Thin Solid Films 460 (2004) 101–115
Plasma spraying of nanostructured partially yttria stabilized zirconia powders J.F. Lia,1, H. Liaoa,*, X.Y. Wanga,2, C. Coddeta, H. Chenb, C.X. Dingb a
´ ´ Laboratoire d’Etudes et de Recherches sur les Materiaux, les Plasmas et les Surfaces, Universite´ de Technologie de Belfort-Montbeliard, 90 010 Belfort Cedex, France b Shanghai Institute of Ceramics, Chinese Academy of Sciences, Shanghai 200050, PR China Received 14 March 2003; received in revised form 21 November 2003; accepted 6 January 2004 Available Online 27 March 2004
Abstract Nanosized partially yttria stabilized zirconia particles, prepared using a co-precipitation method, were reprocessed into agglomerate powders using two methods for plasma spraying. The first method was to make micrometer-sized agglomerates directly following the grinding of the calcined yttria–zirconia agglomerates. The second method was to reconstitute the nanosized particles into micrometer agglomerates using spray drying. The deposition efficiency, porosity, microhardness and average grain size of the deposits made from these two reprocessed powders were studied. Distinct results related to the process parameters were obtained for the two types of powders. The second type of powder was more suitable for plasma spraying than the first one. Using the second type of powder, some unique results distinguished from those of the conventional partially yttria stabilized zirconia powders were observed and an optimized coating with a porosity of 3.8%, Hv0.3 of 953 and mainly consisting of 1–3 mm columnar grains in the columnar direction and smaller than 100 nm in their cross-sections was achieved. 䊚 2004 Elsevier B.V. All rights reserved. Keywords: Plasma processing and deposition; Coatings; Nanostructures; Statistical experiment
1. Introduction Thermal spray using 10–100 mm sized metallic and ceramic feedstock powders has been widely used for many years to deposit coatings applied in modifying surface properties of engineered components w1,2x. Recently, the uses of nanostructured powders and solution-precursors as thermal spray feedstocks have appeared and have been increasingly investigated to deposit nanostructured coatings w3–14x. These innovations are driven and developed by the two following reasons: as regards thermal spray, when finer-sized *Corresponding author. Tel.: q33-38458-3242; fax: q33-384583030. E-mail address:
[email protected] (H. Liao). 1 J.F. Li is presently at Laser Processing Research Centre of Department of Mechanical, Aerospace and Manufacturing Engineering and Corrosion and Protection Centre, University of Manchester Institute of Science and Technology (UMIST), P.O. Box 88, Manchester M60 1QD, UK. 2 X.Y. Wang is presently at Department of Textile, UMIST, P.O. Box 88, Manchester M60 1QD, UK.
particles or solution-precursors are used as feedstock, coatings with novel structures and improved properties may be prepared. For example, nanostructured coatings prepared through these innovations may reduce the amount of interlamellar boundariesycracks or even be absent from the interlamellar boundariesycracks. Consequently, these coatings may give increased thermal corrosion and wear resistance, as well as a better surface and other useful features w15–18x. As regards materials preparation, they may make use of the quick solidification during thermal spray to avoid extensive grain growth in the coatings so as to provide a new approach to prepare consolidate nanomaterials w16,19x. The use of solution-precursor feedstock in plasma spray offers several potential advantages over the conventional powder feedstock. These include a better control over the chemistry of the coating, the ability to deposit compositionally-graded coatings with ease, the ability to deposit coatings that are inherently nanostructured and processing versatility w12–14x. However, the use of nanostructured powders as feedstock was pre-
0040-6090/04/$ - see front matter 䊚 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2004.01.078
102
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
vailing during investigation on thermal sprayed nanostructured coatings w3–11x. This may be due to the fact that the use of powder feedstock is a well-established process in thermal spray. It is essentially impossible to melt individual nanometer-sized particles without evaporating during their flight in high-temperature stream, such as plasma jet for example Ref. w18x. In order to overcome this problem, the nanometer-sized grains or particles were introduced within larger micrometer-sized particles by mechanical milling w3–5x or reconstituted into micrometer granules using spray drying w6–14x. It has been demonstrated from the extensive research that thermal spraying of nanostructured powders can deposit high quality stainless steel, WCyCo and Al2O3 –TiO2 coatings w3,8,9,19x. However, some negative results have also been reported on thermal spraying of nanostructured WCyCo powders w6,7x. In fact, the microstructures and properties of thermal sprayed coatings are significantly influenced by feedstock characteristics, such as morphology, size distribution and flowability, and the spraying process conditions w20–26x. These negative results on the thermal spraying of nanostructured WCyCo powders may be due to improper powder characteristics or spraying process conditions. Therefore, it is necessary to optimize both the nanostructured powders and thermal spray process parameters to control the coating microstructures and properties. There are two approaches to identify the effects of the process parameters on the coating microstructure and properties w23x. One approach is directly to study how process parameters affect coating microstructure and properties w20x. The other approach is to develop on-line particle monitoring system to measure the temperature and velocity of in-flight particles and correlate them with the coating microstructure and properties w23–26x. In this paper, the investigation on the plasma spraying of nanostructured partially yttria stabilized zirconia powders was carried out using the first approach with a statistical experiment, the uniform design experiment w27x. Plasma sprayed partially yttria stabilized zirconia (YSZ) coatings are increasingly used as thermal barrier coatings (TBCs) for gas turbines and diesel engines w14,21–26,28,29x. Deposition efficiency is closely related to the physical andyor chemical interaction of nanostructured powders with high-temperature plasma stream w20x. High deposition efficiency is also very important to lower the cost and shorten the process time w28x. A porous coating is useful to lower the thermal conductivity w29x. Relative to micrometer grains, nanosized grains may increase thermal scattering due to increased grain boundaries. Dense nanostructured YSZ coatings, while not increasing the thermal diffusivityyconductivity appreciably, in some cases, could effectively improve their erosion and corrosion resistance as well as the
