Desulfurization

by

Eugene Pretorius and Helmut Oltmann

Process Technology Group – LWB Refractories
1 Introduction
In most steel grades sulfur has to be minimized below a certain level. Sulfur control is important
throughout the iron and steelmaking process, from careful choice of raw materials, hot metal
desulfurization, limited removal in primary steelmaking and the primary desulfurization process
in the ladle. The main sulfur removal mechanism is the transfer of sulfur from the metal to the
slag phase. This section reviews the fundamental principles of sulfur removal, with specific
elaboration on the kinetics of desulfurisation in the ladle and thoughts on sulfur behavior in the
EAF.
It is generally known that the following parameters are important for good sulfur removal from
the metal to the slag.
• Liquid and fluid slag with a high dissolved lime content – high sulfur/sulfide capacity (Cs).
• Higher temperature, which improves both thermodynamics and kinetics.
• Oxygen content of the steel and oxidation state of the slag – lower is better.
• Decent slag volume – higher slag volume removes more S, for same distribution ratio.
• Mixing (stirring) of the steel – kinetics of mass transfer to slag metal interface.
• The extent of slag and metal inter-mixing at the interface (stirring intensity and slag fluidity)

2 Thermodynamic principles of sulfur removal

2.1 Sulfur removal reaction and sulfide capacity

The following reaction is generally accepted to describe sulfur transfer from the metal to the
slag:
[S] + (O 2− ) ⇔ (S 2− ) + [O] (1)

Where [X] indicates X dissolved in the metal (infinite dilution standard state) and (X) indicates
species dissolved in the slag. The equilibrium constant for this reaction cannot be evaluated
directly, however, due to difficulty in describing the activity of the ionic species in the slag (O2-
and S2-) and has to be determined experimentally. Experimental determination is usually by gas-
slag equilibration techniques based on the following reaction:

1 S + O 2− = S 2− + 1 O (2)
2 2 2 2

Slag samples of known composition are equilibrated with a gas of controlled pS2 (partial
pressure of sulfur) and pO2 and the equilibrium sulfur content of the slag determined by analysis.
The sulfide capacity (CS) can then be calculated according to equation (3), which is based on the
expression for the equilibrium constant of reaction (2).

1
p O2 a 2−
CS = 2
(%S) = K ( 2) O (3)
1
p S2 γ S 2−
2

LWB Refractories 2 Process Technology Group

The term sulfide capacity (CS) is used to describe a completely liquid slag's potential ability to
remove sulfur and is a function of slag composition and temperature. Sulfide capacities for
various slags have been experimentally determined and examples are shown in Figures 1 and 2.
Sulfide capacities are usually expressed in log units, for example: log CS = -3.2. The less
negative the logarithmic unit (larger the actual number), the better the sulfur removing capacity
of the slag, i.e., a slag with log CS = -1 has a better sulfur removing capacity than a slag with a
log CS = -3. Generally the sulfide capacity increases as the CaO-content increases in these slags.

Figure 1. Iso-sulfide capacity curves for CaO-Al2O3-MgO slags at 1600°C.

(Numerical values show - log CS.)

Figure 2. Iso-sulfide capacity curves for CaO-Al2O3-SiO2 slags at 1600°C.

(Numerical values show - log CS.)

Industrial slags are invariably more complex than the slag systems of which the sulfide capacity
has been determined experimentally and a general expression for Cs as a function of slag
composition and temperature would be of great use. The correlation of sulfide capacity with

LWB Refractories 3 Process Technology Group

Optical Basicity (OB) has been shown to have a good correlation over a wide range of slag
compositions and is therefore widely used.

2.2 Optical basicity

Table 1 shows optical basicity (Λ) values for the most common slag components.

Table 1. Values of optical basicity (Λ) of slag components.

Oxide Optical Basicity (Λ)

Na2O 1.15
CaO 1.0
MgO 0.78
CaF2 0.67
TiO2 0.61
Al2O3 0.61
MnO 0.59
Cr2O3 0.55
FeO 0.51
Fe2O3 0.48
SiO2 0.48

The average optical basicity (Λ) for a slag of any composition can be calculated by means of the
expression:
Λ = XAoxΛAOx + XBOyΛBOy + ... (4)
where,
mole fraction of component * number of oxygen atoms in oxide molecule (5)
X=
Σ mole fraction of component * number of oxygens in oxide molecule of all components

Thus, in a CaO-Al2O3-SiO2 slag for example:

N CaO
X CaO =
N CaO + 3N Al2O3 + 2 N SiO 2
3N Al2O3
X Al2O3 =
N CaO + 3N Al2O3 + 2 N SiO 2

2 N SiO 2
X SiO 2 =
N CaO + 3N Al2O3 + 2 N SiO 2

While the optical basicity is currently the best compositional parameter available that can be
used to describe the composition (basicity) of a slag, it does not say anything about the physical
properties of a slag. For example, consider the following "slag" composition:

LWB Refractories 4 Process Technology Group

 % CaO 62
% MgO 8
% SiO2 30

The optical basicity of this slag can be calculated as Λ = 0.756, which could be considered
sufficiently basic to ensure good sulfur removal and minimum refractory wear. However, the
calculation of this number is meaningless because the slag listed above is completely solid at
steelmaking temperatures, and it will only start to melt at 1790°C and will be fully molten at
1950°C.

