Thermal barrier coatings technology: critical review,progress update,remaining challenges and prospects

Yttria stabilised zirconia top coat

Yttria stabilised zirconia has high melting point (not of much significance to current usage), low thermal conductivity, high oxygen permeability and relatively high coefficient of thermal expansion compared to other oxides and the 6–8 wt-%Y₂O₃ composition is relatively stable during high temperature exposure. An important attribute of YSZ for turbine design is that thermal conductivity is essentially temperature invariant. Another important attribute of YSZ is that ZrO₂ and Y₂O₃ have similar vapour pressures allowing deposition of YSZ with uniform composition by the EB-PVD process. This is unlike alternate TBC materials containing oxides of different vapour pressures yielding undesirable, non-uniform and layered structures. Yttria stabilised zirconia is applied either by APS or EB-PVD techniques. These processes make YSZ ‘strain tolerant’ by depositing a structure that contains numerous pores, gaps, microcracks and interfaces (Fig. 7). Strain tolerant structures are necessary to accommodate thermal expansion misfit between the substrate and top layer. The high oxygen permeability of the YSZ requires that the bond coats be highly resistant to oxidation and hot corrosion attack. The bond coat compositions are rich in aluminium to form a protective, TGO scale of α-Al₂O₃, typically 1–6 μm. The NiCoCrAl (Y, Hf, Si and Re) type overlay bond coats, typically 100–200 μm, are mainly based on the γ-Ni and β-NiAl phases. The Pt platinum modified diffusion aluminide bond coats, typically 50 μm, mainly consist of a β-NiAl–Pt phase. In addition to enhancing oxidation resistance, TGO serves to bond the ceramic top coat to the substrate/bond coat system. Thickening and buckling of the TGO scale and subsequent cracking is one of the several failure mechanisms of TBC. The adhesion and mechanical integrity of the TGO scale is dependent on the composition and impurity levels of the bond coat.

OPEN IN VIEWER

Figure 7.

Types of TBC porosity

A pseudobinary ZrO₂–Y₂O₃ phase diagram modified by Levi1 is shown in Fig. 8. In this figure, the Y₂O₃ content in the x axis is labelled as mole fraction YO_1·5 since wt-%Y₂O₃ roughly translates to 0·5 mol.-%Y₂O₃ in ZrO₂–Y₂O₃ compositions. Numerous modifications to precisely define phase boundaries have been proposed.15 As opposed to the equilibrium tetragonal phase t (t-ZrO₂), a closely related non-equilibrium, metastable phase t (t'-ZrO₂) is obtained in the as deposited TBC due to high rate deposition by either APS or EB-PVD. This phenomenon is highly desirable for achieving optimised lives. Though thermodynamically driven to partition into equilibrium t+c, the t' phase does not transform easily to tetragonal and monoclinic phases under normal circumstances until at least ∼1200°C,16,17 which is adequate in current use. Stability will become an issue for higher temperature operations. Transformation to monoclinic phase at lower Y₂O₃ contents is undesirable due to significant volume change (∼3 to 4%) causing cracks and TBC spallation. Higher Y₂O₃ contents have been shown to have lower fracture toughness, a possible cause of lower erosion and impact resistance to be discussed later.

OPEN IN VIEWER

Figure 8.

The zirconia rich corner of the ZrO₂–YO_1·5 binary phase diagram. Yttria stabilised zirconia composition and operating temperature range are shown by the hatched region. The tetragonal prime phase is metastable within the T _o bounds shown as dashed lines superimposed on the phase diagram (courtesy: Professor Carlos Levi, UCSB)

During the initial TBC development originating at NASA during the 1970s, it was demonstrated by Stecura18 ^– 20 in APS TBC that about 6–8 wt-% (∼3 to 4·5 mol.-%) Y₂O₃ addition to ZrO₂ provided the longest TBC thermal cycle life in burner rig tests, even though ZrO₂ was not fully stabilised. This conclusion was drawn with different test temperatures and various bond coat compositions. Time to first crack related to subsequent TBC spallation in thermal cycle tests as a function of Y₂O₃ content and four composition variations of NiCrAlY bond coat is shown in Fig. 9. Later, Stecura evaluated Yb₂O₃ containing ZrO₂ and concluded that maximum life is obtained between 12·4 and 14·7 wt-%Yb₂O₃. Thermal barrier coating life decreased very rapidly at concentrations exceeding these optimum values. It must be pointed out that, after extensive research for the last two decades to identify better top coat compositions, 6–8 wt-%Y₂O₃ content in ZrO2 continues to be optimum. Alternate oxide stabilisers evaluated primarily to reduce thermal conductivity have shown TBC lives to be inferior to YSZ to be discussed in a later section. To date, YSZ continues to be a material system of choice. It is interesting to note that Engel et al. 21 observed on ‘single-crystal’ ZrO₂ that maximum fracture toughness and fracture strength are obtained around 4–5 wt-%Y₂O₃ (Fig. 10).

OPEN IN VIEWER

Figure 9.

Time to first crack and subsequent TBC spallation in a thermal cycle test as a function of Y₂O₃ content for APS TBC (Stecura18 ^– 20)

OPEN IN VIEWER

Figure 10.

Flexure strength and fracture toughness of ZrO₂ single crystals as a function of Y₂O₃ content (Engel et al.21)

It has been proposed by Virkar22,23 and Baither et al. 24 based on experiments on sintered ZrO₂–3 mol.-%Y₂O₃ specimens that the t' phase microstructure consists of large grains on the order of 100 μm or greater. These grains contain very small domains (∼0·1 μm) arranged in three mutually orthogonal orientations in each grain making t' phase extremely non-transformable. However, these domains reorient under stress that provides high toughness, a phenomenon called ferroelastic domain switching. Mercer et al. 25 and co-workers have argued based on indentation and TEM studies that the nucleation of domains contributes to the toughness of YSZ TBC, and not the switching. This assertion was drawn by examining the material in the wake of an indentation induced crack by using TEM and by interferometry.

Yttria stabilised zirconia compositions are thermally stable up to the current use temperatures of TBC which is ∼1200°C. Yttria stabilised zirconia has shown minimum partition to the cubic+tetragonal phases which in turn transform to tetragonal+monoclinic phases respectively, upon cooling. It is another desirable attribute since destabilisation and subsequent monoclinic formation can lead to TBC spallation. Investigation on the thermal stability of YSZ and ZrO₂ containing various stabilisers has been conducted more recently by Rebollo et al. 16 At higher temperatures, in addition to destabilisation, sintering can modify pore structure influencing thermal conductivity as well as strain tolerance.

