1. Introduction

Si quantum dots (Si QDs) have attracted much attention because of their novel optical and electronic characteristics, which can be utilized to develop the advanced nano-devices, such as non-volatile memories, solar cells and light emitting devices [1–10]. As a typical indirect bandgap semiconductor, bulk Si has been excluded as a good choice for the efficient light source. But, unlike their bulk counterpart, Si QDs exhibit efficient luminescence due to the improved radiative recombination probability. So far, light emission from Si QDs have been extensively studied and the emission wavelength usually falls into the visible to near-infrared (NIR) range which is above the bandgap of bulk Si (1.12eV) due to the quantum size effect [5–8,11]. However, the on-chip optoelectronic integrations call for efficient light sources working in the wavelength range of 1200nm-1600nm. So it is highly desired to get the light emission with wavelength compatible with the optical telecommunications by modifying the electronic structure of Si QDs.

Doping in semiconductors is an effective way to change the electric properties and it has been widely used in bulk Si. For Si QDs, impurity doping is also an important method to change their electric and optical characteristics [12–14]. Fujii et al. reported that an additional photoluminescence (PL) peak at ~0.9eV was observed in the phosphorus-doped (P-doped) Si nanocrystals dispersed in phosphosilicate glass thin films when the temperature was as low as 5K [15, 16]. They attributed the emission to the radiative recombination of electron hole pairs in P_b centers. Besides, a subband light emission was also detected in the boron/phosphorus co-doped (B/P co-doped) Si nanocrystals and it was considered to arise from transition between the donor and acceptor states [17,18]. In our previous work, we prepared P-doped Si QDs embedded in SiO₂ host matrix and observed the sub-band light emission (~1.3μm) at room temperature [19]. The dependence on the annealing temperatures and doping concentrations was briefly studied. However, the origin and detailed luminescent behaviors of this sub-band light emission are still not fully understood and further investigation is necessary.

The time-resolved photoluminescence (TRPL) and temperature dependent PL measurements are powerful tools to study the PL characteristics and to give deep insight into the luminescence mechanism as well as the radiative and/or non-radiative recombination processes of Si nanocrystals [20–24]. In this paper, the luminescence behaviors of P-doped Si QDs/SiO₂ multilayers with ultra-small dot sizes are systematically studied. It is found that the two independent emission bands, one centered at 890nm and another centered at 1250nm, exhibit different behaviors, which indicates that they have different origins. The P doping concentrations can affect the luminescence properties obviously. Moreover, the TRPL measurements are performed to clarify the competing relationship between the two bands. The temperature dependent PL measurements are also used to determine the origin of the 1250nm emission.

2. Experimental

The phosphorus-doped (P-doped) hydrogenated amorphous Si/SiO₂ (a-Si:H/SiO₂) stacked structures were fabricated on p-type mono-crystalline Si wafers (1-3Ω/cm) and quartz substrates in a conventional plasma enhanced chemical vapor deposition(PECVD) system by alternatively repeating the a-Si deposition and in situ plasma oxidation process. The RF power was 50W and the substrate temperature was fixed at 250°C during the deposition process. The a-Si:H layers were obtained by using SiH₄ with the flow rate of 5 sccm. The thicknesses of a-Si layers were varied from 2nm to 11nm by changing the deposition time of SiH₄. Meanwhile, PH₃ was added into the precursor gas to realize P doping in a-Si layers. The flow rate of PH₃ was varied to obtain different P doping concentrations and the nominal P concentration was calculated by the gas ratio of PH₃ and SiH₄. As a reference, we also prepared un-doped and boron-doped (B-doped) a-Si/SiO₂ multilayers under the same conditions. The as-deposited samples were dehydrogenated (DH) at 450°C and subsequently annealed at 900°C for crystallization of a-Si layers. Both the dehydrogenation and annealing processes last for 1h under N₂ ambient.