The application of the optical basicity concept to steelmaking slags is only useful if the slags are
completely melted.
From the above illustration it is clear that significantly more information, such as solidus and
liquidus phase relations and viscosity data, are required in order to completely evaluate slags that
would be suitable in, for example, the ladle furnace.

2.3 An optical basicity calculation example

Calculate the optical basicity of the following slag:

% CaO - 45, % SiO2 - 35, % MgO - 10, % Al2O3 - 10

Solution:

Optical basicity (Λ) Molecular weight

CaO 1 56.08
SiO2 0.48 60.09
MgO 0.78 40.31
Al2O3 0.61 101.96

Moles CaO: 45/56.08 = 0.802

Moles SiO2: 35/60.09 = 0.582
Moles MgO: 10/40.31 = 0.248
Moles Al2O3: 10/101.96 = 0.098
Total moles = 1.730

Mole fraction (N) CaO: 0.802/1.73 = 0.464

Mole fraction (N) SiO2: 0.582/1.73 = 0.336
Mole fraction (N) MgO: 0.248/1.73 = 0.143
Mole fraction (N) Al2O3: 0.098/1.73 = 0.057

Total Base value = NCaO + 2NSiO2 + NMgO + 3NAl2O3

= 0.464 + 2(0.336) + 0.143 + 3(0.057)
= 1.45

LWB Refractories 5 Process Technology Group

XCaO = NCaO/Total Base 0.464/1.45 = 0.320
XSiO2 = 2NSiO2/Total Base 2(0.336)/1.45 = 0.463
XMgO = NMgO/Total Base 0.143/1.45 = 0.099
XAl2O3 = 3NAl2O3/Total Base 3(0.057)/1.45 = 0.118

Optical Basicity = (XCaO*ΛCaO) + (XSiO2*ΛSiO2) + (XMgO*ΛMgO) + (XAl2O3*ΛAl2O3)

= (0.320 * 1) + (0.463 * 0.48) + (0.099 * 0.78) + (0.118 * 0.61)
= 0.691

2.4 Optical Basicity correlation with sulfide capacity

The relationship between sulfide capacity and optical basicity (Λ) is given by:

[22690 − (54640 * Λ )]
log CS = + [(43.6 * Λ) − 25.2] (6)
T

Figure 3 shows the correlation between sulfide capacity and calculated optical basicity of slags at
1500°C (2732°F).

Figure 3. Correlation between sulfide capacity and calculated

optical basicity of slags at 1500°C.

LWB Refractories 6 Process Technology Group

In an effort to improve on the accuracy of the correlations with optical basicity in specific slag
composition regimes, various adjustments or corrections have been proposed. For example to
improve accuracy with CaF2-containing slags a modified optical basicity has also been proposed,
which ignores the effect of CaF2 on the optical basicity, but includes a correction term in the
calculation of log Cs. Alternative correlations not based on optical basicity are also available to
relate CS to the slag composition, but the applicability of these relations is probably limited to
slag compositions close to the original compositions used to derive these correlations. In
summary, the optical basicity correlation is easy to use and a good estimate over a wide range of
slag compositions, but alternative sulfide capacity data can also be used in their applicable range
of compositions.

2.5 Importance of liquid slag

One of the most important aspects that is frequently overlooked when evaluating and discussing
slags, is that most slags consist of two fractions, i.e., a liquid fraction and a solid fraction. Slag
fluidity in basic slags is controlled by the liquid and solid fractions of the slag. The higher the
solid fraction of the slag - the lower the fluidity of the slag (higher viscosity). A basic slag that is
completely liquid has maximum fluidity. Only the liquid portion of the slag is active in
desulfurization, a high solid content effectively reducing the slag volume.

The following example illustrates the importance of fluidity and effective slag volume.

Consider the following two slags at 1600°C (2912°F):

Slag O (Figure 4) Slag K (Figure 4)

% SiO2 38 35
% CaO 45 50
% MgO 17 15
Phases present 100% liquid 56% liquid, 39% Ca2SiO4,
(Optimum slag) 5% MgO

In many calculations of sulfur removal, the total composition of the slag is sometimes mistakenly
considered. It is the lime dissolved in the liquid fraction of the slag that is removing sulfur from
the steel and not lime as determined by the chemical analysis of the "total" slag. The
undissolved lime in the slag does not remove any sulfur. The liquid fraction of slag K has the
identical composition as slag O. The addition of more lime than required (slag O to slag K), only
resulted in a decrease in the amount of liquid available for desulfurization (liquid composition
stayed constant) and an increase in the viscosity of the slag.

