Fingerprints in Iron: Identifying the Production Origins of Iron Artifacts with Major Elemental Analysis of Slag and Slag Inclusions Abstract Iron was an important component in many preindustrial Old World economies and was frequently cited as a material of trade. The ability to accurately identify the site or region of an iron oďjeĐt’s primary production can offer new insights into the relationships between economic agents, whether individuals producers or whole societies. A multivariate statistical strategy for comparing the major elemental fingerprints of bloomery iron production slag with those of the slag inclusions found in the metal provides an effective means of achieving this goal. Application of the strategy to the products of three experimental smelts demonstrates its efficacy. Its archaeological potential is shown through two case studies from the British Iron Age. The first considers the production origins of an iron object assemblage from Beckford. The second considers the provenance of Iron Age swords. Bloomery Iron production systems Bloomery, or direct process, iron production involves the solid state reduction of iron oxides into a mass of metallic iron called a bloom. During, the smelting process, the majority of the volatile elements of the ore and charcoal fuel fume out of the furnace, while most of the reduced compounds join the bloom, and non-reduced compounds from a liquid slag. Melting furnace walls may also contribute non-reduced compounds to the slag. The slag provides a medium through which reduced iron particles sinter and form the bloom. As a consequence, some slag inevitably remains entangled within the bloom. Most of the entrapped slag is removed as the bloom is smithed via a series of heating and hammering tasks that ultimately transform the metal into its final form. Some slag remains as small inclusions in the iron through out all stages of smithing. New inclusions may accumulate from smithing additives or fuel ash and clay contaminants from the smithing hearth. A simplified model of the system is shown below. Michael F. Charlton (mike.charlton@ucl.ac.uk), UCL Institute of Archaeology Eleanor Blakelock, Archaeometallurgy Tim Young, GeoArch Marcos Martinón-Torres, UCL Institute of Archaeology Janet Lang, British Museum Sarah Paynter, English Heritage Fingerprinting bloomery iron Just as human beings possess similar, albeit unique , patterns of friction ridges on their fingertips, so bloomery slag produced at different sites and regions possess similar, but no less unique, patterns of elemental composition. The chemical discrimination of iron smelting slag from different sites and regions has been repeatedly demonstrated in multiple archaeological contexts (Buchwald 2005; Charlton 2007; Morton and Wingrove 1972; Paynter 2006; Serneels 1993). The chemical fingerprints of bloomery sites and regions are then transferred via slag inclusions to every piece of iron produced at the smelting centers in the same way criminals leave traces of their illicit activities at the scene of a crime. Of course, it is impossible to identify the criminal without also having examples of a known fingerprint, an unmarred fingerprint at the crime scene and a sufficient number of comparable points of reference on the prints themselves. Again, the same is true when investigating the production origins of bloomery iron. There must be a suitable reference collection of smelting slag, an assurance that the analyzed slag inclusions are uncontaminated smelting slag, and a means of comparing multiple elements simultaneously. Charlton et al. (2012) introduced a multivariate strategy for identifying the primary production origins of bloomery iron objects. First, a Linear Discriminant Analysis (LDA) is conducted on a matrix of element compositional data obtained from slag samples with known production origins. A kernel density estimation (KDE) is conducted and the result projected into the multivariate space created by the LDA. The KDE defines the provenance fields, or fingerprints, of the bloomery iron. Next, a matrix of compositional data obtained from an oďjeĐt’s slag inclusions are filtered to remove non-smelting and contaminated inclusions. The data from the remaining smelting-derived slag inclusions are then projected into the LDA space and compared to known smelting provenances. This strategy was applied to materials from three experimental ironmaking campaigns (XP17, XP23 and XP26) conducted at the National Museum of Wales in St. Fagans. While the general parameters of the experiments were similar in all cases, XP17 used a manual bellows blown furnace and was charged with a Welsh Blaenafon siderite ore. XP23 and XP26 used a continuous fan blower and was charged with a South African Sishen hematite. The iron, slag and slag inclusions produced by the these campaigns were selected by Blakelock et al. (2009) for chemical analysis. The LDA of smelting slag compositional data using seven elements (represented as oxides) resulted in good discrimination of XP17 from XP23 and XP26 smelting slag, and partial separation of XP23 and XP26. Projected into the same multivariate space, the chemical fingerprints of the object slag inclusions were correctly assigned to the experiment from which they derived. A graph highlighting the results is shown below. Case study 1: The iron objects of Beckford Paynter (2006) compiled a robust matrix of slag chemical compositions derived from bloomery sites across Pre- Roman and Roman Iron Age Britain, exposing a number of regional patterns. Prior research by Hedges and Salter (1979) showed that slag inclusions from currency bars excavated at Beckford, UK formed a coherent chemical group. Here, we use the chemical fingerprints from a partial set of Paynter’s (2006) bloomery