Systematic Study of the SiOx Film with Different Stoichiometry by Plasma-Enhanced Atomic Layer Deposition and Its Application in SiOx/SiO2 Super-Lattice

Atomic scale control of the thickness of thin film makes atomic layer deposition highly advantageous in the preparation of high quality super-lattices. However, precisely controlling the film chemical stoichiometry is very challenging. In this study, we deposited SiOx film with different stoichiometry by plasma enhanced atomic layer deposition. After reviewing various deposition parameters like temperature, precursor pulse time, and gas flow, the silicon dioxides of stoichiometric (SiO2) and non-stoichiometric (SiO1.8 and SiO1.6) were successfully fabricated. X-ray photo-electron spectroscopy was first employed to analyze the element content and chemical bonding energy of these films. Then the morphology, structure, composition, and optical characteristics of SiOx film were systematically studied through atomic force microscope, transmission electron microscopy, X-ray reflection, and spectroscopic ellipsometry. The experimental results indicate that both the mass density and refractive index of SiO1.8 and SiO1.6 are less than SiO2 film. The energy band-gap is approved by spectroscopic ellipsometry data and X-ray photo-electron spectroscopy O 1s analysis. The results demonstrate that the energy band-gap decreases as the oxygen concentration decreases in SiOx film. After we obtained the Si-rich silicon oxide film deposition, the SiO1.6/SiO2 super-lattices was fabricated and its photoluminescence (PL) property was characterized by PL spectra. The weak PL intensity gives us greater awareness that more research is needed in order to decrease the x of SiOx film to a larger extent through further optimizing plasma-enhanced atomic layer deposition processes, and hence improve the photoluminescence properties of SiOx/SiO2 super-lattices.

opto-electronic technologies in the future [3]. Since the discovery of efficient photoluminescence in the red region in porous Si at room temperature [4], silicon nanocrystals (Si-NCs) were extensively studied in the last decade [5,6]. Particularly in recent years, Si-NCs have aroused significant attention in the applications of light sources and "all-silicon" devices like silicon lasers, light emitting diode (LED), flash memories [7,8], and tandem solar cells. Among the different techniques of Si-NCs fabrication in a solid matrix, super-lattices were considered to be very effective in producing size-, density-, and shape-controlled Si-NCs [9][10][11].
A large number of studies have been reported on Si-NCs embedded in silicon dioxide (SiO 2 ) [12][13][14]. Based on the quantum confinement theory [15][16][17], the correlation between the photoluminescence (PL) properties and the size of the nanocrystals has been established. Control over all three parameters (size, density, spherical shape) can be reached by depositing thin alternating layers of stoichiometric and Si rich dielectrics in the form of a superlattice (SL). This approach has been well established for silicon nanocrystals in SiO 2 matrix. However, most of the Si-NCs and superlattices in previous studies are fabricated by chemical vapor deposition (CVD) [18][19][20], sputtering [21][22][23], or other deposition techniques [24], these techniques are good at depositing thick film but weak in thin film. Which hamper the further study on the control of Si-NCs in superlattice with ultrathin dielectric film or barrier layer. Atomic layer deposition (ALD) is a promising technology for advanced thin film deposition as it offers excellent control at the atomic scale over the thickness and uniformity of the film [25][26][27]. It allows the precise preparation of size-and distribution-controlled silicon nanocrystals. So, it will be a wonderful opportunity to fabricate silicon oxides and related superlattice by ALD technique, and study its photoluminescence properties.
There have been a certain amount of studies on the growth of SiO 2 film using ALD [28][29][30], but rarely on the control of its stoichiometry and optical/electrical properties. Thus, in this study, plasmaenhanced atomic layer deposition (PEALD) was used to deposit SiO x film with different stoichiometry. In detail, ALD deposition parameters were firstly studied to understand the effect of deposition parameters like precursor pulse time and temperature on the growth properties of SiO x film. Upon choosing suitable deposition parameters, three kinds of film (SiO 2 , SiO 1.8 , and SiO 1.6 ) with different stoichiometry were acquired successfully. Film properties like physical, chemical, and photo-electrical properties of SiO x film with different stoichiometry were studied systematically. Specifically, basic film characteristics including microstructure, density, and roughness were evaluated, while the chemical bonding character of the obtained film was discussed by Si 2p and O 1s in detail. Furthermore, the energy band-gap of the SiO x film with different stoichiometry was identified by both spectroscopic ellipsometry (SE) data analysis and X-ray photo-electron spectroscopy (XPS) measurements. Finally, to examine the properties of Si rich SiO x obtained by ALD in this study, SiO 1.6 /SiO 2 superlattice was fabricated and its photoluminescence property was characterized by PL spectra.

Film Preparation
SiO x films were deposited on p-type (1-10 Ω·cm), single polished, Si(100) wafers in a BENEQ TFS200 ALD system (BENEQ, Finland) at a vacuum degree of 1 mbar. Prior to deposition, the Si wafers were cleaned by a standard RCA process followed by a deionized water rinsing and drying in N 2 . During the deposition process, precursors of Si and O were tris(dimethylamino)silane (TDMAS) and O 2 plasma. TDMAS was purchased from Fornano company and maintained at 20 • C in a stainless bottle. The O 2 plasma was activated at 200 W. The schematic diagram of one ALD cycle of SiO x growth utilized in this work is illustrated in Figure 1a. Each ALD cycle contains four steps: TDMAS pulse, Ar purge, plasma processing, and Ar purge.

