Introduction

The need for increasing of power and efficiency of gas turbines for land and aviation applications is inevitably connected with combustion of fuel at high temperatures. Superalloys that make up the turbine components are not able to withstand these high temperatures therefore thermal barrier coatings (TBC) have been used for several decades in order to provide thermal protection (Ref 1). However, conventional air plasma sprayed yttria-stabilized zirconia coatings are reaching the limit of their capabilities. Development of new generation coatings is required to push the technology to higher efficiency levels (Ref 2).

Crucial factor for new TBCs is lowering of thermal conductivity of the coatings. It has been demonstrated that thermal conductivity of zirconia can be lowered by substitution of a stabilizer; instead of standard stabilizing with yttrium oxide, another rare earth oxide can be used (Ref 3). Usage of dysprosium was shown to be suitable because of its greater ionic volume than yttrium, causing increasing phonon scattering and therefore lowering thermal conductivity, especially at high temperatures (Ref 4). Another way of lowering thermal conductivity can be done by introducing a high level of porosity into the coating (Ref 5). Porous structure of the topcoat is not only beneficial because of low thermal conductivity, but the presence of pores also enables the coating to accommodate to the stresses generated during high temperature cyclic exposure (Ref 6).

Besides low thermal conductivity of the topcoat; other important requirements are long coating lifetime and stable properties for the whole service life. The exposure of coatings to high temperatures during the service can lead to sintering, which causes densification of the coating and therefore loss of its beneficial thermo-mechanical properties (Ref 7). Reduction of the sintering rate is a key goal for coating development and previous work demonstrates it can be reduced by lowering the impurity level of the powder feedstock (Ref 8) and by introducing large globular pores, which can resist sintering better than delaminations and small pores (Ref 9).

Previous studies by the authors had demonstrated promising lifetime and low thermal conductivity of coatings generated with large globular pores. These pores were created by the use of a spray powder containing a polymer-pore former (Ref 5, 6, 9). While such mixed powders have been used in the past for high temperature abradable coatings, extensive work to optimize the coatings for TBC applications had not been carried out.

Therefore, the aim of this paper has been to study the influence of spraying parameters during preparation of high porosity TBCs generated with a porosity former. A design of experiments approach has been used to evaluate spray parameters that influence both thermal conductivity and lifetime of the thermal barrier coating.

Materials and Methods

Sample Production

For all the APS experiments reported in this study, a F4-MB plasma torch (Sultzer Metco, Wholen, Switzerland) was used for deposition. The design of experiments approach was taken for creation of the test matrix with a 2-level full factorial design being used. Three factors were chosen to alter in the spray process:

  • Spray distance

  • Powder feed rate

  • Hydrogen Flow

The matrix results in a 9 run design with an additional two repeat runs at the centre point; thus giving a total of 11 spray runs as represented in Table 1. The experiments 9-11 served to verify the stability of the process and they were sprayed using parameters close to standard industrial deposition parameters of YSZ (Ref 4). In the other experiments, deposition parameters were either lowered (−) or increased (+) in comparison to the standard conditions 9-11.

Table 1 Parameters of the DoE
Full size table

Bond coats for all samples were produced using standard AMPERIT 410 NiCoCrAlY powder (H.C. Starck, Munich, Germany), sprayed with F4-MB gun to the thickness 170 µm. Substrates used in this study were Hastelloy X in dimensions 25.4 × 25.4 × 2 mm for microstructure and thermal conductivity measurements. For thermo-cyclic fatigue test samples in dimensions 50 × 30 × 5 mm were used and coupons Ø25.4 × 5 mm were coated for additional tests. All the substrates used for spraying were fixed to rotational fixture and grit-blasted and bond-coated in one run to achieve identical conditions for later application of the topcoat.

Ceramic coatings were produced with agglomerated & sintered (A&S) high-purity dysprosia-stabilized zirconia powder mixed with polymer porosity former (Sultzer Metco, Wholen, Switzerland). Polymer particles are added to the powder to create microstructure with porosity not achievable with conventional powder. All the ceramic topcoats were sprayed to achieve 700 µm thick coating.

