Applied Surface Science 239 (2004) 25–35
Remote nitrogen plasma treatment of a polyethylene powder Optimisation of the process by composite experimental designs Brigitte Mutela,*, Muriel Biganb, Herve´ Vezinb a
Laboratoire de Ge´nie des Proce´de´s d’Interactions Fluides re´actifs-Mate´riaux (UPRES-EA 2698), Universite´ des Sciences et Technologies de Lille, 59655 Villeneuve d’Ascq Cedex, France b Laboratoire de Chimie Organique et Macromole´culaire (UMR 8009), Universite´ des Sciences et Technologies de Lille, 59655 Villeneuve d’Ascq Cedex, France Received 29 January 2004; accepted 17 April 2004 Available online 21 August 2004
Abstract A coupling between fluidised bed and far cold remote nitrogen plasma technologies is used to treat a polyethylene powder in order to increase its wettability evaluated from capillarity rising measurements. An optimisation of the process is performed thanks to face-centred central composite experimental designs. It is shown that treatment duration and oxygen rate added to nitrogen are the more influent parameters. Electron spin resonance analyses revealed the presence of a mixture of alkyl and allyl radicals on the treated powder. Reactions leading to the incorporations of new chemical functions on the polyethylene surface are proposed. # 2004 Elsevier B.V. All rights reserved. PACS: 52.75.Rx; 81.20.Fw; 81.65.-b; 82.65.Yh Keywords: Cold plasma; Polyethylene powder; Wettability; Capillarity rising; Electron spin resonance; Surface modifications
1. Introduction Taking into account their low surface energy, polymers often require a modification of their surface properties before any use. Cold plasma treatments are a convenient way to obtain these modifications by introducing new surface chemical groups without affecting the bulk properties. In particular, the efficiency of the far cold remote nitrogen plasma technology to modify polymers surfaces has been clearly evidenced [1–4]. This medium, located far from the excitation zone and *
Corresponding author. Tel.: þ33 3 20 436540; fax: þ33 3 20 434158. E-mail address:
[email protected] (B. Mutel).
exempt of ions and electrons is characterised by a gas temperature about the ambient [5]. However, a good contact between the polymer surface and the plasma is an important factor to operate efficiently the surface modifications. So, conventional plasma reactor cannot be used for powder materials because it is necessary to treat the whole surface and aggregation of the powder has to be eliminated. To reach this aim, the fluidisation technique can be used: fluidised bed reactors, circulating or not, are characterised by a good mixing of particles and high rate of transfer phenomena. However, by now, only very few works deal about the treatment or the coating of powders by fluidised bed plasma process and all of them involve discharge plasma [6–15]. Among them, only two papers deal about the treatment of
0169-4332/$ – see front matter # 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2004.04.021
26
B. Mutel et al. / Applied Surface Science 239 (2004) 25–35
polymer powders [14,15]. Inagaki et al. [14] showed that by using a discharge oxygen plasma treatment, it was possible to increase the wettability of a polyethylene (PE) powder. The contact angle with water, higher than 908 for the untreated PE, was equal to 518 after a treatment during 6 h. By X-ray photoelectron spectroscopy (XPS), Park and Kim [15] showed that the surface of untreated PE powder contained a low oxygen ratio (O/C 0.03). After a treatment in an oxygen plasma discharge for 3 h at 150 W discharge power, the O/C ratio increased from 0.03 to 0.30. In the same conditions, Inagaki et al. [14] observed an O/C increase from 0.07 to 0.15, which was due to the incorporation of C¼O and C(O)–O– groups at the outermost layer of the powder. The originality of the process described in the present paper is the use of the far cold remote nitrogen plasma both to fluidise and to treat the powder. In previous works [16,17], it was clearly demonstrated that the treatment increases the powder wettability; and the increase is more important when small amounts of oxygen are added to the nitrogen plasma gas. In such a process, involving a great number of operating parameters, trying to achieve the maximum wettability value can be an exhausting task by changing one parameter at a time. Moreover, only an apparent optimum is generally obtained and interac-
tions between the involved parameters are ignored. To overcome this problem, the experimental design technique can be used. Based on simple statistics, it allows reducing the number of experiments to be performed and to know interactions among variables in the studied range. In the present work, a composite factorial design of experiments was chosen in order to determine plasma treatment conditions leading to the best wettability. This method leads to a mathematical model that allows obtaining response surfaces versus the influent parameters. Electron spin resonance (ESR) investigated modifications induced on the polyethylene powder surface treated in these conditions. Reactions leading to the incorporations of new chemical functions on the polyethylene surface are proposed.
