1. Introduction

LEDs based on III-Nitride system and InGaN system are regarded as the most important light sources in next-generation solid-state lighting owing to their high quantum efficiency and multiple applications [1–6]. However, these state-of-the-art LED technologies are not fully compatible with current complementary metal-oxide semiconductor (CMOS) techniques. In the past decade, highly luminescent silicon-based materials have attracted a great deal of interest owing to their potential application in Si-based monolithic optoelectronic integrated circuits [7–17]. To engineer Si into a more efficient light-emitting material, different approaches such as low-dimensional Si systems and Si modification have been developed [7–15]. Among them, enormous efforts have been devoted to the silicon oxide system in which efficient photoluminescence as well as optical gain has been achieved [8, 16, 17]. However, it is found that the silicon oxide system is not suitable for the fabrication of stable and efficient electroluminescent devices due to the wide bandgap of SiO₂ (9 eV) that causes the difficulty in efficient electrical injection. Silicon nitride with respect to silicon dioxide features narrower tunable band gaps in the range of 2.0–5.0 eV, rendering it useful for the design of efficient light emitting devices. So far, an effective EL at a low driving voltage has been achieved in the SiN-based light-emitting devices (LEDs) [18–25]. However, light emission efficiency of SiNx-based LEDs is still so low that it can't meet the demands of silicon-based light sources. The insufficient EL efficiency is due to the unbalanced carrier injection and strong nonradiative recombination in silicon nitride system. In recent years, much attention has been paid on the silicon oxynitride system [26–30]. It has been reported that silicon oxynitride exhibits strong green light emission as a result of the formation of Si-O localize states [27, 31]. Moreover, silicon oxynitride in comparison to silicon nitride and silicon oxide can more effectively improve the equivalent carrier injections in LEDs and thus significantly increase the carrier recombination probability [30]. In addition, silicon oxynitride also has a narrower tunable band gaps as compared to silicon oxide, making it favorable for carrier injections in LEDs at a low driving voltage [29, 30]. Although recent studies have explored efficient light emission from the silicon oxynitride system, the progress is still slow. In particular, a well-control full-color emitter based on silicon oxynitride is still far from being established. The partial reason for this is the lack of information available to correlate PL characteristics to the influential factors such as oxygen content.

In this paper, we will show that silicon oxynitride films can be finely tuned to emit light in different colors by means of modulating the oxygen content. Photoluminescence (PL) measurements combined with XPS analysis reveal that the increase of the oxygen content from 8% to 61% results in the red, orange-yellow and white switching photoluminescence. Furthermore, it is interesting to find that these very bright red, orange-yellow and white light emissions can be clearly observed with the naked eye in a bright room. The intense tunable light emission is discussed and suggested from the chemical bonds reconstruction as a result of the increasing oxygen content in the films.

2. Experimental

The amorphous silicon oxynitride films were prepared on Si (100) wafers and quartz in very high frequency plasma enhanced chemical vapor deposition (VHF-PECVD) system using the gas mixture of SiH₄, NH₃, and O₂ as the precursor. The power density of 0.6 W/cm² was used in the experiments. The depositions were carried out at a radio frequency power density of 0.6 W/cm², a low temperature of 250 °C and a pressure of 20 Pa. The flow rates of SiH₄ and NH₃ were kept at 5 and 3 sccm, respectively, while the flow rates of O₂ were varied from 0.6 to 3 sccm. Samples were named by Sx (x = 1, 2, 3, 4) for the O₂ flow rates at 0.6, 1, 2, and 3 sccm, respectively. PL measurements were performed on a Jobin Yvon fluorolog-3 spectrophotometer with PMT detector at room temperature. Photoluminescence time decay measurements were performed on the Edinburgh FLS 920 spectrometer equipped with a laser diode head as excitation source (20 MHz repetition rate, λ_exc = 401 nm, 5 ps time resolution upon deconvolution). Optical absorption spectra were conducted by Shimadzu UV-3600 spectrophotometer on quartz samples and used to estimate the optical bandgaps of samples. A Fourier transform infrared (FTIR) spectroscope was employed to record the bonding configurations of the samples. And x-ray photoelectron spectroscopy (XPS) was used to investigate the Si, N, and O contents in the samples.

