Availableonlineatwww.sciencedirect.com Science Direct E噩≈RS ELSEVIER Joumal of the European Ceramic Society 28(2008)1405-1419 www.elsevier.comlocate/jeurceramsoc The influence of oxides on the performance of advanced gas turbines A G. Evans D R. Clarke. C G. Lev Materials Department, University of California, Santa Barbara, Santa Barbara, CA 93106-5050, USA Available online 28 January 2008 Zirconia and alumina have been successfully incorporated into turbines used for propulsion and power generation. They exert a crucial influence on the fuel efficiency. The roles of these oxides within the overall system are described, relative to those for the other constituents, and their most important properties are outlined. The mechanisms that govern their properties are presented ar hes for adjusting them in desirable directions are discussed. Opportunities for new materials with potential for superior performance are 2007 Elsevier Ltd. All rights reserved. Keywords: Thermal barrier coatings; Alumina; Zirconia; Interfaces: Thermal properties 1. The motivation excess of 1200C)within an oxidizing environment. A single material would be incapable of satisfying these requir emen Oxides are present in turbines used for propulsion and power The viable solution is an oxide/metal multilayer(Fig. 2). The generation. Their benefits are manifest in a substantial outer oxide imparts thermal protection: while the metallic layer in the longevity of various hot section components (bond coat) affords oxidation protection through the formation technology demonstrates how oxides can be used to of a second oxide, as well as plastic accommodation of strain. -8 structural members that experience environmental extremes At the technology inception, the preferred insulating oxide Documenting the principles that underlie this success facilitates was determined to be yttria-stabilized zirconia (YSZ), chosen dissemination to other systems. The technology involves choices because of its low, temperature-invariant, thermal conductivity of materials and spatial configurations, as well as survivability(Fig 3a). The most desirable phase was ascertained by conduct upon extreme temperature cycling without loss of functionality. ing laboratory-based thermal cycle tests to seek the composition durability: that is, the lar ber of l.I. Materials and configurations cycles before the coating spalls(Fig. 3b). The outcome w 7wt. o yttria-stabilized zirconia(7-YSZ). This composition The following considerations have motivated the choice of still used. &e 1-20 It remains the material of choice because e the discovery of lower thermal conductiv materials and their spatial configurations. The thermal require- Ity opt ig.1).By directing air through other properties(especially toughness2-2)are also crucial channels, the structural alloy is internally cooled: with heat On rotating components, the layer thickness is important. It i transfer coefficient determined by the flow rate and the chan a compromise between having sufficient thickness to achieve nel geometry. Subject to a combustion temperature, Tgas, and the desired temperature drop, yet thin enough to avert exces- an external heat transfer coefficient, superposing an external sive inertial loads, due to the extra mass. The outcome is insulting oxide allows Tgas to be raised while retaining the thickness in the range 100= Htbc 3250 umOn stationary com- alloy at an allowable maximum temperature. Remarkably, insu- ponents, such as shrouds and combustors, the mass is less critical lating oxides deposited onto geometrically complex structural and much thicker layers can be used. The choice is typically, components, such as airfoils, remain attached for extended peri- 500um sTbc I mm. ods despite cycling through an enormous temperature range (in gov materials for oxidation protec tion are straightforwar with nuanced implementation (i)A thermally grown GO) forms at the bond coat sur- orresponding author. face by reaction with the combustion gas. The preferred TGO E-mail address: agevans @engineering. ucsb.edu(A G. Evans) should have the lowest possible oxygen ingress at the temper 0955-2219/S-see front matter o 2007 Elsevier Ltd. All rights reserved. doi: 10.1016/j-jeurceramsoc 2007 12.023
Available online at www.sciencedirect.com Journal of the European Ceramic Society 28 (2008) 1405–1419 The influence of oxides on the performance of advanced gas turbines A.G. Evans ∗, D.R. Clarke, C.G. Levi Materials Department, University of California, Santa Barbara, Santa Barbara, CA 93106-5050, USA Available online 28 January 2008 Abstract Zirconia and alumina have been successfully incorporated into turbines used for propulsion and power generation. They exert a crucial influence on the fuel efficiency. The roles of these oxides within the overall system are described, relative to those for the other constituents, and their most important properties are outlined. The mechanisms that govern their properties are presented and approaches for adjusting them in desirable directions are discussed. Opportunities for new materials with potential for superior performance are assessed. © 2007 Elsevier Ltd. All rights reserved. Keywords: Thermal barrier coatings; Alumina; Zirconia; Interfaces; Thermal properties 1. The motivation Oxides are present in turbines used for propulsion and power generation. Their benefits are manifest in a substantial increase in the longevity of various hot section components.1–8 The technology demonstrates how oxides can be used to protect structural members that experience environmental extremes. Documenting the principles that underlie this success facilitates dissemination to other systems. The technology involves choices of materials and spatial configurations, as well as survivability upon extreme temperature cycling without loss of functionality. 1.1. Materials and configurations The following considerations have motivated the choice of materials and their spatial configurations. The thermal requirements are straightforward (Fig. 1). By directing air through channels, the structural alloy is internally cooled: with heat transfer coefficient determined by the flow rate and the channel geometry. Subject to a combustion temperature, Tgas, and an external heat transfer coefficient, superposing an external insulting oxide allows Tgas to be raised while retaining the alloy at an allowable maximum temperature. Remarkably, insulating oxides deposited onto geometrically complex structural components, such as airfoils, remain attached for extended periods despite cycling through an enormous temperature range (in ∗ Corresponding author. E-mail address: agevans@engineering.ucsb.edu (A.G. Evans). excess of 1200 ◦C) within an oxidizing environment. A single material would be incapable of satisfying these requirements. The viable solution is an oxide/metal multilayer (Fig. 2). The outer oxide imparts thermal protection: while the metallic layer (bond coat) affords oxidation protection through the formation of a second oxide, as well as plastic accommodation of strain.1–8 At the technology inception, the preferred insulating oxide was determined to be yttria-stabilized zirconia (YSZ), chosen because of its low, temperature-invariant, thermal conductivity9 (Fig. 3a). The most desirable phase was ascertained by conducting laboratory-based thermal cycle tests to seek the composition affording greatest durability: that is, the largest number of cycles before the coating spalls (Fig. 3b).16 The outcome was 7 wt.% yttria-stabilized zirconia (7-YSZ). This composition is still used, despite the discovery of lower thermal conductivity options.4,17–20 It remains the material of choice because other properties (especially toughness21–25) are also crucial. On rotating components, the layer thickness is important. It is a compromise between having sufficient thickness to achieve the desired temperature drop, yet thin enough to avert excessive inertial loads, due to the extra mass. The outcome is thickness in the range 100 ≤ Htbc ≤ 250m. On stationary components, such as shrouds and combustors, the mass is less critical and much thicker layers can be used. The choice is typically, 500m ≤ Htbc ≤ 1 mm. The principles governing the materials for oxidation protection are straightforward: albeit with nuanced implementation. (i) A thermally grown oxide (TGO) forms at the bond coat surface by reaction with the combustion gas. The preferred TGO should have the lowest possible oxygen ingress at the temper- 0955-2219/$ – see front matter © 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.jeurceramsoc.2007.12.023
