1 1

[MS] ⫹ O2 ⫽ (MO) ⫹ S2 [1]

Communications 2 2
where [MS] represents metal sulfide in the matte (i.e., Ni3S2,
Nickel, Copper, and Cobalt Slag Cu2S, or CoS) and (MO) represents metal oxides (i.e., NiO,
Cu2O, or CoO) in the slag. In practice, the sulfur and oxygen
Losses during Converting pressures are determined mainly by the exchange of iron
because iron is a substantial part of both the slag and matte.
P. TOSCANO and T.A. UTIGARD The exchange of metal therefore is best considered by the
following reaction:[5–8]
The objective of this investigation was to determine the
slag and matte compositions during laboratory converting [MS] ⫹ (FeO) ⫽ (MO) ⫹ [FeS] [2]
of nickel/copper flash smelting matte. The experiments were where [FeS] represents ferrous sulfide in matte and (FeO)
carried out at 1250 ⬚C with the injection of air or oxygen- represents ferrous oxide in the slag.
enriched air. In the case of copper, on average, 56 pct During converting, the oxygen pressure increases as the
reported to the slag as entrained matte. The entrained parti- matte FeS content decreases and the slag magnetite content
cles have a higher Cu-to-Ni ratio than the corresponding increases. Several investigators[9,10,11] have found that the
bulk matte, indicating that the entrained matte particles are nickel solubility in fayalitic slag increases with increasing
further oxidized in the slag. In the case of cobalt, 98 pct partial pressure of oxygen. Therefore, it is not surprising
reported to the slag as dissolved CoO, while for nickel, 76 that slag losses increase as the converting process proceeds.
pct was lost as oxide. At least 30 to 41 pct of the oxide Lee[18] determined that for a given matte grade, metal
nickel loss is due to the oxidation of entrained matte droplets. losses to slags decrease at lower sulfur pressures. Font et
When the matte iron content decreases below about 3 wt al.[12,13] investigated the copper, nickel, and cobalt solubilit-
pct, the slag losses and the magnetite content start to ies between FeOx-SiOs-MgO slag and Cu2S-Ni3S2-FeS matte
increase rapidly. under SO2 pressures of 0.1, 0.5, and 1 atm. At a given matte
Nickel produced in Canada is extracted from sulfide ores grade, the solubility of copper in the slag was found to be
where pentlandite ((Ni, Fe)9S8), chalcopyrite (CuFeS2), and independent of pSO2 while that of nickel and cobalt increased
pyrrhotite (Fe7S8) are the major minerals. Mined sulfide ores with increasing pSO2. These results imply that dissolved metal
contain typically about 1 pct nickel and are concentrated to losses can be decreased by decreasing oxygen, sulfur, or
7 to 20 pct nickel by various mineral processing methods. SO2 pressures. However, in industrial matte converting oper-
The concentrate may then be dried and is ready for one of ations, one does not have the freedom to vary these parame-
three pyrometallurgical processes: ters at will. Operational variables include the oxygen
enrichment in the blast, the amount and type of flux used,
(1) partial roasting/electric furnace smelting/converting,[1,2]
and the temperature. By varying the oxygen content in the
(2) flash smelting/converting,[3] and
blast, the SO2 pressure in bubbles formed within the matte
(3) direct Outokumpu nickel process (DON) process.[4]
and slag also varies.
Molten matte from the electric or flash smelting furnace The experiments were designed to simulate the converting
is transferred to a Peirce Smith converter for the removal process by converting flash furnace matte to low levels of
of the final FeS to form a Bessemer matte with about 0.5 iron. In these tests, 500 g of the flash furnace matte (26.2
to 2 wt pct Fe. This is achieved by blowing air or oxygen- pct Fe, 22.4 pct Ni, 22.7 pct Cu, 0.74 pct Co, 27.1 pct S,
enriched air into the matte. It is during this stage with high and 2.86 pct O) was added to a MgO crucible with an inner
oxygen pressures that the greatest amounts of valuable metal diameter of 7.5 cm and a height of 15 cm. The converting
loss occur. The slag produced is invariably treated, either in was accomplished by injecting air or oxygen-enriched air
a separate slag cleaning step or by recycling back to the into the liquid matte through a lance that was immersed into
smelting furnace. This recycling may lead to problems in the matte with the tip at a distance of 0.5 cm from the
the primary smelting furnace as well as limit the furnace bottom of the crucible. Silica additions were made so that
capacity to treat fresh feed. In order to minimize these losses, the oxidized iron would form a fayalitic slag. All the experi-
it is important to determine their form and origin. Therefore, ments were carried out at 1250 ⬚C.
the purpose of this investigation was to determine the amount An electrical resistance furnace was used to conduct all
of entrained and dissolved Ni, Co, and Cu losses as a function experimental test work. One R-type thermocouple was used
of the matte iron content during the converting process. to control the furnace temperature and another R-type ther-
During smelting and converting of mattes, equilibrium mocouple was used to measure the actual temperature of
tends to be quickly approached with the slag.[5] Nickel and the melt itself. An alumina sheath protected the second ther-
cobalt report to slag predominately as oxides and the equilib- mocouple and the tip of the thermocouple was maintained
rium between the matte and the slag can be represented by about 2 cm below the surface of the melt.
the following general reaction: Nitrogen gas (99.998 pct) was flowing above the matte
while it was being melted to prevent any preoxidation. Once
the matte was melted and heated to 1250 ⬚C, it was converted
to low iron levels by the injection of air, oxygen-enriched
P. TOSCANO, Engineer, is with Minnovex Technologies Inc., Toronto, air, or oxygen (21, 40, 60, and 95 pct O2). The air/oxygen
ON, Canada M3K 2A2. T.A. UTIGARD, Professor, is with the Department
of Materials Science and Engineering, University of Toronto, Toronto, ON, gas was injected into the matte through a 1.3-mm i.d. ⫻ 5-
Canada M5S3E4. Contact e-mail: utigard@[Link] mm o.d. SiC/SiN tube. It was found that this particular
Manuscript submitted December 6, 2001. material had a much longer life than alumina or mullite

