An impedance spectroscopy study of high-temperature oxidation of thermal barrier coatings

An impedance spectroscopy study of high-temperature oxidation of thermal barrier coatings

Materials Science and Engineering B97 (2003) 46 /53 www.elsevier.com/locate/mseb An impedance spectroscopy study of high-temperature oxidation of th...

388KB Sizes 5 Downloads 34 Views

Materials Science and Engineering B97 (2003) 46 /53 www.elsevier.com/locate/mseb

An impedance spectroscopy study of high-temperature oxidation of thermal barrier coatings Shenhua Song, Ping Xiao  Manchester Materials Science Centre, University of Manchester/UMIST, Grosvenor Street, Manchester M1 7HS, UK Received 25 March 2002; accepted 23 September 2002

Abstract Oxidation of air-plasma-sprayed (APS) thermal barrier coatings (TBCs) was carried out in air at 950 8C and was investigated using impedance spectroscopy coupled with scanning electron microscopy and X-ray diffraction. After oxidation for between 500 and 3000 h, a continuous alumina scale was formed at the bond coat/top coat interface in the TBCs, and was evaluated using impedance spectroscopy. The impedance spectra of the oxidised TBCs showed that there were four relaxation processes, which were attributed to the yttria-stabilised zirconia (YSZ) bulk of the TBCs, the thermally grown alumina scale, the YSZ grain boundary, and the metal electrode effect. Impedance analysis showed that the resistivity of the alumina scale increased with increasing oxidation time, demonstrating the thermally grown oxide (TGO) growth. The resistance of the YSZ grain boundaries also increased after oxidation for 2000 h, suggesting the composition change at grain boundaries after a long-term oxidation. # 2002 Elsevier Science B.V. All rights reserved. Keywords: Impedance spectroscopy; High-temperature oxidation; Thermal barrier coatings

1. Introduction In gas turbines, the maximum tolerable temperature for advanced superalloys is up to around 1000 8C. A thermal barrier coating (TBC) system is usually applied to increase the operation temperature and to protect the turbine components like blades [1 /5]. A TBC system consists typically of an oxidation-resistant metallic bond coat on a superalloy substrate and a heat-insulating ceramic top coat attached on the bond coat. An yttriastabilised zirconia (YSZ) (6 /8 wt.% Y2O3, balance% ZrO2) is usually employed as the top coat material. Normally, the bond coat alloy is an MCrAlY (M /Ni and/or Co), which protects the superalloy substrate from oxidation by forming a continuous protective oxide scale (usually a-Al2O3) at the metal /ceramic interface. As the thermally grown oxide (TGO) grows at the YSZ/bond coat interface due to oxidation of the bond coat, internal stresses are built up at the top coat/

 Corresponding author. Tel.: /44-161-200-5941; fax: /44-161200-3586 E-mail address: [email protected] (P. Xiao).

bond coat interface since the oxidation is accompanied with volumetric expansion. The thermal mismatch among the bond coat, TGO, and YSZ further increases the stress level during thermal cycling of TBCs. These stresses can cause cracking in the YSZ and lead to failure of TBCs [2,6 /8]. It has been found that the growth of the TGO layer is an important factor in the degradation of TBCs [2,9]. However, it is difficult to examine the degradation of TBCs non-destructively, which is essential for predicting the lifetime and failure of TBCs on engine components. Acoustic emission has been used for the examination of cracking in TBCs [10], where acoustic emissions are pulses of elastic strain energy released spontaneously during deformation. Infrared thermography has been used [11,12] for evaluating the delamination of TBCs. In this case, a pulse of heat is applied to one side of a specimen and then the spatial distribution of the heat flux on the opposite surface, which is dependent on the homogeneity of the specimen measured. Piezospectroscopy has been developed to examine the stress level and failure in TBCs non-destructively, based on the analysis of the shape of the luminescence spectra [13]. Although these non-destructive techniques can be used to char-

0921-5107/02/$ - see front matter # 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 1 - 5 1 0 7 ( 0 2 ) 0 0 3 9 7 - 5

