Analysis of Spinel Based Inclusions During the Last Stage of The Steelmaking Process of SAE 52100

2020 Al 2 O 3 (alumina) and MgO·Al 2 O 3 (spinel) inclusions cause valve obstruction (clogging) in continuous casting and can deteriorate the quality of the final product. In this context, industrial heats of the bearing steel SAE 52100 was examined. Samples were collected in the final steps of the steelmaking process, both after vacuum treatment and during continuous casting. A scanning electron microscope (SEM) equipped with a energy-dispersive spectrometer (EDS) and automated particle characterization analysis was used to characterize the inclusions present in the steel samples. Thermodynamic calculations were performed with the commercial software FactSage TM 7.2. Based on thermodynamic predictions, parameters such as solid fraction, liquid fraction, MgO saturation point of the slags, content of dissolved elements in steel (Al, Mg, Ca,...) and the construction of a phase stability diagram were determined. The results in this study showed a tendency for increase in MgO content in the inclusions with the decrease of %FeO and SiO 2 contents in the slag, an increase of binary basicity (%CaO/%SiO 2 ). It is verified that the MgO contents in the slag were close to the saturation, increasing the probability for the formation of inclusions rich in MgO and/or spinel. On the other hand, stability diagrams confirm the formation of spinel inclusions for each of the heats analyzed. During the final step of the steelmaking process, there is a tendency for re-oxidation, which is verified by an increase in the density of inclusions (or total oxygen TO The present on the analysis of slag composition (%FeO content, %SiO 2 content, binary basicity, liquid fraction and saturation point in MgO), steel composition, characterization and modification by calcium of stages the refining of the from Based on thermodynamic calculations, equilibrium diagrams we and control


Introduction
Bearing steels are known for their quality and performance, guaranteed by high toughness and low levels of non-metallic inclusions. In addition, these steels have good mechanical strength and high fatigue life. These characteristics result in mechanical properties are achieved through a rigid steelmaking process in the melt shop, where the liquid steel undergoes a series of refining reactions, including the removal of non-metallic inclusions (NMIs) through their transport, absorption and dissolution to the slag phase [1][2][3][4][5][6][7][8][9][10][11] . However, although all efforts made in the steel industry to strictly meet the cleanliness requirements (clean steels), it is still observed in the metallurgical processing of this steel, the residual of alumina (Al 2 O 3 ) inclusions, which are the product of deoxidation in Al-killed steels and spinel (MgO·Al 2 O 3 ) inclusions, formed by the presence of Mg in the liquid steel. Both are considered to be harmful to the mechanical properties of the steels [12][13][14] .
As described by Costa and Silva 1 , while there is still controversy concerning the effect of the individual types of NMIs on the fatigue life of bearing steels, some general agreement seems to prevail regarding calcium aluminates and spinel inclusions that are detrimental. The solid phases of higher occurrence in the inclusions are spinel, calcium aluminate and corundum. The formation of these inclusions is a sign that their modification was unfinished. The control of slag composition is always one of the key points in controlling NMIs in bearing steels. Furthermore, the reversion of magnesium from the slag, which may lead to the formation of spinel inclusions, must be avoided.
Following the same idea in regards to the effect of slags, Okuyama et al. 15 showed a strong relationship between slag parameters and the control of NMIs, indicating that an increase of basicity and the CaO/Al 2 O 3 ratio in the slag favors the formation of spinel inclusions in steel. Park and Todoroki 16 also report some alternatives (direct or indirect supply of calcium, reduction of basicity (%CaO/%SiO 2 , %CaO/%Al 2 O 3 ratio) and content of Al 2 O 3 and MgO in the slag, minimizing melt re-oxidation in conjunction with low Al content and others) for better control of spinel during the steelmaking process. According to Yang et al. 17 , modification of inclusions with calcium is not always effective, generating partly modified inclusions with spinel nuclei, which are detrimental to the properties of the steel due to their low deformability during hot rolling. Finally, studies on the formation of spinel inclusions have been reported through thermodynamic calculations of phase stability diagrams of the Al-Mg-O system in steel to predict the formation of inclusions 12,18-23 .
