Transport Phenomena in Electric Smelting of Nickel Matte:

Part I. Electric Potential Distribution

Y.Y. SHENG, G.A. IRONS, and D.G. TISDALE

An electric potential probe was constructed so that simultaneous, multiple measurements of electric
potential could be made in a six-in-line electric furnace for smelting nickel calcine having a maximum
transformer capacity of 36 MVA. When the electric potential distributions were compared with those
calculated from the solution of the Laplace equation, it was evident that there was significant electric
potential drop at the electrode surface, 100 to 120 V for an applied potential of 180 to 230 V and
currents of 20 to 30 kA. The Soderberg electrodes were continuously oxidized in the slag, likely
creating carbon monoxide. The electric potential drop at the surface was attributed to arcing through
the carbon monoxide. Thus, heat was released in the immediate vicinity of the electrode due to
arcing, as well as in the bulk of the slag by Joule heating. The proper distribution of heat dissipation
is required for the transport model, developed in Part II of this series.

I. INTRODUCTION purchased consumable. For example, the Falconbridge fur-

nace modeled in the present study is the single largest load
ELECTRIC smelting is used for a wide variety of ma- on the public utility grid in Ontario, Canada. Consequently,
terials such as copper, nickel calcine, nickel laterites, steel,
even small improvements in electrical use can have enor-
prereduced iron ore, and electric arc furnace dust. There are
mous environmental and financial benefits. Surprisingly,
at least four distinct smelting techniques.[1]
there have been few published studies of electrical power
(1) In immersed-electrode smelting, the electrode tips are use in electric furnaces, as will be reviewed in Section II.
dipped into molten slag, where electrical energy is gen- This lack of understanding has prompted the present study.
erated by Joule heating. This practice is employed In Part I of this two-part article, the electric potential
mainly for copper-nickel calcines and nickel laterites. distribution inside a full-scale operating furnace was mea-
The present study focuses on this practice. sured. The measurements revealed a substantial electric po-
(2) Open-arc smelting uses an arc between the electrode tential drop in the immediate vicinity of the electrodes,
and the bath to generate heat. This technique is most which had not been previously reported. The electric po-
commonly used for melting metals such as steel. It is tential distribution is one of the most important factors in
important that the heat radiated by the arc be captured determining the heat use distribution. In Part II, this infor-
by the melt; the development of foaming slag practices mation is used to develop a comprehensive model for fluid
has improved the energy recovery in modern ultra- flow, heat transfer, and electric current distribution in Fal-
high-power furnaces. conbridge’s electric furnace.[4]
(3) Submerged-arc smelting is the traditional method for
reduction smelting with coke in furnaces producing fer-
roalloys, silicon, and phosphorus. Electric energy is lib- II. LITERATURE REVIEW
erated by arcing between coke particles floating on the
In immersed-electrode smelting furnaces, heat is supplied
bath.
to the bath by an alternating current flowing between elec-
(4) Shielded-arc smelting was developed in Falconbridge’s
trodes immersed in the molten slag. The heat generation
Dominican Republic operations for laterite smelting.
and transfer process is determined by the electrical con-
Long arcs are covered with charge to ensure efficient
ductivity of the molten slag and the electric potential ap-
use of electrical energy for smelting.[1]
plied to electrodes, as well as the specific geometry of the
Most of the world’s nickel is produced either by flash furnace, which is represented by a geometric factor defined
smelting or by electric smelting.[2] The major advantages of by the following equation:[5,6]
the electric smelting processes are the high sulfur capture
and high metal recoveries.[3] This process requires large fg
Rs 5 [1]
amounts of electric energy, which is often the most costly ss
where Rs is the total resistance between two electrodes, ss
is the electrical conductivity of slag, and fg is the geometric
Y.Y. SHENG, formerly Post Doctoral Fellow with the Department of
Materials Science and Engineering, McMaster University, is Research factor. To date, most efforts have concentrated on the de-
Scientist, Noranda Technology Centre, Pointe Clairs, PQ, Canada H9R termination of appropriate algebraic expressions for the ge-
1G5. G.A. IRONS, DOFASCO/NSERC Professor of Process Metallurgy, ometric factor in various furnace configurations. Urquhart
is with the Department of Materials Science and Engineering, McMaster et al.[7] investigated the electrical characteristics of the Cu-
University, Hamilton, ON, Canada L8S 4L7. D.G. TISDALE, Process
Engineer, is with the Smelter Complex, Falconbridge Limited, Ni electric smelting process by measuring slag temperature,
Falconbridge, ON, Canada P0M 1S0. slag composition, and the electrical parameters. They also
Manuscript submitted April 12, 1996. conducted experiments with small-scale analog models

