Geochemical Link Between Slag Inclusion and Hammerscale, a Potential New Tool to Better Understand the Iron Supply of an Archaeological Workshop

Hammerscale is the waste formed during hammering of hot iron on the anvil. It is mainly composed of iron oxides (wüstite, magnetite, hematite) with a variable amount of silica-rich component. This research focuses on the potential geochemical link between hammerscale and smelting derived slag inclusion contained in the metallic iron. In the frame of previous research, the smithing practice of African Dogon craftsmen (Mali) has been recorded in details. Different samples were collected including, slag, hammerscale, metal. Two different groups of iron bars produced by different bloomery process and more than one hundred experimental pieces of hammerscale were chemically investigated in situ by Energy Dispersive X-ray Spectroscopy (SEM–EDS) and by X-ray Fluorescence spectrometry (WD-XRF) in Fribourg / Switzerland. A small part of them was also analyzed by Laser Ablation Inductively Coupled Plasma Mass Spectrometry (LA-ICP-MS) in Orléans / France. The results show a strong contamination of the hammerscale by the working environment (flux, ashes, sand, etc.) impacting the concentrations of most of non-reducible elements, in particular the most abundant ones (Si, Al, K, Ca). However, several in situ analyses revealed abnormal high contents of Ti, Mn, Zr in contrast to the chemical composition of the forging environment. The concentrations of these elements are affected by contamination but the ratios, in this context, remain unchanged. TiO2/Zr ratios measured in slag inclusion are consistent with the ratios measured in hammerscale produced from the same iron type. Those results support that hammerscale could still carry the chemical signature of the smelting system and thus the geochemical fingerprint of the iron worked. Applying this methodology on archeological hammerscale would make it possible to assess the chemical diversity of worked irons and to provide information about the primary ore sources. Improving our current knowledge of the workshop iron supplies and therefore the circulation of iron in the past times.


Introduction
During the smelting of iron ore by the bloomery process, a solid bloom and a liquid waste, the slag, are produced. Some of these slags are trapped in the bloom. After the smithing stage some slag inclusions remain in the metal product. Several studies have shown that there is a chemical link between these slag inclusions in the metal and the smelting system they origin from [1][2][3]. From a chemical point of view the smelting system is the stock of chemical elements involved in the process. This stock is composed by the ore, the fuel (mainly charcoal), and eventual contaminant like sandy materials. The different chemical elements contained in the stock do not behave in the same way under the temperature and oxygen fugacity conditions required for the bloomery process. Fe oxides are partially reduced to the metallic state.

Materials and Methods
All analyzed hammerscale is the result of the work of blacksmiths from Pays Dogon, Mali [7]. They were produced from three different forging operations N°07, N°08, N°105 (Table 1). During each forging operation one type of iron bar was forged (Fig. 2). The African iron 1 and African iron 2 worked during the operation N°07 and operation N°08 were produced by bloomery process [7]. They have been bought in a specialized shop in Bruxelles. The iron ore used and the smelting conditions are unknown. These African iron bar could be related to these bars which have a function of currency, and which are found from the east of Nigeria to the west of Cameroon, between the 16 th and the twentieth century [20]. Straight African iron 1 bars are found among the Yoruba, an ethnic group of Nigeria whereas bipyramid African iron 2 bars are found among Fulani, Jukun, Mumuye ethnic groups. Both of African iron type, 1 and 2, are between 38 and 40 cm long. African bars type 1 show a poorly compacted appearance in contrast to African bars type 2 which show a laminated appearance and several weld lines [7]. The modern steel worked during operation N°105 has been bought from an industrial company. The hammerscale produced from the modern steel will be use as a comparison tool. The aim is to follow the influence of slag inclusion by comparing hammerscale potentially affected by slag inclusion participation (African iron) and hammerscale non-affected (modern steel). Each forging operation has been described in detail [7], the working temperature during forging, the technical processes (addition of flux The study of the products provides information on the smelting system (ore, flux addition, fO2). The aim of the study is to explore if a link exists between hammerscale and smelting slag through slag inclusions during the forging, welding or not) or even the configuration of the working team. During each forging, a certain number of hammerscale have been produced and collected. They have been, washed and sorted according to the forging sequence, "stage 1," beginning of forging, corresponding to the shaping of the iron bar (pre-forming) and "stage 2," end of forging, the final step to obtain the object (forming).

