupled with Energy Dispersive X-ray Spectroscopy (SEM-EDS) and Metallography methods. The results show that all the vessels were made of Cu-Sn alloy with variable amount of tin. On the other hand, arsenical copper is used to make the metallic core of the base of two vessels. The microscopic observations show that the vessels were shaped by cycles of working and heat treatment operations that are obviously apparent from the microstructure including recrystallized and twinned grains. Significant amount of elongated sulphidic inclusions as well as lead globules are visible in the metallic microstructure. In fact, the results show the use of tin bronze alloy in South-western Iran in the Early/Middle Bronze Age and show the development of bronze metallurgy in the 3rd Millennium BC in the southwest of Iran at the Zagros region.

 * Department of Conservation of Historic Properties, Faculty of Conservation, Art University of Isfahan, P.O.Box: 1744, Isfahan, Iran. ** Department of Archaeology, University of Zabol, Zabol, Iran.

Keywords: Bronze Age, Tin Bronze, Arsenical Copper, Deh Dumen, Metallography, SEMEDS Resumen La reciente excavación arqueológica en la necrópolis de la Edad del Bronce de Deh Dumen, en la montaña de Zagros, suroeste de Irán, ha permitido recuperar diversos objetos, incluyendo objetos de metal. Esta necrópolis se data en la Edad del Bronce Medio temprano (III AC). Para identificar el proceso de fabricación de los objetos metálicos de Deh Dumen, se seleccionaron 10 muestras de 8 vasos rotos. Se analizaron por microscopía electrónica de barrido con energía dispersiva de rayos X (SEM-EDS) y se hicieron metalografias. Los resultados muestran que todos los vasos estaban hechos de aleación de Cu-Sn con cantidad variable de estaño. Por otra parte, el cobre arsenical se utiliza para hacer el núcleo metálico de la base de dos recipientes. Las observaciones microscópicas muestran que los vasos fueron conformados mediante ciclos de trabajo de forja y de tratamiento térmico tal como revela su microestrúctura con granos recristalizados y gemelos, cantidad significativa de inclusiones alargadas de sulfuro, así como glóbulos de plomo. De hecho, los resultados muestran el uso de la aleación de bronce al estaño en el sur-oeste de Irán en la edad de bronce temprana / media durante el III milenio antes de Cristo en la región de Zagros, en el suroeste de Irán. Palabras clave: Edad del Bronce, bronce binario, cobre arsenicado, Deh Dumen, Metalografia, SEM-EDS.

105

INTRODUCTION The Bronze Age begins at the end of the 4th and early 3rd Millennium BC and ends in the middle of the 2nd millennium BC in the Iranian Plateau (Oudbashi et al. 2012). The importance of the Bronze Age in Iran metallurgy has not been extensively studied and thus very little is known about the alloys used to produce metallic artefacts. From the metallurgical point of view, the Bronze Age has an important role in development history of metals in Iran. Especially, because of the transformation from arsenical copper in Chalcolithic/Early Bronze Age (EBA) periods to tin bronze during the third millennium BC of Iran (Oudbashi et al. 2012; Thornton 2009), analysing metal artefacts from Early and Middle Bronze Age (MBA) sites can help us to better understand the metallurgical changes during this period. In fact, this period is the time of emergence of a new alloy that influenced metallurgical activities. Although, tin bronze was identified and used in the beginning of third millennium BC in western and south-western Iran and Mesopotamia, there is no evidence of using this alloy in other parts of Iran until the second millennium BC (Pigott 2004; Thornton 2009, Nezafati 2006). In winter 2013, a rescue archaeological project was conducted at the prehistoric graveyard near Deh Dumen village, about 70 km northwest of the city of Yasuj, the capital of Kohgiluyeh and Boyer-Ahmad province, in southwest of Iran (Fig. 1). These archaeological excavations were carried out as a part of large-scale emergency excavations of various sites at risk due to the dam project over Khersan River. Many metal finds were found among different archaeological objects in the graves. Based on the graves’ form and structure and archaeological findings such as metals and potteries, the site is dated to EBA/MBA of western/ south-western Iran, third millennium BC (Oudbashi et al. 2016). The graves and burial goods of Deh Dumen have many comparisons in other Early/Middle Bronze Age graveyards excavated in Pusht-i Kuh of Luristan, western Iran such as Bani Surmeh graveyard (Haerinck and Overlaet 2006). For studying and analysing some of the metal objects from Deh Dumen Bronze Age graveyard, a research project was established, funded by the research office of the Art University of Isfahan. The identification of alloy’s composition in these samples was the major aim of this research. Thus, in order to provide more analytical results for this period, we studied the composition and microstructure of the metallic artefacts of Deh Dumen site, determining the types of alloys used and microstructural characterization of the metal samples, for identifying the traditions of metalworking in Deh Dumen objects. The aim of this paper is to present the analytical and microscopic results of the study of some recently excavated metal objects from an EBA/MBA graveyard located at Zagros Mountain, southwest of Iran. This metallic objects were discovered in archaeological excavations carried out in the Deh

Dumen graveyard. Although some analytical studies were already performed on this samples (Oudbashi et al. 2016), the current work tries to present additional new data about the metal working and the manufacturing of bronze objects in 3rd Millennium BC, using microanalytical and microscopic methods. MATERIALS AND METHODS To study the composition and manufacturing process of copper alloy objects from the Bronze Age, 8 broken vessels or parts of broken vessels excavated in different graves of Deh Dumen graveyard were examined (Fig. 2). In two vessels, it was possible to take samples from the base of the vessel (D.P.190 and D.P.287). In these samples a large and separated piece of metal was used into the metallic sheet that formed the body of vessels to form the base of Vessel (Fig. 2). These vessels were found broken into pieces and it allowed us to take a sample from each internal metallic piece. Thus, 8 samples were selected from the vessels’ body sheets and two samples were selected from the large base pieces. Finally, the 10 metal samples were selected and prepared from eight metallic artefacts. A small part of each sample was mounted in epoxy resin, ground with abrasive paper (from 180 to 3000 grid) and then polished with 3 and 0.5 micron diamond pastes respectively to prepare cross section. The microstructural observations and semi-quantitative chemical microanalyses were carried out by Scanning Electron Microscopy coupled with Energy Dispersive X-ray spectroscopy (SEM-EDS). The polished cross sections were analysed by Field Emission Scanning Electron microscope (FE-SEM), MIRA3 model, TESCAN coupled with an EDS analyser. The samples were also studied by optical microscopy to characterize the microstructure and manufacturing processes before and after etching with alcoholic FeCl3 solution (Scott 1991). The metallographic observations were performed with a microscope POL/POLK model manufactured by Alltion.

106

Figure 1.  Map of Iran showing location of Deh Dumen site and other archaeological sites described in the paper.

Cu

Sn

Pb

As

Ni

Zn

P

S

D.P.-Non

82.66

14.21

1.01

0.79

0.40

0.64

0.09

0.19

D.P.-102

79.39

16.27

3.29

0.00

0.32

0.73

0.00

0.00

D.P.-144

82.20

14.34

1.59

0.50

0.39

0.72

0.11

0.16

D.P.-150

82.36

14.43

1.25

0.71

0.27

0.68

0.15

0.14

D.P.-190/1

92.60

0.56

1.47

4.21

0.29

0.59

0.10

0.19

D.P.-190

82.63

14.78

0.76

0.41

0.35

0.95

0.00

0.12

D.P.-222

81.65

16.13

2.22

0.00

0.00

0.00

0.00

0.00

D.P.-283

85.41

12.54

1.40

0.64

0.00

0.00

0.00

0.00

D.P.-287

79.35

17.70

0.61

0.92

0.26

0.75

0.18

0.23

D.P.-287/1

95.10

0.68

1.99

2.23

0.00

0.00

0.00

0.00

Table 1. Results of SEM-EDS analysis on alloy compositin in ten metallic samples (wt %)

RESULTS AND DISCUSSION Alloy Composition The results of the semi-quantitative SEM-EDS analysis on 10 metallic samples from Deh Dumen are presented in Tab. 1. The chemical analysis indicates that the body of the 8 metallic vessels were made of copper-tin alloy with different amounts of Sn (Tab. 1). The Sn amount varies between 12.54% and 17.70% in these samples. In samples of the internal metallic piece of two vessels’ base (samples D.P.190/1 and D.P.287/1), tin was detected only as minor element (0.56 and 0.68 wt%) while arsenic is measured as the main alloying component at 4.21 and 2.23 wt% respectively. Arsenic is present in eight of the bronze vessels’ body as minor/trace element only. Fig. 3 shows the scatter plot of Sn versus As. It proves that the bodies were made of tin bronze while the bases were manufactured from arsenical copper. Although, tin content is determined in higher amount in the semi-quantitative analysis results, but it seems that tin amount should be lower than these measured amounts as are measured by quantitative analysis in another report (Oudbashi et al. 2016). Lead is detected in significant amount (more than 1%) while zinc and nickel are measured in minor levels only. Phosphorus and sulphur are measured in minor/trace content in some samples. Based on the analytical results, we could conclude that the bronze vessels from EBA and MBA site of Deh Dumen were made of variable tin containing bronze alloy. Tin bronze artefacts with considerable tin amount are an important find in the archaeometallurgy of the Bronze Age of Iran. The first evidences of tin bronze production in the Iranian Plateau date back to the end of 4th and the 3rd millennium BC. Some bronze objects were found among various copper and arsenical copper objects discovered in EBA and MBA graveyards in

western Iran. The results of analysis of 58 metallic objects from 7 archaeological sites dated to the end of 4th and the 3rd Millennium BC from Luristan region (western Iran) showed that they are made of different copper alloys: 19 objects are tin bronze, 32 objects are arsenical coppers and 9 samples are made of unalloyed pure copper (Begemann et al. 2008). On the other hand, the results of chemical analysis of 41 objects from the EBA and MBA phases of Kalleh Nisar graveyard (phases A-I, A-II and C) dated to the end of 4th and 3rd millennium BC indicated that 24 objects were made of tin bronze. 5 of the bronze objects belong to A-I phase of Kalleh Nisar that could be considered as the earliest observed tin bronze objects in the Iranian Plateau (Pigott 2004; Fleming et al. 2005; Nezafati et al. 2006). Also, 30 objects belonging to EBA and MBA (3300-1900 BC) of Godin Tappeh, western Iran were analysed and the results show that only 5 objects were made of tin bronze while 23 samples are arsenical coppers and 2 objects are manufactured from unalloyed copper (Frame 2010). Other examples of tin bronze production in the third millennium BC in the Iranian Plateau are found in Susa, Giyan, Sialk and Tappeh Yahya (Moorey 1982; Nezafati 2006; Ghirshman 1938; Thornton and Lamberg-Karlovsky 2004). The absence among these samples of a significant amount of arsenic in the composition of copper/bronze objects is also indicative. Results of different studies on prehistoric copper metallurgy proved that the Chalcolithic and EBA/MBA copper alloy objects from Iranian Plateau have significant amount of arsenic while the Middle/ Late Bronze Age (LBA) as well as the Iron Age are mainly tin bronzes usually contains arsenic as minor or trace element. Interesting examples of the transition from arsenical copper to tin bronze may be observed in copper metallurgy in western Iran where the manufacturing of different metals with arsenical copper in Chalcolithic was changed to low tin bronze production

107

Figure 2. Eight broken metallic vessels from Deh Dumen; ten samples were selected from these objects including eight samples from vessels’ body as well as two samples from internal pieces of vessels’ base (D.P.190/1 and D.P.287/1).

108

beside arsenical copper in EBA and MBA and continued with tin Bronze objects in the Bronze Age and as well as the Iron Age, some examples from different sites from Luristan and Godin Tappeh (Frame 2010; Begemann et al. 2008). For example, in Godin Tappeh the EBA and MBA copper objects contain 1-4 wt% of arsenic while the LBA objects have less than 2 percent of As (Frame 2010). As noted above, analysis of different EBA and MBA metallic objects from western Iran showed that the tin bronze objects have low amount of As while many of low tin or non-bronze objects contain significant amount of arsenic variable up to 10 percent in weight (Fleming et al. 2005; Begemann et al. 2008). It obviously shows that during the third millennium BC, a passing stage from arsenical copper to tin bronze metallurgy occurred. For this reason, arsenical copper objects are observed beside tin bronzes in analysis of metallic objects from the same period in these sites. The Deh Dumen metal objects could be considered as the stage of transition from arsenical copper to tin bronze in the Iranian Plateau. This is visible from the use of arsenical copper and tin bronze in the production of the two bronze vessels. Considering the limited amount of metal finds from Deh Dumen and some limitations of sampling intact objects, the conclusions from the present data about arsenical copper and bronze alloy in this site should be limited to the examined samples and more analyses are needs for better clarifying the technological picture. Metallography and microstructure The cross section of the samples were observed and examined by optical and electron microscopy. The microstructure of the samples in OM micrographs includes a single-phase bronze matrix in which many elongated or partially round Cu-S inclusions are visible. Also, some corrosion is visible near the surface areas (Fig. 4).

Figure 3. Two base pieces (white circles) are made of copper arsenic alloy while eight bodies (black circles) are manufactured by tin bronze.

Observation of microstructure of the Deh Dumen samples in SEM-BSE show many elongated dark inclusions scattered in the single-phase bronze alloy matrix. In many cases, these dark inclusions were elongated in the length direction of the cross sections. EDS analysis on one of dark elongated inclusions in sample D.P.190/1 (Fig. 4 and Tab. 2, analysis A) indicated that copper is measured as 54.27 wt% and sulphur is detected as 18.43 wt%. Other main elements are lead and iron. The analytical results show that it is made of copper sulphide compounds with significant amounts of iron. According to the composition, it is obvious that the dark elongated inclusions are copper sulphide phases with some impurities. Further, some bright globules also are visible in the high magnification in the alloy matrix in SEM-BSE micrographs. To identify their chemical composition and nature, two bright globules from two samples, (D.P.190/1 and D.P.287/1) were analysed by SEM-EDS. The analysis of these globules shows that they are Pb-rich phases with circa 70-80% of lead (Fig. 5, Tab. 2, analysis B and D). Lead is immiscible in copper and occurs as segregated fine globules scattered in the copper matrix (Scott 1991). In addition, some metallic fine phases are visible in FE-SEM micrographs of sample D.P.287. In the first glance, they may appear as two-phase compounds. Analysis of one of them showed that it is a tin-rich compound with 35% of tin and 55% of copper (Fig. 5, Tab. 2, analysis C). Based on EDS analysis, these are some Cu-Sn, tin-rich phases with about 30% of tin and may be segregated α + δ eutectoid phases that are formed during solidification of the bronze melt and are remained unchanged after probable thermomechanical operation on the original cast microstructure of the bronze (Chase 1994; Scott 1991; Oudbashi and Davami 2014). The microstructure of the samples after etching with alcoholic FeCl3 solution was observed by optical microscope. According to the metallographic observation, it is clear that the microstructure of all vessel s’ bodies consist of worked and annealed alpha solid solution grains containing twinning and slip bands (Fig. 6a-f). It proves that the original cast bronzes were mechanically treated in cycles of cold-working and annealing to shape the thin sheets that formed the bronze vessels. Presence of twin lines as well as slip bands in one sample reveals that the vessel has been worked in cold state at the final stage of its manufacturing process (Caron et al. 2004; Scott 1991; Oudbashi and Davami 2014). Nevertheless, in some cases, the working and heat treatment was not able to remove all segregations occurred during the solidification of the molten alloy and some eutectoid phases may be observed in the microstructure, as shown in the SEM-BSE micrographs (Scott 1991).

109

Cu

Pb

Sn

As

Zn

S

Fe

Ni

P

Se

Sb

O

D.P.-190/1-A

54.27

12.85

0.78

0.42

0.00

18.43

9.70

0.00

0.00

3.55

0.00

0.00

D.P.-190/1-B

14.74

82.72

0.00

0.66

1.29

0.00

0.00

0.49

0.10

0.00

0.00

0.00

D.P.-287-C

57.94

2.10

35.24

1.15

2.15

0.00

0.00

0.85

0.57

0.00

0.00

0.00

D.P.-287/1-D

5.18

72.41

0.95

5.06

0.00

0.00

2.60

0.00

0.00

0.00

1.21

12.59

Table 2. Results of SEM-EDS microanalysis on some fine phases observed in the microstructure of samples (wt %)

Figure 4. OM micrograph of some samples before etch; a) D.P.144; b) D.P.287; c) D.P.190, d) D.P.150. The microstructure of samples show scattered dark inclusions in the bronze matrix as well as some corrosion attacks near the surface of samples.

In fact, the metallographic studies of the body of the bronze vessels show that all 8 samples were shaped by a cycle of cold working and annealing to reach their final shape. This is apparent from the equi-axed and recrystallized grains. The grain size is variable in different samples. A smaller grain size may be due to more times of working and heating cycles. In fact, the amount of thermo-mechanical operation on the bronze pieces leads to different grain size in the final products (Scott 1991). Some variations are evident in the size of the grain in the microstructure of the samples, for example, two different grain size are visible in sample D.P.144 showing dissimilar amounts of working and annealing in different parts of the vessel. Also, the grain size is smaller near the surface of the samples.

