Advanced Ancient Steelmaking Across the Arctic European Landscape

3.1 Chronology

Previously, radiocarbon dating of charcoal samples from and around construction remains has been carried out at 19 of the 42 studied sites. These sites were primarily iron production sites, although, at the time of the dating, they were generally not yet known as either smelting or smithing sites. The majority of these sites have been attributed to the pre-Roman Iron Age, representing the initial phase of iron technology in Arctic Europe (as detailed in Tables 2–4). In our current study, we have expanded this dataset by conducting new radiocarbon dating on metallurgical remains at three previously undated sites and seven previously dated sites. We achieved this by sampling charcoal trapped in slag and extracting carbon from iron (steel). Prior to radiocarbon dating, wood species analyses were conducted to assess the intrinsic age of the charcoal samples and address potential errors such as the “old wood effect.” Additionally, contamination from fossil carbon was ruled out due to the absence of calcium carbonates in the local geology (Gassman & Schäfer, 2018). This approach allowed us to obtain reliable dating for the actual ironworking and confirm the established chronology (as summarized in Tables 2–4). However, 20 of the analyzed sites remain undated with respect to archaeometric analyses, and the chronology at these sites remains uncertain. The challenge lies in making fine-scale chronological attributions due to the long-term use of these sites. This complexity is particularly pronounced for interior sites, which, owing to land uplift, may contain traces of settlements extending back over 10,000 years. To address this, we have approximated the dating of metallurgical material at these sites through a careful analysis of its contextual relation to typologically dated asbestos-tempered ceramics (Hulthén, 1991; Jørgensen & Olsen, 1988). Consequently, these sites are assigned to a broader Late Bronze Age/pre-Roman – Roman Iron Age phase.

3.2 Data Collection, Sampling Strategy, and Constraints

A stepwise strategy was employed to collect data related to prehistoric metallurgical material in Arctic Europe. The process began with a comprehensive review of excavation reports, museum collection databases, and published works on various sites. The primary focus was on sites representing the initial phase of iron technology, specifically those dated to the Late Bronze Age, pre-Roman Iron Age, and Roman Iron Age. The initial selection method was primarily based on available radiocarbon dating and the depositional contexts of the finds, which were dated by the associated pottery and so-called “Stone Age finds” (certain asbestos-tempered pottery and lithics). Sites were also selected according to geographical spread (to cover the widest possible area). Additionally, sites were chosen based on their association with Kjelmöy and Risvik ceramics, which, according to previous literature, are linked respectively to hunter-gatherer and agrarian economies (Bennerhag et al., 2023). Subsequently, material studies were conducted at the archival institutions holding the metallurgical material: in Finland at the Finnish Heritage Agency in Helsinki, in Norway at the Arctic University Museum of Norway in Tromsø, and in Sweden at the Silver Museum in Arjeplog and the County Museum of Norrbotten in Luleå. This thorough approach allowed us to refine the chronology.

Our selection process at each institution was guided by a visual (macroscopic) examination of the finds. Specifically, we focused on slag samples, aiming to create a representative collection that encompassed different morphological types. This allowed us to identify various production processes associated with ironworking. The most common type of slag, as observed macroscopically, was included in our selection. Additionally, we deliberately chose samples of slag that deviated in form, texture, and magnetism. The degree of magnetism served as an indicator of the specific type of slag. Moreover, iron fragments lost in the production process or parts of objects/semi-finished objects were especially targeted to establish the different types and qualities of iron used and produced. The most magnetic pieces were selected with the aid of a magnet. In addition to slag and iron fragments, we also considered technical ceramics. Our selection criteria for technical ceramics were based on identifiable shape and macroscopic observations of temperature gradients. By studying these ceramics, we aimed to identify variations in raw materials (such as clay), understand the construction details of furnace shafts and smithing hearths, and explore aspects related to refractivity and firing temperatures. In some instances, the sampling was limited to a few technical ceramic fragments overall recovered from excavation.

In addition, samples were taken from fresh ore found in the vicinity of the Swedish lakes Kakel and Storavan. The purpose was to assess whether these ores were viable for iron production. These ore samples were originally collected during archaeological surveys conducted in the 1970s–1980s due to hydropower expansion in the area. At the time of collection, the approach followed an unrestricted sampling design. This meant that artifacts and anomalous objects were gathered without regard to their cultural or natural affiliation, as well as the historical period they represented (sometimes even misidentified and referred to as slag).

