Vous êtes sur la page 1sur 34



Manganese and silicon are crucial constituents in steelmaking, as deoxidants, desulphurizers

and alloying elements. Silicon is the primary deoxidizer. Manganese is a milder deoxidizer
than silicon but enhances the effectiveness due to the formation of stable manganese silicates
and aluminates. It also serves as desulphurizer. Manganese is used as an alloying element in
almost all types of steel. Of particular interest is its modifying effect on the iron-carbon
system by increasing the hardenability of the steel.

In steelmaking SiMn is preferred over HCFeMn and FeSi as an additive because the liquid
silicate product that forms during deoxidation is easier to remove from the metal than the
solid SiO2 that forms when using FeSi. Standard SiMn also has lower levels of trace elements
(i.e. P, C, Al, and N) than standard HCFeMn and FeSi (Olsen, Tangstad, and Lindstad, 2007).
Figure 1 depicts the rise in consumption of SiMn and decrease in consumption of HCFeMn
over the 10-year period.

Silicomanganese is produced by carbothermic reduction of oxidic raw materials in electric

submerged arc furnaces. The same type of furnaces is used for FeMn and SiMn alloys.
Operation of the SiMn process is often more difficult than the FeMn process because higher
process temperature is needed. The size of the SiMn furnaces is usually in the range 15-40
MVA, giving 80-220 tonne of alloy per day.

Standard silicomanganese with 16.5-17.5 % Si and about 65 - 68% Mn is produced from a

blend of different manganese ore with about 28% to 45% MnO, quartzite, SiMn-remelts or
off grade, and reductants. Some production plants add HCFeMn slag and minor amounts of
MgO-containing minerals are added, e.g. dolomite [CaCO3⋅MgCO3] or olivine
[(MgO)2⋅SiO2]. The discard slag from the SiMn process normally contains 5 to 10% MnO.

To avoid misunderstandings it may be necessary to define some terms used in this report
(Tangstad, 1996):

 Slag - Oxide mixture where the manganese is present as MnO.

 Charge - Mixture of raw materials feed to the furnace.
 Flux - Material added to the charge for changing environment, e.g. basicity and
 Pre reduction zone - Part of the furnace where the flux and ore are mainly solid. This
is above the electrode tip
 Coke bed - Part of the furnace where the ore is melted in a surplus of carbonaceous
agent that is below the electrode tip. It could also be defined as the area in the furnace
where the electric current runs.
 Basicity – Is the chemical property, it is defined as the ratio basic oxides to acidic

Manganese ores normally contain unwanted elements that cannot be removed in the mining
and processing stages. Of special importance is phosphorus due to the strict demands in
respect of this element both in the FeMn and SiMn alloys. Iron, phosphorus and arsenic are
reduced more easily than manganese and will consequently go first into the metal. Their
content in the final alloy must therefore be controlled by selection of ores. The HCFeMn slag
is a very pure source of manganese because the easily reduced impurities in the ores have
been taken up by the HCFeMn metal in the preceding process step. The content of impurities,
like phosphorus, in SiMn alloys is therefore controlled, not only by the selection of
manganese ores, but also by the relative amounts of manganese ores and HCFeMn slag in the
raw material mix.

A process temperature of 1600 to 1650°C is necessary to obtain metal with sufficiently high
content of Si and discard slag with low MnO. FeMn slag has a relatively low melting
temperature (about 1250°C) compared with Mn-ores. When the Mn-ore starts melting at
around 1350°C, it will contain a mixture of a solid and a liquid phase, where the solid phase
is MnO. Further heating and reduction to 1550°C or more is necessary before the melting ore
will mix with the slag and flow freely. With a high share of Mn-ore in the mix, the surface
temperature and process temperature in the coke bed zone will be higher.
The specific power consumption for production of standard SiMn from a mixture of Mn-ores,
slag and Si-rich metallic remelts, can typically be 3900-4500 kWh/tonne metal, dependent
first of all on the amount of metallics added to the feed. The power consumption will increase
with the Si-content of the metal produced, and also with the amount of slag per tonne of
SiMn. Each additional 100 kg slag produced will consume additionally about 50 kWh electric
energy. About 100 kWh per tonne of metal and some coke will be saved if the ore fraction in
the charge is reduced to MnO by CO gas ascending from the smelt reduction zone.

Heat needed for smelting and reduction is supplied by electrical energy through the
electrodes. Manganese ores, briquettes and sinters are the main Mn sources and quartz is the
main Si-source in the alloy. Quartz is also added to adjust the properties of the final slag.
Several carbon sources are used as reductants, coke and coal are the most common and the
major source used in South Africa. SiMn is produced at temperatures around 1600 – 1650oC.
The process and reactions are described in detail by Olsen et al.

During production of manganese alloys, the submerged arc furnace can be divided into two
main reaction zones as illustrated in Figure 3. Their existences have been confirmed by
excavation of industrial furnace. In the upper part of the furnace, in the prereduction zone, the
raw materials are solid and MnO2, Mn3O4 and Mn2O3 are reduced to MnO by ascending
CO-gas. In the lower part of the furnace, in the coke bed, slag and metal are liquid and the
final reduction of MnO to Mn and in silicomanganese production, SiO2 to Si, takes place by
reaction (1)–(3).

MnOslag + C = Mnalloy + COg (Eq 1)

2SiO2 + 2C=2 Si + 2CO2 (Eq 2)

2MnO +Si = 2Mn + SiO (Eq 3)

Previous pilot experiments have shown that most of the reduction takes place at the top of the
coke bed. Knowledge about the conditions in this zone is therefore important when
production of manganese alloys is discussed. Several aspects of melting and reduction at the
top of the coke bed have been studied at Sintef and NTNU and reported in various

Ore chemical composition

The production of high-grade manganese ore is decreasing every year ( figure 3). It was
283,679 ton in 2008; 273,008 ton in 2009; 30,478 ton in 2012; and only 4,411.53 ton in 2013
. Thus the utilization of low-grade manganese ore for making ferromanganese and
silicomanganese is done due to the depletion and high cost price of this high-grade
manganese ore.
Description Mn Fe MgO Al2O3 SiO2 CaO P2O5
Briquettes 33.98 6.37 7.03 0.68 11.44 10.66 0.051
UMK Low Grade 32.94 4.89 3.82 0.13 4.06 16.54 0.009
Kudumane 37.03 4.76 2.88 0.15 4.4 15.34 0.012
Umk_28% 29.68 5.28 4.26 0.14 3.93 17.34 0.009
Sinter 42.2 7.33 7.96 0.9 8.02 14.04 0.014
Quartz 0.83 99.16

Most of the low-grade manganese ore are very low in Mn/Fe ratio. In Silicomanganese
production, the minimum requirement of Mn/Fe ratio in manganese ore is >6 for resulting
silicomanganese, containing at least 65% Mn and 16% Si contents . Thus, the beneficiation
process should be conducted on low-grade manganese ore to improve its Mn/Fe ratio.
Reduction roasting process followed by magnetic separation are commonly used for
beneficiation of lowgrade manganese ore.(Proceedings of the 1st International Process
Metallurgy Conference (IPMC 2016) AIP Conf. Proc. 1805,


Basicity is important parameter in the production of SiMn because it affects ;

 manganese recovery
 slag and metal separation and
 Fluidity of the slag

The economics of silicomanganese smelting is enhanced by minimising the loss of

manganese as metal inclusions whereby basicity plays a major role, as MnO dissolved in the
slag, and by production of metal high in silicon and low in carbon.

