Mineral Processing and Extractive Metallurgy

Transactions of the Institutions of Mining and Metallurgy: Section C

ISSN: 0371-9553 (Print) 1743-2855 (Online) Journal homepage: http://www.tandfonline.com/loi/ympm20

Understanding ferronickel smelting from

laterites through computational thermodynamics
modelling
D. R. Swinbourne
To cite this article: D. R. Swinbourne (2014) Understanding ferronickel smelting from laterites
through computational thermodynamics modelling, Mineral Processing and Extractive
Metallurgy, 123:3, 127-140, DOI: 10.1179/1743285514Y.0000000056
To link to this article: http://dx.doi.org/10.1179/1743285514Y.0000000056

Published online: 22 May 2014.

Submit your article to this journal

Article views: 1661

View related articles

View Crossmark data

Citing articles: 3 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=ympm20
Download by: [36.84.226.236]

Date: 05 December 2016, At: 08:14

Understanding ferronickel smelting from

laterites through computational
thermodynamics modelling
D. R. Swinbourne*
The smelting of nickel laterite ores to ferronickel alloy is unique in extractive metallurgy. It treats
feed that is very low grade with respect to the target metal and, as a result, produces much more
waste slag than valuable metal product. The energy consumption per tonne of product is
therefore high and requires sustained research and design development in an effort to improve
the economics of laterite smelting. In this work, the main characteristics of nickel laterite smelting
are reviewed, and then a simple and transparent computational thermodynamics model of the
electric furnace smelting step is developed. This model predicts the nickel grade, nickel recovery
and FeO content of the slag as functions of the iron recovery to ferronickel satisfactorily. It
correctly predicts that the carbon and silicon contents in ferronickel increase sharply at high iron
recoveries. However, in common with more sophisticated models, it incorrectly predicts the iron
recovery at which this increase is observed in practice. It is concluded that the model provides an
accessible and a satisfactorily accurate vehicle for understanding the relationships between
process variables and process outcomes during nickel laterite smelting.
Keywords: Computational thermodynamics, Smelting, Laterites, Ferronickel

Introduction
Nickel is an important metal, with its use as an alloying
element in steelmaking dominating all other uses. Most
nickel is produced as the commercially pure metal, but
approximately one-third of the worlds new nickel is
ferronickel. About 98% of all ferronickel produced is
used for stainless steelmaking, and demand for stainless
steel is forecast to continue to increase to meet the needs
of the developing nations. World annual production of
ferronickel is around 250 000 t, with the two largest
producers being the Australian company BHP Billiton
and the French company ERAMET, through its
subsidiary Societe Le Nickel (Cartman, 2010).
Most of the worlds accessible nickel reserves,
amounting to approximately 72% of the total, are oxidic
ores called laterite (Sudol, 2005). Nickel is also
recovered from sulphide ore deposits, but the proportion
produced from laterites has steadily increased until it
reached 60% in 2013 (Cluzel, 2013). Laterites are the
result of oxidation by chemical weathering, and supergene enrichment under tropical conditions, of mafic/
ultramafic rocks. They vary greatly in depth, nickel
grade, chemical composition and mineralogy (Dalvi
et al., 2004). Being surface deposits, most laterites
are cheaply mined using conventional earthmoving
School of Civil, Environmental & Chemical Engineering, RMIT University,
124 Latrobe Street, Melbourne, Vic. 3000, Australia
*Corresponding author, email drs@rmit.edu.au

2014 Institute of Materials, Minerals and Mining and The AusIMM

Published by Maney on behalf of the Institute and The AusIMM
Received 15 November 2013; accepted 4 April 2014
DOI 10.1179/1743285514Y.0000000056

machinery. The upper layers are known as limonite

and are characterised by nickel contents from 0?8 to
1?5 wt-%, high iron contents and low magnesia and
silica contents. The lower layers beside the bedrock are
called saprolite and have nickel contents from 1?8 to
3 wt-%, relatively low iron contents but high magnesia
and silica contents.
The differing mineralogy of laterites requires different
processing techniques, as a result of the form in which
nickel is found in the two types of laterite. Nickel is loosely
bound to goethite in limonitic ores but is incorporated into
the lattice of magnesium silicate minerals in saprolitic ores
(Bergman, 2003). As a result, limonites are suited for
processing by hydrometallurgical methods whereas saprolites require treatment by pyrometallurgical methods
(Cartman, 2010). Approximately 40% of all laterite mined
is processed by pyrometallurgical methods (Dalvi et al.,
2004), with about 90% of that involving the production of
ferronickel alloys and the remainder being processed to
nickel matte (Crundwell et al., 2011).
Crundwell et al. (2011) summarised the physicochemical principles and technology of smelting to ferronickel.
Laterite is mined by open cut methods, minimally
upgraded by screening to remove unaltered low-nickel
bedrock, then crushed. It contains about 35 wt-% of free
water so the ore is dried in a rotary kiln (RK), with the
product still containing approximately 1013 wt-%
water, reported as loss on ignition (LOI). Most of this
water is chemically bound within such minerals as
garnierite (Mg,Ni)3Si2O5(OH)4 and goethite FeO?OH so

Mineral Processing and Extractive

Metallurgy (Trans. Inst. Min Metall. C)

2014

VOL

123

NO

127

Swinbourne

Ferronickel smelting from laterites

1 Generalised owsheet of a laterite to ferronickel smelter

temperatures of 700 to 900uC are needed to remove it.

The dried material, with some added coal, passes to
large RKs where a natural gas, oil or coal flame heats
the material. The coal volatiles and some of the fixed
carbon partially reduce the ore. The remaining fixed
carbon acts as the reductant in the following smelting
step. The nature and reactivity of the added coal has a
dramatic effect on kiln operations (Nelson et al., 2007).
The use of unreactive coal, such as anthracite, results in
very little kiln reduction and a red coloured product,
while more reactive bituminous coals achieve the desired
partial reduction and a black coloured kiln product.
There is disagreement in the literature regarding the
extent of partial reduction achieved in kiln operations,
probably as a result of local variations in operating
practice. Crundwell et al. (2011) stated that most iron is
reduced to FeO and that the extent of reduction of the
oxides to metal is typically 25% for nickel, 20% for
cobalt and 5% for iron. Daenuwy and Dalvi (1997)
report that at the P.T. Inco smelter in Indonesia all of
the Fe2O3 in the feed is reduced to Fe3O4 and FeO, with
approximately 40% of iron present as Fe3z, 55% as
Fe2z and 5% as metallic iron. About 50% of nickel and
20% of cobalt are present in the metallic form, with the
remainder being oxides. However, Solar (2013a) disagrees with their extents of reduction to the metallic
form and believes that only 20% of nickel, 10% of cobalt
and 12% of iron are reduced to the metals.
Hot calcine is fed to an electric furnace (EF) where
virtually all of the remaining Fe3O4 is reduced to FeO and
the NiO and CoO, together with part of the FeO, are
reduced to molten ferronickel by the carbon in the
calcine. The gangue oxides form slag. Finally, the molten
ferronickel is refined to remove phosphorus and sulphur
and, if necessary, to adjust the carbon and silicon
contents in order to meet market specifications before
casting or shotting to a 1015 mm diameter product.
The flowsheet described above and shown as Fig. 1 is
commonly referred to as the RKEF process (Walker
et al., 2009) because of its use of RK and EFs.
Smelting to ferronickel achieves a high recovery of
nickel. However, energy consumption is high and this
dictates that ore grade must be relatively high and

128

electricity must be cheap if a ferronickel smelter is to be

economic (Crundwell et al., 2011).

