Engineered Coatings for Ni Alloys in High Temperature Reactors ELIZABETH A. CLARK, JAMES Y. YANG, DEEPAK KUMAR, GARY S. WAS, and CARLOS G. LEVI Alloy 617 is a candidate material for the intermediate heat exchanger of the He-cooled very high temperature reactor. At target temperatures ‡1223 K (950 C), low level impurities in the gas stream may cause carburization, decarburization, and/or oxidation of 617 with deleterious effects on its mechanical properties. The chromia scale formed naturally by 617 does not provide adequate protection in the expected environment. Alpha alumina offers a greater potential as an effective diffusion barrier with superior stability, but it requires modification of the alloy surface. This work explores two approaches to surface modification based on aluminizing, either alone or in combination with FeCrAlY cladding, followed by pre-oxidation to form alpha alumina. Both approaches yield coatings with promising diffusional stability on alloy 617. Initial corrosion studies in impure He environments reveal that the alpha alumina is stable and protects the underlying substrate in both carburizing and decarburizing environments. DOI: 10.1007/s11661-012-1411-2 The Minerals, Metals & Materials Society and ASM International 2012
I.
INTRODUCTION
A leading concept in GenIV nuclear plants is the Hecooled very high temperature reactor (VHTR) with ultimate target temperatures in the primary gas loop ‡1223 K (950 C) and a plant design service life requirement of 60 years.[1] In addition to the reactor, a critical component in the primary He loop is the intermediate heat exchanger (IHX) that transfers the energy to a secondary working fluid for use in power generation, chemical processes, etc. While the VHTR reactor concept is generally acknowledged as technically viable,[1] implementation and full exploitation of its potential are critically limited by materials that can withstand the extreme combination of temperature, thermal/mechanical stresses, and a He environment containing low levels of deleterious impurities (O2, CO/CO2, H2/H2O, CH4). Candidate materials of current interest are based on Ni, typified by alloy 617 (UNS N06617).[1–3] Previous work has shown that Ni alloys in the impure He environment at temperatures of order 1173 K to 1273 K (900 C to 1000 C) can be susceptible to oxidation, carburization, and/or decarburization,[4–6] with deleterious effects on their toughness[7] and creep strength.[8] Moreover, the interplay between the environment and thermo-mechanical cycles arising from fluctuations in power or flows of the heat exchange fluids can lead to ELIZABETH A. CLARK, Graduate Student, JAMES Y. YANG, Research Specialist, and CARLOS G. LEVI, Professor, are with the Materials Department, University of California, Santa Barbara, CA 93106-5050. Contact e-mail:
[email protected] DEEPAK KUMAR, Graduate Student, is with the Department of Materials Science and Engineering, University of Michigan, Ann Arbor, MI 48109-2104. GARY S. WAS, Professor, is with the Department of Nuclear Engineering and Radiological Sciences, University of Michigan. Manuscript submitted October 25, 2011. Article published online September 26, 2012 METALLURGICAL AND MATERIALS TRANSACTIONS A
cracking by the creep-fatigue phenomena.[9] In the absence of suitable environmental protection, the leading candidate alloys are unlikely to maintain their structural integrity at temperatures above ~1123 K (850 C). The success of the VHTR concept depends critically on the development of strategies that improve the stability of alloys in impure He for temperatures up to 1273 K (1000 C) and for long exposure times. Effective approaches to environmental protection of alloys at high temperatures are generally based on thin surface layers of dense, stable, and thermo-chemically compatible oxides. Ni alloys of interest are natural chromia formers, but this oxide is susceptible to reduction and/or conversion to carbide depending on the relative O and C potentials.[4] Alumina is considerably more stable than chromia against hydro- and carbo-thermic reduction at all temperatures of interest, e.g.[10] From a kinetic perspective, alpha alumina (a-Al2O3) is well established as the most efficient diffusion barrier to protect Ni- and Fe-based alloys against oxidation and corrosion in extreme environments such as that of gas turbines.[11] a-Al2O3 has also demonstrated potential as an effective carburization barrier in Ni alloys,[12] and as a highly effective barrier to tritium permeation in steels.[10,13] Tritium barriers have been largely investigated in the context of fusion applications, but leakage into the secondary loop of the IHX is also a concern for VHTR plants.[14] Moreover, the thermal conductivity of a-Al2O3 at 1273 K (1000 C) is ~25 pct that of 617, but because its thickness is so much smaller, its presence should not impair the capabilities of the system in high thermal flux applications. Alpha alumina thus emerges as the leading candidate barrier layer for protection of Ni alloys in the VHTR environment with potentially added benefits in tritium containment. There are two fundamental barriers to the implementation of this concept: (1) The alloys of interest are not VOLUME 44A, FEBRUARY 2013—835
natural a-Al2O3 formers, requiring chemical modification of their surfaces to enable the in situ evolution of the protective scale and its stability over the desired life of the system; and (2) a-Al2O3 does not form readily by oxidation at temperatures below 1273 K (1000 C), often being superseded by transient (but persistent) metastable forms of alumina and/or other oxides which are generally less protective.[15–18] In essence, the prospective VHTR operating temperature falls in the gap between the useful range of chromia scales and that typical of a-Al2O3. The research described in this paper explores two surface modification concepts to enable the formation of a-Al2O3 on alloy 617. Both concepts involve aluminizing, in one case as the final treatment and in the other as a foundation for subsequent cladding with a thin layer of FeCrAlY, a ferritic alloy known to form alpha alumina with relative ease.[19] The subsequent sections describe the experimental implementation of these concepts, their relative ability to form a-Al2O3, their diffusional stability, and their behavior under carburizing and decarburizing environments.
II.
EXPERIMENTAL APPROACH
A. Materials Alloy 617 was supplied as 11-mm-thick plate (Special Metals, Corp., Huntington, WV). The FeCrAlY for cladding was procured in the form of 50 and 100 lm foils (Goodfellow, Oakland, PA). The compositions of both materials are given in Table I. All substrates and cladding materials were polished to 800 grit finish and cleaned ultrasonically before processing. Materials for aluminizing included Al powder (<44 lm, 99.5 pct, Alfa Aesar, Ward Hill, MA) as the source, Al2O3 powder (500 grit, Norton Co, Worcester, MA) as the filler, as well as CrCl3Æ6H2O and NH4Cl (both> 99.5 pct purity), AlCl3 and AlCl3Æ6H2O (both reagent grade) as activators, all from Alfa Aesar. B. Aluminizing A high activity, low temperature aluminizing process was selected for this study. However, the prospect of cladding the aluminized specimens required some modifications of the typical approach[20] to facilitate surface preparation for diffusion bonding. First, the aluminizing temperature selected for this study was 973 K (700 C), which is lower than normal practice ~1073 K ± 50 K (~800 C ± 50 C). This was done primarily to facilitate comparison with a parallel study on barrier layer development on ferritic/martensitic steels (to be reported separately). A second change involved wrapping the alloy coupons in Nextel 610 Alumina Fiber Fabrics (3M, MI) to preclude direct contact with the granular aluminizing bed that was found to lead to rougher surfaces, requiring more extensive cleaning and partial removal of the aluminized layer prior to cladding. (Over-the-pack aluminizing would provide an improved surface finish, but the rate was considerably 836—VOLUME 44A, FEBRUARY 2013
Table I.