G. Evans et al. Joumal of the European Ceramic Sociery 28(2008)1405-1419 misfit differs for each of the layers. It is least important for the external oxide because this layer need not be dense: it serves only to insulate the underlying alloy and does not provide oxi- dation protection. It is designed with a microstructure having spatially configured porosity that affords low in-plane stiffness and strain tolerance 40-44 This strategy cannot be used for either the tGo or the bond coat because to serve their functions both need to be dense(minimal porosity). The TGO misfit can- can be managed by limiting its thickness. The misfits between the bond coat and substrate are more nuanced: they occur not only from thermal expansion, but also phase transformations and swelling. 45 Understanding these misfits, ascertaining their importance to system durability, and finding means to control them, has been an important research focus from engines(Fig 4). Small diameter spalls can be tolerated, because backside cooling and boundary layer effects still allow the exposed surface to be protected by the(surrounding) intact oxide. Degradation only becomes a concern after an appreciable Fig. 1. A schematic of an airfoil and a magnified view of a surface zone with area fraction of the coating has been removed. Actual spall for- the TBC and bond coat layers identified. The thermal conditions are defined. mation is preceded by smaller cracks that extend and coalesce atures of interest(900-1150oC), with correspondingly small or at the interface between the TGO and the bond coz de layer along delamination planes located either within the oxid The ensuing article highlights the roles of the oxide con- counter-diffusion of the metallic elements. (ii) The bond coat stituents. It is organized as follows. The constituent materials should have sufficient thermo-chemical compatibility with the and their salient thermo-mechanical prope rues are o structural alloy that the basic composition, microstructure and The spectrum of mechanisms governing the performance and usability of hot section components are described, thereby singular solution is an alloy thatforms a-Al2O3 upon oxidation. illuminating the oxide functionalities. With reference to these To achieve this, near its surface, the alloy must contain sufi- mechanisms. the dominant characteristics of the oxides are cient Al that the primary oxidation product is, indeed, a-Al2O discussed, with associated mechanistic understanding. In turn and, moreover, acts as a reservoir for re-formation of a-Al203 these mechanisms reveal opportunities for new research on should spallation occur. The common choices are alloys based oxides that might further enhance the fuel efficiency on Ni(Al) with various additions(such as Cr, Co, Pt, Y and hf Other requirements are more nuanced. They dictate competi 2. The constituents and their thermo-mechanical tive advantage, through key aspects of system performance and properties durability. In practice, three categories of bond coat have been implemented, differentiated by the phases present and the alloy additions.(a)One category consists of a single B-phase usually The requirements imposed on each layer( Fig. 2)dictate th constituent property attributes. In current implementations, the made by inter-diffusing Al and Pt with Ni adjacent to the surface structure and composition of the substrate and the insulating of the superalloy. 4.(b)A second consists of a two-phase y/p- oxide are largely fixed. Options exist for the bond coat, which EB-PVD 29-31(c)The third is a two-phase yhy alloy made by affect the formation of the ensuing TGO infusing Pt(and Hf) into the substrate 32.33 Systems made using 2.1. Insulating oxide ese bond coats perform differently with durability governed The thermal expansion coefficient of this layer, atbc, is appreciably lower than that for the substrate, asub: the dif- 1. 2. Performance and durability ference is about.athe-asub≡△abc≈-3ppm/K. To prevent spontaneous delamination due to this misfit, the in-plane mod- o survive extreme thermal cycling the misfit st rains between ulus of the layer, Etbe, must be controlled, as illustrated by arise due to differences in thermal expansion coefficient, as well creep in the Ysz causes it to become stress-free. Subsequer as phase transformations and inter-diffusion. They cause resid- cooling induces residual stress through the thermal expan ual stresses upon temperature cycling, which activate inelastic sion misfit. If the Ysz were fully dense(Etbe=200 GPa, mechanisms that, in turn, limit durability. The importance of the Tbc N0. 2), for a typical value of the average temperature
1406 A.G. Evans et al. / Journal of the European Ceramic Society 28 (2008) 1405–1419 Fig. 1. A schematic of an airfoil and a magnified view of a surface zone with the TBC and bond coat layers identified. The thermal conditions are defined. atures of interest (900–1150 ◦C), with correspondingly small counter-diffusion of the metallic elements. (ii) The bond coat should have sufficient thermo-chemical compatibility with the structural alloy that the basic composition, microstructure and properties are retained for the expected life of the system. The singular solution is an alloy that forms -Al2O3 upon oxidation. To achieve this, near its surface, the alloy must contain suffi- cient Al that the primary oxidation product is, indeed, -Al2O3 and, moreover, acts as a reservoir for re-formation of -Al2O3 should spallation occur. The common choices are alloys based on Ni(Al) with various additions (such as Cr, Co, Pt, Y and Hf). Other requirements are more nuanced. They dictate competitive advantage, through key aspects of system performance and durability. In practice, three categories of bond coat have been implemented, differentiated by the phases present and the alloy additions. (a) One category consists of a single -phase usually made by inter-diffusing Al and Pt with Ni adjacent to the surface of the superalloy.26–28 (b) A second consists of a two-phase, /- alloy, usually deposited onto the substrate by plasma spraying or EB-PVD.29–31 (c) The third is a two-phase / alloy made by infusing Pt (and Hf) into the substrate.32,33 Systems made using these bond coats perform differently with durability governed by different mechanisms. 1.2. Performance and durability To survive extreme thermal cycling the misfit strains between the layers must be understood and managed.34–39 These strains arise due to differences in thermal expansion coefficient, as well as phase transformations and inter-diffusion. They cause residual stresses upon temperature cycling, which activate inelastic mechanisms that, in turn, limit durability. The importance of the misfit differs for each of the layers. It is least important for the external oxide because this layer need not be dense: it serves only to insulate the underlying alloy and does not provide oxidation protection. It is designed with a microstructure having spatially configured porosity that affords low in-plane stiffness and strain tolerance.40–44 This strategy cannot be used for either the TGO or the bond coat: because, to serve their functions, both need to be dense (minimal porosity). The TGO misfit cannot be independently controlled, but its adverse consequences can be managed by limiting its thickness. The misfits between the bond coat and substrate are more nuanced: they occur not only from thermal expansion, but also phase transformations39 and swelling.45 Understanding these misfits, ascertaining their importance to system durability, and finding means to control them, has been an important research focus. Ultimately the durability is governed by spalling of the external insulating oxide, as deduced from components removed from engines (Fig. 4). Small diameter spalls can be tolerated, because backside cooling and boundary layer effects still allow the exposed surface to be protected by the (surrounding) intact oxide. Degradation only becomes a concern after an appreciable area fraction of the coating has been removed. Actual spall formation is preceded by smaller cracks that extend and coalesce along delamination planes located either within the oxide layer or at the interface between the TGO and the bond coat. The ensuing article highlights the roles of the oxide constituents. It is organized as follows. The constituent materials and their salient thermo-mechanical properties are outlined. The spectrum of mechanisms governing the performance and durability of hot section components are described, thereby illuminating the oxide functionalities. With reference to these mechanisms, the dominant characteristics of the oxides are discussed, with associated mechanistic understanding. In turn, these mechanisms reveal opportunities for new research on oxides that might further enhance the fuel efficiency. 2. The constituents and their thermo-mechanical properties The requirements imposed on each layer (Fig. 2) dictate the constituent property attributes. In current implementations, the structure and composition of the substrate and the insulating oxide are largely fixed. Options exist for the bond coat, which affect the formation of the ensuing TGO. 2.1. Insulating oxide The thermal expansion coefficient of this layer, αtbc, is appreciably lower than that for the substrate, αsub: the difference is about, αtbc − αsub ≡ αtbc ≈ −3 ppm/K. To prevent spontaneous delamination due to this misfit, the in-plane modulus of the layer, Etbc, must be controlled, as illustrated by the following simple argument. At the highest temperature, creep in the YSZ causes it to become stress-free. Subsequent cooling induces residual stress through the thermal expansion misfit. If the YSZ were fully dense (Etbc = 200 GPa, vtbc ≈ 0.2), for a typical value of the average temperature
