ISIJ International, Vol. 59 (2019), No. 5

© 2019 ISIJ839

ISIJ International, Vol. 59 (2019), No. 5, pp. 839–847

* Corresponding author: E-mail: pbd150@uowmail.edu.auDOI: https://doi.org/10.2355/isijinternational.ISIJINT-2018-683

1. Introduction

Phosphorus generally has a negative effect on the mechanical properties of steel.1–3) Low phosphorus iron ores used in the production of steel come at a cost premium. This cost premium is a significant driver for dephosphorisation research. The ability to utilise low cost, low-quality ores high in phosphorus and other impurities such as titanium oxides potentially offers advantages for integrated steelmak-ers. However, an improved understanding of the effect of impurities such as TiO2 on dephosphorisation is required if steelmakers are to utilise these ores economically.

The dephosphorisation reaction in basic oxygen steelmak-ing (BOS) is commonly represented by either the ionic reac-tion in Eq. (1) or the molecular reaction in Eq. (2)

[ ] [ ] ( ) ( )P O O PO� � �� �5

2

3

22

43 .................. (1)

2 5 2 5[ ] [ ] ( )P O P O� � ......................... (2)

where ( ) denotes solution in slag and [ ] denotes solution in liquid iron. Unless otherwise stated the reference state for

Phosphorus Partition and Phosphate Capacity of TiO2 Bearing Basic Oxygen Steelmaking Slags

Phillip Brian DRAIN,1)* Brian Joseph MONAGHAN,1) Raymond James LONGBOTTOM,1) Michael Wallace CHAPMAN,2,3) Guangqing ZHANG1) and Sheng Jason CHEW2,3)

1) ARC Research Hub for Australian Steel Manufacturing & School of Mechanical, Materials, Mechatronic and Biomedical Engineering, University of Wollongong, Wollongong, NSW, 2522 Australia.2) BlueScope, Ironmaking Technology, PO Box 1854, Wollongong, NSW, 2505 Australia.3) ARC Research Hub for Australian Steel Manufacturing, University of Wollongong, Wollongong, NSW, 2522 Australia.

(Received on October 12, 2018; accepted on December 4, 2018; J-STAGE Advance published date: February 9, 2019)

The phosphorus partition (LP) and phosphate capacity (CPO43− ) were measured for TiO2 bearing basic oxy-

gen steelmaking slags in the CaO–SiO2–MgO–FetO–(TiO2–MnO–Al2O3–P2O5) system at 1 650°C. The effect of slag and metal additions were tested by varying the TiO2 content from 0.0 to 18.0 mass% and the [Ti] content from 0.009 to 0.301 mass%. A recently published LP model was used to assess the experimental LP data using the measured slag composition and temperature. Experimental LP data from this study and literature data were used to modify the published model to include titania.

Increasing the TiO2 concentration of the slag was found to decrease the LP and CPO43− of basic oxygen

steelmaking (BOS) slags. Capacity values in the range of 2.2×1016 to 1.5×1018 at 1 650°C were obtained. An empirical model for determining CPO4

3− was developed for BOS slags using a large dataset of published slag and pO2 data for relevant slag systems, including published data for titania bearing slags. The pre-dicted CPO4

3− from the empirical model was found to agree with the experimentally determined CPO43− data

from this study.

KEY WORDS: basic oxygen steelmaking (BOS); dephosphorisation; phosphorus partition; phosphate capacity; slags; TiO2.

all metal solutes is the 1 mass% standard state. For (P2O5) the hypothetical pure molten phase is the reference state.

The standard Gibbs free energy (ΔG°) for Eq. (2) is given in Eq. (3)4)

�G T J mol� � � �– . ·( )832 384 632 65 1 ............ (3)

where T is the temperature in K.From consideration of the reactions given in Eqs. (1)–(2)

and ΔG° in Eq. (3),4) it can be shown that the dephosphorisa-tion reaction is favoured by high [P], high oxygen potential (represented by [O] in Eq. (2)), high basicity (O2−) and low temperature (T). The [O] can be related to the oxygen partial pressure (pO2) by consideration of the reaction given in Eq. (4) and its respective ΔG° in Eq. (5).4)

1

22O Og( ) [ ]= ............................... (4)

�G T J mol� � � � � �115 750 4 63 1. ( ) ............. (5)

Dephosphorisation in the integrated steelmaking process is typically carried out in either hot metal pre-treatment or the BOS converter. However, there is no ideal location where all the conditions for dephosphorisation are optimised e.g. hot metal pre-treatment has a characteristic low [O] and the BOS vessel has a high T.

ISIJ International, Vol. 59 (2019), No. 5

© 2019 ISIJ 840

The oxygen potential for dephosphorisation may be rep-resented by a reaction or balance of competing reactions that control delivery of oxygen in Eqs. (1) and (2). In the BOS process, the oxygen potential for dephosphorisation is generally considered to be controlled by the balance between decarburisation and FeO reactions, given in Eqs. (6) and (7). This oxygen potential changes throughout the steelmaking process.

Fe O FeOl g( ) ( ) ( )� �1

22 ........................ (6)

[ ] ( ) ( )C O COg g� �1

22 ......................... (7)

For dephosphorisation of liquid iron containing [Ti], the oxidation reactions of [Ti], given in Eqs. (8)5) and (10),5) may also affect the dynamic oxygen balance for dephos-phorisation.

