- Email: [email protected]
The effect of preoxidation atmosphere on oxidation behavior and thermal cycle life of thermal barrier coatings
Materials Science and Engineering A 441 (2006) 119–125
The effect of preoxidation atmosphere on oxidation behavior and thermal cycle life of thermal barrier coatings Mineaki Matsumoto a,∗ , Kazuyuki Hayakawa a , Satoshi Kitaoka a , Hideaki Matsubara a , Hiroshi Takayama b , Yukio Kagiya b , Yuuji Sugita b a
Materials Research and Development Laboratory, Japan Fine Ceramics Center, 2-4-1 Mutsuno, Atsuta-ku, Nagoya 456-8587, Japan b Electric Power R&D Center, Chubu Electric Power Co. Inc., 20-1 Kitasekiyama, Ohdaka-cho, Midori, Nagoya 459-8522, Japan Received 24 May 2006; received in revised form 10 July 2006; accepted 18 August 2006
Abstract The effects of oxygen partial pressure (pO2 ) of the preoxidation atmosphere on the growth of thermally grown oxide (TGO) and thermal cycle life of plasma-sprayed thermal barrier coatings (TBCs) were investigated. The pO2 of the preoxidation atmosphere was controlled by using a solidstate electrochemical oxygen pump system. The purity and microstructure of continuous Al2 O3 layer formed on the bond coat during preoxidation at 1050 ◦ C were highly influenced by the pO2 of the atmosphere. The specimen preoxidized at 1050 ◦ C under a pO2 of 10−14 to 10−15 atm, which is around the dissolution pressure of (Co, Ni)(Al, Cr)2 O4 spinel, showed the lowest growth rate of TGO and the longest thermal cycle life. © 2006 Elsevier B.V. All rights reserved. Keywords: Thermal barrier coating; Thermally grown oxide; Preoxidation; Oxygen partial pressure
1. Introduction Development of advanced thermal barrier coatings (TBCs) is the most promising way to increase the efficiency of gas turbines. Current state-of-the-art TBCs typically consist of an Y2 O3 stabilized ZrO2 (YSZ) top coat and a metallic bond coat (MCrAlY, M = Co, Ni). The top coat acts as a thermal insulator, while the bond coat provides oxidation protection for the underlying superalloy by forming protective oxide scale. It is generally accepted that bond coat oxidation is a critical factor controlling the life of TBCs. When TBCs are used at high temperatures, a thermally grown oxide (TGO) forms at the top coat/bond coat interface. The TGO growth leads to increase residual stress, which accelerates the spallation of TBCs. Recent research has shown that failure of TBCs occurred when the TGO attained a critical thickness in the range of 3–10 m [1]. The oxidation resistance of a bond coat relies on the ability of the alloy to produce a stable, continuous, slow growing and adherent TGO on its surface. Formation of pure ␣-Al2 O3 as
∗
Corresponding author. Tel.: +81 52 871 3500; fax: +81 52 871 3599. E-mail address: [email protected] (M. Matsumoto).
0921-5093/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.msea.2006.08.099
a protective layer on the bond coat surface is preferred due to the low diffusion rate of oxygen and metal ions through it, as well as its high chemical and thermal stability [2]. However, the oxidation of MCrAlY usually accompanies fast growing, non-protective oxide phases such as (Co, Ni)(Al, Cr)2 O4 and (Co, Ni)O, which are believed to promote the spalling of TBCs [3–5]. Much effort has been directed toward improving oxidation resistance of bond coats to prolong the lives of TBCs. The incorporation of thin protective ␣-Al2 O3 between a top coat and a bond coat before use is one of the most promising ways to reduce the rate of TGO growth. A protective interlayer can be formed uniformly by applying heat treatments before or after the deposition of a top coat [6–9] or by using another method such as CVD [10,11]. There are patents which propose that the preoxidation of a bond coat under a low oxygen partial pressure (pO2 ) before the deposition of a top coat is effective in suppressing the oxidation due to protective ␣-Al2 O3 formation [8,9]. However, no experimental results showing the optimum pO2 for the preoxidation atmosphere are available in the open literature. The purpose of this study is to clarify the effect of pO2 during the preoxidation of bond coats on the growth of TGO at high
120
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
temperature. Thermal cycle lives of the TBCs under controlled preoxidation condition are also determined. 2. Experimental Co–Ni–Cr–Al–Y alloy (in mass% of Co–32%Ni–16%Cr– 8%Al–0.5%Y) was vacuum plasma sprayed to a thickness of 150 m on grit-blasted substrates (Inconel 738LC) with dimensions of 20 mm × 20 mm × 3 mm. Preoxidation heat treatments of the substrates before the deposition of top coats were performed at 1050 ◦ C for 4 h in an Ar flow with a controlled pO2 by using a furnace equipped with a solid-state oxygen pump system. Oxygen in the Ar flow was pumped out in a CaO stabilized ZrO2 tube with an imposed voltage of up to 1.3 V at 800 ◦ C, and then the Ar flow with a controlled pO2 was introduced into the furnace. The pO2 in the furnace during the heat treatment was monitored using an oxygen sensor located at the outlet of the furnace. The sensor was calibrated under various pO2 values by using another oxygen sensor which was inserted directly into the furnace operated at 1050 ◦ C. Samples fabricated in this study are listed in Table 1. Sample A was preoxidized in air, while sample B was preoxidized in an Ar flow without controlling the pO2 . The flow rate of Ar was 2 × 10−4 m3 /min. The purity of Ar was >99.9999% with a pO2 of about 10−4 atm. The pO2 of the atmosphere during preoxidation of sample B was measured to 10−12 to 10−13 atm. Samples C and D were annealed in an Ar flow with a reduced pO2 of 10−14 to 10−17 atm. A sample without preoxidation (sample E) was also prepared for comparison. After the preoxidation, ZrO2 –8 wt.