Sponsor TBD Report No.

AISI/DOE Technology Roadmap Program

Final Report

Behavior of Phosphorus in DRUHBI During

Electric Furnace Steelmaking

Richard J. Fruehan
Christopher P. Manning

October 5,2001

Work Performed under Cooperative Agreement

NO.DE-FC07-97ID13554

Prepared for
US.Department of Energy
Prepared by
American Iron and Steel Institute
Technology Roadmap Program Office
Pittsburgh, PA 15220

1
 DISCLAIMER

"This report was prepared as an account of work sponsored by an Agency of the United
States Government. Neither the United States Government nor any agency thereof, nor
any of their employees, makes any warranty, express or implied, or assumes any legal
liability or responsibility for the accuracy, completeness, or usefulness of any
information, apparatus, product, or process disclosed, or represents that it use would not
infringe privately owned rights. Reference herein to any specific commercial product,
process, or service by trade name, trademark, manufacturer, or otherwise, does not
necessarily constitute or imply endorsement, recommendation, or favoring by the United
States Government or any agency therof. The views and opinions of authors expressed
herein do not necessarily state or reflect those of the United States Government or any
agency thereof."

"This report has been reproduced from the best available copy. Available in paper copy
and microfiche"

Number of pages in report: 2 15

DOE and DOE contractors can obtain copies of this report

FROM: Office of Scientific and Technical Information, P.O.
Box 62, Oak Ridge, TN 37831. (615) 576-8401

This report is publicly available from the department of Commerce, National

Technical Information Service, 5285 Port Royal Road, Springfield, VA 22161.
(703) 487-4650.

..
11
 TABLE OF CONTENTS

...
Table of Contents 111

List of Figures V

List of Tables xi

List of Symbols xii

...
Executive Summary x111

Acknowledgements xv

Chapter 1: Introduction 1
1.2 Research Objectives 4

Chapter 2 : Literature Review 5

2.1 Behavior of Phosphorus in DFU/HBI 5
2.2 -
Slag Metal Phosphorus Equilibrium 8
2.3 Kinetics of Phosphorus Transfer 15

Chapter 3: Experiment Procedure 33

3.1 Metallographic Observations 33
3.2 Fast Melting Experiments 33
3.3 Controlled Melting Observations 36
3.4 Laboratory Kinetic Experiments 37

Chapter 4 Results and Discussion 41

4.1 Metallographic Observations 41
4.2 Fast Melting Experiments 47
4.3 Controlled Melting Observations 55
4.4 Kinetics of Phosphorus Transfer 56
4.4.1 Mathematical Model of Phosphorus Transfer
4.4.2 Laboratory Kinetic Experiments

Chapter 5 Plant Trials 94

5.1 Pennsylvania Steel Technologies, Inc - Bethlehem Steel Mass 94
Transfer Trials
5.2 North Star Steel - Texas EAF trials with DRI/HBI 110
5.3 BHP - Sydney Steel Mill EAF trials with HBI 128

...
111
Chapter 6 Process Model 134
6.1 Slag and Temperature Model 134
6.1.1 Iron Oxide Formation and Reduction During EAF Steelmaking 135
6.1.2 Flux Additions and Slag Make During EAF Steelmaking 142
6.1.3 Furnace Temperature During EAF Steelmaking 150
6.2 Numerical Process Model for Phosphorus Transfer During EAF 153
Steelmaking
6.2.1 NSST Electric Furnace Model 154
6.2.2 BHP Electric Furnace Model - Continuous Feeding of HBI 158
6.3 Value of a Predictive Process Model 164
6.3.1 Influence of Mass Transfer Kinetics Upon Metal Phosphorus at Tap 164
6.3.2 Effect of Melt-In FeO Upon Metal Phosphorus at Tap 167
6.3.3 V-Ratio and Phosphorus Mass Transfer 171
Chapter 7: Conclusions 175
7.1 Investigation of Commercial DRIMBI 175
7.2 Laboratory Phosphorus Kinetic Study 176
7.3 EAF Plant Trials 177
7.4 Process Model for Phosphorus Control in the EAF 178

Chapter 8: Future Work 180

8.1 Laboratory Experiments 180
8.2 Plant Trials 181
8.3 Process Model 182

Appendices 183

References 186

iv
 List of Figures

Chapter 2:
Figure 1: Secondary electron image of HBI adapted from Trotter et al. [ I ] 5
Figure 2: SEM oxygen map of HBI adapted from Trotter et al. [l] 6
Figure 3: SEM phosphorus map of HBI adapted from Trotter et al. [ 13 6
Figure 4: Equilibrium phosphorus partition ratio as a function of temperature as 14
predicted by three correlations assuming a slag chemistry of 25%Fe0 -
7%Mg0 - 46%Ca0 - 22%Si02
Figure 5: Equilibrium phosphorus partition ratio as a function of FeO as predicted by 14
three
correlations assuming T = 1600°C and constant basicity, CaOl SiOz= 2.1.
Figure 6: Schematic representation of the film model 16
Figure 7: Schematic representation of the motion of a packet or element of fluid to and 18
away from the interface according to the penetration theory.

Figure 8: Experimental apparatus used for Phosphorus kinetic study by Mori et al. [34] 22
Figure 9: Dephosphorization kinetic data adapted from Mori et al. [34] plotted with 24
calculated curves for best fit overall mass transfer coefficients, ko. The thee
data sets shown were collected for approximately the same slag chemistry at
different experimental temperatures.
Figure 10: Dephosphorization kinetic data adapted from Mori et al. [34] plotted with 24
calculated curves for best fit overall mass transfer coefficients, ko. The thee
data sets shown were collected for different slag chemistries at the same
experimental temperature(164OOC).
Figure 11: rephosphorization kinetic data adapted from Mori et al. [34] plotted with 25
calculated curves for best fit overall mass transfer coefficients, ko. The thee
data sets shown were collected for different slag chemistries at different
experimental temperatures.
Figure 12: Rephosphorization kinetic data adapted from Mori et al. [34] plotted with 26
calculated curves for best fit overall mass transfer coefficients, ko. The thee
data sets shown were collected for the same slag chemistry at different
experimental temQeratures.

Figure 13: Calculated ks versus (1/?l)ln. Adapted from Monaghan et al. [40]. 31

ChaDter 3:
Figure 14: Experimental apparatus used for quick melting and quenching of DRI and 35
HBI.
Figure 15: Experimental apparatus for kinetic experiments. 39
Chapter 4
Figure 16: SEM micrograph of polished Midrex DRI cross section. 41
Figure 17: SEM micrograph of polished Midrex DRI cross section. 42
Figure 18: SEM micrograph of polished Midrex HBI cross section. 42
Figure 19: SEM micrograph of polished Midrex HBI cross section. The approximate 43
chemical composition of the indicated oxide feature was measured via EDS.

V
Figure 20: SEM micrograph of polished Midrex HBI cross section. The approximate 43
chemical composition of the indicated oxide feature was measured via EDS.
Figure 21: Optical micrograph of polished BHP HBI cross section. 44
Figure 22: SEM micrograph of polished BHP HBI cross section. 45
Figure 23: SEM micrograph of polished cross section of BHP HBI. The approximate 46
of the large oxide feature was measured via EDS.
Figure 24: SEM micrograph of polished cross section of BHP HBI. The 47
approximate composition of the large oxide feature was measured via
EDS.
Figure 25: Results of fast melting experiments with six commercial DRI and HBI 48
materials.
Figure 26: Calculated metal phosphorus content versus time assuming a liquid phase 49
mass transfer controlling model.
Figure 27: Calculated gangue phosphorus content versus time assuming a liquid phase 50
mass transfer controlling model
Figure 28: Metal phosphorus as a function of time for Circal HBI melted into a 51
magnesia saturated liquid slag with basicity =1 at 1600°C.
Figure 29: Metal phosphorus as a function of time for BHP - HBI melted into a 52
magnesia saturated liquid slag with basicity = I at 1600°C.
Figure 30: Fast melting experiments with Circal HBI with additions of 1 % and 1.5 % 53
CaO and 1.78% CaC03.
Figure 31: Calculated metal phosphorus versus time for the experimental parameters of 61
this study assuming an overall mass transfer coefficient equal to 0.001cdsec.
Figure 32: Calculated slag phosphorus versus time for the experimental parameters of 61
this study assuming an overall mass transfer coefficient equal to 0.00lcdsec.
Figure 33: Metal phosphorus versus time for dephosphorization experiment #12. 64
Continuous curve calculated via equation [46] assuming a best fit ko.
Figure 34: Metal phosphorus versus time for dephosphorization experiment #13. 64
Continuous curve calculated via Equation [46] assuming a best fit ko.
Figure 35: Metal phosphorus versus time for dephosphorization experiment #14. 65
Continuous curve calculated via Equation [46] assuming a best fit ko.
Figure 36: Data from dephosphorization experiment #12 replotted as the RHS of 66
equation [48] versus time.
Figure 37: Data from dephosphorization experiment #13 replotted as the RHS of 66
equation [48] versus time.
Figure 38: Data from dephosphorization experiment #14 replotted as the RHS of 67
equation[48] versus time.
Figure 39: Adapted from Mori et al. [34]. Data from dephosphorization experiment 68
shown in Figure 12 replotted as the RHS of equation 48 versus time.
Figure 40: Adapted from Mori et al. [34]. Data from rephosphorization experiment 68
shown in Figure13, and replotted here as the RHS of equation 48 versus time.

Figure 41 : Metal phosphorus versus time for dephosphorization experiment #18. The 70
solid line was calculated via Equation [46] assuming a constant, best fit ko.
Figure 42: Metal phosphorus versus time for dephosphorization experiment #19. The 70
continuous curve was calculated via Equation [46] assuming a constant, best
fit ko.
Figure 43: Metal phosphorus versus time for dephosphorization experiment #20. The 71
continuous curve was calculated via Equation [46] assuming a constant, best
fit ko.
Figure 44: Metal phosphorus versus time for dephosphorization experiment #22. The 71
continuous curve was calculated via Equation [46] assuming a constant, best
fit ko.
Figure 45: Data from dephosphorization experiment #18 replotted as the RHS of 72
equation[48] versus time.

vi
 Figure 46: Data from dephosphorization experiment #19 replotted as the RHS of 72
equation[48] versus time.
Figure 47: Data from dephosphorization experiment #20 replotted as the RHS of 73
equation[48] versus time.

Figure 48: Data from dephosphorization experiment #22 replotted as the RHS of 73
equation[48] versus time.
Figure 49: Results of rephosphorization experiment #16 plotted as metal phosphorus 74
versus time. The continuous curve was calculated via equation [45]
assuming a best fit ko.
Figure 50: Data from Rephosphorization Experiment #16 plotted as the RHS of 75
equation [47] versus time.
Figure 5 1: The apparent mass transfer parameter, A*ko, as a function of time for 77
dephosphorization experiments 18 - 22.
Figure 52: Adapted from Gaye et al. [49]. A schematic representation of the interfacial 78
tension, y, (upper graph) and metal phosphorus content (lower graph) as
functions of time.
Figure 53: Optical micrograph of the slag-metal interface, quenched 4 minutes after the 81
start of dephosphorization. White spots are metal droplets suspended in the
slag.
Figure 54: Composite of several optical micrographs of the slag metal interface 81
quenched 4 minutes after the start of dephosphorization. White spots are
metal droplets suspended in the slag phase.
Figure 55: X-ray images adapted from sessile drop experiments with Fe-O.l%P 82
conducted by Seetharaman et al. [50].
Figure 56: Metal phosphorus versus time data for dephosphorization experiment 83
conducted by Seetharaman et al. [50]
Figure 57: Experimental dephosphorization data from Figure 56, replotted as the RHS 84
of equation [48] versus time. Data are adapted from Seetharaman et al. [50]
Figure 58: Simulated phosphorus mass transfer during the sessile drop experiments of 85
Seetharaman et al. [50]
Figure 59: Metal phosphorus versus time for dephosphorization experiment #2 1. 87
Continuous curve calculated via Equation E461 assuming a constant, best fit
ko.
Figure 60: Data from dephosphorization experiment #21 replotted as the RHS of 88
equation[48] versus time.
Figure 61: Oxygen transfer kinetic data adapted from Mori et al. [34] The continuous 89
curve indicates only the trend not a fundamental calculation.
Figure 39: Adapted from Mori et al. [34]. Data from dephosphorization experiment 90
shown in Figure 10 replotted as the RHS of equation 48 versus time.
Figure 62: Adapted from Mori et al. [34]. Data from dephosphorization experiment 91
shown in Figure 12 replotted as the RHS of equation 48 versus time.
Figure 63: Oxygen transfer kinetic data adapted from Mori et al. [34] The continuous 91
curve was plotted assuming liquid phase mass transfer controlled kinetics
and a best fit mass transfer parameter.
Figure 64: RHS of equation 53 versus time. 92
Chapter 5
Figure 65: Bath temperature versus time. 97
Figure 66: Slag chemistry versus time. All components are in weight percent. 97
Figure 67: Slag and metal phosphorus content versus time. 98
Figure 68: Calculated equilibrium and observed slag-metal phosphorus distribution 99
versus time
Figure 69: Measured bulk metal phosphorus and calculated metal interface phosphorus 100
versus time

vii
 Figure 70: RHS of equation 55 versus time. 101
Figure 71: Continuous evaluation of equation 55 along continuous curves in Figure 67. 102
Figure 72: Bath temperature versus time 103
Figure 73: Slag chemistry versus time. All components are in weight percent. 103
Figure 74: Slag and metal phosphorus versus time. 104
Figure 75: Measured bulk and calculated interfacial metal phosphorus versus time 105
Figure 76: Temperature versus time. 106
Figure 77: Slag chemistry versus time. An addition of 100 lbs was made at t = 3 106
minutes.
Figure 78: Measured bulk slag and metal phosphorus content versus time. 107
Figure 79: Measured bulk and calculated interfacial phosphorus content versus time. 108
Only the final four samples of this heat are considered due to uncertainties in
the amounts of solids in the steel and slag phase.
Figure 80: Continuous evaluation of equation 55 according to the data in Figure 79. 109
Figure 8 1: Degree of final reduction of residual iron oxide and deposited carbon in 112
commercial DRI. Data adapted from Goldstein et al. [54]
Figure 82: Temperature profile within DRI pellets as modeled by Goldstein.[54] 112
Figure 83: Wt% of slag component X versus time for a 100% scrap heat. The first 116
sample was taken 15 minutes after the dropping of the third bucket.
Figure 84: Wt% of slag component X versus time for a 25% DRI - 75% scrap heat. 117
The first sample was taken 7 minutes after the dropping of the third charge.
Figure 85: Final wt% FeO in the slag versus pounds of Circal in the charge. 120
Figure 86: Final wt% FeO in the slag versus oxygen injected through the door lance and 121
super sonic injector.
Figure 87: Final wt% FeO in the slag versus the moles of iron oxide in Circal plus the 122
moles of injected oxygen minus the moles of carbon contained in the pig
iron for the trial heats with Circal HBI.
Figure 88: Observed phosphorus distribution at tap versus the equilibrium distribution 123
as calculated by the Ide et al. correlation.
Figure 89: Observed phosphorus distribution at tap versus the equilibrium distribution 123
as calculated by the Turkdogan correlation.
Figure 90: Observed phosphorus distribution at tap versus the equilibrium distribution 124
as calculated by the Zhang et al. correlation.
Figure 9 1: Furnace temperature versus time during the refining period for one heat with 125
25% DRI - 75% scrap.
Figure 92: Slag chemistry versus time during the refining period for one heat with 25% 126
DRI - 75% scrap.
Figure 93: Slag chemistry versus time during the refining period for one heat with 25% 126
DRI - 75% scrap.
Figure 94: Measured bulk metal phosphorus and calculated metal interface phosphorus 127
versus time
Figure 95: Slag chemistry versus time for a single two bucket batch charged EAF heat. 129
The initial sample was taken between the first and final scrap charges to the
furnace.
Figure 96: Slag chemistry versus time for a trial heat with 60 tons of continuously fed 131
HBI.
Figure 97: Phosphorus content of the slag versus time for the trial heat with 60 tons of 132
continuously fed HBI shown in Figure 96.
Chapter 6
Figure 84: Wt% of slag component X versus time for a 25% DRI - 75% scrap heat. 138
The first sample was taken 7 minutes after the dropping of the third charge.

viii
Figure 98: Iron oxide content of the slag versus time according to equation 42. 140
Figure 99: Interfacial area between the injected carbon and slag as function of time 141
during refining.
Figure 100: FeO in slag as function of time during refining. 142
Figure 101: Slag model 0 - constant CaO, MgO, and SiOz, with changing FeO and no 144
slag flushing
Figure 102: Slag Model 1 - Constant FeO and SiOz, with increasing CaO and MgO and 145
no slag flushing
Figure 103: Slag Model 2- Constant FeO and SiOz, with increasing CaO and MgO with 146
slag flushing
Figure 104: Slag Model 3 - FeO increase and reduction, SiOz increase, CaO and MgO 147
increase, and slag flushing
Figure 96: Slag chemistry versus time for a trial heat with 60 tons of continuously fed 148
HBI.
Figure 105: Slag Model 3.1- Simulation of HBI continuous feeding. 149
Figure 106: Steel temperature profile during a hypothetical EAF heat with a three 150
bucket charge and a 75 minute tap to tap time.
Figure 107: Steel temperature versus time during the refining period of the heat shown 151
in Figure 106
Figure 108: Steel bath temperature versus time for a 25% DRI - 75% scrap heat. The 152
first sample was taken 7 minutes after the dropping of the third charge.
Figure 109: Temperature profile during continuous feeding of HBI into the BHP - 152
Sydney electric furnace.
Figure 110: Steel temperature profile during a hypothetical continuously fed EAF heat. 153
Figure 111: Predicted slag chemistry profile during the refining period of an electric 155
furnace heat at NSST.
Figure 112: Slag-Metal equilibrium phosphorus distribution predicted by the Ide et al. 156
correlation for the slag chemistry and temperature data of Figures 111 and
108.
Figure 113: Metal phosphorus content predicted by the numerical process model for an 157
EAF heat at NSST. The two discrete points are actual furnace chemistries.
Figure 114: Slag phosphorus content predicted by the numerical process model for an 157
EAF heat at NSST. The three discrete points are actual measured furnace
slag chemistries.
Figure 115: Modeled slag chemistry profile for the BHP electric furnace trial with 159
continuously fed HBI. The actual slag chemistry profile is shown in Figure
96.
Figure 116: Simulated temperature profile for the BHP electric furnace trial with 160
continuously fed HBI. The actual temperature profile is shown in Figure
109.
Figure 17: Slag-MetaI equilibrium phosphorus distribution predicted by the Ide et al. 161
correlation for the slag chemistry and temperature data of Figures 115 and
116.
Figure 18: Metal phosphorus content predicted by the numerical process model for an 162
EAF heat, which was continuously fed with 60 tons of HBI.
Figure 19: Slag phosphorus content predicted by the numerical process model for an 162
EAF heat, which was continuously fed with 60 tons of HBI. The discrete
points are actual measured furnace slag chemistries.
Figure 120: -Slag phosphorus content predicted by-the numerical process model for an 163
EAF heat, which was continuously fed with 60 tons of HBI. The upper
continuous curve was calculated assuming all phosphorus from the HBI
initially reports to the metal phase. The lower continuous curve was
calculated assuming all phosphorus from the HBI initially entered the slag
phase. The discrete points are actual measured furnace slag chemistries.

ix
Figure 121: Metal temperature profile for the simulated EAF heats. 165
Figure 122: Slag chemistry profile for simulated EAF heats. 166
Figure 123: Equilibrium Phosphorus distribution ratio for simulated EAF heats based 166
upon the temperature and slag chemistry profiles of Figures 121 and 122.
Figure 124: Simulated EAF metal phosphorus content versus time for three different 167
mass transfer conditions.
Figure 125: Slag chemistry profile for simulated heat with an initial FeO content of 169
30wt%
Figure 126: FeO profiles for three simulated heats. 169
Figure 127: Calculated equilibrium phosphorus distribution for three heats with varying 170
initial FeO contents in the slag.
Figure 128: Predicted metal phosphorus content during the three simulated EAF heats. 171
Figure 129: Equilibrium phosphorus distribution ratios for electric furnace simulations 173
with different V-ratios.
Figure 130: Predicted metal phosphorus content during the three simulated EAF heats 173

X
 List of Tables

Chapter 3:
Table 1: Chemical Composition of Investigated Commercial Materials 33
Chapter 4
Table 2: Calculated Flux due to Oxygen Transfer 79
Chapter 5
-
Table 3: Operating Parameters of PST EAF 96
Table 4: Results of Mass Transfer Trials at PST 110
Table 5: Trial Heats Conducted at NSST 113
Table 6: Chemistries of Scrap Substitute Materials used in NSST Furnace Trials. 114
Table 7: Operating Parameters of NSST - EAF 115
Table 8: Average Cast Tons Yield and Final Furnace FeO for the Various Scrap 118
Charge Mixes of the NSST Trials
-
Table 9: Operating Parameters of BHP EAF 129
Table 10: Chemical composition of HBI used in the BHP - Sydney trials 130
Chapter 6 Table 11: Conditions of EAF V-ratio and Mass Transfer Parameter Simulations 172

xi
 List of Svmbols

Slag-metal phosphorus distribution ratio, dimensionless

wt% of species X in the slag phase
wt% of species X in the metal phase
Slag-metal reaction area, cm2 (unless otherwise noted)
Mass transfer parameter, cm3/s
Roultian activity coefficient of species a
w t %CaO/w t %Sio, (B-ratio)
Concentration of Species A, moles or grams of species A per cm3
Thickness of theoretical stagnant film near interface, cm
Diffusion Coefficient of species A, cm’/s
Activity coefficient species x in the lwt% state
Activity coefficient of species x assuming roultian standard state
mass or molar flux of species A, g/cm’s or moles/cm’s
Reaction equilibrium constant
Phosphate capacity
Mass transfer coefficent of the slag and metal respectively, c d s
Overall mass transfer coefficient, including resistances in the slag and metal, c d s
Characteristic length an element of liquid will travel at the interface before it is re-
entrained into the bulk, cm
Characteristic thickness of an element of liquid at the interface, in which mass transport
is controlled by steady state or non-steady state atomic or molecular diffusion, cm
Viscostiy
Characteristic time to establish steady state diffusion across a surface element of liquid
with characteristic thickness of L, seconds
Residence time of an element of liquid at the interface before it is re-entrained into the
bulk, seconds
Characteristic velocity of an element of liquid at the interface, c d s
Mass of slag and metal respectively, g (unless otherwise noted)
Distance from the slag metal interface, cm

xii
 Executive Summary

In modem electric arc furnace (EAF) steelmaking, ore based scrap substitute materials
are often used to control the chemistry of the steel produced and/or improve the
efficiency of the process. At the present time, the most common forms of direct reduced
scrap substitute materials are direct reduced iron (DRI) and hot briquetted Iron (Hl3I).
DRI is produced from a combination of pelletized and lump ore, which is reduced by
reformed natural gas in a shaft furnace such as the Midrex furnace. HBI is currently
produced by hot pressing either reduced DRI pellets or reduced iron ore fines into dense
metallic iron briquettes. Due to the great similarities of most direct reduced materials, this
large group of materials will be referred to generically as DRI/HBI unless greater
specification is required. Most commercially available forms of DRI/HBI contain
significantly higher levels of phosphorus as compared with scrap. As increasing
quantities of these materials are used in EAF steelmaking for the production of higher
quality steels, control of phosphorus levels in the metal will become a concern. The
goals of this research include the development of a more complete understanding of the
behavior of phosphorus in DRI/HBI during EAF steelmaking. A thorough investigation
of the kinetics and thermodynamics of phosphorus transfer in the EAF has been
conducted based upon laboratory and plant experiments and trials.
Metallographic investigation of commercial DRI/HBI materials indicates that before
melting, phosphorus in DRI/HBI is contained as a calcium phosphate phase in the
unreduced oxide “gangue” portion of the material. Laboratory experiments have shown
that phosphorus reversion from gangue to metal occurs significantly in less than one
minute for all commercial duect reduced materials tested in this study.
Based upon the work of previous researchers and the experimental results of the
present study, it was assumed that the kinetics of phosphorus transfer between slag and
metal are limited by liquid phase mass transfer of phosphorus in the slag, metal, or both
phases. An overall mass transfer coefficient, ko, was defined, which includes the effects
of mass transfer in both the slag and the metal. Fundamental laboratory lunetic
experiments indicate that either the slag-metal interfacial area, A, and/or the overall
liquid phase mass transfer coefficient, ko, change during dephosphorization experiments.

...
Xlll
Because the contributions of the reaction area and the mass transfer coefficient to the
overall rate are difficult to separate, experimental results were analyzed in terms of the
mass transfer parameter, A*ko. The liquid phase mass transfer parameter for
dephosphorization was found to range between 10" to 1x10" cm3/sfor different
experimental conditions.
Plant trials were conducted to directly evaluate the conditions of mass transfer in the
electric furnace. Controlled mass transfer experiments were conducted on a 150 ton DC
electric arc furnace. The mass transfer parameter, A*ko, for this furnace was determined
to be between 1 . 7 ~ 1 and
0 ~ 3 . 5 ~ 1 0cm3/s.
~ An additional series of plant trials were
conducted on a 150 ton AC electric furnace to examine the effects of different scrap
substitute materials upon the slag chemistry, the behavior of phosphorus in the steel, and
upon furnace yield. The data from these trials were also used to develop empirical
models for the slag chemistry and furnace temperature as functions of time during a
single heat.
The laboratory and plant data were used to develop a numerical process model to
describe phosphorus transfer in the EAF. The experimental data from this study and the
results of previous researchers were used to test the process model and reasonable
agreement was observed. The process model was used to evaluate the influence of
various EAF operating parameters upon phosphorus mass transfer between the slag and
metal.

xiv
 ACJiNO WLEDGEMENTS

The authors would like to acknowledge the AISI-DOE for financial support of this
research. In addition, the member companies of the Center for Iron and Steelmaking
Research at Carnegie Mellon University have supported this research both financially and
with guidance and information. Specifically, BHP, Midrex, and Cliffs Reduced Iron
supplied commercial material samples and invaluable information regarding commercial
direct reduced iron and hot briquetted iron products. The plant trials conducted in
conjunction with this project were among the most critical and truly interesting
components of the research. This success of this project was due in large part to the
cooperation of many individuals at North Star Steel - Texas and Pennsylvania Steel
Technologies-Bethlehem Steel regarding these plant trials. The authors would also like
to thank Dr. Sridhar Seethuaman and Dr. Alan Cramb of Carnegie Mellon University,
and Bhaskar Yalamanchilli of North Star Steel for their participation on the thesis
committee of the doctoral thesis, which resulted from this research. Their contribution to
the direction and quality of this work is greatly appreciated.

xv
 CHAPTER 1

INTRODUCTION
Very often, large amounts of ore based scrap substitute materials such as direct
reduced iron (DRI) pellets, hot briquetted iron (HBI), iron carbide, etc. are used when
producing steels with low residuals via EAF steelmalung. At the present time, the most
common forms of direct reduced scrap substitute materials are DRI and HBI. DRI is
produced from a combination of pelletized and lump ore, which is reduced in a shaft
furnace such as the Midrex furnace by reformed natural gas. The product DRI is
therefore a pellet material with a large amount of internal porosity and a very low bulk
density. HBI is currently produced via two major process routes. By one process route,
reduced DRI pellets can be fed from the reducing furnace into a briquetting machine.
Physically compacting the pellets into larger briquettes greatly decreases the porosity of
the material and increases the bulk density. HBI is also produced directly from ore fines
via processes such as Finmet and Circored in which the ore fines are reduced in a series
of fluidized bed reactors and then compacted into dense briquettes. Generally, the
chemical properties of DRI and HBI sold commercially today are very similar. Nearly all
commercial direct reduced materials contain between 80 to 90 percent metallic iron, 5 to
10 percent iron oxide, 5 to 10 percent gangue, and 0 to 4 percent carbon. The term
gangue is used to describe the unreduced components of the mineral ore, including silica,
alumina, lime, magnesia, etc. Due to the great similarities of most direct reduced
materials, this large group of materials will be referred to generally as DRIM[BI for most
of this paper unless greater specification is required.
Whereas DRI/HBI products have low contents of copper, nickel, nitrogen, sulfur,
etc., they can contain large amounts of phosphorus. Depending upon the quality of ore
used to produce the DRI/HBI, these products can contain five to ten times as much
phosphorus as scrap. Most commercially available DRI/HBI materials commonly
contain between 0.03 to 0.09% phosphorus as compared with 0.015 to 0.02 % for scrap.
DRI/HBI produced from lower grades of ore can contain as high as 0.15%P.
In normal EAF steelmaking, the phosphorus distribution ratio between slag and metal
is low as compared to a BOF. The phosphorus partition ratio (Lp) can be defined as:

1
 (%P)sZn 1
Lp = [%P]mf?A

The L, observed at tap for common EAF slag chemistries ranges from 5-50 as compared
with 50-150 in oxygen steelmaking. The observed phosphorus partition ratios for both
EAF and BOF steelmalung do not necessarily reflect the equilibrium Lp for the slag
chemistries and furnace temperatures of the given system. In both processes, the
approach toward equilibrium is limited by the kinetics of phosphorus transfer. In
addition to relatively low observed phosphorus partition ratios, lower slag volumes are
used in EAF steelmaking. Therefore, thermodynamics and a mass balance predict steels
of high phosphorus content when using DRI/HBI. However, indications from steel
producers are that phosphorus contents in actual practices are much lower than predicted.
One hypothesis has been presented in the literature as a possible explanation of the
unexpectedly low metal phosphorus contents observed in industry. Previous researchers
[1,2,3] have shown that the phosphorus contained in direct reduced iron products is
present as an oxide component gangue phase. The fact that the phosphate in the gangue
is not reduced is consistent with the thermodynamics for the conditions present during the
common direct reduction processes, Le. Midrex, Hyl 111, Fior, etc. It was theorized that
during melting, the phosphorus in the DRI/HBI gangue first physically separates to the
slag phase creating an initially high phosphorus slag. It was believed that slow lunetics
of phosphorus reversion as well as slag flushing resulted in the unexpectedly low
phosphorus levels in the steel [1,2]. However, the kinetics of phosphorus transfer
between the gangue and metal phases of DRL/HBI are unknown even on a microscopic
scale. Furthermore, the thermodynamics and kinetics of phosphorus transfer between the
bulk slag and metal phases have not been studied in detail for the electric arc furnace.
Therefore, the above hypothesis relies upon several significant assumptions regarding the
rate of phosphorus transfer within the DRI/HBI material during melting, and the rate of
phosphorus transfer between the bulk slag and steel during EAF steelmaking.

1.2 RESEARCH OBJECTIVES

2
 In order to efficiently control the phosphorus in ore based scrap substitutes in the
EAF, a more sophisticated understanding of the phenomena of gangue-metal separation
during melting and phosphorus transfer the EAF must be developed. The goals of this
research are to contribute to a more complete understanding of both the behavior of
phosphorus in DRI and the kinetics and thermodynamics of phosphorus transfer in metal
- slag (or DRI/HBI gangue) systems. The end result of this work should be a simple real-
time process model that can be used by the plant operator to design slag chemistry and
flushing practices, which optimize phosphorus retention in the slag phase for any type or
amount of DRI/HBI charge.
This study includes three major components: laboratory experiments, plant
experiments and trials, and development of a process model. The laboratory experiments
include the physical observation of the gangue - metal microstructure of several types of
DRI/HBI via optical and scanning electron microscopy. Also, the precise distribution of
phosphorus between the gangue and metal phases in DRI has been determined. The
behavior of phosphorus, and the rate of gangue - metal separation in DRI upon melting
was investigated. Also, phosphorus distribution is a strong function of the FeO content of
the slag. Therefore, the effect of carbon content and degree of metallization of DRI/HBI
upon slag make and phosphorus distribution is also examined.
The fundamental thermodynamics and lunetics of phosphorus transfer in slag -
metal systems were reevaluated through a laboratory scale study in order to develop the
appropriate rate equations. Plant experiments were conducted to examine the specific
rates of mass transfer between slag and metal for the EAF. Both the stirring conditions
and the slag - metal interfacial area present during electric furnace steelmaking are
unknown. Therefore, it is extremely difficult to transfer data and observations obtained
in the laboratory to the industrial scale. From the experimental work, a process model
was developed, which can accurately model slag-metal phosphorus transfer in the EAF.
The accuracy of the process model was tested using the plant trial data. The value of the
process model in evaluating various operating conditions for the EAF is exhibited
through several select examples.

3
 CHAPTER 2

LITERATURE REVIEW
-
2.1 The Behavior of Phosphorus in DRI/HBI A joint investigation by BHP and
CASIMA[ 11 examined the behavior of phosphorus in two types of HBI during EAF
steelmalung. The study compared plant data regarding metal phosphorus levels achieved
when using two types of HBI feed: HBI # I produced from Mt. Newman, Australia fines
via the FIOR process (0.065wt%P), and HI31 #2 produced from Venezuelan ore via the
Midrex process (0.095wt%P). The study reported that low phosphorus melts could be
obtained using large amounts of both types of feed. Other steel producers have reported
similar observations.[ 5 ] The BHP/CASIMA study included a SEM metallographic
investigation of the two types of HBI. Elemental mapping of the HI31 shown in Figures
1-3 indicates that the phosphorus is preferentially associated with the oxide phase present
in the HI31 rather than with the reduced iron.

1
I
Figure 1: Secondary electron image of HBI adapted from Trotter et al. [ 11

4
Figure 2: SEM oxygen map of HBI adapted from Trotter et al. [ 11

al. [l]

This finding is supported by chemical analysis performed by other investigators.[2]

Unfortunately, no scale was provided with the original pictures to indicate the relative
sizes of the imaged gangue features. However, based upon metallographic results of the
present study, which will be presented later in this paper, these features are most likely on
the order of 50 to 150 microns in size.