Slag basicity (B) is one of the main parameter in silicomanganese production. According to
et al Tangstan Figure 1: Main reactions in SAF for SiMn production the basicity 0.76-0.95
can produce silicomanganese with approximately 66% Mn and 17% Si. The optimum of slag
basicity for smelting high-grade manganese ore is 1.1 5 while for smelting medium-grade
manganese ore is 0.7. The optimum of slag basicity for smelting beneficiated low-grade
manganese ore is less studied. Thus, in this present study, the impact of ore basicity on
silicomanganese furnace efficiency from smelting the beneficiated low-grade manganese ores
& medium grade ores with Mn/Fe ratio 6.0 in submerged electric arc furnaces would be
investigated clearly.
Component %
MnO 11.5
SiO2 43
MgO 8
CaO 31.5
FeO 0.3
Al2O3 4.5
K2O 0.24
TiO2 0.08
SUM 99.12

Therefero the objective of this project is to invistigates the impact of ore basicity on the
efficiency of SiMn furnaces.The project would include invistigating:

 Melting point of ore.

 Quartz addition to maintain slag basicity.
 Calcining reactions.
 Power efficiency
 Ore consumption
 Slag to metal ratio and
 Mass balance

Melting point of ore

Results from several different works which have mostly been published earlier are presented.
Details about the experimental methods can be found this cited publication: Ringdalen, E.,
Gaal, S., Tangstad, M. et al. Metall and Materi Trans B (2010) manuscripts . The methods
used by Brynjulfsen are described briefly below. These are: determination of melting
properties by sessile drop apparatus, bulk melting experiments in induction furnace and
investigation of reaction mechanisms in pilot scale furnace.

Sessile drop experiments

When melting properties are measured in the sessile drop apparatus, a 10-50 mg ore sample
placed on a graphite substrate is heated in CO-atmosphere. In this case it was heated to 1100
C and held at this temperature for 30 minutes to complete pre-reduction, then heated further
5 oC/ min until complete melting. Changes of sample shape is recorded by video and coupled
to measured temperatures.
The melting properties are determined by visual analysis of the recorded data:

 Start of melting - when the shape of the particle has lost the sharp edges
 Final melting - when the droplet is round, with no artefacts

Bulk softening and melting

When melting properties are measured by bulk softening and melting, a 500g sample
consisting of 10-15mm ore particles mixed with coke is placed on the top of a coke bed in a
graphite crucible. The set up shown in Figure 2 is used. The crucible was heated to 1200 oC
and held at this temperature for 30 minutes to complete pre-reduction, then heated around 20
/min to the designated temperature and then cooled down. The cold crucible was filled with
epoxy and across section cut through the centre. Amount of unreduced, softened and reduced
ore in the coke bed was determined by area analysis as illustrated in Figure 5. A series of
experiments were stopped at different temperatures to determine the change in melting and
reduction with increasing temperature.

Figure 2: Set up for bulk melting experiments


Conditions at the top of the coke bed

The coke bed consists of solid coke mixed with liquid metal and slag. Heat is generated by
ohmic resistance in solid coke, and most of the electrical energy supplied to the furnace is
consumed in this zone. At the top of the coke bed initial slag is formed by melting and
reduction of the Mn-sources and dissolution of the fluxes. When their viscosity is low
enough, the materials will drain into the coke bed. Furnace excavations and pilot exper-
iments have shown that most of the reduction takes place at the top of the coke bed (see
Figure 3). Conditions here as temperature, compositions of liquid phases and rate for
dissolution of solid phases affects the degree of reduction at the top of the coke bed and the
composition of the phases in the coke bed.

The temperature at the top of the coke bed is believed to be determined by the temperature
needed for the slag to drain into the coke bed. This has not yet been shown experimentally. In
pilot experiments with SiMn production temperatures around 1100-1200oC were measured
just above the coke bed and around 1550–1650oC inside the coke bed just below the top,
varying with the type of raw materials used. The measurements have low accuracy, but
confirm the large temperature gradient at the top of the coke bed indicated by the large MnO
reduction observed industrially and in pilot experiments.
The temperature required to obtain a slag with low enough viscosity to drain down in the
coke bed, depends on the viscosity of the liquid phase and its amount of solids. The solid
particles can either be solids in equilibrium with the liquid phase or solid particles that due to
kinetics have not dissolved in the liquid phase.

Experiments in sessile drop furnace have shown that metal will drain down in the coke bed
independent of the slag and therefore not affect the conditions in this zone. Reduction of
MnO to alloy will reduce the total MnO content in the slag and amount of solid MnO-
containing phases in equilibrium with the slag. reduction path under equilibrium conditions
for a charge mixture with certain composition reaches liquidus, and the slag would be a
mixture of solid MnO particles dispersed in the liquid phase. The slag will drain down in the
coke bed when the amount of solids is sufficiently low given that the liquid phase has low
enough viscosity. The temperature where this is achieved is determined by the composition of
the charge mixture. At equilibrium, the temperature at the top of the coke bed is therefore
determined by the composition of the charge mixture. If equilibrium is not reached, the
temperature will be higher.

Figure 4: Main reactions in SAF for SiMn production (based on drawing from Tangstad)

A higher temperature at the top of the coke bed than given by the liquidus temperature of the
charge mixture can be caused by slow dissolution of fluxes in the slag phase. Another
probable cause is that the solid phases in the ore are not equilibrated during heating and that a
higher temperature than liquidus is required to form the first liquid phase. Too slow
dissolution of solid MnO in the liquid slag to obtain equilibrium ratio between the two phases
will also give a higher temperature for drainage of the slag into the coke bed. In a comparison
of liquidus temperatures for typical charge mixtures and measured temperature for complete
melting of the ores, the liquidus temperature for the charge mixture were lower than the
melting temperature of the ores.

Melting properties and especially the temperature of incipient melting of the ores, are
assumed to be of importance for the conditions at the top of the coke bed. Melting properties
of various types of manganese ores have been measured by different methods. Since melting
and reduction takes place simultaneously, both properties are included in the measurements.

Addition of dolomite increased the melting temperature while it decreased when quartz was
added. The experiments also showed that heat-treated agglomerates started to melt at lower
temperatures than the raw ores with the same chemical composition. During the heat
treatment low melting phases as tephroite (Mn2SiO4) and galaxite (MnAl2O3) had formed
and the temperature for initial melting was lowered. The difference in melting temperature
for the ores with same chemical composition shows that initial melting does not take place
under equilibrium. The temperature for initial melting can then not be calculated from the
chemical composition of the ore. It is affected by the phases present in the solid ore as well as
the size of the individual phases.

MnO is reduced from the liquid phase simultaneously with further melting and dissolution of
solid particles in the liquid slag. Reduction during melting was investigated in bulk smelting
experiments simulating the conditions at the top of the coke bed.

