Formation Mechanism of Secondary Inclusions in Fe-36mass%Ni Alloy Using a Novel Combination Analysis Technique

Abstract

Controlling the size, number, and composition of secondary inclusions is vital in the production of high-quality steels. In this study, experimental and computational investigation of the relationship between secondary inclusion formation in Fe-36mass%Ni alloy and cooling rate was carried out. Assuming the case of large ingots, solidification experiments using various cooling rates (0.17 to 128 K/min) were employed and the size, number, composition, and distribution of inclusions were analyzed by SEM-EDS automatic inclusion analysis. Like previous studies, inclusion number density increased with increasing cooling rate, while inclusion size decreased with increase of cooling rate. On the contrary, oxide inclusion area fraction was found to have little relationship with the cooling rate and was instead found related with oxygen content of the sample. As a new attempt to investigate the relationship between microsegregation and secondary inclusion formation, a combination of SEM-EDS analysis and EPMA mapping analysis was carried out. By superimposing information of microsegregation and inclusions, it was found that high-Al₂O₃ inclusions formed during the early stage of solidification, whereas low-Al₂O₃ inclusions formed during the later stage of solidification. These findings suggest that Al₂O₃ inclusions formed in the early stage of solidification reacted with the remaining Si-enriched liquid steel and changed into low-Al₂O₃ inclusions. Experimental results were also confirmed by thermodynamic calculations. Present work made it possible to understand deeper the relationship between microsegregation and secondary inclusion formation.

1. Introduction

It is well known that non-metallic inclusion formation in steels significantly influence both quality of steel and steelmaking process. Non-metallic inclusions in steels are classified as either primary or secondary inclusions depending on the timing of their formation. Primary inclusions are formed during steelmaking and refining processes such as the basic oxygen furnace (BOF), ladle furnace (LF), RH degassing, and so on. Meanwhile, secondary inclusions are formed during casting process due to cooling and solidification of steel such that they inevitably remain in the final steel product. Although the presence of these inclusions is widely known to have negative impacts on steel quality such as toughness, fatigue, and corrosion resistance, recent studies suggest that secondary inclusions can be utilized to act as nucleation sites to improve mechanical properties of steels. To effectively utilize the benefits of these inclusions while minimizing their negative effects, it is important to control the size, composition, and distribution of secondary inclusions.

The following reports have suggested that both cooling rate and microsegregation play an important role in the formation of secondary inclusions. Goto et al.^1,2) initially investigated inclusion formation in Ti/Al deoxidized low carbon steels produced with cooling rates of 6 to 560 K/min using actual continuous casting slabs. They found that the number of inclusions increased with cooling rate while inclusion size decreased with increasing cooling rate. Following their investigation, the effect of cooling rate on secondary inclusion formation in various steels began to gain attention. Zhang et al.³⁾ and Luo et al.⁴⁾ experimentally studied inclusion formation with cooling rates of 2 to 600 K/min for pipeline steel and non-oriented electrical steel, respectively. Inclusion formation in 304 stainless steel was also experimentally investigated by Wang et al.⁵⁾ with cooling rates of 10 to 10980 K/min, and Suzuki et al.⁶⁾ with cooling rates of 1.1 to 150 K/min using actual steel slabs and laboratory ingots. Kim et al.⁷⁾ studied inclusion formation in Si-Mn/Ti deoxidized low carbon steels with cooling rates of 32 and 40200 K/min. Yu et al.⁸⁾ studied inclusion formation in 12%Cr steel with cooling rates of 3 to 60×10⁶ K/min using conventional casting, laser remelting, and melt spinning to obtain high cooling rate. Finally, Kikuchi et al.^9,10) studied inclusion formation in Al/Ti deoxidized low carbon high manganese steel with cooling rates of 120 to 5000 K/min by employing ingots and confocal scanning laser microscope (CSLM). In the above-mentioned studies, although steel grade and inclusion species were different, tendency of the relationship between cooling rate and inclusion size/number was the same. The cooling rates employed in these studies were from 1 K/min to very fast cooling rates since their studies simulated cooling conditions during continuous casting or welding. However, in the case of large steel ingots, the cooling rate can be as low as about 0.1 K/min,¹¹⁾ particularly inside of the ingot, which is an order of magnitude lower than the cooling rates in previous studies.

The effect of microsegregation on the formation of secondary inclusions have been evaluated by many model calculations, such as the pioneering work by Matsumiya et al.¹²⁾ and as summarized by Yu et al.¹³⁾ However, there have been a few studies on experimental observations of the relationship between microsegregation and secondary inclusion formation. Sawai et al.¹⁴⁾ investigated the relationship between inclusions and matrix composition on Ti/Zr deoxidized low carbon steel using computer-aided X-ray microanalyzer (CMA). The application of CMA to inclusion analysis provided an insight into the relationship between inclusion distribution and microsegregation. This methodology was also employed in a series of studies^15,16,17,18) on MnS precipitation on Ti/Zr oxides. However, composition information and minimum detection size by CMA is quite limited. Meanwhile, the development of automated inclusion analysis using SEM-EDS over the last two decades has made possible to obtain comprehensive information on inclusions from a relatively wide scanning area. This technique has become widely employed in many inclusion studies, but there are no experimental reports associating automated inclusion analysis data with microsegregation.

