Microstructure characterisation of ALD-grown epitaxial SnO2 thin films

Microstructure characterisation of ALD-grown epitaxial SnO2 thin films

ARTICLE IN PRESS Journal of Crystal Growth 260 (2004) 191–200 Microstructure characterisation of ALD-grown epitaxial SnO2 thin films J. Lua, J. Sundq...

1MB Sizes 0 Downloads 89 Views

ARTICLE IN PRESS

Journal of Crystal Growth 260 (2004) 191–200

Microstructure characterisation of ALD-grown epitaxial SnO2 thin films J. Lua, J. Sundqvistb, M. Ottossonb, A. Tarrec, A. Rosentalc, J. Aarikc, A. Ha( rstab,* b

a ( Microstructure Laboratory, The Angstr om . Laboratory, Uppsala University, Box 521, SE-751 21 Uppsala, Sweden ( The Angstr om . Laboratory, Department of Materials Chemistry, Uppsala University, Box 538, SE-751 21 Uppsala, Sweden c Institute of Physics, University of Tartu, Riia 142, 51014 Tartu, Estonia

Received 10 May 2003; accepted 14 August 2003 Communicated by D.T.J. Hurle

Abstract The microstructures of epitaxial SnO2 (rutile) thin films deposited by atomic layer deposition on a-Al2O3 (0 1 2) substrates at 600 C using either SnCl4 or SnI4 as tin precursor have been investigated by X-ray diffraction and transmission electron microscopy. It is shown that the film/substrate interface is flat without any visible additional phase, while the film structure and surface morphology depend on the metal precursor. Typical of the SnCl4-process films is a rough surface and a high density of defects, including (0 1 1) and (1 0 1) twins and anti-phase boundaries (APBs). In contrast, the SnI4-process films with their uniform thickness, flat surface and low density of APBs approach the perfect single crystal. Irrespective of the processing, all films have (1 0 1) texture, parallel to the substrate surface, with the epitaxial relationships ½0 1 0SnO2 J½1 0 0a-Al2 O3 and ½1 0 1% SnO2 J½1% 2% 1a-Al2 O3 : r 2003 Elsevier B.V. All rights reserved. PACS: 81.15.Gh; 68.55.Jk; 68.55.Ln; 61.16.Gh Keywords: A1. Planar defects; A1. Transmission electron microscopy; A3. Atomic layer epitaxy; B1. Oxides

1. Introduction Tin dioxide (SnO2) films with the rutile structure have been widely used as catalysts for CO oxidation [1], dye-sensitised solar cells [2] and semiconductor gas sensors [3] owing to their electrical, optical, electrochemical properties and *Corresponding author. Tel.: +46-18-4713723; fax: +46-18503056. E-mail address: [email protected] (A. H(arsta).

high chemical stability. Many techniques have been utilised to deposit the films, such as pulsed laser ablation [4], RF magnetron sputtering technique [5], sol–gel [6,7], chemical vapour deposition (CVD) [8–10] and atomic layer deposition (ALD) [11]. ALD is a pulsed CVD technique, by which the deposition is performed through selflimiting surface reactions from alternately supplied gaseous precursors [12]. If self-limitation is realised, the films will be uniform over large areas and have a thickness which is proportional to the

0022-0248/$ - see front matter r 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2003.08.042

ARTICLE IN PRESS J. Lu et al. / Journal of Crystal Growth 260 (2004) 191–200

192

number of completed cycles. A critical step in the application of ALD is the selection of precursors. For tin dioxide deposition, a common tin precursor is SnCl4. However, in the SnCl4 process the gaseous by-product HCl can etch the growing film which in turn might result in a rough surface. An alternative tin precursor is SnI4. It is then possible to use a hydrogen-free oxygen precursor, e.g., O2, which eliminates the risk for formation of a hydrogen halide. Furthermore, SnI4 is thermally less stable than SnCl4, which should make it possible to reduce the deposition temperature. SnO2 with the rutile structure can be epitaxially grown on a-Al2O3 (0 1 2) substrates. In most cases, the resulting films had many defects, such as twin boundaries and stacking faults. The defects strongly affect the electrical transport in the films [13]. The deleterious effects can be reduced by fabricating less defective films, ultimately films having single-crystalline structure. Dominguez et al. [4] have successfully synthesised singlecrystalline SnO2 films using the femtosecond pulsed laser deposition technique. However, numerous (1 0 1) anti-phase boundaries (APBs) were present in these films. Such a structure may be inadmissible in some applications. Recently, Sundqvist et al. [14] successfully grew SnO2 films by ALD using the SnI4–O2 precursor combination. For ultrathin films, these SnO2 films were sensitive to CO in air [15]. In the present paper, structural characterisation by X-ray diffraction (XRD) and transmission electron microscopy (TEM) has been carried out for SnO2 films prepared by ALD using four different precursor combinations, SnCl4–H2O, SnCl4–H2O2, SnI4–H2O2 and SnI4–O2. The aim is to in detail investigate the microstructure and morphology of the films and also to correlate the film structure to the deposition process employed.

