Tracing of Deoxidation Products in Ti‐Stabilized Interstitial Free Steels by La and Ce on an Industrial and Laboratory Scale

The continuous casting of Al‐killed Ti‐stabilized interstitial free steels is often affected by clogging. By today, the mechanism behind this phenomenon is not entirely clarified. The active tracing method, which involves the direct addition of rare‐earth elements (REEs), enables tracking of nonmetallic inclusions (NMIs) over process time. Due to the high oxygen affinity of these elements, preexisting NMIs are partially reduced and marked by this tracer. Using scanning electron microscopy with energy‐dispersive spectroscopy, it is possible to differentiate such labeled NMIs from particles formed at later stages in the process by the absence of these tracing elements. Active tracing is used in industrial and laboratory settings to trace preexisting alumina NMIs with La or Ce. In the investigation, it is revealed that the number of small NMIs increases after the addition of REEs, and subsequently, the size increases again after FeTi is alloyed. In both settings, traced and untraced Al–Ti oxides are found. The separation tendency of traced NMIs is studied over time by analyzing the composition of the slags. Furthermore, the impact of deoxidation products on the formation of the clogging layer within the submerged entry nozzle is investigated, indicating that these NMIs contribute to its formation.

to other phases and may act as a subsequent buildup site for NMIs in the steel phase.
One way to minimize clogging is to modify inclusions using a calcium treatment. [11]In this process, the existing alumina population is modified to form liquid or partially liquid calcium aluminates, which show a decreased driving force for agglomeration and adhesion to the casting system.However, this option is only available to a limited extent for ULC and IF steels, as their ladle slag often contains high amounts of wustite (FeO).The interaction between this FeO-rich slag and the steel can notably reduce the calcium yield, leading to an incomplete inclusion modification. [4]Consequently, the types of inclusions formed under these circumstances may further intensify the clogging tendency.
As previously mentioned, the factors influencing clogging can be multifaceted and complex.A significant factor in reducing clogging is determining the actual sources of inclusions that are involved in the formation of the clogging layer.Consequently, countermeasures can be taken to suppress these inclusions formation in advance.As described in the review by Costa e Silva, [12] NMIs continuously interact chemically with their environment (steel melt, slag, and refractory) throughout the entire steelmaking process.This interaction can modify the morphology and composition of these particles to such an extent that tracking these NMI back to their original source is no longer possible.The state-of-the-art tracer method for tracking NMIs over the steelmaking process is the direct addition of rare-earth elements (REEs) to the steel melt.
In addition to using these REEs as tracers, these metals can also be deliberately used for inclusion modification.As described by Li et al. [13] the NMIs can be modified by REEs; their melting points fall below the liquidus temperature of the steel.This can achieve a similar effect as the conventional Ca treatment.The work of Wang et al. [14] can be used as a reference for practical application of the inclusion modification by REEs on an industrial scale.In this experiment, a Ce-Fe alloy was added to a highstrength IF steel during the Ruhrstahl Heraeus (RH) treatment in the secondary metallurgy.The existing alumina inclusions were modified to Ce-containing oxide or oxide-sulfide types during this process.The modification by Ce significantly reduced the effective inclusion size and achieved a spherical inclusion shape.Furthermore, the tendency to cluster was suppressed.
There are also risks associated with using REEs for inclusion modification.According to research by Li et al. [15] inappropriate contents of these elements can form additional solid inclusions, increasing the proneness to clogging.Furthermore, REEs have the potential to react with refractory materials and alter their interfacial properties. [16]or the present study, trials are performed both in an industrial setting and on a laboratory scale.Several Ti-IF heats are alloyed with either La or Ce in the industrial trial.The laboratory trials simulate the production process and track the modification of NMIs.In both cases, steel samples are taken and analyzed regarding cleanness.Since the steel melts are sampled over the processing time, the changes in morphology due to REEs and FeTi addition can be analyzed.In the industrial experiment, tundish and mold slag samples are taken and investigated for their REE content to study the separation tendency of traced deoxidation products.Furthermore, the formed clogging layer on the wall of the SEN is also part of the investigation, which is analyzed with automated scanning electron microscopy with energy-dispersive spectroscopy (SEM/EDS).Thus, it is possible to investigate the involvement of preexisting NMIs in forming the clogging layer.

Ti-IF Steel
The industrial experiments were conducted at the integrated steel mill of voestalpine Stahl GmbH in Linz, Austria.The Ti-IF steel grade was selected for these experiments due to its higher sensitivity to SEN clogging.Table 1 presents an exemplary analysis of a Ti-IF steel sample obtained from a tundish of the steel plant of voestalpine Stahl GmbH.

Input Material for Laboratory Trial
Pure iron was used as a raw material for the laboratory experiments since it has a high purity and a carbon content similar to a Ti-IF steel.The chemical composition of pure iron is shown in Table 2.An O-rich master alloy with around 0.2 wt% O was added for these trials to achieve a comparable oxygen content as slightly before the Al killing, where ≈300 ppm O were in the steel.The prominent inclusions in the raw material were SiO 2 and MnO.The further input materials were Al, La, Ce, and FeTi70, with an O total of 0.18 wt%.

