CMOS-Compatible Learning Device for Neuromorphic Synapse Application using Adjustable Hot Carrier Injections

Abstract

In this work, we study the kinetics of charging by hot carrier injection (HCI) in CMOS-Compatible floating gate(FG) cell under variety of pulse conditions. The adapted pulse which makes weight update linearly and symmetrically, is generated by adding amount of threshold voltage shifted. To implement the adapted pulse, we propose a 3T1C Metal-Insulator-Metal(MIM) FG with simple yet effective charge sharing scheme. With this scheme we can sample the present threshold voltage and store it in a capacitor. Stored voltage in a capacitor is then coupled to required voltage. Write operation sequence in both cell and arrays is successively simulated in a 180nm CMOS technology. To improve accuracy, we added pre-charging phase which reduced charge sharing error up to by 87.7%.