Introduction

Most high-temperature wrought Ni-based alloys rely for protection against oxidation attack on the formation of a slowly growing, adherent chromia-base surface oxide or scale, which forms due to the selective oxidation of Cr.1 The most common wrought alloys have no or minimal amounts of γ′ and rely on solid solution strengthening by Cr, Mo, Co and W2,3 as well as carbide precipitates (e.g., of the type M23C6 or M6C) for creep strength.4,5 Especially at lower temperatures (e.g., 700°C), intermetallic phases also may contribute to the creep strength of the alloy.6 However, long-term aging at lower temperatures may result in excessive formation of intermetallic and carbide precipitates, which may adversely affect material ductility.6–8

Especially when oxidized at high temperatures (≥900°C), the oxidation resistance of wrought Ni-base alloys will gradually deteriorate due to the preferential removal of Cr from the alloy due to Cr consumption by oxide scale growth and re-growth if the scale spalls. The Cr depletion eventually results in a decrease of the Cr concentration at the alloy-oxide interface to such a low level that the growth of the protective chromia scale can no longer be sustained. This leads to the formation of rapidly growing, less protective oxide phases such as NiO and/or NiCr2O4 and is frequently termed “breakaway oxidation”.9

Before breakaway occurs, the oxidation-induced Cr depletion may also affect material properties (e.g., creep strength, ductility) due to changes in the alloy microstructure,10–12 i.e. the amount and distribution of strengthening precipitates. Obviously, these depletion-induced processes will be more significant in thin-walled components. A commonly encountered example is the subsurface dissolution of Cr-rich carbides,13–18 and the effect of a carbide-free subscale zone on alloy mechanical properties has been previously discussed.8 Thus, to predict component performance in long-term applications, the model must quantify Cr depletion as a function of time and temperature but also its effect on the microstructure.

The present investigation on alloy 230 was part of a larger project to model the lifetime of thin-walled NiCr components. Alloy 230 is a commonly used high-temperature alloy for combustor liners, honeycomb structures and heat exchangers.19 This paper focuses on the microstructural changes as a function of specimen thickness, oxidation time and temperature. Additional results and more theoretical considerations related to oxidation-induced lifetime limits will be presented elsewhere.20

Experimental

The alloy 230 specimens were all fabricated from a 1.2-mm-thick sheet in the solution-annealed condition, supplied by Haynes International. Bulk chemical analyses were performed by inductively coupled plasma–optical emission spectroscopy (ICP–OES) after completely dissolving the metal in aqua regia. The concentrations of C, N, S and O were determined using combustion–infrared analysis. Table I shows that the composition was very close to the nominal values; however, the Co and Fe contents were much lower than the maximum values and the La content was not detected by the ICP-OES process. Specimens of 20 × 10 mm2 were ground to nominal thicknesses of 0.2 mm, 0.3 mm and 0.5 mm and then polished to a P1200 surface finish. Prior to oxidation, the specimens were degreased in ethanol. Oxidation exposures at 950°C, 1000°C and 1050°C in laboratory air were conducted for 1000 h or 3000 h. For the 1000°C and 1050°C experiments at Forschungszentrum Jülich, the specimens were exposed in 50-h cycles and hung on alumina rods in alumina crucibles, which enabled the determination of the specimen mass change, amount of spalled oxide and the total mass gain (i.e. specimen + spalled oxide, in the following termed as “total mass gain”). At 950°C, the specimens were exposed in 10-h cycles with 10 min cooling between cycles and weighed once per week at 16-cycle intervals at Oak Ridge National Laboratory. In the automated vertical cyclic rig, the specimens were held on alumina rods with a Pt-10Rh wire and only the specimen mass changes were determined.21

Table I Chemical composition (in wt.%) of the studied alloy 230 heat determined by ICP-OES and combustion infrared analysis compared with nominal composition supplied by the manufacturing data sheet24
Full size table

After oxidation, the specimens were sputtered with a thin gold layer followed by Ni plating in a NiSO4 bath before standard metallographic cross-section preparation. Sectioned specimens were then analyzed using light optical microscopy, scanning electron microscopy (SEM), energy dispersive x-ray spectroscopy (EDX) using a field emission gun SEM (Zeiss Supra 50VP) and electron backscatter diffraction (EBSD; Zeiss Merlin). Selected specimens were analyzed by glow discharge optical emission spectroscopy (GDOES) whereby quantification procedures described in references22,23 were used.

Thermodynamic calculations of the phase equilibria were carried out with the software package Thermo-Calc v.5 using the database TTNI7. The volume fraction of precipitates as well as the depth of the precipitate-free zone was calculated using the computer software AnalySIS pro 5.0.

Results

Figure 1 shows an example of the mass change data at 1000°C in laboratory air illustrating the typical effect of specimen thickness on oxidation kinetics. The oxidation kinetics at other temperatures showed similar trends and are presented elsewhere.20 The mass change at 1000°C was independent of specimen thickness until ~400 h of exposure (Fig. 1). After longer exposure times, the specimen and total mass changes for the 0.3- and 0.5-mm specimens began to deviate, indicating scale spallation. The total mass gains for both specimens remained similar, but the extent of scale spallation was slightly more pronounced for the thicker specimen.

