The effect of applied dc bias voltage on the properties of a-C:H films prepared in a dual dc–rf plasma system

Applied Surface Science 227 (2004) 364–372

The effect of applied dc bias voltage on the properties of a-C:H films prepared in a dual dc–rf plasma system H.X. Li, T. Xu, J.M. Chen*, H.D. Zhou, H.W. Liu State Key Laboratory of Solid Lubrication, Lanzhou Institute of Chemical Physics, Chinese Academy of Sciences, Lanzhou 730000, China Received 1 October 2003; received in revised form 10 December 2003; accepted 10 December 2003

Abstract A dual direct current and radio frequency (dc–rf) plasma system was used to deposit hydrogenated amorphous carbon (a-C:H) films from methane plasma. It has the advantages of separately controlling ion density and ion energy by rf power and dc bias, respectively, over conventional simply capacitive-coupled rf-PECVD system. Thus the a-C:H films were obtained at different applied dc biases using CH4 plus H2 as the feedstock. The structural, mechanical and tribological properties of the resulting aC:H films were investigated as a function of the applied dc bias voltage in the range 0–500 V. The results showed that the structure and properties of a-C:H films strongly depended on the applied dc bias voltage. With an increase in the applied dc bias voltage, the deposition rate and the bonded hydrogen content decreased, and the hardness and sp3 content (according to Raman spectra) were shown to reach their maximum values simultaneously at an applied dc bias voltage of 100 V. At the same time, the smallest friction coefficient and the best wear-resistance were obtained. # 2004 Elsevier B.V. All rights reserved. Keywords: Hydrogenated amorphous carbon (a-C:H) films; Applied dc bias voltage; Chemical vapor deposition (CVD)

1. Introduction Hydrogenated amorphous carbon (a-C:H) films have been attracting considerable interest due to their unique combination of properties, such as high hardness, low friction coefficient, high electrical resistivity, optical transparency, and chemical inertness. These properties make a-C:H films suitable for numerous potential applications in low-friction and wearresistant coatings, protective optical and biomedical coatings, electro-luminescent materials, and field emission devices [1–3]. Various techniques have been *

Corresponding author. Tel.: þ86-931-8277491; fax: þ86-931-8277088. E-mail address: [email protected] (J.M. Chen).

developed to preparing a-C:H films, including ionbeam deposition [4], reactive magnetron sputtering [5], pulsed laser deposition [6], radio frequency (rf) plasma enhanced chemical vapor deposition (rfPECVD) [7], electron cyclotron resonance chemical vapor deposition (ECR-CVD) [8]. Among the abovementioned deposition techniques, rf-PECVD is accepted as the most popular laboratory method to prepare a-C:H films from a hydrocarbon precursor. It is known that in the deposition of a-C:H films, the bias voltage charged on the negative electrode, which determines the ion energy impinging onto the growing surface, plays an important role in controlling the structure and properties of the films [1,9–12]. Extensive studies have investigated the effects of the rf induced self-bias voltage on the film growth process

0169-4332/$ – see front matter # 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2003.12.013

H.X. Li et al. / Applied Surface Science 227 (2004) 364–372

and the film properties [13–16]. However, using the rf induced self-bias is not capable of independently controlling the plasma density and the ion energy, since they both vary with the rf drive power. Furthermore, the rf induced self-bias depends not only on the rf power, but also strongly on the deposition conditions (such as deposition pressure), resulting in difficulty to control the induced self-bias voltage. Apart from the rf induced dc self-bias, an applied dc voltage can also supply a negative biasing effect. The bias voltage can be exactly and easily controlled by applying a dc power to the rf-PECVD system. It might be feasible to obtain quality a-C:H films using the socalled dc–rf-PECVD technique [17,18]. However, few papers have been published on the growth of a-C:H films making use of the dc–rf-PECVD technique. The purpose of this work is to investigate the effect of applied dc bias voltage on the structure, mechanical and tribological properties of a-C:H films deposited using the dual dc–rf-PECVD technique.