5
 The BHP/CASIMA investigators concluded that the phosphorus present in the gangue
initially enters the slag phase during melting. Slow kinetics of phosphorus reversion and
slag flushing practices result in the observed unexpected low phosphorus levels in the
metal. This basic theory represents the original hypothesis of the present work.
A much earlier study of phosphorus in direct reduced sponge iron by Queens [3],
also concluded that phosphorus is not reduced during direct reduction. In the study by
Queens, an electron microscope investigation of the gangue phase identified 20 to 30
micron particles containing 21%wt P - 37%wt%Ca. A component of the study includes a
series of 2 ton bulk melting experiments with several grades of sponge iron containing up
to 0.07wt%P. Queens concluded that with special fluxing modifications, which ensured a
basic, oxidizing slag, the high phosphorus material could be used to produce low
phosphorus steel. However, this researcher made no specific comments on the kinetics of
phosphorus transfer either to or from the slag phase.
Between 1965 and 1975, Lukens Steel conducted an extensive series of trials with a
number of scrap substitute materials, including hhdrex DRI, Hyl DRI, Fior HBI, and
various synthetic materials. Portions of the findings of that study were published in the
open literature[5, 6,7]. Review of internal literature describing this study, which was
shared by Lukens with the present author, gives a much better notion of the true extent of
the research conducted by Lukens. In that study, more than 800 trials of both batch and
continuously fed electric furnace heats were conducted.@ - 101 The effects of scrap
substitute usage were evaluated with respect to a wide range of operating parameters. Of
special interest to the present study were trial heats conducted with varying amounts of
HyLSA prereduced pellets, which contained 0.362wt% phosphorus.[8] Ten trial heats
were conducted in a 165 ton AC-EAF during which 17 tons of HyLSA pellets were
continuously fed during refining. Incredibly, only one of the 10 heats was tapped with an
unacceptably high phosphorus content of 0.022wt%P.
In addition to the heats with HyLSA pellets, a larger number of trials were conducted
with Fior HBI produced by Esso from Venezuelan iron ore.[9] The reduced Fior material
contained 0.14wt% phosphorus. For most of the trial heats, approximately 25 tons of
briquettes were continuously fed into the furnace during the refining period. Chemical
analysis of slag samples from before and after continuous feeding showed that the slag
phosphorus increased from 0.294 wt% to 0.618 wt%. Despite the increase in slag

6
phosphorus, all heats were tapped at, or below O.Olwt% P in the steel. In 24 trials with a
third material, 25 tons of HBI containing 0.096wt%P were continuously fed into the 165
ton furnace over a period of 35 minutes.[lO] Metal samples were taken from the furnace
immediately before and after continuous feeding. Comparison of the average before and
after chemistries for the 24 heats revealed a 0.001% decrease in the metal phosphorus
after the continuous feeding period. Despite these observations, it is unlikely that the low
metal phosphorus contents measured by Lukens were the result of slow kinetics.
Continuous feeding of a dense carbon and iron oxide containing material like the Fior
HBI will greatly enhance slag metal mixing. Also, because these trials were conducted
prior to the days of powerful UHP transformers, the tap to tap time on most of the trial
heats exceeded 3 hours. It is far more likely that the observed low metal phosphorus
contents were the result of favorable equilibrium conditions. With a low to medium
powered furnace (by today’s standard) producing steel in an ingot cast shop, it is very
likely that most heats were tapped at much lower temperatures and lower FeO contents
than are common today. For these conditions, is very likely that the high observed
phosphorus partition ratios of Lukens trials are near or below the equilibrium distribution
ratio. In such a case, reversion of phosphorus from the slag to the metal will not occur.

2.2 Slag-Metal Phosphorus Equilibrium - Although equilibrium thermodynamics is

not the focus of this study, it is the deviation from the equilibrium phosphorus
distribution between slag and metal that represents the driving force for any phosphorus
transfer reaction. Therefore, a thorough understanding of the relevant equilibrium
thermodynamics is necessary for an accurate evaluation of the kinetics of phosphorus
transfer. The review that follows represents only a brief introduction to a long and
extensive history of thermodynamic studies [refs. 11 - 331.
Winkler and Chipman [27], Suito et al. [18, 191 and others have carried out basic
thermodynamic studies of phosphorus distributions over wide ranges of temperature and
slag chemistry. Fuwa [l I] and Wijk [26] provided overviews of the phosphorus problem
in the steel industry and Stubbles [17] examined industrial practice in some detail.
Turkdogan [20], Suito et al. [ 191, Zhang et al. [28], Ide and Fruehan [29] and others have
provided interpretations of available data. The accurate prediction of the phosphorus
partition ratio based upon readily accessible parameters such as temperature and slag

7
composition represents an extremely useful, yet difficult capability. The studies
conducted so far have covered the range from very theoretical evaluations to very
empirical modeling of experimental and industrial data. Most thermodynamic studies to
date have attempted to evaluate phosphorus in the slag - metal system using either
classical molecular thermodynamics or an approach that recognizes the ionic nature of
molten slags.
Some of the earliest studies described the phosphorus reaction as:

Assuming low solute concentrations in the metal phase, the equilibrium constant for this
reaction can be written:

In quantifying phosphorus equilibrium many researchers have used the concept of a

phosphate capacity which contains experimentally accessible parameters. In the case of
equation 3, many researchers have described the phosphate capacity as:

Based upon laboratory data, Balajiva et al. [24] established the following empirical
relationship between Kp and weight percent CaO in the slag:

logK, = 11.801og(%CaO)-C 5

Where C is an experimentally determined, temperature dependant variable parameter.

Equation [5]can be expanded such that:

8
 6

Where N, is the mole fraction of species i and yr is the Roultian activity coefficient.
Comparison of equations [4] and [6] suggests that the activity coefficient terms in
equation [6] are the source of the compositional dependence of Kp.
Later studies [21] attempted to directly evaluate these activity coefficient terms
assuming an imaginary pure liquid P,O, as a standard state. Correlations have been
developed for Kp by several researchers which include the compositional dependence of
the activity coefficient terms on basic slag components, i.e. CaO, CaF,, FeO, MgO, MnO,
etc.
More recently, several attempts have been made to reevaluate phosphorus equilibrium,
recognizing the true ionic nature of molten slags. [19,23, 281 The most recent literature
is in agreement that for basic slags, the slag metal phosphorus reaction can be written as
equation [2]:

5 3
[P]+,[O]+,(O -2
)= (Poi3)
7
L L

The equilibrium constant (Kp) for this reaction can be expressed as:

ap o i 3
K p = a p * a 512 312
o *ao-2
8

where ai is the activity of species i. An important difference should be noted between

equations 8 and 4. Due to the form in which the phosphorus reaction is expressed in
equation 2, the activity (or weight percent) of phosphorus in the metal appears as a
squared term in equation 4. However, due to the form of the phosphorus reaction in
equation 7, the activity of phosphorus appears to the first power in equation 8. At low
metal phosphorus contents, this represents a significant difference in the equilibrium
constant.
If this expression is expanded:

9
WhereA represents the one-weight percent standard state activity coefficient of species i.
When rearranged in the following manner:

N
poi3
[%PI [%O]
-
SI2 -
fP fO5l2
Y Po43
312
No-2
3/2
Yo-2
.- 1
K“4
10

Again, the terms that are most accessible experimentally are placed on the left hand side
of the equation, and the terms remaining on the right represent a formidable collection of
largely unknown quantities.
Several researchers have attempted to apply a formalism for describing ionic melts,
established by Flood, Forland, and Grjotheim [30, 311 to the phosphorus system. By this
method, an ideal anionic equilibrium constant for equation [7] is defined as:

-
-
Np o i 3
11
Kp04 [%PI [%0]2.5
e NkJ2

It was shown by Flood and Grjotheim that:

12
i

Where N’,i is the equivalent cationic fraction of the cation Mi (Le. Ca”, Fe”, MgZ*,
Mn2+,etc.) and KMiis the equilibrium constant for the formation of pure MJPO,), by the
reactions:

10
 2[ PI + 5[01 + 3 Ca 0 = Ca, (PO,)
2[ PI + 5[01 + 3 ~ g =oMg3 (PO&
2 [ P ] + 5 [ 0 ] + 3 M n 0 = M n 3 ( P 0 4- ) 3
...
2[P1+5[01 + 3 M i 0 = M i 3 ( P 0 4 ) 2
In equation [12], f(y) represents the sum of all the activity coefficient terms. In order to
convert K,, to a phosphorus distribution, equations [ l l ] and [12] can be combined and
rearranged such that:

The term N’M,can be calculated from the slag chemistry. The KMiterms can be calculated
from the experimentally determined free energies of formation for the different species
present. In most cases it can be assumed that the dissolved oxygen content is controlled
by the reaction:
Fe+[O]=(FeO)
Therefore, [O] can be correlated to the activity of FeO in the slag. Some researchers have
developed empirical relations that describe No-2as a function of Si02. [28] Other
correlations have been developed which relate No-2to optical bacisity or other
compositional parameters.
The f(y) term is unknown and can only be determined experimentally. Several
researchers have also attempted to develop correlations that relate this term to some
compositionally dependent term such as optical basisity. The f(y) term can be defined as
follows:

14

11
Where y,‘refers to the activity coefficient of phosphate or free oxygen ions with respect
to each cation, i. From equation [13] it is apparent that f(y) is very similar to the ynOj/
(yFeO)’term which was evaluated in earlier studies. Belton et al. [32] have shown an
apparent henrian behavior of the phosphate ion in slags with basicities >1.3. These
researchers were then able to develop a correlation that expressed f(y) as a function of
concentrations of ferrate ion and free oxygen ion and temperature. The researchers
demonstrated that this correlation agreed fairly well with several independent data sets,
though only for a limited range of slag basicities.
All correlations to date describing equilibrium phosphorus distribution have relied to
some extent upon empirical evaluation of new or existing data sets. Systematic errors or
differences within these data sets, as well as fundamental differences in the approaches of
the various researchers, have resulted in significant quantitative disagreement between
these different correlations. Despite this quantitative disagreement within the literature,
there is general agreement upon which factors most strongly influence phosphorus
equilibrium, namely:
Increasing temperature reduces Lp.
Highly basic slags increase Lp.
Oxidizing conditions increase Lp.
Several correlations developed by different researchers for the prediction of phosphorus
distribution are listed in Appendix A. Figures 4 and 5 highlight both the similar trends
and numerical differences for four different correlations.

12
 1000
900
800
700
600
n
500
400
300
200
1 00
0
1500 1550 1600 1650 1700 1750 1800
Temperature, "C
Figure 4: Equilibrium phosphorus partition ratio as a function of temperature as predicted by three
correlations assuming a slag chemistry of: 25%FeO - 7%Mg0 -
46%Ca0 - 22%Si02

600

500

400

300

200

100

0
0 10 20 30 40 50
wt%FeO
Figure 5 : Equilibrium phosphorus partition ratio as a function of FeO as predicted by three
correlations assuming T = 1600°C and constant basicity, CaO/ SiOz= 2.1.
Different correlations more accurately predict phosphorus equilibrium at different slag
chemistries and temperatures. Therefore, when attempting to predict the equilibrium
phosphorus distribution for a given slag - metal system, it is imperative to select the
model which best describes the system of interest. A recent study by Ide and Fruehan
[29] reevaluated the phosphorus equilibrium for the slag chemistries and temperatures
directly applicable to this study. In that study, the basic framework of the correlation of

13
Suito et al. was used, and focused experiments were conducted to evaluate and refine
some of the coefficients of the equation. The basic trends of the improved correlation
follow those of Suito et al. The Ide et al. correlation is also listed in Appendix A.

2.3 Kinetics of Phosphorus Transfer- As with the results of thermodynamic studies,

there is also some disagreement amongst the results of previous kinetic studies. This is
due in part, to the fact that different equilibrium phosphorus distribution models have
been used in these studies. As mentioned above, most kinetic studies to date have
examined the rate of oxidation of phosphorus from metal to slag (dephosphorization).
Mori et al. [34] and Motlagh [35]specifically looked at the kinetics of phosphorus
reduction from slag to metal (phosphorus reversion or rephosphorization). Most previous
studies of phosphorus transfer kinetics have attempted to identify the rate limiting step of
this slag-metal reaction. All studies reviewed by the present author either concluded or
assumed that the kinetics of the phosphorus reaction are limited by the rate of mass
transfer of phosphorus from the bulk of either the slag or metal to the slag-metal
interface. Such reactions are said to be controlled by liquid phase mass transfer in the
slag, metal or both phases. The basic assumptions of this model are:
The chemical reaction at the slag metal interface is rapid, thus the interface is
always at equilibrium
The driving force for mass transfer is the difference in the concentration of
phosphorus in the bulk liquid phase and the equilibrium concentration at the
interface
The rates of physical transport of phosphorus atoms in the metal or phosphate
ions in the slag are limited by some resistance
The liquid phase mass transfer model is typically expressed as:
15

Where J, is the flux of species A, k, is the mass transfer coefficient, and C', and C, are
the concentrations of A at the interface and bulk respectively. Because various units of
both flux and concentration are commonly reported in the literature, the mass transfer
coefficient is often reported in inconsistent units. This confusion can be easily avoided
by adding the appropriate terns to the flux equation, which are required for unit

14
conversion. However, these terms are often rolled into the mass transfer coefficient. In
the present paper, ‘mass transfer coefficients’ will always be defined in units of c d s .
When equation 15 is rearranged, the simplest definition of the mass transfer coefficient is
obtained:
J, 16
k --
A - (CA -CAI

This definition can be useful from an engineering stand-point, in the description of

complex systems involving turbulent fluid flow. However, the nuances of the
fundamental phenomena, which are described by the mass transfer coefficient is a topic
of some debate within the literature. Several different physical and mathematical
justifications of the mass transfer coefficient have been developed in the literature. A
brief review of these arguments will be presented in the following.
One of the oldest theoretical descriptions of the mass transfer coefficient is often
attributed to Lewis and Whitman [36]. This model, commonly referred to as the film
theory, suggests that all of the resistance to mass transfer in a fluid phase is limited to a
thin unmixed film of fluid near the fluid interface. This model is shown schematically in
Figure 6.

Phase 1
Interface

I
I

Figure 6: Schematic representation of the film model

The film region is bounded on one side by the interface and on the other by a completely
mixed bulk region. Figure 6 fully depicts only one side of the two phase interface. In the
case a slag-metal interface, films would exist on both sides of the interface in the slag and
the metal. Mass transfer through the stagnant film region occurs purely by atomic or
molecular diffusion. Diffusive flux can be described by Fick’s first law, which can be
written as:
17

15
Where J, is again the flux of species A, D, is the diffusion coefficient of species A, and
*C,/*x is the concentration gradient of species A in the x direction. If steady state
conditions are initially assumed, the film theory suggests that the entire concentration
difference between the interface and the bulk fluid occurs across the length of the film
thickness, *. By this argument the flux equation can be re-written as:
18

By rearranging equation 18 and combining it with our definition of the mass transfer
coefficient in equation 16, we obtain:

k, =$D 19

Equation 19 suggests that if the mass transfer coefficient and diffusion coefficient of a
species are known, a film thickness can be calculated, or if the film thickness and
diffusion coefficient are known, the mass transfer coefficient can be calculated.
However, as many researchers have pointed out, the existence of a truly stagnant film
near the interface of a turbulent liquid is not likely to exist for most conditions.
Regardless of the accuracy of the physical model used to justify the mathematics, the film
theory suggests that the mass transfer coefficient is directly proportional to the diffusion
coefficient.
Although it is not relevant to the present study, it is worth pointing out that under
conditions of fluid flow past a solid, a laminar momentum boundary layer is likely to
exist near the interface of the solid and fluid. [37] If the concentration boundary layer of
Figure 6 is contained entirely within the laminar momentum boundary layer, it is
probable that atomic or molecular diffusion would represent the dominant mechanism of
mass flux normal to the solid-fluid interface. Although not entirely accurate, the basic
model of Figure 6 is relevant for mass transfer across a solid-fluid interface. A
significant effort has been made to develop correlations for the mass transfer coefficient
based upon different combinations of dimensionless groups for various geometries of
solid-fluid interfaces. Examples of such correlations can be found in many heat and mass
transfer reference books. [37,3 8,39,401

16
 In contrast to a fluid-solid interface, a liquid-liquid interface such as a slag-metal
interface, offers significantly less resistance to shear stresses. For this reason, momentum
boundary layers near fluid-fluid interfaces are much less stable than near solid-fluid
interfaces, and it is much easier for eddy currents to penetrate right up to the interface.
l3gbie [41] suggested that turbulent eddy currents continuously deliver new elements or
packets of liquid from the well mixed bulk to the interface. The residence time, 8, of a
liquid packet at the interface can be viewed in terms of the interfacial velocity, vi,of the
packet and a characteristic length, 1 , i t travels across the interface before it is re-
entrained into the bulk.
20

Because a given packet of liquid is only at the interface a short time, 9, before it is swept
back into the bulk, the packet is subject to unsteady-state diffusion or penetration of
species A while it is at the interface. This theory is generally referred to as the
penetration theory and is represented schematically in Figure 7.

Phase 1 Phase 2
CA
CiA

CiA

Figure 7: Schematic representation of the motion of a packet or element of fluid to and away from the
interface according to the penetration theory.

A key assumption of the penetration theory is that the time a packet of fluid is at the
interface, 8, is shorter than the time required to establish steady state diffusion across
thickness of that packet. The solution for non-steady state diffusion into a semi-infinite
body can be found in most mass transport text books. [37,39] For the present case, non-

17
steady state diffusion into the liquid packet yields the following solution for the
concentration of A as a function of time and distance from the interface:
21

Using Fick’s first law to solve for the flux at interface (x=O):
22

The instantaneous flux at the interface is given by:

23

The average flux across the interface during the residence time of a liquid packet at the
interface can be calculated by integrating equation 23 with respect to time from t=O to
t=8 and dividing by 8:

24

25

Therefore, in contrast to the film theory, which predicts a linear dependence of the mass
transfer coefficient with the diffusion coefficient, the penetration theory predicts the mass
transfer coefficient to be equal to:
26

Like the film thickness, *, in the film theory, the residence time, 8, of the penetration
theory is a function of the stirring conditions and the physical properties of the interface,
and is extremely difficult to quantify experimentally.
A variation of the penetration theory was introduced by Danckwertz.[42] This
model, often termed the surface renewal model, includes a surface age distribution rather
than a constant residence time, 8, for all surface elements. The mathematics of this
model are more complex than the basic penetration model as it includes a probability

18
distribution function for the residence time rather than a simple constant. The surface
renewal model results in the same proportionality of the mass transfer coefficient with the
square root of the diffusion coefficient.
Experimental results from various fluid systems and different stirring conditions
have indicated that the mass transfer coefficient is proportional to D” where n ranges
between 0.5 and 1.[37] This suggests that, at least mathematically, the film theory and
penetration/surface renewal theories may represent the boundaries of a continuous
relationship between D and k. Toor and Marchello [42] proposed the film-penetration
model, which approaches the film and penetration models as limiting cases of a single
phenomenon. The film-penetration model suggests that surface elements of an average
thickness, L, are circulating between the interface and bulk fluid with some average
residence time at the interface. Within the thickness, L, of the surface elements, mass
transport normal to the interface is dominated by atomic or molecular diffusion. At
distances greater than L form the interface, mass transfer is dominated by turbulent
transfer and the concentration is approximately constant.
Toor and Marchello [42] suggest that elements of liquid at the interface are different
from those in the bulk liquid as they are more stable due to the effects of interfacial
tension. Because the rotational and translational motion of interfacial elements are more
restricted relative to bulk elements, a concentration gradient and diffusional mass transfer
can be established within the element. This suggests that the residence time of an
element of liquid at the interface is somehow proportional to the interfacial tension;
however, the authors make no speculation regarding the specifics of that relationship.
The time, 2, required to establish steady state diffusion within an interfacial element
of thickness L can be roughly approximated as:
L2 27
r=-
D
The film-penetration theory recognizes that the key difference between the film and
penetration models lies in the relationship between the average residence time of the
surface elements, 0, and the time required to establish steady state diffusion, z. When the
residence time, 0, is much greater than 7, diffusion across the interfacial elements
proceeds in a steady state manner, and the mass transfer coefficient is proportional to the
diffusion coefficient. When the residence time is much less than z, diffusion into the

19
interfacial element proceeds in a non-steady state manner before the element is swept into
the bulk. In the latter case, the mass transfer coefficient is proportional to the square root
of the diffusion coefficient. Assuming a distribution of residence times about an average,
when the average residence time and the time to establish steady state are similar, the
mass transfer coefficient will be proportional to D”,where n is between 0.5 and 1. Toor
and Marchello offer a continuous graphical solution of the proportionality of the flux to
the diffusion coefficient as a function of 0 /zwith the solutions of the film and
penetration theories as asymptotes.
The film-penetration theory offers a very interesting framework from a conceptional
standpoint. However, the inaccessibility of 0 and z,limit the usefulness of this model in
generating meaningful calculations. The conclusion one must draw from the literature, is
that the precise physical and mathematical description of the mass transfer coefficient for
liquid-liquid reactions remains somewhat unclear.
As mentioned above, Mori et al. [34] conducted an extensive fundamental study of
slag-metal phosphorus transfer kinetics. These researchers employed a similar
experimental approach as proposed in the present study to produce fairly internally
consistent data. Figure 8 shows the experimental apparatus used by Mori et al. in that
study.

20
 I M t
A: Sanipliiig hole
€3: U’dtt.-codedCIIP
C : Silica tube
D: Cas inlct
E: Craphitc ring
F: Protcctioti crucible
C: MgO crucible
H: Slag
I: XIetEII
J: Supporter
K. Thermocouple
L: Casoutlet
31: Slag holder
Y: Induction coil

Figure 8: Experimental apparatus used for Phosphorus kinetic study by Mori et al. [34]

’ In their study, the researchers melted approximately 250 g of iron or iron-phosphorus

alloy in a magnesia crucible in an induction furnace. Once the experimental temperature
was reached, approximately 40 g of prefused synthetic slag contained in a pure iron slag
holder (shown as item M in Figure 8) was lowered to the metal surface. The experiment
was judged to begin when the slag holder melted and the slag flowed out onto the liquid
metal surface. The slag chemistries and experimental temperatures studied by Mori et al.
are listed in Appendix B. Samples were taken from the metal and slag repeatedly once
the slag and metal were in contact. Metal samples were analyzed for oxygen and
phosphorus, and slag samples were analyzed for CaO, SiO,, FeO, MgO, and P,O,.
In the discussion of their results, these researchers noted some difficulty in choosing
an adequate thermodynamic model that was consistent with their results. Without
extensive justification, Mori et al. [34] assumed that liquid phase mass transfer in either

21
the slag or metal was controlling the kinetics of phosphorus transfer. The researchers
developed the flux equations for phosphorus in the slag and metal assuming constant
mass transfer coefficients in the slag and metal, a constant area of slag metal contact, and
with concentrations in weight percent. The flux equations for phosphorus in the metal
and the slag are given as equations 28 and 29 respectively:

d[%P] Ap k 28
-- rn [[%Pi]-[%PI]
at Wm

29

Where, A is the slag - metal surface area,

k, and k, are the mass transfer coefficients in the slag and metal,
p, and p, are the slag and metal densities,
W, and W, are the weights of slag and metal, and
(%I") and [%PIare the phosphorus contents of the slag and metal at the sladmetal
interface .
Mori et al. reasoned that due to the conditions of their experimental apparatus, the
kinetics of dephosphorization would be controlled entirely by mass transfer in the metal.
Using a numerical solution to equation 28 to analyze their experimental results, the
researchers calculated metal liquid mass transfer rates in the range of 0.33 x l o 2to 1.09
x lo-*c d s e c for the conditions studied. In Figures 9 and 10, Data from this study by
Mori et al. [34] are replotted. In addition to the raw data, the solid lines shown were
calculated based upon the liquid phase mass transfer model with the conditions stated
above, i.e. constant km,constant A, etc.

22
 0.4
T = 1680°C
0.35
caclulated
I
-
I
0.3 A T = 1640°C

T = 1570°C I
0.25
n o
Y
8
u
0.2
B 0.15

0.1

0.05

0
0 5 10 15 20
Time (minutes)

Figure 9: Dephosphorization kinetic data adapted from Mori et al. [34] plotted with calculated curves for
best fit overall mass transfer coefficients, ko. The thee data sets shown were collected for approximately
the same slag chemistry at different experimental temperatures.

0a45
0.4
0.35
a ---_
A
slag 2
calculatec
slag 3
calculatec
0.3 o slag 4
n
u
P 0.25 ------ calculatec
8
0.2
3
c,

0.15

0 1 I 1 I

0 5 10 15 20
Time (minutes)
Figure 10: Dephosphorization kinetic data adapted from Mori et al. [34] plotted with calculated curves for
best fit overall mass transfer coefficients, ko. The thee data sets shown were collected for different slag
chemistries at the same experimental temperature(1640°C).

23
In Figures 9 and 10, the calculated continuous curves were generated assuming a constant
driving force (e.g. constant Lp), a constant overall mass transfer coefficient, and constant
slag metal area. For these calculations, it was assumed that the final slag and metal
chemistries provided in the paper by Mori et al., and listed in Appendix B of this paper,
represented the equilibrium phosphorus partition, Lp. Clearly, however, there are
indications in these and other data presented by Mori et al. that at least some of the
experiments had not yet reached equilibrium by the final sample.
In both Figures 9 and 10, the experimental data exhibit an initial fast rate, which is
faster than the calculated curve at very short experimental times. After 3 to 5 minutes,
the transfer rate appears to slow down, becoming slower than the calculated curve.
In Figures 11 and 12, rephosphorization data from the same study are replotted.
Again the continuous curves represent the calculated metal phosphorus versus time
assuming a liquid phase mass transfer model and a best fit overall mass transfer
coefficient.

0.16 Slag 5, 1600°C

calculated
-
0.14 ------ ko = 0.009cm/s
------ ----
0 Slag6, 1615°C
calculated
0.12 - 0
Slag 7, 1595°C
+ X
/ / /

0 ------ calcula
0.1
E
0.08
5 0.06
0.04

0.02

0 5 10 15 20 25
Time (minutes)
Figure 11: rephosphorization kinetic data adapted from Mori et al. [34] plotted with calculated curves for
best fit overall mass transfer coefficients, ko. The thee data sets shown were collected for different slag
chemistries at different experimental temperatures.

24
 ~

calculated
0.08 - ------ ko = 0.0075cm/s
A Slag 9,1650°C
0.07 - / /
5- ---- calculated
A'

0.06
),009cm/s
0.05

0.04
I /

0.03

0.02

0.01

o ! I

0 5 10 15 20 25
Time (minutes)
Figure 12: Rephosphorization kinetic data adapted from Mori et al. [34] plotted with calculated curves for
best fit overall mass transfer coefficients, ko. The thee data sets shown were collected for the same slag
chemistry at different experimental temperatures.

The general trends observed in the dephosphorization experiments regarding an initial

fast rate followed by a slower rate, were also observed in the rephosphorization
experiments. Mori et al. identified an apparent dependence of the mass transfer
coefficient upon the equilibrium phosphorus distribution ratio, e.g. the driving force for
mass transfer. However, none of the theoretical descriptions of the mass transfer
coefficient presented earlier in this section even suggest a dependence of k upon
concentration or driving force parameters. Despite these very interesting trends, Mori et
al. did not offer an extensive discussion of these points in their analysis. The above
mentioned trends in the data of Mori et al. [34] are also reflected in the data of the
laboratory lunetic study presented later in this paper. Therefore, the data of these
experiments by Mori et al. will be revisited in the discussion of the present data.
In the study by Mori et al. [34], most dephosphorization rates were observed to be
fairly rapid, e.g. equilibrium was obtained within 15 minutes. Within the experimental
technique used by these researchers a significant transient period existed at the start of
each experiment. During dephosphorization and rephosphorization experiments, the
experiment was judged to start when the slag holder, shown in Figure 8, melted,
presumably near 1537"C, and the slag flowed out onto the liquid iron or Fe-P alloy. For

25
experiments at higher temperatures, a finite time must have passed before the slag
reached the experimental temperature. This would result in an initially higher Lp, and
hence driving force, which would decrease until the target temperature was obtained.
The experiments conducted by Mori et al. were extremely well executed; however, the
experimental technique employed by these researchers may have introduced some
unnecessary complexity to an already difficult and dynamic system.
Several studies have been reported recently investigating the kinetics of phosphorus
transfer between slag and either hot metal or high carbon iron. [43 - 471 Several attempts
have been made to predict the behavior of highly complex systems involving multiple
simultaneous multi-phase reactions. [44,45] Several studies have been conducted where
these models have been applied to experimental data collected for systems of various
sizes. [43, 461 Generally, such multi-component models require the simultaneous fitting
for a large number of experimental parameters. This can be very useful for modeling
plant data. Also, if the fundamental mechanisms and rate constants for the separate
reactions are known with a high degree of confidence, multicomponant lunetic models
are a useful framework for evaluating overall rate constants in complex systems. It is the
opinion of this researcher that the kinetics of phosphorus transfer are not well understood
for even very simple and well-defined systems. Hot metal or high carbon phosphorus
oxidation experiments can be very difficult to evaluate since the system never reaches
true equilibrium. Most importantly, huge gradients in oxygen potential between the slag
and metal, and a dynamic slag composition make it very difficult to accurately define the
driving force for phosphorus transfer. Several dephosphorization studies have estimated
the equilibrium phosphorus partition to be the final or maximum partition achieved
during the study. Other studies have attempted to correlate the oxygen potential at the
interface to the Fe2'/Fe3' ratio in highly oxidized slags. Both of these estimations require
significant assumptions and may not reflect the true driving force for dephosphorization.
In a recent study, Wei et al. [44] examined the effects of stirring rate, slag chemistry,
and experimental atmosphere on the rate of phosphorus transfer between slag and high
carbon iron. In this study, both dephosphorization and rephosphorization lunetics were
studied for an A1,0, - CaO - SiO, - Li,O - Fe,O slag / Fe - 4.4%C system at 1300°C
under an Ar-0, atmosphere. The system was mechanically stirred by an eccentric
alumina stirrer at rotation speeds from 0 to 300 rpm.

26
 The researchers attempted to develop a model to describe the various rates of the
different simultaneous reactions of the system. In this model the researchers linked the
multiple simultaneous reactions to the oxygen potential at the slag - metal interface. By
assuming mass flux balance at the gas - slag and slag - metal interfaces, the researchers
numerically solved the flux equations for the various components for short increments of
time. For the calculations within the model, it was assumed that the phosphorus partition
ratio, thus the driving force for phosphorus transfer, only varied with oxygen activity at
the slag - metal interFace. Although this allowed the authors to directly link the
phosphorus reaction to the Fez' me3' ratio in the bulk slag, it neglects all other
compositional variation of Lp. Since up to fifty percent of the starting FeO content of the
slag was consumed during these experiments, the above assumption probably introduced
some degree of error into the analysis. A mixed slag and metal liquid phase mass transfer
controlled model was developed to describe the kinetics of phosphorus transfer. A
numerical value was assigned to the overall mass transfer coefficient for phosphorus.
However, the system evaluated by Wei et al. 1441 was extremely complex and dynamic.
Consequently, the overall model contained many variable parameters, which may have
distorted the true mass transfer kmetics of phosphorus.
In the analysis of their results, Wei et al. [44] noted that the mechanical stirring rate
had a negligible effect on the apparent rate of dephosphorization. It was proposed that
this was the result of a balanced effect of a simultaneous increase in the metal phase mass
transfer coefficient, ko, and a decease in Lp. Lp is decreased by an increased rate of
decarborization resulting in a decrease in the oxygen activity at the interface. However,
the unique balancing of these two terms seems unlikely. Two other more likely
explanations include:
Liquid phase mass transfer can not be accurately described as a constant mass transfer
coefficient in the highly complex system studied by Wei et al. E371
Stirring in the slag phase, resulting from CO evolution due to the reaction of FeO in
the slag and carbon in the metal, was dominating the mass transfer conditions for the
system. That is to say that the additional stining energy provided by the mechanical
stirring was small relative to the stirring energy of the evolving gas.
If phosphorus transfer is controlled by slag phase mass transfer, and stirring conditions in
the slag are controlled by the CO evolution mentioned above, increasing the mechanical

27
stirring rate would have only a small effect on the rate of phosphorus transfer.
Furthermore, due to the large amount of mechanical and gas stirring during the
experiments, it is unlikely that a constant slag-metal area based upon the crucible cross-
sectional area accurately reflects the true slag - metal reaction area. The effects of
mechanical stirring on the amount of slag metal mixing and rate of separation are
uncertain at best. Despite this uncertainty, the apparent rate of phosphorus transfer
observed by Wei et al. [44] is relatively fast. Almost all of the experimental data
presented in that study appeared to reach a steady state phosphorus level within 5 to 10
minutes. Many of the observations reported by Wei et al. were admittedly the effect of
various experimental parameters on the driving force for phosphorus transfer. Therefore,
it is difficult to draw solid conclusions regarding the fundamental mechanism and rate of
phosphorus transfer.
In a very recent study, Monaghan et al. [47] attempted to evaluate the kinetics of
dephosphorization of carbon saturated iron. In these experiments, a highly oxidizing slag
(up to 66wt% Fe,O,) was used. Like Wei et al. [44], these researchers assumed that
oxygen activity at the slag metal interface dominated the driving force for
dephosphorization. It was also assumed that the slag phase controlled the oxygen activity
at the interface. This latter assumption, made by both Wei et al. and Monaghan et al.
implies that either mass transfer of carbon in the metal or chemical lunetics of the
reaction:

are slow. Subscripts i indicate oxygen and carbon at the interface. The fact that any
dephosphorization occurs in these experiments validates this assumption to some extent,
as the equilibrium oxygen activity in carbon saturated iron is very very low.
Monaghan et al. [47] attempted to evaluate phosphorus transfer kinetics by
separating driving force effects from their experimental data. By assuming a liquid phase
mass transfer controlled model, the researchers attempted to isolate driving force effects.
An unfortunate limitation of the approached used by Monaghan et al. [47] is the inability
to define the phosphorus partition ratio, thus the driving force, with a high degree of
confidence. Despite this fact, several interesting observations were made. The
researchers reported with some confidence that mass transfer in the slag phase controlled

28
the rate of dephosphorization. In conjunction with this conclusion, a correlation was
demonstrated between slag viscosity and the apparent rate of mass transfer. Three slag
components, Fe,O,, CaF2,and CaCI2,were varied to produce slags with a range of
viscosities. In general, the rate of mass transfer decreased with increasing slag viscosity.
Monaghan et al. assumed a penetration theory description of the slag mass transfer
coefficient as described above:

k , = 2 n eF
30

Where ks is the slag mass transfer coefficient, D is the diffusivity of phosphorus, and 0 is
the contact time of an element of slag with the slag - metal interface. Also, since
diffusivity and viscosity, q, are inversely related:

1 31
D=-

It can be said that:

32

In the analysis of their results Monaghan et al. [47] included a plot of k, versus ( 1 1 ~ ) ' ~
adapted below as Figure 13:

29
 n
0.12 'r
CaF2
0.1
Fe203
A CaC12
0.08

0.06

0.04

0.02

0
1 5

Figure 13: Calculated ks versus Adapted from Monaghan et al. [47].