S. Song, P. Xiao / Materials Science and Engineering B97 (2003) 46 /53

acterise the formation of defects in TBCs when cracking occurs, these techniques cannot be used to determine the extent of degradation and predict the failure in TBCs. None of these techniques can be used to reflect the composition change and growth of the TGO, even though the growth of the TGO is a critical factor in controlling the degradation of TBCs [14,15]. Impedance spectroscopy is a non-destructive technique, which has been used extensively to determine the electrical properties of materials [14 /17]. Using impedance spectroscopy, Wang et al. [18] and Ali et al. [19] have studied the TGO growth in TBCs after oxidation at 1100 8C. Impedance measurements were used to examine the formation of the continuous alumina layer and mixed oxide layer (TGO) in TBCs due to the oxidation. Two relaxation processes appeared in the impedance spectra corresponding to the alumina and YSZ top coat when a discontinuous alumina layer was formed between the top and bond coats. When a continuous alumina layer and a mixed oxide layer were formed between the top and the bond coats, three relaxation processes were found in the impedance spectra, corresponding to the alumina layer, the mixed oxide layer, and the YSZ layer. Ali et al. [19] also investigated the degradation of air-plasma-sprayed (APS) TBCs at 1150 8C. The spallation of the YSZ top coat was found to be caused by the microstructural development of the TGO. The change in the resistivity of the TGO was found to be associated with the change in its microstructure and microchemistry. The resistivity of the TGO decreased rapidly during oxidation from 10 to 1000 h, which corresponded to an increase in porosity within the TGO and a compositional change of the TGO from alumina to a mixture of chromia and (Ni,Co)(Cr,Al)2O4 spinel. It is worth mentioning that there were only two relaxation processes emerging in the impedance spectra, representing the YSZ and TGO layers. The alumina could not be separated from the mixed oxide in the impedance spectra because the impure alumina is close to the mixed oxide in electrical properties. Recently, Ali and Xiao [20] have also used impedance measurements to monitor crack propagation in TBCs during thermal cycling. Most of the previous studies by our team were focused on the oxidation of TBCs at temperatures above 1000 8C, which is higher than the operation temperature for superalloy components in engines. The aim of the present work is to use impedance spectroscopy to investigate oxidation of the TBCs at a lower temperature of 950 8C, which is in the temperature range at which superalloy components in turbine engines usually operate. Impedance measurements of TBCs in this work have shown four relaxation processes, where three of them correspond to YSZ grains, YSZ grain boundaries, the TGO, and the metal electrode effect. The electrical resistance of the TGO clearly increased with increasing

47

oxidation time whereas the capacitance decreased. The electrical properties of the YSZ grain boundaries also showed a clear change after the specimens were oxidised for more than 2000 h, indicating some changes occurring in the YSZ grain boundaries after a long-term oxidation.

2. Experimental Specimens with a coating area of 20 /20 mm2 were prepared by APS by Sermatch Ltd., Lincoln, UK, with both a bond coat ( /150 mm thick) and a top coat (/ 250 mm thick) on Haynes-230 superalloy plates (3 mm thick). The chemical composition of the superalloy (wt.%) was 22Cr, 14W, 3Fe, 2Mo, 0.02La, 0.015B, and 0.1C with nickel as the balance. Chemical composition of the bond coat (wt.%) is 32Ni, 38.5Co, 21Cr, 8Al, and 0.5Y. The YSZ contained 8 wt.% Y2O3 and 92 wt.% ZrO2. The specimens were heated in air to 950 8C with a rate of 5 K min 1 and oxidised for 100, 500, 1000, 2000, and 3000 h, respectively. Three to five specimens were used for impedance measurements for each oxidation condition. For impedance measurements, the metal side of the oxidised specimen was mechanically polished to remove the oxide layer and acted as one electrode. The TBC side was coated with a silver paint, which served as the other electrode. In order to consolidate the silver paint and enhance its adhesion to the specimen surface, the paint was cured at 400 8C for 30 min. Impedance measurements were conducted at 400 8C using a Solarton SI 1255 HF frequency response analyser coupled with a Solarton 1296 Dielectric Interface, which is computercontrolled. Spectra analysis was performed using Zview impedance analysis software (Scribner Associates, Inc., Southern Pines, NC) to extract the electrical and dielectric properties of TBCs. In the measurements, alternating current (AC) amplitude of 100 mV was employed and the AC frequency was in the range 1 to 1 /107 Hz. The measured results were represented by the mean values of data points obtained together with the standard deviation. Microstructures of the specimens were examined with the use of a J840 scanning electron microscope equipped with an LINKS energy dispersive X-ray microanalyser (EDX).

