形成的热喷涂涂层由于毛孔不完整的填充图案表面的裂缝

Plasma Chem Plasma Process(2007)27:647–657

DOI10.1007/s11090-007-9097-8

O R I G I N A L P A P E R

Formation of Pores in Thermal Spray Coatings due to Incomplete Filling of Crevices in Patterned Surfaces

M.Xue?S.Chandra?J.Mostaghimi?H.R.Salimijazi

Received:6May2007/Accepted:5July2007/Published online:25July2007

óSpringer Science+Business Media,LLC2007

Abstract Molten particles in a thermal spray land on a rough surface,coalesce with each other and freeze to form a coating.Surface tension prevents liquid splats from completely ?lling crevices in the substrate,forming pores.An analytical model is developed to esti-mate the volume of such pores by calculating the equilibrium shape of a liquid meniscus pressing down on a surface asperity.Predictions from the model are compared with experimental results for the volume of voids formed under plasma sprayed yttria stabilised zirconia(YSZ)particles(average diameter18l m)landing with an average velocity of 250m/s on patterned silicon surfaces that had vertical posts on them.The model predicted, to within an order-of-magnitude,the volume of voids on a surface in which the posts were tall(3l m high)and closely spaced(1l m apart),where pores were principally formed by incomplete?lling of gaps.

Keywords Thermal spray coatingáFormation of poresáIncomplete?llingáPatterned surfaceáSurface tension

Nomenclature

e length o

f meniscus

g gravity acceleration

h liquid height

h s height of deformed liquid bottom

i number of post,counting from point of droplet impact

p pressure

r radial direction

R splat radius

R a average surface roughness

s theoretical height of deformed liquid surface of the underside of a splat

V impact velocity

M.Xue(&)áS.ChandraáJ.MostaghimiáH.R.Salimijazi

Centre for Advanced Coating Technologies,University of Toronto,Toronto,Canada

e-mail:michelle.xue@utoronto.ca

V in Volume of incompletely?lled voids

x Cartesian coordinate

y Cartesian coordinate

z Cartesian coordinate

a capillary length

b radius of the surface asperities

h angle of perfect contact

r surface tension

q liquid density

/v meniscus contact angle

/t meniscus contact angle

Introduction

Thermal spray coatings are formed by the impact of molten droplets onto a solid surface that fuse together to form a thin protective layer.Such coatings are not perfectly dense,but contain pores and cracks that may or may not be desirable,depending on the function of the coating.Coatings that shield components from wear or corrosion should have low porosity,since that improves adhesion strength and makes the coating impervious.The insulating properties of a thermal barrier coating are improved,however,by the presence of air pockets within it.In either case it is important to be able to predict and control porosity during coating deposition.

Several attempts have been made to develop computer models that can predict the structure of thermal spray coatings[1–3].All of these models postulate a set of rules that determine the size and locations of pores,based on observations made in experimental studies.Several different mechanisms have been identi?ed that can create porosity:curling up of splats due to thermal stresses,entrapment of gas under impacting particles and incomplete?lling of cavities in the already deposited coating.Protuberances may already exist on a rough substrate,or they may be created during spraying by the presence of unmelted particles in the spray,or as a result of satellite droplets detaching from impacting droplets and solidifying on the surface.

Pores formed by gas entrapment are typically very small,and found at the interface between splats in thermal spray coatings.Based on transmission electron microscopy of plasma-sprayed Al2O3coatings,McPherson and Shafer[4]showed that the interfaces between lamellae consist of regions of perfect contact alternating with gaps of0.01–0.1l m which probably arise from absorbed or entrapped gas between impinging droplets and previously solidi?ed layers.