Processing methods for top coat

Thermal barrier coatings are porous with overall porosities generally in the range of 10–25%. In the EB-PVD process, an ingot (or ingots) of a YSZ composition is vapourised in a vacuum chamber using a focused electron beam. Before deposition, the samples are preheated to ∼1000°C in a low oxygen partial pressure environment resulting in the growth of a thin (∼0·05 to 0·1 μm) TGO layer. The parts are coated by manipulation within the vapour cloud on the preheated substrate surface. The EB-PVD process is especially effective in tailoring microstructures and inplane modulus. For example, EB-PVD can yield a very desirable columnar microstructure (Figs. 7 and 11) to result in a low inplane modulus that provides strain accommodation for the coefficient of thermal expansion mismatch between the metallic substrate and the ceramic top coat. The air plasma spray process yields a microstructure with horizontal splats of YSZ (Fig. 11). The interfaces/gaps between the splats provide additional reduction in thermal conductivity and thermal expansion compliance with the substrate. The pore structure characteristics of EB-PVD and APS are schematically shown in Fig. 12 (reproduced from Ref. 26). Various versions of the plasma spray processes are used as an alternative to the expensive EB-PVD process. Excellent reviews of the APS process and processing parameters, and the control of coating characteristics and properties are provided by Sampath.26,27 During preheating and top coat deposition, and during subsequent heat treatments, the bond coat generates an adherent α-alumina layer (TGO) between the bond coat and the YSZ top coat. Formation of this α-alumina of optimal thickness and crystallography is critical to achieve reproducible TBC life and prevent premature spallation of the TBC. A TGO thickness of about 0·3–1·0 μm is considered optimum right after deposition.28 The TGO thickens during the TBC use, eventually contributing to TBC spallation. Thermal barrier coating failure mechanisms are discussed in the section on ‘Temperature and thermal cycle dependent failure’.

OPEN IN VIEWER

Figure 11.

Thermal barrier coating top coat microstructures

OPEN IN VIEWER

Figure 12.

An illustrative description of microstructures of TBCs across length scales. The defect architectures of plasma sprayed coatings are differentiated from those of EB-PVD coatings26

Process selection considers performance, capital cost, per-part process cost, thickness requirements and control, composition requirements and control, surface roughness requirement, line of sight versus non-line of sight characteristics and degree of cooling hole closure. The EB-PVD process provides longer TBC life, smoother coating with excellent composition and thickness control, and causes minimum cooling hole closure. This process is preferred for blades and vanes in spite of much higher cost compared to the APS processes.

Air plasma spray processes are used for combustor parts, HPT shrouds and on most power generation components due to lower cost and requirement for thicker coatings. A variation of the APS process that is increasingly being used is ‘dense vertically cracked’ (DVC) TBC coating process.29 Vertical cracks are introduced to provide strain tolerance and serve similar function as intercolumnar gaps in the EB-PVD coatings at much lower cost. However, the performance is compromised in this lower cost version of the EB-PVD process.

Promising newer processing methods that are attempting to compete with EB-PVD and APS include directed vapour deposition, a variation of the EB-PVD process where a supersonic gas jet is used to direct vapours onto components being coated to increase process efficiency and deposition rate. The process acts as a non-line of sight process, and is able to coat multiple components with complex geometry.30 ^– 32 Composition uniformity and microstructural flexibility are other beneficial attributes. Suspension plasma spray or liquid solution precursor plasma spray33 ^– 35 creates microstructures with high segmented crack density and finer pores for reduced thermal conductivity and improved thermal cycle performance. The process employs suspensions of submicrometre particles as feedstock. This circumvents flow issues with finer particles in a powder feedstock, thereby, improving process efficiency. Microstructure and composition flexibility are two other advantages. A recent review of the developments in this technology is provided by Killinger et al. 36

Bond coat

Before introduction of TBC, turbine blades and vanes were protected from oxidative and corrosive environments with either diffusion coatings such as single phase, simple, β-NiAl or Pt modified β-NiAl or two phase overlay γ+β MCrAlX coatings (where M represents Ni or Co and X represents Y, Hf, Si, or other minor additives such as Re). The choice between these coating types by turbine engine manufacturers was based on experience, design philosophy, performance and cost. Although designed originally for environmental resistance, these coatings have been used as bond coats for TBC applications without any specific composition modifications and process development.

The MCrAlX type bond coats usually contain (in wt-%) 15–25%Cr, 10–15%Al and 0·2–0·5%Y, and consist of a β-NiAl phase in a γ-Ni matrix. These coatings, typically 100–150 μm thick, are applied using a variety of overlay processes: APS, high velocity oxyfuel, low pressure plasma spray, cathodic arc/ion plasma deposition or EB-PVD. The overlay coating processes are about 2–4 times more expensive than the diffusion processes, but offer much greater control of the coating composition since the coating composition is dictated essentially by the composition of the coating source. This attribute provides flexibility in depositing bond coat layers of various different chemical compositions. In actual practice, a knowledge of the vapour pressure, deposition efficiency and spatial distribution of all the chemical elements of the coating is required. The β-NiAl bond coats contain (in wt-%) 20–30%Al, and are generally applied by a diffusion process. Electroplating of Pt on the surface of the alloy precedes the diffusion cycle in the Pt modified β-NiAl bond coats. The diffusion processes enrich the component surface with Al. The composition of a diffusion coating is dictated by thermodynamic and kinetic constraints, and there are limitations on what multicomponent compositions can be deposited in reproducible and cost effective manner. The thickness is typically limited to about 40–60 μm. The β-NiAl coatings consist of an interdiffusion zone typically about the thickness of the top layer sandwiched between the alloy and the β-NiAl layer. A typical microstructure of TBC with a β-NiAl type bond coat and a TGO layer between the bond coat and the YSZ top layer is shown in Fig. 4a .

It must be strongly emphasised that environmental coatings form and reform the TGO layer as the coating is thermally cycled. In TBC applications, TGO spallation leads to TBC spallation. A significant level of understanding has been developed over the last two decades on various stages of TGO formation, growth and adhesion. The effects of major elements (aluminium, chromium and platinum) and minor elements additions (Y, Hf, Zr, Si, Pt and Re) have been studied extensively in bulk alloys and coatings. Of these, additions of small amounts of Hf, Y, Zr, Si and the elements (e.g. Ti, S, Hf and W) diffusing into the coating from the substrate play a critical role.6,37 ^– 59 A thin layer of Pt over a grit blasted NiCoCrAlHfSi coating surface has demonstrated a beneficial effect,44 though, this observation would have limited industrial application due to high cost of Pt. The effect of grit blasting,57 preoxidation58 and post-oxidation60 has been studied. Some of these modifications alter transformation kinetics of transient alumina such as θ-Al₂O₃ to α-Al₂O₃, faster establishment of α-Al₂O₃ being desirable.59 A remarkable effect on initial stages of oxide formation, transformation and growth kinetics has been demonstrated by Gleeson when Y, Hf and Si are added synergistically (codoping) at very low levels (Fig. 13).61 An excellent summary of research on the oxidation mechanism and bond coat compositions is provided by Gleeson.5 Impurities such as S and C in the starting materials and introduced during processing have a critical role on TGO scale adhesion, sometimes leading to infant mortality. The impurity effect will be discussed in section TBC scatter. A recent review on alumina scale formation is given by Heuer et al. 10

OPEN IN VIEWER

Figure 13.

