Archaeology, Experience and Experiment :

Further tests of Icelandic iron smelting furnaces

Darrell Markewitz, with contributions by Neil Peterson and Rey Cogswell


In 2021, a re-evaluation based on the available descriptions of the Viking Age bloomery iron smelting site at Háls in Iceland was made by a team of independent researchers from Ontario, Canada. This new assessment was largely focused on what had been learned from eight previous experimental tests, plus the considerable smelting experience gained since this specific project was started in 2007. Two additional smelt events were then undertaken, one in June and one in September of 2021. This paper will detail this process, report the results, and illustrate how archaeological evidence was interpreted through practised skills into functioning experiments.

Author's Note :

This is an expanded version of this paper, from the original draft as prepared and submitted for inclusion in the upcoming volume Can These Bones Come to Life? : Insights from Reconstruction, Reenactment and Re-creation - Volume 3, M. Cramer & K. Mondschein (editors).
The final published version is likely to be considerably shorter, with fewer illustrations in black and white, than is seen here. This version  includes all the full colour illustrations, plus a number of detailed 'mosaic grid' and other photographs only found here.
Regular readers will find elements taken from a number of other reports related to Phase Three, found on this web site and on the blog. The various references have links in the Notes and References sections at the end of the text.

Experimental Background

Starting in 2007, a small group of independent researchers undertook an experimental investigation guided by the archaeology of a Viking Age bloomery iron smelting site located at Háls in Iceland. Descriptions of this site were presented by Kevin P. Smith in his paper Ore, Fire, Hammer, Sickle: Iron Production in Viking Age and Early Medieval Iceland (Smith 2005). The team, drawn from members of the Dark Ages Re-Creation Company from Ontario, Canada, was lead by the primary author, and had a long professional and personal relationship with Smith.

This series started with the preparation of a concept overview, attempting to match a theoretical furnace design proposed by Smith into a functional system. (Markewitz, 2008) The planned experimental work would physically test how various elements would work individually and then finally into a functional whole, of that proposed furnace layout. There were a total of eight furnace builds and full smelt cycles, that can loosely be broken down into two phases :

In 2021, it was decided to return to work on the Háls system, as a number of major questions remained unanswered. An important factor would be re-evaluating the archaeological evidence, in light of significantly more practical experience with small scale direct process bloomery furnaces that had been acquired since this series started (over 60 additional smelts).

The intentions for Phase Three of the overall Háls Icelandic experimental series:

Re-Evaluating Háls

Háls is located in Mid-Western Iceland.(Smith, 2005) The number of furnace bases and large size of the associated slag piles all suggest an 'industrial' operation, with many firings over an extended period of time. Use date can be estimated from a volcanic ash layer within the turf blocks used in furnace construction which was deposited in the early 870's, Smith suggests smelting activity dated to the 870's to mid 950's, with two separate 'campaigns' of use. (Smith, 2005) The original excavations exposed a number of furnace bases, four with slag bowls from their last use still in place. Additionally the excavation process exposed a cross section along the edge of one furnace. (1)


Figure1: “ Stacked furnace bases (Ia - VII) within a silt and turf enclosure (Feature 1) excavated in 2000 at the Viking Age iron production complex (Háls 5-6) at the site of Háls, Kollslækur, Borgarbyggð, western Iceland."

    " Slag rims (dark brown) and slag bases (light brown), fire-damaged rocks (dark gray) and rocks without evident fire damage (light gray) are shown in the plan view at bottom with hache-marks identifying areas of charcoal staining and charcoal concentrations.

    " Profile drawing of the western wall of the excavation unit, at top, shows from base to surface: undisturbed, pre-occupation sediments (yellow); silt from rebuilt enclosures surrounding the furnace bases (beige); charcoal and slag lenses (black); slanting turf blocks from furnace stack construction (light beige) with couplets of the "landnám tephra (877±1 AD) shown inside the cut turf blocks; cross section through the edge of furnace III (gray); sediments from post-iron-production agricultural activities (ca. 1000-1250 AD) at the site (brown); and the rooted modern surface soil (green).”
Text and image used with permission of Kevin P. Smith.

    The following elements can be established from Smith’s original 2005 report :

    Most unfortunately, further archaeological investigations of this site were not possible. Importantly to this discussion, chemical analysis of the ore used, the slag remaining, or the clay recovered, could not be made.

Matching Clay

    The major question about Háls is why the distinctive furnace construction (turf cone possibly supporting a thin clay liner) was used. Access to clay, a standard building material for bloomery furnaces, is problematic for Iceland, where true clays are both uncommon, and where available, limited in volume. Key understanding why this construction method was employed are the characteristics of the clay that was used. As has already been stated, no detailed analysis of the actual clay fragments recovered was available. This also included any specifics about possible additions of secondary materials (temper) to modify the base characteristics to be more suitable for use in furnace construction (possible sand or organics).

    Importantly, a good sized clay bank had been located within a ‘reasonable’ travel distance (about 15 km, considered an ’overnight walking trip’) by Michelle Hayeur-Smith.(2) A sample of that clay was made available, but it is unknown if this material does in fact match the clay fragments recovered at Háls. This sample was analyzed by team member Marcus Burnham, a professional geologist. 

Result type Average Mixture
SiO2 47.61 51.63
TiO2 2.24 0.03
Al2O3 13.44 13.81
Cr2O3 0.02 0
Fe2O3 13.14 13.1
MnO 0.21 0
MgO 5.63 5.81
CaO 9.1 9.08
BaO 0.01 0
Na2O 2.08 1.36
K2O 0.8 0
P2O5 0.3 0
SO3 -0.07 0
Other 0 0.57
LOI (volatiles) 5.23 4.58
Sum Of Conc. 99.74 99.98

Table A : Analysis of Icelandic clay sample.

    Based on recommendations by Burham, the following components were purchased from a commercial pottery supplier to create about 45 kg dry weight of a simulated ‘Icelandic’ clay :

Bentonite 22.7 kg
Mullite Grog 2.3 kg
Talc (Amal C-98) 6.8 kg
Wollastonite 7.7 kg
Black Iron Oxide 5.0 kg

Table B : Commercial Components for simulated clay.

    The dry clay components were blended using a mechanical cement mixer. Small batches with differing additives were hydrated to plastic, pressed into 4.5 cm diameter by 10 cm long cylinders, then dried and heated to at least approximate furnace firing temperatures (1070 C)in a propane forge. (3)

The furnace built for these experiments used approximately 19.5 kg (dry) of the ‘Icelandic’ clay. If the original workers at Háls were able to utilize the clay bank located by Hayeur-Smith, shortage of supply should not actually have been a problem, as unless clay was immediately close to hand, the Viking Age gathering method would certainly have been by ‘the wagon load’.

Based on the firing tests, and considerable previous experience, the dry thirds of clay / sand / dry manure mix was used for the furnace construction. Two points should be noted :

  1. It remains unknown what, if any, tempering additives were employed at Háls.

  2. The sand used was Ontario, granite based (72 % silica), rather than the Icelandic, basalt based (45-52 % silica) type. (Geology Science, 2018) It is unknown if this would have a significant effect.