Chemical Compositions of Alloy 617 and Fecralloy Fecralloy
Alloy 617* Element
Wt. pct
At. Pct
Ni Fe Cr Co Mo Al Ti Si Mn Cu C Y Zr
bal. 0.68 21.6 11.5 9.4 0.9 0.3 0.07 0.03 0.02 0.09
bal. 0.71 24.22 11.38 5.71 1.94 0.37 0.15 0.03 0.02 0.44
Wt. pct
At. Pct
bal. 22
bal. 22.03
5
9.65
0.3 0.2
0.56 0.19
0.02 0.1 0.1
0.09 0.06 0.06
*Analysis by Sherry Labs, Daleville, IN. Impurities in alloy 617 include <0.0005 pct O, 0.007 pct N, 0.007 pct P, <0.005 pct S.
slower in the experimental setup utilized.) The third modification pertains to the selection of the activator. A thin scale of relatively poor adherence was found to evolve on the as-aluminized surface upon annealing in low pO2. This scale must be removed and the surface polished prior to cladding. The ease of that operation was found to depend on the activator, presumably because the adherence depends on the initial surface microroughness and Al content. Four different activators reported in the literature (NH4Cl, AlCl3, AlCl3Æ6H2O, and CrCl3Æ6H2O) were investigated with regard to yield, stability, and surface quality after aluminizing. Of these, AlCl3 and AlCl3Æ6H2O were found to give loose scales after heat treatment, but the aluminizing rates were too slow at the temperatures selected, especially for the latter. NH4Cl gave the best yield, but the scale was significantly more difficult to remove, while CrCl3Æ6H2O offered a slightly lower yield, but easier surface cleaning after aluminizing relative to NH4Cl. Notably, both activators gave the same aluminized thickness and comparable microstructure for a given weight gain, albeit at somewhat different rates, and both were used in this investigation. The selected processing route involved aluminizing alloy coupons, 11 9 5 9 1 mm, by pack cementation using a powder mixture of 4Al/4(CrCl3Æ6H2O or NH4Cl)/92Al2O3 (in weight percent). The coupons were wrapped in alumina fabric and embedded in the pack, which was subsequently sealed in an alumina crucible and placed in a tube furnace under flowing high purity Ar (XO £ 10 5). The pack was heated to 973 K (700 C) and held for 2–8 hours depending on the activator and desired thickness. After aluminizing, the specimens were annealed at 1273 K (1000 C) in flowing, gettered Ar (with a measured oxygen content of XO £ 10 13 produced by passing the high purity Ar through a Centorr 2A Inert Gas Purifier) for times ranging from 1 to 16 hours to fully develop the structure of the surface layers. Samples for He exposure or FeCrAlY cladding were aluminized to form an approximately 20-lm-thick initial layer, followed by heat treating for 16 hours at METALLURGICAL AND MATERIALS TRANSACTIONS A
1273 K (1000 C). Any scale formed during this process was removed and the surface was polished down to an 800 grit finish to leave a clean, flat metal surface. C. Cladding with FeCrAlY A subset of specimens was clad with FeCrAlY foil by diffusion bonding in a high vacuum hot press for 1 hour at 1273 K (1000 C) under a pressure of 20 MPa.* Clad *Cladding by diffusion bonding is used here only for proof of concept. It is acknowledged that surface layers of FeCrAlY could be applied by other means if the concept was to be implemented into practice.
specimens consisted of 617 coupons aluminized as described earlier, sandwiched between two 100-lm-thick sheets of FeCrAlY foil. The aluminized + clad specimens were heat treated at 1273 K (1000 C) for up to 168 hours in Ar to assess their diffusional stability. A separate set of specimens intended for assessing the diffusional stability of the system without the aluminized interlayer consisted of bonding one 50 lm sheet of FeCrAlY foil between two 1-mm-thick 617 coupons. In all cases, sheets of Fiberfrax ceramic paper were placed between the carbon platens of the hot press and the sample to minimize the possibility of reaction or bonding between them. Prior to use, the ceramic paper was held for 2 hours at 873 K (600 C) to remove any volatile additives. The clad specimens were also polished down to an 800 grit finish in preparation for pre-oxidation. D. Corrosion Studies in Impure He Environments Specimens for evaluation in impure He environments were first pre-oxidized at 1273 K (1000 C) for 2 hours in gettered Ar (XO £ 10 13) to form a thin, continuous, dense a-Al2O3 scale. The alumina phase(s) present was identified using photostimulated luminescence spectroscopy (PSLS).[18] A Horiba Jobin Yvon LabRAM Aramis Raman equipped with a 633 nm He-Ne laser was used for the PSLS analyses. Exposures were performed at 1273 K (1000 C) for 100 and 500 hours in flowing He environments with carefully controlled CO:CO2 ratios of ~9 (13.5CO:1.5CO2 in ppm) and ~1272 (1908CO:1.5CO2 in ppm), corresponding to decarburizing and carburizing conditions relative to bare 617, respectively, using specially designed equipment described elsewhere.[21] E. Microstructural Analysis Microstructural characterization of the different specimens was performed using a variety of instruments. Samples were first sectioned using a slow speed diamond saw, mounted in phenolic resin, and polished to 0.25 lm diamond finish. Scanning electron microscopy (SEM) was performed using an FEI XL40 Sirion FEG Digital SEM equipped with an Oxford energy dispersive x-ray spectrometer (EDS). Concentration profiles across the different layers were measured using a Cameca SX-50 electron microprobe (EMPA). Specimens for examinaMETALLURGICAL AND MATERIALS TRANSACTIONS A
tion by transmission electron microscopy (TEM) were extracted from polished cross sections using either an FEI DB235 or a Helios 600 dual beam focus ion beam (FIB) systems. TEM was performed on an FEI Tecnai G2 Sphera TEM, also equipped with an Oxford EDS detector. Phase identification was performed using selected area diffraction (SAD) in the TEM as well as by x-ray diffraction (XRD) using a PANalytical X’PERT powder diffractometer.
III.
