- Email: [email protected]
microstructure of thermally grown oxide on the failure of EB-PVD thermal barrier coating with NiCoCrAlY bond coat
Surface & Coatings Technology 200 (2006) 5869 – 5876 www.elsevier.com/locate/surfcoat
Effects of phase constituents/microstructure of thermally grown oxide on the failure of EB-PVD thermal barrier coating with NiCoCrAlY bond coat J. Liu, J.W. Byeon, Y.H. Sohn ⁎ Advanced Materials Processing and Analysis Center and Department of Mechanical, Materials and Aerospace Engineering, University of Central Florida, Orlando, FL 32816-2455, United States Received 18 August 2005; accepted in revised form 31 August 2005 Available online 25 October 2005
Abstract A correlation among thermal cycling lifetime, bond coat surface preparation, phase constituents and microstructure of thermally grown oxide (TGO) was examined for electron beam physical vapor deposited (EB-PVD) thermal barrier coatings (TBCs) consisting of ZrO2–7 wt.% Y2O3 (YSZ) ceramic topcoat, NiCoCrAlY bond coats and CMSX-4 superalloy. Variation in the bond coat surface is characterized based on surface roughness modification and pre-oxidation (1100 °C at PO2 of 10− 8 atm up to 4 h) carried out prior to YSZ deposition by EB-PVD. TBC specimens with pre-oxidized bond coats exhibited longer lifetimes than those without pre-oxidation, especially for metallographically polished bond coats. Given a surface roughness of NiCoCrAlY bond coats, TBC lifetime was observed to increase with an increase in the amount of α-Al2O3 in the initial thermally grown oxide (TGO) scale, which was detected by photostimulated luminescence (PL). Using focused ion beam in-situ lift out technique (FIB-INLO), site-specific preparation of thermally cycled TBC specimens for transmission electron microscopy (TEM) was successfully carried out. In addition to the presence of α-Al2O3, a small particulate phase identified as cubic Y2O3 was observed within the TGO scale by electron diffraction analysis on TEM. © 2005 Elsevier B.V. All rights reserved. Keywords: Thermal barrier coatings; Surface modification; Photoluminescence; Thermally grown oxide; Thermal cycling
1. Introduction Thermal barrier coatings (TBCs) have helped to improve the high temperature durability and performance of hot section components in advanced gas turbine engines [1–5]. Generally, TBC system consists of four layers: ceramic topcoat, thermally grown oxide (TGO), metallic bond coat, and superalloy substrate. Electron beam physical vapor deposition (EB-PVD) has provided TBCs with advantageous properties, such as improved durability, smooth surface finish and good erosion resistance [6]. In EB-PVD TBCs, characteristics of bond coat surface and the growth of TGO are considered to be the crucial factor influencing the failure of TBCs [7–9]. The initial defects and surface irregularities of the bond coats can give rise to inplane and out-of-plane tensile stresses. TGO growth results in a
⁎ Corresponding author. Tel.: +1 407 882 1181; fax: +1 407 882 1462. E-mail address: [email protected] (Y.H. Sohn). 0257-8972/$ - see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2005.08.140
constrained volume expansion that leads to compressive “growth” stresses (< 1 GPa) persisting at all temperatures. Upon cooling, the thermal expansion mismatch between the TGO and bond coat leads to a high compressive residual stress in the TGO [10–14] that provides the strain energy that can drive spallation, typically within and/or near the TGO scale. Phase constituent of the TGO is considered to be a critical factor influencing the adhesion at the TGO/YSZ interface. The formation of the metastable θ- and/or γ-Al2O3 and its conversion to the stable α-Al2O3 in the TGO has been reported to have a profound effect on the structural integrity of the TGO during thermal cycles [15,16]. The polymorphic transformation of Al2O3 may affect the residual stress in the TGO due to volumetric constraint and the nucleation of sub-critical cracks [17,18]. Thus, the formation of an “optimum” TGO that only consists of α-Al2O3 prior to the deposition of topcoat may help to improve the durability and reliability of TBCs. In this study, TBCs, produced by the EB-PVD as a function of NiCoCrAlY bond coat surface roughness, were examined to
5870
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
Table 1 Summary of bond coat surface roughness, Ra, the fracture surface characteristics and lifetime of TBCs as a function of bond coat surface preparation Bond coat surface preparation
As-sprayed b Hand-polished c Barrel-finished d
Pre-oxidation a
No Yes No Yes No Yes
Average bond coat surface roughness (Ra, nm) Average
St. Dev.
1321.3 1321.3 4.3 302.6 320.1 491.2
601.6 601.6 8.4 21.1 10.3 22.9
Fracture path
Percentage of TGO on the exposed bond coat %
Average lifetime of TBCs (thermal cycles)
YSZ/TGO and TGO/Bond coat YSZ/TGO and TGO/Bond coat * TGO/Bond coat YSZ/TGO TGO/Bond coat
70.5 58.9 * 18.0 98.9 24.1
87.5 101.0 0.0 95.0 25.0 50.0
*No observable TGO. Fracture between the YSZ and bond coats prior to thermal cycling. a Heat treatment at PO2 ≌ 10− 8 atm and 1100 °C for up to 4 h to form continuous TGO layer. b Surface finish as low pressure plasma was sprayed. c Surface finish by metallographic polishing. d Media tumble treatment for 90 min.
investigate the influence of bond coat surface modification and pre-oxidation heat treatment on their lifetime, microstructural development and failure characteristics. 2. Experimental procedure Twenty-six disk-shaped specimens of CMSX-4 (Ni–9.0 Co– 6.5 Cr–6.5 Ta–6.0 W–5.6 Al–3.0 Re–1.0 Ti–0.6 Mo–0.1 Hf in wt.%) superalloys coated with low pressure plasma sprayed (LPPS) NiCoCrAlY (PWA276) bond coat and EB-PVD ZrO2– 7 wt.% Y2O3 (YSZ) topcoat were employed in this study. Various surface preparation and heat treatment were carried out before the deposition of the top YSZ coatings as listed in Table 1. The bond coat surface can be categorized based on processing technique: as-sprayed (as low-pressure plasma sprayed), handpolished metallographically down to 0.25 μm, and barrelfinished (media tumble treatment for 90 min). Moreover, after each kind of the bond coat surface processing, selected specimens were pre-oxidized in an oxidizing environment (PO2 ≌ 10− 8 atm) at 1100 °C for up to 4 h. Prior to YSZ deposition, bond coat surface roughness and phase constituents within the initial oxide layer were
examined by optical profilometry (OP) (WYKOTM NT 3300 optical profilometry) and photostimulated luminescence spectroscopy (PL) (Renishaw™ System 1000B Ramanscope™), respectively. From the bond coat surface roughness profiles, the average value of the roughness Ra was calculated using: Ra ¼
n 1X Ri n i¼1
ð1Þ
where Ri is the surface roughness of each lateral resolution (0.3 μm) with vertical resolution of 0.1–3.0 nm. Each measurement contained 1000 spatial sampling (n). After the topcoat deposition, changes in the phase constituents of the TGO scale were examined by using photostimulated luminescence (PL). Intensity ratio of αAl2O3 to the total intensity, Iα/T was determined by: Iα=T ¼
IR IR þ IN þ Im
ð2Þ
where IR, IN, and Im refer to the integrated luminescence intensities of α-Al2O3, N-luminescence, and metastable alumina luminescence, respectively.
