Integrated approach for low-temperature synthesis of high-quality silicon nitride films in PECVD using RF–UHF hybrid plasmas

As explained earlier, it is important to monitor the plasma parameters to study their role in the plasma processes. It is also well known that, in an RF plasma environment, the LP I–V (current–voltage) characteristic is distorted owing to oscillations of the plasma potential (V_p) at the RF frequency and its higher harmonics [33, 34]. Garscadden and Emelus [33] reported that the general feature of this distortion is to alter the LP I–V characteristic towards a more negative voltage, providing incorrect readings for the LP floating potential (V_f) and V_p. For sinusoidal excursions of V_p and an LP biased to retard electrons, the electron temperature (T_e) is determined accurately as long as the potential variations remain within the exponential regime of the LP characteristic (i.e. do not take the LP into the electron saturation region). Large excursions of V_p that can occur in high-density plasma can provide an overestimation of the T_e and underestimation of the plasma density (n₀). Numerous studies [15, 34–40] with different RF compensation schemes have been proposed and reported to eliminate this distortion. For the high-frequency UHF (320 MHz) source operation, one can use an LP without RF compensation. However, in the RF frequency and the hybrid plasmas one has to use RF compensated LP. With this consideration, a new LP structure and tuning procedure, similar to that in [34], is designed, and the RF compensated LP is fabricated. The details of the RF LP and its design concept are presented elsewhere [15]. For the LP data acquisition, in the present studies, a commercial Hiden power supply system (UK) is used. The LP system uses window-based graphics ESPsoft application software, which allows acquisition, storage, and analysis of the LP I–V characteristics.

LP analysis from the I–V data determines various plasma parameters such as plasma density (n₀), T_e, V_p, electron energy distribution function (EEDF), etc. The details of LP analysis, the analysis procedure, and its validity and usage are discussed [41, 42] in great detail in the literature.

For the OES diagnostics (figure 1(a)), the spectral data relevant to different optical emission lines are acquired through an optical fibre (Ocean Optics) with an Acton Spectra Pro 500i spectrometer. The spectrometer has minimum resolution  ≈  0.05 nm, a 10 μm wide entrance slit, and a 1200 grooves mm⁻¹ grating, and can be used in conjunction with a PIMAX Princeton Instruments ICCD camera connected to a PC. For the data acquisition the window based graphics software WinSpec32^TM is used.

The VUVAS system (figure 1) is essentially a combination of a VUV monochromator (ARC VM-502) and a high-pressure micro-discharge hollow cathode lamp (MHCL), symmetrically positioned with respect to the chamber. Figure 1(a) also shows that a magnesium fluoride (MgF₂) lens (plano-convex) is placed on each side through the chamber atmosphere. To generate VUV light from the MHCL an on–off modulated dc power of about 10 Hz is supplied. VUV light from the MHCL is focused parallel to the MgF₂ lens and is made to pass through the monochromator across the plasma chamber. The emitted radiation relevant to the resonant emission of the N atoms (in the MHCL) is absorbed by the N atoms (ground state) in the plasma. The light transmitted through the plasma is then focused on the monochromator by another MgF₂ lens. This UV light is converted to visible radiation by the sodium salicylate scintillator of the monochromator. This light is then detected by a photomultiplier tube, and the amplitude of this is recorded as a voltage signal in the digital oscilloscope.

For the VUVAS measurements, it is necessary to select and detect the relevant, resonant emission of the N atoms of the light source, which is absorbed by the ground state N in the process plasmas. The emission profile of the light source also has to be determined to estimate the absorption profile of N and the N atom density (n_A). Additionally, one has to consider the effective absorption path length accurately. In the process plasmas, the guess is, however, not trivial due to the nature of the spatial distribution of N atom density. Also, one needs the information of the gas temperature in the plasma and MHCL light source. Therefore, these aspects should be considered for the development of the VUVAS diagnostic system to make an accurate measurement of n_A.