Fig. 1
figure 1

Typical example of mass change of alloy 230 (with specified specimen thicknesses) during oxidation in laboratory air at 1000°C. The open and closed symbols indicate the total and specimen mass changes, respectively

Full size image

For the 0.2-mm-thick specimen, there was only a minor difference between the specimen and total mass gains indicating that little or no scale spallation occurred up to 3000 h. However, after ~500 h, the total mass gain deviated from near-parabolic behavior to almost linear kinetics.

Figure 2 shows Cr profiles measured by EDX for specimens of different thickness after 1000 h and 3000 h exposure at 1000°C. It should be noted that the EDX point analyses measured the Cr concentration in the alloy matrix and do not take into account the Cr tied up in the precipitate phases which possesses a size that is clearly visible in the SEM image. After 1000 h, a subscale Cr depletion was apparent and the Cr mass fraction at the scale–alloy interface for all of the specimens shown in Fig. 2a was ~17%. After 3000 h (Fig. 2b), the Cr concentrations at the scale–alloy interface for the thicker specimens were quite similar, i.e. ~15%. However, for the 0.2-mm specimen, the depletion profile reached the center of the sample. With a severely depleted Cr reservoir, the result was a substantial decrease in the Cr interface concentration, as expected from theoretical considerations.25

Fig. 2
figure 2

Cr profiles measured by EDX (measured in γ-Ni matrix) for specimens of different thicknesses after (a) 1000 h and (b) 3000 h exposure at 1000°C (Dashed line the original Cr concentration in alloy 230)

Full size image

In order to obtain information about the alloy microstructure without it being affected by the subscale depletion process, a 0.5-mm specimen was isothermally exposed at 1000°C for only 100 h. Figure 3 shows that the alloy exhibited a four-phase microstructure consisting of a face-centred cubic γ-Ni matrix and three types of precipitates. EDX revealed one of the precipitates (black arrows) to be rich in W and Ni, likely M6C,6,26 whereas the other type of precipitate (white arrows) was a Cr-rich nitride with typical composition with respect to metallic constituents in atomic percent: 37Cr-34Ni-15W-4Mo-10Si. Only traces of a Cr-rich carbide (black dashed arrows), probably M23C6 27 were found, mainly at alloy grain boundaries. These precipitates were too small for quantitative analysis by EDX at this stage.

Fig. 3
figure 3

BSE image of bulk alloy 230 (0.5 mm thick sample) showing microstructure after 100 h isothermal exposure in synthetic air at 1000°C (Black, black dashed and white arrows W-rich carbides, Cr-rich carbides and Cr-rich nitrides (indicated as π) respectively)

Full size image

The typical microstructural features of alloy 230 specimens after prolonged oxidation is shown in Fig. 4, using the 0.3-mm specimen after 3000 h exposure at 1000°C as an example. Apart from the internal aluminum oxide, three types of precipitates were observed in the alloy (in the BSE image appearing white and gray, respectively). These precipitates were absent beneath the scale and were mainly present in the center of the specimen. Comparing Figs. 3 and 4, the volume fraction of precipitates appeared to increase with increasing exposure time.

Fig. 4
figure 4

BSE image of 0.3-mm-thick alloy 230 after reaction for 3000 h in laboratory air at 1000°C. Designations of phases are based on EDX and EBSD results in Figs. 5 and 6. Labeled precipitates in the lower part are Ni/W-rich M6C carbides, Cr-rich M23C6 carbides and Cr-rich nitrides (indicated as π)

Full size image

The precipitates formed after prolonged exposure could be unequivocally identified by EDX and EBSD (Fig. 5). Figure 5a shows a BSE image from the center of the 0.3-mm specimen oxidized at 1000°C for 3000 h, and Fig. 5b–g are element maps from this region. The carbide appearing bright in the BSE image is rich in Ni and W (similar to Figs. 3 and 4) with some Mo present, whereas the other carbide is Cr-rich.

Fig. 5
figure 5

(a) BSE image with (b–g) EDX element maps from the center of a 0.3-mm-thick alloy 230 specimen oxidized for 3000 h at 1000°C in laboratory air. The location used for the measurement corresponds to the left side of Fig. 4)

Full size image

The result of the EBSD phase map for the same specimen is shown in Fig. 6a, confirming the presence of M23C6 and M6C phases after long-term oxidation at 1000°C. EDX analyses indicated typical compositions of the metallic constituents in the two carbide precipitates (in wt.%) to be 63W-20Ni-12Cr-5Mo for M6C and 40Cr-32Ni-24W-4Mo for M23C6.

