The innovation of iron and the Xiongnu – a case study from Central Mongolia

This article presents the oldest iron smelting furnaces of the Xiongnu Empire period in central Mongolia and argues that a significant smelting center existed at the site of Baga Nariĭn Am. Five iron smelting furnaces and four smelting installations were excavated, with a total 26 furnaces further identified through SQUID magnetometry. In combination with a review of data on iron production in contemporary Mongolia, the Transbaikal region, Tuva, the Minusinsk Basin and the Altai, we argue that this new data alters existing narratives of the adoption of iron in eastern Eurasia. While iron smelting as such was adopted from the Minusinsk Basin, where the oldest iron smelting furnaces in eastern Eurasia are currently found, we suggest that the driving force behind the massive boom in iron metallurgy from the second century BCE onward was the Xiongnu Empire. During the course of the Xiongnu Empire, the development of more efficient iron technology is evident, with the steppe empire also inventing a new furnace type. These findings are significant for understanding the dynamics of iron industries in the eastern Eurasian Steppe and paves the way for necessary work on metallurgical installations in the Minusinsk Basin and Tuva.


Introduction
Research on the economies of eastern Eurasian pastoralists focuses mostly on subsistence economy regardless of the discipline, be it archaeology or history. This is even more true when it comes to research on the Mongolian Plateau. Targeted studies on craft production in pastoralist communities and the steppe empires are rare (Brosseder and Miller 2011, 27-28;Reichert 2018Reichert , 2020Ishtseren et al. 2020;Vodyasov et al. 2021), despite the fact that it is widely recognized as an important field of research. Iron metallurgy is no exception, despite huge advances over the past years ; Reichert 2020, 99-110;Vodyasov et al. 2021).
Viewed from a global perspective, iron metallurgy arrived late in eastern Eurasia, appearing at the end of the third century BCE, roughly one millennium after it initially spread from northern India (Turner 2020). The existence of iron metallurgy should not be confused with the incipient stage of iron use, when a few objects made of iron are found but no slags or production debris are known. This is true for both the Near East and Europe, as well as eastern Eurasia and Asia (Wagner 2008, 91-93;Bebermeier et al. 2016;Erb-Satullo 2019, esp. 564-566;Kašuba et al. 2019). The ways in which this new material and technology were adopted were, however, wildly diverse across Eurasia, as metal craftsmen also attempted independent approaches (Kašuba et al. 2019, 199).
Explanations given for the spread of iron technology, especially with regards to the Near East, have been discussed recently (Erb-Satullo 2019, 576-583). Erb-Satullo sees two main strands of arguments: the first group that focuses on material properties pursue the argument that iron, especially with carburization, is harder than bronze, which eventually led to the replacement of bronze. However, the synopsis of case studies in the Near East indicates that the evidence is ambiguous and suggests that "increasing hardness may have been the consequence of the widespread use of iron, rather than its initial impetus" (Erb-Satullo 2019, 579). The second group of explanations focuses on the organization and economics of iron production in relation to bronze. One of the widespread and often repeated, but also criticized hypothesis, is the suggestion that a tin shortage drove the rise of the iron industry (Snodgrass 1971;Waldbaum 1980).
Only with more abundant research on workshops and production debris over the past two decades, research can start investigating the link of the spread of iron and the social context and oganization of iron (and copper alloy) metallurgy and the cultural and political factors that enabled the spread (Erb-Satullo 2019, 583; 593).
Turner, who studied the spread of iron metallurgy through Afro-Eurasia identified several stages, with the first being the time "when iron becomes a material used for multiple object types (military and agricultural tool use, plus at least one of another used type such as construction, utilitarian or ornamental) with increasing frequency…" (Turner 2020, 23). He characterizes the second stage as the time period when "iron first almost completely replaces an older material for a critical object type …" (Turner 2020, 23). In our geographical area of interest, the Mongolian Plateau, these first two stages cannot be differentiated due to the low resolution of the current chronology. But unlike several other regions of Eurasia, the moment when this new material came into regular use can be identified clearly, and it coincides with the early Xiongnu Empire from the second century BCE onward. At that time, a variety of objects were made from this new material: weaponry -especially arrowheads -knives, parts of horse gear, and agricultural tools. These objects are found abundantly in burials. At the same time also debris from local iron production, slags, and iron furnaces are attested. As is the case in other regions of the world, both the debris and increase in the numbers of iron artifacts indicate a new period "in the trajectory of iron adoption" (Erb-Satullo 2019, 567). The contemporaneous processes of empire formation and the iron metallurgy boom highlight the link between innovation and socio-economic conditions.
In the eastern Eurasian Steppe, metallurgy has been studied abundantly in the Minusinsk Basin, where not only a very developed Late Bronze Age metallurgical industry thrived but also a flourishing iron industry. Iakov Sunchugashev, who studied iron smelting sites in Tuva and the Minusinsk Basin for over three decades between the 1960s and late 1990s, asked the crucial question that remains valid for the region today: Since iron smelting had already been mastered during the Tagar period (c. 800-300/200 BCE), why was there only a surge in iron smelting activity during the 2nd to first centuries BCE? (Sunchugashev 1979, 22-23). Because of the abundance of metallurgical sites in the Minusinsk Basin this region has been viewed as the center, from where iron-smelting technology was adopted in Mongolia, before it was further developed locally (Sasada andAmartuvshin 2014, 1023;Sasada 2015, 50;Sasada and Ishtseren 2020, 12). In addition, while it has been noted that the iron's total adoption in Eastern Eurasia is associated with the Xiongnu polity, this process is dated to the late first century BCE and the early first century CE, thus only during the empire's later phase (Amzarakov 2015a;Vodyasov et al. 2020Vodyasov et al. , 2021.
A rapidly growing number of investigated Xiongnu period iron smelting furnaces, together with a larg series of radiocarbon dates from the past decade in Mongolia beg for revisting the state of research on iron metallurgy in Mongolia and its neighboring regions. In this article, we take a closer look at the evidence, question the primacy of the Minusinsk Basin with respect to the spread of the iron technology, suggest that the spread of iron metallurgy in the eastern Eurasian Steppe was driven by the Xiongnu polity from its very beginning. While it has been acknowledged that there are three different types of furnaces in the Xiongnu Empire, among them an underground furnace with a tunnel construction, which reflects a distinct way of applying iron technology in eastern Eurasia. We suggest, based on the current data, that the processes that drove the rise of the iron industry in the Xiongnu Empire also led to the invention of this new furnace type. Our discussion is used to better contextualize several furnaces published for the first time in this article from the Orkhon Valley in central Mongolia, which represent the oldest, most securely dated furnaces discovered so far.

