Silicon oxynitride films deposited by reactive high power impulse magnetron sputtering using nitrous oxide as a single-source precursor

Silicon oxynitride thin films were synthesized by reactive high power impulse magnetron sputtering of silicon in argon/nitrous oxide plasmas. Nitrous oxide was employed as a single-source precursor ...


I. INTRODUCTION
2][3][4][5][6][7][8] One of the beneficial aspects of silicon oxynitride (SiO x N y ) thin films is the possibility to tailor the properties by adjusting the relative amounts of oxygen and nitrogen in the material.This yields properties ranging from and beyond those of amorphous silicon (a-Si), silicon nitride, and silicon oxide, or a mixture thereof.These possibilities are especially useful in optoelectronics, where SiO x N y has applications in graded refractive index layers, 5,9 antireflection coatings, 2 and optical waveguides. 10,11Various deposition methods have been used to grow SiO x N y films by varying oxygen to nitrogen ratios.These include chemical vapor deposition (CVD), 4,5,7 laser ablation, 9,12 plasma nitridation, 13,14 and magnetron sputtering. 3,6The use of CVD methods is limited due to hydrogen-containing precursors, as hydrogen has a deteriorating effect on the optical properties of SiO x N y thin films through the formation of N-H bonds. 10,11Conventionally two reactive gases, O 2 and N 2 , have been used to supply each element when SiO x N y has been synthesized by reactive magnetron sputtering. 1,2The two-gas approach is challenging as accurate control of both reactive gas flows is difficult due to nonlinear target effects as a function of the reactive gas flow rate. 15A reactive gas pulsing process can be used to overcome these nonlinear effects, but requires additional instrumentation to control the pulsing of the reactive gas. 16,17n order to tailor the reactive deposition process further, we chose reactive high power impulse magnetron sputtering (rHiPIMS) as the deposition method.High power impulse magnetron sputtering (HiPIMS) is an ionized physical vapor deposition technique based on conventional direct current magnetron sputtering. 18,19In HiPIMS, short high voltage pulses are delivered to the cathode, resulting in highly ionized and dense plasmas for target metals. 191][22] HiPIMS also offers a possibility to affect the film properties by adjusting the pulse-related deposition parameters, namely, the pulse frequency and energy per pulse. 19,23Reduced or even eliminated hysteresis effects were also reported for HiPIMS. 24][27][28][29][30][31][32] To eliminate the nonlinear effects of having two reactive gases, we employed nitrous oxide (N 2 O) as a singlesource precursor gas in the rHiPIMS deposition of SiO x N y .The behavior of nitrous oxide during the reactive sputter deposition was not yet investigated, but its effects on the plasma chemistry can be predicted by consideration of the ionization and dissociation pathways of the molecule.
Table I summarizes relevant data for the Si/Ar/N 2 O discharge.As can be seen, the ionization energy of N 2 O [E P;N 2 O ¼ 12:9 eV (Ref.35)] is lower than the ionization energy of Ar [E P,Ar ¼ 15.8 eV (Ref.34)].Upon electron impact, the N 2 O molecule is either ionized or undergoes dissociative ionization along the following channels: a) Electronic mail: tuoha@ifm.liu.se As seen from the appearance energies of the N 2 O dissociation products in Table II, the ionized fragments containing oxygen, like NO þ and O þ , appear at lower energies than the nitrogen species N þ 2 and N þ .Additionally, the partial electron impact ionization cross sections presented in Table II imply that low electron energies favor the production of oxygen-containing species.Furthermore, the production of O À ions through dissociative electron attachment has also been shown to contribute to the splitting of the N 2 O molecule 40 These ions were shown to exhibit relatively long lifetimes during the discharge afterglow between the HiPIMS pulses. 41The presence of O -is observed already at zero electron energies at gas temperatures above the room temperature. 42The reaction cross section increases with increasing gas temperature and values are of the same order of magnitude as the electron impact ionization cross sections for N 2 O. 42 As can be understood from the dissociation scheme, the presence of multiple ion species and precursor fragments can be expected in the rHiPIMS plasma.
In this article, we investigate the effects of rHiPIMS process parameters on the resulting SiO x N y thin film properties.Furthermore, the effects of the reactive gas flow, pulse frequency, and energy per pulse on the chemical composition of the films and their chemical bonding structure are presented and discussed.

