Contact us

My IOPscience

Silicon oxide particle formation in RF plasmas investigated by infrared absorption
spectroscopy and mass spectrometry

This content has been downloaded from IOPscience. Please scroll down to see the full text.
1998 J. Phys. D: Appl. Phys. 31 74
View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address:
This content was downloaded on 10/05/2014 at 14:26

Please note that terms and conditions apply.

Switzerland § Eindhoven University of Technology. A small proportion of silane gas produces nanometric stoichiometric particles whereas a large proportion produces larger under-stoichiometric particles. CH-1015 Lausanne. Introduction Deposition of thin films of silicon dioxide is widely used in semiconductor device technology. Ecole Polytechnique Fed de Lausanne. C Courteille†. Post-oxidation of hydrogenated silicon particles trapped in an argon plasma was obtained by adding oxygen. Experimental oxidation studies are usually performed on films. but only a few investigations on oxidized silicon particles [5–7] have been reported. was positioned at the reactor wall behind a diffusor plate minimizing the influence on the powder dynamics by gas drag. S M Scholz‡. The gas inlet. 31 (1998) 74–84. Powder grown in an argon–silane plasma and then trapped in a pure argon discharge can be postc 1998 IOP Publishing Ltd 0022-3727/98/010074+11$19. Otherwise. The composition of SiOx particles produced in various argon–silane–oxygen plasmas will be discussed. MX-D-Ecublens. F-38402 Saint Martin d’Heres Cedex. Absorption spectroscopy was sufficiently sensitive to reveal particles too small to be visually observed by laser light scattering. Phys. 1. G M W Kroesen§. The second part of this paper discusses the composition of the initial SiOx ionic clusters produced in a silane–oxygen plasma as measured by mass spectrometry. Mass spectrometry of negative and positive ions showed an extensive range of ionic clusters which may be at the origin of the observed particle formation. PPB-Ecublens. D: Appl. Printed in the UK PII: S0022-3727(98)86665-5 Silicon oxide particle formation in RF plasmas investigated by infrared absorption spectroscopy and mass spectrometry Ch Hollenstein†.56 MHz with 130 mm diameter electrodes. The distance between the stainless steel electrodes was 26 mm. A model based on an iterative reaction sequence gives a good agreement with the measured positive ion mass spectrum. Phys. 9] and experimental studies [10–12] have demonstrated the possible role of plasma species such as negative ions in the oxidation process. A A Howling†. Universite´ Joseph Fourier-Grenoble 1. In situ Fourier transform infrared absorption spectroscopy has been used to study the composition of particles formed and suspended in radio-frequency discharges of silane–oxygen-argon gas mixtures. France Received 7 August 1997. the oxygen was injected directly into the reactor through a calibrated . The silane gas consumption was observed by infrared absorption. Many theoretical [8. for argon and silane. D Magni†. The Netherlands ´ k Laboratoire de Spectrometrie Physique. the existence of particles in the plasma is generally ignored. The experimental arrangement The experiment was performed in a RF capacitively coupled plasma at 13. In the first part of this paper we demonstrate the in situ monitoring of nanometre sized oxidized silicon particles by infrared absorption spectroscopy measurements of powder suspended in a RF plasma. Switzerland ´ erale ´ ‡ Laboratoire de Technologie des Poudres. W de Zeeuw§ and W Schwarzenbachk ´ erale ´ † Centre de Recherches en Physique des Plasmas. Ecole Polytechnique Fed de Lausanne. although they may play an important role in the composition and quality of the deposited silicon oxide. It has been studied intensively during the last 30 years. To avoid reaction of silane with oxygen in the gas line. in final form 19 September 1997 Abstract.50 oxidized by admixture of oxygen to this argon plasma. 140 Avenue de la Physique. with a considerable effort to model the fundamental oxidation processes [1–4]. The stoichiometry of the produced particles depends on the silane flow rate and was compared with commercial colloidal silica. CH-1015 Lausanne. 2. N Simons§. BP 87.J. 5600 MB Eindhoven.