Beyond the artifact: debris from iron production in the archaeological record
Before we investigate iron-smelting on the Mongolian Plateau and its neighboring regions, it is worth considering the kind of remains in the archaeological record that attest to iron production. Each phase in the iron production process creates its own unique debris ( Fig. 1): the first step is ore extraction (prospecting and gathering ore). The second is smelting (converting ore to bloom, which creates smelting slags in the process), followed by primary smithing (converting bloom to bars/billets, which creates primary slags) and secondary smithing (commonly known as forging or blacksmithing; converting bars/billets into iron objects, which creates smithing slags and hammerscale in the process) (Arnoldussen and Brusgaard 2015, 115). The most archaeologically visible and important sources of information on the metallurgical process are slags (Arnoldussen and Brusgaard 2015, 115;Hauptmann 2020, 199-201;Rijk 2007), which reflect both the smelting and processing methods (Fig. 1). Beyond general distinctions between flowing slags, furnace slags, and dome-shaped slags, slag typologies are often idealistic compared to realworld specimens that are not clear-cut. Numerous factors influence the appearance of slags, with slags from the same smelting or processing methods often turning out differently, but also those from different processing methods appearing similar. Above all, slags are often preserved fragmentarily, which makes a typology difficult (Rijk 2007, 114). However, single features of slags point to either smelting or primary smithing: dome-shaped slags are generally produced from iron processing, while the flowing structure is produced from smelting (Rijk 2007, 114-19). Additionally, analyses of the chemical composition and properties can support distinctions, and issues caused by slags' heterogeneous structures can be mitigated by examining larger numbers of samples, since a single analysis is subject to too many uncertainties (Rijk 2007, 113-14).
Iron furnaces are less abundant in the archaeological record and vary considerably across the globe (Pleiner 2000, 141-94). When Pleiner wrote his overview, nothing was known about iron furnaces in the Mongolian Plateau, but Pleiner noted that in regions next to our area of interest -the Minusinsk Basin and Tuva -underground tunnel-type furnaces are characteristic. He points out that such furnaces Fig. 1 Schematic overview over the steps in the iron production process from ore extraction to secondary smithing (Arnoldussen and Brusgaard 2015, 116 Fig. 1) are extremely well insulated and mechanically very stable, though the disadvantage was the limit "to which production capacity could be increased" (Pleiner 2000, 188-89). He hypothesizes that this well-insulated underground type was developed in areas with hard climatic conditions, although at the time he wrote the book, he could not trace its origin (Pleiner 2000, 189). He did not know about remains even further east at Ivolga in the Transbaikal region, where such an underground tunnel furnace from the Xiongnu period between the late third century BCE and end of the first century CE has been known since the 1950s (Davydova 1956, 273-274 Fig. 7;Davydova 1995, pl. 176).

Before the Xiongnu: iron smelting furnaces in Eastern Eurasia
It is quite clear that iron metallurgy in Mongolia was adopted from outside. Xiongnu iron technology and furnace construction techniques were not, however, adopted from the regions of Kazakhstan and China. 1 Iron smelting is known much earlier in Kazakhstan than on the Mongolian Plateau, as attested by iron slags found in the metallurgical center of Kent at the settlement of Alat, which dates no later than the twelfth century BCE (Varfolomeev et al. 2016, 8) and represents the earliest experiments with iron production in central Kazakhstan (Varfolomeev et al. 2017). In the lower layer of Alat only copper smelting was attested, while in the upper layer four iron furnaces, several pits, and two buildings came to light (Žauymbaev 2013;Evdokimov and Zhauymbaev 2013;Varfolomeev et al. 2017). The excavated iron kilns were constructed differently than the ones known from Mongolia (Zhauymbaev and Evdokimov 2008;Evdokimov and Zhauymbaev 2013, 436). Besides furnaces for iron-smelting, kilns for roasting ore were also identified (Žauymbaev 2013, 437).
The widespread use of iron in China is known from the third and second centuries BCE (Wagner 2008, 112). Since the earliest iron artifacts are known from Xinjiang (Wagner 2008, 91-93), it was suggested that bloomery iron technology was transmitted to China from the northwest, the Fergana Valley, or the Eurasian Steppe (Wagner 2008, 97;Qian and Huang 2021, 4). However, from the Warring States period (457-221 BCE), blast furnaces were used in China to produce cast iron, whereas the bloomery smelting process did not catch on (Wagner 2008, 105-7;Qian and Huang 2021).
The Evidence from the Minusinsk Basin. Sasada and Ishtseren point out that the Xiongnu adopted iron metallurgy from the Minusinsk Basin, where the metal industry was already well developed in the Bronze Age, and this has been the subject of much research (Sunchugashev 1969(Sunchugashev , 1975(Sunchugashev , 1979(Sunchugashev , 1993. Copper smelting furnaces of the Tagar and Tes' periods in the Minusinsk Basin are very different in overall construction to the iron smelting furnaces (Sunchugashev 1975, 87 Fig. 28;94 Fig. 34). But with its long flourishing bronze industry, these experts in pyrotechnology had gathered experience in managing heat and airflow to furnaces for centuries prior to the Xiongnu. By 1979, Iakov Sunchugashev had identified 34 sites from different periods associated with iron production in the Minusinsk Basin and excavated many of them (Sunchugashev 1979, 13 Fig. 1). However, little research has been conducted on this issue after him (Murakami 2013;Amzarakov 2015a, b;Tulush 2017).
The oldest iron smelting furnace in the Minusinsk Basin is found at Ulus Zimnik, and it is said to belong to the Tagar period because diagnostic Tagar pottery was found there (Sunchugashev 1993, 70). The kiln was truncated by a modern canal and not preserved completely ( Fig. 2.1). From the image published (Sunchugashev 1993, 70 Fig. 63), it appears that this furnace was a rectangular pit measuring 60 × 30 cm and 40 cm deep. Similar pit furnaces were uncovered in Sarala ( Fig. 2.2), and Sunchugashev dated the undecorated pottery to the late Tagar period (Sunchugashev 1993, 89-96), which he assigned to the third to second centuries BCE (Sunchugashev 1993, 91). Since his publication, the absolute dates and cultural attributions have changed: for example, it was still customary in the 1990s to assign the Tes' phase to the Tagar culture (Parzinger 2006, esp. 621). Since neither Sunchugashev's descriptions of the pottery nor his black and white drawings allow for a more refined chronological attribution, the sites could either be what is currently defined as late Tagar (fourth to second centuries BCE) or the subsequent Tes' culture (second to first centuries BCE) (Amzarakov 2015b).
Most of the furnaces that Sunchugashev investigated, however, were attributed to the Tashtyk period (ca. first to fourth centuries CE by current definitions) based on pottery unearthed at some of the sites (Sunchugashev 1979;1993). The installations were regularly accompanied by finds of clay tuyères (Sunchugashev 1993). The dominant furnace type attributed to the Tashtyk phase is an underground furnace with an attached tunnel; however, its appearance varies considerably (Sunchugashev 1979, 30 Fig. 3;34 Fig. 7;40 Fig. 13). Whether this variation indicates changes over time or different kilns being used concurrently remains to be determined. In any case, it remains uncertain when such underground tunnel-type furnaces came into use, since we lack a decent series of modern radiocarbon dates. There are two modern excavations at iron smelting sites that provide insight also into the dates. In Tolcheya, a Russian-Japanese team has uncovered a central rectangular pit measuring 1.5 × 1.2 m surrounded by five iron smelting kilns (Amzarakov 2015a, b). The kilns were connected to the central pit by underground tunnels (Fig. 2.3). In one instance, a ceramic pipe was preserved, indicating that the bellows were operated from the central pit (Amzarakov 2015a, 96).
Based on the excavations of these two sites, Amzarakov suggests that the types of furnaces with underground air ducts at Tolcheya and the later pit furnaces from Troshkino represent a technological advance at the turn of the era (Amzarakov 2015a, 98). If accurate, this development is similar to the changes to iron-smelting furnaces seen in Mongolia, as shown below. Overall, one has to bear in mind that there is so far no evidence for underground tunnel-furnaces prior to the Tes' period in the Minusinsk Basin, parallel to the early Xiongnu period on the Mongolian Plateau.

