Ultrathin Nitride Ferroic Memory with Large ON/OFF Ratios for Analog In-Memory Computing

Area of Science:

• Materials Science
• Solid-State Electronics
• Non-volatile Memory Technologies

Background:

• Digital von Neumann systems face energy and complexity challenges.
• Current ferroelectric resistive memories exhibit limitations in ON/OFF ratios, imprint, and semiconductor compatibility.
• Need for advanced memory solutions for analog computing and edge devices.

Purpose of the Study:

• To demonstrate ferroelectric and analog resistive switching in a novel epitaxial nitride heterojunction.
• To bridge the performance and compatibility gap in ferroelectric memory.
• To explore the potential of scandium aluminum nitride (ScAlN) for advanced computing architectures.

Main Methods:

• Fabrication of ultrathin (≈5 nm) ScAlN nitride ferroelectric layers within a heterojunction.
• Characterization of resistive switching behavior in a metal/oxide/nitride ferroelectric junction.
• Evaluation of memory characteristics including ON/OFF ratio, uniformity, retention, and cycling endurance.
• Demonstration of multi-state operation, linear analog computing, and image processing capabilities.
• Neural network simulations for image recognition using memory weight update characteristics.

Main Results:

• Achieved high ON/OFF ratios (up to 10^5), excellent uniformity, good retention (>10^5 s), and cycling endurance (>10^4) in the nitride ferroelectric junction.
• Demonstrated non-volatile multi-level programmability enabling analog computing and accurate image processing.
• Neural network simulations showed high image recognition accuracy (92.9%) for Modified National Institute of Standards and Technology datasets.

Conclusions:

• The developed ScAlN-based nitride heterojunction offers a promising solution for high-performance, compatible ferroelectric memory.
• This work provides landmark evidence for constructing advanced memory/computing architectures using emerging nitride ferroelectrics.
• The technology promotes the development of integrated functional edge devices beyond silicon.