Experimental Ironmaking Once More: Combining Theory and Fi Nd Material

DISCUSSION

Q Fig. 1 Smelting according to Evenstad’s manuscript from year 1782, printed in 1790. The 60 pages long manuscript covers every aspect of contemporary ironmaking. Drawing H. Ness Experimental ironmaking once more: combining theory and fi nd material

In the previous paper the Th e large number of sites, the older Borre in Vestfold, however not re- excellent quality of Norwegian ones very large with several furnac- ported in the same thorough way as Roman Age and Medieval es and up to 50 tons of slag, express those fi rst mentioned. blooms was documented. The that the process was reproducible. number of sites tells us that the In this paper I will suggest that the Smelting by the Evenstad technique method was reproducible, and processes were complex, but not by following his cookbook- like ad- that the blacksmith could make complicated. vice from the year 1782 is another, fi nished items from the blooms but successful story. By 14C-dating with a high output, very Th e quality is remarkable. A basic we know that it was introduced diff erent from many reported question is how the sintering of as- around year 1400, probably less experiments. sumed primary small fragments to than 50 years aft er the Black Death. large lumps of solid, slag-free iron Pictures, redrawn from experimen- Q Arne ESPELUND was achieved. Th e carbon control tal smelting are shown in Fig. 1. (NO) can be attributed to human eff orts while the sintering must be a pure- As shown, it is a step-wise process. ly physical phenomenon. A driving Wood is at fi rst transformed into Introduction force for the sintering process is the charcoal in the conical-shaped fur- In the previous paper A new look at lowering of the specifi c surface area, nace. It should be noted that a pri- experimental ironmaking (Espelund, as is also known for snow crystals, mary over-reduction to a carburised in EuroREA 2009) I presented the growing in time when left undis- metal is followed by an oxidising re- excellent quality of Roman age and turbed. In the year 1794 the Swedish fi ning by adding roasted ore at a late Medieval Norwegian blooms, most metallurgist Carl Rinman clearly ex- stage, when only a little remains in of which are now kept and preserved pressed that iron oxide must remain the furnace. Th is is a modern inter- in Norwegian archaeological muse- in the slag in order to fl ux it and also pretation of his very practical advice. ums. Because they are, so to speak, combine förena, the ductile grains of Th e product of the smelting is fi nal- slag-free, no smithing in order to re- iron (Rinman 1794:17). He inserted ly removed with a shovel. Some of it move slag was necessary. Th eir den- a rod of iron into the tuyere of his represents a bloom, while the rest is sities are around 6, in contrast to furnace in order to test the quality of thrown onto the slag heap. Th e proc- 7.8, which indicate that they are in the slag, a measure also known from ess therefore is direct only in the a primary state and were not infl u- the Catalan process. sense that it was performed in the enced by heavy smithing. Th is view- same furnace. It consists of reduc- point is also supported by the gen- Because of their quality, the out- tion followed by refi ning. eral lack of hearths and anvil stones put upon smithing of the primary at Roman and Medieval bloomery bloom into fi nished objects is as- As an example of a misleading ex- sites in Norway. Th ey number some sumed to represent some 90 %. periment, I would like to present 500 and 10,000 respectively. Th e Accordingly, the blooms indicate the following case study: Th e Nor- lack of entrained slag is expressed a process that is diff erent from the wegian TV NRK I on January 2nd chemically by the very low contents standard experiments of modern 2009 showed smelting which was of silicon (Si) and manganese (Mn) bloomery ironmakers with their claimed to be successful. Th e tap- (representing oxides) found by con- yield of some 15 to 20%. In particu- ping of slag went well while the iron ventional analysis of the metal, and lar I refer to reported experiments produced appeared to be brittle. their low densities due to empty mi- in a medieval type of furnace in However, when asked for details I cropores. Th ey appear to be evenly Sweden and Denmark. In Norway was told that bones had been added distributed and present no problem such experiments have been carried to roasted bog iron ore. Bones con- upon smithing. out at Kittilsbu in Oppland and at tain calcium oxide (CaO) and also