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 29B, FEBRUARY 1998—77

 Fig. 3—Plan view of the Falconbridge electric furnace for nickel smelting,
showing the internal dimensions in meters.

in the slag follows Ohm’s law, misrepresentation of the slag

resistance results in an error in the calculated power gen-
eration rate in the furnace.
Fig. 1—Measured resistance between electrodes in the Falconbridge The electrical power release rate is also dependent on the
electric furnace as a function of electrode immersion for two different slag
depths.[9] The hatched region shows that the previous correlations[6,7,8] electrical conductivity of slag, which is a strong function
underestimate the resistance. The dotted line is the ohmic resistance of of both slag temperature and composition. Recently, Jiao
the slag calculated in the present work. and Themelis[10] proposed some useful correlations for slag
conductivity. They carefully analyzed existing data from
both laboratory and plant measurements and performed
multiple linear regression analyses to relate the slag con-
ductivity to the molar concentrations of Ca+2, Mg+2, and
other cations. The electrical conductivity of Falconbridge’s
electric furnace slag was measured by Stubina et al.[11] As
Figure 2 shows, the measured conductivities were slightly
higher than those from the correlations proposed by Jiao
and Themelis. The data and the correlations show a strong
dependency of slag conductivity on temperature and on the
Fe/SiO2 ratio of the slag. However, variations in the slag
conductivity due to fluctuations in slag composition or tem-
perature cannot account for the large discrepancy in Figure
1 between the measured slag resistance and that calculated
using the correlations. Therefore, there are some additional
factors that influence the resistance in the furnace. The pres-
ent experiments were undertaken to comprehend them.

III. PROCESS DESCRIPTION

In 1978, Falconbridge Limited commissioned a new pro-
cess in Falconbridge, Ontario, which consists of two slurry-
fed fluidized-bed roasters for partial sulfur elimination and
two electric furnaces for smelting the hot calcine. Cleaned
roaster gases are treated in a single-contact acid plant.
Fig. 2—The electrical conductivity of the slag in the present study with Matte from the electric furnaces is processed in Peirce–
an iron-to-silica ratio of 0.95. Stubina et al.[11] measured the conductivity,
whereas Jiao and Themelis[10] developed correlations based on Smith converters. A semifinished matte is shipped to Fal-
measurements. conbridge’s Norwegian refinery for further treatment.[12]
Top and side views of the electric furnace are shown in
to study the effect of the electrode geometry on current Figures 3 and 4, respectively. The incoming electric power
partition in the slag, and they computed the power gener- at 115 kV is stepped down to 13.8 kV in the primary trans-
ation distribution in the bath with mathematical models former. The electric furnace has three secondary 12 MVA
based on measured voltage and current data. Again, using transformers, which are located in vaults adjacent to the
analog models, Jiao and Themelis[8] developed their own furnace and have on-load tap changers to produce voltages
correlations, which they claimed to cover more electrical between 137 and 254 V in the star connection and between
current paths and therefore to be more accurate than pre- 238 and 440 V in the delta connection. The electrical con-
vious correlations. figuration of the furnace is shown in Figure 5 and other
These studies have contributed greatly to our understand- operational parameters are listed in Table I.
ing of the heat generation and transfer process in an electric
furnace. However, in the Falconbridge furnace, it was IV. MEASUREMENT OF ELECTRIC POTENTIAL
found that there was a large difference between the mea- DISTRIBUTION IN SLAG
sured slag resistance and these correlations,[9] as shown in
A. The Electric Potential Probe
Figure 1. (Calculations from the present model are also
shown for later discussion.) The discrepancy is particularly The electric potential distribution in the slag was mea-
severe at shallow electrode immersions. Since heat release sured with a newly designed electric potential probe. A sec-

78—VOLUME 29B, FEBRUARY 1998 METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 4—Schematic end view of the Falconbridge electric furnace for nickel
smelting.

