'Evidence of Absence' : The Erosion of Bloomery Furnaces

Being a loose draft for what might become a formal study?

(latest edit - June 1, 2021)

One of the huge problems in practical attempts to use archaeological evidence of ancient bloomery iron smelting furnaces is simply - not much of those furnaces survive.
In most cases, all that can be expected to be recovered, after hundreds (if not thousands) of years, will be the vary base level of the furnace, basically the slag bowl below tuyere level, with perhaps the base level wall section that surrounded that. Usually there is no sign of the tuyere, or any indication of how it might have been mounted. If the walls had been constructed some clay mixture, there may be some indication by way of that inner part of the walls that had reached high enough temperatures to sinter into a ceramic. The action of rain has simply washed away any raw clay that composed the outer section of the original walls. Even the sintered portions are sure to have been subjected to years of alternating freeze and thaw cycles, further fragmenting even these remains.

Right : A series of images showing the stages of excavation of one of the furnaces at Skogar, Iceland. This furnace, showing only about the bottom 30 cm of height recovered and its clear lower base construction of stacked stones, served as the conceptual prototype for the experimental furnace seen below. (1)
furnace at

Image taken from : St. Sigur∂arson, G., & Zoëga, G., 2013, 'Skógar í Fnjóskadal - Fornleifarannsókn 2011-2012', Bygg∂asafn Skagfir∂inga Rannsóknaskýrslur, 2013/140 : page 29 (available as a PDF)

2020 start
2021 remains
Condition of the stone block furnace after the winter of 2019-20, at preheat of the November 2020 smelt. (View to the west.) Remains of the same furnace, after the breaking down for extraction, and left over the winter of 2020-21. (View to the south.)

The images above make a dramatic example. This specific furnace, part of the ongoing Icelandic series, was constructed originally in June of 2019. One important difference between this test furnace and the prototype from Skogar is that the ancient remains are only of a shallow, stone lined pit, below ground level, with no superstructure recovered. This would be a 'slag room' type in operation (slag draining into this lower chamber). The test furnace was built totally above ground level, and is a slag tapping type. It had been used for a total of five complete smelts, the last being the '65 for 65' in October 2020. The furnace had been protected by wrapping it in plastic over the winter of 2019 - 2020. For the extraction during '65 for 65', the upper portion of the furnace was broken down by pulling off each of the hot stones, which were placed clear of the working area. This dismantled the furnace down to the top of the slag bowl (roughly just below tuyere level). The jumble of stones remaining show heat effects, most with slag attached to their inner surfaces. The slag bowl itself is mainly intact, with the depression inside where the bloom had been pulled free.

Our experimental work at Wareham is typically investigating variations on furnace layout, typically related to specific historic furnaces or methods. To that end, most typically a given furnace build is often only used once or twice. (2).
As I have commented on many times over the years, despite millennia of human iron making, involving what most certainly must be tens of thousands of furnaces, it is quite rare to find much more than a slag bowl (relatively in-destructible glass) surrounded by a ring of stones or vague circle of clay. (3) 

But just how does a furnace, after it is abandoned, disintegrate over time?

Well, I had any number of used furnaces at Wareham, and have set about undertaking at least some loose observations. So far this has involved a total of seven individual furnaces, built using a number of differing materials. (4)

Group of aging furnace shafts (May 24. 2021).
The two oldest furnaces ( 2004 / 2004) are located separately.
Also not visible is June 2018, hidden behind 'DD1 Test'.

TABLE : Furnaces Under Erosion
Comparison images of use, as moved, current condition
Links back to original smelt reports
June 2004 November 2005 June 2007
'Hot Swap'
October 2007
'Icelandic 1'
(image by Neil Peterson)
(image by Neil Peterson)
11/05 furnace
- straight clay
- 'Blue Mountain Red' locally dug
- 10 cm + wall as fired
- fired 4 1/2 hours, (inconsistent)
- original location
- clay / sand
- ’Albany Brown’ river clay
- 7 cm wall as first fired
- 2 uses of furnace
- fired total 14 hours
- sections of wall, moved
- clay / chopped grass
- Bell Dark clay
- 7.5 cm wall as first fired
(hole from use)
- 5 uses (‘double swap’)
- fired total 21 hours
- complete, moved
- clay as ‘bricks’
- Bell Dark clay
- 6 (+) cm wall as fired
- fired 4 hours
- partial lower section only,
view to south
placed swap
image Spring 2010
mass before moving
(April 2021)
view to west
2021 swap

