Materials Science and Engineering C 43 (2014) 45–49

Contents lists available at ScienceDirect

Materials Science and Engineering C

j ourna l homepage: www.e lsev ie r .com/ locate /msec

Growth and surface characterization of TiNbZr thin films deposited bymagnetron sputtering for biomedical applications

D.A. Tallarico a, A.L. Gobbi b, P.I. Paulin Filho c, M.E.H. Maia da Costa d, P.A.P. Nascente c,⁎a Federal University of Sao Carlos, Materials Science and Engineering Graduation Program, Via Washington Luis km 235, CEP 13565-905 Sao Carlos, SP, Brazilb Brazilian Nanotechnology National Laboratory, Rua Giuseppe Máximo Scolfaro 10.000, CEP 13083-100 Campinas, SP, Brazilc Federal University of Sao Carlos, Department of Materials Engineering, Via Washington Luis km 235, CEP 13565-905 Sao Carlos, SP, Brazild Pontifical Catholic University of Rio de Janeiro, Department of Physics, CEP 22451-900 Rio de Janeiro, RJ, Brazil

⁎ Corresponding author. Tel.: +55 16 33518528; fax: +E-mail address: nascente@ufscar.br (P.A.P. Nascente).

http://dx.doi.org/10.1016/j.msec.2014.07.0130928-4931/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 6 December 2013Received in revised form 2 June 2014Accepted 2 July 2014Available online 9 July 2014

Keywords:Titanium alloysThin filmsBiomaterialsXPSNanoindentation

Low modulus of elasticity and the presence of non-toxic elements are important criteria for the development ofmaterials for implant applications. Low modulus Ti alloys can be developed by designing β-Ti alloys containingnon-toxic alloying elements such asNb and Zr. Actually,most of themetallic implants are producedwith stainlesssteel (SS) because it has adequate bulk properties to be used as biomaterials for orthopedic or dental implantsand is less expensive than Ti and its alloys, but it is less biocompatible than them. The coating of this SS implantswith Ti alloy thin films may be one alternative to improve the biomaterial properties at a relatively low cost.Sputtering is a physical deposition technique that allows the formation of nanostructured thin films. Nanostruc-tured surfaces are interesting when it comes to the bone/implant interface due to the fact that both the surfaceand the bone have nanoscale particle sizes and similar mechanical properties. TiNbZr thin films were depositedon both Si(111) and stainless steel (SS) substrates. The TiNbZr/Si(111) film was used as a model system, whilethe TiNbZr/SS film might improve the biocompatibility and extend the life time of stainless steel implants. Themorphology, chemical composition, Young's modulus, and hardness of the films were analyzed by atomic forcemicroscopy (AFM), X-ray photoelectron spectroscopy (XPS), energy-dispersive X-ray spectroscopy (EDS), andnanoindentation.

55 16 33611160.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

Materials for bone replacement might mimic the architecture of thebone [1].Most implants aremetals: stainless steels, Co–Cr systemalloys,and titanium alloys [2]. The Young's moduli of those biomaterials aremuch greater than that of cortical bone, then bone resorption occurs[2]. That is because when the bone is stressed, the bone-producingcells called osteoblasts are stimulated into generating more bone [3].So if the bone is replaced by a metallic counterpart that is stiffer thanthe original bone, the replacementwill tend to bear a greater proportionof the load, shielding the surrounding skeleton from its normal stresslevels. Thus, bone replacements cemented to neighboring bonemay be-come loose over time, as the surrounding bone is resorbed [3]. There-fore, the presence of non-toxic elements and low modulus of elasticityare the two most important criteria for the development of materialsfor implant applications. Lowmodulus Ti alloys can be developed by de-signing β-Ti alloys containing non-toxic alloying elements such as Nb,Zr, and Ta [4–6]. The β titanium alloys consist mainly of low modulussingle β phase and therefore the modulus of these alloys is low [7].Therefore, the elastic modulus is an important parameter, as a value

close to that of the bone material leads to a better transfer of functionalloads to the bone, enhancing the stimulation for new bone growth[8–10].