Slag Inclusions
Several sections extracted from the 3 different bar types have been previously prepared for metallographic studies. Each sample come from the extremity of an iron bar, in order to keep the iron bars in their entirety. They were then fixed in a conductive epoxy resin before being polished to 1/4 µm using diamond suspensions. Finally, in order to reveal the iron microstructures a 3% nital etch was performed [7]. Slag inclusions were analyzed by Energy Dispersive Spectrometry (EDS) coupled with a Scanning Electron Microscope (SEM) FEI XL 30 Sirion FEG at the Department of Earth Sciences of the Université de Fribourg, Switzerland. Acquisition and processing of X-ray spectra was processed with the AZtec software developed by Oxford Instruments. All analyses were performed after calibration of the instrument with a cobalt standard sample. Measurement settings remained constant during analyses, with a working distance of 8.5 mm, a current of 20 kV, a dead time of 30%, and an acquisition time ranging from 60 to 100 scd. For high concentrations (> 1%) uncertainty considered is 5%, it can be as high as 10% for low concentrations (< 1%). Detection limit is 0.15 wt.%. For each bar, 40 inclusions distributed over the entire surface of the section were studied following the methodology developed by [2].

Hammerscale
Hammerscale was prepared for SEM-EDS and Laser Ablation Inductively Coupled Plasma Mass Spectrometry (LA-ICP-MS) analysis. Hammerscale was fixed in epoxy resin and ground to 600 µm and then 250 µm with abrasive silicon carbide disks to achieve the desired sections. Polishing to 9, 3, 1 and ¼ µm was then performed with diamond suspensions. A 10 nm carbon layer was finally deposited on the polished surface. The SEM-EDS analytical parameters used for the analysis of slag inclusions were equally applied for the analysis of hammerscale. LA-ICP-MS analyses were performed at the Centre Ernest Babelon (UMR-7065 IRAMAT CNRS, Orléans, France). Laser pulse was set to 6 Hz with an ablation time of 33 s and a 100 µm spot diameter. Analyzed hammerscale was ablated between 2 and 5 times on the sample surface and the results were averaged for data evaluation. The variability coefficient of the chemical elements concentrations measured in a hammerscale is lower than 0.1 for 46 chemical elements when iron content is lower than 95% (Fig. 3). Some elements (Ag, In, Sn, Au, Pb, Bi) are quite variable and therefore have not been taken into account for the interpretation of the results. These high variation coefficients, up to 1.3 for In or Au are the consequence of the contamination of the device by routine analysis performed on metallic coins enriched in these elements. We used the quantification procedure developed by [21] and [22]. A second part of hammerscale was prepared for total analysis by X-ray fluorescence spectrometry (WD-XRF). Hammerscale produced during different forging operations were divided into subgroups of 0.8 g. Each group was then crushed with an agate mortar. After calcining the samples at 900 degrees, 0.7 g of sample powder was mixed with 6,65 g of lithium tetraborate (Li2B4O7) and 0.35 g of lithium fluoride (LiF) to produce a glass tablet. The specimens were analyzed with a Zetium X-ray WDS (Malvern-PANalytical, Malvern, UK) at the Department of Earth Sciences of the Université de Fribourg. The major elements (Si, Al, Fe, Mg, Ti, Ca, Na, K, P, Ti) and some of the trace elements (Cr, Zr, V, Sr, Y, Ni) were measured with the routine program NewBasalt calibrated with 36 international references [23]. The uncertainty for major elements is about 1% and for trace elements 10%.