II. EXPERIMENTAL METHODS
All films were deposited with the industrial coating system CC800/9 (CemeCon AG, Germany). 43In our experiments, one rectangular silicon target (area 440 cm 2 ) was sputtered in Ar/N 2 O atmosphere keeping a constant deposition pressure of 400 mPa.The cathode was operated in power-regulated HiPIMS mode with a pulse width of 200 ls for all deposition processes.The substrates faced the target at a distance of 60 mm during the depositions.Moreover, a pulsed bias voltage (V b ) of À100 V, synchronized with the cathode pulse, was applied to the substrate table.Depositions were performed at a substrate temperature of 350 C. The films were deposited on boron doped Si(100) substrates, using the above-mentioned settings.Three different deposition series were prepared; (1) variation of the percentage of nitrous oxide in the plasma, (2) variation of the pulse frequency, and (3) variation of the pulse energy.
For (1), the amount of N 2 O in the working gas was varied by adjusting the flows of Ar and N 2 O where f N 2 O is the flow of nitrous oxide and f Ar is the flow of argon, respectively.The flow of N 2 O was varied from 0 to 25 sccm, corresponding to f N 2 O=Ar of 0% to 6.2%.Here, a pulse frequency of 600 Hz and an average cathode power of 2400 W, resulting in pulse energies of $4 J were used.For (2), different frequencies of 200, 400, 600, and 800 Hz were studied, using average cathode powers of 800, 1600, 2400, and 3200 W, respectively.The power was varied in order to maintain a pulse energy of $4 J.The reactive gas flow was kept at 10 sccm, corresponding to an f N 2 O=Ar of 2.4%.For (3), the pulse energy was varied from 2 to 6 J in 1 J steps by changing the average target power from 1200 to 3600 W in 600 W steps, while the frequency was kept at 600 Hz.Here, an f N 2 O=Ar of 2.4% was used.Target current and voltage waveforms during depositions were recorded with a Tektronix DPO4054 500 MHz bandwidth digital oscilloscope.Cross-sectional scanning electron microscopy (SEM, LEO 1550 Gemini, Zeiss, Germany) was used to measure film thicknesses and to investigate the film morphology.To assess the structural properties of the films, x-ray diffraction (XRD) was carried out.A Philips powder diffractometer (PW 1820) equipped with a Cu(Ka) radiation source was operated at 40 kV and 40 mA, in order to record h/2h scans.The residual film stresses were assessed by wafer curvature method, using XRD (PANalytical Empyrean) operated at 45 kV and 40 mA. 44The Stoney formula for anisotropic single crystal Si(100) was applied to relate the measured substrate curvature to the residual thin film stress, assuming uniform plane stress in the film 45 Species Appearance energy (eV) 12.9 (Ref.where r f is the in-plane stress component in the film, t f is the film thickness, h is the substrate thickness, M Si ð100Þ is the biaxial modulus of Si(100) (180.3GPa), and R is the radius of the curvature of the substrate.The same instrument was utilized to perform x-ray reflectivity (XRR) measurements.PANalytical X'Pert Reflectivity software was used to iteratively fit the measured XRR curves to evaluate the film density.A layered model containing the substrate, the film, and oxides both on the film and the substrate was assumed.
Time-of-flight elastic recoil detection analysis (ERDA) was carried out to obtain the elemental composition of the films. 46,47A 36 MeV 127 I 8þ ion beam with an incident angle of 22.5 relative to the sample surface was used.To study chemical bonding in the films, x-ray photoelectron spectroscopy (XPS, Axis Ultra DLD , Kratos Analytical, Manchester, UK) with monochromatic Al(Ka) x-rays (h ¼ 1486.6 eV) was employed.The base pressure in the analysis chamber remained below 1 Â 10 À7 Pa during acquisition.Core level spectra of the Si 2p, N 1s, O 1s, Ar 2p, and C 1s regions were recorded on as-deposited samples and after a sputter clean of 120 s with 2 keV Ar þ beam rastered over 3 Â 3 mm 2 surface area at an incident angle of 70 with respect to the surface normal.Sputter cleaning resulted in a decreased resolution of the Si 2p core level spectra components, suggesting structural modification to the films due to the sputter ion impact.Therefore, the core level spectra acquired from the as-deposited samples were analyzed.Automatic charge compensation was used during the acquisition of all spectra.A Shirley-type background was subtracted from all spectra prior to peak fitting.Voigt peak shapes with the Lorentzian contribution of 30% were used to model the chemical structure of the SiO x N y films.In order to ensure a reliable peak fit model, Si 2p spectra were recorded for thermally grown silicon dioxide (SiO 2 ), CVD-grown silicon nitride (Si 3 N 4 ), and pure Si deposited by comparable deposition parameters serving as internal references for the SiO x N y films.These reference samples were used to identify the number of components in the Si 2p region.All spectra were referenced to C-C bond, at 284.8 eV. 48pectroscopic ellipsometry was used to study the optical constants of the films.The measurements were performed with a variable angle spectroscopic ellipsometer (J.A. Woollam Co.) at four incident angles 45 , 55 , 65 , and 75 over a wavelength range of 245-1690 nm.The data were analyzed with a COMPLETEEASE software version 4.72 and fitted with a Tauc-Lorentz model for amorphous films to assess their optical properties. 49

