`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`P1: GJB
`10.1146/annurev.matsci.33.011403.113718
`
`Annu. Rev. Mater. Res. 2003. 33:383–417
`doi: 10.1146/annurev.matsci.33.011403.113718
`Copyright c(cid:176) 2003 by Annual Reviews. All rights reserved
`First published online as a Review in Advance on April 18, 2003
`
`MATERIALS DESIGN FOR THE NEXT GENERATION
`THERMAL BARRIER COATINGS
`
`D.R. Clarke and C.G. Levi
`Materials Department, College of Engineering, University of California, Santa Barbara,
`California 93106-5050; email: clarke@engineering.ucsb.edu
`
`Key Words
`zirconia, materials science, high temperature
`n Abstract The emphasis in this short review is to describe the materials issues
`involved in the development of present thermal barrier coatings and the advances
`necessary for the next generation, higher temperature capability coatings.
`
`INTRODUCTION
`
`The development of today’s gas turbine engines has been the result of continual
`improvements in a wide variety of engineering skills including turbine design,
`combustion, and materials. One measure of the substantial improvements over the
`past five decades is the increase in the maximum gas temperature at a turbine
`airfoil afforded by these improvements, as shown in Figure 1. The increase in
`airfoil temperature has been facilitated by three principal materials developments:
`dramatic advances in alloy design to produce alloy compositions that are both
`more creep resistant and oxidation resistant; advances in casting technology that
`have facilitated not only the casting of large single-crystal superalloy blades and
`vanes but also the intricate internal channels in the blades to facilitate cooling; and
`the development of a viable coating technology to deposit a conformal, thermally
`insulating coating on turbine components. The advances and developments in the
`first two areas have been reviewed extensively elsewhere (1). Less well known
`is the development of thermal barrier coatings (TBCs), even though in the last
`decade their use has enabled a dramatic increase in airfoil temperature, far greater
`than that enabled by the switch from cast alloy blades to single crystal blades over
`approximately 30 years.
`As originally envisaged, the primary function of a TBC is to provide a low
`thermal conductivity barrier to heat transfer from the hot gas in the engine to the
`surface of the coated alloy component, whether in the combustor or the turbine
`(Figure 2). The TBC allows the turbine designer to increase the gas temperature,
`and thereby the engine efficiency, without increasing the surface temperature of
`the alloy. Subsequently, it has been recognized that a TBC also confers additional
`benefits, for instance, providing protection to rapid thermal transients such as occur
`
`0084-6600/03/0801-0383$14.00
`
`383
`
`GE-1010.001
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`384
`
`CLARKE ¥ LEVI
`
`Figure 1 Increase in turbine airfoil temperature over the last six decades through com-
`binations of materials advances and associated developments in cooling techniques.
`Since this diagram was constructed, the shaded region has extended to the present year,
`and the use of uncooled silicon nitride remains for the future.
`
`due to flame out, and as a means to even out local temperature gradients. Indeed,
`in some cases, the use of a TBC has simplified the design of blades by minimizing
`thermal distortions of the blade. However, undoubtedly the biggest benefit of TBCs
`has been to extend the life of alloy components in the hottest sections in an engine
`by decreasing their surface temperatures.
`Present day TBCs generally consist of a yttria-stabilized zirconia (YSZ) coat-
`ing deposited onto an oxidation-resistant bond-coat alloy that is first applied to a
`nickel-based superalloy component (Figure 2). In diesel engine applications where
`the temperatures are usually lower, the YSZ coating is generally applied directly
`onto the alloy. Two main types of coating are in use. For relatively small compo-
`nents such as blades and vanes in aerospace turbines, the coatings can be applied
`by electron-beam physical vapor deposition (EB-PVD). For larger components
`such as the combustion chambers and the blades and vanes of power generation,
`stationary turbines, the coatings are usually applied by plasma-spraying (PS). In
`many respects, the choice of materials and their production represent a mature
`materials technology. While improvements in their capabilities continue, there is
`a growing realization that new TBC systems will be required for the next genera-
`tion turbines presently being designed. To set the stage for coming developments,
`we first review the selection of materials used in present YSZ coatings, some
`of the new insights that have been gained in understanding how YSZ coatings
`
`GE-1010.002
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`THERMAL BARRIER COATINGS
`
`385
`
`Figure 2
`Schematic illustration of a TBC
`and the associated bond-coat on a superalloy
`in a thermal gradient.
