Inﬂuence of Pt Addition and Manufacturing Process on the Failure Mechanisms of NiCoCrAlYTa-Base Thermal Barrier Coating Systems under Thermal Cycling Conditions

: The cyclic oxidation of NiCoCrAlYTa-base thermal barrier coating systems was investigated at 1100 ◦ C. The inﬂuence of the NiCoCrAlYTa deposition process, the coating modiﬁcation by a Pt-overlayer, and the surface preparation steps were studied. Thermal cycling results showed that the addition of a Pt-overlayer, a dense and oxide-free bond-coating microstructure, together with a smooth NiCoCrAlYTa surface prior to Pt deposition and a suitable surface preparation before thermal barrier deposition all increase the lifetime. Degradation mechanisms are proposed to explain how coating defects develop during thermal cycling and how the fabrication process inﬂuences both failure and lifetime. systems after thermal cycling and proposes degradation mechanisms. TBC system after 290 cycles. For the MCNG-Trib-Pt/2 TBC system, the bond-coating was single-phased γ -Ni, with a composition in at % equal to 49.7Ni-17.3Co-16.5Cr-9.9Al-3.4Pt-0.8Ta-0.4Ti-0.9W-0.4Mo-0.3Hf-0.3Y close to the surface, based on EDS analyses. γ ’-Ni 3 Al regions were only visible at the BC/superalloy interface. Most carbides had disappeared and the remaining ones were localized along the interface with the superalloy.


Introduction
Thermal barrier coating (TBC) systems have been developed to reduce the temperature of turbine blades. They consist of a Ni-base superalloy (the blade), an Al-rich coating (the bond-coating, BC) and an insulating top coat (TBC) made of yttria-stabilized zirconia (YSZ). For aeronautic applications, the top coat is usually deposited by electron-beam physical vapor deposition (EB-PVD) to obtain a columnar microstructure, making it capable of accommodating strain. Because YSZ is permeable to oxygen, it cannot protect the superalloy against oxidation. To ensure protection, the Al-rich bond-coating has to form a layer of α-Al 2 O 3 , a protective, adherent and slow-growing oxide, called thermally grown oxide (TGO).
In service, TBC systems undergo mechanical and environmental stresses under high temperatures. Whether it is due to foreign object damage [1,2], calcium-magnesium-aluminosilicate (CMAS) deposition [3,4] or thermo-mechanical damage of the TBC system [5], environmental stresses lead to damage and/or loss of the thermal barrier, which is harmful for the TBC system lifetime. A key point of TBC system thermo-mechanical damage is TGO formation and growth. Indeed, this oxide scale, forming between the bond-coating and the TBC, undergoes stresses resulting from: (i) the transformation from transient alumina to α-alumina; (ii) the oxide growth, and (iii) the thermal expansion coefficient (CTE) mismatch. While the transient alumina to α-alumina transformation leads to tensile stresses [6,7], the other two phenomena result in compressive stresses. Due to a great difference between the alumina volume and the volume of consumed metal, compressive stresses reaching 200 MPa at 1100 • C could be generated within the oxide scale [7]. Regarding

Materials and Methods
To study the effect of Pt addition on the degradation mechanisms of TBC systems with a NiCoCrAlYTa bond-coating, two samples were manufactured, one with and one without Pt. To understand the influence of the bond-coating fabrication process, NiCoCrAlYTa coatings were made either by composite electroplating (Tribomet ® process) or vacuum plasma spray (VPS).

Materials and Coatings
Three Ni-base superalloys were used for the fabrication of the TBC systems: the 1st generation superalloy AM3, the 2nd generation superalloy CMSX-4, and the 4th generation superalloy MCNG. CMSX-4 superalloy contains Re and some Hf, as MCNG superalloy which also contains Ru. Their compositions are given in Table 1. Three batches of NiCoCrAlYTa-base TBC systems were manufactured. The NiCoCrAlYTa coating was made using either the Tribomet ® process (Praxair ST, Oldmixon, UK) or vacuum plasma spraying (Surface Engineering Research Laboratory, LERMPS, Montbéliard, France), while Pt was deposited either by electroplating (Praxair ST, Indianapolis, IN, USA) or by sputtering (Cranfield University, Bedfordshire, England, UK). The thermal barrier was always deposited by EB-PVD but by various manufacturers: Praxair ST (Indianapolis, IN, USA), Ceramic Coating Center (Châtellerault, France), and Surface Engineering and Nanotechnology Institute (Cranfield University, England). All the details regarding the specimen geometry, the deposition techniques, the thicknesses, the surface preparations,  Table 2. The first batch was provided by Praxair ST and was composed of AM3-base TBC systems both with and without Pt. To coat the entire surface of the rectangular AM3 specimens with NiCoCrAlYTa coating, two passes were required, generating a locally thicker coating. In addition, a hole was drilled at one end of the AM3 substrate to coat the entire sample surface with the thermal barrier. The fabrication of the second batch, on AM3 and MCNG superalloys, involved Praxair ST for the bond-coating and Ceramic Coating Center for the thermal barrier. A stem made of Hastelloy W was welded to one edge of the rectangular substrate in order to hold the sample during the coating process. This allowed the deposition of a uniform coating. For batch 3, only one face of the substrate was coated. The manufacturing of the VPS NiCoCrAlYTa coating was performed at LERMPS using AMDRY 997 powder (wt %: Ni-23Co-20Cr-8.5Al-4Ta-0.6Y; at %: Ni-20.9Co-20.9Cr-16.9Al-1.2Ta-0.4Y). The Pt layer and the thermal barrier were deposited at the Surface Engineering and Nanotechnology Institute, Cranfield University. Table 2. Details on the geometry and manufacturing process of the three batches of thermal barrier coating (TBC) systems. 't' represents the thickness, 'r' the curvature radius of the edges. provided by Praxair ST and was composed of AM3-base TBC systems both with and without Pt. To coat the entire surface of the rectangular AM3 specimens with NiCoCrAlYTa coating, two passes were required, generating a locally thicker coating. In addition, a hole was drilled at one end of the AM3 substrate to coat the entire sample surface with the thermal barrier. The fabrication of the second batch, on AM3 and MCNG superalloys, involved Praxair ST for the bond-coating and Ceramic Coating Center for the thermal barrier. A stem made of Hastelloy W was welded to one edge of the rectangular substrate in order to hold the sample during the coating process. This allowed the deposition of a uniform coating. For batch 3, only one face of the substrate was coated. The manufacturing of the VPS NiCoCrAlYTa coating was performed at LERMPS using AMDRY 997 powder (wt %: Ni-23Co-20Cr-8.5Al-4Ta-0.6Y; at %: Ni-20.9Co-20.9Cr-16.9Al-1.2Ta-0.4Y). The Pt layer and the thermal barrier were deposited at the Surface Engineering and Nanotechnology Institute, Cranfield University.    Between batches 1 and 2, the grit blasting before EB-PVD and the thermal barrier coating deposition were not made by the same companies. However, the main difference was the addition of a smoothing step on the NiCoCrAlYTa surface before Pt deposition for batch 2. To reduce the high coating roughness resulting from the VPS process, the NiCoCrAlYTa coating from batch 3 was slightly ground using P1200 SiC abrasive discs. Grinding was stopped when the proportion of ground surface, estimated by image analysis on optical microscopy images, was close to that of a partially machined surface (≈24-34% of the entire surface) [33].