In this study, to understand the relationship between secondary inclusion formation and cooling rate especially under slow cooling conditions assuming large ingots, solidification experiments employing various cooling rates (0.17 to 128 K/min) were carried out. For the experiments, Fe-36mass%Ni alloy, which is also known as the invar alloy, was employed due to the following reasons: a) Fe-36mass%Ni alloy solidifies into a single γ phase from the beginning to the end of solidification, b) the deoxidation equilibrium of Fe–Ni alloy system has been well studied, c) it is a practical steel alloy that does not contain too many alloying elements, and d) due to its low coefficient of thermal expansion, it is widely used as a precision mechanical material that requires high cleanliness, and therefore, the control of inclusions is strongly demanded. The selection of the Fe-36mass%Ni alloy allows for easier modeling of the inclusion formation and more reliable deoxidation equilibrium calculations. Finally, as a new attempt to investigate inclusion formation, the combination of electron probe microanalyzer (EPMA) and SEM-EDS automatic inclusion analysis was employed in this study to correlate inclusion data obtained from SEM-EDS inclusion analysis with the microsegregation data obtained from EPMA analysis.

2. Experiments

Master alloy with Fe-36mass%Ni composition prepared from reagent materials by vacuum induction melting and electroslag remelting was used in this study. Chemical composition of prepared master alloy is shown in Table 1.

Table 1. The chemical composition of the prepared master alloy (in mass%).

C	Si	Mn	Ni	Al	O
0.0013	0.073	0.24	36.0	0.0008	0.0016

Heating experiments were conducted using a vertical heat resistance furnace in an Ar atmosphere (300 ml/min) as shown in Fig. 1. About 300 g of the master alloy was placed in an Al₂O₃ crucible (ID: 40 mm, OD: 50 mm, H: 100 mm), and heated to temperatures above 1773 K. Thermocouple with mullite sheath (ID: 4 mm, OD: 6 mm) was immersed in the sample melt and positioned 15 mm from the bottom of the crucible. In some experiments, metallic Si (purity >99.9%) was added into the melt through the Al₂O₃ tube to achieve a Si content of 0.25 mass% and stirred by the thermocouple sheath for 1 min. After adding metallic Si, the sample melt was kept at 1753 K (±2 K) for 1 hour before the cooling step. The holding temperature was set to 15°C above the liquidus temperature to ensure that solidification does not begin before the cooling step. In this study, several cooling procedures were applied to obtain widely different cooling rates; (a) Quenched: Just after the immersed thermocouple was pulled out from the melt, the crucible containing the sample was taken out from the furnace and cooled in the air. (b) Furnace-cooled: The power to the furnace was cut off and the sample was cooled inside the furnace. (c) Controlled cooling: The furnace temperature was gradually decreased at a cooling rate of 0.

Fig. 1.

Schematic diagram of the experimental apparatus.

2 to 3 K/min. After the sample was completely solidified, furnace power was turned off and the sample was cooled to room temperature under an Ar atmosphere. Specific experimental conditions such as the addition or no addition of Si, cooling conditions, and measured cooling rates (see below) are listed in Table 2.

Table 2. Experimental cooling conditions and measured chemical composition of the samples (* Estimated from SDAS).

No.	Si addition	Cooling condition	Cooling rate R [K/min]	Chemical Composition [mass%]
				C	Si	Mn	Ni	Al	O
1	None	Controlled (0.2 K/min)	0.17	0.001	0.07	0.23	36.38	0.0008	0.0026
2	None	Controlled (0.2 K/min)	0.17	0.002	0.09	0.21	35.83	0.0010	0.0011
3	None	Controlled (0.4 K/min)	0.30	<0.001	0.05	0.20	36.26	0.0015	0.0029
4	None	Controlled (1 K/min)	0.40	<0.001	0.04	0.18	36.09	0.0021	0.0042
5	None	Controlled (2 K/min)	1.07	0.002	0.02	0.14	35.98	0.0019	0.0076
6	None	Controlled (3 K/min)	1.02	0.005	0.05	0.20	36.12	0.0005	0.0042
7	None	Furnace-cooled	2.60	<0.001	0.05	0.20	36.21	0.0007	0.0052
8	None	Furnace-cooled	2.99	0.005	0.07	0.23	36.00	0.0013	0.0031
9	None	Furnace-cooled	6.86	0.001	0.04	0.17	35.75	0.0021	0.0062
10	None	Quenched	60.2*	0.001	0.04	0.18	36.14	0.0021	0.0091
11	None	Quenched	128*	0.001	0.06	0.21	36.20	0.0009	0.0066
12	Add	Controlled (0.2 K/min)	0.15	<0.001	0.16	0.18	35.76	0.0012	0.0020
13	Add	Controlled (2 K/min)	0.65	0.005	0.25	0.23	35.99	0.0018	0.0015
14	Add	Furnace-cooled	3.29	<0.001	0.19	0.20	36.06	0.0009	0.0029
15	Add	Furnace-cooled	5.27	0.004	0.21	0.21	35.49	0.0034	0.0023

In the experiments of “Furnace-cooled” and “Controlled cooling”, the temperature of the sample during solidification was continuously measured by the immersed thermocouple. For example, the temperature measurement of sample No. 3 (Controlled 0.4 K/min) is shown in Fig. 2. Solid line shows the measured temperature T [K], red dashed line shows a time derivative of temperature dT/dt [K/min] and bule dashed line shows second-order time derivative of temperature d²T/dt² [K/min²]. The temperature decreased just after the cooling step started. After a supercooling and recovery of temperature were observed, the temperature decreased again. As shown in Fig. 2, solidification start time t₁ [min], solidification finish time t₂ [min], solidification start temp. T₁ [K], and solidification finish temp. T₂ [K] are determined as the time of maximum undercooling, the time when d²T/dt² takes an extremum value, the maximum temperature after undercooling, and the temperature at the time when d²T/dt² takes an extremum value, respectively. In this study, the local solidification time Δt [min] and the average cooling rate R [K/min] (here after simply referred to as “cooling rate”) are defined as Eqs. (1) and (2), respectively.   

$$\mathrm{\Delta }t=t_2-t_1$$(1)

  

$$R=\frac{T_1-T_2}{\mathrm{\Delta }t}=\frac{T_1-T_2}{t_2-t_1}$$(2)

Fig. 2.