Fig. 1. Thermal cycling lifetime of TBCs as a function of bond coat surface finish and pre-oxidation heat treatment. Values in parenthesis represent the number of specimens tested for the lifetime determination.
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
A furnace thermal cycling test, that consisted of 10-min heatup to 1121 °C, 40-min hold at 1121 °C, followed by 10-min forced air-quench, was carried out for all TBC specimens. During the test, specimens were periodically withdrawn for PL measurements at room temperature to assess the residual stress and phase transformations within the TGO as a function of thermal cycles. The specimens were considered to have failed when the YSZ spallation area was greater than 50%. Microstructure and phase constituents of the fracture surfaces (i.e., the top surface of the bond coat where the YSZ had spalled, and the bottom surface of the spalled YSZ) and cross-sections were analyzed by scanning electron microscopy (SEM) and Xray energy dispersive spectroscopy (XEDS). Focused ion beam in-situ lift out technique (FIB-INLO) [19] was employed for the preparation of site-specific transmission electron microscopy (TEM) specimens with an emphasis on the TGO scale. TEM and scanning TEM (STEM) analysis included high angle annular dark field (HAADF) imaging, electron diffraction, and XEDS. 3. Results Fig. 1 shows the average thermal cycling lifetime of TBC specimens as a function of bond coat surface preparation. The reported lifetimes are low for all TBCs when compared to those available in numerous literatures. In addition, without preoxidation, TBCs with hand-polished bond coats failed before any thermal cycling (i.e., 0 lifetime). The lifetime of all TBCs increased after pre-oxidation. The most significant improvement in TBC lifetime was observed for specimens with handpolished bond coats that were pre-oxidized prior to the EB-PVD of YSZ. Table 1 reports the bond coat surface roughness values. The roughest surface was observed for as-sprayed bond coat, while the smoothest surface was observed for hand-polished bond coat. Fig. 2 presents the typical PL spectra from the TGO scale
Fig. 2. Typical PL spectra from the TGO scale developed on the TBC specimens with barrel-finished bond coats with and without pre-oxidation and TBC specimen with hand-polished bond coat and pre-oxidation, showing N-luminescence.
5871
Fig. 3. Magnitude of compressive residual stress in the α-Al2O3 scale for TBCs, determined from PL measurements, as a function of thermal cycle. The bond coats for these TBCs were not pre-oxidized.
developed on the TBC specimen with barrel-finished bond coats with and without pre-oxidation. Similar spectra were obtained for TBC specimens with as-sprayed bond coat. The Npeaks of typical PL spectrum from the TGO scale developed on the TBC specimen with hand-polished bond coats with preoxidation was also shown in Fig. 2. Figs. 3 and 4 present the magnitude of compressive residual stress in the α-Al2O3 scale as a function of thermal cycle for TBCs produced without and with pre-oxidation heat treatment, respectively. In Fig. 3, the magnitude of the initial compressive residual stress is higher for TBCs with as-sprayed bond coat, and lower for the TBCs with barrel-finished bond coat. However, during the thermal cycling test, the compressive residual stress in the α-Al2O3 scale increased rapidly for TBCs with barrel-finished bond coats after 5 cycles. In Fig. 4, the magnitude of the initial compressive residual stress is the highest for TBCs with hand-polished bond coats, followed by as-sprayed and barrel-finished. During thermal cycling test, the magnitude of compressive residual stress in the α-Al2O3 scale
Fig. 4. Magnitude of compressive residual stress within the α-Al2O3 scale for TBCs, determined from PL measurements, with pre-oxidized bond coats as a function of thermal cycle.
5872
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
Fig. 5. Cross-sectional backscattered electron micrographs of failed TBC specimens with (a) hand-polished and pre-oxidized bond coat, showing Y and Hf-rich oxides embedded in the TGO; (b) Cross-sectional backscattered electron micrograph of TBC specimen with as-sprayed and pre-oxidized bond coat; showing Y and Hf or Ni, Co and Cr-rich particles embedded in the TGO.
of TBCs with hand-polished and as-sprayed bond coats remained high compared to that of TBCs with barrel-finished bond coats as shown in Fig. 4. The standard deviation of the compressive residual stress during thermal cycling is significantly larger for TBCs with barrel-finished bond coat. The initial rise in compressive residual stress after few thermal cycles, as shown in Figs. 3 and 4 can be attributed to the
coverage of TGO scale from discontinuous to continuous [20,21]. In Gell's studies, the initial oxide scale did not form a continuous layer until after at least 10, 1-h thermal cycles at 1121 °C. Cross-sectional microstructure of TBCs with pre-oxidized bond coats are shown in Fig. 5. Fig. 5(a) presents the crosssectional microstructure of failed TBCs with hand-polished
Fig. 6. (a) HAADF images of Y2O3 embedded in the TGO (α-Al2O3) for failed TBC specimen with hand-polished and pre-oxidized bond coats. Small precipitates at the Y2O3/α-Al2O3 interface boundaries are very rich in Hf; (b) Diffraction pattern of Y2O3 (cubic) particles marked as (b) in (a); (c) Diffraction patterns of α-Al2O3 marked as (c) in (a); (d) HAADF image of TGO of as-sprayed and pre-oxidized bond coat.
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
bond coat. Particles rich in Yand Hf were frequently found in the TGO scale. Fig. 5(b) shows the cross-sectional microstructure of failed TBCs with as-sprayed bond coats. Particles rich in Y and Hf were also frequently found within the TGO, which is penetrated into the bond coats (i.e., pegging). Oxides rich in Ni, Co, and Cr were frequently found near the fracture interfaces, which may have led to the final failure. Fig. 6(a) shows a HAADF image obtained from STEM of particles embedded within the TGO layer parallel to the YSZ/ TGO/bond-coat interfaces for hand-polished and pre-oxidized bond coats. The TGO primarily consisted of α-Al2O3 as confirmed by the diffraction pattern in Fig. 6(c). XEDS shows that the particles embedded in α-Al2O3 are rich in Y or Hf. Yrich particles were identified as cubic-Y2O3 by the diffraction pattern shown in Fig. 6(b). Hf concentration was observed to be higher at Al2O3 grain boundaries and Al2O3/Y2O3 interphase boundaries. Specifically, grain boundaries of Al2O3 gave rise to brighter contrast in HAADF images in Fig. 6 (a). For TBCs with as-sprayed and pre-oxidized bond coats, the TGO also primarily consisted of α-Al2O3. Y2O3 and Cr-rich particles embedded in α-Al2O3 TGO were also observed as shown in Fig. 6(d). Backscattered electron images from the bottom surface of the spalled YSZ and the top surface of the bond coats where
5873
YSZ had spalled are presented in Fig. 7. For TBC specimen with barrel-finished bond coat without any pre-oxidation heat treatment, the phase constituents on the fracture surfaces suggest that the spallation occurred primarily along the interfaces between the YSZ and TGO as shown in Fig. 7(a) and (b). However, after pre-oxidation, the phase constituents of the fracture paths changed, along with the increase in thermal cycling lifetime. For TBC specimens with barrelfinished bond coat surface with pre-oxidation heat treatment, the phase constituents on the fracture surfaces indicate that the spallation occurred primarily along the TGO/bond coat interface as presented in Fig. 7(c) and (d). Characteristics of fracture paths and lifetimes of TBCs as a function of bond coat surface preparation are summarized in Table 1. For TBCs with hand-polished and barrel-finished bond coats after pre-oxidation, more TGO remained on the bottom surface of the spalled YSZ with increases in the thermal cycling lifetime. The fracture path remained similar for TBCs with assprayed bond coat (i.e., both interfaces of YSZ/TGO and TGO/ bond coat as well as within the TGO) after the pre-oxidation heat treatment: the improvement in the lifetime of the TBCs with as-sprayed bond coats was not significant after the preoxidation heat treatment.