According to Mitchell and Zemansky [43], the density of the absorbers (or the radical density, n_A) can be connected to the parameter α called the background absorption coefficient in the plasma, by integrating α over the absorption line profile as

$$\begin{eqnarray}&&\mathop{\int}^{}\alpha \,\text{d}\nu ~=~\frac{\lambda _{0}^{2}\,{{g}_{2}}}{8\pi \,{{g}_{1}}}{{A}_{21}}\,{{n}_{\text{A}}}\end{eqnarray} \tag{ 1 }$$

where λ₀, A₂₁, g₁, and g₂ are respectively the line center wavelength, the transition probability from an upper to a lower level, and the statistical weights of the lower and upper states of the transition. Note that the coefficient α is dependent on the plasma conditions. Additionally, depending upon the particular radicals or absorbing species and their interaction with other species in the background plasma, the parameter α can vary. In particular, at a given pressure, plasma density, and gas temperature, it can be accurately predicted according to the theory of line broadening in plasmas. In the present study, one can calculate the electron–neutral collision frequency at pressure  =  500 mTorr. Considering a typical plasma process with electron temperature, T_e  =  4 eV (velocity V_e  =  1.2  ×  10⁶ m s⁻¹), neutral density (at 500 mTorr, gas temperature  =  300 K) ~ 1.6  ×  10²² m⁻³), and collision cross section [24], σ ~ 13  ×  10⁻²⁰ m², the collision frequency ~σV_e (neutral density) can acquire a value of 1.5  ×  10⁹ Hz (a few GHz). In this sense, and from the spectroscopy standpoint, the present operating conditions can be treated as high-pressure conditions. In the high-pressure environment, the line absorption has to be described by a Voigt profile (see [24] for more details), which is a combination of Gaussian and Lorentzian profiles. Further, the absorption coefficient A of the absorbed light in the plasma is a function of various parameters such as absorption path length, gas temperature in the plasma, gas temperature in the MHCL light source, and line absorption profiles [44]. Note that the details of the VUVAS theoretical formulation, the measurement, and the analysis procedure are reported elsewhere [45]. In the present study, the reference line used for the VUVAS measurement is 120.7 nm.

During the deposition process, images of the flow of radical species in plasma are simultaneously recorded (figure 1(a)) using an ICCD camera (PI-MAX, Princeton Instruments) with the objective UV-NIKKOR 105 mm. We use two bandpass filters with FWHM (full width at half maximum) of 7.5 nm. The center wavelengths of the filters are 656.2 nm (F10-656.2, Korea Electro-Optics) and 354.7 nm (F10-354.7, Korea Electro-Optics). They are respectively used for measuring the two-dimensional (2D) gaseous flows of hydrogen and nitrogen with the ICCD camera. A cage filter wheel is used to install these filters so that they can be changed quickly and easily without disturbing the setup. The ICCD camera used has a resolution of 512 pixels  ×  512 pixels. It is focused on a typical field of view of 25 mm  ×  25 mm, which gives a magnification of about 20.5 pixels mm⁻¹. The delay time between the plasma generation and the actual measurements is an essential parameter. For all the measurements, the time delays are individually adjusted with respect to the laser Q-switch delays having an accuracy of 2 ns. The ICCD camera is operated in shutter mode to allow very small (up to 2 ns) and precise integration times.

For the reference, we consider the optimized power conditions of the RF (13.56 MHz) and UHF (320 MHz) sources investigated earlier [28, 29] for deposition of nanocrystalline silicon films using PECVD processes. For this purpose, a gas mixture of SiH₄ (10 sccm) and H₂ (2000 sccm) is utilized. Then both the plasma sources are operated independently by increasing applied power, and the emission intensities from the various excited states of atomic hydrogen (H_α (656.2 nm), H_β (486.1 nm)) and molecular hydrogen along with the Si (288 nm) and SiH (410–425 nm) emission lines are monitored using the OES diagnostic. Measurements [28, 29] reveal (not shown here) that line intensities produced by the RF source are much lower than those of the UHF source. It is seen that the emission intensities increased with increasing RF and UHF power up to 180 W, and then saturated beyond 180 W. This experiment clearly suggests that, at high power (beyond 180 W), the individual operation of these (RF and UHF) sources would not be effective. However, one can combine RF and UHF sources simultaneously in dual-frequency (hybrid) operation for the PECVD process. Note that, in the hybrid operation, the total power is about twice that of the individual (RF/UHF) source power. However, for efficient radical and plasma production, the hybrid operation would be quite useful for the plasma process. Additionally, the emission intensities measured (not shown [28, 29]) in the hybrid plasmas are significantly (about two to four times) higher than that of individual source operated at high powers. Therefore, in the present experiment hybrid plasma refers to the simultaneous operation of both RF and UHF sources at about 180 W power.

