Slag–metal reactions during ladle treatment
with focus on desulphurisation
M. Andersson, M. Hallberg, L. Jonsson, and P. Jönsson
Within several cooperative projects, KTH (Royal
Institute of Technology), Ovako Steel AB, and
MEFOS have investigated the desulphurisation of
bearing steel during vacuum degassing. The work
includes thermodynamic calculations of the slag–
metal equilibrium, CFD modelling of slag–metal
reactions, and plant trials. Results from the various
studies are presented and discussed in this paper.
Models for predicting slag properties (sulphide
capacity, viscosity, and oxide activities) in liquid
slags as functions of slag composition and
temperature have been used for the calculation of
data which have been employed in static and
dynamic modelling of sulphur refining. The results
from static modelling show that the method allows
fast and easy evaluation of the theoretical
desulphurisation conditions during degassing at
Ovako Steel AB, as well as theoretical
determination of the parameters that have the
greatest influence on the equilibrium sulphur
distribution. The conclusion from dynamic
modelling is that the vacuum degassing operation
can be described dynamically with the present
knowledge of sulphide capacity, sulphur
distribution, viscosity, and oxide activities of ladle
slags if this knowledge is combined with fluid flow
modelling to derive the overall kinetics. The
presented model approaches have been found
useful in understanding the sulphur refining
process at Ovako Steel AB. The dynamic modelling
concept is also believed to have potential for
dynamic descriptions of other slag–metal reactions
in steelmaking. I&S/1684
Dr Andersson (maggan@metallurgi.kt h.se), Dr Jonsson, and Dr
Jönsson are in the Division of Metallurgy, Royal Institute of
Technology, SE–100 44 Stockholm, Sweden. Dr Jonsson is also
with MEFOS, Box 812, SE–971 25 LuleaÊ, Sweden. Dr Hallberg
is with Ovako Steel AB, SE–813 82 Hofors, Sweden. Based on a
presentation at the 6th International Conference on ‘Molten slags,
fluxes, and salts’ held in Stockholm and Helsinki on 12–16 June
2000.
© 2002 IoM Communications Ltd.
INTRODUCTION
At Ovako Steel AB, for many years attention has been
focused on producing clean steel with respect to non-
metallic inclusions and tramp elements. This has resulted
in signiŽ cantly reduced average total oxygen content in bear-
ing steel from 11 ppm in 1985 to 5 ppm in 1999.1 However,
to be able to survive in the future, new applications for
clean special steels must be found, which means that the
improvement of steel quality has to continue. Sulphides in
steel may also act as crack initiators (similar to calcium
aluminate oxides) when there is a high load in diVerent
directions. Consequently, sulphur control during ladle treat-
ment increases in importance. Increasing demands on pro-
224 Ironmaking and Steelmaking 2002 Vol. 29 No. 3
ductivity in the steel plant mean that further improvement
of the processes that determine the quality of the steel is
even more vital.
Today, 100% of the steel being produced at Ovako is
vacuum degassed. Vacuum degassing is performed in order
to reduce the sulphur and hydrogen contents as well as
the amount of non-metallic inclusions in the molten steel.
By controlling the ladle treatment, especially the vacuum
degassing operation, the quality of the steel is controlled.
Within several cooperative projects,2 – 4 KTH (Royal Institute
of Technology), Ovako Steel, and MEFOS have therefore
investigated the desulphurisation of bearing steel during
vacuum degassing. Since the top slag acts as a recipient
for sulphur, optimisation of the degassing operation will
result in an optimisation of the top slag with respect to
thermodynamics, kinetics, and  uid dynamics.
The present paper focuses on vacuum degassing and
covers basic thermodynamic and thermophysical research
of the slag–metal equilibrium, CFD modelling of slag–metal
reactions, and plant trials. Top slag properties, such as
sulphide capacity, viscosity, and oxide activity, are described
theoretically along with measurements and modelling results.
The theory behind calculation of the equilibrium sulphur
distribution between slag and metal phases is presented.
Plant trials performed at Ovako Steel are described including
a process description, the experimental procedure for carry-
ing out the plant trials, and results from the trials. Results
from static and dynamic modelling of the sulphur distri-
bution between the slag and metal are presented. Finally,
the results are summarised and some conclusions are drawn.
SLAG PROPERTIES
The ability to describe critical thermodynamic and thermo-
physical properties of the participating phases as functions
of slag composition and temperature is of utmost importance
when developing models for simulation and control of
diVerent metallurgical processes. In the present paper, the
authors have chosen to highlight some of the slag properties
which have been modelled and measured in the Division
of Metallurgy, KTH.
Sulphide capacity
When calculating slag–metal reactions with respect to
desulphurisation, the sulphide capacity is most important.
The sulphide capacity CS for a liquid slag was deŽ ned by
Fincham and Richardson5 as
A B
K 1 aO 2 pO 2 1 /2
CS = –= (% S )s l a g . . . . . . . (1)
fS 2 pS 2
–
where a O 2 is the activity of oxygen in the slag phase, p S 2
and p O 2 are– the partial pressures of S2 (g) and O2 (g), fS 2
is the activity coeYcient of sulphur in the slag phase,–
(%S)s l a g is the sulphur content in the slag*, and K 1 is the
equilibrium constant for the reaction
0´5S2 (g)+ (O2 – )s la g =0·5O2 (g)+(S2 – )s la g . . . . (2)
* All compositions are given in weight per cent throughout unless
speciŽ ed otherwise.
DOI 10.1179/030192302225004106