As part of our study, we also surveyed the raw material supplies in the local vicinity of the excavated iron production site in Vivungi, Sweden. Our goal was to gain insights into the availability and quality of both local clay and ore. This work involved close collaboration with local residents, who actively participated in the surveys and shared invaluable knowledge about the local raw material sources they were familiar with from participating in modern taskscapes. Our surveys revealed numerous sources of limonite ore in the local lakes. However, locating high-quality clays proved to be quite challenging. The nearest clay source was more than 15 km away.

Of particular interest for our sampling was the large assemblage from metalworking collected between the 1940s and 1980s in Arctic Sweden, Finland, and Norway. These collections resulted from surveys and excavations of ancient settlements associated with lake regulations and hydropower expansion. While the number of slag fragments found at each site was limited, the aggregated material from all these sites represents an extensive landscape of metalworking. Intriguingly, this material was found at early hunter-gatherer sites, alongside Stone Age artifacts – including lithics, projectile points, and pottery – indicating an early utilization of iron. However, during the time of recovery, these sites posed challenges due to persistent evolutionary ideas. These ideas revolved around the dichotomy between hunter-gatherers and farmers and the dominant chronological model of the Three-Age system, with the assumed evolutionary stages inherent in the progression from the Stone Age – the Bronze Age and the Iron Age. The material context did not align with these evolutionary ideas, and as a result, a general idea of a “delayed Stone Age” (rather than an “Early Iron Age”) was established, which in turn meant that the metallurgical remains were severely neglected, and no methods were employed for the identification of the metallurgical material, such as archaeometric analyses including radiocarbon dating of carbon in steel and charcoal trapped in slag – methods since long established in European iron research.

The preconceived notions regarding the Arctic regions of Scandinavia as a periphery with a relatively late settlement history have significantly influenced interpretations of the metallurgical material associated with lake regulations. Unfortunately, the documentation of this material collection often lacks contextual records. This is particularly pronounced when it comes to slags, which were traditionally regarded as waste material with limited information, meaning that some slag finds were not even collected. Furthermore, misidentifications have significantly impacted our understanding of the material, typically downplaying variation, i.e., almost all the metallurgical material has been unilaterally recorded as slag and, in rare cases, as burnt clay. While a few archaeological features and structures related to metallurgical activities have been identified, there is a notable absence of hammer scales (magnetic flakes detached from metal during smithing) among the collected material, although these should have been part of the material, given the now-identified smithing activities. Unfortunately, several of the sites are now submerged underwater and exposed to erosion due to the construction of dams and hydroelectric power stations. As a result, much of the microscopic evidence of production activity has been irretrievably lost. Unfortunately, these conventional views persist within the material-holding institutions, leading to selection-related restrictions (such as sampling) that were occasionally imposed before we could study the material. Additionally, parts of the collections remain stored in inaccessible find magazines. The institutions were rather uncomprehensive by the fact that slags could provide interesting information if explored more comprehensively, as it (according to the traditional view) was considered waste/mass material.

Overall, our study encompasses material collected over a 70-year period (including our recent excavations) with all that this entails in terms of varying interpretative and methodological perspectives, practical research techniques, and levels of ambition. Despite the serious limitations, we have managed to distinguish both reduction and smithing slag alongside technical ceramics (such as furnace wall and hearth lining), ore, iron waste, and iron fragments across a broad geographical area. These aggregated sites form a rich and extensive landscape of metalworking, allowing us to draw initial but substantial conclusions about the technology and organization involved. Slags and other refuse materials provide more informative insights into the metal craft than merely examining the morphology of the final products. This shift in focus from product to process represents a departure from the traditional approach in North Scandinavian iron research. Consequently, this material offers a valuable synchronic perspective on Early Iron Age metallurgical traditions in the Arctic European area.

3.3 Set of Samples, Methods, and Detailed Analyses

In total, 237 finds were selected for the present study: 65 from Finland, 33 from Norway, and 139 from Sweden (some of which included more than one underlying find item and various material types) (for details, see Supplementary S.5–S.7). Most of these finds consist of reduction or smithing slags (including both primary and secondary smithing) and technical ceramics (furnace walls and hearth linings). Additionally, the collection includes iron waste, fragments of iron objects, fresh and ancient ores, heat-affected stones, stones with slag cover, and red-brown magnetic particles and lumps. Initially, a visual inspection and morphological description were conducted on all the sampled material. This process included a special recording of technical ceramics and a characterization of slag finds based on their degree of magnetism. Subsequently, in collaboration with archaeometric fellow researchers, we selected samples for more detailed analyses.

