2014 Volume 54 Issue 12 Pages 2813-2820
In the present study, inclusions in calcium-treated steel after RH treatment, in the tundish and in bloom were studied. Only two types of inclusions were detected in all steel samples, namely liquid calcium aluminate inclusions and inclusions of two phases with spinel in the center surrounded by the liquid calcium aluminate. The attachment of the inclusions on the inner surface of SEN was investigated for two types of refractory materials. The results indicated that liquid calcium aluminate inclusions could attach on the wall of SEN, when the refractory had big grain size and big cavities on the surface. On the other hand, tiny grain size and smooth surface of the SEN showed no attachment of the inclusions. The different behaviors of the two types of SEN were well explained using the results of flow calculation in the nozzle. The mechanism of the attachment was also discussed based on the experimental results and the CFD calculation. The tiny nodules formed on the surface of the inner nozzle due to inclusion attachment could be a source of macro inclusions.
To improve the cleanliness of steel and control of steelmaking process, a lot of attention has been paid to the inclusion attachment on submerged entry nozzle (SEN), which may cause clogging and affect castability and steel quality.
Many researchers1,2,3,4,5,6,7,8,9,10,11,12,13,14) investigated the clogging behavior and obtained valuable information. Clogging was classified into four types by Thomas et al.1,2) according to its origins: deoxidation products, solidified steel, complex oxides and reaction products. Several investigators analyzed the composition of clogging during the casting of aluminum-killed steel, and found that alumina was the main phase. Some of them concluded that these alumina inclusions formed by deoxidation were the main reason for clogging,3,4,5,6,7) while some found that the reaction between the steel and the nozzle refractory could also cause clogging.8,9,10,11) Besides, reoxidation were also considered as a source of clogging.12,13,14) However, most researchers believed that the attachment of the solid alumina inclusions already presented in the steel before casting was the primary source of SEN clogging.1,15)
Researchers have been trying to develop new techniques to change the solid inclusions into liquid, because most scholars believed that liquid inclusions would not attach on SEN. Calcium treatment is one of the methods. It is commonly used as an efficient way to reduce clogging. Actually, sometimes clogging still occurred after calcium treatment in industrial practice. Different complex oxides were in fact found to attach on the surface of SEN, such as solid calcium aluminates and calcium sulfides.16,17,18,19) The researchers believed that the solid inclusions caused by the non-suitable calcium treatment were the main reason of clogging. The existence of solid calcium aluminates, such as CaO·2Al2O3 and CaO·6Al2O3, suggested that more calcium was needed to get liquid oxide inclusions in liquid steel. However, to the authors’ knowledge attachment usually appeared on the inner wall of the SEN for calcium-treated steel in some steel plants. The steelmakers attempted to control the calcium content to reduce them, but got unsatisfactory results. Preliminary investigation indicated that the attachment was in fact liquid calcium aluminate.
So far, there is no report about the attachment of liquid calcium aluminate inclusions on the wall of SEN. The present work investigates the possibility of attachment of liquid calcium aluminate on the inner wall of SEN. The second objective of the work is to study the main factor(s) that result(s) in the attachment of liquid inclusion.
Industrial study was carried out in the steel plant of Xingtai Iron & Steel Co. Ltd focusing on the steel grade SCM435. This steel grade was produced by the process route: blast furnace → hot metal desulfurization → BOF (80t) steelmaking → LF (80t) refining → RH (80t) refining → Bloom continuous casting. The details of the process could be found in a previous publication.20) It is important to mention that calcium wire was fed into ladle for calcium treatment after RH degassing. The addition of calcium was followed by 10 minutes of soft blowing. Thereafter, the ladle was sent to the continuous caster for casting.
The typical composition of SCM435 steel is listed in Table 1. The composition range of the ladle slag along with the compositions of tundish powder as well as mold powder are presented in Table 2. Two different types of SEN (Type-A and Type-B) were used during the casting. The total chemical compositions of the refractories of the two types were very similar (both consist mainly of alumina and graphite). On the other hand, the morphologies of the two types of materials were very different. Detailed description about the difference between Type-A and Type-B is given in the result part.