[ ] [ ] ( )Ti O TiO� �2 2 .......................... (8)

�G T J mol� � � � � �646 500 224 555 1. ( ) .......... (9)

[ ] / [ ] / ( )Ti O Ti O� �3 2 1 2 2 3 .................. (10)

�G T J mol� � � � � �1 023 000 342 581 1. ( ) ........ (11)

In addition to affecting the oxygen potential, the titanium oxide formed will also change the slag basicity. Slag basic-ity is used to imply the degree of polymerisation of the slag structure. The formation of both Ti4+ (i.e. TiO2) and Ti3+ (i.e. Ti2O3) has been observed in steelmaking slags.6) The formation of Ti3+ may be significant as it exhibits basic behaviour,7,8) while the effect of Ti4+ is less clear as it has been reported to behave as an amphoteric oxide.7,8) Stud-ies in CaO–SiO2–TiOx slag have found the Ti3+/Ti4+ ratio increases with decreasing pO2 and basicity, and increasing temperature.9,10) In BOS slags, the titanium oxide may be expected to be predominately in the form of TiO2 due to the high oxygen partial pressure and basicity. The exchange between Ti3+ and Ti4+ in slags can be represented by the reaction given in Eq. (12)7) and evaluated from its ΔG° given in Eq. (13).7)

TiO Ti O Os s g2 2 3 21 21

4( ) ( ) ( )/� � ............... (12)

�G T J mol� � � � �189 954 48 5 1. ( ) ............. (13)

The phosphorus partition (LP) and phosphate capac-ity (CPO4

3−),11) defined in Eqs. (14) and (15), are common approaches for assessing the dephosphorisation ability of a slag,

LP

PP =

(% )

[% ] ............................... (14)

C

PO

p pPO

P O43

2 2

43

1 2 5 4� �

�(% )/ /

.......................... (15)

where (%P) and [P] are the mass% of P, (% )PO43− is the

mass% of phosphate ion in the slag, and pi is the partial pressure of species i.

There are many published models that predict the LP from slag composition and temperature. These models often account for the oxygen potential using a proxy such as Fet or FetO,12,13) where Fet is the total iron oxide in slag repre-sented by elemental Fe in mass%. Use of the phosphorus gas solution reaction as given in Eq. (16), its associated ΔG° given in Eq. (17),4) and the equilibrium constant K determined via Eq. (18) allows the CPO4

3− to be obtained from the LP through Eq. (19)

1

22P Pg( ) [ ]= .............................. (16)

�G T J mol� � � �– . ·( )122 173 19 25 1 ........... (17)

KG

RTP[ ] exp�

� ����

���

� ........................ (18)

CL K M

p f MPO

P P PO

O P P43

43

2

5 4�

�

�[ ]

/ ....................... (19)

where R is the universal gas constant (8.314 J·K −1mol −1), Mi is the molecular weight of species i, fP is the activity coefficient of phosphorous in the steel, and K[P] is the equi-librium constant for the reaction given in Eq. (16).

The reported CPO43− of various slags have been reviewed

in Thermodynamic Data for Steelmaking14) and the Slag Atlas.15) Although the dephosphorisation literature is exten-sive, only three studies16–22) report data for titanium oxide containing steelmaking slags.

Two of these studies (Selin;16–20) Usui et al.22)) are of particular interest in understanding the effect of TiO2 on dephosphorisation. The third titanium study by Kor21) is of limited use as titanium oxide in the slag is reported as a combined total of TiO2 and Al2O3.

Usui et al.22) used slags with low titanium oxide (≤1.3 mass%, TiO2) concentrations whereas Selin16–20) focused on electric arc furnace (EAF) steelmaking slags with v-ratios (as defined in Eq. (20)) of 1.25–2.5 and at tempera-tures ~1 600°C. Though BOS slags typically have higher v-ratios (2 to 5) than EAF slags and tempertures as high as 1 700°C,16–20) Selin’s data are still of use in understaning BOS behaviour.

vCaO

SiO=(% )

(% )2 ............................. (20)

where (%CaO) and (%SiO2) are in mass%.In this equlibrium slag-steel study, the effect of TiO2 on

the dephosphorisation ability of synthetic slags in the CaO–SiO2–MgO–FetO–TiO2 system are evaluated at temperatures and compositions representative of BOS steelmaking.

2. Experimental

The slag-steel LP experiments were carried out in a verti-cal tube furnace under a controlled (99.999%) Ar atmo-sphere for 12 hours at 1 650°C. A schematic of the set-up used is shown in Fig. 1. In the experiments a mass of 4 g of slag and 30 g of ~0.18 mass% [P] alloy were melted in a high purity dense MgO crucible. The slag was held in a Fe foil cup (~0.16 g) during heating to keep the slag and

ISIJ International, Vol. 59 (2019), No. 5

© 2019 ISIJ841

[P] alloy separate until both were liquid. The high purity Ar gas was scrubbed prior to entering the furnace by passing it through ascarite, drierite and then copper turnings at 300°C.

At the end of the experiment, the crucible was rapidly cooled in the furnace to room temperature. The slag and steel were separated by hand in preparation for composition analysis. The metal was dissolved in nitric acid for analysis using an Agilent 710 series inductively coupled plasma opti-cal emission spectroscopy (ICP-OES) instrument. The slag was crushed and analysed using an AMETEK SPECTRO XEPOS energy dispersive polarisation x-ray fluorescence (XRF) spectrometer.

A further series of experiments were conducted with the pO2 controlled by the CO/CO2 ratio in the gas phase. These experiments were carried out to validate the CPO4

3− model developed in this study. The controlled gas pO2 experi-ments were conducted under an Ar–CO–CO2 atmosphere with a CO/CO2 ratio of 0.060 and 0.114. This gave a pO2 of 1.6×10 −9 atm to 5.6×10 −9 atm at the experimental temper-ature of 1 650°C. The Ar–CO–CO2 gas was dried by passing through a bed of drierite before entering the furnace. The CO/CO2 ratios were obtained by mixing calibration grade gas mix of 25%CO2-Ar and CO gas (99.2%CO, 0.35% N2, 0.07%Ar, 0.35% other hydrocarbons, O2>5 ppm).

2.1. Slag PreparationSynthetic slag pellets in the CaO–SiO2–MgO–FetO–

TiO2 system were made from high purity oxide powders (>99.9%) and pre-fused slag. The pre-fused slag was pre-pared by mixing dried (or calcined) oxide powders in the required amounts, followed by melting the mixed powder in a Pt crucible for ~20 minutes at 1 600°C using a muffle furnace. The liquid slag was then quenched and crushed to fine powder using a ring grinder. This process was repeated to ensure slag homogeneity. Slag pellets were prepared by mixing dried (or calcined) powders and pre-fused slag in the required amounts for pressing into a 4 g pellet. The pellets were then sintered for 12 hours at 1 000°C in a muffle fur-nace. XRF analyses of the sintered slag pellets and the pre-

fused slag are given in Table 1. Specific iron and titanium ions in the slag were not measured in this study and have been reported as FetO and TiO2 respectively.