We employed archaeometric analyses to extract relevant metallurgical information about the metal smelting and smithing material. These analyses also shed light on the technologies used to reconstruct craft practices. Furthermore, we aimed to gain insights into the economic organization, resource acquisition, and the overall social context of metal technology. Understanding the production processes is crucial for establishing the various activities and practices related to iron production at each site. Additionally, it allows us to explore how metalworking was integrated into the organization of society and the overall spatial layout of the crafting landscape (Erb-Satullo, 2022). Variations in resource acquisition and production practices may also indicate different levels of participation among distinct social groups across different communities.

Several analytical methods were employed for the metallurgical finds. Metallographic analyses were conducted on iron waste to establish its quality, determining whether it is ferritic iron, steel, or, for instance, phosphoric iron. Additionally, petrographic and bulk chemical analyses were performed on slag samples. The primary purpose was to define the process stage during which the slag was formed and to determine its composition in terms of major, minor, and trace elements. This allowed for a comparison of slags from various contexts (both within and between sites) and provided insights into the types of ores used. Furthermore, samples of fresh iron ore underwent test-roasting and were subjected to petrographic and chemical analysis. These assessments aimed to evaluate the quality of the local ore and uncover possible prehistoric connections and uses. Additionally, technical ceramics (primarily from furnace walls and hearth linings) were classified and recorded macroscopically. A subset of these ceramics underwent further analysis, including petrographic microscopy and thermal analyses. These investigations helped establish raw material choices and provided insights into construction and curation strategies. Moreover, they shed light on the thermal and mechanical properties of the clays (Supplementary S.1).

Radiocarbon dating has been performed on iron waste and samples of charcoal included in slag. Before dating the charcoal, wood species analyses were conducted to assess the intrinsic age of the sample and identify possible sources of error. The radiocarbon dates consistently fall within the pre-Roman to Roman Iron Age period, except for a sample from Norway that was radiocarbon dated to the Viking Age (AD 800–1050). Tables 2–4 provide a compilation of all radiocarbon dating samples, including previous data from excavated iron production and smithing sites.

4.1 Iron Smelting

Analytical results confirm the production of iron at 10 previously excavated production sites (Bennerhag et al., 2021; Forsberg, 2012; Jørgensen, 2010), with 16 furnaces spread throughout Arctic Norway, Sweden, and Finland. Radiocarbon dating spans the period from 300 BC to AD 300 (Figure 1, Tables 2–4). These sites include Hemmestad and Flakstadvåg in Norway, Sangis and Vivungi in Sweden, and Riitakanranta, Kotijänki, Kemijärvi, Äkälänniemi, Mikkeli, and Lahti in Finland. Additionally, two smelting sites have been confirmed in the Lake Kakel area in northern Sweden. These sites, known as the Revi site (SMA 3319) and the Nåtti site (without a number), were identified through detailed petrographic and chemical analyses of stearin-like slag. Notably, these sites currently lack any traces of furnace construction. As they belong to the lake-regulation material, they are submerged and further excavations are not possible. The Revi slag has been radiocarbon-dated (using charcoal inclusion in the slag) to the pre-Roman Iron Age (Figure 1 and Table 4).

Interpretations in the literature regarding iron smelting in the Arctic European area have often centered around a dichotomic division of furnace remains into a western and eastern cultural sphere attributed to different smelting technologies and traditions. The argument stems from reconstruction models of the furnaces, which, unfortunately, rely on insufficient knowledge regarding the necessary criteria for the smelting process to function effectively. For instance, the aerodynamic perspective emphasizes the fundamental role of furnace shaft height, as well as the need for adequate room above the slag mass for the bloom to form. In the literature, the eastern furnaces (in our study including the north Swedish and north Finnish sites) are typically associated with a relic and archaic bowl-furnace technology, lacking superstructures (Kotivuori, 2013; Lavento, 1999). These furnaces are often seen as part of a hunter-gatherer economy. In contrast, the western furnaces (in our study including the Norwegian sites) are considered part of a supposedly superior European shaft furnace tradition (Jørgensen, 2010) related to agrarian societies (Pleiner, 2000). However, our findings diverge from this prevailing view. Substantial evidence from clay-built furnace shafts at all our smelting sites suggests a consistent smelting technology across the entire region – from Finland in the east to Norway in the west. Notably, recent years have witnessed discussions questioning the overall validity of the typological division of furnace constructions at a broader European level. Consequently, the tendency to downgrade furnace types for certain groups extends beyond Arctic Europe (Dungworth, 2014; Hjärthner-Holdar et al., 2018; Rondelez, 2017).