Oxidation Test
Thermal cycling tests under laboratory air were carried out on the TBC systems using a vertical cyclic oxidation rig (CIRIMAT, Toulouse, France). A cycle was composed of a 1 h dwell at 1100 • C (including fast heating) and a cooling of 15 min to ambient temperature. The fast cooling rate was obtained with a high flow of air, cleaned from oil and pollution, directed towards the samples from the four sides of the sample holder. Drilled plates, maintained on a ceramic rod, were used to position the specimens. Thanks to their holes, AM3-Trib and AM3-Trib-Pt/1 TBC systems were hung to the plates using Pt wires. Samples from batches 2 and 3 were placed on a sample-holder made of a porous ceramic brick, itself placed on a plate (AM3-Trib-Pt/2, MCNG-Trib-Pt/2, AM3-VPS-Pt, and CMSX-4-VPS-Pt TBC systems). The geometry of the sample-holder was such that the thermal barrier coated surfaces were well exposed to the air flow for cooling. Samples were removed from the furnace either when spallation spread across at least 25% of the thermal barrier coated surface (AM3-Trib, AM3-Trib-Pt/1, AM3-VPS-Pt, and CMSX-4-VPS-Pt TBC systems) or at a given number of cycles (AM3-Trib-Pt/2 and MCNG-Trib-Pt/2 TBC systems). The estimate of the percentage of spalled area was done visually.
In the discussion section, thermal cycling results are compared to isothermal oxidations performed on AM3-Trib-Pt/2 and MCNG-Trib-Pt/2 TBC systems. Details on these oxidation tests can be found in [23].

Characterization
Observations by scanning electron microscopy (SEM) of sample surfaces and polished cross-sections, using the secondary electron imaging mode (SE) or the backscattered imaging mode (BSE), were done using an LEO 435VP microscope (ZEISS, Cambridge, UK). Energy-dispersive X-ray spectroscopy (EDS) analysis was performed using a PGT IMIX-PC system (Princeton Gama Tech, Princeton, USA), with real standards for quantification. Some observations of the bond-coating surface were also made using a JEOL JSM 6700F (JEOL, Japan), using the SE mode. X-ray diffraction (XRD) analyses were done using a Seifert 3000TT apparatus (Rich Seifert and Co, Ahrensburg, Germany) with a copper anti-cathode (λ = 1.54056 Å) and a grazing incidence of 4 • (steps: 0.04 • , time per step: 6 s).
The lifetimes of AM3-Trib, AM3-Trib-Pt/1, AM3-VPS-Pt, and CMSX-4-VPS-Pt TBC systems have already been published [31]. The present study details the microstructural characterization of the systems after thermal cycling and proposes degradation mechanisms.

