gas jet assisted vapor deposition of yttria stabilized zirconia · 2013-04-24 · gas jet assisted...

Gas jet assisted vapor deposition of yttria stabilized zirconiaD. D. Hass and H. N. G. Wadleya�

Department of Materials Science and Engineering, School of Engineering and Applied Science,University of Virginia, Charlottesville, Virginia 22903

�Received 29 August 2008; accepted 26 January 2009; published 27 February 2009�

A gas jet assisted electron beam evaporation process for synthesizing yttria stabilized zirconia�YSZ� coatings has recently been reported. The process uses a rarefied inert gas jet to entrain andtransport vapor to a substrate. The gas jet enables the lateral spreading of the flux to be controlledand large fractions of the vapor to be deposited on samples of relatively small size. When the gaspressure is high, coatings grown at 1050 °C and below have a columnar structure and a high porefraction. The total pore volume fraction, the morphology of the inter- and intracolumn pores and thecoating texture are all observed to be a strong function of the gas pressure in the chamber withincreasing chamber pressure leading to larger intercolumnar pore spacings, wider pores, a highertotal pore volume fraction, and a reduction in the coating texture. A direct simulation Monte Carlosimulation approach has been used to investigate vapor transport for the various gas pressuresexplored in this study. The simulation indicates that as the gas pressure increases, binary scatteringevents between the vapor and background gas broaden the vapor molecule incidence angledistribution. This intensifies flux shadowing and results in the incorporation of voids in the coating.Increasing the gas pressure also results in a rapid increase in the vapor phase nucleation of YSZclusters. This observation coincides with a transition from a �200� textured columnar morphology atmoderate pressures to a nanogranular structure with no texture and a very high nanoscopic pore
volume fraction at high pressures. © 2009 American Vacuum Society. �DOI: 10.1116/1.3085725�
I. INTRODUCTION

Yttria stabilized zirconia �YSZ� thermal barrier coatings�TBCs� are widely employed in the thermal protection sys-tems used to enhance the high temperature life of hot sectioncomponents in gas turbine engines.1,2 These superalloy com-ponents are protected from high temperature oxidation by acoating system consisting of a metallic bond coat based uponplatinum modified nickel aluminides3 or MCrAlY alloys,4 athermally grown �ideally pure �-alumina� oxide �TGO�layer5 and the YSZ thermal protection layer. Yttria partiallystabilized zirconia containing about 7 wt % Y2O3 �7YSZ� isusually chosen as the thermal barrier material because of itslow thermal conductivity,6 high melting point,7 thermal ex-pansion coefficient close to that of the substrate metal,8 andits thermochemical compatibility with the TGO layer that itcontacts during service. These 7YSZ coatings are usuallyapplied to hot section engine components by either by airplasma spray9 �APS� or electron beam physical vapor depo-sition �EB-PVD�.10 However other gas jet processes are alsobeing explored including electron beam evaporation com-bined with gas jet enabled directed vapor deposition11 and asolution precursor plasma spray �SPPS� process.12 The per-formance of a TBC coating is measured by its ability toreduce the temperature of the TGO layer �thereby reducingits growth rate�, its resistance to spallation during thermalcycling and erosion/foreign object damage by small �or

a�Author to whom correspondence should be addressed; electronic mail:
[email protected]
404 J. Vac. Sci. Technol. A 27„2…, Mar/Apr 2009 0734-2101/2009

larger� particle impacts and its susceptibility to reaction withcalcium aluminum magnesium silicates that accumulateupon the coating surface during use.

When thermal transport through the TBC occurs by con-duction, the coating thermal resistance is controlled by itsthermal conductivity. The TBC layer thermal conductivitydepends upon that of the material used to make the layer andthe volume fraction and morphology of the pores within thelayer.13 In APS coatings, repeated molten ceramic dropletimpingement, spreading and rapid solidification leads to theformation of long pores parallel to the substrate. The orien-tation of these pores is highly effective at impeding the flowof the heat through the coating and the through thicknessthermal conductivity of 7YSZ coatings made this is very low�0.8–1.0 W /mK at 25 °C�.14 However, pores in this orien-tation provide little compliance to accommodate in the planedifferential strains created between the coating and substrateby their thermal expansion mismatch. The coatings made bythis approach are therefore susceptible to spallation duringthermal cycling. A recent variant of the APS deposition pro-cesses has resulted in dense vertically cracked structureswhich are more resistance to thermal cycling. The SPPS pro-cess also results in structures with mixtures of pore shapesand orientations15 and those containing voids and cracks inthe through thickness direction are claimed to offer signifi-cant increases in thermal cycle lifetimes.16 An ultrafast vari-ant of the plasma deposition technique has also been recentlyproposed for TBC deposition.17 In this approach some of theceramic particles are evaporated during their travel withinthe plasma and vapor condensation, gas phase nucleated
nanocluster impacts and liquid droplet impingement occur
404/27„2…/404/11/$25.00 ©2009 American Vacuum Society

405 D. D. Hass and H. N. G. Wadley: Gas jet assisted vapor deposition of yttria stabilized zirconia 405

simultaneously resulting in a structure of low conductivity.The cyclic spallation life of this novel coating has not beenreported.