RESULTS
The primary focus of this paper is to assess whether the proposed environmental barrier concepts offer significant advantages in durability and stability relative to the baseline alloy 617. Results are thus presented for their oxidation and interdiffusion behaviors, as well as their performance in impure flowing He. Since the focus here is on ‘‘proof of concept,’’ this paper would be followed by others describing in-depth studies in the relevant areas as well as on the effect of static and cyclic thermo-mechanical loads in impure He. A. Oxidation Behavior The environmental barrier concepts proposed rely on the formation of a thin, dense layer of alpha alumina by pre-oxidation, and the ability of the substrate to reform the protective layer in the event of cracking or spallation. Hence, an important first step is to assess the oxidation behavior of the aluminized and clad surfaces in low pO2 environments. (The self-healing ability will be addressed as part of a broader oxidation study in a forthcoming publication). There are also broader fundamental reasons to understand oxidation of these alloys in low pO2, to be addressed in the discussion. After one hour of oxidation in a gettered Ar (XO £ 10 13) environment at 1273 K (1000 C), PSLS revealed the clad + aluminized sample had a continuous scale consisting of alpha alumina alone, while the aluminized-only sample had a mixed scale of alpha, theta, and possibly gamma alumina (Figure 1). After 2 hours of oxidation, however, the scale was identified to be completely alpha, so this longer pretreatment was selected for samples to be tested in impure He environments. The morphologies of the thermally grown alumina corresponding to the PSLS patterns in Figure 1 are depicted in Figure 2 and compared with those produced by oxidation of FeCrAlY in air at 1273 K (1000 C) for 1 hour. The latter is also shown to be all alpha alumina by PSLS, but the morphology is remarkably different from the others, comprising of clusters of lamellar crystals underlaid by a continuous layer of oxide. The scale formed on the aluminized surfaces in air is not included here, but is well known to exhibit anisotropic growth, albeit as whiskers rather than plates.[18] B. Interdiffusion To assess the feasibility of direct cladding FeCrAlY on alloy 617, a thin foil of the former was diffusion VOLUME 44A, FEBRUARY 2013—837
Fig. 1—Comparison of the luminescence spectra from oxide scales formed upon exposure to flowing Ar with Xo ~ 10–13 at 1273 K (1000 C) for 1 h. (a1) to (a6) correspond to various locations on the surface of an aluminized 617 sample, while (c) is from a typical location on the surface of FeCrAlY diffusion bonded to aluminized 617. All locations on the clad surface show the same spectra in contrast with the variability of oxide phases on the aluminized sample.
bonded between two coupons of the latter as described earlier. The initial 617/FeCrAlY/617 bond is remarkably sound as illustrated in Figure 3(a). However, the 50 lm FeCrAlY foil dissolved in the Ni alloy after 168 hours (1 week) at 1273 K (1000 C), Figure 3(b). Concomitantly, intergranular second phases identified by SAD in the TEM as M23C6 carbides were also found after this treatment. EDS showed concentrations of Cr and Mo in these regions, Figures 3(c) and (d). The precipitates formed largely along the grain boundaries as well as within the region of the original FeCrAlY layer. Clearly, a thin clad layer alone would not be a viable long-term solution to maintain the formation of alpha alumina on the surface. Because FeCrAlY is still attractive for its ability to form alpha alumina readily, its implementation would only be feasible through the incorporation of a suitable diffusion barrier between cladding and substrate to achieve the requisite diffusional stability of the system at 1273 K (1000 C). Similar studies on aluminized 617 reveal a much slower interdiffusion between the surface layer and the bulk substrate than that of FeCrAlY at the same temperature as shown in Figure 4. An initial aluminized layer of slightly less than 10 lm evolves only to approximately 28 lm after 16 hours (Figure 5), with changes in its morphology, but no significant effect on the underlying substrate chemistry (Figure 6). The microstructural evolution of the coatings produced by 838—VOLUME 44A, FEBRUARY 2013
Fig. 2—Surface morphologies of samples oxidized for 1 h at 1273 K (1000 C). (a) FeCrAlY in air, (b) FeCrAlY in gettered Ar (Xo £ 10–13), and (c) Aluminized 617 in gettered Ar. All samples were polished to 800 grit finish before oxidation. The magnification is the same in all micrographs.
aluminizing 617 is outlined in Table II to facilitate description, to follow shortly, and comparison with the aluminized + clad specimens. The as-aluminized layer was identified primarily as Ni2Al3 using XRD, although Al:Ni ratios as high as 3:1 were detected by EDS near the surface. After heat treatment, the modified surface is predominantly an NiAl (B2)-based solid solution with precipitates rich in Cr and Mo. Two Cr-rich layers evolve at the interface between the main NiAl layer and the substrate (Figure 6). SAD and EDS in the TEM have shown the upper layer to consist of M23C6 carbides, whereas the one closer to the substrate is sigma (r) phase with a (Cr + Mo):(Co + Ni) ratio of ~3:2. The M23C6 + r double layer is interpenetrated by a small fraction of metal comprised of extensions of the bounding B2-NiAl and 617 alloy phases. It is evident from Figure 4 that the thickness of all layers increases METALLURGICAL AND MATERIALS TRANSACTIONS A
Fig. 3—FeCrAIY foil diffusion bonded between two 617 sheets for 1 h at 1273 K (1000 C) (a) and subsequently heat treated for 168 h (1 week) at the same temperature in gettered Ar (b). Note that the as-bonded microstructure has been etched for contrast. After heat treatment, the EDS elemental maps show relatively uniform concentrations of Ni, Fe, and Al across the original interfaces, except for intergranular regions where a second phase rich in Cr (c) and Mo (d) is clearly evident (bottom).