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, FEBRUARY 2003—121

tubes when in contact with slag, matte, and oxidizing gas. rates used. It is seen that within the composition range stud-
The SiC/SiN tubes were slightly porous so they were coated ied, this regression equation can with reasonable accuracy
with alumina cement to prevent gas leakage. predict the cobalt distribution.
Before sampling of matte and slag, the lance was removed Figure 6 gives the total nickel losses (dissolved plus
from the melt and converting was halted. The matte and entrained) in the slag. Although there is a significant degree
slag were allowed to settle for approximately 2 minutes and of scatter in the data, the nickel losses increase as the iron
then the slag was sampled using a steel rod. The cold rod content in the matte decreases. This scatter is caused by
was dipped into the slag and a small amount of slag froze entrained matte in the slag. Based on a regression analysis
onto the surface of the rod. The slag was then quenched in of the tests carried out, on average, 76 pct of the nickel in
water and the slag sample broken off. The slag was sampled the slag occurs as dissolved nickel oxide.
until there was approximately 8 g of slag. The samples In the case of copper, as seen in Figure 7, the scatter is
were subsequently ground to 100 pct passing 100 mesh and even more noticeable. Copper tends to report to the slag to
assayed at the INCO research laboratory using the bromine a large extent as entrained matte (average of 56 pct) and the
alcohol leach method. The leach solution contains all sulfides iron content in the matte does not significantly affect the
and metallics (i.e., entrained matte) and the residue contains losses. On average, only 44 pct of the copper losses were
the dissolved oxides. The matte was sampled through a associated with dissolved copper in the slag.
quartz tube by applying suction. Approximately 4-g samples Distribution coefficients (also known as partition coeffi-
of matte were taken. Matte samples were broken up into 1- cients) have been defined[5,7,15] as
to 4-mm pieces and sent to the INCO research laboratory
[wt pct M]M
for analysis. The matte samples were not ground to pre- DM ⫽ [6]
vent oxidation. (wt pct M)S
The main experimental results are summarized in Table
Some other researchers such as Font et al.[12,13] and
I. Figure 1 plots the iron removal from the matte vs time
Matousek[16] used the inverse version to depict distribu-
for five tests with various oxygen enrichments and flow
tion coefficients.
rates. It is seen that the higher the oxygen content in the
Figures 8 and 9 show the distribution coefficients of cobalt
blast, the faster iron is removed. However, in the tests with
and nickel vs the iron content in the matte. It is seen that
40 and 60 pct O2 enrichment, the oxygen efficiency was
the Co distribution coefficients correspond very well with
only about 60 pct. The reason may be that due to the high
published industrial and laboratory data. Our data are higher
flow rates used in these two particular tests, some gas may
than those from industrial converting[14] and slightly lower
have leaked out through the pores of the SiC injection tube.
than those obtained by Bustos et al.[17] in a plant trial at