Thermodynamics of smelting
The oxide components of the feed to the EF are mainly
NiO, FeO, SiO2 and MgO and the reductant is carbon.
The Ellingham Diagram, given in Fig. 2, shows that at
15001600uC under standard state conditions, there is a
thermodynamic driving force for the reduction of NiO
and FeO by carbon, but that SiO2 and MgO are too
stable to be significantly reduced. It is also apparent that
preferential reduction of NiO should be possible.
However, NiO is not present in the furnace at unit
activity but is dissolved in slag at low activity. The
activity of NiO in molten calcine is estimated to be
about 0?01 while in the tapped slag it will be 0?001,
based on 90% recovery of nickel. The activity of nickel
in molten ferronickel ranges from 0?2 to 0?4, i.e. not far
from unity, so the oxygen potential of the reaction
2NilzO2 g~2NiOs

(1)

is given to a close approximation by

RT ln pO2 &DG o z2RT lnaNiO

(2)

The lines representing the equilibrium oxygen potential

of the Ni/O2(g)/NiO reaction at NiO activities from 0?1 to
0?001 are also shown in Fig. 2. Activity corrections are
not made to the line representing the Fe/O2/FeO reaction,
because the activity of FeO in slag and iron in ferronickel
are not far enough from unity to affect the general
conclusions. It is now apparent that the initial driving
force for the reduction of NiO is a little more favourable
than that for FeO, but that FeO reduction is favoured
when nickel recovery is high. Nickel laterite cannot be
smelted to nickel, but to a ferronickel alloy.
The competing reductions of NiO and FeO can be
represented by the reaction
NiOzFe~NizFeO

(3)

where ( ) refers to species in the slag and [ ] to species in

the alloy. It follows from the equilibrium expression for
this reaction that

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

Swinbourne

Ferronickel smelting from laterites

ferronickel ranges from approximately 10 to 30 so that

typical nickel recoveries vary from 90 to 95%.
The standard oxygen potential required for SiO2
reduction is only a little lower than that for the C/O2(g)/
CO(g) equilibrium. Silicon will also be present in
ferronickel at very low activity so a little silica reduction
to silicon is expected at the higher extents of reduction.
Under abnormally strong reducing conditions, SiO(g),
and even a little Mg(g), can form and condense as SiO2,
and MgO, in the off-gas system (Nelson et al., 2007).
The amount of energy which must be supplied for
reduction is indicated by the standard enthalpy changes
for the two main reduction reactions at the processing
temperature of approximately 1550uC

2 Ellingham diagram, with activity corrections, for reactions during nickel laterite smelting

NiOszC~NizCO g DH o ~132:47 kJ

(7)

FeOszC~FezCO g DH o ~138:81 kJ

(8)

The thermodynamic data were taken from the

database of the HSC v7?1 software (Outotec Research
Oy, 2013). Both reactions are strongly endothermic so
the required energy input will be large, being typically
about 500 kWh/t of calcine (Warner et al., 2006).

Industry surveys
aNi =aNiO ~K aFe =aFeO

(4)

It is usual in industry to express the activity ratio of

the nickel species as a partition coefficient (Solar et al.,
2008)
PCNi ~wt-% Ni=wt-% Ni

(5)

and to express the activity of FeO in slag as proportional

to wt-%FeO since, this is the way the composition of the
slag is conventionally reported, therefore
PCNi !wt-% Fe=wt-% FeO

(6)

It follows that the recovery of nickel to the alloy will

increase as the iron content of the alloy increases and
therefore the FeO content of the slag decreases, i.e. as
the nickel grade of the alloy decreases. The value of
PCNi typically varies from approximately 175 to 250
(Solar et al., 2008) so very high nickel recoveries might
be expected. However, nickel recovery also depends on
the masses of ferronickel and slag produced and the slag
mass is always much greater than the ferronickel mass.
Solar et al. (2008) reported that the mass ratio of slag/

Bergman (2003) presented and described the flowsheets

of several nickel smelters while Warner et al. (2006)
compiled extensive operating data. This was added by
Solar et al. (2008) and included nickel, cobalt and iron
distribution ratios and furnace recoveries, as well as
product mass ratios. Furnace recovery is distinct from
overall smelter recovery, which includes losses in all unit
operations and unaccounted losses. Table 1 contains
some important parameters on seven selected smelters.
Several ferronickel smelters have commenced operations
since 2006, but their operational data are yet to be made
public.
The quality of laterite feed is usually measured by the
Fe/Ni and SiO2/MgO ratio and the quality of the
ferronickel by its nickel grade, but Solar et al. (2008)
showed that these three parameters are not useful for
characterising smelter operations. The iron recovery to
ferronickel is a much better indicator of the reducing
conditions experienced in the EF. The first four smelters
listed in Table 1 have relatively low iron reductions and
so use mild reducing conditions, i.e. smaller proportions
of carbon reductant in the calcine feed. However, the Fe/
Ni ratio of their laterites varies widely from 7 to 29 and
the ferronickels produced vary from 15 to 38?5 wt-%Ni.
The final three smelters achieve much higher iron

Table 1 Selected data for ferronickel smelters (from Warner et al., 2006; Solar et al., 2008)
Laterite feed

1.
2.
3.
4.
5.
6.
7.

Falcondo
Codemin
Cerro Matoso
Loma de Niquel
Doniambo
Pomalaa
Pamco

Alloy Ni grade

Furnace recoveries

Fe/Ni

SiO2/MgO

wt-%Ni

Slag/alloy mass ratio

Ni%

Co%

Fe%

10.5
11.7
7.0
11.5
4.8
6.1
6.1

1.6
1.6
2.8
1.3
1.75
1.6
1.6

38.5
28
35
22.5
25
19
18.5

27.8
19.2
13.5
17.2
10
10.9
8.1

90.2
91.8
92.8
92.2
94.9
95.1
97.0

76.3
58.7
65.6
56.9
71.1
69.8
75.4

13.4
19.8
24.3
27.1
54.9
59.1
65.0

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

129

Swinbourne

Ferronickel smelting from laterites

recoveries, but it is evident that they utilise only highquality laterites and produce lower grades of ferronickel.
Solar et al. (2008) determined that low reducing
conditions producing high-grade ferronickel are economically preferable. Although such operations have
lower nickel recoveries than smelters using more
strongly reducing conditions, they are characterised by
lower power and reductant requirements, smaller off-gas
volumes, higher furnace throughputs and lower refining
costs.
It can also be seen in Table 1 that ferronickel smelting
from laterites is characterised by the production of much
more slag than alloy. This is in contrast to almost all
other smelting processes where the slag mass is less than
the metal or matte mass produced. The result is that the
slag exerts a powerful influence on furnace operations.

Composition of the calcine

The composition of the partially reduced calcine feed to
the EFs was not given by Warner et al. (2006), but can
be calculated from the composition of the dried feed to
the calcination/reduction kilns for various smelters. This
information was reported and is listed in Table 2.
The nickel and cobalt contents of laterites do not vary
much, but the Fe/Ni ratios vary considerably. The SiO2/
MgO ratio averages 1?8, although it was unusually high
for the Cerro Matoso laterite that was being smelted at
the time of the survey. The water content, given as
LOI, is typically 1013 wt-%.