A.G. Evans et al. Journal of the European Ceramic Society 28(2008)1405-1419 1407 EXPLODED DESIGNATION REQUIREMENTS VIEW T, >T THERMAL Low Thermal Conductivity. OXIDE Microstructural Stability Chemical Compatibility MAy30MAYMAT2 Scale Change -A2O3 THERMALLY Minimum Thickness GROWN OXIDE TGO) Controlled defects Adherent With BC TGO Chemically Homogeneous orms o-A22O Devoid Of Segregants Creep Resistant. (b) TBC ond Coat Superalloy 50u on zone 20 um Fig. 2. An exploded view of the tri-layer thermal barrier system indicating the functionalities of each of the layers. Cross sections of actual systems are included
A.G. Evans et al. / Journal of the European Ceramic Society 28 (2008) 1405–1419 1407 Fig. 2. An exploded view of the tri-layer thermal barrier system indicating the functionalities of each of the layers. Cross sections of actual systems are included
A G. Evans et al Joumal of the European Ceramic Sociery 28(2008)1405-1419 lay with ideal"bond coat would have the following attributes: (i)resis- tant to inter-diffusion with the substrate, (ii) minimal strain misfit YSZ [101 with the substrate(based on thermal expansion, phase transfor- mations and minimal inter-diffusion-induced swelling) and (iii) EB-PVD YSZ [11 high creep strength with adequate ductility. All of these pref erences cannot be realized simultaneously. The challenge has been to identify those attributes having the greatest importance Gdzr207[11 La2zr2o7 2.3. Thermally grown oxide EB-PVD Gd2Zr2O7 [11] The characteristics of the TGO are controlled largely by the nd coat microstructure and microchemistry, but modulated 00 by impurities, water vapor and dopants. Upon initial oxida tion, transient phases of alumina generally form. Later, these 2500 convert into a-Al2O3. -s The bond coats used in practice develop this phase at a relatively early stage within the cyclic life, minimizing adverse influences of the phase transforma- tion on durability. As the a-Al2O3 layer grows, it develops a small(but significant) compressive stress. 54.5 Upon cooling. the compression increases dramatically, due to thermal expan sion misfit with the substrate: atgo -asub Aatgo -7 ppm/K, 1500 such that igo N-4 GPa at ambient 56-58 Consequently, even though the TGo may be relatively thin at the end of the cyclic life (htgo 6 um), the energy stored/area is quite large. Utgo=tgohtgo/2Etgo 80J/m- and contributes substantiall to the potential for delamination at the tgo/bond coat interface (Fig. 5) 2.4. Inte 70 While interfaces between metals and oxides involve funda mentally strong(covalent and ionic)bonds, 9-6I their adhesion mole fraction YO,5 can be compromised by minor impurities(S is especially detrimental). To inhibit such degradation, there has been a (a)The thermal conductivity of several insulating, ternary oxides as a long history in the industry of systematically lowering the S of temperature. ( b)A binary phase diagram for the ZrOz-YO1s level in superalloys, as well as using selected alloy additions(Y, nowing the phases expected. A line representative of the cyclic dura- Pt, Hf, etc. ) to tie-up remnant S drop(ATN1100C), the residual stress at ambient would 3. Mechanisms limiting the durability of hot section be,oR≈Etc△the△T(l-h)≈-8GPa. For thickness, components hc100μm,. the stored energy/area,Ubc≡哏hh/2Ebc2 160J/m", would substantially exceed the mode I toughness (rtbe 45 J/m2 for 7-YSZ2), rendering the system prone systems was the difficulty in realizing laboratory tests that repro- to spontaneous delamination. To obviate this problem, duced the conditions that arise in an operating turbine. Furnace deposition methods have been developed that create a non- cycle and burner rig tests were widely used, but the spalling dense microstructure with appreciably lower in-plane modulus, mechanisms were not always representative of those found in Ebc<50 GPa. 40,4 In this modulus range, the stored energy airfoils, shrouds or combustors removed from actual engine becomes of order the toughness(typically, Utbc 45 J/m2 for service. As the body of information acquired on component Htbe= 150 um), enabling implementation. The columnar struc- accumulated, this concern became less problematic. A remain- ture developed by EB-PvD is especially effective. 42 ing issue is the merit of purported failure mechanisms presented in the literature, obtained on specimens tested in a laboratory 2.2 Bond coat setting. To eliminate the concern, each of the mechanisms pre- sented below has been carefully scrutinized and correlated with The relationships between the properties of the bond coat and engine experience. Namely, the mechanisms are those that the ystem durability are much more nuanced, because of the highly authors deem reproducible and verifiable, on the basis of engine
1408 A.G. Evans et al. / Journal of the European Ceramic Society 28 (2008) 1405–1419 Fig. 3. (a) The thermal conductivity of several insulating, ternary oxides as a function of temperature.10–14(b) A binary phase diagram for the ZrO2–YO1.5 system showing the phases expected.15 A line representative of the cyclic durability is superposed.16 drop (T ≈ 1100 ◦C), the residual stress at ambient would be, σR ≈ EtbcαtbcT/(1 − vtbc) ≈ −0.8 GPa. For thickness, Htbc ≥ 100m, the stored energy/area, Utbc ≡ σ2 RHtbc/2Etbc ≥ 160 J/m2, would substantially exceed the mode I toughness (Γ tbc ≈ 45 J/m2 for 7-YSZ23), rendering the system prone to spontaneous delamination.22 To obviate this problem, deposition methods have been developed that create a nondense microstructure with appreciably lower in-plane modulus, Etbc ≤ 50 GPa.40,41 In this modulus range, the stored energy becomes of order the toughness (typically, Utbc ≈ 45 J/m2 for Htbc = 150m), enabling implementation. The columnar structure developed by EB-PVD is especially effective.42 2.2. Bond coat The relationships between the properties of the bond coat and system durability are much more nuanced, because of the highly non-linear interplay with the substrate and the TGO.36,37 The “ideal” bond coat would have the following attributes: (i) resistant to inter-diffusion with the substrate, (ii) minimal strain misfit with the substrate (based on thermal expansion, phase transformations and minimal inter-diffusion-induced swelling) and (iii) high creep strength with adequate ductility. All of these preferences cannot be realized simultaneously. The challenge has been to identify those attributes having the greatest importance. 2.3. Thermally grown oxide The characteristics of the TGO are controlled largely by the bond coat microstructure and microchemistry, but modulated by impurities, water vapor and dopants. Upon initial oxidation, transient phases of alumina generally form. Later, these convert into -Al2O3. 46–53 The bond coats used in practice develop this phase at a relatively early stage within the cyclic life, minimizing adverse influences of the phase transformation on durability. As the -Al2O3 layer grows, it develops a small (but significant) compressive stress.54,55 Upon cooling, the compression increases dramatically, due to thermal expansion misfit with the substrate: αtgo − αsub ≡ αtgo ≈ −7 ppm/K, such that σtgo ≈ −4 GPa at ambient.56–58 Consequently, even though the TGO may be relatively thin at the end of the cyclic life (htgo ≈ 6m), the energy stored/area is quite large, Utgo = σ2 tgohtgo/2Etgo ≈ 80 J/m2 and contributes substantially to the potential for delamination at the TGO/bond coat interface (Fig. 5). 2.4. Interfaces While interfaces between metals and oxides involve fundamentally strong (covalent and ionic) bonds,59–61 their adhesion can be compromised by minor impurities (S is especially detrimental).61 To inhibit such degradation, there has been a long history in the industry of systematically lowering the S level in superalloys, as well as using selected alloy additions (Y, Pt, Hf, etc.) to tie-up remnant S. 3. Mechanisms limiting the durability of hot section components An early challenge in the implementation of thermal barrier systems was the difficulty in realizing laboratory tests that reproduced the conditions that arise in an operating turbine. Furnace cycle and burner rig tests were widely used, but the spalling mechanisms were not always representative of those found in airfoils, shrouds or combustors removed from actual engine service. As the body of information acquired on components accumulated, this concern became less problematic. A remaining issue is the merit of purported failure mechanisms presented in the literature, obtained on specimens tested in a laboratory setting. To eliminate the concern, each of the mechanisms presented below has been carefully scrutinized and correlated with engine experience. Namely, the mechanisms are those that the authors deem reproducible and verifiable, on the basis of engine