Samples D.P.190/1 and D.P.287/1 are made of arsenical copper and have formed the internal metallic piece of two vessels’ base. Their microstructure also shows worked and annealed grains. Nevertheless, some differences are visible in these samples. In sample D.P.190/1, individual grains could be identified and remnant of coring is clearly visible as parallel dark and light bands (Fig. 5g). These bands are called flow lines (Louthan Jr. 1986). This microstructure shows that the coring has been distorted into near parallel bands by hammering that has superimposed recrystallized grains, some of which contain annealing twins. It proves that the metal piece has been worked and annealed to remove micro-segregation and to shape the metallic piece (Philip et al. 2003). It had led to produce

110

twinned grains but had left the distorted remnant coring due to not enough annealing in time or temperature to remove all coring in the microstructure. Dungworth (2013) has done an experimental study to identify the effect of annealing time and temperature for the production of different microstructure in bronze alloys from no recrystallized to fully homogenous. He states that there are a range of times and temperatures in which a cored and recrystallized microstructure can be produced. In fact, the presence of cored and recrystallized microstructure together in this arsenical copper sample shows that it was probably annealed between 500 and 700 °C in a short time (Dungworth 2013; Louthan Jr. 1986). Sample D.P.287/1 shows also a superimposed grain structure including worked and recrystallized grains with some remnant of coring (flow lines), but the amount of retained coring in this samples seem to be less than the other arsenical copper sample (Fig. 6h). The metallic pieces that were inserted into the base of the vessels were cast from arsenical copper and then were hammered and partially annealed to reach their final shape. Then the bronze sheets of the vessels bodies were

placed upon them, hammered and annealed to produce a thinned body vessel with a heavy thick base. Although the vessels are broken in pieces, it is clearly evident that the body of the vessels was originally made of a one-piece bronze sheet and there is no evidence of joining operation to produce the vessels’ body. It shows that the Bronze Age metal smiths were able to produce metallic vessels in a high performance level. CONCLUSION A microscopic and micro-analytical study was performed on some recently excavated vessels from the Early/Middle Bronze Age (EBA/ MBA) graveyard of Deh Dumen (south-western Iran) to characterize alloying and manufacturing processes of copper alloys during that time. Based on the results, the vessels’ bodies of Deh Dumen were made of variable Sn-containing tin bronze alloys. These tin bronze objects with considerable tin amount are significant findings in the archaeometallurgical history of the Bronze Age of Iran because of the lack of large amounts

Figure 5.  SEM-BSE micrograph of different fine phases scattered in the matrix and their EDS peaks. The elongated dark inclusions are Cu-S compounds (A) while the fine and bright phases are lead and silver globules (B and D). Analysis C shows an alpha + delta eutectoid.

111

Figure 6. Microstructure of some samples from Deh Dumen after etching in FeCl3 solution. a) D.P.Anon, b) D.P.144, c) D.P.150, d) D.P.190, e) D.P.283, f) D.P.287, g) D.P.190/1 and h) D.P.287/1; The microstructure of vessels’ body are including worked and annealed grains of alpha solid solution with twin line and slip bands (a-f) while the vessels’ base show a different microstructure consisting of remnants of coring superimposed on the worked and annealed grain structure.

112

of identified tin bronze objects from the third millennium BC in this region. Although, the absence among the same samples of a major amount of arsenic in the composition of the bronze bodies is also important, the use of arsenical copper to make some parts of two vessels (the metallic piece of the internal part of the base) is another important discovery. Based on the literature, in many cases the Chalcolithic and EBA copper objects from Iran have significant amounts of arsenic, but it is changed in the MBA and LBA to tin bronzes containing arsenic as minor or trace element only. It is apparently visible from the use of arsenical copper and tin bronze in the manufacture of a single bronze vessel that the Deh Dumen metal objects could be considered as being produced in the stage of transition from application of accidental/deliberate Cu-As to Cu-Sn alloy on the Iranian Plateau during the third millennium BC. Interesting examples of such transition may be observed in western Iran, where manufacturing objects with arsenical copper in Chalcolithic era was changed to low tin bronzes as well as arsenical copper in EBA and MBA in different sites from Luristan and western Iran. The microscopic studies show that all samples (tin bronzes and arsenical coppers) were shaped by thermo-mechanical processes, although it has not been enough to make fully recrystallized or homogenous microstructure in some samples. It can be revealed from the coring as well as eutectoid phases that are observed in the microscopic structure of some samples. The next step of this research project will be to conduct provenance analysis on the samples and to do some comparative studies and characterization of the probable origin of the tin bronze metallurgy in the third millennium BC in western/ south-western Iran. ACKNOWLEDGEMENTS The authors are thankful to B. Rahmani and S. Nikroo, Razi Applied Science Foundation, Atefeh Shekofteh and Banafsheh Heidarpour, Art University of Isfahan and Media Rahmani, for their valuable helps and comments in archaeological and experimental studies. The archaeological excavation of Deh Dumen was done under authority of Iranian Centre for Archaeological Research (ICAR) and ICHTO office of Kohgiluyeh and Boyer-Ahmad province. The experimental and technical works presented in this paper has been carried out in the framework of the research project No. 9210/5, financed by Research Office of Art University of Isfahan, Iran in 2013-2014.

REFERENCES BEGEMANN, F.; HAERINCK, E.; OVERLAET, B.; SCHMITT-STRECKER, S. and TALLON, F. 2008: «An Archaeo-Metallurgical Study of the Early and Middle Bronze Age in Luristan, Iran». Iranica Antiqua XLIII: 1-66. CARON, R. N.; BARTH, R. G. and TYLER, D. E. 2004: «Metallography and Microstructures of Copper and its Alloys», In ASM handbook 9: Metallography and Microstructures. ASM International: 775-788. CHASE, W. T. 1994: «Chinese Bronzes: Casting, Finishing, Patination, and Corrosion». In D. A. Scott; J. Podany and B. Considine (eds.): Ancient and Historic Metals: Conservation and scientific research: proceedings of a symposium organized by the J. Paul Getty Museum and the Getty Conservation Institute, (Los Angeles 1991). Getty Conservation Institute. Los Angeles: 85-117. DUNGWORTH, D. 2013: «An experimental study of some early copper smithing techniques». In D. Dungworth and R. C. P. Doonan (eds.). Accidental and Experimental Archaeometallurgy. London: 149152. FLEMING, S. J.; PIGOTT, V. C.; SWANN, C. P. and NASH, S. K. 2005: «Bronze in Luristan: Preliminary Analytical Evidence from Copper/bronze Artifacts Excavated by the Belgian Mission in Iran». Iranica Antiqua XL: 35-64. FRAME, L. 2010: «Metallurgical investigations at Godin Tepe, Iran, Part I: the metal finds». Journal of Archaeological Science 37: 1700-1715. GHIRSHMAN, R. 1938: Fouilles de Sialk. Librairie Orientaliste Paul Guenthner. Paris. HAERINCK, E. and OVERLAET, B. 2006: Luristan Excavation Documents Vol. VI: Bani Surmah: An Early Bronze Age Graveyard in Pusht-i Kuh, Luristan. Acta Iranica 43. Peeters Publishers. Leuven. LOUTHAN Jr. M. R. 1986: «Optical Metallography». In R. E. Whan (ed.): ASM Handbook 10: Materials Characterizations: 299-308. MOOREY, P. R. S. 1982: «Archaeology and Pre-Achaemenid Metalworking in Iran: A Fifteen Year Retrospective». Iran 20: 81-101. NEZAFATI, N. 2006: «Au-Sn-W-Cu-Mineralization in the Astaneh-Sarband Area, West Central Iran, including a comparison of the ores with ancient bronze artifacts from Western Asia». PhD Dissertation. Der Geowissenschaftlichen Fakultät, Der EberhardKarls-Universität Tübingen. German. Unpublished. NEZAFATI, N.; PERNICKA, E. and MOMENZADEH, M. 2006: «Ancient Tin: Old Question and a New Answer». Antiquity 80: 308.

113

OUDBASHI, O.; NASERI, R. and MALEKZADEH, M. 2016: «Technical Studies on the Bronze Age Metal Artefacts from the Graveyard of Deh Dumen, South-Western Iran (Third Millennium BC)». Archaeometry 58: 947-965.

PHILIP, G.; CLOGG, P. W., DUNGWORTH, D. and STOS, S. 2003: Copper Metallurgy in the Jordan Valley from the Third to the First Millennia BC: Chemical, Metallographic and Lead Isotope Analyses of Artefacts from Pella». Levant 35: 71-100.

OUDBASHI, O. and DAVAMI, P. 2014: «Metallography and microstructure interpretation of some archaeological tin bronze vessels from Iran». Materials Characterization 97: 74-82.

SCOTT, D. A. 1991: Metallography and Microstructure of Ancient and Historic Metals. Getty Conservation Institute. Los Angeles.

OUDBASHI, O.; EMAMI S. M. and DAVAMI P. 2012: «Bronze in Archaeology: A Review of the Archaeometallurgy of Bronze in Ancient Iran». In L. Collini (ed.): Copper Alloys-Early Applications and Current Performance-Enhancing Processes. InTech Open Access Publication Rijeka: 153-178. PIGOTT, V. C. 2004: «On the Importance of Iran in the Study of Prehistoric Copper-Base Metallurgy». In T. Stöllner; R. Slotta and Vatandoust A. (eds.): Persia’s Ancient Splendour: Mining, Handicraft and Archaeology. Deutsches Bergbau-Museum. Bochum: 28-43.

THORNTON, C. P. 2009: «The Emergence of Complex Metallurgy on the Iranian Plateau: Escaping the Levantine Paradigm». Journal of World Prehistory 22: 301-327. THORNTON, C. P. and LAMBERG-KARLOVSKY, C. C. 2004: «Tappeh Yahya and the Prehistoric Metallurgy of South-Eastern Iran». In T. Stöllner; R. Slotta and Vatandoust A. (eds.): Persia’s Ancient Splendour: Mining, Handicraft and Archaeology. Deutsches Bergbau- Museum. Bochum: 264-273.

114

COPPER PRODUCTION IN EARLY BRONZE AGE THASSOS: NEW FINDS AND EXPERIMENTAL SIMULATION IN THE CONTEXT OF CONTEMPORANEOUS AEGEAN METALLURGICAL PRACTICES Nerantzis Nerantzis*, Yannis Bassiakos**, Myrto Georgakopoulou***, Elena Filippaki**, Georgios Mastrotheodoros**

Abstract Evidence for early copper production on Thassos dating to the Early Bronze Age derives from Limenaria and Aghios Antonios. Both sites are situated at the SW part of the island, where most of the Thassian Cu-ore outcrops also occur while the local treatment of such ores is corroborated by the analysed smelting slags and other relevant archaeometallurgical residues. A recent experimental simulation was carried out in order to approach questions regarding copper ore reduction in a northern Aegean context. For the duration of the smelt a temperature of 1220 °C in average was maintained for about 90 minutes during which period large pieces of furnace slag formed near the tuyère openings. Analysis of the experimental slag has shown significant microstructural similarities with the archaeological slag, although the chemical and mineralogical composition of the major phases exhibits some noticeable differences. By attempting an assessment of the conditions prevalent during the experimental smelt we pursue an evaluation of the copper reduction performance on EBA Thassos. Furthermore, we initiate a technological comparison for investigating any recognizable links and differences in early copper production between Thassos and the southern Aegean islands.

   * Université de Lille 3 SHS-UMR 8164 HALMA, France.  **  Laboratory of Archaeometry, NCSR «Demokritos» 15310 Aghia Paraskevi Attiki, Greece. *** UCL-Qatar, PO Box 25256, 2nd Floor, Georgetown Building, Hamad bin Khalifa University, Doha, Qatar.

Keywords: copper, metallurgy, smelting, slag, experimental simulation, Aegean, Bronze Age Resumen Las pruebas de producción de cobre durante la Edad del Bronce en Thassos proceden de los yacimientos de Limenaria y Agios Antonios. Ambos se sitáan en el SW de la isla, donde se localizan la mayoría de los recursos minerales de cobre, mientras que el tratamiento local de esos minerales está confirmado por la presencia de escorias de reducción y otros residuos arqueometalúrgicos relevantes. Se realizó una experimentación con el fin de comprobar cuestiones relacionadas con la reducción del cobre en el contexto del área del norte del Egeo. La duración de la reducción a una temperatura media de 1220 °C fue de 90 minutos, periodo en el que se formaron grandes fragmentos de escoria de horno cerca de las toberas. El análisis de las escorias experimentales muestra microestructuras similares con las escorias arqueológicas, aunque la composición química y mineralógica de las fases principales muestra notables diferencias. Al intentar una evaluación de las condiciones prevalentes durante la reducción experimental se busca valorar el rendimiento de la reducción de cobre durante la Edad del Bronce Antiguo en Thassos. Además, iniciamos una comparación tecnológica para identificar las semejanzas y diferencias entre Thassos y las islas del sur del Egeo. Palabras clave: cobre, metalurgia, reducción, escoria, experimentación, Egeo, Edad del Bronce.

115

1. INTRODUCTION The mineral wealth and metallurgical tradition of Thassos, the northernmost Aegean island, is mentioned in mythological accounts and ancient historical sources as a significant strand of the island’s economy. The most well-known myth is connected with the abduction of Europa by Zeus, disguised as a bull, from the shores of Phoenicia to Greece and beyond. Her two brothers Cadmus and Thassos were the leaders of a Phoenician expedition to claim their sister back, and it was Thassos who gave his name to a colony established on the island of Odonis, henceforth known as Thassos. According to Herodotus [Histories VI, 46-47] and Strabo [Geography XIV, 5.28], the Phoenicians settled the island in an attempt to secure access to its gold and silver mines (Marincola and De Selincourt 2003; Roller 2014). They allegedly mined for gold at a location near Ainyra and Kinyra, opposite Samothrace. In particular, Herodotus remarked that a »whole mountain had been turned upside down during the process» [Histories VI, 47] (Marincola and De Selincourt 2003). A joint archaeological-geological project has identified such a site at Klisidi, near Kinyra in eastern Thassos, where gold mining is evident in underground shafts and tunnels dating to the Classical and Hellenistic periods (Weisgerber and Wagner 1988; Koukouli-Chrysanthaki 1988). Despite the suggestions in myths and historical sources that Phoenicians were the first explorers of the North Aegean’s mineral wealth, there is so far no tangible archaeological evidence for their presence on Thassos or the opposite Thracian mainland. In fact the earliest archaeological finds relating to metal extraction on Thassos date back to the Final Neolithic period (ca. 4500-3200 BC) and the beginnings of the Early Bronze Age (ca. 3200-2000 BC) arguing strongly for the development of an early, indigenous metal working tradition on the island and the opposite shores (Papadopoulos 2008; Bassiakos 2012; Nerantzis and Papadopoulos 2013). The physical remains of mining and metallurgy practiced over long periods of time at numerous localities on the island have been the subject of archaeological and geological research since the 1970s. Ancient shafts, mining galleries and substantial slag heaps of historical periods are the major features that have been studied so far (Wagner and Weisgerber 1988; Vavelidis et al. 1988; Muller 2011). It is only quite recently that the beginnings of metallurgy on Thassos have attracted scientific research and the available results point to similarities of the Thassian EBA metallurgy with its contemporary Cycladic tradition (Georgakopoulou et al. 2011; Nerantzis et al. 2016). In the last fifteen years, accumulat-

ing evidence for early copper production based on indigenous cupreous ores has come to light at the SW part of the island, where most of the Thassian Cu-ore outcrops also occur.

Figure 1. Map of Thassos showing the Early Bronze Age coastal settlements discussed in the text.

The available evidence derives mainly from coastal settlements and consists of infrequent lead/silver in addition to more abundant copper production residues. In particular, excavations at Limenaria (Figure 1), yielded evidence for leadsilver extraction of the early 4th millennium BC and copper smelting of the Early Bronze Age I phase (Papadopoulos 2008; Bassiakos 2012). Crucibles and moulds used for the melting and casting of copper objects derive from Skala Sotiros (EBA II) and arsenical-copper contained in smelting slag in the form of prills has been found at Aghios Antonios Potos of the EBA I (Nerantzis and Papadopoulos 2013; Maniatis et al. 2015). Arsenical copper artefacts derive from Agios Ioannis, Limenaria, Aghios Antonios and Skala Sotiros (Nerantzis et al. 2016). Since the provenance of these implements has not been established so far, an ongoing programme of Lead Isotope Analysis and Trace Elements analysis aims to investigate the possibility of local manufacture from Thassian smelted copper ores. 2. STUDY OF THE ARCHAEOMETALLURGICAL EVIDENCE The current, ongoing study of mineral ores, slags, moulds and crucible remains as well as copper alloy artefacts from the above locations aims to provide a synthetic view of Thassian copper production during the Bronze Age. Results from

116

the analysis of copper ores that were sampled from mining locations on SW Thassos are useful in attempting to investigate the possible sources of raw materials used by the smelters. Two types of ores were studied by SEM-EDS, namely limonite and hematite hosting oxidized copper minerals from two locations (Nerantzis et al. 2016, 575). Thus the limonitic ore from Koumaria was found to contain inclusions of Fe-Cu-Cl composition within which Cu reaches 25% and inclusions of Sb, Cu-Zn and Ag. The samples of hematite from Mavrolakas revealed distinct phases rich in Cu in contents up to 55% (Nerantzis et al. 2016, 576). Such ores could have potentially been used for copper smelting after beneficiation but the scarcity of remaining copper reserves and the lack of mining evidence make it difficult to establish the sources of raw minerals that supplied the Early Bronze Age settlements.

