Suspension High‐Velocity Oxy‐Fuel–Sprayed Dense Vertically Cracked and Suspension‐Plasma‐Sprayed Columnar Yttria‐Stabilized Zirconia Coatings: Calcia Magnesia Alumino Silicates Infiltration and Thermal Cycling Performance

The quest to increase the surface temperatures and resistance to the corrosive environment of thermal barrier coatings topcoats mean that newer coating design strategies are needed. In this study, a performance evaluation of suspension high‐velocity oxy‐fuel (SHVOF)‐sprayed dense vertically cracked (DVC) and suspension‐plasma‐sprayed (SPS) columnar structure (CS) topcoats is conducted. The calcia magnesia alumino silicate (CMAS) evaluation is conducted at 1300 °C for 30 min, whereas the furnace cycling tests (FCT) is conducted at 1135 °C for 45 min cycle dwell time. The CMAS infiltrates down to the bond coat layer, but does not induce partial or complete topcoat spallation on all studied topcoat layers. In terms of CMAS infiltration, the CMAS appears to be restricted along the vertical cracks. The FCT of the SPS CS structure indicates a failure largely at the thermally grown oxide (TGO)–topcoat interface, while the DVC topcoat layers indicate a mix‐mode failure, i.e., both material‐associated cracking and localized spallations at the TGO–topcoat interface. Overall, the SHVOF‐sprayed ethanol‐based DVC topcoat seems to offer a balanced trade‐off, i.e., a majority of the topcoat is still intact after 100 thermal cycles and exceeds the material durability and performance offered by the SPS CS structure.


Introduction
[3][4] In the case of aircraft engines, the incorporation of the ceramic TBC coatings, in conjunction with the advances in superalloy compositions and casting technology, has allowed the engine manufacturers to increase turbine inlet temperatures as high as 1600 °C, improving the engine's performance and fuel efficiency and the lifetime of the underlying superalloys. [5,6]The technology is also at the forefront of NetZero2050 materials development to enhance localized surface temperatures to increase overall engine efficiency.Typically, yttriastabilized zirconia (t'-YSZ, tetragonal) containing 7-8 wt% Y 2 O 3 is used as the preferred material composition for the TBCs in gas turbines, owing to their low thermal conductivity, high-temperature applicability, inertness, and compatibility with the thermal expansion coefficient of underlying superalloy substrate. [7]The most common methods to deposit TBCs are atmospheric plasma spraying (APS) and electron beam physical vapor deposition (EB-PVD). [8,9]The APS coatings deposit in a lamellar structure and are typically used in power-generation engines and aviation to coat combustor chambers and vanes, whereas the EB-PVD coatings deposit in a feathery structure and are typically used in aviation to coat turbine blades and vanes.In many respects, ceramic TBCs are designed to withstand the most demanding conditions a material can withstand in today's technology; however, despite the promise offered by these ceramic materials, they are prone to cracking and spallation during high-temperature exposure and thermal cycling. [1]he failure phenomena may include 1) high-temperature sintering of porous YSZ, which makes the coating stiffer and less compliant over time, [10] 2) transverse cracking (branching of vertical cracks) caused by thermal shock upon cooling, [11,12] 3) the destabilization of tetragonal metastable phase, t', through a rather complex transformation process, inducing undesirable stresses due to a %4% volume change, [3,13] and 4) the unsustainable, time-dependent growth of thermally grown oxide (TGO) scale, which may likely induce internal stresses due to thermal expansion mismatch at the topcoat/bond coat interface. [5,14]The large stresses in the material system, due to the YSZ sintering and TGO growth, cause the formation of interfacial cavities in the TGO layer, which could lead to the propagation/coalescence of interfacial cracks and, subsequently, the delamination of the topcoat. [15]In addition to these material cavities and cracks, the internal stresses also cause the coating to buckle, which accelerates the spallation of the YSZ topcoat. [16,17][20][21][22][23] These calcia magnesia alumino silicate (CMAS) particles, owing to their low viscosity at the turbine-operating temperatures and excellent wetting characteristics with a TBC material, infiltrate via columnar/horizontal cracks and/or internal porosities.The infiltrated CMAS then reacts chemically to severely degrade the coatings, which may cause a premature delamination and expose the critical superalloy to hot, corrosive gases. [20]dvancements to induce a synergic high-temperature thermal cycling and CMAS resistance require fundamental changes in the coating design, which may include the use of newer feedstock processing techniques to produce dense structure, with a degree of control over TBCs vertical/horizontal crack density and vertical crack width (ensuring thermal cyclic compliance while restricting CMAS infiltration).For example, the associated thermal shock within the TBCs topcoat system following a hightemperature exposure can be significantly reduced by introducing vertical surface cracks. [24,25]These improvements in obtaining a more strain-tolerant TBC may not contribute to inducing CMAS resistance.CMAS resistance, in contrast, could be introduced by restricting the CMAS infiltration in the vertical cracks, denser microstructures as compared to CS and feathery columnar structures (CSs), and reducing the number and extent of horizontal cracks (in other words, the branching of vertical cracks).A dense vertically cracked (DVC) TBC design may offer a balanced trade-off to thermal cycling and CMAS degradation by introducing a controlled number of vertical cracks, where the crack width is optimized to induce a synergetic solution to both high-temperature thermal cycling and CMAS challenges.The dense microstructures separated by vertical cracks act as a barrier to CMAS penetration and may mitigate/regulate the CMAS interactions in the interior of the TBC; however, limited studies have been conducted to explore and understand the CMAS penetration and thermal cycling degradation on the DVC versus CS structures. [26,27][30][31] In addition, the selection of a suspension thermal spray method, the suspension medium, suspension solid loading, and spray parameters also dictates the final microstructure of the coatings.For instance, the suspension high-velocity oxy-fuel (SHVOF) is known to deposit denser, well-adhered coatings (as the molten splats are comparably thinner in high-velocity oxy-fuel (HVOF) flame due to higher kinetic energy generated by the fuel-oxygen combustion resulting in a supersonic speed of flame) as compared to those coatings produced by suspension plasma spray (SPS). [31,32]Furthermore, the choice of suspension medium influences the particle in-flight behavior, heat flux to the substrate, and droplet size distribution, which subsequently affects the overall flame enthalpy, vertical crack formation, and coating characteristics, such as hardness, porosity, and thickness. [33]urther to our previous study on the vertical microcrack formation for a DVC microstructure, [34] this study investigates the suspension thermal-sprayed 7.5YSZ (YSZ with 7.5 wt% Y 2 O 3 ) coatings to understand the CMAS infiltration and thermal cycling behavior.The SPS was used to produce a columnar TBC structure (CS), and SHVOF thermal spray to deposit a DVC TBC structure.The DVC TBC topcoat layers were deposited using two types of suspension medium (water and ethanol) in the SHVOF thermal spray in an attempt to optimize the topcoat characteristics for durability enhancement.No comparative examinations in terms of the effect of suspension medium for DVC structures have been reported in literature.Finally, all three coating design strategies were subjected to isothermal high-temperature CMAS and thermal cycling tests, and their microstructural performances, phase stability, and degradation mechanisms are compared and discussed.Thus, this study examines the DVC and CS structures in terms of their durability and performance.

