Вы находитесь на странице: 1из 35

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/288727257

Basic principles of sulfide smelting and converting with oxygen-rich gas

Article · January 2002

CITATIONS READS
3 202

3 authors, including:

Hong Yong Sohn Manuel Pérez-Tello


University of Utah Universidad de Sonora (Unison)
725 PUBLICATIONS   6,345 CITATIONS    40 PUBLICATIONS   230 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Development of a Novel Flash Ironmaking Technology View project

Multidimensional simulation of ventilation in an underground mining face View project

All content following this page was uploaded by Manuel Pérez-Tello on 06 January 2016.

The user has requested enhancement of the downloaded file.


BASIC PRINCIPLES OF SULFIDE SMELTING AND CONVERTING
WITH OXYGEN-RICH GAS

Kimio Itagaki1), Hong Yong Sohn2), and Manuel Pérez-Tello3)

1)
Institute of Multidisciplinary Research for Advanced Materials,
Tohoku University, Katahira 2-1-1, Aoba-ku 980-8577, Sendai, Japan
2)
Department of Metallurgical Engineering, University of Utah,
Salt Lake City, Utah 84112, U.S.A.
3)
Department of Chemical Engineering and Metallurgy, University of Sonora,
Hermosillo, Sonora 83000, Mexico

Abstract

Thermodynamics and rate process fundamentals are two of the important principles
involved in nonferrous production processes. Process configuration establishes whether
thermodynamics or rate process behavior is most relevant in a particular application. In this
paper, thermodynamic theory and rate process fundamentals are used to analyze the main
characteristics of smelting and converting processes. The effect of increased oxygen potential is
examined with emphasis in view of the increased use of highly oxygen-enriched gas in modern
smelting processes.

The phase equilibrium and the distribution of minor elements between copper or nickel
matte and FeOx-SiO2 based or FeOx-CaO based slag are reviewed. The use of oxygen-rich gas
has no serious effects on the recovery of valuable copper and silver in the matte phase, while the
recovery of nickel and cobalt is adversely affected. The oxygen content in the gas also strongly
affects the distribution behaviors of arsenic, antimony and bismuth.

Analysis of the rate processes occurring in the shaft of flash smelting and converting
furnaces by a 3-D computational fluid-dynamics model is also discussed. After being validated
with experimental data collected in a large laboratory furnace, the computer model has been
used to simulate the main features of an industrial flash converting operation. The results
present quantitative features of the improved mixing of the particle-gas suspension and
increased oxygen efficiency obtained with a distributor-cone burner compared with those with a
single axial entry burner. The role of basic principles in the engineering analysis of
high-intensity matte smelting processes is also discussed.
Introduction

Many pyrometallurgical processes have been and still are carried out at or close to
equilibrium conditions due to the high process temperature and the intermittent nature of
operation. Modern smelting processes must allow higher productivity and must thus operate at
higher intensity, which may not allow time for equilibrium. Therefore, a large number of reports
on rate processes including chemical kinetics and transport phenomena related to
pyrometallurgical operations have started to appear in the literature. High intensity has been
achieved by the application of high levels of oxygen enrichment, continuous operation, and
improved equipment reliability. New high-intensity processes have been developed for a
number of smelting operations. Although these processes are all complex in nature and each
contains distinctive features, they can all in principle be analyzed in terms of established
fundamental laws. The study of high-intensity smelting processes from a fundamental
standpoint is relevant for the control and optimization of industrial operations. In addition, such
studies provide a fundamental understanding of the mechanisms governing these systems, and
the accumulated knowledge is useful in the development of new smelting technologies.

Of particular relevance in high intensity smelting processes is the development of


mathematical formulations capable of reproducing the important phenomena occurring in the
process units. Mathematical tools of this kind are useful because they provide a quantitative
description of the process behavior, thus eliminating subjective and often fallible human
judgment. Upon validation with experimental data, well-constructed mathematical models can
be used to simulate the process behavior under different operating conditions, thus replacing
experimentations that are costly and time-consuming in commercial or pilot plants. A review of
the recent advances in the mathematical modeling of various smelting technologies has been
presented by one of the authors [1]. Mathematical modeling of pyrometallurgical processes has
become necessary because they have become continuous and larger with greatly increased
throughput rates. Therefore, even a small increase in the efficiency and/or yield now means a
much greater profit than before. Simultaneously, large and fast computers have become
available so that the solution of extremely complicated governing equations can now be
accomplished within a reasonable amount of computing time.

Recent efforts in the mathematical modeling of sulfide smelting systems have included
the following topics [1]:

(1) Flash smelting and converting reactor models,


(2) QSL-type channel reactor models,
(3) Thermodynamic equilibrium models, and
(4) Minor element behavior models.

The principles used in the analysis of the above systems are strongly dependent on the
process configuration. Flash smelting and converting reactors are mostly controlled by rate
processes, such as turbulence, heterogeneous reaction kinetics, and heat and mass transfer
occurring simultaneously in the reaction shaft. On the other hand, bath smelting processes are
mostly controlled by mass transfer processes with equilibrium among the phases. Fundamental
relationships involved in the mathematical models of these systems must be carefully chosen
and the models validated with experimental data before they are used for design, optimization,
and scale-up of large-scale operations. In this paper, the use of fundamental laws in the analysis
of high-intensity sulfide smelting and converting processes is discussed. The first discussion
addresses the thermodynamic aspects of the production of molten mattes of copper and nickel
using oxygen-rich gas, including minor element behavior. The second discussion involves the
analysis and simulation of the rate processes in flash smelting and converting reactors.

Thermodynamic Study of Matte Smelting With Oxygen-Rich Gas

The use of oxygen-rich gas for smelting sulfide concentrates enables intensification of
the smelting processes and affords a high productivity. Therefore, a combination of intensive
reactors with the use of oxygen-rich gas has led to a number of improved copper and nickel
smelting processes. The phase equilibrium relations for slag and copper or nickel matte under
high partial pressures of SO2 are of practical importance for oxygen-smelting processes. The
distribution behavior of minor elements in the copper or nickel concentrates between the slag
and matte phases under high partial pressures of SO2 is also of great concern. However, few data
are reported on the phase relations and the distribution of minor elements under high partial
pressures of SO2 - greater than 0.1 atm. Hence, a series of experimental and analytical studies of
the slag/matte equilibria [2-11] has recently been conducted to extend the slag/matte
equilibrium data from the air blowing condition at about 0.1 atm pSO2 to the oxygen-smelting
condition, the results of which are reviewed in this paper.

Oxygen-Sulfur Potential Diagram

An oxygen-sulfur potential diagram is useful for fundamental thermodynamic


understanding of copper and nickel matte smelting [12]. The log pO2 - log pS2 relationship at
specified matte grades (mass % Cu or Ni in the matte) is given by Equation (2):

FeS(l) + 1/2O2(g) = FeO(l) + 1/2S2(g) (1)

log pO2 = log pS2 - 2 log K1 + 2 log (aFeO/aFeS) (2)

where K1 is the equilibrium constant of Reaction (1), and aFeO and aFeS are the activities of FeO
and FeS, respectively. The log pO2 - log pS2 relationship in a region of pSO2 higher than 0.1 atm is
of practical importance for understanding copper and nickel smelting with oxygen-rich gas.
Although some experimental studies [13-17] have been conducted in a limited range of pSO2
around 0.1 atm, there are few data for higher pSO2. Tavera and Davenport [18] studied the phase
equilibrium between the fayalite slag and copper mattes under partial pressures of SO2 up to 1
atm but without controlling the partial pressures of sulfur and oxygen.