A. Process characteristics
Figure 1 shows a typical current and voltage waveform recorded during the rHiPIMS discharge of Si in Ar/N 2 O plasmas.For the investigated range of processes, the target current and voltage waveforms did not change significantly with respect to their onset of rise, slope, and peak values.As shown in the inset of Fig. 1, increasing nitrous oxide flows during deposition results in slightly lower peak target currents, which is commonly ascribed to target poisoning. 50owever, within the range of studied reactive gas flows, the deposition rates did not change significantly, indicating that all processes were carried out in the metallic or transition region of the reactive discharge.Thus, the target was not poisoned and metallic target surface conditions can be assumed. 51This suggests that the drop in the peak target current is mainly caused by a decreased secondary electron emission yield (c SE ) as a consequence of increased N 2 O flows. 52The secondary electron emission yield was calculated according to the below equation: 53 where E P is the first ionization energy of the arriving ion and / is the work function of the target surface.When estimating c SE for Si discharges in N 2 O and Ar using E P values of 12.9 and 15.8 eV, respectively, and a value of 4.6 eV (Ref.54) for the silicon target work function, the secondary electron yield by Ar þ bombardment is approximately four times higher compared to the case when pure N 2 O is used.Therefore, the drop of 3 A in peak target current can be ascribed to the increase in N 2 O, as 25 sccm corresponding to a f N 2 O=Ar of 6.2% is used.Moreover, the plasma density may be slightly reduced upon introduction of N 2 O, due to its dissociation into precursor fragments and their further ionization.This may contribute to reduced peak target currents.However, considering the low N 2 O ionization energy of 12.9 eV and comparing this to ionization and appearance energies of Ar þ (cf.that the direct ionization of the N 2 O molecule is most probable.Hence, the effect of a reduced plasma density on the peak target current upon introduction of 25 sccm N 2 O is considered to be minor.The decrease in target current is accompanied by an increase in the target voltage (not illustrated), since the processes were carried out in power-regulated mode.A decrease of the peak target current is not observed for depositions with varying pulse frequencies at an N 2 O flow of 10 sccm, implying equal target surface chemistries and secondary electron yields for these processes.Raising the energy per pulse results in inherently increasing peak target currents and voltages.Here, the target voltage waveforms show an increasing voltage drop as the target current reaches its maximum.To conclude, an altered target surface or secondary electron yields cannot be drawn from these waveforms.