%Y2 O3 top coats were airplasma sprayed on the substrates to a thickness of about 300 m. Samples without top coats were also produced for analysis of the TGO. Samples were subjected to an oxidation test at 1200 ◦ C in air. The thickness of the TGO formed by oxidation was measured by cross-sectional images. A scanning electron microscope (SEM, Hitachi S-4500) equipped with an energy dispersive X-ray spectrometer (EDS) was employed to investigate the microstructure of the TGO. TGO phases were analyzed by X-ray diffraction (XRD, Philips, PW1877) using Cu K␣ radiation. Photostimulated Cr3+ luminescence spectroscopy (PSLS) [12] was also used for phase identification in the initial TGO. The element distribution as a function of depth below the initial TGO surface was determined by secondary ion mass spectrometry (SIMS, Physical Electrons ADEPT1010). The depth profiling was performed using a 5 keV Cs+ beam over an area
of about 88 m × 144 m up to a depth of 3 m. The results were quantified using a profile of an Al–Cr–Ni–Co–O sintered compact as a standard. Thermal cycle tests of the samples were performed using a vertical furnace. The heating time was 10 min at 1150 ◦ C, then the samples were moved into the water at 25 ◦ C for 2 min. Before the testing, samples were oxidized in air at 1200 ◦ C for 50 h. The lives of the coatings were determined by the number of cycles at which the coating failure occurred. 3. Results and discussion 3.1. Analysis of initial TGO formed by preoxidation Samples that had not been overlaid with top coats were analyzed by XRD, SEM and SIMS in order to investigate the phases in the initial TGO. Fig. 1 presents XRD patterns of the bond coat surface of samples after preoxidation. Besides Co–Ni–Cr–Al–Y peaks (␥/␥ and β), ␣-Al2 O3 and (Co, Ni)(Al, Cr)2 O4 spinel are identified in sample A, showing that the initial TGO consists mainly of these two phases. In contrast, only ␣-Al2 O3 is identified as a phase of the initial TGO in samples B–D. Fig. 2 shows the surface morphologies of the initial TGO of samples A–D. Blade-like crystals are observed on the surfaces of samples A and B, whereas such crystals are not seen in samples C and D. PLPS measurements showed evidence of both and ␣-Al2 O3 on the surfaces of samples A and B, showing that the blade-like crystals are transient Al2 O3 formed during the transformation from metastable ␥- and/or -Al2 O3 to ␣-Al2 O3 as reported previously [13,14]. The transient Al2 O3 is believed to be deleterious because the growth rate of -Al2 O3 as a TGO is an order of magnitude higher than that of ␣-Al2 O3 [15]. The large decrease in volume that accompanies transformation from ␥- and/or -Al2 O3 to ␣-Al2 O3 at high temperature will also be detrimental to the TBC life [2,13,16]. On the other hand, only ␣-Al2 O3 was identified by PLPS for samples C and D. It is noteworthy that the formation of transient Al2 O3 shows dependence on the pO2 of the preoxidation atmosphere. The fundamental
Table 1 Samples used in this study Sample
pO2 during preoxidation at 1050 ◦ C
A B C D E
0.2 atm (in air) 10−12 to 10−13 atm 10−14 to 10−15 atm 10−16 to 10−17 atm Without preoxidation
Fig. 1. XRD patterns of the bond coat surfaces of samples before overlaying with top coat: (1) sample E, (2) sample A, (3) sample B, (4) sample C and (5) sample D.
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
121
Fig. 2. Surface morphology of each sample in the as-processed condition.
reasons for this phenomenon, however, are not understood, and are now under investigation. Fractured cross-sectional images of the initial TGO can be compared in Fig. 3. The initial TGO of samples A and B apparently consists of two layers. The outer layer contains blade-like crystals and equiaxed grains, which are believed transient Al2 O3 and (Co, Ni)(Al, Cr)2 O4 spinel, respectively. The inner layer is composed of columnar grains which are believed to be ␣Al2 O3 . In contrast, the initial TGO of samples C and D consists almost entirely of large columnar grains, which are ␣-Al2 O3 as identified by XRD and PSLS. It is apparent that the grain
size in the lateral direction of the columnar Al2 O3 of samples C and D (∼0.4 m) is much larger than that of samples A and B (∼0.2 m). The initial TGO was also analyzed by SIMS in order to determine impurities that were present in the TGO. Fig. 4 presents the depth profiles of samples B and C. In both cases, Al and O are the dominant species at the surface. It should be noted that a few % of Co and Ni is concentrated on the surface of sample B, and the amounts of these elements near the surface of sample C are about one tenth lower than those for sample B. It is believed that a small amount of (Co, Ni)(Al, Cr)2 O4 is present on the surface
Fig. 3. Fractured cross-sectional images of the initial TGO of each sample. The arrows indicate the TGO surface and the TGO/bond coat interface. Note the different magnification of A.
122
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
Fig. 5. Calculated phase diagram as a function of pO2 in Co–32%Ni–16%Cr– 8%Al–0.5%Y system at 1050 ◦ C.
