Investigation of Thermal Shock Behavior of Multilayer Thermal Barrier Coatings with Superior Erosion Resistance Prepared by Atmospheric Plasma Spraying

Gadolinium zirconate (GZ) has become a promising thermal barrier coating (TBC) candidate material for high-temperature applications because of its excellent high-temperature phase stability and low thermal conductivity compared to yttria-stabilized zirconia (YSZ). The double-ceramiclayered (DCL) coating comprised of GZ and YSZ was confirmed to possess better durability. However, the particle-erosion resistance of GZ is poor due to its low fracture toughness. In this study, a novel erosion-resistant layer, an Al2O3-GdAlO3 (AGAP) amorphous layer, was deposited as the top layer to resist erosion. Three triple-ceramic-layer (TCL) coatings comprised of an Al2O3-GAP layer as the top layer, a GZ layer, a GZ/YSZ composite layer, and a rare-earth-doped gadolinium zirconate (GSZC) layer as the intermediate layer, and a YSZ layer as the base layer. For comparison, an AGAP-YSZ DCL coating without a middle layer was prepared as well. Under the erosion speed of 200 m/s, only a small amount of spallation occurred on the surface of the Al2O3-GAP layer, indicating a superior particle-erosion resistance. In the thermal shock test, the Al2O3-GAP layer experienced glass transition and the glass transition temperature was close to 1500 ◦C. The hardness of the Al2O3GAP coating after glass transition increased ~170% compared to the as-sprayed Al2O3-GAP coating. Moreover, The DCL TBC and TCL TBCs exhibited different failure mechanisms, which illustrated the necessity of the middle layer. The finite element model (FEM) simulation also shows that the introduction of the GZ layer can obviously reduce the thermal stress at the TC/BC interface. In terms of coating with a modified GZ layer, the AGAP-GZ/YSZ-YSZ coating and AGAP-GSZC-YSZ coating showed a similar failure model to the AGAP-GZ-YSZ coating, and the AGAP-GSZC-YSZ coating exhibited better thermal shock resistance.


Introduction
Thermal barrier coatings (TBCs) are widely employed on hot-section components of turbine engines due to their excellent thermal insulation performance. The higher engine-operating temperature could result in improved efficiency and specific power; therefore, growing requirements are put forward to TBCs for better high-temperature performance [1][2][3]. Typical TBC systems consist of a super-alloy substrate, a metallic bond coat (MCrAlY, M = Ni or/and Co), and a ceramic topcoat [4,5]. In general, TBCs are deposited by atmospheric plasma spraying (APS) with a lamellar microstructure and electron-beam physical vapor deposition (EB-PVD) with a columnar microstructure [4]. Additionally, some emerging techniques are investigated and developed, such as solution precursor plasma spray (SPPS) and plasma spray-physical vapor deposition, etc. [6][7][8]. However, APS applications are more predominant owing to their high spray efficiency and versatility. In terms of TBCs applied to more severe service environments, the search for new candidate ceramic materials and structure designs for ceramic layers are two important indicators [9][10][11][12].
Up until now, 6 wt.%-8 wt.% Y 2 O 3 stabilized ZrO 2 (YSZ) is the commonly used topcoat material that possesses appropriate thermal expansion, excellent thermal insulation, and crack resistance. However, the operating temperature of YSZ TBC is limited below 1200 • C. It is reported that the thermal shock lifetime of YSZ TBC reduces rapidly under experimental conditions exceeding 1100 • C due to the undesired phase transform and sintering of YSZ [13,14]. Moreover, the failure of the bond coat and substrate materials can cause spallation when operating temperatures exceed their service limits. With the higher turbine inlet temperature of developed gas engines, higher surface temperatures and larger thermal gradients in TBC systems are required. In order to meet this ambitious goal, a new generation of TC candidate materials has been explored in recent decades. Rare-earth co-stabilized zirconia [15,16], pyrochlore or fluorite structure compounds (A 2 B 2 O 7 ) [17], hexaaluminate compounds with magnetoplumbite structures (LnMgAl 11 O 19 ) [18,19], and perovskite structure compounds (ABO 3 ) [20] have been put forward as potential materials for ceramic layers. In particular, gadolinium zirconate with excellent sinter resistance, thermal stability, and low thermal conductivity, is considered to be the most promising ceramic layer material for ultra-high-temperature environments [9,21,22]. However, its low fracture toughness leads to poor durability both in thermal shock tests and erosion tests [23][24][25]. Therefore, the toughening of Gd 2 Zr 2 O 7 is crucial to realizing its further application in TBC.
