E ﬀ ect of Niobium Addition on the High-Temperature Oxidation Behavior of 22Cr25NiWCoCu Stainless Steel in Air

: The oxidation behavior of 22Cr25NiWCoCu stainless steel with 0–0.86 wt.% Nb addition was investigated at 900 ◦ C in air. With an increase in Nb addition, the volume fraction and average particle size of Nb (C, N) increased, but the grain size decreased. The results of grazing incidence X-ray di ﬀ raction indicated that the steels with Nb additions promoted Cr 2 O 3 formation. After 10–100 h of oxidation, Cr depletion occurred at the region with Nb (C, N) precipitates because of its high Cr consumption rate, leading to Fe oxide generation. Finally, after 100 h of oxidation, the outer layer of Fe 2 O 3 and the inner layer of (Cr, Fe, Mn) 3 O 4 were formed.


Introduction
According to a global energy report from the International Energy Agency, the amount of CO 2 emissions increased from 23 gigatons (Gt) to 32.5 Gt from 2000 to 2017 [1]. Reducing CO 2 emissions is necessary to decrease their impact on the environment. CO 2 emission is generally reduced in two ways: one is to improve thermal efficiency, and the other is to capture CO 2 from the end-of-pipe fuel gas [2]. Given that thermal efficiency can be improved by increasing the working temperature, the materials applied as parts of power generators may face harsh working conditions such as high temperatures (700-760 • C) and high steam pressures (30)(31)(32)(33)(34)(35). Accordingly, the materials must have good high-temperature mechanical properties and oxidation resistance. In general, the best candidate materials are Ni-based superalloys (for example, Inconel and Incoloy alloys) because of their superior mechanical properties (tensile stress and creep resistance) and corrosion resistance (hot corrosion and high-temperature oxidation); however, their costs are also relatively high. Thus, austenitic stainless steels with high chromium (Cr, 16-30 wt.%) and nickel (Ni, > 20%) content should be the focus in the development of high-temperature applicable materials.

Materials and Sample Preparation
The chemical compositions of the steels studied are listed in Table 1. The steels were melted based on the major compositions of the commercial Sanicro 25 steel with varying amounts of Nb. The nitrogen content of the molten metal was analyzed using an oxygen and nitrogen elemental analyzer (HORIBA EMGA-920 O/N analyzer), and the other alloying elements were analyzed by an inductively coupled plasma atomic emission spectroscope ARL-4460 Metals Analyzer (Thermo Fisher Scientific, Waltham, MA, USA). The materials were produced using a vacuum induction-melting furnace. The ingot was hot rolled at 1250 • C and finished at an end-rolling temperature of 950 • C with a final reduction ratio of 85%. The hot-rolled plate was normalized at 1050 • C for 30 min to remove residual stress and annealed at 1300 • C for 2 h to ensure that the steel microstructure was homogenous.
The specimens for the oxidation experiments with dimensions of 14 × 9 × 1 mm 3 were cut from the as-annealed plates, and a hole with a diameter of 1 mm was subsequently drilled at the top of the specimen for suspension. The prepared specimens were ground with SiC abrasive paper up to #1200 and ultrasonically cleaned with acetone and ethanol. The specimens were then suspended in a ceramic crucible with a platinum wire for the oxidation tests conducted by TGA and by the atmospheric box furnace.

High-Temperature Oxidation Tests
Continuous isothermal weight changes were analyzed using TGA (NETZSCH STA 449 F3 Jupiter ® , Selb, Germany). The samples were suspended in a ceramic crucible with a platinum wire. The chamber was heated to 900 • C with a heating rate of 40 • C/min and then held for 150 h with a gas flow of 100 mL/min (laboratory air, 0.21 atm oxygen partial pressure).
Specimens for the observation of oxides were exposed to the ambient atmosphere, suspended in the ceramic crucible with a platinum wire, and directly placed into the atmospheric furnace preheated to 900 • C with exposure times of 1, 10, 25, 50, 75, 100, and 150 h. The specimens with the crucible were then removed from the furnace and air cooled at the end of the exposure. The weight change of the oxidized specimens was measured with an electronic balance (±0.01 mg).
GI-XRD (Bruker D8 Advance Series II, Bremen, Germany) was employed for the phase identification of oxidation products at 2 • X-ray incidence angle. GI-XRD results were analyzed by using the software TOPAS for calculating the ratio of phases. The microstructure characterization was examined by the backscattered electron mode of the SEM (Hitachi SU-1500, Hitachi, Tokyo, Japan). The specimens for cross-sectional observation were plated with a nickel layer to avoid damaging the oxide layers during grinding and polishing. Phase distributions and crystallographic orientations of the oxide layer of the specimens oxidized for 150 h were observed using a Zeiss Supra 55 field-emission electron microscope, together with Oxford Symmetry: Symmetry CMOS Detector (Oxford instrument, High Wycombe, UK) for EBSD data collection with 20 kV accelerating voltage at 12 mm working distance under 70 • tilt angle. Chemical composition characterization of the oxide layer was conducted using EDS (EDAX, Octane Elect EDS System) at 30 kV accelerating voltage and a 10 mm working distance. Figure 1 shows the optical micrographs of the 22Cr25NiWCoCu stainless steel specimens with Nb addition. In the case of 0-Nb, no precipitate was observed, and the average grain size was 426 ± 107 µm. For 0.29-Nb addition, the average grain size decreased to 146 ± 25 µm, and the fine precipitates of Nb (C, N) were randomly distributed in the matrix. The average grain size of 0.58-Nb steel was 114 ± 12 µm, and the Nb (C, N) precipitates were fine and large sized. When the Nb addition was raised to 0.86 wt.%, the volume fraction of large-sized precipitates of Nb (C, N) increased significantly, few fine particles were observed, and the average grain size decreased to 90 ± 12 µm.

Microstructure Analysis of As-Reannealed Specimens
These results show that when the Nb addition increases in 22Cr25NiWCoCu stainless steels, the average grain size decreases, and the precipitates of Nb (C, N) tend to cluster. These phenomena were also observed in our previous study [13], in which the fine and coarse precipitates of Nb (C, N) simultaneously appeared. The hardness values of the four steels were~210 HV, as shown in Table 2. These results indicated that alloyed Nb can promote the grain refinement and precipitation of Nb (C, N), but the hardness of the four steels did not exhibit substantial differences.   Figure 2 shows the weight gain per unit area for each specimen. Continuous measurements were conducted using TGA to investigate the change in oxides during oxidation. Discontinuous measurements were performed using an atmospheric furnace in the same oxidation conditions, and the discontinuous weight changes of individual specimens were recorded. TGA results are shown in Figure 2a, in which the solid lines represent the continuous weight changes over time, and the dotted lines are the fitting curves using the power law:

High-Temperature Oxidation Behavior
where ΔW is the weight gain per unit area; KP, n, and C are the constants for the power law; and t is the exposure time. The fitting curves are listed in Table 3. The results show that the oxidation behaviors of 0-Nb and 0.29-Nb steels had n values of approximately 2, and those of 0.58-Nb and 0.86-Nb steels had n values larger than 3. Furthermore, 0.58-Nb steel had the highest weight gain before 46 h, but 0.86-Nb steel had the highest weight gain after 46 h, which is clearly observed in Figure 2a. Table 3. Fitting parameters of thermal gravimetric analysis (TGA) data using the power law.   Figure 2 shows the weight gain per unit area for each specimen. Continuous measurements were conducted using TGA to investigate the change in oxides during oxidation. Discontinuous measurements were performed using an atmospheric furnace in the same oxidation conditions, and the discontinuous weight changes of individual specimens were recorded. TGA results are shown in Figure 2a, in which the solid lines represent the continuous weight changes over time, and the dotted lines are the fitting curves using the power law:

High-Temperature Oxidation Behavior
where ∆W is the weight gain per unit area; K P , n, and C are the constants for the power law; and t is the exposure time. The fitting curves are listed in Table 3. The results show that the oxidation behaviors of 0-Nb and 0.29-Nb steels had n values of approximately 2, and those of 0.58-Nb and 0.86-Nb steels had n values larger than 3. Furthermore, 0.58-Nb steel had the highest weight gain before 46 h, but 0.86-Nb steel had the highest weight gain after 46 h, which is clearly observed in Figure 2a. Table 3. Fitting parameters of thermal gravimetric analysis (TGA) data using the power law. reveals that all specimens had three oxidation stages. First, a weight gain following the parabolic growth law was observed at 10 h. A low oxidation rate was then reached. Finally, the oxidation rate increased again for 0.58-Nb and 0.86-Nb steels, where the most weight gain can be found in 0.86-Nb steel after 150 h of oxidation. Although the test conditions for the continuous and discontinuous measurements obtained from TGA and the atmospheric furnace were similar, the weight gain curves obtained were different, as shown in Figure 2. The same specimens were employed for continuous measurements in TGA, whereas those for the atmospheric furnace were oxidized after a given time and then cooled to obtain the weight gain. The 0.58-Nb and 0.86-Nb steels showed similar measurements after long oxidation times and had higher weight gains than 0-Nb and 0.29-Nb steels. In addition, 0.86-Nb steel gained the highest weight in both tests.  Figure 2b shows the weight gain per unit area obtained from discontinuous measurements and reveals that all specimens had three oxidation stages. First, a weight gain following the parabolic growth law was observed at 10 h. A low oxidation rate was then reached. Finally, the oxidation rate increased again for 0.58-Nb and 0.86-Nb steels, where the most weight gain can be found in 0.86-Nb steel after 150 h of oxidation.

Specimen
Although the test conditions for the continuous and discontinuous measurements obtained from TGA and the atmospheric furnace were similar, the weight gain curves obtained were different, as shown in Figure 2. The same specimens were employed for continuous measurements in TGA, whereas those for the atmospheric furnace were oxidized after a given time and then cooled to obtain the weight gain. The 0.58-Nb and 0.86-Nb steels showed similar measurements after long oxidation times and had higher weight gains than 0-Nb and 0.29-Nb steels. In addition, 0.86-Nb steel gained the highest weight in both tests. Figure 3 shows the surface morphology after 10 and 150 h of oxidation. After 10 h of oxidation, some differences were visible between the steels with and without Nb addition. In the case of 0-Nb steel, the surface morphology was homogeneous and flat, and the oxides grew along the direction of scratches, as shown in Figure 3a. Figure 3c,e,f shows that on the surface of the steels with Nb additions of 0.29, 0.58, and 0.86, some nodules were observed with average sizes of 1, 3, and 6 µm, respectively. Furthermore, the oxides formed on the nodule-free sites of Nb-added steels were homogeneous and flat, similar to those in 0-Nb steel. The surface morphology of 0-Nb steel after 150 h oxidation is shown in Figure 3b, in which the oxides remained homogeneous but with a slight increase in the oxide grain size. Figure 3d shows that the nodules of 0.29-Nb steel grew after 150 h of oxidation. Some nodules formed relatively close to the others, and the sites without nodules remained homogeneous. A significant change in the morphology of oxide was found in 0.58-Nb and 0.86-Nb steels. In Figure 3f, a porous, sponge-like oxide can be clearly seen around the nodules of the 0.58-Nb steel and is connected with the other adjacent sponge oxides. In Figure 3h, the total surface of the 0.86-Nb steel was covered with sponge-like oxide, and partial oxides started to transform into coarse oxides.

Analysis of Oxide Surface Morphology
Metals 2019, 9, x FOR PEER REVIEW 6 of 16 Figure 3 shows the surface morphology after 10 and 150 h of oxidation. After 10 h of oxidation, some differences were visible between the steels with and without Nb addition. In the case of 0-Nb steel, the surface morphology was homogeneous and flat, and the oxides grew along the direction of scratches, as shown in Figure 3a. Figure 3c,e,f shows that on the surface of the steels with Nb additions of 0.29, 0.58, and 0.86, some nodules were observed with average sizes of 1, 3, and 6 μm, respectively. Furthermore, the oxides formed on the nodule-free sites of Nb-added steels were homogeneous and flat, similar to those in 0-Nb steel. The surface morphology of 0-Nb steel after 150 h oxidation is shown in Figure 3b, in which the oxides remained homogeneous but with a slight increase in the oxide grain size. Figure 3d shows that the nodules of 0.29-Nb steel grew after 150 h of oxidation. Some nodules formed relatively close to the others, and the sites without nodules remained homogeneous. A significant change in the morphology of oxide was found in 0.58-Nb and 0.86-Nb steels. In Figure 3f, a porous, sponge-like oxide can be clearly seen around the nodules of the 0.58-Nb steel and is connected with the other adjacent sponge oxides. In Figure 3h, the total surface of the 0.86-Nb steel was covered with sponge-like oxide, and partial oxides started to transform into coarse oxides.  Table 4 shows the volume fractions of Cr2O3, spinel (Cr, Mn, Fe)3O4, and Fe2O3 phases calculated using the TOPAS software. The oxides of 0-Nb steel were composed of ~65-74% Cr2O3 and ~21-32% M3O4, but Fe oxide was not detected. The oxides of 0.29-Nb steel had no Fe2O3, but the volume fraction of Cr2O3 increased to ~76-87% and that of M3O4 decreased to ~8-20%. The oxides of 0.58-Nb steel consisted of ~76-92% Cr2O3 and ~8-17% M3O4 within 100 h, and ~40% Fe2O3 was found after 150 h of oxidation. The 0.86-Nb steel was mostly composed of Cr2O3 (~76-93%) before 75 h of exposure, but the volume fraction of Fe2O3 in oxidized specimens increased from ~9 to ~53% in 100-150 h of exposure and became the major corrosion product in the outer oxide layer. GI-XRD results revealed that the volume fraction of Cr2O3 increased with Nb addition, and Fe2O3 was formed. Table 4. Volume fractions of Cr2O3, M3O4, Fe2O3, and γ phases in 22Cr25NiWCoCu stainless steels with 0-0.86% Nb after oxidation in ambient air at 900 °C for an annealing time of 10-150 h obtained from grazing incidence X-ray using software TOPAS.    , and γ phases in 22Cr25NiWCoCu stainless steels with 0-0.86% Nb after oxidation in ambient air at 900 • C for an annealing time of 10-150 h obtained from grazing incidence X-ray using software TOPAS.

Analysis of Oxide Layers at Cross-Section
The microstructures of the oxide layer at cross-section after 10-150 h are shown in Figure 4. A significant difference in the microstructure was found between the steels with and without Nb addition. Nodules were observed in steels with Nb addition, as shown in Figure 4b-d, but not in 0-Nb steel as shown in Figure 4a. Furthermore, oxide layers with nodules were thicker than those without nodules, and precipitates were found around the nodules. After 150 h, 0-Nb steel had a 3 µm-thick oxide layer without nodules. The formation of nodules was observed in 0.29-Nb steel, and the 10 µm-thick oxide layer of the nodule became loose. The oxide layers of 0.58-Nb and 0.86-Nb steels had thicknesses of~12 and~15 µm, respectively. The 0.86-Nb steel had many porous structures, whereas the 0.58-Nb steel had fewer.

Analysis of Oxide Layers at Cross-Section
The microstructures of the oxide layer at cross-section after 10-150 h are shown in Figure 4. A significant difference in the microstructure was found between the steels with and without Nb addition. Nodules were observed in steels with Nb addition, as shown in Figures 4b-d, but not in 0-Nb steel as shown in Figure 4a. Furthermore, oxide layers with nodules were thicker than those without nodules, and precipitates were found around the nodules. After 150 h, 0-Nb steel had a 3 μm-thick oxide layer without nodules. The formation of nodules was observed in 0.29-Nb steel, and the 10 μm-thick oxide layer of the nodule became loose. The oxide layers of 0.58-Nb and 0.86-Nb steels had thicknesses of ~12 and ~15 μm, respectively. The 0.86-Nb steel had many porous structures, whereas the 0.58-Nb steel had fewer.    Mn, Cr, and O but no Fe oxide, as shown in Figure 5f. Furthermore, the oxides of Cr2O3 and spinel (Cr, Mn)3O4 were columnar and did not show a preferred orientation, as displayed in Figure 5c,d. The phase map in Figure 6e and the EDS line scans in Figure 6f,g show that the major oxide of 0.29-Nb steel was Cr2O3, and spinel was found at the steel's outer layer. Results of EDS line scan also show an increase in Fe and a decrease in Cr at the bottom of the nodule in Figure 6f. Some of the oxide grains were almost columnar, but some were almost equiaxed where the oxides Cr2O3 and spinel exhibited small grain sizes and random orientations, as shown in Figure 6c,d. The phase map in Figure 6e and the EDS line scans in Figure 6f,g show that the major oxide of 0.29-Nb steel was Cr 2 O 3 , and spinel was found at the steel's outer layer. Results of EDS line scan also show an increase in Fe and a decrease in Cr at the bottom of the nodule in Figure 6f. Some of the oxide grains were almost columnar, but some were almost equiaxed where the oxides Cr 2 O 3 and spinel exhibited small grain sizes and random orientations, as shown in Figure 6c,d. In the case of 0.58-Nb steel, the phase map in Figure 7e shows that some large-sized spinel formed inside the nodules. In addition, the EDS line scan in Figure 7f shows that Fe signals were detected in the center and top surface of the nodule. A low content of Cr was also detected at the center of the nodule. These observations revealed that the high consumption of Cr led to Cr depletion around the Nb precipitates and, subsequently, led to iron oxide formation. Fine equiaxed grains of Cr2O3 and spinel at the oxide layer were observed, and no preferred orientation was found in the oxide grains, as shown in Figure 7c,d. In the case of 0.58-Nb steel, the phase map in Figure 7e shows that some large-sized spinel formed inside the nodules. In addition, the EDS line scan in Figure 7f shows that Fe signals were detected in the center and top surface of the nodule. A low content of Cr was also detected at the center of the nodule. These observations revealed that the high consumption of Cr led to Cr depletion around the Nb precipitates and, subsequently, led to iron oxide formation. Fine equiaxed grains of Cr 2 O 3 and spinel at the oxide layer were observed, and no preferred orientation was found in the oxide grains, as shown in Figure 7c,d. In 0.86-Nb steel, the major oxides in the outer layer were hematite and magnetite, the inner layer contained (Cr, Fe, Mn)3O4, and a transition layer had Cr2O3 between both layers, as displayed in

Effect of Nb Addition on Oxidation Behavior
An increase in Nb addition caused a decrease in the average grain size, as shown in Figure 1. According to TGA curves shown in Figure 2a, the steels with Nb additions had higher oxidation rates before 50 h than those without Nb addition. The reduction of grain size by Nb addition increased the density of the grain boundary, thus offering many outward diffusion paths for Cr and, consequently, enhancing the oxidation rate.
In 1957, Hart [25] proposed an effective volumetric diffusion that was associated with matrix and dislocation. Following Hart's concept, Smeltzer [26] proposed the effective diffusion constant as follows:

Effect of Nb Addition on Oxidation Behavior
An increase in Nb addition caused a decrease in the average grain size, as shown in Figure 1. According to TGA curves shown in Figure 2a, the steels with Nb additions had higher oxidation rates before 50 h than those without Nb addition. The reduction of grain size by Nb addition increased the density of the grain boundary, thus offering many outward diffusion paths for Cr and, consequently, enhancing the oxidation rate.
In 1957, Hart [25] proposed an effective volumetric diffusion that was associated with matrix and dislocation. Following Hart's concept, Smeltzer [26] proposed the effective diffusion constant as follows: where D eff is the effective diffusion, f is the parameter associated with grain boundaries and grain size, D L is the lattice diffusivity, and D GB is the grain boundary diffusivity. Here, f is equal to 2δ/d, in which δ is the grain boundary width at approximately 0.5 nm, and d is the average grain size. Equation (2) can be rewritten as: According to self-diffusion in face-centered cubic metals [27], D GB is~6 orders of magnitude higher than D L at an oxidation temperature of 900 • C. Hence, Equation (3) can be simplified as: With the help of Equation (4), the D eff for all steels is calculated as follows: Apart from refining the grain size, Nb addition increased the formation of Cr 2 O 3 and reduced that of M 3 O 4 , as shown in Table 4. Here, the formation of Cr 2 O 3 and M 3 O 4 lies in the inner and the outer oxide layers, as shown in Figures 5-8. Compared with the diffusion coefficients of Cr and Mn [28], that of Mn was 2 orders of magnitude higher than that of Cr. Therefore, spinel (Cr, Mn) 3 O 4 formed at the outer oxide layer in 0-Nb steel. GI-XRD results show that the formation of Cr 2 O 3 increased with Nb addition before 100 h. Cr 2 O 3 was the major oxide in Nb-containing and 0-Nb steels. The reduction of grain size has a beneficial effect on the formation of Cr 2 O 3 [14][15][16][17][18][19].

Effect of Precipitates on the Oxidation Behavior
According to the above-mentioned results in Section 3.5, 0.86-Nb steel should have a higher oxidation rate than 0.58-Nb steel; however, the latter obtained the highest oxidation rate. This finding indicates that the grain size did not affect the oxidation rate. According to a previous study [13], the volume fractions of Nb (C, N) precipitates for 0-Nb, 0.29-Nb, 0.58-Nb, and 0.86-Nb were 0, 3.1, 5.2, and 3.5 wt.%, respectively.
According to the microstructure of the surface ( Figure 2) and cross-sectional analysis (Figures 4-8), nodules were only found in the steels with Nb addition. Small-sized nodules disappeared, and large-sized ones had grown with prolonged oxidation. Porous, sponge-like oxides initially formed at the nodules and then immediately transformed into large-sized oxides. From the cross-sectional microstructure of the oxide layer, a Nb (C, N) precipitate was always found embedded in the nodule, indicating that the formation of nodules was induced by these Nb (C, N) precipitates. These phenomena may be attributed to the high Cr consumption rate at the interface of precipitates and substrate. Therefore, the effect of precipitates contributes to the oxidation rate. This finding explains the higher oxidation rate of 0.58-Nb steel at the initial stage.
At the initial stage, a protective Cr 2 O 3 layer existed on the surface, and the Cr 2 O 3 oxide was the major component of oxide layer before 100 h, as shown in Table 4. Considering the effects of grain size and precipitates, 0.86-Nb steel should have the highest weight gain per unit area. After a long oxidation time, the 0.86-Nb steel had the highest weight gain per unit area. After 150 h, the Fe 2 O 3 oxide became the major component of 0.58-Nb and 0.86-Nb steels. The high consumption of Cr for 0.86-Nb steel led to Cr depletion around Nb precipitates, as shown in Figure 7f. Furthermore, the Nb (C, N) at the center of the nodules led to the breakdown of the oxide layer. Thus, Cr cannot be continuously supplemented from the matrix due to Cr depletion, that is, when the concentration of Cr is lower than~15 wt.% [14]. Fe may then start to form into a porous oxide layer of Fe 2 O 3 , resulting in the increased oxidation rate compared with that of the compact oxide layer of Cr 2 O 3 . The formation of nodules can be induced by the concentration of the protective oxide former [29][30][31][32], inhomogeneity of the chemical composition of alloy [33], surface roughness [34], breakdown of the protective oxide layer [35][36][37], and grain boundaries [38,39]. Allen et al. reported that the mounds (nodules) result from the phase transformation of Nb (C, N) during high-temperature oxidation [11]. In Tayler's study, Nb (C, N) was found at the center of these mounds during phase transformation [12].
According to the above discussions, a schematic diagram for the formation of the oxide layer in terms of Nb (C, N) precipitates is summarized in Figure 9 to understand the growth evolution of the oxide layer. At Stage I, a Cr 2 O 3 oxide layer formed instantly on the specimen surface.
According to the above discussions, a schematic diagram for the formation of the oxide layer in terms of Nb(C, N) precipitates is summarized in Figure 9 to understand the growth evolution of the oxide layer. At Stage I, a Cr2O3 oxide layer formed instantly on the specimen surface.
Oxide clusters were observed at the region with Nb(C, N) particles, which were imbedded in nodules. The oxide layer grew faster in the region with nodules than in the area without nodules. At Stage II, the region with nodules grew at a high Cr consumption rate, resulting in Cr depletion. Consequently, Fe oxides were initially formed at Cr-depleted zones around precipitates. Finally, at Stage III, a thick oxide layer consisting of Fe2O3 at the outer layer and (Cr, Fe, Mn)3O4 at the inner layer was formed.

Conclusions
An increase in Nb content refined the grain size from 426 to 90 μm, promoted the formation of Cr2O3, and avoided the formation of M3O4. In addition, the volume fraction of Nb(C, N) precipitates increased from 0% to 5.2% when the Nb content increased from 0-0.58 wt.%, but it decreased to 3.5% when the Nb content was 0.86 wt.%.
After 0-10 h, a Cr2O3 oxide layer formed, and oxide-Nb(C, N) particle clusters appeared in nodules. After 10-100 h, a high Cr consumption rate was observed in the nodule-forming regions, resulting in Cr depletion. Consequently, Fe oxides were generated at the Cr-depleted zones with Nb(C, N) precipitates. Finally, after 100 h, a thick oxide layer composed of an Fe2O3 outer layer and (Cr, Fe, Mn)3O4 inner layer was synthesized.  Oxide clusters were observed at the region with Nb (C, N) particles, which were imbedded in nodules. The oxide layer grew faster in the region with nodules than in the area without nodules. At Stage II, the region with nodules grew at a high Cr consumption rate, resulting in Cr depletion. Consequently, Fe oxides were initially formed at Cr-depleted zones around precipitates. Finally, at Stage III, a thick oxide layer consisting of Fe 2 O 3 at the outer layer and (Cr, Fe, Mn) 3 O 4 at the inner layer was formed.

Conclusions
An increase in Nb content refined the grain size from 426 to 90 µm, promoted the formation of Cr 2 O 3, and avoided the formation of M 3 O 4 . In addition, the volume fraction of Nb (C, N) precipitates increased from 0% to 5.2% when the Nb content increased from 0-0.58 wt.%, but it decreased to 3.5% when the Nb content was 0.86 wt.%.
After 0-10 h, a Cr 2 O 3 oxide layer formed, and oxide-Nb (C, N) particle clusters appeared in nodules. After 10-100 h, a high Cr consumption rate was observed in the nodule-forming regions, resulting in Cr depletion. Consequently, Fe oxides were generated at the Cr-depleted zones with Nb (C, N) precipitates. Finally, after 100 h, a thick oxide layer composed of an