How Dense Are You : Recording bloom density from experimental iron smelting Darrell Markewitz, with contributions by Neil Peterson

‘How Dense Are You?’ :
Recording bloom density from experimental iron smelting.

INTRODUCTION

Charting the density of a sample group of 30 iron blooms, created over a span of roughly 15 years during experiments aimed at understanding iron smelting furnaces, and comparing these to a number of artifact samples.

The focus of the bloomery iron smelting undertaken by the team from the Dark Ages Re-Creation Company, here at Wareham, Ontario, Canada, has been towards Northern European prototypes, primarily those from what has been alternately called the ‘Migration Era’ or ‘Dark Ages’; so ‘Post Roman to Early Medieval’ era (roughly 400 - 1100 AD). At the risk of being overly simplistic, those ancient furnaces were generally smaller diameter (25 – 50 cm interior), short to medium shafts ( estimated at 50 to 100 cm), bellows blown, and slag tapping. Over two decades of experimental investigations, I have personally undertaken or been directly involved with 90 bloomery iron smelts, with dozens more observed. Tests have included a number of extended series investigating Viking Age at Vinland (total 8), Viking Age Icelandic (total 12), Pictish (total 7) and Celtic Iron Age (total 4). (1)
It is important to remember that the initial work here was aimed at testing individual elements leading to an effective furnace design and establishing a working method, that together would lead to dependable production of iron blooms. Later, individual modern aspects would be replaced with more historic ones, in an attempt to reverse engineer back towards potential ancient systems and methods.

IRON SMELTS

Figure 1 : A fairly typical furnace used in this experimental series. (experiment # 90 - ‘Wind & Weathering’, October 2021)

The furnaces built and smelts undertaken have the following characteristics :
•   Furnaces are a ‘short shaft’ type, typically 25 - 30 cm interior diameter (ID), and in the range of 60 - 70 cm tall.
•   Primarily clay cobb construction, most using a standard mix of powdered clay / course sand / shredded horse manure (dry thirds by volume).
•   Slag management has been primarily ‘slag tapping’ (although some tests of slag contained and slag pit types have been undertaken).
•   Most typically insert style tuyeres, earliest of steel pipe, from 2005 prepared ceramic tubes, and since 2012 almost exclusively heavy copper. There have also been a number of experiments utilizing the ‘bellows plate / blow hole’ arrangement. (2)
•   A variety of ore types have been used, but from 2008 onwards mainly using a ‘bog ore analog’ consisting of red oxide powder (Fe2O3) mixed with whole wheat flour as a binder, with Fe concentrations in the range of 51 - 54 %.
•   Locally available commercial Maple or Oak charcoal as the fuel.
•   Final bloom extractions have both been bottom / side (via a prepared arch) or from the top, normally without breaking down the furnace structure.
•   A fairly standard method of pre-heating, adding charges of charcoal and ore, to final burning down before extraction is followed. (3)

    During the early years, particularly working with Lee Sauder and others at the annual Smeltfest workshops, individual smelts were in the 45 kg of ore range, which in turn resulted in blooms at 8 - 12 kg. Tests done here in Wareham (where accessing ore was a problem) more typically ran 25 - 30 kg ore, with resulting blooms averaging between 3 - 5 kg. It was considered true that if a smaller bloom could be produced, continuing to a larger one would primarily just require additional materials, time and effort. It should be noted that past experience has found that with larger ore volume smelts, overall yields and bloom densities have also been found to increase (at least before compaction effects).
    Another reason for keeping to smaller blooms was ease of handling during the secondary compaction of blooms down into working bars. There is a physical limit to how much impact force (largely untrained!) workers can apply with hand sledge hammers. Even historically, once blooms got much over about 5 - 8 kg, they were commonly cut up into smaller chunks to allow effective compaction into the finished iron bars. (4) There are functional limits at roughly 8 – 10 kg, imposed by human power limits; bellows operation to required air volume, raw hammering force, and the problems of subsequent re-heating. (Although larger blooms certainly can, and definitely were, produced.) (5)
    Sauder and Skip Williams had investigated the use of high volume air in similar sized bloomeries, clearly illustrating how this impacted by increasing both bloom density and overall yields (6), and this work had a large impact on the Wareham experiments.
Most smelts use an electric blower for air supply with which the expected yields of iron from ore (in smelts as described above) is in the range of 20%. It is well understood that air delivery from high volume electric blowers will not accurately duplicate that from human operated bellows in a number of aspects. There have been a number of tests using various human powered bellows types (total 15). The most dependable bellows unit delivers in the range of 500 - 550 litres per minute (LpM). (7) Electric blowers are regularly used primarily because of the general difficulty to organize the significant labour pool required using a specifically Viking Age type, twin chamber bellows.

