A New Look at Experimental Ironmaking (PDF)

ITEMS / AME The 2006 symposium on Ancient Metallurgical Experiments A new look at experimental ironmaking

The author discusses current between. Th e fourth side was “open”, mal size bloomery furnace (while in experiments in reconstructing with a tap hole for slag and a hole for large blast furnaces with cold blast, ancient ironmaking and sug- the mouthpiece of the bellows. Th e temperatures above 1300 °C are gests that serious experiments height is assumed to have been some achieved). Th ere is hardly any dan- and more complex methods, 80 cm. Th e slag was tapped at inter- ger of running the bloomery furnace which take proper slag forma- vals while the solid bloom was taken at a too high temperature, which tion into account, are needed. up with some sort of tongs. Th e Ro- might lead to liquid cast iron. man age furnace were diff erent, and Q Arne ESPELUND so was the type of furnace used be- Th e reduction of the iron ore, as- tween AD 1400-1800. However, most sumed to consist of Fe O , is stepwise (NO) 2 3 experiments aim at a re-enacting of via Fe3O4 and FeO to metallic iron the Medieval process based upon Fe. Th e equilibrium gas pressure at Th e modern blacksmith can buy his the type described above, which was the fi nal reduction step is represent-

or her iron and steel as a cheap com- common in Europe. ed by a gas ratio of pCO/pCO2 near 1 : modity in any hardware store. Still it 1, while the gas emerging from the is there: the dream of making one’s Most of such experiments are hap- surface of glowing charcoal hardly

own iron, well supported by the fact penings, performed on special oc- contains any CO2. When a normal that ancient bloomery iron had a casions at archaeological open air ore is applied, one can only expect very good reputation. In my country museums. Many are not reported a reduction of the two elements iron Norway it was said by blacksmiths at all, but is is claimed by some that Fe and phosphorus P, the latter dis- that it fl owed on the anvil. Th ey good iron is obtained. However, solved in to the metallic phase. El- would gladly trade fi nished objects, for this author they oft en miss two ements such as silicon, aluminium such as horse shoes, in exchange points: a comparison with ancient, and manganese are retained as ox- for an ancient bloom (known from well defi ned bloomery iron, and ides and end up in the slag. Fyresdal about 1930). also some words about the output, relative to the amount of ore and Th e chemical analysis of bloomery Th e remains of bloomery furnaces the slag produced. In the following slags from sites expressing a large tell you that the construction was I shall present a few points, which and successful smelting show about

simple. A medieval furnace in my should be relevant for experiment- 25 w % SiO2, while the individual country consisted of a shaft made ers. Admittedly, my information values for FeO and MnO present a from clay with an inner diameter of stems mainly from my home coun- wide scatter. However, as these two about 30 cm. It was supported on try Norway. However, with some oxides behave in the same way in a three sides with vertical slabs of stone 10 000 slag heaps representing three slag, it is permissible to add them. with an insulating layer of sand in or more specifi c bloomery methods, Th is gives values around 62 w% for spread in time and space, it is fair to the sum. Th e solidifi ed slag shows say that my country has very rich above all the presence of the com-

material for such archaeo-metallur- pound fayalite (Fe, Mn)2SiO2, for gical studies, which should involve which the value FeO+MnO is about archaeologists, metallurgists as well 70%. As will be shown, a high as blacksmiths. I am eager to reach amount of FeO in the slag is re- experimenters with my ideas. quired to secure fl uidity and a low content of carbon in the metal, re- Th e combustion in a bloomery fur- quired for good, malleable iron. nace can be represented by the equa- tion. Th e morphology of Norwegian bloomery slags can broadly speaking

