Non-metallic Inclusion Evolution during Argon Oxygen Decarburization and Ladle Processing

Abstract

In this work we use thermodynamic and kinetic models to predict evolution of spinel and MgO non-metallic inclusions after argon oxygen decarburization (AOD) processing of an aluminum-deoxidized stainless steel. Inclusions form during liquid steel processing and are generally detrimental to downstream processing and steel properties. Models predicting inclusion evolution pathways have typically been applied to lower alloy content carbon steels. In this work we investigated a steel composition with alloying elements such as Cr, Co, Mo, and Nb, processed in the AOD, which utilized higher stirring rates and temperatures compared to ladle refining. Calculations were made with FactSage, with effective equilibrium reaction zone (EERZ) kinetic calculations made using the macro processing feature. Results were compared to plant sample compositions and automated inclusion analysis. The high temperatures and stirring rates present in AOD processing were observed to facilitate Mg transfer to the steel and inclusions. Modeling without the Mg*O associate in FactSage more closely fit the observed data. Additional alloying elements did not significantly change slag/metal reactions or inclusion compositions in modeling, but the strong stirring led to inclusion generation processes (reoxidation or slag entrainment) that were not accounted for in the present work. The reduction stage of AOD processing was found to be vital to inclusions in this process route.

1. Introduction

Endogenous non-metallic inclusions form during liquid processing of steels via reactions with dissolved species such as O, S and N. Control of inclusion compositions, sizes, shapes, and amounts can measurably affect steel properties. It has become essential to engineer inclusions for the required purpose of the steel.¹⁾

Much of inclusion engineering is accomplished during ladle processing, with many previous studies on inclusion/metal and slag/metal reactions.^{2,3,4,5,6,7,8,9,10,11)} However, processing before and after ladle treatment can affect inclusions in the final product, particularly with some of the additional steps found in specialty steelmaking. The important process variables for inclusion evolution are temperature, time, steel composition, slag composition, and refractory composition. For example, high temperatures increase the solubility of species such as Mg in the melt, increasing the driving force for certain slag/metal reactions. Spinel formation has been noted in previous studies of inclusions in stainless steels.^{12,13,14,15,16,17,18)} This subject has been reviewed by Park and Todoroki¹³⁾ in detail. Additionally, in a previous study,¹⁷⁾ the present authors characterized inclusions from an aluminum-deoxidized, martensitic stainless steel during ladle processing, finding spinel inclusions immediately prior to tap of the argon oxygen decarburization (AOD) vessel and evolution to MgO prior to Ca treatment. It was apparent that slag/metal reactions had already occurred in the AOD.

AOD is a common specialty steelmaking process to prevent loss of valuable alloying elements such as Cr at low C contents^19,20) by controlling the partial pressure of O₂ (p_O2). There are two main stages to AOD processing: decarburization and reduction. During the reduction stage oxidized metals (particularly Cr) are recovered. In addition, deoxidation, desulfurization, and alloying are performed, followed by ladle treatment for final alloy additions and temperature control. AOD processing typically involves significantly higher temperatures than ladle processing combined with, during the reduction stage, low dissolved oxygen and deoxidized slags that are MgO-saturated.²¹⁾

The kinetics of AOD processing have been modeled in many studies, most of which focus on the decarburization stage, as reviewed by Visuri et al.,^22,23) or Wei et al.^24,25) Kinetic models of the reduction stage are far more limited. A kinetic model has been proposed by Visuri^22,26) focusing on the rate of Cr recovery from slag, combining work on slag emulsification with AOD reduction. High stirring rates are often present to accelerate recovery of Cr from the slag, resulting in significant slag emulsification and relatively fast reaction kinetics.²²⁾ However, the effect of the high temperatures and high stirring rates in AOD processing on non-metallic inclusions was not the focus of these modeling studies.

The impact of slag/metal reactions on inclusion composition, particularly Mg content in aluminum-deoxidized steels in contact with MgO-saturated slags, is well documented.^{16,17,18,27,28,29)} Mg-containing inclusions can be detrimental, as they can impact castability and final properties.^13,15,30,31) Sources of Mg in inclusions include reactions at the slag/metal interface (because an Al-deoxidized steel with Al₂O₃ inclusions is not in equilibrium with an MgO-saturated slag) and entrainment of slag as the steel and slag mix during stirring.³²⁾ The stirring rates during AOD processing are expected to lead to expected significant emulsification and entrainment.^22,26) Concurrently, reactions between the liquid metal and MgO-containing refractory materials present can also affect the amount of Mg in the steel. However, assuming that the slag is saturated in MgO, these reactions should not affect the overall equilibrium and are typically slower than reactions with the slag.

