GaN nanowires have been grown by molecular beam epitaxy either catalyst-free or catalyst-induced by means of Ni seeds. Under identical growth conditions of temperature and V/III ratio, both types of GaN nanowires are of wurtzite structure elongated in the Ga-polar direction and are constricted by M-plane facets. However, the catalyst-induced nanowires contain many more basal-plane stacking faults and their photoluminescence is weaker. These differences can be explained as effects of the catalyst Ni seeds. Open image in new window

On Si(1 1 1) and Si(0 0 1), GaN nanowires (NWs) form in a self-induced way without the need for any external material. On sapphire, NW growth is induced by Ni collectors. Both types of NWs exhibit the wurtzite crystal structure and grow in the Ga-polar C-direction perpendicular to the substrate. The NW sidewalls are M-plane facets, although on the Ni-induced NWs also A-plane segments form, if the growth temperature is low. Both self-induced and collector-induced NWs are free of strain and epitaxially aligned to the substrate, but in particular the former show a significant spread in tilt and twist caused by a mostly amorphous interfacial layer of Si-N. The self-induced NWs are virtually free of extended defects, but the collector-induced NWs contain many stacking faults. The photoluminescence of the former is significantly brighter and sharper. The spectra of single, dispersed, self-induced NWs contain extremely sharp excitonic lines. Significant emission is caused by excitons bound to donors close to the surface whose binding energy is reduced compared to the bulk value. In comparison, both the microstructure and optical properties of the self-induced NWs are superior. The limited material quality of the collector-induced NWs can be explained by detrimental effects of the collector.

The epitaxial interface formed by a thin buffer layer of AlN deposited by molecular beam epitaxy on (0001) ${\mathrm{Al}}_2{\mathrm{O}}_3$ is investigated using electron microscopy techniques. Plan-view observations display a two-dimensional translational moiré pattern resulting from the difference in lattice parameters between the two crystals. The effective misfit of $-0.1091$ suggests the presence of a network of $60°$ misfit interfacial dislocations. These are, most of the time, introduced every 8 atomic planes of the AlN lattice or 9 atomic planes of the ${\mathrm{Al}}_2{\mathrm{O}}_3$ lattice, which is directly verified by cross-section high-resolution electron microscopy (HREM). The density of threading dislocations terminating at the buffer/substrate interface is in agreement with the density of threading dislocations near the surface of the GaN film. This in conjunction with terminating fringes observed in the moiré patterns provides strong evidence that threading dislocations are connected with the interfacial misfit dislocations. Plan-view HREM images reveal that a threading dislocation is directly related to two extra prismatic AlN half-planes, the missing parts of which are associated with two misfit dislocations in the interfacial network. The Burgers vector of the threading dislocation equals the sum of the Burgers vectors of the two misfit dislocations and therefore is of the a type. Although misfit dislocations relax the larger fraction of the misfit strain, the AlN buffer layer is still under a compressive residual strain $(-8.2\times {10}^{-3}),$ since the effective misfit is smaller than the natural misfit between the two lattices.