Oxidation behaviour of novel ODS FeAlCr intermetallic alloys

(1)

Oxidation behaviour of novel ODS FeAlCr intermetallic alloys

M.A. Montealegre

^a

, G. Strehl

^b

, J.L. Gonza´lez-Carrasco

^a,

*, G. Borchardt

^b

aCentro Nacional de Investigaciones Metalu´rgicas, CENIM-CSIC, Avda Gregorio del Amo 8, Madrid 28040, Spain

bInstitut fu¨r Metallurgie, TU Clausthal, Clausthal-Zellerfeld D-38678, Germany Received 1 July 2004; received in revised form 16 November 2004; accepted 15 February 2005

Available online 7 April 2005

Abstract

The isothermal oxidation behaviour of novel oxide dispersion strengthened (ODS) FeAlCr intermetallic alloys was investigated at 900 and 1100 8C for up to about 150 h, with special emphasis on the oxidation kinetics, scale morphology and scale adherence. Comparisons with other alumina forming ferritic alloys and intermetallics are included. At 1100 8C, the new FeAlCr alloys show lower oxidation rates whereas at 900 8C they exhibit higher oxidation rates, which is associated with the formation of metastable q-alumina scales. All intermetallic alloys are prone to scale spallation at 1100 8C during the cooling stage, which results from the elevated compressive residual stresses developed as a consequence of the large difference of thermal expansion coefficients between alumina and the metal substrate. In addition, the novel alloys suffer from subscale void formation. It is proposed that a decrease in the Al concentration beneath the scale yields a smaller number of thermal vacancies in equilibrium and excess of vacancies condense as voids. The absence of such oxidation induced voids in the Cr-free Fe40Al-Grade 3, having the same grain size and volume fraction of dispersoid, highlights the role of Cr in the vacancy behaviour.

Keywords: A. Iron aluminides (based on FeAl); B. Oxidation; B. Alloy design

1. Introduction

Iron aluminides are being proposed as engineering materials for high temperature applications and as possible substitutes for stainless and special steels at room temperature Their major advantages are their low density (!6 g cm^K3), low raw materials cost, and low content of strategic elements. The major problem is their brittleness at room temperature[1], which creates difficulties during their production, and their poor strength and creep resistance above 600 8C[2]. Efforts to improve their ductility at room temperature have been made by appropriate processing, microalloying, and oxide dispersion strengthening (ODS), which have led to grain boundary strengthening and grain size reduction [3,4]. Due to the good results, these materials have attracted increasing attention in recent years.

Iron aluminides are, generally, considered to be oxidation resistant because of their ability to form a surface alumina scale. A recent model proposed by Evans and Strawbridge [5] suggests that materials with low creep strength will be able to dissipate strain energy during cooling by deforming and, therefore, iron aluminides should form very adherent scales. However, relative to other alumina forming alloys, iron aluminides are prone to suffer scale spallation at high temperatures (O1000 8C)[6,7]even when adding reactive elements [8–11] that are expected to enhance scale adhesion. Additionally, iron aluminides can suffer from subscale void formation during oxidation [12–14]. These can be as deep as the thickness of the scale above them. Therefore, void formation must be considered as an important aspect, for instance, when using the alloy for thin components or for protective coatings.

The major drawback of iron aluminides is the suscepti- bility to environmental embrittlement at room temperature due to reaction of the Al with water vapour, which produces atomic hydrogen[15]. Alloying with Cr has been shown to be an effective method to minimize this effect [16–18].

Moreover, Cr is currently added to many engineering alloys in order to increase their oxidation resistance.

www.elsevier.com/locate/intermet

doi:10.1016/j.intermet.2005.02.003

* Corresponding author. Tel.: C34 915 538 900; fax: C34 915 347 425.

E-mail address: jlg@cenim.csic.es (J.L. Gonza´lez-Carrasco).

(2)

The investigated yttria dispersed FeAlCr intermetallic alloys have been developed as part of the European Research Project ALUSI.¹ Designing alloys for a given application always considers a balance of the material properties. To assess the effect of the alloy elements, two levels of Cr, Al, and yttria were selected in such a way that it would be possible to determine the role of these elements on the chemical and physical properties. Addition of yttria is aimed both to produce a fine-grained microstructure and to improve the oxidation behaviour, while still retaining useful mechanical properties. Alloying with Cr is intended to increase the corrosion resistance at ambient temperature in very aggressive environments. Previous studies have investigated the effects of Cr additions on the oxidation resistance of aluminides type Fe₃Al [19–22]. However, information on the effect of Cr on the oxidation behaviour of FeAl is scarce and limited to coatings[23].

In this study, the oxidation behaviour of the novel FeAlCr intermetallic alloys was investigated at 900 and 1100 8C, with special emphasis on the oxidation kinetics, scale morphology and scale adherence. When appropriate, comparisons with alumina forming alloys known by their oxidation resistance, like Fe40Al-Grade 3 intermetallic[10, 11,13,24,25]and PM 2000[26–29], are included. Two stage oxidation experiments under ¹⁶O and ¹⁸O were used to determine the oxide scale growth mechanism at 1100 8C.