The sand used was dug from a local natural deposit. This was screened using a 6 mm grid to remove stones, individual particles varied considerably (inclusion of limestone pebbles). The horse manure was gathered from a neighbour’s farm, selecting well dried ‘pucks’ that had been aging for a considerable time. These were shredded by rubbing between the hands. The proportionsof dry clay / sand / manure were estimated by eye, then hand mixed to blend uniformly. Water was added to create the desired plasticity, attempting to keep the batch uniform and to a consistency known to be suitable based on past building experience. The material was prepared in two separate batches, so some variation in moisture content would be expected between batches. A very approximate measurement was made of volumes used, with roughly 10 litres each of dry clay, sand and manure, mixed with about 10 litres of water, per batch.

Timber Framing?

    The diagonal stacking of turf exposed at Háls certainly suggested an alternative to a self supporting thick clay furnace, instead using a main cone shaped structure built of turf with only a relatively thin fire proofing layer of clay. Smith had further proposed that this structural cone was surrounded by a timber box, with the space between filled with earth to create a flat upper working surface. (Smith, 2005, p. 192)


Figure 2 : The 2007 interpretation of the furnace construction, modified for improved function,
based on Smith's original concept)

    Our previous test of a full scale, timber framed and earth packed build (October 2016) suggested there was no advantage to this build. The tools used for adding both ore and charcoal almost through necessity require long handles to protect workers from the extreme heat at the top of the furnace. In practice, it proved no problem to simply stand outside the raised box when adding materials to the furnace, something that is required every two or three minutes.   
    One possible reason for a flat upper surface would be if undertaking a top extraction, working down into the shaft while standing on the earthen platform around the furnace. Past experience with top extraction methods has clearly demonstrated how physically punishing this method is for the workers involved. Even when standing beside a typical 60 – 70 cm tall furnace shaft, where the legs are partially protected by the walls, the upper body, and significantly the face, is exposed to the full heat of the roughly 1100 + C interior. A worker attempting to stand on top of the proposed earth platform would have the entire body exposed to this extreme, even if the face would be further removed. Also, any of the iron tools required for extraction (chisels, hooks, tongs) would need to be excessively long to reach down that extra distance to the bottom of the furnace shaft.

    In addition to these practical considerations, one critical factor that wouldhave been expected to show larger traces in the archaeology is the raw volume of earth that would be required to create such a level working platform. With a typical 65 – 70 cm shaft height, and a turf cone at the suggested 200 cm base diameter, a (very) rough estimate is that some 1.6 cubic metres of earth fill would be required. As wooden framing collapsed over time, it would be expected that this material would be most thickly deposited at the outside edges of the framing, certainly not the situation recorded in the excavation diagram as indicated above. Having noted this however, it also important to remember the overall site at Hals remained a working farm – for centuries, so alterations due to plowing could certainly be the explanation.
    Most importantly, use of a top extraction leaves a complete circular slag bowl, more of an ‘offset bagel’ shape, rather than the broken 'C' shapes exposed at Háls. (Markewitz, 2006-2021)

    Although suitable timber and quantity of earth had been gathered in preparation for the Phase 3 furnace construction, in the end it was decided to test a build using only the sod cone structure, without any additional framing.

    As a quantity of cut grass sods (see discussion below) would be needed, the first location where these where cut was a 200 by 275 cm area where the furnace was to be located, stripped to a depth of about 10 cm. In anticipation of much later examination of the furnace through a process of annual aging, a clean sand base was created by simply filling the space where this sod was removed. One important aspect here is that no additional framing was required to establish the clear straight line edges from were the sod strips were removed. This is considered reflective of the roughly straight edges exposed at Háls, which were found to be composed of a layer of deposited material extending downwards roughly 5 cm at the edges. (Smith 2005, page 204 / see figure 1)

Building with Sod

    In this experimental series overall, a clear distinction is being made between Icelandic turf and Ontario sod. In both cases, what is being referred to is a strip cut from the uppermost natural growth layer, to a depth where the root system of the vegetation has penetrated and able to hold a thickness of the underlaying dirt in place. Smith indicated that the individual turf strips exposed at Háls had an average thickness of 10 cm. (Smith, pers. comms) There are significant differences between the natural environments of Iceland and Central Ontario, and this clearly has an impact on the qualities of both the soils and vegetation. Grass sod is used intentionally here to refer to the form of ground cover existing at the work location at Wareham, which is not expected to be an identical match for the turf structure existing at Háls in Iceland. (4)
    In this construction, the sods used were loosely of two different vegetation types, as cut from various locations around the property at Wareham, all to a rough thickness of 10 cm. First was material composed primarily of what had originally been (some 30 years ago!) been commercially available ‘grass sod’ strips, placed over a thin layer of topsoil, in turn over fill composed of a mix of sand, clay soil, and (many!) stones. This lawn grass had never been maintained, so also included clover and random Ontario weed plants. Strips of this material would form the interior three layers of the structure. The outer two layers were strips composed of primarily Quack Grass (Elymus repens), which has a much different root size and density (larger roots, but fewer of them).

    In interpreting the archaeology at Háls into the test furnace used for Phase Three, the overall construction was primarily taken from furnace III, while details of the front tuyere and extraction portions mainly from furnace VII.


Figure 3 : Structure from the cross section of furnace III, reconstructed into a possible working furnace.
(Roughly placed on a 10 cm grid.)

    To get some concept of the possible original size and details for furnace III, the scaled drawing above was prepared. The slumping into the shaft is seen straightened to vertical, and the lower silt layer corrected to horizontal. The shaft diameter isset to 28 cm, consistent to both the rough dimensions of existing slag bowls Háls 5 / 6 / 7. The lengths of individual turf layers were then mappedon to this framework, attempting to match both length and tapering thickness as recorded in the excavation cross section. As seen in the original profile, the inner three layers sat on a higher earthen layer than the outer two did. The turf lines were then extended upwards, so the outside layer would remain roughly 10 cm thick at its upper margin. This suggested a total shaft height estimated at 65 cm, a height considered functional based on previous experiments. These measurements would guide the actual construction.
    The test furnace was placed roughly centred side to side and slightly to one end of the prepared sand work space to mimic the overall placement suggested by Smith originally, and echoing the placement of furnace I. The additional space, located to the ‘front’ of the furnace, would allow for some comparison to the ‘charcoal tongue’ feature recorded in the excavation plan.


Figure 4 : Plan view, showing the placement of the furnace shaft in the working area.