3. Results and discussion 3.1. Microstructure and oxidation kinetics Fig. 1 shows the cross-sections of the as-received and oxidised specimens. There is no detectable TGO present between the bond and top coats in the as-received

48

S. Song, P. Xiao / Materials Science and Engineering B97 (2003) 46 /53

Fig. 1. Scanning electron micrographs for the cross-sections in the TBCs as-received (a) and oxidised in air at 950 8C, for (b) 100 h, (c) 500 h, (d) 1000 h, and (e) 3000 h.

specimen. In the 100 h oxidised specimen, the TGO layer was formed but is not continuous. After oxidation for 500 h, a continuous TGO layer was formed at the bond coat/top coat interface. The SEM image of the TGO layer exhibits no apparent change even after 3000 h oxidation as compared with that after 500 h oxidation. With provision for the limited spatial resolution of EDX in SEM, it is difficult to conduct EDX microanalysis on the 500 h oxidised specimen because of an insufficient thickness of the scale. Energy-dispersive X-ray microanalysis has therefore been performed only on the centre of the TGOs for the specimens oxidised for 1000 and 3000 h. The results show that the TGOs contain 100%

alumina. The EDX microanalyses have also been carried out in regions of the TGOs 0.5 mm away from the scale/ YSZ interface and these indicate that a small amount of Ni and Cr ( B/5 wt.%) was present for the 3000 h oxidised specimen. However, it is difficult to ensure that it is not from the superalloy substrate. It appears that there is no new phase being formed in the areas near the YSZ. The oxidation kinetics is represented in Fig. 2, where the oxide layer thickness is plotted as a function of the square root of oxidation time. Clearly, there is a linear relationship between the oxide thickness and the square root of oxidation time, indicating that the oxidation

S. Song, P. Xiao / Materials Science and Engineering B97 (2003) 46 /53

Fig. 2. Oxidation kinetics of the TBCs at 950 8C.

follows a parabolic law. However, there are two segments separated at 1000 h oxidation with different slopes and the line slope for oxidation times shorter than 1000 h is greater than that for oxidation longer than 1000 h. This demonstrates that the reaction rate parameter is larger during the initial oxidation stage than it is for oxidation times longer than 1000 h as the slope is equal to the square root of the reaction rate parameter [21]. There have been many studies on the degradation of TBC systems at temperatures between 1000 and 1150 8C [2,18,22 /27]. As shown in these studies, after long-term oxidation the TGOs are comprised of a-Al2O3 and spinel while the a-alumina is a major product with the spinel between the YSZ and the alumina. However, our recent work [28] indicates that after long-term oxidation at 1150 8C the a-alumina would disappear. The TGOs consist mainly of a-Cr2O3/(Ni,Co)(Cr,Al)2O4 spinel mixed oxides so that the protective capability of the TGOs is substantially reduced. Experimental results also demonstrate that the content of Al in the TGOs decreases with increasing oxidation time and meanwhile the content of Cr and Ni in the TGOs increases. In the present work, since the oxidation temperature is low when compared with the others mentioned above, the oxidation kinetics is slow. As a consequence, there are no mixed oxides present in the TGOs although they may contain some Ni and Cr in regions near the YSZ after 3000 h oxidation. Concerning the oxidation kinetics (see Fig. 2), as described above, the TGO growth appears to follow the parabolic law. However, there are two reaction rate parameters, one of which corresponds to oxidation times shorter than 1000 h and the other to oxidation times longer than 1000 h. The oxidation kinetics may be described by [29] x2  kc t;