Synergistic beneficial effect of Hf, Si and Y in a β-NiAl coating composition61

Thermal barrier coating operating at higher temperatures will require improved bond coats capable of operating at higher temperatures. This is especially true if TBCs are used as a prime reliant coating. Localised TBC loss due to foreign object impact will probably be difficult to avoid. In this case, TBC loss will expose and rapidly oxidise the bond coat, causing rapid degradation of underlying substrate. However, it must be recognised that bond coat materials and the underlying substrate have similar thermal conductivities, and, therefore, improvement in bond coat temperature capability will be limited by the creep strength of the substrate. It should be pointed out that the extent of coating/substrate interdiffusion will increase as the operating temperature of the turbine increases. An ideal bond coat should have the following attributes:

The newer generation Ni based superalloys contain high levels of rhenium (Re) and relatively higher levels of Al. The higher levels of alloying elements tend to cause microstructural instability. Specifically, Re has been identified as a key alloying element promoting an SRZ layer beneath the coating layer (Fig. 14). The SRZ layer consists of a γ' matrix containing γ and P phase (TCP) needles (as opposed to a γ matrix containing γ' of a typical Ni based superalloy). The SRZ formation has been demonstrated to be detrimental to the creep and rupture properties of the alloys.62 To address the SRZ issue, Rickerby et al. 63,64 and Gleeson et al. 65 have developed lower aluminium containing bond coats chemically compatible with the Ni based superalloy substrate. Rickerby applies a thin layer of Pt followed by a diffusion heat treat to create a Pt modified γ+γ′ coating layer. With investigation of phase equilibria at 1100 and 1150°C in the Ni–Al–Pt system, Hayashi et al. 66 have identified γ+γ′ bond coat compositions that have demonstrated oxidation performance matching the conventional MCrAlX and a β-NiAl coating. To address the high cost of Pt, Gleeson61 has developed coating compositions with lower Pt levels in combination with low level additions of Hf, Y and Si.

OPEN IN VIEWER

Figure 14.

An SRZ is created between the coating and a high Re containing Ni based superalloy

A modified version of a β-NiAl bond coat has been developed by Darolia et al. 67 ^– 71 at GE Aviation using overlay processes such as ion plasma deposition, APS and EB-PVD. The bond coat is predominantly a β phase, and contains (in wt-%) about 20–25Al, 5Cr, 0·25–0·5Zr and/or 0·25–0·5Hf. The bond coat was designed based on the author’s extensive knowledge gained during the development of NiAl based intermetallic alloys for turbine applications.72 ^– 74 A large library of NiAl compositions is available in single crystal form to evaluate oxidation resistance and TBC coated life. Compositions were selected based on the knowledge of strengthening behaviour in single crystal NiAl to develop an oxidation resistant and ‘strong’, creep resistant bond coat. Thermal expansion data were used to match thermal expansion of the coating with the underlying substrate. Reactive element doped NiAl coupons coated with TBC and tested to spallation failure showed that minor additions of Hf or Zr to NiAl provided substantial increases in the TBC spallation life as seen in Fig. 15. The improvement was linked, in part, to increase in the room and high temperature strengths provided by solid solution strengthening by Cr in solution within the β phase matrix of the coating and dispersions of α-Cr particles, and by the β′-Heusler phase Ni₂Al(Hf, Zr) precipitates within the β phase matrix of the coating (Figs. 16– 18). In addition, TGO scale adherence was provided, in part, by Zr or Hf oxide pegging.

OPEN IN VIEWER

Figure 15.

Thermal barrier coating spallation life for TBC coated single crystal NiAl compositions containing small amounts of Zr, Hf and Ti (compositions in at-%) compared with stoichiometric NiAl (50/50)

OPEN IN VIEWER

Figure 16.

Effects of small amount of alloying additions on the <100> room temperature fracture strength of single crystal NiAl compositions

OPEN IN VIEWER

Figure 17.

Improvement in rupture strength (life) at 982°C/124 MPa in the <110> orientation with alloying additions such as Hf, Ti and Ta to a NiAl (50Ni–49·3Al–0·5Hf–0·2Ga) single crystal composition. The compositions are shown in at-%

OPEN IN VIEWER

Figure 18.

Microstructure of a NiAlCrZr bond coat composition showing Cr particle dispersions and β' precipitates that contribute to strengthening of the bond coat

A programme was initiated to translate the improved performance of the TBC coated NiAl single crystal alloys into bond coat compositions. The result was development of an overlay β phase NiAlCrZr bond coat strengthened to prevent rumpling. As an example, dependence of TBC life in a NiAlCrZr bond coating as a function of Zr content is shown in Fig. 19. Thermal expansion match with the substrate is demonstrated as shown in Fig. 20. In addition to increased oxidation resistance and TBC life in laboratory and engine tests, the bond coat has demonstrated reduced wall consumption of the turbine blade. Reduced rumpling resulted in underlying substrate running cooler by maintaining a smooth air flow boundary layer on the surface (Fig. 21). The NiAlCrZr bond coat has been introduced in GE Aviation’s new engines such as GEnx for the Boeing 787 Dreamliner airplane.

OPEN IN VIEWER

Figure 19.

Thermal barrier coating spallation life results from furnace thermal cycle testing of NiAlCrZr coatings produced by EB-PVD method as a function of Zr content68

OPEN IN VIEWER

Figure 20.

Coefficient of Thermal Expansion match between a second generation single crystal Ni based superalloy (Rene′ N5) and NiAlCrZr bond coat (courtesy Professor Kevin Hemker, JHU, unpublished work)

OPEN IN VIEWER

Figure 21.

Engine test results with NiAlCrZr bond coat compared with NiAlPt bond coat. a NiAlPt bond coat and b NiAlCrZr bond coat. Bond coat rumpling seen in the NiAlPt bond coat c is absent in the NiAlCrZr bond coat d. Lower oxidation in the NiAlCrZr bond coat and retention of γ+γ' structure in the underlying superalloy was also observed

Diffusion barriers

Owing to increased interdiffusion expected in higher temperature applications, and to avoid SRZ formation, extensive development effort has been made to identify diffusion barrier layers applied between the bond coat and the substrate to block interdiffusion. Multilayer bond coats incorporating a combination of diffusion and overlay processes have been evaluated. Diffusion barrier materials have included metallic (i.e. Re and Ru), intermetallic (i.e. sigma and laves phases) and oxides.75 ^– 78 Majority of the work has been on Ru containing layers. Ru and Re are demonstrated to be effective barriers, but possess poor oxidation resistance. Alloying to adequately impart oxidation resistance to Ru and Re has not been demonstrated. Oxide barriers have durability issues during thermal cycling. So far, a diffusion barrier layer that meets all of the critical requirements, namely, good adhesion (due to thermal expansion difference), oxidation resistance (or does not oxidise such as an oxide layer), TBC durability during thermal cycling and practical and cost effective processing is yet to be developed, and continues to challenge materials scientists.