    As indicated in figure 4, an additional raised earth pad was created, this of dirt screened to remove stones (via a 1.2 cm grid) laid to a rough depth of 7 cm. This pad was square during the first part of the build, held to place with a frame of standard 2 x 4 lumber to ease construction. At that point the dirt pad was 94 cm north to south ( tuyere to rear) and 96 cm east west. Once the lower part of the clay liner had been built up, and the first two layers of diagonally stacked sods were placed, this frame was removed. Next the earth pad was cut to a circular shape.
    In keeping with the structure indicated by the profile at Háls, a total of five full layers of sod were laid. There needed to be an additional small block of sod placed inner most and basically vertical to the liner, to create the start of the desired diagonal lines. T The individual sod strips had been cut as rectangles, roughly 20 cm long by 50 – 75 long, and were placed with their top edge against the outside surface of the inner clay liner as that was built upwards. To generate the required conical shape, pieces were cut into various triangular shapes and lengths to fill the gaps. The resulting structure was slightly irregular, and had base measurements of roughly 70 cm to either side and a total 160 cm front (tuyere side) to back. The total amount of cut sods used was roughly 11 square metres, a volume of roughly 1.1 cubic metres.
    If the full timber frame and earth fill had been used, the frame would be constructed after the full sod cone was established, then the earth fill added last.

Building with Clay

    One aspect that had been tested through the Phase Two builds was a steady decrease in the thickness of clay liner walls. The construction sequence had been to stack up a number of layers grass sod to a height of roughly 30 cm, leaving a cylindrical hole in the centre. The sides of this hole would then be plastered with clay mix. This process was repeated at least twice, allowing for building upwards sufficiently to create the required shaft height (typically 60 cm). As the clay layers were reduced in thickness with individual builds, it became more and more difficult to get the clay to remain in place against the loose earth edges of the sod layers. The failure point was found to be between 4 to 3 cm thickness.
    One solution proposed to the ‘sticking to the sod’ problem was to reverse the order of the layering during construction. The clay walls would be built up against an internal form, then the sod laid against this liner as the furnace was built higher. Use of a form allowed control of both the internal diameter and overall shape. (5)

    During the Viking Age, there would be a number of ways of doing this :

    For the Phase Three furnace, a standard cylindrical sheet metal form was used. This was covered with two layers of newsprint as a separator. The form was 28 cm ID, and 40 cm tall, the top fitted with a handle to make removal easier.

    Observations of the weathering process of previously used furnaces have been proceeding on a casual basis at Wareham since the first smelting attempts were started here (since 2004). This includes (to date of writing) a group of eight furnaces, all exposed to the full range of summer rains and winter freezing, the majority for over a decade at this point. The general result has been a reduction in the remaining wall thickness from an average starting thickness of 7 cm to a current average of 3.7 cm. (Markewitz, 2021A). This certainly indicates that even if originally significantly thicker clay walls are built, the exterior portions, which are essentially ‘dried mud’, are washed away by rain over even a short time frame. Only the inner / lower portions, subjected to the full heat of the furnace, can be expected to be completely sintered into a ceramic. The durability of this ceramic will be dependent both on the nature of the clay body used, and what additional tempering ingredients where in the mix.
    The combination of experience in effective building and uncertainty of what might have been the original wall thickness at Háls lead to a decision to undertake the Phase Three construction using a 4.5 cm thickness. In practice, this was controlled by using a ‘two fingers width’ measure as the liner was built up.
    From the larger batch of mixed clay, individual lumps of material where pulled, averaging between ‘orange to grapefruit’ size and pressed into rough brick shapes, typically about 10 x 15 cm by the desired 4.5 cm thickness. Each was added to the developing cylinder, with care taken to blend new bricks into the edges of those already in place, while at the same time maintaining a fairly consistent wall thickness.
    The first batch of mixed clay proved enough to build the liner to 35 cm height. Once the clay was constructed to this level, provision for the front extraction arch (and possible slag tapping) had to be made. Working from the Háls excavation diagram and the one photograph available, a set of individual stones were selected that most closely resembled the sizes and shapes of those indicated. These stones were primarily gniess (basalt is not available in Central Ontario), with fairly flat top and bottom surfaces which made arranging them much easier. One piece of basalt (gathered earlier from Nova Scotia, some 1700 km distant) was on hand, a larger slab that was positioned to frame a lower open arch space. This served as the support for the sod layers which would be placed above to cover the front wall of the furnace. This stone would also serve as a support for the tuyere, to be added later. It should be noted that all these stones were placed outside the clay liner, and so were not expected to exposed to much (if any) heat damage effects. This arrangement would also not allow for any slag from the developing slag bowl to adhere to their surfaces.


Figure 5 : First clay build completed to 35 cm, stones placed for the extraction arch,
with two layers of sod stacked to support.

    Once these stones were placed, the first two layers of supporting sod were positioned. The metal form used to control the interior diameter was then pulled up and removed. Any major flaws in blending over the clay surface revealed were repaired, then the interior was filled with a dry mixture of roughly 50 / 50 sand and wood ash. This fill would both support the clay structure, but also serves help pull moisture out of the damp walls. The metal form was re-positioned for the next build layer. The overall sequence of mixing and construction steps was then repeated, eventually leading to the clay liner built up to a total of 65 cm shaft height. On completion of the full clay liner, the remaining layers of sod were placed, to a finial height of 65 cm.
    After allowing the clay to dry and harden for two days, the ash sand mix was carefully scooped from the inside. Some shrinking of the total wall height was obvious, as well as settling of the sod layers themselves. Using a dry wall saw, the clay was cut along the line available by the interior spacing of the framing stones. This created an available extraction arch opening 20 cm tall by 23 wide. Although a cut line was made, this section of clay wall was left completely in place. A smaller opening was cut into the bottom centre, creating a port for the slag tapping that was expected to be required. (7) This was a distinctive upside down U shape, 8 cm tall by 6 cm wide. The clay here was pulled free, and slightly re-shaped to allow it to be removed or replaced as might be needed.
    After the tuyere was installed (discussion below), the distance of tuyere tip centre above the existing hard base of the furnace was found to be an acceptable 23 cm. Unfortunately, with the slightly reduced total shaft * height now 63 cm, this also allowed for only 40 cm of functional stack * distance. This has come to be considered a bare minimum for effective ore reduction in furnaces of this size and type. To ensure a working furnace stack, and to provide a bit of extra against expected complications, an additional 10 cm of clay wall was added to the top of the liner (so creating a stack height of 50 cm) This extra material would stand well above the existing line of the diagonal sods however.

