1.Introduction

Despite commercial success of III-nitride-based devices for application in visible and ultraviolet optoelectronics and in high-power and high-frequency electronics, their performance is significantly limited due to the lack of native substrates [1,2]. The use of foreign substrates such as sapphire, Si, SiC, GaAs or GaP leads to problems with high-dislocation density, mosaic crystal structure, biaxial induced stress, and wafer bowing, etc [3,4]. Of course, these drawbacks can be minimized by the use of homoepitaxy onto bulk GaN or AlN substrates. Unfortunately, despite many years of intense research, the high equilibrium pressures and temperatures of the III-nitrides so far have rendered the growth of bulk substrates for homoepitaxy with production grade sizes smaller. Among all the bulk growth techniques for nitrides under investigation today, hydride vapor phase epitaxy (HVPE) is considered to be a practical method to obtain thick GaN layers since it utilizes a process at more favorable conditions,namely,low pressure and relatively low growth temperature [5]. Combined with laser lift-off (LLO) [6,7] techniques, a larger size free standing GaN-substrate can be obtained. The biggest advantage of the technique is its ability to produce high quality material with a largely reduced density of dislocations as demonstrated by the observation of numerous sharp exciton features in the photoluminescence (PL) spectrum.

It is known that there have been a number of photoluminescence studies on bulk GaN prepared by different growth techniques [8–13]. Most of them focus on identifying the origin of exciton emission peaks. Though similar properties have been observed in these GaN samples, the intensities of the exciton emissions vary from sample to sample depending on the crystal quality. Since HVPE is the method that dominates the GaN substrate market today, a more detailed understanding of the optical quality of GaN substrate after removing the foreign substrate is required for later design of GaN-based devices.

In this study, we report the temperature-dependent PL study of the Ga-/N-faces of a free-standing GaN substrate, fabricated by using HVPE and LLO process. A different effect of the strong excitonic luminescence ($D_2^0$X) at 3.473 eV for the Ga-/N-faces has been observed. Meanwhile, a detailed comparative analysis on the temperature-dependent ratio of the peak intensity of $D_2^0$X to $X_A^{n=1}$ for both faces has been also given. In addition, the influence of dislocation density on exciton features has been further investigated by Cathodoluminiscence (CL) in a scanning electron microscope (SEM).

2. Experimental

The sample used in our work was unintentionally doped GaN grown on sapphire using home-made horizontal HVPE. The growth system was equipped with both hot wall quartz tube reactor and a resistive furnace. During HVPE growth, the temperature was kept at 1040°C and the growth pressure was 760 Torr. HCl and Ammonia were used as input active gases. Before GaN deposition, HCl flowed over Ga source to form gaseous gallium chlorides at 830°C. The mixture of H2 and N2 (mixed ratio of 1:1) were used as carrier gas. Finally, the GaN layer with a thichness of 350μm was formed. The thick GaN layer was then separated from the sapphire substrate by LLO process and was grinded and double polished. In the PL measurement, the sample was mounted in a variable temperature cryostat operating from 3 to 300 K. The 325 nm line of a He-Cd laser was used for excitation. CL measurement was performed at room temperature using a CL system (MonoCL3 + ) attached to a field emission scanning electron microscope (Quanta 400 FEG).

3. Results and discussions

The structure quality of bulk GaN has been reported elsewhere [14]. Here, the structural properties of the Ga-/N-faces of a free-standing GaN substrate have been characterized by high- resolution X-ray diffraction (HRXRD). The full widths at half maximum of the 2θ-ω scan of the (0002) reflexes for the Ga-/N-faces are 30 and 41 arcsec, respectively, which indicates high quality for both Ga-face and N-face.

Figure 1 shows a typical PL spectrum of the Ga-face and N-face of a free-standing HVPE GaN sample at 3 K. It is found that both faces have well-defined excitonic emission lines. The near-band-edge (NBE) emission of both faces is dominated by the neutral donor bound-exciton transition ($D_2^0$X) [8]. The FWHMs of the $D_2^0$X for the Ga-/N-faces are 2.4 and 1.9 meV under low power excitation, respectively. This indicates high quality of the sample investigated in the present work. Other PL lines due to the free-excitonic line$X_A^{n=1}$, the neutral donor bound-exciton transition ($D_1^0$X) for both faces, and the transition involving the excited 2s-like state of the donor ($D_2^0$X)_{n = 2} for Ga-face are also seen here [9].

 

Fig. 1 Low temperature PL spectrum of the Ga-/N-faces of a free-standing HVPE GaN.

Download Full Size | PDF

Figure 2 shows the temperature-dependent PL spectra in the temperature range of 3―140 K for both faces of the same sample. The intensity scale is the same for spectra show in Figs. 2(a) and 2(b). A pronounced relative intensity change of the neutral donor bound-exciton ($D_2^0$X) and the free exciton $X_A^{n=1}$transitions is observed as the temperature increases. As can been clearly seen from Figs. 2(a) and 2(b), the relative intensity of the $D_2^0$X for both faces decreases with increasing temperature while the relative intensity of the $X_A^{n=1}$ increases. Such feature has been observed earlier and attributed to thermalization of the bound excitons into free excitons [15,16]. Here, it is interesting to see that the temperature dependence of the $D_2^0$X from the N-face (see Fig. 2(b)) is apparently different from that of the Ga-face (see Fig. 2(a)). When the temperature is raised to 50 K, the $D_2^0$X of the N-face loses its dominant position. However, even though the temperature is raised up to 80 K, the $D_2^0$X of the Ga-face still maintains its dominant position among the other peaks. The mechanism behind this phenomenon is not clear.