The result of temperature measurement of sample No. 3.

After the sample has completely cooled down, the solidified sample was removed from the crucible and cut horizontally 10 mm from the sample bottom, while the top part was cut longitudinally in half. The bottom part was used for the chemical composition analysis. The Si, Mn, and Ni contents were obtained by analyzing the sample cross section using an X-ray fluorescence analysis (Shimadzu MXF-2400). The C, Al, and O contents were obtained by analyzing the specimen taken from the center of the cross section using a combustion analyzer (LECO CS-800), an atomic absorption spectrophotometer (Hitachi Z-2010), and an infrared absorptiometry (LECO ON836), respectively. Half of the sample taken from the top was used for the measurement of the secondary dendrite arm spacing. The vertical section was polished and etched with an aqueous copper ammonium chloride solution for 1 minute. Then, the secondary dendrite arm spacing was taken using a digital microscope (KEYENCE VHX-6000) as the average of more than 30 measurements per sample. The other half of the sample taken from the top, after mounting in a resin, and polished, was used for the inclusion analysis using an automatic SEM-EDS inclusion analysis (JEOL JSM-6610LA). The detail of the inclusion analysis is described in our previous paper.¹⁹⁾ In this study, the analysis area and the minimum detection size were set be about 50 mm² and 1 μm², respectively. By using this technique, inclusion number, composition, size, and distribution were comprehensively obtained.

In some samples (No. 3, 11, and 15), EPMA analysis (JEOL JXA-8530F) was carried out just after SEM-EDS inclusions analysis without any treatment on the surface. The area analyzed by EPMA analysis (2 mm²) was included in that of the SEM-EDS inclusion analysis (about 50 mm²), allowing to combine the results of the EPMA and the SEM-EDS analyses.

3. Results and Discussions

3.1. Secondary Dendrite Arm Spacing (SDAS) and Cooling Rate

Examples of the microstructure (Samples No. 2, 6, and 14) photographed using digital microscope are shown in Fig. 3. With increasing cooling rate, the observed microstructure and dendrite arm spacing became smaller. Measured sample cooling rates, R [K/min], and secondary dendrite arm spacings, SDAS [μm], are shown in Fig. 4. From these, the relationship SDAS=AR^−B, where A and B are constants, was satisfied as shown in Eq. (3):   

$$SDAS=243R^{-0.389}$$(3)

The above relationship was found similar with the reported equation for Fe-26mass%Ni alloy,²⁰⁾ suggesting reliability of the current work. The actual cooling rates of the “Quenched” samples cannot be accurately measured during cooling, so the cooling rates were estimated from the measured SDAS using Eq. (3) from above.

Fig. 3.

Microstructures of etched samples photographed by using digital microscope (×30) for the SDAS measurement. (Online version in color.)

Fig. 4.

Relationship between secondary dendrite arm spacing SDAS and cooling rate. (Online version in color.)

The measured liquidus temperature T_L and solidus temperature T_S (T₁ and T₂ in Fig. 2) are shown in Fig. 5 together with equilibrium calculation data using FactSage7.3^TM with FSstel database (Fe-0.1Si-0.2Mn-36Ni in mass%). The measured values of T_L were nearly constant regardless of the cooling rate and agree well with that of the equilibrium calculation. On the other hand, the measured values of T_S decreased with increasing cooling rate and were lower than the values obtained from the equilibrium calculation. This difference between the measured and calculated T_S values might be explained by segregation. Under slower cooling rates, measured T_S values were close to the calculated values since there was enough time for diffusion to occur. On the other hand, under higher cooling rates, the effect of segregation was more significant because of the lesser influence of diffusion.

Fig. 5.

Measured liquidus temperature T_L and solidus temperature T_S. (Online version in color.)

3.2. Chemical Compositions

The measured chemical compositions of each sample are shown in Table 2 and the relationship between Al and O contents are shown in Fig. 6. Reported experimental results of the Al deoxidation equilibrium in Fe-36mass%Ni at 1773 K^19,21) and calculated lines of the Al deoxidation equilibrium using their proposed parameters^21,22) at 1753 K, which is the holding temperature before the cooling step, are also described in the figure. Despite the dispersion of Al and O contents in this study, in general, it was consistent with the Al deoxidation equilibrium at 1753 K. The relationship between aluminum deoxidation product, logK′_Al (=log[mass%Al]²[mass%O]³), and cooling rates are shown in Fig. 7. It was found that logK′_Al decreases with cooling rate, suggesting that at slow cooling rates, inclusions formed during cooling had sufficient time to float and separate. On the other hand, at faster cooling rates, inclusions became more difficult to separate and the equilibrium state before cooling was apparently maintained.

Fig. 6.

Relationship between Al and O contents and reported Al deoxidation equilibrium at 1773 K and the calculation at 1753 K. (Online version in color.)