2. Experimental procedure The film depositions were carried out in two different flow-type low-pressure ALD reactors, A [16] and B [14]. Reactor A was used to deposit samples 1–3 and reactor B to deposit sample 4 and the experimental parameters are listed in Table 1. Three of these ALD processes used have been described in more detail elsewhere, SnCl4–H2O2 [17], SnCl4–H2O [18] and SnI4–O2 [14], but not the SnI4–H2O2 process. All films were deposited at a deposition temperature of 600 C on a-Al2O3 (0 1 2) substrates, which prior to deposition were ultrasonically cleaned in methanol. The pulse times, t1 2t2 2t3 2t4 ; where t1 2t4 denote the duration of the tin precursor pulse, the first purge pulse, the oxygen precursor pulse and the second purge pulse, respectively, were 2–2–2–2 s for samples 1–3 and 4–6–4–6 for sample 4. A Philips X’Pert Pro MRD diffractometer using CuKa radiation with a parallel beam set-up was used to record the pole figures. The cross-sectional specimens for TEM examination were prepared as follows. The substrates with deposits were cut into 3  5 mm2 pieces with a low-speed diamond saw. Pairs of pieces with the deposits face to face were bonded together using epoxy, polished into a rod with 2.5 mm diameter and subsequently inserted into a brass tube. The rod was cut to a 0.5-mm thick slice with a diamond saw. At last, the structures were ground from both sides to a thickness of 0.1 mm, dimpled at the centre to 10 mm thickness and ion-milled to electron transparency. The high-resolution TEM (HRTEM) characterisation was carried out using a field emission gun TECNAI F30 ST operated at ( The 300 kV with a point resolution of 2.05 A. low-magnification images and selected area electron diffraction (SAED) patterns were obtained

Table 1 ALD process parameters and the resulting film thicknesses, obtained from TEM cross sections Sample no.

Tin precursor

Tvap ( C)

Ptot (Torr)

Oxygen precursor

No. of cycles

Thickness (nm)

1 2 3 4

SnCl4 SnCl4 SnI4 SnI4

0 0 84 115

1.8 2.4 1.9 10

H2O2 H2O H2O2 O2

3000 3000 1000 1000

93 117 41 117

ARTICLE IN PRESS J. Lu et al. / Journal of Crystal Growth 260 (2004) 191–200

using a JEOL2000FXII at a working voltage of 200 kV.

3. Results 3.1. X-ray diffraction results The SnO2 {1 0 1} pole figures for the films deposited from the SnI4–O2 and the SnCl4–H2O2 processes are shown in Fig. 1. The pole figures for the films deposited using the other two precursor combinations are not shown, since the pole figure

193

from the SnI4–H2O2 film is similar to the SnI4–O2 film and the pole figure from the SnCl4–H2O film is similar to the SnCl4–H2O2 film. For the film deposited using the SnI4–O2 process, it can be seen that only reflections from the {1 0 1} set of planes appear (Fig. 1a). The corresponding in-planar orientational relationship is ½0 1 0 SnO2J½1 0 0a-Al2 O3 and ½1 0 1% SnO2J½1% 2% 1a-Al2 O3 : In contrast, for the film deposited using the SnCl4–H2O2 process, additional reflections can be seen (Fig. 1b). The main in-planar orientational relationship is still the same, which can be seen from the intensities since they are more than 10 times

Fig. 1. {1 0 1} pole figures for films deposited from the SnI4–O2 (a) and the SnCl4–H2O2 (b) processes.