It is difficult to control high partial pressures of SO2 and S2 by conventional methods


with a gas mixture of CO-CO2-SO2-N2 or H2-CO2-SO2-Ar. Improved control of the partial
pressures of SO2 and S2 in the gas phase was achieved by using a gas mixture of Ar-SO2-S2 [10].
Consequently, pO2 was also fixed at a given value based on the equilibrium relationship for the
reaction 0.5S2(g) + O2(g) = SO2(g). The mixture of SO2 and Ar gases was passed through a
reservoir with liquid sulfur that was kept at a given temperature. After passing the reservoir, the
gas was led to the slag/matte reaction system at 1250 or 1300oC through a glass tube that was
heated at 145-305oC to prevent the condensation of sulfur. The relationship between the partial
pressure of S2 at 1250 or 1300oC and the temperature of the reservoir was determined in advance.
In the present studies, an FeOx-SiO2 based slag saturated with silica or an FeOx-CaO slag was
equilibrated with copper or nickel mattes in an MgO crucible at 1250 or 1300oC. The MgO
content in the equilibrated FeOx-SiO2 based slag was about 5 mass % at 1300oC, while that in
the FeOx-CaO slag was about 2 mass% at 1250oC. The slags initially contained no other
components. For the SiO2-FeOx slags, x changed from 1.02 at pO2 of 10-9 atm to 1.12 at pO2 of
10-6 atm; while, for the CaO-FeOx slags, x changed from 1.07 at pO2 of 10-9 atm to 1.38 at pO2 of
10--6 atm.

The relationship between matte grade and pS2 as well as pO2 for the equilibration of the
FeOx-SiO2 based slag with Cu2S-FeS mattes and Ni3S2-FeS mattes are shown for given pSO2 of
0.1, 0.5 and 1 atm in Figures 1 and 2, respectively. It is noted in Figure 1 that pS2 decreases with
increasing matte grade, while pO2 increases, and the decrease in pS2 and the increase in pO2 are
remarkable when the matte grade is over 75 mass % Cu. At a given matte grade, pO2 and pS2
Figure 1. Relation between oxygen or sulfur Figure 2. Relation between oxygen or sulfur
partial pressure and copper content in the partial pressure and nickel content in the
Cu2S-FeS matte equilibrated with the silica- Ni3S2-FeS matte equilibrated with the silica-
saturated FeOx-SiO2 based slag under pSO2 of saturated FeOx-SiO2 based slag under pSO2 of
0.1, 0.5 and 1 atm at 1300ºC. 0.1, 0.5 and 1 atm at 1300ºC.

increase with increasing pSO2. For the slag/Ni3S2-FeS matte equilibration, pS2 decreases and pO2
slightly increases with increasing matte grade up to 45 mass % Ni and both change considerably
at higher matte grades approaching 73.3 mass % Ni (Ni3S2). At a specified matte grade, pO2 and
pS2 increase with increasing pSO2. The relationships for the Cu2S-FeS and Ni3S2-FeS mattes
under pSO2 of 0.1 atm agree well with those reported by Kaiura et al. [13] and Celmer [17],
respectively.

Sulfur-oxygen potential diagrams at 1300oC, obtained from the experimental results of


Figures 1 and 2, are illustrated in Figures 3 and 4 for the equilibration of the FeOx-SiO2 based
slag with Cu2S-FeS matte and Ni3S2-FeS matte, respectively. For each matte grade, the log pO2 -
log pS2 relationship is essentially a straight line with a slope of unity. This indicates that the
terms (aFeO/aFeS) and (pO2/pS2) in Equation (2) are constant against pSO2 at each matte grade, and
has an important implication, as described in the following discussion on the dissolution of
copper and nickel in the slag.
Figure 3. Sulfur-oxygen potential diagram Figure 4. Sulfur-oxygen potential diagram
for the Cu-Fe-S-O-SiO2 system with for the Ni-Fe-S-O-SiO2 system with
specified matte grades in relation to pSO2 at specified matte grades in relation to pSO2 at
1300ºC. 1300ºC.

Solubility of Copper and Nickel in Slag

The solubility of copper in the FeOx-SiO2 based slag was determined at 1300oC under
pSO2 of 0.1, 0.5 and 1 atm for various matte grades, as shown in Figure 5. The copper content in
the slag first increases with increasing matte grade, reaches a maximum at around 20 – 30 % Cu
in the matte, and decreases slightly between 30 and 60 % Cu. The copper content then increases
abruptly as matte grade increases further. It is noteworthy that the solubility is relatively
independent of pSO2 within the scatter of the data. The present results under pSO2 of 0.1 atm agree,
as a whole, with those by Yazawa and co-workers [19], who reported a maximum at about 35 %
Cu in the matte. As shown in Figure 6, the nickel content in the FeOx-SiO2 based slag, which
was determined at 1300oC under pSO2 of 0.1, 0.5 and 1 atm, increases slightly with matte grade
up to about 38 % Ni, and rapidly above this matte grade. A similar relationship was found by
Celmer [17] at 1250oC under pSO2 of 0.1 atm. It is noteworthy that the nickel solubility is
dependent on pSO2 and increases with increasing pSO2 at a specified matte grade.

The solubilities of copper and nickel in the FeOx-CaO slag equilibrated with Cu2S-FeS
and Ni3S2-FeS mattes are shown for various matte grades in Figures 7 and 8, respectively. It was
found that phase separation into matte and slag does not occur at matte grades lower than the
lowest value for each pSO2 value shown in these figures, and that the lowest matte grade at which
phase separation occurs increases with increasing pSO2. It is also noteworthy for the FeOx-CaO
slag that copper solubility is independent of pSO2 at a specified matte grade, while nickel
solubility is dependent.
Figure 5. Solubility of copper in the Figure 6. Solubility of nickel in the
silica-saturated FeOx-SiO2 based slag silica-saturated FeOx-SiO2 based slag
equilibrated with the Cu2S-FeS matte equilibrated with the Ni3S2-FeS matte
against matte grade under pSO2 of 0.1, against matte grade under pSO2 of 0.1, 0.5
0.5 and 1 atm at 1300ºC. and 1 atm at 1300ºC.

Figure 7. Solubility of copper in the Figure 8. Solubility of nickel in the


FeOx-CaO slag equilibrated with the FeOx-CaO slag equilibrated with the
Cu2S-FeS matte against matte grade Ni3S2-FeS matte against matte grade
under pSO2 of 0.1, 0.5 and 1 atm at under pSO2 of 0.1, 0.5 and 1 atm at
1250ºC. 1250ºC.
The observed difference in the solubilities of copper and nickel is worth examining on
the basis of thermodynamics. The solubility of copper in the slag, (%Cu), is considered to be
given by the sum of an oxidic part, (%Cu)o, and a sulfidic part, (%Cu)s, as follows [20, 21].
From the mono-metallic reaction given by

CuS0.5(l) + 1/4O2(g) = CuO0.5(l) + 1/4S2(g), (3)

the oxidic part can be expressed as

(%Cu)o = aCuS0.5(A/(γCuO0.5))(pO2/pS2)1/4 (4)

where aCuS0.5 is the activity of CuS0.5 component, (γCuO0.5) is the activity coefficient of CuO0.5 in
the slag, and A = MCu nt K3, where MCu is the atomic weight of copper, nt is the total number of
moles of the components (mono-metallic species) in 100 g of the slag, and K3 is the equilibrium
constant of Reaction (3). On the other hand, the sulfidic part is given by

(%Cu)s = aCuS0.5(B/(γCuS0.5)), (5)

where (γCuS0.5) is the activity coefficient of CuS0.5 component in the slag, and B = MCu nt. By
combining Equations (4) and (5), the solubility of copper is given by:

(%Cu) = aCuS0.5[(A/(γCuO0.5))(pO2/pS2)1/4 + B/(γCuS0.5)]. (6)

When both (γCuO0.5) and (γCuS0.5) in the slag equilibrated with a given matte are constant
with pSO2, (%Cu) will be independent of pSO2 because the (pO2/pS2)1/4 term in Equation (6) does
not vary with pSO2, as shown in Figure 3. Hence, (%Cu) will be dependent only on aCuS0.5, in
other word, matte grade. This thermodynamic prediction is in accordance with the results shown
in Figures 5 and 7 in that (%Cu) is independent of pSO2 for a fixed matte grade. Conversely, it
may be concluded from the present results that (γCuO0.5) and (γCuS0.5) in the slag equilibrated with
a given matte are constant with pSO2.

