Proceedings: Tenth International Ferroalloys Congress; 1 4 February 2004

INFACON X: Transformation through Technology Cape Town, South Africa

ISBN: 0-9584663-5-1 Produced by: Document Transformation Technologies
THE BEHAVIOUR OF COKE IN SUBMERGED ARC FURNACE
SMELTING OF FERROMANGANESE


S. Kamalpour
1
and W.J. Rankin
2

1
Purdue University Calumet, Department of Chemistry and Physics; 2200 169th Street Hammond,
IN 46323-2094, USA. E-mail: sxkama@aol.com
2
CSIRO Minerals, PO Box 312, Clayton South, VIC 3169, Australia. E-mail: john.rankin@csiro.au


ABSTRACT

It has been observed that the productivity of ferromanganese furnaces, and the stability of metal and slag
compositions, vary considerably according to the type of coke used as reductant. However, the desired
properties of coke for optimum furnace performance are not well understood. In this investigation, the
behaviour of four cokes, which exhibited a wide range of performance in ferromanganese production, was
investigated by means of laboratory simulations of reducing conditions in the furnace.

The Stationary Charge in Controlled Environment (SCICE) technique was used to simulate the conditions in
the furnace shaft. Batches of raw materials, mixed in the same proportions as used in practice, were heated
at fixed rates to 900, 1100 and 1300C then cooled. The mass loss was used as a measure of the degree of
reduction. Conditions in the lower part of the furnace were simulated by means of melt-reduction
experiments in which lumps of coke were drilled out to form crucibles in which enriched ferromanganese
slags were reduced at 1500 and 1550C then cooled.

An examination of the raw materials and the cooled, reacted charges using a variety of analytical and
characterization techniques enabled the sequence of reaction and changes in morphology of the raw
materials to be followed. This enabled an understanding of the behaviour of the cokes during reduction to
be developed, which was then used to explain the observed performance of the cokes in practice. By relating
the observed behaviour of the cokes in the laboratory tests to measurable properties of the unreacted cokes
and to their observed performance in practice, it was shown that to achieve high productivity and good
furnace performance, cokes should be selected to have moderately high reactivity towards CO
2
(Boudouard
reactivity) relative to conventional iron blast furnace cokes.

1. INTRODUCTION

The reactions and changes in the burden materials during the submerged arc furnace production of high
carbon ferromanganese have been extensively investigated through both plant and laboratory scale studies.
The excavation of a frozen furnace at the Meyerton ferromanganese plant in South Africa in 1977 has
provided considerable understanding of the changes in the burden materials and the nature of the reaction
zones. Details of the excavation procedure are given by Barcza et al. [1], and studies by Koursaris [2-4] on
the excavation products revealed that the interior of the furnace could be divided into nine zones of which
three appear to be of most importance in reduction and metallurgical processing. These are: a cone of fast
descending burden around each electrode; a coke bed directly below the electrodes; and a mixed coke-slag
layer beneath the coke bed.

In some early laboratory studies of carbonaceous reduction of iron ores [5-7] a technique to simulate the
processes that occur in the shaft of the iron blast furnace was developed which became known as the
Stationary Charge In Controlled Environment (SCICE) technique. Urquhart [8] used the SCICE technique to
study the production of high-carbon ferrochromium in a submerged arc furnace and the technique was
subsequently used by Grimsley [9], Dewar and See [10] and Koursaris and See [11] to investigate the
reduction of South African Mamatwan manganese ore at relatively low temperatures using Delmas and
ISCOR cokes.
381
Examination of samples from the dig-out the Meyerton furnace [1] by Koursaris and Finn [4] revealed that
the mineralogical and morphological changes in the ore were similar to those that occurred in the SCICE
investigations [11].

The major findings of that study were that between 1300 and 1600C reduction occurred in three main
stages:
rapid reduction of higher manganese oxides to MnO and of hematite to metallic iron (in the solid state)
by carbon monoxide and formation of a primary slag consisting mainly of CaO and SiO
2
;
dissolution of MnO into the slag and reduction at the surface of ore lumps to form metallic beads; and
reduction by lump carbon in contact with molten slag.