Fig. 6
figure 6

EBSD phase map for alloy 230 in two different places (a, b) in the center of the 0.3-mm specimen after 3000 h oxidation at 1000°C in laboratory air. In the EBSD phase map red corresponds to the γ-Ni matrix, yellow to M23C6, blue to M6C and green to π-Ni2W3N. The location used for the measurement in (a) corresponds to the left side of Fig. 4; the location used for the measurement in (b) corresponds to the right side of Fig. 4

Full size image

Additionally, nitride precipitates could be identified (Figs. 5c and 6b) which were isostructural with Ni2W3N (π-phase)28,29 with typical concentrations of the metallic constituents (in wt.%) 39Cr-32Ni-24W-4Mo. Unlike the nitride phase observed in the specimen exposed for only 100 h (Fig. 3), no Si was found in the π-phase precipitates. In some specimens and/or specimen areas, the Cr-rich carbides and nitrides coexisted (Fig. 5a) whereas in some places only carbides or nitrides were present.

The microstructural evolution of alloy 230 as a function of time and specimen thickness at 1000°C is shown in Fig. 7. Only one temperature was chosen for the illustration because the trends were found to be qualitatively similar for all three temperatures. In all cases, the scale was predominantly Cr2O3 with minor amounts of alumina internal oxidation, which preferentially nucleated along alloy grain boundaries. This morphology is very common for commercial Ni-Cr alloys with high (>20%) Cr contents and low Al contents.30 (The Al may be intentional or an impurity associated with commercial alloy manufacturing.) On top of the chromia layer, a very thin Mn-rich oxide, likely MnCr2O4, was found to be present. However, due to the very limited Mn reservoir in the specimens, no unequivocal indications for spinel formation were found after longer exposures, especially in the thinnest specimens. The scale thickness varied from 6 μm to 30 μm. The scale on the 0.2-mm specimens was substantially thicker than that on the 0.3- and 0.5-mm specimens, in agreement with the gravimetric data (Fig. 1). Additionally, minor void formation was found in the subsurface zone, probably related to the Kirkendall effect.1 Figures 4 and 7 clearly show that the width of the region, which was free of Cr-rich carbide and nitride precipitates (blue dashed lines), is slightly wider than that of the M6C depletion zone (red dashed line). For the 0.2-mm specimen, almost all the M6C precipitates were consumed after 3000 h at 1000°C (Fig. 7f).

Fig. 7
figure 7

BSE images of (a, d) 0.5 mm, (b, e) 0.3 mm and (c, f) 0.2-mm-thick alloy 230 after reaction for (a–c) 1000 h and (d–f) 3000 h in laboratory air at 1000°C (blue and red dashed lines indicate the width of the zones depleted in M6C and Cr-rich precipitates, respectively).

Full size image

Table II summarizes the measured volume fractions, expressed as a percentage, of M6C and the Cr-rich precipitates at 1000 h and 3000 h for the three specimen thicknesses. Particularly for the thin specimens, the quantitative estimation of the volume fractions of the various precipitates in the specimen center was difficult, due to the limited width of the precipitate-containing zone. Also, in the BSE images, it was difficult to distinguish between the two types of Cr-rich precipitates (M23C6 and Cr-nitrides) in the AnalySIS pro 5.0 software due to the similarity in contrast. Thus, these precipitates were summed in Table II. In spite of the quantification difficulties, the results show that the volume fraction of Cr-rich precipitates in the center of the specimen increased with exposure time and the volume fraction of M6C tended to decrease. This was also found at the other temperatures as will be discussed later in the present section. As illustrated in Figs. 5 and 6, Cr-rich M23C6 and π-phase co-existed in the core of some specimens. However, there was a clear trend that with increasing exposure time and decreasing specimen thickness the relative amount of nitride increased at the expense of M23C6, until in the thinnest specimen after 3000 h the Cr-rich precipitates consisted solely of nitrides (Fig. 7f).

Table II Volume fraction (%) of respective precipitates in the specimen center after oxidation at 1000°C (see text for detailed explanation)
Full size table

The width of the M6C-free subsurface zone in the 0.5- and 0.3-mm-thick specimens was smaller than that of the zone free of Cr-rich precipitates (Fig. 7a–e). In the 0.2-mm specimen, a similar trend was observed in the specimen oxidized for 1000 h (Fig. 7c). However, the area in which Cr-rich precipitates (here π-phase) were present after 3000 h was limited to a narrow zone of approximately 50 μm around the center of the specimen. Hardly any precipitates of the M6C were found (Fig. 7f).

Table III summarizes the widths of the carbide/nitride free subsurface zones after the 1000°C exposures shown in Fig. 7. The data illustrate that the widths of the zones which were free of Cr-rich precipitates increased with increasing exposure time and specimen thickness. An exception is the thickest specimen in which the Cr-rich precipitate-free zone after shorter exposure time was larger than after 3000 h. That, however, may be caused by the fact that the Cr-rich precipitates in the specimen oxidized for 1000 h were very small and that the measurements might be subject to a larger error.