The solubility of nickel in the slag equilibrated with a Ni3S2-FeS matte is also given by
the sum of the oxidic part, (%Ni)o, and the sulfidic part, (%Ni)s. Based on the following reaction
that is written by taking the Ni3S2 component in a mono-metallic form of NiS2/3,

NiS2/3(l) + 1/2O2(g) = NiO(l) + 1/3S2(g), (7)


the oxidic part is given by

(%Ni)o = aNiS2/3(C/(γNiO))(pO21/2/pS21/3) (8)

where aNiS2/3 is the activity of NiS2/3 component, (γNiO) is the activity coefficient of NiO
component in the slag and C = MNi nt K7. On the other hand, based on the reaction

NiS2/3(l) + 1/6S2(g) = NiS(l), (9)

the sulfidic part is given by

(%Ni)s = aNiS2/3(D/(γNiS)) pS21/6 (10)

where (γNiS) is the activity coefficient of NiS component in the slag and D = MNi nt. By
combining Equations (8) and (10), the solubility of nickel in the slag is given by

(%Ni) = aNiS2/3[(C/(γNiO))(pO21/2/pS21/3) + (D/(γNiS))pS21/6]. (11)

As suggested in Figure 2, the terms (pO21/2/pS21/3) and pS21/6 increase with increasing pSO2
at a specified matte grade. This results in the increase of (%Ni) according to Equation (11). The
result of this thermodynamic analysis is also in accordance with the present experimental results,
as shown in Figures 6 and 8.

In summary, the difference in the dependency on pSO2 of copper and nickel solubilities is
ascribed to the difference in the chemical formulas of the mono-metallic sulfides in that one
copper atom overall combines with a half sulfur atom, while one nickel atom does with 2/3 atom
of sulfur.

Minor Element Distribution between Slag and Matte

The distribution ratios of several minor elements X between the FeOx-SiO2 based slag
and matte, defined as Lxs/m = (mass % X in slag)/(mass % X in matte), have been determined at
1300oC under pSO2 of 0.1, 0.5 and 1 atm. The results for silver and arsenic are shown in Figures
9 and 11 for the Cu2S-FeS matte, and Figures 10 and 12 for the Ni3S2-FeS matte, respectively.
As shown in Figures 9 and 10, LAgs/m shows minimum values at high matte grades and, at a
Figure 9. Distribution ratio of Ag Figure 10. Distribution ratio of Ag
between the silica-saturated FeOx-SiO2 between the silica-saturated FeOx-SiO2
based slag and the Cu2S-FeS matte based slag and the Ni3S2-FeS matte
against matte grade at 1300ºC. against matte grade at 1300ºC.

Figure 11. Distribution ratio of As Figure 12. Distribution ratio of As


between the silica-saturated FeOx-SiO2 between the silica-saturated FeOx-SiO2
based slag and the Cu2S-FeS matte based slag and the Ni3S2-FeS matte
against matte grade at 1300ºC. against matte grade at 1300ºC.

specified matte grade, it is independent of pSO2 within the scatter of the data. As shown in
Figures 11 and 12, LAss/m values are much smaller at higher matte grades than at lower matte
grades and, at a given matte grade, they increase with pSO2. It was clarified by Itagaki and
coworkers [6, 10] that LSbs/m presents a behavior similar to that of LAss/m, while LBis/m increases
with matte grade at high matte grades and, at a given matte grade, both LSbs/m and LBis/m increase
with pSO2. It was reported by Itagaki and coworkers [6, 11] that the distribution ratio of cobalt
between the FeOx-SiO2 based slag or the FeOx-CaO slag and the Ni3S2-FeS matte increases with
increasing pSO2 at a fixed matte grade.

The behaviors of the distribution ratio at different partial pressures of SO2 can be
analyzed on the basis of thermodynamics, as follows. When the metallic, sulfidic and oxidic
dissolution of a minor element X in the slag and matte phases is taken into consideration, the
corresponding reactions are given by

{XSυ/2} = (XSυ/2); (aXSυ/2) / {aXSυ/2} = 1 (12)

{XSυ/2} + υ/4 O2(g) = (XOυ/2) + υ/4S2(g); (aXOυ/2)/{aXSυ/2} = K13(pO2/pS2)υ/4 (13)

{X} + υ/4O2(g) = (XOυ/2); (aXOυ/2)/{aX} = K14pO2υ/4 (14)

{X} + υ/4S2(g) = (XSυ/2); (aXSυ/2)/{aX} = K15pS2υ/4 (15)

where ( ) and { } denote the slag and matte phases, respectively, and υ is the valence. It was
clarified in the present study that the change of pSO2 at a specified matte grade results in the
change of pO2 as well as pS2, as shown in Figures 1 and 2, while the (pO2/pS2) ratio is constant
with pSO2, as shown in Figures 3 and 4. Equations (12) through (15) provide a thermodynamic
basis for determining which component prevails in the matte phase, as illustrated by Font et al.
[6]. It was suggested that, for As, Sb and Bi, the metallic species are more stable in the matte
phase than the sulfide species, while the sulfide was for Ag. Yazawa [22] reported that the
metallic species do not exist in the slag phase under high pO2 around 10-8 atm, which
corresponds to the present equilibration between the slag and matte phases. Hence, the metallic
dissolution in slag is excluded in the present thermodynamic analysis.

Similarly to Equations (3) and (5) for copper, Equations (12) and (13) were selected for
silver in the present discussion. By rearranging these equations and converting the mole
fractions of AgS0.5 and AgO0.5 to mass %, the following equation is obtained:

LAgs/m = ((nt)/{nt}){γAgS0.5}(1/(γAgS0.5) + K13(pO2/pS2)1/4 /(γAgO0.5)) (16)

The first and second terms on the right-hand side of Equation (16) represent the contributions of
sulfidic and oxidic dissolution, respectively. Here, {nt} and (nt) are the moles in 100 g of matte
and slag phases, respectively. As shown in Figures 3 and 4, (pO2/pS2) is constant against pSO2 at a
fixed matte grade. Hence, it may be predicted from Equation (16) that LAgs/m is constant with
pSO2 at a given matte grade, provided that the activity coefficients {γAgS0.5} and (γAgO0.5) are
independent of pSO2. The present experimental results, shown in Figures 9 and 10, follow this
prediction.

For arsenic, antimony and bismuth, oxidic dissolution prevails in the slag, and the
distribution ratio is given by

LXs/m = ((nt)/{nt}){γX}(K14pO23/4/(γXO1.5)) (17)

where X represents As, Sb or Bi. As shown in Figures 1 and 2, pO2 in Equation (17) increases
with pSO2 when matte grade is fixed. Hence, it may be predicted from Equation (17) that LXs/m
for As, Sb and Bi increases with increasing pSO2, provided that the activity coefficients {γX} and
(γXO1.5) are independent of pSO2. The present results, shown in Figures 11 and 12, again follow
this prediction.

Fractional Distribution of VA Elements among Gas, Slag and Matte

The fractional distribution of arsenic, antimony and bismuth among the gas, slag and
matte phases during a smelting stage was thermodynamically evaluated with regard to the effect
of using oxygen-rich gas for blowing for mattes of Cu2S-FeS, Ni3S2-FeS and Cu2S-Ni3S2-FeS
with a (Ni/Cu) molar ratio of unity. Detailed methodology for this thermodynamic calculation is
described in Reference [23]. The main reaction taking place in smelting sulfide concentrates,
FeS(l) + 3/2O2(g) = FeO(l) + SO2(g), was finely divided into n-increments before iron-free
matte is obtained. Thermodynamic equilibrium was assumed at the end of each reaction stage
and the mass balance for an impurity element was taken among the gas, slag and matte phases.
The thermodynamic data obtained in the present study and those reported in the literature [23]
were used in the calculation.

The fractional distribution of arsenic is shown in relation to the oxygen enrichment in


blowing gas in Figure 13. The calculation was performed for a charge with 38.1 mass % Fe and
0.3 mass % As smelted at 1300oC to produce a matte with 16.5 mass % Fe. The content of
arsenic in the produced matte is also shown with a dash-dot line in Figure 13. At a specified
oxygen enrichment, the proportion into the gas phase, (As)g, and that into the slag phase, <As>s,
increase in the order of Ni3S2-FeS, Ni3S2-Cu2S-FeS and Cu2S-FeS mattes. {As}m and arsenic
content in the produced matte decrease in the same order. These behaviors are attributed mainly
to the extremely low activity coefficients of arsenic in the nickel and nickel-copper mattes in
Figure 13. Fractional distribution of As among the gas, FeOx-SiO2 based slag and
matte phases against degree of oxygen-enrichment in blowing gas when a charge
with 38.1 mass% Fe and 0.3 mass% As is smelted at 1300ºC to produce a matte
with 16.5 mass% Fe: (a) Ni3S2-FeS matte, (b) Ni3S2-Cu2S-FeS (Cu/Ni mole ratio
of 1) matte, (c) Cu2S-FeS matte systems.

comparison with that in the copper matte [9]. It is noteworthy that (As)g decreases while <As>s
increases with increasing oxygen enrichment. The decrease of (As)g is ascribed to the decrease
in the amount of exhaust gas, while the increase of <As>s is due to the increase in pO2 with
increasing pSO2 (which increases with oxygen enrichment) as suggested in Figures 1 and 2, and
the resultant increase of LAss/m is according to Equation (17).

