Sunday, June 27, 2021

Icelandic Clay Mix Test - the full smelt

The following is a fast overview of the bloomery iron smelt at Wareham, June 19.

This experiment has the DARC team return to our experimental work based on elements of the 'industrial' Viking Age site at Hals, Iceland, originally excavated by Kevin P. Smith (1). The earlier work can be loosely divided into two phases (2) :

- Phase One / four experiments / 2007 and 2008 / testing individual design elements

- Phase Two / four experiments / 2012 to 2016 / testing use of turf builds, combining elements from phase 1

The current work, as Phase Three, centres on:

a) testing clay mixtures based on those from Iceland

b) testing durability of a turf (sod) supported furnace structure over time

An earlier blog post ("Sticking to It - a clay mix for Icelandic?") detailed the logic behind the clay mixture itself.

Part way through the build process

Based on experience from the earlier furnaces of this series, the wall thickness was set at 4.5 cm, measured against 'two fingers' as shown. An internal form was used, metal in this case, keeping the interior diameter consistent to 28 cm. (This replacing what could have been a wooden straight sided 'barrel' form historically). The clay was constructed to the rough top of the form (about 30 + cm), then the supporting cut sod strips were placed in a circle on the outside. The form was pulled upwards, with the exposed interior filled with a mix of half sand / wood ash. This method has been used many times, the ash mix both supports the thin clay and also helps to dry the walls. 

Built at 35 cm

After the first full course was established, a set of small stones, used like bricks, were placed to frame an extraction arch for the front of the furnace. One large piece of basalt on hand was used as a lintel to eventually support the tuyere and upper sod pieces.

With air system, at first charcoal addition

The build continued to raise the total stack height to 75 cm. The supporting sods were laid in five layers, to a height of roughly 70 cm (at the start of the smelt). The cone created was a bit irregular, extending out 72 cm to the left side as seen, to 56 cm to the right side. Although materials (timber and earth) had been labouriously gathered, it was decided not to box the sod cone and back-fill to a flat upper surface.

The extraction arch was 23 cm wide at the furnace wall, opening to 33 cm at the edges of the framing stones. The lintel slab sat on a slight diagonal, at roughly 20 cm high. There was additionally a small tapping arch cut, 6 cm at the base and 8 cm above the hard base. 

There was no specific attempt to match the air system to evidence from Hals, we chose to use our proven heavy copper tube, inset 5 cm beyond the interior wall. As supported on the lintel slab, and with our proven 22 degree down angle, this set the tuyere tip a bit high in the furnace. In turn this reduced the effective stack height to only 40 cm (normally considered a minimum distance). As this position would also leave a lot of space under the tuyere, a 'soft base' was created with wood ash from the drying fire and a layer of charcoal fines on top, raising the set base to 15 cm below the tuyere. The simplest method to support the normal viewing port and air input connection was to suspend it from a metal rod set well clear of the working opening in the sod cone. 

The primary smelting team :

Darrell Markewitz - smelt master

Neil Peterson - ore and records

Rey Cogswell - charcoal / ore

Kay Burnham - charcoal / ore

Richard Schwitzer - compaction

Kelly Probyn-Smith - safety

Travis Sweet - photography

Slag tap. Earth falling free of the upper sods can be seen as light reddish pieces.

Over the 7 1/2 hours of the main smelt sequence, it proved necessary to make at least four major slag taps to lower internal slag levels threatening to 'drown' the air blast. 

Later into the smelt, it was clear that furnace heat was seriously effecting the sod structure. Earth was baking dry, and pulling free from the root systems and falling downwards between any cracks. This would cause the whole sod cone to slump into itself by the end. There was no venting of furnace gases observed however.

Air flow had been purposefully reduced for this smelt, down to a range considered possible for a human operated bellows of Norse type. Although the long serving electric blower was used, air flow was initially set to roughly 700 LpM, increased about the mid way point to 800 Lpm. (3) This would create an average burn rate at 8 minutes per kg over the main sequence. 

