Improved Adhesion of Different Environmental Barrier Coatings on Al2O3/Al2O3‐Ceramic Matrix Composites

In high‐temperature combustion atmospheres, well‐adhering environmental barrier coatings (EBCs) are required to protect the underlying ceramic matrix composites (CMCs) from corrosion. Herein the adhesion mechanisms of three different coatings produced by atmospheric plasma spraying (APS) on an Al2O3/Al2O3‐CMC are investigated. In particular, the influence of surface structuring by laser ablation prior to coating production is investigated. Y2O3, yttria‐stabilized zirconia (YSZ), and Gd2Zr2O7 are chosen as potential EBCs. The coating adhesion on CMC‐substrates with and without surface structuring is analyzed by furnace cycling, pull‐adhesion tests, and burner‐rig tests with gradient. Special interest is paid to the interactions at the coating–substrate interface before and after heat treatment and their effect on the coating adhesion and lifetime. Two different adhesion mechanisms are found: adhesion promoted by chemical reaction and adhesion promoted by mechanical interlocking.


Introduction
Among other things, the efficiency of a gas turbine is limited by its maximum service temperature, which is determined by the thermomechanical capabilities of the components in the hightemperature section of the turbine. Decades of research are therefore aiming for solutions that enable a further increase in the turbine inlet temperature. In the past years, the maximum service temperature could be drastically increased by the use of thermal barrier coatings (TBCs) and complex cooling systems. Unfortunately, the extra work that is needed for cooling causes high-efficiency losses. Calculations showed that operation at 1300 C without cooling could increase the output power by almost 50%. [1,2] Another factor, especially relevant to applications in aviation, is the reduction of weight. Any weight reduction reduces fuel consumption not only by the amount of the original mass which does not need to be accelerated, but also by the mass of the saved fuel. Therefore, the desire to further optimize the process generates the demand for lighter materials with a higher thermomechanical stability compared with the singlecrystal superalloys that are frequently used up to now.
Ceramic matrix composites (CMCs) based on oxides, for example, alumina, are suitable materials to meet the increasing requirements for high-temperature applications in gas turbine environment. This is due to their excellent temperature stability, [3,4] as well as due to their reduced weight compared with current state-of-theart nickel-based alloys. Although alumina is quite inert, water vapor that is present in the aggressive turbine atmosphere (%10%) [5,6] can cause severe corrosion, as it leads to the formation of volatile hydroxides at temperatures above 1200 C. [7] Therefore, the application of suitable environmental barrier coatings (EBC) to mitigate this corrosion is inevitable for long-term use.
Obviously, the ideal EBC should have high corrosion resistance and should be stable at high temperatures. To identify potential EBC candidates, corrosion rates of several materials have been studied. [8][9][10][11][12][13][14] Water vapor corrosion rates of an Al 2 O 3 / Al 2 O 3 -CMC and possible coating materials are shown in Table 1.
In addition to a high water vapor resistance, a dense microstructure is preferred to suppress the diffusion of water through the coating. However, a dense coating can cause high stresses during thermal cycling, leading to premature failure. To achieve a long service life, a low coefficient of thermal expansion (CTE) mismatch between coating and substrate is advantageous. Considering that, materials such as Y 2 O 3 and yttrium aluminates appear to be more suitable than yttria-stabilized zirconia (YSZ) and Gd 2 Zr 2 O 7 .
