Search

Two methods for GaN growth on sapphire by metal-organic vapour phase epitaxy are discussed. The first involves only two-dimensional (2D) growth, and results in a high dislocation density, but also a high electrical resistivity. The second involves initial growth of three-dimensional (3D) islands employing a low V:III ratio, followed by island coalescence at a high V:III ratio. It is often assumed that threading dislocations (TDs) form via the coalescence of 3D islands, but detailed atomic force microscopy studies on partially coalesced samples find no evidence of an increased TD density at coalescence boundaries, suggesting that other possible origins for TDs should be considered. The 3D–2D growth method allows TD densities as low as 1.1 × 108 cm−2 to be achieved, but unlike the 2D growth samples these layers are not highly resistive. Scanning capacitance microscopy is used to demonstrate the presence of an unintentionally doped layer close to the GaN/sapphire interface. To simultaneously achieve a reduced TD density compared to 2D growth samples and a high resistivity, a high temperature AlN buffer layer may be employed.

High-resolution transmission electron microscope (HRTEM) lattice fringe images and lattice parameter maps are used to reveal the rapid modification of InGaN quantum wells by the electron beam in a TEM. Images acquired within seconds of first irradiating a region of quantum well do not exhibit the strong nanometre-scale strain contrast which has been reported to signify the presence of very indium-rich regions in InGaN quantum wells. However, after a very short period of irradiation by a relatively low electron beam current density, images of the specimen could be interpreted as indicating the presence of these indium “clusters”. The beam damage is shown to occur for the InGaN quantum wells grown in our lab as well as those in a commercial blue light emitting diode (LED) and in TEM specimens prepared only using mechanical polishing. Possible mechanisms for the beam damage are discussed and we make suggestions as to what may cause exciton localisation in quantum wells that do not contain gross composition fluctuations.

Semi-polar (1122) GaN films were grown on (1100) m-plane sapphire substrates. Growth demonstrated surface striations aligned perpendicular to the in-plane GaN m-axis. SiNx interlayers were incorporated into the as-grown films with the purpose of decreasing the density of defects in the material. Inclusion of interlayers increased the characteristic length of surface striations and feature size. X-ray rocking curves widths are shown to be correlated to specific threading dislocation geometry. Skew-symmetric omega scan peak broadening suggests a decrease in the proportion of screw-type dislocations to edge-type dislocations with increasing number of interlayers.

Fast-turnaround, accurate methods for the assessment of threading dislocation densities in GaN are an essential research tool. Here, we present an in-situ surface treatment for use in MOVPE (metal-organic vapour phase epitaxy) growth, in which GaN is exposed to a SiH4 flux at 860 °C in the presence of NH3. Subsequent characterisation by atomic force microscopy shows that the treatment is effective in increasing edge and mixed/screw dislocation pit sizes on both n- and p-type material, and on partially coalesced GaN layers.

Gross well-width fluctuations have been observed in single InGaN/GaN quantum wells (QWs) grown by two methods: (i) a “two-temperature” method, where the GaN barrier has been grown at a higher temperature than the InGaN layer, (ii) a single-temperature method, where the InGaN was annealed in a flow of NH3/N2, prior to capping. InGaN epilayers analogous to these QW structures have been grown and characterized by atomic force microscopy (AFM) and transmission electron microscopy (TEM). AFM reveals a network of interlinking InGaN strips. Compositional analysis across the InGaN strips has shown that the centres of these strips to be indium-rich compared to the strip-edges. Furthermore, we have correlated the position of threading dislocations (TDs) in relation to these InGaN strips and determined that 90 % of the TDs terminate at the channels between the InGaN strips or near the edges of the strips. We propose that confinement of charge carriers occurs in the indium-rich regions at the centre of the InGaN strips, preventing non-radiative recombination losses at dislocation sites.

Misfit dislocations (MDs) have been observed using transmission electron microscopy (TEM) in InGaN/GaN quantum well (MQW) structures grown under different metal-organic vapour phase epitaxy (MOVPE) regimes and with In-contents equal to or higher than 20%. These dislocations are even observed in a single quantum well 3 nm thick with an In-content of 22%. Conversely, no MDs were observed in QW structures with an In-content of 16%. The presence of MDs in the QW stack leads to strain relaxation which has been confirmed in the indium-rich structures by high resolution X-ray diffraction (HRXRD).

