Vacuum 86 (2012) 1195e1202

Contents lists available at SciVerse ScienceDirect

Vacuum
journal homepage: www.elsevier.com/locate/vacuum

Optical properties and crystallinity of hydrogenated nanocrystalline silicon (nc-Si:H) thin lms deposited by rf-PECVD
Goh Boon Tong a, *, Zarina Aspanut a, Muhamad Rasat Muhamad a, b, Saadah Abdul Rahman a
a b

Low Dimensional Materials Research Centre, Department of Physics, Faculty of Science, University of Malaya, Lembah Pantai, 50603 Kuala Lumpur, Malaysia Chancellory, Multimedia University, 63100 Cyberjaya, Selangor, Malaysia

a r t i c l e i n f o
Article history: Received 15 September 2011 Received in revised form 1 November 2011 Accepted 2 November 2011 Keywords: nc-Si:H LBL Optical properties Crystallinity HRTEM Ionic constant

a b s t r a c t
Hydrogenated nanocrystalline silicon (nc-Si:H) thin lms prepared in a home-built radio-frequency (rf) plasma enhanced chemical vapour deposition (PECVD) system have been studied. The rf powers were xed in the range of 5 We80 W. The optical properties and crystallinity of the lms were studied by Xray diffraction (XRD), Micro-Raman scattering spectroscopy, high resolution transmission electron microscope (HRTEM), and optical transmission and reection spectroscopy. The XRD and Micro-Raman scattering spectra were used to investigate the evidence of crystallinity in order to determine the crystallite sizes and crystalline volume fraction in the lms. The HRTEM image of the lm was used to correlate with the crystallinity that was determined from XRD and Micro-Raman scattering spectra. Optical constants such as refractive index, optical energy gap, Tauc slope, Urbach energy and ionic constants were obtained from the optical transmission and reectance spectra. From the results, it was interesting to found that the optical constants showed a good correlation with the crystallinity within the variation of rf power. Also, the ionic constants of the lms showed an indication of the degree of crystallinity in the lms. The variation of the optical energy gap with the rf power based on structure disorder and the quantum connement effect is discussed. 2011 Elsevier Ltd. All rights reserved.

1. Introduction Hydrogenated nanocrystalline silicon (nc-Si:H) thin lms have been extensively studies in recent years especially in opto-electronic device applications such as solar cells, thin lm transistors (TFTs) and sensor devices [1e3]. The nc-Si:H is a biphasic material consisting of a small amounts of a crystalline volume fraction of nanocrystallites embedded within an amorphous matrix. Presence of this crystalline phase inclusions improves the structure of the lms by enhancing the solar cell conversion efciency and reducing the light induced effect, improving the stability of the TFT devices and increasing the mobility of carriers to improve the efciency of devices [1e3]. Also, this material is a promising candidate due to its wider band gap and near-direct band gap nature [4]. This characteristic is very much dependent on crystallite size and crystalline volume fraction in lms and those parameters controlling the optical properties of the lms. However, the main difculty with nc-Si:H is precise control of the size of crystallite and their uniform distribution, otherwise it is very difcult to get the desired properties. The optical properties are important, as they provide direct information on the electronic characteristics of the lms such as
* Corresponding author. Tel.: 60 3 79674147; fax: 60 3 79674146. E-mail addresses: gohbt@um.edu.my, boontong77@yahoo.com (G.B. Tong). 0042-207X/$ e see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.vacuum.2011.11.001

dielectric and ionic constants. These optical constants can be determined from the optical reectance and transmission spectra such as refractive index, no, optical energy gap, EG, Tauc slope, B, Urbach energy, EU, and ionic constant, b. Actually, these constants have strong correlation with the crystallinity of the lms such as crystallite size, crystalline volume fraction, X, even the distribution of the crystalline grains on the lms. Understanding this correlation will provide better control in the deposition of the device quality of nc-Si:H thin lms for opto-electronic applications. In this work, nc-Si:H thin lms have been deposited by continuous deposition using a home-built rf-PECVD system at different rf powers. The optical properties and crystallinity of the lms were characterized by X-ray diffraction (XRD), micro-Raman scattering spectroscopy, high resolution transmission electron microscope (HRTEM), and optical transmission and reection spectroscopy. The objective of this work is to study the relation of the presence of nc-Si embedded within an amorphous matrix with optical properties of the lms. We also aim to study the effect of substrate on the crystallinity of the lms under different variations of rf power. 2. Experiments Hydrogenated nanocrystalline silicon (nc-Si:H) thin lms were deposited on p-type (111) crystal silicon (c-Si) wafers and quartz