Despite the apparent linear behavior expressed in this plot when only one slag component
is varied, there is no overall trend relating slag viscosity and the calculated slag mass
transfer coefficient. Furthermore, slag mass transfer coefficients on the order of 0.1 c d s
are extremely unusual. Based on these observations it is unlikely that the kSterm
presented in Figure 13 is a true mass transfer coefficient containing only hnetic
information. Several other parameters can also be correlated to slag viscosity, which are
also likely to influence the plot in Figure 13. These include:
1. Slag foamability
2. Susceptibility to slag-metal mixing, and rate of droplet separation.
3. Slag composition and therefore Lp, i.e. the driving force.
Imprecise knowledge of the reaction area due to slag metal mixing, or uncertainty
regarding the driving force for the reaction would significantly affect the calculated term
k,. Furthermore, little is known about the mass transfer of reacting species within a
foamy slag. The trends reported by Monaghan et al. [47] are interesting, though they
must be viewed with a high degree of skepticism. Investigation of a simpler system is
necessary to truly understand their observations.
One important point of the phosphorus studies conducted on high carbon systems
should be restated, as it is particularly relevant to process model developed later in the
paper. The fact that dephosphorization of high carbon iron via an oxidizing slag can

30
occur suggests that the oxygen potential of the slag, not the metal dominates the oxygen
potential at the interface. This suggests that in evaluating the equilibrium phosphorus
distribution for slag - steel systems, the equilibrium can be calculated based entirely on
the slag chemistry and temperature without detailed consideration of the bulk metal
oxygen potential.

31
 Chapter 3

Experimental Procedures
3.1 Metallographic Observations
It is an objective of this study to investigate the microstructures of several types of
direct reduced iron products from various sources. A metallographic investigation has
been conducted via optical and scanning electron microscopy examining several types of
direct reduced products. These materials are listed in Table 1.
Table 1: Chemical Composition of Investigated Commercial Materials
Material Method of Fet FeO C SiO, A1,0, CaO P
Reduction
Midrex DRI Midrex 93.13 5.87 1.89 1.92 --- 1.59 0.028
Midrex HBI Midrex 92 10 c 0.5 1.8 --- --- 0.06
BHP - HBI Finmet 92 8 1.3 1.6 --- 0.1 0.045
Mobile DRI Midrex 91.2 7 2.43 1.14 0.36 0.3 0.037
Circa1 HBI Circored 97 12 0 1.03 0.32 --- 0.028
Sidor DRI Midrex 93.76 7.6 >3 1.67 --- 2.11 0.054
opco HBI Midrex 91.5 8.4 0.8 -2 --- --- 0.053

The analytical x-ray capabilities of the electron microscope were used to provide rough
estimates of the chemistries of the different phases, which were observed.

3.2 DRI Fast Melting Experiments

Several experiments were designed and conducted where DRI material was quickly
melted and quenched. The goal of these experiments was to allow the gangue and metal
phases to physically separate while in the liquid state, but to quench the material before
significant phosphorus transfer could occur. . A standard procedure for these
experiments was developed, which allowed for:
1. Reasonable Temperature Control - Phosphorus equilibrium is a strong function of
temperature. Poor temperature control after melting could introduce significant
error.

33
 2. Efficient Phase Separation - The micrographs of HBI and DRI shown in the
results section, illustrate the heavily intermingled nature of the metal and gangue,
Gas evolved upon melting DRI/HBI also tends to re-entrain gangue particles.
Efficient phase separation is necessary to ensure that the true metal phosphorus
content is being measured.
3. Accurate Analysis - The chemical analysis for these experiments was conducted
by the author, with special attention paid at all stages to ensure accurate analysis
of phosphorus contained in the metal phase only.
In these experiments 20 to 30 grams of either H B I or DRI were lowered into a hot
empty magnesia crucible in a high purity argon atmosphere. The MgO crucible was
heated within a molybdenum crucible via an induction furnace. For all experiments, the
temperature was held at 160O+1O0Cas measured by a two color optical pyrometer aimed
at the bottom of the MgO crucible. The accuracy of the pyrometer was initially checked
for this system against a type B thermocouple and found to be accurate within +lO°C. A
schematic of this apparatus is shown Figure 14. Once the material was in the furnace,
heating to the melting temperature took between 30 and 120 seconds depending upon the
density and gangue composition of the material. The time from when liquid was first
observed to form until melting was completed took 20 to 40 seconds for all materials.
For all materials containing carbon, rapid gas evolution resulted in a completely
emulsified melt for the first 30 seconds after melting. Pin samples were taken from the
melt via a 3 mm LD. quartz tube as a function of time as measured from first observed
melting. After 5 minutes had passed the furnace was shut off and the melt was cooled at
approximately 300"C/min. The final sample, the slug of metal remaining in the crucible,
was estimated to be solid at 6 min.

34
 Optical __+

P ym meter

~4 Silic-
Rxn Tube
1
0
Ma- ia 0

Crucible
40
0
0

DRI Pellets 2

Mo
-
Susceptor
Crucible

U
I

Pedestal Argon In
Figure 14: Experimental apparatus used for quick melting and quenching of DRI and HBI.

It was found that some of the samples, particularly those from early in the experiment,
were a mixture of slag and metal. Significant effort was made to separate the oxide and
metal prior to analysis. Approximately 0.3 to 0.5 grams of oxide free metal was isolated
from each sample for analysis via the phospho-molybdenum-blue photometric method.
Several additional fast melting experiments were conducted where HBI material was
dropped into a liquid slag phase instead of an empty crucible. For these experiments 30
grams of HBI were dropped into 15 grams of magnesia saturated liquid slag with a
basicity of 1 and zero initial iron oxide at 1600°C. The intended purpose of the slag

35
phase was only to capture and dilute the early melting components of the gangue, prior to
bulk melting of the HBI material. For this reason the slag initially contained no iron
oxide, and thus had no phosphorus capacity. If the early melting portion of the gangue
contained significant amounts of phosphorus, and could effectively separate to an
external slag prior to bulk melting, the initial metal samples should show a relatively low
phosphorus content, which would increase with time.
One final set of fast melting experiments were conducted where the gangue chemistry
of the Circal material was altered by adding CaO or CaCO, to the material. For these
experiments, several hundred grams of the Circal HBI was crushed to powder of less than
40 mesh. The crushed HI31 was then blended with small amounts of reagent grade CaO or
CaCO, powder. Both the CaO and CaCO, were very fine powders. The material was
then repressed to a density of approximately 5 g/cm3. The Circal material contains
approximately 1% silica, 5 to 12% FeO, and very little else other than pure iron. In the
modified material, just enough lime was added to give a basicity of 1 and 1.5. Rapid
melting experiments were subsequently conducted where 30 grams of material were
lowered into a hot empty crucible. The fast melting procedure was the same as described
above.

3.3 Controlled Melting Observations

In order to develop a better understanding of the behavior of DRI/HBI during melting,
several controlled melting experiments were conducted. A hot stage confocal
microscope was used to observe two HBI materials microscopically during melting. In
the microscope, 4mm diameter samples of either Midrex HBI or BHP-HI31 were heated
until complete melting occurred. Samples of each material were prepared with a single
polished surface. The temperature was controlled manually, and increased from room
temperature to approximately 1550°C at a rate of roughly 10 to 25 "C per minute. The
entire heating and melting cycle was recorded on a S-VHS cassette via a CCD camera for
subsequent review.
Both DRI and HBI materials were also observed macroscopically during melting in an
induction furnace. For these experiments, sectioned and polished samples of either DRI
or HBI were placed in MgO crucible contained within a molybdenum crucible. The
Molybdenum crucible was heated via the induction furnace at approximately 10 "C per

36
minute as measured by a two color optical pyrometer aimed at the polished surface of the
DRI or HI31 sample. The heating and melting cycle was recorded via a CCD camera and
S-VHS videocassette recorder.

3.4 Laboratory Kinetic Experiments

It was originally hypothesized that the phosphorus in DRI is primarily present in the
oxidized state in the gangue. It was believed that slow kinetics coupled with frequent
slag flushing resulted in the lower than expected metal phosphorus levels, which have
been reported by some EAF steel producers. Thus, a fundamental investigation of the
kinetics of phosphorus reversion was originally planned as an integral part of this study.
The present understanding is that, even if phosphorus is initially present exclusively in
the oxide phase, phosphorus reversion from gangue to metal will be very rapid upon
melting due to the enormous metal - oxide interfacial area present in DRI pellets. If this
is the case, the kinetics of both phosphorus reversion and removal are important to
understanding the observed industrial phenomenon.
Laboratory experiments designed to elucidate the rate controlling mechanisms and
rate parameters of phosphorus transfer were conducted. These experiments are designed
to measure the rate of transfer from slag to metal and metal to slag as a function of slag
chemistry and temperature. It was initially assumed that the rate of phosphorus transfer
is controlled by liquid phase mass transfer in both the slag and the metal.
Dephosphorization experiments were conducted for slag chemistries representative of
common EAF practices:
CaO - 30-55%
SiO, - 10-30%
MgO - 8-12%
FeO - 5-40%
In addition, phosphorus reversion experiments were conducted using a slag phase
representative of typically observed DRI gangue chemistries, e.g. low basicity, high
phosphate, and variable FeO. The temperature of lunetic investigations ranged from
1550°C to 1650°C. The experiments were conducted in a 10 kW induction furnace under
known stirring conditions for the metal. Experiments were conducted in 3 cm I.D. high
purity, high density magnesia crucibles unless otherwise specified. Samples of the metal
were taken via a 3mm I.D. quartz tube as a function of time and analyzed for phosphorus.

37
Slag samples were take using a cold iron rod and were analyzed for CaO, MgO, SiO,,
Fe,O, and phosphorus. Phosphorus in both the slag and metal was analyzed by the
phospho-molybdenum-blue photometric method. The specific results and method of
analysis for slag samples taken from kinetic experiments appear in Appendix D. Initially,
in these laboratory experiments the simplest case was studied, where no carbon is
present, and the reaction:
(FeO) = [Fe] +[O]
controls dissolved oxygen in the metal. Due to the extreme complexity of coupled slag
metal reactions such as the phosphorus reaction, it is necessary to reduce of the number
of system variables present in a real EAF environment to a manageable few. Under these
controlled experimental conditions, basic mass transfer and kinetic parameters can be
determined and then applied to more realistic systems.
The experimental apparatus for these experiments is shown in Figure 15.

38
 Argon Out
n
- : . r : I
I I
Fused Silica
\

-
I
Reti-actory Lid
0
Mo Wire 0
Susceptor3
0
MgO Crucible 0

2
\
0
Alumina

TypeB
Thermocoupl e
I

u Argon In
I

Figure 15: Experimental apparatus for kinetic experiments.

For these experiments, the iron metal (approximately 150 to 200 grams) was inductively
heated and stirred. However, the slag (approximately 60 to 80 grams) was indirectly
heated via a molybdenum susceptor. In order to be sure thermal gradients were

39
minimized, the surface temperature of the slag was continuously measured using a two
color optical pyrometer and the metal temperature was measured via a type B
thermocouple in contact with the bottom of the magnesia crucible. The pyrometer was
calibrated for the slags used in the experiments with a type B thermocouple. Special
attention was paid to ensure that the slag and metal temperatures stayed within 10 degrees
of one another. Because the molybdenum susceptor coupled very efficiently with the
induction coil, the temperature of the slag could be adjusted by +50"C by altering the
vertical position of the experimental apparatus within the coil, and/or by making subtle
changes to the spacing of the coils. Continuous monitoring and adjustment of the
temperature in this manner ensured a fairly isothermal system.
Several experimental techniques were explored initially for dephosphorization and
rephosphorization kinetic experiments. It was found that the addition of small, high
phosphorus additions, which dissolved rapidly in either the slag or metal allowed for the
shortest transient period at the start of each experiment. For dephosphorization
experiments, pure electrolytic iron and the experimental slag were allowed to equilibrate
at the experimental temperature for 30 minutes. At that time, an addition of 2 to 4 grams
of a Fe-9.5%P alloy was made to the metal. The ferrophosphorus addition was made by
attaching the material to a 2 mm diameter pure iron wire, which was submerged through
the slag layer to the bottom of the crucible until the wire and the ferrophosphorus melted.
The experiment was judged to begin at the moment the iron wire began to collapse into
the melt. For rephosphorization experiments, phosphorus additions were made to the slag
in the form of a partially sintered calcium phosphate pellet. Experiments were judged to
begin when the pellet was visually observed to melt into the slag phase. For both
dephosphorization and rephosphorization experiments, the temperature of the slag andor
metal phase were observed to fluctuate by less than 10 "C at the time of addition.

40
 ChaDter 4

Results and Discussion

4.1 Metallographic Observations
Although the chemistries, densities, and other physical properties of the materials
investigated in this study differ greatly, several common features were observed. The
"gangue phase" is actually a complex mixture of several discrete oxide phases.
Phosphorus was observed to be associated only with the oxide phases containing
significant amounts of lime. In general, the oxide feature sizes observed in the two
Midrex based materials were very small, generally less than 50 p.The oxide particles
found to contain measurable amounts of phosphorus were extremely small, usually less
than 5pm. Figures 16 and 17 are micrographs of Midrex pellet materials, which were
sectioned, polished, and observed via scanning electron microscopy.

41
 -. _ _ _ _ ~ _ _ _ _
Figure 17: SEM micrograph of polished Midrex DRI cross section.

Figures 18 - 20 are micrographs of Midrex HBI material, which was sectioned, polished,
and observed via scanning electron microscopy.

42
 chemical

Figure 20: SEM micrograph of polished Midrex HI31 cross section. The approximate chemical
composition of the indicated oxide feature was measured via EDS.

43
For both the DRI and HBI materials examined, three general types of oxide features are
observed in Figures 16 - 20:
1. The darker gray areas embedded in the lighter gray iron matrix were found to
be primarily unreduced iron oxide.
2. The larger black particles, indicated in Figure 19, were primarily SiO, with
only trace amounts of other oxides and no elevated concentration of
phosphorus
3. The very fine oxide features, indicated in Figure 20, were found to be a
mixture of several oxide components, including significantly elevated levels
of phosphorus and lime.
Samples of BHP HBI produced via the Finmet process were also examined.
Microstructurally, this material is very different than Midrex based materials. This
is due to the fact that the Finmet process uses a series of fluidized beds to reduce
ore fines. Alternatively, in the Midrex process, a combination of lump ore and
pelletized ore are reduced in a shaft furnace. The Midrex pellets produced from
pelletized ore were generally observed to have a very fine grain structure, whereas
the briquetted fines based Finmet material exhibited a much courser structure.
Figure 21 is optical micrograph of BHP HBI, which was sectioned and polished.

Figure 21: Optical micrograph of polished BHP HBI cross section.

44
Clearly, the largest gangue features observed in this material are on the order of
thousands of microns as opposed to tens of microns as seen in Figures 16-20.
However, when these very large particles are examined in closer detail in the SEM,
one finds that they are in fact a composite of several oxide phases. Figure 22 is an
SEM micrograph of the same sample of BHP-HBI. An approximate overall
chemistry is listed for the large particle shown in the image as measured via the
EDS unit of the microscope.

Figure 22: SEM micrograph of polished BHP HBI cross section.

As mentioned, this "macro" particle is actually a composite of various phases, thus

the chemistry listed does not reflect the composition of a single specific phase.
With the exception of these very large oxide features, which were not observed in
pellet based materials, the finer gangue particles were similar to what was observed
in the Midrex DRI and HBI. The larger and most common oxide features,
excluding unreduced iron oxide, include a very high silica - alumina phase. One
such particle is shown in Figure 23.

45
Figure 23: SEM micrograph of polished cross section of BHP HBI. The approximate
composition of the large oxide feature was measured via EDS.

In general, these high silica particles, as well as the regions of unreduced FeO, and
the "macro1'composite particles do not appear to contain elevated levels of
phosphorus as compared with the surrounding metal matrix. As with the Midrex
based materials, particles containing significant amounts of phosphorus also
exhibited much higher contents of lime, supporting the conclusion that the
phosphorus in direct reduced materials is most likely in the form of a calcium
phosphate phase. Although the high phosphorus particles identified in the BHP-
HBI are 2 to 5 times larger than those identified in the pellet based materials, even
the largest of these particles were no larger than 50 microns in diameter. A typical
high phosphorus particle found in the BHP-HBI material and the approximate
chemical composition of the particle is shown below in Figure 24.

46
Figure 24: SEM micrograph of polished cross section of BHP HBI. The approximate
composition of the large oxide feature was measured via EDS.

It should be noted that the various discrete phases observed in the solid state are
unlikely to separate from the metallic iron phase as discrete particles.. The CaO-
P,O, pseudo-binary system exhibits phases, which melt well below 1200°C. The
FeO rich side of the FeO-SO, system also contains low melting phases. During
controlled melting experiments reported below, first melting was observed to occur
at 1100°C or less. For the most part, the various dlscrete oxide phases contained in
DRI/HBI material are loosely connected by strings of unreduced or reoxidized
FeO. It is unlikely that a gangue particle of the chemistry shown in Figure 24
would still exist as a discrete phase at the time of bulk melting.

4.2 Fast Melting Experiments

The results of fast melting experiments with six commercial DRI and HBI materials are
shown in Figure 25.

47
 0.07

0.06 - A A A
--
A A
E - -
1
0
0.05 -
n
rn Y Y
r --t- Midrex DRI
0.04-
m m

ir\,
U
-t- Midrex HBI
EB 0.03 - A Sidor DRI
O----* n - - 0 -

-
-
L ^

0.02 -O-- Circal HBI

-W- BHP HBI

0.01 - OPCO HBI

0 2 4 6 8

Minutes
Figure 25: Results of fast melting experiments with six commercial DRI and HI31 materials.

Table 1 from Section 3.1 listing the chemical compositions of the materials used in the
fast melting experiments is shown again below:

Table 1: Chemical Composition of Investigated Commercial Materials

Material Method of Fet FeO C SiO, A1,0, CaO P
Reduction
Midrex DRI Midrex 93.13 5.87 1.89 1.92 --- 1.59 0.028
Midrex HBI Midrex 92 10 ~ 0 . 5 1.8 --- --- 0.06
BHP - HBI Finmet 92 8 1.3 1.6 --- 0.1 0.045
Circal HI31 Circored 97 12 0 1.03 0.32 --- 0.028
Sidor DRI Midrex 93.76 7.6 >3 1.67 --- 2.11 0.054
opco HBI Midrex 91.5 8.4 0.8 -2 --- --- 0.053

If the data in Figure 25 are compared with the total phosphorus values for each material
(shown in Table 1) it is clear that nearly all the phosphorus is transferred from the gangue
to the metal in less than a minute after melting. Furthermore, only modest fluctuations in
the metal phosphorus occur at times longer than a minute.
The reason for this rapid reversion can be better understood by making some
fundamental calculations. The metallographic investigation of DRI/HBI indicated that
the oxide gangue features are on the order of tens of microns or less in dimension.

48
Therefore, the area of contact between the gangue particles and the surrounding iron
matrix is enormous. An extremely conservative estimate of this area of contact can be
calculated based on the mass fractions of metal and oxide provided by chemical analysis
and assuming a dispersion of spherical oxide particles 50 microns in diameter. The
assumption of spherical gangue particles significantly underestimates the contact area as
this geometry represents the lowest possible surface area to volume ratio. Such a very
conservative calculation predicts an area of contact on the order of 120 cm2/cm3of HBI
material or roughly 600 cm2/gramof HBI. This estimate of metal - oxide interfacial area,
combined with calculated estimates for the equilibrium phosphorus distribution for the
gangue material (Lp -1) and the overall mass transfer coefficient (ko-lo"), can be used
in the kinetic model derived later in this paper to predict an equilibration time of
significantly less than one second. Based upon these approximations, Figures 26 and 27
show the calculated change in metal and gangue phosphorus levels respectively assuming
all of the phosphorus is initially contained in the gangue phase.

0.07

0.06
0.05

a
R

Y
0.04
Be -50pm particle diameter
* 0.03
> -200vm particle diameter
0.02

0.01

0
0 0.5 1 1.5 2 2.5 3
Time (seconds)
Figure 26: Calculated metal phosphorus content versus time assuming a liquid phase mass
transfer controlling model.

49
 0.6

0.5

h
0.4
n
v
0.3
*
0.2

0.1

0
0 0.5 1 1.5 2 2.5 3
Time (seconds)
Figure 27: Calculated gangue phosphorus content versus time assuming a liquid phase mass
transfer controlling model.

In Figures 26 and 27 two cases were modeled, assuming:

1. An average gangue particle size of 50 p.
2. And, an average gangue particle size of 200 p.
These calculations agree with the experimental observation that all of the phosphorus is
transferred from the gangue phase to the metal within the time scale required for melting.
Fast melting experiments in which a liquid slag was present were conducted with the
Circal and BHP - HBI materials. Figure 28 shows the results of a fast melting
experiment where 30 grams Circal was added to 15 grams of liquid magnesia saturated
liquid slag with a basicity of 1 and zero initial iron oxide at 1600°C.

50
 0.014
4
0.012

0.01
n
0.008
8 -Circa1 HBI w/ slag
g 0.006
0.004

0.002

0
0 1 2 3 4 5 6 7

Minutes
Figure 28: Metal phosphorus as a function of time for Circa1 HI31 melted into a magnesia
saturated liquid slag with basicity = I at 1600°C.

The results shown in Figure 28 are very interesting for several reasons. Because
this material contains no carbon, the FeO contained in this material reported to the
slag generating a slag with some phosphorus capacity. The indication of the data is
that phosphorus transfer kinetics are fast in this system and that the system is near
equilibrium in under 1 minute. The data suggest that the phosphorus is slightly
decreasing with time. This would imply that full reversion occurs simultaneously
with melting and that the phosphorus is then transferred from the metal to the slag
until equilibrium is approached (or the experiment is terminated.)
The results of the fast melting experiments with the BHP - HI31 material in slag
are shown in Figures 29.

51
 0.035 -.
~ B H PHBI W/ slag I
x
Be
c 0.02 - wt%(P)
3
0.015 -

0 2 4 6 8 10 12 14 16
Time
Figure 29: Metal phosphorus as a function of time for BHP - HBI melted into a magnesia
saturated liquid slag with basicity = I at 1600°C.

At the time of the experiment shown in Figure 29, the degree of metallization of the
BHP material had degraded somewhat from the time at which chemical analysis in
Table 1 was completed. Therefore, the 1.38wt% carbon contained in the material
was not sufficient to consume all of the iron oxide present. Therefore, like the
Circal material, the BHP - HBI contributed some FeO to the liquid slag upon
melting, generating a slag with some phosphorus capacity. As was seen with the
Circal material, roughly half of the phosphorus in the material had transferred to
the slag by the time of the final sample.
The metal samples from the experiment in Figure 29 were also analyzed for
carbon. The data from the experiment with the BHP material indicate that by the
time of the first sample 99% of the carbon contained in the material has been
consumed. The continued reaction between the remaining carbon and the FeO in
the slag is apparent in the data.
The results of the fast melting experiments with the modified Circal material to
which lime was added are shown below in Figure 30.

52
 0.035 I

0.03 -
I Total phosphorus = 0.03%

0.025 -

0.02 -
n

w
Y

5 0.015 -
0.01

0.005 -
-
- Circal HBI w/ 1.5% CaO

As the Circal material contains approximately 1wt% silica, additions of 1 wt% CaO and
1.5 wt% CaO should have resulted in a gangue chemistry with a V-ratio of 1 and 1.5
respectively. The 1.78% CaCO, should have given the molar equivalent of 1% CaO.
None of these additions were found to significantly affect the metallic iron yield of the
HBI (around 86%.)
There are several interesting results in the data of Figure 30.
1. Small additions on the order of 1 % cut the melt-in metal phosphorus in half.
2. Although the both the Sidor and Midrex DRI materials exhibit gangue chemistries
with V-ratios greater than 1 (see Table 1) the carbon in these materials consumes
the residual iron oxide upon melting. The resultant gangue upon melting contains
almost no iron oxide and will therefore have a very low phosphorus capacity.
3. The data in Figure 30 show a lot of scatter. This is because a high phosphorus
gangue actually forms when enough lime is added to give the gangue some
phosphorus capacity. Small amounts of gangue entrained in metal samples
strongly affected the analysis.
4. There appears to be little difference whether the lime is added in the form of CaO
or CaCO,. The compromise presented with CaCO, is that a greater mass of
material is required for the same number of moles of Ca.

53
 5 . There also seems to be little difference between the experiments with additions of
1% and 1.5%CaO. This is most likely due to the fact that, in the case of the Circa1
material the gangue, which is formed will be more than 80% FeO. At such a high
iron oxide contents, small shifts in the V-ratio will have much less of an effect on
phosphorus equilibrium than is normally expected.

The use of prefluxed or basic ore pellets in h4drex type reactors or the addition of small
amounts of lime into fines based reactors may represent a simple and cost effective way
of controlling the phosphorus in direct reduced materials. However, in materials with
very high carbon content, the effect upon melt-in phosphorus will be negligible.
Although, the metallic yield of a fluxed material will be reduced by 1 or 2%, several
benefits may be generated by the added components. Most electric furnace producers
modify their standard furnace flux practice when using large amounts of DRIM[BI to
account for the additional silica in the material. A DRVHBI material, which melts in with
a gangue chemistry similar to the target slag chemistry, is more easily substituted in to
the standard scrap melting practice. Furthermore, large flux additions in the form of lime
or dolomite often dissolve slowly in the furnace and may not be effective in controlling
the furnace chemistry until late in the heat. Lime contained within DRI or HI31 will melt-
in far more quickly than top charged lime and may generate a good steelmalung slag
more efficiently than conventional fluxing practices. Ultimately, a pre-fluxed material
would reduce the number of operating variables when using DRI, guaranteeing greater
consistency in the furnace slag chemistry from heat to heat and during individual heats.
Greater consistency offers many operational benefits, which will not be elaborated upon
here.

4.3 Controlled Melting Observations

Several important observations were made based on the controlled melting
experiments:

54
1. First melting of the oxide gangue phase occurs long before the melting temperature of
the bulk metal phase is approached. First melting has been observed to occur on a
microscopic scale at approximately 1050°C for the HBI material. A liquid phase can
be observed on a macroscopic scale at approximately 1200°C. Most or all of the
oxide phase appears to be molten at approximately 1350°C.
2. The molten gangue appears to react with the surrounding metal to form a gaseous
product at temperatures as low as 1050°C. This can be verified by the formation and
release of bubbles in the liquid oxide phase.
3. When the material is heated slowly (10 - 20"/min) from 1000 "C to the metal melting
temperature at approximately 1530"C, the reaction between the metal and gangue
appears to be complete before the sample reaches 1200°C. When the sample is
heated rapidly (50 - 100"/min), the reaction resulting in bubble formation in the
gangue continues until the entire sample is molten.

The analytical software package Thermocalc was also used to model the melting
behavior of an oxide phase with a chemistry representative of DRI/HBI gangue.
The model predicted first melting to occur at 1051°C and complete melting to
occur at 1233°C. This prediction is supported by the above mentioned physical
observations. These findings are of interest as they may offer a possible
mechanism by which gangue - metal separation might occur during an EAF
heating cycle. When melted in the presence of a liquid slag phase, the phosphorus
in the low melting gangue might be physically "captured' and diluted in the slag
phase long before bulk melting of the metal phase occurred. Both the driving force
and lunetics of reversion from a bulk slag to the metal would be reduced as
compared with the direct reaction between the gangue and metal. However, the
efficiency of the escape of the gangue from the solid metal matrix is controlled by
several factors:
1. The inter-connectivity of the oxide phase within the metal matrix.
2. The driving force for phase separation, both surface tension effects and gas
pressure from internal reaction within the material.
3. The precise melting temperatures of both the oxide and the metal phases.

55
In spite of the above arguments, the results of the fast melting experiments with and
without a liquid slag present suggest that the achieved degree of gangue separation
at low temperatures is not sufficient to result in significantly reduced metal
phosphorus levels at the time of bulk melting.

4.4 Kinetics of Phosphorus Transfer

4.1.1 Mathematical Model of Phosphorus Transfer - In order to evaluate the
experimental results, it is necessary to introduce the framework of the mathematical
model used in this study to describe the kinetics of phosphorus transfer. Most recent
literature is in agreement that the slag-metal phosphorus reaction is correctly expressed
in the form:

[ P ]5 + p 3+ p-2 ) = ( P o i 3 ) 33

Where [ ] indicates metal phase and ( ) indicates slag phase. The thermodynamics of
phosphorus transfer was discussed in detail in Section 2.2. If this reaction is accepted,
several possible rate controlling steps become apparent. In the case of dephosphorization
these steps include:
1. Metal phase mass transfer of [PI to the slag metal interface
2. Metal phase mass transfer of [O] to the slag metal interface, or supply of [O] from
the slag to the interface via the reaction: (FeO) = Fe + [O]
3. Slag phase mass transfer of (02-)
to the interface
4. Slag phase mass transfer of the away from the interface
5. Chemical reaction of the formation of
The possible rate limiting steps of rephosphorization are obviously linked, differing only
in the direction of the transferred species and the dissociation rather than the formation of
Based upon the fact that most slag - metal chemical reactions are rapid at high
temperatures, most previous researchers have assumed the kinetics of phosphorus transfer
to be limited by liquid phase mass transfer. Since [O] can be supplied to the interface
both from the bulk metal and via the dissociation of FeO at the slag-metal interface, it is
most likely that the kinetics of phosphorus oxidation are limited by the mass transfer of

56
phosphorus either in the slag or in the metal. Previous researchers have shown that the
oxidation of phosphorus from high carbon iron via a highly oxidized slag is possible. [44-
471 This suggests that the oxygen potential of the interface is controlled by the slag phase
and that the rate of oxygen delivery to the interface by the slag is faster than the rate of
mass transfer in the metal phase. As a result, it is likely that the kinetics of phosphorus
transfer are controlled by mass transfer of phosphorus in the slag, metal, or both phases.
The results of previous researchers are divided as to whether mass transfer in the slag or
metal is the rate controlling step. This is due to the fact that it is experimentally difficult
to separate the individual effects of mass transfer in the two phases. Therefore, it was
initially assumed that phosphorus transfer between the slag and metal is controlled by
liquid phase mass transfer in both phases.
Several analytical and numerical models of phosphorus transfer have been developed
by other researchers and were discussed within the literature review of this study. For
this study, kinetic expressions for phosphorus transfer from slag to metal and metal to
slag were developed based upon liquid phase mass transfer correlations developed for
sulfur transfer by Fruehan [4]and others.
In order to account for the effects of mass transfer in both the slag and the metal an
overall mass transfer coefficient, ko, must be defined. Assuming that phosphorus is
conserved within the slag-metal system, the flux of phosphorus out of the metal must
equal the flux into the slag and vise versa. This can be expressed mathematically as:

d(%P)
w, = - d[%P]wnl 34
dt dt
Where the flux equations for phosphorus in the metal and the slag are respectively:

28

--
a(%p) - - [(WP) - (%Pi)] 29
at ws

Then equations 28,29,34 can be combined to give:

35

57
Within the liquid phase mass transfer model, it is assumed that the chemical reaction is
fast. Therefore, the phosphorus contents of the slag and metal at the interface are given
by the equilibrium phosphorus distribution ratio, according to:

(%Pi) =L,[%P'] 36

By combining equations 34 and 35, it is possible to solve for [%PI] as a function of the
bulk concentrations:

PmkmE%Pl+ Psks (%PI 37

[%Pi]=
Pmkm + P s k s L ~

If this definition of [%Pi] is put back into the flux equation for phosphorus in the metal:

pmkm[%PI + psk, (%P)l 38

1
=-
at Wm Pmk, + P A L P

Equation 38 can be simplified and rearranged to give the result:

39

58
Equation 39 suggests that an overall mass transfer coefficient, ko, can be defined such

=m
that:
1
40
ko

As defined by equation 40, ko has the units of c d s . A similar exercise can be performed
to arrive at and expression for the flux in the slag phase:

41

In this case, ko, is defined as:

1
k, =
42

The two definitions for ko expressed in equations 40 and 42 differ by the ratio of the
densities of the metal and slag. Although it is a nuisance to have two definitions for ko, it
is necessary to maintain the units of ko as c d s . Because the independent contributions
of the slag and metal mass transfer coefficients are difficult to separate experimentally,
the flux equations for phosphorus in the slag and metal will generally be expressed in
terms of an overall mass transfer coefficient in the present work.
Under some conditions, it is possible to obtain an analytical solution to the flux
equations for phosphorus in the slag and/or metal. The phosphorus content of the metal
and slag are related through the mass balance:
W 43
[( %P)- (%PO)>]
= -[[%PI - [% P O ] ] - Q
WS

For the rephosphorization experiments of this study, the initial metal phosphorus [%Po]
level will be essentially zero. Thus, equation 43 can be combined with equations 41 and
42 such that:

59
 44

And then integrated to produce an expression, which computes (%P)as a function of

time:

An equivalent expression for dephosphorization can also be developed. In the case of

phosphorus transfer from metal to slag, the expression takes the form:

46

From this equation, the metal phosphorus level can be calculated directly as a function of
time. Within the solution of equation 45 and 46 it is assumed that neither the mass
transfer coefficient, ko, nor the interfacial area, A, are functions of time or concentration.
However, it will be shown in the experimental results of this work and of previous
studies, that either A or ko do vary as a function of time. This point will be discussed in
the following section.
Figures 31 and 32 show metal phosphorus and slag phosphorus levels versus time as
calculated by the above equations for conditions, which are relevant to this study.