In order to investigate the influence of the gas mixture on the powder composition. which tends to favour powder production in silane plasmas [13]. For the in situ infrared absorption measurements. ZnSe windows were mounted on diametrically opposite flanges. needle valve located at the top of the reactor. All the experiments were carried out with a RF power of 10 W and a pressure of 0.20 mbar controlled by a throttle valve. FTIR measurements Two types of in situ oxidation experiments will be described: (i) production of oxidized silicon powder in argon– silane–oxygen plasmas and (ii) post-oxidation of silicon powder trapped in an argon plasma. the measurements were performed by averaging over 100 spectra. A constant total flow was maintained by adjusting the argon flow. a Balzers Plasma Process Monitor type PPM 421 was used to detect neutral radicals and ions (positive and negative) with a mass range up to 2000 amu. A schematic top view of the experimental arrangement. The reactor was kept at room temperature. The parallel beam leaving the spectrometer was directed through the plasma (single pass) and was focused by an off-axis parabolic gold mirror onto a liquid-nitrogen-cooled mercury–cadmium–telluride (MCT) detector (Graseby type FTIR-W24).1% of the source intensity. in the range 2000–2300 cm−1 . through the gas and the 75 . the FTIR measurement. The Fourier transform of the detector signal yielded the spectrum in the 600– 4500 cm−1 range with a chosen spectral resolution of 1 cm−1 . The extraction orifice used for these experiments was of 100 µm diameter. The infrared absorption measurements were performed under steady state conditions of the powder production as monitored by laser light scattering. The minimum absorption which could be detected by the system was of the order of 0. showing the plasma reactor. Reference spectra with the plasma off were acquired before and after each measurement to check for the possible influence of deposition on the windows and for variations in the source intensity or the detector sensitivity. To obtain a reasonable signal-to-noise ratio. The spectra were measured using a Bruker IFS-66 spectrometer with a glow bar source and an external detector. Figure 2 shows the infrared transmitted signal. Light scattering of a He–Ne laser beam was used to visualize the powder produced in the plasma. 3. For the mass spectrometer measurements.Particle formation in RF plasmas Figure 1. The reactor was described in more detail in [14]. (i) Oxidized silicon powder was produced in diluted argon–silane–oxygen discharges. the mass spectrometer measurement and the He–Ne laser scattering diagnostic. the silane gas flow was altered while the oxygen flow was kept unchanged. The probe head was positioned adjacent to the electrode gap 10 mm from the plasma boundary. Visual observation was through the window perpendicular to the laser beam path (figure 1).

7 sccm of oxygen. different peaks in figure 3(a) originating from Si—H vibrations (peaks A and C) and from Si—O excitation (peaks B.1 sccm of silane plus 7.56 MHz. powder was visually observed using the He–Ne laser light scattering system (particle size >λ/10 ' 70 nm). from the presence of broad 76 features in the infrared absorption spectrum of figure 3(b). Referring to table 1 (for a non-exhaustive list of possible mode assignments).2 mbar and RF plasma power 10 W at 13. it must be concluded that powder was nevertheless present in the plasma. but with a particulate size below the detection limit ('70 nm) of the He–Ne laser scattering system used. However. Large differences between these absorption spectra were observed.Ch Hollenstein et al Figure 2. The spectra show the typical absorption spectrum of silane (2050–2300 cm−1 ). in the range 600–1400 cm−1 .2 sccm of argon plus 3. using the two previous silane to oxygen flow rates ratios corresponding to the figure 2. for two different argon–silane–oxygen gas mixtures: (a) 35. only Si—O features appear (peaks E and F). whereas. Infrared transmission spectra recorded through the gas (subscript 1) and through the plasma. plasma using two different silane:oxygen:argon flow rates. in figure 3(b). under steady state conditions (superscript 2). The absorption spectroscopy technique is therefore sufficiently sensitive to reveal particles too small to be visually observed using laser light scattering. The plasma parameters were pressure 0. no powder could be observed in the discharge.4. D and E) are present. From table 1 the large peak centred near 1085 cm−1 in figure 3(b) seems to reveal the presence of stoichiometric (SiO2 ) powder whereas in figure 3(a) the broad peak (E) indicates an under- . A small amount of silane can be totally consumed by the oxygen.3 sccm of argon plus 17 sccm of silane plus 7. The differences among the IR spectra demonstrate the influence of the silane:oxygen ratio through the amount of silane gas consumed in the process. whereas at low silane flow rates (figure 3(b)). The reactor was at room temperature. With a high silane flow (figure 3(a)). Figures 3(a) and (b) show the measured absorbance spectra.7 sccm of oxygen and (b) 49. With decreasing silane:oxygen ratio the obvious silane absorption finishes in complete depletion of the silane gas in the discharge at a flow ratio of 0.