Figure 2. SEM/EDX photomicrograph of arsenical-copper dagger from Aghios Antonios, Thassos: almost pure silver micro-incorporations (white) hosted in a matrix of cupreous phases with varying arsenic content (grey to dark-grey).

ern Thassos, and the Acropolis mine to the north (Muller 1979; Kozelj and Muller 1988) and also at Mavrokorfi and Asimotrypes on Mount Pangaeon (arsenopyrites) of the opposite, mainland coast (Eliopoulos and Kilias 2011). The polymetallic deposits at the above mining locations had been exploited during antiquity (Vavelidis et al. 1988; Muller 2011) but the possibility of an earlier phase of mining these ore deposits for their copper contents has not been confirmed and needs further investigation. Slag pieces have been recovered during excavations at two EBA sites, namely Limenaria (Bassiakos 2012; Bassiakos et al. in press) and Aghios Antonios (Nerantzis and Papadopoulos 2013; Nerantzis and Papadopoulos 2016). They were analyzed using optical microscopy and SEM-EDS and found to contain fayalite and magnetite, spheroid copper prills, matte inclusions, a glassy matrix and barite (BaSO4), the latter characteristically associated with the Ba-bearing local mineralisation (Voreadis 1954; Vavelidis 1984). Most analysed matte inclusions contain 65-70% Cu, 20-23% S and around 4% Fe and their presence hints to the possible use of secondary copper minerals, which contain traces of residual sulphides (Nerantzis et al. 2016, 578). The presence of copper prills enclosed in the slag matrix indicates certain losses of copper into the slag (Figure 3). Arsenic is also present in most analysed copper prills in contents between 2 and 4% and close to 3% in matte phases. So far the data point to the smelting primarily secondary oxidised ores with few residual sulphides in a process thought to be roughly similar to that as suggested at Chrysokamino (Bassiakos and Catapotis 2006) in eastern Crete. In an attempt to better understand the process of smelting and produce comparable data, an experimental project was recently undertaken, the results of which are presented here for the first time.

Previous Lead Isotope Analysis (LIA) studies have attempted to establish base-line data for the distinction between various sources of mineral ores utilized for copper smelting. Our recent analysis of Thassian copper-based artefacts from three EBA settlements confirmed the presence of low silver contents (Nerantzis et al. 2015, 578) but their provenance has not been established so far (Figure 2). Yet the Thassian LIA field of silver has been discriminated from those of Laurion and Siphnos based on the previous analyses (Wagner and Weisgerber 1988; Stos-Gale and Gale 1992). To that respect, the detection of minute silver –and scarcer gold according to our newer observations– inclusions in some of the analyzed Thassian copper artefacts points to the use of a silver and even gold bearing copper ore (Nerantzis et al. 2016). Such ores occur at the mines of west117

Figure 3. SEM/EDX photomicrograph of slag from Aghios Antonios, Thassos: laths of fayalite (dark grey), glassy matrix (mid grey), copper prill (white), matte prill on the left (light grey).

3. COPPER SMELTING EXPERIMENTS BASED ON BRONZE AGE AEGEAN EXAMPLES Previous experiments based on the examples from Chrysokamino on Crete and Kythnos in the Cyclades have offered significant data on the smelting performance of furnaces replicating those used across the EBA Aegean. In the case of Chrysokamino, where fragments of perforated clay-furnace walls were found, experimental smelts allowed testing the technological choice of perforations (Pryce et al. 2007). These small holes on the furnace walls were used for exploiting the air draught, easier achieving the required high temperatures and controlling the increase of oxygen to counter the reduction of iron minerals due to reducing conditions within the furnace (Catapotis et al. 2008). Operating with four, hand-driven bellows, in addition to air intake from the perforations, the experimenters were successful in producing free-running slag and reduce/extract most of the metallic copper contained in the ores. Remains of conical-shaft bowl-type furnaces for copper production, either with or without perforations, have been found in the southern Aegean islands of Kythnos, Seriphos, Siphnos and elsewhere (Bassiakos and Philaniotou 2007; Georgakopoulou et al. 2011; Bassiakos et al. 2013). Copper ore reduction in clay-bowl furnaces usually produced a nugget of copper at the furnace bot-

tom. Based on morphological features and analytical studies, slag became free-running and was tapped out of a bowl furnace while a significant amount formed in the furnace bottom, encasing in this way droplets, prills or nuggets of copper. Since most slag recovered from the Bronze Age smelting sites and settlements contain varying amounts of copper, either in the form of minute prills or comprising an elemental ingredient of the slags, it has been suggested that this was due to certain inefficiencies during operation of the bowl-type furnace. In order to address questions of the Thassian pyrometallurgical practices, the hereupon experiment has been designed based on previous experience and attempting to –possibly– simulate the conditions prevalent during copper ore reduction in a northern Aegean context. 4. EXPERIMENTAL COPPER ORE REDUCTION: THE THASSIAN EXAMPLE The experimental smelting of copper ores was conducted during June 2014 at the yard of N.C.S.R. «Demokritos», in the foothills of Hymettus mountain. Our major goal was to reconstruct and operate a furnace based on the Thassian metallurgical example and in particular to produce smelting remains similar to those found at Aghios Antonios on Thassos. Since no

Figure 4. Raw materials used in the experimental replication.

118

archaeological furnace fragments were found so far in any prehistoric Thassian site that could account for the actual shape and dimensions, it was decided to conduct the experiment by using the Cycladic-type conical-shaft furnace, the one without perforations. Excavations have shown that this device represents a rather prevailing metallurgical accessory in other contemporaneous Aegean EBA smelting sites for copper production. For the construction of the furnace stack, a refractory clay paste resembling prehistoric examples was prepared. Earthenware studio clay was used with the addition of coarse quartz (< 0.80 cm grain size) as anti-plastic agent and broken pine needles at a ratio of 2:1:1. Their incorporation is crucial in order to increase the refractory capacity of the clay and its resistance to thermal shock. After kneading the paste manually the chimney superstructure was formed by pressing the clay against the internal walls of an inverted furnace mould in the shape of a truncated cone. The mould was lined with a plastic bag to avoid adhesion. After two days of drying at room tem-

perature the clay stack was removed from the mould and left to dry for another three days in a laboratory fume hood ventilation shaft. The dimensions of the stack were 40cm diameter at the base and 25cm at the top, with a height of 40cm and wall thickness of 2.50-3.00 cm. The hearth was prepared in the open to facilitate the stack by digging a shallow cavity of 10cm depth and 60cm diameter, which was then lined with refractory clay and left to dry for five days. Finally a clay tuyère was formed from the same paste used for the furnace. Its length was 34 cm with an air channel of maximum diameter 5.50 cm and minimum, at the nozzle edge, 2.50 cm. After drying, three holes were opened at the perimeter of the furnace base to allow for setting the clay tuyère and hoses used in air flow supply. A larger hole was opened to function as a tapping arch for any free-running slag to flow out during smelting and smaller holes for natural wind inflow. The raw materials used for the charge were malachite-bearing schist that was crushed and handpicked to gather the heavier and richer in ore particles and crumb cuprite.

Figure 5. Experimental copper ore reduction in progress.

119

Three flux constituents were used to lower the ores’ melting points and allow for liquid slag formation. These were pre-treated magnetite in fine fragments (< 1.0 cm), quartz in the shape of gravel and fine grit calcite (Figure 4). The ratio of fuel to ore, magnetite, quartz and calcite was 4:1:1:0.70:0.66 for each charge. During conduct of the experiment that lasted for about six hours the pretreated charge was added 6 times with numerous intervening stages of charcoal additions. Additional charcoal was necessary to fill the furnace up to its rim in order to maintain high temperatures. The air supply was provided by an electric air compressor directed through hoses towards the three openings near the base of the furnace at a low wind speed of 4 to 5 degrees in the Beaufort scale. One of the hoses was fitted to the clay tuyère and the remaining two were adjusted with the use of tiles towards the openings (Figure 5). Firing of the charcoal filled stack begun in the morning at 11:00 am while the first charge was introduced half an hour later. The temperature within the furnace was monitored and recorded using a thermocouple at 20 minute intervals. In order to acquire accurate readings the probe was placed through the frontal holes above the tapping arch and near the air inflow channels. By the end of the first hour and after having added the third charge the temperature reached 960 °C (Figure 6). Then for the next hour it peaked at 1070 °C with some fluctuations. With the addition of the fifth charge it dropped to 900 °C and 800 °C after two hours of constant operation. Such fluctuation was possibly due to inadequate air supply and it was decided to raise slightly the wind speed and add more charcoal. These actions resulted in raising the temperature to 1235 °C by the time when the sixth charge was introduced. From that point onwards we managed to maintain a stable temperature of 1220 °C, in average, for about 90 minutes during which time charcoal was added at intervals of twenty minutes. This was the crucial period when most of the ore should have reacted and become partially reduced and the forming copper consolidated. No liquid slag had formed, probably due to the shorter than required time of high and stable temperatures but it was determined that large pieces of furnace slag formed near the tuyère openings. After about four hours of constant operation it was decided to stop further charcoal addition and leave the furnace to cool since one of the experiment’s main goals, i.e. to produce furnace slag, had been achieved. In most smelting operations reproducing ancient examples, metallic copper in form of prills is entrapped in the slag. Crushing the slag to reveal any entrapped copper was a common method used by prehistoric metalworkers as sug-

gested by the broken pieces found in smelting sites. It was one of the experiment’s purposes to produce slag with entrapped copper prills which could be then compared with the archaeological examples. The following day the furnace was demolished to reveal a substantial amount of slag that had formed mainly on the front side where the air supply in the blast zone was stronger. It was obvious that part of the ore had not been reduced and remained mixed with slag that was too viscous to become free-running. At the same time small lumps of metallic copper, around 0.5-1.0 cm in length, were observed attached to the slagged conglomerate. Therefore, based on such macroscopic characteristics, it could be concluded that the process of experimental smelting succeded to partially reproduce the prehistoric operation as suggested by the similarities in slag formation. Although it was not possible to produce a lump of copper separated from the slag at the bottom of the furnace, copper coalesced in large prills visible to the naked eye. Selected slag pieces were taken for laboratory analysis, chosen based on the differences in their macroscopic features. Although the slag material, coming out of the experimental process, was rather uniform, samples, including different kinds of inclusions like copper prills, fragments of charcoal or partly reacted charge-ingredients, were analysed in order to get as much information as possible. Phase composition of the selected samples was conducted, either using Optical and Scanning Electron Microscope or X-ray diffraction. In addition, where necessary, chemical analyses, using the Energy Dispersive Spectrometer (EDS) that is attached to the Scanning Electron Microscope, were carried out for individual phases, so as to provide their chemical composition. The preliminary results are presented here.

120

Figure 6. Temperature profile of the experimental smelting.

5. PHASE COMPOSITION OF EXPERIMENTAL SLAG The microstructure of the experimental slags, when first observed under the optical miscroscope, seemingly showed similatities in texture with examples of archaeological slag from Aghios Antonios that have arisen due to similarity in their production processes, as well as the composition of the raw materials used (Figures 7 and 8). However, the bulk chemical analyses of the experimental slags showed in some cases exceedingly high levels of Cu-contents, varying from 5% to 60% (Table 1). This copper content is mainly present as kidney shaped copper globules or as elongated formations representing the most characteristic features of the slags, giving the impression of slagged ore rather than smelting slag (Figure 9). Magnetite bands develop through crystallisation from the raw material either in the form of fairly large crystals or as skeletal chaplets (Figure 10a), some of them incorporating inclusions of copper (Figure 10b), while others are subjected to an eutectoid decay forming laths of delafossite (Hauptmann 2007) (Figure 11). X-ray diffraction analysis identified the phase of cuprospinel (CuFe2O4), interpreting semi-reducing conditions at 900-1000 ºC, during which the solubility of Cu2+ in magnetite rises, leading to the formation of CuFe2O4 Fe3O4solid solution (Gadalla et al. 1966, Hauptmann 2007). In addition, cuprite (Cu2O) identified in all analysed slags by X-ray diffraction, and also observed under the microscope and SEM/EDS analyses (Figure 12), indicates a partial melt under a moderate to weak reducing gas atmosphere (Hauptmann 2007). As far as the silicates are concerned (regarded to be amorphous), their composition ranges widely from calciumrich silicates to copper-rich ones, as shown from the analyses reported in Table 2. Nonetheless, pure SiO2 phases were also identified. A closer examination of the glassy matrix reveals the formation of microcrystalline phases (Figure 13). Despite the small number of samples analysed, the coexistence of magnetite and delafossite, the high copper content in the slags and the existence of cuprospinel phases suggests that the smelting conditions were far from stable. Even though the temperature had reached the 1200  ºC threshold for a period of ca. 90 minutes, resulting in the production of lumps of metallic copper, the smelting process was only partially sucessful. The microstructural phases hint to slightly reducing conditions, adequate to reduce copper oxides to copper metal, but not the trivalent iron oxides to divalent, thus precluding the formation of iron silicates. The heterogeneity of the smelting operation is reflected in the different slag components. The process appears

to have been terminated before the reaction was completed, at least in some parts of the furnace. This is also corroborated by the absence of any liquid slag tapped out of the furnace.

121

Figure 7. Optical microscope image of experimental slag.

Figure 8. Optical microscope image of slag from Aghios Antonios, Thassos.

Figure 9. Kidney shaped globules of copper along with elongated delafossites. (SEM image, backscattered mode).

(a)

(b)

Figure 10. (a) Skeletal magnetite dispersed into the glassy matrix (b) Crystals of magnetites including copper. (SEM image, backscattered mode).

Na2O

MgO

Al2O3

SiO2

K2 O

CaO

Fe2O3

CuO

Total

TES 1

2.27

2.37

7.54

29.37

1.34

6.21

45.77

5.13

100

TES 2

bdl

1.99

2.09

40.42

1.07

26.94

22.88

4.60

100

TES 3

bdl

0.90

1.03

18.04

0.50

6.86

8.17

64.50

100

TES 4

bdl

bdl

2.14

40.19

0.85

10.79

9.23

36.80

100

Table 1. Averages of two bulk analyses of experimental slag samples using EDS. Results are given as oxides expressed as weight percent (wt%). Elements detected below 0.1% are not included (bdl, below detection limit)

Figure 11. Metallic copper surrounded or framed by columnar delafossite, the latter exhibiting also exsolution lamelae (SEM image, backscattered mode).

Figure 12. Copper prills either oxidised into cuprite (the light grey rim), or surrounded by magnetite remnants (the dark grey rim). The background consists of a glassy matrix, neoformations of silicates and lamellae of delafossites (SEM image, backscattered mode).

122

magnetites or cuprospinels

delafossites

silicates MgO

0.65

6.55

Al2O3

bdl

2.17

SiO2

52.18

36.17

1.29

CaO

42.19

25.29

Fe2O3

38.41

Fe2O3

3.15

14.93

4.99

CuO

56.36

CuO

1.83

14.39

100

Total

100

Total

100

100

MgO

1.75

Al2O3

0.55

SiO2

0.46

SiO2

3.39

CaO

0.4

CaO

Fe2O3

92.4

CuO Total

Table 2. Averages of chemical analyses of silicates, magnetites (or cuprospinels) and delafossites using EDS. Results are given as oxides expressed as weight percent (wt%). Elements detected below 0.1% are not included (bdl, below detection limit)

Figure 13. Lamellae of delafossites. In between there is a glassy ground mass hosting microcrystalline copper-bearing composite silicate phases. (SEM image, backscattered mode).