1196

G.B. Tong et al. / Vacuum 86 (2012) 1195e1202

substrates using a home-built 13.56 MHz radio-frequency (rf) plasma enhanced chemical vapour deposition (PECVD) system from the discharge of mixed silane (SiH4) and hydrogen (H2) gases described in our earlier publication [5]. The reactor was capacitively coupled by two parallel electrodes with a distance of 5 cm and area of 28 cm2. Prior to deposition, hydrogen plasma treatment was done on the surface of the substrates for 10 min at a chamber pressure of 0.8 mbar to clean the substrate surface from any contamination. The deposition pressure and substrate temperature were maintained at 0.8 mbar and 250  C respectively, during the deposition. The SiH4 and H2 gases ow-rates were xed at 2 and 100 sccm respectively, producing a H2 to SiH4 ow-rate ratio of 50. The total deposition time was xed at 60 min. Five sets of samples were prepared with different rf powers of 5, 15, 25, 60 and 80 W which correspond to rf power densities of 179, 536, 893, 2143 and 2857 mW/cm2. The X-ray diffraction (XRD) measurement was carried out using SIEMENS D5000 X-ray diffractometer, with Cu Ka X-ray radiation (l 1.5418 ). The Raman spectra of the lms were recorded using a Horiba Jobin Yvon 800 UV Micro-Raman Spectrometer with Ar laser (Excitation wavelength of 514.5 nm). The lm for high resolution transmission electron microscope (HRTEM) micrographs was obtained by means of a TEM (JEOL JEM-3100F) operating at 300 kV. The optical transmission and reectance spectra of the lms were obtained using a JASCO V570 ultra-violet visible nearinfrared (UVeVISeNIR) spectrophotometer. The XRD and Raman scattering spectroscopy were performed on lms deposited on c-Si and quartz substrates while the optical spectroscopy was measured only on the lms deposited on the quartz substrate. 3. Results and discussion Fig. 1 shows the X-ray diffraction (XRD) spectra of lms deposited on c-Si and quartz substrates at different rf powers. For the lms deposited on c-Si substrate, all the lms except for the lm deposited at rf power of 25 W show diffraction peaks at 28.4, 47.3 and 56.1 corresponding to c-Si orientations of (111), (220) and (311) planes respectively [6]. The absence of a diffraction peak at 47.3 for the lm deposited at rf power of 25 W could be due to less dominance of the c-Si (220) plane contributing to the diffraction. However, for the lms deposited on the quartz substrate, there are only small c-Si diffraction peaks appearing at 28.4, 47.3 and 56.1 for the lms deposited at rf powers above 25 W. For the lms deposited at rf powers below 25 W, there is a very small shoulder at 28.4 appearing from the broad amorphous peak at about 22 suggesting that the presence of very small crystallites embedded

within the amorphous matrix for the lms. From this observation, we nd that there is a difference in the effect of the substrate in inducing the crystallization of the deposited lm under variation of rf power. The c-Si substrate as a semiconductor shows a crystalline orientation of c-Si (311) plane in the gure, which induces the crystalline growth in the deposited lms with preferred orientation of c-Si (311) plane. Moreover, this crystalline in the deposited lm was formed at the lower rf powers. For the quartz substrate as insulator, the crystalline growth behaved naturally in the deposited lm with preferred orientation of c-Si (111) plane in highly diluted hydrogen of rf-PECVD deposition [7]. However, this crystalline plane only appears signicantly for the lms deposited at rf power above 25 W. An increase in rf power increases the intensity of the diffraction peaks for both lms deposited on c-Si and quartz substrates indicating the increase in crystallinity of the lms. By using the Scherrers equation, the crystallites size can be estimated from the relation of D kl=bcosq, where k, l, b and q are the Scherrers constant, the wavelength of X-ray, the full width at half maximum (FWHM) and Bragg angle of the diffraction peak respectively [8]. The values of the crystallite sizes for the lms deposited on c-Si and quartz substrates are tabulated in Table 1. The variations of the crystallite sizes for the both lms are consistent with the XRD spectra as a function of rf power. Raman scattering spectra of lms deposited on c-Si and quartz substrates at different rf powers is shown in Fig. 2. Note that, all the lms show appearance of asymmetric Raman peaks in the range of 500e512 cm1 indicating that the lms are mixed phases of nanocrystalline and amorphous structures [9]. However, the lms deposited on c-Si substrates at rf powers of 5 and 15 W show appearance of a sharp crystalline TO peak at about 520 cm1. Due to the thickness of the lms (less than 500 nm) and laser wavelength of 514.5 nm used in the measurement, the sharp crystalline TO peak suggests the single c-Si at 520 cm1 is due to the effect of c-Si substrate [10]. The variation of these Raman peaks demonstrates a similar trend with the XRD diffraction peak intensities as a function of rf power. Moreover, the Raman peaks shifted to a lower wave number from the single c-Si peak at 520 cm1 agreeing with the XRD result that has been suggested, namely that there are very small crystallites embedded within the amorphous matrix in all the deposited lms. From the Raman scattering spectra, it is well known that the Raman peak can be deconvoluted into three components corresponding to amorphous, grain boundaries and nanocrystalline components at around 480, 500 and 515 cm1 respectively. A typical Raman peak deconvolution into four components for the lms deposited on c-Si substrates is shown in Fig. 2(c). The additional component at around 520 cm1 is