In the other tests with air and 95 pct O2 enrichment, the
Falconbridge. The laboratory equilibrium data by Lee[18] are
oxygen efficiency is nearly 100 pct confirming that the
higher than our data due to the low sulfur pressures used in
experimental technique is working well.
his tests. As the sulfur pressure decreases, the matte becomes
Figure 2 shows the sum of the Ni, Cu, and Co concentra-
deficient in sulfur and enriched in metallics. This is known
tions in the matte during converting. Also shown is a regres-
to improve the recovery of cobalt. Based on a regression
sion line of industrial converting data in a Peirce Smith
analysis of our data, the cobalt distribution coefficient
converter.[14] There is good correspondence between our
depends on the matte iron contents as follows:
laboratory converting process and that taking place in indus-
trial converters. As seen in this graph, the oxygen concentra- DCo ⫽ 0.229 ⫻ wt pct Fe ⫺ 0.0043 ⫻ (wt pct Fe)2 [7]
tion in the blast has little effect on the stoichiometry of
the matte. As seen in Figure 9, in the case of the nickel partition
Figure 3 gives the total (dissolved and entrained) cobalt coefficient, due to matte entrainment, the data are somewhat
slag losses vs the matte iron content. It is clear that as more scattered than for cobalt. Our data correspond very
the matte grade increases, more cobalt reports to the slag. well with the data by Bustos et al.[17] Again, the equilibrium
Although not as significant, it appears that increasing the laboratory studies by Lee[18] show increasing partition coeffi-
O2 content in the blast may lead to increased losses. Figure cients with decreasing sulfur pressure. Based on a regression
4 shows the ratio of dissolved cobalt in the slag vs the total analysis of our data, the nickel distribution coefficient is
amount of cobalt in the slag. On average, 98 pct of the given by
cobalt slag losses occur as dissolved oxides and very little DNi ⫽ 15 ⫻ ln (wt pct Fe) ⫹ 3.7 [8]
as entrained matte. This means that the cobalt losses can be
described by the following exchange reaction: In order to improve the understanding of the effect of
entrained matte in the slag, the ratios of Cu/Ni (1) in the
CoS ⫹ FeO ⫽ CoO ⫹ FeS [3]
matte, (2) total in the slag, (3) dissolved in the slag, and (4)
or entrained particles in the slag were calculated and plotted
CoS ⫹ 1/2Fe2SiO4 ⫽ 1/2Co2SiO4 ⫹ FeS [4] in Figure 10. It is seen that for the matte, the ratio increases
slightly as the converting process proceeds. In the slag, the
Based on a regression analysis of results from all the tests, total Cu/Ni ratio decreases during converting because nickel
the following Co distribution ratio was found: oxidizes more readily than copper as the wt pct Fe in the
pct CoS ⫻ pct FeM matte decreases. This is clearly seen by the decreasing ratio
QCo ⫽ ⫽ 0.1 ⫹ 0.023 ⫻ pct FeM [5] of dissolved Cu/Ni in the slag as the iron is removed. On
pct CoM ⫻ pct FeS
the other hand, the Cu/Ni ratio for entrained particles is
This relation is plotted in Figure 5 together with the experi- larger than one and varies quite significantly. The fact that
mental data for the various oxygen enrichments and flow this ratio is higher than the corresponding ratio in the matte

