A Vertical Molecular Synaptic Transistor with Redox-Induced Analog States

Area of Science:

• Molecular electronics
• Neuromorphic engineering
• Materials science

Background:

• Traditional computing faces limitations in energy efficiency and processing complex tasks.
• Neuromorphic hardware aims to mimic the brain's structure and function for efficient computation.
• Molecular transistors offer potential for miniaturization and novel functionalities in electronic devices.

Purpose of the Study:

• To develop a three-terminal molecular transistor with synaptic plasticity.
• To investigate the role of redox-active molecules in enabling nonvolatile switching and analog conductance.
• To demonstrate the potential of molecular synaptic transistors in neural network applications.

Main Methods:

• Fabrication of a three-terminal device using ferrocene-terminated alkanethiolate, graphene, and gold electrodes.
• Utilizing an ion-gel gate to modulate channel conductance and induce postsynaptic-like responses.
• Characterization of synaptic plasticity, including short-term and long-term plasticity, and multilevel conductance states.
• Simulation of a neural network using the molecular synaptic transistor for MNIST pattern recognition.

Main Results:

• The molecular transistor exhibited neuroinspired plasticity, including paired-pulse facilitation and a transition to long-term plasticity.
• Ferrocene moiety enabled nonvolatile switching via redox and ion trapping, leading to programmable conductance changes.
• Control devices without ferrocene showed only transient responses, highlighting the importance of the redox-active component.
• Achieved multilevel conductance states crucial for learning processes.
• Demonstrated ~88% accuracy in MNIST recognition with a simulated neural network.

Conclusions:

• Vertical molecular transistor systems are promising for molecular-level neuromorphic hardware.
• The three-terminal, read/write-decoupled architecture overcomes limitations of two-terminal memristive devices.
• This work paves the way for developing energy-efficient, brain-inspired computing systems at the molecular scale.