1. Introduction

Big progress in the fabrication technology of InGaN/GaN multiple quantum wells (MQWs) in the recent decades has resulted in great success in optoelectronics devices for a wide range of applications, including display and general illumination [1–6]. Despite the existence of a large number of crystallographic defects, such as threading dislocations, the light emission efficiency of InGaN-based light-emitting devices is unexpectedly high. The localization of free carriers in the localized potential minima, induced by the energy band fluctuation in InGaN QWs, is believed to isolate carriers from dislocations, resulting in suppression of nonradiative recombination process [7]. However, the explicit role of localization on luminescence mechanism remains ambiguous. Recently, the atomistic tight-binding simulations show that the optical matrix elements and radiative recombination coefficients decrease due to random potential fluctuations in InGaN alloys [8]. In contrast, by considering the occupation of all electron and hole states for realistic carrier densities, it is predicted that the localization can give rise to an overall net increase of the radiative recombination coefficients [9]. It is also demonstrated that the localized electrons and holes are distributed independently in polar structures, as the localization effect is far less pronounced for electrons [10,11]. On the other hand, experimentally, the localization-induced anomalous variation of spectral features of InGaN/GaN MQWs still attract numerous investigations [12–14]. The temperature-dependent S-shaped behaviors of luminescence peak energy, which is considered as a fingerprint of carrier localization, differ significantly for different samples [15]. And the spectral properties may also be remarkably influenced by the carrier transfer process between different localization states [16,17]. Thus, further studies are necessary to gain an insight on the localization-related emission mechanism for InGaN/GaN MQWs.

In this work, two InGaN/GaN MQW samples with the same epitaxial structures are grown by adjusting the growth rate and time of the MQW active regions. The temperature-dependent photoluminescence (TDPL) spectra of both samples are compared, and an anomalous increase of peak PL height with increasing temperature are observed and analyzed carefully. Our study reveals that the carrier hopping dynamics between different localized potential traps may play an important role on the promotion of the luminescence efficiency of InGaN QWs.

2. Experimental process

Two InGaN/GaN MQW samples, designated as A and B, were grown on c-plane sapphire substrates via metalorganic chemical vapor deposition (MOCVD) system with close-coupled showerhead vertical reactor. All the samples consisted of a 2-μm thick Si-doped GaN layer, a two-period unintentionally doped InGaN/GaN MQW active region, and a 150-nm Mg-doped GaN layer. During the epitaxial growth of MQW region, the triethylgallium (TEGa), trimethylindium (TMIn) and ammonia (NH₃) were used as precursors for Ga, In and N sources, respectively. The growth temperatures of InGaN well and GaN barrier layers were 710 and 830 °C, respectively. The growth rates of GaN barrier layers were modulated by simply changing the TEGa flow rate. While for the growth of InGaN well layers, both the TEGa and TMIn source flow rates were changed carefully to modify the growth rate of InGaN layer, and meanwhile to keep a constant indium content of InGaN alloy in both samples [18,19]. For the growth of sample A, the growth rates of InGaN and GaN layers were 0.0065 and 0.0068 nm/s, while they were accelerated to 0.012 and 0.03 nm/s during the growth of sample B. Correspondingly, to maintain constant well and barrier widths, the growth time of GaN and InGaN layers of sample B was reduced. In short, compared with sample A, the MQW active region of sample B was grown with relative faster growth rates and shorter growth time.

The high-resolution X-ray diffraction (XRD) and TDPL measurements were used to characterize the MQW structural parameters as well as the optical properties. During the PL measurements, a 405-nm semiconductor violet laser with output power of 20 mW was used as excitation source so that only the carriers in InGaN QWs can be optically excited. The temperature was controlled by a closed-cycle refrigerator of CTI Cryogenics in the range from 20 to 300 K.

3. Results and discussion

Since the growth parameters are intentionally adjusted to keep the MQW epitaxial structures of two samples identical, it is reasonable that the (0002) XRD ω/2θ scan curves of samples A and B are almost the same as shown in the inset of Fig. 1, where the scan curve of sample B is vertically shifted for clarity. The presence of high-order superlattice satellite peaks clearly indicates that a well-defined MQW periodic structure is formed in both samples. According to the measured XRD scan curves, the GaN barrier thickness, the well thickness and indium content of InGaN layers are approximately 12.0 nm, 4.6 nm and 9%, respectively, for both samples. In general, a larger volume of active regions is highly advantageous to promote the performance of optoelectronic devices, such as high-power LEDs and laser diodes [3,4]. Therefore, it is meaningful to investigate the properties of thick InGaN QWs from the view of practical applications, and thus the InGaN QWs are relatively thicker in our work. However, in Fig. 1 where the room temperature PL spectra of samples A and B are plotted, the PL intensity of sample B is much higher than that of A, and the PL peak wavelength of both samples are nearly the same, i.e. about 461 nm.

 

Fig. 1 Room-temperature PL spectra of samples A (black) and B (red) measured under excitation power of 20 mW. The inset shows the (0002) high resolution XRD ω/2θ scan of both samples.