A.G. Evans et al. Journal of the European Ceramic Society 28(2008)1405-1419 1409 NT verging Cracks m N RUMPLING/RATCHETING EDGE-DELAMINATIONVOID FORMATION RINS Bond Coat .Delamination IMPACT DAMAGE Substrate MOLTEN DEPOSITS Fig 4. A summary of the various mechanisms that can cause spalling of the TBC on turbine airfoils. The intrinsic mechanisms are governed by strain misfits between the constituent layers upon thermal cycling. The extrinsic mechanisms are determined by external factors. Also shown at the left is an airfoil removed from engine ervice that contains various spalled regions. experience. They reside within two basic categories: intrinsic ings, three different intrinsic mechanisms have been identified, and extrinsic(Fig. 4). Those in the intrinsic category are not differentiated in terms of the surface exposed by the spall (i)Or especially sensitive to the presence of a thermal gradient in the mechanism exposes zirconia and some alumina on both delani- component and vice versa. The intrinsic category is character- nation surfaces. Cross sectioning indicates that it is accompanied ized by a group of mechanisms that arise because of the strain by rumpling (or ratcheting) of the TGO, manifest as undt misfits associated with the constituent materials. These mecha- tions that, locally, penetrate into the bond coat(Fig. 6). 62-64 nisms can often be reproduced in well-executed furnace cycle This mechanism arises primarily in systems with B-phase bond and burnerrig tests The failures are ultimately manifest as spalls, coats. (ii) A second exposes the bond coat, with periodic islands usually present in hot sections. In systems with EB-PVD coat- of TGO and some entrained zirconia. The bond coat exhibits imprints of the grains in the TGO, suggesting brittle failure by loss of adhesion at the metal/oxide interface. Cross sections affirm that the failure occurs primarily by delamination along the interface, with local extension through thickness heterogeneities in the tGo 1)A third but now with superposed features indicative of voids formed times 65 All intrinsic mechanisms have a characteristic Toughness TGO thickness, herit, at the incidence of spalling. However, hcrit depends on the bond coat composition and microstructure, as well as the thermal cycling history. In itself, it is not an useful metric for characterizing failure across a range of bond coats Delamination and cycling scenarios. The extrinsic category cannot be repro- duced in furnace cycling or conventional burner rig tests. The mechanisms include damage induced by particle impact(ero- sion and foreign object damage),66-69 delaminations enabled by the penetration of deposits of calcium-magnesium-alumino- Delamination silicate(CMAS) formed from the ingress into the engine of sands and dust in the atmosphere-as well as those introduced by thermal gradients. All are dominated by the microstructure and properties of the insulating oxide. The manifestations in TGO Thickness, htgo(um turbine hardware are as follows. Foreign object damage(FOD) Fig. 5. The energy release rates for delamination along either the TGO/bond is apparent as spalls at the leading edges of airfoils. Less severe ce, as a function of TGO thickness, or internally, within the TBC. particle impacts cause the gradual thinning of the TBC,by ero- is an estimate of the mode ll toughness of the interface. sion: also in the vicinity of the leading edges. CMAS damage
A.G. Evans et al. / Journal of the European Ceramic Society 28 (2008) 1405–1419 1409 Fig. 4. A summary of the various mechanisms that can cause spalling of the TBC on turbine airfoils. The intrinsic mechanisms are governed by strain misfits between the constituent layers upon thermal cycling. The extrinsic mechanisms are determined by external factors. Also shown at the left is an airfoil removed from engine service that contains various spalled regions. experience. They reside within two basic categories: intrinsic and extrinsic (Fig. 4). Those in the intrinsic category are not especially sensitive to the presence of a thermal gradient in the component and vice versa. The intrinsic category is characterized by a group of mechanisms that arise because of the strain misfits associated with the constituent materials. These mechanisms can often be reproduced in well-executed furnace cycle and burner rig tests. The failures are ultimately manifest as spalls, usually present in hot sections. In systems with EB-PVD coatFig. 5. The energy release rates for delamination along either the TGO/bond coat interface, as a function of TGO thickness, or internally, within the TBC. Also shown is an estimate of the mode II toughness of the interface. ings, three different intrinsic mechanisms have been identified, differentiated in terms of the surface exposed by the spall. (i) One mechanism exposes zirconia and some alumina on both delamination surfaces. Cross sectioning indicates that it is accompanied by rumpling (or ratcheting) of the TGO, manifest as undulations that, locally, penetrate into the bond coat (Fig. 6).62–64 This mechanism arises primarily in systems with -phase bond coats. (ii) A second exposes the bond coat, with periodic islands of TGO and some entrained zirconia. The bond coat exhibits imprints of the grains in the TGO, suggesting brittle failure by loss of adhesion at the metal/oxide interface. Cross sections affirm that the failure occurs primarily by delamination along the interface, with local extension through thickness heterogeneities in the TGO (Fig. 7).31 (iii) A third exposes the bond coat, but now with superposed features indicative of voids formed at longer times.65 All intrinsic mechanisms have a characteristic TGO thickness, hcrit, at the incidence of spalling. However, hcrit depends on the bond coat composition and microstructure, as well as the thermal cycling history. In itself, it is not an useful metric for characterizing failure across a range of bond coats and cycling scenarios. The extrinsic category cannot be reproduced in furnace cycling or conventional burner rig tests. The mechanisms include damage induced by particle impact (erosion and foreign object damage),66–69 delaminations enabled by the penetration of deposits of calcium–magnesium–aluminosilicate (CMAS) formed from the ingress into the engine of sands and dust in the atmosphere70–72 as well as those introduced by thermal gradients.21 All are dominated by the microstructure and properties of the insulating oxide. The manifestations in turbine hardware are as follows. Foreign object damage (FOD) is apparent as spalls at the leading edges of airfoils. Less severe particle impacts cause the gradual thinning of the TBC, by erosion: also in the vicinity of the leading edges. CMAS damage