Figure 2 : Ken Cook using the second test ‘smelter’ bellows built in 2008, shown in use for experiment 42 (‘Vinland 3’, Nov. 2009)

This is considered a significant element to this report, as the there is no archaeological evidence for exact design and sizes for bellows specifically used for iron smelting during the reference period. (8) There are some general estimates that have been made on what air volumes may have been required (into the range of 350 LpM for a furnace at 25 cm ID). (9) The experience at Wareham has been that at those lower suggested volumes, iron is most certainly produced, but the blooms created show both low overall yields and a lacy consistency (which becomes problematic at the bloom to bar phase) (10)

BLOOMS

The cornerstone reference work, Radomir Pliener’s ‘Iron in Archaeology - the European Bloomery Smelters’ gives the following :

“About 90 blooms or parts of blooms had been cut through in order to observe the consistency of the metal. In 34% of the cases, the iron was unconsolidated, the block consisting more or less of sintered-together iron sponge with many isolated grains, metallic fibres and nodules embedded in a slag matrix. Some 30% of bloom finds were partly consolidated, and had evidently undergone primary reheating and were partly or wholly forged and shaped mostly into the loaf form, sometimes with flattened sides, but with their interiors still full of entrapped slag, so that further forgings would have been necessary to obtain a workable material for the manufacture of artifacts.” (Pleiner, 2000, pg 244)

Figure 3 : Disk cut cross section of a typical bloom measured in this study, half of 6.8 kg.
(experiment # 22 - ‘Redemption’, Nov. 2006) Note the relatively few internal voids exposed.

Generally, all the blooms produced here at Wareham are all 'finished' to the same level, which does make at least comparisons between them valid. Given the small working team, with normally only myself as having any significant blacksmithing experience, (11) we hammer work any bloom through only the one heat cycle available at extraction. This process involves first knocking off any clinging slag, then very rough compaction to force the loose exterior metal into the core. As temperatures are rapidly dropping while all this is happening, there is rarely very much actual hammer welding taking place, more the collapsing of larger voids and forcing out still fluid slag. Although different individuals may take part as strikers between individual smelts, the hammers used are identical, so the variation in overall force applied is likely to be relatively consistent.
Starting in 2012, there was a 30 ton hydraulic press available at Wareham (a modified log splitter). In some cases, after the initial hand hammering, blooms were rushed to the workshop and the mass was given several compressions, then if possible, cut, all via the press. Given the initial slag removal and compression is still undertaken by hand, and the distance from the smelting area to the press, blooms are at best ‘into the low orange’ visible heat range by the time they are subjected to the press. It is only the massive force available that allows much further modification to be possible. It is worth remembering that although an extracted bloom is ‘as hot as it ever is going to get’, this temperature is quickly dropping over this described sequence.

Figure 4 : Initial hammering of a freshly extracted bloom. (L – R) Ken Cook, Sam Fallezone, Darrell Markewitz. (experiment 42 - ‘Vinland 3’, Nov. 2009) Image by Vandy Simpson

    All this suggests that the blooms detailed below should generally conform to the ‘30% of finds’ described by Pleiner (above) as ‘partially consolidated’.
    Although it is most certainly true that making any evaluations of artifact blooms from published photographs is chancy at best, knowing how process has resulted in surface features after working / observing the creation of so many blooms at least gives some indications.
    There are three characteristics that might be considered important to assessing the overall effectiveness of an individual iron smelt, and the ‘quality’ of the iron bloom produced :