Fig. 6—Cross section of a section of the probe used for the electric
potential measurement, showing the details of construction of two of the
five tungsten electrodes, not to scale.

electrodes made of tungsten rod with a diameter of 3.18

mm and a length of 44.5 mm were positioned in the re-
fractory and the pipe. The distance between the electrodes
was 254 mm. Nichrome alloy wire, protected by a high-
temperature ceramic sleeve, was used to connect each elec-
trode to a digital voltmeter. The lowest electrode was im-
Fig. 5—Electrical circuit of the Falconbridge electric furnace for nickel mersed just below the matte/slag interface to act as the
smelting. ground for the electric potential measurement; all measure-
ments are reported with respect to the matte potential.
Table I. Electric Furnace Test Conditions The temperature distribution in the bath was measured
with a separate holding rod containing a K-type thermo-
Power (MW) 25 to 40 couple. The thermocouple had a stainless steel sheath with
Electric potential (V) 180 to 230 an outer diameter of 6.35 mm and a length of about 7 m.
Current (kA) 20 to 28
Electrode diameter (mm) 1400
Electrode immersion (mm) 200 to 600 B. The Location of Measurement
Electrode consumption 1 to 1.8
(kg C/tonne calcine) (all six electrodes)
The locations for the electric potential measurement in
Slag depth (mm) 1200
slag were quite restricted since the furnace was enclosed,
Matte depth (mm) 900
except for several ports in the roof for temperature mea-
Slag temperature (7C) 1220 to 1300
surement. Two such ports on either side of an electrode
Matte temperature (7C) 1150 to 1200
were chosen for the electric potential measurement, and
their locations relative to the electrode are shown in Figure
7 as locations A and B. Because the probes were inserted
tion of the probe is shown in Figure 6. The probe was made at an angle, they sagged if they were inside the furnace for
of a steel pipe with an outer diameter of 25.4 mm and a an extended time. To overcome this problem, some meas-
length of 3.0 m. The bottom half of the pipe was covered urements were taken vertically downward at location C.
with a 25.0-mm-thick layer of alumina-rich refractory. Five The probe assembly was lowered on a carriage to multiple

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 29B, FEBRUARY 1998—79

 Fig. 8—Electric potential in the slag with respect to the matte as a function
Fig. 7—Plan and end views of the locations for electrical potential of distance from the matte at location A in furnace 1. v Data correspond
measurements. Dimensions are in meters. to 180 and 130 V applied to electrodes 1 and 2, respectively, 28 kA, and
8.7 MW between electrodes 1 and 2 with a slag depth of 250 mm. ¶
positions, so that more than five measurements could be Data correspond to 230 and 180 V applied to electrodes 1 and 2,
respectively, 32 kA, and 13 MW between electrodes 1 and 2 with a slag
made at each location. The measurements were carried out depth of 300 mm. The lines were computed from the model for a slag
in furnaces 1 and 2; however, the locations in each furnace depth of 250 mm, and the various potentials applied to the slag, Vs,
were different. In furnace 1, location A (Figure 7) was be- indicated.
tween electrodes 1 and 2 (phase I, Figure 5). In furnace 2,
locations B and C were between electrodes 2 (phase I) and model results, which will be discussed in Section V. In each
3 (phase II). figure, the uppermost curve represents the electric potential
profile that would be observed if the electric potential that
C. Experimental Procedure the slag experiences, Vs, were the electric potential applied
to the electrode. Clearly, the measured electric potentials
Preceding the measurements, the furnace was maintained are substantially lower. The other curves represent lower
in as steady an operation as possible. The procedure was electric potentials for the slag. Note that the electric poten-
as follows: tials that the slag experiences are some 100 to 120 V lower
(1) measurement of the immersion of the furnace elec- than those applied to the electrode.
trodes by first raising them to the slag-calcine interface A temperature profile for location A is shown in Figure
(indicated by a zero current reading) and then returning 11. The temperature was reasonably uniform, except near
them to the designated positions; the matte interface. The average slag tapping temperature
(2) measurement of the vertical temperature profile in slag only varied from 1230 7C to 1280 7C in the present exper-
and matte with the furnace power off at the same lo- iments. If the power were off for more than 30 minutes for
cations as the electric potential measurement; a temperature measurement, then the slag became thermally
(3) measurement of the depths of furnace bottom buildup, stratified. The data in Figure 11 were gathered over a period
matte, and slag, respectively, at the same locations as of less than 30 minutes.
the electric potential measurement with a sounding bar;
and
(4) electric potential measurement with the electric poten- V. DISCUSSION
tial probe. A. Calculation of Electric Potential Distribution in Slag
The details of the mathematical model used for the cal-
D. Experimental Results
culation of the electric potential distribution are presented
The measured electric potential distributions for three in Part II of this article, where a simplified Maxwell’s equa-
different locations are shown in Figures 8 through 10. The tion[13,14] is solved together with the flow and temperature
curves superimposed on the diagram are the mathematical fields. As will be shown, the electromagnetic force field is