Curling up at the edge of splats is frequently observed in thermal spray coatings and may be a major source of porosity.The SEM micrograph in Fig.1(a)shows solidi?ed7% yttria stabilized zirconia(YSZ)splats deposited on a polished stainless steel substrate by thermal plasma spray.The particle temperature and velocity upon the impact were2900°C and250m/s,respectively.A section(Fig.1(b))through the splat along the line marked A–A0,made using focused ion beam(FIB)milling shows a gap between the splat and substrate,where the edge of the splat had curled up.Splat curl up is cause by residual

stresses in the splat as it cools and shrinks after being sprayed on the surface.The bottom surface of the splat is attached to the substrate and cannot shrink,while the upper surface is free to contract.The resulting stresses are relieved either by the edges of the splat curling up or the splat cracking.The upper surface of the splat in Fig.1(a)shows a network of cracks caused by such stresses.Metallic splats are much more ductile than ceramic splats and do not crack easily:curl-up is therefore the only means of providing stress relief.Curling-up is therefore a much more important source of porosity in metallic coatings than in ceramic coatings.

Cirolini et al.[2]developed a model for the deposition of a thermal barrier plasma-sprayed coating assuming that curling was caused by the temperature drop across the splat when the solidi?cation front just reached the top.Xue,Chandra and Mostaghimi [5]developed a model to predict the angle of curl-up around the edges of metallic splats and showed that predictions agreed reasonably well with measurements from splats formed by the impact of both large (2–3mm)and small (30–50l m)diameter droplets of molten metals.When a molten droplet lands on a rough surface,not all of the voids under it will be completely ?lled.The pressure inside the molten drop may not be enough to overcome surface tension and force liquid into small cavities.Then,as the splat freezes,pores will form wherever there is incomplete ?lling.This mechanism is complex since it depends on many factors,including pressure distribution in the impacting drop,substrate topography and the rate of solidi?cation.

Fukanuma [6]developed a model for porosity formation during thermal spray coating processes by considering deformation of a molten particle,and showed that most of the porosity is near the periphery of the splat,starting at a distance from its centre of about 0.6times the spat radius (R ).Porosity was sensitive to particle velocity,ambient gas pressure,particle diameter,and molten material viscosity.

Sobolev and Guilemany [7]derived a set of analytical formulae to describe the pressure distribution in a ?attening droplet along the droplet-substrate interface during thermal spraying.Droplet-substrate micro-adhesion and coating porosity along the radius were shown to depend on the ratio of the pressure developed upon impact to the capillary pressure.Cai and Lavernia [8]modelled porosity formation during spray forming and studied the effects of processing parameters on porosity in spray formed

materials.

Fig.1(a )SEM micrographs of YSZ single splats deposited on a polished stainless steel substrate at 300°C and (b )cross-section through splat at line line A–A 0showing splat/substrate interface in Particle temperature and velocity at impact were 2900°C and 250m/s

The objective of this paper is to develop an analytical model to estimate the volume of voids formed by incomplete ?lling of the interstices under a splat on rough surfaces.Experiments were done in which molten particles of yttria stabilised zirconia (YSZ)were plasma sprayed onto silicon surfaces that had a pattern of vertical posts,1–3l m high,etched on them.Predictions from the model were compared with measurements of pores formed due to incomplete ?lling of the voids between posts on the substrate.

Incomplete Filling of Interstices on Rough Surfaces

Ideal Model of Rough Surface

Heichal and Chandra [9]developed a theory to predict contact resistance between a liquid metal and a rough solid surface.The rough surface was idealized as consisting of a series of semi-circular asperities,all having the same radius b .Surface roughness R a is de?ned as the arithmetic average of the height of surface peaks above a hypothetical perfectly smooth plane.Figure 2shows an idealized representation of liquid in contact with a solid substrate on which there is a single asperity.For the idealized surface shown in Fig.2the average roughness is:

R a ?R 2b 0y ex Tdx

2b ?R p 0b 2sin 2ew Td w

2b ?pb

4e1T

Surface tension prevents the liquid from completely wetting the surface peak.Timsit

[10]derived a model to calculate the equilibrium pro?le of the liquid meniscus by using variational methods to determine the shape that minimized the total potential and surface energy of the system.Figure 2shows the geometry of the contact meniscus y (x )formed with a liquid in contact with a single semi-circular asperity with pro?le z (x ).The origin of the coordinate system used in the 2-D analysis is located at the point of contact of the meniscus with the horizontal surface that forms the base of the asperity.If we know the shape of the liquid meniscus and the pro?le of the asperity we can integrate the

difference Fig.2Geometry of the contact meniscus formed with an asperity

between them over the length e to calculate the volume of incompletely ?lled voids (labelled V in in Fig.2)