Table III Depth of the subsurface zones (μm) which were free of M6C and Cr-rich precipitates (carbides and nitrides) in specimens oxidized at 1000°C
Full size table

To illustrate the effect of temperature, Fig. 8 shows the microstructures of 0.2-mm-thick alloy 230 specimens oxidized for 3000 h at 950°C and 1050°C. At both temperatures, a subscale region free of carbide and nitride precipitates developed as a result of Cr consumption by oxide scale formation, like at 1000°C. However, the widths of these precipitate-free zones were much less pronounced at the lowest oxidation temperature. The enrichment of the Cr-rich precipitates in the specimen center at 950°C (Fig. 8a) was, as expected, far less significant than that observed after exposure at 1000°C (Fig. 7f) and at 1050°C (Fig. 8b). (Note that the 1050°C specimen was stopped after 2000 h due to breakaway oxidation at the specimen corners.20)

Fig. 8
figure 8

BSE images for 0.2-mm-thick alloy 230 specimens after reaction for (a) 3000 h at 950°C h and (b) 2000 h at 1050°C in laboratory air (blue and red dashed lines indicate the width of the zones free of M6C and Cr-rich precipitates, respectively)

Full size image

Both the 0.2- and 0.5-mm-thick specimens exposed at 950°C contain M6C and Cr-rich precipitate phases (Fig. 9). The images in Fig. 9 indicate that the volume fraction of the Cr-rich precipitates near the specimen center increased with increasing oxidation time (compare Fig. 9a with b, and c with d) and decreasing specimen thickness (compare Fig. 9a with c, and b with d) as was observed for the specimens oxidized at 1000°C. It is interesting to note that, for the 0.2-mm specimens after 1000 h and 3000 h as well as for the 0.5-mm specimen after 3000 h, M6C as well as Cr-rich carbide and nitride were found. However, for the 0.5-mm specimen after 1000 h, no indications of the presence of Cr-rich carbides could be detected.

Fig. 9
figure 9

BSE images for (a, b) 0.2-mm and (c, d) 0.5-mm-thick alloy 230 specimens after reaction for (a, c) 1000 h and (b, d) 3000 h at 950°C in laboratory air

Full size image

At 1050°C, the dissolution of the M6C carbides in the subscale region was similar to the observations at 1000°C. A subsurface zone depleted in carbide and nitride was clearly apparent after the 1000 h exposure (Fig. 10). The Cr-rich precipitates in the center after 3000 h exposure mainly consisted of nitride. There was a tendency that some M6C precipitates appeared again after 3000 h (Fig. 10) in the subsurface zone. A similar observation was made for the 0.3-mm specimen. For the 0.2-mm specimen after 2000 h, only a very narrow Cr-nitride layer was present in the specimen center, showing similar features as after 3000 h at 1000°C (Figs. 7f and 8b).

Fig. 10
figure 10

BSE images of 0.5-mm-thick alloy 230 after reaction for (a) 1000 and (b) 3000 h at 1050°C in laboratory air (Blue and red dashed lines indicate the width of the zones free of M6C and Cr-rich precipitates, respectively)

Full size image

Discussion

Oxidation Kinetics

In interpreting the mass change data (Fig. 1), it has to be borne in mind that oxidation of chromia-forming alloys in high pO2 gases (such as air) will be affected by the formation of volatile Cr species, such as CrO3 and CrO2(OH)2 in the presence of H2O.31–33 Based on relevant data,34 the mass loss induced by this effect in static laboratory air in the temperature range considered would be in the range of a few tenths of a mg/cm−2 for a pure chromia scale up to the maximum exposure time and thus should not fundamentally alter the mass change kinetics in the present study. It also should be mentioned that the Mn in alloy 230 (Table I) is known to form a thin outer layer of (Cr, Mn)3O4 on top of the chromia scale.35 This spinel layer has been shown36 to reduce the formation of volatile Cr species. Thus, formation of volatile Cr species certainly occurs, but it is not expected to substantially affect the measured kinetics and related subscale depletion phenomena.

The results in Fig. 1 show, as expected, increasing oxidation rates with increasing temperature. Considering the O/Cr mass ratio and the well-known density of chromia (5.2 g cm−3),37 it is easy to derive that, for growth of a pure chromia scale, an area-specific O uptake of 1 mg cm−2 corresponds to a scale thickness of 6.07 μm. Ignoring the very minor amounts of spinel phase and the O uptake by internal oxidation, the scale thicknesses derived from weight gains may be compared with the actually prevailing scale thicknesses visible in the metallographic cross sections (Fig. 7). As an example: the 0.2-mm specimen exhibited a mass gain of 1.95 mg cm−2 after 1000 h at 1000°C,which would correspond to a scale thickness of 11.8 μm. This is in good agreement with the value seen in the metallographic cross-section (Fig. 7c). It is easy to verify that, in spite of the mentioned sources of inaccurcy, in all cases very reasonable agreement between calculated and measured values prevails.

