A general molecular-scale dynamic memristor model based on non-steady-state charge transport kinetics and its...

Area of Science:

• Molecular electronics
• Neuromorphic computing
• Materials science

Background:

• Non-steady-state molecular dynamics, involving coupled electron transport and chemical evolution, are crucial for molecular memristors.
• A general model linking these dynamics to neuromorphic computing is currently lacking.

Purpose of the Study:

• To develop a dynamic memristor model integrating electron transport and slow chemical kinetics.
• To demonstrate the model's ability to reproduce experimental data and emulate synaptic functions.
• To explore the model's application in reservoir computing for optimized performance.

Main Methods:

• Integration of Landauer and Marcus electron transport theories with kinetics of slow processes (e.g., ion migration, conformational changes).
• Reproduction of experimental conductance hysteresis and emulation of synaptic plasticity (short-term and spike-timing-dependent).
• Incorporation of the dynamic model into a reservoir computing architecture.

Main Results:

• The dynamic memristor model successfully reproduces experimental conductance hysteresis.
• The model emulates key synaptic functions, including short-term and spike-timing-dependent plasticity.
• Computational performance in a reservoir computing setup is optimized when input frequency and bias mapping align with molecular system kinetics.

Conclusions:

• A chemistry-centric, bottom-up dynamic model provides a theoretical foundation for molecular-scale neuromorphic computing.
• Non-steady-state molecular dynamics can effectively drive information processing for future computing paradigms.
• This approach offers a pathway for developing advanced neuromorphic devices in the post-Moore era.