3. Results and discussion

Figure 1(a) presents PL spectra of silicon oxynitride samples deposited at different oxygen flow rates. It can be seen that the samples excited by the 325 nm line from Xe lamp show tunable light emission in the visible range when varying the oxygen flow rate from 0.6 sccm to 3 sccm. For the sample S1, the dominant PL is peaked at 1.65 eV (~750 nm). With the increase of oxygen flow rate, the dominant PL peak is gradually shifted toward the high energy side, and it is observed at 1.82 eV (~680 nm) in sample S2 and at 1.97 eV (~630 nm) in sample S3. With increasing oxygen flow rate up to 3.0 sccm, the PL band from sample S4 splits into two peaks. One is related to the blue band at 2.82 eV (~440 nm), the other corresponds to the green band at 2.29 eV (~540 nm). The PL intensity tends to increase with increasing the oxygen flow rate. Among the oxygen flow rates investigated, 2 sccm results in the highest PL intensity, as is demonstrated in Fig. 1(b). It is interesting to find that intense red, orange, yellow and white switching light emissions from the samples are perceptible to be observed with the naked eye in a bright room when excited with a wavelength of 325 nm, as is shown in the inset of Fig. 1(a). Our experimental results indicate that oxygen plays an importance role in the light emission of silicon oxynitride.

 

Fig. 1 (a) PL spectra of silicon oxynitride samples fabricated at different oxygen flow rates: 0.6 sccm (S1), 1.0 sccm (S2), 2.0 sccm (S3), and 3.0 sccm (S4). Inset shows the photography of the samples under an excitation wavelength of 325 nm. (b) The integrated PL intensity of the samples as a function of oxygen flow rate.

Download Full Size | PDF

Figure 2 shows optical bandgap energy (E_opt) of the samples as a function of oxygen flow rate. The E_opt is calculated according to the Tauc equation (αhν)^1/2 = Β (hν−Ε_opt), where α is the absorption coefficient and B is a constant [32]. One can see that the E_opt increases from 2.28 to 3.81 eV with the increase of the oxygen flow rate from 0.6 to 3 sccm. The increase in the E_opt indicates an evolution of the band structure in the samples, which can be attributed to the increasing oxygen concentration in the samples as revealed in the following Fig. 4(b). From Fig. 2, it is also found that the PL peak energy of the samples is smaller than the corresponding E_opt. This indicates that the tunable PL in our case is not from the band-to-band recombination. To clarify the PL characteristics, the effect of excitation wavelengths on PL of the samples were investigated as shown in Fig. 3.One can see that the peak positions for all PL bands show a negligible shift as the excitation wavelengths vary from 300 nm to 400 nm. This behavior is identical to that of defect-related PL which peak position is independent of the excitation wavelengths because of the narrow distribution of defect-related localized states [27]. Therefore, in our case the PL bands may come from the defect-related luminescence centers.

 

Fig. 2 Optical bandgap and PL peak energy of silicon oxynitride samples as a function of oxygen flow rate. Inset shows the tauc plots of S1.

Download Full Size | PDF

 

Fig. 3 PL peak energy versus the excitation wavelength for S1, S2, S3, and S4.