Timsit [10]provided a solution for the equilibrium shape of the liquid interface,assuming that b /h <<1and e/a <<1and where the liquid does not wet the surface,i.e./t ?/v ?180 .The solution showed that the pro?le of the liquid meniscus is:

y ex T?2h sin 2x 2a

e2T

The capillary length is de?ned as a ????????????r =q g p ,where r is the surface tension of the

liquid.The angle subtended by the area of contact between the liquid and the surface asperity is:

h ?2b h

a t

b h 1=2e3T

and the horizontal length of the void between the liquid and substrate is:

e ?a 22b h ea 2tb h T !1=2e4T

The surface of the substrate is de?ned by

z ex T??????????????????????????????????????????????b 2àee tb sin h àx T2

q x [e tb sin h àb z ex T?0

x \e tb sin h àb 8<:e5TWhen the single semi-circular asperity in Fig.2is transformed into a series of asper-

ities,the origin of the meniscus moved to the valley between the asperities.In this case

Eq.(5)should be modi?ed to z ex T????????????????????????????b 2àeb àx T2q .

If we assume the asperity is a hemisphere,the volume of the 3-D void between it and the liquid is:

V in ?Z e

02p eb àx Ty ex Tàz ex T? dx e6T

A computer program was written to evaluate Eq.(6).

Pressure Distribution along a Splat Radius

Application of Eq.(6)to calculate void volume requires the hydrostatic head h forcing liquid down onto asperities on the surface.In the case of an impacting droplet the hydrostatic head can be replaced by an equivalent impact pressure:

p ?q gh e7T

Numerical simulations of droplet impact were conducted to study the pressure distri-butions along the splat radius.We used a 3-D numerical code of ?uid ?ow and heat transfer

[11]to simulate impact of an 18l m diameter droplet of zirconia with an initial temper-ature of 2900°C landing with a velocity of 250m/s on a silicon surface.The droplet spread out into a splat of 55l m in diameter.Pressure at the splat-substrate interface was found to decay with distance from the splat centre,and with time after impact.The maximum pressure experienced under the splat was recorded at several locations and is shown by the data points in Fig.3.An exponential curve of the form:

p er T?A 1e àr 2A 2e8T

was ?t through the data to describe the radial pressure distribution,where p is the maxi-mum pressure and r the radial location.A 1and A 2are functions of droplet impact velocity,V and splat radius,R .The maximum impact pressure at the centre of the droplet is given by

p er ?0T?12q V 2

e9T

We assume that the pressure decays to a very small value within a distance equal to two radii:

p er ?2R T%10à10Pa e10T

Note that the small value,10à10,is used here instead of zero,because exponential functions asymptotically approach zero but never reach it.Substituting Eqs.(9)and (10)into Eq.(8)gives expressions for A 1and A 2so that:

p er T?12q V 2áe àr 2A 2e11T

where

Fig.3Maximum pressure distribution along the radial location after an 18l m diameter YSZ particle impacted on a substrate with impacting temperature and velocity of 2900°C and 250m/s,respectively.The maximum splat radius for this case is 55l m

A 2?à4R 2ln 10

à1012q V 2

!"#à1e12T

Experimental Procedure

To observe incomplete ?lling of voids under plasma sprayed particles impacting on a rough surface,single splats of crushed 7%Y 2O 3–ZrO 2powders (Amperit 825-0,à22.6/+5.6l m from H.C.Starck,Germany)were collected after impact on patterned Si surfaces.To obtain a narrow particle size distribution,Amperit 825-0powders were sieved to à25l m.The average particle diameter was 18l m.

The substrates were made of silicon wafers on which vertical posts with square cross-sections were etched.Two different patterns were used:3l m in height,4l m in width,with 1l m spacing in between and 1l m in height,4l m in width,with 5l m spacing in between.Surfaces were kept at a temperature of 300°C during spraying to vaporize any adsorbed contaminants on them.