Steel samples were taken at different stages of the process, namely (1) before calcium treatment (after RH degassing), (2) after calcium treatment (after RH treatment), (3) in the tundish (positioned at the middle tundish), and (4) in bloom.
After casting, the submerged entry nozzle was cooled in air to room temperature. The nozzle was visualized first, before its inner layer were cut off and prepared for microscopic analyses. A scanning electron microscope (SEM, HITACHI S-3700N) attached with energy dispersive spectrometer (EDS) was employed to study both the morphology and phases in the inner layer of the nozzle.
The inclusions in the steel samples were also analyzed by SEM and EDS. In addition, about 2 kg steel sample of bloom was electrolyzed by Slime Method.21) The obtained macro inclusions were analyzed by SEM and EDS as well.
The total oxygen contents of these steel samples were determined by fusion and infrared absorption method. The nitrogen contents were analyzed by thermal conductimetric method after fusion under an atmosphere of inert gas.
Two types of inclusions are observed throughout the process after RH refining as shown in Table 3. Inclusions of Type-1 consist of only calcium aluminate phase with small amounts of SiO2 and MgO. Type-2 inclusions consist of two phases, namely MgO·Al2O3 spinel in the center surrounded by the calcium aluminate phase. Table 3 shows that these two types of inclusions are the only types detected before calcium treatment, after calcium treatment, in the tundish and in bloom. Figures 1 and 2 present the SEM microphotograph and the element mappings of the two types of inclusions. Note that these types of inclusions are also found in the liquid steel after LF and after RH refining.20)
Element mappings of a typical Type-1 inclusion.
Element mappings of a typical Type-2 inclusion.
Table 4 lists the composition ranges of the calcium aluminate phase in both Type-1 and Type-2 inclusions in tundish and bloom. The table shows that both the SiO2 content and MgO content in calcium aluminate phase are low, only about 1–2 mass% in average. The table also shows that the calcium aluminate phase in Type-2 inclusions is basically the same phase as in Type-1. The composition indicates that the calcium aluminate phase is in the liquid region based on CaO–Al2O3 phase diagram.22) The liquid nature of the calcium aluminate phase in both Type-1 and Type-2 is further confirmed by the globular shape of the two types of inclusions shown in Figs. 1 and 2.
The inclusions of both types found in the samples are usually in the size range of smaller than 10 μm. However, macro inclusions are also detected when electrolysis technique is used. Figure 3 presents an example of the macro inclusions. The composition range of the calcium aluminate phase is Al2O3 62.6–63.6 mass%, CaO 27.1–33.0 mass% and MgO less than 2.2 mass% as well as SiO2 1.2–6.7 mass%. These macro inclusions are usually larger than 100 μm. The inclusion shown in Fig. 3 has a diameter of about 300 μm.
SEM image of macro calcium aluminate inclusion.
The total oxygen and nitrogen contents after RH and in bloom are given in Table 5. It can be seen that the total oxygen contents are relatively low with the average value around 8 ppm. The table also shows that the total oxygen content in bloom are somewhat lower than that after RH treatment in most heats. In contrast to the case of oxygen, an increase of nitrogen content is evidently seen when comparing the total nitrogen content in the bloom and in the liquid steel after RH. The average nitrogen pick-up is usually above 10 ppm. However, in some heats, the nitrogen pick-up is substantial, e.g. Heat 6.
Figures 4(a) and 4(b) present the SEM images of the inner surfaces of nozzle Type-A and nozzle Type-B, respectively. Both refractories are alumina based with graphite addition (about 20 mass%) and silica as impurity. The grain sizes of the oxide particles in the region at the surface area of Type-A nozzle are large, typically a few hundred micrometers. The refractory has also many big holes, hundred micrometers in size. On the other hand, the grain size in the Type-B refractory is small. Consequently, no big hole are seen in the material. Though the total compositions of the refractory materials are very similar, the morphologies of the inner surfaces of the two types of SEN are considerably different.
