Journal of Materials Research and Technology

w w w. j m r t . c o m . b r

ORIGINAL ARTICLE

Estimating Viscosities in Iron and Steelmaking

Slags in the CaO-Al2O3-MgO-SiO2-(TiO2) System with
Basis on a Thermodynamic Model
Andr Costa e Silva1,*
1
EEIMVR, Universidade Federal Fluminense, Volta Redonda, Brazil.

Manuscript received May 13, 2012; in revised form September 6, 2012.

Slags are critical components in several iron and steelmaking processes. In steelmaking, different slags with various properties are used
in the different processing steps from melting and rening to continuous casting. Thermodynamic properties, viscosity and surface
tension are among the most important properties considered in the design of slag compositions for steel processing. A brief literature
review of the approaches at modeling viscosities shows that viscosity correlations with chemical composition, basicity (both chemical
and optical), and more recently with slag constitution have been attempted with different degrees of success. In this work, a model
based on the Riboud viscosity model combined with the thermodynamic Kappor-Frohberg-Gaye cell model is described and presented as
a potential alternative to obtaining slag viscosity estimates directly from computational thermodynamics software. The rst results of
the adjustment and application of the model to the CaO-Al2O3-MgO-SiO2 system are presented. The quality of the adjustment is discussed
as well as its current limitations. Future steps in the model development involve its extension to include FeO, TiO2, and CaF2.

KEY WORDS: Viscosity; Slag; Computational thermodynamics; Steelmaking.

2012 Brazilian Metallurgical, Materials and Mining Association. Published by Elsevier Editora Ltda. All rights reserved.

1. Introduction phosphides, and uorides) have very complex structures,

Slags have a fundamental role in all steelmaking processes,
from primary rening to continuous casting. Further to its
chemical behavior, normally dened by the slag thermody-
namic properties, some physical properties are also of par-
amount importance for the adequate performance of the
slag in different processes[13]. Among the physical proper-
ties, viscosity and surface tension are specially important.
However, oxide based slags (and those including suldes,
*Corresponding author.
E-mail address: andre.costaesilva@gmail.com (A. Costa e Silva)
and their properties, in particular viscosity, can vary in a
wide range of values. These facts make predicting or model-
ing slag viscosities a challenge[4]. As the capacity to properly
designing slags with basis on their desired thermodynamic
behavior improves, the capacity of being able to predict its
viscosity to properly adjust this property to the envisage
process becomes even more important.
In the present work, the state-of-the-art in viscosity
modeling is briey reviewed and a new model is presented.
This model is developed by introducing data from slag con-
stitution, calculated by a thermodynamic model that is sup-
ported by a physico-chemical structural model in an empiri-