(1)

where x is the thickness of the oxide layer, t the oxidation time, and kc the reaction rate parameter,

49

which is related to the diffusion rate of reacting elements. In the light of the experimental data shown in Fig. 2, kc can be acquired as approximately 1/1014 cm2 s 1 for oxidation times shorter than 1000 h and 2/10 15 cm2 s 1 for oxidation times longer than 1000 h. Due to the fact that the central area of TGOs are composed of pure alumina with no impurity detected by EDX microanalysis, the oxidation rate should be controlled by diffusion of reacting elements in the alumina. Allowing for the two different reaction rate parameters, one may infer that in the earlier stage of oxidation, the alumina is less dense, leading to a larger reaction rate parameter. Pint et al. [30] studied the oxidation of PtAl bond coats at 1150 8C, where the alumina layer was formed from oxidation of the bond coat. They found that the reaction rate parameter was about 1/10 12 cm2 s 1. Our recent work on TBCs with the same bond coat as in the present work also indicated that before the alumina layer was completely replaced by mixed oxides, the reaction rate parameter was of the same order of magnitude as that acquired by Pint et al. [30]. Since the oxidation temperature used in this work is 200 8C lower than 1150 8C, it is difficult to compare the reaction rate parameters obtained in the present case with those obtained from the previous studies described above. Nevertheless that our values are much lower than those measured at 1150 8C should be reasonable from the standpoint of oxidation kinetics.

3.2. Electrical properties and their relation to microstructural features The electrical properties of different specimens have been measured using impedance measurements. There are different presentations of impedance data, in which Nyquist and Bode plots are most frequently used. In a Nyquist plot, the impedance is represented by a real part Z ? and an imaginary part Z ?? with the formula Z (v )/ Z ?/jZ ??, where v is the angular frequency and j /â/ 1. In a Bode plot, the modulus of the impedance and the phase angle are both plotted as a function of frequency. For a simple resistor /capacitor (R /C ) circuit, the Nyquist plot is characterised by a single semicircle. Usually, the Nyquist plot is used to determine the major parameters, such as resistance and capacitance corresponding to an electrochemical system by fitting the measured spectra according to an equivalent circuit, which corresponds to a physical model of the system. As described in Refs. [15,31,32], for an oxide scale system, the measured capacitance response is often not ideal, i.e. not a pure capacitor. This deviation can be modelled by the use of a constant phase element (CPE) instead of an

S. Song, P. Xiao / Materials Science and Engineering B97 (2003) 46 /53

50

ideal capacitance element in the equivalent circuit. The impedance of a CPE, ZCPE, is given by [15,31,32] ZCPE 

1 ; A(jv)n

(2)

where A is a parameter independent of frequency. When the exponential factor n /1, the CPE functions as an ideal capacitor and A is therefore equal to the capacitance C . In most cases, n is less than 1. It should be noted that n is a mathematical factor without physical meaning (i.e. it acts as a fitting parameter), but it represents an effective approach to complicated relations among several elements (R , C , etc.) [32]. In general, a CPE is associated with chemical inhomogeneity and geometrical non-uniformity, which cause a frequency dispersion [32]. In the case of no ideal capacitive response, the value of A cannot be used to represent the capacitance of the system. Here we adopt an equivalent capacitance C , which may be acquired by [33] C R(1n)=n A1=n ;

(3)

where R is the resistance. The phase angles are represented in Fig. 3 as a function of frequency. Clearly, there are three relaxation processes for the as-sprayed and 100 h oxidised specimens while there are four for the specimens oxidised for 1000 h or more. For all the specimens, the highest- or second highest-frequency relaxation has a similar relaxation frequency. It has been confirmed [34,35] that the lowest-frequency relaxation stems from the electrode effect rather than the material being investigated. For the as-received specimen, since there is no TGO layer between the YSZ top coat and the metallic bond coat (see Fig. 1), the two relaxation processes except for the electrode effect should be responses from the grains and grain boundaries of the YSZ. This phenomenon has also been confirmed in a study regarding the densification of