LWB Refractories 7 Process Technology Group

 SiO2
1600°C (2912°F) Isothermal Section
10 90

20 80
Sea area (all liquid)
S+L
30 70 Land area (all solid)

Swamp area (liquid + solid)

40 60

50 50
M2S + L
Mg2SiO4
60 O 40
C2S + L
Ca2SiO4 M+L M2S + L + M
K
70 C2S + L + M 30
Ca3SiO5
80 20

90 10

10 20 30 40 50 60 70 80 90
CaO MgO
Figure 4. Isothermal section of the CaO-MgO-SiO2 system at 1600°C

The errors that can be made by not considering only the liquid portion of a slag is illustrated by
the following example:

Table 2. Example of possible errors in the calculation of final metal sulfur content
Considering the Considering the Considering only
total slag chemistry total slag the liquid fraction
(Slag O) chemistry (Slag K) of slag K
Temperature (°C) 1600 1600 1600
Si (wt%) 0.035 0.035 0.035
WM (tons) 100 100 100
WS (lb.) 2000 2000 (2000*0.56) = 1120
Λ 0.698 0.717 0.698
[% O] (ppm) 15 15 15
log CS -3.01 -2.74 -3.01
LS 17.83 32.84 17.83
SF (wt%) 0.0297 0.0263 0.0318
Calculated result Correct Wrong!! Correct

LWB Refractories 8 Process Technology Group

To summarize: the sulfide capacity can be defined as the ability of a liquid slag to absorb sulfur,
determined experimentally as the equilibrium sulfur content for certain conditions (temperature,
pS2 and pO2). The sulfide capacity is a function of temperature and slag composition and can be
calculated according to correlations based on the concepts of optical basicity.

3 Sulfide capacity in slag-metal systems.

3.1 Log Ls

The concept of sulfide capacity was explained and experimentally determined according to the
gas-slag reaction (2), but we would now like to apply it to the slag-metal reaction (1). While in
the gas-slag system the partial pressure of O2 and S2 are important, in the slag-metal system it is
the O and S in solution in the metal that determines the equilibrium sulfur content of the slag.

[S] + (O 2− ) ⇔ (S 2− ) + [O] (1)

1 S + O 2− = S 2− + 1 O (2)
2 2 2 2

By subtracting reaction (2) from reaction (1), we obtain reaction (7) for which the equilibrium
constant is calculated as in (8).

[S] + 1 2 O 2 = 1 2 S 2 + [O] (7)

log K(5) = -935 / T + 1.375 (8)

The derivation is not shown in detail here, but involves a substitution of the expression for
sulfide capacity (3) into the equilibrium constant for (7) and converting to the logarithmic form,
to arrive at equation (9). This equation relates the sulfide capacity and other known
thermodynamic properties of the metal, to the one parameter of prime importance for further
evaluation – the sulfur distribution (Ls).

(%S) − 935
log Ls = log = + 1.375 + log Cs + log f S − log a O (9)
[%S] T

f
In this equation, S is the activity coefficient for sulfur in the metal according to the equation for
a f a
the activity of S in dilute solution: S = S * [%S] and similarly for the activity of oxygen: O =
f f
O * [%O]. The activity coefficient of S in metal S can be calculated based on the metal
composition and interaction parameters. For common carbon steel grades, the activity coefficient
f
is close to unity (log S ≈ 0) and this assumption thus has little effect on the calculations. The
activity and activity coefficient of oxygen are more variable and need to be considered more
carefully.

LWB Refractories 9 Process Technology Group

If the sulfide capacity (Cs) for a particular slag is known, then the sulfur distribution ratio (Ls)
f a
can be calculated for a particular slag-metal system in which the values of S and O are known.
For best sulfur removal a large value of Ls is required, which can be obtained by high
a f
temperature, large Cs and low oxygen activity O. The activity coefficient of sulfur S is fixed by
the metal composition and not really a controlled variable. Note that the Ls ratio is not dependant
on the relative slag and metal mass or initial sulfur content.

a
3.2 Log O
The activity of oxygen can be measured or calculated and has an important impact on the
magnitude of Ls. Generally for deoxidized metal in the ladle, the higher the concentration of
deoxidizer in the metal, the lower the oxygen activity. The oxygen activity is also typically lower
for Al-killed, than for Si-killed steel. The lower the activity of the corresponding oxide (in slag
or in metal as inclusions), the lower the oxygen activity. According to (9), a lower oxygen
activity in the metal will result in higher Ls and thus improved sulfur removal. The oxygen
activity for the desulfurisation calculation, can be determined in one of two ways:
• Actual measurement of oxygen activity and temperature in the ladle by an oxygen probe
produces an electro-motive-force or emf value (measured as a voltage), relative to a reference
cell of known oxygen potential. The emf is usually re-calculated or converted by the
measuring unit into an oxygen content in parts per million (ppm) and sometimes carbon
content. (For the oxygen content: [%O] * 10000 = ppm[O]).
• Calculation based on slag-metal reaction equilibrium of the primary deoxidizer specie
(usually Al or Si). The difficult part of this calculation is to estimate the activity of the oxide-
specie in the slag (Al2O3 or SiO2), which can be done by correlation with experimentally
determined activities in similar slag systems or slag models. With knowledge of the
temperature, activity coefficient (and concentration) of the deoxidizer specie (Al or Si in
a
metal) and activity of the oxide, the oxygen activity O can be calculated for use in equation
(9).
•
Note that the oxygen activity of the bulk metal is measured by the probe, which can differ from
the conditions at the slag-metal interface, which is calculated by considering slag-metal
equilibrium. When the oxygen content is determined, it should be combined with the activity
f
coefficient O (which can be calculated based on the metal chemistry of a sample) to obtain the
a
activity O. In most low carbon steel compositions, the effect of Al, Si and C in the metal are
f
most important to determine the value of the activity coefficient of oxygen ( O).

