lable at ScienceDirect

Acta Materialia 106 (2016) 1e14

Contents lists avai

Acta Materialia

journal homepage: www.elsevier .com/locate/actamat

Full length article

Response of ytterbium disilicateesilicon environmental barriercoatings to thermal cycling in water vapor

Bradley T. Richards a, Kelly A. Young a, Foucault de Francqueville b, Stephen Sehr b,Matthew R. Begley b, Haydn N.G. Wadley a, *

a Department of Materials Science and Engineering, University of Virginia, Charlottesville, VA 22903, USAb Departments of Mechanical Engineering and Materials, University of California, Santa Barbara, Santa Barbara, CA 93106, USA

a r t i c l e i n f o

Article history:Received 19 October 2015Received in revised form22 December 2015Accepted 30 December 2015Available online 8 January 2016

Keywords:Environmental barrier coatingsThermal spray depositionYtterbium silicatesSilicon steam volatility

* Corresponding author.E-mail address: haydn@virginia.edu (H.N.G. Wadle

http://dx.doi.org/10.1016/j.actamat.2015.12.0531359-6454/© 2016 Acta Materialia Inc. Published by E

a b s t r a c t

A preliminary study of a promising bi-layer environmental barrier coating (EBC) designed to reduce thesusceptibility of SiC composites to hot water vapor erosion is reported. The EBC system consisted of asilicon bond coat and a pore-free ytterbium disilicate (YbDS; Yb2Si2O7) topcoat. Both layers weredeposited on a-SiC substrates using a recently optimized air plasma spray method. The two layers of thecoating system had coefficients of thermal expansion (CTE) that were well matched to that of the sub-strate, while the YbDS has been reported to have a moderate resistance to silicon hydroxide vaporforming reactions in water vapor rich environments. Thermal cycling experiments were conducted be-tween 110 �C and 1316 �C in a flowing 90% H2O/10% O2 atmospheric pressure environment, and resultedin the formation of a thermally grown (silica) oxide (TGO) at the silicon-ytterbium disilicate interface.The TGO layer exhibited linear oxidation kinetics consistent with oxidizer diffusion through the ytter-bium silicate layer controlling its thickening rate. The effective diffusion coefficient of the oxidizingspecies in the YbDS layer was estimated to be 2 � 10�12 m2s�1 at 1316 �C. Slow steam volatilization of theYbDS topcoat resulted in the formation of a thin, partially protective, high CTE ytterbium monosilicatelayer on the outside of the YbDS coating. Progressive edge delamination of the coating system wasobserved with steam exposure time, consistent with water vapor volatilization of the TGO edges thatwere directly exposed to the environment. This was aided by outward bending of the delaminated regionto relax TGO and YbMS surface layer stresses developed during the cooling phase of each thermal cycle.

© 2016 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

1. Introduction

Environmental barrier coatings (EBCs) will be needed to protectSiC-based ceramic components in future gas turbine engines [1].Their primary purpose is to eliminate the rapid volatilization of SiCin water vapor rich environments [1e6], while also inhibitingoxidation of SiC-based ceramic matrix composites (CMCs) [7e12].The design of these EBCs is therefore driven by a different combi-nation of objectives to those of the thermal barrier coating (TBC)systems that are widely used to extend the life of superalloy airfoils[13]. The objective of TBC designs is to reduce a metal component'ssurface temperature while delaying delamination failure fromoxidation [14,15], reactions with molten silicate deposits [16,17]and various forms of hot corrosion [18e23]. However, unlike TBC

y).

lsevier Ltd. All rights reserved.

systemswhere coating spallation results in damage, but usually notloss of the airfoil, the premature failure of an EBC could have con-sequences that are much more serious, since the life of an unpro-tected SiC component in a combustion environment is likely to beshort [2,3,24e26]. Since no sufficiently durable EBC system for usein thermo-cyclic, water vapor rich environments has been reportedto date, the development of such a coating system is of considerableimportance for the implementation of SiC-based CMCs in advancedgas turbine engines.

For non-rotating applications involving several thousand hoursof environmental exposure at temperatures up to 1316 �C (2400 �F),research has focused on multilayered EBC systems that use a Si“bond coat” applied to the SiC. The design objective for the bondcoat is similar to that of the aluminum-rich metallic layer used inTBC systems [27]. It serves as a sacrificial layer that forms a pro-tective thermally grown oxide (SiO2) upon exposure to oxidizingspecies, thereby inhibiting their access to the SiC-based component

Fig. 1. Schematic illustration of a bi-layer Yb2Si2O7/Si (YbDS/Si) coated a-SiC substrate.

B.T. Richards et al. / Acta Materialia 106 (2016) 1e142

surface. However, volatilization of the protective oxide occurs by itsreaction with water vapor to form gaseous silicon hydroxides suchas Si(OH)4. The bond coat must therefore be covered (environ-mentally protected) by materials that impede the diffusion of ox-ygen and water vapor to the Si surface, while being highly resistantto silicon hydroxide forming reactions with water to avoid steamvolatilization during engine operation [1,25,28e38].

These design objectives must be achieved subject to numerousconstraints. For example, the various layers of the EBC systemmusthave a coefficient of thermal expansion (CTE) similar to SiC to avoidcracking or delamination during heating and cooling over the widetemperature range experienced during operation of a gas turbineengine [32,33]. The EBC materials must be stable and thermo-chemically compatible so that deleterious reaction products are notformed between them or with the substrate. In addition, they mustmeet other demands of the engine environment which includeresistance to erosion by fine (dust) particles [39e42], impact bylarger foreign objects [41e43] and survival of reactions withmoltencalcium-magnesium-aluminum-silicates (CMAS) [44e48]. Finally,when used on rotated components, EBCs will be required to sustainsignificant static and cyclic loads that could cause creep deforma-tion and fracture [22,35e37,49e52]. This use environment istherefore one of the most severe ever envisioned for an advancedmaterial system.

The environmental stability, CTE, and elastic modulus of manycandidate EBC materials have been reviewed [1,30,34,53]. Theytterbium monosilicate (Yb2SiO5)/mullite (Al6Si2O13)/Si tri-layersystem has received significant recent attention due to the ther-mochemical compatibility between its component materials andthe very low steam volatility of Yb2SiO5 (YbMS) [53]. However, theCTE of YbMS has been found to be substantially higher than that ofSiC (7.5 � 10�6 �C�1 for YbMS versus 4.7 � 10�6 �C�1 for SiC) [1,34].During cooling from 1316 �C, this was found to result in thedevelopment of a high (biaxial) tension in the YbMS layer, whichwas relieved by channel (mud) cracking [32,33]. The presence ofcracks through the YbMS and mullite (which also has a higher CTEthan that of SiC) layers then provided oxidizing species a gas phasetransport path to the Si bond coat, resulting in the rapid growth of ab-cristobalite (SiO2) phase on the outer silicon surface. The ther-mally grown oxide (TGO) also underwent a large reduction involume (~4.5%) during its cubic to tetragonal (inversion type) b/ aphase transformation upon cooling through ~220 �C, resulting insevere TGO microcracking, loss of its oxidation protection qualities,and premature spallation of the EBC. These observations indicatethat high CTE topcoat candidates (including YbMS, other rare earthsilicates with high CTE, and aluminum-silicates such asmullite), arelikely to be unsatisfactory for the diffusion-impeding layers of EBCsystems.

Ytterbium disilicate (YbDS), a line compound in the Yb2O3eSiO2binary phase diagram with composition Yb2Si2O7, is a promisingalternate material for environmental barrier layer applications. Itsrecently measured temperature dependent CTE is about4.1 � 10�6 �C�1 [53] (compared to 4.7 � 10�6 �C�1 for SiC) [47].YbDS therefore satisfies one of the primary design requirements ofan EBC application. However, its resistance to volatilization by re-actions with water vapor is significantly less than that of its mon-osilicate counterpart [53], and its thermochemical stability incontact with a Si bond coat has not been reported. A recent studyhas led to the identification of an optimized atmospheric plasmaspray (APS) deposition method for applying YbDS coatings to Sibond coat protected SiC substrates [54].

The objective of the preliminary study reported here is to beginan investigation of the thermomechanical and environmentalresponse of an APS deposited YbDS/Si EBC system on a SiC substrateas it is subjected to thermal cycling between 110 �C and 1316 �C in a

flowing steam/oxygen environment. The study reports and char-acterizes the growth of a TGO layer on the silicon bond coat duringsteam cycling for up to 2000 h, and estimates the diffusion coeffi-cient of the oxidizing species through the YbDS coating. It reportsand investigates the process by which the topcoat is partiallyvolatilized during steam cycling, and explores the mechanism bywhich edge coating failure eventually occurs during thermalcycling in a water vapor rich environment. This EBC system is thefirst to be reported with sufficient resistance to volatilization andthermal cycling for an engine application.

