Simultaneous Analysis of Soluble and Insoluble Oxygen Contents in Al-Killed Steels of Various C Contents and Supersaturation Phenomena in the Steel

Abstract

Soluble and insoluble O contents ([S. O] and [I. O], respectively) in Al-killed steels of various C contents (0 to ~2 mass pct.) were simultaneously measured using “two-stage” inert gas fusion infrared absorptiometry, where majority of the inclusions was alumina. Several steel specimens were used for the O content analysis where the specimens were either taken from a steel plant or prepared in this laboratory. In addition to this, several pretreatment methods of the specimens were tested in order to provide reliable analysis results. It was found that mechanical grinding of the specimen’s surface should be carefully carried out in order not to induce unwanted surface oxidation. This resulted in overestimation of the O content of the specimen. [S. O] and [I. O] in the steel specimens were successfully analyzed using the “two-stage” gas fusion method in the context of inert gas fusion infrared absorptiometry. Considerable portion of total O content ([T. O] = [S. O] + [I. O]) was the [S. O], which was higher than the equilibrium O content of the known Al deoxidation equilibria. Therefore, the supersaturation in the Al deoxidation was confirmed. The analyzed [I. O] was independently validated by measuring area of the exposed alumina inclusions on the polished section of the specimen using SEM. It was also found that increasing C content in the steel lowered the [S. O] while [I. O] hardly changed. It is concluded that C in the steel, mechanical stirring, and remelting, could relieve the supersaturation.

1. Introduction

To satisfy customer’s demand on steel product quality and its performance, controlling the cleanliness of the steel is essential. Among several impurities in the steel, O is probably the most difficult impurities to control for steelmakers. Deoxidation during tapping and secondary refining processes, and reoxidation during casting process result in generation of oxide inclusions. Without strict control of the inclusions, it is still far away from producing so-called clean steel. Retained inclusions in the steel after casting would cause serious troubles on deterioration of mechanical properties, crack initiation, surface degradation, and fatigue, etc.¹⁾ Clogging nozzle during the continuous casting by suspending oxide inclusions is also well recognized,²⁾ although some other mechanisms were reported.^3,4,5,6,7)

To assess the cleanliness of steel, various terms may be considered: total O content ([T. O]), soluble Al content ([S. Al]), number density of inclusions, etc. Nitrogen pick-up may be a clue for reoxidation by air entrapment.⁸⁾ All of these are related with the existence of oxide inclusions in the steel. Among these, [T. O] has been often used to assess the cleanliness due to its character of simple and rapid analysis procedure by a conventional inert gas fusion infrared analysis.⁹⁾ In many circumstances in steelmaking plants, the [T. O] has been regarded as the oxide inclusions in the steel, thereby representing insoluble O content ([I. O]). On the other hand, many laboratory scale investigations showed that there was also considerable amount of O which is chemically dissolved in liquid steel. This accounts for the soluble O content ([S. O]) out of the [T. O]. Therefore, it is desirable to distinguish both [I. O] and [S. O]. However, the current approach for the O content analysis is usually carried out to obtain [T. O], and this is regarded either [I. O] (in the practical operations) or [S. O] (in the academic investigations).

Recently, one of the present authors has proposed a method to analyze both [S. O] and [I. O] simultaneously using the inert gas fusion infrared absorptiometry.¹⁰⁾ This method can be used with commercially available equipment by modifying the heating pattern in “two-stage”, contrary to the conventional “one-stage” heating. Therefore, it was referred to as “two-stage” gas fusion analysis.¹¹⁾ It is easy and fast in the analyzing both [S. O] and [I. O] in steel specimens when majority of inclusions was alumina. Possibility to be applicable for other types of oxide inclusions were recently discussed in terms of thermodynamic considerations.¹¹⁾

In the present study, the proposed method was applied for Al-killed steels of various C contents. Previously, unexpected high [S. O] results in the Ultra-Low C (ULC) steel specimens were observed.¹⁰⁾ It was shown that the high [S. O] could be decreased by remelting the specimens, while [I. O] in the same specimens increased at the same time. It was thought that this phenomenon might be due to a presence of supersaturation. This phenomenon was revisited in the present study. Moreover, validity of the analysis method was confirmed by two independent trials: one validation for [T. O] by comparison between conventional method and the proposed method and the other validation for [I. O] by comparison between cross-sectional inclusion analysis and the proposed method. A reason of the high [S. O] was accounted for by supersaturation phenomena in Al deoxidation.^{10,12,13,14,15,16)} Proper pretreatment of steel specimen before the gas fusion analysis was also supplemented.

2. Principles of Oxygen Analysis in Steel Sample

2.1. Conventional Inert Gas Fusion Infrared Absorption Method

Currently accepted method for [T. O] in steel specimen is 1) joule heating of a steel specimen in a graphite crucible in a gas fusion chamber, 2) generating CO (and possibly CO₂) gas by carbothermic reduction, which is carried by a carrier gas (typically He gas), 3) measuring an extent of absorption of infrared ray’s intensity by NDIR (Non Dispersive InfraRed) sensor, 4) converting the extent to the amount of O in the specimen, and 5) reporting the [T. O] in the specimen.⁹⁾ In the case of Al-killed steel, it is safe to assume that majority of the oxide inclusion is alumina. Therefore, following two reactions are considered:   

$$\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3\left(\mathrm{s}\right)+3\underset{\_}{\mathrm{C}}=3\mathrm{C}\mathrm{O}\left(\mathrm{g}\right)+2\underset{\_}{\mathrm{A}\mathrm{l}}$$(1)

  

$$\underset{\_}{\mathrm{O}}+\underset{\_}{\mathrm{C}}=\mathrm{C}\mathrm{O}\left(\mathrm{g}\right)$$(2)

where O and C are the soluble elements already in the liquid steel. C can be additionally dissolved from the graphite crucible. These two reactions occur simultaneously at a high temperature in a “one-stage”. The evolved CO gas is detected for its composition in the carrier gas, and is converted to the O mass in the steel specimen along with proper calibration. Therefore, O content in the specimen is obtained. It is obvious that the “one-stage” heating method only provides [T. O].