2. Experimental details and methods 2.1. Coupling between the far cold remote nitrogen plasma and the fluidised bed technologies The experimental set-up is shown in Fig. 1. The nitrogen flow created by a continuous pumping (33 Nm3/h), was excited by an electrodeless discharge
Fig. 1. Experimental set-up.
B. Mutel et al. / Applied Surface Science 239 (2004) 25–35
by means of a microwave generator (2450 MHz) that could deliver an incident power up to 1.5 kW. The discharge was produced in a quartz tube (32 mm diameter) connected to the fluidised bed reactor through a 908 elbow. This reactor was constituted of a vertical cylindrical Pyrex glass tube (height: h ¼ 1 m; inner diameter: D ¼ 0.15 m) with a porous plate at the bottom to support the powder bed. The distance (d) between the discharge and the porous plate was 0.65 m. The plasma surrounding this plate is a far cold remote nitrogen plasma characterised by a yellow afterglow. The main reactive species are atomic nitrogen in the ground state N(4S), and vibrationally and/or electronically excited nitrogen molecules such as N2ðX1 Sg þ Þv and N2ðA3 Su þ Þ. The long lifetime of nitrogen atoms (10 s) due to a redissociation of nitrogen molecules [18] allows an expansion of the far cold remote nitrogen plasma in the whole fluidised bed reactor. It is interesting to note that the addition of oxygen (with percentage lower than 5%) to the nitrogen flow allows producing O(3 P) atoms without destroying the plasma volume expansion. Of course, the porous plate has to be made with a material that does not destroy the reactive species of the plasma. In this work, it was a polyethylene porous plate (pore diameter 17 mm). Nitrogen and oxygen (Industrial quality, U grade, Air Liquide) flow rates (QN 2 and QO 2 , respectively) were adjusted by mass flow regulators. The pressure was measured at the top of the column using a Pirani type vacuum gauge. The powder to treat was PE (GUR X 132 provided by Hostalen). The density or specific gravity (rp ¼ 960 kg/m3), the BET surface area (Ss ¼ 0.5 m2/g) and the granulometry (280 mm) of the powder were respectively measured with a helium gas pycnometer (Micrometrics, AccuPyc 1330), by nitrogen gas adsorption (Micrometrics, Asap 2000) and with a light-scattering laser instrument (BeckmanCoulter LS230), respectively. A given mass of the PE powder was filled in the reactor and then the nitrogen (or nitrogen þ oxygen) flow was injected. For low flow rates, the PE powder did not move. When the flow increased, the powder started to move and when the pressure counterbalanced the weight of the bed, the fluidisation was initiated. In this work, the fluidisation of the PE powder appears with a gaseous flow equal to
27
0.35 N l/min. Prior to the treatment, the powder was always fluidised with such a nitrogen flow during 5 min. 2.2. Wettability measurements The wettability of the PE powder was determined with a tensiometer (Kru¨ ss K12) from dynamic capillarity rising measurements according to the Washburn method [19]. A 5 mm inside diameter aluminium tube, bottom closed with a porous paper, was filled with 1 g of powder. The bottom of the tube was then put in close contact with a chosen liquid which penetrates up into the PE powder column by capillarity rising. The weight m of the penetrating liquid was measured as a function of time t after contact of the bottom tube with the liquid. The capillarity rising is given by the Washburn equation: m2 ðKrL2 gL cos yL Þ ¼ 2ZL t
(1)
where rL, ZL and gL are the specific gravity, the viscosity and the surface energy of the liquid, respectively. yL is the contact angle between the liquid and the powder. The empirical constant K, depends on the particles size and on their degree of packing. It could be experimentally determined using a liquid that perfectly wetted the powder, for which the contact angle against the powder is assumed to be 08 (cos yL ¼ 1). In this work, K was determined with heptane. Surface energy of liquids was always checked by tensiometric measurements (with Kru¨ ss tensiometer), according to du Nouy method. 2.3. ESR study ESR measurements were made at 100 K in a Brucker ELEXYS 580E spectrometer operating at 9 GHz microwave frequency. Spectra were recorded at non-saturating microwave power (3.1 mW) and with a frequency and a modulation of the magnetic field respectively equal to 100 kHz and 10 G. Each ESR spectrum consists in three accumulations and corresponds to the derivative of the absorption spectrum. The absolute magnitude of the free radicals population of the treated powder was determined by comparison of the area under the ESR absorption