2. Experimental

2.1. Coating deposition

Bi-layer YbDS/Si coatings were deposited onto surface rough-ened 25.4 mm � 12.7 mm x 4.8 mm thick a-SiC Hexoloy™ sub-strates (Saint Gobain Ceramics, Niagara Falls, NY), Fig. 1. Thesubstrate edges were first ground forming a 45� chamfer to facili-tate over-coating the edges of the samplewhere delaminationmostoften initiates [32,32]. The target thickness of the silicon layer was50 mm while that of YbDS layer was 125 mm. The plasma spraydeposition parameters used for each layer are summarized inTable 1. The structure of as deposited and stabilization annealedYbDS layers deposited using this spray parameter combination hasbeen recently reported by Richards et al. [54] These spray param-eters resulted in a dense YbDS coating with as deposited grain sizeof several hundred nanometers. During thermal cycling at 1316 �C,this gradually increased to ~5 mm. These spray conditions werechosen as a compromise to ensure deposition of low void contentcoatings while limiting the loss of SiO from the liquid dropletsduring transit through the plasma plume. This SiO loss is a result ofits higher vapor pressure (compared to ytterbium containing vaporspecies) at the droplet temperatures encountered in plasma spraydeposition [34]. It results in the formation of a two phaseYbDS þ YbMS coating observable by a difference in gray scalecontrast when imaged in the scanning electron microscope usingbackscattered electrons (BSE mode imaging), Fig. 2(a). In thecoatings studied here, the YbMS volume fractionwas ~15%, and can

Table 1Deposition parameters for air plasma spray deposition of YbDS/Si layers.

APS Layer Torch power (kW) Arc current (A) Primary Ar (slm) Secondary H2 (slm) Powder feed rate (g/min) Ar carrier gas flow rate (slm)

Yb2Si2O7 11.2 275 84.95 0.94 41.5/lower 5.90Si 22.7 500 77.87 0.94 31.0/upper 5.90

Fig. 2. BSE images of a stabilization annealed YbDS/Si bi-layer coating deposited ontoa-SiC. (a) Low magnification image showing two phases in the YbDS topcoat. (b)Higher magnification view of the YbDSeSi interface showing interface structure andsmall Yb2SiO5 (YbMS) precipitates (of lighter contrast) in the YbDS topcoat.

Fig. 3. Schematic illustration of the testing geometry and environment local to thesamples during the hot part cycle of the steam-cycle. The furnace was raised after 1 hreducing the sample temperature from 1316� to 110 �C.

B.T. Richards et al. / Acta Materialia 106 (2016) 1e14 3

be seen in Fig. 2(a) as elongated (ellipsoidal) regions of a lightercontrast phase oriented in the plane of the coating. YbMS pre-cipitates can also be seen in the YbDS topcoat, Fig. 2(b), consistentwith prior studies of this system [54]. Though the topcoat is amixture of YbDS and YbMS, it will be referenced as the “YbDS layer”for convenience.

The APS powders used for deposition were the same as thoseused in prior EBC research [54], and consisted of a specially pro-duced stoichiometric YbDS powder (Treibacher Industrie Inc., Tor-onto, ON) and APS grade SL-111 Si powder (Micron Metals,Bergenfield, NJ). The YbDS powder size distribution was the sameas that previously reported (from 20 to 50 mm). The Si powder wassieved before use to remove fine and coarse particles so that itspowder diameter ranged between 70 and 110 mm. The spray pa-rameters for Si deposition were adjusted to result in a low porositySi bond coat similar to that previously reported [33,54], Table 1. Itsrough surface, Fig. 2(a) and (b) was intended to improve adhesionbetween the two coating layers.

Both the Si and YbDS layers were deposited onto substrates

placed in a box furnace held at a temperature of 1200 �C. To reduceoxidation of the SiC and Si bond coat surfaces during deposition, anAr/H2 reducing gas mixture was continuously flowed through thefurnace at a rate of 20 slm. After deposition of the Si layer, thereducing gas flow was terminated and the YbDS layer was thendeposited into the furnace. The coated samples were finally stabi-lization annealed (to transform metastable phases) at a tempera-ture of 1300 �C in air for 20 h [1,25,28,30e34].

2.2. Steam cycling

Coated substrates were isothermally cycled in a steam-cyclingfurnace. The thermal cycle consisted of a 60 min hot (1316 �C)and a 10 min cold (110 �C) cycle conducted within an atmosphericpressure, 90% H2O/10% O2 gas flow with a flow speed of 4.4 cm/s(equivalent to a volumetric flow of 4.1slm). The environment localto these substrates during steam cycling is shown in Fig. 3. For thetesting configuration used here, 4 samples were initially cycledsimultaneously, with one being removed after 250, 500, 750 and1000 cycles. This was followed by the testing of a fifth samplesubjected to a multi-sample exposure for 500 cycles and 1500additional cycles with just a single sample in the furnace (for a totalof 2000 cycles). During steam cycling, all the samples were exam-ined optically every 100 cycles to assess damage.

The environmental flow conditions were similar to those pre-viously used for steam-cycling studies [32,33], and approximate the

B.T. Richards et al. / Acta Materialia 106 (2016) 1e144

H2O partial pressure during lean hydrocarbon combustion at apressure of 10 atm. Saturation of Si-bearing vapor (Si(OH)4 species)can occur in this furnace arrangement. Because the steam flowspeed was low, and the area of (uncoated) SiC exposed to theenvironment was large (2800 mm2) when 4 samples were present,a significant Si(OH)4 concentration is expected from the water va-por reactionwith the SiO2 scale grown on the exposed SiC surfaces.The recession rate of SiO2 was calculated for the conditions usedhere by the method of Golden and Opila [55], and found to be44.65 nm/h. Since the surface area of SiO2 exposed to the envi-ronment was 700mm2 per sample, 1.35� 10�9 mol of Si(OH)4 wereformed every second while four samples were in the furnace. Usingexperimental thermochemical data for the similar YeSieO system[56], the equilibrium partial pressure of Si(OH)4,was found to be4.97 � 10�6 atm. Since the gas flow rate through the steam-cyclingfurnace was 4.1 slm, the Si(OH)4 concentration was 40% of thesaturation limit with four samples in the furnace, but only 10% ofthe saturation limit with a single sample present. Testing the 2000cycle sample individually for its last 1500 cycles in a low Si(OH)4concentration environment then enabled the surface volatilizationof the YbDS layer to be observed and characterized. The singlesample 2000 cycle test was therefore intended to investigate thecombined effects of steam volatilization of the YbDS layer andthermal cycling while the other four samples enabled study of thethermomechanical response of the EBC system and the effects ofsteam reactions with edge exposed TGO.

2.3. Coating characterization

X-ray diffraction (XRD) measurements of the YbDS topcoat us-ing X'Pert Pro MPD software (PANalytical, Westborough, MA) wereperformed on some samples. These samples were sectioned, pol-ished, and examined with a Quanta 650 field emission scanningelectron microscope (FEI, Hillsboro, OR) operating in the back-scattered electron (BSE) mode. All images were collected underlow-vacuum conditions. A gamma correction was applied to theimages to enhance relative contrast between high and low-Z ma-terials [32,54]. Energy dispersive spectroscopy (EDS) was used forelemental microanalysis (X-MaxN 150 SDD, Oxford Instruments,Concord, MA). EDS spectrawere captured using a 10 kV acceleratingvoltage and small spot size to minimize fluorescence interactions[47].

Average TGO layer thicknesses were determined from 50different areas equally spaced across the coating cross-section. Thethickness of oxide scale perpendicular to the local silicon bond coatsurface of each sampled area was measured at about hundredevenly spaced locations across the coating cross-section and aver-aged. The standard deviation of these average oxide scale thicknesswas 100 nm or less, however, as shown later, the difference be-tween the thinnest and thickest TGO thickness was much largerthan this. It should also be noted that the measurements used tocompute an average thickness were collected from locations wherethe TGO layer was pore-free and flat. Raman Spectroscopy, using aninVia microscope (Renishaw, Hoffman Estates, IL, was used toidentify the TGO phase [23,24]. Raman analyses were performedusing a 50� lens with a numerical aperture of 0.5. An argon laser(488 nm) was used as the light source. Approximately 99% of theGaussian distributed incident light of the source resided within a1 mmdiameter circle on the sample, yielding a spatial resolution forthis technique of ~1 mm.

2.4. Energy release rate and residual stress calculations

Thermomechanical analyses of coating residual stresses, andassociated strain energy release rates (ERR) during debonding of

the elastic coating system, were calculated using the LayerSlayer[57] software package. All ERR calculations assumed that thecoating layers remained elastic during cooling and ignored stressrelief by mechanisms such as cracking or creep which can be sig-nificant at temperatures above 800 �C [58]. The ERR analysis wastherefore similar to that previously applied to tri-layer EBC systems[32,33]. The thermophysical properties and the small residualstresses calculated for this elastic coatingesubstrate bi-layer aftercooling from the stabilization (stress free) annealing temperatureof 1300 �C (before TGO growth had commenced) are listed inTable 2 [59e68].