In Al-deoxidized steels, the reaction occurring between the liquid steel and slag is shown in Eq. (1):   

$$\begin{array}{c}2{[\mathrm{A}\mathrm{l}]}_{Steel}+3{(\mathrm{M}\mathrm{g}\mathrm{O})}_{Slag}\to 3{[\mathrm{M}\mathrm{g}]}_{Steel}+{(\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3)}_{Slag}\end{array}$$(1)

Mg transfer has been observed in steels with high alloy Al levels^27,33) and those that are deoxidized with levels of Al below 500 ppm.³⁴⁾ Temperatures well above 1600°C, as in the AOD, have been observed to lead to increased Mg pickup.³³⁾ Once dissolved Mg is in the steel, the following reaction can occur at inclusions, Eq. (2).   

$$\begin{array}{c}\begin{array}{l}3{[\mathrm{M}\mathrm{g}]}_{Steel}+4{(\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3)}_{Inclusion}\to \\ 2{[\mathrm{A}\mathrm{l}]}_{Steel}+3{(\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3*\mathrm{M}\mathrm{g}\mathrm{O})}_{Inclusion}\end{array}\end{array}$$(2)

Sufficient Mg levels can even drive the reaction toward Eq. (3) at low total oxygen levels, leading to MgO inclusions.^5,27)   

$$\begin{array}{c}3{[\mathrm{M}\mathrm{g}]}_{Steel}+{(\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3)}_{Inclusion}\to 2{[\mathrm{A}\mathrm{l}]}_{Steel}+3{(\mathrm{M}\mathrm{g}\mathrm{O})}_{Inclusion}\end{array}$$(3)

Modeling of ladle treatment serves a useful purpose in examining the impact on inclusion populations of a range of processing conditions and compositions. Slag/metal reactions, inclusion generation and evolution, and overall mass transfer kinetics in ladle refining have been studied across a wide range of low alloy steels.^{3,6,8,9,11,35,36)} Effective Equilibrium Reaction Zone (EERZ) models have been widely used for this purpose, as they incorporate both thermodynamics and kinetics during treatment. An EERZ model applied to ladle treatment assumes local equilibrium at the slag/metal interface, with mass transfer to the interface in the steel the limiting factor.⁴⁾

Steels produced via the AOD generally have more alloying elements and are exposed to higher temperatures and higher stirring rates compared to lower alloy steels. In this work, industrial data combined with equilibrium and kinetic calculations were used to understand the evolution of inclusions during the final stages of AOD processing. Additional alloying elements were not expected to affect inclusion populations because of the high affinity for Al and Mg for dissolved oxygen, but AOD processing conditions were expected to exert significant influence over inclusion populations. An EERZ model was applied to the reduction stage of AOD treatment with modified steel compositions to observe predicted inclusion evolution and slag/metal reactions. The goal was to evaluate the gaps in a basic EERZ that must be filled to accurately forecast inclusions and to understand the additional thermodynamic and kinetic considerations for high temperature AOD process routes.

2. Materials and Methods - Industrial Samples

The intent of this study was to evaluate slag/metal reactions and inclusion evolution in a 10.5Cr steel for an electric arc furnace (EAF)-to-AOD-to-trim ladle processing route. The AOD size was 50 tons, with approximate slag weight of 5 tons. During the reduction stage, Al additions were made both for deoxidation and for fuel (with O₂ stirring) to increase temperature. Temperatures at reduction were typically near 1750°C, as shown in the “T_max” column of Table 1 and then declined until AOD tap (T_min). Average temperatures in the ladle were closer to 1650°C.

Table 1. Maximum, minimum, and average temperature data for six heats between deoxidation Al addition and tap during AOD processing.

Heat	Tmax (°C)	Tmin (°C)	Tavg (°C)
1	1763	1701	1732
2	1708	1648	1679
3	1761	1689	1725
4	1706	1686	1696
5	1745	1673	1709
6	1768	1705	1736
Avg.	1742	1684	1713

After Cr reduction, deoxidation, and desulfurization, there was approximately three minutes of Ar stirring at 0.5–0.6 Nm³/min per ton of steel. During trim ladle treatment, minor alloying additions and FeCa treatment were performed while waiting for the heat to reach pouring temperature. The metal was then bottom poured and bottom teemed into ingot molds.

Bulk composition sample data was available for six heats of the studied steel at the end of AOD processing, reported in Table 2. Steel samples were taken via spoon from the AOD or ladle and solidified in silica/alumina refractory molds of approximately 3 cm inner diameter and 8 cm height. Metal samples were analyzed for composition by X-ray fluorescence (XRF) and optical emission spectroscopy (OES) at the plant. One slag sample was available, taken concurrently with the metal sample from Heat 1 and shown at the bottom of Table 2. Slag samples were analyzed at an outside laboratory by XRF.

Table 2. Steel compositions at the end of AOD treatment for the industrial heats studied in this work. All values are in weight percent, except * columns, which are mass ppm. Also included is the slag composition at the end of AOD from Heat 1.