Materials
The 7.5YSZ (7.5 wt% of yttria in zirconia) in an ethanol-based suspension and as a dried powder (which was later used to make water-based suspension) was supplied by Imerys Murg GmbH, Germany.To help with the stabilization of YSZ particles in the water-based suspension, ammonium hydroxide which was formulated within the lab was used to increase the suspension pH and D-3005 polyacrylic acid (PAA) supplied by Dow, USA, was used as a dispersant.Praxair, UK, supplied the MCrAlY powder (CO-210-24, À45 μm/þ16 μm, CoNiCrAlY, Co-32.0Ni-21.0Cr-8.0Al-0.5Y in wt%).CMAS powder (48 mol% Silica, 35 mol% calcia, 10 mol% magnesia, and 7 mol% alumina) was supplied from Oerlikon Metco, UK.

Feedstock Preparation
Two types of feedstock suspensions were sprayed in this study, i.e., YSZ with 7.5 wt% Y 2 O 3 at a 20 wt% solid loading in a) deionized water and b) ethanol.The YSZ powder particles are shown in Figure 1a, and the particle size distributions are shown in Figure 1b.The particle size distribution was measured using a laser scattering particle size distribution analyzer LA-960, Horiba Jobin Yvon, Japan.Overall, the particles had d 10 , d 50 , and d 90 of 1.69, 3.34, and 5.79 μm, respectively.However, Figure 1a indicates much finer particles as opposed to the PSD measured by the laser particle size analyzer.The supplier indicated a size range of <2 μm.The inconsistency in the particle sizes could be due to the equipment limitation or the dispersion of YSZ in deionized water, as no surfactants were added for PSD measurement.The YSZ water suspension for spraying was prepared by mixing the dried YSZ powder at a 20 wt% solid loading in the balance of deionized water.The PAA, at a 0.5 wt% of the solid loading, and a suspension pH of 9 was maintained to stabilize the dispersed phase and mitigate particle aggregation in the water-based suspension.The ethanol-based suspension was used as received.Both suspensions were mechanically rolled for 30 min, followed by an hour of simultaneous ultrasonic dispersion and mechanical stirring.

Substrate Preparation
The bond coat was deposited on disc-shaped Inconel 718 substrates of diameter 12.7 and 3 mm thickness.The nominal composition of the substrates was Ni-19.0Cr-18.5Fe-5.1Nb-3.0Mo-0.9Ti-0.5Al-0.04C(in wt%).The Inconel substrates were grit basted on a grit blaster supplied from Guyson, UK, at a 6 bar pressure using F100 brown alumina (0.125-0.149 mm).The grid-blasted substrates were ultrasonically cleaned before the bond coat deposition with industrial methylated spirit and deionized water thrice and were dried using compressed air.

