A

Addmvpi©

K

1

fihbnwpftecwJat

C

h0

Available online at www.sciencedirect.com

ScienceDirect

Journal of the European Ceramic Society 34 (2014) 3069–3083

Plasma spray deposition of tri-layer environmental barrier coatings

Bradley T. Richards ∗, Haydn N.G. WadleyDepartment of Materials Science and Engineering, University of Virginia, Charlottesville, VA 22903, United States

Received 13 January 2014; received in revised form 15 April 2014; accepted 18 April 2014Available online 14 May 2014

bstract

n air plasma spray process has been used to deposit tri-layer environmental barrier coatings consisting of a silicon bond coat, a mullite inter-iffusion barrier, and a Yb2SiO5 top coat on SiC substrates. Solidified droplets in as-deposited Yb2SiO5 and mullite layers were discovered to beepleted in silicon. This led to the formation of an Yb2SiO5 + Yb2O3 two-phase top coat and 2:1 mullite (2Al2O3*SiO2) coat deposited from 3:2ullite powder. The compositions were consistent with preferential silicon evaporation during transient plasma heating; a consequence of the high

apor pressure of silicon species at plasma temperatures. Annealing at 1300 ◦C resulted in internal bond coat oxidation of pore and splat surfaces,

recipitation of Yb2O3 in the top coat, and transformation of 2:1 mullite to 3:2 mullite + Al2O3. Mud-cracks were found in the Yb2SiO5 layer andn precipitated Al2O3 due to the thermal expansion mismatch between these coating phases and the substrate.

2014 Elsevier Ltd. All rights reserved.

eywords: Environmental barrier coatings; Ytterbium silicate

ascTtciciip(

th

. Introduction

Improvements to the fuel efficiency and specific thrust ofuture gas turbine engines will require advances in many areasncluding: (i) increasing the gas turbine inlet temperature, (ii)igher overall pressure ratios of engines, (iii) increasing they-pass ratio and (iv) incorporation of more efficient thermody-amic cycles into novel engine core designs.1–3 Developmentsill be paced by the emergence of new propulsion materials,rotective coatings, thermal management concepts, and manu-acturing tools for economical component fabrication. Arguablyhe most difficult materials challenge will be encountered in thengine’s hot core section.2,4–6 Here, severely stressed rotatingomponents will be subjected to ever increasing temperatureshose upper limit is set by stoichiometric combustion of the

et-A fuel. At a combustor pressure of 1 MPa and an air temper-◦ 7

ture of 500 C, the NASA CEA code predicts this fuel burn

emperature to be 2255 ◦C.

∗ Corresponding author at: 395 McCormick Road, P.O. Box 400745,harlottesville, VA 22904-745, United States. Tel.: +1 804 901 3070.

E-mail address: btr4we@virginia.edu (B.T. Richards).

ioipSSrc

ttp://dx.doi.org/10.1016/j.jeurceramsoc.2014.04.027955-2219/© 2014 Elsevier Ltd. All rights reserved.

Recent increases to the turbine inlet temperature have beenchieved by the use of internal air and surface film cooleduperalloy components protected against oxidation and hotorrosion by thermal barrier coating (TBC) systems, Fig. 1.his strategy is approaching its limit because coating sin-

ering leads to a loss of compliance and increased thermalonductivity,9–13 the bond coat oxidation rate (and thus delam-nation risk) increase rapidly with temperature,9,14–16 andalcium–magnesium–aluminum–silicate (CMAS) melting andnfiltration cause premature coating failure.17–19 Even if thesessues were resolved, TBCs can be eroded by fine (dust)articles19–22 or chipped away by larger foreign object damageFOD).19,22,23

These considerations have stimulated the development ofurbine components made from ceramic materials with muchigher temperature capability.24–27 Monolithic ceramics havensufficient fracture toughness, so development has focusedn damage tolerant fiber reinforced ceramic matrix compos-tes (CMCs) with weak fiber/matrix interfaces.28–31 The mostromising CMCs are based upon woven fabrics of BN coatediC fibers (such as Hi-Nicalon S and Sylramic fibers) and

iC matrices incorporated by chemical vapor infiltration withesidual pores filled by silicon slurry infiltration followed byarburization. While oxide–oxide CMC systems are chemically

3070 B.T. Richards, H.N.G. Wadley / Journal of the European Ceramic Society 34 (2014) 3069–3083

Fig. 1. The evolution of turbine blade materials, coatings, cooling conceptsand turbine inlet T4.1 gas temperature plotted against the year of entry. Currentengines utilize internal and film cooled single crystal (SX) superalloy bladespf

ifiahlp

sdwgTfllgrbno

oifomzctaiCtrev

aH

F

naieufstmwCTgmhast

pidaoutucc

2

2

oiwoFbAfuu

rotected by thermal barrier coatings in the hottest regions of the engine (adaptedrom Ref. 8).

nert in oxidizing environments, the current fibers have insuf-cient creep rupture strengths at temperatures of interest,32–34

nd are only used in lightly loaded applications.35–38 The mosteavily stressed components in future engines are likely to uti-ize SiC fiber/SiC matrix CMC materials for robust mechanicalerformance.27,39–43

When SiC is exposed to oxidizing environments a protectiveilica scale forms with a thickness that exhibits parabolic oxi-ation kinetics.44 Unfortunately, the protective scale reacts withater vapor in hydrocarbon combustion environments to createaseous silicon hydroxides via several reaction pathways.45–48

he temperature, pressure, incident water vapor flux, and localow conditions during exposure determine the observed para-

inear oxidation kinetics.46–49 SiC recession rates significantlyreater than 1 �m/h have been observed in the 1300–1350 ◦Cange.45 Consequently, CMC components need to be protectedy environmental barrier coatings (EBCs) to prevent compo-ent recession during operational lifetimes of tens of thousandsf hours.

The first EBCs consisted of a 3:2 mullite (3Al2O3*2SiO2r Al6Si2O13) coating applied directly to the substrate.50 Dur-ng prolonged high temperature exposure, silica was volatilizedrom the mullite leaving an alumina scale that readily spalled,ffering only short protection times.51 The next EBC develop-ent focused on bi-layer concepts in which an yttria-stabilized

irconia (YSZ) top coat was used to prevent volatilization andhemical attack of the underlying mullite.52 Thermal cyclingests in water vapor revealed that the coating quickly crackednd delaminated at either the YSZ/mullite or the mullite/SiCnterfaces. The rapid failure was attributed to the substantialTE mismatch between YSZ, the underlying mullite layer and

he substrate.51,53–56 Subsequently, several alkaline earth andare earth silicates were identified with well-matched CTE’s thatxhibit little weight change during high temperature, high waterapor pressure environmental exposure,57,58 Table 1.

Barium–strontium aluminosilicate (BSAS) has a low silicactivity and a CTE that is well matched with substrates, Table 1.owever, BSAS EBCs react with thermally grown silica,

t

s

ig. 2. Schematic illustration of an APS deposited tri-layer EBC structure.

ecessitating a tri-layer BSAS/mullite/silicon EBCrchitecture65 to eliminate failure at the coating/substratenterface and increase EBC life.72 Unfortunately, low meltingutectics occur near 1300 ◦C and limit the useful lifetime tonder 1000 h for higher temperatures,65 indicating the needor top coats with low volatility and high temperature phasetability. Ytterbium and lutetium silicates exhibit some ofhe lowest volatilities and are monomorphic. Ytterbium has

onosilicate (Yb2SiO5) and disilicate (Yb2Si2O7) variantshereas lutetium has only a monosilicate variant. The reportedTE of these and other candidate materials are summarized inable 1. The data for Yb2SiO5 is inconsistent. One report58

ives a value of 3.5–4.5 × 10−6 ◦C−1 which is reasonablyatched to SiC. However, a more recent report57 gives a

igher value of 7–8 × 10−6 ◦C−1. Like BSAS, the RE-silicatesre unstable as single layer coatings and so a tri-layer RE-ilicate/mullite/silicon coating approach analogous to the BSASop coat EBC has been proposed.58

There have been few studies of the tri-layer EBC depositionrocess or investigations of the relationships between coat-ng deposition conditions, coating structure, and thermo-cyclicurability. To gain insight into these relationships, we describen air plasma spray (APS) process to deposit tri-layer coatingsn silicon carbide substrates with Yb2SiO5 top coats, and havesed a variety of methods to characterize the coatings. We reporthe presence and identify the formation mechanism of previouslynidentified phases in the Yb2SiO5 and mullite layers, and dis-uss the implications of their presence upon the propensity forracking.

. Experimental

.1. Coating deposition

Three-layer Yb2SiO5/mullite/silicon EBCs were depositednto SiC substrates using an APS approach. The general coat-ng architecture is schematically illustrated in Fig. 2. Coatingsere deposited on 2.54 cm × 1.27 cm × 0.32 cm �-SiC couponsbtained from HexoloyTM (Saint Gobain Ceramics, Niagaraalls, NY). Prior to EBC deposition, the substrates were gritlasted using ∼270 �m diameter SiC grit (Black SiC, Whitebrasives, Niagra Falls, ON CA) resulting in an average sur-

ace roughness Ra = 1 �m, as measured by surface profilometrysing a 500 �m sampling distance. The substrates were thenltrasonically cleaned in ethanol to remove surface contamina-

ion.