Fig. 7.

Relationship between the aluminum deoxidation product logK′_Al and cooling rate. (Online version in color.)

3.3. SEM-EDS Inclusion Analysis

Results of SEM-EDS automatic inclusion analysis are summarized in Table 3, where inclusion number density N_A [–/mm²], area fraction f_A [–], arithmetic mean equivalent circle diameter (ECD) d_Ari [μm], and volumetric number density N_V [–/mm³] were obtained from Eqs. (4), (5), (6), (7), (8). Here, Eq. (8) is based on Fullman’s work,²³⁾ which is also expressed as DeHoff’s equation.²⁴⁾   

$$N_A=\frac{n}{A}$$(4)

  

$$f_A=\frac{\sum S_i}{A}\times {10}^6$$(5)

  

$$d_{Ari}=\frac{1}{n}\sum d_i=\frac{1}{n}\sum \sqrt{\frac{4S_i}{\pi }}$$(6)

  

$$d_{Har}=\frac{n}{\sum \frac{1}{d_i}}=\frac{n}{\sum \sqrt{\frac{\pi }{4S_i}}}$$(7)

  

$$N_V=\frac{2N_A}{\pi d_{Har}}\times {10}^3$$(8)

From the above equations, A, n, S_i, d_i, and d_Har are analysis area [mm²], number of inclusions in analysis area [–], size of inclusion i [μm²], ECD of inclusion i [μm], and harmonic mean ECD d_Har [μm], respectively. EDS analysis of the inclusions indicate that they were mainly composed of oxides such as MgO, Al₂O₃, SiO₂, and MnO. Other kind of non-metallic inclusions such as sulfide, nitride, and carbide were not observed. The presence of other oxides indicates that although Al deoxidation equilibrium was dominant and the thermodynamic activity of Al₂O₃ was almost unity before cooling, such was not the case during solidification. This will also be explained further in the next section. Since Mg was not intentionally added both in the master alloy production and in the solidification experiment, it was assumed to have entered from refractory materials used in the production of the master alloy using vacuum induction melting.

Table 3. Results of SEM-EDS automatic inclusion analysis.

No.	Number density NA [–/mm2]	Area fraction fA ×106 [–]	Arithmetic mean ECD dAri [μm]	Volumetric number density NV [–/mm3]	Chemical Composition (mass%)
					MgO	Al2O3	SiO2	MnO
1	1.80	175	9.81	165	<0.1	11.6	75.0	13.2
2	0.70	41	7.73	88	0.2	4.0	87.8	7.9
3	3.35	238	8.64	314	0.3	4.0	87.8	8.0
4	9.16	303	5.67	1349	<0.1	2.4	91.4	6.1
5	11.21	443	6.73	1226	0	1.0	41.2	57.8
6	9.65	249	5.19	1490	<0.1	3.0	94.4	2.5
7	32.74	312	3.17	7827	0.5	5.3	79.1	15.1
8	26.38	148	2.49	7746	3.9	21.9	48.5	25.4
9	34.05	389	3.49	7509	<0.1	5.8	78.8	15.1
10	69.99	442	2.31	25694	0	2.4	64.1	33.1
11	48.62	360	2.79	13360	0.1	6.7	83.1	10.0
12	0.76	49	7.33	121	0.1	6.9	90.5	2.4
13	5.35	129	4.25	1195	5.0	37.7	39.8	17.4
14	31.32	198	2.59	9109	0.1	31.2	64.8	3.7
15	22.28	132	2.54	6503	5.6	39.2	37.8	17.1

Examples of observed inclusion images and compositions analyzed by SEM-EDS automatic inclusion analysis are shown in Fig. 8. Here, It is note that inclusion images are a black-white inversion from the original back scattered electron (BSE) image and EDS analysis in SEM-EDS inclusion analysis analyzes the compositional composition of the entire inclusion, even if there are multiple constituent phases in a single inclusion. The constituent phases of inclusions changed from Sp (MgO·Al₂O₃), Sp + L (MgO–Al₂O₃–SiO₂–MnO), L, L+SiO₂, and finally SiO₂ with increasing SiO₂ content. All inclusions were independent (not clustered) and circular, although they differed in composition and phase.

Fig. 8.

Examples of SEM-EDS automatic inclusion analysis. (a) BSE image of inclusions in sample No. 7. (b) BSE image of inclusions in sample No. 8. (c) Inclusion compositions. (Online version in color.)

The relationship between inclusion composition and equivalent circle diameter (ECD) using three different cooling rates (R=0.17, 1.028, and 128 K/min) are shown in Fig. 9. It was observed that the relationship between SiO₂ concentration and ECD changed depending on the cooling rate. Under a fast cooling rate (No. 11), SiO₂ contents gradually decreased with increasing ECD. Under a medium cooling rate (No. 6), the decrease in SiO₂ content with increasing ECD became smaller, and under the slowest cooling rate (No. 1), SiO₂ contents did not change with ECD anymore. This will be discussed later.

Fig. 9.

Relationship between inclusion composition and equivalent diameter under three different cooling rates (R=0.17, 1.02 and 128 K/min).