The fractional distribution of antimony is shown in Figure 14, which was calculated
under the same condition as for the arsenic distribution. At a specified oxygen enrichment, (Sb)g
and <Sb>s also increase in the order of Ni3S2-FeS, Ni3S2-Cu2S-FeS and Cu2S-FeS mattes. It is
noted that the content of antimony in the produced matte does not change significantly with the
oxygen enrichment because the increase in the amount going into the slag phase is comparable
to the decrease in (Sb)g. The distribution of bismuth, as shown in Figure 15, is similar to that of
arsenic in Figure 13. The content of bismuth in the produced Cu2S-FeS matte increases with
oxygen enrichment due to the decrease in (Bi)g and a moderate increase in <Bi>s. These results
suggest that elimination of arsenic, antimony and bismuth by volatilization becomes less
favorable with increasing oxygen enrichment in the blowing gas, while elimination by slagging
becomes more favorable.

Using other sources [24-26] of thermodynamic data, Kim and Sohn [27] carried
out a computer simulation for predicting the minor-element behaviour in copper smelting
and converting at up to 84.5 % oxygen enrichment in comparison with the condition of air
Figure 14. Fractional distribution of Sb. (The conditions are the same as for Figure 13.)

Figure 15. Fractional distribution of Bi. (The conditions are the same as for Figure 13.)

blowing. Although the calculations were made under different temperatures, feed compositions,
and minor elements contents, the results showed clear dependencies on the oxygen enrichment
that are in close agreement with the presentresults shown in Figures 13-15. It was predicted that
the overall elimination of arsenic and bismuth decreased greatly because of a reduction in gas
volume, but elimination by slagging increased, as the degree of oxygen enrichment increased.
The effecton antimony was similar but more moderate.

Computational Fluid Dynamics Modeling of Flash Smelting


and Flash Converting Furnaces

The flash smelting process has reached a mature status in the nonferrous smelting
industry over the last twenty years. There are currently over forty Outokumpu flash smelting
furnaces in operation producing about 40 % of the world’s primary copper [28]. The process is
also used for the smelting of nickel concentrates. Most recently, the principles of the
Outokumpu flash smelting process have been applied to the converting step of coppermaking.
The new process developed by Kennecott and Outokumpu is called the Kennecott-Outokumpu
Flash Converting Process [29]. The flash converting process overcomes the limitations inherent
in handling molten matte in the widely used Peirce-Smith converters [30]. The flash converting
process is based on solidifying the molten matte from the smelting step and feeding this solid
matte to an oxygen-fed flash furnace. The first commercial operation of the flash converting
process was started up in June 1995 [31].

In spite of the increasing industrial stature of the flash smelting process, it has been
relatively recently that realistic computer simulations of the process have been made. The main
reasons for this are the complexity of the process and the fact that the computing capacity to
solve complicated equations describing such a process has become available only recently. For
flash converting, few computational studies have been developed during the course of its
development and implementation at industrial scale.

The configurations of the flash smelting and flash converting furnaces are essentially the
same. Fine concentrate or matte particles mixed with flux particles are injected into the furnace
shaft together with a high-strength oxygen-enriched gas stream. In the reaction shaft, the
particles are quickly dispersed by the turbulent gas jet, heated by the reactor walls and the gas
until ignition, partly oxidized, and melted. The molten particles settle in the furnace bottom,
where incompletely oxidized sulfides react with metal oxides to produce matte or blister copper,
and the matte or blister copper and slag phases are separated. In both systems, the residence time
of the particles in the reaction shaft is of the order of seconds, whereas the residence time of the
molten bath in the settler is of the order of hours. As a result, the behavior of the reaction shaft is
mostly controlled by rate processes, whereas the molten bath may be assumed to reach chemical
equilibrium. Here, attention is focused on the analysis of the processes occurring in the particles
from the moment they enter the furnace shaft until they reach the molten bath or are carried
away by the off-gas stream towards the uptake shaft. Further, the characteristics of the gas-phase
surrounding the particles in the shaft, such as gas temperature, oxygen composition, and flow
patterns, are elucidated.

Although in this example only the flash converting furnace is analyzed, the fluid
dynamics and particle flow results would be similar for a flash smelting reactor. On the other
hand, the oxidation reactions experienced by copper concentrate particles and copper matte
particles are different. Thus, appropriate modifications should be made in the present
formulation to incorporate the reaction kinetics of the oxidation reactions of copper concentrate
particles in flash smelting.
Model Formulation

A detailed description of the model formulation and early results have been presented in
previous reports by Perez-Tello et al. [32, 33]. In the present paper, the principles of the
mathematical formulation and recent computational results obtained with the computer code are
presented. A schematic representation of a flash smelting or flash converting furnace for
mathematical modeling purposes is shown in Figure 16. A 3-D framework was used. To
describe the phenomena occurring in the particle-laden turbulent gas jet, the conservation
equations for the transport of mass, momentum, and energy were written for each phase. A
distinctive feature of the present formulation is the incorporation of the particle cloud model
developed by Baxter and Smith [34] to describe particle dispersion. For the flash smelting of
chalcopyrite concentrates, the reaction terms are given in the paper by Hahn and Sohn [35]. For
flash converting, the reaction terms are developed for the reaction path shown in Figure 17.

COPPER M A TTE
PARTICLES

y OXYGEN-ENRICHED AIR

z
x

PARTICLE CLOUDS
O2

REACTION SHAFT

SO2

TO UPTAKE
SHAFT

M O LTEN BATH
SURFACE

Figure 16. Schematic representation of the flash smelting and flash converting
furnace shafts for mathematical modeling purposes.

In Figure 17, the following assumptions have been made: (1) Particles are spherical and
nonporous; (2) the inlet particle composition is represented by the general formula:
Cu2 S • yFeS x + inerts; (3) the oxidation products are Cu2 O , CuO , and Fe3O4 ; (4) particle
temperature is uniform, and (5) once the particles are ignited, the intrinsic kinetics is very fast
compared with the mass transfer rate of oxygen to and within the particle; thus, the mass transfer
rate of oxygen controls the overall reaction rate throughout the reaction path. Two
phase-changes were assumed to occur along the reaction path: (1) The melting of the unreacted
sulfide core at temperature Tmc , and (2) the melting of the oxide shell at temperature Tmo . The
latter temperature is of particular relevance because it represents a shift in the reaction
mechanism. This is because the shrinking-core scheme used before oxide melting is no longer
valid once the particle becomes completely molten. When this happens, the mass transfer rate of
oxygen gas to the particle surface was assumed to control the particle oxidation rate.

MATTE PARTICLE
(Cu 2S• y FeSx )

• • •• • •• • • MATTE CORE
• •• • •• •
•• •
•• • • ••
p < p II ••
• •

• • Cu O + F e 3 O4
A ,s A ,e •• •
••
• •
•• • 2
• • • •
• • • • • • •• •
• •• •

MATTE CORE
• •

• • • •
• • •
• • •
• • • •

• •• • •• • ••• ••
• • •


• • • • •

•••
• • • •

• •• •

••
• •
• •
•• •

Cu 2O + F e 3 O 4
< • •

Tp Tm c •




• ••
•••
•••






•• • • • •• ••
• •
• • •

•• •
• • •
• •
• • •

CuO + F e 3 O4
• • • • •
• • • •
• •
• • • •
• •










• •


• •

MOLTEN MATTE
• •• • •• • • •• •
• • •


• • • • •

•••
• • • •


•• •

••
• •

Tm c < Tp < Tm o ••

Cu 2O + F e 3O4
• •

• •

••••
• • •



• •• •
• •

•••


• •

•• • • • •• ••

• • •

•• •
• • •
• •
• • •

CuO + F e 3 O4
• • • • •
• • • •
• •
• • • •
• •

Tp > Tm o MOLTEN PARTICLE

Figure 17. Proposed reaction path for the oxidation of copper matte particles.

