Tailored aluminium oxide layers by bipolar current ...

Surface & Coatings Technology 201 (2007) 8677 – 8682 www.elsevier.com/locate/surfcoat

Tailored aluminium oxide layers by bipolar current adjustment in the Plasma Electrolytic Oxidation (PEO) process F. Jaspard-Mécuson a,b , T. Czerwiec b , G. Henrion b,⁎, T. Belmonte b , L. Dujardin a , A. Viola a , J. Beauvir c b

a Messier-Bugatti, 5 rue A. de St Exupéry, 67123 MOLSHEIM cedex, France Laboratoire de Science et Génie des Surfaces (UMR CNRS-INPL 7570)- Parc de Saurupt - CS 14234- 54042 NANCY Cedex, France c Ceratronic recherche, 13 rue des jardins, 67380 LINGOLSHEIM, France

Available online 16 October 2006

Abstract The plasma electrolytic oxidation process of aluminium alloys is investigated for two different current waveforms. It is shown that particular conditions may be established which strongly reduce the arcing that usually cause detrimental defects in the oxide layer for treatment time greater than typically 40–50 min. This results in a “softer” process. As a consequence thick homogenous layers may be grown with no large discharge channels. Through the presented results, the importance of the negative charge density relative to the positive one is evidenced thus pointing out the need of using a pulse bipolar current supply. © 2006 Elsevier B.V. All rights reserved. Keywords: Plasma Electrolytic Oxidation (PEO); Oxidation; Aluminium; Aluminium oxide; Growth kinetics

1. Introduction Plasma electrolytic oxidation (PEO) [1], also known as micro-arc oxidation (MAO) [2], anodic spark oxidation [3] or spark anodizing [4] has gained growing interest for the recent years in light-weight metal (Al, Ti, and Mg alloy) oxidation [5]. Compared with conventional anodising, PEO treatments are usually achieved by using high voltage, low frequency (around 50 Hz) AC supply in dilute alkaline electrolytes [1]. The oxide layers obtained by PEO processing of aluminium alloys are thick, hard and well-adherent to the substrate. As a consequence, the surface properties obtained after PEO treatments are promising for industrial applications. The oxide layers achieved on aluminium alloys can be subdivided in two sublayers [6–14]: • The superficial porous layer contains γ−Al2O3 with hardness between 500–1000 Hv and poor mechanical characteristics (especially in terms of wear). This layer will be referred to as the “porous layer” along this paper, since it is generally known as so. ⁎ Corresponding author. Tel.: +33 383 584 255. E-mail address: [email protected] (G. Henrion). 0257-8972/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2006.09.005

• The internal layer, lying between the porous layer and the substrate, contains γ−Al2O3 and also α−Al2O3 with hardness between 900–2000 Hv and very good mechanical properties. This layer is usually described as the “compact layer”; this denomination will be used in the present work. Owing to its poor mechanical behaviour, the porous outer layer needs usually to be removed prior to industrial applications. Therefore, many efforts are done to limit or to suppress the growth of the porous layer during a PEO treatment. A way to comply this requirement is to optimize the electrolyte composition. Another approach consists in using special current regimes such as bipolar current pulses [15–20]. Indeed, Timoshenko and Magurova [18] showed that by pulsing the current in both negative and positive semi-periods significantly improves the properties of the oxide layer formed on magnesium alloy. They also evidenced the importance of the pulse initiation delay and duration when using pulsed bipolar current supply. Yerokhin et al. [19] recently reported a comparison between aluminium oxide layers properties obtained by 50 Hz AC and pulsed bipolar current (PBC) modes of PEO. They showed that the PBC-PEO process is able to improve the oxide layer morphology, especially by reducing the thickness of the porous outer layer.