Modern Steel and African Iron Slag Inclusions
Modern steel bars were not analyzed further since information of the chemical composition was already available. According to the manufacturing company, Mn is the most abundant alloying element (Mn = 1.4 wt.%).
Slag inclusions measured in 9 African iron bars type 1 (B101, B102, B103, B105, B106, B108, B112, B117, B118), show overall high titanium and zirconium contents (up to 35 wt.% TiO 2 for the B106 bar and up to 6 wt.% Zr for the B108 bar) (Fig. 4). This high content can be explained Fig. 2 The three different bar types African iron bar type 1, African iron bar type 2 and modern steel bar by the use of an ore rich in these elements, like a placer deposit that accumulate heavy density minerals, magnetite (Fe 3 O 4 ), rutile (TiO 2 ), zircon (ZrSiO 4 ) … The ratios of certain elements measured (Zr, Ti, Mn, Si, Al, K) for different slag inclusions in the same iron bar show a small variation, B105 varies from TiO 2 /Zr = 4.4 to TiO 2 /Zr = 3. They are two exceptions (B108, B102). There is however a slight variability of the ratios between the different iron bars studied. This most probably reflects the use of different batches of the same ore.
Slag inclusions measured in 6 African iron bars type 2 (B202, B205, B210, B214, B210, B217) show a more heterogeneous chemistry between the individual bars (Fig. 5). On the one hand B210, B212, B214 are very concentrated in MnO (up to 85 wt.%) while TiO 2 and Zr are absent. On the other hand, slag inclusions analyzed in B217, B205, B202 are concentrated in TiO 2 (up to 50 wt.%) and Zr (up to 8.5 wt.%) while MnO is almost absent. TiO 2 /Zr is relatively constant for slag inclusions measured in the same iron bar but quite variable between the different bars (B217: TiO 2 / Zr = 16; B205: TiO 2 /Zr = 4; B202: TiO 2 /Zr = 0.5). The heterogeneous chemistry of slag inclusions from iron type 2 will allow a discrimination considerably less precise than the hammerscale from iron type 1 or from modern steel.

Hammerscale
Hammerscale have different morphologies, they can be distinguished in three categories, flake, spherical and irregular (Fig. 6). Their formation modes are discussed in literature  [15,24,25]. Oxide particles, formed by solid state diffusion of oxygen through the metal, can be detached from the iron metal by desquamation or by hammering. In most cases it is flake particles that are ejected, they were collected in highest abundance [7]. They are composed of three superposed layers of iron oxides, a layer of hematite Fe 2 O 3 (1%), a layer of magnetite Fe 3 O 4 (4%) and a layer of wüstite FeO (95%) ( Fig. 6) with columnar crystals [26]. In archeological samples studied [14] a silicate glass is often present at grain boundaries, which is not observed in modern hammerscale [27]. Spherical hammerscale is associated with the expulsion of a liquid present on the surface of the worked metal, which cools in contact with air. They are also composed of iron oxides and generally a higher abundance of silicate glass. This have been observed in samples studied and in literature [14,15]. Irregular hammerscale seems to be the result of a mixture between a solid and a melted silica component. In all types of hammerscale it is possible to find metallic pieces inherited from the worked iron bar. Different morphologies of hammerscale can be formed under the same hammer blow [15]. Nevertheless, during the work of a dirty iron containing lot of slag inclusions or in the case of fluxes addition like sand, the liquid component on the metal surface is more important than in normal forging situation. Spherical hammerscale formation in these cases is then favored.