Yttria stabilised zirconia composition modifications including codoping with other stabilisers

Incorporation of Y₂O₃ to the ZrO₂ fluorite crystal structure creates a defect structure (solute Y atoms and oxygen vacancies). In baseline YSZ, one oxygen vacancy is created for every two yttrium cations. The amount and stability of this defect structure and the porosity created during deposition control the as deposited and exposed thermal conductivity acting as phonon scattering centres. In baseline YSZ, the easiest approach would be to vary the Y₂O₃ content.

Substitution and addition of ternary and high order dopants of metal (M) oxides of the form MO, MO₂ or M₂O₃ with varying ionic radius or atomic mass have been evaluated while maintaining the t' phase. There are over 50 issued patents related to low conductivity with subtle differences in compositions and processing. Developments with the most relevance, uniqueness and impact are discussed next.

Numerous codoped YSZ compositions have shown lower thermal conductivity. One of first research to lower thermal conductivity conducted with ceria stabilised zirconia (CeSZ) as an alternate to YSZ or added to YSZ demonstrated lower thermal conductivity.92 Processing due to vapour pressure differences and inferior erosion resistance has been major issue on the ceria containing TBC. In spite of this limitation, development in this area is continuing.87 Rickerby and Tamarin,93 Nicholls et al.,94,95 Rigney and Darolia96 and Schulz et al. 97 have investigated dopants such as NiO, MgO, CaO, Nd₂O₃, Gd₂O₃, Er₂O₃,Yb₂O₃, Dy₂O₃, CeO₂, SrO, Sc₂O₃, Eu₂O₃, Fe₂O₃, In₂O, Sm₂O₃, Ho₂O₃, HfO₂ and Ta₂O₅ with various combinations and levels. Majority of these dopants have shown to reduce thermal conductivity from about 25 up to 40%. A clear trend, however, of the effects of ionic radius and/or mass of the dopants is difficult to establish due to scatter in data, mainly due to processing difficulties, microstructural differences in the as deposited coatings and errors associated with conductivity measurement techniques.

An interesting ‘multicomponent defect cluster’ concept investigated by Zhu and Miller98 ^– 101 has also shown to lower thermal conductivity. Thermal barrier coatings with oxides clusters of ZrO₂–Y₂O₃–Nd₂O₃–Yb₂O₃ (Gd₂O₃, Sm₂O₃ and Sc₂O₃) have been investigated on both plasma sprayed and EB-PVD coatings. In one study, the effect of the cluster dopant type with compositions ranging from YSZ only, YSZ plus a single Nd₂O₃ or Yb₂O₃ dopant, YSZ plus both the Nd₂O₃ and Yb₂O₃ in varying relative concentrations was investigated. It was observed that ZrO₂–Y₂O₃ with paired dopant additions (Nd₂O₃+Yb₂O₃) had lower thermal conductivity than those of YSZ, or YSZ with a single cluster dopant, Nd or Yb. The coatings with equal amount of cluster dopants (Yb₂O₃/Nd₂O₃ = 1) often showed the lowest conductivity at a given total dopant concentration. Optimum dopant concentration was established at ∼10 mol.-%. The results of their findings are summarised in Fig. 23. The TEM observations indicated nanoscale clustering of the smaller and larger cations in different regions, with Y being uniformly distributed. It has been postulated that the clusters contribution to the phonon scattering is responsible for larger reductions in conductivity than those found with Y alone or codoped with only one of the cations at comparable total dopant concentration. Higher thermal stability and TBC thermal cycle life comparable to that obtained with YSZ TBC have been shown.

OPEN IN VIEWER

Figure 23.

Thermal conductivity of various oxide cluster TBCs as a function of total dopant concentration after 5 or 20 h laser high heat flux tests at 1316°C (Zhu and Miller98 ^– 101)

Yttria stabilised zirconia with HfO₂ additions

Yttria stabilised zirconia compositions containing hafnia have also been evaluated by Peters et al.,4 Zhu and Miller,102 Gorman et al.,103 Singh et al.104 and Matsumoto et al.,105 since the crystal lattice of zirconia and hafnia is isomor phous, and a complete solubility exists. Yttria stabilised zirconia ingots generally contain about 1–2 wt-% hafnia. Larger additions of hafnia (e.g. 40 wt-% zirconia+40 wt-% hafnia+20 wt-% yttria) further reduce thermal conductivity but the largest effect of 30% reduction at high temperature was reported for zirconia free 27 wt-% yttria stabilised hafnia.104 The latter showed a much denser and fine columnar microstructure and was less susceptible to sintering. Similar favourable lower shrinkage rates have been found for EB-PVD 7·5 wt-% yttria–hafnia samples that were not rotated during deposition. Peters et al. 4 have demonstrated similar results with 32 wt-% yttria–hafnia without notable processing problems with nearly the same composition in both ingot and coating.

Pyrochlore type (A₂B₂O₇) structures

Unlike the t' YSZ based compositions, a new class of low conductivity compositions based on the pyrochlore type zirconate structure [(Gd, Eu, Sm, Nd, La)₂Zr₂O₇] have emerged, of which Gd₂Zr₂O₇, La₂Zr₂O₇ and Sm₂Zr₂O₇ have been studied the most.106 ^– 110 Conductivity reductions up to 50% have been reported. In addition to lower conductivity, Gd₂Zr₂O₇ TBC has been claimed to provide calcium–magnesium–alumino-silicate (CMAS) attack mitigation,111 though, independent verification and understanding of this benefit is yet to be conducted. Several other beneficial attributes include phase stability up to at least 1500°C,16 and sluggish sintering kinetics.109,112 Some of these compositions are rather difficult to deposit, especially by EB-PVD, uniformity of composition and microstructure being the major issue.109 Owing to chemical incompatibility with TGO, a thin, about ∼25 to 50 μm, underlayer of YSZ is necessary113 to negate reduced TBC life times observed with a just a pyrochlore type zirconate top coat layer. Such underlayers have been reported for majority of ‘low κ TBC’ compositions.4,114 Thermochemical compatibility issues between alumina and ZrO₂–Gd₂O₃ are discussed by Leckie et al. 115 and an assessment of the thermodynamic parameters in the ZrO₂–Y₂O₃–Al₂O₃ system is provided by Fabrichnaya and Aldinger.116 Significant conductivity reductions have been reported by Vassen et al. 109 on zirconate variations such as La_1·7Dy_0·3Zr₂O₇, La_1·4Gd_0·6Zr₂O₇, La_1·4Eu_0·6Zr₂O₇ and La_1·4Nd_0·6Zr₂O₇ applied by APS.

To negate conductivity increase during service, Darolia et al. 117,118 have evaluated combinations of one or more dopant oxides for conductivity reduction along with carbides such as YbC₂, NdC₂ and LaC₂ of the dopant metals which are either evaporated from a carbide ingot source or formed by reaction during or after the EB-PVD processing. In another approach, insoluble alumina by Rigney and Darolia,119 or alumina combined with lanthanum oxide, chromium oxide and/or yttrium chromate by Ackerman et al. 120 have been evaluated. Insoluble particles or dispersions in the TBC structure provide additional sites for phonon scattering as well as stabilise the structure during subsequent exposure.