SEM images of different SEN inner walls: (a) Type-A; (b) Type-B.
Overviews of the used SENs are presented in Figs. 5(a) and 5(b), for Type-A and Type-B respectively. In the case of Type-A, an attached layer is evidently seen. Since the SEN was cooled down to room temperature in air after usage, this attached layer had separated from the refectory due to its different physical properties from the bulk of the refractory. In strong contrast to Type-A, no attached layer is noticed on the inner surface of the Type-B SEN. The inner surface is pretty smooth after usage.
Photos of used SENs: (a) Type-A; (b) Type B.
Figure 5(a) also shows that the inner surface of the attached layer in Type-A SEN is extremely uneven. To illustrate it clearly, the vertical cross section and horizontal cross section of the used SEN are presented in Figs. 6(a) and 6(b), respectively. Figures 6(a) and 6(b) show clearly the roughness of the surface. A careful visualization indicates that the surface consists of super-cooled oxide liquid phase together with solid oxide particles. Metal droplets are also seen here and there embedded in the frozen liquid. Note that some vacancies are found in the frozen liquid oxide. These holes could be either due to the volume change associated with solidification or due to formation of CO gas.
Vertical (a) and cross (b) sections of used Type-A SEN.
Figure 7 presents the element mappings of the surface area in the attached layer. The boundary between the layer and mounting resin in these mappings was the boundary between the attachment layer and liquid steel during the continuous casting process. It can be seen in Fig. 7 that the surface area consists of mostly calcium aluminate phases. The calcium aluminate phases contain also small amounts of SiO2. Inside these calcium aluminate phases, many tiny MgO rich islands are seen. It is interesting to mention that a very thin Na2O rich layer is detected at the surface of the attached layer.
Element mappings of attachment layer.
A number of phases are detected in this attached layer. Figure 8(a) presents the SEM microphotograph of the area near the surface. The phases along with their typical compositions are presented in Table 6. In general, two calcium aluminate phases are found. One of them (Phase-I in the table, light color in Fig. 8) is super-cooled liquid; and the second one (Phase-II, dark color in Fig. 8) is a solid calcium aluminate, possibly CaO·Al2O3. The third phase (Phase-III, darker color in Fig. 8) found in the layer is spinel phase, MgO·Al2O3. The positions where the EDS analyses are made are marked in Fig. 8(a). The numbers of the positions correspond to the numbers in Table 6. The shape of Phase-II confirms that this phase is solid. In view of the importance of MgO distribution in the attached layer in understanding the attachment process, Fig. 8(b) presents the SEM microphotograph of the area containing Phase-III (spinel) with high magnification. It can be seen that the spinel phase is surrounded by Phase-I and Phase-II, and its size is less than 10 μm.
SEM images of attachment layer.
The total chemical composition in the layer varies with the position. Several area analyses are made in the layer. The compositions of these areas are given in Table 6.
It is extremely important to identify the source of the attachment. Dekkers23) reports that the mold powder can be entrapped into the nozzle. As shown in Table 2, the mold powder is based on CaO–SiO2 system with the Al2O3 content of less than 5 mass% and fluorine. On the other hand, the attached layer consists of mostly calcium aluminates without any F. The absence of F and the big difference in composition rules out the possibility that the layer is formed due to entrapped mold powder.
Figures 7 and 8 as well as Table 6 evidently show that the surface region of the attached layer consists of mainly calcium aluminate phases and spinel particles. The fact that the refractory of the SEN has only trace of MgO and not any spinel phase rules out any possibility of the refractory being the suppler of the spinel phase. It is well known that the spinel is not stable, when there is a trace of calcium in liquid steel. The dissolved Mg content therefore would not result in the formation of the spinel phase from the liquid metal. Moreover, Table 2 shows evidently that the spinel phase cannot be from the ladle slag, tundish powder or mold powder. The remaining two possibilities for the formation of the attached layer are (1) inclusion from tundish and (2) reoxidation product.
