**Environmental Barrier Coatings Made by Di**ff**erent Thermal Spray Technologies**

**Robert Vaßen 1,\*, Emine Bakan 1, Caren Gatzen 1, Seongwong Kim 1,2, Daniel Emil Mack <sup>1</sup> and Olivier Guillon 1,3**


Received: 6 October 2019; Accepted: 18 November 2019; Published: 22 November 2019

**Abstract:** Environmental barrier coatings (EBCs) are essential to protect ceramic matrix composites against water vapor recession in typical gas turbine environments. Both oxide and non-oxide-based ceramic matrix composites (CMCs) need such coatings as they show only a limited stability. As the thermal expansion coefficients are quite different between the two CMCs, the suitable EBC materials for both applications are different. In the paper examples of EBCs for both types of CMCs are presented. In case of EBCs for oxide-based CMCs, the limited strength of the CMC leads to damage of the surface if standard grit-blasting techniques are used. Only in the case of oxide-based CMCs different processes as laser ablation have been used to optimize the surface topography. Another result for many EBCs for oxide-based CMC is the possibility to deposit them by standard atmospheric plasma spraying (APS) as crystalline coatings. Hence, in case of these coatings only the APS process will be described. For the EBCs for non-oxide CMCs the state-of-the-art materials are rare earth or yttrium silicates. Here the major challenge is to obtain dense and crystalline coatings. While for the Y2SiO5 a promising microstructure could be obtained by a heat-treatment of an APS coating, this was not the case for Yb2Si2O7. Here also other thermal spray processes as high velocity oxygen fuel (HVOF), suspension plasma spraying (SPS), and very low-pressure plasma spraying (VLPPS) are used and the results described mainly with respect to crystallinity and porosity.

**Keywords:** environmental barrier coatings; thermal spray methods; atmospheric plasma spraying; suspension plasma spraying; very low-pressure plasma spraying; high velocity oxygen fuel spraying

#### **1. Introduction**

Ceramic materials often show unique high temperature capability. However, monolithic ceramics suffer from an intrinsic low fracture toughness and, hence, for demanding high temperature applications as blades, vanes, shrouds or transition ducts of turbine engines reinforced ceramics have to be used [1]. Even sensitive parts can be made from such ceramic matrix composites (CMCs) and this has been demonstrated in the last years by different industries especially by GE [2]. For applications with higher mechanical loads, typically, non oxide CMCs as SiC/SiC are used due to their excellent high temperature strength and creep properties. However, also oxide/oxide composites which are mainly based on alumina and typically have reduced high temperature properties are of interest for specific applications e.g., combustion chamber applications [3]. In contrast to their metallic counterparts, CMCs do not have to be protected against oxidation as SiC forms a well-protecting silica scale and the oxide-based ceramics are anyhow stable at high temperatures in oxidizing atmosphere. However,

these oxides tend to be rather unstable in the gas turbine atmosphere, namely water vapour containing combustion gases having velocities above 100 m/s and temperatures above 1100 ◦C. In this environment volatile (oxy) hydroxides are formed which remove the scale and can lead to a massive damage of the hole component [4,5]:

$$\text{SiO}\_2 + 2\text{ H}\_2\text{O} \to \text{Si(OH)}\_4 \uparrow \text{ or } \text{Al}\_2\text{O}\_3 + 3\text{ H}\_2\text{O} \to 2\text{ Al(OH)}\_3\uparrow\tag{1}$$

This water vapor recession can be addressed by coatings using highly corrosion-resistant materials (so-called environmental barrier coatings (EBCs)). Many interesting materials have been tested in high velocity and high-temperature and water vapor environments. The recession rates and coefficients of thermal expansion of most of the EBC materials and CMCs of interest in this study are summarized in Figure 1. The recession rate and the thermal expansion coefficient (TEC) is plotted for different materials and the values for two CMC materials are given, an SiC/SiC and an alumina-based Ox/Ox composite. Additionally, Gd2Zr2O7 (CTE = 10.6 <sup>×</sup> 10−6·K<sup>−</sup>1) was considered and was shown to have the lowest recession rate among pyrochlore materials successfully used for thermal barrier coating (TBC) application [6,7]. Generally, suitable EBC materials should have low corrosion rates and TECs close to the used CMC.

**Figure 1.** Recession rate and thermal expansion coefficient for EBC materials relevant to this study (blue) and the SiC/SiC and Al2O3/Al2O3 composites (red) (Data from [8–10]).

Compared to thermal barrier coatings (TBCs) EBCs should have a low porosity level to avoid the access of water vapour to the CMC. This will lead to rather high Young's moduli *E*EBC in the EBCs close to the bulk values. A high Young's modulus is not favourable with respect to failure of the coatings as it increases the stored elastic energy; if this so-called energy release rate *G* exceeds a critical value *G*c, a crack can propagate and failure occurs. The energy release rate can be calculated under plane stress conditions for a homogeneous coating stress σ and an infinite crack as:

$$G = \frac{\sigma^2 h}{2E\_{\rm EBC}}\tag{2}$$

with *h* being the coating thickness. If the stress in the coating relaxes at high temperature, the stress at room temperature is for a thin coating determined by the mismatch of thermal expansion coefficients:

$$
\sigma = E\_{\rm EBC} \left( \alpha\_{\rm EBC} - \alpha\_{\rm CMC} \right) \,\Delta T \tag{3}
$$

As the square of TEC mismatch is determining the energy release rate in (2) a large mismatch should be avoided. Thickness and Young´s modulus have a linear influence on it.

The most often used technology to apply EBCs is thermal spray. Thermal spray is a versatile coating process in which the feedstock material is accelerated by a process gas and in most cases heated in powder form and then deposited on the substrate [11]. In thermal spray processes, the bonding between coating and substrate is mainly due to mechanical interlocking at asperities of the substrate surface. A sufficient bonding is typically achieved by a grit-blasting process prior to the deposition. However, it was found that for low strength materials as oxide-based composites this treatment can lead to significant degradation of the material. Hence, new methods as laser ablation to structure the surface will be discussed in the present paper.

In the thermal spray process, the molten particles will impinge on the substrate, the deformed droplets ("splats") will cool down quickly and generate large tensile stresses which typically introduce cracks [12]. These cracks form an open network which allow the diffusion of gas species through it and hence, such coatings will delay the water vapor corrosion but can hardly avoid it. In order to form gas-tight coatings rather hot spraying conditions with high substrate temperatures (above 500◦C) and complete melting of the feedstock are beneficial. Such conditions can lead to a re-melting of already deposited splats and by this improve the bonding in the coating. Such approaches have been used to produce gas-tight electrolytes with a thickness below 50 μm from yttria stabilized zirconia (YSZ) for solid oxide fuel cells [13]. It was also demonstrated that high particle velocities at impact (>300 m/s) are favourable. Similar process conditions are also used to deposit segmented TBCs. In this process very dense coatings are produced by APS leading also to high tensile stresses in the coatings [14]. If a certain thickness (typically about 100 μm) is reached, the energy release rate exceeds the critical one and segmentation cracks are formed. This is of course not wanted for EBC applications. The segmentation cracks are driven by the cooling of the splats and hence by the absolute TEC of the ceramic coating material and not by the mismatch to the substrate. As a result EBCs for SiC/SiC should be less prone to segmentation cracks then those for Ox/Ox CMCs assuming the same other materials properties (Young´s modulus, toughness).

In addition to the difficulty to produce very dense coatings another issue is often important for EBCs. The materials used tend to form an amorphous phase when cooled down quickly from the molten state to the substrate temperature. This is detrimental, as the coatings will typically crystallize during operation, which is accompanied by a volume reduction and pore/crack formation in the coating. So, a highly crystalline coating in the as-sprayed condition is desirable. Previous results indicate that high deposition temperatures are favourable as then the time at temperature for crystallization is sufficient. It was proposed to deposit coatings in a furnace [15], however, we try in our approaches to avoid it as it introduces additional process steps and hence costs.

For the deposition of the EBCs we have investigated four different thermal spray processes (see Figure 2). Basic requirements for the processes are the capability to reach high feedstock temperatures and velocities.

