We report correlated nanoscale mapping of the structure, composition, and properties of regrown GaN p-n junctions to identify how etching and non-planar regrowth processes limit diode performance via the introduction of unintentional dopants and defect states. p-GaN was selectively regrown in n-GaN trenches with SiO2 masks of variable mask-to-trench-width ratio. Dilute Al layers were periodically introduced during regrowth as markers of the growth interface. Correlated nanoscale mapping of doping, conductivity, and dopant complexes was achieved with atom probe tomography (APT), scanning spreading resistance microscopy (SSRM), and cathodoluminescence (CL) spectroscopy, respectively. The Al marker layers, detected by APT, enabled reconstruction of the faceted growth interface and correlation of the dopant concentration with position and time. The p-GaN growth rate is proportional to the mask-to-trench width ratio while the dopant incorporation rate is invariant. At trench edges, magnesium incorporation is suppressed, and oxygen incorporation enhanced, due to preferential incorporation on the semi-polar growth surface, leading to compensation and less abrupt p-n junctions; the SiO2 mask is a source of oxygen. Residual etch damage below the regrowth interface induces n-type and p-type conductivity, creating leakage pathways. The non-uniform Mg incorporation is driven by crystal anisotropy and is thus inherent to non-planar regrowth, but can be mitigated by engineering the regrowth interface and process parameters. The unprecedented integration of spatially resolved mapping of dopants, impurities, conductivity, and carrier type is a powerful approach to discriminating distinct factors that limit the performance of regrown diodes, enabling the rational optimization of process and device design.
Read full abstract