TBC coatings with the best thermal cycling life are usu-ally made by the EB-PVD process.18 The 7YSZ source ma-terial is vaporized by impingement of an electron beam upona source ingot and transported under low pressure��10−3 Pa� conditions to a heated air foil where condensa-tion occurs. By rotating the air foil in the vapor plume, aporous coating can be produced at deposition temperatures inthe 1000–1050 °C range. These vapor deposited YSZ coat-ings have highly �200� textured, columnar microstructureswith primary intercolumnar pores oriented in the coatinggrowth direction. These help to accommodate the �in-plane�coefficient of thermal expansion mismatch between the coat-ing and the metal substrate thereby enhancing the cyclicspallation life. However, these intercolumnar pores are ori-entated perpendicular to the plane of the coating and there-fore only modestly reduce the coatings thermalconductivity.19 Finer scale �initially feathery� pores are alsopresent within the columns and are much more effective atimpeding the heat flow through the coating. The throughthickness thermal conductivity of EB-PVD YSZ coatings isstill significantly higher than that of the APS coatings �it liesin the 1.5–1.9 W /mK range at 25 °C� �Ref. 20� but can bevaried by modifying the rotation rate and substrate tempera-ture which effect the pore fraction and morphology.14 Recentmodeling has begun to establish the causal linkages betweenthe pore morphology and performance in these systems21 andhas led to the identification of novel zig-zag �or herringbone�morphologies that seek to achieve both in plane complianceand low through thickness thermal conductivity.22 Thesestudies indicate that control of the pore morphology is essen-tial if optimal coating performance is to be achieved.23

Recent experimental and atomistic modeling studies indi-cate that the morphology in vapor deposited coatings is af-fected by many processing variables. These include the va-por atoms kinetic energy,24,25 its incidence angle,26,27 thesubstrate’s temperature,28,29 the deposition rate,28,30 the back-ground pressure,28,31,32 vapor plume composition,33 and thesubstrate’s roughness.34 The experimental studies clearlyshow that porous, columnar morphologies are associatedwith low kinetic energy, oblique vapor incidence angles, lowsubstrate temperatures, high deposition rates, rough sub-strates, and high chamber pressures.24,26,28,34,35 Atomisticmodels generally support these observations36 and link poreformation to flux shadowing under conditions of restrictedadatom mobility.37

Several groups have attempted to use atomistic simula-tions to investigate the relationships between the pore vol-ume fraction and morphology and the conditions used togrow a vapor deposited film. The most fundamental ap-proach based upon density functional theory is too computa-tionally constrained for the problems encountered in eventhin film growth. Instead, the interatomic bonding has beencoarse grained and represented by an interatomic potential.

38
Potentials for many materials have been proposed, but
JVST A - Vacuum, Surfaces, and Films

none as yet have reliably addressed the interatomic forcesthat control the structure and properties of yttria stabilizedzirconia. Molecular dynamics methods have been utilizedwith some of the potentials proposed for metallic systems toinvestigate pore control.39 Unfortunately computational con-straints force the use of greatly accelerated deposition ratesfor the simulations and an underestimation of the contribu-tion of diffusional processes upon the atomic assembly of acoating.

Gilmer et al. pioneered the use of a kinetic Monte Carlomodeling approach to vapor deposition that does not requireacceleration of the deposition rate.40 In this approach, theactivation barriers for diffusion of the species of interest areprecalculated �for example, by density functional theory orby molecular statics using interatomic potentials� and thenthe system of interest is assembled atom by atom allowingdiffusion to occur at time steps defined by the smallest acti-vation energy diffusive event. A two-dimensional version ofthe approach has been used to explore the multidiffusionpath assembly of thin nickel films using activation barriersobtained from an embedded atom potential for nickel.41 Theapproach successfully predicted the pore structures andtrends in pore fraction observed experimentally in metal sys-tems. Cho et al. extended the approach to investigate thegrowth of columnar structures in coatings that were rotatedin a vapor plume and showed that local high points on thesurface shadowed the incident flux and this controlled thegrowth of the both the primary and feathery pore structuresobserved in TBC coatings.42 These atomistic modeling ap-proaches reveal that the pore fraction and morphology can becontrolled by the molecular weight and incidence angle dis-tribution of the depositing flux, the diffusion rate on thegrowth surface and the initial surface upon which depositionoccurs.

The low chamber pressures ��10−4 Pa� used to facilitateelectron beam propagation in the EB-PVD process results incollisionless vapor transport to the substrate. When the sub-strate is far from the vapor source, this collisionless transportresults in a narrow vapor angle of incidence distributionupon a flat, substrate positioned perpendicular to the flux.Very few oblique atom impacts with the substrate occur andshadowing is therefore reduced. While the vapor atoms ki-netic energy is low ��0.2 eV� the porous, columnar mor-phology is only achieved when the flux incidence angle dis-tribution is broadened to promote flux shadowing. In the EB-PVD process, this is normally accomplished by rotating thesubstrate. This rotation is also utilized to ensure uniformcoating of air foil components. The ability to exert greatercontrol of the pore shape and distribution during the vapordeposition of TBCs is limited by limitations to the design ofcurrent EB-PVD systems which provide little ability to con-trol the incidence angle of the vapor other than by samplerotation.

There are other ways of manipulating the angle of inci-dence distribution of the vapor atoms/molecules. One ap-proach is to entrain the flux in a supersonic gas jet that is

43
directed to the substrate. Within the jet, the pressure can be


high �10–100 Pa� and the mean free path is greatly reducedso that many vapor phase collisions occur. For example, themean free path of zirconia molecules in helium is predictedby binary collision theory to decrease from 10 m to 1 mmwhen the background pressure is increased from10−3 to 10 Pa.44 Multiple scattering events during the trans-port of vapor both widens the angular incidence distributionand thermalizes the flux. However, increasing the pressurealso increases the frequency of three body collisions andtherefore the probability of nucleating vapor phase clusters.

An electron beam directed vapor deposition �DVD� pro-cess has been proposed as a means for depositing zirconiabased coatings at high pressure �10–100 Pa�.45 Preliminarystudies indicated that YSZ layers with porous, columnarmorphologies could be deposited onto stationary substratesusing this technique.46 Subsequent experiments have shownthat TBCs with very high pore fractions and very low ther-mal conductivities �comparable to APS coatings� can bemade by combining gas phase scattering with substrate rota-tion. However, a detailed study of the dependence of thecoating morphology on pressure has not been conducted.Here, the morphology, texture and density of YSZ films de-posited over a range of pressures is systematically investi-gated. Altering the chamber pressure is observed to greatlyaffect the characteristics of the coatings. A direct simulationMonte Carlo �DSMC� approach is used to estimate the flowfield characteristics for the range of pressures used in theexperiments. These flow fields are then used to simulate mul-tiple scattering of the vapor flux and to deduce the angle ofincidence distribution of the vapor flux at the stationary sub-strate. This allows the mechanisms responsible for the largechanges in the coatings pore morphology and microstructureto be identified.