The Umk ore without fluxes and with quartz and with dolomite is compared. The lower initial
melting temperatures for heat-treated ores found in the sessile drop experiments were
confirmed by bulk melting tests. The temperature needed to reduce all the ore vary between
the different cases. At the temperature where MnO is reduced below a certain amount the slag
will drain into the coke bed and this is assumed to be the temperature at the top of the coke
bed. Level of MnO reduction needed for drainage is not clear: If it is assumed that the limit is
MnO< 10% the following temperatures can be found. While Umk ore will drain into the coke
bed at around 1580oC, heat-treated ores as sinter and pellets will drain into the coke bed at
around 1550oC. Addition of quartz will lower the temperature for drainage of Umk ore
to1425oC while addition of dolomite will increase it to 1600oC.

The cases with lowest softening temperatures for the ores had also the lowest reduction
temperature, but the differences between the cases were larger for softening temperature than
for reduction temperature. Based on the bulk smelting experiments, the degree of reduction at
different temperatures is calculated and modelled . The temperature for 50% reduction of the
ores found in bulk smelting experiments were compared to temperature for initial melting
measured with sessile drop. Ores that start to melt at lower temperatures are reduced and
drain into the coke bed at lower temperature. It was also found that ores that melt at a low
temperature will have a higher reduction rate than ores that melt at a higher temperature.

Industrially, it has been shown that the slag is tapped close to the liquidus composition as
illustrated in the case in Figure 5. A higher temperature is required if either the dissolution of
the solid phase in the liquid slag or reduction from the liquid slag is slow. The solid phase in
equilibrium with the liquid can in addition to MnO be Mn2SiO4, MnAl2O4 or other
depending on the composition of the ore. Dissolution rate for the solid is expected to be
affected by the phase. Analysis of industrial data and thermodynamic calculations by
Ostrovski showed that in HCFeMn production the solid phase was close to equilibrium with
the liquid phase while the liquid slag phase and alloy were not in equilibrium. If this
reduction is the slowest step, the reduction rate and the temperature for reduction of the ores
will be determined by the rate for reduction of MnO from the liquid slag.

The current investigations indicate that initial melting of the ores is slower. The correlation
between reduction temperature and temperature for initial melting indicate that formation of a
liquid phase and not the reduction of MnO from the liquid phase is rate determining step. The
difference between ore pellets and sinter indicates that this is not a result of the chemical
composition of the ore, but the phases are present. The decrease in melting temperature with
quartz addition and the increase with dolomite addition show that although the temperature
for initial melting is mainly a result of charge in composition it is also affected by the solid
phases that are present. Reactions in solid state between flux and ore can have formed new
phases that affect the melting conditions.
Figure 5: Ternary phase diagram of SiO2, MnO, CaO and MgO

Dissolution of the fluxes in the first liquid that is formed and changes in the melting
temperature due to this is another and may be more probable explanation. The decrease in
melting temperature by addition of quartz then indicate formation of a liquid phase in
amounts too low to be detected by the experimental method at temperatures lower than
reported for initial melting in the experiments. Temperature for initial melting and reduction
temperature are correlated when fluxes are added similarly as for ores. So also when fluxes
are added, formation of liquid phase seems to be the step determining the temperature for
melting and reduction. In SiMn production where quartz is added slag and alloy are
according to Ostrovski contrary to in HCFeMn production close to equilibrium.

The results shows that the temperature where solids in an mixture of solid and liquidus
reduced so the slag will drain into the coke bed, vary with the temperature for initial smelting
of the ore. It is mainly a result of the liquidus temperature of the mixture. But the temperature
will have some deviation from this and, temperature will increase with deceasing amount of
acid oxides in the ore. Addition of quartz to the charge mixture will decrease the temperature
while it is increased by addition of dolomite.
The top of the coke bed is defined as the interface between coke mixed with solid ore and
coke mixed with a liquid phase. The temperature at the top of the coke bed is therefore
assumed. The temperature in the bulk coke bed is determined by the energy density in the
coke bed, e.g. the size of the coke bed and the energy input. The temperature in the coke bed
will hence be the same or higher than the temperature at the top of the coke bed. Most of the
reduction takes place at the top of the coke bed, and the temperature here is therefore crucial
for furnace performance and products.

Pilot experiments by A. Muan and E.F. Osborne et al 1965 comparing HCFeMn

production from Comilog ore and from Comilog sinter showed that with use of low melting
raw materials as sinter, the slag will not necessarily be tapped at liquidus composition. With
the use of Comilog ore, the temperature at the top of the coke bed was high enough for CaO
to dissolve and a basicity of 0.65 was obtained in the slag. It was then reduced to a liquidus
MnO content of 33 % and the tapped slag had this MnO content. When Comilog sinter with
the lower melting temperature was used, the temperature at the top of the coke bed was too
low to dissolve CaO, the liquidus MnO content was 55% and slag with this MnO content
drained into the coke bed. The slag dissolved CaO while draining into the coke bed and a
basicity of 0.65 was reached in the tapped slag. But while draining through the coke bed
MnO in the slag was not reduced to the new liquidus composition. When low melting raw
materials were used, MnO content in the slag was 40%.

In SiMn production, SiO2 must be reduced to Si and a higher temperature is required. In the
experiments by Brynjulfsen, quartz dissolved easily in the initial slag. Analysis by Ostrovski
showed that equilibrium between slag and alloy was achieved while the slag was not in
equilibrium with quartz and Maroufi showed that mass transfer in the slag was rate
determining. A higher temperature at the top of the coke bed is in both cases assumed to be
beneficial to allow more of the SiO2 reduction to take place here. This hypothesis still needs
to be verified.

The use of raw materials with higher temperature for initial melting will give a higher
temperature at the top of the coke bed. This effect has been seen industrially in HCFeCr
production as a correlation between melting properties of ores and Si-content in alloy. Ores
with a low content of acid oxides have high melting temperatures and would from this be
expected to have an advantage in SiMn production. Heat treatment of the ores as sintering
and pelletising lowers the melting temperatures and is expected to give a disadvantage.
Addition of quartz that is needed in SiMn production will lower the melting temperature and
reduce some of the advantage of the basic ores. Experimental studies of the effect of various
ores and ore properties is needed to clarify their effect on the temperature on the top of the
coke bed and how this will affect SiMn production.

Melting properties of raw materials for Mn-alloy productions

Melting properties of industrial Mn-sources can, based on the discussion above, affect
furnace performance and product quality in the production of Mn-alloys, especially SiMn
production. Temperature for initial melting is one of the most important parameters. Melting
properties have been measured at Sintef and NTNU for several industrial Mn sources. Some
of the published results are presented in Table 1 to Table 4 to give an overview over melting
properties than can be expected for the different ores. Results from the different
investigations are presented in separate tables since there might be small differences in
measurement procedure between the different measurement campaigns. The reported
temperatures for initial melting vary from 1220oC to 1622oC. The melting properties are
presented to indicate the differences that can be expected in industrial manganese sources.
The background for the observed differences between these sources is thus not further
discussed here.