As-Processed Bond-Coatings and TBC Systems
An important characterization work on the effect of Pt on the microstructure of NiCoCrAlYTa coatings was made and discussed in a previous article [32], based on AM3-Trib-Pt/2 and AM3-VPS-Pt systems, as well as the AM3-Trib system from a batch different to the one of the present study. This previous work, combining XRD analyses, SEM observations, EDS analyses, and transmission electron microscopy, highlighted martensite formation and Al uphill diffusion with Pt addition. Indeed, while the Trib coating, on AM3 superalloy, contained nearly 19 at % in its upper part, the AM3-Trib-Pt/2 system exhibited more than 20 at % of Al below the surface, with a peak at 24 at %. Concerning the AM3-VPS-Pt system, an Al concentration between 31 and 34 at % was measured below the surface. This section summarizes the main results of this characterization work.
At ambient temperature, the Trib coating was mainly composed of a β-NiAl phase within a γ-Ni matrix [32,34,35]. In addition, the coating exhibited a few oxide precipitates and numerous Ta carbide (TaC) precipitates [32]. Close to the superalloy, TaC carbides also contained Ti, as this element diffused fast from the superalloy towards the bond-coating surface during high temperature exposures.
At ambient temperature, the Trib-Pt/2 bond-coating consisted of an outer part, rich in Pt, composed of L1 0 -NiAl martensite, γ'-Ni 3 Al, and γ-Ni, and an inner part made of γ-Ni, containing little Pt [32]. TaC precipitates were also present close to the interface with the superalloy and many oxide particles were dispersed in the inner zone of the coating. Small oxides were also located at the position of the original Pt/NiCoCrAlYTa interface. The oxide quantity was higher for AM3-Trib-Pt/2 and MCNG-Trib-Pt/2 TBC systems compared to AM3-Trib-Pt/1 TBC system, manufactured earlier.
Pores were also observed on the as-processed surface of the bond-coating of AM3-Trib-Pt/2 and MCNG-Trib-Pt/2 TBC systems, after heat treatment under vacuum (Figure 1a,b). These open pores are referred to as wells in the following.
At ambient temperature, a VPS NiCoCrAlYTa coating is composed of the β-NiAl, γ'-Ni 3 Al, γ-Ni phases, and TaC uniformly dispersed throughout the coating thickness [36][37][38]. The addition of Pt disrupted the homogeneous microstructure [32]. The outer part of the coating was composed of a Pt-rich martensite (L1 0 ) layer over a γ'-Ni 3 Al layer (a few γ'-Ni 3 Al grains were also observed within the martensite layer). The inner part of the coating was Pt-rich γ-Ni/γ'-Ni 3 Al, with the γ-Ni phase being the matrix. As for the Trib-Pt/2 coating, small oxides were located at the initial position of the Pt/NiCoCrAlYTa interface and open pores were visible on the bond-coating surface after the heat treatment under vacuum. However, no oxides were dispersed in the inner zone. Indeed, while the Trib coating, on AM3 superalloy, contained nearly 19 at % in its upper part, the AM3-Trib-Pt/2 system exhibited more than 20 at % of Al below the surface, with a peak at 24 at %. Concerning the AM3-VPS-Pt system, an Al concentration between 31 and 34 at % was measured below the surface. This section summarizes the main results of this characterization work. At ambient temperature, the Trib coating was mainly composed of a β-NiAl phase within a γ-Ni matrix [32,34,35]. In addition, the coating exhibited a few oxide precipitates and numerous Ta carbide (TaC) precipitates [32]. Close to the superalloy, TaC carbides also contained Ti, as this element diffused fast from the superalloy towards the bond-coating surface during high temperature exposures.
At ambient temperature, the Trib-Pt/2 bond-coating consisted of an outer part, rich in Pt, composed of L10-NiAl martensite, γ'-Ni3Al, and γ-Ni, and an inner part made of γ-Ni, containing little Pt [32]. TaC precipitates were also present close to the interface with the superalloy and many oxide particles were dispersed in the inner zone of the coating. Small oxides were also located at the position of the original Pt/NiCoCrAlYTa interface. The oxide quantity was higher for AM3-Trib-Pt/2 and MCNG-Trib-Pt/2 TBC systems compared to AM3-Trib-Pt/1 TBC system, manufactured earlier.
Pores were also observed on the as-processed surface of the bond-coating of AM3-Trib-Pt/2 and MCNG-Trib-Pt/2 TBC systems, after heat treatment under vacuum (Figure 1a,b). These open pores are referred to as wells in the following.
At ambient temperature, a VPS NiCoCrAlYTa coating is composed of the β-NiAl, γ'-Ni3Al, γ-Ni phases, and TaC uniformly dispersed throughout the coating thickness [36][37][38]. The addition of Pt disrupted the homogeneous microstructure [32]. The outer part of the coating was composed of a Pt-rich martensite (L10) layer over a γ'-Ni3Al layer (a few γ'-Ni3Al grains were also observed within the martensite layer). The inner part of the coating was Pt-rich γ-Ni/γ'-Ni3Al, with the γ-Ni phase being the matrix. As for the Trib-Pt/2 coating, small oxides were located at the initial position of the Pt/NiCoCrAlYTa interface and open pores were visible on the bond-coating surface after the heat treatment under vacuum. However, no oxides were dispersed in the inner zone. The decrease in roughness resulting from the smoothing step added in the fabrication process of the Trib-Pt/2 coating is presented in Figure 1c,e. The influence of the NiCoCrAlYTa deposition process (Tribomet ® or VPS) is illustrated in Figure 1d,f. The surface of the Trib-Pt/2 coating was smooth while the surface of the VPS-Pt coating exhibited many convex and concave areas. In concave areas, the YSZ columns were tilted in relation to the general column orientation. Cross-sections of failed AM3-Trib and AM3-Trib-Pt/1 TBC systems were observed by SEM ( Figure 3). While cracking occurred at the TGO/TBC interface with the Trib bond-coating, the thermal barrier was still adherent to the TGO when Pt was added, except in concave areas where debonding occurred in the YSZ (Figure 3a-c). The surface of the Trib-Pt/1 coating was undulated and numerous conical defects were visible in the TBC, in concave areas, Figure 3a,c. Such defects, called 'corn kernel' by Birks et al. [39] or 'pinch-off' by Mumm et al. [40], were also observed in a previous study [41] after thermal cycling of TBC systems composed of a VPS NiCoCrAlYTa bond-coating. According to EDS analyses, the TGO formed on the Trib coating was mainly composed of Al2O3 but also contained other oxides (including Y-Al-rich oxides), Figure 3d. On the contrary, only Al2O3 could be detected in the oxide layer that developed on the Trib-Pt/1 coating, Figure 3e.

Influence of Pt on the Cyclic Oxidation Resistance of TBC Systems
Because of Al consumption by oxidation, the β-NiAl phase of the Trib coating vanished, leading to a single-phased γ-Ni layer of at least 20 µm after only 36 cycles, Figure 3b. Also, the number of Ta carbides (white precipitates on BSE SEM pictures) dropped compared to the as-received coating and were mainly localized close to the superalloy (in the β-NiAl-rich zone). Ta-rich particles were also visible close to and within the TGO, Figure 3b,d. Wherever they were within the coating, the Ta-rich particles contained Ti after 36 cycles. After 290 cycles, the Trib-Pt/1 coating had evolved towards a nearly completely single-phased γ-Ni microstructure, Figure 3, with a composition in at % equal to 43.6Ni-18.1Co-20.8Cr-10.0Al-3.3Pt-0.9Ta-0.9Ti-0.8W-0.9Mo-0.7Y close to the surface based on EDS analyses. The γ'-Ni3Al phase remained in small proportions in a few regions below the surface and at the BC/superalloy interface due to interdiffusion. Wherever it was, the γ'-Ni3Al phase contained Ti (2.4 at % below the bond-coating surface, 2.7 at % at the interface with the superalloy, based on EDS analyses). Like the Trib, only a few carbides remained along the BC/superalloy interface of the AM3-Trib-Pt/1 TBC system, Figure 3c. The decrease in roughness resulting from the smoothing step added in the fabrication process of the Trib-Pt/2 coating is presented in Figure 1c,e. The influence of the NiCoCrAlYTa deposition process (Tribomet ® or VPS) is illustrated in Figure 1d,f. The surface of the Trib-Pt/2 coating was smooth while the surface of the VPS-Pt coating exhibited many convex and concave areas. In concave areas, the YSZ columns were tilted in relation to the general column orientation. Cross-sections of failed AM3-Trib and AM3-Trib-Pt/1 TBC systems were observed by SEM ( Figure 3). While cracking occurred at the TGO/TBC interface with the Trib bond-coating, the thermal barrier was still adherent to the TGO when Pt was added, except in concave areas where debonding occurred in the YSZ (Figure 3a-c). The surface of the Trib-Pt/1 coating was undulated and numerous conical defects were visible in the TBC, in concave areas, Figure 3a,c. Such defects, called 'corn kernel' by Birks et al. [39] or 'pinch-off' by Mumm et al. [40], were also observed in a previous study [41] after thermal cycling of TBC systems composed of a VPS NiCoCrAlYTa bond-coating. According to EDS analyses, the TGO formed on the Trib coating was mainly composed of Al 2 O 3 but also contained other oxides (including Y-Al-rich oxides), Figure 3d. On the contrary, only Al 2 O 3 could be detected in the oxide layer that developed on the Trib-Pt/1 coating, Figure 3e.