Iron Kilns of the Xiongnu period in Mongolia, the Transbaikal area, and Tuva
The first Xiongnu-period iron smelting furnace was excavated in the 1950s by Antonina Davydova at the fortified settlement of Ivolga in the Transbaikal area (Davydova 1956, 273-74;1995, 51). Under a square mound measuring 30 × 30 m and 1.65 m in height in the center of the settlement, an iron-smelting furnace was excavated ( Fig. 3.1). It consists of two pits connected by an underground channel with an additional small pit nearby ( Fig. 3.2). The furnace itself is a rectangular pit measuring 37 × 37 cm with a depth of 35 cm. Its corners are rounded and the walls covered with clay. In the pit and the entrance to the tunnel, slags and charcoal were recovered. On the surface to the southwest of the furnace, parts of a heavily burnt clay cupola were found (Davydova 1956, 273-74). Directly bordering the furnace to the southeast, there is another round pit with a diameter of ca. 25 cm and a depth of 50 cm. During a research stay in 2016 in Buryatia, Tomotaka Sasada confirmed that clay tuyères were found in this context as well (Sasada and Ishtseren 2020, 12). While the two pits connected by a tunnel can be interpreted as the remains of a bloomery furnace with tuyères, the third pit was either another furnace or an installation with a different function that cannot be determined based on the publication. In Tuva, iron furnaces have already been investigated in the 1960s (Sunchugashev 1969, 112). At that time, Sunchugashev identified 30 sites associated with iron metallurgy (Sunchugashev 1969, 44 Fig. 24). There are basically two types of furnaces: a pit-type furnace and an underground one with a tunnel connecting the furnace to a pit. Generally, tuyères were found in the smelting installations at all sites. Because of similarities to the furnaces from the Minusinsk Basin, Sunchugashev dated the iron furnaces of the Baĭ-Siut valley in Tuva (Turlig, Chinge-Aksy, Tardan, Surug-Chem) to the Tashtyk period (Sunchugashev 1969, 126). But he also noted their similarity to local copper melting kilns and the furnace at Ivolga, which would suggest a somewhat earlier date for the furnaces. Iron production certainly began earlier than the Tashtyk period, as the most ancient iron mines belong to the third and second centuries BCE (Sunchugashev 1969, 106), and one belongs to the period between the late second century BCE to the first century CE, as typical Xiongu pottery was found there (Kara-sug, Sunchugashev 1969, 107-108, Fig. 51). This early study, however, does not resolve which furnace type was used during the second and first centuries BCE. To our knowledge, no modern radiocarbon dates exist for the iron smelting furnaces in Tuva from the periods of interest.

Khustyn Bulag, Kherlen Valley
Research on iron furnaces during the Xiongnu period on the Mongolian Plateau has been catapulted forward since 2011 by a Mongol-Japanese team excavating in Khustyn Bulag in the Kherlen River valley (Amartüvshin et al. 2012;Sasada and Amartuvshin 2014;Sasada and Ishtseren 2020;Ishtseren et al. 2020). Over the years, the team has investigated a total of 12 installations connected to iron-smelting, and so far this is the largest iron-smelting site known from Mongolia ( Fig. 4). Three furnace types were identified at Khustyn Bulag (Ishtseren and Sasada 2018;Sasada and Ishtseren 2020;Ishtseren et al. 2020) Type 1 (kiln nos. 4, 5, 7, 10, 12) consists of a small square pit about 50 × 50 cm, with a larger oval or square slag disposal pit attached to it ( Fig. 5.1).
Type 2 (kiln no. 1) is just one large slag pit that measures about 1 × 0.5 m, in which a large amount of slag was found ( Fig. 5.2).

3
Type 3 (kiln nos. 3, 6, 8, 9) also consists of two parts, but the slag pit of the furnace is connected to the larger slag disposal pit by a tunnel (Fig. 5.3a-b).
All three furnace types yielded fragments of tuyères. Besides the remains of walls, fragments of bellows and slags were discovered, as well as stone hammers, a stone anvil, and a few pieces of pottery (Sasada andAmartuvshin 2014, 1018;Sasada and Ishtseren 2020). The stratigraphy points to a succession of furnace types, with type 3 belonging to the older horizon and type 1 to the upper layer. Furthermore, in the upper layer, more than one clay tuyère was used for a single furnace, which testifies to improvement to the blasting technology for the air supply. Also the use of charcoal in the younger phase instead of wood in the older phase points to technical improvement (Sasada andAmartuvshin 2014, 1023), however a charcoal pile was not identified during the excavations. The evidence of a different type of furnace (type 1) in the later Xiongnu period, as well as the use of more tuyères associated with this younger group and the more abundant use of charcoal, shows that the bloomery process became more efficient (Sasada and Ishtseren 2020, 12). The series of radiocarbon dates from Khustyn Bulag confirms this temporal succession of furnace types (Sasada and Ishtseren 2020, 12). The settlement of Ivolga with the location of the iron kiln in its center with schematic drawing of the kiln and other indicators of metallurgical (bronze) processing (Davydova 1995, Fig. 2;pl. 176) Besides furnaces pits for roasting ore were identified in Khustyn Bulag and wood was used as fuel for this step of the iron production (Sasada and Amartuvshin 2014, 1023). 5

Baga Nariĭn Am, Orkhon Valley
In central Mongolia, in the Orkhon Valley 6 km south of Karakorum, five iron-smelting furnaces that can be attributed to the Xiongnu period, as well as four more undated ones, were excavated at the site Baga Nariĭn Am between 2012 and 2013. The site is located on the left bank of the Orkhon River, where the river flows from the Khangai Mountains into the wide plains of the middle Orkhon Valley (Fig. 6). During two surveys in spring and fall 2009, members of the Mongolian-German project "Geoarchaeology in the Steppe", headed by Jan Bemmann, collected substantial materials (coins, metal artifacts, pottery) from the Mongol period (Bemmann et al. 2011). In the nearby capital of the Mongol Empire, Karakorum, secondary iron processing was identified (Reichert 2020, 99-110). Finds of metal slags and a geophysical survey showing strong anomalies in the SQUID (Superconducting QUantum Interference Device) magnetograms suggested potential Mongol period metalworking at Baga Nariĭn Am (Fig. 7;Pohl et al. 2012, 54-56). With the intention of investigating metal production, we opened one trench where the magnetometry had indicated metalworking (Fig. 8).