2.2. Two-Stage Heating to Separate Oxygen Signal

Hong and Kang proposed to modify the carbothermic heating in two-stage: one at low temperature where only the Reaction (2) takes place, and the other at high temperature where the Reaction (1) takes place subsequently.¹⁰⁾ In this way, the O in the oxide inclusion (alumina in this case) and the other O dissolved in the steel can be separated in the course of the carbothermic reaction. This generates two-peak CO evolution intensity vs the extraction time. Integrating area under each peak yields [S. O] and [I. O] respectively.¹¹⁾ Therefore, sum of these two finally gives [T. O] (= [S. O] + [I. O]). Its validity was previously confirmed by independently measuring [T. O] and [I. O], respectively.¹⁰⁾ In addition to this, thermodynamic basis of this “two-stage” method was assessed by the present authors.¹¹⁾ It provides isothermal carbothermic reduction temperature for each oxide inclusion, which was found to be dependent on steel composition and the gas fusion chamber pressure. In the previous case¹⁰⁾ and the present study which will be shown later, the Al-killed steel containing Al in a few hundred ppm levels contains alumina inclusion. For example, when [pct Al] = 0.04, the alumina inclusion is reduced by the carbothermic reaction (Reaction (1)) at a temperature above a range of 1518°C to 1565°C.¹¹⁾ Therefore, the first heating at the “low” temperature should be carried out at a temperature lower than this temperature range, but also should not be carried out too low temperature in order to proceed the reaction rapidly. The latter was also confirmed by separated experimental investigations.^11,17) For typical Al-killed steels where [pct Al] varies 0.03 to 0.04, the proposed temperatures for the two reactions (1530°C for the Reaction (1) and 2050°C for the Reaction (2)) in the previous study¹⁰⁾ were reasonable.

3. Method and Materials

Detailed procedure was reported previously.¹⁰⁾ Those only relevant to the present study are mentioned in this section.

3.1. Sample Preparation

Contrary to the previous study,¹⁰⁾ both plant steel specimens and in-house specimens were employed in the present study. Detailed information of the specimens is listed in Table 1. One of the plant specimens (TD1) was ULC grade steel, and the other (TD2) was a LC grade steel. Both were sampled in tundish processes using lollipop samplers.

Table 1. Steel specimens employed in the present study and their composition analysis results. C contents of TD1, TD2 were as received from POSCO. Those of the others (AK1 to AK4) and Al contents of all the specimens were analyzed in the present study. [S. O], [I. O], and [T. O] were analyzed in the present study using “two-stage” gas fusion analysis. Condition of SEM analysis for counting I. O from inclusions are also listed. “–” means negligible (not added).

Sample No.	Composition			Two-stage heating			SEM analysis					Note
	C	S. Al	I. Al	[S. O]	[I. O]	[T. O]		[I. O] (ppm)		Aobs	dmin
	(pct)			(ppm)			# of inc.	MD	AF	(mm2)	(μm)
TD1	0.002	0.034	0.0028	20.0 ± 2.0	2.4 ± 2.4	22.4 ± 2.1	41	3.74	13.22	0.825	0.23	0.06Si-0.14Mn-0.05P-0.006S-0.001Ti
TD2	0.033	0.017	0.0018	17.5 ± 1.3	1.5 ± 1.7	19.0 ± 2.6	98	10.74	13.47	0.443		0.01Si-0.23Mn-0.01P-0.006S-0.001Ti
AK1	–	0.038	0.0026	18.0 ± 5.2	5.6 ± 3.4	25.5 ± 5.2	155	7.98	10.56	1.170		–
AK2	0.133	0.047	0.0022	12.3 ± 0.3	3.1 ± 0.7	15.4 ± 0.8	65	3.36	8.60	0.888		–
AK3	0.398	0.048	0.0031	11.2 ± 1.5	3.5 ± 1.1	14.7 ± 2.5	76	2.98	12.30	0.885		–
AK4	1.830	0.050	0.0025	7.8 ± 0.7	2.6 ± 0.4	10.4 ± 0.9	77	4.53	5.82	1.033		–
AK5	–	0.036	0.0060	37.9 ± 2.8	5.9 ± 0.5	43.8 ± 2.6	N. A.					Before stirring
AK5’	–	0.032	0.0040	18.3 ± 2.4	4.6 ± 1.5	22.9 ± 3.6						After stirring
AK1’	N. A.			15.2 ± 4.4	15.3 ± 2.9	30.5 ± 5.9						After remelting
AK3’				3.0 ± 1.2	3.1 ± 0.6	6.1 ± 1.7

Other specimens were prepared in the present authors’ laboratory. An alumina crucible (60 × 55 × 100 mm) with 500 grams of electrolytic iron (typically contains impurity O as much as ~80 ppm) was charged into a high-frequency induction furnace (Insung, maximum power 30 kW). Subsequently, it was heated up to 1600°C to melt the electrolytic iron. Ar gas (> 99.999 pct.) was purified by Drierite and heated Cu and Mg chips at 500°C to remove moisture and residual O. It flowed into the induction furnace chamber. In the case of C-bearing specimen, appropriate amount of graphite powder was added. After the electrolytic iron was completely melted, Al shots enveloped by an iron foil were added. The Al-killed liquid steel was homogenized for 30 minutes. Then, the liquid steel was sampled by using a quartz tube (6 × 4 × 60 mm) and the specimen was quenched in a water bath at an approximated cooling rate of 100°C sec⁻¹ to 200°C sec⁻¹. These are referred to as AK1 to AK4 in the present study (see Table 1). The C contents in the AK 2 to AK4 specimens were analyzed using LECO CS-844 (St. Joseph, MI, USA).

Additional specimens were prepared to produce almost inclusion-free steel specimens. This was intended to have steel specimens of nearly homogeneous in O content, without interrupted by the presence of oxide inclusions. Almost similar procedure was employed except for the use of 55 pct CaO – 45 pct Al₂O₃ synthetic flux placed on top of the liquid steel, which was contained in a smaller alumina crucible (25 × 21 × 60 mm). The flux rapidly melted to absorb floating alumina inclusions. After the 30 minutes, the crucible with the liquid steel was quenched by the water. This is referred to as AKS in the present study (see Table 2).

Table 2. Steel specimens prepared with the synthetic flux (55CaO – 45Al₂O₃).

Sample No.	Composition	Case No.		Two-stage heating
	S. Al			[S. O]	[I. O]	[T. O]
	(pct)			(ppm)
AKS	0.063	1	No pretreatment	10.0 ± 2.0	3.5 ± 1.5	13.5 ± 3.4
		2	Surface grinding under air	17.4 ± 7.4	3.2 ± 4.2	20.6 ± 3.2
		3	Surface grinding under Ar shield gas	14.7 ± 1.7	8.1 ± 0.3	22.8 ± 1.4
		4	Electrolytic polishing	8.8 ± 0.5	2.0 ± 0.5	10.8 ± 1.0
		5	Surface grinding in ethanol bath	9.5 ± 0.4	1.5 ± 0.3	11.1 ± 0.7

3.2. Pretreatment of the Specimens

All the specimens were machined to be suitable for the subsequent gas fusion analysis. Those were cut into several pieces of rectangular shape (TD1, TD2, AKS) or of rod shape (AK1 to AK4). Before those were used for the gas fusion analysis, the following five pretreatment tests were carried out for the AKS specimens:

• Case 1: no special pretreatment was done.