2. Experimental details 2.1. Film deposition A direct current and radio frequency PECVD system was used to deposit a-C:H films. The system consisted of a plasma reactor equipped with a loadlock chamber and a gas supply and pumping system. The cylindrical stainless-steel vacuum reactor chamber was 50 cm in diameter and 50 cm in height, with two planar electrodes (400 and 900 cm2 in area, respectively) being placed vertically and 7 cm apart from each other. A rf power source (13.56 MHz) was supplied on to the larger, upper electrode to excite the plasma and control the plasma density. A separate dc power source (1000 V) was supplied on to the smaller, lower electrode to bias the substrate and control the ion energy. A base pressure of 105 Pa was attained in the chamber with a turbomolecular pumping system. a-C:H films were deposited on polished n-type Si(1 0 0) wafers at different applied dc bias voltages, using CH4 and H2 gases as the feedstock. The deposition conditions are detailed in Table 1. Prior to deposition, the substrates were cleaned with Ar plasma sputtering at a bias voltage of 400 V for 20 min so as to remove the native oxide on the Si surface. During

365

Table 1 Summary of a-C:H film deposition conditions Item

Parameter

CH4 gas flow rate (sccm) H2 gas flow rate (sccm) Deposition pressure (Pa) Radio frequency power (W) Applied dc bias (V)

10.3 20 5 200 0, 100, 200, 300, 400 and 500

the film deposition, the substrate was only heated by the plasma and the substrate temperature was measured by a thermocouple mounted on the substrate holder. 2.2. Characterization of a-C:H films The surface morphology of the a-C:H films was observed on a SPM-9500 atomic force microscope (AFM). The film deposition rate was estimated by measuring the thickness using surface profilometry. Micro-Raman backscattering spectra of the a-C:H films were recorded on a Jobin Yvon T64000 spectrometer operating with 514.5 nm Ar laser as the excitation source. The Raman spectra were fitted based on two Gaussian curve shapes with a curve-fitting software. Fourier transformation infrared (FTIR) spectra of the a-C:H films were recorded on a FTS165 spectrometer so as to detect the changes of the hydrogen bonding in the films. 2.3. Mechanical properties The hardness of the a-C:H films were determined on a nano-indenterII (MTS Systems Corporation) using a Berkovich diamond tip and continuous stiffness option, with the maximum indentation depth being kept at 40 nm to minimize the substrate contribution. Five replicate indentations were made for each film sample and the hardness was calculated from the load– unloading curves. The adhesion strength was measured using a scratch tester. The radius of a diamond stylus in the scratch test was 0.4 mm and the measurements were carried out at a stable speed of 3 mm/min and a load increasing rate of 1 N/min. Film failure is determined by an acoustic emission detector. The load at which fracture takes place is defined as the critical load, and has been

H.X. Li et al. / Applied Surface Science 227 (2004) 364–372 180 160

o

proposed to be a quantitative measure for practical adhesion.

Deposition temperature ( C)

366

2.4. Tribological behavior The friction and wear behaviors of the films were evaluated on a commercially available ball-on-disk tribometer by sliding the coated substrate against a Si3N4 ceramic ball (diameter 4 mm, HV 1400–1700) at a sliding velocity of 60 mm/s and a load of 2 N. The friction test rig provides a reciprocating-sliding configuration. All the tests were conducted at room temperature and a relative humidity of 40–45%. It was assumed that lubrication failure of the films occurred as the friction coefficient rose sharply to a higher and stable value similar to that of a cleaned bare Si substrate against the same counterface. The wear tracks of the films were observed on a JSM-5600LV scanning electron microscope (SEM).

140 120 100 80 60 40 0

100

200

300

400

500

Applied DC bias voltage (V) Fig. 1. Deposition temperature as a function of applied dc bias voltage.

seen that the deposition temperature increases linearly with the increase in applied dc bias voltage. This may be due to an increased heat generated by greater collision of higher-energy ion species arising from the increase in the applied dc bias voltage.

3. Results and discussion 3.2. Surface morphology using AFM 3.1. Deposition temperature of a-C:H films In the film deposition process, the substrate was only heated by the plasma and the substrate temperature was measured by a thermocouple mounted on the substrate holder. Fig. 1 shows the deposition temperature as a function of applied dc bias voltage. It can be

Fig. 2 shows the typical 3D AFM images of the aC:H films deposited at an applied dc bias voltage of 100 V (Fig. 2(a)) and 400 V (Fig. 2(b)). It is seen that the films are uniform and smooth, and are composed of small and compact spheres. The size of the spherical grains and the root mean square (rms) roughness

Fig. 2. AFM images of the morphologies of a-C:H films deposited at applied dc bias of 100 V (a) and 400 V (b).