4 euroREA 7/2010 Experimental iron making once more DISCUSSION

phosphorus( P), detrimental for the Th e removal, or exclusion, of car- quality of the iron produced. By re- bon in γ-iron can be expressed by placing iron oxide (FeO) in the slag, the equation: this (CaO) can increase the output as well as the fl uidity of the slag, C + FeO ↔ Fe + CO however, at the cost of metal quality. Th e archaeologist T. Gansum has Th is is a chemical equilibrium, also added bones to his charge – in where C expresses carbon dissolved his case deliberately (Gansum & et in the metal. A high content of FeO al 2002). However, analysis of hun- corresponds to a low value for dis- dreds of ancient slag samples have solved carbon. only rarely revealed contents of CaO above 1% (Espelund 2009a). Th e idea should therefore be to main- tain iron oxide (FeO) together with In this paper I shall go into theoret- the metal produced, if possible “at ical details, show some analyses, ex- birth”, in the statum nascendi. How- press something about the ore sit- ever, at the operating temperature of Q Fig. 2 The Fe-C-diagram expressing stability ranges uation and present my ideas about a bloomery furnace – some 1100 °C for the different phases. Abscissa % C, ordinate tem- how a two-step process can be car- – FeO is solid. Its chemical function perature. A vertical line is drawn at % C = 0.2, a mate- ried out. can be taken over by FeO dissolved in rial of the best possible quality for general purposes in a liquid iron silicate, so-called fayalite the smithy. This iron must be produced as a solid in the Basics of the iron-carbon slag. Th is slag is fl uid at normal op- γ-state around 1100 °C, which is the operating tempera- erating temperatures and is therefore ture of a bloomery furnace -oxygen system separated from the solid iron. Th e re- Th e iron-carbon diagram is, from a maining question now is how we can metallurgist’s point of view, indis- secure a proper slag. Over-reduction pensable when the formation and to a carburized metal without the the quality of iron and steel are dis- benefi cial eff ect of FeO dissolved in a cussed. It is shown in Fig. 2. silicate slag seems to be the problem.

Noteworthy is the line Acm, rep- From fi nds to theory resenting saturation with iron car- and back again bide Fe3C (or elemental carbon C), which reaches 2% at 1100 °C. Among archaeometallurgists in Carbon atoms are accommodated Norway the character of the slag in the face-centred crystalline cu- now is regarded as very signifi cant. bic structure of γ-Fe, also named Th e Roman type is vesicular, present austenite. A metal with 2 % carbon also as large lumps, expressing that it cannot be forged. solidifi ed in a slag pit. Organic material (wood) in the pit was imbed- Th e Baur-Glässner diagram, shown ded in liquid slag and later decayed. in Fig. 3 expresses both the relations Th e Medieval slag is mostly dense between gas mixtures, such as those and fl at with a smooth upper and a created in a bloomery furnace, and rough lower side, expressing tapping Q Fig. 3 The Baur-Glässner diagram for a combined pres- the phases α- and γ-Fe, carbon and sideways at ground level, while the sure of CO + CO2 = 0.4 atm., balance nitrogen N2. This is wustite (FeO). “Evenstad” type is porous, and that the typical gas created by combustion of charcoal in

gas evolution, due to an addition of a bloomery furnace. Abscissa ratio CO/(CO + CO2), ordinate Th e lower curve shows the gas com- ore at a late stage, took place during temperature. Dotted lines in the γ-field express the carbon position in chemical equilibrium with solidifi cation. Because of its specif- content at different temperatures and gas mixtures. solid carbon. Notice that the content ic character, 14C-dating of charcoal of carbon monoxide (CO) rises dras- from a slag heap will rarely be a sur- tically between 650 and 800 °C. As prise to the experienced archaeo- seen, no reduction of FeO to metallic metallurgist. Th e three categories Fe can take place below a tempera- normally represent periods of max- ture of about 660 °C, which is a red- imum production around year 200, glow heat. Th e right curve expresses 1100 and 1600 AD, respectively. the gas ratio for coexistence of the lowest oxide FeO and the metal. Th is Wherever slag is present in large metal is practically carbon-free, well amounts, it expresses a large and suc- suited for forging, while the values cessful production. Irrespective of for γ-iron in contact with carbon, site, period, type of furnace, process Q Fig. 4 Slag from Nystaul NIKU No. shown by means of the left curve, are and ore, the “end” slag contains 63 99/31. Sample No. 2 in the table 1. high – the same as those in the Fe-C- ± 3% (FeO and MnO) and 23 ± 3 % White metallic iron, rounded parti- diagram. Th is value is much too high SiO2. Th e individual values for FeO cles FeO, and dark matrix fayalite- for malleable iron. and MnO show a wide scatter, but like, probably glass

