The evolution of residual stress in the thermally grown oxide on Pt diffusion bond coats in TBCs

Abstract

The evolution of residual stress in the thermally grown oxide (TGO), formed on bond coats (BC) produced by diffusion of Pt into a Ni-based superalloy, has been studied as a function of thermal cycling using photo-luminescence piezo-spectroscopy through the yttria-stabilised zirconia (YSZ) thermal barrier coating. The luminescence spectra were analysed in terms of a high stress and a low stress contribution to the residual stress sampled at a single analysis point (approx. 20 μm in size); the low stress contribution being caused by either local TGO fractures or interfacial delamination The mean compressive residual stress increases during the first two thirds of the lifetime due to gradual stiffening of the non-planar TGO as it thickens and becomes more resistant to bending. In the last third of life the mean stress tends to decrease due to relaxation by increasingly numerous local damage events. This is more pronounced as a dramatic increase in the parameters that quantify the extent of the low stress contributions. Spatial mapping of the low stress contributions shows that they are isolated at first, but become more numerous towards the end of life and coalesce into larger regions of damage a few hundred μm in size. Eventual spallation is by unstable delamination at the BC/TGO interface and the stored elastic energy released is between 40 and 80 J m⁻². The lifetime is governed by the growth of the local damage regions until one (probably near an edge) reaches a critical size above which it propagates in a highly unstable way over most of the interface.

Introduction

Thermal barrier coatings (TBCs) are widely used in gas-turbine engines in order to protect metal components such as turbine blades from hot combustion gases. The increased turbine inlet gas temperature that TBCs permit results in higher engine efficiency and fuel economy. The TBC systems applied on turbine blades typically comprise an oxidation-resistant (alumina-forming) metallic bond coat (BC) and a thermally insulating top coat, which is usually zirconia partially stabilised by 8 wt% yttria (YSZ) deposited by electron-beam evaporation (EBPVD). The BC and YSZ layers are typically about 50–200 μm in thickness. During service the BC is slowly oxidised to form a protective thermally grown oxide (TGO) alumina layer at the interface between the BC and the YSZ. It is important to understand the processes that lead to failure of the TBC system in service in order to be able to make reliable lifetime predictions and improve lifetimes. Many different processes can contribute to failure of the TBC by spallation and the dominant one will depend on the details of the TBC system and the thermal cycling regime[1]. Nevertheless, it is generally accepted that stresses generated by thermal expansion mismatch between the different constituents of the coating play a major role in driving the failure. It is also usually assumed that the stresses at high temperature are relatively low because they can be relaxed by creep and that the largest stresses occur in the TGO on cooling to room temperature.

Christensen et al. [2] showed that the residual stress in the TGO could be measured in-situ beneath the YSZ using photo-luminescence piezo-spectroscopy from Cr³⁺ impurities in solid solution in the TGO. The stress-induced frequency shift from the TGO was found to be consistent with the large in-plane biaxial compressive stress expected from thermal expansion mismatch between alumina and the superalloy substrate. This technique has excited much interest since it also offers a potential non-destructive method for monitoring the stresses in engine components and assessing their residual life.

In previous work [3] we applied this technique to measure the residual stress in the TGO formed after isothermal oxidation of a TBC system comprising approximately 200 μm 8 wt% yttria-stablised zirconia (YSZ) grown by electron beam physical vapour deposition (EBPVD) onto a Pt diffusion bond coat (BC) [4]. Those experiments showed that in this system the apparent residual stress in the TGO on specimens without the YSZ top coat varied considerably from point to point; and the mean stress was much lower than the expected thermo-elastic plane stress. This was partly explained by surface roughness and partly by local fracture of the TGO. When the YSZ was present the mean stress was approximately equal to the estimated thermo-elastic plane stress, but the luminescence line-width was too large to be explained by interface roughness. It was subsequently realised that this was compounded by an over-simplistic analysis of the luminescence spectra. A constrained fitting procedure was developed [5] for deconvoluting the spectra based on the observation that there were two contributions to the stress distribution; one corresponding to a high stress in the range 2.0–4.5 GPa, and the other to a low stress ranging from 1 GPa compressive to 1.5 GPa tensile. On specimens without YSZ top coats the low stress regions could be identified with fractures in the TGO and thus the low stress component is potentially a means of quantifying local damage in the TGO. The quantification was demonstrated by mapping the spectra on rectangular grids of up to 2601 analysis points. Nychka and Clarke [6] proposed a similar constrained deconvolution approach.

The aim of the work described in this paper was to apply this approach to study how the residual stress in the TGO grown on the Pt diffusion BC under EPVD YSZ evolves with thermal cycling, particularly as failure by spalling is approached, and to interpret this in terms of a failure mechanism.