Excavated furnaces
In the western trench (HC 38), we excavated five ironsmelting kilns (features 9/12,10,11,13,15,Figs. 9,10,11,12,13,and 14). 6 These installations are very similar to Sasada's type 3, which is characterized by a tunnel connecting the furnace with a pit (Fig. 5.3). After removal of the humus layer, each furnace appeared as an oval-almost kidney-shaped dark coloration in the soil next to a larger roundish pit (Figs. 11.1, 12.1, 13.1, and 14.1). The shape of the furnace itself may have been rectangular with rounded corners, similar to furnace 9, which was documented at a deeper level (Fig. 10.1.3). In the cases of features 9/12, 11, 13, and 15, each furnace was connected to one pit; only feature 10 had two furnaces connected to one and the same pit ( Fig. 11.1). The walls of the underground furnaces were inclined toward the larger pit (Figs. 11.2.3, 12.2, 13.1, and 14.2). Only one slag block remained at the bottom of furnace 9 ( Fig. 10.4), while in all other features a mixture of slag fragments and charcoal had accumulated in the larger pit, as well as in the furnace itself (Figs. 11.3,12.3,and 14.3). No tuyères were found during the excavations.
Although underground tunnel-type furnace have long been known, questions regarding the function/s of the underground tunnel remain. Pleiner assumed that its primary function was to act as an air duct into the combustion chamber, but he admits that "questions concerning the operation, air supply and blowing of these furnaces" remain (Pleiner 2000, 185 Fig. 50;189). Vodyasov questions this but does not provide an alternative explanation (Vodyasov et al. 2021, 4). At Baga Nariĭn Am, the furnace walls opposite the underground tunnel were consistently discolored red from the heat (Figs. 11.2.3,12.2,13.1,and 14.2). This may indicate that the greatest heat accumulated here, which would corroborate that the tunnels were used as air ducts. In simulations of heat distribution in later shaft furnaces for cast iron, the greatest  Sasada and Ishtseren 2020, 9 Fig. 4), type 2 (2, Sasada and Ishtseren 2020, 10 Fig. 7) and the tunnel-type furnace 3 (3-4, Sasada 2015, 53 Fig. 8;Sasada and Ishtseren 2020, 9 Fig. 5) 1 3  (1) with overlay of the SQUID magnetogram (2). Google Earth, location 48 T N331939m E5227040, image from 8/10/2018, Image ©Maxar Technologies. SQUID Magnetogram © Sven Linzen heat accumulated opposite of the air duct which supports this hypothesis (Qian and Huang 2021). Another function of these tunnels, as they led from the pit downward to the bottom of the furnace, was potentially to tap and remove the slag into the working pit, as larger amounts of slag fragments are found in the tunnel and in the working pit connected to the tunnel when a slag block was not found in situ in the lower part of the furnace.
The radiocarbon dates show that the kilns at Baga Nariĭn Am are the oldest known iron-smelting furnaces for the Xiongnu period (Fig. 18). 7 They are also the first from central Mongolia.
A second trench was placed along the eastern edge of the terrace (HD 84-94). Four features were investigated. Features 1 and 4 are similar in construction to the furnaces mentioned above but smaller (Figs. 15,16,and 17). They present as rectangular pits with rounded corners and burnt pit walls (Figs. 16.1.3 and 17.1.3). Attached to the pit is a small duct (Figs. 16.2 and 17.4). In all features, fewer slag fragments were found than in the larger installations in the western trench. Feature 3 is probably just a less well-preserved example of this type of furnace, while the larger feature 2 was only partially investigated and cannot be assessed accurately.
Currently, there are no absolute dates for these features. Their tunnel constructions are reminiscent of the kilns excavated in the western trench; however, they are smaller and less charcoal and slag was generally found. Because of the underground components, which point to a specific furnace construction and firing practice currently known only in Mongolia during the Xiongnu period, we think it is likely that these installations also belong to that period. The structure of the furnace hearths possibly points to an additional step in the iron smelting process. Confirmation of the dates and exact function of the furnaces will hopefully be provided through further indepth analyses of the slags. Overall, the five iron-smelting furnaces excavated and the other four smaller potential smelting installations at Baga Nariĭn Am allow a second iron smelting center of the Xiongnu period to be tentatively identified.

Analysis of the magnetometry reveals iron smelting center
After having verified by excavations that the strong magnetic anomalies are iron-smelting furnaces the question arose how many furnaces can be identified through the magnetogram. Our analysis of the point-shaped magnetic anomalies is based on the magnetogram data shown in Fig. 8.1, as well as data from additional SQUID gradiometers (S. Linzen et al. 2007), applied to the Baga Nariĭn Am area. Depth calculations of the magnetic signal origin are thus possible which enhance the classification of the detected buried archaeological objects. The magnetic anomalies of the iron-smelting furnaces are characterized primarily by their high amplitudes of up to 180 nT/m. Thus, they are easily visible within a special magnetographic representation, which suppresses small signals with a grey level scale of ±50 nT/m (e.g. Figure 8.1). If we reduce the magnetogram amplitude scaling to a usual value of ±10 nT/m (see Fig. 7.2), a lot more anomalies become visible, which, however, correspond mostly to archaeological remains of other periods. Moreover, the furnace anomalies themselves are more complex in shape than those generated by a simple magnetic point source. Double anomalies of very different intensities often correspond to two concentrations of variable amounts of slag. A large and compact concentration of slag -like in feature 9 -results in a strong magnetic anomaly. Whether the magnetic properties of other materials nearby or mixed in with the slags contribute significantly to the total of furnace anomalies detected is the subject of ongoing discussions.
In sum, a total of 26 similar anomalies in the same area of Baga Nariĭn Am are detectable in the SQUID magnetogram data (Fig. 8.2 red and blue circles). Aside from the five already excavated, a further 10 magnetic anomalies ( Fig. 8.2 red circles) can be classified with relatively high confidence as furnaces of the same type. Another 11 anomalies (Fig. 8.2 blue circles) are remarkably similar to those excavated and are possibly also furnaces. Furthermore, anomalies theoretically generated by even smaller furnaces or installations associated with other steps in iron production are found beyond the main concentration of anomalies ( Fig. 8.2 green circles).
Overall the magnetogram allows to identify with certainty a minimum of 15 iron smelting furnaces, possibly up to 26 in total. This substantial number of iron-smelting furnaces at Baga Nariĭn Am suggest that an iron-smelting center like that at Khustyn existed here. Together with the recent identification of Khargany Dörvölzhin (aka the Dragon City or Longcheng) in the Orkhon Valley, this contributes to a growing body of evidence for the region's importance within the Xiongnu Empire.

Dating of iron kilns of the Xiongnu period
Apart from these two iron smelting centers, more iron kilns have been excavated in western and eastern Mongolia over recent years (Fig. 18) (Ishtseren et al. 2020, 77-79;91 Fig. 1). Most of them belong to the tunnel-type (Sasada's type 3), and a few belong to type 2 or type 1. Interestingly, two furnaces are located in the Mongolian Altai, while in the Russian Altai the evidence for iron smelting during the Xiongnu period is scarce, with only one piece of iron slag collected from an oval pit dated to the Pazyryk and Xiongnu periods (457-46 cal BCE), indicating that iron smelting took place. The exact structure of the furnace remains, however, unknown. A piece of charcoal collected from an iron smelting installation at Barun-Khal II in the Cis-Baikal region has also produced a date within the Xiongnu period (166 cal BCE -59 cal CE), but again, no information about the furnace type is available.
We have gathered all available radiocarbon data from the literature on iron smelting furnaces from Mongolia for which the furnace type could be identified (Fig. 19). Types 2 and 3 were used during the same time period, mainly the second and first centuries BCE, while type 1 was dominant from the first century BCE through the first century CE. So far, no estimates for the potential outputs of these furnace types have been attempted, but observations at Khustyn Bulag indicate that the abandonment of the tunnel-furnace type was part of improvements to the smelting process (Sasada andAmartuvshin 2014, 1023;Sasada and Ishtseren 2020, 12).