Influence of Pt on the Cyclic Oxidation Resistance of TBC Systems
Because of Al consumption by oxidation, the β-NiAl phase of the Trib coating vanished, leading to a single-phased γ-Ni layer of at least 20 µm after only 36 cycles, Figure 3b. Also, the number of Ta carbides (white precipitates on BSE SEM pictures) dropped compared to the as-received coating and were mainly localized close to the superalloy (in the β-NiAl-rich zone). Ta-rich particles were also visible close to and within the TGO, Figure 3b,d. Wherever they were within the coating, the Ta-rich particles contained Ti after 36 cycles. After 290 cycles, the Trib-Pt/1 coating had evolved towards a nearly completely single-phased γ-Ni microstructure, Figure 3, with a composition in at % equal to 43.6Ni-18.1Co-20.8Cr-10.0Al-3.3Pt-0.9Ta-0.9Ti-0.8W-0.9Mo-0.7Y close to the surface based on EDS analyses. The γ'-Ni 3 Al phase remained in small proportions in a few regions below the surface and at the BC/superalloy interface due to interdiffusion. Wherever it was, the γ'-Ni 3 Al phase contained Ti (2.4 at % below the bond-coating surface, 2.7 at % at the interface with the superalloy, based on EDS analyses). Like the Trib, only a few carbides remained along the BC/superalloy interface of the AM3-Trib-Pt/1 TBC system, Figure 3c.

Degradation of TBC Systems Containing a Trib-Pt/2 Coating
During the thermal cycling of TBC systems containing a Trib-Pt/2 coating, an unusual degradation was observed at one end of the AM3-base system. Therefore, this section is divided into two parts to distinguish the regular behavior of the TBC systems from the catastrophic one.

Healthy Area
After 300 cycles at 1100 °C, AM3-Trib-Pt/2 and MCNG-Trib-Pt/2 TBC systems exhibited spallation, Figure 2. The spalled area proportion was much larger when AM3 was the superalloy. For both systems, SEM observations on spalled areas revealed oxidized bond-coating surface with TGO flakes, Figure 4. These TGO flakes exhibited thermal barrier remains that did not look like corn kernels but rather like the broken base of YSZ columns. Oxidized wells were also visible on the bondcoating surface of the MCNG-Trib-Pt/2 TBC system, Figure 4b,c. However, cross-section observations indicated that the TGO layer was still adherent to the thermal barrier after spallation, Figure 5a,b. After 300 cycles of 1 h at 1100 °C, the bond-coating microstructure of the AM3-Trib-Pt/2 TBC system was mainly composed of the γ-Ni phase, with γ'-Ni3Al phase regions present below the surface and at the BC/superalloy interface, Figure 5c,d. The composition, in at %, of the γ-Ni and γ'-Ni3Al phases close to the surface were 44.4Ni-18.8Co-20.6Cr-9.2Al-4.2Pt-0.2Ta-1.0Ti-0.8W-0.9Mo and 52.4Ni-10.3Co-5.9Cr-16.2Al-7.4Pt-2.2Ta-3.2Ti-0.9W-0.7Mo-0.8Y respectively, based on EDS analyses. A few γ'-Ni3Al regions were also observed within the internal coating zone with no TaC carbides, which was not the case for the AM3-Trib-Pt/1 TBC system after 290 cycles. For the MCNG-Trib-Pt/2 TBC system, the bond-coating was single-phased γ-Ni, with a composition in at % equal to 49.7Ni-17.3Co-16.5Cr-9.9Al-3.4Pt-0.8Ta-0.4Ti-0.9W-0.4Mo-0.3Hf-0.3Y close to the surface, based on EDS analyses. γ'-Ni3Al regions were only visible at the BC/superalloy interface. Most carbides had disappeared and the remaining ones were localized along the interface with the superalloy.

Degradation of TBC Systems Containing a Trib-Pt/2 Coating
During the thermal cycling of TBC systems containing a Trib-Pt/2 coating, an unusual degradation was observed at one end of the AM3-base system. Therefore, this section is divided into two parts to distinguish the regular behavior of the TBC systems from the catastrophic one.

Healthy Area
After 300 cycles at 1100 • C, AM3-Trib-Pt/2 and MCNG-Trib-Pt/2 TBC systems exhibited spallation, Figure 2. The spalled area proportion was much larger when AM3 was the superalloy. For both systems, SEM observations on spalled areas revealed oxidized bond-coating surface with TGO flakes, Figure 4. These TGO flakes exhibited thermal barrier remains that did not look like corn kernels but rather like the broken base of YSZ columns. Oxidized wells were also visible on the bond-coating surface of the MCNG-Trib-Pt/2 TBC system, Figure 4b,c. However, cross-section observations indicated that the TGO layer was still adherent to the thermal barrier after spallation, Figure 5a,b. After 300 cycles of 1 h at 1100 • C, the bond-coating microstructure of the AM3-Trib-Pt/2 TBC system was mainly composed of the γ-Ni phase, with γ'-Ni 3 Al phase regions present below the surface and at the BC/superalloy interface, Figure 5c,d. The composition, in at %, of the γ-Ni and γ'-Ni 3 Al phases close to the surface were 44.4Ni-18.8Co-20.6Cr-9.2Al-4.2Pt-0.2Ta-1.0Ti-0.8W-0.9Mo and 52.4Ni-10.3Co-5.9Cr-16.2Al-7.4Pt-2.2Ta-3.2Ti-0.9W-0.7Mo-0.8Y respectively, based on EDS analyses. A few γ'-Ni 3 Al regions were also observed within the internal coating zone with no TaC carbides, which was not the case for the AM3-Trib-Pt/1 TBC system after 290 cycles. For the MCNG-Trib-Pt/2 TBC system, the bond-coating was single-phased γ-Ni, with a composition in at % equal to 49.7Ni-17.3Co-16.5Cr-9.9Al-3.4Pt-0.8Ta-0.4Ti-0.9W-0.4Mo-0.3Hf-0.3Y close to the surface, based on EDS analyses. γ'-Ni 3 Al regions were only visible at the BC/superalloy interface. Most carbides had disappeared and the remaining ones were localized along the interface with the superalloy.

Area of Catastrophic Degradation
After only 300 cycles, an unusual degradation was observed on the bond-coating surface of the AM3-Trib-Pt/2 TBC system, as pointed out by the arrows on Figure 2b,c. Next, SEM observations of the bond-coating surface and cross-section were done on this specific area, Figure 6. On the surface, particles, either Pt-rich or Ta-rich (containing also some Ti for the latter), were dispersed in (Ni, Co, Cr, Al)-rich oxides. An XRD analysis done on this specific area confirmed the presence of other oxides than alumina (like NiO, NiCr 2 O 4, and TiTaO 4 ) and revealed that the Pt-rich particles were pure Pt, Figure 7. Cross-section observations showed that this unusual degradation concerned the entire circumference of one sample end, Figure 6c. (Ni, Co, Cr, Al)-rich oxides containing Pt and Ta-rich particles did not spread over the entire surface, Figure 6d, while an Al 2 O 3 layer dividing the bond-coating into two zones, an outer zone and an inner one, was present along the entire sample circumference, Figure 6c. No Al was detected by EDS in the region situated between the external oxide scale and the inner Al 2 O 3 layer. Such catastrophic degradation was not observed for the MCNG-Trib-Pt/2 TBC system despite oxide penetrating within the bond-coating in a few areas, Figure 8.