When Mn alloys are produced in SAF, most of the reduction takes place at the top of the coke
bed. The temperature in this zone is determined by the temperature where the slag has
viscosity low enough to drain into the coke bed. When no flux is added, this temperature has
been found to depend on the temperature for initial melting of the Mn sources. Addition of
quartz will decrease the temperature for initial melting while it will increase when dolomite is

Temperature for initial melting of Mn-sources is not only determined from their composition
as the liquidus temperature for the mixture. Heat-treated ores have lower melting
temperatures than the ore they are made from. The temperature for initial melting will
decrease with increasing amount of acidic oxides in the ore. For industrially used Mn-
sources, the temperature varies from below 1250oC to above 1600oC. Materials that melt at a
low temperature will have a higher reduction rate and obtain high enough reduction to drain
into the coke bed a lower temperature than materials with a higher temperature for initial
melting. Melting and reduction properties were determined by two different methods that
were correlated.

Dolomite and limestone will not dissolve in the slag at the top of the coke bed if the
temperature is too low. This will give a low basiscity and high MnO content in the slag at the
top of the coke bed. This led in pilot experiments to higher MnO content in the tapped slag.
In two pilot experiments comparing HCFeMn production from ore with sinter made from the
same ore, use of sinter gave a higher MnO content in the tapped slag.
Addition of quartz

Quartz to maintain slag basicity

Important properties of quartz

 Quartz and quartzite are the sources for silicon in the carbothermic process.
 Quartzite is mainly used for Silicomanganese and ferrosilicon metal production. This
is because the purity of quartzites is usually lower than for other types of quartz
deposits and thus is not suitable for silicon production. Nevertheless, relative pure
quartzites may be used in the silicon process as well, especially if mixed with other
 Furnaceability is a common international industrial quality term. A definition is given
by Kallfelz (2000b): A quartz is of good “furnaceability” when all its chemical and
physical criteria are such to make it an appropriate Silica raw material for the
production of first grade Silicon Metal at high rates of process performance.

The industry has defined a list of absolute requirements to the quartz raw material that are
necessary to achieve for the process to be optimized (Schei et al., 1998):

 Chemistry (e.g. Al, Ti, B, P, Fe and Ca)

 Lump size (typically 10 – 150 mm)
 Mechanical strength
 Thermal strength
 Softening properties

Chemistry and size are the most common specifications used by all ferrosilicon and silicon
metal producers. The requirements to chemistry are related to the content of impurity
elements especially elements such as Al, Ti, B, P, Fe and Ca. Generally, elements more noble
than Si (e.g. Al and Ca) end up in the product, whereas the volatile components follow the
off-gas (e.g. Schei et al., 1998). However, the reactions in the furnace are much more
complicated than that, and the distribution of the elements in the raw materials also
determines where the elements go. Some elements, especially alkalis (e.g. Na and K) may
actually lower the melting point of quartz (Seltveit, 1992).

Generally, the requirements to the raw materials are connected to the requirements to the
products of the furnace. The sizing requirements may vary for the different plants and 10-150
mm is a general requirement.

However, some producers have specifications for narrower sizing. Some silicomanganese and
silicon metal producers focus on, or measure the mechanical- and thermal strength, although
these are not included in the specifications to the supplier. Additionally, some producers
focus on the softening properties of the quartz. Further, additional requirements may be
defined by the individual producer, according to what is most optimal for the specific
operation. The mechanical properties of quartz affect the size reduction of the raw materials
during production in the mine, transport and storage before charging. The generated fine
material creates problems for the carbothermic process as it may lower the permeability of
the charge and obstruct the gas flow from the lower parts of the furnace to the upper parts
where SiO-gas reacts with the unreacted carbon in the charge to form SiC, which is an
important reaction in the furnace. Additionally, some of the SiO-gas condensates to form a
sticky mixture of SiO2 and liquid Si. Loss of SiO-gas through unwanted gas channels and
lowered Si-recovery may be the results of the low permeability charge.

Fines may be defined by two different criteria. In this context, fines are defined as material
less than 2 mm size, which is the most critical for the process. Fines <2 mm will lower the
permeability of the charge. Fines may also be defined as the material of lump size below
specifications (e.g. 10 mm). Too much fines in the quartz raw material is also a problem for
the quartz producers from an economical view and generate waste disposal problems, as the
fine material in many cases is useless for other customers as well.

The softening properties of the quartz are another side of the thermo-mechanical properties.
The softening temperature, or softening interval, is the temperature on which the quartz starts
to melt. This is lower than the melting point of quartz at 1723 °C. The softening temperature
should be as close to the melting temperature of quartz as possible to achieve the ideal
process where quartz move down to the cavity walls before it starts to melt and droplets of
molten quartz drip from the cavity wall into the cavity, where Si forming reactions take place.
Alkali-elements (and to a lesser extent alkali earths) are known to affect the melting
temperature of the quartz (Seltveit, 1992). According to e.g. Kallfelz (2000a), quartz that
starts to soften or even melt too high up in the furnace, will create a sticky mass, which will
agglomerate with other particles and become electrically conductive and alter the electric
paths in the furnace and even reduce the power of the arc.

Knowledge of the rate of dissolution of quartz into the slag is important for developing
understanding of the reduction reactions. No data on the dissolution rate and diffusivity of
silica in industrial SiMn slag were found in literature. Dolan and Johnston [3] studied
diffusion of Mn, Fe, Ca and Si in molten silicate slag (Al2O3, SiO2, CaO) over the
temperature range of 1400-1600°C using the semi-infinite capillary method. They found a
relationship between “molecular structure” of melts and diffusivity. A decrease in the average
size of the silicate anions resulted in increasing diffusivity of all species. Silicon exhibited the
lowest effective diffusivity (10-6 cm2/s) among the examined species .

Liang et al. used an isotope tracer method to measure self diffusion of calcium, aluminum,
silicon and oxygen in molten CaO-Al2O3-SiO2 at 1500 °C and 1 Gpa. According to their
results, the self diffusion coefficients of species increased with the decreasing melt viscosity.
They also found that self diffusion coefficient of silicon was the lowest (1.8-39.2 Χ 10-8
cm2/s) while calcium had the highest self diffusivity.

Keller and Schwerdtfeger also used the capillary technique to measure the tracer diffusivity
of Si in CaOSiO2 melts at 1600 °C. According to their results, the diffusivity of Si decreased
with increasing silica content. Mellberg and Lundstrom studied the transfer of Si between a
liquid Fe-Si alloy and molten CaO-Al2O3-SiO2 slags with a small amount of CaF2 at
1350°C and 1450°C using the electrochemical galvanostatic method. They found that the rate
of transfer of Si was controlled exclusively by diffusion in the slag. According to their report,
chemical diffusivity of silica in the slag increased from 10-7 cm2/s at 1350°C up to 10-6at

The relationship between the structure of melts and their transport properties was examined
by a number of investigators. Mills examined the relationships between structure of silicate
slags and their viscosity, diffusivi-ty and density; he demonstrated a good correlation
between transport properties of melts and the depolymerisation parameter in such melts.

The present paper studied dissolution rate and diffusivity of silica in industrial SiMn slag at
temperatures in the range of 1400 to 1550°C. Carbothermal reduction of silica from the
manganese ore-quartz mixture was also examined at 1600 oC in CO atmosphere.