Coating Deposition
The coatings were sprayed in two stages into a two-layer material system: topcoat and bond coat on the substrates.First, the Inconel substrates were sprayed with the MCrAlY bond coat and second, the bond-coated substrates were sprayed with the YSZ topcoat layer.A kerosene fuel-based HVOF thermal spray MetJet IV torch (Metallization, Dudley, UK) was used to spray the bond coat.The bond coat was sprayed at the gas flow rates representing a 100% stoichiometric combustion ratio, i.e., 476 mL min À1 kerosene and 920 L min À1 oxygen, and at 356 mm stand-off distance.More details on the bond coat deposition are described elsewhere. [35]he topcoat YSZ was deposited using SPS and SHVOF thermal spray.A Mettech Axial III system was used for SPS, and the system was a three-cathode high-power, high-enthalpy system with axial feedstock injection directly in the plasma plume.The suspension was injected using a Mettech NanoFeed system at a 100 mL min À1 feed rate.The nanofeed system operated in a closed-loop system, which means the suspension was continuously recirculated, allowing an easy switchover of spray conditions.The YSZ feedstock was injected axially in a relatively high-enthalpy flame (12.5 kJ L À1 ) at a 122 kW torch power and the suspension was atomized at 20 L min À1 using nitrogen gas.The flow of atomizing gas, which was injected before the suspension enters the plasma plume, was used to control the size of the in-flight droplets.This might alter the heat transfer and subsequently the formation of the final deposition. [36,37]t was imperative to mention that the injection of nitrogen as the carrier or atomizing gas was expected to improve the in-flight particle temperatures by enhancing the overall flame efficiency.The suspension was injected using a 0.02 in (0.51 mm) diameter injector and a 3/8 in (9.52 mm) exit nozzle was used for the torch.The SHVOF thermal spray was carried out using a modified TopGun SS torch, supplied by GTV GmBH, Germany.The suspension was injected axially using a 0.3 mm diameter injector nozzle at a 50 mL min À1 suspension feed rate.The spray torch had a 22 mm combustion chamber.The suspension chamber was pressurized and maintained at a 5 bar pressure and the flow was regulated using an ultrasonic feedback-loop flowmeter, supplied by Bronkhorst, UK.More details of the setup can be found elsewhere. [38]n both systems, the suspension was continuously stirred using a pneumatic mechanical stirrer.Both front and back cooling were supplied during spraying using compressed air jets.The spray torches were mounted on a 6-axis robot, supplied by ABB IRB 2400, UK, which allowed a greater control of the spray deposition and imparted repeatability.The step size was the torch movement length between the consecutive steps when the coating was being deposited (Table 1).
The naming convention used in this work is as follows: A-B-C, where A is the thermal spray method, B is the type of coating structure, and C is the suspension medium (whereas W for water and E for ethanol).For example, the naming convention for a DVC coating produced using the water-based YSZ suspension and SHVOF thermal spray method would be SHVOF-DVC-W.

Sample Preparation
The as-sprayed coatings were cross-sectioned on a precision cutting saw, supplied by Metprep, UK, using a SiC cutting disk.A room temperature cure two-part epoxy matrix was used to cold mount the sectioned samples under a vacuum to allow thorough impregnation.The mounted, cured samples were ground using SiC grinding paper and in the following grit size order: P240, P400, P800, and P1200.The ground samples were polished down to 1 μm using a diamond pad.

Microstructural and Phase Characterization
The X-Ray diffraction (XRD), via a D8 Advance Da Vinci diffractometer, supplied by Bruker, Germany, was used to identify the phase composition of the feedstock powder and the coatings.The measurements were carried out in Bragg-Brentano geometry from 10°to 90°2θ at a dwell time of 0.2 s step À1 , a step size set to 0.02°, and a copper cathode (λ kα = 1.5406Å).The diffraction patterns were analyzed using the EVA 5.1 program package supported by the PDF-2 database international centre for diffraction data-powder diffraction file (ICDD-PDF).
An FEI Quanta 600 scanning electron microscope (SEM), FEI Europe, The Netherlands, was used to capture microstructural images and to conduct energy-dispersive X-Ray (EDX) elemental characterization.The microstructural images (both backscattered and secondary electron imaging [BSE and SE]) of the developed coatings and as-received powder were captured at a working distance of 10 mm and under high-vacuum conditions.Three topcoat SEM images captured at 250Â magnification were used to measure the crack density (number of vertical cracks) of the as-sprayed coatings.The color contour of the EDX elemental maps indicates the elemental percentile within the field of view, depending on the concentration of the element.
The coating thickness was measured on different five SE coating cross-sectional images, captured at an Â250 magnification covering 500 microns.For porosity evaluation, five BSE images captured at an Â3000 magnification covering 20 μm were used.The coating thickness and porosity were measured using the ImageJ analysis suite, NIH, USA.The Raman spectroscopy was carried out in LabRAM HR Spectrometer, Horiba Jobin YVON, Japan, with a modified XYZ stage supplied by Marzhauser, Germany.The calibration of the instrument was carried out on a standard Si (100) sample at a reference band of 520.7 cm À1 .The spectra were obtained with a 532 nm green laser and the measured spectra were normalized by applying linear baseline subtraction using LabSpec 6 software, Horiba Jobin YVON, Japan.

High-Temperature CMAS Exposure
The CMAS exposure was carried out in isothermal conditions in a BFF14/5 3216CP furnace supplied by Elite Thermal Systems Limited, UK.The CMAS powder, which was a mixture of raw inorganic compounds, was added in deionized water at a 10 wt% solid loading and was applied on top of the as-sprayed coatings using a brush.The samples were then dried on a heating plate at 80 °C and the dried samples were weighed to get 15 mg cm À2 CMAS.The CMAS-applied samples were then heated at a 10 °C min À1 rate to 1300 °C and the temperature was maintained for 30 min.The dwell time was carefully selected to allow a balance between exploring the CMAS infiltration in the vertical cracks of the deposited coatings and the formation of reaction products.The CMASexposed samples were then cooled down to room temperature at 10 °C min À1 .