In-flight properties of the particles were measured using DPV 2000 (Tecnar Automation, Montreal, Canada). Measurement of temperature, diameter, and velocity of individual particles entering the plasma jet were made for about 5000 particles in the spray stream center that was performed by an automated routine for each condition. Mapping of the spray stream was performed using 5 × 5 array with equispaced points with 5 mm separation and measuring approximately 300 particles in each location.

Isothermal Heat Treatment

In order to burn out the polymer from the coating and form a complex structure of pores, all the samples were heat treated at 550 °C for 2 h in the air atmosphere. Additionally, one sample from each run was heat treated at 1150 °C for 50 h, which leads to sintering of the structure, changes in thermal properties of the coating and formation of TGO at the interface. In this case, heat treatment is a simplification of aging which appears in real service.

Thermo-Cyclic Fatigue

Thermo-cyclic fatigue testing is performed as a form of lifetime assessment. Testing involves placing test samples in the hot zone of a furnace at 1100 °C for 60 mins. After the high temperature cycle is complete, the samples are removed from the hot zone to a cooling stage where compressed air is used to force cool the samples to 100 °C in under 10 mins. Failure criterion is deemed to be 20% spallation of the ceramic coating. In total, 3 samples were tested for each experimental run.

Thermal Conductivity

Thermal properties of As-sprayed samples (with burned out polymer) and Heat-treated samples were evaluated using the laser flash technique. A Netzsch Laser Flash Analysis 427 (Netzsch Gebrätebau GmbH, Germany) was used to measure the diffusivity of the system at room temperature.

Two round disks Ø10 mm were water cut from coated samples for each condition (As-sprayed and Heat-treated) and used to evaluate thermal properties. As zirconia is transparent to light of the wavelength produced by LFA laser, all the samples were coated with a thin layer of graphite before placing them to the testing device. During the test, a laser pulse is fired at the back of the sample, heat generated by laser pulse is transferred through the sample and a measurement of the temperature change is made on the front side of the sample by infra-red detector. Each measurement is an average of five shots. Diffusivity α (m2/s) of the system can be calculated using the following formula (Ref 10):

$$\upalpha = \frac{0.138 \cdot L^{2}}{t(0.5)},$$

where L is the sample thickness and t(0.5) the time corresponding to the half maximum increase in temperature. Thermal conductivity λ (W·m−1·K−1) can be easily calculated from thermal diffusivity when specific heat capacity C p (J·kg−1 K−1) and density ρ (kg/m3) are known using the formula (Ref 10):

$$\uplambda = \upalpha \cdot C_{\text{p}} \cdot \rho$$

In this case, measurement of thermal properties of the full system consisting of 3 layers, i.e., substrate, bond coat, and topcoat was performed. The contribution of thermally grown oxide layer for the heat-treated samples to the whole is very small, owing layers low thickness (usually about 5 µm), and can therefore be neglected (Ref 9). To separate thermal properties of the topcoat from the measurement, it is necessary to know thermal conductivities of the other layers; then thermal properties of the topcoat can be calculated using data for the other layers and the rule of mixtures. The data for substrate and bond coat layer used for the calculation were taken from the previous study (Ref 5, 9).

Microstructure and Porosity Analysis

Samples produced for microstructure were sectioned using a diamond wafering blade. The cross section of each sample was mounted in low viscosity resin and polished semi-automatically using Buehler PowerPro 5000 (Buehler, USA) and a standard TBC polishing method. Morphology of the samples was assessed using optical and scanning electron microscopy Hitachi TM3000 (Hitachi, Japan).

Porosity measurements were carried out using image analysis routine developed previously (Ref 11). In this procedure, 25 images at a magnification 800× (using SEM) were taken along the cross section of the coating to fully represent the microstructure. The images are then processed using Aphelion image analysis software (ADCIS, France). The image is converted into binary format and an auto-thresholding technique is applied to avoid errors caused by operator. The software is able to differentiate between cracks and pores using the height-to-length ratio, evaluate size distribution of pores, and also calculate the orientation of the cracks.