3 cm (left side)
5.7 cm (right side)
3.8 cm (left side)
June 2008
DD Test 3
October 2018
Slag Pit 3
October 2018
Arrow Filming
October 2020
'65 for 65'
6-08 slagpit 10-18 stone
- mix not recorded
(likely chopped straw cobb)
- thickness not recorded
- fired 3 1/3 hours
- complete, moved
- clay / sand / manure
(in metal shell)
- EPK with shredded horse
- 5 cm as fired
- 3 uses of furnace
- fired 18 hours total
- base walls and slag pit mass,
- clay / sand / manure
- EPK with shredded horse
- 7 cm as fired
- fired 6 hours,
- base and broken wall segments,
- stone block / clay fill
- gneiss / EPK mix
- variable thickness (+ 10 cm)
- 5 uses of furnace
- fired 32 3/4 hours total
- base walls and slag bowl
mass before moved
larger pieces seen were moved
image by Neil Peterson

stone mess
all remains April 2021
(April 2021)
image May 2021
stone bowl
moved - May 2021
upper seen above
5.7 cm

lower right wall
clay 5 cm + slag
total 8 cm
lower wall at tuyere
clay 3.8 cm + slag
total 6.5 cm

to far left seen above

upper wall segment
2 cm (top) / 2.7 cm (bottom)

This whole process was undertaken originally without much of a systematic approach:
In any firing, internal temperatures into the range of 1150 - 1250 C are certainly produced (to allow effective reduction of ore to creation of a metallic bloom). 
How this temperature is applied through the thickness of the clay mixture walls remains unknown. Outside surface temperatures of the furnaces are generally not measured.
Clays generally start to vitrify into ceramics starting in the 900 C + range, with the recommendation to 'bisque' fire into about 1000 C. (5) Depending on the specific clay (and additives) the optimum 'full firing' temperature is can be + 200 C (depending on the clay type).
In the chart above, the early use of naturally dug clays ('Blue Mountain Red' and 'Albany Brown') proved a problem, as both of these are a low fire temperature clays. There was thus excessive melting of interior walls, resulting in large amounts of clogging high silica slag. Although mainly done for reasons of reducing raw labour (placed against low cost!) the switch to the higher firing 'Bell Dark' commercial clay solved much of those problems. Durability and price being considered, our standard build has become using the 'high fire' clay EPK, which normally would be fired to the 1200 - 1350 C range, again about the same internal temperatures we have recorded.

In conversation with my experimental partner, Neil Peterson, some important considerations were raised :

- In any furnace firing, there is certainly a marked difference between the internal wall surface (to, if not above the sintering temperature) and the outside surface. The outer side of the wall is not likely to be raised high enough (plus 870 C) to generate actual sintering. The transition of temperatures through the wall thickness will also graduate upwards in a tall furnace.

- There is likely a limit to the possible penetration through the walls, even with repeated firings. It actually unlikely that those furnaces detailed above that undertook multiple firings will have a greater sintering effect than those with single uses. Most single firings last an average of 4 - 5 hours at temperature.
The one visible exception to this was seen with the furnace used for the 'hot swap' smelt. Here, one full cycle was undertaken to bloom extraction, then a second working team immediately re-filled and undertook a second full smelt sequence. This furnace ran over a 9 hour operation total, and one result was that a large hole was burned through the wall around the highest temperature area at the tuyere point.

One of the things that becomes apparent in terms of erosion over time is the effects of a freeze / thaw cycle against relative porosity of specific mixtures (where water absorbed by the ceramic freezes, greatly expands, then effectively shatters the material).
There are three main types :
- Straight Clay : seen in June 2004 / October 2007
The best effect of the use of sold clay is seen with the October 2007 build, when solid 'bricks' were cut from fully compacted clay blocks. The solid composition is appearing to let very little moisture penetrate the fully sintered ceramic, resulting in very little frost damage, even with over a decade of exposure to Ontario winters.
- Clay / Sand : (The standard used by Lee Sauder, which takes more care in construction, but does result in a very durable construction.)
- Clay / Organic Cobb (5) :  A) Copped Straw : November 2005 / June 2007 (as grass) / June 2008
                        B) Shredded Horse Manure : October 2018 / October 2018
The use of chopped straw leaves easily visible, fairly large (several mm wide) 'tunnels' through the sintered wall material. These most easily allow water to penetrate into the remaining walls, which leads to shattering of the material with the effects of frost. This penetration to expansion effect of freezing is less dramatic with the use of shredded horse manure, where the individual organic pieces are an order of magnitude smaller in length and diameter.

So, what is the point?

Section of the furnace wall after the November 2005 smelt.
Slag adhering to the inner surface, with a fragment of the ceramic tube tuyere in place
The graduation in colour, indicating the progression from an interior sintered ceramic (dark brown)
through to dried clay on the outside surfaces, is clearly seen.