Fig. 7. (a) Backscattered electron image of the bottom surface of spalled YSZ and (b) the top surface of the bond coat for TBC specimen with barrel-finished bond coat without pre-oxidation. The spallation has occurred after 20 cycles along the YSZ/TGO interface. (c) Backscattered electron image of the bottom surface of spalled YSZ and (d) the top surface of the bond coat for TBC specimen with barrel-finished and pre-oxidation bond coat. The spallation has occurred after 75 cycles along the TGO/ bond coat interface.
5874
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
4. Discussion Fig. 8 correlates the bond coat surface preparation technique, the average bond coat surface roughness (Ra), and the average thermal cycling lifetime of TBCs. For TBCs without pre-oxidation heat treatment, a longer TBC lifetime was observed with rougher bond coats as presented in Fig. 8 (a). However, such a simple relation does not exist for TBCs after the pre-oxidation heat treatment as presented in Fig. 8 (b). TBCs with hand-polished and pre-oxidized bond coats yielded higher durability than TBCs with barrel-finished and pre-oxidized bond coat. Clearly, the surface roughness of the bond coat alone is not the only factor that affects the lifetime of the TBCs. For TBC specimens produced without any pre-oxidation heat treatment, Fig. 9(a) shows the correlation among the initial residual stresses of the TGO scale (α-Al2O3), bond coat surface preparation and the thermal cycling lifetime. The magnitude of initial compressive residual stress in the α-Al2O3 is only slightly higher for TBCs with as-sprayed bond coats, which has the longer lifetime. For TBCs with pre-oxidized bond coats, the magnitude of the initial compressive residual stress in the αAl2O3 was similar for the TBCs with as-sprayed and barrelfinished bond coats. However, the lifetime of these two types of
Fig. 8. Correlation among the bond coat surface preparation method, the bond coat surface roughness and the lifetime of EB-PVD TBCs with NiCoCrAlY bond coat: (a) without pre-oxidation heat treatment; and (b) with pre-oxidation heat treatment.
Fig. 9. Correlation among the initial residual compressive stress within the αAl2O3 scale, bond coat surface preparation and the thermal cycling lifetime for TBCs specimens: (a) without pre-oxidation heat treatment and (b) with preoxidized bond coats.
specimens had a great difference as shown in Fig. 9(b). Throughout thermal cycling, for TBCs without pre-oxidation heat treatment, the magnitude of compressive residual stress within the TGO was observed to be lower for the specimens with as-sprayed bond coat. These TBCs had a longer lifetime. For TBC specimens with barrel-finished bond coat and a shorter lifetime, the magnitude of the compressive residual stress in the TGO was higher as shown in Fig. 3. For TBCs with preoxidized bond coats, the magnitude of compressive residual stress within the TGO remained higher during thermal cycling for the TBCs with longer lifetime as shown in Fig. 4. Clearly, the initial compressive residual stress within the TGO is not the only factor that can be directly and simply related to the lifetime of the TBCs. A longer lifetime was obtained for specimens with higher relative luminescence intensity from the equilibrium α-Al2O3 in the initial TGO layer for a given surface preparation/roughness as shown in Fig. 10. Phase transformation from the metastable to stable phase may be accompanied by the formation of voids due to volumetric constraints. For example, there is about 4.7% volume contraction for θ- to α-Al2O3 phase transformation [18]. Thus, these phase transformations can create damages at the YSZ/TGO interface. However, controlled pre-oxidation heat treatment that can promote the formation of α-Al2O3 and stop premature failure at the YSZ/TGO interface can increase the lifetime of the TBCs. According to Czech's study [22], a continuous α-Al2O3 layer formed on the hand-polished surface after oxidation at 950° and 1000 °C. In this study, N-luminescence from TGO scale was observed from the pre-oxidized TBC specimens with
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
hand-polished bond coat as shown in Fig. 2, presumably due to a significant presence of Cr3+ in α-Al2O3. The formation of Cr2O3 has been reported to assist the nucleation of α-Al2O3 [18]. In addition, hand-polishing may introduce residual stresses near the bond coat surface, which can provide additional strain energy that can aid the equilibrium phase nucleation and/or transformation. Moreover, the hand-polished process can remove the absorbed contaminants that can adversely affect the phase nucleation/transformation. These contaminants can also influence the integrity of TGO/bond coat interface [23]. Initial phase constituents within the TGO, not only affected the lifetime of TBCs, but also influenced the initial compressive residual stress within the α-Al2O3. Before thermal cycling, the higher relative luminescence intensity from the equilibrium α-Al2O3 corresponded to the larger magnitude of initial compressive residual stress for TBCs with pre-oxidized bond coats as shown in Fig. 11. A significant amount of spinels were observed at fracture interfaces as shown in Fig. 5. According to Lee's and Haynes's studies [24,25], they found that the fracture in the bond coat oxidation products often takes place in the region of spinels. Spinels are considered to be brittle material with low fracture toughness that weakens the interface [26,27]. The difference in the thermal expansion coefficient of the bond coat and TGO is much larger than that of the TGO and YSZ.[28] Therefore, after a significant repeated thermal fatigue, fracture would prefer to propagate at the TGO/bond coat
5875
Fig. 11. Correlation among the bond coat surface preparation method, initial residual stress within the α-Al2O3 scale, and initial relative intensity ratio of the equilibrium α-Al2O3 for pre-oxidized TBC specimens.
interface. For specimen with shorter lifetime, spallation occurs at the interface of YSZ/TGO, due to a poor adhesion between the YSZ and the TGO. Thus, TBCs with the longest durability can be achieved by having the fracture paths restricted to the TGO/bond coat interface when an optimum TGO scale with a good adhesion between YSZ and TGO is established via preoxidation heat-treatment. Similar spallation results were found by Mumm and Evans using EB-PVD TBCs [29]. Based on their research, after short time (< 10 h) of isothermal exposure at 1100 °C, delamination occurs primarily within the TGO and YSZ. After longer exposures (< 100 h), delamination occurs principally along the interface of TGO and bond coat. According to Wright [30], reactive elements (RE) (e.g., Y) inhibit S segregation at the metal-oxide interface. Also, RE segregate to oxide grain boundaries, where they can significantly reduce the outward transport of Al, hence decrease the rate of oxidation (now mostly by oxygen transport), drastically change the oxide morphology and contribute to the improved scale adherence (reduced interfacial void formation) [23]. However, oversupply of the RE can be detrimental. Increased concentration of RE in the bond coat can cause internal oxidation and formation of RE oxides [24] (e.g., HfO2, Y2O3 and YAG) within the TGO scale as shown in Fig. 6. Also, the formation of RE oxides can act as fast diffusion paths for oxygen and cause mechanical disruption if incorporated into the TGO scale. Extensive formation of Y2O3 and other RE oxides may be responsible for overall poor performance of lifetime observed in this study.
