Properties of PECVD-deposited thermal barrier coatings

Properties of PECVD-deposited thermal barrier coatings

Surface and Coatings Technology 142᎐144 Ž2001. 835᎐842 Properties of PECVD-deposited thermal barrier coatings Boris Preauchat, Stefan DrawinU ´ ONERA...

1MB Sizes 0 Downloads 41 Views

Surface and Coatings Technology 142᎐144 Ž2001. 835᎐842

Properties of PECVD-deposited thermal barrier coatings Boris Preauchat, Stefan DrawinU ´ ONERA-Metallic Materials and Processing Department, BP 72, 92322 Chatillon Cedex, France ˆ

Abstract Thick, partially yttria-stabilised zirconia coatings have been deposited by PECVD. Evolution of the morphology and phase composition has been studied as a function of the annealing temperature and duration Ž1100᎐1400⬚C, for up to 1000 h.. The as-deposited columnar morphology is similar to EBPVD and in good agreement with the PVD structural zone models. Evolution of the phase composition follows the phase diagram. The overall thermal stability observed is encouraging for TBC applications. 䊚 2001 Elsevier Science B.V. All rights reserved. Keywords: Growth models; Structural zones 1,2,T; Scanning electron microscopy ŽSEM.; X-Ray diffraction; Plasma-assisted chemical vapour deposition ŽPACVD.; Zirconium oxide

1. Introduction Thick Ž) 100 ␮m. zirconia-based coatings are used as thermal barrier coatings ŽTBCs. for hot-section turboengine components w1x. Two techniques are used on industrial scale, plasma spraying ŽPS. and electron beam physical vapour deposition ŽEBPVD.. The properties of the widely used 8 wt.% yttria partially stabilised zirconia TBCs deposited by both techniques have been widely discussed in the literature w2x. However, because it induces longer lifetimes, the columnar morphology of EBPVD coatings is preferred for the most thermally and mechanically loaded components Žblades and vanes.. ONERA has developed plasma-enhanced chemical vapour deposition ŽPECVD. as an alternative technique to deposit columnar TBCs in conditions that could circumvent the main drawbacks of EBPVD, namely high investment and running costs and low capability for coating complex-shaped parts Ži.e. pronounced line-of-sight feature .. Morphological and crystallographic properties of PECVD coatings deposited at various temperatures Ž700 and 900⬚C. will be presented, for both as-deposited and isothermally aged samples Ž1100᎐1400⬚C for up to 1000 h.. U

Corresponding author. Fax: q33-1-46-73-41-64. E-mail address: [email protected] ŽS. Drawin..

2. Experimental

2.1. Deposition reactor

The characteristics of the PECVD reactor developed at ONERA ŽFig. 1. have been detailed elsewhere w3,4x. Zirconia-based coatings, mainly 3᎐20 mol.% YO1.5stabilised ZrO 2 , have been deposited at rates reaching 150 ␮mrh on various substrates Žalumina or superalloys; diameter, 10᎐25 mm; thickness, 0.6᎐2.5 mm.. The precursors, zirconium tetrachloride ŽZrCl 4 ; Alfa Aesar, 99.5q % pure. and yttrium tris-2,2,6,6-tetramethyl-3,5-heptanedionate wYŽthd. 3 ; Inorgtech Ltd, 99.9% purex, are sublimated in separate furnaces Ž170᎐230⬚C.. A carrier gas Žargon, 99.9996% pure. transports the vapours to a mixing chamber, where secondary argon and oxygen Ž99.9995% pure. flows are added. The gases are then injected into the deposition chamber, a 100-mm-diameter quartz tube maintained at low vacuum, in which a surfaguide-type microwave discharge is produced. The substrate temperature can be adjusted between 300 and 900⬚C, by thermal transfer from the plasma or by an additional resistive heater. The temperature is measured using a thermocouple in the mid-plane of the substrate; it has been verified that

0257-8972r01r$ - see front matter 䊚 2001 Elsevier Science B.V. All rights reserved. PII: S 0 2 5 7 - 8 9 7 2 Ž 0 1 . 0 1 2 1 1 - 7

B. Preauchat, S. Drawin r Surface and Coatings Technology 142᎐144 (2001) 835᎐842 ´

836

Fig. 1. Schematic diagram of the PECVD reactor.