3.2. Oxidation behavior and thermal cycle life
Fig. 4. SIMS depth profiles of selected elements through the bond coats of samples B and C after preoxidation: (a) sample B and (b) sample C.
of sample B as observed by other researchers [6,17], although it was not identified by XRD as shown in Fig. 1. In contrast, it is clear that highly pure Al2 O3 is formed in sample C. Fig. 5 shows a relationship between the pO2 and the equilibrium phases of Co–32%Ni–16%Cr–8%Al–0.5%Y at 1050 ◦ C under 1 atm obtained by calculation, where optimization of the thermodynamic parameters and calculation of the phase equibria are carried out using the Factsage program set. At pO2 values in the range of 0.2–10−14 atm, (Co, Ni)O, (Co, Ni)(Al, Cr)2 O4 and a small amount of Y3 Al5 O12 (not shown in the figure) as well as γ phase are stable. For samples A and B, the pO2 values of the preoxidation atmospheres are above 10−13 atm, where (Co, Ni)(Al, Cr)2 O4 and (Co, Ni)O are stable phases. Thus, the initial TGO of samples A and B can contain these phases. At pO2 values in the range of 10−22 to 10−14 atm, (Al, Cr)2 O3 corundum as well as ␥/␥ are stable, which predicts that the initial TGO of samples C and D will consist of pure (Al, Cr)2 O3 with few impurities such as Co and Ni. The calculation results are in good agreement with the XRD and SIMS results as mentioned above.
Fig. 6 shows microstructures of samples after the deposition of top coats. Initial thickness of TGO of the samples A–D formed during the preoxidation are about 2, 1, 0.8 and 0.1 m, respectively. It is clear that the initial TGO thickness decreases as the pO2 of the preoxidation atmosphere is decreased. Fig. 7 shows the TGO thickness of the samples as a function of thermal exposure time at 1200 ◦ C in air. Although the thickness of the TGO was largely scattered due to local oxidation of unmelted particles in the bond coat, only the average thickness is shown in Fig. 4 for simplification. The TGO growth of sample E, which did not undergo preoxidation, shows the fastest growth rate, followed by samples A, B, D and C, in that order. The relatively low growth rate for samples with preoxidation shows that the Al2 O3 layer formed by the preoxidation heat treatment reduces the rate of TGO growth. Of the samples with preoxidation, the lowest growth rate of TGO is observed for sample C, showing that controlling the pO2 of the preoxidation atmosphere is an effective way to suppress oxidation of a bond coat. The microstructures of the samples after oxidation in air at 1200 ◦ C for 50 h are compared in Fig. 8. The TGO of sample E consists of two layers: the upper layer is porous and has relatively bright gray contrast, and the bottom layer is dense and has dark contrast. SEM-EDS analysis showed that the upper layer of porous oxides contains large amounts of Ni, Co and Cr as well as Al and O. It is believed that the porous oxide is mainly (Co, Ni)(Al, Cr)2 O4 spinel as discussed below. The EDS analysis also revealed that a high concentration of Al and O, and low concentrations of other atoms in the bottom layer, suggesting that this later is composed mainly of Al2 O3 . For samples with preoxidation, the TGO does not show a layered structure such as sample E. However, porous gray oxides (spinel) are often observed for samples A and B, although the amounts are smaller than in sample E. In contrast, porous oxides are scarcely found in sample C. The TGO formed on the surfaces of bond coats after the oxidation test were also analyzed by XRD to clarify the changes in TGO during oxidation. Fig. 9 shows XRD patterns obtained from
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
123
Fig. 6. SEM images of each sample after overlaying with top coat.
the surfaces of samples without top coats oxidized at 1200 ◦ C for 10 h in air. Peaks of ␣-Al2 O3 and (Co, Ni)(Al, Cr)2 O4 spinel are found in every specimen, however, the intensities of these phases differ between the samples. Namely, strong spinel peaks are observed for samples A and B, while spinel peaks are much weaker in sample C. The volume ratio of (Co, Ni)(Al, Cr)2 O4 in the TGO of samples A, B, C, D and E estimated from the XRD result are 25%, 22%, 7%, 14% and 45%, respectively, where XRD patterns of powder mixtures of ␣-Al2 O3 and CoAl2 O4 were used to quantify the phases. It is widely accepted that the formation of transient oxides such as spinel at a top coat/bond coat interface plays a significant role in initiating cracks at the interface because of their high growth rates and porous structure [6–9]. In this study, the amount of spinel in the TGO for sample C is the lowest, which means that the quality of the TGO of sample C is superior to that of the TGO of the other samples.
Fig. 7. TGO thickness of each sample as a function of thermal exposure time at 1200 ◦ C.