ARTICLE IN PRESS 194

J. Lu et al. / Journal of Crystal Growth 260 (2004) 191–200

higher than for any other orientations. The strongest additional contribution (denoted T1 in Fig. 1) is from the 180 rotation of the main (1 0 1) orientation, corresponding to the ½0 1% 0SnO2 J½1 0 0a-Al2 O3 and ½1% 0 1SnO2 J½1% 2% 1a-Al2 O3 orientation. There are also contributions from twins with (0 1 1) and ð0 1% 1Þ as the twinning planes (denoted T2, see TEM results). Finally, a weak contribution from a tilt of about 46.5 in the ½1% 0 1 direction (denoted T3 in Fig. 1) can also be seen. 3.2. TEM results 3.2.1. SnO2 films from the SnCl4 process Figs. 2a and b are TEM bright field images of the SnCl4-process samples (No. 1–2 in Table 1) made from two different oxygen precursors, H2O2 and H2O. The films have a flat interface, a rough

surface and a high density of defects. As can be seen from the micrograph, a common defect is a planar defect with about 45 inclining angle to the substrate surface. The electron diffraction investigation demonstrated that the films consisted of only the SnO2 rutile phase (see Fig. 3a) with the (1 0 1) plane parallel to the substrate surface. The planar defect is a twin with (0 1 1) as twinning plane (Fig. 3b). Since the (1 0 1) plane is parallel to the substrate surface, the angle between the twinning plane and the substrate surface is the same as the angle between the (0 1 1) and (1 0 1) planes, which can be accurately calculated from the following formula: cos f ¼

ðh1 h2 þ k1 k2 Þ=a2 þ l1 l2 =c2 f½ðh21 þ k12 Þ=a2 þ l12 =c2  ½ðh22 þ k22 Þ=a2 þ l22 =c2 g1=2

Fig. 2. TEM images of SnO2 rutile films: (a) sample no. 1, and (b) sample no. 2.

Fig. 3. SAED patterns: (a) SnO2 film matrix, (b) (0 1 1) twin, and (c) (1 0 1) twin.

;

ARTICLE IN PRESS J. Lu et al. / Journal of Crystal Growth 260 (2004) 191–200

[19] where f is the angle between the planes (h1 k1 l1 ) and (h2 k2 l2 ), a and c are the cell parameters of SnO2. The angle between (1 0 1) and (0 1 1) was calculated to be 46.5 , which agreed well with the result measured directly from the micrographs. This (0 1 1) twin can also be seen in the {1 0 1} pole figure (T2 in Fig. 1b). In addition to the (0 1 1) twin, another defect namely a (1 0 1) twin indicated by white lines in Figs. 1 was also observed. The SAED pattern (Fig. 3c) showed that the twinning plane (1 0 1) is parallel to the film surface. The structural relationship between the (1 0 1) twin and the film matrix is a 180 rotation in the plane of the substrate surface. As discussed above, this (1 0 1) twin is also seen in the pole figure (T1 in Fig. 1b). In addition to the growth twins, stacking faults (or so-called APBs which will be discussed in the next section) have also been found in both films. It can thus be concluded that the films made from both the SnCl4–H2O2 and SnCl4–H2O processes have similar mediocre-quality microstructure and morphology. Figs. 4a and b are HRTEM images of film– substrate interfaces viewed along the ½1 1 1SnO2 direction. The blur lattice image in the a-Al2O3 (0 1 2) plane can be explained by the deviation in angle between the ½0 1 2a-Al2 O3 direction and the ½1 1 1SnO2 direction. There is no visible additional phase in the sharp interfaces. Fig. 5 shows the atomic structure of the (0 1 1) twin viewed in the [1 1 1] direction. The straight twin boundary is coherent consisting of Sn atoms shared by the two neighbouring components of the twin, which has been described earlier in the literature [20]. It was reported that the Sn sublattice is mirror symmetric with respect to the coherent twin boundary (CTB), while the oxygen atoms forming octahedral cages around the Sn atoms are not mirror symmetric to the CTB. The atomic structure of the (1 0 1) twin boundary has also been characterised by HRTEM. The (1 0 1) twin boundary is curved and incoherent (see Fig. 6). The pseudo-superstructure (about three times the (1 0 1) spacing) in the twin boundary could be a moire! pattern caused by overlap of the twin and the film matrix. To confirm this, fast fourier transformation (FFT) analysis was performed on the moire! pattern, which exhibited the

195

Fig. 4. HRTEM images of the SnO2/substrate interfaces of: (a) sample no. 1, and (b) sample no. 2.

Fig. 5. HRTEM image of a coherent (0 1 1) twin boundary from sample no. 1.

same pattern (inset in Fig. 6) as the SAED pattern of the (1 0 1) twin. No superstructure reflections can be seen in the pattern.