Fig. 1. XRD spectra of nc-Si:H thin lms deposited on (a) c-Si and (b) quartz substrates at different rf powers.

G.B. Tong et al. / Vacuum 86 (2012) 1195e1202 Table 1 Crystallite size, crystalline volume fraction, lm thickness and refractive index of nc-Si:H thin lms deposited on c-Si and quartz substrates, at different rf powers. rf Power (W) XRD Crystallite size, Dc (nm) 5W 15 W 25 W 60 W 80 W 3.4 3.7 3.7 5.7 5.8 Crystallite size, Dq (nm) e e e 5.7 6.6 Raman Crystallite size, dc (nm) 2.5 3.2 3.1 5.1 5.1 Crystallite size, dq (nm) 2.3 2.0 2.6 3.9 5.0 Crystalline volume fraction, Xc 27 28 25 51 55 Crystalline volume fraction, Xq 23 7 13 43 43 Optical Film thickness, t (nm) 125 201 388 786 903

1197

no 2.58 2.53 2.81 3.02 2.99

Di, di and Xi represent the estimated crystallite size from XRD, crystallite size from Raman and crystalline volume fraction from Raman respectively. i c and q represent the parameters of the lms that deposited on c-Si and quartz substrate respectively.

contributed by the signal from c-Si substrate. The crystalline volume fraction, X of the lms was estimated using the relation of X Ib Ic =aIa Ib Ic , where Ia, Ib and Ic are integrated intensities of the Raman peaks corresponding to amorphous, grain boundaries and nanocrystalline components respectively. The factor a is generally is equal to 1 for nc-Si lms [11]. Also, the crystallite size of the nanocrystallites in the lms can be estimated p using the relation of d 2p B=Du, where B is 2.24 cm1 nm2 for

Si and u is the value of shift nanocrystalline component to single crystal Si peak located at 521 cm1 [12]. The values of the estimated crystalline volume fraction and crystallite size for the lms deposited on c-Si and quartz substrates are tabulated in Table 1. As can be seen from the table, the values of crystalline volume fraction are higher for the lms deposited on c-Si and quartz substrates at higher rf powers (60 and 80 W). However, the values of crystalline volume fraction for the lms deposited on c-Si

Fig. 2. Raman scattering spectra of nc-Si:H thin lms deposited on (a) c-Si and (b) quartz substrates at different rf powers. (c) A typical of Raman peak deconvoluted into four components for the lms deposited on c-Si substrates. The tted components labelled as a, b, c and d represents the amorphous, intermediate, nc-Si and c-Si components respectively. The light grey line indicates the sum of the all components.

1198

G.B. Tong et al. / Vacuum 86 (2012) 1195e1202

Fig. 3. HRTEM image of nc-Si:H thin lm deposited at rf power of 80 W. Inset (a) shows the image with lower resolution showing the spherical nanoparticles and inset (b) demonstrate the corresponding electron diffraction pattern.

substrate are higher as compared to the lms deposited on a quartz substrate. On the other hand, the crystallite size increases with increase in rf power for the lms deposited on both the substrates. This shows good agreement with the XRD results. An increase in the crystallite size of nc-Si depending on the deposition conditions is compatible with previous observations [13e15]. In this case, prehydrogen plasma treatment about 10 min before the deposition created a lot of energetic atomic hydrogen removing the oxide contaminants on the substrate surface, which then leads to the creation of the dangling bonds. These dangling bonds provide preferable growth sites for crystallization at the early stage of the lm growth. During the deposition, an increase in rf power produces more atomic hydrogen from the dissociation of SiH4 and H2 molecules. The presence of atomic hydrogen plays an important role in the creation of nucleation sites on the surface by breaking weak bonds and forming rigid bonds after the abstraction. Along with the local heating induced by this process, this leads to an increment in crystallite size and density of the crystalline regions that formed during the deposition and it can even enhance the agglomeration of adjacent nc-Si during growth. Thus, for this deposition condition bigger nc-Si with a higher crystalline volume