In our current work directed to replicating the Viking Age furnaces excavated at Hals, Iceland, by Kevin Smith :
" Although neither standing furnace walls nor clay linings were preserved in situ, small frost-shattered fragments of fire-hardened and vitrified silty clay have been found scattered throughout the slag heap.  At least one of these preserved the edge of what appears to be a circular, vitrified opening for a bellows nozzle. These, the first ceramic furnace linings reported from Iceland, were recovered only by water-screening bulk soil samples from the slag heaps.  Although small, they are clearly remnants of once more abundant, expediently produced, fragile ceramic linings that have not withstood the intense freeze-thaw cycles characteristic of the Icelandic setting. The remnants of these ceramic linings from Háls are no more than 1-3 centimeters thick and would certainly not have allowed the furnaces to be self-standing.  Nor do sections through the furnace remnants suggest that thicker, unfired clay walls were once present.  However, slag lumps found adhering to samples of this clay-like matrix as well as to fire-reddened slabs of basalt suggest that both materials may have been used to line the furnace shafts.  " (6)

Bold added for my own emphasis

Taking reference from the various eroding furnaces above :

Furnace Firing
Starting Thickness (cm)
Current Thickness (cm)
June 2004
solid clay
10 +
not observable
November 2005
straw cobb
not observable
June 2007
grass cobb
3 - 5.7
October 2007
solid clay
June 2008
straw cobb ?
October 2018
manure cobb
October 2018
manure cobb
2 - 3.8


It is clear, that even over a 'short' time frame, that :

One question most certainly is raised by the last (highlighted) part of the quote above :
" Nor do sections through the furnace remnants suggest that thicker, unfired clay walls were once present."
I personally most certainly lack training in, or experience with, archaeological excavation technique. Saying that, is is unclear to me how easy it would be for a field examination to tell the difference between what originally had been 'plain dirt' and 'raw clay', most especially after 1000 years of weather effects.

In our own test series attempting to replicate the 'framed turf cone' system represented at Hals, one of our testing variables has been to gradually reduce the thickness of the clay liner. It was found at thickness less than 5 cm, there were significant problems with both manipulating the clay itself, and in durability of that liner, even over a single smelt cycle. At 3 cm, it became extremely difficult to get the clay to actually stick in place against the fragile dirt surface of the stacked turf pieces. (7) Although a change in the build process may help to solve the construction aspect, the serious problem of these thin clay walls failing (to destruction) still remains. (8)

One clear indication from the examples presented may be that in fact the original walls at Hals were built to a more functional 'usual' thickness. It is the combined effects of a limited sintering process, coupled with the combined effects of water and freezing cycles, that are then most likely to have resulted in the small diameter wall fragments recovered archaeologically.

End Notes:

1) Obviously, the Viking Age furnace (#23) from Skogar only remains as the below ground construction. This furnace was operated as a 'slag room' type (a chamber excavated, well below tuyere level, to contain any drained slag, rather than slag tapping). There is no clear indications of the upper structure. So what originally had been constructed was not necessarily built of stone blocks. The test furnace was completely above ground level, and operated as a slag tapping type. On complete excavation of the furnace after the last use, our experimental furnace also had a ring of stones attached around the final slag bowl (similar to that indicated in the archaeology).

2) To date of writing, there have been a total of 49 smelts undertaken here at Wareham. For these, a total of 37 individual furnaces have been constructed. (If I have my count correct!) This difference in approach is significant to that seen with most others in the Early Iron movement here in North America, where most typically the stress is on production, with furnaces built to withstand as many uses as possible.

3) I would refer interested readers to the 'gold standard' - Radomir Pleiner, 'Iron in Archaeology, the European Bloomery Smelters'. Once almost impossible to find, this cornerstone reference is now available as a free PDF download (!!)

4) It could be considered 8 in total, if the stone block build seen above could be included. That furnace has been recorded via scaled and grid photographs in May 2021, but only after a single winter's exposure. The plan is to clear off the remains, to free one of the main working spaces in the smelting area for future experiments.

5) Getting easy access to hard data on the firing dynamics for a specific clay body has turned out to be much more difficult (even in the world of the internet) than expected. I found the simple and direct commentaries provided by Leslie Milne at The Pottery Wheel extremely helpful!

6) Adding organic materials provides three effects to the wall structure. Depending on the size of the individual pieces, especially during the initial drying and shrinking phase, the organic material acts like re-bar in concrete, holding the clay material together and reducing potential cracking. Next, as the walls are first heated, remaining moisture flashes (explosively!) to expanding steam. The holes inside chopped straw especially, gives this steam some place to vent. As the wall material starts to heat beyond the roughly 250 + C range, the organic material burns away, leaving a somewhat porous texture. These air spaces create an insulation effect. (Although it should be noted that the dynamics of this effect has never been measured specifically.)

6) ' Ore, Fire, Hammer, Sickle: Iron Production in Viking Age and Early Medieval Iceland ‘ Kevin P. Smith - 2004, pg 13

7) Details of the 8 experimental tests in this series are found in 'Now with 70% Less Clay! Experiments with Viking Age, Icelandic, turf walled iron smelting furnaces', Smith, Markewitz & Peterson' (pending). This paper was prepared as part of the 2021 EAC12 Conference, with a video overview available : https://youtu.be/7Ltz5NG2BP0
The individual smelt reports are also published (marked in tan) on the full documentation web site : www.warehamforge.ca/ironsmelting

8) A separate report dealing with potential build methods for the clay walls, specifically as this relates to constructing the Hals type furnace, is under preparation.

Unless otherwise credited
Images and Text © 2021 Darrell Markewitz