Moreover, solid particle erosion is another important factor that causes the failure of TBCs [26]. Erosion occurs due to the impaction of solid particles ingested from the air stream when aero gas turbines operating in sandy environments or solid products are generated in the combustion chamber. Subsequently, the mass loss and even spallation of the ceramic coating leads to the failure of thermal protection of the substrate material. The factors that influence the erosion resistance of TBCs include their microstructure, porosity, and materials [27]. EB-PVD coatings with a columnar microstructure possess better erosion resistance than splat-like microstructure coatings prepared by APS as different erosion mechanisms occur during particle erosion [28]. However, the high cost and low deposition efficiency limit the further application of EB-PVD coatings. Additionally, the porosity of an APS coating significantly determines its erosion resistance owing to the dependency of the coating fracture toughness on the lamellar bonding ratio [29]. Moreover, the intrinsic fracture toughness of different materials can influence their erosion resistance [30].
In our previous study, two kinds of toughened GZ coatings were proposed, which exhibited better fracture toughness than conventional GZ coatings [31]. The particle-erosion test also proved that the coatings with higher fracture toughness possess better erosion resistance simultaneously. However, the erosion rates of GZ/YSZ coatings and GSZC coatings are still too high to fulfil the service requirements. It seems to be difficult for a single material to meet all the requirements for a TBC. To improve the durability of TBCs under complex service conditions, it is of great significance to design novel architecture, especially multilayer TBCs [32,33]. The multilayered TC is composed of an erosion-resistant outer layer, a thermal barrier middle layer, and a crack-resistant bottom layer. Based on this multilayer system, a triple-ceramic-layer (TCL) coating was put forward in the present study. A novel Al 2 O 3 -GdAlO 3 (AGAP) amorphous coating was deposited as the outer layer on the original double-ceramic-layer (DCL) coating aiming to resist erosion. The middle GZ layer works as a thermal barrier layer and the bottom YSZ layer can reduce thermal mismatch stress between the GZ layer and the BC. The particle-erosion resistance and thermal shock behavior were investigated in this study. For comparison, the AGAP-YSZ double-ceramic-layer (DCL) coating was prepared and tested as well.

Coating Deposition
Prior to spray, GH3230 superalloy substrates with a dimension of Φ 25.4 mm × 3 mm were grit-blasted and then ultrasonically cleaned with alcohol in order to increase the adhesion strength. In the present research, all coatings were prepared by the F4 APS system (Oerlikon Metco, Zurich, Switzerland). The bond coat was deposited using a commercially available NiCrAlY powder (45-106 µm, Beijing Sunspraying new material Co., Ltd., Beijing, China). As for the multi-layered topcoat, the materials of the bottom YSZ layers and middle zirconate layers were the same as in our previous study [31], and an Al 2 O 3 /Gd 2 O 3 granulated powder was utilized to prepare the top ceramic layer, namely AGAP. For simplicity, the middle layers were prepared with Gd 2 Zr 2 O 7 powder, a mixed powder of Gd 2 Zr 2 O 7 and YSZ, and (Gd 0 . 925 Sc 0 . 075 ) 2 (Zr 0 . 7 Ce 0 . 3 ) 2 O 7 powders are named after GZ, GZ/YSZ, and GSZC, respectively. Figure 1 shows the morphologies and particle size distribution of Al 2 O 3 /Gd 2 O 3 granulated powder, which exhibited a subsphaeroidal shape with a D 50 of~40 µm. In the surface micrograph of Al 2 O 3 /Gd 2 O 3 granulated powder under backscattered electron (BSE) mode, numerous light and dark regions are observed, which demonstrated that the material is composed of several phases. During the preparation of the topcoat, the preheating process was carried out before spraying each layer in order to remove surface stains and obtain a well-bonded interface. To ensure the comparability of the thermal shock test, the thickness of the bond coat of TBCs was 120 µm and the total thickness of the multilayered top coating was~480 µm. The detailed spraying parameters of multilayer TBCs are listed in Table 1. resistance and thermal shock behavior were investigated in this study. For comparison, the AGAP-YSZ double-ceramic-layer (DCL) coating was prepared and tested as well.