II. EXPERIMENT

A. Deposition process

In a DVD process, material is vaporized using a highvoltage/medium power �60 kV /10 kW� axial e-beam gun.By incorporating differential pumping of the gun column anda very small ��3 mm diameter� e-beam exit opening into theprocess chamber the gun can function in a high pressureenvironment �up to 200 Pa in helium�. Figure 1 shows aschematic illustration of the process.47 The gas jet was cre-ated by supersonic expansion through a nozzle near the siteof evaporation. This was achieved by maintaining anupstream/downstream pressure ratio in excess of two. Thegas jet entrained the vapor and transported it to the substrate.The gas jet was introduced into the chamber via a 1.27 cmnozzle opening. In the present experiments an upstream/processing chamber pressure ratio of between 2.5 and 10.0was used. The process chamber pressure was varied from13.3 to 106.4 Pa �Table I�. The substrates onto which depo-sition occurred were 25.4 mm diameter IN100 buttons with anickel aluminide bond coat �supplied by GE Aircraft�. Thesubstrate surface had a root mean square �rms� roughness of

48
405.10 nm and a mean roughness �Ra� of 106.66 nm. Sta-
J. Vac. Sci. Technol. A, Vol. 27, No. 2, Mar/Apr 2009

tionary substrates were positioned 12.0 cm from the nozzle�x-direction� and 9.5 cm from the source. The height�y-direction� of the substrate was set so that the bottom of thesubstrate was level with the top of the crucible. This resultedin the center point of the substrate being slightly lower thanthe nozzle center point and maximized the deposition effi-ciency for many of the process conditions used.48

7YSZ vapor was created by evaporation from a porous7YSZ source rod �Transtech, Inc.�. To compensate for oxy-gen lost from the source during evaporation, a helium-oxygen �3.0 vol % oxygen� carrier gas was used to create thegas jet. The substrate was heated to 1050 °C using a resistiveheater placed behind the substrate. The substrate temperaturewas monitored using a thermocouple inserted in a drilledhole centered with respect to the substrate thickness. Thislimited errors from convective cooling of the thermocoupleby the carrier gas flow and by radiant heating of the thermo-couple from by the heater. Evaporation was allowed to con-tinue until coating thicknesses of between 50 and 100 �mwere achieved. The deposition rate was between 2.0 and6.0 �m min−1.

FIG. 1. Experimental setup used for the deposition of partially stabilizedzirconia �containing 7 wt % yttria� on platinum modified nickel aluminiedcoated IN100 substrates at a temperature of 1050 °C.

TABLE I. DVD process conditions employed.

Chamberpressure

�Pa�Pressure

ratio

Gasflow�slm�

Depositionrate

��m /min�

13.3 10.0 5.0 2.026.6 10.0 8.0 3.839.9 6.6 8.0 5.853.2 5.0 8.0 7.766.5 4.0 8.0 6.979.8 3.4 8.0 4.993.1 2.5 8.0 2.6


B. Coating characterization

The coating density was measured directly from the sub-strates weight gain and dimensional changes. For many ofthe conditions investigated, elongated, intergrowth columnpores were observed that extended from the substrate to thecoating surface. Finer, intracolumnar porosity was also ob-served. Figure 2 schematically illustrates the coating mor-phology. The pore morphology has been characterized bydetermining the primary pore width, �, and the intercolumnarspacing, � on mechanically polished cross sections of thecoatings. All pore parameters were measured within6.35 mm of the center of the substrate. The pore spacing wasmeasured at the midplane of the coating. The pore width wasmeasured at three positions: 10 �m below the coating sur-face, at the coating midpoint, and 10 �m above the substrate.For the measurement of the pore width it was recognized thatthe polished surface resulted in an expanded width due to therange of angles that the pores intersected the polished crosssection. These angles varied from 0° to 70°. The error wascorrected by multiplying the measured value of the spacingby a correction factor of 1.22. This corresponds to an inter-section angle of 35° �the midpoint between two extreme val-ues of 0° and 70°�. The coating texture was determined byx-ray diffraction through comparison of the peak intensitieswith those of a randomly orientated sample. Error bars ofone standard deviation are shown on all graphs of data.

III. RESULTS

Figure 3 shows the cross-section �a� and surface views

FIG. 2. Schematic illustration defining the parameters used to quantify themorphology of the porosity found in 7 YSZ coatings.

��b� and �c�� of a coating deposited at the lowest investigated


pressure �13.3 Pa�. This and all other coatings deposited hada columnar structure. The primary growth diameters did notgenerally vary with thickness. The columns were approxi-mately parallel sided and not highly tapered like those seenin EB-PVD samples.13 Wide, intercolumnar pores up to5 �m in width defined the primary columns. These extendedfrom the substrate to the coatings outer surface, Fig. 4�A�.The primary columns had a fractal structure with each col-umn subdivided by smaller intracolumnar pores.

The intercolumnar pore width, the intercolumnar porespacing, and the total pore volume fraction of the yttria sta-bilized zirconia coatings all varied with chamber pressure. InFig. 4, the effect of chamber pressure on the intercolumnarpore spacing and width is shown. The pore spacing steadilyincreased with chamber pressure. The pore width increasedwith chamber pressure until a maximum width was observedat an intermediate pressure. Further pressure increases led toa reduction in pore width.