First of all, atmospheric plasma spraying (APS) will be considered as a kind of standard baseline process. Here a fast variant (high velocity APS, HV-APS) will be described. The APS process shows a high temperature of the process gases (>10000 K at the nozzle exit) due to the efficient heating by electric arcs. Hence, a good and fast heat-transfer to the particles is possible, which allows complete particle melting, however can also lead to loss of those constituents in the powder with high vapor pressure. Typical velocities are in the range of 200–300 m/s, however specific conditions as small nozzle diameter and larger amounts of process gas allow higher velocities (up to 500 m/s, HV-APS)

While in the APS process particle in the range of several 10 micrometers (see splat in Figure 2) are used, the suspension plasma spraying allows much finer droplets and splats due to the use of suspensions. Here the droplet size is determined by a complex atomization process. The droplets typically in the size range of micrometers can easily reach the temperature and velocities of the process gas, hence both are typically higher than in APS. If the right suspension properties are chosen this can lead to improved densities, however also some more significant loss of volatile species can be expected due to the higher surface area of the droplets.

The plasma guns can also be operated in inert gas environment. For that a vacuum chamber is evacuated and then refilled by inert gas to typical pressures of about 50 mbar. The process is often. used to deposit metals to avoid in-flight oxidation. Due to the expansion of the plume into a chamber with reduced pressure, the gas velocities are higher than in APS. In our used equipment, powerful pumping units allow to maintain even lower chamber pressures in the mbar range during operation (Very low-pressure plasma spraying-VLPPS). This leads to even higher gas and particle velocities (see the large expanding plume in Figure 2) with corresponding high coating densities. In addition, the used powerful gun and the reduced cooling by convection allows homogenous high substrate temperatures which support dense and crystalline coating formation. This could be demonstrated for the deposition of dense ceramic membranes [16].

The last process which will be considered is a combustion process, the high velocity oxygen fuel (HVOF) spraying. The maximum gas temperature is limited to temperatures slightly above 3000 ◦C, so the heating of the feedstock is limited. On the other hand, the use of a de Laval nozzle allows extremely high, supersonic gas velocities (see shock diamonds in Figure 2) and corresponding high particle velocities. With such high velocities dense coating might be possible even without complete melting of the feedstock. Such a particle with non-molten core is visible in Figure 2

An additional more detailed description of some of the results can be found in our previous papers [17–20].

**Figure 2.** Characteristics of the 4 investigated thermal spray processes (Axial III used for SPS, TriplexPro210 for APS, O3CP for VLPPS, Dj2700 for HVOF) showing plume temperatures and particle velocities, photos of the plumes plus conditions used (process gases, current in case of plasma processes, and stand-off distances), and photos of typical splats.

#### **2. Materials and Methods**

#### *2.1. Feedstock Materials, Suspensions and Substrates*

Different feedstock powders have been used for the investigations, Table 1 gives an overview of the used materials.


**Table 1.** Description of the used feedstock materials.

For SPS purposes, YDS powder was milled in ethanol with the addition of polyethylenimine (PEI) and zirconia milling balls (*d* = 3 mm). The mixture was milled on roller cylinder (60 min−1, 48 h) in order to produce a homogeneously-dispersed suspension (30 wt % in solid content). After milling, the suspension was diluted with ethanol to 10 wt % of solid loading and the particle size distribution was *d*10 = 1.3 μm, *d*50 = 3.1 μm, and *d*90 = 5.0 μm.

As substrates different materials have been taken as carbon steel, graphite, monolithic SiC and partly CMCs as 2d C/SiC (Schunk CF 226 P 75), SiC/SiCN (provided by DLR Stuttgart [21]) and an alumina based CMC FW12 (Walter E. C. Pritzkow Spezialkeramik, Filderstadt-Sielmingen, [22]). In the case of YS, a bond-coated IN738 was used as substrate for preliminary tests.

#### *2.2. Thermal Spray Facilities*

At IEK-1 different thermal spray systems for the deposition of ceramic coatings with high particle velocities are available. The "work horse" is a MultiCoat system (Oerlikon Metco, Wohlen, Switzerland) which was operated in this study for APS with the three-cathode TriplexPro 210 and for HVOF with the Diamond Jet 2700 spray torches both mounted on a six-axis robot (IRB 2400, ABB, Zurich, Switzerland). For the suspension plasma spraying a different spray booth equipped with the Axial III plasma torch with three separate cathode-anode pairs (Mettech Northwestern Corp., North Vancouver, BC, Canada) which allowed the axial injection of the liquid feedstock was used. As feeding system a home-made device was employed.

Vacuum plasma spraying was performed in a MultiCoat platform (Oerlikon Metco, Wohlen, Switzerland) with a 6 m<sup>3</sup> tank volume. Here the powerful O3CP torch (Oerlikon Metco, Wohlen, Switzerland) was chosen as plasma torch. The deposition temperatures were monitored by infrared pyrometers in each processing method.

An overview of the employed spray parameters is found in Table 2. The parameter used are based on long-year experience in the thermal spray of ceramics. Some further optimization trials have been made especially for EBC materials for oxide-based CMCs but they are not shown here. As shown in the results, the coating microstructure appears rather dense and often crystalline, no efforts to further optimize the microstructure were made. In the case of YDS the APS coating properties were not satisfying, here also other techniques have been investigated.

#### *2.3. Surface Treatment*

All samples were ultrasonically cleaned before spraying. To increase the adhesion properties, some of the Al2O3/Al2O3 CMC samples were pre-treated mechanically or by laser ablation. The mechanical pre-treatment was carried out with 80-grit sandpaper or by grit blasting with F 36 powder.


**Table 2.** Overview of the used thermal spray conditions (YS = Y2SiO5, YBDS = Yb2Si2O7)

A pulsed laser Trumpf TruMark 5020 with a wavelength of 1062 nm (Nd:YAG), 50 ns pulse duration and a maximum peak power of 15 kW was used for surface structuring (Trumpf GmbH + Co. KG, Ditzingen, Germany). The used laser parameters are given in Table 3. For a more detailed description of the ablation process the reader is referred to Gatzen et al. [23]



#### *2.4. Characterization*

After spraying the samples were sectioned, polished, and examined with a scanning electron microscope (Carl Zeiss NTS GmbH, Oberkochen, Germany) combined with an energy-dispersive

X-ray INCAEnergy355 spectrometer (EDS, Oxford Instruments Ltd., Abingdon, Oxfordshire, UK). Alternatively, SEM images were taken with a Hitachi TM3000 (Hitachi, Krefeld, Germany). Acquired SEM images were employed to assess the porosity in the coatings by means of image analysis using an image thresholding procedure with the analysis pro software (Olympus Soft Imaging Solutions GmbH, Germany). The analysis was performed on 10 SEM micrographs (2000× magnification) per sample, each with a resolution of 1280 × 1100 pixels and covering a horizontal field width of 126 μm. Crack density values of samples manufactured by suspension plasma spraying were calculated from the number of vertical cracks by using 15 SEM images (with ×300 magnification and width of 600 μm) as well.

X-ray diffraction analysis was performed with a D4 Endeavor with Cu-Kα radiation (λ = 1.54187 Å) & TOPAS software V4.2., Bruker AXS, Germany. In addition to the Rietveld method, also amorphous content and quantitative phase analysis (QPA) was used to evaluate crystalline phases and amorphous contents in the coatings (see [17] for details). High-temperature XRD (HT-XRD) was performed at the PANalytical Empyrean diffractometer in Bragg-Brentano geometry using a Cu Kα anode and an environmental heating chamber HTK1200N (Anton Paar GmbH, Ostfildern, Germany) between room temperature and 1400 ◦C.

The surface profiles were measured by white light interferometry (cyberTechnologies, CT350T, Eching, Germany).

The EBC systems on oxide/oxide composites were also thermally cycled at 1200◦C in a furnace. The heating and cooling rates were +/− 10 K/min, immediately after reaching room temperature the next cycle was started. 4 of these cycles have been made for each coating system. Thermal cycling is seen as a tool to characterize the bonding in the case of the EBC for oxide-based CMCs and hence the efficiency of the surface treatment.

#### **3. Results and Discussion**

#### *3.1. APS Coatings on Ox*/*Ox CMCs*

Coatings of Y2O3, Gd2Zr2O7, Y3Al5O12 and Yb2Si2O7 were applied on the untreated and cleaned substrates. XRD measurements of as-sprayed samples were carried out. The measurements are plotted in Figure 3, signals that cannot be attributed to the desired phase were marked with an asterisk (\*). The XRD measurements of Y2O3 and Gd2Zr2O7 show the diffraction patterns of the corresponding phases, no signals from secondary phases were observed. The coatings appear to have a high crystallinity. In contrast to this, the XRD measurements of Y3Al5O12 and Yb2Si2O7 coatings show humps, which indicate the presence of large amounts of amorphous phases. Due to the low crystallinity not all phases could be identified. The XRD measurements of the Y3Al5O12 coating shows the diffraction pattern of Y3Al5O12 and YAlO3. The formation of the alumina depleted phase YAlO3 is attributed to evaporation of alumina during the spraying process, since the vapor pressure of Al2O3 is higher than that of Y2O3 [24]. The XRD measurement of the Yb2Si2O7 coating shows no reflections of the desired phase. Besides the large amorphous hump, the diffraction pattern of Yb2O3 was observed. The high amorphous content of the Y3Al5O12 and Yb2Si2O7 coatings might cause crystallization stress in the coating during thermal treatment.