oxidation resistance of the metallic bond-coat. It can, hence, extend the service life of TBCs w17,30x. Retaining nanometer-sized grains within the dense coatings is also a substantial challenge for preparation of nanostructured coatings. The hardness of nanomaterials is closely related to their grain size w31x. Therefore, during the statistical experiment, the deposition efficiency, porosity, microhardness and average grain size were considered as the objective functions related to the five investigated process parameters, the arc current, argon flow rate, hydrogen flow rate, spray distance and powder feed rate. Note that although the plasma spraying of nanostructured partially yttria stabilized zirconia particles has been reported before w10,11x, these studies were mainly concerned with the determination of phases and grain size of the coatings. In the present study, the plasma spraying of the nanostructured powders was carried out covering a wider range of spraying process parameters by statistical experiment. In addition to the phases and grain size, attention was especially paid to the results of the deposition efficiency, porosity and microhardness of the deposited coatings distinguished from those of the coatings sprayed using the conventional partially yttria stabilized zirconia powders. 2. Experimental procedure 2.1. Preparation of nanostructured yttria–zirconia powders Nanosized yttria–zirconia particles were synthesized using co-precipitation method. The synthesis mechanism is briefly described as follows: zirconium oxy-chloride (ZrOCl2Ø8H2O) solution, doped with yttrium nitrate wY(NO3)3Ø6H2Ox in corresponding ratio of 7wt.% Y2O3 within the Y2O3 –ZrO2 powder, was slowly neutralized with ammonia solution to form a precipitate. The formed precipitate was washed, dried and then calcined at 1000 8C for 2 h to produce nanostructured Y2O3 –ZrO2 agglomerates. X-Ray diffraction (XRD) results showed that the nanostructured Y2O3 –ZrO2 agglomerates mainly consisted of non-transformable tetragonal zirconia with a small amount of monoclinic phase. The average grain size of the agglomerates determined according to XRD peak broadening using the Scherrer formula w8,11x was 37.6 nm, and a transmission electron microscope (TEM) micrograph of the nanostructured particles is shown in Fig. 1. Two procedures were employed to reprocess the nanostructured agglomerates for plasma spraying. For the first procedure, nanosized particles were directly kept within micrometer agglomerates, following suitably grinding the calcined yttria–zirconia agglomerates of several millimetres in size. Such reprocessed agglomerates were referenced to as powder A. The other proce-
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
103
Fig. 1. A TEM micrograph of the starting nanostructural particles. Fig. 3. The particle size distribution curves of the two powders.
dure was first to grind the calcined agglomerates into nanosized particles, the nanosized particles were then reconstituted into micrometer powders by spray drying. Such reconstituted powders were referenced to as powder B. The scanning electron microscope (SEM) micrographs and particle size distributions of the two kinds of powders are shown in Figs. 2 and 3, respectively. The particle size distributions were measured using a laser particle size analyser COULTER-LS130. The major difference between the two powders was that the powder A had a broader particle distribution and poorer flowability than the powder B. 2.2. Plasma spraying Plasma spraying was carried out using a Sulzer-Metco F4-MB plasma gun (Sulzer-Metco, Switzerland) mounted on an ABB robot (ABB, Sweden). Stainless steel plates with the dimensions of 120=50=2 mm3 were used as substrates. Prior to spraying, the stainless steel plates were grit-blasted with alumina abrasive. During spraying, compressed air was applied towards the back of the plates to cool the substrate. Simultaneously, compressed air was also applied to the side from torch, impinging on the coating surface, approximately 15 mm behind the plasma jet, so as to increase the cooling and
Table 1 Spray trials of the process parameters arranged for powder A according to a Table U10(10153) of uniform design experiment Spray trial
Current, x1 (A)
Ar, x2 (lymin)
H 2, x 3 (lymin)
Spray distance, x4 (mm)
ST1 ST2 ST3 ST4 ST5 ST6 ST7 ST8 ST9 ST10
530 540 550 560 570 580 590 600 610 620
35 45 30 40 50 30 40 50 35 45
14 14 12.5 12.5 11 11 9.5 9.5 8 8
100 110 130 90 100 120 130 90 110 120
solidifying rate of molten splats, but not to affect the plasma jet. The coating thicknesses were 0.5–1.2 mm. The spraying trials were arranged according to uniform design experiment, a method of statistical experiment. Uniform design experiment as well as the conventional statistical experiments, such as the Taguchi and orthogonal design methods, have been developed to reduce the experimental trials for experiments involving lots of factors w20,27,32,33x. In this study, the powder feed rate of the powder A was fixed to 11.5 gymin for
Fig. 2. The SEM micrographs of the two powders: (a) powder A; and (b) powder B.
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
104
Table 2 Spray trials of the process parameters arranged for powder B according to a Table U10(10253) of uniform design experiment Spray trial
Current, x1 (A)
Ar, x2 (lymin)
H 2, x 3 (lymin)
Spray distance, x4 (mm)
Powder feed rate, x5 (gymin)
ST1 ST2 ST3 ST4 ST5 ST6 ST7 ST8 ST9 ST10
530 540 550 560 570 580 590 600 610 620
40 35 50 30 50 35 40 45 30 45
9.5 14 8 11 12.5 12.5 8 14 9.5 11
120 100 100 130 110 90 120 130 110 90
7.7 16.0 12.5 13.7 18.4 8.9 17.2 10.1 11.3 14.9
different spray trials; otherwise, the feed rate became unstable. The Table U10(10153) of uniform design experiment w27x was employed to arrange the investigated process parameters of the powder A as listed in Table 1, and the Table U10(10253) was employed to arrange the investigated process parameters of the powder B as listed in Table 2. Except for the powder feed rate, the design scope of the process parameters in Table 1 was the same as those in Table 2. The parameters in Tables 1 and 2 were graded into ten or five levels and incorporated into ten spray trials, the other process parameters were fixed as constants. The deposition efficiency of each spray trial was calculated from the coating mass and the corresponding powder mass fed to the gun. The former was directly measured using a balance with a resolution of 0.01 g. The latter was determined from the powder feed rate and the spraying time on the substrate. 2.3. Metallographic preparation Metallographic cross-sections of the coatings were prepared for the porosity and microhardness measurements. The samples were first cut to the dimensions of 28=10=2 mm3 and then mounted with epoxy resin under vacuum at pressure below 30 mbar following suggested metallographic preparation of plasma sprayed yttria stabilized zirconia coatings w29x. The mounted samples were successively ground with 600, 1200 and 2400 grit SiC papers and eventually polished using diamond slurries of 9 mm, 3 mm and 1 mm for 15 min, 15 min and 10 min, respectively. 2.4. Porosity and microhardness measurements Porosity measurements used an image analysis method. The measurements consisted of the following steps: (a) to observe a 321=246 mm image of the metallographic microstructure using an optical microscope at magnification of =500; (b) to obtain the grey level image of 768=512 pixel using a CCD camera; (c) to enhance the image, to eliminate all interferences and
clearly identify the pores; (d) to produce a binary image from the enhanced image using a suitable threshold to extract the pores from the background; and (e) to calculate the porosity rate according to the Delesse’s stereological protocol, i.e. the area fraction of the pores on the image w34x. The latter three steps were accomplished using the image analysis toolbox of the MATLAB 6.2 (Mathworks Inc.). The threshold to extract the pores from the background was manually selected from grey level of 0–1, by comparing the produced binary image with the original grey level image. As optical microscopy cannot distinguish pores as small as ;1 mm, only big globular pores were measured using the present image analysis method. See Ref. w35x for the simplified concept of globular pores existing in plasma sprayed ceramic deposits. For each coating sample, 10 images were randomly chosen on the metallographic cross-section and were measured. The microhardness measurements were done by indenting the coating cross-sections at a load of 2.942 N for 15 s using a Leitz RZD-DO Vickers hardness tester (Leitz, Germany). For each metallographic crosssection, the measurement series included 20 indentations, and the distance between indentations was kept three times greater than the indentation diagonal to prevent the effects of the stress field of nearby indentations. 2.5. Average grain size estimation X-Ray diffraction was used to identify the phase and estimate the average grain size of the coatings sprayed using the powder B. It was carried out using a Siemens D5000 X-ray diffractometer (Siemens, Germany) with CoKa radiation. All coatings sprayed using the powder B were found predominately to consist of non-transformable tetragonal phase. The average grain size of the coatings was estimated based on XRD peak broadening according to the Scherrer formula w8,11x bs