Industrial Trial
The production process for Ti-IF and Ti-ULC steels mainly involved using a basic oxygen furnace, ladle treatment at the ladle  furnace, refining at the RH degasser, and subsequent casting on a continuous slab caster.A schematic illustration of the Ti-IF steel production route, including the REE addition, is shown in Figure 1.
Typically, 180 tons of steel were tapped into the steel ladle, containing ≈600 ppm of oxygen and 300 ppm of carbon.Subsequently, the temperature of the melt was adjusted for later casting, and slag formers were liquified using the ladle furnace.The melt was then transferred to the RH vacuum facility for decarburization and afterward killed with Al.Following a brief RH circulation phase of 6 min, 6.5 kg of metallic La or Ce was added through the vacuum chamber of the RH degasser.The purpose of the REE addition was to mark primary deoxidation products.The employed amount was adopted from other industrial trials documented in the literature. [15,17,18]A compromise was sought to achieve the highest possible REE content within the steel without negatively impacting the castability.Subsequently, another circulation period of 8 min ensured proper homogenization of La or Ce within the melt and promoted inclusion removal, thereby ensuring a high level of cleanness.Toward the end of the treatment, FeTi was alloyed to reduce nitrogen and carbon to a minimum, thus improving these steels' aging resistance.Following this secondary steel treatment, the melt was transferred to the continuous caster.In this particular case, the Ti-IF heats were cast using a singlestrand continuous casting machine with a tundish capacity of 32 tons.As illustrated in Figure 2a, the exceptional inclusion treatment with Ce or La was carried out on an entire Ti-IF casting sequence.The sequence consisted of 6 heats, with the first and last not receiving any REE treatment.La was used as a tracer for the second and fourth heat, while Ce was an alternative tracer for the third and fifth heat.This alternating addition of REEs was intended to assess the individual contribution of each melt, particularly in the formation of clogging layers in the SEN.
The utilized SEN was changed multiple times throughout the casting sequence, as indicated by the local minima of the stopper rod position in Figure 2b.The first SEN (SEN 1) was picked to investigate the buildup of clogging further.As depicted in Figure 2c, both tundish and mold slag samples were taken throughout the experiment.Tundish slag samples were taken approximately every 10 min after each ladle exchange operation.Mold slag sampling primarily focused on the 20th and 40th  minutes of casting from each respective heat.Black diamonds denote tundish slag samples, while mold slag samples are represented by white triangles in Figure 2c.
Additional standard lollipop samples were taken within the steel production to analyze the modification of preexisting deoxidation products over the process.The initial state of the deoxidation products before the REE treatment was assessed by taking a steel sample immediately after the Al deoxidation.Further sampling was done after the REE treatment and the addition of FeTi alloy to monitor the modifications caused by these elements.Additionally, the melt in the tundish was sampled to investigate changes of the NMIs after an extended reaction time.

Laboratory Trial
The two laboratory trials were performed under an inert atmosphere in a resistance-heated Tammann-type furnace (Ruhrstrat HRTK 32 Sond.).The schematic of the used furnace is illustrated in Figure 3.A detailed description of this furnace can be found in Dorrer et al. [8] The benefit of this aggregate was that sampling and alloying were possible during the experiment.Hence, the changes in the inclusion landscape after REE and FeTi addition can be studied.
The laboratory trials were based on the industrial experiment.Thus, the time frames between deoxidation by Al, the REE addition, and the FeTi alloying were similar.A detailed schedule of the laboratory trials, including sampling and alloying, is shown in Figure 4.In both experiments, 300 g of pure iron and 10 g of the O-rich master alloy were heated up to 1600 °C.Adding an O-rich master alloy was essential to achieve a comparable O content of ≈250 ppm as in industry at the RH degasser directly after the vacuum decarburization.The target value for Al was 400 ppm in the remaining melt.A thermodynamic calculation was necessary to determine the amount of La and Ce for an effective tracing success.For this melt, 160 ppm of REEs were required.The recovery rate for the REEs of 30% was assumed based on the findings of Thiele et al. [19] where different alloying concepts for REEs were tested on a laboratory scale in a highfrequency remelting furnace.The REEs were wrapped in an Al foil to prevent oxidation.The aspired Ti/Al ratio was 1.0.Since this aggregate had only a weak bath movement, stirring with a quartz rod was applied after Al, La/Ce, and FeTi addition.