`
`fail, and then describe approaches to the development of the next generation TBC
`systems.
`
`PRINCIPAL REQUIREMENTS OF A THERMAL
`BARRIER COATING
`
`The turbine designers’ primary requirement of a TBC is that it have a low ther-
`mal conductivity and, for rotating components, preferably also a low density to
`minimize centrifugal loads. At the materials design level this translates into three
`additional requirements. First, the material must have strain compliance so as to
`withstand the strains associated with thermal expansion mismatch between the
`coating and the underlying alloy on thermal cycling. The use cycle, both the max-
`imum temperature and the times at temperature, of course, varies between aircraft
`and power generation turbines, but nevertheless the coating must accommodate
`the large strains associated with thermal cycling. The need for strain compliance
`is illustrated in Figure 3, where the thermal expansion coefficients of zirconia,
`
`GE-1010.003
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`386
`
`CLARKE ¥ LEVI
`
`Figure 3 The thermal expansion coefficients and thermal conductivity of
`a range of materials illustrating the differences in thermal expansion and
`conductivity of the principal components in TBC systems.
`
`alumina, and a number of alloys including nickel-based superalloys are cross-
`plotted against thermal conductivity. In the absence of any strain compliance, for
`instance due to a decreased elastic modulus, the large elastic mismatch would gen-
`erate very large stresses and lead to spontaneous failure on cooling. Second, the
`coating material must exhibit thermodynamic compatibility with the oxide, usually
`aluminum oxide, formed on the bond-coat alloy at high temperatures. Third, with
`the continual quest to run engines at higher temperatures and the increasing dif-
`ficulty of increasing the metal temperature, it is increasingly likely that designers
`will seek “prime reliant” coatings, namely ones that can be used with assurance
`that they will not fail. Prime reliant thermal coatings are ones that are necessary
`to prevent the temperature of the metal from exceeding its maximum temperature,
`much in the same way that the tiles on the space shuttle prevent the underlying
`aluminum airframe from being exposed to temperatures in excess of their melting
`temperature on re-entry.
`Because weight is at a premium in aircraft engines, thin coatings with the lowest
`possible thermal conductivity are required. In contrast, in stationary, ground-based
`engines where weight is less of a consideration, a desired temperature drop can be
`
`GE-1010.004
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`THERMAL BARRIER COATINGS
`
`387
`
`achieved through simply increasing the TBC thickness. In practice, in components
`in both types of engine, the thickness of the TBC usually is varied from place to
`place to provide the desired thermal insulation.
`Erosion of the coating by both ingested particles, such as sand, from the op-
`erating environment and particles that come loose from the combustor liners as it
`degrades is a perennial source of concern, especially when the particles are large
`enough to cause impact damage of the coating. In some cases, inborn fine parti-
`cles, primarily dust and sand, melt into the coating as a wetting silicate while it is
`hot and can degrade the coating. These silicates, usually variants of Si-Al-Mg-Ca
`oxides that are the principal elements in sands, are often referred to collectively as
`CMAS.
`A recently recognized requirement of many materials exposed to high tempera-
`tures in gas turbines is a long-term stability in the presence of steam. This is partly
`a direct result of the generation of water during the combustion process, but in a
`number of designs it is a consequence of the use of steam injection to enhance
`turbine efficiency. Little is known about the effects of long-term exposure to steam
`on turbine materials. However, tests have revealed that many silicon-based com-
`pounds, including SiC, are unstable to the formation of volatile SiO, which results
`in the slow retraction of the material as evidenced by the reduction in thickness
`of components over long operating periods. This active oxidation and evaporation
`phenomenon precludes the use of silicon compounds in coatings unless protected
`by another coating.