during the heat treatment after aluminizing, from ~6 to ~13 lm for the top Ni2Al3/NiAl layer decorated with precipitates and from ~3.5 to ~10 lm for the combined M23C6 + r layer. The formation of the M23C6 carbide layer does not appear to affect the carbides already present in the substrate, e.g., those evident as bright second phases in Figures 4(a), (b), and (e). The microstructure of the bulk 617 after aluminizing is similar to that of the as-received, uncoated 617. The previous results present an interesting opportunity. While NiAl does not form alpha alumina as readily as FeCrAlY at the temperatures of interest (cf. Figure 1), it exhibits superior stability vis-a`-vis interdiffusion with the substrate. One might then hypothesize that the NiAl layer could act as an effective diffusion barrier between FeCrAlY and 617, as long as interdiffusion between NiAl and the outer FeCrAlY layer does not impair the alumina forming ability of the latter. To assess that hypothesis, samples of alloy 617 that were previously aluminized and heat treated to form NiAl were then clad with FeCrAlY. The initial bond between the FeCrAlY cladding and the aluminized 617 appears also quite sound, Figure 7(a). After 24 hours at 1273 K (1000 C) in gettered Ar, there is some growth of the aluminized layer in addition to a recession of the B2-NiAl layer from the original bonding line, Figure 7(b). This recession, however, does not increase significantly between 24 and 168 hours, Figure 7(c). A closer examination of METALLURGICAL AND MATERIALS TRANSACTIONS A
Figures 7(b) and (c) reveals that the band of M23C6 + r phases underlying the B2 layer now contains a third constituent, appearing brighter in these BSE images. This was identified by TEM as Mo-rich M6C carbide. It is further noted in Figure 7(c) that some M6C carbides have also evolved within the 617 matrix, presumably from the predominantly M23C6 in the original material. Microprobe analyses corresponding to the micrographs in Figures 7(a), (b), and (c) are given in Figures 7(d), (e), and (f), respectively. It is first noted in these figures that the Al content of the NiAl layer remains high and relatively constant throughout the 168 hours of exposure. A comparison of Figures 7(d) and (e) reveals early diffusion of Ni into the FeCrAlY, with concomitant dilution of the Fe content in the latter, as well as counter-diffusion of Fe into the NiAl. However, the ratio of (Ni + Fe):(Al + Cr) in the B2 layer is reasonably constant from 24 to 168 hours, Figures 7(e) and (f). The Cr content in this layer remains relatively low in the same period in spite of an observed reduction in the population of Cr-rich precipitates within the B2 layer. The interdiffusion of Ni and Fe between the NiAl and the overlaying FeCrAlY is accompanied by gradual changes in the microstructure of the latter, summarized in Table II. The effects are first evident in the region between the upper NiAl boundary and the original bond line, Figures 7(b) and (c), and are characterized by a VOLUME 44A, FEBRUARY 2013—839
Fig. 4—Back scattered electron images showing the evolution of the aluminized surface layer(s) on alloy 617 at 1273 K (1000 C) in getterd Ar: (a) as-aluminized, (b) 1 h, (c) 4 h, (d) 9 h, and (e) 16 h. The aluminization conditions for this sample were 700C/4 h using CrCl3Æ6H2O as the activator. The thickness of the modified layer evolves in an approximately parabolic fashion, cf. Fig. 5.
Fig. 5—Evolution of the aluminized interlayer at the surface or under the FeCrAlY cladding during sequential thermal exposures at 1273 K (1000 C). The points below 16 h (t1/2 = 4) correspond to the initial treatment after aluminizing. The typical thickness in the as-clad specimens is then indicated. The points at longer times correspond to heat treatments in gettered Ar (empty symbols) or impure He (filled symbols) for aluminized-only (circles) and aluminized + clad specimens (squares). 840—VOLUME 44A, FEBRUARY 2013
Fig. 6—Microstructure (top) and composition profile across the aluminized layer and neighboring 617 substrate region after 16 h at 1273 K (1000 C) in gettered Ar. METALLURGICAL AND MATERIALS TRANSACTIONS A
Table II. Microstructural Evolution of Aluminized and FeCrAlY Clad + Aluminized Coatings on Alloy 617, Upon Exposures to 1273 K (1000 °C) Over Times Up to 500 h
in Figure 5, was consistent with that expected from the initial trend for a sample that had seen 17 hours total at 1273 K (1000 C), i.e., 16 hours during the heat treatment after aluminizing and 1 hour during diffusion bonding.) At longer times, the thickness of the NiAl layer appears to stabilize at ~40 lm. In principle, the interlayers produced by aluminizing and heat treatment should continue to dissolve slowly into the substrate and the cladding, especially when Al is removed at the surface by oxidation, but the evolution appears to proceed slowly as evidenced by the stability of these layers at times as long as 500 hours, noted in Table II and discussed later. The thickness of the cladding is also nominally stable, although this may not be immediately obvious in Figures 7(a), (b), and (c) because the aluminized surfaces were polished to slightly different depths prior to diffusion bonding. C. Long-Term Exposure to Impure He
The initial condition is indicated at the top. The aluminized layer was heat treated prior to cladding, so the initial condition is essentially that indicated by 1 h at 1273 K (1000 C).
rapid outward drop in the Ni, Co, and Al compositions, and the concomitant rise in the Fe and Cr content, Figures 7(e) and (f). Ordered precipitates identified as having B2 structure by SAD in the TEM are found in the disordered bcc (A2) FeCrAlY matrix below and around the original bond line, Figure 8(a). The details regarding the formation and implications of these precipitates will be discussed in more depth in a forthcoming publication. By 168 hours, the Ni distribution within the FeCrAlY shows a lower average near the surface than away from it, Figure 7(f), corresponding to the two regions of different contrast in Figure 7(c). The region with the higher Ni content nearest the NiAl layer has evolved from the dispersion of B2 precipitates in A2 to a nominally single-phase fcc (A1) solid solution, Figure 8(b), confirmed by SAD in TEM. However, this layer seems to grow slowly as inferred from the comparison of Figure 7(c) with the specimens held for 500 hours, described in the following section. The presence of the cladding does not appear to affect substantially the microstructural changes accompanying the interdiffusion between the NiAl layer and the substrate. The evolution of the thickness of the aluminized layer depicted in Figure 5 reveals that after cladding, there is an initial period of further broadening, following essentially the trend established prior to cladding, indicated by the dashed line. (Note that the initial aluminized thickness in the samples that were to be clad was significantly higher than in Figure 4(a) to allow for the necessary removal of the surface roughness prior to bonding. In consequence, the thickness of an as-clad sample, representing the initial condition for all the longer diffusion treatments and denoted by an arrow METALLURGICAL AND MATERIALS TRANSACTIONS A
The aluminized samples performed remarkably well in relation to uncoated 617,[5] both in carburizing and decarburizing environments. A comparison of weight gain measurements with the aluminized + clad samples could not be made rigorously because of significant edge degradation in the former which led to significant weight gains, especially during carburization. (Note that the edges were not clad and experienced damage during sample preparation prior to the experiment, which led to partial delamination of the cladding). However, the aluminized samples showed less than 10 pct of the increase in weight per unit area shown by the uncoated 617 after 500 hours, regardless of environment; the gain was consistent with an overall thickening of the alumina scale from ~100 to 200 nm to ~1 lm. Notwithstanding the uncertainty in the weight measurements for the clad samples, the micrographs in Figures 9(a) and (b) and 10(a) and (b) reveal that the thin alpha alumina film produced by pre-oxidation still appears continuous and dense, acting as a barrier to oxygen diffusion inward, and presumably C diffusion inward or outward, depending on the C potential in the environment. Figures 9(c) and (d) and 10(c) and (d) show significant evolution of the aluminized and clad layers during the longer treatment, compared with the as-processed conditions in Figures 4(e) and 7(a). The original layers are still identifiable after the long-term exposure at 1273 K (1000 C), but have clearly experienced significant evolution in their constitution (Table II). Compositional profiles of the aluminized samples after 500 hours at 1273 K (1000 C) are presented in Figures 9(e) and (f) with the continued phase evolution of the coating given in Table II. It is evident that the Ni:Al ratio in the near-surface layer has evolved to values as high as ~3, e.g., Figure 9(f), consistent with the gradual evolution of Ni3Al (L12) domains within the NiAl layer. The Ni3Al phase appears brighter in the BSE images of Figures 9(c) and (d) because of its lower Al content. Note also that the NiAl phase is still continuous in Figure 9(c), but is discontinuous in the thinner aluminized layer of Figure 9(d). Nevertheless, both specimens retain enough Al to preserve the stable VOLUME 44A, FEBRUARY 2013—841
Fig. 7—SEM images of the microstructure evolution of aluminized 617 that has been clad with FeCrAlY and corresponding EMPA composition profiles. As-clad (a, d) and the diffusion bond after 24 (b, e) and 168 (c, f) hour heat treatments. Note that the aluminized layer grows initially, but then appears to be relatively stable after 24 h.