New ceramic compositions

Other suggested compounds121,122 with very limited conductivity data as a coating are: hafnates such as La₂Hf₂O₇, monazites such as LaPO₄, magnetoplumbite structures such as LaMgAl₁₁O₁₉ and LaLiAl₁₁O_18·5, garnets such as Y3AlxFe5−xO12, Y₃Fe₅O₁₂, W₃Nb₁₂O₄₄, mullite, TiO₂, ZrSiO₄, ZrTiO₄, perovskite structures such as SrZrO₃, BaZrO₃ and rare earth aluminates such as RE₂SrAl₂O₇.122 Of these compounds, Ca_xMg_1–xZr₄(PO₄)₆ (abbreviated as CMZP) has been reported with a conductivity of 0·7 W m⁻¹ K⁻¹ at 1000°C.123 Perovskites (ABO₃) can accommodate a wide variety of different ions in solid solution including ions with large atomic mass. Some of these compounds are stable to very high temperatures, and have the potential to be developed as future low conductivity materials. Processing as a coating with controlled uniform composition, however, will be a serious challenge.

An excellent summary of reported thermal conductivity data is provided by Levi,1 and reproduced in Fig. 24. There appears to be a minimum value obtainable around 610 wt-% of the total stabiliser additions. Thermal barrier coating life as determined in furnace thermal cycle tests superimposed in the ZrO₂–Y₂O₃ phase diagram by Levi is shown in Fig. 25. It is interesting to note that the TBC life has a maximum value around 6–10 wt-% of the total stabiliser additions (irrespective of the type of stabiliser) to ZrO₂. Non-transformable tetragonal (t') phase is the predominant phase in this composition range.

OPEN IN VIEWER

Figure 24.

Comparative summary of thermal conductivity values for EB-PVD coatings reported in the literature for a variety of zirconia based materials (Levi1)

OPEN IN VIEWER

Figure 25.

A line representative of the cyclic durability of TBC is superimposed on a binary phase diagram for the ZrO₂–YO_1·5 showing phases expected (courtesy of Levi, UCSB)

Processing modifications

Processing modifications to incorporate porosity is the easiest approach to reduce the thermal conductivity of YSZ. However, the porosity reduction is only effective if the porosity is retained via improved sintering resistance. For EB-PVD, in addition to porosity within the columns, their feathery structure and gaps between the columns play a significant role. For APS, splat boundaries which are essentially perpendicular to the heat flow cause conductivity reductions as low as 50% of fully dense YSZ in contrast to ∼30% for EB-PVD YSZ. The EB-PVD processing parameters that have been demonstrated for microstructure modifications include the following:

It must be pointed out that high temperature exposure during service will densify the coating structure via pore coarsening and sintering, negating the porosity effect. Thermal barrier coating durability for majority of new TBC compositions under thermal cycling is reduced and the erosion rate is increased. The decreased performance has been attributed to lower toughness rendering the system susceptible to delamination internally in the TBC upon thermal cycling.8 Non-compatibility with TGO is also a major contributor.

Temperature and thermal cycle dependent failure

Not surprisingly, engine thermal cycles (e.g. temperature, dwell time, rates of excursions, gradients and cycling frequency) play a significant part in TBC failures. The oxidation of the bond coat results in the formation and growth of the TGO that induces stresses and displacements at the TGO/TBC/bond coat interfaces eventually leading to TBC spallation. Extensive research has been carried out to establish the various mechanisms and factors that control TBC spallation in both EB-PVD and APS TBC.124 ^– 144 In spite of over 50 publications on this subject, arguments still persist. The relative roles of plasticity of TGO and the bond coat as well as phase transformations in the bond coat due to Al depletion are still being debated. During engine operation, several interrelated thermal cycle dependent phenomena take place. The wide variation of engine operating conditions is one of the reasons why TBC failure mechanisms are not completely understood and agreed upon. Additionally, TBC in service is a very dynamic system with continuously changing composition, crystalline phases and microstructures. These changes lead to changes in physical and mechanical properties, so creating models becomes rather difficult. The other complicating factor is that current TBC systems show a wide distribution in life, with a significant proportion failing at much earlier times primarily due to processing variations, and sometimes lack of adequate process control. Additionally, the bond coat surface gets contaminated with a minute amount of impurities such as S, originating from the raw materials or by diffusion from the underlying substrate. The incidence of lower life TBC is avoidable with better raw material control and cleaner heat treat furnaces. Laboratory furnace and rig tests used as screening tests, unfortunately, are not generally simulative of engine operating conditions that are difficult to duplicate in laboratory settings.

A brief summary of our understanding of the factors that control thermal cycle dependent TBC spall in case of EB-PVD TBC with a β phase NiAlPt bond coat follows. The failure mechanisms for the EB-PVD TBC with MCrAlX bond coat and APS TBC differ somewhat,132 and will not be elaborated in this paper. The TGO thickens with time at operating temperature resulting in a constrained volume expansion that leads to inplane compressive ‘growth’ stresses. Additionally, upon cooling, the thermal expansion mismatch with the substrate leads to very high ‘thermal’ compressive residual stresses in the TGO reaching about 3–6 GPa at ambient temperature. During thermal cycling, TGO seeks to relieve compression, by means of out of plane displacements, preferentially into the bond coat since it is relatively soft at high temperature, and prone to creep deformation. During repeated thermal cycling, progressive roughening of the bond coat/TGO/top coat interfaces occurs due to cyclic creep of the bond coat as shown in Fig. 27. Such roughening/undulations are often termed ‘ratcheting’, ‘rumpling’ or ‘bucking’, though, arguments have persisted on the exact terminology. Once the stored energies at the TGO/TBC/bond coat interfaces exceed a threshold value, cracks and separation occur. Initial surface imperfections and rough bond coat surfaces are considered to be contributing to the out of plane stresses, and are the origins of initial displacements and cracks. Large scale buckling of TBC and spallation is preceded by smaller cracks that extend and grow either at the TGO/bond coat or TGO/TBC interface, though, separation within the TGO, and within the TBC at the TGO/TBC interface is often observed.

OPEN IN VIEWER

Figure 27.

Thermally grown oxide undulations referred to as rumpling or ratcheting in a thermally cycled TBC (YSZ top coat and NiAlPt bond coat)

The growth of the TGO and interdiffusion between the bond coat and the substrate results in the depletion of Al in the bond coat. The Al depletion can promote the formation of other oxides, such as Ni containing spinels, that accelerates localised oxidation by providing fast oxygen diffusion paths. Al depletion also leads to phase transformations such as β to γ' and/or martensite formation (for a β NiAlPt bond coat). Large stresses generated due to transformations in the bond coat have been suggested as contributing factors in promoting undulation growth. Thus, thermal cycling combined with TGO stresses, strain misfit (due to thermal expansion and phase transformations), bond coat creep deformation and initial imperfections leads to TBC failure. Loss of Al by oxidation and by interdiffusion between the bond coat and the substrate also plays a significant role by changing the bond coat properties, and sometimes, creating voids at the bond coat/substrate interface. Several of the concurrent mechanisms operating related to use temperature and thermal cycling during engine service are schematically depicted in Fig. 28 (reproduced from Clarke and Levi2). During the early days of TBC application, TGO thickness of ∼6 μm was considered in industry to be maximum allowable thickness to avoid TBC spallation. Thermally grown oxide thickness up to 17 μm has been observed before failure in the β-NiAl bond coat by Darolia et al. 67 ^– 71 at GE Aviation. The concept of critical TGO thickness is definitely not a very useful guide since the critical TGO thickness strongly depends on bond coat composition, impurities, microstructure and thermal cycling history.

