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

net/publication/257761515

Optimizing Peirce–Smith Converters Using Thermodynamic Modeling and

Plant Sampling

Article in JOM: the journal of the Minerals, Metals & Materials Society · May 2012
DOI: 10.1007/s11837-012-0329-x

CITATIONS READS

22 1,646

5 authors, including:

Nubia Cardona Valencia Pascal Coursol

Deltamet Consultants Glencore
7 PUBLICATIONS 213 CITATIONS 24 PUBLICATIONS 265 CITATIONS

SEE PROFILE SEE PROFILE

Roberto Parra
University of Concepción
82 PUBLICATIONS 778 CITATIONS

SEE PROFILE

All content following this page was uploaded by Nubia Cardona Valencia on 22 July 2014.

The user has requested enhancement of the downloaded file.

JOM, Vol. 64, No. 5, 2012
DOI: 10.1007/s11837-012-0329-x
Ó 2012 TMS

Feature Case Study

Optimizing Peirce–Smith Converters

Using Thermodynamic Modeling and
Plant Sampling
N. Cardona, P.J. Mackey, P. Coursol, R. Parada, and R. Parra

The performance of pyrometal- the quality of converter slag and as- ers (PSC) for converting to blister
lurgical slag cleaning furnaces at sess the overall condition of the con- copper. Flash furnace slag is tapped
many primary copper smelters is de- verter operation. and transferred to one of the two
pendent in part on the quality of the fuel-air injected rotary slag cleaning
INTRODUCTION
converter slag, commonly produced furnaces (SCF) for treatment prior
in the batch-wise Peirce Smith con- The Chagres smelter, part of An- to discard. Slag blow converter slag
verter (PSC). In order to understand glo American PLC, is located 80 is also treated in the SCF. Cleaned
the impact of converter slag chemistry km north of Santiago, Chile and has slag is skimmed and discarded, while
and at the same time help optimize the a nominal capacity of 600 kt of dry the produced matte (~72% Cu) is ac-
converter operation, thermodynamic concentrates per year. The copper cumulated over several cycles be-
modeling of molten slag (including production process commences with fore skimming and transferred to the
any contained slag solid fractions) mixing and drying concentrate feed PSC. A simpli¿ed Àow diagram of the
was carried out on slag produced at with silica Àux and internally generat- smelter is provided in Figure 1.
the Chagres smelter in Chile. Phase ed revert material in two rotary kilns. The PSC cycle at the Chagres
characterization studies on actual The dryer product is smelted at an smelter typically consists of two slag
plant slag samples were also carried average rate of 80 tph with oxygen- blows, followed by the copper blow.
out. The results are provided in the enriched air in the Outokumpu Àash After matte charging, oxygen-en-
present paper. This work is also con- smelting furnace (FSF) to produce riched air (23% to 26% O2) is injected
sidered as a case study example to il- matte (~61%Cu) and slag. The matte through submerged tuyeres. In each
lustrate the type of work that can be is periodically tapped and transferred of the ¿rst two blows, silica Àux and
performed to fairly quickly diagnose by ladle to the Peirce-Smith convert- plant reverts are added to the charge;
at the end of each blow the slag is
skimmed while the matte, approach-
ing white metal grade, remains in the
vessel. The copper blow also includes
addition of some silica Àux and plant
reverts; the produced molten oxide
slag is returned to a converter (nor-
mally in the second slag blow).
Unduly high magnetite and cop-
per levels in PSC slag can be prob-
lematic to the SCF process, leading to
higher copper levels than target (<1
wt.%), thus affecting the overall cop-
per recovery at the smelter. In order
to better understand the impact of the
converter slag quality on the SCF pro-
cess, and help in optimizing the con-
verter operation, the physical chem-
istry and the composition of phases
present in PSC slag were studied us-
Figure 1. Chagres ing thermodynamic modeling, sup-
Smelter flow dia-
gram.1 ported by measurements on selected
plant samples.

