1. Introduction

For achieving silicon (Si) based monolithic optoelectronic integration, the most crucial issue urgent to be solved is the high efficient Si-based light source. Unfortunately, Si is a typical indirect band gap material, which limits its efficiency as a light emitter. Thus, Si-based luminescent materials were actively investigated with the main purpose of improving the performance of the PL quantum yields (PL QYs), the electroluminescence (EL) devices, and the related radiative recombination mechanisms in more than past two decades [1–19].

Generally, Si-based luminescent materials exhibit two typical types, Si nano-structured materials, and Si-based compounds. As summarized by Canham [1], the improved PL QY from Si nano-structured materials, such as porous Si [2], colloidal Si quantum dots (Si QDs) [3–5], nanocrystal-Si (NC-Si) embedded thin films [6–9], has been reported previously. By passivating the surface of Si QDs with organic molecules, the PL quantum yields of Si QDs as high as 43% [3] and 60% [4] were obtained. Recently, Shen et al. reported that the measured PL QY of colloidal Si NCs has been enhanced significantly from 23.6% to 55.8% by prolonging the laser ablation time from 30 to 120 min [5]. Negro et al. and Walters et al. reported the PL QY of 7% and PL internal quantum efficiency of 59% ± 9% from NC-Si embedded silicon nitride (SiN_x) and silicon dioxide (SiO_x) thin films, respectively [7,8]. For EL properties from Si nano-structured materials, Cheng et al. realized an optimized NC Si-organic light-emitting device [10]. Liu et al. obtained the EL QYs of 6.2% from colloidal Octyl-Si QDs LEDs [11]. However, the exciting results from Si-based compounds is still lacking in the open literature [12–15]. In our previous work, we reported the oxygen impurity doped into silicon nitride networks induces the significant enhancement of PL intensity [16]. Then we identified the new luminescent N-Si-O bonding related defect states in the band gap [17,18]. Furthermore, we realized the highly efficient visible LEDs based on phosphorus doped n-a-SiN_xO_y/p-Si heterojunction structures, and found that the EL peak energy is coincided with that of PL of a-SiN_xO_y, suggesting the consistent radiative recombination origin in a-SiN_xO_y systems [19].

In this study, we systematically investigated the absolute PL quantum yields and the related radiative recombination mechanisms for tunable luminescent a-SiN_xO_y films. The PL QY at PL peak energy (E_PL) range of 2.12-2.91 eV was directly measured, and the value of 8.38% has been achieved at E_PL = 2.55 eV. By performing the FTIR, XPS and EPR measurements, the existence of N-Si-O bonding states has been confirmed. The PL QY values for varying stoichiometric a-SiN_xO_y films are proportional to the concentration of N-Si-O bonding related N_x defects, which indicated the dominant contribution of luminescent N-Si-O bonding states to PL QY in radiative recombination processes. We also precisely performed the temperature dependent time resolved PL (TD TR-PL) to monitor the ns-PL lifetimes evolution profile as a function of detected emission wavelengths, thus further verifying the luminescent N-Si-O defect states are responsible for the high efficient PL.

2. Experimental

With adequate adjustment of the gas flow ratio R (R = NH₃/SiH₄), the varying stoichiometric a-SiN_xO_y films were deposited on p-Si and quartz substrates by the PECVD (OXFORD Plasmalab 80Plus), using source gases include 5% SiH₄ in N₂ plus NH₃ and N₂, and then subsequently oxidized in situ by oxygen plasma treatment. The detailed flow parameters of source gases were described in Table 1. The radio frequency power, reaction pressure, and substrate temperature were kept constant at 20 W, 620 mTorr, and 250 °C. The Tauc’s optical band gap (E_opt) was determined from the transmittance measurements. The steady-state PL spectroscopy were recorded by a Fluorolo-3 system in computer-controlled Delta 9023 oven under 8 K, using a He–Cd laser as excitation source. The absolute PL QY measurements were performed through a calibrated integrating sphere method under the excitation wavelength of 375 nm at room temperature [16]. The chemical compositions and bonding configurations were checked by performing the XPS and FTIR measurements. The EPR measurements (Bruker EMXplus X-band spectrometer) were further used to identify the atomic scale defects generated in the band gap. A FLS980 (Edinburgh Instrument) equipped with a EPL375 pulse diode laser (pulse width ~53 ps, repetition rate = 20 MHz, λ_exc = 375 nm, pumping fluence W_PF = 5 mJ/cm²), and a time-correlated single photo counting (TCSPC) system (time resolution ~100ps) were used to record the characteristics of the temperature dependent time resolved photoluminescence (TD-TRPL) of a-SiN_xO_y films with different R.

Table 1. Summary of the fabricating parameters, E_opt, E_PL and the PL QY.