60
 0.04
0.035
0.03 ko = 0.001cm/s
A =7cm2
0.025 Lp =52
Metal = 190grams
0.02 Gangue = 65grams
0.015
0.01
0,005
0
0 10 20 30 40
Minutes

Figure 31: Calculated metal phosphorus versus time for the experimental parameters of this study
assuming an overall mass transfer coefficient equal to 0.00 l c d s e c .

2.02

1.98 -I\
sB
m ko = 0.001cm/s
A =7cm2
.w
1s96 - Lp =52
v
1.94 - Metal = 190grams
Gangue = 65grams
1.92 -

1.88 ! I I I

0 10 20 30 40
Minutes

Figure 32: Calculated slag phosphorus versus time for the experimental parameters of this study assuming
an overall mass transfer coefficient equal to 0.001cdsec.

61
Equations 45 and 46 are also very useful for analyzing experimental data. When slightly
rearranged as shown below:

And:

All of the driving force or constant terms can be isolated to the right hand side or RHS of
these kinetic expressions. When this RHS of equation 47 is plotted versus time, the result
should be a straight line with a constant slope equal to koA. This product has been
previously termed the “mass transfer parameter.”[4] It is often necessary to group the
interfacial area and the mass transfer coefficient together in this way, because it can be
very difficult to calculate either term independently under some experimental conditions.
Specifically, under circumstances where significant slag-metal intermixing or
emulsification occurs, it is impossible to calculate a meaningful interfacial area.
In previous evaluations of steelmalung vessels, the entire term:

49

has also been used a convenient lunetic parameter. In oxygen steelmalung vessels, the
oxygen blowing rate often scales linearly with the size of the vessel or the weight of the
contained metal, W,. Also, both the slag-metal area, A , due to intermixing and the mass
transfer coefficient, ko, tend to increase with increasing blowing rate. Therefore, the term
expressed in equation 49 remains relatively constant for different oxygen steelmaking
vessels of varying size. However, this balancing relationship between reactor size and
gas usage is not expected to hold for electric arc furnace vessels, and may be less useful

62
in describing this system. The term in equation 49 has the units of reciprocal time and is
very similar to a reaction rate constant. Defining this term as a ‘mass transfer rate
constant,’ however, is an inaccurate description of the properties of the parameter as it
likely to vary significantly as a given furnace heat proceeds. During both oxygen
steelmaking and electric arc furnace steelmalung the physical conditions, which influence
equation 49, i.e. the stirring conditions, slag-metal area, and liquid slag and metal masses,
are functions of time during any given heat. Therefore, the term in equation 49 should
not be expected to remain constant for the duration of an entire heat.

4.4.2 Laboratory Kinetic Experiments

A large number of dephosphorization and rephosphorization experiments were
conducted. Early dephosphorization experiments were conducted by a slightly different
technique than the standard method described in Chapter 3. Initially, an Fe-0.6%P alloy
and a slag containing no iron oxide were melted and allowed to equilibrate at the
experimental temperature. In theory, a slag containing no iron oxide should have a
phosphorus partition ratio near zero. The experiments were initiated when an addition of
FeO was made to the slag, thus increasing Lp. The large slag addition required for these
experiments raised some concerns regarding the transient period near the beginning of the
experiments. However, some reproducible results were obtained. Below, Figures 33,34,
and 35 show the results of three dephosphorization experiments plotted as weight percent
phosphorus in the metal versus time. The continuous curve shown in the plot was
calculated via equation [46] for the experimental conditions assuming a best fit ko.

63
 h
0.7

Dephosphorization Experiment #12

CIS= 1.7
FeO = 30%
0.5 MgO= 9%
o.6 T = 1600°C

0.1 -
----Assuming ko = 0.02 cm/s

0 I
* -
a

I
*
I
A
I
A
I
4

0 10 20 30 40 50 60 70
Time (minutes)

Figure 33: Metal phosphorus versus time for dephosphorization experiment #12. Continuous curve
calculated via equation [46] assuming a best fit ko.

0.6- Dephosphorization Experiment #13

CIS = 1.9
FeO = 20%
0.5 - MgO= 9%
 Dephosphorization Experiment #14
c/s = 2.0
FeO = 30%
MgO= 9%
T =1650°C

----Assumingko = 0.015cm/s
6 +
6
0 I I I 1 I

0 10 20 30 40 50 60
Time (minutes)

Figure 35: Metal phosphorus versus time for dephosphorization experiment #14. Continuous curve
calculated via Equation [46] assuming a best fit ko.

These three experiments (12 - 14) are typical of the general trends, which were observed
for early experiments. The relatively high initial metal phosphorus level decreased
quickly. After 10 to 15 minutes, very little phosphorus transfer was observed. This
initial fast rate followed by a slower rate of phosphorus transfer is consistent with the
results of Mod et al. [34] In Figures 36 - 38, the data from experiments 12-14 are
replotted as the right hand side, RHS, of equation [48] versus time. Therefore, the slope
of the data in these RHS versus time plots is equal to the mass transfer parameter A*ko
with the units cm3/s

65
 100 -

75 -
-
50 -

c m3/s

Dephosphorizat ion Experiment #12

I I I I I I

0 5 10 15 20 25 30 35
Time
Figure 36: Data from dephosphorization experiment #12 replotted as the RHS of equation [48] versus time.

75

Dephosphori.zat ion Experiment #13

0 1 I I I I I I

0 5 10 15 20 25 30 35
T ime (minut es)
Figure 37: Data from dephosphorization experiment #13 replotted as the RHS of equation [48] versus time.

66
 50

25
I
A*ko=0,14cm3/s

Dephosphorizat ion Exper iment#14

01 I I I

0 5 10 15 20 25
Time
Figure 38: Data from dephosphorization experiment #14 replotted as the RHS of equation [48] versus time.

Although there is some scatter in the data presented in Figures 36 - 38, it is clear that the
apparent rate of reaction is changing. Because it is difficult to determine whether the
observed trend is the result of a systematic change in the reaction area or in the mass
transfer coefficient, the results are analyzed in terms of the mass transfer parameter. The
continuous curve indicates the general trend of the data but has no quantitative
significance. The two rates near the beginning and end of the experiments are only meant
to indicate the magnitude of the mass transfer parameter at these times. The author does
not intend to suggest that these two rates indicate two separate regions with different
mass transfer mechanisms and rates. All three experiments display a decrease in the
mass transfer parameter by a factor of 20 to 30 from the start of the experiment to the
end. In Figures 39 and 40 the kinetic data of two experiments from Mori et al. [34]
which were presented in Chapter 2.3, are replotted as the RHS of equations 48 and 47
respectively.

67
 10

+Slcg4, 1640°C
fl

0
0 5 10 15 20
T ime (minut es)
Figure 39: Adapted from Mori et al. [34]. Data from dephosphorization experiment shown in Figure
10 replotted as the RHS of equation 48 versus time.

0.5'

0-4-

0.3'

0,2'

- s
i *SIag5, 1615" C
0.1 ko = 0.002 c m/s

n
U
4 I I I I

0 5 10 15 20 25
Time (minutes)
Figure 40: Adapted from Mori et al. [34]. Data from rephosphorization experiment shown in Figure
11, and replotted here as the RHS of equation 48versus time.

68
The trends of the present study are in agreement with those observed in the data of Mori
et al. [34]. Both the general decrease in the apparent rate and the numerical values of the
initial and final rates (when crucible size differences are accounted for) are very similar
for the present work and the work of Mori et al. [34].
However, the experimental procedure followed in early experiments was very similar
to that of Mori et al. [34] and therefore shared many of the problems of those
experiments. In these initial experiments, the slag and metal temperatures were not
measured independently. The experimental temperature was monitored only by the
optical pyrometer. Later calibrations suggested that a significant temperature gradient
may have existed between the slag and metal phases. Also, there was some concern that
the relatively large slag additions (20 to 30 wt%) were resulting in an unacceptably long
transient period at the start of the reaction. Furthermore, due to the apparently
discontinuous change in the reaction rate at short experimental times, as displayed in
Figures 36 and 38, it was desired to develop an experimental technique, which would
allow for a precise reaction start time, a minimum transient period, and accurate sampling
at short times. For these reasons, the standard method described in Section 3.4 was
adopted.
The results of four dephosphorization experiments are shown below in Figures 41 -
44. Again, the continuous curves were calculated via Equation [46] assuming a constant,
best fit ko.

69
 Dephosphorization Experiment#1 8
cis = 1.9
0.2 FetO = 20%
MgO =9.5%

-
&
Y
0.15
T = 1600°C

Be
c’
3: 0.1

0.05 Assuming ko = 0.008cm/s

0 1 I I I

0 20 40 60 80
Time (minutes)

Figure 41: Metal phosphorus versus time for dephosphorization experiment #18. The solid line was
calculated via Equation [46] assuming a constant, best fit ko.

0.16

0.14 Dephosphorization Experiment # 7 9

cis=2.1 .
FeOt=20% .
0.12
MgO=8,5% .

-n 0.1
T = 1600°C .

5 0.08

Y
3:
0.06

0.04

0.02
+ + +
0 10 20 30 40 50 60 70 80

Time (minutes)

Figure 42: Metal phosphorus versus time for dephosphorization experiment #19. The continuous curve
was calculated via Equation [46] assuming a constant, best fit ko.

70
 R
0.14 I
Dephosphorization Experiment#20
c/s = 2.0
FetO = 25%
0.1 MgO =9.0%
Os’* T = 1650°C

0.02 Assuming ko = O.O08cm/s

o ! I 1 I

0 20 40 60 80
Time (minutes)
Figure 43: Metal phosphorus versus time for dephosphorization experiment #20. The continuous curve was
calculated via Equation [46] assuming a constant, best fit ko.

0.12

Dephosphorization Experiment #22

0.1
CIS= 1.4 .
FeOt= 18%
0.08 MgO= 12% .
T = 1600°C I

8 0.06

U
ti
0.04

0.02

n
- J
.
0 10 20 30 40 50 60 70 80

Time (minutes)
Figure 44: Metal phosphorus versus time for dephosphorization experiment #22. The continuous curve was
calculated via Equation [46] assuming a constant, best fit ko.

71
Again, the data from these experiments were replotted as the right hand side of equation
. The resulting graphs are shown below as Figures 45 -47:
70 1

60-
60

50 -
A*ko = 0.006 cm3/s
40 -

/
J
30

20 A*ko = 0.1 cm3/s

10
lo L Dephosphorization Experiment #18 .
0
O l
0 10 20 30 40 50 60 70 80

Time (minutes)
Figure 45: Data from dephosphorization experiment #I8 replotted as the RHS of equation [48] versus time.

80 ,
70

60 \
A*ko = 0.007 c m 3/s
50

J{{
40
K

30

20 A*ko = 0.07 cm 3 / ~

10
Dephosphorization Experiment #19 .
0
0 10 20 30 40 50 60 70 80
Time (minutes)
Figure 46: Data from dephosphorization experiment #I9 replotted as the RHS of equation [48] versus time.

72
 A
v

35

30 A*ko = 0.0007 cm 3 / s

25

15
A*ko = 0.091 cm 3 / s
10

5 Dephosphorization Experiment #20 .

O I I I I

0 20 40 60 80
Time (minutes)
Figure 47: Data from dephosphorization experiment #20 replotted as the RHS of equation [48] versus time.

60

50

40

vl
30
U

A*ko=0.049 cm 3 / s

Dephosphorization Experiment #22

0 10 20 30 40 50 60 70 80

Time (minutes)

Figure 48: Data from dephosphorization experiment #22 replotted as the RHS of equation [48] versus time.

73
 These results clearly confirm the trends observed in early experiments and also in the
data of Mori et a1 [34]. Figures 45 -48 show that what appeared to be a discontinuous
rate change in early experiments is actually a continuous decrease in the rate as a function
of time. Again, experiments showed a reproducible drop in the apparent mass transfer
parameter by approximately one to two orders of magnitude. For most
dephosphorization experiments the average overall mass transfer parameter for the first 4
minutes was found to be approximately 0.07 c d s . For the final 45 to 60 minutes of the
experiment the average A*ko was typically 0.007 cmls for dephosphorization
experiments.
In addition to the dephosphorization experiments reported above, several
rephosphorization or phosphorus reversion experiments were conducted. The results of a
phosphorus reversion experiment are shown in Figure 49. In this experiment, a slag
chemistry representative of DRI gangue was used:

I
0.14 - Rephosphorization
Rephosphorization Experiment
Experiment #16
#16
CIS =0.32
0.12 - FeO =25%
MgO = 10%
0.1 - (%P)0=2%
T = 1600°C

0.08

0,06

0.04 ----Assuming ko = 0.0003cm/s

0.02

0 I

0 20 40 60 80 100
Time (minutes)
Figure 49: Results of rephosphorization experiment #16 plotted as metal phosphorus versus time. The
continuous curve was calculated via equation E451 assuming a best fit ko.

In agreement with the results of Mori et a1.[34], it was found that rephosphorization
proceeded at a slower rate than dephosphorization under seemingly similar experimental
conditions. The experiment shown in Figure 49 had not reached a clear equilibrium

74
phosphorus distribution after 90 minutes. The data from Experiment #16 are replotted as
the RHS of equation [47] in Figure 50 below:

2.5

2- 4

1.5 A*ko = 2x10 -4cm3/s

0
VI
I
er
1
A*ko = 0.002
0.002 cm 3/s
0.5

Rephosphorization Experiment #16

0
0 20 40 60 80 100
Time (minutes)

Figure 50: Data from Rephosphorization Experiment #16 plotted as the RHS of equation [47] versus time.

Although the average rate in Figure 50 is significantly slower than the rates of
dephosphorization presented earlier, the data exhibits a pronounced bend toward a slower
rate at longer times. Again the initial and the final rates differ by approximately one
order of magnitude. As was discussed in Section 4.4.1, the overall mass transfer
coefficient for rephosphorization should differ from that of dephosphorization by the
ratio of the slag density over the metal density, approximately 0.5 in this case. The data
of the present study indicate that the rate of rephosphorization is 20 to 50 times slower
than the observed rates of dephosphorization.
Two significant differences exist between the conditions of the rephosphorization and
dephosphorization experiments, which should be noted. The pre-logarithmic terms in
equations 47 and 48 are artifacts of the integration of the flux equations. For the
conditions of the experiments conducted in the present study as well as the experiments
of Mori et al. [34], the pre-logarithmic terms were consistently different for
dephosphorization and rephosphorization experiments. As a result of the experimental

75
conditions, the pre-logarithmic term of equation 47 was consistently a factor 2 to 8
greater than that for equation 48. As a result, the RHS of equation 48 develops two to
eight times faster than the RHS of equation 47 increasing the slope by the same amount.
The second inconsistency between the dephosphorization and rephosphorization
experiments lies in the physical properties of the slags used in the different experiments.
In the present study, rephosphorization experiments were conducted with slags with very
low basicities. As a result, the slags used in those experiments consistently exhibited
much higher viscosities as compared with the slags used in dephosphorization
experiments. Several models of the mass transfer coefficient were presented in Section
4.4.1. In each model, slag viscosity is likely to have some inverse proportionality with
the mass transfer coefficient; however, the exact form of the proportionality is not
precisely known. Each model explicitly describes the proportionality of the mass transfer
coefficient to the diffusion coefficient. However, the film thickness of the film theory
andor the surface residence time of the penetration theory are also likely to be functions
of the slag viscosity.
For the reasons outlined above, it is believed that observed slower rate of
rephosphorization relative to dephosphorization was the result of the experimental
conditions and not a fundamental difference in the mass transfer mechanism. The
observation of a variable mass transfer parameter during both dephosphorization and
rephosphorization experiments supports the hypothesis that the two reactions are
fundamentally similar.

Variable Mass Transfer Parameter:

The apparent overall mass transfer parameter was calculated as a function of time for
dephosphorization experiments 18 - 22. This calculation was made by determining the
slope in the RHS versus time plots (Figure 45-48) for each increment of time. Where
local scatter in the data produced physically meaningless values, the slope was calculated
over a longer time increment, which was representative of the overall trend. Figure 51
shows the result of this calculation:

76
 t”
Q)
%
E
0.35 -. A Experiment #18
a Experiment #19
Experiment #20
x Experiment #22

The very near correlation of the instantaneous mass transfer coefficients for experiments
18 - 20, suggests that the observed phenomenon is consistent and reproducible.
Furthermore, the fact that a similar trend can be observed in the data of Mori et al. [34]
suggests that this is not a systematic error inherent to the experimental design. This
observed change in the apparent mass transfer parameter, suggests one or a combination
of several things:
1. The reaction area is not constant, i.e. some dynamic interfacial phenomenon
such as emulsification is occurring.
2. The overall mass transfer coefficient is changing as a function of time.
3. The dnving force for the reaction expressed in equations 47 and 48, does not
correctly reflect the true system.

Dynamic Interfacial Phenomena:

It is possible that the observed change in reaction rate is not caused by a change in the
overall mass transfer coefficient, but by a change in the slag-metal interfacial area. It is
possible that some dynamic interface phenomenon is causing a distortion of the interface

77
such that at short times, when the driving force and mass flux are large, the reaction area
is temporarily increased. It is widely accepted that emulsification can occur during
periods of intense mass transfer between two immiscible liquids. [39] The mechanisms
and criteria for emulsification and other interfacial phenomena between liquid iron-slag
systems have been studied by a number of researchers.[48,49]. Cramb and Chung[48]
and Gaye et al. [49] and others have shown that high rates of mass transfer of surface
active elements can result in a deep reductions in slag-metal interfacial tension, y. These
researchers have also shown that as the mass transfer reaction approaches equilibrium
and the mass flux at the interface decreases, the interfacial tension again returns to a
higher equilibrium value. This is shown schematically in Figure 52.
I

Time

Figure 52: Adapted from Gaye et al. [49]. A schematic representation of the interfacial tension, y, (upper
graph) and metal phosphorus content (lower graph) as functions of time.

It has been proposed that during this transient period of reduced interfacial tension,
turbulent interfacial fluid flow, driven by Marangoni forces can result in emulsification.
Under less dramatic conditions, intensified local stimng at the interface and a distortion
of the slag-metal meniscus without complete emulsification could significantly enhance
the rate of mass transfer.
Gaye et al. [49] measured the dynamic interfacial tension during a wide range slag-
metal reactions including dephosphorization. For the various oxidation reactions studied,
Gaye et al. established criteria for dynamic interfacial tension reduction in terms of the
associated oxygen flux across the interface. In the case of the dephosphorization

78
reaction, these researchers found that when the oxygen flux due to the phosphorus
reaction exceeded 0.03 g atom.m.'.s-' the interfacial tension was reduced form an initial
high equilibrium value (1200mJ /m') to a very low value c 300 mJ/m2. Furthermore,
when the oxygen flux exceeded 0.5 g.atom.m".s.' the interfacial tension was reduced to
such a value very near zero. The metal alloy studied by Gaye et al. initially contained
from 1 to 4 percent phosphorus. Also, the slags used in that study were ternary calcium-
alumina-wustite slags. Therefore, the slag-metal systems examined in the present study
differed somewhat from those used by Gaye et al. Despite these differences, the initial
oxygen flux due to the phosphorus reaction was calculated for the first five minutes of the
experiments shown in Figure 51. The results of those calculations are tabulated below.

Initial Oxygen Flux (g.atom.m".s")

Due to the Measured Rate of
Phosphorus Transfer
Experiment #18 1.38
Experiment #19 0.83
Experiment #20 1.1
Experiment #22 0.43

According to the criteria established by Gaye et al., all of the dephosphorization

experiments shown in Figure 51 were likely to under go a large reduction in the
interfacial tension during the first 10 to 12 minutes of reaction. Interestingly,
Experiments 19 and 22 resulted in the smallest initial oxygen fluxes. These experiments
also exhibited a smaller difference between the initial and final mass transfer parameters
relative to Experiments 18 and 20. These results suggest that the mass transfer parameter
is related to the mass flux through the interfacial tension. However, it remains unclear
whether the dynamic interfacial tension is affecting the slag-metal reaction area or the
mass transfer coefficient itself.
An attempt was made to find clear evidence of slag metal interfacial distortion during
dephosphorization. Cramb and Chung [48] examined the effect of silica reduction via
Fe-AI droplets upon the slag-metal interface. Several experiments were conducted by
those researchers in which Fe-AI droplets were allowed to react with the slag for various

79
lengths of time before the experiment was quenched. After cooling, the samples were
sectioned, polished and examined metallographically for indications of slag-metal
emulsification. From these experiments, the researchers were able to obtain clear visual
evidence of slag penetration into the metal phase and metal droplet separation into the
bulk slag phase.
In attempt to produce similar results for the slag metal system of the present study,
several experiments were conducted with essentially the same slag metal system as
experiment #18, but were terminated after approximately 4 minutes by quenching to
room temperature. Unfortunately, the slag chemistry of these experiments undergoes
'slag falling' or spontaneous disintegration due to low temperature phase transformations
as it is cooled. Thus, a large portion of the bulk slag phase could not be recovered.
However, in the vicinity of the slag-metal interface, the several millimeters of the slag
phase remained as a monolithic solid. The remaining slag and metal were then sectioned
vertically, polished, and examined under the optical microscope for evidence of
emulsification. Figure 53 shows the slag-metal interface at approximately 50x
magnification.

Void

Slag metal interface N

80
Figure 53: Optical micrograph of the slag-metal interface, quenched 4 minutes after the start of
dephosphorization. White spots are metal droplets suspended in the slag.

Figure 54 is a composite of several micrographs showing the slag metal interface.

Figure 54: Composite of several optical micrographs of the slag metal interface quenched 4 minutes after
the start of dephosphorization. White spots are metal droplets suspended in the slag phase.

Metal droplets can be observed in the slag phase in the micrographs shown in Figures 53
and 54. Investigation of the entire interface revealed that a “halo” of metal droplets had
formed at a constant distance above the interface in the slag phase. No evidence was
found of slag entrainment into the metal. The observation of metal droplets in the slag
phase supports the hypothesis that the slag metal interface is unstable during
dephosphorization. However, this investigation cannot be substituted for an in situ
observation of the interface while the reaction is occurring.
In a recent study, Seetharaman et al. conducted a series of sessile drop experiments
with Fe-P droplets in a highly dephosphorizing slag.[50,5 11 During these experiments,
the behavior of the slag-metal interface was observed in situ during dephosphorization
via x-ray radiography. In these experiments, an Fe-O.l%P sample of 3 - 7 grams was
placed in an MgO crucible. The sample was allowed to equilibrate at 1550°C. A pellet
of slag weighing approximately 3 grams with the initial composition:
CaO 40%
Si02 30%
Fe2Os 30%

was then added to the furnace. After the slag addition, X-ray images were taken of the
sessile drop at a rate of 1 second per minute for a ten minute period. X-ray images from

81
one of the sessile drop experiments conducted by Seetharaman et al. [50]are reproduced

I(
in Figure 55:

L
t = 0 seconds t = 10 seconds

m -
t = 70 seconds t = 231 seconds

- w-
I -
t = 278 seconds t > 300 seconds
Figure 55: X-ray images adapted from sessile drop experiments with Fe-O.l%P conducted by Seetharaman
et al. [50].

In Figure 55, the first frame at t = 0 shows the Fe-O.l%P droplet before the slag addition.
The second frame shows the sample 10 seconds after the slag addition. The droplet has
clearly changed shape dramatically, significantly increasing the surface area of the drop.
The third image, taken at t = 70 seconds, shows that the droplet has formed a neck on the
left-hand side. At this stage, of the experiment, the interfacial tension between the
droplet and the slag appears to be very low. Under conditions of more energetic fluid
flow, partial emulsification of the slag and metal is quite possible. By t = 231 seconds
the droplet has begun to pull back together, and at times greater than 278 seconds the
drop returns to a shape with a high contact angle, indicating an increase in the interfacial
tension.
As component of their study, Seetharaman et al. [5 13 conducted a dephosphorization
experiment similar to those of the present study, for the same slag-metal system of the
sessile drop experiments. The results of one dephosphorization experiment from that
study are shown blow.

82
 0.12

0.1

n 0.08
eL
Y
0.06
w
b
0.04

0.02

0
0 5 10 15 20
Time (minutes)

Figure 56: Metal phosphorus versus time data for dephosphorization experiment conducted by
Seetharaman et al. [51]

The data in Figure 56 show similar trends as were observed in the present study, i.e. a
best fit ko under predicts the rate of dephosphorization at short times and over estimates
the rate at longer times. The data in Figure 56 were replotted as the RHS of equation [48]
versus time in Figure 57.

4J

40
35
30
25
20
15
10
5
0
0 5 10 15 20
Time
Figure 57: Experimental dephosphorization data from Figure 5 1, replotted as the RHS of equation [48]
versus time. Data are adapted from Seetharaman et al. [43]

The mass transfer parameter in Figure 57 is consistent with the results of the present
study both in magnitude and the observed behavior. In the calculation of the RHS of

83
equation [48] the final slag chemistry for the experiment, as reported by Seetharaman et
a]., [51] was used to calculate the equilibrium phosphorus distribution ratio for the
system.
From the data in Figure 57, and the reported cross sectional area of the crucibles used
by Seetharaman et al., an average apparent mass transfer coefficient was calculated to be
0.01 1 c d s . An attempt was made to roughly calculate the rate of phosphorus transfer
from the sessile drop shown in Figure 55. For this simulation, several important
assumptions were made. First, the change in reaction area relating to the shape change of
the droplet was ignored. The observed shape change would result a change in a reaction
area by roughly a factor of four. The intent of this simulation is only to generate an order
of magnitude calculation. Other sources of error, including temperature and slag
chemistry variation near the beginning of the experiment will easily result in fluctuations
of the driving force for phosphorus mass transfer of least a factor of four. Also, a
constant and average mass transfer coefficient was assumed. In the calculation, a
hemispherical cap of 5 grams of Fe-O.l%P alloy was brought into contact with 3 grams of
slag of the experimental chemistry at t = 0. The numerical phosphorus mass transfer
model, described in Chapter 6, was used solve for the phosphorus content of the metal
droplet as function of time. The result is shown in Figure 58.

0.1 -

0.08 -
n
er.
Y
0.06 -
+
>
0.04 -
0.02 -

0 I I I I I

0 1 2 3 4 5 6
Time (minutes)
Figure 58: Simulated phosphorus mass transfer during the sessile drop experiments of Seetharaman et al.
~511

84
The results of the simulation in Figure 58 coincide very well with the observed changes
in the shape of the sessile drop shown in Figure 55. Most of the phosphorus transfer is
completed within 3 minutes for the conditions of the sessile drop experiments. During
the sessile drop experiments, the contact angle of the drop was observed to begin
increasing again at times greater than 120 seconds.
These results strongly support the hypothesis that a deep reduction of the interfacial
tension will occur for the conditions of phosphorus mass transfer studies in the present
work. The observed variation of the mass transfer parameter, may be tied to actual
changes in the reaction area due to slag metal emulsification or some less dramatic form
of interfacial distortion. However, this is not the only possible explanation for observed
change in the mass transfer parameter.
In Section 2.3 several physical and mathematical descriptions of the mass transfer
coefficient were discussed. The film theory proposes that mass transfer occurs via
molecular diffusion across a stagnant film of thickness:

a=-D 50
k
Based upon the calculated mass transfer coefficients for the present experiments, a
theoretical film thickness of approximately 10.’ to cm can be calculated to exist in
both the slag and metal phases. The time required to establish steady state diffusion
across such a small film thickness would be significantly less than a second. Therefore,
the transition from non-steady state to steady state diffusion could be ignored. However,
as was discussed in section 2.3, the instability of the slag-metal interface makes it
unlikely that the film theory will provide an accurate description of the mass transfer
coefficient. Alternatively, the film-penetration model [42] predicts that when the average
residence time of an element of liquid at the interface is short, mass transfer will proceed
via a combination of non-steady state diffusion into the surface elements and turbulent
mixing between the interface and bulk.
Toor and Marchello [42] suggest that the residence time of elements of liquid at the
interface is some function of the interfacial tension. Interfacial elements of liquid are
different from those in the bulk liquid because they are more stable in position and
orientation with respect to the interface. This added stability and the residence time of a
liquid element at the interface are related to the interfacial tension. Therefore, the

85
observed dynamic reduction of the interfacial tension during dephosphorization [49, 501
could have a direct effect upon the mass transfer coefficient. During periods of high
mass flux and low interfacial tension, the average residence time of interfacial elements
could be reduced by several orders of magnitude. During periods of low mass flux and
high interfacial tension, the residence time would also increase. According the
penetration model, the mass transfer parameter is proportional to the inverse square root
of the residence time. In the extreme case, the film-penetration model predicts the
possibility of transition from non-steady state diffusion at low interfacial tension and
short surface residence times to steady state diffusion at high interfacial tension and long
surface residence times. Such a transition would result in a change in the proportionality
of the mass transfer coefficient with the diffusion coefficient from a square root
dependence to a linear dependence. Such a shift could easily cause an order of
magnitude or greater change in the mass transfer coefficient.
The above argument regarding the effect of a dynamic interfacial tension upon the
mass transfer coefficient is very hypothetical in nature. Explicit relationships between
key parameters such as the interfacial tension and the surface residence time have not
been established in the literature. Although it is difficult to quantify these effects, a
mechanism can be imagined by which a dynamic interfacial tension could result in a
dynamic mass transfer parameter without emulsification or a change in the slag-metal
area.
Variable Driving Force for Phosphorus Transfer:
The observed decrease in the apparent rate of mass transfer could also be the result of
an inadequately or incorrectly defined driving force or mechanism for phosphorus
transfer. Mori et al. [34] noted a dependence of the overall rate upon the equilibrium
partition ratio, Lp. In order to reproduce this effect in the present study, an experiment
was conducted using slag with a significantly lower Lp than for the slags of previous
experiments. Figure 59 shows the results of dephosphorization experiment #21 plotted as
metal phosphorus versus time.

86
 DephosphorizationExperiment#21
0.14 c/s= 1.5
FetO = 9.0 %
0.12 MgO =8.0 %
T = 1600°C

n
0.1
L
5
Y

0.08
s + 4
0.06
Assuming ko = 0,003cm/s
_____-
0.04

0.02

0
0 20 40 60 80
Time (minutes)
Figure 59: Metal phosphorus versus time for dephosphorization experiment #21. Continuous curve
calculated via Equation [46] assuming a constant, best fit ko.

The rate and extent of phosphorus transfer was clearly less in experiment #21 than for
other experiments. Figure 60 replots the data of experiment #21 as the RHS of
equation[48] versus time.

45 - ' Dephosphorization Experiment #21

40 -

35 -
30 - A*ko = 0.01 cm 3 / s
L4
I 2 5 -
p:
20 -

15 -

10 -
A*ko = 0.02 cm 3 / s

0 10 20 30 40 50 60 70 80

Time (minutes)
Figure 60: Data from dephosphorization experiment #21 replotted as the RHS of equation [48] versus time.

87
It is apparent from Figure 60, that both the average overall rate and the change in the rate
are significantly less in experiment #21 as compared with other dephosphorization data
presented above. However, although it is only by a factor of two, there is an
unmistakable bend in the data toward a slower rate. The mass flux of oxygen associated
with the phosphorus transfer shown in the first 12 minutes of Figure 59 is on the order of
0.2 g.atom.m-2.s-1.According to Gaye et al. [49] this flux is within the range where a
reduction of the interfacial tension is likely to occur, but the interfacial tension will not
necessarily go to zero.
In review of the data of Mori et al. [34] for phosphorus transfer, these same trends are
also observed. Those researchers observed an apparent dependence of the average mass
transfer coefficient upon Lp. Although Mori et al. did not comment specifically on the
systematic change in rate, the degree to which the mass transfer coefficient changed from
beginning to end of those experiments appears to be related to the driving force for
dephosphorization as well. In experiments where the Lp was large, ko showed a large
decrease from the beginning to the end of the experiment. In dephosphorization
experiments where Lp was low, ko showed much less variation from start to finish.
In the study by Mori et al., samples taken from the bulk metal for phosphorus
analysis were also analyzed for oxygen. For most dephosphorization experiments the
oxygen content of the metal increased from an initial value typically below 50 ppm to a
final value in excess of 1000 ppm. As reported by Mori et al. [34], the bulk metal oxygen
content increased faster, but on the same time scale as the rate of change of phosphorus.
Despite this fact, these researchers assumed that oxygen mass transfer was significantly
faster than phosphorus such that the slag metal interface was always at equilibrium with
respect to oxygen. In Figure 61, metal oxygen data from the dephosphorization
experiment shown in Figures 9 and 39 are replotted.

88
 0.2
0.18
0.16
0.14
0 0.12
c1

Y
Be 0.1
3 0.08
Y
(slag4, T = 1640°C
0.06
0.04
0.02
0
0 5 10 15 20
Time (minutes)
Figure 61: Oxygen transfer kinetic data adapted from Mori et al. [34] The continuous curve indicates only
the trend not a fundamental calculation.

The data in Figure 61 show an interesting peak in the metal oxygen at approximately 5
minutes. This also correlates to the time in Figure 39, shown below, when the
phosphorus mass transfer coefficient appears to slow down.

10

e S l a g 4 , 16 40°C
4

0 5 10 15 20
T ime (minut es)
Figure 39: Adapted from Mori et al. [34]. Data from dephosphorization experiment shown in Figure
10 replotted as the RHS of equation 48 versus time.

89
This trend of a peak in metal oxygen followed by a decay was common to all
dephosphorization experiments conducted with slags, which exhibited a large equilibrium
phosphorus distribution ratios. The experiments in which Lp was large were also the
experiments, which exhibited the largest changes in the phosphorus mass transfer
coefficient. The Lp for the experiment shown in Figures 10 and 39 was equal to 69. By
comparison, the Lp for the experiment conducted with slag 1 at 1540°C shown in Figure
9 was 34. The data from this experiment are replotted as the RHS of equation 48 in
Figure 62.