2012 Brazilian Metallurgical, Materials and Mining Association. Published by Elsevier Editora Ltda. All rights reserved.

J. Mater. Res. Technol. 2012; 1(3):154-160

Estimating Viscosities in Iron and Steelmaking Slags in the
CaO-Al2O3-MgO-SiO2-(TiO2) System with Basis on a Thermodynamic Model
cal equation traditionally used to predict viscosity based on
the slag chemical composition. The results of the applica-
tion of this proposed model to a system of great interest to
steel processing slags (CaO-Al2O3-MgO-SiO2) are presented
and discussed.
2. Models for Slag Viscosity Prediction and
their Relations to Structure and Temperature
Sridhar[5] has recently reviewed the most relevant models
for describing the viscosity of slags and liquid metals. In this
context, the usual denition for viscosity as the ability of
a uid to resist shearing forces is used. Most slag systems
are Newtonian uids, where viscosity does not depend on
the shear rate, thus:
dv
W K
dy
where W is the shear stress [Pa], is the uid viscosity (a
coefcient given [Pas]) and dv is the velocity (v) gradient
dy
normal to the shear stress [s1].
2.1 Temperature Dependence of Viscosity
As shear should involve the relative movement of layers of
atoms, ions or molecules (or species) in a structure, and
that should depend on their bonding strength, it is reason-
able to assume that this force should follow an Arrhenius
type relation with temperature. This is assumed by several
authors[3,5]. So, viscosity could be described as:
E
K A exp A (1)
RT
where A is a pre-exponential coefcient and EA is the ac-
tivation energy for the phenomenon. However, due to the
complex nature of the species present in the slag and due to
the fact that the slag constitution (i.e. the species present)
may also vary with temperature[4,6], Eq. (2) has also been
proposed for the description of slag viscosity by Urbain[3,7]
based on the Weymann-Frenkel kinetic theory of liquids.
E cies in the quasi-chemical thermodynamic model for the
K A T exp A (2)
RT
2.2 Chemical Composition Dependency of Viscosity
Due to the complexity of slag structure and hence the spe-
cies present in the slag as a function of chemical compo-
sition, many different ways of incorporating the effect
of chemical composition in viscosity models have been
proposed. The simplest approach, classied as empirical
models by Kondratiev et al.[7], has been to t polynomial
descriptions to the relation between viscosity, tempera-
ture, and composition, not even considering Eqs. (1) or (2).
These models have only been successful for interpolation
(and minor extrapolations) in limited ranges of composi-
tion and temperature. On the other hand, both Eqs. (1) and
(2) can have their A and E coefcients tted directly to
J. Mater. Res. Technol. 2012; 1(3):154-160
155
the slag chemical composition, without taking in consider-
ation the species present in the slag, only its constituent
compounds. In an attempt to consider the behavior of the
various slag forming compounds, Urbain and other authors[4]
have proposed the grouping of the slag forming oxides in:
glass forming, modiers, and amphoteric. Silica (SiO2) is a
glass forming oxide. Silica is characterized by strong bond-
ing, polymerization (chains of tetrahedral), and high viscos-
ities. Modiers are oxides containing Na+, K+, Ca2+, Fe2+, and
Mg2+ ions that modify the polymeric glass structure breaking
the strong bonds (or chains) and decreasing the viscosity.
Oxides such as Al2O3 or Cr2O3 can display both behaviors,
depending on their concentrations and the other oxides
present in the slag[4]. This approach is very similar to the
simplied description of steelmaking slags, based on basic-
ity concepts[6]. In this approach[3], A and EA in Eqs. (1) and
(2) are expressed as shown in Eqs. (3) and (4).
In Aw A0  AX
component
i i
oxides i
(3)
EW E0  Ei X i
component
oxides i
where A is expressed in Pa.s.K-1 and E in K.
(4)
Aw exp[ A0  Amod ( X CaO  X MnO  X MgO  X FeO ) 
 A fluoride X CaF2  Aalkaline mod ( X Na2O  X K 2O )  AamphXXAl2O3 ]
EW E0  Emod ( X CaO  X MnO  X MgO  X FeO )
 E fluoride X CaF2  Ealkaline mod ( X Na2O  X K 2O )  EamphXXAl2O3
Eq. (4) is just a regrouping of terms of Eq. (3) for com-
pounds with the same approximate behavior[3].
More advanced models try to correlate viscosity with
structural characteristics or parameters of the slag. Thus,
viscosity has been correlated to optical basicity, for in-
stance[4,8]. Kondatriev and Jak[9], on the other hand, have
attempted a correlation between viscosity and the concen-
tration of species in the slag, given as the fraction of spe-
slag. These authors have adopted an intermediate tempera-
ture dependence as given in Eq. (5):
E
K AT 3 / 2 exp A (5)
T
Using the quasi-chemical model in FACTSAGE software
and slag databases they adjusted the coefcients A and E
as a function of species concentration obtained from ther-
modynamic calculation.
3. The Present Model
3.1 Model Formulation
In the present work, a model based on Eq. (2) and in a mod-
ication of Eq. (3) to consider the concentration of cells
156 Costa e Silva