Methods of Analysis
The standard lollipop samples taken during the industrial experiment were analyzed for bulk chemistry and the content of REEs using spark optical emission spectroscopy (Spark-OES).For this purpose, the ARL iSpark 8880 instrument (Thermo Fisher Scientific, Waltham, USA) was used.The composition Figure 3. Schematic illustration of resistance-heated Tammann-type furnace (modified).Reproduced under the terms of the CC-BY license. [8]Copyright 2019, The Authors.Published by Wiley-VCH GmbH.  of the different slag samples obtained from the ladle, tundish, and mold was determined using inductively coupled plasma OES (ICP-OES, SPECTRO ARCOS ICP-OES, SPECTRO Analytical Instruments GmbH, Kleve, Germany).Before the slag analysis, the samples were ground, dried, and homogenized.An open digestion process utilizing HCl, HNO 3 , and H 2 O 2 was employed to dissolve the slag components in an aqueous solution.
The standard lollipop samples and the steel samples obtained from the laboratory trials were analyzed for their micro-cleanness to study the modification of the deoxidation products during various process steps.For the characterization of NMIs, a JEOL 7200F field-emission SEM (JEOL Germany GmbH, Freising, Germany) equipped with a 100 mm 2 silicon drift detector (SDD) EDS detector (Oxford Instruments Ultim Max 100; Oxford Instruments GmbH NanoAnalysis, Wiesbaden, Germany) was used.Only particles with an equivalent circle diameter (ECD) greater than 1 μm were detected at the automated measurement.The duration of EDS analysis for each particle was set to 1 s.This method enabled an evaluation of the existing inclusion population regarding chemical and morphological characteristics.
For the inclusion evaluation of this REE-traced steel, using an additional MATLAB tool for the specific assessment of REEcontaining inclusions was necessary.Due to the backscattered electron (BSE) coefficient being a function of the mean atomic number of the interacted volume, [20] REE-containing phases appeared brighter in the BSE imaging mode compared to the surrounding Fe-based matrix.In contrast, regular oxide particles exhibited lower BSE yield than the surrounding steel, resulting in a lower BSE intensity (darker appearance in the BSE image) than the matrix.The observed inclusions often have a heterogeneous structure, with regions containing higher REE content and phases with no REEs.As a result, heterogeneous REE-traced NMIs often have multiple contrasting gray values, which can be mistakenly identified as individual particles by standard software.A software was developed to reassemble these particles into a single multiphase inclusion to address this issue.The principle of the recombination software has been previously published by Thiele et al. [21] As mentioned earlier, the formed clogging layer structure within the used SEN was also examined in detail.For this purpose, a section from the central part of the straight area of each SEN was cut out using an angle grinder, which was then halved (Figure 5a).Each sample of the SEN refractory material and the corresponding clogging layer counterpart were further cut to an appropriate size (Figure 5b), washed with ethanol, and then dried.Subsequently, the sample was embedded using a Sn-Bi alloy, to prevent the separation of the inner wall of the SEN and the clogging material during the following polishing.The polishing procedure of the sample was performed analog to the sample preparations of the particle analysis but was additionally sputtered with Au to prevent electrical charging during SEM/ EDS inspections.As shown in the overview BSE image in Figure 5b, specific sample area was defined to study the layer formation in detail.The selected area (Figure 5c) was investigated with a TESCAN CLARA (TESCAN GmbH, Dortmund, Germany) with an additional 80 mm 2 SDD EDS detector (Oxford Instruments X-Max N 80, Oxford Instruments GmbH NanoAnalysis, Wiesbaden, Germany).
The SEN consisted of an anti-clogging layer and zirconiabased materials in the slag zone to prevent clogging and refractory wear.Table 3 presents the chemical composition of the anti-clogging layer of the SEN, revealing that it mainly consisted of alumina silicate.

Chemical Composition of Steel and Slag Samples
Table 4 presents the chemical compositions of each IF steel heat after the RH treatment (H1-H6) and the two remaining melts from the laboratory trials (LM1 and LM2).As explained in the previous chapter, 6.5 kg of La or Ce were added in the second to fifth heats.The chemical composition of the steel samples was determined using Spark-OES.The content of La or Ce was analyzed by using ICP mass spectrometry (ICP-MS).
Figure 6 illustrates the changes of La and Ce concentrations in the analyzed lollipop steel samples from the industrial trials.Due to the incomplete filling of the lollipop mold during the sampling process, the second tundish sample from the fourth heat was unavailable for analysis.
The chemical analysis of the steel samples showed that the highest REE content was observed immediately after the addition.Afterward, a steady reduction in the REE concentration was observed.In the case of the fifth heat, Ce traced, a lower concentration was measured immediately after REE addition.Compared to the taken sample after REE addition, the subsequent lollipop sample (after RH) had a higher Ce content.
Figure 7 provides an overview of the changes in the chemical compositions of the tundish and mold slag samples throughout the entire Ti-IF casting sequence.The REE content in the tundish slag shows a stepwise increase, correlating different heats with distinct tracers in the tundish at specific times (Figure 7).During the casting of the La-traced heats, a significantly higher increase in the La concentration was observed within the slag (Figure 7b-black diamonds) compared to the Ce concentration at the Ce-traced melts inside the tundish (Figure 7c-black squares).The variations of La and Ce concentrations within the mold slag are depicted as white triangles in Figure 7b,c.The REE concentration fluctuated periodically depending on the heat and which particular tracer was added.