**Figure 3.** XRD measurements of Y2O3, Gd2Zr2O7, Y3Al5O12 and Yb2Si2O7 coatings in the as-sprayed state, peaks of secondary phases are marked with \*.

The as-sprayed samples were subjected to furnace thermal cycling, the used temperature program consists of four 20 h cycles at 1200 ◦C. SEM images of cross sections of samples in the as-sprayed state and after thermal cycling are presented in Figure 4. Additionally, higher magnification images of the interface region of the coatings are presented in Figure 5.

**Figure 4.** SEM-images of Y2O3, Gd2Zr2O7, Y3Al5O12 and Yb2Si2O7 APS coatings on Ox/Ox CMCs before (left) and after thermal cycling (right).

**Figure 5.** SEM images of the coating substrate interface of Y2O3, Gd2Zr2O7, Y3Al5O12 and Yb2Si2O7 coatings on Ox/Ox CMCs after thermal treatment.

The obtained Y2O3 coatings were dense and fully crystalline. The adhesion of the Y2O3 samples seems to be quite good, after heat treatment no signs of cracks or delamination are observed. The good adhesion can be attributed small TEC mismatch and phase stability, resulting in low driving force for delamination. Another reason for the good adhesion is the formation of yttrium-aluminates at the coating substrate interface leading to chemical bonds between coating and substrate [25]. In Figure 5 the formation of several yttrium-aluminates known from the phase diagram [26] can be observed.

The SEM images of the Gd2Zr2O7 coatings show dense and homogenous coatings in the as-sprayed state. Figure 5 shows the coating substrate interface in detail. After thermal cycling no sign of a reaction between Gd2Zr2O7 and Al2O3 was observed. This is in contrast to Lakiza et al. [27] and Leckie et al. [28] The phase diagram of Gd2Zr2O7, presented by Lakiza et al. [27], suggests formation of gadolinium-aluminates such as GdAlO3 and Gd4Al2O9. The reaction between Gd2Zr2O7 and Al2O3 to GdAlO3 was observed in a study by Leckie et al. [28]. The absence of this reaction in this study can be explained by bad wetting which causes only small contact areas to the substrate, as a consequence the reaction is inhibited. However, after heat treatment the coating was delaminated at the interface (marked by arrows in Figure 4). This can be explained by the TEC mismatch (ΔCTE <sup>=</sup> <sup>3</sup> <sup>×</sup> <sup>10</sup>−<sup>6</sup> <sup>K</sup>−<sup>1</sup> , see Equations (2) and (3)) and bad wetting of the relative smooth substrate (*R*a = 2.6 μm).

The SEM images of the Y3Al5O12 coating confirm the presence of secondary phases within the coating. Furthermore, large round pores can be observed in the Y3Al5O12 coatings. The high porosity is not favourable for an EBC, as pores and cracks increase the permeability for water vapor. Despite the low TEC mismatch (ΔCTE = 3 <sup>×</sup> 10−<sup>6</sup> K−1), the Y3Al5O12 coatings tend to fail during thermal cycling. This can be explained by the high amorphous content of the APS-Y3Al5O12 coatings. Thermal treatment leads to crystallization and phase transformation of the coating; this causes stress within the material leading to the formation of segmentation and delamination cracks. The formation of large pores in Y3Al5O12 coatings was also reported by Weyant et al. [29], a dependence between crystallization, substrate temperature and porosity was assumed. According to the phase diagram [30] of Y2O3 and Al2O3, a reaction between coating and substrate is not expected. The SEM image of the Y3Al5O12 - Al2O3 interface (Figure 5) shows that unlike with Y2O3, no reaction occurs between substrate and coating.

The cross-sections show that the disilicate coatings have a lot of pores and consist of several phases, which is, again, not favourable for an EBC. Furthermore, crystallization problems occurred, as described for the Y3Al5O12 coatings. Due to the high TEC mismatch (ΔCTE = 5.2 <sup>×</sup> 10−<sup>6</sup> K−1, see Equations (2) and (3)) and the stresses that arise during crystallization, these coatings were delaminated during thermal treatment. According to the phase diagram of Yb2O3-SiO2-Al2O3 [31], the formation of mullite (Al6Si2O13) and ytterbium-aluminates (Yb4Al2O9, Yb3Al5O12) during thermal cycling is possible. However, no reaction between coating and substrate was observed (see Figure 5). The absence of a reaction layer may be attributed to the loose contact between coating and substrate as well as the high TEC mismatch and the resulting premature coating delamination.

Due to the high coating crystallinity, phase purity and coating density, Gd2Zr2O7 and Y2O3 were chosen for further experiments. Although Gd2Zr2O7 has a higher CTE mismatch, it is a desirable top coat due to its excellent CMAS stability. In order to increase the coating adhesion different methods of surface preparation were carried out before spraying.

First, samples of the Al2O3/Al2O3 CMC were mechanically treated to increase the surface roughness. Some samples of the Al2O3/Al2O3 CMC were ground, others were grit-blasted. Both methods were able to increase the surface roughness from 2.6 μm to 5.1 μm and 15 μm, respectively. The samples were coated and subsequently furnace cycled for 4 × 20 h at 1200 ◦C. SEM images of cross-sections of the as-sprayed Y2O3-coatings as well as furnace cycled samples are shown in Figure 6. The results show that the mechanical pre-treatment damaged the ceramic substrate to such an extent that even Y2O3 coatings failed. Unlike in the study of Gérendas et al. [32], mechanical pre-treatment was not able to achieve sufficient coating adhesion on this CMC.

**Figure 6.** SEM images of APS-Y2O3 coatings on Ox/Ox CMCs with and without mechanical pre-treatment before and after thermal cycling.

Laser ablation was used for surface structuring without damaging of the substrates. The positive effect of laser surface structuring of Al2O3/Al2O3 CMCs on coating adhesion was recently demonstrated by Gatzen et al. [23]. Two different structures were chosen for this study: honeycomb and cauliflower structure. Both structures are illustrated in Figure 7.

**Figure 7.** Surface profiles, white light topography and SEM images of ox/ox CMC before (top) and after structuring by laser ablation (middle, bottom).

The untreated substrate has a roughness of about 2.6 μm and a homogenous smooth surface. Few cracks and pores at the surface offer possibilities for clamping. The cauliflower like surface structure comprises an irregular and inhomogeneous surface of re-solidified alumina. An average roughness of 10.8 μm was achieved with this pattern. The honeycomb structure shows well- defined holes close to each other. A roughness of about 5.7 μm was measured for this pattern. Both surface structures offer more possibilities for clamping, therefore, an increase in lifetime is expected especially for Gd2Zr2O7 coatings.

After structuring the samples were coated with Y2O3 and Gd2Zr2O7. SEM images of the coated and thermally cycled samples are shown in Figure 8. Both coatings could infiltrate the voids in between the substrate structure. The surface structures, especially the cauliflower structure, allow interlocking of the coating. After thermal cycling no cracks or delamination occurred in the Y2O3 coating-substrate-system. This shows that in contrast to grit-blasting and grinding, laser structuring of the CMC does not weaken the ceramic matrix. Furthermore, after thermal cycling of the laser-structured Gd2Zr2O7 coated samples, the coating-substrate interface shows no delamination cracks. This is a significant improve, compared to the coating on untreated substrates. Although the energy release rate is high in these coatings with high mismatch (see Equations (2) and (3)), the laser structuring of the samples proofed to be beneficial for the coating adhesion (e.g., the critical energy release).

**Figure 8.** SEM images of APS Y2O3 and Gd2Zr2O7 coatings on laser structured Ox/Ox CMC substrates after thermal cycling (4 × 20 h at 1200 ◦C).

In addition, a third method of adhesion improvement was tested for Gd2Zr2O7: the usage of an Y2O3-bondcoat. For this, the untreated substrate was first coated with Y2O3 and then coated with Gd2Zr2O7. This double layer coating system is referring to the double layered TBCs as published by Vaßen et al. [33]. The Y2O3-bondcoat helps to buffer to some extent especially at edges the CTE mismatch between the CMC and Gd2Zr2O7. Furthermore, the Y2O3 coating is known to be good adherent on Alumina CMCs [25] and offer a rough surface for the Gd2Zr2O7 coating. Using Y2O3 as an interlayer between substrate and Gd2Zr2O7 also proofed to increase the coating adhesion as well.

