Neuromorphic engineering in GaN HEMTs exploiting dendritic dislocations for neuromodulation behaviors and adaptive intelligent power forecasting systems

Abstract

Gallium nitride (GaN), a prominent III-V semiconductor, plays a pivotal role in the advancement of sophisticated power systems. This study presents an innovative neuromorphic engineering approach utilizing the unique dendritic dislocations extending from bulk to channel within GaN High Electron Mobility Transistors (HEMTs) for three-dimensional charge modulation. This approach dynamically alters charge states and channel potentials in response to multiple parametric stimuli, effectively mimicking neural processing and transmission functions, as well as emulating direct and indirect neuromodulation behaviors observed in human nervous system. The advanced neuromodulation emulation enhances biomimetic diversity, laying a solid foundation for the design of intelligent systems. Comprehensive material and electrical analyses reveal the charging and discharging behaviors of the defect energy levels, underlying the neuromorphic working mechanisms. Additionally, by incorporating artificial neural network algorithms, an intelligent power forecasting system that leverages the device’s diverse neuromorphic traits is developed for real-time and adaptive power management, optimizing efficiency and accuracy in balancing supply and demand, thereby enhancing the cost-effectiveness and sustainability of power systems. The findings highlight the neuromorphic capabilities of GaN devices and open avenues for further exploration into the intelligent power management.