The suitability of Y 2 O 3 as EBC for oxide-based WHIPOX CMCs has been studied by Mechnich et al. and was also demonstrated on real combustor components. [23,24] Prior to coating by atmospheric plasma spraying (APS), a reaction-bonded aluminium oxide (RBAO) bond coat was applied to increase the adhesion of the coatings to the substrate and reduce the gas permeability of the sample. [25] The formation of a reaction layer consisting of the series of yttrium aluminates (Y 4 Al 2 O 9 , YAlO 3 , DOI: 10.1002/adem.202000087 In high-temperature combustion atmospheres, well-adhering environmental barrier coatings (EBCs) are required to protect the underlying ceramic matrix composites (CMCs) from corrosion. Herein the adhesion mechanisms of three different coatings produced by atmospheric plasma spraying (APS) on an Al 2 O 3 / Al 2 O 3 -CMC are investigated. In particular, the influence of surface structuring by laser ablation prior to coating production is investigated. Y 2 O 3 , yttria-stabilized zirconia (YSZ), and Gd 2 Zr 2 O 7 are chosen as potential EBCs. The coating adhesion on CMC-substrates with and without surface structuring is analyzed by furnace cycling, pull-adhesion tests, and burner-rig tests with gradient. Special interest is paid to the interactions at the coating-substrate interface before and after heat treatment and their effect on the coating adhesion and lifetime. Two different adhesion mechanisms are found: adhesion promoted by chemical reaction and adhesion promoted by mechanical interlocking. and Y 3 Al 5 O 12 ) was observed after exposure to high temperatures. As the yttrium aluminates have similar CTEs compared with coating and substrate, the reaction layer was classified as not detrimental. On the contrary, the good bonding between coating and substrate was attributed to the formation of a reaction scale. [23] Recently, water vapor corrosion tests of an Y 2 O 3 -coated Al 2 O 3 / Al 2 O 3 -CMC demonstrated its high potential as EBC. [16] Following the arguments of well-matching CTEs, the yttrium aluminates Y 3 Al 5 O 12 and YAlO 3 are promising coating candidates as well. However, it is known from literature that the application of yttrium aluminate coatings by APS is quite challenging. [26,27] This is due to the unfavorable volume shrinkage of partially amorphous coatings during recrystallization. In a recent study of YAlO 3 EBCs on Keramikblech FW12, we demonstrated that problems such as phase segregation and pore formation can be overcome using very low pressure plasma spraying (VLPPS). [26] Using low-pressure conditions, wellmolten and fast particles, as well as higher substrate temperatures and lower cooling rates, can be achieved, resulting in dense and fully crystalline coatings. Similar to the Y 2 O 3 coatings, the formation of a reaction layer between EBC and alumina-based CMC substrate was observed. In contrast to the APS process, in the VLPPS process, the formation of a reaction scale was observed directly after spraying. Pull-adhesion tests (PATs) showed that the adhesion of YAlO 3 -VLPPS coatings was significantly higher than for APS coatings, which lacked a reaction layer. Therefore, it is likely that the good coating adhesion can be associated with the formation of reaction scale at the coatingsubstrate interface. Furthermore, VLPPS-YAlO 3 coatings showed a long service life in burner-rig tests with temperature gradients. It should be noted that no additional bond coats or surface treatments were required in our study.
The impact of the reaction scale at the coating-substrate interface on the coating adhesion/lifetime is especially interesting in the context of Gd 2 Zr 2 O 7 coatings. Gd 2 Zr 2 O 7 is a commonly used material in the TBC area and is known for its high resistance against fly ash attack. [28] As the material has already been wellinvestigated and a resistance against steam and hot corrosion was found, it is an interesting candidate for EBCs despite its large CTE mismatch (Δ CTE ≃ 3.9 Â 10 À6 K À1 ). Studies in the TBC area showed that a reaction between Al 2 O 3 and Gd 2 Zr 2 O 7 to GdAlO 3 can occur. [21,29,30] In contrast to the EBCs where a reaction of Y 2 O 3 or yttrium aluminate coatings with the substrate is considered to promote adhesion, the formation of GdAlO 3 at the Al 2 O 3 -Gd 2 Zr 2 O 7 interface was associated with premature failure of the coating. For this reason, there are only few studies concerning Gd 2 Zr 2 O 7 as EBC. Recently, Gatzen et al. [31] showed that the coating application of Gd 2 Zr 2 O 7 coatings on Al 2 O 3 /Al 2 O 3 -CMCs (Keramikblech FW12) by APS is possible. However, the coating adhesion on untreated substrates was quite poor which was explained by bad wetting behavior of the coating. The adhesion was drastically increased by surface roughening using laser ablation. Contrary to the experiments with YAlO 3 , no reaction between coating and substrate was observed directly after spraying. [32] Because no high-temperature tests were conducted, whether a potential reaction at the interface favors adhesion remains unanswered.