Figure 2(a) presents various optical emission intensities, over a wide wavelength range of 300  −  700 nm, relevant to different excited species in the RF and hybrid plasma processes with the gas mixture of N₂ (1000 sccm) and SiH₄ (10 sccm). Spectra acquisitions are made from a viewport (figure 1(a)) with an exposure time of 0.5 s to diagnose the plasma emission using the spectrometer. It can be seen that the line intensities of the emission lines (figure 2(a)) produced in the hybrid plasma are much higher than those of the RF plasma. The OES diagnostic is carried out further (figure 2(b)) with the sequential addition of SiH₄ gas from 10 sccm to 100 sccm in the N₂ (1000–930 sccm) plasmas to investigate the optimize process conditions in the SiH₄–NH₃ processes. For clarity, only N₂ (337.1 nm) and N₂ (357.7 nm) line intensities are monitored. It can be seen from figure 2(b) that emission intensity rapidly increases up to the SiH₄ flow rate of 20 sccm. Further, the intensity does not change much, and it tends to saturate, at the higher flow rates. Accordingly, the gas flow rates of N₂ at 1000 sccm and SiH₄ at 20 sccm are considered as the optimized/favorable conditions and marked as a dashed line in figure 2(b) for the N₂–SiH₄–NH₃ deposition process. Then, NH₃ gas is added successively with different flow rates to the reactive mixture of N₂ (1000 sccm) and SiH₄ (20 sccm). Figure 2(c) presents the overall variation of intensities (of N₂ (337.1 nm) and N₂ (357.7 nm) emission lines) as a function of NH₃ flow rates. It is evident from figure 2(c) that even a small addition of NH₃, about 1% (10 sccm), can reduce the N₂ intensity significantly. Thus, the OES diagnostic suggests that the mixture of N₂ (1000 sccm)  +  SiH₄ (20 sccm)  +  NH₃ (10 or 20 sccm), would be the favorable conditions for the SiN_x:H process. Note also that the higher emission intensities, in figure 2, of the hybrid plasmas than of RF plasmas, correspond to more excitation and dissociation of radicals and molecules in the process plasmas. This nature of variation in the N₂–SiH₄–NH₃ plasmas is further investigated using LP and VUVAS diagnostics, which will be discussed later.

Zoom In Zoom Out Reset image size

Additionally, during the deposition process by addition of N₂, SiH₄, and NH₃, there will also be formation of hydrogen (H₂ gas and H radicals) and nitrogen species in the plasmas. Thus, to examine the gaseous flows or distribution of various species such as N₂ and H, we carry out measurements using the ICCD camera. As explained above, image intensities are measured using the camera corresponding to the favorable process conditions for the mixture of N₂ (1000 sccm) and SiH₄ (20 sccm). Then the effect of NH₃ addition to the N₂ (1000 sccm) and SiH₄ (20 sccm) mixture is also monitored. Figures 3(a) and (b) show the image intensities in two dimensions using filters of nitrogen and hydrogen, respectively. In the figures, the upper colored scale bar represents the magnitude of image intensities of the gaseous flows. All figures are shown with the same scale in the range from 1000 to 90 000 (in arbitrary units) to make a visual comparison between the RF and hybrid plasmas. The figure also depicts the relevant positions A, B, and C for the showerhead (of the RF source), UHF electrodes, and substrate, respectively. One can locate the relative intensities from the background color corresponding to flows of N₂ and H, represented by the black dotted circles, at the top of each scale bar.

Zoom In Zoom Out Reset image size

Figure 3(a) clearly shows the effectiveness of using hybrid plasmas. The two panels on the left-hand side present the image intensities in the absence of NH₃ (0 sccm flow rate). The panel at the bottom shows distinct regions of yellow and red colors that represent high image intensities of nitrogen flows generated by the hybrid plasmas. The corresponding intensities in RF plasmas, represented by the green and light-blue colors, are lower than those of hybrid plasmas. However, at the substrate location (point C) the figure shows moderate intensity (with light-green background) in the hybrid plasmas. With the addition of NH₃ with flow rate 10 sccm, there is a major change in the intensity. The yellow and red colors are replaced by a green background above the UHF electrodes. Further, the green background under the UHF electrodes is replaced by a semi-blue background (with the dotted circle in the substrate location), which corresponds to the reduction in the image intensity. With further increase of flow rate to 20 sccm, the N₂ species are confined in between the showerhead and the UHF electrodes. There is a noteworthy reduction in the intensity (represented by the dark-blue color) at the higher flow rate of 30 sccm. Note that these 2D plots represent images of the flow of radical species in plasmas during the deposition process. Further, the appreciable change in the intensity indicates that the plasma is only confined to nearby regions of the RF and UHF electrodes with higher NH₃ flow rates. Note also that the reduction in the image intensities is consistent with the OES data, as explained in figure 2(c) above. This modification in the image intensities indicates a considerable change in the plasma parameters that is to be explained further by LP data. In contrast to the intensities relevant to the N₂ flows, the corresponding intensities using the hydrogen filter are much lower, as shown in figure 3(b). In the absence of NH₃, there is a small trace of hydrogen intensities (represented by the green and yellow colors), which is confined to the UHF electrode only. Figure 3(b) shows that most parts of the figures are filled with the blue background color, which corresponds to a very low value of intensity in the scale bar.