Analyses of process-related parameters at sites with construction remains demonstrate, at odds with literature, that smelting operations took place in small bloomery furnaces (inner diameter of app. 0.3 m) with an upright cylindrical shaft made of clay, and with a rectangular stone framed base and underlying slag pit, in a few cases lined with clay (Figure 2). Analyses further demonstrate that the furnaces were used several times. Roasted limonites in terms of lake and bog ores along with charcoal (as fuel) from pine and birch were used in the smelting process; the charcoal provided the high temperatures and the reducing atmosphere needed for the iron ore to reduce into metallic iron. Air-inlet holes, visible on furnace wall fragments, were identified at all sites with construction remains, indicating that air was supplied using bellows (Figure 3). The bellows were placed at the bottom of the furnace shaft and mounted on the stone frame – the hottest zone of the furnace, where temperatures slightly exceeded 1,250°C. Within this zone, the viscous slag (Figure 4) separated from the solid iron and accumulated at the bottom part of the furnace in the slag pit. Subsequently, the slag was removed through an arch in the stone frame or through the top of the furnace. Since the smelting occurred below the melting temperature of iron, the iron formed in a solid state, creating a spongey mass known as a bloom just below the blowing zone. Remarkably, blooms are rarely found in archaeological material globally. However, at 7 of our 12 smelting sites, small fragments of iron waste (detached parts of blooms showing no signs of additional consolidation or mechanical processing) were discovered around the furnaces (Figure 5). Overall, the ancient Arctic furnace technology aligns with the shaft furnace technology – essentially the same basic technology used in other contemporary and later iron production furnaces (Bennerhag et al., 2021; Hjärthner-Holdar et al., 2018).

Figure 2

Bloomery furnaces at the Vivungi site, northern Sweden. Within one of the furnaces, a circular charcoal feature was discovered, indicating that the inner part of the furnace shaft was round (photograph by Carina Bennerhag © Norrbottens museum).

Figure 3

A wall fragment from one of the furnaces at the Vivungi site, comprising a section of the original clay shaft (with an air inlet hole), and parts of the stone frame (photograph by Staffan Nygren © Norrbottens museum).

Figure 4

Stearin-like reduction slag from the iron production site in Vivungi, featuring well-preserved charcoal imprints (photograph by Staffan Nygren © Norrbottens museum).

Figure 5

Iron waste from the Vivungi site. The etched sample shows a pearlite texture, which demonstrates a high carbon content (photograph and micrograph by E. Ogenhall © The Archaeologists, Swedish National Historical Museums).

4.2 Iron Smithing

Evidence from smithing is more abundant than evidence from iron smelting in the analyzed material. Twenty-seven of the 42 studied sites show indications of smithing, with a significant portion spatially connected to large settlement sites. These sites are spread across a vast geographic area covering northernmost Norway and Sweden (Figure 1 and Table 1). However, regarding Finland, language barriers and lack of time have contributed to the fact that the Finnish smithing material (although smithing finds are abundant also in the Finnish material) has not yet been selected and analyzed.

Physical remains of ancient smithing hearths are rare, and overall, there is limited knowledge in the literature regarding the organization of European iron workshops and roofed forges during the Early Iron Ages (for exceptions, see Berranger & Fluzin, 2014; Dieudonné-Glad, 1997; Englund & Hjärthner-Holdar, 2017). Historically, the focus has been on the exchange and trade of metal objects rather than the organization of metal production. If any physical remains exist, they are generally limited to the spatial distribution of slags, iron waste (bloom pieces), hammering scales (indicating the location of the anvil) (Figure 6), and remnants of the clay lining of slag pits. Against this backdrop, our relatively abundant discoveries of smithing installations in Arctic Europe stand out as remarkable.

Figure 6

Hammering scales produced during smithing, the distribution indicating the location of the anvil (photo © Norrbottens museum).