0.9l dcosu
(1)
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
105
Table 3 Results of stepwise regression analysis that regressed the experimental deposition efficiency data as the third order polynomial equations of the investigated process parameters Source
Degrees of freedom
Mean square
Overall F
Confidence level
Powder A
22.16y7.13=10
y5 2 2 4
Regression Residual
1 8
148.64 3.82
38.86
)0.99
Powder B
68.90y4.76=10y2x3x4q4.58=10y6 x21x5y5.71=10y4x32q1.28=10y3x2x3x4
Regression Residual
4 5
234.44 10.09
23.24
)0.99
Powder
Equation xx
where b is the corrected full width at half-maximum of the diffraction peak at the diffraction angle 2u, l is the wavelength of the radiation and d is the average grain size. The diffraction peak of (111) planes at the 2u approximately 358 was used, and the correction of the full width at half-maximum of the diffraction peak was done by a standard silica specimen to remove the instrument broadening. The contribution from the internal strains of the coatings was neglected because the broadening due to the internal strains should be negligible when compared to that caused by the fine crystallites in the coating samples w11x. 2.6. Microstructure characterization The microstructures of some representative coatings were observed using a JEOL JSM-5800LV (JEOL, Japan) scanning electron microscopy on the polished and fractured cross-sections. The coatings were also removed from the substrate, ground and Arq ion beam thinned for the TEM observation with a JEM200CX (JEM, Japan) transmission electron microscopy. 3. Regression analysis Experimental results like deposition efficiency, porosity, microhardness and average grain size of the coatings obtained can be regressed as the polynomial equations of the investigated process parameters w36x. According to the principles of statistical experiments w20,27,32,33x, the regressed equations would reflect the effects of investigated process parameters on the deposition efficiency, porosity, microhardness and grain size of the coating only if the experimental data were sufficiently accurate and reliable, and could hence be used to optimize the process parameters. The deposition efficiency as well as porosity, microhardness and average grain size were all regressed as the third order polynomial equations of the investigated process parameters using the stepwise regression analysis as detailed in the previous work w32x. Since the uniform design experiment requires much less experimental trials than the conventional statistical experiments, such as Taguchi and orthogonal design methods, for covering an identical design scope of experimental parameters, stepwise regression analysis must be used
to found the regressed equations. During the stepwise regression analysis, the element that affects an objective function most significantly is first selected from the first to third order polynomials of the investigated process parameters, using the F test. More elements are then tested and selected into the regressed equation step by step. In the final regressed equation, only those polynomial elements statistically significant to the objective function are retained. In addition, a residual degree of freedom not less than 4 is usually required to assure that the final regressed equation reflects the dependence of the objective function on the process parameters w27x. It should be noted that these regressed equations were not used to develop models, but mainly to predict the effects of the process parameters on the deposition efficiency, porosity, microhardness and average grain size, as a relatively simple and practical method. The statistical significance of the final regressed equation, i.e. the total retained elements of the process parameters, can be judged with a confidence level determined from the overall F value and the regression and residual degrees of freedom, by using the F test w36x. The closer to unity the confidence level, the more significant is the regressed equation. The prediction precision of the equations is measured by the residual mean square value, e.g. if the process parameters are x1, x2, x3, x4 and x5, there would be a probability of 0.95 that the predicted objective function is among f(x1, x2,«,x5)"2=(residual mean square)1y2. 4. Results 4.1. Deposition efficiency Table 3 lists the results of the stepwise regression analysis for the experimental deposition efficiency data of the two powders. The two regressed equations in Table 3 were both highly significant with the confidence levels higher than 0.99 w36x. The regressed equations listed in Table 3 revealed that the deposition efficiencies of the two powders exhibited different trends dependent on the investigated plasma spray process parameters. The deposition efficiency of the powder A appeared to be significantly related only to the argon flow rate and the spray distance. The lower the argon flow rate and the shorter the spray distance, the larger was the depo-
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
106
Table 4 Comparison of the deposition efficiency results of the uniform design experiment with the values calculated from the regressed equations listed in Table 3 Spray trial
ST1 ST2 ST3 ST4 ST5 ST6 ST7 ST8 ST9 ST10
Powder A
Powder B
Experimental result (%)
Value from equation (%)
Relative error (%)
Experimental result (%)
Value from equation (%)
Relative error (%)
14.4 6.3 17.6 12.4 4.1 12.0 7.4 6.8 9.7 4.7
13.4 6.3 13.8 11.9 4.3 14.5 7.3 6.1 12.6 4.8
y6.4 0.4 y21.5 y3.9 6.7 20.9 y0.2 y9.7 29.9 3.9
48.4 62.2 28.9 59.5 46.5 50.6 61.5 54.9 64.0 56.6
46.9 61.9 29.4 59.9 47.8 55.0 63.8 54.3 61.3 52.9
y3.2 y0.6 1.9 0.7 3.0 8.5 y3.2 y1.2 y4.2 y6.5
sition efficiency. Within the design scope of the investigated process parameters, the predicted largest deposition efficiency was only 16.4% at the argon flow rate of 30 lymin and the spray distance of 90 mm. However, the deposition efficiency of the powder B was significantly related to all the five investigated process parameters. The larger the arc current and the powder feed rate, and the smaller the argon flow rate and the spray distance, the higher was the deposition efficiency. At lower argon flow rate, the deposition efficiency decreased with increasing the hydrogen flow rate. At higher argon flow rate, the deposition efficiency increased with increasing the hydrogen flow rate. Within the design scope of the investigated process parameters, the predicted largest deposition efficiency was 79.2% at the arc current of 620 A, the argon flow rate of 30 ly min, the hydrogen flow rate of 8 lymin, the spray distance of 90 mm and the powder feed rate of 18.4 gy min. Table 4 compares the values calculated from the regressed equations listed in Table 3 with the experimental results of deposition efficiency. It can be seen that the calculated values were in good agreement with the experimental results, with consideration of the predictive precision of the two equations in Table 3.