#### *3.2. Development of APS Y2SiO5 coatings*

According to Figure 1 Y2SiO5 has an excellent water vapor resistance and is therefore considered as appropriate EBC material [34,35]. In this investigation only rather hot APS spraying conditions are investigated for this material. The density of all the coatings manufactured by the investigated conditions (see Table 2) has been rather high, more interesting is the phase content which shows remarkable differences (see Figure 9). Compared to the He as secondary gas, hydrogen increased the plasma enthalpy which leads to higher substrate temperatures (750◦C) and increases the amount of crystallization. The comparable amount of SiO2 in all coatings might indicate no significant differences in the silica loss during spraying (although the amorphous content is not considered).

Due to the highest crystallinity, the condition D with hydrogen and long stand-off distance (condition D) was used for further studies and coatings on C/SiC substrates were prepared. In Figure 10 SEM micrographs of the coatings in the as-sprayed condition and after annealing for 12 h at 1350 ◦C. First of all, the coating is also after the heat-treatment well-adherent. This indicates that the TEC mismatch (see Equations (2) and (3)) between coating and substrate seems to be not extremely detrimental (although the material would fit even better to Ox/Ox CMC substrates). In addition, also the phase transformation at 850 ◦C and the strongly anisotropic expansion of Y2SiO5 [36] obviously does not significantly affect coating integrity.

**Figure 9.** Phase evaluation of 4 Y2SiO5 coatings (caption indicate power, stand-off distance and secondary gas).

**Figure 10.** SEM micrographs of plasma-sprayed Y2SiO5 coatings (condition D) in the as-sprayed (**a**,**b**) and annealed (**c**,**d**, 1350◦C, 12h) condition.

In addition, the size of the segmentation and micro cracks is largely reduced due to sintering, some segmentation cracks are even no longer penetrating through the whole coating. Also the phase

distribution changed significantly. While in the as-sprayed condition rather large regions with low and high silicon content appear (stemming from the impinging particles which have lost silicon during spraying mainly at their surface), the annealed coating shows much finer phases consisting of cubic Y2O3 (bright) and monoclinic Y2SiO5 (darker). The evolution of phases is shown in Figure 11 measured by XRD and using Rietveld refinement. Obviously, a short-term annealing at 1000 ◦C is not sufficient to form the equilibrium phases. As the coating is rather dense, the Y2O3 as a second phase has a reasonable corrosion resistance (Figure 1) and the microstructure probably will show some particulate toughening effects, the found coating is expected to perform good as EBC.

**Figure 11.** Phase evolution in APS Y2SiO5 coatings (condition D) after heat-treatment.

#### *3.3. Yb2Si2O7 Coatings Manufactured by Di*ff*erent Thermal Spray Techniques for SiC*/*SiC Substrates*

Figure 12 shows the phase composition of APS, HVOF and VLPPS coatings deposited from different particle size fractions of the same Yb2Si2O7 feedstock. The main differences between the different methods are melting degree of the sprayed particles and deposition temperatures which define the crystallinity of the coatings as well as the degree of Si evaporation during spraying.

In APS and VLPPS, as a result of the high heat transfer from plasma to the particles, particles melt, impact on the substrate in the molten state and rapidly solidify on the substrate. High plasma powers were selected in the deposition of EBCs with APS (57 kW) and VLPPS (90 kW) as a dense microstructure consists of well-flattened particles with good interfacial contact is aimed. While high plasma powers ensure melting of particles in the plume, it also leads to a significant amount of Si-evaporation from Yb2Si2O7 during spraying. As a result of that, Si-depleted secondary phases such as Yb2SiO5 and Yb2O3 are found in the as-sprayed coatings as shown in the Figure 12a. Furthermore, due to quenching of the molten particles on the substrate, that is at nearly room temperature if not pre-heated by plasma plume, crystallization of the glass forming silicates is suppressed and the amorphous phase is formed. The APS experiments carried out in order to increase the SiC substrate temperature by heating with plasma prior to deposition revealed that for a substrate size of 50 <sup>×</sup> 50mm2, it is possible to increase the substrate temperatures up to about 800–900 ◦C however the temperature rapidly goes down to 500–550 ◦C till the deposition starts. As shown in Figure 12a, this deposition temperature range was found to be not sufficient to activate and complete the crystallization of the coating. According to HT-XRD analysis of amorphous plasma-sprayed Yb2Si2O7 particles, which were collected in water and dried subsequently at 70 ◦C, the crystallization temperature of the material was found to be above 1000 ◦C, which explains the amorphous structure of the coating deposited at about 550 ◦C. Aiming at a higher substrate temperature as well as slower cooling rates in order to provide higher energy and longer time for the atoms to rearrange into the crystalline state, VLPPS experiments were conducted in the controlled atmosphere chamber (2 mbar). It was found that it is possible to reach substrate

temperatures higher than 1000 ◦C (Figure 13) owing to higher plasma power of the O3CP torch as well as to retain it till the deposition starts due to reduced heat transfer under vacuum. To prolong the high-temperature phase after deposition for some minutes, the coating was kept to be heated using the plasma plume and by this procedure, highly crystalline coatings could be produced as shown in Figure 12a. Nevertheless, it should be mentioned that Si-evaporation remains a problem as more than 10 wt.% Yb2SiO5 was detected in the VLPPS coatings.

**Figure 12.** (**a**) Quantitative phase composition of the as-deposited coatings from Yb2Si2O7 feedstock using different thermal spray methods, (**b**) HT-XRD diffractograms of atmospheric plasma sprayed Yb2Si2O7 feedstock.

The HVOF process yields lower flame temperatures in comparison with the plasma spray processes as the flame is generated by combustion. As a result of that, particle temperatures as well as the deposition temperature are lower. Nonetheless, using the HVOF process, Yb2Si2O7 coatings with higher crystallinity in comparison with the APS could be manufactured because of the deposition of un-melted or partially molten particles. This was avoided in the plasma spray processes as un-melted particles increase the porosity levels in the coatings due to imperfect contact regions as well as the porous morphology of the particle itself. In the HVOF process, however, not the porosity stemming from the particle morphology, but the porosity caused by the bad contact can be minimized thanks to very high flame velocity and, thus, the high momentum transfer to the particles in the flame. If the brittle oxide particles are completely un-melted, they break upon high-velocity impact on the

substrate, however, if only their core is un-melted and the outer surface is molten, they can adhere on the substrate or on the previously deposited layers. To reach such particle conditions, sensitive process optimizations are required in terms of fuel–oxygen stoichiometry, total gas flow, powder feed rate, stand-off distance, and particle size distribution. Details of these investigations can be found elsewhere [18] and HVOF microstructure will be further discussed in the following. Consequently, these coatings own higher crystallinity regardless of the lower deposition temperature in the process as un-melted part of the particles remain crystalline. An added benefit of low flame temperature is diminished Si-evaporation from the particles during spraying. It can be seen from Figure 12a that the detected Yb2SiO5 content from about 50 wt.% crystalline HVOF coating is about 5 wt.% which fits well to the feedstock composition although of course the large amorphous content has to be considered.

**Figure 13.** Recorded temperature measurements using a thermocouple during VLPPS deposition of Yb2Si2O7 on monolithic SiC substrate.

The microstructure of the coatings sprayed from Yb2Si2O7 feedstock with APS, HVOF and VLPPS processes can be seen in Figure 14. The two APS coatings (a,b) that were deposited at different temperatures (Figure 14a is the "standard" APS sample and b is deposited at a higher temperature, further details are discussed below) have vertical cracks running through the thickness of the coatings which is associated with the higher thermal stresses in these coatings due to the significant amount of Yb2SiO5 content (about 30 wt.% in these coatings determined after crystallization heat treatment). Because thermal expansion coefficient of Yb2SiO5 (7.5 <sup>×</sup> <sup>10</sup>−<sup>6</sup> ◦C<sup>−</sup>1) is larger than that of the Yb2Si2O7 (4.7 <sup>×</sup> 10−<sup>6</sup> ◦C−1) as well as the SiC/SiC substrate (5.1 <sup>×</sup> 10−<sup>6</sup> ◦C−1) [37]. This leads to greater tensile thermal stresses in the oxide coating after cooling, along with the tensile quenching stresses, which induce the vertical cracking. The high amorphous content in the APS layer would even further increase the CTE mismatch, however simultaneously due to a reduced Young´s modulus of the amorphous state, the effect on the stress level might be limited. Therefore, as cracks are not desired in the EBC microstructure, it is of crucial importance to minimize Si- evaporation during spraying.