B. Thin film characterization
Cross-sectional SEM shows a dense and featureless morphology with a smooth surface structure for all investigated films.The films show gray and shiny appearance without visible adhesive failures upon ocular inspection.A typical cross-sectional SEM image is presented in Fig. 2 for a SiO x N y film deposited with a pulse frequency of 200 Hz and an N 2 O flow of 10 sccm.The deposition rates scale with the frequency and pulse energy.Specifically, pulse energies of 3 and 6 J yielded 1.8 and 3.0 nm/s, respectively, while the deposition rate at 2 J was only 1.0 nm/s.According to h/2h scans, the films are x-ray amorphous.Residual film stresses and densities did not show significant dependencies on the investigated parameters.The compressive stresses ranged between À650 and À960 MPa while a density of 2.45 6 0.15 g/cm 3 was determined for all films.The obtained film densities are closer to that of a-Si ($2.3 g/cm 3 ) 55 than to a-SiO x ($2.1 g/cm 3 ) 56 or a-SiN x ($3.0 g/cm 3 ), 57 which agrees with the silicon-rich atomic composition of the SiO x N y films (see next paragraph).
The atomic concentrations of oxygen and nitrogen in the films as obtained by ERDA are shown in Figs.3(a)-3(c).The overall film oxygen and nitrogen content is influenced by the N 2 O flow, while the film O/N-ratio is tuned with the pulse frequency and pulse energy.As shown in Fig. 3(a), the O and N content of the SiO x N y films increase as the N 2 O flow increases.The O/N-ratio is not significantly affected by the N 2 O flow.In Fig. 3(b), the effect of increasing frequencies from 200 Hz to 800 Hz and average target powers from 0.8 to 3.2 kW on the film composition is shown.Here, a strong reduction of the O content by $8 at.% is observed, while the N content is hardly affected.The change in composition corresponds to O/N-ratios between $1.8 and $0.8. Figure 3(c) shows the O and N content with increasing pulse energies.Increasing the pulse energy from 2 6 J results in a decreased film O content from $13 to $5 at.%, while the N content does not show a strong dependence on the pulse energy.The decreasing O content results in a reduction of the film O/N-ratio from $1.9 to $0.9.In addition, ERDA reveals that the films contain argon ($3 at.%) and trace amounts of carbon (Շ0.1 at.%).
The oxygen surplus at low pulse frequencies and pulse energies is ascribed to a preferred formation of film forming O-containing precursor species.Low energies per pulse and low average target powers promote the creation of NO þ and O þ , but not N þ and N þ 2 , since the oxygen-containing species show lower appearance energies (cf.Table II).The inherently high reactivity of oxygen, elevated sticking coefficient, as well as the higher electronegativity compared to nitrogen, also contributes to the surface reactions at the substrate and the target, leading to an oxygen surplus in the films. 51,58At the same time, for the depositions using 10 sccm N 2 O, changes regarding the peak target current were not observed when varying pulse frequencies while keeping the energy per pulse at 4 J.This suggests that the target surface is not affected noticeably by a lowered average power and increased pulse off-times during the depositions.The film oxygen surplus is most distinct when low pulse frequencies and thus extended pulse off-times or low pulse energies and thus reduced sputter rates are applied.Here, a higher percentage of the sputtered material is available for chemical reactions with reactive gaseous species at the substrate.The composition of the films is supposed to be mainly determined by reactions at the substrate surface as the ionization mean free path of the sputtered target material is estimated to be over twice as large as the target-substrate distance (60 mm). 59This decreases the probability of gas-phase reactions.In this context, it should be noted that the f N 2 O=Ar was only 2.4% during depositions, reducing the probability for gas-phase reactions further.The composition of SiON thin films deposited at pulse frequencies >500 Hz and energies >4 J is determined by the raised sputter rate of atomic Si.Consequently, the O and N contents in the films are reduced.Additionally, high energies per pulse lead to gas rarefaction in front of the target.Gas rarefaction lowers the collision probability of the sputtered material with the gaseous species further and yields even less amount of deposited compound material. 59In contrast to the considerably reduced oxygen contents in the films, the nitrogen content is hardly affected by the change in energetics of the deposition process.According to appearance energies and energies for dissociative ionization (cf.Table II and dissociation reactions), the steady N content in the films at increased pulse energies is understood to be a consequence of the favored formation of N þ and N þ 2 .The chemical bond structure of the SiON films was assessed by the evaluation of the XPS Si 2p core level spectra obtained from as-deposited samples.Due to the amorphous nature of the films, it can be assumed that Si is bond to Si, O, and N in a random manner.Contributing to this assumption is the fact that HiPIMS processes are far from thermal equilibrium, 19 supporting a random bond structure in the films. 60As a consequence of the stochastic nature of the bond formation process and thus the lack of repeating unit cells, as well as the existence of nearest-neighbor effects, broad bond contributions of up to eV in FWHM are observed.Due to the comparatively low O and N contents in the films, all O and N is bond to Si.This is also corroborated by corresponding N 1s and O 1s core level spectra, presenting one broad, featureless peak, indicating no O-N bonding in the films.
Figures 4-6 show the Si 2p core level spectra of the SiO x N y films deposited with varying N 2 O flows, pulse frequencies, and pulse energies, respectively.Based on the information acquired from our reference samples, the Si 2p core level spectra were deconvoluted into five components, two for the characteristic Si 2p doublet (Si 2p 3=2 at 98.85 6 0.10 eV, Si 2p 1=2 at 99.45 6 0.10 eV), one component assigned to the Si-N bond at 99.90 6 0.10 eV, another to the Si-O bond at 102.70 6 0.10 eV, and an intermediate contribution at 101.10 6 0.15 eV between the peaks referred as Si-O/N.This contribution is assigned to Si-N, which is affected by O as next neighbor.The SiO x N y deposited at f N 2 O=Ar of 1.2% [Fig.4(a)] is an exception with respect to the intermediate peak position as this peak is found at 101.60 eV.The deviation is attributed to surface oxide as this particular film presents the least O and N contents of all films.As the spectra were obtained from as-deposited samples, the relative increase of O at the surface contributes to a shift toward higher binding energies.Due to the amorphous nature of the films, the peaks assigned to O and N bond contributions are comparatively broad and thus set a limit to the accuracy of the peak fit model regarding the different bonding states in the films.The bonding configuration in amorphous SiO x N y films have been shown to be many fold, and the choice of applicable bond models depends also on the deposition temperature as well as on other parameters such as ion bombardment affecting film growth dynamics during deposition. 60,61he relative contributions of the three elements Si, N, and O to the total Si 2p area are shown in Fig. 7.For this purpose, the Si-O/N component was split between these two based on the O/N-ratio of the film.Increasing N 2 O flows result in an increased number of Si-O and Si-N bonds, correlating with the increasing atomic concentrations of both elements [cf.Fig. 7(a)].This relation is not as obvious for the frequency and pulse energy series; as the O/N-ratio in the films ranges between 1.9 and 0.8, the peak position of the intermediate peak is shifted toward lower binding energies with decreasing O/N-ratio, i.e., decreasing frequency or pulse energy (cf.Figs. 5 and 6).The shift of the contribution assigned to Si-O/N can be explained by reduced influence of O as the O/N-ratio decreases.A more detailed analysis of the bonding is not pursued, as our model contains only three components for Si-O/Si-N bonds.