OPEN IN VIEWER

Figure 28.

Schematic summary of the several of the concurrent processes occurring in the bond coat, TGO and TBC during use at high temperatures (Clarke and Levi2)

The above scenario suggests the following guidelines for longer TBC life:

The other temperature dependent TBC degradation is sintering and densification of the porous structure of the top coat resulting in increased thermal conductivity and decreased stain tolerance. Phase destabilisation or changes are not currently limiting issues since TBC operating temperatures are below the threshold of transformation, though, it is becoming an issue for the power generation turbines that run for extended periods of time (e.g. 20 000+ h), metastable phase t' being transformed to the monoclinic phase.

Erosion and impact damage

Thermal barrier coating compaction, plastic deformation, spallation or gradual thinning by particulates in the gas turbine environments generally termed ‘extrinsic failures’ have limited full TBC utilisation. The damage is assumed to be primarily dependent on the microstructure and properties of the top coat. It is quite likely that the TBC/TGO and TGO/bond coat interface properties also play a significant role. For a given TBC system, impinging particle size, mass, velocity, temperature, rotation speed of the component, impact location and incidence angles relative to the specific locations in the component are deciding factors to differentiate between the impact or erosion damage. Particles of various sizes impinge turbine blades at a variety of angles. The leading edge of the blade could experience a 90° impingement whereas the other locations are subjected to particle impingement at lower angles. The impingement angles to the TBC coated turbine combustor liners are about 20° or lower. It has been demonstrated in the laboratory tests as well as during service that the EB-PVD TBC is generally 7–10 times more erosion resistant than the APS TBC in majority of the particle impingement conditions.145 A typical particle size distribution in a dust collected from a high pressure turbine blade is shown in Fig. 29. Particle size ranges from 10 to 100 μm. Sand, ash and dirt ingested are typically smaller in size whereas debris from upstream engine components such as combustor can be of much larger size, and their frequency, fortunately, is low. Different size particles have definite trajectory in the hot flow path rendering impact and erosion damage combined with particle size location specific. Turbine blade locations that are prone to either impact or erosion damage are shown in Fig. 30.

OPEN IN VIEWER

Figure 29.

Particle size distribution in a turbine environment

OPEN IN VIEWER

Figure 30.

Locations in an HPT blade prone to impact (locations 1 and 2), erosion (locations 2 and 4) and CMAS (location 5). No TBC degradation at location 3

Impacts from ‘large’ particles can cause plastic deformation, kink bands, compaction and brittle fracture of the columns and complete spallation of TBC. Examples of impact damage are given in Figs. 31 and 32. Areas of impact, as noted by compressed and fractured PVD TBC columns, can be seen in Fig. 32, and in some areas shear fractures propagating from the impact site to the bond coat/alumina interface are evident. This fracture, upon reaching the interface, can then propagate along the interface and result in a local spall with an appearance similar to typical thermal cycle spallation. Terminology for this type of damage as ‘foreign object damage’ is misleading since debris from TBC spalled from combustor liners often lands on the blades. An example is shown in Fig. 33. The impact event is unpredictable; however, from field experience, the leading edges on high pressure blades have been found to be susceptible to impact related TBC spallation because of high rotation speed, sharp curvature, relatively higher thermal stresses and their exposure to high velocity particulates. In certain applications, experience has shown that impact related spallation or erosion extends further onto the convex side of the blades. Thermal barrier coating loss at these locations must be accounted for in the design of blades to avoid local hotspots that can accelerate degradation of the underlying bond coat and the superalloy blade. Additional cooling air is required in such locations negatively affecting engine efficiency.

OPEN IN VIEWER

Figure 31.

Electron beam physical vapour deposition TBC delamination at a leading edge of an HPT blade by particle impact

OPEN IN VIEWER

Figure 32.

Electron beam physical vapour deposition TBC deformation, fracture and delamination in an HPT blade by particle impact

OPEN IN VIEWER

Figure 33.

Air plasma spray TBC spalled from combustor can cause impact and erosion damage in EB-PVD TBC applied to a turbine blade

Typical examples of TBC erosion indicated by gradual thinning of TBC are shown in Fig. 34. Erosion/thinning of TBC generally occurs on the pressure side pocket, and on the suction side near the leading edge of the blade. The design of the blade can account for TBC thinning (e.g. thicker TBC) on blade locations susceptible to erosion on the basis of field experience.

OPEN IN VIEWER

Figure 34.

Typical EB-PVD TBC erosion locations and microstructural details of TBC thinning in blade locations indicated by circles

Development of particle impact resistant TBC continues to be a major development activity for a ‘prime reliant’ TBC. The activities have focused on: capability to reproduce the damage modes observed in service in controlled laboratory tests; understanding of the mechanisms of various forms of TBC damage caused by the impinging particles; and improvements in impact and erosion resistance by TBC composition and microstructure modifications.

In order to assess impact resistance under conditions simulative of turbine blades in service, tests developed by Bruce146 at GE Aviation and Wellman et al. 147 ^– 152 at Cranfied University have evaluated erosion rates and impact events with different sizes and types of particles introduced into the gas stream of a combustor burner rig at high temperatures and high velocities. Progress towards a mechanistic understanding has been limited by the absence of well controlled experiments capable of duplicating the conditions expected in turbine engines. The challenges are associated with the high temperatures (typically 1100°C) and high impact velocities (up to 300 m s⁻¹), as well as the relatively small particles involved (∼20 to 500 μm) and their compositions (usually calcium–alumino-silicate: CMAS). Currently, there are no tests capable of single particle impacts that can reasonably reproduce engine operating environments.

Evans et al. 153 ^– 155 have analysed the erosion and impact events on field returned blades and laboratory tested specimens. The effort was to establish trends of the material removal rates with the properties of the columnar microstructure EB-PVD TBC. A limitation for creating such models is that the size and velocity of the impacting particles responsible for specific damage sites are unknown. The exact temperature of the component surface during particle impingement is also unknown whenever engine hardware is used for analysis. Other complication arises from the fact that it is unknown whether the damage comprises a single event or an accumulation of multiple events of various energies, though, erosion involves a sequence of nanosecond impingement and TBC thinning.