1 Calculated sulphide capacity of Al2O3–CaO–8MgO–
7SiO2 slag (wt-%): solidification of slag was not
considered
The sulphide capacity is dependent only on the temperature
and the slag composition. It describes the potential ability
of an arbitrary homogeneous molten slag to remove sulphur
and it could be used to compare the desulphurisation
characteristics of diVerent slags.
A model for calculation of the sulphide capacity for
multicomponent slags at diVerent temperatures has been
developed at the Division of Metallurgy, KTH.6 The model
enables the prediction of the sulphide capacities of multi-
component slags from the data of lower order systems.
The expression of sulphide capacity from equation (1) is
represented by the following relationships
A B
DG °
K 1 =exp ­ . . . . . . . . . . . (3)
RT
and
A B
aO 2 æ (X i ji ) +jm i x
–= exp ­ . . . . . . . (4)
fS 2 RT
– to be a function of both temperature and composition of
where DG ° is the Gibbs free energy of reaction (2), R is
the gas constant, T is the thermodynamic temperature. In
equation (4), subscript i represents component i and X i is
the molar fraction of component i in the multicomponent
system. The term ji is expressed as a linear function of
the temperature for each component in the slag in the
absence of interaction between diVerent species. The term
jm ix represents the mutual interaction ( binary and ternary)
between diVerent species in the slag and is dependent on
slag composition and temperature. In order to express the
mixing, the liquid slag is described by a modiŽ ed Temkin7
approach, which considers the mixing of cations and anions
within each of their subgroupings. More details regarding
the equations and the assessed parameters necessary for
the calculation have been presented elsewhere.6
Pure liquid FeO is chosen as the standard in the model,
for which the ratio aO 2 / fS 2 is taken as unity. The standard
Gibbs free energy DG ° (–J mol– – 1 ) is calculated from sulphide
capacity measurements of pure liquid FeO as
DG °= 118535 ­ 58´815T . . . . . . . . . . (5)
In Fig. 1 the calculated sulphide capacity is shown as
function of Al2 O3 content and temperature for an Al2 O3 –
CaO–8MgO–7SiO2 slag. Saturation of CaO or MgO in
the slag was considered. The predictions were carried out
using the KTH model.6
Viscosity
Viscosity is a thermophysical property that aVects the
kinetic conditions during ladle treatment. It is essential that
Andersson et al. Slag–metal reactions during ladle treatment 225
2 Measured and calculated (KTH model) dynamic viscosity
of slag of composition (wt-%) 40Al2O3–45CaO–6MgO–
9SiO2
the viscosity of both liquid slag and metal are known for
dynamic modelling of the process. While viscosity values
of liquid iron at diVerent temperatures have been reasonably
well established,8 accurate data of viscosities of ironmaking
and steelmaking slags are still lacking. At KTH, viscosity
measurements of diVerent metallurgical slag systems have
been incorporated into a slag model,9 which enables extra-
polation of slag viscosities as functions of temperature as
well as composition in the case of multicomponent systems.
A set of model parameters is optimised based on the
experimental data up to ternary systems.
In the KTH model, the dynamic viscosity g is expressed
by an Arrhenius relation
A B
hN r DG*
g= exp . . . . . . . . . . . (6)
M RT
where h is Planck’s constant, N is Avagadro’s number, and
r and M are the density and molecular weight of the liquid
slag, respectively. The term DG* in equation (6) is the
Gibbs energy of activation for viscosity, which is considered
the liquid slag. In the case of multicomponent systems, the
Gibbs activation energy can be represented by
DG* = æ X i DG *i +DG *m ix . . . . . . . . . (7)
where DG*i is the Gibbs energy of activation of the pure com-
ponent i in the liquid state. The second term in equation (7)
is due to the mutual interactions between diVerent species
and is expected to be a function of composition and tem-
perature. In the case of ionic solutions (slags), the Temkin
description is adopted. More detailed information concern-
ing the viscosity model can be found in published literature.9
In Fig. 2, the measured viscosity as a function of tem-
perature for a 40Al2 O3 –45CaO–6MgO–9SiO2 slag is shown
together with the calculated viscosity. The viscosity was
measured using the rotating cylinder method.9 The model
calculation is an extrapolation from lower order systems.
The results show that the model calculation gives a lower
viscosity (about 10–15%) than the measured value.
Oxide activities
Thermodynamic properties of metallurgical multicompon-
ent slags are of great importance when studying slag–metal
reactions and developing models for process control. The
activities of oxides in the molten slag and alloying elements
in the molten metal together with the process temperature
determine the equilibrium oxygen potential in the system.
Many measurements of oxide activities in diVerent slag
systems have been made, but it is very diYcult to determine
properties for all compositions that are encountered in
practice. The thermodynamic modelling of silicate melts
Ironmaking and Steelmaking 2002 Vol. 29 No. 3
226 Andersson et al. Slag–metal reactions during ladle treatment