The silicon powder used for bond coat deposition wasupplied by Micron Metals (Bergenfield, NJ) with a particle

B.T. Richards, H.N.G. Wadley / Journal of the European Ceramic Society 34 (2014) 3069–3083 3071

Table 1Thermal expansion coefficients of EBC materials, adapted from Ref. 57.

Material CTE (×10−6 ◦C−1) Melting point (◦C) Application

Y2SiO5 5–6 198059

Topcoat

Er2SiO5 5–7, 7–8a 58 198059

Yb2SiO5 3.5–4.5, 7–8a,58 5.7–9.1b 195059

Lu2Si2O7 3.860 UnknownSc2Si2O7 + Sc2O3 5–6 1860c 61

Yb2Si2O7 4–662 185059

Yb2O3 6.8–8.463 2415d 64

BSAS (monoclinic) 4–5 130065

BSAS (hexagonal) 7–8 130065

Mullite 5–6 180066

Intermediate coatAlumina (�) 6.0–8.467 207268

Si 3.5–4.5 140069 Bond coatSiC, SiC/SiC 4.5–5.5 254570

SubstrateSi3N4 3–4 187571

a More recent value reported.b Value determined in Section 3.1 of this article.c Lower-melting eutectic at 1770 ◦C on Sc2O3-rich side and 1660 ◦C on SiO2-rich side.62

d Undergoes pre-melt phase transformation at 2360 ◦C.65

Table 2Measured composition of received ytterbium monosilicate powder, and predic-tions for ytterbium silicates and ytterbia. Data is also shown from individualsplats in an as-deposited coating.

Spectrum location Yb (at%) Si (at%) O (at%)

Powdera 21.0–22.1 13.3–13.6 64.4–65.6Yb2SiO5 25 12.5 62.5Yb2Si2O7 18.2 18.2 63.6Yb2O3 40 0 60Point 1 37.3 4.0 58.7Point 2 32.9 7.0 60.1Point 3 31.6 8.4 60Point 4 29.6 9.9 60.4Point 5 27.4 12.5 60.1P

dbdTofipsoCcT

tcs1wEFH

Fig. 3. Schematic illustration of the APS and substrate heating system used tod

smrddTump1awaTtc

oint 6 26.5 12.4 61.1

a 95% confidence interval.

iameter range of 10–44 �m. The mullite powder providedy Saint Gobain Ceramics (Worcester, MA) had a particleiameter of 16–53 �m. The Yb2SiO5 powder was supplied byreibacher Industrie Inc. (Toronto, ON) with a particle diameterf 20–50 �m. The composition certified powders were identi-ed by their respective manufacturers as fully reacted and phaseure, and confirmed by XRD prior to deposition. The compo-ition of the Yb2SiO5 powder was also measured using EDSn polished cross sections. Using a PB-ZAF algorithm (Brukerorporation, Ewing, NJ), the composition was found to be verylose to that of Yb2SiO5 with variability of about ±0.5 at%,able 2.

The coatings were deposited using two air plasma spray sys-ems at the NASA Glenn Research Center.50 The silicon bondoat was deposited using an ambient substrate temperature APSystem with a Sulzer Metco 9MB torch at a standoff distance of00 mm. The two other coating layers were deposited at 1200 ◦Cithin a box furnace, using a plasma torch originally designed by

lectro-Plasma Inc. (EPI) with a standoff distance of 190 mm,ig. 3. Each spray system had a single Plasmatron Roto Feedopper (Model 1251) with standard powder feed wheel that was

tia

eposit the environmental barrier coatings.

et to 2.2 rpm. Based on a similar powder feed system, we esti-ate the Yb2SiO5 feed rate was ∼27 g/min and the mullite feed

ate was ∼13 g/min. The fineness of the silicon powder led toifficulties with catch can experiments, but we project (based onifferently sized Si powder) that the feed rate was ∼10 g/min.he silicon bond coat was deposited onto unheated substratessing a torch power of 50 kW. The plasma forming gas was a 2:1ixture of Ar and N2 with a flow rate of 3400SLM. The plasma

lume was traversed 3 times across the sample at a speed of250 mm/s using six-axis robotic manipulation. For the mullitend Yb2SiO5 top coat deposition, the silicon coated substratesere heated to 1200 ◦C inside a box furnace containing a small

perture through which APS depositions were performed, Fig. 3.he mullite layer was deposited using a plasma power of 56 kW;

he Yb2SiO5 was deposited at a lower power of 35 kW. Bothoatings were deposited using a 2:1 mixture of Ar and He forhe plasma forming gas and a flow rate of 763SLM. These coat-

ng layers were deposited by traversing the plasma plume twicecross the samples at a speed of 950 mm/s. After fabrication,

3 European Ceramic Society 34 (2014) 3069–3083

s2

2

tttcs3Dnamtutfpui

bctimbtb(e

η

wBrv

uclaeprCsw

Fd

2

uwhwwftawse

3

3

ssscTTwSh7bsi

072 B.T. Richards, H.N.G. Wadley / Journal of the

ome samples were stabilization annealed in air at 1300 ◦C for0 h.

.2. Characterization

Both the as-fabricated and annealed samples were subjectedo XRD analysis using a PANalytical X’Pert PRO MPD diffrac-ometer (Westborough, MA) and Cu K� radiation to identifyop coat phases and volume fractions. Residual stress was cal-ulated for the top coat from XRD measurements using thein2ψ technique73 with χ tilts of 0◦, 18.43◦, 26.56◦, 33.21◦,9.23◦, and 45◦ for diffraction peaks (2θ) between 55◦ and 65◦.iffraction peaks of this low angle reflection were used becauseeither Yb2O3 nor Yb2SiO5 had high angle peaks of measur-ble intensity. The low angle of the peaks prohibits preciseeasurements, but reported figures are reasonable approxima-

ions of the stress state. Peak center fitting was accomplishedsing a Gaussian fitting algorithm (Solver in Microsoft Excel)hat included a slanting background fit. Additionally, both as-abricated and annealed samples were sectioned and diamondolished to 0.25 �m finish, and in some cases vibrapolishedsing 0.02 �m silica colloid suspension. They were then exam-ned by SEM in BSE mode, and with EDS.

The difference in electron backscattering coefficient η

etween regions of disparate composition was utilized duringoating characterization. It relates the flux of backscattered elec-rons emitted from the sample to the flux of incident electronsn the probe. This factor can be calculated for a multicomponent

ixture using a weighted sum: ηmix = �ηiCi where ηi is theackscatter coefficient for the ith element in pure form, and Ci

he weight fraction for the ith element in the compound.74 Theackscatter coefficient for each pure element may be estimatedfor a broad range of accelerating voltages) using an empiricalxpression75:

i = −0.0254 + (0.016)Zi − (1.86 × 10−4)Z2i

+ (8.3 × 10−7)Z3i (1)

here Zi is the atomic number of the ith element. Contrast inSE images was adjusted using gamma-correction over multiple

anges of the 16-bit spectrum to preserve detail of compositionalariation within each coating layer.

The EDS spectra were standardized (except where noted)sing an aggregate spectrum from the respective powders foralibration. To form the calibration standard, spectra were col-ected from a minimum of 20 randomly selected particles of thes-received powder with greater than 2.5 × 105 total counts forach spectrum. The net integrated intensity under the respectiveeaks was summed to attain a total standard spectrum for theespective compound with greater than 5 million total counts.

rack and pore distributions were calculated by image analy-

is using an average from 3 different areas of the same samplehere a 100 �m length of the layer was examined.

3

s

ig. 4. Thermal expansion and CTE data for hot-pressed Yb2SiO5 with a relativeensity of ∼70%.

.3. Dilatometry

We have measured the CTE of the ytterbium monosilicatesing a Netzsch 402 C Dilatometer (Burlington, MA) calibratedith high purity alumina. The Yb2SiO5 sample was prepared byot-pressing the as-received Yb2SiO5 powder for 3 h at 1650 ◦Cith ∼40 MPa load. The hot pressing resulted in a blank thatas ∼70% dense. The Yb2SiO5 dilatometry sample was cut

rom the blank with dimensions of 4 mm × 4 mm × 14 mm. Theests were performed between 20 ◦C and 1450 ◦C using heatingnd cooling rates of 3 ◦C/min. The expansion displacement dataas recorded at 0.05 ◦C steps and used to compute the expan-

ion strain and CTE. CTE curves have been smoothed whereasxpansion curves have not.