In all previous studies of reduction of manganese ores to form ferromanganese, most attention has been paid
to the behaviour of the ore and little attention has been paid to that of the coke. Pistorius [12] recently
reviewed the role of carbon in ferro-alloy production and highlighted the importance of the kinetics of
carbon dissolution in the metal.

The objective of the present work was to investigate the behaviour of coke in ferromanganese production in
order to identify the desirable properties of coke for optimum furnace performance. The investigation was
undertaken to attempt to explain the different furnace performance observed by metallurgists at the TEMCO
plant then operated by BHP Limited (now BHP Billiton Limited) in Tasmania, Australia. For
ferromanganese production, TEMCO operates two Elkem closed top, rotating arc furnaces rated at 27,000
kVA and 29,000 kVA and an Elkem 36,000 kVA stationary closed furnace. The ferromanganese is produced
from metallurgical grade lump manganese ore from Groote Eylandt in Australias Northern Territory, sinter
produced from manganese ore fines on a Lurgi down-draught sintering machine at TEMCO, and iron ore for
grade control.

In the late 1980s, TEMCO metallurgists ran a series of campaigns using cokes from different sources and
observed a range of furnace performance.

For some cokes, the furnace was stable and productivity was high while for others productivity was low and
the furnace behaved poorly; for example,
The temperature of the top of the furnace increased by more than 300C.
The furnace off-gas was high in hydrogen and sometimes contained volatilised manganese.
The slag and alloy composition varied over a wide range.
Graphitised coke was tapped from the furnace with the slag.
Coke, electrode and power consumption rates were high.

A selection of coke samples which exhibited a range of behaviours from good to bad furnace performance
was supplied by TEMCO for the purpose of attempting to identify the characteristics of the cokes which
caused the variation in performance. The investigation involved characterising the cokes to determine their
range of properties and performing a series of experiments in the laboratory to simulate the reduction of
burden materials.

2. COKE CHARACTERISATION

The composition of the coke samples is given in Table 1. The Newcastle, Port Kembla and Whyalla cokes
are blast furnace cokes prepared at the respective steel works in Australia (the Newcastle plant has
subsequently closed). The source of the Japanese coke is not known. Use of the Japanese and Newcastle
cokes was found by TEMCO metallurgists to result in better furnace performance than use of the Port
Kembla and Whyalla cokes.

A microscopic examination revealed that the Japanese coke contained large quantities of isotropic and
anisotropic fusible materials. The anisotropic material had the widest variety of mosaic textures, ranging
from fine to medium and course grain. It was concluded that the Japanese coke was made from a blend of
low, medium and high rank coals and coke breeze. Newcastle coke contained a smaller amount of fusible
material than the Japanese coke. The fusible materials also contained a narrower variety of mosaic textures;
382
there was little isotropic material and no flow/patch mosaic texture. This indicates that the Newcastle coke
was made from a blend of low and medium rank coals with no added coke breeze. The Port Kembla and
Whyalla cokes were micro-texturally similar and both contained larger quantities of non-fusible materials
than the Newcastle and Japanese cokes. The blend from which they were produced contained a larger
quantity of medium to high rank coal as revealed by the coarser mosaic textures. The non-fusible
components in the Port Kembla, Whyalla and Newcastle cokes exhibited both isotropic and anisotropic
properties. Fewer non-fusible grains were present in the Japanese coke than the Port Kembla and Whyalla
cokes, while Newcastle coke contained fewer non-fusible grains than Japanese coke.

Table 1. Composition of Coke Samples (wt%).

Japanese Newcastle
Port
Kembla
Whyalla
Fixed Carbon 87.9 86.6 88.5 85.7
Volatile Matter 0.60 0.50 0.30 1.04
SiO
2
6.07 3.72 6.54 7.58
Al
2
O
3
3.07 2.01 4.35 3.91
CaO 0.32 0.50 0.26 0.40
Fe
2
O
3
2.20 1.24 1.14 1.37
S - 0.40 - 0.36
P 0.09 0.07 0.05 0.04
Mn 0.09 0.32 0.01 0.05
K
2
O 0.13 0.17 0.13 0.15
MgO 0.12 0.10 0.02 0.11
TOTAL 101.6 95.6 101.7 99.3

Rounded, oval and elongated pores were present in all the cokes. The rounded and oval pores were formed
by degasification during coking while the elongated pores were formed by interconnection of degasification
pores, also during coking, or originated from the original coal. The Japanese coke contained the largest
degasification pores and their walls were relatively thin. Pore size in the Whyalla coke was larger than in
Port Kembla coke but smaller than in the Newcastle and Japanese cokes. The pores in the Japanese coke
occurred predominantly in the fusible components while they were present in both the fusible and non-
fusible grains in the other three cokes.