regions to assess the provenance of additional iron artifacts excavated at Beckford using slag inclusion data compiled by researchers at English Heritage. The sites used to define the geochemical regions are plotted on the map of southern Britain to the left. The location of Beckford is also plotted and is observed to be in close proximity to the Forest of Dean Carboniferous and Bristol-Mendip (FDCBM) geochemical region. The graph of Linear Discriminant axes 1 and 2 (right) displays the chemical fingerprints defined by KDEs for each geochemical region as well as a projection of the average object slag inclusion chemistries into the same space. The majority of the objects plot within or near the FDCBM boundaries. Approximately 70 % of the objects can be identified to the local bloomery region while 30 % of the objects are non-local in origin. Sword Date Group – tradition Probable Origin River Thames (9) 4 th – 3 rd BC A – southern Southern Britain?? Orton Meadows (31) 4 th – 3 rd BC A – southern Southern Britain?? Isleham (101) 1 st BC – AD 50 D – southern FDCBM Orton Meadows (122) 1 st BC – AD 50 D – southern MJ and SHTS Grimthorpe (177) 2 nd BC E – northern KTS Melsonby (199) 1 st AD F – northern KTS and MJ Sadberge (207) 1 st AD F – northern SW Stanwick (245) Late 1 st AD H – merged FDCBM Case Study 2: British Iron Age Swords Lang (2006) presented a detailed technological study of a series of British Iron Age swords housed in the British Museum. The swords date from the middle 4 th century BC to the late 1 st century AD and are drawn from the two dominant stylistic distributions (southern and northern). One sword represents the merged tradition of the 1 st century AD. As part of the investigation, slag inclusion chemistries were analyzed by SEM-EDS. Here, we make a preliminary assessment of sword provenance using the chemical fingerprints detected by Paynter (2006) for British bloomery regions. Results are shown in the surrounding graphs and summarized in the table below. Tentative provenance hypotheses are constructed for all swords. Additionally and in agreement with Lang’s (2006) assessment, it is clear that some swords are constructed from iron from multiple batches with more than one regional origin. Discussion and Conclusion: The chemical fingerprints imparted to bloomery iron via slag inclusions provides the essential evidence for hypothesizing the production origins of preindustrial iron objects. The investigator need only apply an effective multivariate lens to resolve the identifying characteristics linking the iron to its source. This proves true even when considering a small number of major elements. The efficacy of this approach is demonstrated above by its application to the products of experimental iron smelting campaigns. The potential benefits of the multivariate provenance approach are highlighted by its application to the iron assemblage of Beckford and the series of iron age swords. However, there are three important caveats regarding the analysis of the archaeological case studies. First, sample sizes for the inclusion data are too small in many cases to provide a clear chemical fingerprint—without data for at least 30 inclusions, deciding which are contaminated becomes a matter of interpretation. It also becomes more difficult to quantify the distribution of the inclusions and calculate a true mean. Second, the regions used in this provenance analysis are not exhaustive. The consideration of additional bloomery sites and regions could add greater clarity to our analysis. Finally, the analyses used here were not conducted with provenance determinations in mind. As a consequence there may be some instrument biases applied to the elemental compositions that are difficult to assess. Nonetheless, it is clear that the overall patterns are consistent. The further development of bloomery fingerprinting will provide archaeologists with greater insight on the preindustrial iron economy and shed new light on extinct exchange networks. Bellows Bloom Object Bar Billet Product 1 Smelting slag Smithing slag Secondary Tertiary Charge Smithing Ore Charcoal References: Blakelock, E., Martinón-Torres, M., Veldhuijzen, H.A., Young, T., 2009. Slag inclusions in iron objects and the quest for provenance: an experiment and a case study. Journal of Archaeological Science 36, 1745-1757. Buchwald, V.F., 2005. Iron and Steel in Ancient Times. The Royal Danish Academy of Sciences and Letters, Copenhagen. Charlton, M.F., 2007. Ironworking in Northwest Wales: An Evolutionary Analysis. Unpublished PhD thesis, University College London. Charlton, M.F., Blakelock, E., Martinón-Torres, M., Young, T., 2012. Investigating the production provenance of iron artifacts with multivariate methods. Journal of Archaeological Science 39, 2280-2293. Charlton, M.F., Crew, P., Rehren, Th., Shennan, S.J., 2010. Explaining the evolution of ironmaking recipes —an example from northwest Wales. Journal of Anthropological Archaeology 29, 352-367. Hedges, R.E.M., Salter, C.J., 1979. Source determination of iron currency bars through analysis of the slag inclusions. Archaeometry 22, 161-175. Lang, J., 2006. Appendix 1: The technology of some of the swords, in I. M. Stead, British Iron Age Swords and Scabbards, London, British Museum Press, 85-114. Morton, G.R., Wingrove, J., 1972. Constitution of bloomery slags: part II: medieval. Journal of the Iron and Steel Institute 210, 478–488. Paynter, S., 2006. Regional variations in bloomery smelting slag of the Iron Age and Romano-British periods. Archaeometry 48, 271-292. Serneels, V., 1993. Archéométrie des Scories de Fer. Recherches sur la Sidérurgie Ancienne en Suisse Occidentale. Collection Créée par Colin Martin, Lausanne. Slag inclusions in an iron matrix (Blakelock et al. 2009)