respectively). For comparison the absorption spectrum of a commercial colloidal silica (Ludox LS) is shown in figure 3(c). hydrogenated silicon powder was initially produced in a 4% silane-inargon (48 sccm of argon and 2 sccm of silane) discharge. According to Hu [15] the (LO) mode becomes active for particles smaller than 360 nm. 77 . (ii) For the post-oxidation experiments.Particle formation in RF plasmas Figure 2. The similarity between the commercial SiO2 powder and the plasma-produced powder (figure 3(b)) demonstrates that nanoscale stoichiometric silica particles can be produced within a plasma. (Continued) stoichiometric powder composition. Hu [15] attributed the peak G at 1230 cm−1 to the longitudinal optical (LO) phonon mode of SiO2 . under-stoichiometric powder (figure 2(a)) was produced in a plasma containing a significant amount of undissociated silane whereas stoichiometric nanoparticulates were produced in a silanedepleted plasma (figure 2(c)). Micrometrics) and by a particle diameter of 10 nm according to dynamic light diffusion (Malvern Zetasizer 4). The negative peaks in the range 900–950 cm−1 are silane depletion gas features. the infrared spectrum was obtained using a Nicolet 510 FTIR spectrometer in transmission mode with a KBr lozenge used as mechanical support for the SiO2 powder. the composition of the resulting particle is therefore influenced by the silane-to-oxygen flow ratio. For this specific powder. but can become active for platelet particles smaller than λ/(2π n) (where λ and n are the wavelength of the incident radiation and the refractive index of the embedding matrix. This size-calibrated SiO2 powder is characterized by a specific surface area of 200 m2 g−1 as measured by gas absorption (using the BET method. This can explain why the G peak is present in the nanoscale Ludox powder spectra and under the plasma conditions of a low silane flow rate but not in figure 3(a). which is normally infrared inactive. To summarize. As summarized in table 1. for which visible particulates are probably larger than the threshold diameter limit.

all of them lead to the conclusion that the powder produced was only composed of amorphous hydrogenated silicon. The continuous increase of the absorbance signal for wavenumbers above 1400 cm−1 . of particles created in two different gas mixtures: (a) 35. which can be entirely attributed to the oxidation of the hydrogenated silicon powders. Despite there being various possible attributions for the absorption peaks in the range 600–900 cm−1 (see table 1). After steady state production of the hydrogenated silicon powder had been achieved. in the 600–2000 cm−1 range.7 sccm of oxygen and (b) 49. The absorbance signal of the previously trapped powder changed considerably when oxygen was added to the argon plasma (figure 4(b)). oxygen was added to this plasma with a typical flow rate of 7. The absorbance spectra. The spectrum for the powder trapped in the argon plasma (figure 4(a)) shows three distinct peaks.3 sccm of argon plus 17 sccm of silane plus 7.2 sccm or argon plus 3. Finally.1 sccm of silane plus 7. of powder trapped in the pure argon plasma before (figure 4(a)) and after (figure 4(b)) the addition of oxygen. the oxidation of the powder found in their experiment was probably due to a leak or to oxygen originating from water from the reactor chamber. no significant scattering contribution was found since figure 4(b) shows . However. in contrast to the results of Kroesen et al [5]. resulting in the trapping of some powder in the remaining pure argon plasma. The plasma parameters were the same as those for figure 2. No Si—O-related vibrations were found in spectra of this 78 trapped powder.7 sccm.Ch Hollenstein et al Figure 3. as monitored by laser scattering. under steady state conditions.7 sccm of oxygen. in contrast to figure 4(a). originated from Rayleigh scattering of the infrared beam by the particles suspended in the plasma. in figure 4(a). The original peaks were replaced by a broad peak at 950–1250 cm−1 . Figure 4 represents the steady state absorbance signals. who saw oxidation features in infrared absorption spectra of powder produced under similar conditions. the silane gas feed was stopped. (c) The FTIR absorbance spectrum of ‘Ludox LS’ colloidal SiO2 powder.