High-Temperature Thermal Cycling Evaluation
The thermal cycling tests were conducted in isothermal conditions in a 051030 Rapid Temp Furnace, supplied by CM Furnaces Inc., USA.Each thermal cycle comprised heating up to 1135 °C in 10 min, dwelling the samples at the temperature for 45 min, and ending the cycle with forced compressed air cooling down to 50 °C.A total of three samples were tested for each topcoat type.The furnace oven had a programmable stage that moved the sample in and out of the heating chamber.A high-definition Webcam (Logitech C930e) was used to capture the images of each sample after each cycle (a picture after every minute until the test is concluded).In this study, the coatings were considered to have failed when a spallation area of more than 25% was reached, and the images captured were analyzed (% material loss) using ImageJ analysis suite (NIH, USA) to evaluate the spalled area based on the top-view of the samples.In our setup, a camera recorded images 24/7 when the FCT was running, except when the samples were inside the hot zone of the furnace (1135 °C) that was not visible by the camera.It was imperative to mention that the spalled area was examined during the cooled stage of the samples, and it was possible that cracks/delaminations might be present underneath the coatings which appeared well Bonded from the top surface.3. Results

Microstructural Characterization of As-Sprayed Coatings
The cross-sectional BSE images of the as-sprayed SPS CS and SHVOF thermal-sprayed DVC topcoat layers are shown in Figure 2. The diffraction patterns of the as-received YSZ powder and the as-sprayed topcoat layers of the same powder are shown in Figure 4a.The as-received YSZ powder indicated a t'-YSZ phase (PDF-82-1242) and no major phase transformation was observed in all as-sprayed topcoats, regardless of the spray method and suspension medium used.

High-Temperature CMAS Exposure
The cross-sectional BSE images and EDX analysis of the CMASexposed SPS-CS-E topcoat layer are shown in Figure 5a.No apparent topcoat delamination was caused by the CMAS exposure.The TGO growth along the topcoat-bond coat interface is evident and the TGO layer was %2-4 μm thick.Figure 5b indicates that the top surface, which was in direct contact with CMAS, degraded more, with Y depleted down to 2.9 wt% (EDX 1).Whereas, the EDX point 2 in the interior of the coating and away from the top surface gives a 7.3 wt% Y.The EDX elemental analysis may not directly indicate the composition of YSZ present; thus, the Y/Zr ratio is used as an indicator of yttria depletion.The as-sprayed coatings, being the nominal composition, have a Y/Zr ratio of around 0.13, computed based on the molecular weights.The EDX point 1 indicates a Y/Zr ratio of 0.05, suggesting an yttria-depleted zone on the top part of the coatings.The EDX point 2, being in the interior of the coating and away from the CMAS interacted zone, indicates a slightly lower Y/Zr ratio as compared to that of the pristine composition.In addition, the elements Si and Ca are referred to as evidence of CMAS penetration in the coating as silica and calcia constitute major components of CMAS powder used for this study.Figure 5c shows that the infiltrated CMAS reacted within the vertical cracks, with a Y/Zr ratio down to 0.07 and the elemental Si and Ca penetration at 4.2 and 3.0 wt%, respectively (EDX 3). Figure 5d also indicates the initiation of horizontal cracks along the vertical cracks.An elemental reading of 11.6 wt% Si and 10.2 wt% Ca at the EDX point 4 in Figure 5d shows that the CMAS infiltrated throughout the topcoat and down to the bond coat layer; whereas, the Y/Zr ratio remains around the pristine composition ratio.
The cross-sectional BSE images and EDX analysis of the CMAS-exposed SHVOF-DVC-E topcoat layer are shown in Figure 6.Overall, the CMAS was largely restricted along the vertical cracks, with no apparent horizontal cracking.Figure 6b,c shows the CMAS mainly reacted on the top surface and along the vertical cracks of the SHVOF-DVC-E topcoat.The EDX point analysis (EDX points 1-3) suggests that the interior of the topcoat remains largely unreacted, indicating a near nominal composition and a Y/Zr ratio of 0.06 at the CMAS interacted top part of the YSZ layer.Figure 6d indicates that the vertical cracks did not show horizontal cracking near the bond coat layer, while some horizontal cracking can be seen on the SPS-CE-S topcoat layer, as shown in Figure 5c.The EDX point 4 (Figure 6d) indicates the CMAS infiltrated through the coating thickness; however, the elemental presence of CMAS was nearly halved as compared to that of SPS-CE-E topcoat CMAS infiltration and the YSZ composition seems to remain unaltered.
The cross-sectional BSE images and EDX analysis of the CMAS-exposed SHVOF-DVC-W topcoat layer are shown in Figure 7.The SHVOF-DVC-W topcoat developed vertical cracks following high-temperature exposure.It is imperative to mention that no clear vertical cracks were observed on the as-sprayed SHVOF-DVC-W topcoat layer, shown in Figure 2c.Some horizontal cracks in the bond coat layer are also evident as shown in Figure 7a.A clear distinction can be drawn between the CMASreacted top surface and the interior of the topcoat as shown in Figure 7b.The EDX point 1 indicates yttria depletion down to 0.05 Y/Zr ratio, and the CMAS penetration at 3.4 wt% Si and 2.4 wt% Ca.Whereas, the EDX point 2 in the interior of the topcoat shows Y at a slightly lower Y/Zr concentration as compared to the pristine YSZ composition.The CMAS penetration was severely restricted along the vertical cracks, shown in Figure 7c, and a limited CMAS penetration in the interior of the topcoat, indicated in the EDX point 3.As the CMAS infiltrated through the topcoat and toward the bond coat layer, shown in Figure 7d, the CMAS reactions along the vertical crack reduced and the EDX point 4 indicated a minimal Y depletion as compared to that of the EDX point 1 on the top surface of the same topcoat.Overall, the CMAS seems to largely restrict along the vertical cracks and no apparent horizontal cracks are evident on the SHVOF-DVC-W topcoat layer.The yttria depletion around the CMAS-interacted top part of the coatings remains nearly the same on all three studied coatings.It is imperative to mention that the EDX is a semiquantitative characterization technique and these are preliminary observations based on the available results.Thus, accurate estimations could not be established as there is insufficient data to ascertain such claims.
The diffraction patterns of the CMAS-exposed topcoat layers are shown in Figure 4b.Four main phases were identified across the studied topcoats: t'-YSZ (yttria-rich, PDF-82-1242), monoclinic YSZ (yttria-lean, PDF-81-1315), diopside (CaMgSiO 7 , PDF-75-1092), and wollastonite (CaSiO 3 , PDF-27-0088).The wollastonite phase was only detected in the SPS-CS-E topcoat layer.The Raman analysis on all three topcoat layers after CMAS exposure is shown in Figure 5e, 6e, and 7e.Both the ethanol-based topcoat layers (SPS-CS-E and SHVOF-DVC-E) indicated monoclinic zirconia spectra at the top part of the topcoat, i.e., at the top CMAS/topcoat interface, and no such spectra were detected on the SHVOF-DVC-W topcoat layer.It is imperative to mention that the Raman spectroscopy is a site-specific characterization technique.
An EDX map of the CMAS infiltration throughout the topcoat thickness for all three topcoats is shown in Figure 8. Overall, the molten CMAS infiltrated via the vertical cracks up to the bond coat layer on all three topcoat layers and the formation of a TGO layer is evident near the bond coat layer.As compared with the SPS-CS-E and SHVOF-DVC-W topcoat layers, the infiltration of CMAS in the vertical cracks of the SHVOF-DVC-E topcoat layer was significantly reduced.Despite the formation of uncontrolled, dissimilar vertical cracks in the SHVOF-DVC-W topcoat layer (no vertical cracks in the as-sprayed SHVOF-DVC-W topcoat shown in Figure 2c), the CMAS attack was substantially mitigated and a complete or partial spallation was not observed across the coating.The EDX maps of Figure 8a,b also indicate the formation of spinels (red Mg deposits along the TGO growth).Figure 8a may indicate the CMAS infiltration via the horizontal cracks and/or interpass porosities.
To further examine the horizontal infiltration of CMAS, SEM images and EDX elemental mapping of CMAS penetration along the vertical cracks were analyzed, shown in Figure 9. Figure 9a shows the horizontal cracking and the subsequent CMAS penetration in these cracks.Some isolated CMAS readings (away from the vertical cracks and at the top right of the EDX map) could indicate the CMAS penetration in the interior of the bulk topcoat layer via passages created due to the horizontal cracking of other vertical cracks.As compared to the SPS-CE-S topcoat layer, the DVC structures (both water and ethanol) indicate a reduction in horizontal cracking along the vertical cracks and CMAS penetration is restricted along the top surface and in the vertical cracks.Among the DVC topcoat layers, the SHVOF-DVC-E topcoat seems most promising in terms of CMAS resistance.Overall, the Y-depleted regions are also evident along the CMAS interacting areas and correlate with the CMAS elemental mapping and the EDX point analysis carried out in this section.