Sample Characterization
The surface morphology was observed using an atomic force microscope (AFM, Bruker, icon), and a typical 5 μm × 5 μm area was investigated using non-contact mode. Transmission electron microscopy (TEM) (FEI, TECNAI G2 F20) was used to analyze the micro-structure and interface composition of the silicon oxides. The chemical bonding character of the obtained film was characterized by XPS (SPECS, Berlin, Germany) using a mono-chromatic Al Kα source (hν = 1486.6 eV). A narrow scan resolution of 0.1 eV was used. The adventitious C 1s peak, arising from traces of hydrocarbon in the spectrometer, was used as a reference for evaluating the peak positions because of static charging of samples. The C 1s peak position was observed together with other peaks (Si 2p, N 1s, and O 1s) of the spectrum, and all the XPS spectra were calibrated by the C 1s peak at a binding energy of 284.6 eV. The micro-structure and morphology of the film were characterized by X-ray reflection (XRR) (Bruker, D8, Billerica, MA, USA). SE measurements were carried out on a rotating analyzer ellipsometer (SOPRA, GES-5E, Annecy, France). The incident angle was 75°. The spectral wavelength range from 190 to 800 nm, the system measured the spectra of Ψ and Δ as functions of wavelength (λ). The resulting spectra was fitted with WinElli_II software. PL spectroscopy was used to investigate the optical properties of Si-NCs produced within the SiOx/SiO2 super-lattice. The PL was excited by the 325 nm line of HeCd laser. The PL signal was focused into a single monochromator and detected by a nitrogen-cooled charge-coupled device (CCD) camera. All spectra were corrected for the spectral response of the measurement system.

Film Fabrication and Growth Rate Experiments
In order to optimize the growth parameters needed for the self-limiting deposition of SiOx thin film, the effect of temperature, TDMAS dose, and O2 plasma duration was studied. At first, with the purpose of studying the effect of temperature on growth rate, 100 cycles with 2 s TDMAS and 8 s O2 plasma was deposited at different temperatures (100-250 °C). The growth rate increased initially as the temperature was increased from 100 to 175 °C. However, it stayed nearly constant at an ALD temperature window from 175 °C to 250 °C, as seen in Figure 1b

Sample Characterization
The surface morphology was observed using an atomic force microscope (AFM, Bruker, icon), and a typical 5 µm × 5 µm area was investigated using non-contact mode. Transmission electron microscopy (TEM) (FEI, TECNAI G2 F20) was used to analyze the micro-structure and interface composition of the silicon oxides. The chemical bonding character of the obtained film was characterized by XPS (SPECS, Berlin, Germany) using a mono-chromatic Al Kα source (hν = 1486.6 eV). A narrow scan resolution of 0.1 eV was used. The adventitious C 1s peak, arising from traces of hydrocarbon in the spectrometer, was used as a reference for evaluating the peak positions because of static charging of samples. The C 1s peak position was observed together with other peaks (Si 2p, N 1s, and O 1s) of the spectrum, and all the XPS spectra were calibrated by the C 1s peak at a binding energy of 284.6 eV. The micro-structure and morphology of the film were characterized by X-ray reflection (XRR) (Bruker, D8, Billerica, MA, USA). SE measurements were carried out on a rotating analyzer ellipsometer (SOPRA, GES-5E, Annecy, France). The incident angle was 75 • . The spectral wavelength range from 190 to 800 nm, the system measured the spectra of Ψ and ∆ as functions of wavelength (λ). The resulting spectra was fitted with WinElli_II software. PL spectroscopy was used to investigate the optical properties of Si-NCs produced within the SiO x /SiO 2 super-lattice. The PL was excited by the 325 nm line of HeCd laser. The PL signal was focused into a single monochromator and detected by a nitrogen-cooled charge-coupled device (CCD) camera. All spectra were corrected for the spectral response of the measurement system.

Film Fabrication and Growth Rate Experiments
In order to optimize the growth parameters needed for the self-limiting deposition of SiO x thin film, the effect of temperature, TDMAS dose, and O 2 plasma duration was studied. At first, with the purpose of studying the effect of temperature on growth rate, 100 cycles with 2 s TDMAS and 8 s O 2 plasma was deposited at different temperatures (100-250 • C). The growth rate increased initially as the temperature was increased from 100 to 175 • C. However, it stayed nearly constant at an ALD temperature window from 175 • C to 250 • C, as seen in Figure 1b, where deposition rate was constant at 0.07 nm/cycle. Figure 1c,d display the growth rate of the SiO x thin film as a function of plasma time and TDMAS pulse time at a deposition temperature of 200 • C. The O 2 flow rate was fixed at 50 sccm. The growth rate was determined here in terms of the film thickness divided by the total number of applied ALD cycles. The film thickness was 7.08, 9.8, 11.96, 13.56, 14.02, 14.13, and 14.21 nm for the samples with the plasma pulse time of 2, 4, 6, 8, 10, 12, and 14 s, respectively. The total ALD cycles were kept at 200 for the simulation. It can be observed from Figure 1c that the growth rate increases initially as plasma time increases. It remained constant at 0.071 nm/cycle when the plasma time was greater than 8 s. To obtain the saturated TDMAS pulse time, the plasma pulse time was kept at 8 s. The film thickness became 3.8, 5.4, 6.7, 7.08, 7.16, and 7.12 nm when the TDMAS pulse time was set as 0.5, 1, 2, 4, 6, and 8 s, respectively. As seen in Figure 1d, deposition rate was saturated for TDMAS pulse time starting from 2 s. These results suggest that the SiO x thin film was grown in a self-limiting manner when using the PEALD technique. Figure 2 shows the 5 µm 2 × 5 µm 2 AFM images for SiO 2 thin films with different deposition temperatures. It is clear that the morphology changes with different temperatures. As the deposition temperature increases, the resulting morphology transforms from an island structure into a slim needle like structure. The surface root mean-square (RMS) roughness becomes lower with the increase in deposition temperature. For SiO 2 thin film deposited at 100 • C, it presents the maximum roughness of 0.52 nm, then the value decreases to 0.43, 0.27, and 0.26 nm for the films deposited at 150, 200, and 250 • C, respectively. It was found that the surface roughness of all the films prepared in this work is very small, indicating that the films have a smooth surface.  