Abstract
Environmental Barrier Coatings (EBCs) are used as protective coatings for silicon carbide components in gas turbine engines to inhibit surface reactions with water vapor released by hydrocarbon combustion. In this high temperature environment, oxidizing species such as H2O and O2 react with SiC to form gaseous CO and a solid SiO2 layer on the SiC surface. However, subsequent reaction of the normally protective silica layer with water vapor leads to the formation of gaseous Si(OH)4, and volatilization of the ceramic at a rate that increases with temperature, water vapor pressure and flow rate over the surface. All three of these parameters are high in modern gas turbine engines, leading to severe volatility issues for SiC-based ceramics. The primary objective of an environmental barrier coating (EBC) then is to impede access of oxidizing species to the underlying component while maintaining adherence over long durations (in the range of 5,000 hours). The target use temperature of current EBCs is 1316 °C (2400 °F) with repeated cycling between this and the ambient temperature. While several EBC concepts have been proposed, there has been no fundamental study of their response to thermocyclic loading in water vapor environments, and little is known about the mechanisms that govern their performance and durability. Materials property data is also unavailable for many of the candidate materials in their bulk (fully dense) and coating structure forms, and little is known about the relationships between coating microstructure and the way in which the coating is applied.
One promising coating system with very low steam volatility is the ytterbium monosilicate/mullite/silicon (Yb2SiO5/Al6Si2O13/Si) tri-layer EBC. EBCs are generally applied by an air plasma spray (APS) deposition process in which particles of the coating material are melted in an atmospheric pressure, Ar-based plasma jet, and the molten droplets sprayed onto the component surface. In collaboration with the NASA Glenn Research Center, this coating system was deposited on α-SiC substrates using a high power APS approach, and the relation between the coating structure and the process environment characterized. Some of the coatings delaminated from the substrate upon cooling, which was indicative of significant variation in the adherence between the SiC substrate and Si layer. The EBC coated α-SiC substrates that did not delaminate were then tested in a steam-cycling furnace designed to reproduce some aspects of the thermocyclic and environmental conditions found during lean combustion engine operation. The thermal cycles consisted of a 60 min 1316 °C hold time in a flowing 90 % H2O and 10 % O2 environment followed by cooling to 110 °C with a hold for 10 min. All of the coatings delaminated from their substrate with coating lifetimes ranging from less than one hundred to several hundred cycles.
The primary factor contributing to rapid steam cycling failure of the coatings was rapid penetration of oxidizing species through channel (mud) cracks in the Yb2SiO5 and Al6Si2O13 layers that terminated at the Si bond coat. An analysis of the residual stresses resulting from differences in the coefficients of thermal expansion (CTE) of the three materials used in the coating and that of the SiC substrate indicated that mud cracks were formed as a result of large in-plane biaxial tensile stress in the ytterbium monosilicate and (to a lesser extent) mullite layers. The mud cracks provided a fast transport path to the silicon layer that resulted in rapid formation of a thermally grown oxide (TGO) identified to be the cristobalite phase of SiO2. The TGO layer was then found to undergo severe microfracture as it underwent a β (high) → α (low) cristobalite phase transformation on cooling through ~220 °C that was accompanied by a 4.5 vol% contraction upon cooling. The repetition of this phase transformation with cracking led to rapid oxidation of the silicon layer and failure of the system by spallation of the Yb2SiO5 and Al6Si2O13 layers above it.
An improved APS system was designed and installed at the University to enable the deposition process to be studied carefully, and an optimized coating process to be developed. In this revised process, the Si layer was deposited onto the SiC at high temperature (1200 °C) under a reducing environment to improve its adherence with the substrate and ensure the layer had a high relative density (only a few isolated pores). This eliminated premature delamination issues experienced during cooling for some of the high power deposited coatings. The effect of spray parameter selection upon the microstructure of Yb2SiO5 and its disilicate (Yb2Si2O7) counterpart (which has a CTE more closely matched to the substrate) was investigated. Secondary phases were identified and shown to result from SiO evaporation from the powder particles during plasma heating. Their volume fractions, as well as other microstructural and defect features of the layers were all investigated. Though mud cracking was again observed in Yb2SiO5 layers, no such cracking was identified in Yb2Si2O7 layers. This difference in cracking behavior was consistent with differences in the residual stress developed during cooling of the two ytterbium silicates.