**Figure 14.** Microstructure of the coatings sprayed from Yb2Si2O7 feedstock using (**a**,**b**) APS, (**c**) HVOF and (**d**) VLPPS process. Note that both APS coatings (**a**,**b**) were sprayed with 520 A current and 90 mm spray distance but deposition temperatures were about 550 ◦C and 900 ◦C at a and b, respectively (see text).

Both APS coatings (Figure 14a,b) were sprayed at similar conditions but sample size in (b) was very small (3 <sup>×</sup> 3.8 <sup>×</sup> 36 mm3) and as a result of that a maximum deposition temperature of about 900 ◦<sup>C</sup> could be reached for this particular sample. Therefore, in contrast to standard size APS sample that was analysed in the preceding section and shown in Figure 14a, the APS coating shown in Figure 14b is highly crystalline (76 wt.%). The porosity content of the two APS coatings are also dissimilar, i.e., the coating (a) is significantly denser than the coating (b). This difference between the microstructures was associated with the higher crystallinity in the coating (b). Seemingly, heterogeneous nucleation takes place at the splat boundaries during cooling while the centre of the splats remains amorphous. As density increases in the crystallizing zone, the volume is reduced which induces elastic tensile stresses within the amorphous region. Pores within the splat, therefore, nucleate when these stresses are large enough. Other relevant theories to the formation of pores are discussed elsewhere [20].

No vertical cracking was observed in the fairly dense HVOF layer (Figure 14c), presumably due to lower Yb2SiO5 content and hence reduced thermal stresses in the process as a result of the lower particle and deposition temperatures, respectively. However, isothermal thermal cycling experiments revealed that the adhesion of the HVOF oxide layer on the Si bond coat is poor because the HVOF layer partially delaminated only after few cycles [18]. The short lifetime of the layer was attributed to the presence of un-melted particles at the interface wherein the cracks can easily propagate. The un-melted particles can be avoided for instance by decreasing the particle size of the feedstock but that means higher amorphous content and Yb2SiO5 phase in the layer at the same time. Figure 14d shows the microstructure of a VLPPS layer that is free of vertical cracks, dense and also crystalline in the as-sprayed state, which makes it a very promising method developed for EBC manufacturing in Jülich. High deposition temperatures (min. 1000 ◦C) and moderate cooling rates (approx. 55 K/min) in the process as shown in Figure 13 evidently helps for crystallization as well as for stress relaxation in the layer. Further investigations are ongoing in this direction to understand the effect of deposition temperature and cooling rates on crystallization kinetics and related mechanical properties in the layer, as well as the stress state in the EBC system.

While a variety of coating morphologies can be obtained with thermal spray processes, characteristics of the coatings, typically the size of microstructural features, is controlled by the used feedstock [38]. For example, the size of intersplat pores is directly depending on the size and shape of powdery feedstocks. Since a well flowable powder in the size range of about 10–100 μm is usually applied for conventional plasma spraying, there is a minimum size for application of powdery feedstock. Suspension plasma spray would be one of alternative processes for conventional ones in this context. Suspension plasma spraying is one of the rather new thermal spray techniques, which has a liquid feed stock and relatively higher plasma power compared with conventional ones. Especially, SPS could supply a great diversity in microstructural features from columnar and porous to bulk-like dense ones [39,40]. Main processing parameters, such as gun power, spraying distance, roughness, and particle size distribution in suspension, would play an important role in controlling the microstructures fabricated by SPS.

Figure 15 shows the cross-sectional microstructures of Yb2Si2O7 coatings fabricated by using the SPS technique with different spraying distances of 50 and 70 mm. As seen especially in the higher magnification images, bulk-like dense microstructure can be obtained within this work regarding gun power, spraying distance and particle size distribution in suspension feedstock. Vertical cracks in the coating microstructure were inevitable because of tensile stress from rapid cooling after spraying as can be seen in Figure 15c.

**Figure 15.** Microstructure of the coatings sprayed from Yb2Si2O7 suspension by using SPS with a gun distance of (**a**,**b**) 50 mm and (**c**,**d**) 70 mm.

Figure 16 reveals how the stand-off distance influences the substrate temperature and by that e.g., the degree of crystallization. Shorter stand-off distances give higher temperatures and by that also better splat bonding and reduced relaxation, which give a higher segmentation crack density. Figure 17 shows the X-ray diffraction patterns of Yb2Si2O7 coatings fabricated by using the SPS technique with different spraying distance. Considerable amounts of the amorphous phase were observed from EBC with long spraying distance (90 mm). In the case of short spraying distance (50 mm), the degree of

crystallinity was around 80%, which was calculated from the areal intensity ratio between crystalline peaks and amorphous humps of XRD.

**Figure 16.** Substrate temperature and crack density as function of the spraying distance.

**Figure 17.** X-ray diffraction patterns from YDS EBCs fabricated by using suspension plasma spraying with spraying distance of 50, 70, or 90 mm. Arrows indicate amorphous baselines in XRD.

#### **4. Conclusions**

This study summarizes insights into the development of EBCs for both oxide- and non-oxide-based CMCs. Results of coating experiments with different ceramic powders on an Al2O3/Al2O3 CMC were presented; the samples were furnace cycled to test the high temperature behaviour. Coatings of Y2O3 and Gd2Zr2O7 showed promising results, as they were crystalline and dense in the as-sprayed state. Y2O3 coatings showed excellent adhesion due to the formation of chemical bonds between coating and substrate. In contrast to this, Gd2Zr2O7 coatings tend to fail during cycling because of bad contact between coating and substrate. The coating adhesion could be significantly increased by laser structuring of the CMC before coating.

Y2SiO5 coatings could be prepared with rather high crystallinity (nearly 70%). Heat treatment led to the formation of a fine-grained microstructure with the major phase Y2SiO5 and a considerable amount of Y2O3 without delamination after cooling.

Yb2Si2O7 coatings have been produced by different thermal spray process, namely, APS, HVOF, VLPPS, and SPS. APS can deliver dense coatings with a high degree of amorphous phase content. HVOF gives rather crystalline coatings with acceptable porosity levels, VLPPS gives even better coatings. Furthermore, SPS can give a high degree of crystallinity, however segmentation cracks are difficult to avoid.

A summary of the outcome reflecting the actual situation as seen in our institute is given in Table 4. So, our message is that also other thermal spray methods than APS can be used to obtain promising EBCs. Certainly, further improvements of processes can change the ranking in this evaluation.

**Table 4.** Comparison of the used thermal spray processes for the deposition of EBCs (o reflects average, - (– very) bad, + (++ very) good results with respect to the criteria).


**Author Contributions:** R.V.: writing—original draft preparation, supervision, review and editing; E.B.: investigation, writing—original draft preparation; C.G.: investigation, writing—original draft preparation; S.K.: investigation, writing—original draft preparation; D.E.M.: supervision, review and editing; O.G.: supervision, review and editing.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors acknowledge the contributions of the following colleagues in our institute: Ralf Laufs, Frank Kurze and Mr. Karl-Heinz Rauwald for the invaluable assistance during plasma spraying and Georg Mauer for valuable discussions of thermal spray results. We also would like to thank Doris Sebold for SEM analysis and Yoo Jung Sohn for the extended XRD analysis. We also appreciate the support of Sigrid Schwartz-Lückge and Mark Kappertz in sample preparation and characterization.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **YAlO3—A Novel Environmental Barrier Coating for Al2O3**/**Al2O3–Ceramic Matrix Composites**

#### **Caren Gatzen \*, Daniel Emil Mack, Olivier Guillon and Robert Vaßen**

Forschungszentrum Jülich GmbH, Institute of Energy and Climate Research, Materials Synthesis and Processing (IEK-1), 52425 Jülich, Germany

**\*** Correspondence: c.gatzen@fz-juelich.de

Received: 30 August 2019; Accepted: 23 September 2019; Published: 25 September 2019

**Abstract:** Ceramic matrix composites (CMCs) are promising materials for high-temperature applications. Environmental barrier coatings (EBCs) are needed to protect the components against water vapor attack. A new potential EBC material, YAlO3, was studied in this paper. Different plasma-spraying techniques were used for the production of coatings on an alumina-based CMC, such as atmospheric plasma spraying (APS) and very low pressure plasma spraying (VLPPS). No bond coats or surface treatments were applied. The performance was tested by pull–adhesion tests, burner rig tests, and calcium-magnesium-aluminum-silicate (CMAS) corrosion tests. The samples were subsequently analyzed by means of X-ray diffraction, scanning electron microscopy, and energy-dispersive X-ray spectroscopy. Special attention was paid to the interaction at the interface between coating and substrate. The results show that fully crystalline and good adherent YAlO3 coatings can be produced without further substrate preparation such as surface pretreatment or bond coat application. The formation of a thin reaction layer between coating and substrate seems to promote adhesion.