For all studied temperatures, oxide adherence was found to increase with decreasing specimen thickness. This is in qualitative agreement with previous results and theoretical considerations related to the effect of specimen thickness on oxide scale adherence.38 Whereas no indication for oxide spallation for the 0.2-mm specimens was found, the 0.3- and 0.5-mm specimens exhibited the onset of scale spallation after approximately 300 h at 1050°C, 600 h at 1000°C (Fig. 1) and 2000 h at 950°C. At the times at which the thicker specimens started to exhibit scale spallation, the 0.2-mm specimen showed oxidation kinetics which were faster than parabolic. This enhanced oxidation might be related to relaxation of growth stresses by plastic deformation of the metallic substrate,39,40 rapid exhaustion of the reservoir of the reactive element La39,41 or, more likely, to tensile microcracking occurring upon re-heating after repeated specimen cooling, as described elsewhere.38,42

Modelling M6C Dissolution and M23C6 Formation

It is well established that long-term, high-temperature exposure of Ni base alloys generally results in precipitate coarsening.43 In the present case, however, the precipitate size and morphology were mainly affected by subscale changes in alloy composition, especially due to the Cr depletion for these relatively thin specimens oxidized at high temperature. The experimental results show that the Cr depletion leads to formation of precipitate-free subscale zones, and that the widths of these zones for the various precipitate types (carbides and nitrides) differ substantially. Also, the volume fraction of Cr-rich precipitates (M23C6 and Cr-nitrides) in the specimen center increased with time. In most samples, the amount of M23C6 was larger after 1000 h than after 3000 h, where mainly nitrides were found. These effects increased with temperature and decreasing specimen thickness. At the extreme of the 0.2-mm-thick specimen after 3000 h at 1000°C and 2000 h at 1050°C, the M6C carbide tended to vanish completely.

Thermodynamic calculations revealed that, for the C content (i.e. 0.09%) prevailing in the as-received alloy 230, only γ-Ni and M6C exist in addition to a very minor amount of Cr-rich nitride. Based on the experimental findings (Fig. 6), it was assumed in the calculations that the π-phase was the only Cr-rich nitride which may be formed. If the C content was increased to ~0.16%, the calculation then predicted M23C6 formation with a corresponding decrease in M6C fraction as the C content further increased (Fig. 11). As the major precipitate phase in the as-received material is M6C carbide, the first part of the discussion focuses on the carbides and initially ignores the nitrides.

Fig. 11
figure 11

Molar fractions of different phases in alloy 230 as a function of C content at 1000°C (calculated by Thermo-Calc using database TTNI7). For details: see text

Full size image

To assist in understanding the mechanism of the observed microstructural changes in alloy 230 during oxidation, a section of the Ni-Cr-W-Mo-C phase diagram at 1000°C was calculated with fixed 14 wt.% W and 2 wt.% Mo. The trends discussed for 1000°C are qualitatively similar for the range of temperatures studied. The phase diagram together with C isoactivity lines are shown in Fig. 12. (An isoactivity line is a locus of compositions in the phase diagram for which the chemical potential (i.e. activity) of the selected element (C) is constant.) Point A in Fig. 12 indicates that the bulk composition of alloy 230 is located in a two-phase field in which the Ni-matrix is in equilibrium with the W-rich M6C carbide, as shown in Fig. 11.

Fig. 12
figure 12

Section of the Ni–Cr-W-Mo-C phase diagram at 1000°C for constant W and Mo contents of 14% and 2%, respectively, calculated in Thermo-Calc using the TTNI7 database. The dashed line indicates isoactivity for which the chemical potential (activity) of the C is constant

Full size image

During high-temperature oxidation, the Cr concentration at the scale-alloy interface decreases due to the continuous Cr depletion related to the scaling process (Fig. 2). Moving from point A to point P in Fig. 12, as the Cr content decreases, results in an increase in the solubility of the M6C carbide. This may at first sight be surprising as the carbide mainly contains Ni, W and some Mo, but hardly any Cr. The reason for this dependence of M6C solubility on Cr content is related to the thermodynamic interactions of the alloying elements in the γ-Ni matrix, specifically, the effect of Cr content on the activity of C and W. The thermodynamic calculations in Fig. 13 show that the W activity decreases with decreasing Cr content, thus destabilizing the M6C carbide. The effect of Cr on the thermal stability of M6C in a Ni-base alloy has been extensively discussed.44 Figure 13 also shows that Cr depletion leads to an increased C activity. The Cr-depleted near-surface region of the alloy is thus an area of high C activity compared to that in the middle of the specimen. The resulting C activity gradient creates a driving force for C to diffuse from the Cr-depleted surface region to the specimen center in order to maintain equilibrium. In Fig. 12, this means that the diffusion path for this transformation will be close to the isoactivity line from point A to point B.

Fig. 13
figure 13

Activities of different elements in alloy 230 as a function of Cr content at 1000°C, calculated by Thermo-Calc. Areas indicated I–IV relate to phase fields γ + M23C6, γ + M23C6 + M6C, γ + M6C +M(N,C) and γ + M6C + π, respectively

Full size image

Back diffusion of C as a result of oxidation-induced Cr-depletion has been described in previous studies, e.g., for the Ni-base alloy 617.7,16,45–47 The back diffusion has generally been explained to be the result of the Cr-depletion-induced dissolution of the Cr-rich carbide. The dissolution of the carbide obviously results in an increase of the C activity in the matrix,23 thus providing a composition gradient driving force for C diffusion towards the specimen center. The mechanism as to why the higher C activity in the subsurface regions results in C back diffusion rather than in decarburization in this case is discussed elsewhere.48

The back diffusion of C occurring during air oxidation of alloy 230 is not related to a composition gradient, as frequently described for cases in which Cr depletion results in dissolution of Cr-rich carbides.7 Here, the back diffusion occurs due to an activity gradient driving diffusion. The increased C activity is mainly due to the effect of Cr loss on M6C stability and the thermodynamic interaction among Cr, C and W in the γ-Ni-matrix.