The present study focuses on the analysis of slag composition (%FeO content, %SiO 2 content, binary basicity, liquid fraction and saturation point in MgO), steel composition, characterization and modification by calcium of inclusions during the last stages of the refining process of the SAE 52100 obtained from industry. Based on thermodynamic calculations, equilibrium diagrams and fundamental theory found in the literature, we aim to contribute to a better understanding and increase the control of cleanliness (re-oxidation) and quality of this steel.
During the manufacturing process the stages were as follows: electric arc furnace (EAF) -processing time about 60 minutes, ladle furnace (LF) -processing time about 50 minutes, vacuum degassing tank type (VD) -processing time about 20 minutes (pressure 1x10-4 bar; average gas flow rate 125 Nl/min) and solidification in a continuous casting (CC) machine with three strands. The average temperature at the end of the secondary refining process and at the beginning of the casting is 1550 ºC. The deoxidation process applied includes aluminum (Al) addition during steel tapping and at the end of the vacuum degassing (VD). Instead of adding CaSi wire for inclusion treatment, the modification of inclusions were controlled by an indirect supply of calcium via slag phase to the metal during the ladle refining process. A total of nine (9) heats of SAE 52100 were monitored in this study. Steel samples were collected during the final step of the vacuum degassing treatment (samples X1), and during the continuous casting process at the tundish (samples X2). Furthermore, during the vacuum degassing step, samples of slag (samples Y1) were simultaneously collected in the same CC sequencing. The chemical compositions of the steel samples were determined by optical emission spectroscopy (ARL 3560 spectrometer) and total oxygen (TO) using LECO TC-436. The [Mg] content in the bath was not measured. The collected steel samples were prepared metallographically. Afterwards, the chemical composition, morphology, size and number of inclusions in the samples were determined using a scanning electron microscope (SEM-ASPEX Explorer) with automated feature analysis (AFA) and equipped with a dispersive energy spectrometer (EDS). The beam is centered on the inclusion using a rotating chord algorithm and an EDS spectrum is collected. During acquisition, the EDS spectrum is quantified and the metric (including average, maximum and minimum diameters, orientation and centroid) and quantification data are used to classify the feature according to user defined rules. A classification rule file is a series of Boolean expressions that define a set of classes (a list of elements in which you are interested) into which each feature is assigned 29 .
For this study, the analyzed area of the steel samples varied in the range of 49.7 to 66.9 mm2. Inclusions smaller than 2.5μm were disregarded, since, for inclusionary control of bearing steel, inclusions larger than 4μm are considered 8,25,28. The operating parameters were an accelerating voltage of 20 keV, 16-18 mm focus, backscattering mode, inclusion diameter detection limit was size 4 μm and maximum size 225 μm. Time process EDS:1s (min.) and 2s (máx). The slag compositions were measured by X-ray fluorescence spectroscopy (FRX).

Thermodynamic calculations
Thermodynamic calculations were performed with the aid of a computational thermochemical software package named FactSage TM 7.2 30 . In order to determine the solid fraction, liquid fraction and MgO saturation of the slags used, the chemical composition of the final slag (sample Y1) for each heat was taken. The temperature of 1550°C was considered and the FactPS and FToxid databases were selected 29 . This same methodology was also used to verify the influence of the increase of the silica in the slag.
In the Equilibrium module and from the chemical composition of the steel (sample X2), including the total oxygen and slag (sample Y1) of each heat, the  Table 1 summarizes the results of chemical analyses for the slag and steel and thermodynamic calculations of the steelmaking process in the analyzed heats. The results in this table will be explained in the course of the discussion of the results. For a better analysis of the results, the data obtained from the heats were separated into two sequence groups (A and B) to find out their effect on the variable of interest in the process and formation of inclusions in steel.