Porosity and pore size distribution was also determined by mercury intrusion porosimetry (MIP) carried out on PoreMaster 33 GT (Quantachrome Corporation, Boynton Beach, FL). For the measurements, coatings were detached from the substrates using nitro-hydrochloric acid. Substrate side of the coating surface was polished to remove substrate residues. Prior to the measurement, coatings were heat treated in a furnace at 120 °C to remove adsorbed water from the inner surfaces. Due to the small sample volumes, blank cell correction was applied to increase accuracy of the measurements.

Results and Discussion

Particle In-Flight Data

The graph in the Fig. 1 shows the distribution of temperature for conditions N2 (Low Hydrogen flow), N10 (standard), and N7 (High Hydrogen flow). The light-gray area in the graph represents temperature below melting point of DySZ powder.

Fig. 1
figure 1

Temperature distribution of particles for different conditions

Full size image

It can be observed that as hydrogen flow is increased, a greater proportion of the spray particles have temperatures above the melting point. Hydrogen improves the heat transfer within the plasma stream. As a result of the three levels of hydrogen flow, the results can be grouped into “hot”, “cold”, and “standard” depositions conditions.

Porosity and Microstructure

Porosity level of the topcoats varies from 20% porosity for the coating N7 up to 29% porosity for the coating N4. Total porosity and presence of cracks of the single coatings are displayed in the Fig. 2. There can be clearly seen that the highest porosity levels are obtained for “cold” conditions (experiment numbers N1-N4). Similarly low porosity level was achieved for “hot” conditions; standard conditions give the coatings with medium porosity level about 26%. Despite the significant differences in total porosity, only marginal changes were observed in variation of number of cracks in the coatings presented in Fig. 2.

Fig. 2
figure 2

Porosity level of the topcoats

Full size image

Cross sections of the samples showing the differences in porosity are presented in the Fig. 3, 4, and 5. Presence of large globular pores is characteristic of the use of a pore former and is apparent for coatings with high porosity levels, mean size of the pores decreases with increasing the hydrogen flow.

Fig. 3
figure 3

Cross section of the coating N7 containing low level of porosity (20%)

Full size image
Fig. 4
figure 4

Cross section of the coating N10 containing medium level of porosity (26%)

Full size image
Fig. 5
figure 5

Cross section of the coating N2 containing high level of porosity (29%)

Full size image

Total thickness of the topcoats was desired to be 700 µm and the total number of passes of the torch needed to achieve full coating thickness was calculated from a few first spraying passes. The real thickness of the topcoats is shown in the Fig. 6, together with thickness increase after a pass of the spray torch.

Fig. 6
figure 6

Total thickness of the topcoats for single experiments

Full size image

Thermal Property Results

Thermal conductivity of the As-sprayed (AS) and Heat-treated (HT) samples evaluated by LFA at room temperature and their relative increase caused by sintering processes during the heat treatment can be read from the Table 2. The study presumes that porosity level of the coatings does not change significantly with heat treatment, based on previous study of the coatings generated with porosity former (Ref 9). This behavior is caused preferentially by presence of large globular pores in the coating, which are resistant to the sintering effect. The key factor, which causes the increase of thermal conductivity of the heat-treated samples, is micro-crack healing. Especially, the loss of horizontal cracks gives an increase in thermal conductivity.

Table 2 Thermal conductivity of As-sprayed and heat-treated samples
Full size table

The lowest thermal conductivity in the As-sprayed state is not necessarily an indication of the best performance. The crucial factor is the sintering resistance of the coatings, which varies significantly for presented coatings. The relative increase in the thermal conductivity after the isothermal heat treatment for the coating N2 (48%) compared to N4 (102%) that is shown in Table 2, demonstrates a large range in sintering rate for the coatings presented in this study.

The dependence of thermal conductivity on total porosity level of the coatings is presented in Fig. 7, showing that the level of total porosity is not the key factor for achieving of the coatings with extremely low thermal conductivity, nor in maintaining that after thermal exposure.