Area of Catastrophic Degradation
After only 300 cycles, an unusual degradation was observed on the bond-coating surface of the AM3-Trib-Pt/2 TBC system, as pointed out by the arrows on Figure 2b,c. Next, SEM observations of the bond-coating surface and cross-section were done on this specific area, Figure 6. On the surface, particles, either Pt-rich or Ta-rich (containing also some Ti for the latter), were dispersed in (Ni, Co, Cr, Al)-rich oxides. An XRD analysis done on this specific area confirmed the presence of other oxides than alumina (like NiO, NiCr2O4, and TiTaO4) and revealed that the Pt-rich particles were pure Pt, Figure 7. Cross-section observations showed that this unusual degradation concerned the entire circumference of one sample end, Figure 6c. (Ni, Co, Cr, Al)-rich oxides containing Pt and Ta-rich particles did not spread over the entire surface, Figure 6d, while an Al2O3 layer dividing the bondcoating into two zones, an outer zone and an inner one, was present along the entire sample circumference, Figure 6c. No Al was detected by EDS in the region situated between the external oxide scale and the inner Al2O3 layer. Such catastrophic degradation was not observed for the MCNG-Trib-Pt/2 TBC system despite oxide penetrating within the bond-coating in a few areas, Figure 8.

Degradation of TBC Systems with a VPS-Pt Bond-Coating
The lifetimes of AM3-VPS-Pt and CMSX-4-VPS-Pt TBC systems are presented in Figure 2. The average lifetime of the AM3-base systems was shorter than the average lifetime of the CMSX-4-base systems.
After thermal cycling, bond-coating surfaces and cross-sections were observed by SEM, Figure  9. The features are shown for the AM3-VPS-Pt TBC system that lasted 92 cycles but they were identical whatever the superalloy. First, many corn kernels were visible on the bond-coating surface after spallation, Figure 9a,b,d. Second, the oxide scale was an almost exclusive Al2O3 layer but contained Ti, according to EDS analyses, and it exhibited cavities, Figure 9b. The bond-coating surface was deformed, Figure 9b-d, and some oxide penetrations within the bond-coating were visible, as for the CMSX-4-VPS-Pt TBC system that lasted 199 cycles, Figure 9d. The bond-coating was no longer composed of the martensitic phase, Figure 9c,d. A thick discontinuous γ'-Ni3Al layer, Ti-and Ta-rich, was present close to the surface and in the interdiffusion zone. The composition of this γ'-Ni3Al layer was, in at %, equal to 49.2Ni-11.3Co-6.1Cr-17.9Al-8.1Pt-3.5Ta-1.8Ti-0.9W-0.5Mo-0.7Y and 49.7Ni-11.7Co-5.4Cr-16.4Al-8.4Pt-4.2Ta-1.2Ti-1.2W-0.1Mo-0.2Re-0.3Hf-1.2Y for AM3-base and CMSX-4-base systems respectively, based on EDS analyses. The inner part of the coating was composed of γ-Ni. A

Degradation of TBC Systems with a VPS-Pt Bond-Coating
The lifetimes of AM3-VPS-Pt and CMSX-4-VPS-Pt TBC systems are presented in Figure 2. The average lifetime of the AM3-base systems was shorter than the average lifetime of the CMSX-4-base systems.
After thermal cycling, bond-coating surfaces and cross-sections were observed by SEM, Figure  9. The features are shown for the AM3-VPS-Pt TBC system that lasted 92 cycles but they were identical whatever the superalloy. First, many corn kernels were visible on the bond-coating surface after spallation, Figure 9a,b,d. Second, the oxide scale was an almost exclusive Al2O3 layer but contained Ti, according to EDS analyses, and it exhibited cavities, Figure 9b. The bond-coating surface was deformed, Figure 9b-d, and some oxide penetrations within the bond-coating were visible, as for the CMSX-4-VPS-Pt TBC system that lasted 199 cycles, Figure 9d. The bond-coating was no longer composed of the martensitic phase, Figure 9c,d. A thick discontinuous γ'-Ni3Al layer, Ti-and Ta-rich, was present close to the surface and in the interdiffusion zone. The composition of this γ'-Ni3Al layer was, in at %, equal to 49.2Ni-11.3Co-6.1Cr-17.9Al-8.1Pt-3.5Ta-1.8Ti-0.9W-0.5Mo-0.7Y and 49.7Ni-11.7Co-5.4Cr-16.4Al-8.4Pt-4.2Ta-1.2Ti-1.2W-0.1Mo-0.2Re-0.3Hf-1.2Y for AM3-base and CMSX-4-base systems respectively, based on EDS analyses. The inner part of the coating was composed of γ-Ni. A

Degradation of TBC Systems with a VPS-Pt Bond-Coating
The lifetimes of AM3-VPS-Pt and CMSX-4-VPS-Pt TBC systems are presented in Figure 2. The average lifetime of the AM3-base systems was shorter than the average lifetime of the CMSX-4-base systems.
After thermal cycling, bond-coating surfaces and cross-sections were observed by SEM, Figure 9. The features are shown for the AM3-VPS-Pt TBC system that lasted 92 cycles but they were identical whatever the superalloy. First, many corn kernels were visible on the bond-coating surface after spallation, Figure 9a,b,d. Second, the oxide scale was an almost exclusive Al 2 O 3 layer but contained Ti, according to EDS analyses, and it exhibited cavities, Figure 9b. The bond-coating surface was deformed, Figure 9b-d, and some oxide penetrations within the bond-coating were visible, as for the CMSX-4-VPS-Pt TBC system that lasted 199 cycles, Figure 9d. The bond-coating was no longer composed of the martensitic phase, Figure 9c,d. A thick discontinuous γ'-Ni 3 Al layer, Ti-and Ta-rich, was present close to the surface and in the interdiffusion zone. The composition of this γ'-Ni 3 Al layer was, in at %, equal to 49.2Ni-11.3Co-6.1Cr-17.9Al-8.1Pt-3.5Ta-1.8Ti-0.9W-0.5Mo-0.7Y and 49.7Ni-11.7Co-5.4Cr-16.4Al-8.4Pt-4.2Ta-1.2Ti-1.2W-0.1Mo-0.2Re-0.3Hf-1.2Y for AM3-base and CMSX-4-base systems respectively, based on EDS analyses. The inner part of the coating was composed of γ-Ni. A few isolated γ'-Ni 3 Al grains were also visible within the inner part, below the discontinuous γ'-Ni 3 Al layer, as few isolated γ-Ni grains were present above the thick γ'-Ni 3 Al layer. few isolated γ'-Ni3Al grains were also visible within the inner part, below the discontinuous γ'-Ni3Al layer, as few isolated γ-Ni grains were present above the thick γ'-Ni3Al layer.