Formation of slag

Chemical Composition and its Effect on Properties and Slag Structures

The slag is a liquid oxide melt and it can be divided in basic and acid oxides. Basicity
describes the ratio between acid and basic oxides in the slag. MgO and CaO are the most
important basic oxides. SiO2 and Al2O3 are the acid oxides in Mn rich slag. Al2O3 may also
act as basic oxide in some slags. MnO is basic but it is not used when the basicity ratio is
calculated because MnO is reducible. (Brynjulfsen, 2013) MgO and CaO are basic oxides and
have similar effect on equilibrium, liquidus composition, viscosity and electrical resistivity.
Silica and alumina as both are acid oxides will also have similar effect on slag properties.
Viscosity is a physical property of the slag. The slag structure affect the physical properties,
and in this chapter the viscosity is used as an example of how the chemical composition of
the slag affect the properties and slag structure. Figure 2.2 show that high amount of acid
oxides in the slag makes the slag very viscous.

The reason for the high viscosity is that the SiO2 makes big networks producing anions in
manganese silicate slags:
SiO2 +2O2- = SiO4- 4 (2.13)

Figure 2.3 shows how the SiO 4 – 4 anions may produce long range structures of complex

Basic oxides like CaO and MnO will break this network producing cation in slags:

CaO = O2° + C a2+ (2.14)

MnO = O2° + Mn2+ (2.15)

Equation 2.16 is one of the formulas used a lot for calculating the basicity. This formula is
mostly used for steel production.

B1 =( MgO +CaO)/SiO2

Slag compositon and slag basicity

Typical slag components

Oxide Melting Point (oC)

SiO2 1720
CaO 2600
MnO 1850
MgO 2800
FeO 1370
Al2O3 2030

From this table above it can be seen that the melting point of most of the oxides fall above
normal Silicomanganese production temperatures. The exception is of FeO which have the
melting point of 1370 oC.

With respect to slags, consideration must be given to desired components. Since SiMn
electric furnace slagline refractories are acid , then SiO2 saturation is required to minimize
chemical attack of the refractories. Additionally, saturation of this slag with this oxides is an
essential requirement for good foaming properties.

For the sake of simplicity the components in table can be divided into two categories, namely
refractory oxides and fluxing oxides. The refractory oxides are CaO and MgO, and the
addition of these oxides beyond liquidus composition results in an increase effective
viscousity of slag. The fluxing oxides consist of SiO2, Al2O3 and FeO. Addition of the
oxides will increase the fluidity of slag.

Therefore the secret to slag engineering for submerge electric furnace is to achieve balance
between refractory oxides and fluxing oxides. This balance will ultimately yield slags that are
both compatible with refractories and capable of attaining the desired slag properties. The
complexity in achieving this balance is that fluxing oxides are not equal in strength and will
have different effects on solubility of CaO and MgO in the slag. For example replacing SiO2
with Al2O3 in the slag will increase the solubility of CaO but decrease the solubility of MgO.
Increasing FeO content in the slag decreases MgO solubility, but have little effect on CaO

The concept of slag basicity is an attempt to define the balance between refractory oxide and
fluxing oxides. The most commonly used expression for slag basicity is the B2 ratio, which is
the ratio of CaO and SiO2 in the slag (%CaO/%SiO2). Other indices used to define slag
basicity are:

 B5 = (CaO + MgO)/ SiO3

 B4 = (CaO + MgO)/ (SiO3+Al2O3)
 B3 = CaO /SiO3+Al2O3

In silicomanganese production, Tangstaard et al (1999) recommended the basicity B5 of 0.7

for optimum slag properties. In steelmaking Wunsche and Simcoe et al 1986 recommended
B5 ratio of 1.8, in order to maintain optimum voltage gradient in the arc coloumn. Also in
steelmaking Selin et al 1979 and co-workers discouraged the use of basicity ratios, but used
graphical methods and a model to dertimine MgO saturation levels in the slag as afunction of
CaO and FeO content in the slag.

In this project B5 indices will be used for the determination of ore/slag basicities.

Transalloys Slag Composition and it's effect on performance

K. Bezemer “Infacon preceedings vol VII 1997” discovered that the prevailing high slag to
alloy ratios, relatively small changes in the MnO content of the slag affect the manganese
recovery significantly. It was recognised at an early stage that the slag should contain at least
7% MgO to improve slag fluidity and manganese recovery. Serpentine was used initially and
with reasonable success. However, the combined silica in this mineral did not take part in the
reduction reaction and just diluted the slag. Quartzite was still required to produce silicon and
to flux the excess lime. Several alternatives to obtain a higher MgO content in the slag
without any undesirable side effect were considered. Eventually it was decided to evaluate
the replacement of serpentine by naturally occurring magnesite.

Insert table

It was found that improved recoveries and better power consumption could be achieved and
further work was done to optimise the quantity of magnesite used in the process. In this
respect, there are several practical implications that affect the limit of the quantity of
magnesite that can be added. This limit is not only defined by the liquidus temperature and
viscosity of the slag but also by economic factors such as reductant consumed by the
carbonate present in the magnesite and the overall energy consumption. In our work, the
MgO content of the slag was limited to approximately 20%, as higher contents made the
operation unstable and had a marked adverse effect on energy requirements.

Turning to the base/acid ratio, which is here defined as (%CaO + MgO)/ %Si02 , the MnO
decreased with increasing base/acid ratio as could be expected. It was observed, however,
that replacing a proportion of the MgO by CaO to obtain a similar base/acid ratio almost
invariably yielded higher MnO contents in the slag.

The factors influencing the manganese recovery beside temperature are slag basicity
(CaO+MgO/Si02), the CaO/MgO ratio and the Ah03 content in the slag. Increasing
temperature favours the endothermic reduction reaction both thermodynamically and
kinetically. The effect of slag chemistry on manganese oxide reduction is more complex. The
manganese recovery is higher for the basic slag, due to higher MnO activity coefficients in
the basic slag. The addition of lime to the raw material mix decreases the MnO saturation
concentration and increases the MnO activity in the slag for the same MnO content. The
result is a decreasing equilibrium MnO concentration in the slag and an increasing rate of
reduction. An increase in the slag basicity above 1.1 has a less significant effect on the
manganese recovery. Much higher CaO and MgO contents result in higher slag viscosity
which impairs the reduction of manganese. Ah03 as well increases the slag viscosity; this can
slow down the reduction reaction of manganese. To keep the slag well flowable, the
concentration of Ah03 in the slag should not exceed 20%.

Phase Diagrams In this chapter ternary phase diagrams for the most relevant slag systems are
presented. Phase diagrams are good tools for looking at which phases it should be in the
system at different temperatures at equilibrium. It is possible to look at phase diagrams with
more than three compounds, but here it is focused on ternary diagrams with three species.

MnO-SiO2-CaO System (printed from Allibert et al. (1995))

Figure 2.11 shows the phase diagram for the MnO-SiO2-CaO system.

As mentioned earlier CaO and MnO have similar impact to the slag system, and Ca and Mn
have great solubility in each other, which is shown with an big area of (Ca,Mn)O phase in the
lower part of the phase diagram. Typical charges for SiMn production have about 40 % silica.
That means the total composition of the charge will be inside the 1300 °C isothermal plan
and be fully liquid above 1300 °C. The charge may also be liquid at about 1200 °C because
of the 1200 °C isotherm in this area in the phase diagram. These temperatures fit good to the
temperatures in the MnO, SiO2, Al2O3 phase diagram.