Organization of the iron industry in the Xiongnu empire
Despite major advances in identifying and excavating ironsmelting furnaces the reconstruction of the full châine operatoire and our knowledge of the spatial organization of the iron industry is fragmentary. In the following we draw together the existing evidence, also from neighboring regions and other time periods in order to provide the fullest possible picture on the iron industry in our area of interest. Ore extraction is the first step. Ancient mining, however, has not been the subject of study for any time period on the Mongolian Plateau, so we know little about it. A factor that exacerbates this situation can be highlighted by looking at Europe, where there has been more research. While a large number of iron smelting sites is known, the evidence for mining is much more sparse, as lower demand in the early stages of iron exploitation means that little to no trace appears in the archaeological record (Pleiner 2000, 93). This is well reflected in maps in the well researched iron production landscape of the Siegerland or in the Taunus (Garner and Zeiler 2020, inserts 1-7; Posluschny and Schade-Lindig 2019, 201 Fig. 7). The maps show clearly that iron smelting sites are much more numerous than ore mining sites. It also shows that smelting sites are generally close to places of ore extraction. Smithing, however, is oftentimes spatially separated from the smelting sites.
In our general area of interest, in the Minusinsk Basin of the Tashtyk period, mining sites have been excavated close to the river Chernyĭ Iius, with the nearest smelting sites at a distance of 2-5 km (Sunchugashev 1993, 86 Fig. 80) and thus seem comparable to the examples from Germany. For the Xiongnu iron smelting center Khustyn Bulag excavated ore corresponds to samples found almost 30 km from the Fig. 9 Overview over the western trench HC38 with furnaces (F). Graphics Ernst Pohl and Anna Stefanischin site. But the authors suggest also another ore source only 15 km away may also have provided the ore, which is not yet corroborated by scientific analyses (Sasada and Amartuvshin 2014, 1018 Fig. 2; 1020-1021). For our site of Baga Nariĭn Am, no iron deposits in the vicinity are currently known and no analyses of slags has tried to match the ore to known iron deposits from the Orkhon Valley.
And for the site of Katylyg 5 in Tuva which post-dates the Xiongnu Empire, researchers even suggest that ore from different mines that are located as far as 50-70 km away were smelted in the furnaces, which seems very far. Research is ongoing to determine whether ore was indeed transported over such distances or whether ores closer to the smelting site were exploited (Vodyasov et al. 2021, 11-12).
Ore smelting was separated from settlement contexts, except for the site of Ivolga, where the iron smelting furnace was found within the enclosed area of the settlement. At all the other smelting sites, there is no indication of a nearby

Graphics Ernst Pohl and Anna Stefanischin
Fuel is another topic that is not yet sufficiently understood for the Xiongnu period which also ties into a lack of high-resolution data for environmental reconstruction during that time period. In Khustyn Bulag charcoal was used as fuel (Sasada andAmartuvshin 2014, 1023), however, no charcoal pile for charcoal production has yet been identified at any smelting site of the Xiongnu period. Only in Tuva, several charcoal piles have been investigated: One large pit measuring 2 × 2 × 0.7 m with two layers of charcoal is undated (Sunchugashev 1969, 121 Fig. 61). And just outside the post-Xiongnu settlement Katylyg 5, two pits measuring 2.8 m in diameter and more than one meter deep testify to the production of charcoal (Vodyasov et al. 2021, 2-3 Fig. 2). Whether charcoal was even the main fuel or whether wood was used a fuel as well and whether that changed over time, still needs to be determined.
Another question regarding the organization of the iron industry is where primary and secondary smithing was taking place. Small slag fragments have been found in habitation contexts at the permanent settlement of Boroo, mainly Most of the small pieces were adhered to the wall of a small bowl, thus Serneels proposed that the bowl was used in a small forge. The small size and the small quantity of debris point to low intensity activity at Boroo (Serneels 2013, 195-96). The scarcity of remains related to metallurgical production can perhaps be explained by the use of an itinerant blacksmith who only spent a few hours on site as opposed the presence of a permanent metallurgical installation (Ramseyer and Pousaz 2013, 217-18). The surface collections of slags at the open settlement of Dureny shows that some processing of iron took place there but it cannot be determined whether the slags were the result of the smelting or smithing processes (Davydova and Miniaev 2003, 10-13). In the foothills of the Khanui valley, Houle and his team were surprised by the amount of evidence for iron production in the form of small slag fragments, which they recovered from test excavations in six occupation areas with no evidence for permanent settlements (Houle and Broderick 2011, 150).
Similar to Boroo, the slags are indicative of iron processing in domestic contexts on a small scale. This is also attested for pastoralists' dwellings in north-central Mongolia, albeit for a later period (Park et al. 2020).
Current research suggests that more iron-smelting sites and centers will become known in different parts of Mongolia. With the exception of Ivolga, iron-smelting is not connected to or associated with settlement sites. One has to bear in mind, however, that extensive settlement research has not yet taken place. Slags from seasonal habitation sites, as well as settlements with permanent buildings, point to small-scale yet widespread iron-working, probably primary smithing.

After the Xiongnu
A recent and comprehensive study of iron metallurgy in Tuva concerns the period after the demise of the Xiongnu Empire (Vodyasov et al. 2021). The fortified Kokel' culture site of Katylyg-5 was used between the second and fourth centuries CE (Sadykov 2018, 72 Table 1). 9 Though iron slags were found in varying concentrations across the site, the northwestern area, where nine iron furnaces were excavated, was used for smelting, while smithing occurred in the southern part of Katylyg 5, clearly separated from the former (Fig. 20). Charcoal from the furnaces and smithing area indicate they were mainly in use in the third and fourth centuries CE (Vodyasov et al. 2021, 6 Fig. 8). The furnaces are trapeziodal in shape, about 1 m deep, and were connected to a slag pit by an underground channel (Vodyasov et al. 2021, 5 Fig. 5).
These are similar to iron smelting furnaces that have been investigated over the past decades along the northwestern shore of Lake Baikal and on Olkhon island (Kozhevnikov et al. 2001;Kharinsky and Snopkov 2004;Kozhevnikov and Kharinsky 2005;Kozhevnikov et al. 2019), where a total of six iron production sites have been identified (Kozhevnikov et al. 2019, 105 Fig. 2). More comprehensive excavations of iron smelting installations have been conducted at three sites that were also accompanied by geophysical surveys: Barun-Khal II, Barun-Khal III, and Kurminskoe ozero 1 (Kharinsky and Snopkov 2004;Kozhevnikov et al. 2019). At Barun-Khal, slags were found over an area of about 0.15 km2, which indicates iron production on a large scale (Kozhevnikov et al. 2003, 526). The chemical and mineral compositions of the slags points to the use of the bloomery process (Kozhevnikov et al. 2001). At all three sites, similar furnaces were excavated: they are all underground constructions with large working pits and additional, elongated trenches, sometimes partitioned with bricks (Fig. 21). Attached to the working pits/trenches are trapezoidal  (Snopkov et al. 2012), several furnaces were also investigated (Kharinsky and Snopkov 2004). Again, around a larger working pit several furnaces are connected to the pit by a tunnel. However, next to the large smelting construction, there are also three furnaces that resemble Sasada's type 3, with two pits and a tunnel connecting them ( Fig. 21.1, furnaces 6-8). The radiocarbon dates from these three sites indicate that smelting took place repeatedly over a long time span; most of the dates belong to the period after the collapse of the Xiongnu Empire (Fig. 22). No settlement with further iron-working installations can be associated with those described above.
Excavated furnaces from the Russian Altai belong also mostly to the period after the demise of the Xiongnu Empire (Ziniakov 1988(Ziniakov , 2019. 10 In the Iustyd valley, Ziniakov has identified a type of furnace which has a D-shaped furnace pit, between 60-80 cm deep, and a larger slag disposal pit in front of it (Fig. 23.1) (Vodiasov and Zaĭtseva 2020, 132 Fig. 4). At a depth of ca. 30 cm, remains of one or two clay tuyères were found (Vodiasov and Zaĭtseva 2020, 132). Although no analogies are known for this exact furnace type (Vodiasov and Zaĭtseva 2020, 133) from neighboring regions, it appears to be roughly similar to the pit type of the Tashtyk period in the Minusinsk Basin (Murakami 2013). From the excavations of the Iustyd-type furnaces, one block of slag has been identified, and its charcoal inclusion produced a radiocarbon date of 261-557 cal CE (95.4%).
In the Chuya-Kurai area, there are numerous metallurgy sites, which feature a specific Kosh-Agach type of bloomery iron furnace. This box-shaped furnace consists of a rectangular pit lined with stones ( Fig. 23.2). Characteristically, one finds holes for air in the furnace wall. 11 The typological characteristics and ceramics associated with it led the excavator to conclude that iron smelting in the southeastern Altai took place between the sixth and tenth centuries CE (Ziniakov 1988, 51;Agatova et al. 2018a, b, 94;Vodyasov et al. 2020, 8). The recently refined dating of iron furnaces in the Russian Altai allows the Kosh-Agach type of furnace to be dated between the late third century through the ninth century CE (Fig. 24) (Agatova et al. 2018a, b;Murakami Fig. 17 Furnace 3 and 4. Furnace 3 with schematic drawing of top-view (1) and profile (3), furnace 4 with schematic drawing of top-view (1) and profile (2) and photo (3). Graphics Ernst Pohl and Anna Stefanischin et al. 2019;Vodyasov et al. 2020;Vodiasov and Zaĭtseva 2020). No clay tuyères have been found in association with this type of kiln.
A recent review of iron furnaces in the late first and second millennium CE shows that the Minusinsk Basin and the area along the northern shore of Lake Baikal has shown similarities in the development of iron smelting sites (Kharinskiĭ and Snopkov 2020, 88). Iron furnaces of the second millennium CE were constructed very differently from earlier ones. In the Cis-Baikal region, iron-smelting furnaces were shallow rectangular boxes lined with stone slabs. This type is also known from the Minusinsk Basin, though the documentation is not as detailed nor is the dating as specific (Sunchugashev 1993, 103 Fig. 99). The iron furnace documented by Gero von Merhart might also belong to this period (Merhart 1929). 12 The development of the iron industry in the Altai and on the Mongolian Plateau from the late first and in the second millennium CE, is currently not known. Sunchugashev reports accumulations of iron slags in Uyghur and Mongol-period settlements in Tuva (Sunchugashev 1969, 111), but it remains unknown whether the slags mentioned are smelting or primary smithing slags. From settlements of the Mongol Empire period in Mongolia, no accumulations of slags have been reported to date.