The accuracy in measuring/calculating the oxygen activity that actually controls/determines the
sulfur distribution ratio is critical – especially at very low oxygen activity or oxygen content in
the steel. The calculation of Ls is very sensitive to oxygen activity at low oxygen activity, so that
small deviations have a major impact on the resulting Ls as shown in Figure 5. A slight
difference between the true oxygen activity determining Ls and the estimated or calculated value,
can result in large deviations between the calculated and true S-behavior. For the curves in
f a
Figure 5, the activity coefficient of oxygen was assumed to be one O = 1, so that O = [%O] =
ppm[O]/10000. At very low oxygen activity there is also a possibility that the conditions at the

LWB Refractories 10 Process Technology Group

slag-metal interface differ considerably from the conditions in the bulk of the metal – both due to
concentration gradients and oxide activity. The activity of Al2O3 or SiO2 are lower at the
interface due to solution into the slag, whereas in the bulk of the metal, the pure oxides with
activity of unity could be present (as inclusions).

1000

900

800

700
Ls = (%S)/[%S]

600

500

400

300

200

100

0
0 5 10 15 20 25 30 35

Oxygen content - ppm O

logCS=-2 logCS=-1.8 logCS=-1.6

Figure 5: Sensitivity of Ls to oxygen content.

3.3 Effect of oxidized slag.

The effect of the amount of reducible oxides (FeO, MnO and Cr2O3) in ladle slags on the
desulfurizing efficiency is shown in Figure 6 and low reducible oxide content is thus preferable
for good desulfurization. For this reason the slag is deoxidized by addition of aluminum shred,
FeSi or CaC2 when desulfurization is critical. The definition of the %desulfurization in this
widely quoted Figure is not quite clear, but it is reasonable to assume that the reducible oxides
act as a source of oxygen at the slag metal interface, which will decrease the effective Ls ratio to
a lower value than calculated by considering the bulk oxygen content. A simple correction of the
Ls ratio for oxidized species in the slag can thus be made in the calculation.

LWB Refractories 11 Process Technology Group

 100

% Des u lf u rizat io n 80

60

40

20

1 2 3 4
(FeO + M n O ), %

Figure 6: Effect of (FeO & MnO) in slag on desulfurization.

3.4 Equilibrium Sulfur content.

In addition to the Ls-ratio, the slag and metal weight (Wm and Wslag) and initial sulfur
concentration in the metal (%S°) do affect the final equilibrium sulfur content (%Seq) according
to:

[%S]° * Wm
[%Seq ] = (10)
Wm + Ls * Wslag

Only initial sulfur content in the metal is considered here and if the slag contains significant
sulfur at the start of treatment it should be included, but in most cases it has very little influence
on the final sulfur content. (The top half of the equation then becomes: {[%S]°*Wm +
(%S)°*Wslag}) According to equation (10), to achieve lower final sulfur content in the metal, the
slag volume (Wslag or slag to metal ratio) or Ls can be increased, or alternatively the initial sulfur
content in the metal decreased. As shown earlier, the value of Ls is affected mainly by the slag
composition (sulfide capacity), temperature and oxygen content of the metal.

3.5 Slag design

Published phase diagram and liquidus slag models can be used to design slags for specific
metallurgical requirements. The choice of fluidizers (SiO2, Al2O3, and/or CaF2) in the design of
these slags is very important because the solubility of CaO and MgO is strongly dependent on
the type of fluidizer used. As seen in the optical basicity calculation, CaO has the highest optical
basicity of the common ladle slag components and will have the most impact on the optical
basicity and thus sulfide capacity. In order of decreasing basicity the order is CaO, MgO, Al2O3
and SiO2. It follows that replacing Al2O3 with SiO2 as fluidizer will lower the sulfide capacity,

LWB Refractories 12 Process Technology Group

both because the optical basicity of SiO2 is lower than Al2O3 and because lime solubility
decreases. In practice we cannot use any slag composition (or temperature), but are restricted by
slag fluidity and compatibility with the ladle refractories.