Calculating manganese alloys is not an easy task due to the number of variables involved.
The aim of the process optimisation model for FeMn/SiMn production presented in this paper
is to find the raw material mix which optimises a target function such as costs per ton of
product, energy consumed per ton of product, slag produced per ton of product, Mn recovery
etc. The model is capable to optimise any amounts of manganese bearing materials, reducing
agents, slag formers (lime/limestone, dolomite, quartz) and iron sources.

The model is designed as a targets/constraints optimisation model. Targets and constraints are
the values that have to be achieved, with more or less flexibility. These include product
specification and process parameters. The product specifications the are composition of alloy
to be produced (carbon and silicon content in case of FeMn), tapping weight, and temperature
of the melt; slag has to be ~ x% liquid and within a certain composition (slag basicity, FeO
and J\D3 content, CaO to MgO relation), tapping weight and temperature range. The process
parameters are reduction degrees (Cr, P, Pb, S), Mn distribution between metal and slag, slag
to metal ratio, evaporation factors (Ba,~ Mn. Na, Pb, Si, S), mechanical dust loss of raw
materials and electrode consumption. Optimisation is performed using the MATLAB®
Optimisation Toolbox. Criteria for optimisation are costs per ton of product, energy
consumed per ton of product, slag produced per ton of product, Mn recovery etc. The
possibility to change variables to constants and vice versa is provided by a special algorithm
which allows both the determination of the necessary raw materials input for predefined
targets/constraints and calculation of the output for the predefined input. In the first case,
calculation type named "calculate material", the calculation is carried out in the opposite
direction to the material flow, see Navigator in figure 4. As calculation input serve the
product specification

process parameters and raw materials analysis. As calculation output the weight of charged
raw materials, the weight and the composition of metal, slag, dust and off gas are obtained. In
the case of "set material" calculation type the weight and the composition of metal, slag, dust
and off gas can be calculated without defined product specifications. Here the weights and
analysis of raw materials as well as process parameters are used as calculation input.

Quartz contains chemical gangue besides the conglomerates of surface sticking mud coat and
earth materials which contributes for slag forming tendencies. Other important parameter is
its ability to withstand the high temperature crumbling called the thermal stability.
When the thermal stability is poor the quartz disintegrates before it is reduced. Quartz
typically contains 1-2% impurities such as Al203, FeO, CaO, etc. Mud and earth materials
also contribute towards slag formation. Thermal stability of quartz is very important and
quartz with poor thermal stability disintegrates in the furnace burden forming slag. The
impurities embedded in the quartzites are not accessible for reduction before the quartzite
lumps decripate or melt. In the latter case, they become dissolved in a glassy phase where
they become more and more Ferro Silicon Operation at IMFA – A Critical Analysis 375
enriched as the silica is consumed in the formation of SiO(g) and Si(l).

The metal oxide activities increases and the silica activity decrease during this stage, and the
temperature needed for further consumption of silica increases. Eventually the consumption
of silica slows down and stops, leaving some silica behind tied up in a very viscous melt that
is highly enriched in alumina. This melt tends to accumulate in the furnace unless it gets the
opportunity to react with calcium in the metal as expressed by the reaction 3 Ca +
(Al2O3)slag = 2 Al + 3 (CaO)slag This exchange reaction with calcium makes the slag
sufficiently fluid to be tapped together with the metal.[1] 2

This suggests that the SiO2 loss increased with increase of the impurities and increase in the
volume of slag


The effects of quartz addition on manganese recovery, silicon recovery and metallic yield are
given in Figure 9, 10 and 11, respectively. The addition of quartz in the charge mix was
found to be associated with decreased metallic yield. The decrease in metallic yield was due
to decreasing Mn recovery with the addition of quartz. Although, there was increase in Si
recovery but not enough to compensate the losses due to decreased Mn recovery.

Figure 8. Chemical composition obtained at different quartz addition (Heat no. 9-12 and 2)
Figure 9. Effect of quartz addition on Mn recovery
Figure 10. Effect of quartz addition on Si recovery

In a melt, reduction of manganese and silicon mainly takes place from their combined oxides
as shown below, which is rather difficult than reduction from their free oxides[8].

Presence of basic oxides such as CaO and MgO form stabler silicates and
generate free MnO available for reduction by carbon. However, with the addition of silica in
slag, MnO.SiO2 type complex is formed, which limits the activities of free MnO and hence
affect MnO reduction to Mn. Thus, an increase in silica addition resulted in increased silicon
content and decrease in manganese recovery. Addition of silica also increases the slag
liquidus, which intern lead to higher bath temperature favorable for Si recovery.

Industrial manganese ores vary significantly in chemistry and mineralogy. It is well-known

that the properties of a manganese ore have a large influence on the technology and
efficiency of the production of manganese alloys.

For the prereduction of manganese ore, the carbothermic reduction has been investigated
from various points of view. Ostrovski and Webb [8] investigated the reduction mechanism
of a siliceous manganese ore by graphite and found there were two stages of reduction. In the
first stage, iron oxide was reduced to iron and manganese oxide to MnO by CO. During the
second stage, MnO was reduced to Mn by carbon dissolved in the metallic phase. In
comparison with the two-stage mechanism, Eric and Burucu [9] observed in their study, in
which Mamatwan manganese ore was reduced by powder graphite under argon atmosphere
between 1000 and 1350˚C, the reduction went through three stages. During the first stage,
higher manganese and iron oxides were reduced to MnO and FeO, respectively. In the second
stage, metallic nuclei rich in iron were formed as randomly distributed small globules. The
third stage was the reduction of remaining MnO and FeO by carbon dissolved in the carbide

Hansen et al. [10] studied the reducibility of three different manganese ores (siliceous,
carbonate, oxide) with coke, and found that they reacted differently upon heating. Samples
that were essentially manganese oxides readily reacted to form the lower oxides at lower
temperature (700 ~ 800˚C). The ores that were primarily carbonates reduced to the divalent
oxides at higher temperature (1150 ~ 1200˚C). Manganese silicate ores were least amenable
to reduction. The extent and kinetics of manganese ore reduction also strongly depend on the
ore composition.

It is reported by Yastreboff [11] that silica has a strong retarding effect on the carbothermic
reduction of manganese ore. This is in a good agreement with the result obtained by Rankin,
[12] who also found that the addition of Fe2O3 increased the rate of reduction, and the effects
of CaO, MgO, and Al2O3 were very small, while there was a marked increase in reduction
rate when K2CO3 and Na2CO3 were present. Besides,

Antonov and Chufarov [13] reported that additions of Fe2O3 and K2CO3 accelerated the
reduction of manganese ore. Dewar and See [14] investigated the influence of different
carbonaceous reducing agents on the rate of reduction of Mamatwan manganese ore in the
temperature range 850 to 1500˚C under neutral atmosphere. Although there was no big
difference in the mode of reduction with different reducing agents, both the extent and rate of
reduction were enhanced when the reactivity of reducing agents toward CO2 increased. At
the same time, it is evidently shown in the study conducted by Kor [15] that the reduction rate
of manganese oxides by coconut char is much faster than that of pure graphite at temperature
between 900 and 1200˚C.