Another material known in the TBC field is YSZ. Unlike Gd 2 Zr 2 O 7 , Y 2 O 3, and the yttrium aluminates, no reaction between YSZ and the alumina-based substrate is expected. Despite its rather high CTE mismatch (Δ CTE ≃ 3.2 Â 10 À6 K À1 ), YSZ was tested for its suitability as an EBC, e.g., by Gerendás et al. [33] and Mechnich et al. [34][35][36] The latter showed the promising performance of YSZ coatings produced by electron-beam physical vapor deposition (EB-PVD). Nevertheless, the WHIPOX substrates used in that study had to be pretreated by sandblasting or grinding to ensure a good bonding; afterward, an RBAO [37] bond coat had to be applied to improve the coating adhesion. Gerendás et al. [33] reported a surprisingly long lifetime of APS-YSZ coatings on sandblasted UMOX substrates. In contrast to this, significantly shorter lifetimes were found for YSZ coatings on sandblasted WHIPOX and OXIPOL substrates. This result demonstrates that the results cannot be easily transferred from one coatingsubstrate system to another. Therefore, the question remains whether YSZ is a suitable EBC candidate for the CMC (Keramikblech FW12) used in this study and whether a surface treatment, respectively, which one, can be applied.
However, Vaßen et al. [32] showed recently that mechanical treatment of the brittle ceramic matrix in CMCs may lead to severe damage and premature coating failure. It was suggested to use surface structuring by laser ablation instead. Another suggested option is the usage of an APS-Y 2 O 3 layer as bond coat. [32] Summarizing , we can say that the formation of reaction layers at the interface, as the examples of Y 2 O 3 and YAlO 3 show, may have a positive effect on the adhesion of the coating. Experiments with coatings where no reaction is possible, such as YSZ, show that good surface pretreatment is necessary to increase adhesion. A good substrate pretreatment may allow the use of materials with moderate CTE mismatch. If possible, a method that is not harmful to the substrate should be chosen, such as laser ablation.
Some questions, however, arise from these observations which are as follows: 1) Is it possible to further increase the adhesion of coatings that form a reaction scale (e.g., Y 2 O 3 ) by laser structuring of the substrate?
Experiments in this study aim to answer these questions. For this, a commercially available alumina-based CMC (Keramikblech FW12) was coated by APS. Y 2 O 3 , Gd 2 Zr 2 O 7 , and YSZ coatings were chosen as coating materials. For a better understanding of the adhesion mechanism, some samples were subjected to a surface pretreatment by laser ablation prior to coating application. Samples with and without laser structuring were coated with each coating material. PATs, furnace cycling tests, and burner-rig tests with gradient were conducted to study the coating adhesion. The samples were analyzed by means of X-ray diffraction (XRD) and scanning electron microscopy (SEM). By comparing materials with different potentials to form reactions with the substrate and with clearly different CTE mismatches, it should be possible not only to answer the aforementioned questions, but also to clarify the interaction of the different adhesion mechanisms.

Experimental Section
The CMC used for this study was the commercially available material FW12 (Pritzkow Spezialkeramik, Germany), consisting of alumina fibers (Nextel 610) that are embedded in a porous matrix of 85% alumina and 15% YSZ. All samples were ultrasonically cleaned before coating application.
A Nd:YAG laser with a wavelength of 1062 nm and a maximum peak power of 15 kW was used for the surface structuring of the substrate. Two different surface structures were chosen for surface structuring ( Table 2). The surface structuring is described in detail in a previous study. [31] Different coating materials were chosen as EBC: Y 2 O 3 , Gd 2 Zr 2 O 7 , and YSZ. APS was used for coating manufacture; the coating parameters are shown in Table 3. An Oerlikon Metco MultiCoat facility was used, equipped with a TriplexPro-210 gun that was mounted on a six-axis robot. The substrate temperatures during the coating procedure were measured with an IR pyrometer.
