Assessment of Reactivity between Submerged Entry Nozzle (SEN) and Ultra‐low C Liquid Steel: Comparison between Oxide‐Based SEN and Oxide‐Less SEN

Submerged entry nozzle (SEN) clogging is investigated with an emphasis on the interfacial reaction between the nozzle refractory and the liquid steel. Within the context of carbothermically produced CO(g) being the main cause of the early stage SEN clogging, various SEN refractory materials are assessed in view of their impact on the initial clog deposit formation. A series of high‐temperature experiments are carried out to observe the interfacial reaction between the SEN refractory materials and the liquid steel: three oxide‐based and one oxide‐less refractory materials are employed. Ti‐free and Ti‐added ultra‐low C steel are reacted with these refractory materials. The reactivity of the refractories is assessed by 1) the morphology, composition, and thickness of the initial clog deposit and 2) the evolution of steel compositions. Decreasing SiO2 content in “oxide‐based” refractory is recommended. The “oxide‐less” refractory composed of AlON–BN–AlN is shown to be promising to suppress the early stage of SEN clogging. The present results would provide directions for designing the SEN refractory constitution to suppress the SEN clogging.

DOI: 10.1002/srin.202400025Submerged entry nozzle (SEN) clogging is investigated with an emphasis on the interfacial reaction between the nozzle refractory and the liquid steel.Within the context of carbothermically produced CO(g) being the main cause of the early stage SEN clogging, various SEN refractory materials are assessed in view of their impact on the initial clog deposit formation.A series of high-temperature experiments are carried out to observe the interfacial reaction between the SEN refractory materials and the liquid steel: three oxide-based and one oxide-less refractory materials are employed.Ti-free and Ti-added ultra-low C steel are reacted with these refractory materials.The reactivity of the refractories is assessed by 1) the morphology, composition, and thickness of the initial clog deposit and 2) the evolution of steel compositions.Decreasing SiO 2 content in "oxide-based" refractory is recommended.The "oxide-less" refractory composed of AlON-BN-AlN is shown to be promising to suppress the early stage of SEN clogging.The present results would provide directions for designing the SEN refractory constitution to suppress the SEN clogging.
as a result of Al deoxidation. [3,20]Once the network alumina inclusions form, subsequent "build-up" alumina attachment is triggered.This scenario is schematically shown in Figure 1. [3]eaction (3) has been valid in the case of low C Al-killed steel. [4]Lee et al. expanded this mechanism to SEN clogging during continuous casting of ultra-low C (ULC) Al-killed steel where Ti is often added to secure the deep drawing quality of the steel product. [21]A significant build-up of the skull is reported. [17,18,22,23]Reaction (3) was expanded as where the stoichiometry of the reaction is indefinite.Not only Al but also Ti and Fe simultaneously oxidize to form the complex oxide, Fe t O-Al 2 O 3 -TiO x (hereafter referred to as "FAT"), in addition to Al 2 O 3 .The FAT is liquid at the casting temperature; therefore, it adheres to the SEN and the liquid steel, both.Fe t O is then reduced by Al and Ti in the liquid steel, leaving Fe drops inside the clog deposit. [17,18]All these previous investigations reported that CO(g) due to the carbothermic reaction (Reaction (1)) is the origin of the early-stage SEN clogging. [3,17]Therefore, suppressing the CO(g) action is required.The present authors assessed the characteristics of the carbothermic reaction of four SEN refractories (AGS, AG, CL, and CSG), which are based on various oxides with C. [19] Their constitution reproduced from the present authors' previous research is listed in Table 1, [19] except for CSG which is not discussed in the present article.Reaction (1) was confirmed.The extent of CO(g) emission varied by the constitution of each refractory.In the light of the stoichiometry of Reaction (1), the following index was derived [19] where n i is the number of moles of i in SEN refractories.It was shown that Γ was linearly proportional to the volume of the emitted CO(g). [19]In order to suppress the early stage SEN clogging by minimizing the volume of the emitted CO(g), it is necessary to decrease either SiO 2 or C in the SEN refractories.
The other countermeasure was also proposed by Lee et al. [13] The emitted CO(g) could be absorbed and left in the SEN by adding B 4 C as "CO absorber."They showed that adding B 4 C in AGS (alumina-graphite-silica mixed) refractory up to 3 mass pct.decreased the thickness of the initial clog deposit (δ product ) from 60 to 10 μm. [13]Although the effect of such inorganic additives on the ceramic material may need additional tests, [24] it was also possible to suppress the oxidation of Fe and Ti (Reaction ( 5)).This is an important step to suppress initial SEN clogging for Ti-added ULC steel continuous casting.
Another countermeasure could be blocking the cause of the carbothermic reaction (Reaction (1)) at its source.The present authors reported that the oxide-less SEN refractory composed of AlON-BN-AlN did not emit CO(g) but emitted only N 2 (g).Moreover, the volume of the emitted N 2 was significantly less than that from any of the oxide-based refractories (2-14 times less depending on the refractory types). [19]This was very promising that the oxide-less SEN refractory could suppress the early stage SEN clogging by suppressing the network alumina formation (Reaction (3)) and the FAT formation (Reaction ( 5)).The present article reports the subsequent results of the interfacial reaction between liquid steel and SEN refractories in order to reveal the role of SEN's components on the initial clog deposit development.Three oxide-based refractories and one oxide-less refractory used in the present authors' previous research [19] were employed and were reacted with Ti-free ULC steel and Ti-added ULC steel by a finger-rotating method.The interfacial reaction was assessed in view of the interfacial reaction products (morphology, composition, and thickness) and the evolution of the liquid steel compositions.