Obtaining quantitative experimental confirmation of the C enrichment in the specimen center is not possible from EDX or WDX measurements, e.g., due to the multiphase character of the microstructure. A suitable method for measuring the total C content is GDOES. However, depth profiling by sputtering through the surface scale and the relatively wide subscale depletion zone result in a very deep sputter crater. It is well established that the measurement accuracy may substantially decrease with increasing crater depth, because of crater side effects and non-flatness of the sputter crater. Therefore, a different approach was used to measure the C and N contents in the specimen center. Specimens of 0.5 mm thickness after exposure at 1000°C for 1000 h and 3000 h (compare microstructures in Fig. 7a and b) were carefully ground until reaching the center of the specimen. Due to the limited accuracy of the grinding method and the very limited width of the precipitate-containing zones (Fig. 7), this procedure was unfortunately not possible for the 0.3- and 0.2-mm specimens.

The ground surfaces, corresponding to the center of the 0.5-mm specimens, were analyzed by GDOES. For obtaining the relative sensitivity factors22 for the various elements, a non-exposed alloy 230 specimen was analyzed using the same sputtering parameters as used for the two oxidized specimens, taking into account the known bulk alloy composition (Table I). The quantified depth profiles as a function of sputtering time for the specimen exposed for 3000 h in Fig. 14 shows that the average C content in the specimen center was 1.03 at.% (compared to 0.48 at.% present in the as-received material). For the 1000 h (results not shown here), the C content in the specimen center was 0.74 at.%. Figure 14 also indicates that, for the 3000 h specimen at a depth corresponding to a sputtering time of approximately 1300 s, the C content decreased, eventually reaching values which were even lower than the C content in the as-received material. The sputter time at which this effect occurs thus corresponds to the subscale carbide free zone shown in Fig. 7d. These results clearly illustrate that indeed a C back diffusion occurred as a result of the mechanisms described above. After the 3000 h exposure at 1000°C, this resulted in an approximate doubling of the C content in the center of the 0.5-mm specimen. As the C diffusivity is orders of magnitude larger than that of the carbide forming elements (here mainly Cr and W) and most of the C in the specimen center is tied up in form of carbides, the C concentrations are mainly governed by profiles of the mentioned metallic elements, i.e. they are relatively flat.

Fig. 14
figure 14

Element concentrations as function of sputter time for 0.5-mm alloy 230 sample after 3000 h exposure at 1000°C. Before the GDOES measurement, the specimens was ground to half thickness. Thus, the left part of the diagram (sputter time equals 0 s) corresponds to the center of the specimen

Full size image

Specimen or component thickness is an important consideration for the discussion of the oxidation-induced microstructural changes. In thin specimens, as studied here, the Cr reservoir was limited. In addition, the C diffusing from the surface will enrich the center. For a component several millimeters thick, the C released from the subsurface depletion zone will only make a minor change in the C content of the wide zone at the center of the component. For these thin specimens, the width of the subsurface depletion zone becomes similar or larger than the width of remaining carbide-containing zone after long times or at the highest temperatures, e.g. Fig. 7f. Even for the 0.5-mm specimen, the C content in the specimen center after a 3000 h exposure approximately doubled compared to that in the as-received material, as shown in Fig. 14. In Fig. 12, this increase in C content would correspond to movement on the isoactivity line to point Q. Once the C concentration exceeds 0.14 wt.%, Cr-rich M23C6 becomes stable with a resulting decrease in the Cr concentration in the γ-Ni matrix. For simple geometrical reasons, it is apparent that the relative enrichment of C in the specimen center will tend to be even larger for the 0.3- and 0.2-mm specimens.

For the case in which the W-rich M6C carbides completely dissolve in the 0.2-mm-thick specimen after 3000 h at 1000°C (Fig. 7f), the Cr profile showed significant Cr depletion across the entire specimen, such that the average Cr content in the specimen had dropped from 22 down to ~12 wt.% (Fig. 2b). In Fig. 12, that depletion corresponds to the red arrow path from point A to point D and a transition from the three-phase field γ-Ni + M6C + M23C6 region towards the two-phase field γ-Ni + M23C6 region with no M6C present. One assumption in Fig. 12 is that the concentration of W and Mo remain at the original levels of 14% and 2%, respectively. In reality, however, the Cr depletion may affect the activities of these elements (Fig. 15), thus establishing W and Mo activity gradients between the subscale depletion zones and the specimen center.