Role of Contamination in Hammerscale Composition, Geochemical Modeling
Hammerscale from those three forging operations where no fluxes were intentionally used show a clear evidence of contamination from the forging environment. Quartz has been found in many hammerscale samples and sometimes in abundance. They might be incorporated when hot hammerscale falls on the ground. This hypothesis can be supported by several observation that show adhering quartz grains on the bottom side of the particles (Fig. 7). The quartz grains show disequilibrium structures especially at their boundaries. This is due to a reaction with iron oxides contained in the hammerscale. This reaction well described in literature [28], will produce either dendritic fayalite (Fe 2 SiO 4 ) or glass. The driving parameter is the cooling rate for dendrite formation. The forging environment influences the chemical composition of hammerscale in a significant way. This contamination seems to be mostly siliceous. In our case there is no petrological evidence of CaO, K 2 O, Al 2 O 3 enriched particles incorporated in hammerscale that could be related to the hearth clay walls or the charcoal ashes used during forging operations (Table 2). But in other situation these potential contaminants, quite concentrated in some major elements such as CaO, K 2 O, MgO, Al 2 O 3 could irremediably affect the chemical signature of the potential slag inclusions for these specific elements. Hammerscale produced during the three experimental forging processes show an unexpected high concentration in Ti, Mn, Zr, Cr (ble 3) in comparison with the elements present in the contaminating environment. Hammerscale produced during the forging of modern steel show a high MnO concentration in both glass and iron oxides. Those made on iron bars resulting from the forging of the African bar type 1 and type 2 do not show any high concentration in manganese, but rather in titanium and zirconium. However, the titanium concentration appears to be much higher for the hammerscale produced from iron type 2 than the hammerscale from African iron type 1. As the potential contaminating materials show a minor concentration in TiO 2 and Zr ( Table 2) and very rare natural materials of the lithosphere reach such concentrations [29], these geochemical anomalies reported in hammerscale can be associated with confidence to the inheritance of the slag inclusions contained in the African iron worked. Nevertheless, it is necessary to assess if the chemical signature for elements with high contrast in concentration between slag inclusion and contaminating environment has a chance to be preserved despite contamination. Geochemical models (Figs. 8,9) provide an understanding of this issue. They were built using mixing equations [30,31] (Eqs. 1,2,3). The aim of the models Fig. 8 is to assess the limiting conditions where the geochemical signature inherited from the slag inclusion is too altered by the contaminant to be considered. The idea is to compare a contaminated mix concentration X CM calculated   (dilution). %Variation is the percentage of the variation between the contaminated mix (X CM ) and the reference.
The lower the %variation, the more the slag signature is getting closer from its initial signature. We made the calculation with a variable initial concentration of TiO 2 in the slag inclusion varying between 30 wt.% and 1 wt.% and with different percentage of participation of the slag inclusion (1, 5, 10, 20, 30, 40%). The concentration of the contaminant is constant and fixed to 0.1 wt.% TiO 2 which is the concentration of the local sand measured in this study. The last %Variation still considered as an unaltered geochemical signature is fixed at 10%. With a SEM-EDS or WD-XRF (10% uncertainty) no matter if a contaminated mix or an uncontaminated mix is analyzed the uncontaminated reference concentration value will almost always still be in the uncertainty range if the contaminated mix is not different from more than 10% from the uncontaminated mix. The model shows (1) the more the slag participates and the lower the %variation and (2) the more the TiO 2 concentration in the slag is high and the lower the %variation. In other terms, the signature is not or slightly affected when the slag inclusions participation is important and when TiO 2 concentration in the slag is high. For a low content of TiO 2 (2 wt.%) the participation must be at least of 30% to preserve the geochemical signature (< 10% Variation). For high content of TiO 2 (28 wt.%) the participation may be low as 5% to preserve the signature (< 10 wt.%). In our case study the average of TiO 2 concentration measured in slag inclusions from African iron 1 is 9 wt.%. The signature must be preserved up to 10% of slag inclusions participation.
Concentration of contaminated mix

Concentration of uncontaminated mix
Variation of a contaminated mix compared to an uncontaminated mix Model Fig. 9 allow to estimate at which minimum concentration of the contaminated mixture the information will be derived from the uncontaminated reference concentration X REF . The signature appears to be preserved (< 10% variation) down to 1,1 wt.% TiO 2 for low participation (5%) of the slag inclusions in the contaminated mix, in higher slag participation calculations (40%) the signature is preserved down to 0,7 wt.% TiO 2 . The typical TiO 2 slag inclusion concentration could be then preserved from contamination even for relative low content. Nevertheless, it must be noticed that what is measured in a hammerscale never correspond to a binary mix (slag inclusions + contaminant) but a ternary mix (slag inclusions + contaminant + iron oxides). As the iron oxide inherited from the iron worked contain only siderophile elements like Fe, Ni, Co, As, it does not make a deviation of the concentration from the reference but only a dilution.
(1)   It means ratios are conserved but concentrations are highly diluted in comparison with the concentration of the initial slag inclusions. Therefore, for elements such as Ti, Zr, V, Mn, Cr and some other trace elements, the geochemical signature of the slag inclusions must be preserved even if concentrations are measured in low abundance in the final mix (slag inclusion + contaminant + iron oxides = hammerscale). The essential factor for the conservation of the signature is to have a high concentration of an element in the slag and a low concentration of the same element in the contaminant ((X I Slag/X I contaminant) > > 1). As long as this point is verified, the concentration of the chemical elements in hammerscale is only important for the analytical resolution of the instrument used.

