Xiwen Liu1,John Ting1,Fiagbenu Fiagbenu1,Jeffrey Zheng1,Dixiong Wang1,Pariasadat Musavigharavi1,Eric Stach1,Troy Olsson1,Deep Jariwala1
University of Pennsylvania1
Xiwen Liu1,John Ting1,Fiagbenu Fiagbenu1,Jeffrey Zheng1,Dixiong Wang1,Pariasadat Musavigharavi1,Eric Stach1,Troy Olsson1,Deep Jariwala1
University of Pennsylvania1
The deluge of sensors and data generating devices has driven a paradigm shift in modern<br/>computing from arithmetic-logic centric to data centric processing. At a hardware level, this<br/>presents an urgent need to integrate dense, high-performance and low-power memory units<br/>with Si logic-processor units. However, data-heavy problems such as search and pattern<br/>matching also require paradigm changing innovations at the circuit and architecture level to<br/>enable compute in memory (CIM) operations. CIM architectures that combine data storage yet<br/>concurrently offer low-delay and small footprint are highly sought after but have not been<br/>realized. Here, we present Aluminum Scandium Nitride (AlScN) ferroelectric diode (FeD)<br/>memristor devices that allow for storage, search and neural network-based pattern recognition<br/>in a transistor-free architecture. Our devices can be directly integrated on top of Si processors<br/>in a scalable, back-end-of-line process. We leverage the field-programmability, non-volatility<br/>and non-linearity of FeDs to demonstrated circuit blocks that can support search operations in-<br/>situ memory with search delay times < 0.1 ns and a cell footprint < 0.12 µm 2 . In addition, we<br/>demonstrate matrix multiplication operations with 4-bit operation of the FeDs. Our results<br/>highlight FeDs as promising candidates for fast, efficient, and multifunctional CIM platforms.