The impact of the choice of slag composition or main fluidizers is clearly illustrated by the slags
in Table 3. All of the slags have a high fluidity and are just saturated with respect to CaO and
MgO ("creamy") and are therefore compatible with magnesia and dolomite refractories.
Different slag composition choices, result in different desulfurisation potential and equilibrium
sulfur content for the same conditions of oxygen content of the metal and slag mass. Using
available phase diagrams or liquidus models, various combinations of fluidizing agents can be
used so that an infinite number of slags with varying sulfide capacities can be designed to attain
specific metallurgical goals.

Table 3. Compositions of slags that are just CaO or CaO and MgO saturated at 1600°C
(Λ=optical basicity, Cs = sulfide capacity, and Sf = final sulfur)

Slag 1 Slag 2 Slag 3 Slag 4

% CaO 45 52 53 57
% MgO 17 10 13 8
% Al2O3 17 23
% SiO2 38 21 23 9
% CaF2 11 3
B5 Ratio 1.63 1.63 1.94 1.86
Λ 0.698 0.742 0.758 0.785
-Log Cs 3.01 2.38 2.14 1.76
Sf* 0.0424 0.0285 0.0217 0.0121
*
Using 100 tonnes of steel, 1000 kg of slag, Si = 0.05%, O = 15 ppm

Slags can also be “designed” for re-sulfurized steel grades, that is not to remove much sulfur –
e.g. Slag 1 in Table 3. This usually requires substitution of MgO for CaO (and SiO2 for Al2O3) in
order to generate slags that are still fluid but with a decreased thermodynamic sulfide capacity,
while maintaining basicity for basic refractory compatibility. These slags will have lower
CaO/SiO2 ratios that will increase the solubility of MgO in the slag. All these slags are 100%
fluid at steelmaking temperatures and compatible with magnesia slag line refractories.

We have so far only considered the thermodynamic factors that affect the potential for sulfur
removal – i.e. the best case scenario of what can be achieved for the conditions after sufficient or
infinite time. The kinetics or rate of sulfur removal are, however, also important and will be
considered next.

LWB Refractories 13 Process Technology Group

4 Kinetics of sulfur removal in the ladle.

4.1 Rate of desulfurization

While thermodynamic describe the driving force and extent or limits of sulfur transfer from
metal to slag, the actual change achieved in the allotted treatment time depends on a number of
kinetic factors. It has been established that mass transfer of sulfur in the metal to the slag-metal
interface, limits the overall kinetics of sulfur transfer to the slag. This means that the rate at
which sulfur is transferred from the bulk metal to the slag-metal interface and transfer across the
boundary layer close to the interface, are the slowest and thus rate limiting step – which is
mainly affected by stirring in the metal. This implies that the slag-metal reaction is relatively
rapid and mass transfer in the slag is not limiting – i.e. a well-mixed liquid slag. The mixing of
the metal is thus the main limiting parameter, as controlled by the inert gas stirring through
porous plugs or electro magnetic stirring.

The following equations describe the kinetic theory of sulfur removal and equation (11)
describes the rate of change in sulfur content as a function of the sulfur content at time t (%S),
the initial sulfur content (%S°) and the equilibrium sulfur content (%Seq). The driving force for
sulfur transfer is the difference between the current sulfur content at time t and the equilibrium
sulfur content or (%S-%Seq) and k is the overall kinetic constant.

d%S
= −k (%S − %S eq ) (11)
dt

Integration of equation (11) between time 0 and t results in the following:

(%S − %Seq )
ln = −kt (12)
(%S° − %S eq )

It should be clear that the kinetic constant determines how rapidly the sulfur content in the metal
decreases or sulfur is transferred from the metal to the slag. The larger the kinetic constant, the
quicker will the sulfur removal be. As the rate is controlled by mass transfer in the metal, the
most important parameter determining the value of k, would be the stirring intensity in the metal.
The magnitude of the kinetic constant can be determined from actual plant conditions by
sampling and analysis of metal sulfur content over time intervals. It can also be estimated by
published correlations, which relate the kinetic constant to the intensity of stirring as a result of a
particular gas flow rate.

4.2 Correlations of kinetic constant (k) and gas flow rate.

Two different correlations between gas flow rate and stirring intensity were found in literature.
Both are based on the concept of stirring energy due to gas stirring, incorporating the
displacement and drag effects of the rising bubbles. The stirring energy is then correlated by
empirical measurements to first the mass transfer coefficient and then the kinetic constant or
directly to the kinetic constant. While the correlations might be only strictly valid for the
particular conditions under which the measurements were done, it is nevertheless a good estimate

LWB Refractories 14 Process Technology Group

of the expected rate of sulfur removal. Specific conditions which can impact the applicability
are:
• Type and position of porous plug in ladle.
• Ladle dimensions and slag/metal weight.
• Slag-metal emulsification behavior at the interface (slag fluidity).
Extrapolation of these ‘empirical’ relations to other plants can only be viewed as a rough
estimation and should be checked with detailed sampling and correlation. Extrapolation of these
correlation for bottom stirring through purge plugs, to stirring by top-lance should be done with
care.