View Table

3. Results and discussion

As shown in Fig. 1(a), with the increased R range from 0.3 up to 5, the E_PL exhibits a blue shift from 2.12 eV to 2.91 eV, which were listed in Table 1, and the bright purple-orange light emission was perceptible by the naked eye [Fig. 1(a) insert]. Then we directly measured the absolute PL QYs in a calibrated integration sphere. Figure 1(b) represents the PL QY measurement processes of a-SiN_xO_y samples with R = 1.5. We directly measured the total incident excitation photons ${\varphi }_0$ (black line); the unabsorbed excitation photons $\varphi {}_1$ and corresponding emitted photons ${\phi }_1$ when the excitation source is direct onto samples (blue line); the unabsorbed excitation photons ${\varphi }_2$ and corresponding emitted photons ${\phi }_2$ when the excitation source is indirect onto samples (red line). The PL QY is given by

 

Fig. 1 (a) Normalized PL spectra of a-SiN_xO_y films for various R. The insets are luminescent photographs of the related samples. (b) The PL QY measurement processes of a-SiN_xO_y samples for R = 1.5.

Download Full Size | PDF

Where ${\phi }_{ext}$ and ${\varphi }_{abs}$ are the numbers of the emitted photons emitted out of the samples and the numbers of absorbed photons at an excitation wavelength, respectively. A is the film absorbance when the excitation light is direct onto the samples. By using Eq. (1), for R = 1.5, the PL QY of 8.38% has been achieved at E_PL = 2.55 eV, which is higher than those of reported NC-Si embedded SiN_x films [7]. For the PL peak range of 2.12-2.91 eV, the PL QY values exceeding 1.5% has been achieved in a-SiN_xO_y samples, which were listed in Table 1 and plotted as a function of R in Fig. 3(b).

The presence of Si, N, and O was measured from the binding energies of Si 2p, N 1s, and O 1s peaks in the obtained XPS spectra. After the top layer (~60 nm) was removed away by Ar ion beam, the relative concentrations of O slightly change and have an average value about 3.96% for various R, and then tend to stabilize with further increasing sputtering times, which is a direct evidence of the incorporated O into a-SiN_x networks. The binding energies of Si 2p peaks are located between 101.9 eV of Si₃N₄ and 103.4 eV of SiO₂, which indicates the existence of the N-Si-O bonding configurations in a-SiN_xO_y films [20,21]. To identify the N-Si-O bonding configurations, we employed the FTIR spectra of a-SiN_xO_y films (R = 0.8) and the controlled a-SiN_x films deposited on double polished Si substrates. As shown in Fig. 2(a), the distorted Si-Si mode (460 cm⁻¹), Si-N stretching mode (900 cm⁻¹), N–H rocking mode (1170 cm⁻¹), Si–H stretching mode (2150 cm⁻¹), and N–H stretching mode (3350 cm⁻¹) vibration peaks are distinctly clear [22]. However, the Si-O stretching mode (1070 cm⁻¹) cannot be seen in both of the two FTIR spectra, suggesting the incorporated oxygen atoms just like doped impurity. From the enlarged Si-N stretching mode vibration peaks, we found that the Si-N stretching peaks of a-SiN_xO_y films slightly shift to high wavenumber and exhibit a shoulder peak, both of which should be origin from the incorporation of O into a-SiN_x and form the N-Si-O bonding configurations, as shown in Fig. 2(b).

 

Fig. 2 (a) The FTIR spectra of both a-SiN_xO_y films (R = 0.8) and the controlled a-SiN_x films. (b) Enlarged Si-N and Si-O stretching mode in both of the two FTIR spectra.

