The impact of nanoporous SiNx interlayer growth position on high-quality GaN epitaxial films

The impact of nanoporous SiNx interlayer growth position on high-quality GaN epitaxial film was elucidated from the behavior of dislocations. The best quality GaN film was achieved when a nanoporous SiNx interlayer was grown on a rough layer, with the high-resolution X-ray diffraction rocking curve full width at half maximum for (1102 ) reflection decreasing to 223 arcs, and the total dislocation density reduced to less than 1.0×10 cm. GaN films were grown on sapphire substrates by metal organic chemical vapor deposition. The quality of these films was investigated with high-resolution X-ray diffraction, atomic force microscopy, and cross-sectional transmission electron microscopy. A preference for the formation of half-loops to reduce threading dislocations was observed when an SiNx interlayer was grown on a rough layer. A growth mechanism is proposed to explain this preference.

GaN and related compounds are playing an indispensable role world-wide in optoelectronic device applications, such as high-brightness light-emitting diodes (LEDs), laser diodes (LDs) and detectors, because of their large band-gaps, high thermal conductivity and chemical inertness [1,2]. However, the large density of threading dislocations (TDs) in GaN is still one of the main challenges hindering the development of nitride-based optoelectronic devices, especially for shortwavelength light-emitting devices, such as LDs and ultraviolet (UV) LEDs [3]. Generally, heteroepitaxial growth of GaN films is based on sapphire, and large dislocations are inevitably induced by the large lattice mismatch between GaN and sapphire [4]. Many efforts have been made to improve the quality of GaN grown on these substrates. Patterned template technologies such as epitaxial lateral overgrowth (ELOG), pendeo-epitaxy (PE) and lateral overgrowth from trenches (LOFT) have proved effective in dis-location reduction. These methods can reduce the dislocation density to the order of magnitude 10 5 cm −2 [5]. Unfortunately, contaminants are inevitably introduced into the epitaxial film by ex-situ lithography and etching during the entire film preparation process. Moreover, thicker GaN films are required to form well-defined smooth surfaces, which results in a more time-consuming process.
In-situ nanoporous SiN x interlayer deposition is a much simpler and cleaner way to tailor the patterned template for GaN epitaxy. This method has been verified as a convenient and efficient way to reduce TDs to 10 7 -10 8 cm −2 in GaN grown on sapphire substrates [6][7][8][9][10][11][12][13][14][15]. The method has been used to improve the performance of devices such as blue and green LEDs, UV-LEDs, Schottky barrier photo-detectors, and metal-semiconductor-metal photo-detectors [16][17][18][19][20]. However, the behavior of dislocations, including their formation, termination, threading up and bending, occur at different growth stages of GaN films. If the SiN x interlayer was inserted at different growth stages, the quality of the GaN film would be changed because of the different effects of the SiN x interlayer on the behavior of dislocations. However, there have been few reports on the effect of SiN x interlayer position on the crystal quality of GaN film. In the present study the SiN x interlayer was inserted into different stages of the GaN growth process by metal organic chemical vapor deposition (MOCVD), and the influence of different SiN x interlayer positions on the quality of GaN was systematically investigated. Finally, GaN films with improved crystal quality and morphology were obtained when SiN x interlayer was grown on a rough layer. The rocking curve full width at half maximum (FWHM) for the asymmetrical plane (1 1 02) reflection was 223 arcs and the dislocation density of the epitaxial film decreased to less than 1×10 8 cm −2 .

Materials and methods
The samples were grown on c-plane sapphire substrates using a modified two-step method, during which a Veeco Pioneer P125 vertical flow metal organic chemical vapor deposition (MOCVD) reactor was employed. Trimethylgallium (TMG) and ammonia (NH 3 ) were used as the precursors of Ga and N. Mixtures of H 2 and N 2 were used as carrier gas during the whole GaN film preparation process.
The sapphire substrate was first exposed to a hydrogen flow at 1060°C. A nominal 25 nm thick GaN nucleation layer was then deposited at 490°C, and annealed at 980°C for about 1 min. After that, a GaN layer of enhanced 3Dgrowth mode was grown at the same temperature, as the socalled rough layer for which the growth conditions were controlled to grow the annealed nuclei up to GaN islands that just fully covered the surface. Finally, a 2-μm hightemperature GaN layer was then grown at 1050°C to obtain a coalesced surface and high-quality epitaxial film. An in-situ SiN x layer, grown for 6 min using 200 ppm silane and NH 3 under the same conditions, was inserted at various positions during GaN deposition. In samples A, B and C, SiN x was grown on annealed sapphire substrates, on the rough layer, and on the high-temperature GaN layer with well defined surface, respectively.
The crystal qualities of these three samples were characterized by high-resolution X-ray diffraction (HRXRD) with a Bede D1 system equipped with a four-bounce channel-cut Si (220) monochromator, delivering a pure CuKα 1 line of wavelength 0.154056 nm. The surface morphology was imaged by atomic force microscopy (AFM) with a CSPM 4000 system. Dislocations were observed with a JEOL 2010 transmission electron microscope (TEM) operating at 200 kV: the specimens were prepared by mechanical polishing followed by ion milling.