Yield : At it’s simplest, a measure of effort and raw materials input against metal output. It has become clear that there is a certain amount of ore required to create the working slag bowl system before any metal can functionally accumulate. (Those researchers who concentrate only on slag and it’s chemistry most often completely fail to understand this.) Early production of lacy iron film inside a slag matrix may indicate an iron reduction process, but does not in any way illustrate actual bloom production. Our own experience is that roughly 8 kg of ore (modified by ore iron concentration / silica content), within furnace diameters used in these tests, is required to create a truly functional slag bowl system. (12) As mentioned earlier, potential yield is greatly effected by total ore additions. Once bloom collection begins, additional ore continues to add more mass, so a doubling ore beyond that threshold not only will double the bloom weight, but also double the effective yield. Ancient iron makers could be expected to want the very best returns from their efforts (adjusted by the real problems of working massive blooms).

Density : Beyond compaction, how dense a given bloom may be will be is the result of two primary factors; the number and size of any internal voids, and how much slag remains trapped within the metallic mass. It must be remembered that the desired end product is not the bloom itself, but the working bar compacted from that bloom. Early experimenters had reported bloom to bar returns ranging from 80 % to as low as 40% of the starting ‘bloom’ weight. (13) Obviously, the more dense any bloom is, the less work will be required to weld closed existing voids and expel remaining slag. In practice, an ancient smith is most likely to assess density through appearance and experience in ‘heft’. It should be noted that one secondary effect of the segmented blooms common through the Viking Age is that the cut surfaces expose the consistency over the centre of a bloom.

‘Hardness’ (Carbon Content) : This is suggested to be essentially a modern concept. It always needs to be remembered that the end product desired by ancient makers would be primarily lower carbon, so therefore easier to forge, ‘soft’ iron. Although higher carbon contents (still less than 1%) are better for durable (hard) cutting edges, creation of ‘bloomery steels’ would have been accidental at best. (14) If nothing else, the common method of inserting hard ‘steel’ edges into larger tool bodies composed of softer iron should indicate this. It needs to be remembered that the only way an ancient smith could assess carbon content would be through experience, based on subtle colour changes to the metal and the ‘feel’ of differing materials under hammer strokes.

One major problem in this report is that finding actual recorded densities for artifact blooms has proved extremely difficult. Weights are certainly available, but at best only the roughest dimensions are given (typically just a maximum length and width), but as blooms are complex three dimensional shapes, these do not allow for any calculations. (15) In some general (private) discussions with both working archaeologists and scientifically trained individuals, the reasons that this might be the case become clear:

The simplest method of measuring volume is through water displacement, in fact the method undertaken for this report. The problem is that emersion in water is most certainly likely to promote even further corrosion of what are at the very least ‘uncommon’ artifacts. (16) One potential intervention (suggested by Tim Young) would be to place the artifact bloom inside some combination of both durable (blooms have ragged, often sharp, edges, depending on amount of compaction) yet flexible, wrapping. This however is certainly to result in at best ‘ball park’ estimates, as any voids or surface imperfections would not be included.

A variation on liquid displacement might prove possible by replacing water with some volatile liquid that would quickly evaporate, perhaps something like acetone (?), however this is more a thought exercise than a practical suggestion. A concern that introducing any chemical to the surface of the bloom would seriously impact on the accuracy of later detailed analysis. (For the liquid suggested for example, any remaining charcoal embedded would have potential carbon 14 dating rendered useless.)

A suggestion given by David Wentz would be an ‘inert gas displacement pyconometery system’. This equipment is certainly beyond the reach of almost any field archaeologist. (17)

Another suggestion (by several individuals) was the use of a 3-D scanner. Combined with suitable computer software, these systems should be able to easily produce at least an accurate estimate of the surface into volume. ‘Small scale / economy’ units are however fairly recent. I would expect increasingly universities and museums would have by now invested in such units. Balanced against this is the fact that the majority of artifact blooms were excavated well before the existence of the related technologies. (18)