546
Optimizing Peirce–Smith Converters Using Thermodynamic Modeling and Plant Sampling 547

The results of this study are report- ied between 0.9 and 3.0 to cover the withdrawing. This technique provided
ed in the present paper. In addition, the complete range of possible slag com- conditions for obtaining a rapidly so-
authors consider this work as a type of positions and converter operations. lidi¿ed slag sample. The bulk chemi-
case study illustrating a new approach The liquidus lines were delimited con- cal composition of the samples was
towards achieving better converter sidering the likely solid phases to be determined by atomic absorption. In
slag and converter optimization at formed in each particular system: spi- addition, three samples were mounted,
other plants and achieving improved nel ([Fe2+, Zn2+, Mg2+]{Fe3+, Al3+}2O4), polished and examined by light opti-
overall plant copper recoveries. The olivine ([Mg2+, Fe2+, Zn2+, Ca2+]2SiO4), cal microscopy (LOM) and electron
authors welcome comments and sug- pyroxene([Fe2+, Mg2+, Ca2+]SiO3) and microprobe (EPMA). The microanaly-
gestions in this regard. tridymite (SiO2). The thermodynamic ses of phases were performed using a
properties of each component were JXA JEOL-8900L, with 3 Pm electron
taken from the FactSageTM database; microprobe (WDS system), accelera-
METHODOLOGY AND
the molten phases (copper, matte, and tion voltage 20 kV, beam current 20
FUNDAMENTAL
slag) and the solubility of copper in nA and counting time 20 s for each el-
CONSIDERATIONS
liquid slag were described using the ement. The procedure for the analysis
The thermodynamic modeling of modi¿ed quasichemical model,4,5 with of copper-bearing phases in the solidi-
converter slag was performed using the the sul¿de solubility according to the ¿ed industrial slag samples has been
FactSageTM 6.2 software.2 This pack- Kang and Pelton quadruplet approxi- described by Cardona et al.8
age is well suited to perform calcula- mation.6 A detailed methodology has
RESULTS AND DISCUSSION
tions involving slag and matte compo- been described in Cardona et al.7
sitions relevant to the copper smelting The chemical activities of Cu2S and
Impact of Fe/SiO2 Ratio on the
processes.3 This approach assumes FeS in slag were ¿xed according to the
Liquid Slag from First and
that thermodynamic equilibrium is es- average matte composition considered
Second Slag Blows
tablished between all phases present for each slag blow (Table I). In each
in the system due to high temperature case, the p(SO2) was ¿xed at 0.16 atm, At Chagres, silica Àux is added
and effective mixing conditions in the based on the average level of oxygen to the converter at the beginning of
vessel. The following parameters were enrichment in tuyere air (25%O2). In slag blows to produce an iron silicate
adopted for examination in the pres- the copper blow, the chemical activity fayalite-type slag. When the Àux is ef-
ent work, in part due to applicability of liquid copper and the oxygen poten- fectively absorbed into the melt, the
for the present objective and also their tial (pO2) were ¿xed. tendency to form excess magnetite
importance in process optimization: A converter sampling campaign was can normally be controlled according
the liquidus temperature, the copper also carried out. Slag samples were to reaction (1):
solubility and the proportion of solid collected by inserting a 1 cm thick
Fe3O4+2SiO2 (slag)
phases in slag from the ¿rst slag blow, Àat steel plate, a15 cm by 15 cm, into
= Fe2SiO4 (slag) + 3/2 SO2 (1)
the second slag blow and the copper the molten stream and then quickly
blow.
In mapping the PSC slag chemistry,
Table I. Base Values for Slag, Matte and Copper Compositions1
the number of degrees of freedom in
the present system for each blow was Phase wt.% 1st Slag Blow 2nd Slag Blow Cu blow
¿xed according to the *ibbs phase Slag Total Cu 2.1–7.7 2.1–7.7 35–40
rule.3 Average analyses based on
smelter data and also data from the Total S 0.1–0.5 0.1–0.5 0.04
special sampling campaign under- Fe/SiO2 1.4–2.5 1.4 v 2.3 1.3–1.6
taken as part of the present study were Al2O3 0.8 0.9 1.1
used to de¿ne the base values (Table
ZnO 0.6 0.5 0.0
I). Minor elements such as Pb, As, Sb,
Bi, Cr, and Ni were neglected in this CaO 0.4 0.3 0.6
work as their level was considered too MgO 0.2 0.4 0.2
low (less than 0.%) to signi¿cantly
Na2O +K2O 0.5 0.6 0.6
affect the slag chemistry under the Matte/ White Metal
converter conditions. (average composition) Cu 73.0 79.0 —
The following parameters were Fe 5.5 0.7 —
¿xed for each computation according S 21.5 20.3 —
to the reference values given in Table I: Blister Copper
[%Al2O3]slag, [%MgO]slag, [%CaO]slag, Cu — — 99+
[%Na2O]slag, the total pressure (1 atm) O — — 0.2–0.7
and temperature (1,180–1,250qC). S — — 0.05–0.01
The ratio [%Fe/%SiO2]slag was var-
548 Cardona, Mackey, Coursol, Parada, and Parra