9 -
a-
7 -
6-
to 5 -
I ko = 0.015 cm/s
4 -

0 2 4 6 a 10
Time (minutes)

Figure 62: Adapted from Mori et al. [34]. Data from dephosphorization experiment shown in Figure 9
replotted as the RHS of equation 48 versus time.

The mass transfer coefficient in Figure 62 only varies from 0.033 c d s at the beginning
of the experiment to 0.011 cm/s toward the end. The data for the oxygen content of the
metal for this experiment are replotted in Figure 63.

90
 0.12

0.1

n
0.08
0
Y

Be 0.06
*
3
0.04

0.02

0
0 5 10 15
Time (minutes)

Figure 63: Oxygen transfer kinetic data adapted from Mori et al. [34] The continuous curve was
plotted assuming liquid phase mass transfer controlled kinetics and a best fit mass transfer parameter.

The data in Figure 63 do not show the peak in the metal oxygen content, which was
observed in Figure 61. In order to calculate the continuous curve in Figure 63, it was
assumed that liquid phase mass transfer was controlling oxygen transfer into the bulk
metal, and that the final metal oxygen content represented equilibrium. Under this
assumption, the flux equation for oxygen is given by:

51

The term [%O], in equation 51is the oxygen content of the metal at the slag metal
interface. Assuming that the interface is at equilibrium, and that the final metal oxygen
represents that equilibrium, equation 5 1 can be rewritten in terms of the final metal
oxygen:

-
d%[O] - Akop ([%0Is- [%O])
- -- 52
dt wm

Equation 52 can be integrated analytically to give the equation:

91
 1
[%OlS -[%O] 53
- kot = w,{I
Ap [%OIf - [%OI,
The continuous curve in Figure 63 was calculated according to equation 53. The data of
Figure 63 were replotted as the RHS of equation 53 versus time in Figure 64.

25

20

15
r
cr)
p:
10

0
0 5 10 15
Time (minutes)
Figure 64: RHS of equation 53 versus time.

Figure 64 is typical of the results for oxygen transfer in the metal for dephosphorization
experiments conducted by Mori et al. with low phosphorus distribution ratios. Clearly,
the assumptions necessary for equations 52 and 53 to be used are not true for experiments
that exhibited a peak and decline in the metal oxygen. Equation 51 should still be valid;
however, the interfacial metal oxygen is changing during the heat, and the final bulk
metal oxygen content does not describe the condition at the interface at all times during
the experiment.
The focus here by the author on oxygen is for several reasons.
1. The oxygen potential at the slag-metal interface is one of the strongest influences of
the equilibrium phosphorus reaction.
2. Oxygen is surface active in iron, where as phosphorus is not particularly so.
Accordingly, previous researchers examining the effects of phosphorus transfer upon
the interfacial tension have done so with respect the associated oxygen flux, not the
phosphorus flux. [49] The mechanism by which oxygen is exchanged across the slag-
metal interface should have a significant effect upon the properties of the interface
during the reaction.

92
3. Clearly, the data of Mori et al. [34]show that something unexpected is occurring with
respect to oxygen. If it is real, the peak and decay in metal oxygen content for
dephosphorization experiments with a large Lp, cannot be explained by conventional
hypotheses regarding the transfer of reactants to and products away form the interface
during dephosphorization.
One simple explanation of the observed peak and decay in metal oxygen content can be
rationalized. In the experiments for which this phenomenon was observed, the initial
oxygen mass flux, calculated directly from the oxygen data, was on the order of 8
g.atom.m-'.s-'. This is a massive flux and would undoubtedly result in significant
emulsification. Therefore, metaI samples taken during the first 5 minutes of reaction may
have contained large amounts of entrained slag, which would significantly elevate the
observed total oxygen content. The decay in oxygen content after 5 minutes can be
rationalized as resulting from the separation of entrained slag inclusions from the metal
phase. This explanation is as likely or more likely than any other.

93
 ChaDter 5

Plant Trials
Laboratory experiments can be an effective tool in evaluating the fundamental
equations, mechanism, and rate parameters for chemical reactions, which occur during
steelmaking. However, the mass transfer conditions of the EAF or any industrial reactor
differ significantly from conditions, which can be effectively simulated in the laboratory.
Differences in the stirring conditions and degree of slag metal mixing in the EAF versus
those in laboratory experiments can result in variations in the mass transfer parameter by
several orders of magnitude. Therefore, a number of plant trials were conducted in order
to gain a better understanding the parameters, which influence mass transfer during EAF
steelmaking. The results of these industrial scale experiments, as well as the results of
previous researchers, will be discussed in the following sections.
5.1 Pennsylvania Steel Technologies, Inc-Bethlehem Steel Mass Transfer Trials
Several controlled trial heats were conducted in order to evaluate the mass transfer
parameter for the EAF. It should be noted that differences in specific melting practices
from one plant to the next will lead to variations in the mass transfer parameter for
different operations. The degree of slag foaming and amount of slag metal mixing are
most strongly effected by the evolution of gas, primarily carbon monoxide, within the
steel melt, at the slag metal interface, andor within the slag layer. The location, extent,
and rate of gas evolution is affected by several furnace operating parameters, such as:
Amount and physical characteristics of charge carbon and injected carbon
Bath carbon at melt-in, e.g. scrap mix plus additional carbon
Slag chemistry at melt-in, e.g. flux practice, scrap characteristics, melting practice
Amount, characteristics, and method of addition of DRI/HBI materials
Amount and method of oxygen injection
Use of bottom stirring
The mass transfer parameter of a given furnace will be influenced by the amount of slag-
metal mixing, the physical dimensions of the furnace, the furnace power and melting rate,
as well as the electromagnetic characteristics of the arc. Irons et al. [52, 531 have
attempted to evaluate the relative influence of these parameters on stirring in the steel

94
bath through computational fluid dynamic modeling of the EAF. Due to the extreme
complexity of a high temperature three (or more) phase system with moving boundaries,
multiple simultaneous multi-phase chemical reactions, and large chemical, temperature,
and surface tension gradients it is extremely difficult to precisely characterize the EAF
through fluid dynamic modeling. As a parametric comparison, the fluid dynamic
modeling of Irons et al. indicates that, in the absence of bottom stirring, CO evolution
within the bath is the most significant driver for fluid flow in the steel in an AC furnace.
Also, the evolution of gas bubbles at or below the steel-slag interface significantly
increases the slag-metal interfacial area by enhancing the amount of slag-metal inter-
mixing. For this reason, it is assumed that, in the absence of bottoms stirring, the degree
of slag-metal mixing effected by the various influences listed above, will dominate the
mass transfer parameter. Therefore, the mass transfer parameter should scale with some
proportionality to the superficial gas velocity at the imaginary flat-bath slag-metal
interface. However, gas generated within or above the slag layer will have limited effect
on the mass transfer parameter. For the above reasons, it is recognized that no universal
mass transfer parameter can be defined for all electric furnaces. However, the results of
the trials reported here should be characteristic of a modern electric furnace of similar
physical dimensions with an effective foamy slag practice.
Controlled industrial scale mass transfer experiments were conducted on a 150 ton
DC electric furnace by Pennsylvania Steel Technologies, Inc-Bethlehem Steel (PST).
Some important operating parameters of the furnace used are given below:

95
Table 3: Operating Parameters of PST - EAF
Power UHP-DC
Capacity 150 tons 3 bucket charge
Diameter 23 feet
Tap to Tap Approx. 60 minutes
Typical Oxygen Usage Melting - 89,040 SCF
Refining - 46,200 SCF
Total - 1200 SCF/Ton
Injected Carbon* 600 - 800 lbs
Charge Carbon 9000 Ibs in 3rdbucket charge
Flux - Pre-blended" 77.2%Ca0 - 6562 lbs
14.6%Mg0 - 1241 lbs
4.6% Si0,- 391 lbs
Total - 8500 lbs *Addition completed by
melt-in of 2'ldbucket
Slag Flushing During Melting =: 0
During Refine - variable.. .O to 10,000 lbs
Hot Steel Heel 37 tons
Slag Heel 4 to 12 inches -- 8600 lbs to 22,000 lbs

For these plant trials, temperature readings and samples were taken of the slag and
metal as a function of time as soon as a reliable temperature reading could be made
(approximately 10 to 15 minutes after the final charge.) This resulted in the collection of
approximately 4 slag, metal, and temperature samples during the 10 to 15 minute refining
period. The slag and metal samples were then analyzed. Due to the focus of this study,
it was intended to examine the rate of phosphorus transfer between the metal and slag in
order to determine the mass transfer parameter for this furnace. As was discussed at
length earlier, the equilibrium phosphorus distribution is a strong function of temperature
and slag chemistry. Therefore, even in the unlikely case where the slag and steel melt in
at the equilibrium phosphorus distribution, the dynamic variation of that equilibrium with
changes in bath temperature and slag chemistry during refining will drive further
phosphorus mass transfer. The lag in the actual measured phosphorus distribution behind
the calculated dynamic equilibrium distribution should give a good indication of the
phosphorus mass transfer parameter for this furnace.
In order to calculate the dynamic phosphorus equilibrium, the slag chemistry and bath
temperature must be known as functions of time during the period of interest. Figures 65
and 66 show the actual temperatures and slag chemistries, which were measured for one
experimental heat.

96
 1660
Heat #760

1640
u
d

*L
~
.
I 1620 -
a
1600 -
*L

2* 1580
c
1560

1540 I I I I I

0 2 4 6 a 10 12

Time (minutes)

Figure 65: Bath temperature versus time.

35

30 --A- FeO

25
'' //------ +
-c-

" / - - / e ,

0
0 2 4 6 8 10 12

Time (minutes)

Figure 66: Slag chemistry versus time. All components are in weight percent.

The temperature profile shown in Figure 65 is fairly typical. As melt-in of the final
charge nears completion, the temperature of the bath begins to increase rapidly. The slag
chemistry profile indicated in Figure 66 shows similar trends to those seen in other
furnaces, as will be discussed in the following sections. The initial slag formed during
melting is high in FeO due to rapid oxidation of the scrap pile by injected oxygen. The

97
FeO content of the slag subsequently decreases due to chemical reduction via bath
carbon, charge carbon, and injected carbon, as well as due to dilution by late melting
components of the slag. Although all flux additions are made to the furnace prior the
addition of the third scrap bucket, a significant portion of the slag can be expected to
freeze off when the final charge is dropped. The continuous increase in the V ratio of the
slag in Figure 66 indicates the continuous dissolution of solid lime. Both the reduction of
iron oxide from the slag and the continuous dissolution and melting of solid slag
components in the furnace will result in a dynamic total slag mass during the refining
period. This point is further complicated by slag flushing.
Figure 67 shows the measured slag and metal phosphorus for the above heat.

0.025 0.25

0.02 0.2

0.015

0.01 0.1

0.005 0.05 5
rn
W
U
0 0
0 2 4 6 8 10 12
Time (minutes)

Figure 67: Slag and metal phosphorus content versus time.

Clearly, the metal phosphorus content is decreasing during the refining period for this
heat. However, if negligible phosphorus losses to the gas phase are assumed, the initial
simultaneous decrease in the slag and metal phosphorus content confirms that the slag
weight and possibly the metal weight are increasing. In Figure 68, the slag chemistry and
temperature data of Figures 65 and 66 were used to calculate the equilibrium phosphorus
distribution according to two different correlations. The observed slag-metal phosphorus
distribution is also shown.

98
 5 10.0 - -+-,---#A

4
II
5.0 ‘ r-------’r---

0.0 I I I I I

Figure 68: Calculated equilibrium and observed slag-metal phosphorus distribution

versus time

Figure 68 indicates that the observed slag-metal phosphorus distribution is below the
equilibrium value during the entire heat. The flux equation for phosphorus in the metal is
given by equation 28:

a[%P]
--_ - ApmkrnB P I - F P i I] 28
at Wm

Where the driving force for mass transfer is proportional to the difference between the
bulk phosphorus content and the phosphorus at the metal-slag interface. It can be
assumed that the slag-metal phosphorus distribution at the interface will be at the
equilibrium dictated by the temperature and slag chemistry as calculated in Figure 68. In
order to calculate the actual interfacial phosphorus values, some assumptions must be
made regarding the masses of liquid steel and slag, which are present and regarding the
total amount of phosphorus in the system. As a first approximation, it was assumed that
all of the steel, 150 tons, is liquid and 10 tons of liquid slag is present. Based upon the
scrap charge, it was estimated that the 150 tons of steel carried 69 pounds of phosphorus
into the furnace, and the slag heel contained an additional 32 pounds of phosphorus.
Based upon these assumptions the interfacial metal phosphorus content was calculated

99
according to the Ide et al. equilibrium correlation. The results of this calculation and the
measured bulk metal phosphorus content are shown in Figure 69.

0.04 -
n
-.--t- Measured bulk wt??(P)
k 0.035 -
eR - Calculated interfacial wt%(P)
+

B 0.03
0.025
0.02
0.015

0.01
0.005

0
0 2 4 6 8 10 12
Time (minutes)

Figure 69: Measured bulk metal phosphorus and calculated metal interface phosphorus
versus time

As discussed earlier, due to the variation of the equilibrium interfacial phosphorus value,
the flux equation for phosphorus transfer in the steel can not be integrated analytically.
However, the flux equation can be integrated numerically as a sum of integrals over
discrete time intervals according to the following equation:

54

As with the analytical solution presented in Chapter 4, equation 54 can be rearranged

such that the mass transfer parameter is isolated on the left hand side of the equation:

55

Therefore, by plotting the right hand side of equation 55 versus time, the slope of the
resulting line should be equal to the mass transfer parameter. As mentioned in Section
4.4.1, the term on the left hand side of equation 54 has been used in previous lunetic
evaluations of oxygen steelmalung vessels. The term:

100
 49

has been shown to remain roughly constant for different oxygen steelmaking vessels.
This is due to the scaling effect of vessel size and oxygen blowing rate in oxygen
steelmaking vessels. The same scaling effect is not expected to hold for electric arc
furnaces. Thus, this term is not as useful in defining mass transfer in the EAF.
In Figure 70, the summation expressed in equation 55 was plotted versus time.

4.E+07 /
/

3.E+07

Ak = 8.4*10 4 cm3/s
W 2.E+07

1 .E+07

Ak = 2.4*10 cm3/s
O.E+OO’ 1

0 2 4 6 8 10 12
Time (minutes)

Figure 70: RHS of equation 55versus time. The observed slope is seen to vary between
2.4*104cm3/sand 8.4*104cm3/s.

When mass transfer parameters calculated in Figure 70 are converted to the form of the
rate parameter expressed in equation 49, it is found that rate ranged between 0.066 min-
and 0.24 min-I. By comparison, equation 49 has been evaluated for the BOF and was
found to be on the order of 1 min-I. The average value of 0.15 m i d for this trial is in
good agreement with the prediction that stirring in a modern EAF will be similar to, but
less than that in a BOF. When the interpolated continuous curves in Figure 70 are
considered in addition to the discrete experimental points, the resultant plot is given in
Figure 7 1.

101
 4.E+07

3.E+07

w 2.E+07

1 .E+07

O.E+OO
0 2 4 6 8 10 12
Time (minutes)

Figure 7 1: Continuous evaluation of equation 55 along continuous curves in Figure 69.

The curve in Figure 71 suggest that in the late stages of melt-in and in the early stages of
the refining period, mass transfer is significantly slower as compared with later in the
heat. This is expected as the slag viscosity is continually decreasing as the temperature is
increasing. Unmelted solids in both the steel and slag will inhibit fluid flow in both
phases. In addition, the rate of oxygen injection is increased by approximately 30%
during the refining period as compared with during melting. The oxidation of carbon in
the melt by FeO in the slag and injected oxygen will be enhanced by a more fluid slag
and increased oxygen blowing. Increased carbon monoxide generation will in turn
further increase fluid flow and slag metal mixing leading to a self amplification of the
mass transfer parameter.
The temperature and slag chemistry profiles of a second heat are shown in Figures 72
and 73.

102
 1600

1590
e. 1580
8

+
5 1570
m
1560
m
E
Q
1550
I-
1540

1530
0 2 4 6 8 10
Time

Figure 72: Bath temperature versus time

35

30

25

20 -+-/- .--+--- CaO

Be
5 15

10

0
0 2 4 6 8 lo
Time

Figure 73: Slag chemistry versus time. All components are in weight percent.

The temperature profile in Figure 72 indicates an additional complication of these

experiments. The 35 "C drop in temperature between the first and second sample is most
likely the result of a cave in of unmelted 'cold' scrap, which had been stuck against the
side wall of the furnace. For many common grades produced by PST, up to 12% iron is
included in the scrap mix. This material is typically among the more dense components
of the scrap mix and can contain up to O.lwt%P. The effect of late melting scrap on the

103
bath temperature and chemistry will also have a secondary effect on the slag chemistry.
This is shown in Figure 73. The lower V ratios indicated in Figure 73 as compared with
Figure 65, suggest that a greater portion of the lime remained solid within the furnace
until later in the heat. Based upon a heat balance of only the solid and liquid steel phases
(neglecting the slag phase), the observed temperature drop correlates to approximately
14,000 lbs of late melting scrap. The metal and slag phosphorus are plotted as a function
of time for this heat in Figure 74.

0.025 - -0.25
,-E--, ,a
0.02 - / / #
# \
\ + # #
H 0
-0.2
6

'0.15

-0.1
- fl- - wt%p)
0,005 -0,05

01 -0

Figure 74: Slag and metal phosphorus versus time.

The slag and metal phosphorus increase between the first and second samples in Figure
74 is consistent with the melt-in of approximately 14,000 lbs of scrap or iron with
0.1wt%P. The equilibrium interfacial metal phosphorus content was calculated assuming
a constant slag mass but with a changing liquid steel mass. In the calculation of the first
point, 14,000 lbs of scrap with O.l%P were not included in the phosphorus balance for
the furnace. It was assumed that all scrap and the contained phosphorus was melted in at
the second data point at 4 minutes. The results of the interfacial phosphorus calculation
are shown in Figure 75.

104
 a
n

-
Y
0.025
> calc . wt??[ P) i
+I'

e?
0.02 -
A

CL
rn __---- /---
---__
a
0 0.01 -__--+----

a
-
I

(R 0.005
+
E 0 I I I I

0 2 4 6 8 10
Time (minutes)

Figure 75: Measured bulk and calculated interfacial metal phosphorus versus time.
Clearly, the uncertainties in the masses of the liquid steel and slag introduce
significant error in the first 4 minutes of Figure 75. The metal phosphorus increases
between the first and second samples despite the clear driving force for
dephosphorization, i.e. the interfacial phosphorus is lower than the bulk phosphorus. The
continuous smoothed curves in Figures 74 and 75 above are probably poor
representations reality. It is more likely that the initial metal phosphorus increase, due to
the late melting slag, was actually greater than is shown, and that some phosphorus
transfer to the slag had already occurred by the time of the second sample. It is possible
to develop a more complex scrap melting model in order to accurately evaluate these
data. However, such a complex model will not be used here in the determination of an
experimental parameter.
For one trial heat, a controlled addition of 100 lbs of commercial ferrophosphorus was
made to the furnace shortly after melt-in of the final bucket. Temperature readings and
samples were taken of the slag and metal immediately prior to the addition and as a
function of time thereafter. The added material in this trial was an Fe-22%P lump
material, which was expected to cause an instantaneous increase of 0.007 wt%P in the
metal bath. The temperature and slag chemistry profiles for this heat are shown in
Figures 76 and 77.

105
 640
630
620
610
1600
s
1590
L
1580
E
QI
1570
1560
1550
1540 ! I I 1

0 5 10 15 20
Time (minutes)

Figure 76: Temperature versus time.

100 Ibs of FeP

Heat #853
o ! I I I I

0 5 10 15 20
Time

Figure 77: Slag chemistry versus time. An addition of 100 lbs was made at t = 3
minutes.

The measured slag and metal phosphorus contents are plotted versus time in Figure 78.

106
 0.03 I I

n
L
E 0.025
b
-P

m, 0.02
m
3
/'
a 0.015 7 I
a
m
9
a 0.01
L
I

0.005
r Heat #853

0 5 10 15 20
Time
Figure 78: Measured bulk slag and metal phosphorus content versus time.

The variations in the slag and metal phosphorus shown in Figure 78 are somewhat
unexpected and indicate that continuous melting of solids in both the slag and metal was
occurring after the addition. From the temperature profile in Figure 76, it is likely that a
significant amount of unmelted scrap was present until the time of the third sample at t =
8 minutes. The simultaneous decrease in the slag and metal phosphorus after t = 10
minutes suggests that continuous dissolution of unmelted slag is leading to the dilution of
the slag phosphorus. During this particular heat, very little slag flushing occurred.
Therefore, an approximate conservation of mass for phosphorus can be assumed for the
furnace for this heat. In order to maintain a constant total mass of phosphorus in the
furnace across the final 4 samples, the assumed mass of slag was increased by 25% in
each of the last two points. The behavior of the melt in the first 8 minutes is too complex
to model at this time without making significant assumptions. Therefore, only the final
four samples were used in evaluating the mass transfer parameter during this heat. The
measured bulk and calculated equilibrium interfacial metal phosphorus contents are
plotted versus time in Figure 79.

107
 -K
n
0.016 1
L
0.014
3
* 0.012

2
5

0.01
L
0.008
a
a
0.006
*Measured bulk wt%[Pl
-
m
0.004 ~~~

Calculated interface wt#[P]

g
r
0.002

0
0 5 10 15 20
Time (minutes)
Figure 79: Measured bulk and calculated interfacial phosphorus content versus time.
Only the final four samples of this heat are considered due to uncertainties in the amounts
of solids in the steel and slag phase.

The calculated phosphorus content at the steel-slag interface shown in Figure 79 is

significantly influenced by the assumption of a varying slag weight. Furthermore, the
ability to accurately determine the deviation of the bulk phosphorus from that at the
interface is critical in determining the driving force for mass transfer. Uncertainty in the
driving force for mass transfer leads to obvious difficulties in determining the mass
transfer parameter. In Figure 80 the RHS of equation 53’ is plotted versus time assuming
the concentrations indicated in Figure 79.

108
 5.E+08

4.E+08

3.E+08
w
2.E+08

1 .E+08

O.E+OO
0 5 10 15 20
Time (minutes)

Figure 80: Continuous evaluation of equation 55 according to the data in Figure 79.

As in Figures 70 and 7 1, Figure 80 shows a continuously varying mass transfer

parameter. It is believed that the nearly all of the steel charge was liquid during the
period, which is evaluated in Figure 80. Thus, the initial mass transfer parameter is
higher than was calculated in the early stages of Figure 80. Between 10 and 12 minutes
the rapid increase in temperature results in a drop in the equilibrium phosphorus
distribution and interfacial metal phosphorus. Thus, there is a temporary slowing of the
rate of mass transfer as the bulk and equilibrium metal phosphorus contents converge. In
the final portion of the sampling period, the temperature increase leads to an in crease in
both the lime content of the slag and in the total slag mass. In this final period, the
calculated mass transfer parameter is very rapid, approaching that which is observed
during oxygen steelmaking.
However, this calculation should be viewed with a high degree of skepticism. The
equilibrium interfacial metal phosphorus content is inversely proportional to the liquid
slag mass. The deviation of the bulk metal phosphorus from that at the interface defines
the driving force for mass transfer. Under estimation of the liquid slag mass, and thus the
driving force, would tend to exaggerate the mass transfer parameter. Therefore, the
stated uncertainty in the precise amount of liquid slag in the furnace is likely to cause
some degree of uncertainty in the calculations shown in Figure 80.

109
 Table 4 lists the results of the trials at PST in terms of the three measures of mass
transfer kinetics, which have been used in the present study, the apparent mass transfer
coefficient, the mass transfer parameter, and mass transfer time constant of equation 49:

Mass Transfer Mass Transfer AP&,

Coefficient, ko Parameter, A*ko w m
(assuming furnace area)
cmfs cm3/s min.'
High Value 0.99 3.5*105 1
Low Value 0.05 1.7" 104 0.05
Best Fit Value 0.25 8.7*104 0.25
BOF --- --- 1
Laboratory Expts. = 0.01 = 0.07 = 0.16

In Table 4, the apparent mass transfer coefficient was calculated for the trials at PST
assuming a slag-metal area equal to the cross sectional area of the furnace. The
calculated apparent mass transfer coefficient for the PST furnace ranged between 0.05
and 1 cmfs. The laboratory kinetic experiments performed according to the procedure
outlined in Section 3.4 were conducted in a very well mixed system. It is likely that the
bulk fluid flow in the EAF is much lower than that in the laboratory scale experiments.
Based on this consideration and the calculated values in Table 4, it is likely that the actual
slag-metal area is roughly 50 to 100 times greater than the cross sectional area of the
furnace. Considering the volume of gas, which is injected and/or produced in the
furnace, it is reasonable to expect that slag-metal mixing will result in a two order of
magnitude increase in the slag-metal area.
5.2 North Star Steel - Texas EAF Trials with DRUHBI
A series of plant trials with various scrap substitute materials were completed at North
Star-Texas. The purpose of the trials at NSST was to determine the effects of DRUHBI
metallization and carbon content, upon slag chemistry, and thus upon phosphorus control
during EAF steelmaking. The primary influence of these factors was predicted to be
upon the melt-in FeO content of the slag. Although, the contribution of unreduced iron
oxide from DRVHBI to the EAF slag would result in a larger total slag mass and highly
oxidizing conditions, excessive amounts of FeO in the slag can dilute the basic
components of slag resulting in a low equilibrium phosphorus distribution. The
thermodynamics of the phosphorus reaction were discussed in detail in previous sections.

110
In addition, excessive amounts of iron oxide in the slag can interfere with slag foaming,
increase electrode consumption, accelerate refractory wear, and result in costly yield
losses.
All commercial direct reduced materials contain some amount of unreduced iron oxide
within the final product. In addition, most conventional materials contain some quantity
of carbon, deposited late in the direct reduction process, in the form of either graphite
andor iron carbide. Both the amount of metallization obtained, and the specific amount
of deposited carbon vary from one commercial product to the next. Most materials are
engineered such that the minimum quantity of deposited carbon will balance the moles of
unreduced iron oxide in the material. Additional carbon beyond what is required for the
final reduction of unreduced iron oxide will either enter the liquid steel melt, react with
the slag, or be available for combustion depending upon the method of addition and
melting in the electric furnace.
Goldstein and Fruehan [54], as well as others, have conducted investigations of the
final reduction of unreduced iron oxide by deposited carbon in commercial direct reduced
pellet products. Qualitatively, this reaction has been shown to be very fast at
temperatures above 1100°C. Goldstein et al. concluded that the rate of this final
reduction during rapid heating was most likely limited by heat transfer. In the
experiments of Goldstein et al., both commercial and more homogenous repressed pellets
were dropped into a liquid slag or a hot empty crucible. The rate of the reaction was
measured by measuring the rate of evolved gas. Under all conditions studied, this
reduction took between 10 to 20 seconds, which is similar to the time required for the
heating of the pellets from 800 to 1500 "C. Figure 81 shows the results of two
experiments by Goldstein et a1.[54] where the fraction of conversion of the unreduced
iron to metallic iron is measured as a function of time. Figure 82 shows the results of the
heat transfer model used by Goldstein et al. showing that the heating rate is very close to
the rate of final reduction within the pellets during heat up.

111
 e Homogeneous
1.0 - o Commercial Q o8 o 8 8 8 * o o 8 c y
0O O O i P
0.8 - 0 0. -
. e'
0O

.
u 0.6 - 80
e -
0
c 0
e8 -
0
'E 0.4 -
U
E eo
0.2 -
'0 -
- 8*e 0
0
0 . 0 : : ' I ' I ' I ' I ' I '

Figure 81: Degree of final reduction of residual iron oxide and

deposited carbon in commercial DRI. Data adapted from Goldstein et al.
[541

3800 f I I I 1 1
I

A CentralNode
Internal Node
o SuifaceNode

0 4 8 12 16 20
Time (s)
Figure 82: Temperature profile within DRI pellets as modeled by
Goldstein. [54]

From the results of Goldstein and Fruehan [54],it can be concluded that the final
reduction of iron oxide by deposited carbon in DRI will occur prior to melting during
either batch or continuous charging into the electric furnace. Therefore, for materials,
which are oxygen limited or contain enough carbon to balance the quantity of unreduced
iron in the material, no iron oxide should remain in the material at the time of melting.
The fate of excess deposited carbon in direct reduced materials is somewhat less clear.

112
For the conditions of continuous charging, excess carbon contained in DRI may react
directly with FeO contained in the slag. For batch charging practices, the excess carbon
will most likely first enter the liquid steel melt and then react with the slag layer or
injected oxygen during the refining period.
The Circal HBI material produced by Cliffs Reduced Iron, contains no carbon as it is
reduced via hydrogen. However, this material contains unreduced iron oxide similar to
other conventional direct reduced materials. Due to the quality of the ore from which it is
produced, the total number of iron units available in this material is very high (>97%
total iron.) However, the total number of recoverable units may be lower than for carbon
containing material. As described in Chapter 4,several laboratory experiments were
conducted where Circal HBI was melted in a high purity argon atmosphere. In these
melts, the Circal material produced a metallic yield of approximately 85%. The
remaining 15% of the material formed a very high FeO slag layer, which did not react
appreciably with the metal. Based on the discussion above, it is likely that when the
Circal material is melted in the electric furnace, the residual FeO in the material will first
report to the slag phase to be reduced andor flushed from the vessel during melting and
refining. The potential contribution of additional FeO to the slag by DRVHBI, and the
subsequent rate and degree of reduction of those iron units by carbon in the steel bath and
injected carbon comprised the focus of the trials at NSST.
Similar to the trials at PST, slag and metal chemistry and temperature data were
collected as a function of time during the refining period for the trials at NSST. Data
were collected for heats with the following scrap charges:

## of Trial Heats Scrap Pig Iron AIR-DRI Circal HBI

2 100% 0% 0% 0%
10 75% 0% 25 % 0%
10 72% 14% 0% 14%
5 70% 10% 0% 20%
5 80% 0% 0% 20%

113
The portion of the charge referred to generically as ‘scrap’ was a combination of plate
and structural, #2 shredded, and #2 heavy melting scrap steels. Within each set of trials,
the source and ratios of the different scrap materials were held constant in attempt to
maintain some consistency within the trials. The pig iron, DRI, and HBI materials used
in these trials possessed significantly different carbon and phosphorus contents. The
chemistries of these materials are listed below:

Table 6: Chemistries of Scrap Substitute Materials used in NSST Furnace Trials.

The AIR-DRI contains more than twice as much carbon as is required for the reduction of
the contained FeO. A one to one weight ratio mixture of Circal HBI to Pig Iron also
provides twice the amount of carbon required for the reduction of the FeO contained in
the Circal. A two to one weight ratio of Circal HBI to pig iron results in an approximate
molar balance of the iron oxide in Circal with the carbon in the pig iron.
For the two trial heats with a 100% scrap charge and the 10 trial heats with 25% DRI,
a minimum of three slag samples were taken during the refining period. A minimum of
two metal samples and two temperature readings were taken as soon as the most of the
metallic charge had melted in. For some heats, more samples were obtained. In all cases
the sampling period ranged from 10 to 35 minutes in duration with 3 to 8 minutes
between samples. For the 20 trial heats with HBI and Pig Iron, only two samples were
obtained from both the metal and slag. Some important operating parameters of the
furnace used in the NSST trials are listed below:

114
Table 7: Operating Parameters of NSST EAF -
Power AC
Capacity 150 tons 3 bucket charge
Diameter 23 feet
Tap to Tap Approx. 85 minutes
Typical Oxygen Usage 42,000 SCF
Total =: 300 SCF/Ton
Injected Carbon* 2000 Ibs *Injected during refining period
Charge Carbon* 2000 Ibs *charged after 1"bucket
Flux CaO- 70001bs - added during melt-in of lst,2'ldand 3rd
buckets

Mg0-28001bs -1/2 back charged during turn around

and 1/2 charged during refining period
Slag Flushing Kghly variable =: 10,000 to 18,000 lbs
Hot Steel Heel 5 to 10 tons
Slag Heel 4 to 6 inches = 8600 lbs to 11,000 lbs

As explained above, the purpose of the trials at NSST was to attempt to determine the
effects of DRUHBI metallization and carbon content, upon slag chemistry. Multiple slag
samples during each heat were desired in order to determine the bulk changes in the slag
chemistry from melt-in until tap. Typically, the power on time between the dropping of
the third bucket and the time at tap is on the order of 40 minutes. A bulk liquid slag
phase is formed approximately 10 to 15 minutes after the dropping of the third bucket.
As was exhibited for the trial heats with 100% scrap and a scrap - DRI charge, the
maximum length of the slag sampling period is on the order of 20 to 30 minutes.
Unfortunately, for the trial heats with HI31 and Pig Iron, the time interval between the two
samples ranged from 3 to 15 minutes. Therefore, the heats with Circa1 HBI and Pig Iron
were not as useful in evaluating the dynamics of the slag system as had been hoped. In
addition, it should be pointed out that a significant modification of the furnace oxygen
injection system was made between the completion of the trials with DRI, and the trials
with HBI and Pig Iron. This unfortunately complicates direct comparison of the DRI and
HBI heats. The equipment modification was also accompanied by an upgrade of the
control and monitoring system, which allowed for the detailed recording of the oxygen
and natural gas usage for each heat.

115
 Some very interesting observations were made from the first sets of trial heats
conducted with 100% scrap and 25%DRI - 75%scrap. In contrast to a BOF, the data
from these heats indicate that the particular furnace practice used at NSST melts-in at a
fairly high FeO level in the slag, which is reduced as the heat proceeds. A similar trend
can be seen clearly in the data from PST shown in Figure 66. This decline in the FeO
content of the slag is a combination of actual reduction of the FeO from the slag via
reaction with the steel melt and injected carbon, as well as continuous flushing and

Fl
dissolution of lime and MgO additions. The slag chemistries of two heats as are plotted
versus time are shown in Figures 83 and 84.

45 !+ - - - -- --- L
.

------.
- \ - \

-
40 - - --. - - FeO
Nm-
-t-Si02
35 -
30 - P205
X
25 -
b
Y

20 -> c
-
-
15

10
- - r).

5
-
A

0-7
Y

I
-
Y

I
.. I
m!

