Reservoir Computing Enabled by Polymer Electrolyte-Gated MoS₂ Transistors for Time-Series Processing.

Area of Science:

• Materials Science
• Neuromorphic Engineering
• Solid-State Electronics

Background:

• Reservoir computing (RC) requires specialized hardware for processing temporal data.
• Molybdenum Disulfide (MoS2) exhibits unique electronic properties suitable for transistor applications.

Purpose of the Study:

• To develop a novel reservoir computing system using polymer electrolyte-gated MoS2 transistors.
• To demonstrate the efficacy of MoS2 transistors as reservoir nodes for complex temporal pattern processing.

Main Methods:

• Fabrication of polymer electrolyte-gated MoS2 transistors utilizing lithium ion intercalation.
• Implementation of a time-multiplexed virtual node architecture for the RC system.
• Evaluation of the system's performance on spoken digit recognition and chaotic time-series prediction tasks.

Main Results:

• MoS2 transistors exhibited dynamic conductance modulation via reversible phase transitions (2H to 1T').
• The RC system achieved 95.1% accuracy in spoken digit recognition (NIST TI-46 dataset).
• Chaotic time-series prediction (Lorenz system) yielded a normalized root mean square error of 0.04.

Conclusions:

• Polymer electrolyte-gated MoS2 transistors are effective building blocks for efficient RC systems.
• The proposed system demonstrates practical applicability and enhanced scalability for neuromorphic computation.
• This approach offers a promising pathway for processing complex temporal dynamics in electronic devices.