The stirring energy according to Sundberg can be calculated by:

 298 P 
E [ W ] = 6.183 * V * T * 1 − + ln bottom  (13)
 T  Ptop 
  

and the mass transfer coefficient (ms) and overall kinetic constant (k) [Ishii et.al.]:

m S [m / s] = 8.33 *10 −4 (E / Area) (14)

m S * ρ * Area  Wm 
k[s −1 ] = 1+  (15)
Wm  Wslag * Ls 
 

Alternatively the relations by Plushkell as used by Turkdogan can be used:

6.183 * V * T  Pbottom 

ε [ W / ton ] = ln (16)
Wm  Ptop 
 
k [min −1 ] = 0.013 * (ε) 0.25 forε < 60W / ton (17)

k [min −1 ] = 8 *10 −6 * (ε) 2.1 forε > 60 W / ton (18)

In the above, symbols have the following meaning:

V = flow rate of stirring gas [Nm3/min]
T = temperature [K]
ρ = density of liquid steel at T [kg/m3]
Area = surface area of slag-metal interface [m2]
Ptop = pressure above ladle [atm]
Pbottom = Pressure at bottom of ladle = Ptop + ρgh [atm]

In general the correlations show that increased rate of desulfurization can be achieved by:
• Higher temperature (usually increases kinetics)
• Higher gas stirring rate (V)
• Decreasing the pressure above the bath by vacuum (Ptop)

LWB Refractories 15 Process Technology Group

The other parameters are related by the size of the ladle, for example increasing the planar
surface area by increasing the ladle diameter, would decrease the ladle depth and consequently
Pbottom also, for constant metal weight. The effect of slag mass (Wslag) in equation (15) might be
contrary to ‘gut feeling’ at first glance, as higher slag mass seems to hinder desulfurisation by
decreased rate (k). However, the slag mass also affects the equilibrium %S, so that the result is
that it would take longer to reach the equilibrium sulfur content (lower k), but the equilibrium
sulfur content is also much lower (according to (10)).

In the second correlation, the kinetic constant increases much more drastically with stirring
intensity above a critical value of 60 W/ton - equation (18), than below this critical intensity as in
equation (17). This is due to emulsification of the metal-slag interface at higher flow rates
(intensity of stirring), which increases the mass transfer at the boundary layer and the surface
area significantly. At lower flow rates the metal-slag interface is relatively stagnant and mass
transfer is slower. This is another reason why a fluid slag (creamy, but no major solid content) is
important as the emulsification is important to achieve high rates of desulfurisation demanded by
current production schedules.

Another interesting aspect of the equations, is that decreasing the pressure above the vessel (e.g.
in degasser) increases the stirring intensity for identical dimensions and gas flow rates – thus
increases mass transfer and the rate of desulfurisation. However, it does not affect the sulfur
distribution ratio or equilibrium sulfur content, just the rate at which it is achieved.

For electro-magnetic stirring (EMS) emulsification of the slag-metal interface does not occur as
readily as for gas stirring (physical effect of bubble is important) and consequently sulfur transfer
is expected to be slower. The controlled variable for EMS would be the induction current (as gas
flow rate is for gas stirring), but no correlation for sulfur removal could be found in literature.
Data for inclusion removal for different EMS-current setting is available, but would be plant
specific and does not allow correlation with stirring intensity (W/t) and desulfurization.

The position of the purge plugs in the ladle can have a significant effect on the sulfur removal
kinetics and thus accuracy of these correlations. Intermixing or formation of a slag-metal
emulsion at the interface is promoted by the stir plug in the center of the ladle – this provides
increased surface area and less boundary layer resistance for the desulfurisation reaction to
occur. However, a central position also promotes the formation of dead zones in the lower part of
the ladle and metal mixing is not efficient. Mixing efficiency is improved (mixing/homogenizing
time reduced) by moving the purge plug away from the center of the ladle, also enabling the
opening of an eye for alloy additions. The closer the plug is to the refractory lining however, the
more slag line erosion is expected. Modeling work and practical experience suggests that the best
position for a compromise between the above effects be with the stir plug on one third of the
ladle diameter. (Gas stirring should also not be directly underneath an electrode position).

The kinetic equations have been incorporated into the Baker Kinetic de-S model for estimation
of desulfurization rates and correlation with plant data.

LWB Refractories 16 Process Technology Group

4.3 Change of %S with time.
With knowledge of the overall kinetic constant, the sulfur content variation with time toward
equilibrium can be determined according to equation (19) derived from equation (12).

%S = %Seq + (%S° − %Seq )e − kt (19)

The typical form of the exponential function in equation (19) is shown graphically in Figure 7 for
different k-values, representing different stirring gas flow rates. The other parameters, the initial
and equilibrium sulfur content are the same for both graphs.

0.03

0.025

0.02
%S in metal

0.015

0.01

0.005

0
0 5 10 15 20 25 30 35 40 45

Time (minutes)

k=0.05 k=0.10

Figure 7: Example of %S versus time for different k-values.