Coating Deposition
Prior to spray, GH3230 superalloy substrates with a dimension of Ф 25.4 mm × 3 mm were grit-blasted and then ultrasonically cleaned with alcohol in order to increase the adhesion strength. In the present research, all coatings were prepared by the F4 APS system (Oerlikon Metco, Zurich, Switzerland). The bond coat was deposited using a commercially available NiCrAlY powder (45-106 μm, Beijing Sunspraying new material Co., Ltd., Beijing, China). As for the multi-layered topcoat, the materials of the bottom YSZ layers and middle zirconate layers were the same as in our previous study [31], and an Al2O3/Gd2O3 granulated powder was utilized to prepare the top ceramic layer, namely AGAP. For simplicity, the middle layers were prepared with Gd2Zr2O7 powder, a mixed powder of Gd2Zr2O7 and YSZ, and (Gd0.925Sc0.075)2(Zr0.7Ce0.3)2O7 powders are named after GZ, GZ/YSZ, and GSZC, respectively. Figure 1 shows the morphologies and particle size distribution of Al2O3/Gd2O3 granulated powder, which exhibited a subsphaeroidal shape with a D50 of ~40 μm. In the surface micrograph of Al2O3/Gd2O3 granulated powder under backscattered electron (BSE) mode, numerous light and dark regions are observed, which demonstrated that the material is composed of several phases. During the preparation of the topcoat, the preheating process was carried out before spraying each layer in order to remove surface stains and obtain a well-bonded interface. To ensure the comparability of the thermal shock test, the thickness of the bond coat of TBCs was ~120 μm and the total thickness of the multilayered top coating was ~480 μm. The detailed spraying parameters of multilayer TBCs are listed in Table 1.

Particle-Erosion Test
Solid particle-erosion test was conducted via a homemade jet tester, where coating samples are subjected to continuous impinging of abrasives at room temperature. The impingement angle can be changed by adjusting the clamp, and this study choose an erosion angle of 90 • in order to produce higher erosion rates. Irregular Al 2 O 3 powder with a size range of 40-60 µm was chosen as erodent. During the erosion process, the samples were exposed to successive increments of abrasives at a feed rate of 2 g/min for 5 min. Erosion speed was controlled at 200 m/s by adjusting the carrier gas pressure. Sample masses were measured before erosion and after every 10 g of abrasives using a sensitive weighting balance with a precision of 0.01 mg. Prior to each mass measurement, ultrasonic cleaning was performed in order to remove the loosely bound erosion debris after the test as it could influence the results. The slope of the function of coating mass loss versus erodent exposure is taken as the erosion rate. Two samples of each TBC variation were tested, which were used for top surface analysis and cross-sectional analysis, respectively.

Thermal Shock Test
Burner rig thermal shock (BRTS) test was carried out using a gas burner rig test setup (Shaanxi Dewei Automation Co., Ltd., Xi'an, China). During the thermal shock process, the coating side of the sample was heated by high-temperature flame produced by a propane-oxygen gun and, simultaneously, the substrate side of the sample was cooled by compressed air. Therefore, a large temperature gradient formed inside the TBC system, which better simulated the engine operating environment. This temperature load can be set by adjusting the distance between the sample and propane-oxygen gun as well as the flow rate of compressed air. In this study, an extremely severe temperature condition was adopted to estimate the thermal shock resistance of samples. Each cycle is composed of heating the coating surface to 1500 • C (the temperature of substrate surface was about 1050 • C), maintaining this high temperature for 1000 s, and subsequently rapidly quenching in deionized water. A noncontact infrared thermometer with a range of −18 • C to 1650 • C was used to monitor the coating surface and substrate surface temperatures of the samples. Under such harsh experimental conditions, the performance of TBCs will deteriorate rapidly, so thermal shock lifetimes are not an appropriate evaluating indicator in this study. Therefore, three thermal cycles were conducted on each sample in order to obtain samples in a failure state. Three samples of each TBC variety were tested and the failed samples were investigated and compared systemically.

Characterization
The as-sprayed TBCs and failed TBCs were cold-mounted using liquid epoxy resin after cutting, followed by a metallographic polishing procedure to prepare the crosssectional sample of TBCs. The surface and cross-sectional microstructure of samples were investigated by scanning electron microscopy (SEM, HITACHI S-3400 N, Tokyo, Japan). The energy dispersive spectrometer (EDS) equipped in the SEM was used to analyze the chemical composition of samples. Phase analysis of the top surface of coating was performed using X-ray diffraction (XRD, D/Max, 2550VB/PC, RIGAKU., Tokyo, Japan) with filtered Cu Kα radiation. Diffraction angle was set in the range of 10-80 • at a scan rate of 12 • /min. The microhardness of the coating was measured by indenter method on metallographically polished cross-section of the sample using a micro-Vickers Indenter (BUEHER MICROMET5104, Akashi Corporation, Osaka, Japan) with a load of 300 g for 15 s. A total of 15 points were randomly selected for each sample to increase the reliability of the data.

