High precision tuning of state for memristive devices by adaptable variation-tolerant algorithm

Area of Science:

• Materials Science
• Electrical Engineering
• Computer Science

Background:

• Memristive devices, particularly titanium dioxide thin films, exhibit unique properties for electronic applications.
• Resistive switching in memristors offers potential for nonvolatile memory and analog computing.
• Variability in memristor switching behavior poses a challenge for precise state control.

Purpose of the Study:

• To design a simple write algorithm for precise conductance tuning in titanium dioxide memristive devices.
• To achieve high-precision, nonvolatile states in memristors despite inherent switching variations.
• To demonstrate the utility of precisely tuned memristors in analog computing applications.

Main Methods:

• Development of a novel write algorithm to control device conductance at a specific bias point.
• Utilizing the inherent filamentary nature of resistive switching in titanium dioxide for stable states.
• Fabrication and testing of hybrid circuitry integrating memristive devices with an analog summing amplifier.

Main Results:

• Achieved 1% relative accuracy (seven-bit precision) in tuning memristor conductance within its dynamic range.
• Demonstrated that the high-precision state is nonvolatile and likely sustainable for nanoscale devices.
• Successfully performed analog multiply-and-add (dot-product) computation using a hybrid circuit with two memristive devices.

Conclusions:

• The developed write algorithm enables precise and nonvolatile state control in titanium dioxide memristors.
• Memristive devices with high-precision tuning are attractive for low-precision analog computing.
• The demonstrated hybrid circuitry showcases the potential of memristors for efficient information processing, particularly for bottleneck operations.