The highest oxidation resistance, which was obtained in sample C, can be attributed to the formation of large ␣-Al2 O3 grains during preoxidation as shown in Fig. 3. It is well known that scales consisting of ␣-Al2 O3 grains grow predominantly by the diffusion of oxygen along oxide grain boundaries [2]. The larger grains mean there is less grain boundary area for oxygen diffusion, which results in a low TGO growth rate. Recently, it was revealed by transmission electron microscopy observation that the TGO of plasma-sprayed TBC annealed in Ar was thin and consisted of pure Al2 O3 with large grains, whereas the TGO of the TBC annealed in air was thick and consisted of Al2 O3 and the other oxides containing Al, Cr, Ni and Co with small equiaxed fine grains [6]. It was pointed out that the TGO of pure Al2 O3 with large grains contributes to reduce oxidation because the grain boundaries act as oxygen inward diffusion paths as well as cation (Co, Ni) outward diffusion paths [6]. In this case, the initial TGO of samples A and B also consist of ␣-Al2 O3 with a small grain size as well as spinel and transient Al2 O3 . The relatively high pO2 of the preoxidation atmosphere of the samples leads to a high Co and Ni content in the TGO, which will inhibit grain growth of ␣-Al2 O3 due to drag and/or pinning effects. In contrast, the initial TGO of sample C is highly pure corundum as shown in Fig. 4. This is because the pO2 of the preoxidation atmosphere is 10−14 to 10−15 atm, which is around the dissociation pressure of (Co, Ni)(Al, Cr)2 O4 . The low impurity content in the initial TGO of sample C will lead to grain growth in the lateral direction, resulting in high oxidation resistance. On the contrary, the TGO growth rate of sample D with preoxidation under a pO2 of 10−16 atm is faster than that of sample C. In terms of thermodynamics, the purity of the initial TGO is the highest for sample D which is preoxidized in a pO2 of under the dissociation pressure of (Co, Ni)(Al, Cr)2 O4 . The relatively
124
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
Fig. 8. Microstructures of each sample after heat treatment at 1200 ◦ C for 50 h.
poor oxidation resistance for sample D suggests that the initial TGO is not sufficiently thick or continuous to suppress inward diffusion of oxygen and outward diffusion of cation. Fig. 10 shows the thermal cycle lives of the samples, where each result is the average of eight measurements. The life of sample E could not be determined because TBC failure occurred during the oxidation heat treatment prior to the thermal cycle test. The average thermal cycle life of sample C is twice as long as that of samples A, B and D, which is in good agreement with the TGO growth rate as mentioned above. The long life of sample C can be attributed to its thin TGO with low spinel content. It is concluded that the optimum pO2 during preoxidation is around the dissociation pressure of (Co, Ni)(Al, Cr)2 O4 . Fig. 9. XRD patters of the bond coat surfaces of samples without top coats after heat treatment at 1200 ◦ C for 10 h: (1) sample E, (2) sample A, (3) sample B, (4) sample C and (5) sample D.
4. Conclusions The effects of preoxidation heat treatment at 1050 ◦ C on oxidation behavior and thermal cycle life of plasma-sprayed TBC were investigated in terms of the pO2 of the preoxidation atmosphere. The conclusions are summarized as follows:
Fig. 10. Furnace cycle life of each sample.
1. The initial TGO formed on bond coats during preoxidation consists of transient Al2 O3 , (Co, Ni)(Al, Cr)2 O4 spinel and ␣-Al2 O3 with relatively small grains under a pO2 of 0.2–10−13 atm, whereas highly pure ␣-Al2 O3 with large grain size was formed on bond coats under a pO2 of 10−14 to 10−17 atm. 2. The best oxidation resistance and the longest thermal cycle life was obtained for the sample preoxidized in a pO2 of 10−14 to 10−15 atm, which is around the dissolution pressure of (Co, Ni)(Al, Cr)2 O4 spinel. The initial TGO with large grain size was considered to reduce the growth rate of TGO due to less grain boundary area for oxygen inward diffusion and cation outward diffusion.
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
3. The oxidation resistance and thermal cycle life of the specimen preoxidized under a pO2 of 10−16 to 10−17 atm were relatively poor. This is because the total amount of oxygen in the preoxidation atmosphere is too small to form a protective Al2 O3 layer of adequate thickness. References [1] A.G. Evans, D.R. Mumm, J.W. Hutchinson, G.H. Meier, F.S. Pettit, Prog. Mater. Sci. 46 (2001) 505. [2] R. Prescott, M.J. Graham, Oxide Met. 28 (1992) 233. [3] N.M. Yanar, G.H. Meier, F.S. Pettit, Scripta Mater. 46 (2002) 325. [4] E.A.G. Shillington, D.R. Clarke, Acta Mater. 47 (1999) 1297. [5] M.J. Stiger, N.M. Yanar, M.G. Topping, F.S. Petti, G.H. Meier, Z. Metallkd. 90 (1999) 1069. [6] S. Takahashi, M. Yoshiba, Y. Harada, Mater. Trans. 44 (2003) 1181.
125
[7] D. Strauss, G. Muller, G. Schmacher, V. Engelko, W. Stamm, D. Clemens, W.J. Quaddakers, Surf. Coat. Technol. 135 (2001) 196. [8] R. Darolia, US Patent 6,607,611. [9] M.J. Weimer, B.A. Nagaraj, A. Bangalore, J.C. Schaeffer, J.A. Heaney, US Patent 6,532,657. [10] J.H. Sun, E. Chang, C.H. Chao, M.J. Cheng, Oxide Met. 40 (1993) 465. [11] Y.F. Su, L.F. Allard, D.W. Coffey, W.Y. Lee, Metall. Mater. Trans. A 35A (2004) 1055. [12] R.J. Cristensen, D.M. Lipkin, D.R. Clarke, Appl. Phys. Lett. 69 (1996) 3754. [13] D.M. Lipkin, H. Schaffer, F. Adar, D.R. Clarke, Appl. Phys. Lett. 70 (1997) 2550. [14] S. Taniguchi, A. Andoh, Oxide Met. 58 (2002) 545. [15] V.K. Tolpygo, D.R. Clarke, Mater. High Temp. 17 (2000) 59. [16] S. Laxman, B. Franke, B.W. Kempshall, Y.H. Sohn, L.A. Giannuzzi, K.S. Murphy, Surf. Coat. Technol. 177–178 (2004) 121. [17] T.J. Nijdam, L.P.H. Jeurgens, W.G. Sloof, Acta Mater. 54 (2005) 1643.