The total pore volume fraction �deduced from the macro-scopic density data� exhibited a large increase �from2 to 50 vol %� as the pressure was increased from13.3 to 106.4 Pa, Fig. 5. The intercolumnar pore volume

FIG. 3. SEM micrographs showing �A� the cross section and �B� the topsurface of a 7 YSZ coating deposited using a chamber pressure of 13.3 Paand a pressure ratio of 10.0. Note the intercolumnar pores that extend fromthe substrate to the coating surface.

fraction is also shown. The largest intercolumnar pore vol-


ume fraction occurred at the lowest investigated pressure�13.3 Pa� due primarily to the close spacing of the interco-lumnar pores in this condition. The difference between thetotal pore volume and the intercolumnar pore volume gives ameasure of the intracolumnar pore volume. This systemati-cally increased with pressure. The pore volume fraction dataare summarized in Table II. The intracolumnar pore volumefraction data corresponded well with the observed coatingmorphology. At the lowest pressure �13.3 Pa� the columnsappeared relatively dense. As the pressure increased the col-umn sides developed a feathery appearance. At the highestgrowth pressure �106.4 Pa� the columns had a granular ap-pearance and contained fewer intercolumnar pores, Fig. 6.

The coating texture was examined by x-ray diffraction.The peak intensities normalized by the �111� peak intensityare given in Table III and plotted versus pressure in Fig. 7.Coatings deposited at the lowest pressure �13.3 Pa� had a

FIG. 4. Plot showing the change in �A� the intercolumnar pore width and �B�the intercolumnar pore spacing with chamber pressure. Note the systematicincrease in both parameters as the chamber pressure increased.

�200� preferred orientation. This �200� texture disappeared as


the chamber pressure was increased. A randomly orientated,nanocrystalline structure resulted at the highest pressure�106.4 Pa�.

IV. DIRECT SIMULATION MONTE CARLOSIMULATIONS

A DSMC technique was used to simulate binary collisionsduring transport of the vapor to the substrate thus estimatethe vapor incidence angle distribution �LAD�.49 A two-stepmodeling approach was used. In step 1 the gas jet supersonicexpansion and its interaction with a flat substrate was ana-lyzed. The code output was the gas jet flow field in the re-gion near the substrate. This was used in step 2 to analyzethe interaction of the gas jet with the vapor flux and to esti-mate the incidence angle distribution at the substrate. Detailscan be found in Ref. 48.

Representative simulations of the gas jet velocity fieldcomponents in the axial and radial directions during the su-

FIG. 5. Plot showing the measured change in �A� the total pore volumefraction and �B� the intercolumnar pore volume fraction in YSZ coatingsdeposited using a range of chamber pressures �13.3–106.4 Pa�. Note thedramatic increase in pore volume fraction as the chamber pressure in-creased. The difference between the curves in �A� and �B� is a measure ofthe intracolumnar pore volume fraction.

TABLE II. Pore parameters for each pressure condition.

Chamberpressure

�Pa�

Interclumnarpore

spacing ��m�

Intercolumnarpore

width��m�

Intercolumnarpore

volume�%�

Totalpore

volume�%�

13.3 3.3 0.4 12.1 ¯

26.6 11.8 0.5 4.2 7.039.9 15.5 0.9 5.8 ¯

53.2 24.9 1.4 5.6 31.566.5 48.4 1.9 3.9 ¯

79.8 69.0 1.9 2.7 38.593.1 133 1.4 1.1 45.6


personic expansion are shown in Fig. 8. In the axial directionthe maximum velocity decreased as the chamber pressureincreased, Table IV. This was the result of a decrease in thepressure ratio with higher chamber pressure. The interactionof the gas with the substrate resulted in the formation of astrong radial component to the velocity �i.e., the wall jetvelocity� near the substrate. The maximum wall jet velocityis also given in Table IV.

FIG. 6. High magnification SEM micrograph showing cross sections of zir-conia coatings deposited using �A� a chamber pressure of 13.3 Pa and apressure ratio of 10.0 and �B� a chamber pressure of 106.6 Pa and a pressureratio of 2.5. Note the smoother and denser appearance of the low pressurecondition and the porous, granular appearance of the high pressure case.

TABLE III. Relative peak intensities from XRD patterns of the yttria stabi-lized zirconia films deposited with different chamber pressures.

Chamberpressure �Torr� I�111� I�200� I�220� I�311� I�400�

13.6 1 0.56 0.55 0.53 0.2326.6 1 0.32 0.37 0.33 0.1039.9 1 0.25 0.38 0.34 0.0753.2 1 0.30 0.43 0.36 0.0966.5 1 0.26 0.47 0.37 0.09

Random 1 0.20 0.56 0.38 0.07


The IAD at a given location on the substrate �near to theregion where porosity measurements were performed� wasestimated for the various process conditions. The results, Fig.9, show that many of the molecules impinged the substrate atoblique angles of incidence. To characterize changes in thedistribution, two parameters were defined and are given inTable IV: �i�The peak maximum angle, �m, defined as theincident angle at which the maximum intensity, Io, occurredand �ii� the peak dispersion width, Pw, defined as the widthof the distribution at 0.5 Io.

In Fig. 10, Pw is plotted as a function of chamber pres-sure. As the chamber pressure increased Pw broadened. �m

FIG. 7. XRD data showing the change in the relative intensity of the �200�peak with respect to the �111� peak. High �200� peak intensities indicate thepresence of a �200� preferred orientation. Note the increased �200� peakintensity as the chamber pressure was decreased.

FIG. 8. DSMC simulations showing the speed of a helium-3.0% oxygencarrier gas in �A� the axial direction and �B� the radial direction. The axialcomponent was reduced in the region near the substrate and significant
radial component developed �i.e., the wall jet�.


was observed to decrease from −18° at the lowest pressure�13.3 Pa� to 0° at the highest pressure �106.4 Pa�.