the difference from the substrate surface temperature is less than 10⬚C. 2.2. Substrates The coatings for the morphology study were deposited on 20-mm-diameter sintered ␣-alumina disks ŽDegussa AL23 grade. having a R a value of approximately 0.5 ␮m Žmeasured with a Taylor-Hobson Talysurf 5-120 profilometer.. For the crystallographic study, smoother, 10-mm-diameter sintered ␣-alumina disks Ž R a ; 0.05 ␮m. were used ŽCoors Ceramics SUPERSTRATE䊛 996.. The choice of alumina is justified by the fact that TBCs are currently deposited on MCrAlY ŽM s Ni andror Co. or NiAl-coated superalloys exhibiting a ca 1-␮m-thick alumina scale. The use of pure alumina substrates instead of coated superalloys eliminates possible failures at the metal᎐alumina interface or within the alumina scale during long-duration heat treatments. 2.3. Coatings The standard deposition conditions for 8 mol.% YO1.5-stabilised ZrO 2 , which enable deposition rates of approximately 100 ␮mrh, are given in Table 1. The coating thickness is in the range 65᎐100 ␮m. The coatings are white at all deposition temperatures.

MfrM T s 0.88⭈ I Ž 400. fr Ž I Ž 004. T q I Ž 400. T .

Ž1.

with Mf q M T s 1

Ž2.

where M and I Ž hkl . are the phase mole fraction and integrated Ž hkl .-peak intensity, respectively. Mm is taken to be 0, because no monoclinic peaks were detected. In the original formula, Miller and co-workers considered the contribution of only one tetragonal phase Žt⬘.. In our XRD patterns, we had to consider the existence of an additional tetragonal phase Žt. to allow a good fit; thus, in Eq. Ž1., the subscript T stands for the sum of the t and t⬘ peak intensities. The mole fraction of the t phase is extracted from M T using Eq. Ž3.: MtrM T s Ž I Ž 004. t q I Ž 400. t . r Ž I Ž 004. T q I Ž 400. T .

Table 1 Typical operating conditions Total pressure ŽPa. Total gas flow Žm3 sy1 . O2 mole fraction Mass of sublimated ZrCl4 Žkg. Mass of sublimated YŽthd.3 Žkg. Microwave power ŽW. Substrate temperature Ž⬚C.

The yttria content was controlled by electron-probe microanalysis ŽEPMA. with the same apparatus ŽCAMECA Camebax Datanim. as reported by Bisson et al. w5x. The measurement of the Y and Zr content is not straightforward, because of the high porosity level and the non-conductivity of the oxide. However, Bisson and co-workers were able to define satisfactory analysis conditions and adequate corrections in their work on single crystals with various yttria contents. According to the Y2 O 3 ᎐ZrO 2 diagram w6x, four phases can be observed in the coatings: a high-yttriacontent fluorite phase f, a monoclinic phase m and two tetragonal phases labelled t Žlow yttria content. and t⬘ Žhigh yttria content.. The latter is a metastable phase existing at room temperature, denoted ‘non-transformable’ in comparison with the t phase, which transforms into the m phase during cooling, with an important volume change that is detrimental to the integrity of the coating. To estimate the destabilisation rate of the coatings, we have quantified the mole fractions of the phases appearing during high-temperature treatments with the help of Eq. Ž1. proposed by Miller et al. w7x, by fitting ŽPhilips APD software. the peaks on the X-ray diffraction patterns in the  4004 -region ŽPhilips PW1730 diffractometer, CuK ␣ radiation, integration time of 90 s per 0.02⬚ 2␪-step.:

Ž3. 106.4 1.33= 10y5 0.5 2᎐2.5= 10y3 0.4᎐0.6= 10y3 1700 700, 900

M T s Mt q Mt⬘

Ž4.

The values obtained by EPMA were compared to those given by the relations between lattice parameters a and c and the yttria mole fraction x extracted from

B. Preauchat, S. Drawin r Surface and Coatings Technology 142᎐144 (2001) 835᎐842 ´

Scott’s work w6x and reported by Kisi and Howard w8x for bulk ceramic: xs y Ž c t⬘rat⬘ y 1.0225. r0.001311 for 0.04- x- 0.12 Ž5. at⬘ s 0.50801q 0.03582⭈ x for 0.04- x- 0.13

Ž6.

a f s 0.51159q 0.01547⭈ x for 0.12- x- 0.25

Ž7.