2. Experimental

The materials used have been produced through a powder metallurgy route consisting of mechanical alloying followed by thermomechanical treatments at elevated temperatures. The alloys were obtained from two different sources. The new experimental FeAlCr intermetallic alloys, hereafter ALUSI alloys, and the PM 2000 were supplied by Plansee (Reutte, Austria) as hipped or hot rolled bars (35 mm diameter), respectively. The Fe40Al-Grade 3 alloy was supplied by DTEN, CEA (Department Technique Energie Nouvelle, Grenoble, France) as hot extruded bars of 15 mm in diameter. Nominal compositions of the materials are given in Table 1. In the as-received condition, the intermetallic alloys present a fine-grained microstructure with an average grain size of about 2 mm (ALUSI alloys) and 1 mm (Fe40Al-Grade 3). In the case of the PM 2000, small grains of about 1 mm in diameter, slightly elongated in the longitudinal bar direction, were observed. The ALUSI alloys, in addition to the Y₂O₃particles, contain a fine Al₂O₃ dispersion, due to the pick up of oxygen during milling (carried out in argon), and a large amount of finely dispersed Cr-rich precipitates. The Fe-based intermetallic alloys are

non-ferromagnetic at room temperature, whereas the PM 2000 alloy exhibits soft ferromagnetic behaviour[30].

Thermal expansion measurements were performed up to 1200 8C under Helium in a LK02 dilatometer by using rods of 2 mm diameter and 12 mm long. A heating and cooling rate of 0.3 8C s^K1was used.

Isothermal-oxidation experiments were performed in a thermobalance at 900 and 1100 8C for exposures of up to about 150 h. The sample support was made from alumina.

Through the balance and the furnace tube synthetic air flowed at a rate of 5 ml min^K1, mainly to protect the balance from ascending hot gas. The accuracy of the employed Mettler Toledo AT261 DeltaRange balance has been set to 0.1 mg, because this is at the same time the range of measurement fluctuations induced by gas convection inside the tube. Because the mass gain of the samples, especially at the lower temperature, was only a few times larger than the resolution of the balance, determination of the true initial mass plays an important role for the calculation of mass gain per unit area from thermogravimetric raw data. A very reliable method to do this is to determine the difference in the buoyancy effect at the end of the measurement. Fig. 1 demonstrates that this method produces mass gain curves that are in good agreement with data determined by discontinuous oxidation in a conventional furnace.

For the oxidation tests, cylindrical samples with a diameter of 15 and 1 mm thickness were cut by electrospark erosion. Surfaces were abraded on successively finer silicon carbide papers, also rounding the corners, and mechanically polished with 1 mm diamond paste to a mirror like finish.

Finally, cleaning in a Soxhlet reflux condensation apparatus with isopropanol was carried out. Sample dimensions and weight were recorded before and after oxidation.

To elucidate the alumina scale growth mechanism two- stage oxidation experiments were performed at 1100 8C for two durations on selected specimens. The samples were first oxidised in¹⁶O under about 200 mbar pressure for 3 h.

The oxidant atmosphere was then evacuated (2 min) without cooling the samples and replaced by ¹⁸O (isotopic purity of 91%) under about 200 mbar pressure. After a period of 9 h, the ¹⁸O was pumped out, then the specimen was pulled out of the furnace. The isotopic distribution in the alumina scale was monitored by Secondary Neutral Mass Spectrometry (SNMS). Secondary Ion Mass Spec- trometry (SIMS) was used to determine the depth profile of major elements in the oxide scale. To understand the differences observed between the tested samples, oxidation products on selected specimens were characterised by X-ray diffraction (XRD) and scanning electron microscopy (SEM) with a field emission gun (FEG) coupled with an energy dispersive X-ray (EDX) system for chemical analysis. To reveal the grain size pattern of the alumina scale, selected specimens were broken at ambient temperature after immersion for several minutes in liquid nitrogen and inspected by SEM.

1Growth Programme ‘Development of alumina forming ODS ferritic superalloys as new biomaterials for surgical implants (ALUSI)’ GRD1- 1999-10659.

(3)

3. Results

3.1. Thermal expansion measurements

Fig. 2 shows the evolution of the mean thermal expansion coefficient as a function of temperature for the investigated alloys. Differences in the curves obtained in the cooling run (not displayed in the figure) were slight compared with those obtained during heating and, therefore, no phase changes are believed to occur during heating. For comparative purposes, values for PM 2000 [31] and polycrystalline alumina [32] are also incorporated in the plot. It is apparent that expansion for the iron aluminides is higher than for the PM 2000 alloy, being values much higher for both types of alloys than for polycrystalline alumina. When comparing the values for the intermetallic alloys, it is evident that Cr addition considerably increases the thermal expansion coefficient (up to 50% for the ALUSI 2 compared with Fe40Al-Grade 3), especially at temperatures above 400 8C. The effect of Al (see values for the ALUSI 1 and 2) is also to increase the thermal expansion coefficient, although in this case a smaller contribution of about 20% was observed. A decrease in the yttria content seems not to have any effect.

3.2. Oxidation behaviour

Fig. 3shows the mass gain behaviour for the investigated alloys at 1100 8C (Fig. 3a) and 900 8C (Fig. 3b) for about 150 h exposure in air. Comparative analysis of the oxidation kinetics at 1100 8C shows that mass gains decrease in the order PM 2000, Fe40Al-Grade 3 and ALUSI alloys. Within the ALUSI alloys, the lowest mass gain corresponds to the ALUSI 2, irrespective of the oxidation time. It, generally, appeared that mass gain for the ALUSI 3 is somewhat lower than for the ALUSI 1 and 4, which show similar behaviour.

At 900 8C, however, mass gains for the ALUSI alloys are higher than for the PM 2000, which shows somewhat higher values than the Fe40Al-Grade 3 alloy. Differences in mass gain between the ALUSI alloys are insignificant and fall within the scatter range.