Heat	Mo	Nb	Ni	Co	Mn	Cr	V	Ca*	S*	Si	Al*	Mg*	C	Fe
1	0.74	0.28	0.48	5.72	0.82	10.58	0.22	4	<30	0.31	534	20	0.10	Bal
2	0.67	0.33	0.43	5.63	0.69	10.68	0.21	6	<20	0.41	189	14	0.10	Bal
3	0.71	0.32	0.47	5.55	0.80	10.96	0.26	8	<30	0.49	673	27	0.09	Bal
4	0.73	0.30	0.44	5.36	0.81	11.50	0.26	5	<30	0.30	286	15	0.13	Bal
5	0.71	0.30	0.43	5.53	0.80	11.09	0.23	15	<20	0.35	229	18	0.11	Bal
6	0.72	0.32	0.40	5.60	0.83	11.29	0.25	6	<20	0.32	365	13	0.10	Bal

	Heat 1 Slag	CaO	Al2O3	MgO	SiO2	FeO	MnO	Cr2O3
		48.5	36.9	10.6	3.3	0.2	0.13	0.34

Six total liquid steel and slag samples were also taken from Heat 1 after AOD tap and during ladle processing (further details provided in Huck et al.¹⁷⁾). Steel samples were analyzed for inclusion contents and composition using automated feature analysis (AFA) on an ThermoFisher/FEI Aspex Explorer SEM according to the parameters in Table 3. Detected features were filtered to only those containing <400 counts of C and either Al counts > 500, Mg counts > 500, or S counts > 300 with Mn or Ca counts > 200. Bound O was calculated by converting the areal detected concentration into mass ppm.

Table 3. General AFA analysis parameters for inclusion detection.

Parameter	Value
Acceleration Voltage (kV)	10
Steel matrix brightness	170
Aluminum tape brightness	45
Detection Threshold	115
Spot Size	45%

3. Thermodynamic and Kinetic Modeling

Modeling was performed to examine both slag/metal and inclusion/metal equilibria as well as the kinetics of both slag/metal reactions and inclusion evolution. Emphasis was placed on processing conditions that approximated the final stages of the AOD, where a deoxidized steel was in contact with MgO-saturated slag. Specific focus was also given to the effect of alloying elements such as Cr, V and Nb.

Modeling was performed using of FactSage^TM 8.0³⁷⁾ with databases FTmisc, FToxid, and FactPS. The FTmisc database was intended specifically for deoxidation equilibria, and utilizes fictitious associate species to account for strong interactions between cations such as Al, Ca, Mg, Cr, and others, and solution O.³⁸⁾ Based on the results of previous investigations,^6,39) the Ca*O associate species was excluded for all calculations in this investigation. Additionally, it became apparent during the course of modeling and from prior investigations¹⁷⁾ that the Mg*O associate has a significant impact on the amount of Mg pickup from predicted from the slag at temperatures above 1600°C, so modeling with and without this associate was completed for comparison.

Inclusion/metal equilibrium calculations were performed for Heat 1 using three steel and two slag composition sets, as shown in Table 4. The intent of this was to evaluate the effect of additional alloying elements, particularly Cr, on the overall equilibria calculated from FactSage^TM to understand if further thermodynamic experiments and assessments at these temperatures are necessary to fit observed data. The “Full” composition included all alloying elements and elements associated with deoxidation and inclusions (Al, Mg, Ca, Mn, Si). The “Add Cr” composition included those associated with deoxidation and inclusions as well as Cr. The “Basic” composition included only those typically associated with deoxidation and inclusions. O levels and temperature were varied to understand their impact on inclusion compositions.

Table 4. Compositions used for construction of inclusion stability diagrams based on Heat 1 composition. Values are in weight %, with values followed by * in mass ppm. “Average” line denotes the average composition from the six heats in Table 2.

Comp.	Mo	Nb	Ni	Co	Mn	Cr	V	Ca*	S*	Si	Al*	Mg*	C	Fe
Full	0.74	0.28	0.48	5.72	0.82	10.58	0.22	4	30	0.32	550	20	0.10	Bal
Add Cr	0.00	0.00	0.00	0.00	0.82	10.58	0.00	4	30	0.32	550	20	0.10	Bal
Basic	0.00	0.00	0.00	0.00	0.82	0.00	0.00	4	30	0.32	550	20	0.10	Bal
Average	0.71	0.31	0.44	5.56	0.79	11.02	0.24	7	5	0.36	379	18	0.11	Bal

Additionally, slag/metal equilibrium calculations were performed using the average composition of Heats 1–6 and the slag composition in Table 2, equivalent to the “Add Cr” line in Table 5. For equilibrium calculations, the variables of interest were Al and Mg levels and temperature. The effect of including and excluding the Mg*O associate species was also investigated.

Table 5. Slag compositions used for modeling of AOD treatment.