The coatings deposited for this study had the composition xYO1.5 ᎐ Ž1 y x .ZrO 2 with x s 6.4᎐8.8 Ž"0.7. mol.% ŽEPMA values.. The accuracy of the position of the most intense XRD peaks, such as Ž400.t⬘, is good at "0.005⬚, i.e. "0.5 mol.%YO1.5 , because of the excellent convergence of the fitting procedure. For each deposition temperature, one 20-mm-diameter sample was cut into five parts for scanning electron microscope analysis ŽSEM, Zeiss DSM 982 Gemini, with field effect electron gun. in the as-deposited state and after annealing at 1100, 1200, 1300 and 1400⬚C for

837

Table 2 Evolution of column properties for coatings deposited at 700⬚C Heat treatment

Platelet size Ž␮m.

Intercolumnar spacing Ž␮m.

As-deposited 1100⬚C Ž100 h. 1200⬚C Ž100 h. 1300⬚C Ž100 h. 1400⬚C Ž100 h.

0.2᎐0.4 0.3᎐0.6 0.4᎐1 0.6᎐1 ᎐

᎐ 0.5᎐1 1᎐2 1᎐3 1᎐5

100 h in laboratory air; for XRD analysis, four 10-mmdiameter samples were annealed at the same temperatures, but periodically air-quenched for analysis, without any visible spallation.

3. Results 3.1. Coatings deposited at 700⬚C 3.1.1. Morphology The 700⬚C morphology is presented in Fig. 2. According to the models of Movchan and Demchishin w9x or Thornton w10x, it belongs to the zone 2 side of the transition zone T, with bundles of fine and irregularly aligned columns, characteristic of the zone T, but with interlinked platelets observable at the surface, more characteristic of zone 2. Column sintering still occurs ŽTable 2.: as the annealing temperature increases, the intercolumnar spacing increases, the size of the platelets increases and their amount decreases; the column tip becomes smoother, but the width of columns varies little and the widest cracks in the coatings extend over less than 5 ␮m. 3.1.2. Crystallography The as-deposited coatings exhibit the single t⬘ phase with a strong w200x texture. During annealing at 1100⬚C, a significant amount of fluorite phase appears only after 1000 h ŽTable 3.. This duration is reduced to 100 h at 1200⬚C. At 1300⬚C, the destabilisation of the t⬘ phase is already at an advanced stage after 50 h, and even earlier Ž20 h. at 1400⬚C, comparable with data reported in the literature w7,11x. 3.2. Coatings deposited at 900⬚C

Fig. 2. As-deposited coating ŽTsubstrate s 700⬚C.: Ža. fractograph; and Žb. surface micrograph.

3.2.1. Morphology The as-deposited 900⬚C coatings exhibit well-defined columns with four-sided pyramidal heads ŽFig. 3. corresponding to zone 2 described by Movchan and Demchishin w9x and Thornton w10x as a consequence of high surface diffusivity. However, rift-like growth defects parallel to the column axis and generally perpendicular

838

B. Preauchat, S. Drawin r Surface and Coatings Technology 142᎐144 (2001) 835᎐842 ´

Table 3 Phase distribution and lattice parameters after annealing for coatings deposited at 700⬚C a Yttria content of as-deposited coating Žmol.% YO1.5 .b

Heat treatment

Phases

Mole fraction Ž%.

Lattice parameter Žnm.

Yttria content of the phase Žmol.% YO1.5 .c

8.0r8.5

1100⬚C Ž1000 h.

t t⬘ t⬘1 f

25 66 9

at s 0.5089 at ⬘ s 0.5098 as 0.5132

-3 4.9 14.5 10.5d

8.0r8.5

1200⬚C Ž500 h.

t t⬘ t⬘1 f

66 20 14

at s 0.5093 at ⬘ s 0.5098 as 0.5129

3.7 5.1 13.7 8.5d

8.5r8.5

1300⬚C Ž50 h.

t t⬘ t⬘1 f

45 21 34

at s 0.5093 at ⬘ s 0.5102 as 0.5126

3.7 6.1 13.0 6.9 d

8.5r8.5

1400⬚C Ž20 h.