Experimental Section
The present experiments were carried out to simulate the interface of the inner wall of an SEN when liquid steel was charged into a casting mold.A finger-rotating method was employed by rotating immersed SEN refractory in the liquid steel.A schematic  [3] Copyright 2021, The Authors.Published by Springer Nature.figure of the experimental apparatus is shown in Figure 2. The experiments were designed to simulate the formation of "network" alumina or "FAT" corresponding to the early stage of clogging (initial layer) and the formation of "build-up" alumina layer corresponding to the late stage (growth).

Sample Preparation
All of the SEN refractories used in the present study were supplied by Chosun Refractories Co. Ltd., Korea (Table 1).They were machined into the form of a bar (10 Â 10 Â 160 (mm)).
The steel sample was prepared by melting 600 g of electrolytic iron in an alumina crucible in an induction melting furnace.A quartz tube (OD 80 Â ID 70 Â H 1000 (mm)) sealed by brass end caps on both sides was used as the reaction chamber for the steel sample preparation and the subsequent refractory-steel reaction.The crucible charged in a graphite susceptor was loaded into an induction melting furnace, which was heated to 1565 °C.The temperature was measured by a B-type thermocouple located beneath the graphite susceptor.To remove oxygen as an impurity in the electrolytic iron, an Ar-4 pct H 2 gas mixture flowed over the liquid iron for 4 h.The O content in the steel decreased to 20 ppm, analyzed by inert gas fusion infrared absorptiometry (LECO-ON836, St-Joseph, MI, USA).Appropriate amounts of Al pellet (5 N grade, Kojundo, Japan) and Ti sponge (99.9 mass pct., Kojundo, Japan) were added to the liquid steel to make Fe-0.03pct Al or Fe-0.03 pct Al-0.07 pct Ti steel samples.This represents Ti-free ULC steel and Ti-added ULC steel, respectively.The melts were allowed to homogenize in the crucible.