0–10 percent by weight) in the slag.

The slag from the second blow can
Figure 2. Impact of the contain solids from a %Fe/%SiO2 ratio
Fe/SiO2 ratio on the cal- of 1.5 and upwards at a temperature of
culated PSC slag liqui- 1,200qC (see Figure 4). The conditions
dus from 1st and 2nd blows
at Chagres Smelter would be optimal between % Fe/SiO2
(p(SO2)=0.16 atm., First ratios of 1.6 to 1.8 at a temperature be-
blow: >Fe@matte=5.5wt.% tween 1,200q to 1,220qC. A validation
and >Cu@matte=71.0 wt.%,
Second blow: >Fe@ at the plant would be required to con-
matte
=0.7wt.% and >Cu@ ¿rm the optimal proportion of solid to
matte
=79.0wt.%. L+ Trid: limit entrainment while maintaining
Liquid with tridymite sol-
ids. Present operational the refractory protection.
conditions are normally Solid magnetite crystals formed in
within the dotted square. the molten slag from the ¿rst blow
were identi¿ed with a distinctive mor-
phology on solidi¿ed slag samples
as shown in Figure 5. It is noted that
other magnetite crystals (dendrites)
The addition of more Àux than re- solid spinel formed in the converter present in the industrial samples (re-
quired is undesirable as the slag can slag (essentially as magnetite) at the ferred as Magnetite II) are formed dur-
tend to become more viscous and end of the ¿rst and second slag blows. ing slag solidi¿cation as a result of the
additional smelting heat is required. As shown in Figure 3, a temper- oxidation of soluble ferric iron in the
Thus, maintaining the correct Fe/SiO2 ature between 1,180 and 1,200qC, at a liquid (as Fe2O3), as discussed by Car-
ratio in converter throughout the cycle %Fe/%SiO2 ratio of 1.8 to 2.0 appears dona et al.8 Both forms of magnetite
is therefore quite important for the optimal to obtain a reasonable level of are accounted by the magnetite mea-
production of a good quality slag. Fig- solid magnetite (for example, between surement indicated by the Satmagan®
ure 2 presents the computed liquidus
temperature of slag produced after the
¿rst and second slag blows at Chagres
smelter.
Based on Figure 2, the liquidus Figure 3. Calcu-
increases steeply with the increas- lated solid fraction
ing %Fe/%SiO2 ratio (towards the in slag from 1st blow
(p(SO2)=0.16 atm.,
high SiO2 range), until a minimum at >Fe@ matte =5.5wt.%,
%Fe/%SiO2 ratio of 1.4 for the ¿rst >Cu@ matte=71.0wt.%,
slag blow and 1.35 for the second slag (>Al2O3@slag=0.8wt.%,
>CaO@ slag=0.4wt.%,
blow. A transition occurs between the >N 2 O+K 2 O@ slag =0.5
formation of tridymite and spinel as wt.%, >MgO@slag=0.2
the primary solid phase in the liquid wt.%, >ZnO@slag=0.6
wt.%).
slag. At %Fe/%SiO2 ratios higher than
these values, the liquidus temperatures
increases by 11.5°C per 0.1 increment
of the %Fe/%SiO2 ratio (for ratios be-
tween 1.4 and 2.1).
Solids fraction in the slag. The for-
mation of a solid phase in converter
slag is well known.9 When properly
Figure 4. Calculated
controlled, this condition is also use- solid fraction in slag
ful since the solids present can help from the 2nd blow
form a protective layer on the refrac- (p(SO2)=0.16 atm.,
>Fe@matte, = 0.7wt.%,
tory lining of the vessel. On the other >Cu@matte, =79.0wt.%.
hand, an excessive amount of solids (>Al 2 O 3 @ slag =0.9wt.%,
increases the slag viscosity, thus pre- > C a O @ slag= 0 . 3 w t . % ,
>N2O+K2O@slag=0.6wt.%,
venting good matte–slag separation > M g O @ slag= 0 . 4 w t . % ,
and increasing the amount of entrained >ZnO@slag=0.5 wt.%).
matte in the slag.
Figures 3 and 4 show the effect of
the Fe/SiO2 ratio on the percentage of
Optimizing Peirce–Smith Converters Using Thermodynamic Modeling and Plant Sampling 549