Reaction was assumed to start only after a particle reaches an ignition temperature Tcr
estimated as: Tcr = 0.5(Tin + Tmc ) ; here, Tin is the temperature of incipient reaction measured in
a previous study by Perez-Tello et al. [36], and Tmc is the melting point of the matte. The
computation of Tcr by the above procedure was adopted because it best fitted the experimental
results. Tin was computed from the correlations reported in reference [36]. The melting point of
the matte Tmc was set to 1390 and 1350 K for the 72 and the 58% matte, respectively [37].

Upon ignition, particle oxidation was assumed to occur according to the following
reaction:
3 2 y
Cu2 S • yFeS x + [ + y ( x + )]O2 → Cu2 O + Fe3O4 + (1 + xy ) SO2
2 3 3

∆H1r = −400 − 600 y kJ/mol Cu2 S • yFeS x (18)

Reaction (18) was assumed to take place topochemically, forming a porous crust of
Cu2 O , Fe3O4 , and inerts surrounding the unreacted core of Cu2 S • yFeS x . The porosity of this
layer was estimated from the following expression:
1 / ρ Cu2O + ( y / 3) / ρ Fe3O4 + z / ρ in
ε I = 1− (19)
1 / ρ Cu2 S + y / ρ FeSx + z / ρ in

Along the reaction path, if the oxygen partial pressure at the particle surface became
higher than the equilibrium oxygen partial pressure at the Cu2 O / CuO interface, a second layer
of CuO + inerts was assumed to form according to the following reaction:
1
Cu2 O + O2 → 2CuO ∆H2r = −130 kJ/mol Cu2 O (20)
2
The porosity of this layer was estimated from the following expression:

2 / ρ CuO + ( y / 3) / ρ Fe3O4 + z / ρ in
ε II = 1 − (21)
1 / ρ Cu2 S + y / ρ FeS x + z / ρ in

The shrinking-core expressions to compute the rate of oxygen consumption from


Reactions (18), (20), and total, are summarized in Table I. The derivation of these expressions is
detailed elsewhere [32, 33]. As the particle travels through the flow field, it can shift from one
case in Table I to another.

Model Validation

The validation of the computer model was performed by comparing the model
predictions with the experimental data collected in a large laboratory furnace [38]. A large
number of validation runs were conducted, and a detailed discussion of the results is presented
elsewhere [32]. A representative sample of the validation runs is shown in Table II. Table III
shows the characteristics of the matte particles tested in the experiments. Table IV shows the
parameters for the computer simulations of the laboratory furnace.
Table I. Shrinking-Core Expressions to Compute the Rate of Oxygen Consumption

CuO Surface R A,T R A,1 R A,2


Present Conditio ( r = rp ) ( r = rc ) ( r = ri )
n
Ap (C A,b − C AI ,e )
(a) No p A,s < p AII,e 1 rp (rp − rc )
R A,T 0
+
k m, A rc DAI ,eff

Ap (C A,b − C AII,e ) Ac (C AII,e − C AI ,e )


(b) Yes p A,s ≥ p AII,e rc (ri − rc ) R A,T − R A,1
1 rp (rp − ri )
+
k m, A ri D AII,eff ri D AI ,eff

Ap (C A,b − C AI ,e )
(c) Yes p A, s < p AII,e  rp   rp  (r − r ) rp (rp − ri )
R A,T 0
1
+    i I c +
k m, A  rc   ri  D A,eff ri D AII,eff

Table II. Parameters for the Validation Runs


Parameter Run No.
1 2
Matte grade, % Cu 72 58
Particle size fraction, µm 74-105 40-75
Oxygen concentration 70 70
in process gas, vol.%
Oxygen-to-matte ratio, 0.25 0.3
kg O2/kg matte
Particle size distribution in feed: 0.136 0.365
volume fraction of the indicated (74) (53)
particle size in µm given in 0.495 0.463
parentheses. (90) (74)
0.277 0.172
(127) (149)
0.092
(179)
-1
Solid feed rate, kg h 3 3
Gas feed rate, kg h-1 1.03 1.24
Gas volumetric flow rate 1.06 1.27
@ 86.1 kPa and 298 K, m3 h-1
Inlet gas axial velocity at nozzle, m s-1 2.53 3.05
Table III. Characteristics of the Matte Particles Used in all the Validation Runs
Characteristic High-grade Low-grade
matte (72% Cu) matte (58% Cu)
Cu2S content, mass fraction 0.928 0.764
FeSx content, mass fraction 0.072 0.236
FeSx/Cu2S molar ratio, y 0.14 0.54
S/Fe molar ratio, x 0.93 0.94
Melting point of sulfide phase, K 1390 1350
Melting point of oxide phase, K 1850 1850

Table IV. General Parameters for the Computer Simulations of the Laboratory Furnace
Parameter Value
Inlet/outlet and reactor geometry:
Primary stream diameter (particles), m 0.02
Secondary stream diameter, m 0.002*
Secondary stream injection angle, degrees 17 †
Outlet diameter, m 0.05
Reactor diameter, m 0.24
Reaction shaft length, m 1.4

Wall temperature, K:
Top wall 1350
Side wall 1350 + 218x - 269x2 £
Bottom surface 710 ‡
Surface emissivity (inlets, outlet, walls, and bottom) 0.85

Primary stream (particles):


Temperature, K 298
Particle velocity, m s-1 2.3 §
Particle density, kg m-3 5400
Number of injection points for cloud model 9
Standard deviation of incoming clouds, m 0.002
Secondary stream (process gas):
Temperature, K 298
Pressure, kPa 86.1
Turbulent intensity 0.2 ¤
Outlet gas temperature, K 710 ‡

* Nozzle diameter (six total)


† Angle with respect to centerline
‡ For radiation calculations only
£ Variable x is the axial distance from top, m
§ Settling velocity computed from Stokes’s law
¤ Estimated
The predicted and experimental values of fractional conversion in the laboratory furnace
are shown in Figure 18(a). The present formulation predicted the results observed in run 1
reasonably well. However, the rate of oxygen consumption for run 2 was overpredicted. Figure
18(b) shows that the agreement between the computed and observed values is reasonable for
Run 1 in terms of the fraction of sulfur remaining in the particles. It is important that the model
predicted the correct axial position where rapid removal of sulfur took place. However, for run 2
the removal of sulfur from the particles was overpredicted..

1.6
Run 1,
Fractional Conversion

1.4
predicted
1.2
Run 1,
1
experimental
0.8
Run 2,
0.6 predicted
0.4 Run 2,
0.2 experimental
0
0 0.5 1 1.5 2
Axial Distance from Top, m

(a)
1
Fraction of Sulfur Remaining

0.8 Run 1,
predicted
0.6 Run 1,
experimental
Run 2,
0.4
predicted
Run 2,
0.2 experimental

0
0 0.5 1 1.5 2
Axial Distance from Top, m

(b)
Figure 18. Predicted and experimental values of (a) fractional conversion, and (b) sulfur
remaining in the particles, along the centerline of the laboratory furnace.

Computer visualization of the oxygen and sulfur dioxide concentration contours showed
that most of the oxidation reactions in the laboratory furnace took place within 40-60 cm below
the burner tip. The computed gas flow field (Figure 19) indicates a strong recirculation pattern
extending 60 cm below the burner tip. Such recirculation promotes backmixing of the gas phase
and the dispersion of the particles in the reaction shaft. The hottest spots in the furnace were
predicted within 60 cm below the burner tip, and corresponded to the reaction zone in which the
reacted particles reached their highest temperature at about 1650 K.

Figure 19. Cross-sectional view of the gas velocity vectors in the laboratory furnace,
six-jet entry system. (The largest vector represents 1.32 m/s.)