The analysis has postulated two major damage modes, erosion mode and impact mode, with an intermediate mode in between as described in next sections on ‘Erosion mode’ and on ‘Impact mode’. The transition between the two modes depends on impact velocity, part rotation speed, impact angle, particle size, temperature, contact area relative to the column diameter as well as the TBC constituent properties and microstructure.

Erosion mode

For low kinetic energy mainly with small particles, during initial impact, to accommodate the projectile as it penetrates, elastic stress waves that are transmitted down the columns, can give fracture at tops of columns, at mid-depth and at TBC/TGO interface. For example, induced elastic bending waves can cause preexisting flaws at the column perimeter (e.g. column feathers) to form cracks that extend across the columns, beneath the surface, resulting in an array of column sized cracks (Fig. 35). Once the cracks link, small amounts of material are removed. Elastic waves also reflect off the bottom of the columns becoming tensile waves that propagate back to the surface. These waves could also cause cracks to form and extend across the columns.153,154

OPEN IN VIEWER

Figure 35.

Particle erosion mechanism of crack initiation and propagation with small kinetic energy mainly with small particles

Impact mode

With larger particles, combinations of high kinetic energy and high temperature cause YSZ to be susceptible to large scale plastic deformation and densification around the contact site (Fig. 32a ). Outside the densified zone, kink bands form and extend diagonally downward, toward the TGO interface. Such bands have been identified at a variety of different angles to the interface (Fig. 32b ). Within the bands, the columns are plastically bent, causing the boundaries to crack, weakening the material (Fig. 32c ). In some cases, the bands reach the TGO interface. When this happens, they nucleate a delamination that extends outward from the impact site, along a trajectory within the TBC, just above the TGO. Such delaminations provide a mechanism for creating large scale spalls (Fig. 32d ).

Mechanism maps153 for the onset of material removal by particle impact with the plane indicating the transition from elastic to plastic response superposed are shown in Fig. 36. Evans et al. 153 analysis for initiation and threshold has concluded that increasing the TBC toughness should have the most pervasive influence, through its role in elevating the threshold and, in some cases, decreasing the removal rate (Fig. 35). However, increased toughness may not provide the desired benefit since TBC is quite ductile at temperatures >900°C.156 In fact, doped YSZ with increased toughness, to be discussed next, have not proven to be more resistant relative to baseline compositions in laboratory testing.157 The models also imply that softer materials (at high temperature) should have a substantially higher threshold as well as a slightly lower erosion rate above the threshold. A systematic assessment of the deformation mechanisms and of trends in yield strength with composition and temperature would be beneficial. Additionally, reducing the diameter of the columns should be beneficial in affecting material removal in the elastic range. Designed experiments, preferably using single particle impacts, and further analysis of field returned hardware are required to further aid our understanding.

OPEN IN VIEWER

Figure 36.

Mechanism maps proposed by Evans et al. 153 for the onset of TBC removal by particle impact with the plane indicating the transition from elastic to plastic response superposed

Approaches for improved impact and erosion resistant TBC

Composition modifications and composite toughening

Low κ compositions

Though lacking in ‘desired’ erosion and impact resistance, YSZ has demonstrated, at least in the laboratory tests with the newer ‘low κ’ TBC compositions, superior particle erosion and impact resistance. Over 200 combinations of compositions and microstructural variations have been tested at GE Aviation under various test conditions to find a low conductivity TBC composition or microstructure with erosion and impact resistance equal to or better than that of baseline YSZ. This effort has generally been unsuccessful except in a couple of cases (composite/layered and modulated structured TBC, discussed later). Wellman et al. 149 have erosion tested ten different TBC compositions under various test conditions including at room temperature and high temperatures. The erosion behaviour of several of the tested low conductivity compositions of doped and codoped YSZ was compared with that of baseline YSZ. Their findings are: all the EB-PVD TBC had an erosion rate lower than that of the APS TBC; aging during high temperature exposure decreases impact and erosion resistance, and becomes more pronounced as the time and/or temperature of aging increases; YSZ is superior to all the ‘low κ compositions’ investigated; and thinner inclined columns are more erosion resistant. Dopant additions to YSZ increase the erosion rate of EB-PVD YSZ significantly as shown in Fig. 37 (reproduced from Ref. 147); this increase is then further magnified when the samples are aged before erosion testing. Engine experiences with the doped YSZ, if any, are considered proprietary, and not available in the open literature.

OPEN IN VIEWER

Figure 37.

Erosion rates of YSZ modified with Dy₂O₃ and Gd₂O₃ (reproduced from Wellman and Nicholls147)

Fracture toughness improvement

The superior erosion and impact behaviour of YSZ is likely related to its higher fracture toughness as shown in Fig. 38 (reproduced from Ref. 8) that compares fracture toughness of zirconia coatings containing different Y₂O₃ levels. It is interesting to look at fracture toughness data (circa 1983) of YSZ in the single crystal and polycrystalline forms. Fracture toughness was found to be maximum around 6 wt-%Y₂O₃ as shown in Fig. 10.21 Impact and erosion test data shown in Fig. 39 obtained at GE Aviation158 on TBC coatings show a remarkable resemblance to the fracture toughness plot in Fig. 10. Research activities, therefore, have sought to identify TBC compositions with fracture toughness higher than that of YSZ. A limited number of oxide dopants, generally with tetravalent (Ti⁴⁺) and pentavalent (Ta⁵⁺) cations can increase tetragonality of zirconia and significantly increase fracture toughness as reported by Kim159,160 by incorporating small amounts of Ta₂O₅. Greater than twofold increase in toughness at room temperature with substitution of Ti⁴⁺ for the larger Zr⁴⁺ cation into single phase tetragonal (t′) YSZ increasing the tetragonality of the unit cell has been demonstrated by Schaedler et al. 161,162 Ferroelastic toughening is postulated to be the underlying mechanism for the increase. As mentioned earlier, laboratory testing with these new compositions has not demonstrated any improvement. Fracture toughness measurements at elevated temperatures (about 1100–1200°C) are essential to further elucidate the role of fracture toughness.

OPEN IN VIEWER

Figure 38.

Fracture toughness of 7 wt-%YSZ and 20 wt-%YSZ coating.8 Also shown data for parially stabilised zirconia in the sintered ‘bulk’ form (TZP)

OPEN IN VIEWER

Figure 39.

Impact resistance of YSZ coatings containing various levels of Y₂O₃.158 The fracture toughness data superimposed from Ref. 21

Composite toughening

By incorporating dispersion of ceramic particles such as alumina or chromia in the YSZ microstructure to explore composite toughening, Darolia and Rigney163,164 have demonstrated a 50% improvement in impact and erosion resistance with alumina dispersion. In another approach by Rigney and Darolia,165 the top layer consists of layers of particle free YSZ alternating with layers of YSZ with dispersions of alumina particles.