. Results and discussion

.1. Yb2SiO5 thermal expansion coefficient

The thermal expansion strain versus temperature relation-hips during heating and cooling of hot pressed Yb2SiO5 arehown in Fig. 4. The hysteresis upon heating and cooling isimilar to that previously reported for RE tungstate’s of lowrystal symmetry, attributed to anisotropic thermal expansion.76

he CTE deduced from the strain data is also shown in Fig. 4.he CTE during heating ranged from 5.7 to 9.1 × 10−6 ◦C−1,hile during cooling it varied from 2.8 to 10.9 × 10−6 ◦C−1.tandard convention is to report the CTE measured during theeating phase of the test, and we report an average CTE of.4 × 10−6 ◦C−1, which supports the most recent value reportedy Lee,57 Table 1. While hysteresis of the thermal expansiontrain was observed, no permanent change in length occurred,ndicating that sintering did not contribute to the measured CTE.

.2. Tri-layer EBC structure

A BSE image of an as-deposited tri-layered EBC system ishown in Fig. 5(a). Detailed measurements over the 1.27 cm

B.T. Richards, H.N.G. Wadley / Journal of the European Ceramic Society 34 (2014) 3069–3083 3073

F icatedr

woaTrtftodstFpta

aFwisooFiwatr

3

adbp

5ct(t1

cbihgcaocep1tfcFw3acthhsps

e

ig. 5. BSE micrographs of Yb2SiO5/Al6Si2O13/Si three layer EBC: (a) as-fabregions where higher magnification micrographs are subsequently analyzed.

ide coating indicated a variation in silicon bond coat thicknessf 75–150 �m, a mullite layer thickness variation of 25–100 �m,nd an Yb2SiO5 top coat thickness between 25 and 100 �m.hese variations resulted from fluctuations in the powder feed

ate (“powder pulsing”, presumably due to low feed RPM) andemperature variation of the particles which impacted the sur-ace: those that had not sufficiently melted did not adhere tohe sample, and those that were too hot may have “splashed”ff. Mud-cracking of the top coat was observed in the as-eposited coatings, Fig. 5(a). These mud-cracks were spacedeveral hundred micrometers apart, and frequently extended intohe mullite layer during the stabilization annealing treatment,ig. 5(b), but were arrested at the silicon bond coat confirmingrevious observations.58 All three layers contained porosity inhe as-deposited condition; it was not removed by stabilizationnnealing.

Adhesion between the silicon bond coat and the SiC substrateppeared good, except in areas with high interfacial porosity,ig. 5(a). Though the pores on the interface were small, theyere numerous with 5–10 pores of 3–5 �m diameter intersect-

ng each 100 �m length of interface. The interface between theilicon bond coat and mullite also contained isolated regionsf weak bonding that appeared to result from incomplete flowf the liquid mullite droplets over the rough silicon bond coat,ig. 5. The length of these regions ranged from a few to approx-

mately 30 �m. On average these more weakly bonded regionsere spaced 100 �m apart. The interface between the mullite

nd the Yb2SiO5 had very little porosity. The several microme-er long pores observed were typically located in areas where theadius of curvature of the underlying mullite coat was smallest.

.3. Yb2SiO5 top coat

Splat-to-splat adherence in the ytterbium monosilicate coat

ppeared very good, though some spherical pores (with poreiameter of <0.5 �m) and more elongated porosity were seen atoth splat-to-splat boundaries and within splats, Fig. 6(a). Thelanar density of these pores in cross-section was approximately

bdw

and (b) after a 20 h stabilization anneal in air at 1300 ◦C. Dashed boxes indicate

0 pores per 1000 �m2 of coating. Pores of diameter <0.5 �montributed about 0.2% to the total porosity while the contribu-ion from larger spherical pores was around 0.7%. Filling defectsnon-spherical/jagged pores) contributed roughly 0.6% to theotal porosity, resulting in a total void volume fraction of about.5%.

Large vertical mud-cracks were present in as-depositedoatings, Fig. 5(a). These mud-cracks were not deflected by splatoundaries or pores, and were oriented nearly normal to the coat-ng’s surface. The mud-cracks were periodically spaced severalundred micrometers apart. The cracks in as-deposited coatingsenerally terminated at the Yb2SiO5/mullite interface and hadrack face opening widths (at ambient temperature) greater than

micrometer. Additional 1–10 �m long microcracks were alsobserved to originate at splat interfaces, Fig. 6(a). Some of theracks were confined to a single solidified droplet while oth-rs had penetrated several solidified droplets. The cross-sectionlanar density of these microcracks was roughly 5 cracks per000 �m2 of top coat examined. Most were oriented normal tohe coating surface, and therefore appear to be a micro-scaleorm of mud-cracking. While this population of micro-mud-racks did not extend during stabilization annealing at 1300 ◦C,ig. 6(b), a new population of equiaxed microcracks developedithin many of the splats. These microcracks had lengths of–10 �m, and the combined planar density of all the microcracksfter annealing had doubled to 10 cracks per 1000 �m2 of topoat, Fig. 6(b). No delamination cracks were found in eitherhe as-deposited or annealed Yb2SiO5 layer. When viewed atigher magnification, Fig. 7(a), many of the splats appear toave deposited with fine dendritic structures with intermediatepacing around 100 nm; these structures stabilized into a two-hase structure upon annealing similar to that expected for thelow cooling of proeutectic phase structures, Fig. 7(b).

The presence of both circular (undeformed) and highlylongated splat shapes in the as-deposited coatings indicates

oth partially and fully melted particles were present duringeposition of the layer, Fig. 6. This variation is consistentith a substantial spread in particle trajectories (and therefore

3074 B.T. Richards, H.N.G. Wadley / Journal of the European Ceramic Society 34 (2014) 3069–3083

F n I in as-deposited micrograph Fig. 5(a), and (b) region IV in the annealed conditionm

tdpTfiYTtbdt

afsTtTcitW

Table 3Atomic compositions for as-deposited ytterbium silicate topcoat calculated fromEDS point probe spectra at 550 rectangular grid points distributed over sixdifferent collection areas.

Parameter Composition (at%)

Average Yb content 24.4 ± 2.7Yb max/min 36.1/17.4Average Si content 10.6 ± 2.3Si max/min 14.3/1.6Average O content 65.0 ± 0.9O

hLsvfio

ig. 6. High magnification micrographs of the Yb2SiO5 topcoats from: (a) regioicrograph Fig. 5(b).

hermal profiles) through the plasma plume.77–81 BSE contrastifferences between individual splats in Fig. 6(a) indicate com-ositional variations in the as deposited ytterbium silicate layer.he contrast is directly related to the electron backscatter coef-cient η of a composition. We calculate (using Eq. (1)) that forb2O3, η = 0.75. It is 0.67 for Yb2SiO5, and 0.61 for Yb2Si2O7.he large differences in η between these three phases indicate

hat small compositional differences in the Yb2O3–SiO2 pseudo-inary system are readily observed by BSE imaging, and theifferences in contrast of the various splats in Fig. 6(a) are dueo variations of their composition.

To quantify the compositional differences of splats in thes-deposited coating, energy dispersive spectra were collectedrom the points indicated in Fig. 6(a). A summary of the compo-ition data as well as that of the as-received powder was given inable 2. EDS analysis indicated significant loss of silicon from

he lightest contrast splats. A summary of the EDS analysis,able 3, indicated some splats contained as little as 2 at% sili-

on. The deposited stoichiometry was on average 0.6 at% lowern Yb, 1.9 at% lower in Si and 2.5 at% higher in oxygen thanhe starting powder, indicating loss of Si with respect to Yb.

e note that preferential loss of silicon during APS processing

spt

Fig. 7. High magnification views of (a) region VII in the as-deposited coatin

max/min 68.9/59.1

as also been suggested for the RE silicate apatite phasea9SrSi6O26.5.82 When the as-deposited splats with significantilicon loss were examined at higher magnification, Fig. 7(a), aery fine, two-phase dendritic structure could be identified. Theneness of this structure precluded accurate EDS identificationf the composition of each phase.

Upon stabilization annealing (1300 ◦C, 20 h, air), many of the

olidified droplets appeared to have undergone a phase decom-osition reaction with a light BSE contrast phase appearing onhe grain boundaries. This can be clearly seen for the center top

g shown in Fig. 6(a), and (b) annealed top coat near a splat boundary.

B.T. Richards, H.N.G. Wadley / Journal of the European Ceramic Society 34 (2014) 3069–3083 3075

Fig. 8. EDS line scan of annealed ytterbium silicate top coat. Dashed linesindicate atomic compositions for Yb2SiO5. Note 40 at% Yb–60 at% O is thecomposition of ytterbia. Location of scan line is indicated on accompanyingm

ssetasTeittctoaito

dvptri(hciB

Fig. 9. Calculated partial pressure versus temperature plot for primary gaseousspecies in the Yb–Si–O system. Curves are reported for an equilibrated system.