80—VOLUME 29B, FEBRUARY 1998 METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 9—Electric potential in the slag with respect to the matte as a function Fig. 10—Electric potential in the slag with respect to the matte as a
of distance from the matte at location C in furnace 2. The data correspond function of distance from the matte at location B in furnace 2. The data
to 230 and 210 V applied to electrodes 2 and 3, respectively, 23 kA, and correspond to 230 V applied to electrodes 2 and 3, 21 kA, and 12 MW
12 MW between electrodes 2 and 3 with a slag depth of 300 mm. The between electrodes 2 and 3 with a slag depth of 300 mm. The lines were
lines were computed from the model for a slag depth of 350 mm, and the computed from the model for a slag depth of 350 mm, and the various
various potentials applied to the slag, Vs, indicated. potentials applied to the slag, Vs, indicated.

weak and the magnetic Reynolds number, Rem, is small;

hence, the fluid flow and current flux equations may be
decoupled. The conservation of current may be written as
¹ z J 5 ¹ z (ss¹V) 5 0 [2]
which is the familiar Laplace equation. As noted in Section
IV, the temperature is uniform throughout most of the slag;
thus, the conductivity will be uniform. If the conductivity
is uniform, then it may be dropped from Eq. [2]. This equa-
tion was solved in three dimensions for various potentials
in contact with the slag in Figures 8 through 10. The bound-
ary conditions were the same as those described in Part II.

B. Electric Potential Drop at Electrode Surface

It is indeed striking that the measured electric potentials Fig. 11—Measured vertical temperature profile in the electric furnace at
only agree with those calculated by Eq. [2] if the electric location A.
potential applied to the slag is 100 to 120 V less than that
applied to the electrodes. The electric potential applied to electric potential drop occurs in the melt, but is not asso-
the electrodes is measured above the furnace roof. Based ciated with ohmic electric potential drop in the slag. The
on a conductivity of the Soderberg electrodes of 1.5 3 1025 most plausible explanation for this finding is that the elec-
V m,[15] the resistance of an electrode is less than 0.1 mV, tric potential drop occurs at the surface of the electrode due
much lower than the resistance in the slag, which is of the to arcing. Each electrode is consumed at a rate of 5.8 kg/h
order of 10 mV. Therefore, ohmic losses in the electrode at a smelting rate of 35 tonne/h by the reduction of oxides
are not responsible for the discrepancy. in the slag. The electrode carbon is most likely converted
The power associated with the difference in electric po- to carbon monoxide, yielding gas evolution rates of about
tential applied to the electrode and that experienced by the 1.66 3 1022 m3/s at each electrode. The current must find
slag must be used in the furnace. If this were not the case, its way through or around the gas. Given high temperatures,
then the heat demands of the smelting processes could not voltages, and current densities, the gas is likely ionized to
be met. Therefore, one is left with the conclusion that the create an arc or plasma. Figure 12 shows the equivalent