The relationship between inclusion number density and ECD using three different cooling rates (R=0.17, 1.02 and 128 K/min) are shown in Fig. 10. Under a fast cooling rate (No. 11), the inclusion number density was maximum at ECD less than 2 μm, and the number density decreased as ECD increased. Under a medium cooling rate (No. 6), the peak position of the inclusion number density shifted to an ECD of 5 μm and the peak of inclusion density decreased. Under a slow cooling rate (No. 1), no clear peak in inclusion number density was observed, and was at the highest value for an ECD of 12 μm and larger. These results indicate that the number of inclusions and the ratio of small inclusions were higher under fast cooling rate conditions. This is also clearly confirmed in Fig. 11 showing the effect of the cooling rate on number density and arithmetic mean ECD, respectively. Although steel grade and inclusion species were different in each study^1,3,4,5,8,9) and some scattering was observed at high cooling rates, which may have been caused by the detection size limit, a consistent trend of increasing number and decreasing size of inclusions with increasing cooling rate can be clearly observed. Present study also confirms that these trends are satisfied up to a cooling rate of 10⁻¹ K/min. It must also be noted that Fig. 11 includes some cited data for which it is unclear whether they are arithmetic mean or harmonic mean, but the overall trend of decreasing inclusion size with increasing cooling rate is considered unaffected.

Fig. 10.

Relationship between number density and equivalent diameter under three different cooling rates (R=0.17, 1.02 and 128 K/min).

Fig. 11.

Effect of the cooling rate on a) inclusions number density N_A and b) arithmetic mean ECD d_Ari.

Unlike inclusion number and size, area fraction f_A [–] and volume fraction f_V [–] of inclusions are reported by Goto et al.¹⁾ and Zhang et al.³⁾ to have little relationship with the cooling rate. This was also confirmed in the present work as shown in Fig. 12(a). On the other hand, as shown in Fig. 12(b), area fraction was found to correlate with the oxygen content. Here, oxygen contents in inclusions, [mass%O]_Inclusion, was estimated by Eq. (9) and compared with the total oxygen content measured by combustion analysis. In this estimation, it was assumed that the inclusion composition, which was taken as the average composition, is uniform; that for inclusion density, the additivity relationship holds; and that the area and volume fractions of the inclusions are equal (f_A=f_V).   

$${\left[\text{masss\%O}\right]}_{Inclusion}=\frac{f_A\sum \frac{y{M}_{\text{O}}}{M_{{\text{M}}_x{\text{O}}_y}}\left(\%{\text{M}}_x{\text{O}}_y\right)}{{\rho }_{Metal}\sum \frac{\left(\%{\text{M}}_x{\text{O}}_y\right)}{{\rho }_{{\text{M}}_x{\text{O}}_y}}}\times 100$$(9)

Here, (%M_xO_y), ρ_i, and M_i are average composition of component M_xO_y [mass%], density of component i [g/cm³], and molar mass of component i [g/mol], respectively. Densities used in this calculation are listed in Table 4.²⁵⁾ From this estimation, it was found that the total oxygen content is almost equal to the oxygen content in inclusions as shown in Fig. 13. It can be considered that dissolved oxygen in liquid alloy completely changed into oxide inclusions during solidification and dissolved oxygen content in solidified alloy is negligibly small. Therefore, the area fraction of inclusions, which indicates the amount of inclusion, was controlled by oxygen content as shown in Fig. 12(b).

Fig. 12.

Relationship between a) cooling rate, b) oxygen content and area/volume fraction.

Table 4. Densities used in this study (g/cm³).

ρMetal	ρMgO	ρAl2O3	ρSiO2	ρMnO
8.2	3.58	3.98	2.65	5.45

Fig. 13.

Relationship between the total oxygen content and oxygen content in inclusions.

3.4. SEM-EDS Inclusion Analysis and Combination with SEM-EDS Inclusion Analysis

EPMA analysis of samples No. 3, 13, and 15 (R=0.30, 0.65, and 5.27 K/min) were carried out and EPMA mapping of Ni, Si, and Mn of sample No. 15 are shown in Fig. 14. It was clearly observed that microsegregation occurred during solidification of the sample. Now, since EPMA analysis was carried out just after the SEM-EDS automatic inclusion analysis and the area of both analyses overlaps, the results of these two analyses can be superimposed like a cell animation. In other words, the inclusion information such as composition, size, and distribution obtained by the SEM-EDS can be combined with the information of microsegregation obtained by the EPMA analysis.

Fig. 14.

Result of EPMA analysis of sample No. 15.

The superimposed image of inclusions distribution by SEM-EDS and Ni mapping image by EPMA for sample No. 15 is shown in Fig. 15. The composition of the inclusions observed within that area is shown in Fig. 16. Here, inclusions are classified in two groups, one with an inclusion composition of Al₂O₃+MgO ≥ 40 mass% (described as high-Al₂O₃ inclusion) and the other with that of Al₂O₃+MgO < 40 mass% (described as low-Al₂O₃ inclusion). The same superimposed image was also obtained for samples No. 3 and 13 as shown in Fig. 17. Although there are a few exceptions, it was found that high-Al₂O₃ inclusions were in the low-Ni region and low-Al₂O₃ inclusions were in the high-Ni region. This can be clearly observed in Fig. 18(a), which shows the relationship between Al₂O₃+MgO content of inclusions and Ni content of the matrix at the location of inclusions.

Fig. 15.

Inclusions distribution on the Ni mapping image for sample No. 15.

Fig. 16.

Compositions of inclusions located in the analyzed area of the EPMA analysis (No. 15).

Fig. 17.

Inclusions distribution on the Ni mapping image for samples No. 3 and 13.

Fig. 18.

Relationship between Ni content of matrix and a) Al₂O₃+MgO content in inclusions and b) estimated solid fraction f_S. (Online version in color.)