28
B. Mutel et al. / Applied Surface Science 239 (2004) 25–35
curve with the one of a reference material, a pitch in KCl standard. The ESR spectra were simulated using Winsim 20001, the hyperfine splitting constants and the calculation of the radicals ratio were performed using automatic simplex algorithm implemented in the software. After the plasma treatment, the PE powder was exposed to the ambient air for a duration denoted by d (ageing time), then, dipped in liquid nitrogen in order to stop the ageing. To operate at d 0, the powder was removed from the reactor under nitrogen atmosphere and was immediately dipped in liquid nitrogen. 2.4. Experimental design In order to determine the influential factors as well as their interactions, a face-centered central composite design was chosen [20–26]. This experimental design includes: the full factorial design 2k or the fractional factorial design 2k1 for a number of factors k 5. The fractional factorial design allows reducing the number of experiments. Therefore, 2k or 2k1 experiments are required to cover all possible combinations of factor levels; The low and high levels are codified Xi ¼ 1 and Xi ¼ þ1, respectively. axial points (or ‘‘star points’’) placed on the axe of each factor in order to encircle the experimental domain. They answer to a particular criterion of optimisation: the error on answer forecast is the same for all the points of a sphere (or hyper sphere) centred at the origin of the experimental domain. It is the criterion of rotability. For a design without replicas, the distance a between the axial points and the centre of the domain is given by: a ¼ ½nf 0:25 where nf is the number of summits of the studied domain. Here nf was equal to 2k or 2k1. the central point, repeated and conducted for estimation of the experimental error. The relation between the coded and the original scales is given as follows: Ui ¼ Ui 0 þ DUi Xi
with Ui, the original variable, Ui0, the midpoint of original interval, Xi, the coded variable and DUi, the interval of origin range. The central composite experimental design is represented by a mathematical model obtained by multiple regressions and fitted with a second-order polynomial function according to the following form: y ¼ bo þ
X X X b i Xi þ bii Xi 2 þ bij Xi Xj þ e i
i
i;j
y is the matrix of the answers. Coefficient of the effects relating to each factor are determined by matrix algebra according to the relation: b ¼ ðX t XÞ1 X t y X is the experiment matrix in coded variables; Xt is the transposed experiment matrix and (XtX)1 is the reverse of the matrix product of Xt by X, bi is the coefficient of the effect of factor i, bij is the coefficient of interaction effect between factors i and j, bi and bij were calculated using software Modde 4 [17].
3. Results 3.1. Process optimisation A first design of experiments (denoted design no. 1) was carried out to determine the influence of five factors on the treatment efficiency: the gaseous flow (U1), the oxygen rate (U2), the treatment time (U3), the transmitted microwave power (U4) and the mass of powder (U5). Table 1 lists these five factors, their abbreviations and their levels. The response y of each experiment was evaluated from capillarity rising of water and was quantified from m2/t. This parameter was preferred to the contact angle one as it allows to follow the wettability of the powder with a given liquid in a continuous way. Indeed, Washburm method does not allow to evaluate yL values higher than 908, and, as soon as the surface energy of the powder is higher than the liquid one, the yL value is always equal to 08. The experimental design and the responses are gathered Table 2.