As shown in Fig. 2, Type-2 inclusion consists of two phases, with spinel phase in the center and calcium aluminate around. If inclusions of this type are attached on the surface of the nozzle, the MgO·Al2O3 spinel will be distributed in the calcium aluminate phase. This situation is evidently seen in Fig. 8(b). Note that the nozzle was cooled in air after usage. It is expected that the liquid calcium aluminate would go through phase transformation during cooling, leading to therefore the precipitation of the solid calcium aluminate phase (Phase-II). It is important to mention that the liquid calcium aluminate phase found in the attached layer has much more SiO2 content than the inclusions in the tundish. The considerably higher SiO2 could be well explained by the dissolution of both SiO2 and mullite phase originally in the refractory into the liquid calcium aluminate phase originating from the inclusions. An examination of Tables 4 and 6 shows that the mass ratio of Al2O3/CaO of the liquid calcium aluminate phase in the inclusion in tundish compares reasonably well with the ratio of the area analysis, although this ratio varies with position. This comparison would further support the reasoning that the layer found on the inner surface of the nozzle of Type-A is formed by the attachment of liquid calcium aluminate. A detailed discussion about the dissolution mechanism will be given in the later section.
As shown in Table 1, the steel contains about 0.03 mass% aluminum. Formation of alumina particles due to reoxidation (reaction of Eq. (1)) is reported to be an important reason for the attachment of oxide on the inner wall of nozzle.12,13,14)
It is logical to expect that at least some pure alumina particles should be found at the surface zone of the attached layer, if this argument is also true in the present case. However, Figs. 7 and 8 as well as Table 6 ruled out evidently the existence of pure alumina particles at the surface of the attached layer. This observation implies that the reaction of Eq. (1) is not appreciable in the nozzle in the present casting process.
It should be mentioned that there is definitely air dissolution in the steel after RH treatment, as indicated by the increase of nitrogen content (see Table 5). In general, the increase of nitrogen is about 10 ppm after RH treatment. This would correspond to about 2 ppm oxygen pickup. This oxygen pickup would result in reoxidation of dissolved Al. However, due to the great super saturation required for homogenous nucleation of alumina, the nuclei of pure Al2O3 cannot be formed in the nozzle. The existing calcium aluminate inclusions would act as heterogeneous nuclei. Consequently, the reaction of Eq. (1) would take place on the surface of the calcium aluminate inclusions. The formed alumina will further be dissolved in the liquid oxide solutions. The increase of Al2O3 content in the calcium aluminate inclusions from the tundish to the bloom (see Table 4) strongly supports this reasoning. The absence of pure alumina inclusions in the bloom is also in line with this discussion. Similar phenomenon is also reported by Yang and his co-workers.24) The surface of the nozzle could also be favorable seats for aluminum reoxidation by the reaction of Eq. (1). Since the calcium aluminate phase in the surface layer (Table 6) is not saturated by Al2O3, the reoxidation product would therefore be dissolved into the calcium aluminate phase. Hence, the reoxidation of aluminum on the inner surface of the nozzle would not form pure alumina particles as shown in Figs. 7 and 8 as well as Table 6. The above discussion suggests that reoxidation would increase the Al2O3 content in the calcium aluminate phase in attached layer. However, reoxidation cannot be the main source for the formation of the attached layer on the inner surface of Type-A nozzle.
The attachment of inclusions on the surface of inner nozzle has been reported by many researchers.3,4,5,6,7) However, only solid particles are found to attach on the refractory in all these publications. The present study evidently shows that liquid calcium aluminate inclusions can also attach on the refractory surface, e.g. Type-A. It is very important to point out that phase transformation takes place during cooling of the SEN after usage. Solid calcium aluminates would precipitate from liquid oxide solution. This aspect must be considered when analyzing the source of the attachment based on the phases found in the SEN.