fraction are formed at higher rf powers, however, the crystallization is more signicant for the lms deposited on c-Si substrate. Fig. 3 shows a HRTEM image of the nc-Si:H thin lms deposited at rf power of 80 W. In the gure, the image demonstrates a homogeneous amorphous network with a presence of small crystalline grains surrounding the amorphous matrix. From the rough estimation on the image, the diameter of the crystalline grain is about 5 nm. This crystallite size is about the same with the average crystallite sizes that was estimated from XRD and Raman scattering spectra for the lms deposited on c-Si and quartz substrates at the same rf power. The estimated lattice spacing is about 0.31 nm which corresponds to the c-Si (111) plane. This supports the preferred orientation of c-Si (111) plane that was shown by the XRD results. The inserted gure labelled (a) showed that the crystalline grains in the amorphous matrix are clustering together forming spherical nanoparticles with the average particle diameter being about 23 nm. The corresponding electron diffraction pattern as shown in the inserted gure labelled as (a) shows the evidence of the crystallinity of the lm. The weak electron diffraction signal could be due to small amounts of crystalline grains present in this scanning spot. Besides, the crystalline volume fraction in the lm is considered low (43%) for the weak electron diffraction signal. Fig. 4 shows the optical reectance and transmittance spectra of the lms deposited at different rf powers. In Fig. 4(a), the reectance spectra exhibits interference patterns at the wavelength above 400 nm due to the effects of coherent multiple reections on the lms. On the other hand, the end point of the interference fringe of the reectance spectrum shifted signicantly towards a longer wavelength from 390 to 540 nm with increase in rf power. In addition, appearance of clear peaks at 274 nm in the UV region of the reectance spectra for all the lms suggests that the lms are crystalline in structure which has been indicated by previous works [6,16]. In Fig. 4(b), the transmittance spectra show interference fringes at the wavelength above 400 nm which is similar to the interference fringes of the reectance spectra. These spectra present a good transparency in the near-infrared region. However, this transparency degrades in the visible region which can be explained by the optical absorption characteristics of the lms. Generally, the region of 400e600 nm, where the transmittance decreases gradually to zero indicates the absorption edge of the optical energy gap. In Fig. 4(b), the absorption edge shifted to a longer wavelength with increase in rf power. This is similar with the shift of the end point of the interference fringe in the reection spectrum with increase in rf power. Therefore, these changes in the reectance and transmittance spectra suggest that the optical energy gap decreases with increase in rf power.

120 100

80 W 60 W

b 100
90 80

80 60

Transmission (%)

70 60 50 40 30 20 10 5W 15 W 25 W 400 500 60 W 80 W 600 700 800 900

Reflection (%)

25 W

15 W 40 20 0 200 5W

300

400

500

600

700

800

900

0 300

Wavelength (nm)

Wavelength (nm)

Fig. 4. Optical (a) transmission and (b) reectance spectra of nc-Si:H thin lms deposited at different rf powers.