2.2. Alloy PreparationA Fe–P master alloy and three [Ti] bearing alloys with

contents of 0.05, 0.14, 0.3 mass%, and [P] content of ~0.18 mass% were prepared for this study. The alloys were prepared using a ferrophosphorus alloy (25.7 mass% [P]), a low carbon, aluminium killed alloy (0.015 mass% [P]) and a high purity Ti rod (99.99%). The alloys were melted under an Ar atmosphere at 1 600°C in an alumina crucible. After solidification the alloys were cut and remelted again at 1 600°C to ensure uniform Ti and P distribution. The compositions of these alloys are given in Table 2. All elements except P and Ti were measured by electric arc optical emission spectroscopy (ARC-OES). P and Ti were measured via ICP-OES.

2.3. Equilibrium TimeIn the experiments carried out under Ar, the equilibrium

time was established by carrying out a series of runs for up to 12 h using slag B2 with the Fe–P master alloy. The changes in the slag and steel phosphorus level were deter-

Fig. 1. A schematic of the experimental furnace set up used.

Table 1. Synthetic slag compositions as measured by XRF.

Description MgO SiO2 CaO TiO2*) FetO*) v-ratio

Pre-fused slag 13.1 43.9 43.1 0.0 0.0 1.0

B2#) 9.7 21.5 45.5 0.0 23.3 2.1

R^) 9.9 17.6 49.4 0.0 23.1 2.8

TiO2 9.1 17.5 49.0 1.2 23.2 2.8

TiO2 9.2 15.8 46.8 5.4 22.8 3.0

TiO2 8.9 15.4 45.3 8.8 21.6 2.9

TiO2 8.8 15.7 44.1 9.9 21.5 2.8

TiO2 8.6 15.0 43.2 13.0 20.2 2.9

TiO2 7.1 14.0 39.8 20.6 18.4 2.8

TiO2 – FetO 10.1 16.0 37.4 9.2 27.4 2.3

TiO2 – FetO 9.8 17.9 38.9 10.9 22.4 2.2

TiO2 – FetO 9.9 16.8 42.3 11.0 20.1 2.5

TiO2 – FetO 10.3 16.1 45.2 11.1 17.3 2.8

TiO2 – FetO 10.3 15.2 49.4 11.1 14.1 3.2*) All titanium oxides reported as TiO2, **) all iron oxides reported as FetO, #) slag used for equilibrium time experiments, ^) base slag used in tem-perature and reference series experiments. Trace levels (< 0.1 mass%) of Al2O3, P2O5, and MnO were detected in the unreacted slags.

Table 2. Alloy compositions in mass% using ARC-OES and [P] results from ICP-OES.

Description C P Mn Si Al Ti S

Fe–P Master alloy 0.033 0.181 0.217 0.010 0.038 0.009 0.010

[Ti] = 0.05 mass% 0.026 0.170 0.219 0.009 0.031 0.045 0.010

[Ti] = 0.14 mass% 0.033 0.176 0.220 0.010 0.036 0.139 0.010

[Ti] = 0.30 mass% 0.029 0.171 0.221 0.010 0.048 0.301 0.010

Trace levels (< 0.003 mass%) of V were detected in the experimental alloys.

ISIJ International, Vol. 59 (2019), No. 5

© 2019 ISIJ 842

mined and the analysis is given in Fig. 2. It was found that after >9 h reaction the measured (P2O5), [P] and LP each approached a constant value. After this time, it was assumed that the system was at equilibrium. The >9 h to achieve equilibrium is consistent with the 8 to 12 h equilibrium times other researchers have reported.23–25) To ensure equi-librium was achieved in this study, all experiments were run for at least 10 h, the majority being run for 12 h.

In the CO/CO2 experiments, the initial slag composi-tion was approximated to the composition required for gas-slag-steel equilibrium at the experimental temperature. This was to facilitate a fast equilibration time. The required slag composition in equilibrium with the CO/CO2 gas ratio was estimated using the MTDATA thermodynamic soft-ware package.26) MTDATA is a thermodynamic software package that uses Gibbs energy minimisation to calculate complex multi-component phase equilibria in gas-liquid-solid systems. The primary limitation of this estimate was that the MTDATA databases did not contain any titanium oxide data.

In the CO/CO2 experiments the heating-cooling schedule was similar to that of the experiments carried out under high purity Ar. The crucible containing slag and steel was heated to experimental temperature under Ar. The sample was then held at this temperature for 5 minutes prior to the gas being changed over to an Ar–CO–CO2 gas mixture and held for 12 h. After 12 h, the gas was changed back to high purity Ar and the experiments were cooled as per the Ar atmosphere slag-steel experiments.

3. Results and Discussion

3.1. Evaluation of the LP DataThe experimental slag and metal compositions and the

LP values obtained are given in Table 3. A series of seven experiments were conducted using the R series slag com-

position (Table 1) and experimental temperature (1 650°C) to test reproducibility (Table 3). The mean measured LP of the reference slag was 189 with a standard deviation of 7 (3.6% of the LP value).

3.1.1. Use of the Assis et al.27) LP ModelThe intention was to evaluate the experimental LP data

using the recently published LP model by Assis et al.27) (Eq. (21))

logLCaO MgO P O MnO

P �

� � �� �

�

0 068 0 42 1 16 0 2

11570

2 5. % . (% ) . (% ) . (% )( )

TTlog( Fe )t� �10 52 2 5. . %

........................................ (21)

where T is the temperature in K. This model was found to be in excellent agreement with previous work of the authors for slags in the CaO–SiO2–MgO–FetO–MnO–Al2O3–P2O5 system.28)

The advantage of the Assis et al.27) model is that the data used in its development were representative of a wide range of industrial BOS slag compositions and temperatures. The model is essentially an updated version of the Suito et al.29–31) model expanded to include the data of Basu et al.23,24) and Assis et al.27,32) However, the model does not include an input for titania.

To assess whether the Assis et al.27) LP model was suit-able for predicting LP in slags containing significant quanti-ties of TiO2, LP values were calculated from the model and compared with those measured in this study (Table 3) and those reported by Selin16,17) using two separate approaches:

1) Assuming TiO2 has an equivalent effect on LP to SiO2 allowing the use of the as-measured slag composition, as shown in Fig. 3(a),

2) Assuming no TiO2 in the slag and normalising the slag composition to 100 mass%, as shown in Fig. 3(b).