Pt Effect on the Cyclic Oxidation Resistance of TBC Systems with a Trib Bond-Coating
AM3-Trib and AM3-Trib-Pt/1 TBC systems were thermally cycled at 1100 °C (batch 1). A large increase in lifetime was obtained with Pt addition (from 36 to 290 cycles). Even though only one sample per system was tested, this confirms the beneficial effect of Pt on the oxidation resistance of TBC systems. However, 290 cycles is a short lifetime for turbine blade applications, even for a system based on a 1st generation superalloy. Therefore, it is necessary to understand the reasons for these short lifetimes.
For the AM3-Trib TBC system, crack propagation occurred at the TGO/TBC interface and the TGO layer was mainly composed of Al2O3, but contained other oxides. The presence of oxides other than Al2O3 on top of the TGO was responsible for the weak adherence of the thermal barrier to the TGO layer and explains the very short lifetime of the system. Y-Al-rich oxides and Ta-Ti-rich particles were also present on the TGO layer. During bond-coating oxidation, oxides of reactive elements, such as Hf or Y, can form and lead to peg formation. Pegs act as mechanical anchorage points and favor the TGO adherence to the bond-coating [42][43][44][45][46]. However, peg formation leads to local deformations of the BC/TGO interface, an increase in TGO thickness, and heterogeneities in TGO composition that result in stress concentrations during thermal cycling. Therefore, a high peg density causes a shorter lifetime [17,[47][48][49][50]. The presence of Y-Al-rich oxides within the TGO after only 36 cycles suggests that the Y content of the bond-coating is too high. Concerning Ti, most authors agree that it is harmful with regard to oxidation [34,35,[51][52][53][54][55][56]. One way to reduce its detrimental effect on oxidation consists in adding Ta to the NiCoCrAlY because Ta traps Ti and C diffusing from the superalloy towards the oxide scale by forming (Ta, Ti) C carbides [9,18,57,58]. Therefore, Ta decreases the superalloy effect on the oxidation behavior [34] since Ti can no longer diffuse towards the TGO. Regarding the AM3-Trib-Pt/1 system, the TBC was still adherent to the TGO after 290 cycles and only Al2O3 could be detected in the oxide scale. This last observation confirms that Pt favors Al selective oxidation, as

Pt Effect on the Cyclic Oxidation Resistance of TBC Systems with a Trib Bond-Coating
AM3-Trib and AM3-Trib-Pt/1 TBC systems were thermally cycled at 1100 • C (batch 1). A large increase in lifetime was obtained with Pt addition (from 36 to 290 cycles). Even though only one sample per system was tested, this confirms the beneficial effect of Pt on the oxidation resistance of TBC systems. However, 290 cycles is a short lifetime for turbine blade applications, even for a system based on a 1st generation superalloy. Therefore, it is necessary to understand the reasons for these short lifetimes.
For the AM3-Trib TBC system, crack propagation occurred at the TGO/TBC interface and the TGO layer was mainly composed of Al 2 O 3 , but contained other oxides. The presence of oxides other than Al 2 O 3 on top of the TGO was responsible for the weak adherence of the thermal barrier to the TGO layer and explains the very short lifetime of the system. Y-Al-rich oxides and Ta-Ti-rich particles were also present on the TGO layer. During bond-coating oxidation, oxides of reactive elements, such as Hf or Y, can form and lead to peg formation. Pegs act as mechanical anchorage points and favor the TGO adherence to the bond-coating [42][43][44][45][46]. However, peg formation leads to local deformations of the BC/TGO interface, an increase in TGO thickness, and heterogeneities in TGO composition that result in stress concentrations during thermal cycling. Therefore, a high peg density causes a shorter lifetime [17,[47][48][49][50]. The presence of Y-Al-rich oxides within the TGO after only 36 cycles suggests that the Y content of the bond-coating is too high. Concerning Ti, most authors agree that it is harmful with regard to oxidation [34,35,[51][52][53][54][55][56]. One way to reduce its detrimental effect on oxidation consists in adding Ta to the NiCoCrAlY because Ta traps Ti and C diffusing from the superalloy towards the oxide scale by forming (Ta, Ti) C carbides [9,18,57,58]. Therefore, Ta decreases the superalloy effect on the oxidation behavior [34] since Ti can no longer diffuse towards the TGO. Regarding the AM3-Trib-Pt/1 system, the TBC was still adherent to the TGO after 290 cycles and only Al 2 O 3 could be detected in the oxide scale. This last observation confirms that Pt favors Al selective oxidation, as demonstrated in one of our previous studies [23] and as already observed by others [19,20]. This is explained by the decrease in Al activity due to the presence of Pt, as shown by others [59,60], which led to an Al-uphill diffusion during the vacuum heat treatment. The AM3-Trib-Pt/1 TBC system failure was not due to heterogeneity of the TGO in terms of composition and thickness. Rather it was due to the presence of numerous corn kernel defects, characteristic of a surface exhibiting concave areas before thermal barrier deposition by EB-PVD. When the TGO/TBC interface is strong, the thermal barrier prevents the development of bond-coating surface undulations [20,[61][62][63]. Because corn kernel defects are lightly bonded to the thermal barrier coating, the bond-coating surface deformation is easier in these specific areas. Furthermore, TGO undulations in the corn kernel areas lead to stress concentration and to fracture, according to Evans et al. [5]. In the present study, the presence of many corn kernels shows that the Pt electroplating process did not smooth the rough NiCoCrAlYTa bond-coating surface.
By forming an exclusive Al 2 O 3 layer, the Trib-Pt coating was closer to Pt-modified nickel aluminide coating whose degradation is mainly due to rumpling. Various explanations have been given for rumpling development, among them the martensite transformation of β-NiAl [64] coupled with coating creep [65]. With an outer zone rich in martensite, the Trib-Pt coating is more affected by rumpling than the NiCoCrAlYTa. This could constitute a drawback for such a coating, but this needs to be confirmed for longer lifetimes. Indeed, this may depend on the duration of the martensite presence below the oxide scale.