Fig. 3. Phase angle as a function of frequency for the specimens asreceived and oxidised at 950 8C for different times.

the YSZ [36]. In the two relaxation processes, the higher-frequency one is related to the grains and the other to the grain boundaries. Accordingly, the grain and grain boundaries in the YSZ are primarily responsible for the highest- and second highest-frequency relaxations. According to the microstructure (see Fig. 1), the third relaxation process from the right for the specimens oxidised for 500 h or longer should originate from the continuous thermally grown alumina layer. Therefore, the presence of the discontinuous TGO should be the reason for not having this relaxation process in the 100 h oxidised specimen. However, if the chemical composition in the alumina changes or a mixed oxide layer (which comprises alumina, chromia, spinel, and nickel oxide) is formed at the metal /YSZ interface, making the TGO less insulative than pure alumina, the fourth relaxation process may overlap with the YSZ grain boundary response. In this case, the second highest-frequency relaxation is a combined effect from both YSZ grain boundaries and TGO [19,34,35]. Typical Nyquist plots for the as-sprayed and oxidised specimens for different times are represented in Fig. 4. With the exception of the electrode effect, a tail shown in the low-frequency range, two semicircles are present for the as-received and 100 h oxidised specimens and three for the specimens oxidised for 500 h or longer. As in the phase angle /frequency diagram (see Fig. 3), the two semicircles (or arcs) in the former are mainly caused by the grain and grain boundaries of the YSZ and the

Fig. 4. Typical Nyquist plots for the specimens as-received and oxidised at 950 8C for different times, measured at 400 8C.

S. Song, P. Xiao / Materials Science and Engineering B97 (2003) 46 /53

third semicircle for the latter by the continuous TGO layer. To obtain the electrical properties of different layers, it is necessary to establish an equivalent circuit model to fit the measured impedance spectra. Two semicircles in the spectra of the as-received and 100 h oxidised specimens can be simulated based on a model of two R /C components with a series connection, which correspond to YSZ grains and YSZ grain boundaries. Three semicircles in the spectra for the specimens oxidised for 500 h or more can be simulated based on a model of three R /C components, corresponding to YSZ grains, YSZ grain boundaries and the TGO (Fig. 5). In the model, CPE represents the constant phase element as described earlier. With the above models, a typical fitted result for the 1000 h oxidised specimen is shown in Fig. 6, together with the measured spectrum except for the electrode effect range. Obviously, the fitted and measured results fit very well to each other, showing that the results are reliable. However, it should be appreciated that impedance measurements can be used to determine the response of the TGO only when it is continuous. The resistance and capacitance of the TGO are represented in Fig. 7 as a function of thickness with the oxidation time given for each data point. Clearly, the resistance of the TGO increases and the capacitance decreases with increasing TGO thickness. Nevertheless, the above relationships are not absolutely linear, indicating that the resistivity and dielectric constant vary during oxidation. The resistivity and dielectric constant of the TGO are shown in Fig. 8 as a function of oxidation time. The resistivity increases with increasing oxidation time and the dielectric constant shows no apparent change before 1000 h oxidation but decreases thereafter. The TGO may be densified with oxidation, giving rise to an increase in resistivity. Normally, alumina is a p-type semiconductor. Its resistivity increases with decreasing oxygen partial pressure that

51

Fig. 6. (a) A typical fitted impedance spectrum together with the measured one for the 1000 h oxidised specimen and (b) the highfrequency range in (a).

Fig. 7. Resistance and capacitance of the TGO as a function of oxidation time at 400 8C, determined by impedance analysis.