Download Full Size | PDF

To analyze the origin of PL characteristics, FTIR was employed to examine the local bonding configurations of the samples. Figure 4(a) displays the FTIR absorption spectra of all samples. These spectra mainly show the following vibrational bands [28, 33]: The predominant band between 830 cm⁻¹ and 1060 cm⁻¹ is mainly caused by the vibration of Si-N, O-Si-N and Si-O groups. The ~2140 cm⁻¹ band is connected with the Si-H stretching vibration. The ~3350 cm⁻¹ band is associated to N-H stretching vibration. A remarkable feature for the FTIR spectra in Fig. 4(a) is that the predominant bands strongly depend on the oxygen flow rates. The predominant band in S1 appears at ~836 cm⁻¹, corresponding to the Si-N stretching vibration. With increasing the oxygen flow rate, this band becomes broaden and shifts to ~855 cm⁻¹ with a shoulder at ~920 cm⁻¹ in S3. According to earlier FTIR analysis on a-SiN_xO_y films reported by Dong [33], the ~920 cm⁻¹ band is from the vibration of O-Si-N bond. With increasing oxygen flow rate up to 3.0 sccm, the predominant band moves to ~1060 cm⁻¹, which is assigned to Si-O-Si stretching vibration. Apparently, the increase of the oxygen flow rate results in chemical bonds reconstruction in the samples. Combining with the PL results mentioned above, the chemical bonds reconstruction induced by oxygen seems to be responsible for the strong tunable light emission.

 

Fig. 4 (a) FTIR spectra of silicon oxynitride samples fabricated at different oxygen flow rates, respectively. (b) Si, N and O concentration as a function of the oxygen flow rate. (c) Deconvoluted Si 2p XPS spectra for S1, S2, S3, and S4, respectively.

Download Full Size | PDF

The chemistry of the samples was further characterized by XPS. The Si, N and O content in the samples were estimated from XPS spectra as indicated in Fig. 4(b). At the oxygen flow rate of 0.6 sccm, the contents of Si, N and O in S1 are 60%, 31% and 8%, respectively. This means that the dominant phase in S1 is silicon nitride. With increasing the oxygen flow rate up to 3 sccm, the O content rapidly increases to 61%, while the contents of Si and N simultaneously decrease down to 29% and 8%, respectively. This indicates the dominant phase in S4 changes into silicon oxide. To gain more insight on the phase transition in the samples, Si 2p core level spectra, which are an indication for the coexistence of different ionic states of Si atoms [34], were analyzed as shown in Fig. 4(c). One can see that the binding energy spectra are typically composed of various Si phases, those are untreated Si, SiN_x, SiO_xN_y and SiO_x. At low oxygen flow rates (0.6 and 1.0 sccm), the samples are dominated by SiN_x phase. It means that the oxygen may be doped as defect states located in the SiN_x energy band. In previous work, it has been reported that the introduction of oxygen can create the localized states related to the N-Si-O bonds in the SiN_x and give rise to strong light emission [27, 33]. Moreover, the location of the oxygen-induced luminescent defect states in the band gap is found to be at around 0.63 eV from the absorption edge. In our case, we note that the values between E_opt and E_PL in S1 and S2 are about 0.63 eV and 0.61 eV,respectively, which is in good agreement with the value reported by Dong [33]. This strongly indicates that the PL emissions in S1 and S2 are originated from the radiative recombination in the localized states related to the N-Si-O bonds. From Fig. 4(c), it is also found that there is a peak located at around 99.50 eV for S1 and S2, which is related to the elemental Si components, indicating a formation of Si clusters in S1 and S2. With the increase of oxygen flow rate from 1.0 to 2.0 sccm, one can see that the dominant phase in the S3 changes into silicon oxynitride. Therefore, according to the Wong’ work, the 1.96 eV PL band in S3 is caused by the ≡Si₂N· defects in silicon oxynitride [35]. It is worth noting that this kind of defects strongly relies on the oxygen content in the film and can be removed by increasing the oxygen content to some extent [35]. This well accounts for the fact that the 1.96 eV PL band observed in S3 disappears after increasing the oxygen flow rate to 3.0 sccm. It is found that in this case high oxygen content (61%) leads to the dominant phase of S4 transforming into silicon oxide, as is indicated in Fig. 4(c). Compared with the PL spectra in Fig. 1(a), one can see that the PL band in S4 synchronously shifts to high energy side and splits into two peaks at 2.29 eV and 2.82 eV, respectively. In silicon oxide, 2.29 eV PL band is unambiguously attributed to the radiative recombination through defects such as the nonbridging oxygen hole centers [36]. On the other hand, we notice that the 2.82 eV PL band is consistent with the previous observation reported by Noma [37] where 2.6-2.9 eV PL band is ascribed to the Si-N defect states located at the top of the valence band in silicon oxide structure. The vibration band at ~860 cm⁻¹ shown in Fig. 4(a) further presents a direct evident of the existence of Si-N in S4. Therefore, the 2.82 eV PL in our case is suggested from the recombination of holes and electrons in localized states corresponding to Si–N bonds. From the above results, it is obvious that oxygen plays a decisive role in the chemical bonds reconstruction, especially the transformation of the kinds of defect luminescence centers as well as the dominant phase structures in silicon oxynitride films.