Fig. 15
figure 15

Comparison of Cr, W and Mo concentration profiles in the γ-Ni matrix measured by EDX (symbols) in the studied 0.3-mm-thick specimen of alloy 230 with DICTRA calculations (lines) after 1000 h oxidation at 1000°C in laboratory air. Measured oxidation kinetics used as input for the calculation were taken from the gravimetric data in Fig. 1

Full size image

For developing a more detailed description of the mechanism of the M6C/M23C6 transformation discussed above, a set of calculations was performed in DICTRA to simulate the experimentally observed oxidation-induced phase transformations. Similar to the thermodynamic calculations, five elements were taken into account, i.e. Ni, Cr, W, Mo, and C. A time-dependent outward Cr flux, derived from the experimentally determined oxidation data, was set as a boundary condition to simulate the oxidation-induced Cr depletion, using the procedure described elsewhere.23 Figure 15 shows the calculated and measured concentration profiles in a 0.3-mm-thick specimen after oxidation for 1000 h at 1000°C. Good agreement between the measured and calculated Cr profiles was obtained, which implies that the Cr depletion kinetics were accurate. However, the calculated W profile did not completely match with the experimental one in the near-surface region, although the W subsurface enrichment was qualitatively correct. The reason for the inaccuracy is the inability of DICTRA to take into account the interface displacement due to metal consumption by the scaling process. While this effect is virtually negligible for thick specimens, the moving boundary effect distorts the depletion calculations in a 0.3-mm-thick specimen. For a better quantitative description of the depletion kinetics, interface displacement has to be implemented into the model. However, commercial software packages like DICTRA do not so far provide this option. Modeling approaches developed to account for this effect have been described elsewhere,16,23 but fall outside the scope of the present investigation.

According to the thermodynamic calculations (see Fig. 11), the M23C6 carbide is not stable in the as-received condition in alloy 230. Based on the above considerations, it forms in the specimen center after back diffusion has resulted in an overall C content higher than the critical value of 0.16 wt.%. The C migration from the Cr-depleted zone towards the specimen core is a dynamic process and it requires a certain time until the C enrichment in the center of the foil exceeds the threshold for M23C6 formation, which is 0.14 wt.% at 1000°C, according to the calculated phase diagram (Fig. 12). For a given oxidation condition (i.e. the depth of the Cr-depleted zone), the critical C concentration required for carbide formation will be reached after shorter oxidation times in a thin than in a thick specimen. This is in agreement with the 950°C results presented in Fig. 9.

The DICTRA calculations in Fig. 16 thus at least qualitatively correctly describe (1) the formation of a M6C depleted zone (2) the initially unstable Cr-rich M23C6 carbide forms due to the subscale Cr depletion and the C back diffusion, and (3) the accompanied decrease in the M6C volume fraction.

Fig. 16
figure 16

DICTRA calculation profiles showing volume fractions of M6C and M23C6 in the 0.3-mm-thick specimen of alloy 230 after (a) 1000 h and (b) 3000 h at 1000°C in laboratory air. Input kinetics from Fig. 1, starting alloy composition from Table I, and calculations in Fig. 11

Full size image

The total calculated decrease in volume fraction of M6C was smaller than experimentally observed. This discrepancy is related to the formation of nitride along with oxidation Cr depletion in the specimen center, as will be illustrated in the following section.

Modelling Nitride Formation in the Center of the Specimen

The high volume fractions of nitride in the (thin) specimen after long exposure time are indicative of nitridation. Apart from back diffusion of N similar to that of C, there was also substantial N ingress from the environment. In the current type of experiments involving oxide growth, oxide spallation and N uptake, it was virtually impossible to accurately estimate the nitridation kinetics and the incubation time (nitridation occurs after the occurrence of a certain Cr depletion). With so many experimental unknowns, diffusion calculations would be extremely inaccurate and inconclusive.

Therefore, another attempt was made to rationalize the experimentally observed microstructural changes, using phase equilibria calculated by ThermoCalc and the data base TTNI7. Based on the activity data in Fig. 13, the subscale Cr depletion would result in a back diffusion of N in a similar manner as described for C. This was confirmed by the GDOES results shown in Fig. 14. Although the measured concentrations for N were nearer to the detection limits (as indicated by the larger signal to noise ratio) than is the case for C, the results clearly revealed that an increase of the N content in the specimen center occured whereby the N content increased with increasing exposure time. The database TTNI7 considers various types of nitride phases, i.e. mainly Cr2N. The nitride phase present after a very short (100 h) exposure time (Fig. 3) could not, be unequivocally identified, as described in previous sections. The nitride in the specimen center, however, could clearly be correlated with π-phase (see Fig. 6b). The composition of this phase at 1000°C for the alloy 230 composition in the data base TTNI7 is 57Cr-36Ni-2Mo-5N (in wt.%). Compared to the actually measured nitride composition, e.g., in Fig. 4 (39Cr-32Ni-24W-4Mo in wt.%), the database TTNI7 did not take into account W solubility; however, Mo solubility was correctly described.