5. Conclusion Effects of surface preparation methods for a NiCoCrAlY bond coat on the thermal cyclic lifetime and failure of an EBPVD TBC were investigated in this study. Findings from this investigation are listed below: Fig. 10. Correlation among the relative initial intensity ratio of α-Al2O3, bond coat surface preparation, and the thermal cycling lifetime of TBCs specimens: TBCs with (a) as-sprayed bond coat; (b) hand-polished bond coat; (c) barrelfinished bond coat.
• A longer lifetime during 1-h thermal cycling at 1121 °C was observed for EB-PVD TBCs with as-sprayed NiCoCrAlY bond coats regardless of pre-oxidation and TBCs with handpolished bond coats only after pre-oxidation.
5876
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
• Lifetime of EB-PVD TBCs, in general, was observed to increase when a pre-oxidation heat treatment at 1100 °C with PO2=10− 8 atm was carried out prior to the YSZ deposition. With pre-oxidation heat treatment, relative photostimulated luminescence intensity of the α-Al2O3 increased. Thus, the improvement in TBC lifetime can be correlated with the increase in luminescence intensity of α-Al2O3 in the TGO scale given a surface preparation/roughness. The lifetime improvement due to the pre-oxidation was particularly significant to TBCs with hand-polished NiCoCrAlY bond coat. • Spallation-fracture paths depended on the lifetime of TBCs. Premature spallation of TBCs occurs at the interface between the YSZ and TGO. Longer durability can be achieved by restricting the fracture paths to the TGO/bond coat interface. • Small particulate phase observed through the TGO scale was identified as Y2O3 (cubic) by diffraction analysis on TEM. While the addition of Y in the NiCoCrAlY bond coat helps the adhesion of the TGO scale, excessive alloying can lead to deleterious effects. Acknowledgements The authors would like to thank Drs. F.S. Pettit and G.H. Meier at University of Pittsburgh for pre-oxidation heat treatment of TBC specimens, and Mr. K.S. Murphy at Howmet for processing the TBC. Authors would also like to express their gratitude to Drs. M. Gell and E.H. Jordan for financial support of this study, which was carried out as a part of University Turbine Systems Research (No. 01-01-SR091) awarded by South Carolina Institute for Energy Studies administered by Dr. R.A. Wenglarz. References [1] N.P. Padture, M. Gell, E.H. Jordan, Science 296 (2002) 280. [2] D.J. Wortman, B.A. Nagaraj, E.C. Duderstadt, Mater. Sci. Eng., A Struct. Mater.: Prop. Microstruct. Process. 120–121 (1989) 433.
[3] R. Miller, J. Therm. Spray Technol. 6 (1997) 35. [4] A.G. Evans, D.R. Mumm, J.W. Hutchinson, G.H. Meier, F.S. Pettit, Prog. Mater. Sci. 46 (2001) 505. [5] D. Zhu, R.A. Miller, MRS Bull. 25 (2000) 43. [6] F.C. Toriz, A.B. Thakker, S.K. Gupta, Surf. Coat. Technol. 39–40 (1989) 161. [7] P.K. Wright, Mater. Sci. Eng., A Struct. Mater.: Prop. Microstruct. Process. 245 (1998) 191. [8] J. Cheng, E.H. Jordan, B. Barber, M. Gell, Acta Mater. 46 (1998) 5839. [9] A.G. Evans, M.Y. He, J.W. Hutchinson, Prog. Mater. Sci. 46 (2001) 249. [10] D.M. Lipkin, D.R. Clarke, Oxid. Met. 45 (1996) 267. [11] V.K. Tolpygo, D.R. Clarke, Oxid. Met. 49 (1998) 187. [12] C. Mennicke, E. Schumann, C. Ulrich, M. Ruehle, Mat. Sci. Forum 389 (1997) 251. [13] V. Sergo, D.R. Clarke, J. Am. Ceram. Soc. 81 (1998) 3237. [14] R.J. Christensen, V.K. Tolpygo, D.R. Clarke, Acta Mater. 45 (1997) 1761. [15] T.F. An, H.R. Guan, X.F. Sun, Z.Q. Hu, Oxid. Met. 54 (2000) 301. [16] M.W. Brumm, H.J. Grabke, Corros. Sci. 33 (1992) 1677. [17] C. Mennicke, D.R. Mumm, D.R. Clarke, Z. Metallkd. 90 (1999) 1079. [18] D.R. Clarke, V. Sergo, M.Y. He, in: J.M. Hampikian, N.B. Dahotre (Eds.), Elevated Temperature Coatings: Science and Technology, vol. III, TMS, Warrendale, PA, 1999, p. 67. [19] K. Gamo, Nucl. Instrum. Methods Phys. Res., B Beam Interact. Mater. Atoms 121 (1997) 464. [20] L.D. Xie, Y.H. Sohn, E.H. Jordan, M. Gell, Surf. Coat. Technol. 176 (2003) 57. [21] S. Sridharan, L.D. Xie, E.H. Jordan, M. Gell, Surf. Coat. Technol. 179 (2004) 286. [22] N. Czech, Surf. Coat. Technol. 108–109 (1998) 36. [23] B.A. Pint, Oxid. Met. 45 (1996) 1. [24] E.Y. Lee, “Life prediction and failure mechanisms for thermal barrier coatings”, Ph.D. dissertation, Worcester Polytechnic Institute (1991). [25] J.A. Haynes, E.D. Rigney, M.K. Ferber, W.D. Porter, Surf. Coat. Technol. 86–87 (1996) 102. [26] C.A. Calow, I.T. Porter, J. Mater. Sci. 6 (1971) 156. [27] R.G. Vardiman, Mater. Res. Bull. 7 (1972) 699. [28] A.G. Evans, D.R. Mumm, J.W. Hutchinson, G.H. Meier, F.S. Pettit, Prog. Mater. Sci. 46 (2001) 505. [29] D.R. Mumm, A.G. Evans, Acta Mater. 48 (2000) 1815. [30] I.G. Wright, B.A. Pint, First International Conference on Industrial Gas Turbine Technologies CAME-GT Brussels, 10–11 July, 2003.