Finite Element Analysis
In terms of multilayer TBCs, not only the TC/BC interface but also the interfaces inside the TC are subjected to complex thermal stress during thermal shock owing to the mismatch of material properties and uneven temperature distribution in the TBC system. A finite element model (FEM) was established to study the influence of the multilayered TC design on the stress distribution of TBC during thermal shock by ABAQUS 2016. The geometric model with two kinds of multilayered TC designs was extracted from the cross-sectional morphologies with a width of 1 mm, as shown in Figure 2. The temperature load was consistent with that of the thermal shock test, and the highest temperature of the coating surface and substrate surface were 1500 and 1050 • C, respectively. Sequentially coupled thermal stress analysis was applied in this simulation. Firstly, the temperature distribution was calculated by transient heat transfer analysis. The heat exchange between the sample and its surrounding environment was achieved by convection, and the convective heat transfer coefficients of the heating stage and cooling stage were taken as 110 and 1000 W/m 2 · • C, respectively [34]. Additionally, heat transfer inside the sample was carried out by conduction. Subsequently, the stress analysis was performed under the previously calculated temperature field. Meanwhile, the left boundary of the model was set as a symmetric boundary, and the vertical displacement of the bottom boundary was set to zero. The material properties used in this FEM analysis were obtained from the literature [35][36][37][38].
mismatch of material properties and uneven temperature distribution in the TBC system. A finite element model (FEM) was established to study the influence of the multilayered TC design on the stress distribution of TBC during thermal shock by ABAQUS 2016. The geometric model with two kinds of multilayered TC designs was extracted from the crosssectional morphologies with a width of 1 mm, as shown in Figure 2. The temperature load was consistent with that of the thermal shock test, and the highest temperature of the coating surface and substrate surface were 1500 and 1050 °C, respectively. Sequentially coupled thermal stress analysis was applied in this simulation. Firstly, the temperature distribution was calculated by transient heat transfer analysis. The heat exchange between the sample and its surrounding environment was achieved by convection, and the convective heat transfer coefficients of the heating stage and cooling stage were taken as 110 and 1000 W/m 2 ·°C, respectively [34]. Additionally, heat transfer inside the sample was carried out by conduction. Subsequently, the stress analysis was performed under the previously calculated temperature field. Meanwhile, the left boundary of the model was set as a symmetric boundary, and the vertical displacement of the bottom boundary was set to zero. The material properties used in this FEM analysis were obtained from the literature [35][36][37][38].

Characterization
The cross-sectional microstructures of the as-sprayed TBCs are shown in Figure 3. All multilayer coatings exhibit clear interfaces, which demonstrates that the multilayered TCs are in a well-bonded state. The typical lamellar structure accompanied by some unmelted particles, pores, and interlaminar cracks can be seen in all layers. It is these characteristics that ensure the superb thermal properties of TBCs. Additionally, these layers exhibit slightly different porosity due to differences in the materials and spraying processes. The average porosity of the Al2O3-GAP, GZ, GZ/YSZ, GSZC, and YSZ layers evaluated from the SEM micrographs using ImageJ software (three typical images for each layer) are approximately 6.48%, 8.55%, 13.53%, 14.68%, and 12.03%, respectively. The Al2O3-GAP layer is relatively dense in spite of the distribution of some micro-melted particles. Owing to the high spraying power and a high traverse speed of the plasma gun, the Al2O3/Gd2O3 granulated powders will melt better during the spraying process and a wellbonded splats interface will be formed during solidification. Moreover, the dense microstructure is beneficial for improving the particle-erosion resistance of the coatings. As seen in Figure 4a, the main diffraction peaks in the XRD spectrum of the Al2O3/Gd2O3 granulated powders could be attributed to α-Al2O3 and c-Gd2O3. It can be seen that obvious amorphous scattering peaks occurred in the as-sprayed Al2O3-GAP coating, as shown in Figure 4b, which indicates that a large number of amorphous phases formed during the spraying process. Additionally, a small number of α-Al2O3 and GAP phases existed in the amorphous matrix. The microstructure and phase composition of the middle zirconate layers and bottom YSZ layers are similar to our previous study [31]. The low thermal

Characterization