Considering that the π-phase is the only nitride formed, alloy 230 should at 1000°C exhibit a microstructure consisting of a γ-Ni matrix, W-rich M6C and very minor amounts of π-phase (Fig. 11). A doubling of the C content and N content (see Fig. 14) by back diffusion resulted in the formation of a four-phase microstructure consisting of a γ-Ni matrix, W-rich M6C, Cr-rich M23C6 and π-phase (Fig. 17a).

Fig. 17
figure 17

Phase fractions in alloy 230 as function of temperature for various C and N contents. For details, see text

Full size image

Considering the densities of the mentioned phases (9 g cm−3 for γ-Ni,49 14.2 g cm−3 for M6C50 and 6 g cm−3 for π-phase51) reveals that the volume fraction of M6C in the alloy not affected by oxidation-induced depletion processes is approximately 2.7 vol.% whereas the volume fraction of π-phase is extremely small. If one assumes that the back diffusion of N would in the 0.3- and 0.2-mm specimens result in a N content of 0.18 wt.%, i.e. even higher than found for the 0.5-mm specimen after 3000 h (Fig. 14), the volume fraction of the π-phase would be 3.9%. This is substantially lower than that experimentally observed for the thinnest specimens after 3000 h at 1000°C (Table III). This leads to the conclusion that, upon increasing exposure time and decreasing specimen thickness, the high volume fractions of nitride in the specimen center was not only related to back diffusion of N but additionally to uptake of N from the atmosphere (Fig. 17b). This is probably related to the fact that the substantial decrease in Cr interface concentration substantially lower than ~15% which occurred during long-term exposure (Fig. 2) resulted in a loss of protective properties of the surface oxide scale, which then may become prone to transport of gas molecules such as N.

Assuming a doubling of C content (due to back diffusion) and an even greater than doubling of the N content in the specimen core, the resulting phase fractions are shown in Fig. 14. Here, the C content was set at twice the original value and the N content arbitrarily to a high value of 0.4%, i.e. substantially higher than twice the original value used in Fig. 14. This resulted in a substantial increase in the phase fraction of π-phase and, at 1000°C, a slight decrease in phase fraction of M6C and a vanishing of the M23C6, which is in agreement with the experimental observations (Fig. 7e and f). The dissolution of the Cr-rich carbide will obviously result in an increase of the C activity in the center of the specimen. This is expected to result in a redistribution of C in the specimen. This may be the qualitative explanation for the observations illustrated in Fig. 10 where after 1000 h a subscale M6C-free zone was clearly found, whereas after 3000 h this was not the case.

The thermodynamic calculations in Fig. 17b do not predict that the formation of large volume fractions of Cr-rich nitrides after long periods in the thinnest specimens is accompanied by the experimentally observed vanishing of M6C (Figs. 7f and 8b). The effect occurs when the overall Cr content in the specimen core starts to decrease (Fig. 2). The thermodynamic calculations in Fig. 12 predict a decreasing volume fraction of M6C upon substantially decreased Cr content; however, a complete vanishing of the M6C was quantitatively not correctly described. The discrepancy might be related to the fact that the π-phase in the database TTNI7 does not consider W solubility. This is contrary to the experimentally observed composition in which, typically, 24 wt.% W was present in the π-phase. Due to this high W solubility, formation of the π-phase is expected to substantially decrease the W content in the γ-matrix which obviously should have a destabilizing effect on the Ni/W-rich M6C phase. The vanishing of the M6C phase (Figs. 7f and 8b) in the specimen core is thus likely caused by a combination of the substantial oxidation-induced Cr loss and the consumption of W by formation of the W-containing nitride.

Summary and Conclusion

Thin (0.2–0.5 mm) specimens of alloy 230 oxidized at 950–1050°C in laboratory air for 1000–3000 h exhibited a microstructure which consisted of a γ-Ni matrix, Ni/W-rich M6C and Cr-rich precipitates (M23C6 and Cr-nitrides). The Cr loss due to oxidation resulted in a Cr-depleted subsurface zone in which the Cr-rich precipitates dissolved. In addition, the Cr-lean M6C was also depleted in this zone. The width of the M6C-free zone was always smaller than that of the zone which was free of Cr-rich precipitates (carbides and/or nitrides). In the center of the specimens, Cr-rich M23C6 and π-nitrides were formed. These effects increased with oxidation time and temperature and decreasing specimen thickness. Using thermodynamic and kinetic modeling, a mechanism was developed to explain these observations wherein the Cr depletion in the subsurface zone increased the local C and N activity, resulting in an activity gradient that drove back C and N to the center of the specimen. Because the specimens were thin, the relative change in C and N content is sufficient to stabilize the Cr-rich carbide (M23C6) and nitride (π-phase) near the specimen center. DICTRA calculations using the measured oxidation kinetics as input data successfully predicted the measured Cr depletion profile and qualitatively predicted the other microstructural changes. With increasing exposure time and decreasing specimen thickness, precipitate formation in the specimen center gradually became affected by N uptake from the environment. This process resulted in enhanced nitride formation at the expense of the Cr-rich carbides, in qualitative agreement with the thermodynamic calculations. Ongoing Cr depletion additionally resulted in dissolution of the M6C precipitates.