Failure Mechanism of AM3-Trib-Pt/2 and MCNG-Trib-Pt/2 TBC Systems
SEM observations of the Trib-Pt/2 bond-coatings after 300 cycles revealed that spalled areas were oxidized (Figure 4) while the TGO layer was still adherent to the thermal barrier (Figure 5a,b). As spallation occurred before 300 cycles, cracking probably happened along the BC/TGO interface and cycling continuation led to the bond-coating surface oxidation under the detached TGO. For these TBC systems, an additional surface preparation step was added to smooth the Trib coating surface before Pt deposition. For both systems, no corn kernel defect was visible after spallation. This highlights the relevance of smoothing the bond-coating surface before EB-PVD in order to minimize thermal barrier defects. However, despite this additional surface preparation step, the AM3-Trib-Pt/2 TBC system lifetime was not improved when compared to the one of AM3-Trib-Pt/1 TBC system. This is likely due to the catastrophic degradation observed on the AM3-Trib-Pt/2 TBC system.
After heat treatment under vacuum, the surface of Trib-Pt/2 bond-coatings exhibited a deep open porosity, for both superalloys (AM3 and MCNG). In addition, these bond-coatings contained numerous oxides in the inner zone [23]. This may mean that oxygen was trapped during coating fabrication because of CrAlYTa particle oxidation. Thus, during heat treatment under vacuum, oxides could limit interdiffusion between the CrAlYTa particles and the (Ni, Co) matrix, leading to oxidized and not very dense areas (at the former CrAlYTa/(Ni, Co) interfaces) and therefore creating a porosity network. This porosity network and the wells play a major role in the proposed failure mechanism, Figure 10. During the thermal cycling test under laboratory air, the bond-coating surface (including the planar areas and the wells) oxidized to form an Al 2 O 3 layer. The volume expansion (due to the formation of intergranular Al 2 O 3 ) added to the large thermal stresses generated during the rapid coolings and heatings of the thermal cycling test (due to coefficient of thermal expansion mismatch [5]), favored bond-coating micro-cracking at the well points. Because of micro-cracking, the oxide progressively penetrated within the bond-coating during cycling. When the oxide layer developed in these wells reached the former CrAlYTa/(Ni, Co) interfaces, its propagation parallel to the coating surface was favored by the oxides already there. Its propagation was also mainly favored by the lack of coating cohesion at these particular zones. Finally, intergranular oxidation progressed along this porosity network within the bond-coating. As Al was consumed at the surface but also at the wells and at the CrAlYTa/(Ni, Co) interfaces, it was no longer available to ensure the growth of an external protective Al 2 O 3 scale. This was made even worse by the Al 2 O 3 that developed internally, parallel to the bond-coating surface, blocking the Al up-hill diffusion towards the external surface. When the Al content and the Al flux below the TGO became inferior to the critical values required for Al 2 O 3 formation, other oxides formed. These oxides, rich in Ni, Co, Cr, and Al, developed during this breakaway oxidation stage and led to the system failure. This proposed degradation mechanism is therefore a mix between the Intrinsic Chemical Failure and the Mechanically Induced Failure [66]. catastrophic degradation. Sulfur, contained in the superalloys, is well-known for its detrimental effect on TGO adherence [67][68][69][70][71][72]. In the present study, the difference in lifetime is unlikely to be due to the sulfur content of the superalloy as both superalloys contained less than 0.5 ppmw (concentrations determined by glow discharge mass spectroscopy, GDMS). The only known differences are the Pt quantity [32] and the oxidation rate constant [23] of the coating as well as the superalloy Ti content, all higher for the AM3-base system, and the Hf doping of MCNG superalloy. A greater Pt content is generally thought to be beneficial. Nevertheless, a higher Pt quantity led to a larger martensite quantity in the external layer, as observed on cross-sections. As mentioned previously, the martensitic transformation is known to favor rumpling apparition in nickel aluminide coatings. The presence of martensite may well favor propagation within the bond-coating and therefore catastrophic degradation. Also, a higher oxidation rate and the absence of Hf in the AM3-Trib-Pt/2 TBC system could have led to larger stresses within the wells and easier metal/Al2O3 interfacial crack propagation.  Although the MCNG-Trib-Pt/2 and the AM3-Trib-Pt/2 TBC systems were fabricated at the same time and tested for the same number of cycles, the MCNG-base TBC system did not exhibit catastrophic degradation. Sulfur, contained in the superalloys, is well-known for its detrimental effect on TGO adherence [67][68][69][70][71][72]. In the present study, the difference in lifetime is unlikely to be due to the sulfur content of the superalloy as both superalloys contained less than 0.5 ppmw (concentrations determined by glow discharge mass spectroscopy, GDMS). The only known differences are the Pt quantity [32] and the oxidation rate constant [23] of the coating as well as the superalloy Ti content, all higher for the AM3-base system, and the Hf doping of MCNG superalloy. A greater Pt content is generally thought to be beneficial. Nevertheless, a higher Pt quantity led to a larger martensite quantity in the external layer, as observed on cross-sections. As mentioned previously, the martensitic transformation is known to favor rumpling apparition in nickel aluminide coatings. The presence of martensite may well favor propagation within the bond-coating and therefore catastrophic degradation. Also, a higher oxidation rate and the absence of Hf in the AM3-Trib-Pt/2 TBC system could have led to larger stresses within the wells and easier metal/Al 2 O 3 interfacial crack propagation.
Even if no catastrophic degradation was observed for the MCNG-Trib-Pt/2 TBC system, oxidation penetration was encountered locally. This appeared similar to the onset of the catastrophic degradation that started on the edge for the AM3-Trib-Pt/2 TBC system. The higher surface/volume ratio at edges and corners explains the faster consumption of the Al reservoir contained within the material. Thermo-mechanical stresses are also higher and the quality of surface preparation is lower in these regions. All these reasons explain why degradation is greater at the sample edges. An increase in degradation in areas with a low curvature radius has previously been observed [73][74][75].
Catastrophic degradation occurred for the AM3-Trib-Pt/2 TBC system after 300 cycles at 1100 • C while this was not encountered for the AM3-Trib-Pt/1 TBC system that lasted 290 cycles at the same temperature. As explained previously, these TBC systems were not fabricated at the same time. The AM3-Trib-Pt/1 TBC system had a hole and a bond-coating over-thickness, while the AM3-Trib-Pt/2 TBC system had a welded stem. The thermal barrier thickness also differed between these batches. Due to these differences, the TBC system lifetimes must be compared with caution. Although the inner part of the Trib-Pt/1 bond-coating contained oxides, the oxide quantity was much lower than the one of the Trib-Pt/2 bond-coating. This prevented catastrophic degradation for the AM3-Trib-Pt/1 TBC system. Such an issue has already been mentioned by Subanovic et al. [76]. In their study, they showed that when the oxygen partial pressure within the VPS chamber was high enough, oxides rich in reactive elements were present through VPS NiCoCrAl(Y/Hf) coatings that exhibited poor oxidation behavior. Two other differences between these systems have to be mentioned: the surface preparation before Pt electroplating and before thermal barrier deposition. While for the AM3-Trib-Pt/1 TBC systems the bond-coating surface preparation was made at Praxair ST (the Tribomet ® NiCoCrAlYTa coating manufacturer, Oldmixon, UK), the bond-coating surface of the AM3-Trib-Pt/2 TBC system was grit-blasted at Safran Aircraft Engines (Châtellerault, France), according to parameters optimized for Pt-modified nickel aluminide coatings. It is then possible that the grit blasting carried out at Safran Aircraft Engines was insufficient to close the wells present at the Trib-Pt/2 surface. On the contrary, a surface preparation suitable for Trib coatings closed this open porosity and therefore prevented catastrophic degradation, Figure 10.
Finally, when the AM3-Trib-Pt/2 TBC system was oxidized isothermally [23], no thermal stress developed. Therefore, no micro-crack formed, oxide did not penetrate through the bond-coating depth, and no catastrophic degradation was observed.

