A specific composition of the CMAS was prepared based on the literature [10] [16] [17] [21] . The following chemical reagents were used: CaO (DINÂMICA QUÍMICA, 95% purity), MgO (CAAL, 95% purity), Al₂O₃ (NEON, 95% purity), and SiO₂ (ADICEL, 94% purity). The oxides were placed in a Petri dish and mixed with a spatula until uniform coloration was achieved. Table 1 details the percentage by weight of each component in the manufactured CMAS. It is important to mention that the constitution of the CMAS may vary depending on the place of origin [18] .

The original TBC sample was fabricated on a carbon steel substrate. The APS technique in an argon-hydrogen atmosphere, using an F4 plasma torch (Oerlikon Metco, F4MB), configured with a 6 mm nozzle and 1.8 mm injector, was employed to deposit both the bond coat and the top coat. The bond coat was made of NiCrAl alloy and the ceramic layer was ZrO₂-7%Y₂O₃ (Oerlikon Metco). A PACE cutter, model PICO 155 Precision Saw, was used to cut the original sample into six smaller specimens. Each of these specimens was approximately 17 × 17 × 4 (Dimensions in mm). Of these six samples obtained, one was left as a spare and the other five were used for the study and were numbered 0, 1, 2, 3, and 4.

For the isothermal treatment cycles, a Lindberg/Blue M Tube Furnace was used. On each isothermally treated sample, approximately 10 mg/cm² of CMAS was manually deposited on its surface before being placed in the furnace. To facilitate reading and understanding of the work, the samples were renamed according to the isothermal treatment applied. Sample 0 did not receive isothermal treatment. The other samples (1 to 4) were initially heated with a linear ramp of 10˚C/min from room temperature (25˚C) to a temperature of 1000˚C. Subsequently, there was a secondary heating process that used a linear ramp of 5˚C/min. Finally, the samples were slowly cooled to room temperature. Table 2 shows the secondary heating process, the length of stay and the designation adopted for the samples.

After isothermal treatment, the samples in Table 2 were cut, embedded in Bakelite, sanded, and polished to obtain a flat, mirror-like surface and evaluate their respective cross sections. To visualize the chemical components on the surfaces of the polished samples, the energy dispersive x-ray spectroscopy (EDS) technique was used. This technique was performed using a JEOL scanning electron microscope (SEM), model JSM-6510.

Photographs taken with conventional equipment show the top view of the TBC samples to analyze the wettability of the CMAS (Figure 1). Figure 1(a) presents sample 0, untreated. Figure 1(b) illustrates sample 1-1200-1 with the deposition of 10 mg/cm² of CMAS before the isothermal treatment. Images of samples 1-1200-1, 2-1200-5, 3-1250-1, and 4-1250-5 after the isothermal treatments are in Figures 1 (c)-(f), respectively.

After the isothermal treatments, the particle size of the deposited CMAS was modified. The material maintained a light coloration, but the grains joined together, visually increasing the grain size of the CMAS. Cracks visible to the naked eye appeared in samples 2-1200-5 and 4-1250-5 (Figure 1(d) and Figure 1(f)). These samples were subjected to a longer exposure time (5 h) at their

respective isothermal treatment temperatures, therefore, the exposure time was more detrimental to the coating than the applied temperature. Wettability was not observed under the tested conditions. The area occupied by the deposited CMAS remained approximately the same before and after isothermal treatments (Figures 1(b)-(f)). Thus, it can be assumed that no silicate was formed or that only a small amount was formed. Two different ways to produce CMAS are found in the literature. In the former, the components are mixed in appropriate proportions and applied to the surface of the samples [16] [22] [23] [24] [25] . In the second, after mixing the components, a heat treatment is done before applying the material to the samples. This heat treatment aims to ensure the formation of silicate [18] [26] [27] [28] [29] . In this study, the first methodology was adopted because it is closer to the operation of gas turbines. It is important to mention that some of the applied material may have become silicate and penetrated the YSZ. Thus, a cross-sectional analysis of the samples is necessary to determine whether or not the molten silicates infiltrated into the TBC.

Figure 2 shows the conventional equipment images (frontal and isometric) of samples 1-1200-1, 2-1200-5, 3-1250-1, and 4-1250-5 after their respective isothermal treatments. Images of sample 0 (no treatment) are also shown for comparison.

The height of the samples in Figure 2 was measured with a caliper. Samples 2-1200-5 and 4-1250-5 had a greater increase in height (around 20%) and consequently in volume. A comparison of samples 2-1200-5 and 3-1250-1 demonstrated that exposure time had a greater influence than the temperature on

volume increase. Buckling can be observed in sample 2-1200-5. Severe corrosion was also observed on the steel substrates (indicated by arrows in Figure 2). This is because the sides of the obtained samples were exposed to the environment after cutting the original sample, causing hot oxidation on the steel substrates during the isothermal treatments.

CMAS can increase the volume of the TBC by penetrating the YSZ. In the presence of molten silicates, the YSZ grains dissolve easily. As a result, a phase transformation occurs in YSZ, which originally had a tetragonal structure that precipitated into a monoclinic structure. This phase transformation is accompanied by an increase in volume. Furthermore, once inside the coating, the CMAS can separate the crystals from the structure, implying a further increase in volume [21] [22] [30] . However, to infiltrate the TBC structure through pores and microcracking, CMAS must melt [31] to form silicate. However, because no wettability was observed, silicate formation probably did not occur, and thus penetration did not occur, or if it did, it was a small amount of material. Therefore, the oxidation corrosion of the substrates is the main factor related to the observed volume increase, since iron oxidation involves the absorption of oxygen from the atmosphere, which implies an increase in the mass of the oxidized substrate.