The surface characterization of implant materials is a topic of mainimportance since the surface plays a key role in the living tissue re-sponse to the metal presence. There are several aspects of the biomate-rials surface–livingmedia interactions that are still not solved. This topichas been extensively studied in titanium and its alloys [11–14]. In thesematerials, it was found that both the topography and the chemical sur-face composition have a strong influence in the early stages of theosseointegration process. The surface–biological media interactionsare a multiscale problem, with a broad range from a few microns,where the surface topography changes the effective contact area be-tween the implant and the surrounding bone [11,12], to tens of nano-meters, where the influence of chemical species present on the surfacecanmodify the nucleation and growth of Ca–P compounds that precedethe formation of hydroxyapatite [13,14].

Sputtering is a physical deposition technique that has been widelyused in several industrial applications with great success, and allowsthe formation of nanostructured thin films [15]. In the biomedicalfield, however, this is a novel technique for whichmany potential appli-cations are prompting extensive research [16–19]. Nanostructured sur-faces are interestingwhen it comes to the bone/implant interface due to

Fig. 1. Non-contact tapping mode 3D images for TiNbZr/Si(111).

Table 1Values of roughness and grain sizes for TiNbZr films.

Roughness Ra (nm) Grain size (nm)

TiNbZr/Si(11) 3.0 53TiNbZr/SS 2.8 70

46 D.A. Tallarico et al. / Materials Science and Engineering C 43 (2014) 45–49

the fact that both the surface and the bone have nanoscale particle sizesand similar mechanical properties [20]. The high surface energy ofnanostructured materials leads to desirable cellular responses sincebone-forming cells generally attach themselves to surfaces with rough-ness in the nanometer range [20]. The combination of these characteris-tics causes an increase of fracture resistance and biocompatibility for theimplants. In addition, the particles generated by nanostructured im-plants are not immunoreactive and therefore less harmful to thehuman body than the microparticles of conventional implants [21].

Stainless steel (SS) has adequate bulk properties to be used as bioma-terials for orthopedic or dental implants and is less expensive than Ti andits alloys, but it is less biocompatibility than them. In this study, TiNbZrthin films are proposed as a surface coating for SS. The TiNbZr alloyswere deposited by magnetron sputtering on Si(111) and SS substrates.The TiNbZr/Si(111) coating film was used as a model system. The mor-phology, chemical composition, Young's moduli, and hardness of thecoating films were analyzed by AFM, XPS, EDS, and nanoindentation.

2. Materials and methods

The titanium, niobium, and zirconium films were sputter depositedonto cleaned (111) silicon and stainless steel substrates at roomtemperature from zirconium, titanium, and niobium metal targets(0.060 m diameter × 0.003 m thick, 99.9% pure); the substrates weremechanically clamped to the dc magnetron cathode of a conventionalsputtering system (Balzers BA510). Argon (99.999% pure) was used asthe sputtering gas. The target–substrate separation was 0.260 m. Thefilms were deposited under conditions: 380 W (Ti), 90 W (Nb), 50 W(Zr) power density, 500 mA current, 500 V voltage, 2300 s depositiontime, and 2 × 10−5 Pa base vacuum. The film elements were depositedsimultaneously and the films thickness was 460 nm for both substrates.

Fig. 2. Non-contact tapping mode 3D images for TiNbZr/SS.

Surface morphology and roughness were determined using acommercial atomic force microscope (MultiMode 8 Bruker AXS).The instrument was operated in tapping mode, and the image sizeswere 500 × 500 nm2. To characterize the roughness, the averageroughness (Ra) was obtained as well as the standard deviation ofthe amplitude (Rq) and the maximum peak (Rmax).

The X-ray photoelectron spectroscopy (XPS) analyses wereperformed under ultrahigh vacuum (low 10−7 Pa range) employing aThermo Electron ESCALAB 250 with an Al Kα (hν= 1486.6 eV) mono-chromatized focused X-ray source. The spectrometer was calibratedagainst the reference binding energies (BEs) of clean Cu (Cu 2p3/2 at932.6 eV), Ag (Ag 3d5/2 at 368.2 eV) and Au (Au 4f7/2 at 84.0 eV) sam-ples. The analyzed area had a diameter of about 500 μm. In addition tothe survey spectrum (pass energy of 100 eV, step energy of 1 eV), thefollowing core levelswere systematically recorded at higher energy res-olution (pass energy of 20 eV): C 1 s, O 1 s, Ti 2p, Zr 3d, Nb 3d and Si 2p(step energy of 0.1 eV). To take into account surface charging effects,core levels were referenced by setting the lowest BE component of theresolved C 1 s peak (corresponding to adventitious carbon in a hydro-carbon environment) to 284.8 eV. Core level peak decompositionswere performed with the CasaXPS© program. All peaks were fittedusing a Shirley background, by using a 70% Gaussian/30% Lorentzianpeak shape. Argon ion sputtering (partial pressure of 8 × 10−5 Pa)was used to profile the samples in a 2 mm × 2 mm area.

