[Paper Review] Memristive, Spintronic, and 2D-Materials-Based Devices to Improve and Complement Computing Hardware

Dovydas Joksas, AbdulAziz AlMutairi+6 more|arXiv (Cornell University)|Mar 11, 2022

Advanced Memory and Neural Computing207 references46 citations

TL;DR

This perspective paper proposes memristive, spintronic, and 2D-materials-based devices as transformative alternatives to conventional CMOS technology, enabling energy-efficient, non-volatile memory and neuromorphic computing. By leveraging resistance switching in ReRAM, spin-dependent transport in MTJs, and tunable electronic properties in 2D heterostructures, these technologies address the memory wall and energy inefficiency of von Neumann architectures, with key advances in scalability, low-power operation, and compatibility with existing fabrication processes.

ABSTRACT

In a data-driven economy, virtually all industries benefit from advances in information technology -- powerful computing systems are critically important for rapid technological progress. However, this progress might be at risk of slowing down if we do not address the discrepancy between our current computing power demands and what the existing technologies can offer. Key limitations to improving energy efficiency are the excessive growth of data transfer costs associated with the von Neumann architecture and the fundamental limits of complementary metal-oxide-semiconductor (CMOS) technologies, such as transistors. In this perspective article, we discuss three technologies that will likely play an essential role in future computing systems: memristive electronics, spintronics, and electronics based on 2D materials. We present how these may transform conventional digital computers and contribute to the adoption of new paradigms, like neuromorphic computing.

Motivation & Objective

• Address the growing energy and performance gap in AI and machine learning workloads driven by increasing data demands.
• Overcome fundamental limitations of CMOS technology, including power density and scaling constraints.
• Explore how memristive, spintronic, and 2D-materials-based devices can improve conventional digital computing and enable new paradigms like neuromorphic computing.
• Identify key materials and engineering challenges—doping, dielectric integration, contact resistance, and scalable fabrication—for practical deployment.
• Assess the potential for coexistence and synergy between emerging technologies and conventional CMOS-based systems.

Proposed method

• Review physical principles of memristors based on resistance switching in ReRAM, PCM, and MRAM, emphasizing non-volatile, low-power operation.
• Analyze spintronic devices such as magnetic tunnel junctions (MTJs) and spin-transfer torque (STT) effects for non-volatile memory and spin-based logic.
• Examine 2D materials (e.g., transition metal dichalcogenides, hBN) for field-effect transistors and heterostructures with tailored electronic and dielectric properties.
• Integrate these technologies into system-level architectures, including crossbar arrays for analog AI accelerators and neuromorphic spiking networks.
• Evaluate device performance using metrics like energy-delay product, retention time, switching speed, and resistance window.
• Assess compatibility with existing CMOS processes and scalability through materials engineering and fabrication techniques like atomic layer deposition and CVD growth.

Experimental results

Research questions

• RQ1 How can memristive devices overcome the memory wall and energy inefficiency of von Neumann architectures?
• RQ2 What are the key physical mechanisms enabling spintronic devices to function as non-volatile memory and logic elements?
• RQ3 What are the primary materials and fabrication challenges hindering the integration of 2D materials into high-performance computing systems?
• RQ4 How can memristors, spintronics, and 2D materials be combined to enable scalable neuromorphic computing systems?
• RQ5 What system-level architectures are required to fully exploit the potential of these emerging technologies in real-world AI and edge computing applications?

Key findings

• Memristive devices such as ReRAM, PCM, and MRAM offer non-volatile, low-energy, and scalable alternatives to traditional flash memory, with resistance switching enabling multi-level storage and in-memory computing.
• Spintronic devices like MTJs enable non-volatile memory with fast switching speeds (sub-ns) and low static power, suitable for high-density storage and spin-based logic operations.
• 2D materials such as transition metal dichalcogenides (TMDs) and hexagonal boron nitride (hBN) enable atomically thin transistors and heterostructures with tunable bandgaps and high-κ dielectrics, though doping and contact resistance remain critical challenges.
• The integration of 2D dielectrics like oxidized HfO2 and ZrO2 on TMDs enables near-perfect interfaces and high-κ properties, improving gate control and device performance.
• Hybrid architectures combining memristors, spintronic devices, and 2D materials in crossbar arrays show promise for analog AI accelerators, achieving energy efficiency improvements of up to 10x over conventional CMOS-based systems.
• Despite progress, challenges in scalable, defect-free, and reproducible fabrication—especially for 2D materials and high-κ dielectrics—remain significant barriers to industrial adoption.