7/2010 euroREA 5 DISCUSSION Experimental iron making once more

an intermediate oxide. It is possible of ore, a partial reduction of Fe2O3 that 4% SiO2 may be the minimum to FeO, followed by reaction of FeO required for the desired slag forma- and SiO2 to fayalite Fe2SiO4 ? tion (Espelund 2009b), while 20 % or more of SiO2 in the ore will give Th e slag marked S2 was also studied fayalite slag and little or no metal. by SEM-analysis, resulting in a picture very much like that in Fig. 4. Q Fig. 8 An ore spear (malmspyd) At a few sites two kinds of slag have Th e site is regarded as a place where used for finding bog ore, marked BG been found, as shown in the table 1 only step 1 of a two-step process (Berger Gjermundsen) (1702 - 1755). (R represents the ratio [%FeO + was carried out. He lived in Rendalen, county of % MnO] / % SiO2 with molar val- Hedmark, an area where bloomery ues. For fayalite the value of R = 2. A site near Granhaugsetra in Sollia, ironmaking still was performed at Th e slags marked 1, 3 and 6 therefore municipality of Stor-Elvdal, county that time. Notice that the spear is contain fayalite with excess FeO.) Hedmark similar to the site Røst is smooth, suitable for finding ore now being studied. of the kind shown in Fig. 9. Photo: Th e slag No. 2 has been analysed P. Fredriksen with SEM. Fig. 4 shows back-scat- How was step 1 tered electrons. performed? they can safely be added because of A thorough search for bloomery Fig. 4 is dominated by the lowest similarity as slag components. Sam- sites by Mr. Asbjørn Ryen in the oxide of iron FeO, while specks of ples with such analyses have been community Os, county of Hed- metal also occur. Th e formation of gathered from Roman age, Medieval mark, led to a site at Røst in Dals- this mixture must take place in an and newer sites in Norway, Medieval bygda, shown in Fig. 5. environment where the combus- sites from Iceland, a Roman age site tion of carbon (charcoal) delivers in Austria, as well as Catalan sites in Th e only visible structures are two heat, while a chemical equilibrium the Pyrenees (Espelund 2009a). rings of stones, reminiscent of places between the solid or liquid slag for an open fi re, and also two pieces with carbon or the equivalent CO- For chemical analysis, representa- of burnt clay. However, in front of rich gas must be avoided. tive of pieces that express solidifi - Mr. Skare we found roasted ore and cation from a homogeneous liquid, charcoal, and behind the junipers to In Fig. 6 the standard medieval should be selected. A few samples the left a few pieces of slag. 14C-dat- bloomery furnace has been modifi ed ing AD 1165-1225 (charcoal from by making holes for admission of air of slag containing around 10% SiO2 seem to express a story diff erent birch). No charcoal pit or large slag in the shaft , which is only half full. from the one above. Was it a fail- heap was found in the vicinity. ure? Or was it produced on purpose It is conceivable that the wanted ra- as a semi-product? It seems to be A sample of roasted ore was acquired tios of CO/(CO + CO2) near 0.75 – necessary to cast a look on the ore by magnetic separation. Chemical see Fig. 3 – can be created in this and to follow it all the way to a good analyses of this ore and of three piec- way. Such secondary combustion metal. In Fig. 1 in the previous ar- es of slag are given in table 2. Notice by admission of air is essential for ticle bog iron ore containing alter- the similarity for S 1-3 with slags nos. the puddling process for the refi n- natively 4, 8 and 12 % SiO2 was fol- 2, 4 and 7 in table 1. Could it be that ing of pig iron invented by Cort in lowed through four steps via FeO as the slags represent a pre-treatment England. Th e heat of combustion