The effect of atmosphere on the carbothermic reduction of manganese oxide and manganese
ore were studied by several researchers. Kononov R. and Ostrovski O. [16] studied the
carbothermic reduction of manganese oxides in hydrogen, helium, and argon at different
temperatures. It was found that the reduction rate of manganese oxide in hydrogen was higher
than in helium, and in helium higher than in argon. Ding W. [20]8 examined the effect of the
partial pressure of CO on the manganese ore reduction by carbon. He stated that lowering the
CO partial pressure enhanced the rate of carbothermic reduction of manganese ore. Similar
results were obtained by Skjervheim [21] and Yastreboff. [22]

In addition, Nathaniel et al. [23] conducted the reduction of manganese ore in a gas
containing methane in the temperature range 1000 ~ 1200˚C. This study found that the rate of
reduction increased with increase in the level of CH4 in the gas.

The investigations discussed above employed solid carbon as the reducing agent in almost all
cases. In comparison with solid carbon, the reduction of manganese ore by CO may result in
a faster rate even in a lower temperature, owing to the advantages of gaseous reduction over
carbothermic reduction. [20] Up to now, however, very little research has been carried out for
the gaseous reduction of manganese ore by CO, and more intensive studies on prereduction
by CO are essential to develop the effective and suitable techniques for the beneficiation of
ferruginous low-grade manganese ores.

Thermal Characteristics

Combined thermo-gravimetric (TG) and differential thermal (DT) analysis in an inert gas
atmosphere over the temperature range of 20oC to 1,200oC, as shown in Figure 1, has
identified that:

• A weight loss of 35% is attained in two distinct stages. Weight loss of 31% is achieved in
the temperature range of 415oC to 730oC. A further 4% of weight loss occurs between
730oC and 1,200oC.

• The endothermal peaks at 622oC, 645oC and 791oC indicate that three different reactions
take place during decomposition. These reactions are expected to be related to the
dissociation and calcination of various combinations of carbonates, according to the
following reactions:

MgCO3 = MgO + CO2 (1)

MnCO3 = MnO + CO2 (2)

FeCO3 = FeO + CO2 (3)

CaCO3 = CaO + CO2

The temperature for calcination of natural limestone is significantly higher than that at which
CaCO3 contained in Nsuta carbonate actually decomposes because the carbonate phase is a
solid solution of (Mn,Ca,Mg)CO3.

2.5 Loss on Ignition

The loss on ignition value for Nsuta carbonate ore is commonly given as 29% at 900oC. This
standard test is carried out in a muffle furnace without an inert gas atmosphere. Chemical

The most representative chemical analysis for the carbonate ore has been collected from a
large number of samples and is given in Table 4.

The general characteristics of the Nusta carbonate ore are that:

• % CaO is the same as or greater than the % MgO

• SiO2 :Al2O3 ratio is approximately constant

• Mn 2+ (or MnCO3) varies inversely with the SiO2


Since the decomposition of natural carbonate ore in a smelting furnace occurs in a reducing
environment, this calcination process allows the manganese to remain in its lowest oxidation
state until it is reacted with carbon [5].

The sequence of reactions is as follows:

MnCO3 = MnO + CO2 (gas) (5)

MnO + C = Mn + CO (gas) (6)

This sequence of reactions is superimposed onto a furnace schematic as shown in Figure 2.

The Boudouard reaction:

CO2 (gas) + C = 2CO (7)

is strongly endothermic and only proceeds at any meaningful rate at a temperature in excess
of 1,000oC.[10,11] Since the Nsuta carbonate decomposes in a reducing environment at a
temperature below 700oC, it would appear that there is no possibility for the CO2 gas which
is released during calcination to react with carbon in the coke, but will be discharged directly
to the stack. As a result, the reduction reactions for carbonate and oxide ores may be
compared as follows, assuming that higher oxides have been prereduced by carbon monoxide
(CO) gas:

from oxides: Mn3O4 + 4C = 3Mn + 4CO (8)

from carbonates: 3MnO + 3C = 3Mn + 3CO (9)

It can be seen that carbonate ore needs 25 percent less carbon per unit of manganese when
compared with the standard mixed oxide, Mn3O4. The lowering of the fixed carbon per unit
of manganese processed increases furnace charge resistivity and therefore enhances furnace
productivity. This unique feature of carbonate ore must be considered when making the
transition from regular oxide smelting to carbonate ore processing.
Optimum placement of the electrode tips (the point of highest temperature and chemical
activity) on or in the coke bed, (and not in or through the coke bed), will of course permit
operators to optimise the quantity of reductant carbon consumed from coke and coal
compared to electrode carbon that might inadvertently be used as a reductant The melting and
reduction behaviour of ores are defined by their chemistry, mineralogy and physical
properties, which change as the ore is heated in a reducing atmosphere in a ferroalloy

Mineralogy and geology of manganese ores have been presented in a number of publications
which are reviewed by Varentsov and Grasselly [1]. Wessels and Groote Eylandt ores were
studied by Gutzmer and Beukes [2] and Ostwald [3] correspondingly, among others.
However, the change in the ore chemistry and mineralogy upon heating has only been
examined to a limited extent. Kang et al. [4] studied equilibrium phases in the CaO-MnO-
Al2O3-SiO2 system. They also analysed experimental data presented in literature. A number
of phases were identified in this system [4]: slag (molten oxide phase), olivine, tephroite,
monoxide, rhodonite, galaxite, anorthite and others, depending on the temperature and
chemistry. Zhao et al. [5] examined equilibrium phases in the CaO-MnO-Al2O3-SiO2- K2O
system. They demonstrated that K2O strongly affected the melting properties of the oxide
system. Experimentally, these systems were analysed by equilibrating with a metal phase
(Mn [4] or Mn-Si alloy [5]) at a given temperature in an inert atmosphere and quenching.
These conditions were strongly reducing which can be referred to the smelting/reduction
conditions in the ferroalloy furnace.

The focus of this paper is on the phase composition of the ore in the temperature range 800-
1200°Cwhen the ore is exposed to the CO gas; under these conditions manganese oxides are
reduced only to MnO. The paper also studies the phases formed in this temperature range in
air, as these conditions are relevant to the sintering process. This paper analyses chemical and
phase composition of CVRD, Gabonese, Wessels and Groote Eylandt ores and their change
upon heating in different gas atmospheres. Equilibrium phases formed upon heating in air and
in a reducing atmosphere were analysed using FACTSage software. The aim of this study is
to establish the behaviour of the ore upon heating and define a temperature range in which the
ore is solid.

Wessels and Groote Eylandt ores were examined at the University of New South Wales
(UNSW), Australia; CVRD and Gabonese ores were studied at SINTEF and the Norwegian
University of Science and Technology (NTNU), Norway. And found that Manganese ores of
different origin have a different chemistry and mineralogy, which affect their melting and
reduction properties. Literature covers the mineralogy of original ores[1-3] and composition
and melting properties of slags [4,5], while changes in the phase composition of manganese
ores upon heating have not been reported. The slag liquidus temperature, TL, changes in a
complex manner with changing chemistry. For example, an increase in the silica
concentration to 10 wt% increases TL for CaO/Al2O3 ratio 0.5-1 and decreases TL when the
CaO/Al2O3 ratio is 2 [4] the effect of K2O on the liquidus temperature depends on
manganese oxide content and CaO/Al2O3 ratio [5]. Understanding the melting properties of
ore requires data on the phase composition and its change during the process of heating,
which depends on the gas atmosphere.