Experimental Procedure
The SEN refractory bar was connected to the equipment designed for submerging/rotating by stepping motors through a brass jig.During the reaction between the refractory and steel, the experiment was carried out under the Ar gas atmosphere (500 mL min À1 ) purified by passing through a CaSO 4 column, Cu chips at 500 °C, and Mg chips at 500 °C.After the liquid steel sample was prepared and homogenized (Section 2.1), a small portion of the liquid steel was taken by a quartz tube.This moment was set to t = 0 min.Subsequently, the refractory was immersed in the liquid steel by 40 mm in depth, and was subsequently rotated at a rate of 100 rpm, which could simulate the flow rate of liquid steel in the SEN. [25]The rotation was carried out for 30 min to proceed with the interfacial reaction.During the reaction, the steel samples were periodically taken by a quartz tube (t = 5, 10, 20, and 30 min).After taking the last steel sample at t = 30 min, the refractory was withdrawn from the liquid steel.This test corresponds to Nos. 1-8 in Table 2.
In separate runs, the "build-up" alumina layer was intentionally promoted.To produce the suspending alumina inclusion in liquid steel, 0.29 g of Al pellet and 0.85 g of Fe 2 O 3 powder as additives were prepared, which aimed to produce ≈900 ppm of Al 2 O 3 in 600 g of the melt.The additives were wrapped in an iron foil and placed inside the quartz tube using a magnet prior to the melting of the liquid steel.A quartz tube containing the additives was sealed with a rubber stopper to prevent the infiltration of the atmosphere.The reaction between the refractory and suspending inclusion proceeded in two stages.First, the reaction between the refractory and the liquid steel was carried out to produce the initial reaction product as described in the previous paragraph.After that, the rotation of the refractory was stopped, and the refractory was temporarily withdrawn from the melt.Then, the magnet was removed to drop the additives onto the liquid steel to produce alumina inclusions.As soon as the additives were dropped, the refractory was reimmersed in the melt and was rotated.It was carried out for 10 min.After that, the reaction was terminated, and the refractory was withdrawn from the melt.The evolution of steel compositions and interfacial reaction products were analyzed from the steel samples and the collected refractory, respectively.The detailed procedure can be found elsewhere. [13]The collected refractory was cut to see a crosssection parallel to the rotating direction.It was mounted with an epoxy resin, polished, and gold-coated to observe the interfacial reaction product by using field emission scanning electron microscope (FE-SEM, JEOL-7100 F, JEOL, Tokyo, Japan).Energy dispersive spectrometer (EDS) analysis was carried out by Oxford-X-Max 50 (Oxford Instruments, Oxford, UK) with AZtec 3.1.For the chemical analysis of steel compositions, 2 g of each steel sample was cut and analyzed using inductively coupled plasma-atomic emission spectroscopy (Thermo-Fisher Scientific ICAP 6500, Waltham, MA, USA) for Al and Ti contents and by LECO CS-844 (St.Joseph, MI, USA) for C content.

Interfacial Reaction Between Liquid Steel and SEN Refractory to Simulate Network Alumina or FAT Formation
Figure 3 shows photographs of four SEN refractories used in the present study: (a-d) in the middle row for the refractories before immersion, (e-h) in the upper row for the refractories after immersion in Ti-free steel for 30 min, and (i-l) in the lower row for the refractories after another immersion in Ti-added steel for 30 min, without generating alumina inclusions.All the experimental conditions are listed in Table 2.

Ti-Free ULC Steel: Nos. 1-4
Visual inspection of the refractory surface resulted in the following observations.Figure 3e-g shows gray-colored deposits on the refractory surface.Figure 3h shows a similar deposit, but the covered area is smaller.According to Fukuda et al. [20] the deposits shown in Figure 3e-g were most likely "network" alumina formed by Reaction (3).A typical appearance is depicted in Figure 4, illustrating a typical step for the sample analysis, after the reaction between a refractory and Ti-free ULC steel in this case (No. 3 in Table 2).A small steel drop was attached to the refractory, containing numerous alumina particles were seen.Elemental mapping analysis using EDS allowed for the identification of the steel drop, refractory, and alumina inclusions.
Interface morphologies of the refractories reacted with Ti-free ULC steel for 30 min reveal the steel, inclusions inside the steel, and inclusions attached beneath the steel (Figure 5, Nos.1-4).While Figure 5c,d shows the refractory part, Figure 5a,b does not display the refractory part, which was separated during sample handling.The inclusions were identified as alumina, based on the EDS elemental mapping.The alumina inclusions in Figure 5a-c appear polygonal, with a large size (up to a few tens μm).In contrast, the alumina inclusions in Figure 5d (ALBN refractory) were smaller in size and total number.As reported by the present authors, [19] the oxide-less refractory does not emit CO(g).It is evident that the oxide-less SEN refractory is superior in suppressing Reaction (3) by eliminating Reaction (1).