(species) in the Gaye-Kapoor-Frohberg[1012] thermodynamic
model for the slag was developed. This way, no correla-
tion with the slag forming components (composition) was
attempted. This stems from the observation most models
that try to correlate composition and viscosity has limited
success. It is the assumption of the present work that this
limitation is due to the deviations between slag constitu-
tion and composition as discussed above.
The thermodynamic model used in the present viscosity
modeling was developed by Gaye et al.[10,11] based on the
Kapoor-Frohberg[12] proposal. In this model, liquid oxides
are described as symmetric (i-O-i) and asymmetric (i-O-j)
cells. Other cations can also be considered in an extension
of the model[11]. The model considers the cell formation en-
ergies as well as the energies of interactions between cells.
The Gibbs energy of symmetric cells corresponds to that of
the equivalent pure oxide. Thus, for instance:
G Al:O : Al N 0G Al2O3
where the factor adjusts for the number of oxygen atoms
in the cell and in the oxide formula.
Asymmetric cells have their Gibbs energies given as:
Gi:O: j Wij  N i 0 GixO y  N j 0 G j xO y
where Wij can be composition-dependent. Further-
more, the model has dened rules for the cell interaction
energies:
a) All interactions between similar cells are considered
as part of their Gibbs energy;
b) Interaction between different cells obey the follow-
ing additivity rules:
cies on the cell concentrations calculated according to this
model. Thus, if yij is the concentration of cell (in cell frac-
tion or mole fraction of cell) i-O-j in slag (where i and j are
cations), the viscosity coefcients in Eq. (2) become:
Aw exp( A0  y ij Aij )
i j ti
(10)
Ew E 0  y ij Eij
i j ti
In the present case, the yij are calculated using Thermo-
calc[14] and the Gaye-Kapoor-Frohberg model as implement-
ed in the database SLAG3[15] with basis on Gayes assess-
ments[10,11].
Access to structural information of the slag constituents,
however, is very limited. The insight proposed by Richard-
son in 1954, that there must exist three types of oxygen in
slag, namely those bonded to two silicon atoms (O0), those
bonded to one silicon atom (O-1), and those unattached to
(6) silicon (O-2)[16] has paved the way for several slag structure
models[17,18] (including the Kapoor-Frohberg approach) ap-
plied to steelmaking slags. The concentration of these dif-
ferent types of oxygen can be directly measured by X-ray
photoelectron spectroscopy (XPS) if one assumes that oxygen
bonds are preserved when a liquid slag is quenched to a vit-
reous sample[19]. Comparison of the measured concentrations
(7)
of O0, O1, and O2 should, thus, give interesting insights on
slag structure and adequacy of thermodynamic models. For
this reason, the measurements of Park and Rhee[19] in CaO-
SiO2 mixtures were compared with the results of the thermo-
dynamic model discussed above. Fig. 1 compares their mea-
surement results, their calculated concentrations obtaining
by tting the degree of dissociation of CaO to the XPS results
and using Toops data with the calculated concentrations (Y) of

Eii  jj 2 Eij ii (8a)

1 0.10
Si-O-Si, Si-O-Ca, O-, O0 concentration (mole fraction)

0.9 0.09
Eij  kk Eik  kk  E jk  kk (8b)

Ca-O-Ca and O-2 concentration (mole fraction)

0.8 0.08

The Eij factor can be dependent on the concentration of 0.7 0.07

the highest cation charge concentration. 0.6 0.06
Thus, the Gibbs energy of the liquid slag can be calcu-
lated as: 0.5 0.05

0.4 0.04

G L
x G
i i
ii i  o i  xij Gi o  j 0.3 0.03

(9) 0.2 0.02

 N i  j Eij  RT xij lnxij 0.1 0.01

i j i j ti
0 0.00
0.4 0.45 0.5 0.55 0.6 0.65 0.7
where the x coefcients are the cell concentrations. Their Mole fraction SiO2
variation is limited by the concentration of the oxides form-
ing the slag and determined through the minimization of Y(Si-O-Si) Oo calc.[19] Y(Ca-O-Ca)
the Gibbs energy of the liquid phase for a given slag com- Y(Si-O-Ca) O- exp.[19] O-2 calc.[19]
O- calc.[19] Oo exp.[19] O-2 exp.[19]
position. The coefcients Ni-j are related to the number of
interacting pairs of cells and are determined in the same
way. Despite their different formulations, according to Pel- Fig. 1 Concentrations of O0 (), O-1(), and O-2 () experimentally
ton[13], this model and the quasi-chemical model are for- measured by XPS at 1,873 K, together with their calculated concentrations
tted to the data by Park and Rhee (see original reference for further
mally similar. discussion[19]). Also presented are the cell concentrations (mole fraction) of
It is then possible to describe viscosity by adapting Ca-O-Ca, Si-O-Si, and Si-O-Ca cells calculated using Thermo-calc and the
Eq. (2) with A and E factors that have linear dependen- SLAG3 database, in the CaO-SiO2 system