• Case 2: surface of the specimen was ground by a motor-driven hand grinder to peel off any surface scale, under air as usual.

• Case 3: the same as the Case 2, except for the use of Ar gas to prevent the surface oxidation by air. The specimen was shielded by a flow of Ar gas injected through a stainless tube (ID 4 mm) during the grinding.

• Case 4: the specimen was polished in an electrolytic bath of 90 pct acetic acid + 10 pct perchloric acid with a Pt mesh electrode for 60 sec under 1 A.

• Case 5: the same as the Case 2, except for the use of ethanol bath to prevent the surface oxidation by air.

After each pretreatment, any dust and metallic fragment were removed in an ultrasonic bath of ethanol. The specimens were then dried and analyzed by inert gas fusion infrared absorptiometry. In the present study, LECO ON-836 (St. Joseph, MI, USA) were used.

3.3. Carbothermic Extraction of Oxygen

The equipment (LECO ON-836) was calibrated for O content and temperature. Detailed procedure can be found elsewhere.¹⁰⁾ In the present study, O content was calibrated using certified standard samples, supplied by LECO. Temperature was calibrated by melting temperature of Cu, Ni, and Ti at C saturation. The calibrated results are shown in Fig. 1. The temperature inside the gas fusion chamber of LECO ON-836 was controlled either by heating power or could be controlled directly in a recent updated software. For the alumina inclusion, 1530°C and 2050°C were chosen to extract soluble O and insoluble O in the Al-killed steel specimen, as was reported in the previous study.^10,11) These are marked by closed symbols in the figure.

Fig. 1.

Relationship between the heating power of LECO ON-836 and melting point of Cu, Ni, and Ti, all saturated by C. Two target temperatures for the two reactions (Reaction (1) and Reaction (2)) were marked by red solid circles. (Online version in color.)

3.4. Procedures to Validate the Analysis

Validation of the present analysis results for [I. O] was carried out in two independent ways. As was done in the previous study,¹⁰⁾ the analyzed [I. O] should be proportional to the amount of oxide inclusions. The number density of the inclusions (N_A) on a polished surface of the steel specimen was measured using a Field-Emission Scanning Electron Microscope (FE-SEM, JEOL-7100F, JEOL, Tokyo, Japan) equipped with an Energy Dispersive Spectrometer (EDS) and automated feature analysis (detector: Oxford X-Max 50, software: AZtec 3.1 Oxford Instrument, Oxford, UK). 15 kV, WD 10 mm, 2 sec/features over 20000 signal/sec were employed. Detailed procedure to convert the analyzed N_A to [I. O] was given elsewhere^10,18) As was mentioned, majority of the oxide inclusions was found out to be alumina. Detailed evidence is given in Appendix 1. In addition to the SEM analysis, insoluble Al content ([I. Al]) was obtained to account for the [I. O] assuming all the insoluble Al was due to the insoluble O. The [I. Al] was obtained by subtracting soluble Al content ([S. Al]) from total Al content ([T. Al]). Both [S. Al] and [T. Al] were analyzed by Inductively Coupled Plasma - Atomic Emission Spectroscopy (Thermo-Fisher Scientific ICAP 6500, Waltham, MA, USA). Detailed analysis procedure for the [T. Al] and the [S. Al] can be found elsewhere.¹⁹⁾

4. Results

4.1. Optimization of the Pretreatment Procedure

Figure 2 shows the analyzed [S. O] and [I. O] of the AKS specimen treated by five different methods as described in Sec. 3.2. [T. O] was also shown, which was the sum of [S. O] and [I. O]. Since these specimens were prepared with the intention of removing alumina inclusions in the liquid steel, it was expected to have very low [I. O]. Except for the Case 3, the analyzed [I. O] was 1.5 (± 0.3) to 3.5 (± 1.5) ppm. It is seen that the amount of inclusion was low, and its level was consistent each other. Case 3 may be considered as an outlier. The [S. O]s of the Cases 1, 4, and 5 were relatively consistent each other (10.0 ± 2.0, 8.8 ± 0.5, and 9.5 ± 0.4, respectively). On the other hand, those in the Cases 2 and 3 were somewhat higher (17.4 ± 7.4 and 14.7 ± 1.7, respectively). Moreover, those two cases also showed higher uncertainty, compared to other cases. It is likely that the specimen treated in the Case 2 might have been oxidized during the pretreatment. It was most likely surface oxidation, which might generate Fe oxide on its surface. Therefore, its content could be included in the [S. O] during the two-stage gas fusion analysis.²⁰⁾ The Case 3 also showed slightly higher [S. O]. Griding even under the Ar shield gas seems not enough to prevent the surface oxidation. It is to be noted that these two cases showed [S. O] even higher than that of the Case 1 which was not treated. Indeed, the specimen was cut from the quenched ingot by a cutting wheel with cooling water (Case 1). But the two specimens (Case 2 and Case 3) were treated additionally with the hand grinder which generated considerable amount of tribological heat. This could be an energy of the surface oxidation. In order not to cause the surface oxidation, the electrolytic polishing (Case 4) and the surface griding in cold ethanol bath (Case 5) were subsequently carried out. Both tests show consistent and reliable results. In view of practical convenience, further analysis was carried out with the surface griding in cold ethanol bath, which is thought to effectively cool the specimen (Case 5).

Fig. 2.

O contents analyzed by the “two-stage” gas fusion analysis for the AKS specimen by the five different pretreatment methods. (Online version in color.)

4.2. Types of Oxygen in Al-Killed Steels of Various C Contents

Figure 3 shows the analyzed O content in the specimens with their C content ([pct C]). Those are all Al-killed steels and their [S. Al]s were not very different: TD2 has the lowest [S. Al] (= 0.017 pct) and AK4 has the highest [S. Al] (= 0.050 pct). [T. O] of TD1 and that of AK1, both have very low C contents and similar Al contents, are similar each other. Both specimens show that major portion of [T. O] was [S. O]. It means that O in the alumina inclusion in the specimens is only a part of the [T. O].

Fig. 3.

O contents analyzed by the “two-stage” gas fusion analysis for the plant specimens (TD1 and TD2) and for the in-house specimens (AK1 to AK4). [S. Al] of the specimens vary from 0.017 to 0.050 (see Table 1). (Online version in color.)

