Microscopic impact of isolated and clustered threading dislocations on GaN-on-GaN Schottky diode performance

Threading dislocations (TDs) strongly influence the performance and reliability of GaN power devices, yet their local impact on material properties and device operation remains insufficiently understood. In this work, we investigate two freestanding GaN substrates grown by hydride vapor phase epitaxy containing distinct dislocation configurations-clustered TDs (Wafer A) and predominantly isolated TDs (Wafer B)-using low-temperature hyperspectral cathodoluminescence (CL)combined with x-ray diffraction and electrical measurements of vertical GaN-on-GaN Schottky diodes. Hyperspectral CL mapping reveals pronounced optical and electronic inhomogeneities associated with dislocation clusters, including near-band-edge emission shifts of ∼6 meV, linewidth broadening of ∼0.4 meV, and strong luminescence quenching due to enhanced non-radiative recombinations. These effects are consistent with strain-induced segregation of donor impurities such as oxygen and silicon at dislocation cores. Donor mapping derived from CL linewidth analysis indicates local doping levels of 2.5-3.0 × 10¹⁶cm^-3within clusters, compared with approximately 1.5 × 10¹⁶cm^-3in defect-free regions. In contrast, isolated dislocations produce significantly weaker strain fields and impurity segregation. These local inhomogeneities correlate with degraded device performance: diodes fabricated on substrates containing dislocation clusters exhibit breakdown voltages (BV) that are approximately 40%-50% lower and show larger statistical dispersion compared with devices fabricated on substrates containing isolated dislocations. The reduced BV is attributed to trap-assisted leakage conduction and local electric-field distortion induced by impurity-rich dislocation networks. This work provides microscopic insight into the role of dislocation configuration in determining the electrical reliability of GaN power devices and highlights hyperspectral CL as a powerful tool for directly correlating microscopic defect structures with macroscopic device performance.