`More difficult to design against are the effects of corrosion, especially airborne
`species and those, such as sulfur and vanadium, in the fuel itself. The majority of
`land-based turbines operate on natural gas, but there is increasing interest in using
`alternative fuels, such as coal gas, that are much dirtier. The consequences of using
`such alternative fuels and their effects on coatings are only now beginning to be
`investigated.
`
`THE THERMAL BARRIER COATING SYSTEM
`
`From a materials engineering perspective, it is necessary to consider the TBC as
`an integrated materials system rather than simply a thermally insulating material
`coating on a structural alloy component. A representative cross-section of a com-
`mercial coating, shown in Figure 4, illustrates the multilayered nature of a coating
`after high-temperature exposure. There are three principal layers in addition to the
`superalloy and the low-conductivity coating. Between the alloy and the coating is
`the bond-coat, so called because in the initial development stages in producing a
`viable coating, it was found that the superalloy had to be first covered with a bond-
`coat to ensure that the YSZ coating remained adherent upon oxidation. Between
`the bond-coat and the YSZ coating—sometimes referred to as the overcoat—is the
`oxide formed during high-temperature exposure. Finally, during the formation of
`the bond-coat and the YSZ coating as well as subsequently during use, a reaction
`layer forms as a result of inter-diffusion between the bond-coat and the superalloy.
`
`GE-1010.005
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`388
`
`CLARKE ¥ LEVI
`
`Figure 4 Cross-section of a TBC deposited by electron beam evaporation. Note the
`columnar microstructure of the zirconia coating. The white band is a reaction layer
`formed by interdiffusion during use between the Al-rich bond-coat above and the Ni-
`rich superalloy, below. The thermally grown oxide (TGO) is too thin to be discernible
`in this micrograph.
`
`In practice, as mentioned above, there are two distinct types of zirconia coat-
`ings reflecting different approaches to creating the strain compliance essential to
`withstand thermal cycling. The two types of coatings are EB-PVD coatings and
`plasma-sprayed coatings. In EB-PVD coatings, the lateral strain compliance re-
`sults from the columnar structure and inter-columnar gaps produced by rotation of
`the component during deposition. The columnar structure can be seen in the mi-
`crograph of Figure 4. Transmission electron microscopy reveals that the individual
`columns also contain microscopic porosity that reduces the thermal conductivity
`of the coating. In plasma-sprayed coatings, the lateral strain compliance and re-
`duced thermal conductivity is conferred by the incorporation of porosity between
`“splats” of successively deposited material. This porosity is illustrated in Figure 5,
`where the splats in a plasma spray bond coat have preferentially oxidized and
`consequentially appear as dark veins.
`Two major classes of bond-coat alloys have also evolved over the years, but
`both were developed to form an aluminum oxide (fi-Al2O3) on exposure to air
`at high temperatures. This is important for several reasons. One is that Al2O3
`is phase compatible with YSZ, ensuring long-term thermodynamic stability of
`the coating. Uncoated, the majority of nickel-based superalloys form complex,
`
`GE-1010.006
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`THERMAL BARRIER COATINGS
`
`389
`
`Figure 5 Cross-section of a TBC deposited by plasma-spraying. The plate-
`like porosity is evident in the coating as the dark veins in the center of the
`micrograph.
`
`multilayered nickel oxide, nickel-chromium spinels and chromium oxide, in addi-
`tion to alumina, and these are not thermodynamically stable with YSZ (2). Further-
`more, alumina is usually considered to be the slowest growing high-temperature ox-
`ide on account of it having the smallest oxygen diffusivity (3). The rationale for the
`selection of bond-coat alloys is really a subject of another review, but the bond-coat
`has to perform a number of disparate functions. It must provide a bond between the
`deposited TBC and the underlying alloy. In the early days of TBC development, the
`bonding to the alloy was a major concern, particularly plasma-sprayed, hence
`the term bond-coat. Because zirconia is a fast-ion oxygen conductor, the bond-coat
`must also be able to form a protective, stable, and slow-growing oxide to prevent
`oxidative attack of the alloy. As is described below, one of the principal forms of
`failure is associated with failure of the protective aluminium oxide. The bond-coat
`must also have sufficient morphological stability so that on heating and cooling,
`as well as at high temperature, it does not distort and introduce incompatibilities
`that can also cause the introduction of interface defects.