Increasing [pct C] gradually lowered the [T. O]. To be strict, [S. O] decreased while [I. O] did not change noticeably. Both the plant specimens and in-house specimens show consistent trend that 1) major portion of [T. O] was [S. O] and 2) increasing [pct C] lowered [S. O]. A reasonable explanation will be given in Sec. 5.1.

In the previous study,¹⁰⁾ the analyzed O contents were validated using conventional “one-stage” gas fusion analysis for the [T. O] and SEM analysis for the [I. O]. It was shown that the analyzed O contents using “two-stage” gas fusion analysis were in good agreement in the validation. In the present study, the analyzed [I. O] were again validated by using SEM, as was described in Sec. 3.4. [I. O] can be estimated from the SEM analysis by measuring inclusion size on the polished surface of the steel specimen:   

$$\left[\mathrm{I}.\mathrm{\hspace{0.17em} }\mathrm{O}\right]\left(\mathrm{p}\mathrm{p}\mathrm{m}\right)=f_V\times \frac{{\rho }_{\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3}}{{\rho }_{\mathrm{F}\mathrm{e}}}\times \frac{3{\mathrm{M}}_{\mathrm{O}}}{2{\mathrm{M}}_{\mathrm{A}\mathrm{l}}+3{\mathrm{M}}_{\mathrm{O}}}\times {10}^6$$(3)

where f_V, ρ_Fe, ρ_Al2O3, M_O, and M_Al are the volume fraction of inclusions, the density of Fe (7860 kg m⁻³), the density of alumina (3950 kg m⁻³), the atomic weight of O (0.0159 kg mol⁻¹), and that of Al (0.0270 kg mol⁻¹), respectively. Here, f_V were estimated by Mean Diameter (MD) method, which was shown to be more reliable than Area Fraction (AF) method.^{10,18,21,22,23)} Details can be found in Appendix 2. The comparison with the MD method is shown in Fig. 4. The calculated [I. O]s by the MD method of four specimens ([pct C] = 0.002, 0.133, 0.398, 1.83) were in favorable agreement with the measured [I. O]s using the two-stage gas fusion analysis, while those of two other specimens ([pct C] = 0.0, 0.033) were higher than the analyzed [I. O]s by the two-stage gas fusion analysis. Except for these two specimens, the [I. O] measured by the two-stage combustion method is thought to be reliable.

Fig. 4.

Comparison between [I. O] of the specimens (TD1, TD2, AK1 to AK4) measured by two different methods: an estimation from volume fraction of inclusions (MD) and the two-stage gas fusion analysis method. Numbers near the square symbols are [pct C] of each specimen. (Online version in color.)

In addition to this, [I. Al] was also used to validate the present results. [I. O] can be estimated from the [I. Al] as:   

$$\left[\mathrm{I}.\mathrm{\hspace{0.17em} }\mathrm{O}\right]\left(\mathrm{p}\mathrm{p}\mathrm{m}\right)=\frac{3{\mathrm{M}}_{\mathrm{O}}}{2{\mathrm{M}}_{\mathrm{A}\mathrm{l}}}\times \left[\mathrm{I}.\mathrm{\hspace{0.17em} }\mathrm{A}\mathrm{l}\right]\left(\mathrm{p}\mathrm{p}\mathrm{m}\right)$$(4)

The estimated [I. O] was compared with that obtained by the two-stage gas fusion analysis in Fig. 5. The estimated [I. O] from the analyzed [I. Al] is significantly higher than that of two-stage gas fusion analysis for all the specimens. Indeed, the estimated [I. O] values by the Eq. (4) in some specimens (TD1, AK2, AK3, AK4) were even higher than their [T. O] values. It is thought that the ICP analysis for [T. Al] and [S. Al] to give the [I. Al] by their difference was not an adequate method in the present case. Therefore, it is concluded that the [I. O] analyzed by the two-stage gas fusion analysis was reliable by the validation using SEM analysis, but further refinement on the ICP analysis should be envisaged.

Fig. 5.

Comparison between [I. O] of the specimens (TD1, TD2, AK1 to AK4) measured by two different methods: other estimations from [I. Al] analyzed by ICP-AES and the two-stage gas fusion analysis method. (Online version in color.)

The estimated [I. O] by the SEM shown in Fig. 4 does not show meaningful relationship with [pct C], marked near each symbol. Figure 3 also showed that the analyzed [I. O] did not show a dependence on [pct C]. Therefore, it is further concluded that the increase of [pct C] is not relevant to the direct inclusion removal, but is responsible for the decrease of soluble O in the liquid steel.

5. Discussions

5.1. Supersaturation in Al Deoxidation in Liquid Steel

Figure 6 shows the calculated Al deoxidation equilibria in liquid steel at 1600°C by thick lines. The calculations were carried out by FactSage thermochemical software^24,25) with FSstel database where the Modified Quasichemical Model was adopted for the deoxidation equilibria developed by one of the present authors.^26,27) In order to show the effect of C on the deoxidation equilibria, the deoxidation equilibria were calculated at [pct C] = 0 (solid line) and 1.83 (dashed line), respectively. The measured data ([S. Al] vs [S. O]) were plotted together by symbols. It should be stressed that, in principle, the deoxidation equilibria must be compared with soluble contents of solutes. Most of the previous deoxidation investigations employed [T. Al] and [T. O] assuming that those represent [S. Al] and [S. O].^12,28,29,30) In the present study, true [S. Al] and [S. O] were used to compare those with the model calculation. As seen in the figure, all the data show higher O content than that in equilibrium with alumina, regardless of the [pct C]. Since the experimental data represent true [S. O], the discrepancy cannot be accounted for by entrapment of any suspending alumina inclusions. This should be seen as the supersaturation in the deoxidation. The supersaturation of Al-killed steel was first reported by Elliott and co-worker,^31,32) and was extensively investigated by Japanese scholars.^{12,13,14,15,16,33)} Suito et al. observed significant extent of the supersaturation in Al deoxidized liquid Fe, and they found that mechanical stirring could resolve the supersaturation.¹²⁾ In their work, they eliminated suspending alumina inclusion as much as possible to regard analyzed [T. O] as the [S. O]. According to the classical nucleation theory,³¹⁾ a nucleus may overcome the nucleation activation energy by absorbing an energy (heating or mechanical stirring) or having heterogeneous interface.

Fig. 6.

Relationship between the analyzed [S. O] and [S. Al] of the specimens (TD1, TD2, AK1 to AK4), showing considerably higher [S. O] than that of the equilibrium O content shown by a full line ([pct C] = 0) and a dashed line ([pct C] = 1.83).^24,25,26) Thin vertical lines represent differences between the analyzed [S. O] ([S. O]_ss) and the equilibrium [S. O] ([S. O]_eq), respectively. (Online version in color.)