8678

F. Jaspard-Mécuson et al. / Surface & Coatings Technology 201 (2007) 8677–8682

However, as far as the oxide layer grows, the micro-discharges that develop across the material surface become more and more intense and can cause detrimental defects in the oxide layer [16,21]. Consequently, specific process conditions must be developed in order to reduced or suppress the arcing effect. This paper reports on some results obtained with the CERATRONIC® pulsed current supply [15]. In particular, it will be shown that strong arcing can be significantly reduced by adjusting the relative shape of the negative and positive current pulses, and that thick oxide layers with different morphology can be obtained with these special working conditions. The associated micro-discharges characteristics, as performed by using digital camera and optical emission spectroscopy, are also discussed. 2. Experimental procedure The experimental device consists basically of a conventional electrolysis tank (Fig. 1). The electrolytic solution consists of KOH and Na2SiO3 diluted in distilled water; the resulting electrolyte conductivity ranges from 2.3 to 2.8 mS/cm; its pH is 11.7. A cooling system maintains the electrolyte temperature below 30 °C during the process. The rectangular samples (100⁎90⁎6 mm3) immersed in the electrolyte are made of 2214T6 aluminium alloy. They are cleaned with acetone before each treatment to avoid poisoning of the electrolyte bath and to ensure a reproducible initial surface contamination for each sample. Two stainless steel counter-electrodes face both sides of the sample which are treated simultaneously. Two alternative current generators (CERATRONIC® process [15]) are used, which deliver current to the substrate with amplitude in the range 0–36 A and 75–150 A respectively. Whatever the generator used, the current frequency is set at 100 Hz. Process parameters, such as the period (T), the different duration of each step in the

current waveform (Ti), the amplitude of the positive (Ip) and negative (In) current can be adjusted over a wide range. For the presented results, the value of I p was fixed at 77 A, corresponding to a current density of 38 A/dm2. We define the parameter R as being the positive to negative charge quantity ratio (see Fig. 1 for the definition of Ti) : Z T1 þT2 þT 3 Z T qp R ¼ with qp ¼ Ip d dt and qn ¼ In d dt qn 0 T1 þT2 þT3 Depending on the working conditions, Ti values and In amplitude are adjusted to match the desired R value. The microdischarges occurring on the sample surface during the PEO process are characterized by means of optical emission spectroscopy (OES) with a spectral resolution of 0.3 nm. The light emitted by the micro-discharges is collected by an optical fibre immersed in the electrolyte, and located a few centimetres from the sparks to optimise the collected light intensity and to reduce absorption by the electrolyte bath. A video camera is used to observe the evolution of the micro-discharge aspect over the process duration. Except otherwise mentioned, the integration time of the camera is fixed at 8–10 ms. The layer characteristics are investigated by use of scanning electron microscopy (SEM). Measurement of the layer thickness results from the statistical average of at least 40 measurements taken randomly on a cross-section of the oxide layer. 3. Results and discussion 3.1. Process with arcs (R = 1.57) Some results which are typical from a conventional PEO process will be presented first. For that purpose, the positive

Fig. 1. Schematic view of the experimental device together with the imposed pulsed current shape over one period.


8679

Fig. 2. Side-view pictures of aluminium alloy samples at different time of the PEO process (R = 1.57). The integration time of the camera is set between 8 and 10 ms.

current density is adjusted to 38 A dm− 2 and the charge quantity ratio R is set to 1.57. The evolution of the micro-discharge aspect as a function of the treatment time is depicted in Fig. 2 that corresponds to sideview pictures of the sample during PEO processing of aluminium alloy. For these views, the integration time of the camera was set between 8 and 10 ms. Meanwhile, the time variation of the Al (λ = 396.2 nm) emission line intensity along the PEO process is reported in Fig. 3. From Figs. 2 and 3, it is possible to recognize the usual steps of a PEO process as described earlier [16,22]. At the early stage of the process, the growing oxide film breaks down due to an increase in the applied voltage. This was called step 1 in Refs. [16,22], which is characterized by sparks flashing randomly all over the aluminium alloy surface (Fig. 2a). Following this step, a maximum in Al emission intensity (IAl) is observed around 2 min (Fig. 3). From that time, IAl starts an exponential decrease with time constant τ, while sparks progressively change to micro-arcs (step 2, Fig. 2c). Finally, after the end of the rapid IAl decrease, step 3 (arcs regime, Fig. 2d,e) is characterized by an aluminium signal close to zero with random intensity peaks