Analysis of changes in the phase composition of Wessels and Groote Eylandt ores at UNSW
was qualitative; it established that MnO2 was reduced to Mn2O3 and Mn3O4 when the ore
was heated in air and to MnO when it was heated in a hydrogen atmosphere. Iron oxides were
reduced by hydrogen to metallic iron. The same changes were observed in the phase
composition of CVRD and Gabonese ores at NTNU/SINTEF. Quantitative XRD analysis was
not accurate as some phases had poor crystallinity; nevertheless it gave more detailed
information on phase change in the process of heating of ores, particularly in a reducing CO

In general results of equilibrium calculations using FACTSage in a reducing atmosphere and

XRD analysis were in a good agreement. This indicates that the ore samples studied in a
reducing atmosphere were close to equilibrium. A good correlation was also observed
between the experimentally determined and calculated major phases in the ore heated in air,
although a discrepancy was observed for minor phases. This can be explained by poor
crystallinity of some phases and insufficient holding time in air to reach equilibrium, as can
also be expected during the industrial processing of manganese ores.

In the original ore, manganese naturally existed in the oxidising Mn4+ state, in the form of
pyrolusite, nsutite, bixbyite or braunite. In Groote Eylandt, CVRD and Gabonese ores which
contain K2O (0.9-1 wt% K2O in CVRD and Gabonese ores and 1.5 wt% K2O in GE ore), a
significant fraction of Mn4+ was found in the form of cryptomelane. Upon heating, Mn4+
was reduced to Mn3+ and Mn2+, with a significant effect on the ore phase composition.
Potassium oxide reacted with silica and alumina forming leucite KAlSi2O6 and/ore
kaliophilite KAlSiO4; these compounds have a high melting temperature (1693°C for leucite
[7]). When the ore was heated in a reducing atmosphere, higher manganese oxides were
reduced to MnO which with silica and other oxides formed tephroite with a relatively low
melting temperature. Potassium oxide, partially dissolved in this silicate phase, decreased its
melting temperature, forming liquid slag.

Both CVRD and Gabonese ores contained a high concentration of alumina. In a reducing
atmosphere, alumina with MnO formed galaxite, detected by XRD analysis and predicted by
the calculation of equilibrium phases using FACTSage. No liquid phase was observed or
predicted in CVRD ore heated up to 1200°C in air or reducing atmosphere. XRD analysis of
sintered CVRD ore detected tephroite phase and an unidentified phase; both of them were
molten at the sintering temperature (≈1450°C). Formation of liquid slag in ores at equilibrium
was predicted for Gabonese and GE ores in air at 1200°C. Under reducing conditions, molten
slag was predicted in Groote Eylandt and Gabonese ores at 1000-1200°C, although a major
phase was solid monoxide (predominately, MnO). A liquid phase, even a small amount, can
affect material flow in the furnace. No MnO reduction to metal phase is expected at these
The conclusion of SINTEF and the Norwegian University of Science and Technology
(NTNU), Norway. Analysis of Wessels, Groote Eylandt, CVRD and Gabonese manganese
ores revealed their diverse chemical composition and complex mineralogical structure.

The following phases in the original ore were identified by XRD and EPMA analyses at
UNSW and quantitative XRD analysis at NTNU/SINTEF.

Upon sintering in air, MnO2 was reduced to Mn2O3 and Mn3O4, sintering of manganese
ores in hydrogen reduced higher manganese oxides to MnO and iron oxides to metallic iron.
Calculated equilibrium phases using FACTsage for ores heated under a reducing gas
atmosphere were in agreement with XRD analysis. Formation of liquid slag in ores at
equilibrium was predicted for Wessels ore at 1200°C in air; Groote Eylandt ore at 1200°C in
air and 1000-1200°C in a reducing atmosphere; Gabonese ore at 1200°C in air and 1000-
1200°C in a reducing atmosphere; no liquid slag was observed in equilibrium phases in
CVRD ore.

The ore analysis was reworked to a mixture of ideal minerals based on XRD and XRF
analysis (in this case braunite, bixbyite, hausmannite, calcite, dolomite, quartz). This is done
to ensure that the different amounts of oxygen associated with manganese could be captured
for the mass balance and the correct enthalpy data could be used for the energy balance.

The slag and metal temperature was assumed to be 1550°C and the off-gas temperature
450°C (a significant amount of CO2 leaves the process compared with, for example, a
ferrochromium process). The model is implemented (as a first pass) in Microsoft Excel®.
The model inputs are the %FeO and (%CaO+%MgO)/%SiO2 ratio in the slag as well as the
%C and %Si in the metal. Furthermore, the %Mn lost to the off-gas is also an input, as well
as the energy losses (in this case assumed to be 15%). The mass balance is solved iteratively
to calculate the amount of pure carbon and SiO2 required to satisfy the specified conditions,
as well as the metal, slag, and gas compositions.

As per the approach by Broekman and Ford (2004), a correlation was drawn from 490 taps of
plant data to estimate the amount of MnO in the slag as a function of the other slag
components. This correlation does not include temperature, hence the scatter observed in the
actual fitted model, but gave reasonable results. The correlation was found to be %MnO =
46.89 – 1.03 (%CaO+%MgO) with a regression coefficient of 0.73. When data for
temperature becomes available, further refinement will be made to the model, since the
activity of MnO in the slag is a function of both temperature and basicity (Holappa and Xiao,
2004). Based on a mass balance on 410 data points for one furnace, the average loss of
manganese to the off-gas was 9.03 mass%. The 95% confidence on the mean for this figure is
1.31%. For the energy balance, enthalpy data was sourced mainly from FactSage 7.0 (FactPS
database for pure species and gas, FTOxid for the slag enthalpy, and SGTE solutions
database for the enthalpy of SiMn metal).

Nadir (2015) calculated the equilibrium slag composition as function of temperature. The
start composition for this slag can be seen from Figure 2.8. This model is based on these

 100 % of CaO, MgO and Al2O3 goes into the slag phase.
 100 % of iron oxide is reduced to metallic iron at 1200 °C.
 MnO is not reduced until 1200 °C and SiO2 is not reduced until 1400 °C.
 Only Mn and Si are distributed between the slag and the metal phase.

The unreducible oxides shows a continues increase in concentration in the slag phase with
increased temperature, and that is naturally because of reduction of MnO and SiO2. From
1200 °C to 1400 °C, only MnO is reduced which can be seen in the figure. Above 1400 °C
and until 1500 °C, both MnO and SiO2 are reduced, however the reduction rate for silica is
quite low at these temperatures and hence the MnO content in the slag is observed to
constantly decrease and silica content in the slag still increase. Above 1500 °C the reduction
rate for SiO2 also increase which can be seen from the decreasing amount of silica in the
slag. For calculating the end slag, the amount of unreducible oxides are used.