Threshold adaptive transistor realized with RRAMs for neuromorphic circuits

We proposed a novel threshold adaptive transistor (TAT) for neuromorphic circuits. The threshold adaptive transistor is achieved based on RRAM devices, which guarantees the compatibility with current CMOS technology. By embedding RRAM material stack between the gate electrode and gate dielectric, remarkable nonvolatile threshold adaptation could be obtained. The embedded RRAM device is kept at high resistance state, which makes it act as a nonvolatile capacitor. The threshold could be nonlinearly adjusted by the voltage pulses applied on the gate of the transistor. We quantitatively estimate the range of the capacitance variation of the RRAM device. The threshold of the TAT is simulated and show expected variation. The simulated output of an inverter including a TAT shows a nonlinear adaptive controllability.