For a better understanding of the plasma and radical formation during the PECVD process, figures 4(a)–(e) present the LP and VUVAS data taken in N₂–SiH₄ plasmas by using RF and hybrid power. The process condition is changed by varying the SiH₄ flow rate at a fixed operating pressure of 500 mTorr. It is obvious from figure 4(a) that T_e increases linearly from 3.0 to 9.0 eV, and 2.0 to 7.5 eV, in RF and hybrid plasmas, respectively. Figure 4(a) also shows that the value of T_e in RF plasmas is higher than that in hybrid plasmas, and the overall behavior of T_e variation is quite similar in the two cases. Note that the measured T_e is a thermalized population in the thermal equilibrium, which means that electrons with energy higher than T_e are present in the process plasmas. Additionally, to maintain the ionization and to sustain the discharge one requires the minimum energy to be approximately the ionization potential of nitrogen (~18 eV) [24]. However, the measured T_e values are much less than this ionization energy. In this sense, one can expect here the role played by the high-energy electrons to maintain the required ionization and dissociation processes. Note that, in the previous work [15], it has been shown that there is a significant high-energy electron tail in the EEDF using the hybrid plasmas (by adding the UHF source) that enables dissociation of N₂ and ionization of N atoms much better in the hybrid plasma than in the RF plasma. It can be seen that n_A (figure 4(b)) has a lower value in the RF plasmas than in hybrid plasmas. As expected, the absorption coefficient A (figure 4(c)) also acquires a lower value in the RF plasmas than in the hybrid case. Also, the profiles of n_A follow the corresponding profiles of A. Figure 4(d) shows that the n₀ value is much higher in the hybrid plasmas. Thus, LP and VUVAS measurements clearly show that both the ionization and dissociation in the hybrid plasmas are quite effective as compared to the RF plasmas. There is also similar variation of V_p (figure 4(e)) from about 108 to 125 V and 105 to 115 V in the RF and hybrid plasmas, respectively. One can translate this V_p into an equivalent energy (=  eV_p; e is the electron charge) for ions in plasmas. Now, consider the typical case of 20 sccm SiH₄ flow rate of figure 4. From figure 4(e), one can see that V_p  ≈  115 and 111 V in RF and hybrid plasmas, respectively. The corresponding ion energies (=  eV_p) are about 115 and 111 V in RF and hybrid plasmas, respectively. Note that the cross sections for charge exchange collisions of N ions (N⁺) with N₂ can be of three types: slow ions, $\text{N}_{2}^{+}$, and fast N (see figure 1 and table 1 of [46]). It can be seen (figure 1 of [46]) that the charge exchange cross section (σ_i) of slow ions is significantly higher (about one order of magnitude) than that of $\text{N}_{2}^{+}$. Also, below ion energy ~600 eV, σ_i is negligible [46]. Therefore, in the present study the ion charge exchange collision will mostly generate slow N⁺ ions. Further, the value of σ_i relevant to ion energy ~111 eV and 115 eV is about 5  ×  10⁻¹⁹ m², and the value of neutral nitrogen (N_neutral) at an operating pressure of 500 mTorr is about 1.6  ×  10²² m⁻³. Using these values of σ_i and N_neutral, one can calculate the ion charge exchange path length for nitrogen [46] as L_ch,_ex  =  1/(σ_iN_neutral)  ≈  0.1 mm. Corresponding to the plasma parameters, the values of Debye length (λ_d  =  √ (ε₀T_e/(n₀e²) and ε₀ is the permittivity of free space) are about 0.07 mm and 0.04 mm in the RF and hybrid plasmas, respectively. Considering a sheath thickness (w_s)  ≈  2–3 λ_d, one can see that L_ch,_ex ~ w_s. As mentioned earlier, the substrate is grounded in the present experiments, and there is no externally applied bias. Thus, the accelerated ions (by the V_p) crossing the sheath will undergo charge-exchange collisions in the sheath and do not affect the substrate much by losing their charge effectively.