Effects of the slag chemical composition on
MgO content of inclusions Figure 2 shows the different compositions of the industrial slags of the analyzed heats represented in the pseudo-ternary system CaO-SiO 2 -Al 2 O 3 -11% MgO at the temperature of 1550°C.
As shown in Figure 2 all slag compositions are located in a MgO saturated region, (MgO+ slag liquid), with MgO saturation values (Table 1), ranging from 3.6 to 6 mass%. As can be seen in Table 1, the Al 2 O 3 , CaO, SiO 2 contents of these slags range between 15-25%, 47-52% and 13-17.6% by mass, respectively. Typical slag compositions for secondary refining of high basicity (2.8-3.6), saturated in CaO and MgO and with high content of alumina 25,36,37 . MgO content in the slag is in the range of 9 to 16 mass%. Values of 75 to 100% of the liquid fraction were observed. With the exception of heats with 100% liquid fraction, the solid slag fraction ranges from 7% to 25% due to the precipitation of saturated phases (C 2 S and MgO).
In the present work, [Al] contents range (calculated) in the steel varied between 0.0089-0.0130 mass% (89-130 ppm) and are considered minimum levels to those found in the literature 4,37,40,41 . From Figure 4a it is observed that with the decrease of the amount of FeO in the slag the content of MgO in the inclusions tends to be greater. That is, deoxidized slags promote the reduction of MgO in the slag by increasing the [Mg] dissolved in the steel. According to results of other researchers 38,39 slags with low reductive oxides (such as %FeO and %MnO) may promote inclusions with higher MgO content. Although not shown in this study, the relationship between MgO in the inclusions considering the contents of (%FeO+%MnO) in the slag did not promote significant changes in the tendency of the results presented only with FeO. The average MnO contents in the slags were (Heats A: 0.128%, Heats B: 0.085%). Figure 4b shows a good correlation between the dissolved [Mg] content and the aMgO/aFeO ratio at the end of the process. According to Mendez et al. 38 , the effect of the composition of the slag appears in the ratio between    Okuyama et al. 15 investigated the influence of the slag composition on the formation of spinel inclusions (MgAl 2 O 4 ) and concluded that by reducing the basicity (%CaO/%SiO 2 ) and (%CaO/%Al 2 O 3 ) ratio in the slag, the MgO content in alumina inclusions would be reduced. Deng et al. 11 reported that with the binary basicity at levels around 3 to 4, together with a slag with 20 wt% in Al 2 O 3 , the alumina activity decreases while that from MgO increases in the slag. This means that the dissolution of Mg into the steel bath is difficult, being favorable for the control of spinel formation. Figure 6 shows the different ranges of SiO 2 contents in the slag for the analyzed heats.
The values of the data, which vary in the range of 13.6 to 17.6 mass%, show the tendency of reduction in the MgO inclusion content with the increase of the SiO 2 content in the slag. This fact is in accordance with experimental results presented by Park and Todoroki 16      It is important to consider that the basicity of the slag is reduced with increasing silica (acid slag). The implications of this will be: aluminum as the most active deoxidation agent can reduce silica from a slag that increases these element contents in a metal; the bigger silica content in a slag, the bigger will be the losses of aluminum in the metal phase, forming alumina inclusions. Figure 7 shows the relationship between the MgO content in the inclusions and the liquid fraction in the slag of the analyzed heats.
As shown in Figure 7, there is an increase in the MgO content in the inclusions with increasing liquid fraction in the slag. This suggests that the slag/steel interaction is favored when the slag is more liquid and consequently increasing the transfer of Mg from the slag to the bath and subsequently to the inclusions. Similar results were obtained by Bartosiaki et al. 25 and Pereira et al. 29 .
According to Bartosiaki 25 , alumina acts as a flux in basic slags and high alumina contents are used to promote an abundant liquid phase without compromising the removal and absorption of nonmetallic inclusions. On the other hand, the binary basicity of the slag must be high enough to ensure the indirect treatment of the inclusions with calcium.