1410 A G. Evans et al Joumal of the European Ceramic Sociery 28(2008)1405-1419 1mm The pre-eminent infuences of the oxide constituents emerge in the context of the foregoing mechanisms. The alumina influ ences all of the intrinsic mechanisms. When the spalling is dictated by rumpling of the TGo, three of the most important influences are attributed to the alumina: (a) its thickening rate, (b)the magnitude of the stress induced as it thickens and(c)the thermal expansion misfit with the substrate. Small values of all three are most desirable. The microstructure and properties of the insulating layer have greatest influence upon the extrinsic mechanisms. The salient properties include:(a) the toughness (b)the yield strength at high temperature, (c)the in-plane mod- ulus and(d)its densification rate. The incidence and extent of all of the cracking and delamination mechanisms scale directly with the toughness. Indeed a central attribute of 7-YSz is its rela tively high toughness. The yield strength plays a role through the ability of plasticity to dissipate the kinetic energy from particles in the airfoils. The modulus affects the level of residual stress induced as the system thermally cycles because of the thermal expan Crack Nucleation sion misfit with the substrate. In turn. the modulus is affected by the microstructure, established by the method of deposition, and modified by sintering or CMAS penetration 4. The role of the alumina 50 um 4.1. Thickening and elongation Over the thickness range that dominates intrins ity, the salient phenomena are dictated by the a-Al2O3 phase with columnar grain structure(Fig 8). At the temperatures of interest(1000-1125C), the growth involves counter-diffusion of oxygen and aluminum along the a-Al2O3 grain boundaries (Figs.8 and 9). 73 The diffusion flux, and hence, the parabolic rate constant, is dictated by the grain size of the a-Al2O3 as well as by impurities and by dopants that diffuse to the tGo and become entrained. While there is a wealth of empirical with a ridge crack along the on this topIc, the fundamentals remain elusive. Two issues top. This buckle initiated at the upper-free edge.(b and c)Cross sections away rom the buckle revealing the development of multiple(small)cracks in the TBC above the tgo (i) Phase transformation from amorphous alumina(formed during initial stages of heating)through various transient is found in the hottest sections of airfoils, especially along the polymorphs to the stable a-Al2O3. The rate of transfor- pressure surface, and in shrouds. Once molten, the CMAs is mation determines the nucleation rate of the stable phase evident through a yellow coloration associated with transition Is well as its lateral growth rate. In turn, this determines elements(such as Fe and Mn), in the deposits. The mechanism the grain size of the a-Al2O3, thereby controlling the sub- operates in the presence of a thermal gradient that plays a dual sequent thickening rate (as oxidation proceeds by grain role.2(a)It enables the CMAS (once molten)to penetrate to a boundary diffusion). The a-Al2O3 forms by means of a specified depth into the oxide. (b) It causes the surface to experi- nucleation and growth process. Understanding the role of ence residual tensile stress upon cooling. In turn, these stresses certain dopants on this process is important because(a) Cr provide the energy release rate, G, that enables internal delani nation(substantially elevated by the increase in stiffness caused earths such orporated for corrosion resistance and(b)rare as hf and y are used to control interface adhe- y CMAs penetration). Because it has not yet been possible sion, as well as the rumpling propensity(discussed below ) to adequately duplicate these mechanisms in laboratory scale It is known that doping with Hf and Y retards the growth tests, the mechanistic understanding has been based on obser- rate of the a-Al2O3 once nucleated. This benefit happens rations and measurements made on components removed from because these large cations are soluble in engines. The implementation of new testing facilities will rectify but not in alpha, causing them to be rejected from the grow this shortcoming ing a. Consequently, the interface cannot advance into the
1410 A.G. Evans et al. / Journal of the European Ceramic Society 28 (2008) 1405–1419 Fig. 6. (a) A large-scale buckle in a TBC coating with a ridge crack along the top. This buckle initiated at the upper-free edge. (b and c) Cross sections away from the buckle revealing the development of multiple (small) cracks in the TBC above the TGO. is found in the hottest sections of airfoils, especially along the pressure surface, and in shrouds. Once molten, the CMAS is evident through a yellow coloration associated with transition elements (such as Fe and Mn), in the deposits. The mechanism operates in the presence of a thermal gradient that plays a dual role.21 (a) It enables the CMAS (once molten) to penetrate to a specified depth into the oxide. (b) It causes the surface to experience residual tensile stress upon cooling.72 In turn, these stresses provide the energy release rate, G, that enables internal delamination (substantially elevated by the increase in stiffness caused by CMAS penetration). Because it has not yet been possible to adequately duplicate these mechanisms in laboratory scale tests, the mechanistic understanding has been based on observations and measurements made on components removed from engines. The implementation of new testing facilities will rectify this shortcoming. The pre-eminent influences of the oxide constituents emerge in the context of the foregoing mechanisms. The alumina influences all of the intrinsic mechanisms. When the spalling is dictated by rumpling of the TGO, three of the most important influences are attributed to the alumina: (a) its thickening rate, (b) the magnitude of the stress induced as it thickens and (c) the thermal expansion misfit with the substrate. Small values of all three are most desirable. The microstructure and properties of the insulating layer have greatest influence upon the extrinsic mechanisms. The salient properties include: (a) the toughness, (b) the yield strength at high temperature, (c) the in-plane modulus and (d) its densification rate. The incidence and extent of all of the cracking and delamination mechanisms scale directly with the toughness. Indeed, a central attribute of 7-YSZ is its relatively high toughness. The yield strength plays a role through the ability of plasticity to dissipate the kinetic energy from particles circulating in the turbine that impinge onto the rapidly rotating airfoils. The modulus affects the level of residual stress induced as the system thermally cycles because of the thermal expansion misfit with the substrate. In turn, the modulus is affected by the microstructure, established by the method of deposition, and modified by sintering or CMAS penetration. 4. The role of the alumina 4.1. Thickening and elongation Over the thickness range that dominates intrinsic durability, the salient phenomena are dictated by the -Al2O3 phase with columnar grain structure (Fig. 8). At the temperatures of interest (1000–1125 ◦C), the growth involves counter-diffusion of oxygen and aluminum along the -Al2O3 grain boundaries (Figs. 8 and 9).73 The diffusion flux, and hence, the parabolic rate constant, is dictated by the grain size of the -Al2O3 as well as by impurities and by dopants that diffuse to the TGO and become entrained. While there is a wealth of empirical data on this topic, the fundamentals remain elusive. Two issues are critical: (i) Phase transformation from amorphous alumina (formed during initial stages of heating) through various transient polymorphs to the stable -Al2O3. The rate of transformation determines the nucleation rate of the stable phase, as well as its lateral growth rate. In turn, this determines the grain size of the -Al2O3, thereby controlling the subsequent thickening rate (as oxidation proceeds by grain boundary diffusion). The -Al2O3 forms by means of a nucleation and growth process. Understanding the role of certain dopants on this process is important because (a) Cr is always incorporated for corrosion resistance and (b) rare earths such as Hf and Y are used to control interface adhesion, as well as the rumpling propensity (discussed below). It is known that doping with Hf and Y retards the growth rate of the -Al2O3 once nucleated. This benefit happens because these large cations are soluble in gamma and theta but not in alpha, causing them to be rejected from the growing . Consequently, the interface cannot advance into the