Discussion
This overview of iron smelting furnaces in eastern Eurasia coupled with new data from central Mongolia allows narratives of the spread and adoption of iron metallurgy in Mongolia and its neighboring regions to be updated and developed. Through the data generated from our fieldwork in Baga Nariĭn Am, as well as the analyses of the published literature, the following picture emerges.
The earliest iron smelting furnace in eastern Eurasia probablay belong to the late Tagar period (4th-2nd c. BCE) while iron objects were circulating already earlier. That the earliest furnaces are located in the Minusinsk Basin is not surprising, considering the region's long and flourishing bronze industry -the necessary experience and pyrotechnological knowledge was already present. However, the Xiongnu polity established at the end of the third century BCE in the Mongolian Plateau was the driver behind the expansion of eastern Eurasia's iron industry during the second and first centuries BCE (Vodyasov et al. 2021). But, while this boom has previously been noted for the first century BCE and first century CE (Vodyasov et al. 2021), the new data from Baga Nariĭn Am in central Mongolia underscores that this process had already started in the second century BCE and correlates with the establishing of the steppe empire.
Furthermore, the detailed analysis of the magnetic SQUID prospection data showed that many more iron furnaces existed and revealed thus a larger iron production site at Baga Nariĭn Am. The recorded quality of the buried furnaces' magnetic anomalies in combination with the knowledge of the excavations bear potential for future simulations of the occurring variety of furnace constructions with their different depositions of slags, whether as conglomerate of slag fragments mixed with soil (e.g., furnace 10, Figs. 8 and 11) or as a deposit of a larger compact slag block (e.g., furnace 9, Figs. 8 and 10). Such an improved understanding of the magnetic anomaly appearances will support the locating of further furnaces of the investigated type as well as the differentiation from ones of other constructions and ages.
We also suggest that the metallurgists of the Xiongnu Empire invented the tunnel-type furnaces (Sasada's type 3) as the earliest furnaces of this type are known from the Mongolian Plateau and belong to the second century BCE. A further technological improvement of the iron-producing process is noted in the later Xiongnu Empire during the late first century BCE and first century CE, when the underground tunnel-type furnaces were replaced by furnaces that appear simply as two pits in the archaeological record, with no tunnel connecting the pit and furnace. Observations at Khustyn Bulag led Sasada and his team to conclude that this change of furnace type, which is accompanied by a larger number of tuyères per furnace and the use of charcoal instead of wood as fuel, indicate technical improvement. Together with Pleiner's assessment that the underground tunnel-type furnaces are not scalable in their output, we hypothesize that this change in furnace type was fuelled by greater demand for iron in the later phase of the Empire (late first century BCE and first century CE). Burial furnishing probably reflect such a shift in demand. During the early phase of the Empire, in the second and first centuries BCE, bronze belt plaques appear, while later burials have belt plaques with iron backings. Also, in the satellite burials of large terrace toms in the late Xiongnu period, such as tomb 1 of Gol Mod 2, iron artifacts dominate, though both materials were in use simultaneously. Overall, the mass of iron fragments excavated from Xiongnu burials indicates a high demand for this material in the later period.
Thus, we suggest that the boom of the iron technology as well as further steps of technological improvement in eastern Eurasia can generally be linked to the Xiongnu steppe Empire. And thus it is all the more interesting that tunneltype furnaces continued to be used after the demise of the Empire in the Minusinsk Basin and on the northwestern shore of Lake Baikal. This ties into discussions about the connection of the adoption and spread (or rejection) of a new technology with the social context (Hansen et al. 2016, 780-782;Erb-Satullo 2019, 583). In other world regions, in the Levant, for example, the expansion of the iron economy is driven by imperial consolidation in Mesopotamia and eastern Anatolia in the ninth to seventh centuries BCE (Erb-Satullo 2019, 584-590; 593). The innovation and spread of Fig. 20 Iron Processing site of Katylyg-5 in Tuva. 1 drawing and profile section of a typical iron smelting furnace, 2 map of the site with smelting area and smithing area indicated. To the southwest two nearby charcoal piles (Vodyasov et al. 2021, 3 Fig. 1, 5 Fig. 5B) iron technology is coupled to political authority, similar to the situation in the eastern Eurasian steppes. Situating the metallurgical system within its broader social contexts requires also the consideration of the copper alloy production system. In Mongolia, we know little about bronze production. A few copper smelting sites are known, but they are not comprehensively published (Eregzen 2015). As far as can be judged, stones were used to build small, shallow smelting installations (Eregzen 2015, 33 Fig. 4). At Ivolga, numerous bronze slags and crucibles have been found (Fig. 3), but no workshop devoted to bronze production has been identified (Reichert 2018). The only other way to learn about the bronze metallurgy of the Xiongnu period is alloy composition data that reveal regional differences in production and point to an opportunistic approach to bronze production without standardization (Miniaev 1980(Miniaev , 1983Park et al. 2011;Park et al. 2018;Brosseder and Hsu 2022). This speaks to a decentralized organization of bronze production.
Based on the current data, we hypothesize that iron and copper metallurgy was organized in similar ways: both were decentralized; smelting/production occurred in several central regions; and the secondary production of iron as well as bronze-casting occurred in settlements (Reichert 2018). At the household level, small-scale need for metal was possibly met by itinerant craftsmen. To further our understanding, a comprehensive understanding of Xiongnu period settlements in northern Mongolia is needed, allowing also a Fig. 21 Iron smelting furnaces from the Cis-Baikal area. 1 Kurminskoe ozero (Kharinsky and Snopkov 2004, 170 Fig. 4), 2 Barun-Khal 2 (Kozhevnikov et al. 2019, 112 Fig. 10), 3 Barun-Khal 3 (Kozhevnikov et al. 2019, 116 Fig. 16)