Inclusion characterization
The total oxygen (TO) content, composed in the sum of the oxygen dissolved in the steel and the oxygen present in the oxides (inclusions), can be used as a cleaning parameter in steels 46 . Figure 8 shows a comparison between the inclusion density and the total oxygen content in samples X1 (after vacuum) and X2 (in the tundish).
As can be observed after vacuum and in the tundish there is an increase in the density of inclusions for the analyzed heats. In parallel, the same tendency was found for the total oxygen (TO) variations between samples X1 and X2. In Figure 8, each sample presents, in particular, its total oxygen variation. However, there is a relationship between increasing the density of inclusions with the TO content. The positive variations of TO indicate reoxidation of the liquid steel (there was an increase in the total oxygen content). In the case of negative variations, most of them are close to 0 (without variation), according to samples A4, A5, B1 and B2, but it does not mean that there has not been a severe reoxidation as in the others. This behavior in the increase of the inclusions densities and total oxygen may suggests that reoxidation events occurred in many of these heats. In this case, re-oxidation is merely one of the possibilities for increasing the density of inclusions in steel.
It is known from literature that after secondary refining and during casting, liquid steel can be exposed to uncontrolled sources of oxygen (air). Materials listed among the sources include, the refractory material and oxygen passage through the slag from the tundish, re-oxidation of the steel during transfer between the ladle and tundish and especially during the ladle pouring to the tundish during the start of heats and after changing ladles 47,[48][49][50] . According to Shu-feng et al. 51 , any increase in oxygen potential since re-oxidation may lead to the formation of secondary spinel inclusions. On the other hand, according to Park and Todoroki 16 and Yanyan et al. 52 , during the steelmaking process the steel presents a temperature drop promoting the formation of new inclusions of alumina and/or spinel and, in the meantime, increasing the non-metallic inclusion densities.
The typical maximum cleaning requirements specified for bearing steels correspond to a total oxygen value (TO) <10 ppm. However, as observed in this study, it was very difficult to control TO at these levels, even reporting TO very high up to 22 and 29 ppm in some of the heats. Howewer, to obtain low oxygen in steel, high basicity refining slag with lower Fe t O content and low C/A ratio ranges during ladle furnace refining is necessary. There are new technologies to prevent oxygen contamination of molten steel: the prevention of slag invasion into the tundish by controlling the amount of molten steel remaining in the ladle; the prevention of air contamination of molten steel flow from ladle to tundish by adopting the tundish seal box or use of protective shrouding to protect from re-oxidation; the prevention of air contamination of molten steel by a perfect seal between the tundish and immersion nozzles and the prevention of the involution of mold powder by the control of immersion depth of immersion nozzles 8,36,37 . Figure 9 illustrates the variation of relative fraction (population) in the quantitative analyis of each type of inclusion in molten steel at 1550°C in the tundish. It can be  observed that almost all the samples presented similar species of inclusions groups including, other inclusions (calcium alumino-silicate, sulfides and others), spinels, complex and alumina, the last three being those not completely transformed into calcium aluminates liquid. Alumina inclusions are present in samples A1, A2, A5, B1, B3 and B4. In regards to particle counting, it is important to identify samples A3 and B2 as the samples with the lowest indices, appearing with less than 20 particles detected.
The observed inclusions for the complex as well as the partially modified inclusions have an average maximum size of 5.96 μm with most of the bands of diameters above 35% as opposed to diameters of 4.19 and 5.21 μm that were below this ratio. Regarding the inclusions of pure spinel with a average maximum size of 8.05 μm, a maximum proportion to 35% in the average diameter ranges of 3.77 and 4.79 μm is observed, and below of 22% for the rest of the inclusion diameters. In geral, the proportion of spinel inclusions observed in these heats are in low proportion to the amount of complex inclusions. Table 2 shows the average compositions of all inclusions taken from the distributor (sample X2), together with the CaO/Al 2 O 3 (C/A) ratios in the inclusion.
According to Bao et al. 41 , with the aim of modifying the solid inclusions into liquids (C 12 A 7 ), the CaO/Al 2 O 3 ratio of the inclusions should be approximately 0.86 and the MgO content should be controlled below 10%.