The supersaturation ratio S with respect to alumina is defined as:¹²⁾   

$$S=\frac{{\left(a_{\underset{\_}{\mathrm{A}\mathrm{l}}}^2{a}_{\underset{\_}{\mathrm{O}}}^3\right)}_{\mathrm{s}\mathrm{s}}}{{\left(a_{\underset{\_}{\mathrm{A}\mathrm{l}}}^2{a}_{\underset{\_}{\mathrm{O}}}^3\right)}_{\mathrm{e}\mathrm{q}}}$$(5)

where “ss” and “eq” refer to “supersaturation” and “equilibrium”, respectively. From Fig. 6, the analyzed [S. O] can be regarded as [S. O] in the supersaturation, and the calculated [S. O] on the line can be regarded as [S. O] at the equilibrium. Assuming the activity coefficient of O, f_O, is not very different between the two [S. O] values at the same [S. Al] where (a_Al)_ss = (a_Al)_eq, the S can be approximated as:   

$$S=\frac{{\left(a_{\underset{\_}{\mathrm{A}\mathrm{l}}}^2\right)}_{\mathrm{s}\mathrm{s}}{\left(f_{\underset{\_}{\mathrm{O}}}\times \left[\mathrm{S}.\mathrm{\hspace{0.17em} } \mathrm{O}\right]\right)}_{\mathrm{s}\mathrm{s}}^3}{{\left(a_{\underset{\_}{\mathrm{A}\mathrm{l}}}^2\right)}_{\mathrm{e}\mathrm{q}}{\left(f_{\underset{\_}{\mathrm{O}}}\times \left[\mathrm{S}.\mathrm{\hspace{0.17em} } \mathrm{O}\right]\right)}_{\mathrm{e}\mathrm{q}}^3}\approx {\left(\frac{{\left[\mathrm{S}.\mathrm{\hspace{0.17em} } \mathrm{O}\right]}_{\mathrm{s}\mathrm{s}}}{{\left[\mathrm{S}.\mathrm{\hspace{0.17em} } \mathrm{O}\right]}_{\mathrm{e}\mathrm{q}}}\right)}^3$$(6)

Lower limit of thin vertical lines in Fig. 6 means the equilibrium [S. O] ([S. O]_eq) calculated for each sample at its [pct C] and [pct S. Al]. Therefore, the length of the vertical lines is proportional to the S.

The approximated S by the Eq. (6) for the present specimens is shown in Fig. 7, along with the [S. Al] at each [pct C]. It is clearly seen that the S gradually decreased as [pct C] increased, while [S. Al] was almost the same. It should be stressed that both the plant specimens and the in-house specimens were in the supersaturation in Al deoxidation. Suito et al. reported a range of S of Fe–Al alloys of various Al contents.¹²⁾ Those relevant to the present study (similar [S. Al]) are shown by a vertical line in the figure. The shaded area in Fig. 7 was drawn to guide the eye, based on the [S. O]s of the four AK specimens and that of one TD specimen (TD1 of lower [pct C]). This is because 1) [S. Al] in the five specimens (TD1 and AK1 to AK4) were in a similar range (−1.30 to −1.49 in a logarithmic scale) compared to that in the other specimen (TD2, log [S. Al] = −1.77), and 2) these five specimens showed a monotonous trend of decreasing log S as [pct C] increases. log S of the other specimen (TD2) was a bit off from the trend (the shaded area): it was due to lower [S. Al] which resulted in a higher equilibrium [S. O]_eq, as seen in Fig. 6.

Fig. 7.

Supersaturation ratio S and [S. Al] of the specimens (TD1, TD2, AK1 to AK4) plotted for the [pct C] of each specimen. The S reported by Suito et al.¹²⁾ for some their specimens ([pct C] = 0 and the similar [S. Al] as that of the present study) is shown by a vertical line. The S in the present study over wide [pct C] is marked by the shaded area. (Online version in color.)

5.2. Relief of the Supersaturation by Mechanical Stirring and by Remelting

To confirm and to relieve the supersaturation, two trials were attempted in the present study. One was executing the mechanical stirring in the supersaturated melt and the other was remelting of the supersaturated melt with wider contact area to alumina wall, as was done in the previous study.¹⁰⁾

An Al-killed steel with no C was prepared as was described in Sec. 3.1 without the flux. After homogenizing the liquid steel, some portion of it was sampled using the quartz tube. Subsequently, the liquid steel was stirred mechanically by an alumina rod for 10 minutes, and was sampled similarly. Compositions of the specimens before (AK5) and after the stirring (AK5’) were analyzed as described earlier, and are listed in Table 1. Figure 8 shows the analyzed O content in the steel specimen before (left) and after (right) the mechanical stirring. [T. O] decreased after the stirring. Most of the decrease was due to the decrease of [S. O]. This tells that the supersaturated O and Al reacted to form alumina and floated up during the mechanical stirring. Therefore, [S. O] decreased while [I. O] did not vary much, and the supersaturation was partly relieved (still above the equilibrium in view of the [S. O]).

Fig. 8.

Change of the O contents of the specimen before the stirring (AK5) and after the stirring (AK5’). (Online version in color.)

The other trial was remelting the supersaturated specimens (AK1 and AK3) in small size alumina crucible (9.4 × 9.0 × 5 mm). This was intended to facilitate heterogeneous nucleation of alumina, if there was the supersaturation in the deoxidation. Approximately 0.3 g of AK1 and AK3 specimens, which were already shown to be supersaturated, were charged in the alumina crucible, respectively. Those were remelted in a halogen lamp heated gold-image furnace. Ar gas purified as described in Sec. 3.1 was used to protect the specimen during the melting. In addition to this, a Ti foil was placed near the specimen to absorb any trace of oxygen in the Ar gas. The specimen was heated to 1600°C at a rate of 150°C min⁻¹, was kept for 3 minutes, and rapidly cooled down to room temperature (~−300°C min⁻¹). The remelted specimens (AK1’ and AK3’) were analyzed for their O contents, and the results are shown in Fig. 9 (also see Table 1). Both specimens showed the [S. O] decreased. In the case of the AK1 specimen (ULC steel), the [I. O] increased, in consistent with the previous study.¹⁰⁾ On the other hand, the analyzed [I. O] in AK3 specimen (mid-C steel) did not change much after the remelting. However, it is seen that [T. O] decreased as much as the [S. O] decreased. Therefore, it was postulated that the supersaturation was relieved, but the formed alumina might be lost during collecting the specimen after the remelting. In the case of the AK1’, the specimen seemed slightly oxidized, generating more alumina than the decrease of the [S. O]. However, the analyzed [T. O]s before and after the remelting were within analytical uncertainty. For the specimen with higher [pct C], the relief of the supersaturation was evident.