2 C + O2 + 4 N2 = 2 CO + 4 N2 be divide into three types: large pieces with cavities expressing the decay Th is expression represents the fact of organic material (from the slag pit that air consists of about four parts furnaces of the Roman iron Age), fl at of nitrogen and one part of oxygen, and dense pieces from the Medieval Q Fig. 1 Column I initial weight 1000 g, showing in the and that the nitrogen just passes side-tapped furnaces and porous slag lower part a total of some 10 – 20 % of oxides other than through the furnace. Enthalpy cal- from the most recent period (with culations express that the maximum the Evenstad type of furnace: dug-in Fe2O3. Column II weight loss when Fe2O3 is reduced to FeO. Column III metallic iron is produced, with a fur- (adiabatic) temperature of this com- with no tapping of slag). ther weight loss. In column IV the iron is consolidated. bustion taking place in a shaft fur- Column V shows the remaining slag. M/S expresses the nace with an excess of charcoal lies It follows from this that only ores at around 1400 °C (Espelund 1996, with a sum of FeO + MnO well above ratio metal/slag. SiO2 “steals” FeO and has a marked ef- fect on output. However, a minimum of silica is required 2009). Due to heat loss it is fair to the value for a good slag can give any for slag formation, necessary as a solvent for FeO. With assume that 1200 °C is the maximal metal. In fi g. 1 the output of iron operating temperature for a nor- and the amount of slag produced are some 25% SiO2 in the ore no metal can be expected!

50 euroREA 6/2009 The 2006 symposium on Ancient Metallurgical Experiments ITEMS / AME shown for three samples of roasted shows ferritic iron with very small ful smelting can only be explained ore with alternative values 4, 8 and amounts of slag at the boundaries by assuming a slag control, repre-

12% of SiO2 and a minor amount of where three grains meet. sented by the chemical equilibrium MnO and Al2O3. Initial weight of ore 1000 g. Resulting slag assumed to be of the fayalite type. FeO + C ← → Fe + CO

Th is expresses that the ore provid- When liquid slag is present, near sat- ed for the experimental smelting in uration with wustite FeO, the carbon Eindhoven, the Netherlands, 2006, content of the corresponding metal containing some 73% Fe2O3 and will be negligible (Espelund 2008b). 23% SiO2 could hardly give any iron Th e successful smelting is represent- in a bloomery process. If this iron ed by a replacement of carbon con- ore upon smelting resulted in iron, trol by slag control. In the English- there is reason to question the qual- Q Fig. 4. Micrograph of a sample tak- speaking world it was said: “Make ity of this metal. Notice that an anal- en from the bloom from Skeibrok. your slag and the metal looks aft er it- ysis is useless unless it includes sili- Etching in 2% nital. It shows ferrit- self.” Th e slag must be a silicate slag, ca, if possible also MnO and Al2O3, ic iron with very small amounts of commonly named fayalite slag. which at times are signifi cant. slag at the boundaries where three grains meet. It now only remains to achieve prop- In Norway there are many museum er slag formation, in addition to a re- pieces of bloomery iron. Th ey are of Th e analysis of the bloom from Li duction from oxides to metallic iron. two types: shows the following values: In a properly run bloomery furnace, using a rich ore, it is likely that iron Rather fl at, weighing some 2 - 12 kg, Element % by weight oxide is reduced all the way to met- found in particular in the county of al, without allowing enough FeO for Telemark, and rounded, weighing Si 0.09 the reaction with silica in the ore some 18 kg, found mainly in Trøn- Mn 0.16 and a refi ning. delag. Both types have been cut C 0.43 Q Fig. 3 Blooms from Skeibrok halfway through with an axe, clear- S 0.011 In the Evenstad process, fully de- ly in the red-hot state, probably in scribed in contemporary writing a-b (weight order to test the quality. P 0.015 from the year 1782, the primary 4.02 kg) and product is a carbon-rich metal. Th is Li c-d (weight Th is author has had a chance to study Microscopy revealed large diff er- is refi ned in the furnace by an addi- 2.83 kg). Shown some such blooms. Th eir densities ences in composition, in the range tion of 4 l of ore towards the end of a are also the cut have been measured by immersion 0.7 - 0.1 %C. Th e value for car- smelting cycle, lasting some 5 hours surfaces. From in water, found to lie between 5 and bon therefore must be regarded as (Evenstad 1968). Th e other process Martens (1979). 6.5, in contrast to 7.8 for pure iron, a compromise. Still the content of described in contemporary writing: expressing that the blooms have an slag is very low. the Catalan process from the Pyr- evenly distributed porosity. Th is also tells that the blooms are primary, a Th e two blooms described are rep- fact that is supported by the external resentative for Medieval ironmak- dimensions. ing in Norway. No secondary slag removal was required, therefore the Drillings from the bloom from output during the processing to fi n- Skeibrok were analysed chemically, ished objects must have been above giving the following 90 %. Th is is in great contrast to the results of experimental ironmaking in Sweden and Denmark (Englund Element % by weight 2002, Lyngstrøm 2008) Si 0.01 Mn < 0.01 Th e Fe-C diagram expresses the equilibrium phase relations as a C < 0.01 function of temperature. As seen, S 0.008 the iron created in a bloomery furnace in the presence of solid car- Q Fig. 2 Histograms according to weight and thickness of P 0.006 bon, runs at a temperature of about ancient blooms, found in Norway, now mostly in muse- 1100 °C should be of the austenite ums. F denotes fellujern, B blåsterjern, in agreement with Th e fi gures tell us that this is an ex- type, containing some 2% carbon. two kinds mentioned with prices in the medieval law book tremely pure and slag-free iron (Si On the anvil this iron is hard and Jonsbok from AD 1281. Values from Martens (1979). Blåster- and Mn refl ect slag inclusions). Th is brittle, very diff erent from the mal- jern is a primary metal, while fellujarn has been compact- is confi rmed by metalloghraphy, as leable iron, desired by blacksmiths. ed by moderate smithing. Blooms marked T, added in the shown in fi g. 4. A micrograph of a A majority of the primary blooms present figure, have been found mainly in Trøndelag and sample taken from the bloom from depicted in fi g. 2 contain little or no stem from the Roman Iron Age type of furnace. As the lat- Skeibrok. Etching in 2% nital. It carbon. Th e result of such success- ter blooms are rounded, thickness has no meaning.