Table 2. Average force (and Standard Deviation) required to break coke
particles under compression.

Coke Type Size
Range
(mm)
Force (N) S.D. (N)
Japanese 50-200 760 185
20-50 200 75.0
6-19 100 40.0
Newcastle 50-200 530 200
20-50 110 60.0
6-19 125 55.0
Port Kembla 50-200 940 285
20-50 210 60.0
6-19 105 40.0
Whyalla 50-200 2060 315
20-50 185 105
6-19 170 85.0

383
2.1 Strength
A Tinius Olsen hardness testing machine was used to determine the strength of coke particles. The apparatus
had two plates which were free to rotate in all directions so that the force exerted on particles was through
their flattest faces. The tests were carried out on coke samples between 6 and 200 mm in diameter. Particles
were tested individually and, for each coke size fraction, 10 measurements were made then averaged. The
values are shown in Table 2.

The magnitude of the standard deviation indicates there was a large variation in strength, which is to be
expected for non-homogenous materials of irregular shape. The averaged results show that, for each coke
type, strength increased considerably with increasing size of particles. In the size range used by TEMCO (6
19 mm) Whyalla coke was the strongest followed by Newcastle coke with the Japanese and Port Kembla
cokes having about the same strength.

2.2 Porosity
Porosity in coke is of two forms: open (pores that are interconnected and which lead to the surface of the
grain) and closed. Open porosity was measured using a mercury porosimeter. Preliminary measurements
indicated that the bulk of open porosity was in the size range 10 100 m. Closed porosity was measured
using an IBAS 2000 optical image analyser to examine polished sections of coke. The pores were
discriminated from coke on the basis of reflectivity and closed pores were defined as all pores with radius
less than 10 m. One hundred images were examined for each sample, which were processed separately then
averaged. The results are shown in Table 3.

The total porosity of all the cokes was around 40 percent, with open porosity being the larger component in
all cases. There was not much variation between the four coke types of the open and closed porosity though
the Japanese and Newcastle cokes had slightly lower open porosity and slightly higher closed porosity than
the Port Kembla and Whyalla cokes.

Table 3. Open and closed porosity of the cokes as received (vol. %).


Open Porosity (%)

Closed Porosity (%)

Total
Japanese 24.6 15.3 39.9
Newcastle 22.6 16.5 39.1
Port Kembla 27.2 12.2 39.4
Whyalla 29.3 13.4 42.6

2.3 Reactivity
The reactivity of the cokes with respect to carbon dioxide (the Boudouard reactivity),

CO
2
(g) + C(s) = 2 CO(g) (1)

was determined at 1000C on two size ranges of particles. The results are presented in Table 4. Reactivity is
defined as the volume of CO
2
(at STP) that reacts with 1.0 g of coke per second. In practice the reaction was
carried out over a period of time and the reactivity averaged over the first 30 minutes and over the first 45
minutes are given in the table. The weight loss of the coke samples after 55 minutes of reaction with CO
2
is
also given. The results show there are no significant differences between the reactivities of the cokes
determined after 30 and 45 minutes, nor are there any significant differences between the reactivities
determined on the two size fractions except for the Whyalla coke; it is slightly more reactive in the smaller
size range. Japanese coke is clearly the most reactive followed by Newcastle coke. The Port Kembla coke is
more reactive than the Whyalla coke in the smaller particles size range but is about equal in reactivity in the
smaller size range. The trends in weight loss after 55 minutes reaction are similar to the trends in reactivity.
384
Table 4. Coke Reactivity Index at 1000 C for the as received cokes.