F

ch

eipSg1cmdfhtYhleoea

droplets can be understood with the aid of the ytterbia-silica

icrograph.

plat of Fig. 6(b). After stabilization annealing, the dendritictructure of the splats in Fig. 7(a) has been replaced with anquiaxed polycrystalline structure with extremely fine precipi-ates, Fig. 7(b). Again, the precipitate structure was too fine tonalyze via EDS, but compositional fluctuations within the bulktructure after annealing were assessed using a line scan, Fig. 8.he variations in Si, Yb and O are shown together with the ref-rence concentrations of these elements in Yb2SiO5. The resultsndicate that the darker contrast droplets had compositions closeo that of Yb2SiO5 and precipitated a smaller volume fraction ofhe second phase than the silicon depleted droplets. The mediumontrast splats had low silicon contents and decomposed into awo-phase mixture with approximately equal volume fractionsf each phase. The lightest contrast splats precipitated smallmounts of the gray phase in some instances (as can be seenn Fig. 7(b)), but remained largely free of gray phase precipi-ates. The composition of the lightest regions approached thatf ytterbia (Yb2O3).

The silicon content of the solidified droplets in the as-eposited ytterbium monosilicate layer showed considerableariability, Table 2, but was almost always less than that of theowder used for its fabrication. Examination of Fig. 5 showshat while some of the larger diameter powder particles hadetained a spherical shape, indicating they had barely meltedn the plasma plume, others with much lighter BSE contrastindicative of a higher fraction of high atomic number elements)ad flowed significantly upon impingement with the depositingoating. This suggests they had been heated well above the melt-

ng point of ytterbium monosilicate (TM = 1950 ◦C). The lighterSE contrast and the EDS compositional characterization is

pY

ig. 10. Ytterbia-silica pseudo-binary phase diagram (redrawn from Ref. 59).

onsistent with preferential silicon evaporation from the mosteated particles during the plasma spray process.

The partial pressure of the highest vapor pressure species inquilibrium with a mixture of Yb2O3 + SiO2 and plasma form-ng gases at 1 MPa can be calculated using the FactSage softwarerogram.85 Assuming ideal interactions between Yb2O3 andiO2, and 80 g of Yb2SiO5 in 761 standard liters of an inertas mixture with an Ar:He gas volume ratio of 2:1 containing0 ppm of O2, it is possible to calculate the partial pressure of theonstituents as a function of temperature, Fig. 9. Recent thermalodels of plasma spray processes predict temperatures of pow-

er particles in plasma spray plumes in the 2000–3500 ◦C rangeor spray parameters similar to those used here.86–90 Particleseated into this temperature range have SiO activities (vapor par-ial pressures) that are many orders of magnitude above those ofb or any of its oxides. We therefore conclude that during super-eated molten particle transport through the plasma plume, theoss of SiO significantly exceeds that of Yb. This will be exac-rbated for particles that are heated to the highest temperaturesr for the longest times; i.e. smaller diameter particles or thosentrained in the core of the plasma plume where temperaturesre the highest.86–90

The phases that form during solidification of silicon depleted

seudo-binary phase diagram, Fig. 10.59 Silicon loss fromb2SiO5 line compound shifts the composition into a two phase

3076 B.T. Richards, H.N.G. Wadley / Journal of the European Ceramic Society 34 (2014) 3069–3083

F depoR rn in

YtwTicdcwYssStgsc

hpt

abbpXo2ltF

ig. 11. XRD patterns for the nominally ytterbium monosilicate topcoat: (a) as-eference peaks for both phases in (a) and calculated Rietveld refinement patte

b2O3 + Yb2SiO5 region of the phase diagram. This region con-ains a eutectic point, and so the as-solidified structure that formsill be dependent upon a droplet’s composition. Examination ofable 3 shows that some particles contained only a tenth the sil-

con fraction of the original powder. Under equilibrium coolingonditions, these would nucleate solid proeutectic Yb2O3 den-rites, and a eutectic Yb2SiO5 + Yb2O3 inter-dendritic structureonsistent with the result shown in Fig. 7(a). Liquid dropletsith hypereutectic compositions would precipitate proeutecticb2SiO5 and an inter-dendritic eutectic structure. The rapid

olidification rate of the droplets is likely to result in a verymall dendrite spacing that would be difficult to resolve with theEM; but on annealing, these two microstructures are expected

o evolve to the two types of coarser structure seen in micro-

raphs such as Figs. 6(b) and 7(b). The lightest contrast (mostilicon depleted) regions consisted of a continuous Yb2O3 phaseontaining particles of Yb2SiO5, Fig. 7(b), whereas those with

rbi

sited and (b) annealed. Peaks are indexed for Yb2O3 in (a) and Yb2SiO5 in (b).(b) are also shown.

yper-eutectic compositions consisted of equiaxed Yb2SiO5olycrystals with grain boundary Yb2O3 particles as seen inhe darker contrast splats in Fig. 6(b).

XRD patterns of both the as-deposited and annealed top coatre shown in Fig. 11. The reflection peaks in both patterns cane indexed using the diffraction patterns of monoclinic ytter-ium monosilicate and cubic ytterbia. A hump in the diffractionattern at low 2θ (below 20◦, not shown) was present in theRD patterns of as-deposited top coats indicating the presencef amorphous material. After the annealing, the absence of a lowθ hump in the patterns indicated the coating had fully crystal-ized. A Rietveld analysis83 of the annealed coating indicateshe composition to be 85 vol% Yb2SiO5 and 15 vol% Yb2O3,ig. 11(b). The as-deposited ytterbia exhibited peak shifts with

espect to the ytterbia reference pattern; for example the ytter-ia {2 2 2} peak was shifted +0.5◦. This peak shift is not seenn annealed coatings and corresponds to a contraction in lattice

B.T. Richards, H.N.G. Wadley / Journal of the European Ceramic Society 34 (2014) 3069–3083 3077

Table 4Material parameters and calculated thermal stress at 20 ◦C for coating layers assuming 1300 ◦C stress free temperature and �-SiC substrate.

Coating layer Young’s modulus (GPa) CTE (� × 106) ◦C−1 Poisson ratio ν Thermal stress (MPa)

Yb2SiO5 (monoclinic) 172a 7.5 (58, Section 3.1) 0.27b 893 (tensile)Yb2SiO5 (monoclinic) 172a 4.057 0.27b 211 (compressive)Yb2O3 (cubic) 18895 7.663 0.304c 1013 (tensile)3:2 Mullite (orthorhombic) 22097 5.350 0.2898 246 (tensile)Alumina (�) 40267 7.267 0.23167 1693 (tensile)Silicon (dcc) 16369 4.169 0.22369 153 (compressive)Silicon carbide (�) 430 4.67 0.14 N/A

a Determined by nano-indentation.

pr

cctfimstvc

tsfiwsaaaPitlal

σ

wmCdthmsm

opt

t7mtTtYi

ctiss(dsimammdbtb

dfclCpecmtrYa

b Value taken for Y2Si2O7.94

c Value taken for Y2O3.96

arameter in the out-of-plane (coating normal) direction withespect to the reference crystal lattice parameter.

A XRD residual stress measurement of the as-deposited topoat was performed using the ytterbia {6 2 2} peaks, and indi-ated the ytterbia in the top coat at room temperature to have aensile residual stress of 425 MPa with a sin2ψ correlation coef-cient of R2 = 0.74. For the annealed top coat only ytterbiumonosilicate peaks had high enough intensities for stress mea-

urements. However, since Yb2SiO5 is monoclinic,59 the roomemperature stress state between individual grains is expected toary widely, and the sin2ψ measurement technique was incon-lusive.

The observed mud-cracking is consistent with the substan-ial CTE difference between the top coat and the rest of theystem, as suggested by a recently reported CTE measurementor Yb2SiO5

57 and confirmed by our dilatometry measurementsn Section 3.1, Table 1. To investigate the origin of cracking,e performed an elastic thermomechanical calculation of the

tress state in the coating system using a previously describedpproach that used a thin film approximation.91 The thin filmpproximation assumes curvature of the composite system and

biaxial stress state in coating layers whose elastic moduli,oisson’s ratio and CTE are different. It is also assumes that the

nteractions between coating layers are small in comparison tohe interaction with the substrate, and the stress in each coatingayer is therefore independent of that in adjacent layers. Thispproach leads to a thermal residual stress in the ith coatingayer (prior to relaxation by cracking) given by:

i = (Ei/(1 − νi))(αs − αi) T (2)

here σi is the thermal stress in the ith layer, Ei is the elasticodulus of the ith layer, νi is its Poisson’s ratio, αs and αi are theTE of the substrate and ith layer respectively, and T is theifference in temperature between the stress free (deposition)emperature of the ith layer and the temperature of interest. Weave verified the validity of this approach for our system using aore accurate solution formulated by Hsueh92 where the layer

tresses are interdependent: the stresses calculated using the twoethods were identical.