From Fig. 3(a) it can be seen that the Assis et al.27) model under predicts the LP value when TiO2 is assumed to be equivalent to SiO2. The under-prediction (deviation from the 1 to 1 correspondence line) was found to be an average of 11.9% for data measured in this study and 18.6% for Selin’s data. This indicates TiO2 cannot be assumed to be equivalent to SiO2.

From Fig. 3(b) it can be seen that the Assis et al.27) model over-predicts the LP value when there is assumed to be no TiO2 in the slag. The over-prediction (deviation from the 1 to 1 correspondence line) was found to be an average of 14.6% for data measured in this study and 20.3% for Selin’s data. This indicates the effect of TiO2 cannot be disregarded without introducing significant error into the LP prediction.

In a previous study by the current authors28) using simi-lar slag compositions but without TiO2 an excellent fit was achieved using the Assis et al.27) model for LP in the CaO–SiO2–MgO–FetO–MnO–Al2O3–P2O5. Given this, it can be argued that the deviation from the model shown in this study is primarily due to the effect of TiO2. The effect of TiO2 may be accounted for by expanding the model to include TiO2.

The LP data from both this study and Selin’s16,17) investi-gation were used to determine a coefficient for a new TiO2 term to obtain the LP model in Eq. (22). The compositions

Fig. 2. Composition versus time plots for a) (P2O5) measured via XRF and [P] measured via ICP-OES and b) LP from mea-sured (P2O5) and [P].

ISIJ International, Vol. 59 (2019), No. 5

© 2019 ISIJ843

used were those given in Table 3 and reported by Selin.16,17) In developing the modified Assis et al. model, the original composition and temperature coefficients of the Assis et al. model27) were held constant. The adapted model output com-pared to the measured LP in this study and that by Selin16,17) are given in Fig. 4. From Fig. 4, it can be seen that the adapted Assis et al model27) represents the data well, with an average error of 0.5% and 3.9% of the LP for the TiO2 bearing slags in this study and Selin’s16,17) data, respectively.

logL CaO MgO P O

MnO

P � � ��� �

0 068 0 42 1 16

0 20 0 087

2 5. % . (% ) . (% )

. (% ) .

( )

((% )

. . %

TiO

log( Fe )T

t

2

1157010 49 2 5

�

� � �

........................................ (22)

3.1.2. The Effect of TiOx on LP

The measured LP was found to decrease with increasing TiO2 content of the slag. This is shown in Figs. 5(a) and 5(b). The lines on the figures represent predictions by the modified Assis et al.27) model.

The data in Fig. 5(a) are for the R, TiO2, and TiO2– FetO series. The data used in the modified LP model in Eq. (22)

Table 3. Experimental results for CaO–SiO2–MgO–FetO–(MnO–Al2O3–TiO2–P2O5) slags.

Exp. Time Temp (MgO) (Al2O3) (SiO2) (P2O5) (CaO) (TiO2)* (MnO) (FetO)# [P] LP log(pO2)No./Series h °C mass% mass% mass% mass% mass% mass% mass% mass% mass% ratio log(atm)

1 R 10 1 650 9.8 0.4 17.0 1.4 47.2 0.1 0.9 23.3 0.003 201 – 2 R 11 1 650 12.4 0.4 16.5 1.3 45.7 0.0 0.8 22.9 0.003 194 – 3 R 12 1 650 17.3 0.3 15.2 1.3 43.0 0.0 0.8 22.1 0.003 185 – 4 R 12 1 650 11.0 0.5 17.7 4.0 47.8 0.1 1.0 18.0 0.009 193 – 5 R 12 1 650 11.3 0.6 16.4 3.9 45.0 0.1 1.1 21.6 0.009 189 – 6 R 12 1 650 11.8 0.4 16.2 3.7 46.2 0.1 1.1 20.6 0.009 178 – 7 R 12 1 650 9.6 0.4 18.0 3.8 49.4 0.0 1.1 17.7 0.009 186 – 8 TiO2 12 1 650 10.7 0.4 17.1 3.8 46.0 1.1 1.1 19.7 0.009 189 – 9 TiO2 12 1 650 10.4 0.5 15.7 3.7 44.5 5.1 1.2 19.1 0.012 130 –10 TiO2 12 1 650 10.9 0.5 14.8 3.9 42.1 8.2 1.1 18.5 0.018 96 –11 TiO2 12 1 650 9.9 0.6 16.4 3.7 41.5 9.7 1.1 17.2 0.024 67 –12 TiO2 12 1 650 12.4 0.4 12.9 3.7 40.9 12.4 1.2 16.1 0.027 60 –13 TiO2 12 1 650 15.9 0.5 13.7 3.4 35.3 18.0 1.1 12.1 0.090 17 –14 TiO2 12 1 650 9.9 0.6 15.3 1.4 45.1 0.9 0.9 25.9 0.003 185 –15 TiO2 12 1 650 11.1 0.5 16.5 1.4 40.6 9.0 0.8 20.0 0.011 58 –16 TiO2– FetO 12 1 650 18.4 0.5 18.7 2.7 39.7 10.5 1.0 8.5 0.096 12 –17 TiO2 – FetO 12 1 650 17.5 0.5 18.1 3.1 38.5 10.7 1.0 10.6 0.071 19 –18 TiO2 – FetO 12 1 650 15.9 0.4 16.2 3.3 36.1 10.4 1.2 16.7 0.040 36 –19 TiO2 – FetO 12 1 650 19.0 0.4 16.2 3.4 35.4 10.1 1.1 14.3 0.050 29 –20 TiO2 – FetO 12 1 650 17.2 0.6 13.6 3.9 38.8 9.2 1.0 15.7 0.030 56 –21 TiO2 – FetO 12 1 650 25.6 0.3 12.1 2.9 27.9 8.7 1.0 21.5 0.040 31 –22 [Ti] 12 1 650 9.8 0.5 17.3 3.8 47.2 0.5 1.0 19.8 0.009 191 –23 [Ti] 12 1 650 10.8 0.6 17.9 3.9 49.1 1.6 1.0 15.1 0.011 148 –24 [Ti] 12 1 650 10.8 0.6 17.3 3.9 47.6 3.1 0.9 15.7 0.013 131 –25 CO/CO2 12 1 650 6.1 0.3 7.8 1.4 24.3 0.0 0.8 59.2 0.016 38 −8.2526 CO/CO2 12 1 650 9.8 0.2 9.5 2.1 26.7 0.6 0.9 50.4 0.016 58 −8.2527 CO/CO2 12 1 650 11.8 0.2 7.7 2.0 22.0 4.8 0.9 50.6 0.025 34 −8.2528 CO/CO2 12 1 650 13.2 0.3 13.4 2.9 39.4 0.0 1.2 29.4 0.009 142 −8.8029 CO/CO2 12 1 650 10.1 0.3 13.5 3.0 41.1 1.0 1.2 29.8 0.008 175 −8.8030 CO/CO2 12 1 650 7.5 0.3 15.0 1.8 39.8 6.1 0.1 29.4 0.008 100 −8.80

*total titanium oxides reported as TiO2, #total iron oxides reported as FetO.