Species	CaO	Al2O3	MgO	SiO2	FeO	MnO	Cr2O3
Basic	48.7	37	10.6	3.3	0.2	0.14	0
Add Cr	48.5	36.9	10.6	3.3	0.2	0.13	0.34

An EERZ model was used to examine the slag/metal reaction kinetics using the macro processing feature³⁵⁾ of FactSage^TM and the influence of additional alloying elements. Two slag compositions, based off an end of AOD Heat 1 sample and shown in Table 5, were used, with the “Add Cr” composition containing a small amount of Cr₂O₃ used when Cr was included. Steel compositions were as in Table 4 except for starting Mg levels reduced to 3 mass ppm and starting O levels of 50 ppm to explicitly model pickup of Mg. For the period of interest in this work, during the end of the reduction stage after Al+O₂ injection for heating, the slag/metal reactions in the AOD vessel were assumed to occur as they do in the ladle.³⁾ Modeling in the ladle for Heat 1 has been covered in a previous work.¹⁷⁾

A schematic for the modeling process can be seen in Fig. 1. In each timestep, a certain amount of steel and slag are equilibrated, with metallic products returned to the metal and oxide products to the slag. The amount of slag and metal that react are based on an overall mass transfer coefficient and interfacial area for each phase, as shown in Eq. (4) for steel and Eq. (5) for slag.   

$$\begin{array}{c}{V}_{Steel}={(mA)}_{Steel}{\rho }_{Steel}\mathrm{\Delta }t\end{array}$$(4)

  

$$\begin{array}{c}{V}_{Slag}={(mA)}_{Slag}{\rho }_{Slag}\mathrm{\Delta }t\end{array}$$(5)

Fig. 1.

Schematic view of EERZ model (not to scale). Zones 1–4 are respectively: 1-bulk steel (includes 2), 2-interfacial or reaction volume steel, 3-interfacial or reaction volume slag, 4-bulk slag (includes 3).

The (mA) parameter is an effective mass transfer coefficient that accounts for both the kinetic parameter and the reaction area, a simplification over other models, such as that of Visuri.^22,26) Inclusion removal occurred at the same rate as slag/metal reactions, with no special rate constant assigned for flotation. No alloy additions or reoxidation was included. Further details of this model can be found in previous work.^3,4,36)

The (mA) parameter is a key input parameter for this model. In previous studies it has been calibrated to experimental data or based on correlations that include stirring rate.^{3,6,40,41,42)} This work used an empirical correlation fitted to the data of Ishii et al.⁴³⁾ that yields an (mA) term for the steel of 0.44 m³/s. Ishii et al.⁴³⁾ calculated stirring energy based on ladle depth, temperature, and Ar flow rates and plotted this stirring energy divided by the ladle cross-sectional area against the mass transfer coefficient obtained from S removal. An equation was fit to this data to provide an empirical prediction for (mA) where A is the cross-sectional area, rather than the total reaction area (which is unknown). Because of the high stirring rates in the AOD vessel, the fitted correlation was extrapolated beyond those used to simulate slag/metal reactions in a ladle.⁴⁰⁾

To assess the validity of this value, it was compared to that used in a study of AOD kinetics during the reduction stage from Visuri et al.,^22,26) which explicitly modeled slag droplet surface area and slag/steel mass transfer. In this previous work,²⁶⁾ assuming a 150-ton AOD with a 1.5 m plume diameter, total slag droplet surface area was found to be approximately an average of 1000 m² in the first three minutes of stirring after reduction and the slag/steel mass transfer coefficient was calculated to be 7.34–7.81 × 10⁻⁴ m/s. Assuming the 50-ton AOD in the current study had a plume diameter one-half that of the 150-ton AOD reported by Visuri et al.,²⁶⁾ the total droplet surface area would be 500 m² from the reported equations,²²⁾ as plume diameter and droplet surface area were linearly related. Using the lower reported value of the mass transfer coefficient, the calculated mA for the conditions of this study would be 0.37 m³/s, close to that calculated from the Ishii correlation. Mass transfer in the slag (m_slag) was assumed to be 0.1 m_steel, again based on previous results.⁴¹⁾ AOD and ladle steel masses were assumed to be 50 tons, with a slag mass of 5 tons.

4. Results – Inclusions in Industrial Samples

The inclusion compositions of samples from Heat 1 are shown in Fig. 2. These ternary diagrams are in mole fraction of the inclusion cation species, assumed to be in their respective oxide species. Triangle sizes correspond to the area fraction of inclusions detected in that composition, with the largest triangle as a percentage of the total area shown at the upper right of each diagram.

Fig. 2.

Inclusion compositions for samples from Heat 1. a) is just before AOD tap, b) in the ladle after AOD tap and c) after treatment with FeCa. Diagram axes are cation mole fractions.

The composition of deoxidizing elements through these samples can be seen in Fig. 3. Total Mg levels were initially about 20 ppm and total Ca levels were less than 5 ppm. An increase in Ca occurred only after Ca treatment. Some Al fade was observed upon transferring from the AOD to the ladle.

Fig. 3.

Change in composition of Mg (solid line, squares), Al (Dotted line, triangles), and Ca (dash-dot-dot, circles) during Heat 1. Al is plotted on the left axis, with Mg and Ca at the right. Sample letters correspond to Fig. 2.