magnetic analyzer, based on sample

magnetic susceptibility. This instru-
ment is commonly used for magnetite
measurement and process control at
copper smelters.
Calculated and Observed
Distribution of Copper Levels
in PSC Slag Figure 5. BSE Image
of slag sample from
In order to better understand the sol- the 1st blow contain-
uble and entrained losses copper loss- ing excessive solid
magnetite phases: 1:
es in the Chagres PSC slag, thermo- Matte, 2: Copper, 3:
dynamic calculations were performed Magnetite I, 4: Mag-
and selected plant samples were ana- netite II.
lyzed. The results are presented in
100 μm
Tables II and III.
It is well known that the copper sol-
ubility in slag increases exponentially Figure 6. Impact of
the Fe/SiO2 ratio
as the %Fe in the matte approaches on the calculated
zero.4,9 This type of increase is il- PSC slag líquidus.
lustrated in Table II as judged by the ( > O @ Blister= 0 . 5 w t . %
>Al 2O 3@ slag=1.1wt.%,
copper content in slag increasing from >CaO@ slag =0.6wt.%,
about 0.5% to about 1.9% Cu in slag >MgO@ slag =0.2wt.%,
from the ¿rst slag blow to the second > N 2O + K 2O @
=0.6wt.%, ). L+
slag blow. From the perspective of slag
Trid: Liquid with Trid-
copper losses, it is considered desir- ymite solids. Present
able not to blow the matte to less than operational condi-
tions are normally
1 wt.% Fe at the end of second slag within the dotted
blow. This allows maintenance of an square.
acceptable level of soluble copper in
the slag, thus helping the slag cleaning
step in the SCF.
It can be seen that the soluble cop-
per estimates presented in Table II are The amounts of solid magnetite slag is treated in two fuel-air injected
in good agreement with the measured (magnetite I) in the samples shown in rotary vessels. The presence of solid
results presented in Table III. It is con- Table III were measured by light op- magnetite in the slag hinders cleaning
sidered that this result in part reÀects tical microscopy (LOM) and image and requjires a higher operating tem-
the effective bath mixing conditions analysis techniques. They appear rea- perature (1,240qC+)..
in the converter and the resulting ap- sonable considering the apparent Àu- Matte entrained in slag is generally
proach to near thermodynamic equi- idity of the slag observed at the plant. easy to recover through a pyrometal-
librium conditions. As mentioned above, the converter lurgical process (involves mainly time/
settling at the correct temperature),
Table II. Calculated Soluble Copper Levels in PSC Slag as Calculated Using FactsageT™M whereas lowering the level of soluble
Temperature copper ¿rstly requires chemical reduc-
tion. Settling of the resulting ¿ne par-
PSC Slag %Fe in Matte ( ºC) Fe/SiO2 Soluble Copper
ticles can be problematic; hence it is
First Slag Blow 5.5 1,180–1,250 1.9–2.3 0.5–0.9 preferable to minimize the amount of
Second Slag Blow 0.7 1,180–1,250 1.6–2.2 1.6– 1.9 soluble copper in the ¿rst place. Table