Furnace Thermal Cycling (FCT) Behavior
A detailed evaluation of high-temperature thermal cycling behavior on the studied topcoat layers is listed in Table 2 and the SEM images and EDX elemental maps are shown in Figure 10 and 11.In this study, the coatings are categorized to have failed if the coating degradation reaches >25% (analyzed after each thermal cycle and on a cooled sample).The mechanism of topcoat layer spallation is not well defined; however, the growth of microscopic cracks leads to a percolating structure and the linking up and coalescing of these cracks cause the coating to spall.Before a complete spallation of topcoat layers occurs, the topcoat shows macroscopic, localized spallations, which is termed "flaking" in this study.Overall, a complete topcoat failure was observed on SPS-CS-E (85th thermal cycle) and SHVOF-DVC-W (36th thermal cycle) topcoat layers, whereas the SHVOF-DVC-E layer did not experience a complete topcoat spallation up to 100 thermal cycles.It is imperative to mention that a direct comparison between the SPS and SHVOF coating structures could not be drawn due to the difference in the coating thicknesses.Thus, different mechanisms of topcoat degradation can be explained for the DVC and CS topcoat structures.
For the DVC topcoat layers (both ethanol and water) shown in Figure 10a, the topcoat layer indicated flaking (in other words, slow degradation and chipping of the topcoat layer); whereas, for the SPS-CS-E topcoat layer shown in Figure 11a, the topcoat spalled instantly, i.e., without any apparent flaking.The cross section of the spalled SPS-CS-E topcoat layer is shown in Figure 11b.A closer inspection of the EDX elemental maps of the same topcoat layer shown in Figure 11c indicates traces of Al and O elements on the spalled topcoat layer, which highlights a failure mainly at the topcoat/bond coat interface.The cross-sectional microstructure of the thermal cycled SHVOF-DVC-E topcoat, shown in Figure 10b, indicates a mixed-mode failure of topcoat degradation and localized topcoat spallations at the TGO-topcoat interface.The formation of horizontal cracks along the vertical cracks is prevalent; however, a majority of the coating remains intact despite obvious flaking across the observed area, as shown in Figure 10a.
The diffraction patterns of the studied coatings following high-temperature FCT are shown in Figure 4c.In all three FCT-exposed topcoats, no phase transformation was observed as compared to that of the as-sprayed topcoat layers.In this study, a zirconia TBC with 7.5 wt% yttria was examined, which typically crystallizes as the metastable, non-transformable tetragonal (t 0 ) phase and remains supersaturated at all temperatures of interest, up to 1400 °C. [39]The results are in agreement with another relevant study, [40] which indicated a stable tetragonal phase at prolonged high-temperature exposure, i.e., at 1340 °C and up to 72 h.Another study indicated a small <1% monoclinic phase transformation [39] ; however, no such change is observed in this study.