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 29B, FEBRUARY 1998—81

 the resistance of the arc increases with the applied potential.
Figure 1 shows a comparison between the resistance of the
slag alone and the measured resistance; the resistance due
to arcing is not a strong function of immersion depth.
There is no published work on electric potential drops in
carbon monoxide arcs; recent unpublished work at Mintek
in South Africa[16] indicates that the potential drop over a
0.05-m gap of carbon monoxide in a smelting environment
may be 70 to 190 V. While it is not possible to directly
compare their work with the present, it demonstrates that
arcing results in significant potential gradients.
As outlined in Section I, immersed-arc smelting is the
only technique generally considered not to arc. The findings
of the present investigation strongly suggest that there is an
arc. However, it would be very difficult to make measure-
ments near the electrode to obtain direct evidence of the
arc and its shape. Furthermore, arcing may occur in other
furnaces in which a graphite electrode is consumed because
gas must be created, and under smelting conditions, it is
likely ionized.

C. Assessment of Analog Correlations

Previous correlations[6,7,8] were developed by fitting re-
gression equations to results obtained by the measurement
of resistance in low-temperature, small-scale analog models
of full-scale furnaces. In those models, there was no arcing
at the electrodes, so these models cannot accurately portray
the real electrodes in that respect.
Fig. 12—Illustration of the equivalent circuit for the electric furnace In principle, there should be no difference between the
showing that extra resistance at the electrode surface, Re, results in a
difference between the potential applied to the electrode, Ve, and the data from the analog model and from the solution of the
potential experienced by the slag, Vs. Laplace equation (Eq. [2]) for the determination of the
ohmic portion of the resistance of the slag. Figure 1 shows
a comparison between the analog correlations and the so-
lution of Eq. [2] for the slag. The agreement is good for
the deeper immersion depths, but is poor for shallow ones.
This is probably a result of extrapolation of the correlations
outside their range of validity. There are other potential
problems with these correlations.
(1) In situations where the conductivity of the slag varies
greatly, due to composition or temperature gradients,
the constant conductivity of the analog correlations lim-
its their applicability.
(2) Where the geometry is significantly different from the
analog model, there may also be difficulties, such as
the shallow immersion shown in Figure 1. The resis-
tance is considerably larger in such cases, due to higher
current density.
Fig. 13—Resistance as a function of the potential applied to the electrodes.
The hatched region shows the range of resistance between electrodes Therefore, it is preferable to solve the appropriate current
measured in the plant. The line shows the ohmic resistance of the slag equations rather than employ analog models, particularly
calculated with the model. The difference between the two is attributed
to the arcing. when noting the ease with which this can be done with
modern computers.
circuit for this phenomenon. There is a large electric po-
D. The Measurement Technique
tential drop at the electrode surface, (Vc 2 Vs), due to the
resistance between the slag and the electrodes. This electric This is the first time that the electric potential distribution
potential drop will dissipate heat in the immediate vicinity in an electric smelting furnace has been reported. The probe
of the electrodes. Figure 13 shows the resistance calculated designed for the electric potential measurement worked
for the slag alone with Eq. [2] compared with plant data. well. For the normal slag temperature of 1250 7C, the probe
The calculated resistance of the slag alone is virtually con- functioned in the slag for up to 1 hour. Its life was limited
stant, whereas the actual resistance increases with increas- by erosion of the lowest electrode in the matte; erosion of
ing electric potential applied to the electrodes. Apparently, the electrodes in the slag was negligible.

82—VOLUME 29B, FEBRUARY 1998 METALLURGICAL AND MATERIALS TRANSACTIONS B

 The accuracy of the electric potential measurement was Re extra resistance between slag and electrode
affected by several factors such as the location of the elec- (ohm)
tric potential probe and the choice of the ground point. The Rs resistance of slag alone (ohm)
building ground was not suitable; much more stable read- Rem magnetic Reynolds number, LU/am (—)
ings were obtained when the electrode immersed in the Vc electric potential applied to electrode (V)
matte was used as the reference. Some problems in repro- Vs electric potential applied to slag (V)
ducibility were encountered in exactly repositioning the U velocity scale (m/s)
probe at the same location due to the length of the probe. am magnetic diffusivity 1/ssmm
As will be shown in Part II, the electric potential is very ss electrical conductivity of slag (mho/m)
sensitive to location. mm magnetic permeability (Henry/m)