Fig. 5. Equivalent circuits for (a) two R /C components for the specimens as-received and oxidised for 100 h and (b) three R /C components for the specimens oxidised for 500 h or longer in series connection.

usually decreases with decreasing porosity within the material [21]. Therefore, the denser the alumina, the smaller is the oxygen partial pressure within the material, leading to a higher resistivity. The reason for the dielectric constant decreasing is not clear and needs to be studied further. As mentioned in Section 1, members of our team [19] studied the degradation of thermal barrier coatings at 1150 8C and found that the resistivity of the TGO decreased with increasing oxidation time after 10 h oxidation because of the degradation

52

S. Song, P. Xiao / Materials Science and Engineering B97 (2003) 46 /53

Fig. 10. Resistance and capacitance of the YSZ grain boundaries at 400 8C, determined by impedance analysis. Fig. 8. Resistivity and dielectric constant of the TGO as a function of oxidation time at 400 8C, determined by impedance analysis.

of the TGO in terms of microstructure and composition. Accordingly, the increase in resistivity found in present study demonstrates that the composition of the TGO has not apparently changed and the metallic substrate is effectively protected by the continuous TGO layer. The resistance and capacitance of the YSZ bulk are represented in Fig. 9 as a function of oxidation time. Obviously, both resistance and capacitance show no apparent change during oxidation. The resistance and capacitance of the YSZ grain boundaries are plotted in Fig. 10 as a function of oxidation time. The boundary capacitance in the YSZ shows no apparent change during oxidation whereas the boundary resistance increased after 2000 h oxidation. X-ray diffraction showed no apparent phase transformation in the YSZ after up to 3000 h oxidation at 950 8C [37]. The increase in the resistance of the YSZ grain boundaries might therefore be caused by yttria segregation at grain boundaries. Further study needs to be carried out to

understand the evolutions of the composition and electrical properties of YSZ grains and grain boundaries.

4. Summary Oxidation of APS-prepared TBCs has been carried out in air at 950 8C. When the oxidation time is shorter than 500 h, the scale (alumina) formed between the top and bond coats is not continuous. In this scenario, impedance measurements cannot be used to detect the alumina scale because the YSZ is much more conductive than the alumina, leading to most of the current passing through the YSZ rather than the alumina. After oxidation for 500 /3000 h, impedance measurements show that a continuous alumina scale is formed. In the impedance spectra, there are four relaxation processes, which correspond to the YSZ grains, the YSZ grain boundaries, the TGO, and the metal electrode effect. Impedance analysis demonstrates that the resistivity of the alumina scale increases with increasing oxidation time. This implies that the protective capability of the scale is enhanced with increasing oxidation time within the oxidation time regime used in the present work. The electrical properties of YSZ grain boundaries changed after the TBCs were oxidised for more than 2000 h. This indicates that there is some change at the grain boundaries, even if we cannot detect it using conventional SEM and energy-dispersive spectroscopy.

Acknowledgements Fig. 9. Resistance and capacitance of the YSZ bulk at 400 8C, determined by impedance analysis.

This work was supported by EPSRC (UK) under Grant No. GR/M86743. The authors thank Prof. B.

S. Song, P. Xiao / Materials Science and Engineering B97 (2003) 46 /53

Ralph of Brunel University for commenting on the manuscript.

References [1] W.J. Brindley, J. Thermal. Spray. Technol. 6 (1997) 3. [2] J.A. Haynes, E.D. Rigney, M.K. Ferber, W.D. Porter, Surf. Coat. Technol. 86-87 (1996) 102. [3] R.A. Miller, J. Thermal. Spray. Technol. 6 (1997) 35. [4] S. Bose, J. DeMasi-Marcin, J. Thermal. Spray. Technol. 6 (1997) 99. [5] V. Sergo, D.R. Clarke, J. Am. Ceram. Soc. 81 (1998) 3237. [6] R.L. Jones, Mater. High Temp. 9 (1991) 228. [7] R.C. Hendricks, G. McDonald, R.L. Mullen, Ceram. Eng. Sci. Proc. 4 (1983) 802. [8] R.L. Jones, D. Mess, Surf. Coat. Technol. 86 /87 (1996) 94. [9] M.Y. He, A.G. Evans, J.W. Hutchinson, Mater. Sci. Eng. A 245 (1998) 16. [10] C.C. Berndt, J. Eng. Gas Turbine Power 107 (1985) 142. [11] S. Harada, Y. Ikeda, Y. Mizuta, Y. Sugita, A. Ito, The inspection of delaminated defects in ZrO2-coated material by infrared thermography techniques, in: Proceedings of the International Symposium on Thermographic NDT and E Techniques, 1995, pp. 47 /52. [12] M. Saitoh, Y. Itoh, J. Ishii, Evaluation of thermal barrier coating using infrared thermography, in: Proceedings of the International Symposium on Thermographic NDT and E Techniques, 1995, pp. 53 /58. [13] X. Peng, D.R. Clarke, J. Am. Ceram. Soc. 83 (2000) 1165. [14] F. Mansfeld, Corrosion 36 (1981) 301. [15] J.R. MacDonald (Ed.), Impedance Spectroscopy, Wiley, Chichester, 1987. [16] N.D. Cogger, N.J. Evans, An introduction to electrochemical impedance measurement, Technical Report Number 006, Solartron Ltd., Farnborough, UK, 1998. [17] A.J. Bard, L.R. Faulkner, Techniques Based on Concepts of Impedance, Electrochemical Methods (Fundamentals and Application), Wiley, Chichester, 1980 (Chapter 9).