3. Results

Six low porosity YbDS on Si bi-layer coatings were deposited onSiC substrates using the optimized APS approach. A photograph ofthe coated surface of one of the samples after deposition of itscoating is shown in Fig. 4(a) and after stabilization annealing inFig. 4(b). The average thickness of the YbDS layer for each sample isreported in Table 3. Slight variations in thickness of the YbDS layerwere observed between individual coatings and over the cross-section of a single coating. The thickness of the majority of thetop coat for each sample generally varied within ±10 mm of thereported average. However, small areas of substantially thicker orthinner thickness (up to ± 20% of the average) were occasionallyobserved which is typical for an APS process [54]. The thicknessvalues reported in Table 3 are the average of fifty measurementstaken normal to the substrate surface (10 values sampled from fivedifferent regions across the coating). One of the samples was usedto characterize the coating composition and structure while theremainder were steam cycled between 110 �C and 1316 �C.

While edge damagewas evident upon optical inspection of mostsamples, the damage did not appear to have affected the samplecenter even after 2000 cycles, Fig. 4(c), and none of the coatings hadspalled in locations other than the coating edges. Fig. 5 shows BSEmode SEM images of the as-deposited and annealed coating crosssections, and a selection of those that had been steam cycled.Minimal changes in layer thicknesses and coating structure wereapparent in these low magnification images at the coating center.Examination of these images indicated the gradual (and very mi-nor) formation of a new TGO phase at the YbDSeSi interface, Fig. 5,and the appearance of a layer of porous and microcracked materialat the outer surface of the YbDS layer in the sample removed fromthe furnace after 2000 steam cycles, Fig. 5(f).

3.1. Thermally grown oxide (TGO)

When examined by SEM in the BSE imaging mode, a darker grayphase was found at the YbDSeSi interface of all of the steam-cycledcoatings, Fig. 6. This dark gray phase was observed to be dense butof variable thickness in each sample. An EDS analysis of the layerrevealed that it contained only Si and O. Since there are 11 crys-talline polymorphs of silica and two other non-crystalline phases[69], Raman analysis of the layer was performed to identify whichphase had formed. The only spectral peaks present not belonging toSi hadwavenumbers of 230 cm�1 and 416 cm�1, indicating the layerto be a-cristobalite with SiO2 stoichiometry [70e72]. This a-cris-tobalite had presumably formed as b-cristobalite at 1316 �C andtransformed on cooling through 220 �C to the a phase. Cristobalitehas previously been identified as the TGO phase formed upon a Sibond coat during thermal cycling at this temperature [32,33], andon SiC oxidized at comparable temperatures [2,73], even though itis not the equilibrium phase at this temperature [74]. The Yb2O3 e

SiO2 phase diagram relevant to this EBC system has only fourstoichiometric phases at temperatures up to the YbDS melting

Table 2Thermophysical properties of EBC system components.

Material CTE ( � 106 �C-1) Young's Modulus (GPa) Poisson ratio n Layer thickness Thermal residual stressa (MPa) Application

YbDSEAPS

4.1 180b 90c 0.27d 125 mm �170�88

Top coat

Cristobalite-aCristobalite-b

96e

3.1[36]65 [29]70f

�0.164[35]�0.042[37]

0e5 mm >4300 Oxidation product

SiliconEAPS

4.1 163 [34] 70g 0.223 50 mm �145�65

Bond coat

SiC (a) 4.67 430 0.14 4.76 mm 2412

Substrate

a Calculated using LayerSlayer and the bulk mechanical properties within the table.b Value from nanoidentation.c 50% reduction in modulus assumed for dense APS structure, similar to prior APS work [28].d Based on Y2Si2O7 [33].e Average of values reported on the 20e200 �C interval [31,32].f Based on Young's modulus ratio of a and b quartz [30] and a cristobalite.g Value for APS material from Richards, Zhao, and Wadley [28].

Fig. 4. Optical images of YbDS/Si coated SiC substrates. (a) As-deposited, (b) annealed,and (c) after 2000 steam cycles at 1316 �C with the gas flow direction indicated. Thelighter contrast region around the edge of the sample in (c) correspond to regions ofthe coating that had partially delaminated from the substrate.

B.T. Richards et al. / Acta Materialia 106 (2016) 1e14 5

point (1650 �C): YbDS, YbMS, and the two terminal oxides [75]. This

indicates that there is no mixing of Yb2Si2O7 and SiO2 at temper-atures up to ~1650 �C, and it is therefore concluded that theYb2Si2O7 e SiO2 interface is thermochemically stable, in agreementwith prior determinations [76].

Examination of Fig. 6 indicates that the average thickness of thethermally grown oxide (TGO) increased with high temperatureexposure time, reaching a thickness of a few micrometers after2000 steam cycles. The TGOwas dense and adherent in all samples.The average measured thickness, h' of the TGO near the center ofthe steam-cycled samples is plotted as a function of hot exposuretime (number of cycles times 1 h) in Fig. 7(a). The measured oxidethickness depended upon accumulated time, and h'(t) at 1316 �Ccould be well fitted to a linear relation: h'(t) ¼ ho þ kl t where thelinear rate constant kl ¼ 1.2 nm/h and ho ¼ 140 nm. This ho valueagreed with other observations of annealed coatings (not shown)that had a TGO thickness before steam cycling of 100e200 nm.

Careful examination of the TGO layer of the sample exposed to2000 steam cycles, Figs. 6(d) and 8(a) shows that it containedclosely (~2 mm) spaced vertical (channel type) cracks that fullypenetrated the oxide layer. Occasionally, one of these crackedsegments had also delaminated from the underlying Si layer,especially in regions where the local TGO thickness varied, Fig. 8(a).Channel type cracking of the TGO was also observed in the sampleexposed to 1000 steam cycles, Fig. 8(b), but to a much lesser extentthan the 2000 cycle sample. This channel cracking was onlyobservable in the SEM in samples exposed to 1000 or 2000 steam-cycles whose TGO thicknesses were 1 mm or more. The separationdistance between the crack faces was measured to be in the rangeof 3e10 nm implying that the crack face separations accommo-dated an inplane equivalent strain (500 cracks per mm� 5 nm/percrack) of 0.25%.

3.2. YbDS volatilization

The 2000 steam-cycle topcoat was observed to have suffered asurface reaction with the steam environment, Fig. 5(f). X-ray MassAttenuation Coefficients [77] indicate the maximum depth ofpenetration of 45 kV X-rays is ~75 mm for YbDS and ~56 mm forYbMS. Since the majority of the diffracted signal in an XRD exper-iment therefore originates from a region within a depth of 20 mmfrom the surface, XRD is a suitable technique for a phase analysis ofthe surface volatilized region. Fig. 9 shows an XRD pattern of thesurface of the sample subjected to 2000 cycles and compares it tothat of a sample in the annealed condition. All indexed peaks couldbe attributed to one of three phases previously identified in thissystem [54]. The XRD patterns for the thermally cycled surfaceindicated a significant increase in the volume fraction of (the I2/a

Table 3YbDS layer thicknesses of samples. Thicknesses range ±10 mm.

Coating condition As-deposited Annealed 250 Cycles 500 Cycles 750 Cycles 1000 Cycles 2000 Cycles

Thickness (mm) 125 135 125 155 130 135 125

Fig. 5. BSE mode SEMmicrographs of YbDS/Si coated system before and after steam cycling. (a) As-deposited, (b) annealed, (c) after 250 steam cycles, (d) 500 cycles, (e) 1000 cycles,and (f) 2000 cycles.

Fig. 6. BSE micrographs of the YbDSeSi interface after steam cycling showing the development of a thermally grown oxide (TGO). (a) After 250, (b) 500, (c) 1000, and (d) 2000cycles at 1316 �C with 90%H2O/10%O2 flowing at 44 mm/s.

B.T. Richards et al. / Acta Materialia 106 (2016) 1e146

Fig. 7. Plots of SiO2 TGO thickness as a function of accumulated exposure time at1316 �C for a bi-layer EBC. (a) Average thicknesses measured from multiple sampleswith varying YbDS layer thickness. (b) Average thicknesses normalized for thermallygrown silica beneath a 100 mm thick YbDS layer.

Fig. 8. BSE micrograph of the thermally grown cristobalite (SiO2) oxide at the YbDSeSiinterface during steam cycling. (a) TGO at the YbDSeSi interface after 2000 steamcycles, (b) after 1000 steam cycles.