* Shaft is defined as distance from ground level to top of furnace.
Stack is defined as distance from centre of tuyere tip to top of furnace.
(Markewitz, 2021B)


    A critical element to effective furnace operation are the dynamics of air supply. Within the archaeology of Viking Age Iceland there has been nothing recovered indicating directly the use of ceramic, copper alloy or iron tuyere tubes. (Smith, 2005, pg. 191)
    Tuyeres inserted through furnace walls are all subjected to the full internal temperatures during smelting. Clay tubes will sinter into (relatively) durable ceramic, at least on the furnace ends, further subjected to visible heat erosion effects, and thus can be found archaeologically (Pliener, 2000). Heavy walled copper tuyeres have been long used by this team, and have been found to have almost no damage effects over many smelts. Outside a possible archaeological recovery of what would certainly have been an ‘expensive’ and multiple use piece of equipment, a copper tuyere is unlikely to leave any visible signs. (8) A third possibility would be a very thick walled forged iron tuyere. Although some recent tests have been undertaken using this type, limited information is available on what signatures might remain. (Sauder, 2020.)
    The only remains found at Háls suggestive of the air supply system was a single clay fragment suggesting “... a circular, vitrified opening ...” (Smith, 2005, pg 191) which was later described as indicating a 5 – 6 cm diameter. (Smith, pers. Comm.)This potentially could be clay initially packed around an insert type tuyere, but later allowing hot gasses to escape. Alternately the fragment could be the opening for some variation of the ‘blow hole’ method (where the air supply tube is placed just touching, even proud of, a hole cut into the furnace wall). This arrangement would not create any significant damage to metal or ceramic tuyeres, and might even allow the use of bone, possibly even wood, tubes. With past test uses of a blow hole (Phase Two and others), there has been found to be a serious impact on air penetration into the furnace, in turn changing the dynamics of how heat locates in the furnace. This results in higher, shallower, slag bowls, typically only extending part way across the furnace diameter, requiring more careful slag management (so frequent draining / tapping). More heat washes back on to the wall surface around and above the insertion point, creating more damage to those surfaces. With a reduced size slag bowls also comes smaller blooms. It needs to be remembered that any variation of the blow hole system requires a fully exposed and self supporting front furnace wall section, which would necessitate thicker clay walls, at least in this area.
    There is an advantage to placing any solid tuyere tube so it rests on top of a lintel stone, both to maintain position and the weight. It wouldalso be possible to mount tuyere tube directly through a stack of turf. The turf layers would certainly be less stable, and are likely to result in shifting of at least entry angle over the course of the smelt. Also any metal tuyeres rely on freely exposed surfaces to radiate off excessive heat, thus keeping their interior portions from melting. At the very least, the depth of surrounding (and insulating) turf would have to be limited as much as possible.
    The use of some variation of the blow hole method, although possibly indicted in the archaeology, was considered an additional complication at this point in theseries. As the intent of this experiment was to test the simulated Icelandic clay mixture, it was decided to use a long proven forged copper tuyere. The total length is 45 cm, the size tapering from 5 cm ID at 3 mm walls at the air supply end to 2 cm ID at 6 mm thick on the furnace end, with a weight of 3.4 kg.
    To reduce as much as possible strain on the clay liner, the heavy copper tuyere was placed so its bottom edge would rest on the top of the lintel slab. This placed the tuyere above the extraction arch, instead of our standard mounting of the air input at right angles to the arch. In keeping with our past experience, the tuyere was set with an overall downwards angle of 22 degrees and with the inner tip placed extending 5 cm beyond the interior wall.
    The final connection to the air system was made using a set of steel pipe fittings and hose linking to ahigh volume electric blower supplying air. (9) To allow for a clear access to the tapping port and extraction arch, these fittings were hung from a set of steel rods, with uprights located to either side of the front ‘slot’ in the stacked sods. Although certainly differing in detail from the Norse use of a human powered bellows system, it is our feeling that tubes made of heavy leather hung from wooden poles could easily allow for the same type of required offset to the air supply.

Ore and Charcoal

    Icelanders have utilized locally available primary bog iron ore over their history. The exact iron oxide, silica (as sand), and other components will vary by deposition type and environmental factors, not only between specific locations, but also over time. Although traces are found across Iceland, finding a deposit large enough, and with a high enough iron (FeOOH) content to allow for useful exploitation in direct process bloomery furnaces is based on chance. The amount of silica present directly impacts on the amount of slag formed, which may be significant to understanding the process at Hals, where less evidence of tap slag has been found.
    Importantly to these experiments, there are no naturally occurring bog ores within several hundred kilometres travel of the test location in Ontario. Starting with the ‘Vinland’ series in 2008, a suitable ‘bog ore analog’, had been developed, originally based on the fragments of ore recovered during excavations at L’Anse aux Meadows, Newfoundland. (Markewitz, 2014)
    Our decision was to proceed with our well proven, DD1 analog mix as ore (commericial Fe2O3 red oxide + 10% whole wheat flour as binder, at 53% iron content).This is mixed as dry powder, then water is added to create a paste like consistency. This is then spread on shallow trays and allowed to sun dry before collecting. As with the early phases, these tests would continue using roughly 25 kg ore amounts per smelt. In addition, an initial amount of iron rich tap slag was added as broken pieces. It has been proven over a large number of earlier smelts that this sequence will quickly establish a working slag bowl system, speeding the formation of the metallic bloom and also resulting in increased overall yield. The exact size of the blooms produced at Háls remains unknown, but it was considered that the 3 to 5 kg blooms normally expected from this combination would certainly prove the basic effectiveness of the smelting furnace and sequence employed.
    The second variation of materials was the type of charcoal to be utilized. The most likely species used at Háls would be birch, although it does need to be pointed out that identification of the exact type was never undertaken. The experimental work at Wareham has used commercial hardwood charcoal, either maple or oak, as available. For these specific smelts, white oak charcoal was used. There is a clear relationship between wood species and charcoal density, the same volume of oak containing more contributing carbon than the same volume of less dense birch. This suggests that to provide the same reactive chemistry, a slightly taller stack height would be required for a furnace using all birch charcoal. (It must be noted that it is unknown if any specifically documented experimental tests have been undertaken to measure this.)

Smelting Sequence

    A fairly standard sequence was undertaken both times. A period of pre-heat with wood splits is followed by filling the furnace with rough charcoal. As the interior fully ignited, this is changed to graded fuel, broken and screened to pieces ranging from .5 to 2.5 cm. Charcoal is measured using a standard ‘bucket’, roughly 1.8 kg, (later one ‘bucket full’ is weighed to calculate the total amount consumed). This is added to the furnace as the top level has dropped roughly 5 cm, about 500 gms per addition. Once the entire charcoal column has fully ignited, air volume is adjusted to set what was considered a suitable burn rate, typically aiming for a rate of 8 to 12 minutes per bucket. Slag charges were added first, followed by first charges of ore. Slag and ore is pre-measured by weight, and those amounts added evenly through each of the quarter charcoal bucket units. As the reaction progresses, individual ore charges are increased, normally reaching between 2 - 3 kg per charcoal measure. Slag is tapped as required to ensure the air blast remained clear, primarily determined by changes in sound, but also by observing the tuyere tip via a glass view port. Once all available ore is added, additional charcoal is added to ensure all the reduced ore is deposited onto the bloom, then the remaining charcoal allowed to burn down to close to tuyere level. At this point the bottom arch is opened and the bloom mass is broken free of the slag bowl and extracted. The mass is then hammered with multiple sledge hammers through a single ‘heat’ cycle to knock off clinging slag and partially compress the bloom. Final weight is taken afterwards. (10) Use of a standard method is considered important, as it allows comparisons between previous experiments.