B. Mutel et al. / Applied Surface Science 239 (2004) 25–35
29
Table 1 Experimental domain and coding of variables for experimental design no. 1 Variables
Factors
X1 X2 X3 X4 X5
U1, U2, U3, U4, U5,
Levels
gaseous flow (NL/min) oxygen rate (%) treatment time (min) transmitted power (W) mass of powder (g)
1
þ1
a
þa
2.20 0.19 10 400 20
5.70 0.57 25 800 35
0.45 0 2.5 200 12.5
7.45 0.76 32.5 1000 42.5
The central composite experimental design was established in three parts: a fractional factorial design 251 ¼ 16 (five factors at two levels) corresponding to experiments 1–16 (Table 2); Table 2 Experimental design no. 1 matrix and results of capillarity rising Experiments
X1
X2
X3
X4
X5
m2/t (g2/s)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
1 þ1 1 þ1 1 þ1 1 þ1 1 þ1 1 þ1 1 þ1 1 þ1 a þa 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 1 þ1 þ1 1 1 þ1 þ1 1 1 þ1 þ1 1 1 þ1 þ1 0 0 a þa 0 0 0 0 0 0 0 0 0 0 0 0
1 1 1 1 þ1 þ1 þ1 þ1 1 1 1 1 þ1 þ1 þ1 þ1 0 0 0 0 a þa 0 0 0 0 0 0 0 0 0 0
1 1 1 1 1 1 1 1 þ1 þ1 þ1 þ1 þ1 þ1 þ1 þ1 0 0 0 0 0 0 a þa 0 0 0 0 0 0 0 0
þ1 1 1 þ1 1 þ1 þ1 1 1 þ1 þ1 1 þ1 þ1 1 þ1 0 0 0 0 0 0 0 0 a þa 0 0 0 0 0 0
2.73 16.60 9.95 17.95 16.00 23.15 47.10 35.90 1.30 1.00 10.50 1.00 5.80 25.60 42.90 56.80 3.90 19.30 0.38 4.53 0.79 47.10 5.05 5.07 4.70 3.10 36.90 41.70 39.50 34.00 46.00 33.00
the axial points (experiments 17–26, Table 2) with a ¼ [16]0.25 ¼ 2 were added to the fractional factorial design to provide for estimation of curvature model; six replicates (experiments 26–32, Table 2) at the design centre were used to allow estimation of the pure error sum of squares. Computer program Modde 41 [27] was used to generate results. The rather good correlation between experimental y values and y ones calculated from the mathematical model, shown Fig. 2, is confirmed by Statistical data shown Table 3. Table 4 shows effects or influence of each factor (b). Treatment time (b3 ¼ 11.1) and oxygen rate (b2 ¼ 6.97) are the more influent parameters. Gaseous flow effect is slightest (b1 ¼ 3.78). In comparison with the effects of these three factors, mass of powder and transmitted power ones are insignificant (b4 ¼ 0.24, b5 ¼ 0.25). However a light first-order interaction between time of treatment and oxygen rate (b23 ¼ 4.74) can be noticed. From the mathematical model it is possible to access to the response surfaces. Curves shown Fig. 3 illustrate perfectly the influence of the treatment time
Table 3 Statistical data for experimental design no. 1 and 2 Experimental design no. 1
Experimental design no. 2
R2 ¼ 0.993 R2 Adj. ¼ 0.972 Q2 ¼ 0.951 R.S.D. ¼ 2.8585 CI ¼ 95%
R2 ¼ 0.998 R2 Adj. ¼ 0.996 Q2 ¼ 0.987 R.S.D. ¼ 0.4639 CI ¼ 95%
R2, coefficient of determination; R2 Adj, adjusted coefficient of determination; Q2, (SS–PRESS)/SS; SS, sum of square of y corrected for a mean; PRESS, prediction residual sum of squares; R.S.D., square mean of residual; CI ¼ confidence interval.
30
B. Mutel et al. / Applied Surface Science 239 (2004) 25–35
Fig. 2. Correlation between y (capillarity rising) experimental values and y values calculated from the mathematical model—design no. 1.