Peak Wavenumber (cm−1 ) Mode assignments References A 630–720 [20–28] B 730–830 SiH bending. SiH3 rocking (SiH2 )n shear and rocking 730 cm−1 rocking and 830 cm−1 bending mode of O2 Si—H2 780 cm−1 is a Si—H (bending motion perpendicular to Si—H bond) + Si—O—Si (bending + small stretching). 24] [30] [4] [27] [20. Triple-SiH rocking Si—H bending in SiH2 and SiH3 (SiH2 )n wagging and scissors + peak at 2090 cm−1 : SiH2 isolated stretching Peak arising from structural combination of Si—(O2 Si2 ). 21. 24] [33.Particle formation in RF plasmas Table 1. 79 . O and H bonded to common Si site 800 cm−1 : O—Si—O bending 814 cm−1 : Si—O—Si bending of a thermal SiO2 layer —(SiH2 )n wagging. 32] [23. —(SiH2 )n wagging. The absorbance spectra of silicon powder formed in an argon (48 sccm)–silane (2 sccm) plasma and trapped in discharges of (a) 48 sccm of argon and (b) 48 sccm of argon and 7. Various possible peak assignments. Si—(O3 Si) and Si—(O4 ) Relationship between IR absorption peak position and the stoichiometry of SiOx (x = 2 at 1085 cm−1 ) 940 cm−1 : isolated Si—O—Si stretching and 980 cm−1 : oxygen asymmetrical stretching of Si—O—Si in cis and trans configurations and 1030 cm−1 : Si—O—Si stretching mode in which the silicon atoms are bonded to other oxygen atoms 1042 + 870 + 800 cm−1 : Si2 O3 layer feature (intermediate between SiO and SiO2 ) SiO surface mode Longitudinal optical (LO) phonon mode of SiO2 only active for platelet particles smaller than λ/(2π n ) C 835 and 843 D 850 and 890 875 E 920–1085 F G 1150 1230 [29] [23.7 sccm of oxygen. 28] [26] [23] [31] [31] [31. The plasma parameters were the same as those for figure 2. wagging SiH2 . O—Si—H cis configuration. rocking. 30] [34] [15] Figure 4.

An oxidation post-treatment of trapped hydrogenated silicon powder by addition of oxygen to an argon plasma has thereby been demonstrated. The Balzers mass spectrometry measurements. Two possible causes are proposed to explain the loss of scattering signal caused by the addition of the small amount of oxygen. mass spectrometry has been used to investigate the plasma composition during production of nanometre sized particles . are uncorrected for any mass-dependent falling off in sensitivity. under steady state conditions. First. a flat absorbance signal in the range above 1400 cm−1 . 4. powder detrapping due to changes in the plasma potential topography caused by modifications in the plasma composition [16]. Mass spectrometry measurements In addition to infrared absorption spectroscopy.1 mbar at RF frequency 13. The ground electrode was heated to 200 ◦ C. the positive ions and the neutral species in pure silane plasma respectively. The plasma parameters were pressure 0. Here (a).Ch Hollenstein et al Figure 5. for a modulated plasma with a running time (plasma on) of 110 µs and a pause (plasma off) of 300 µs. acquired with low mass resolution.56 MHz. powder detrapping due to additional gas drag. and secondly. 80 but further experiments will be necessary to determine whether a SiOx coating of an a-Si:H core formed or whether complete oxidation occurred. (b) and (c) represent the negative ions. (d). the positive ions and the neutral species in a diluted silane–oxygen (10% in volume) plasma respectively. (e) and (f) represent the negative ions. These raw data.