0 5 10 15 20
Time (from first sample)

Figure 83: Wt% of slag component X versus time for a 100% scrap heat. The first
sample was taken 15 minutes after the dropping of the third bucket.

116
 50 -

40 -
8
4

C 30 -

0)
A
., Y
I 1 I I I

0 5 10 15 20 25 30 35
Time (from first sample)

Figure 84: Wt% of slag component X versus time for a 25% DRI - 75% scrap heat. The
first sample was taken 7 minutes after the dropping of the third charge.

The trends observed in Figures 83 and 84 will be discussed in greater detail in a later
section of this paper. These data are presented here only to give the reader an indication
of the trends, which were observed in these trials. Toward the end of the refining period,
the slag chemistry appears to approach a steady state composition. No flux additions
were made late in the refining period for any of the heats at NSST. It is likely that all of
the metallic charge, and most of the flux additions were molten by the time of the final
slag sample. Also, due to the consistent availability of final slag chemistry data for all
trial heats, an attempt was made to identify trends in the final FeO contents of the slags
for the different scrap charges. The average yield and average final FeO content of the
slag for each of the different charge mixes are shown in Table 8.

117
Table 8: Average Cast Tons Yield and Final Furnace FeO for the Various Scrap
Charge Mixes of the NSST ‘rials
Scrap Mix Yield Final wt % Ratio of
FeO in the Slag Carbon : FeO in Scrap
Substitute Materials
100% Scrap 91.6 % 42.5 wt% N/A
25% AIR-DRI - 75% Scrap 89.5 % 37.4 wt% 2: 1
14%Circal - 14%Pig Iron - 86.4 % 37.4 wt% 2: 1
72% Scrap
1O%Pig Iron - 20% Circal - 79.5 % 42.1 wt% 1:1
70% Scrap
20% Circal - 80% Scrap 85.12 % 46.5 wt% 0:1

The yield listed in Table 8 was calculated as the number of quality cast tons per ton of
charge material. However, significant additional yield losses can occur between the
furnace and the billet yard, as suggested by the apparent lack of correlation between iron
loss to the slag and cast ton yield. A yield term based upon cast tons of good billets,
averaged over a large number of heats, is a very good indicator of overall steel plant
performance. However, this term will not be useful in evaluating the relative
performance of the electric furnace for different charge materials, unless the study is
conducted over a large number of heats in a single sequence. For a study such as the
present one, a much better indicator of furnace performance is the number of tons of
liquid steel tapped to the ladle per ton of charge material. Unfortunately, ladle weights
could not be measured during these trials.
Another furnace performance parameter, which is closely related to furnace yield is
the FeO content of the slag. Loss of iron units to the slag during melting that are not later
recovered, represents the largest single source of yield loss for the electric furnace.
Therefore, in Table 8, there does appear to be a correlation between the final FeO content
of the slag with the scrap charge mix. As mentioned earlier, additional oxygen injection
capacity was added to the furnace between the trials with DRI and those with HBI.
However, a simultaneous increase in the amount of charge carbon should have more than
compensated for the extra oxygen in the system. Due to these inconsistencies between
the trials, only qualitative comparisons will be made between these trials. According to
the data in Table 8, as the molar ratio of carbon to iron oxide decreases from 2: 1 to 1:1 to
0: 1, the amount of iron oxide in the slag increases. In this ratio, only additional carbon
introduced with the metallic charge, either in the DRI itself or in pig iron, is considered.

11s
For all of the trial heats with Circal HBI, the same charge carbon practice was used.
Therefore, charge carbon was not considered a variable in evaluating these trials. The
amount of injected carbon, which was used in each heat was regulated by the furnace
operator. Although this was a variable in these trials it is not a quantity, which is
measured at NSST. Therefore, the amount of injected carbon was assumed to be
approximately constant from heat to heat. It will be demonstrated later, injected carbon,
may not play a major role in determining final FeO content of the electric furnace slag.
Therefore, it was assumed that by neglecting the variations in the amount of injected
carbon significant error would not result. On average, the 10 heats with 25% AIR-DRI
and the 10 heats with 14% HBI-14% pig iron showed a decrease in the in the final FeO
content of the slag as compared to the heats with a 100% scrap charge. Due to the scatter
in the data and the described changes in furnace equipment and practice, the fact that
these two scrap mixes showed the same average final FeO content should be considered
coincidental. However, the decrease in final FeO content as compared with heats with
100% scrap can be explained. Both DRI and HBI will significantly densify batch charges
of loose scrap. A more dense scrap charge will be much less susceptible to excessive
oxidation by injected oxygen during melting. Also, these scrap mixes provided an excess
of more than 200 pounds of carbon per ton of scrap substitute material, and should have
significantly increased the melt-in carbon content of the steel bath.
Also of interest is the fact that the heats with 10% pig iron and 20% Circal HBI
exhibited a very similar final FeO content as the heats with a 100% scrap charge.
Although the HBI and pig iron should have helped to densify the scrap charge, this
combination should have contributed almost no excess carbon to the steel bath.
The heats with 20% Circal HBI and no pig iron showed an increase in the final FeO
content of the slag as compared with all other heats. The four percent increase in FeO for
these heat compared with the heats with 100% scrap shown in Table 8, actually under
represents the increase in the loss of iron units to the slag. For the heats with 20% HBI
and no pig iron, the flux additions were increased from 7000 lbs of CaO and 2800 lbs of
MgO to 9700 lbs of lime and 3000 Ibs of MgO. This modification in flux practice was
made in order to prevent excessive refractory wear by a highly fluid slag, and to ensure
sufficient lime was present to control phosphorus and sulfur levels in the steel. Ignoring
the effects of slag flushing, the increase in the wt% of FeO in the slag and the increase in

119
the total slag weight result in an increase in the actual pounds of iron oxide in the slag by
nearly forty percent! The assumption of no slag flushing is a poor representation of the
actual system. Therefore, the calculated 40% increase in iron units in the slag represents
an upper limit of the possible change. Again, the apparent lack of correlation between
the quality cast tons yield, which is listed in Table 8, and the iron losses to the slag
suggests that this calculated yield is a poor indicator of furnace performance.
Because the 20 trial heats containing HBI were conducted in a sequence, and very few
parameters were varied, an attempt was made to evaluate these data in a more
quantitative manner. In Figure 85, the final iron oxide content of the slag is plotted
versus the actual pounds of Circal HBI, which were charged in the furnace for each heat.

y = 0.0003~+ 26.503
R2 = 0.2612

30000 40000 50000 60000 70000

Ibs of Circal HBI in the Charge

Figure 85: Final wt% FeO in the slag versus pounds of Circal in the charge.

A distinct increase in the final FeO content of the slag as a function of HBI in the charge
can be detected. However, the large amount of scatter in the data suggest that other
factors are influencing the slag chemistry.
In most modern electric furnace operations oxy-fuel burners and direct oxygen
injection are used to supplement electrical energy input in order to increase melting rate
and furnace productivity. At NSST both oxy-fuel burners and direct oxygen injection are
used. Oxygen is injected during melting and refining via a door lance and a super sonic
oxygen injector. In Figure 86, the final iron oxide content of the slag is plotted versus the
amount of oxygen injected via the door lance and super sonic burner for each heat.

120
 uu

50 -
to
I

,c 40 -
!
B
30 -
y = 0 . 0 0 3 ~+ 10.501
-
Y

20 - R2 = 0.6177
10
e
z 10 -
0 I I I I I

7000 8000 9000 10000 11000 12000 13000

Lance 0 2 + Super Sonic 0 2 (SCF)

Figure 86: Final wt% FeO in the slag versus oxygen injected through the door lance and
super sonic injector.

In Figure 86, oxygen injected through the oxy-fuel burners was neglected. Although the
oxidizing combustion products of these burners would contribute to oxidation of the
scrap, the gas usage through these burners was fairly consistent from heat to heat.
Clearly, the correlation between injected oxygen and the FeO content of the slag is much
tighter than the relationship between HBI in the charge and the slag chemistry. A
comparison of Figures 85 and 86 suggests that the quantity of injected oxygen has a
much greater effect upon the FeO content of the slag, than the amount of oxygen
introduced by the FeO in the Circa1 material.
In addition to the amount of oxygen entering the furnace with the HBI and the injected
oxygen, the amount of carbon introduced into the furnace via the pig iron also varied for
the trial heats. In an attempt to account for this variation in carbon, the moles of carbon
contained in the charged pig iron were subtracted from the moles of oxygen atoms
contained in the HBI and the injected lance and super sonic oxygen. In Figure 87, final
iron oxide content of the slag is plotted versus the resultant ‘excess oxygen,’ which was
not balanced by carbon in the pig iron.

121
 50 - e,

1
y = 1 E - 0 4 ~+ 36.553
R 2 = 0.5501
10
2o

-20000 0 20000 40000 60000 80000 100000 120000

-
+ moles of Circal FeO moles
Moles of injected '0'
of Pig Iron Carbon
Figure 87: Final wt% FeO in the slag versus the moles of iron oxide in Circal plus the
moles of injected oxygen minus the moles of carbon contained in the pig iron for the trial
heats with Circal HBI.

Clearly, a great deal of scatter still remains in the data as plotted in Figure 87. For some
of the intermediate points in Figure 86, representing the heats with the 10% pig iron -
20% Circal charge, significant furnace delays occurred during the trial heats. For these
heats, it can be assumed that significant oxidation of the partially melted scrap pile
occurred during the delays. This explains some of the scatter in the data of Figures 85
through 87. Another significant source of scatter is related to slag flushing. The time and
amount of slag flushing was highly variable during the 20 trial heats with Circal HBI.
The data of Figures 69, 83, and 84 suggest that electric furnace slags typically melt in
with fairly high FeO contents. As will be argued later, the FeO content of the slag
decreases as the heat proceeds as a result of chemical reduction of the iron oxide and
dilution by continuously dissolving lime and magnesite. Flushing during early melting,
high FeO slag from the furnace early in the heat will remove proportionately more iron
units from the furnace as compared with flushing later in the heat. Therefore the timing
and amount of slag flushing can have significant effects on the FeO content of the slag at
tap. Another important goal of the trials at NSST was to evaluate any possible effects of
the different scrap substitute materials upon the behavior of phosphorus in the furnace.
Generally, all the mixes studied melted in at fairly low metal phosphorus levels
0.003wt%P to 0.014wt%P.). During the refining period, the metal phosphorus content
either increased or decreased by no more than 0.004 wt% for all heats and usually by less

122
than O.OOlwt% for most heats. It was also desired to determine, which equilibrium
phosphorus distribution correlation best matched the data from the NSST trials. In
Figures 88, 89, and 90 the observed phosphorus distribution at tap is plotted versus the
calculated equilibrium distribution based on the correlations of Ide et al. [29],
Turkdogan[20], and Zhang et a1.[28] respectively.

60
n

50
-m
fi
a 40
Y

II
a 30
4
0
p
U
20
L
2
8
10
0
0
0 10 20 30 40 50 60
Ide et al. Calculated Lp = (P)/(Pl

Figure 88: Observed phosphorus distribution at tap versus the equilibrium distribution as
calculated by the Ide et al. correlation.

a 50
n

u E)u
%
P4

Y
40
II
30

p
U
20
L
m 10

0 10 20 30 40 50 60
Turkdogan Calculated Lp = (P)/(P)
Figure 89: Observed phosphorus distribution at tap versus the equilibrium distribution as
calculated by the Turkdogan correlation.

123
 -
60
a
n

50
-a

5
A

40
II
= 30
d

g
U
20
L

m
10
0
0
0 10 20 30 40 50 60
Zhang et al. Calculated Lp = (P)/(Pl

Figure 90: Observed phosphorus distribution at tap versus the equilibrium distribution as
calculated by the Zhang et al. correlation.

As plotted in Figures 88 through 90, the experimental data show a great deal of scatter.
As explained in section 6.1, the equilibrium phosphorus distribution is extremely
dynamic during the course of a single heat. Therefore it is not expected that the furnace
should reach equilibrium at the time of tap. However, because many of the equilibrium
correlations in the literature can either over or under predict equilibrium by several orders
of magnitude for some temperatures and slag chemistries, it is a worthwhile exercise to
plot the data versus these correlations to confirm that they are close. A major limitation
of both the Ide et al. and the Turkdogan correlations is the fact that neither can account
for the effect of manganese oxide in the slag upon phosphorus equilibrium. All of the
slag samples for the trials at NSST contained from 6 to 9 percent manganese oxide.
These slags also contained up to 3 percent alumina. Both alumina and manganese will
have an effect on the free oxygen activity in the molten slag and will, therefore have
some effect on the equilibrium phosphorus distribution. The Zhang et al. correlation does
contain a term, which accounts for manganese oxide dependence of phosphorus
equilibrium. Although the data in Figure 90 does show a consistent deviation from
equilibrium, the data are somewhat more coherent and less scattered than in Figures 88
and 89. Despite this apparent improvement, this correlation significantly over predicts
equilibrium for some of the heats with DRI with higher basicities. These data are not
shown in Figure 90 as the calculated equilibrium is clearly in error.

124
 There is a consistent discrepancy between the data from the trials with HBI and those
with DRI. Nearly all the trials with HI31 showed an observed final phosphorus
distribution, which was higher than the equilibrium value as calculated by any of the
correlations. However, the observed final phosphorus distribution from the trials with
DRI consistently fell nearer or below the calculated equilibrium. An actual measured
phosphorus distribution, which is below equilibrium describes the conditions necessary
for phosphorus reversion from slag to metal. In agreement with this fact, 11 out of 20
heats with HBI showed phosphorus pick up in the metal from the first to final sample. Of
the remaining 9 heats, only one showed a decrease in metal phosphorus and the rest
showed no change between the two samples. The trials with AIR-DRI melted in at very
low metal phosphorus values and showed only minor fluctuations during the heats. Four
heats showed a minor 0.001 wt% decrease in metal phosphorus, two heats showed an
increase, and six showed no change. Overall, these observations are in best agreement
with Figure 88, suggesting that out of these choices, the Ide et al. correlation is the most
accurate.
An attempt was made to evaluate the conditions of phosphorus mass transfer for the
trials at NSST as was done for the trials at PST. In Figures 91 and 92, the slag chemistry
and temperature are plotted as a function of time for one of the trial heats.

3020
3000
2980
o”* 2960
2940
2 2920
Y

k 2900
+1 2880
2860
2840
2820
0 2 4 6 8 10
Time (from first sample)
Figure 9 1: Furnace temperature versus time during the refining period for one heat with
25% DRI - 75% scrap.

125
 45
40
35
30
25
5 20

0 2 4 6 8 10
Time (from first sample)

Figure 92: Slag chemistry versus time during the refining period for one heat with 25%
DRI - 75% scrap.

In Figure 93, the observed slag-metal phosphorus distribution and the equilibrium
phosphorus distribution according to the Ide et al. correlation are plotted versus time.

60.0 1

50.0
n
40.0
%
fi
30.0
II
a 20.0
4

10.0

0.0
0 2 4 6 8 10
Time (from first sample)

Figure 93: Calculated equilibrium and observed slag-metal phosphorus distribution

versus time

The general behavior of both the calculated equilibrium and observed phosphorus
distribution shown in Figure 93, is fairly representative of the trends observed for most of
the trial heats with AIR-DRI. The equilibrium phosphorus distribution initially increases
during the refining period in response to the increasing lime content of the slag, but then

126
decreases with the rapid increase in temperature toward the end of the heat. In Figure
94, the measured bulk and calculated interfacial metal phosphorus contents are plotted
versus time.

0.008

0.007 - /

-
---- ---
0

-
0.006

a
Y
0.005

0.004
47--

-
-- A
7

0.

w
> 0.003 -
0.002 -C- Measured bulk wI%[P)

0.001 Calculated Interfacial

0 ~

0 2 4 6 8 10
Time (from first sample)
Figure 94: Measured bulk metal phosphorus and calculated metal interface phosphorus
versus time

The behavior of the measured bulk metal phosphorus plotted in Figure 94 is very typical
of what was observed for all the trial heats with AIR-DRI. The first metal sample in
Figure 92 was taken 27 minutes after the dropping of the third bucket. By the time of the
first metal sample, the bulk metal had reached a very low metal phosphorus content for
all heats, which varied by only f 0.001 wt% during the refining period. Because the
measured metal phosphorus content, as well as the deviation from the calculated
interfacial phosphorus content were very low, the driving force for phosphorus transfer,
as calculated by equation 54, is also very small. Under these conditions, the calculation
of the mass transfer parameter becomes very susceptible to errors in the estimated slag
mass and calculated equilibrium, as well as errors in the chemical analysis. For the heats
where the experimental data allowed evaluation of equation , the term:

AP, k, 49
wn,
was found to be on the order of 0.05 - 0.15 min-'. This result is in fair agreement with the
complimentary results from the electric furnace at PST. It is not unreasonable that the
mass transfer parameter of the furnace at NSST is lower than that of the furnace at PST.

127
For various reasons, the standard furnace practice at PST results in a much higher melt-in
carbon content than is used at NSST. As a result, much less injected carbon is used by
PST to maintain a foamy slag, Carbon monoxide evolved at the slag-metal interface as a
result of metal decarborization will generate more slag-metal stirring than gas evolved
from solid injected carbon suspended in the slag layer. The higher rate of oxygen usage
at PST relative to NSST should also result in better mixing conditions in the furnace.
Because only two slag and metal samples and one temperature reading were obtained
for the trial heats with HBI, quantitative evaluation of the mass transfer parameter during
these heats is not permitted. However, qualitative evaluation of the slag and metal
chemistry data, indicate reasonable agreement with the calculations above for the heats
with DRI.

5.3 BHP - Sydney Steel Mill EAF Trials with HBI

Researchers at BHP [l],previously conducted a large number of electric furnace trials
with HBI. A great deal of the data from those trials were shared with the present author.
Similar to the trials conducted at PST and NSST for this study, the BHP researchers
examined the fluctuations of the slag and metal chemistries during the course of the trial
heats. In fact, the amount of slag sampling performed during the BHP trials was far more
extensive than that of either the PST or NSST trials. Samples were taken of the liquid
slag at various stages during the melting and refining sequence for those heats. Table 9
lists some of the operating parameter of the BHP - Sydney furnace in which these trials
were conducted.

128
Table 9: Operating Parameters of BHP EAF -
Power AC
Capacity 80 tons 2 bucket charge
or 1 bucket + continuously fed HBI
Diameter ??
Tap to Tap Approx. 70 minutes
Typical Oxygen Usage Total - 81 0 SCF/Ton
Injected Carbon* As required for slag foaming
Charge Carbon 0 - 2000 Ibs
Flux As required to maintain a V ratio = 2.5
Typical usage: Lime - 4400 - 6600 lbs
Dolomite - 3300 - 4400 Ibs
Slag Flushing ???
Hot Steel Heel ???
Slag Heel ???

In Figure 95, the slag chemistry of one heat from the BHP trials is plotted as a function of
time. In this figure, the first slag sample was taken immediately before the final scrap
charge was dropped.

50

45

40

35

30
x
E 25
20

15

10

Figure 95: Slag chemistry versus time for a single two bucket batch charged EAF heat.
The initial sample was taken between the first and final scrap charges to the furnace.

The data in the last 30 minutes of Figure 95 show very similar trends to the data from
PST and NSST. The slag sample at t = 0 was taken between melt down of the first
bucket and the addition of the second and final scrap charge. The changes in slag

129
chemistry from the first to the second sample in Figure 95 highlight the sharp increase in
slag FeO during melt down. Also, the decrease in slag basicity between the first and
second samples suggests that a portion of the lime has re-solidified due to the sharp
temperature decrease, which results from the cold scrap addition. As was seen in the data
from NSST, a continuous increase in the basicity of the liquid slag is observed, even
when no flux additions are made late in the heat. This is due to the continuous

Total Iron 92.6%

FeO 9%
Carbon 1.38%
Phosphorus 0.053%
SiO, 1.52%
A403 0.69%
MgO 0.21%

During the continuously fed heats, approximately 60 tons were continuously fed into
furnace at a target rate of 1.5 tons per minute. Lime and dolomite were fed concurrently
with the HBI to balance the silica content of the material and maintain a V ratio of 2.5.
The slag chemistry for one of the continuously fed heats is plotted versus time in
Figure96.

130
 40

35

30

X
25
-
-*-
Si02
CaC
MgC
* FeO
15

10
-t Start cts feed t
5 - 60 tons HBI End cts
feed HBI

Time
Figure 96: Slag chemistry versus time for a trial heat with 60 tons of continuously fed
HBI.

In Figure 97, the slag FeO increases by nearly 20 wt% as a result of the continuously fed
HBI. Based on the compositional data in Table 10, the continuously fed HBI material
contains slightly less carbon than is required to reduce the iron oxide in the material. The
imbalance results in an excess 7.2 kg of FeO per ton of HBI. For the 60 tons of charged
HBI this generates a total of 430 kg of FeO, which must be absorbed by the slag. Based
on the observed change in the FeO content of the slag, the actual contribution of FeO to
the slag by the HBI was probably more than 7.2 kg per ton. It is quite possible that the
chemistry reported in Table 10 did not account for degradation of the metallization of the
HBI due to aging or possible moister pickup. Either source of oxygen would further shift
the carbon - oxygen ratio away from the stoichiometric balance.
The continuously fed HBI heats are somewhat parallel, though on a much larger scale,
to the fast melting experiments of the present study, which were reported in Chapter 4.
Unfortunately, data regarding the metal chemistry were not available to the present
author. However, data regarding the phosphorus content of the slag for the heat shown in
Figure 96 are plotted in Figure 97 versus time.

131
 0.45 i

-
t t
A
0.25
Y
ge
>
* o'2 - Startctsfeed
End cts
feed HBI
60 tons HBI
0.15 -

O ' l l
0.05

0 10 20 30 40 50 60 70
Time (minutes)

Figure 97: Phosphorus content of the slag versus time for the trial heat with 60 tons of
continuously fed HBI shown in Figure 96.

In Figure 97, after an initial increase, the phosphorus content of the slag decreases during
the period of continuous feeding. When continuous feeding is terminated, the slag
phosphorus begins to increase again, despite a rapid increase in furnace temperature.
During the continuous feeding period, the slag volume increases significantly due to the
contribution of the HI31 gangue material and the balancing flux additions. Also, a large
amount of iron oxide reports to the slag as a result of the melting HBI. The BHP
researchers estimated that the slag weight would increase by 25% at the level of HBI
usage of this heat. If the phosphorus in the HBI material reported directly to the slag with
the gangue material, 3 1 kg of phosphorus would enter the slag. This much phosphorus
would more than double the amount of phosphorus in the slag. Therefore, even
accounting for the diluting effects of the increased slag weight, the slag phosphorus
content should increase significantly. The fact that the phosphorus content of the slag
decreases for a significant portion of the continuous feeding period supports the earlier
conclusion that the phosphorus in HBI does not remain in the gangue during melting.
The increase in the slag phosphorus content after continuous feeding is stopped suggests
that the slag metal phosphorus distribution is below the equilibrium value, and the
dephosphorization of the metal is occurring. For these reasons, the data in Figure 97 are
consistent with the theory that phosphorus reversion from HBI gangue to the surrounding

132
metal is rapid, and that the phosphorus in the steel is then transferred to or from the
external slag according to the equilibrium of the slag chemistry and furnace temperature.

133
 Chapter 6

Process Model
6.1 Slag and Temperature Model
In the previous chapter, experimentally determined temperature and slag chemistry data
were used to evaluate the mass transfer conditions of two electric arc furnaces. The end
goal of this project is to develop a process model, which can be used in a predictive
manner to evaluate various raw materials and operating conditions. An electric arc
furnace represents an extremely complex and dynamic reactor. During EAF steelmaking
almost all system parameters are varying simultaneously with varying degrees of
interdependency. Continuous oxygen and carbon injection, flux addition and dissolution,
slag foaming and flushing, as well as continuous variation of the temperature are just a
few of the complexities of the EAF environment. Continuous dissolution and/or melting
of scrap and slag materials, cause difficulties in determining even some of the most basic
system parameters such as the steel and slag weight. A process model should be useful in
designing flux practice such that phosphorus retention in the slag is optimized without
compromising yield considerations. The slag-metal system during electric steelmaking is
an extremely dynamic system, which is sensitive to many operating parameters. In order
to model specific chemical reactions such as decarborization or dephosphorization, the
dynamics of the slag chemistry and furnace temperature must be known with a fair
degree of accuracy. The slag chemistry can change dramatically during the course of a
single heat as the result of several factors:
Oxidation of metallic charge materials during melting

Reduction of FeO from the slag

Continuous addition of CaO and MgO as well as dissolution of undissolved slag

components as a result of temperature increase and changes in FeO

Changing slag mass due to the above factors, as well as due to slag flushing

A basic understanding of the general trends in temperature and slag chemistry during
EAF steelmalung and the fundamental phenomena, which govern these changes, must be
established before a predictive model can be developed for this system. In the previous

134
chapter a large amount of industrial data from three different electric arc furnaces was
introduced. Some common trends were identified in the behaviors of the slag chemistry
for all three furnaces. In this section, an attempt will be made to roughly model some of
the basic phenomena, which influence the behavior of the bulk slag chemistry during
electric furnace steelmaking.

6.1.1 Iron Oxide Formation and Reduction During EAF Steelmaking

Like most modern electric furnace shops, the three electric furnaces evaluated in
Chapter 5 operate with high melting efficiency and shorter tap to tap times as priorities.
As a result of aggressive melting practices using large amounts of injected oxygen, the
slag-make at melt-in generally contains large amounts of iron oxide. Assuming
relatively constant slag and metal heal weights, the total amount of iron units lost during
a given heat will be a function of the iron oxide content of the slag, as well as the period,
rate, and quantity of slag flushing obtained for a given slag practice. In Section 5.2,
correlations were identified between the final iron oxide content of the slag and the
amount of injected oxygen and the carbon - oxygen balance of the charge materials.
Although the melt-in chemistry of the slag could not be evaluated due to a lack of data, it
is most likely that the melt-in iron oxide content of the slag is most strongly affected by:
The amount of iron oxide in the slag heal
The amount of injected oxygen
The physical characteristics of the scrap charge, i.e. bulk density
The amount of unbalanced oxygen in direct reduced iron materials
Furnace delays occurring partially through the melting sequence
For a furnace employing a hot heal practice, rapid oxidation of the very low carbon liquid
steel heal will occur during the turn-around period while the furnace roof is off. A slag
heal of 5 tons could easily contain 3 tons of iron oxide. The amount of iron oxide formed
as a result of injected oxygen is related to both the quantity of oxygen used and the
characteristics of the scrap. A scrap charge with a low bulk density will have a large
amount of surface area available for oxidation. Ignoring other influences, the data from
NSST trials indicated an approximate linear relationship between the final FeO content of
the slag and the amount of injected oxygen according to the relationship:

135
 wt%(FeO) = 0.003 * (SCFof injected 0,)+ const. 56
A similar relationship is expected to exist for the melt-in FeO content of the slag. The
value of an expression like equation 56 and the analogue for melt-in FeO, can not be
understated. Such expressions are unique to the particular furnace and operating practice
for which they are developed. In order to optimize a furnace for all parameters, it is
necessary to know, at least empirically, the relationships between the various operating
variables and performance indicators. If the empirical relationships, which link the
different parameters, can be established for a given furnace, it is simple to develop a
matrix analysis technique to optimize the furnace practice with respect to the different
performance indicators.
Scrap substitute materials like the Circal HBI, which contain insufficient carbon to
reduce the iron oxide contained within them will also contribute iron oxide units to the
slag at melt-in. In the specific case of the Circal material, roughly 240 pounds of FeO per
ton of HBI or 120 kg of FeO per metric ton of HBI will be added to the slag when this
material is melted.
After melt-in, iron oxide in the slag will be reduced via:
Reaction with injected carbon
Reaction with the steel bath
Reaction with entrained liquid steel droplets in the slag
Additional oxygen injected during refining to promote slag foaming and bath
decarborization will form new iron oxide units. However, the summation of the various
oxidation and reduction reactions most often results in a net reduction of FeO from the
slag. For a given melt-in FeO content, the total amount of iron units recovered from the
slag will be dependant upon a large number of variables including:
The specific details of the slag flushing practice
The melt-in steel carbon content, which is a function of the scrap mix and amount
of charge carbon
The mass transfer conditions of the furnace
The amount and efficiency of injected carbon
Most modern electric furnace operations employ a foamy slag practice. After the
majority of the scrap pile is melted down and the arc is no longer surrounded by scrap,

136
the quantity of heat radiated to the furnace side-walls increases and heat transfer
efficiency to the steel decreases. By burying the arc in a foamed slag, radiation to the
furnace walls is reduced and heat transfer to the steel bath is greatly increased.
Furthermore, increased arc stability during foamy slag operation allows for higher
furnace power settings among many other benefits. For many furnace operations, foamy
slag practice is directly related to slag flushing. A slag with good foamability will
increase in volume by a factor of four. For furnaces with an insufficient amount of free
board, the rapid increase in slag volume associated with foaming will result in large
amounts of slag flushing. If excessive slag foaming and flushing occur early in the heat,
many otherwise recoverable iron oxide units may be unnecessarily lost from the furnace.
It is extremely difficult to separate the precise contributions of the various
mechanisms of iron oxide reduction from the slag for actual plant data. Slag flushing,
coupled with continuous lime and magnesia addition andor dissolution make it very
difficult to track either the precise slag volume or the exact causes of observed changes in
slag chemistry. However, comparison of the fundamental reactions to actual plant data,
can give a good indication of the limitations of the various reduction mechanisms. The
slag chemistry data from the NSST electric furnace trials, which were introduced in
Chapter 5 are reprinted below.

137
 -
I<

-
\
60 \

'.'. -
\
\
\
- --. - - FeO
\ Si02
50 -

40 -
'-.-- ---_-- --------a
P205

51
d

C 30 -
I

0 5 10 15 20 25 30 35
Time (from first sample)

Figure 84: Wt% of slag component X versus time for a 25% DRI - 75% scrap heat. The
first sample was taken 7 minutes after the dropping of the third charge.

As mentioned in Chapter 4, the trends observed in the slag chemistry from the NSST
trials are very similar to those in the final 30 minutes of the BHP data. For all the trial
heats at NSST, significant slag flushing commenced approximately 5 minutes after the
time of the first slag sample. In addition, the continuous dissolution of lime and
magnesite added during melt-in of the first two buckets, contributed to the dilution of the
iron oxide present at the time of the first sample. Therefore, uncertainty in the total slag
weight, as well as the precise amount of dissolved lime and MgO at any given instant,
make it difficult to determine a reliable rate of iron oxide reduction from these numbers.
Very careful inspection of the changes in the slag FeO with comparison to changes in the
basicity suggest that some reduction of FeO from the slag does occur during the refining
period.
The first sample obtained from the metal, correlating to the second slag sample in
Figure 84, contained 0.026 wt% carbon. Based upon the composition of the scrap charge,
which included 25% AIR-DRI, the theoretical melt-in carbon content of the steel is
around 0.5 wt% C. This assumes that approximately half of the carbon in the DRI
material is consumed by the iron oxide contained within the pellets and half reports to the
steel melt. This calculated melt-in carbon probably represents an upper limit, and the

138
actual value is likely to be lower. Under these conditions, the rate of reaction between
the slag and either entrained steel droplets or the bulk metal bath will most likely be
limited by mass transfer of carbon to the slag metal interface. The rate of decarborization
is therefore likely to be governed by the equation:
57
--
dt
Where [%CIi designates the carbon content at the slag metal interface. If it is assumed
that the chemical reaction rate is fast, the [%C], can be assumed to be in equilibrium with
the FeO content of the slag. Under the oxidizing conditions of this system, the
equilibrium and interfacial carbon content would be effectively zero. Therefore it is very
simple to integrate equation 57 giving the result:
58

This integration implicitly assumes that the mass transfer parameter is a constant. Both
laboratory and plant experiments have shown, that the mass transfer parameter is not
constant and may be a function of time andor concentration. Under some conditions,
these variations can be ignored, and a best fit, constant value can be assigned to the mass
transfer parameter. At the extreme limits of the reaction, when the bulk concentration is
very far from equilibrium and as the bulk concentration becomes very near equilibrium,
the assumption of a constant mass transfer parameter may result in significant error.
For the calculation of the reaction rate at intermediate times a constant mass transfer
parameter may be acceptable. In such a case the term:

49

should be the same order of magnitude as the value calculated in Chapter 5 for
phosphorus transfer. Therefore, it is a simple matter to calculate the amount of iron oxide
reduction, which will occur as a result of reaction with carbon in the steel bath. Figure 98
shows the calculated change in wt% FeO in the slag versus time according to equation
58.

139
 70

60 -

50 -
m
0
QI
LL
40 -

l o0 1
0 5 10 15 20 25 30
Time (minutes)
Figure 98: Iron oxide content of the slag versus time according to equation 58.

Despite the many assumptions made in the calculation of Figure 98, the agreement of the
calculation with the experimental data shown in Figure 84 is remarkably good.
At NSST, slag foaming is significantly enhanced by the direct injection of carbon into
the slag layer via a door lance. Carbon injected directly into the slag during the refining
period should have a significant effect on the FeO in the slag. The reduction of FeO in
slag by solid carbon will proceed via the gaseous intermediates CO/CO,. At high FeO
levels in the slag, this reaction is most likely limited by the combustion of CO at the gas
carbon interface. The reactivity of the specific type of carbon used and the surface area
available for reaction, will have a strong effect on the observed rate of reduction of FeO
from the slag. Typically, approximately 2000 pounds of carbon is injected through the
door lance per heat during the refining period. By contrast, the entire charge of 150 tons
of scrap steel and DRI only contained 1650 pounds of carbon. One ton of injected carbon
should be capable of reducing 6 tons of iron oxide from the slag. Assuming the total slag
weight for this operation ranges between 13 and 17 tons, a major reduction in the slag
FeO should be achieved as a result of injected carbon. It should be pointed out that
during refining oxygen is simultaneously injected with the carbon through the door lance.
However, the maximum amount of injected oxygen observed for any of the trial heats at
NSST would only combust approximately 15 to 20 percent of the injected carbon.
Therefore, the quantity of injected carbon, which is available to react with the slag is of
the same order of magnitude as the amount of carbon contained in the steel bath.