Electric furnace design

Hot calcine is smelted in EFs in which the heat is
generated by the passage of large electric currents
through the arc plasma and slag between suspended
carbon electrodes. Some furnaces still operate in
resistive mode, i.e. without an arc gap. The furnace is
lined with high-quality MgO refractories and is fitted
with embedded water-cooled copper elements in the
sidewall. Calcine is charged through feed pipes in the
roof. Such a furnace typically consumes up to
4000 t day21 of hot calcine, produces up to 250 t day21
of ferronickel and 3700 t day21 of slag (Crundwell et al.,
2011).
Electric furnaces are either circular in shape, with
three electrodes positioned at the corners of an
equilateral triangle, or rectangular, with six inline
electrodes. The recently constructed Societe du Nickel
de Nouvelle-Caledonie et Coree Pty Ltd (SNNC) and
Jusikhoesa Poseuko (POSCO) joint venture smelter in
Korea has a state-of-the-art circular furnace which is
22?2 m diameter by 7?7 m tall, has 1?8 m diameter

electrodes with a power rating of 94 MW and is

equipped with 27 feed pipes (Rodd et al., 2010).
Rectangular furnaces are less common, although may
become more popular as furnace power levels increase.
Having six electrodes, they generate half as much
reaction gas per electrode as the circular furnaces and
this improves feed behaviour around the electrodes. A
typical rectangular furnace, at the Falcondo smelter, is
24?3 m long, 8?8 m wide and 7?3 m high with 1?4 m
diameter electrodes and a power rating of 56 MW
(Warner et al., 2006).
Earlier electric arc furnaces operated in the immersed
electrode mode, i.e. the electrodes dipped into the slag
and all heat was generated by Joule heating of the slag.
Heat was transferred from the hot slag to the overlying
calcine (Crundwell et al., 2011). Higher throughputs
require higher power input. Increased power levels have
the benefits of a decreasing proportion of the input
energy being lost and a significantly lower capital cost
per tonne of calcine smelted (Walker et al., 2010);
however, higher power levels with immersed electrode
smelting result in operating difficulties. Increased power
dissipation in the slag raises the slag bath temperature,
resulting in higher sidewall heat flux and refractory
wear, higher electrode consumption because of the
higher currents and the need for physically larger and
more costly transformers, busbars and electrodes
(Voermann et al., 2004).
The problems with immersed electrode smelting were
overcome during the 1960s and 1970s by the development of the shielded arc method. In this mode, the
electrodes sit above the slag and current passes by arcing
through a plasma gap, then through the slag and finally
through another plasma gap to the matching electrode
(Janzen et al., 2004). This mode of operation is shown in
Fig. 3.
The total power to the furnace, Pt, can be expressed as
the sum of the power dissipated in the arcs, Pa, and that
dissipated in the slag, Pb
Pt ~Pa zPb

(9)

Voermann et al. (2004) showed that for the simplest

case of a single-phase operation and a power factor of 1


Pa =Pb ~ Vt2 =Pt Rb {1
(10)
where Vt is the total voltage and Rb is the slag bath
resistance. It is desired that most power be dissipated for
calcine melting in the arcs, so Pa/Pb is typically about 5.
Such a ratio requires that the voltage be high. Bath

Table 2 Composition of dried feed to calcination/reduction kilns for several smelters (from Warner et al., 2006)
wt-%

1. Falcondo
2. Codemin
3. Cerro Matoso
4. Loma de Niquel
5. Doniambo
6. Pomalaa
7. Pamco
Average
SD

130

wt-%/wt-%

Ni

Co

Fe

SiO2

MgO

LOI

Fe/Ni

SiO2/MgO

1.38
1.44
2.20
1.48
2.70
2.20
2.30
1.96
0.52

0.04
0.04
0.09
0.08
0.06
0.05
0.08
0.06
0.02

14.5
16.9
15.3
17.0
13.0
13.4
14.0

37.3
36.3
44.0
35.6
42.0
38.8
37.0

24.1
22.5
16.0
27.3
24.0
23.6
23.5

12.5
9.6

11.5

11
11.5

10.5
11.7
7.0
11.5
4.8
6.1
6.1

1.5
1.6
2.8
1.3
1.8
1.6
1.6
1.8
0.5

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

Swinbourne

Ferronickel smelting from laterites

3 Cross-section of an electric arc furnace operating in shielded arc mode

ratio of arc power (Pa) to bath power (Pb), which is

achieved by manipulation of the arc length and control
of the slag depth (Crundwell et al., 2011). The hot
calcine feedrate must also be carefully controlled. If too
low then the arcs will be uncovered and the furnace roof
will overheat, while if too high then the furnace could be
filled to the roof and this severely impedes the flow of
waste gas (Janzen et al., 2004).

power only need to be sufficient to provide metal and

slag temperatures which are adequate for tapping. A
furnace operating at 75 MW uses currents ranging from
11 to 35 kA and electrode voltages over 1000 V to
provide arc lengths up to 50 cm.
Power density, i.e. the power per square metre of
hearth area, is an important factor in furnace design,
because the larger the power density, the smaller the
heat losses relative to the production rate (Rodd et al.,
2010). However, sidewall heat flux is increased so very
efficient water-cooled copper elements in the furnace
sidewalls are required (Voermann et al., 2004). The
SNNC furnace has a high power density of 270 kW m22
(Rodd et al., 2010), but Walker et al. (2009) believed that
a 120-MW furnace with a power density as high as
375 kW m22 could be successfully built and operated.
The reduction of calcine to ferronickel results in the
production of a large volume of waste gas, largely
carbon monoxide, e.g. the SNNC furnace produces up
to 85 000 m3 h22. Calcine must be charged close to the
electrodes so that it can receive heat from the arcs, so
this gas flow is also concentrated around the electrodes
(Rodd et al., 2010). This can result in fluidisation of the
calcine and interrupted feeding into the arc zone, so
calcine must be carefully sized to maintain adequate gas
permeability. If the calcine shield around the electrodes
is not maintained then lower power efficiency and
possible roof and feed pipe damage results (Nelson
et al., 2007).
A ferronickel arc furnace has several control requirements. It must smelt calcine at the required rate to
produce the target nickel grade of alloy with the
required slag and metal temperatures to allow both the
maintenance of a slag freeze layer on the refractories and
easy tapping of the slag. This requires an appropriate

Furnace products: ferronickel

The target nickel grade of the ferronickel produced is
largely a function of perceived customer preferences
(Solar et al., 2008) and can range from 17 to almost 40
wt-%Ni. Nickel grade is not determined by the quality of
the laterite, as expressed by the Fe/Ni ratio. Cerro
Matoso laterite had an Fe/Ni ratio of 7 and produced a
35 wt-%Ni alloy, while Pamco smelts laterite with a
similar ratio of 6?1 but to an 18?5 wt-%alloy.
The composition of ferronickel alloys produced by
several smelters is given in Table 3. Ferronickels from
low iron reduction smelters (14) are very close to binary
FeNi alloys, whereas those from high iron reduction
smelters (57) contain significant amounts of carbon,
silicon and chromium, whose concentrations are very
similar. The high silicon content makes these alloys
unsuited for direct use in stainless steelmaking, where
the silicon specification of the steels is low, so they need
to be refined to remove most of the silicon before sale
(Crundwell et al., 2011). Refining also removes most of
the carbon and chromium.
The liquidus temperature of ferronickel is strongly
affected by its composition and can be estimated from
the FeNiC phase diagram shown in Fig. 4. Alloys
from low iron reduction smelters (14) have liquidus
temperatures in the range 14501460uC but alloys from

Table 3 Ferronickel compositions and tapping temperatures (from Warner et al., 2006)
wt-%

1.
2.
3.
4.
5.
6.
7.