The overall surface morphology of the specimens was observed byscanning electron microcopy (SEM) (JEOL JSM-6460LV) at 15 kV. Sur-face elemental analysis was simultaneously determined by energy dis-persive X-ray spectroscopy (EDS) (EDAX Genesis XM4-Sys 60). TheEDS analysis was carried out at least in two different regions per sample,and a mapping of selected regions as well as punctual analysis of differ-ent morphologies of interest was included.

Nanoindentation tests were carried out with a TriboIndenternanoindenter (Hysitron Inc.) using a Berkovich diamond tip. Each spec-imen was tested at room temperature and the measurements weredone at extremely small penetration depths with 5 μN maximum in-denter load. The statistical analysis of the measured results allowedfor the hardness (H) and Young's modulus (E) values. To eliminate

Fig. 3. Long scan spectra for TiNbZr films deposited on Si(111).

Fig. 4. Long scan spectra for TiNbZr films deposited on stainless steel.

47D.A. Tallarico et al. / Materials Science and Engineering C 43 (2014) 45–49

any influence of the substrate material in the measurement of Young'smodulus, the penetration range of the indenter was limited to a depthb0.2d, where d is the film thickness.

3. Results and discussion

3.1. AFM analysis

Figs. 1 and 2 display the morphology of the films deposited onSi(111) and SS substrates, respectively. The roughness and grain sizecan be analytically described by the parameters (Ra and grain size) ex-tracted from the image analyses, presented in Table 1.

The average roughness (Ra) of the sampleswas established from theAFM images before and after deposition, over a surface region of 500 nm× 500 nm. Each samplewas analyzed at five randomly chosen locations.Statistical analyses were performed using the standard deviation. In thecase of the silicon single crystal, the alloy film showed a larger rough-ness (Ra = 3.0 nm) than the bare substrate (Ra = 0.6 nm) [22], andin the case of stainless steel, the film presented a lower roughness(2.8 nm) than the raw substrate (Ra = 7.6 nm) [22]. Both depositedsamples exhibit statistically equal roughness (their difference are with-in 7%), showing that the roughness of the deposited thin film is inde-pendent of the substrate roughness for this thickness deposition(460 nm), likewise, the grain size magnitude for both samples is veryclose (their difference are within 32%) (see Table 1).

Fig. 5. Spectra for TiNbZr/Si(111): (a

According to Geetha et al. [20], the interest in nanostructuredma-terials for medical applications is based on the fact that human bonesconsist of inorganic minerals of grain sizes varying from 20 to 80 nmin length and 2 to 3 nm in diameter. The variation in the surface en-ergy due to the surface roughness leads to desirable cellular re-sponses on nanostructured titanium and other materials resultingin high osseointegration [23]. Nanograined materials made of Ti(Cp), Ti–6Al–4V, CoCr, and ceramic biomaterials such as alumina, ti-tania, hydroxyapatite also exhibit increased cell adhesion [24,25].When the grain sizes decreased from 167 to 24 nm, the osteoblastadhesion increased by 51% and the fibroblast adhesion responsiblefor encapsulation was reduced by 235% [25,26].

3.2. XPS and EDS analysis

Figs. 3 and 4 display the long scan spectrum to TiNbZr films deposit-ed in the Si(111) and stainless steel substrates respectively. Both spectrashow that the film surface has C, O, Ti, Nb and Zr, confirming the forma-tion of the alloy. It should be mentioned that carbon is commonly de-tected by XPS and it is due to adsorbed hydrocarbon molecules;oxygen also is detected and it is due to adsorbed CO and/or CO2, and/or metallic oxides formed on the surface. No trace of the substrate wasdetected, suggesting a good coating.

