Vertical Self-Rectifying Memristive Arrays for Page-Wise Parallel Logic and Arithmetic Processing.

Area of Science:

• Computer Science
• Materials Science
• Electrical Engineering

Background:

• Conventional von Neumann architectures face data transfer bottlenecks.
• Memristor-based logic-in-memory (LIM) systems offer a promising solution by integrating computation within memory arrays.
• Non-volatile switching and dense integration capabilities make memristors suitable for LIM.

Purpose of the Study:

• To implement a page-wise LIM architecture using a 3D vertical resistive random-access memory (VRRAM) array.
• To perform Boolean functions and arithmetic operations entirely within the memory array.
• To demonstrate a scalable and energy-efficient in-memory computing solution.

Main Methods:

• A 3D VRRAM array composed of self-rectifying Pt-Ta2O5-Al:HfO2-TiN memristors was utilized.
• Two logic primitives, 1M and 2M logic, were employed for intra-page and inter-page operations.
• A memristive arithmetic logic unit (mALU) was designed and implemented for arithmetic functions.

Main Results:

• Core Boolean functions were executed entirely within the memory array via resistive state transitions.
• A 2-bit full adder was implemented with a minimal footprint of three cells and completed in 12 steps.
• Experimental demonstration of intra- and inter-page operations and complete mALU functionality with high reproducibility.

Conclusions:

• The proposed page-wise LIM architecture significantly reduces spatiotemporal cost.
• This architecture holds promise for scalable and energy-efficient in-memory computing.
• Memristor-based LIM systems are a viable approach to overcome von Neumann bottlenecks.