A.G. Evans et al. Journal of the European Ceramic Society 28(2008)1405-1419 1411 4 Cross sections through a burner rig test specimen comprising an MCrAlY bond coat after thermal cycling to full life. (a) A section remote from the spalled showing the intact interface. (b) A region adjacent to the spalled zone showing that the delamination follows the interface except where it extends across a thickness heterogeneity. metastable phase until the excess solute precipitates, since 4.2. The stresses in the TGo the local driving force is insufficient. Conversely, Cr and Fe (cations soluble in a)accelerate the growth rate: a feature If all of the new a-Al2O3 formed at the interface with the elucidated by polishing the surface of bond coat alloys with bond coat, the ensuing volume increase would be accommodated Cr and Fe oxides and noting preferential growth of alumina by upward(rigid body) motion of the prior TGo, obviating on heterogeneous nuclei created at the polishing features growth stress. Instead, the outward counter-fiux of Al causes (scratches, grit-blast grooves, etc. )(Fig. 10). Given these some new a-Al2O3 to form at dislocations/ledges along the counteracting influences of the dopants, a compromise in transverse grain boundaries, as well as that formed on the the relative dopant levels is required. However, because a face of the TGO(Fig 9). That formed at the boundaries must be quantitative model is lacking, reliance has been placed on accommodated by lateral deformation of the neighboring grains empiricism, with adverse consequences for progress toward causing a compressive growth stress(Fig. 11). Since a-Al2O3 optimal doping strategies is susceptible to plastic deformation at the growth tempera (i) The relative inward and outward diffusive fluxes along the ture, the stress attains a"steady-state, wherein the strain-rate grain boundaries in the ensuing a-Al2O3 influences both induced by the growth is balanced by the creep-rate. The mag- thickening and elongation. The grain boundaries governing nitude of this stress has been measured in situ for the TGo these effects are clean and devoid of amorphous interphases formed on several different bond coats. It is measured to be of (although sub-monolayers of Hf or Zr can be entrained). The order, a growth -300MPa4(Fig. 12). While this stress level conventional picture is that, once a-Al2O3 is formed, oxide appears reasonable based on deformation mechanism maps for hickening is dominated by inward diffusion of O. In prac- a-Al2O3 and stress relaxation rates in a typical TGO, it remains tice, there is a counter-flux of cations, 73, 74 demonstrated by to develop a quantitative model. Moreover, the influences of using the following protocol. The alloy is oxidized to form role of dopants on growth strains and relaxations are poorly a continuous TGO, which is then polished at an angle, and understood. Upon cooling, because of its relatively low thermal re-oxidized. Some of the new oxide forms as ridges along expansion coefficient(relative to the substrate)a large in-plane locations where the oxide grain boundaries intersect the sur- compression develops(Fig. 12). At ambient, whenever the TGO face(Fig 9). There are corresponding ridges where the grain remains planar(no rumpling), the compressive stress is in the boundaries intersect the interface. Measurement of the vol- range, -35<0tgo<-6GPa, depending on the thermal expan- ume of these ridges allows assessment of the relative anion sion coefficient for the substrate. 8 When rumpling occurs, the and cation fluxes (note that, in Fig. 9 the inward and outward stress diminishes because of bending and elongation of the fluxes must be of comparable magnitude). Such measure- TGO 37.57 ments reveal that the counter-fluxes depend sensitively Rumpling is highly non-linear phenomenon, involving inter dopants such as Hf and Y entrained in the grain boundaries ctions between the tgo. bond coat and substrate and reliant of the growing oxide. The unresolved question is how to on many thermo-mechanical properties of the layers. The inte think about the atomic mechanisms of counter-diffusion, actions have been unearthed through the development of a code and how this process determines elongation strain. These and its validation by incisive experiments. While the rumpling fundamentals have not been addressed in the literature rate is strongly influenced by the strain misfits between the sul
A.G. Evans et al. / Journal of the European Ceramic Society 28 (2008) 1405–1419 1411 Fig. 7. Cross sections through a burner rig test specimen comprising an MCrAlY bond coat after thermal cycling to full life. (a) A section remote from the spalled region showing the intact interface. (b) A region adjacent to the spalled zone showing that the delamination follows the interface except where it extends across a thickness heterogeneity. metastable phase until the excess solute precipitates, since the local driving force is insufficient. Conversely, Cr and Fe (cations soluble in ) accelerate the growth rate: a feature elucidated by polishing the surface of bond coat alloys with Cr and Fe oxides and noting preferential growth of alumina on heterogeneous nuclei created at the polishing features (scratches, grit-blast grooves, etc.) (Fig. 10). Given these counteracting influences of the dopants, a compromise in the relative dopant levels is required. However, because a quantitative model is lacking, reliance has been placed on empiricism, with adverse consequences for progress toward optimal doping strategies. (ii) The relative inward and outward diffusive fluxes along the grain boundaries in the ensuing -Al2O3 influences both thickening and elongation. The grain boundaries governing these effects are clean and devoid of amorphous interphases (although sub-monolayers of Hf or Zr can be entrained). The conventional picture is that, once -Al2O3 is formed, oxide thickening is dominated by inward diffusion of O. In practice, there is a counter-flux of cations,73,74 demonstrated by using the following protocol. The alloy is oxidized to form a continuous TGO, which is then polished at an angle, and re-oxidized. Some of the new oxide forms as ridges along locations where the oxide grain boundaries intersect the surface (Fig. 9). There are corresponding ridges where the grain boundaries intersect the interface. Measurement of the volume of these ridges allows assessment of the relative anion and cation fluxes (note that, in Fig. 9 the inward and outward fluxes must be of comparable magnitude). Such measurements reveal that the counter-fluxes depend sensitively on dopants such as Hf and Y entrained in the grain boundaries of the growing oxide. The unresolved question is how to think about the atomic mechanisms of counter-diffusion, and how this process determines elongation strain. These fundamentals have not been addressed in the literature. 4.2. The stresses in the TGO If all of the new -Al2O3 formed at the interface with the bond coat, the ensuing volume increase would be accommodated by upward (rigid body) motion of the prior TGO, obviating a growth stress. Instead, the outward counter-flux of Al causes some new -Al2O3 to form at dislocations/ledges along the transverse grain boundaries, as well as that formed on the surface of the TGO (Fig. 9). That formed at the boundaries must be accommodated by lateral deformation of the neighboring grains, causing a compressive growth stress (Fig. 11). Since -Al2O3 is susceptible to plastic deformation at the growth temperature, the stress attains a “steady-state”, wherein the strain-rate induced by the growth is balanced by the creep-rate. The magnitude of this stress has been measured in situ for the TGO formed on several different bond coats. It is measured to be of order, σgrowth ≈ −300 MPa54 (Fig. 12). While this stress level appears reasonable based on deformation mechanism maps for -Al2O3 and stress relaxation rates in a typical TGO, it remains to develop a quantitative model. Moreover, the influences of role of dopants on growth strains and relaxations are poorly understood. Upon cooling, because of its relatively low thermal expansion coefficient (relative to the substrate) a large in-plane compression develops (Fig. 12). At ambient, whenever the TGO remains planar (no rumpling), the compressive stress is in the range, −3.5 < σtgo < −6 GPa, depending on the thermal expansion coefficient for the substrate.58 When rumpling occurs, the stress diminishes, because of bending and elongation of the TGO.37,57 Rumpling is highly non-linear phenomenon, involving interactions between the TGO, bond coat and substrate and reliant on many thermo-mechanical properties of the layers. The interactions have been unearthed through the development of a code and its validation by incisive experiments.36 While the rumpling rate is strongly influenced by the strain misfits between the sub-