t t⬘ t⬘1 f

39 30 31

at s 0.5093 at s 0.5115 as 0.5126

3.5 9.8 12.7 6.3d

a

See text for further explanations. Left number: EPMA result; right number: XRD result using Eq. Ž5.. c XRD result using Eq. Ž6.. d SRD result using Eq. Ž7.. b

to the pyramid’s edges Ži.e. along  1004 cleavage planes. are observed. The morphology varies in a column, with a dense core and feather-like flanks ŽFig. 3a,c.; the latter feature is a deviation from the original model and is also observed in EBPVD coatings w12x, but to a lesser extent. Macroscopic ledges, approximately 150 nm wide and characteristic of some surface diffusivity, are observed on column tips at high magnification ŽFig. 3d.. The morphology at 900⬚C is little altered by annealing up to 1300⬚C, the ledges and the feathers disappearing after 100 h at 1100⬚C and being replaced by smooth column tips and sintered branches. Moreover, no further evolution is noted at 1200 and 1300⬚C. Annealing at 1400⬚C induces further smoothing of the column edges and development of the first cracks in the coatings.

from the normal direction ŽFig. 5., which corresponds to the angle between the w100x and w111x directions in a fluorite crystal Ž54.74⬚.. These elements describe a typical fibre texture along the w200x axis normal to the substrate.

3.2.2. Crystallography The sole t⬘ phase is observed in as-deposited coatings with a stronger w200x texture than observed on the 700⬚C coatings. Fig. 4 presents the  4004 -region patterns for different annealing times at 1100⬚C. In the 900⬚C coatings, the emergence of the fluorite peak between Ž004.t⬘ and Ž400.t⬘ peaks is similar to the 700⬚C coatings. w200xt⬘ and w111xt⬘ pole figures have been measured on a thicker coating Ž; 200 ␮m.. The former presents a strong peak with a FWHM of approximately 15⬚. The latter displays a regular annulus at approximately 55⬚

4.1. Growth mechanisms

4. Discussion The first interesting result in this study is the presence at any substrate temperature of the sole metastable tetragonal phase t⬘ in as-deposited coatings according to the phase diagram w6x. This phase, which exhibits good thermomechanical resistance, is required for gas turbine applications. In all cases, no monoclinic phase has been detected.

The types of coating morphology observed are in good agreement with the structures described by Movchan and Demchishin w9x and Thornton w10x, explained mainly by increasing surface and volume diffusion with increasing temperature. The number of columns per unit area decreases with increasing coating thickness for all substrate temperatures. Van der Drift w13x ascribed this phenomenon to competitive column growth, depending on the growth rates of facets. Moreover, he predicts, in the case of infinite diffusivi-

B. Preauchat, S. Drawin r Surface and Coatings Technology 142᎐144 (2001) 835᎐842 ´

839

Fig. 3. As-deposited coating ŽTsubstrate s 900⬚C.: Ža. fractograph showing column heads and a secondary nucleus on one facet; Žb. surface micrograph; Žc. micrograph of a polished cross-section; and Žd. surface micrograph showing a column tip at high magnification.

ties, that a cubic crystal growing along the ²100: directions presents a four-faced geometry with  1114 facets, as observed for the 900⬚C coatings. The small pyramidal nuclei observed on 900⬚C coating column heads

ŽFig. 3a and Fig. 6. due to secondary nucleation indicate that the first growth stages follow his theoretical model. The appearance of the column boundaries is

Fig. 4. Diffraction patterns obtained on 900⬚C PECVD as-deposited coatings and after annealing at 1100⬚C for 125 h, 250 h, 500 h and 1000 h.

Fig. 5. 1114t⬘-pole figure obtained on a 200-␮m-thick 900⬚C PECVD as-deposited coating showing an annulus typical of a fibre texture along the w200xt⬘ axis.