140
 In a recent study, Ken- and Fruehan [ S 5 ] investigated the rate of reaction of a slag
containing 10% FeO and solid graphite. These researchers found the rate of reaction for
their system to be on the order of 10.’ lb/cm2.sec. Using this fundamental reaction rate,
one can model the rate of reduction of FeO from the slag for different conditions. A
model was constructed, which allow for variation of
0 The grain size of the injected fines
0 Injection rate
Total quantity injected
Initial slag FeO content
Specific gravity of the injected carbon
Within this model, a system of ordinary differential equations is solved simultaneously
using a numerical solver. This method allows for the evaluation of a number of
interdependent parameters. The simulation shown below was generated assuming that
carbon injection into the slag starts at t = 0. This basically describes the scenario where
the carbon is turned on at the beginning of the refining period. Figure 99 displays the
surface area of carbon available for reaction as a function of time assuming the injection
of 2mm fines at a rate of 150 lbs/min.

15’107 I I \ I I I I I I

1.10~

5’ lo6

I I \ I I I I I I
0
0 2 4 6 8 10 12 14 16 18 20
l i m e (rnmutes)
Figure 99: Interfacial area between the injected carbon and slag as function of time
during refining.

After 20 minutes the reaction area approaches a steady state value as the rate of injection
begins to balance the rate of consumption, which is in turn, a function of reaction area.
This suggests that even the slag - carbon reaction area is a transient during the refine

141
period. Figure 100 displays the slag FeO as a function of time assuming an initial slag
composition with 50wt% FeO.

- --
60 I I I I I I I I I

.----I---.

*--
---*-
--
_I--

- *=.-* -
-- ---
40
L
*
h-l

---c
-m

20 - -

1 I I 1 I I I I 1
0

Figure 100: FeO in slag as function of time during refining.

The slag and metal weights used in the model were based on the NSST furnace. The
model predicts a lag in the reaction rate, which is related to the area dependence of this
reaction. A similar phenomenon was frequently observed during the trials at NSST.
Often, when flat bath was reached, and carbon injection was commenced, there was an
initial delay during which slag foaming was moderate at best. After several minutes of
carbon injection, the slag - carbon reaction rate would suddenly surge, resulting in
massive flushing of slag from the furnace. Even when the flow of carbon was stopped,
the surge in the slag-carbon reaction took a minute or two to subside. This simulation
most likely represents a fairly optimistic prediction of the actual conditions.
Despite this observation, the degree of iron oxide reduction predicted for carbon
injected was not observed for any heats. Furthermore, the predicted trend of an
accelerating rate of FeO reduction was also not observed. For these reasons, it is the
opinion of the current author that solid carbon injected through a door lance is actually a
very inefficient means of reducing carbon from the slag. In fact, careful observation of
the slag, which was flushed from the furnace slag door at both PST and NSST revealed
that the slag was ‘burning’ as it left the furnace. The most reasonable explanation for this
phenomenon is that the injected carbon is only partially consumed before it is swept back
out the slag door with the exiting slag. It is likely that the volume elements of slag,
which benefit most from the injected carbon and obtain the highest degrees of FeO

142
reduction end up in the slag pit under the furnace. The lack of correlation between the
actual slag chemistry data, and the calculated degree of reduction, which should be
effected by injected carbon, suggest that this mechanism of iron oxide reduction has only
a minor effect on the bulk slag chemistry.

6.1.2 Flux Additions and Slag Make During EAF Steelmaking

In addition to the formation and reduction of iron oxide during EAF steelmaking,
behavior of the slag chemistry is complicated by several factors. For many electric
furnace operations, flux additions of lime, dolomite, andor magnesite are made early
during the melting cycle to ensure the protection of furnace refractories from acid
components of the slag make. In most cases, a large portion of these flux additions
remains solid until well into the refining stage of the heat. Frequently, researchers
estimate the total slag weight in a given furnace by dividing the total mass of flux
additions by their mass fraction as determined chemical analysis. Because most slag
sampling techniques will only give the chemical composition of the liquid slag phase, it
becomes very difficult to estimate the total slag weight when a significant portion of the
flux additions are likely in the solid state. At higher temperatures, late in the heat, nearly
all flux additions will have dissolved for most flux practices. However, for many electric
furnaces, significant slag flushing or slopping commences as soon as flat bath is achieved
and slag foaming begins. All of these factors, contribute to make the simplest slag mass
calculation, a complex guessing game.
For the above reasons, a series of crude models were constructed in attempt to mimic
the observed trends in the experimental data presented in Chapter 5.
Slag Model 0
In the first model shown below, only the influence of FeO was considered:

0 The total amounts of CaO, MgO, and SiO, were held constant
0 The mass of FeO in the slag was increased at t = 15, then reduced assuming
reduction by carbon in the steel bath only
0 The total slag mass was unrestrained and allowed to vary with the changes in FeO

143
 60

50

40
X

5 30 %MgO
s %CaO

o ! I I I I I

0 10 20 30 40 50 60
Time

Figure 101: Slag model 0 - constant CaO, MgO, and SiO,, with changing FeO and no
slag flushing

The important attributes of Model 0 include:

The change in the FeO is equal to the sum of the change of all other components
The basicity and CaOMgO and MgO/SiO, ratios remain constant

Slag Model 1
In the next case, the increase in FeO during melting was ignored, and the effect of
increasing CaO in the final 30 minutes of the heat was studied. The thermodynamic
temperature dependence of the solubilities of CaO and MgO, the kinetics of the
dissolution phenomena, and the specific heating rate will all combine to determine the
actual rates of dissolution of these components into the bulk slag. For the purposes of
this exercise, detailed scientific predictions of the dissolution rates of CaO and MgO were
not attempted. In this model:
The masses of FeO and SiO, were held constant
The mass of CaO increased proportionally to the square root of time, and the mass of
MgO increased linearly with time.
The slag weight was unrestrained and allowed to increase with the changes in lime
and magnesia

144
 -
%Si02
%MgO
-_- %CaO
%FeO

10

I I I I

0 10 20 30 40 50
Time

Figure 102: Slag Model 1 - Constant FeO and SO,, with increasing CaO and MgO and
no slag flushing

The important attributes of Figure 102 are:

The basicity is increasing

0 The ratio of FeO/SiO,remains constant.
Significant reductions in the observed wt % FeO can be effected by simply increasing
the total lime

Slag Model 2
Slag Model 2 is the same as Slag Model 1 shown above, however, the total slag weight
was restrained to simulate slag flushing. A weighted increment of each slag component
was removed from the total slag weight at each time step in order to accommodate the
additional CaO and MgO.

145
 60

50

40
%Si02
K
%MgO
5 30
--- %CaO
s
%FeO
20

10

0 1 I I I I

0 10 20 30 40 50
Time

Figure 103: Slag Model 2- Constant FeO and SiO,, with increasing CaO and MgO with
slag flushing

The important attributes of Slag Models 1 and 2 are the same. Slag flushing will result in
only a subtle shift in the final values of the different slag components.

Slag Model 3
In Slag Model 3, the variables modeled in Models 0-2 were combined. In this case:

The weight of FeO in the slag was initially increased to represent scrap oxidation
during melting, then allowed decay assuming reduction via carbon in the steel bath
The weight of SiO, was increased at t = 15 then held constant.
The mass of CaO increased proportionally with the square root of time, and the mass
of MgO increased linearly with time.
The total slag mass was restrained as in Simulation I11 to simulate slag flushing

146
 60.000

50.000

40.000

30.000

20.000

10.000

0.000
0 10 20 30 40 50 60
Time

Figure 104: Slag Model 3 - FeO increase and reduction, SiO, increase, CaO and MgO
increase, and slag flushing

In Slag Model 3, the mass of the dissolved lime in the slag roughly doubled and the mass
of FeO decreased by roughly 40% in the last 45 minutes of the simulation. Although the
absolute values of the slag components do not exactly match, the general trends observed
in the data from the BHP furnace trials are fairly well reproduced in Figure 104.

Slag Model 3.1 - Continuous HBI Feeding

The effects of continuous feeding of HBI on the dynamics of the slag chemistry are also
of great interest. The actual slag chemistry data from the BHP electric furnace trial with
continuously fed HBI are reprinted below:

147
 25 -
X
520 - --A- MgO
3
15 -

t
A

10 -
\
+-L--*’
Start cts feed
5 - 60 tons HBI
t
End cts
feed HBI
04 I
0 10 20 30 40 50 60 70
Time

Figure 96: Slag chemistry versus time for a trial heat with 60 tons of continuously fed
HBI.

The above data were simulated according to the following assumptions:

0 After an initial transient period, the slag silica and FeO increase in proportion to the
rate of HBI injection and the relative weight percents of those components in the HBI
The mass of silica in the system remained constant after continuous feeding stopped
FeO was reduced from the slag by carbon in the metal assuming a carbon content of
0.3 wt%C in the steel at the end of continuous feeding.
Continuous lime and magnesia injection continued at a constant rate until the last 10
minutes of the heat
0 The total slag weight was restrained to simulate slag flushing

148
 HBI cts feed from t=O to t = 35

%Si02
x 30
Be %MgO
+ --- %CaO
20
%FeO

10

0
0 10 20 30 40 50 60 70 80
Time

Figure 105: Slag Model 3.1- Simulation of HBI continuous feeding.

The results of continuous feeding slag model are not bad. Unfortunately, because no
metal samples were available to compare with the changes in the slag chemistry, the
carbon content of the steel was estimated based upon the rate and extent of iron oxide
reduction.
The above exercise, though rather crude, made several important points clear:

The initial rise in the slag wt% FeO, observed during melting is due primarily
to an actual increase in the mass of iron oxide in the slag

0 The decrease in the slag % FeO is due to both a decrease in the mass of iron
oxide in the slag and an increase in the mass of dissolved lime and magnesia.

0 It is very difficult to tell whether dilution, or actual reduction is dominating

the slag chemistry.

0 Tracking the ratios of the various slag components offers the most reliable
way of deciphering what is happening.

The primary mechanism of FeO reduction from the slag is believed to be via
reaction with dissolved carbon in the steel bath. The observed rates of steel
decarborization and slag reduction roughly agree with evaluation of the mass
transfer conditions, which was presented in Chapter 5.

149
6.1.3 Furnace Temperature During EAF Steelmaking

Large thermal gradients are almost always present in the slag-metal system during
EAF steelmaking. Therefore, it should be recognized that both experimental temperature
measurements andor thermal models, which assign a single temperature to the entire
furnace are clearly not representative of the entire system. No attempt is made in the
present study to model the thermal gradients between the slag and metal during
steelmaking. However, separate measurements of the steel and slag temperatures
indicate that the slag temperature can be more then 100°C higher than that of the metal
during the refining period. However, because the steel temperature is the more important
and more commonly monitored parameter during steelmaking, all “furnace” temperatures
discussed in this paper imply the temperature of the metal bath. The general temperature
profile during an EAF heat is shown schematically in Figure 106:

I I
I I I
1600 I I I

y 1400
I
I
I
8* 1200 I
I
I
Y
1000 I
iR I
L 800 I
u I
a I
E 600 I
I
8
I- 400 I I
I I I

200 1st I 2nd 3rd I

1 buckel I bucket I buckel

I
I Refining
0
0 20 40 60 80
Heat Time (minutes)

Figure 106: Steel temperature profile during a hypothetical EAF heat with a three bucket
charge and a 75 minute tap to tap time.

During the melting stages of the cycle shown in Figure 106, heat transfer to the scrap pile
from both oxy-fuel burners and the arc is very efficient, and the temperature of the scrap
increases rapidly. As the heat approaches flat bath, heat transfer to the metal from the arc
decreases. Also, the radiative heat transfer efficiency from the oxy-fuel burners to the
scrap decreases rapidly as the furnace temperature increases. During the final melting
stages of the heat, the temperature remains relatively constant until most of the steel is

150
molten. Toward the end of the refining period, nearly all the flux additions and all of the
metallic charge are liquid and the temperature begins to increase rapidly once again. The
last 30 minutes of Figure 106 are re-plotted in greater detail in Figure 107.

1625

yh I600 -
Q)
L
3
5L 1575 -
a
Q)

E
I-
Q) 1550 -

40 50 60 70 80
Refining Time (minutes)

Figure 107: Steel temperature versus time during the refining period of the heat shown in
Figure 106

The specific shape of the curve in Figure 107 will be influenced by several factors
including the physical and chemical characteristics of the scrap charge and the power
rating of the furnace. For a very high powered furnace such as the one at PST, the
temperature increase at the end of the refining period will be fairly sharp as was seen in
Figure 74 in Chapter 5. By contrast the furnace at NSST is a lower powered furnace and
the temperature increase during the refining period was measured to be more linear. The
measured steel temperature of one of the trial heats at NSST is plotted versus time in
Figure 108.

151
 1700 I

0 5 10 15 20 25 30 35
Time (from first sample)

Figure 108: Steel bath temperature versus time for a 25% DRI - 75% scrap heat. The
first sample was taken 7 minutes after the dropping of the third charge.

During most DRUHBI continuous feeding practices, the feed rates of the iron bearing
material and the required continuous flux additions are regulated to maintain a steady
state condition of continuous melting at a constant temperature at a high power setting.
The measured furnace temperature profile for the continuously fed heat shown in Figure
96 is plotted versus time in Figure 109 below.
1 660

1640

1620
u
1600
L
3
1580
n
QI

E 1560
c
1540

1520

1500 4 I

0 10 20 30 40 50 60
Time

Figure 109: Temperature profile during continuous feeding of HBI into the BHP -
Sydney electric furnace.

152
A schematic representation of the temperature profile.for the duration of the entire heat of
a continuously fed practice is shown in Figure 110.

1800 7
-
I I
1600 I I
/

cs
I I
I I
0 I I

*
L
I
I
I
I
a I
I
I
I
Y
I I
*L

F* Begin
Continous
End
Continous
c
Feeding Feeding
I I I I

0 10 20 30 40 50 60
Heat Time (minutes)
Figure 110: Steel temperature profile during a hypothetical continuously fed EAF heat.

Continuous feeding of a portion of the furnace charge in the form of DRI or HBI has
been shown to offer many operating advantages over a completely batch fed practice. [5]
One major effect of continuous feeding is to reduce the tap to tap cycle by extending the
time during which the furnace is operated at full power. In such a practice, the melting
and refining periods occur simultaneously. This generally extends the period of time at
which the furnace is continuously at high temperature and slag metal reactions are in
progress.

6.2 Numerical Process Model for Phosphorus Transfer During EAF Steelmaking
In order to evaluate the relative influence of different system parameters on the
behavior of phosphorus in the electric furnace, a process model was developed. The
mathematics of the process model are fundamentally very simple. Within the model, the
two ordinary differential equations for phosphorus flux in the metal and slag given by:

a [%PI -
--_
at
ApmkrrLF P I - b p i
Wm
I 28

and.

153
 29

are solved simultaneously. If it is assumed that the interface is always at equilibrium, the
interfacial phosphorus contents in the slag and metal are linked through the equilibrium
phosphorus distribution. The equilibrium phosphorus distribution can be calculated as a
function of temperature and slag chemistry according to any of the correlations listed in
Appendix B. In the present model, the correlation of Ide et al. was used. As described in
previous sections, both the slag chemistry and the temperature vary with time during a
given heat. In Section 6.1 simple slag and temperature models were described, which can
be used to express those system variables as algebraic functions of time. These algebraic
expressions can then be combined with the Ide et al. correlation to calculate the slag-
metal phosphorus distribution as a function of time. The simple system of the two
differential equations expressed above, and the complex algebraic expression for
phosphorus equilibrium can be solved using most commercially available numerical math
software programs. For the present study, the symbolic mathematics software Mathcad'
was used. The use of a symbolic math solver allowed for various iterations of the model
to be developed and evaluated quickly. A representative worksheet from the model is
shown in Appendix E. Because the equations are solved numerically, for specified
increments of time, it is possible to vary the driving force for phosphorus transfer as a
function of time, and solve for the slag and metal phosphorus concentrations in a very
dynamic system.

6.2.1 NSST Electric Furnace Model

In order to test the numerical process model, an attempt was made to simulate the
conditions of one of the trial heats from the DRI trials at NSST. Based upon the
measured melt-in slag chemistry of the trial heat shown in Figure 111 and the actual
scrap charge and flux additions, the slag model generated the following slag profile for
the heat.

154
 60 I
\

.. .\--
\
50 \ \

---_---- -----
\

40 z

X
8 30
5
20

10

0 I I I I I I

0 5 10 15 20 25 30 35
Time (minutes)

Figure 111: Predicted slag chemistry profile during the refining period of an electric
furnace heat at NSST.

Based upon the temperature profile data for the NSST furnace, which was presented in
Section 6.1.3, a simple linear temperature increase during the refining period was
approximated. The equilibrium phosphorus distribution predicted by the Ide et al.
correlation for the above temperature and slag chemistry profiles is plotted in Figure 112.

155
 90
80
70

5 60
5
I-
L
50

8-
40

Ly
30
20
10
0
0 5 10 15 20 25 30 35
Time

Figure 112: Slag-Metal equilibrium phosphorus distribution predicted by the Ide et al.
correlation for the slag chemistry and temperature data of Figures 111 and 108.

The flux equations for phosphorus transfer in the slag and metal were solved assuming
the dynamic equilibrium phosphorus distribution shown in Figure 112. For this particular
simulation, the effect of slag flushing was ignored even though flushing did occur during
the actual heat. Also, the kinetic term:

AP,k, 49
Yn
was assumed to be 0.13 min". The calculated metal and slag phosphorus contents are
plotted below versus time in Figures 113 and 114.

156
 0.016 1

0.002 -
0 I i i i i i

0 5 10 15 20 25 30 35
Time (minutes)

Figure 113: Metal phosphorus content predicted by the numerical process model for an
EAF heat at NSST. The two discrete points are actual furnace chemistries.

0
0.4 - 0

o ! I 1 1 1 1 1

0 5 10 15 20 25 30 35
Time (minutes)
Figure 114: Slag phosphorus content predicted by the numerical process model for an
EAF heat at NSST. The three discrete points are actual measured furnace slag
chemistries.

157
The agreement between the predicted and calculated chemistries in Figures 113 and 114
is reasonably good. As described in Section 5.2, all of the early metal samples from the
DRI trial heats at NSST showed fairly low metal phosphorus contents with only minor
subsequent fluctuations. The simulation in Figure 113 suggests that a significant portion
of the phosphorus mass transfer occurs early during the refining period. It is also
interesting that with respect to phosphorus equilibrium, the positive influence of the slag
chemistry changes, e.g. increasing CaO and decreasing FeO, more than make up for
negative influence of the increasing temperature. As a result, the phosphorus distribution
ratio remained high during the heat, and no period of reversion was observed.

6.2.2 BHP Electric Furnace Model - Continuous Feeding of HBI

The plant data from the BHP electric furnace trials with continuously fed HBI were also
used to test the process model. A continuously fed electric furnace represents one of the
most complex case studies, which can be evaluated by a model such as the present
process model. During the actual and the simulated furnace heat, 60 tons of HBI were
continuously fed into a 30 ton hot heal at an average rate of 1.7 tons per minute. Flux
additions of lime and dolomite were fed concurrently with the metallic charge in order to
balance the silica content of the HBI and maintain a slag V ratio of 2.5. The charging
rate was selected such that the furnace power input matched the heat requirements for the
heating and melting all of the charge materials. Therefore, during the period of
continuous charging the furnace temperature nearly constant. When continuous feeding
was terminated, the temperature increased exponentially. In order to model this system
the following assumptions were made:
Slag:
The initial slag weight was based upon an assumed 75 kilograms of slag per
ton of steel on the heel.
The initial slag chemistry was approximated from the experimental data in
Figure 95.
The mass of silica in the slag increased proportionally with the rate of feeding
of HBI and the amount of silica in the HBI material according to Table 8.

158
 0 The mass of lime in the slag increased at a rate of 2.5 times the increase in
slag silica.
The mass of MgO of the slag increased linearly with time.
0 Due to a lack of data regarding the chemistry of the steel heel, the change in
the weight of iron oxide in the slag was approximated in a best fit manner.
For the data in Figure 95, it was found that a harmonic equation could be used
to roughly represent the profile of the FeO.
0 The late in the heat, the above assumptions resulted in approximately 200 kg
of slag per ton of metal. For a significant portion of the heat, however, the
furnace was only partially filled. It was assumed that slag flushing would
only become significant late in the heat when nearly all of the metallic charge
had been fed into the furnace.
The above assumptions resulted in the following calculated slag profile for the
continuously fed heat:

60

50

40 - - -- - - - - - - -
X
8
s
30

20

I I
10

o f 1 I 1 I I I I
0 10 20 30 40 50 60 70
Time (minutes)
Figure 115: Modeled slag chemistry profile for the BHP electric furnace trial with
continuously fed HBI. The actual slag chemistry profile is shown in Figure 97.

Figure 115 represents the major features of the actual measured slag chemistry profile
reasonably well.

159
Temperature:
The temperature was modeled after the general behavior observed in the in the
actual data from the heat shown in Figure 109. The temperature basically
followed a slowly developing exponential equation. The temperature profile
used in this simulation is shown in Figure 116.

1610

1600

1590
v
f 1580
3
w
1570
a
Q1

E 1560

1550

1540

1530
0 10 20 30 40 50 60 70
Time (minutes)

Figure 116: Simulated temperature profile for the BHP electric furnace trial with
continuously fed HBI. The actual temperature profile is shown in Figure 109.

Based upon the slag chemistry and temperature profiles in Figures 115 and 116, the
equilibrium phosphorus distribution was calculated according to the Ide et al. correlation.

160
 vu

50 -
P
40 -
E3
'E 30 -
P,
s 20 -
I
0-

LLI

10 -

0 10 20 30 40 50 60 70
Time

Figure 117: Slag-Metal equilibrium phosphorus distribution predicted by the Ide et al.
correlation for the slag chemistry and temperature data of Figures 115 and 116.

Due to the continuous addition of phosphorus containing HBI, several additional

assumptions had to be made. Based upon the laboratory results of the present research, it
is believed that phosphorus reversion from HI31 gangue to the surrounding bulk metal
occurs on the same time scale as melting. Therefore, within the numerical model, the
flux equation for phosphorus in the metal was modified to account for the phosphorus
entering with the HBI. The flux equation for the slag phase remained the same. The
changes in the liquid steel and slag weights were also incorporated into the model. The
modified phosphorus flux equations were then solved. The calculated metal and slag
phosphorus contents are plotted versus time in Figures 118 and 119 respectively.

161
 0.035

0.03

0.025
n
0,02
8
5 0.015

0.01

0.005

0
0 20 40 60 80
Time (minutes)

Figure 118: Metal phosphorus content predicted by the numerical process model for an
EAF heat, which was continuously fed with 60 tons of HBI.

0.45
0.4

= 0.35
0.3
0.25
5 0.2
0.15 -
0.1 -
0.05 -
0 I I I

0 20 40 60 80
Time (minutes)
Figure 119: Slag phosphorus content predicted by the numerical process model for an
EAF heat, which was continuously fed with 60 tons of HBI. The discrete points are
actual measured furnace slag chemistries.

The agreement between the calculated and experimentally measured slag phosphorus
content as shown in Figure 119 is excellent. Figure 119 suggests that extremely complex
slag-metal systems can be described very accurately using the present simple model if the
time dependant slag chemistry and temperature behaviors are known and can be

162
mathematically represented. With the appropriate data, the behavior of the slag
chemistry during a given heat could be calculated from first principals accurately enough
to suit the present model.
The calculations represented in Figures 118 and 119 were made assuming that all of
the phosphorus in the continuously charged HBI initially enters the liquid metal melt.
Phosphorus is then transferred between the slag and metal in response to the equilibrium
conditions at each time step. As was discussed previously, many researchers believe that
when HBI is melted, the phosphorus contained in the material will initially enter the slag
phase and then transfer to the metal in response to equilibrium. The basic model used in
the simulation shown in Figures 118 and 119 was modified such that all of the
phosphorus contained in the HBI initially entered the slag phase rather than the metal.
The calculated slag phosphorus content for the two cases, e.g. phosphorus initially
entering the metal phase and phosphorus initially entering the slag phase, are plotted
along with the experimental data in Figure 120.

0.5 'I

Phosphorus initially enters slag

fl
0 0 z

t
Phosphorus initially enters metal

0.1 -
0.05 -
0 I

0 20 40 60 a0
Time (minutes)
Figure 120: Slag phosphorus content predicted by the numerical process model for an
EAF heat, which was continuously fed with 60 tons of HBI. The upper continuous curve
was calculated assuming all phosphorus from the HBI initially reports to the metal phase.
The lower continuous curve was calculated assuming all phosphorus from the HBI
initially entered the slag phase. The discrete points are actual measured furnace slag
chemistries.

163
Figure 120 suggests that there is almost no difference between whether the phosphorus
contained in the HBI initially reports to the slag or metal. The HBI feed rate of 1.7 tons
per minute delivers, roughly 0.92 kg of phosphorus per minute to either the slag or metal.
Alternatively, during the first 10 minutes of the simulation, phosphorus is being
transferred from the metal to the slag at a rate of approximately 0.4 to 0.5 kg per minute.
The net build-up of .42 kg per minute represents only a 0.01 wt%(P) build-up in the slag.
Under conditions where liquid phase mass transfer is extremely slow, andor the feeding
rate is very high, a significant build up in either phase might occur. However, continuous
feeding of HBI results in rapid evolution of carbon monoxide at or just below the slag
metal interface. As a result, the stirring conditions and the mass transfer parameter of a
given furnace will be strongly related to the rate of continuous HBI charging. Therefore,
it is likely that the conditions of low slag-metal stirring and high HBI feed rate will be
mutually exclusive.

6.3 Value of a Predictive Process Model

A numerical process model for phosphorus transfer in the electric furnace was
developed and tested in the previous sections. The general trends in slag chemistry and
bath temperature, which were observed for three different electric furnace practices were
also discussed in previous sections. A numerical process model as was described in
Section 6.2 should be extremely useful in evaluating the influence of various furnace
operating parameters upon the behavior of phosphorus during a given heat. Model
simulations can be conducted as an alternative to expensive furnace trials in order to
better optimize furnace practice for phosphorus control. Several comparative heat
simulations are presented below, primarily as examples of how the process model can be
used.
6.3.1 Influence of Mass Transfer Kinetics Upon Metal Phosphorus at Tap
Mass transfer kinetics will vary from one electric furnace to another. The present
author speculates that the mass transfer parameter of the electric furnace can vary by at
least an order of magnitude. The stirring conditions of the electric furnace will be most
strongly influence by such factors as the use of bottom stirring, the melt-in metal carbon
and slag FeO contents, continuous DRI/HBI feeding, etc. Under typical EAF

164
steelmaking conditions, some degree of dephosphorization will proceed during the
refining stage of the heat. As tap to tap times become shorter, it may be necessary for
some furnace operators to improve the mass transfer conditions of their furnaces in order
to achieve acceptable phosphorus levels in the steel.
The process model was used to simulate three heats with different mass transfer
parameters. Temperature and slag chemistry profiles were generated, which reflect the
observations of actual plant data as discussed in previous sections. The temperature and
slag chemistry profiles for the three simulated heats are shown in Figures 121 and 122.

1520 ! 1 I I I 1 I

0 5 10 15 20 25 30 35
Time (minutes)
Figure 121: Metal temperature profile for the simulated EAF heats.

165
 60

50

40
X
8
5
30
-------__-______
20

10

0
0 5 10 15 20 25 30 35

Time (minutes)

Figure 122: Slag chemistry profile for simulated EAF heats.

The equilibrium phosphorus distribution ratio was calculated for the above temperature
and chemistry profiles, and is plotted versus time in Figure 123.

:
120

100
P
4 80

20

0 1 1 I I I I

Time

Figure 123: Equilibrium Phosphorus distribution ratio for simulated EAF heats based
upon the temperature and slag chemistry profiles of Figures 121 and 122.

The equilibrium phosphorus distribution shown in Figure 123, is particularly useful in

comparing mass transfer conditions because it peaks at some time in the middle of the
heat. For very fast mass transfer conditions, the slag-metal system will closely follow the
equilibrium profile. For slower mass transfer conditions, the specific shape of the

166
dynamic equilibrium profile will be less discernable. The numerical process model was
used to calculate the metal phosphorus content versus time for three mass transfer
conditions, which represent the limits of what was observed in the present study.

0.025

0.02

0.015

0.01
I

0.005

1 Akp/W = 0.5 rnin-1

0 5 10 15 20 25 30 35
Time (minutes)
Figure 124: Simulated EAF metal phosphorus content versus time for three different
mass transfer conditions.

The results of Figure 124 suggest that the kinetics of phosphorus mass transfer in the
electric furnace has a strong effect on the metal phosphorus content of the steel during the
course of a heat. For a tap to tap cycle with a 30 minute flat bath period, the three mass
transfer conditions modeled in Figure 124 result in very similar final metal phosphorus
contents in the steel. A 30 minute flat bath period is typical for a furnace with a 80 to
100 minute tap to tap time. If the refining period is reduced in length by 50 percent, the
final metal phosphorus content will vary significantly for the three mass transfer
conditions. For some electric furnace operations where the mass transfer conditions of
the furnace can not be improved, the push toward shorter tap to tap cycles may be
accompanied by restrictions in the chemical composition of the charge materials.

167
6.3.2 Effect of Melt-in FeO Upon Metal Phosphorus at Tap
Oxygen is used in electric furnaces today in ever increasing amounts in order to
reduce electrical usage and accelerate melting. However, the amount of oxygen will have
a direct effect upon the melt-in chemistry of the slag. In addition to compromises
regarding metallic yield, increased oxygen usage may have an effect upon phosphorus
control in the metal.
The behavior of FeO in electric furnace steelmalung slag was discussed at length in
Section 6.1.1. It has been concluded by the present author that, for the furnaces
examined in this study, the melt-in FeO content of the slag is most strongly influenced by
the amount of injected oxygen used during melting. This assertion assumes DRI/HBI
usage is less than 30% of the metallic charge, and that none of the charge materials
exhibit exceedingly low metallizations. Subsequent to melt-down, iron oxide is reduced
from the slag primarily via reaction with carbon in the steel bath. The degree of FeO
recovery obtained for a given heat depends upon the slag flushing practice, the initial
FeO content of the slag, the total slag weight, the melt-in carbon content of the steel, and
the lunetics of mass transfer in the furnace.
Three electric furnace heats where simulated were the initial melt-in FeO content of
the slag was varied from 30 wt%, 50.wt %, to 65 wt%. The final FeO content of the slag
was held constant at 20wt% FeO, which implies that the heats also melted in with
different metal carbon contents. The behavior of the other slag components followed the
typical behavior observed in the plant trials of this study, e.g. increasing basicity due to
continuous dissolution of lime and relatively constant magnesia content. The furnace
temperature was increased from 1540°C near the beginning of the refining period to
1640°C at tap according to a slowly developing exponential function. The basic slag
chemistry behavior for the heat containing 30 wt% FeO at melt-in is shown in Figure
125:

168
 0 I I I I I

0 5 10 15 20 25 30 35
Time (minutes)

Figure 125: Slag chemistry profile for simulated heat with an initial FeO content of
30wt%

The other two heats with higher initial iron oxide contents showed the same basic slag
chemistry profiles, e.g. the other slag components maintained the same ratios to one
another. However, the slag FeO profiles for these heats differed as shown in Figure 126.

70

60

50
s9 40
v

8 30
5 20

10

0
0 5 10 15 20 25 30 35
Time (minutes)

Figure 126: FeO profiles for three simulated heats.

169
Based upon the described temperature and slag chemistry profiles, the equilibrium
phosphorus distribution was calculated as a function of time for the three heats.

120

100 ---------
P
80
E
3

-
P
'E 60
---
Ly
s
I-

40
30wt% FeO a t t = 0
50wt% FeO at t = 0
65wt% FeO a t t = 0
20

0
0 5 10 15 20 25 30 35
Time

Figure 127: Calculated equilibrium phosphorus distribution for three heats with varying
initial FeO contents in the slag.

Figure 127 shows that as the melt-in FeO content of the slag increases, the equilibrium
phosphorus distribution early in the heat also decreases. The decrease in Lp toward the
end of the heats is due to the rapid temperature increase just prior to tapping.
The numerical process model was used to calculate the phosphorus content of the steel
as a function of time during the heat. The results for all three heats are shown in Figure
128.

170
 0.025

0.02 50wt% FeO at t = 0

0.015
Y

Be
cr
s 0.01

0 005

0 I I I I I 1

0 5 10 15 20 25 30 35
Time (minutes)
Figure 128: Predicted metal phosphorus content during the three simulated EAF heats.

Figure 128 indicates that despite differences in the phosphorus equilibrium early in the
heat, the phosphorus content of the metal is not compromised if the iron oxide content of
the slag is reduced to a reasonable level by the time of tap. Obviously, if the iron oxide
content of the slag can not be reduced in the time allowed during the refining period, the
equilibrium phosphorus distribution and metal phosphorus content at tap will be
negatively effected. Because the kinetics of decarborization and dephosphorization are
both liquid phase mass transfer limited for the conditions of this simulation, changes in
the mass transfer conditions of the furnace should not significantly change the result of
this simulation. If the slag is reduced primarily via direct carbon injection into the slag,
the results of the present simulation may become invalid, though those conditions could
just as easily be modeled.

6.3.3 V Ratio and the Phosphorus Mass Transfer

The relationship between slag V ratio and phosphorus control is common knowledge
among those who make quality steels. Metal phosphorus decreases as a strong function
of slag V ratio. Therefore, when making phosphorus sensitive grades from high
phosphorus raw materials such as pig iron or DRUHBI, a logical process modification is
to increase the flux rate to the furnace in order to increase the V ratio of the slag.