`The two classes of bond-coat alloys that have been developed are the platinum-
`modified nickel aluminide (PtNiAl) and MCrAlY alloys (M here refers to one or
`more of the elements Co, Ni, and Fe). The selection of these two classes of alloys
`is largely based on their prior use as oxidation- and corrosion-resistant coatings for
`protecting high-temperature alloys before the advent of TBCs. For instance, the
`
`GE-1010.007
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`390
`
`CLARKE ¥ LEVI
`
`PtNiAl was originally developed as an alternative oxidation-resistant coating for
`protecting alloys at higher-temperature operation than the MCrAlY alloys available
`at the time.
`Different methods of applying the bond-coat alloys have been developed largely
`to meet production goals. Typically, PtNiAl bond-coats are formed by first electro-
`depositing Pt onto the superalloy component and then annealing it in an aluminum-
`rich vapor atmosphere. In this second step, aluminum diffuses into the surface of
`the alloy while nickel diffuses out where it reacts with the aluminum and platinum
`to form the PtNiAl aluminide coating. Depending on the quality of the coating
`required, the aluminum is provided in a pack-process or in a CVD reactor from an
`AlCl3 source. In contrast, the MCrAlY coatings are commonly deposited by one of
`a number of variants of plasma-spraying. These processes are particularly attractive
`for coating large components and are, of course, cheaper than EB deposition. Also,
`as plasma-spraying does not involve a diffusion process, thicker bond-coats can be
`deposited than with the aluminizing process used to form the PtNiAl bond coats.
`It remains uncertain at this time which of these coating types is best for different
`applications. In large part this is because it is not yet known which combination
`of materials properties leads to the longest, high-temperature life of the coating.
`To provide the largest reservoir of aluminum one would expect that the thicker the
`bond-coat and the higher its aluminum content the better. However, one would also
`expect that the bond-coat should have as large a yield stress as possible at high tem-
`perature with as closely matched thermal expansion mismatch with the superalloy
`as possible to avoid thermal expansion mismatch stresses on thermal cycling.
`
`FAILURE MECHANISMS
`
`Investigation of the ways in which present YSZ coatings fail has provided consid-
`erable insight into the underlying mechanisms that limit coating life. Although, as
`with failure analysis in other areas of complex material systems, there are many
`complications, the findings nevertheless point toward methods of producing coat-
`ings that can withstand longer lives at temperature and higher use temperatures.
`One of the chronic problems is that the life of present TBC coatings invariably
`shows a wide distribution, with a high proportion of the population clustered about
`a median value but with a significant proportion failing at much earlier times.
`There is substantial circumstantial evidence to suggest that many of the TBC
`failures are associated with the oxidation of the bond-coat (4). Indeed, a number
`of manufacturers are believed to use an oxidation criterion as a basis for predicting
`average life. One such criterion is the combination of time and temperature to lead
`to a critical thickness of the TGO1. Another, embodied in the Coatlife software,
`is an aluminum depletion criterion based on the combined time and temperature
`
`1The concept of a critical thickness, of the order of 6 „m at 1100
`C, appears to have
`originated from observations of coatings that failed under isothermal testing conditions.
`
`–
`
`GE-1010.008
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`THERMAL BARRIER COATINGS
`
`391
`
`for the concentration of aluminum at the bond-coat surface to fall below a critical
`value. In the case of MCrAlY bond-coats, the rationale for this is that when the
`Al concentration falls to »8 a/o, aluminum oxide is no longer the thermodynamic
`preferred phase and other oxides, notably spinels, form (5). These other oxides
`do not form such a protective scale, and consequently the alloy oxidizes faster. In
`addition, the formation of these oxides is associated with an increase in volume
`that can be disruptive and possibly have lower fracture energies, although this has
`yet to be unequivocally demonstrated. Nevertheless, there are reports that when
`the bond-coat is porous, and at low-oxidation temperatures, failure follows such
`aluminum depletion (6).