53

[18] X. Wang, J.F. Mei, P. Xiao, J. Eur. Ceram. Soc. 21 (2001) 855. [19] M.S. Ali, S.-H. Song, P. Xiao, J. Eur. Ceram. Soc. 22 (2002) 101. [20] M.S. Ali, P. Xiao, Acta. Mater., submitted for publication. [21] N. Birks, G.H. Meier, Introduction to High-temperature Oxidation of Metals, Edward Arnold, London, 1983. [22] M.Y. He, A.G. Evans, J.W. Hutchinson, Mater. Sci. Eng. A 245 (1998) 168. [23] J.A. Haynes, M.K. Ferber, W.D. Porter, E.D. Rigney, Oxid. Met. 52 (1999) 31. [24] W.J. Brindley, R.A. Miller, Surf. Coat. Technol. 43 /44 (1990) 446. [25] L. Lelait, S. Alperine, R. Mevrel, J. Mater. Sci. 27 (1992) 5. [26] K. Ogawa, D. Minkov, T. Shoji, M. Sato, H. Hashimoto, NDT&E. Int. 32 (1999) 177. [27] J.A. Haynes, M.K. Ferber, W.D. Porter, E.D. Rigney, Mater High Temp. 16 (1999). [28] M.S. Ali, S.-H. Song, P. Xiao, J. Mater. Sci. 37 (2002) 2097. [29] M. Schutze, Protective Oxide Scales and their Breakdown, Wiley, New York, 1997, p. 49. [30] B.A. Pint, I.G. Wright, W.Y. Lee, Y. Zhang, K. Prussner, K.B. Alexander, Mater. Sci. Eng. A 245 (1998) 201. [31] J.G. Fletcher, A.R. West, J.T.S. Irvine, J. Electrochem. Soc. 142 (1995) 2650. [32] S.T. Amaral, I.L. Muller, Corrosion 55 (1999) 17. [33] T. Jacobsen, B. Zachau-Christiansen, L. Bay, S. Skaarup, SOFC cathode mechanisms, in: F.W. Paulsen, N. Bonanos, S. Linderoth, M. Mogensen, B. Zachau-Christiansen (Eds.), High-temperature Electrochemistry: Ceramics and Metals, Proceedings of the 17th Riso International Symposium on Materials Science, Riso, National Laboratory, Roskilde, Denmark, 1996, pp. 29 /40. [34] X. Wang, Impedance spectroscopy evaluation of ceramic materials, Ph.D. Thesis, Brunel University, UK, 2001. [35] M.S. Ali, Degradation of thermal barrier coatings, Ph.D. Thesis, Brunel University, UK, 2001. [36] M.C. Steil, F. Thevenot, M. Kleitz, J. Electrochem. Soc. 144 (1997) 390. [37] S.-H. Song, P. Xiao, Unpublished work, Manchester Materials Science Centre, University of Manchester/UMIST, Manchester, UK.