Plasma spray experiments were carried out with a custom-made plasma torch working with a gas mixture of CO 2/CH 4.The plasma torch-substrate distance was maintained at 50mm during deposition.The typical atmospheric plasma spray parameters and conditions used for collecting YSZ single splats are summarized in Table 1.A custom-made jig was used to reduce the number of splats landing on the substrate,so that individual splats could be identi?ed.The jig consists of a sliding steel shutter with a slit (approximately 2mm wide)placed at a distance of approximately 1mm in front of the substrate.During spraying,the shutter was released so that it fell under its own weight,allowing a few droplets from each section of the spray to impinge on the substrate as it traversed the plasma spray.

Particle conditions upon the impact were monitored by a DPV-2000system from Tecnar Ltd,Montreal,Canada.Variation in particle properties across the cross-section of the plasma plume were characterized by measurements at 81points across the cross-section of the plume in a square 9·9grid with 3mm distance between nodes.The average temperature and velocity of the particles at the centre of the plasma plump were T =2900°C and V =250m/s,respectively.

Splats deposited were photographed using scanning electron microscope (SEM).In order to study pore formation,cross-sections through individual splats were made by FIB

Table 1Atmospheric plasma spray process parameters

Gun

CO 2/CH 4Current,A

268Voltage,V

178Total gas ?ow rate,(CO 2/CH 4),slpm

30/7.1Powder carrier gas (Ar),slpm

10Feed rate,g/min

8Spray distance,mm

50Substrate temperature (°C)300

milling(Micrion Corporation,Peabody,MA,USA).Micrographs from the cross-sections were taken by SEM(Hitachi S-570,Tokyo,Japan).Image analysis software,Clemex (Clemex Technologies Inc.,Longueuil,Canada),were employed to measure the porosity across the splat/substrate interfaces.

Results and Discussion

Figure4(a)shows a micrograph from a YSZ splat after it impacted on the patterned surface kept at300°C.The square posts on this surface were3l m high and4l m width,with 1l m spacing in between.Figure4(b)shows a cross-section through the centre of the splat, made by focused-ion beam milling.The interface between the splat and substrate can be seen in the micrograph,as well as the gaps in the substrate.The gaps are labelled in Fig.4(b),with numbers starting at the splat centre and increasing with radial distance from the centre.The inset in Fig.4(b)shows enlarged images of the gaps in the substrate,where spaces near the centre of the splat are seen to be more?lled than those near the edges.The incompletely?lled interstices are those located where the pressure under the splat falls off sharply(see Fig.3).

The porosity inside each interstice in Fig.4(b)was measured based on the area of the void in the micrograph using image analysis software.The porosity values in the two interstices at both sides of each rectangular post were averaged to obtain a porosity value of the post.Assuming that the void had uniform cross-section all around each post,the area void was then converted to a volume void by multiplying the average porosity value by the

Fig.4SEM micrograph of(a)the top view and(b)the central cross-section,of a splat formed by an18l m diameter YSZ droplet impacting on the silicon patterned surface with particle temperature and velocity of 2900°C and250m/s,respectively.The substrate temperature was300°C.The posts on the substrate are 3l m high,4l m wide,with1l m spacing in between

volume of interstice around each post.For example,the volume of interstice around each post in Fig.4(b)was 27l m 3,and the porosity value of post two from measurement was 3%.The volume of void around post two was then calculated as 27·3%=0.81l m 3.The volumes of voids around each rectangular post,numbered from 1to 6,along the splat radius were calculated and listed in Table 2.The voids in the interstice on the other side of the impact point,numbered 10–60,were also measured and listed in Table 2.

The volumes of void caused by incomplete ?lling of the each of the spaces 1–6and 10–60between the posts were calculated using Eq.(6).Eq.(11)was used to predict the pressure distribution under an impacting droplet and calculate the pressure p (r i ),in the i th gap located at radius r i ,measured from the droplet impact point.The calculated values are listed in Table 2.Figure 5shows a comparison between the measured and calculated values of void volumes.Predicted void sizes were of the same order of magnitude as those measured in experiments.