Discussion
The durability of a TBC system, which is initially a three-layered system: substrate, bond coat, and the topcoat (shown in Figure 2), extends to depend on a combination of several phenomena and their interaction on a four-layered system (shown in Figure 8), i.e., an addition of a TGO layer at the bond coat/topcoat interface.These phenomena may include the following: [41] 1) unsustainable, parabolic, and time-dependent TGO growth along the bond coat/topcoat interface, 2) generation of internal stresses due to their thermal expansion mismatch, 3) crack/buckling initiations, growth, and their coalescence, 4) depletion of aluminum in the bond coat and degradation of bond coat strength following high-temperature thermal cycling, and 5) the bond coat rumpling.Triggering a specific combination of interactions and their subsequent failure mechanisms depend on a number of physical factors, such as testing temperature, cycle duration, contaminations, coating processing techniques, material chemistry and their surface characteristics, and lastly, the coatings physical attributes, such as porosity, microstructures, and vertical cracks, etc. [42] 4.1.High-Temperature CMAS Exposure A CMAS attack at temperatures <1400 °C on a stabilized YSZ composition (7-8 wt% yttria) may induce a series of phenomena: [43][44][45][46] 1) penetration of molten CMAS in the open porosities and cracks within the TBC structure, 2) Y 3þ and Zr 4þ react with CMAS, 3) leading to a precipitation of Y-lean ZrO 2 grains that occurs (i.e., tetragonal, t 0 , to monoclinic m phase transformation upon cooling), and lastly, 4) the grain coarsening in fully cooled TBC layer.A schematic of CMASinfiltrated TBC structures is shown in Figure 12.
The CMAS infiltration on the studied topcoats can be seen in the SEM images shown in Figure 5-7, which is further supplemented with EDX point analysis carried out on Y-depleted regions in the same figures indicating Y-lean zirconia grains.The t 0 -m transformation occurs due to a higher solubility of Y 3þ in CMAS as compared to that of Zr 4þ (Zr 4þ diffuses relatively rapidly and the solubility limit is reached), which leads to depletion of yttrium and formation of Y-depleted monoclinic phases upon cooling. [20,46]The Y-depleted regions on the top surface and along the vertical cracks where CMAS reacted can be seen in EDX elemental maps shown in Figure 8 and 9, and the monoclinic zirconia phase was also confirmed via XRD and Raman analysis, indicated in Figure 4-6.The Raman spectra on SHVOF-DVC-W topcoat layers did not indicate the presence of a monoclinic zirconia phase, and this could be due to the CMAS reacted area being limited as the CMAS penetration was largely restricted in the interior of the topcoat layer.Furthermore, the CMAS acts as a sink for Y 3þ , i.e., the Y content in the CMAS composition is relatively low, and the absorbed content continuously diffuses into the wider CMAS deposits.Thus, CMAS deposits maintain a low Y content locally, which continues to deplete the YSZ grains as it penetrates deep inside the TBC structure. [40]he penetration of CMAS in the horizontal openings along the CS is shown in Figure 9a and 12a.The Y 3þ solubility does not considerably affect the composition and flowability of CMAS.On the flowability of CMAS, it is imperative to indicate that the molten CMAS has a low viscosity (≤3 Pa s À1 ) at temperatures ≥1300 °C [47] and it is expected the CMAS would penetrate the entire thickness within minutes, provided the microstructures do not induce any hindrance.The penetration of CMAS up to  or near the region of the bond coat layer can be seen in EDX elemental maps shown in Figure 8.The CMAS penetration in the interior of the topcoat layer along the vertical cracks differentiated among the water-and ethanol-based DVC topcoat layers, as depicted in Figure 12b,c, respectively.The CMAS mainly reacted on the top surface and along the vertical cracks in the case of SHVOF-DVC-W topcoat layers, which is debated in length in Section 3.2 and the microstructures are shown in Figure 7.However, in the SHVOF-DVC-E topcoat layers, the CMAS penetration along the vertical crack widened as CMAS infiltrates down to the bond coat layer.The phenomenon is depicted in Figure 12b (dotted black and orange lines).This could be due to a higher porosity of the SHVOF-DVC-E topcoat layer as compared to the SHVOF-DVC-W topcoat layer.
The SHVOF-DVC-W topcoat was densest among the studied layers, which could be associated with larger in-flight droplets.The behavior of in-flight liquid atomization can be ascertained by two dimensionless parameters, Reynolds number, Re, and Weber number, We.Ethanol has a higher value of both parameters, [48] which induces enhanced atomization, leading to finer droplets and increased porosity in the final deposits.A denser microstructure was expected of the water-based suspension as the molten splats are comparably larger than that of the ethanolbased suspension. [48]In terms of ethanol-based suspensions, the HVOF thermal spray splats are comparably thinner and are accelerated at much faster speeds as compared to the SPS splats, thus, depositing in denser, less porous structures. [34]urthermore, the absence of vertical cracks in the as-sprayed SHVOF-DVC-W topcoat layer, shown in Figure 2c, could be due to insufficient flame temperature and time available for the topcoat layers to expand and contract during coating deposition, which mainly forms vertical cracks in the resultant topcoat layers.The vaporization of water (29.75MJ kg À1 ) consumes significantly more heat as compared to ethanol (2.26 MJ kg À1 ), which decreases the overall flame temperature and increases the chamber pressure.An enhanced chamber pressure amplifies the overall gas velocity and results in faster in-flight droplets. [33]