In our study, we have identified various types of smithing remains, with slag cakes – each corresponding to a single work session – being the most common debris found at smithing sites in both Sweden and Norway. Most of these slag cakes result from object fabrication and consequently exhibit oxidizing conditions (Serneels & Perret, 2003). However, we also find smithing slag cakes displaying reducing conditions, along with separate lumps of metallic iron (either concentrated within slag or as distinct pieces). These findings suggest that bloom consolidation (primary smithing) (Erb-Satullo & Walton, 2017) occurred right at the smithing sites. Interestingly, and contrary to existing literature, this implies iron blooms (rather than semi-finished products like iron bars) served as raw material for smithing. The prevailing organizational model of iron working in Europe (primarily based on materials from the Late Iron Age) is the primary forging of the raw iron was carried out at the smelting site before being distributed to various workshops (for exceptions see, e.g., Bauvais & Fluzin, 2009). The Arctic organizational way of iron working appears to have required the close proximity of smithing and smelting sites, as blooms are unwieldy to transport over long distances. This is supported by three of our 27 studied smithing sites (specifically the Swedish Revi, Nåtti, and Sangis sites), where the distance between smithing and smelting ranges from 10–500 m. Additionally, the proximity between smelting and smithing offered logistical advantages, especially regarding the use of charcoal, essential for both processes, since there is a great risk for charcoal to shatter while transported.

At two of our excavated sites, Sangis and Vallen, remarkable evidence of hearth installations – characterized by oval features containing charcoal – has been uncovered (Figure 7). What makes this discovery even more significant is that one of these hearths was found within the remains of a hut. Consequently, the pre-Roman Vallen site now stands as the oldest known remains from a roofed forge in all of Scandinavia. Before this, the oldest known iron workshop with a roof (dating to the Roman Iron Age) was located in southern Sweden, approximately 1,000 km south of Vallen (Englund & Hjärthner-Holdar, 2017). In the broader context of Arctic Europe, only a limited number of metallurgical contexts with evidence of forging installations have been previously excavated, primarily from later periods (around AD 400 and beyond) (Jørgensen, 2012). Interestingly, workshops are not even mentioned in the literature concerning Arctic Europe during earlier periods. Finds of smithing slag are often interpreted simply as evidence for tool maintenance (Sundquist, 1999). However, our research underscores the critical role of controlled lighting for ancient Arctic blacksmiths. This lighting allowed them to observe color changes in iron – an essential aspect for understanding the metal’s temperature and behavior during heating. Consequently, it strongly suggests that forging took place under a roof, shielding the process from direct sunlight and rain, as well as the harsh Arctic winters. This concept likely applies to smelting as well, where the smelter requires optimal conditions to monitor the process (Stepanov et al., 2022). Unfortunately, several of our sites are now submerged underwater due to dam construction, making it challenging to further substantiate these hypotheses regarding our sites.

Figure 7

Smithing hearth structure in cross section, recovered at the Sangis site, northern Sweden (photo © Norrbottens museum).

Additionally, finished metal artifacts have been unearthed at various Swedish and Norwegian smithing sites. Among these artifacts, knives constitute the majority of finds. Notably, two knives and an axe from Sangis – whose chemical composition indicates they were made from local ores – have been analyzed and radiocarbon-dated. These artifacts span the period from 200 BC to AD 100. Given the context, it is likely that the yet-to-be-dated knives also originate from this same time frame.

4.3 Raw Materials

The knowledge and availability of raw materials significantly influence the technological choices and chaîne opératoire of the ironworking process. This impact is particularly pronounced for pyrotechnical ceramics (furnace walls and hearth linings). The functionality of these ceramics is significantly constrained by the thermo-mechanical properties of the clay, including plasticity and refractoriness (i.e., performance at extremely high temperatures). Selecting and supplying clay that can endure high temperatures while remaining sufficiently plastic for constructing a maintainable shaft is a crucial but often overlooked aspect of iron manufacturing (Martinón-Torres & Rehren, 2014; Stilborg, 2023). Constructions used to smelt iron ores – often built wholly or partly of clay – must withstand temperatures of up to 1,200–1,300°C for extended periods. Most natural clays in Scandinavia possess adequate refractivity for bloomery process temperatures. However, there are exceptions. Calciferous clays and clays with large iron oxide concentrations or iron-rich rock fragments may lead to instability and, in the worst-case scenario, collapsing furnace walls. To enhance refractivity, tempering the clay with non-plastic, often silicate-rich materials (such as quartz), can be beneficial.