Falcondo
Codemin
Cerro Matoso
Loma de Niquel
Doniambo
Pomalaa
Pamco

Ni

Co

Fe

Si

Cr

Tapping temp. (uC)

38.5
28
35
23
25
19.2
18.5

0.92
0.56
0.77
0.45

0.3
0.5

60.3
71.2
63.6
78
71.3
73.7
75.5

Trace
0.02
0.01
0.26
1.5
2.4
2

Trace

0.03
0.15
2
2.45
1.5

0.03

0.05
0.06

1.7
1.5

1455
1490
1460
1550
1500
1450
1450

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

131

Swinbourne

Ferronickel smelting from laterites

4 Liquid temperatures of FeNiC alloys (Lebrun and Perrot, 2008), together with the composition of ferronickel from
selected smelters. 1: Falcondo; 2: Codemin; 3: Cerro Matoso; 4: Loma de Niquel; 5: Doniambo; 6: Pomalaa; 7: Pamco

high iron reduction smelters (57) have liquidus

temperatures from 1350uC to as low as 1250uC.
The required alloy temperature for ease of tapping
must be at least 25uC above the metal liquidus, i.e. the
metal superheat (DTsuper) must be about 25uC
(Voermann et al., 2004). However, the minimum furnace
operating temperature is set by the slag, because it
typically has a liquidus temperature above 1550uC. It
follows that the molten ferronickel is highly superheated. This has a significant impact on furnace design,
because the heat flux to the sidewall in contact with the
molten alloy is proportional to DTsuper4/3, so that there is
a need for very strong cooling of the sidewall in contact
with the molten alloy (Voermann et al., 2004). For low
iron reduction smelters, where the alloy liquidus
temperature is high, the heat flux is approximately
1200 kW m22, but for high iron reduction smelters,
where the alloy liquidus temperature is low, the heat flux
is typically 5000 kW m22. High alloy superheat does
have the benefit, however, of limiting the formation of
solidified layers in the ladle when the alloy is transferred
to refining and minimising additional heating required
during refining (Nelson et al., 2007).
Finally, the target nickel grade affects the metallurgical
performance of the smelter. Producing high-grade ferronickels involves less reduction of FeO to iron and so leads
to increased throughput, lower overall energy consumption and lower reductant consumption. However, these
benefits come at the expense of a decreased nickel recovery

in the EF (Crundwell et al., 2011). Falcondo is a highgrade ferronickel producer with a calcine throughput of
0?84 t h21 m22 at an energy consumption of 379 kWh/t
and a furnace nickel recovery of 90%, while Pomalaa is a
low-grade ferronickel producer with a throughput of
0?19 t h21 m22, a power consumption of 600 kWh/t and
a furnace nickel recovery of 95% (Warner et al., 2006).

Furnace products: slag

The major components of laterite smelting slags are
FeO, SiO2 and MgO, while there are also small amounts
of Al2O3 and CaO. A ferronickel furnace is mainly a
producer of slag, which comprises over 90% of the
furnace output. Modification of the slag composition
through the addition of fluxes would require large
amounts of flux and so is rarely economic (Utigard,
1994). It follows that the properties of the slag are
determined by the SiO2/MgO ratio of the laterite ore and
the concentration of unreduced FeO, which depends on
the Fe/Ni ratio of the laterite and the extent of reduction
achieved in the furnace. The slag compositions and
tapping temperatures for the selected smelters (Warner
et al., 2006) are summarised in Table 4.
The slag composition can be projected onto a
convenient form of the FeOSiO2MgO ternary phase
diagram to allow estimates to be made of the liquidus
temperature of the slag. Figure 5 shows that the FeO
SiO2MgO system is characterised by a low liquidus
temperature valley. On the low silica side, the liquidus

Table 4 Slag compositions and tapping temperatures (from Warner et al., 2006)
wt-%

1.
2.
3.
4.
5.
6.
7.

132

Falcondo
Codemin
Cerro Matoso
Loma de Niquel
Doniambo
Pomalaa
Pamco

SiO2

MgO

FeO

Al2O3

CaO

NiO

CoO

Tapping temp. (uC)

43.4
44.5
56
45.3
55.8
57.2
54

29.3
28.7
20
36.6
31.9
31
35

17.9
19.2
18.9
15.2
7.3
5.9
6.4

2.6
3.9
2.8
2.1
0.9
2.3
1.6

0.7
0.4
0.55

6.7
0.7

0.19
0.17
0.25
0.11
0.14
0.09
0.07

0.01

0.04
,0.02
0.02

0.02

1550
1600
1560
1650
1600
1550
1550

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

Swinbourne

5 Liquidus temperatures of SiO2MgOFeO slags in equilibrium with iron (from Levin et al., 1964). The normalised compositions for several smelters are shown. 1:
Falcondo; 2: Codemin; 3: Cerro Matoso; 4: Loma de
Niquel; 5: Doniambo; 6: Pomalaa; 7: Pamco

temperatures increase steeply with decrease in the SiO2/

MgO ratio and are almost independent of the FeO
content of the slag. On the high silica side, liquidus
temperatures are strongly dependent on the FeO content
of the slag.
It can be seen that most smelters have slags with
liquidus temperatures in the range 15501600uC and lie
close to the bottom of the valley on the phase diagram.
Loma de Niquel has an unusually high liquidus
temperature because of the low SiO2/MgO ratio of its
laterite and this could not be significantly reduced by
decreasing the extent of reduction in the furnace, i.e.
increasing the FeO content. Cerro Matoso slag had a
relatively low liquidus temperature, which would
increase rapidly if more FeO was reduced from the slag,
i.e. Cerro Matoso could not have produced low-grade
ferronickel alloy without substantially increasing the
liquidus temperature of its slag.
The presence of a small amount of Al2O3 in all slags
affects the liquidus temperature predictions. Jak and
Hayes (2010) showed that the presence of 3 wt-%Al2O3
in slag lowered the liquidus temperature by up to 50uC.
These decreased liquidus temperatures increase the slag
superheat, i.e. the difference between the furnace
operating temperature and the slag liquidus temperature. Increased slag superheat significantly increases the
difficulty of containing the slag within the furnace.
Slag superheat also has a strong effect on furnace
design. Voermann et al. (2004) showed that the rate of
heat transfer to the sidewalls is directly proportional to
the slag superheat and the power density, i.e. the power
dissipated in the bath (Pb) per unit area of bath crosssectional area. It is very important that a frozen layer of
slag be maintained on the sidewalls to protect the
refractories so furnaces with high heat fluxes require
very effective cooling, such as that provided by Hatch
waffle coolers (Hatch and Wasmund, 1974).
Slag viscosity is a very important parameter. Slag
tapping temperature must be at least 50uC above the slag
liquidus to ensure sufficient slag fluidity for good metal/
slag separation and slag tapping (Voermann et al.,
2004). Viscosity also affects the rate of slag attack on the
refractories, as does the SiO2/MgO ratio. The higher the

Ferronickel smelting from laterites

6 Viscosity of slags containing 3 wt-%Al2O3 and various

SiO2/MgO ratios at 1500uC (Jak and Hayes, 2010)

SiO2/MgO ratio, the more severe the attack on MgO

refractories (Dalvi et al., 2004). The propensity for a slag
to foam as the large volume of waste gases vents through
the slag is also a function of slag viscosity. Finally, the
ferronickel droplet residence time in the slag increases
with slag viscosity and this leads to higher mechanical
losses of ferronickel to the slag. The viscosity data of Jak
and Hayes (2010) are shown in Fig. 6, where it can be
seen that viscosity increases with decreasing FeO
content, especially for slags with high SiO2/MgO ratio.
Droplet residence time is also a function of the
interfacial tension of the slag and this decreases as the
FeO content of the slag increases, so mechanical alloy
losses to slag will also decrease with increasing FeO
content (Broadbent and Machingawuta, 1993).

Methodology
HSC simulation
HSC Chemistry for Windows v.7?1 software package
(Outotec Research Oy, 2013) is a convenient way to
determine the equilibrium state of a multicomponent
multiphase system and has been successfully used to
examine metallurgical processes such as ferromanganese
smelting (Vanderstaay et al., 2004), lead concentrates
smelting (Yamaguchi et al., 2005), the argon oxygen
decarburisation (AOD) process for stainless steelmaking
(Swinbourne and Kho, 2012) and copper flash converting (Swinbourne and Kho, 2012). The software takes the
specification of all likely phases in the system as inputs:
the species which are, or likely to be created, in each
phase; the activity coefficients of all species in each
phase; the masses of all input species and the temperature and pressure of the system. A Gibbs free energy
minimisation routine is employed to output the quantities of all species in all phases at equilibrium. As such,
it provides valuable information on the effects of process
variables on the amounts and compositions of molten
phases as a function of temperature or the amount of
one or more of the inputs.