Micro-Cleanness Evaluation
The results of the automated SEM/EDS analysis to determine the micro-cleanness of the lollipop samples of the industrial trial and the laboratory samples are shown in this section.The inclusions were categorized into different classes based on their nonmetallic bonding partner (O, S, N, oxide-sulfide, oxide-nitride, nitridesulfide, and oxide-nitride-sulfide).The specific typification of these inclusions depends on the metallic bonding partner (e.g., Al, Mg, or Ca).
To understand the ongoing inclusion modification resulting from the addition of REEs, SEM/EDS analysis was performed on samples from two melts in an industrial-scale experiment and two melts in a laboratory-scale experiment.The aim was to compare and highlight the differences between the industrial and laboratory scales, for instance, the influence on the achievable inclusion tracing rates.A representative La-traced melt (Heat 2 and LM1) and a melt in which Ce was added as a tracer (Heat 3 and LM2) were examined for each setting.This section focuses on the La-modification results, as similar findings were observed for the Ce-traced samples.In the four investigated samples taken directly after Al deoxidation, mainly Al-containing oxides were detected, regardless of whether they originated from the laboratory or industrial setting.In Figure 8, the results of the micro-cleanness evaluation of the two La-traced melts, Heat 2 and LM1, are presented.This overview includes only the three main types of inclusions that were predominantly found, with the addition of TiN for completeness, despite it not being part of the study.The reported mean ECD contains the entire population of traced and untraced inclusion types.The samples within each column correspond to their respective counterparts from the industrial-and laboratory-scale experiments taken during the same process period.As shown in Figure 8a,e, similar-sized alumina particles were found in both melts.For Heat 2, the Al 2 O 3 particles had a mean ECD of 1.69 μm with a standard deviation of AE0.93 μm, while for sample P1, a mean ECD of 2.29 μm AE 1.04 μm was observed.In the case of the laboratory melt (LM), a higher proportion of MgO-Al 2 O 3 and CaO-MgO-Al 2 O 3 oxides was detected.Significant differences in the tracing rates between the industrial and laboratory trials were revealed during the investigation of the samples taken immediately after the addition of La.In the case of the LM (P3), a substantial proportion of the existing alumina particles (70%) were traced and modified.In contrast, only 17.6% of the occurring alumina particles were marked within the second heat's industrial lollipop sample (after REE).All samples, taken directly after the REE addition, showed an increase in inclusion number per unit area.A change in the mean ECD was also observed.In the lollipop sample from Heat 2, the ECD of the untraced alumina was approximately 1.48 AE 0.28 μm, while for the traced, it was 1.52 μm AE 0.34 μm.Similar findings were observed in the laboratory sample, where the mean ECD of pure alumina was slightly smaller (1.49AE 0.33 μm) compared to La-modified Al 2 O 3 (1.87AE 0.96 μm).In both cases, the La-modified alumina NMIs showed in the manual investigation a heterogeneous structure and, in most cases, a spherical shape.In the industrial samples, aside from the dominant inclusion type of traced and untraced Al 2 O 3 , single SiO 2 -Al 2 O 3 and MnO-Al 2 O 3 were also identified.Following the alloying of FeTi, the content of alumina decreased, accompanied by a raised presence of Al 2 O 3 -TiO x and TiN.This increase is more distinct in the industrial-scale experiment.The highest proportion of Al 2 O 3 -TiO x and TiN was observed in the tundish lollipop samples.

Investigation of SEN Clogging Layer
The utilized SEN and their clogging layers were also subjected to investigate the correlation between the deoxidation products and the formation of these clogging structures.For this purpose, the used nozzles and their corresponding clogging layers were analyzed using SEM/EDS.Figure 9 displays the results of the elemental mapping of the first SEN (changed after the third heat), providing insight into the spatial distribution of elements.On the left side, the SEN refractory is depicted.As shown in the EDS mapping, the refractory material was infiltrated by the liquid steel, extending to a depth of approximately 1 mm into the material.
Within the clogging layer of SEN 1, three distinct zones can be identified.The first section extends from approximately 0.25 to 2.5 mm, measured from the inner wall of the SEN (Figure 9-Zone 1).Aside from pure alumina, Al 2 O 3 -TiO x and solidified steel were found within this zone.A significant portion of the observed oxide network is marked with La in the second region (Figure 9-Zone 2), which extends between 2.5 and 4.5 mm.The last section (Figure 9-Zone 3), which is located approximately 4.5-8.7 mm from the inner wall of the   SEN, is characterized by a notable presence of Ce-traced alumina inclusions within the clogging structure.In the second and third zone, areas with solidified steel can also be observed, although their size and quantity decrease with increasing distance from the SEN refractory.In contrast to the part of the clogging layer of zone 1, only small amounts of Ti-containing particles could be detected within the sintered clogging networks of the second and third clogging zone.

Recovery Rates
The addition of La and Ce led to high losses in both cases, the industrial and the laboratory trials.In the industrial experiment, the amount of REEs was approximately 40 ppm, similar to the studies of Gao et al. [17] and Geng et al. [18] where Ce was used as a tracer.The amount was significantly lower compared to the laboratory trials since the previous investigation of Tian et al. [22] found that REEs may increase the clogging tendency.The effective amount of La was about 23 ppm for the second heat and 26 ppm for the fourth.In comparison, only 15 and 8 ppm Ce were measured in the third and fifth heat.Hence, the losses for Ce were significantly higher than for La.Several studies used ferrocerium with a Ce content of approximately 10 wt% and a recovery rate of 50%, leading to lower losses of this element. [14,17,18,23]urthermore, the amount of La and Ce decreased over the process, as shown in Figure 6, due to the deposition of traced NMIs into the slag.
Compared to the industrial experiments, higher losses of the tracers were expected for the laboratory trials since only a low mass of these highly oxygen affine elements had to be alloyed.Hence, a recovery rate of 30% based on the findings of Thiele et al. [19] was set.The results of the ICP-MS investigation of all samples led to the assumption that the losses at the resistanceheated Tammann-type furnace were higher than expected.Only 5.4% of the added amount of La was found in the remaining melt of the trial LM1 and 11.3% of Ce at LM2.The reason for the lower recovery rates is the longer contact time between the REEs and the atmosphere at the resistance-heated Tammann-type furnace compared to the high-frequency remelting furnace used for the study of Thiele et al. [19] Besides the higher amount of REEs in the melt, a higher tracing rate for Al-containing NMIs results for the laboratory trials.The tracing rate in the laboratory was 70.59% for La and 92.82% for Ce.In the industry, the values of the tracing rates were about 10.23% and 34.26% for La and Ce in the example of the second and third heat.The reasons for the higher REE content were the assumption of a recovery rate for the laboratory experiment compared to the industrial ones and the presence of slag in the industry.The tracing rate was higher for Ce at the laboratory due to the higher losses of La.For the industrial trial, the tracing rate was also higher for Ce although Ce's measured content was lower than La's.In this case, the interaction between slag and steel as well as the lower uptake of traced inclusions into the slag must be considered.