V. HIGH PRESSURE VAPOR DEPOSITION

When vapor deposition is performed using chamber pres-sures greater than �0.1 Pa, the mean free path for vaporatoms collisions with the background gas becomes less thanthe typical source-to-substrate distance �30–50 cm�. Thesecollisions can alter the coating morphology in two ways. Thefirst is by altering flux shadowing mechanisms. Binary colli-sions scatter the vapor onto various trajectories resulting inchanges to the IAD of the depositing flux. This controlsshadowing phenomena. The second is the nucleation andgrowth of multiple atom clusters. The impingement of clus-ters on the substrate has been observed to affect the coatinggrowth process.50,51 These mechanisms are discussed belowin detail.

A. Flux shadowing effects

Flux shadowing results in porosity when oblique atomarrivals are “shadowed” by fast growing grains or surfaceasperities. This creates local reductions in the density of va-

FIG. 9. Change in the IAD distribution with chamber pressure is given.Higher chamber pressures resulted in broader distributions and peak posi-tions close to the substrate normal.

TABLE IV. DSMC simulation results for the gas jet e

Chamberpressure�Torr�

Maximumvelocity

Z-dir �m/s�

Wall jetvelocity

�m/s�

13.3 1455 14726.6 1635 25539.9 1482 20053.2 1300 16766.5 1270 12679.8 1175 94


por flux that impinges on the substrate. A reduced localgrowth rate results in the formation of pores in the coating.The degree to which this occurs depends upon the IAD andthe adatom surface mobility. A high fraction of oblique ada-tom arrivals and a low surface mobility promote porosity. Ahigh surface mobility will compensate the local flux deple-tion by surface diffusion.

An increased chamber pressure during DVD depositionalters the mean free path of vapor molecules as well as thevelocity distribution of the background gas. This combina-tion controls the distribution of incidence angles of the de-positing molecules. Atomistic models have revealed thatchanges to these distributions will result in changes to thecoating microstructure.36

B. Cluster formation

The advent of binary and ternary collisions during vaportransport will, in some cases, promote gas phase clustering.The formation and ensuing deposition of clusters is impor-tant as they can affect the pore morphology and texture ofvapor deposited films.50,51 The effect of a cluster impact de-pends strongly on its impact energy and size. High energycluster impacts lead to denser coating morphologies and

sion and IAD distribution.

erageADdeg�

Peakmaximum, Pm

�deg�

Peak width,Pw �at 0.5Pm�

�deg�

6.5 18 618.0 22 686.7 15 929.5 14 925.9 14 1014.4 5 101

FIG. 10. Change in the peak dispersion width with chamber pressure isgiven. The anticipated effect of cluster formation on Pw is schematically

xpan

AvI�

111

shown.


lower energy cluster result in porous, granular coatings withequiaxed grain structures.52 To further explore the probabil-ity of cluster formation in a DVD environment a kineticapproach given by Bensen53 is considered.

Generally speaking, three body collisions are required toform clusters in the vapor phase in order to conserve the lawsof momentum and energy for elementary reactions.54 Eitherthree vapor species or two vapor species and a backgroundgas atom or molecule may facilitate a reaction. The reactionkinetics is therefore a strong function of the density of thevapor species as it controls the likelihood that a third bodyimpacts the transition state of a metastable dimer. The vapordensity is a related to the evaporation rate, the system geom-etry and the operating chamber pressure.

For the current study, it is believed that the probability ofcluster formation is greatly affected by the presence of mo-lecular evaporation and a “reactive” carrier gas �containingoxygen�. Thus, instead of single atoms, ZrO and ZrO2 mol-ecules are the predominate species leaving the vapor source.Oxygen molecules are also present in the carrier gas stream.The presence of these molecules alters the cluster formationkinetics. Examples of binary collisions that could enable theformation of stable clusters are given below:

ZrO + O2 ↔ ZrO2 + O*, �Step I� , �1�

ZrO + ZrO2 ↔ 2ZrO + O*, �Step IIA� , �2�

ZrO2 + ZrO2 ↔ 2ZrO + O2, �Step IIB� , �3�

2ZrO + ZrO2 ↔ 3ZrO + O*, �Step III� , �4�

�N − 1�ZrO + ZrO2 ↔ �N�ZrO + O*, �Step IV� . �5�

The oxygen atoms �O*� provide one mechanism to re-move the latent heat of fusion. Helium/ cluster collisionsprovide a second mechanism. The important result is thatthree-body collisions are not necessarily required to nucleateclusters. As a result, the nucleation of clusters is not consid-ered rate limiting for the process conditions used here andonly the subsequent cluster growth from continued collisionevents with the vapor molecules, oxygen molecules, and thehelium carrier gas need be considered.

Cluster growth can be estimated using

dN

dT=

ZAN−A

�AN�= �A��aAN

+ �aA

22�8kT

��1/2

,

where �A� is the volume concentration of the vapor species,�aAN is the hard sphere diameter of the cluster, �a is the hardsphere diameter of the vapor species, is the Boltzman con-stant, T is the gas temperature, and � is the reduced mass ofcollision, defined by 1 /�=1 /mAN+1 /mA. Note that clustergrowth is largely dependent on the vapor species concentra-tion. The average size of the cluster can then be determined
using

Nav =1

0

N�t�d ,

where is the time of flight of the vapor species.Using this approach, the effect of chamber pressure on the

cluster size was determined. Estimates of the key parametersare given in Table V. The zirconia concentration in the vaporwill increase with chamber pressure due to increased vaporfocusing by the gas jet. Visualization studies are used toobserve these effects.48 Values for are determined fromDSMC calculations. The effect of pressure on the cluster sizefor a range of �aA values is shown in Fig. 11. The value of�aA is not precisely known for ZrO and ZrO2 molecules andwill increase as the cluster grows. Thus, an estimated rangeof �aA values is given in Fig. 11. The increased zirconiaconcentration and longer time of flight that results when thepressure is increased results in a higher probability of largeclusters �clusters greater than 100 molecules in size appearpossible�. The effect of these impacts on the growth andmorphology of the yttria stabilized zirconia coatings is con-sidered below.