Figure 1c that the growth rate increases initially as plasma time increases. It remained constant at 0.071 nm/cycle when the plasma time was greater than 8 s. To obtain the saturated TDMAS pulse time, the plasma pulse time was kept at 8 s. The film thickness became 3.8, 5.4, 6.7, 7.08, 7.16, and 7.12 nm when the TDMAS pulse time was set as 0.5, 1, 2, 4, 6, and 8 s, respectively. As seen in Figure 1d, deposition rate was saturated for TDMAS pulse time starting from 2 s. These results suggest that the SiOx thin film was grown in a self-limiting manner when using the PEALD technique. Figure 2 shows the 5 μm 2 × 5 μm 2 AFM images for SiO2 thin films with different deposition temperatures. It is clear that the morphology changes with different temperatures. As the deposition temperature increases, the resulting morphology transforms from an island structure into a slim needle like structure. The surface root mean-square (RMS) roughness becomes lower with the increase in deposition temperature. For SiO2 thin film deposited at 100 °C, it presents the maximum roughness of 0.52 nm, then the value decreases to 0.43, 0.27, and 0.26 nm for the films deposited at 150, 200, and 250 °C, respectively. It was found that the surface roughness of all the films prepared in this work is very small, indicating that the films have a smooth surface. Then, the three sample sets (S1, S2, and S3) were prepared in order to obtain SiOx film with different stoichiometry. In ALD deposition, the temperature was fixed at 250 °C. The precursor pulse time was changed in order to obtain silicon oxides of stoichiometric (SiO2) and non-stoichiometric (SiOx, x < 2). To be specific, TDMAS pulse time was 2, 2, and 4 s for S1, S2, and S3 sample, respectively, O2 plasma duration was 8, 4, and 2 s for S1, S2, and S3 film, respectively. The purge time was always kept at 4 s no matter whether it was after the TDMAS pulse or after the O2 plasma. Meanwhile, in order to decrease the oxygen content in the film to form Si rich silicon dioxide, the O2 flow rate was also adjusted, it was 50 sccm, 50 sccm, and 25 sccm on the deposition of S1, S2, and S3 film, respectively.
After deposition, all samples were tested by XPS to determine the stoichiometric. Figure 3 displays the XPS wide scan spectrum of S1, S2, and S3 film after Ar + ion treatment. The determined XPS peaks of Si, O, and C are stated in the figure. All the peaks of elements are marked and recognized in detail. By comparing the content of Si and O element in three samples, it is found that the stoichiometries are x = 2, 1.8, and 1.6 for S1, S2, and S3 film, respectively. Which means that the S1 sample is the silicon dioxide film of stoichiometric and S2 and S3 samples are silicon dioxide of nonstoichiometric. Then, the three sample sets (S1, S2, and S3) were prepared in order to obtain SiO x film with different stoichiometry. In ALD deposition, the temperature was fixed at 250 • C. The precursor pulse time was changed in order to obtain silicon oxides of stoichiometric (SiO 2 ) and non-stoichiometric (SiO x , x < 2). To be specific, TDMAS pulse time was 2, 2, and 4 s for S1, S2, and S3 sample, respectively, O 2 plasma duration was 8, 4, and 2 s for S1, S2, and S3 film, respectively. The purge time was always kept at 4 s no matter whether it was after the TDMAS pulse or after the O 2 plasma. Meanwhile, in order to decrease the oxygen content in the film to form Si rich silicon dioxide, the O 2 flow rate was also adjusted, it was 50 sccm, 50 sccm, and 25 sccm on the deposition of S1, S2, and S3 film, respectively.
After deposition, all samples were tested by XPS to determine the stoichiometric. Figure 3 displays the XPS wide scan spectrum of S1, S2, and S3 film after Ar + ion treatment. The determined XPS peaks of Si, O, and C are stated in the figure. All the peaks of elements are marked and recognized in detail. By comparing the content of Si and O element in three samples, it is found that the stoichiometries are x = 2, 1.8, and 1.6 for S1, S2, and S3 film, respectively. Which means that the S1 sample is the silicon dioxide film of stoichiometric and S2 and S3 samples are silicon dioxide of nonstoichiometric.

Properties of SiOx Film with Different Stoichiometry
To further analyze the concentration, microstructure, chemical valence, and composition of the film and elements, the Si 2p and O 1s spectra of the SiOx sample were examined by high-resolution XPS, as shown in Figure 4. From the Si 2p spectra of the S1 film ( It is worth pointing out that the intensity and the area of the Si-OHx and Si-Si related sub-peaks increased compared to the strength in S1 sample, which means more content of silicon-related clusters became residual in the film. For the S3 sample, the four sub-peaks are centered at 104.2, 103.3, 101.2, and 100.6 eV. The intensity and area of the Si-OHx related sub-peak further increased, which is induced by the decrease in O2 plasma duration and flow rate. By comparing the O 1s spectra of S1 (Figure 4a2), S2 (Figure 4b2) and S3 sample ( Figure  4c2), it can be found that the O 1s spectra in all three films can be fitted mainly by two sub-peaks located at ~534 eV and ~533.3 eV. For S1, it is 534 eV and 533.3 eV, for S2, it is 534.3 eV and 533.3 eV, then for S3, it is 533.7 eV and 533.2 eV. When we compare these two sub-peaks with the sub-peaks related to Si−O bonds in Si 2p peak (for a better view, they were both marked as the same red and blue color, respectively), we find that they present the same variation tendency in these three samples. Evidently, in Si 2p peaks, the sub-peak related to Si-O bonds with higher binding energy (red line) changed from 104.4 eV to 104.2 eV when reducing the oxygen content in SiOx film. As a matter of fact, the sub-peak related to O-Si bonds with higher binding energy (red line) also shifted from 534 eV to 533.7 eV. Therefore, it is supposed that the two sub-peaks related to O-Si bonds in the O 1s peak (Figure 4a2-c2) correspond with the two sub-peaks related to Si-O bonds of the matching sample found in Si 2p (Figure 4a1-c1). Moreover, the distance of the two sub-peaks related to Si-O bonds decreased from 1.0 eV to 0.9 eV, while, the value of the two sub-peaks related to O-Si bonds decreased from 0.7 to 0.5 eV as well. In addition to these variations, the relative intensity of these two sub-peaks corresponding to high and low binding energy also had the same variation trend both in Si 2p and O 1s. All these results indicate that the chemical binding energy and microstructure are different in these SiOx film with different stoichiometry.