G.B. Tong et al. / Vacuum 86 (2012) 1195e1202

1199

The thicknesses of the lms were determined from interference fringes of the reectance and transmittance spectra in Fig. 4, using the conventional optical expressions which have been employed by Manifacier et al. [17] and Swanepoel [18] since 1980. The calculated thicknesses of the lms are tabulated in Table 1. The lm thickness increases with increase in rf power. This is consistent with increase of the number of interference fringes of the reectance and transmittance spectra against rf power which is usually used as an indication of the lm thickness. Generally, increase in rf power enhances the dissociation and ionization of silane radicals and thus a resulting increase of the ux of depositing precursors to the substrate surface. This has contributed to the increase of the deposition rate [19e21]. From the obtained data of the reectance, R and transmittance, T, the n versus l plots of the lms can be obtained by tting the n values obtained at the extreme points of the interference fringes to the Cauchy 2 relation [22] of n no ko =l , where no is the static refractive index of the lm and ko is a constant tting parameter. The values of static refractive index of the lms are in the range between 2.5 and 3. Also, the data of reectance, R and transmittance, T, can be used to determine absorption coefcient, a of the lms using the relation [23] of a 1=tln1 R=T, where the t is lm thickness. The calculated a was plotted against the photon energy, E to determine the Urbach energy, EU using the relation [24] of aE ao expE E* =EU , where ao is a constant and E* is the onset of the band tail. The gure of the variation of a against photon energy of the lms deposited at different rf powers is not shown. The calculated a in the region of the absorption edge was substituted into the Tauc relation [25] of aE B2 E EG 2 , where the EG is optical energy gap which is determined from the extrapolation of the linear region of the Tauc plot of (aE)1/2 versus E from the Tauc relation, the B is proportional to the slope of the band edge and E is photon energy. The region of the energy (1.8e3.0 eV) was selected due to the region of density of states of nc-Si:H thin lm where the absorption edge occurs. Fig. 5 shows the Taucs plots of the lms deposited at different rf powers. In the gure, it clearly shows that the absorption edge shifted to lower photon energy with increase in rf power. Also, the absorption intensity increases with increase in rf power. These are agreeable with the observation in Fig. 4(b). Variation of the optical energy gap with rf power will be discussed in the next part of the text. Variations of the Urbach energy, EU and Tauc slope, B of the lms with rf power are shown in Fig. 6. EU also called Urbach tail band
800 5W 15 W 600 25 W 60 W 80 W

1200

800

1000 600 800

B (cm.eV)-0.5

600

400

400 200 200

20

40

60

80

0 100

Power (W)
Fig. 6. Variations of the Tauc slope, B and Urbach energy, EU with rf power for the lms deposited at different rf powers.

400

200

0 1.8

2.0

2.2

2.4

2.6

2.8

3.0

E (eV)
Fig. 5. Plot of (aE)1/2 as a function of photon energy, E for nc-Si:H thin lms deposited at different rf powers.

width denes the width of the exponential tail of density of states, which extends into the band gap. Usually, the exponential tail of the density of states was referring to the localized states which are an indication of structure disorder, where a higher value indicates higher disorder [26]. As in EU, B is an indicator of the structural order, however, it is inversely proportional to the structural disorder where a high value of B indicates a more ordered lm structure with a large band edge width [27]. The B decreases gradually at lower rf powers (5 We25 W) and is nearly constant at low values with further increase in rf power. However, the EU showed an opposite trend as compared to the B. These trends indicate that the lms are more ordered at lower rf powers and decreased in structural order with further increase in rf power. Generally, the lms are more disordered in nature due to the increase in the deposition rate with increase in rf power [7]. In this work, the increase in the crystallite size and simultaneous increase in crystalline volume fraction with increasing rf power, as observed from XRD and Raman scattering spectroscopy studies respectively, conrm the formation of an increasing number of nc-Si grains at higher rf power. So, the increase of the band edge width with increasing rf power signies the growth of defects and disorder [23] which may be correlated with the increasing number of nc-Si grains and the associated grain boundaries which are known to be normally highly defective [21]. The dispersion curve obtained from the Cauchy relation was described above and can be used to determine the dispersion energy, Ed and single oscillator energy, Eo using the Wemple and 2 DiDenomico relation [28,29] of n2 1 Ed Eo =Eo E2 , where Eo and E are the single oscillator energy and photon energy respectively. Furthermore, the obtained values of no, EG and Ed were substituted into Bhattacharya relation [30] of b Ed =6EG n2 1, to estimate the ionic constant, b of the lms. The value of b is depends on the chemical bonding character of the material. Fig. 7 shows the variations of the dispersion energy, Ed and ionic constant, b of the lms with rf power. Generally, the Ed is a measure of the strength of the inter-band optical transitions which indicates the characteristics of electronic transition from valence band to conduction band in the lm. The value of Ed is related to the structural disorder in the lm [31] and the structural disorder of ncSi:H lms were mainly contributed by the grain boundaries of nanocrystallites embedded within the amorphous matrix [31]. As can be seen from the gure, the Ed increases with increase in rf power indicates that the lms are more disordered at higher rf

( E)1/2 (cm-1/2.eV1/2)

Eu (meV)

1200

G.B. Tong et al. / Vacuum 86 (2012) 1195e1202

0.35
25 0.30

2.4

50

20

2.2

40

Ed (eV)

EG (eV)

0.25

2.0 20 1.8

10 0.20 5

10

20

40

60

80

0.15 100

1.6

20

40

60

80

0 100

Power (W)
Fig. 7. Variations of the dispersion energy, Ed and ionic constant, b with rf power for the lms deposited at different rf powers. Inset shows the variation of crystalline volume fraction with rf power for the deposited lms.