A G. Evans et al Joumal of the European Ceramic Sociery 28(2008)1405-1419 thickness that causes delamination(Fig. 5). In practice, this approach to predicting the critical thickness has not been real- ized, because the mode II toughness is a notoriously difficult property to measure. Instead, estimates based on the mode I and mixed-mode delamination toughness have been invoked However, these results demonstrate order of magnitude varia- ons that depend on th used to form the interface. Instead of placing reliance solely on measurements, a simulation scheme that distinguishes the factors dominating the adhesion(Fig. 13)is being pursued. It has two basic ingredients. (i) The traction/separation characte istics during bond rupture at the interface are ascertained using a first principles approach based on density functional theory Wuum(i) These results are input to an embedded process zone(EPz) simulation of interface crack extension that captures the mul- tiplicative influence on the toughness of the plastic dissipation occurring in the bond coat. In general, the traction/separation GB curves are found to depend on the termination plane(stoichiom- etry)of the a-Al2O3, as well as the presence of dopants and impurities. For y-Ni(Al) alloys, the termination has been ascer- tained to be a mix of stoichiometric and al-rich the former is the least adherent(Fig. 14). Moreover, when present, S seg regates to this interface and further decreases the adhesion(by up to 70%), because the interfacial covalent-ionic Ni-O bonds eplaced with weaker ior with Hf obviates the detriment, especially when it segregates on interstitial sites(Hfn)(Fig. 14), because Hf-Ni and Hf-o bonds effectively knit the surfaces together. When integrated(Fig. 13), hese results establish a rationale for designing tough interfaces. Recall that Hf has the additional benefit that it affects the creep strength of the TGO when it segregates to the grain boundaries. 5. The insulating oxide Beyond the basic requirement that the thermal conductiv ity be low and(preferably) temperature invariant, the following properties are critical to system performance. Toughness affects all of the extrinsic mechanisms. Remarkably, the range realiz GB able among all(non-fibrous) oxides is fully encompassed by YSZ across the composition range between cubic and tetrag Fig 8.()Fractured cross section of a TGo illustrating the inner, columnar onal(Fig. 15). The cubic materials(c-ZrO2-20-YSZ)are portion of the oxide formed by inward diffusion of o and the outer, equiaxed exceptionally brittle(toughness, Ia6J/m), while partially portion formed by outward diffusion of AL. (b)Schematic diagram showing the stabilized tetragonal materials( t-Zr02-3-YSZ), which expe- rience a martensitic transformation to the monoclinic phase (m-ZrO,), are among the to strate and bond coat, as well as its creep strength, the TGo is formation mechanism is inapplicable for two related reasons also important. From a TGO perspective, the growth stress and(a) It is thermodynamically forbidden at elevated temperatures, the thickening rate are most influential specifically those above To(t/m)(Fig. 16a), wherein there is no driving force for the partitionless t->m transformation.(b) 4.3. Interface adhesion Repeated cycling across the To(t/m) results in disruptive vol- ume changes every time the t-Zro2 transforms to m-zrO2 on When rumpling is suppressed, durability is limited by delam- cooling and regenerates upon heating, with concomitant micro- ination along the interface between the TGO and the bond coat. cracking Compositions within the non-transformable tetragonal The energy release rate enabling this mechanism(Fig. 5)is com- () phase field, bound by the compositions for which To(t/m)is municated to the interface as a mode Il(shear)delamination. below ambient and To(c/t)is below the maximum operating tem- In principle, equating the energy release rate to the mode ll perature, provide the best performance Because tetragonality toughness of the interface predicts a lower bound on the tGo typically decreases with increasing dopant content the pre
1412 A.G. Evans et al. / Journal of the European Ceramic Society 28 (2008) 1405–1419 Fig. 8. (a) Fractured cross section of a TGO illustrating the inner, columnar portion of the oxide formed by inward diffusion of O and the outer, equiaxed portion formed by outward diffusion of Al. (b) Schematic diagram showing the flux paths. strate and bond coat, as well as its creep strength, the TGO is also important. From a TGO perspective, the growth stress and the thickening rate are most influential. 4.3. Interface adhesion When rumpling is suppressed, durability is limited by delamination along the interface between the TGO and the bond coat. The energy release rate enabling this mechanism (Fig. 5) is communicated to the interface as a mode II (shear) delamination. In principle, equating the energy release rate to the mode II toughness of the interface predicts a lower bound on the TGO thickness that causes delamination (Fig. 5). In practice, this approach to predicting the critical thickness has not been realized, because the mode II toughness is a notoriously difficult property to measure. Instead, estimates based on the mode I and mixed-mode delamination toughness have been invoked. However, these results demonstrate order of magnitude variations that depend on the presence of segregants and the method used to form the interface. Instead of placing reliance solely on measurements, a simulation scheme that distinguishes the factors dominating the adhesion (Fig. 13) is being pursued. It has two basic ingredients. (i) The traction/separation characteristics during bond rupture at the interface are ascertained using a first principles approach based on density functional theory. (ii) These results are input to an embedded process zone (EPZ) simulation of interface crack extension that captures the multiplicative influence on the toughness of the plastic dissipation occurring in the bond coat. In general, the traction/separation curves are found to depend on the termination plane (stoichiometry) of the -Al2O3, as well as the presence of dopants and impurities. For -Ni(Al) alloys, the termination has been ascertained to be a mix of stoichiometric and Al-rich. The former is the least adherent (Fig. 14). Moreover, when present, S segregates to this interface and further decreases the adhesion (by up to 70%), because the interfacial covalent-ionic Ni–O bonds are replaced with weaker ionic-covalent S–Al bonds. Doping with Hf obviates the detriment, especially when it segregates on interstitial sites (HfI) (Fig. 14), because Hf–Ni and Hf–O bonds effectively knit the surfaces together. When integrated (Fig. 13), these results establish a rationale for designing tough interfaces. Recall that Hf has the additional benefit that it affects the creep strength of the TGO when it segregates to the grain boundaries. 5. The insulating oxide Beyond the basic requirement that the thermal conductivity be low and (preferably) temperature invariant, the following properties are critical to system performance. Toughness affects all of the extrinsic mechanisms. Remarkably, the range realizable among all (non-fibrous) oxides is fully encompassed by YSZ across the composition range between cubic and tetragonal (Fig. 15).22 The cubic materials (c-ZrO2 →20-YSZ) are exceptionally brittle (toughness, Γ ≈ 6 J/m2), while partially stabilized tetragonal materials (t-ZrO2 →3-YSZ), which experience a martensitic transformation to the monoclinic phase (m-ZrO2), are among the toughest (Γ > 300 J/m2). The transformation mechanism is inapplicable for two related reasons. (a) It is thermodynamically forbidden at elevated temperatures, specifically those above T0(t/m) (Fig. 16a), wherein there is no driving force for the partitionless t→m transformation. (b) Repeated cycling across the T0(t/m) results in disruptive volume changes every time the t-ZrO2 transforms to m-ZrO2 on cooling and regenerates upon heating, with concomitant microcracking. Compositions within the non-transformable tetragonal (t ) phase field, bound by the compositions for which T0(t/m) is below ambient and T0(c/t) is below the maximum operating temperature, provide the best performance. Because tetragonality typically decreases with increasing dopant content75 the pre-
