Damage Evolution in Thermal Barrier Coatings with Thermal Cycling Bauke Heeg Lumium, Leeuwarden, The Netherlands Vladimir K. Tolpygo Honeywell Aerospace, Phoenix, Arizona 85034 David R. Clarke w School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138 Thermal barrier coatings typically fail on cooling after pro- longed thermal cycling or isothermal exposure. The mechanics of spalling requires that first a critical sized portion of the coat- ing separates from the underlying material, then buckles and finally spalls away. The critical size for buckling depends on the thickness of the coating but is several millimeters for typical zirconia coatings 150 lm thick. As-deposited coatings do not have interface separations but they form on thermal cycling as described in this work based on observations of coating cross- sections combined with the stress redistribution in the thermally grown oxide imaged using a piezospectroscopic luminescence method. Analysis of the images reveals that small, isolated regions of damage initially form and then grow, linking up and coalescing to form percolating structures across the coating until the buckling condition is attained, the buckle ex- tends and failure occurs by spallation. The piezospectroscopic imaging of the stresses in the thermally grown oxide formed by oxidation beneath thermal barrier coatings provides a form of ‘‘stress tomography’’ enabling the subcritical separations to be monitored. I. Introduction O NE of the major themes of Tony Evans’ research in the last decade, and one which we were privileged to collaborate with him, was the prime reliance of thermal barrier coating sys- tems. Thermal barrier coatings have been in widespread use in commercial and military gas turbine engines for several decades providing thermal protection to superalloy components in the hottest sections of engines and in the last decade or so have been deemed ‘‘prime reliant.’’ In some advanced engines, the coatings are in contact with gases that exceed the melting temperature of the superalloy blades and vanes. 1 In many respects, thermal barrier coatings must withstand the most demanding conditions that any ceramic component is subject to in today’s technology but being ceramic materials, the coatings are prone to fail. The central scientific and engineering issue is to understand the fail- ure mechanisms under different engine operating conditions and use that information to help identify conditions under which the coatings can be considered prime-reliant. This was the focus of an ONR Multi-University Research Initiative that Tony Evans led. The research described in this manuscript is a part of that quest. Failure of TBCs is not well defined but is usually taken to have occurred when a portion of the coating has buckled and spalled away to expose a portion of the underlying component. Usually, this originates from edges but can also occur, especially on 1 in. diameter test samples, away from an edge. Irrespective of the origin, local buckling occurs before spallation as illus- trated in the optical micrographs in Fig. 1. From images such as these and cross-sections of partially spalled coatings, failure occurs by an accumulation of damage until at some critical condition, yet poorly defined, the coating buckles, the buckle propagates and the coating then spalls away, leaving the under- lying superalloy unprotected. The damage usually forms in the vicinity of the interface region between the thermal barrier coat- ings and the superalloy. Several mechanisms have been identi- fied that can cause the onset of local damage but the mechanics of buckling 2–4 requires that a tensile stress perpendicular to the interface is necessary to form local separations or delaminations. In this contribution, we describe and quantify observations of the subcritical damage evolution obtained using microstructural observations made of polished cross-sections and by a nonde- structive imaging method based on photo-stimulated lumines- cence piezospectroscopy (PSLS). The latter reveals the spatial distribution of the mean stress in the thermally grown oxide that forms between the thermal barrier oxide and the bond-coat. This, in turn, is sensitive to the extent of local damage in and around the TGO, including the TBC itself. II. Materials The 7 wt% yttria-stabilized zirconia (YSZ) coatings investigated in this work were deposited by electron beam evaporation onto flat platinum-modified nickel aluminide coated Rene´ N5 super- alloy coupons by Howmet Research Center. The coatings were all similar to other state-of-the-art coatings made the same way but they were distinguished by all being from the same large batch of coatings deposited on 25 mm diameter, 3 mm thick superalloy disks all wafered from the same single crystal casting. All the coatings were subjected to thermal cycling between room temperature and 11501C with 1 h holds at the high tem- perature as has been described previously. 5 A sufficiently large number of coatings were tested that the average life of the coat- ing batch under these thermal cycling conditions was estab- lished. This was determined to be 185 cycles. Some of the coatings were cycled to intermediate fractions of life rather than to spallation failure. Most of these were sectioned to char- acterize the damage at different fractions of life while four were subject to further analysis by PSLS 6 described in this work. The thermal diffusivities of these same coatings were also evaluated using a novel test based on the time delay in luminescence. 7 A. Heuer—contributing editor Part of the experimental work was conducted while at MetroLaser Inc., Irvine, CA. w Author to whom correspondence should be addressed. e-mail: clarke@seas.harvard. edu Manuscript No. 28809. Received October 21, 2010; approved February 11, 2011. J ournal J. Am. Ceram. Soc., 94 [S1] S112–S119 (2011) DOI: 10.1111/j.1551-2916.2011.04496.x r 2011 The American Ceramic Society S112