H.X. Li et al. / Applied Surface Science 227 (2004) 364–372 Table 2 The thickness and deposition time of a-C:H films produced at different applied dc bias voltage Applied dc bias voltage (V)

Film thickness (nm)

0 100 200 300 400 500

375 425 400 390 370 360

     

Deposition time (min)

10 25 45 50 40 60

100 120 150 200 240 340

of the a-C:H films increase with increasing applied dc bias voltage and might be attributed to the increasing number of sp2 bonded carbon sites [19]. 3.3. Deposition rate of the films The average thickness and the deposition time of the a-C:H films produced at different applied dc bias voltage are summarized in Table 2. It is found that the film thickness lies in the range of 360–425 nm. Then the average deposition rate of the films is evaluated. Fig. 3 shows the relationship between the deposition rate of a-C:H films and the applied dc bias voltage. Reduction in the deposition rate is observed with an increase in applied dc bias voltage. It is thought that the reason for the decrease in deposition rate can possibly be due to the increase in deposition temperature with increasing applied dc bias, as shown in Fig. 1. Yoon et al. [20] has explained this phenomenon by the

367

behaviour of the adsorbed layer model [14,21]. According to this model, the collision of neutral species with the substrate forms a physisorbed layer in the first step by an attractive dipole–dipole interaction between the substrate and the adsorbed species. The physisorbed species can then be chemisorbed in the second step by the interaction with impinging ions, leading to the formation of new chemical bonds and film growth. An increase in the substrate temperature reduces the residence time of the physisorbed layer, thus reducing the deposition rate. 3.4. Raman analysis Raman spectroscopy is popularly used to probe the quality of carbon films due to its ability to distinguish between different bonding types and domain sizes [22,23]. Usually, the Raman spectra of a-C:H films are characterized by a G peak around 1550 cm1 and a D shoulder around 1350 cm1. Many previous studies [1,24–26] have shown that the G peak shift and the D peak shift in Raman spectra can provide some information about the a-C:H film structure. Generally speaking, the G peak and D peak of a conventional a-C:H film will shift towards a lower wavenumber with increasing sp3 bonding in the films, which corresponds to the increased diamond-like characteristics of the films. Fig. 4 shows the Raman spectra of the a-C:H films deposited at different applied dc bias voltage. The peak positions of the D and G line signals were

5.0 1577

500V

4.0

Intensity (a.u.)

Deposition rate (nm/min)

4.5 3.5 3.0 2.5 2.0

1391 1570

400V

1384 1563

300V

1378 1557

200V

1377 1548

100V

1.5

1373

0V

1.0 0.5

1000

0.0 0

100

200

300

400

500

1200

1400 1600 -1 Raman shift (cm )

1800

Applied DC bias voltage (V) Fig. 3. Deposition rate of a-C:H films as a function of applied dc bias voltage.

Fig. 4. Raman spectra of a-C:H films deposited at different applied dc bias voltage showing the shifts in the D peak and G peak positions.

H.X. Li et al. / Applied Surface Science 227 (2004) 364–372

determined by fitting the Raman spectra using two Gaussian distributions. From the spectra, it can be seen that as the applied dc bias is changed from 100 V to 500 V, the G peak position shifts towards higher wavenumber from 1548 to 1577 cm1, and the D peak position also shifts upwards from 1373 to 1391 cm1. It should be pointed out that the Raman spectra of the film deposited at zero applied dc bias exhibits strong fluorescence and photoluminescence when excited optically, suggesting the formation of an organic polymerized structure with high hydrogen content [15]. The film shows visible emission from the surface during Raman measurement and this could have masked the Raman signal, resulting in a strong background emission. The variation of the Raman spectra may parallel the transition in a-C:H film structure from polymer-like to diamond-like, and then to graphite-like. This can be explained using the subsurface implantation model reported by Silva and Amartunga [27] and Lifshitz et al. [28]. Accordingly, at zero applied dc bias voltage, the ion energy is too much small and most of the hydrocarbon species are trapped on the surface, resulting in the formation of the polymer-like film. With increasing applied dc bias voltage, the ion species will have sufficient energy to penetrate into the subsurface and lead to the formation of dense C–C networks, thus the film has the most diamond-like character (at an applied dc bias of about 100 V). However, at too high dc bias voltage (higher than 100 V), the dissipation of the excess heat generated by the impinging ions could relax the high compressive stress or excess density phase and lead to loose carbon networks [14,29]. Furthermore, an increase in the substrate temperature at higher applied dc bias (see Fig. 1) could also result in the formation of more graphitic components in the films. 3.5. IR absorption IR spectroscopy is widely used to characterize the C–H bonding in a-C:H films [1,30–32]. The IR absorption characteristics of the films as a function of applied dc bias voltage are shown in Fig. 5. It can be seen that the C–H absorption peaks centered at 2920 cm1,which is associated with the majority presence of hydrogen bonds in the form of sp3-CH and sp3-CH2 asymmetric stretching mode, and two smaller shoulder peaks occur at 2950 and 2860 cm1 corre-