A.G. Evans et al. Journal of the European Ceramic Society 28(2008)1405-1419 1413 um 1 um REO Fig. 9. Illustration of the same area of an alumina TGO on re-oxidation after smoothly polishing the TGO formed in the first oxidatio New oxide forms the grain boundaries of t ally formed TGO and the amount increases with further oxidation. Schematic of the counter-diffusion of O and Al along the grain boundaries leading to a thickening of the oxide above and below the oxide on either side of the boundary. In between grain boundaries, the oxide is signifi inner as indicated by the arrows. ferred compositions within this range are those at the lower dense 7-YSZ) are representative: consistent with a microstruc- end, pre-eminently exemplified by 7-YSZ. Typically, the mode ture comprising columns bonded at periodic attachment sites. I toughness is in the range, 40<I<50J/m223-25,76: sufficient Yielding. The ability of the layer to yield when impacted by for to prevent spalling after manufacturing and to suppress large- eign objects in the engine is a significant attribute. The plastic scale delamination when exposed to thermal gradients. Such response upon impact at high temperature is reflected in the rel- compositions are ferro-elastic 77-79 Upon crack extension, dissi- atively low yield strength of 7-YSZ above 900C(Fig. 18a) pation occurs through the formation(or switching)of nano-scale The associated plastic dissipation serves to absorb much of the domains, resulting in toughening that scales with the tetrago- kinetic energy from the impact and thereby, diminish the ampli nality, c/a, and the coercive stress 2324. The concept that tude of the elastic waves that propagate through the layer2 Also the tetragonality governs the toughness of tetragonal oxides has of interest is the role of yielding in the development of kink bands led to the development of new compositions with appreciably in systems having columnar microstructure(Fig. 18b). The plas- higher toughness than7-YSZ24 25.80 The most prominent exam- tic bending of the columns is apparent, as well as the incidence ple resides within the ZrO2-YO1 5-TiO2 ternary phase field of cracks wherever the bending induces large tensile strains.68. 76 ig. 16), having toughnes 24 As engine temperatures continue to increase, two factors lus. Given that the in-plane modulus of the layer is so important, will inevitably limit the capability of 7-YSZ and tetragonal it is remarkable that measurements are still sparse. The only compositions derived from it. One limitation arises from the comprehensive results are those published by Johnson et al. 40 metastable nature of non-transformable t'phases, since the Values in the range 20<E<40 GPa(compared with 200 GPa for amount of solute required for non-transformability typically
A.G. Evans et al. / Journal of the European Ceramic Society 28 (2008) 1405–1419 1413 Fig. 9. Illustration of the same area of an alumina TGO on re-oxidation after smoothly polishing the TGO formed in the first oxidation step. New oxide forms along the grain boundaries of the initially formed TGO and the amount increases with further oxidation. Schematic of the counter-diffusion of O and Al along the TGO grain boundaries leading to a thickening of the oxide above and below the oxide on either side of the boundary. In between grain boundaries, the oxide is significantly thinner as indicated by the arrows. ferred compositions within this range are those at the lower end, pre-eminently exemplified by 7-YSZ. Typically, the mode I toughness is in the range, 40 < Γ < 50 J/m2 23–25,76: sufficient to prevent spalling after manufacturing and to suppress largescale delamination when exposed to thermal gradients. Such compositions are ferro-elastic.77–79 Upon crack extension, dissipation occurs through the formation (or switching) of nano-scale domains, resulting in toughening that scales with the tetragonality, c/a, and the coercive stress.23,24,77–79 The concept that the tetragonality governs the toughness of tetragonal oxides has led to the development of new compositions with appreciably higher toughness than 7-YSZ.24,25,80 The most prominent example resides within the ZrO2–YO1.5–TiO2 ternary phase field (Fig. 16), having toughness, Γ ≈ 90 J/m2 24 (Fig. 17). Modulus. Given that the in-plane modulus of the layer is so important, it is remarkable that measurements are still sparse. The only comprehensive results are those published by Johnson et al.40 Values in the range 20 < E < 40 GPa (compared with 200 GPa for dense 7-YSZ) are representative: consistent with a microstructure comprising columns bonded at periodic attachment sites. Yielding. The ability of the layer to yield when impacted by foreign objects in the engine is a significant attribute. The plastic response upon impact at high temperature is reflected in the relatively low yield strength of 7-YSZ above 900 ◦C (Fig. 18a). The associated plastic dissipation serves to absorb much of the kinetic energy from the impact and thereby, diminish the amplitude of the elastic waves that propagate through the layer.21 Also of interest is the role of yielding in the development of kink bands in systems having columnar microstructure (Fig. 18b). The plastic bending of the columns is apparent, as well as the incidence of cracks wherever the bending induces large tensile strains.68,76 As engine temperatures continue to increase, two factors will inevitably limit the capability of 7-YSZ and tetragonal compositions derived from it. One limitation arises from the metastable nature of non-transformable t’ phases, since the amount of solute required for non-transformability typically
1414 A G. Evans et al Joumal of the European Ceramic Sociery 28(2008)1405-1419 EXPLODED Al2O3 GROWTH New 50μm Fig. 10. Illustration of preferred nucleation of alpha-alumina islands on hetero- eneous sites, such as scratches, on the alloy surface. Optical micrographs. △D/D exceeds the solubility limit for most viable stabilizers. The Fig. 11. A schematic indicating the elastic compression of the TGO grains implication is that all these compositions have a durability ulti- needed to accommodate the formation of alumina at the internal grain mately limited by the partitioning of the supersaturated t' into boundaries. the equilibrium phases, and the potential for undesirable trans- formation of the depleted tetragonal phase into monoclinic on ooling. 81-87 Small windows of opportunity exist in ternary systems. The composition exhibiting highest toughness resides in the ZrO2-YO15-TiO2 system having a depleted tetragonal phase that is non-transformable upon cooling. 4 An interest ing opportunity is presented by the ZrO2-YO1.5-TaO25 system trigonal field Growth and exhibits a regime in which compositions are both immune toughness comparable to(or slightly higher than)7-YSZ 25 This s, to phase separation(up to at least 1500C)and also exhibit system has yet to be synthesized in coating form and evaluated for durability Heatin compositions)is the degradation by CMAS penetration. 7288 a Rare earth zirconates, notably Gd2Zr2O7, offer a solution as they react with the CMAs melt and induce its crystallization at tem peratures well above the melting point of the original deposit. 89 Because the reaction and crystallization kinetics are competitive with that for infiltration, the process effectively seals the surface Time at Oxidation Temperature(min) against further penetration. However, RE Zirconates are typl- Fig. 12. An illustration of the stress in the TGO measured in situ in the sy cally cubic ( pyrochlore or d-phase)and exhibit the same low chrotron during a single thermal cycle from ambient to 1125%C and back to toughness as cubic zirconia compositions. Moreover, they are ambient
1414 A.G. Evans et al. / Journal of the European Ceramic Society 28 (2008) 1405–1419 Fig. 10. Illustration of preferred nucleation of alpha-alumina islands on heterogeneous sites, such as scratches, on the alloy surface. Optical micrographs. exceeds the solubility limit for most viable stabilizers.4 The implication is that all these compositions have a durability ultimately limited by the partitioning of the supersaturated t’ into the equilibrium phases, and the potential for undesirable transformation of the depleted tetragonal phase into monoclinic on cooling.81–87 Small windows of opportunity exist in ternary systems. The composition exhibiting highest toughness resides in the ZrO2–YO1.5–TiO2 system having a depleted tetragonal phase that is non-transformable upon cooling.24 An interesting opportunity is presented by the ZrO2–YO1.5–TaO2.5 system wherein the tetragonal field penetrates deeply into the ternary and exhibits a regime in which compositions are both immune to phase separation (up to at least 1500 ◦C) and also exhibit toughness comparable to (or slightly higher than) 7-YSZ.25 This system has yet to be synthesized in coating form and evaluated for durability. A second major limitation of 7-YSZ (and its derivative t’ compositions) is the degradation by CMAS penetration.71,72,88 Rare earth zirconates, notably Gd2Zr2O7, offer a solution as they react with the CMAS melt and induce its crystallization at temperatures well above the melting point of the original deposit.89 Because the reaction and crystallization kinetics are competitive with that for infiltration, the process effectively seals the surface against further penetration. However, RE zirconates are typically cubic (pyrochlore or -phase) and exhibit the same low toughness as cubic zirconia compositions. Moreover, they are Fig. 11. A schematic indicating the elastic compression of the TGO grains needed to accommodate the formation of new alumina at the internal grain boundaries. Fig. 12. An illustration of the stress in the TGO measured in situ in the synchrotron during a single thermal cycle from ambient to 1125 ◦C and back to ambient