A series of Al0.47Ga0.53N/GaN heterostructures with different AlN interlayer thicknesses ranging from 1nm to 50nm has been examined. It was found that when the interlayer thickness is greater than ∼5nm, it becomes possible to grow 250nm of Al0.47Ga0.53N without cracking. The interlayers are then believed to be sufficiently relaxed to place the AlGaN under compressive strain. The mechanisms for this relaxation have been studied using high angle annular dark field (HAADF) imaging, conventional transmission electron microscopy (TEM), energy-filtered TEM (EFTEM) and electron energy loss spectroscopy (EELS). It is found that relaxation takes place through both the small-scale cracking of the interlayer and the generation of misfit dislocations at the GaN/AlN interface. EELS and EFTEM have been used to probe the Al and Ga content of both the material filling the interlayer cracks, and the interlayer itself. This chemical analysis suggests Ga-rich AlGaN areas inside the interlayer cracks and also significant compositional variations in defect-free interlayer regions. It is observed that relaxation by the generation of misfit dislocations results in an increase in the threading dislocation density of the AlGaN layer, in part due to the bending up of misfit dislocations at crack walls.

The strain in Al0.80Ga0.20N/GaN distributed Bragg reflectors (DBRs) grown by metalorganic vapour phase epitaxy (MOVPE) on Si(111) was studied using high resolution X-ray diffraction (XRD). Previous studies have shown that a 50nm Al0.80Ga0.20N layer induced a compressive strain in Al0.12Ga0.88N capping layers and prevented crack formation. A seven period Al0.80Ga0.20N/GaN DBR was grown, but this was found to be cracked at room temperature, despite compression in the GaN layers. This problem was solved by growing an identical structure with the addition of a 650nm GaN cap, and due to the compression in this layer, the structure was crack-free.

The reduction of threading dislocations (TDs) in GaN is very important as TDs act as non-radiative recombination centres and generally reduce the luminescence efficiency of GaN-based optoelectronic devices, particularly UV emitting devices. In order to reduce the density of TDs, we have used an in-situ ELO technique by depositing a non-uniform SiNx mask (as revealed by STEM-EDX) using simultaneous flow of silane and ammonia gas, either on a sapphire substrate or on a GaN pseudo-substrate. The effect of several growth parameters, such as V/III ratio, duration of Si/N treatment, pressure and temperature, on the morphology, structural quality and reduction mechanisms of TDs in GaN have been investigated and the effects of the first two parameters are reported here. To start the epilayer growth with a low V/III ratio gives rise to predominantly 3D growth, forming islands with multiple inclined facets and a vanishing top (0001) facet. A high V/III ratio enhances the lateral growth, and islands with inclined {-2112} facets together with wide flat top (0001) facets are formed. The duration of the Si/N treatment and the low V/III ratio growth regime are optimised in order to reduce the TD density down to 8×107 cm−2, as measured from plan-view TEM images. It is observed that TDs bend through 90° at inclined side facets. The role of atomic ledges moving across inclined facets which are responsible for bending the line direction of the TDs is briefly discussed.

We have studied the distribution of indium in InxGa1-xN single quantum wells (SQWs) with compositions of In0.18Ga0.82N and In0.22Ga0.78N. Based on lattice parameter maps extracted from HRTEM lattice fringe images, we do not find evidence for strong nanometre-scale fluctuations in the composition of the wells (indium “clusters”). Z-contrast (high-angle annular dark field) scanning TEM (STEM) and X-ray reflectivity (XRR) results both show that the width of the interface at the top of the quantum well is slightly greater than that at the bottom, which is quite abrupt. The quantum wells exhibit bright photoluminescence so these results suggest that the conventional explanation for the high efficiency of light emission from InGaN LEDs may not apply in these structures.

In the Claus process hydrogen sulfide reacts to elemental sulfur. Because the Claus reaction is thermodynamically limited, sulfur compounds are still present in Claus tailgas. To avoid air pollution, the tailgas has to be treated.