Kinetic evaluation is useful to determine the how the important parameters (LS, slag volume,
stirring rate, etc.) affect the time required to lower the sulfur content of the metal as tapped, to a
certain aim value (specification). The time available for ladle treatment is limited and different
scenarios with different the parameters can be evaluated theoretically with this calculation,
before testing it in the plant.

The sulfur change with time can also be evaluated for changed conditions – i.e. vigorous/hard
stirring for 10 minutes, followed by alloy trimming and ‘softer’ rinsing stir for 15 minutes for
inclusion flotation. The different conditions for each period can be determined and the expected
sulfur change with time calculated for the two successive periods. This would be useful to
determine the minimum required “hard stir time” to reach the specified %S after soft rinsing. The

LWB Refractories 17 Process Technology Group

conditions to limit sulfur removal (i.e. re-sulfurized grades), can also be evaluated with these
relations.

It is important to realize that the desulfurisation reactions continue to occur after treatment at the
ladle furnace, even if at a lower rate due to reduced stirring. For a long delay before casting, the
sulfur content can still reduce significantly after ladle treatment – can be critical for close spec or
lower limit sulfur grades. This of course depends on how far from equilibrium the sulfur
distribution is – i.e. the remaining driving force.

4.4 Additional comments on kinetics:

• The conditions in the ladle will change with time (temperature, slag composition, activity of
oxygen) and will affect the driving force or Ls value and thus also the kinetics. If the changes
are minor, average or constant values can be used for the calculations.
a
• The difference between the O at the slag-metal interface and in bulk of the metal (where it is
usually measured) can significantly affect sulfur removal.
• Excessive stirring and opening of a large “bubble eye” can lead to significant re-oxidation –
a
affecting both the O at the interface and slag composition (also by inclusion flotation).
• The effect of reducible oxides (FeO & MnO) is difficult to accurately include in the
calculations and is expected to lower the sulfur distribution ratio by way of its effect on the
interfacial oxygen activity. Significant reducible oxide content will result in fading of the
deoxidizer and resulting increase in the bulk oxygen activity.
• Although not common, it is possible at high total sulfur in the system (sulfur load) and high
Ls, that the sulfur content of the slag exceeds the maximum solubility and solid CaS is
formed. It is believed that this will lead to a lowering of the effective Ls, due to lowering of
the sulfide capacity (Cs) as the lime content of the liquid slag is reduced. It is not clear
whether experimentally determined values for Cs are valid for conditions that would exceed
S-solubility in the slag. This can also be a localized effect if poor mixing conditions exist in
the slag phase.
• Initial sulfur content in EAF transfer slag and in ladle skulls or remaining slag can be
important, as they contribute to the total sulfur content of the system (sulfur load) –
especially if both low S and re-S grades are produced in the same shop (ladles).

5 Sulfur behavior in the EAF

5.1 EAF conditions and Cs.

The basic theory for sulfur removal (Ls as in equation (9)) indicates that reducing conditions
a
(low O) and highly basic slags are best to transfer sulfur from metal to slag. These are the
conditions during secondary metallurgical treatment in the ladle. This implies that sulfur removal
in the EAF at oxidizing conditions and less basic slags will be less effective. The advantage in
the EAF is the relatively large slag volume (in the region of 200 lb per ton of metal for EAF, vs.
20 lb per ton of metal in the ladle). Due to the large slag volume, the sulfur removed can still be
significant, even at low LS. The ideal slag composition in the EAF is not dictated by
desulfurisation, but by more important process aspects like foaming. Sulfur removal in the EAF

LWB Refractories 18 Process Technology Group

is limited by other process requirements and the best method of controlling sulfur will still be
selection of input materials and good desulfurisation in the ladle.

Not much experimental data on sulfide capacity (Cs) of EAF type slags is available, as most
investigations have concentrated on ladle slags. Generally the log Cs of an EAF slag will be
much lower than that of a synthetic ladle slag due to lower CaO content and significant SiO2 and
MgO content. The classic relations of logCs to optical basicity are not necessarily valid, as the
correlations were developed for ladle slags. There is also uncertainty regarding the optical
basicity values to be used for FeO and MnO, which are present in much large quantities than in
ladle slags and can significantly affect the optical basicity (e.g. values of either 0.51 or 1.0 for
FeO can be used). Recent research results have indicated that sulfur capacity increases with
increasing FeO or MnO content, but this also depends on which specie is replaced in the slag
(e.g. CaO versus SiO2). However, the oxygen activity of the system will also increase with
increasing FeO content and the change in Ls as a result of the opposing changes is not quite
clear.