Effects of phase constituents/microstructure of thermally grown oxide on the failure of EB-PVD thermal barrier coating with NiCoCrAlY bond coat J. Liu, J.W. Byeon, Y.H. Sohn ⁎ Advanced Materials Processing and Analysis Center and Department of Mechanical, Materials and Aerospace Engineering, University of Central Florida, Orlando, FL 32816-2455, United States Received 18 August 2005; accepted in revised form 31 August 2005 Available online 25 October 2005
Abstract A correlation among thermal cycling lifetime, bond coat surface preparation, phase constituents and microstructure of thermally grown oxide (TGO) was examined for electron beam physical vapor deposited (EB-PVD) thermal barrier coatings (TBCs) consisting of ZrO2–7 wt.% Y2O3 (YSZ) ceramic topcoat, NiCoCrAlY bond coats and CMSX-4 superalloy. Variation in the bond coat surface is characterized based on surface roughness modification and pre-oxidation (1100 °C at PO2 of 10− 8 atm up to 4 h) carried out prior to YSZ deposition by EB-PVD. TBC specimens with pre-oxidized bond coats exhibited longer lifetimes than those without pre-oxidation, especially for metallographically polished bond coats. Given a surface roughness of NiCoCrAlY bond coats, TBC lifetime was observed to increase with an increase in the amount of α-Al2O3 in the initial thermally grown oxide (TGO) scale, which was detected by photostimulated luminescence (PL). Using focused ion beam in-situ lift out technique (FIB-INLO), site-specific preparation of thermally cycled TBC specimens for transmission electron microscopy (TEM) was successfully carried out. In addition to the presence of α-Al2O3, a small particulate phase identified as cubic Y2O3 was observed within the TGO scale by electron diffraction analysis on TEM. © 2005 Elsevier B.V. All rights reserved. Keywords: Thermal barrier coatings; Surface modification; Photoluminescence; Thermally grown oxide; Thermal cycling
1. Introduction Thermal barrier coatings (TBCs) have helped to improve the high temperature durability and performance of hot section components in advanced gas turbine engines [1–5]. Generally, TBC system consists of four layers: ceramic topcoat, thermally grown oxide (TGO), metallic bond coat, and superalloy substrate. Electron beam physical vapor deposition (EB-PVD) has provided TBCs with advantageous properties, such as improved durability, smooth surface finish and good erosion resistance [6]. In EB-PVD TBCs, characteristics of bond coat surface and the growth of TGO are considered to be the crucial factor influencing the failure of TBCs [7–9]. The initial defects and surface irregularities of the bond coats can give rise to inplane and out-of-plane tensile stresses. TGO growth results in a
⁎ Corresponding author. Tel.: +1 407 882 1181; fax: +1 407 882 1462. E-mail address: [email protected] (Y.H. Sohn). 0257-8972/$ - see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2005.08.140
constrained volume expansion that leads to compressive “growth” stresses (< 1 GPa) persisting at all temperatures. Upon cooling, the thermal expansion mismatch between the TGO and bond coat leads to a high compressive residual stress in the TGO [10–14] that provides the strain energy that can drive spallation, typically within and/or near the TGO scale. Phase constituent of the TGO is considered to be a critical factor influencing the adhesion at the TGO/YSZ interface. The formation of the metastable θ- and/or γ-Al2O3 and its conversion to the stable α-Al2O3 in the TGO has been reported to have a profound effect on the structural integrity of the TGO during thermal cycles [15,16]. The polymorphic transformation of Al2O3 may affect the residual stress in the TGO due to volumetric constraint and the nucleation of sub-critical cracks [17,18]. Thus, the formation of an “optimum” TGO that only consists of α-Al2O3 prior to the deposition of topcoat may help to improve the durability and reliability of TBCs. In this study, TBCs, produced by the EB-PVD as a function of NiCoCrAlY bond coat surface roughness, were examined to
5870
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
Table 1 Summary of bond coat surface roughness, Ra, the fracture surface characteristics and lifetime of TBCs as a function of bond coat surface preparation Bond coat surface preparation
As-sprayed b Hand-polished c Barrel-finished d
Pre-oxidation a
No Yes No Yes No Yes
Average bond coat surface roughness (Ra, nm) Average
St. Dev.
1321.3 1321.3 4.3 302.6 320.1 491.2
601.6 601.6 8.4 21.1 10.3 22.9
Fracture path
Percentage of TGO on the exposed bond coat %
Average lifetime of TBCs (thermal cycles)
YSZ/TGO and TGO/Bond coat YSZ/TGO and TGO/Bond coat * TGO/Bond coat YSZ/TGO TGO/Bond coat
70.5 58.9 * 18.0 98.9 24.1
87.5 101.0 0.0 95.0 25.0 50.0
*No observable TGO. Fracture between the YSZ and bond coats prior to thermal cycling. a Heat treatment at PO2 ≌ 10− 8 atm and 1100 °C for up to 4 h to form continuous TGO layer. b Surface finish as low pressure plasma was sprayed. c Surface finish by metallographic polishing. d Media tumble treatment for 90 min.
investigate the influence of bond coat surface modification and pre-oxidation heat treatment on their lifetime, microstructural development and failure characteristics. 2. Experimental procedure Twenty-six disk-shaped specimens of CMSX-4 (Ni–9.0 Co– 6.5 Cr–6.5 Ta–6.0 W–5.6 Al–3.0 Re–1.0 Ti–0.6 Mo–0.1 Hf in wt.%) superalloys coated with low pressure plasma sprayed (LPPS) NiCoCrAlY (PWA276) bond coat and EB-PVD ZrO2– 7 wt.% Y2O3 (YSZ) topcoat were employed in this study. Various surface preparation and heat treatment were carried out before the deposition of the top YSZ coatings as listed in Table 1. The bond coat surface can be categorized based on processing technique: as-sprayed (as low-pressure plasma sprayed), handpolished metallographically down to 0.25 μm, and barrelfinished (media tumble treatment for 90 min). Moreover, after each kind of the bond coat surface processing, selected specimens were pre-oxidized in an oxidizing environment (PO2 ≌ 10− 8 atm) at 1100 °C for up to 4 h. Prior to YSZ deposition, bond coat surface roughness and phase constituents within the initial oxide layer were
examined by optical profilometry (OP) (WYKOTM NT 3300 optical profilometry) and photostimulated luminescence spectroscopy (PL) (Renishaw™ System 1000B Ramanscope™), respectively. From the bond coat surface roughness profiles, the average value of the roughness Ra was calculated using: Ra ¼
n 1X Ri n i¼1
ð1Þ
where Ri is the surface roughness of each lateral resolution (0.3 μm) with vertical resolution of 0.1–3.0 nm. Each measurement contained 1000 spatial sampling (n). After the topcoat deposition, changes in the phase constituents of the TGO scale were examined by using photostimulated luminescence (PL). Intensity ratio of αAl2O3 to the total intensity, Iα/T was determined by: Iα=T ¼
IR IR þ IN þ Im
ð2Þ
where IR, IN, and Im refer to the integrated luminescence intensities of α-Al2O3, N-luminescence, and metastable alumina luminescence, respectively.