The steady-state photoluminescence (PL) spectra were measured by a HORIBA Jobin Yvon synapse CCD detector in the short wavelength range (600nm-1000nm) and a liquid N₂ cooled InGaAs detector in the long wavelength range (1000nm-1600nm). The excitation source was a 30 mW He-Cd laser with the wavelength of 325 nm. The temperature dependent photoluminescence spectra in the range of 20K-300K are obtained by using a Sumitomo Cryogenics HC-4E cooling machine. The time-resolved photoluminescence (TRPL) spectra were measured by using an Edinburgh instruments FLS980fluorescence spectrophotometer equipped with a Xenon flash lamp (λ_exc = 325 nm, 1-2μs pulse width and repetition rate 100Hz).

3. Results and discussion

Figure 1 shows the cross-sectional TEM images for P-doped Si/SiO₂ multilayers after annealing at 900°C. The inset is the magnified image of a single Si QD. In this figure, the Si and SiO₂ layers can be clearly identified and the periodical structure is remained after 900°C annealing. The thickness of the Si and SiO₂ layer is 2.1nm and 3.6nm, respectively. The high-density Si QDs are formed in the Si layers and they are perfectly confined by the SiO₂ layers. The average size of the Si QDs is 1.99 ± 0.09nm, which is consisted with the thickness of the Si layers. With increasing the depositing time of the SiH₄, the Si layers will be thicker, and the size of Si QDs will be increasing accordingly, which has been revealed in our previous work [6, 25]. In the inset we can find that the diameter of the Si QD is consistent with the thickness of the Si layer. Meanwhile, the inter-planar spacing distance is 0.31nm, which indicates the <111> crystalline structure of Si.

 

Fig. 1 Cross-sectional TEM images for P-doped Si/SiO₂ multilayers after 900°C annealing. Inset is the magnified image for a single Si QD.

Download Full Size | PDF

Usually, only a broad emission band around 800nm-900nm can be observed for un-doped Si QDs/SiO₂ multilayers which can be attributed to the radiative recombination of photo-excited electron-hole pairs via the luminescent states at the Si QDs/SiO₂ interfaces and/or the conduction-to-valence transition [2,8]. However, for P-doped Si QDs/SiO₂ multilayers with the dot-size of about 2nm, there are two independent emission bands: one is centered at 890nm (denoted as Band I), which almost has the same peak position and shape as that of un-doped samples; another is centered at 1250nm (Band II) which cannot be detected in the un-doped samples. Figure 2(a) shows the room temperature PL spectra of P-doped samples in the short-wavelength range (650nm-1000nm) after 900°C annealing (blue line). For comparison, the PL spectra of un-doped and B-doped samples are also presented. We can find that the PL peak location and shape of Band I for P-doped samples are almost the same as that of the un-doped ones, which indicates that they have the same origin due to the radiative recombination centers at the Si QDs/SiO₂interface region. Meanwhile, the intensity of Band I for the P-doped samples is enhanced in comparison with the un-doped ones. This is because the surface defects, such as the Si dangling bonds, are passivated by P atoms [15,26], which will improve the radiative recombination. For the B-doped samples, only a weak PL peak located at ~780nm is observed. Recently, the PL signal from B-doped Si nanocrystals was reported by Veettil et al. The authors suggested that the dangling bonds at the Si/SiO₂ interface were passivated by B atoms and the PL signal was enhanced accordingly [18]. In our samples, this peak is similar to the band-to-band emission (750nm) of 2.2nm Si QDs due to the theoretical calculation [27]. It is possible that the surface states of the Si QDs are strongly modified by the B atoms and only the band-to-band recombination remains. Figure 2(b) shows the room temperature PL spectra of un-doped, P-doped and B-doped Si QDs/SiO₂ multilayers in the wavelength range 1000-1600nm after 900°C annealing. Here we can see that a broad luminescence band centered at ~1250nm (Band II) appears in the P-doped Si QDs/SiO₂ multilayers. For the B-doped Si QDs, a broad PL signal below the bulk Si bandgap was also reported and it was attributed to the acceptor relating states [28]. However, no low-energy PL signal can be found in our B-doped samples. It is worth pointing out that the emission band centered at 1250nm can only be detected in P-doped Si QDs/SiO₂ multilayers with ultra-small dot sizes (<3nm). For samples with the dot size larger than 4nm, the luminescence peak cannot be observed any more.