Properties of SiO x Film with Different Stoichiometry
To further analyze the concentration, microstructure, chemical valence, and composition of the film and elements, the Si 2p and O 1s spectra of the SiO x sample were examined by high-resolution XPS, as shown in Figure 4. From the Si 2p spectra of the S1 film (Figure 4a1 , and Si-Si (Si substrate) bonds, respectively. It is worth pointing out that the intensity and the area of the Si-OH x and Si-Si related sub-peaks increased compared to the strength in S1 sample, which means more content of silicon-related clusters became residual in the film. For the S3 sample, the four sub-peaks are centered at 104.2, 103.3, 101.2, and 100.6 eV. The intensity and area of the Si-OH x related sub-peak further increased, which is induced by the decrease in O 2 plasma duration and flow rate. By comparing the O 1s spectra of S1 (Figure 4a2), S2 (Figure 4b2) and S3 sample (Figure 4c2), it can be found that the O 1s spectra in all three films can be fitted mainly by two sub-peaks located at~534 eV and~533.3 eV. For S1, it is 534 eV and 533.3 eV, for S2, it is 534.3 eV and 533.3 eV, then for S3, it is 533.7 eV and 533.2 eV. When we compare these two sub-peaks with the sub-peaks related to Si-O bonds in Si 2p peak (for a better view, they were both marked as the same red and blue color, respectively), we find that they present the same variation tendency in these three samples. Evidently, in Si 2p peaks, the sub-peak related to Si-O bonds with higher binding energy (red line) changed from 104.4 eV to 104.2 eV when reducing the oxygen content in SiO x film. As a matter of fact, the sub-peak related to O-Si bonds with higher binding energy (red line) also shifted from 534 eV to 533.7 eV. Therefore, it is supposed that the two sub-peaks related to O-Si bonds in the O 1s peak (Figure 4a2-c2) correspond with the two sub-peaks related to Si-O bonds of the matching sample found in Si 2p (Figure 4a1-c1). Moreover, the distance of the two sub-peaks related to Si-O bonds decreased from 1.0 eV to 0.9 eV, while, the value of the two sub-peaks related to O-Si bonds decreased from 0.7 to 0.5 eV as well. In addition to these variations, the relative intensity of these two sub-peaks corresponding to high and low binding energy also had the same variation trend both in Si 2p and O 1s. All these results indicate that the chemical binding energy and microstructure are different in these SiO x film with different stoichiometry. For further analysis, the TEM test was used to verify the microstructure like surface roughness and the interfacial composition of these three films. The surface morphology of the three samples can be observed by the low-resolution TEM images shown in Figure 5a1−c1. It can be seen that all films present a smooth surface. This finding agrees with the surface morphology and roughness data shown in Figure 2c, since all films were deposited at 200 °C. The high-resolution TEM images shown in Figure 5a2−c2 exhibit different film structure. As shown in Figure 5a2, the S1 sample is just composed of one layer, which is obviously the SiO2 film. For S2 sample, two sublayers are detectable in Figure 5b2, layer 1 is assigned to the SiO2 owing to the native oxidation on the substrate surface. It also comes from the contribution of residual oxygen in the reactor chamber. It should be noted that residual oxygen is unavoidable since the vacuum is just about 1 mbar. Therefore, at the beginning of the ALD process, the oxygen flow was unstable and not the value that we desired. As a result, SiO2 was generated more or less on the surface of the Si substrate. After few ALD cycles, the residual oxygen in the chamber ran out. Then, silicon oxides with the desired stoichiometry could be obtained when the oxygen flow began to stabilize. Based on this, layer 2 in Figure 5b2 is assigned to the SiO1.8 layer. Moreover, for the S3 sample shown in Figure 5c2, a bilayers structure can also be found. Same as the S2 sample, the layers 1 and 2 are assigned to SiO2 and SiO1.6, respectively. This finding is in agreement with the variation of the Si−O bonds in the Si 2p spectra and O−Si bonds in the O 1s spectra. To be specific, as seen in Figure 4, both of Si−O and O−Si bonds shift to lower binding energy with the Si/O ratio changing from 2 to 1.6. According to reports in previous research, higher valence of Si ion corresponds to higher binding energy [32,33]. On the basis of this finding, there is reason to believe that a certain amount of Si ion turns to lower valence in S2 and S3 sample. This can be verified both by the Si 2p spectra in Figure 4b1-c1 and the TEM structure in Figure 5b2-c2. Meanwhile, because of the existence of more valence states in S2 and S3 sample, the Si 2p and O 1s spectra in these two samples present broaden profile than the value in S1 sample, as seen in Figure 4. Figure 6a shows the measured (scatter line) and simulated (solid line) XRR curve of the SiOx film with different stoichiometry. Figure 6b,c shows the film density and roughness obtained by XRR simulation. The density of SiO2 (S1) sample was ~2.49 g/cm 3 , and it became 2.46 g/cm 3 for the SiO1.8 (S2) film. Then, the value decreased to 2.25 g/cm 3 for SiO1.6 (S2) film. This result indicated that the mass density of silicon dioxide decreases with the decrease of oxygen concentration, in other words, the film mass density becomes lower for Si rich silicon dioxide (SiO1.6) than SiO2. Meanwhile, surface roughness obtained by XRR and AFM is also shown in Figure 6c. It was found that the surface roughness of all the films prepared in this work is very small, which means that the film is very smooth. It should be noted here that both XRR and AFM data present a similar tendency of variation For further analysis, the TEM test was used to verify the microstructure like surface roughness and the interfacial composition of these three films. The surface morphology of the three samples can be observed by the low-resolution TEM images shown in