1. Introduction

The tunability of the fundamental gap of indium gallium nitride (InGaN) across the full visible spectrum has led to the development of a variety of optoelectronic devices that use this alloy, including blue, green, red, and white light-emitting diodes, blue and green laser diodes, and solar cells [1–4]. Recent works have also shown the potential of using the InGaN material as substrates for reducing the charge separation effect especially for high performance laser diode emitting in the range of yellow to green spectral regions [5–7]. Still, substantial work is required to better understand the link between material quality and the optical and electronic properties of InGaN epilayers, in order to improve the performances of these kind of devices. A particular aspect of InGaN alloys is the presence of compositional fluctuations, both predicted by theory and observed experimentally [8]. Ponce et al. [9] and Bertram et al. [10], in particular, observed macroscopic inclusions in SEM micrographs of InGaN epilayers thicker than 100 nm and with compositions ranging between 10 % and 32 %. The inclusions were systematically found to have a higher indium concentration and to luminesce at a different wavelength than the surrounding InGaN matrix they were embedded in. Inclusions with similar properties have also been reported for InGaN/GaN multiple quantum well (MQW) stacks [11–13]. While these inclusions have been thoroughly described in the above-mentionned papers, an explanation to the mechanism behind their formation, the origin of their difference with the surrounding matrix, and their impact on the growth of InGaN epilayers has yet to be proposed.

The present paper focuses on the InGaN epilayers that present similar inclusions. A combined investigation of the morphological, structural, and optical properties of these inclusions using scanning electron microscopy (SEM), atomic force microscopy (AFM), high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM), bright field STEM (BF-STEM), energy-dispersive x-ray spectroscopy (EDX), and cathodoluminescnence (CL) is proposed. The inclusions are shown to systematically surround threading dislocations already present in the GaN buffer. Two distinct types of inclusion are clearly identified and are shown to correspond to distinct steps in a localized transition from two-dimensional to three-dimensional growth. A mechanism that describes these localized transitions is proposed.

2. Experiment

The InGaN epilayers discussed in the present study were grown in a T-shape reactor by MOVPE [14]. Nitrogen (N₂) was used as the carrier gas and trimethylgallium (TMGa), trimethylindium (TMIn), and ammonia (NH₃) were used as precursors to elementary gallium, indium, and nitrogen. All epilayers were grown at 800 °C, with a reactor pressure of 100 Torr. The ratio of TMIn to the sum of TMIn and TMGa in the vapour phase, TMIn/III, was 33.3 %, and the V/III ratio was 6500. The epilayers were 77 nm thick and were grown on 3.5 μm thick GaN/sapphire templates. High-resolution x-ray diffraction was used to determine the indium content and strain state of the InGaN epilayers. These were obtained by combining symmetric (00.2) plane ω−2θ scans and reciprocal space maps (RSMs) of the asymmetric (11.4) plane. The surface morphology of the epilayers was investigated using SEM and AFM. Sections of the samples were prepared for observation in a STEM using focused ion beam (FIB) etching. To preserve the sample surface during the FIB etching process, the samples were coated with a 50 nm thick carbon layer, followed by a 150 nm thick layer of Si₃N₄. The sections were oriented along the 〈112̄0〉 zone axis. HAADF-STEM, BF-STEM, and EDX experiments were then performed in a dedicated, aberration-corrected JEOL 2200FS microscope working at 200 kV with a probe current of 150 pA and a probe size of 0.12 nm at the FWHM.

The optical emission properties were investigated by CL which provide the possibility to analyze microscopic areas of the layer surface and to investigate in-depth the samples by adjusting the electron beam energy from 0.5 to 25 keV. The room temperature CL measurements were performed in a digital scanning electron microscope (SEM) (Zeiss supra^TM 55VP). The CL emission is detected via a parabolic mirror collector and analyzed by a spectrometer (iHR320) with a focal length of 320 mm using 1200 grooves/mm grating and a spectral resolution of 0.06 nm. The signal is then registered by a HORIBA JOBIN YVON Instr., 1024 × 256 Symphony charge-coupled Liquid N₂-cooled with a CCD camera device. Hyperspectral mapping was performed by synchronizing the scanning of the electron beam to the spectrometer. A complete CL spectrum is obtained for each point on the sample and two-dimensional maps is reconstructed from specific spectral features.