171
However, it is most likely that there is a limit, beyond which further increases in the V
ratio may negatively impact the final phosphorus content of the steel.
For a significant portion of the electric furnace steelmaking cycle, the furnace
temperature is below 1600 "C. As was discussed in previous sections, investigation of
actual furnace slag chemistry data shows that solid lime charged early in the melting
cycle is continuously dissolving into the liquid slag late into the refining period. This is
due in part to the kinetics of lime dissolution into the slag, but is primarily due to the
temperature dependent solubility limit of lime in the CaO-SiO,-FeO-MgO slag system.
The maximum amount of lime present in the liquid portion of the slag will always be
limited by this solubility limit. Lime additions in excess of the CaO solubility limit will
result in either solid flux agglomerations frozen to the furnace side-wall or in second
phase particles suspended in the slag. Both phenomena are seen in actual steelmalung
practices, and both will have a direct effect upon mass transfer in the furnace. Large solid
agglomerations in the slag or slag crusting, will significantly reduce the amount of
stirring in the slag phase. Second phase particles have a strong effect on the bulk
viscosity of slags. The dependence of the bulk viscosity of steelmaking slags upon
second phase particles is fairly well know due to its relevance to slag foaming. [48] For
the temperatures and V-ratios sometimes observed in EAF steelmaking, second phase
particles may be responsible for a factor of 5 increase in the bulk viscosity of the slag.
Because the slag mass transfer coefficient is inversely proportional to square root of
viscosity, suspended second phase particles may appreciably reduce the rate of mass
transfer in the EAF.
A series of three simulations were conducted where the V-ratio of the slag and the
phosphorus mass transfer parameter were varied. The specific conditions of the
simulations were as follows:
Table 11: Conditions of EAF Variable V-ratio and Mass Transfer Parameter
Simulations
APrnkQ
Simulation V-Ratio
w m
A 2 0.2
B 3 0.1
C 4 0.05

172
Assuming the temperature and iron oxide profiles used in Section 6.3.1 above, the
equilibrium phosphorus distribution ratios were calculated as functions of time for the
three slag conditions listed in Table 11.

600 Slag A. V = 2

I
Slag 5, V = 3
500 Slag C, V = 4
a
-I

.-3
A 300

I’
-.-E 400 .
]’
I /
g 200
I-

W __-----
___----- --
___----
~

_---
\

100

o ! I I I I I I

0 5 10 15 20 25 30 35
Time
Figure 129: Equilibrium phosphorus distribution ratios for electric furnace simulations
with different V-ratios.

Based on these calculated equilibrium conditions and the mass transfer conditions listed
in Table 11, the phosphorus content of the steel bath was calculated using the numerical
model. The results of the three simulations are shown in Figure 130.

0,025

0.02

0.015
Y
8
5 0.01

0.005

0
0 5 10 15 20 25 30 35
Time (minutes)
Figure 130: Predicted metal phosphorus content during the three simulated EAF heats.

173
Despite the very favorable phosphorus equilibrium conditions of Slag A, the slower mass
transfer kinetics related to the suspended solids in the slag result in a final metal
chemistry similar to that obtained with Slag C. Slag B, representing a compromise
between viscosity and equilibrium phosphorus distribution ratio generates the lowest
metal phosphorus content at tap. The relationship between V-ratio and the mass transfer
parameter, which is proposed in Table 11 is not a fundamental calculated correlation.
Therefore, the present author does not propose that slag B represents the ideal
dephosphorizing slag. The simulations in Figure 130 are intended to be representative of
phenomena, which are real, but not a precise and quantitative evaluation of those
relationships. A diligent review of the literature would likely yield the data necessary to
develop a more accurate correlation between excess solid lime and slag viscosity, and
therefore mass transfer kinetics.

174
 CHAPTER 7

CONCLUSIONS
The primary goal of the present research was to contribute to a better understanding of
the behavior of phosphorus contained in DRI/HBI upon melting. The primary
application of this research was aimed toward improved control of phosphorus in the
EAF when using large amounts of DRI/HBI in the charge. The results of this research
can be broken down into four basic components:

7.1 Investigation of Commercial DRVHBI

An extensive investigation of the physical and chemical attributes of 7 different
commercial DRI/HBI materials was completed. All components of this investigation
were aimed toward the development of a fundamentally based understanding of the
behavior of phosphorus in these materials upon melting. Some of the specific findings of
this study include:
e Scanning electron microscopy indicates that phosphorus in DRI/HBI is present as
a calcium phosphate phase.
e SEM investigation has shown that the gangue feature sizes present in DRI/HBI
are very small. (= 20-200 pm in size) The enonnous area of contact between
these fine gangue particles and the surrounding metal provides a large reaction
area for phosphorus transfer from the gangue to metal upon melting.
e Fast melting experiments with DRI and HBI indicate that a significant amount of
phosphorus is transferred from the gangue to the metal phase upon melting and
that little phosphorus transfer occurs at times greater than one minute. The
lunetics of phosphorus transfer from gangue to metal are much faster then the rate
at which liquid oxide inclusions can physically separate from the liquid metal.
e Since the rate of equilibration between metal and the gangue phases in DRI/HBI
is very fast, the melt-in phosphorus content of the metal can only be controlled by
modifying the gangue chemistry to improve the equilibrium phosphorus
distribution ratio of the gangue.

175
 Small additions of basic components to DRI/HBI materials can alter the melt in
chemistry of the metal. However, this is also heavily dependant upon the amount
of iron oxide and carbon in the material. The ideal DRUHBI material with respect
to phosphorus would melt in with a gangue chemistries very similar to that of a
good steelmaking slag.
In such an ideal material, enough carbon should be present to reduce 80% of the
iron oxide present in the material at the time of use. As necessary flux additions
should be made to the DRI or HBI such that the overall gangue material exhibits a
V-ratio of 1.65 and an iron oxide content of 26%. Only minor flux additions
would be required to from a good steelmaking slag as the DRI/HBI material
melted.
When DRI/HBI is continuously fed to the EAF, excess carbon in material can be
very effective in promoting a foamy slag. When DRI/HBI is batch charged with
scrap, the primary result of excess carbon in DRI is to increase the melt-in carbon
content of the steel. In both cases excess carbon will reduce the FeO content of
the gangue to effectively zero as the material is melted. This will result in a
gangue/slag-make with a very low Lp (Lp < 1.) In the case of continuously fed
DRI, the operational benefits of the excess carbon may out-weigh phosphorus
considerations. In the case of a batch charged process, cheaper sources of carbon
relative to excess carbon in DRI can be obtained. A DRUHBI material, which
contains enough carbon to consume 80% of the iron oxide in DRI and melts-in
with a low metal phosphorus content, should be of greater value to a batch
charged process as compared to a material, which contains excess carbon.
7.2 Laboratory Phosphorus Kinetics Study
A fundamental laboratory study was conducted for phosphorus transfer between metal
and slag.
Laboratory kinetic experiments have shown that the apparent rate of phosphorus
transfer decreases as the dephosphorization and rephosphorization reactions
progress.
Both the average value of the apparent mass transfer coefficient and the extent to
which it varies during dephosphorization, appear to be functions of interfacial

176
 tension, which is in turn a function of the mass flux under conditions of intense
mass transfer.
0 For the present experimental conditions the apparent overall mass transfer
coefficient for dephosphorization was on the order of 0.Olcds at the start of
reaction and decreased to approximately 0.001 c d s toward the end of the
reaction.
The results of the present kinetic study are in good agreement with those of Mori
et a1.[34] and Seetharaman et al. [50]
Review of the literature indicates that the observed variance of the mass transfer
parameter may be related to a dynamic interfacial phenomenon, which is likely to
enhance mass transfer when mass flux across the interface is high and to become
less significant as phosphorus approaches equilibrium.
A transient reduction in the slag-metal interfacial tension during
dephosphorization could lead to either an increase in the reaction area due to
emulsification, an enhancement of the mass transfer coefficient, or both. Based
on data of the present study and information contained in the literature, it is
impossible to isolate the effects of area changes or changes in the mass transfer
coefficient within the mass transfer parameter.

7.3 EAF Plant trials

A series of electric furnace trials were conducted at Pennsylvania Steel Technologies in
order to determine the mass transfer parameter for that furnace. Additional trials were
conducted by North Star Steel - Texas to examine the effects of different types of
DRUHBI materials upon the furnace yield, slag chemistry, and metal phosphorus
behavior.
The mass transfer conditions of a 150 ton DC furnace were evaluated based on
controlled furnace trials. The results of the trials are summarized in terms of
several different kinetic parameters in Table 4 (shown here again):

177
Table 4: Results of Mass Transfer Trials at PST
Mass Transfer Mass Transfer APmko
Coefficient, ko Parameter, A*ko Wm
(assuming furnace area)
cds cm3/s min-'
High Value
Low Value
Best Fit Value
0.99
0.05
0.25
3.5*105
1.7*104
8.7*104
1 1
0.05
0.25
BOF --- --- 1
Laboratory Expts. =: 0.01 =: 0.07 =: 0.16

It was determined that when enough operating variables were held constant or
accurately measured, reasonable analysis of the experimental data yielded very
realistic and reproducible numbers for the phosphorus mass transfer parameter.
The most difficult variables to either control or measure, were the instantaneous
weights of the liquid portion of the slag and metal.
The trials at NSST suggest that the amount of injected oxygen used during
melting has a more significant effect upon the FeO in the slag for a given heat,
than does the FeO contained in DRI materials.
Several general trends were observed in the NSST slag chemistry data behavior as
a function of time during a given heat.
The data from the NSST trials indicate that the unbalanced iron oxide units in
Circa1 can be recovered by increasing the melt-in carbon content of the bath with
pig iron additions. Indications are good that any source of carbon, which first
melts or dissolves into the steel bath will be effective in recovering iron oxide
units from the slag.
Carbon, which remains suspended in the slag may be less effective in recovering
iron units from a high FeO slag. This includes foamy-slag or injected carbon.
Indications are that phosphorus mass transfer between the slag and metal is fast
enough that the details of whether phosphorus from DRI/HBI first enters the slag
or the metal is less relevant to the final metal phosphorus than are other operating
parameters. Factors such as iron oxide content of the slag, excess superheat at
tap, and slag carry over to the ladle have a much stronger influence on the final
phosphorus content of the cast steel.

178
7.4 Process Model for Phosphorus control in the EAF
A process model based upon the plant trials and laboratory experiments was developed
and tested using the plant data from the trials at NSST and data from trials conducted at
BHP Steel Sydney.
A semi-empirical model with a basis in fundamental phenomena was developed to
describe the slag chemistry as a function for a given heat.
It was found that the process model was able to reproduce the observed phosphorus
mass transfer in both the NSST and BHP plant data when the slag chemistry and
temperature profiles were known with accuracy.
In the simulation of a heat, which was continuously fed with HBI for more than 60%
of the charge, the process model indicated that it made very little difference whether
the phosphorus in HBI first melts into the slag or the steel. After feeding 60 tons of
material in a period of 35 minutes, the metal phosphorus content was predicted to be
only 0.002wt% higher, when the phosphorus first entered the metal versus if it
initially enters the sag.
For illustrative purposes, the process model was used to simulate various operating
conditions. The intent of the simulations was to show that the model could be used to
evaluate the effects of various operating parameters on the behavior of phosphorus in
the furnace.

179
 CHAPTER 8

FUTURE WORK
8.1 Laboratory Experiments
A better understanding of the fundamental phenomena, which lead to a dynamic mass
transfer parameter during dephosphorization should be developed.
A more precise definition of the mass transfer parameter, including the precise
influence of parameters such as the slag-metal interfacial tension is needed.
A better understanding of the structural aspects of molten salts is needed. With
respect to the present study, a better understanding of the fundamental mechanism of
phosphate ion formation at the slag-metal interface, and the precise mechanism of
oxygen delivery to the interface from the slag is necessary to develop a better
understanding of dynamic interfacial phenomena.
A fundamental lunetic study should be conducted regarding the mass transfer of
phosphorus, and other species in foamy slags. Most modern electric furnaces take
advantage of the many operational benefits of a foamy slag practice. It is likely that
the kinetics of mass transfer will be different in a highly agitated slag with 25% of its
theoretical density.
During UHP-DC electric furnace steelmalung, more than lo5coulombs per second
(>lo0 kA) of charge are passed through the slag and across the slag metal interface.
Alternatively, the charge transfer involved during typical conditions of
dephosphorization is on the order of lo4or lo3coulombs per second. Because slag-
metal reactions are electrochemical by nature, it would be interesting to examine the
possible effects of arc current upon the thermodynamics and lunetics of slag-metal
reactions important to steelmaking.

8.2 Plant Trials

The plant trials conducted in conjunction with this study were an invaluable
component of this project. Specific knowledge regarding the trends in slag chemistry

180
 and temperature as functions of time during the steelmaking cycle are critical in
arriving at the right answer.
Efforts should continue to develop an understanding of the phenomena, which most
strongly affect the bulk changes in slag chemistry which are observed in EAF
steelmaking. The rate of scrap pile oxidation by oxidizing gasses in the furnace needs
to be better understood. For these experiments, early sampling of the first slag to
form in the furnace is required. Such samples were obtained for a few of the heats
conducted at NSST. If good correlations can be established between various melting
parameters, such as oxygen injection and the iron oxide content of the slag at melt-in,
a more systematic evaluation of the reduction of iron oxide from slag can be realized.
In addition to FeO behavior, a systematic study should be performed of the behavior
of lime and magnesia in EAF slags. The specific relationships, which govern the rate
of lime dissolution need to be better understood.
If the flux dissolution and iron oxide formation and reduction processes are well
understood for a given furnace, it should be possible to optimize the furnace practice
with respect to yield, energy, carbon, and oxygen usage and melting time.
If the normal process parameters, which affect the bulk slag chemistry are well
understood a more systematic and controlled technique can be developed for
comparing the performance of various raw materials in the furnace. With a more
systematic evaluation process, more information can be gathered from fewer trial
heats.
Very large temperature gradients exist in the EAF during melting and refining. A
better understanding of the temperature profiles in the slag and metal during an
individual heat would very useful with respect to modeling of chemical reactions in
the furnace

8.3 Process Model

A useful tool was developed for evaluating the effects of various process parameters
upon the behavior of phosphorus in the steel.
The process relies heavily upon semi-empirical relationships describing the observed
behavior or the bulk slag chemistry of the EAF during steelmaking, i.e. the
dissolution rates of lime and MgO, and the melt in FeO content of the slag. Once the

181
fundamental processes, which govern the behavior of the slag chemistry are known,
they should be included in the process model.
In its present form, the process model can be executed in a symbolic math program.
The model allows for a great deal of flexibility in how the slag chemistry and
temperature data are introduced into the model. If the behavior of the slag chemistry
can be understood from a fundamental perspective, such that its behavior becomes
dependant upon only a few process parameters, the need for this flexibility may be
diminished. Depending upon how the model is to be used in the future, it may be
valuable to code the model such that it is a stand alone executable program. As an
executable program, the model would be much more useful to electric furnace
operators who wish to evaluate various conditions on their own.
For furnaces with very short tap to tap cycles, a significant portion of the refining
processes occur simultaneously with melting. The model should be modified to
account for a continuously melting scrap charge.
In its present form, the process model can be used to predict the phosphorus content
of the steel as functions of time during a heat. With the basic framework of the
phosphorus process model developed, it should be very simple to expand the model to
predict the behavior of other slag metal reactions such as desulfurization.
This model could be combined with others, developed for nitrogen transfer in steel to
provide a very powerful tool for the optimization of electric furnace steelmaking.

182
 Appendix A

Equilibrium Phosphorus
Distribution Correlations

183
 Phosphorus Distribution Equations
Investigator Log (%P)/[%P]

22,350
Healy + 0.08(%CaO) + 2.5. log(%Fet) - 16.00
T

Turkdogan 21,740
+ 0.071 (%CaO + 0.3%Mg0) + 2.5.log(%FeO) - 9.87
T

10,730
Suito et al. - + 4.110log(%CaO+ 0.3%Mg0+%CaF2 -0.05%Fet0)+ 2.50 log(%FeO) +O,S%*log(%P205)- 13.87
T

11,000 1
Zhang et al. + -[162(%CaO) + 127.5(%MgO) +28.5(%MnO)]+ 2.5*log(%FeO) - 6 . 2 8 ~ ( % S O 2 ) - 10.76
T T

Ide and 10,730 + 4.110 log(%CaO+ 0.15%Mg0+%CaF2 -O.OS%Fe,O)+ 2,50log(%FeO) +0.S%*log(%P205)- 13.87
T
Fruehan

184
 Appendix B

Chemical Composition of Slags

Used by Mori et al. [34]

185
Designation Designation
By Mori et al. in Present
Study
FetO cao sio, PP, MgO LP,,"
D 1 33.2 34.9 30.6 0 12.0 100.6
E 2 45.0 18.1 28.5 0 27.9 21.7
C 3 23.1 40.2 35.2 0 10.7 75.3
A 4 42.0 35.3 20.7 0 6.5 164.1
L 5 25.1 25.8 43.8 5.1 23.6 45.45
I 6 69.0 12.0 11.6 7.4 7.5 8.61
F 7 47.3 29.8 16.9 5.3 8.0 245.4
H 8 48.1 25.7 18.6 7.0 9.4 160.1
H 9 48.1 25.7 18.6 7.0 9.7 161.3

186
 Appendix C

Chemical Composition of Commercial DRUHBI Materials

Used in the Present Study

187
Chemical Composition of Investigated Commercial Materials
Material Method of Fe, FeO C SiO, A1,0, CaO P
Reduction
Midrex DRI Midrex 93.13 5.87 1.89 1.92 --- 1.59 0.028
Midrex HBI Midrex 92 10 <OS 1.8 ___ --- 0.06
BHP - HBI Finmet 92 8 1.3 1.6 --- 0.1 0.045
Mobile DRI Midrex 91.2 7 2.43 1.14 0.36 0.3 0.037
Circa1 HBI Circored 97 12 0 1.03 0.32 --- 0.028
Sidor DRI Midrex 93.76 7.6 >3 1.67 --- 2.11 0.054
opco HBI Midrex 91.5 8.4 0.8 -2 ___ --- 0.053

Only limited compositional data were available from the suppliers of these materials.
Shaded data were provided by the supplier of the material. The unknown quantities will
be determined and reported in the final version of this report.

188
 Appendix D

Methods and Results of Chemical Analysis of

Experimental Slags

189
SiO, Gravimetric (SiF, vaporization)
Total Iron Stannous chloride reduction and dichromate titration
CaO KMnO, titration
MgO EDTA titration
P205 Phospho-Molybdenum-Bluephotometric method

sample Time %Si02 %CaO CIS %MgO %FetO wt%(P) LP

(min)
12.11 11 22.58 48.36 2.14 error 20.36 1.41 46.32
12.6 60 22.82 44.48 1.95 error 12.55 1.23 116.63

sample Time %Si02 %CaO CIS %MgO %FetO wt%(P) LP

(min)
13.11 11 24.07 46.21 1.92 error 19.45 1.38 37.50
13.32 32 23.92 34.48 1.44 error 15.12 1.38 111.78

Slag Analysis for Dephosphorization Experiment #18

sample time (min) %Si02 %CaO CIS %MgO %FetO wt%(P) LP
18.a 0 24.47 45.58 1.86 error 19.19 0.04 10.20
18.b 0.5 24.22 45.36 1.87 error 19.89 0.17 0.82
18.c 31 25.74 44.77 1.74 error 13.35 0.81 23.36
18.d 75 23.16 44.10 1.90 10.93 15.58 0.90 38.67

Slag Analysis for Dephosphorization Experiment #19

190
 sample time (min) %Si02 %CaO CIS %MgO %FetO wt%(P) LP
19.e 0 19.82 43.42 2.19 8.03 27.76 0.44 56.62
20.a 0.5 20.79 40.42 1.94 8.90 25.63 0.52 5.16
20.b 31 20.46 39.49 1.93 8.93 24.41 0.84 50.93
20.c 76 21.04 47.99 2.28 8.79 19.64 0.80 60.70

Slag Analysis for Dephosphorization Experiment #21

sample time (min) %Si02 %CaO CIS %MgO %FetO wt%(P) LP
22a 0 25.71 43.63 1.70 12.88 20.30 0.42 44.40
22b 30.5 27.40 38.95 1.42 12.19 17.63 0.69 29.38
22c 75 27.83 37.70 1.35 13.76 17.49 0.71 45.57

sample time (min) %Si02 %CaO CIS %MgO %FetO wt%(P) LP

16a 8.3 60.1 12.62 0.21 5.36 8.52 2.46 171.OO
16b 60 38.3 21.01 0.55 17.23 17.18 1.68 65.97
16c 90 43.7 18.31 0..42 15.25 18.73 1.13 11.24

191
 Appendix E

Example Worksheet for Phosphorus

Process Model

192
 Numerical Kinetic Model of NSST EAF Trials

t := 0,1..30

T ( t ) := 1813 + 0.1.e*23't

C(t) := T(t) - 273

FeO(t) := 25 + 20.e-.12't
B = CIS

B ( t ) := (1.5 + 0.00001.t3)

MgO := 6

(90 - MgO - FeO(t))

Si02(t) :=
+ 1)

CaO ( t) : = Si02 ( t ) . B ( t )

+ .072.(CaO(t)+ 0.15.MgO) + 2 S , l o g ( F e O ( t ) )- 10.5

[WI
11570
Lp(t) := 10

Initial Conditions (cgs)

ko := 38 Dm : = 7 Wm := 150.106 Pm := .02

A := 2.8*105 DS := 2.7 w s := 8.106 Ps := .2

ko
A.Dm.- = 0.497 Mass Transfer Parameter (min-1)
Wrn

Interfacial Metal Phosphorus

P :=

D ( t , P ) := Slag and Metal Flux Equations

1 . 4 . D s . E ) .(Pi ( t ) .Lp(t) - P1)

Z := rkfixed ( P , 0 , 3 0 , 3 0 0 , D ) n := 0 . . 300

193
 50 I I I I I

-
40- * - - -

cao(t)
- - - - _-
FeO( t )
- _. . . . . . . . . . . . . . . . . . . . . . . -
--
S102(t) " - ~ _ x

-
20 -
""I

-MgO
-
-
10 -
I

I I I I I
0

1800 1 Temperature Profile

1.8 I I

t
0
1.6 -

1400
0 1.5 15 22.5 30
1.4 '
0
I
10
I
20
I
30
Time (minutes) t

c -
4
h
v

50 -

0 I I I I I
0 5 10 15 20 25 30
Time (minutes)

Interfacial Metal Phosphorus

0.008 I I I I I

-E
.-
s". 0.006 -

r
-
0.005

0.004 I I I I I
0 5 10 15 20 25 30
Time (minutes)
194
 0.48 - e-...---- _ _ _ _ - - - - - - - - - - - - -- - - T
/ +

h
0.36
b
e
5
0.24

0.12

0 I I I I I
0 5 10 15 20 25 30
Time (minutes)

0.02

0.015
a
3

e
z
0.01

L c II c
0.005 LI
I

01 I I I I I
0 5 10 15 20 25 30
Time (minutes)

195
 REFERENCES:
[ 11 M. Lee, D. Trotter, and 0. Mazzei, “Production of Low Phosphorus and Nitrogen
Steels Using HBI in the EAF”, 1997 Electric Furnace Proceedings, 1997, pp. 467-
472.

[2] “Phosphorus and DRI”, New Steel, April 1994, p l l .

[3] C. Queens, “The Behavior of Phosphorus at Melt-Down of Sponge Iron”, Arch

Eisenhuttenwes, Vol. 46, 1975, pp. 695-700..

[4] R. J. Fruehan and A. W. Cramb, “Fundamental Issues of Steel Refining,” Procedings

of the Julian Szekely Memorial Symposium, TMS, 1997, pp. 209-230.

[5] R. Smailer, H. Jensen, W. Scott “Utilization of Direct Reduction Iron in Electric

Steelmaking”, Proceedings of the Third Interregional Symposium on the Iron and
S tee1 Industry, Brasilia, Brazil, October, 1973.

[6] R. Smailer, J.Price “Operating Benefits of Continuously Charging Prereduced Ore

Briquetts to a 150-Ton Electric Furnace”, OpenHearthProceedings, 1970, pp. 38-43.

[7] R. Smailer, H. Jensen, W. Scott “Utilization of Direct Reduction Iron in Electric

Steelmaking”, AIS1 82”dGeneral Meeting, May 22, 1974.

[8] K. McEuen, “History of Electric Furnace Trials Using Prereduced Iron Bearing
Materials”, Lukens Steel Company - Research and Development Report 73-2, April
1973.

[9] V. Joshi, “Value analysis of Direct Reduction Iron for Electric Furnace Steelmaking”,
Lukens Steel Company - Research and Development Report 76-3, March 1976.

[ 101 A. Lingras, “Some Aspects of Steelmaking with FIOR Briquettes”, Lukens Steel
Company - Research and Development Report 79-9, March, 1979.

Ell] T. Fuwa, “Dephosphorization in Liquid Iron and Steel”, I&SM, June 1981, pp. 18-
25, I&SM, July 1981, pp. 25-31.

[ 12]R. Inoue and H. Suito, “Phosphorus Distribution Between Soda- and Lime-based
Fluxes and Carbon-saturated Iron Melts”, Trans. ISIJ, Vol. 25, 1985, pp. 118-126.

[13] H. Ishii and R.J. Fruehan, “Dephosphorization Equilibria Between Liquid Iron and
Highly Basic CaO-Based Slags Saturated with MgO”, ISS Transactions, Feb. 1997,
pp. 47-54.

196
 K. Ito and N. Sano, “Phosphorus Distribution Between Basic Slags and Carbon-
saturated Iron at Hot Metal Temperatures,” Trans. ISIJ, Vol. 25, 1985, pp. 355-362.

P. Johnston, EfSect of CaF, on Slags used for the External Dephosphorization of Hot
Metal, Masters Thesis, Department of Materials Engineering, The University of
W 011 ongong, 1997.

T. Mori, “On the phosphorus Distribution Between Slag and Metal”, Trans. of the
Japan Institute of Metals, Vol. 25, No. 11, 1984, pp. 761-771.

R.J. Stubbles, “Phosphorus and Sulfur in Steelmaking”, I&SM, March 1986 pp. 34-
35, April 1986, pp. 44-45, May 1986 pp. 50-51, July 1986 pp. 60-61, Aug. 1986 pp.
47, Sept. 1986 pp. 68-70, Nov. 1986 pp. 50-51, Dec. 1986 pp. 34-35, Jan 1987 pp.
52-53, Feb. 1987 pp. 41, March 1987 pp. 45-46, April 1987 pp. 40-41.

[18] H. Suito and F. Inoue, “Effect of Calcium Fluoride on Phosphorus Distribution

Between MgO Saturated Slags of the System Ca0-Mg0-FeOx-Si02and Liquid
Iron,” Trans. ISIJ, Vol. 22, 1982, pp. 869-877.

[19] H. Suito, F. Inoue and M. Takada, “Phosphorus Distribution Between Liquid Iron
and MgO Saturated Slags of the System CaO-MgO-FeOx-SiO2”,Trans. ISIJ, Vol.
21, 1981, pp. 251-259.

[20] E.T. Turkdogan, Steelmaking Fundamentals, Institute of Materials, London, 1996,

pp. 186-189.

[21] E.T. Turkdogan and J. Pearson, “Activities of Constituents of Iron and Steelmaking
Slags”, JISI, Vol. 173, 1953, pp. 217-223; vol. 175, 1953, pp. 393-
401; vol 176, 1954, pp. 59-63.

[22] G.J.W. Kor, “Effect of Fluorspar and Other Fluxes on Slag-Metal Equilbria
Involving Phosphorus and Sulfur”, Metall. Trans., Vol. 8B, 1978, pp. 107-113.

[23] G.W. Healy, “A New Look at Phosphorus Distribution”, JISI, Vol. 208, 1970, pp.
664-668.

[24] K. Balajiva, A.G. Quarrell, and P. Varjagupta, “A Laboratory Investigation of the

Phosphorus Reaction in the Basic Steelmaking Process”, JISI, Vol. 153, 1946,
pp. 115-150.

[25] K. Balajiva, A.G. Quarrell, and P. Varjagupta, “The Effect of Temperature on the
Phosphorus Reaction in the Basic Steelmaking Process”, JISI, Vol. 155, 1947, pp.
563-567.

[26] 0. Wijk, “The Phosphorus Problem in Iron and Steelmaking”, Scandinavian Journal
of Metallurgy, Vol. 22, 1993, pp. 130-138.

197
[27] T. Winkler and J. Chipman, “An Equilibrium Study of the Distribution of
Phosphorus Between Liquid Iron and Basic Slags”, Trans. AIME, Vol. 167, 1946.
pp. 111-133.

[28] X.F. Zhang, I.D. Sommerville, and J. M. Tiger, “An Equation for the Equilibrium
Distribution of Phosphorus Between Basic Slags and Steel”, ISS Trans. Vol. 6, 1985,
pp. 29-34.

[29] K. Ide, Evaluation of Phosphorus Reaction Equilibrium Steelmaking and Kinetics in

Foaming Slag, Masters Thesis, Department of Materials Science and Engineering,
Carnegie Mellon U., 1999.

[30] T. Forland, and K. Grjotheim, Met. Trans. B., Vol. 8, 1977, pp. 645-650.

[31] H.Flood and K. Grjotheim, J.Iron and Steel Inst., Vol. 171, 1952, pp. 64-70.

[32] C.Garlic, G.R. Belton and S. Jahanshahi, “A Thermodynamic Study of Metal/Slag

Phosphorus Equilibria,”

[33] P. Johnston, EfSect of CaF, on Slags used f o r the External Dephosphorization of Hot
Metal, Masters Thesis, Department of Materials Engineering, The University of
Wollongong, 1997.

[34] K. Mori, S. Doi, T. Kaneko, and Y. Kawai, “Rate of Transfer of Phosphorus

Between Metal and Slag”, Trans. ISIJ, Vol. 18, 1978, pp. 261-268.

[35] M. Motlagh, “Dephosphorization of Steel During Melting”, ISS Trans. Vol. 6, 1985,
pp. 202-213.

[36] W.K. Lewis and W. Whitman, Industrial and Engineering Chemistry. Vol. 16, 1924
pp 1215.

[37] D.R. Poireir and G. H. Geigger, Transport Phenomena in Materials Processing.

TMS, Warrendale-PA, 1994, pp. 476-538.

[38] E.L. Cussler, Difision - Mass Transfer in Fluid Systems. Cambridge University
Press, New York, 1984, pp. 215-307.

[39] A.H.P. Skelland, DifSusionaZ Mass Transfer. John Wiley &Sons, New York, 1974,
pp, 29-39 & 83-307.

[40] T.K. Sherwood and R.L. Pigford, Boundary Layer Theory., 7d1ed., McGraw-Hill,
New York, 1975.

[41] R. Higbie, Trans. AIChEJ. Vol. 31, 1935, p 365.

[42] H. Toor and J.M. Marchello, “Film-Penetration Model for Mass and Heat Transfer”,
AIChEJ, Vol. 4, No. 1 1958, pp. 97-101.

198
[43] S. Mukawa, and Y. Mizukami, “Effect of Stirring Energy and Rate of Oxygen
Supply on the Rate of Jot Metal Dephosphorization”, ISIJ International, Vol. 35,
1995, pp. 1374-1380.

[44] P. Wei, M. Sano, M. Hirasawa, and K. Mori, “Kinetics of Phosphorus Transfer

Between Iron Oxide Containing Slag and Molten Iron of High Carbon Concentration
Under Ar-02 Atmosphere”, ISIJ International, Vol. 33, No. 4, 1993, pp. 479-487.

[45] S. Ohguchi, D.G. Robertson, B. Deo, P. Grieveson, and J.H. Jeffes, “Simultaneous
Dephosphorization and Desulfurization of Molten Pig Iron”, Ironmaking and
Steelmaking, 1984, Vol. 11, No. 4, pp. 202-213.

[46] S. Kitamura, T. Kitamura, K. Shibata, Y. Mizukami, S. Mukawa, and J. Nakagawa,

“Effect of Stirring Energy Temperature and Flux Composition on Hot Metal
Dephosphorization,” ISIJ International, Vol. 31, 1991, No. 11, pp. 1322-1328.

[47] B.J. Monaghan, R.J. Pomfret, and L.S. Coley, “The kinetics of Dephosphorization of
Carbon-Saturated Iron Using an Oxidizing Slag” Met. Trans B, Vol. 29B, 1998
pplll-118.

[48] Yongsug Chung and Alan Cramb, “Direct Observation of Spontaneous

Emulsification and Associated Interfacial Phenomena at the Steel-Slag Interface,”
Phi. Trans. Royal Society of London, Vol. 356, 1998, pp.981-993.

[49] H. Gaye, L.D. Lucas, M. Olette, and P. V. Riboud, “Metal-Slag Interfacial

Properties: Equilibrium Values and ‘Dynamic’ Phenomena,” Canadian
Metallurgical Quarterly, Vol. 23, No. 2, 1984, pp.179-191.

[50] A. Jakobsson, M. Nasu, J. Mangwiru, K.C. Mills, S. Seetharaman, “Interfacial

Tension Effects on Slag-Metal Reactions,” Phi. Trans. Royal Society of London,
Vol. 356, 1998, pp.995-1001.

[51] A. Jakobsson, M. Nasu, B. Monaghan, K.C. Mills, S. Seetharaman, “Effect of

Slagmetal Interfacial on Kinetics of Dephosphorization,” Iron and Steelmaking,
Vol. 26, NO. 5 , 1999, pp.353-357.

[52] L. Gu, G. Irons, “Physical and Mathematical Modeling of Oxygen Lancing and Arc
Jetting in EAF Furnaces,” 1998 Electric Furnace Proceedings, 1998, pp. 269-277.

[53] D. Guo, L. Gu, G. Irons, “Evaluation of Stirring in Electric Arc Furnaces,” 2000
Electric Furnace Proceedings, 2000, pp. 223-233.
[54] D. Goldstein, Nitrogen Reactions In the EAF and OSM, Doctoral Thesis,
Department of Materials Science and Engineering, Carnegie Mellon U., 1996.

[55]J.Kerr, and R.J Fruehan, “Foamability of Stainless Steelmaking Slags in the EAF,”
2000 Electric Furnace Proceedings, 2000, pp. 1049-1063.

199
[56]R.J Fruehan, ed., The Making Shaping and Treating of Steel - I l l h Edition-
Steelmaking and Refining Volume, AISE, Pittsburgh, 1998, P. 103.

200