J. Mater. Res. Technol. 2012; 1(3):154-160

Estimating Viscosities in Iron and Steelmaking Slags in the
CaO-Al2O3-MgO-SiO2-(TiO2) System with Basis on a Thermodynamic Model
Ca-O-Ca, Si-O-Si, and Si-O-Ca cells using the above thermo-
dynamic model. It is evident that the model describes quite
well the concentration of the different types of oxygen via
the cell concentration. Furthermore, this can give important
insights on the relevant cells for the viscosity modeling, as
discussed in the next section.
4. Results and Discussion
4.1 First Stage of Modeling
The rst stage of modeling was focused on conrming the
viability of the PARROT (Parameter Optimization) routine of
Thermo-calc[14] in optimizing through least-squares, non-ther-
modynamic data, and viscosity in the present case. In that
stage[20] data concerning viscosities in the SiO2-Al2O3-MgO-CaO
system were collected from various sources (for binaries,
mostly from Kondratiev and Jak[9] and for ternaries data from
Toplis and Dingwell[21]). Data with log 2 has been elimi-
nated since the viscosity for these glasses frequently require
two step dependence on temperature.
Data for the quaternary system from Shankar et al.[22] as
well as Nakamoto et al.[23] were used. Overall, the data were
in the following approximate range of compositions: 7%80%
SiO2, 0%70% Al2O3, 0%16.5% MgO, and 8%70% CaO. The
range of temperatures in the data set was 16232273 K.
Thus, the equations to be adjusted were:
(11)
A A0  yij Aij A0  y SiSi ASiSi  y SiCa ASiCa  y SiAl ASiAl
i j ti
 y SiMg ASiMg  yCaCa ACaCa  yCaAl ACaAl  yCaMg ACaMg 
y AlAl AAlAl  y AlMg AAlMg  y MgMg AMgMg
Ew E0  yij Eij E0  y SiSi ESiSi  y SiCa E SiCa 
i j ti
y SiAl ESiAl  y SiMg ESiMg  yCaCa ECaCa  yCaAl ECaAl 
yCaMg ECaMg  y AlAl E AlAl  y AlMg E AlMg  y MgMg EMgMg
4.1.1 Results at the rst stage of modeling
The strategy for the optimization of the various parameters
is discussed elsewhere[20]. It should be noted, however, that
the model has only 2x3 adjustable parameters per binary
system and no higher order parameter. It is hence expected
that optimization should converge to consistent param-
eters. In the rst stage of modeling A and E coefcients
depending on y2Si-O-Si were introduced and only 8 variables
were actually used. For the range of MgO contents present
in the data only coefcients for ySi-O-Mg were used. The re-
sults of this stage of optimization of the model coefcients
are summarized in Fig. 2.
The results of the optimization of the coefcients shown
in Fig. 2 indicate that deviation exists for the adjustment
of some experimental values but the overall agreement
is good for a wide range of viscosity values, associated to
temperature and composition variations.
J. Mater. Res. Technol. 2012; 1(3):154-160
157
2.5
2.0
Calculated values of log ( in Pa.s)
1.5
1.0
0.5
0
-0.5
-1.0
-1.5
-1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0 2.5
Experimental values of log ( in Pa.s)
Fig. 2 Calculated values of logarithm of viscosities [log K (K in Pa.s)]
compared to experimental values of the same variable. Calculation
performed using the rst attempt at modeling viscosity given by Eq. (11)
(see text) in the CaO-MgO-SiO2-Al2O3 system[20]
4.2 Second Stage of Modeling
As the optimizing strategy has been proved adequate in the
rst stage of modeling, the second stage expanded the num-
ber of experimental data and attempted to introduce an ad-
ditional oxide, TiO2 in the viscosity model. Additional data
were selected from Shankar et al.[22], Nakamoto et al.[23],
and Saito et al.[24]. Special care was taken to eliminate data
in which a solid phase could be present. This was done via
the equilibrium calculation for the bulk slag composition
and temperature in Thermo-calc. This is specially impor-
tant since the effect of the presence of solid particles on
the apparent viscosity is well known[25].
Fig. 3 presents the results of the new optimization for
the quaternary system[26]. In this optimization no term
2
on y Si O  Si was used.
The results in Fig. 3 were obtained with 16 adjustable
variables. As at least four of them have low correlation, as
indicated in Parrot, it should be possible to achieve similar
results with only 12 adjustable variables.
4.2.1 Slags containing TiO2
In order to extend the model to slags containing TiO2, a rst
attempt was made to identify which cells containing Ti+y ions
would be more relevant. Ti-O-Ca and Si-O-Ti, for the range
of slags explored, were the Ti containing cells with higher
concentration. Coefcients for these cells were included in
Aw and Ew [Eq. (10)]. Fig. 4(a) shows that the adjustment of
the model was not good. The deviations were concentrated
in the titanium containing slags. Examining the constitution
of these slags, two important facts were observed: (a) the
concentration of the Ti containing cells is never high, in the
range of TiO2 contents of 0%10% mass; and (b) the addition
 158 Costa e Silva