6/2009 euroREA 51 ITEMS / AME The 2006 symposium on Ancient Metallurgical Experiments

enees – Basque country, uses a shaft Preliminary studies of fi nd mate- Norway and the process of turning it into which is divided vertically into two rial indicate that the Medieval type iron and steel. Translation from Danish parts: In the one part normal reduc- of furnace could be used for both of chapters 1 – 9 by Niels L. Jensen. Bull. tion takes place while fi ne grained ore process No. 1 and 2. Hist. Met. 2. London Hjärthner-Holdar, E., 1993: Järnets och is supplied continuously to the slag järnmetallurgins introduktion i Sverige. bath from the other part. Both proc- Concluding remarks Aun 16. Uppsala. esses are very specifi c and at the same Lyngstrøm, H., 2008: Dansk Jern. Det Kgl. time express sound metallurgy, which It is most likely that blacksmiths Nordiske Oldskrift selskab. København clearly is based upon the principle of will continue their attempts to Martens, I., 1979: Blåsterjern og slag control. It follows that processes make good iron in a simple, di- fellujern – en lite påaktet funngruppe. such as the two mentioned in con- rect process by charging the fur- Oldsaksamlingen. Oslo. temporary writing would never have nace with roasted ore and charcoal. been developed if iron could be made I hope that with this contribution I Summary in the simplest possible way. can convince them that an ore with Ein neuer Blick auf die experimentelle a high ratio of iron oxide to silica Herstellung von Eisen In a lecture held in 2001 I present- is a primary requirement. It is re- ed alternate ways to produce good markable that the good ironmakers Eine große Zahl experimentalarchäolo- iron by proper slag control (Espe- in Tranemo, Sweden went some 300 gischer Aktivitäten zur Eisenherstellung werden nicht publiziert, und wenn, lund 2008a). During recent years I km north to fetch their good ore in dann werden oft zwei Punkte vergessen: have come across many samples of Härjedalen (Espelund 2008b). Der Vergleich mit bekannten Funden slag containing 5 – 10% SiO2 while archäologischer Eisenobjekte, aber a normal “end” slag should con- It is likely that some iron will be cre- auch das quantitative Resultat des tain some 25%. Admittedly at most ated in experiments of this kind. Such Schmelzprozesses im Vergleich zur Menge places only slag of the latter kind has was the situation also for our forefa- des eingesetzten Erzes und der anfallenden been found. However, at sites such thers. Th e development of very spe- Schlacke. Eine gute Publikation sollte as Møsstrond in Telemark (Espelund cifi c and complex methods can only alle Schritte des Herstellungsprozesses 2004), Fagstad near Lillehammer be explained by a moderate success beschreiben: Die Verbrennung in einem and also near Uppsala (Hjärthner- in standard experiments of the kind Schmelzofen, die Eisenreduzierung und die chemische Analyse des gewonnenen Holdar 1993) slags of two kinds have seen in Eindhoven. Th e reproduc- Eisens als auch der Schlacke. Es ist in been reported (Espelund 2009). ibility of the high quality of iron and jedem Fall notwendig, die Ergebnisse also the output expressed by materi- der Experimente mit archäologischen It is therefore claimed that one al fi nds are remarkable. No modern Objekten zu vergleichen. Weiterhin