the comparison of the values calculated from the regressed equations with the experimental data of the two powders. In Tables 6 and 8, the S.D. values were standard deviations of the experimental data, and the xy95 and xq95 were, respectively, the lower and upper bonds of the 95% confidence interval of the means. The four regressed equations in Tables 5 and 7 all had confidence levels higher than 0.99, in other words, significance levels less than 0.01 w36x. From these tables, it can be seen that both the porosity and microhardness of the coatings sprayed using the two powders were affected by the investigated process parameters in different modes. The porosity of the coatings sprayed using the powder A decreased with decreasing both the arc current and the argon flow rate, but was not sensitive to the other investigated process parameters. The predicted lowest porosity was 20.6% at the arc current of 530 A and the argon flow rate of 30 lymin. However, the porosity of the coatings sprayed using the powder B was statistically affected by all five investigated process parameters. The larger the arc current, the hydrogen flow rate and the powder feed rate, and the smaller the argon flow rate and the spray distance, the lower the porosity was. The predicted lowest porosity was y8.8% at the arc current of 620 A, the argon flow rate of 30 lymin, the hydrogen flow rate of 14 lymin and the powder feed rate of 18.4 gymin. A negative porosity is impossible for any materials, thereby it is not possible to extrapolate the porosity values
4.2. Porosity and microhardness Tables 5–8 present the results of the regression analysis for the porosity and microhardness as well as
Table 5 Results of stepwise regression analysis that regressed the experimental porosity data as the third order polynomial equations of the investigated process parameters Powder
Equation y3
x 1 x2
Powder A
y1.12q1.363=10
Powder B
19.65y6.99=10y7x12x2y6.66=10y6x12x3 y2.89=10y6x1x4x5q2.27=10y4x22x4
Overall F
Source
Degrees of freedom
Mean square
Confidence level
Regression Residual
1 8
374.46 5.85
63.99
)0.99
Regression Residual
4 5
460.70 0.54
846.53
)0.99
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
107
Table 6 Comparison of the porosity results of the uniform design experiment with the values calculated from the regressed equations listed in Table 5 Powder
Spray trial
Result of statistical analysis of experimental data Mean (%)
S.D. (%)
xy95 (%)
xq95 (%)
Value calculated from the regressed equation (%)
Relative error (%)
Powder A
ST1 ST2 ST3 ST4 ST5 ST6 ST7 ST8 ST9 ST10
24.7 30.0 21.5 29.5 36.6 20.6 36.6 40.9 27.6 34.7
4.5 7.7 6.8 5.5 7.3 7.1 8.8 10.0 5.5 12.4
21.5 24.5 16.7 25.5 31.4 15.5 30.3 33.8 23.7 25.8
27.9 35.5 26.4 33.4 41.9 25.7 42.9 48.0 31.6 43.6
24.1 32.0 21.4 29.4 37.7 22.6 31.0 39.7 28.0 36.9
y2.3 6.6 y0.9 y0.3 3.0 9.7 y15.3 y2.8 1.1 6.3
Powder B
ST1 ST2 ST3 ST4 ST5 ST6 ST7 ST8 ST9 ST10
36.1 11.2 46.6 13.8 40.9 7.4 30.0 31.1 8.3 18.7
3.2 1.6 2.0 2.0 2.4 1.0 2.9 5.0 1.9 4.1
33.8 10.1 45.2 12.4 39.2 6.7 27.9 26.5 6.9 15.8
38.4 12.4 48.0 15.2 42.6 8.1 32.0 33.7 9.7 21.6
35.7 11.6 46.8 13.0 40.2 7.5 30.8 31.7 8.6 18.1
y1.0 3.4 0.4 y5.4 y1.7 1.4 2.6 2.0 3.5 y3.2
outside the region of the parameters tested in the experiment. Like the porosity, the microhardness of the coatings sprayed using the powder A was also related only to the arc current and the argon flow rate. The smaller the arc current and the argon flow rate, the higher the microhardness was. The predicted highest microhardness was 482 kgymm2 at the arc current of 530 A and the argon flow rate of 30 lymin. The microhardness of the coatings sprayed using the powder B was influenced by the arc current, the argon flow rate as well as the hydrogen flow rate. The larger the arc current and the hydrogen flow rate, and the smaller the argon flow rate, the higher the microhardness was. The predicted highest microhardness was 893 kgymm2 at the arc current of 620 A, the argon flow rate of 30 lymin and the hydrogen flow rate of 14 lymin. Except that the calculated microhardness of the coatings sprayed using the spray trials ST1 and ST2 of the powders A were slightly outside the 95% confidence intervals of the experimental means, all the porosity
values and the other microhardness values calculated from the regressed equations were within the 95% confidence intervals of the experimental means (Tables 6 and 8). 4.3. Average grain size Fig. 4 presents the XRD patterns of some coatings and the powder B used to estimate the average grain size. Tables 9 and 10 are the results of the regression analysis for the average grain size and the comparison of the values calculated from the regressed equation with experimental data. The regressed equation also had a confidence level of 0.99, and the values calculated from the regressed equation appeared to be consistent with the experimental results. According to the regressed equation, the average grain size of the coatings appeared to be affected by the arc current, the argon flow rate and the spray distance, but was not sensitive to the hydrogen flow rate and the powder feed rate. The larger the arc current and the