Results and discussion
The FWHMs of HRXRD rocking curves were determined.
The symmetric (0002) and asymmetric (1 102 ) reflection results for the three samples are listed in Table 1. A GaN film without SiN x interlayer was also grown using the same conditions and characterized. The FWHMs for (0002) and (1 102 ) of this film were 241 and 348 arcs respectively. Referring to the FWHMs in Table 1, we can conclude that the quality of the GaN film was improved by inserting an SiN x interlayer. The FWHM for (1 102 ) plane of sample B was the smallest of the three samples. The 223 arcs for sample B is one of the best results that has been reported for the in-situ SiN x interlayer deposition method, noting that the thickness of our sample was only 2.5 μm [12]. However, it is shown in the Table 1 that the FWHMs for the (0002) reflections of the three samples were almost the same, indicating that the influence of the SiN x interlayer position on screw dislocation density is negligible. These (0002) and (1 102 ) rocking curves are related to the mosaic structure of GaN. The mosaic structures, including tilts and twists, are always accommodated by screw and edge dislocations, respectively. Consequently, the dislocation densities of the samples could be accounted for by these FWHMs for (0002) and (1 102 ) reflection results [21]. However, dislocation densities obtained from HRXRD measurement are always overestimated. That is due mainly to the fact that mixed dislocations which consist of the screw and edge type together are calculated repeatedly, as well as the strain influence of the tested materials and the equipment broadening from X-ray dispersion. In any case, the dislocation densities of all three samples were reduced by SiN x interlayers, especially in the case of sample B.
To examine the surface morphology of GaN films, AFM measurements were conducted with the results shown in Figure 1(a)-(c), from which the terrace morphology on the GaN surface is clearly visible. The surface features are closely related to the presence of microstructural defects in materials [22]. The root-mean-square (RMS) roughness values in an area of 2 μm×2 μm for the three samples were 0.150, 0.095 and 0.195 nm, respectively. Thus, the AFM for sample B exhibited the best quality and morphology. To characterize the dislocation density more directly, etching was performed after growth. The surfaces of the three samples were etched in a mixture of H 3 PO 4 and H 2 SO 4 in mole ratio 3 : 1 at 250°C for 10 min. The morphology after etching was  imaged by AFM as shown in Figure 1(d)-(f). The etched pit densities of the three samples are shown in Table 1: sample B had the smallest number of etched pits. The pits have been proved to originate from screw threading dislocations or mixed threading dislocations with screw components, while few pure edge dislocations cause pits [23]. Since the density of mixed dislocations is usually an order of magnitude larger than that of the screw type in GaN film, most of the pits are mixed type dislocations. The etched pit density of sample B was 5.5×10 7 cm −2 and the total dislocation density of sample B was less than 1×10 8 cm −2 . Both AFM and HRXRD measurement revealed that the quality of sample B was superior to the quality of the other two samples, with the lowest threading dislocation density. To understand the dislocation reduction mechanism further, bright-field cross-sectional TEM analysis of samples B and C was performed. Sample A was not included because of its large threading dislocation density. Cross-sectional specimens were observed along the [1 100 ] direction. The images of dislocation evolution are shown in Figure 2, with a composite of some micrographs so that a panorama was obtained that can be studied systematically. The SiN x interlayer appears as a faint line in all images. It can be observed that most of the dislocations are blocked from threading up. With the two-beam extinction: g · b=0 where g is the direction of diffraction and b is the Burgers vector, the dislocations with different Burgers vectors could be out of contrast when g changed. So when g=[0002 Figure 2(a) and (b), respectively, and for sample C in Figure 2(c) and (d), respectively. Comparing the images with different g, some screw dislocations that are not seen in Figure 2(b) and (d) are present on the top of images as shown in Figure 2(a) and (c). Hence, it is easy to reach the conclusion that edge dislocations are more effectively stopped from threading up by the SiN x interlayer than are screw dislocations. This is in agreement with the results of HRXRD measurement which indicate that the (0002) plane FWHMs are almost the same, while the (1120 ) plane FWHMs are different. Analyzing the two specimens reveals that in sample B, except for dislocations blocked by SiN x interlayer, many dislocations were bent towards the in-plane direction during growth. Subsequently some of them with the opposite Burgers vector were annihilated by the formation of half-loops. The result was creation of dislocation half-loops with increased probability of decreasing the threading dislocation density. These dislocation half-loops are not seen in sample C. Consequently, the effect of the SiN x interlayer on decreasing the TD density is much more obvious in sample B than in sample C.