The curves ofFig. 3were fitted to the power law

Dm Z ktⁿ (1)

Where Dm is the mass gain per unit area, t is the exposure time, n is the rate exponent, and k the corresponding oxidation rate constant. The relevant kinetic parameters for the specimens oxidised at 1100 and 900 8C are given in Table 2. As can be seen, at 1100 8C the oxidation kinetics are best described by a parabolic growth law. The variations

Table 1

Chemical composition (nominal values) of the investigated alloys in mass%

Fe Al Cr Y2O3 Ti S (ppm)^a C %

ALUSI 1 Bal. 20 12 1 – 10–15 0.014

ALUSI 2 Bal. 25 12 1 – 8 !0.010

ALUSI 3 Bal. 20 8 1 – 11–13 0.017

ALUSI 4 Bal. 20 8 0.3 – 12–18 0.012

Fe40Al-Grade 3 Bal. 24 – 1 – 4–7 !0.010

PM 2000 Bal. 5.5 19 0.5 0.5 40 %0.010

a Sulfur was determined on mechanically alloyed powders by combustion (IR detection). Data below 50 ppm must be considered semi-quantitative.

200 400 600 800 1000 1200

10 20 30 40

ALUSI 3 ALUSI 4 ALUSI 1

CTE (10-6/˚C)

Temperature ˚C

Alumina PM 2000

ALUSI 2

Fe 40Al-Grade 3

Fig. 2. Coefficient of Thermal Expansion as a function of temperature for intermetallic alloys, PM 2000[31]and polycrystalline alumina[32].

1 10 100

1E-5 1E-4 1E-3

∆m/A [g/cm2]

time [h]

Fig. 1. Comparison of the mass gain measured continuously (line) in a thermobalance and discontinuously (symbols) in a conventional furnace.

The values obtained by the two methods are in good agreement.

(4)

in the rate exponent can be explained by its sensitivity to the mass gain data at the earlier oxidation stages, where mass gain values are very sensitive to the determination of the initial mass because of the Archimedes effect during heating and turbulences. At 900 8C, mass gain of Fe40Al-Grade 3 can still be described as parabolic. The ALUSI alloys, however, follow a non-parabolic growth law with a rate exponent of approximately nz0.4, whereas the oxidation of PM 2000 follows a cubic rate law.

3.3. Characterisation of oxide scales

Surface examination of the selected ALUSI specimens oxidised at 1100 8C reveals the formation of a homogeneous

and even scale with signs of spallation, as shown inFig. 4a for the ALUSI 4 after 3 h exposure. The remaining scale seems to be adherent to the substrate. The absence of reoxidation at the spalled zones, confirmed by the presence of zones of white contrast on backscattered electronic images at higher magnifications, indicates that spallation occurred during the cooling stage. Quantitative metallo- graphic techniques applied to assess the area fraction of the spalled zones on SEI images of ALUSI specimens oxidised for 3 h area reveals about 6% of bare metal. Since flakes of the spalled scale often remain adhered to the surface, an overestimation of the total area of the spalled zones is expected. At the non-spalled zones a network of ridges delimiting cells with an average size of about 1–2 mm in diameter is clearly visible at higher magnifications (see Fig. 4b). The ridge-like cell morphology of the scale is indicative of a-Al₂O_3. The extent of spalled zones increased with increasing exposure time.

A closer examination of the spalled zones on the ALUSI alloys oxidised at 1100 8C revealed the presence of interfacial voids after very short exposures, Fig. 4b. The average size of the voids is about 3–4 mm. With progressing oxidation, the voids become interconnected yielding to larger sizes. In the void interior coarse yttria particles are usually observed. SEM examination of specimens oxidised at 1050 8C,Fig. 5, revealed a heterogeneous distribution of voids of smaller size. From its distribution on the sample surface it is deduced that void formation is linked to the boundaries of the original powder particles where coarse particles act as preferential nucleation sites.

At 900 8C, the alumina scale developed on the ALUSI alloys was more distinctive and contained platelet-like oxide, typical of q-Al₂O₃ scales, the length of which increased with increasing exposure. Fig. 6a illustrates the surface morphology for the ALUSI 4 alloy after 120 h of oxidation. Scale remains adherent without signs of spallation except at the sharp edges where large cavities beneath the spalled scale are also observed,Fig. 6b.

The scale pattern observed at 1100 8C is similar to that developed on the Fe40Al-Grade 3, except that void formation at the scale/metal interface was not observed, Fig. 7a. The most significant difference with respect to the ALUSI alloys during oxidation at 900 8C is that a flat and even scale is formed, without signs of metastable alumina (Fig. 7b). For PM 2000, there is some variation in the morphology of the alumina scale with small oxide nodules enriched in Fe, Cr, Ti or Y at the outermost part of the scale.

The distinctive feature for this alloy is the absence of scale spallation at the investigated temperatures and times. A more detailed microstructural characterisation of the oxidation products at these temperatures has been previously reported for Fe40Al-Grade 3 [10,11,25] and PM 2000[29].

To further characterise the oxide microstructure, XRD was used to determine the scale phase structures. This confirms that the ALUSI alloys develops an exclusively

0 20 40 60 80 100 120 140

0 4x10^-4 8x10^-4

Fe40Al-Grade 3

ALUSI 3 ALUSI 4

ALUSI 2 Mass gain (g cm-2)

Time (h)

ALUSI 1 PM 2000

0 20 40 60 80 100 120 140

0 1x10^-4 2x10^-4 3x10^-4

ALUSI 3

ALUSI 2

ALUSI 4

Fe40Al-Grade 3 Mass gain (gcm-2)

Time (h)

ALUSI 1

PM 2000 (b)

(a)

Fig. 3. Mass gain variation as a function of exposure time during oxidation in air at (a) 1100 8C and (b) 900 8C.