Based on these results, in the case of heats A, heats A3 and A2 show a high C/A ratio of inclusions, which means that they were quite modified compared to heats A1, A4 and A5. For these heats A, the MgO content is high in heats A1, A2 and A4 with values of 11.47, 13.86, 15.21 and, for heats A3 and A5 with values of 6.8 and 9.61. As can be seen, heat A2 has a C/A ratio of 0.8 with a high final average MgO content in the inclusions, in which case a partial modification of the inclusions may be considered. On the other hand, for heats B, only heats B2 and B4 reach the maximum modification of the inclusions with C/A values of 0.80 and 0.84, unlike heats B1 and B3 both with a value of 0.46, also showing that their modification was insufficient. The high MgO content is in the heats B1 and B3, with 18.78 and 16.72 as opposed to the B2 and B4 heats with values of 6.87 and 9.01 (and close to those of heats A3 and A5).
According to the studies of other researchers 34,39,49,50,53 , this behavior is associated with the transformation mechanism of inclusion following the sequence The low C/A ratio in the inclusions shows itself as one of the reasons why spinel inclusions are frequently observed in these heats 51 . An estimation of the Ca contents required for the spinel modification was performed using thermodynamic calculations, Table 3, through the variation of the Ca content in the samples collected at the distributor (X2). Comparing the total Ca contents, the amount of calcium required is high for heats A1, A2, B1 and B3 and to a lesser extent for heats A4 and B2 and nonexistent for heats A3, A5 and B4.
Some of these heats, such as the sequential heats, A1 (1st), A2 (2nd), B1 (2nd), B3 (4th) required more calcium to be added. There were almost constant additions to the heats A3, A4, A5, B2 and B4. According to Geldenhuis and Pistorius 55 the amount of calcium to be added depends on the concentration of alumina inclusions in the steel. Recently, Kumar and Pistorius 56-58 emphasizes the low rate of transfer of calcium by slag to the inclusions. On the other hand, Yang Wen et al. 34 considered re-oxidation of molten steel and precipitation of CaS as factors for consuming large amounts of calcium that was expected to be used to modify inclusions to calcium aluminates with low melting points. According Costa e Silva 1 the amount of calcium needed to modify oxides depends on the total oxygen content in the steel. To date, there is no method of determining total oxygen in time to use it for a decision on the amount of calcium to , which may explain the appearance of these inclusions in the steel. However, one can choose one of the following countermeasures from the stability diagrams; first, for a range of 89-130 ppm Al dissolved in the steel, the stability of the spinel will decrease significantly with increasing Mg content in the steel greater than about 0.001mass% (10ppm). Second, and as previously shown in Table 3, add of several ppm Ca for the formation of liquid inclusions rather than spinel. Figure 11 shows a comparison of the inclusion composition distribuition of different sizes between two typical heats A3 and B3 in the ternary diagrams CaO-SiO 2 -Al 2 O 3 (sample X1-after vacuum) and CaO-Al 2 O 3 -MgO (sample X2-tundish) system.  Two conditions were selected that illustrate the two groups of inclusions. These conditions represent the best and worst scenario, considering steps X1 and X2 ( Figure 11). For the A3 heat, inclusions of CaO and alumina, as well as larger inclusions of silicates, are observed after vacuum. Subsequently in the tundish, most of the inclusions were concentrated in the liquid region (region of low melting point). However for the B3 heat, the inclusions were calcium alumino silicates and alumina after the vacuum. In the tundish, inclusions of spinel, complex and some inclusions in the liquid region are observed. As shown by Figure 11 in sample A3_X2 the MgO is lower than in the condition shown by sample B3_X2, this suggests the formation of MgO (spinel) rich inclusions.
Some typical morphologies of inclusions found in these heats are shown in Figure 12. The types of inclusions observed are based in alumina and calcium aluminate silicate as well as by calcium aluminate silicate with high MgO content.

Conclusions
From analyzing the results of the heats, we conclude: