New insight on early oxidation stages of austenitic stainless steel from in situ XPS analysis on single-crystalline Fe–18Cr–13Ni

Abstract In situ X-ray photoelectron spectroscopy real-time measurements and angular-dependent high resolution core level analysis were used for the first time to investigate the Cr enrichment and oxide growth mechanisms on a model 304 austenitic stainless steel surface in the very initial stages of oxidation leading to pre-passivation. The oxidation kinetics was followed for increasing oxygen exposure and temperature, revealing an early nucleation regime (for exposure


Introduction
Stainless steels (FeCr-based alloys) are widely used because of their high corrosion resistance resulting from the formation of a continuous and protective surface oxide layer, the passive film. Numerous surface analytical studies have shown that the passive film is only a few nanometers thick and markedly enriched in Cr(III) oxide/hydroxide species [1][2][3][4][5][6][7][8][9][10][11][12][13]. Recently, it has been suggested from nanometer scale studies on austenitic stainless steels that and around 300 • C [20,22,37], the latter because of its industrial relevance. However, no oxidation study of Fe-Cr-Ni alloys was conducted systematically at different temperatures.
Regarding oxygen exposure, experiments were carried out at values larger than 1 000 L (1 L = 1.33×10 −6 mbar·s), under oxygen pressure usually higher than 10 −6 mbar. According to our kinetics data presented hereafter, this is far beyond the initial stages of oxidation, defined as the range of exposure for which the surface varies the most. The understanding of the mechanisms of oxide growth and Cr enrichment in the initial stages of oxidation is essential to understand the structure and properties of ultra-thin oxide films. However, to the best of our knowledge, there was no investigation of Fe-Cr-Ni alloy surfaces during and/or after very low oxygen exposure (< 100 L), neither of the mechanism of ultra-thin oxide film formation.
Nitrogen segregation has been observed in some studies performed on N-containing stainless steel [22,[38][39][40] and resulted in the formation of surface chromium nitride compounds.
It was deduced by some authors that the co-segregation of chromium with nitrogen could increase the proportion of surface chromium oxides [41].
Here, we bring new insight on the oxidation kinetics and evolution of the surface composition and stratification during the initial oxidation of austenitic stainless steel. Nitrogen segregation is also discussed. The study was performed by in situ XPS on a (100)-oriented Fe-18Cr-13Ni single crystal (a model of the most common 304 grade of austenitic stainless steel) during the initial stages of oxygen exposure (< 100 L) at RT, 150 • C and 250 • C.
The work is aimed at providing new comprehensive knowledge for the surface treatments of Fe-Cr-Ni alloys for improved resistance to initiation of localized corrosion of stainless steel.

Experimental
Firstly, the clean surface and nitrogen segregation were characterized. Then, the evolution of the oxygen uptake during oxygen exposure was monitored in situ to determine the oxidation kinetics. For that purpose, a novel specific XPS acquisition mode was adopted to record the real-time evolution of the photoelectron spectra during oxidation. In this way, the continuity of signal variation was achieved which strengthened greatly the kinetics monitoring. Afterwards, oxidation was conducted at exposures selected according to the kinetics data. The high resolution spectra were decomposed in order to determine the chemical states and their proportions. Based on angle-resolved data, oxide models were proposed to discuss the surface composition and stratification, as well as the thickness and formation mechanism of the oxide films and their Cr enrichment. A complementary Scanning Tunneling Microscopy (STM) study of the local topographical and structural alterations of the surface brought by initial oxidation will be reported separately [42].
The experiments were performed in a multi-chamber UHV system including a preparation chamber and two analysis chambers (XPS chamber and STM chamber), maintaining a base pressure < 10 −10 mbar. The preparation chamber is equipped with an Ar + cold cathode ion gun for sputtering and resistive heating filament for annealing. Samples are heated from back side and temperature is controlled through a thermocouple attached on the manipulator in contact with the sample.
A (100)-oriented Fe-18Cr-13Ni single crystal with 99.999% purity was used. The sample was mechanically and electrochemically polished before introduction in the UHV system (procedure described elsewhere [23]). After introduction, the sample surface was cleaned by cycles of Ar + ion sputtering (1 kV, 10 µA, 10 minutes) and annealing (700 • C, 10 minutes) in the preparation chamber. Sputtering and annealing cycles were applied between each oxidation experiment. The chemical and structural quality of the surface were controlled by XPS and low energy electron diffraction (LEED), respectively.
Gaseous oxygen (99.999% purity) was introduced via a leak valve directly connected to the preparation chamber or via an oxidation cell installed in the XPS analysis chamber.
XPS analysis was conducted with the Argus spectrometer, using the monochromatic XM1000 Mkll Al Kα X-ray source (1486.6 eV) both from Scienta Omicron. The Argus detector with 128 channels has a high spatial resolution and a high counting rate that allow to record electron spectra over short time period. The in situ XPS investigations were carried out using the Snapshot detection mode, recording a spectrum every 10 seconds in real-time monitoring during oxidation. Thousands of spectra were recorded to describe the variation 4 of oxygen (or nitrogen) peak intensity during oxidation (or annealing). Most of oxidation treatments were carried out at RT, 150 • C and 250 • C with O 2 pressure ≤ 10 −8 mbar. Such low oxygen pressure was kept in order to monitor the 0-100 L exposure range with good precision. The pressure in the chamber was continuously recorded and the exposure to O 2 was calculated from the integration of oxygen pressure as a function of oxidation time.
According to the in situ kinetics data, series of oxidations under very low oxygen exposures were carried out for detailed chemical analysis of the surface evolution. Survey and high resolution spectra of Fe 2p, Cr 2p, Ni 2p, O 1s, N 1s and C 1s core levels were recorded after each oxidation treatment, with constant pass energy of 50 eV and 20 eV, respectively.
Both take-off angles of 45 • and 90 • were studied in order to determine the distribution of species along the direction normal to the surface. All obtained XPS spectra were referenced to the Fermi edge. Three exposures less than 10 L and nine exposures above were chosen for decomposition analyses. More than 400 high-resolution spectra were recorded and analyzed in this work. After exposure to about 300 L at 250 • C, no more significant change in the substrate signal was observed.
Data analyses were performed using the CasaXPS software (Version 2.3.17) [43]. For reconstruction of the high resolution spectra, several constraints were considered to achieve accuracy and consistency.
The width of analyzed regions was carefully adjusted to ensure a suitable background contribution. For Fe and Cr, the 2p core levels were fitted by adjusting the intensity ratio between 2p 3/2 and 2p 1/2 components to values matching the theoretical ratio. For nickel, as there was no overlapping between the 2p 3/2 (852.8 eV) and 2p 1/2 (869.6 eV) peaks, only Ni 2p 3/2 was considered. In order to improve the accuracy of the quantitative determination, only the 2p 3/2 data were used to calculate the concentration and to establish the oxide model as 2p 1/2 components have larger uncertainties.
Background subtraction methods can influence the quantitative values of Fe and Cr chemical states concentrations [25,31,44,45]. It was found that the chemical states proportions extracted by using different background subtraction can vary by ±5% for iron oxide and ±3% for chromium oxide [25]. To minimize the effect of background subtraction on components concentration, an adjustable Shirley background [46] was adopted for our peak fitting.
A Lorentzian asymmetric line shape, with the same asymptotic behavior as Doniach-Sunjić component [47], convoluted by a Gaussian shape was used for fitting the metallic component peaks. The indexes of asymmetry follow the order: metal > oxide > satellite ≈ 0. A simple product of Gaussian and Lorentzian line shapes was used for the non-metallic components.
Binding energy (BE) and full width at half maximum (FWHM) are two important parameters affecting the reconstruction of the synthetic curve shape. Although metal compounds states have been largely studied by XPS [25][26][27][28][29][30][31]35], BE and FWHM can be different according to the used sample and spectrometer. There are few data in the literature about the BE position of 2p 1/2 core level peaks and shake-up satellite peaks. As for FWHM, it depends on the chemical environment. Normally, the FWHM values of the different peaks follow the order: satellite > oxide(nitride) > metal and 2p 1/2 > 2p 3/2 .
Once the fitting parameters were determined, they were fixed for all the high-resolution spectra recorded at take-off angles of 45 • and 90 • . Only the relative peak intensities were treated as the variable for peak fitting. Table 1 compiles the fitting parameters determined from the XPS spectra reconstruction.

XPS fitting parameters
They are consistent with typical values found in other works [25,28,29]. The chemical shifts of the different Fe oxides with respect to Fe 0 metal are 1.7 eV, 2.6 eV and 3.5 eV for the components assigned to Fe 2+ (in Fe 3 O 4 matrix), Fe 2+ (in FeO-like matrix) and Fe 3+ (in Fe 2 O 3 matrix), respectively. Spin orbit splitting is 13.1 eV between the Fe 2p 3/2 and Fe 2p 1/2 main peaks and 13.5 eV between the Fe 2p 3/2 and Fe 2p 1/2 shake-up satellite peaks, which is close to the values from recent studies [25,26,28,31,35]. BE separations between the Fe 2p 3/2 main peaks and Fe 2p 3/2 satellite peaks are 4.5 eV and 5.9 eV for Fe 2+ and Fe 3+ , respectively. . These values are of the order reported in the literature [5,25,27,28,30]. Spinorbit splitting between the Cr 2p 3/2 and Cr 2p 1/2 peaks are 9.4 eV, 9.5 eV, 9.6 eV and 9.0 eV for Cr 0 , Cr 3+ (CrN), Cr 3+ (Cr 2 O 3 ) and Cr shake-up satellite peaks, respectively. BE separation for Cr satellite peak is 14.5 eV which is 2-3 eV smaller than reported from a Cr 2p curve fitting study of a pure Cr 2 O 3 powder sample [25].
The 2p 3/2 and 2p 1/2 doublet intensity ratios for Fe and Cr vary from 0.49 to 0.52, which is close to the theoretical value of 0.5.
As for the FWHM, compared to the metallic peaks, it is confirmed that the oxide peaks are wider. Satellite peaks have larger FWHM values in the range about 3.5-4.3 eV due to the strong overlap with the main peaks and the background. The metallic peaks contribute a lot to the asymmetry of the spectra.

Clean surface characterization and nitrogen segregation
After Ar + sputtering the alloy surface, only the Fe 2p, Cr 2p and Ni 2p core levels were identified indicating the absence of contaminants. No carbon or nitrogen were detected. It proves that the nominal amounts of these elements are below the detection limit of the XPS (< 0.5 at%). The surface composition (at%) was determined as 75Fe-15Cr-10Ni, depleted in Cr at the surface compared to the nominal bulk value (69Fe-18Cr-13Ni). However, after annealing at 700 • C for 10 minutes, a N 1s feature appeared at around 397.2 eV, indicating that nitrogen, below the detection limit after sputtering, segregated on the surface upon thermal treatment. The details of the Fe 2p, Cr 2p, Ni 2p and N 1s spectra are discussed in Section 3.4. The composition of the surface obtained after annealing was (at%) 54Fe-29Cr-9Ni with 8 at% N. It should be noted that these composition values were deduced from quantitative analysis of the XPS spectra assuming an identical surface distribution for all elements. Their relative changes evidence that Cr and N segregate at the surface upon annealing. The form under which these elements segregate will be discussed in section 3.4. Fig. 1 depicts the real-time segregation of nitrogen at the surface as illustrated by the plot of the normalized area of N 1s peak versus annealing time ( Fig. 1 (a)) and the variation of the segregation rate as a function of annealing temperature ( Fig. 1 (b)). The data were acquired in three successive experiments (as seen from the two discontinuities in the plots) using the XPS Snapshot mode.  about 620 • C before dropping quickly to zero after surface saturation at 680 • C, showing that nitrogen segregation depends effectively on temperature. Since nitrogen starts to segregate at 400 • C, it is nearly impossible to prepare an annealed and well-structured metallic surface that is nitrogen-free. As shown in Fig. 1 (b), the nitrogen segregation kinetics plot can be fitted by the following equation: where k is the Boltzmann constant and E the activation energy for N segregation. The fitted value of E is 72±5 kJ/mol. This is less than the value found for N segregation in a ferritic Fe-Mo alloy [48]. The N saturation of the oxide-free surface would therefore be easier to achieve in our case of an austenitic alloy.
Previous studies have discussed the influence of nitrogen on oxidation and wet corrosion of stainless steel [5,22,38,40,41]. It was reported that Cr and N co-segregate at the surface, as confirmed in the present work, which may promote the selective oxidation of metallic Cr enriched at the co-segregated surface. Cr and N co-segregate in the form of chromium nitride compounds [22,40,41], which is discussed in the following.

Surface oxidation kinetics
In order to determine the influence of oxygen pressure and temperature during the initial oxidation of the Fe-18Cr-13Ni(100) surface, oxidation runs were performed at RT, 150 • C and 250 • C at different oxygen pressures on the sample surface pre-annealed at 700 • C. Fig. 2 presents the evolution of the normalized O 1s peak area at these three temperatures as a function of O 2 exposure (100 L) as measured in real-time by in situ XPS. Fig. 2 (b) enlarges the region of the first 25 L of exposure. This is the first observation of oxidation kinetics at the very early stage. In previous studies, oxidation experiments of iron or stainless steel were carried out at P O 2 > 10 −6 mbar [20,21,24,39,49]. At these pressures, the detailed information on oxide film growth in the very first stages is lost. This is why our study was limited to 10 −8 mbar for low exposures (< 100 L) and did not exceed 10 −6 mbar for larger ones, as shown by the insets in Fig. 2 (a). Since there was no evidence of discontinuity in the kinetic curves recorded at RT and 150 • C, the change of oxygen pressure is concluded to have no apparent influence on the oxidation kinetics in the conditions tested.
At a fixed temperature, the O 1s intensity first increases rapidly before slowing down and leveling. At RT, the oxygen uptake reaches saturation when oxygen exposure exceeds 10 L ( Fig. 2 (b)). At 150 • C or 250 • C, the oxygen uptake is reduced after exposure to 20 L ( Fig. 2 (a)), but no oxidation saturation is observed upon oxygen exposure up to 81 L and to 35 L, respectively. The inset in Fig. 2 (b) presents the oxygen uptake rate ( dO dt ) recorded in situ in the 0-10 L exposure range at the three temperatures. It shows that the oxidation rate reaches a maximum at 2-3 L exposure and then decreases to stabilize to around zero at 10 L at RT. At 150 • C and 250 • C, a slow oxidation rate is still observed after 10 L exposure.
As evidenced in Fig. 2 (b), as the temperature increases, the initial oxygen uptake is reduced. This is consistent with the decrease of the physical adsorption time (Arrhenius law) with, as a consequence, the decrease of the reaction rate at very low exposure (< 10 L). At RT, a component was detected at 531.4 eV BE in the O 1s region, close to the value attributed to adsorbed molecular oxygen [50]. This component, distinct from the oxide component peaks, was not observed in the spectra obtained at 150 and 250 • C ( Fig. S1 in supplementary data). This is consistent with the initial reduction of the sticking coefficient at these temperatures. In contrast, Fig. 2 (b) shows that for exposures larger than 10 L, the oxygen uptake and its saturation level increase at higher temperature. This is consistent with higher temperature promoting atomic displacement at the surface and/or atomic transport between surface and sub-surface of the sample, thus resulting in the formation of more surface oxide by oxide growth.
Studies reported in the literature were carried out at 10 −6 mbar or higher pressure for exposures larger than 1 000 L [20,21,24], overlooking the initial stages of oxidation.
Oxidation saturation was also observed by XPS on Fe(111) as recently reported by Davies et al. [21]. It was found that surface oxygen saturates after an oxygen exposure of about 100 L at 280 K (under P O 2 ≈ 10 −6 mbar), whereas about 10 L are required to reach saturation on our alloy surface at RT. On Fe(111), the rate of oxygen uptake was almost the same at 280 K and 500 K, but the oxygen coverage reached at 500 K is higher [21] in agreement with the increasing saturation level measured at 150 • C and 250 • C in the present work.
Thus, it is shown by in situ XPS analysis of the initial stages of the Fe-18Cr-13Ni surface oxidation that there is no apparent effects of oxygen pressure on the oxidation kinetics in the tested pressure range (< 10 −6 mbar) and that the effects of temperature are different depending on oxygen exposure. The exposure of about 10 L can be estimated as the threshold below which the oxidation rate would be essentially limited by the sticking coefficient suggesting a nucleation phase reaching saturation at this exposure. Beyond 10 L, atomic displacement and/or transport would become rate limiting suggesting that the oxidation process has entered the growth phase leading to the increase of the surface oxide formation with increasing temperature. contamination was detected. The reconstruction of all Fe 2p, Cr 2p and N 1s spectra (more than 300 spectra) was performed using the fitting parameters compiled in Table 1. Fig. 3 shows the reconstruction of Fe 2p, Cr 2p and N 1s spectra measured at 45 • take-off angle after exposures of 0 L, 3 L, 6 L and 14 L at RT.
For the sake of clarity, the main peaks and satellites have been summed up for each oxidation state of Fe and Cr. The Fe 2+ and Fe 3+ components were identified based on the spectra obtained at higher temperature where they are more prominent. Comparatively, the evolution of the Cr 2p spectra was more marked at RT (as seen on Fig. 3), enabling the Cr 3+ oxide and nitride components to be better identified.
The spectra presented in Fig. 3 show that the main cations in the surface oxide grown  and N m (CrN surf ace , CrN bulk , N bulk and N minority in Table 1) were also distinguished before and after oxygen exposure. For all core levels, there are no significant differences between the spectra at 6 L and 14 L, indicating that the surface reaction reached saturation at RT in agreement with the in situ kinetics data presented in Fig. 2.
Even if one considers the uncertainty in the peak fitting, the spectra presented in Fig.   3 [17,18] addressing the conditions of FeO formation as compared to the phase diagram for the iron-oxygen system.
Graat et al. [36] proposed the presence of FeO at RT for low oxygen exposure of 22 L. Our finding of Fe 2+ species is in agreement with this work [36] and that of Roosendaal et al. [18] but we also find Fe 3+ species. Lin et al. [17] indicated that FeO x , the FeO-like oxide, was the nucleating oxide formed after low oxygen exposure (10 L to 50 L) at RT.  Fig. 3 that chromium oxidizes more rapidly than iron because of its higher oxide to metal intensity ratio when oxygen exposure is lower than 14 L, i.e. in the nucleation phase. This is consistent with the results of Lince et al. [16].    Fig. 4

Layered surface composition
Although previous studies [16,19,20,22,24] have shown that no abrupt interface is formed between oxide and metal, layered models have been widely applied to describe the stratification of oxidized surfaces and quantify their composition. In this work, we have performed comparative analysis of the XPS intensities measured at 45 • and 90 • take-off angles of the emitted photoelectrons for each chemical component in order to define its relative distribution along the surface normal and thus to determine the in-depth distribution of the surface species.