alumina scale on the surface as evidenced by Figures 9(a) and (b). It is also noted that the nearly continuous r phase layer formed during the shorter heat treatment after aluminizing, e.g., Figure 4(e), has now been largely replaced by a layer of M6C carbides right above the 617 substrate, shown brighter in the BSE images of Figures 9(c) and (d). Some M6C carbides associated with the M23C6 carbides are also evident within the substrate, e.g., in Figure 9(d). The EMPA profiles in Figures 9(e) and (f) show major Cr peaks corresponding to the M23C6 carbides and Mo peaks at the location of the M6C carbides. (The C content is not included in the analysis, so the Cr and Mo concentrations are based only on the overall metal content). These carbide layers are still interpenetrated by minor fractions of metallic B2 (NiAl) and A1 (617) phases. In the case of the aluminized + clad samples, Figures 10(c), (d), (e), and (f), the NiAl layer is still 842—VOLUME 44A, FEBRUARY 2013
present and has not grown substantially compared with the observations after the 24 and 168 hour heat treatments, Figure 5. The Al content of the B2 layer is slightly lower than that in Figures 7(d), (e), and (f), but significantly higher than the samples that were aluminized, but not clad. The NiAl phase is still continuous after the 1273 K (1000 C)/500 h exposure in both He environments, but there is a significant fraction of a second phase different from Ni3Al present in both cases. The M23C6 layer under the NiAl is still present and relatively continuous, but the r layer is less evident in Figures 10(c) and (d) and only discrete M6C grains are observed at this interface (cf. Table II). There are, however, M6C carbides clearly distinguishable within the substrate. The diffusion profiles for the FeCrAlY cladding in Figure 10(e) still show the step in the distribution of Ni associated with the change from B2/A2 to c (A1) microstructures noted in the cladding METALLURGICAL AND MATERIALS TRANSACTIONS A
preliminary tests should arguably be preventable by improvement in the sample preparation procedures; hence, those results are not further discussed here. In this context, the discussion is limited to the oxidation and interdiffusion phenomena, the underlying fundamentals and the implications for the implementation of the barrier concepts. A. Barrier Formation and Stability in Low pO2
Fig. 8—SEM images of the region near the FeCrAlY/NiAl interface in clad + aluminized 617 samples after heat treatment at 1273 K (1000 C) for (a) 24 h and (b) 168 h—details from Figs. 7(b) and (c). The precipitates above the interface in (a) were identified as B2 (Ni,Fe)(Al,Cr), but are absent in the same region at longer times, (b) Concurrently, the b.c.c (A2) matrix evolves into an f.c.c (A1) solid solution.
after 168 hours, Figure 7(f), but not in Figure 10(f). In this case, the cladding was polished down to a thinner layer that has fully transformed to c. While the Ni concentration increases in the section of FeCrAlY that has transformed to c, cf. Figures 10(e) and 7(f), the extent of this layer has not increased significantly when compared to the treatments at 168 hours. It is also noted that specimens exposed in both carburizing and decarburizing He environments show lower Al contents throughout the FeCrAlY layer than the specimens in Figure 7.
IV.
DISCUSSION
The salient finding from this investigation is that the barrier layer concepts proposed perform much better than the uncoated alloy 617 in impure He environments over a wide range of C/O potentials,[5] preventing the ingress/egress of C that can lead to undesirable changes in mechanical properties.[7,8] In areas where the alumina scale is not breached, there is no evidence of a connection between the environmental exposure and the observed microstructural changes, Figures 9 and 10, except for the depletion of Al due to oxidation. Interdiffusion between the layers appears to play a much stronger role in their evolution as discussed below. Where differences are observed between the carburizing and decarburizing environments, they appear to be primarily related to differences in the thickness of the aluminized and/or clad layers owing to the variability in sample preparation, and not to the differences in C or O activity in the environment. When damage occurred in the samples prior to the impure He exposures, as in the clad coupons, there was obvious degradation in the damaged areas that did depend on the specific environment. While this suggests that the effect of damage needs to be further investigated, the anomalous degree of damage in these METALLURGICAL AND MATERIALS TRANSACTIONS A
The rationale for understanding the oxidation behavior of the engineered surface concepts proposed in low pO2 environments is self-evident. The He in the reactor circuit is expected to be pressurized to ~8 MPa[1–3] and contains ppm or lower concentrations of H2/H2O, CH4/CO/CO2 that, in principle, determine the carbon and oxygen potentials.[4] For chromia formers, it has been argued that typical impurity concentrations in reactor He can lead to a combination of high C activity and low pO2 to destabilize the scale, leading to extensive formation of carbides.[4,22] For alumina formers, one can calculate that in the range of 1073 K to 1273 K (800 C to 1000 C), the oxide is stable on pure Al for oxygen concentrations of ~10 44 to ~10 35. The Al activity in FeCrAl at 1123 K (850 C) is reported as ~1.2 9 10 4[23] and ranges from 1.4 9 10 5 to 1.8 9 10 4 for Ni60Al40 from 1073 K to 1273 K (800 C to 1000 C) and from 3.7 9 10 4 to 3.3 9 10 3 for Ni50Al50 in the same range.[24] Taking the values for the lower Al content, one may estimate that the upper bound of the dissociation pO2 for Al2O3 would be within ~10 38 and ~10 30 bar (~10 33 to ~10 26 Pa). All gas environments in this investigation had oxygen concentrations well above this limit, from gettered Ar with XO [ 10 13 to XO 10 21 for a decarburizing environment (CO:CO2 ~9 and CO + CO2 ~15 ppm) to XO 10 24 for a carburizing environment (CO:CO2 ~1272 and CO + CO2 ~1910 ppm). This is consistent with the observation that a thin, stable alumina forms upon pre-oxidation in Ar and is preserved in both impure He environments. What is less obvious is whether the formation of alumina is sufficiently competitive kinetically with the formation of Cr or Mo carbide for the higher CO:CO2 environments in the event of cracking or spallation. Carbide formation could, in turn, have important implications for the susceptibility of the structure to creep-fatigue damage.[9] This issue is currently under investigation. The selection of a low pO2 environment instead of air for building the alumina scale prior to He exposure deserves further discussion. Note in Figure 2(a) that preoxidation in air leads to a scale with a dense underlayer and a rather open overlayer of clusters with the typical lamellar morphology of transient aluminas.[15–18] While these are fully transformed to alpha after 1 hour at 1273 K (1000 C) for FeCrAlY, the rough morphology is arguably not desirable for surfaces that will be exposed to high velocity gas flows. The clusters may break away under the aerodynamic friction and locally detach the underlying dense layer. The smoother surfaces resulting from the low pO2 pre-oxidation are much more desirable in this regard. More importantly, the transient aluminas grow significantly faster than the alpha phase, promoting rapid depletion of Al from VOLUME 44A, FEBRUARY 2013—843