Power (W)

2.7 2.5 2.3

powers. The Ed agrees with the B and EU under the same variation of rf power. The values of single oscillator energy, Eo, are nearly constant for all the deposited lms. The values of b are in the range of 0.2e0.27 for all the deposited lms indicate the atomic bonds in the lms have ionic characteristics [32]. This also suggests that all the lms containing the same anions species. From previous report [28], the b increased with the crystallinity of the ionic solids which contained a single anion species. Therefore, the trend of b with rf power for all the deposited lms suggests that high values of b could be an indication of the crystallinity in the lms. This agrees with the crystalline volume fraction of the lms as tabulated in Table 1. The inset of the gure showed the same trend occurs to the XC under the same variation of rf power which agrees with the prediction above. In addition, the dispersion energy, Ed can be related to hydrogen content, CH using the following relation [33] of p CH % 1=3nv =ns 4 Ed =2:8, where ns is the electron density 23 of c-Si taken as 2 10 electrons cm3, and nv is the valence electron density in the lm which is determined using the following relation 2 [34] of nv 0:0143Ed =3 0 1 1023 electrons cm3 , where 3 (0) is taken as no, which is the long-wavelength limit of the refractive index [32]. The variations of the optical energy gap, EG, and hydrogen content, CH of the lms with rf power is shown in Fig. 8(a). EG decreases gradually with increase in rf power to 60 W and showed nearly no change with further increase in rf power up to 80 W. On the other hand, the CH shows no signicant variation with the rf power. In general, the variation of EG in nc-Si:H thin lms has been reported as dependent on the hydrogen content, CH [35], the structural disorder [36] or the crystallite size of nc-Si. In a-Si:H lms, the optical energy gap is known to increase nearly proportional to the hydrogen content [35]. The values as high as 2e2.5 eV have been reported at w40% hydrogen content [37]. The values of optical energy gap measured in our nc-Si:H lms are in the range of 1.82e2.27 eV with hydrogen content about w34e39%. However, the optical energy gap of nc-Si:H lms showed no dependence on the hydrogen content indicating that the hydrogen could be accumulated in the amorphous phase. By neglecting of the role of the hydrogen content, the variation of the optical energy gap could be due to the structural disorder and the crystallite size of nc-Si. The results from the variations of the EU, B and Ed with rf power as described above explains that the highly disordered lm structure narrows the optical band gap for the lms deposited at higher rf powers. This is supported by previous work

2.1

EG (eV)

1.9 1.7 1.5 1.3 1.1 2 3 4 Crystallite size (nm) 5 6

Fig. 8. (a) Variations of the optical energy gap, EG and hydrogen content, CH with rf power for the lms deposited at different rf powers, (b) The plot of optical energy gap, EG against the crystallite size of nc-Si:H thin lms. The line is drawn to guide the eye.

which studied the relation of the optical energy gap with the structural disorder in the nc-Si:H thin lms [27]. According to the model that has been described by Santana et al. [38], the presence of nc-Si embedded within the amorphous matrix enhanced the quantum connement effect which is responsible for the widening of the band gap. The model of the quantum connement effect follows the relation of EG Ec c=a2 , where c is the connement constant, Ec is the bulk silicon band gap and a is the diameter of crystallite. From this relation, it clearly indicates that the optical energy gap, EG depends on crystallite size of nc-Si embedded within the amorphous matrix. Moreover, the crystallite size of the nc-Si must be smaller than 10 nm in order to enhance the quantum connement effect. Fig. 8(b) shows the variation of EG against the crystallite size of the nc-Si in the lms. The values of crystallite sizes plotted in the gure, refers to the crystallite sizes that were determined from the XRD and micro-Raman scattering spectra for the lms deposited on quartz substrates. This is to ensure the consistency of the results which can be understood that the values of EG were determined from the lms deposited on quartz substrates. From the gure, a decrease of crystallite size of nc-Si shows gradual increase in EG which shows good agreement with the connement model as described above. At least, these crystallite sizes of nc-Si could be responsible for the quantum connement effect which narrows the optical band gap of nc-Si:H thin lms at higher rf powers.