122—VOLUME 34B, FEBRUARY 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

 Table I. Composition (Weight Percent) of Matte and Slag during Converting Tests
Pct Metallics in Matte Pct Dissolved Metals in Slag Total Amount in Slag
Pct Fe Pct Cu Pct Ni Pct Co Pct Cu Pct Ni Pct Co Pct Cu Pct Ni Pct Co Pct Fe
21 pct O2 15.4 29.7 28.6 0.82 0.29 0.34 0.29 0.56 0.55 0.30 42.9
2.0 L/min 11.9 31.6 30.5 0.80 0.26 0.49 0.34 0.61 0.63 0.34 40.7
9.3 33.2 32.0 0.78 0.31 0.57 0.39 0.53 0.61 0.40 40.9
7.3 34.7 33.3 0.74 0.38 0.68 0.47 0.63 0.79 0.47 40.9
6.4 35.5 33.8 0.69 0.40 0.82 0.53 0.60 0.84 0.53 41.4
5.3 36.4 34.5 0.66 0.44 0.96 0.59 1.01 1.34 0.59 41.5
4.1 37.9 35.6 0.61 0.53 1.21 0.67 0.90 1.42 0.67 41.7
2.9 38.9 36.2 0.51 0.30 1.49 0.72 1.07 1.82 0.75 41.3
Pct Fe Pct Cu Pct Ni Pct Co Pct Cu Pct Ni Pct Co Pct Cu Pct Ni Pct Co Pct Fe
95 pct O2 15.0 29.6 28.9 0.81 0.34 0.55 0.35 0.97 0.93 0.36 45.9
1.5 L/min 12.7 30.3 30.9 0.81 0.46 0.68 0.39 0.93 0.98 0.41 44.8
9.7 32.4 32.5 0.78 0.42 0.82 0.47 0.93 1.09 0.48 45.4
7.0 34.0 33.9 0.73 0.43 1.07 0.58 0.76 1.16 0.59 46.1
5.2 35.4 35.2 0.68 0.53 1.33 0.66 2.68 3.15 0.69 43.7
3.0 37.2 36.4 0.57 0.65 2.16 0.89 1.15 2.28 0.91 45.1
1.6 38.2 37.7 0.40 0.81 3.60 1.08 6.91 8.76 1.09 38.3
1.2 40.0 36.7 0.36 0.29 4.42 1.19 2.12 5.15 1.18 42.0
Pct Fe Pct Cu Pct Ni Pct Co Pct Cu Pct Ni Pct Co Pct Cu Pct Ni Pct Co Pct Fe
60 pct O2 19.6 26.9 26.5 0.79 0.26 0.49 0.27 1.64 1.53 0.29 41.8
3.0 L/min 17.5 27.8 27.5 0.81 0.31 0.47 0.30 0.70 0.64 0.30 45.1
17.0 28.7 28.4 0.80 0.24 0.53 0.33 1.43 1.36 0.34 44.1
13.7 29.1 28.8 0.86 0.34 0.56 0.36 0.63 0.63 0.36 44.9
12.7 31.2 31.1 0.88 0.24 0.62 0.39 0.73 0.78 0.40 44.6
10.8 32.4 32.0 0.81 0.42 0.77 0.46 0.66 0.82 0.46 44.7
7.6 33.9 32.9 0.78 0.27 0.96 0.54 2.49 2.74 0.55 41.7
5.4 36.2 34.5 0.69 0.18 1.18 0.63 0.84 1.19 0.63 44.6
Pct Fe Pct Cu Pct Ni Pct Co Pct Cu Pct Ni Pct Co Pct Cu Pct Ni Pct Co Pct Fe
40 pct O2 19.2 27.6 27.4 0.81 0.24 0.42 0.29 0.89 0.70 0.29 44.8
3.0 L/min 15.2 29.1 28.5 0.81 0.18 0.41 0.28 1.06 0.95 0.31 43.1
14.5 30.1 29.8 0.82 0.23 0.47 0.31 2.08 2.07 0.34 41.7
13.3 30.9 30.8 0.82 0.27 0.51 0.34 1.52 1.48 0.36 42.4
12.2 31.6 31.5 0.81 0.21 0.54 0.36 1.14 1.15 0.38 43.8
10.6 32.5 31.8 0.79 0.26 0.60 0.39 1.28 1.89 0.65 38.3
8.8 33.2 32.9 0.80 0.30 0.68 0.42 0.66 0.76 0.41 41.1
6.9 35.1 34.9 0.75 0.40 0.71 0.45 0.72 0.85 0.46 42.1
4.9 35.5 36.5 0.70 0.41 1.09 0.54 0.88 1.18 0.56 41.6
Pct Fe Pct Cu Pct Ni Pct Co Pct Cu Pct Ni Pct Co Pct Cu Pct Ni Pct Co Pct Fe
21 pct O2 17.8 27.3 26.3 0.84 0.25 0.35 0.27 2.07 1.71 0.29 43.7
2.0 L/min 13.1 30.1 28.7 0.83 0.36 0.45 0.33 0.65 0.57 0.34 44.3
11.4 31.5 30.2 0.83 0.37 0.53 0.38 0.53 0.56 0.38 44.0
9.3 33.5 31.5 0.81 0.42 0.64 0.44 0.71 0.79 0.49 43.2
7.7 34.2 32.3 0.76 0.41 0.76 0.49 0.57 0.78 0.50 43.5
6.8 34.7 32.6 0.73 0.40 0.81 0.53 0.58 0.84 0.54 43.6
5.8 36.0 33.4 0.68 0.46 0.96 0.59 1.08 1.41 0.60 43.2
4.5 36.6 34.0 0.65 0.51 1.33 0.68 0.67 1.29 0.69 43.9
4.0 37.5 34.2 0.60 0.58 1.48 0.76 0.96 1.66 0.75 43.1
3.4 38.2 34.3 0.54 0.43 1.83 0.76 0.89 2.03 0.78 43.0