Table 2

Oxidation rate constants of investigated alloys at 1100 and 9008C

Alloy 1100 8C 900 8C

n k (g cm^K2) s^Kn n k (g cm^K2) s^Kn

ALUSI 1 0.54 5.55!10^K9 0.43 1.11!10^K8

ALUSI 2 0.45 5.55!10^K9 0.37 1.66!10^K8

ALUSI 3 0.50 2.77!10^K9 0.35 1.66!10^K8

ALUSI 4 0.43 8.33!10^K9 0.37 1.38!10^K8

Fe40Al-Grade 3 0.47 1.39!10^K8 0.53 2.77!10^K9

PM 2000 0.35 4.17!10^K8 0.34 1.11!10^K8

(5)

a-alumina layer at 1100 8C, whereas at 900 8C several peaks associated with metastable alumina, likely q-type, were observed.

To reveal the grain size pattern of the scale, a set of notched ALUSI specimens oxidised at 900 and 1100 8C for 120 h were deformed at ambient temperature after immersion for various minutes in liquid nitrogen. At 1100 8C, the scales are about 2 mm thick in ALUSI 3 and have well defined columnar microstructures, Fig. 8a, whereas specimens oxidised at 900 8C,Fig. 8b, show a two-layer scale with the outer part consisting of blades of the untransformed metastable alumina phase. Cross-sectional examination of polished specimens failed to reveal pores within the alloy, whether on the grain interior or in areas immediately beneath the scale-alloy interface.

Qualitative depth profiles of major elements in the oxide scale were performed after oxidation at 1100 8C. As an example,Fig. 9shows the variation of Fe, Cr, Al and O for specimens of ALUSI 4 oxidised for 12 h,Fig. 9a, and 120 h, Fig. 9b. The scale consisted of alumina with a Cr peak close to the oxide–gas interface followed by a slight maximum in the Fe concentration. After the longest exposure the Cr peak disappears leaving a single peak of Fe similar to that for the shorter exposure.

3.4. Two stage-oxidation experiments

To understand the growth mechanism of the alumina scale at the highest temperature, two-stage sequential oxidation and subsequent SNMS analysis was performed with the ALUSI 4. The samples were first oxidised in¹⁶O for 3 h, then the oxidant atmosphere was evacuated (2 min) without cooling the samples and replaced by¹⁸O (isotopic purity of 91%). After a period of 9 h the¹⁸O was pumped out, then the specimen was pulled out of the furnace. The SNMS profiles of¹⁶O and¹⁸O are presented inFig. 10. The oxygen isotope profiles clearly indicate the presence of two major peaks of¹⁸O: one peak is located at the oxide surface and the second main ¹⁸O peak is situated within the pre- existing oxide scale close to the metal-oxide interface, which denotes predominant inwards oxide growth (anion diffusion) with progressing oxidation.

4. Discussion

Oxidation experiments have shown that ALUSI alloys are able to form an exclusively alumina scale, disregarding the outermost part of the scale that contains limited amounts of Cr and Fe. Formation of a nearly pure alumina scale is expected since the aluminium content is well in excess of the critical levels (14 at.%) for alumina scale formation at 800 8C[33]. With respect to the temperature dependence of oxidation, it can be seen that mass gain at 900 8C is similar to that at 1100 8C. For the longer exposures, differences in mass gain between both temperatures for the ALUSI alloys are very small. In the case of PM 2000 and Fe40Al-Grade 3 alloys, however, the higher the temperature the higher the mass gains. The different behaviour for the ALUSI alloys at 900 8C is related to the formation of metastable q-alumina phase, which is known to grow faster than a-alumina[8,12, 27,34–36]. The transformation into a-phase is an irrevers- ible phenomenon that depends on exposure time and temperature[37]. At higher temperature the transformation process is faster. For oxidation temperatures below 1000 8C, the transformation rate is slow and metastable aluminas are thus usually observed in most alumina forming alloys.

Fe40Al-Grade 3 forms a nearly exclusively a-alumina scale at 900 8C, even after very short exposures, which is the main reason for its lower oxidation kinetics. The better oxidation

Fig. 4. Surface SEM examination of ALUSI alloys after oxidation (a) ALUSI 4 1100 8C/3 h and (b) detail of a void.

Fig. 5. SEM micrograph of the interfacial voids formed on the ALUSI 2 (1050 8C/24 h).

(6)

resistance results from a favoured nucleation of a-alumina due to the presence of the dispersoids. The fact that the ALUSI alloys, having a grain size and dispersoid distribution similar to that for the Cr-free intermetallic, exhibit a massive formation of metastable alumina at 900 8C highlights the role of Cr on the alumina growth mechanism and oxidation kinetics, as will be discussed later.

At 900 8C, the ALUSI alloys showed a platelet-shaped structure at the top of the scale associated with the outward scale growth, whereas at 1100 8C the scale was flat and even. The results of the two-stage sequential oxidation experiments (Fig. 10) indicate a predominant inward diffusion of oxygen with some outward diffusion of cations.

The small peaks of Fe and Cr at the outer part of the scale after the shortest exposure,Fig. 9a, and of Fe after oxidation for 120 h, Fig. 9b, indicate that these elements are either remnants from the very first oxidation stage or have diffused through the scale during oxidation. The absence of Cr for the specimen with the longer exposure can result from Cr oxide evaporation [38], which may occur above about 850 8C. The rough scale–gas interface observed during SEM examination, Fig. 8a, provides some evidence of such evaporation. Since magnitude of the Fe peak seems to be similar after 3 and 120 h exposure, we conclude that cation transport plays a more significant role during the earlier stages of oxidation. The fine surface ridges observed at the gas interface,Fig. 4b, are indicative of partial Al transport by outward diffusion, as are the platelets-like oxide developed at 900 8C[6,39].