VI. CONCLUSIONS
REFERENCES
A novel multielectrode probe was developed to simul-
taneously measure the voltage at several locations in the 1. A.G. Matyas, R.C. Francki, K.M. Donaldson, and B. Wasmund: CIM
slag of an electric furnace. Consequently, this is the first Bull., 1993, vol. 86 (972), pp. 92-99.
2. J.C. Taylor and H.R. Traulsen: World Survey of Nonferrous Smelters,
report of such measurements. By comparison with com- Proc. Symp. on World Survey of Nonferrous Smelters at the TMS-
puted voltage distributions, it was deduced that there is a AIME Annual Meeting, Jan. 25–29, 1988, Phoenix, AZ.
substantial voltage drop at the electrode interface, 100 to 3. C. Diaz, B.R. Conard, C.E. O’Neill, and A.D. Dalvi: CIM Bull., 1994,
120 V for applied potentials of 180 to 230 V for the Fal- vol. 87 (98), pp. 62-68.
4. Y.Y. Sheng, G.A. Irons, and D.G. Tisdale: Metall. Mater. Trans. B,
conbridge furnace. This potential drop was attributed to the 1998, vol. 29B, pp. 85-94.
creation of an arc from the carbon monoxide evolved at the 5. S.S. Attwood: Electric and Magnetic Fields, John Wiley & Sons, New
electrodes due to chemical erosion. Therefore, heat is York, NY, 1949, pp. 88 and 160.
evolved both at the electrode surface and in the bulk of the 6. J.A. Perssan and D.G. Treilhard: J. Met., 1973, Jan., pp. 34-39.
slag by ohmic heating. This improved understanding of the 7. R.C. Urquhart, M.S. Rennie, and C.C. Rabey: Proc. Int. Symp. on
Copper Extraction and Refining, Feb. 22–26, 1976, Las Vegas, NV,
heat evolution was used in Part II. A.K. Biswas and W.G. Davenport, eds., TMS-AIME, Warrendale,
PA, 1976.
8. Q. Jiao and N.J. Themelis: Metall. Trans. B, 1988, vol. 19B, pp. 133-
ACKNOWLEDGMENTS 40.
9. G. O’Connell: ‘‘Relationship between Slag Depth, Electrode
The authors are very grateful to Falconbridge, Limited, Immersion, and Electrode Resistance in Falconbridge’s Electric
for funding this project and permitting publication of this Furnaces,’’ Inter-office Technical Report, Metallurgical Technology
article. The encouraging comments from Falconbridge per- Centre. Falconbridge, Sudbury, ON, Canada, May 31, 1993.
10. Q. Jiao and N.J. Themelis: Metall. Trans. B, 1991, vol. 22B, pp. 183-
sonnel including Mr. Ian Cameron, Mr. Chai Tan, and Dr. 92.
Gary O’Connell are greatly appreciated. The assistance of 11. N.M. Stubina, G.H. Kaiura, H. Tseng, and J.M. Toguri: ‘‘To Develop
Mr. Owen Kelly and Falconbridge’s smelter technicians a Clean Smelting Process for Highly Roasted Copper-Nickel Calcines
and furnace operators during the experiments is also grate- and to Improve Metal Recovery, Part D: Electrical Conductivity
fully acknowledged. Measurements of Electric Furnace Smelting Slags,’’ DSS File No.
15SQ.23440-3-9172, University of Toronto, Toronto, 1988.
12. N. Stubina, J. Chao, and C. Tan: CIM Bull., 1994, June, pp. 57-61.
13. J.D. Kraus and K.R. Carver: Electromagnetics, 2nd ed. McGraw-Hill
NOMENCLATURE Book Company, New York, NY, 1973.
fg geometric factor of the electric furnace (l/m) 14. W.F. Hughes and F.J. Young: The Electromagnetodynamics of Fluids,
He electrode immersion in slag (m) Wiley, New York, NY, 1966.
Hs slag thickness (m) 15. J.R. Jones, M.A. Laughton, and M.G. Say: Electrical Engineer’s
Reference Book, 15th ed., Butterworth-Heinemann Ltd., Oxford,
I electrical current (A) United Kingdom, 1993.
J conduction current density (A/m2) 16. R. Jones: Mintek, Randburg, South Africa, private communication,
L length scale (m) 1996.

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 29B, FEBRUARY 1998—83