Zoom In Zoom Out Reset image size

Note that, at 100 sccm SiH₄ flow rate, there is a significant improvement in n_A (figure 4(b)) due to the rapid enhancement of the absorption (high value of A, figure 4(c)) in the plasma. The high value of n_A can be attributed to high values of T_e. However, in figure 4(d), n₀ values of both RF and hybrid plasmas are close to each other at a very high flow rate (100 sccm) of SiH₄. To obtain a better understanding of the PECVD processes, one can examine the plasma process by the power balance estimation. It is obvious that, during the electron impact ionization and dissociation processes, power in the plasma is coupled through collisions. We recall that, in the process plasmas, the main precursor is the N₂ gas. Consider that an electron undergoes a collision with a neutral nitrogen gas molecule. After the collision event, its oscillatory motion will be affected, and momentum will also be randomized. In the steady state condition, the average energy (electron temperature)  ≈  T_e (in the unit of eV) is such that the energy imparted to the neutral gas in a collision is equal to the energy gained by electrons in the time between two successive collisions. Accordingly, the former will acquire a value of about T_e(2m_e/M_n) (here M_n and m_e are respectively, the mass of an electron and N₂). The latter can be written as ~P_a/ν_c (where P_a and ν_c are the power absorbed from the applied ac fields and the collision frequency respectively). Equating these terms, one will obtain P_a  =  T_e(2m_e/M_n)ν_c. Then, the power loss in the plasma per unit volume (P_loss) through collision can be expressed as

$$\begin{eqnarray}&&{{P}_{\text{loss}}}=\frac{2\,{{m}_{\text{e}}}}{{{M}_{\text{n}}}}{{T}_{\text{e}}}\,e\,\,{{\nu}_{\text{c}}}{{n}_{0}}\end{eqnarray}$$

To determine the power loss or power coupling in the plasma, consider the case of electron–neutral collision at a pressure of 500 mTorr (relevant to the N₂ flow rate variation from 1000 to 930 sccm for a SiH₄ variation of 0 to 100 sccm). Using the collision cross section data [24] for nitrogen and plasma parameters (figures 4(a) and (d)), one can determine the value of P_loss. For a simple calculation, one can assume the plasmas to be uniform and homogeneous. Relevant to the chamber height of about 22 cm and diameter about 27.2 cm, the volume of the plasma is V  =  0.013 m³. The product of P_loss and V can give the value of power loss P_col (in the unit of W) due to collision in plasmas. We recall here that the applied RF and hybrid power are 180 W and 360 W respectively. Figure 4(f) presents the estimated power loss in the plasma due to electron–N₂ collisions. For simplicity, the power loss in the coaxial transmission line (coaxial cable) and heating of chamber wall are ignored in the power estimation. Figure 4(f) shows that power loss in the plasma increases with increase in SiH₄ flow rate up to 20 sccm, and then it tends to saturate at a higher flow rate of SiH₄. It indicates that, with an increase in flow rate, most of the power will be spent in the dissociation of SiH₄ gas. This may be the reason why the n₀ values (figure 4(d)) of both RF and hybrid plasmas are somewhat closer at 100 sccm of SiH₄ due to closer values of P_col (figure 4(f)). From figure 4(f), it is seen that the significant (>60% of applied power) power loss takes place at a SiH₄ flow rate of 20 sccm. Accordingly, NH₃ is sequentially mixed with a fixed background of 20 sccm of SiH₄ for the N₂–SiH₄–NH₃ process, as shown in figure 5.

Zoom In Zoom Out Reset image size

Additionally, note that the information on plasma parameters is determined from I–V curves below the electron saturation region using the LP analysis procedure reported earlier [41]. Further, with an increase in the gas flow rate of SiH₄ there is a decrease in n₀ values and increase in T_e values (figure 4(a)). Moreover, the value of P_col (figure 4(f)) also decreases beyond the flow rate of 20 sccm. This variation indicates that, apart from the electron impact process with N₂, substantial power is also spent for the dissociation of SiH₄. Note also that sequential addition of SiH₄ would increase the electronegativity and produce dust particles. Thus, the negatively charged dust would enhance the bulk electric field and cause the electron energy (T_e) to increase [47, 48]. However, there is no trace of any oscillatory nature in the probe current of the I–V characteristic (not shown here) observed near the electron saturation region (which can occur due to negatively charged dust). Thus, the concern of dust particles in the LP data is believed to be not serious.

