Process Optimization for Low Power HfO2-Based Resistive Switching Devices for In-Memory Computing Application

In this work we investigate four different HfO2-based switching devices with low switching power and conductance quantization for in-memory computing. We have compared the characteristics of HfO2 treated with hydrogen plasma at the midpoint of deposition with that of stoichiometric HfO2, HfO2/Al2O3 bilayer and HfZrO as the switching layer. The device fabrication was focused on engineering the defect distribution or oxygen vacancy concentration in the switching layer. PVD TiN/5 nm ALD TiN/5 nm Ru was used as the top electrode and the bottom electrode was 10 nm Ti/50 nm TiN for all the devices. We have evaluated the minimum compliance current required for switching, forming voltage, and an optimized SET pulse operation. It was observed that the switching power requirement is significantly lower in a plasma treated HfO2 resistive switching device and the devices demonstrated excellent conductance quantization.