Separation Behavior of NMI (Steel and Slag Samples)
As previously mentioned, a steady decline of the REE content in the industrial samples could be observed after adding these tracers at the RH degasser.This decline could indicate the separation behavior of the traced deoxidation products during this period.The micro-cleanness evaluations of these lollipop samples show a similar trend (Figure 10).A lower Ce concentration was observed in the sample taken after the Ce addition compared to the sample after the FeTi addition at the fifth heat.This difference could be attributed to inadequate homogenization within the molten steel during the sampling procedure at the RH facility.According to Yang et al. [24] the decrease in the number of inclusions during the RH treatment is primarily attributed to their coagulation through collisions caused by the intense turbulence within the steel melt during its circulation.These particles coalesce into clusters, increasing their susceptibility to removal through buoyancy forces.Moreover, their high surface-to-volume ratio facilitates the attachment of the particle clusters to the argon bubbles' surface, supporting their transport to the slag-steel interface. [12]The REE content decreased between the last sample taken at the RH degasser (after RH) and the subsequent tundish steel samples.This can be attributed to the inclusion removal process that occurs within the steel ladle during the transport of the melt to the continuous caster and the subsequent holding time at the ladle turret.In the absence of injected stirring gas in this process period, temperature stratification occurs within the steel melt, resulting in a certain degree of bath circulation.The occurring circulation subsequently facilitates the transport of existing particles to the slag-steel interface. [25]According to Zhang et al. [26] the main mechanism for inclusion removal in the continuous casting tundish is the particle growth through turbulence collision, followed by floating of particles to the free surface.Another possibility of inclusion removal is by adhesion to solid surfaces, such as the refractory lining of the tundish. [26]n addition to the decrease in REE content within the lollipop samples, these elements were gradually enriched in the respective process slags.The analysis of the ladle slag samples taken immediately after the RH treatment indicated that approximately 12-15% of the total La or Ce added to the steel was detected in the slag within approximately 10 min.Considering the La and Ce content of the tundish slag samples, a successive enrichment of these elements can also be observed throughout the entire casting period, indicating the separation behavior of these traced particles into the slag over the casting time.The analysis of the mold slag samples led to similar results.The contents of REE follow a distinctive oscillating pattern, which can be attributed to the interaction between the continuous separation of inclusions and the addition of new casting powder to the mold slag.At the same time, a part of the emulsion formed by the molten casting powder and preexisting mold slag is entrapped in the gap between the oscillating copper mold and the solidifying steel shell.A characteristic periodic variation of the REE concentration, depending on which heat is being cast at a given time, results.
These findings highlight the method's effectiveness in tracking the steel's deoxidation products throughout the entire production process and their deposition within the respective process slags.In this study, REEs were added following a standard production procedure of Ti-IF steels.The tracers in the steel plant can be used for a wide range of purposes, including assessing the effectiveness of various measures (such as gas stirring operations, tundish inserts, casting parameters, etc.) aimed at improving the cleanness of the existing steelmaking process.However, one critical aspect needs to be considered when utilizing this method.As stated in the review of Zheng et al., [27] the interfacial properties of the inclusions play a significant role in the removal process.These properties affect the agglomeration behavior, the capacity to adhere to bubbles, and the potential of NMIs to pass the slag-steel interface for subsequent dissolution into the slag phase.
Based on this background, ongoing research is focusing on developing an alternative method that does not require the active addition of tracer elements.In the study conducted by Thiele et al. [28] patterns of natural trace element distribution, including REEs, were analyzed in input materials such as alloying elements, auxiliaries, and refractory materials using ICP-MS and laser-ablation ICP-MS.The mass fractions of REEs were then normalized to a reference material and compared to the REE pattern obtained from the analysis of an SEN clog.During this investigation, a clear correlation was found between Al granules, used as deoxidizer, and the clogging material within the used SEN.