Input amounts of species

Thermodynamic modelling of a process ideally requires
that there are reliable industrial data available for
comparison with the predicted outcomes. It was found
that the data for smelters listed in Table 1 contained

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

133

Swinbourne

Ferronickel smelting from laterites

many inconsistencies as indicated by element mass

balances that failed to close, which indicates inaccurate
data, and that no information was provided on the
composition of the partially reduced calcine entering the
EFs. As a result, representative feeds were determined
based on the data in Table 2.
It was assumed that calcine feeds contained only Ni,
NiO, CoO, FeO, Fe3O4, SiO2 and MgO, because the
Al2O3 content is too small to be significant, and the
addition of burnt lime as a flux is uncommon. Typical
calcine feeds were therefore considered to contain 2 wt% total Ni, have Fe/Ni (wt-%/wt-%) ratios of 5 and 10
and have a SiO2/MgO (wt-%/wt-%) ratio of 1?8. A small
amount of CoO was included so that the behaviour of
cobalt could be examined. Nickel metallisation was
assumed as 20% while that of cobalt was ignored
because the mass of CoO is very small and the carbon
requirement in the EF would not be significantly
affected by neglecting it. Iron oxides were assumed to
comprise 40%Fe3z and 60%Fe2z based on the data of
Daenuwy and Dalvi (1997). Rather than creating a
separate phase for the calcine feed, the oxides were
allocated to an initial slag phase and the metals were
allocated to an initial metal phase. The assumed phases,
species in each phase, initial masses of each species
(rounded to integer values) and their activity coefficients
are given in Table 5. The activity coefficient values are
justified below.

Activity coefficient values

The gas phase can be considered as an ideal solution at
high temperatures so that the activity coefficients of all
gas species were considered as unity.
Selection of appropriate values for the activity
coefficients of all species in the molten phases is the
most difficult aspect of modelling. HSC 7?1 does not
contain solution databases, so activity coefficients must
be found from the literature. Swinbourne (2011) has
argued that this is a distinct advantage of the HSC
package when used in the educational context, because it
requires users to find and critically evaluate appropriate
data themselves. Activity coefficients are a function of
the composition of the phase concerned, but HSC only
provides for constant values of the activity coefficient so
representative values for the composition ranges considered
Table 5 Representative feeds to the electric furnace
Mass of species (kg)
Phase Species Fe/Ni55
O2(g)
N2(g)
CO(g)
CO2(g)
Slag NiO
CoO
FeO
Fe3O4
SiO2(CR)
MgO
Metal Ni
Co
Fe
C
Si
Pure C
Gas

134

Fe/Ni510

20
20
1
1
50
100
85
170
540
455
300
250
4
4

0 to 40, step 0.7

Activity coefficient
1
1
1
1
3
3
1
1
1
1
0.65
1.1
1
1.3
0.003
1

7 The activity coefcient of iron and nickel in ferronickel

alloys at 1600uC (Conard et al., 1978)

likely must be selected. This limitation does not detract from

the value of the modelling for predicting trends in smelter
outcomes as a function of process variables. It is also not
essential to have precise values of activity coefficients
because small variations in value have little impact on the
predicted process outcomes.

Ferronickel
The activities of iron and nickel in ferronickel alloys
were determined by Conard et al. (1978) at 1600uC using
mass spectrometry, and they also reviewed the earlier
work of Zellars et al. (1959), Speiser et al. (1959) and
Belton and Fruehan (1967). Their results are shown in
Fig. 7. It can be seen that the activity coefficient of iron
is close to unity for alloys of relevance to laterite
smelting. There is greater disagreement regarding the
activity coefficient of nickel, but a value of 0?65 is a
reasonable estimate and would not be expected to be
significantly different at 1550uC.
Belton and Fruehan (1967) showed that the limiting
activity coefficient of cobalt in ironcobalt alloys at
1590uC is very close to 1?1.
The activity coefficients of carbon and silicon in
molten ferronickel were estimated using the dilute
solution model described by Sigworth and Elliott
(1974), using the tabulation of Lupis (1983) for the
interaction coefficients. The activity coefficients are
given by the following equation
0

ln ci ~lnci zSj eji Xj

(11)

where ci is the Raoultian activity coefficient, c0i is the

Raoultian activity coefficient at infinite dilution, eji is the
binary first-order interaction coefficient between species
i and j, and Xj is the mole fraction of species j. Table 4
shows that alloys containing high nickel contents have
very little carbon, hence only the calculated activity
coefficients of carbon (graphite) in molten iron and
20 wt-%NiFe are shown in Fig. 8. Elliott et al.s (1963)
data for FeC alloys are also shown and are in good
agreement with the calculated values. It can be seen that

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

Swinbourne

Ferronickel smelting from laterites

8 Activity coefcient of carbon in ferronickel alloys as a

function of nickel and carbon contents at 1550uC

the activity coefficient of carbon increases as the nickel

content of the alloy and the carbon content increases. A
representative activity coefficient of carbon was assumed
to be 1?3. It was also found that this value was
satisfactory over the range 14501650uC.
The SigworthElliott model (Sigworth and Elliott,
1974) was also used to calculate a value for the activity
coefficient of 0?003 for silicon. Vogel and Palme (2004)
determined the activity coefficient of silicon in carbonfree ironnickel alloys at 1300 to1400uC and their data
indicate, by extrapolation, that a value of 0?003 is
reasonable.

9 Activity coefcients of FeO(l) and SiO2(cr) in the SiO2

MgOFeO system at 1600uC

Slag
Kojima et al. (1969) determined the activity of FeO in
FeOMgOSiO2 slags at 1600uC. These data were used
to calculate the activity coefficient of FeO over the liquid
field and this is shown in Fig. 9. The value of the activity
coefficient is not very sensitive to composition at low
FeO concentrations, varying from 0?8 at the silicasaturation boundary to about 1?4 at the olivinesaturation boundary. A value of unity was considered
to be a satisfactory representation for typical ferronickel
slags.
The activity coefficient of Fe3O4 in slag was considered as unity because, under the reducing conditions
in the EF, it would not be present in the final slag. As a
result, the activity coefficient assigned to it is not
important.
Experimental activity data for SiO2 in FeOSiO2
MgO slags could not be found, so FactSage 6?3?1
software (Bale et al., 2002) using the FToxid solution
database for liquid slag and FSstel solution database for
the liquid iron were used to calculate values. The
standard state was pure solid cristobalite. The results
are also shown in Fig. 9, where it can be seen that values
vary from about 1?75 near the silica-saturation boundary to between 0?75 and 1 along the olivine-saturation
boundary. A value of unity was also considered to be a
satisfactory representation of the activity coefficient of
SiO2(cr) for typical ferronickel slags.
The activity coefficient of NiO(s) in FeOxMgOSiO2
slags at 1500uC was determined by Henao et al. (2001).
Their results are scattered between 2 and 7, but an
average value for the activity coefficient of NiO(s) of 3?5,
independent of the FeOx content of the slag, is reasonable. This is close to the value of 3 reported by Pagador
et al. (1997). In ferronickel smelting, the temperature of

the slag is 15501600uC, and an increased temperature

would shift the value of all activity coefficients towards
unity, so the activity coefficient of NiO used in this work
was assumed as 3.
Henao et al. (2001) reported data for CoO in FeOx
SiO2MgO slags containing approximately 10 wt%Al2O3. The activity coefficient of CoO(s) in such slags
was very similar to that reported for NiO(s), although
again the scatter is considerable, so the activity
coefficient of CoO(s) in Al2O3-free slags was assumed
to be the same as that of NiO(s).