123456 7 M S 1S 2S 3 MnO 1.71 3.84 2.8 3.53 FeO 81.4 85.1 81.25

Fe2O3 93.92 Al 0 1.31 2.58 2.19 2.47 Niku99/31 End slag * Medieval Møsstrond Niku 99/31* Hellegryte slag Fagstad Slag D** Fagstad Slag E** E Fagstad SEM ** AMI 2 *** Lyckebo AMI 1 *** Lyckebo 2 3

MnO 1.280 0.250 12.610 1.990 1.969 0.250 0.110 SiO2 2.82 9.52 8.05 9.97 FeO 64.460 90.50 54.10 81.010 76.101 61.940 76.400 CaO 0.17 1.64 1.02 1.40 MgO 0.01 0.39 0.23 0.41 SiO 24.080 6.330 21.610 8.510 10.481 25.500 9.200 2 K2O 0.08 0.38 0.33 0.29

Al2O3 4.910 0.630 6.790 2.000 2.556 5.900 2.800 P2O5 0.058 0.088 0.045 0.061

P2O5 0.322 0.138 2.410 4.186 6.278 0.190 0.230 TiO2 0.01 0.05 0.06 0.08 Sum of 10 oxides 98.062 98.268 100.120 99.696 100.000 *99.360 *92.490 BaO 0.17 0.23 0.20 0.21 Sum 100.3 100.1 100.0 99.7 FeO+MnO 65.740 - 66.710 - - 62.190 - R 25.5 7.46 9.11 7.09 R 2.280 11.950 2.580 8.130 - 2.050 6.930 Q Table 2 Chemical analyses of

Q Table 1 Analyses from sites with two kinds of slag. Notice the differences for % SiO2 and R, both roasted ore M and of three pieces of clearly at two markedly different levels. As seen, the parallel values for FeO are not so distinctly slag S1 – S3 from the site Røst different as the contents of silica SiO2

6 euroREA 7/2010 Experimental iron making once more DISCUSSION

CO + CO (N ) of carbon to CO2 is almost 3 times ore of a satisfactory quality. Th is pat- has changed to ombrotrophic, i.e. 2 2 larger than the primary combustion tern was also expressed in 2007 by aff ected by rainwater only. Accord- Fe2O3 (SiO2) to CO, so that the charcoal required Swedish ironmakers from Tranemo ingly there is no longer any contact C in this case is much less than in a near Göteborg, who went 300 km between dissolving minerals in the true reduction furnace. north to fi nd good ore. I claim that bottom and oxygen in the air. O (N ) good ore of the kind described by 2 2 Th e problem of tapping or of some Ole Evenstad in the year 1782 has It can be mentioned that the con- CO + ½ O = CO other method for retrieving slags of been consumed and not renewed. He sumption of this type of bog ore dur- 2 2 the type shown in Fig. 4 from the fur- used a metallic stick in order to lo- ing the Roman Iron Age in Trønde- nace remains to be solved. In Fig. 6 cate good ore, mostly found as lumps lag has been assessed to at least some tapping of such a mixture is indicat- in a layer some 20 cm below the top- 25 000 tons (Espelund 2009c). C + ½ O2 = CO ed. But it seems unlikely that liquid, soil. Such a stick is shown in Fig. 8. FeO-rich slag can pass through a lay- However, visible ore formation er of incandescent charcoal. However Th e development of bog ore for- takes place also nowadays in ochre if the temperature can be maintained, mation can be explained in the fol- wells (raudveller). Ground water O2(N2) FeO + Fe2SiO4 such tapping is conceivable aft er the lowing way: Th e Scandinavian Pe- at some 4 °C containing dissolved charcoal has been consumed. Th e ninsula was largely covered by ice divalent ions of iron Fe++ when Q Fig. 6 present author is eager to discuss such until about 10 000 years ago. Th e emerging and meeting oxygen in Modified a process with experienced smelters. slow movement of this ice removed the air will precipitate ochre as a medieval the soil and the vegetation, leaving running process. Chemically it can furnace, in One question is whether the pro- in places moraine material. Barren be represented by FeOOH. In con- which air is posed practice was performed only rocks on exposed surfaces dominat- trast to the sloping mires, a well is admitted in the in Norway. A micrograph of slag ed. A humus layer was gradually built created below the bog. When dried shaft, allowing from England, presented without up during the millennia to follow. this ochre is extremely fi ne-grained secondary any further description, shows ex- Where sloping mires (bakkemyrer) and fl uff y, suitable as a pigment. It combustion of actly the same pattern as interme- were created the conditions for ore appears to have been avoided as a CO to CO2. Such diate slag in Norway, such as shown formation were favourable, provided raw material for iron production a process might in Fig. 4 above. a reasonable thickness and precipi- in the past, due either to fi neness, transform the tation. Th e formation took place un- too much silica or phosphorus (in- silica in iron ore derground over decades or centuries troduced by bacteria which thrive into fayalite. The ore situation and resulted in lump-size limonite. in such an environment). I have Concept A. E. Attempts to make experiments and Th e botanists name such mires min- found some twenty such ore wells produce bloomery iron have been erotrophic. A rare fi nd of such ore is in my country. In Fig. 10 the ochre met with diffi culties in fi nding a documented in Soknedalen, county well at Namlåmyra in the commu- good ore. Broadly speaking, in my of Sør-Trøndelag, shown in Fig. 9. nity Midtre Gauldal, county of Sør- country we now hardly fi nd bog iron Trøndelag is shown. Th e analysis of a lump of this ore af- ter roasting at 650 °C ran as follows: An analysis of the fresh precipitate aft er roasting at 600 °C gave the following result: Sørstuv MnO 0.45 FeO Namlå