Fig. 3. Measured LP versus LP calculated using the Assis et al. LP Model27) a) the measured slag composition, and b) the mea-sured slag composition assuming no TiO2 and normalised to 100 mass%. The 1 to 1 correspondence between the mea-sured and calculated LP data is shown by the solid line.

ISIJ International, Vol. 59 (2019), No. 5

© 2019 ISIJ 844

were average experimental conditions of 1 650°C, 10.5 mass% MgO, 3.2 mass% P2O5, 1 mass% MnO, 19 mass% FetO, and 0.5 ±0.2 mass% Al2O3.

The model data in Fig. 5(b) were similar to those in Fig. 5(a) except for FetO and TiO2, where the contents ranged from 10–40 mass% and 0–8 mass%, respectively. The mea-sured LP data in Fig. 5(b) (experiments 4–5, 7–10, 16–18) were constrained to similar conditions (1 650°C, 10.5±1 mass% MgO, 3.2±1 mass% P2O5, 0.5±0.2 mass% Al2O3).

Increasing TiO2 concentration was found to cause a drop off in the LP measured, as shown in Fig. 5. This effect may in part be explained by consideration of the effect of TiO2 on basicity. A lime equivalence term, enclosed by square brackets in Eq. (22), is used to represent basicity in the modified Assis et al.27) model. TiO2 has a lower effective basicity than CaO, as shown by the respective coefficients of 0.087 and 1.0. The lower basicity caused by increasing TiO2 concentration inhibits dephosphorisation, driving the reac-tion as given in Eq. (1) to the left, decreasing the removal

of phosphorus to the slag.It can be seen from Fig. 5(b) that the maximum LP is

achieved at approximately 20 mass% FetO. This is in good agreement with previous studies.33–36) This behaviour is best understood by consideration of the reaction in Eq. (1), where phosphorous removal from steel to slag (i.e. increasing LP) is maximised by the combination of high basicity and high oxygen potential at a fixed temperature. An increase of Fet above ~20 mass% causes a decrease in LP due to the dilution of the basic components of the slag.

3.1.3. The Effect of [Ti] on LP

The effect of initial [Ti] concentration on the final slag composition is shown in Fig. 6. The lines represent linear regression of the data.

Increasing the initial [Ti] concentration was found to decrease the final FetO content and increase the final TiO2 content of the slag. This effect may at least in part be explained by [Ti] reduction of FeO to Fe as given in Eq. (23).

2 2 2( ) [ ] ( )( )FeO Ti Fe TiOl� � � ............... (23)

Furthermore, [Ti] is known to decrease the activity of phosphorus and oxygen in liquid iron, thereby inhibit-ing dephosphorisation. This may be accounted for using the Henrian solution model, Eqs. (24) and (25), and peer reviewed first order interaction coefficient (ePTi = −0.04 and eOTi = −1.12) data.14)

h f PP p[ ] [% ]= ............................. (24)

logf e P e ip PP

Px� ��[% ] [% ] .................. (25)

where fp = Henrian activity coefficient [P], [%P] = mass% of Phosphorus.

The corresponding LP measurements for the data in Fig. 6 are shown in Fig. 7 against the final TiO2 content. As before, the measured LP was found to decrease with increas-ing TiO2. This TiO2 entered the system as [Ti]. Therefore, it can be seen that increasing TiO2, added to the system as [Ti], has a similar effect to adding TiO2 to the slag.

3.2. Development and Testing of an Empirical CPO43−−

ModelThe CPO4

3− data were primarily determined using the experimental slag composition and an empirical CPO4

3− model developed in this study. This approach was used due to the limited availability of thermodynamic data for titanium oxides in the CaO–SiO2–MgO–FetO system. Consequently,

Fig. 4. Measured LP versus LP calculated using the modified Assis et al. LP Model in Eq. (22) and the measured slag composi-tion. The 1 to 1 correspondence between the measured and calculated LP data is shown by the solid line.

Fig. 5. Measured LP versus a) TiO2 at 1 650°C (R, TiO2, and TiO2–FetO series) b) FetO at 1 650°C for 10.5 ±1 mass% MgO, 3.2 ±1 mass% P2O5, 0.5 ± 0.2 mass% Al2O3, and a v-ratio of 2.7–2.8. The lines represent calculated LP values using the modified Assis et al. LP Model in Eq. (22).

Fig. 6. Measured TiO2 and FetO versus Initial [Ti] concentration.

ISIJ International, Vol. 59 (2019), No. 5

© 2019 ISIJ845

the pO2 required for the CPO43− calculation (Eq. (19)) was not

readily determined via thermodynamic calculation.The pO2 is commonly determined experimentally via

Fe2+/Fe3+ speciation. However, the Fe2+/Fe3+ speciation titration method is adversely affected by the presence of other multi-valence oxides, limiting its use for slags contain-ing both FetO and TiO2. Although the pO2 was known in the CO/CO2 gas experiments, it was not practical to conduct all the experiments under these conditions. The inability to use both of these experimental methods for pO2 determination necessitated the development of the empirical CPO4

3− model for use on the titanium oxide containing slags of this study.

The empirical CPO43− model was developed using compiled

equilibrium CPO43− data for the CaO–SiO2–MgO–FetO–

MnO–Al2O3–TiOx–P2O5 system from reviewed and recent data sources.14–17,21,22,24,29–31,33,37–67) These data include the TiO2 bearing slags data of Selin16,17) but not any experimen-tal data from this study.