The effect of preoxidation atmosphere on oxidation behavior and thermal cycle life of thermal barrier coatings Mineaki Matsumoto a,∗ , Kazuyuki Hayakawa a , Satoshi Kitaoka a , Hideaki Matsubara a , Hiroshi Takayama b , Yukio Kagiya b , Yuuji Sugita b a
Materials Research and Development Laboratory, Japan Fine Ceramics Center, 2-4-1 Mutsuno, Atsuta-ku, Nagoya 456-8587, Japan b Electric Power R&D Center, Chubu Electric Power Co. Inc., 20-1 Kitasekiyama, Ohdaka-cho, Midori, Nagoya 459-8522, Japan Received 24 May 2006; received in revised form 10 July 2006; accepted 18 August 2006
Abstract The effects of oxygen partial pressure (pO2 ) of the preoxidation atmosphere on the growth of thermally grown oxide (TGO) and thermal cycle life of plasma-sprayed thermal barrier coatings (TBCs) were investigated. The pO2 of the preoxidation atmosphere was controlled by using a solidstate electrochemical oxygen pump system. The purity and microstructure of continuous Al2 O3 layer formed on the bond coat during preoxidation at 1050 ◦ C were highly influenced by the pO2 of the atmosphere. The specimen preoxidized at 1050 ◦ C under a pO2 of 10−14 to 10−15 atm, which is around the dissolution pressure of (Co, Ni)(Al, Cr)2 O4 spinel, showed the lowest growth rate of TGO and the longest thermal cycle life. © 2006 Elsevier B.V. All rights reserved. Keywords: Thermal barrier coating; Thermally grown oxide; Preoxidation; Oxygen partial pressure
1. Introduction Development of advanced thermal barrier coatings (TBCs) is the most promising way to increase the efficiency of gas turbines. Current state-of-the-art TBCs typically consist of an Y2 O3 stabilized ZrO2 (YSZ) top coat and a metallic bond coat (MCrAlY, M = Co, Ni). The top coat acts as a thermal insulator, while the bond coat provides oxidation protection for the underlying superalloy by forming protective oxide scale. It is generally accepted that bond coat oxidation is a critical factor controlling the life of TBCs. When TBCs are used at high temperatures, a thermally grown oxide (TGO) forms at the top coat/bond coat interface. The TGO growth leads to increase residual stress, which accelerates the spallation of TBCs. Recent research has shown that failure of TBCs occurred when the TGO attained a critical thickness in the range of 3–10 m [1]. The oxidation resistance of a bond coat relies on the ability of the alloy to produce a stable, continuous, slow growing and adherent TGO on its surface. Formation of pure ␣-Al2 O3 as
∗
Corresponding author. Tel.: +81 52 871 3500; fax: +81 52 871 3599. E-mail address: [email protected] (M. Matsumoto).
0921-5093/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.msea.2006.08.099
a protective layer on the bond coat surface is preferred due to the low diffusion rate of oxygen and metal ions through it, as well as its high chemical and thermal stability [2]. However, the oxidation of MCrAlY usually accompanies fast growing, non-protective oxide phases such as (Co, Ni)(Al, Cr)2 O4 and (Co, Ni)O, which are believed to promote the spalling of TBCs [3–5]. Much effort has been directed toward improving oxidation resistance of bond coats to prolong the lives of TBCs. The incorporation of thin protective ␣-Al2 O3 between a top coat and a bond coat before use is one of the most promising ways to reduce the rate of TGO growth. A protective interlayer can be formed uniformly by applying heat treatments before or after the deposition of a top coat [6–9] or by using another method such as CVD [10,11]. There are patents which propose that the preoxidation of a bond coat under a low oxygen partial pressure (pO2 ) before the deposition of a top coat is effective in suppressing the oxidation due to protective ␣-Al2 O3 formation [8,9]. However, no experimental results showing the optimum pO2 for the preoxidation atmosphere are available in the open literature. The purpose of this study is to clarify the effect of pO2 during the preoxidation of bond coats on the growth of TGO at high
120
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
temperature. Thermal cycle lives of the TBCs under controlled preoxidation condition are also determined. 2. Experimental Co–Ni–Cr–Al–Y alloy (in mass% of Co–32%Ni–16%Cr– 8%Al–0.5%Y) was vacuum plasma sprayed to a thickness of 150 m on grit-blasted substrates (Inconel 738LC) with dimensions of 20 mm × 20 mm × 3 mm. Preoxidation heat treatments of the substrates before the deposition of top coats were performed at 1050 ◦ C for 4 h in an Ar flow with a controlled pO2 by using a furnace equipped with a solid-state oxygen pump system. Oxygen in the Ar flow was pumped out in a CaO stabilized ZrO2 tube with an imposed voltage of up to 1.3 V at 800 ◦ C, and then the Ar flow with a controlled pO2 was introduced into the furnace. The pO2 in the furnace during the heat treatment was monitored using an oxygen sensor located at the outlet of the furnace. The sensor was calibrated under various pO2 values by using another oxygen sensor which was inserted directly into the furnace operated at 1050 ◦ C. Samples fabricated in this study are listed in Table 1. Sample A was preoxidized in air, while sample B was preoxidized in an Ar flow without controlling the pO2 . The flow rate of Ar was 2 × 10−4 m3 /min. The purity of Ar was >99.9999% with a pO2 of about 10−4 atm. The pO2 of the atmosphere during preoxidation of sample B was measured to 10−12 to 10−13 atm. Samples C and D were annealed in an Ar flow with a reduced pO2 of 10−14 to 10−17 atm. A sample without preoxidation (sample E) was also prepared for comparison. After the preoxidation, ZrO2 –8 wt.