Fig. 1. Thermal cycling lifetime of TBCs as a function of bond coat surface finish and pre-oxidation heat treatment. Values in parenthesis represent the number of specimens tested for the lifetime determination.
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
A furnace thermal cycling test, that consisted of 10-min heatup to 1121 °C, 40-min hold at 1121 °C, followed by 10-min forced air-quench, was carried out for all TBC specimens. During the test, specimens were periodically withdrawn for PL measurements at room temperature to assess the residual stress and phase transformations within the TGO as a function of thermal cycles. The specimens were considered to have failed when the YSZ spallation area was greater than 50%. Microstructure and phase constituents of the fracture surfaces (i.e., the top surface of the bond coat where the YSZ had spalled, and the bottom surface of the spalled YSZ) and cross-sections were analyzed by scanning electron microscopy (SEM) and Xray energy dispersive spectroscopy (XEDS). Focused ion beam in-situ lift out technique (FIB-INLO) [19] was employed for the preparation of site-specific transmission electron microscopy (TEM) specimens with an emphasis on the TGO scale. TEM and scanning TEM (STEM) analysis included high angle annular dark field (HAADF) imaging, electron diffraction, and XEDS. 3. Results Fig. 1 shows the average thermal cycling lifetime of TBC specimens as a function of bond coat surface preparation. The reported lifetimes are low for all TBCs when compared to those available in numerous literatures. In addition, without preoxidation, TBCs with hand-polished bond coats failed before any thermal cycling (i.e., 0 lifetime). The lifetime of all TBCs increased after pre-oxidation. The most significant improvement in TBC lifetime was observed for specimens with handpolished bond coats that were pre-oxidized prior to the EB-PVD of YSZ. Table 1 reports the bond coat surface roughness values. The roughest surface was observed for as-sprayed bond coat, while the smoothest surface was observed for hand-polished bond coat. Fig. 2 presents the typical PL spectra from the TGO scale
Fig. 2. Typical PL spectra from the TGO scale developed on the TBC specimens with barrel-finished bond coats with and without pre-oxidation and TBC specimen with hand-polished bond coat and pre-oxidation, showing N-luminescence.
5871
Fig. 3. Magnitude of compressive residual stress in the α-Al2O3 scale for TBCs, determined from PL measurements, as a function of thermal cycle. The bond coats for these TBCs were not pre-oxidized.
developed on the TBC specimen with barrel-finished bond coats with and without pre-oxidation. Similar spectra were obtained for TBC specimens with as-sprayed bond coat. The Npeaks of typical PL spectrum from the TGO scale developed on the TBC specimen with hand-polished bond coats with preoxidation was also shown in Fig. 2. Figs. 3 and 4 present the magnitude of compressive residual stress in the α-Al2O3 scale as a function of thermal cycle for TBCs produced without and with pre-oxidation heat treatment, respectively. In Fig. 3, the magnitude of the initial compressive residual stress is higher for TBCs with as-sprayed bond coat, and lower for the TBCs with barrel-finished bond coat. However, during the thermal cycling test, the compressive residual stress in the α-Al2O3 scale increased rapidly for TBCs with barrel-finished bond coats after 5 cycles. In Fig. 4, the magnitude of the initial compressive residual stress is the highest for TBCs with hand-polished bond coats, followed by as-sprayed and barrel-finished. During thermal cycling test, the magnitude of compressive residual stress in the α-Al2O3 scale
Fig. 4. Magnitude of compressive residual stress within the α-Al2O3 scale for TBCs, determined from PL measurements, with pre-oxidized bond coats as a function of thermal cycle.
5872
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
Fig. 5. Cross-sectional backscattered electron micrographs of failed TBC specimens with (a) hand-polished and pre-oxidized bond coat, showing Y and Hf-rich oxides embedded in the TGO; (b) Cross-sectional backscattered electron micrograph of TBC specimen with as-sprayed and pre-oxidized bond coat; showing Y and Hf or Ni, Co and Cr-rich particles embedded in the TGO.
of TBCs with hand-polished and as-sprayed bond coats remained high compared to that of TBCs with barrel-finished bond coats as shown in Fig. 4. The standard deviation of the compressive residual stress during thermal cycling is significantly larger for TBCs with barrel-finished bond coat. The initial rise in compressive residual stress after few thermal cycles, as shown in Figs. 3 and 4 can be attributed to the
coverage of TGO scale from discontinuous to continuous [20,21]. In Gell's studies, the initial oxide scale did not form a continuous layer until after at least 10, 1-h thermal cycles at 1121 °C. Cross-sectional microstructure of TBCs with pre-oxidized bond coats are shown in Fig. 5. Fig. 5(a) presents the crosssectional microstructure of failed TBCs with hand-polished
Fig. 6. (a) HAADF images of Y2O3 embedded in the TGO (α-Al2O3) for failed TBC specimen with hand-polished and pre-oxidized bond coats. Small precipitates at the Y2O3/α-Al2O3 interface boundaries are very rich in Hf; (b) Diffraction pattern of Y2O3 (cubic) particles marked as (b) in (a); (c) Diffraction patterns of α-Al2O3 marked as (c) in (a); (d) HAADF image of TGO of as-sprayed and pre-oxidized bond coat.
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
bond coat. Particles rich in Yand Hf were frequently found in the TGO scale. Fig. 5(b) shows the cross-sectional microstructure of failed TBCs with as-sprayed bond coats. Particles rich in Y and Hf were also frequently found within the TGO, which is penetrated into the bond coats (i.e., pegging). Oxides rich in Ni, Co, and Cr were frequently found near the fracture interfaces, which may have led to the final failure. Fig. 6(a) shows a HAADF image obtained from STEM of particles embedded within the TGO layer parallel to the YSZ/ TGO/bond-coat interfaces for hand-polished and pre-oxidized bond coats. The TGO primarily consisted of α-Al2O3 as confirmed by the diffraction pattern in Fig. 6(c). XEDS shows that the particles embedded in α-Al2O3 are rich in Y or Hf. Yrich particles were identified as cubic-Y2O3 by the diffraction pattern shown in Fig. 6(b). Hf concentration was observed to be higher at Al2O3 grain boundaries and Al2O3/Y2O3 interphase boundaries. Specifically, grain boundaries of Al2O3 gave rise to brighter contrast in HAADF images in Fig. 6 (a). For TBCs with as-sprayed and pre-oxidized bond coats, the TGO also primarily consisted of α-Al2O3. Y2O3 and Cr-rich particles embedded in α-Al2O3 TGO were also observed as shown in Fig. 6(d). Backscattered electron images from the bottom surface of the spalled YSZ and the top surface of the bond coats where
5873
YSZ had spalled are presented in Fig. 7. For TBC specimen with barrel-finished bond coat without any pre-oxidation heat treatment, the phase constituents on the fracture surfaces suggest that the spallation occurred primarily along the interfaces between the YSZ and TGO as shown in Fig. 7(a) and (b). However, after pre-oxidation, the phase constituents of the fracture paths changed, along with the increase in thermal cycling lifetime. For TBC specimens with barrelfinished bond coat surface with pre-oxidation heat treatment, the phase constituents on the fracture surfaces indicate that the spallation occurred primarily along the TGO/bond coat interface as presented in Fig. 7(c) and (d). Characteristics of fracture paths and lifetimes of TBCs as a function of bond coat surface preparation are summarized in Table 1. For TBCs with hand-polished and barrel-finished bond coats after pre-oxidation, more TGO remained on the bottom surface of the spalled YSZ with increases in the thermal cycling lifetime. The fracture path remained similar for TBCs with assprayed bond coat (i.e., both interfaces of YSZ/TGO and TGO/ bond coat as well as within the TGO) after the pre-oxidation heat treatment: the improvement in the lifetime of the TBCs with as-sprayed bond coats was not significant after the preoxidation heat treatment.