The reported (P), [P] and [O] (or when reported the pO2) from the compiled data were used to determine the CPO4

3− using Eq. (19), where K[P] was determined using Eq. (16) and ideal Henrian behaviour was assumed for [P] (i.e. fP = 1). The pO2 was determined from the reported [O] using the methodology detailed by Ishii and Fruehan,40) where the free energy of the formation of FeO (Eq. (27)) and the activity of FeO in the slag (aFeO) were derived from the experimen-tal dissolved oxygen in iron [O] data (Eq. (29)) using the Henrian solution model.

Fe O FeOl g l( ) ( ) ( )� �12 2 ..................... (26)

�G T J mol� � � � � �234 620 46 90 1. ( ) .......... (27)

Ka

pFeO

FeO

O

( ) /=

2

1 2 ............................. (28)

af O

f OFeO

O

O satsat

=[ ]

[ ]. .

.......................... (29)

log ( / . )[ ] . [% ]f T O PO � � � �1 750 0 76 0 006 ..... (30)

logf T OO satsat . ( / . )[ ] .� � �1 750 0 76 ............. (31)

log[O] Tsat. / .� �6 320 2 734 ................. (32)

The aFeO is derived from experimental data using the Taylor and Chipman relation,68) fO and [%O] are the activity coefficient and mass% of oxygen in liquid iron, respectively.

Table 4. Slag systems covered in CPO43− database.

No. Studies Slag System References

7 CaO–MgO–P2O5–SiO2–FetO 24,30,31,49,55,64,69)

2 CaO–MgO–P2O5–SiO2–FetO–MnO–Al2O338,45)

3 CaO–MgO–P2O5–SiO2–FetO–Al2O337,39,40)

1 CaO–P2O5–FetO–Al2O341)

1 CaO–P2O5–SiO2–FetO–Al2O343)

1 CaO–P2O5–SiO2–FetO 44)

2 CaO–P2O5–SiO2–FetO–MnO 46,56,58)

1 MgO–P2O5–SiO2–FetO–MnO 47)

3 CaO–MgO–P2O5–SiO2–FetO–MnO–Al2O3–TiO2

16–22)

1 CaO–MgO–P2O5–FetO 51,53)

1 CaO–MgO–P2O5–SiO2–FetO–MnO–Al2O3–TiOx–V2O5

16–20)

1 CaO–MgO–P2O5–SiO2–FetO–MnO 63)

Table 5. The maximum, minimum, mean and standard deviation for temperature, slag composition, [P], pO2 and CPO4

3− covered in CPO4

3− database.

Variable Mean Standard Deviation Minimum Maximum

Temp 1 608 38 1 530 1 700

(CaO) 40.6 16.6 0.00 62.9

(MgO) 6.06 7.8 0.00 40.4

(P2O5) 11.87 13.1 0.00 49.0

(SiO2) 9.02 10.8 0.00 42.6

(FetO) 28.30 19.1 0.1 91.8

(MnO) 1.53 6.4 0.00 70.1

(Al2O3) 2.40 7.3 0.00 44.6

(TiO2) 0.08 0.7 0.00 12.6

(V2O5) 0.06 0.7 0.00 12.7

[P] 0.170 0.450 0.00 2.950

LP 248 301 0 2 290

log(pO2) −9.09 0.53 −11.81 −8.01

log(CPO43−) 18.09 1.09 14.37 20.20

Fig. 7. Measured LP versus TiO2 entering the system as [Ti].

The subscript sat., refers to oxygen saturation in iron.From this approach a database of 1 382 laboratory mea-

surements of CPO43− was produced. The slag systems and

characteristic data (mean, standard deviation, minimum and maximum) of key variables covered in this database are detailed in Tables 4 and 5 respectively.

The CPO43− database was used to develop the empirical

model for CPO43− given in Eq. (33).

log CT

CaO MgO

P O

PO( ) . ( ) . ( )

. ( ) .

43

46 1410 020 0 010

0 077 0 1022 5

� � � �

� � (( ) . ( ) . ( )

. ( ) . ( ) .

SiO Fe O MnO

Al O TiOt2

2 3 2

0 059 0 040

0 084 0 079 0

� �� � � 0080 3 4752 5( ) .V O �

........................................ (33)

ISIJ International, Vol. 59 (2019), No. 5

© 2019 ISIJ 846

A comparison of the measured data with those calculated from the CPO4

3− model in Eq. (33) is given in Fig. 8. Also shown in the figure is the coefficient of determination (R2) of 0.89. The Selin data16–20) are highlighted in this figure to demonstrate that the model is a good representation of the CPO4

3− of titanium oxide containing slags. It was not treated as a separate series in the derivation of Eq. (33).

To validate the empirical CPO43− model, the model predic-

tions using Eq. (33) were tested against experimental CPO43−

data generated in this study and calculated from Eq. (19). To calculate the CPO4

3− using Eq. (19) a pO2 is required. For TiO2 free slags carried out under Ar (R series in Table 3), the pO2 was determined using MTDATA26) with inputs of the slag and steel compositions. This approach was explained in detail in a previous study.28) For the experiments using CO/CO2 gas, the pO2 of the gas phase was fixed (CO/CO2 series in Table 3). The results are shown in Fig. 9. It should

be noted that these experimental CPO43− were not used in the

development of the CPO43− model. The high R2 value of 0.98,

indicates that the model can predict the CPO43− of the titanium

oxide containing slags in this study well.

3.2.1. The Effect of TiO2 on CPO43−

The data in Fig. 9 have been replotted in Fig. 10 so that the effect of TiO2 on CPO4

3− can be more readily assessed.The CPO4

3− was observed to have a weakly decreasing trend with increasing TiO2 content of the slag. The experimental data shown in Fig. 10 were constrained to slags containing 10.5 ± 1 mass% MgO, a v-ratio of 2.7±0.1, and 19 ± 3 mass% FetO. The modelling inputs were the average experi-mental conditions 1 650°C, 10.5 mass% MgO, 3.2 mass% P2O5, 1 mass% MnO, v-ratio of 2.7, and 19 mass% FetO. The scatter observed in the experimental results around the model line in Fig. 10 may partly be attributed to variations in the slag FetO, v-ratio, and MgO.