Ti-Added ULC Steel: Nos. 5-8
Interface morphologies of the refractories reacted with Ti-added ULC steel for 30 min (Figure 6, Nos.5-8) exhibit differences from those of Ti-free ULC steel.The inclusions inside the steel From another perspective, the interface morphologies of the oxide-based refractories at a lower magnification (Â60) revealed an interesting phenomenon: the reaction product on the AG refractory was discontinuous (Figure 7b), whereas on the CL refractory, it was continuous (Figure 7c).In the case of AGS, the continuity of the reaction product was observed to be in between the previous two cases.Sasai and Mizukami reported that the reaction product on an AGS-type refractory with Al-killed steel was dense and continuous, while that with Ti-killed steel, it was discontinuous and porous. [8]The present study demonstrated that different types of SEN refractory resulted in different morphologies, even with the same type of molten steel.It is speculated that the initial reaction product (alumina and FAT [9] ) exhibited a preferential wetting tendency to a specific part of the AG refractory used in the present study, potentially causing the discontinuous reaction product layer as shown in Figure 7b.However, this speculation requires further validation.Although the validation is beyond the scope of the present study, this phenomenon can impact various aspects of the interfacial reaction between the SEN refractories and the molten steel, with the direct contacting area between the AG refractory and the molten steel being the largest, followed by AGS and CL.This will be discussed in Section 3.3.2.To simulate the "build-up" alumina layer, separate runs were conducted using AGS, AG, CL, or ALBN refractory with Ti-added ULC steel.The interfacial morphology of no. 9 (AGS) is shown in Figure 8a, revealing two layers: an alumina layer on top of an Al-Ti oxide layer.According to thermodynamic analysis to estimate possible inclusion formation, the addition of Al to the present liquid steel sample (Fe-0.03pct Al-0.07 pct Ti) should produce alumina as long as there is an appropriate amount of O source ("Region A" in Figure 9 [26] ).The alumina layers shown in Figure 8a are believed to be alumina produced by adding Al and Fe 2 O 3 , which was intentionally added to generate the "buildup" alumina.No. 10 (AG) and No. 11 (CL) exhibited similar behavior.
However, it is interesting to note that an Al-Ti oxide layer was also observed in Figure 8b (No. 12, ALBN with the "build-up" alumina generation), which was not seen in Figure 6d (No. 8, ALBN refractory without the "build-up" alumina generation).As the ALBN refractory did not reoxidize the liquid steel, it was likely formed by excess reoxidation due to the addition of Fe 2 O 3 to promote "build-up" alumina.Therefore, the Al-Ti oxide layer shown in Figure 8a (No. 9, AGS with the "build-up" alumina generation) might also be formed partly by the addition of Fe 2 O 3 .Nevertheless, the number density of the alumina inclusions seen inside the steel in Figure 8a,b appears higher than those in Figure 6a,d.Therefore, the alumina found in Figure 8a,b was likely formed by the deoxidation of Al, and the alumina layer on top of the Al-Ti oxide layer seems to be the "build-up" alumina.
Cases Nos.4-8 also demonstrated that the oxide-less SEN refractory shows promise in suppressing network alumina or FAT growth, which is responsible for the early stage SEN clogging. [3,18]However, if the liquid steel containing Ti undergoes considerable reoxidation, as was for the case for No. 12, Ti in the liquid steel can be oxidized with Al (also with Fe) to form the Al-Ti oxide layer.Then, it could become the site for further development of the alumina layer.This underscores the importance of not only controlling the SEN constitution but also managing the cleanliness of liquid steel in preventing SEN clogging.