ARTICLE IN PRESS 196

J. Lu et al. / Journal of Crystal Growth 260 (2004) 191–200

surface of the as-grown film and therefore no straight CTBs could be obtained. Above the (1 0 1) twin II, there is a stacking fault namely an APB. Under the (1 0 1) twin II, a (0 1 1) twin initiated from the substrate surface and terminated in the film. It is obvious that the (1 0 1) twin II has hindered the growth of the (0 1 1) twin. As can be seen in the inset of Fig. 7, a twin boundary jumped from one Sn atomic plane to another plane and created a step (marked by a dashed line). The step height is usually not larger than two metal sublattice spacings of the (0 1 1) plane. In the same inset, the twin dislocations have been marked by arrows.

Fig. 6. HRTEM image of a curved (1 0 1) twinning boundary from sample no. 1. The inset is an FFT pattern taken from the twinning boundary region.

The stress of a thin film mainly exists in the region near the film–substrate interface. Thus the majority of the stress is released by generating a number of defects close to the interface [21]. In film no. 2, several defects, two (1 0 1) twins, one (0 1 1) twin and a stacking fault (APB) were found in the vicinity of the interface (see Fig. 7). Both straight and curved (1 0 1) twinning boundaries are observed in this image. A (1 0 1) twin I nucleated and grew coherently from a thin matrix. The straight CTBs of the (1 0 1) twin are parallel to the substrate surface. Another (1 0 1) twin II at the centre of the figure is completely surrounded by the SnO2 matrix with curved twinning boundaries. This (1 0 1) twin nucleated directly from the rough

3.2.2. SnO2 films from the SnI4 process Figs. 8a and b are micrographs of SnO2 rutile films (No. 3–4 in Table 1) deposited from the SnI4– H2O2 and SnI4–O2 processes. These films are clearly different from the SnCl4-process films although all films have the same epitaxial relationships with the a-Al2O3 substrate. Firstly, the SnI4process films have a uniform thickness, a flat surface and a low density of defects. The difference in surface morphology between the films from the different halides can be clearly seen in the HREM images (Figs. 9a and b). The rough surface of the SnCl4-process film is facetted not only with the (1 0 1) plane but also with other planes such as the (0 1 1) plane and consequently forms surface stages. A nucleus of a (1 0 1) twin was also found on the rough surface. In contrast, the SnI4-process film has a flat surface without any stages. Furthermore, the SnI4-process films are seemingly single crystalline without any twin structure. The appearance of contrast in the films is here caused by stress. The SnO2/a-Al2O3 interface of the films (Figs. 10a and b) is also sharp without any additional phase. A few stacking faults (APBs) in the (0 1 1) plane were observed and the APBs could initiate from the substrate surface (Fig. 10a). In Fig. 11, an APB initiating from inside the film is shown, which most probably is caused by stress. It can be mentioned that the density of APBs in the SnI4process films is much lower than that in films synthesised by femtosecond pulsed laser deposition [4]. Finally, no significant difference in the

ARTICLE IN PRESS J. Lu et al. / Journal of Crystal Growth 260 (2004) 191–200

197

Fig. 7. HRTEM image from the vicinity of the SnO2/substrate interface. In the inset, a step is marked by a dashed line through the twin boundary and three twin dislocations are indicated by arrows.

Fig. 8. TEM images of SnO2 rutile films: (a) sample no. 3, and (b) sample no. 4.

structure and morphology has been found between the films made by the two different oxygen precursors except for the film thickness. This difference was caused by the different deposition conditions employed (see Table 1).

4. Discussion There are several types of twin in the rutile structure: {3 0 1}, {1 2 1} and {0 1 1} twins [22]. The (0 1 1) and (1 0 1) twins found in this study belong

ARTICLE IN PRESS 198

J. Lu et al. / Journal of Crystal Growth 260 (2004) 191–200

Fig. 9. (a) HRTEM image of a rough surface of a SnCl4process film containing a stage facetted with a (0 1 1) plane and a (1 0 1) twin nucleus, and (b) HRTEM image of a flat surface of a SnI4-process film.