High-Temperature Thermal Cycling Behavior
The mechanisms in the TBC multilayered system during a hightemperature thermal cycling exposure can be explained via 1) physical attributes and growth of the TGO layer, 2) mechanics within the TGO layer, and 3) vertical and horizontal crack widening due to expansion and contraction of TBCs with every other thermal cycle.These phenomena are depicted in Figure 13.Apart from these factors, the material displacements within the TBC structure are also dependent on heating and cooling rates in thermal cycling, especially in the testing temperatures above 1100 °C. [49,50]The TGO layer forms due to the diffusion of aluminum in the bond coat with the porous TBC topcoat during the high-temperature exposure. [15]At an early stage of thermal  cycling, the fracture toughness of TGO increases, up to %2.5-3.5 MPa p m, with an increase in their thickness and the damage is limited to internal, small cracks within the layer.However, when the critical thickness of %5 μm is reached, the fracture toughness of TGO rapidly decreases to %2.0 MPa p m and the coalescence of smaller cracks takes place to form long cracks that may lead to delamination. [10]The CS topcoat layers fail due to crack propagation at or above the TGO interface and/or due to reaching the critical TGO thickness.The porous CS coatings have a higher segmentation density (crack density), which means the crack has a shorter distance it needs to propagate.The DVC coatings benefit from a reasonable segmentation density, as the microstructure becomes more tolerant of thicker TGO growth as it is more difficult to propagate the horizontal cracks.In this study, the SPS-CS-E topcoat layer failed largely at the topcoat/TGO interface, which may suggest that the critical thickness of TGO was reached (%4.5 μm in this study).
The mean stresses in the TGO change considerably with the evolving thermal cycles and are initially high due to increasing compressive stresses induced by the growth of the TGO layer and the thermal expansion mismatch with the underlying alloy. [15]In addition to TGO growth, the mixed oxide growth, shown in Figure 10 and 11, is also detrimental to the spallation lifetime.As the growth of the evolving TGO layer is constrained by the underlying bond coat surface and the TBC topcoat, the TGO deforms to conform to the shape of the interface and a change in shape generates the triaxial stresses from the initial biaxial compressive stresses. [15]The triaxial stresses mainly consist of existing compressive stress and the out-of-plane tensile stress, which force separations within the TGO layer, or in other words, the failure of the TGO layer.
Overall, as compared to the SPS cracked CS, the DVC structure was more effective in relieving the interfacial tensile stresses (reducing overall stiffness in the coatings), thus reducing the TBC tendency to crack and spall. [51,52]It is imperative to mention that the coating thickness has a strong impact on FCT tests in the case of APS TBCs, and the thicker coatings are likely to spall sooner. [53]Furthermore, at present, the topcoat spallation in FCT after each cycle is determined on the cooled samples [54][55][56] and IR thermal imaging during the testing temperature and cooling cycle could indicate the true extent of the delamination after each cycle.The volumetric changes associated with hightemperature exposure led to horizontal cracking along the vertical cracks, or in other words, branching of the vertical cracks in the DVC structure.The lateral elongation of these horizontal cracks initiated a vertical crack-to-vertical crack spallation, shown in Figure 10 and depicted in Figure 13b.In the case of the water-based DVC structure, the likelihood of a TGO failure increased with an increase in the number of localized spallations (flaking) and their coalescence with succeeding thermal cycles.Furthermore, the material-associated degradation in the SHVOF-DVC-W topcoat layer could have accelerated due to a more dense structure as it leaves less room for thermal expansion in case of high-temperature exposure and induces more cracking.
To conclude, the CMAS attack on the SHVOF-DVC-E topcoat layer may suggest a slightly higher infiltration as compared to that of the SHVOF-DVC-W topcoat layer; however, the ethanolbased DVC topcoat produces an optimal microstructure that offers a balanced trade-off between the CMAS and thermal cycling performance.

Conclusions
The suspension thermal-sprayed 7.5YSZ TBCs were investigated to understand the CMAS infiltration and thermal cycling behavior.A comparison was drawn to study the use of SHVOF thermal spray to deposit DVC TBC coatings and the SPS columnar TBC coatings.Furthermore, the effect of the suspension medium on the final SHVOF DVC structure was also analyzed in terms of their CMAS and furnace cycling test (FCT) performance.
The molten CMAS infiltrated down to the TGO/bond coat layer in all as-sprayed topcoats; however, the CMAS penetration did not induce a partial or complete topcoat spallation.The post-CMAS characterization indicated monoclinic YSZ phases, which imply a martensitic transformation upon cooling, i.e., from the stabilized t'-YSZ phase to the yttria-lean monoclinic phase.In FCT, the SHVOF-sprayed ethanol-based DVC topcoat indicated a 22% material loss at the 100th cycle (the failure criterion was set at >25% material loss).Whereas, the SHVOF water-based DVC topcoat showed a 74% material loss at the 50th cycle.Both DVC topcoats induced flaking, whereas an instantaneous topcoat spallation (complete removal, without apparent flaking) was observed on the SPS columnar at the 76th cycle.The characterization of the spalled topcoat appears to indicate instant spallation at the TGO-topcoat interface.
The thermal spraying method, coating design, and choice of suspension medium seem to indicate a considerable change in the CMAS and thermal cycling performance.The ethanol-based SHVOF DVC topcoat demonstrated promising performance, i.e., the coating sustained 100 thermal cycles.The topcoat did not completely delaminate in the case of FCT and CMAS infiltration and severely restricted the CMAS infiltration in the vertical cracks.The promising performance could be attributed to optimized porosity and DVC structure, i.e., relieving the interfacial tearing stresses without further cracking or widening of the existing DVC structure managing the TBC tendency to crack and spall.Therefore, it can be posited that the SHVOF DVC coatings represent a promising avenue for advancing coating technology beyond the limitations of the conventional lamellar APS and columnar EB-PVD coatings.