**Keywords:** atmospheric plasma spraying (APS); very low pressure plasma spraying (VLPPS); environmental barrier coating (EBC); YAlO3; yttrium aluminum perovskite (YAP); ceramic matrix composite (CMC); Al2O3

#### **1. Introduction**

The efficiency of a gas turbine is determined by its pressure ratio and its maximum service temperature. A way to increase the efficiency is to increase the maximum temperature. However, the maximum service temperature is limited by the physical and chemical stability of the components in the high-temperature section. A significant temperature increase was achieved by the use of single-crystal super alloys, thermal barrier coatings, and complex cooling systems. Unfortunately, the complex cooling of the components leads to high efficiency losses. Therefore, there is a need for new high-temperature materials with higher temperature capabilities, so that the cooling of the components can be reduced or even omitted [1,2].

Ceramics, such as SiC and Al2O3, offer high temperature and chemical stability and therefore are promising materials for high-temperature applications. Reinforcing of the ceramic matrix with ceramic fibers avoids catastrophic failure due to crack deflection and bridging mechanisms and leads to a pseudo-plastic failure behavior of the ceramic material [3]. These so-called ceramic matrix composites (CMCs) combine high chemical and thermal stability with moderate creep rates and high strength. Among the CMCs, those based on SiC and Al2O3 are most common. Compared to SiC, Al2O3 offers higher resistance against oxidation and corrosion [3]. For this reason, an alumina-based CMC was chosen in this study.

Nevertheless, at temperatures above 1200 ◦C, the water vapor in the combustion atmosphere causes corrosion reactions and the formation of volatile hydroxides. Environmental barrier coatings (EBCs) are needed to protect the material from degradation [4].

$$\text{Al}\_2\text{O}\_3\text{ (s)} + 3\text{H}\_2\text{O}\text{ (g)} \rightleftharpoons 2\text{Al}(\text{OH})\_3\text{ (g)}\tag{1}$$

An EBC should have a high resistance against water vapor at elevated temperatures; therefore, the corrosion rates of many materials were studied in the past (see Fritsch et al. [5–8], Herrmann et al. [9]). Besides the corrosion resistance, the coefficient of thermal expansion (CTE) is a crucial parameter for the performance of an EBC. Large CTE differences between substrate and EBC cause stresses during heating and cooling and may lead to premature failure. Therefore, different materials were categorized according to their corrosion rates and their CTE and presented in Figure 1. For better orientation, data for SiC/SiC and Al2O3/Al2O3 CMCs were added. It can be seen from Figure 1 that different CMCs require different type of coatings. Rare-earth disilicides are promising coating candidates for SiC/SiC composites [10,11], while rare-earth monosilicides and yttrium aluminates are potential EBC candidates for Al2O3/Al2O3 CMCs.

Some of these materials have been studied as potential EBC for oxide/oxide CMCs. Among these materials, yttria-stabilized zirconia (YSZ) coatings have been intensively investigated, since YSZ is the material of choice for thermal barrier coatings (TBCs). The suitability of plasma-sprayed, sputtered, and electron-beam physical vapor deposition (EB-PVD) YSZ coatings on wound highly porous oxide (WHIPOX) [12] has been studied by Braue et al. and Mechnich et al. [13–15]. To increase the bonding between substrate and coating, a reaction-bonded alumina (RBAO) [16] bond coat had to be used. Because of the CTE difference between CMC and YSZ, the coating adhesion was quite weak. Furthermore, it has been reported by Vassen et al. [17] that mechanical treatment of CMCs can cause severe damage to the matrix or even its failure.

**Figure 1.** Corrosion rates and coefficient of thermal expansion of selected materials (blue boxes) and ceramic matrix composites (CMCs) (red circles) (data from references [5,7,18]).

Mechnich et al. [19] studied the adhesion of atmospheric plasma-sprayed (APS) Y2O3 coatings on WHIPOX substrates that were previously coated with an RBAO bond coat. The coating adhesion was tested by furnace cycling tests at 1200 ◦C. The results showed a good adhesion and no failure after 500 cycles. SEM investigations revealed the formation of yttrium aluminates at the coating–substrate interface. The formation of this reaction layer is believed to result in a good coating performance.

Atmospheric plasma-sprayed Gd2Zr2O7 coatings on an Al2O3-based CMC has been studied by Gatzen et al. [20]. Pull adhesion tests showed weak adhesion of Gd2Zr2O7 coatings on substrates without pretreatment, while the coating adhesion on laser-structured substrates was significantly increased.

Gerendas et al. [21] published a study on the performance of several APS-coated systems (YSZ, spinel, YSZ/silicate, mullite/silicate) on different oxide/oxide CMCs (UMOXTM, WHIPOXTM, OXIPOLTM (Oxidic CMC based on Polymers)). All substrates were sandblasted before coating application, and in the case of WHIPOX, an RBAO bond coat was additionally applied. The coating performance was tested with burner rig tests, and the results suggested a strong dependence of adhesion on the used CMC material.

All of the above-mentioned coating strategies involve surface preparation by sandblasting, laser structuring, or even the application of a bond coat. Furthermore, the used coatings offer only moderate corrosion rates, but as Figure 1 suggests, there are materials with corrosion rates one or two orders of magnitude lower and with more suitable CTEs. An example is YAlO3 (YAP, yttrium aluminum perovskite), which has a very low corrosion rate and a CTE close to that of the used CMC material. Because of the great chemical similarity of YAlO3 and Y2O3 and according to the phase diagram [22], a reaction layer may be formed at the interface (Y3Al5O12, YAG, yttrium aluminum garnet). Studies on Y2O3 coatings [19] have shown that this reaction layer is very beneficial for coating adhesion. Therefore, coatings with good adhesion and long service life are expected. However, there are no studies yet on the suitability of YAlO3 either as TBC or as EBC.

In the past, the production of plasma-sprayed yttrium aluminate coatings (YAG, Y3Al5O12) proved to be particularly challenging [23]. The challenging behavior of Y3Al5O12 may also be relevant for YAlO3 coatings. Nevertheless, YAlO3 is a promising EBC candidate. Therefore, plasma-sprayed YAlO3 coatings as EBC for an Al2O3-based CMC will be discussed in this study, paying special attention to the phase evolution during plasma spraying.

Because of the expected chemical bonding between coating and substrate, which is believed to cause good coating adhesion and to minimize the processing effort, no additional bond coat or surface treatment was used in this study. The coating adhesion of the plasma-sprayed coatings was studied by means of pull–adhesion tests (PAT) and thermal cycling tests. Furthermore, the resistance of the YAlO3 coatings against CMAS corrosion was tested.

#### **2. Materials and Methods**

The CMC used for this study was the commercially available material FW12 (Pritzkow Spezialkeramik, Filderstadt, Germany), consisting of alumina fibers (Nextel 610) that are embedded in a porous matrix of 85% alumina and 15% yttria-stabilized zirconia (YSZ). Before coating application, the samples were ultrasonically cleaned.

The coatings were applied by plasma spraying. The coatings were produced by means of APS and very low pressure plasma spraying (VLPPS). The coating parameters are described in Table 1. For the production of the APS coatings, an Oerlicon Metco multicoat-facility (Wohlen, Switzerland) was used, equipped with a TriplexPro-210 gun that was mounted on a six-axis robot. The VLLPS coating runs were carried out in a vacuum chamber (2 mbar), using an O3CP gun (Oerlicon Metco, Wohlen, Switzerland). The substrate temperatures during processing were measured by IR pyrometry.


**Table 1.** Used coating parameters for the production of YAlO3 coatings.

SLPM = Standard Liter Per Minute.

In both processes, a spray-dried YAlO3 powder with a particle size of 37 μm and a purity of 96% was used. X-ray diffraction (XRD) measurements (Figure 2) revealed the presence of Y4Al2O9 and Y3Al5O12 as secondary phases.

**Figure 2.** XRD measurement (**a**) and SEM images (**b**,**c**) of the used YAlO3 powder. The diffraction pattern of YAlO3 is shown in red.