B.T. Richards et al. / Acta Materialia 106 (2016) 1e14 7

structure) YbMS phase which was present in addition to (the C2/mstructure) YbDS. Only a trace amount of a metastable (P21/cstructure) YbMS phase was identified.

Fig. 10(a) shows a higher magnification BSE image of the crosssectional surface of a steam volatilized YbDS region: the environ-mentally attacked YbDS layer was 5e15 mm thick. Yb and Si EDS dotmaps are presented in Fig. 10(b) and (c). These dot maps and thelighter surface BSE contrast indicate an increased Yb (and reducedSi) content in the surface volatilized layer compared to the unaf-fected material below. Based upon the crystallographic identifica-tion of only YbMS and YbDS by XRD and the differences in Yb and Sicontent indicated by EDS, this surface volatilized layer was identi-fied as predominantly YbMS, consistent with steam jet studies [78].

Fig. 11 shows three higher magnification BSE mode SEM imagesof the 2000 cycle sample surface. They show that the surface layerhad also become porous. The pores had roughly circular crosssections but were long in length, and many were connected to thesurface of the coating. The porosity of this surface layer, andreduced SiO2 content when compared to the as-deposited topcoat,are consistent with loss of SiO2 from the YbDS layer during steamcycling, leaving YbMS that has exhibited a greater resistance tovolatilization [30,78]. These observations suggest that a YbDStopcoat might be able to provide better than anticipated environ-mental protection under the test conditions used here.

Several forms of mechanical damage are evident in the samplesurface exposed to 2000 h of flowing steam and oxygen at 1316 �C,Fig. 11. Some areas underwent vertical cracking of the adherent, butporous YbMS surface layer, Fig. 11(a), with very large crack openingdisplacement. In other areas, Fig. 11(b), delamination cracks hadresulted in spalling of the compositionally modified surface region.Elsewhere, channel cracks propagated deeply into the unaffectedYbDS region and intersected a delamination crack within the un-affected YbDS region, Fig. 11(c). The damage was consistent withinplane shrinkage due to SiO2 loss and residual stresses that ariseupon cooling; a consequence of the large CTE mismatch betweenYbMS (7.5 � 10�6 �C�1) and the YbDS (4.1 � 10�6 �C�1) [33].

3.3. Edge delamination

Only the edges of the coatings were observed to undergodelamination, the severity of which increased with exposure timeas shown in Fig. 12. It was manifested as a lifting-off of the coatingedge. This edge delamination occurred despite attempts to extendthe coating over the sides of the substrate by chamfering the cor-ners, as can be seen in Fig. 12(a). The edge failure mode began withoxidation of the thin Si bond coat at the side of the substratesforming the darker gray TGO phase that can be seen with BSE im-aging, Fig. 12(b). EDS and Raman analyses of this dark gray phaseconfirmed it to be a-cristobalite (SiO2). This edge SiO2 layer un-derwent rapid steam volatilization and mechanical spallation,Fig. 12(c). As this consumed the Si bond coat at the edge of the

Fig. 9. X-ray diffraction patterns of the annealed and 2000 steam-cycle sample surfaces. All peaks of the patterns have been indexed using equilibrium YbDS (C2/m) and equilibriumI2/a and metastable P21/c YbMS peaks. The slightly elevated background of the 2000 cycle pattern results from a very thin layer of epoxy on the surface that was used to preventcoating spallation during handling.

B.T. Richards et al. / Acta Materialia 106 (2016) 1e148

samples, it allowed the YbDS layer to bend away from the substratesurface. The distance of oxygen penetration from the sides of thesample, and therefore lateral distance of TGO growth and coatingdelamination from the edge of the sample, all increased withaccumulation of steam cycles. After 2000 cycles, the coatingshowed a reverse curvature, Fig. 12(d).

4. Discussion

The results presented above indicate that the use of an APSdeposited bilayer Si/YbDS EBC provided up to 2000 h of protectionfrom volatilization during steam cycling between 110� and 1316 �Cin an atmospheric pressure, 90% H2O/10% O2 environment. TheYbDS layer impeded the transport of oxidizing species to the Sibond coat resulting in slow TGO layer thickening. Prolongedthermo-cyclic testing in slowly flowing steam containing a lowpartial pressure of Si(OH)4 resulted in the gradual loss of SiO2 fromthe outer surface of the YBDS top coat accompanied by its con-version to YbMS. There was no evidence of residual stress drivenchannel cracking or delamination of the coating system, consistentwith the good CTE matching in the substrate coating system.However, significant edge delaminationwas observedwhich slowlyadvanced inwards from the periphery of the substrates. This edgedelamination was initiated by oxidation of the environmentexposed bond coat at the sides of the test coupons, and was aidedby the development of a bending moment in the coating as theYbDS outer surface was converted to YbMS with a significantlyhigher CTE coefficient.

4.1. Oxide growth at the bond coat surface

After 2000 h of steam and oxygen exposure at 1316 �C, thethickness of the cristobalite TGO oxide layer formed in the YbDS/Sisystem had only reached 2.5 mm. This contrasted sharply withsimilar studies of YbMS protected silicon layers on the same sub-strate where TGO thicknesses of 15e150 mm (depending upon thecoating structure) were observed after less than 400 h of identicalenvironmental exposure [32,33]. In the YbMS coated system, veryrapid oxidation occurred because channel cracks fully penetratedthe YbMS layer and arrested in the silicon bond coat. This provideda very rapid, gas-phase transport path for both oxygen and H2Ovapor to the silicon surface. Deal and Grove [79] have shown thatthe rate of TGO growth rate on silicon was much higher for H2Othan dry O2. The slow TGO growth rate observed far from thesample edges of the channel crack free YbDS coated samples in-dicates that solid-state diffusional transport of the oxidizing species

through the YbDS layer governed the rate of oxidation of the Sibond coat surface. Since there has been little experimental study ofthe transport kinetics of candidate oxidizing species (H2O, OH� orO2�) in rare earth silicates, we illustrate our estimation of thediffusivity of the oxidant by taking it to be O2� (formed by thesurface dissociation of O2). Its most likely transport path would bevia grain boundaries since the grain diameter of the YbDS is small(several hundred nanometers after deposition to ~5 mm afterenvironmental exposure for 2000 h, Fig. 6)(d).

The silica TGO thicknesses measured here, Fig. 7(a), were ob-tained from samples of slightly differing YbDS thicknesses, Table 3.To compensate for this, the data has been corrected to that for aconstant YbDS layer thickness of 100 mm. Since there exists con-stant oxygen partial pressure at the environment/YbDS and YbDS/SiO2 interfaces, diffusion can be modeled by the one-dimensionalform of Fick's First Law. The flux of the oxidizing species, J(x), thatpasses through a plane in the YbDS layer a distance x below theouter surface of the layer depends upon the (unknown) oxygenYbDS diffusion coefficient, DO, and the gradient in concentration ofthe oxidant:

JðxÞ ¼ �DOdc=dx (1)

where c is the concentration of the oxidizer (presumed here to beoxygen anions).

The partial pressure of O2 at the exterior sample surface duringsteam cycling was 0.1atm. Using the ideal gas law, this partialpressure can be converted to an oxygen concentration (number ofmoles of O2 (n) per unit volume, V);

nV¼ P

RT¼ 0:1atm

0:0821 l$atm=ðmole$KÞ � 1589K¼ 0:767 mole

.m3

where P is the O2 partial pressure, R the Universal gas constant andT the absolute temperature. The oxygen partial pressure at theYbDS-SiO2 interface is established by local thermochemical equi-librium, and was calculated, using the FactSage software program[80], to be 4.29� 10�10 atm at 1316 �C. The concentration of oxygenanions at the bond coat surface was therefore very small comparedto that at the outer surface. If DO is constant throughout the YbDScoating, this implies that the oxidizing flux, J(x) reaching a growingSiO2 layer is inversely related to the thickness of the YbDS layerthrough which it has diffused. If all this flux was consumed in silicaformation, the TGO thickness will vary inversely with the thicknessof the diffusion barrier. The topcoat thickness therefore provides apractical means for controlling the rate of growth of the TGO layer.

Fig. 10. (a) BSE micrograph of a surface volatilized region of the topcoat. (b) A dot mapfor Yb, and (c) a dot map for Si of the same region shown in (a) from EDS analysis.

Fig. 11. BSE cross-section micrographs showing SiO2 volatilization from the YbDSsurface after 2000 steam cycles at 1316 �C. (a) Adherent area of YbMS with porosityand microcracking. (b) Spalled region of surface volatilized layer. (c) Delaminated re-gion of coating containing steam-reduced YbMS and region of less affected YbDS.