Smelt A – June 20, 2021


Figure 6 : Front elevation of completed furnace, ready for experiment A

    Details :

Darrell Markewitz was smelt master, with Neil Peterson as lead hand and record keeping, Rey Cogswell as primary assistant.

    This smelt generally ran as was expected. The one modification from standard practice was the use of a slightly lower air volume, set intentionally to more closely mimimic amounts possible through the use of human powered bellows systems. This substantially increased the burn time per kilogram (where something closer to 4 – 5 minutes is considered more ideal). The most obvious impact of this reduced air flow was that the slag bowl initially developed too high and potentially too close to the tuyere side of the furnace. This necessitated an early tap to prevent ‘drowning’ the air blast. Shortly into the addition of ore charges, some of the lower charcoal fines were scraped out, allowing tapped slag to flow below the bottom of the existing bowl. This method has been found to soften the bottom of a slag bowl, allowing it to slump downwards, in effect lowering the entire complex so as to leave more space both for the accumulating liquid slag and the developing bloom. Over the course of the entire smelt, the furnace would also self tap a number of times, a total of 7.9 kg tap slag would be recovered.

    Another impact of the lower air rate was that the final bloom was larger overall, but with a noticeably less dense ‘spongy’ texture, visibly with considerable slag still included in patches. Although this bloom has a high yield number (especially for a smelt of this ore amount), there is expected to be a high loss at the important bloom into bar phase. (11)
    It proved easy enough to break the front part of the interior slag bowl clear. The size of the extracted mass did result in also pulling away the lintel stone, which in turn caused the sod strips being supported by this stone to also fall away. The line of breakage to the interior liner was slightly above the original cut for the extraction arch, running along just at the lower edge of the tuyere point.
    During the extraction, the pieces of broken sod and any hot slag were shovelled up and discarded to one side, primarily to keep the working area clear. he fallen lintel block was shifted slightly further away from the furnace, to serve as a fulcrum for the long bar used to lever the bloom mass free. The overall extraction process created a triangular shaped spray of small slag fragments and partially burned charcoal, leading away from the front of the furnace. The portions of the slag bowl remaining inside the furnace were a C shape, pointed towards the extraction arch, with a clear depression marking the location (and size) of the removed bloom mass. All these features are considered relevant, as the exposed remains of the last furnace fired at Háls (number VII), showed much the same kind of features.


Figure 7 : Furnace just after extraction, June experiment.

    Inside the furnace itself, there was considerable erosion of the clay liner, following the expected oval pattern centred around the tuyere. At some places (especially close to and above the tuyere) the wall thickness had been reduced from the starting 4.5 cm to as little as an estimated 1 cm thick, all now heavily slag coated. The Icelandic clay mix did survive a full smelting cycle, but with more erosion damage than was the case with the EPK clay (higher firing temperature) normally used. This depth of erosion suggested that if the wall thickness had been much less than the 4.5 cm used, failure due to melting through the walls would certainly have occurred. On examining the overall damage to the furnace structure as it slowly cooled, it was felt that it should prove fairly easy to replace the lower sections that had broken away, and layer fresh clay onto the eroded areas inside.

1e 2e

(additional) Photo mosaic of furnace after Smelt A - bottom is marginally 'north'.

In the early hours of the morning following experiment A,
a roughly 20 cm long snapping turtle had crawled into the
still warm furnace! (12) The tuyere had been torn free, and
a section of the front wall above this, roughly 30 cm tall
by 15 cm wide, had been knocked out.

(additional) Turtle as discovered the following mornng.
In the image to the right, the red line shows the lower lin
of the furnace wall - after extraction.


Repairs for Experiment B

    The portions of slag bowl remaining after extraction were left in place. A metal cover was placed over the top opening of the furnace liner, but otherwise the sod structure was left exposed to natural weathering over the summer months, with the intent to schedule experiment B in early October. By August, however, with concern that the existing furnace might be suffering too much damage under late summer rains to be easily repaired for re-use, it was decided to move the scheduled second experiment forward. Repairs were undertaken the week of August 9 to 13. The remaining slag bowl was removed, with any lumps of slag still adhering to the inner walls broken free. The interior was cleaned down to the original starting hard base (dirt) level.


Figure 8 : View down the interior, after slag clinging to walls and remaining lower bowl was removed.
Extraction arch to the top of this image.

    In figure 8, the considerable heat erosion of the clay liner is clear. Seen at the bottom of the image is the smooth line of original clay surface, protected by the base layer of charcoal fines. The heat effects are concentrated to the central portion of the shaft, the hottest part of the interior during firing. The upper portions of the furnace (although somewhat distorted here by perspective) are again seen to be relatively free from any damage. The overall erosion was both more extensive and considerably deeper than had been seen in previous clay walled furnaces. This was considered to be primarily the result of the significantly lower melting point of the simulated Icelandic clay body.
    First a new extraction arch was created, with the framing build up using the same stone blocks and lintel stone as before. One change was that the arch space was filled with a number of smaller flat pieces of stone (gneiss), which were placed to form the lower front wall into the interior of the furnace. Gaps between the stones were filled with the clay mixture.
    A new set of grass sod strips were cut and placed above the lintel stone, to help define the shape of the (missing) front part of the furnace. Fresh clay mix (left over from the initial build) was applied to the inner side of the furnace, first to re-establish the missing portion from the earlier turtle damage. Next, clay was added over top of the eroded wall, attempting to even the surface out back to the original 4.5 cm thickness and smooth contours.


Figure 9 : Interior after patching, larger areas indicated. At the top is seen the straight line of the lintel stone, which will support the tuyere tip when that is installed. The edges of the extraction arch stones are seen below the lintel.

    Overall the supporting grass sods had slumped considerably, so an additional layer of sod was added to the top, building up to roughly the same height as the remaining clay liner. The last part of the repair process was installing the same copper tuyere, positioned and supported much as it had been for experiment A.

Smelt B – September 4, 2021


Figure 10 : Front elevation details of repaired furnace, ready for experiment B.

    With re-use of Furnace A, most of the layout details remained the same, with these changes :

Neil Peterson was the smelt master, with Dave Cox as lead hand, Rey Cogswell undertaking the extraction, Marcus Burnham as primary assistant (Darrell Markewitz supervising).

    Again in Smelt B, potential air volumes were reduced towards what was considered possible via human powered bellows systems. Again the first impact of this was to cause the initial slag bowl to form too high, raising the level of liquid slag upwards to where there the air blast was being effected quite early in the process. This was partially corrected by using a steel rod (9.5 mm diameter) to probe vertically, about 1/3 of the furnace diameter back from the tuyere tip. This probe was forced down through the still fragile bottom of the slag bowl, allowing hot liquid slag to drain down into the charcoal fines layer, a process repeated three times early in the smelt. Significantly, a burning through of the front furnace wall, about 10 cm above the tuyere, was observed at roughly 90 minutes into the primary smelting cycle. This suggested that the lower volume air blast was not fully penetrating into the furnace, but instead washing back against the furnace wall at this location. Failure at the joint between the new added clay and the original wall was also likely.
    Roughly after the addition of the first two ore charges, part of the extraction arch was opened, in this case by the removal of one of the lower stone filling pieces. As in smelt A, some of the lower charcoal fines were scraped out, allowing any self tapping slag to flow below and soften the slag bowl, a process that would proceed without any further intervention for the remainder of the smelt.
    The air rate was set slightly higher than for Smelt A, certainly indicated by the shorter time per kilo recorded.