Temperature
The electric arc furnace is not an isothermal reactor. The
temperature in the arc cavity is high but the residence
time of alloy/slag droplets in that region is short. Both
phases are in contact for the maximum time at the alloy/
slag interface so the temperature of this region should be
used for equilibrium calculations. The average tapping
temperature of ferronickel for the selected smelters given
in Table 3 is 1480uC, while that of the slag is 1580uC. As
a result, the appropriate temperature for modelling will
be considered as 1550uC.

Results and discussion

The nickel grade and recovery of nickel to the alloy are a
function of the amount of carbon present in calcine feed
to the EF (Crundwell et al., 2011), so the amount of
carbon in calcine will be used as the independent
variable in the modelling studies. The recoveries of
nickel, cobalt and iron are shown in Fig. 10. The curves
for nickel and cobalt are identical for both calcines, but

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

135

Swinbourne

Ferronickel smelting from laterites

10 Recoveries of nickel, cobalt and iron as a function of

carbon in calcine feed (Fe/Ni510)

136

12 Minor components of ferronickel as a function of carbon in calcine having Fe/Ni55

the iron recovery is strongly affected by the Fe/Ni ratio

because of the different amounts of iron in them.
There is no metallic phase present at very low carbon
additions, although there was metallic nickel in the feed
to the EF as a result of pre-reduction in the kiln. Nickel
initially reacts with the small quantity of Fe3O4 present
to produce FeO and NiO, the reaction being driven by
the low activity of NiO in slag. Carbon then preferentially reduces the remaining Fe3O4 to FeO. The recovery
of nickel is close to 100% at 20 kg/t of carbon, with the
cobalt recovery being about 90%. Iron recovery
increases almost linearly with the quantity of carbon in
calcine.
The composition of the ferronickel is shown in
Fig. 11. High-grade alloys containing 3540 wt-%Ni
require approximately 1520 kg of carbon per tonne of
calcine, i.e. 1?52?0 wt-% carbon in calcine. Low-grade
alloys containing 1720 wt-%Ni require approximately
2530 kg/t of calcine, i.e. about 2?53?0 wt-% carbon in
calcine. These carbon contents are in good agreement
with those used in practice (Crundwell et al., 2011).
While the Fe/Ni ratio of the calcine has a minimal
effect on the concentrations of nickel and iron in
ferronickel, it has a strong effect on the carbon and
silicon contents. The carbon and silicon contents of
ferronickel are shown in Fig. 12, together with the
cobalt content, for the calcine having an Fe/Ni ratio of
5. The cobalt concentration quickly reaches a maximum

and then decreases as more iron is reduced into the

alloy. There is a steep increase in both silicon and
carbon contents at high levels of carbon in calcine,
which explains the small decrease in iron content of the
alloy above 30 kg/t of carbon as shown in Fig. 11.
Results for the calcine having an Fe/Ni ratio of 10 are
shown in Fig. 13. The silicon content of the ferronickel
alloy is negligible for all carbon additions. The carbon
content increases steadily with increasing carbon in the
calcine, but is about 30 times less at high carbon
additions than that observed when the calcine had an
Fe/Ni ratio of 5.
These qualitative trends in composition are confirmed
by the data in Tables 1 and 3. Smelters having calcines
with Fe/Ni ratios from 5 to 7 produce low carbon
ferronickel when the nickel grade is high, as at Cerro
Matoso, but high carbon ferronickel when the nickel
grade is low, as at Pomalaa. When the Fe/Ni ratio in
calcine is above 10, the alloy is almost carbon-free for
high-grade alloys, as seen for Falcondo, and contains
only a little carbon when the nickel grade is low, e.g.
Codemin.
The smelting of calcines with low Fe/Ni ratios results
in significant carbon and silicon contents in the
ferronickel because when low-grade alloys are produced,
most of the iron is in the alloy. As a result, the FeO
content of the slag is much lower than when calcines
with high Fe/Ni ratios are smelted, as can be seen in

11 Composition of ferronickel as a function of carbon in

calcine feed

13 Minor components of ferronickel as a function of carbon in calcine feed (Fe/Ni510)

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

Swinbourne

14 Partial pressure of oxygen as a function of the carbon

in calcine, at 1550uC

Table 4. The partial pressure of oxygen in the system is a

function of the concentration of the most unstable oxide
in slag, i.e. FeO. The Fe/FeO/O2(g) equilibrium is given
by the following reaction
2FeszO2 g~2FeOl K~5:42|108 @ 15500 C (12)
Rearranging the equilibrium constant expression and
taking logarithms gives
log p2O ~{9:55z2log aFeO {2log aFe

(13)

Ferronickel alloys contain 60 to 80 wt-% iron so that

the activity of iron is not far from unity, and hence the
logarithm of the iron activity is close to zero. Assuming
that the activity of FeO is proportional to its concentration, a decrease in the FeO concentration by a factor of
10 results in a decrease in the partial pressure of oxygen
by a factor of 100 times. This decrease in partial pressure
causes some SiO2 to be reduced to silicon and carbon
dissolving into the ferronickel. These trends in partial
pressure of oxygen as a function of the carbon in calcine
are shown in Fig. 14.
Finally, the effect of temperature on the silicon and
carbon contents of ferronickel was examined for a
calcine with an Fe/Ni ratio of 5 and 30 kg/t of carbon in

Ferronickel smelting from laterites

16 Comparison of the predicted nickel grade of ferronickel with industrial data, where the Fe/Ni ratio in the
calcine is given in parentheses. 1: Falcondo; 2:
Codemin; 3: Cerro Matoso; 4: Loma de Niquel; 5:
Doniambo; 6: Pomalaa; 7: Pamco

calcine. The results are shown in Fig. 15. The carbon

content decreases with increase in temperature, but the
silicon content increases rapidly. This accounts for the
process upset known as silicon reversion, which only
occurs under very reducing conditions (Walker et al.,
2010). Ferronickel droplets formed in the arc craters are
hotter than the average slag temperature and have
silicon contents well above the equilibrium value
appropriate for alloys at that average temperature. As
these droplets pass through the slag layer, FeO in slag
reoxidises the silicon in the alloy droplets by a reaction
that is very exothermic
Silz2FeOl~2FelzSiO2 cr

(14)

DH0 (15500 C)~{433 kJ

The now superheated alloy droplets join the bulk of

the ferronickel and raise the temperature of the metal
bath. This results in increased heat transfer to the
refractories and penetration of the refractories by the
alloy, which severely compromises the refractories.
Furnace power has to be reduced, which decreases
throughput (Walker et al., 2010).

Comparisons with plant data

15 Carbon and silicon contents of ferronickel as a function of temperature for Fe/Ni55 laterite and 30 kg of
carbon per tonne of calcine feed

Comparison of the model predictions with industrial

data is not possible on the basis of the amount of carbon
in calcine, because this figure is rarely reported. Solar
et al. (2008) used the iron recovery in the ferronickel as a
measure of the extent of reduction, and this permits
useful comparisons to be made. The relationship
between nickel grade and iron recovery is shown in
Fig. 16. The agreement between the model predictions
and the industrial data, taken from Table 1, is seen to be
excellent.
The FeO content of the slag is a function of the iron
recovery, but is also affected by the SiO2/MgO ratio in
the calcine and the presence of other oxides that report
to the slag, such as Al2O3 and CaO. A comparison of the
model prediction of the FeO content of slag and
industrial data is given in Fig. 17. The agreement is
very good for the smelters operating with high recoveries
of iron, but less satisfactory for those operating with low

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

137

Swinbourne

Ferronickel smelting from laterites

17 Comparison of the predicted FeO content of slag with

industrial data, where the Fe/Ni ratio in the feed calcine is given in parentheses. 1: Falcondo; 2:
Codemin; 3: Cerro Matoso; 4: Loma de Niquel; 5:
Doniambo; 6: Pomalaa; 7: Pamco