C. Film optical properties
The optical properties of the films show a strong correlation to the percentage of both oxygen and nitrogen in the films.The film refractive indices n and extinction coefficients k were obtained from the Tauc-Lorentz model fitted to the ellipsometric data collected for each sample. 62efractive indices and extinction coefficients for all films at the wavelength of 633 nm are shown in Fig. 8. Refractive index values for amorphous Si (dashed line at n ¼ 4.5), SiO 2 (dashed-dotted line at n ¼ 1.45), and SiN x (short dashed line at n ¼ 2.0), as well as the extinction coefficient value of a-Si (dotted line at k ¼ 0.38) are indicated as horizontal lines in the subfigures.Values of a-Si and SiO 2 were recorded for the reference samples used in XPS, n for SiN x from Ref. 63.
As is shown in Figs.8(a)-8(c), increased N 2 O flow rates as well as decreased pulse frequencies and energies result in equivalent film optical properties.Both n and k follow the film elemental composition; increased total O and N contents yield lower values.The shape of the n and k dispersion curves remains the same due to comparable elemental composition and morphology of the films. 63As the films are Si-rich the n and k values remain still closer to a-Si than to SiO x N y , i.e., the films have nonzero extinction coefficients and high refractive indices. 63Due to the comparatively low film O and N concentrations the optical properties are