Spray-dried powders were used for coating manufacture ( Table 4). The XRD measurements of the powders are shown in Figure 1. The measurements reveal the high purity of the used powders. The coated samples were furnace cycled in air for 4 Â 20 h at 1200 C. XRD measurements were conducted before and after thermal aging. A Bruker D4 Endeavor (Karlsruhe, Germany) with Cu Kα radiation (λ ¼ 1.54187 Å) was used for XRD measurements.
Coating adhesion of as-sprayed coatings was measured by PATs. For this, an Elcometer 510 (Aalen, Germany) was used according to ASTM D4541. Test dollies with 10 mm diameter were glued to the coated surface with an Araldite two-part epoxy adhesive.
The coating performance during thermal cycling with a thermal gradient was tested by burner-rig tests, and the used setup is described by Traeger et al. [42] The samples were mounted in a ceramic sample holder and subjected to cycles of 5 min of heating followed by 2 min of cooling. The temperature was measured with pyrometers from the front and back side. The used temperature program consisted of 510 cycles at 1200 C, followed by 500 cycles at 1300 C, followed by 230 cycles at 1400 C. The backside temperature was set to 600 C.
Materialographic cross sections were prepared from samples in the as-sprayed state and after testing. The polished samples were sputtered with platinum (Leica EM ACE200, Vienna, Austria) and analyzed by SEM (Hitachi, TM3000).

Coating Formation and Interface Reactions
First, the cleaned CMC substrates were coated without additional pretreatment. XRD measurements of the as-sprayed and heattreated samples are shown in Figure 2. The coatings are highly crystalline and phase pure even in the as-sprayed state. The measurements show no significant difference between the as-sprayed and heat-treated state. Therefore, neither crystallization nor phase transformation are expected to take place during heat treatment. The diffraction pattern of Gd 2 Zr 2 O 7 -coated samples shows the presence of traces of ZrO 2 and Gd 2 O 3 . During plasma spraying, Gd is vaporized, changing the stoichiometry of Gd 2 Zr 2 O 7 and causing demixing. This phenomenon is well known for this material. [46] A typical Y 2 O 3 splat is shown in Figure 3a. The splat has a disklike shape, which is characteristic for the ideal splat morphology, indicating a good wetting behavior of the coating-substrate system. [47,48] Therefore, smooth and good adherent coatings are expected. In contrast to this, the splat of the YSZ coating (Figure 3b) is not well flattened and unmolten/partial molten parts can be observed. The presence of unmolten particles can be attributed to the larger particle size of YSZ powder, compared with Y 2 O 3 and Gd 2 Zr 2 O 7 powders. The low flattening indicates  www.advancedsciencenews.com www.aem-journal.com poor wetting and thus bad contact between coating and substrate. The reduced flattening can be explained by the incomplete melting and the lower substrate temperature during thermal spraying. [49][50][51] As a result of the worse contact and the lower degree of melting, a weak adhesion and higher porosity of YSZ coatings is expected. The SEM image of the Gd 2 Zr 2 O 7 splat is shown in Figure 3c. The splat is irregular shaped and cracked. The high CTE difference between Gd 2 Zr 2 O 7 and substrate causes high stresses during quenching, leading to partial delamination of the splat. A reduced level of adhesion is, therefore, expected. SEM images of the coated samples before and after heat treatment are shown in Figure 4. As expected from the splat morphology, the SEM images of the Y 2 O 3 coatings (Figure 4a) show a dense and homogenous coating microstructure. Only few pores can be observed. After heat treatment, no signs of delamination are observed. In contrast to this, the YSZ coatings offer a more inhomogeneous coating microstructure with a fine distributed porosity (Figure 4b). As a result of the bad contact between coating and substrate, the coating was partly delaminated even in the as-sprayed state. After furnace cycling, the coating was completely delaminated. The SEM images of the Gd 2 Zr 2 O 7 coatings (Figure 4c) reveal dense and homogenous coating microstructures. Although the analysis of the splats suggested otherwise, the coating seems to adhere to the substrate. After heat treatment, a gap between coating and substrate is observed and the coating is partly delaminated.