Note that, for the LP measurements, the LP is inserted into the discharge chamber from the side port (figure 1(a)). Note also that the LP is symmetrically placed with respect to the UHF and RF electrodes at the mid-plane of the chamber. Specifically, the LP is positioned in between the UHF electrodes (at a distance of 2.0 cm from each electrode of the UHF source and 4 cm from the shower head), which is well outside the sheath. In this sense, in the hybrid discharge, the LP is located near the UHF power source. One may consider here the possibility of the UHF source position compared to the LP position. However, our previous plasma diagnostic study [29], in the plasma processes (for the synthesis of crystalline), shows that the n₀ and T_e values measured independently in the RF and UHF plasmas are quite comparable. In particular, in the hybrid discharges, n₀ values are significantly higher, and T_e acquires lower values than that of RF and UHF discharges. Of course, the total applied power in the hybrid plasma is about twice that for individual operation of RF and UHF plasmas.

The plasma chemistry in the hybrid discharges can be visualized further by considering the concept of the plasma sheath in CCPs. A CCP can be modeled (not shown) as a series LCR resonance circuit [48, 49]. In this sense, one can consider the sheaths as capacitors in series with a resistive load (not shown). The LCR circuit consists of an inductor (L) representing electron inertia and sheaths as capacitors (C) in series with a resistive load (R) resembling power dissipation (ohmic heating) via elastic collisions. The characteristic resonance frequency (ω_res) of the circuit is given by [50]

$$\begin{eqnarray}&&{{\omega}_{\text{res}}}={{\omega}_{\text{pe}}}\cdot \sqrt{\frac{\delta}{{{L}_{\text{eff}}}}}\end{eqnarray} \tag{ 3 }$$

where the terms δ, L_eff, and ω_pe are respectively the mean sheath thickness, the effective bulk length between the electrodes, and the electron plasma frequency. Considering the situation of plasma excitation directly at resonance frequency ω_res, where the plasma can be efficiently heated, for typical n₀ ~ 10¹¹ cm⁻³ (in the present experiment), δ ~ 11 cm (=  c/ω_pe, c speed of light in vacuum), and discharge geometries (with L_eff ~ 9 cm), ω_res ~ 314 MHz. In this case, ω_res is much higher than the RF exciting frequency at 13.56 MHz. Therefore, electron resonance heating by low-frequency RF (13.56 MHz) can be avoided in the present scenario. This value (314 MHz) is closer to the UHF excitation frequency (320 MHz) used for the present case. Thus, there can be resonant heating by the UHF source when operated in the hybrid plasmas. Indeed, the EEDF in the case of hybrid plasmas has a significant high-energy tail, as reported earlier [15]. Also, efficient heating of the plasmas at ω_res depends on the nonlinearity (plasma sheath) of the system, which can allow the generation of higher-order harmonics of the RF (at fundamental) driving frequency in the plasma currents [51]. The detailed investigation of the electron heating in CCPs is beyond the scope of this paper.

In contrast to N₂–SiH₄ process plasmas (figure 4), T_e values (figure 5(a)) in N₂–SiH₄–NH₃ processes rapidly decrease on adding NH₃ even for a small amount of 10 sccm. At higher NH₃ flow rate (>10 sccm), T_e tends to saturate. Accordingly, n_A and A values in figures 5(b) and (c) are also much lower. Moreover, n₀ values, in both RF and hybrid plasmas (figure 5(d)), monotonically decrease with increasing NH₃ flow rate. Similar variations of V_p are also observed in figure 5(e). The power expended or power lost (P_col) by the electron with N₂, in figure 5(f), has a similar profile to T_e (figure 5(a)). The saturation of P_col at higher NH₃ flow rates (figure 5(f)) indicates that most of the power will be spent in the dissociation of NH₃ due to its lower dissociation energy. Further, with the increase in the NH₃ flow rate, T_e becomes small. Probably for this reason the n₀ value decreases significantly with the enhancement of NH₃ flow rate. Note that the high-energy tail of the EEDF is drastically suppressed with the addition of NH₃ [15]. This result suggests that films deposited with higher NH₃ flow rate would contain a higher fraction of –NH or –NH₂ groups. Note moreover that LP and VUVAS measurements are also consistent with those measured by the ICCD camera (figure 3), where the image intensities of the gas flows are very small even at low flow rate of NH₃. Therefore, the LP and VUVAS result indicates that low content of NH₃ would be a favorable condition for the synthesis SiN_x:H film.