The observed decreasing trend in CPO43− with TiO2 may

be in part be explained by considering the effect of TiO2 on basicity. In the CPO4

3− model (Eq. (33)), basicity is rep-resented by a lime equivalence coefficient for each oxide. A lower coefficient represents a more acidic or less basic oxide, for example the coefficient for SiO2 is −0.102. TiO2 has a lower coefficient (−0.079) than CaO (0.02), MgO (−0.01), and FetO (−0.059). Hence, additions of TiO2 to a fixed CaO–MgO–FetO–SiO2 base slag composition decreases the slag basicity and consequently the predicted CPO4

3− .

3.3. Industrial RelevanceImproved understanding of the effect of TiO2 on the

CPO43− allows steelmakers to optimise hot metal and steel

processing to minimise flux usage and re-processing of out of specification heats with varying titanium concentration. For example, in the BOS, steelmakers often use the simple v-ratio of the predicted slag composition from a heat and mass balance of all charged materials (e.g. hot metal, scrap, fluxes, iron ore…) to determine the CPO4

3− of a slag. If the TiO2 concentration is high, the CPO4

3− is likely to be lower than predicted by simple v-ratio relations. Correct predic-tion of the CPO4

3− of high TiO2 containing slags enables the steelmakers to add the correct amount of flux to achieve the required grade [P] specification.

Fig. 8. Measured log(CPO43−) versus log(CPO4

3−) calculated from the empirical model (Eq. (33)) for published experimental data.14–17,21,22,24,29–31,33,37–67) The 1 to 1 correspondence between the measured and calculated log(CPO4

3−) data is shown by the solid line.

Fig. 9. Measured log(CPO43−) versus log(CPO4

3−) calculated from the empirical model (Eq. (33)) for the experimental data. The 1 to 1 correspondence between the measured and calcu-lated log(CPO4

3−) data is shown by the solid line.Fig. 10. log(CPO4

3−) versus TiO2 mass% under Ar. The line repre-sents the calculated log(CPO4

3−) using Eq. (33).

ISIJ International, Vol. 59 (2019), No. 5

© 2019 ISIJ847

4. Conclusions

Slag-steel equilibrium experiments were performed using CaO–SiO2–MgO–FetO–TiO2 synthetic slags and Fe–P–Ti alloys representative of BOS. The (TiO2) and [Ti] concen-trations were varied to investigate their effect on the LP and CPO4

3− . Increasing (TiO2) and [Ti] were found to lower both the LP and CPO4

3− in CaO–SiO2–MgO–FetO–TiO2 slags, which was attributed to lowering the slag basicity.

The Assis et al.27) LP model was modified to include the effect of TiO2 on slag basicity to give good agreement with the LP data from this study and Selin’s study.16,17) Further, an empirical CPO4

3− model was developed to predict the capacities in the CaO–SiO2–MgO–FetO–TiO2 slag system. The CPO4

3− model was developed using a large dataset of published slag and pO2 data for slag systems relevant to the BOS process and was validated using experimental CPO4

3− data from this study. The empirical CPO43− model was

developed to overcome issues associated with the limited availability of thermodynamic data for TiOx in the CaO–SiO2–MgO–FetO system and the difficulties in determining the pO2 experimentally.

AcknowledgementsFunding from the Australian Research Council Industrial

Transformation Research Hubs Scheme (Project Number IH130100017) is gratefully acknowledged. The authors would also like to acknowledge Ian Webb of the University of Wollongong for helping with the experimental program.

REFERENCES

1) T. A. Boom, D. R. Foshnacht and D. M. Haezebrouck: Iron Steel-maker, 17 (1990), 35.

2) T. A. Boom, D. R. Foshnacht and D. M. Haezebrouck: Iron Steel-maker, 17 (1990), 59.

3) B. J. Monaghan, R. J. Pomfret and K. S. Coley: Metall. Mater. Trans. B, 29 (1998), 111.

4) E. T. Turkdogan: ISIJ Int., 40 (2000), 964.5) Y. Morizane, B. Ozturk and R. J. Fruehan: Metall. Mater. Trans. B,

30 (1999), 29.6) J. J. Pak, J. O. Jo, S. I. Kim, W. Y. Kim, T. I. Chung, S. M. Seo, J.

H. Park and D. S. Kim: ISIJ Int., 47 (2007), 16.7) C. Ariyo, P. Gonzalez and L. Holappa: Steel Res. Int., 76 (2005), 284.8) C. O. Ariyo and L. E. K. Holappa: Scand. J. Metall., 31 (2002), 385.9) B. Yan, J. Zhang and Q. Song: Molten 2009: VIII Int. Conf. on

Molten Slags, Fluxes and Salts, ed. by M. Sanches et al., GECAMIN, Santiago, (2009), 309.

10) F. Tsukihashi, A. Tagaya and N. Sano: Trans. Iron Steel Inst. Jpn., 28 (1988), 164.

11) C. Wagner: Metall. Trans. B, 6 (1975), 405.12) P. Drain, B. J. Monaghan, G. Zhang, R. Longbottom, I. Murgas and

M. W. Chapman: Chemeca 2016: Chemical Engineering - Regen-eration, Recovery and Reinvention, Engineers Australia, Melbourne, (2016), 223.

13) P. B. Drain, B. Monaghan, G. Zhang, R. J. Longbottom, M. W. Chapman and S. J. Chew: Ironmaking Steelmaking, 44 (2017), 721.

14) M. Hino and K. Ito: Thermodynamic Data for Steelmaking, Tohoku University Press, Sendai, (2010), 208.

15) D. Kirner and D. Janke: Slag Atlas, 2nd ed., Verlag Stahleisen GmbH, Düsseldorf, (1995), 268.

16) R. Selin: Scand. J. Metall., 20 (1991), 318.17) R. Selin: Scand. J. Metall., 20 (1991), 279.18) R. Selin, Y. Dong and Q. A. Wu: Scand. J. Metall., 19 (1990), 98.19) R. Selin: 3rd Int. Conf. on Molten Slags and Fluxes, The Institute of

Metals, East Kilbride, Scotland, (1988), 317.20) R. Selin: Ph.D. thesis, KTH - The Royal Institute of Technology,

Stockholm, (1987).