The equilibrium constant in equation (8) K 8 is expressed

as1 7
935
log K 8 = ­ + 1´375 . . . . . . . . . . (9)
T

The equilibrium constant K 8 can also be written as

A B
aO pS 2 1 /2 (% S )s l a g a O
K8= = . . . . . (10)
aS pO 2 [% S ]m e t a l fS C S

where a O and a S are the activities of oxygen and sulphur

in the metal phase, fS is the activity coeYcient for sulphur in
the metal phase, and (%S)s l a g and [%S]m e t a l are the sulphur
contents in the slag and metal phases.
By combining equations (1), (9), and (10), the following
expression for the equilibrium sulphur distribution LS
between the slag and metal phases is obtained1 7
3 Comparison of Al2O3 activities for typical ladle furnace
slag15 calculated using the KTH model6 and Ohta and (%S)s l a g 935
Suito equation16 log LS = log =­ + 1´375
[%S]m e t a l T

+log C S +log fS ­ log a O . . . . . . (11)

has thus become one way of providing metallurgists with
necessary activity data for multicomponent slags within The activity coeYcient of a dissolved element in the steel
wide composition ranges. bath can be calculated by using Wagner’s equation provided
Several empirical or semi-empirical models for the pre- that the molten metal can be treated as a dilute solution.
diction of thermodynamic properties of slags have been Wagner’s equation is written as
developed over the years. Among them are the IRSID
model1 0 ( based on the work of Kapoor and Frohberg1 1 ) and log fj = æ (eij [%i ] ) . . . . . . . . . . . (12)
the two-sublattice model by Hillert et al.1 2 In the Division
of Metallurgy, KTH, a mathematical model has recently where fj is the activity coeYcient for element j in the
been developed1 3 which predicts the oxide activities in liquid molten steel, subscript i represents the dissolved elements
multicomponent slags. The model uses only experimental in the molten steel, and eij is the interaction parameter for
information from the corresponding binary subsystems. element j.
Furthermore, the model applies Temkin’s description of The oxygen activity in the steel bath aO can be calculated
entropy of ionic melts7 coupled with Lumsden’s description by solving the equilibrium reactions between slag and metal
of silicate and aluminate melts.1 4 Lumsden considered the with respect to dissolved alloying elements in the steel melt
ionic melt as an oxygen ion matrix with the cations, including and oxides in the slag phase. The general form of the
Si4 + , distributed in it. A hypothetical standard state for reactions determining the oxygen activity is
silica is adopted, where the SiO2 network is dissociated MeX O= X Me +O . . . . . . . . . . . (13)
into Si4 + and O2 – ions. More details on the model can be
found in published literature1 5 on the subject. To simplify the calculation it can be assumed that the
Another approach was taken by Ohta and Suito,1 6 equilibrium between dissolved aluminium and oxygen in
who used measured activity data to derive expressions by the steel bath and alumina in the top slag controls the
multiple regression analysis for activities of Al 2 O3 and SiO2 oxygen activity. Except for the dynamic modelling where
and activity coeYcients for FeO and MnO. The slag system reoxidation is considered, this assumption has been used
considered was Al2 O3 –CaO–MgO–SiO2 at 1600°C. Their in the study documented in this report. It is a reasonable
work resulted in equations by which activities and activity simpliŽ cation, since the oxygen aYnity of aluminium is very
coeYcients were easily calculated using the slag composition high compared to other alloying elements. The equilibrium
of interest. In the present work, the expressions by Ohta reaction is
and Suito were adopted for the calculation of oxide activities
in most cases. In future work, the aim is to test the use of 2Al +3O= Al2 O3 (s) . . . . . . . . . . . (14)
the KTH model for calculation of oxide activities in the
slag phase. DG °=­ 1205115 + 386´714T . . . . . . . . (15)
In Fig. 3, a comparison between the calculated activity of The data for the change of standard Gibbs free energy
Al2 O3 using data from the KTH model1 5 and the Ohta and (J mol – 1 ), equation (15), were taken from a compilation
Suito equation, is shown. The results reveal a discrepancy, by Hayes.1 8 Solid alumina was chosen as the standard
which decreases when the basicity increases. It should state. The equilibrium constant for equation (14) K 1 4 could
also be pointed out that basicity above 5 corresponds to be written as
SiO2 contents in the slag below 10%. The Ohta and Suito

A B
equation1 6 for alumina activity was not veriŽ ed below ­ DG ° aA l2 O 3
10%SiO2 , which means that extrapolation above basicity K 1 4 = exp = . . . . . . . . (16)
RT a2A l a3O
5 should be observed with care. The present authors have,
however, used the Ohta and Suito equation1 6 at SiO2 where a A l 2 O 3 is the activity of alumina in the slag phase
contents below 10%. and aA l is the activity of aluminium in the metal phase.
In order to calculate the oxygen activity a O from the
SULPHUR DISTRIBUTION above equation, the activities of aluminium in the molten
steel and alumina in the top slag need to be estimated. The
In order to relate the sulphide capacity to the equilibrium
activity coeYcient of aluminium in the molten steel fA l can
sulphur distribution between the slag and metal phases the
be calculated using Wagner’s equation (12). The aluminium
following reaction should be considered
activity is the product of fA l and the aluminium content in
S m e t a l +0·5O2 (g)= O m e t a l +0·5S2 (g) . . . . . . (8) the molten steel.
Ironmaking and Steelmaking 2002 Vol. 29 No. 3
 Andersson et al. Slag–metal reactions during ladle treatment 227