The model was less effective in predicting void volume when the spacing between posts was much larger,since other mechanisms for creating voids,besides incomplete ?lling of interstices,became important.Figure 6(a)shows a YSZ splat on a patterned surface with square posts 1l m high,4l m wide,and 5l m spaces in between.Measured void volumes are compared with those predicted by Eq.(6)in Table 3.The measured voids are much larger than those obtained from calculations.However,the model only accounts for experimental measurements and calculation using the analytical model

Radial location

123456Total volume 102030405060Measurements (l m 3)

0.00.8±0.3 4.1±0.3 6.7±0.39.6±0.312.2±0.333.4±0.30.00.0 1.8±0.3 4.7±0.314.8±0.317.7±0.339.0±0.3Analytical calculation (l m 3)0.10.1

0.20.9 4.214.2

19.7Fig.5Volume of voids along the radius of the YSZ splat shown in Fig.4

incomplete ?lling of spaces next to the posts.Some of the largest voids seen in Fig 6(b)are those located equidistant from the posts.These could not have been created by surface tension preventing the liquid meniscus entering a small crevice.Rather,they must have been created by another mechanism,such as gas entrapment.Note that the corresponding interstices on the other side of the splat in Fig.6(b)(labelled 10and 20)are almost com-pletely ?lled,indicating that the ?lling was not controlled by pressure alone,since the pressure distribution would have been expected to be symmetrical on the two sides of the splat.

Summary

When a molten droplet impacts a rough surface,pressure in the liquid drives it into the crevices between asperities on the surface.However,surface tension prevents the liquid from completely ?lling these gaps,leaving voids.An analytical model is developed to calculate the volume of these voids.Predictions from the model are compared with experimental results for the volume of voids formed under plasma sprayed YSZ particles landing on patterned silicon surfaces with vertical posts on it.The model predicted,to within an order-of-magnitude,the volume of voids on a surface in which the posts were tall experimental measurements and calculation using the analytical model

Radial location

123Total volume Measurements (l m 3)

0.0 1.2±0.6 4.0±0.6 5.2±0.6Analytical calculation (l m 3)0.0020.0040.020

0.028

Fig.6SEM micrograph of (a )the top view and (b )the central cross-section,of a splat formed by an 18l m diameter YSZ droplet impacting on the silicon patterned surface with particle temperature and velocity of 2900°C and 250m/s,respectively.The substrate temperature was 300°C.The posts on the substrate are 1l m in high,4l m wide,with 5l m spacing in between

(3l m high)and closely spaced(1l m apart),where the main reason for formation of pores was incomplete?lling of gaps.The model did not work well on a surface where the posts were short(1l m high)and widely separated(5l m apart).In this case it appeared that other mechanisms,such as gas entrapment or formation of cracks in the splat were responsible for pore formation.

Acknowledgement The authors would like to acknowledge the support and assistance of GE Global Research Centre,NY,USA,in providing the patterned surfaces and FIB micrographs.

Reference

1.Knotek O,Elsing R(1987)Surf Coat Tech32:261–271

2.Cirolini S,Harding JH,Jacucci G(1991)Surf Coat Tech48:137–145

3.Ghafouri-Azar R,Mostaghimi J,Chandra S,Charmchi M(2003)J Therm Spray Tech12:3–69

4.McPherson R,Shafer BV(1982)Thin Solid Films97:201–204

5.Xue M,Chandra S,Mostaghimi J(2006)J Therm Spray Tech15:531–536

6.Fukanuma H(1994)J Therm Spray Tech3(1):33–44

7.Sobolev VV,Guilemany JM(1996)J Mater Letter28:331–335

8.Cai WD,Lavernia EJ(1997)Mater Sci Eng A226–228:8–12

9.Heichal Y,Chandra S(2005)J Heat Transf127:1269–1275

10.Timsit RS(1982)Wear83:129–141

11.Pasandideh-Fard M,Chandra S,Mostaghimi J(2002)Int J Heat Mass Tran45:2229–2242