Fig. 9.

Change of the O contents of the specimen before the remelting (AK1 and AK3) and after the remelting (AK1’ and AK3’). (Online version in color.)

From these two tests, it can be additionally concluded that the Al-killed C steel specimens employed in the present study were supersaturated. The supersaturation could be relieved again by either the mechanical stirring or the heterogeneous nucleation by enlarging contact area between the liquid steel and the alumina. This conclusion could be seamlessly drawn thanks to the developed analysis method for the soluble/insoluble O in steel specimen in the present study.

5.3. Supersaturation in Al-killed Steel and Role of C in the Steel

Nozzle clogging and defect in cast slab were more frequently reported in ULC grade steel than mid-C or high C grade steel. In the study of Fukuda et al.,³⁴⁾ the thickness of alumina buildup on the surface of alumina/graphite refractory nozzle was very thin for 4 pct C containing Al-killed steel, while it was considerably thick for 0.2 pct C containing Al-killed steel. The O source of the alumina build up was thought to be silica in the nozzle refractory, which underwent a carbothermic reaction.^2,3,4,5,6,7) This implies that high C in the liquid steel suppresses the supersaturation, and low C circumstances in liquid steel is prone to the supersaturation. Therefore, the supersaturation in relatively low C Al deoxidation was relieved at the interface between the nozzle and the liquid steel which was supersaturated.

In this section, the role of C on the relief of the supersaturation is discussed. According to the classical homogeneous nucleation theory,³¹⁾ the critical supersaturation ratio S^* is given as:   

$$\mathrm{R}T\mathrm{\hspace{0.17em} } ln{S}^{*}=-{\left(\frac{16\pi }{3}\right)}^{\frac{1}{2}}{\left({\sigma }_{\mathrm{F}\mathrm{e}-\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3}\right)}^{\frac{3}{2}}{v}_m{\left\{\mathrm{k}Tln\frac{I}{\mathrm{A}}\right\}}^{-\frac{1}{2}}$$(7)

where R and k have the usual meaning, σ_Fe-Al2O3, v_m, I and A are the interfacial energy between liquid iron and alumina (J m⁻²), the molar volume of alumina (m³ mol⁻¹), the nucleation rate (m⁻³ s⁻¹), and the constant depending on the complexity of the molecular species forming and on the character of the means by which atoms and molecules are transported to and from the site where the new interface is form.³²⁾ When this equation is applied to liquid steel with various C contents, all other terms are very similar except for the σ_Fe-Al2O3. Therefore, the following relation is derived:¹⁴⁾   

$$log{S}^{*}={\left(\frac{{\sigma }_{\mathrm{F}\mathrm{e}-\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3}}{\sigma {°}_{\mathrm{F}\mathrm{e}-\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3}}\right)}^{\frac{3}{2}}logS{°}^{,*}$$(8)

where with and without the superscript “°” mean those in liquid steel without C and with C, respectively. This relationship provides the effect of the σ_Fe-Al2O3 between liquid steel and alumina on the degree of S^*. The σ_Fe-Al2O3 was estimated using Young’s equation:   

$${\sigma }_{\mathrm{F}\mathrm{e}-\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3}={\sigma }_{\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3}-{\sigma }_{\mathrm{F}\mathrm{e}}\times cos\theta$$(9)

where σ_Al2O3, σ_Fe, and θ are the surface energy of alumina (J m⁻²), that of liquid Fe (J m⁻²), and the contact angle (°), respectively. σ_Al2O3 is reported to be 0.75 J m⁻².³⁵⁾ σ_Fe at 1600°C was 1.834 J m⁻².³⁶⁾ Two experimental results for the θ at 1550 to 1570°C were found,^37,38) and these two were considered to evaluate the σ_Fe-Al2O3. The obtained σ_Fe-Al2O3 was inserted in the Eq. (8) to get the S^*/S^*,° as a function of [pct C]. The calculated S^*/S^*,° was shown in Fig. 10. The S^* was seen to decrease by the increase of [pct C]. It depends on the reported θ data significantly in the two independent studies.^37,38) The decreasing tendency of S^* is consistent with that of S shown in Fig. 7. This is also consistent with the report by Lee and Suito¹⁵⁾ that supersaturation in their Fe–Al alloy was not relieved after a reoxidation by Fe_tO in top slag, but that in the Fe–Al–1 pct C alloy under the same condition was relieved after 60 minutes.

Fig. 10.

Effect of C content on (a) the interfacial energy between alumina and liquid steel of various C content (σ_Fe-Al2O3)^37,38) and (b) the normalized critical supersaturation ratio (S^*/S^*,°). The shaded area is taken from Fig. 7. (Online version in color.)

It was clear that the Al-killed steel specimens were in the supersaturation, as was previously reported.^12,13,14,15) It was shown in the present study that the supersaturation could be relieved by either mechanical stirring or the presence of C in the liquid steel, although the supersaturation was not fully relieved in most cases. Only the AK3’ specimen showed that the supersaturation could be fully relieved where ~0.4 pct C containing Al-killed steel was remelted. Mild mechanical stirring was effective for the relief, but the supersaturation was not fully relieved under the condition. Wasai and Mukai pointed out that the supersaturation was one of the reasons for the excess O in liquid steel above the equilibrium value.¹⁶⁾ The other reason was clearly suspending alumina inclusion,¹⁶⁾ which could not be distinguished by the conventional “one-stage” gas fusion analysis. This could be successfully distinguished in the present study. By doing this, the degree of the supersaturation could be measured.

5.4. A Viewpoint to Improve Steel Cleanliness

The existence of this excess O, in the form of the soluble O, is important, because this excess soluble O will finally crystallize as alumina inclusions upon cooling.¹²⁾ Depending on the cooling rate and retained [S. O], size and amount of the crystallized inclusions will be determined. Its consequence on the steel cleanliness is to be further investigated. As was pointed out by the previous study,¹⁰⁾ the liquid steel in a large-scale vessel (such as ladle or tundish) may be in the supersaturation state. Many of the Al-killed specimens employed in the previous study¹⁰⁾ and in the present study were taken either the ladle or the tundish, and those were shown to be in the supersaturation state. Depending on the type of O in the steel, the way of improving the cleanliness of the liquid steel may be different. Contrary to the procedure to remove suspending alumina inclusions in the liquid steel, the removal of the excess soluble O should be carried out first by relieving the supersaturation. Strong mechanical stirring and increasing heterogeneous nucleation site would be thought. It will be followed by removal of the newly generated alumina inclusion which are then suspending in the liquid steel. This raises a necessity of further investigations to confirm the existence of the supersaturation in the liquid steel in a large-scale vessel, and its consequences on the steel cleanliness.