The activity of alumina in the top slag is estimated by

using the expression for activity of Al2 O3 in an Al2 O3 –
CaO–MgO–SiO2 slag at 1600°C by Ohta and Suito1 6
{­ 0´275 (%CaO)+ 0·167(%MgO)}
log aA l 2 O 3 =
(%SiO2 ) 4 Sampling procedure for the trials (sample 2 was
+0·033(%Al2 O3 )­ 1·560 . . . . . (17) excluded for heat A)

PLANT TRIALS AT OVAKO STEEL AB normalised composition of 32%Al2 O3 , 53%CaO, 7·5%MgO,

Process description and 7·5%SiO2 was used. Trials were performed during
vacuum degassing for both stirring with only argon (heats
Ovako Steel manufactures bearing steels and low alloy
A–C) and argon stirring in combination with inductive
speciality steels for highly stressed applications. The scrap
stirring ( heats D and E) in order to compare gas stirring
based steel plant is situated in Hofors. The annual raw steel
by itself with combined stirring.
production capacity is 500 kt for the year 2000.
Sampling was carried out before and after vacuum
The scrap is melted in a 100 t oval bottom-tapped
degassing for all the heats (samples 1 and 3 in Fig. 4). How-
(OBT) electric arc furnace. After adjusting the steel to the
ever, in order to study the variation of the sulphur concen-
desired phosphorous, carbon, and temperature levels, the
tration with time, the degassing practice was interrupted
steel is tapped into a ladle while undergoing predeoxidation.
after 3 min for heats B–E and an additional sample was
The ladle is then transported by an overhead crane to
taken, sample 2. After sample 2 was taken, the chamber
the deslagging station. Afterwards, the steel enters the
was again evacuated and the degassing process continued.
ASEA–SKF furnace from a ladle car.
Vacuum was reached within 2–3 min. Separate calcu-
The secondary reŽ ning process consists of three main
lations indicated that this period with continuously decreasing
steps. First, induction stirring is used during the alloying,
pressure approximately corresponded to 30 s under vacuum
deoxidation, and melting of the synthetic top slag. Second,
evacuation. To be able to compare heat A, which was not
gas stirring enhances the vacuum degassing operation, where
interrupted, with heats B–E, the period from when vacuum
hydrogen and sulphur reŽ ning are done. Argon gas is injected
was reached to when degassing was completed was pro-
through two porous plugs during vacuum degassing. Third,
longed from 10+3 min to 14·5 min for this heat. The total
induction stirring is used again, after the vacuum degassing
vacuum time then became 15 min (0·5+14·5 min) for heat
operation is completed, in order to further promote the
A and 14 min (0·5+ 3+ 0·5+ 10 min) for heats B–E. In
separation of inclusions from the steel. Upon completion
Table 2, total vacuum time is given together with the gas
of ladle treatment, the steel is cast using up-hill teeming
 owrate for porous plugs 1 and 2 and sulphur results from
into twenty-four 4·2 t ingots.
the analysis. It should be mentioned that the vacuum
Several full scale trials have been carried out at Ovako
pressure in the chamber varied between 1·3 and 1·8 torr
Steel. The Ž rst presented trials were carried out in order to
(173–240 Pa) for the heats.
investigate the in uence of slag properties, such as sulphide
capacity, viscosity, and oxide activity, on desulphurisation.
The results were used in the static modelling work. The RESULTS AND DISCUSSION
following plant trials were made to verify the dynamic Static modelling
modelling work. Parameter study
The eVect of the following parameters on the equilibrium
Static modelling work trials sulphur distribution LS was calculated: %Al and %C in the
Experiments were conducted for a high carbon, chromium molten steel, temperature, and (%Al2 O3 )/(%CaO) ratio in
bearing steel grade ( 1%C and about 1·4%Cr) with the the top slag.3 It was concluded from an earlier investi-
objective to evaluate how diVerent slag properties aVect the gation1 9 that the SiO2 and MgO components in the top
desulphurisation during ladle treatment.3 The top slag con- slag usually changed very little during the ladle reŽ ning
stituents varied during the trials as follows: 27–44%Al2 O3 , operation. Thus, suitable contents of SiO2 and MgO were
39–58%CaO, 3–16%MgO, and 4–11%SiO2 . A commercial chosen and kept constant during the calculations of the
synthetic slag mixture was used together with lime. In some whole parameter study. A typical bearing steel composition
cases (for investigating high alumina slags) pure alumina was chosen for the calculation, with the following major
was also added to the slag. Note that these slag com- alloying elements: 1%C, 0·28%Si, 0·28%Mn, and 1·4%Cr.
positions do not correspond to those used in the normal Results from the parameter study are shown in Fig. 5.
production.
Slag and steel samples were collected before and after
Table 1 Intervals of composition analysis, wt-%, of slag*
degassing and after the Ž nal stirring period. The temper- and steel samples BD, AD, and FS taken before
ature of the molten steel was measured at each sampling and after degassing and after final stirring period
occasion. Slag samples were collected manually with a slag respectively (Tm is measured temperature of
spoon. The temperature was measured and steel samples molten steel)
were taken using the automatic sampling equipment installed
Sample BD Sample AD Sample FS
at the ASEA–SKF ladle furnace. A dual thickness (‘lollipop’)
sampler from Rescon-Electro Nite was used to sample the Slag
liquid steel to determine the amount of sulphur as well as Al2 O3 27–42 30–44 30–44
other elements that dissolved in the steel. The sulphur con- CaO 45–58 39–54 40–54
tent was determined using a Leco CS244 (ASTM 1019–94). MgO 3–11 5–16 5–14
SiO 2 6–11 4–11 4–11
The results are given in Table 1. S 0·30–0·74 1·2–2·7 1·2–2·7
(1%C–1·4%Cr) steel
Dynamic modelling work trials Si 0·24–0·29 0·24–0·29 0·24–0·29
Mn 0·26–0·33 0·26–0·33 0·26–0·33
Experiments were conducted for Ž ve heats: A, B, C, D, and S 0·016–0·028 0·007–0·014 0·007–0·015
E, of high carbon, chromium bearing steel (1%C, 1·4%Cr) Al 0·039–0·086 0·021–0·051 0·025–0·050
in order to verify the model with respect to desulphurisation Tm , °C 1565–1618 1497–1552 1530–1544
during vacuum degassing.2 A synthetic top slag with a *Intervals of nomalised oxide compositions.