negative and positive ion and neutral species spectra of pure silane and of a diluted silane– oxygen plasma at degraded mass resolution are compared in figure 5. The negative ion mass spectra at high resolution in a pure silane plasma in the ranges 20–100 amu for (a) and 365–455 amu for (c). The presence 81 . under both plasma conditions only cations of lower mass and neutral radicals are detected. In (b) and (d) spectra of the negative ions in a diluted silane–oxygen (10% in volume) plasma in the same mass ranges as (a) and (c) are shown. To overcome the trapping of the negative ions in the RF discharge. In both cases. In the diluted plasma each single atomic mass unit from 100 amu up to at least 1000 amu is represented in silane–oxygen plasmas. In these experiments the grounded electrode was heated to 200 ◦ C and a gas mixture of 10% oxygen in silane at a total pressure of 0. can be observed. However. positive ions and negative ions up to 2000 amu. 3) are presented whereas. The plasma on-time (110 µs) and offtime ('300 µs) gave the optimal negative ion signal. In figure 6 the negative ion spectra at high mass resolution of pure silane and of a silane–oxygen plasma are compared for two different mass ranges. intermediate groups ascribed to Si—H—O bonding have appeared. in figure 6(b). the rich variety of Si—O—H radical chemistry completely changes the regularly structured spectrum of pure silane plasmas. 2. a power modulation technique was employed to measure the negative ion component of the plasma [17]. In figure 6(a) only the groups Sin H− x (n = 1. The Balzers PPM 421 plasma monitor allows the determination of neutral species. The spectrum of the diluted oxygen–silane plasma shows Six Hy O+ z clusters up to a similar mass limit to that observed in reaction studies [18] performed by Fourier transform mass spectrometry. For comparison. above 1000 amu. respectively. For higher masses (figures 6(c) and (d)). The plasma parameters were the same as those for figure 5. in pure and oxygen-diluted silane plasmas.1 mbar was used. anions of very high mass.Particle formation in RF plasmas Figure 6.

to some extent. On the other hand. 82 First of all we consider the following cluster growth reactions (1) and (2) to be the most probable: SiH2 insertion + Sin Hx O+ m + SiH4 → Sin+1 Hx+2 Om + H2 elimination (1) and O insertion + Sin Hx O+ m + H2 O → Sin Hx Om+1 + H2 elimination (2) reactions (1) and (2) being the insertion of 30 amu (SiH2 ) and 16 amu (O) respectively and corresponding to the energetically favourable elimination of H2 [18]. In diluted silane–oxygen plasmas. possible: Si insertion + Sin Hx O+ m + SiH4 → Sin+1 Hx Om + 2H2 elimination (3) and O insertion + Sin Hx O+ m + H2 O → Sin Hx−2 Om+1 + 2H2 elimination (4) . in which 2H2 is eliminated instead of H2 only [18]. we adapted a simple model based on six empirical rules. Hollenstein et al [19] have recently shown that a statistical model for the distribution of hydrogen within the anionic clusters can reproduce the regular form of the measured negative ion spectra of pure silane plasmas. chemical structure effects will have to be included to understand the more intense peaks in the negative ion spectrum. the regular structure of the cation spectrum (figure 5(e)) seems to show no special ‘magic number’ effect like those for anions and cations in pure silane.Ch Hollenstein et al Figure 6. Then we consider the parallel reactions (3) and (4). (Continued) of certain masses with higher abundance in the mass range 400–600 amu might indicate clusters with particular stability. statistical reactions also appear to be important but. Following their observations. Mandich and Reents [18] have proposed that addition of H2 O promotes clustering of positive ions in silane plasmas.