Fig. 22
Overview over the radiocarbon dates from the Cis-Baikal area. References mentioned in the text Fig. 23 Types of iron smelting furnaces from the Rusian Altai. 1 Iustyd type (Vodiasov and Zaĭtseva 2020, 132 Fig. 4), 2 Kosh-Agach type (Vodyasov et al. 2020, 4 Fig. 4) better reconstruction of the spatial organization of the metal industry.
It is equally important to clarify the chronological relationship between the iron-smelting industries in the Minusinsk Basin and at Xiongnu sites, not only to corroborate (or falsify) with modern radiocarbon dates our reconstruction of events, but also to further our understanding of the so-called Abakan palace, a unique adobe building in the Minusinsk Basin. This building was incorporated into a larger planned settlement or city (Kyzlasov 2001;Postică Kyzlasov 2010, 183-245) with a regular layout (Postică Kyzlasov 2010, 197). It was suggested that the palace was built in the context of obtaining raw metals (Parzinger 2003, 362). Whether this was the function of this possible Xiongnu outpost making use of the long-established metal industry, is unclear, but appealing. Metal objects from the Minusinsk Basin, such as belt plaques, despite local characteristics, show a strong connection to the Mongolian Plateau, whereas grave structure, burial rites, and bronze metallurgy all point to local traditions and practices. A better understanding of the spatial organization of the Xiongnu Empire and its production sector as well as of Xiongnu period settlements on the Mongolian Plateau will allow for deepening our understanding of the iron production in the Xiongnu Empire. Fig. 24 Overview over radiocarbon dates from the Russian Altai (Agatova et al. 2018a, b;Vodyasov et al. 2020) Although further data are required to refine our reconstruction of the iron production system, there is a clear relationship between the growth of the iron industry and the early Xiongnu Empire, which continued over time. The empires that were repeatedly newly established on the Mongolian Plateau, of which the Xiongnu was the first, provide a good opportunity to study the interconnectedness of the socio-political conditions and the production sector of the economy.

Description of the excavated features of Baga Nariĭn Am
The features presented here were excavated in the context of an archaeological project during summer seasons in 2012 and 2013. In 2012, the remains of small iron smelting installations (features 1-4) were excavated in the areas HD 84 and 94. During 2012 and 2013, five iron-smelting furnaces were investigated (features 9/12, 10, 11, 13, 15) in area HC 38. The area designations refer to a local excavation network, the hypothetical zero point of which was assumed to be southwest of the terrace of Baga Nariĭn am. The excavation areas usually have dimensions of 5 × 5 m, the area plans were laid out on the basis of artificial strata. Deviating from this, isolated features were also excavated and documented according to natural strata. The charcoal recovered from features 10-13 and 15 was analyzed by K.-U. Heußner from the German Archaeological Institute in Berlin and examined in relation to dendrochronological curves. Since the curve could not be connected to a master curve at the time, the oldest and the youngest tree ring were sent to the laboratory for 14C dating. A complete excavation report with the Mongolian-period finds and features of all excavation campaigns 2011-2013 is in preparation. Fig. 9) 7.1.1 Feature 9/12 (Fig. 10) Features 9 and 12 were excavated and documented in two campaigns in 2012 and 2013. With our excavation of the northern section HC 38/1-45 in 2013, we realized that feature 9, which we had documented one year earlier, was the furnace for pit F. 12. Since we had removed F. 9 in 2012, a cross section to clarify the situation and reveal the underground tunnel connecting both features was no longer possible. The profiles that we established on both sides of the baulk between sections HC 38/1-45 (excavated 2013) and HC 38/51-95 (excavated 2012) show inclined profiles of the features.

Western trench HC 38 (
Feature 9: After removing the topsoil, an irregular round pit came to light on pl. 4, some 0.4 m below the surface. The pit was filled with brown silty sand, including dark grey spots. This continued until pl. 6, 0.55 m below the surface, where the pit appears as a rectangular feature of 0.9 × 0.7 m with rounded corners. An outer ring of orange-burnt clayey sand shows that this place saw higher temperatures. Inside, there was another ring-shaped structure consisting of dark brown silty sand. The pit was filled on this level with light brown silty sand and one small black, burnt sand structure on the southeastern wall.
When troweling the feature, a big slag deposit about 30 cm wide and more than 45 cm high was found, starting some 0.2 m below pl. 6. This deposit included remains of burnt wood as charcoal.
Feature 12: On pl. 1, some 50 cm below the surface, we saw an irregular circular structure of at least 1.6 m in diameter with an fill alternating from light yellowish soil with small stones to a dark grayish-brown compact soil with charcoal inclusions.
On pl. 3 the structure became rectangular of at least 95 cm × 80 cm with a dark gray-brown compact filling. Below this, a dark gray compact layer of up to 20 cm thick with slag fragments becomes visible below which we documented a compact light brown layer. On the bottom of the fire channel to a depth of 1.05 m loose dark gray-brown soil with charcoal.  . 11) Two iron furnaces (features 10B, 10C) attached to a larger central pit (feature 10A) by tunnels.
Pl. 1 revealed a nearly circular pit of 1.4 m in diameter with a dark gray-brown, compact sandy fill with slate rubble and lumps of slag. The two kilns are visible on the surface of pl. 1 as oval to slightly kidney-shaped features at a distance of ca. 30-40 cm from the central pit. In order to reveal their structure we decided to cut a profile, first from feature 10A to 10 B, second from feature 10A to feature 10C.
Furnace 10B measures 1 × 0.6 m. The surrounding soil was discolored orange by heat up to 10 cm thick. Its fill was of dark brown compact sand with charcoal inclusions, but not stones.
Furnace 10C measures 1.1 m × 0.65 m. The surrounding soil at south end was discolored orange up to 6 cm by heat. The feature was filled with dark brown compact sand and dense slate rubble.
In the central pit (10A) below the slate rubble a 30 cm thick layer with fragments of slag was excavated, followed by an up to 30 cm thick, trough-shaped filling of irregular charcoal strips and sandy soil, which is colored orange on the sides of the charcoal due to the effect of heat. The pit is connected to the furnace feature 10B by a tunnel. In the upper part, but also on the bottom of the tunnel the heat has discolored the outside soil red up to 6 cm thick.
The filling of the kiln encountered at the surface can be traced to a depth of 20 cm. Below, a loose dark gray-black filling was noted that consists almost entirely of pieces of slag in the upper part and charcoal pieces until the bottom. This patch of charcoal was separated from the central pit (f. 10A) filling by a compact grayish-yellow to light brown filling.
The second kiln, feature 10C, is separated from the central pit (f. 10A) by a baulk up to 25 cm wide. Both features are connected by a tunnel. The fill of feature 10C that was encountered on pl. 1 to a depth of 25 cm. Below this, first a dark gray fill about 15 cm thick with charcoal was documented, then two compact grayish yellow and light brown fills; the latter interspersed with charcoal particles. On the bottom of the furnace channel a thick dark gray-black densely packed with charcoal reached a depth of 30 cm. The profile shows that the working pit cuts off the channel to the furnace. Kiln 10C is therefore the older one in this context. Finds and samples: slag and charcoal; 14C-sample: Ugams 46,973, 2130 ± 20 BP, 342-54 calBC.