The area density and number density of inclusions in Heat 1 detected by AFA can be seen in Fig. 4. Densities both markedly decreased between the sample before AOD tap and the first ladle sample.

Fig. 4.

Area and number density of detected inclusions through treatment of Heat 1. Samples letter correspond to Fig. 2.

5. Results - Modeling

Mg levels observed in the six heats of Table 2 measured by OES were related to measured Al levels, as can be seen in Fig. 5. FactSage equilibrium predictions for the average composition of the heats with and without the Mg*O associate enabled are also shown for comparison at 1700°C and 1750°C. Note that the curve for the results with the associate at 1750°C is well above the plotted axes and is not shown. To construct this plot, the slag activity of MgO was fixed at 1 and the activity of Al₂O₃ was varied to simulate a range of slag compositions.

Fig. 5.

Mg and Al levels measured by OES for ten different heats, with AOD tap for 6 heats shown. Also shown are equilibrium calculations from FactSage of dissolved species with and without Mg*O enabled and at two different temperatures. Curve with Mg*O at 1750°C is well off scale.

An inclusion stability diagram at 1750°C, showing total stable inclusion amounts with varying total O levels in the steel, for the three compositions listed in Table 4 is shown in Fig. 6. Mg*O was excluded for these calculations. The most significant difference was that the addition of Cr reduced total inclusion amount at the same O content compared to the “Basic” composition. The stable phase fields (not pictured) remained essentially the same but were shifted slightly in total O concentration. At 1650°C, which is not shown, there was no discernible difference between the compositions.

Fig. 6.

Amount of predicted inclusions vs. total O at 1750°C. Changes in slope correspond to phase fields, which were included in Fig. 7.

A set of inclusion stability diagrams for the “Full” composition at 1650°C and 1750°C are shown in Figs. 7(a) and 7(b), respectively. Total inclusion content, shown in the bold top line, increased at 1650°C and compared to 1750°C and the MgO phase field greatly expanded. Again, Mg*O was excluded in these calculations.

Fig. 7.

Inclusion stability diagrams using the “Full” composition at 1650°C (a) and 1750°C (b). Labeled regions are solution phases of mostly Al, Mg, Ca, O and S.

Results from EERZ modeling comparing inclusion evolution and overall mass transfer of Mg and O using the three different compositions are shown in Fig. 8. Mg*O was excluded in these models. These calculations focused on comparing compositional effects and timescales. As seen below, changes occur very rapidly under the mass transfer conditions simulated. Plume turbulence and slag emulsification will result in reoxidation and slag entrainment in the steel, which were not incorporated.

Fig. 8.

(a) EERZ modeling results at 1750°C of inclusion composition for progression of mass %Mg to total %Mg and %Al. (b) Corresponding results showing pickup of Mg (increasing) and removal of O (decreasing), with O leaving the melt principally with inclusions.

Inclusions evolved to a stoichiometric spinel composition and then towards MgO as Mg transfer occurred and inclusion removal reduced total O. Values of Mg and O close to equilibrium were achieved by about one minute into modeling. This was reasonably consistent with the kinetics of Cr recovery shown by Visuri et al.,²⁶⁾ with near complete recovery of chromia from the slag within three minutes in that study.

6. Discussion

Industrial data from AOD and ladle treatment of a 10.5Cr Al-deoxidized specialty steel was coupled with FactSage^TM modeling to understand inclusion evolution and slag/metal reactions during this process route. Thermodynamic and kinetic models were evaluated in context of these experimental results.

Observed inclusion evolution in the studied Al-deoxidized specialty steel stayed almost entirely within the CaO–MgO–Al₂O₃ system, much as in other low carbon aluminum-deoxidized steels. This was in accord with the results of Park,¹⁴⁾ with Cr₂O₃ only appearing in inclusions in a 16.5Cr steel at Al levels below 60 ppm. The ternary diagrams shown in Fig. 2 represented more than 95% of the cationic species in the detected oxides, with very few sulfide inclusions observed due to low S of these steels.

Spinel inclusions were observed in the AOD and MgO inclusions were observed in the first sample after tapping into the ladle. Reoxidation during sampling likely occurred. Previous studies^31,44) have found that reoxidation inclusions are typically spinel in Al-deoxidized, Mg-containing steels, suggesting that the inclusions in Fig. 3(b) near spinel composition were a result of reoxidation. The inclusion stability diagram in Fig. 7 indicated that MgO forms because of an expansion of the stable phase field with a drop in temperature from 1700–1750°C in the AOD to 1650°C in the ladle. This may be further influenced by the drop in bound O reflected in the drop in total inclusion area in Fig. 4, which should reflect a drop in total O, moving left in Fig. 7(a) or 7(b). This points to a drop in temperature and total O in the steel as the cause for the appearance of MgO inclusions in this processing route.