The scale formed at 1100 8C on ALUSI alloys spalled at various zones. Scale spallation is also observed for the Fe40Al-Grade 3 alloy, on both thick [10,11] and thin

Fig. 6. SEM micrograph of the ALUSI 4 after oxidation at 900 8C for 120 h showing (a) the platelet-like oxide of q-Al2O3and (b) oxidation induced voids formed at the specimen edges.

Fig. 7. SEM surface morphology of Fe40Al-Grade 3 intermetallic alloy after oxidation at (a) 1100 8C/200 h and (b) 900 8C for 120 h.

(7)

specimens [24], but not for the PM 2000. The steady increase of the mass gains with increasing time denotes that scale spallation on intermetallics does not occur during high temperature exposure but during cooling. This observation is supported by the fact that reoxidation is not observed at the spalled zones. Scale spallation results[8,10] from the considerable misfit of the thermal expansion coefficients (TEC) between the substrate and the alumina,Fig. 2, which yields a very high level of compressive residual stresses in the oxide. These residual thermal stresses can be estimated using the equation[40]

s_TEZ ðE0=ð1 Kn0ÞÞDa DT (2)

where E₀and n₀are the Young’s modulus and the Poisson’s ratio, respectively, of the oxide, Da is the thermal expansion coefficient difference, and DT, the temperature change. A crude estimation of the residual stresses can be obtained using properties for bulk alumina (Young’s modulus of about 360 GPa [40] and Poisson’s ratio of about 0.27).

Based on the thermal expansion coefficients of Fig. 2, compressive residual stresses in the range of 11–12 GPa for the ALUSI alloys, and 8 and 4 GPa for Fe40Al-Grade 3 and PM 2000 alloys, respectively, are obtained. Values for

the intermetallic alloys clearly exceed 5 GPa, reported as the compressive strength of a-Al2O3[41]. The implications of a large misfit in the TEC on the spallation resistance are clear. However, a direct correlation between the TEC misfit and the magnitude of spallation is not expected since various other physical and chemical properties could vary.

Obviously, the higher the thermal expansion of the intermetallic alloy, the smaller the temperature decreases for spallation. The reasons for the large TEC for ALUSI alloys and their abnormal behaviour between 400 and 450 8C are unexplained and need further investigation.

In addition to the elevated residual stresses, subscale void formation on the ALUSI alloys after oxidation at 1100 8C diminishes the contact areas between the oxide and the substrate, deteriorating the interface integrity and reducing the capability of the scale to withstand the residual stresses.

The absence of voids uniformly distributed in the sample cross-section easily discards the thermal induced porosity that is often observed in powder metallurgy alloys after high temperature exposure due to the assembly of entrapped gas from the material production.

Most accepted mechanisms accounting for void formation induced by oxidation, particularly for Ni-base

Fig. 8. SEM micrographs of the fracture cross-sections of the scales formed on ALUSI 3 after oxidation at (a)1100 8C/120 h and (b) 900 8C/120 h. (Me denotes metal substrate).

(8)

superalloys[42], are based on condensation of vacancies at the scale/metal interface or in the grain interior. Flux of vacancies is assumed to come either from the outer surface (vacancy injection) or from the metal interior (Kirkendall effect). The former mechanism considers that new oxide is

formed at the gas/scale interface, thus vacancies are injected to counterbalance the outward cation diffusion[43]. In the present case, however, results of the two-stage oxidation experiment performed at 1100 8C provide evidence that, at least after 3 h exposure, a predominant inwards scale growth occurs, i.e. anion diffusion, consistent with previous investigations on the growth mechanism of a-alumina scales developed on a Fe-25 wt% Al[44]. Void formation is also not observed on specimens oxidised at 900 8C, where q-alumina forms growing predominantly by outward aluminium transport [36]. The absence of voids at this temperature suggests that injected vacancies are annihilated at the large number of grain boundaries and dispersoid–

metal interfaces, which are known to be very effective sinks for vacancies [45]. The large cavities observed at the specimen’s edges, Fig. 6b, are explained by the specimen geometry that imposes a high volume to surface ratio, inducing strong outward aluminium transport[46].

The mechanism based on the Kirkendall effect has been proved for Ni₃Al and NiAl[34,47,48]. The mechanism is related to the outwards flux of vacancies from the interior of the substrate resulting from the relative diffusion rate of Fe to Al into the substrate. Therefore, compositional gradients beneath the scale should be present, at least before the reduction in the oxidation rate associated with the parabolic growth law, which allows equilibrium in the chemical composition [42]. Compositional depth profiles performed by SEM and SIMS after about 3 and 120 h of oxidation, however, failed to reveal gradients in the chemical composition of ALUSI alloys. It should be remarked that although several researchers have noted that constitutional vacancies form in aluminides[see recent review in 49], but FeAl appears to be the only aluminide to exhibit a high concentration of thermal vacancies. Hehenkamp et al.[50]

have proved that in the range of 30–38 at.% Al all diffusivities (vacancy mobility) increase with decreasing Al content, at least in the range of 800–1100 8C. Therefore, the loss of Al during oxidation will allow a fast chemical equilibrium beneath the scale and, consequently, Kirkendall effect does not need to be invoked to explain void formation in Fe-40Al type iron aluminides.