Chemical Comparison of Hammerscale Produced from Different Bars
It has been previously shown (3.2.1) that in this case study the elements Ti, Zr, Mn have the highest probability of being representative of the initial chemical ratios of the slag in inclusion in the metal. These elements are enriched in slag inclusions in contrast to the sandy contaminant. Hammerscale produced from the two different African iron types should therefore show chemical differences for these particular elements. Hammerscale from modern steel should also show differences but this is not linked with slag inclusion. Figure 10 is the result of about 500 SEM-EDS analyses on about 50 hammerscale samples produced during the 3 forgings F07 (iron bar type 1), F08 (iron bar type 2) and F105 (modern steel bar). These analyses include measurement made on iron oxides, glass and fayalite. A majority of analyses performed on hammerscale produced during the forging of modern steel (F105) show a high concentration of MnO. Those from forging F07 seem to be concentrated in Zr and slightly in TiO 2 while those from forging F08 seem to be concentrated in TiO 2 and very slightly in Zr. This qualitative comparison allows us to highlight interesting chemical differences. Nevertheless, to avoid a contamination bias it is necessary to compare the different chemical ratios. The TiO 2 , MnO, Zr ternary diagram (Fig. 11) combining WD-XRF and LA-ICP-MS analyses confirm that there are chemical differences between hammerscale produced from the modern steel, iron bar type 1 and iron bar type 2. Hammerscale from F07 forging show a lower TiO 2 /Zr than those from F08 forging. Among the three elements Ti, Zr, Mn, hammerscale produced during F105 forging are mainly composed by Mn whereas for hammerscale produced during forging F07 and F08 it represents a very small proportion. Observations made by SEM-EDS on TiO 2 , MnO, Zr concentrations (Fig. 10) correlate well with differences in ratios observed (Fig. 11). This is because the different bars were worked under similar forging conditions (no intentionally addition) and therefore the participation of contaminant in the composition of hammerscale produced was not drastically different between the three forgings. V/Zr ratios  Table 2). No evidence of contamination by ashes and hearth have been reported. They seem to play a minor role in the contaminant composition. Shaded red area, define above 10% variation, represents the zone where the typical signature of slag inclusions is considered altered. On the contrary, shaded green area define under 10% variation represents the zone where the typical signature is considered non-altered. Most of the calculations plot in the green zone, except for low TiO 2 wt.% slag inclusions and low % SI participation made on iron oxides and glasses but also area measurements with oxide and glass (< 3% of measurements). Y-axis represents the number of analyses. n = sum of analysis measured on hammerscale produced from the two African iron types (F07 and F08) show a good linear correlation (Fig. 12). However, the chemical distribution of these analyses is quite distinct with a V/Zr ratio between 0.12 and 0.2 for hammerscale produced from African iron type 1 (F07) and V/Zr = 0.5 to 0.8 for hammerscale produced from African iron type 2 (F08). It must be noticed that hammerscale produced during the first forging step, the pre-forming, are more concentrated in V, Zr, TiO 2 than the hammerscale produced during the second forging step, the forming (Figs. 12,13,14). This indicates a difference in the participation of the slag in the chemical composition of hammerscale between the beginning and the end of the forging operation. The TiO 2 /Zr and V/Zr ratios measured on individual hammerscale by LA-ICP-MS plot in the range of measurements made on hammerscale groups. This means that hammerscale are relatively homogeneous in terms of chemical ratios for these given elements.