CH (%)

15

30

G.B. Tong et al. / Vacuum 86 (2012) 1195e1202 Table 2 Summary results of the various deposition methods, their deposition conditions and lm properties. Deposition methods rf-PECVD (1) Deposition conditions Power density 179e2857 mW/cm2 (5e80 W), Ts w 250  C, Pressure w 0.8 mbar, Hydrogen dilution, R FH2 =FSiH4 50 a Power 10e45 W, Ts w 220  C, Pressure w 1.2 mbar, Hydrogen dilution, R FH2 =FSiH4 50 Power density 340 mW/cm2, Ts w 250  C, Pressure w 1.33 mbar, Hydrogen dilution, R FH =FSiH4 100e300 a Power 80 W, Ts w 400  C, Pressure w 0.003 mbar, Hydrogen ow-rate, FH 0.002 mbar Ts w 200  C, Pressure w 0.013e0.067 mbar, Hydrogen dilution, R FH =FSiH4 15 Crystallite size and crystalline volume fraction d w 2.0e6.6 nm XC w 7e43% d w no change XC w 64e79% d w 8.3e13 nm XC w 22e58% d w 45e64 nm Optical energy gap EG w 1.86e1.95 eV Reported works This work

1201

rf-PECVD (2)

EG w 1.86e1.95 eV

Badran et al. (Ref. [39])

rf-PECVD (3)

EG w 1.85e1.88 eV

Deyan et al. (Ref. [40])

rf Magnetron Sputtering

EG w 1.47e1.72 eV

Tovar et al. (Ref. [41])

HWCVD

d w 2.0e2.5 nm XC w 10e54%

EG w 1.7e2.1 eV

Jadkar et al. (Ref. [42])

a Only power is shown in the table for the deposition conditions where the geometry of electrode is not provided. FH2 ; FSiH4 and Ts represent the hydrogen gas ow-rate, silane gas ow-rate and substrate temperature respectively, of the each deposition methods.

Table 2 tabulates the summary results of the various deposition methods, their deposition conditions and lm properties. The result in this work shows almost similar with the work reported by Badran et al. [39]. Higher deposition pressure generally produced higher crystalline volume fraction that supported by Badran et al. work [40]. Higher hydrogen dilution ratios was carried out by Deyan et al. using similar deposition method however the crystalline volume fraction and the crystallite size are signicantly higher as compare to our work. The magnetron sputtering method produced larger crystallite size which does not exhibit quantum connement effect as shown by Tovar et al. [41]. HWCVD method showed a slightly higher crystalline volume fraction with smaller crystalline size at the lower substrate temperature. However, the lower deposition pressure was applied for achieving these lm properties. Moreover, no ion bombardment effect in the HWCVD processes might leads to the small crystallite size in the lms. 4. Conclusions In this work, we have studied the effects of rf power on the optical properties and crystallinity of nc-Si:H lms deposited on crystal silicon and quartz substrates by rf-PECVD. Increase in rf power increases the crystalline volume fraction and crystallite size in the lms due to the enhancements of the hydrogen etching effect and the local heating effect, respectively. The enhancement of the crystallinity in the lms with rf power was more signicant for the lms deposited on crystal silicon substrates, even at lower rf powers. Furthermore, increase in rf power decreases the Tauc slope and increases the Urbach energy producing a large band edge width associated with high structural disorder in the lms. The variations of the dispersion energy and ionic constants with rf power correlated well with the increase of structural disorder and the crystallinity in the lms, respectively. Nevertheless, the variation of the optical energy gap with rf power has been explained by the disorder structure in the lms and the crystallite size that enhances the quantum connement effect which is responsible for the widening of the optical energy gap at lower rf powers. Finally, the crystallite size estimated from HRTEM image demonstrated good agreement with the crystallite sizes determined from the XRD and Raman spectra. Acknowledgements This work was supported by the Ministry of Higher Education for Fundamental Research Grant Scheme (FRGS) of FP002/2010B