Micro-Cleanness of Steel Samples
To gain insight into the changes in the inclusion population caused by the addition of REEs, SEM/EDS analysis was performed on lollipop samples from the industrial trials and steel samples from the laboratory experiments.The analysis aimed to provide more information about the reaction between the existing inclusion population and the REEs.To achieve this objective, a comparison was made between La-traced melts (Heat 2 and LM1) and melts with Ce added as a tracer (Heat 3 and LM2).The purpose was to observe differences between the occurring NMIs at the laboratory and industrial trials regarding their morphology and composition after REE and FeTi addition.Similar types of inclusions could be found in all four samples, which were taken immediately after the Al deoxidation.The majority of the detected inclusions consisted of pure Al 2 O 3 .An increased presence of MgO-Al 2 O 3 (MA) spinel inclusions was also seen in the laboratory sample analysis.The presence of MA spinels is due to residuals in the used pure iron where these particles were formed due to a possible reaction between alumina and the refractory lining of the crucible. [29]After the addition of REEs, existing inclusions were modified, but the achieved tracing rates varied significantly.In the industrial-scale experiment, specifically in Heat 2 with La addition, the tracing agent was detected in 17.6% of the alumina inclusions in the taken sample after REE addition.Figure 10 depicts approximately 42.5% of the Al 2 O 3 inclusions being marked with Ce.
Compared to the results from the industrial trials, the laboratory-scale experiments achieved a higher tracing rate, which can be attributed to the higher concentration of REEs used in the laboratory setting.Nonetheless, a similar pattern is observed when comparing the La-and Ce-traced melts in both the industrial and laboratory settings.In both cases, a significantly higher tracing rate could be achieved by using Ce as a tracer.A similar pattern is observed when comparing the La-and Ce-traced melts in the laboratory experiments, with a higher tracing rate measured in the case of Ce addition.In the case of the industrial trial, many submicroscopic inclusions were found in the manual SEM/EDS analysis of the taken samples after Ce addition.The minimum size detected at the automated SEM/EDS measurement was 1 μm.Hence, these presumably newly formed inclusions were not analyzed.Similar findings appeared in the industrial sample after La addition.The inclusions were slightly larger compared to the sample after Ce addition, and thus, more newly formed NMIs could be measured.An explanation for this increase in NMIs after REE addition provides the study of Geng et al. [30] They found that REEs lead to an input of O and S in the melt and, as a result, to reoxidation.Since more newly formed NMIs were measured after the La addition, the number of particles per square millimeter increased, and further, the tracing rate of Al-containing NMIs decreased.
According to calculations performed by Thiele et al. [19] it is predicted that the following inclusion phases become thermodynamically stable with increasing La content after the Based on the added amount of REEs (La/Ce), the alumina inclusions will mainly form complex Al-REE oxides (REEAl 11 O 18 and AlREEO 3 ).Furthermore, in both cases, the REE content is insufficient to form pure REE oxides or REE sulfides.The prediction of similar inclusion modification steps in the case of Ce was reported by Wang et al. [23] and in the case of La by Mao et al. [31] Figure 11 shows the ternary system of Ce-Al-O including the detected NMIs of the industrial sample after RH and the remaining melt of the laboratory trial LM2.In addition to these results, the positions of CeAl 11 O 18 and AlCeO 3 are marked as reference points.
These theoretical considerations are consistent with the experimental observations.The analysis of the industrial and laboratory samples revealed that most of the REE-traced alumina consisted of complex REE-Al oxides, primarily in the form of REEAl 11 O 18 .One La-modified example of such an inclusion type is shown in Figure 12c.As shown in Figure 11, REEAlO 3 inclusions were mainly found in the laboratory setting, which is consistent with the higher concentration of added REEs.Most REE-modified NMIs were heterogeneous multiphase inclusions and showed a strong segregation tendency.This has also been documented in other publications, such as Ren et al. [32] Single inclusions of Al-REE oxide-sulfide (Figure 12b) were observed in the industrial lollipop samples taken immediately after adding the tracer at the RH degasser.In the later stage of the process, no inclusions of this type were found.This inclusion phase was likely formed immediately after the addition of REEs, as it may have led to the formation of a region with elevated REE concentration within the melt at that time.As the steel is homogenized during the ongoing RH circulation and the local REE activity decreases, the formation of these types of inclusions is suppressed as they become thermodynamically unstable.In contrast, these oxidic-sulfidic phases were not found in the samples of the LMs, which is likely due to lower sulfur contents within the pure iron.
The chosen tracing technique is possible, as these REE metals have a higher oxygen affinity at the given steelmaking temperatures than classic deoxidants (such as Al).REEs can partially reduce oxides formed during deoxidation.These REEs are incorporated into the inclusion complex in this process, effectively marking the existing NMIs. [14]The chemical reactions describing the formation of REE-Al oxides through the interaction between preexisting alumina and La and the corresponding free Gibbs energies are expressed by Equation ( 1) and (2) [33] ½La þ As shown in Figure 10, there was a significant decrease in the measured ECD of alumina particles immediately after the addition of Ce (1.54 AE 0.60 μm), compared to the sample taken directly after deoxidation (2.55 AE 1.66 μm).This change in inclusion size was also reported by Li et al. [34] Their study demonstrated that the Ce treatment influenced the size distribution and quantity of NMI per unit area.According to the documented results, there was a decrease in the percentage of inclusions larger than 5 μm as the Ce content increased.Nevertheless, a critical threshold of 0.034 wt% Ce was reported, above which the inclusion size increased once again.Similar observations were made in the present study.In the case of the industrial trial with a lower REE addition, a significant reduction in ECD was achieved.However, only a slight change in the inclusion size was observed in the laboratory experiments with higher La/Ce content.Li et al. [34] concluded that there exists an optimal Ce addition content that can facilitate the formation of finer and more dispersed NMIs.Since the results of the Ce and La treatments showed a similar pattern, the same likely applies to La.As presented in Figure 10, a significant increase in Al 2 O 3 -TiO x was observed after adding FeTi.Dorrer et al. [8] demonstrated that immediately after the addition of FeTi, this inclusion type was formed.Similar to the findings in this work, it was observed that the newly formed Al 2 O 3 -TiO x inclusions were relatively small, and the amount of pure alumina significantly decreased.The presented mean ECD values in Figure 8 and 10 are strongly influenced by the fraction of traced Ticontaining oxides, which exhibited a larger ECD in this case.Figure 12d represents an example of an untraced Al 2 O 3 -TiO x inclusion.However, it should be noted that the X-ray intensity map displays a misleading La concentration due to the overlapping X-ray peaks of the La-L α and the Ti-K α lines.Manual SEM/ EDS measurements of the traced Al 2 O 3 -TiO x revealed that it consisted partly of agglomerated REE-Al oxides and Al-Ti oxides (Figure 12e).Furthermore, it was observed that submicroscopic Ti-containing phases formed on the surface of existing REE-Al oxides.A trend observed in both the industrial and laboratory settings is the increase in the tracing rate over process time.The rise of traced NMIs may be attributed to the previously described changes in interfacial properties.As Zheng et al. [27] reported, REE oxides have a decreased wetting angle compared to regular alumina, which lowers their driving force to form clusters or chance to attach to argon bubbles.Additionally, the ongoing modification by REEs leads to an increase in their density.Based on the findings of Waudby et al. [35] REE oxides have a similar density to liquid steel.The reduction of the resulting buoyancy force, combined with the altered wetting properties, decreases the deposition probability compared to untraced alumina inclusions.
In summary, similar results were obtained regarding inclusion modification and size distribution when using La or Ce as tracers.The only noticeable difference observed in both settings was that the tracing rate using Ce was significantly higher compared to La.Further investigation is required to determine the underlying cause for the observed discrepancy between Ce and La.