20
Layered models were applied to quantify the surface composition and equivalent oxide film thickness. The resulting values do not account for the inhomogeneity of the growth process that can be expected due to local structural and chemical variations at the alloy surface, in particular in the nucleation phase. This aspect will be discussed separately in an article based on STM data. Still, the values reported hereafter allow us to discuss the highly relevant aspect of chromium enrichment and its variation with the build-up of the oxide film.  Fig. 5 (b). The equivalent thickness values were calculated from the intensity ratios between the metal and oxide components of the Fe and Cr 2p 3/2 core level based on the exponential attenuation of the photoelectron intensity with depth [15,55,56].
These thicknesses are given in Fig. 6.
These composition and thickness results should be taken with caution as they are calculated from a simplified model. However these quantities are relevant for comparison with previous studies and relative changes in concentrations help us to understand the underlying mechanism that govern the oxide growth process.
Based on an analysis of angular dependent photoemission spectra, a more precise model of the surface may be proposed. Fig. 6 presents the schematic view of the species distribution as a function of oxygen exposure at RT, 150 • C and 250 • C. This is the first complete stratification scheme of oxidized surface of a stainless steel at the initial oxidation stage.
The equivalent thicknesses of the oxide regions ranges between 1.1-2.2 nm, calculated based on a simplified homogeneous model including the interfacial oxide/metal mixed region.
This is consistent with oxide films reported to be 1-4 nm thick when formed on Fe or Fe-base alloys [16, 17, 19, 20, 22-24, 36, 57]. At RT, the calculated thickness is stable at 1.6-1.7 nm after the 6 L exposure, in agreement with the saturation observed in the oxygen uptake ( Fig. 2 and 4) and with 1.8-2 nm thickness reported for the native oxide film formed on the same austenitic stainless steel surface [23]. At 150 and 250 • C, the oxide equivalent thickness increases with exposure also following the same trend as the oxygen uptake (Fig.   2). However, the increase observed in the oxygen uptake at saturation with temperature increasing to 150 and 250 • C (Fig. 4) is According to previous studies [22,24], there would be no clear separation between Cr oxide and Fe oxide in the oxidized surface. An amorphous oxide FeCr 2 O 4 -like phase could exist after a large exposure at a higher temperature (1000 L, 300 • C) [20]. However, our angle- Upon oxidation at 250 • C, the same layered structure of oxide film is found with a mixed Cr 3+ /Fe 3+ layer above the Fe 2+ -containing interfacial layer. Moreover, the results show that the structure is determined not only by the competitive oxidation of Cr and Fe, but also by their atomic mobility. At 6 L (Fig. 6), metallic Cr is found preferentially in the interfacial region containing Fe 2+ and Cr-nitrides, suggesting its transport from the modified alloy region in order to sustain Cr oxidation since the oxide film is equally enriched in chromium 24 (the cation fraction is 59 at% for Cr 3+ , 13 at% for Fe 2+ and 28 at% for Fe 3+ in Fig. 5 (b)).