Feature 11 (Fig. 12)
Furnace (F. 11B) with a pit (F. 11A) is situated in the northeast corner of the section and was not excavated completely.
On pl. 1 we saw an irregular circular pit in the northeastern corner oft he trench, feature 11A, that measures ca. 1.85 m in diameter. On pl 2-3 the pit is reduced in diameter to approx. 1.7 m. It was interspersed with dark gray-brown to dark brown compact sandy fill and slag lumps. The fragments of slags in pit f. 11A can be traced in profile up to a depth of 10 cm. Below this is a trough-shaped filling, max. 45 cm thick, made of irregular charcoal strips and sandy soil, which is colored orange on the sides of the charcoal due to heat effect.
Feature 11B was visible on pl. 1 as irregular oval discoloration of 1.0 m × 0.68 m. It was filled with dark brown compact sand with broken slate. The surrounding soil was discolored orange up to 6 cm thick by heat. Feature 11B was separated from feature 11A by a 40 cm strip of sterile soil. Both features were connected by a tunnel. The outer side of feature 11B, the tunnel and the sole of the tunnel are discolored red up to 5 cm due to heat. In the profile, the filling of dark brown, compact sand can first be traced up to a depth of 30 cm. Below is a filling made entirely of charcoal, which can be traced up to 80 cm deep below the baulk. Finds and samples: slag and charcoal. 13 (Fig. 13) Kiln (f. 13 B) with one pit (f. 13 A) connected by a tunnel. Some 40 cm below the surface, a nearly circular pit became visible (f. 13A) on pl. 2. It measures 0.95 m in diameter, filled in the majority with dark gray-black soil with slags and small slate rubble. The filling of the pit, feature 13A, can be traced in the profile as a trough-shaped filling with slag fragments and slate rubble to a depth of 38 cm. Below this is a sterile layer up to 20 cm thick of light brown, compact sandy material with small stone inclusions. A tunnel connects feature 13A with 13B.

Feature
Ca. 40 cm further southwest from pit f. 13A, a slightly kidney-shaped stain of 1.4 m × 0.6 m became visible (f. 13B). The fill in the northern part consists of loose, dark grayish brown sand with slag and slate fragments. The surrounding soil is discolored orange up to 6 cm by heat. The filling of loose, dark grayish-brown sand visible at the surface can be traced to a depth of 12 cm. Below this follows a filling of charcoal and slag, which extends to a depth of 85 cm below the surface of pl. 2. The sole of the tunnel is also discolored red by heat.
Finds and samples: slag and charcoal; 14C-sample: Ugams 17,009, 2160 ± 25; 357-111 calBC. 7.1.5 Feature 15 (Fig. 14) Pit (F. 15A) with furnace (F. 15B). On level 2-3 an irregularly round pit of max. 1.65 m diameter was noted (f. 15A). Its filling consists of compact medium-dark brown soil with little slate rubble, and a light, granular yellow streak in the southeast. Feature 15B was seen as a slightly kidney-shaped discoloration measuring 1.0 m × 0.6 m. Its filling consists of darker brown compact soil. Both features were connected by a tunnel.
Feature 15A: The filling of the furnace chamber is visible in the profile as a trough-shaped structure to a depth of 35 cm. The furnace itself has a filling which is inclined toward the tunnel and the pit f. 15A. On its outside wall the surrounding soil is discolored due to heat, which is also true for the sole of the tunnel. The filling of feature 15 B is brown and compact in its upper part ut to a depth of max. 20 cm. Below it, two fillings, each about 20 cm thick, slope obliquely down to the kiln pit. The upper one is dark graybrown with a charcoal band and little slate rubble, the lower one compact gray-brown. Underneath, a charcoal layer up to 30 cm thick extends to a depth of 80 cm below the natural soil that separates both features.
Finds and samples: slag and charcoal. HD 84-94 (Fig. 15) The creation of two additional trenches on the southern terrace surface was triggered by finds of small pieces of slag that had been picked up in higher frequencies on the surface at the edge of the terrace sloping down to the river. The trenches are located in squares HD 84/56-100 and HD 94/6-50. Four small technical iron-smelting installations (feature 1-4) were unearthed immediately below the surface. F 1, 3-4 were negatively excavated. We documented feature 2 only incompletely as it is located in the northwestern corner of the trench and extends beyond the trench's borders. 1 (Fig. 16) Furnace pit with lateral lower air duct on the terrace edge sloping to the southeast. Rectangular plan with rounded corners of max. 48 cm × 34 cm and a max. depth of 44 cm. Outside the pit the heat had burnt the surrounding soil up to 10 cm thick. The channel opened to the southeast with a length of 48 cm and a max width of 32 cm. The furnace pit and the channel are filled with two layers. On the bottom a max. 10 cm thick layer of gray-black burnt loose sand, the upper layer contains a fill of hard, brown to gray sand with fragments of slag, burnt clay (roof of the kiln?) and stones.

Feature 2 (Fig. 16)
Partly excavated iron furnace in the northwest corner of the trench. On level 1 immediately below the surface a rectangular pit with rounded corners that measures 1.05 m in length became visible. Width to northern edge of trench max. 36 cm. In the drawing of the northern profile, a pit of 1.2 m length is visible, extending further west beyond the trench. The depth of the pit measures max. 36 cm in the western part, while the eastern part is only 20 cm deep. The fill consists of two layers: the lower layer, up to 16 cm thick, consists of gray-black burnt loose sand with black streaks of ash and charcoal in its upper part. The upper layer consists of up to 23 cm thick of hard brown to gray sand and contains fragments of slag and burnt clay.

Feature 3 (Fig. 17)
After having cleared the top soil the outline of a elongated pit extending northwest-southeast became visible. The southeastern part is incompletely preserved because of the sloping terrain of the terrace. The pit measures 70 × 60 cm. The northern part of the pit has been excavated. Its northwestern and northern border is of reddish color and burnt up to 6 cm thick. The southern profile reveals a maximum depth of 22 cm. The fill of the pit consists of gray to dark brown sand, in its southeastern part, the lower part of the fill consists of a max. 8 cm thick layer of gray-black burnt sand. This feature represents the remains of a small furnace pit with a channel to the southeast. No findings, but 30 cm to the south pieces of slag were found. 4 (Fig. 17) Ca. 1 m further south of feature 3 we excavated another small furnace pit with lateral duct on the terrace edge sloping to the southeast. The pit we found was rectangular with rounded corners and measures max. 52 cm × 42 cm and reaches a depth of 44 cm. The borders of the pit were red and the soil burnt up to 8 cm thick. The duct to the southeast had a length of 52 cm and a width of 38 cm. The furnace pit and the channel were filled with three layers. On the bottom a max. 6 cm thick layer of gray-brown sand (discoloration of the subsoil?), above it a max. 6 cm thick layer of gray-black burnt loose sand. And the upper fill consists of hard brown to gray sand which contained fragments of slag and burnt clay.