Slag Balances - June 14, 2008
With math assistance by Neil Peterson

At the Heltborg symposium, metallurgist Arne Espelund stressed several times that he considered measurements of the slags produced during an iron smelt to be be critical to evaluating the process. We rarely actually do this, as the raw mechanics (and general chaos) of an experimental smelt in the field make exact measurements difficult, if not totally impossible.
That being said, I was presented with a unique opportunity with the June 14, 2008 smelt here in Wareham. This smelt was inside our standard clay cobb 'Norse short shaft' style smelter, and was a full scale test of the 'DARC Dirt 1' bog ore analog material. (go on to the Smelt Report / Experiment Data / Smelter Layout )

In brief:
Charcoal consumption - average of 10 minutes per 1.85 +/- kg
Air rate - measured at 90 - 95 KpH = 750 - 800 LpM
Ore charges - roughly 1 : 1 to charcoal (2 kg per charcoal bucket)
Total ore added - 20 kg

This then represents most of raw the input numbers. The most important in this consideration is the total ore addition. This figure needs to be modified by the water content of the ore.
The ore body used for this smelt was our own experimental 'DARC DIRT 1'. This material was developed by Gus Gissing and myself as an analog for a primary bog ore. It had been decided at the test stage to use a commercially available red iron oxide that most closely resembled the Fe2O3 created after roasting a natural ore. We chose 'Spanish Red' purchased in 25 KG bags from a local pottery supply house.
These are typically much finer powders ( listed as average 1.5 microns) than are really useful in the smelter. Others who have attempted to use oxide powders have reported that the venting of gas out of the top of the smelter tends to blow this fine material right back out again. Our solution was to mix a small amount of whole wheat flour (regular baking flour) as an organic binder. The decision to add a small amount of silica was based on the almost total lack of that component in the sample material. (Silica is required for generating the required slag bath).

DARK Dirt One:
Oxide - 80%
Silica - 10%
Flour - 10%
The dry powders were mixed roughly by hand, then enough water added to make a dough like paste. This required the addition of about 50 % by weight of water. The paste was dried before it was then broken up for addition to the smelter.
As added to smelter:
Solids - 90%
Water - 10%

This remaining water in the ore body was determined by taking a 500 gm sample and baking it in a propane gas forge at roughly 950 C (by colour) for about 10 minutes. The time used was not considered long enough to burn off a significant weight of the organic binder, but enough to remove all the water. The sample was then again weighed - the result was a net loss of 47 gms. This value is likely to vary from batch to batch, and possibly from piece to piece.

The specifications of the 'Spanish Red' source material (product number - MIOR) :
Fe203 - 81 %
SiO2 (silica) - 5 %
Al2O3 (alumina) - 2.6 %
CaO (calcium oxide) - 2.3 %
MgO (magnesium Oxide) - 2 %
LOI * ('loss on ignition' - water & C02) - 7 %

At the mixing phase (not considering water) :
Fe203 - 64.8 % (81 % of 80 %)
SiO2 (silica) - 14 % (10 % plus 5 % of 80%)
Al2O3 (alumina) - 2.1 % (2.6 % of 80 %)
CaO (calcium oxide) - 1.84 % (2.3 % of 80 %)
MgO (magnesium Oxide) - 1.6% (2 % of 80 %)

That suggests the resulting Fe total is on the low side (based on past experience). With only 64.8 % Fe2O3, that means only 45.4 % Fe available within the ore. This value is right at the bottom limit on Fe concentration normally considered suitable for our type of smelting equipment and process.In comparison, our original samples of primary bog ore (from L'Anse aux Meadows and St Lunaire) had been tested :

St Lunaire ( test by M. Burnham):
Fe203 - 64.04 % (80.7)
SiO2 (silica) - 2.24 % (2.8)
Al2O3 (alumina) - 3.35 % (4.2)
CaO (calcium oxide) - .69 % (.9)
MgO (magnesium oxide) - .08 % (.1)
Mn0 (manganese oxide) - .62 (.8)

On this sample, there was LOI of some 26 %. That suggests for easy comparison, the numbers should be really be adjusted upwards (the second figures). If considering the LOI potential in the analog (about 20 % in flour and water) the DARC Dirt One is a very close mimic of the St Lunaire material.

L'Anse aux Meadows (test by R. Hansen)
Fe203 - 89.5 %
SiO2 (silica) - 1.24 %
Al2O3 (alumina) - 2.45 %
CaO (calcium oxide) - .47%
MgO (magnesium oxide) - .05 %
Mn0 (manganese oxide) - 5.33

There is another variable to be considered when comparing total ore additions to a furnace and comparing the use of DARC Dirt One with other ore types. As the material is not roasted before use, there is a significant amount of the recorded weight contained in volatiles - in this case water and the flour binder. I would suggest we should be at least subtracting the water weight (about 10%) from our yield calculations. So we really only are putting in 90 % of the 20 kg as actual 'ore' - and of that only 81% was actually the Fe2O3.
From the working (charge) weight of ore - the total iron oxide content is only 64.8 %.