SAMPLE ARTIFACTS

For comparison, three artifact blooms are reported by Arne Espelund in ‘The evidence and the secrets of ancient bloomery ironmaking in Norway’, under the chapter heading “Blooms from the Early Iron Age” (all are from Norway):

a) Lake Hartevatn, Bykle, county of Aust-Agder
Reference : #00/187 - University Museum of Antiquities
now at The Hovden Museum of Iron Production, Bykle
https://hovden.com/tellus/hovden-museum-of-iron-production/?lang=en
Date is not given
Dimensions not given
Weight of 22.54 kg
Density of 4.6 gm/cc

Figure 5 : From Espelund, 2013, pg 47

Looking at the image presented, the top surface of the bloom is shown, with what is most likely the tuyere side to the bottom of the image. The outside surface shows many irregular ‘spikes’, which certainly suggests this bloom was not compressed by hammering beyond removal of any clinging slag. The slight depression in the upper surface suggests the bloom was sitting quite high in the furnace, so that the upper part was exposed to erosion effects from the air blast hitting it.

b) Modum, County Buskerud
Reference : #AKS-61 - Museum of Cultural History, University of Oslo
now in a private collection ‘at Kjølstad’
Date is not given
Dimensions : thickness at 12.1 – 12.4 cm, diameter at 23.1 – 23.8 cm
Weight of 17.89 kg
Density of 5.36 gm/cc

Figure 6 : From Espelund, 2013, pg 48

Again, the orientation of the image is with top shown, tuyere likely to be to the bottom of the image. The upper surface is flatter, so likely a better position within the furnaces in terms of the effect of the air blast. The surface does show some flatter areas that might be the result of initial hammering, but still the overall shape is quite ‘lumpy’, which suggests minimal compaction at best.

c) Hira / Asket
Reference : #T-21 175 – NTNU University Science Museum, Trondheim
Date is given (via C-14) as 760 – 410 BC
Dimensions are not given, but estimated from scale as about 21 cm wide by 24 cm long
Weight of 17.2 kg
Density of 4.9 gm/cc

Figure 7 : From Espelund, 2013, pg 128 (described pg 48-49)

    This bloom is shown from the bottom side, and it obviously has a more bowl shaped central portion that trails away thinner to one side. The tuyere point is likely to to right of the image. Generally this bloom has a smoother surface, and there are several areas that appear to show flattening (seen on what was the bottom and to the right side)
    Although three measurements is an extremely small sample size, the average density of these artifact blooms is 4.95 gm/cc.

METHOD

    The first step was weighing each of the full blooms or combined pieces. For masses up to 6 kg, a simple platform digital scale was used, which provided measurements accurate to one gram. Over this amount, blooms were placed in a bucket and measured with a luggage type digital scale, accurate to 10 grams. (then the weight of the bucket subtracted).

Figure 8 : Two different scales used for bloom weight measurements.

As discussed, the simplest method for calculating volume was used, being water displacement. Given the wide range of physical sizes (from 1 to 15 kg), three different containers were used, selected for closest fit to the bloom being measured. These were filled to ‘just overflowing’ to determine their volume capacity, so 2090 / 8520 / 12375 millilitre. The largest container, used for the largest full blooms, was wider than high, which given the sampling method, is likely to have resulted in the lowest accuracy (used for only 2), with those measured in the smallest (used for 17) being the most accurate. Measurements of water volume were made using a graduated cylinder, adding 1000 ml amounts, or fractions, added again with bloom in place until just overflowing. These measurements were at best + / - 5 ml of accuracy. Warm water was used (bath temperature), both because of the low temperature of the working space, and also to speed evaporation of remaining moisture on bloom surfaces after immersion.

Figure 9 : Measuring volume, a bloom placed in the mid sized container, with the graduated cylinder.

The densities were calculated, and are recorded at two decimal places, although given the simple nature of the process used, one place would be more reasonable.

SAMPLED BLOOMS

Figure 10 : The blooms measured in this study, including full blooms (rear) and those that had been cut into smaller pieces.

For this report, a total of 30 blooms were measured, either as existing full blooms (total 13) or the pieces composing one single, or still remaining from, an original bloom (total 17).