In addition to the oxidation of the substrate, it should also be mentioned the appearance of some cracks in the samples after the isothermal treatments. As previously mentioned, the original structure of a TBC system consists of two layers (bond coat and top coat) deposited on the substrate. In Figure 3 this structure can be seen by the scanning electron microscope (SEM) image.

The samples tested for a 5-hour interval showed the appearance of cracks in their structures, which is also one of the reasons for the observed increase in volume (Figure 4).

The cross-section of the samples was examined using EDS to observe the permeability of the molten silicates inside the ceramic layer of the TBC. For each analysis elements with a lower percentage of weight and/or irrelevant for this study were disregarded. Figure 5 illustrates the overview of sample 0, which allows the constituent elements of TBC to be identified.

Three distinct regions can be seen in Figure 5. At the bottom, the intense amount of Fe demonstrates that this is the steel substrate. In the upper region, the Zr element appears more intensely, indicating the top coat region (YSZ). The bond coat can be identified by the presence of the elements Ni, Cr, and Al (although Al is distributed throughout the region, a greater concentration is noticeable in the bond coat region). Therefore, a NiCrAl-type alloy is evident in the bond coat.

Figure 6 illustrates an EDS image focused on the ceramic coating of sample 0. In this image, the presence of Y is visible, due to the higher magnification. Oxygen also appears throughout the region. This element is present in zirconia (ZrO₂) and yttria (Y₂O₃). However, the sample surface could also have oxidation, as the sample had contact with the atmosphere after polishing and before being inserted into the sample holder of the SEM.

The ceramic coating of sample 1-1200-1 presented in Figure 7 illustrates the ceramic layer (Zr), bond coat (Ni and Cr), and part of the substrate (Fe). Therefore, it was not possible to identify regions with infiltration of molten silicates in sample 1-1200-1. The short exposure time may not have been sufficient to melt CMAS and penetrate the coating structure; therefore, the deposited material remained on the YSZ surface [31] .

An EDS image of sample 2-1200-5 is in Figure 8. A linear concentration of Ca is recorded in the upper part of the plate, indicating that CMAS is present on the surface of the ceramic layer. However, the upper right region shows a higher concentration of both Ca and Mg. In the same region, no presence of Zr and Y

appears in lower concentrations, which indicates the accumulation of CMAS in the region of YSZ discontinuity. However, this discontinuity may have been caused by the CMAS itself through a chemical corrosion process [16] [20] . The small amount of Y in this region reinforces this hypothesis since a discontinuity due to mechanical friction would imply the complete absence of both Zr and Y. At an intermediate depth, a greater concentration of Ca is noted on the right than in the left region, evidencing that the high concentration in the upper region allowed greater infiltration, including in regions with Zr and Y. This is because the amount of CMAS has a large effect on the severity of the corrosion. Factors that influence the penetration of CMAS inside the TBC are temperature, time, the surface roughness of the YSZ, and the number of molten silicates. The higher the amount of CMAS in the region, the greater the observed degradation effects and the greater the penetration depth achieved [32] [33] .

The ceramic layer of sample 3-1250-1 is illustrated in Figure 9.

The constituent elements of yttria-stabilized zirconia (Zr, Y, and O) are easily identifiable in Figure 9. The elements of interest—Ca, Mg, and Al (CMAS constituents)—appear at higher concentrations above the ceramic coating.

However, a comparison between Zr and Ca indicates an apparent overlap between these elements at the CMAS interface with the YSZ. This means that a small amount of CMAS infiltration occurred in the surface region of the YSZ. The infiltration process of CMAS has the following sequence: 1) Through wettability, the molten CMAS spreads along the surface of the TBC. During this propagation process on the surface, the dynamic contact angle decreases dramatically over time. 2) Upon reaching a steady state of spreading, the infiltration process begins, which is believed to start through the surface defects, such as microcracks formed during the coating deposition process. 3) Once inside the ceramic layer, the molten CMAS propagates through microstructure defects, mainly wide-open cracks [32] . Figure 10 shows the beginning of penetration (i.e., stage 2).

Figure 11 shows the EDS image for sample 4-1250-5. The constituent elements of the YSZ (Zr, Y, and O) appear to be well distributed in the analyzed region. However, no relevant quantities of the participating CMAS elements (Ca, Mg, Al, and Si) were found, indicating the absence of silicate in the analyzed region. The exposure time influences the penetration depth of the CMAS into the ceramic layer, which may reach the bond coat or even the substrate [22] [31] [34] [35] . However, the literature indicates that 5 h at 1250˚C is sufficient for complete penetration of CMAS into the YSZ [21] . This also evidences that silicate formation did not occur or only a small amount was formed.

i.e., as mentioned earlier, some authors performed a heat treatment on the mixture of the CMAS constituents before depositing it onto TBC samples. This heat treatment requires a high temperature and exposure time to ensure silicate formation, such as 1400˚C for the time interval of 30 min [18] or 1 h [21] . Another possibility is to focus a higher magnification on the surface of the ceramic layer since CMAS accumulates on the surface of the top coat when there is no penetration [31] . An investigation in the pore and microcrack regions, which are pathways for CMAS [36] , could be evidence of the

beginning of penetration.

Figure 12 shows the EDS image that was focus on the surface layer of sample 4-1250-5. In the left upper region is possible to observe a great amount of Ca, Mg and Al (constituents of CMAS) indicating the presence of these elements, but these tree elements (together) are above on the surface of YSZ when compared to Zr. On the other side, the comparison between only Ca and Zr allow to observe that in some regions a superficial penetration has occurred how is showed in Figure 13. This penetration was very superficiality, similar to that which occurred in sample 3-1250-1 (Figure 10).