The cross-sectional microstructures of the as-sprayed TBCs are shown in Figure 3. All multilayer coatings exhibit clear interfaces, which demonstrates that the multilayered TCs are in a well-bonded state. The typical lamellar structure accompanied by some unmelted particles, pores, and interlaminar cracks can be seen in all layers. It is these characteristics that ensure the superb thermal properties of TBCs. Additionally, these layers exhibit slightly different porosity due to differences in the materials and spraying processes. The average porosity of the Al 2 O 3 -GAP, GZ, GZ/YSZ, GSZC, and YSZ layers evaluated from the SEM micrographs using ImageJ software (three typical images for each layer) are approximately 6.48%, 8.55%, 13.53%, 14.68%, and 12.03%, respectively. The Al 2 O 3 -GAP layer is relatively dense in spite of the distribution of some micro-melted particles. Owing to the high spraying power and a high traverse speed of the plasma gun, the Al 2 O 3 /Gd 2 O 3 granulated powders will melt better during the spraying process and a well-bonded splats interface will be formed during solidification. Moreover, the dense microstructure is beneficial for improving the particle-erosion resistance of the coatings. As seen in Figure 4a, the main diffraction peaks in the XRD spectrum of the Al 2 O 3 /Gd 2 O 3 granulated powders could be attributed to α-Al 2 O 3 and c-Gd 2 O 3 . It can be seen that obvious amorphous scattering peaks occurred in the as-sprayed Al 2 O 3 -GAP coating, as shown in Figure 4b, which indicates that a large number of amorphous phases formed during the spraying process. Additionally, a small number of α-Al 2 O 3 and GAP phases existed in the amorphous matrix. The microstructure and phase composition of the middle zirconate layers and bottom YSZ layers are similar to our previous study [31]. The low thermal conductivity of the middle zirconate layers guarantees the performance of TBCs for high temperatures. Meanwhile, the bottom YSZ layer was designed to improve the cracking resistance of the coating and solve the thermochemical compatibility issues of gadolinium zirconate and TGO [38]. conductivity of the middle zirconate layers guarantees the performance of TBCs for high temperatures. Meanwhile, the bottom YSZ layer was designed to improve the cracking resistance of the coating and solve the thermochemical compatibility issues of gadolinium zirconate and TGO [38].    conductivity of the middle zirconate layers guarantees the performance of TBCs for high temperatures. Meanwhile, the bottom YSZ layer was designed to improve the cracking resistance of the coating and solve the thermochemical compatibility issues of gadolinium zirconate and TGO [38].    However, the amorphous Al 2 O 3 -GAP layer still exhibits an absolute advantage in erosion resistance (~0.077 mg/g) compared to the middle layers. Therefore, the deposition of the Al 2 O 3 -GAP layer upon the coating to improve the erosion resistance of TBC systems is a reasonable choice. For further study, its erosion mechanism will be investigated in the following section.

Particle-Erosion Resistance
in Figure 6a. Furthermore, these splats spread fully during deposition as particles have melted completely in the high-power plasma flame and few vertical cracks inside the splats formed during solidification. Extraordinarily, a relatively smooth surface morphology occurs after the erosion test (shown in Figure 6b), which is entirely different from the fracture feature of the GZ coating after the impaction of abrasives in our previous study [31]. The excellent crack resistance of the Al2O3-GAP grain boundary in combination with the few primary cracks inside the Al2O3-GAP splats contributed to this failure morphology. As a result, the Al2O3-GAP splats tend to peel off due to cracks propagated along the splat boundaries. On the other hand, as for the coating fabricated from the fine powders, these droplets tend to stack on the surface in a flat manner and, subsequently, the splat interfaces are smooth [39]. Therefore, a relatively smooth surface morphology was exposed after the erosion test due to cracks propagated along the smooth splat interfaces, which can be confirmed by the cross-sectional morphologies, as shown in Figure 6c,d. In general, the superior erosion resistance of the Al2O3-GAP coating can be attributed to the well-bonded splat interfaces.    Figure 6a. Furthermore, these splats spread fully during deposition as particles have melted completely in the high-power plasma flame and few vertical cracks inside the splats formed during solidification. Extraordinarily, a relatively smooth surface morphology occurs after the erosion test (shown in Figure 6b), which is entirely different from the fracture feature of the GZ coating after the impaction of abrasives in our previous study [31]. The excellent crack resistance of the Al 2 O 3 -GAP grain boundary in combination with the few primary cracks inside the Al 2 O 3 -GAP splats contributed to this failure morphology. As a result, the Al 2 O 3 -GAP splats tend to peel off due to cracks propagated along the splat boundaries. On the other hand, as for the coating fabricated from the fine powders, these droplets tend to stack on the surface in a flat manner and, subsequently, the splat interfaces are smooth [39]. Therefore, a relatively smooth surface morphology was exposed after the erosion test due to cracks propagated along the smooth splat interfaces, which can be confirmed by the cross-sectional morphologies, as shown in Figure 6c,d. In general, the superior erosion resistance of the Al 2 O 3 -GAP coating can be attributed to the well-bonded splat interfaces.