Initially 5 kg of previously gathered iron rich tap slag was added to establish a working slag bowl system. This may have also contributed to the need for later tapping. A total of 25 kg of the DD1 analog ore (51% Fe) was used with combined average addition rate of 11.8 minutes per kg.

Pulling the bloom mass free. Image by Travis Sweet

The bloom mas proved much larger than expected! In pulling it clear of the furnace, the supporting lintel stone was pulled away, resulting in the upper sod layers at the front of the furnace also collapsing, exposing the front furnace liner.

Classic compaction shot - Darrell holding, Neil on sledge : Image by Travis Sweet
With loose gromps hammered clear, before cutting on the press : Image by Travis Sweet
To everyone's great surprise, after some first stage compaction and rough cutting, the resulting bloom, although a bit loose in texture, weighed in at 8.9 kg. This represents a 35% return from ore (4).


Furnace, just after extraction / compaction, with remaining slag bowl seen.

On a quick examination of the furnace (after our ritual post smelt Guinness) :

- The furnace still contained about 2/3 of the slag bowl, completely filling the interior save for a large piece broken away at the extraction arch to expose the bloom, creating a clear C shape. The depression towards the air blast side clearly frames where the bloom was pulled free. The slag bowl remaining is of a size and shape quite similar to those exposed at Hals.

- The front portion of the furnace has clearly broken away, when the supporting stone lintel pulled off, as the bloom mass was finally pulled free. The image above shows it's natural fall position. The sod that had been above this has pretty much totally crumbled away into it's containing dirt.

- There is considerable slumping of the encasing sod cone. The clay liner had actually lifted slightly several times while working the bloom free, but had remained as a solid structure overall. 

Largest of the roughly compressed and cut bloom pieces : 3 kg

to come : looking closely at how the clay walls survived the smelt


1) Smith, K.P., 2005, "Ore, Fire, Hammer, Sickle: Iron Production in Viking Age and Early Medieval Iceland", AVISTA Studies in the History of Medieval Technology, Science, and Art, Volume 4, USA

Also available as PDF on line :

2) A overview of both the Hals site and these experiments is currently under preparation, co authored by Smith, Markewitz and Peterson (‘Now with 70% Less Clay! Experiments with Viking Age Icelandic Turf walled Iron Smelting Furnaces’) A short video overview was presented at the recent EAC 12 virtual conference, available on line :

3) For a furnace at 28 cm, the normal air flow would be set to closer to 900 - 1000 Lpm. see : Air Flow Rates

4) This impressive result may be the effect of the 5 kg of iron slag added as a first step. Our past results for smelts in the 35 - 30 kg ore range have more typically been roughly 5 kg (or less).  

Tuesday, June 15, 2021

'Sticking to It' - a clay mix for Icelandic?

 Part of our ongoing experimental series investigating a possible bloomery iron smelting furnace, as suggested by the archaeology of the site at Hals, Iceland. ( 1 )

The clay mix being tested was based on an 'in team' analysis of materials gathered by Michelle Hayeur-Smith from a natural deposit 'fairly close' to Hals. (Noting that it remains unknown if this was the clay material actually utilized for the furnace builds there). ( 2 ). Team member Marcus Burnham suggested a set of components available at Pottery Supply House, which would in combination approximate at least the chemistry of the Icelandic sample. ( 3 )

These individual components were dry mixed by hand, giving roughly 50 kg of the clay powder (hopefully enough for two builds with reduced wall thickness). ( 4 )

I made up a total of four small batches for next step testing :
- clay mix with water
- wet clay mix plus powdered slag (66 / 33 slag) - this 'copper shot' sand blast slag ( 5 )
- wet clay mix plus fine sand (50 / 50 by weight)
- dry mix / fine sand / shredded horse manure by volume (ie our current proportions using the EPK clay base)

In each case, the various mixtures of materials holds together well, without being too 'sticky', and all are nicely plastic.
It is not expected to have any physical problems building a furnace wall structure with any of these mixes.