Fig. 9—Cross sections of the aluminized 617 samples and EPMA composition profiles after 500 h exposures to He environments at 1273 K (1000 C) with CO:CO2 ratios corresponding to decarburizing (a, c, e) and carburizing (b, d, f) conditions. The dense, continuous alumina film produced by pre-oxidation in low pO2 environment thickens during exposure and remains adherent. The morphological evolution of the aluminized layer does not appear to be related to any chemical interaction with the environment.
the outer surface. This may be less critical for the aluminized layer, which represents a rather large reservoir, but is known to be detrimental in thin cross sections of FeCrAlY.[23] Conversely, the transformation of transient aluminas into alpha takes longer for the aluminized surfaces, Figure 1, and that difference is accentuated at lower temperatures. 844—VOLUME 44A, FEBRUARY 2013
The effects of oxidation at low pO2 have been explored in the past,[23,25] but the underlying link to the competition between metastable and stable alumina forms has not been elucidated. It is well established that corundum (alpha) is the stable form of alumina at all temperatures, but that metastable spinel-based (c, h) forms typically evolve first in a wide variety of processes, METALLURGICAL AND MATERIALS TRANSACTIONS A
B. Interdiffusion While the temperature + lifetime demands of the VHTR system remain a paramount challenge to surface engineering concepts involving layered architectures susceptible to interdiffusion, the relative stability of the configurations explored in this work bodes well as a starting point for a durable solution. The crucial role of the B2-NiAl layer is self-evident, acting as the Al reservoir to sustain the formation of a-Al2O3 as well as a nominal barrier to the interdiffusion of the FeCrAlY cladding with the substrate, which was found to be quite rapid at 1273 K (1000 C). However, it can be shown from multicomponent phase diagrams calculated from existing thermodynamic databases** that the amount of **All references to thermodynamic calculations in this paper pertain to the Pandat 8 software used in conjunction with the PanNi8 database, both commercially available from CompuTherm, Inc. In the interest of space, these issues will be discussed in greater detail in a separate publication, but the calculations should be readily accessible to any reader familiar with this approach.
Fig. 10—Cross sections of the aluminized + clad 617 samples and EPMA composition profiles after 500 h exposures to He environments at 1273 K (1000 C) with CO:CO2 ratios corresponding to decarburizing (a, c, e) and carburizing (b, d, f) conditions. The dense, continuous alumina film produced by pre-oxidation in a low pO2 environment remains adherent and thickens during exposure. Note the more significant diffusion of Ni into FeCrAlY relative to the shorter exposures.
from oxidation to crystallization of amorphous oxides. It has been proposed that the corundum structure is much less tolerant to disorder than the spinel forms of alumina. The inference is that under conditions of rapid growth, where defects are more likely to be introduced, the spinel form is favored over corundum.[26] It then follows that slow growth conditions, such as those promoted by low pO2,[27] would increase the competitiveness of alpha over the transient aluminas. The evidence supports this view, but it is also clear that the transient aluminas are not fully suppressed, especially in the aluminized-only surfaces. The net effect is to minimize their growth prior to the nucleation of alpha, allowing the latter to establish its predominance over the oxidation kinetics, while the scale is still thin. METALLURGICAL AND MATERIALS TRANSACTIONS A
Al in the initial NiAl surface layers, 2 9 ~25 lm with ~38 at. pct Al (Figure 6), could be completely dissolved even in the relatively thin (1 mm) 617 substrate used here without changing the phase constitution of the latter at 1273 K (1000 C). The system is thus inherently metastable, but the multilayer architecture evolves sufficiently slowly for times longer than ~40 hours to make it relatively stable to at least 500 hours, Figure 5. The sluggish evolution has been observed in other aluminized Ni alloy systems and is ascribed to the presence of interphases that act as diffusion barriers between the B2 phase and the substrate alloy, cf. Figure 6, notably the r phase.[20,28,29] It is therefore instructive to examine how the system evolves toward this relatively stable configuration. To the best of the authors’ knowledge the only article in the open literature dealing with high activity aluminizing of alloy 617 is of a single-stage high temperature (1273 K (1000 C)) process[30] rather than the more common two-stage low temperature (~800 C ± 50 C) process, like the one discussed here. The behavior in the first stage is exemplified by the aluminizing of elemental Ni, which for thinner coatings (~35 lm) is reported to form only Ni2Al3 of relatively uniform composition above the c-Ni substrate.[20] The rationale is that inward Al diffusion is much faster than outward Ni diffusion in Ni2Al3,[20] but the broader implication would be that all other intermetallics in the binary Al-Ni system are initially suppressed under these conditions. Available thermodynamic databases confirm the feasibility of a metastable binary equilibrium between Ni2Al3 and c-Ni(A1) with extended solubilities at 973 K (700 C) of 50 pct Ni and 27 pct Al, respectively, compared with equilibrium values of 41 pct Ni and 10 pct Al. What may lead to the kinetic suppression of the other phases, however, is not clear. Thicker coatings (~220 lm) formed on pure Ni at longer times (6 h/1123 K (850 C)) do exhibit the expected thin layers of L12 and B2 (with distinct Ni-rich and Al-rich regions) phases VOLUME 44A, FEBRUARY 2013—845
