Mechanical Integrity of Thermal Barrier Coatings: Coating Development and Micromechanics

The coatings were applied with high velocity oxyfuel spray (HVOF). HVOF produces relatively dense coatings (porosity <1%) with a good adhesion and an oxide content <1% [16, 25, 26].

The materials for the HVOF process were fed in powder form. The bond coat material is a newly developed NiCuCrAl alloy [15], the top-coat material is a Ni-based superalloy Rene80 [27]. The compositions and the sizes of the powder are shown in Table 1. A WokaStar 610 gun from Sulzer Metco (now known as Oerlikon Metco) was used for coating application. The coating parameters were established in preliminary studies (see e.g. [28]) and are shown in Table 2.

For the wind tunnel tests, 200 mm \(\times \) 200 mm \(\times \) 1 mm Incoloy 800 H plates were coated on one side. To increase the coating adhesion and avoid the growth of delamination cracks, the samples were diffusion heat treated for 0.5 h at 800 \(^{\circ }\)C.

The arc-heated wind tunnel L3K at DLR, Cologne [29], was used to obtain the required aerothermal loads. The experiments were conducted at Mach 7.7, a free stream temperature of 477 K, a pressure of 50.3 Pa, and a flow velocity of 3756 m/s. The model was subjected to the flow for 120 s at an angle of attack of 20\(^{\circ }\). The panel surface temperature was measured using an infrared camera. For details on the sample geometry and the experiments see [30, 31].

The coatings were tested on a realistic cooling channel geometry in a hot-gas flow in a subscale combustion chamber. The rectangular combustion chamber is designed to disassemble and exchange a test segment as part of the chamber wall. This test segment is made of the copper alloy CuCr1Zr and has a realistic cooling channel geometry for nitrogen as coolant. For details on the test bench and the experimental conditions, see [3].

The coating was applied on the hot-gas side of the test segment. Prior to coating application, the surface of this segment was ground with P800, rinsed in acetone, and grit blasted to remove thick oxide layers and impurifications. The front edge oriented towards the hot-gas flow was rounded (0.5 mm radius) to enhance the coating adhesion on this critical edge. The sample was masked using metal sheets and adhesive tape (aluminium tape for grit blasting, fibre-reinforced high-temperature tape during coating application). The sample was cooled with pressurized air in the internal cooling channels during coating process. Tests with prototypes showed that this cooling is sufficient to keep the temperature in the copper substrate below 200 \(^{\circ }\)C during the coating process.

After the spray process, the coating surface was ground and polished to achieve a roughness of Ra 0.5 \(\upmu \)m. Afterwards, the sample was diffusion heat treated for 6 h at 700 \(^\circ \)C in an argon (Ar 4.8) atmosphere to enhance the coating adhesion.

The coated test panel was tested in the combustion chamber for 36 cycles. One cycle consists of pre cooling, 27.5 s hot-gas phase, and post cooling.

Figure 1 shows the in-plane elastic strain and the temperature in a 150 \({\upmu }\)m thick coating during one simulated combustion cycle of the 1000 kN chamber. The elastic strain is plotted for a point near the hot-gas surface since the maximum loads were observed here.

The cycle starts at room temperature and \(\varepsilon _{el.11}=0\) (Point (A) in Fig. 1). Due to the pre cooling prior to ignition, small tensile strains are induced in the coating, caused by the different coefficient of thermal expansion (CTE) of coating and substrate. The transient maximum of these stresses is caused by the fact that the thin combustion chamber wall cools down faster than the surrounding nickel jacket. After ignition, the coating’s temperature increases rapidly whereas the copper liner is still cooled, so that large compressive strains are induced in the coating (B). At (C), the yield strength of the coating is exceeded. With further heating, the temperature dependent yield strength decreases, and plastic deformation reduces the elastic strain although the overall mechanical strain is still increasing. At maximum temperature (D), the elastic strain is \(\varepsilon _{el.11}<0.1\%\), whereas the plastic strain is \(\varepsilon _{pl.11}\approx 2\%\). After the end of the hot-gas phase, the coating cools down and thermal contraction leads to large elastic tensile strains. At point (E), the yield strength is exceeded and plastic deformation is calculated in the FEM simulations. In reality, it can be expected that the coatings crack instead of plastic deformation, but this could not be considered in the FEM simulations due to the lack of reliable crack-related material parameters. The critical elastic strain for vertical cracks in the coatings is approximately 0.55% (see introduction), which is exceeded here. Consequently, vertical cracks are inevitable. Even thinner coatings showed tensile strains larger than the critical value for crack growth.