Evaluation of Refractory's Reactivity
To evaluate the reactivity of each refractory with liquid steel, two types of experimental results were analyzed: the thickness of the reaction product between the refractory and the steel and the chemical composition change of the liquid steel during the reaction with the refractory.

Thickness of Reaction Product
The thickness of the reaction product (δ product ) was measured, as it could correspond to the extent of clog deposit growth during the continuous casting process.Among all refractory samples obtained in the present study, those shown in Figure 5 for Ti-free ULC steel were not used because the reaction product layer was not clearly defined.Therefore, the refractory samples (AGS, AG, CL, and ALBN) reacted with Ti-added ULC steel were used.In the case of Nos.5-8, δ product of each refractory, which had reacted with liquid steel for 30 min, is shown in Figure 10 by blue-open square.In the case of Nos.9-12, δ product of each refractory, which had reacted with liquid steel for 40 min (30 min of reaction between the refractory and the liquid steel and 10 min of additional reaction with the build-up alumina), is shown in Figure 10 by red-filled circles.In this case, where δ productÀ1 is due to Reaction (5) (FAT), and δ productÀ2 is due to build-up alumina.
relatively free from errors that might be caused by subjective decisions during the judgment of the boundary of the reaction layer.k-means clustering is a method of vector quantization, originally developed for partitioning many data into a small set of clusters ("k" sets). [27,28]Each data is eventually allocated to one of the k clusters with the nearest mean (cluster centroids).Each centroid is regarded as a prototype of a subset of the data allocated to each center.The k-means clustering minimizes within-cluster variances, requiring efficient algorithms to converge quickly to a minimum. [29]In the present study, the SEM image was digitalized and the brightness of each pixel was stored as raw data.Along the direction passing the interfaces of the steel, the reaction product, and the resin, the brightness data were obtained.By setting k = 3 (steel, reaction product, and resin), all the pixel data along the direction were quantized into three different clusters.The thickness of a clustered dataset corresponding to the reaction product was obtained, which corresponds to the thickness of the reaction product layer.For further detailed procedure, refer to ref. [29] Figure 10 shows the δ product ð¼ δ productÀ1 Þ at t = 30 min by Ti-added ULC steel without the build-up inclusion (Nos.5-8).Because the reaction product developed irregularly, the uncertainty limit of the δ product is noticeable.Nevertheless, an attempt was made to measure the thickness over a wide range to obtain an average thickness.This was more significant when the reaction product was discontinuous, such as that shown in Figure 7b.On average δ product varied from 60 to 80 μm for the oxide-based (AGS, AG, and CL) refractories.In contrast, δ product for the oxideless (ALBN) refractory was significantly lower (≈5 μm).The uncertainty was also low: the reaction product growth was limited.This result can be expected from Figure 6.
Figure 10 also shows the δ product ð¼ δ productÀ1 þ δ productÀ2 Þ at t = 40 min by Ti-added ULC steel after the build-up inclusion was generated (Nos.9-12).There was no significant difference in δ productÀ2 ð¼ δ product À δ productÀ1 Þ among the oxide-based refractories.However, δ product of the oxide-less refractories was also significantly lower than that of the oxide-based refractories.This should be mostly due to the lower δ productÀ1 .