Comparison Between Slag Inclusions and Hammerscale
The chemical comparison between slag inclusions and hammerscale allows to control if the TiO 2 /Zr ratios of hammerscale produced during the forging of the two African iron bar types correspond to the TiO 2 /Zr ratios measured in slag inclusions of the iron bars. TiO 2 /Zr ratios measured on hammerscale produced from each African iron bar types show a good linear correlation (Fig. 13). But, ratios between hammerscale produced from the two iron bar types are nevertheless clearly distinct. TiO 2 /Zr = 5.8 to 3.5 for hammerscale produced from African iron bar type 1. TiO 2 /Zr = 31 to 19 for hammerscale produced from African iron bar type 2. Slag inclusions in African iron type 1 show a variation range of TiO 2 /Zr, between 6 and 1. Therefore, hammerscale resulting from the work of the African iron bar type 1 match with the TiO 2 /Zr chemical  Fig. 12 Zr (ppm) vs. V (ppm) concentration measured by WD-XRF and LA-ICP-MS for hammerscale from F07 and F08 operations. Blue and red shaded area corresponds to the chemical range describe by hammerscale respectively produced from African iron bar type 1 and type 2. Captions, colors and shapes are the same as for Fig. 11 range of slag inclusions measured in African iron bar type 1 (Fig. 13). Whereas, hammerscale from african iron bar type 2 does not match. Slag inclusions in African iron bars type 2 show a wide range of TiO 2 /Zr, 40 to 0.5. Nevertheless, 3 iron bars show a relative distinct ratio, B217 (TiO 2 /Zr = 17), B214 (TiO 2 /Zr = 4), B202 (TiO 2 /Zr = 0.5). As well as hammerscale from iron type 1, hammerscale from iron type 2 match with the TiO 2 /Zr range describe by the slag inclusion (Fig. 14). The range of TiO 2 /Zr measured in these hammerscale are close to the range of those from slag inclusions of B217. However, hammerscale from African iron type 1 match also with the range describe by slag inclusion measured in African iron type 2. In this case, the wide range of TiO 2 /Zr in slag inclusion from African iron type 2 do not allow a proper discrimination of hammerscale produced from iron bar type 1.
It would have been interesting to compare the V/Zr ratios of hammerscale from both iron type (Fig. 12) but unfortunately V contents in the slag were too small (< 0.15 wt.%) to be measured by SEM-EDS.