 

Fig. 2 Room temperature PL spectra of 900°C annealed Si QDs/SiO₂ multilayers in (a) 600-1000nm range and (b) 1000-1600nm range. (c)Integrated PL intensity of P-doped Si QDs/SiO₂ multilayers annealed at 900°C with various doping concentrations.

Download Full Size | PDF

The light emission below the bandgap of bulk Si was previously observed in P-doped Si QDs at the temperature as low as 5K which was attributed to the radiative recombination of electron-hole pairs at the P_b centers [15]. It was found that the sub-band light emission intensity was gradually reduced with increasing the P doping concentration in the films and the emission finally disappeared completely. However, in our case, the PL intensity of P-doped Si QDs/SiO₂ multilayers shows the different behaviors depending on P doping concentrations. Figure 2(c) shows the integrated PL intensity of un-doped and P-doped Si QDs/SiO₂ multilayers annealed at 900°C as a function of P doping concentrations. For the un-doped sample, Band I is relatively low and Band II can’t be observed. For the P-doped samples, the intensity of Band I is enhanced due to the passivation effect of P atoms, which has been discussed above. Meanwhile, Band II centered at ~1250nm emerges in the NIR range. It is found that the intensity of Band I is monotonously decreasing with the P concentration increasing, and it is almost completely quenched when the doping concentration is about 4%. However, the luminescence intensity of emission band at 1250nm is gradually enhanced with increasing the P doping concentration from 0.2% to 3%. When further increasing the concentration to 4%, the PL intensity of Band II is greatly reduced. It looks like that the competing relationship exists between the two emission bands. The quantum yield of two emission bands was measured by using a Xe lamp (Xenon short ARC) working at 325 nm as an exciting source and the photomultiplier detector (Hamamatsu 928 PMT) with an integrating sphere and calibrated by the reference sample. The quantum yield (QY) is about 0.5% (Band I) and 0.81% (Band II), respectively for 3% P-doped sample [29].

It is interesting to study the origin of the sub-band light emission because the peak location is exactly in the near infrared range and is compatible for the optical telecommunication. Recently, Pi et al. [26] built a theoretical model to study the location of P impurities in Si QDs embedded in SiO₂ host matrix. They found that, for the oxygen-passivated Si QDs with the dangling bonds on the surface, part of P atoms can be incorporated into the inner side of Si QDs at the substitutional sites. If the dot size is small enough, the simulation results also point out that the deep levels will be generated in the gap of Si QDs and they can act as the luminescent centers to emit the sub-band light. Our experimental results are in good agreement with the theoretical expectation. With the concentration increasing, it is natural that more P induced deep levels will be generated. Since the photo-excited electron-hole pairs can recombine radiatively via either the interface states or the P-induced deep levels, it is reasonable that more electron-hole recombination will occur via the P-induced deep levels when the doping concentration is higher. As a consequence, the intensity of the emission band at 1250nm will be enhanced. On the other hand, the suppression of both emission bands at the P concentration of 4% can be explained in two aspects. One reason is the strong Auger recombination process in the heavily doped Si QDs/SiO₂ multilayers. Since the recombination time of the Auger process is much shorter than that of the radiative recombination [30], both PL intensities will be significantly suppressed. Another possibility is the damage of crystalline structures caused by the large amount of incorporated impurities in the ultra-small Si QDs [31]. The intensively moving impurities will generate a great deal of defect states, which can act as non-radiative recombination centers and suppress the emission intensity accordingly.