Analysis of SEN Clogging Layer
The SEM/EDS investigations of the clogging layer within SEN 1 unveiled the presence of three distinct zones.The first zone, situated closest to the inner wall of the SEN, was characterized by a composition primarily consisting of pure alumina, Al 2 O 3 -TiO x , and solidified steel.No REE-marked oxides were detected in this region since the first heat was a standard heat without adding tracers.Steel droplets in the clogging layer have been observed in previous studies. [36,37]The presence of steel in the layer is attributed to changes in the wetting characteristics between the liquid steel and the refractory material of the SEN.A local concentration peak of Ti appeared within the first clogging zone in the area surrounded by solidified steel droplets (Figure 13).Further the elemental mappings indicate the formation of a CaO-Al 2 O 3 -TiO x (CAT) boundary layer at the interface between the steel and the alumina clogging network.
The formation of such a layer structure has also been addressed and discussed in the study of Lee et al. [36] based on their previous findings documented in ref. [7].Their work [36] proposed a mechanism involving a reaction between carbon and SiO 2 within the SEN refractory.This reaction leads to the formation of CO, which acts as a local source of reoxidation and oxidizes Al, Ti, and Fe.The resulting complex FeO-Al 2 O 3 -TiO x (FAT) compound is liquid at the prevailing temperatures and exhibits high wettability, serving as a binder between refractory, steel, and inclusions.Additionally, it is described that a reduction of this FAT oxide compound can occur through the dissolved Al in the steel or through reactions with CaO from the SEN refractory.As a result, the formation of the CAT boundary layer and the reduction of Fe droplets may occur.This mechanism could have also contributed to the formation of the initial clogging layer in the present study, especially during its early stage.
In the second zone, situated within the central part of the developed clogging layer, a notable amount of the observed oxide network is marked by La.An exemplary EDS mapping from an area with elevated La concentration is shown in Figure 14.Considering the absence of any additional La source in this experimental series, it can be presumed that the observed alumina oxides of zone 2 were formed during the Al killing process and subsequently modified by the presence of REE.Specifically, it can be attributed to the deoxidation products originating from the second heat of the Ti-IF casting sequence, as only this melt was alloyed with La and subsequently cast using the first SEN.
The SEM/EDS analysis was performed on the inner layer (closest to SEN center) of the clogging structure, as depicted in Figure 15.Within this specific region, alumina networks marked with Ce were observed.Similar to the second zone, the found coral-shaped sintered alumina particles are deoxidation products originating from the third and final heat, which was cast using this SEN.According to Michelic et al. [4] the primary mechanism causing the attaching and sintering of deoxidation products to the preexisting clogging layer is the combination of surface tension and fluid flow effects.A comprehensive analysis of the underlying mechanisms that contribute to the continued growth of NMIs on the existing layer structure is presented by Barati et al. [1] and Solórzano et al. [38] As shown in the overview of the investigated clogging layer, illustrated in Figure 9, overlapping regions were observed within the La-or Ce-marked clogging structure.This overlap can be attributed due to the mixing between the respective heats during the ladle exchange process within the tundish.The absence of distinct boundaries between these La-or Ce-traced oxide regions, especially in the inner part of the clogging network, indicates a  nonuniform layer formation with overlapping and intertwined clogging areas.Such a type of branched clogging structure is also predicted by model calculations performed by Barati et al. [39]