6. Conclusion

The previously proposed method to analyze soluble and insoluble O contents in steel specimens¹⁰⁾ were applied to several Al-killed steel of various C contents. “Two-stage” gas fusion analysis in the context of the inert gas fusion infrared absorptiometry was used. In addition to this, a reliable pretreatment method to prepare the steel specimens was proposed to suppress unwanted surface oxidation during surface griding of the specimens. The analyzed insoluble O content, which represents the amount of oxide inclusions (alumina in the present study), was additionally validated by SEM analysis which provided the inclusion size and the number density. Therefore, the proposed method was validated for its accuracy.

It was found that there were considerable supersaturation phenomena in Al deoxidation in the steel specimens, regardless of its origin (tundish in a steelmaking plant or the present authors’ laboratory). Degree of the supersaturation was highest when there was no C in the steel. Increasing C content decreased the supersaturation, which is in consistent with the report by Lee and Suito.¹⁵⁾ In the present study, it was experimentally confirmed for the first time. The decrease of the supersaturation was interpreted by the decrease of the critical supersaturation due to decreasing interfacial energy between liquid steel and alumina upon the increase of C content. In addition to this, execution of mechanical stirring or remelting in a smaller size alumina crucible, which increases the possibility of heterogeneous nucleation, was shown to be working to relieve the supersaturation. A necessity of further investigations to confirm the existence of the supersaturation in the liquid steel in a large-scale vessel, and its consequences on the steel cleanliness were proposed.

Acknowledgement

The authors appreciate the financial support from POSCO, Pohang, Korea.

References

• 1)   L.  Zhang and  B. G.  Thomas: ISIJ Int., 43 (2003), 271. https://doi.org/10.2355/isijinternational.43.271
• 2)   S.  Ogibayashi: Taikabutsu Overseas, 15 (1995), 3.
• 3)   J.-H.  Lee,  M.-H.  Kang,  S.-K.  Kim and  Y.-B.  Kang: ISIJ Int., 58 (2018), 1257. https://doi.org/10.2355/isijinternational.ISIJINT-2018-164
• 4)   J.-H.  Lee,  S.-K.  Kim,  M.-H.  Kang and  Y.-B.  Kang: BHM Berg-Hüttenmänn. Monatsh., 163 (2018), 18. https://doi.org/10.1007/s00501-017-0687-3
• 5)   Y.-B.  Kang and  J.-H.  Lee: Metall. Ital., (2019), No. 1, 5.
• 6)   J.-H.  Lee,  M.-H.  Kang,  S.-K.  Kim,  J.  Kim,  M.-S.  Kim and  Y.-B.  Kang: ISIJ Int., 59 (2019), 749. https://doi.org/10.2355/isijinternational.ISIJINT-2018-672
• 7)   J.-H.  Lee and  Y.-B.  Kang: ISIJ Int., 60 (2020), 426. https://doi.org/10.2355/isijinternational.ISIJINT-2019-384
• 8)   B.  Coletti,  B.  Gommers,  C.  Vercruyssen,  B.  Blanpain,  P.  Wollants and  F.  Haers: Ironmaking Steelmaking, 30 (2003), 101. https://doi.org/10.1179/030192303225001694
• 9)  ON836 Oxygen/Nitrogen Analyzer Instruction Manual, Ver. 1.9.x ed., LECO Corporation, St. Joseph, MI, (2013).
• 10)   H.-M.  Hong and  Y.-B.  Kang: ISIJ Int., 61 (2021), 2464. https://doi.org/10.2355/isijinternational.ISIJINT-2020-596
• 11)   Y.-B.  Kang,  Y.-M.  Cho and  H.-M.  Hong: Metall. Mater. Trans. B, (2022). https://doi.org/10.1007/s11663-022-02520-7
• 12)   H.  Suito,  H.  Inoue and  R.  Inoue: ISIJ Int., 31 (1991), 1381. https://doi.org/10.2355/isijinternational.31.1381
• 13)   H.  Suito: Proc. Ethem T. Turkdogan Symp., ISS, Warrendale, PA, (1994), 141.
• 14)   J.  Tanabe and  H.  Suito: Metall. Mater. Trans. B, 26 (1995), 95. https://doi.org/10.1007/BF02648982
• 15)   K. R.  Lee and  H.  Suito: Metall. Mater. Trans. B, 27 (1996), 423. https://doi.org/10.1007/BF02914907
• 16)   K.  Wasai and  K.  Mukai: Metall. Mater. Trans. B, 30 (1999), 1065. https://doi.org/10.1007/s11663-999-0112-8
• 17)   R.  Khanna,  M.  Ikram-Ul Haq,  Y.  Wang,  S.  Seetharaman and  V.  Sahajwalla: Metall. Mater. Trans. B, 42 (2011), 677. https://doi.org/10.1007/s11663-011-9520-7
• 18)   A.  Karasev and  H.  Suito: Metall. Mater. Trans. B, 30 (1999), 259. https://doi.org/10.1007/s11663-999-0055-0
• 19)   Y.-M.  Cho,  W.-Y.  Cha and  Y.-B.  Kang: Metall. Mater. Trans. B, 52 (2021), 3032. https://doi.org/10.1007/s11663-021-02220-8
• 20)   T.  Ise,  Y.  Nuri,  Y.  Kato,  T.  Ohishi and  H.  Matsunaga: ISIJ Int., 38 (1998), 1362. https://doi.org/10.2355/isijinternational.38.1362
• 21)   R. L.  Fullman: JOM, 5 (1953), 447. https://doi.org/10.1007/BF03398971
• 22)   R.  DeHoff and  F.  Rhines: Trans. Metall. Soc. AIME, 221 (1961), 975.
• 23)   J.  Hillard: Quantitative Microscopy, McGraw-Hill, New York, (1968), 45.
• 24)   C. W.  Bale,  E.  Bélisle,  P.  Chartrand,  S. A.  Decterov,  G.  Eriksson,  K.  Hack,  I.-H.  Jung,  Y.-B.  Kang,  J.  Melançon,  A. D.  Pelton,  C.  Robelin and  S.  Petersen: Calphad, 33 (2009), 295. https://doi.org/10.1016/j.calphad.2008.09.009
• 25)   C. W.  Bale,  E.  Bélisle,  P.  Chartrand,  S. A.  Decterov,  G.  Eriksson,  A. E.  Gheribi,  K.  Hack,  I.-H.  Jung,  Y.-B.  Kang,  J.  Melançon,  A. D.  Pelton,  S.  Petersen,  C.  Robelin,  J.  Sangster,  P.  Spencer and  M.-A.  Van Ende: Calphad, 54 (2016), 35. https://doi.org/10.1016/j.calphad.2016.05.002
• 26)   M.-K.  Paek,  J.-J.  Pak and  Y.-B.  Kang: Metall. Mater. Trans. B, 46 (2015), 2224. https://doi.org/10.1007/s11663-015-0369-z
• 27)   Y.-B.  Kang: ISIJ Int., 60 (2020), 2717. https://doi.org/10.2355/isijinternational.ISIJINT-2020-101
• 28)   M.-K.  Paek,  J.-M.  Jang,  Y.-B.  Kang and  J.-J.  Pak: Metall. Mater. Trans. B, 46 (2015), 1826. https://doi.org/10.1007/s11663-015-0368-0
• 29)   M.-K.  Paek,  K.-H.  Do,  Y.-B.  Kang,  I.-H.  Jung and  J.-J.  Pak: Metall. Mater. Trans. B, 47 (2016), 2837. https://doi.org/10.1007/s11663-016-0728-4
• 30)   Y.  Kang,  M.  Thunman,  D.  Sichen,  T.  Morohoshi,  K.  Mizukami and  K.  Morita: ISIJ Int., 49 (2009), 1483.
• 31)   M. L.  Turpin and  J. F.  Elliott: J. Iron Steel Inst., 204 (1966), 217.
• 32)   G.  Forward and  J. F.  Elliott: J. Met., 19 (1967), 54.
• 33)   T.  Isono,  K.  Ohnuki and  K.  Umezawa: Prof. 4th Int. Conf. on Molten Slags and Fluxes, ISIJ, Tokyo, (1992), 493.
• 34)   Y.  Fukuda,  Y.  Ueshima and  S.  Mizoguchi: ISIJ Int., 32 (1992), 164. https://doi.org/10.2355/isijinternational.32.164
• 35)   K.  Ogino,  K.  Nogi and  Y.  Koshida: Tetsu-to-Hagané, 59 (1973), 1380 (in Japanese). https://doi.org/10.2355/tetsutohagane1955.59.10_1380
• 36)   Y.  Kim,  J.  Lim,  J.  Choe and  J.  Lee: Metall. Mater. Trans. B, 45 (2014), 947. https://doi.org/10.1007/s11663-014-0033-z
• 37)   F. A.  Halden and  W. D.  Kingery: J. Phys. Chem., 59 (1955), 557. https://doi.org/10.1021/j150528a018
• 38)   I.  Jimbo and  A. W.  Cramb: ISIJ Int., 32 (1992), 26.

