Reconfigurable Compute-In-Memory on Field-Programmable Ferroelectric Diodes

TL;DR

This paper introduces a reconfigurable, transistor-free compute-in-memory architecture using ferroelectric diodes that enables storage, search, and neural network operations with high density and scalability on silicon microprocessors.

Contribution

The work presents a novel, scalable CIM architecture based on ferroelectric diodes that combines storage, search, and neural network functions on a single platform.

Findings

Demonstrated search operations with cell footprint < 0.12 um2 at 45-nm technology.

Achieved neural network operations with 4-bit precision.

Integrated device and circuit designs compatible with silicon microprocessors.

Abstract

The deluge of sensors and data generating devices has driven a paradigm shift in modern computing from arithmetic-logic centric to data-centric processing. Data-centric processing require innovations at device level to enable novel compute-in-memory (CIM) operations. A key challenge in construction of CIM architectures is the conflicting trade-off between the performance and their flexibility for various essential data operations. Here, we present a transistor-free CIM architecture that permits storage, search and neural network operations on sub-50nm thick Aluminum Scandium Nitride ferroelectric diodes (FeDs). Our circuit designs and devices can be directly integrated on top of Silicon microprocessors in a scalable process. By leveraging the field-programmability, non-volatility and non-linearity of FeDs, search operations are demonstrated with a cell footprint < 0.12 um2 when projected onto 45-nm node technology. We further demonstrate neural network operations with 4-bit operation using FeDs. Our results highlight FeDs as candidates for efficient and multifunctional CIM platforms.