Across the six studied heats, there was a clear relationship between Al and Mg levels. This trend was quite clear in Fig. 5, despite variations in slag compositions, temperatures, and other treatment conditions between the different heats shown and between the different conditions in the ladle and AOD vessel. What is significant in this processing route is the degree of Mg transfer at relatively low total Al in the metal (100–800 mass ppm), with levels often greater than 20 mass ppm Mg. The combination of high temperatures at the end of AOD processing (typically 1700–1750°C), MgO saturated slag, and reducing conditions facilitated this high Mg transfer.

Given the high Mg levels, this data enabled evaluation of thermodynamic parameters and models, particularly of the Fe–Mg–O system. Deviations in inclusion/metal behavior in the Fe–Mg–O system from varied thermodynamic data have been reported by Pretorius et al.³¹⁾ and were investigated in a previous study of ladle samples.¹⁷⁾ Additionally, in FactSage, the data used for the Mg*O associate is based on data collected at 1650°C and below.^45,46) Figure 5 showed that including the presence of Mg*O at high temperatures increased the amount of predicted Mg pickup far beyond observed levels. This implies either that this associate model is not sufficiently accurate at these temperatures, or that the AOD vessel is far from slag/metal equilibrium. This would only occur with slow kinetics not typically observed in secondary steelmaking (such as desulfurization kinetics^41,42)) and far slower than previously studied AOD kinetics.^22,24,25,26) While not shown, a similar effect is present for solution O since it must be paired with Mg in associate form. At 1750°C, calculated O soluble in the liquid steel for the “Full” composition, which includes O in associate species, was 51–56 mass ppm at equilibrium, an unrealistically high value that exceeds detected bound O calculated from the area fraction of Fig. 4. Without Mg*O, soluble O was 6–9 ppm for the same conditions. Further experiments at 1700°C and above are necessary to validate the Fe–Mg–O system thermodynamics, but the current fit of the data without Mg*O was superior, which justified its absence in the remainder of modeling in this study. Submerged industrial sampling, small scale induction furnace experiments similar to those by Liu et al.⁵⁾ but at higher temperatures, and comparison to published thermodynamic data and models will help to define the true system thermodynamics.

The effect of adding additional alloying elements in specialty steels appeared relatively minor comparing curves in the inclusion stability diagram shown in Fig. 6. Cr had the most effect, decreasing total inclusion content at the same total O. This is mostly a result of Cr*O and Cr₂*O, the two associate species, which consumed some solution O and reduced the amount available to form inclusions. Additional elements in the “Full” composition move total inclusion content back toward the “Basic” composition. While not shown, the phase fields maintained relatively the same sizes and positions for all three compositions (see field in Fig. 7(b)), suggesting the effect of these additional elements on modeling overall is minor.

EERZ modeling results, shown in Fig. 8, also supported this conclusion, with inclusion evolution proceeding in an identical manner in with only slight differences in timing. The timing differences can largely be attributed to initialization of all three simulations with the same O content. During the simulations, the associates (particularly Cr) reduce available solution O, which accelerates pickup of Mg by the inclusions. This is apparent in Fig. 8(b), with eventual solution O levels from the “Add Cr” composition roughly twice those in the “Basic” composition. Aside from differing prediction in dissolved O, slag/metal reactions in (b) show only minor difference from the different compositions, suggesting the additional elements do not play a strong role in results. Experiments using lab setups with much lower mass transfer rates, such as induction furnace experiments, may provide data that could evaluate differences in reaction kinetics.

It is important to note that the predicted inclusion removal that proceeds as fast as overall mass transfer does not fit observed AOD results, with many inclusions left after the vigorous stirring of deoxidation in the industrial data. After one minute in Fig. 8, virtually all inclusions have been removed and only dissolved O remains, which is certainly not the case in Fig. 2 or 4. This also results in MgO inclusions at these low O levels, which was not the result observed in the AOD data. It is likely that the strong emulsification important to facilitating Cr recovery kinetics either explicitly re-adds inclusions to the steel directly through emulsification or provides an opportunity for reoxidation by atmosphere at the surface of the steel around the plume. More study of this is necessary to accurately predict inclusion evolution in the AOD vessel at high stirring rates with an EERZ model.

Modeling of specialty steels appears amenable to treatment with FactSage and an EERZ model, with some adjustments to solution thermodynamics necessary and some additional effects incorporated. In less sensitive cases, some elements can be ignored as they only appear to cause small changes. However, it is apparent that the differential processing conditions present in some specialty alloys may play a much greater role in inclusion/metal and slag/metal reactions, which may not be appropriately accounted for in current thermodynamic models. This is particularly true for the Fe–Mg–O system, which needs further evaluation at high temperatures, as it appears to be extremely important to inclusion compositions and slag/metal reactions.

7. Conclusion

The CaO–MgO–Al₂O₃ system appears to be of utmost importance for inclusions in Al-deoxidized specialty steelmaking, much as in carbon steels. The additional alloying elements in the studied steel may have marginally changed the overall thermodynamics and kinetics but did not foster a significant shift in inclusion compositions. The observed inclusions during AOD and ladle treatment contained little evidence of additional alloying elements. Modeling results supported this conclusion, with the cation components present in inclusion phases predicted to be 99% Ca, Al, and Mg.