Residual thermal stresses have been calculated for all layers

f the coating and secondary phases using the thermophysicalroperties of each layer and assuming the stress free temperatureo be the annealing temperature (1300 ◦C), Table 4. If we assume

cF

he top coat to be pure ytterbium monosilicate with a CTE of.5 × 10−6 ◦C−1 (as proposed by Lee57 and consistent with oureasurements), we calculate that the top coat develops a biaxial

ensile stress of 893 MPa upon cooling to ambient temperature.he stress calculated for an Yb2O3 enriched Yb2SiO5/Yb2O3

op coat would be somewhere between the values calculated forb2SiO5 and Yb2O3: the stress will increase moderately with

ncreasing Yb2O3 content, Table 4.The mud-cracking observed in as-deposited and annealed

oatings is consistent with the high tensile stress calculated forhe top coat using Eq. (2). If such a tensile stress exists upon cool-ng from deposition, the lattice parameter of all phases should belightly larger in the in-plane directions due to the biaxial tensiletress state, and slightly contracted in the out-of-plane directionsdue to the Poisson effect). This is consistent with analysis of theiffraction pattern of Fig. 11(a). The measured residual tensiletress of 425 MPa in the Yb2O3 phase of the as-deposited coat-ng is about half that calculated in Table 4, presumably because

ud-cracking had partially relaxed the strain in this layer. Fornnealed top coats, only diffraction from monoclinic ytterbiumonosilicate could be measured. However, the anisotropic ther-al expansion properties of this crystal class lead to angular

istortions93 and a room-temperature stress state which is notiaxial at the grain size scale: accurate stress measurement usinghe sin2ψ method was not possible, so stress evolution could note monitored using this technique.

Micro-mud-cracking was observed in the most siliconepleted droplets, Fig. 6(a). If we assume a rule of mixturesor the CTE and elastic modulus of such splats, then the lightestontrast splats will have a slightly greater CTE and elastic modu-us than dark contrast splats (Table 4). When the slightly higherTE is combined with the increased elastic modulus of theseredominantly ytterbia splats, the small, ytterbia rich splats willxperience a state of triaxial tension (even in the mud-crackedoating) upon cooling due to their higher CTE. The strain energyay then be released by microcracking. We therefore conclude

hat in the most intensely heated, light contrast splats, evapo-ation of silicon in the molten state leads to the formation ofb2O3 which is susceptible to microcracking due to high macro

nd local tensile thermal residual stresses.

Upon annealing the top coat, additional “equiaxed” micro-

racking was observed in the darker BSE contrast droplets,ig. 6(b). Increases in microcracking upon annealing have been

3078 B.T. Richards, H.N.G. Wadley / Journal of the European Ceramic Society 34 (2014) 3069–3083

F (a) region II in as-deposited Fig. 5(a) and (b) region V in annealed Fig. 5(b).

octpcmut

3

odStIdflcwptctsnrg

td1dr(lhi

Table 5Atomic compositions for as-deposited mullite coat calculated from EDS pointprobe spectra at 550 rectangular grid points distributed over six different collec-tion areas. All values are in at%.

Parameter Composition (at%)

Average Al content 34.9 ± 1.6Al max/min 39.9/27.1Average Si content 10.5 ± 1.6Si max/min 19.2/5.2Average O content 54.6 ± 0.8O

lpettry

wtaFcpsS1bmr

im

ig. 12. High magnification micrographs of the mullite intermediate layer from

bserved previously in EBCs and have been linked to volumehanges during phase transformation.50,65,99–101 We note thathe droplets where microcracking was most prevalent consistedrimarily of Yb2SiO5. If a substantial difference in CTE withrystallographic orientation exists as in Y2SiO5

102 and otheronoclinic structures,76 anisotropic thermal stresses will occur

pon cooling between Yb2SiO5 crystals that may have con-ributed to the observed cracking.

.4. Mullite intermediate layer

A higher magnification micrograph of region II from Fig. 5(a)f the as-deposited mullite layer is shown in Fig. 12(a). The as-eposited mullite layer contained a variety of pore structures.ome larger, rounded pores had diameters of several microme-

ers while flattened pores existed at some inter-splat boundaries.nter-splat pores have been reported to result from filling defectsuring plasma spray processing.84 We calculate that rounded andattened pores contributed roughly 1% to total porosity of theoating layer. The majority of the mullite coat was nanoporousith a nanopore diameter of 10–100 nm and a cross-sectionalore density of approximately 7000 pores per 1000 �m2, con-ributing 7% to total porosity of the layer. The total mulliteoat porosity was ∼8%. The as-deposited mullite structure waswo-phase with a plate-like morphology second phase. Aftertabilization annealing, the number of plates had increased sig-ificantly and the spacing between plates decreased. Dense,ecrystallized areas of the mullite matrix with columnar crystalrowth were evident in the microstructure, Fig. 12(b).

EDS analysis of the as-deposited mullite coats indicateshey were oxygen and silicon deficient, Table 5. The averageeposited composition of the layer was 6.3 at% higher in Al,

at% higher in Si and 7.3 at% lower in O than the starting pow-er, indicating volatilization of Si and O. The deposited Al:Siatio lies between 3:2 and 2:1 mullite, having a ratio of 3.5:1

2:1 mullite is 4:1 Al:Si and 3:2 mullite is 3:1 Al:Si). The plate-ike features seen in Fig. 12(a) were slightly rich in aluminum,aving aluminum contents 1–3 at% higher than the surround-ng matrix. An EDS line scan analysis of the annealed mullite

dp3T

max/min 56.9/51.4

ayer, Fig. 13, indicates the stoichiometric difference betweenlate-like features and the surrounding matrix increased consid-rably during annealing. The EDS analysis indicated the plateso be composed of Al2O3 containing 1–5 at% silicon, whilehe surrounding matrix composition approached 3Al:1Si cor-esponding to 3:2 mullite. Regions with no precipitated aluminaielded EDS compositions close to 3:2 mullite.

The presence of alumina plates in APS mullite is consistentith previous XRD and SEM observations.50 Higher resolu-

ion analysis showed the plates were composed of sub-plates oflumina with fine bands of a silicon-rich species interspersed,ig. 14. Analysis of a single area at high magnification indi-ated that the plate region was composed of 76 vol% aluminalatelets, 22 vol% silicon-rich species, and 2% porosity. Theilicon-rich bands were too small to analyze with EDS in theEM, having widths ranging from several nanometers to roughly00 nanometers, but BSE imaging contrast indicates that theseands may be similar in composition to the surrounding mulliteatrix. Several plates in the annealed coating were microc-

acked, Figs. 12(b) and 13.We have observed that the APS melting and deposition of

nitially stoichiometric 3:2 mullite powder resulted in an alu-ina rich, near 2:1 mullite coating, Table 5. The composition

eposited here is similar to that of previous studies on melt

rocessed mullite,66,103 but was less silicon deficient than the:1 stoichiometry coating previously reported for EBCs.50,104

he FactSage software85 can again be used to determine the

B.T. Richards, H.N.G. Wadley / Journal of the European Ceramic Society 34 (2014) 3069–3083 3079

Fig. 13. EDS line scan of annealed mullite layer. The initial 3Al2O3:2SiO2 molarmullite composition is indicated by dashed lines. Note that 40 at% Al–60 at% Ois the composition of alumina. The location of the line scan is indicated on themicrograph.

Fs

eoitoacdsoda

Fig. 15. Calculated partial pressure vs. temperature plot for primary gaseousspecies in the Al–Si–O system. Data are reported for a fully equilibrated system.

Ft

FpTtdiTptomscgtpast

es

ig. 14. High magnification view of phase decomposition into alumina plateeparated by silicon rich material in the annealed mullite coat.

quilibrium partial pressures of the gas species for a mixturef 3Al2O3 + 2SiO2 and the plasma forming gas assuming idealnteractions. Fig. 15 shows the calculated vapor pressure versusemperature relations for the components of a system consistingf 35 g of Al2O3 and SiO2 (with a stoichiometry of Al6Si2O13)nd 761 standard liters of 2:1 Ar:He volume fraction ratio gasontaining 10 ppm O2. The particle temperatures during plasmaeposition lie in the 2000–3500 ◦C range where the vapor pres-ure of SiO is several orders of magnitude higher than those

f any aluminum containing vapor species. This vapor pressureifference therefore results in preferential evaporation of SiOnd formation of a compositionally modified mullite layer.

2dt

ig. 16. Alumina–silica pseudo-binary phase diagram. Dashed lines indicatehe boundaries of metastable phases (modified and redrawn from Ref. 65).

A phase diagram for the SiO2–Al2O3 system is shown inig. 16.66 Dashed lines for 2:1 and 3:1 mullite indicate com-ositional boundaries for the formation of metastable mullites.he silicon deficient 2:1 mullite we have deposited lies in

he alumina + mullite two-phase field of the equilibrium phaseiagram. Examination of Fig. 12(a) indicates that a predom-nantly single-phase 2:1 composition mullite was deposited.his then transformed upon annealing at 1300 ◦C into a two-hase structure consisting of Al2O3 + mullite, consistent withhe as-deposited composition and low temperature metastabilityf non-stoichiometric mullites.105 After annealing, areas in theullite coat also have dense, columnar recrystallized grains with

imilar BSE contrast to the surrounding matrix. High aspect ratioolumnar mullite grains have been previously observed in fine-rained (recrystallized) 3:2 mullite with growth textured alonghe “c” axis.106 These grains are interspersed with a very finelate-morphology phase that appears similar in BSE contrast tolumina. The occurrence of such recrystallized regions is con-istent with a phase transformation from metastable 2:1 mulliteo equilibrium 3:2 mullite + alumina.