%Y2 O3 top coats were airplasma sprayed on the substrates to a thickness of about 300 m. Samples without top coats were also produced for analysis of the TGO. Samples were subjected to an oxidation test at 1200 ◦ C in air. The thickness of the TGO formed by oxidation was measured by cross-sectional images. A scanning electron microscope (SEM, Hitachi S-4500) equipped with an energy dispersive X-ray spectrometer (EDS) was employed to investigate the microstructure of the TGO. TGO phases were analyzed by X-ray diffraction (XRD, Philips, PW1877) using Cu K␣ radiation. Photostimulated Cr3+ luminescence spectroscopy (PSLS) [12] was also used for phase identification in the initial TGO. The element distribution as a function of depth below the initial TGO surface was determined by secondary ion mass spectrometry (SIMS, Physical Electrons ADEPT1010). The depth profiling was performed using a 5 keV Cs+ beam over an area
of about 88 m × 144 m up to a depth of 3 m. The results were quantified using a profile of an Al–Cr–Ni–Co–O sintered compact as a standard. Thermal cycle tests of the samples were performed using a vertical furnace. The heating time was 10 min at 1150 ◦ C, then the samples were moved into the water at 25 ◦ C for 2 min. Before the testing, samples were oxidized in air at 1200 ◦ C for 50 h. The lives of the coatings were determined by the number of cycles at which the coating failure occurred. 3. Results and discussion 3.1. Analysis of initial TGO formed by preoxidation Samples that had not been overlaid with top coats were analyzed by XRD, SEM and SIMS in order to investigate the phases in the initial TGO. Fig. 1 presents XRD patterns of the bond coat surface of samples after preoxidation. Besides Co–Ni–Cr–Al–Y peaks (␥/␥ and β), ␣-Al2 O3 and (Co, Ni)(Al, Cr)2 O4 spinel are identified in sample A, showing that the initial TGO consists mainly of these two phases. In contrast, only ␣-Al2 O3 is identified as a phase of the initial TGO in samples B–D. Fig. 2 shows the surface morphologies of the initial TGO of samples A–D. Blade-like crystals are observed on the surfaces of samples A and B, whereas such crystals are not seen in samples C and D. PLPS measurements showed evidence of both and ␣-Al2 O3 on the surfaces of samples A and B, showing that the blade-like crystals are transient Al2 O3 formed during the transformation from metastable ␥- and/or -Al2 O3 to ␣-Al2 O3 as reported previously [13,14]. The transient Al2 O3 is believed to be deleterious because the growth rate of -Al2 O3 as a TGO is an order of magnitude higher than that of ␣-Al2 O3 [15]. The large decrease in volume that accompanies transformation from ␥- and/or -Al2 O3 to ␣-Al2 O3 at high temperature will also be detrimental to the TBC life [2,13,16]. On the other hand, only ␣-Al2 O3 was identified by PLPS for samples C and D. It is noteworthy that the formation of transient Al2 O3 shows dependence on the pO2 of the preoxidation atmosphere. The fundamental
Table 1 Samples used in this study Sample
pO2 during preoxidation at 1050 ◦ C
A B C D E
0.2 atm (in air) 10−12 to 10−13 atm 10−14 to 10−15 atm 10−16 to 10−17 atm Without preoxidation
Fig. 1. XRD patterns of the bond coat surfaces of samples before overlaying with top coat: (1) sample E, (2) sample A, (3) sample B, (4) sample C and (5) sample D.
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
121
Fig. 2. Surface morphology of each sample in the as-processed condition.
reasons for this phenomenon, however, are not understood, and are now under investigation. Fractured cross-sectional images of the initial TGO can be compared in Fig. 3. The initial TGO of samples A and B apparently consists of two layers. The outer layer contains blade-like crystals and equiaxed grains, which are believed transient Al2 O3 and (Co, Ni)(Al, Cr)2 O4 spinel, respectively. The inner layer is composed of columnar grains which are believed to be ␣Al2 O3 . In contrast, the initial TGO of samples C and D consists almost entirely of large columnar grains, which are ␣-Al2 O3 as identified by XRD and PSLS. It is apparent that the grain
size in the lateral direction of the columnar Al2 O3 of samples C and D (∼0.4 m) is much larger than that of samples A and B (∼0.2 m). The initial TGO was also analyzed by SIMS in order to determine impurities that were present in the TGO. Fig. 4 presents the depth profiles of samples B and C. In both cases, Al and O are the dominant species at the surface. It should be noted that a few % of Co and Ni is concentrated on the surface of sample B, and the amounts of these elements near the surface of sample C are about one tenth lower than those for sample B. It is believed that a small amount of (Co, Ni)(Al, Cr)2 O4 is present on the surface
Fig. 3. Fractured cross-sectional images of the initial TGO of each sample. The arrows indicate the TGO surface and the TGO/bond coat interface. Note the different magnification of A.
122
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
Fig. 5. Calculated phase diagram as a function of pO2 in Co–32%Ni–16%Cr– 8%Al–0.5%Y system at 1050 ◦ C.