Fe2O3 84.07 Fe2O3 76.86

Al2O3 1.63 MnO 3.36

SiO2 12.62 SiO2 15.9

CaO 0.36 Al2O3 0.01

Q Fig. 5 The archaeologist K. Skare at MgO 0.07 P2O5 0.055 the site Røst K2O 0.09 CaO 3.1

P2O5 0.001 Mg0 0.01

TiO2 0.04 Ba0 0.2

BaO 0.12 Ti02 0.01

Sum 99.4 K20 0.01 R 5.0 Sum 99.5 R 3.8 Th e value for R = 5.0 (%FeO+%MnO) / %SiO2 (molar) expresses that it is Th e temperature of the emerging acceptable as a raw material. Notice water was found to be + 4 °C, while Q Fig. 7 Micrograph of slag from the low content of phosphorus. pH was 6.84. England (Bayley & al., 2008). Notice FeO as a dominant phase. A nearly During recent centuries or mil- Th is type of current ochre forma- similar sample is shown by Pleiner lennia these mires have grown in tion does not appear to occur on (2000), as in picture 2 on Plate XXIII thickness so that their character the Continent. On the other hand,

7/2010 euroREA 7 DISCUSSION Experimental iron making once more

his remarkable book on contempo- Rinman, C., 1794: Korrt underrättelse om rary ironmaking. His method is sat- sättet att smälta sjö och myrmalmer uti isfactory from a modern metallurgi- blästerugnar. (Short explanation on how cal point of view. It would not have to smelt lake and bog ores in bloomery been revealed by the fi nds only. furnaces). Translation by L. J. Hukkinen. Jernkontoret H 66 Stockholm 1997 It is conceivable that a pre-treatment Summary of ore, rendering a semi-product Noch einmal experimentelle composed of mainly iron oxide and Eisenherstellung – Zur Kombination iron silicate, would be an excellent von Th eorie und Fundmaterial raw material to feed into a normal Im vorigen Artikel wurde die gute Quali- reduction furnace as an alternative tät der norwegischen Luppen aus der Rö- to a material composed of iron oxide merzeit und dem Mittelalter beschrieben. and silica. Th e presence of fayalite Die Anzahl der Fundstellen zeigt, dass should warrant a metal that met the die Herstellung reproduzierbar war, und Q Fig. 9 Limonite ore in a ditch near Sørstuvollen in demands of the blacksmith. auch, dass der Schmied Fertigprodukte Soknedal, community of Midtre Gauldal, county Sør- mit hoher Ausbeute herstellen konnte, Trøndelag. On the picture the local informant Arnt O. Th e author is looking forward to a sehr unterschiedlich von vielen beschrie- Solberg and Magne Rokne. Due to road building, drain- practical cooperation with black- benen Experimenten. age of water has taken place at this spot. The overall im- smiths who are willing to incorpo- Das Fe-C-Diagramm zeigt, dass ein pression is that of a sloping mire (bakkemyr) rate a serious interpretation of fi nds Kohlenstoff -Gehalt von 2 % bei angenom- material, as well as modern metallur- mener Arbeitstemperatur im Rennofen gical thinking in their experiments. zu erwarten ist. Der erwünschte niedrige It is a pity that facilities for analysis Inhalt kann nur durch Schlackenkontrolle of ores, slag and metal hardly can be erzeugt werden. Im Evenstad-Verfahren said to be available as a service to the wird das Primär-Eisen durch einen Zusatz many experimenters. von extra Eisenerz gefrischt. Im Mittel- alter-Verfahren schlägt der Verfasser vor, dass die Herstellung zwei Stufen umfasst. Acknowledgement In der ersten Stufe wurde ein Gemisch von FeO und Fayalit hergestellt. Das