The as-deposited mullite layer had not mud-cracked. How-ver, some areas of delamination between this layer and theilicon bond coat were observed, Fig. 12(a). During annealing

5–100 �m long intra-layer microcracking between solidifiedroplets was observed, as well as mud-crack extension fromhe top coat, Fig. 12(b). These observations are consistent with

3080 B.T. Richards, H.N.G. Wadley / Journal of the European Ceramic Society 34 (2014) 3069–3083

F III in as-deposited Fig. 5(a), and (b) region VI after annealing in air (20 h at 1300 ◦C)f

eachsiaa7fith(

3

cccppp

wadnicetat

te

Table 6Atomic compositions reported by PB-ZAF (non-standardized) analysis of pointprobe spectra in Fig. 17(b).

Spectrum Si (at%) O (at%)

Point 1 27.0 73.0Point 2 26.5 73.5Point 3 27.3 72.7Point 4 93.7 6.3a

Point 5 94.2 5.8a

Point 6 93.8 6.2a

Silica (SiO2) 33.3 66.7

a Trace oxygen resulted from native oxide growth after sample preparationand oxygen detection inaccuracy in PB-ZAF analysis.

Fig. 18. High magnification view corresponding to region VIII in Fig. 5(b)sf

t

ig. 17. High magnification micrographs of the silicon bond coat from (a) regionrom Fig. 5(b).

arlier findings of APS deposited mullite coatings.50,65 The over-ll crack density was low with roughly 1 crack per 1000 �m2 ofross section examined. Previous work in APS deposited mulliteas suggested that mullite layers can microcrack as a result ofecond phase alumina or silica precipitation.51 The microcrack-ng of the alumina plates after annealing, Figs. 12(b) and 13,ppears to be a consequence of the CTE mismatch betweenlumina and mullite since �-alumina has an average CTE of.2 × 10−6 ◦C−1, which is substantially higher than that reportedor surrounding mullite (5.5 × 10−6 ◦C−1, Table 1). Upon cool-ng to ambient temperature from the annealing temperature,his CTE difference would induce triaxial tensile stresses muchigher than those experienced in the rest of the mullite coatTable 4) that are conducive to alumina microcracking.

.5. Silicon bond coat

The region marked as III in Fig. 5(a) of the as-deposited bondoat is shown at higher magnification in Fig. 17(a). The bond coatontained significant porosity located at splat boundaries. Typi-al pore diameter was 3–5 �m. There were approximately 5–10ores per 100 �m of inter-splat interface. The many debonds andores between the splats resulted in a network of interconnectedores and debonds within the bond coat.

Upon annealing in air, the interconnected pore-crack net-ork became internally covered in a darker contrast phase. This

ppeared to be silica and was verified by EDS, Table 6. Oxi-ation of the pore surfaces resulted in an interconnected silicaetwork throughout the bond coat, Fig. 17(b). The measured sil-ca thickness was ∼500 nm throughout the majority of the bondoat, though greater thicknesses were observed near the coatingdges. Mud-cracks that had extended from the top coat throughhe mullite layer had terminated at the silicon/mullite interfacend did not show any sign of oxidation, Fig. 18, indicating that

hese cracks formed upon cooling after the stabilization anneal.

The presence of porosity both within the Si coating and athe SiC/Si interface resulted from several causes. The pres-nce of inter-splat voids, Fig. 17(a), is consistent with both the

fistA

howing a mud-crack penetrating the mullite coat and terminating in the upperew micrometers of the silicon bond coat.

emperature of some powder particles at impact being insuf-cient to allow complete liquid droplet flow over the coating

urface and others being so hot that porous regions form dueo droplet “splashing” that creates micrometer sized particles.dditionally, because the silicon bond coat was deposited in

Euro

arsot

saRssGbftpptidtotwapt

4

biS

1

2

3

4

A

McbfwlKoN

R

B.T. Richards, H.N.G. Wadley / Journal of the

rgon shrouded air, oxygen can be entrained in the plasma plumeesulting in rapid formation of a native oxide on liquid particleurfaces that impedes inter-splat bonding.107 The combinationf these effects has resulted in a less than ideal bond coat struc-ure.

After annealing, the appearance of a roughly 500 nm thickilica layer was observed on the interior surface of many voidsnd inter-splat debonds within the silicon bond coat, Fig. 17(b).egions closest to the edges of the sample coupon had silica

cale thicknesses exceeding 600 nm. The growth of such silicacales is well documented and accepted to follow the Deal androves model.108 Using the Deal and Groves analysis, we haveack-calculated the oxygen partial pressure during exposure toorm a 500 nm silica scale in 20 h at 1300 ◦C. The oxygen par-ial pressure calculated is 0.13 atm, or 60% of the oxygen partialressure found in dry air at atmospheric pressure (101 kPa): thisartial pressure is too great to result from bulk diffusion throughhe coating. An oxygen partial pressure lower than the surround-ng environment implies that oxygen was forced to follow a gasiffusion route as opposed to a convective route to the bond coat;he relatively high partial pressure back-calculated is reflectivef the rapid gas diffusion kinetics of oxygen at high tempera-ures. The necessity of molecular diffusion transport combinedith the long diffusion path through silicon channels that are

ctively absorbing oxygen may have yielded lower O2 partialressures in pores located far from coating edges and close tohe substrate.

. Conclusions

An APS process was used to deposit a tri-layer environmentalarrier coating system consisting of a silicon bond coat, a mullitenterdiffusion barrier, and an Yb2SiO5 top coat on monolithic �iC substrates. We find that:

. While the APS process is able to deposit tri-layer coatings,many defects are introduced during the process. Theseinclude micro and nanoscale porosity formed by incompletemolten droplet flow during splat impact, inter-splat delami-nation within the silicon bond coat forming an interconnectedpore and crack network, and mud-crack formation in theYb2SiO5 top coat resulting from in-plane biaxial thermalcontraction during cooling.

. The majority of the solidified droplets in the as-depositedYb2SiO5 and mullite layers were depleted in silicon. Thisled to the formation of two-phase top coats consisting ofa mixture of Yb2SiO5 and Yb2O3 as well as 2:1 mullitelayers deposited from 3:2 mullite powders. The loss of sili-con was consistent with preferential silicon evaporation fromsuperheated powder particles; a consequence of the highervapor pressure of silicon bearing species at the temperaturesexperienced during transient plasma heating.

◦

. Annealing in air at 1300 C resulted in mud-crack exten-sion from the top coat into the mullite layer and oxidation ofthe interconnected crack and pore network in the bond coat.The precipitation of Yb2O3 combined with the anisotropic

pean Ceramic Society 34 (2014) 3069–3083 3081

thermal expansion of Yb2SiO5 in the top coat resulted inmicrocracking.

. Upon annealing, the 2:1 as-deposited mullite transformedto a mixture of 3:2 mullite + alumina consistent with phaseequilibria in this system. Additional microcracking occurredwithin the precipitated alumina plates due to significant CTEmismatch between alumina and the coating system.

cknowledgements

We express our gratitude to Dennis Fox, Bryan Harder, andichael Cuy at NASA Glenn who helped us to deposit the APS

oatings described in this research. We are also grateful to Eliza-eth Opila for advice on thermochemistry and Md Shamsujjohaor aid in residual stress measurement. Some imaging in thisork was performed at ICTAS NCFL. Rolls Royce (Indianapo-

is) supported some aspects of this work and we are grateful to. Lee of that organization for helpful discussions. The Officef Naval Research (ONR) also supported this study under grant00014-11-1-0917 (program manager Dr. David Shifler).

eferences

1. Boyce MP. Gas turbine engineering handbook. Butterworth-Heinemann;2011.

2. Giampaolo T. The gas turbine handbook: principles and practices. MarcelDekker Incorporated; 2003.

3. Walsh PP, Fletcher P. Gas turbine performance. Blackwell ScienceLimited; 2004.

4. Perepezko JH. The hotter the engine, the better. Science2009;326(5956):1068–9.

5. DeMasi-Marcin JT, Gupta DK. Protective coatings in the gas turbineengine. Surf Coat Technol 1994;68–69:1–9.

6. Williams JC, Starke Jr EA. Progress in structural materials for aerospacesystems. Acta Mater 2003;51(19):5775–99.

7. McBride BJ, Gordon S. Computer program for calculation of complexchemical equilibrium compositions and applications, II. Users manual andprogram description. NASA reference publication 1311; 1996. p. 84–5.

8. Schafrik R. Personal Communication; 2008.9. Evans AG, Mumm D, Hutchinson J, Meier G, Pettit F. Mechanisms

controlling the durability of thermal barrier coatings. Prog Mater Sci2001;46(5):505–53.

10. Evans A, Hutchinson J. The mechanics of coating delamination in thermalgradients. Surf Coat Technol 2007;201(18):7905–16.

11. Zhao H, Yu Z, Wadley HN. The influence of coating complianceon the delamination of thermal barrier coatings. Surf Coat Technol2010;204(15):2432–41.