IV. CONCLUSIONS
Silicon-rich silicon oxynitride films were synthesized by rHiPIMS at a constant process pressure of 400 mPa using different N 2 O/Ar flow ratios.The amount of oxygen and nitrogen in the films can be controlled by adjusting the N 2 O flow to affect the concentrations of both oxygen and nitrogen.The film O/N-ratio can be tuned in the range of 0.8-1.9 by changing the pulse frequency between 200 and 800 Hz while maintaining a pulse energy of 4 J, or by changing the energy per pulse between 2 and 6 J, with higher frequencies and pulse energies resulting in lower O/N-ratios.Under these deposition conditions, the films are amorphous and exhibit random chemical bonding.Optical properties of the films are governed by their Si-rich nature, resulting in refractive indices and extinction coefficients that are closer to a-Si values than to silicon oxynitride.The control of O/N-ratio by pulse frequency and energy poses pathways to tailor the film chemical composition from O-rich SiON to N-rich SiON.

FIG. 2 .
FIG. 2. Cross-sectional SEM image of a film deposited with pulse frequency of 200 Hz at N 2 O flow of 10 sccm.Charging effects due to the insulating properties of the film are visible in the image.

FIG. 4 .
FIG.4.XPS Si 2p core level spectra of the films deposited with different N 2 O/Ar flow ratios.Flow ratios, core level components, as well as component positions are indicated.

FIG. 5 . 6 J
FIG.5.XPS Si 2p core level spectra of the films deposited with different pulse frequencies.Frequencies, core level components, as well as component positions are indicated.

FIG. 8 .
FIG.8.Refractive indices n and extinction coefficient k for SiO x N y films deposited by variation of (a) the f N2O=Ar , (b) the pulse frequencies, and (c) the pulse energies.The refractive indices for a-Si (dashed line at n ¼ 4.5), SiO 2 (dashed-dotted line at n ¼ 1.45), and SiN x (short dashed line at n ¼ 2.0) as well as the extinction coefficient of a-Si (dotted line at k ¼ 0.38) are indicated as horizontal lines in the subfigures.The inset of (b) shows the n-dispersion for films deposited with varying pulse frequencies whereas the inset of (c) shows the k-dispersion for films deposited with different average powers.

TABLE I .
First ionization energies of Si, Ar, and N 2 O plus total electron impact cross sections for Ar and N 2 O.

TABLE II .
Appearance energies and partial electron impact cross sections for N 2 O fragments upon electron impact.