    Rey Cogswell undertook the full extraction effort, the first time attempting this task. Lack of experience would result in the extraction to be considerably drawn out, all the while the furnace slowly cooling. The net effect was that the bloom did not easily pull free of the encompassing slag bowl, and a much larger mass than normal had to be pulled from the furnace. The raw size of this mass would result in considerably more damage to the lower front of the furnace than was expected. The stones supporting the front of the furnace where all torn free. As the strips of sod above the fallen lintel block collapsed, it was obvious there had been a major burn through of the front wall, with fused dirt now attached in a thick band in the area above the tuyere. The entire tuyere and air supply system would pull off, dropping into the fan of hot debris at the front of the furnace.
    Between the extra mass now attached to the upper wall, the removal of the supporting stones, and the force of removing the large bloom encased in additional slag, the entire clay furnace liner would slump outwards as one unit, pulling partially clear of the rear sod structure.


Figure 11 : Debris at the front of the furnace, after extraction. Note the mass of fused dirt, the result of the burn through of the front wall above the tuyere.The large stone in the foreground was the lintel block.

    Although the total mass extracted was larger, after one cycle of hammering to strike away clinging slag and partial compaction, the bloom was roughly the same size as that from experiment A. Once again it had a noticeably less dense ‘spongy’ texture and visibly retained considerable slag. Again the yield recorded against ore volume is disproportionately high, with excessive losses expected at the later bloom into bar compaction phase.
    One addition to the initial compaction work here was the first inclusion of a large wooden ‘troll hammer’, as observed in use by many European teams. This proved quite effective for the initial phases of hammering off the clinging slag ‘mother’ from the bloom itself.


Figure 12 : Marcus Burnham with the ‘Troll Hammer’, Richard Schwitzer with a standard sledge, Dave Cox holding. The slag shows as the darker material still encasing the hot metallic bloom core, more obvious in a photograph than in real life.

3a 3c 3c
2a 2b 2c 2d
1a 1b 1c

(additional) Photo mosaic of furnace after Smelt B - bottom is marginally 'north'.

Showing clay liner slumped towards tuyere side.
Note gap between rear sod layers.
Image down interior, tuyere side to bottom.
Note how interior slag bowl completely missing.

Air Rates

    A major addition to instrumentation had been made, and employed for this smelt. This was the purchase of an Omega HHF1000 flow meter. (13) This unit allowed for far more accurate measurements of the actual air volumes being injected into the furnace, over the roughly calibrated ‘sliding blast gate’ numbers recorded in the past. (14) One of the problems with the earlier measurements was that these were the potential total air available from the air equipment, and so did not account for the restrictive effect of the materials inside the furnace stack, certain to vary considerably over time. As this was the first use of this gauge, only infrequent measurements were made :

Time Event Plate Meter Calculated
10:11 fill rough charcoal      700

10:15 air increase 900

11:19 full graded charcoal
11.5 865
11:23 air reduced 700 not recorded 550
12:00 air increased 800 8.8 660
12:44 air increased 900 9.6 720
Table C : Air volumes as estimated / recorded / calculated

    The overall implication of these measurements is that earlier reported air volumes are only significant as they are compared to each other, not an accurate indication of exact volumes. Roughly, these earlier numbers are likely to be 20 % higher than the actual air amounts into the furnace. In working practice however, the physical burn rate (time for consumption of a given amount of charcoal) is used to determine the best functional air volume at any given point.



Figure 13 : Top edge of the furnace liner, above the tuyere – condition after 5 weeks from June firing.

    The temperature gradient that is produced by the firing of a smelting furnace has been seen to effect clay composition walls in differing ways, which in turn results in variations in damage effects as a furnace is exposed to weather over time. There had been an unusual amount of rain over early Summer at Wareham (a period where normally there had been little, if any, rain in past years). The fracture line seen in figure 12 (image taken July 17) marks the division of heat effects between the sintered interior and the ‘baked mud’ exterior, along the top edge of the clay liner. By early August, the outer portions had broken free along this line. There had also been considerable overall slumping of the sod structure, again due to the exposure to rain.
    The plan for Smelt C was to leave the furnace standing, with a roughly 100 cm diameter round wooden cover (‘shield sized’) placed over the top. Normally there is a smelt experiment in later June, so this schedule would subject the furnace to a full Ontario winter of freeze and thaw, then the expected spring rains. (Freezing temperatures can be expected at Wareham by later October, snow by mid November, eventually accumulating to a metre or more in depth.)
    It was clear however, even at the end of Smelt B, that a significant rebuild of both the inner clay liner and the surrounding sod structure would be required before any attempt to use the existing furnace again. September in Central Ontario also saw a huge increase in rainfall over past decades however, clearly accelerating these effects. (15)


Figure 14 : Rain erosion, September 24 - compare with post smelt image seen above (figure 11).


Figure 15: Side view, showing slumping of the furnace towards the extraction side (red lines).
Slag moved from the working area (while hot) in the foreground.

    Figure 15 clearly shows the slumping forward of the remaining clay liner, a total of roughly 17 degrees off the original vertical, over the first 20 days since the last firing. Any attempt to re-use this furnace would most certainly require as a first step clearing away the falling dirt, and restoring the line of the liner to vertical. Given the overall fragile condition of the clay, especially along the mass of fused dirt on the front wall, it is considered unlikely that such a re-adjustment would even prove possible.
    It was decided instead to leave the furnace remains openly exposed to weather, basically the start of observing the effects of environmental processes on this specific furnace type. It is the long term plan to continue recording the effects of natural weathering on this furnace, the hope is over at least a decade.

Time Study

    In an effort to get some concept of the labour involved in building the furnace described, then operating this through a full smelting cycle, some attempt to keep time / labour records was made. The build work was undertaken by a single, well experienced individual, the time indicated is ‘person hours’. It should be noted that digging / cutting was all undertaken using modern steel bladed tools.

    This puts the total build time for this furnace at 17 ‘person hours’. Importantly, this is not including gathering required materials.
    One significant labour factor that is missing is the time and effort related to making charcoal. Commercial charcoal was used, purchased as ten, 8 kg bags per smelt.
    Another major labour component not included here is that required to find, gather, and potentially roast, a suitable quantity of natural bog iron ore. The DD1 analog used for these smelts requires infrequent moving over the drying period (several days, dependant on weather).