3. Results and discussion

The InGaN layer is shown to be strained on GaN, as InGaN diffraction spot is located close along the vertical line corresponding to pseudomorphic InGaN on GaN (Fig. 1a). In addition, the (00.2) XRD ω − 2θ scan revealed several diffraction satellites which fit well with 11.6 % indium composition and 77 nm InGaN layer thickness as shown in Fig. 1b. SEM and AFM surface images revealed a 2D morphology, as shown in Figs. 2a and 2b. The root mean square (RMS) roughness obtained for a 5 × 5 μm² surface is of around 2 nm. Despite the good macroscopic properties of the alloy mentioned above, one can clearly observe two distinct microscopic surface defects features which mostly form close loop with or without hillocks inside (Fig. 2a). For clarity, both types of defects are labeled hereafter as inclusion#1 (Ic1) and inclusion#2 (Ic2), respectively. The AFM profiles taken along a line across the two different types of inclusion reveal that Ic2 are very rough as compared to Ic1 (Fig. 2b). Such surface morphology has already been reported by Bertram et al. [10] for single InGaN films and by Bruckbauer et al. and Kumar et al. [11, 12] for InGaN/GaN MQW structures.

 

Fig. 1 (a) (11.4) reciprocal space map and (b) (00.2) XRD ω − 2θ scan and simulation fitted by X’Pert Epitaxy software.

Download Full Size | PDF

 

Fig. 2 (a) Typical SEM and (b) 5 × 5 μm² AFM surface images. The solid lines on the 2D AFM image correspond to the AFM profile of Ic1 and Ic2 reported in the inset.

Download Full Size | PDF

The HAADF-STEM and BF-STEM cross sectional images revealed that inclusions are systematically centered around threading dislocation terminations, as shown in Fig. 3. The average diameter of these 3D domains is around 400 nm in agreement with SEM and AFM measurements.

 

Fig. 3 (a) Cross-section HAADF-STEM and (b) BF-STEM images.

Download Full Size | PDF

HAADF-STEM and EDX were performed to investigate the composition of the different regions of the sample. Between inclusions, the layer is homogeneous with an indium content of around 12 % (Fig. 4). In Fig. 5a, the rough 3D domains (Ic2) show HAADF contrast variations indicating of the indium fluctuation content. EDX measurements, taken for several lines along the growth direction reveal two regions of different indium content (Fig. 5b). For the first 30 nm beyond the InGaN/GaN interface (InGaN#1) the composition is constant and around 12 %. Close to the surface, the indium incorporation increases especially for surface pyramidal features (L5). As expected, directions other than 〈0002〉 tend to incorporate more indium. In addition, a slight depletion of indium in the surrounding of V-defect (L6, L7 and L8) is observed in agreement with the results of Ponce et al. [9].

 

Fig. 4 (a) Cross-section HAADF-STEM image between inclusions and (b) corresponding indium content measured by EDX through line scan L1 and L2.

Download Full Size | PDF

 

Fig. 5 (a) Cross-section HAADF-STEM image for an area containing 3D domains (Ic2) and (b) corresponding indium content measured by EDX through lines scan L3 to L9.