21) G. J. W. Kor: Metall. Trans. B, 8 (1977), 107.22) T. Usui, K. Yamada, Y. Kawai, S. Inoue, H. Ishikawa and Y. Nimura:

Tetsu-to-Hagané, 77 (1991), 1641 (in Japanese).23) S. Basu, A. K. Lahiri and S. Seetharaman: Metall. Mater. Trans. B,

38 (2007), 357.24) S. Basu, A. K. Lahiri and S. Seetharaman: Metall. Mater. Trans. B,

38 (2007), 623.25) A. N. Assis, M. A. Tayeb, S. Sridhar and R. J. Fruehan: Metall.

Mater. Trans. B, 46 (2015), 2255.26) R. H. Davies, A. T. Dinsdale, J. A. Gisby, J. A. J. Robinson and S.

M. Martin: Calphad, 26 (2002), 229.27) A. N. Assis, R. J. Fruehan and S. Sridhar: Iron and Steel Technology

Conf. (AISTech 2012), AIST, Warrendale, PA, (2012), 861.28) P. B. Drain, B. J. Monaghan, R. J. Longbottom, M. W. Chapman, G.

Zhang and S. Chew: ISIJ Int., 58 (2018), 1965.29) F. N. H. Schrama, E. M. Beunder, B. Van den Berg, Y. Yang and R.

Boom: Ironmaking Steelmaking, 44 (2017), 333.30) H. Suito, R. Inoue and M. Takada: Trans. Iron Steel Inst. Jpn., 21

(1981), 250.31) H. Suito, R. Inoue and M. Takada: Tetsu-to-Hagané, 67 (1981), 75

(in Japanese).32) A. N. Assis: Ph.D. thesis, Carnegie Mellon University, Pittsburgh,

PA, (2014), http://repository.cmu.edu/dissertations/464, (accessed 2015-11-18).

33) K. Balajiva, A. G. Quarrell and P. Vajragupta: JISI, 153 (1946), 115.34) K. Balajiva and P. Vajragupta: J. Iron Steel Inst., 155 (1947), 563.35) P. Vajragupta: J. Iron Steel Inst., 158 (1948), 494.36) R. G. Ward: An Introduction to The Physical Chemistry of Iron and

Steel Making, Edward Arnold, London, (1962), 121.37) W. A. Fischer and H. Vom Ende: Stahl Eisen, 72 (1952), 1398.38) C. Garlick, G. R. Belton and S. Jahanshahi: A Thermodynamic Study

of Metal/slag Phosphorus Equilibria, Report CRL_R_48_86, BHP Central Research Laboratories, Shortland, N.S.W., (1986).

39) M. Hino, S.-y. Yamamoto and S. Ban-Ya: Tetsu-to-Hagané, 79 (1993), 1039 (in Japanese).

40) H. Ishii and R. J. Fruehan: Iron Steelmaker, 24 (1997), 47.41) S. Ban-Ya, M. Hino, A. Sato and O. Terayama: Tetsu-to-Hagané, 77

(1991), 361 (in Japanese).42) N. Bannenberg: Stahl Eisen, 111 (1991), 71.43) J. C. Wrampelmeyer, S. Dimitrov and D. Janke: Steel Res., 60 (1989),

539.44) S. Nakamura, F. Tsukihashi and I. Sano: ISIJ Int., 33 (1993), 53.45) I. H. Jung, J. D. Seo and S. H. Kim: Steel Res., 71 (2000), 333.46) H. Knuppel and F. Oeters: Stahl Eisen, 81 (1961), 1437.47) Y. Kobayashi, N. Yoshida and K. Nagai: ISIJ Int., 44 (2004), 21.48) K. Kunisada and H. Iwai: Trans. Iron Steel Inst. Jpn., 27 (1987), 332.49) C. M. Lee and R. J. Fruehan: Ironmaking Steelmaking, 32 (2005),

503.50) X. M. Yang, J. Y. Li, M. Zhang and J. Zhang: Ironmaking Steelmak-

ing, 43 (2016), 39.51) R. Nagabayashi, M. Hino and S. Ban-Ya: Tetsu-to-Hagané, 74

(1988), 1770 (in Japanese).52) R. Nagabayashi, M. Hino and S. Ban-ya: Tetsu-to-Hagané, 74 (1988),

1585 (in Japanese).53) R. Nagabayashi, M. Hino and S. Ban-Ya: Tetsu-to-Hagané, 74

(1988), 1577 (in Japanese).54) Y. Shirota, K. Katohgi, K. Klein, H. J. Engell and D. Janke: Trans.

Iron Steel Inst. Jpn., 25 (1985), 1132.55) T. Ting, H. G. Katayama and A. Tanaka: Tetsu-to-Hagané, 72 (1986),

225 (in Japanese).56) G. Trömel and W. Fix: Arch. Eisenhüttenwes., 33 (1962), 745.57) G. Trömel and W. Fix: Arch. Eisenhüttenwes., 32 (1961), 209.58) G. Trömel, W. Fix and H. Werner Fritze: Arch. Eisenhüttenwes., 32

(1961), 353.59) A. Werme and P.-A. Lundh: Scand. J. Metall., 16 (1987), 33.60) R. W. Young, J. A. Duffy, G. J. Hassall and Z. Xu: Iron Steelmaking,

19 (1992), 201.61) R. W. Young: Use of the Optical Basicity Concept for Determining

Phosphorus and Sulphur Slag/metal Partitions, European Commis-sion, Luxemburg, (1991), 1.

62) H. Suito and R. Inoue: ISIJ Int., 35 (1995), 258.63) H. Suito and R. Inoue: ISIJ Int., 35 (1995), 266.64) H. Suito and R. Inoue: Trans. Iron Steel Inst. Jpn., 24 (1984), 40.65) H. Suito and R. Inoue: Trans. Iron Steel Inst. Jpn., 24 (1984), 257.66) M. Y. M. Ahmed: Ph.D. thesis, University of Utah, (2013), https://

www.researchgate.net/publication/257960806, (accessed 2017-07-09).67) J. Zhou, X. Bi, R. Yue and F. Yang: ISIJ Int., 57 (2017), 706.68) C. R. Taylor and J. Chipman: Trans. AIME, 154 (1942), 228.69) K. Kunisada and H. Iwai: Trans. Iron Steel Inst. Jpn., 27 (1987), 263.