Equilibrium and EERZ modeling, with some modifications, were useful in understanding slag/metal and inclusion/metal reactions. Equilibrium calculations showed the potential pathway from spinel inclusions in the AOD to MgO inclusions in the ladle. EERZ modeling demonstrated observed Mg levels were a result of slag/metal reactions at high temperatures during AOD reduction. Further processes will need to be incorporated to accurately model inclusion evolution.

References

• 1)   S. V.  Subramanian and  D. A. R.  Kay: Proc. Int. Symp. on the Physical Chemistry of Iron and Steelmaking, Canadian Institute of Mining and Metallurgy, Westmount, QC, (1982), 30.
• 2)   E. B.  Pretorius,  H. G.  Oltmann and  B. T.  Schart: AISTech 2013 Proc., (Pittsburgh), AIST, Pittsburgh, PA, (2013). http://digital.library.aist.org/pages/PR-364-097.htm, (accessed 2018-12-14).
• 3)   D.  Kumar: Ph.D. thesis, Carnegie Mellon University, (2018), https://doi.org/10.1184/R1/6715898.v1, (accessed 2019-07-29).
• 4)   D.  Kumar,  K. C.  Ahlborg and  P. C.  Pistorius: Metall. Mater. Trans. B, 50 (2019), 2163. https://doi.org/10.1007/s11663-019-01623-y
• 5)   C.  Liu,  X.  Gao,  S.  Ueda and  S.  Kitamura: ISIJ Int., 59 (2019), 268. https://doi.org/10.2355/isijinternational.ISIJINT-2018-584
• 6)   C.  Liu,  D.  Kumar,  B. A.  Webler and  P. C.  Pistorius: Metall. Mater. Trans. B, 51 (2020), 529. https://doi.org/10.1007/s11663-020-01774-3
• 7)   N.  Verma,  P. C.  Pistorius,  R. J.  Fruehan,  M.  Potter,  M.  Lind and  S.  Story: Metall. Mater. Trans. B, 42 (2011), 711. https://doi.org/10.1007/s11663-011-9516-3
• 8)   A.  Harada,  N.  Maruoka,  H.  Shibata and  S. Y.  Kitamura: ISIJ Int., 53 (2013), 2118. https://doi.org/10.2355/isijinternational.53.2118
• 9)   A.  Harada,  N.  Maruoka,  H.  Shibata,  M.  Zeze,  N.  Asahara,  F.  Huang and  S.  Kitamura: ISIJ Int., 54 (2014), 2569. https://doi.org/10.2355/isijinternational.54.2569
• 10)   A. L. V.  da Costa e Silva: J. Mater. Res. Technol., 7 (2018), 283. https://doi.org/10.1016/j.jmrt.2018.04.003
• 11)   J. H.  Shin,  Y.  Chung and  J. H.  Park: Metall. Mater. Trans. B, 48 (2017), 46. https://doi.org/10.1007/s11663-016-0734-6
• 12)   J. H.  Park and  Y.  Kang: Steel Res. Int., 88 (2017), 1700130. https://doi.org/10.1002/srin.201700130
• 13)   J. H.  Park and  H.  Todoroki: ISIJ Int., 50 (2010), 1333. https://doi.org/10.2355/isijinternational.50.1333
• 14)   J. H.  Park and  D. S.  Kim: Metall. Mater. Trans. B, 36 (2005), 495. https://doi.org/10.1007/s11663-005-0041-0
• 15)   J. H.  Park: Metall. Mater. Trans. B, 38 (2007), 657. https://doi.org/10.1007/s11663-007-9066-x
• 16)   H.  Todoroki and  K.  Mizuno: Iron Steelmaker, 30 (2003), 60.
• 17)   A.  Huck and  B.  Webler: AISTech 2020 Proc., (Virtual), AIST, Warrendale, PA, (2020), 1545. https://doi.org/10.33313/380/166
• 18)   G.  Okuyama,  K.  Yamaguchi,  S.  Takeuchi and  K. I.  Sorimachi: ISIJ Int., 40 (2000), 121. https://doi.org/10.2355/isijinternational.40.121
• 19)   W. A.  Krivsky: Metall. Mater. Trans., 4 (1973), 1439. https://doi.org/10.1007/BF02667991
• 20)   R. J.  Fruehan: Metall. Mater. Trans. B, 6 (1975), 573. https://doi.org/10.1007/BF02913850
• 21)   R. J.  Fruehan: The Making, Shaping, and Treating of Steel, Steelmaking and Refining Ed., AIST, Pittsburgh, PA, (1998), 136.
• 22)   V.  Visuri,  M.  Järvinen,  P.  Sulasalmi,  E.  Heikkinen,  J.  Savolainen and  T.  Fabritius: ISIJ Int., 53 (2013), 603. https://doi.org/10.2355/isijinternational.53.603