(Fig. 3) due to the few strong remaining arcs that appear on the surface. Micro-arcs and arcs occurring during steps 2 and 3 cause irreversible damages to the oxide layer as illustrated in Figs. 4 and 5. From the SEM cross-section of the oxide layer obtained after a 140 min PEO treatment (R = 1.57), discharge channels may be observed that pass through the overall layer. Such discharge channels are due to strong arcs occurring during step 3. With such process conditions after 140 min of PEO treatment, the oxide layer appears as inhomogeneous, rather thin, with an irregular metal–oxide interface and exhibits some porosities or empty inclusions (Fig. 4). The layer is divided into two sub-layers: an outer layer constituted of plates and pores and an inner more compact layer. The corresponding SEM top surface view shown in Fig. 5 reveals that discharge channels are always surrounded by craters described as “pancakes” by Sundararajan and Rama Krishna [23]. From the mechanism proposed by this group [23,24], the layer growth results from molten aluminium which is oxidized when flowing out through the discharge channels that are created due to the oxide layer breakdown. By this way, alumina is formed which contributes to the layer when being ejected from the channels and rapidly cooled at the surface–electrolyte interface (pancake and plate layer formation). From a comparison between Figs. 4 and 5, it appears that the mechanism proposed in Refs. [23,24] for the

Fig. 3. Time variation of the Al (λ = 396.2 nm) emission line intensity along the PEO process (R = 1.57).

Fig. 4. SEM micrograph cross-section after a 140 min PEO treatment (R = 1.57). Description of the different sub-layers.

8680


Fig. 5. SEM micrograph of the top surface after a 140 min PEO treatment (R = 1.57).

pancake formation also holds for the formation of the outer layer constituted by plates and pores.

Fig. 7. Time variation of the Al (λ = 396.2 nm) emission line intensity along the PEO process (R = 0.89).

3.2. Process with reduced arcing (R = 0.89) For treatments at the same positive current density (38 A dm− 2), but with a charge quantity ratio R = qp / qn = 0.89, the initial stage of the PEO process (from 0 to 40 min) is the same as the one described in Section 3.1. Pictures represented in Fig. 6 show an optical aspect of the discharges similar to the one reported in Fig. 2. However, for treatment times longer than typically 40 min, the surface activity of the discharges turns to decrease (Fig. 6d). In fact, it must be noted that the integration time was set 200 times higher for picture 6d than for pictures 6a, 6b and 6c (2s against 8–10 ms). This points out a total modification of the process regime which switches from an “arc” regime to a “softer” one which does not exhibit the strong arcs that were reported as detrimental to the oxide layer [16,22]. Consequently for treatment time higher than 40 min, the acoustic emission and OES intensities are strongly lowered than those observed in the process with arcs (Section 3.1). This is illustrated in Fig. 7, where the variation of aluminium line intensity is plotted as a function of the treatment time. As already mentioned, for treatment time shorter than typically 35 to 40 min the process behaves as in the previous case, and the variation of IAl as a function of treatment time is very similar to

the one reported in Fig. 3 with conditions R = 1.57. On the other hand, IAl falls down to zero for treatment time longer than 40 min indicating a strong decrease of the discharge activity and efficiency to produce and excite aluminium vapour. At this point we must precise that the integration time used for OES measurements cannot be set higher than few tens of ms in order to avoid the CCD detector to saturate during the first steps of the PEO process (around 2 min). Owing to the low intensity and random distribution of the remaining discharges, and considering the solid angle of the optical fibre, the OES aluminium signal is so low that it cannot be detected under these conditions. Within these particular current conditions and after 60 min of PEO treatment, the oxide layer thickness becomes more homogenous and the layer–substrate interface is smooth. This homogenisation, which occurs very locally for treatment times around 50 to 60 min extends to the whole surface as the process duration exceeds 80 min (Fig. 8). At those times, the oxide layer is still composed of two sub-layers but the plates that formed the top layer in the process with arcs (R = 1.57; Section 3.1) have disappeared giving place to a porous layer as described in the literature [6–14]. From the SEM observation of the top surface

Fig. 6. Side-view pictures of aluminium alloy samples at different time of the PEO process (R = 0.89). Integration time is set between 8 and 10 ms for pictures a, b and c; Integration time is set at 2000 ms for picture d.


8681

Fig. 10. Dependence of the average total layer thickness on the PEO treatment time. R = 1.57 (squares) and R = 0.89 (triangles).

