Synergistic effect of residual elements on oxidation rates and oxide/metal interface characteristics in a low-carbon steel oxidized at 1180°C for 3 hours

ABSTRACT Residual elements (Cu, Ni and Sn) of various contents have been added to a low-carbon steel to simulate the scenario of increased use of scrap during steel production. The samples were oxidized in air to represent reheating before hot rolling, using thermogravimetry analysis (TGA). Scanning electron microscopy (SEM) and energy dispersive spectrometry (EDS) were employed to investigate the microstructure and the enrichments at the oxide/metal interface. Results showed the residual elements had an impact on the roughness of the scale/metal interface, internal Fe oxides formation, the amount of Ni&Cu enriched Fe phase and Cu&Sn enriched liquid phase at the interface, and the depth and spacing of the liquid phase penetrations along the grain boundaries. Thermodynamic simulation predicted the effect of residual elements on liquid phases formation at the interface, with good agreement between prediction and experimental results. The influence of residual elements on the oxidation rate was discussed.


Introduction
Recycling steel scrap for steel production via the Electric Arc Furnace (EAF) route or Blast Furnace -Basic Oxygen Steelmaking (BF-BOS) route can significantly reduce CO 2 emissions [1][2][3].However, the steel scrap often contains residual elements that cannot be removed by the steelmaking process because they are nobler than Fe, such as Cu, Ni and Sn.The content of residual elements depends on the quality of the scrap and the amount of scrap used as feedstock in steelmaking [4]; and the content will accumulate in final steel products with an increase in recycling cycles.The effect of impurities has been investigated with regard to microstructure and mechanical properties in a low-carbon dual phase steel [5].
During heat treatment at high temperatures, which is inevitable for most steels, Fe will be oxidized to form oxide scales while Cu will enrich at the oxide/metal interface [6][7][8][9][10].When the Cu content exceeds its solubility limit in the matrix above 1100°C, Cu-rich liquid phase forms.The major problem the Cu-rich liquid phase can cause is hot shortness when the liquid phase penetrates the Fe grain boundaries and leads to surface cracking during the subsequent hot rolling [11].It has been found that Sn lowers the solubility of Cu in the steel and lowers the solidus temperature of the Cu-Fe alloy, promoting the formation of Cu liquid phase; meanwhile, the presence of Ni can counteract the effect of Sn by expanding the Cu solubility range in Fe [12][13][14][15][16][17][18][19].In Imai et al.'s work [12,13], a mild steel containing 0.3 wt-% Cu (in wt pct) was subjected to oxidation for 2 h at 1000-1300°C followed by tensile testing with a strain of 40% at the oxidation temperature.Surface cracking due to the Curich liquid phase only occurred at 1100°C.With an addition of 0.04 wt-% Sn, the cracking was found to occur over a wider temperature range of 1000-1100°C and the crack number increased.An addition of 0.3 wt-% Ni was found to totally suppress cracking during the tensile test.Note that steel scraps contain Ni, however, the Ni content is often quite low [4], and intentionally increasing the amount of Ni in steel during production increases the cost.
The amount of Cu-rich liquid phase at the interface was reduced at higher oxidation temperatures, i.e. above 1200°C [12,13,15,20].The reason given in [12,13] is that the Cu-rich liquid phase was occluded into the scale at high temperatures.Garza et al. [20] reported the surface hot shortness of a Cu-containing oxidized forging steel was mitigated at oxidation temperatures above 1200°C.It was suggested that, at higher temperatures, Cu back diffusion into the matrix prevented an accumulation of liquid Cu along the grain boundaries.
The phase constituents and their fractions at the oxide/ metal interface depend on the degree of enrichment of the different residual elements, which is related to the bulk composition and the oxidation temperature and time [16][17][18][19]21,22].A numerical model coupled with Thermo-Calc software was developed by Webler et al. [17] to predict the concentrations of different elements in the Fe phase and Cu liquid phase at the interface as a result of selective oxidation of Fe.It assumes that, after oxidation, the residual elements would leave the oxidized Fe and diffuse to a layer of liquid phase by the interface and then to the Fe matrix.However, the model is only valid when no residual elements are lost to the scale due to occlusion, which occurs during longtime high-temperature oxidation.
The oxidation rates of steels can be affected by residual elements.Ali et al. [23] showed that a higher level of Ni (2.92 wt-% compared to 0.50 wt-%) reduced the oxidation rates and lowered the oxidation activation energy in a high strength steel subjected to oxidation at 1000-1200°C for up to 60 mins.Yin et al. [22] reported the effect of various residual elements on the oxidation kinetics in Fe at 1150°C for up to 10 mins.It was found that 0.3 wt-% Cu did not change the oxidation rate, but when Ni was present the oxidation rate decreased.However, no relationship was observed between the Ni content and oxidation rate.It was also mentioned that the addition of Sn (0.03-0.15 wt-%) lowered the oxidation rate, however there was no clear relationship between the content and oxidation rate [21].
In previous studies, the effects of residual elements on the oxidation behaviours and hot shortness have been investigated, and the benefits of Ni on suppressing the liquid Cu phase formation have been documented.However, previous studies have mainly focused on short oxidation times, and there is a lack of information regarding the oxidation rates and comprehensive characterization of the metal/oxide interface in steels subjected to longer oxidation times that represents typical slab reheating during steel production.In this paper, residual elements (Cu, Ni and Sn) have been added to a low-carbon steel to simulate the recycling of various grades of scraps during steel production.The levels of these elements were chosen to represent the scenarios of increased recycling of various quality grade scraps, including high Cu and high Sn grade scrap.Thermogravimetry analysis (TGA) and scanning electron microscopy/energy dispersive spectroscopy (SEM/EDS) were carried out to investigate the effect of residual elements on the oxidation behaviour and characteristics of the metal/oxide interface.Fine EDS mapping of the residual elements at the interface showed significant enrichments of the residual elements and revealed the distributions of the liquid CuSn phase and the solid Fe(CuNi) phase, allowing for accurate measurements of their amounts and compositions.In addition, the penetration of the liquid phase along the grain boundaries was quantified by measuring the depth and spacing of the penetrations.The current work provides quantitative information that can be linked to hot shortness cracking in the future.

Materials and experimental methods
The chemical compositions of the studied steels are shown in Table 1.The compositions of the residual elements are in bold.The compositions were measured using optical emission spectroscopy (OES) for metal elements and a Carbon/ Sulphur Analyzer for C and S contents.Steel 1 represents the base composition of the Al-killed low-carbon steel produced via the BF-BOF steelmaking route.Impurity elements, such as Cu, Ni and Sn, were added to the base composition to simulate the compositions when increased scraps are used for production.Steel-2 to Steel-5 contain different levels of impurity elements to provide a flexible and large margin for impurity increase.
Ingots of 25 kg were cast using a vacuum induction melter in Tata Steel IJmuiden.Ingots of 100 mm thick were homogenized at 1220°C for 60 min, hot rolled to 32 mm transfer bars in seven passes.Rectangular samples were machined out from the transfer bars, with dimensions of 30 × 22 × 10 mm.Samples were polished with 600-grit SiC paper.The Quartz tube in the TGA furnace is 6.5 cm in diameter and 88 cm in height, and the sample is within the uniform temperature zone.The furnace was filled with Ar and pre-heated to the target temperature, i.e. 1180°C.The temperature of the hot zone was calibrated by a K-type thermocouple.The sample was lowered to the centre of the hot zone of the furnace and hung from a balance using a platinum wire.After 2 min, Ar was switched off and air was filled into the furnace at a flow rate of 5 L/min (corresponding to a velocity of 12 cm/s at 1180°C), and the sample was held isothermally for 3 h.In the end, the sample was immediately transferred to a tank filled with Ar and cooled to room temperature.Repeats were made to confirm the trend of oxidation rates in steels of various levels of residual elements.
The samples were cold-mounted, sectioned and polished.For the final polishing stage, 0.05 µm colloidal silica was used to achieve a smooth surface finish.Scanning Electron Microscopy (SEM) in backscattered electron mode (BSE-SEM) was performed using an FEI-Versa system to reveal the microstructural features of the scales and interfaces after oxidation.In the BSE-SEM, the oxidation-induced phases were distinguished by their brightness.The CuSn phase appears brighter than the Fe(CuNi) phase, and the Fe (CuNi) phase is brighter than the Fe matrix, as the atomic weights are in the order: Fe < Ni < Cu < Sn.The voltage, current, working distance were 15keV, 8n-10nA and 10-12 µm, respectively.Energy dispersive X-ray (EDS) was used to reveal the elemental information at the regions of interest.Equilibrium phase calculation using FactSage 8.2 and Thermo-Calc with a TCFE10 database was applied to the current Fe-Cu-Ni-Sn system to support the experimental observations.

Oxidation rate
The mass changes per surface area (Δm/A, mg/cm 2 ) against time for the steels are plotted in Figure 1(a).All the TGA cures show that the oxidation rates were relatively fast in the initial stage, but slowed down thereafter.Parabolic fitting was carried out ([Δm/A] 2 vs time) for the TGA curves, and the parabolic rate constant for each TGA curve is given in Figure 1(b).Steel 1 with the base composition, showed the slowest oxidation rate.Increasing the impurity contents led to higher oxidation rates in Steel 2 and 3. Compared to Steel 3, Steel 4 has 0.30 wt-% more Cu and 0.15 wt-% less of Ni, and it showed a slightly lower oxidation rate than Steel 3. Compared to Steel 4, Steel 5 contained 0.15 wt-% more Ni, and the oxidation rate decreased.

Scale
The panoramic SEM micrographs of the oxidation scales of the steels are given in Figure 2. All the oxide scales consist of three major sub-layers of wustite, magnetite and haematite, in that order from the oxide/metal interface to external surface.Note that the relative thickness fraction of wustite in all scales deviates from 0.95, especially in Steel 1 and 2, indicating the interruption of Fe cation diffusion through wustite during oxidation [24], which is discussed in section 4.3.In addition, many large pores /cracks can be seen in the scale on Steel 1 and 2 samples, but much less so for Steels 3-5.Meanwhile, magnetite precipitates could be observed in the wustite layer in all the scales, which is due to the decomposition of wustite during cooling [22,25], an example is shown in Figure 3.Moreover, a large gap was present between the scale and the steel in all samples.It is likely that during cooling, interfacial stress was generated as the scale and metal have different thermal expansion coefficients, causing the scale to lift from the surface and potentially spall off.

Interface morphology
The oxide/metal interface morphologies are displayed in Figure 4(a-e) for the steels after oxidation.The interface for Steel 1 was nearly planar.The addition of residual elements led to uneven interfaces.Interface roughness, which was defined as the measured length of the interface divided by the length of the image, was determined for the steels and the values are given in Figure 4(f).The total sample length used for the analysis was 5400 µm for each steel.The interfaces became rougher as the contents of Cu and Ni increased up to 0.30 wt-% Ni and to 0.30 wt-% Cu in Steel 3. Compared to Steel 3, Steel 4 and 5 contained a higher content of Cu and exhibit interfaces of lower roughness.

Oxidation products
The BSE-SEM images and the corresponding EDS mapping near the oxide/metal interface for Steel 1 after oxidation are shown in Figure 5. Internal Fe oxides could be found by the scale/metal interface.They contain about 65 wt-% of Fe and 10 wt-% Mn, measured by EDS.In addition, spherical oxides beneath the interface are observed.EDS mappings in Figure 5 indicate Mn, Al, Si and S are present in those internal oxides.The Fe and Mn contents in those spherical oxides are around 40 and 35 wt-%, respectively.Fine EDS scans were conducted on some of the internal oxides and one representative case is given in Figure 6.This oxide exhibits complex structures: in addition to the presence of Mn, Al and Cr coexist in the middle of the particle, and Si appears on the periphery.Also, S enrichment at the interface is observed.The amount and the distribution of each element present in each oxide particle examined varied; however, the following conclusions can be drawn based on the observations: a. Mn presents throughout the oxides; b.Al appears in 90% of the oxides; c, Cr appears in 63% of the oxides, it always coexists with Al; d. for 87% of the oxides, Si appears at the outer layers; and e. for 77% of the oxides, S appears at their interfaces with metal is expected to MnS.
Phase equilibrium calculation was carried out using Thermo-Calc for a system at a pressure of 1 bar at 1180°C.The input composition for the calculation, indicated in Table 2, was determined by EDS map scans that covered a region 80 µm in length and 40 µm in depth beneath the scale/metal interface.The equilibrium phases and their corresponding compositions are given in Table 2.In addition to the matrix, several oxide types were predicted, such as the halite phase rich in Mn, the spinel phase rich in Al, Cr and Mn, and the olivine phase rich in Mn and Si.The MnS phase was also predicted to be present.It appears that the elemental observations from EDS in the particles can be correlated to the phases calculated by Thermo-Calc.
The additions of impurity elements had an impact on the oxide/metal interface characteristics.The Cu and Ni maps of the oxidized Steel 2 sample, Figure 7, show that a continuous layer of Ni and Cu-rich Fe phase (referred as Fe(CuNi) phase) was formed along the scale/metal interface, and some Fe (CuNi) phase was occluded into the scale.This continuous layer of Fe(CuNi) phase was not seen in the steel containing Cu and Ni after short-time oxidations, suggesting a higher degree of residual element enrichment after long-time oxidation [18,19].Meanwhile, large Fe oxides appeared close to the scale interface.
With an increase in the contents of residual elements in Steel 3, Figure 8, a larger amount of the Fe(CuNi) phase was produced at the interface, and another distinct phase, rich in Cu and Sn (referred as CuSn phase), was observed that coexisted with the Fe(CuNi) phase.This phase is expected to be liquid at the oxidation temperature of 1180°C [12].In Steels 4 and 5, the Cu content was further increased to 0.6 wt-%, leading to increases in the amount of both Fe(CuNi) phase and CuSn phase, Figure 9     The compositions of the Fe(CuNi) phase and CuSn phase are summarized in Figure 13.In the Fe(CuNi) phase, the Ni content is around 8.0 and 14.5 wt-% when the bulk content is 0.15 wt-% and 3.0 wt-%, respectively.The Cu content in  Fe(CuNi) also increases with its content in the bulk.Meanwhile, Fe(CuNi) phase with the highest Sn content was found in Steel 3. Regarding the CuSn phase, Steel 3 only had Type One formed along the interface.While in Steel 4 and Steel 5, both Type One and Type Two appeared, and each type has very similar compositions in the two different steels.In Steel 4, the Cu and Sn contents in Type One were 50% and 34%, compared to 75% and 14% in Type Two, respectively.Based on the EDS measurement, the Type One and Two CuSn phase are probably the δ and α (Cu) phase, respectively [26].Note that the measured Cu and Sn contents in Type One deviated from those in δ from the binary phase diagram.This is because the phases studied contained other elements in addition to Cu and Sn.Also, the accuracy of EDS is compromised when analysing the contents of small secondary phases.

Elemental enrichment
EDS analyses revealed the enrichments of Cu, Ni and Sn near the scale/metal interface.Ni, Cu and Sn are more noble than Fe, therefore oxidation of those elements did not occur; instead, they were rejected from the oxides and accumulated at the interface with the steel.When the Cu enrichments exceed the Cu solubility in the steel at the oxidation temperature, liquid Cu will form at the oxide/metal interface.As reported, Ni increases the solubility limit of Cu in the austenite phase and suppresses the formation of the liquid phase; while Sn counteracts the effect of Ni by decreasing the Cu solubility in the austenite phase [12,13].To understand the synergistic effects of the residual elements on the liquid phase formation, the Fe-Cu-Ni-Sn quaternary system at  1180°C was projected on the Fe-Cu-Ni ternary system using FactSage (Database: FSstel) software, and solidus lines at different Sn contents were imposed on the phase diagram, as shown in Figure 14(a).It shows that small amounts of Sn can greatly expand the solid + liquid phase and required large amounts of Ni to prevent the liquid phase formation.The contents of the Cu, Ni and Sn at the interface were obtained by calculating the weighted average contents of those elements in the Fe(CuNi) and CuSn phases by the interface, as shown in Figure 14(b).It is important to note that the magnitude of enrichment is different for each element in each steel; therefore, the Cu/Ni/Sn ratio at the interface was different from that in the bulk.For Steel 2, the selective oxidation of Fe led to the increase in the contents of Cu, Ni and Sn to 4.0, 7.9 and 0.2 wt-%, respectively, at the interface.In this case, the interface composition is far away from the liquid-solid two-phase region (Figure 14(a)), which supports the observation that the liquid phase was absent in Steel 2. Steel 3 doubled the residual level compared to Steel 2, resulting in 10.1 wt-% Cu, 14.4 wt-% Ni and 2.4 wt-% Sn at the interface, which falls close to the solidus line by the two-phase side; meanwhile, SEM/EDS detected some of the expected liquid phase compositions in Steel 3. Steel 4 and 5 had relatively large amounts of Cu in the bulk, and the Cu was enriched to 23.4 and 18.9 wt-%, respectively, at the interface leading to the formation of liquid phase in both steels.Compared to Steel 5, Steel 4 contained a lower amount of Ni, which reduced the enrichment of Ni and increased the enrichment of Sn, moving the composition deeper into the two-phase region.The phase diagram therefore indicates a higher amount of liquid phase would form at the interface in Steel 4, which agrees with the observation that the normalized liquid phase thickness is higher in Steel 4 than in Steel 5. The analysis led to an important implication: the formation of the liquid phase is determined not only by the bulk contents of the residual elements, but also by their enrichments at the interface.Here, the challenge is that a model is not available to predict the enrichment of the residual elements taking into account the interface roughness and occlusion.As such, it is difficult to predict the magnitude of the liquid phase formation with the basic parameters: bulk composition, temperature and time.The current work provides information for the development of a model that considers interface roughness and occlusion during long-time oxidation.
In Si-containing steels, it has been reported that adding Si (between 0.1 and 0.2 wt-%) resulted in rough scale/metal interfaces after oxidation [27,28].However, the Si content in  the current steels was negligible.On the other hand, uneven interfaces caused by Ni enrichment have been reported in Nicontaining steels subjected to oxidation over 1100°C [14,29,30].Therefore, the rough interfaces observed in Figure 4 were caused by Ni rather than Si.Akamatsu et al. [29] observed smooth oxide/metal interfaces in steels containing 1 wt-% Cu and 0.5 wt-% Sn after 1200°C/30min oxidation.However, with the addition of 0.5 wt-% Ni, irregular interfaces were obtained after oxidation.Fukagawa et al. [14] showed a small addition of 0.023 wt-% Ni was enough to cause irregular interfaces after oxidation at 1200°C.On the other hand, it was proposed by Chen et al. [30] that Cu could modify the interface morphology when the Cu-rich liquid phase forms.The liquid phase could flatten the interface, as observed in Fe-Cu steels [18].This may explain why the interface in Steel 5 with 0.6 wt-% Cu shows a lower roughness than in Steel 3 with 0.3 wt-% Cu when both steels contain the same Ni content (0.3 wt-%).
It is known that the addition of Al to stainless steels helps to increase their oxidation resistance by forming a continuous and compact Al 2 O 3 layer at the metal/scale interface.However, this effect requires relatively high Al contents (above 1 wt-%) and low oxidation temperatures (<1100°C ) [31][32][33].In the current steels, Al was found to exist in the internal oxides, but a continuous layer of Al oxides at the scale/metal interface has not been detected.The Al content was insufficient to provide a continuous layer of Al 2 O 3 , and the Al content did not vary between the steels, making it difficult to draw conclusions about the effect of Al on the oxidation behaviour in these steels.

Internal oxidation
Oxidation of Fe is controlled by the diffusion of iron cations through the wustite, and local equilibrium is established at the scale/metal interface [8].The Fe activity is approximately unity, and the oxygen activity is determined by the dissociation pressure of FeO.However, when the interface is covered by a layer of Fe(CuNi) phase, the local equilibrium conditions will change, which has important implications on the oxidation behaviours at the interface [16].The Fe activity as a function of Fe content in the Fe-Cu-Ni system at 1180°C was evaluated using FactSage, and the results are given in Figure 15.Here, Ni/Cu ratios relevant to those measured in the Fe(CuNi) phases in various steels are considered.The calculations show the magnitude of Fe activity decreases when Cu and Ni are present.Note that at a given Fe content, the Fe activity decreases more when the Ni content is higher, indicating that Ni is more effective in reducing the Fe activity than Cu.During oxidation experiments, as the Fe activity reduces due to the Ni and Cu enrichments at the interface, the oxygen activity at the interface increases to maintain the thermodynamic equilibrium.Thus, oxygen would diffuse into the matrix and internal Fe oxidation occurs.Consequently, the internal oxides grow and link together, and also link with advancing external oxides, occluding the Fe(CuNi) and CuSn phases [16].In the current work, internal Fe oxides were rarely observed in Steel 1.In Steel 2, those internal Fe oxides frequently formed at the interface.In Steel 3-5, with high residual levels, large amounts of internal Fe oxides can be observed, which means that occlusion of Fe(CuNi) and CuSn phases occurs more readily.

Over-oxidization of the scale
It has been widely reported that the thickness ratio of wustite/magnetite/haematite is 95:4:1 in the oxide scales formed during Fe oxidation in air between 700-1250 °C [34].The wustite thickness ratio deviates from 0.95 when the diffusion of Fe cations through the wustite layer is interrupted.In the current work, the wustite ratio in all the steels deviated from the ideal 0.95; especially for Steel 1 and Steel 2 (Figure 2).The decreased wustite ratios indicate that Fe diffusivity through the wustite layer was suppressed, which indicates the occurrence of detachment between the oxide and metal during oxidation [27] (the detachments discussed here are not the gaps observed in Figure 2, which were probably formed during cooling).The detachment effectively blocked iron diffusion across the scale, resulting in the virtual cessation of Fe oxidation [25].However, oxidation continued in the wustite and magnetite layers changing the oxides, thereby increasing the thickness of the magnetite and haematite layers and decreasing the thickness of the wustite layer.The 'over-oxidization' of the scales is tentatively explained as follows.
Generally, the specific volume of the metal is different from that of the oxides [8].If the scale growth is solely achieved by the outward diffusion of metal cations, the new oxide would form at the scale surface without any constraint, thus no stress will be developed.It has been proposed that inward oxygen diffusion is possible via grain boundaries and microcracks in the scale; thus, new oxides could form in the scale [35].Evidence of inward oxygen diffusion has been reported in multiple systems [36][37][38].When oxides form inside the scale, stresses arise due to the volume differences between the oxide and the metal.As Fe oxides exhibit larger specific volumes than iron, when the new oxides form in scale interior, lateral stress can develop.If the stress exceeds the critical scale/metal bonding strength, the scale and metal lose contact, leading to porosity formation and local scale detachment.However, when the steels contain Ni, scale/ metal interfaces became rougher with oxide protrusions, which could link with internal oxidations.As a result, the scale was anchored and the adhesion between the scale and metal became stronger.On the other hand, this anchoring mechanism was either unavailable in Steel 1 or weak in Steel 2. It is postulated that due to the weak scale/metal adhesion, the oxidation was slowed down in Steel 1 and 2 associated with the detachment during oxidation.
In the current work, it is ascertained that the enrichment of Ni can effectively suppress the formation of liquid phase.However, the uneven interface is expected to make descaling difficult.If the scale is not properly removed, during the subsequent rolling, the rolls could press the oxides into the base metal, causing damage and reducing the surface quality of the products.

Conclusions
Oxidation of steels with different levels of residual elements was carried out at 1180°C for 3 h, and the oxidation rates and interface characteristics were assessed for the steels.The main conclusions are: 1. Selective oxidation led to the enrichment of Cu, Ni and Sn at the interface.Fe(CuNi) formed at the interface and its amount was proportional to the Cu + Ni level.Sn promoted the liquid CuSn phase formation, while Ni decreased the amount of CuSn phase.Cu and Sn grain boundary penetrations were observed in steels with high residual levels (Steels 3-5).2. Ni enrichment caused unevenness of the interface between the oxide scale and metal.Meanwhile, Ni enrichment, along with Cu, reduced the Fe activity, resulting internal Fe oxidation.The rough interface and internal oxides are expected to increase scale adhesion to the metal.It is postulated that the relatively strong adhesion caused less porosity/detachment at the interface, and hence faster oxidation rates during oxidation in the steels with high residual levels.3. FactSage was used to predict the phase constituents at the interface, using the compositions at the interface as the inputs.The predictions for the formation of CuSn phase agree with the experimental observations.

Disclosure statement
No potential conflict of interest was reported by the author(s).
and Figure10.It is observed that the CuSn phase can be categorized into two types.Type One contains a higher fraction of Sn than Type Two (34 wt-% vs 14 wt-%).It was measured that the area ratios of Type One and Type Two are 0.1:0.9 in Steel 4 and 0.05:0.95 in Steel 5, respectively.They could be the products of the transformation of the CuSn liquid phase during cooling[26].The Fe (CuNi) and CuSn shown in the images would eventually be occluded after internal Fe oxides grow and encompass them as the oxidation process proceeds.Examples of occlusions of the CuSn phase are given in Figure10, as indicated by arrows.Moreover, in Steels 3-5, the CuSn phase was observed penetrating the grain grains (GBs).Examples are given in

Figure 2 .
Figure 2. BSE-SEM images showing the scales for the steels oxidized at 1180°C for 3 h.Scale bar, 3 mm.

Figure 3 .
Figure 3. BSE-SEM images showing the examples of magnetite precipitates in wustite in the scale of Steel 1.

Figure 11 .
Figure 11.The average depths of the GB penetration in Steel 3, 4 and 5 were 22 ± 6, 33 ± 15 and 38 ± 16 µm, respectively; and the average spacings of the Cu&Sn enriched GBs were 360, 189 and 257 µm, respectively.The measurements cover a total distance of 3600 µm along the interface for each steel.To quantify the amount of the Fe(CuNi) phase and CuSn phase at the scale/metal interface, normalized phase thickness was defined as the measured area of the phase divided by the length of the image from 10 images covering a total length of 1000 µm.The normalized phase thicknesses measured for the Fe(CuNi) phase and the CuSn phase in the steels are summarized in Figure 12.It is indicated that the amount of Fe(NiCu) phase formed during oxidation was proportional to the level of Cu + Ni contents (see the embedded image in Figure 12(a)).Meanwhile, the largest amount of CuSn was found in Steel 4 (0.60 wt-% Cu).Increasing Ni to 0.30 wt-% in Steel 5 suppressed the CuSn formation, showing the influence of the Cu and Sn to Ni ratio.Decreasing the Cu content to 0.30 wt-% in Steel 3 further reduced the CuSn levels.In Steel 2, the presence of 0.15 wt-% Ni has prevented the CuSn formation when 0.15 wt-% Cu and 0.03 wt-% Sn were present.The compositions of the Fe(CuNi) phase and CuSn phase are summarized in Figure13.In the Fe(CuNi) phase, the Ni content is around 8.0 and 14.5 wt-% when the bulk content is 0.15 wt-% and 3.0 wt-%, respectively.The Cu content in

Figure 4 .
Figure 4. (a-e) BSE-SEM images showing the morphologies of the scale/metal interfaces for the steels oxidized at 1180°C for 3 h.Scale bar, 30 µm.(f) Roughness of the interfaces for the oxidized steels.

Figure 5 .
Figure 5. BSE-SEM image and elemental maps of the Steel 1 (Base) oxidized at 1180°C for 3 h.

Figure 6 .
Figure 6.BSE-SEM image and elemental maps of one representative inclusion near the scale/metal interface of the oxidized Steel.

Figure 11 .
Figure 11.BSE-SEM image showing the enrichments of residual elements at the grain boundaries in (a) Steel 3, (b) Steel 4 and (c) Steel 5. EDS maps of (d) Cu, (e) Ni and (f) Sn corresponding to (b).

Figure 12 .
Figure 12.Normalized thickness of (a) Fe(CuNi) phase and (b) CuSn liquid phase in the oxidized steels.The normalized thickness of Fe(CuNi) against Cu + Ni level is embedded in (a).

Figure 14 .
Figure 14.(a) Fe-Cu-Ni-Sn quaternary system at 1180°C was projected on the Fe-Cu-Ni ternary system using FactSage.Solidus lines at various Sn contents are given.(b) contents of Ni, Cu and Sn at the scale/metal interface.

Figure
Figure Fe activity in the Fe-Cu-Ni system at 1180°C, calculated by FactSage.
Figure 1.(a) Weight gain and (b) Parabolic oxidation rate constants for the steels oxidized at 1180°C for 3 h.

Table 2 .
Equilibrium oxidation phase calculation using Thermo-Calc at 1180°C.The compositions are in wt-%.The input composition for the calculation was determined by EDS map scans near the scale/metal interface of the oxidized Steel 1.