Table 7 Results of stepwise regression analysis that regressed the experimental microhardness data as the third order polynomial equations of the investigated process parameters Powder
Equation
Source y2
x 1x 2
Powder A
833.58y2.21=10
Powder B
494.38q9.67=10y5x21x3y4.51=10y3x32
Degrees of freedom
Mean square 4
Overall F
Confidence level
Regression Residual
1 8
9.84=10 1.55=103
63.54
)0.99
Regression Residual
2 7
1.49=105 1.17=103
127.04
)0.99
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
108
Table 8 Comparison of the microhardness results of the uniform design experiment with the values calculated from the regressed equations listed in Table 7 Powder
Spray trial
Result of statistical analysis of experimental data Mean (kgymm2)
S.D. (kgymm2)
xy95 (kgymm2)
xq95 (kgymm2)
Value calculated from the regressed equation (kgymm2)
Relative error (%)
Powder A
ST1 ST2 ST3 ST4 ST5 ST6 ST7 ST8 ST9 ST10
364 228 530 397 212 422 308 156 340 228
91 51 95 44 26 95 67 28 78 47
322 204 485 376 200 378 277 143 303 206
407 252 574 418 224 467 308 169 376 250
424 297 469 339 204 449 312 171 362 217
16.4 3.1 y11.4 y14.7 y3.8 6.4 1.5 9.9 6.6 y4.7
Powder B
ST1 ST2 ST3 ST4 ST5 ST6 ST7 ST8 ST9 ST10
467 685 175 684 332 678 456 549 778 533
127 111 28 84 64 86 87 114 86 123
404 630 162 643 300 635 413 492 735 472
530 740 189 726 364 720 500 605 821 594
468 683 178 714 323 701 482 575 717 494
3.1 y2.9 1.4 4.4 y2.6 3.4 5.8 4.9 y7.9 y7.2
4.4. Process optimization
Fig. 4. The XRD patterns of some coatings and powder B used to estimate the average grain size.
smaller the argon flow rate and the spray distance, the smaller the average grain size was. The predicted smallest average grain size was 39.5 nm at the arc current of 620 A, the argon flow rate of 30 lymin and the spray distance of 90 mm.
The powder A had too low deposition efficiency, the predicted largest deposition efficiency was 16.4%. Therefore, only the optimized process parameters of the powder B according to the regressed equations listed in Tables 3, 5, 7 and 9 as aforementioned were experimentally tested. These optimized process parameters are further summarized in Table 11. The spray trial AO1 was for the largest deposition efficiency, and the spray trial AO2 was for the smallest porosity and the largest microhardness. Considering the lifetime of the plasma gun, a hydrogen flow rate of 12.5 lymin, slightly smaller than the largest 14 lymin within the design scope of the investigated hydrogen flow rate, was chosen for the optimization of the porosity and microhardness. As the average grain size appeared to be insensitive to the hydrogen flow rate, both the two spray trials were correspondent to the predicted smallest average grain size.
Table 9 Results of stepwise regression analysis that regressed the experimental grain size data of the coatings sprayed using powder B as the third order polynomial equation of the investigated process parameters Equation 49.7y6.32=10
y8 3 1
x q2.01=10
y5
2 2 4
xx
Source
Degrees of freedom
Mean square
Overall F
Confidence level
Regression Residual
2 7
59.82 6.29
9.50
0.99
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
109
Table 10 Comparison of the grain size results of the uniform design experiment with the values calculated from the regressed equation listed in Table 9 Spray trial
Experimental result (nm)
Value from the regressed equation (nm)
Relative error (%)
ST1 ST2 ST3 ST4 ST5 ST6 ST7 ST8 ST9 ST10
52.7 45.3 47.8 46.4 51.8 47.0 49.3 51.5 44.1 38.5
51.8 47.0 49.2 48.7 50.2 43.1 48.3 51.4 42.6 42.0
y1.6 3.7 2.9 5.1 y3.1 y8.2 y2.1 y0.1 y3.4 9.1
Table 11 The spray trials used to asses the validity of the regressed equations and optimize the deposition efficiency, porosity, microhardness and average grain size of the coatings sprayed using the powder B Spray trial
Arc current (A)
Ar (lymin)
H2 (lymin)
Spray distance (mm)
Powder feed rate (gymin)
Optimization objective
AO1 AO2
620 620
30 30
8 12.5
90 90
18.4 18.4
Deposition efficiency, grain size Porosity, microhardness, grain size
Table 12 Comparison of the experimental results of the spray trials detailed in Table 11 with the values calculated from the regressed equations listed in Tables 3, 5, 7 and 9 Spray trial
AO1
AO2
*
Parameters*
Result of statistical analysis of experimental data Mean
S.D.
xy95
xq95
Deposition efficiency Porosity Microhardness Average grain size
74.1 6.3 706 35.3
–
–
–
1.4 119 –
5.3 647 –
7.3 765 –
Deposition efficiency Porosity Microhardness Average grain size
68.2 3.8 953 48.0
–
–
–
1.0 96 –
3.1 905 –
4.5 1001 –
Value from the equations
Relative error (%)
79.2 6.7 670 39.5
6.9 6.3 y5.1 11.9
75.4 y4.8 837 39.5
10.6 y226.3 y12.2 y17.7
The units of the parameters: refer to Tables 4, 6, 8 and 10.