implies that matte particles entrained in the slag get further the Cu in the particles is oxidized, it is possible to determine
oxidized while remaining in the slag. The cobalt (in particu- how much nickel oxide is formed during matte particle
lar) and nickel will oxidize selectively over copper. This oxidation in the slag. From a mass balance of the dissolved
further means that a part of the dissolved cobalt and nickel and entrained nickel and copper in the slag, it was found
losses in the slag originate from the oxidation of entrained that for the various tests, of the total dissolved nickel in the
matte particles. This oxidation may be due to reactions with slag, at least 30 to 41 pct was formed by the oxidation of
magnetite in the slag or by oxygen gas bubbling through these entrained matte particles. Because some of the copper
the slag or by gases in contact with the surface of the slag. in the entrained particles probably also gets oxidized, this
By assuming that initially the entrained matte particles nickel loss is probably even higher. This suggests that one
have the same composition as the matte and that none of should try to minimize entrainment in order to decrease the

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, FEBRUARY 2003—123

 Fig. 4—Dissolved over total cobalt in the slag vs matte iron content.

Fig. 1—Wt pct Fe vs time during converting of flash furnace matte.

Fig. 5—Cobalt distribution ratio vs matte iron content.

Fig. 2—Matte grade vs wt pct Fe during converting of flash furnace matte.