Failure Mechanism of TBC Systems with a VPS-Pt Bond-Coating
Despite a significant difference in lifetime between AM3-VPS-Pt and CMSX-4-VPS-Pt TBC systems, the uncertainty domains had to be taken into account. The lifetimes of both TBC systems tend to be different, but additional data would be necessary to confirm. Nevertheless, from the present data, it is reasonable to say that the superalloy may influence the cyclic oxidation behavior of TBC systems with a VPS-Pt coating.
The S content of AM3 and CMSX-4 superalloys was measured by GDMS; 0.28 ppmw in AM3 against 1.4 ppmw in CMSX-4. The higher S content should have accelerated the degradation of the CMSX-4-VPS-Pt TBC systems but thermal barrier spallation occurred more rapidly with AM3 superalloy. Hf, contained in CMSX-4, is a reactive element known to mitigate the detrimental effect of S [67,[77][78][79]. Therefore, the presence of Hf compensated the higher S content of CMSX-4 and the difference in TBC system lifetime could not be explained by the difference of S concentration. In a previous study, a Trib-Pt/2 coating deposited on MCNG, Hf-rich, exhibited a slightly lower oxidation rate than the same coating on AM3, which is Hf-free [23]. Thus, the presence of Hf in CMSX-4 and its possible effects on the oxidation rate and oxide scale adherence could have been beneficial for the TBC system lifetime. However, further analyses are necessary to understand why the CMSX-4-based TBC system had a better behavior.
In both systems, only Al 2 O 3 could be detected in the oxide scale that also contained some Ti, as shown in [31]. This confirms the beneficial effect of Pt on Al-selective oxidation and the rapid Ti diffusion from the superalloy toward the oxide scale. Cavities were also observed within the TGO, which was not the case for the TBC systems with a Trib-Pt coating. Parameters used during Pt deposition by sputtering influence the microstructure and the degradation of Pt-modified nickel aluminide coatings. Non optimized sputtering parameters could lead to cavities within the TGO [80].
Numerous corn kernel defects were visible at the bond-coating surface after spallation, proof that the bond-coating surface was too rough before EB-PVD. It can be concluded that the light grinding made after VPS did not lead to a surface finish suitable for EB-PVD. However, a complete grinding of the VPS NiCoCrAlYTa was not recommended before sputtering a 7 µm Pt layer since thick layers may come off from a smooth substrate during deposition or may generate cavities during diffusion heat treatment.
After vacuum heat treatment, deep concave zones were visible at the surface of VPS-Pt coatings. After thermal cycling, their surface was greatly deformed and at some regions oxide penetrated within the bond-coating and could even reach the original Pt/NiCoCrAlYTa interface. Like for the cavities present within the TGO, the large bond-coating deformation could be due to the sputtering process. It could also be due to the martensitic transformation of β-NiAl. Complementary tests would be necessary to better understand this.
Despite deep oxide penetration, up to the former Pt/NiCoCrAlYTa interface, oxidation did not propagate parallel to the bond-coating surface. This can be explained by a lower number of cycles but it may also be due to the absence of oxide and porosity in the inner zone of VPS bond-coatings after fabrication. Even if the oxidized wells penetrated deep inside the bond-coating during thermal cycling, they could not follow any porosity network and therefore did not propagate parallel to the surface, Figure 10.
It is interesting to point out that Jackson et al. [22] noticed deep oxide protrusions in NiCoCrAlY coatings deposited by argon-shrouded plasma spraying after 2067 cycles at 1100 • C (cycles with a 45 min hot dwell). This was observed only in the case of high Y-content coatings (0.5 wt % versus 0.1 wt %) with a rough surface. They proposed that the protrusions were initiated by preferential oxidation of Y in concave regions of the coating surface. They did not claim any mechanism relative to the protrusion propagation due to a lack of data, but mentioned the possibility of cracking.

Conclusions
In this study, many TBC systems that differ by fabrication process (Pt addition, deposition techniques, surface preparation steps) were oxidized. The results confirm that Pt is beneficial in terms of TBC system lifetime under thermal cycling conditions as it favors Al selective oxidation on a Trib coating, by martensitic (Ni, Pt) Al formation. On the other hand, Pt seems to destabilize TaC carbides that trap Ti. The beneficial effect of Pt on the cyclic oxidation resistance of Trib-base TBC systems is then mitigated when used in systems based on Ti-rich superalloys. Moreover, martensite formation led to stress development in the coating.
While the short lifetime of AM3-Trib-Pt/1 TBC systems was attributed to numerous corn kernel defects because of a rough bond-coating surface before EB-PVD, AM3-Trib-Pt/2 and MCNG-Trib-Pt/2, TBC systems failed because the Trib coating contained numerous oxide precipitates and was not dense enough. Concerning AM3-VPS-Pt and CMSX-VPS-Pt TBC systems their short lifetime was due to non-optimized sputtering parameters and a great density of corn kernel defects. This highlights the relevance of smoothing the NiCoCrAlYTa surface before Pt deposition to obtain, after heat treatment under vacuum, a bond-coating roughness suitable for EB-PVD.
Finally, the present work shows that the TBC system lifetime is increased when: (i) the NiCoCrAlYTa coating is dense, well-interdiffused and oxide-free; (ii) the NiCoCrAlYTa surface is smoothed before Pt deposition; and (iii) defects on the bond-coating surface are removed by a suitable surface preparation before EB-PVD.