Table 12 compares the optimized deposition efficiency, porosity, microhardness and average grain size predicted from the regressed equations listed in Tables 3, 5, 7 and 9 with the experimental results. Apparent deviation existed for the unreasonable negative porosity and the average grain size. The average grain size of the coating AO1 was clearly smaller than that of the coating AO2. However, the other values predicted from the regressed equations agreed very well with the experimental results if only the predictive precision of these equations was taken into account. 4.5. Coating microstructure Fig. 5 presents the SEM micrographs of the polished cross-sections of the coatings sprayed using the spray
trials AO1 and AO2. There were more pores in the coating AO1 than in the coating AO2. Fig. 6 shows the SEM micrographs of the fractured cross-sections of the two coatings. These micrographs, however, revealed that the coatings still exhibited the typical lamellar structure w37x and were mainly composed of flat plate-like lamellae as well as some partly molten or equiaxed grain particles. The lamellae were typically 1–3 mm thick with columnar grains extending through their thickness. There also appeared more partly molten or equiaxed grain particles in coating AO1 than in coating AO2. TEM observation found that except for a very few of equiaxed grains with size of 100–200 nm, the two coatings were mainly comprised of the equiaxed grains smaller than 100 nm (Fig. 7a,c) with smaller amounts of columnar grains from several hundred nanometers to
110
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
Fig. 5. The SEM micrographs of the polished cross-sections of the two coatings sprayed using spray trials: (a), (b) AO1; and (c), (d) AO2.
micron in the columnar direction but smaller than 100 nm in the cross-section (Fig. 7b,d). Considered the TEM sampling direction, most of the equiaxed grains should be in the direction of the cross-sections of the columnar grains. Thus it can be concluded that the columnar grains in the coatings were nanostructural in their crosssection.
5. Discussion 5.1. Powder A The smaller the argon flow rate, the higher were the deposition efficiency and the microhardness, and the lower was the porosity for the powder A. It seems
Fig. 6. The SEM micrographs of the fractured cross-sections of the two coatings sprayed using spray trials: (a), (b) AO1; and (c), (d) AO2.
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
111
Fig. 7. The TEM micrographs of the typical grain morphologies in the two coatings sprayed using spray trials: (a), (b) AO1; and (c), (d) AO2.
related to a better melting of the powder particles. This is because a lower ratio of argon to hydrogen in the plasma gases corresponds to a higher enthalpy and thermal conductivity of the plasma jet w38x. It has been confirmed that a higher deposition efficiency and smaller coating porosity was associated with a better melting state of conventional partially yttria stabilized zirconia powders w24,26x. The deposition efficiency was insensitive to the arc current and hydrogen flow rate, the porosity increased and the microhardness decreased with the increase in the arc current. That is probably due to the complex interdependence of the process parameters for this powder w20x. For example, for such a widely distributed and
nanostructured powder, a higher arc current would be useful to a better melting of larger particles, but might also lead to an evaporation of smaller particles. The complex interdependence may shield and disturb some direct relationships between the process parameters and the deposition efficiency, porosity and microhardness. Regardless of the validity of the regressed equations, it was also clear that the powder A had much lower deposition efficiency (Table 4) than the conventional powder. For the general sets of plasma parameters, the coatings sprayed using this type of reprocessed nanostructured powder had obviously larger porosity (Table 6) and somewhat lower microhardness (Table 8) than the coatings sprayed using the conventional powders
112
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
w23,24x. For the conventional micrometer-sized powders, narrow particle size distribution and good flowability are generally considered as the excellent powder performance for thermal spraying. This seems also the cases for the nanostructured powders used for thermal spraying as shown by the poor performance of the current powder A. 5.2. Powder B 5.2.1. Deposition efficiency Within the design scope of the process parameters, the efficiency of the powder B decreased with increasing the hydrogen flow rate at a lower argon flow rate, then increased with increasing the hydrogen flow rate at a higher argon flow rate (Table 3). It is suggested that a plasma jet with moderate energy intensity should be necessary for the highest deposition efficiency of the powder B w28,38x. This may be reasonably ascribed to the facts that the powder B had a small particle size and density and the nanostructured particles were relatively easy to be molten in the plasma jet. The largest deposition efficiency of the powder B was 74.1%. Such a high deposition efficiency is unusual for plasma spraying of partially yttria stabilized zirconia powders as reported by a large body of literature w24,28,29x. The deposition efficiency data collected by Kucuk et al. w24x from the existing literature for conventional yttria stabilized zirconia powders was in the range of 20–65%. 5.2.2. Porosity The negative lowest porosity value predicted from the regressed equation may be attributed to the insufficiently accurate porosity data for regression analysis. This was evident from the relatively wide confidence intervals of the porosity means of some coatings (Table 6), which resulted from the heterogeneous distribution of pores within these coatings. However, the coating sprayed using the optimized process parameters was significantly denser than the coatings sprayed using the spray trials listed in Table 2 (Fig. 5, Tables 6 and 7), it can be hence expected that the regressed equation of the porosity could soundly predict the relative trend of porosity size of the different process parameters although not providing an absolute value of the porosity. From the equation in Table 5, it can be seen that a lower porosity corresponded to a plasma jet with higher energy intensity and a better melting of the powder particles. Such relationships between the porosity of the coating and energy intensity of plasma jet and the melting state of the powder particles is the same as those of the coatings sprayed using the conventional powders w20,28,29,38x. As aforementioned, however, the highest deposition efficiency was consistent with a plasma jet with moderate energy intensity. This can be
Fig. 8. Plot of porosity vs. deposition efficiency for the powder B (R value is the coefficient of correlation of the data fitting, and the error bars represent the 95% confidence intervals of the porosity means).
ascribed to the two following aspects: (i) the very small powder particles were difficult to penetrate into a plasma jet with higher dynamic energy w21,22,25x; and (ii) the very small powder particles were evaporated during flight in a plasma jet with higher energy intensity. For most spray trials listed in Table 2, the coatings sprayed using the powder B had larger porosity values (Table 6) than those sprayed using the conventional powder w23,24x. But for optimized process parameters, the coating sprayed using the powder B was denser and had smaller porosity (Table 12) than that sprayed using the conventional powder. Unlike the conventional powders for which the porosity correlates with the deposition efficiency w29x, no significant linear relationship was observed between the porosity and deposition efficiency of the powder B as shown in Fig. 8. Such a result may be due to some bad measurement data of porosity. It was noticed that the porosity of plasma sprayed partially yttria stabilized zirconia coatings measured using image analysis method is easily disturbed by the metallographic preparation w29x. However, since the effects of the process parameters on the porosity of the present coatings are the same as those of the coatings sprayed using the conventional powders w20,28,29,38x, the current result of the relationship between the porosity and deposition efficiency should be justifiable. This can be also confirmed by the experimental data of porosity and deposition efficiency of the coatings AO1 and AO2. The coating AO1 had both higher deposition efficiency and larger porosity than the coating AO2. Therefore, this type of nanostructured powder is attractive because it could provide the advantage to deposit porous coatings with a higher deposition efficiency. 5.2.3. Microhardness The regressed equation of the microhardness showed that a higher microhardness was also associated with a
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
113
microhardness increased with decreasing the argon flow rate, similar trend was also visible form the result of Vickers microhardness in Ref. w10x, if the confidence intervals of measured data are taken into consideration, although it was stated that the microhardness increased with increasing the argon flow rate. A distinguished difference is the effect of the spray distance on the microhardness. In the present work, the spray distance did not significantly affect the microhardness (Table 7). In Ref. w10x, it was observed that the microhardness increased with the decrease in spray distance.