Figure 5a1-c1. It can be seen that all films present a smooth surface. This finding agrees with the surface morphology and roughness data shown in Figure 2c, since all films were deposited at 200 • C. The high-resolution TEM images shown in Figure 5a2-c2 exhibit different film structure. As shown in Figure 5a2, the S1 sample is just composed of one layer, which is obviously the SiO 2 film. For S2 sample, two sublayers are detectable in Figure 5b2, layer 1 is assigned to the SiO 2 owing to the native oxidation on the substrate surface. It also comes from the contribution of residual oxygen in the reactor chamber. It should be noted that residual oxygen is unavoidable since the vacuum is just about 1 mbar. Therefore, at the beginning of the ALD process, the oxygen flow was unstable and not the value that we desired. As a result, SiO 2 was generated more or less on the surface of the Si substrate. After few ALD cycles, the residual oxygen in the chamber ran out. Then, silicon oxides with the desired stoichiometry could be obtained when the oxygen flow began to stabilize. Based on this, layer 2 in Figure 5b2 is assigned to the SiO 1.8 layer. Moreover, for the S3 sample shown in Figure 5c2, a bilayers structure can also be found. Same as the S2 sample, the layers 1 and 2 are assigned to SiO 2 and SiO 1.6 , respectively. This finding is in agreement with the variation of the Si-O bonds in the Si 2p spectra and O-Si bonds in the O 1s spectra. To be specific, as seen in Figure 4, both of Si-O and O-Si bonds shift to lower binding energy with the Si/O ratio changing from 2 to 1.6. According to reports in previous research, higher valence of Si ion corresponds to higher binding energy [32,33]. On the basis of this finding, there is reason to believe that a certain amount of Si ion turns to lower valence in S2 and S3 sample. This can be verified both by the Si 2p spectra in Figure 4b1-c1 and the TEM structure in Figure 5b2-c2. Meanwhile, because of the existence of more valence states in S2 and S3 sample, the Si 2p and O 1s spectra in these two samples present broaden profile than the value in S1 sample, as seen in Figure 4. Figure 6a shows the measured (scatter line) and simulated (solid line) XRR curve of the SiO x film with different stoichiometry. Figure 6b,c shows the film density and roughness obtained by XRR simulation. The density of SiO 2 (S1) sample was~2.49 g/cm 3 , and it became 2.46 g/cm 3 for the SiO 1.8 (S2) film. Then, the value decreased to 2.25 g/cm 3 for SiO 1.6 (S2) film. This result indicated that the mass density of silicon dioxide decreases with the decrease of oxygen concentration, in other words, the film mass density becomes lower for Si rich silicon dioxide (SiO 1.6 ) than SiO 2 . Meanwhile, surface roughness obtained by XRR and AFM is also shown in Figure 6c. It was found that the surface roughness of all the films prepared in this work is very small, which means that the film is very smooth. It should be noted here that both XRR and AFM data present a similar tendency of variation for roughness. Obviously, the RMS result simulated from XRR data is higher than the value obtained by AFM, which is in agreement with the result reported in other work [34]. As seen from RMS values obtained by XRR, the SiO 2 film exhibits the highest RMS roughness value of about 0.64 nm, and the value of the other film is around 0.4-0.5 nm. Nanomaterials 2019, 9, x FOR PEER REVIEW 7 of 13 for roughness. Obviously, the RMS result simulated from XRR data is higher than the value obtained by AFM, which is in agreement with the result reported in other work [34]. As seen from RMS values obtained by XRR, the SiO2 film exhibits the highest RMS roughness value of about 0.64 nm, and the value of the other film is around 0.4-0.5 nm.  In order to achieve in-depth knowledge on the evolution of the optical properties of SiOx film with different stoichiometry, SE measurements have been performed at room temperature immediately after removing the samples from the deposition system to avoid surface contamination. for roughness. Obviously, the RMS result simulated from XRR data is higher than the value obtained by AFM, which is in agreement with the result reported in other work [34]. As seen from RMS values obtained by XRR, the SiO2 film exhibits the highest RMS roughness value of about 0.64 nm, and the value of the other film is around 0.4-0.5 nm.  In order to achieve in-depth knowledge on the evolution of the optical properties of SiOx film with different stoichiometry, SE measurements have been performed at room temperature immediately after removing the samples from the deposition system to avoid surface contamination. In order to achieve in-depth knowledge on the evolution of the optical properties of SiO x film with different stoichiometry, SE measurements have been performed at room temperature immediately after removing the samples from the deposition system to avoid surface contamination. The measured (scatter line) and simulated (solid line) Ψ and ∆ of SiO x thin films for angles of incidence 75 • are shown in Figure 7. From this data, the refractive index of the SiO x can be deduced through a fitting-based analysis. Because of the excellent transparency of the SiO x in the considered wavelength range, it is valid to use a Cauchy relation to model the refractive index of the SiO x : n(λ) = A + B/λ 2 + C/λ 4 . As can be seen in Figure 7, an excellent fit (solid lines) is achieved. Then, the optical constants of SiO x thin film is extracted and obtained from the fitted Ψ and ∆. Figure 8a shows the changes of the refractive index (n) as a function of wavelength in the range of 190-800 nm. Evidently, this fitting procedure yielded a refractive index value (n) of 1.54 at a wavelength of 632.8 nm for SiO 2 , in agreement with the value reported in the literature [35]. Then for SiO 1.8 and SiO 1.6 , it becomes 1.51 and 1.42, respectively. Therefore, the refractive index of the SiO x film can be affected by varying the oxygen concentration. The