Chapple-Sokol J D and Gordon R G 1990 J. A comparison between this model and measurements at high mass resolution of the positive ion spectrum is presented in figures 7(a) and (b) for a representative part of the total positive ion mass spectrum. Zhang J F and Eccleston W 1993 Semicond. The overall features are clearly reproduced by In situ FTIR and mass spectrometer measurements were used to investigate silane–oxygen reaction products in a RF plasma. Sci. in which the isotopic abundances of silicon and oxygen have also been taken into account. Electrochem. McCormick T L. namely SiOx with x = 1 for initial stoichiometry. References [1] Landsberger L M and Tiller W A 1990 J. The number ratio [H]:[Si] of H atoms to Si atoms within the cluster after a large number of consecutive silane reactions therefore tends to 2f n:n = 2f . Acknowledgments The authors thank M Korner from Balzers AG for the help in mass spectrometry. Phys. We introduce f as the branching ratio for the ‘probable reaction’ (1) and hence 1 − f is the branching ratio for the ‘possible reaction’ (3). whereas a small silane flow appears to produce nano-scale stoichiometric particles quite similar to the commercial nanometric SiO2 powder. Kouhli M. The fifth rule forbids any reactions with atomic H elimination [18] and so a sequence of ‘odd’ masses develops independently of an ‘even’ mass sequence.Particle formation in RF plasmas the model. Lett. A statistical reaction scheme for the positive initial cluster growth appears to be in reasonably good agreement with the experimental results. Appl. 18 554 83 . 64 6515 [9] Taylor S. Eccleston W and Barlow K J 1988 J. Res. A post-treatment oxidation of hydrogenated silicon particles was obtained by addition of oxygen to the argon plasma with a-Si:H particles trapped in it. Boufendi L. a modification in the particle chemistry with dilution could be seen. the transmission FTIR signals have illustrated that a small proportion of silane can be totally depleted by the oxygen. additional reaction mechanisms must also be involved in this case. Technol. An equal branching ratio for H2 and 2H2 elimination was also supposed for the oxygen addition reactions (2) and (4). 8 1426 [10] O’Hanlon J F and Pennebaker W B 1971 Appl. reactions (3) and (4) giving increments of 28 and 14 amu respectively. Soc. Conclusions Figure 7. Torres A. Bouchoule A and de Hoog F J 1996 J. As far as the gas phase is concerned. a high silane flow being able to produce large non-stoichiometric particles of SiOx . An imposed sixth rule allows only alternate addition of Si and O to the clusters and so imposes concentration equality [Si] = [O]. According to previous measurements [19] in pure silane plasmas. den Boer J H W G. 137 2825 [2] Giunta C J. which retrospectively supports the choice of equal branching ratios between the H2 and 2H2 elimination reactions for additions of silicon (reactions (1) and (3)) and oxygen (reactions (2) and (4)). Soc. Aceves M and Osorio R 1995 Thin Solid Films 270 114 [5] Kroesen G M W. the experimentally observed [H]:[Si] ratio of 43 and therefore the branching ratios of the assumed reactions (1) and (3) for H2 and 2H2 elimination are f = 23 and 1 − f = 13 respectively. Insofar as the created particulates are concerned. Huang J Y and Huang J G 1995 J. for n consecutive reactions with silane molecules.0136 (for Brite–Euram project BE-7328) and BEW 9400051. 10 139 [8] Taylor S. The occurrence of a rich variety of Si—O—H ionic reactions was demonstrated also by means of mass spectrometry measurements. Miotello A and Kelly R 1994 Thin Solid Films 241 383 [4] Calleja W. we consider the case of cluster growth reactions for pure silane plasmas when only reactions (1) and (3) are involved. Mater. Vivet F. Technol. Iteration of the reaction sequences (1)–(4) with these branching probabilities. 137 3237 [3] Verdi L. Electrochem. Vac. 5. First. Koch C C. A comparison of the positive ion mass spectra in a silane–oxygen (10% in volume) plasma obtained from (a) the modelling and (b) the experimental measurements. gives the calculated positive ion mass spectrum presented in figure 7(a). Therefore. n silicon atoms and 2f n hydrogen atoms are added because only a fraction f of the n reactions are of type (1) which add H2 to the cluster. Phys. This work was funded by Swiss Federal Research Grants BBW 93. Judging by the complexity of the negative ion spectrum for silane–oxygen plasmas (figures 5(d) and 6(d)). Falcony C. The reasonable agreement obtained lends some support to the assumed iterative reaction sequence scheme which is essentially a statistical model for cluster growth. starting from the measured abundances of the Si2 Hx O+ 2 group. A 14 546 [6] Baraton M-I 1995 Nanostructured Materials 5 179 [7] Shen T D. Sci. Nemanich R J.