Fig. 5 shows the Ti 2p, Nb 3d, and Zr 3d core level spectra for the as-deposited and oxidized TiNbZr deposited on Si(111). The Ti 2p corelevel was fitted with two doublets (Fig. 5(a)), located at 453.9 eV and458.6 eV which corresponds to metallic Ti and TiO2 respectively. TheNb 3d core level spectrum (Fig. 5(b)) was decomposed with three dou-blets, located at 202.4 eV, 204.5 eV, and 207.2 eV. These components areassigned to metallic Nb, NbO, and Nb2O5 respectively. The Zr 3d corelevel presents up to two doublets (Fig. 5(c)), the more intense locatedat 182.2 eV and at 178.2 eV. The more intense doublet corresponds toZrO2. The onewith low intensity corresponds to metallic Zr in the layer.

Fig. 6 shows the Ti 2p, Nb 3d, and Zr 3d core level spectra for the as-deposited and oxidized TiNbZr deposited on SS. The Ti 2p core level wasfitted with two doublets (Fig. 6(a)), located at 453.8 eV and 458.6 eVwhich corresponds to metallic Ti and TiO2 respectively. The Nb 3dcore level spectrum (Fig. 6(b)) was decomposed with two doublets, lo-cated at 205.2 eV, and 207.2 eV. These components are assigned NbO,and Nb2O5 respectively. The Zr 3d core level presents only one doublet(Fig. 6(c)) located at 182.4 that corresponds to ZrO2.

Table 2 presents the TiTiþNbþZr,

NbTiþNbþZr, and

ZrTiþNbþZr atomic ratios for the

TiNbZr films deposited on Si(111) and stainless steel substrates obtain-ed by XPS. According to Table 2we can infer that the surfaces of the filmcompositions (in atomic %) are approximately Ti30Nb20Zr/Si (111) andTi16Nb21Zr/SS.

) Ti 2p, (b) Nb 3d, and (c) Zr 3d.

Fig. 6. Spectra for TiNbZr/SS: (a) Ti 2p, (b) Nb 3d, and (c) Zr 3d.

48 D.A. Tallarico et al. / Materials Science and Engineering C 43 (2014) 45–49

Table 3 presents the TiTiþNbþZr,

NbTiþNbþZr, and

ZrTiþNbþZratomic ratios for the

TiNbZr films deposited on Si(111) and stainless steel substrates obtain-ed by EDS. According to Table 3 we can infer that the bulk film compo-sitions (in atomic %) are approximately Ti18Nb18Zr/Si (111) andTi16Nb20Zr/SS. For the film deposited on silicon substrate, XPS andEDS analyses indicated that the thin film has 28% less titanium, 60%more Nb, and 11% more Zr on the near-surface region, probed by XPS,as compared to the bulk composition,measured by EDS. For thefilm de-posited on SS substrate, XPS and EDS analyses showed that the thin filmhas 1.5% less titanium, the sameNb concentration and 5%more Zr in thissurface, with compared with the bulk. Therefore, the compositionremained almost constant in depth for the stainless steel substrate,which makes this sample very interesting for further biomolecularinteraction.

3.3. Nanoindentation analysis

The elastic Young's modulus (E) and the hardness (H) of TiNbZr/Si(111) and TiNbZr/SS thin films obtained by using the nanoindentationtechnique are depicted in Fig. 7 as a function of indentation depth. Eachvalue is an average of 10–15measurements. Formetallic biomaterials, itis desirable low elastic modulus, close to the values of the Young'smod-ulus of nature bone (10–40 GPa), to enable a better bone-implant me-chanical cohesion [27]. Fig. 7 shows a non-uniform behavior near thesurface (approximately first 40 nm). Since these differences were ob-served for the three thin films used in this study, they were associatedwith the experimental conditions. As suggested by Cáceres et al. in asimilar study with titanium alloys [28], this could be a combination oferrors in depth determination at the very small contact displacementand the greater effect of the surface roughness, which can set a lowerlimit to the useful nanoindentation size. After the first nanometers,the measurements showed a constant value (plateau region).

The results showed in the Fig. 7 are represented in the Table 4.Table 4 shows that both films presented similar values of Young's mod-ulus (their difference is within 5.3%) and hardness, their difference iswithin 1.7%, independent of the substrate.

Table 2Atomic ratios obtained by XPS for TiNbZr films deposited on Si(111) and stainless steelsubstrates.