Ironmaking and Steelmaking 2002 Vol. 29 No. 3

228 Andersson et al. Slag–metal reactions during ladle treatment

6 Calculated equilibrium sulphur distribution plotted

5 Calculated equilibrium sulphur distribution as function against sulphur distribution determined from slag and
of Al2O3 content (wt-%) in top slag3 steel analyses3

Figure 5 indicates that when the temperature and the sulphur distribution. However, the results in Fig. 6 show
dissolved aluminium content in the molten steel both decrease that the equilibrium sulphur distribution tends towards an
at the same time, as they do in the parameter study, there equilibrium value during vacuum degassing.
is a minor eVect on the equilibrium sulphur distribution It can also be noted that the demands of Ovako Steel
between slag and metal. However, if the slag composition on the limits of sulphur content in the steel were always
changes at the same time, the in uence on the equilibrium fulŽ lled during the trials, although in many cases the slag
sulphur distribution could be substantial. For example, an composition was not optimal for sulphur reŽ ning.
increase in the Al2 O3 content in the top slag from 30 to
35% would theoretically decrease the equilibrium sulphur Dynamic modelling
distribution by about 50%. (Note that the values represent
The driving force for developing a dynamic model of the
equilibrium calculations.)
vacuum degassing operation is basically a lack of pro-
duction information (actually tools) for determination of
Evaluation of plant trial results steel cleanness during treatment. Today, it is very diYcult
As mentioned earlier, the main conclusion from the para- to take samples during the degassing operation. A model
meter study3 was that a change in the %Al2 O3 /%CaO would help the operators to decide on the required treatment
ratio has the greatest in uence of the studied parameters time with respect to the prevailing operational conditions.
on the sulphur distribution ratio. Therefore, speciŽ c plant Figure 7 shows a schematic layout of the dynamic vacuum
trials were performed3 as described above, where the degassing model. At the bottom of the Ž gure, there are
alumina content before vacuum treatment was allowed to
vary from 27 to 44%.
The equilibrium sulphur distribution L S was estimated
for all heats by using sulphide capacities calculated by the
KTH model for the diVerent slag samples. The oxygen
activities were estimated as described earlier. The results
are shown in Fig. 6, where the estimated equilibrium sulphur
distributions are plotted against the sulphur distributions
from analysis of the slag and steel samples. It can be seen
that there was no equilibrium between the top slag and the
molten steel with respect to sulphur content, before degassing.
Just after degassing, the estimated equilibrium sulphur distri-
butions agreed reasonably well with the analysis deter-
mined sulphur distributions. The agreement even somewhat
improved after the Ž nal heating and stirring period at the
end of ladle treatment. It should be noted that this com-
parison is based on equilibrium conditions. In reality, the
experimental data at the end of vacuum degassing may not
represent an equilibrium situation. Moreover, a previous
study1 9 showed that slag sampling aVects the experimentally
determined slag sulphur content and consequently also the 7 Diagram showing vacuum degassing model

Table 2 Sulphur content of steel melts for heats A, B, and C (gas stirring only) and heats D and E (combined stirring)
during vacuum treatment: interrupted sampling (after 3 min) was not carried out for heat A
Total time at vacuum pressure Gas flowrate, NL min –1 Sulphur content [%]
Heat Approx., min Used in model, s Ar(1) Ar(2) Before 3 min After

A(G) 15 900 200 130 0·023 … 0·007

B(G) 14 848 130 140 0·023 0·015 0·012
C(G) 14 840 150 50 0·022 0·018 0·012
D(C) 14 840 110 50 0·019 0·014 0·009
E(C) 14 816 130 130 0·022 0·014 0·010
(G): gas stirring, (C): combined stirring.