Discussion
Tracing the geochemical fingerprint of the iron worked through the metallurgical waste that remains in a workshop is quite difficult with our current methodologies. The first one, based on the analysis of slag inclusion in metallic iron waste/object need to study a large amount of non-corroded iron to be representative of the forging activity. However, metallic iron wastes are most of the time, found in small abundance [32] and in advanced state of corrosion. Moreover, the measurement of slag inclusion composition in iron need a large amount of time to process. This methodology seems optimal for great archeological discoveries like iron bars found in Saintes-Maries-de-la-Mer shipwrecks [33], building [34,35] or iron object collections [36,37] but not for the study of a common archeological workshop. The second one is based on the study of plano-convex-bottom slag. When the metal under processing is clean, containing a minimal amount of derived smelting slag inclusions, the contribution is too small and very difficult to detect. But, if the raw iron is rich in slag inclusions, its contribution can be detected. It will be the case, for instance, when a plano-convex-bottom slag is the result of the processing of a dirty raw bloom. However, their formation can be the result of several hours of work and several different pieces of iron could have been involved resulting in mixed signature. Moreover, the numerous macroscopic and microscopic observations of plano-convex-bottom slag demonstrate beyond any doubt their heterogeneity. As a consequence, the representativity of a chemical analysis is always questionable, especially as minor components are considered. Unlike iron objects hammerscale are metallurgical wastes that are frequently found in abundance on archeological sites where there was a forging activity [15,[38][39][40][41][42][43]. Moreover, they are formed from a tiny part of a single piece of iron unlike plano-convex-bottom slag that are a potential mix of different processing iron.
The results of this study showed that there can be an identifiable chemical link between slag included in the worked metal and hammerscale produced. Chemical elements concentrated in the slag inclusions and in low abundance in the contaminating sandy materials are good geochemical tracers. It is frequent that iron ores are enriched in various elements that are not enriched in the common sandy-clay materials. Indeed, many iron ores are the result of quite specific metallogenic processes. Those particular processes taking place during the formation of the rock favor the enrichment in Al, P, Ti, Mn, V, Cr, As, Ni, Co, REE. It happens that associated gangue minerals can provide other substances, like Zr, or Ba. The enriched nonreducible chemical elements in the ore are transmitted to the smelting slag and then to the slag inclusions.
During the iron working the non-reducible elements enriched slag inclusions take part in the geochemical composition of the hammerscale. Looking at these geochemical enriched concentrations, in contrast to sandy materials, inherited from slag inclusions, allow to fingerprint the smelting system where the iron worked comes from. In an archeological hammerscale assemblage, it would therefore be possible to chemically differentiate hammerscale produced from iron smelted with different smelting system. Moreover, as these geochemical anomalies are inherited from the ore, it allows comparison with potential provenance areas. The risk of mixing slag inclusions of different origins in a hammerscale is minimal, as a hammerscale is the sampling of a very tiny part of the worked iron. However, this method has its own limitations. Indeed, geochemical anomalies used to differentiate hammerscale may not be always visible. It depends of two essentials factors, the participation of slag inclusions in the composition of hammerscale and the concentration of slag inclusion in specific elements. Participation of slag inclusion is (1) affected by fluxes additions. This technique has been used for a long time by blacksmiths, some operations, especially welding, require silica additions to protect the metal from oxidation [28]. These intentionally additions of sandy materials dilute the geochemical signal of slag inclusion. Participation of slag inclusion is (2) affected by the progression of forging. At the beginning of the forging there is more inclusion of slag in the iron and therefore they participate more in the composition of the hammerscale than at the end. In both situations if the participation of slag in the composition of the hammerscale is too low, because either the fluxes additions are massive or the hammerscale have been produced at the end of the forging, the geochemical signal of slag inclusion transmitted should be in the chemical background noise of the contaminant materials. The chemical information would be then difficult to interpret. The optimal conditions to use this method is to perform measurement on hammerscale with high slag participation like hammerscale produced from a relative dirty iron. In terms of chemistry, it means that the specific elements (Ti, Mn, Zr etc.…) measured in the hammerscale must be at least 2 or 3 times more concentrated than an average sandyclay material. The main analytical techniques SEM-EDS and WD-XRF used for this experimental study would be restrictive in an archeological context. For the SEM-EDS it is the detection limit that is much too high (0.1 wt.%) to consider trace elements content. For WD-XRF it is the minimum sample mass required, 0.7 g, for an analysis. A hammerscale has an average mass of 0.05 g, in a known context where hammerscale are produced from the same iron, it is possible to mix them in order to reach a mass of 0.7 g. However, in an archeological context where hammerscale are potentially produced from different iron types, it is not an option, the result would be an average of the signature of different type of worked irons and not the signature of one part of a single iron. The most suitable analytical device to make measurements combining a low detection limit and a small mass requirement is the Laser Ablation Inductively Coupled Plasma Mass Spectrometry (LA-ICP-MS).

Conclusions
This study highlighted the potential geochemical link between slag inclusions in iron and hammerscale. During the iron working, slag inclusions are transmitted to hammerscale. Studying the chemical elements enriched in slag inclusions and depleted in the sandy environment allowed to differentiate hammerscale produced from iron smelted from different smelting system. The specific ratios of some chemical elements in slag inclusions are preserved from contamination and are transmitted to hammerscale. The archeological application seems promising. It would allow a better understanding of the iron supply of a workshop. Different questions can be address: How many chemical iron types have been worked? Are the same irons that have been worked over time? Is the chemical signature of the same iron found on the workshop nearby? Moreover, as the chemical elements used to trace the geochemical signature is inherited from the ore that have been smelted, it allows comparison with potential provenance areas. Nevertheless, using this method need to study a large number of samples. For a meaningful analysis of one hammerscale it required a minimum participation of slag inclusion. These conditions should not be met in all hammerscale and therefore some of analyses should not be usable. The LA-ICP-MS seems to be the most suitable analytical method to study hammerscale individually. Analysis time is short (1 mn/hammerscale) and sample preparation relatively simple. Making the study of an archeological hammerscale assemblage feasible in an efficient way.
Acknowledgments Many thanks to Bernard Gratuze for the discussions we had and for his contribution during LA-ICP-MS analysis. We also would like to thanks Ildiko Katona Serneels for sharing some of SEM pictures, and Alexandre Salzmann, Patrick Dietsche for their contribution during samples preparation. Our last thanks go to the two anonymous reviewers, their helpful comments improved the quality of this manuscript.
Funding Open access funding provided by University of Fribourg.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.