Procurement analyses reveal a complex raw material composition at many of our sites, with deliberate selections of clays possessing special characteristics. Notably, the Finnish Neitilä, Äkkälänniemi, and Riitakanranta smelting sites exhibit close similarities to the Swedish Sangis and Vivungi melting sites. At these locations, smelters employed medium to very coarse-grained clays for constructing furnace shafts. Although these clays had a high volume of sand (predominantly quartz, enhancing thermal properties), they were somewhat limited in plasticity. However, to gain a deeper understanding of the choice of clay, we require additional information about the types of clay available in the areas surrounding these sites. This would help determine whether clay choices were primarily influenced by a shared craft tradition or driven by geological factors. Interestingly, our analyses indicate that curation strategies were adapted to address the insufficient plastic qualities of the clay. Instead of rebuilding the furnaces, cracks were repeatedly repaired, and layers of clay were applied between runs. After each repair, the furnaces continued to be used for additional smelting cycles. This practice underscores widespread knowledge of handling this specific type of clay. Furthermore, the sintering interval of these clays exhibits deformation and melting between 1,150 and 1,250°C, indicating narrow temperature windows for carrying out metallurgical processes. Consequently, ancient smelters skillfully balanced working temperatures on the very edge of the furnace’s thermal capability. Interestingly, nearly all analyzed smithing sites with refractory remains in terms of hearth linings – including most Swedish sites and approximately half of the Norwegian smithing sites – also employed the same coarse-grained clays as those used for the furnaces. These clays were well-suited for high-temperature operations.

Analyses reveal alternative choices of clay at the Swedish Vivungi and the Finnish Neitilä sites. These sites show the use of finer-grained clay with higher plasticity (while maintaining refractivity). This choice indicates a different curation strategy, where the furnace shaft was torn apart and re-built after only a few smelts. While the direct reason might have been that the furnace builders found a larger or closer clay bed with better plasticity, the fact that they could readily shift to a different curation strategy further testifies to their solid knowledge of the raw material.

At the Hemmestad and Flakstadvåg sites in Norway, as well as the Lahti site in southern Finland, clays with clearly low-refractory qualities were used. These included calciferous clay and clay stone, both unsuitable for high temperatures (a conclusion further supported by slag analyses indicating oxygen leakage into the furnaces caused by cracks). Interestingly, despite the repeated use of furnaces at the Hemmestad and Flakstadvåg sites, there are no signs of repairing the troublesome cracks that appeared in the furnace walls. The utilization of clays with low-refractory qualities, coupled with the lack of curation strategies at these sites, suggests a partly different technological system compared to the north Swedish and Finnish smelting sites. In an upcoming article, we intend to explore whether this divergence could be interpreted as a result of partly different knowledge acquisition.

The availability of viable ores was another essential component of smelting technology. Ancient ores are rarely found during excavations of iron production sites, and this study is no exception, with the Vivungi site being the sole outlier (Bennerhag et al., 2021). However, bulk chemical analysis of slag recovered from various iron smelting sites provides valuable information about the type of ancient iron ore used in the furnaces. These analyses exclusively show the use of limonite ores, which are generally abundant in the Arctic European geography (Aschan, 1906), across all 12 smelting sites. We infer that these ores are primarily lake ores. Consequently, all 12 sites are strategically located near lakes or watercourses, ensuring a steady supply of lake ores. This offers an advantage over bog ores, which can be depleted, as lake ores naturally renew every 25–50 years (Naumann, 1922). Additionally, we conducted further sampling and analysis of fresh ores (often mistaken for slag) collected during the lake regulation surveys in northern Sweden. Some of these fresh ores were found to be viable as iron ores, with a high iron oxide content exceeding 70% and the presence of manganese (2–4% MnO). Furthermore, the chemical composition of the ore collected from the lake adjacent to the Vivungi site correlates with bloomery slags from that same site, confirming that ancient smelters indeed sourced their ores from the lake.

Slag analyses further imply that ancient smelters employed ores with varying elemental compositions, which significantly impacted the quality of the manufactured iron. Specifically, the Finnish sites of Neitilä, Äkälänniemi, Riitakanranta, and Mikkeli utilized phosphorus ores (containing up to 1.7% P₂O₅), while the Swedish sites of Sangis and Vivungi, along with the Finnish sites mentioned earlier, used manganese-containing ores (ranging from 0.4 to 5 wt% MnO). The presence of phosphorus in the ores contributes to the production of ductile iron, albeit with some hardness. On the other hand, manganese enables the direct production of steel within the bloomery furnace (see, e.g., Bennerhag et al., 2021; Crew et al., 2011; Hjärthner-Holdar et al., 2018). This results in a harder product that has the potential for various heat treatments, further enhancing its material properties. Notably, this practice was not unique to our analyzed sites but was also widespread elsewhere in Europe during that period (Hjärthner-Holdar et al., 2018).