3.2. Oxidation behavior and thermal cycle life
Fig. 4. SIMS depth profiles of selected elements through the bond coats of samples B and C after preoxidation: (a) sample B and (b) sample C.
of sample B as observed by other researchers [6,17], although it was not identified by XRD as shown in Fig. 1. In contrast, it is clear that highly pure Al2 O3 is formed in sample C. Fig. 5 shows a relationship between the pO2 and the equilibrium phases of Co–32%Ni–16%Cr–8%Al–0.5%Y at 1050 ◦ C under 1 atm obtained by calculation, where optimization of the thermodynamic parameters and calculation of the phase equibria are carried out using the Factsage program set. At pO2 values in the range of 0.2–10−14 atm, (Co, Ni)O, (Co, Ni)(Al, Cr)2 O4 and a small amount of Y3 Al5 O12 (not shown in the figure) as well as γ phase are stable. For samples A and B, the pO2 values of the preoxidation atmospheres are above 10−13 atm, where (Co, Ni)(Al, Cr)2 O4 and (Co, Ni)O are stable phases. Thus, the initial TGO of samples A and B can contain these phases. At pO2 values in the range of 10−22 to 10−14 atm, (Al, Cr)2 O3 corundum as well as ␥/␥ are stable, which predicts that the initial TGO of samples C and D will consist of pure (Al, Cr)2 O3 with few impurities such as Co and Ni. The calculation results are in good agreement with the XRD and SIMS results as mentioned above.
Fig. 6 shows microstructures of samples after the deposition of top coats. Initial thickness of TGO of the samples A–D formed during the preoxidation are about 2, 1, 0.8 and 0.1 m, respectively. It is clear that the initial TGO thickness decreases as the pO2 of the preoxidation atmosphere is decreased. Fig. 7 shows the TGO thickness of the samples as a function of thermal exposure time at 1200 ◦ C in air. Although the thickness of the TGO was largely scattered due to local oxidation of unmelted particles in the bond coat, only the average thickness is shown in Fig. 4 for simplification. The TGO growth of sample E, which did not undergo preoxidation, shows the fastest growth rate, followed by samples A, B, D and C, in that order. The relatively low growth rate for samples with preoxidation shows that the Al2 O3 layer formed by the preoxidation heat treatment reduces the rate of TGO growth. Of the samples with preoxidation, the lowest growth rate of TGO is observed for sample C, showing that controlling the pO2 of the preoxidation atmosphere is an effective way to suppress oxidation of a bond coat. The microstructures of the samples after oxidation in air at 1200 ◦ C for 50 h are compared in Fig. 8. The TGO of sample E consists of two layers: the upper layer is porous and has relatively bright gray contrast, and the bottom layer is dense and has dark contrast. SEM-EDS analysis showed that the upper layer of porous oxides contains large amounts of Ni, Co and Cr as well as Al and O. It is believed that the porous oxide is mainly (Co, Ni)(Al, Cr)2 O4 spinel as discussed below. The EDS analysis also revealed that a high concentration of Al and O, and low concentrations of other atoms in the bottom layer, suggesting that this later is composed mainly of Al2 O3 . For samples with preoxidation, the TGO does not show a layered structure such as sample E. However, porous gray oxides (spinel) are often observed for samples A and B, although the amounts are smaller than in sample E. In contrast, porous oxides are scarcely found in sample C. The TGO formed on the surfaces of bond coats after the oxidation test were also analyzed by XRD to clarify the changes in TGO during oxidation. Fig. 9 shows XRD patterns obtained from
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
123
Fig. 6. SEM images of each sample after overlaying with top coat.
the surfaces of samples without top coats oxidized at 1200 ◦ C for 10 h in air. Peaks of ␣-Al2 O3 and (Co, Ni)(Al, Cr)2 O4 spinel are found in every specimen, however, the intensities of these phases differ between the samples. Namely, strong spinel peaks are observed for samples A and B, while spinel peaks are much weaker in sample C. The volume ratio of (Co, Ni)(Al, Cr)2 O4 in the TGO of samples A, B, C, D and E estimated from the XRD result are 25%, 22%, 7%, 14% and 45%, respectively, where XRD patterns of powder mixtures of ␣-Al2 O3 and CoAl2 O4 were used to quantify the phases. It is widely accepted that the formation of transient oxides such as spinel at a top coat/bond coat interface plays a significant role in initiating cracks at the interface because of their high growth rates and porous structure [6–9]. In this study, the amount of spinel in the TGO for sample C is the lowest, which means that the quality of the TGO of sample C is superior to that of the TGO of the other samples.
Fig. 7. TGO thickness of each sample as a function of thermal exposure time at 1200 ◦ C.
The highest oxidation resistance, which was obtained in sample C, can be attributed to the formation of large ␣-Al2 O3 grains during preoxidation as shown in Fig. 3. It is well known that scales consisting of ␣-Al2 O3 grains grow predominantly by the diffusion of oxygen along oxide grain boundaries [2]. The larger grains mean there is less grain boundary area for oxygen diffusion, which results in a low TGO growth rate. Recently, it was revealed by transmission electron microscopy observation that the TGO of plasma-sprayed TBC annealed in Ar was thin and consisted of pure Al2 O3 with large grains, whereas the TGO of the TBC annealed in air was thick and consisted of Al2 O3 and the other oxides containing Al, Cr, Ni and Co with small equiaxed fine grains [6]. It was pointed out that the TGO of pure Al2 O3 with large grains contributes to reduce oxidation because the grain boundaries act as oxygen inward diffusion paths as well as cation (Co, Ni) outward diffusion paths [6]. In this case, the initial TGO of samples A and B also consist of ␣-Al2 O3 with a small grain size as well as spinel and transient Al2 O3 . The relatively high pO2 of the preoxidation atmosphere of the samples leads to a high Co and Ni content in the TGO, which will inhibit grain growth of ␣-Al2 O3 due to drag and/or pinning effects. In contrast, the initial TGO of sample C is highly pure corundum as shown in Fig. 4. This is because the pO2 of the preoxidation atmosphere is 10−14 to 10−15 atm, which is around the dissociation pressure of (Co, Ni)(Al, Cr)2 O4 . The low impurity content in the initial TGO of sample C will lead to grain growth in the lateral direction, resulting in high oxidation resistance. On the contrary, the TGO growth rate of sample D with preoxidation under a pO2 of 10−16 atm is faster than that of sample C. In terms of thermodynamics, the purity of the initial TGO is the highest for sample D which is preoxidized in a pO2 of under the dissociation pressure of (Co, Ni)(Al, Cr)2 O4 . The relatively
124
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
Fig. 8. Microstructures of each sample after heat treatment at 1200 ◦ C for 50 h.