Figure 11 : A 3/4 view of a full bloom, the top surface towards the camera.
The original tuyere placement to the right. (experiment # 9 - ‘OABA’, May 2005)

The # 9 – May 2005 bloom provides a nice comparison to the artifact blooms illustrated above. This bloom had been only somewhat compacted, and had one corner (upper left in the image) broken off for further working. The top surface shows the same concave surface caused by air blast erosion seen in the bloom from Lake Hartevatn (figure 5). Considered in cross section, this bloom also has a similar shape to the one from Hira / Asket; a central more bowl shaped core, with a thinner trailing edge opposite the tuyere side. The surface of this bloom remains quite irregular, with a granular texture seen in other blooms that used the same ore type (granular hematite, a commercial blasting grit). Although less than half the total weight of the Hira / Asket artifact (7 kg against 17.2 kg) the actual calculated density is slightly higher, at 5.3 gm/cc against the reported 4.9 gm/cc.
Three samples were from experiments where it was not possible to extract the hot bloom, so the metallic mass was later broken free of the entire slag bowl via cold hammering. This has resulted in a different appearance to the bloom, with a rough ‘spikey’ surface to the top and sides, and slag containing charcoal still attached to the bottom portions. The effect of this on calculated density is most clearly seen on Iceland 6, which despite one of the larger weights (at 8.8 kg) has one of the lowest density numbers (at 4.5 gm/cc)

Figure 12 : Bottom area of a bloom hammered free from surrounding slag bowl when cold, showing the matrix of slag with embedded charcoal still remaining. (experiment # 60 - 'Icelandic 6', June 2015)

Almost all the blooms had been compacted to some degree, either by hand sledge hammers (total 16), or using the hydraulic press (total 11).
The sectioned blooms had been hot cut either using an axe driven by sledge hammers (total 5) or via a cutting blade on a hydraulic press (total 9). The alternative was cutting cold, using a ‘zip’ disk on an angle grinder (total 3 – most clearly seen in figure 3 above). Most commonly cutting rendered full blooms into roughly 1 – 2 kg pieces and was done for the purely functional reasons of both the on-hand equipment sizes, and simple hand hammering ability, leading into the next stage of the overall iron making process, rendering raw blooms into finished working bars.

Figure 13 : Bloom sectioned using an axe driven by hand sledge hammers. Note that one piece had been further compacted. (experiment # 38 - ‘Icelandic 3’, Oct. 2008)

P3A

Figure 14 : Bloom first roughly hand hammered, then sectioned using the hydraulic press, with two pieces (bottom) further compacted via the press. (experiment # 88 - ‘Icelandic P3-A’, June 2021)

The DATA

    Two blooms have been removed from the final averages, reducing the data set to 28 samples. An earlier error in labelling had resulted in two different objects marked ‘Oct. 2016’. Both appear as complete blooms, (a combined total of 6 kg) but more detailed notes describing that smelt indicate a much larger bloom mass (total at 8.8 kg), that photographs clearly show as almost completely sectioned into two haves. The calculated density for sample marked Oct. 2016 A (a low value at 4.3 gm/cc) both falls well short of the overall range, and most significantly is quite different from that marked Oct. 2016 B (a high value at 7.1 gm/cc). This is certainly unlikely if these were half portions of the same bloom.
    The 15.9 kg bloom created from experiment # 87 (65 for 65) has been removed from the calculation of average bloom weight (only). This is because of the large total amount of ore used was over double the typical, at 65 kg (over a wide mixture of ore types). (19) It is included in the other averages however.
    The full chart of weight and volume to density, with data on ore, air and bloom type is available as Appendix One. The data for individual smelt / blooms is seen arranged in order of increasing density, with Oct. 2016 A, Oct. 2016 B (red) and 65 for 65 (blue) separated. The three artifact blooms are indicted (green) Those values considered estimates shown in italics. The blooms illustrated here indicated (tan).

    Preliminary results : Bloom Average Density of 5.98 gm/cc (28 samples) against an average weight of 4.6 kg and average volume of 850 cc. Over this sample set, the average ore to bloom yield is 26%.

Looking at Plotted Data

    In the following discussion, the scatter plot diagrams were prepared by Neil Peterson from the larger data set.

