3D Investigation of InGaN Nanodisks in GaN Nanowires

Abstract

Nanoscaled structures like nanowires (NWs) can influence device characteristics (e.g. higher internal efficiency [1]), making them very suitable for the application in optoelectronic devices. Complex structures like InGaN nanodisks (NDs) embedded in GaN NWs can for example be used as active regions of tunable color and white light LEDs [2,3]. The GaN NWs investigated in this work have been grown via plasma‐assisted molecular‐beam epitaxy on n‐type Si (111) substrates and contain 10x InGaN NDs. The growth direction was [000‐1] and the NWs exhibit a hexagonal base with (1‐100) planes (m‐plane) forming the side facets. The TEM analysis of InGaN/GaN NWs grown under comparable growth parameters is reported in [4]. There it was shown that the ~4nm thick InGaN NDs exhibit a truncated pyramidal shape consisting of a (0001) central facet that is delimited by declining sixfold {10‐1 l } facets, where l can be ‐1, ‐2 or ‐3. In addition to this declined facets a continuation of the (0001) central facet towards the m‐plane sidewalls of the NWs could be observed but is not described in the published STEM images [4]. Since these images contain projected information they do not deliver knowledge on the real shape of the NDs. To obtain a deeper insight into the three dimensional geometry of the embedded InGaN NDs electron tomography is the method of choice. Conventional sample preparation (spread NWs over a carbon film lying on a TEM grid) just allows tilting in a range of approximately +/‐70° leading to a strong missing wedge effect [5]. The concomitant reduction of the resolution makes a meaningful reconstruction of the InGaN NDs impossible. To overcome this problem the sample needs to be prepared in such a way that a sample tilt of +/‐90° is permitted. For this procedure we used a SEM (JIB 4601F, JEOL) with an integrated manipulator needle (Kleindiek). First of all a conventional FIB‐lift‐out‐grid (Pelco, Ted Pella) with four narrow posts was trimmed with a scalpel in such a way, that the grid width was reduced from 3mm to less than 1.5mm and that just one post was left. On top of this post an electron beam induced, turret shaped tungsten structure was deposited in the SEM to create an exposed position on which the NW could be attached without any risk of shadowing effects during the tilt series in TEM. Using the manipulator needle a few NWs have been detached from the Si substrate and transferred to the FIB‐lift‐out‐grid (cf Figure 1a). The NW that is most suitable oriented was brought closer to the top of the tungsten deposition. Since the attractive force between needle‐NW and tungsten deposition‐NW, respectively, are strong a deposition for connection is not necessarily required (cf Figure 1b). Electron tomography measurements were performed in STEM mode (JEM 2200FS, JEOL using a model 2030, Fischione tomography holder) with a tilt range of ‐90° until +82° (the tilt angle limitation is due to a restriction of the TEM stage and not related to the sample geometry) and a tilt step of 2°. For reconstruction of the data the software IMOD [6] was used. Figure 2 shows a section through the middle of a NW running parallel to the (11‐20) plane (a‐plane). Within this image two features can be observed. First the faceting of the InGaN NDs with an increased steepness of the inclination angle of the side facets for higher lying NDs can be seen. The inclination angle fits well to the above mentioned {10‐1 l } planes. Second the yellow circle marks a region where the afore‐noted split of NDs appears. This is particularly interesting since this structure could be easily attributed to projection artifacts in conventional STEM images. This example shows, that selecting a sample geometry which allows a tilt angle range as high as possible, is essential for obtaining the required resolution.