The relationship between inclusion composition and microsegregation was obtained by combining the EPMA analysis and the SEM-EDS inclusion analysis. However, further investigation of inclusion formation during solidification may require consideration of the solid and/or liquid fraction. Since Ni content of both liquid and solid phase at a solidification interface is expected to increase as solidification proceeds, Ni content in the later solidified position should be higher than that in the earlier solidified solid position. Considering adjacent points A and B, and point A solidified earlier than point B, Ni contents in each position (C_A and C_B) have a relation of C_A < C_B upon solidification. However, after solidification, the difference between C_A and C_B should decrease because of diffusion. Now, assuming the concentration gap will not be reversed and the solid fraction distribution over the analytic area is uniform, the solid fraction f_S [–] of a certain point when they are solidified can be estimated by using Eq. (10).   

$$f_S=\frac{n_i}{N}$$(10)

Here, n_i and N are the order of Ni content from lowest to highest analysis point in the EPMA mapping and the number of analyzed points in the EPMA mapping (in this study, N=160000). The estimated relationship between Ni content of matrix and solid fraction is shown in Fig. 18(b) and the relationship between the estimated solid fraction and inclusion number is shown in Fig. 19. It can be confirmed that high-Al₂O₃ inclusions were in the lower solid fraction region while low-Al₂O₃ inclusions were in the high solid fraction region, indicating that low-Al₂O₃ inclusions formed during the later stage of solidification. The peaks of low Al₂O₃ inclusions were found in f_S = 0.8–1.0 and those of high-Al₂O₃ inclusions were in f_S = 0.1–0.3.

Fig. 19.

Relationship between estimated fs and inclusion number. (Online version in color.)

Similar phenomenon was also reported by experimental investigation and numerical analysis. Gamutan et al.²⁶⁾ experimentally studied the inclusion composition change of Fe-0.2Si-0.8Mn-7.4Cr-0.1Al (in mass%) by quenching samples from different temperatures during solidification at a cooling rate of 1 K/min. In the sample quenched at liquidus temperature, Al₂O₃ inclusions were mainly observed. But in the sample at solidus temperature, Al₂O₃–SiO₂–MnO inclusions were mainly observed. They concluded that this composition changes might be caused by microsegregation. Choudhary and Ghosh²⁷⁾ calculated the inclusion formation and composition changes of Mn–Si deoxidized low carbon steel by using the Clyne–Kurz model²⁸⁾ and FactSage. In their calculation, pure Al₂O₃ and MnO–SiO₂–Al₂O₃ inclusions are generated in early stage of solidification and SiO₂ and SiO₂ rich MnO–SiO₂–Al₂O₃ inclusions are generated in later stage of solidification.

3.5. Thermodynamics Calculation and Comparison with the Experimental Result

Using FactSage7.4^TM thermodynamic software with Scheil-Gulliver cooling employing the database of FSstel for metallic phase and FToxid for oxide phase, inclusion formation during cooling from 1753 K to 1673 K was calculated for Fe-0.05/0.20Si-0.2Mn-36Ni-0.001Al-0.0011O (in mass%). Here, oxygen contents were assumed to be in equilibrium with Al₂O₃ at 1753 K (before cooling). Calculation results are shown Fig. 20. In both conditions of low-Si and high-Si, Al₂O₃, Slag (Al₂O₃–SiO₂–MnO slag liquid), Mullite and SiO₂ were formed in sequence. The formation of inclusions can be divided into three stages. Stage I is in the temperature range of 1753 K to T_L. Since it was intended to be in Al deoxidation equilibrium at 1753 K, Al₂O₃ was formed with decreasing temperature. Stage II is in the temperature range of T_L to 1724 K, where the amount of Al₂O₃ formation rapidly increased as solidification started due to the increasing oxygen content in the liquid phase due to solidification (solute ejection from the solid/liquid interface). Stage III begins from the temperature below 1724 K, where the formation of new Al₂O₃ inclusions stopped and the formation of Mullite, Slag (Al₂O₃–SiO₂–MnO liquid) and SiO₂ started. This can be explained by the depletion of dissolved Al as it was already consumed in the earlier stages and the increasing Si and Mn contents due to the solidification, respectively.

Fig. 20.

Calculation of inclusion formation using FactSage with Scheil-Gulliver cooling for low-Si as Fe-0.05Si-0.2Mn-36Ni-0.001Al-0.0011O and high-Si as Fe-0.2Si-0.2Mn-36Ni-0.001Al-0.0011O (in mass%).

For samples No. 11 and 15, thermodynamic calculation results agree well with the experimental result, of which high-Al₂O₃ inclusions formed in the early stage of solidification while low-Al₂O₃ inclusions formed in the later stage of solidification. On the other hand, for sample No. 3, there were no high-Al₂O₃ inclusions and only low-Al₂O₃ inclusions. This could be explained by the following two reasons: (1) Since the cooling rate of No. 3 was lowest among those three samples, there was enough time for the high-Al₂O₃ inclusions that formed in the early stage of cooling to float up. This is also supported by Fig. 6, where the logK’_Al was found to decrease with cooling rate; or (2) High-Al₂O₃ inclusions were not able float up because they were inhibited by the dendrites but instead reacted with the Si-enriched liquid and then changed into low-Al₂O₃ inclusions. Figure 21 shows the calculated Si and Al contents during cooling and the phase stability of inclusions at 1673 K. Si and Al contents changed from Al=0.001 mass% and Si = 0.05/0.2 mass% to low-Al and high-Si contents. The stable inclusion phase also changed from Al₂O₃ to Mullite, Slag and finally SiO₂. This calculation does not consider the reaction between the generated solid phase (including inclusions) and the liquid phase, so the generated Al₂O₃ remains without any reaction, but the calculation of stable inclusion phase predicts that the initially formed Al₂O₃ will react with the Si-enriched liquid and finally change to SiO₂. This is also supported by Fig. 9 indicating the dependence of cooling rates on the relation between inclusion SiO₂ content and ECD. Under a fast cooling rate, although small inclusions can be changed into low-Al₂O₃ inclusions, large inclusions did not have sufficient time to completely change composition and remained low in SiO₂ content. On the other hand, under a slow cooling rate, there was enough time for inclusion composition to change even in the larger inclusions.