Fig. 7. (a) Backscattered electron image of the bottom surface of spalled YSZ and (b) the top surface of the bond coat for TBC specimen with barrel-finished bond coat without pre-oxidation. The spallation has occurred after 20 cycles along the YSZ/TGO interface. (c) Backscattered electron image of the bottom surface of spalled YSZ and (d) the top surface of the bond coat for TBC specimen with barrel-finished and pre-oxidation bond coat. The spallation has occurred after 75 cycles along the TGO/ bond coat interface.
5874
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
4. Discussion Fig. 8 correlates the bond coat surface preparation technique, the average bond coat surface roughness (Ra), and the average thermal cycling lifetime of TBCs. For TBCs without pre-oxidation heat treatment, a longer TBC lifetime was observed with rougher bond coats as presented in Fig. 8 (a). However, such a simple relation does not exist for TBCs after the pre-oxidation heat treatment as presented in Fig. 8 (b). TBCs with hand-polished and pre-oxidized bond coats yielded higher durability than TBCs with barrel-finished and pre-oxidized bond coat. Clearly, the surface roughness of the bond coat alone is not the only factor that affects the lifetime of the TBCs. For TBC specimens produced without any pre-oxidation heat treatment, Fig. 9(a) shows the correlation among the initial residual stresses of the TGO scale (α-Al2O3), bond coat surface preparation and the thermal cycling lifetime. The magnitude of initial compressive residual stress in the α-Al2O3 is only slightly higher for TBCs with as-sprayed bond coats, which has the longer lifetime. For TBCs with pre-oxidized bond coats, the magnitude of the initial compressive residual stress in the αAl2O3 was similar for the TBCs with as-sprayed and barrelfinished bond coats. However, the lifetime of these two types of
Fig. 8. Correlation among the bond coat surface preparation method, the bond coat surface roughness and the lifetime of EB-PVD TBCs with NiCoCrAlY bond coat: (a) without pre-oxidation heat treatment; and (b) with pre-oxidation heat treatment.
Fig. 9. Correlation among the initial residual compressive stress within the αAl2O3 scale, bond coat surface preparation and the thermal cycling lifetime for TBCs specimens: (a) without pre-oxidation heat treatment and (b) with preoxidized bond coats.
specimens had a great difference as shown in Fig. 9(b). Throughout thermal cycling, for TBCs without pre-oxidation heat treatment, the magnitude of compressive residual stress within the TGO was observed to be lower for the specimens with as-sprayed bond coat. These TBCs had a longer lifetime. For TBC specimens with barrel-finished bond coat and a shorter lifetime, the magnitude of the compressive residual stress in the TGO was higher as shown in Fig. 3. For TBCs with preoxidized bond coats, the magnitude of compressive residual stress within the TGO remained higher during thermal cycling for the TBCs with longer lifetime as shown in Fig. 4. Clearly, the initial compressive residual stress within the TGO is not the only factor that can be directly and simply related to the lifetime of the TBCs. A longer lifetime was obtained for specimens with higher relative luminescence intensity from the equilibrium α-Al2O3 in the initial TGO layer for a given surface preparation/roughness as shown in Fig. 10. Phase transformation from the metastable to stable phase may be accompanied by the formation of voids due to volumetric constraints. For example, there is about 4.7% volume contraction for θ- to α-Al2O3 phase transformation [18]. Thus, these phase transformations can create damages at the YSZ/TGO interface. However, controlled pre-oxidation heat treatment that can promote the formation of α-Al2O3 and stop premature failure at the YSZ/TGO interface can increase the lifetime of the TBCs. According to Czech's study [22], a continuous α-Al2O3 layer formed on the hand-polished surface after oxidation at 950° and 1000 °C. In this study, N-luminescence from TGO scale was observed from the pre-oxidized TBC specimens with
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
hand-polished bond coat as shown in Fig. 2, presumably due to a significant presence of Cr3+ in α-Al2O3. The formation of Cr2O3 has been reported to assist the nucleation of α-Al2O3 [18]. In addition, hand-polishing may introduce residual stresses near the bond coat surface, which can provide additional strain energy that can aid the equilibrium phase nucleation and/or transformation. Moreover, the hand-polished process can remove the absorbed contaminants that can adversely affect the phase nucleation/transformation. These contaminants can also influence the integrity of TGO/bond coat interface [23]. Initial phase constituents within the TGO, not only affected the lifetime of TBCs, but also influenced the initial compressive residual stress within the α-Al2O3. Before thermal cycling, the higher relative luminescence intensity from the equilibrium α-Al2O3 corresponded to the larger magnitude of initial compressive residual stress for TBCs with pre-oxidized bond coats as shown in Fig. 11. A significant amount of spinels were observed at fracture interfaces as shown in Fig. 5. According to Lee's and Haynes's studies [24,25], they found that the fracture in the bond coat oxidation products often takes place in the region of spinels. Spinels are considered to be brittle material with low fracture toughness that weakens the interface [26,27]. The difference in the thermal expansion coefficient of the bond coat and TGO is much larger than that of the TGO and YSZ.[28] Therefore, after a significant repeated thermal fatigue, fracture would prefer to propagate at the TGO/bond coat
5875
Fig. 11. Correlation among the bond coat surface preparation method, initial residual stress within the α-Al2O3 scale, and initial relative intensity ratio of the equilibrium α-Al2O3 for pre-oxidized TBC specimens.
interface. For specimen with shorter lifetime, spallation occurs at the interface of YSZ/TGO, due to a poor adhesion between the YSZ and the TGO. Thus, TBCs with the longest durability can be achieved by having the fracture paths restricted to the TGO/bond coat interface when an optimum TGO scale with a good adhesion between YSZ and TGO is established via preoxidation heat-treatment. Similar spallation results were found by Mumm and Evans using EB-PVD TBCs [29]. Based on their research, after short time (< 10 h) of isothermal exposure at 1100 °C, delamination occurs primarily within the TGO and YSZ. After longer exposures (< 100 h), delamination occurs principally along the interface of TGO and bond coat. According to Wright [30], reactive elements (RE) (e.g., Y) inhibit S segregation at the metal-oxide interface. Also, RE segregate to oxide grain boundaries, where they can significantly reduce the outward transport of Al, hence decrease the rate of oxidation (now mostly by oxygen transport), drastically change the oxide morphology and contribute to the improved scale adherence (reduced interfacial void formation) [23]. However, oversupply of the RE can be detrimental. Increased concentration of RE in the bond coat can cause internal oxidation and formation of RE oxides [24] (e.g., HfO2, Y2O3 and YAG) within the TGO scale as shown in Fig. 6. Also, the formation of RE oxides can act as fast diffusion paths for oxygen and cause mechanical disruption if incorporated into the TGO scale. Extensive formation of Y2O3 and other RE oxides may be responsible for overall poor performance of lifetime observed in this study.