Download Full Size | PDF

Twenty local CL spectra were taken along a line across each type of inclusion to study their luminescence properties. For clarity, only the most representative ones are presented in Figs. 6a and 6b. The main characteristic of Ic1 luminescence is a decrease in intensity and no shift of the CL InGaN peak from the edge to the middle of such inclusion (Fig. 6a). The latter observation indicates no significant variation of the strain and indium composition in agreement with the results of Kumar et al. [12]. On the contrary, Bruckbauer et al. [11] have observed emission inside the inclusion to be more intense and redshifted as compared to the InGaN matrix. In Fig. 6b, CL measurements taken from the outside to the middle of Ic2 show a drastic change in luminescence behavior: the main InGaN CL peak wavelength decrease in intensity and is redshifted by 15 nm (108 meV). On the other hand, additional InGaN peaks are observed at higher wavelength for spectra taken at the edge and the middle of Ic2. The shift or the splitting of the CL peak wavelength can be explained in terms of strain relaxation and/or indium composition fluctuation. We attributed Ic2 to microscopic relaxed regions, as local 3D growth is observed. In addition, the large splitting of the 432 nm and 481 nm CL peaks with the emission wavelength at around 410 nm cannot be explained only by strain relaxation but by an increase of the In content.

 

Fig. 6 Most representatives local CL spectra taken for twenty points along a line across the different surface defects features: (a) Ic1 and (b) Ic2 (the inset shows the line scan and the location of the four spots).

Download Full Size | PDF

Refering to CL spectra shown in Fig. 6b, we have recorded CL hyperspectral mapping of Ic2. Green, blue, and red colors reported in Fig. 7 correspond to normalized luminescence intensity in the range of 392–424 nm, 426–454 nm and 457–507 nm, respectively. The luminescence behavior clearly confirms higher indium content inside Ic2 as compared to the InGaN matrix, in agreement with EDX measurements. Furthermore, series of mapping reveal that the indium composition changes from one inclusion to the other and seems to increase with their in-plane size.

 

Fig. 7 CL hyperspectral mapping taken for 10 x 10 spots with an integration time of 2s/point and an acceleration voltage of 4 keV. Green, blue, and red colors correspond to normalized luminescence intensity for wavelength λ = 407 nm, λ = 432 nm and λ = 477 nm, with a peak bandwidth of 32 nm, 28 nm and 50 nm, respectively.

Download Full Size | PDF

Depth-resolved CL measurements were also performed for Ic2, as shown in Fig. 8. The electron energy beam was varied from 4 keV to 5 keV which corresponds, according to the Monte Carlo simulation [15], to calculated sample depths of maximum energy loss of 23 nm and 33 nm, respectively. For the surrounding planar region, the InGaN CL peak position (410 nm) does not depend on the acceleration voltage (Fig. 8a). At the Ic2 position, the 410 nm CL peak is only observed as the electron beam energy increases to 5 keV (Fig. 8b). Thus, Ic2 originates from the InGaN top surface layer and is attributed to the presence of In-rich areas embedded in a lower InGaN composition matrix.

 

Fig. 8 Most representative local CL spectra taken at 4 keV and 5 keV electron beam energy: (a) outside and (b) inside of Ic2.

Download Full Size | PDF

From our observation the formation of Ic2 can be explained as follows. After the first 30 nm, spatial fluctuation of the composition is observed and local In-rich 3D domains are shown to emerge systematically around threading dislocations terminations. Indeed, indium adatoms are prone to accumulate around V-defects and additional facets appear. Directions other than 〈0002〉 enhance the growth, leading to In-rich pyramidal features. A slight depletion of the indium composition in the neighborhood of Ic2 was also observed. We assume Ic1 to be the first step in the formation of Ic2. Finally, we believe that the In-rich local 3D regions contribute to the transition from two-dimensional to three-dimensional growth observed in thicker epilayers [16–18].

4. Conclusion

In the present work, we have reported on the luminescence properties of InGaN surface defects features. Especially, the local and spatial variations of indium composition and microstructure through inclusions were investigated by STEM, EDX and cathodoluminescence in scanning and spot modes. Between inclusions and within the first 30 nm beyond the InGaN/GaN interface, the film is shown to be homogeneous both in strain and composition. As the growth proceeds, 3D pyramidal In-rich domains emerge around threading dislocations terminations leading to a slight depletion on indium in their neighborhood. We believe inclusions to correspond to distinct steps in a localized transition from two-dimensional to three-dimensional growth.