The (optimized) low power Yb2SiO5/Al6Si2O13/Si EBCs were again tested in steam-cycling, but no statistically significant improvement in coating life was observed. The improved Si layer adherence and microstructure eliminated the primary delamination mode seen in the high power deposited coatings, but a new failure mechanism was observed in the optimized ytterbium monosilicate protected system. As opposed to mud crack termination at the Si bond coat, mud cracks in the low power tri-layer system bifurcated either within the Al6Si2O13 layer or at one of its interfaces. The bifurcated cracks continued to propagate towards the Si bond coat where the crack ligaments turned and propagated as horizontal delamination cracks through the mid-plane of the Si layer. Rapid oxidizing species access to the interior of the EBC still occurred through these mud cracks, but instead of oxidizing the Al6Si2O13 – Si interface, the faces of the delamination cracks within the Si layer were oxidized to form cristobalite. These cracks advanced sequentially during steam cycling until spallation occurred. In collaborations with researchers at UCSB, the thermomechanical competition between the single channel and bifurcated crack damage modes was investigated using Finite Element Analysis (FEA) combined with J-integral methods to calculate the stored elastic strain energy release rate (ERR) during fracture by the two crack propagation paths.
Simpler bi-layer Yb2Si2O7/Si EBCs were also deposited on the same SiC substrates using the optimized, low power APS process. The ytterbium disilicate layer in these coatings contained about 15 vol% Yb2SiO5 but did not mud crack upon cooling. Neither spallation failure of the coatings nor cracking of the ytterbium silicate layer was observed during steam furnace testing for up to 2,000 cycles (2,000 hours at 1316 °C). A cristobalite TGO layer was found to form slowly at the Yb2Si2O7 – Si interface, and the thickness of this layer was measured by sectioning samples extracted from the furnace after 250, 500, 750, 1,000, and 2,000 steam cycles. The maximum TGO thickness reached only ~2.5 μm after 2,000 hours of exposure, even though it had begun to microcrack and therefore had lost its protective properties. The TGO layer thickness exhibited a linear dependence upon high temperature exposure time, consistent with growth being limited by diffusion of the oxidizing species through the ytterbium disilicate layer. The linear oxidation rate constant was used to calculate a monatomic oxygen flux and effective oxygen diffusion coefficient through the Yb2Si2O7 layer. Some volatilization of Si from the surface of the Yb2Si2O7 layer was also observed, and the coating edges suffered from preferential attack. However, the steam-cycling performance of this coating was deemed sufficient to merit further investigations of the system for stressed (rotating) environments.
Large stand-alone plates of Yb2Si2O7 and Si were deposited by APS. These plates were precision ground into mechanical test specimens through collaboration with technical staff at the NASA Glenn Research Center. The specimens were used to measure the elastic modulus, fracture toughness, flexure strength at low and high temperatures, and obtain estimates of the materials’ creep resistance. The mechanical properties of the ytterbium disilicate material were significantly lower than values for fully dense materials. Reductions of 2-5x were observed across all quasi-static properties. The creep rates of APS Yb2Si2O7 were found to be very rapid (1 % creep strain accumulated in only 25 h at 900 °C and 16MPa). The activation energy for creep in the 10 % porous APS material was in the 130-150 kJ/mol range. The creep rate of the APS Si samples was also high, but in this case similar to bulk polycrystalline silicon. The mechanical properties of the APS deposited ytterbium disilicate/Si system appear insufficient for future application on rotating components.
The implications of the various failure mechanisms observed and the material properties measured and calculated for these EBC systems are discussed. The materials selection, processing, and performance relationships are interpreted in the context of further EBC development, and several suggestions are proposed for future work to address the technical challenges of this evolving field.