6.  CONCLUDING REMARKS Summing up, the archaeological and analytical data from the study of archaeometallurgical remains of Early Bronze Age Thassos adds to a growing number of relevant evidence from the southern Aegean islands, namely in the Cyclades and Crete (Bassiakos and Philaniotou 2007; Pryce et al. 2007; Georgakopoulou 2007; Georgakopoulou et al. 2011). The type of Thassian polymetallic ores provides hints that this process might allow arsenical-copper production a fact that has been already verified for Chrysokamino on Crete. Tap and furnace bottom slags have been found in EBA Thassos, corroborating the similarities with the corresponding types of slag from the southern Aegean sites. Copper prills, mattes and Cu-containing components detected in the Thassian metallurgical residues

indicate a partially incomplete furnace control; therefore somewhat low efficiency in metal production. This feature applies to most studied Aegean Early Bronze Age copper production centres. The experimental simulation has been effective, in terms of Cu-globules formation and glass matrix existence. However, further microstructural characteristics indicate that the whole process was only partly successful. Further analytical studies for a better identification of the slag components are planned for the future as well as a new series of experimental simulations with manually operated bellows. Moreover, the experiment and the available analytical data indicate that there are numerous differences between the early Thassian and contemporaneous southern Aegean metallurgical practices, when examined in detail. Considering the evidence from the Cyclades, Crete and elsewhere, the picture that starts to emerge is one of numerous production sites across the Aegean utilising similar technological practices for the extraction of copper. The similarities refer to: a) utilization of secondary/oxidised copper ores (i.e. malachite hosting schist, containing also iron-ore, the latter functioning as self-fluxing agent), b) slightly inneficient copper reduction/retrieval causing certain loss of the desirable metal, entrapped in the discarded smelting slags, c) accession and dependance to the indigenous coupreous raw material; this practice is often seen as witness to the nativeness of the raw coupreous materials from the accompanying «tracers» i.e. Ba, Zn, Ag and Au for the Thassian ores. The local geological features characterising each metalliferous area gave rise to variations on the produced alloys i.e arsenical copper with characteristic trace elements (Ba, Zn, Ni, Sb etc.) while differential access to certain «exotic» raw materials, such as tin played an important role. Beyond these broad similarities, however, there are emerging technological differences, such as the use of different types of furnaces, that open up interesting prospects for considering the dis-

123

tribution of shared technological knowledge and practices opening up the potential of metallurgical technology to inform on social interactions in the Early Bronze Age Aegean. Our attempted experimental replication of the Thassian copper smelting technology, admittedly still based on limited data with regards to local furnace typologies, has offered new perspectives on parameters that cannot be directly assessed based on the laboratory analyses; in this case specifically we believe a consideration of the speed and quantity of air inflow, as well as possibly smelting times. Continuous experiments will be required to achieve slag compositions and textures that correspond to those attested in the archaeological samples, and thus to fully understand the practicalities of ancient copper smelting on Thassos and its neighbouring regions. ACKNOWLEDGEMENTS The authors would like to thank the Institute for Aegean Prehistory (INSTAP) for funding the project. Thanks are also due to the Aghios Antonios excavator Dr. Stratis Papadopoulos and the Ephorate of Antiquities of Kavala-Thassos. We also thank Dr. Evangelos Tsakalos, member of the Archaeometry Lab/‘Demokritos’ staff, for his help during conduct of the experimental simulation. BIBLIOGRAPHY

Σίφνο, Πρακτικά Δ΄ Διεθνούς Σιφναϊκού Συμποσίου», Society for Sifnian Studies, Athens, 45-64. BASSIAKOS, Y.; PAPADOPOULOS, S. and NERANTZIS, N. (in press): «Early copper and silver production at prehistoric Limenaria», Thasos: a technological approach, in Bassiakos Y. (ed.) Prehistoric Metal Production in the Aegean: Material Evidence and Analysis, Springer-Verlag, Heidelberg. CATAPOTIS, M.; PRYCE, O. and BASSIAKOS, Y. 2008: «Preliminary Results from an Experimental Study of Perforated Copper-Smelting Shaft Furnaces from Chrysokamino (Eastern Crete)», in I. Tzachili (ed.) Aegean Metallurgy in the Bronze Age: Proceedings of an international symposium held at the University of Crete, Rethymnon, Ta Pragmata, Publications, Athens, 113-122. DOONAN, R. C. P.; DAY, P. M. and DIMOPOULOU-RETHEMIOTAKI, N. 2007: «Lame excuses for emerging complexity in Early Bronze Age Crete: the metallurgical finds from Poros-Katsambas and their context», In P. M. Day and R .C. P. Doonan (eds.) Metallurgy in the Early Bronze Age Aegean, Sheffield Studies in Aegean Archaeology, 7, Oxbow Books, 98-122. DUNGWORTH, D. and DOONAN, R. C. P. 2013: «Accidental and Experimental Archaeometallurgy», HMS Occasional Publication N.º 7, London. ELIOPOULOS, D. G. and KILIAS, S. P. 2011: «MarbleHosted Submicroscopic Gold Mineralization at Asimotrypes Area, Mount Pangeon, Southern Rhodope Core Complex, Greece». Economic Geology 106, 751-780.

BACHMANN, H. G. 1982: The identification of slags from archaeological sites, Occasional Publication No. 6, Institute of Archaeology, London.

GADALLA, A.; FORD, W. F and WHITE, J. 1962/63: «Equilibrium relationships in the system CuOCu2O-SiO2». Trans British Ceramic Soc. 62: 45-66.

BASSIAKOS, Y. 2012: «Πρώιμη παραγωγή χαλκού και αργύρου στα Λιμενάρια Θάσου: Μια τεχνολογική προσέγγιση. Πρακτικά Ημερίδας Δέκα Χρόνια Ανασκαφική Έρευνα στον Προϊστορικό Οικισμό Λιμεναρίων Θάσου», ΥΠΠΟΤ-ΙΗ’ ΕΠΚΑ, Θεσσαλονίκη, 197-225.

GEORGAKOPOULOU, M. 2007: «Metallurgical activities within Early Cycladic settlements: the case of Daskaleio-Kavos», In Day P. M. and Doonan R. C. P. (eds.) Metallurgy of the Early Bronze Age Aegean, Sheffield Studies in Aegean Archaeology 7, Oxford: Oxbow Books, 123-134.

BASSIAKOS, Y. and CATAPOTIS, M. 2006: «Reconstruction of the copper smelting process at Chrysokamino based on the analysis of ore and slag samples», in P.P. Betancourt (ed.) The Chrysokamino metallurgy workshop and its territory, Hesperia Supplement 36, Princeton, 329-353.

GEORGAKOPOULOU, M.; BASSIAKOS, Y. and PHILANIOTOU, O. 2011: «Seriphos surfaces: a study of copper slag heaps and copper sources in the context of Early Bronze Age metal production». Archaeometry 53(1), 123-145.

BASSIAKOS, Y. and PHILANIOTOU, O. 2007: «Early copper production on Kythnos: Archaeological evidence and analytical approaches to the reconstruction of metallurgical process», in P. M. Day and R. C. P. Doonan (eds.) op. cit., 19-56. BASSIAKOS, Y.; GEORGAKOPOULOU, M. and WAGNER, G. A. 2013: «Οι νέες αρχαιομεταλλουργικές μελέτες στη

HAUPTMANN, A. 2007: The archaeo-metallurgy of copper, Evidence from Faynan, Jordan, Berlin, Springer. KOUKOULI-CHRYSANTHAKI, CH. 1988: «Die archäologischen Funde aus den Goldgruben bei Kinyra», in G. A. Wagner and G. Weisgerber (eds.) Antike Edel und Buntmetallgewinnung auf Thasos, Der Anschnitt, Beiheft 6, Bergbau-Museum, Bochum, 173-179.

124

KOZELJ, T. and MULLER, A. 1988: «La mine d’or de l’acropole de Thasos», in G. A. Wagner and G.Weisgerber (eds.) Antike Edel- und Buntmetallgewinnung auf Thasos, Der Anschnitt, Beiheft 6. Bochum: Deutsches Bergbau Museum, Bochum, 180-197. MANIATIS, Y.; NERANTZIS, N. and PAPADOPOULOS, S. 2015: «Radiocarbon dating of Aghios Antonios, Potos and inter-site variability», in south Thassos, Greece, Radiocarbon 57(5), 807-823. MARINCOLA, J. M. and DE SELINCOURT, A. 2003: Herodotus, The Histories. Penguin Classics. MULLER, A. 1979: «Le mine de l’acropole de Thasos, Thasiaca: Paris», Bulletin de Correspondance Hellenique, Suppl. V, 315-344. MULLER, A. 2011: «Les Minerais, le Marbre et le Vin, Aux Sources de la Prospérité Thasienne», Revue de Etudes Grecques, Tome 124, 179-192. NERANTZIS, N. and PAPADOPOULOS, S. 2016 «Copper production during the Early Bronze Age at Aghios Antonios, Potos in Thassos», In E. Photos-Jones in collaboration with Y. Bassiakos, E. Filippaki, A. Hein, I. Karatasios, V. Kilikoglou and E. Kouloumpi (eds.) Proceedings of the 6th Symposium of the Hellenic Society for Archaeometry, BAR S2780, British Archaeological Reports Ltd, 89-94. NERANTZIS, N., BASSIAKOS, Y. and PAPADOPOULOS, S. 2016: «Copper metallurgy of the Early Bronze Age» in Thassos, north Aegean, Journal of Archaeological Science: Reports 7, 574-580. NERANTZIS, N. and PAPADOPOULOS, S. 2013: «Reassessment and new data on the diachronic relationship of Thasos Island with its indigenous metal resources: a review», Journal of Archaeological and Anthropological Sciences, Volume 5, Issue 3, 183-196. PAPADOPOULOS, S. 2008: «Silver and Copper Production Practices in the Prehistoric Settlement at Limenaria, Thassos». in I. Tzachili (ed.) Aegean Metallurgy in the Bronze Age, Proceedings of an International Symposium held at the University of Crete,

Rethymnon, Greece, 19-21/11/2004, Ta Pragmata Publications, Athens, 59-67. PRYCE, T. O.; BASSIAKOS, Y.; CATAPOTIS, M. and DOONAN, R. C. P. 2007: «‘De Cerimoniae’ Technological choices in copper-smelting furnace design at Early Bronze Age Chrysokamino, Crete», Archaeometry 49(3), 543-557. ROLLER, D. W. 2014: The Geography of Strabo: An English Translation with Introduction and Notes, Cambridge University Press. STOS-GALE, Z. and GALE, N. 1992: «Sources of copper used on Thassos in Late Bronze and Early Iron Age», in Ch. Koukouli-Chrysanthaki (ed.) Πρωτοϊστορική Θάσος, Τα νεκροταφεία του οικισμού Καστρί, Αθήνα: Δημοσιεύματα του Αρχαιολογικού Δελτίου, αρ. 45, 782-793. VAVELIDIS, M. 1984: Neue beobachtungen zur genese der Schichtgebundenen Pb-Zn-Be-Ba-As-Ag-Cu) und der Au-Vorkommen auf Thasos mit einem Beitrag zur Geologie, Petrographie und zum Metamorphosegrad des Gesteinskoplexes des Insel, Doktor Areit, Universität Heidelberg. VAVELIDIS, M.; PERNICKA, E. and WAGNER, G. A. 1988: «Die Goldvorkommen von Thasos», in G. A. Wagner and G. Weisgerber (eds.) Antike Edel- und Buntmetallgewinnung auf Thasos, Der Anschnitt, Beiheft 6, Deutsches Bergbau Museum, Bochum, 113-124. VOREADIS, G. D. 1954: Γεωλογικαί και κοιτασματολογικαί έρεναι εν Θάσω, Ι.Γ.Μ.Ε., Αθήνα, 227-282. WAGNER, G. A. and WEISGERBER, G. 1988: Antike Edelund Buntmetallgewinnung auf Thasos, Der Anschnitt, Beiheft 6, Bochum: Deutsches Bergbau Museum, Bochum. WEISGERBER, G. and WAGNER, G. A. 1988: «Der antike Goldbergbau auf dem Gipfel des Klisidi bei Kinyra», in G. A. Wagner and G. Weisgerber (eds.) Antike Edel und Buntmetallgewinnung auf Thasos, Der Anschnitt, Beiheft 6, Bergbau-Museum, Bochum, 154-172.

125

COMPARATIVE STUDY OF SLAGS FROM TWO DIFFERENT COPPER SMELTING SITES IN THE SOUTHERN ‘ARABA VALLEY, ISRAEL Shana Shilstein**, Uzi Avner***, Tal Kan-Cipor - Meron****, Sariel Shalev*

Abstract The results of XRF analysis of different copper smelting slags from two ancient Cu smelting sites are presented, compared and discussed. Yotvata. Using XRF we found a clear chemical compositional difference between two kinds of slags: the small ones are lower in iron and manganese in comparison to copper (in values of [I(Fe)+I(Mn)]/I(Cu) ratios) and higher in their inhomogeneity, while the ratios of intensities of slag «cakes» are higher in favor of Fe+Mn, with higher homogeneity (Fe+Mn concentrations are ≈20-30%, and Cu concentrations are ≈1.5% and ≈0.2% correspondingly). The results of 14C dating of these slags show that the small ones belong to the Iron Age (1000 BCE) and the bigger ‘cake’ shape ones were produced in the Early Islamic Period (700 AD). Amram 54/1. Here, as well we found two different types of slags that could be dated by their typical composition. Small slags with Cu concentration (6-9%) higher then Fe (≈2%), representing a self-fluxing process as known in the Chalcolithic Period. Tapping and parts of ‘cake’ slags, relatively inhomogeneous, containing about

**** Department of Archaeology and School of Marine Science, University of Haifa, Haifa 31905, Israel. **** Department of Particle Physics and Astrophysics, Weizmann Institute of Science, 76100 Rehovot, Israel. **** The Dead Sea- Arava Science Centre and the Arava Institute, Eilat, Israel. **** Department of Archaeology, University of Haifa, Haifa 31905, Israel.

15% Fe and ≈1-2 % Cu and could be dated by their composition to Iron Age or a bit before. The results of our work show that with XRF analysis in the field, we could show direct correlation between different slag shape and composition and ‘insert’ it chronologically into the history of technology of a specific site. Key words: Israel; ancient copper smelting; XRF; slag content; technology improvement. Resumen Se presentan resultados de análisis por XRF de escorias provenientes de dos sitios de fundición de Cu antiguos. Yotvata. Mediante XRF hallamos una clara diferencia entre los dos siguientes tipos de escoria: las más pequeñas, de menor contenido de hierro y manganeso en relación al de cobre (en valores de [I(Fe)+I(Mn)]/I(Cu)) y de mayor grado de inhomogeneidad, mientras que la relación de intensidades en escorias en forma de «torta» presentan mayor contenido de Fe+Mn y mayor homogeneidad (las concentraciones de (Fe+Mn) representan  ≈20-30%, mientras que las de Cu son  ≈1.5% and ≈0.2%, respectivamente). Los resultados de datación por 14C indican que estas escorias datan del 1000 a.C. (edad de hierro), las más pequeñas, y del 700 d.C. (islámico temprano), las de forma de «torta». Amram 54/1.  En los pequeños trozos de escoria la concentración de Cu  (6-9%)  es mayor que la de Fe  (≈2%), representando un  proceso autofundente, conocido en los tiempos calcolíticos. Las fracciones de escoria tamizada y la de

127

«torta» son relativamente inhomogeneas, conteniendo alrededor del 15% de Fe y ≈1-2 % de Cu, pudiendo ser datadas en la Era de Hierro o levemente más antiguas. Por lo tanto, el estudio de la composición de escorias demuestra directamente el progreso tecnológico en el tiempo, en este área específica. Palabras clave: Israel; fundiciones antiguas de Cu; XRF; contenido en escoria; avance tecnológico.

Timna (Yotvata and site 54/1 in Amram Valley). In both cases it is easy to differentiate even visually the coexisting of two different kinds of slag. In one site, Yotvata, the slags are dated by 14C, and in the other site, Amram 54/1 the slags are not dated. Therefore these two sites are good for experimenting in XRF compositional clustering of the different groups of slags and in the possibility of using the chemical clustering for preliminary, on site, rough dating. ARCHAEOLOGICAL MATERIAL

INTRODUCTION Ancient extractive copper metallurgy in the ‘Araba valley has a very long history – from Chalcolithic up to medieval and modern times. In the main centers of copper production (Feinan and Timna) the development stages of metallurgical process have a long history of intensive study, starting by pioneers’ investigations of international team leading by B. Rothenberg in Timna (i.e. Rothenberg 1996), and continued by T. Levy et al in Feinan (Ben Yosef et al 2010) with specific slag studied of G. Bachman (Bachmann 1980), A. Huptman (Hauptmann 2000), E. Ben Yosef and others (Ben Yosef 2010). In other, much smaller smelting sites in the region, the dating of the industrial activities there are not so clear. At the same time, in many cases in these sites, clear indications of a sequence of different technological processes in time are observed. For instance, in our study of ancient copper-smelting site near En Yahav (Yekutiely et al 2005, Shalev et al 2006) bellow the place of copper production in the Bronze Age, a later production was identified, dated to the Mameluk Period. In all these relatively small ancient copper-smelting sites in the Araba Valley, the main visible archaeological remains are slag heaps and scattered slag debris. In this study, we wish to suggest a method to analyze these slags on site, as part of routine site survey for receiving a compositional profile and rough dating of the different kinds of slags. Such possibility was already particularly demonstrated during analytical research of the slag in Feinan (Hauptmann 2000, Bachmann 2008, Ben Yosef et al 2010, Ben Yosef 2010) and carried out in study of slag in some sites in Timna (Shalev et al 2007, Ben Yosef 2010, Shilstein et al 2014) and in several sites around Timna (Shalev et al. in press), presented in details in Figure 1, where a comparison of different slags from different periods and sites is presented. In this research and as a part of the major aim described above, we wish to demonstrate the existence of different stages of copper production and its possible identification by XRF analytical study of slags from two sites close to

Yotvata. A low pile of fragments of larger tapping slag «cakes» (Yotvata I, Fig. 2, up to 10 cm size), ca.6m across, was found some 200m west of the Late Roman fortress. The site was dated by Rothenberg as Byzantine; however, it contains distinctive Nabataean pottery shards (1st century BCE to 4th century CE). 14C dates made by Dr. L. Cummings, Paleo Research Institute, Golden Colorado, gave the date of 700 AD to this site. Adjacent to this pile of tapping slags, is a larger area scattered with small and tiny pieces of slags (Yotvata II, Fig.3, up to 1cm size), unnoticed by Rothenberg and is currently dated by 14 C analysis of several charcoal samples, to about 1000 BCE. Site Amram 54/1 is located on a hilltop outside the closed valley of Nahal Amram, 2.3 km ESE of the miners’ camp. No construction or installation is discernible on the hill, only a scatter of copper slag and copper ore. On the slopes there are also small stones and chunks of clay maybe remains of furnaces lining, coated with slag or green glazed. No datable artifacts were found yet in the site. Here two different kinds of slags were found: small (up to 1 cm, covered by corrosion and contained Cu inclusions); broken parts of tapping slag «cakes» (originally circa 5 cm thick and 8x8 cm in diameter). The slag pieces were divided to two groups by weight (more and less than 12 g). ANALYTICAL METHOD Slag samples from the two groups of each site were analyzed by energy dispersive X-ray fluorescence method (XRF), using a portable bench top EX–Calibur (Xenemetrix). The intensities (I) of XRF lines for Ca, Mn, Fe, Cu, Pb, Sr, Ba were measured at 45kV,0.25mA (Mo filtering), for 60 seconds, with a primary beam of 2mm in diameter for slag pieces or (in several samples) for powder produced by mechanical crushing. The method is described in details in Shalev et al., 2006 and in Shalev et al (in press).