Conclusion
The following findings can be drawn from this research: 1) In both the industrial and laboratory trials, preexisting deoxidation products were traced.The tracing rates in the remaining melts were significantly higher for the laboratory trials, with 70.59% for La and 92.82% for Ce, compared to 10.23% for La and 34.26% for Ce in the tundish sample of Heat 2 and 3.The separation of traced NMIs into the used slags played an essential role in this case in addition to the higher added amount of REEs in the laboratory.2) During its application at the steel plant of voestalpine Stahl GmbH, the method proved effective in tracking the deoxidation products in Ti-IF steel throughout the production process.A steady decline of the REE content in the industrial samples could be observed during the period between the RH treatment and casting.Simultaneously, a successive enrichment of these elements was measured in the respective slags, indicating the separation tendency of NMIs into the slag phase.
3) Micro-cleanness evaluations of steel samples revealed that due to the addition of REEs, existing alumina inclusions were modified to the complex heterogeneous REE-Al oxides, REEAl 11 O 18 and AlREEO 3 , similar to previous studies.An increase in the inclusion number per unit area was observed, but the ECD decreased significantly.For example, in Heat 2 of the industrial trial, the addition of La resulted in a reduction of the ECD from 1.69 AE 0.93 to 1.49 AE 0.29 μm.In the laboratoryscale experiment (LM2), a decrease from 2.29 AE 1.04 to 1.70 AE 0.83 μm was observed.In both settings, the inclusions' size increased again after adding FeTi, and Al-Ti oxides were formed.When comparing changes in inclusion modification and size distribution due to the REEs, no significant differences were found between using La and Ce as tracers.4) Through SEM/EDS examinations of the SEN, it was shown that the REE-marked preexisting deoxidation products were involved in the formation of the clogging layer.The clogging layer network could be split into three distinct zones.The formation of each zone within the clog could be attributed to the specific heats based on their corresponding tracer (or the absence of one).

Figure 1 .
Figure 1.Production route of Ti-stabilized interstitial free (Ti-IF) steel with additional rare-earth elements (REE) treatment.

Figure 2 .
Figure 2. Ti-IF casting sequence and sampling procedure: a) tundish weight over casting time and division of heats, b) stopper rod position over casting time and changes of SEN, and c) sampling of tundish and mold slag over time.

Figure 4 .
Figure 4. Schedule of the two laboratory trials.

Figure 5 .
Figure 5. a) Cross-sectional view of halved submerged entry nozzle (SEN); b) backscattered electron (BSE) image: embedded SEN and clogging structure; and c) BSE image: magnified view of SEN refractory and clogging deposits.

Figure 7 .
Figure 7. a) Tundish weight during casting sequence; b) La and c) Ce concentrations within tundish and mold slag samples.

Figure 8 .
Figure 8.Comparison of inclusion number and size over process time in La-traced melts between the industrial trial and LM: a-d) Heat 2 and e-h) LM1.

Figure 9 .
Figure 9. a-f ) Elemental mapping of SEN refractory and clogging layer and g) corresponding EDS line scan concentration profiles for the measured section.

Figure 10 .
Figure 10.a-d) Comparison of inclusion number, size, and tracing rate over process time: Ce-traced Heat 3.

Figure 11 .
Figure 11.Ternary system of Ce-Al-O, including the detected NMIs of the third heat from the industrial trial, the remaining melt of the LM2, and the theoretical composition of CeAl 11 O 18 and AlCeO 3 .

Figure 12 .
Figure 12.Change of NMIs due to La and FeTi on the example of the industrial trial: a) Al 2 O 3 after Al; b) La-traced Al 2 O 3 after La; c) La-traced Al 2 O 3 after RH; d) Al 2 O 3 -TiO x after RH; and e) La-traced Al 2 O 3 of 1. Tundish sample.

Figure 13 .
Figure 13.CaO-Al 2 O 3 -TiO x layer on the interface of Fe droplet and oxide network: a) BSE image of SEN refractory and clogging structure; b) BSE image of the region of interest; and c-g) corresponding elemental mappings.

Figure 14 .
Figure 14.La-traced deoxidation products within the middle layer of clogging deposits: a) BSE image of SEN refractory and clogging structure; b) BSE image of the region of interest; and c-g) corresponding elemental mappings.

Table 1 .
Chemical composition of a Ti-IF steel sample provided by the steel plant of voestalpine Stahl GmbH.

Table 2 .
Chemical composition of pure iron for the laboratory trials.

Table 3 .
Chemical composition of the anti-clogging refractory in wt%.

Table 4 .
Chemical composition of industrial steel samples (H1-H6) at the end of RH refining and the remaining melt of the laboratory trials (LM1 and LM2) in wt%.
Figure 6.Changes of La and Ce content in steel samples during Ruhrstahl Heraeus (RH) treatment and casting process.www.advancedsciencenews.com l www.steel-research.desteel research int.2024, 95, 2300665