An interesting feature is that voids are observed at the subscale developed on the ALUSI 2 alloy, but not on the Cr-free Fe40Al-Grade 3 intermetallic alloy having the same aluminium content. Moreover, both alloys develop a a-alumina scale growing by a similar mechanism. There- fore, subscale void formation must be influenced by other unidentified factors. The enthalpy for vacancy formation depends on the Al content, with smaller values and larger thermal vacancy concentrations found near FeAl stoichi- ometry[51]. Therefore, the fast initial oxidation rate yields a significant Al consumption that causes a local super- saturation of vacancies. The high migration energy relative to the formation energy that characterises iron aluminides [52] would account for the condensation in the substrate of the Cr-free Fe40Al-Grade 3 intermetallic alloy, where

0.0 0.5 1.0 1.5 2.0 2.5 3.0

0 3x10⁴ 5x10⁴ 8x10⁴ 1x10⁵ 1x10⁵ 2x10⁵

Counts s-1

Depth µm (a)

oxide Metal

0 1 2 3

0 1x10⁴ 2x10⁴ 3x10⁴ 4x10⁴ 5x10⁴ 6x10⁴

Counts s-1

Depth µm

Fe

Cr

O Al (b)

oxide Metal

Fig. 9. SIMS depth profiles of oxide scales formed on ALUSI 4 after oxidation at 1100 8C for (a) 12 h and (b) 120 h.

0.0 0.5 1.0 1.5

0 2000 4000 6000 8000

Counts s-1

Depth (µm) O 16

O 18

Fig. 10. Oxygen isotope profiles after two-stage oxidation experiments for ALUSI 4 at 1100 8C (3 h¹⁶OC9 h¹⁸O).

(9)

the high densities of grain boundaries and dispersoid act as effective sinks for vacancies[45], rather than condensation at the scale-metal interface. Accordingly, the ALUSI alloys, having an oxidation rate smaller than that of Fe40Al-Grade 3, will experience a lesser consumption of aluminium and, therefore, a smaller void area fraction, which is contrary to the experimental observations. The high content of Cr in the ALUSI alloys suggests the importance of this element on the subscale void formation, but it is not clear why.

One possibility is that Cr influences the transport properties in the substrate. Previous investigations[52,53]

report that Cr addition (2 at.%) to cast FeAl causes a decrease in the energy for vacancy migration. As the alloy composition (Al and Cr) changes during the earlier stages of oxidation two factors may affect the void formation rate: the speed at which vacancies migrate and the number of sinks.

Assuming that voids form by the condensation of vacancies, and that the number and separation of sinks in Fe40Al- Grade 3 and ALUSI alloys are similar, the rate of void formation will be determined by vacancy formation and mobility. No definitive conclusions about the role of Cr on the vacancy behaviour can be drawn at the present time.

Preliminary results have shown that heat treatment of ALUSI 4 at 1100 8C for 10 min, followed by water quenching, yields a hardness increase from 450 to 511 Hv1, which might be associated with vacancy hard- ening. Further ageing at 400 8C for 140 h caused a softening to the initial value of 450 Hv1, which could be related to a process of vacancy annihilation. The fact that the same thermal cycle does not yield to any significant variation in hardness of the Cr-free Fe40Al-Grade 3 intermetallic provides evidence of the role of Cr. Further experiments are in progress to determine the role of this element.

With respect to the detailed mechanism for void nucleation, a recent analysis [14] of interfacial void formation on cast FeAl oxidised at 1000 8C has shown that nucleation and growth occurred very early, even before the sample reached the test temperature. In the present work, voids are observed on specimens oxidised at 1050 and 1100 8C. Void nucleation on the ALUSI alloys may be favoured by the coarse particles decorating the mechanically alloyed particles, as deduced from the void pattern observed at 1050 8C, Fig. 5. With progressing exposure, void enlargement may be assisted by the tensile stresses beneath the scale counterbalancing the compressive stresses associated to the scale growth[54]. Growth of voids, is also well possible by surface diffusion or evaporation of Al from the metal substrate [48,55], with the former being more important form smaller voids [14]. Additional growth during cooling may occur since vacancy concentration in equilibrium is known to decrease with decreasing temperature. Acceleration of the void growth by fast sulphur segregation to the void surface [56–58] may be ruled out since the uniform dispersion of fine yttria particles should be an effective method to trap any sulphur[57,59,60]. Precise measurements determining volume fraction of voids as

a function of the oxidation temperature, exposure time, and cooling conditions should be desirable.

The remaining issue concerns the effect of alloying elements on the oxidation behaviour of the intermetallic alloys. From the comparative analysis of mass gains for the ALUSI 1 and 2 it follows that an increase of the Al content from 20 to 25 wt% causes a significant decrease in the oxidation rate at 1100 8C. At 900 8C, however, differences are very small, within the scatter of results. In the literature, the rates of oxidation of iron aluminides have been reported to decrease with increase in the Al content of the intermetallic [61]and also for the Cr-alloyed intermetallics[7].

With regards to the effect of chromium, it is well known that additions to alloys with low Al content, like PM 2000, play a beneficial role by favouring the selective oxidation of Al allowing the development of an exclusively alumina scale. For the ALUSI alloys, where the Al content is well above the 14 at.% required to form an exclusively alumina scale [33], additions of Cr have shown different effects on isothermal oxidation kinetics. At 1100 8C Cr decreases considerably the oxidation rate, as seen when comparing results for the ALUSI 2 (12% Cr) and the Cr-free intermetallic Fe40Al-Grade 3 alloy. A decrease in the Cr content from 12 (ALUSI 1) to 8 wt% (ALUSI 3) seems not to have a significant effect. At 900 8C, however, the intermetallics containing Cr exhibit higher oxidation rates, clearly associated with the delayed transformation from q-alumina to of a-alumina.

The beneficial role of Cr at 1100 8C agrees with previous results on oxidation at the same temperature of FeAl coatings with different Cr contents [23]. Overall, Cr additions (6.5 wt%) yielded a significant decrease in the oxidation rate. Nuclei of a-Cr₂O₃are believed to form at the initial stages of oxidation[34]. Their presence might favour nucleation of a-alumina, accelerating the transformation of metastable alumina. The larger number of nuclei resulted in a fine ridge-type alumina scale (less than 1–2 mm). This explanation is, however, not valid for the present work because, although the ridge structure is also fine (about 3 mm), the Cr-containing specimens oxidised at 900 8C showed retarded transformation of metastable alumina.