TiNbZr film TiTiþNbþZr

NbTiþNbþZr

ZrTiþNbþZr

Si(111) substrate 0.50 0.30 0.20SS substrate 0.63 0.16 0.21

According to Liu et al. [27] it is of interest that themodulus of elastic-ity of thematerial used tomanufacture implants is as close as possible tothe value of the bone. This closeness of values of the module facilitatesthe transfer of functional loads between implant and bone and helpsprevent damage to the tissue around the recovered material. From thepoint of view of wear, it is important that the implant surface has highhardness to prevent abrasionwaste generated during functional loadingto be released in the body. Thus, it is important to associate a lowmod-ulus of elasticity with high hardness values. The value of Young's mod-ulus is approximately 100 GPa and ismuch smaller than values found tostainless steel (210 GPa) and Co–Cr alloys (204–240 GPa) [29–31]. Thevalue of hardness (8GPa)was greater than the hardness of commercial-ly pure titanium, titanium alloys such as traditional Ti6Al4V (4.9 GPa),and stainless steel 316 L (2 to 6.7 GPa) [30–32]. Both thin films present-ed good values of elastic modulus and hardness which are better thanvalues reported for commercial titanium alloys (some containing niobi-um and zirconium) used in implants [27–32].

4. Conclusions

The roughness, grain size magnitude, Young's modulus, and hard-ness of the deposited thin films were the same for both depositions, in-dependent of the substrate characteristics, for a thickness of 460 nm.Furthermore, the composition remained almost constant in depth, par-ticularly for the stainless steel substrate, showing the reliability of thisdeposition technique.

The surfaces of the thin films were oxidized, and this is highly desir-able for implant materials since an oxidized surface layer providesgreater corrosion protection. The nanostructured grain sizes obtainedfor the films allowed a variation in the surface energy that could leadto desirable cellular responses resulting in high osseointegration.

The Young's modulus of the film deposited on stainless steel wasabout 50% smaller than the modulus of the substrate and the hardnessof the film deposited on stainless steel was about 100% greater thanthe hardness obtained for the substrate. Thus, the deposition of TiNbZrcoating films on the stainless steel implants could yield in Young's

Table 3Atomic ratios obtained by EDS for TiNbZr films deposited on Si(111) and stainless steelsubstrates.

TiNbZr film TiTiþNbþZr

NbTiþNbþZr

ZrTiþNbþZr

Si(111) substrate 0.64 0.18 0.18SS substrate 0.64 0.16 0.20

Fig. 7. (a) Young's modulus and (b) hardness for the TiNbZr/Si(111) and TiNbZr/SS thin films.

Table 4Young's modulus and hardness for TiNbZr films deposited on Si(111) and stainless steelsubstrates.

TiNbZr film Young's modulus (GPa) Hardness (GPa)

Si(111) substrate 95 8.03SS substrate 100 8.17

49D.A. Tallarico et al. / Materials Science and Engineering C 43 (2014) 45–49

modulus more suitable to the bone value. Regarding wear, it is impor-tant that the implant surface has high hardness to prevent abrasionwaste generated during functional loading to be released in the body.

Acknowledgments

This work has been supported by the Brazilian Synchrotron LightSource (LNLS) under proposals LMF-8912 and LMF-10475. Grants andscholarships from the Brazilian agencies FAPESP (process 2009/17055-7), CAPES, and CNPq (processes 302001/2010-7 and 470816/2007-0) are gratefully acknowledged.

References

[1] G.E. Ryan, A.S. Pandit, D.P. Apatsidis, Biomaterials 29 (2008) 3625–3635.[2] M. Niinomi, Sci. Technol. Adv. Mater. 4 (2003) 445–454.[3] A.E. Aguilar Maya, D.R. Grana, A. Hazarabedian, G.A. Kokubu, M.I. Luppo, G. Vigna,

Mater. Sci. Eng. C 32 (2012) 321–329.[4] L. You, X. Song, Mater. Lett. 80–1 (2012) 165–167.[5] T. Akahori, M. Niinomi, H. Fukui, M. Ogawa, H. Toda, Mater. Sci. Eng. C 25–3 (2005)

248–254.[6] M. Niinomi, T. Akahori, T. Takeuch, S. Katsura, H. Fukui, H. Toda, Mater. Sci. Eng. C

25–3 (2005) 417–425.[7] C.J. Boehlert, C.J. Cowen, C.R. Jaeger, M. Niinomi, T. Akahori, Mater. Sci. Eng. C 25–3

(2005) 263–275.[8] S. Sienz, S. Mändl, B. Rauschenbach, Surf. Coat. Technol. 156 (1–3) (2002) 185–189.