4.4 Iron Alloys

We gain valuable insight into what was achievable in the furnaces from the small fragments of iron waste – detached parts of blooms measuring 10–40 mm – that were recovered at seven of the ten smelting sites with construction remains. At two of these seven sites (both located in northern Sweden), over 1 kg of iron waste was discovered. However, at the other three smelting sites – the Finnish Lahti and the Norwegian Hemmestad and Flakstadvåg sites – no metallic iron debris was found. This absence of iron waste may be attributed to problems observed during the reduction process.

At all seven smelting sites with iron waste, metallographic analyses demonstrate a highly similar pattern in terms of high levels of carbon (more than 0.8% C). Consequently, this study establishes the presence of high-quality steel as a key iron alloy. At six sites (Swedish Sangis and Vivungi, Finnish Neitilä, Äkälänniemi, Riitakanranta, and Mikkeli), this can be attributed to the use of manganese-containing ores (evidenced through chemical analyses of slag), indicating deliberate steel production directly in the bloomery furnace. Even cast iron (with more than 2% C), typically associated with blast furnace production (but also claimed to be an inevitable by-product of high-carbon steel production in bloomery furnaces, Crew et al., 2011), is revealed in a few lumps at the Swedish (Vivungi) and Finnish (Riitakanranta and Mikkeli) sites. The iron waste discovered at 6 of the 16 confirmed Swedish smithing sites (Sangis R730 and R797, Sandudden NA 54 and NA 82, Hoppot NA 36, and Vallen R 90) (Table 1), most likely originating from primary smithing, provides additional evidence of steel as a desired product. Furthermore, the presence of phosphoric iron observed at two Finnish smelting sites (Riitakanranta and Mikkeli) suggests a demand for hard and ductile metals. This is further confirmed by the chemical analyses of slag at these sites, as well as at two other Finnish sites (Neitilä and Äkälänniemi), which reveal the use of ores with a phosphorus content. The production of phosphoric iron in the Arctic European region is also indicated through the recovery of an iron knife at the Swedish Sangis smithing site, combining both phosphoric iron and steel.

4.5 Fabrication and Composition of Artifacts

Another crucial aspect of the technological system lies in how the smelted iron was shaped into desired products. Our analytical results reveal that the most typical fabrication techniques employed were on par with the cutting-edge technology of Europe during that era, both in terms of expertise and specialization. At the Sangis smithing site, artifact analyses demonstrate welding techniques that combined different steel and iron qualities (including both phosphorus and soft ferritic iron) into multilayered tools. These tools were further enhanced through various heat treatments, such as quenching and tempering. The analyzed objects – primarily knives and an axe – consistently featured the hardest steel along the edges, emphasizing the desire for tough and durable cutting tools. The fabrication techniques and iron alloys used were carefully tailored to specific tool types. Additionally, one of the analyzed knives provides evidence of an intent for visual and stylistic effects, achieved through lamination techniques that combine different layers of steel and phosphoric iron (Figure 8).

Figure 8

Knife composed of multiple layers of phosphoric iron and low-carbon steel, resembling pattern welding techniques commonly used in the Late Iron Age, found at the Sangis smithing site in northern Sweden (photograph by Staffan Nygren © Norrbottens museum).

In literature, the welding and laminating procedures observed in ancient Arctic Europe are widely regarded as the most challenging operations undertaken by ancient European smiths (Gilmour, 2017). The difficulty arises from the fact that the different iron alloys to be joined often have distinct optimal smithing temperatures (Güder et al., 2017), resulting in narrow temperature windows for successful attachment. Consequently, the ancient blacksmith’s ability to determine these temperatures using their senses – such as assessing the color of the glowing metal – becomes crucial. Previous literature suggests that these methods were attempted in Europe (particularly its central regions) as early as 500 BC. However, they did not gain widespread adoption until Roman times, eventually reaching the northern European continent during the Middle Ages (but never really to its Arctic parts) (Pleiner, 2006). Remarkably, the present study reveals that over 2,000 years ago, advanced smithing techniques were already established in Arctic Europe. These techniques persisted for 400–500 years, as documented at the Swedish Sangis site and in the Kakel and Storavan areas.