Overall, an assessment of the present formulation indicated that a reasonable


representation of the processes occurring in the shaft of the laboratory furnace was obtained.
The complexity of the system being modeled led to a number of simplifications in the treatment
of the gas, particle, and radiation equations. The discrepancy observed between the predicted
and experimental values in Figure 18 is of the same order of magnitude to the uncertainties of
the mathematical simplifications and experimental measurements. Thus, the present
formulation was deemed to be suitable for the analysis and optimization of industrial flash
converting operations.

Simulation of an Industrial Flash Converting Furnace

The present formulation has been used to simulate the behavior of an industrial flash
converting furnace under various operating conditions. Early simulations reported by
Perez-Tello [32] were aimed at analyzing the effects of burner configuration on the overall
features of the flash converting reactor. A number of simulations were made in which two types
of burner configurations (single axial entry and distributor cone which is assumed to inject the
process gas like a showerhead with a maximum 60º angle to the axis at the injector periphery)
were tested. The results showed that the 60º-distributor-cone burner is more efficient than a
single axial entry burner in terms of particle dispersion and overall reaction rate achieved in the
reactor. Further simulations of the flash converting reactor were carried out to obtain additional
results, using the same formulation but a finer computational mesh to describe the system
geometry. Some of the computational results so far obtained are discussed in this paper.

The main goal of the present study was to clarify in greater detail the effect of burner
configuration on the general features predicted for the flash converting furnace. Early
calculations by Perez-Tello [32] indicated that burner configurations providing a strong radial
velocity component to the inlet gas was likely to promote a more efficient use of the reactor
volume. Thus, it is expected that a 90º-distributor-cone burner would be the most efficient of all.
This case corresponds to a horizontal slot through which the process gas is injected radially into
the furnace, similar to the distribution cone developed by Outokumpu. In the present study, a
flash converting furnace featuring an 85º-distributor-cone burner is analyzed and the results
compared with those obtained for a single axial entry burner. (It was computationally difficult to
inject the gas at exactly 90º and the results for 85 to 89º distributor cones were similar.) Tables V
and VI show the general parameters for the computer simulations presented here. The
parameters in these tables were common to all the simulation runs, and approximately
correspond to typical operating conditions in the commercial unit [31]. High-grade (72% Cu)
matte particles with the characteristics reported in Table III were the solid feed. The gas feed
rate reported in Table VI corresponds to the oxygen requirement to oxidize all sulfur in the
particles to sulfur dioxide, and all the iron to magnetite. A computational mesh consisting of
230,500 cells was used to represent the reaction shaft and a portion of the settler freeboard.

Figure 20 shows the velocity fields predicted for both systems. For the single axial entry
burner (Figure 20 (a)), the largest velocity values are located along the centerline. In the regions
between the gas jet core and the furnace walls, the gas moves more slowly. This is indicated by
both the length of the velocity vectors and the grey scale used to represent the gas velocity
magnitude. Most of the gas jet flows directly towards the uptake shaft upon reaching the molten
bath surface, and recirculation flow is limited to the closed end of the free board.
Table V. General Parameters for the Computer Simulations of the Industrial Furnace
Parameter Value
Inlet/Outlet and Reactor Geometry:
Reaction Shaft Diameter, m 4.25
Reaction Shaft Length, m 6.5
Settler Freeboard Length, m 1.0
Settler Freeboard Width, m 4.45

Wall Temperature, K†:


Top and Side Walls 1570
Molten Bath Surface 1620
Surface Emissivity (Inlets, Outlet, Walls, and Molten Bath) 0.85

Primary Stream (Particles):


Temperature, K 298
Particle Axial Velocity, m s-1 2.3§
Particle Density, kg m-3 5400
Number of Injection Points for Cloud Model 16
Standard Deviation of Incoming Clouds (σ0), m 0.03
Stream Diameter, m 0.4

Secondary Stream (Process Gas):


Temperature, K 298
Pressure, kPa 86.1
Turbulent Intensity 0.1§
Outlet Gas Temperature, K 1570 ‡
Stream Diameter, m 0.6
Injection Velocity, m/s
Axial component 16.3
Radial component (0; 85º) 0; Varies from 0
at the center to
186 at the edge

† Approximate experimental values


‡ For radiation calculations only
§ Estimated

Table VI. Operating Conditions for the Computer Simulations of the Industrial Furnace
Parameter Value
Particle Size, µm 50 to 150
Oxygen Concentration in Process 70
Gas, vol.%
Solid Feed Rate, ton h-1 54
-1
Gas Feed Rate, ton h 16.8
Gas Injection Angle, degrees 0, 85
(a) (b)

Figure 20. Gas velocity fields in an industrial flash converting furnace: (a) single axial entry
burner, (b) 85º-distributor cone burner. (The numbers indicate velocity magnitude in m/s.)

In contrast, Figure 20(b) shows the velocity field predicted when the
85º-distributor-cone burner is used. In this system, the inlet gas velocity is about 125 m/s and it
rapidly falls upon entering the furnace. The radial component in the inlet gas velocity produces
large recirculation zones that extend through the entire reactor volume. It is of interest to note
that the velocity of the gas flowing towards the uptake shaft is much smaller than most of the
velocities predicted within the shaft volume. Thus, the 85º configuration strongly promotes
backmixing of the particle-gas suspension, which in turn results in uniform distribution of
particles and oxygen. Further, the gas velocity in the free board is lower in this case than with a
single axial injection due to the higher extent of reaction, which removes oxygen from the gas
phase to form metal oxides.

The flow patterns in the shaft substantially affect the overall features of the flash
converting operation, in particular the distribution of particles. Figure 21 shows the particle
number density predicted for both systems. When the single entry burner is used (Figure 21(a)),
the particles are concentrated along the centerline and do not disperse until they travel beyond
the upper half of the reaction shaft length. Thus, a large portion of the reactor volume is not used
for the particle-gas contact. Particles falling along the centerline follow a straight trajectory and
are surrounded by the main portion of the gas jet core. Because this gas flows at very high
velocity and the particles and oxygen are not evenly distributed, the reaction rates are poor.
(a) (b)

Figure 21. Particle number density in an industrial flash converting furnace: (a) single axial
entry burner, (b) 85º-distributor cone burner. (The numbers indicate the logarithm
with base 10 of the particle number density in number of particles per cubic meter.)

On the other hand, Figure 21(b) shows the predicted distribution of particles for the 85º-
cone burner. The large recirculation zones in this system promote a strong backmixing of the
particle-gas suspension, and as a result the particles get dispersed throughout the reactor volume.
Although this characteristic is favorable in terms of particle oxidation rate, a potential drawback
is the high possibility of particle deposition on the reactor walls. This is indicated in Figure
21(b) by large particle number density values in the vicinity of the reactor walls. In the present
formulation, particles hitting the reactor walls were assumed to get stuck and not to react any
further.

This type of 3-D model will enable one to predict the extent of dust carry-over through
the uptake, which is an important practical operating concern. The behavior discussed in Figures
20 and 21 is expected to be valid for both flash converting and flash smelting furnaces because
the conditions in both reactors are similar and the system geometry is essentially the same.
However, other gas and particle characteristics such as temperature and composition are
dependent upon the heterogeneous reaction kinetics, which are different in matte and
concentrate particles.

Figure 22 shows the gas temperature contours predicted for a flash converting furnace
shaft for the two burner configurations tested. When the single axial entry burner is used (Figure
22(a)), the gas jet flowing through the centerline exchanges little heat with the surroundings.
Along the centerline, gas temperature only starts to increase close to the bath surface where
mixing and heat generation by oxidation reactions occur. Mixing also occurs in the areas
between the gas jet core and the reactor walls (Figure 20(a)), and thus gas temperature increases
in this region. It is of interest to note a long cylindrical volume around the gas jet core in which
very steep temperature gradients occur, thus indicating uneven exchange of heat in this system.

(a) (b)

Figure 22. Gas temperature in an industrial flash converting furnace: (a) single axial entry
burner, (b) 85º-distributor cone burner. (The numbers indicate temperature in Kelvins.)

On the other hand, Figure 22(b) shows the gas temperature contours predicted for the
85º-distributor-cone system. In this case, gas temperature is more uniform than in the previous
case and temperature gradients in the reaction shaft are smooth over the entire reactor diameter.
It is noted that the gas is heated almost immediately upon entering the reaction chamber. This is
mostly due to the large recirculations occurring in this system, which mix the cold gas entering
the furnace with hot gases coming from the bottom (Figure 20(b)). No hot spots occur. This
indicates that particles are oxidized evenly in the reaction chamber.