poor oxidation resistance for sample D suggests that the initial TGO is not sufficiently thick or continuous to suppress inward diffusion of oxygen and outward diffusion of cation. Fig. 10 shows the thermal cycle lives of the samples, where each result is the average of eight measurements. The life of sample E could not be determined because TBC failure occurred during the oxidation heat treatment prior to the thermal cycle test. The average thermal cycle life of sample C is twice as long as that of samples A, B and D, which is in good agreement with the TGO growth rate as mentioned above. The long life of sample C can be attributed to its thin TGO with low spinel content. It is concluded that the optimum pO2 during preoxidation is around the dissociation pressure of (Co, Ni)(Al, Cr)2 O4 . Fig. 9. XRD patters of the bond coat surfaces of samples without top coats after heat treatment at 1200 ◦ C for 10 h: (1) sample E, (2) sample A, (3) sample B, (4) sample C and (5) sample D.
4. Conclusions The effects of preoxidation heat treatment at 1050 ◦ C on oxidation behavior and thermal cycle life of plasma-sprayed TBC were investigated in terms of the pO2 of the preoxidation atmosphere. The conclusions are summarized as follows:
Fig. 10. Furnace cycle life of each sample.
1. The initial TGO formed on bond coats during preoxidation consists of transient Al2 O3 , (Co, Ni)(Al, Cr)2 O4 spinel and ␣-Al2 O3 with relatively small grains under a pO2 of 0.2–10−13 atm, whereas highly pure ␣-Al2 O3 with large grain size was formed on bond coats under a pO2 of 10−14 to 10−17 atm. 2. The best oxidation resistance and the longest thermal cycle life was obtained for the sample preoxidized in a pO2 of 10−14 to 10−15 atm, which is around the dissolution pressure of (Co, Ni)(Al, Cr)2 O4 spinel. The initial TGO with large grain size was considered to reduce the growth rate of TGO due to less grain boundary area for oxygen inward diffusion and cation outward diffusion.
M. Matsumoto et al. / Materials Science and Engineering A 441 (2006) 119–125
3. The oxidation resistance and thermal cycle life of the specimen preoxidized under a pO2 of 10−16 to 10−17 atm were relatively poor. This is because the total amount of oxygen in the preoxidation atmosphere is too small to form a protective Al2 O3 layer of adequate thickness. References [1] A.G. Evans, D.R. Mumm, J.W. Hutchinson, G.H. Meier, F.S. Pettit, Prog. Mater. Sci. 46 (2001) 505. [2] R. Prescott, M.J. Graham, Oxide Met. 28 (1992) 233. [3] N.M. Yanar, G.H. Meier, F.S. Pettit, Scripta Mater. 46 (2002) 325. [4] E.A.G. Shillington, D.R. Clarke, Acta Mater. 47 (1999) 1297. [5] M.J. Stiger, N.M. Yanar, M.G. Topping, F.S. Petti, G.H. Meier, Z. Metallkd. 90 (1999) 1069. [6] S. Takahashi, M. Yoshiba, Y. Harada, Mater. Trans. 44 (2003) 1181.
125
[7] D. Strauss, G. Muller, G. Schmacher, V. Engelko, W. Stamm, D. Clemens, W.J. Quaddakers, Surf. Coat. Technol. 135 (2001) 196. [8] R. Darolia, US Patent 6,607,611. [9] M.J. Weimer, B.A. Nagaraj, A. Bangalore, J.C. Schaeffer, J.A. Heaney, US Patent 6,532,657. [10] J.H. Sun, E. Chang, C.H. Chao, M.J. Cheng, Oxide Met. 40 (1993) 465. [11] Y.F. Su, L.F. Allard, D.W. Coffey, W.Y. Lee, Metall. Mater. Trans. A 35A (2004) 1055. [12] R.J. Cristensen, D.M. Lipkin, D.R. Clarke, Appl. Phys. Lett. 69 (1996) 3754. [13] D.M. Lipkin, H. Schaffer, F. Adar, D.R. Clarke, Appl. Phys. Lett. 70 (1997) 2550. [14] S. Taniguchi, A. Andoh, Oxide Met. 58 (2002) 545. [15] V.K. Tolpygo, D.R. Clarke, Mater. High Temp. 17 (2000) 59. [16] S. Laxman, B. Franke, B.W. Kempshall, Y.H. Sohn, L.A. Giannuzzi, K.S. Murphy, Surf. Coat. Technol. 177–178 (2004) 121. [17] T.J. Nijdam, L.P.H. Jeurgens, W.G. Sloof, Acta Mater. 54 (2005) 1643.