-1

2920cm

-1

Absorbance (a.u )

368

2950cm -1

2860cm

0V 100V 200V

-1

3220cm

300V 400V

2600 2700 2800 2900 3000 3100 3200 3300 3400 -1

Wavenumber (cm ) Fig. 5. IR spectra of a-C:H films deposited at different applied dc bias voltage showing the change in the intensity of C–H absorption.

sponding to sp3-CH3 asymmetric stretching mode and sp3-CH3 symmetric model, respectively [30,31]. This suggests that most of the hydrogen is bonded to sp3hybridized C atoms. Thereafter, an overall reduction in the intensity of these modes is seen in the films with an increase in applied dc bias voltage. Since the intensity of the C–H bonding, i.e., the intensity of the C–H absorption peak in the IR spectra, is an indication of the bonded hydrogen content in the aC:H films, it is therefore deduced that the bonded hydrogen content in the films decreases with increasing applied dc bias voltage. The reduction in the bonded hydrogen content in the films may be the result of a more severe breakage of the C–H bonds in the presence of an enhanced substrate bombardment by ion species with higher energy at an increased applied dc bias [15]. It is noted that the film deposited at zero dc negative bias voltage is obviously polymeric character in IR spectra: C–H stretching modes are well resolved and characteristic water band is observed at around 3220 cm1 which may be due to water diffused into the open columnar structure of the film as was reported by Stief et al. [30]. 3.6. Mechanical properties 3.6.1. Film hardness The hardness of a-C:H films is very important to their use as protective coatings. Fig. 6 shows the changes in the film hardness as a function of the applied dc bias voltage. It is seen that the film deposited at zero

H.X. Li et al. / Applied Surface Science 227 (2004) 364–372

369

5

16 14

Critical load (N)

Hardness (GPa)

4 12 10 8 6 4

3 2 1

2

0

0 0

100 200 300 400 Applied DC bias voltage (V)

500

0

100 200 300 400 Applied DC bias voltage (V)

500

Fig. 6. Film hardness as a function of applied dc bias voltage.

Fig. 7. Critical loads of the a-C:H films as a function of applied dc bias voltage.

applied dc bias voltage is the softest with a hardness of 2.8 GPa due to the polymer-like structure caused probably by higher hydrogenation. With an increase in the applied dc bias voltage, the film hardness increases at first to a maximum value of 14.3 GPa at an applied dc bias of 100 V, and then decreases as

the applied dc bias voltage exceeds 100 V. The variation of film hardness with applied dc bias voltage is consistent with the Raman spectra seen in Fig. 4. It is worth pointing out that our result, which shows an optimum applied dc bias voltage for maximum film hardness, is in contrast to that of other reports [33,34]

Fig. 8. Friction coefficient of the films deposited at different applied dc bias voltage.

370

H.X. Li et al. / Applied Surface Science 227 (2004) 364–372

in which a continuous increase in film hardness was observed with increasing bias voltage. This can be attributed to an increase in the substrate temperature and higher ion energy amorphizing the films at higher applied dc bias voltage, as discussed in the section of Raman spectra.

ness [35,36]. Therefore, a slight decrease in critical load is observed at higher applied dc bias voltage and may be due to the decrease in film hardness and film thickness.