A closer view at the Y 2 O 3 coating-substrate interface before and after thermal cycling is shown in Figure 5. The formation of a continuous reaction scale can be observed at the interface. This reaction scale is present directly after spraying and grows during furnace cycling. After furnace cycling, a thickness of about 0.3 μm is reached. The formation of three different phases was observed. Based on the Y 2 O 3 -Al 2 O 3 phase diagram [52] and the findings of Mechnich et al., [23] it is assumed that these are YAlO 3 , Y 3 Al 5 O 12 , and Y 4 Al 2 O 9 . Mechnich et al. [23] showed the formation of these phases at the Al 2 O 3 -Y 2 O 3 interface after annealing. In this study, we can show that the reaction of Y 2 O 3 with Al 2 O 3 starts already during the coating process. This is in accordance with our study concerning VLPPS-YAlO 3 coatings. [26] The formation of a reaction layer in the as-sprayed state can be explained by the high substrate temperature during coating and the good wetting of Y 2 O 3 splats. The yttrium aluminates, which are formed at the interface, have similar CTEs, compared with coating and substrate; therefore, coating failure due to the formation of a reaction layer is not expected. In the study by Mechnich et al., [23] it was also assumed that the reaction layer could lead to an improvement in adhesion in this special case. In contrast, no mixing phase is known from the Al 2 O 3 -ZrO 2 phase diagram. [30] As the formation of a reaction phase at the interphase is not expected, it can be assumed that the adhesion of YSZ is based on mechanical interlocking and is not affected by reactions at the interface.
www.advancedsciencenews.com www.aem-journal.com 1200 C, no reaction or diffusion zone can be observed. However, according to the phase diagram [53] and results of Leckie et al., [29] the reactions of Gd 2 Zr 2 O 7 with Al 2 O 3 are likely. Leckie et al. [29] reported a reaction between EB-PVD-Gd 2 Zr 2 O 7 coatings and Al 2 O 3 substrates, leading to the formation of GdAlO 3 and Gd 4 Al 2 O 9 at the interface. The absence of a reaction layer in our experiments might be explained by poor contact between coating and substrate, which slows the diffusion. The SEM image of the Gd 2 Zr 2 O 7 splat (Figure 3c) supports this assumption. The partly delamination of splats hinders interdiffusion and the formation of a reaction layer between coating and substrate. The quite high CTE mismatch between coating and substrate (Δ CTE ≃ 3.9 Â 10 À6 K À1 ) is another driving force for coating spallation and thus prevents the formation of a reaction layer.

Coating Formation on Laser Structured Surfaces
Two different structures were chosen for surface pretreatment: mixed structure and dot structure. [31] The surface structures of both are shown in Figure 7. The so-called mixed structure comprises an irregular wavy and rough surface. In contrast to   www.advancedsciencenews.com www.aem-journal.com this, the dot structure offers a regular surface profile with valleys in defined distances and parts of the untreated surface can be found in between. The initial surface roughness of untreated samples was R a ¼ 2.6 μm. Surface structuring by laser ablation could increase the surface roughness, a value of R a ¼ 3.9 μm was measured for the dot structure. A significant increase to R a ¼ 17.5 μm was achieved for the mixed structure. Nevertheless, both generated structures should offer manifold possibilities for the mechanical interlocking of the coating. Furthermore, the cavities and bumps of the surface are obstacles at which cracks can be deflected or stopped. [54][55][56] Both effects contribute to an increased coating adhesion. The formation of a reaction layer between Y 2 O 3 coating and the untreated Al 2 O 3 substrate did not lead to delamination. On the contrary, the reaction is a sign of good contact between coating and substrate. The presence of a reaction layer indicates that interdiffusion must have occurred at the interface. Because covalent and ionic bonds are the strongest attractive forces, the reaction between coating and substrate helps to promote adhesion and outweighs the effect of interdiffusion. [57] As a consequence, a good adhesion strength during tensile tests and burner-rig tests is expected. SEM images of the structured surfaces coated with Y 2 O 3 are shown in Figure 8a,b. It can be clearly seen that the coating was able to infiltrate the cavities. Again, the Y 2 O 3 coatings exhibit good adhesion. A homogenous coating microstructure is observed with no signs of delamination after furnace cycling.