Chart A : Blooms by Mass and Density

There is a clear relationship shown between decreasing bloom weight to increasing bloom density. As suggested in the earlier practical discussion, this would be expected as an effect of hammer compression, where the force applied remains relatively constant (sledge hammer blows) but the resistance to compaction will decrease as blooms get smaller. An unexpected observation in the charted data is what appears to be three groupings (illustrated by coloured ovals).

Chart B : Compaction Effects

The effect of the amount of compression is also seen when the after extraction treatment of individual blooms is charted against density. There were only two blooms in the sample set which did not experience any consolation process (both broken free of encased slag after cold) and both of these are found on the extreme lower end of density numbers. Both those blooms worked with hand sledge hammers (total 14), and those worked with both hand hammers plus the use of the hydraulic press (total 10), show considerable range in density numbers. Those subjected to the second compression effect of the press (even if at lower heat ranges) show a clear drift to the higher calculated densities. This is pretty much the pattern that would be expected.

In an attempt to further understand the variation in densities, the mass against density data was plotted against a number of other elements considered possible influencers.

Chart C : Ore Iron Content

It is known that iron content of the ore will directly impact on the overall production of blooms, but in this series, only ore of suitable qualities had been used. Very loosely, there were two main groupings of ores used for the measured blooms. The most frequent ore used (after Spring 2008) had been several slightly different mixtures of the team’s red iron oxide (Fe2O3) ‘bog ore analog’, with iron concentrations averaging 52 % (ranging from 51 to 54 % over 14 samples). There is a second grouping of ores of various types, all of which are recorded at 65 % iron content (6 samples). One outlier used granular hematite at 68 % (Fe3O4), with a number of samples not having recorded iron content.
From this chart, it would appear that there is a very slight indication that the ores with the higher iron content had in fact resulted in blooms in the lower density side of the range. This is unexpected, enough so that other factors are suspected as the cause. Typically the smelts using the higher iron content ores were also those using larger ore quantities, which in turn created the larger individual blooms, so those least to experience additional compaction through hammering. This might also suggest the largest blooms in this report are most similar to the even larger artifact samples.

Chart D : Effect of Air Volumes

It has been mentioned how Sauder & Williams had demonstrated that higher air volumes, even with all other factors remaining constant, will greatly increase not only furnace yields, but also the density of blooms created. Their ‘ideal’ had been given as 1.2 to 1.5 litres each minute per square cm of cross sectional area at tuyere level (L/min/cm2).(Sauder & Williams 2002) The influence of meeting and working with the pair, starting in Fall of 2002, is clearly seen in the use of high volume air in almost all of this team’s experimental work.
When the surveyed bloom densities are charted against the rough average air volumes used to create those blooms, this effect is the most clearly seen of any of the potential effecting elements. Those smelts with air volumes below 1.2 L/min/cm2 are grouped towards the lower densities, those in the range of 1.2 to 1.4 L/min/cm2 show grouped to the mid range, and those with air above 1.4 L/min/cm2 tend to the higher densities.

Chart E : Experience

    The blooms measured ranged in creation date from March 2005 (experiment # 8) to October 2021 (experiment # 90). A general assumption was, that over time, the working team would gain experience, so that blooms should increase in ‘quality’. As already detailed, employing similar furnaces, ores and air amounts, the two aspects that might be expected to improve would be overall yields and effective densities. Looking at the plotted densities against creation dates, there is a bit of a weight of the earliest smelts to the low end and the later efforts to the higher end. The spread within each of the 5 year groupings still remains quite wide however. This spreading effect may have as much to do with the nature of the individual and series investigations progressing into more and more specialized furnace designs as time progressed.

OVERALL CONCLUSIONS

    The density of solid wrought iron bar is given as 7.75 gm/cc (20)  Obviously this is a fully compacted state, so depending on what losses could be expected rendering blooms into bars needs to be reflected in any comparison.
    The sample size of artifact blooms with reported density is extremely small (Although it should be noted that a number of museums were contacted with a request for measurements from their collections, no further information proved available.) At least superficially, the artifact blooms reported here show a range in density of 4.6 to 5.36 gm/cc, an average of 4.95 g/cc. The blooms measured in this report range from 4.45 gm/cc to 7.28 gm/cc, an average of 5.98 gm/cc (over 28 samples).
Comparing total mass to calculated density, it is clear that the largest blooms measured are to the lower density values. In this it might be most realistic to compare the artifact samples to the 7 blooms ranging from 6.97 to 9.14 kg, although bearing in mind that the artifacts are all double these weights. This group ranges in density from 4.41 to 5.52 gm/cc, an average of 4.94 gm/cc, basically identical to the artifact samples.