The fact that a radial velocity component of the injected feeds increases the uniformity
of the particle and gas distribution in the shaft can be expected even without a complex 3-D fluid
dynamic model; what is important is that such a model yields quantitative effects of operating
conditions.

The above features have a major impact on the rate at which the oxidation reactions
occur, and thus the extent of oxygen consumption. Figure 23 shows the oxygen concentration
contours predicted for the single axial entry and the 85º-distributor-cone burner, respectively. As
expected from the results discussed previously, oxygen consumption in the 85º configuration is
substantially larger than that predicted for the single-entry system. Oxygen concentration in the
gas flowing towards the uptake shaft is predicted to be 42 mole pct for the single-entry system,
and 20 mole pct for the 85º system. The concentration contours also show no sharp reaction
front in the 85º system, whereas in the single-entry system the reaction front is located in the
lower portion of the reaction shaft in the vicinity of the centerline.

(a) (b)

Figure 23. Oxygen concentration in an industrial flash converting furnace: (a) single axial
entry burner, (b) 85º-distributor cone burner. (The numbers indicate
concentration in mole fraction.)

The extent of oxygen consumption predicted by the computer simulations is lower than
those observed in the commercial plant. Oxygen concentration in the off-gas stream entering the
waste-heat boiler is about 5-7 pct by mole in a commercial furnace [39]. There may be a number
of reasons for the discrepancy between the predicted and the observed values, including the fact
that in the present formulation, particles hitting the reactor walls were assumed to get stuck and
not to react any further. The neglected amount of oxygen consumption by these particles may be
substantial in the 85º configuration, because in this system particle number density is high near
the wall and the deposition rate is significant. Further improvements to the formulation should
take this into consideration.

Concluding Remarks

In this paper, basic principles were applied to analyze the behavior of two high-intensity
smelting processes. The examples included the production of molten mattes of copper and
nickel concentrates, which is mostly characterized by thermodynamic factors, and the flash
smelting and flash converting processes, for which rate processes are important. Appropriate
mathematical models were developed for each case and used to analyze the main features in
these systems. The results discussed in this paper are summarized below.

To thermodynamically understand the copper or nickel matte smelting processes with


oxygen-rich gas, phase equilibrium and distribution experiments were conducted for the
equilibration of an FeOx-SiO2 based slag or an FeOx-CaO slag with Cu2S-FeS or Ni3S2-FeS
mattes at 1250 or 1300oC under controlled partial pressures of S2 and O2 with pSO2 of 0.1, 0.5
and 1 atm. Conclusions from this study are as follows:

(1) The effect of pSO2 on the solubility of copper and nickel in the slag is quite different
between the Cu2S-FeS and Ni3S2-FeS mattes. At a specified matte grade, the copper
solubility is constant against pSO2, while the nickel solubility increases with pSO2.
(2) The distribution ratios of arsenic, antimony, bismuth and cobalt at a specified matte
grade increase with increasing pSO2, while those of copper and silver are almost
constant against pSO2.
(3) The fractional distribution of arsenic, antimony and bismuth among the gas, slag and
matte phases, which was evaluated by using the obtained thermodynamic data,
suggests that elimination of these minor elements from the matte phase by means of
volatilization becomes less favorable when the degree of oxygen enrichment in the
blowing gas is increased, while elimination by slagging becomes more favorable.
(4) These results suggest that the use of oxygen-rich gas for blowing copper and nickel
concentrates has no serious effect on the recovery of valuable copper and silver in
the matte phase, while it lowers the recovery of nickel and cobalt. It is also indicated
that the distribution behavior of arsenic, antimony, bismuth and cobalt is strongly
affected by the use of oxygen.

A three-dimensional fluid dynamics model of the flash smelting and flash converting
reactors was developed. The model incorporates the transport of momentum, heat, mass, and
reaction kinetics between the gas and particles in a particle-laden turbulent gas jet. The
computer model was used to simulate a flash converting operation with tonnage oxygen for
laboratory and industrial scales of operation. The results from the computational studies are
summarized as follows:

(5) The model represents with reasonable accuracy the major phenomena occurring in
the large laboratory furnace. Reasonable agreement between the model predictions
and the experimental data was obtained for the fractional conversion and sulfur
removal from the particles in the reaction shaft.
(6) The model can be used to obtain quantitative effects of burner configuration, which
strongly affects the behavior of industrial flash converting and flash smelting
furnaces including such factors of practical importance as dust carry-over. Overall, a
greatly improved distribution of particles, higher reaction rate, and higher oxygen
consumption is obtained with an 85º-distributor-cone burner than those obtained
with a single axial entry burner.
(7) The present formulation can be applied to the analysis of a flash smelting furnace by
incorporating the appropriate terms describing the reaction kinetics of copper
concentrate particles.

The authors wish to emphasize the key role that basic principles play in the design,
analysis, and optimization of modern high-intensity smelting processes. With continuing
advances in computing technology, it is expected that more sophisticated mathematical models
will be developed in the near future that will help us better understand the behavior of these
processes from a fundamental standpoint.

Acknowledgments

The authors express their gratitude to Reaction Engineering International, Salt Lake City,
Utah, for providing the grid generation software IGNITE used in the computer simulations of
the industrial flash furnaces.
Nomenclature

Ac m2 Outer area of the unreacted core


Ap m2 External surface area of a particle
CA Mol m-3 Oxygen concentration
C AI ,e , C AII,e Mol m-3 Equilibrium oxygen concentration at the Cu2S/Cu2O
and Cu2O/CuO interfaces, respectively
D A,eff m2 s-1 Oxygen effective diffusivity in the particle
km m s-1 Mass transfer coefficient in the bulk gas
L s/m
X
_ Distribution ratio of minor element X between slag and
matte phases
nt mol Total number of moles in 100 g of mixture
p A, s Pa Oxygen partial pressure at the particle surface
p AI ,e Pa Equilibrium oxygen partial pressure at the Cu2S/Cu2O
interface
p AII,e Pa Equilibrium oxygen partial pressure at the Cu2O/CuO
interface
pO2 , pS 2 atm Partial pressure of oxygen and sulfur, respectively
rc m Unreacted core radius
ri m Radius of the Cu2O/CuO interface
rp m Particle radius
R A,T mol s-1 Total rate of oxygen consumption in a particle
R A,1 , R A,2 mol s-1 Rate of oxygen consumption in a particle by reactions 4
and 6, respectively
Tcr K Characteristic temperature of reaction
x - Moles of sulfur per mole of iron in FeSx
y - Moles of FeSx per mole of Cu2S in matte particle
z - Moles of inerts per mole of Cu2S in matte particle
Greek Symbols
εI - Porosity of phase I
ε II - Porosity of phase II
υ - Element valence
Subscripts
b Bulk gas
e Equilibrium
g Gas
i i-th species
in Inerts
o Oxidic part
s Slag
X Minor element
Superscripts
c unreacted core
p Particle
I phase I (Cu2O/Fe3O4)
II phase II (CuO/ Fe3O4)
Special symbols
{} Mass percent in the matte phase
<> Mass percent in the slag phase
() Mass percent in the gas or slag phase