(minutes) Team Size Pre-Heat to Temperature Main Sequence Burn Down Extraction Event Total

Smelt A 3 127 28 393 58 20 626

Smelt B 3 82 86 156 40 30 394

105 57 275 49 25 510

314 171 824 147 75

LABOUR minutes 1530

as hours 25.5

Table D : Working Times (as minutes)

    Over the smelting process, it has been found that the most effective working team is comprised of three individuals. The normal allocation is :

    During the initial pre-heat phase, charcoal is also broken and screened for size.This usually occupies two individuals for roughly one hour (at the volumes used here). Tap slag and ore (depending on type), are also broken to size. During the extraction phase, one individual frees the bloom, with one closely assisting, and the third maintaining the furnace and safety. Compaction normally involves one person holding the mass, two striking. It should be noted that different individuals may take on these various tasks over the course of the long hours required.
    The smelt sequence records made do not include the full time required for the burn down, or the exact time required for extraction or the initial consolidation process. Most commonly, work on a typical iron smelt day starts at 8 am for the team leader, with final walking off the work site at roughly 6 pm (for a smelt of this size).
    There is always work setting up tools and general preparations at the start of a smelt, then an initial clean up of the work site afterwards. Roughly these combine as one person over 2 hours total. (and are not included below)    
    That all being stated, the average labour given for the actual smelt at 25.5 person hours is at least a reasonable estimate. It is important to remember that this represents a total using the faster burn rates possible via an electric blower. As well taking longer for the entire sequence, there is also a major labour component to the continual operation of a human powered bellows system of suitable volume / size. Past experience of this team suggests an addition of three individuals, working in short rotation over an expected 8 to 10 hours.
    Within the limitations given, this paces total labour for one smelt at roughly 44 + person hours for the creation of a roughly compacted 8 kg bloom (not counting bellows workers). It should be remembered that this is still not to the desired end product, fully compressed ‘currency bars’.


Primary Test A : The function of simulated clay body,
based on the best possible duplication of the material recovered close to Háls.

    Initial heating tests of samples of the simulated Icelandic clay had suggested that there might be some problems with melting of this material when subjected to the normal firing temperatures inside a bloomery iron smelting furnace, expected to be as high as 1250 + C.
    The decision to build the clay walls to a thickness of 4.5 cm was primarily based on experience gained earlier during the Phase 2 experiments. This turned out to be a functional minimum, at least as demonstrated by Smelt A. Through that fairly standard firing sequence, the normal erosion effects that are concentrated around the tuyere had reduced the wall thickness to as little as 1 cm in some spots.
    Another expected result of the lower melting point exhibited by this clay body was an increase in slag produced (over the amounts expected from the ore analog). This proved to be the black (iron rich) and importantly, fluid type. This did require both tapping out and other management processes over the course of both smelts to ensure the air blast through the tuyere was not interrupted.
    Used as component of a well understood furnace layout and employing a standard firing process, the simulated Icelandic clay proved quite suitable through a single firing.

Primary Test B : How does this furnace perform with repairs and a second firing?

    It was clear however, that excessive internal erosion had resulted during the first firing. After breaking clear the first slag bowl that remained, generally it proved fairly easy to apply new clay on to the inner surfaces, in an attempt to restore the original 4 – 4.5 cm wall thickness.
    The effectiveness of needed repairs was further complicated by two factors (both beyond the control of the experiment). The first of these was the additional damage done to the middle front wall by the turtle. The second was a two week delay and wet weather, between the undertaking of repairs, and the mounting of Smelt B. (15) Both these factors may have contributed to less than ideal bonding between the existing wall and the repairs applied. This most likely the reason for the burn through failure experienced in Smelt B.
    It did prove easily possible to repair a previously used furnace and effectively undertake a second full smelt.

Primary Test C : How does a well designed furnace, on the Háls pattern, endure over time?

    The effects of the additional turtle damage, plus that caused by less experienced hands during extraction in Smelt B, certainly accelerated overall impacts to both the clay liner and the surrounding sod structure.
    The original outline for Phase 3 called for the furnace to be somewhat covered, and otherwise exposed over an Ontario winter. In light of the damage at the end of Smelt B, coupled with the effects of unusual rainfall over the month of September, it has been decided not to attempt to re-use this furnace. The impact of overwintering and a third firing thus remains untested.
    Instead, the furnace and surrounding work area will remain in place and be exposed to the natural cycle of weather over an extended period into the future – ideally many years. As it degrades, it is hoped that records of that process may serve to have some value to archaeologists when interpreting ancient smelting furnaces.

    One additional observation from the two smelt experiments of Phase Three lays in the ore into iron results. Both smelts used identical ore analog, ore amounts and charcoal type. Air volumes applied were roughly similar. The overall working sequence both times was also similar, outside the fine detail of needed slag control. The end results where found to be roughly identical as well, with unusually high bloom sizes and yield numbers, with the blooms created of similar quality.

    This all proves the basic viability of this reconstruction of the Háls working system, at least for a limited set of firing cycles. The placement of several furnaces in close proximity at Háls suggests that these also were short use cycle structures, tore clear and re-made as each failed.

Future Experiments?

    At this point in this extensive experimental series, it is felt most of the major elements of furnace design have been tested through full smelting sequences. Obviously, the major overall discrepancy remains the difference between the materials available at the Ontario location, and those at Háls in Iceland. The most obvious solution would be to undertake at least a limited experimental series in Iceland, but simple logistics (primarily funding) make this unlikely. There are three elements understood as being important :

    As detailed, the exact type of tuyere used at Háls, and its method of mounting, remains unknown. The larger series has tested both ceramic and copper tuyeres, using primarily inserted mounts, but also variations of the ‘blow hole’ method. Two further tuyere types remain untested by this team :

    The impact of modern electric blowers versus human bellows systems for supplying air still remains to be tested in detail. The exact form of bellows used by Viking Age Icelanders in the iron smelting process remains unknown, as there is no direct artifact evidence for smelting equipment that has been found. There is no doubt various human operated bellows systems, applied to this type of ‘short shaft’ bloomery furnace, will result in the production of metallic iron, as many teams of iron makers (including this group) have certainly proven. Most of these efforts are experiential however, reporting (when done at all) has been more qualitative than quantitative. Research into air systems remains ongoing.


    The author would like to acknowledge the huge amount of enthusiastic, volunteer, labour contributed now for over two decades by members of the Dark Ages Re-Creation Company. In addition to the primary workers credited for their larger roles over these two experiments, the following also directly participated :

Richard Schwitzer

Marcus Burnham

Kay Burnham

Kelly Probyn-Smith

    Considerable thanks goes to Kevin Smith, who has become a close advisor over 20 years of learning and investigation to the Norse of the North Atlantic. It is largely due to his encouragement and constant assistance that all this work into bloomery iron smelting had shifted from the experiential to the truly experimental.

    A special thanks to the additional financial contributions of Neil Peterson, who donated all the charcoal used in the 2021 experiments.


1) During the 2000 excavations, a two by two metre square was excavated down to the top surface of the furnace slag bowls, and exposing only part of what may be a larger iron making area. (Smith, 2005) Unfortunately, Smith's work on the site was halted at this point.

