Simulation of integrate-and-fire neuron circuits using HfO2-based ferroelectric field effect transistors

• Published in: 2019 26th IEEE International Conference on Electronics, Circuits and Systems (ICECS) pp. 229 - 232
• Main Authors: Suresh, Bharathwaj, Bertele, Martin, Breyer, Evelyn T., Klein, Philipp, Mulaosmanovic, Halid, Mikolajick, Thomas, Slesazeck, Stefan, Chicca, Elisabetta
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.11.2019
• Subjects: ferroelectric FET (FeFET); hafnium oxide; integrate-and-fire (I&F) neurons; leaky integration; neuromorphic circuits
• DOI: 10.1109/ICECS46596.2019.8965004

Inspired by neurobiological systems, Spiking Neural Networks (SNNs) are gaining an increasing interest in the field of bio-inspired machine learning. Neurons, as central processing and short-term memory units of biological neural systems, are thus at the forefront of cutting-edge research approaches. The realization of CMOS circuits replicating neuronal features, namely the integration of action potentials and firing according to the all-or-nothing law, imposes various challenges like large area and power consumption. The non-volatile storage of polarization states and accumulative switching behavior of nanoscale HfO 2 - based Ferroelectric Field-Effect Transistors (FeFETs), promise to circumvent these issues. In this paper, we propose two FeFET-based neuronal circuits emulating the Integrate-and-Fire (I&F) behavior of biological neurons on the basis of SPICE simulations. Additionally, modulating the depolarization of the FeFETs enables the replication of a biology-based concept known as membrane leakage. The presented capacitor-free implementation is crucial for the development of neuromorphic systems that allow more complex features at a given area and power constraint.