`Although related to the oxidation behavior of the bond-coat, neither the con-
`cept of a critical thickness nor aluminum depletion can account for the wide dis-
`tribution in failure lives, especially under thermal cycling conditions. Indeed, in
`–
`C, the alu-
`the majority of materials examined after failures above about 1000
`minum concentration, although depleted somewhat, has not fallen to the critical
`value (7). Similarly, the short-lived coatings have failed before the TGO thickness
`has reached the thickness of its counterparts that have shown the longest lives.
`Together these findings indicate that failure occurs due to extrinsic factors arising
`during oxidation.
`The prevailing mode of failure is one in which part of the coating buckles and
`spalls away from the alloy, typically on cooling down to room temperature (8, 9).
`A typical buckling failure, in this case nucleated from the edge of a test coupon, is
`illustrated in Figure 6. Such buckling and subsequent spallation is a common mode
`of failure of all films and coatings under compression, generally associated with
`the development of compressive residual stresses in the coatings as a result of the
`difference in thermal expansion coefficient between the coating and the underlying
`alloy. The mechanics of the failure by buckling of a thin, elastically isotropic
`film under compression from a flat surface is well understood (10), provided an
`unbonded region of a critical size, db, exists at the interface (Figure 7). For a fixed
`film thickness and residual stress, the stress at which buckling will occur is given
`by the relation:
`
`¶2
`
`:
`
`1.
`
`h d
`
`b
`
`(cid:181)
`
`(cid:190)=E D 4:8
`
`Thin film buckling is entirely analogous to the standard Euler buckling condition
`of a column—a bifurcation phenomenon. The striking feature of this relation is that
`the flaw size depends linearly on the thickness of the film.2 Since even the thinnest
`of TBC is over 100 „m thick, the critical size to which an interface flaw must grow
`before buckling can occur can be several millimeters. As interface separations of
`this large size are not usually present after coating, one of the major unresolved
`
`2Because they are designed to have in-plane strain compliance, TBCs are not usually
`isotropic elastic solids. The buckling condition is modified from that in Equation 1 to
`account for the elastic anisotropy.
`
`GE-1010.009
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`392
`
`CLARKE ¥ LEVI
`
`Figure 6 (Top) Incipient buckling of a TBC coating viewed under reflected light.
`(Bottom) The failure surface revealed by spallation of the TBC consists of a mixture
`of local failure between the TGO and the bond-coat (appearing dark) and in the TBC
`itself (light regions).
`
`questions is how interface separations first form and then grow to such a large
`size. Such progressive failure consisting of nucleation of local interface separation
`and their subsequent growth has indeed been observed (11). Recent mechanics
`calculations have shown that interface perturbations from flatness can decrease
`the critical size at which buckles can initiate and then grow in size to form a spall
`
`GE-1010.010
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`THERMAL BARRIER COATINGS
`
`393
`
`Figure 7 Schematic illustration of the buckling of a compressed film above
`a pre-existing defect of diameter db.
`
`(12). Nevertheless, localized flaws must first initiate and then grow for spallation
`failure to occur. Understanding the nucleation of these flaws, their growth, and
`progressive linking together is essential before realistic models for predicting life
`can be developed.
`Insight into the formation of flaws comes from microstructural examination of
`coating cross-sections after high-temperature exposure but prior to spallation. Four
`examples are shown in Figure 8, each from a YSZ TBC-coated PtNiAl bond-coat
`(13). In each case, the coating was deposited conformally onto the surface of a
`flat bond-coat so that the coating/bond-coat interface was initially flat and intact.
`As three of the micrographs illustrate, the surface of the bond-coat roughens
`and separations form with the TBC even though the bottom surface of the TBC
`remains flat. These separations are the interface flaws that progressively grow in
`size and link together with adjacent ones to allow buckling and spallation (9, 11).