840

B. Preauchat, S. Drawin r Surface and Coatings Technology 142᎐144 (2001) 835᎐842 ´

to investigate the evolution of the Ž111. peak intensity with coating thickness. 4.2. E¨ aluation of coating stability

Fig. 6. Surface micrograph of as-deposited coating ŽTsubstrate s 900⬚C.. Nucleus grown by secondary nucleation on a column head facet.

characteristic of a dendritic growth, so we can assume that some competition exists between both fastest growth directions of the cubic crystal,  1114 and  1004 . The existence of a weak Ž111.t⬘ peak in ␪᎐2␪ patterns w I Ž111.t⬘rI Ž200.t⬘ ; 7%x is partly attributed to the presence of small pyramidal nuclei grown on the column head facets, one  1114 facet being parallel to the substrate. After polishing to half-thickness, the coating presents a strongly attenuated Ž111.t⬘ peak intensity w I Ž111.t⬘rI Ž200.t⬘ ; 0.2%x. Further work is in progress

A problem encountered with our coatings is the presence of a systematic shift of the XRD peak positions towards higher angles with respect to the expected positions. The yttria contents given in Tables 3 and 4, calculated using Eq. Ž6. for tetragonal phases and Eq. Ž7. for the fluorite phase, are therefore slightly underestimated. For this reason, the fluorite phase exhibits calculated yttria contents not included in the validity domain of Eq. Ž7.. It is unlikely that the shift is due to thermal stresses originating from the thermal expansion mismatch between the coating ŽCTE; 11 = 10y6 Ky1 for bulk zirconia; slightly different values have been found on EBPVD coatings along the c and a axes w14x. and substrate ŽCTE; 8 = 10y6 Ky1 for bulk alumina., because the columnar structure Žwith columns perpendicular to the substrate surface . does not transmit in-plane tensile stresses efficiently. Growth stresses Žcalculated to be approx. 1 GPa, assuming a Young’s modulus for direction w100x of approx. 350 GPa w15x., which include all other phenomena present in a thermodynamically ‘non-equilibrium material’, are presumed to be responsible for the shift. Nevertheless, Eqs. Ž6. and Ž7. can be used to give indicative coating

Table 4 Phase distribution and lattice parameters after annealing for coatings deposited at 900⬚C a Yttria content of as-deposited coating Žmol.% YO1.5 .b

Heat treatment

Phases

Mole fraction Ž%.

Lattice parameter Žnm.

Yttria content of the phase Žmol.% YO1.5 .c

6.4r8.2

1100⬚C Ž1000 h.

t t⬘ t⬘1 f

38 59 3

at s 0.5091 at ⬘ s 0.5101 as 0.5130

-3 5.9 14.0 9.3d

6.6r7.9

1200⬚C Ž500 h.

t t⬘ t⬘1 f

63 15 22

at s 0.5091 at ⬘ s 0.5101 as 0.5121

3.2 5.8 11.5 3.6 d

8.8r8.7

1300⬚C Ž50 h.

t t⬘ t⬘1 f

56 14 30

at s 0.5094 at ⬘ s 0.5105 as 0.5122

3.9 6.9 11.8 4.1d

8.8r8.7

1400⬚C Ž20 h.

t t⬘ t⬘1 f

34 26 40

at s 0.5094 at ⬘ s 0.5108 as 0.5124

3.9 7.9 12.2 5.0 d

a

See text for further explanations. Left number: EPMA result; right number: XRD result using Eq. Ž5.. c XRD result using Eq. Ž6.. d XRD result using Eq. Ž7.. b

B. Preauchat, S. Drawin r Surface and Coatings Technology 142᎐144 (2001) 835᎐842 ´

Fig. 7. Fitting result on XRD pattern in the 4004-region for a coating deposited at 900⬚C and annealed for 500 h at 1200⬚C. Note the non-linear y-scale.

841

ments. VanValzah and Eaton w17x focused on the monoclinic phase in requenched PS TBCs and have shown that low yttria-content t phase can be retained at room temperature by quenching; a 1-h treatment at approximately 300⬚C is enough to complete the transformation t ª m. If we compare the 900⬚C PECVD coatings with the PS coatings w11x, the crystallographic stability seems to be of the same order of magnitude. Few systematic investigations have been published on the influence of heat treatment on EBPVD coatings. The stability of the 900⬚C PECVD TBCs compares favourably to Schulz’s EBPVD TBCs w12x, if the respective fluorite phase contents, which are less sensitive to the cooling route and is different in the two studies, are considered.