References

1. H.Y. Sohn and V. Ramachandran, “Advances in Sulfide Smelting – Technology, R&D, and
Education,” in Sulfide Smelting’98: Current and Future Practices, J.A. Asteljoki and R.L.
Stephens (Eds.) (TMS, Warrendale, Pennsylvania, 1998), 3-37.
2. G. Roghani, M. Hino, and K. Itagaki, "Phase Equilibrium between Calcium Ferrite Slag and
Copper Matte at 1523 K under High Partial Pressures of SO2," Mater. Trans. Japan Inst.
Metals, 37 (1996), 1431-1437.
3. G. Roghani, J.M. Font, M. Hino, and K. Itagaki, "Distribution of Minor Elements between
Calcium Ferrite Slag and Copper Matte at 1523 K under High Partial Pressures of SO2,"
Mater. Trans. Japan Inst. Metals, 37 (1996), 1574-1579.
4. G. Roghani, M. Hino, and K. Itagaki, "Phase Equilibrium and Minor Element Distribution
between SiO2-CaO-FeOx-MgO Slag and Copper Matte at 1573 K under High Partial
Pressures of SO2," Mater. Trans. Japan Inst. Metals, 38 (1997), 707-713.
5. J.M. Font, Y. Takeda, and K. Itagaki, "Phase Equilibrium between Iron-Silicate Base Slag
and Nickel-Iron Matte at 1573 K under High Partial Pressures of SO2," Mater. Trans. Japan
Inst. Metals, 39 (1998), 652-657.
6. J.M. Font, M. Hino, and K. Itagaki, "Minor Elements Distribution between Iron-Silicate
Slag and Ni3S2-FeS Matte under High Partial Pressures of SO2, " Mater. Trans. Japan Inst.
Metals, 39 (1998), 834-840.
7. J.M. Font, G. Roghani, M. Hino, and K. Itagaki, "Solubility of Copper or Nickel in
Iron-Silicate Base Slag Equilibrated with Cu2S-FeS or Ni3S2-FeS Matte under High Partial
Pressures of SO2," Metallurgical Review of Min. Mater. Proc. Inst. Japan, 15 (1) (1998),
75-86.
8. J.M. Font, M. Hino, and K. Itagaki, "Phase Equilibrium and Minor Elements Distribution
between Iron-Silicate Base Slag and Nickel-Copper-Iron Matte at 1573 K under High
Partial Pressures of SO2," Mater. Trans. Japan. Inst. Metals, 40 (1999), 20-26.
9. J.M. Font, M. Hino, and K. Itagaki, "Thermodynamic Evaluations of Distribution Behavior
of VA Elements in Nickel Matte Smelting," Metallurgical Review of Min. Mater. Proc. Inst.
Japan, 15 (2) (1998), 202-220.
10. G. Roghani, Y. Takeda, and K. Itagaki, "Phase Equilibrium and Minor Element Distribution
Between FeOx-SiO2-MgO Based Slag and Cu2S-FeS Matte at 1573 K under High Partial
Pressures of SO2," Metall. Mater. Trans. B, 31B (2000), 705-712.
11. J.M. Font, M. Hino, and K. Itagaki, "Phase Equilibrium and Minor Element Distribution
between Ni3S2-FeS Matte and Calcium Ferrite Slag under High Partial Pressures of SO2,"
Metall. Mater. Trans. B, 31B (2000), 1231-1239.
12. A. Yazawa, "Thermodynamic Consideration of Copper Smelting," Can. Met. Quart. 13
(1974), 443-453.
13. G.H. Kaiura, K. Watanabe, and A. Yazawa, "The Behavior of Lead in Silica-Saturated,
Copper Smelting Systems," Can. Met. Quart. 19 (1980), 191-200.
14. C. Acuna and A. Yazawa, "Behaviors of Arsenic, Antimony and Lead in Phase Equilibria
among Copper, Matte and Calcium or Barium Ferrite Slag," Trans. Japan Inst. Metals, 28
(1987), 498-501.
15. H.K. Jalkanen, E.K. Holappa, and J.K. Makinen, "Some Novel Aspects of Matte-Slag
Equilibria in Copper Smelting," in Advances in Sulfide Smelting, Vol. 1 Basic Principles, H.
Y. Sohn et al. (Eds.) (TMS, Warrendale, Pennsylvania, 1983), 277-292.
16. U. Kuxmann and H. Bussman, "Untersuchungen zu den Schmelzgleichgewichten zwischen
Kupfer, Stein und Eisenoxidschlacken in Kalk- und Quarztiegeln," Erzmetall, 27 (1974),
353-365.
17. R. S. Celmer, "Study on the Distribution of the Minor Elements in Nickel Matte Smelting,"
(Ph.D. thesis, University of Toronto, 1987), 1-341.
18. F.J. Tavera and W.G. Davenport, "Equilibrium of Copper Matte and Fayalite Slag under
Controlled Partial Pressures of SO2," Metall. Trans. B, 10B (1979), 237-241.
19. A. Yazawa, M. Oida, and Y. Nishikawa, "Distribution Equilibria for Ni, Co, As, Sb and Cu
between Matte and Slag," J. Min. Metall. Inst. Japan, 98 (1982), 963-968.
20. M. Nagamori, "The Behavior of Sulfur in Industrial Pyrometallugical Slags," J. Metals, 46
(8) (1994), 65-71.
21. Y. Takeda, "The Effect of Basicity on Oxidic Dissolution of Copper in Slag," in
Metallurgical Processes for Early Twenty-first Century, Vol. 1 Basic Principles, H. Y. Sohn
(Ed.) (TMS, Warrendale, Pennsylvania, 1994), 453-466.
22. A. Yazawa, "Distribution of Various Elements between Copper, Matte and Slag," Erzmetall,
33 (1980), 377-382.
23. K. Itagaki and A. Yazawa, "Thermodynamic Evaluation of Distribution Behavior of
Arsenic, Antimony and Bismuth in Copper Smelting," in Advances in Sulfide Smelting, Vol.
1 Basic Principles, H. Y. Sohn et al. (Eds.) (TMS, Warrendale, Pennsylvania, 1983),
119-142.
24. M. Nagamori and P. C. Chaubal, "Thermodynamics of Copper Matte Converting: Part III.
Steady-state Volatilization of Au, Ag, Pb, Zn, Ni, Se, Te, Bi, Sb and As from Slag, Matte
and Metallic Copper," Metall. Trans. B, 13B (1982), 319-329.
25. P. C. Chaubal and M. Nagamori, "Thermodynamics for Arsenic and Antimony in Copper
Matte Converting - Computer Simulation," Metall. Trans. B, 19B (1988), 547-556.
26. M. Nagamori and P. J. Mackey, "Thermodynamics of Copper Matte Converting: Part II.
Distribution of Au, Ag, Pb, Zn, Ni, Se, Te, Bi, Sb and As between Copper, Matte and Slag
in the Noranda Process," Metall. Trans. B, 9B (1978), 567-579.
27. H. G. Kim and H. Y. Sohn, "Minor-element Behaviour in Copper Smelting and Converting
with the Use of Tonnage Oxygen," Trans. Instn. Min. Metall. Sect. C, 107 (1998), C43-C59.
28. I.V. Kojo and P. Hanniala, in Metallurgical Processes for the Early Twenty-First Century,
Vol. 2. Technology and Practice, H.Y. Sohn (Ed.) (TMS, Warrendale, PA, 1994), 89-102.
29. J.A. Asteljoki, L.K. Bailey, D.B. George, and D.W. Rodolff, J. Metals, 20 (5) (1985), 20-23.
30. K.J. Richards, D.B. George, and L.K. Bailey, in Advances in Sulfide Smelting, Vol. 2.
Technology and Practice, H.Y. Sohn et al. (Eds.) (TMS, Warrendale, PA, 1983), 489-498.
31. D.B. George, R.J. Gottling, and C.J. Newman, Proc. Copper 95-Cobre 95 International
Conference, Vol. IV (CIM, Montreal, Canada, 1995), 41-42.
32. M. Perez-Tello, Ph.D. Dissertation, University of Utah, Salt Lake City, Utah, 1999.
33. M. Perez-Tello, H.Y. Sohn, and P.J. Smith, Accepted for publication in Metall. Mater. Trans.
B, (2001); in Fluid Flow Phenomena in Metals Processing, El-Kaddah et al. (eds.) (TMS,
Warrendale, Pennsylvania, 1999), 101-116.
34. L.L. Baxter and P.J. Smith, Energy & Fuels, 7 (1993), 852-859.
35. Y.B. Hahn and H.Y. Sohn, Metall. Trans. B, 21B (1990), 959-966.
36. M. Perez-Tello, H.Y. Sohn, and J. Löttiger, Min. Metall. Process., 16(2) (1999), 1-7.
37. W.G. Davenport and E.H. Partelpoeg: Flash Smelting: Analysis, Control, and Optimization;
(Pergamon: Oxford 1987).
38. M. Perez-Tello, H.Y. Sohn, K. St. Marie, and A. Jokilaakso, Accepted for publication in
Metall. Mater. Trans. B, (2001).
39. D.B. George, Kennecott Utah Co., personal communication, 1997.

View publication stats

Вам также может понравиться