Thermal Shock Resistance
In this test, an extreme temperature condition was carried out to estimate the thermal shock resistance of four kinds of TBCs. Thus, the thermal shock lifetimes of TBCs are not focused on in this section, and their thermal shock behavior is investigated after three

Thermal Shock Resistance
In this test, an extreme temperature condition was carried out to estimate the thermal shock resistance of four kinds of TBCs. Thus, the thermal shock lifetimes of TBCs are not focused on in this section, and their thermal shock behavior is investigated after three cycles of thermal shocks.
The macro and microscopic images of the AGAP-YSZ coating after the thermal shock test are given in Figure 7. It can be seen in the macroscopic image that the coating peeled off slightly at the edge of the sample. More obviously, the coating bulged at the center of the sample and a glassy layer formed after the thermal shock test. Generally, the initial cracking originated from the edge of the samples due to the edge effect and gradually spread to the adjacent regions during the conventional thermal shock test [40,41]. However, an obvious bulge at the center instead of spallation at the edge of the sample occurred in this test, which indicated a different failure mechanism. From the cross-sectional morphologies, some details can be gained visually. The TC gradually bulges at area 1 and a nearly 500 µm gap exists between the TC and BC at area 2. Owing to the absence of a middle heat-insulation layer with low thermal conductivity, the TC bulged dramatically under the action of a tremendous thermal expansion mismatch between the TC and BC during such a severe thermal shock test. Additionally, some vertical cracks propagated inside the YSZ layer due to radial stress generated during deformation.
Coatings 2022, 12, x FOR PEER REVIEW 9 of 16 was slightly lower due to the higher temperature of the flame center. The unique vitrification at the center of the sample indicates that the glass transition temperature of Al2O3-GAP coating is close to 1500 °C. In addition, the Al2O3-GAP layer is significantly thinned in Figure 7f, indicating that the Al2O3-GAP layer may be difficult to operate in higher temperatures.   Another unanticipated finding was that the Al 2 O 3 -GAP layer experienced a series of interesting transitions during the thermal shock test. Figure 7d shows the microstructure of the Al 2 O 3 -GAP layer in the peripheral area of the sample after the thermal shock test. Some short horizontal and vertical cracks were propagated inside the Al 2 O 3 -GAP layer, whereas the interface of the Al 2 O 3 -GAP layer and YSZ layer remained well-bonded. In contrast, a highly dense microstructure occurred in the central area of the sample after the thermal shock test, as shown in Figure 7e. Additionally, the top and bottom of the Al 2 O 3 -GAP layer exhibited completely different states. Compositional segregation seems to be occurring at the bottom of the Al 2 O 3 -GAP layer, which finally resulted in glass transition at the top of the Al 2 O 3 -GAP layer. To gain more information, element mapping of the vitrification area was detected by EDS, as shown in Figure 8. It is obvious that the Al element is enriched in the darker area and the Gd element is enriched in the brighter area at bottom of the Al 2 O 3 -GAP layer, whereas the Al and Gd elements are distributed uniformly in the vitrification area at the top of the Al 2 O 3 -GAP layer. It seems that the glass transition of the Al 2 O 3 -GAP layer is quite sensitive to temperature. In our thermal shock test, the temperature of the sample center reached 1500 • C, whereas the temperature of the sample outside was slightly lower due to the higher temperature of the flame center. The unique vitrification at the center of the sample indicates that the glass transition temperature of Al 2 O 3 -GAP coating is close to 1500 • C. In addition, the Al 2 O 3 -GAP layer is significantly thinned in Figure 7f, indicating that the Al 2 O 3 -GAP layer may be difficult to operate in higher temperatures.  