and of the oxygen rate. So, for a gas flow ¼ 3.95 NL/ min; a transmitted power ¼ 600 W and a mass of powder ¼ 27.5 g, maximum oxygen rate and treatment time values are required to obtain the maximum capillarity rising. Table 4 Coefficients of the factors and interactions on the capillarity rising responses for experimental design no. 1 b0 b1 b2 b3 b4 b5 b11 b22 b33 b44 b55
37.02 3.78 6.97 11.10 0.25 0.24 6.35 7.36 3.20 7.99 8.28
b12
1.30
b13
2.24
b14
0.76
b15
2.63
b23
4.74
b24
2.71
b25
3.18
b34
3.79
b35
0.58
b45
0.57
Nevertheless optimisation was continued on the two most influential factors (treatment time and oxygen rate) of the previous study and a second experimental design (denoted by design no. 2) was defined. It was again a central composite design. The three less influent factors (gas flow, mass of powder and transmitted power) were kept constant and equal to 3.95 NL/min, 300 W and 20 g, respectively. The domain of variation of the two others (treatment time and oxygen rate) is shown Table 5. The experimental design was established such as defined and used previously. The response y of each experiment was evaluated from capillarity rising of ethylene glycol (gLV ¼ 46.7 mN/m, K ¼ 4:9 105 cm5). Results are gathered in Table 6. Table 7 shows that the treatment time is the more influent factor (b1 ¼ 2.57) while the oxygen rate influence is rather weak (b2 ¼ 0.063). This negative value indicates that an increase of the oxygen rate in higher level produces a decrease of the response. The good correlation between experimental and calculated y values is shown Fig. 4. Statistical values shown Table 3 show Table 5 Experimental domain and coding of variables for experimental design no. 2 Variables
Factors
Levels þ1
a
þa
X1 X2
U1, treatment time (min) U2, oxygen rate (%)
15 60 0.5 2
5.7 0.2
69.3 2.3
1
B. Mutel et al. / Applied Surface Science 239 (2004) 25–35
31
Fig. 3. Response surface for capillarity rising for design no. 1. U1 ¼ 3.95 N l/min; U4 ¼ 600 W; U5 ¼ 27.5 g; U3: 0!32.5 min; U2: 0!0.76%.
that the fitted model was adequate: 99.8% of the variability in the response could be explained by this model which was also used to plot response curves shown Fig. 5. A rather strong degree of curvature can be observed and the optimum can be easily determined Table 6 Design matrix and results of capillarity rising for design n82 Experiments
X1
X2
m2/t (g2/s)
1 2 3 4 5 6 7 8 9 10
1 þ1 1 þ1 a þa 0 0 0 0
1 1 þ1 þ1 0 0 a þa 0 0
20.1 26.7 20.6 24.7 19.9 26.9 22.2 22.9 39.2 39.2
in the experimental range. The plasma treatment of polyethylene powder method is efficient for a treatment time value equal to 42 min and for an oxygen rate equal to 1.3 percent. In these conditions, the contact angle with ethylene glycol is equal to 08; while it is higher than 908 for the untreated PE. 3.2. ESR analysis No ESR signal was detected on the untreated PE. For the PE treated in conditions previously determined to get the best wettability, ESR spectra was Table 7 Coefficients of the factors and interactions on the capillarity rising responses for experimental design no. 2 b0
b1
b2
b11
b22
b12
39.200
2.570
0.063
7.880
8.310
0.620
32
B. Mutel et al. / Applied Surface Science 239 (2004) 25–35
Fig. 4. Correlation between y (capillarity rising) experimental values and y values calculated from the mathematical model—design no. 2.
Fig. 5. Response surface for capillarity rising for design no. 2. Gaseous flow: 3.95 N l/min; transmitted power: 300 W; treated mass: 20 g; U1: 15!60 min; U2: 0!2%.
B. Mutel et al. / Applied Surface Science 239 (2004) 25–35
33
Fig. 6. Evolution of the ESR integrated signal (a.u.) versus ageing.
characterised by a six lines pattern. At first, the influence of d on the ESR integrated signal was studied. Fig. 6 shows that the radical concentration decreases very quickly and reaches 50% as soon as d ¼ 2 min. For d values higher than 2 h, the integrated ESR signal was very weak. It was no more detected after an ageing time equal to 12 h. However, the spectra are not modified during the ageing in open
air and the sextet pattern remained. The peroxy radical, described by Hori et al. [28] as typically a broad asymmetric single line (10 G) was never detected in our conditions. The six lines pattern is similar to the one already obtained by O’Neill et al. [29] by g irradiation (11.25 Mrad) of PE surfaces. According to these authors, the spectrum obtained is presumed to be the result of either the alkyl free
Fig. 7. ESR spectra. Comparison between the experimental spectrum and the simulated one based on a combination of alkyl (65%) and allyl (35%) radicals.