Fig. 8. SEM micrograph cross-sections at different PEO process duration (R = 0.89). a) 10 min; b) 40 min; c) 90 min.

of the sample (Fig. 9), it appears that the pancakes are still present but the discharge channels have disappeared or are difficult to observe. In place of the previously observed discharge channels, honeycomb-like structures, due to a lot of very small holes (less than 1 μm in diameter) are observed at the oxide surface (Fig. 9). One may reasonably think that these holes are related to the pores (less than 1 μm in diameter) observed in Fig. 8. 3.3. Comparative discussion of the two processes

treatments [16,21,22,23,24,25]. This is consistent with the growth mechanism proposed by Sundararajan et al. [23,24] considering the growth of the oxide layer as a coating process resulting from molten aluminium that is oxidized when flowing out through the discharge channels. By this way, alumina is formed which contributes to the coating when being ejected from the channels and rapidly cooled at the surface-electrolyte interface. The slope of the linear variation of the layer thickness as a function of the treatment time was shown to be determined by the positive current density [16–21]. So the same slope observed in Fig. 10 for short treatment times (b 40 min) can be explained by the same positive current density used in the two processes. This is also consistent with the proposed growth mechanism because discharges leading to micro-arcs or arcs only occur during the positive current alternation [21,26,27]. For treatment time higher than 40 min the layer thickness evolution depends on the positive to negative charge quantity ratio R. In the process with high R value (R = 1.57), the growth kinetics exhibits a rather regular behaviour that may still be described by a linear variation. On the other hand, when the

The variation of the average total layer thickness with the process time for both cases (R = 1.57 and R = 0.89) are reported in Fig. 10. It can be observed that for short treatment time (b40 min) the growth kinetics is the same for the two processes, exhibiting a linear kinetics commonly observed in PEO

Fig. 9. SEM micrograph of the top surface after a 60 min PEO treatment (R = 0.89).

Fig. 11. Evolution of the mean pancake diameter as a function of the PEO treatment duration R = 1.57 (squares); R = 0.89 (circles).

8682


negative charge quantity is set greater than the positive one (R = 0.89), the growth kinetics exhibits two different regimes as pointed out in Fig. 10. The growth kinetics presents a linear to parabolic transition between 50 and 70 min. This kinetics transition is connected to the switching from arc to soft regime described earlier. It is noteworthy that there is an enhancement of the growth rate at the kinetics transition, while a diminution is generally expected [28]. Such an effect is probably due to the vanishing of a loss term in the growth kinetics for R = 0.89. As a matter of fact, arcs contribute to the growing of the layer as described earlier but they also destroy the forming layer. Therefore, the growth of the oxide layer results from a competition between these two phenomena. To illustrate this purpose, the mean value of the pancake diameter as a function of the PEO process duration is reported in Fig. 11 for the two conditions discussed here. If the pancake diameter is similar in both cases at short process times, it clearly appears that the pancake size remains at a constant value for treatment times higher than 40 min when using negative quantity of charges greater than the positive one (R = 0.89). This clearly indicates that the removing of the layer is minimized with these conditions and is thus consistent with an increase in the growth rate. It must be noted that such a “soft” PEO discharge regime was previously reported in the literature [10,29]. Xue et al. [10] noted that big sparks spots turn to very fine sparks that can be observed only in the dark. However, the electrical conditions are not precisely described in this work. Malyshev [29] explained that it is necessary either to reduce the current density or to increase the cathodic to anodic current ratio in order to improve the layer quality. He associated the “soft” regime established in those conditions to a self-organisation of the specimen. 4. Conclusion The PEO processing of aluminium alloys (2214-T6) has been investigated for different shapes of the applied current waveform. A particular attention was paid to two specific conditions by changing the relative quantity of the positive charge density with respect of the negative one. From OES discharge investigations and SEM analysis of the processed samples, it appeared that the PEO process behaviour is similar in both conditions during the first steps of the treatment, typically up to treatment time close to 40 min. On the other hand, it has been established that applying a higher negative charge quantity than the positive one to the substrate greatly improves the resulting oxide layers, especially in terms of thickness and homogeneity over the whole surface, strongly reduces the presence of large discharge channels in the final layer. This last aspect has been related to the growth kinetics which exhibits an acceleration when working with R = 0.89 at process time higher than 50–60 min corresponding to a switch from linear to parabolic growth regime. The “softer” discharges that can hardly be observed within these conditions results in lowering the loss term associated to strong arcs occurring during the treatment with R = 1.57 and responsible of the creation of large discharge channels. These results point out the importance of the negative current alternation in the PEO process and the