A mechanical property measurement was carried out to evaluate the effect of glass transition on the coating property using the indentation method. Figures 8 and 9 show the typical indentation morphologies and Vickers hardness of the Al2O3-GAP coating in different states, respectively. The hardness of the thermal-shocked Al2O3-GAP coating (723 HV) is slightly higher than the as-sprayed Al2O3-GAP coating (586.3 HV). However, the hardness of the Al2O3-GAP coating in segregation and vitrification increased ~170% compared to the as-sprayed Al2O3-GAP coating, which indicates that glass transition is beneficial to the mechanical properties of the Al2O3-GAP coating and even better erosion resistance of the Al2O3-GAP coating may be obtained after glass transition. Unexpectedly, the Al2O3-GAP coating in vitrification becomes brittle as cracks propagated obviously on both sides of the indentation. A mechanical property measurement was carried out to evaluate the effect of glass transition on the coating property using the indentation method. Figures 8 and 9 show the typical indentation morphologies and Vickers hardness of the Al 2 O 3 -GAP coating in different states, respectively. The hardness of the thermal-shocked Al 2 O 3 -GAP coating (723 HV) is slightly higher than the as-sprayed Al 2 O 3 -GAP coating (586.3 HV). However, the hardness of the Al 2 O 3 -GAP coating in segregation and vitrification increased~170% compared to the as-sprayed Al 2 O 3 -GAP coating, which indicates that glass transition is beneficial to the mechanical properties of the Al 2 O 3 -GAP coating and even better erosion resistance of the Al 2 O 3 -GAP coating may be obtained after glass transition. Unexpectedly, the Al 2 O 3 -GAP coating in vitrification becomes brittle as cracks propagated obviously on both sides of the indentation. The morphologies of the AGAP-GZ-YSZ coating after failure are displayed in Figure  10, which shows a completely different failure mode compared to the AGAP-YSZ coating. Besides the spallation of the TC at the edge, obvious cracks occurred at the Al2O3-GAP/GZ and GZ/YSZ interfaces. The existence of the GZ layer with low thermal conductivity pro- The morphologies of the AGAP-GZ-YSZ coating after failure are displayed in Figure 10, which shows a completely different failure mode compared to the AGAP-YSZ coating. Besides the spallation of the TC at the edge, obvious cracks occurred at the Al 2 O 3 -GAP/GZ and GZ/YSZ interfaces. The existence of the GZ layer with low thermal conductivity provides considerable heat insulation, which reduced the thermal expansion mismatch between the BC and YSZ layer during the quenching process. As a result, the TC tended to peel off from the edge due to the edge effect. Additionally, a continuous and uneven oxide layer with a thickness range of 2-10 µm has been formed at different locations of the TC/BC interface. Figure 11 shows the element mapping of the TC/BC interface, which indicates that the enrichment of Cr inside the oxide layer appears rather than the enrichment of the Al. This element composition indicated that the oxide layer is not a traditional thermally grown oxide (TGO) layer. Other studies also reported that the formation and thickening of the TGO layer are not obvious in the BRTS test [42,43]. The absence of TGO is due to the relatively low temperature of the BC layer and an insufficient thermal exposure time at a high temperature in the BRTS test. The thicker oxide layer in Figure 10d was caused by the accelerated oxidation of the BC layer after the peeling of the TC layer. Therefore, this BRTS test can reduce the effect of TGO growth on the cracking and failure of TBCs and assess the performance of TBCs under extreme temperature gradients. Additionally, the existence of the GZ layer introduces more interfaces in a multilayer TC and these interfaces are prone to cracking under thermal shock due to thermal mismatch. What is worth noting is that component segregation also occurred at the center of the Al 2 O 3 -GAP layer in this sample; however, it seemed that the Al 2 O 3 -GAP layer has peeled off before a complete glass transition.
Coatings 2022, 12, x FOR PEER REVIEW 11 of 16 in the AGAP-GZ-YSZ coating, compared to the AGAP-YSZ coating. However, the thermal stress at Al2O3-GAP/GZ and GZ/YSZ interfaces increases simultaneously due to the differences in thermophysical properties of materials. It seems that the introduction of the GZ layer can decrease the thermal expansion mismatch stress between the BC and TC layer but, unexpectedly, it exacerbates the thermal expansion mismatch inside TC at the same time. As a result, premature spallation tends to occur inside the TC.