between the c-Ni and the Ni2Al3.[31] When other elements are present in solid solution in the Ni, e.g., Cr and Mo, their mobility is marginal and their solubility in Ni2Al3 is quite limited at the aluminizing temperatures. They are thus trapped behind the advancing Ni2Al3 front and precipitate out, typically as a-(Cr) (A2) or aluminides, e.g.[32] When carbides are present, they are also retained in their original location behind the aluminizing front.[20,28,30] As the Ni2Al3 layer thickens, the Al transport to the interface slows down. NiAl eventually nucleates and grows at the expense of the original Ni2Al3, combining Al diffusion from the latter with Ni diffusion from the substrate, which predominates on the Ni-rich side of the B2 field.[33,34] However, this second stage usually occurs in a heat treatment at higher temperature, of order 1273 K (1000 C), after aluminizing, although similar behavior ensues when the high activity aluminizing is performed at 1273 K (1000 C).[30] The outward Ni diffusion has two important implications. It gives rise to the two distinct NiAl layers evident in Figure 6, one containing the precipitates present in the original Ni2Al3 and the other essentially single phase. With the conversion of Ni2Al3 into NiAl, the original intermetallic precipitates within the former change to A2. More importantly, as Ni is removed and some Al diffuses inward, the affected substrate region evolves through the sequence c-(Ni) fi c + c¢ fi c¢(L12) fi c¢ + b fi b (B2) with a progressive loss of solubility of C and refractory elements (Cr, Mo, etc.).[20,35] These form a band of topologically close-packed (TCP) phases (and carbides when C is present in significant amounts) interpenetrated by alloy, typically known as the interdiffusion zone (IDZ). Because Al transport through these layers is essentially limited to the fraction of metallic phase in them plus grain boundary contributions, the IDZ can, in principle, play a significant role in preserving the high Al concentration within the surface layers.[28] Among the TCP phases, r has attracted particular attention as a potential diffusion barrier.[29] Cr and Mo form r in binary combinations with Fe or Co, but not with Ni, although thermodynamic calculations predict a stable r phase in the ternary Cr-Mo-Ni at ~30 pct Ni, from ~20 to ~65 pct Cr. Accordingly, the precipitates in the IDZ for aluminized Ni-Cr alloys are simply a-Cr.[36] The addition of a modest amount of Fe in alloy 600 yields r in the IDZ during the first stage, but it decomposes into M23C6 and B2 when heated to a higher temperature in the second stage.[28] When Co and Mo are added in alloy 617, the r phase forms in the heat treatment after aluminizing, Figure 4, and remains stable for at least 168 hours at 1273 K (1000 C), Figure 7. However, r is replaced after 500 hours by discontinuous M6C, Figures 9 and 10, at variance with the behavior of 617 aluminized in one step at 1273 K (1000 C) where the M23C6 + r bi-layer remained stable after 480 hours at 950 C.[30] The thickness of the IDZ is reported to evolve from ~10 lm after aluminizing to ~40 lm after the exposure above, with both layers growing. In contrast, the M23C6 and r layers here grow to a thickness of ~10 lm during the early heat treatment, Figure 4, and ~16 lm after 846—VOLUME 44A, FEBRUARY 2013
168 hours, Figure 7, with the relative fraction of M23C6 increasing gradually, but is subsequently reduced after 500 hours, e.g., Figure 9(a) as r disappears. The reasons for the discrepancy in durability of the r layer between this study and that reported in the literature are not clear. The somewhat lower temperature for the extended heat treatment in the latter could play a role, but the single step process appears to have led to a somewhat different composition of the r phase with a significantly higher Cr content and lower Ni content. The inference is that tailoring the chemical composition of this layer can enhance its stability. The concept has been explored in the literature by Re deposition on the surface prior to aluminizing to yield a relatively continuous r layer, but the implementation is cumbersome since the diffusion barrier also precludes the outward diffusion of Ni.[29,37] It is notable that in spite of the different behaviors of the putative diffusion barriers between this study and the previous one in the literature,[30] both aluminide layers evolve toward a relatively steady thickness as seen from comparing Figure 5 with Figure 6 in Narita et. al.[29] This suggests that the carbide layer may be just as effective a diffusion barrier as r, although the former does not normally occur in the higher temperature superalloys and there is obviously a concern with a relatively continuous layer(s) of brittle phases between the coating and the substrate. Preliminary observations have shown that whenever microcracks appear in the carbide layer, they seem to arrest at the metal pockets within it, notwithstanding the fact that the metal phase is arguably brittle (B2). Hence, the overall issue of diffusion across the IDZ layers and the mechanical integrity of the phases in this region remain under investigation. When cladding is present, the B2 layer represents an effective barrier to the diffusion of Cr, the solubility of which is limited in NiAl,[35] and so this is not considered a major issue. The larger concern is the potential for Al and Ni migration from NiAl into the ferritic A2 layer. The former is desirable at a rate sufficient to sustain the a-Al2O3 growth at the surface (or healing if the scale is damaged), but not more rapidly since that would deplete the NiAl reservoir faster than needed and may change the phase constitution of the FeCrAlY. The evolution of the Al concentration profiles in Figures 7 and 10 suggests that this goal is met. However, the same figures indicate that Ni diffusion into the FeCrAlY is faster than that of Al and leads to significant microstructural changes, cf. Table II. The first one is the precipitation of fine coherent B2 phases in the A2 matrix, Figure 8, which is reminiscent of the c-c¢ microstructure in superalloys and could lead to improvements in the mechanical properties of the surface layer, a desirable target if creep-fatigue issues are a concern. Unfortunately, the continued migration of Ni leads to the conversion of the B2/A2 mixture into c (A1) solid solution in the vicinity of the NiAl layer. However, the transformation front appears to slow down over time, perhaps because of the rejection of Cr, which is an A2 stabilizer, into the top layer, e.g., Figure 10(a). Note, however, that the FeCrAlY is almost completely converted into c in Figure 10(b) where the initial thickness METALLURGICAL AND MATERIALS TRANSACTIONS A
of the cladding was smaller due to additional material removed during sample preparation. Research continues on these issues.
V.