5. Conclusion compositions Žwhich remain coherent with EPMA bulk results . and to follow their evolution after annealing. Another difficulty is the strong w200x texture and the peak superposition that often hinders the accurate positioning of the Ž004.t,t⬘ peaks ŽFig. 7. and prevents the use of the more precise Eq. Ž5., for which the lattice parameter c is needed. The fit results are, however, coherent: the same texture is observed for t⬘ and t phases. This is not really surprising, because their tetragonality is low Ž cra ; 1. and many authors w8,11,12,16x reported the decomposition of the t⬘ phase at high temperatures into small t and f precipitates in a t⬘ matrix. Our experimental data do not allow the rigorous distinction of the fluorite phase from a hypothetical very high yttria-content tetragonal phase1 t⬘1 , because the corresponding peaks are widened by the small precipitate size. Furthermore, because of the shift observed in the diffraction patterns, the yttria contents calculated with Eqs. Ž6. and Ž7. are far out of the validity domain for both equations Žsee Tables 3 and 4., and thus cannot help to dispel the ambiguity. However, a comparison with published data shows that it is the fluorite phase which is likely to be formed during the destabilisation process. It is evident that the yttria content of the t⬘ phase decreases with annealing duration for all temperatures, except for 1400⬚C for both 700⬚C and 900⬚C coatings; annealing of a new set of samples should confirm or refute this behaviour. The absence of the monoclinic phase is ascribed to quenching of the samples after the various heat treat-

1

Note that the same ambiguity has been reported by Lelait on PS TBCs w11x.

Good morphological and crystallographic properties were found for PECVD thick coatings deposited at 900⬚C. Coatings with relatively dense columns are desired for TBC applications: both PECVD and EBPVD processes allow the deposition of such coatings, which exhibit good resistance towards sintering. Other properties relevant to this type of application are currently being investigated, such as thermal properties and lifetime in cyclic high-temperature conditions.

Acknowledgements The DGA is greatly acknowledged for supporting the thesis of B. Preauchat. Thanks are due to Mrs Sanchez ´ for X-ray analysis, to Y. Pioche for EPMA analysis, to Mrs Raffestin for providing SEM micrographs. References w1x R.A. Miller, Surf. Coat. Technol. 30 Ž1987. 1᎐11. w2x G.W. Goward, Surf. Coat. Technol. 108r109 Ž1998. 73᎐79. w3x S. Chevillard, M.H. Vidal-Setif, ´ S. Drawin, AGARD Report R-823, Ž1998. 11.1᎐11.9. w4x B. Preauchat, S. Drawin, S. Landais, J. Phys. IV Fr. 10 Ž2000. ´ Pr4r149᎐Pr4r154. w5x J.F. Bisson, D. Fournier, M. Poulain, O. Lavigne, R. Mevrel, J. ´ Am. Ceram. Soc. 83 Ž8. Ž2000. 1993᎐1998. w6x H.G. Scott, J. Mater. Sci. 10 Ž1975. 1527᎐1535. w7x R.A. Miller, J.L. Smialek, R.G. Garlick, in: A.H. Heuer, L.W. Hobbs ŽEds.., Science and Technology of Zirconia I, Advances in Ceramics, 3, The American Ceramic Society, 1982, pp. 241᎐253. w8x E.H. Kisi, C.J. Howard, Key Eng. Mater. 153r154 Ž1998. 1᎐36. w9x B.A. Movchan, A.V. Demchishin, Fiz. Metal. Metalloved. 28 Ž1969. 83᎐90. w10x J.A. Thornton, J. Vac. Sci. Technol. 7 Ž1977. 239᎐260. w11x L. Lelait, These, ` Universite´ d’Orsay, 1991.

842

B. Preauchat, S. Drawin r Surface and Coatings Technology 142᎐144 (2001) 835᎐842 ´

w12x U. Schulz, J. Am. Ceram. Soc. 83 Ž2000. 904᎐910. w13x A. van der Drift, Philips. Res. Rep. 22 Ž1967. 267᎐288. w14x N. Czech, H. Fietzek, M. Juez-Lorenzo, V. Kolarik, W. Stamm, Surf. Coat. Technol. 113 Ž1999. 157᎐164. w15x R.P. Ingel, D. Lewis III, J. Am. Ceram. Soc. 71 Ž4. Ž1988. 265᎐271.

w16x A.H. Heuer, R. Chaim, V. Lanteri, in: S. Somiya, N. Yamamoto, H. Hanagida ŽEds.., Science and Technology of Zirconia III, Advances in Ceramics, 24A, The American Ceramic Society, 1988, pp. 3᎐20. w17x J.R. VanValzah, H.E. Eaton, Surf. Coat. Technol. 46 Ž1991. 289᎐300.