As the introduction of the GZ layer exhibits poor compatibility with the Al2O3-GAP layer and YSZ layer, two kinds of modified GZ layers (GZ/YSZ and GSZC) were used to replace the GZ layer. Figures 14 and 15 display the macroscopic and cross-sectional morphologies of the failed AGAP-GZ/YSZ-YSZ and AGAP-GSZC-YSZ coatings, respectively, which shows a similar failure model with the AGAP-GZ-YSZ coating. Spallation of the TC at the edge and the obvious formation of oxide at the BC/TC interface occurred in the AGAP-GZ/YSZ-YSZ coating. However, the GZ/YSZ layer remained well-bonded with the Al2O3-GAP top layer during thermal shock. The propagation of cracks inside the GZ/YSZ layer resulted in the spallation of the GZ/YSZ layer together with the Al2O3-GAP top layer. The GZ-GSZC-YSZ coating seems to exhibit better thermal shock resistance as the whole TC did not peel off significantly after the thermal shock test. The GSZC layer with even lower thermal conductivity further reduced the thermal expansion mismatch between the BC and YSZ layer, which delayed the spallation of the TC. Additionally, the GSZC layer exhibits excellent compatibility with the Al2O3-GAP top layer as few cracks were propagated at the Al2O3-GAP/GSZC interface during thermal shock. However, the unexcepted spallation of the GSZC layer together with the Al2O3-GAP top layer still occurred, which is related to the relatively low fracture toughness of the GSZC layer.  The AGAP-YSZ and AGAP-GZ-YSZ coatings exhibited completely different failure models in the thermal shock test. The AGAP-YSZ coating failed due to a tremendous bulge of the TC, while the propagation of interfacial cracks and spallation of the TC at the edge both occurred in the AGAP-GZ-YSZ coating. Now, the failure behaviors of two kinds of TBCs will be analyzed from stress distribution by FEM simulation. The thermal stress distribution of AGAP-YSZ and AGAP-GZ-YSZ TBCs after quenching are shown in Figure 12. The maximum S22 (stress perpendicular to the interfaces) is located at the YSZ/BC interface in the AGAP-YSZ model, and obvious stress concentrations appear simultaneously at the Al 2 O 3 -GAP/GZ, GZ/YSZ, and YSZ/BC interfaces in the AGAP-GZ-YSZ model. The simulation results can support the experimental results as the position of spallation is consistent with that of the maximum stress.     In order to observe the simulation results more intuitively, the thermal stress distribution along the coating thickness at the right boundary in the TBC system with two kinds of TC designs after quenching are plotted in Figure 13. Evidently, it suggests that the thermal stress at YSZ/BC interface reduces significantly with the introduction of the GZ layer in the AGAP-GZ-YSZ coating, compared to the AGAP-YSZ coating. However, the thermal stress at Al 2 O 3 -GAP/GZ and GZ/YSZ interfaces increases simultaneously due to the differences in thermophysical properties of materials. It seems that the introduction of the GZ layer can decrease the thermal expansion mismatch stress between the BC and TC layer but, unexpectedly, it exacerbates the thermal expansion mismatch inside TC at the same time. As a result, premature spallation tends to occur inside the TC.   As the introduction of the GZ layer exhibits poor compatibility with the Al 2 O 3 -GAP layer and YSZ layer, two kinds of modified GZ layers (GZ/YSZ and GSZC) were used to replace the GZ layer. Figures 14 and 15 display the macroscopic and cross-sectional morphologies of the failed AGAP-GZ/YSZ-YSZ and AGAP-GSZC-YSZ coatings, respectively, which shows a similar failure model with the AGAP-GZ-YSZ coating. Spallation of the TC at the edge and the obvious formation of oxide at the BC/TC interface occurred in the AGAP-GZ/YSZ-YSZ coating. However, the GZ/YSZ layer remained well-bonded with the Al 2 O 3 -GAP top layer during thermal shock. The propagation of cracks inside the GZ/YSZ layer resulted in the spallation of the GZ/YSZ layer together with the Al 2 O 3 -GAP top layer. The GZ-GSZC-YSZ coating seems to exhibit better thermal shock resistance as the whole TC did not peel off significantly after the thermal shock test. The GSZC layer with even lower thermal conductivity further reduced the thermal expansion mismatch between the BC and YSZ layer, which delayed the spallation of the TC. Additionally, the GSZC layer exhibits excellent compatibility with the Al 2 O 3 -GAP top layer as few cracks were propagated at the Al 2 O 3 -GAP/GSZC interface during thermal shock. However, the unexcepted spallation of the GSZC layer together with the Al 2 O 3 -GAP top layer still occurred, which is related to the relatively low fracture toughness of the GSZC layer. Figure 13. Thermal stress (S22) distribution along coating thickness in TBC system with two kinds of TC design after quenching.

Conclusions
In this paper, a novel erosion-resistant coating (Al 2 O 3 -GAP coating) was proposed. Subsequently, a multilayered architecture was designed for TBCs under harsh erosion and high-temperature serving. The erosion resistance and thermal shock behavior of these TBCs were investigated and analyzed. The major useful conclusions can be drawn as follows: 1.
The Al 2 O 3 -GAP amorphous coating with a dense microstructure was fabricated by Al 2 O 3 /Gd 2 O 3 granulated powder under an appropriate plasma-spraying process.
In the particle-erosion test, Al 2 O 3 -GAP coating exhibited superior erosion resistance compared to the middle layers and protected the integrity of the coating structure. It shows that the Al 2 O 3 -GAP coating is a promising erosion-resistant layer under severe erosion conditions. 2.
In the thermal shock test, the Al 2 O 3 -GAP layer experienced a series of transitions. Obvious compositional segregation and vitrification occurred at the central area of the sample after the thermal shock test, which indicated that the glass transition temperature of the Al 2 O 3 -GAP coating is close to 1500 • C. The hardness of the Al 2 O 3 -