As shown in Figs. 5(a) and 5(b), the behaviors of the two types of nozzles are very different. In the case of Type-A, an attached layer is evidently seen, while no attached layer is on the inner surface of the Type-B SEN. It is important to point out that the two types of SEN have very similar chemical compositions. The main difference between these two types of SEN is the grain size and morphology (see Fig. 4). Type-A SEN has a large grain size with obvious holes, while Type-B has a much compacter matrix. The roughness of the surface of Type-A refractory seems to hold the key for the easy attachment. In order to understand this aspect, a simple model calculation is made.
A steady state and 2-D axisymmetric model is employed for this purpose. The calculating domain along with its dimensions is schematically shown in Fig. 9. The continue equation and momentum conservation equations (Navier-Stokes Equations) are considered. To describe the turbulent nature of the flow, the two-equation model k-ε model is employed. Commercial software COMSOL Multi-physics 4.4 is used to carry out the simulation. The boundary conditions are surmised below.
Sketch map of mold and SEN.
(1) At the wall: Wall function is considered.
(2) At the inlet: Velocity is set as 0.7m/s, which is calculated from casting speed.
(3) At the free surface: Symmetry is chosen.
(4) At the outlet: Pressure is set as zero.
The parameters of the k-ε model are referenced from Ref. 25), while the density of liquid steel is considered as 7100 kg/m3 with dynamic viscosity of 0.0055 Pa·s.
In the simulation, three different situations are considered, namely (1) smooth surface, (2) single cavity on the surface and (3) a number of cavities. The results are presented in Figs. 10(a)–10(c), respectively. Note that only a small region from Fig. 9 is presented in Figs. 10(a)–10(c) to illustrate the situation better. This region, as marked in Fig. 9 is near the outlet of the nozzle.
Modelling results of different attachment situations. (a) Smooth surface; (b) Single cavity on the surface; (c) A number of cavities.
It can be seen in Fig. 10(a) that when the inner wall is smooth, there is no backflow in the SEN. On the other hand, backflow would take place when the surface of the wall is not smooth (Figs. 10(b) and 10(c)). More cavities would create more serious backflow. The backflow would give the inclusions more time to stay close to the wall. The longer resistance time of the inclusion in the nozzle and the uneven surface of the refractory would consequently lead to the situation, where the inclusions are stopped by the cavity and attach on the surface. In fact, this process has a nature of self-acceleration. More attached inclusions would result in more unevenness of the surface and therefore better condition for attachment, as illustrated in Figs. 10(b) and 10(c). The results shown in Figs. 10(a)–10(c) explain very well the situations presented in Figs. 5 and 6. The agreement between Figs. 5(b) and 10(a) is very convincing. The surface of Type-B SEN is smooth with small grains and no big cavities. No obvious attachment is expected. The big cavities on the surface of Type-A SEN result in the backflow of the metal and longer resistance of the inclusions and consequently their attachments near the cavities. The previously attached inclusions further increases the roughness of the surface, which accelerates the process of attachment. Figure 10(c) is in good agreement with the photograph of the inner surface in Fig. 6.
The backflow of the liquid metal is even supported by the presence of sodium on the surface of the attached layer shown in Fig. 7. The matrix of the refractory of the SEN contains very little Na. The presence of Na only at the surface of the attached layer strongly suggests sodium is from the liquid metal. Note that the mold powder contains high content of sodium oxide (Table 2). The dissolved aluminum in the metal would reduce the Na2O at the interface between metal and the mold powder according to the reaction shown in Eq. (2).
The backflow of metal, though not strong, would bring the dissolved Na into the nozzle. Since the activity of Na2O in the attached layer is almost zero, sodium would be oxidized following the thermodynamic constraint. The attached layer actually contains Na2O throughout the whole thickness (see Fig. 8 and Table 6). It is somewhat difficult to show the presence of Na2O in the whole layer very clearly by the element mapping in Fig. 7, due to the low concentration. The fact that the surface layer contains bigger amount of Na2O could be explained by instant reaction shown in Eq. (2). The Na2O at the surface would diffuse in the liquid layer, when new oxide inclusions are entrapped by the layer.