0.6 1.2 mm Particles 1.00 3.00 mm Particles
Reactivity (cm
3
g
-1
s
-1
) Weight
loss (%)
Reactivity (cm
3
g
-1
s
-1
) Weight
loss (%)
30 min 45 min 55 min 30 min 45 min 55 min
Japanese 0.36 0.34 13.5 0.32 0.29 10.40
Newcastle 0.23 0.21 7.40 0.20 0.20 8.40
Port Kembla 0.15 0.15 4.40 0.15 0.14 3.50
Whyalla 0.18 0.19 3.90 0.08 0.08 3.10

3. EXPERIMENTAL INVESTIGATION OF REDUCTION OF SINTER BURDEN

The coke characterisation results suggest that strength and initial porosity of the cokes are not related directly
to the performance of the cokes in the furnace. However the reactivity of the cokes correlates well with the
behaviour of the cokes in the furnace, the more reactive cokes being reported to perform better than the less
reactive cokes. An experimental program was designed to assess whether coke reactivity could affect the
reduction behaviour of the manganese burden. The program involved laboratory scale SCICE tests to
examine the reduction of furnace burden up to 1300C and a series of reduction-from-melt experiments to
investigate reduction in the higher temperature mixed slag-coke zone of the furnace.

3.1 SCICE Experiments
The objective of the SCICE test is to provide conditions that simulate a batch of raw material descending
through the upper part of the furnace shaft. For a test, a batch of approximately 500 g of charge, made up of
the raw materials used by TEMCO and in the same proportion, was heated in an alumino-silicate crucible in
an induction furnace at the rate of 350C per hour to a predetermined temperature (up to 1300C), held at
temperature for one hour, then cooled under nitrogen. To obtain a reasonably homogenous charge the
particle sizes of the raw materials used in the experimental program were approximately one tenth that used
in practice at TEMCO.

The composition of the raw materials used in the tests is given in Table 5 and the charge mixture used for
standard conditions is shown in Table 6. The experimental set up is illustrated in Figure 1. Considerable
effort was devoted to developing a procedure that gave reproducible temperature profiles. The crucible, with
charge and thermocouple, was weighed before each test and again after cooling at the completion of each
test.

Table 5. Composition of raw materials used in the SCICE tests (mass percent).

Constituent Sinter Mn Ore Fe Ore Limestone
MnO 56.30 48.10 0.55 0.08
SiO
2
6.80 7.10 2.43 2.00
Al
2
O
3
4.60 4.20 0.84 0.30
CaO 0.09 0.07 0.30 52.70
MgO - 0.12 0.16 1.80
BaO 1.60 0.93 - -
Na
2
O 0.10 0.22 - 0.10
K
2
O 1.20 1.65 - 0.10
Fe 5.90 4.40 65.60 0.30
TiO
2
0.23 - 0.04 0.02
P
2
O
5
0.09 0.09 - <0.01
Cr
2
O
3
- - - -
ZnO - - - <0.01
Free H
2
O - - 1.26 -

385
At the completion of a test, the contents of the crucible were emptied onto a metal tray and sieved to separate
out the unreacted coke, which could be easily separated due to its smaller size fraction. Limestone,
identified by its white colour, was then separated by hand, and the unreacted iron ore was separated using a
magnet. Sinter and manganese lump ore were not easily distinguishable and generally both were greenish in
colour after reaction. However the porous structure of the sinter helped to distinguish it from manganese ore.
Each component recovered was weighed and representative samples of each were taken for analysis and
characterisation.

Table 6. Masses and sizes of raw materials used in the SCICE experiment.

Charge Mass (g) Wt% Size (mm)
Sinter 342.4 68.6 2.36 7.5
Mn lump ore 20.7 4.5 2.36 7.5
Limestone 32.5 6.5 2.36 7.5
Iron ore 18.0 3.6 2.36 7.5
Coke 103.0 16.8 0.6 1.2
TOTAL 516.6 100.0

3.1.1 Mass loss results
The mass loss results for the crucibles and contents due to reactions releasing volatiles, predominantly CO
2

and CO, are shown graphically in Figure 2. Coke mass loss is shown in Figure 3. As expected, the total
mass loss, and mass loss by the coke, increased with the final temperature achieved but, interestingly, there is
a significant difference in the mass loss achieved at each temperature depending on the coke type used in the
experiment. The trends for total mass loss and coke mass loss are consistent and strongly indicate that the
rate of reduction of the burden materials in the experiments was controlled by the Boudouard reaction since
the order of the degree of reduction is similar to the order of the reactivity of the coke towards CO
2
at
1000C (Table 4).



Figure 1. Schematic diagram of the crucible assembly used for the SCICE experiments.

386

Figure 2. Total mass loss of crucible contents in the
SCICE experiments.
Figure 3. Mass loss of the coke component of the
SCICE experiment.

3.1.2 Changes in limestone, iron ore, and sinter
Examination of the products from the 900C series of tests revealed that all the limestone had decomposed to
CaO and the iron ore (which was initially hematite and goethite) had partially transformed to magnetite and
wustite. At 1100C some reduction of the iron ore to metallic iron had occurred in all cases. For Japanese
coke all the iron ore had been reduced to metal and for Newcastle coke iron oxide that remained unreduced
had converted fully to wustite, whereas for the other cokes some magnetite remained. At 1300C all the iron
ore had been reduced to globules of metal for all four cokes. The metal globules typically contained about
16 wt% Mn and 4 wt% C.

The unreacted sinter consisted of Hausmanite (Mn
3
O
4
), Jacobsite (Mn,Fe)
3
O
4
, tephroite (Mn
2
SiO
4
) and some
MnO. At 900 C the Hausmanite and Jacobsite in the sinter from Japanese and Newcastle coke tests had
converted to MnO while those reacted with the other cokes had only partially converted. At 1100C, all
remaining Hausmanite and Jacobsite had converted and some metallic prills of Fe, containing 10 30 wt%
Mn, had formed. These were typically 35 50 m in diameter and the relative amount of metal was highest
for the Japanese and Newcastle cokes. At 1300C partial fusion of the sinter had occurred, the amount of
metal had increased further (though the relative amounts were higher for Japanese and Newcastle cokes), and
the metallic prills had increased in size to typically 50 100 m and contained 30 40 wt% Mn.
3.1.3 Changes in coke properties
At 900C, Japanese coke had the greatest and Port Kembla coke the least changes in micro-structure and
texture. Both the fusible and non-fusible materials of the Japanese coke showed signs of reaction whereas in
the other cokes only the fusible materials showed signs of reaction. At 1100C, the micro-structural changes
were greater than at 900C and the pores sizes were usually much larger. Serration occurred in the fusible
and non-fusible materials of the Japanese coke while in the other cokes only the lower rank non-fusible
material became more reactive. At 1300C, most of the Japanese coke had eroded to small fragments due to
serration on the pore peripheries. The non-fusible materials in the other cokes were the initial points of
reaction leading to erosion and fragmentation. The change in open porosity of the reacted cokes is shown in
Figure 4.

The reactivity towards CO
2
(at 1000C) of the four cokes reacted at 1100 and 1300C was determined and
the results are presented Table 7 as mass loss after 55 minutes reaction with CO
2
. The results show that the
cokes from the 1100C SCICE series had reactivities about 2 to 3 times higher than the as received cokes
and that the reactivities increased further in the 1300C series but tended to converge in value. This is
consistent with the results in Figure 3 which show that there was greater spread in the mass loss values of the
four varieties of coke at 1100C compared to 1300C (9 20 g compared with 23 27 g, respectively). This
would be expected if the Boudouard reaction was rate-limiting in the temperature range 900 to 1300C.
387

Figure 4. Open porosity of reacted cokes determined by mercury porosimetry.

Table 7. Reactivity, as % weight loss of coke after 55 minutes reaction with CO
2
at 1000C,
of coke samples from the SCICE experiments.