Download Full Size | PDF

Figure 2 displays the PL spectra of both samples measured at varied temperatures. For both samples at very low temperatures, e.g. 30 K, a small side peak can be seen at the lower energy side of PL spectra. The energy difference of peak positions between InGaN main peak and side peak is about 100 meV for both samples, which is roughly the optical-phonon energy of GaN [20]. Thus, the weak side peak can be ascribed to the one-phonon-assisted transitions, i.e. the phonon replica of main peak, where a phonon participates in the radiative recombination process of InGaN main peak. For sample A, the PL spectral shape is almost symmetric and there is only one main luminescence peak in the whole temperature range, except the weak phonon replica at very low temperatures. It means that the potential distribution of localized luminescent centers is relatively uniform in sample A. However, for sample B, an additional PL peak located at the higher-energy side of the main peak appears at very low temperatures. The carriers generate randomly in the whole energy landscape of InGaN QWs. And then, those photo-generated carriers may stay and recombine radiatively where they are generated due to their limited thermal mobility. Therefore, the appearance of two PL peaks at low temperatures means that the energy potential of luminescent centers in sample B is very inhomogeneous and can be roughly divided into two groups, i.e. the deep localized potential traps (D-traps) where the main PL peak originates, and the shallow localized traps (S-traps) which are the origin of the higher-energy peak. Particularly, at a very low temperature of 30 K, the intensity of the higher-energy peak is almost the same as that of the main peak. It indicates that a large number of carriers can be confined in the S-traps and then recombine by emitting a higher-energy photon, as the carrier mobility is very low and the migration process between different localized traps can be ignored, giving rise to a significant emission of higher-energy peak. However, when the temperature increases higher than about 120 K, the higher-energy peak vanishes. This may be mainly caused by the loss of thermally excited carriers hopping out from the S-traps, since at higher temperatures the carriers will be thermally activated and can be hardly confined in the shallow potential traps.

 

Fig. 2 PL spectra of samples A (left) and B (right) at several different temperatures, measured under excitation power of 20 mW. The arrows indicate the main luminescence peaks and the colored circles denote the variation of main peak position. At very low temperatures, the phonon replica of main peak can be seen for all samples, and there is an additional higher-energy peak at the right side of main peak only for sample B.

Download Full Size | PDF

The appearance of higher-energy peak at low temperatures also leads to an anomalous variation of integral PL intensity for sample B, as shown in Fig. 3 where the variation of integral intensity and main peak height of both samples with increasing temperature are depicted. First, it is noticed that both the PL integral intensity and main peak height at low temperatures are lower for sample B than for A, but they become higher for sample B than for A at high temperature. It implies that the thermal stability of sample B is better, which may be related to the existence of the additional S-traps in InGaN QWs of sample B and will be analyzed later. On the other hand, in Fig. 3(a) the integral intensity, i.e. an integral of the spectral intensity in the whole spectrum, shows a monotonous thermal decay as predicted by Arrhenius equation for both samples, which is widely attributed to the enhanced non-radiative recombination process induced by the elevated temperatures [21]. On the contrary, in Fig. 3(b) it is interesting to notice that when the temperature increases from 20 K (0.05 K⁻¹), the main peak height of sample B increases first and reaches the maximum at about 100 K (0.01 K⁻¹), and then decrease quickly toward room temperature.

 

Fig. 3 Temperature dependencies of integral PL intensity (a) and main peak height (b) for samples A and B measured under excitation power of 20 mW.

Download Full Size | PDF

For detailed analysis of the anomalous increase of main peak height with increasing temperature, the PL spectra of sample B measured at several different temperatures are shown together in Fig. 4. It is obvious that the temperature-induced spectral variations of main peak and higher-energy peak are different from each other. The higher-energy peak intensity monotonously reduces with increasing temperature 30 to 150 K. However, the intensity of main peak increases firstly to a maximum at 90 K, and then decreases monotonously, which is similar to the results shown in Fig. 3(b). The detailed peak height variations of main peak and higher-energy peak in temperature range from 20 to 150 K are depicted together in the inset of Fig. 4. It is found that the height of the additional higher-energy peak decreases rapidly from 20 to 90 K, accompanied with a slight increase of the main peak height. It should be pointed out that the fast attenuation of higher-energy peak cannot be ascribed to the temperature-induced enhancement of non-radiative recombination process alone, since the integral PL intensity of sample B is almost stable in this temperature range as shown in Fig. 3(a). On the other hand, the slight increase of main peak height of sample B implies that additional carriers are supplied into the D-traps. However, the generation rate of photo-generated carriers in InGaN QWs is stable because of the constant excitation power of 405-nm laser used in TDPL measurements. Therefore, it is only possible that those additionally supplied carriers in D-traps are transferred from the S-traps.

 

Fig. 4 PL spectra of sample B at several selected temperatures, measured under the laser power of 20 mW. The inset shows the variations of peak height of main peak (black squares) and higher-energy peak (red circles) with temperature, respectively.