To better understand the recombination mechanism in P-doped Si QDs/SiO₂ multilayers, the time-resolved photoluminescence (TRPL) measurements are performed. The multi-exponential function (1) was used to fit the PL decay curves which was usually used in amorphous and nano-crystalline Si-based luminescent materials though [32, 33].

where A_i and τ_i represent the amplitude and decay time of each exponential component. In our case, two exponential components were needed to well fit the measurement data. The typical TRPL spectrum with fitting curve of band Ι and band Π for 1% P-doped sample are shown in Fig. 3(a) and 3(b), respectively. The fitting parameters were also given in the figures. Figure 3(c) show the intensity-weighted averaged PL lifetime which can be determined by [34]:

 

Fig. 3 TRPL spectrum with fitting curve of band Ι (a) and band Π (b) for 1% P-doped sample; (c) The intensity-weighted averaged PL lifetime of the two emission bands of the Si QDs/SiO₂ multilayers with various doping concentrations.

Download Full Size | PDF

For the un-doped samples, the decay time of the ~890nm PL is about 58.36μs. Due to the indirect bandgap of bulk Si, the lifetime of the electron-hole recombination in Si nanocrystals is usually in the order of 10-100μs, which is relatively higher than direct bandgap materials. In our samples, the lifetime of the PL is in good agreement with previous reports and it can be attributed to the intrinsic electron-hole recombination in Si QDs. For the P-doped samples, the decay time for the two emission bands can be detected. For the 0.2% P-doped sample, the decay time of Band I is 61.03μs. It is found that with the doping concentration increasing, the decay time is decreasing monotonously and it is only 24.37μs for the 3% P-doped sample. This phenomenon implies that a short lifetime process is enhanced with increasing the concentration. It is possible that the non-radiative Auger recombination process contributes to the reduced lifetime. As the doping concentration increases, more electrons are generated and the Auger recombination process is enhanced, which leads to the reduction of the lifetime of Band I.

Meanwhile, it is found that the lifetime for Band II is almost unchanged (9-11μs) with the P doping concentration, which is one order of magnitude less than that of Band I, suggesting that the new emission route is introduced in P-doped Si QDs/SiO₂ multilayers. As mentioned before, P dopants may create the deep level in the gap of Si QDs, which is responsible for the subband emission Band II. The much shorter lifetime for Band II corresponds to the larger recombination rate, which means that photo-excited electron-hole pairs prefer to recombine via the deep levels instead of the interface states. The gradually enhanced PL intensity of Band II as well as the constant lifetime with the P doping concentrations indicates that more photo-excited electron-hole pairs recombine via the deep levels. The lifetime can explain the origin of the competing relationship of the two emitting mechanisms, and it can support the theoretical assumption that the electrons prefer to relax to the deep level and lead to a NIR light emission.

In order to clarify the nature of Band II, the temperature dependent PL spectra are also obtained. Figures 4(a) and 4(b) show the PL spectra of the 0.2% and 2% P-doped Si QDs/SiO₂ multilayers measured in the range of 20-300K. It is found that the PL peak shifts to the shorter wavelength range when the temperature decreases. The evolution of the PL energy with the temperature can give us the information whether the recombination occurs in the Si QDs or via the interface states [21]. For Si nanocrystals, the bandgap is decreased with the temperature increases [20]. The redshift of Band II with the temperature increasing can be explained by the decreasing of the bandgap, which indicates that the radiative recombination originates from the deep level and the valence band of the Si QDs instead of the interface states.

 

Fig. 4 Temperature dependent PL spectra of Band II for (a) 0.2% and (b) 2% P-doped Si QDs/SiO₂ multilayers (c) Integrated PL intensity of Band II for P-doped Si QDs/SiO₂ multilayers with various concentrations.