and University of Malaya Postgraduate Research Fund of PS323/ 2009C. One of the authors, Goh Boon Tong is also grateful to University of Malaya staff training scheme (SLAI) for the nancial support. References
[1] [2] [3] [4] [5] Shim J-H, Im Seongil, Kim Youn Joong, Cho N-H. Thin Solid Films 2006;503:55. Shirai H, Arai T, Nakamura T. Appl Surf Sci 1997;113/114:111. Kleider JP, Longeaud C, Bruggemann R, Houze F. Thin Solid Films 2001;383:57. Kumar Sushil, Dixit PN, Rauthan CMS, Parashar A, Gope Jhuma. J Phys. Condens Matter 2008;20:335215. Ritikos Richard, Goh Boon Tong, Sharif Khairul Anuar Mat, Muhamad Muhamad Rasat, Abdul Rahman Saadah. Thin Solid Films 2009;517: 5092. Budini N, Rinaldi PA, Schmidt JA, Arce RD, Buitrago RH. Thin Solid Films 2010; 518:5349. Das Chandan, Ray Swati. J Phys D Appl Phys 2002;35:2211. Tong Goh Boon, Gani Siti Meriam Ab, Muhamad Muhamad Rasat, Abdul Rahman Saadah. Thin Solid Films 2009;517:4945. Wang XX, Zhang JG, Ding L, Cheng BW, Ge WK, Yu JZ, et al. Phys Rev B 2005; 72:195313. Chen H, Shen WZ. J Appl Phys 2004;96(2):1024. Han Daxing, Lorentzen JD, Weinberg-Wolf J, McNeil LE, Wang Qi. J Appl Phys 2003;94(5):2930. Tsu R, Gonzalez-Hernandez J, Chao SS, Lee SC, Tanaka K. Appl Phys Lett 1982; 40:534. Monroy BM, Santana G, Aguilar-Hernandez J, Benami A, Fandino J, Ponce A, et al. J Lumin 2006;121:349. Benami A, Santana G, Ortiz A, Ponce A, Romeu D, Aguilar-Hernandez J, et al. Nanotechnology 2007;18:155704. Kim T-Y, Park N-M, Kim K-H, Sung GY, Ok Y-W, Seong T-Y, et al. Appl Phys Lett 2004;85:5355. Roy B, Mahan AH, Wang Q, Reedy R, Readey DW, Ginley DS. Thin Solid Films 2008;516:6517. Manifcacier JC, Gasiot J, Fillard JP. J Phys E Sci Instrum 1976;9:1002. Swanepoel R. J Phys E Sci Instrum 1983;16:1214. Lebib S, Roca i Cabarrocas P. J Appl Phys 2005;97:104334. Chowdhury Amartya, Mukhopadhyay Sumita, Ray Swati. Sol Energy Mater Sol Cells 2008;92:385. Bhattacharya Koyel, Das Debajyoti. Nanotechnology 2007;18:415704. Swanepoel R. J Phys E Sci Instrum 1984;17:896e903. Cremona A, Laguardia L, Vassallo E, Ambrosone G, Coscia U, Orsini F, et al. J Appl Phys 2005;97:023533. Shutzmann M. Handbook in semiconductors, vol. 3A. Amsterdam: NorthHolland; 1994. Wang YH, Lin J, Huan CHA. Mater Sci. Eng B 2003;104:80. Cody GD. Mater Res Soc Symp Proc 2005;862. A1.3.1. Tong GB, Abdul Rahman S. The proceeding of the 2006 IEEE International Conference on Semiconductor Electronics (ICSE2006). 2006, p. 472. Wemple SH, Didomenico M. Phys Rev Lett 1969;23:1156. Wemple SH, Didomenico M. Phys Rev B 1971;3:1338. Krishna MG, Bhattacharya AK. Mater Sci Eng B 2001;86:41. Chowdhury Amartya, Mukhopadhyay Sumita, Ray Swati. J Cryst Growth 2007; 304:352. Dakhel AA. Vacuum 2007;81:1101. Ance C, Ferraton JP, Berger JM, De Chelle F. Phys Stat Sol.(b) 1982;113:105.

[6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33]

1202

G.B. Tong et al. / Vacuum 86 (2012) 1195e1202 [39] Badran RI, Al-Hazmi FS, Al-Heniti S, Al-Ghamdi AA, Li J, Xiong S. Vacuum 2009;83:1023. [40] Wang Desheng, Liu Qiming, Li Fei, Qin Yanli, Liu Dequan, Tang Zeguo, et al. Appl Surf Sci 2010;257:1342. [41] Thaiyalnayaki V, Cerqueira MF, Ferreira JA, Tovar J. Vacuum 2008;82: 1433. [42] Bakr NA, Funder AM, Waman VS, Kamble MM, Hawaldar RR, Amalnerkar DP, et al. J Phys. Chem Solids 2011;72:685.

[34] Ance C, Ravindra NM, Berger JM, De Chelle F, Donnadieu A, Ferraton JP. J NonCryst Solids 1983;59 & 60:173. [35] Cody GD, Wronski CR, Abeles B, Stephens RB, Brooks B. Sol Cells 1980;2:227. [36] Kshirsager ST, Lannin JS. Phys Rev B 1982;25:2916. [37] Hishikawa Yoshihiro, Tsuge Sadaji, Nakamura Noboru, Tsuda Shinya, Nakano Shoichi, Kuwano Yukinori. J Appl Phys 1991;69(1):508. [38] Remolina A, Monroy BM, Garcia-Sanchez MF, Ponce A, Bizarro M, Alonso JC, et al. Nanotechnology 2009;20:245604.