Therefore, it appears that nucleation of a-alumina rather than transformation from metastable alumina is the mechanism of growth of stable alumina at high temperature.

The detrimental role of Cr on the oxidation behaviour at 900 8C of ALUSI alloys is consistent with that reported for the oxidation of Fe₃Al [7,62]. In this case, however, transient oxidation behaviour is different and oxides of all constituents such as FeO are formed during the fast-growth transient period [12]. Why Cr additions delay the transformation of metastable alumina in the ALUSI alloys remains uncertain since this element has shown to accelerate it in other alumina forming intermetallics, like NiAl [34]. As mentioned previously, it is possible that Cr addition to the ODS iron aluminides modifies the transport properties in the substrate.

(10)

There has been much discussion about the role of reactive elements in the oxidation of ODS FeCrAl alloys or iron aluminides either by doping elements or by dispersions of oxides[6,63]. In general, small additions of yttria to alumina forming alloys, typically in concentrations of 0.5 mass%, greatly increase the scale adherence, but higher contents slightly accelerate oxidation[64]. The current results for the ALUSI alloys suggest that RE additions to iron aluminides, of either 1 or 0.3 wt%, are not as effective in improving scale adhesion compared with ODS FeCrAl-based alumina formers.

With regards to the yttria content of these alloys differences in the oxidation behaviour are not significant and, therefore, the optimum dopant level will be determined on the basis of the minimum oxide volume fraction required for strengthening.

5. Conclusions

† Oxidation experiments have shown that, consistent with the high Al content, the ALUSI alloys are able to form an exclusively alumina scale, apart from disregarding the outermost part that contains small amounts of Cr and Fe.

Compared with Fe40Al-Grade 3 intermetallic and PM 2000, a ferritic alloy, the new ODS FeAlCr intermetallic alloys show lower oxidation rates at 1100 8C. At 900 8C, however, they exhibit higher oxidation kinetics, related to the massive formation of metastable alumina phases, likely q-phase.

† An increase of Al content from 20 to 25 wt% causes a significant decrease in the oxidation rate at 1100 8C. At 900 8C, however, differences are very small.

† Additions of Cr to the Al-rich intermetallics have a dual effect on the isothermal oxidation kinetics that depends on the temperature. At 1100 8C Cr decreases considerably the oxidation rate. At 900 8C, however, the intermetallics containing Cr exhibit higher oxidation rates clearly be associated with the decrease of the transformation rate to alpha-alumina from q-alumina.

† Oxide scales on the ALUSI alloys at 1100 8C suffer scale spallation during cooling.

† Yttria additions to iron aluminides, either 1 or 0.3 wt%, are not as effective in improving scale adhesion as in ODS FeCrAl-based alumina formers.

† The FeAlCr intermetallic alloys are prone to fast subscale void formation during oxidation at 1100 8C that deteriorates the scale adherence. The absence of such oxidation induced voids in the Cr-free Fe40Al- Grade 3, having the same grain size and volume fraction of dispersoid, highlights the detrimental role of Cr.

Acknowledgements

The authors would like to thank financial support to the European Commission (ALUSI Project, GROWTH-CT- 1999/00083) and Dr P. Fielitz (TU Clausthal) for support

with the SIMS measurements. Dr M.A. Mun˜oz-Morris (CENIM) is specially acknowledged for useful scientific discussions.

References

[1] Morris DG, Morris-Mun˜oz MA. Intermetallics 1999;7:1121.

[2] McKamey CG, DeVan JH, Tortorelli PF, Sikka VK. J Mater Res 1991;6:1779.

[3] Srinivasa N, Sikka VK. In: Schneibel JH, Crimp MA, editors.

Processing, properties and applications of iron aluminides.

Warrendale, PA: The Materials, Metals and Minerals Society; 1994.

p. 69.

[4] Baccino R, San Filippo D, Moret F, Lefort A, Webb G. Proceedings of the PM’94, powder metallurgy World Congress, Paris, June 1994. vol.

II. Les Ulis: Editions de Physique; 1994 p. 1239.

[5] Evans HE, Strawbridge A. Fundamental aspects of high temperature corrosion VI. In: Schores DA, Rapp RA, Hou PY, editors. Proceeding of v.96–26. Pennington, NJ: The Electrochemical Society; 1996. p. 1.

[6] Pint BA, Tortorelli PF, Wright IG. Mater Corros 1996;47:663.

[7] Tortorelli PF, DeVan JH. Compositional influences on the high temperature corrosion resistance of iron aluminides. In: Schneibel JH, Crimp MA, editors. Processing, properties and applications of iron aluminides. Warrendale: TMS; 1994. p. 257–70.

[8] Smialek JL, Doychack J, Gaydosh DJ. Oxid Met 1990;34:259.

[9] Xu CH, Gao W, Gong H. Intermetallics 2000;8:769.

[10] Montealegre MA, Gonza´lez-Carrasco JL, Morris-Munoz MA, Chao J, Morris DG. Intermetallics 2000;8(4):439.

[11] Montealegre MA, Gonza´lez-Carrasco JL. Intermetallics 2002;11(2):

169.

[12] Grabke HJ. Intermetallics 1999;7:1153.

[13] Dang Ngoc Chan C, Huvier C, Dinhut JF. Intermetallics 2001;9:817.

[14] Hou PY, Niu Y, Lienden CV. Oxid Met 2003;9:41.

[15] Liu CT, Lee EH, McKamey CG. Scripta Metall Mater 1989;23:875.