Ironmaking and Steelmaking 2002 Vol. 29 No. 3

 Andersson et al. Slag–metal reactions during ladle treatment 229

agreement between the predictions and the measurements

was good, but closer inspection revealed some systematic
deviations.
According to the results in Table 3, the model predicted
somewhat too high sulphur reŽ ning rates in the initial stage
of the degassing process ( heats C–E). This may be caused
by the fact that the model did not take reoxidation from
top slag or refractory into consideration, thus using too
low oxygen potentials. The predicted sulphur level at the
end of the operation was also a little too high (all heats).
One reason for this may be solid precipitation of calcium
sulphide in the top slag during the vacuum degassing,
which should result in higher sulphide capacity than was
anticipated. Another reason for deviations at the end of
the operation might be the weight of the top slag input
to the model. A higher amount of top slag would, given a
certain value of the sulphide capacity, be able to capture a
8 Comparison between predicted sulphur content and larger amount of sulphur from the steel.
results from plant trials for heat B (Ref. 2)
Figure 9 demonstrates calculated concentration proŽ les
of sulphur after 3 min of vacuum treatment for two heats
thermodynamic models describing, for example, sulphide (A and D in Table 3). In the Ž rst heat, only inert gas
capacity and activities, along with thermophysical models injection was used as the stirring method. In the second
describing physical properties, such as diVusion constants heat, combined gas stirring and inductive stirring was used.
and viscosity. Together with transport equations for the The results clearly demonstrate the feasibility of the
diVerent phases, reaction models for reactions in uencing modelling approach for the operating conditions existing
the sulphur reŽ ning can be implemented. In this way, the at Ovako Steel. The model predictions were in agreement
dynamic model rests on a scientiŽ c foundation and is able with the experimental data almost to the level of measure-
to capture dynamical changes in temperature and com- ment error, but relied on accurate input data. From experi-
position due to the diVerent reactions. The thermodynamic ence in the project, the parameters causing the greatest
and thermophysical models described earlier in this paper errors in the predictions were the oxygen activity and the
have been used in the dynamic model, except when speciŽ ed slag weight at the start of the operation (used as input to
otherwise in the following sections. the model). Therefore, for accurate prediction of sulphur con-
The oxygen activity in the bulk of the steel melt was centration during vacuum degassing, a rigorous deslagging
calculated by using the activity of aluminium in reaction practice is required together with suYcient knowledge of
(14). In the bulk of the steel melt, it was assumed that the the thermodynamic and physical properties of the synthetic
activity of alumina was unity.2 The steel bulk was deŽ ned slag. With these conditions met, the presented modelling
as liquid metal containing less than 1% top slag. In the approach has the full potential to predict sulphur reŽ ning
slag–metal mixing zone, however, the activity of oxygen during degassing with the accuracy normally needed in
was determined by the equilibrium reactions between the industrial practice.
liquid slag and steel. It was further assumed that the eVect
of changes in slag composition on the sulphide capacity was Simulation study including reoxidation reactions
negligible during degassing. The KTH model6 was applied To investigate the in uence of reoxidation on desulphurisation
to obtain values of the sulphide capacity of the slag phase. the simultaneous consideration of the following slag–metal
In this model, besides the temperature, only Al 2 O3 , CaO, reactions, besides the sulphur–oxygen exchange reaction,
MgO, and SiO2 in the slag phase were assumed to in uence was necessary4
the value of the sulphide capacity.
Fet O( l )= Fe (l )+ O . . . . . . . . . . . (19)
Verification of three-dimensional model MnO(s)=Mn+ O . . . . . . . . . . . (20)
The three-dimensional model is a description of the ladle
used at Ovako Steel AB, with bottom gas stirring through SiO2 (s)=Si+ 2O . . . . . . . . . . . . (21)
two porous plugs and an electromagnetic stirrer. The Al2 O3 (s)=2Al+ 3O . . . . . . . . . . . (22)
oxygen activity in the slag–metal mixing zone was calculated
according to the equilibrium reaction These reactions, together with equation (11), in uence the
oxygen activity and the sulphur and aluminium contents
2Al+3O=(Al2 O3 )s la g . . . . . . . . . . (18) in the steel melt. In order to calculate the activities of these
The activity of Al2 O3 as a function of temperature was oxides in the slag phase, the empirical expressions suggested
obtained from the Tdcalc database2 for a slag containing by Ohta and Suito1 6 were adopted.
32%Al2 O3 , 53%CaO, 7·5%MgO, and 7·5%SiO2 . Any To be able to solve the thermodynamic equations for the
in uence from variations of the slag composition on the transfer of S, O, Al, Si, and Mn at every instant and for
activity of Al2 O3 was neglected. Variations of any other each calculation node, a separate transport equation2 was
elements but Al, O, and S in the molten steel were neglected solved for each of the dissolved elements in the steel phase.