Appendix

Appendix 1. Analysis of Inclusion Composition Using SEM-EDS

EDS analysis was done to confirm the inclusion considered in the present study. Major elements in the inclusions were found to be Al, Fe, and O. Some other elements were rarely found. Table A1 shows the status of the inclusions analyses: number of inclusions analyzed and percentage of inclusions containing other elements than Al, Fe, and O. It is seen that almost all the inclusions were composed of Al–Fe–O. It is commonly observed that Fe intensity arises from the steel matrix during an electron probe analysis. Indeed, the present raw data of the EDS analyses often showed that Fe was the most abundant element of the analyzed results for the inclusions. This might be more severe for small-sized inclusions. However, it is unlikely that the inclusions were composed of Fe-containing oxide such as hercynite (FeAl₂O₄). This is reasonable as [T. O] in the steel specimens were very low (max. 25.5 ± 5.2 ppm). Assuming that the Fe intensity was due to the steel matrix, the inclusion compositions normalized to Fe–Al–O pseudo ternary system were shown in Fig. A1. The abscissa represents mass pct of Fe of the analyzed results, and the ordinate represents molar ratio of O/Al. As seen in the figure, regardless of the Fe content, molar ratio of O/Al is almost 1.5. Al₂O₃ corresponds to this case, while FeAl₂O₄ does not. Therefore, the EDS spectra was seen to be a mixed one between steel matrix (mostly Fe) and alumina inclusion.

Table A1. EDS analysis results for inclusions observed in the steel specimens.

Sample No.	Percentage of inclusion containing other elements than Al, Fe, O	Number of inclusions analyzed
TD1	2.4	41
TD2	1.0	98
AK1	0	155
AK2	1.5	65
AK3	1.3	76
AK4	2.6	77

Fig. A1.

Results of EDS analyses on inclusions in various steel specimens. Molar ratios of O/Al in majority of the EDS analyses are close to 1.5, representing Al₂O₃ inclusion, regardless of Fe content in the EDS results. (Online version in color.)

Appendix 2. Estimation of [I. O] from the SEM Analysis for the Inclusion Size and the Number Density

As was reported in the previous studies and was applied in the previous report by Hong and Kang,¹⁰⁾ a steel specimen was polished, and its surface was subjected to analyze all the inclusions exposed to the surface. In the present study, nearly all the inclusions were found to be alumina. For each inclusion i, the equivalent diameter d_A(i) in the 2-dimensional plane, total number of the inclusions n, the number of inclusion sections per unit area N_A were measured in a specified area (A_obs). The 2-dimensional data were translated into the 3-dimensional data (mean diameter ${\overline{d}}_V$ and the number density per unit volume N_V). This yields the volume fraction of the alumina inclusions (see Table A2), and eventually gives [I. O] as was given in the Eq. (3).

Table A2. Mean diameter (${\overline{d}}_V$), the number of inclusions per unit volume (N_V), and volume fraction of the particles (f_V) estimation using Mean diameter (MD) method and Area Fraction (AF) method.^21,22,23) n, d_A(i), N_A, and A_obs are the number of inclusions, the cross-sectional diameter of an ith inclusion, the number of inclusion sections per unit area, and a given observed area in a cross section, respectively.

Method	${\overline{d}}_V$	NV	fV
MD	$\left(\frac{\pi }{2}\right)\frac{n}{{\sum }_{i=1}^n\frac{1}{d_A\left(i\right)}}$	$\frac{N_A}{{\overline{d}}_V}$	$\frac{\pi }{6}{\overline{d}}_V{}^3{N}_V$
AF	$\left(\frac{\pi }{2}\right)\frac{n}{{\sum }_{i=1}^n\frac{1}{d_A\left(i\right)}}$	$\frac{N_A}{{\overline{d}}_V}$	$\frac{{{\displaystyle \sum }}_{i=1}^n\pi {\left(\frac{d_A}{2}\right)}^2}{A_{\mathrm{o}\mathrm{b}\mathrm{s}}}$