Download Full Size | PDF

Figure 4(c) shows the integrated PL intensity of Band II for P-doped Si QDs/SiO₂ multilayers with various concentrations. We can find that the PL intensity of the 2% P-doped samples is higher than the 0.2% ones due to more P induced deep levels. For the 2% sample, the PL intensity is enhanced with the temperature increasing from 20K to 80K, but the intensity is decreasing monotonously when the temperature is beyond 80K. The evolution trend of the PL intensity for the 0.2% sample is the same, but the temperature for the maximum intensity is smaller (60K). The similar phenomenon has been observed in the un-doped Si nanocrystals previously [21, 35], and their theory can also explain our observation. In the high temperature range, the electron-hole recombination is dominated by the non-radiative process such as the Auger recombination. With the temperature increasing, the non-radiative process is enhanced. So it is reasonable that the intensity is decreasing monotonously.

When the temperature is low enough, the radiative recombination is dominant due to the non-radiative recombination process is significantly suppressed. However, it has been reported that the lifetime of radiative recombination is increased with decreasing the temperature as in un-doped Si QDs [21, 35–37]. Therefore, the corresponding luminescence intensity is reduced gradually with decreasing the temperature since the radiative recombination rate becomes small. It is reported that the PL intensity in the low temperature range can be described with the equation [38],

where N is the total number of Si QDs, τ_r(T) the radiative lifetime, and G = σ•I_ex describes a generation term depending on the absorption cross-section σ and the areal flux density of exciting photons I_ex. When the doping concentration is high, the number of Si QDs with P atoms incorporated is larger, corresponding to a higher value of N. So the PL intensity of 2% P-doped sample is stronger.

It is interesting that the transition temperature for the maximum PL intensity is not the same and it is smaller for the 0.2% P-doped Si QDs/SiO₂ multilayers. In the previous work, the smaller temperature was usually observed at the low areal flux density of exciting photons I_ex which is leading to a smaller value of G [35]. The value G•τ_r(T) = 1 is corresponded to the transition temperature for the maximum PL intensity which is a cut-off point of the low temperature range (radiative recombination dominated) and the high temperature range (non-radiative recombination dominated). In our case, the smaller G is contributed by the low absorption cross-section σ. The value σ is the product of the density of electronic states and the oscillator strength of the optical transition [38]. When the doping concentration is low, the number of deep levels generated by the P atoms is smaller and the density of electronic states is lower and the corresponding σ is smaller. Therefore, the larger lifetime τ_r (T) is needed to reach the point with the maximum PL intensity, which results in the lower transition temperature as seen in the 0.2% P-doped sample compared with that of 2% P-doped one. The temperature dependent PL spectra of Band II exhibit a typical PL behavior of Si nanocrystals and a strong relation to the P doping concentrations, which further demonstrate that Band II originates from the deep levels inside the Si QDs rather than the interface states.

4. Conclusion

We have fabricated the P-doped Si QDs/SiO₂ stacked structures with the dot size of about 2nmby annealing a-Si:H/SiO₂ multilayers at various temperatures. In the P-doped Si QDs/SiO₂ multilayers, two light emission bands are observed: one is centered at 890nm, which is originating from the recombination centers at the Si QDs/SiO₂ interface states, and another is centered at 1250nm due to the P-induced radiative deep levels in the gap of Si QDs. The two emission bands exhibit different behaviors depending on P doping concentrations. The emission intensity of 1250nm band is enhanced with increasing the doping concentration because more P impurities are incorporated into Si QDs. Moreover, the luminescence lifetime of 1250nm emission band is one order of magnitude shorter than that of the 890nm band, indicating the higher recombination rate via P-induced deep levels. The temperature dependent PL spectra show that the recombination of the NIR emission occurs inside the Si QDs instead of the interface states, which further demonstrates the origin of the P-induced deep levels. Our results demonstrate that doping is an effective way to get the emission from Si QDs with the wavelength compatible with the optical telecommunications.