34
B. Mutel et al. / Applied Surface Science 239 (2004) 25–35
radical –CH2–CH –CH2–, the allyl radical –CH2–C(1) H¼C(2)H–C(1)H –CH2– $ –CH2–C(1)H –C(2)H¼C(1) H–CH2– or the polyenyl radical –CH2–(–CH¼CH– )n–CH –CH2–, or a combination of two or more of these. In the case of allyl radical, the unpaired electron spin density on the C(1) and C(2) atoms are respectively equal to 0.622 and 0.231 [30,31]. In our study, several hypotheses were taken into account. Best results were obtained (Fig. 7) for a combination based on 65% of alkyl radicals and 35% of allyl radicals. For the alkyl radicals, the splitting can occur with the four equivalent Hb proton and the Ha proton (the a carbon is the atom on which the unpaired electron is primary located). The splitting constants, respectively equal to aHa ¼ 25.80 G and aHb ¼ 33.05 G agree with literature data [29] which quote the aHa value in the range 12–22 G and the aHb value in the range 28–33 G. For the allyl radical, the splitting constants, respectively equal to aHa ¼ 20.1 G and aHa0 ¼ 9.8 G for Ha and Ha0 protons agree also with data given by O’Neill et al. [29] (to aHa ¼ 20.2 G and aHa0 ¼ 7.5 G). As XPS analyses previously performed [6] give evidence for a nitrogen fixation on PE treated by a cold remote nitrogen plasma mixed with O2, the hypothesis taking into account a splitting with a nitrogen atom was also considered. As no good correlation could be found between simulated and experimental spectrum, this hypothesis was given up. For d ¼ 0, the treated PE–NO contains 5:7 1015 radicals per gram; as the polymer weight is 4 106 g/ mol, it is equivalent to about 0.04 radicals per chain.
larity rising. The quality of the model was verified by a good agreement between experimental and predicted response. Within parameters ranges used in this study, the maximum capillary rising was achieved with an oxygen rate equal to 1.3% and with a treatment time equal to 42 min. While this duration may seem rather long, it remains shorter than the one involved with an oxygen plasma discharge treatment that requires several hours [14,15]. The low surface area of the PE powder used in this work (0.5 m2/g) may explain the insignificant effect of the mass of powder in comparison with the one of other parameters. Including specific area as a factor in the experimental device might be interesting. Works involving PE powder with higher specific areas are now in progress. After such a treatment, ESR analyses reveal the presence of a mixture of alkyl and allyl radicals. These radicals can initiate the fixation of the new functions. In a previous work, performed in a smaller reactor [6], we gave evidence that new nitrogen and oxygen incorporated functions were amine, imine, amide and alcohol, acid and ketone, respectively. For instance, alkyl radicals can react with nitrogen atoms to produce the imine or amine function, with some hydrogen transfer and/or oxygen to produce alcohol and carbonyl groups according to the sequences:
3.3. Conclusion From the results presented in this work, it is confirmed that the coupling between fluidised bed and far cold remote nitrogen plasma processes is efficient to modify surface properties of a PE powder and especially to increase its wettability. A first design of experiments (central composite design) allowed determining that among the five main process factors (treatment time, oxygen rate added to nitrogen, microwave power, gaseous flow and mass of treated powder), the two first were the more influent. Then, a second experimental design involving the two main factors previously determined was performed to optimise the process in order to obtain the higher capil-
The imine function, exposed to wet air after the plasma treatment and before XPS analysis, may also be partly hydrolysed to form ketone group.