In order to further clarify the light emission mechanism of all samples, we have carried out luminescence decay measurements. Figure 5 illustrates normalized RT luminescence decay traces taken from different samples. The decay process can be well fitted with a double exponential function$A_0+A_1*\exp (-t/{\tau }_1)+A_2*\exp (-t/{\tau }_2)$, in which τ_i and A_i (i = 1, 2) represent the lifetime and amplitude of each exponential decay component, respectively, and A₀ is the background level [38]. The intensity-weighted averaged PL lifetimes are then determined by $(A_1\ast {\tau }_1^2+A_2\ast {\tau }_2^2)/(A_1\ast {\tau }_1^{}+A_2\ast {\tau }_2^{})$ [38]. It can be seen that all samples show a fast decay dynamic with the lifetimes between 2.0 to 3.5 nanoseconds as revealed in Fig. 5. Such a luminescent dynamic behavior is consistent with that observed in defect-related luminescent Si-based materials such as SiN_x and SiO_x [38, 39]. We also notice that the luminescent lifetimes in our case are shorter than that reported by Kato [40] where the PL is attributed to optical transitions among the band tail states of silicon oxynitride. These support our argument that the photoluminescence of our samples is from defect states instead of band tail states. It’s worth noting that the lifetimes in different samples show a similar value, almost independent of the oxygen content in the samples. This is completely different from that reported in Si quantum dots (Si-QDs) system where the luminescence lifetime strongly relies on the Si-QD size owing to the quantum confinement effect [41]. Although the luminescence lifetimes in our case are independent of the oxygen content in the samples, the PL intensity of the samples is found to increase with increasing the oxygen flow rate up to 2 sccm, as is shown in Fig. 1(b). For the PL intensity (I_PL), it can be expressed as follows: $I_{PL}={\tau }_R^{-1}/({\tau }_R^{-1}+{\tau }_{NR}^{-1})$, where τ_R and τ_NR are the lifetime of the radiative and nonradiative recombinations, respectively [42]. According to the equation, the value of τ_NR increases with the increase of I_PL due to the similar value of τ_R in different samples. Therefore, the fact that the I_PL increases with the oxygen flow rate can be attributed to the decrease of the nonradiative recombination rate in the samples. This is because the Si–O bonds can effectively relieve the internal stress in silicon oxynitride due to the oxygen taking twofold coordination [40], thus making the structural disorder in silicon oxynitride become smaller.

 

Fig. 5 Room-temperature luminescence decay traces taken from different samples, respectively.

Download Full Size | PDF

4. Conclusion

In summary, strong tunable light emissions from red to white have been achieved in silicon oxynitride films by modulating the oxygen content. The transformation of the kinds of defect luminescence centers as well as the dominant phase structures in silicon oxynitride films controlled by the oxygen content are responsible for the tunable light emissions, while the strong PL intensity is suggested from the nanoseconds recombination lifetime and the decrease of nonradiative recombination rate as a consequence of internal stress reduction in silicon oxynitride. The strong tunable light emission and the fast decay dynamics open the possibility of the applications of the luminescent silicon oxynitride in photonics as well as optoelectronics integration.