As mentioned earlier, the attached layer also contains certain amount of SiO2, which is higher than the SiO2 content in the calcium aluminate inclusions. In order to explain the difference in SiO2 content, the mechanism of attachment of inclusions on the refectory and the reaction between them should be first understood.
On the basis of microscopic investigation in the sample taken from the SEN and the results shown in Fig. 10, the attachment of inclusions on the refractory is schematically shown in Fig. 11. Before casting, the SEN matrix is composited of alumina and graphite with a certain amounts of silica as well as mullite distributed in the matrix, as shown in Fig. 11(a). When the casting starts, the region near the metal-refractory interface is decarburized, plausibly due to carbon dissolution or reduction at steelmaking temperature. A decarburized layer is formed (see Fig. 11(b)). In the case of Type-A SEN, the decarburization would lead to more cavities at the surface. The rougher inner wall increases the opportunity for inclusions attachment, even for the liquid inclusions. As shown in Fig. 2, some of these liquid inclusions contain MgO·Al2O3 spinel in the center. When the liquid inclusions attach on the inner wall of SEN, the liquid inclusions would also bring the spinel islands to the attached layer. The presence of MgO along with Al2O3 in the layer shown in Figs. 7 and 8 evidently supports this reasoning. As discussed above, the attachment is a self-acceleration process, where the attachment increases the roughness of the surface of the wall and therefore enhance further attachment (Figs. 11(c)–11(d)). This process would continue, and thereby resulting in a surface of accretion with tiny nodules (Fig. 11(e)). In the stages shown in Figs. 11(c)–11(e), the liquid aluminate inclusion would dissolve both SiO2 phase and mullite phase. In the SiO2 dissolution process, the silica particles would become smaller and smaller. Finally, all the particles of silica and mullite vanish in the attached layer. The dissolution of the SiO2 containing phases would increase the SiO2 content. Therefore, the composition detected by area analysis in the layer shows higher SiO2 content than that of the calcium aluminate inclusions in the tundish.
Sketch map of inclusion attachment.
The formation of the attached layer would have both positive and negative effect. Since the liquid calcium aluminate inclusions (even the ones with spinel phase in the center) would attach on the nozzle, the total number of inclusions would be reduced. On the other hand, the tiny nodules shown in Fig. 6 could very often be flushed off by the liquid metal and therefor creating macro inclusions. The macro inclusion shown in Fig. 3, which is found in the bloom is very likely formed by this mechanism. A few macro inclusions of this type are enough to disqualify or at least degrade the steel product. In this connection, one can conclude that the smoothness of the inner surface of the SEN is extremely important to avoid the formation of macro inclusions by nozzle attachment, though the attachment does not lead to blocking of the nozzle.
Inclusions in calcium-treated steel after RH treatment, in the tundish and in bloom were studied. Only two types of inclusions were detected in all steel samples, namely liquid calcium aluminate inclusions and inclusions of two phases with spinel in the center surrounded by the liquid calcium aluminate. Two types of SEN were studied. While their compositions were very similar, the surface properties are different. The results indicated that liquid calcium aluminate inclusions could attach on the inner wall of SEN, when the refractory had big grain size and big cavities on the surface. On the other hand, tiny grain size and smooth surface of the SEN showed no attachment of the inclusions. The flow in the SEN was simulated using COMSOL Multi-physics software. The different behaviors of the two types of SENs were well explained using the results of flow calculation in the nozzle. The mechanism of the attachment was also discussed based on the experimental results and the CFD calculation. The tiny nodules formed on the surface of the inner nozzle due to inclusion attachment could be a source of macro inclusions.
The authors are thankful to Mr. Yongchao LI for his help in the industrial trials. Zhiyin DENG expresses his appreciation to China Scholarship Council (CSC) for supporting his study at KTH Royal Institute of Technology. The authors are also grateful for the support of National Natural Science Foundation of China (51134009) and The Specialized Research Fund for the Doctoral Program of Higher Education of China (20110042110010).