Figure 1 .
Figure 1.a) The backscattered electron and scanning electron microscope (BSE SEM) image and b) particle size distribution of the yttria-stabilized zirconia (YSZ) powder.
Figure2ashows the two-layer coating system: the TBC topcoat and MCrAlY bond coat on Inconel 718 as the substrate.Overall, the topcoat layers seem to have well adhered with the bond coat layer, irrespective of the thermal spray methods.The coating thickness has reduced on both SHVOF-sprayed DVC topcoat layers as compared to that of the SPS CS topcoat.The SPS-CS-E topcoat was %280 μm thick (5.6 μm pass À1 ), while the SHVOF-DVC-W and SHVOF-DVC-E topcoat layers were %80 μm thick (2 μm pass À1 ).The disparity in coating thickness and the material deposited per pass could be due to the change in deposition technique or the spray parameters, such as the suspension feed rate, number of passes, step size, and surface speed.The SPS-CS-E topcoat was most porous at 17% porosity, followed by the SHVOF-DVC-E topcoat at 9% and the SHVOF-DVC-W topcoat at 4%.The DVC structures (produced with water and ethanol) indicate a uniform distribution of porosity across the topcoats, whereas the SPS CS structure shows larger pores and interpass porosities throughout the topcoat.The crack density (number of vertical cracks) on the SPS-CS-E topcoat was %8-9 and %6-7 cracks mm À1 on the SHVOF-DVC-E topcoat.Some vertical crack initiation can be seen on the SHVOF-DVC-W topcoat; however, no vertical cracks were formed.The surface morphology of the as-sprayed coatings is shown in Figure3.The SPS-CS-E coatings were composed of larger topographical features as compared to the SHVOF-sprayed DVC coatings.Between the DVC topcoat layers, the SHVOF-DVC-W topcoat layers seem to have much finer features as compared to the SHVOF-DVC-E topcoat layers.The vertical crack initiations were evident on SPS-CS-E and SHVOF-DVC-E coatings, and these initiations seem to be absent on the SHVOF-DVC-W coatings.Under higher magnification, the spherical features indicate different solidification processes, i.e., the SHVOF-sprayed coatings were more densified as compared to the SPS coatings.

Figure 3 .
Figure 3.The surface morphology (top view) of the as-sprayed topcoat layers: a) SPS-CS-E, b) SHVOF-DVC-E, and c) SHVOF-DVC-W.The left-hand-side images were taken with an secondary electron (SE) detector and the right-hand-side images with a BSE detector.

Figure 5 .
Figure 5.The cross-sectional BSE images, energy-dispersive X-Ray (EDX) point analysis (in wt%), and Raman analysis of SPS-CS-E topcoat layer: a) overall microstructure, b) magnified image of inset b,c) magnified image of inset c,d) magnified image of inset d, and e) Raman spectra.

Figure 6 .
Figure 6.The cross-sectional BSE images, EDX point analysis (in wt%), and Raman analysis of SHVOF-DVC-E topcoat layer: a) overall microstructure, b) magnified image of inset b, c) magnified image of inset c, d) magnified image of inset d, and e) Raman spectra.

Figure 7 .
Figure 7.The cross-sectional BSE images, EDX point analysis (in wt%), and Raman analysis of SHVOF-DVC-W topcoat layer: a) overall microstructure, b) magnified image of inset b, c) magnified image of inset c, d) magnified image of inset d, and e) Raman spectra.

Figure 8 .
Figure 8.The cross-sectional BSE images of microstructures, EDX elemental maps, and EDX elemental maps with respect to Si elemental percentile of a) SPS-CS-E, b) SHVOF-DVC-E, and c) SHVOF-DVC-W topcoat layers.

Figure 9 .Figure 10 .
Figure 9.The cross-sectional BSE images of microstructures, EDX elemental maps, EDX elemental maps with respect to CMAS elemental percentiles, and EDX elemental maps with respect to Y elemental percentile of a) SPS-CS-E, b) SHVOF-DVC-E, and c) SHVOF-DVC-W topcoat layers.

Figure 11 .
Figure 11.a) The as-sprayed topcoat SPS layer (before) and the same layers after high-temperature thermal cycling exposure (after 85th cycles).b) The cross-sectional BSE images and c) EDX elemental maps of the SPS-CS-E topcoat layer (at the time of failure, 85th cycle) and SHVOF-DVC-E coatings (after the 100th cycle).

Figure 13 .
Figure 13.A depiction of the high-temperature thermal cycling behavior on a) SPS CS and b) SHVOF-sprayed DVC structure.

Table 1 .
Spray parameters used in this study.