Download Full Size | PDF

Actually, at very low temperature, the photo-generated carriers distribute randomly across the whole InGaN QWs due to the very small thermal mobility of carriers. As temperature rises, the carriers can be thermally activated to populate the higher energy levels in the potential traps, leading to a spectral blueshift as shown in Fig. 4 for the higher-energy peak. The hot carriers occupying higher energy levels are more possible to overcome the energy barrier to transfer between different localized potential traps via hopping process. Therefore, the fast reduction of higher-energy peak can be mainly ascribed to the loss of hot carriers which jump out from S-traps. A portion of the migrated carriers may be captured by the non-radiative recombination centers, while another portion of those carriers may transfer into the D-traps with deeper potential. As a result, when the temperature increases in low temperatures from 20 to 90 K, the additional carriers coming from S-traps can transfer into D-traps, resulting in the promoted emission of the main peak. As the temperature increases further, e.g. beyond 100 K, the non-radiative recombination process gradually becomes predominant, leading to a general reduction of the PL intensity.

As is well known, the localized luminescent centers are mainly originated from the segregation and fluctuation of indium in InGaN alloys. Thus, the existence of two groups of localized states, i.e. D-traps and S-traps, indicates that the indium composition in sample B is more inhomogeneous than in sample A whose TDPL spectra exhibit only one luminescence peak originated from the uniform distribution of indium. Also, it should be noticed that the main peak position of sample A shows a blueshift-to-redshift behavior as temperature increases from 30 to 220 K, whereas there is only a blueshift for sample B, as denoted by colored circles in Fig. 2. It is known that the blueshift process is associated with temperature-induced carrier’s hopping up to higher energy levels in localization states, while the redshift is originated from the bandgap shrinkage with increasing temperature. Thus, the lack of redshift indicates that the carrier’s hopping up process is more significant than the temperature-induced bandgap shrinkage at high temperatures, implying that the localization effect induced by D-traps in sample B is very strong.

In fact, during the epitaxial growth of InGaN/GaN MQWs, when the growth rate increases, the migration time of adatoms on the growth surface reduces, and thus the adatoms may have less chance to achieve the energy minima on the growth surface, which may result in a rougher growth surface of both GaN and InGaN layers. In addition, during the growth of InGaN layers the migration ability of indium adatoms on the growth surface is weakened further due to the lower growth temperature. From the view of growth kinetics, the randomly deposited surface-adsorbed In atoms with weaker migration ability may be less possible to reach the optimal positions on the growth surface, resulting in a more inhomogeneous distribution of In content. Thus, combined with the reduced migration time of indium adatoms and a rougher GaN surface for InGaN growth, the aggregation of indium atoms in InGaN alloy may be very severe [19]. As a consequence, compared with sample A, the distribution of indium atoms in InGaN QWs of sample B with fast growth rate may be more inhomogeneous, resulting in a stronger localization effect, and thus two kinds of localized traps with different potential depth appear.

Therefore, based on the aforementioned discussions, it is reasonable to deduce that for sample B the improved emission efficiency at room temperature, and better thermal stability of luminescence intensity during TDPL measurements, may be mainly ascribed to the more inhomogeneous distribution of localized potential traps, stemmed from the enhanced fluctuation of indium content in InGaN QWs grown with fast growth rate. Furthermore, combined with the results obtained from Figs. 4 and 5, we think that the beneficial effect of inhomogeneous distribution of indium composition in fast-grown InGaN QWs on the light emission may come from two reasons. First, the D-traps formed by the In-rich clusters can suppress the loss of carriers via non-radiative recombination due to the stronger restriction ability on carriers. Second, the S-traps, originated from the lower-In-content localized regions with lower restriction potential barriers, are usually considered as defect free regions, where carriers can be optically generated without being captured by defect-related non-radiative recombination centers. And then, at RT, most of the generated carriers are thermally activated and may transfer into the D-traps and contribute to the main luminescence process, as the relax time of carriers from S-traps into D-traps may be shorter than the carrier life in S-traps due to the small localization potential barriers. In one word, the S-traps may be considered as a kind of carrier generation centers with weak carrier restriction, where carriers can be only optically generated but do not recombine radiatively. Therefore, the existence of S-traps may be also helpful to improve the luminescence efficiency through an indirect way, although there is no apparent light emission coming from S-traps themselves at room temperature.

4. Summary

In summary, the temperature-dependent optical characteristics of two InGaN/GaN MQWs grown with different growth rates are investigated. Compared with sample A, the luminescence intensity of sample B with fast growth rate is significantly improved at room temperature, although the epitaxial structures of both samples are almost identical, which is verified by the XRD scan curves. Besides, an additional higher-energy peak can be seen in the TDPL spectra of sample B at temperatures below about 150 K. It is found that the main peak height of sample B increases anomalously with temperature in the low temperature range, which is attributed to the migration of hot carriers from S-traps into D-traps. It is considered that the enhanced PL intensity of sample B at room temperature may come from both the stronger carrier confinement of D-traps and the carrier transfer process between S-traps and D-traps.