in the model calculation.
In Fig. 8, the predicted average sulphur concentration is Table 3 Comparison between predicted and calculated
plotted as a function of time. The three experimentally sulphur content in steel
determined sulphur concentrations for that heat are also [%S] (t= 0) [%S] (3 min) [%S] (final)
shown (triangles) in order to compare the sampled concen-
Heat Sample Model Sample Model Sample Model
trations with the calculated curve. From the results, it was
evident that the sulphur content in the steel decreased A 0·023 0·023 … 0·014 0·007 0·008
rapidly during the initial stage (3–5 min) of degassing, after B 0·023 0·023 0·015 0·015 0·012 0·012
which the slope declined. The rather prompt drop of the C 0·022 0·022 0·018 0·014 0·012 0·012
reŽ ning rate was usually caused by the decreasing ability D 0·019 0·019 0·014 0·013 0·009 0·010
E 0·022 0·022 0·014 0·012 0·010 0·010
of the slag to absorb sulphur from the melt. The overall
Ironmaking and Steelmaking 2002 Vol. 29 No. 3
230 Andersson et al. Slag–metal reactions during ladle treatment
9 Calculated concentration profiles of sulphur in steel after 3 min of vacuum treatment with a only gas stirring
(heat A) and b combined gas and inductive stirring (heat D)
Separate transport equations for the slag phase were also
solved for the diVerent slag components (Al 2 O 3 , CaO,
MgO, SiO2 , MnO, FeO, and S). Thus, the concentration
proŽ les for the dissolved elements in the steel phase as well
as for oxides and sulphur in the slag phase were determined
at each instant.
When including the above reactions, the model was simpli-
Ž ed to two dimensions due to the need of extensive com-
puter resources. Only gas stirring was considered and the
inert gas was injected through a centrally placed porous plug
in the bottom of the ladle with a  owrate of 130 NL min – 1 .
The total pressure above the melt was held constant during
the calculation at 1·8 torr (240 Pa) in order to simulate
vacuum degassing.2 The slag density was a function of slag
composition and temperature.2 0 The slag was assumed to
be a completely liquid layer, initially without concentration
gradients. Time zero was deŽ ned as the time when the
open eye in the slag layer had been established according
to the  uid  ow calculation. The simulated process time
was 14·1 min. Each simulation was isothermal, i.e. the
temperature remained constant.
In Fig. 10a–d is a depiction of distribution of sulphur
in the slag. Here, the  uid  ow pattern in the slag was
re ected by the concentration proŽ le. The accumulation of
sulphur close to the ladle wall was caused by the presence
of a stagnant zone in the slag. After 9·5 min of vacuum
treatment the local maximum sulphur content in the slag
was about 2%. In Fig. 10e a scanning electron micrograph
obtained from an industrial slag sample is shown. The slag
sample was taken immediately when the vacuum treatment
was Ž nished. In this particular heat, the normalised slag
composition was 32·1%Al2 O3 , 55·5%CaO, 7·6%MgO, and
4·3%SiO2 . The sampling temperature was 1509°C. Pre-
cipitated particles of CaS were identiŽ ed in the slag sample.
If it is assumed that the CaS particles were precipitated
during the vacuum treatment, some explanations as to why
this would occur should be proposed. It is possible that local
enrichment of sulphur in the slag (as seen in Fig. 10a–d)
Ironmaking and Steelmaking 2002 Vol. 29 No. 3
together with a decrease in temperature causes a super-
saturation of sulphur in the slag, which could produce the
precipitation. The CaO content in this particular heat was
also rather high.
In order to evaluate the in uence of the initial FeO
content in the slag on desulphurisation and aluminium
loss, a parameter study was made4 for three diVerent slag
compositions where the initial FeO content in the slag was
0, 2, or 6%. The diVerent initial slag compositions in the
parameter study are speciŽ ed in Table 4. Figure 11 shows
how the model predicted the in uence of diVerent initial
contents of FeO in the top slag on desulphurisation at an
isothermal temperature of 1600°C. At 0% initial FeO, the
average sulphur content decreased from 0·023 to 0·013%
within 10 min. When the initial FeO content increased,
the desulphurisation rate decreased. At 2% initial FeO, the
average sulphur content was 0·015% after 10 min of vacuum
treatment. When the initial FeO content was increased even
further to 6%, the average sulphur content was 0·020%
after 10 min.
CONCLUDING SUMMARY
The results of an investigation of desulphurisation during
vacuum degassing have been reported and discussed. Some
slag properties, which in uence desulphurisation have been
discussed. Furthermore, the use of these slag property data
in static and dynamic modelling has also been presented
and compared with experimental plant data obtained at
Ovako Steel.
Models for predicting the sulphide capacity,6 viscosity9
and oxide activities1 3 ,1 5 ,1 6 in liquid slags as functions of slag
composition and temperature have been presented brie y.
New results from viscosity measurement for an Al 2 O3 –
CaO–MgO–SiO2 synthetic slag (representing a typical ladle
furnace slag) have been reported. The data have also been
compared with predictions using the KTH model and the
agreement was found to be good.
 Andersson et al. Slag–metal reactions during ladle treatment 231