Download Full Size | PDF

To gain more insight in the relationship between the high PL QY and N-Si-O bonding states, we further employed the EPR measurements. The g values for our a-SiN_xO_y films with various R have a range about 2.0025~2.0039, and the total spin densities have a range about 3.4 × 10¹⁷ cm⁻³~8.4 × 10¹⁷ cm⁻³. For the amorphous materials, the structure disorders, defects, and composition etc. will influence the g values. Figure 3(a) represents the measured first derivative EPR absorption spectrum of the a-SiN_xO_y films (R = 0.8) and the controlled a-SiN_x films. The measured zero-crossing g value and Gauss line width (∆H_pp) were listed in Fig. 3(a) insert. We found that the g values in a-SiN_xO_y samples (g = 2.0034) were larger than those in the controlled a-SiN_x films (g = 2.0027), which is as similar as reported in [23]. We precisely considered the possible existence of traditional typical defects in a-SiN_xO_y systems, such as the K⁰ center (·Si≡N₃, g = 2.0028), the $N_2^0$center (Si₂ = N·, g = 2.0038), and the $N_4^{+}$center (Si₄≡N·, g = 2.004~2.0045) are common in a-SiN_x systems, as well as the E_s (·Si≡O₃, g = 2.0008,) and the ${E^{\prime }}_r$ (O₃≡Si···Si≡O₃) are common in a-SiO_x systems, etc. Our g values lie between that of the E_s defect centers (*Si≡SiO₂, g = 2.0014), and the K defect centers (*Si≡Si₂N, g = 2.0047), and are close to the g values of *Si≡SiN₂ (g = 2.0036) [24]. However, the PL intensity of above mentioned defect states are weak [12], and the PL peak energies are not in the range of our observed PL band [13]. Meanwhile, we noticed that O is affecting the g values of the nitroxide like N center (N_x) in a-SiN_x networks, which is the herald of a true a-SiN_xO_y phase [25,26]. Therefore, we decomposed the measured g values into the trivalent Si dangling bonds (DBs) and N-Si-O related N_x defects by using Lorentzian fitting, which were shown in Fig. 3(b) insert (R = 1.5) [17]. The atom scale defects densities in a-SiN_xO_y systems can be obtained by using double integration of measured and fitting EPR signals, thus obtaining the coexisted Si DBs and N_x defects densities, which were plotted as a function of R in Fig. 3(b). We found that the variation tendency of the PL QY values for different R was following with the variation tendency of the relative luminescent N-Si-O bonding states densities, and opposite to that of trivalent Si DBs defects. It can help us to understand the dominant contribution of luminescent N-Si-O bonding states to the high PL quantum efficiency in the ultrafast radiative recombination processes of the a-SiN_xO_y samples, even if the density of Si DBs defect states is higher than that of N_x defect states [12].

 

Fig. 3 (a) The measured first derivative EPR spectra of a-SiN_xO_y films (R = 0.8) and the controlled a-SiN_x films. The inset shows the relative measured parameters. (b) The PL QY and the spin densities of total defects, Si DBs, and N_x defects vs. R. The inset shows the measured EPR and the deconvolved signals of a-SiN_xO_y films (R = 1.5).

Download Full Size | PDF

At last, the temperature dependent ns-PL decay properties of a-SiN_xO_y films have been precisely monitored to analyze the ultrafast recombination processes from luminescent N-Si-O bonding states. Figure 4(a) represents the ns-PL spectra of a-SiN_xO_y samples with R = 1.5 at 300 K. The ns-PL lifetimes were obtained through the well fitted PL decay curves by using$I(t)=A_1\exp (-t/{\tau }_1)+A_2\exp (-t/{\tau }_2)+A_3\exp (-t/{\tau }_3)$ with${\tau }_{meas}(T)={\displaystyle \sum _{i=1}^n\frac{A_i{\tau }_i^2}{A_i{\tau }_i}}$ [27], and have an average value about 7.57 ns. Then we investigated the temperature dependent ns-PL lifetimes [τ_meas(T)] of a-SiN_xO_y samples with R = 1.5, detected at the emission wavelengths (λ_emi) of 430, 455, 470, 490, 505, 520, 535, 550 nm, respectively. As shown in Fig. 4(b), the τ_meas(T) under certain λ_emi keeps nearly stable at T<160 K, indicating radiative recombination dominates the recombination processes in this measurement temperature range [3,28,29]; while when the measurement temperature rises up (160 K~300 K), the increasing domination of nonradiative recombination results in the decreased τ_meas(T). Especially to deserve to be mentioned, the variation tendency of τ_meas(T) as a function of λ_emi under certain temperature is consistent with the spectral profile of stead state PL (see in Fig. 1). This phenomenon once again indicated unique PL properties of luminescent N-Si-O related defect states [30].

 

Fig. 4 (a) The measured ns-TRPL decay spectra and the relative fitted decay curves for a-SiN_xO_y samples with R = 1.5 at 300 K. (b) The ns-PL lifetimes vs. λ_emi for a-SiN_xO_y samples (R = 1.5) under measurement temperatures of 8 K, 80 K, 160 K, 240 K, 300 K, respectively.

Download Full Size | PDF

4. Conclusion

In summary, we have investigated PL quantum yield and the related radiative recombination mechanisms from tunable luminescent a-SiN_xO_y films. The PL QY of 8.38% has been achieved at PL peak 2.55 eV. We found that the PL QY is proportional to the concentration of N-Si-O bonding states, indicating the dominant contribution of luminescent N-Si-O bonding states in radiative recombination processes. We also checked the ns-PL lifetimes evolution profile vs. detected emission wavelengths, and further verified the luminescent N-Si-O defect states are responsible for the observed PL. The obtained high PL QY generated from N-Si-O bonding states suggested the possibility of high efficient Si-based photonic devices in a-SiN_xO_y systems.