TGO/bond coat interface toughening

Analytical models have suggested that toughening the bond with the TGO (i.e. interface) should provide benefit in preventing interfacial delamination and complete loss of TBC. Interface strength/toughness cannot be independently controlled. Bond coat composition dependent interfacial chemistry, in particular segregation of impurities such as S, bond has a strong role. Eberl et al. 166 have initiated a difficult task of the measurement of the interfacial properties while Smith et al.167 ^– 169 have conducted first principles computations of the interfacial adhesion and the role of S and Hf on the interfacial strength: S was calculated to be detrimental and Hf was calculated to be beneficial. Hf was also shown to negate S effect. If S escapes Hf bulk pinning, Hf can mitigate detrimental S effects at the interface. Experimentally, Hou and Priimak53 and Molins et al. 170 have measured S segregation at the TGO/bond coat interface as well as at voids in a NiAl based bond coat composition (44Al–11Pt–37Ni in at-%). S as high as 11 at-% was measured after 50 h at 1100°C as shown in Fig. 40. Hou and Priimak53 have measured interface fracture strength reduction shown in Fig. 41 from ∼80 to ∼10 MPa with 2·5 at-%S at the interface. Experimentally, a large body of literature data exists46,47,51,53 on the detrimental effect of S, and the beneficial effects of reactive elements such as Hf and Y. A fairly recent review of the segregation phenomenon is given by Hou.171 Improved TBC life has been demonstrated by Darolia and Walston172 in Hf containing Ni based superalloys and bond coat compositions.

OPEN IN VIEWER

Figure 40.

S segregation at the TGO/bond coat interface in a NiAl based bond coat composition: 44Al–11Pt–37Ni in at-%170

OPEN IN VIEWER

Figure 41.

Effect of S on interface fracture strength53

Microstructure modifications

Porosity effect

Porosity effects on performance are not clearly understood due to the fact that porosity levels cannot be independently controlled without changing other subtle microstructural features of porosity such as pore diameter and TBC column characteristics such as column diameter, gaps and the feathery structure. Additionally, these features vary through the thickness of the top coat. Coating chamber pressure, deposition rate, deposition temperature, deposition angle, substrate temperature, surface roughness and rotation speed are some of the EB-PVD parameters that influence porosity. Microstructural modifications have to ensure strain compliance of the top coat; maintain a ‘desired’ thermal conductivity; and minimum susceptibility to microstructural changes during high temperature exposure. Experimentally, denser microstructures have demonstrated inferior erosion resistance.147 Porous coatings absorb the impact much more effectively than dense coatings. On the other hand, increased porosity/softness could have inferior erosion performance. Rapid removal and/or compaction leading to impact mode of damage are the issues. For example, ceria doped YSZ deposits as an extremely soft, friable coating with poor room temperature erosion resistance.92

Other approaches such as smoother TBC surface and dense top layers including metallic layers such as Pt would have limited, short term value in actual engine operating environment, roughening of the surface and spallation of dense layers being the underlying issues.

Modulated/wavy TBC structures

A recent breakthrough observation by Darolia et al. 173 from the evaluation of field returned blades showed that a ‘modulated’ (as compared to straight TBC columns) TBC structure can better withstand impact damage as shown in Fig. 42. Evaluation of these blades has shown that the layers between each zone of orientation act as sites for deflection of the impact stress, resulting in removal of only the outer layer(s) rather than the entire coating. In this way, a single layer (perhaps less than 25 μm depending on wave design) is sacrificed rather than the entire coating thickness.

OPEN IN VIEWER

Figure 42.

Observations on field returned HPT blades from the same engine service show that modulated TBC structure is resistant to particle impact damage

Strategy to mitigate CMAS damage

The engine components have been designed with TBC surface temperatures not exceeding 1240°C until a materials solution to mitigate CMAS damage is available. To address this critical issue that limits maximum use temperature, the CMAS mitigation strategy has consisted of depositing a protective coating on the surface of TBC or incorporating sacrificial materials, mostly oxides, in the strain compliant spaces of the coatings. The sacrificial coating reacts with CMAS to increase the melting temperature or viscosity of the contaminants thereby inhibiting infiltration. One of the examples is alumina either as a top layer or codeposited with YSZ.178 ^– 181 The other approach is to deposit a dense, non-cracked and non-porous ceramic or metal impermeable or non-wetting outer layer to inhibit the infiltration of molten CMAS. Examples include Pd, Pt, Pd–Ag and ceramics such AlN and BN.

Kramer et al. 182 have shown that dissolution of gadolinium zirconate, Gd₂Zr₂O₇, into the CMAS melt results in a mixture of crystalline phases, an apatite phase based on Gd₈Ca₂(SiO4)₆O₂ and the fluorite ZrO₂ phase with Gd and Ca in a solid solution, that fills the flow channels, and essentially creates a dense, impervious layer/zone that prevents or significantly slows down further penetration and reaction as shown in Fig. 50. The extent of penetration depends on the relative competitiveness of the infiltration, dissolution and crystallisation kinetics, all of which are dependent on the temperature and the compositions of the melt and the TBC top coat. After an early sealing of the flow channels, reaction continues slowly. The progressive slowing of the attack should be beneficial to the survivability of the reaction layer under thermal cycling conditions. These promising results with gadolinium zirconate have led to evaluation of other rare earth zirconates such as Nd₂Zr₂O₇ and YSZ containing various amounts of Nd₂O₃. The results have shown their effectiveness in mitigating CMAS penetration by the rapid formation of a dense reaction layer in the intercolumnar gaps and the intracolumnar porosity of an EB-PVD TBC.183 As a further improvement of this approach, Darolia et al. 184 have proposed a two layer system consisting of an innerlayer of YSZ and an outer layer of a pyrochlore structure rare earth zirconate. Yttria stabilised zirconia doped with rare earth oxides of Gd, La, Eu, Sm and Nd have also been proposed as the outer layer.

OPEN IN VIEWER

Figure 50.

Dissolution of gadolinium zirconate (Gd₂Zr₂O₇) into the CMAS melt results creates a dense, impervious layer/zone that prevents or significantly slows down further penetration and reaction178

Aygun et al. 185 and Drexler et al. 186 have demonstrated that incorporation of 20 mol.-%Al₂O₃ and 5 mol.-%TiO₂ in the form of a solid solution into YSZ (referred to as YSZ+20Al+5Ti) coatings deposited by APS results in resistance to CMAS attack while conventional APS YSZ was fully penetrated and destroyed by molten CMAS (at 1200°C), under the same conditions. The molten CMAS was found to penetrate only about a third of the YSZ+20Al+5Ti TBC thickness before it was arrested due to the formation of crystalline anorthite. It was also demonstrated that the APS Gd₂Zr₂O₇ TBC is resistant to attack by molten lignite fly ash which penetrated only ∼25% of the TBC thickness due to the formation of an impervious, stable crystalline structure that arrested the penetrating molten fly ash.

Darolia and Fu187 have demonstrated that a top layer of Y₂O₃ or mixtures of YSZ and Y₂O₃ over a YSZ underlayer reacts with CMAS to produce a yttrium–calcium–silicate phase having a needle-like geometry that is very dense and resistant to infiltration of molten CMAS. These coatings were deposited by EB-PVD without any difficulty due to vapour pressure similarity between ZrO₂ and Y₂O₃.