B.T. Richards et al. / Acta Materialia 106 (2016) 1e14 9

This observation also enables the thickness of the TGO layer as afunction of time, h'(t) for samples with variable thickness diffusionbarriers (observed here), to be related to that for a diffusion barrierof constant thickness, h(t) by:

hðtÞ ¼ h0ðtÞ xmeas

xtar(2)

where xmeas is the measured YbDS thickness that gave h'(t) and xtaris the target YbDS thickness that gives h(t). Each measured TGOthickness, h'(t) has been adjusted using Equation (2) to that for a

100 mmYbDS layer thickness (xtar ¼ 100 mm), and these values havebeen plotted as a function of oxidation time in Fig. 7(b). The silicalayer thickness remained linearly related to the oxidation time, butwith a slightly altered linear rate constant kl ¼ 1.44 nm/h.

It is also possible to make a preliminary estimate the oxygendiffusion coefficient, DO in the YbDS layer. If _mO is the mass of theoxygen ions that diffuse through the YbDS per unit area per unittime, and if MO is the molar mass of oxygen atoms, the oxygen fluxincident upon the TGO surface is:

Fig. 12. BSE micrographs of coating edges showing edge attack during steam cycling ofthe bi-layer EBCs. (a) Annealed coating with an over-sprayed edge, (b) edge TGOgrowth after 500 steam cycles, (c) significant edge lifting after 1000 steam cycles, and(d) example of extension of the edge delamination after 2000 steam cycles.

B.T. Richards et al. / Acta Materialia 106 (2016) 1e1410

JO≡ _mO=MO (3)

If _GSiO2is the TGO growth rate (in mass per unit area per unit

time) and 2MO =MSiO2the mass fraction of oxygen anions in SiO2,

the mass of oxygen ions that diffuse through the YbDS per unit areaper unit time is;

_mO ¼ _GSiO2* 2MO

�MSiO2

(4)

Since the TGO mass growth rate is simply the linear growth rateof the TGO, kl, multiplied by the density of cristobalite, rc,combining Equations (3) and (4) gives the oxygen flux incidentupon the TGO surface;

JO ¼ �kl* rc*2

�MSiO2

�(5)

Substituting kl from the experiments into Equation (5) and usinghandbook values for the density and molar mass of cristobalitegives the oxygen flux incident upon the TGO:

JO ¼ 1:44*10�9m3600s

� 2:34� 106g.m3 � 2

60g=mole

¼ 3:12� 10�8mole.�

m2$s�

From Equation (1), the concentration gradient of the oxygenions across a 100 mm thick YbDS layer is:

dcdx

¼ 2� 0:767mole�m3

100� 10�6m¼ 1:534� 104mole

.m4

Thus, the estimated oxygen diffusion coefficient in YbDS at1316 �C is:

DO ¼ JOdcO=dx

¼ 3:12� 10�8mole��

m2$s�

1:534� 104mole�m4 ¼ 2:03� 10�12 m2

.s

This value of DO is high in comparison to monatomic oxygendiffusion coefficients of other oxide materials at this temperature[81], but consistent with relatively fast boundary diffusion of oxy-gen in a YbDS layer with a grain size of 1e2 mm. It was also notedthat as it thickened, the TGO eventually underwent channelcracking, and the protective qualities of the oxide layer may havebeen compromised since these TGO cracks provide a faster trans-port path to the underlying silicon.

4.2. Steam volatilization of YbDS surface

The exposure of YbDS to H2O vapor at 1316 �C in a low Si(OH)4concentration environment has resulted in the formation of aporous YbDS surface layer. The origin of the porosity in the envi-ronmentally affected surface layer is consistent with SiO2 (mass)loss from the YbDS layer by formation gaseous Si(OH)4 leavingbehind a reduced volume region of YbMS. If Yb is assumed toexhibit no volatility, the resulting volume reduction during con-version of YbDS to YbMS is 26%. Since the coating is laterally con-strained, this volume reduction (under a state of biaxial tension in amaterial of very low creep strength [58]) has resulted in a poroussurface layer. The initial stage of this creep enabled pore formationmechanism is schematically illustrated in Fig. 13(a). As exposurecontinues, the pores resulting from SiO2 volatilization grow into thecoating extending the open porous structure. Volatilization canthen occur by a gas path diffusional route involving simultaneoustransport of water vapor into, and Si(OH)4 out of the coating. Theclose spacing of the tunneling pores is such that the solid-statediffusion distance to convert the YbDS to YbMS is small, Fig. 13(b).

4.3. Delamination energy release rate

The various fracture processes observed in the YbDS/Si systemare at first surprising since there are very small differences betweenthe CTE of the substrate, silicon and YbDS layers in this system,Table 2. As a result, the elastic residual stresses needed to drive

Fig. 13. Schematic illustration of the silica volatilization mechanism that results in aporous surface layer of YbMS on the exposed YbDS surface and a TGO layer on thesilicon bond coat after steam cycling. (a) Initial stage of volatilization and (b) late in theexposure process. The growth of a microcracked TGO when the oxidizing speciesreaches the bond coat is also shown.

Fig. 14. Predicted strain energy release rates during delamination of the interfaces inthe YbDSeSi EBC system after cooling from 1316 �C to 20 �C. The calculations in (a)used the elastic properties of solid YbDS and silicon layers while (b) used estimatedvalues for the mechanical properties of YbDS with APS with the APS structure. Allmechanical properties used during the calculations are listed in Table 2.

B.T. Richards et al. / Acta Materialia 106 (2016) 1e14 11

delamination and other fracture processes are small. However, theformation of a b-phase cristobalite SiO2 layer at the YbDS/Si inter-face can result in larger residual stresses being formed upon cool-ing, especially after the b to a transformation upon cooling through220 �C [32,33]. The strain energy release rates (ERRs) available todrive the growth of delamination cracks have been estimated foreach of the interfaces in this system using the LayerSlayer tech-nique [57] and the thermophysical data in Table 2. They are plottedas a function of the fraction of the Si bond coat converted to SiO2

and equivalent TGO thickness in Fig. 14. The ERR calculations wereperformed using both the Young's elastic modulus of fully dense(EBulk) YbDS and a modulus reduced to 50% of this value (EAPS)typical of low porosity APS materials [59,82e86]. For Si, the cal-culations used the polycrystalline bulk elastic modulus and a recentmeasured elastic modulus for a porous APS deposited Si bond coat[59]. Fig. 14(a) uses the elastic moduli of fully dense coating layerswhile Fig. 14(b) uses the lower moduli for the Si and YbDS layersgiven in Table 2. The APS coatings investigated here are expected tolie between these two bounding sets of calculations. These energyrelease rates are 1e2 orders of magnitude lower than thosecalculated for the YbMS/Al6Si2O13/Si tri-layer EBC system [32,33],which is consistent with the significantly better delaminationresistance of the YbDS system.

In the as-deposited and the annealed states, the TGO layer wasvery thin, and the ERR calculated for all interfaces in the systemwere below 13 J/m [2] using bulk elastic moduli, Fig. 14(a), andbelow 8 J/m2 for APS-like elastic modulus. During steam cycling theTGO thickness increased (at the original YbDSeSi interface), theERR for delamination of the YbDS on SiO2 interface remained

Fig. 15. Schematic illustration of the TGO wedging mechanism that caused edge liftingduring steam cycling of a YbDS/Si EBC deposited on SiC. (a) Shows the early stagewhere a SiO2 TGO forms at the edge of the sample and begins to vertically displace(wedges) the EBC above. (b) Shows a later stage of the process where the TGO has beeneroded by steam erosion creating a region of topcoat that is unattached to the bondcoat. The loss of SiO2 from the outer surface of the YbDS layer, results in a high CTEYbMS that causes outward bending of the unattached YbDS topcoat.

B.T. Richards et al. / Acta Materialia 106 (2016) 1e1412

almost constant, independent of the TGO thickness. This arosebecause the strain released by the YbDS layer (when freed from theunderlying system) was very low since there is a good CTE matchbetween Si and SiC, and the TGO layer had only a slight effect uponthe strain energy that remained stored in the portion of the systemthat was still attached to the substrate. Comparison of Fig. 14(a) and(b) shows that the ERR for delamination at the SiO2-YbDS interfacedecreased by a factor of two when the bulk elastic moduli werereplaced by the lower elastic moduli.

It is interesting to note that growth of the TGO initiallydecreased the driving force for failure at the SiCeSi and SieSiO2interfaces, consistent with the excellent delamination resistanceexhibited during the steam cycling experiments. The reason for thisis that upon delamination, the tensile stresses in the oxidecompress the YbDS layer in the delaminated state, implying thatless of its strain energy in the intact state is released duringdelamination. (Conversely, the YbDS layer acts to retain a portion ofthe oxide stress in the delaminated state, such that not all of theoxide's strain energy is released by delamination).