of TiO2 causes important changes in the concentration of y ij y ij

A A0  Aij Ew E 0  Eij
other more relevant cells, as shown in Fig. 5. i j ti yT i j ti yT
For this reason, a change was introduced in the model
and the concentration of Ti containing cells was not consid- where:
ered, while the concentration of the cells already included (12)
in the model were normalized to consider the elimination
of the Ti containing cells, as shown in Eq. (12). Thus, the yT ySiSi  ySiCa  ySiAl  ySiMg  yCaCa  yCaAl 
direct effect of the concentration of Ti containing cells was
not considered in the model. yCaMg  y AlAl  y AlMg  yMgMg

2.5
Fig. 4b shows the improved adjustment of the calcu-
lated values when compared to Fig. 4a. This indicates that
2.0 the most relevant aspect of the structure of the slag (in
what concerns viscosity) is probably the concentration of
Calculated values of log ( in Pa.s)

1.5 the main types of oxygen ions. This apparently conrms

the proposal of other authors[27] that the lower concentra-
1.0 tion oxides in a slag (such as TiO2, in the present case) may
have a strong inuence of the fraction of bridging and
0.5 non-bridging oxygen ions and thus affect the slag prop-
erties. While the agreement between calculated cell con-
centration and the measured concentrations of the differ-
0
ent oxygen types in slags in Fig. 5 is not very good, the
trends displayed by the results of thermodynamic model
-0.5 are consistent with the data and the effect of TiO2 addition
on the concentration of Si-O-Si and Si-O-Ca cells is quite
-1.0 evident. One must speculate how would the concentration
of other cells such as Al-O-Al and Si-O-Mg contribute to the
-1.5 concentration of the different oxygen types measured
-1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0 2.5 via XPS.
Experimental values of log ( in Pa.s)
5. Conclusions
Fig. 3 Calculated values of logarithm of viscosities [log K (K in Pa.s)]
compared to experimental values of the same variable. Calculation A model for the calculation of viscosity of slags based on
performed using Eq. (10) adjusted via optimization using 325 experimental the slag constitution calculated by the cells thermody-
data points

(a) 2.5 (b) 2.5

2.0 2.0

1.5 1.5
Calculated values
Calculated values

1.0 1.0

0.5 0.5

0 0

-0.5 -0.5

-1.0 -1.0

-1.5 -1.5
-1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0 2.5 -1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0 2.5
Experimental values Experimental values
Fig. 4 (a) Calculated values of logarithm of viscosities [(log K (K in Pa.s)] compared to experimental values of the same variable. Calculations performed using
the parameters in Eq. (3) for various temperatures and compositions in the CaO-MgO-SiO2-Al2O3-TiO2 including cells containing Ti+y ions in the model description;
(b) calculated values of logarithm of viscosities [log K (K in Pa.s)] compared to experimental values of the same variable. Calculations performed using the
parameters in Eq. (4) (see text for discussion) in the CaO-MgO-SiO2-Al2O3-TiO2. Improved adjustment of the results when compared to Fig. 4a