`Roughening is more pronounced with thermal cycling but also occurs, albeit more
`slowly, on isothermal exposures (14). The growth of these separations with thermal
`cycling can now be monitored by luminescence piezospectroscopy, as described
`
`GE-1010.011
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`394
`
`GE-1010.012
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`THERMAL BARRIER COATINGS
`
`395
`
`in the next section, which suggests that it has the potential to be used as a viable
`NDE tool (11, 15).
`The micrographs in Figure 8 raise the question as to the underlying mechanisms
`responsible for the observed roughening. At least two new mechanisms have now
`been identified that can lead to such roughening. The roughening has been at-
`tributed to a “ratcheting” phenomenon motivated by the lateral compressive stress
`in the growing TGO and facilitated by thermal cycling (9). Measurements indicate
`that as the TGO grows in thickness with oxidation, it also concurrently develops
`a compressive stress (16, 17). If it were free to expand it would decrease its com-
`pressive stress but because it is attached to the bond-coat, the only way in which
`it can decrease its elastic strain energy is by undulating (Figure 9). In this way, its
`length increases and it remains attached to the alloy. This undulation requires the
`alloy to deform to accommodate the undulation, and the oxide must also deform
`concurrently. According to the ratcheting mechanism, this accommodation is by
`plastic deformation of both the TGO and bond-coat during thermal cycling. As
`the lateral growth of the thickening oxide continues during the high-temperature
`portion of the thermal cycles, it continues to generate compressive stress that is
`relaxed by ratcheting during the thermal cycle so the process is ongoing. Many of
`the essential features of the mechanism have been substantiated by finite element
`computations (18) and are consistent with observations of the increase in length
`of the TGO as the surfaces roughen.
`Another new mechanism shown to cause roughening is the surface displace-
`ment associated with volumetric changes in the bond-coat as aluminum depletion
`occurs. This roughening is illustrated in Figure 10, together with etched cross-
`sections revealing the presence of both (cid:176) 0
`and fl phases in the bond-coat (14).
`After aluminizing and after YSZ deposition, the PtNiAl bond-coat is chemically
`homogeneous and has the fl-NiAl (B2) crystal structure. After high-temperature
`exposure, the initially flat bond-coat is rumpled and etching reveals that the bond-
`coat has partially transformed to (cid:176) 0
`-Ni3Al. In addition, the remaining fl-NiAl phase
`regions often have the characteristic lath structure of a martensite. These two ob-
`servations can be understood as being the result of aluminum depletion from the
`bond-coat and concurrent enrichment of nickel from the underlying superalloy, a
`classic example of interdiffusion. This change in composition is illustrated using
`the binary NiAl diagram in Figure 11. As aluminum is depleted, the average com-
`position of the bond-coat becomes increasingly enriched in nickel until reaching
`
`ˆ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡
`Figure 8 Cross-section of four TBCs illustrating different forms of the local sep-
`aration between the TBC and the bond-coat after thermal cycling. In each case, the
`coating was deposited conformally on the bond-coat so interface separations such as
`these indicate that the underlying bond-coat has changed its surface morphology dur-
`ing high-temperature exposure and thermal cycling. In example (b), no separation has
`occurred and it exhibits the longest life.
`
`GE-1010.013
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`396
`
`CLARKE ¥ LEVI
`
`Figure 9 Schematic illustration of how an initially flat but compressed film (left)
`can lower its elastic strain energy by rumpling (right). The amplitude of rumpling
`is enhanced by thermal cycling and can cause interface separation if a superimposed
`coating cannot deform to follow the displacements of the film.
`
`Figure 10 Microstructure of an initially flat as-aluminized bond coat after 50 1-h
`–
`cycles at 1200
`C: (a) surface rumpling; (b) cross-section showing a rather uniform
`oxide layer and strong surface undulations ((cid:176) 0
`-phase is revealed by etching); (c, d )
`optical micrographs showing etched cross-section before and after cyclic oxidation.
`Dark areas on the optical images correspond to the fl-phase, whereas the (cid:176) 0
`-phase in
`the coating appears white.
`
`GE-1010.014
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`THERMAL BARRIER COATINGS
`
`397
`
`Figure 11 The Ni-Al pseudo-binary phase diagram illustrating the compositional
`range of the fl-NiAl phase and the direction of the change in composition as the bond-
`coat is depleted of Al by interdiffusion and selective oxidation.