Looking at the various data plots, two elements show clear influence on bloom density :

The first is the compaction process upon extraction; basically the more effective this was, the more internal voids could be compressed and contained slag expelled. There is a slight difference seen between those blooms just hand hammer worked at welding temperature, and those that that were then further subjected to a secondary compression at an orange heat via the hydraulic press. As the methods and equipment used is relativity consistent, so will be the amount of raw impact force applied. Any working blacksmith would tell you that impact effect decreases as raw mass increases. So it comes as no surprise that smaller blooms would show more significant compaction effects, thus higher ‘finished’ densities, even after a single working heat cycle.

It had already been demonstrated, and certainly re-enforced by the experiments over the last 20 years by this team, that increased air volumes produced not only higher overall yields (so bloom weights) but also an increase in density. This effect is also seen when density is charted against air volumes used. The average (over 16 data points) is 1.37 L/min/cm2 (an average total of 755 L/min inside furnaces averaging 27 internal diameter). There is a major consideration outstanding over how closely air production from modern, high volume electric blowers matches to air produced by various historic human powered bellows systems. Work on this aspect continues, specifically focused on possible Viking Age prototype bellows. (21)

Taken all together, the information in this report hopefully should serve as a rebuttal to a disparaging statement made at the 2008 'Iron in Thy' (Denmark) symposium :

'Modern experimenters are unable to produce iron blooms of similar quality to those made by the ancient iron masters'

Images :

The three images from ‘The evidence and the secrets of ancient bloomery ironmaking in Norway’ are by Arne Espelund, and have been scanned directly from a print copy of his book (thus the lower quality) These used without permission.

The charts used were prepared by Neil Peterson.

Unless specifically credited, other images by Darrell Markewitz

Notes & Citations

 1) Full documentation of individual experiments found at : Experimental Iron Smelting :

2) An illustration of this method, first demonstrated to me by Micheal Nissen, can be found in the report on Icelandic 3 - Work Dynamic Test, October 2008 :

3) Markewitz, D., 2012 : “If you don’t get any iron ...” - Towards an Effective Method for Small Iron Smelting Furnaces. EXARC Journal, Issue 2012-1

4) Description of ‘split blooms’ can be found in Pliener, R., 2000, ‘Iron in Archaeology - the European Bloomery Smelters’ , pgs 238 – 243
(now available as a downloadable PDF)

5) The reference time frame was chosen to fall between use of ‘Roman’ large sized (passive air draw / slag pit) and Medieval water powered equipments (huge bellows and mechanical trip hammers). Using a furnace similar to the ones detailed here, working in March 2008 Lee Sauder and Michael McCarthy were able to create a massive 80 kg bloom by adding to the top of the same mass over a total of four individual smelts.

6) Sauder, L., & Williams, S., 2002, A Practical Treatise on the Smelting and Smithing of Bloomery Iron, Historical Metallurgy 36 (2), (available as pdf)

7) Markewitz, D., 2022, Wind & Weathering : air delivery and long term erosion - Part 2 : Wind,

8) In fact, there are no known artifact bellows even as blacksmithing equipment from the period. (Pliener, R., 2006, ‘Iron in Archaeology – Early European Blacksmiths’, pg 131) Even as historic illustrations, there are is very little direct evidence – and none for actual iron smelting operations.

9) Rehder, J.E., 2000, ‘The Mastery and Uses of Fire in Antiquity’.