As a result, the modified alloy region is enriched with metallic Fe in its upper part. Thus, the competitive oxidation of iron is promoted gradually after 10 L exposure and Fe oxidation prevails over Cr oxidation at the oxide-metal interface.  Table 2 compiles the data of surface composition and thickness obtained in this work, the data for a native oxide film formed by exposure to ambient air and a passive film formed by anodic polarization on the same alloy surface at RT, and the data for a thermal oxide film formed by exposing a Fe-17Cr alloy surface to oxygen at 327 • C. At RT, the surface saturated under oxygen in the present work is found more enriched in Cr 3+ than the surface aged for 20 h under ambient air [23]. This can be assigned, at least partly, to the Cr-N co-segregation resulting from surface preparation in our work and that would promote the selective oxidation of chromium as previously reported [22]. However, the oxide film grown at RT under oxygen in the present work in much less enriched in Cr 3+ than a passive film produced in acidic aqueous solution [23], which can be assigned to the absence of preferential dissolution of iron oxides promoting chromium enrichment in acidic media, as also found on a ferritic Fe-17Cr alloy oxide film grown by exposure to oxygen [22] and anodic polarization [58]. With the increase of temperature to 250 • C, it is found in the present work that the chromium oxidation is hindered by the depletion of metallic Cr underneath the growing oxide layer, while the diffusion of metallic Fe from the bulk enables to sustain iron oxidation, thus leading to a gradual decrease of the Cr 3+ enrichment with increasing exposure. Compared to the thermal oxidation of a ferritic Fe-17Cr sample (also with surface nitrides) at 327 • C [22], the variation of the Cr 3+ enrichment on our austenitic sample follows the same trend.

Conclusion
Real-time monitoring of the evolution of a model 304 stainless steel monocrystal surface, Fe-18Cr-13Ni(100), in the early stages of oxidation at RT, 150 • C and 250 • C was achieved using in situ XPS in Snapshot mode analysis and complemented by high resolution core level analysis of the growth mechanisms and Cr enrichment. The work revealed, for the first time to our knowledge, the mechanisms of stratification during the early build-up of the surface oxide on a model austenitic stainless steel surface at different temperatures.
The results show that, on the oxide-free surface, nitrogen segregation is detected upon heating at 400 • C and increases significantly above 500 • C to reach a saturation at 680 • C. The activation energy of nitrogen segregation is 72±5 kJ/mol. Marked Cr enrichment of the as prepared metallic surface was observed as a result of Cr segregation and Cr-N co-segregation with N forming mainly three-dimensional CrN b particles with minority two-dimensional CrN s and interstitial nitrogen N b . Surface oxidation destabilizes the 3D chromium nitrides and the oxide films grow above the 2D surface nitrides remaining at the surface.
The temperature dependence of the surface oxygen uptake as measured by real-time monitoring evidences two regimes for the growth of the surface oxide. Up to about 10 L of oxygen exposure, the uptake is faster with a maximum at 2-5 L and decreases with increasing temperature, suggesting a nucleation regime of the surface oxide limited by oxygen adsorption. Beyond 10 L, the uptake is saturated at RT but increases with temperature, suggesting a growth regime of the surface oxide limited by atomic displacement and mobility.
The oxygen pressure in the range below 10 −6 mbar is found to have no effect on the oxidation rate.
In the nucleation regime leading to saturation, Cr 3+ formation is preferential over Fe 3+ formation and Fe 2+ formation is mostly observed in the very first stage of oxidation (0.5 L). Nickel is not oxidized. A surface layer of strongly Cr 3+ -enriched oxide is thus formed over the mainly Fe 2+ oxide (FeO-like) species mixed with the CrN s species in the interfacial layer underneath the Cr 3+ /Fe 3+ mixed layer. Further oxidation beyond saturation at RT causes no marked evolution of the oxide layered structure, composition and thickness. The modified alloy region is still not depleted in metallic Cr and thus able to further grow a Cr 3+ -enriched protective surface layer.
At 150 and 250 • C, the oxide formed in the nucleation regime is also enriched in Cr 3+ but the modified alloy region gets depleted with metallic Cr. As a result, the competitive oxidation of iron is gradually promoted in the oxide growth regime leading to the decrease of the Cr 3+ enrichment in the oxide. Fe oxidation is sustained by diffusion of metallic Fe from sub-surface region to the modified alloy region. At 150 • C, Fe 3+ oxide was found to form at the topmost surface, whereas at 250 • C it was at the oxide-metal interface, suggesting a switch of the oxide growth from predominant iron transport through the Cr 3+ /Fe 3+ oxide at 150 • C to predominant oxygen transport at 250 • C for exposures beyond 35 L.