Fig. 9. Plot of microhardness vs. porosity for powder B (R value is the coefficient of correlation of the data fitting, and the error bars represent the 95% confidence intervals of the means).
plasma jet with a higher energy density. Despite the fact that the microhardness was insensitive to the spray distance and the powder feed rate (Table 7), an significant linear relationship between the microhardness and the porosity was observed (Fig. 9). On the one hand, this signified that the effects of the spray distance and powder feed rate on the porosity might be relatively weak. On the other hand, this revealed that the micrometer-sized defects such as the pores and microcracks were still the factors predominately affecting the microhardness of the coatings. However, provided the identical porosity values, the coatings sprayed using the powder B had a higher microhardness than those sprayed using the conventional powder w28x. XRD results showed that the coatings sprayed using both powders were predominately composed of non-transferable tetragonal phase. Thus it was evident that the nanostructured powder improved the microhardness of the coatings. This could be attributed to the decreased grain size andyor other porous components, intraplat cracks and intersplat lamellar pores w35x that were not characterized in the present work, in the coatings. The microhardness 953 kgymm2 of the optimized coating AO2 was also higher than the limited microhardness data of plasma sprayed partially yttria stabilized zirconia coatings using the conventional powders reported by several other authors w28,39x. Because of different plasma torches, process parameters, nanostructured powder feedstocks and test conditions of microhardness, it is difficult to compare the absolute values of the present microhardness with those of the previous plasma spray deposits of nanostructured partially yttria stabilized zirconia reported in Ref. w10x. The effect of the arc current on the microhardness of the present coatings is in agreement with that in Ref. w10x: the microhardness increased with the increase in the arc current. The present result indicated that the
5.2.4. Average grain size According to the regressed equation, a smaller average grain size was associated with a plasma jet with higher energy intensity and independent of the hydrogen flow rate. This was surprising because a plasma jet with higher energy intensity would more probably destroy the nanostructured feature of the starting powder. The experimental result of the coatings AO1 and AO2 showed that a smaller average grain size corresponded to a smaller hydrogen flow rate (Tables 11 and 12 and Fig. 7). These indicated that the average grain size data determined from the XRD Sherrer formula were not accurate enough for the regression analysis. In fact, some researchers have pointed out that the XRD technique is inadequate to identify the nanostructured features of different coatings w11,13x. To our knowledge, there is still no other technique that can determine the grain size of the different coatings quantitatively and reliably related to the process parameters. As it was assumed that the XRD Sherrer formula could provide an approximate estimation of the average grain size, the combination of the XRD, SEM and TEM results (Tables 10 and 12, Figs. 6 and 7) revealed that the columnar grains of the coatings were nanostructural (smaller than 100 nm) in their cross-section although they were 1–3 mm in the columnar direction. The grain size of the plasma sprayed yttria stabilized zirconia coatings using the conventional powder was in the range of 0.1–0.4 mm in cross-section of the columnar grains w37x and 30–1000 nm of the overall grain size w11x. Due to the considerable scatter of the data, it is difficult to compare them with those of the current coatings. The microstructure of the coatings AO1 and AO2 also showed that a plasma jet with lower energy intensity corresponded to more partly molten particles or fine equiaxed grains within the coating. This was interesting and will be further investigated because incorporating partly molten particles or fine equiaxed grains within the coatings is the general thinking to prepare the nanostructured coating using thermal spraying w8,11,17x. 6. Conclusion Nanosized partially yttria stabilized zirconia particles were synthesized using a co-precipitation method,
114
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
agglomerates were reprocessed to be micrometer-sized either by controlling the grinding of the calcined yttria– zirconia agglomerates or by reconstituting the ground particles using spray drying. The effects of the plasma spray process parameters including the arc current, argon flow rate, hydrogen flow rate, spray distance and powder feed rate on the deposition efficiency, porosity, microhardness and average grain size of the coatings made from these two powders were investigated using a uniform design experiment and subsequent statistical analysis. The powder directly kept within micrometer-sized agglomerates by controlling the grinding of the calcined yttria–zirconia agglomerates had much lower deposition efficiency than the conventional powders. For the general sets of plasma parameters, the coatings sprayed using this powder had significantly larger porosities and lower microhardness. The regressed equations of the deposition efficiency, porosity and microhardness related to the investigated process parameters for this reprocessed nanostructured powder were not straightforward. For the powder reconstituted into micrometer-sized agglomerates using the spray drying, the regressed equations of the deposition efficiency, porosity and microhardness appeared to be more realistic, or at least could soundly predict the relative trend with the process parameters. The