Fig. 21.

Calculated Si and Al contents during cooling for low-Si as Fe-0.05Si-0.2Mn-36Ni-0.001Al-0.0011O and high-Si as Fe-0.2Si-0.2Mn-36Ni-0.001Al-0.0011O (in mass%) and dominant inclusions in Fe-0.2Mn-36Ni at 1724 K.

3.6. Discussion of the Mechanism for Composition Change during Solidification

The results obtained from the combination of SEM-EDS inclusions analysis and EPMA mapping analysis, wherein high-Al₂O₃ inclusions are engulfed at the early stage of cooling and low-Al₂O₃ inclusions are engulfed at the later stage of cooling, can be explained in terms of three possible mechanisms as shown in Fig. 22. Here, Stage I, II and III are basically the same temperature range as that of Fig. 20. In the case of (1) Selective Engulfment, both high and low-Al₂O₃ inclusions already exist from the beginning. So, the difference in terms of the tendency to engulf or push an inclusion causes a difference in the timing of engulfment. Ohta and Suito²⁹⁾ calculated the pushing/engulfment transition of different kind of inclusions by the wetting/interfacial model. They concluded that Al₂O₃ is more easily pushed than MnO–SiO₂ inclusions. This mechanism was found inconsistent with the results of equilibrium calculations (Fig. 20), which showed that the initially formed inclusions are mainly Al₂O₃ and lower SiO₂ inclusions are formed at later stage of solidification. The other mechanisms are (2) Stepwise Formation: high-Al₂O₃ inclusions formed at earlier stage of solifidication and low-Al₂O₃ inclusions formed at late stage of solidification and (3) Composition Change: high-Al₂O₃ inclusions which formed at earlier stage of solidification is changed into low-Al₂O₃ inclusions during solidification. Whether mechanism (2) or (3) is preferred depends on the balance of initial Al and O contents. If the initial Al content is high, large quantities of the initially formed Al₂O₃ is expected to reduce the oxygen content sufficiently thereby reducing inclusion formation during solidification. Whereas if the initial Al content is low, oxygen content remains high enough during solidification to form new inclusions by reacting with the enriched solute. Therefore, further investigation using kinetics analysis and reaction models that consider the reaction between inclusions and residual molten steel are needed to determine which mechanism is more likely to occur. Nonetheless, this study confirms that the composition of inclusions changed during solidification using both experimental investigation and computational thermodynamic analysis.

Fig. 22.

Possible mechanism explaining the experimental results.

4. Conclusion

To obtain fundamental information on the changes in size, number, and composition of secondary inclusions during solidification of steels, solidification experiments using an Fe-36mass%Ni alloy at various cooling conditions (0.17 to 128 K/min) were carried out. The major findings of this study are summarized as follows:

(1) The number density and size of inclusions were significantly influenced by cooling rate. As the cooling rate decreases, inclusions number density decreases and inclusion size increases. Meanwhile, the area fraction of inclusions was found to have little relationship with the cooling rate. Experimental results were found to agree well with other reported studies, confirming that these relationship holds up to the slow cooling rate of 0.17 K/min.

(2) The area fraction of inclusions was related with the total oxygen content in steels. By estimating the oxygen content in inclusions, dissolved oxygen content in liquid alloy completely changed into oxide inclusions during solidification such that dissolved oxygen content in the solidified alloy was negligibly small.

(3) The composition of inclusions changed from high-Al₂O₃ inclusion to low-Al₂O₃ inclusion during solidification. Moreover, the influence of cooling rate on the inclusion composition change differed by inclusion size.

(4) Combining SEM-EDS inclusion analysis with the EPMA analysis to investigate inclusion composition and microsegregation was achieved. It was confirmed that high-Al₂O₃ inclusions were in the lower solid fraction region while low-Al₂O₃ inclusions were in the high solid fraction region, suggesting that low-Al₂O₃ inclusions formed during the later stage of solidification and/or changes in inclusion composition occurred during solidification.

(5) Present experimental results agreed well with the calculation of inclusion formation using thermodynamics software using the Scheil-Gulliver model and the stable phase of inclusions equilibrated with remaining metal liquid. However, further kinetics analysis considering the reaction between precipitated inclusions and remaining liquid metal was also found necessary.