12. Begley MR, Wadley HN. Delamination of ceramic coatings with embeddedmetal layers. J Am Ceram Soc 2011;94(s1):s96–103.

13. Begley MR, Wadley HN. Delamination resistance of thermal bar-rier coatings containing embedded ductile layers. Acta Mater2012;60(6):2497–508.

14. Davis A, Evans A. Some effects of imperfection geometry on the cyclicdistortion of thermally grown oxides. Oxid Met 2006;65(1–2):1–14.

15. Davis A, Evans A. Effects of bond coat misfit strains on the rumpling ofthermally grown oxides. Metall Mater Trans A 2006;37(7):2085–95.

16. Yu Z, Zhao H, Wadley HN. The vapor deposition and oxidation ofplatinum-and yttria-stabilized zirconia multilayers. J Am Ceram Soc

2011;94(8):2671–9.

17. Krämer S, Yang J, Levi CG, Johnson CA. Thermochemical interaction ofthermal barrier coatings with molten CaO–MgO–Al2O3–SiO2 (CMAS)deposits. J Am Ceram Soc 2006;89(10):3167–75.

3 Euro

082 B.T. Richards, H.N.G. Wadley / Journal of the

18. Krämer S, Yang J, Levi CG. Infiltration-inhibiting reaction of gadoliniumzirconate thermal barrier coatings with CMAS melts. J Am Ceram Soc2008;91(2):576–83.

19. Wellman R, Nicholls J. Erosion, corrosion and erosion–corrosion of EBPVD thermal barrier coatings. Tribol Int 2008;41(7):657–62.

20. Nicholls J, Deakin M, Rickerby D. A comparison between the erosionbehaviour of thermal spray and electron beam physical vapour depositionthermal barrier coatings. Wear 1999;233:352–61.

21. Chen X, He M, Spitsberg I, Fleck N, Hutchinson J, Evans A. Mechanismsgoverning the high temperature erosion of thermal barrier coatings. Wear2004;256(7):735–46.

22. Fleck N, Zisis T. The erosion of EB-PVD thermal barrier coatings: thecompetition between mechanisms. Wear 2010;268(11):1214–24.

23. Chen X, Wang R, Yao N, Evans A, Hutchinson J, Bruce R. Foreign objectdamage in a thermal barrier system: mechanisms and simulations. MaterSci Eng A 2003;352(1):221–31.

24. Christin F. Design, fabrication, and application of thermostructural com-posites (TSC) like C/C, C/SiC, and SiC/SiC composites. Adv Eng Mater2002;4(12):903–12.

25. Igawa N, Taguchi T, Nozawa T, et al. Fabrication of SiC fiber reinforcedSiC composite by chemical vapor infiltration for excellent mechanicalproperties. J Phys Chem Solids 2005;66(February–April (2–4)):551–4.

26. Park JY. Fabrication of SiC fiber–SiC matrix composites by reaction sin-tering. J Kor Ceram Soc 2008;45(4):204–7.

27. Naslain R, Christin F. SiC-matrix composite materials for advanced jetengines. MRS Bull 2003;28(9):654–8.

28. Evans AG, Zok FW, Davis J. The role of interfaces in fiber-reinforcedbrittle matrix composites. Compos Sci Technol 1991;42(1–3):3–24.

29. Rühle M, Evans AG. High toughness ceramics and ceramic composites.Prog Mater Sci 1989;33(2):85–167.

30. Marshall DB, Evans AG. Failure mechanisms in ceramic-fiber/ceramic-matrix composites. J Am Ceram Soc 1985;68(5):225–31.

31. Evans AG, Marshall DB. Overview no. 85 The mechanical behavior ofceramic matrix composites. Acta Metall 1989;37(10):2567–83.

32. Wilson DM, Visser LR. High performance oxide fibers for metaland ceramic composites. Compos A: Appl Sci Manuf 2001;32(8):1143–53.

33. Wilson DM, Lueneburg DC, Lieder SL. High temperature properties ofNextel 610 and alumina-based nanocomposite fibers. In: Proceedings of the17th annual conference on composites and advanced ceramic materials:ceramic engineering and science proceedings. John Wiley & Sons Inc.;2008. p. 609–21.

34. Wilson DM, Visser LR. NextelTM 650 Ceramic Oxide Fiber: new alumina-based fiber for high temperature composite reinforcement. In: 24th annualconference on composites, advanced ceramics, materials, and structures:B: ceramic engineering and science proceedings. John Wiley & Sons, Inc.;2008. p. 363–73.

35. Zawada LP, Hay RS, Lee SS, Staehler J. Characterization and high-temperature mechanical behavior of an oxide/oxide composite. J AmCeram Soc 2003;86(6):981–90.

36. Tressler RE. Recent developments in fibers and interphases for hightemperature ceramic matrix composites. Compos A: Appl Sci Manuf1999;30(4):429–37.

37. Ruggles-Wrenn M, Musil S, Mall S, Keller K. Creep behavior of NextelTM

610/Monazite/Alumina composite at elevated temperatures. Compos SciTechnol 2006;66(13):2089–99.

38. Levi CG, Yang JY, Dalgleish BJ, Zok FW, Evans AG. Processingand performance of an all-oxide ceramic composite. J Am Ceram Soc1998;81(8):2077–86.

39. DiCarlo J, Yun H-M, Morscher G, Bhatt R. SiC/SiC composites for 1200 ◦Cand above. In: Handbook of ceramic composites. Springer; 2005. p. 77–98.

40. Zhu S, Mizuno M, Nagano Y, Cao J, Kagawa Y, Kaya H. Creep and fatiguebehavior in an enhanced SiC/SiC composite at high temperature. J AmCeram Soc 1998;81(9):2269–77.

41. Morscher GN, Ojard G, Miller R, et al. Tensile creep and fatigue ofSylramic-iBN melt-infiltrated SiC matrix composites: retained proper-ties, damage development, and failure mechanisms. Compos Sci Technol2008;68(15–16):3305–13.

pean Ceramic Society 34 (2014) 3069–3083

42. Yun HM, DiCarlo JA. Comparison of the tensile, creep, and rupture strengthproperties of stoichiometric SiC fibers. In: Proceedings of the 23rd annualconference on composites, materials and structures. 1999. p. 259–72.

43. Takeda M, Sakamoto J, Saeki A, Imai Y, Ichikawa H. High performancesilicon carbide fiber Hi-Nicalon for ceramic matrix composites. In: CeramEng Sci Proc. 1994. p. 37–44.

44. Costello JA, Tressler RE. Oxidation kinetics of silicon carbide crystals andceramics: I. In dry oxygen. J Am Ceram Soc 1986;69(9):674–81.

45. Opila EJ. Oxidation and volatilization of silica formers in water vapor. JAm Ceram Soc 2003;86:1238–48.

46. Opila EJ, Fox DS, Jacobson NS. Mass spectrometric identification ofSi–O–H(g) species from the reaction of silica with water vapor at atmo-spheric pressure. J Am Ceram Soc 1997;80:1009–12.

47. Opila EJ, Hann Jr RE. Paralinear oxidation of CVD SiC in water vapor. JAm Ceram Soc 1997;80:197–205.

48. Opila EJ, Smialek JL, Robinson RC, Fox DS, Jacobson NS. SiC recessioncaused by SiO2 scale volatility under combustion conditions: II. Thermody-namics and gaseous-diffusion model. J Am Ceram Soc 1999;82:1826–34.

49. Opila EJ. Variation of the oxidation rate of silicon carbide with water-vaporpressure. J Am Ceram Soc 1999;82:625–36.

50. Lee KN, Miller RA, Jacobson NS. New generation of plasma-sprayedmullite coatings on silicon carbide. J Am Ceram Soc 1995;78:705–10.

51. Lee KN. Key durability issues with mullite-based environmental bar-rier coatings for Si-based ceramics. J Eng Gas Turbines Power2000;122:632–6.

52. Lee KN, Miller RA. Development and environmental durability of mulliteand mullite/YSZ dual layer coatings for SiC and Si3N4 ceramics. Surf CoatTechnol 1996;86–87:142–8.

53. He MY, Hutchinson JW, Evans AG. Simulation of stresses and delami-nation in a plasma-sprayed thermal barrier system upon thermal cycling.Mater Sci Eng A 2003;345(1–2):172–8.

54. Evans AG, Hutchinson JW. The mechanics of coating delamination inthermal gradients. Surf Coat Technol 2007;201(18):7905–16.

55. Ambrico JM, Begley MR, Jordan EH. Stress and shape evolution of irreg-ularities in oxide films on elastic–plastic substrates due to thermal cyclingand film growth. Acta Mater 2001;49(9):1577–88.

56. Hsueh CH. Thermal stresses in elastic multilayer systems. Thin Solid Films2002;418(2):182–8.

57. Lee KN. In: (NETL) NETL, editor. The gas turbine handbook. UnitedStates Department of Energy (DOE); 2006.

58. Lee KN, Fox DS, Bansal NP. Rare earth silicate environmental barriercoatings for SiC/SiC composites and Si3N4 ceramics. Corros CeramMatrix Compos 2005;25:1705–15.