The individual mixtures were pressed into small cylinders, filling a standard toilet paper cardboard tube. Each 4 cm diameter by 10 cm long.

- These were placed in a toaster oven set (very roughly) for 90C, over a duration of about 4 hours total.
I'm not sure that temperature was actually reached, because when I pulled them out, I could still handle each with bare (blacksmith's!) hands.
The centre part of the paper cylinders could be depressed - suggesting there was still water remaining. This thought to be the effect of the paper not easily passing moisture (?) .
- The cylinders were then placed in the rear section of my gas forge on a metal tray. (At top operating adjustment, this forge has been found to produce temperatures into the 1100 C range) The forge was lit, with a flow at a greatly reduced level.

Showing results of first (shortened) heat cycle.
Number 1 to left, through to number 4 to right

The material was heated enough to burn off the paper on the upper surface, but this remained solid and attached on the lower side. This suggests the the top had to reach at least 230 C, but the bottom surface (furthest from flame) did not get that hot.
You can see clearly that cylinder 1 - the straight Icelandic mix, has already crumbled.
Cylinder 2 - wet mix plus slag, has also become brittle.
Cylinder 3 (wet / sand) is showing cracking - but remains whole
Cylinder 4 (dry/sand/manure) remains unaffected (that lower crack seen was from
the initial packing of the material)
As it turned out, the forge propane ran out after a short time (20 minutes?). So in this it is hard to estimate just how hot the forge interior would have gotten.
The cylinders were allowed to sit in place inside the forge (so slow cool), overnight.

The following day I refilled all my propane cylinders, then repeated the test.
Again the tray was placed at the very rear of the forge as it was lit. A low propane rate to start. The material was left in place for about 30 minutes (until the forge interior was at full heat) then the tray moved forward to the centre, under the gas jets (where I normally would place metal to heat). ( 6 )

Temperature probe was inserted into the gap between # 3 and #4 (to right).

After 55 minutes total, a temperature reading was taken, the probe wire between the cylinders, tucked slightly underneath. Reading was 980 C.
At that point, you can see that cylinder 1 is composed of crumbled pieces.

The forge was then adjusted for higher temperature (increased gas flow and air combination).
The materials were subjected to another 85 minutes at this temperature, which was measured again to 1070C.

At end of second heating period (open door for image has cooled interior slightly).

You can see that the surface of cylinder 1 has slumped and started to fuse.
Cylinder 2 has developed some serious breaks.

The forge was turned off - and left overnight with the door closed to slowly cool. 

Cylinder 1 (dry mix with water) - has melted
Cylinder 2 (wet / 33% slag) has badly cracked
Cylinder 3 (wet / 50% sand) has developed a number of cracks, but mainly towards the surface. A bit of surface fusing (point directly under the gas jet)
Cylinder 4 (dry mix / sand / manure) shows no significant effects (again, bottom piece was separate at the start)

I took each of # 2 / # 3 / # 4, and applied pressure by flexing between my hands. I attempted to keep the force as close as possible - and applied to what I judged a 'reasonable expectation' level of pressure.

( Cylinder 1 - only the small disk seen earlier could be handled, the result was it easily crumbled )
Cylinder 2 - not much force was required to break along the large cracks - you can see from the dark coloured interior, this crack had broken completely through during the firing.
Cylinder 3 - held together effectively, despite surface cracks
Cylinder 4 - no effect to reasonable force (earlier separated piece shows internal texture)

The conclusion I draw :

Sample #1, our potential Icelandic clay mix alone, is not able to survive temperatures even at the lower end of expected iron smelting range. Combined with it's relative fragility even during the early heating cycle of a furnace, this material on its own is not deemed suitable for furnace construction (without modifications).

Sample #2, with 2/3 wet clay / 1/3 slag (by weight), was considered highly speculative at best. (Included mainly because I had the material!) This offering one possible version of 'Where did the silica come from?'. Iron rich slag could have been on hand (at least after a first smelt attempt). In the test, this mix proved relatively fragile. It might be possible functionally to use this mix, with greater care in mixing, construction and drying. Generally, this is considered quite unlikely to have been used.