TABLE V. Input values for the cluster size calculations.

Chamberpressure�Torr�

Average gasjet axial

velocity �m/s�

Source tosubstrate

distance �m�Time of flight,

��s�Average ZrO2

density �# /m3�

13.3 833.46 0.095 113.98 2.4�10−18

26.6 999.88 0.095 95.01 3.0�10−18

39.9 890.11 0.095 106.72 4.0�10−18

53.2 781.29 0.095 121.95 6.0�10−18

66.5 719.23 0.095 132.08 9.0�10−18

79.8 637.29 0.095 149.06 1.2�10−19

106.4 278.14 0.095 341.55 1.2�10−19

FIG. 11. Estimate of the cluster size for a range of zirconia hard spherediameters. Increased chamber pressures and larger hard sphere diameters
result in larger cluster impacts.


VI. DISCUSSION

Large changes to the morphology of the zirconia coatingsresulted as the chamber pressure was increased during depo-sition. These changes included �i� a larger intercolumnar porespacing, �ii� wider intercolumnar pores, �iii� an increasedvolume fraction of fine scaled, intracolumnar pores, and �iv�a disappearance of the texture typically observed in coatingsof this type. The large multifaceted changes observed indi-cate a fundamental transition in the growth mechanisms ofthese coatings. These changes are related to changes in theflux shadowing and cluster deposition mechanisms discussedabove.

A. Columnar pore structures

Columnar porosity is often present in coatings assembledfrom the vapor phase due to differences in the growth ratesof different crystal surfaces and flux shadowing. The prefer-ential growth of surface nucleated 7 YSZ crystallites orien-tated with their fastest growing directions towards the vaporsource rapidly raises these crystals above others and theyharvest a majority of the flux.55 This then results in astrongly �200� textured, columnar coating morphology. Here,such a �200� texture was observed, but only in the coatingsdeposited at the lowest pressure �13.3 Pa�. The absence oftexture in the higher pressure process conditions indicatesthat competitive growth of this type is not a dominate growthmechanism for those high pressure conditions. Recent ki-netic Monte Carlo simulations indicate that the intercolum-nar pores observed here nucleate at surface asperities on therough substrate and rapidly grow due to flux shadowingmechanisms.56 The role of substrate asperities is important asmany of the intercolumnar pores are eliminated when coat-ing occurs onto atomistically flat substrates.48

Atomistic simulations indicate that pore nucleation at sur-face asperities depends upon the asperity size, the surround-ing surface geometry and the IAD distribution.56 As the IADdistribution becomes broader and more asymmetric, smallerasperity dimensions will nucleate pores �i.e., the IAD distri-bution can alter the population of nucleation sites on thesubstrate�. Pores nucleate when regions on the substrate aredepleted in incident vapor flux with respect to neighboringregions. The depletion is limited by the diffusion of adatomsacross the growth surface into flux depleted regions. Thus,factors that affect the adatom surface mobility �i.e., the vaporspecies translation energy, the latent heat of condensationrelease, and molecular weight together with the substratetemperature, deposition rate, and surface topology� will alsoplay a role.

B. Nanoscopic structures

Interestingly, intercolumnar pores did not form as readilyin coatings as the pressure was increased. This is surprisingas DSMC simulations indicate that a higher pressure in-creased Pw due to a higher probability of repeated binarycollision scattering of vapor atoms and the promotion of ob-
lique vapor incident angles. The resulting increase in flux

shadowing should ease the nucleation of intercolumnar poresand increase the number of intercolumnar pore nucleationsites.36 The fact that this was not observed suggests that adifferent growth process must be occuring.

The most likely possibility is the formation of the gasphase clusters. When clusters impact on a substrate their af-fect on the coating growth process is determined by theirkinetic energy and the cohesion energies of the substrate.Several scenarios exist, including implantation of the clusterinto the substrate, dissociation upon impact, or the formationof mounds on the surface. For the present case, the clustershave a low kinetic energy per atom and low energy clusterimpacts are resulting in small mounds on the surface are thenexpected. When the cluster size is small, the mounds willaffect the adatom surface mobility. For an adatom to diffuseover a mound the activation barrier for an atom jump over aledge �i.e., the Schwoebel barrier� must be overcome. Thisbarrier is much greater than for hopping over a flat surfaceand thus, surface diffusion will be reduced. As the frequencyof cluster impacts and the cluster size increase, the process isno longer governed by the surface adatom diffusion, but in-creasingly by the successive cluster impacts.

The nucleation of intercolumnar pores will be affected intwo ways by the transition from atomistic to cluster deposi-tion. First, as the size of the clusters increase they will beginto approach the size of some of the surface asperities. Thiswill limit shadowing induced pore nucleation, as shownschematically in Fig. 12. Second, DSMC simulations revealthat the IAD peak width �Pw� increases with pressure if ato-mistic deposition is predominant. However, as the clustersize increases with pressure, the rate at which Pw �for thecombined cluster and vapor flux� increases with pressure willbe reduced and eventually decrease. The trajectories of clus-ters increasingly higher mass are increasingly less effectedby collisions with light helium atoms. As the volume fractionand size of the clusters increases �with increasing pressure� agreater fraction of material will impact the substrate at nearnormal incidence. Thus, impacts onto a rough substrate arenot as likely to be shadowed by surface asperities and inter-columnar pore nucleation is partially eliminated. The resultis a reduction in the number of pore nucleation sites on thesubstrate and increased intercolumnar pore spacing.