This mechanism for reduction in ERR, inwhich the strain energyin the delaminated bilayer is maintained to permit less energy to bereleased from the intact state, has been previously elucidated in atheoretical study of a different system [87]. This mechanism reliedcritically on the relative thickness of the layers, and the CTEmismatch between all layers. In the present case, as the oxide layerthickness grows, the balance of compressive and tensile energy inthe delaminated bilayer is lost, and the release of strain energy inthe TGO dominates the behavior.

The importance of the relative bending and stretching stiffnessof the delaminated bilayer is highlighted in the results for the APSsystem; decreasing the modulus of the YbDS decreases the initialvalue of the energy release rates (by a factor of two, as expected),but then proves less effective in maintaining strain energy in thedelaminated state. That is, a topcoat with a lower modulus is lesseffective in retaining strain energy in the TGO after delamination,leading to larger energy release rates. It is also clear that if the TGOundergoes channel cracking before delamination, this provides amechanism for partial relief of its tensile stress, resulting in areduction in the changes to the ERR seen in Fig. 14.

4.4. TGO channel cracking

Channel cracking of the TGO as its thickness increased could beseen in Fig. 6. Vlassak has calculated the strain energy release rate(GSS) during steady state channel cracking of a thin coating [88],using a method developed by Bueth [89] and showed that;

GSS ¼p s2t

�1� n2

�2 E

gða; b; hÞ

where s is the layer stress, t the thickness, n is Poisson's ratio, and Ethe elastic modulus of the TGO layer. The function g represents thecrack opening displacement in terms of the Dundurs parameters (a,b).

The measured TGO-channel crack opening displacement for at ¼ 1.5 mm thick TGO was of 3e10 nm (Section 3.1). Using me-chanical properties presented in Table 2, GSS ranges from 6.8 N/mfor a 3 nm crack opening displacement to 22.5 N/m for a 10 nmcrack-opening dimension at a TGO thickness of 1.5 mm. By com-parison, the critical strain energy for cracking, GIC for a-cristobalitehas been estimated to lie in the range of 7.1e7.7 N/m [90]. There-fore, channel cracking of the TGO becomes energetically favorablefor TGO thicknesses exceeding ~1.5 mm. This is in reasonableagreement with the dependence of the onset TGO channel crackingupon TGO thickness (exposure time) reported in Section 3.1, and is

consistent with the absence of delamination other than at theedges of the coatings in any of the steam cycling experiments.

4.5. Edge delamination

The tests conducted here used samples that were coated on onlyone side. As a result, the edges of the coatings were exposed to thesteam environment during testing. Delamination of the coatingwasobserved to progress from these regions, Fig. 12. There are twoprocesses contributing to edge delamination and the upwardbending of the delaminated coating schematically, Fig. 15. The firstwas oxidation of surface exposed silicon to form SiO2 which incursmore than a 100% volume increase [32,33]. TGO was first formed atthe edge of the sample where the silicon bond coat was in directcontact with the water vapor rich environment, Fig. 15(a). Deal andGrove [79] have shown that the oxidation rate of silicon in steam ismuch faster than of dry oxygen, and this appears to have resulted inthe very rapid edge surface oxidation observed here, Fig. 12. Ex-amination of Fig. 12(d), shows that the entire silicon bond coat wasoxidized and steam volatilized after a 2000 h exposure at 1316 �C.

As oxide growth advanced inwards from the coating edge, thedilation associated with the formation of SiO2 acted as a wedgeinserted at the interface between the Si and YbDS layers, Fig. 15(b).As this wedge extended laterally along the Si bond coat, the YbDStopcoat above was subjected to a bending load which was partiallyrelieved by creep deformation (since YbDS has a low creep flowstrength at the upper testing temperature of 1316 �C) [58]. Thea�cristobalite TGO has a larger CTE than the YbDS layer whichwould create a tensile stress in the TGO on cooling that counteractsthe upward bending. However, the TGO underwent a b / a cris-tobalite phase change on cooling, resulting in TGO microfractureand strain relaxation of the TGO stress during cooling below 220 �C.This also enabled the SiO2 at the exposed edge to spall, and for thatwhich remained to be volatilized more rapidly by the steam envi-ronment (a consequence of its increased specific surface area). Theremoval of material at the exposed back of the wedge also allowed

B.T. Richards et al. / Acta Materialia 106 (2016) 1e14 13

oxidation to continue at the un-oxidized front, thereby permittingthe wedge to rapidly advance inwards towards the coating center,Fig. 15(b). It is also noted that a second component to the upwardbending (and opening of the delaminated edge) was induced by thedevelopment of a YbMS layer on the outer surface of the YbDS layerwith a CTE almost twice that of the YbDS layer.

5. Conclusions

YbDS/Si bi-layer EBCs were deposited onto SiC substrates usinga previously optimized APS approach. Five samples were isother-mally steam cycled in a slowly flowing 90% H2O/10% O2 environ-ment using 60 min hot (1316 �C) and 10 min cold (110 �C) cycle. Itwas found that:

a) Dense APS deposited bi-layers consisting of 125 mm thicknominally YbDS (actually YbDSþ ~15 vol% YbMS) on a 50 mmthick Si bond coat have been found to environmentallyprotect the underlying silicon carbide well during isothermalsteam-cycling to 1316 �C for 2000 cycles.

b) The Si bond coat exhibited linear oxidation kinetics with alinear rate constant of 1.44 nm/h for up to 2000-h exposuresto steam at 1316 �C in a bi-layer system whose topcoatdiffusion barrier layer was 100 mm thick.

c) The oxygen diffusion coefficient for the YbDS layer at 1316 �Cwas calculated from TGO thickening kinetics to beDO ¼ 2 � 10�12 m2s�1, consistent with the oxygen diffusioncoefficients for other complex oxides at this temperature.

d) Linear bond coat oxidation to form microcrack-free cristo-balite has been estimated to terminate at a TGO thickness of~300 nm whereupon the growth rate would transition toparabolic behavior controlled by diffusion through the un-cracked TGO. However, TGO-channel cracking was observedproviding diffusive short-circuits through the scale andextending the linear oxidation regime indefinitely.

e) The estimated ERR for delamination in the system werepredicted to remain below 20 N/m for cristobalite scales upto 5 mm in thickness (twice that after 2000 cycles of oxida-tion), consistent with good adherence of the bi-layer coatingfar from its edges.

f) Oxidation of the bond coat at the edge of the samples initi-ated delamination cracking at the bond coat-YbDS interface.This damage progressed towards the interior of the samplesas the number of cycles increased due to a combination ofTGO microfracture and steam volatilization of the surfaceconnected TGO layer.

g) Loss of SiO2 from an approximately 15 mm thick surface layerof the YbDS layer was observed after prolonged steamcycling at 1316 �C in lower Si(OH)4 partial pressure test en-vironments. This was accompanied by the formation ofcreep-enabled voids that, in combination with channelcracks, relieved the significant shrinkage stress associatedwith SiO2 loss.

The dense YbDS/Si bi-layer EBCs deposited in this study hadexcellent steam-cycling durability while providing total aerialcoverage. This combination of attributes indicates that the systemmay be a viable option for some isothermal applications up to1316 �C. Further analysis of mechanical properties and the effect ofthermal gradients would be necessary to assess the viability of thesystem for use in rotating and internally cooled applications.

Acknowledgments

The authors are grateful to Elizabeth Opila and Hengbei Zhao of

the University of Virginia and Bryan Harder of the NASA GlennResearch Center for helpful discussions of various aspects of thisresearch. This work was supported by the Office of Naval Researchunder grants N00014-11-1-0917 (HNGW/BTR) and N00014-13-1-0859 (MRB) managed by Dr. David Shifler.

References

[1] K.N. Lee, Protective coatings for gas turbines, in: R. Dennis (Ed.), The GasTurbine Handbook, United States Department of Energy (DOE), 2006.

[2] E.J. Opila, Variation of the oxidation rate of silicon carbide with water-vaporpressure, J. Am. Ceram. Soc. 82 (1999) 625e636.

[3] E.J. Opila, Oxidation and volatilization of silica formers in water vapor, J. Am.Ceram. Soc. 86 (2003) 1238e1248.

[4] E.J. Opila, D.S. Fox, N.S. Jacobson, Mass spectrometric identification of Si-O-H(g) species from the reaction of silica with water vapor at atmosphericpressure, J. Am. Ceram. Soc. 80 (1997) 1009e1012.

[5] E.J. Opila, R.E. Hann Jr., Paralinear oxidation of CVD SiC in water vapor, J. Am.Ceram. Soc. 80 (1997) 197e205.

[6] E.J. Opila, J.L. Smialek, R.C. Robinson, D.S. Fox, N.S. Jacobson, SiC recessioncaused by SiO2 scale volatility under combustion conditions: II, thermody-namics and gaseous-diffusion model, J. Am. Ceram. Soc. 82 (1999) 1826e1834.