After coating manufacturing, the samples were furnace-cycled in air for 4 × 20 h at 1200 ◦C. In order to assess the thermal and chemical stability of the coatings, XRD measurements were carried out before and after thermal aging. The XRD measurements were carried out with a Bruker D4 Endeavor (Karlsruhe, Germany), using Cu-Kα-radiation (λ = 1.54187 Å). Rietveld refinements were carried out using FullProf suite [24].

The coating adhesion of the as-sprayed coatings was measured with an Elcometer 510 (Aalen, Germany), according to ASTM D4541 [25]. Test dollies with a 10 mm diameter were used and glued to the sample with Araldite two-part epoxy adhesive. Furthermore, the coating performance during thermal cycling was tested by burner rig tests, using the setup described by Traeger et al. [26]. 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 sides. The used temperature program consisted of 510 cycles at 1200 ◦C, followed by 500 cycles at 1300 ◦C. The CMAS stability of the samples was tested with a special burner rig, described by Steinke et al. [27]. The surface temperature was set to 1250 ◦C to ensure melting of the CMAS components.

Materialographic cross sections of the samples before and after testing were prepared. For this, the samples were embedded into resin, cut, and wet-ground with successively finer abrasive paper down to a grit designation of P4000. Afterwards, the samples were polished with diamond suspensions. The polished samples were sputtered with platinum (Leica EM ACE200, Vienna, Austria) and analyzed by scanning electron microscopy (SEM) (Hitachi, TM3000, Tokyo, Japan).

#### **3. Results and Discussion**

#### *3.1. Coating Formation*

#### 3.1.1. Atmospheric Plasma Spraying

The YAlO3 coatings were produced using APS at varying stand-off distances and resulting different coating temperatures. The XRD-measurements of the APS YAlO3 coatings before and after thermal cycling are presented in Figure 3. The coating sprayed at higher stand-off distance and thus lower substrate temperature was amorphous in the as-sprayed state. The thermal cycling led to the crystallization of the coating. Phase analysis revealed the presence of YAlO3, Y3Al5O12, and Y4Al2O9. The refined phase content of the different phases is given in Table 2. Phase analysis showed that the three yttrium aluminates occurred in almost equal fractions.

**Figure 3.** XRD measurements before (light) and after (dark) thermal treatment of APS YAlO3 coatings; theoretical diffraction patterns of Y4Al2O9 (black) and YAlO3 (red).


**Table 2.** Results of Rietveld refinements of APS YAlO3 coatings.

The coating sprayed at lower stand-off distance and thereby higher substrate temperature was crystalline in the as-sprayed state. The main phase was the desired YAlO3 phase, but significant amounts of Y3Al5O12 and Y4Al2O9 were present. The heat treatment led to further crystallization and phase segregation of the coating. The composition shifted in favor of the desired YAlO3 phase.

To enable crystallization of the YAlO3 coating, a certain energy level must be exceeded. Furthermore, the cooling rate should be slow, so that there is enough time for atomic rearrangement and the formation of a long-range order. The cooling rate can be lowered by using higher plasma power, by shortening the stand-off distance, or by substrate preheating. The splat solidification time [28] is a key factor. A dependence of the splat solidification time on the temperature gradient between splat and substrate has been reported [28,29]. A decreased gradient leads to a longer solidification time, consequently the splats deposited at higher temperatures have more time for crystallization. This explains why the YAlO3 coatings, sprayed at shorter distances (70 mm) and higher substrate temperatures (750 ◦C) were crystalline, while those sprayed at 120 mm distance and a significantly lower temperature (550 ◦C) were amorphous.

The formation of secondary phases can be explained by the evaporation of elements during thermal spraying, causing a shift in the phase diagram [23]. In this case, evaporation of alumina might occur, since alumina has a higher vapor pressure then yttria [30]. A more probable explanation is demixing due to rapid quenching of splats and impure base material (see Figure 2). This is supported by the fact that the coating, which was sprayed at higher distance and lower substrate temperature resulting in a larger temperature gradient between splat and substrate, differed significantly more from the desired phase composition. Furthermore, yttrium aluminates with both high alumina content and low-alumina content were found. If evaporation of alumina was the main mechanism, the phase equilibrium would be shifted to the phase with low alumina content (Y4Al2O9).

SEM images of cross sections of YAlO3 coatings before and after thermal cycling are presented in Figure 4. The amorphous coatings sprayed at higher distances (120 mm) and thus lower substrate temperatures were relatively dense and showed only few cracks. The presence of secondary phases, which was already revealed by XRD-measurements, can also be observed in the SEM images. The SEM images reveal that the coating was delaminated during thermal cycling. Coating delamination might be attributed to the occurring crystallization of the coating.

**Figure 4.** SEM images of the APS YAlO3 coatings before (Coatings sprayed at a distance of 70 mm (**a**) and 120 mm (**b**)) and after thermal treatment (Coatings sprayed at a distance of 70 mm (**c**) and 120 mm (**d**)).

The coatings obtained at a smaller stand-off distance (70 mm) were already crystalline in the as-sprayed state. However, large pores were present in these coatings, especially at the interface between the different coating layers. Sporadic delamination was observed even in the as-sprayed state, and the coatings were prone to form segmentation cracks. Both cracks and pores have a negative effect on corrosion resistance, as they increase the permeability of the EBC to water vapor. Here, the segmentation cracks seemed to continue into the substrate, which could additionally affect the structural integrity of the material.

These results are in good agreement with the results for APS Y3Al5O12 coatings of Weyant et al. [24]. A systematic investigation of the effects of several spraying parameters on the resulting coating microstructure led to the conclusion that the used power and the stand-off distance are the determining factors. Unfortunately, increasing spraying distances led to both lower porosity and reduced crystallinity. Moreover, increased power caused increased crystallinity, but the porosity was increased as well [24]. The same behavior was found in this study for atmospheric plasma-sprayed YAlO3 coating:

The formation of pores seemed to go along with the crystallization of the as-sprayed YAlO3 phase. This can be attributed to the differences in density between melt and solid. It is assumed that the density of amorphous YAlO3 is close to that of the liquid phase, which is slightly below 4.0 g·cm−<sup>3</sup> [31–33]. The density of the crystalline phase is reported to be significantly higher, at 5.35 g·cm−<sup>3</sup> [34]. These differences lead to shrinkage during crystallization and therefore cause the formation of large pores and cracks. In the case of initially dense but amorphous coatings, the tensile stresses that occur during the crystallization processes resulting from the thermal treatment are high, and as a consequence, the coating is delaminated.

The formation of a reaction layer, as reported for Y2O3 coatings, was not observed. It is assumed that the stresses occurring in these coatings were so high that delamination occurred before the reaction could take place. Furthermore, poor wetting could also be a reason for the absence of a reaction layer.

#### 3.1.2. Very Low Pressure Plasma Spraying

Coatings were produced with VLPPS, in order to overcome the aforementioned problems related to crystallization in the APS YAlO3 coatings. VLPPS offers high power, high temperatures and low cooling rates. The results of the VLPPS YAlO3 coatings are presented in Figure 5.

**Figure 5.** (**a**): XRD-measurement of the VLPPS YAlO3 coatings in the as-sprayed and thermally treated state. (**b**,**c**): SEM images of the VLPPS YAlO3 coatings after thermal treatment.

The XRD measurements showed that the YAlO3 coatings were crystalline directly after spraying. The major phase was YAlO3, and only very weak reflections of a secondary phase (yttrium aluminum monoclinic, YAM, Y4Al2O9) were visible. Due to a slight bending of the coated sample, the measured XRD signals were slightly offset from the theoretical ones. The XRD measurement of the thermally treated sample showed hardly any deviations from the original XRD. This shows that there were no phase transformations or significant crystallization processes during heating/cooling, which could cause stresses and cracks in the coating. The formation of nearly single-phase YAlO3 can be attributed to the lower cooling rate. The increased surface temperature led to an increased solidification time. Furthermore, the low chamber pressure led to significantly slower cooling rates of the coating and substrate. As a consequence, the coating had more time to crystallize. Therefore, the resulting coating had a high crystallinity and a high content of the desired YAlO3 phase. Rietveld refinements led to a phase content >88% of the desired YAlO3 phase. This represents a significant increase compared to the coatings produced under atmospheric conditions.