In contrast to YSZ coatings on untreated CMC substrates, the YSZ coatings on laser structured CMC substrates did not fail during furnace cycling. However, the SEM images (Figure 8c,d) reveal that the coatings on dot-structured samples show partly delamination on the untreated parts of the CMC after furnace cycling. At the same time, the coating seems to be still adherent to the substrate in the laser-generated cavities. No delamination cracks can be observed in the SEM image of the YSZ coating on the substrate with the mixed structure. The higher surface roughness seems to offer enough possibilities for clamping and therefore significantly increases the adhesion.
The SEM images of the cross sections of the Gd 2 Zr 2 O 7 -coated samples in the as-sprayed state and after heat treatment are shown in Figure 8e,f. The bonding of the Gd 2 Zr 2 O 7 coatings seems to be improved on the laser pretreated substrates. In contrast to the coatings on untreated substrates, there are no signs of delamination in the coating systems on structured substrates. The coating-substrate interface looks still intact, even after furnace cycling.

Pull-Off Adhesion and Burner-Rig Performance
To analyze the adhesion properties of the different coating materials on structured and unstructured CMC substrates, PATs and burner-rig tests were conducted. The test results are shown in Figure 9. For better orientation, the lower limit of the average cohesion strength of the CMC substrate (8.8 AE 1.1 MPa) was added. The pull-off adhesion tests of the Y 2 O 3 coatings revealed that the adhesion of these coatings is so strong that all coatings had a higher adhesion strength than the CMC itself. The slight increase in the adhesion strength by previous laser structuring of the substrate is not significant in this system, as the measured strength exceeds the average strength of the Al 2 O 3 /Al 2 O 3 CMC. Furthermore, the failure occurred within the CMC or the coating, not at the coating-substrate interface. This leads to the conclusion that the coating-substrate interface is stronger than the cohesion within the substrate and the coating itself.
In contrast to this, the YSZ and Gd 2 Zr 2 O 7 coatings on untreated substrates failed directly after the beginning of the test. The failure occurred directly at the coating-substrate interface. This result corresponds to the weak coating adhesion and poor wetting of these coatings on untreated substrates, which were discussed previously. The YSZ coating on the dot-structured substrate also failed immediately. This result is quite surprising, as the furnace cycling of these samples suggested a better adhesion. However, the dot-structured CMC was not able to increase the adhesion strength and the delamination occurred directly at  www.advancedsciencenews.com www.aem-journal.com the interface, as already indicated in the furnace cycling experiments. This might be attributed to the large untreated parts between the dots, where insufficient contact between coating and substrate prevails. The measurements show that the adhesion strength of YSZ coatings on the mixed structured substrate was significantly increased. The increased adhesion strength at the interface changed the failure mechanism from failure at the interface to failure within the coating. In case of the Gd 2 Zr 2 O 7 coatings, the adhesion strength could be drastically increased by laser pretreatment of the substrate. The measured adhesion strength is in the range of the cohesion strength of the substrate itself and the failure mode was shifted to the inner part of the CMC. The increased adhesion strength is in accordance with the furnace cycling results discussed in the section mentioned previously.
The burner-rig tests confirm these results. Tests with Y 2 O 3 coatings were stopped after 1010 cycles without macroscopic failure. This lifetime corresponds to the average number of cycles reached for state-of-the-art YSZ TBC systems in this rig (shown in yellow in Figure 9b). The SEM images of the Y 2 O 3 coatings (Figure 9c-e) show that all coatings are still well-adherent so that a significantly longer lifetime can be expected. After about an 80 h burner-rig testing, the reaction scale showed a similar thickness after 80 h of furnace cycling. This indicates that the reaction scale at the interface is thermodynamically and mechanically stable and thus promotes adhesion.