References

• 1)   H.  Goto,  K.  Miyazawa,  K.  Yamaguchi,  S.  Ogibayashi and  K.  Tanaka: Tetsu-to-Hagané, 79 (1993), 1082 (in Japanese). https://doi.org/10.2355/tetsutohagane1955.79.9_1082
• 2)   H.  Goto,  K.  Miyazawa,  K.  Yamaguchi,  S.  Ogibayashi and  K.  Tanaka: ISIJ Int., 34 (1994), 414. https://doi.org/10.2355/isijinternational.34.414
• 3)   X.  Zhang,  W.  Yang,  H.  Xu and  L.  Zhang: Metals, 9 (2019), 392. https://doi.org/10.3390/met9040392
• 4)   Y.  Luo,  A. N.  Conejo,  L.  Zhang,  L.  Chen and  L.  Cheng: Metall. Mater. Trans. B, 46 (2015), 2348. https://doi.org/10.1007/s11663-015-0401-3
• 5)   Y.  Wang,  W.  Yang and  L.  Zhang: Steel Res. Int., 90 (2019), 1900027. https://doi.org/10.1002/srin.201900027
• 6)   M.  Suzuki,  R.  Yamaguchi,  K.  Murakami and  M.  Nakada: ISIJ Int., 41 (2001), 247. https://doi.org/10.2355/isijinternational.41.247
• 7)   H. S.  Kim,  H. G.  Lee and  K. S.  Oh: ISIJ Int., 42 (2002), 1404. https://doi.org/10.2355/isijinternational.42.1404
• 8)   H. S.  Yu and  J. G.  Li: Int. J. Miner. Metall. Mater., 22 (2015), 1157. https://doi.org/10.1007/s12613-015-1180-1
• 9)   N.  Kikuchi,  S.  Nabeshima,  Y.  Kishimoto and  S.  Sridhar: ISIJ Int., 48 (2008), 934. https://doi.org/10.2355/isijinternational.48.934
• 10)   N.  Kikuchi,  S.  Nabeshima,  T.  Yamashita, Y kishimoto, S. Sridhar and T. Nagasaka: ISIJ Int., 51 (2011), 2019. https://doi.org/10.2355/isijinternational.51.2019
• 11)   H.  Yamada,  T.  Sakurai and  A.  Takenouchi: Tetsu-to-Hagané, 75 (1989), 105 (in Japanese). https://doi.org/10.2355/tetsutohagane1955.75.1_105
• 12)   T.  Matsumiya: Mater. Trans., JIM, 33 (1992), 783. https://doi.org/10.2320/matertrans1989.33.783
• 13)   D.  You,  S. K.  Michelic,  P.  Presoly,  J.  Liu and  C.  Bernhard: Metals, 7 (2017), 460. https://doi.org/10.3390/met7110460
• 14)   T.  Sawai,  M.  Wakoh,  Y.  Ueshima and  S.  Mizoguchi: ISIJ Int., 32 (1992), 169. https://doi.org/10.2355/isijinternational.32.169
• 15)   M.  Wakoh,  T.  Sawai and  S.  Mizoguchi: Tetsu-to-Hagané, 78 (1992), 1697 (in Japanese). https://doi.org/10.2355/tetsutohagane1955.78.11_1697
• 16)   M.  Wakoh,  T.  Sawai and  S.  Mizoguchi: ISIJ Int., 36 (1996), 1014. https://doi.org/10.2355/isijinternational.36.1014
• 17)   M.  Wakoh,  T.  Sawai and  S.  Mizoguchi: Tetsu-to-Hagané, 82 (1996), 593 (in Japanese). https://doi.org/10.2355/tetsutohagane1955.82.7_593
• 18)   T.  Sawai,  M.  Wakoh and  S.  Mizoguchi: Tetsu-to-Hagané, 82 (1996), 587 (in Japanese). https://doi.org/10.2355/tetsutohagane1955.82.7_587
• 19)   H.  Fukaya,  K.  Kajikawa,  A.  Malfliet,  B.  Blanpain and  M.  Guo: Metall. Mater. Trans. B, 49 (2018), 2389. https://doi.org/10.1007/s11663-018-1336-2
• 20)   M. C.  Flemings,  D. R.  Poirier,  R. V.  Barone and  H. D.  Brody: J. Iron Steel Inst., 208 (1970), 371.
• 21)   S. B.  Lee,  S. M.  Jung, H. G. Lee and C. H. Rhee: ISIJ Int., 42 (2002), 679. https://doi.org/10.2355/isijinternational.42.679
• 22)   H.  Fukaya,  S.  Nakajima,  J.  Gamutan,  S.  Suzuki,  K.  Kajikawa,  K.  Saito and  T.  Miki: ISIJ Int., 61 (2021), 2331. https://doi.org/10.2355/isijinternational.ISIJINT-2020-710
• 23)   R. L.  Fullman: JOM, 5 (1953), 447. https://doi.org/10.1007/BF03398971
• 24)   R. T.  DeHoff and  F. N.  Rhines: Quantitative Metallography, McGraw-Hill, New York, (1968), 128.
• 25)   Y.  Xu,  M.  Yamazaki and  P.  Villars: Jpn. J. Appl. Phys., 50 (2011), 11RH02. https://doi.org/10.1143/JJAP.50.11RH02
• 26)   J. L.  Gamutan,  C.  Fujiwara and  T.  Miki: Steel Res. Int., 93 (2022), 2200285. https://doi.org/10.1002/srin.202200285
• 27)   S. K.  Choudhary and  A.  Ghosh: ISIJ Int., 49 (2009), 1819. https://doi.org/10.2355/isijinternational.49.1819
• 28)   T. W.  Clyne and  W.  Kurz: Metall. Trans. A, 12 (1981), 965. https://doi.org/10.1007/BF02643477
• 29)   H.  Ohta and  H.  Suito: ISIJ Int., 46 (2006), 22. https://doi.org/10.2355/isijinternational.46.22