128

A major obstacle in studying the chemical composition of slag is the great heterogeneity in every slag piece, even in the small ones (XRF lines of each element may differ in intensity up to 100 times). For instance, in the XRF spectrum on Figure 5 (upper) the ratio [I(Fe)+I(Mn)]/I(Cu) is about 50, it means that for a typical content of the fluxing elements (Fe + Mn) of about 20-30%, the Cu concentration is about 0.2-0.3%. The situation of this maximal percentage of readings, which give such high intensities ratios, represents, from our data (Shalev et al, in press) a late periods production like in Early Islamic. Such ratio is not far from ratio of the fluxing components of modern copper smelting production.

In the XRF spectrum on Figure 5 (underneath) this ratio is about 1. It means that for the same content of fluxing elements the copper concentration is not higher than several per cents. The situation of maximal percentage of readings giving such low intensities ratios corresponds to some part of slag samples from Bronze Age (about 2000 BCE) (Yekutieli et al 2005). But even for rather «old» slag samples in some points the spectra similar to the one presented in the upper part of Figure 5 could be observed. In attempt to overcome the effect of heterogeneity, 5-10 analyses of different areas on each piece of slag were conducted. The readings were taken from both the surface and the interior part

Figure 1. Araba map (south of Israel).

129

Figure 2. Slag samples (cake’s parts) from site Yotvata I.

of each slag to avoid a possible effect of surface corrosion, since the X-ray beam penetrates only several tens of microns. In some cases, the rough surface was polished and cleaned before analysis. Generally, the method adopted here includes multiple XRF readings of many slag pieces and of many different spots on each of them, then calculating the percentage of the various ratios of [I(Fe)+I(Mn)] and I(Cu), which are the most important metal elements present in the slag. The Fe and Mn found in the slag originate mainly from the fluxing minerals and the results reflect the relative level of pyrotechnology. As the technology advanced, this ratio increases in favor of Fe and Mn in a higher percentage of readings and the level of homogeneity increases as well, indicating a better use of fluxes, better control of the furnace temperature and better extraction of copper from the smelted ore (Bachmann 1980; Shalev et al. 2006; Ben-Yosef 2010; Shilstein et al. 2014). The main criterion we are using, therefore, is the ratio of intensities of Fe+Mn (fluxing components) and Cu (produced metal) as indicator to the level of copper extraction from the ore

during the smelting process. For a characterizing separately action of each group of slag samples we used the distribution of the number of reading of specific ratio [I(Fe)+I(Mn)]/I(Cu) and sometime even by using just the ratio results as good enough indicator of different groups. This ratio is varied in a broad region from 0.01 (in maximal Cu content and minimal Fe+Mn content in the slag) and up to about 100-200 (in minimal Cu content and maximal Fe+Mn content). Therefore the data will be presented as graphs of percentage of readings depending on the ratio [I(Fe)+I(Mn)]/I(Cu), and the ratio is represented in the logarithmic scale (!). RESULTS AND DISCUSSION Yotvata. In Fig. 6 the graphs show a clear difference between the two types of slag from Yotvata. In the small pieces of slag (Yotvata II) the distribution is ranged from about 0.05 up to about 500 with maximal probability at 30-40. But the significant probability is presented for ratios from 5 up to 200. It means that inhomo-

130

geneity of the copper content found in different parts of this slag could be equal about 40 times (from our data the Fe+Mn content in this region of ratios is varied rather weakly). This distribution has long «tail’ to minimal values of ratio [I(Fe)+I(Mn)]/I(Cu) 0.2 and Cu contents between 0.6 and 2 wt% and the Northern objects by Ag/Au ratios < 0.2 and Cu contents above 1.5 wt%. It must be stressed that the Southern bracelets from

163

Figure 7. Ag/Au ratio as a function of Cu content (in wt%) obtained by XRF for MBA and LBA bracelets from the collection of the NMA. Major elements were normalised to 100 wt%.

Figure 8. Ag/Au ratio as a function of Cu contents (in wt%) obtained by XRF for MBA and LBA bracelets from the collection of the NMA separated by origin (above and below Tagus river). Major elements were normalised to 100 wt%.

the Treasure of Herdade do Álamo as well as the bracelet found in Herdade das Cortes (Beja) are, as expected, contained in the Southern group. This separation could correspond to two different «supplies» that cannot be discussed without an overview of the finds in the whole Iberian Peninsula dated to the BA and before BA, particularly those situated next to gold sources like the Tagus River and the northern mines.

ing 10 rods) were fully analysed by XRF and one of them analysed by µPIXE for the rods and the joins. Data and elemental distribution maps obtained for the gold rods and joining areas confirmed the use of hard-solder and allowed to determine the composition of both the rods and the solders. Data obtained by XRF for the other items from this treasure (attributed by the technologies employed to the IA) showed that necklace Au 189 is made from different alloys with compositions that seem characteristic of the BA, corroborating its attribution by Armbruster and Parreira (1993) to the BA instead of IA. The two values published by Hartmann (1982) using OES for the two composite bracelets from the Treasure of Herdade do Álamo (sampling localisation not provided in this database) are, in addition to insufficient to characterise these objects, far from our XRF and µPIXE results. The high Cu content observed for one of the Hartmann’s samplings suggests that it is representative of a solder instead of a rod.

6. CONCLUSIONS The compositional data obtained using a portable XRF system with 0.5 mm diameter beam for a statistically representative group of 57 BA bracelets from the Portuguese area, belonging to the NMA collection, many of them from treasures with well-established geographical contexts, allowed proposing a chronological reference based on the gold alloys. MBA and LBA bracelets form two distinct groups according to their chronologies, and the MBA bracelets attributed to the Southern Portuguese area show different Ag/Au ratios and Cu contents from those attributed to the Northern area. It should be emphasised that the introduction during the LBA of more complex technological processes of goldworking results in the use of a wider group of alloys. The two composite bracelets from the Treasure of Herdade do Álamo (each made by join-

7. ACKNOWLEDGEMENTS Part of the research leading to the published results has received funding from the Portuguese FCT (Foundation for Science and Technology) under the AuCORRE project PTDC/HISHIS/114698/2009. I. Tissot acknowledges grant SFRH/BDE/51439/2011 from the Portuguese FCT.

164

Ref.

Provenance

Region

Chronology

Link to database*

Au27

Treasure of Colos. Odemira

Beja

South

MBA

110013

Au28

Treasure of Colos. Odemira

Beja

South

MBA

110015

Au29

Treasure of Colos. Odemira

Beja

South

MBA

110030

Au30

Treasure of Colos. Odemira

Beja

South

MBA

110019

Au31

Treasure of Colos. Odemira

Beja

South

MBA

110020

Au37

Vinhós. Fafe

Braga

North

MBA

110031

Au44

Unknown

Beja

South

MBA

110003

Au50

Torgueda. Vila Real

Vila Real

North

MBA

110032

Au57

Treasure of Arnozela. Fafe

Braga

North

MBA

110049

Au58

Treasure of Arnozela. Fafe

Braga

North

MBA

110059

Au59

Treasure of Arnozela. Fafe

Braga

North

MBA

110033

Au60

Treasure of Arnozela. Fafe

Braga

North

MBA

109999

Au61

Treasure of Arnozela. Fafe

Braga

North

MBA

110034

Au62

Treasure of Arnozela. Fafe

Braga

North

MBA

110048

Au63

Treasure of Arnozela. Fafe

Braga

North

MBA

110035

Au64

Treasure of Arnozela. Fafe

Braga

North

MBA

110036

Au65

Treasure of Arnozela. Fafe

Braga

North

MBA

110037

Au66

Treasure of Arnozela. Fafe

Braga

North

MBA

110002

Au67

Treasure of Arnozela. Fafe

Braga

North

MBA

110038

Au68

Treasure of Arnozela. Fafe

Braga

North

MBA

110039

Au69

Treasure of Arnozela. Fafe

Braga

North

MBA

110040

Au70

Treasure of Arnozela. Fafe

Braga

North

MBA

110041

Au71

Treasure of Arnozela. Fafe

Braga

North

MBA

110042

Au72

Treasure of Arnozela. Fafe

Braga

North

MBA

110043

Au73 e 73a

Treasure of Arnozela. Fafe

Braga

North

MBA

110044

Au74

Treasure of Arnozela. Fafe

Braga

North

MBA

110045

Au75

Treasure of Arnozela. Fafe

Braga

North

MBA

110046

Au76

Treasure of Arnozela. Fafe

Braga

North

MBA

110047

Au177

Turquel. Alcobaça

Leiria

North

MBA

110028

Au178

Turquel. Alcobaça

Leiria

North

MBA

110029

Au495

Treasure of "Beira Alta"

"Beira Alta"

North

MBA

110051

Au496

Treasure of "Beira Alta"

"Beira Alta"

North

MBA

110052

Au497

Treasure of "Beira Alta"

"Beira Alta"

North

MBA

110053

Au498

Treasure of "Beira Alta"

"Beira Alta"

North

MBA

110054

Au499

Treasure of "Beira Alta"

"Beira Alta"

North

MBA

110055

Au500

Treasure of "Beira Alta"

"Beira Alta"

North

MBA

110056

Au501

Treasure of "Beira Alta"

"Beira Alta"

North

MBA

110057

Au502

Treasure of "Beira Alta"

"Beira Alta"

North

MBA

110058

Au26

Qui nta da Bouça V N Fa ma l i cã o

Braga

North

LBA

110059

Au32

Santo António. Arraiolos

Évora

South

LBA

110000

Au33

Monforte da Beira. C Branco

Castelo Branco

North

LBA

110021

Au34

Baralhas. Vale de Cambra

Aveiro

North

LBA

110022

Au35

Baralhas. Vale de Cambra

Aveiro

North

LBA

110001

Au36

Vila do Conde

Porto

North

LBA

110023

Au38

Baralhas. Vale de Cambra

Aveiro

North

LBA

110024

Au39

Unknown

Unknown

LBA

110025

Au138

Herdade das Cortes. Alvito

Beja

South

LBA

110004

Au139

Herdade das Cortes. Alvito

Beja

South

LBA

110005

Au142

Herdade das Cortes. Alvito

Beja

South

LBA

110006

Au165

Redondo. Évora

Évora

South

LBA

110009

Au191

Herdade do Álamo. Beja

Beja

South

LBA

110010

Au192

Herdade do Álamo. Beja

Beja

South

LBA

110011

Au193

Cantonha. Guimarães

Braga

North

LBA

110012

Au271

Urra. Portalegre

Portalegre

South

LBA

110014

Au 296

Unknown

Unknown

LBA

110018

Au564

Soalheira. Covilhã

Castelo Branco

North

LBA

110026

Au981

Monte Airoso. Mourão

Évora

South

LBA

110027

* use the link followed by the number for each object http://www.matriznet.dgpc.pt/MatrizNet/Objectos/ObjectosConsultar.aspx?IdReg= Ex: Au502 (http://www.matriznet.dgpc.pt/MatrizNet/Objectos/ObjectosConsultar.aspx?IdReg=110058)

Table 1. List of bracelets considered in this study with provenance and chronology after the NMA database inventory. The link to the online database is obtained by replacing ‘XXXXXX’ in the following address by the reference given in the table: http://www.matriznet.dgpc.pt/MatrizNet/Objectos/ObjectosConsultar.aspx?IdReg=XXXXXX

165

Std 1 M1 M2 Std 2 M1 M2 Std 3 M1 M2 Std 4 M1 M2 Std 5 M1 M2

Au 89 89.3 89.6 89 89.3 89.4 83 82.4 82.4 98 98.1 98.1 99 99.2 99.2

in wt% Ag 11 10.7 10.4 10 9.8 9.8 15 15.9 15.7 0 0 0 0 0 0

HARTMANN, A. 1970: Prähistorische Goldfunde aus Europa. Spektralanalytische Untersuchungen und deren Auswertung. Studien zu den Anfängen der Metallurgie 3, Ed. Mann Verlag, Berlin.

Cu 0 0 0 1 0.8 0.9 2 1.8 1.6 2 1.9 1.9 1 0.8 0.6

— 1982: Prähistorische Goldfunde aus Europa II. Spektralanalytische Untersuchungen und deren Auswertung. Studien zu den Anfängen der Metallurgie 5, Ed. Mann, Berlin. HOOK, D. R. and NEEDHAM, S. P. 1989: «Comparison of recent analyses of British Late Bronze Age goldwork with Irish parallels». Jewellery Studies 3: 15-24. ISSN 0268-2087. LEMASSON, Q.; MOIGNARD, B.; PACHECO, C.; PICHON, L. and GUERRA, M. F. 2015: «Fast mapping of gold jewellery from ancient Egypt with PIXE: Searching for hard-solders and PGE inclusions». Talanta 143: 279-286. doi: 10.1016/j.talanta.2015.04.064.

Table 2. Elemental composition obtained by XRF at experimental conditions for five gold alloy home-made standards used to verify the quantitative results accuracy. Two measures (M1 and M2) were carried out on each sample

MARYON, H. 1941: «I. Welding and Soldering». MAN 41: 118-124. http://www.jstor.org/stable/2791583.

8. REFERENCES

PEREA, A. 1995: «La metalurgia del oro en la fachada atlántica Peninsular durante el Bronce Final interacciones tecnológicas». Complutum 5: 69-78.

McDONALD, A. S. and SISTARE, G. H. 1978: «Part I – Coloured gold metals». Gold Bulletin 11: 128-131. doi: 10.1007/BF03216534.

ALVES, L. B. and REY, B. C. 2009: «Rochas e metais na Pré-história para além da físico-química». In A. M. S. Bettencourt and L. B. Alves (eds.): Dos montes, das pedras e das águas. Formas de interacção com o espaço natural da pré-história à actualidade. CITCEM APEQ, Braga: 37-54. ISBN. 978-989-8351-02-9. ALVES, L. C.; BREESE, M. B. H.; ALVES, E.; PAÚL, A; SILVA, M. R. da; SILVA, M. F. Da and Soares, J. C 2000: «Micron-scale analysis of SiC/SiCf composites using the new Lisbon nuclear microprobe». Nuclear Instruments and Methods in Physics Research Section B 161-163: 334-338. doi: 10.1016/S0168583X(99)00768-5. ARMBRUSTER, B. and PARREIRA, R. (eds.) 1993: Inventário do Museu Nacional de Arqueologia. Colecção de Ourivesaria: 1º Volume do Calcolítico à Idade do Bronze. Secretaria de Estado da Cultura, Instituto Português dos Museus, Lisboa. ISBN 972-957758-7. CORREIA, V. 2007: «The early Iron Age transition in the goldwork of the west Iberian Peninsula». In C. Burgess, P. Topping and F. Lynch (eds.): Beyond Stonehenge essays on the Bronze Age in honour of Colin Burgess. Oxbow Books, Oxford: 90-96. ISBN 9781842172155 . — «Tradições e inovação na ourivesaria arcaica da fachada atlântica peninsular». In M. F. Guerra and I. Tissot (eds.): A Ourivesaria Pré-histórica do Ocidente Peninsular Atlântico Compreender para Preservar. AuCORRE, Lisboa: 21-23. ISBN 978989204470.