COKE AFTER REACTION WITH RAW
MATERIALS
Coke as
received*
Reacted at 1100C Reacted at 1300C
Japanese 13.5 26.4 31.6
Newcastle 7.4 18.7 37.3
Port Kembla 4.4 14.4 27.1
Whyalla 3.9 13.3 34.1
*from Table 4

3.2 Reduction-From-Melt Experiments
To simulate conditions for reduction of MnO-rich slag in the lower part of the furnace, higher temperatures
than could be achieved in the SCICE experiments are needed. An upper limit of about 1300C was imposed
on the SCICE experiments because, above that temperature, the crucibles were excessively attacked by the
slag, contaminating the slag and, at higher temperatures, failing to contain the slag. An approach was used in
which enriched slags of the composition likely to occur in the first-formed slag were reacted in hollows
created by drilling into lumps of coke. In this way the coke acted as the container for the slag as well as
being the reductant. As received coke was used for these tests because the coke recovered after the SCICE
experiments was too porous and weak.

Finely ground ferromanganese slag from TEMCO was doped with MnO
2
and Fe
2
O
3
powders (mixed with a
stoichiometric amount of powdered carbon to form MnO and FeO) and remelted under nitrogen. The
resulting slag, of composition shown in Table 8, was reground and used in the Reduction-from-Melt
experiments. In an experiment, about 10 g of the enriched slag was poured into the hole in a lump of coke
which was placed on a bed of alumina powder in a silicon carbide tray then covered with an inverted
alumino-silicate crucible. The assembly was placed in an electrically heated muffle furnace, programmed to
heat at 1500C/hour, and heated to 1500 or 1550C. It was held at that temperature for 15 minutes, for the
1500C tests, or 30 minutes, for the 1550C tests, then the furnace was switched off. The cooling rate of the
furnace at high temperatures was about 80 C/min which ensured that the crucible contents solidified in less
than 5 minutes after switching off the furnace. The reacted crucibles were sectioned, one section being
used for microscopic examination, the other for chemical analysis. A typical section of a crucible is shown
in Figure 5. An equilibrium experiment was also performed using a Port Kembla coke crucible which was
heated to 1500C and held for 12 hours.

Table 8. Composition of the enriched slag used for reduction-from-melt tests (wt%).

MnO FeO SiO
2
Al
2
O
3
CaO MgO BaO K
2
O TiO
2
P
2
O
5
S
57.4 5.1 11.0 7.6 5.9 2.5 1.7 2.8 0.28 0.13 0.35

388

Figure 5. Photograph of the products of a typical melt reduction test.

Table 9. Composition of alloys produced in the reduction-from-melt experiments (wt%).

1500C 1550C
J
a
p
a
n
e
s
e

N
e
w
c
a
s
t
l
e

P
o
r
t

K
e
m
b
l
a

W
h
y
a
l
l
a

E
q
u
i
l
i
b
r
i
u
m

J
a
p
a
n
e
s
e

N
e
w
c
a
s
t
l
e

P
o
r
t

K
e
m
b
l
a

W
h
y
a
l
l
a

Mn - - - 70.9 - - -
Fe
*
- - - - 13.9 - - -
C
*
7.1 - 7.2 - 4.7 7.2 7.4 -
Si
*
0.1 - 0.6 - 7.6 0.6 0.1 -
Mn/Fe 5.4
+
3.9
+
1.8
+
2.3
+
5.1* 5.1
+
4.1
+
3.8
+
4.0
+

*bulk chemical analysis of alloy recovered from crucible

+
average of 3 microprobe point analyses

The degree of reduction can be followed by the Mn-to-Fe mass ratio as shown in Table 9. FeO is more
easily reduced than MnO and hence the higher the Mn/Fe ratio the greater the degree of reduction achieved.
The results show that for the 1500C series, the degree of reduction follows the initial order of reactivity of
the cokes, with a range of Mn/Fe ratios of 1.8 for Port Kembla coke to 5.4 for Japanese coke. The Mn/Fe
ratio for Japanese coke is similar to that for the equilibrium test. However, for the 1550C series, the effect
of the initial reactivity of coke is small; the range of final Mn/Fe ratios being only 3.8 for Port Kembla to 5.1
for Japanese coke, and closer to the equilibrium value. The results indicate that attainment of the equilibrium
of the reaction

(MnO) + Fe = Mn + (FeO) (2)

is rapid once molten phases are formed, with Japanese coke attaining the equilibrium Mn/Fe ratio at 1500C
within a little over an hour of total reaction and the other cokes approaching the equilibrium ratio in the
1550C tests. Silicon reduction was far from the equilibrium value and this is consistent with the knowledge
that silicon equilibrium in iron making is slow to attain [13]. Note that the carbon content of the alloy from
the equilibrium test is lower than for the other tests because silicon suppresses the solubility of carbon.