J. Mater. Res. Technol. 2012; 1(3):154-160

Estimating Viscosities in Iron and Steelmaking Slags in the
CaO-Al2O3-MgO-SiO2-(TiO2) System with Basis on a Thermodynamic Model 159

(a) 0.45 (b) 0.38

O-
O0 0.36 O-
0.40

Species concentration
Species concentration

O0
0.34
0.35
Y(Si-O-Ca)
0.32
Y(Si-O-Ca)
0.30
0.30
0.25
0.28 Y(Si-O-Si)
0.20 Y(Si-O-Si) 0.26

0.15 Y(Si-O-Mg) 0.24

Y(Al-O-Al)
0.10 0.22
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Mole percent TiO2 Mole percent TiO2

Fig. 5 (a) Calculated cell concentrations using Thermo-calc compared to measured[27] O and O0 concentration in a slag with B = 1.2 (CaO/SiO2), 17%Al2O3,
and 10% MgO, as a function of the SiO, as a function of the TiO2 content; (b) calculated cell concentrations using Thermo-calc compared to measured[27]
O and O0 concentration in a slag with B = 0.8 (CaO/SiO2), 17% Al2O3, and 10% MgO, as a function of the SiO, as a function of the TiO2 content

namic model of Gaye-Kapoor-Frohberg was implemented
and the coefcients adjusted using the least-square rou-
tine in Thermo-calc, Parrot. The importance of censor-
ing viscosity data in which a solid fraction might be pres-
ent was highlighted.
The model was adjusted to viscosities in the
CaO-MgO-SiO2-Al2O3 system in different ranges of composi-
tion and temperature. The agreement was good and the
number of adjustable parameters was kept under 16.
It was possible to demonstrate that the cells concen-
tration calculated in the present thermodynamic model is
related to the types of oxygen in a slag (O0, O1, and O2)
related to the bridging concept of oxygen in slags.
In view of the physical insight gained by searching
the correlation of cells and states of oxygen in the slag,
a novel approach was proposed to describe the effect of
dilute additions to these slags, in which the effect on the
viscosity is modeled by considering only the effect of the
minor oxides on the concentration of the most important
cells in the system.
Acknowledgements
The support of Faperj through a Pronex project coordinated
by F. Rizzo is gratefully acknowledged. The collaboration of
the B Stocco, W Effgen, and B Gandra in different stages of
this work is also acknowledged.
References
1. Ito K, Fruehan RJ. Study on the foaming of CaO-SiO2-FeO slags:
Part I. Foaming parameters and experimental results. Metall
Mater Trans B 1989; 20B(August):50914.
2. Fruehan RJ (ed.). Making, Shaping, and Treating of Steel. Steel-
making and Rening Volume. 11.ed. Pittsburgh PA: AISE Steel
Foundation; 1998.
3. Riboud P, Roux Y, Lucas L, Gaye H. Improvement of continu-
ous casting powders. Fachber. Huttenprax. Metallweiterverarb
1981; 19:85969.
4. Verein Deutscher Eisenhttenleute. Slag Atlas. Dusseldorf: Ver-
lag Stahleisen, 1995; p. 616.
5. Sridhar S. Estimation Models for Molten Slag and Alloy Viscosi-
ties. JOM 2002; 54(11):46-50.
6. Richardson FD. Physical Chemistry of Melts in Metallurgy. Lon-
don: Academic Press; 1974.
7. Kondratiev A, Jak E, Hayes PC. Predicting Slag Viscosities in
Metallurgical Systems. JOM 2002; 54(11):41-5.
8. Nakamura T, Yokoyama T, Toguri JM. Limitations in the metal-
lurgical applications of optical basicity. ISIJ International 1993;
33(1):2049.
9. Kondratiev A, Jak E. A Quasi-Chemical Viscosity Model for Fully
Liquid Slags in the Al2O3-CaO-FeO-SiO2 System. Metall Mater
Trans B 2005; 36B:62338.
10. Gaye H, Welfringer J. Modelling of the thermodynamic prop-
erties of complex metallurgical slags. In: Second International
Symposium on Metallurgical Slags and Fluxes. Lake Tahoe, Ne-
vada: TMS-AIME; 1984.
11. Gaye H, Lehmann J, Matsumiya T, Yamada W. A statistical ther-
modynamics model of slags: applications to systems containing
S, F, P2O5 and Cr oxides. In: 4th International Conference on
Molten Slags and Fluxes. Sendai, Japan: ISIJ; 1992.
12. Kapoor ML, Frohberg MG. Theoretical treatment of activities in
silicate melts. in Chemical Metallurgy of Iron and Steel, Proceed-
ings of a Conference. University of Shefeld: ISI London; 1971.
13. Pelton AD. Solution models. In: Sano N et al. (eds.). Advanced
Physical Chemistry for Process Metallurgy. San Diego, CA: Aca-
demic Press, 1997; p.87116.
14. Sundman B, Jansson B, Andersson JO. The Thermo-Calc data-
bank system. CALPHAD 1985; 9:15390.
15. SLAG3 Database. Thermo-calc Software AB. Stockholm, Swe-
den; 2008
16. Fincham CJB, Richardson FD. The Behaviour of Sulphur in Sili-
cate and Aluminate Melts. Proc R Soc Lond A 1954; 223:4062.
17. Wagner C. The Concept of Basicity of Slags. Metall Mater Trans B
1975; 6B:4059.