11 Influence of different initial FeO contents in top

d slag on desulphurisation (heat size 100 t, isothermal
temperature 1600°C, gas stirring 130 NL min – 1,
S0 =0·023%)

The dynamic modelling of desulphurisation during

degassing including reoxidation (two-dimensional model)
indicated that a local accumulation of sulphur in the slag
could occur, when a recirculation  ow pattern is present.
This could cause a precipitation of calcium sulphide,
especially if the CaO content is high and the temperature
drop is large enough. Investigation of a slag sample by
scanning electron microscopy revealed precipitated particles
of CaS. The model results also showed that an increase in
the initial FeO content in the top slag signiŽ cantly lowered
the desulphurisation rate.

e CONCLUSIONS
10 a–d calculated sulphur concentration profiles in
The following conclusions can be drawn from the work,
slag phase at vacuum treatment times of 1·5, 4, 6·5, when studying desulphurisation during vacuum degassing.
and 9·5 min respectively (heat size 100 t, isothermal 1. Static modelling showed good agreement with plant
temperature 1600°C, gas stirring 130 NL min – 1 ): SEM data made at Ovako Steel AB.
of precipitation of CaS in sample from top slag after 2. A parameter study indicated that the top slag
vacuum treatment is shown in e composition, or to be more speciŽ c, the ratio Al2 O3 /CaO,
may have a substantial eVect on the sulphur distribution
between slag and metal during degassing.
Static modelling of sulphur distribution, based on 3. CFD modelling showed good agreement with plant
equilibrium calculations, was compared with plant trials, data from Ovako steel AB.
carried out at Ovako Steel. The sulphide capacity was 4. CFD modelling including reoxidation from the top
calculated using the KTH model. The oxygen activity was slags indicated that a local accumulation of sulphur in the
estimated using the equilibrium reaction between aluminium slag can occur, which may cause a precipitation of CaS.
and oxygen in the molten steel and aluminium oxide in the The authors would also like to point out some areas of
steel phase. The activity of aluminium oxide was calculated future research:
using an expression by Ohta and Suito.1 6 The agreement 1. The present dynamic modelling concept should be
was found to be good. A parameter study showed that the used for further research on slag–metal reaction during
slag composition (as %Al2 O3 /%CaO) could have a sub- steelmaking.
stantial eVect on the sulphur distribution between slag and 2. The fundamental research work to determine thermo-
metal during degassing. dynamic and thermophysical properties of slags should be
The dynamically predicted Ž nal sulphur contents in the continued.
molten steel are in good agreement with plant data for the 3. SimpliŽ ed versions of dynamic models should be
operating conditions at Ovako Steel AB (three-dimensional developed in order to use them as production tools for
model). The predicted sulphur content though was some- process control.
what lower in the beginning of vacuum treatment. It was
also found that the model was sensitive to the thermo-
dynamic models used to calculate the oxygen activity and REFERENCES
the input value of slag weight. 1. p. sjödin, p. jönsson, m. andreasson, and a. winqvist: Scand.
J. Metall., 1997, 26, 41–46.
2. m. hallberg, l. jonsson, and j. alexis: Proc. Conf. ‘ScanMet I’,
Table 4 Initial slag compositions used in dynamic Vol. 2, 119–169; 1999, LuleaÊ, MEFOS.
parameter study, wt-% 3. m. andersson, p. jönsson, and m. hallberg: Ironmaking
Steelmaking, 2000, 27, 286–293.
FeO Al2 O3 CaO MgO MnO SiO 2 S 4. l. jonsson, m. andersson, and p. jönsson: ‘Modelling of slag–
metal reactions during ladle treatment’, Technical Report
0 29·0 50·0 9·2 0·0 11·2 0·5 TO23–95, Jernkontoret, Stockholm, Sweden, March 2000.
2 28·4 48·9 9 0·2 11·0 0·5
6 27·2 46·9 8·6 0·2 10·5 0·5
5. c. j. b. fincham and f. d. richardson: Proc. R. Soc. A, 1954,
223, 40–62.

Ironmaking and Steelmaking 2002 Vol. 29 No. 3

232 Andersson et al. Slag–metal reactions during ladle treatment
6. m. m. nzotta, du sichen, and s. seetharaman: Metall.
T rans. B, 1999, 30B, 909–920.
7. m. temkin: Acta Phys. Chim. URSS, 1945, 20, 411–420.
8. y. kawai and y. shiraishi (eds.): ‘Handbook of physico-
chemical properties at high temperatures’; 1988, Tokyo, Iron
and Steel Institute of Japan.
9. s. seetharaman, du sichen, and j. y. zhang: JOM, 1999,
51, (8), 38–40.
10. h. gaye and j. welfringer: Proc. 2nd Int. Symp. on ‘Molten
slags and  uxes’, 357–375; 1984, Warrendale, PA, TMS.
11. m. l. kapoor and m. g. frohberg: ‘International metallurgical
chemistry: applications in ferrous metallurgy’, SheYeld, July
1971, The Iron and Steel Institute, 17–23.
12. m. hillert, b. jansson, b. sundman, and j. a·gren: Metall.
T rans. A, 1985, 16A, 261–266.
13. j. björkvall, du sichen, and s. seetharaman: Ironmaking
Steelmaking, 2001, 28, 250–257.
Ironmaking and Steelmaking 2002 Vol. 29 No. 3
14. j. lumsden: ‘Physical chemistry of process metallurgy’, Part 1,
165–205; 1961, New York, Interscience.
15. j. björkvall: ‘Thermodynamic study of multicomponent slags
– a model approach’, Doctoral thesis, TRITA–MET070,
Department of Materials Science and Engineering, KTH,
Stockholm, Feb. 2000.
16. h. ohta and h. suito: Metall. T rans. B, 1998, 29B, 119–129.
17. m. görnerup: ‘Studies of slag metallurgy in stainless steel-
making’, Doctoral thesis, TRITA–TPM053, p. 24, Department
of Metallurgy, KTH, Stockholm, 1997.
18. p. hayes: ‘Process principles in minerals & materials pro-
duction’, 633–652; 1993, Brisbane, Hayes Publishing.
19. m. andersson, p. g. jönsson, and m. m. nzotta: ISIJ Int.,
1999, 39, 1140–1149.
20. ‘Slag atlas’, 2nd edn, 157–160; 1995, Düsseldorf, Verlag
Stahleisen.