2) The size of this bank was given as ‘roughly the size of a tractor trailer truck’. Information provided over a long duration series of private e-mail communications between Kevin Smith, Michelle Hayeur-Smith, Neil Peterson, and the author.

3) A detailed report is available : Markewitz, D., 2021. Sticking To It – a clay mix for Icelandic. [last accessed November 2021]

4) One outstanding concern is how similar in structure the cut grass sod at Wareham in Central Ontario is to the natural Viking Age turf at Háls. The environment in Iceland is radically different, as is the vegetation, even the composition of soil itself. In the Phase 2 tests, there was found to be an obvious difference between grass sods cut at Wareham, and those cut for experiment 8, about 45 km south at Laurel. A fuller discussion of this difference in materials is included in : Smith, P. K., Markewitz, D., and Peterson, N.. Now with 70% Less Clay! Experiments with Viking Age Icelandic Turf Walled Iron Smelting Furnaces. (publication pending).
A short video overview was presented at the EAC 12 virtual conference, March 2021. [last accessed October 2021]

5) See a fuller discussion of building methods and their indications : Markewitz, D., 2021. Stacking Up – on constructing clay furnace walls. [last accessed November 2021]

6) This method (internal wicker form) was used for the first test in the ‘Pictish’ series. See the overview: Markewitz, D., 2014. Late Pictish Late Iron Age Smelt. [last accessed November 2021]

7) It is unknown what, if any, silica (as sand) might have been present in the natural bog iron ore available and used at Háls, and there is limited evidence of tapped slag. It was fully expected that slag tapping would be required in these experiments, based on past use of the DD1 ore analog. It is thus possible that the entire dynamics of slag control, as used here, including the implications to furnace design, may not resemble that used by the Icelanders.

8) Once exception to this are distinctive slag rings, which are formed around the end of the copper tuyere that is exposed inside the furnace. Such ‘end caps’ are quite fragile, but if recovered, can indicate tuyere penetration past the inner wall, insertion angle, inner and outer diameters (thus tube wall thickness). Similar slag rings were excavated at Jamestown Settlement, Virginia, examined by the author in 2012. See : Markewitz, D., 2012. Slag Tuyere Rings. [last accessed November 2021]

9) The blower used is a high volume ‘compression’ type, with a specified maximum volume at 1400 litres per minute. Actual delivery is controlled by a sliding plate across the output. This has been (very!) roughly calibrated to 100 litre hash marks. The whole question of how well any electric blower system can simulate the air produced from probable Viking Age bellows types is considered an entirely separate area of investigation. Although a number of potential Norse style bellows have been tested in past experiments, these all require a considerable number of additional workers, beyond the smelt team itself, to operate effectively. See Markewitz, D., 2012. An Iron Smelt in Vinland. [last accessed November 2021]

10) This is only a brief description of the standard working sequence that has been developed over two decades by this team. A fuller description of the process can be found : Markewitz, D., 2012. “If you don’t get any iron ...” - Towards an Effective Method for Small Iron Smelting Furnaces. EXARC Journal, Issue 2012-1. [last accessed November 2021]

11) It needs to be remembered that the desired end product for ancient iron makers were fully compressed working bars – not semi-product blooms. Modern makers too often report their ore to bloom yields, without reference to starting iron content of ore, or any consideration of bloom density.

12) The snapping turtle (Chelydra serpentina) is native to North America. The smelting area at Wareham is adjacent to a small pond, and a creek runs behind the (rural) property . See : [last accessed August 2021]

13) One of the problems effecting all Independent Researchers is limited access to high quality (and priced!) instrumentation. For details of this instrument, see : [last accessed August 2021]

14) See : Markewitz, D., 2021. Mind the Blast – Correcting Reported Air Volumes.

15) Records from Environment Canada give the past average rainfall for the area around Wareham for the month of September as 7.7 cm total. In 2021, there was a single one day amount of 5.6 cm. Although exact overall total for September were not available at time of writing, September 2021 certainly had the highest amount of overall rainfall, often as heavy downpours, seen over 30 + years of personal observation.

16) The repair work was undertaken on August 18, with the second firing initially scheduled on the 20th. Due to unforeseen events, it was necessary to postpone Smelt B, and the shift of roles during that experiment.

Works Cited :

Geology Science, 2018, Basalt. [last accessed December 2021]

Geology Science, 2018, Granite. [last accessed December 2021]

Markewitz, D., from 2006 (ongoing). Experimental Iron Smelting. [last accessed December 2021].

Markewitz, D., 2008. Working towards an Icelandic Viking Age Smelt Based on the remains at Háls. [last accessed October 2021]

Markewitz, D., 2014. An Iron Smelt in Vinland – Converting Archeological Evidence to Practical Method. in Can These Bones Come to Life? : Insights from Reconstruction, Reenactment, and Re-creation - Volume 2. Cramer, M. (editor), Feelance, Wheaton, Illinois.
The full text of the initial version (V3 - June 2010) : [last accessed November 2021]

Markewitz, D., 2021A. ‘Evidence of Absence’ - The Erosion of Bloomery Iron Furnace. [last accessed October 2021]

Markewitz, D., 2021B, Standardized Reporting of Experimental Iron Smelting – a modest proposal, the EXARC Journal, 2021-1 (also available as pdf) [last accessed January 2021]

Pleiner, R., 2000. Iron In Archaeology – The European Bloomery Smelters. Archeologicky Ustav Cr, Prague.

Sauder, L., 2020 (?- nd.). Forging and using a Wrought Iron Tuyere. [last accessed November 2021]

Smith, K.P., 2005. Ore, Fire, Hammer, Sickle: Iron Production in Viking Age and Early Medieval Iceland, in AVISTA Studies in the History of Medieval Technology, Science, and Art, Volume 4. Bork., R (editor). Ashgate, Aldershot, UK.

Also available as PDF : [last accessed October 2021]

Additional References

The detailed sequence data for these two experimental smelts are available online :

Smelt A, June 20, 2021 :

Smelt B, September 4, 2021 :

[both last accessed November 2021]

Markewitz, D., 2021. Into Phase Three – Continuing Experiments with Iron Smelting, based on Háls, Iceland. [last accessed November 2021]

Peterson, N. 2006 to present. Experiments in the creation of iron using Viking-Era Norse techniques. [last accessed December 2021]

About the Author...

Darrell Markewitz from Ontario, Canada, has been a professional artisan blacksmith since the later 1980’s. Known as a specialist in the Viking Age, he has contributed to a number of major museum exhibits, including extensive work for Parks Canada at L’Anse aux Meadows NHSC. His interest in bloomery iron smelting started there in 2001, and his extensive practical and experimental research has included participation in a number of international projects and academic conferences over the last two decades. As well as a number of published academic papers and journal articles, his past work is well documented on an extensive web site : , plus commentaries on his blog :

Darrell Markewitz
the Wareham Forge
307377 Centre Line A
Proton Station, Ontario

Unless otherwise indicated :
All text and photographs © Darrell Markewitz, the Wareham Forge.