`
`the single-phase boundary at which point further depletion leads to the formation of
`(cid:176) 0
`-Ni3Al. (At even later times, the composition can extend into the (cid:176) region of the
`phase diagram.) Martensitic structures within the fl-NiAl phase field also form as
`the phase boundary is approached. Whereas the martensite start temperature, MS,
`of the pure fl-NiAl compositions is known to be generally around room tempera-
`–
`C (20), the additional Pt, Co, and Cr present in the PtNiAl bond-coat
`ture to 300
`increase the MS temperature, and Hemker et al. have reported MS temperatures of
`»600
`–
`C (21).
`Substantial progress has been made in the past few years in understanding some
`of the mechanisms that lead to flaw initiation and growth during use (9, 14, 18, 22).
`These provide the basis for developing life-prediction models, but nevertheless a
`number of unanswered questions remain. For instance, the micrographs in Figure 8
`were obtained from nominally the same superalloy, with the same bond-coat and
`the same YSZ coating all made by the same manufacturer in the same process
`manner. This difference in interface separation and roughening is particularly
`marked in this figure, but it does suggest that even small, but as yet unidentified,
`concentrations of dopants can have a large effect on life (23).
`Insights gained in the past few years into some of the important processes
`occurring within the coating during use are summarized in Figure 12. Essentially,
`
`GE-1010.015
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`398
`
`CLARKE ¥ LEVI
`
`Figure 12 Schematic summary of the concurrent processes occurring in the bond-
`coat, TGO and TBC, during use at high temperatures. The complexity in failure times
`and failure modes is believed to reflect the competition between hese individual pro-
`cesses.
`
`the TBC system is one that evolves with time at temperature and its evolution
`depends in detail on not only the temperature but also on the thermal cycle history
`and heating and cooling rates, as well as composition.
`
`NON-DESTRUCTIVE EVALUATION
`
`As the design of coatings shifts to a philosophy of prime reliance, the ability to
`non-destructively monitor the coating, identify defects, and evaluate its remaining
`life becomes more important. In addition, there is a growing economic pressure to
`defer maintenance and replacement of parts until really necessary. (The costs are
`staggering: The cost, in replacement electricity alone, of taking a power generation
`turbine out of operation can be of the order of $1 million a day. The cost of replacing
`a single, first-stage turbine blade can also be very high. Depending on its size the
`cost can be as high as $10,000.)
`In the case of large area separations, several millimeters to centimeters, infrared
`imaging provides a direct means of visualizing incipient coating failure provided
`the blade can be accessed. Usually, though, a coating has failed by the time the
`separation reaches such a large size, and thus methods of identifying damage at
`an earlier time, and hence smaller size, are required. No single solution appears
`practical at this stage, although laser-induced acoustic sounding (from Lasson
`
`GE-1010.016
`
`
`
`4 Jun 2003 12:43
`
`AR
`
`AR189-MR33-16.tex AR189-MR33-16.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: GJB
`
`THERMAL BARRIER COATINGS
`
`399
`
`Technologies, personal communication) and higher-spatial resolution imaging us-
`ing polarized scattered light have shown promise in detecting deliberately created
`interface flaws (25). An alternative method is one that utilizes piezospectroscopy,
`the strain-induced shift of luminescence and Raman lines. When illuminated with
`a laser having an appropriate wavelength, luminescence from the aluminum oxide
`TGO formed by oxidation on the bond-coat can be detected through the thickness
`of the TBC (26, 27). The luminescence spectrum, from Cr3C
`ions incorporated into
`the alumina TGO as it grows, is sensitive to the stress state in the TGO. (Because
`zirconia is a wide band-gap material, it is transparent in the visible so illumina-
`tion in the blue or green, e.g., as an argon ion laser, can penetrate to the TGO,
`and the stimulated luminescence, which is in the red, is transmitted back through
`the coating.) The key to the use of photoluminescence as an NDE tool is that the
`frequency of the luminescence lines sh