CONCLUSIONS
The prospective use of alloy 617 in the IHX of the VHTR at the desired operation temperatures ‡1223 K (950 C) is limited by its inadequate resistance to carburization/decarburization that may compromise its structural integrity. This work demonstrated that a stable alpha alumina scale enabled by modifying the surface of alloy 617 by aluminizing, either with or without FeCrAlY cladding, was effective in protecting 617 in impure He environments. The lifetime requirements of the VHTR demand that the coatings implemented are diffusionally stable on the underlying alloy so that they maintain an Al reservoir necessary to sustain a continuous alpha alumina scale. The development of carbide and r interlayers in the aluminized coatings slows the diffusion of Al into the alloy, retaining a suitable Al concentration for maintaining a protective scale. The aluminized layer is also an effective diffusion barrier between FeCrAlY cladding and alloy 617. While both coating concepts were shown to be effective barriers to inward or outward C diffusion when 617 is in contact with impure He, FeCrAlY is a better alpha alumina former, and therefore more likely to readily reform a protective alpha alumina scale if cracking or spalling of the oxide scale were to occur during service.
ACKNOWLEDGMENTS This investigation was supported under DOE-NERI award DE-FG07-07ID14894 and DOE-NEUP award 00102215. The authors are grateful to the late Prof. A.G. Evans for inspiring this work, to Prof. T.M. Pollock for helpful discussions, and to Dr. G.G.E. Seward for his assistance with the EMPA analysis. The research made use of the UCSB MRL Central facilities supported by the MRSEC Program of the National Science Foundation under award No. DMR11-21053.
REFERENCES 1. T. Burchell, R.L. Bratton, R.N. Wright, and J. Wright: INL/EXT06011701, Idaho National Laboratory, Idaho Falls, ID, 2007. 2. T.R. Allen, K. Sridharan, L. Tan, W.E. Windes, J.I. Cole, D.C. Crawford, and G.S. Was: Nucl. Technol., 2008, vol. 162, pp. 342–57. 3. W. Ren and R. Swindeman: J. Press. Vessel Technol., 2009, vol. 131, pp. 044002-1–15.
METALLURGICAL AND MATERIALS TRANSACTIONS A
4. W.J. Quadakkers and H. Schuster: Werkst. Korros., 1985, vol. 36, pp. 141–50. 5. D. Kumar, R.A. Adharapurapu, T.M. Pollock, and G.S. Was: Metall. Mater. Trans. A, 2011, vol. 42A, pp. 1245–65. 6. R. Adharapurapu, D. Kumar, J. Zhu, C.J. Torbet, G.S. Was, and T.M. Pollock: Metall. Mater. Trans. A, 2011, vol. 42A, pp. 1229– 44. 7. J. Klo¨wer and U. Heubner: Mater. Corros., 1998, vol. 49, pp. 237– 45. 8. P.S. Shankar and K. Natesan: J. Nucl. Mater., 2007, vol. 366, pp. 28–36. 9. T.C. Totemeier and H. Tian: Mater. Sci. Eng., A, 2007, vols. 468– 470, pp. 81–87. 10. G.W. Hollenberg, E.P. Simonen, G. Kalinin, and A. Terlain: Fusion Eng. Des., 1995, vol. 28, pp. 190–208. 11. A.G. Evans, D.R. Clarke, and C.G. Levi: J. Eur. Ceram. Soc., 2008, vol. 28, pp. 1405–19. 12. P. Becker and D.J. Young: Oxid. Met., 2007, vol. 67, pp. 267–77. 13. K.S. Forcey, D.K. Ross, and C.H. Wu: J. Nucl. Mater., 1991, vol. 182, pp. 36–51. 14. C.F. McDonald: Appl. Therm. Eng., 1996, vol. 16, pp. 3–32. 15. M.W. Brumm and H.J. Grabke: Corrosion, 1992, vol. 33, pp. 1677–90. 16. J.C. Yang, E. Schumann, I. Levin, and M. Ru¨hle: Acta Mater., 1998, vol. 46, pp. 2195–01. 17. H.J. Grabke: Intermetallics, 1999, vol. 7, pp. 1153–58. 18. V.K. Tolpygo and D.R. Clarke: Mater. High Temp., 2000, vol. 17, pp. 59–70. 19. F.H. Stott, G.C. Wood, and M.G. Hobby: Oxid. Met., 1971, vol. 3, pp. 103–13. 20. G.W. Goward and D.H. Boone: Oxid. Met., 1971, vol. 3, pp. 475– 95. 21. D. Kumar, C.J. Torbet, and G.S. Was: Meas. Sci. Technol., 2009, vol. 20, pp. 095708-1–11. 22. W.J. Quadakkers and H. Schuster: Werkst. Korros., 1885, vol. 36, pp. 335–47. 23. N.S. Jacobson, J.L. Smialek, J.I. Eldridge, and D.L. Humphrey: Oxid. Met., 2008, vol. 69, pp. 343–58. 24. L. Bencze, D.D. Raj, D. Kath, W.A. Oates, L. Singheiser, and K. Hilpert: Metall. Mater. Trans. B, 2004, vol. 35B, pp. 867–76. 25. C.G. Levi, E. Sommer, S.G. Terry, A. Catanoiu, and M. Ru¨hle: J. Am. Ceram. Soc., 2002, vol. 86, pp. 676–85. 26. C.G. Levi: Acta Mater., 1998, vol. 46, pp. 787–800. 27. J. Toscano, A.G. Naumenko, L. Singheiser, and W.J. Quadakkers: Mater. Corros., 2008, vol. 59, pp. 501–07. 28. A. Chien, D. Gan, and P. Shen: Mater. Sci. Eng., 1996, vol. A206, pp. 215–24. 29. T. Narita, K.Z. Thosin, L. Fengqun, S. Hayashi, H. Murakami, B. Gleeson, and D. Young: Mater. Corros., 2005, vol. 56, pp. 923–29. 30. T.S. Jo, S.J. Kim, and Y.D. Kim: Surf. Coat. Technol., 2011, vol. 205, pp. 4743–49. 31. A. Chien, D. Gan, and P. Shen: Mater. Sci. Eng., A, 1999, vol. A272, pp. 207–14. 32. H.L. Huang, Y.Z. Chen, and D. Gan: Mater. Sci. Eng., A, 2002, vol. A328, pp. 238–44. 33. A. Paul, A.A. Kodentsov, and F.J.J. van Loo: J Alloys Compd., 2005, vol. 403, pp. 147–53. 34. Q.C. Xu and A. Van der Ven: Phys. Rev. B, 2010, vol. 81, p. 064303. 35. N.C. Oforka and C.W. Haworth: Scand. J. Metall., 1987, vol. 16, pp. 184–88. 36. H.L. Huang and D. Gan: Mater. Sci. Eng., A, 2008, vol. 485, pp. 550–57. 37. Y. Wu, X.W. Li, G.M. Song, Y.M. Wang, and T. Narita: Oxid. Met., 2010, vol. 74, pp. 287–303.
VOLUME 44A, FEBRUARY 2013—847