Tunable Synaptic Characteristics of a Ti/TiO₂/Si Memory Device for Reservoir Computing

Area of Science:

• Materials Science
• Electrical Engineering
• Computer Science

Background:

• Neuromorphic computing aims to mimic the human brain's structure and function.
• Resistive switching devices offer promising solutions for energy-efficient neuromorphic hardware.
• Understanding the impact of doping on device characteristics is crucial for optimizing performance.

Purpose of the Study:

• To fabricate and characterize Ti/TiO2/Si devices for neuromorphic applications.
• To investigate the long-term and short-term memory effects in these devices.
• To evaluate their potential in pattern recognition and reservoir computing systems.

Main Methods:

• Fabrication of Ti/TiO2/Si devices with varying silicon dopant concentrations.
• Characterization using transmission electron microscopy (TEM).
• Electrical measurements to assess resistive switching behavior and memory characteristics.

Main Results:

• Ti/TiO2/p++Si devices showed interface-type bipolar resistive switching with long-term memory, achieving >85% pattern recognition accuracy in simulations.
• Ti/TiO2/p+Si devices exhibited controllable short-term memory with pulse amplitude-dependent dynamic range and interval-dependent conductance decay.
• A reservoir computing system was built using the short-term effect, differentiating 16 states for pattern recognition.

Conclusions:

• Ti/TiO2/Si devices are viable candidates for neuromorphic systems, offering both long-term and short-term memory functionalities.
• Device performance, including memory type and switching characteristics, can be tuned by silicon doping concentration.
• These devices show potential for advanced computing paradigms like reservoir computing and pattern recognition.