The SEM images of the cross sections of the VLPPS YAlO3 coatings are shown in Figure 5a. The coating had a finely distributed porosity, which was caused by not fully molten particles. The pores were distributed homogenously in the coating. In contrast to APS coatings, no vertical cracks were observed. The VLPPS YAlO3 coatings formed a reaction layer at the interface with the substrate. This reaction zone can be observed even in the as-sprayed samples (Figure 6). This implies that the energy needed for the reaction [35] was reached in these experiments. The higher temperatures might also increase the wetting of the splats and therefore increase the contact between coating and substrate, leading to the formation of a 1–2 μm-thick reaction zone. Because of the position of YAlO3 in the Y2O3–Al2O3 phase diagram [36], the formation of Y3Al5O12 is likely. In contrast to the results for the APS coatings, there was no delamination within the coating–substrate system after heat treatment. Due to the high crystallinity and the formation of chemical bonding between coating and substrate, a strong coating adhesion is expected.

**Figure 6.** SEM image of the VLPPS YAlO3 coating in the as-sprayed state, with a Y3Al5O12 reaction layer at the coating substrate interface.

#### *3.2. Pull–Adhesion Tests*

The coating adhesion was tested with pull–adhesion tests, and the results are illustrated in Figure 7. The adhesion of the APS coatings was very poor, and both coatings failed directly at the beginning of the test. High stresses in the coatings might be the reason for the bad coating adhesion. The adhesion of the VLPPS coatings was significantly higher. The measured strength was in the order of that of the CMC itself.

**Figure 7.** Measured adhesion strength of as-sprayed APS and VLPPS YAlO3 coatings.

The samples were embedded and cut after adhesion testing. SEM images of the polished cross sections are shown in Figure 8. The high stresses in the APS coatings led to failure partly within the CMC and partly at the coating–substrate interface. It is assumed that slight reactions between coating and substrate or clamping occurred, which caused a partial failure within the matrix of the CMC.

The good adhesion between the VLPPS coating and the CMC caused a failure deep within the substrate itself. This is consistent with the measured adhesion strength, as the measured adhesion strength was in the order of the strength of the CMC itself. The high adhesion strength was attributed to the occurrence of a chemical reaction at the coating–substrate interface.

**Figure 8.** SEM images of the APS YAlO3 coatings sprayed at 120 mm (**a**) and 70 mm (**b**) and of the VLPPS YAlO3 coating after pull–adhesion tests PAT (**c**).

#### *3.3. Burner Rig Tests*

The pull–adhesion tests of the VLPPS YAlO3 coatings revealed a promising adhesion strength. On the basis of previous investigations on Y2O3 coatings resulting in excellent stability due to the formation of a reaction layer at the interface [20], a high thermal cycling lifetime of the VLPPS coatings was expected. The APS coatings showed poor adhesion strength, and as a consequence, these coatings were excluded from the following investigations.

The thermal cycling tests were stopped after 1010 cycles (510 cycles at 1200 ◦C and 500 cycles at 1300 ◦C), which corresponds to the upper limit of the average lifetime of standard TBC samples in this test rig. During cycling, no macroscopic failure occurred (see Figure 9). SEM images of the VLPPS YAlO3 coatings after 1010 cycles are shown in Figure 10. The coating still shows a fine distributed porosity. The Y3Al5O12 containing reaction zone at the coating substrate interface was about 1–2 μm-thick. No growth took place within the duration of the test.

**Figure 9.** Photograph of VLPPS YAlO3 coating on FW12 after 1010 cycles of burner rig testing.

**Figure 10.** SEM images of cross sections of VLPPS YAlO3 coatings after 1010 cycles of burner rig testing (**a**), close up of the coating substrate interface (**b**).

No delamination or crack formation was observed after 1010 cycles of burner rig testing; therefore, a significantly longer lifetime can be expected. The long cycling lifetime of the YAlO3 coatings on the FW12-CMC is attributed to the strong bond between the coating and the substrate due to the formation of a thermodynamically stable reaction layer at the interface.

#### *3.4. CMAS Tests*

The coating's resistance to CMAS corrosion was also investigated. The tests were stopped after 274 cycles, which corresponds to about twice the average lifetime of a typical YSZ coating in this rig. After 274 cycles, no macroscopic coating failure was observed (see Figure 11). The XRD measurements of the sample before and after the CMAS test are shown in Figure 12. Besides signals from the CMAS and the coating itself, the formation of calcium–yttrium silicate oxyapatite (Ca2Y8(SiO4)6O2) as a reaction product between YAlO3 and CMAS was observed. This phase was already observed for other Y-containing coatings [37,38]. In addition, the formation of Y-depleted yttrium aluminate Y3Al5O12 was observed accordingly. Turcer et al. [38] recently reported the excellent CMAS resistance of dense YAlO3 pellets due to the fast formation of a protective Ca–Y–Si apatite phase. Our results are in good agreement with those from Turcer et al.

**Figure 11.** Photograph of VLPPS YAlO3 coating on FW 12 after 274 cycles in CMAS test rig.

**Figure 12.** Measured XRD (grey) of VLPPS YAlO3 coating after 274 cycles in CMAS test and theoretical diffraction patterns of YAlO3 (black), Y3Al5O12 (blue), Ca2(Al0.92Mg0.08)((Al0.46Si0.54)2O7) (green), and Ca2Y8(SiO4)6O2 (red).

The SEM images of the resulting cross sections are presented in Figure 13. No indications for crack formation or beginning delamination were found. Because of the porosity of the sample, the molten CMAS was able to infiltrate the upper part of the VLPPS YAlO3 coating (about 46 μm). The SEM images and the corresponding EDS measurements show the sharp borderline of the infiltrated zone. The reaction zone between CMAS and YAlO3 can be clearly seen. The CMAS infiltration and the following reactions led to densification of the infiltrated parts of the coating. The densification of the upper coating may hinder further infiltration and thus extend the lifetime of the coating. Furthermore, the CTE of the CMAS constituents is in the area of 9.7 <sup>×</sup> 10−<sup>6</sup> K−<sup>1</sup> [39,40] and close to the CTE of YAlO3 (8.9 <sup>×</sup> <sup>10</sup>−<sup>6</sup> <sup>K</sup>−<sup>1</sup> [41]); thus, low stresses were expected during testing. Because of the remaining porosity of the lower coating parts, occurring stresses can be relaxed. As a consequence, even after 274 cycles, no spallation occurred.

**Figure 13.** SEM images of the VLPPS YAlO3 coatings after 274 cycles in the CMAS test rig (**a**,**b**). EDS mappings for Ca (**c**) and Y and Al (**d**).

#### **4. Conclusions**

The suitability of YAlO3 coatings as EBC for an Al2O3/Al2O3–CMC was examined. YAlO3 coatings were produced by means of APS and VLPPS, without applying any bond coats. Furthermore, no surface pretreatment was carried out. The APS YAlO3 coating trials resulted in coatings with either poor crystallinity or large pores, both of which are not favorable for an EBC.

YAlO3 coatings produced by VLPPS offered high crystallinity and purity. The VLPPS coatings showed strong adhesion, which was attributed to the formation of chemical bonding between coating and substrate. This thermodynamically stable reaction layer at the coating–substrate interface, probably consisting of Y3Al5O12, was found directly after spraying.

The VLPPS YAlO3 were chosen for further investigations. The thermal cycling lifetime was tested with burner rig tests. The coatings passed the test consisting of 1010 cycles without failure. Furthermore, the hot corrosion behavior was tested with CMAS tests. The coatings showed excellent CMAS resistance, by withstanding more than 274 cycles without failure. The high cycling lifetime was attributed to the densification of the coating due to CMAS infiltration and formation of a reaction product, blocking further infiltration.

The use of VLPPS for the production of YAlO3 coatings led to stable coatings with a microstructure with intermixed porous and dense zones. This led to a high thermal cycling lifetime and high CMAS resistance. Furthermore, the formation of a reaction zone at the interface enabled good coating adhesion. Therefore, VLPPS YAlO3 coatings are promising candidates for EBCs for Al2O3-based CMCs, as they offer high adhesion strength, high thermal stability, and high corrosion resistance.

**Author Contributions:** Conceptualization and methodology, C.G. and D.E.M.; investigation, C.G.; writing—original draft preparation, C.G.; writing—review and editing, D.E.M., R.V. and O.G.; supervision, D.E.M. and R.V.; project administration, R.V.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors would like to thank Ralf Laufs, Frank Kurze, and Karl-Heinz Rauwald for their assistance during plasma spraying. The authors acknowledge Doris Sebold for the SEM analysis. In addition, tha authors would like to thank Volker Bader and Martin Tandler for the execution of the heat treatments, burner rig tests, and CMAS tests.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