The discussion above showed that it is inevitable that coatings crack vertically near the hot-gas surface. Due to the lack of reliable material data for crack propagation in the coatings, it was not possible to model crack growth in the FEM simulations. Instead, the stress gradient through the coating after cooling down was investigated, see Fig. 2: The stress is maximal near the coating surface where the highest temperature and thus the largest plastic deformation occurs during the hot-gas phase. With increasing distance from the surface, the stresses decrease. With higher coating thickness, the stress near the substrate/coating interface becomes less, caused by the lower temperature during the hot-gas phase due to the larger thermal insulation.

It can be estimated with the stress intensity factor whether the cracks will stop near the substrate interface. For plasma sprayed metallic coatings, the critical stress intensity factor for crack propagation is \(K_{Ic}= 1\) to \(10\,\mathrm {MPa\sqrt{m}}\) [34]. Although data for HVOF-sprayed coatings were not available, it can be assumed that \(K_{Ic}\) is even larger due to the lamellar microstructure. For \(K_{Ic}=10\,\mathrm {MPa\sqrt{m}}\), the critical stress for crack propagation is for all crack lengths \(>10\,\upmu \)m below the stress in Fig. 2, so that the cracks would propagate towards the substrate with the assumed \(K_{Ic}=10\,\mathrm {MPa\sqrt{m}}\). A value of \(K_{Ic}>17\,\mathrm {MPa\sqrt{m}}\) would be sufficient to stop the cracks in the bond coat for 150 \(\upmu \)m thick coatings. This shows that future work has to focus on the one hand on the measurements of \(K_{Ic}\), and on the other hand, the bond-coat should be modified to achieve a \(K_{Ic}\) as high as possible.

The coatings designed in Sect. 3 reach surface temperatures of up to 1350 K. It had previously been assumed [37] that the maximum tolerable temperature of the top-coat surface is 1423 K, but experimental evidence was still missing.

The arc heated hypersonic wind tunnel experiment was used to expose the coatings to a supersonic gas-flow and to large surface temperatures to investigate the behaviour of the coatings under these harsh conditions. Figure 3 shows the surface temperature on the coating during the experiments. The hypersonic flow heats the panel to above 1350 K. The local temperature maximum/distribution results from the interaction of surface deformation and flow field [31]. Due to thermal expansion, the coated plate buckled with an amplitude >10 mm causing additional loads on the coatings.

A microscopical inspection of the coating surface after the wind tunnel experiment showed no erosion or coating damage. Figure 4 shows a cut through the coating at \(x=40\) mm and \(y=0\) (according to Fig. 3). The coating morphology did not change, and no degradation by erosion could be observed.

The experiments show that the coatings can withstand surface temperatures of 1350 K under supersonic flow conditions.

As discussed in Sect. 3, vertical cracks are inevitable in combustion chambers due to the large plastic strains at high temperature. It was assumed that these cracks can be tolerated in rocket engines as long as they do not lead to a substrate damage or coating delamination.

To validate these assumptions, the coating system was tested in a subscale combustion chamber. The results from the whole test campaign including also many tests with uncoated test panels are also to be published in [9]; in the following, the coated specimen is investigated in more detail.

Since a coating in an as-sprayed condition showed poor adhesion in preliminary tests and delaminated within one cycle [9], the experiments were carried out on a diffusion heat treated coating. As expected, vertical cracks could be observed in the coating after the first cycle. With additional cycles, more cracks became visible, but the coating was still adhering to the copper substrate even after 36 cycles (Fig. 5). Although a dense crack network formed in the coating, no oxidation or degradation of the copper substrate could be observed. Copper-coloured spots at the front edge (Fig. 5 bottom) are deposits from the upstream part of the combustion chamber.

Since the subscale test chamber was originally designed to provoke a damage of the cooling channel [3], the thin combustion chamber wall buckled in some areas. This caused a partial delamination of the coatings, see Fig. 6.

The heat flux and thus the temperature gradient was small compared to large scale combustion chambers. Consequently, also the stress gradient in the coating was small and the crack propagation cannot be compared to large scale engines as in Sect. 3. However, some of the cracks stopped before reaching the copper substrate, but mostly, the cracks propagated to the substrate interface (Fig. 7 top).

In some cases, the cracks kinked and propagated along the interface, or propagated into the substrate (Fig. 7 bottom). It is expected that this effect is driven by thermomechanical fatigue during the 36 test cycles. Furthermore, the temperature in the copper wall was 975 K [9], far above the maximum service temperature of the copper liner in large scale rocket engines (800 K, see Sect. 3). Thus, thermomechanical fatigue plays a major role in the subscale experiments due to these harsh conditions.