Fig. 6—Wt pct total nickel in the slag vs matte iron content.

laboratory scale using different concentrations of oxygen in

Fig. 3—Wt pct cobalt vs matte iron content. the blast. The results from the laboratory converting tests
simulated industrial results closely.
Based on a regression analysis of our data, the cobalt and
overall nickel losses as well as dissolved losses. In terms nickel distribution coefficients depend on the matte iron
of cobalt oxidized, entrained matte particles contribute only content as follows:
up to about 0.015 pct Co in the slag, which is less than 5
pct of total cobalt losses. pct CoM
⫽ 0.229 ⫻ wt pct Fe ⫺ 0.0043 ⫻ (wt pct Fe)2
High-temperature converting tests were performed on a pct CoS

124—VOLUME 34B, FEBRUARY 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

 Fig. 10—Ratio of Cu over Ni (1) in matte, (2) total in the slag, (3) dissolved
Fig. 7—Wt pct total copper in the slag vs matte iron content.
in slag, and (4) entrained in the slag vs matte iron content.

The financial support from the Centre of Chemical Process

Metallurgy and NSERC is greatly appreciated. The authors
thank INCO Limited for carrying out the chemical analysis
of the matte and slag samples.

REFERENCES
1. S.C.C. Barnett and J.H.E. Jeffes: Trans. Inst. Min. Metal., 1977, vol.
86, pp. C153-C157.
2. D.G. Tisdale and C.G. Ransom: Nickel Cobalt 97, Sudbury, ON,
Canada, Aug. 17–20, 1997, CIM, Montreal, Que., Canada, vol. 3, pp.
35-43.
3. H. Carr, M.J. Humphris, and A. Longo: Proc. Nickel-Cobalt 97 Int.
Fig. 8—Cobalt distribution coefficient vs matte iron content. Symp., CIM, Montreal Que., Canada, 1997, pp. 5-16.
4. T. Makinen and P. Tarkinen: Symp. Proc. Sulfide Smelting, TMS,
Warrendale, PA, 1998, pp. 59-68.
5. P.J. Mackey: Can. Metall. Q., 1982, vol. 21 (1), pp. 221-60.
6. C. Diaz, B.R. Conard, C.E. O’Neill, and A.D. Dalvi: Extractive Metal-
lurgy of Copper, Nickel and Cobalt, Volume 1: Fundamental Aspects,
TMS, Warrendale, PA, 1993, pp. 583-99.
7. P.R. Ammann, J.J. Kim, and T.A. Loose: J. Met., 1979, vol. 31 (7),
pp. 20-25.
8. E.J. Grimsey: Extractive Metallurgy of Gold and Base Metals,
AusIMM, Melbourne, Australia, 1992, pp. 77-80.
9. S.S. Wang, N.H. Santander, and J.M. Toguri: Metall. Trans., 1974,
vol. 5, pp. 261-65.
10. J.R. Taylor and J.H. Jeffes: Trans. Inst. Min. Metall., 1975, vol. 84,
pp. C136-C148.
11. E.J. Grimsey and A.K. Biswas: Trans. Inst. Mining Metall., 1977, vol.
86, pp. C1-C8.
12. J.M. Font, G. Roghani, M. Hino, and K. Itagaki: MMIJ, 1998, vol.
15 (1), pp. 75-86.
13. J.M. Font, M. Hino, and K. Itagaki: Mater. Trans. Jpn. Inst. Met.,
1998, vol. 39, pp. 834-40.
14. Converting Data from Inco, Sudbury, Warren Shellshear private com-
munications, Dec. 2000.
Fig. 9—Nickel distribution coefficient vs matte iron content.
15. M. Nagamori and P.J. Mackey: Can. Metall. Q., 1977, vol. 16, pp.
232-36.
16. J.W. Matousek: Extractive Metallurgy of Copper, Nickel and Cobalt,
vol. 1, Fundamental Aspects, TMS, Warrendale, PA, 1993, pp. 129-42.
17. A.A. Bustos, S.W. Ip, G. O’Connell, G.H. Kaiura, and J.M. Toguri:
pct NiM Extractive Metallurgy of Nickel Cobalt, G.P. Tyroler and C.A. Landolt,
⫽ 15 ⫻ ln (wt pct Fe) ⫹ 3.7 eds., TMS, Warrendale, PA, 1988, pp. 335-54.
pct NiS 18. W.P. Lee: Ph.D. Thesis, University of Toronto, Toronto, 1999.

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, FEBRUARY 2003—125