Dorier J-L and Sansonnens L 1996 J. A 4 689 [33] Pliskin W A and Lehman H S 1965 J. Electrochem. Sci. 2nd edn (London: INSPEC. Viera G. Phys. Joseph J and Irene E A 1991 Appl. Schwarzenbach W. Chao S S. Demond F J and Kalbitzer S 1980 Phys. Phys. 80 1171 [30] Boyd I W and Wilson J I B 1982 J. Status Solidi 100 43 84 [22] Yacobi B G. 543 4166 [31] Nakamura M. 24 101 [17] Howling A A. Rev. 79 8856 [29] Liao W S and Lee S C 1996 J. Chao S S. Dorier J-L. Yang J. Usami K. Chao S S. Howling A A. Appl. Dorier J-L and Hollenstein Ch 1994 J. Courteille C. Hollenstein Ch. Kiermasz A and Eccleston W 1982 Plasma Phys. Rev. 80 2069 [15] Hu S M 1980 J. Appl. Chem. 24 605 [12] Hu Y Z. Vac. Ley L. Courteille C. Tyler J E and Czubatyj W 1984 J. Takagi Y and Lucovsky G 1986 J. Yang J. Phys. Appl. Phys. 59 1353 [13] Dorier J-L. Merad A and Boeuf J-P 1995 IEEE Trans. A 2 353 [27] 1989 Properties of Amorphous Silicon EMIS Datareviews Series No 1. Howling A A. Vac. Hollenstein Ch and Howling A A 1995 J. Sci. Plasma Sci. Lett. Viktorovitch P and Paul W 1981 Phys. Technol. Cardona M and Cuomo J J 1977 Phys. Phys. Soc. Sci. Collins R W. Technol. The Institution of Electrical Engineers) [28] Cicala G. 50 1079 [32] Pai P G. Itoh Y and Nozaki T 1984 Solid State Commun. B 28 3225 [24] Lucovsky G and Pollard W B 1983 Physica B 117&118 865 [25] Hiraki A 1984 Amorphous Semiconductor Technologies & Devices (Amsterdam: North-Holland) [26] Lucovsky G. Mochizuki Y. Sci. A 13 918 [14] Courteille C. Phys. Phys. Schwarzenbach W. 112 1013 [34] Luft W and Tsuo Y S 1993 Hydrogenated Amorphous Silicon Alloy Deposition Processes (New York: Marcel Dekker) . Sansonnens L. A 14 535 [20] Brodsky M H. 75 1340 [18] Mandich M L and Reents W D 1992 J. Appl. Moddel G. Vac. Martins R and Macarico A 1996 J. Bertran E. Hollenstein Ch. B 24 5907 [23] Lucovsky G. Phys. Technol. Fang C J. Appl. Rev. Schwarzenbach W.Ch Hollenstein et al [11] Moruzzi J L. Tyler J E and Czubatyj W 1983 Phys. Technol. Gay P. Appl. 51 5945 [16] Dorier J-L. B 16 3556 [21] Shanks H. Capezzuto P and Bruno G 1996 J. Howling A A. 96 4233 [19] Hollenstein Ch. Cardona M. Vac.