— 2005: «Mecanismos identitarios y de construcción de poder en la transición Bronce-Hierro». Trabajos de Prehistoria 62: 91-104. doi: 10.3989/tp.2005. v62.i2.70. PEREA, A. and ARMBRUSTER, B. 2008: «Tradición, cambio y ruptura generacional. La producción orfebre de la fachada atlántica durante la transición Bronce-Hierro de la Península Ibérica». In S. Celestino Pérez, N. Rafel i Fontanals and X. L. Armada Pita (eds.): Contacto cultural entre el Mediterráneo y el Atlántico: (siglos XII-VII ane): la precolonización a debate. Serie Arqueológica 11, Consejo Superior de Investigaciones Científicas, Madrid: 509-522. ISBN 978-84-00-08689-3. PESSANHA, S.; GUILHERME, A. and CARVALHO, M. L. 2009: «Comparison of matrix effects on portable and stationary XRF spectrometers for cultural heritage samples». Applied Physics A 97: 497-505. doi: 10.1007/s00339-009-5251-x. SENNA-MARTINEZ, J. 2013: «Metals, technique and society. The Iberian Peninsula between the first pleasant societies with metallurgy and the «urban revolution». In M. F. Guerra and I. Tissot (eds.): A Ourivesaria Pré-histórica do Ocidente Peninsular Atlântico Compreender para Preservar. AuCORRE, Lisboa: 11-20. ISBN 978989204470. TAYLOR, J. 1980: Bronze Age goldwork of the British Isles. Cambridge University Press, Cambridge.

166

TISSOT, I.; TISSOT, M.; MANSO, M.; ALVES, L. C; BARREIROS, M. A.; MARCELO, T.; CARVALHO, M. L.; CORREGIDOR, V. and GUERRA, M. F. 2013: «The earrings of Pancas treasure: Analytical study by Xray based techniques – A first approach». Nuclear Instruments and Methods in Physics Research Section B 306: 236-240. doi: 10.1016/j.nimb.2012.11.054. TISSOT, I.; TROALEN, L. G.; MANSO, M.; PONTING, M.; RADTKE, M.; REINHOLZ, U.; BARREIROS, M. A.; SHAW, I.; CARVALHO, M. L. and GUERRA, M. F. 2015: «A multi-analytical approach to gold in Ancient Egypt: studies on provenance and corrosion». Spectrochimica Acta Part B 108: 75- 82. doi: 10.1016/j.sab.2015.03.012 TROALEN, L. G. and GUERRA, M. F. 2016: «Gold from the tomb of Scribe Beri: a comparative analytical approach to the New Kingdom gold grave goods from Riqqa (Egypt)». Applied Physics A 122(3): 1-12. doi: 10.1007/s00339-016-9699-1

TROALEN, L. G.; TATE, J. and GUERRA, M. F. 2014: «Goldwork in Ancient Egypt: workshop practices at Qurneh in the 2nd Intermediate Period». Journal of Archaeological Science 50: 219-226. doi:10.1016/j. jas.2014.07.010. VOÛTE, A. 1995: «Some experiences with the analysis of gold-objects». In G. Mortani and J. P. Northover (eds.): Prehistoric gold in European mines, metallurgy and manufacture. Kluwer Academic Publishers, Dordrecht: 329-340. ISBN 9789401512923. WARNER, R. 2004: «Irish gold artefacts: Observation from Hartmann’s analytical data». In H. Roche, E. Grogan, J. Bradley, J. Coles and B. Raftery (eds.): From megaliths to metal. Essays in honour of George Eogan, Oxbow Books, Oxford: 72-82. ISBN 1 842171518. WARNER, R. and CAHILL, M. 2011: «Analysing ancient Irish gold: an assessment of the Hartmann database». The Journal of Irish Archaeology 20: 45-52.

167

METALLURGICAL AND TECHNOLOGICAL ASPECTS OF EARLY IRON AGE GOLD Barbara Armbruster*, Maryse Blet-Lemarquand**, Bernard Gratuze**, Verena Leusch***, Ernst Pernicka***, Birgit Schorer****, Roland Schwab***

Abstract

INTRODUCTION

Gold ornaments and vessels were important products during the Early Iron Age. This paper deals with the development of goldworking in the West Hallstatt culture (8th-5th cent. BC). It focuses on four aspects of this craft: 1. embossed sheet ornaments; 2. sheet ornaments and rotary motion; 3. technological innovation - soldering and gilding; and 4. workshop identities.

This paper deals with the metallurgical and technological aspects of goldworking in the West Hallstatt culture (8th-5th centuries BC) through the study of craft production.1 This Early Iron Age culture was situated in a geographic area covering south-west Germany, East and Eastcentral France and parts of Switzerland (Fig. 1). It is characterized by fortified settlements and princely burials containing rich prestige goods of the deceased individuals. Gold objects were an important component of these prestige goods. Found mainly in these princely graves, their context is informative for functional, socio-economic and symbolic studies. The male EberdingenHochdorf burial in South-West Germany and the female Vix burial in Burgundy are two representative rich burial mounds containing precious metal jewellery and metal vessels besides luxury items in other valuable, imported and exotic materials, and four-wheeled wagons (Hansen 2010; Rolley 2003). The goldsmith’s work is well suited for demonstrating traditions and local innovations, as well as foreign influences and exchange networks in arts and crafts. Iron Age gold objects are considered as symbols of power and prestige at the aristocratic level of society. Many gold items bear marks of wear that show long-lasting use, while others were produced exclusively for the final deposition during

Key words: gold, Iron Age, technology, workshop. Resumen Los recipientes y adornos de oro fueron objetos importantes durante la Edad del Hierro. Este trabajo aborda el desarrollo del trabajo del oro en la cultura de Hallstatt occidental (siglos VIII-V AC). Se centra en cuatro aspectos de esta artesania: 1. Adornos laminares en relieve. 2.- elementos laminares y con elementos rotatorios, 3.- Innovación tecnológica: soldadura y dorado; y 4.- identidades de taller Palabras clave: oro, Edad del Hierro, tecnología, taller.

    * TRACES – CNRS UMR 5608 – Toulouse (France),   ** IRAMAT-CEB–UMR 5060 CNRS université d’Orléans–Orléans (France)   *** CEZ Archäometrie – Mannheim (Germany) **** Landesmuseum Württemberg–Stuttgart (Germany)

1   The paper presents results from a bi-national project financed by the French ANR and the German DFG, entitled: «West Hallstatt Gold. Rethinking earliest Celtic gold – Economic, social and technological perspectives in the West Hallstatt Culture» 2012-2015 ANR 11-FRAL-0013, directed by B. Armbruster (CNRS-TRACES, Toulouse) and E. Pernicka (University of Heidelberg).

169

Figure 1. Location of the central western zone of the Hallstatt culture and gold finds (Early Iron Age c. 800-500 BC)–Germany, France and Switzerland.

funeral ceremonies. The total weight of about 480 gold objects from the West Hallstatt culture is less than 5 kg. Compared with the Late Bronze Age, Early Iron Age gold production was quite small due to the fact that the majority of West Hallstatt culture gold ornaments, decorative elements and vessels, were made of thin metal sheet. To give an example, the gold hoard from Guînes (Pas-de-Calais) weighs 4. 57 kg. It contains only five ornaments, all solid gold: three neck-rings, a bracelet and a large unidentified object, dating to the 13th-12th centuries BC (Armbruster/ Louboutin 2004). The weight of this Late Bronze Age assemblage equals nearly the total weight of the presently known Early Iron Age goldwork. METHODS The study of metallurgical and technological aspects of ancient gold requires an interdisciplinary approach to metalworking combining archaeology, technology, material sciences and experiments (Armbruster 2011). The first approach to be applied to the gold artefacts is the optical examination by macro- and microscopy for a technical identification of the number of constituent parts, the surface topography and tool marks, and to characterize tool marks and other traces of processing techniques. Some of the technical features observed, like embossing with punches or the manufacturing of cylindrical use of a spinning lathe, were subsequently

reconstructed by experimental archaeology. Technological analysis of the tool marks was conducted by optical microscopy, scanning electron microscopy, macro-photography and by silicon imprints. Many gold objects were analysed by several methods from the material sciences to determine their alloy composition and technology. Objectives of these analytical investigations were to distinguish between natural and artificial alloys, to identify regional or chronological differences, to scrutinize intentional alloying for solder material, and to look for methods of gilding. Different complementary methods were applied: X-ray fluorescence (XRF), laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS), metallography, scanning electron microscopy and associated energy dispersive Xray analysis (SEM-EDX), and radiography (BletLemarquand et al. in print). The gold objects were studied by the two laboratories involved in the project depending where they were kept: the Institut de recherche sur les archéomatériaux, Centre Ernest-Babelon (IRAMAT-CEB) at Orléans, France, and the Curt-Engelhorn-Zentrum Archäometrie (CEZA) at Mannheim, Germany. The two laboratories developed a common analytical protocol enabling comparison of techniques and data between them to ensure complete compatibility of their results. LA-ICP-MS provided full compositional analysis (contents of major, minor and trace elements) of the gold objects and was especially useful to

170

obtain a chemical «fingerprint» of the gold alloy which could discriminate between different metallic stocks and origin (Blet-Lemarquand et al. 2014). The depth profile mode of LA-ICP-MS enabled composition to be determined from the surface to the interior so that the potential surface effects (e.g. copper depletion, contaminations from the burying soil, etc.) could be disregarded. The traces left on the object are invisible to the naked eyes (below 0.1 mm in diameter) and were acceptable to curators. A major limitation of a conventional LA-ICP-MS setup is the restricted sample size given by the dimensions of the ablation cell (up to a few centimetres in maximum size). Two types of innovative analytical cells were developed in Orleans to solve this problem and make it possible to analyse large gold items such as thick bracelets or even torcs. XRF was performed on multi-component objects in order to test the elemental homogeneity and thus decide on the best analytical strategy. This method helped also to characterise solders. It was furthermore used to analyse the gold objects that were too large to enter the LA-ICP-MS ablation cell available in Mannheim. The elements measured using XRF were generally restricted to Au, Ag and Cu, but sometimes significant contents in Sn and Fe were also detected. SEM-EDX was reserved for the investigations of the surface of solders (see below) or to

provide additional information on the surface of gold items. When an invasive approach was allowed, the Mannheim laboratory applied «in situ metallography» to study multi-component metal objects such as those with gilded surfaces (see below). The French laboratory was equipped with the following devices. The SEM is a FEI XL 40 with associated Oxford EDX. The XRF spectrometer is the Bruker ARTAX 1. The HR-ICP-MS is a Thermo Fisher Element XR which can be coupled to two different lasers: a Resonetics Resolution M-50-E (193 nm) or a VG Nd-YAG UV laser probe (266 nm). The German working group in Mannheim used a Fischerscope® XRF system with a silicon drift detector (SDD) and an integrated videomicroscope for object placement, allowing precise measurements with high spatial resolution (ca. 0.2 mm) for non-destructive analyses in the museums stores. A Resonetics laser ablation system with inductively coupled plasma quadrupole mass spectrometer (LA-ICP-QMS) was used for smaller items and selected samples. For technological studies, the objects themselves and some flat polished specimen were studied using optical microscopes (OM) and a Zeiss EVO 60 SEM fitted with a Bruker Quantax system and a SDD with a specified energy resolution better than 127 eV at MnKα.

Figure 2 Examples of decorated sheets from the Eberdingen-Hochdorf burial (photographs a)-c) B. Schorer, © Landesmuseum Württemberg, Stuttgart; d) P. Frankenstein, H. Zwietasch, © Landesmuseum Württemberg, Stuttgart).

171

EMBOSSED SHEET ORNAMENTS The bulk of the gold artefacts, including jewellery, tableware and decorative appliqués, are made of sheet metal. Nearly all of these West Hallstatt sheet ornaments are decorated completely or partially with punched motives (Fig. 2). The sheet ornaments and vessels bear geometric and figurative patterns and symbols produced by chasing and punching. Observation of tool marks with a light microscope, in particular on cylindrical arm –and neck-rings, hair– and ear-rings, pin heads and vessels, showed that the punch marks were mainly executed from the outside of the three dimensional sheet body using so-called negative punches. The decoration of these sheet ornaments by plastic shaping was performed after the form of the object had been produced. Analysis of the punch marks revealed the use of an iconographic repertoire with geometric and figurative patterns known also from other West Hallstatt craft products such as the richly decorated ceramics, bronze objects and textiles. The main repeatedly-occurring motifs are crosses, circles, circular bosses, S- and Z-shapes, meanders, and star-shapes, but stylised horses and masks also appear. A full record of fine metalworking tools of bronze for plastic shaping techniques, including decorative punches, chisels and tracers, is known from the preceding Late Bronze Age (Armbruster et al. 2003). This is also the case with the subsequent Late Iron Age when bronze tools for fine metalworking were still in use. The toolkit of a goldsmith and warrior from the 4th century BC in Iberia comprises the whole panoply of metal tools required (Perea/Armbruster 2011). But unfortunately, decorative punches for embossed sheet ornaments are almost absent from Early Iron Age contexts, like other fine metalworker’s tools. The only probable punches identified for Iron Age goldwork are two decorative punches made from antler, found at the Heuneburg near Hundersingen, Germany (Sievers 1984, Pl. 120). Experimental work with punches of organic material, bone and antler, has yielded very satisfactory results. Punch-mark analyses conducted with detailed measurements and silicon imprints of embossed sheet ornaments using a digital microscope (Keyence) showed evidence of the use of identical tools on items from the same site, like star-shaped punch marks on the dagger and the bracelet from the Eberdingen-Hochdorf burial. Very similar punch marks on the neck-ring from Stuttgart-Bad Cannstatt (burial 1) were undertaken with a different tool. Both sites are contemporary (Hallstatt D2). An interesting example

of embossing methods is shown by the decorative shoe fittings from Eberdingen-Hochdorf. The artisan fabricated the sheet ornaments for both shoes simultaneously by laying two sheets one upon the other and embossing them both at once. Experiments during the project confirmed this previously debatable assumption (Hansen 2010, 44). SHEET ORNAMENTS AND ROTARY MOTION One of the principal typological and technological groups of the West Hallstatt gold consists of cylindrical annular ornaments, about 20 neck-rings and 19 arm-rings. French examples of this group are the neck-rings from Apremont 1 and 2 and from Ensisheim. German examples come from Eberdingen-Hochdorf, Ludwigsburg «Römerhügel», Stuttgart-Bad Cannstatt and Herbertingen-Hundersingen (Fig. 3).

Figure 3. Annular sheet ornaments worked by cold spinning: Torcs from Eberdingen-Hochdorf, Ludwigsburg «Römerhügel», Stuttgart-Bad Cannstatt and HerbertingenHundersingen(photograph P. Frankenstein, H. Zwietasch, © Landesmuseum Württemberg, Stuttgart).

The study of a large number of neck- and arm-rings provides for the first time new insights into the complex manufacturing technique of cold spinning. When examined optically, neckrings show up as closed cylindrical rings whilst arm-rings appear as sheet cylinders which were opened by cutting the sheet with a chisel after

172

Figure 4 A-B. Experimental spinning lathe and tools. a) Wooden lathe and chuck; b) Detail of the corrugated surface of the sheet cylinder (photographs A. Dingeldey).

having carried out the shaping and decoration. Since the sheet cylinders have no joining seam they must have been crafted from a cylindrical cast pre-product. The perfect cylindrical form, as well as the regular corrugated relief, indicate the use of a spinning lathe [relevant] for shaping. Rotary motion in goldworking was known since the Late Bronze Age in Iberia, where solid cast arm- and finger-rings were fabricated on a lathe. In addition to our Iron Age examples of annular sheet cylinders, the Late Bronze Age ornaments were worked as solid wax models on a lathe and subsequently cast in the lost-wax process (Armbruster 2004). The use of a lathe has already been indirectly attested during the Early Iron Age for wood and bronze objects used in woodworking, and to make the wax models used for lost-wax casting (Rieth 1956; Drescher 1987). No archaeological evidence for a lathe is known from prehistoric times. One of the earliest examples known so far relies on the depiction of a woodworking lathe in the Egyptian tomb of Petosiris, c. 300 BC (Lefebure 1924). The use of a lathe can only be deduced from the parallel tool marks and the perfectly circular diameter of the worked piece of wood. Hypothetical reconstitution of the prehistoric lathe relies on analogies from iconographical sources, experimental archaeology and ethnoarchaeological reasoning. The lathe is one of the earliest complex tools and the distant predecessor of complex machines. Construction of a lathe requires a rotary axis pivoted on two wooden posts, manually powered by a looped strip, and a hand-held cutting or pressing tool. The spinning of cold metal used for the West Hallstatt sheet ornaments is a method of forming circular or cylindrical objects with hollow and seamless shapes by rotational symmetry and with great refinement. Cold spinning experiments were performed with antler tools on a reconstructed lathe. The wooden lathe was driven

by a bow while a wooden chuck was fixed on the rotary axis. The desired relief was cut with iron tools into the wooden cylinder to produce a corrugated surface. After that, a sheet cylinder worked by hammering was pulled over this corrugated wooden core. Then the sheet was pressed into the grooves with antler tools (Fig. 4). The decoration was executed from the outside of the cylinder using the wooden core as support for punching. Finally, the wooden chuck had to be removed by burning it out. The temperature of the fire had to be kept low in order to prevent the metal sheet from melting. TECHNOLOGICAL INNOVATION: SOLDERING AND GILDING Apart from the spinning lathe, soldering and gilding techniques were the main technological innovations in goldworking in the West Hallstatt culture. Soldering was possible either with metallic soldering alloys or with copper-salt solders. Copper-salt solder was applied under reducing conditions to form an alloy of gold and copper with a low melting point, especially for the application of small granules and decorative wire elements. The so-called copper-salt or reaction soldering technique is well established on Etruscan and Greek jewellery from the 7th century BC and later (Born et al. 2009, 20). Granulation and filigree decoration techniques appeared also in the Early Iron Age; most probably by Etruscan influences (Eluère 1989). The earliest objects in Western Europe that are supposed to be soldered with copper-salt solders are the pendants from Ins-Grossholz and Jegenstorf-Hurst in Switzerland (Eluère 1989, 50). Similar pendants with soldered filigree ornaments are emerging from the recently-excavated female burial at Bettelbühl in Germany (Krausse/Ebinger-Rist 2011).