ist es wide-spread way to make good bloomery ironmaker has presented wichtig, seriöse Experimente mit solchen iron involved a two-step process. results that can be compared with Methoden durchzuführen, die zur Bildung In the fi rst smelting the reduction the best results of our forefathers. entsprechender Schlacken führen, um Ergebnisse zu erzielen, die auch tatsächlich was moderate. Th e oxide Fe2O3 in the ore was reduced to FeO, which It now remains to make serious ex- mit den Produkten unserer Vorfahren reacted with silica present to form periments with a two-step proc- vergleichbar sind. fayalite according to ess, or some other method, which takes proper slag formation into Une nouvelle approche du travail expérimental du fer account. Perhaps a much reduced Les expériences liées au travail du métal 2 FeO + SiO2 = Fe2SiO4 ratio of charcoal to ore will give the right conditions for the pre- sont aujourd’hui nombreuses, mais n’ap- Th is transition is expressed by the reduction. I hope that some of the portent pas toujours leur contribution à la columns I –II in fi g. 1. During the arduous blacksmiths will try the réfl exion scientifi que à leur juste valeur. La plupart ont lieu lors d’événements particu- smelting step that followed there- proposed two-step process by sys- liers, et nombreuses sont celles qui, malgré fore fayalite would be present, thus tematic experiments! des résultats très encourageants, ne don- securing good iron. nent pas lieu à des publication de résultats Bibliography ou à des comparaisons avec des vestiges As the intermediate product was Englund, L.-E. 2002: Blästbruk. archéologiques. very valuable, it is not likely there Jernkontorets Bergshistoriska Skrift serie Pourtant, il est nécessaire pour une bonne will be much to fi nd. However, at a nr. 40. Stockholm. expérience de prendre en compte toutes les site visited in September this year Espelund, A. 1996: Archäo-Metallurgie, étapes du processus : la combustion dans (Røst in Dalsbygda, Os) , only rich von Norwegen aus betrachtet. Ferrum, le fourneau, la réduction du fer, l’analyse ore and a few pieces of slag contain- No. 68 Eisenbibliothek. Schaffh ausen chimique du métal obtenu ainsi que des Espelund, A. 2004: Jernet i Vest-Telemark ing some 9 % SiO were found. It ap- scories. Il est indispensable également de 2 (with summaries in English). Trondheim. confronter ces résultats avec les observa- pears that only the semi product was Espelund, A. 2008a: Th e mechanism of produced at this place and smelted tions de terrain et les vestiges conservés... bloomery iron production. Th e coming of Ce n’est que par là que le travail expéri- to good iron somewhere else. iron in Eurasia Uppsala. mental des forgerons d’aujourd’hui per- Espelund, A. 2008b: Bondejern i Norge. mettra de retrouver les techniques d’hier. It is tempting to compare a two- (New edition). Trondheim Espelund, A. 2009 (in preparation): step process with a cow chewing of Q Arne Espelund is Professor the cud! We know of this behaviour Bloomery ironmaking and its secrets. Trondheim Emeritus at the Department of because we have seen it, not from Evenstad, O., 1968: A treatise on iron Materials Science and Engineering - fi nds of the intermediate product. ore, as found in the bogs and swamps in NTNU in Trondheim, Norway.

52 euroREA 6/2009