The YSZ coating on the untreated substrate failed after 648 cycles. Considering the weak adhesion and the fact that the coating was already partly delaminated after spraying (Figure 4), this lifetime is surprisingly long. A significantly longer lifetime was observed for the coating on the dot-structured sample (1730 cycles). Although the pull-off adhesion tests suggest a similar weak adhesion as the coatings on the untreated substrates, the furnace cycling indicates a slightly improved adhesion. The results of the burner-rig tests show that the structuring of the substrate almost triples the lifetime. The loose contact between coating and substrate could allow for compensation of the rather large CTE difference between coating and substrate (Δ CTE ≃ 3.2 Â 10 À6 K À1 ), whereas the structuring provides anchor points that prevent delamination. Burner-rig tests of the coating on the mixed structured sample were stopped after 1730 cycles, no macroscopic signs of delamination have www.advancedsciencenews.com www.aem-journal.com been observed. The SEM image of this sample (Figure 9 f-h) shows that the coating is still well-adherent to the substrate. This is in good agreement with the results of the previous tests. According to the increased adhesion strength, a longer cycling lifetime is expected. The Gd 2 Zr 2 O 7 coating on the untreated CMC failed after 520 cycles. The SEM images (Figure 9i) show that delamination took place at the coating-substrate interface. Burner-rig testing of the coatings on the roughened substrates was stopped after 1730 cycles without any failure being detected. The SEM images confirm that no formation of a reaction layer or delamination cracks occurred. Some vertical cracks can be observed in the coating on the substrate with the mixed structure. These cracks seem to be generated at the interface and spread toward the coating surface. The reason for the formation of the cracks is probably the large CTE difference (Δ CTE ≃ 3.9 Â 10 À6 K À1 ) between coating and substrate, leading to tensile stresses within the Gd 2 Zr 2 O 7 coating during cooling. The thermal gradients occurring during thermal cycling (Δ T ¼ 600 C) additionally reinforce these stresses. The lack of these cracks in the coatings on the dot structured samples might be explained by the looser contact between coating and substrate in this system. www.advancedsciencenews.com www.aem-journal.com

Conclusion
The adhesion mechanisms of different EBCs (Y 2 O 3 , Gd 2 Zr 2 O 7 , and YSZ) on an Al 2 O 3 /Al 2 O 3 -CMC were investigated in this study. The coating adhesion was analyzed by means of PATs, furnace cycling tests, and burner-rig tests with thermal gradients. Two different kinds of adhesion were observed: adhesion promoted by chemical bonds between coating and substrate and adhesion promoted by mechanical interlocking. In case of Y 2 O 3 coatings, adhesion is promoted by interdiffusion and formation of a reaction scale between coating and substrate. This layer is formed during plasma spraying and grows during heat treatment to a thickness of about 2 μm. The formation of chemical bonds leads to a high adhesion strength and the long cycling lifetime of Y 2 O 3 coatings on Al 2 O 3 / Al 2 O 3 -CMC substrates, even without pretreatment. Based on the results, no statement can be made regarding the question of whether the adhesion can be improved by laser structuring of the substrate. Alongside the already strong adhesion due to the chemical reaction, the surface structuring seems to have a minor effect.
The formation of chemical bonds between coating and substrate was not observed for Gd 2 Zr 2 O 7 and YSZ coatings, although a reaction between Al 2 O 3 and Gd 2 Zr 2 O 7 would be possible. The absence of a reaction layer is explained by the slow reaction kinetics and the poor wetting between substrate and coating. Because there are no chemical bonds that promote adhesion, mechanical anchoring is the key mechanism. The untreated substrate is relatively smooth and offers few possibilities for clamping, resulting in weak adhesion of Gd 2 Zr 2 O 7 and YSZ coatings. The substrate surfaces were structured by laser ablation prior to coating to increase the roughness and thus improve the coating adhesion. The results show that surface structuring significantly increases the adhesion and thermal cycling lifetime, which allows the use of those materials in EBC application.
The following conclusions can be made: 1) Coating adhesion strongly depends on the surface structure and wettability of the substrate.
2) Chemical bonds might help to drastically increase the coating adhesion.
3) Bad wetting can suppress the formation of a reaction scale at the interface. 4) If mechanical anchoring is the main adhesion mechanism, surface structuring by laser ablation can significantly increase the adhesion strength and lifetime of the coatings.