Tomographic reconstruction and sub-surface imaging of porous nitride semiconductors

Abstract

The research presented here aimed to develop new methods for the structural characterisation of porous nitride semiconductors and apply those methods to further our understanding of these materials. I investigated nanostructured nitrides through porous GaN distributed Bragg reflectors (DBRs), highly-reflective and wavelength-selective mirrors for optoelectronic device applications such as next-generation micro light emitting diode (LED) displays. These photonic structures are fabricated through metal-organic vapour phase epitaxy and a conductivity-selective electrochemical etching process — defect-mediated via nanoscale pipes formed at etched threading dislocations intrinsic to the heteroepitaxial growth process. A non-invasive, sub-surface, plan-view imaging modality that uses the backscattered electron (BSE) signal as generated by a 20 keV primary electron landing energy in the scanning electron microscope (SEM) was proposed and subsequently validated. To this end, I developed a correlative microscopy approach to understanding the spatial resolution, image contrast, and information depth associated with sub-surface BSE-SEM micrographs. In our porous GaN/sapphire DBRs, BSE-SEM micrographs were dominated by the pore morphology associated with the first porous layer (around 100 nm sub-surface) with only diffuse contributions from the second layer (around 200 nm sub-surface). This method provided immediate insights into pore morphology changes driven by variations in crystallographic defect density or type, as well as electrochemical etching conditions, establishing its utility as an accessible, routine characterisation technique alongside cross-sectional SEM. Our sub-surface BSE-SEM imaging technique was applied to study the first porous layer microstructure and optical reflectance spectra of porous GaN DBRs as a function of etching voltage and substrate material. The porosity of the first porous layer increased with etching voltage as well as the proportion of threading dislocations that were participating in the electrochemical process. The highest peak reflectance was achieved at moderate etching voltages, where the first porous layer comprised domains of branched nanoscale pores centred on nanopipes. Here, a moderate etching voltage uniformly porosified highly Si-doped GaN layers whilst minimising porosification in notionally non-porous layers and preventing layer collapse. These findings showed that the defect-mediated etching process initially developed for GaN/sapphire also applies to the burgeoning low-cost, mass-market GaN-on-Si platform and underscores the usefulness of our sub-surface BSE-SEM imaging technique. Serial-section tomography in the focused ion beam scanning electron microscope (FIB-SEM) allowed for volumetric reconstructions of porous GaN DBRs with microscale volume and nanoscale spatial resolution. In our highest-resolution dataset, voxel dimensions of (2 x 2 x 5.1) nm were achieved by leveraging automated tracking and post-process interpolation of slice thickness, enabling unprecedented insight into the three-dimensional porous morphology. The insights provided by tomographic reconstruction allowed us to improve upon the established model for defect-mediated electrochemical etching by proposing a 'cascade' model where the etchant cascades through the material via vertical etching down nanopipes and horizontal etching across branched pores. This framework also considers the inclination of threading dislocations, premature nanopipe termination, and discontinuities in nanopipe formation. Crucially, these volumetric data evidenced generality for the model in successfully etched porous GaN DBRs despite differences in substrate material or etching voltage. Our tomographic reconstruction and sub-surface imaging workflows should be refined in tandem with the evolution of porous nitride semiconductor materials, templates, and devices, alongside their adoption across the broader field of engineering porous materials at multiple scales.