138

19 Comparison of the predicted silicon content of ferronickel with industrial data, where the Fe/Ni ratio in the
calcine is given in parentheses. 3: Cerro Matoso; 4:
Loma de Niquel; 5:Doniambo; 6: Pomalaa; 7: Pamco

extents of iron reduction. The discrepancies do not

necessarily suggest deficiencies in the model. Mass
balances on the data provided by Warner et al. (2006)
for the seven smelters considered in this work do not all
close well, indicating errors in the data reported by the
smelters in the questionnaire sent to them for completion. These errors can be due to sampling issues.
Saprolite ores are very difficult to sample because there
is a strong correlation between particle size and
mineralogy, and therefore to composition. The fines
are richer in nickel and iron, and poorer in MgO, than
the coarse material. Plant samples are typically biased
towards the fine sizes (Solar, 2013b).
The predicted carbon and silicon contents of ferronickels are compared to plant data in Figs. 18 and 19,
where a logarithmic scale is used to expand the low
concentration range. The carbon content for Falcondo
ferronickel cannot be shown because it was given as
trace (Warner et al., 2006). Similarly, the silicon
contents for Falcondo and Codemin ferronickel cannot
be shown in Fig. 19 because they were given as 0
(Warner et al., 2006). The data for the low iron

reduction smelters are in acceptable agreement with

the predictions, but that for the high iron recovery
smelters are not. The modelling predicts that the high
carbon and silicon contents reported in ferronickels
from high iron recovery smelters should not be reached
until iron recoveries of about 90%, whereas in practice
they occur at recoveries of 5565%. This discrepancy has
been found by others using different computational
thermodynamics software. Solar et al. (2014) stated that
Unfortunately, the authors have not been successful in
reproducing the behaviour (of carbon) observed in the
field nor are they aware of any published information
that has. No explanation can be offered for these high
carbon and silicon contents at iron recoveries of only
approximately 60%. Whatever be the cause, it seems to
be common to both carbon and silicon, and is unlikely
to be thermodynamic in origin because both concentrations are little affected by the extent of iron reduction,
i.e. the partial pressure of oxygen in the furnace.
The recovery of nickel as a function of iron recovery is
shown in Fig. 20. The furnace nickel recoveries from the
smelters vary from 2 to 7% below the predicted value,
with those for the high iron reduction smelters being
closest to the predicted values. While the recovery data

18 Comparison of the predicted carbon content of ferronickel with industrial data, where the Fe/Ni ratio in the
calcine is given in parentheses. 2: Codemin; 3: Cerro
Matoso; 4: Loma de Niquel; 5: Doniambo; 6: Pomalaa;
7: Pamco

20 Comparison of nickel recovery with industrial data,

where the Fe/Ni ratio in the calcine is given in parentheses. 1: Falcondo; 2: Codemin; 3: Cerro Matoso;
4: Loma de Niquel; 5: Doniambo; 6: Pomalaa; 7:
Pamco

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

Swinbourne

from the smelters are subject to error because of

sampling difficulties, the discrepancy between predicted
and reported values is mostly likely because of mechanical losses of ferronickel as suspended droplets in slag, as
discussed in detail by Solar (2009). Another contributing
factor is disequilibrium between slag and metal in the
furnace because of poor mixing. Urquhart et al. (1976)
performed tracer studies on an EF used for smelting
coppernickel concentrates and found that the dead
volume in the slag was about 34%. Utigard (1994)
commented that there must be a significant temperature
gradient through the slag in a ferronickel furnace, and it
was expected that the slag in the vicinity of the metal/
matte interface is significantly colder, is fairly stagnant,
and may be of different composition than the bulk of the
slag. He also calculated that the degree of electromagnetic stirring in the slag was not great, leading to the
likelihood that the dead volume in the slag in a
ferronickel furnace is relatively large.

Conclusion
The simple computational thermodynamics model
developed in this work, and applied to two representative model calcine compositions, has been shown to
predict smelter outcomes satisfactorily. The most
notable failure of the model is related to the carbon
and silicon contents of low-grade ferronickels from
calcines with low Fe/Ni ratios. The model predicted that
high silicon and carbon contents should only be found in
ferronickels with much higher iron recoveries than those
observed in practice. This failure is shared by other
models and no reason could be advanced to explain the
discrepancies. Despite this, it is concluded that the
simple model is an effective educational aid to achieving
an understanding of laterite smelting to ferronickel.

Acknowledgement
The author expresses his deep appreciation to Dr
Maurice Solar, Senior Metallurgical Consultant, NonFerrous at Hatch Ltd, for many invaluable comments,
for his insights into laterite smelting, and for reviewing
the draft manuscript. He also thanks the Head of the
School of Civil, Environmental & Chemical Engineering
at RMIT University for the provision of the facilities
needed to carry out the work.

References
Bale, C. W., Chartrand, P., Decterov, S. A., Eriksson, G, Hack, K.,
Ben Mahfoud, R., Melancon, J., Pelton A. D. and Petersen, S.
2002. FactSage thermochemical software and databases. Calphad,
62, 189228.
Belton, G. R. and Fruehan, R. J. 1967. The determination of activities
by mass spectrometry. I. The liquid metallic systems iron-nickel
and iron-cobalt, J. Phys. Chem., 71, 14031409.
Bergman, R. A. 2003. Nickel production from low-iron laterite ores:
process descriptions, CIM Bulletin, 96, (1072), 127138.
Broadbent, C. P. and Machingawuta, N. C. 1993. The physical
chemistry of nickel laterite smelting slags, Proceedings of the Paul
E Queneau International Symposium: extractive metallurgy of
copper, nickel and cobalt, Vol.1 Fundamental aspects, (eds.
R.G. Reddy and R.N. Weizenbach), 316, The Minerals, Metals
and Materials Society
Cartman, R. 2010. An overview of the future production and demand of
ferronickel, 2nd Euro Nickel Conference, (IMM Informa Australia),
Available:
http://www.hatch.com.au/Mining_Metals/Iron_Steel/
Articles/documents/Future_supply_demand_ferronickel.pdf (accessed
on 20th September 2013).