It is well known that a dislocation is terminated at a free surface or when it incorporates with other dislocations that have partly opposite Burgers vectors. Both of these mechanisms led to the decreased dislocation density in sample B, while for sample C only dislocations blocked from threading up were clearly seen. This is demonstrated by the sketch maps of the growing process for samples B and C in Figure  3(a) and (c), and the in-situ monitored reflectivity spectra in Figure 3(b) and (d), respectively. For sample B many hexangular GaN crystalline grains formed on the substrate with (110h ) plane during the rough layer process. The SiN x interlayer tends to deposit at the bottom of the rough layer among the grain boundaries to minimize the surface energy and interface energy, and only cusps of grains were not covered with a SiN x amorphous layer. The SiN x layer prevented the formation of dislocations at the grain boundaries, and blocked some dislocations in the grains from threading up. After SiN x deposition, the cusps of 3D crystalline grain are overgrown laterally in conditions for GaN high-temperature growth, and the horizontal growth rate is faster as demonstrated in Figure 3(a). The shapes of the overgrown grains favor dislocation bending to the horizontal direction because of the inclined free plane, especially the (1 101 ) plane formed during the growth process. Pairs of horizontal dislocations with partly opposite Burgers vectors will in-corporate. In addition, the overgrown grains extend the time of grain coalescence because of their initially 3D shapes, which minimize the formation of dislocations at the coalescence boundary. However, this growth mode is not appropriate for sample C, for which, shortly after the high temperature growth step begins, the surface of GaN film is atomically flat. Hence the SiN x interlayer nucleates around the outcrops of dislocations because of the stress field around the dislocations, then grows mainly with 2D growth mode covering most of the surface. The SiN x interlayer stops many dislocations threading up, as shown in Figure 3(c). However, after SiN x interlayer deposition, conditions for GaN high-temperature growth increase the lateral growth rate. The film coalesces promptly because of the 2D-shaped grains, which increases the probability of forming dislocations at the coalescence boundary. To estimate the coalescence rate, the recovery time was defined using in-situ monitored reflectivity from the minimal reflectivity to the reflectivity of the sapphire, as shown in Figure 3(b) and (d), respectively. It is apparent that the recovery time of sample B was longer than sample C, indicating that the grains coalesced more slowly in sample B. All of the reasons mentioned above made the dislocation density of sample B lower than that of sample C. It can also be proved that the recovery time is shorter for sample A than for sample B. Moreover, in sample A no dislocations are blocked during the growth process because the SiN x interlayer is deposited directly on sapphire.
According to the above analysis, it can be deduced that growing an SiN x interlayer on a rough layer formed initially more effectively decreases dislocation density than an interlayer grown on a flat surface. This is an important factor for growing high quality GaN epitaxial films, because the effect of the SiN x interlayer on decreasing the dislocation density can be optimised provided that the rough layer is accurately controlled. Since the method is easy in practice, optimization of the SiN x position on a rough layer could also be used with other Group III-nitride materials to decrease dislocations.

Conclusion
In summary, we have studied the effect of an SiN x interlayer on the quality of GaN films, by inserting it at different growth positions during the GaN epitaxial process. The best results are obtained when SiN x is grown on a rough layer. The preferential insertion of SiN x on the rough layer not only blocks dislocations formed initially from threading up, but also promotes the 3D growth mode of GaN, delays the recovery time from 3D to 2D growth mode, and minimizes the probability of dislocation formation at the incorporation boundary. The dislocation density of GaN epitaxial film is successfully minimized when the growth position of SiN x interlayer is optimized on the rough layer.