[7] B.N. Cox, F.W. Zok, Advances in ceramic composites reinforced by continuousfibers, Curr. Opin. Solid State Mater. Sci. 1 (5) (1996) 666e673.

[8] F.E. Heredia, J.C. McNulty, F.W. Zok, A.G. Evans, Oxidation embrittlementprobe for ceramicematrix composites, J. Am. Ceram. Soc. 78 (8) (1995)2097e2100.

[9] T.E. Steyer, F.W. Zok, D.P. Walls, Stress rupture of an enhanced nicalon/siliconcarbide composite at intermediate temperatures, J. Am. Ceram. Soc. 81 (8)(1998) 2140e2146.

[10] L.U. Ogbuji, Pest-resistance in SiC/BN/SiC composites, J. Eur. Ceram. Soc. 23 (4)(2003) 613e617.

[11] L.U.J.T. Ogbuji, Oxidative pest degradation of Hi-Nicalon/BN/SiC composite asa function of temperature and time in the burner rig, in: 23nd Annual Con-ference on Composites, Adv. Ceram. Mater. Structures B Ceram. Eng. Sci. Proc.,vol. 230, 2009, p. 105.

[12] K.J. LaRochelle, G. Morscher, Tensile stress rupture behavior of a wovenceramic matrix composite in humid environments at intermediate temper-aturedpart I, Appl. Compos. Mater. 13 (3) (2006) 147e172.

[13] R.A. Miller, Current status of thermal barrier coatings e an overview, Surf.Coatings Technol. 30 (1) (1987) 1e11.

[14] H. Zhao, Z. Yu, H.N.G. Wadley, The influence of coating compliance on thedelamination of thermal barrier coatings, Surf. Coatings Technol. 204 (2010)2432e2441.

[15] M.R. Begley, H.N.G. Wadley, Delamination resistance of thermal barriercoatings containing embedded ductile layers, Acta Mater. 60 (2012)2497e2508.

[16] S. Kr€amer, J. Yang, C.G. Levi, J. Am. Ceram. Soc. 91 (2008) 576e583.[17] H. Zhao, C.G. Levi, H.N.G. Wadley, Molten silicate interactions with thermal

barrier coatings, Surf. Coatings Technol. 251 (2014) 74e86.[18] H. Zhao, F. Yu, T.D. Bennett, H.N.G. Wadley, Morphology and thermal con-

ductivity of yittria stabilized zirconia coatings, Acta Mater. 54 (2006)5195e5207.

[19] M. Yoshiba, K. Abe, T. Aranami, Y. Harada, High-temperature oxidation andhot corrosion behavior of two kinds of thermal barrier coating systems foradvanced gas turbines, J. Therm. Spray. Tech. 5 (3) (1996) 259e268.

[20] D.R. Clarke, M. Oechsner, N.P. Padture, Thermal-barrier coatings for moreefficient gas-turbine engines, MRS Bull. 37 (10) (2012) 891e898.

[21] R. Darolia, Thermal barrier coatings technology: critical review, progress up-date, remaining challenges and prospects, Int. Mater. Rev. 58 (6) (2013)315e348.

[22] N.P. Padture, M. Gell, E.H. Jordan, Thermal barrier coatings for gas-turbineengine applications, Science 296 (5566) (2002) 280e284.

[23] A.G. Evans, D.R. Mumm, J.W. Hutchinson, G.H. Meier, F.S. Pettit, Mechanismscontrolling the durability of thermal barrier coatings, Prog. Mater. Sci. 46 (5)(2001) 505e553.

[24] J. Kimmel, N. Miriyala, J. Price, K. More, P. Tortorelli, H. Eaton, G. Linsey, E. Sun,Evaluation of CFCC liners with EBC after field testing in a gas turbine, J. Eur.Ceram. Soc. 22 (2002) 2769e2775.

[25] K.N. Lee, Current status of environmental barrier coatings for Si-Based ce-ramics, Surf. Coatings Technol. 133e134 (2000) 1e7.

[26] K.N. Lee, Key durability issues with mullite-based environmental barriercoatings for Si-based ceramics, J. Eng. Gas Turbines Power 122 (2000)632e636.

[27] Z. Yu, K.P. Dharmasena, D.D. Hass, H.N.G. Wadley, Vapor deposition of plat-inum alloyed nickel aluminide coatings, Surf. Coatings Technol. 201 (2006)2326e2334.

[28] B.J. Harder, J.D. Almer, C.M. Weyant, K.N. Lee, K.T. Faber, Residual stressanalysis of multilayer environmental barrier coatings, J. Am. Ceram. Soc. 92(2009) 452e459.

[29] K.N. Lee, J.I. Eldridge, R.C. Robinson, Residual stresses and their effects on thedurability of environmental barrier coatings for SiC ceramics, J. Am. Ceram.

B.T. Richards et al. / Acta Materialia 106 (2016) 1e1414

Soc. 88 (2005) 3483e3488.[30] K.N. Lee, D.S. Fox, N.P. Bansal, Rare earth silicate environmental barrier

coatings for SiC/SiC composites and Si3N4 ceramics, Corros. Ceram. MatrixCompos. 25 (2005) 1705e1715.

[31] K.N. Lee, D.S. Fox, J.I. Eldridge, D. Zhu, R.C. Robinson, N.P. Bansal, R.A. Miller,Upper temperature limit of environmental barrier coatings based on mulliteand BSAS, J. Am. Ceram. Soc. 86 (2003) 1299e1306.

[32] B.T. Richards, M.R. Begley, H.N.G. Wadley, Mechanisms of ytterbium mono-silicate/mullite/silicon coating failure during thermal cycling in water vapor,J. Am. Ceram. Soc. (2015) 1e10.

[33] B.T. Richards, S. Sehr, Foucault de Franqueville, M.R. Begley, H.N. Wadley,Delamination of ytterbium monosilicate/mullite/silicon coated SiC duringthermal cycling in water vapor, Acta Mater. 103 (2016) 448e460.

[34] B.T. Richards, H.N.G. Wadley, Plasma spray deposition of tri-layer environ-mental barrier coatings, J. Eur. Ceram. Soc. 34 (12) (2014) 3069e3083.

[35] D. Zhu, S.R. Choi, J.I. Eldridge, K.N. Lee, R.A. Miller, Surface Cracking andInterface Reaction Associated Delamination Failure of Thermal and Environ-mental Barrier Coatings, 2003.

[36] D. Zhu, D.S. Fox, L.J. Ghosn, B. Harder, Creep Behavior of Hafnia and YtterbiumSilicate Environmental Barrier Coating Systems on SiC/SiC Ceramic MatrixComposites, 2011.

[37] D. Zhu, K.N. Lee, R.A. Miller, Thermal gradient Cyclic Behavior of a Thermal/environmental Barrier Coating System on SiC/SiC Ceramic Matrix Composites,ASME Turbo Expo, 2002, pp. 171e178 (Power for Land, Sea, and Air).

[38] D. Zhu, R.A. Miller, Sintering and creep behavior of plasma-sprayed zirconia-and hafnia-based thermal barrier coatings, Surf. Coatings Technol. 108e109(0) (1998) 114e120.

[39] J. Nicholls, M. Deakin, D. Rickerby, A comparison between the erosionbehaviour of thermal spray and electron beam physical vapour depositionthermal barrier coatings, Wear 233 (1999) 352e361.

[40] X. Chen, M.Y. He, I. Spitsberg, N.A. Fleck, J.W. Hutchinson, A.G. Evans, Mech-anisms governing the high temperature erosion of thermal barrier coatings,Wear 256 (7) (2004) 735e746.

[41] R. Wellman, J. Nicholls, Erosion, corrosion and erosionecorrosion of EB PVDthermal barrier coatings, Tribol. Int. 41 (7) (2008) 657e662.

[42] N. Fleck, T. Zisis, The erosion of EB-PVD thermal barrier coatings: Thecompetition between mechanisms, Wear 268 (11) (2010) 1214e1224.

[43] X. Chen, R. Wang, N. Yao, A. Evans, J. Hutchinson, R. Bruce, Foreign objectdamage in a thermal barrier system: mechanisms and simulations, Mater. Sci.Eng. A 352 (1) (2003) 221e231.

[44] Hengbei Zhao, Bradley T. Richards, Haydn N.G. Wadley, Molten Silicate Re-actions with Ytterbium Silicate Coatings 288 (2016) 151e162.

[45] S. Kr€amer, J. Yang, C.G. Levi, C.A. Johnson, Thermochemical interaction ofthermal barrier coatings with molten CaOeMgOeAl2O3eSiO2 (CMAS) de-posits, J. Am. Ceram. Soc. 89 (10) (2006) 3167e3175.

[46] S. Kr€amer, J. Yang, C.G. Levi, Infiltrationeinhibiting reaction of gadoliniumzirconate thermal barrier coatings with CMAS Melts, J. Am. Ceram. Soc. 91 (2)(2008) 576e583.

