Ultra-Efficient Memristive Voltage-Resistance Logic Framework With Integrated Readout Strategy Based on Hafnium Zirconium Oxide.
Abstract
The integration of logic operations and data storage is regarded as a promising strategy for enabling edge computing, a key requirement for real-time, data-intensive scientific and industrial applications. Achieving edge-capable computing, however, requires exceptional spatiotemporal efficiency, and many previously proposed logic schemes still face substantial limitations. Here, we introduce a memristive voltage-resistance (V-R) logic strategy that infers output resistance states without additional incurring area or latency penalties. The approach relies on integrated readout of the discharge behavior into programming of a two-terminal capacitor-type memristor. Implemented in an HfO2-based ferroelectric tunnel junction (FTJ) device with multiple resistive states, this strategy enables a half adder to be computed using a single device in a single step, and a 1-bit full adder using only three FTJs in three steps-achieving previously unattainable spatiotemporal cost for execution of standard benchmarking tasks. Our work provides a pathway to edge computing by maximizing the spatiotemporal efficiency achievable with memristor-based logic-in-memory.