Ferronickel smelting from laterites

Cluzel, D. 2013. Serpentinites: nickel laterite ore deposits: weathered

serpentinites, Elements, 9, 123128.
Conard, B. R., McAneney, T. B. and Sridhar, R. 1978.
Thermodynamics of iron-nickel alloys by mass spectrometry,
Metall. Trans. B, 9B, 463468.
Crundwell, F. K., Moats, M. S., Ramachandran, V., Robinson T. G.
and Davenport, W. G. 2011. Extractive metallurgy of nickel,
cobalt and platinum-group metals, 6784, Oxford, Elsevier.
Daenuwy, A. and Dalvi, A. D. 1997. Development of reduction kiln
design and operation at PT INCO (Indonesia), Proceedings of the
Sudbury, Ontario, Canada Nickel-Cobalt 97 International
Symposium, (eds. C. Diaz, I. Holubec and C. G. Tan), 93113,
Metallurgical Society of CIM Sudbury, Ontario, Canada.
Dalvi, A. D., Bacon, W. G. and Osborne, R. C. 2004. The past and the
future of nickel laterites, PDAC 2004 International Convention,
Trade Show & Investors Exchange, Toronto: The Prospectors
and Developers Association of Canada, 127.
Elliott, J. F., Gleiser, M. and Ramakrishna, V. 1963. Thermochemistry
for steelmaking, vol.II, Addison-Wesley Publishing Co, Reading.
Hatch, G. and Wasmund, B. 1974. Cooling devices for protecting
refractory linings of furnaces, US Patent 3,849,587.
Henao, H. M., Hino, M. and Itagaki, K. 2001. Distribution of Ni, Cr,
Mn, Co and Cu between Fe-Ni alloy and FeOx-MgO-SiO2 base
slags, Mater. Trans. Japan Inst. Metals, 42, (9), 19591966.
Jak, E. and Hayes, P. C. 2010. Slag phase equilibria and viscosities in
ferronickel smelting slags, Proceedings of the 12th International
Ferroalloys Congress, (ed. A. Vartiainen) 631639, Outotec Oy,
Finland.
Janzen, J., Gerritsen, T., Voermann, N., Veloza, E. R. and Delgado,
R. C. 2004. Integrated furnace controls: implementation on a
covered arc (shielded arc) furnace at Cerro Matoso, Proceedings
of the 10th International Ferroalloys Congress, 659669, South
African Institute of Mining and Metallurgy, Cape Town.
Kojima, V. Y., Inoue, M. and Sano, K. 1969. Die aktiviitat des
eisenoxyds in FeOMgOSiO2schlacken bei 1600uC, Arch.
Eisenhuttenwes., 40, 3740.
Lebrun, N. and Perrot, P. 2008. Ternary alloy systems subvolume D:
iron systems, (eds G. Effenberg and S. Ilyenko), 200226, Berlin,
Heidelberg, Springer-Verlag.
Levin, E. M., Robbins, C. R. and McMurdie, H. F. 1964. Phase
diagrams for ceramists, 236, The American Ceramic Society,
Columbus.
Lupis, C. H. P. 1983. Chemical thermodynamics of materials, 524527,
Prentice Hall, Englewood Cliffs.
Nelson, L. R., Geldenhuis, G. M. A., Miraza, T., Badrujaman, T.,
Hidyat, A. T., Jauhari, I., Stober, F. A., Voermann, N.,
Wasmund, B. O. and Jahnsen, E. J. M. 2007. Role of operational
support in ramp-up of the FeNi-II furnace at PT Antam in
Pomalaa, Proceedings of the 11th International Ferroalloys
Congress INFACON XI, (eds R.K. Das and T.S. Sundaresan),
798813, McMillan, Mumbai.
Outotec Research Oy. 2013. HSC Chemistry 7?1, Available: http://
www.outotec.com/Global/Products%20and%20services/Software/
HSC/HSC%20Brochure.pdf (accessed on 5th July 2013).
Pagador, R. U., Hino, M. and Itagaki, K. 1997. Equilibrium between
FeOx-MgO-SiO2 slag and nickel alloy, Proceedings of the NickelCobalt 97, Vol. 2, (eds. C.A. Levac and R.A. Berryman), 6376,
Metallurgical Society of CIM, Montreal.
Rodd, L., Voermann, N., Stober, F., Wasmund, B., Lee, S. H., Lim,
K. Y., Yoo, J. H., Roh, S. J. and Park, J. H. 2010. SNNC: a new
ferronickel smelter in Korea, Proceedings of the 12th
International Ferroalloys Congress, (ed. A. Vartiainen), 697
708, Outotec Oy, Finland.
Sigworth, G. K. and Elliott, J. F. 1974. The thermodynamics of dilute
iron alloys, Met. Sci., 8, 298310.
Solar, M. Y. 2009. Mechanical slag losses in laterite smelting nickel,
Proceedings of the Pyrometallurgy of Nickel and Cobalt 2009
(eds. J. Liu, J. Peacey, M. Barati, S. Kashani-Nejad and B.
Davis), 277292, Montreal, The Metallurgical Society of CIM.
Solar, M. Y. 2013a. Hatch Ltd, Mississauga, ON, Canada (Personal
communication, 24th September 2013).
Solar, M. Y. 2013b. Hatch Ltd, Mississauga, ON, Canada (Personal
communication, 8th November 2013).
Solar, M. Y., Candy, I. and Wasmund, B. 2008. Selection of optimum
grade for smelting nickel laterites, CIM Bull., 11, (1107), 18.
Solar, M. Y., Mostaghel, S., Nicol, S. 2014. Some insights into the
process design of laterite smelters, Conference of Metallurgists,
Vancouver, Canada, September 2014, in preparation

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO

139

Swinbourne

Ferronickel smelting from laterites

Speiser, R., Jacobs, A. J. and Spretnak, J. W. 1959. Activities of iron

and nickel in liquid iron-nickel solutions, Trans TMS-AIME, 215,
185192
Sudol, S. 2005. The thunder from down under, Canadian Mining
Journal, Available: http://www.canadianminingjournal.com/
issues/toc.aspx?edition58/1/2005 (accessed on 15th September,
2013).
Swinbourne, D. R. 2011. Accounting for non-equilibrium effects in the
thermodynamic modelling of processes, Proceedings of the
European Metallurgical Conference, Vol. 4, (ed. J. Harre),
12631276, Clausthal-Zellerfeld, GDMB Informationsges.
Swinbourne, D. R. and Kho, T. S. 2012. Computational thermodynamics modelling of minor element distributions during copper
flash converting. Metall. Mater. Trans. B, 43B, (4), 823829.
Swinbourne, D. R., Kho, T. S., Blanpain, B., Arnout, S. and Langberg,
D. 2012. Understanding stainless steelmaking through computational thermodynamics. Part 3: AOD converting, Min. Proc. Extr.
Met., 121, (1), 2331.
Urquhart, R. C., Rennie, M. S. and Rabey, C. C. 1976. The smelting of
copper-nickel concentrates in an electric furnace, International
symposium on copper extraction and refining, (ed. J. C.
Yannopoulos), 275295, Warrendale, The Metallurgical Society.
Utigard, T. 1994. An analysis of slag stratification in nickel laterite
smelting furnaces due to composition and temperature gradients,
Metall. Trans. B, 25B, 491496.
Vanderstaay, E. C., Swinbourne, D. R. and Monteiro, M. 2004. A
computational thermodynamics model of submerged arc electric
furnace ferromanganese smelting, Min. Proc. Extr. Met., 113, (1),
C38-C44.

140

Voermann, N., Gerritsen, T., Candy, I, Stober, F. and Matyas, A.

2004. Developments in furnace technology for ferronickel
production, Proceedings of the 10th International Ferroalloys
Congress, 455465, Cape Town, South African Institute of
Mining and Metallurgy.
Vogel, I. A. and Palme, H. 2004. Activity coefficients of silicon in ironnickel alloys Experimental determination and relevance for
planetary differentiation, Lunar and Planetary Science XXXV,
Mar 1519, League City, Texas, USA, Abstract No. 1592.
Walker, C., Kashani-Nejad, S., Dalvi, A. D., Voermann N., Candy,
I. M. and Wasmund, B. 2009. Future of rotary kiln electric
furnace (RKEF) processing of nickel laterites, Proceedings of the
European Metallurgical Congress 2009, (ed. J. Harre), 943974,
Clausthal-Zellerfeld, GDMB Informationsges.
Walker, C., Koehler, T., Voermann, N. and Wasmund B. 2010. High
power shielded arc FeNi furnace operation challenges and
solutions, Proceedings of the 12th International Ferroalloys
Congress, (ed. A. Vartiainen), 681696, Finland, Outotec Oy.
Warner, A. E. M., Diaz, C. M. and Dalvi, A. D. 2006. World
nonferrous smelter survey, Part III: laterite J. Met., 58, 11
20.
Yamaguchi, K., Swinbourne, D. R. and Yazawa, A. 2005. Behaviour of
minor species during oxidation smelting of lead concentrate,
Proceedings of the Lead-Zinc 2005, (ed. T. Fujisawa), 12311246,
The Mining and Metallurgical Institute of Japan, Kyoto.
Zellars, G. R., Payne, S. L., Morris, J. P. and Kipp, R. L. 1959. The
activities of iron and nickel in liquid Fe-Ni alloys, Trans. TMSAIME, 215, 181185.

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C)

2014

VOL

123

NO