4. DISCUSSION

Carbothermic reduction of MnO and FeO are endothermic reactions but MnO reduction requires more heat
than FeO reduction,

MnO(s) + C(s) = Mn(l) + CO(g); H = 276 KJ (at 1200C) (3)
389
FeO(l) + C(s) = Fe(s) + CO(g); H = 149 KJ (at 1200C) (4)

MnO reduction also requires a higher minimum temperature (where G = 0): 1400C for MnO and 720C
for FeO. In practice, the thermodynamic activities of Mn and Fe are less than unity due to alloy formation so
the actual minimum temperatures will be slightly lower. In ferromanganese production, therefore, it is
thermodynamically possible for the iron ore to be reduced to iron, the higher manganese oxides to be
reduced to MnO, and some of the MnO to be reduced to Mn dissolved in the metallic iron (at low
thermodynamic activity) in the solid state under the conditions in the shaft using the sensible heat and
reducing capacity of the gases produced in the slag-coke zone. This would leave only the reduction of the
remaining MnO to occur in the high temperature mixed slag-coke zone.

The results of the SCICE and Reduction-from-Melt experiments reveal that the rate of reduction of iron
oxide to metal and the reduction of the higher oxides of manganese to MnO are strongly controlled by the
Boudouard reaction under conditions that exist in the shaft of the submerged arc furnace at temperatures up
to about 1300C. The rate of reduction of MnO to alloy is much less controlled by the Boudouard reaction
since, at temperatures where most of the MnO is reduced (>1300C), the reactivities of the cokes,
irrespective of the initial values, tend to converge to broadly similar values (Table 7).

The height of the shaft of a submerged arc furnace is much less than that of an iron blast furnace and the
residence time of the burden in the shaft is proportionately less. Typically, reduction of oxides to metal must
be performed in a few hours compared with 6 to 8 hours in an iron blast furnace. Use of a reactive coke in
submerged arc furnace production of ferromanganese allows the iron oxide to be reduced to metal, the higher
oxides of manganese to be reduced to MnO and a small amount of the MnO to be reduced to Mn dissolved in
iron by the time the burden reaches the mixed slag-coke layer below the electrodes. The lower height of the
shaft relative to that of an iron blast furnace also means that the strength of coke is a less critical property in
submerged arc furnace smelting of ferromanganese than in iron blast furnace smelting.

It is now possible to postulate a model to explain the observed poor performance at TEMCO when the lower
reactive cokes were used. When a low reactive coke is used, less reduction of iron oxide and of the higher
oxides of manganese will occur in the shaft and a higher proportion of coke will survive and accumulate
around the electrodes resulting in a deeper coke bed. This will decrease the burden resistivity and lead to
higher temperatures and greater heat generation in the hearth. This in turn will lead to volatilisation of
manganese metal, an increased off-gas temperature, and greater reduction of silica. At the same time, power
and electrode consumption will increase and coke particles will graphitise and become less reactive. Due to
the less reactive nature of the coke and the relatively short residence time of charge materials in the hearth,
incomplete reduction of MnO will occur, leading (with the increased silica reduction) to poor alloy and slag
chemistry control, and to unreacted coke being tapped.

5. CONCLUSIONS

Previous studies on the behaviour of burden in the submerged arc furnace production of high carbon
ferromanganese have focussed on the metal-forming constituents. Little attention has been given to the
properties of coke required for good furnace performance.

The overall conclusion of this study is that coke selected for use for ferromanganese production in
submerged arc furnaces should be more reactive (as measured by the conventional Boudouard reactivity test
at 1000C) than those normally used in the iron blast furnace. Use of these cokes should result in stable
furnace operation and good productivity. The reason for this is that more reactive cokes will enable
Boudouard-controlled reactions to be completed higher in the shaft, thereby more effectively utilising the
sensible heat and reducing capacity of the gases from lower in the furnace, and enabling the final, high heat
demand reduction of MnO to Mn to be the major reaction in the high temperature mixed slag-coke zone.
Conversely, less reactive cokes may lead to poor reducing conditions and larger coke beds in the furnace,
resulting in poor furnace control, unstable alloy and slag compositions, and higher power consumption.

