Research on Pt/NiO/WO_3-:Ti/W Multijunction Memristors with Synaptic Learning and Memory Functions

Area of Science:

• Materials Science
• Neuroscience
• Computer Science

Background:

• Artificial synapses are crucial for artificial intelligence (AI) and are inspired by biological synapses.
• Resistive random-access memory (RRAM) devices show promise for artificial synapses due to their tunable synaptic plasticity.
• Existing RRAM devices face challenges like current leakage and limited durability.

Purpose of the Study:

• To develop a novel memristor device with improved characteristics for artificial synapse applications.
• To address the limitations of current two-terminal RRAM devices, specifically current leakage and poor durability.
• To demonstrate the potential for large-scale integration of memristor cross arrays.

Main Methods:

• Fabrication of a Pt/NiO/WO3:Ti/W memristor device featuring a pn-type heterojunction and metal-semiconductor contacts.
• Characterization of the device's electrical properties, including rectification and multi-conductance states under pulse modulation.
• Construction and testing of a 10x10 cross-array to evaluate device performance and leakage current.

Main Results:

• The Pt/NiO/WO3:Ti/W memristor exhibited good rectification characteristics.
• The device demonstrated multi-conductance states and memory characteristics analogous to biological synapses, attributed to internal potential barrier modulation.
• Stable enhancement and suppression of rectification were observed, and the 10x10 cross-array confirmed correct device writing with no apparent leakage current.

Conclusions:

• The developed memristor shows significant potential as an artificial synapse, offering improved performance over existing RRAM devices.
• The device's ability to operate without leakage current in a cross-array configuration is critical for large-scale integration.
• This work provides a promising structural design for advancing memristor-based AI hardware.