Design, Simulation and Analysis of Current-Driven Logic Gates for Neuromorphic Computation-in-Memory

Abstract

To date, numerous studies have focused on voltage-controlled logic-in-memory (LiM) for neuromorphic computing applications. However, only a limited number of works have explored current-controlled operations (SET, RESET, and READ), despite demonstrating certain advantages compared to voltage-controlled approaches. In this work, we propose a NOT and NOR currentdriven MAGIC-based logic-in-memory gates compatible with crossbar architecture. Both gates have been simulated using a threshold-type voltage-based SPICE-compatible model. The results demonstrate its viability, even though full switching of the output memristor is not consistently achieved as desired. We conducted an experimental emulation of the NOT gate with a resistor to corroborate the proposed theory, revealing at least a 2.6μA range of working values for the current pulse to operate the gate, with full switching of the output memristor achieved when desired