390
Based on an examination of the cokes studied it is further concluded that cokes with the desirable attributes
could be produced under appropriate coking conditions from a blend of low, medium and high rank coals.
This will produce cokes of intermediate strength, high porosity and high reactivity (relative to iron blast
furnace cokes).

6. ACKNOWLEDGEMENTS

This work was undertaken in the former G. K. Williams Cooperative Research Centre for Extractive
Metallurgy, a joint venture between CSIRO and the University of Melbourne. Financial support was
provided by AMIRA International, including the provision of a studentship for Saeid Kamalpour.
Appreciation is extended to Dr. Samir Ganguly of TEMCO Limited for his encouragement and technical
advice during the course of the work. Appreciation is also extended to Dr. M. Waladan of the Department of
Chemical Engineering, University of Melbourne for providing advice on the high temperature experimental
techniques. Teresa Pong and Roger Curtain, also of the Department of Chemical Engineering, assisted with
the electron microprobe and scanning electron microscope work, respectively, and Ken Carey, Nicki Agron-
Olshina and Roy Lovel of CSIRO Minerals provided assistance with mercury porosimetry, X-ray diffraction
and image analysis, respectively.

7. REFERENCES

[1] Barcza, N.A., Koursaris A., See, J.B., and Gericke, W.A., The dig out of a 75 MVA high-carbon
ferromanganese electric smelting furnace, Electric Furnace Proceedings, ISS-AIME, Detroit, 1979, pp.
19-33.
[2] Koursaris, A., Aspects of ferromanganese production in the submerged arc furnace, PhD thesis,
University of the Witwatersrand, Johannesburg, 1980.
[3] Koursaris, A., Kleyenstuber, A.S.E., and Finn, C.W.P. A mineralogical investigation of the reduction
of Mamatwan manganese ore with carbon, The Geolog. Soc. South Africa, Pub. no. 7, 1981, pp. 375-
382.
[4] Koursaris, A. and Finn, C.W.P., The reduction of Mamatwan manganese ore in a submerged-arc
furnace, Trans. Iron Steel Inst. Japan, 1985, vol 25, pp. 109-117.
[5] Klemantaski, S., A new experimental technique for studying the blast-furnace process, J. Iron and
Steel Inst., 1957, vol. 185, pp. 237-239.
[6] Johnson, A., Blast-furnace process research by the SCICE Technique, J. Iron and Steel Inst., 1968,
Vol. 185, pp. 237-239.
[7] Woodhouse, W.A.V., Stationary charge in controlled environment (SCICE). A technique for blast
furnace iron making research, Inst. Chem. Engrs., 1968, Symposium series no. 27, pp 203-207.
[8] Urquhart, R.C., Study of the production of high carbon ferrochromium in the submerged arc furnace,
PhD thesis, University of the Witwatersrand , Johannesburg, 1972.
[9] Grimsley, W.D. The study of the reduction characteristics and mechanisms of reduction of manganese
ore by solid carbon, M.Sc. thesis, University of the Witwatersrand , Johannesburg, 1977.
[10] Dewar, K. and See, J.B., The influence of carbonaceous reducing agents on the rate of reduction of
representative manganese and chromium orders, National Institute for Metallurgy, Johannesburg,
South Africa, 1978, Report No. 1968.
[11] Koursaris, A. and See, J.B., Reaction in the production of high-carbon ferromanganese from
Mamatwan ore, J. South African Inst. Min. Metall., 1979, vol 79, no. 6, pp. 149-158.
[12] Pistorius, P.C., Reductant selection in ferro-alloy production: the case for the importance of
dissolution in the metal, J. South African Inst. Min. Metall., 2002, vol 102, no. 1, pp. 33 36.
[13] Turkdogan, E.T., Blast furnace reactions, Met. Trans., 1978, vol. 9B, pp. 163-178.


391