Evolution of Steel Chemistry
During the immersion-rotation of the refractory, the liquid steel (Ti-free or Ti-added ULC steels) reacted with the refractory.Reaction (3) and ( 5) were expected.The C content in the liquid steel would increase gradually during the reactions.Figure 11 shows the measured C content in the Ti-free ULC steel and in the Ti-added ULC steel; therefore, the C dissolution was confirmed.There are two sources of the increase of C content: 1) dissociation of CO(g) in the liquid steel [30][31][32][33] and 2) direct dissolution of graphite in the refractory into the liquid steel. [34]The CO(g) is relevant to the reaction product growth; therefore, it was necessary to selectively determine the C content increase by the dissociation of CO(g).However, distinguishing these contributions to the C content increase was difficult.Therefore, it is not a suitable indicator for evaluating the reactivity of the refractory.
It is to be noted that the C content in the case of Ti-added steel increased slower and less.It was most likely that the C (either from CO(g) or from graphite, or both) reacted with Fe t O in the FAT, thereby not dissolving in the Ti-added liquid steel.Reduction of Fe t O was confirmed in the previous investigation by Lee et al. [18] This usually leaves small Fe droplets inside the reduced FAT (as shown in the Al 2 O 3 -TiO x layer in Figure 6).
The C content in the Ti-added steel reacted with CL refractory was lower than that with either of AGS and AG refractories.The present authors already showed that AG refractory emitted the least CO(g) and AGS emitted the most CO(g) by pyrolysis at high temperature, among AG, AGS, and CL refractories. [19]Therefore, the fact that the C content from the AG refractory was higher than the C content from the CL refractory may provide a conclusion that the majority of C dissolution from the AG refractory was due to direct C(s) dissolution from the AG refractory instead  of CO(g) emission.The discontinuous reaction product (Figure 7b) can partly support this conclusion.The similar level of C content in the Ti-added steel reacted with AGS refractory compared to that with AG refractory is therefore the result of more CO(g) emission [19] and less discontinuity of the reaction product (Figure 7a).
Instead of relying on the C content, the measured Al and Ti contents in the liquid steel were manipulated to derive the apparent rate constants k i (where i = Al, Ti)) from the following simple rate equation k i was obtained by fitting the measured data to the following equation, which was the integrated form of Equation ( 8) In the case of Al, the soluble Al content was used.Figure 12a shows the k Al in the case of Ti-free ULC steel (Nos.1-4), and Figure 12b shows, k Al and k Ti in the case of Ti-added ULC steel (Nos.5-8).k Al and k Ti obtained from both Ti-free and Ti-added ULC steel reacted with ALBN were significantly lower than those in the other cases.Among the cases of oxide-based refractories, AG refractory resulted in lower k Al and k Ti ; therefore, the reactivity of the AG refractory was lower than the others.This is consistent with the finding that AG refractory emitted the lowest amount of CO(g) during pyrolysis of the four oxide-based SEN refractories. [19]

Countermeasures of SEN Clogging during Continuous Casting of ULC Steel
[37][38] In the context of the SEN's carbothermic reaction, the emitted CO(g) plays a critical role (Figure 13a).High-C steel is known to be inert to the development of clog deposits.Fukuda et al. confirmed that high-C Al-free steel (4.0 pct-0.001pct Al) in contact with an AGS-type SEN refractory (49.3 pct Al 2 O 3 -10.9pct.SiO 2 -33.5 pct (SiC þ C), etc.) resulted in virtually no alumina on the surface of the refractory.However, low-C Al-free steel (0.2 pct C-0.001 pct Al) in contact with the same refractory resulted in an alumina deposit on the surface of the refractory. [20]This could be attributed to the fact that Reaction (3) could not proceed effectively because of the high C content in the steel (Figure 13c).On the other hand, low C content in steel would provide favorable conditions for the same reaction (Figure 13a   contact with the low-C Al-free steel also limited the alumina deposit development (Figure 13d). [20]This is due to a limited carbothermic reaction similar to Reaction (1).As the last case showed some limited alumina deposit growth, they proposed the following carbothermic reaction [20] Al 2 O 3 ðs; SENÞ þ 2Cðs; SENÞ ¼ Al 2 OðgÞ þ 2COðgÞ (10)   and Al 2 O(g) and CO(g) were the source of network alumina generation.Decreasing C in the SEN would result in a similar effect (Figure 13e).][41] The presence of Ti in ULC steel is not the only cause of the severe SEN clogging, but a coupled reaction with O in the system causes reoxidation more actively. [42,43]Recently, Lee et al. proposed a likely mechanism for the early stage SEN clogging where Reaction (1) and ( 5) were the major sources of the initial clog deposits. [9]The main cause was CO(g) emission from SEN due to the carbothermic reaction (Reaction (1), Figure 13b).Two countermeasures were already mentioned in Section 1: decreasing SiO 2 content or C content in SEN refractory [19] and absorbing the emitted CO(g) by the CO absorber. [13]he present study focused on the removal of the origin of the carbothermic reaction by replacing C and oxides (SiO 2 , Al 2 O 3 ) with non-oxide type materials: nitride (AlN and BN) and oxynitride (AlON).The present authors' previous study showed that the mixture of AlON-BN-AlN did not emit the oxidizing gas such as CO(g), both by thermodynamic analysis and by pyrolysis with in situ gas analysis using a quadrupole mass spectrometer. [19]Only N 2 (g) was detected as the emitted gas from the SEN refractory.Therefore, the reoxidation could be suppressed (Figure 13f ).In the present study, the oxide-less refractory was shown to be effective in suppressing the initial clog deposit development, most noticeable in Figure 12.It is therefore concluded that the present "oxide-less" SEN refractory is a promising ceramic material for suppressing SEN clogging during continuous casting.