• 23)   V.  Visuri: Ph.D. thesis, University of Oulu, (2017), http://jultika.oulu.fi/files/isbn9789526216713.pdf, (accessed 2021-04-28).
• 24)   J. H.  Wei and  D. P.  Zhu: Metall. Mater. Trans. B, 33 (2002), 111. https://doi.org/10.1007/s11663-002-0091-5
• 25)   J. H.  Wei and  H. Y.  Zuo: Steel Res. Int., 78 (2007), 863. https://doi.org/10.1002/srin.200706299
• 26)   V.  Visuri,  M.  Järvinen,  J.  Savolainen,  P.  Sulasalmi,  E.  Heikkinen and  T.  Fabritius: ISIJ Int., 53 (2013), 613. https://doi.org/10.2355/isijinternational.53.613
• 27)   D.  Tang and  P. C.  Pistorius: Metall. Mater. Trans. B, 52 (2021), 580. https://doi.org/10.1007/s11663-021-02084-y
• 28)   C.  Liu,  M.  Yagi,  X.  Gao,  S.-J.  Kim,  F.  Huang,  S.  Ueda and  S.-y.  Kitamura: Metall. Mater. Trans. B, 49 (2018), 113. https://doi.org/10.1007/s11663-017-1122-6
• 29)   B. A.  Webler and  P. C.  Pistorius: Metall. Mater. Trans. B, 51 (2020), 2437. https://doi.org/10.1007/s11663-020-01949-y
• 30)   L. A.  Frank: Steelmaking Conf. Proc., (Baltimore), Iron and Steel Society, Warrendale, PA, (2001), 403.
• 31)   E. B.  Pretorius,  H. G.  Oltmann and  T.  Cash: Iron Steel Technol., 7 (2010), 31.
• 32)   C.  Dannert,  H.  Köchner,  U.  Strack-Thor,  P.  Valentin,  C.  Bruch,  D.  Sichen,  L.  Holappa,  S.  Louhenkilpi and  P.  Väyrynen: Report: Improvement of ladle stirring to minimise slag emulsification and reoxidation during alloying and rinsing (STIMPROVE), (2010), https://op.europa.eu/en/publication-detail/-/publication/b88c484e-ded1-4719-a0f7-93c1597ab988, (accessed 2021-03-02).
• 33)   H.  Mu,  T.  Zhang,  R. J.  Fruehan and  B. A.  Webler: Metall. Mater. Trans. B, 49 (2018), 1665. https://doi.org/10.1007/s11663-018-1294-8
• 34)   K.  Sakata: ISIJ Int., 46 (2006), 1795. https://doi.org/10.2355/isijinternational.46.1795
• 35)   M. A.  Van Ende and  I. H.  Jung: Metall. Mater. Trans. B, 48 (2017), 28. https://doi.org/10.1007/s11663-016-0698-6
• 36)   S. P. T.  Piva,  D.  Kumar and  P. C.  Pistorius: Metall. Mater. Trans. B, 48 (2017), 37. https://doi.org/10.1007/s11663-016-0764-0
• 37)   C. W.  Bale,  P.  Chartrand,  S. A.  Degterov,  G.  Eriksson,  K.  Hack,  R.  Ben Mahfoud,  J.  Melançon,  A. D.  Pelton and  S.  Petersen: Calphad, 26 (2002), 189. https://doi.org/10.1016/S0364-5916(02)00035-4
• 38)   I. H.  Jung,  S. A.  Decterov and  A. D.  Pelton: Metall. Mater. Trans. B, 35 (2004), 493. https://doi.org/10.1007/s11663-004-0050-4
• 39)   D.  Kumar and  P. C.  Pistorius: Metall. Mater. Trans. B, 52 (2021), 163. https://doi.org/10.1007/s11663-020-02004-6
• 40)   S.  Papadopoli Tonelli Piva: Ph.D. thesis, Carnegie Mellon University, (2018), https://doi.org/10.1184/R1/7235441.v1, (accessed 2018-12-20).
• 41)   D.  Roy,  P. C.  Pistorius and  R. J.  Fruehan: Metall. Mater. Trans. B, 44 (2013), 1086. https://doi.org/10.1007/s11663-013-9813-0
• 42)   D.  Roy,  P. C.  Pistorius and  R. J.  Fruehan: Metall. Mater. Trans. B, 44 (2013), 1095. https://doi.org/10.1007/s11663-013-9888-7
• 43)   A.  Ishii,  M.  Tate,  T.  Ebisawa and  K.  Kawakami: Iron Steelmaker, 10 (1983), 35.
• 44)   J.  Tan and  B. A.  Webler: Iron Steel Technol., 14 (2017), 172.
• 45)   Q.  Han,  D.  Zhou and  C.  Xiang: Steel Res., 68 (1997), 9. https://doi.org/10.1002/srin.199701770
• 46)   J.-D.  Seo and  S.-H.  Kim: Steel Res., 71 (2000), 101. https://doi.org/10.1002/srin.200005697