Fig. 7
figure 7

Dependence of thermal conductivity of the topcoats on total porosity level

Full size image

The thermo-cyclic fatigue lifetime for all tested samples is presented in Fig. 8. Failure in TCF testing is predominately driven by bond coat oxidation and the growth of an oxide layer at the ceramic/bond coat interface. As all coatings in this study were deposited on bond coats of identical composition, produced during the same spray run; relative differences in coating lifetime can be attributed to influence of the ceramic top coat. There is notable spread in results for coatings N9, N10, and N11; although these coatings being repeat points should display similar lifetimes. The coating thickness for N9 samples may result in a slightly higher lifetime in TCF testing relative to the other two centre point; however, a standard deviation of ±30 cycles is common in the TCF test. The accelerated failure of the sample N3 may be attributed to its considerably higher thickness (see Fig. 6), and therefore increased stresses generated during the testing. Overall, the lifetime does not vary greatly within the experimental samples.

Fig. 8
figure 8

TCF life of the samples

Full size image

Changes in phase composition of the topcoats, caused by heat treatment were evaluated using XRD. Diffraction patterns of the feedstock, As-sprayed and Heat-treated coatings are shown in Fig. 9. Tetragonal zirconia stabilized by dysprosia (DySZ) and Baddeleyite (monoclinic phase) were observed within the coatings. Volume fraction of monoclinic phase was calculated from the diffraction patterns using Rietveld analysis. The feedstock material contains about 30% of Baddeleyite, which is reduced during the deposition of the coating to 5% of Baddelyite for As- sprayed samples. During heat treatment, the amount of monoclinic phase increases and saturates at about 16% for Heat-treated samples, such behavior was reported earlier (Ref 12) and is typical for high-purity dysprosia-stabilized zirconia powder. Volume changes connected with phase transformation from tetragonal phase to monoclinic phase cause internal stresses within the coating and may accelerate evolution of spallation during thermo-cyclic fatigue.

Fig. 9
figure 9

Diffraction patterns of the feedstock and topcoats

Full size image

Model Fitting

Analysis of the design of experiments matrix was performed using the commercially available software Modde (Umetri AB, Sweden). Model fitting was performed using partial least squares, with some non-influential interaction effects excluded for better model fitting. The observed versus predicted effects for the three most important responses are shown in Fig. 10 and 11.

Fig. 10
figure 10

Observed versus predicted results for thermal conductivity after heat treatment

Full size image
Fig. 11
figure 11

Observed versus predicted results for Thermo-cyclic fatigue

Full size image

Of general note in the presented results, there is a large variation in results for N9, N10, and N11 spray runs. Since these runs are produced as replicated centre points, the results demonstrate the variability of the process with this particular spray powder. This may be further attributed to the in-homogenous nature of the produced coatings.

The overall fit for deposition rate and porosity is quite high. However, the variability in performance means thermo-cyclic fatigue and thermal conductivity have a lower level of predictability. Improvement of the model may be achieved through a number of additional spray runs based on the edges of the design space rather than only the extreme corners as was carried out here.

Conclusions

Thermal barrier coatings for industrial application are required to have low thermal conductivity after thermal exposure and high cyclic life; both these qualities have to be reached maintaining relatively high deposition efficiency to keep the process cost effective. In the presented study, coatings with porosity level in the range from 20% to 29% were prepared using a porosity former containing powder. By changing deposition parameters, porosity level can be effectively controlled and coatings with favorable properties obtained. It was demonstrated that thermal conductivity of the coatings is highly dependent on porosity level and distribution of cracks and large pores within the topcoat. Thermal conductivity of all the as-sprayed coatings was evaluated to be very low—in the range from 0.51 to 0.84 W/m K, but it increased to the values from 0.95 to 1.4 W/m K after heat treatment for 50 h at the temperature 1150 °C. Microstructural features appear to influence the degree of sintering, and therefore increase in thermal conductivity; though this could not be correlated to image analysis results. Thermo-cyclic fatigue properties of the ceramic coatings showed variable lifetimes within the expected limits of the test. Overall, the influence of process parameters on lifetime beyond thickness changes appears to be low.

Unfortunately, optimisation toward high deposition efficiency of spraying tends toward spray parameters that negatively influence the other performance factors. As an example, coating N2 shows possibly the best performance in the reported tests. However, this coating is produced at a low deposition rate. In order to achieve all aims, a compromise would be required in performance factors and further optimisation work would be required.