to the same {1 0 1} twin but with different twinning plane orientations. The (0 1 1) twinning plane has an angle of 46.5 inclined to the substrate surface while the (1 0 1) twinning plane is parallel to the film surface, and thus leads to that the two twins grow in different orientations. The different growth orientations yield completely different twin morphologies. The (0 1 1) twin is slat-like with a straight coherent twin boundary (TB) while the (1 0 1) twin is generally an equiaxed grain with incoherent TBs. The (0 1 1) twin is a common defect in both SnO2 and TiO2 films, while the (1 0 1) twin is hardly found in TiO2 films [23]. This can be due to the fact that the rectangular in-plane unit cells of these films have different dimensions (cf. [24,25]). In Table 2, the periodicities (d) of the rectangular repeating unit cells of the three materials are listed together with the calculated directional lattice mismatches. In the TiO2/aAl2O3 system, the mismatches in the two directions

Fig. 10. HRTEM images of the SnO2/substrate interfaces of (a) sample no. 3, and (b) sample no. 4.

are comparable. Epitaxial growth of TiO2 film on a-Al2O3 is thus determined by the lattice fits in both the ½1 0 1%  and [0 1 0] directions. In the SnO2/a-Al2O3 system, however, the mismatch in [0 1 0]rutile is much smaller than the mismatch in ½1 0 1% rutile : In the [0 1 0]rutile direction, oxygen atoms continues across the SnO2/a-Al2O3 interface with low deformation energy, while in the ½1 0 1% rutile direction, the lattice fit between SnO2 and a-Al2O3 is considerably worse. The epitaxial

ARTICLE IN PRESS J. Lu et al. / Journal of Crystal Growth 260 (2004) 191–200

199

Fig. 11. HRTEM image of an APB initiating from the inside of the film and terminating at the film surface.

Table 2 ( of the rectangular in-plane unit cells and the corresponding mismatches Edge lengths d (A)

a-Al2O3 TiO2 SnO2

( in ½1 0 1%  d (A) film or ½1% 2% 1a-Al2 O3

( in [0 1 0]film or d (A) ½1 0 0a-Al2 O3

5.12 5.46 5.71

4.76 4.59 4.74

growth of SnO2 on the a-Al2O3 (0 1 2) substrate thus possesses a one-dimensional feature which is dominating. The (1 0 1) twin grain has the same epitaxial relationship in the [0 1 0]rutile direction as the film matrix and is accordingly favoured. The experimental results discussed above show that the morphology and microstructure of the films are not sensitive to the oxygen precursors but strongly dependent on the halide precursors. It is conceivable that the rough surface obtained in the SnCl4 process can be attributed to surface etching by HCl in the SnCl4 process. The HRTEM image (Fig. 9a) clearly illuminates a stage facetted by (0 1 1) planes on the rough surface. A (0 1 1) twin could thus grow epitaxially on the (0 1 1) surface. The appearance of a (1 0 1) twin nucleus in Fig. 9a demonstrates that the rough surface is also

Mismatch of ½1 0 1% film =½1% 2% 1a-Al2 O3 (%)

Mismatch of ½0 1 0film =½1 0 0a-Al2 O3 (%)

6.64 11.52

3.57 0.42

beneficial for the nucleation of a (1 0 1) twin. In contrast, in the SnI4 process no such etching takes place and a uniform twin-free single-crystalline film will be grown on the flat surface. The APBs lie on the (0 1 1) planes with a displacement vector of 12[0 1 1], and if the APBs are terminated in the film partial dislocations with the Burger’s vector 12[a 0 c] will be created. The APBs can be caused by stress in the film or induced by the rough substrate surface. Fig. 10a illustrates an example of the later nucleation mechanism. The shift direction is in the shear plane and thus the stoichiometry is conserved if the APBs terminate on the film surface. However, if an APB terminates in the film, non-stoichiometry would be created by forming a partial dislocation. The partial dislocation contributes to the formation of oxygen