need of using a bipolar current supply with independent adjustable parameters. However, it is necessary to refine this study in order to better define the exact current conditions for the establishment of the “softer” regime. Meanwhile, specific studies must be undertaken to shed light on the influence of the negative current on the PEO process. Works are presently on the way from which we expect to better understand what is really going on during the different phases of the process, depending on the current waveform. Acknowledgments The financial support from the French Research Ministry through “PROXY3A” (RNMP project) is greatly acknowledged. One of us (F.J.M.) would like to acknowledge the National Association for Technical Research (ANRT) getting a grant in the frame of a CIFRE convention. References [1] A.L. Yerokhin, X. Nie, A. Leyland, A. Matthews, S.J. Dowey, Surf. Coat. Technol. 122 (1999) 73. [2] G. Markov, G. Markova, USSR Patent 526 961, Bulletin of Inventions 32 (1976). [3] P. Kurze, W. Krysmann, H.G. Schneider, Cryst. Res. Technol. 21 (1986) 1603. [4] F. Monfort, A. Berkani, E. Matykina, P. Skeldon, G.E. Thompson, H. Habazaki, K. Shimizu, J. Electrochem. Soc. 152 (2005) C382. [5] Special issue on “Plasma-assisted electrochemical techniques for treatment of metal surfaces", Surf. Coat. Technol. 199 (2005). [6] W. Xue, Z. Deng, Y. Lai, R. Chen, J. Am. Ceram. Soc. 81 (5) (1998) 1365. [7] Y.K. Wang, L. Sheng, R.Z. Xiong, B.S. Li, Surf. Eng. 15 (2) (1999) 109. [8] X. Nie, A. Leyland, H.W. Song, A.L. Yerokhin, S.J. Dowey, A. Matthews, Surf. Coat. Technol. 116–119 (1999) 1055. [10] W. Xue, Z. Deng, R. Chen, T. Zhang, Thin Solid Films 372 (2000) 114. [11] J. Tian, Z. Luo, S. Qi, X. Sun, Surf. Coat. Technol. 154 (2002) 1. [12] J.A. Curran, T.W. Clyne, Surf. Coat. Technol. 199 (2005) 16. [13] A.L. Yerokhin, L.O. Snizhko, N.L. Gurevina, A. Leyland, A. Pilkington, A. Matthews, Surf. Coat. Technol. 177–178 (2004) 779. [14] R.C. Barik, J.A. Wharton, R.J.K. Wood, K.R. Stokes, R.L. Jones, Surf. Coat. Technol. 199 (2005) 158. [15] J. Beauvir, Patent WO 01/81658 A1 (2001). [16] F. Mécuson, T. Czerwiec, T. Belmonte, L. Dujardin, A. Viola, G. Henrion, Surf. Coat. Technol. 200 (2005) 804. [17] J. Liang, B. Guo, J. Tian, H. Liu, J. Zhou, W. Liu, T. Xu, Surf. Coat. Technol. 199 (2005) 121. [18] A.V. Timoshenko, Y.V. Magurova, Surf. Coat. Technol. 199 (2005) 135. [19] A.L. Yerokhin, A. Shatrov, V. Samsonov, P. Shashkov, A. Pilkington, A. Leyland, A. Mathews, Surf. Coat. Technol. 199 (2005) 150. [20] B.H. Long, H.H. Wu, B.Y. Long, N.D. Wang, X.Y. Lu, Z.S. Jin, Y.Z. Bai, J. Phys., D. Appl. Phys. 38 (2005) 3491. [21] F. Jaspard-Mecuson, PhD thesis, Institut National Polytechnique de Lorraine, Nancy (France), (2005). [22] F. Mécuson, G. Henrion, T. Czerwiec, L. Dujardin, A. Viola, T. Belmonte, EUROMAT'2005, 5–8 Sept. 2005, Prague (Czech Rep.), 2005. [23] G. Sundararajan, L. Rama Krishna, Surf. Coat. Technol. 167 (2003) 269. [24] L. Rama Krishna, K.R.C. Somaraju, G. Sundararajan, Surf. Coat. Technol. 163–164 (2003) 484. [25] A.L. Yerokhin, L.O. Snizhko, N.L. Gurevina, A. Leyland, A. Pilkington, A. Matthews, Surf. Coat. Techol. 177–178 (2004) 779. [26] A.V. Timoshenko, Y.V. Magurova, Prot. Met. 31 (1995) 474. [27] A.V. Timoshenko, Y.V. Magurova, Rev. Metal. (Madrid) 36 (2000) 323. [28] B.E. Deal, A.S. Grove, J. Appl. Phys. 36 (12) (1965) 3770. [29] V.N. Malyshev, J. Adv. Mater. 4 (1) (1997) 18.