Going back to the June 14, 2008 smelt:
(Chemistry less water)
Fe2O3 = 64.8/.9 = 72% (Fe = 50.4%)
SiO2 = 14/.9 = 15.5%
Al2O3 = 2.1/.9 = 2.3%
CaO = 1.84/.9 = 2%
MgO = 1.6/.9 = 1.8%
(90 % of 20 kg) = 18 kg (removing weight of water)
(80% of 18 kg) = 14.4 kg (working weight of Spanish Red)
(81 % of 14.4 kg) = 11.7 kg total Fe203 added
(70 % of 11.7 kg) = 8.2 kg as actual iron added
Of which we got back 1.8 kg as bloom.

We normally compare dry roasted ore weight against bloom, so our 'relative' yield would be 1.8 bloom from 18 kg working ore = 10 %.
(If we wanted to spin this, we could compare Fe in against Fe out - 1.8 bloom from 8.2 available iron = 22 %)
The June 14 smelt was undertaken with the assistance of Ken Cook and Sam Falazone. It proved quite unusual, in that there was extremely little effect on the structure of the smelter itself. The image shows the inside surface of the smelter, with the area around the tuyere in the upper right.
First, there is very little erosion of the wall material. The area above the tuyere, normally sustains damage, but in this case you can see that the wall has hardly been effected. Curiously, what erosion that took place was just bellow the tuyere. The tuyere itself was also only slightly effected, and other than a slight rounding of the square tip, still remains the same distance proud of the wall surface as at the start of the smelt.
Second, very little material, either slag or partially sintered ore, can be seen remaining attached to the inner wall surface. In many past smelts, a considerable build up of these two materials can be seen.
As this was a newly constructed furnace, it proved possible to gather, sort and record the various slag types produced over the smelting process. The furnace was constructed on a relatively clean base of course sand, allowing any related debris to be easily distinguished. After the smelt, the largest pieces were gathered, and sorted by eye into types. All the remaining debris, including ash and unburned charcoal, where then passed through a 3/16 inch (about 4 mm) screen. The remaining larger pieces were then sorted by eye, using the same skills used to daily clean clinker from my forge fires. Any pieces of slag much larger than about 1.5 cm were sorted by type. (So there will be some loss of smaller particles)
A magnet was passed through the screened debris, extracting any pieces containing enough iron to allow attraction. These 'gromps' were measured separately.

Considerable slag material remained both inside and scattered around the smelter after the bottom extraction of the bloom through the tap arch. This material would have formed the bulk of the slag bowl. These pieces are also a dull medium grey, and often include some imbedded charcoal and ash. Generally the pieces are quite irregular and range around 'walnut' sized.
This smelt had a major 'self tapping' event occur late in the sequence, during the last stages of the burn down phase. This tap slag was fluid, dark olive / black and proved to have no magnetic quality.
As the bloom was extracted, a mass of hard slag remained attached. This material (we call 'mother') cools much more quickly than the iron bloom. It is also brittle, and is shattered off the bloom under the effect of hammering on a wooden stub during the initial consolidation step. Generally this material is quite dense, and a medium matte grey in colour. Some portions of it may prove magnetic (fragments broken off the bloom). Most of the pieces are walnut to fist sized. As this work station is a grassed area just to the side of the smelter platform, any fragments smaller than about 3 cm were surely not recovered.
The total weights of the collected slags:
Magnetic 'Gromps' - .5 kg
Tap slag - 3.2 kg
'Mother' - 2.3 kg
Bowl pieces - 6.6 kg
The bloom itself for this experiment is somewhat smaller than our usual. One factor here is that the weight was taken at a later step in the consolidation process than what is normally the case. (Standard practice is to make a single consolidation heat, striking off the majority of slag 'mother'. In this case, the bloom was subjected to two additional heat / hammer sequences.) The bloom reacted well to hammering, but was somewhat 'lumpy' in texture. Through spark testing, it appears to have some low carbon content (roughly equal to a mild steel).

Weight of Bloom - 1.8 kg
TOTAL INPUT (remove water, organics) - 16.2 KG (oxide 11.6 / silica 4.6)
TOTAL OUTPUT (as recovered) - 15.4 kg (slag 12.6 / bloom 1.8)

This looks pretty close to matching, assuming a reasonable amount of slag material lost in collection and the rough sorting process used.

Special thanks to Gus Gissing of Harder-Gissing Machining for doing the initial materials samples and donating the first raw materials.
Text and photography © 2008, Darrell Markewitz
The image 'Extraction' is by Neil Peterson