Table III. Percentage of Solids and Copper Solubility in PSC Slag (Measured on Plate Samples)
Chem. Analysis Image Analysis (avg.) EPMA (avg.) Distribution of Copper
Sample* (wt.%) (wt.%) (wt.%) (Relative %)
PSC Slag ID Total Cu Fe/SiO2 Solid magnetite Soluble Copper Soluble Entrained
1st Blow EC-6 3.7 2.3 8±1 0.8 ± 0.09 25 75
2nd Blow EC-3 4.3 1.6 4.9 ± 0.5 1.59 ± 0.3 35 65
2nd Blow EC-8 3.8 2.3 7.7 ± 1 2.0 ± 0.46 49 51
* Sampling campaign at Chagres, 18 June to 26 August, 2009
 550 Cardona, Mackey, Coursol, Parada, and Parra

III shows that a signi¿cant proportion fective Àux for the subsequent ¿rst slag such that with the new diagrams and
of the contained copper in the slag blow in the vessel. Such a practice has information produced in this study,
from slag blows is in the form of sol- been described at the former Noranda the plant will evolve further to have a
uble copper. This is especially true for Horne smelter in the late 1970s.11 fully optimized converter operation.
the second blow slag, which can have It is noted any slag remaining in the The authors consider this work as
up to 2wt.% copper as soluble copper vessel before the start of the copper a type of case study illustrating a new
in the slag. blow will increase in amount due to approach towards achieving better
the higher copper oxidation (increase converter slag and converter optimi-
Impact of Fe/SiO2 Ratio on the
in slag mass amount can range up to zation at other plants. The authors
Liquidus of the Slag from the
2 times). In the present FactSageTM welcome comments and suggestions
Copper Blow
calculations, such a condition was not in this regard.
Figure 6 shows the liquidus for considered. However, such a condition
ACKNOWLEDGEMENTS
slag produced from the copper blow can be examined on a well-developed
as a function of the Fe/SiO2 ratio. The mass balance around the cycle con- This work is based on research at
practice at Chagres typically ¿nishes verter operation. the Metallurgical Department of Con-
the blister copper with about 100 ppm cepcion University (DIMET), Chile.
CONCLUSIONS
to 500 ppm in blister and about 2,000 7KH ¿UVW DXWKRU ZLVKHV WR WKDQN WKH
ppm to 7,000 ppm oxygen. It is noted The results from the thermodynam- ‘Comision Nacional de Investigacion
that these data approximately corre- ic modeling indicate that some po- &LHQWÕ¿FD\7HFQRORJLFD &21,&<7 
spond to the earlier data of *erlach et tential improvements at Chagres are RI&KLOHIRULWV¿QDQFLDOVXSSRUWGXU-
al.10 For the present equilibrium cal- still possible, even though the plant is ing her doctoral studies. The authors
culations, an average of 5,000 ppm seen to operate with very good opera- also thank the personnel of Chagres
oxygen in copper was assumed, and tional practices. The %Fe/%SiO2 ra- Smelter; their collaborative attitude
this oxygen level was used to ¿x the tio in the ¿rst blow slag is considered has promoted improved synergies be-
oxygen potential in the bath at the end near-optimal (~1.9). However, the ra- tween fundamentals and applied as-
of the copper blow. It is noted that an tio in the second blow slag could have pects of this project.
oxygen level of 2,000 ppm was adopt- been maintained at a slightly lower
References
ed in the earlier study.1 value than that in the ¿rst blow, so
Under the present operating condi- as to control both the magnetite level 1. N. Cardona, P. Coursol, R. Parada and R. Parra,
TMS 2012, Supplemental Proceedings: Volume 2:
tions in the vessel for the blister ¿n- as the level of entrained matte in slag Materials Properties, Characterization, and Model-
ish (1,220qC and 5,000 ppm oxygen (%Fe/%SiO2 ~1.7). It would also be ing (Warrendale, PA: TMS, 2012), pp.117–124.
in copper), the slag is fully liquid over advisable to continue efforts towards 2. C.W. Bale, E. Bélisle, P. Chartrand, S.A. Dec-
terov, G. Eriksson, K. Hack, I.-H. Jung, Y.-B. Kang, J.
a range of the Fe/SiO2 ratio. The pro- improving matte grade control at the Melançon, A.D. Pelton, C. Robelin, and S. Petersen,
duced slag is high in copper content end of the second slag blow in order CALPHAD, 33 (2009), pp. 295–311.
(30–40% from Table I)—mainly as to minimize the amount of soluble 3. P. Coursol, N. Tripathi, P. Mackey, and T. Leggett,
Can Met. Quart., 49 (3) (2010), pp. 255–262.
soluble copper (as Cu2O). In this range copper in the slag, thereby reducing 4. S. Decterov and A. Pelton, Metall. Trans. B, 30B
of Cu2O content, this component es- the negative impact on the slag clean- (1999), pp. 661–669.
sentially acts as an effective Àux in ing operation. This will help lower 5. A. Pelton and P. Chartrand, Metall. Trans. B, 33A
(2000), pp. 1355–1360.
the converter, thus avoiding magnetite overall smelter copper losses. 6. Y.-B. Kang and A.D. Pelton, Metall. Trans. B, 40B
precipitation. At some plants, this slag Complete control of the chemistry (2000), pp. 979–994.
is referred to as “copper ferrite” slag, of the Peirce–Smith converter is dif- 7. N. Cardona, P. Coursol, Phillip Mackey, and R.
Parra, Can. Met. Quart., 50 (4) (2011), pp. 319–329.
as some CuFeO2 can be formed during ¿cult due to the many parameters in- 8. N. Cardona, P. Coursol, J. Vargas, and R. Parra,
cooling depending on slag composi- volved, some typically not well mea- Can. Met. Quart., 50 (4) (2011), pp. 330–340.
tion. sured and/or not well known: oxygen 9. P.J. Mackey, Can. Met. Quart, 2 (3) (1982), pp.
221–260.
The amount and quality of converter enrichment, input and output matte 10. J. Gerlach, K.P. Kantzer, and F. Pawlek, Metall-
slag, especially in the copper blow, grade, time of blowing and end point wissenschaft und Technik, 17 (11) (1963), pp.
is always a subject of debate at cop- control, Àux and reverts addition, and 1096–1099.
11. J.A. Vogt, P.J. Mackey, and G.C. Balfour, in Cop-
per smelters. The manner of slag re- the recirculation of rich Cu2O slag per and Nickel Converters, ed. R. Johnson (Warren-
moval is also an important subject. At from the copper blow. These parame- dale, PA: The Metallurgical Society of AIME, 1979),
a number of plants including Chagres, ters all inÀuence the mass balance and pp. 357–390.
to avoid slag skimmed along with the the performance of the converting and
blister copper, a small quantity of Àux slag cleaning vessels. With the results N. Cardona is with Kingston Process Metallurgy,
Kingston, Ontario, Canada; P.J. Mackey is with
is added at the converter lip to hold from the present study, Chagres has P.J. Mackey Technology Inc.; P. Coursol is with
back or “bank” the last traces of copper additional information to be included Barrick Gold Corporation; R. Parada is with Cha-
blow slag while skimming copper be- as part of the plant’s continuous im- gres Smelter, Anglo American PLC; R. Parra is
with the University of Concepcion, Chile. Dr. N.
neath the small Àux mass. This Àux by provement projects. The smelter has Cardona can be reached at nubia.cardona@
that time pre-heated, becomes the ef- well-developed mass balance models kpm.ca.

View publication stats