Zwischen- Th e author is grateful to the analyt- produkt wurde dann reduziert, wobei ical laboratories of ERAMET Nor- man gutes Eisen und Schlacke erhielt. Der way in Porsgrunn and Sauda and Verfasser schlägt zuletzt vor, wie das Zwi- to colleagues in my department for schenprodukt hergestellt werden konnte. good help. Nouvelles expérimentations sur le travail du fer: études théoriques et Q Fig. 10. The ore well at Namlåmyra in Singsås county Bibliography matières premières of Sør-Trøndelag Bayley, J., Crossley, D., Ponting, M L’article précédent a documenté l’excellente 2008: Metals and Metalworking. HMS qualité des loups Norvégien de l’époque Occasional Publication No. 6 Romane et Médiévale. Le nombre de sites large lumps of ore used formerly Espelund, A., 2004: Jernet i Vest- trouvés prouvent la reproductibilité de la for house construction in Denmark Telemark (Iron in Western Telemark). méthode, ainsi que l’habilité du forgeron à and Lower Saxonia in Germany are Trondheim faire des produits avec un grand débit des not available in Norway. Espelund A., 2009a: Sant, halvsant og loups, ce qui est très diff érent des récents usant om jernframstilling. (Truth, half résultats d’expériences. Conclusion true and untrue about ironmaking) Le diagramme Fe-C montre que 2% C doit Primitive tider 1. Oslo être envisagé pour un acier produit à la In this and the previous paper, an Espelund, A., 2009b: A new look at température de fonctionnement du four. emphasis is put on a presentation experimental ironmaking. euroREA 6, L’acier à basse teneur en carbone désiré of representative fi nds and also on EXARC, 50-52. peut seulement être obtenu par control Espelund, A., 2009c: paper, presented in a connection to metallurgical theo- du laitier. Dans le procédé d’Evenstad, Bradford in November. le métal primaire carburé est raffi né par ry. One should reckon that it took Evenstad O. (1790) 1968: A treatise some 1000 years for ironmaking to ajout de minerai en fi n de processus. Pour on iron ore as found in the bogs and le procédé Médiéval, l’auteur suggère une reach Scandinavia from a begin- swamps of Norway and the process of première étape consistant à la réduction ning in the Near East around year turning it into iron and steel. (Engl. partielle en FeO et fayalite, suivie d’une 1500 BC. It is very naïve to believe translation by Niels L. Jensen). Bull. réduction de ce semi produit. De petites that the simplest, possible modern Hist.Met. 2 (1968) quantités de laitier trouvées contenant Gansum, T., Hansen, H-J., 2002: Fra jern experiment should lead to a metal 8-10% de SiO2 semblent représenter of comparative quality to that cre- til stål. (From iron to steel). Midgard ce semi produit. L’auteur suggère une ated by the early ironmakers during historisk senter, Borre. méthode pratique pour la première étape – Hjärthner-Holdar, E., 1993: Järnets och the Middle Ages. celle d’une réduction partielle. järnmetallurgins introduktion i Sverige. (Th e introduction of iron in Sweden). Q Arne Espelund is Professor Th e manuscript of Ole Evenstad Aun 16. Uppsala. tells about a complex, discontinuous Pleiner R. 2000: Iron in Archaeology. Th e Emeritus at the Department of practice, which is revealed to us in European Bloomery Smelters. Praha Materials Science and Engineering -

8 euroREA 7/2010