Amide, alcohol and carbonyl functions can also be created from a primary alkyl group according to:
B. Mutel et al. / Applied Surface Science 239 (2004) 25–35
References [1] B. Mutel, O. Dessaux, P. Goudmand, F. Luchier, Rev. Phys. Appl. 25 (1990) 1019. [2] B. Mutel, C.Hoyez, P. Goudmand, O. Dessaux, L. Gengembre, J. Grimblot, in: Proceedings of the 12th International Symposium on Plasma Chemistry, Minneapolis, 1995, p. 167. [3] B. Mutel, J. Grimblot, P. Goudmand, O. Dessaux, Surf. Interface Anal. 30 (2000) 401. [4] B. Mutel, J. Grimblot, V. Moineau, T. Colson, P. Goudmand, O. Dessaux, Surf. Interface Anal. 30 (2000) 41. [5] P. Supiot, O. Dessaux, P. Goudmand, J. Phys. D: Appl. Phys. 28 (1995) 1826. [6] C.R. Wierenga, T.J. Morin, AIChE J. 35 (1989) 1555. [7] T. Rogers, T.J. Morin, Plasma Chem. Plasma Proc. 11 (1991) 203. [8] P. Arnould, J. Amouroux, Bull. Soc. Chim. Fr. 6 (1987) 985. [9] T. Okubo, H. Kawamura, K. Kusakabe, S. Morooka, J. Am. Ceram. Soc. 43 (1990) 1150. [10] M. Karches, P.R. von Rohr, Surf. Coat. Technol. 142–144 (2001) 28. [11] M. Karches, CH. Bayer, Ph. Rudolf Von Rohr, Surf. Coat. Technol. 116–119 (1999) 879. [12] B. Borer, P.R. Rudolf Von Rohr, in: Proceedings of the 16th International Symposium on Plasma Chemistry, Taormina, 2003. [13] M. Heintze, V. Bru¨ ser, W. Brandl, G. Marginean, H. Bubert, S. Haiber, Surf. Coat. Technol. 174/175 (2003) 831. [14] N. Inagaki, S. Tasaka, H. Abe, J. Appl. Polym. Sci. 46 (1992) 595. [15] S.H. Park, S.D. Kim, Polym. Bull. 33 (1994) 249. [16] C. Vivien, C. Wartelle, B. Mutel, J. Grimblot, Surf. Interface Anal. 34 (2002) 575.
35
[17] J.B. Leroy, N. Fatah, B. Mutel, J. Grimblot, Plasma Polym. 8 (1) (2003) 13. [18] O. Dessaux, P. Goudmand, B. Mutel, J. Quant. Spectrosc. Radiat. Transfer 29 (1983) 49; 30 (1983) 31; 30 (1983) 137. [19] E.W. Washburn, Phys. Rev. Ser. 2 (17) (1921) 273. [20] J. Goupy, Tech. Inge´ nieur Traite´ Anal. Caracte´ 1–24 (1996) 230. [21] J. Goupy, La me´ thode des plans d’expe´ riences; Dunod, Paris, 1888. [22] J. Goupy, Plans d’expe´ riences pour surface de re´ ponses; Dunod, Paris, 1999. [23] P. Dagnelie, in Plans d’expe´ riences, Applications a` l’entreprise, in: J.J. Droesbeke; J. Fine, G. Saporta (Eds.), Technip, Paris, 1997, pp 13–16, Chapter 2.1. [24] D. Mathieu, R. Phan Tan Luu, in plans d’expe´ riences, Applications a` l’entreprise, in: J Droesbeke; J.J.; Fine, J.; Saporta, G., (Eds.), Technip, Paris, 1997, pp 466–483, Chapter 9.3. [25] P. Simmerling, J.-C. Sisson, A. Zaidi, Pratique des plans d’expe´ riences; Technique et Documentation; Lavoisier, Paris, 1998. [26] G.E. Box, W.G. Hunter, J.S. Hunter, Response Surface Methods. Statistics for Experimenters; Wiley, New York, 1978. [27] Modde 4.0, UMETRI AB, Box 7960, S-90719 UMEA, Sweden. [28] Y. Hori, S. Shimada, H. Kashiwabara, Polymer 20 (1979) 181. [29] P. O’Neill, C. Birkinshaw, J.J. Leahy, R. Barklie, Polym. Degrad. Stab. 63 (1999) 31. [30] H.C. Lefkowitz, J. Fain, F.A. Matsen, J. Chem. Phys. 23 (1955) 1690. [31] C. Heller, H.K. Mconnell, J. Chem. Phys. 32 (1960) 1535.