The transition to a cluster dominated deposition regimewill also alter the pore size. Kinetic Monte Carlo simulationshave indicated that the pore width is linked to both the IADand the size and geometry of asperities �or discontinuities�on the substrate.56 Broad IAD distributions result in largepore widths due to the increase in the area of the substratethat is flux depleted, Fig. 13. This relationship is observed bynoting that both the IAD peak width �see the projected line inFig. 10 that assumes cluster formation� and the pore width�see Fig. 4�B�� increase until an intermediate pressure�106.4 Pa� is reached after which both values decrease. Thisis further complicated by the observation that the pore spac-ing decreases with increased pressure and thus, the numberof pore nucleation sites is also reduced. Whether a specific
substrate geometry will nucleate a pore is determined by the


IAD and the size of frequency of cluster impacts. For ex-ample, a region on the substrate that nucleates an intercolum-nar pore at low pressure may not necessarily nucleate a poreat higher pressure and vice versa. This is due to changes inthe IAD that result when the chamber pressure is alteredand/or the onset of cluster formation and deposition. Poresthat nucleate at intermediate pressures are observed to have alargest width. This is because large asperities �such as bondcoat grain boundaries� that result in the widest pores are onlyobserved to nucleate pores at intermediate pressures whenPw is large. The thinnest pores are gradually eliminated withincreased pressure presumably because of an increased fre-quency of cluster impacts or loss of the crystallographic tex-ture of the columns in the coating. The disappearance of thethinnest pores together with the formation of wider pores atthe bond coat grain boundaries results in the observed in-crease in the average pore width with pressure until Pw be-gins to decrease due to the increased frequency of clusterimpacts.

While critical for TBC applications, the intercolumnar

FIG. 12. Schematic illustration showing �A� atomistic deposition onto arough substrate resulting in the nucleation of intercolumnar porosity and �B�cluster deposition onto a rough substrate. The clusters shown are of the samesize scale as the substrate roughness leading to the inability to result inshadowing induced pore nucleation.

pore volume fraction typically represents only 5%–10% of


the porosity in these coatings. The total pore volume fractionwas often much higher �up to 50%� and dramatically in-creased with chamber pressure. Kinetic Monte Carlosimulations56 indicate that the level of IAD variation ob-served here cannot account for the large increase in porosity.Other parameters that could affect the intercolumnar porefraction �such as the vapor species translation energy and thevarious contributions an adatom’s surface mobility� are alsonot anticipated to be significantly changed by increasing thechamber pressure since the translation energy of the incidentspecies is low ��0.2 eV per atom� in all cases and the ada-tom surface mobility is primarily controlled by the substratetemperature. However, low energy cluster impacts, such asthose shown schematically in Fig. 12�B� do lead to the for-mation of nanoscale porosity in the coating by a processanalogous to that formed between the much larger “splats” inthe APS process. Molecular dynamics simulations of clusterimpacts57 have shown that nanoscale porosity does form be-tween impacting clusters. This phenomenon becomes impor-tant as the pressure is increased and cluster impacts becomemore frequent.

VII. CONCLUSIONS

Zirconia coatings have been deposited using a directedvapor deposition processing approach. Large effects effectsof deposition pressure upon the microstructure and pore mor-phology were observed. Relatively dense coatings, with tex-tured columns were found at the lowest chamber pressures�13.3 Pa�. As the chamber pressure was increased, wide in-tercolumnar pores were observed. The spacing and width ofthese pores increased with pressure. At the highest pressure�106.4 Pa� large pore volume fractions were measured �up to50%�. However, this porosity consisted almost entirely of

FIG. 13. Schematic illustration showing the effect of the IAD on the width ofintercolumnar pores.

fine scaled �nanoscopic� intracolumnar pores. The morpho-


logical changes have been linked to changes in the incidenceangle distribution of the depositing and the increased occur-rence of low energy cluster impacts at the highest pressure.

ACKNOWLEDGMENTS

The authors are grateful to Timothy Bartel at Sandia Na-tional Laboratories for use of his ICARUS DSMC code. Thisresearch was supported by the Office of Naval Research�Grant No. N0001400-0147� under the program direction ofSteve Fishman and David Shifler.

1NRC Report: Coatings for High Temperature Structure Materials, editedby R. V. Hillary �National Academy, Washington, D.C., 1996�.

2J. W. Fairbanks and R. J. Hecht, Mater. Sci. Eng. 88, 321 �1987�.3I. T. Spitsberg, D. R. Mumm, and A. G. Evans, Mater. Sci. Eng., A 394,176 �2005�.

4J. R. Nicholly, MRS Bull. 28, 659 �2003�.5V. K. Tolpygo, D. R. Clarke, and K. S. Murphy, Surf. Coat. Technol.

146–147, 124 �2001�.6P. Morrell and R. Taylor, High Temp. - High Press. 17, 79 �1985�.7Handbook Chem and Physics �CRC, Boston, 1990�.8R. L. Jones, Report No. NRL/MR/6170-96-7841, 1996.9O. Yamamoto, Electrochim. Acta 45, 2423 �2000�.

10D. J. Wortman, B. A. Nagaraj, and E. C. Duderstadt, Mater. Sci. Eng., A121, 443 �1989�.

11H. Zhao, Fengling Yu, T. D. Bennett, and Haydn Wadley, Acta Mater. 54,5195 �2006�.

12P. R. Strutt, B. H. Kear, and R. Boland, U.S. Patent No. 6025024 �2000�.13O. Unal, J. Am. Ceram. Soc. 77, 984 �1994�.14S. M. Meier, D. M. Nissley, K. D. Sheffler, and T. A. Cruse, J. Eng. Gas.