experimental results showed that the average grain size data determined from the XRD Scherrer formula were not accurate enough for the regression analysis. Such reconstituted nanostructured powder had significantly higher deposition efficiency than the conventional powders. Unlike the conventional powders for which the porosity is correlated with the deposition efficiency, no significant linear relationship existed between the porosity and deposition efficiency for this reconstituted nanostructured powder. It could hence deposit both porous and dense coatings with a relatively higher deposition efficiency. The micrometer-sized defects such as the pores and microcracks were still the factors predominately affecting the microhardness of the coatings. It was also observed that the nanostructured powder improved the microhardness of the coatings. A coating with a porosity of 3.8% and Hv0.3 of 953 kgymm2 was obtained. The coatings sprayed using this powder were mainly composed of the columnar grains with some partly molten or equiaxed grain particles. The columnar grains within coatings were roughly 1–3 mm in the columnar direction and smaller than 100 nm in their cross-sections. Acknowledgments The authors are grateful to Prof. K.T. Fang and Prof. C.Q. Ma for supplying the uniform design table U10(10253), to Miss L. Lahoupe for her help in the
preparation of the metallographic samples and to the Advanced Research Program of Franco-Chinese Cooperation (PRA MX02–03, C. COODDET & C. DING). References w1x R. McPherson, Thin Solid Films 83 (1981) 297. w2x L. Pawlowski, The Science and Engineering of Thermal Spray Coating, John Wiley and Sons Ltd, England, 1995. w3x M.L. Lau, H.G. Jiang, E.J. Lavernia, Synthesis and characterization of nanocrystalline 316-stainless steel coatings by high velocity oxy-fuel spraying, in: C. Coddet (Ed.), Thermal Spray: Meeting the Challenges of the 21st Century, ASM International, Materials Park, OH, 1998, p. 379. w4x H.G. Jiang, M.L. Lau, E.J. Lavernia, Nanostruct. Mater. 10 (2) (1998) 169. w5x M.L. Lau, E. Strock, C.J. Lavernia, E.J. Lavernia, Nanostruct. Mater. 10 (5) (1998) 723. w6x D.A. Stewart, P.H. Shipway, D.G. McCartney, Wear 229 (1999) 789. w7x B.H. Kear, R.K. Sadangi, M. Jain, R. Yao, Z. Kalman, G. Skandan, W.E. Mayo, J. Therm. Spray Technol. 9 (3) (2000) 399. w8x L.L. Shaw, D. Goberman, R. Ren, M. Gell, S. Jiang, Y. Wang, T.D. Xiao, P.R. Strutt, Surf. Coat. Technol. 130 (2000) 1. w9x Y. Wang, S. Jiang, M. Wang, S. Wang, T.D. Xiao, P.R. Strutt, Wear 237 (2000) 176. w10x R.S. Lima, A. Kucuk, C.C. Berndt, Surf. Coat. Technol. 135 (2001) 166. w11x R.S. Lima, A. Kucuk, C.C. Berndt, Mater. Sci. Eng. A 313 (2001) 75. w12x J. Karthikeyan, C.C. Berndt, J. Tikkanen, S. Reddy, H. Herman, Mater. Sci. Eng. A 238 (1997) 275. w13x J. Karthikeyan, C.C. Berndt, S. Reddy, J.Y. Wang, A.H. King, H. Herman, J. Am. Ceram. Soc. 81 (1) (1998) 121. w14x N.P. Padture, K.W. Schlichting, T. Bhatia, A. Ozturk, B. Cetegen, E.H. Jordan, M. Gell, S. Jiang, T.D. Xiao, P.R. Strutt, E. GarCia, P. Miranzo, M.I. Osendi, Acta Mater. 49 (2001) 2251. w15x M. Gell, Nanostruct. Mater. 6 (1995) 997. w16x E.J. Lavernia, L. Kabacoff, J. Therm. Spray Technol. 7 (3) (1998) 291. w17x I. Ahmed, T.L. Bergman, J. Therm. Spray Technol. 9 (2) (2000) 215. w18x S.L. Girshick, High Temp. Mater. Processes 4 (2000) 379. w19x B.H. Kear, P.R. Strutt, Nanostruct. Mater. 6 (1995) 227. w20x R. Kingswell, K.T. Scott, L.L. Wassell, J. Therm. Spray Technol. 2 (2) (1993) 179. w21x P. Fauchais, M. Vardelle, A. Vardelle, L. Bianchi, A.C. Leger, Plasma Chem. Plasma Processing 16 (1) (1996) 99. w22x L. Bianchi, A.C. Leger, M. Vardelle, A. Vardelle, P. Fauchais, Thin Solid Films 305 (1997) 35. w23x M. Prystay, P. Gougeon, C. Moreau, J. Therm. Spray Technol. 10 (1) (2001) 67. w24x M. Friis, C. Persson, J. Wigren, Surf. Coat. Technol. 141 (1– 2) (2001) 115. w25x A. Kucuk, R.S. Lima, C.C. Berndt, J. Am. Ceram. Soc. 84 (4) (2001) 685. w26x A. Kucuk, R.S. Lima, C.C. Berndt, J. Am. Ceram. Soc. 84 (4) (2001) 693. w27x K.T. Fang, Y. Wang, Number-Theoretic Methods in Statistics, Chapman and Hall, London, 1993. w28x E. Lugscheider, F. Ladru, V. Gourlaouen, C. Gualco, Enhanced atmospheric plasma spraying of thick TBCs by improved process control and deposition efficiency, in: C. Coddet (Ed.),
J.F. Li et al. / Thin Solid Films 460 (2004) 101–115
w29x
w30x w31x w32x w33x
Thermal Spray: Meeting the Challenges of the 21st Century, ASM International, Materials Park, OH, 1998, p. 1583. J. Wigren, L. Pejryd, Thermal barrier coatings—why, how, where and where to, in: C. Coddet (Ed.), Thermal Spray: Meeting the Challenges of the 21st Century, ASM International, Materials Park, OH, 1998, p. 1531. K.C. Chang, W.J. Wei, C. Chen, Surf. Coat. Technol. 102 (3) (1998) 197. L. Zhang, J. Mou, Nanomaterials Science (in Chinese), Liaoning Press of Science and Technology, Shunyang, China, 1994. J.F. Li, J.Q. Huang, C.X. Ding, J. Chinese Ceram. Soc. 29 (1) (2001) 49. C.J. Li, G.C. Ji, Y.Y. Wang, Effect of powder type on the relationship between spray parameters and properties of HVOF
w34x w35x w36x w37x w38x w39x
115
sprayed Cr3C2 –NiCr coatings, in: C. Coddet (Ed.), Thermal Spray: Meeting the Challenges of the 21st Century, ASM International, Materials Park, OH, 1998, p. 278. G. Montavon, C. Coddet, C.C. Berndt, S.-H. Leigh, J. Therm. Spray Technol. 7 (2) (1998) 229. A.J. Allen, J. Ilavsky, G.G. Long, J.S. Wallace, C.C. Berndt, H. Herman, Acta Mater. 49 (9) (2001) 1661. N.R. Draper, H. Smith, Applied Regression Analysis, John Wiley and Sons, New York, 1981. G.N. Heintze, R. McPerson, Surf. Coat. Technol. 34 (1988) 15. B. Pateyron, M.-F. Elchinger, G. Delluc, P. Fauchais, Plasma Chem. Plasma Processing 12 (4) (1992) 421. K.M. Jasim, R.D. Rawlings, D.R.F. West, Surf. Coat. Technol. 53 (1992) 75.