ARTICLE IN PRESS 200

J. Lu et al. / Journal of Crystal Growth 260 (2004) 191–200

vacancies and is beneficial for the film conductivity. On the other hand, the grain boundaries in the SnO2 films can trap electrons, form back-to-back Schottky barriers and hence deteriorate the conductivity. Compared with the single-crystalline film made by femtosecond pulsed laser deposition [4], the SnI4-process films contain less APBs, which may explain the observed rather high conductivity of these films [15]. 5. Summary and concluding remarks ALD SnO2 rutile films deposited on (0 1 2) a-Al2O3 substrates from both SnCl4 and SnI4 processes have been characterised by XRD and TEM. All films grew epitaxially on (0 1 2) a-Al2O3 substrates with a (1 0 1) texture. The epitaxial relationships between the films and the substrate are ½0 1 0SnO2 J½1 0 0a-Al2 O3 and ½1 0 1% SnO2 J½1% 2% 1a-Al2 O3 ; with the corresponding mismatches of 0.42% and 11.52% along the ½0 1 0SnO2 and the ½1 0 1% SnO2 directions, respectively. The halide precursor used dramatically affected the film morphology and microstructure. The SnCl4-process films had a rough surface, a non-uniform thickness and contained a high density of defects: growth twins, stacking faults (APBs) and dislocations. In contrast, the SnI4process films approach the perfect single crystal with uniform thickness, flat surface and a low density of APBs. The high quality of the singlecrystal films is likely to contribute to the rather high conductivity of these films. Acknowledgements The authors thank A. Aidla and A.-A. Kiisler for assistance in the growth experiments. This work has been supported in part by the Centre for Advanced Micro Engineering (AME), The ( . Laboratory and the Estonian Science Angstr om Foundation (Grant No. 4205).

References . ’ I’s-li, Z.I. ’ Onsan, [1] N. Akin, G. Kilaz, A.I. Chem. Eng. Sci. 56 (2001) 881. [2] Y. Tachibana, K. Hara, S. Takano, K. Sayama, H. Arakawa, Chem. Phys. Lett. 364 (2002) 297. [3] U. Hoefer, J. Frank, M. Fleischer, Sensors Actuators B 78 (2001) 6. [4] J.E. Dominguez, L. Fu, X.Q. Pan, Appl. Phys. Lett. 81 (2002) 5168. [5] D.H. Kim, S.H. Lee, K.H. Kim, Sensors Actuators B 77 (2001) 427. [6] Y. Ohya, K. Horinouchi, T. Ban, Y. Takahashi, N. Murayama, J. Ceram. Soc. Jpn. 110 (2002) 950. [7] W.P. Tai, J.H. Oh, Sensors Actuators B 85 (2002) 154. [8] P. Montmeat, C. Pijolat, G. Tournier, J.P. Viricelle, Sensors Actuators B 84 (2002) 148. [9] K.M. Chi, C.C. Lin, Y.H. Lu, J.H. Liao, J Chin. Chem. Soc. Taip. 47 (2000) 425. [10] Y. Okajima, T. Ide, K. Kikuchi, K.I. Nakamura, Sensors Mater. 10 (1998) 113. . Thin Solid Films 249 (1994) [11] H. Viirola, L. Niinisto, 144. [12] T. Suntola, Mater. Sci. Rep. 4 (1989) 261. [13] X.Q. Pan, L. Fu, J.E. Dominguez, J. Appl. Phys. 89 (2001) 6056. [14] J. Sundqvist, A. Tarre, A. Rosental, A. H(arsta, Chem. Vapor Deposition 9 (2003) 21. [15] A. Rosental, A. Tarre, A. Gerst, J. Sundqvist, A. H(arsta, A. Aidla, J. Aarik, V. Sammelselg, T. Uustare, Sensors Actuators B 93 (2003) 552. [16] J. Aarik, A. Aidla, H. M.andar, V. Sammelselg, J. Crystal Growth 340 (1999) 110. [17] A. Tarre, A. Rosental, A. Aidla, J. Aarik, J. Sundqvist, A. H(arsta, Vacuum 67 (2002) 571. [18] A. Rosental, A. Tarre, A. Gerst, T. Uustare, V. Sammelselg, Sensors Actuators B 77 (2001) 297. [19] D.B. Williams, C.B. Carter, Transmission Electron Microscopy, Plenum Press, New York, 1996. [20] J.Y. Huang, B.H. Park, D. Jan, X.Q. Pan, Y.T. Zhu, Q.X. Jia, Philos. Mag. A 82 (2002) 735. [21] K. Suzuki, M. Ichihara, S. Takeuchi, Philos. Mag. A 63 (1991) 657. [22] J.G. Zheng, X.Q. Pan, M. Schweizer, F. Zhou, U. Weimar, . W. Gopel, M. Ruhle, J. Appl. Phys. 79 (1996) 7688. [23] M. Schuisky, K. Kukli, J. Aarik, J. Lu, A. H(arsta, J. Crystal Growth 235 (2002) 293. [24] M. Schuisky, A. H(arsta, A. Aidla, K. Kukli, A.-A. Kiisler, J. Aarik, J. Electrochem. Soc. 147 (2000) 3319. [25] A. Tarre, A. Rosental, J. Sundqvist, A. H(arsta, T. Uustare, V. Sammelselg, Surf. Sci. 532–535 (2003) 514.