measured (scatter line) and simulated (solid line) Ψ and Δ of SiOx thin films for angles of incidence 75° are shown in Figure 7. From this data, the refractive index of the SiOx can be deduced through a fitting-based analysis. Because of the excellent transparency of the SiOx in the considered wavelength range, it is valid to use a Cauchy relation to model the refractive index of the SiOx: n λ A B λ ⁄ +C λ ⁄ . As can be seen in Figure 7, an excellent fit (solid lines) is achieved. Then, the optical constants of SiOx thin film is extracted and obtained from the fitted Ψ and Δ. Figure 8a shows the changes of the refractive index (n) as a function of wavelength in the range of 190-800 nm. Evidently, this fitting procedure yielded a refractive index value (n) of 1.54 at a wavelength of 632.8 nm for SiO2, in agreement with the value reported in the literature [35]. Then for SiO1.8 and SiO1.6, it becomes 1.51 and 1.42, respectively. Therefore, the refractive index of the SiOx film can be affected by varying the oxygen concentration.  The refractive index dispersion n of SiOx film was further fitted by the Wemple-DiDomenico model [36] with the purpose of determining the oscillator energy and strength. The refractive index data can be fitted in the spectra below the band-gap to the single oscillator expression [37]: where E represents the photon energy, E0 measures oscillator energy, and Ed represents dispersion energy, which measures the oscillator strength (the strength of inter-band optical transitions). From linear regression of dependence (n 2 − 1) −1 against E 2 (as shown in Figure 8b), the parameters E0 and Ed can be calculated. The values of the fitting constants E0 and Ed are given in Table 1. We can see that for SiO2 sample, E0 is ~12.5 eV, Ed is ~16.5 eV. For SiO1.8 sample, E0 is ~10.5 eV, Ed is ~12.8 eV. Then, for SiO1.6 sample, E0 is ~10.4 eV, Ed is ~9.8 eV. The oscillator energy, E0, is an 'average' energy gap, as The measured (scatter line) and simulated (solid line) Ψ and Δ of SiOx thin films for angles of incidence 75° are shown in Figure 7. From this data, the refractive index of the SiOx can be deduced through a fitting-based analysis. Because of the excellent transparency of the SiOx in the considered wavelength range, it is valid to use a Cauchy relation to model the refractive index of the SiOx: n λ A B λ ⁄ +C λ ⁄ . As can be seen in Figure 7, an excellent fit (solid lines) is achieved. Then, the optical constants of SiOx thin film is extracted and obtained from the fitted Ψ and Δ. Figure 8a shows the changes of the refractive index (n) as a function of wavelength in the range of 190-800 nm. Evidently, this fitting procedure yielded a refractive index value (n) of 1.54 at a wavelength of 632.8 nm for SiO2, in agreement with the value reported in the literature [35]. Then for SiO1.8 and SiO1.6, it becomes 1.51 and 1.42, respectively. Therefore, the refractive index of the SiOx film can be affected by varying the oxygen concentration.  The refractive index dispersion n of SiOx film was further fitted by the Wemple-DiDomenico model [36] with the purpose of determining the oscillator energy and strength. The refractive index data can be fitted in the spectra below the band-gap to the single oscillator expression [37]: where E represents the photon energy, E0 measures oscillator energy, and Ed represents dispersion energy, which measures the oscillator strength (the strength of inter-band optical transitions). From linear regression of dependence (n 2 − 1) −1 against E 2 (as shown in Figure 8b), the parameters E0 and Ed can be calculated. The values of the fitting constants E0 and Ed are given in Table 1. We can see that for SiO2 sample, E0 is ~12.5 eV, Ed is ~16.5 eV. For SiO1.8 sample, E0 is ~10.5 eV, Ed is ~12.8 eV. Then, for SiO1.6 sample, E0 is ~10.4 eV, Ed is ~9.8 eV. The oscillator energy, E0, is an 'average' energy gap, as The refractive index dispersion n of SiO x film was further fitted by the Wemple-DiDomenico model [36] with the purpose of determining the oscillator energy and strength. The refractive index data can be fitted in the spectra below the band-gap to the single oscillator expression [37]: where E represents the photon energy, E 0 measures oscillator energy, and E d represents dispersion energy, which measures the oscillator strength (the strength of inter-band optical transitions). From linear regression of dependence (n 2 − 1) −1 against E 2 (as shown in Figure 8b), the parameters E 0 and E d can be calculated. The values of the fitting constants E 0 and E d are given in Table 1. We can see that for SiO 2 sample, E 0 is~12.5 eV, E d is~16.5 eV. For SiO 1.8 sample, E 0 is~10.5 eV, E d is~12.8 eV. Then, for SiO 1.6 sample, E 0 is~10.4 eV, E d is~9.8 eV. The oscillator energy, E 0 , is an 'average' energy gap, as pointed out in many references [38][39][40]. S.H. Wemple et al. [36] found out that E 0 is related empirically to the lowest direct energy band-gap E g by E 0 ≈ 1.5 E g . The values of E g determined in this way are 8.3, 7.3, and 6.9 eV for SiO 2 , SiO 1.8 , and SiO 1.6 , respectively. Meanwhile, E g of these SiO x film was also determined by XPS O 1s data, which has been reported by many scientists [41][42][43]. As shown in Figure 9, a representative high-resolution scan of the O 1s core level of SiO 2 , SiO 1.8 , and SiO 1.6 film are used to determine the energy band-gap of these three films. We can notice that the E g is 8, 7.5, and 7.3 eV for SiO 2 , SiO 1.8 , and SiO 1.6 film, respectively. As shown in Table 1, the energy band-gap obtained by O 1s peak is very close to the value obtained by the E 0 simulation above. This result confirms that the energy band-gap of SiO x films decrease with decreasing oxygen composition. a Single oscillator and dispersion energies E 0 and E d , respectively, the optical energy band-gap, E g , simulated by E 0 and by O 1s peak, refractive index in the long-wavelength limit n(0), and the refractive index n at 632.8 nm from Figure 8a.