3.6.2. Adhesion of the film to the substrate The average critical loads of the a-C:H films as a function of applied dc bias voltage are shown in Fig. 7. It can be observed that the critical load of the film deposited at zero applied dc bias voltage is low, indicating a poor adhesion between the film and the substrate. On the other hand, the critical loads of the films deposited at various applied dc bias of 100– 500 V increase rapidly to above 3.0 N, and this improvement in adhesion is attributed to the increase in intermixing effect on the interface. Particularly, the critical load of a film depends not only on the mechanical strength (adhesion, cohesion) of the film-substrate system, but also on the film hardness and film thick-

The dynamic friction coefficient of the a-C:H films deposited at different applied dc bias voltages is shown in Fig. 8. It is clear that the film deposited at zero applied dc bias shows the highest friction coefficient in the order of 0.2 and the worst wear durability about only 30 m. Low hardness and poor adhesion of this film would be the main reason for its highest friction coefficient and the worst wear durability. A minimum and stable friction coefficient about 0.06 and a considerably increased antiwear life above 300 m are obtained for the film deposited at an applied dc bias voltage of 100 V. On the other hand, the friction coefficient of the films increases by a factor of two, from 0.06 to 0.13, with an increase in applied

3.7. Tribological behavior

Fig. 9. SEM morphologies of wear tracks of three films: (a) no bias voltage after sliding 30 m; (b) bias voltage of 100 V after sliding 300 m; and (c) bias voltage of 500 V after sliding 200 m.

H.X. Li et al. / Applied Surface Science 227 (2004) 364–372

dc bias voltage from 100 to 500 V, while the antiwear life of the films decreases from above 300 m to about 200 m. The changes in the structure and mechanical properties of the films with increasing applied dc bias voltage, which are shown in Figs. 4–7, may take important positions in controlling the film friction behavior. In order to gain more insights into the friction and wear mechanisms of the films, the worn surfaces of the a-C:H films were studied by SEM. Fig. 9 shows the scanning electron micrographs of the wear tracks for the films deposited at applied dc bias voltage of 0, 100 and 500 V, respectively. The worn surface of the film deposited at zero applied dc bias voltage shows obvious sighs of cracking and spalling at the edge of the wear track. The polymer-like structure with lower hardness and the poor adhesion between the film and substrate cause the cracking and spalling and the early failure of the film during the friction test. The worn surface of the film deposited at an applied dc bias of 100 V are much smooth in appearance and the film remains intact although it is scuffed and deformed, which conforms to the lowest friction coefficient and the highest antiwear life of the film. However, as the applied dc bias voltage increases to 500 V, the worn surface of the film shows obvious flake-like desquamation after sliding about 200 m, which may be attributed to the graphite-like structure of the film with lower hardness.

peak around 1555 cm1. Within an applied dc bias voltage range from 100 to 500 V, the film deposited at 100 V shows the maximum sp3 content and the highest hardness of 14.3 GPa. It records the smallest friction coefficient and the best wear-resistance in sliding against Si3N4 ball as well.

Acknowledgements The authors are grateful to the National Natural Science Foundation of China (Grant No. 59925513 and 5032007), the 863 Program of China (No. 2003AA305670) and ‘‘Top Hundred Talents Program’’ of Chinese Academy of Sciences for financial support. References [1] [2] [3] [4] [5] [6] [7]

[8]

4. Conclusions Hydrogenated amorphous carbon (a-C:H) films were prepared in a dual dc–rf plasma system at different applied dc bias voltages, using CH4 plus H2 gas mixture as the feedstock. Our results show that the applied dc bias voltage has a significant influence on the structure, mechanical and tribological properties of the a-C:H films. The film deposited at zero dc negative bias voltage shows no Raman signal and higher hydrogenation, indicating that the film in this case has a polymerlike structure. This polymer-like structure of the film in turn leads to much low hardness, higher friction coefficient and poor wear-resistance. Contrary to the above, the films deposited at other various dc negative bias voltages show typical diamond-like Raman spectra with a D line peak around 1360 cm1 and a G line

371

[9] [10] [11] [12] [13]

[14] [15] [16]

[17] [18]