Turbine Power Trans. of the ASME 114, 258 �1992�.15Liangde Xie, Xinging Ma, E. H. Jordan, Nitin P. Padture, D. T. Xiao, and

M. Gell, J. Mater. Sci. 39, 1639 �2004�.16M. Gell, Liangde Xie, Xinging Ma, Eric Jordan, and Nitin P. Padature,

Surf. Coat. Technol. 177–178, 97 �2004�.17H. Huang, K. Eguchi, M. Kambara, and T. Yoshida, J. Therm. Spray

Technol. 15, 83 �2006�.18N. Padture, M. Gell, and E. Jordan, Science 296, 280 �2002�.19S. M. Meier and D. K. Gupta, J. Eng. Gas. Turbine Power Trans. of the

ASME 116, 250 �1994�.20C. Levi, Curr. Opin. Solid State Mater. Sci. 8, 77 �2004�.21S. Gu, T. J. Lu, D. D. Hass, and H. N. G. Wadley, Acta Mater. 49, 2539

�2001�.22D. Hass, A. J. Slifka, and H. N. G. Wadley, Acta Mater. 49, 973 �2001�.23U. Schultz, B. Sarukan, K. Fritscher, and C. Leyens, Int. J. Appl. Ceram.

Technol. 1, 302 �2004�.24X. W. Zhou, R. A. Johnson, and H. N. G. Wadley, Acta Mater. 45, 4441

�1997�.25K. H. Müller, Surf. Sci. 184, L375 �1987�.26S. M. Rossnagel, C. Nichols, S. Hamaguchi, D. Ruzic, and R. Turkot, J.

Vac. Sci. Technol. B 14, 1819 �1996�.27J. G. W. van de Waterbeemd and G. W. van Oosterhout, Philips Res. Rep.

22, 375 �1967�.28J. A. Thornton, J. Vac. Sci. Technol. 12, 830 �1975�.


29K. Kennedy, Proceedings of the AVS Vacuum Metallurgical Conference,Los Angeles, 1968 �unpublished�, p. 195.

30K. H. Müller, J. Appl. Phys. 58, 2573 �1985�.31J. J. Cuomo, J. M. E. Harper, C. R. Guarnieri, D. S. Yee, L. J. Attanasio,

J. Angilello, C. T. Wu, and R. H. Hammond, J. Vac. Sci. Technol. 20, 349�1982�.

32T. C. Huang, G. Lim, F. Parmigiani, and E. Kay, J. Vac. Sci. Technol. A3, 2161 �1985�.

33B. V. Movchan and A. V. Demchishin, Phys. Met. Metallogr. 28, 83�1969�.

34D. V. Rigney, R. Viguie, D. J. Wortman, and D. W. Skelly, NASA Conf.Publ. 3312, 135 �1995�.

35K. Robbie, L. J. Friedrich, S. K. Drew, T. Smy, and M. J. Brett, J. Vac.Sci. Technol. A 13, 1032 �1995�.

36Y. G. Yang, R. A. Johnson, and H. N. G. Wadley, Acta Mater. 45, 1455�1997�.

37R. A. Dugdale, Trans. Inst. Met. Finish. 54, 61 �1976�.38M. Finnis, Prog. Mater. Sci. 49, 1 �2004�.39X. W. Zhou and H. N. G. Wadley, Surf. Sci. 487, 159 �2001�.40G. Gilmer, H. Huang, T. dela Rubia, and F. Bauman, Thin Solid Films

365, 189 �2000�.41X. W. Zhou, R. A. Johnson, and H. N. G. Wadley, Acta Mater. 45, 4441

�1997�.42J. Cho, S. G. Terry, R. LeSar, and C. G. Levi, Mater. Sci. Eng., A 391,

390 �2005�.43J. F. Groves, H. N. G. Wadley, A. P. Ritenour, D. D. Hass, and P. L.

Ratnaparkhi, Proceedings of Electron Beam Melting and Refining State ofthe Art, Bakish Materials Corp., Englewood, NJ, 1997 �unpublished�, p.46; D. D. Hass, P. A. Parrish, and H. N. G. Wadley, J. Vac. Sci. Technol.A 16, 3396 �1998�.

44G. A. Bird, Molecular Gas Dynamics and the Direct Simulation of GasFlows �Clarendon, New York, 1994�.

45D. D. Hass, A. J. Slifka, and H. N. G. Wadley, Acta Mater. 49, 973�2001�.

46Hengbei Zhao, Fengling Yu, Ted D. Bennett, and Haydn N. G. Wadley,Acta Mater. 54, 5195 �2006�.

47J. F. Groves and H. N. G. Wadley, Composites, Part B 28B, 57 �1997�.48D. D. Hass, Ph.D. thesis University of Virginia, 2001.49T. J. Bartel and S. J. Plimpton, AIAA Paper No. 92-2860 �1992�.50L. Bardotti, P. Jenson, A. Hoareau, M. Treilleux, B. Cabaud, A. Perez,

and F. Cadete Santos Aires, Surf. Sci. 367, 276 �1996�.51P. Jenson, J. F. Roux, L. Bardotti, B. Cabaud, M. Treilleux, and A. Hoar-

eau, Mater. Sci. Eng., B 60, 51 �1999�.52A. M. Dongare, D. D. Hass, and L. V. Zhigilei, Proceedings of the Inter-

national International Symposium on Clusters and Nano-Assemblies:Physical and Biological Systems, 10–13 November 2003, Richmond, Vir-ginia, USA �unpublished�.

53S. W. Bensen, Foundations of Chemical Kinetics �McGraw-Hill, NewYork, 1960�, Chap. 12.

54S. H. Bauer and D. J. Frurip, J. Phys. Chem. 81, 1015 �1977�.55U. Schulz, S. G. Terry, and C. G. Levi, Mater. Sci. Eng., A 360, 319

�2003�.56D. D. Hass, Y. Y. Yang, and H. N. G. Wadley, J. Porous Mater. �to be

published�.57R. S. Averback, M. Ghaly, and H. Zhu, Radiat. Eff. Defects Solids, 130–

131, 211 �1994�.

gas jet assisted vapor deposition of yttria stabilized zirconia · 2013-04-24 · gas jet assisted...

Documents