Nanomaterials 2019, 9, x FOR PEER REVIEW 9 of 13 pointed out in many references [38][39][40]. S.H. Wemple et al. [36] found out that E0 is related empirically to the lowest direct energy band-gap Eg by E0 ≈ 1.5 Eg. The values of Eg determined in this way are 8.3, 7.3, and 6.9 eV for SiO2, SiO1.8, and SiO1.6, respectively. Meanwhile, Eg of these SiOx film was also determined by XPS O 1s data, which has been reported by many scientists [41][42][43]. As shown in Figure 9, a representative high-resolution scan of the O 1s core level of SiO2, SiO1.8, and SiO1.6 film are used to determine the energy band-gap of these three films. We can notice that the Eg is 8, 7.5, and 7.3 eV for SiO2, SiO1.8, and SiO1.6 film, respectively. As shown in Table 1, the energy band-gap obtained by O 1s peak is very close to the value obtained by the E0 simulation above. This result confirms that the energy band-gap of SiOx films decrease with decreasing oxygen composition. For the dispersion energy an empirical relation is established [36]: Ed = βNcNeZa, where β is a constant, according to Wemple, β has a value of 0.37~0.04 eV. Nc is the coordination number of the nearest neighboring cation to the anion, and Za is the formal chemical valency of the anion, then Ne is the total number of valence electrons per anion [39]. For SiO2, Ne = [(4 valence electrons)·(1 silicon cation) + (6 valence electrons)·(2 oxygen anions)]/2 = 8, and Za = 2. For SiO, Ne = [(4 valence electrons)·(1 silicon cation) + (6 valence electrons)·(1 oxygen anions)]/1 = 10, and Za = 2. Thus, Nc can be calculated from the relationship: Nc = Ed/βNe Za. It is noted that the values of Ed and Nc decreases with the decrease of oxygen concentration in SiOx film. So, we can conclude that SiOx film has a more amorphous structure than SiO2 film. This result correlates with the fact that the SiOx film is less dense and has a lower refractive index [44]. In fact, Figures 6b and 8a demonstrate that the variation of film density and refractive index agrees with this result perfectly. Furthermore, the decrease of Ed from SiO2 to SiOx sample is also supposed to relate to the effective number of valence electrons, the variation of Si-O and Si-Si bonds, as shown in Figure 4. a Single oscillator and dispersion energies E0 and Ed, respectively, the optical energy band-gap, Eg, simulated by E0 and by O 1s peak, refractive index in the long-wavelength limit n(0), and the refractive index n at 632.8 nm from Figure 8a. For the dispersion energy an empirical relation is established [36]: E d = βN c N e Z a , where β is a constant, according to Wemple, β has a value of 0.37~0.04 eV. N c is the coordination number of the nearest neighboring cation to the anion, and Z a is the formal chemical valency of the anion, then N e is the total number of valence electrons per anion [39]. For SiO 2 , N e = [(4 valence electrons)·(1 silicon cation) + (6 valence electrons)·(2 oxygen anions)]/2 = 8, and Z a = 2. For SiO, N e = [(4 valence electrons)·(1 silicon cation) + (6 valence electrons)·(1 oxygen anions)]/1 = 10, and Z a = 2. Thus, N c can be calculated from the relationship: N c = E d /βN e Z a . It is noted that the values of E d and N c decreases with the decrease of oxygen concentration in SiO x film. So, we can conclude that SiO x film has a more amorphous structure than SiO 2 film. This result correlates with the fact that the SiO x film is less dense and has a lower refractive index [44]. In fact, Figures 6b and 8a demonstrate that the variation of film density and refractive index agrees with this result perfectly. Furthermore, the decrease of E d from SiO 2 to SiO x sample is also supposed to relate to the effective number of valence electrons, the variation of Si-O and Si-Si bonds, as shown in Figure 4.
Moreover, the long-wavelength limit of the refractive index, n(0), is given by [44]: n 2 (0) = 1+ E d /E 0 . Then n(0) is estimated to be 1.52, 1.49, and 1.39 for SiO 2 , SiO 1.8 , and SiO 1.6 , respectively, as seen in Table 1. For comparison, the refractive index at 632.8 nm obtained by SE analysis are also listed in Table 1. It is easy to find that the value of n(0) is very close to the refractive index at 632.8 nm, except for a little smaller. The excellent agreement of these results improved the correctness of the model.