5.2 Log Ls
Turkdogan showed a correlation of sulfur data in terms of distribution ratio Ls for oxidizing
processes, without consideration of sulfide capacity. The graphical correlation shows that the
parameter kso = Ls*%O or ks = Ls*%FeO decreases with increasing (%SiO2 + 0.84*%P2O5).
(For some of the data Al2O3 is included in the SiO2 term, with a coefficient of 1.18). The %FeO
term is used as a measure of the oxygen activity in the EAF system. This agrees with the general
theory that more basic slags are better desulfurizers or acidic slags are worse, however, the FeO
content is also important. By application of these relations to plant data from a particular EAF-
shop, Turkdogan showed that similar to the relations in the BOF, the conditions just before tap
are still below equilibrium. A curve was fitted through Turkdogan’s plots of ks and the equations
are:

For equilibrium: ks = 481.2 exp (-0.0939 * (%SiO2 +0.84*%P2O5)) (20)

For typical EAF: ks’ = 434.3 exp (-0.1077 * (%SiO2 + 0.84* %P2O5)) (21)

As for the case in ladles, the practical slag composition is however, restricted by other process
issues and in the EAF slag foaming is of major importance, as is refractory compatibility. By
limiting the slag compositions to those known to be ideal for foaming properties in the EAF, the
sulfur distribution is shown to vary between 1 and 3. For comparison, the slag compositions
close to dual saturation were chosen and the ks and Ls values calculated and the results are
shown in Table 4. Even though the ks value increases with increasing basicity, the FeO content
also has to increase to maintain a liquid slag. More FeO is required to flux the slag, when the
SiO2 content reduces as a result of the increase in basicity. These effects almost cancel out and
the resulting Ls values do not increase much as shown in the table. The values between 1 and 3
are low compared to typical values of 50 to over 500 for ladle slag systems.

LWB Refractories 19 Process Technology Group

 Table 4: Variation of Ls for dual saturation in EAF slags (* ’SiO2’ = %SiO2 + %Al2O3)
Basicity %FeO %MgO %CaO %'SiO2'* EAF-ks LS
1.5 15 13 43.2 28.8 20 1.3
1.65 20 11 43.0 26.0 26 1.3
2 26 9 43.3 21.7 42 1.6
2.5 31 8 43.6 17.4 66 2.1
3 34 7 44.3 14.8 89 2.6
3 45 6 36.8 12.3 116 2.6

The last two rows show a high basicity, defined as CaO/[SiO2 +Al2O3], with low and high FeO
content. The correlation (21) shows an increase in ks for the lower SiO2 content, but the higher
FeO content off-sets the change and the Ls does not change. The higher FeO content results in
increase yield losses and probably departure from the ideal foamy slag composition.

Therefore, in the practical range for EAF foaming slag compositions, the Ls will be between 1
and 3, with a value of 5 being a possible extreme case. The fraction of sulfur remaining for
different Ls-values is indicated as a function of slag volume in Figure 8. At a slag rate of 0.15,
which relates to 150 kg slag per tonne (1000 kg) of metal or alternatively 300 lb of slag per ton
(2000lb) of metal, the fraction of sulfur remaining at equilibrium would be 0.7 for an Ls=3. This
means that 30% of the sulfur was removed from the metal, which is still significant.

Due to the constantly changing conditions in the EAF, equilibrium is rarely achieved and the
application of theoretical calculations is less accurate, than in the ladle. Sulfur distributions
determined in the plant, based on actual metal and slag samples can deviate to either side of the
equilibrium calculations, due to kinetic limitations and late addition of sulfur via scrap melting or
carbon injected.

LWB Refractories 20 Process Technology Group

 1.0

0.9
Fraction S remaining

0.8

0.7

0.6

0.5

0.4
0 0.05 0.1 0.15 0.2 0.25

Slag rate (ton slag per ton metal)

LS=1 LS=3 LS=5

Figure 8: Influence of Ls and slag rate on sulfur remaining/removed.

5.3 Additional comments - EAF:

• Consider the sulfur content of the slag remaining in the furnace after tapping for sulfur
balance of the following heat.
• Most carbon sources for charging and injection (especially coal) contain sulfur, which can be
a significant contributing factor.
• The conditions in the EAF change drastically over time compared to the ladle furnace
(temperature, oxygen activity and slag composition), resulting in a “constant change” of the
thermodynamic driving force for sulfur removal.
• The EAF (or other primary steel making process) is principally oxidizing and thus best suited
to the removal of Si, Mn, C and phosphorous from the melt. The sulfur removed is important,
but changing conditions to maximize sulfur removal, will negatively impact other process
requirements.

6 References
"Ladle Metallurgy Principles and Practices", Fruehan RJ, ISS, Bookcrafters Inc. 1985
"Physicochemical properties of molten slags and glasses", Turkdogan ET, Metals Society
London. 1983.
“Fundamentals of Steelmaking”, Turkdogan ET, The Institute of Materials, London, 1996.

LWB Refractories 21 Process Technology Group

“The Making, Shaping and Treating of Steel”, 11th edition, Steelmaking and Refining Volume,
Fruehan RJ (Editor), 1998, AISE Steel Foundation, Pittsburgh PA.
“Slag Atlas”, Verein Deutscher Eisenhüttenleute, prepared by the Committee for Fundamental
Metallurgy. 1981

LWB Refractories 22 Process Technology Group