173

Copper-salt soldering for joining parts could also have been used for the brooches and for the small rings belonging to the belt-plate from the Eberdingen-Hochdorf tomb, for the strainer spoon from the Heuneburg settlement and for a small ring from burial 2 at Stuttgart-Bad Cannstatt, all in south-western Germany and all dating to Hallstatt D2. This substance-to-substance joining is clearly identifiable by its morphologies, as the metal between the soldered parts is usually molten with dendritic solidification structures (Fig. 5). In most instances a slight increase of the copper content within the molten metal surface could be detected by EDX or LA-ICP-MS measurement. Experimental work with copper-salt solders shows a similar optical appearance, with precise joints and molten surfaces, as well as a significant increase of the copper concentration in the joints. Accordingly, we suggest that copper-salt solders were most probably used for these objects. Other cases of granulation and filigree, all from French funerary contexts, are mostly later than the soldered objects from south-western Germany. They seem to have been made with

metallic solder material, as can be seen on the ornament with a ram’s head from Lazenay, or the ear-rings from Gurgy and Charmoy (Hallstatt D3-La Tène A1). In some cases, as with objects from the Sainte-Colombe and Vix burials, droplets of solder alloy and areas of metallic soldering are visible with the naked eye (Table 1 and Fig. 6a). Eluère (1989, 52) suggested that copper-salt soldering was used for the bracelets from SainteColombe because slightly higher levels of copper were measured on the surface of the solders than in the gold. The recent LA-ICP-MS analysis of solders on these bracelets led to similar chemical patterns, but our conclusions about the joining technique are in disagreement with Eluère’s hypothesis. The solder lump discovered on an ear-ring proves the use of soldering with a gold alloy, while its composition indicates that the solders were enriched with copper. The two bracelets and the two ear-rings from Sainte-Colombe have identical chemical compositions and their solders present similar morphologies. All the gold objects from Sainte-Colombe were soldered using the metallic solder process.

Figure 5. Small rings from Eberdingen-Hochdorf (Germany) and soldered part of a small ring showing dendritic solification structure (photograph B. Schorer, SEM-image R. Schwab, © Landesmuseum Württemberg, Stuttgart).

Figure 6 A-C. Details of soldering. a) Round soldering lump found in vase on an ear-ring from Sainte-Colombe (France), (SEM image M. Blet Lemarquand, © Musée d’Archéologie Nationale). b) Head of the pin from the bracelet of Ensisheim (France). (SEM image M. Blet-Lemarquand, © Musée de Colmar). c) Transverse solder on the ring from Cérilly (France). (SEM image M. Blet-Lemarquand, © Musée d’Archéologie Nationale).

174

SEM-EDX

Au (%)

Ag (%)

Cu (%)

Gold sheet or vase

95.0

5.0

/

Soldering lump

88.2

4.7

7.1

Table 1. Semi-quantitative analysis of an ear-ring from Sainte-Colombe

Without doubt, these items from Vix and Sainte-Colombe were soldered with an intentional gold-silver-copper alloy with a lower melting point than the base metal and made especially for this purpose. In addition, perforations allow the expulsion of gases due to the heating during soldering have been identified on the Sainte-Colombe bracelets and the Vix torc. The head of the pin on the bracelet of Ensisheim (France) is made of several gold granules joined together (Fig. 6b). The elemental composition of all parts of the bracelet is on average 98.0% gold, 1.7% silver and 0.15% copper (LA-ICP-MS data). The solder shows a dendritic structure which had partially disappeared after being filed down. An enrichment in the copper (+ 0.8 %) content was shown for the solder compared with the granules by using [the] SEM-EDX (surface analysis) and [the] LA-ICP-MS (depth analysis). This small but significant discrepancy leads us to believe that the granules were fixed by metallic solder using a gold alloy obtained by melting down a fraction of copper with the gold the bracelet was made of. The ring from Cérilly (France) is closed by a solder (Fig. 6c). This transverse joint has a concave and smooth surface on one side and forms lateral bulges on the other side. The metal of the ring contains 97.7% gold, 1.5% silver and 0.50% copper (LA-ICP-MS data). The surface of the solder is characterised by increases between 2.5% and 4% in copper relative to the other parts of the gold object (SEM-EDX analysis). This chemical fingerprint and the morphology of the solder could be explained either by the use of a copper compound (copper- salt soldering) or by an alloy containing additional copper (metallic solder). Further research on the gold items and archaeological experiments are required to understand fully the joining techniques used during the Early Iron Age. In particular, more detailed analyses are required to understand the obviously different soldering traditions. Gilding is the practice of applying gold onto the surface of a less precious metal and has been used all over the world for thousands of years with various methods (Oddy 1993; Schorer/Schwab 2013). European Bronze Age gilding technology was based on mechanical techniques like wrapping a thin sheet around an object, bending the foil over the edges or fixing the foils with rivets (e.g. Oddy 1993). Most gilded Iron Age objects in

the West Hallstatt culture were still manufactured in this way, like the bronze belt-plate and the dagger, as well as the drinking horns, from the Eberdingen-Hochdorf tomb, the bronze brooches from Stuttgart-Bad Cannstatt and the iron belthook from the Grafenbühl burial in Germany, or French examples like the iron brooch from Vix and the silver brooches from Gurgy. Another much more sophisticated technique of gilding is based on interdiffusion of gold and silver. Heating a silver base tightly covered with a gold foil produces a strongly-bonded gilded surface. Gold and silver are completely miscible in the liquid and solid state and also exhibit a high atomic mobility that results in interdiffusion of the two metals to produce a substance-to-substance bond. The technique is called ‘diffusion bonding’ or ‘fusion-gilding’ (Oddy 1993, 176). Fusion-gilding is well documented for Greek and Etruscan items since the 7th century BC (Schorer/ Schwab 2013, 61). The only Early Iron Age example known until recently was the silver bowl with its gilded omphalos from the Vix tomb (Éluère et al. 1989, 16-27). During the West Hallstatt Gold Project it has been possible to identify several more silver objects that were gilded by diffusion bonding. Examples include small rings from the Gießübel-Talhau cemetery near the Heuneburg settlement and a fragment of a brooch from Châtillon-sur-Seine. Some bronze rivets from the princely grave of Asperg «Grafenbühl» are wrapped in diffusion-gilded silver foil (Schorer / Schwab 2013). WORKSHOP IDENTITIES Several features allow the interpretation of possible workshop identities for gold objects according to their similarity in form, size, decoration, tool-marks and material composition. Ornaments occurring in pairs or sets also clearly reflect workshop identities. In the West Hallstatt culture ear ornaments, arm-rings, brooches and decorative sheets often appear as pairs, while hair-rings or pin-heads occur in sets of several nearly identical items. Local similarities in technology, style and material composition of small rings show regional workshop identities. Although there are differences in the material composition between France and Germany, there are also similarities in shape and technology. This is the case for the small decorated rings from Erkenbrechtsweiler «Burrenhof» in south-western Germany and Haguenau in eastern France, and for another group of rings coming from Nordhouse in Eastern France and Steinheim, Mühlacker and Herbertingen in South-West Germany.

175

Furthermore, the identical use of specific goldworking techniques, such as joining two sheet shells by fitting, illustrates a common workshop area for Germany, France and Switzerland for the production of pin-heads and beads. The decorated hollow pin heads from Urtenen in Switzerland, Ditzingen-Schöckingen in Germany, and Nordhouse in Alsace, as well as the beads from Plichancourt in France attest identical mechanical joining and decoration techniques (Fig. 7a-c). A particularly striking case of actual workshop identities is demonstrated by the identical shape, technology and material of gold threads that survived from garments in two German princely burial mounds. The «Hohmichele» (Altheim-Heiligkreuztal) in the vicinity of the Heuneburg settlement is dated to Hallstatt D1 (around 600 BC), while the «Grafenbühl» near Asperg is dated to Hallstatt D3 (around 490 BC). These burials are about 90 km and 100 years apart, but the flattened gold threads from both sites are chemically and technologically totally identical (Fig. 8). These garments and other luxury textiles were obviously passed down through

several generations, exemplifying the regionally interconnected elite structure of Hallstatt society. SYNOPSIS Early Iron Age gold contains variable concentrations of silver, usually between 10 and 30 %, but also small percentages of copper; so we may conclude that the gold was not yet refined or alloyed and most probably derived from placer deposits. According to Axel Hartmann’s (1970) data the composition of Early Iron Age gold was significantly different from most Bronze Age gold, as well as from La Tène Iron Age gold, which was deliberately alloyed with copper. Gold is available within the territory of the West Hallstatt culture and there is little doubt that most of the goldwork found there was produced by indigenous craftsmen, because of its typology. The relatively small quantity of goldwork, the economic use of the metal in the form of thin sheet, and the high prestige of the objects all suggest a local origin for the gold.

Figure 7 A-C. Pin heads and beads. a) Pin heads from Ditzingen-Schöckingen; b) Pin heads from Nordhouse, c) Beads from Plichancourt. (Photographs a) P. Frankenstein, H. Zwietasch, © Landesmuseum Württemberg, Stuttgart; b)-c) B. Armbruster, © Musée de Colmar.

176

Figure 8 A-B. Gold threads from a) Altheim-Heiligkreuztal «Hohmichele» and b) Asperg „Grafenbühl« (photographs B. Schorer, © Landesmuseum Württemberg, Stuttgart).

The Early Iron Age goldwork is of high quality in technological and aesthetic terms, and was undoubtedly manufactured by metalworkers who were highly specialised and skilled. Using an interdisciplinary approach to reconstruct technological aspects of the chaîne opératoire and the choice of gold alloys, combining technological features like tool-mark analyses with experimental archaeology, the West Hallstatt Gold Project identified technical solutions for manufacturing problems and located workshops of the Hallstatt period. There is evidence for some regional workshops, which seem to have dispersed their products over a wide area, which could indicate that the craftsmen were mobile and were hired exclusively for contract tasks, or that they were permanent workers employed exclusively by regional elites. Rotary motion and the use of the spinning lathe were among the Early Iron Age inventions in goldworking. Technological traditions and innovations point to distinct applications of soldering and gilding techniques. Soldering with an intentional gold alloy has been identified as of «Atlantic» tradition, whereas technical innovations like copper-salt solder and diffusion gilding were [recognised as] «Mediterranean» influences witnessing long distance connections. The presence of silver objects is another indicator of Hallstatt links with southern cultures. The silver could have been imported directly from the Iberian Peninsula, or via Italy. Finally, the flattened gold threads from the Grafenbühl and Hohmichele tombs are the most impressive examples of the longevity and inheritance of property during the Hallstatt period.

pecially we have to thank T. Hoppe from the Landesmuseum Württemberg, Stuttgart and L. Olivier from the Musée d’Archéologie Nationale in Saint-Germain-en-Laye, who supported our studies during the whole project. BIBLIOGRAPHY ARMBRUSTER, B. 2004: Le tournage dans l’orfèvrerie de l’âge du Bronze et du premier Age du Fer en Europe Atlantique. In: M. Feugère/J.-C. Gérold (eds.), Le tournage des origines à l’an mil. Actes du colloque de Niederbronn, octobre 2003. Montagnac. 53-70 Taf. 2. — 2011: Approaches to metalwork - The role of technology in tradition, innovation and cultural change. In: X.-L. Armada Pita/T. Moore (eds.), Atlantic Europe in the First Millennium BC: Crossing the divide. Oxford: 417-438. ARMBRUSTER, B, COMENDADOR REY, B., PEREA, A. and PERNOT, M. 2003: Tools and tool marks. Gold and bronze metallurgy in Western Europe during the Bronze and Early Iron Ages. In: Proceedings of the International Conference «Archaeometallurgy in Europe», Milano 24-26 September 2003, Vol. 1. Milano: 255-265. ARMBRUSTER, B. and LOUBOUTIN, C. 2004: Parures en or de l’Âge du Bronze de Balinghem et Guînes (Pas-de-Calais) : les aspects technologiques. Antiquités Nationales 36 : 133-146.

ACKNOWLEDGEMENTS

BLET-LEMARQUAND, M., NIETO-PELLETIER, S. and SARAH, G. 2014: L’or et l’argent monnayés. In Ph. Dillmann /L. Bellot-Gurlet (eds.): Circulation et provenance des matériaux dans les sociétés anciennes. éditions des archives contemporaines, Paris: 133159.

Many thanks to the colleagues in the institutions that have enabled our investigations. Es-

BLET-LEMARQUAND, M., GRATUZE, B. LEUSCH, V. and SCHWAB, R. (in press): Material Sciences applied

177

to West Hallstatt Gold. In: B. Armbruster (ed.), Iron Age Gold in Celtic Europe, Forschungen zur Archäometrie und Altertumswissenschaft 6, (Rahden/Westf). BORN, H., SCHLOSSER, S., SCHWAB, R., PAZ, B. and PERNICKA, E. 2009: Granuliertes Gold aus Troia in Berlin. Erste technologische Untersuchungen eines anatolischen oder mesopotamischen Handwerks, Restaurierung und Archäologie 2: 19-30. ELUÈRE, CH. 1989: A «Gold connection» between the Etruscans and early Celts? Gold Bulletin 22(2), 1989, 48-55. ÉLUÈRE, CH., DRILHON, F., DUDAY, H. and DUVAL, A.-R 1989 : L’or et l’argent de la tombe de Vix. Bulletin de la Société Préhistorique Française 86 (1): 10-32. DRESCHER, H. 1987: Hallstattzeitliche Blechschmiede, Drechsler und Wagenbauer. In: Prunkwagen und Hügelgrab. Kultur der frühen Eisenzeit von Hallstatt bis Mitterkirchen. Kataloge des Oö. Landesmuseums, NF13, Linz: 41-54. HANSEN, L. 2010: Hochdorf VIII. Die Goldfunde und Trachtbeigaben des späthallstattzeitlichen Fürstengrabes von Eberdingen-Hochdorf (Kr. Ludwigsburg). Forschungen und Berichte zur Vor- und Frühgeschichte in Baden-Württemberg 118. Stuttgart. HARTMANN, A. 1970: Prähistorische Goldfunde aus Europa. Spektralanalytische Untersuchungen und deren Auswertung, Studien zu den Anfängen der Metallurgie 3. Berlin. Krausse/Ebinger-Rist 2011

KRAUSSE, D., EBINGER-RIST, N. 2011: Neues von der «Keltenfürstin» von Herbertingen Archäologische Ausgrabungen in Baden-Württemberg: 113–118. LEFEBURE, M. 1924: Le Tombeau de Petosiris. Le Cairo. ODDY, W. A. 1993: Gilding of Metals in the Old World. In: S. La Niece/P. Craddock (eds.), Metal plating and patination–Cultural, technical and historical development. Oxford: 171-181. PEREA, A. and ARMBRUSTER, B. 2011: Tomb 100 at Cabezo Lucero: new light on goldworking in the fourth-century BC Iberia. Antiquity 85: 158-171. RIETH, A. 1956: Die Bedeutung der Drehbank in vorgeschichtlicher Zeit. In: IV Congreso Internacional Ciencias Prehistóricas y Protohistoricas 1954. Zaragoza:141-144. ROLLEY, C. 2003 (ed.): La tombe princière de Vix. Paris. SCHORER, B. and SCHWAB, R. 2013: Neue Untersuchungen zu Vergoldungstechniken in der jüngeren Hallstattzeit. Restaurierung und Archäologie 6: 5769. SIEVERS, 1984: Die Kleinfunde der Heuneburg. Die Funde aus den Grabungen von 1950-1979. Heuneburgstudien V. Römisch-Germanische Forschungen 42, Mainz.

