In-situ, In-Memory Stateful Vector Logic Operations based on Voltage Controlled Magnetic Anisotropy

Area of Science:

• * Spintronics
• * Computer Engineering
• * Materials Science

Background:

• * Emerging applications like AI and IoT demand high compute power, straining traditional von Neumann architectures.
• * The von Neumann bottleneck limits energy efficiency and throughput due to data transfer between processing and memory units.
• * In-memory computing, performing computations within memory, offers a promising solution.

Purpose of the Study:

• * To propose a novel approach for stateful, in-memory Boolean logic operations.
• * To leverage the voltage-controlled magnetic anisotropy (VCMA) effect in magnetic tunnel junctions (MTJs) for computation.
• * To demonstrate the feasibility of implementing logic gates within existing 1-transistor-1-MTJ bit-cell structures.

Main Methods:

• * Utilizing the VCMA effect in MTJs for in-situ computation.
• * Exploiting voltage asymmetry in the VCMA effect to create stateful IMP (implication) gates.
• * Employing precessional switching dynamics for massively parallel NOT operations.

Main Results:

• * Successful implementation of stateful IMP gates using VCMA effect in MTJs.
• * Demonstration of massively parallel NOT operations through precessional switching.
• * Theoretical basis established for implementing other logic gates (AND, OR, NAND, NOR, NIMP) via multi-cycle operations.

Conclusions:

• * VCMA-based MTJs offer a viable path towards stateful in-memory computing.
• * The proposed method integrates computation into existing manufacturable bit-cell structures without modification.
• * This research addresses the von Neumann bottleneck, paving the way for more energy-efficient and high-throughput computing systems for AI and IoT.