 

Fig. 2 Temperature dependent PL spectra of the Ga-/N-faces of a free-standing HVPE GaN.

Download Full Size | PDF

In order to obtain more information about the effects of temperature on the$D_2^0$X exciton for both faces, the temperature-related ratios of the relative intensities of the $D_2^0$X peak to $X_A^{n=1}$peak for the Ga-face (labeled L1) and the N-face (labeled L2) are given in Fig. 3(a).The ratios obtained here exactly reflect the competition between $D_2^0$X exciton and free exciton at different temperatures. It should be noted that the crucial dot line (labeled L3) equal to 1 is also plotted in Fig. 3(a). Apparently, a monotonic decrease of both L1 and L2 occurs with increasing temperature. The $D_2^0$X transition dominates above the line L3 while the free exciton transition dominates below the line L3. The crossover points for the lines offer the critical point of temperature where the dominance of exciton changes. Here, the critical points of temperature are about 80 and 43 K for the Ga-/N-faces, respectively. According to the Arrhenius fitting, different activation energies for the $D_2^0$X of the Ga-/N-faces are also obtained, shown in Fig. 3(b). It is found that the activation energy for the N-face (21.6 meV) is much smaller than that of the Ga-face (30.7 meV). This activation energy difference may be an important factor accounting for the significant different critical temperature between the Ga-face and N-face. As can been seen from Fig. 3(a), another notable feature shows that the line L1 and L2 are almost parallel. After careful analysis, it is found that the intensity ratio of L1 to L2 is actually a constant and approximately equal to 2.8. The result may be related to different density, distribution, type of defects of both Ga-face and N-face. Consequently, it also indicates that the N-face may contains a larger density of nonradiative recombination centers than that of the Ga-face.

 

Fig. 3 (a) Temperature dependent ratio of the relative intensity of the $D_2^0$X peak to $X_A^{n=1}$ peak for both faces. (b) The variation of the peak intensity of the $D_2^0$X for both faces and their corresponding Arrhenius fittings.

Download Full Size | PDF

To further clarify the mechanism above, AFM and CL images are taken at room temperature for both faces. The AFM images (not shown here) show a smooth surface morphology for both faces. The measured RMS values of surface roughness are about 1.07 nm and 1.49 nm for the Ga-/N-faces, respectively. However, distinct features occur for both faces in the CL images, as shown in Fig. 4.Widely distributed dark spots are observed against the GaN bulk background. This dark spots should be related to the threading dislocation. Since the dislocations should not emit any luminescence, the positions corresponding to dislocations in the Ga-/N-faces regions are relatively dark in the CL images. Due to defects such as threading dislocation that act as nonradiative recombination centers, one would expect that more dark spots are observed in the CL images of the N-face. The average dislocation densities for the Ga-/N-faces (see Fig. 4(a) and Fig. 4(b)) are roughly determined to be 5.8x10⁶/cm² and 12.8x10⁶/cm², respectively. The experimental outcomes are consistent with the expected results. In addition, it is striking that the ratio of the average dislocation density of the N-face to the Ga-face is about 2.2. This value is very close to the the intensity ratio of L1 to L2 above. Thus, it can be inferred that a larger density of nonradiative recombination centers is another major factor which leads to lower dominant temperature of the $D_2^0$X transition for the N-face. Furthermore, the experiment result may also indicate that there exists some correlation between the dislocation density and the thermalization of the bound excitons. The high dislocation densities of the N-face as well as their complex distributions, compared to that of the Ga-face, may give important contributions to lower activation energy of the $D_2^0$X.

 

Fig. 4 (a)-(b) Typical CL images of the Ga-/N-faces of a free-standing HVPE GaN.

Download Full Size | PDF

4. Conclusions

In summary, variable-temperature photoluminescence studies have been performed on the Ga-/N-faces of a free-standing HVPE GaN. The apparent different optical properties for both faces has been found. The $D_2^0$X emission of the Ga-face is dominant among the near-band-edge emission at temperatures up to 80 K while that of N-face dominates below 45 K. Moreover, the temperature-related ratio of the peak intensity of $D_2^0$X to $X_A^{n=1}$ for the Ga-face is found to be about 2.8 times larger than that of the N-face. CL images further reveal that the average dislocation density of the N-face is about 2.2 times larger than that of the Ga-face. The result is in coincidence with that obtained from PL spectra. Such features are ascribed to two main factors: different activation energies for the $D_2^0$X of the Ga-/N-faces and a larger density of nonradiative recombination centers in the N-face. These findings could provide a comparative and effective method to extract useful parameters for bulk GaN, which is important for the later design of GaN-based and related devices.