5. Conclusion Effects of surface preparation methods for a NiCoCrAlY bond coat on the thermal cyclic lifetime and failure of an EBPVD TBC were investigated in this study. Findings from this investigation are listed below: Fig. 10. Correlation among the relative initial intensity ratio of α-Al2O3, bond coat surface preparation, and the thermal cycling lifetime of TBCs specimens: TBCs with (a) as-sprayed bond coat; (b) hand-polished bond coat; (c) barrelfinished bond coat.
• A longer lifetime during 1-h thermal cycling at 1121 °C was observed for EB-PVD TBCs with as-sprayed NiCoCrAlY bond coats regardless of pre-oxidation and TBCs with handpolished bond coats only after pre-oxidation.
5876
J. Liu et al. / Surface & Coatings Technology 200 (2006) 5869–5876
• Lifetime of EB-PVD TBCs, in general, was observed to increase when a pre-oxidation heat treatment at 1100 °C with PO2=10− 8 atm was carried out prior to the YSZ deposition. With pre-oxidation heat treatment, relative photostimulated luminescence intensity of the α-Al2O3 increased. Thus, the improvement in TBC lifetime can be correlated with the increase in luminescence intensity of α-Al2O3 in the TGO scale given a surface preparation/roughness. The lifetime improvement due to the pre-oxidation was particularly significant to TBCs with hand-polished NiCoCrAlY bond coat. • Spallation-fracture paths depended on the lifetime of TBCs. Premature spallation of TBCs occurs at the interface between the YSZ and TGO. Longer durability can be achieved by restricting the fracture paths to the TGO/bond coat interface. • Small particulate phase observed through the TGO scale was identified as Y2O3 (cubic) by diffraction analysis on TEM. While the addition of Y in the NiCoCrAlY bond coat helps the adhesion of the TGO scale, excessive alloying can lead to deleterious effects. Acknowledgements The authors would like to thank Drs. F.S. Pettit and G.H. Meier at University of Pittsburgh for pre-oxidation heat treatment of TBC specimens, and Mr. K.S. Murphy at Howmet for processing the TBC. Authors would also like to express their gratitude to Drs. M. Gell and E.H. Jordan for financial support of this study, which was carried out as a part of University Turbine Systems Research (No. 01-01-SR091) awarded by South Carolina Institute for Energy Studies administered by Dr. R.A. Wenglarz. References [1] N.P. Padture, M. Gell, E.H. Jordan, Science 296 (2002) 280. [2] D.J. Wortman, B.A. Nagaraj, E.C. Duderstadt, Mater. Sci. Eng., A Struct. Mater.: Prop. Microstruct. Process. 120–121 (1989) 433.
[3] R. Miller, J. Therm. Spray Technol. 6 (1997) 35. [4] A.G. Evans, D.R. Mumm, J.W. Hutchinson, G.H. Meier, F.S. Pettit, Prog. Mater. Sci. 46 (2001) 505. [5] D. Zhu, R.A. Miller, MRS Bull. 25 (2000) 43. [6] F.C. Toriz, A.B. Thakker, S.K. Gupta, Surf. Coat. Technol. 39–40 (1989) 161. [7] P.K. Wright, Mater. Sci. Eng., A Struct. Mater.: Prop. Microstruct. Process. 245 (1998) 191. [8] J. Cheng, E.H. Jordan, B. Barber, M. Gell, Acta Mater. 46 (1998) 5839. [9] A.G. Evans, M.Y. He, J.W. Hutchinson, Prog. Mater. Sci. 46 (2001) 249. [10] D.M. Lipkin, D.R. Clarke, Oxid. Met. 45 (1996) 267. [11] V.K. Tolpygo, D.R. Clarke, Oxid. Met. 49 (1998) 187. [12] C. Mennicke, E. Schumann, C. Ulrich, M. Ruehle, Mat. Sci. Forum 389 (1997) 251. [13] V. Sergo, D.R. Clarke, J. Am. Ceram. Soc. 81 (1998) 3237. [14] R.J. Christensen, V.K. Tolpygo, D.R. Clarke, Acta Mater. 45 (1997) 1761. [15] T.F. An, H.R. Guan, X.F. Sun, Z.Q. Hu, Oxid. Met. 54 (2000) 301. [16] M.W. Brumm, H.J. Grabke, Corros. Sci. 33 (1992) 1677. [17] C. Mennicke, D.R. Mumm, D.R. Clarke, Z. Metallkd. 90 (1999) 1079. [18] D.R. Clarke, V. Sergo, M.Y. He, in: J.M. Hampikian, N.B. Dahotre (Eds.), Elevated Temperature Coatings: Science and Technology, vol. III, TMS, Warrendale, PA, 1999, p. 67. [19] K. Gamo, Nucl. Instrum. Methods Phys. Res., B Beam Interact. Mater. Atoms 121 (1997) 464. [20] L.D. Xie, Y.H. Sohn, E.H. Jordan, M. Gell, Surf. Coat. Technol. 176 (2003) 57. [21] S. Sridharan, L.D. Xie, E.H. Jordan, M. Gell, Surf. Coat. Technol. 179 (2004) 286. [22] N. Czech, Surf. Coat. Technol. 108–109 (1998) 36. [23] B.A. Pint, Oxid. Met. 45 (1996) 1. [24] E.Y. Lee, “Life prediction and failure mechanisms for thermal barrier coatings”, Ph.D. dissertation, Worcester Polytechnic Institute (1991). [25] J.A. Haynes, E.D. Rigney, M.K. Ferber, W.D. Porter, Surf. Coat. Technol. 86–87 (1996) 102. [26] C.A. Calow, I.T. Porter, J. Mater. Sci. 6 (1971) 156. [27] R.G. Vardiman, Mater. Res. Bull. 7 (1972) 699. [28] A.G. Evans, D.R. Mumm, J.W. Hutchinson, G.H. Meier, F.S. Pettit, Prog. Mater. Sci. 46 (2001) 505. [29] D.R. Mumm, A.G. Evans, Acta Mater. 48 (2000) 1815. [30] I.G. Wright, B.A. Pint, First International Conference on Industrial Gas Turbine Technologies CAME-GT Brussels, 10–11 July, 2003.