[47] N.L. Ahlborg, D. Zhu, Calciumemagnesium aluminosilicate (CMAS) reactionsand degradation mechanisms of advanced environmental barrier coatings,Surf. Coatings Technol. 237 (2013) 79e87.

[48] K.M. Grant, S. Kr€amer, J. L€ofvander, C.G. Levi, CMAS degradation of environ-mental barrier coatings, Surf. Coatings Technol. 202 (4) (2007) 653e657.

[49] D. Zhu, K. N. Lee, and R. A. Miller, “Thermal Conductivity and ThermalGradient Cyclic Behavior of Refractory Silicate Coatings on SiC/SiC CeramicMatrix Composites,” pp. 443e452 in 25th Annual Conference on Composites,Adv. Ceram. Mater. Struct. B Ceram. Eng. Sci. Proc., Volume 22, Issue 4.

[50] D. Zhu, K.N. Lee, R.A. Miller, Cyclic Failure Mechanisms of Thermal andEnvironmental Barrier Coating Systems under Thermal Gradient Test Condi-tions, 2002.

[51] D. Zhu, R.A. Miller, Determination of creep behavior of thermal barrier coat-ings under laser imposed high thermal and stress gradient conditions,J. Mater. Res. 14 (1) (1999) 146e161.

[52] S.M. Meier, D.K. Gupta, Evolution of thermal barrier coatings in gas turbineengine applications, Trans. ASME J. Eng. Gas. Turbines Power 116 (1994)250e257.

[53] N.S. Jacobson, D.S. Fox, J.L. Smialek, C. Deliacorte, K.N. Lee, in: GRC (Ed.),Performance of Ceramics in Severe Environments, NASA Glenn ResearchCenter, Cleveland, OH, USA, 2005.

[54] B.T. Richards, H. Zhao, H.N.G. Wadley, Air Plasma Spray Deposition of ytter-bium silicates, J. Mater. Sci. 50 (2015) 7939e7957.

[55] R.A. Golden, E.J. Opila, A method for assessing the stability of oxides in high-temperature high-velocity water vapor, J. Am. Ceram. Soc. (2015). SubmittedArticle.

[56] C.W. Bale, P. Chartrand, S.A. Degterov, G. Eriksson, K. Hack, R. Ben Mahfoud,J. Melançon, A.D. Pelton, S. Petersen, FactSage thermochemical software anddatabases, Calphad 26 (2) (2002) 189e228.

[57] R.W. Jackson, M.R. Begley, Critical cooling rates to avoid transient-driven

cracking in thermal barrier coating (TBC) systems, Int. J. Solids Struct. 51 (6)(2014) 1364e1374.

[58] B.T. Richards, L.J. Ghosn, D. Zhu, H. Wadley, Mechanical properties of airplasma sprayed environmental barrier coating (EBC) systems: preliminaryassessments, in: Proceedings of the 39th International Conference andExposition on Advanced Ceramics and Composites, 2015.

[59] B.T. Richards, L.J. Ghosn, D. Zhu, H. Wadley, Mechanical properties of airplasma sprayed environmental barrier coating (EBC) systems: preliminaryassessments, in: Proceedings of the 39th International Conference andExposition on Advanced Ceramics and Composites, 2015.

[60] D. Taylor, The thermal expansion behaviour of the framework silicates,Mineral. Mag. 38 (297) (1972) 593e604.

[61] D. Lakshtanov, S. Sinogeikin, J. Bass, High-temperature phase transitions andelasticity of silica polymorphs, Phys. Chem. Miner. 34 (1) (2007) 11e22.

[62] T.-I. Mah, K.S. Mazdiyasni, Mechanical properties of mullite, J. Am. Ceram. Soc.66 (1983) 699e703.

[63] D.R. Peacor, High-temperature single-crystal study of the cristobalite inver-sion, Z Krist. 138 (1973) 274e298.

[64] Z. Sun, Y. Zhou, J. Wang, M. Li, g-Y2Si2O7, a machinable silicate ceramic:mechanical properties and machinability, J. Am. Ceram. Soc. 90 (8) (2007)2535e2541.

[65] R. Hull, Properties of Crystalline Silicon, IET, 1999.[66] H. Gercek, Poisson's ratio values for rocks, Int. J. Rock Mech. Min. Sci. 44 (1)

(2007) 1e13.[67] E. Bourova, P. Richet, Quartz and cristobalite: high-temperature cell param-

eters and volumes of fusion, Geophys. Res. Lett. 25 (13) (1998) 2333e2336.[68] A. Ballato, Poisson's ratios of auxetic and other technological materials, Ul-

trasonics Ferroelectr. Freq. Control. IEEE Trans. 57 (1) (2010) 7e15.[69] N. Wiberg, Holleman-Wiberg's Inorganic Chemistry, Academic Press, New

York, 2001.[70] J.B. Bates, Raman spectra of a and b cristobalite, J. Chem. Phys. 57 (9) (1972)

4042e4047.[71] F.L. Galeener, Raman and ESR studies of the thermal history of amorphous

SiO2, J. Non Crystall. Solids 71 (1e3) (1985) 373e386.[72] I.P. Swainson, M.T. Dove, D.C. Palmer, Infrared and Raman spectroscopy

studies of the aeb phase transition in cristobalite, Phys. Chem. Miner. 30 (6)(2003) 353e365.

[73] M.J. Guinel, M.G. Norton, Oxidation of silicon carbide and the formation ofsilica polymorphs, J. Mater. Res. 21 (10) (2006) 2550e2563.

[74] P. Patnaik, Handbook of Inorganic Chemistry, The McGraw-Hill Companies,2002, ISBN 0-07-049439-8, p. 344.

[75] I.A. Bondar, Rare-earth silicates, Ceram. Int. 8 (1982) 83e89.[76] D.L. Poerschke, J.S. Van Sluytman, K.B. Wong, C.G. Levi, Thermochemical

compatibility of ytterbiae(hafnia/silica) multilayers for environmental barriercoatings, Acta Mater. 61 (18) (2013) 6743e6755.

[77] J.H. Hubbell, S.M. Seltzer, Tables of X-ray mass attenuation coefficients andmass energy-absorption coefficients 1 KeV to 20 MeV for elements Z¼ 1 to 92and 48 additional substances of dosimetric interest, in: National Inst. OfStandards and Technology-pl, Ionizing Radiation Div, Gaithersburg, MD(United States), 1995.

[78] S. Ueno, T. Ohji, H.-T. Lin, Recession behavior of Yb2Si2O7 phase under highspeed steam jet at high temperatures, Corros. Sci. 50 (1) (2008) 178e182.

[79] B.E. Deal, A.S. Grove, General relationship for the thermal oxidation of silicon,J. Appl. Phys. 36 (12) (1965) 3770e3778.

[80] O. Fabrichnaya, H. Jürgen Seifert, R. Weiland, T. Ludwig, F. Aldinger,A. Navrotsky, Phase equilibria and thermodynamics in the Y2O3-Al2O3-SiO2system, Z. für Met. 92 (9) (2001) 1083e1097.

[81] W.D. Kingery, H.K. Bowen, D.R. Uhlmann, Introduction to Ceramics, Seconded., Wiley, New York, 1976, p. 240.

[82] T.A. Cruse, B. Johnsen, A. Nagy, Mechanical properties testing and results forthermal barrier coatings, J. Therm. Spray. Tech. 6 (1) (1997) 57e66.

[83] J.T. Demasi, M. Ortiz, Thermal barrier Coating Life Prediction Model Devel-opment, Phase 1, 1989.

[84] S.-H. Leigh, C.-K. Lin, C.C. Berndt, Elastic response of thermal spray depositsunder indentation tests, J. Am. Ceram. Soc. 80 (8) (1997) 2093e2099.

[85] C.J. Li, A. Ohmori, Relationships between the microstructure and properties ofthermally sprayed deposits, J. Therm. Spray. Tech. 11 (3) (2002) 365e374.

[86] S.M. Meier, D.M. Nissley, K.D. Sheffler, T.A. Cruse, Thermal barrier coating lifeprediction model development, J. Eng. gas turbines power 114 (2) (1992)258e263.

[87] M.R. Begley, H.N.G. Wadley, Delamination of ceramic coatings with embeddedmetal layers, J. Am. Ceram. Soc. 94 (2011) S96eS103.

[88] J.J. Vlassak, Channel cracking in thin films on substrates of finite thickness, Int.J. Fract. 119 (4) (2003) 299e323.

[89] J.L. Beuth Jr., Cracking of thin bonded films in residual tension, Int. J. SolidsStruct. 29 (1991) 1657e1675.

[90] D. Tromans, J.A. Meech, Fracture toughness and surface energies of covalentminerals: theoretical estimates, Miner. Eng. 17 (1) (2004) 1e15.