4.6 Similarities in Style and Technological Choice

A comparison between our studied sites reveals striking similarities in style and technological choices between the north Swedish and Finnish smelting sites, as well as between the north Norwegian and Swedish smithing sites (Finnish smithing material has not yet been selected and analyzed). Consequently, analyses demonstrate conformity in technological features, including the use of raw materials (clay and ore), curation strategies, and finished products (raw metal and artifacts). Moreover, there are only minor chronological differences between the sites. This “standardization,” which pertains to the mastery of excellent smelting and smithing operations identified at each workshop, strongly indicates a well-established and recognizable technology (possibly including a technological network) across a large geographical area in Arctic Europe.

Overall, our data allow us to reconstruct, for the Early Iron Age (200–50 BC), a departure from existing literature (Pleiner, 2000) by revealing the deliberate production of steel directly in the bloomery furnace. Steel can be traced through all stages of the metallurgical process, strongly indicating a widespread demand for steel as a product. The identification of phosphoric iron at some sites also suggests a demand for other alloys. The predominant use of coarse-grained clays and the repair technology observed at the north Swedish and Finnish smelting sites further point to conformity in technological tradition. While repairs themselves are not necessarily distinctive elements of a specific craft tradition, the fact that furnace shafts with necessary repairs could be reused underscores an established technological practice. A more significant and potentially defining detail lies in the apparent focus on repairing the area around the air-inlet hole(s). Interestingly, coarse-grained clays were also employed by ancient Arctic smiths, as evidenced by the remnants of hearth linings at both Swedish and Norwegian smithing sites. This observation suggests a knowledge base that may have differed from that of the smelters in the Norwegian region, who utilized much finer clays and adopted different curation strategies.

While our results reveal a shared, complex technological system across a large geographical area in Arctic Europe, it does not exclude individual or local differences in technical choices. Morphological characteristics point to certain variations in the construction of smithing hearths at both local and regional scales. For instance, there are differences in the placement of the clay lining: the north Norwegian sites feature wholly lined hearths, while the Swedish sites exhibit both wholly lined and upper-lined hearths. Additionally, the varying sizes and underlying contact materials (sandy subsoil or charcoal bed) for the slag cakes found at the Swedish smithing sites indicate diverse solutions for hearth constructions. These variations are further evident in the dimensions of the hearth lining remnants.

4.7 Landscape Structuring and Organization

The apparent conformity not only pertains to technological features but also extends to the spatial distribution of iron production and smithing operations. When mapped out, distinct clusters emerge around larger lakes and water bodies within forested landscapes (Figure 9). These clusters are closely related to primary resources – namely, ore and wood required for fuel. The spatial organization of smelting and smithing processes is particularly distinctive in Arctic Europe. In fact, the entire metal production cycle often unfolded within the same local context. By clustering different process stages spatially, ancient communities could effectively manage the complete metal production cycle – from ore selection to the manufacturing of finished objects. This sophisticated system relied on an extensive knowledge base held by the relatively small-scale Arctic hunting-gatherer communities, comprising 25–50 people (Kelly, 2013). Similar clustered organizational patterns are also suggested in Siberian iron workshops during the Iron Age (first to third centuries AD) (Vodyasov et al., 2021), as well as in Mongolia, northern Germany, and Denmark during pre-Roman times (Brosseder et al., 2023; Brumlich, 2020; Winther Olesen et al., 2020).

Figure 9

Excavation at the Vivungi site. The spatial clustering of iron production around larger lake systems indicates a strategic location near the primary resources of ore, fuel, and building material (photograph by Staffan Nygren © Norrbottens museum).

The water bodies further typically contained (and still do) an abundance of fish, and we have found large amounts of faunal remains from freshwater fish at the sites where we had access to material context (Sangis and Vivungi). Overall, this points toward coordination of iron production and good supply/utilization of fish. All the Swedish and the Norwegian smithing sites are further spatially connected to large known settlement places, and as a rule also to other craft activities, such as lithic technology, pottery, horn/bone, and skin/fur-craft. Hence, ironworking seems integrated with the existing material technologies and organizationally incorporated with other residential activities (Binford, 1979). Regarding the lithic tradition, we see no signs of a broader technological change at the time of the inception of iron but rather a continuity well before and after. In addition to dismantling the Three-Age system, this indicates that iron was integrated and incorporated seamlessly into the social and economic organization of the Arctic hunter-gatherers.