178

ESTIMATING THE ECONOMIC AND ECOLOGICAL IMPLICATIONS OF AN IRON SMELTING SITE IN BEREKET, SW-TURKEY Eekelers Kim*, Scott Rebecca**, Hodgins Gregory***, Muchez Philippe****, Poblome Jeroen*****, Degryse Patrick******

Abstract In the territory of ancient Sagalassos (SW Turkey), metallurgical waste is found at different localities. During previous surveys, an iron smelting site was found uphill of the Bereket Valley, 20 km SW of the monumental Hellenistic to Byzantine city of Sagalassos. The site consists of two adjoining depressions separated by a narrow plateau. The first depression displays a slag heap of ~110 m². Most slag were identified as tapslag. Chert fragments and limonite minerals were found in the second depression. This area is identified as an ore dressing site. Ore and slag were qualitatively analysed using a Bruker Tracer III-SD handheld portable X-Ray Fluoresence device (HH-pXRF). The chemical composition of the ore and slag were compared to other metallurgical waste in the territory of Sagalassos in order to provide information on potential distributions and exchange patterns. Bereket distin-

******  University of Leuven, Department of Earth and Environmental Sciences, Celestijnenlaan 200E, Leuven, Belgium. ******  University of Leuven, Department of Earth and Environmental Sciences, Celestijnenlaan 200E, Leuven, Belgium. ****** University of Arizona, Accelerator Mass Spectrometry Laboratory, 1118 East Fourth St., Tuscon, AZ 85721-0081, USA. ******  University of Leuven, Department of Earth and Environmental Sciences, Celestijnenlaan 200E, Leuven, Belgium. ****** University of Leuven, Department of Archaeology, Blijde Inkomstraat 21, B-3000 Leuven, Belgium. ******  University of Leuven, Department of Earth and Environmental Sciences, Celestijnenlaan 200E, Leuven, Belgium.

guishes itself by elevated Cr-values compared to the other sites and little to no exchange of raw material between sites is observed. A production estimate was made for the site to quantify the scale of the production. To estimate the sustainability of the iron production, wood consumption is compared to other wood consuming activities in the territory. The ecological impact of the Bereket iron production seems to be minor compared to other craft productions in the territory. This study offers insight into the economic and ecological importance of small iron smelting sites in the Sagalassos territory. Keywords: iron smelting, tap slag, HHpXRF, exchange patterns, production estimates, deforestation Resumen En la zona de la antigua Sagalassos (en el sudoeste de Turquía), se han encontrado residuos de actividad metalúrgica en diferentes localizaciones. En el curso de anteriores prospecciones, se halló un yacimiento de fundición de hierro en la parte alta del valle de Bereket, 20 km al sudoeste de la monumental ciudad helenística y bizantina de Sagalassos. El yacimiento consiste en dos depresiones contiguas separadas por una estrecha meseta. La primera depresión muestra un amontonamiento de escorias de unos 110 m2. En la segunda depresión se encontraron fragmentos de sílex y minerales de limonita. Esta área se identificó como un lugar de refinado de la mena. Menas y escorias se analizaron cualitativamente mediante un instrumento portátil de mano de Fluorescencia de rayos X, un Bruker

179

Tracer III-SD. La composición química de la mena y la escoria se comparó con otros desechos metalúrgicos de la zona de Sagalassos para obtener información sobre las potenciales distribuciones y sobre los patrones de intercambio. Bereket se distingue por los elevados valores de cromo comparados con los de otros yacimientos y se observa un escaso o nulo intercambio de materias primas entre yacimientos. Se hizo también una estimación de la producción del yacimiento para cuantificar la escala de la misma. Los análisis de polen y sedimentos en la zona manifiestan signos de fases de deforestación, sin embargo, éstas no parecen estar correlacionadas con la producción de hierro. Este estudio ofrece un mayor conocimiento de la importancia económica y ecológica de los pequeños yacimientos de fundición del hierro en la zona de Sagalassos.

terpreted as proof of intense contact with the administrative centre of the territory at Sagalassos. Although the valley was subjected to an intense archaeological survey, the iron smelting site uphill was only briefly mentioned in earlier reports. This site, however, is interesting for the current research project which aims to reconstruct the iron craft production in the Sagalassos territory. Since it is the only «intact» smelting site in the territory, it offers the opportunity to characterize the technology, expose potential exchange patterns and assess the impact of this type of production on the local environment. Moreover, as sampling on site is no longer allowed, it thereby offers the opportunity to use a portable X-Ray Fluoresence apparatus (pXRF) in the field to analyze the iron slag. STUDY AREA AND MATERIALS

INTRODUCTION The frequent occurrence of metallurgical waste in several excavation layers, dating from the 1st to the 7th century AD, proves continuous iron working in the Roman to Early Byzantine town of Sagalassos. Vast amounts of smithing slag, dumped in fills, used as road levellers or associated with workshops, are found with rejected hematite ore (Kellens et al. 2003). These slag show a chemical composition similar to other Roman slag, with high FeO, SiO2 and CaO values and moderate Al2O3, MnO, MgO and TiO2 contents (Degryse et al. 2003a). Although traces of a hematite mineralization are found just east of the monumental city, no evidence of smelting activity is found in the close surroundings of Sagalassos. Screes cover the area where the mineralization was found and therefore potential smelting sites. 5 km south of the city, at the Bey Dağları massif, two smelting sites were discovered. Bloom, kiln fragments, furnace cooled slag and tuyères testify to this activity (Kellens et al. 2003). Ceramics date these sites from the 6th to 7th centuries AD. The chemical composition of the slag differs significantly to that of Sagalassos, with elevated TiO2, V2O5 and Zr values (Degryse et al. 2003a). In Dereköy, one of these two sites, an ulvospinel-magnetite placer deposit explains the specific chemical composition of the slag. During a survey campaign in 2013, an iron smelting site uphill of the Bereket valley, 25 km SW of the monumental city, was revisited. In the early nineties, archaeological survey in this valley found funeral remains, as well as a high density of pottery. These remains were dated from the Hellenistic to Roman Imperial period and pointed to the presence of elite elements in the rural territory of Sagalassos (Waelkens et al. 2000; Kaptijn et al. 2013). Moreover, large quantities of Sagalassos red slip tableware were found and in-

The Bereket Valley is located in the southwestern part of the 1200 km² territory of Sagalassos (fig. 1). The valley is bordered by two mountain ridges, the Beșparmak Daǧlari in the northeast and the Kokayanik Tepe in the west. The smelting site is located on the latter at ~1635 m above sea level. The present-day vegetation consists of mountainous forest species like Juniperus excelsa, Pinus brutia, Pinus nigra and Cedrus Libani, together with shrub vegetation (Kaniewski et al. 2007). Winters are long and cold with a dense snow cover, while summers are short and dry (Kaptijn et al. 2013). The local geology consists of a complex amalgamation of tectonic units. Bereket is located on the Lycian Nappe, which overthrust the Bey Dağları Massif in the east (fig. 2) (Degryse et al. 2003b). Associated with this displacement, Cretaceous limestone were thrust over the ophiolitical mélange in the front zone of the Lycian Nappe (Muchez et al. 2003). The contact between these two geological units is characterized by an alteration zone with hematitisation, dolomitisation, silicification and intense ferroan calcite veining. This contact zone is the result of the upward migration of iron-, magnesium- and silica-rich fluids from the ophiolitic mélange into the limestones (Muchez et al. 2003). In the Sagalassos territory, this zone is associated with different types of mineralisations. Around the city of Sagalassos pyrolusite (MnO2), magnetite (Fe3O4), hausmannite [(Mg, Fe)(Al, Fe, Cr)2O4)] and hematite (Fe2O3) are found. On the pass of Başkoy, 20 km east of Bereket, a magnetite, chromite (FeCr2O4) and magnesiochromite (MgCr2O4) mineralisation is located (Degryse et al. 2003b). During the Pliocene, normal faulting caused the deposition of screes and lacustrine marls along the margins of the Beşparmak mountain range. In these marls, a complex badland morphology developed.

180

Figure 1. Upper: Geographical setting of the Sagalassos territory in the Eastern Mediterranean Sea. Lower: Geographical setting of Bereket in the territory of Sagalassos.

181

FIGURE 2. Geological setting: the site located on weathered limestone (after MTA, 1997).

The smelting site consists of two adjoining depressions separated by a narrow plateau. In the northern depression, a slag heap of ~110 m² can be found (fig. 3). The metallurgical waste has an average size of 10-15 cm diameter and weighs between 100 and 1100 g. The surface is characterized by a typical «pahoehoe» lava texture. The slag are dense with large pores limited to the top of the slag. Macroscopically, the interior of the slag seem rather homogeneous. Some slag show iridescent oxidation colors. The slag are identified as tapslag (Cleere 1976) (fig. 3). Uphill of this heap, a walled structure was identified which can be associated with the place where the actual smelting was conducted. Other smelting sites in the territory are also found on the edge of hills, to profit from the natural draft in the smelting furnaces (Rehder 2000). Bloated limestone is found near the walled structure. Chert fragments and discarded limonite (fig. 3) ore were found in the second depression. This depression is recognized as the ore dressing site, where the ore was separated from the gangue material and possibly roasted before smelting. The plateau that separates the two depressions seems to be flattened in comparison to similar

morphologies in this badland area. The plateau is surrounded by large in-situ limestone rocks with smaller blocks in between. The latter ones can not be transported by natural, geomorphological processes. It seems that this plateau is manmade, perhaps for temporary housing. Remarkably, no ceramics or other datable artefacts were found. It is however difficult to assess to what extent the site is intact. Melting water in spring and flash floods in summer cause removal of soil material and probably archaeological material as well. Moreover, slag are found in the recently constructed road. METHODOLOGY ARCHAEOLOGICAL SURVEY The site is divided into three functional areas: the slag heap, the plateau and the ore dressing site. Since sampling is no longer allowed, following recent changes to the Turkish Republic legislative framework, the site was carefully mapped and surveyed. Site boundaries of each functional area were determined based on the density of

182

FIGURE 3. Left: drawing of the slag heap. Right up: limonite ore found in the slag heap. Right under: tapslag found at the site.

the subsurface artifacts. The areas were mapped along a NS-EW trajectory. The area was divided into 15 squares of 6 m². One representative square was selected and along six transects, slag larger than 2 cm diameter were counted and described. CHEMICAL AND STATISTICAL ANALYSIS Handheld portable X-Ray Fluorescence spectroscopy (HH-pXRF, Bruker III-SD) was used to determine the chemical composition of the slag and ore. Samples were crushed by hammering and sieved to a 2 mm mesh. This provided homogenization and showed the most precise results under laboratory conditions (Scott et al. 2016a). To obtain bulk chemical compositions, 23 slag samples, 1 ore and 1 chert sample were first measured under vacuum conditions, at 40 kV and 5 µA. A vacuum reduces the absorption of light element X-rays by air (Kaiser and Wright 2008). Next, samples were analyzed under vacuum at 9 kV and 35 µA, to enhance lighter and minor elements. Artax software was used for further interpretations. Since there is only a calibration available for slag from other sites in the territory (Sagalassos and Bey Da lari) (Scott et al. 2016b) that display a different matrix and chemical composition compared to Bereket, only qualitative chemical results, based on peak areas, are obtained. The quality of the

pXRF measurements are based on the repeatability of the measurement. During the measurement of 25 Bereket samples, 3 «in-house» standards (slag from other archaeological contexts) were measured several times. The relative repeatability error is expressed as %RSD = 100*(standard deviation/mean). A large dataset, including pXRF data of smithing slag from Sagalassos and furnace cooled slag from the Bey Da lari massif, was established during previous campaigns. Also, slag samples brought to Belgium for analysis during earlier campaigns, were measured by pXRF and added to the database. To allow comparison between different sites, the data was ratioed against Rhodium, which is the material the tube is made from and will thus be the most constant over all measurements. To determine the internal structure of the dataset in a multi-element space, Principal Component Analysis (PCA) was selected. The PCA is graphically expressed in a biplot. The two axes are the two main components, which explain the most inherent variability possible. The length of the arrows is proportional to the variables included in the component. The angle between two arrows are a measure of correlation between two variables (Reimann et al 2008). Data were transformed prior to the analysis by using a centred log ratio (clr) transformation to «open» the data and reduce the effect of outliers (Reimann et al 2008). Moreover, Al/Rh was excluded from further analysis. Although measured under vacuum

183

conditions, Al is too light to acquire precise measurements. To graphically present the differences between the different groups obtained by PCA analysis, boxplots were constructed. Boxplots are an informative graphic for displaying the data distribution and are built around the median and the 25th-75th percentile (Reiman et al 2008). All statistical analyses were performed using R software (R Core team, 2012). C measurements

combustion products. δ13C measurements were obtained using a VG Instruments isotope ratio mass spectrometer. CO2 was converted to graphite using the method of Slota et al. (1987) and then measured on a NEC Tandetron Accelerator Mass Spectrometer. 14/13 ratios were measured with respect to OxI and OxII standards and corrected for isotope fractionation. Radiocarbon measurements were calibrated using OxCal4.2.4 and the IntCal13 database (Reimer et al. 2013).

14

One slag sample was exported for analysis to Leuven in the early nineties after previous archaeological surveys in the Bereket Valley. Three sub samples were taken from this one slag and powdered to homogenize the sample. The 14C measurements were conducted by the Arizona AMS laboratory. Crushed samples were leached with 85% phosphoric acid and combusted according to Jull et al. (1993). Samples were placed in an aluminium crucible and mixed with 2 g of high purity iron accelerator. Combustion took place in a radio frequency furnace, in a stream of high purity oxygen. They were heated to ~1700 °C to extract carbon from inclusions within the molten slag. Liberated CO2 was cryogenically purified from oxygen and other

Production estimates The slag heap was quantified by multiplying the slag volume (m³) with the specific density (g/ cm³) of the slag and the packing factor of the heap. Bachman (1982) defined the specific density of a dense tap slag at 3.5 g/cm³. The packing factor chosen was 0.5 as determined by Cleere (1976) for sites with seasonal production. This is assumed to be true for Bereket, considering the long snowy winters in the territory. The quantification of the produced bloom, raw iron and charcoal is based on ratios obtained via experimental archaeology (table 1). The minimum and maximum calculated values are given. Since the site could not be sampled, no charcoal analysis was performed. During previous research it was

Description

Ratio

Metric ton

References

Slag: bloom

3:1

13

Cleere 1971

4:1

10

Joosten et al. 1998; Thomas and Young 1999

4

Crew 1991

45% of the original bloom

4-6

Juleff 1998; Sim 1998, Saunder and Williams 2002

65% of the original bloom

6-8

9:1

88-117

12:1

117-155

16:1

155-207

7:1

612-1450

10:1

874-2072

250 ton branch wood/ha

2.5-9

Rehder 2000; Charlton personal communication

64.03 ton/ha (above ground biomass, young forest)

10-36

Whittaker and Woodwell 1969; Fulé et al. 2008

180 ton/ha (above ground biomass, old forest)

3-13

Woodwell 1969; Fulé et al. 2008

Slag: smithed object 10:1

Charcoal:bloom

Wood:charcoal Hectares of forest

Cleere 1976, Crew 1991; Juleff 1998; Rehder 2000

Cleere 1976; Rehder 2000

Table 1. Production estimates based on ratio’s found in literature

184

determined that slag found at the town of Sagalassos show traces of Pinus and Juniperus (Schoch 1995). The Kokayanik Tepe was originally dominated by Pinus and Q. coccifera. Also Alnus, Ulmus and Juniperus are accounted for (Kaniewski et al. 2007; Bakker et al. 2011). For these reasons, above ground biomass of Mediterranean pine and oak forests (Wittaker and Woodwell 1969; Fulé et al. 2008) are used in the production estimates. RESULTS Bulk chemistry

Ti and Fe show the highest peak area counts in the ore. The slag show a large spread in the peak area counts for Fe (fig. 6). Cr shows elevated counts per second in the chert. The lowest peak area counts for Ca are measured in the ore, followed by the chert. The slag show variable amounts of Ca. The highest peak area counts for K are observed for the chert (with the exception of one slag), while the ore shows the lowest values. Finally Si shows the lowest peak area counts for the ore and chert. The Si peak area counts for the slag samples show a large spread. Slag sample BER 16 shows high peak area counts for Si, Ti, K, V, Cr and Sr and low peak area counts for Fe compared to the other slag samples. To assess whether the chemical composition of the slag reflect the production process (contribution of the furnace lining, potential fluxes) or/and the ore, a spidergram (fig. 7) is plotted. In this spidergram, the average slag (excluding BER 16) is normalized against the ore. The slag is depleted in Al and Ti compared to the ore. Cr content is similar to the ore. Ca and Si are enriched in comparison to the ore.

The XRF spectra of the slag, ore and chert show the presence of Al, Si, P, S, K, Ca, Ti, Cr, Fe, Sr (fig. 4, table 2). The ore also shows a peak matching As. The presence of V is difficult to assess, since this element overlaps with Ti and Cr. The Fe peak at 6.40 kV shows a shoulder, corresponding to trace amounts of Mn. Peaks at 7.46 kV and 8.02 kV and 12.82 kV and 13.47 kV are sumpeaks of Ca and Fe respectively. The Compton scatter shows that ore, chert and slag are matrix matching, with chert and ore being the two endmembers (fig. 5). This implies that a direct qualitative comparison between the chert and ore should be done with caution, since matrix effects need to be taken into account. The relative repeatability error on the peak areas is for most elements