J. Robertson, Mater. Sci. Eng. R 37 (2002) 129. J. Robertson, J. Non-Cryst. Solids 299–302 (2002) 798. A. Grill, Diamond Relat. Mater. 8 (1999) 428. D. Kim, S. Kim, Surf. Coat. Technol. 157 (2002) 66. V. Kulikovsky, P. Bohac, F. Franc, A. Deineka, V. Vorlicek, L. Jastrabik, Diamond Relat. Mater. 10 (2001) 1076. C.L. Cheng, C.T. Chia, C.C. Chiu, C.C. Wu, H.F. Cheng, I.N. Lin, Appl. Surf. Sci. 174 (2001) 251. K. Vercammen, J. Meneve, E. Dekempeneer, J. Smeets, E.W. Roberts, M.J. Eiden, Surf. Coat. Technol. 120–121 (1999) 612. K. Kuramoto, Y. Domoto, H. Hirano, S. Kiyama, S. Tsuda, Appl. Surf. Sci. 113–114 (1997) 227. J. Robertson, Diamond Relat. Mater. 3 (1994) 361. M.A. Tamor, W.C. Vassell, K.R. Carduner, Appl. Phys. Lett. 58 (1991) 592. W. Zhang, A. Tanaka, K. Wazumi, Y. Koga, Diamond Relat. Mater. 11 (2002) 1837. R.G. Lacerda, F.C. Marques, F.L. Freire Jr., Diamond Relat. Mater. 8 (1999) 495. E.H.A. Dekempeneer, R. Jacobs, J. Smeets, J. Meneve, L. Eersels, B. Blanpain, J. Roos, D.J. Oostra, Thin Solid Films 217 (1992) 56. Q. Zhang, S.F. Yoon, Rusli J. Ahn, H. Yang , J. Appl. Phys. 84 (1998) 5538. S.F. Yoon, H. Yang, A. Rusli, J. Ahn, Q. Zhang, Diamond Relat. Mater. 7 (1998) 70. A. Vanhulsel, J.P. Celis, E. Dekempeneer, J. Meneve, J. Smeets, K. Vercammen, Diamond Relat. Mater. 8 (1999) 1193. D.C. Yin, N.K. Xu, Z.T. Liu, H. Yong, X.L. Zheng, Surf. Coat. Technol. 78 (1996) 31. Y. Mokuno, A. Chayahara, Y. Horino, Y. Nishimura, Surf. Coat. Technol. 156 (2002) 328.

372

H.X. Li et al. / Applied Surface Science 227 (2004) 364–372

[19] D. Sheeja, B.K. Tay, S.P. Lau, X. Shi, Wear 249 (2001) 433. [20] S.F. Yoon, H. Yang, A. Rusli, J. Ahn, Q. Zhang, T.L. Poo, Diamond Relat. Mater. 6 (1997) 1683. [21] A. von Keudell, W. Jacob, J. Appl. Phys. 79 (1996) 1092. [22] J. Schwan, S. Ulrich, V. Batori, H. Ehrhardt, S.R.P. Silva, J. Appl. Phys. 80 (1996) 440. [23] A.C. Ferrari, J. Robertson, Phys. Rev. B 61 (2000) 14095. [24] S. Prawer, K.W. Nugent, Y. Lifshitz, G.D. Lempert, E. Grossman, J. Kulik, I. Avigal, R. Kalish, Diamond Relat. Mater. 5 (1996) 433. [25] A.C. Ferrari, B. Kleinsorge, G. Adamopoulos, J. Robertson, W.I. Milne, V. Stolojan, L.M. Brown, A. LiBassi, B.K. Tanner, J. Non-Cryst. Solids 266–269 (2000) 765. [26] M.A. Tamor, W.C. Vassel, J. Appl. Phys. 76 (1994) 3823. [27] S.R.P. Silva, G.A.J. Amartunga, J. Appl. Phys. 72 (1992) 1149.

[28] G.D. Lifshitz, Y. Lempert, E. Grossman, I. Avigal, C. UzanSaguy, R. Kalish, J. Kulik, D. Marton, J.W. Rabalais, Diamond Relat. Mater. 4 (1995) 318. [29] J. Robertson, Diamond Relat. Mater. 2 (1993) 984. [30] R. Stief, J. Scha¨ fer, J. Ristein, L. Ley, W. Beyer, J. Non-Cryst. Solids 198–200 (1996) 636. [31] J. Ristein, R.T. Stief, L. Ley, W. Beyer, J. Appl. Phys. 84 (1998) 3836. [32] W. Jacob, M. Unger, Appl. Phys. Lett. 68 (1996) 475. [33] P. Reinke, W. Jacob, W. Moller, J. Appl. Phys. 74 (1993) 1354. [34] K. Kamata, T. Inoue, K. Sugai, H. Saitoh, K. Murayama, J. Appl. Phys. 78 (1995) 1394. [35] W. Wu, M. Hon, Thin Solid Films 345 (1999) 200. [36] K.W. Xu, N.S. Hu, J.W. He, J.A. Barnard, J. Appl. Phys. 81 (1997) 5396.