59. Bondar IA. Rare-earth silicates. Ceram Int 1982;8:83–9.60. Fukudome T, Tsuruzono S, Karasawa W, Ichikawa Y. Development and

evaluation of ceramic components for gas turbine. In: ASME Paper GT-2002-30627.; 2002.

61. Toropov N, Vasil’eva V. Phase diagram of the scandia—silica binary sys-tem. Zh Neorg Khim 1962;7:1938–45.

62. Chen H, Gao Y, Liu Y, Luo H. Hydrothermal synthesis of ytterbium silicatenanoparticles. Inorg Chem 2010;49(4):1942–6.

63. Stecura S, Campbell WJ. Thermal expansion and phase inversion of rare-earth oxides: bureau of mines. College Park Metallurgy Research Center,Md.; 1960.

64. Zinkevich M. Thermodynamics of rare earth sesquioxides. Prog Mater Sci2007;52(4):597–647.

65. Lee KN, Fox DS, Eldridge JI, et al. Upper temperature limit of environ-mental barrier coatings based on mullite and BSAS. J Am Ceram Soc2003;86:1299–306.

66. Aksaf IA, Pask JA. Stable and metastable equilibria in the systemSiO2–Al2O3. J Am Ceram Soc 1975;58(11–12):507–12.

67. Munro RG. Evaluated material properties for a sintered alpha-alumina. JAm Ceram Soc 1997;80(8):1919–28.

68. Patnaik P. Handbook of inorganic chemistry. The McGraw-Hill Compa-

nies; 2002. p. 344. ISBN 0-07-049439-8.

69. Hull R. Properties of crystalline silicon. IET; 1999.70. Olesinski RW, Abbaschian GJ. The C–Si (Carbon–Silicon) system. Bull

Alloy Phase Diagrams 1984;5(5):486–9.

Euro

1

1

1

1

1

1

1

1

1

G

CB

B.T. Richards, H.N.G. Wadley / Journal of the

71. Ma X, Li C, Zhang W. The thermodynamic assessment of theTi–Si–N system and the interfacial reaction analysis. J Alloys Compd2005;394(1–2):138–47.

72. Lee KN. Current status of environmental barrier coatings for Si-basedceramics. Surf Coat Technol 2000;133–134:1–7.

73. Fitzpatrick M, Fry A, Holdway P, Kandil F, Shackleton J, Suominen L.Determination of residual stresses by X-ray diffraction; 2005.

74. Heinrich KFJ. In: Castaing R, Deschamps P, Philibert J, editors. X-rayoptics and microanalysis, 4th international congress on X-ray optics andmicroanalysis. Paris: Hermann; 1966. p. 1509.

75. Reuter W. The ionization function and its application to the electron probeanalysis of thin films. In: Proceedings 6th international congress on X-rayoptics and microanalysis. 1972. p. 121–30.

76. Sumithra S, Umarji AM. Role of crystal structure on the thermal expan-sion of Ln2W3O12 (Ln = La, Nd, Dy, Y, Er and Yb). Solid State Sci2004;6(12):1313–9.

77. Williamson RL, Fincke JR, Chang CH. A computational examination ofthe sources of statistical variance in particle parameters during thermalplasma spraying. Plasma Chem Plasma Process 2000;20(3):299–324.

78. Remesh K, Yu SCM, Ng HW, Berndt CC. Computational study and experi-mental comparison of the in-flight particle behavior for an external injectionplasma spray process. J Therm Spray Technol 2003;12(4):508–22.

79. Remesh K, Ng HW, Yu SCM. Influence of process parameters on thedeposition footprint in plasma-spray coating. J Therm Spray Technol2003;12(3):377–92.

80. Zhang T, Bao Y, Gawne DT, Liu B, Karwattzki J. Computer model tosimulate the random behaviour of particles in a thermal-spray jet. SurfCoat Technol 2006;201(6):3552–63.

81. Williamson RL, Fincke JR, Chang CH. Numerical study of the relativeimportance of turbulence, particle size and density, and injection parame-ters on particle behavior during thermal plasma spraying. J Therm SprayTechnol 2002;11(1):107–18.

82. Dru S, Meillot E, Wittmann-Teneze K, Benoit R, Saboungi ML. Plasmaspraying of lanthanum silicate electrolytes for intermediate temperaturesolid oxide fuel cells (ITSOFCs). Surf Coat Technol 2010;205(4):1060–4.

83. Pecharsky V, Zavalij P. Fundamentals of powder diffraction and structuralcharacterization of materials. Springer; 2008.

84. Di Girolamo G, Blasi C, Pilloni L, Schioppa M. Microstructural andthermal properties of plasma sprayed mullite coatings. Ceram Int2010;36:1389–95.

85. Bale CW, Chartrand P, Degterov SA, et al. FactSage thermochemical soft-ware and databases. Calphad 2002;26(2):189–228.

86. Ramachandran K, Kikukawa N, Nishiyama H. 3D modeling ofplasma–particle interactions in a plasma jet under dense loading conditions.Thin Solid Films 2003;435(1–2):298–306.

87. Ramachandran K, Nishiyama H. Fully coupled 3D modeling ofplasma–particle interactions in a plasma jet. Thin Solid Films2004;457(1):158–67.

88. Meillot E, Balmigere G. Plasma spraying modeling: particle injection in atime-fluctuating plasma jet. Surf Coat Technol 2008;202(18):4465–9.

89. Li H-P, Chen X. Three-dimensional simulation of a plasma jet

with transverse particle and carrier gas injection. Thin Solid Films2001;390(1–2):175–80.

90. Xiong H-B, Zheng L-L, Sampath S, Williamson RL, Fincke JR. Three-dimensional simulation of plasma spray: effects of carrier gas flow and

ERSX

pean Ceramic Society 34 (2014) 3069–3083 3083

particle injection on plasma jet and entrained particle behavior. Int J HeatMass Transfer 2004;47(24):5189–200.

91. Townsend P, Barnett D, Brunner T. Elastic relationships in layered compos-ite media with approximation for the case of thin films on a thick substrate.J Appl Phys 1987;62(11):4438–44.

92. Hsueh C-H. Modeling of elastic deformation of multilayers due to residualstresses and external bending. J Appl Phys 2002;91:9652.

93. Skinner BJ. Section 6: thermal expansion. Geol Soc Am Memoirs1966;97:75–96.

94. Sun Z, Zhou Y, Wang J, Li M. �-Y2Si2O7, a machinable silicateceramic: mechanical properties and machinability. J Am Ceram Soc2007;90(8):2535–41.

95. Phani KK, Niyogi SK. Elastic modulus-porosity relation in polycrystallinerare-earth oxides. J Am Ceram Soc 1987;70(12):C362–6.

96. Desmaison-Brut M, Montintin J, Valin F, Boncoeur M. Influence ofprocessing conditions on the microstructure and mechanical properties ofsintered yttrium oxides. J Am Ceram Soc 1995;78(3):716–22.

97. Ledbetter H, Kim S, Balzar D, Crudele S, Kriven W. Elastic properties ofmullite. J Am Ceram Soc 1998;81:1025–8.

98. Mah T-I, Mazdiyasni KS. Mechanical properties of mullite. J Am CeramSoc 1983;66:699–703.

99. Harder BJ, Almer J, Lee KN, Faber KT. In situ stress analysis of multilayerenvironmental barrier coatings. Powder Diffr 2009;24:94–8.

00. Harder BJ, Almer JD, Weyant CM, Lee KN, Faber KT. Residual stressanalysis of multilayer environmental barrier coatings. J Am Ceram Soc2009;92:452–9.

01. Harder BJ, Faber KT. Transformation kinetics in plasma-sprayedbarium- and strontium-doped aluminosilicate (BSAS). Scr Mater 2010;62:282–5.

02. O’Bryan HM, Gallagher PK, Berkstresser G. Thermal expansion ofY2SiO5 single crystals. J Am Ceram Soc 1988;71:C42–3.

03. Kriven WM, Pask JA. Solid solution range and microstructures of melt-grown mullite. J Am Ceram Soc 1983;66:649–54.

04. Spitsberg IT, Wang H, Heidorn RW. Method for thermally spraying crack-free mullite coatings on ceramic-based substrates. Google Patents; 2001.

05. Shackelford JF, Doremus RH. Ceramic and glass materials: structure,properties and processing. Springer; 2008.

06. Schneider H, Okada K, Pask JA. Mullite and mullite ceramics. J. Wiley;1994.

07. McPherson R. The relationship between the mechanism of formation,microstructure and properties of plasma-sprayed coatings. Thin Solid Films1981;83(3):297–310.

08. Deal BE, Grove AS. General relationship for the thermal oxidation ofsilicon. J Appl Phys 1965;36:3770–8.

lossary

TE: coefficient of thermal expansionSE: backscattered electronDS: energy dispersive X-ray spectroscopy

E: rare earthEM: scanning electron microscopeRD: X-ray diffraction