[9] M. Takeuchi, Y. Abe, Y. Yoshida, Y. Nakayama, M. Okazaki, Y. Akagawa, Biomaterials24–10 (2003) 1821–1827.

[10] C. Wirth, V. Comte, C. Lagneau, P. Exbrayat, M. Lissac, N. Jaffrezic-Renault, L.Ponsonnet, Mater. Sci. Eng. C 25 (2005) 51–60.

[11] G. Mendonça, D.B.S. Mendonça, F.J.L. Aragão, L.F. Cooper, Biomaterials 29 (2008)3822–3835.

[12] J.E. Davies, Biomaterials 28 (2007) 5058–5067.[13] A.L. Oliveira, J.F. Mano, R.L. Reis, Curr. Opin. Solid State Mater. Sci. 7 (2003) 309–318.[14] F. Barrere, M.M.E. Snel, C.A. van Blitterswijk, K. de Groot, P. Layrolle, Biomaterials 25

(2004) 2901–2910.[15] J.M. Schneider, S. Rohde, W.D. Sproul, A. Matthews, J. Phys. D. Appl. Phys. 33 (2000)

173–186.[16] D.L. Wise, D.J. Trantolo, D.E. Altobelli, M.J. Yaszemski, J.D. Gresser, E.R. Schwartz, En-

cyclopedic Handbook of Biomaterials and Bioengineering, Part B, Applications, Mar-cel Dekker Inc, New York, 1995.

[17] S. Mandl, B. Rauschenbach, Surf. Coat. Technol. 156 (2002) 276–283.[18] P.K. Chu, J.Y. Chen, L. Wang, N. Huang, Mater. Sci. Eng. R. 36 (2002) 143–206.[19] R. Olivares-Navarrete, J.J. Olaya, C. Ramírez, S.E. Rodil, Coatings 1 (2011) 72–87.[20] M. Geetha, A.K. Singh, R. Asokamani, A.K. Gogia, Prog. Mater. Sci. 54 (2009)

397–425.[21] L.G. Gutwein, T.J. Webster, Biomaterials 25 (2004) 4175.[22] D.A. Tallarico, A.L. Gobbi, P.I. Paulin Filho, A. Galtayries, P.A.P. Nascente, J. Vac. Sci.

Technol. A 30–5 (2012) 051505–051508.[23] F.S. Kaplan, W.C. Hayes, T.M. Keaveny, A. Boskey, T.A. Einhorn, J.P. Iannotti, Orthope-

dic basic science, in: S.P. Simon (Ed.), American Academy of Orthopedic Surgeons,Columbus, OH, 1994, pp. 127–185.

[24] K. Dongwoo, L. Jing, Y. Chang, K.M. Haberstroh, T.J. Webster, Biomaterials 29 (2008)970–983.

[25] T.J. Webster, J.U. Ejiofor, Biomaterials 25 (2004) 4731–4739.[26] T.J. Webster, C. Ergumn, R.H. Doremus, R.W. Siegel, R. Bizios, Biomaterials 21 (2000)

1803–1810.[27] X. Liu, P.K. Chu, C. Ding, Mater. Sci. Eng. R 27 (3–4) (2004) 49–121.[28] D. Cáceres, C. Munuera, C. Ocal, J.A. Jiménez, A. Gutiérrez, M.F. López, Acta Biomater.

4 (5) (2008) 1545–1552.[29] O.P. Karasevskaya, O.M. Ivasishin, S.L. Semiatin, Y.V. Matviychuk, Mater. Sci. Eng. A

354 (2003) 121–132.[30] M. Niinomi, Mater. Sci. Eng. A 243 (1998) 231–236.[31] S.L. Dickerson, J.C. Gibeling, Mater. Sci. Eng. A 278 (2000) 121–134.[32] J. Musil, Surf. Coat. Technol. 125 (1–3) (2000) 322–330.