Sample #3, with 50% wet clay / 50% fine sand, certainly would be an effective build material. First note is that the sand used here is of unknown composition, but likely Ontario granite based at about 70% silica (higher than Icelandic basalt based). It has been well demonstrated (Sauder) that high sand mixes can better withstand iron furnace temperatures, but at the same time do require more care during the construction phase to eliminate just the kind of surface cracking seen in the test sample. This is considered a possible effective mixture for the Hals build.

Sample #4, using our developed standard of equal dry volumes of clay mix / fine sand / shredded horse manure, was clearly provided the best properties of the four mixes tested. As proven in the past, the sand provides heat resistance and stability at temperature, the horse manure re-enforcement during the build and drying phases. Even with what is demonstrated as a significantly lower point clay material, this is considered the 'best possible' mixture to use for the upcoming full build test.

1) The original description of the Viking Age, 'industrial' level iron smelting site at Hals was provided by Kevin P. Smith in his 'Ore, Fire, Hammer, Sickle : Iron Production in Viking Age and Early Medieval Iceland' The earliest evaluation towards evolving a working system can be found in 'Working Towards a Viking Age Icelandic Smelt'.
Currently Smith, Peterson and myself are preparing a detailed description of the 2007 - 2016 experimental series, to flesh out our recent presentation 'Now with 70% Less Clay' which was delivered at the 2021 EAC12 Conference.

2) Unfortunately, for a number of reasons, an actual analysis of the any of the clay furnace walls recovered during excavations is available. Such tests were not undertaken at the time, not being within the scope of the original project there. At this point (after some 20 years!) is is quite unlikely to be able to arrange for this currently

3) With the kind financial assistance of Neil Peterson, who donated funds to cover the purchase of the clay materials indicated, plus charcoal to supply the next three full smelts in the current experimental series. 

4) Sorry. Due to problems experienced over the last several years of hard won information derived from the DARC team experimental work, I am intentionally holding back on listing the exact details here. A further paper is under construction, describing our newest interpretation of the excavations at Hals, and how this applies to our current experimental series over 2021 - 2022.

5) This material is composed of 'medium' sized particles, primary components are 53-60% (Iron Oxide), 32-37% (Silicon Dioxide). 

6) Readers need to remember that the camera records light and colour of heated surfaces differently from the human eye. See an earlier blog post.

Wednesday, June 02, 2021

Elora Sculpture Project - 2021


Need a break from COVID containment?

Consider a couple of hours walking around downtown Elora, with a visit to the 'River Walk' in Fergus!

 My own contribution 

An Undiscovered Plant containing a cure for cancer

Is found at site 16 - Fergus River Walk

Tuesday, June 01, 2021

'Evidence of Absence' - Erosion of Furnce walls

Group of aging furnaces at Wareham

Evidence of Absence
Erosion of Bloomery Furnaces

This report is an overview of an ongoing process of observing a number of clay built furnaces as they weather over time, started in 2004 and added to over the years.
This includes some observations that relate back to the continuing experimental series based on the Viking Age furnaces excavated at Hals, Iceland. : Prepared May 2021

Continue to : Evidence of Absence

This commentary is part of the ongoing work related to the Hals / Iceland experimental series. All that background to the current full scale replica build and test smelt set for June 19 at Wareham.

Note to Readers : The 'improvements' made to Blogger (to 'Idiot Proof' submissions) have made any attempt to correctly format complex 'table' contents EXTREMELY frustrating (to anyone who actually knows HTML)! 

After well over four hours attempting to lay out the table of data and images as a blog post, I gave up entirely and moved over to an addition to my primary research web site.


February 15 - May 15, 2012 : Supported by a Crafts Projects - Creation and Development Grant

COPYRIGHT NOTICE - All posted text and images @ Darrell Markewitz.
No duplication, in whole or in part, is permitted without the author's expressed written permission.
For a detailed copyright statement : go HERE