Conclusion
In the context of preventing SEN clogging during the continuous casting of liquid steel, various SEN refractory materials were assessed.Three oxide-based refractories (AGS, AG, and CL) and one oxide-less refractory (ALBN) were employed.Previous investigations aimed at identifying gases emitted from the refractory at high temperatures revealed that the oxide-based refractories mainly emitted CO(g) while the oxide-less refractory did not emit CO(g). [19]In the present study, these SEN refractories were reacted with liquid steel (Ti-free ULC steel or Ti-added ULC steel) at 1565 °C using the finger rotating method.The three oxidebased SEN refractories generated noticeable oxide layers on the surface of the refractory: alumina (with Ti-free ULC steel) or alumina and Al-Ti oxide double layer (with Ti-added ULC steel).On the other hand, the oxide-less SEN refractory generated only a thinner alumina layer, regardless of the presence of Ti in the liquid steel.The thickness of the reaction product was measured using the k-means clustering algorithm: [29] the oxide-less refractory showed a significantly thinner layer than the oxidebased refractory.It was found that suppressing CO(g) emission due to the carbothermic reaction (Reaction (1)) is one of the critical factors in preventing SEN clogging, at least for the early stage SEN clogging. [3]New SEN refractory compositions are suggested, which are free from the carbothermic reaction, such as the ALBN refractory investigated in the present study.

Figure 1 .
Figure 1.Concept of two-stage clogging model.Reproduced under the terms of the Creative Commons CC BY license.[3]Copyright 2021, The Authors.Published by Springer Nature.

Figure 2 .
Figure 2. Schematic figure of the experimental apparatus for the reaction between the refractory and liquid steel.

Figure 3 .Figure 4 .
Figure 3. Photographs of refractories investigated in the present study: a-d) before the reaction with liquid steel, e-h) after 30 min reaction with Ti-free ULC steel, and i) after 30 min reaction with Ti-added ULC steel.a,e,i) AGS, b,f,j) AG, c,g,k) CL, and d,h,l) ALBN refractories.

Figure 8 .
Figure 8. Reaction interface between SEN refractories and Ti-added ULC steel with generated "build-up" alumina inclusions reacted for 40 min at 1565 °C: a) AGS and b) ALBN.

Figure 11 .
Figure 11.C content ([ppm C]) change in ULC steel during the reaction with SEN refractories.

Figure 12 .
Figure 12.Apparent rate constant of metal elements in liquid steel reacted with various SEN refractories: a) Al (k Al ) in Ti-free ULC steel and b) Al (k Al ) and Ti (k Ai ) in Ti-added ULC steel.

Figure 13 .
Figure 13.Schematic figure of various nozzle clogging mechanisms in the view of chemical reaction at the interface between SEN and steel.a) ULC steel, b) Ti-added ULC steel, c) High-C steel, d) SiO2-less SEN, e) C-less SEN and f ) Oxide-less SEN.

Table 1 .
Constitution of the SEN refractories used in the present study.

Table 2 .
Experimental conditions used in the present study."Alumina inclusion" means whether alumina inclusions were intentionally generated or not (see Section 2.2).