J. Mater. Res. Technol. 2012; 1(3):154-160

160
18. Toops GW, Samis CS. Activities of ions in silicate melts. Transac-
tions AIME 224 (October):87887.
19. Park JH, Rhee PC-H. Ionic Properties of Oxygen in Slags. J Non-
Cryst Solids 2001; 282:714.
20. Stoco BN, Martins AAR, Costa e Silva A. Estimativa da viscosidade
em escrias: aplicao no sistema CaO-Al2O3-MgO-SiO2. In: 65 25. Avillez RR, Costa e Silva A, Fortes RAFO. Thermodynamic stud-
Congresso ABM Internacional. Rio de Janeiro, 2010. Anais do 65
Congresso ABM Internacional. So Paulo: ABM, 2010; p.27508.
21. Toplis MJ, Dingwell DB. Shear viscosities of CaO-Al2O3-SiO2 and
MgO-Al2O3-SiO2 liquids: Implications for the structural role of
aluminium and the degree of polymerisation of synthetic and
natural aluminosilicate melts. Geochim Cosmochim Acta 2004;
68(24):516988.
22. Shankar A, Gornerup M, Lahiri AK, Seetharaman S. Experi-
mental Investigation of the Viscosities in CaO-SiO2-MgO-Al2O3
and CaO-SiO2-MgO-Al2O3-TiO2 Slags. Metall Mater Trans B 2007;
38B:9115.
Costa e Silva
23. Nakamoto M, Tanaka T, Lee J, Usui T. Evaluation of Viscosity
of Molten SiO2-CaO-MgO-Al2O3 Slags in Blast Furnace Operation.
ISIJ International 2004; 44(12):21159.
24. Saito N, Ori N, Nakashima K, Mori K. Viscosity of Blast Furnace
Type Slags. Metall Mater Trans B 2003; B34(5):50916.
ies of steelmaking slags with respect to foaming in Electric Arc
Furnace. In: XXXIX CALPHAD. Jeju, Korea, 2010. XXXIX CALPH-
AD Program and Abstracts, 2010; p.146.
26. Oliveira WE, Gandra BF, Costa e Silva A. Estimativa de viscosi-
dade em escrias extenso da aplicao em sistemas a base
de CaO-Al2O3-MgO-SiO2. 67 Congresso Internacional da ABM, Rio
de Janeiro, Brazil, 2012.
27. Sohn I, Wang W, Matsuura H, Tsukihashi F, Min DJ. Inuence of
TiO2 on the Viscous Behavior of Calcium Silicate Melts Contain-
ing 17 Mass% Al2O3 and 10 Mass% MgO. ISIJ International 2012;
52(1):15860.
J. Mater. Res. Technol. 2012; 1(3):154-160