10) See the discussion of the use of a Norse style bellows and the imact on bloom formation in Markewitz, D., 2014, ‘An Iron Smelt in Vinland : Converting Archaeological Evidence into Practical Method’ in"Can These Bones Come to Life?", Cramer (editor) 2014.
(an expanded version is available as a PDF)

11) Experimental partner Neil Peterson has been slowly increasing his skill and knowledge of working with iron blooms. Neil has been coming up and undertaking afternoon workshop sessions were he has been taking bloom pieces in the 500 - 800 gm range and compacting these down into finished working bars. It should be noted that this still is different than sledge hammer work - ‘striking’ is it's own separate skill set (and one that I also have rarely done!) On rare occasions various working blacksmiths have joined in and contributed their skills during initial compaction.

12) This figure of ‘8 kg ore to establish effective bloom creation’ is considered quite important. Far too many archaeologist experimenters undertake smelts using even smaller amounts, then centre their evaluations on the waste slag created, not on the production of a viable mass of iron.

13) “It must be borne in mind that before the iron in any bloom could be transformed into usable artifacts, losses were unavoidable in course of reheating and forging, reducing the amount of metal so some 40 – 80 % of the original weight “ (Pleiner, 2000, pg 245)
Peter Crew, one of the first researchers to undertake repeated and extensive experimental smelts, had initially reported bloom to bar losses in the range of 50%. Crew, P. & Salter, C., 1991, ‘Comparative Data from Iron Smelting and Smithing Experiments’, in Material Archaeology XXVI
(available as a PDF)

The work here at Wareham has so far concentrated on smaller pieces of these blooms, typically 500 – 1000 gms, from a very wide distribution of starting ore types, using the smelting methods described here. In the hands of a skilled blacksmith (40 years experience), the average returns have been ranged from 60 to 80% return of bloom into working bars. It needs to be noted that this work is being undertaken using a coal forge, and with use of an air hammer in the later stages. Also that effectively forging blooms requires its own set of skills. (Tests are still ongoing against a potential future report)

14) This distortion in perception is especially seen in the recent interest in the North American blacksmithing community, were the initial exploration of creation and use of bloomery iron became dominated by blade makers. The question of ‘How can I make ore into iron?’ very quickly became ‘How can I make metal with these specific alloy characteristics to use for knives’.

15) The best example here would again be Pleiner, 2000. Although dozens of blooms are reported by weight and diameter, and many illustrated via scaled drawings, there are no densities listed.

16) Pleiner gives this number as known from Europe :

“To date (2000), more that 500 blooms or their separated parts, coming from 90 archaeological sites have been identified.” (Pleiner, 2000, pg 243).

Given that this reference spans from 8th century BC (‘Late Bronze Age’) to the 14th century (Medieval), this is still not a large number of objects.

17) David suggested the ‘AccuPyc II’ as one such available system. (At a purchase price of about $18,000 CDN!)

18) The effective start of this technology, the CAT scan, dates to the early 1970’s, and even as large hospital use equipment, did not become more wide spread before the early 1980’s. Archaeologist were quick to see the huge value in scanning certain materials, especially mummified remains.
Laser scanners did not become wide spread in engineering / industrial application until into the early 1990’s, (partially because of computing limitations). Those were large view ‘field’ units (now commonly used for survey work at archaeological sites)
(Trying to determine when small scale / desk top type units for object scanning were first commercially produced proved difficult. Units of suitable size are now easily available, and computers powerful enough to run the data also typical. Current price range appears to be roughly $1000 CDN)

19) ‘65 for 65’ was a mounted as a special celebration to mark my 65th birthday. Individual ore types were added roughly in the order in which these had been employed over the previous 20 years of experimental iron smelting, the majority in 6.5 kg total amounts. The total ore amount of 65 kg was 30 % larger than the next largest smelt, and over double that typically used. The bloom created, at 15.9 kg, is 40% larger than the next largest measured here (experiment # 52 - ‘Celtic’, August 2012, at 9.1 kg).

20) https://www.engineeringtoolbox.com/metal-alloys-densities-d_50.html

21) There are two different lines of investigation implied :

Since artifact blooms are clearly large, does this indicate a different overall smelting sequence was followed historically?

Or that there were (unknown) bellows equipments used that actually produced higher air volumes?

This compounded by a possible difference between the ‘pulsing’ air delivery from certain bellows designs, opposed to the constant blast from electric blowers. The most recent experimental test was experiment # 90 - ‘Wind & Weathering', October 2021