[16] McKamey CG, Liu CT. Metall Mater 1990;24:2119.

[17] Klein O, Baker I. Scripta Metall Mater 1992;27:1823.

[18] Balasubramaniam R. Scripta Mater 1996;34:127.

[19] Velon A, Yi DQ. Oxid Met 2002;57:13.

[20] Tortorelli PF, DeVan JH. Mater Sci Eng 1992;153:573.

[21] Deevi SC, Sikka VK. Intermetallics 1996;4:357.

[22] Wang F, Tang Z, Wu W. Oxid Met 1997;48:381.

[23] Liu Z, Gao W. Oxid Met 2000;54:189.

[24] Berlanga M, Gonza´lez-Carrasco JL, Montealegre MA, Mun˜oz- Morris MA. Intermetallics 2004;12:205.

[25] Pedraza F, Grosseau-Poussard JL, Dinhut JF. Intermetallics 2005;

13:27.

[26] Quadakkers WJ, Bongartz K. Mater Corros 1994;45:232.

[27] Prasanna KMN, Khanna AS, Chandra R, Quadakkers WJ. Oxid Met 1996;46(5/6):465.

[28] Lours P, Alexis J, Bernhart G. J Mater Sci Lett 1998;17(13):1089.

[29] Gonza´lez-Carrasco JL, Garcı´a-Alonso MC, Montealegre MA, Escudero ML, Chao J. Oxid Met 2001;55(3/4):209.

[30] Flores MS, Ciapetti G, Gonza´lez-Carrasco JL, Montealegre MA, Multigner M, Pagani S, et al. J Mater Sci Mater M 2004;15:559.

[31] Materials Data Sheet: ODS Superalloy PM2000, Metallwerk Plansee GmbH/Lechbruch; Feb. 1993.

[32] Touloikian YS, Kirby RK, Taylor RE, Lee TYR. Thermal expansion nonmetallic solids. Thermophysical properties of matter. vol. 13. New York: IFI/Plenum; 1970.

[33] Tomaszewicz P, Wallwork G. Rev High Temp Mater 1978;4:75.

[34] Brumm MW, Grabke HJ. Corros Sci 1992;33:1677.

[35] Gonza´lez-Carrasco JL, Pe´rez P, Adeva P, Chao J. Intermetallics 1999;

7:69.

[36] Rybicki GC, Smialek JL. Oxid Met 1989;31:275.

(11)

[37] Doychack J, Ruhle M. Oxid Met 1989;32:431.

[38] Wood GC, Stott FH. Mater Sci Tech 1987;3:519.

[39] Tolpygo VK, Clarke DR. Evidence of outward aluminum grain bounday diffusion in Y- and Zr-doped alumina scales. Proceedings of the electrochemical society 1999 pp. 204–14.

[40] Christensen RJ, Tolpygo VK, Clarke DR. Acta Mater 1997;45:1761.

[41] Do¨rre E, Hu¨brer H. Alumina. Berlin: Springer; 1984 p. 218.

[42] Gonza´lez-Carrasco JL, Guttmann V, Fattori H. Metall Mater Trans 1995;26A:915.

[43] Gibbs GB, Hales R. Corros Sci 1977;17:487.

[44] Prescott R, Mitchell DF, Graham MJ. Corros Sci 1994;50:62.

[45] Montealegre MA, Mun˜oz-Morris MA, Gonza´lez-Carrasco JL, Morris DG. Scripta Mater 2001;44:2673.

[46] Strehl G, Guttmann V, Naumenko D, Kolb-Telieps A, Borchardt G, Quadakkers WJ, et al. The influence of sample geometry on the oxidation and chemical failure of FeCrAl(RE) alloys. In: Schu¨tze M, Quadakkers WJ, Nicholls JR, editors. Lifetime modelling of high temperature corrosion processes. London: European Federation of Corrosion, IOM Communications, EFC Publication No. 34;

2001. p. 107–22.

[47] Kuenzly JD, Douglass DL. Oxid Met 1974;8:139.

[48] Brumm MW, Grabke HJ. Corros Sci 1993;34:547.

[49] Jordan JL, Deevi SC. Intermetallics 2003;11:507.

[50] Hehenkamp TH, Scholz P, Ko¨hler B, Kerl R. Defect and diffusion forum 2001;194–199:389.

[51] Chang YA, Pike LM, Liu CT, Bilbrey AR, Stone DS. Intermetallics 1993;1:107.

[52] Morris MA, Morris DG. Scripta Mater 1998;38:509.

[53] Morris MA, George O, Morris DG. Mater Sci Eng 1998;258:99.

[54] Harris JE. Acta Metall 1978;26:1033.

[55] Hindam HM, Smelzter WW. J Electrochem Soc 1980;127:1630.

[56] Grabke HJ, Wiemer D, Viefhaus H. Appl Surf Sci 1991;47:243.

[57] Pint BA. Oxid Met 1997;48:303.

[58] Hou PY, Pru¨ner K, Fairbrother DH, Roberts JG, Alexander KB.

Scripta Mater 1999;2:241.

[59] Sigler DR. Oxid Met 1993;40:555.

[60] Meier GH, Pettit FS, Smialek JL. Mater Corros 1995;46:232.

[61] Rommerskirchen I, Eltester B, Grabke HJ. Mater Corros 1996;47:

646.

[62] Babu N, Balasubramaniam R, Ghosh A. Corros Sci 2001;43:2239.

[63] Montealegre MA, Gonza´lez-Carrasco JL. Intermetallics 2003;11:169.

[64] Quadakkers WJ, Schmidt K, Grubmeier H, Wallura E. Mater High Temp 1992;10:23.