Photoluminescence Properties of SiO 1.6 /SiO 2 Super-Lattice
To exam the luminescence property of the Si rich SiO x film prepared in this work, SiO x /SiO 2 super-lattice with x = 1.6 was fabricated. The super-lattice consisted of 22 SiO 1.6 /SiO 2 bilayers, and the nominal thickness of SiO 1.6 layer was~1.5 nm, while the SiO 2 layer was fixed at~3 nm, so the total film thickness was about 100 nm. In order to form the Si-NCs in the SiO 2 matrix, thermal annealing was performed in a rapid thermal annealing furnace under nitrogen atmosphere with ramp rates of about 25 • C/s for phase separation and crystallization. In this work, the sample was annealed at 1000 and 1100 • C for 10 s. Figure 10 compares the room-temperature PL spectra of the SiO 1.6 /SiO 2 super-lattice before and after RTP annealing with different temperature. All of the spectra show the similar luminescence profiles, and the PL maximum position does not show any notable features with annealing temperatures, except for an increase in the PL intensity. In detail, three peaks all have a broader profile ranging from 550 nm to 1000 nm, and two small peaks presents around 750 nm and 920 nm. It seems that the spectra is superimposed by some small peaks, for example, three sub-peaks with gaussian shape located at 600, 750, and 920 nm could form a spectra like this. This result presents a familiar PL spectra for Si-NCs in SiO x /SiO 2 super-lattice deposited by other techniques in previous reports [11,[45][46][47][48]. The weak and inapparent luminescence properties of the super-lattice implied that there was not enough Si-NCs formed in SiO 2 matrix. On one hand, it is supposed that the annealing temperature and time are not sufficient to form Si-NCs compared with the conditions in the previous report [49]. On the other hand, the SiO 1.6 layer was not Si-rich enough to form considerable Si-NCs when high temperature annealing. Specifically, the high temperature annealing of such initially amorphous SiO x films results in phase separation described by [9], in SiO x clusters. The phase separation of the SiO x automatically ensures that the nucleated Si nanocrystals/nanoclusters are separated from each other by a SiO 2 shell. Based on the theory above, it is just 20% of SiO x cluster turned to Si nanocrystals when x = 1.6. Also, as reported in previous articles [50], an increase in the stoichiometry parameter x to 1.63 in SiO x /SiO 2 matrix resulted in a drastically reduced density of nanocrystals correlated with a strong decrease in PL intensity. Based on Equation (2), the thickness of the oxide between the Si nanoclusters depends on the stoichiometry of the SiO x as well. The crystallization of SiO x film resulted in randomly distributed nanocrystals with an average size that depends on the original stoichiometry. So both the thickness of SiO x and SiO 2 layer also play a role in this property, which should be systematically studying in further research.
Nanomaterials 2019, 9, x FOR PEER REVIEW 10 of 13 To exam the luminescence property of the Si rich SiOx film prepared in this work, SiOx/SiO2 super-lattice with x = 1.6 was fabricated. The super-lattice consisted of 22 SiO1.6/SiO2 bilayers, and the nominal thickness of SiO1.6 layer was ~1.5 nm, while the SiO2 layer was fixed at ~3 nm, so the total film thickness was about 100 nm. In order to form the Si-NCs in the SiO2 matrix, thermal annealing was performed in a rapid thermal annealing furnace under nitrogen atmosphere with ramp rates of about 25 °C/s for phase separation and crystallization. In this work, the sample was annealed at 1000 and 1100 °C for 10 s. Figure 10 compares the room-temperature PL spectra of the SiO1.6/SiO2 super-lattice before and after RTP annealing with different temperature. All of the spectra show the similar luminescence profiles, and the PL maximum position does not show any notable features with annealing temperatures, except for an increase in the PL intensity. In detail, three peaks all have a broader profile ranging from 550 nm to 1000 nm, and two small peaks presents around 750 nm and 920 nm. It seems that the spectra is superimposed by some small peaks, for example, three sub-peaks with gaussian shape located at 600, 750, and 920 nm could form a spectra like this. This result presents a familiar PL spectra for Si-NCs in SiOx/SiO2 super-lattice deposited by other techniques in previous reports [11,[45][46][47][48]. The weak and inapparent luminescence properties of the super-lattice implied that there was not enough Si-NCs formed in SiO2 matrix. On one hand, it is supposed that the annealing temperature and time are not sufficient to form Si-NCs compared with the conditions in the previous report [49]. On the other hand, the SiO1.6 layer was not Si-rich enough to form considerable Si-NCs when high temperature annealing. Specifically, the high temperature annealing of such initially amorphous SiOx films results in phase separation described by [9], in SiOx clusters. The phase separation of the SiOx automatically ensures that the nucleated Si nanocrystals/nanoclusters are separated from each other by a SiO2 shell. Based on the theory above, it is just 20% of SiOx cluster turned to Si nanocrystals when x = 1.6. Also, as reported in previous articles [50], an increase in the stoichiometry parameter x to 1.63 in SiOx/SiO2 matrix resulted in a drastically reduced density of nanocrystals correlated with a strong decrease in PL intensity. Based on Equation (2), the thickness of the oxide between the Si nanoclusters depends on the stoichiometry of the SiOx as well. The crystallization of SiOx film resulted in randomly distributed nanocrystals with an average size that depends on the original stoichiometry. So both the thickness of SiOx and SiO2 layer also play a role in this property, which should be systematically studying in further research. Figure 10. Photoluminescence spectra of the as-deposited and annealed SiO1.6/SiO2 superlattice.

Conclusions
In this study, the SiO x thin film with different stoichiometry was successfully deposited with TDMAS and O 2 plasma by using PEALD. In order to realize this target, the deposition parameter was firstly studied, and the effect of deposition temperature, plasma pulse time, and gas flow on growth rate was obtained. Then, SiO x thin films with different stoichiometry were fabricated by optimizing the precursor pulse time and O 2 gas flow. Meanwhile, the micro-structure, chemical, and optical properties were systematically investigated by AFM, TEM, XRR, SE, and XPS measurements. It was found that the mass density, refractive index, and optical energy band-gap of SiO x film are all lower than SiO 2 film. Furthermore, SiO 1.6 /SiO 2 superlattice with 150 nm thickness was fabricated based on the deposition parameters obtained above. The photoluminescence test shows that the SiO 1.6 /SiO 2 super-lattice presents a weak and indistinctive peak before and after high temperature annealing, implying that the SiO 1.6 film is still not Si rich enough. Further research is worth exploring to obtain the SiO x film with a smaller x value by optimizing PEALD parameters.