In₂O₃/ZnO heterojunction thin film transistor for high recognition accuracy neuromorphic computing and optoelectronic...

Area of Science:

• Materials Science
• Solid-state Physics
• Neuromorphic Engineering

Background:

• Solid electrolyte-gated transistors offer enhanced chemical stability for microelectronic packaging.
• Metal oxide semiconductors, commonly used as channel materials, face limitations due to defects affecting electrical performance and solution-processability.
• Achieving stable and repeatable transistor performance via solution processing remains a significant challenge.

Purpose of the Study:

• To develop efficient, multifunctional optoelectronic devices using a solution-based fabrication method.
• To create an indium oxide/zinc oxide (In2O3/ZnO) heterojunction structure for improved transistor characteristics.
• To investigate the potential of these devices for neuromorphic applications, mimicking synaptic plasticity and neuron behavior.

Main Methods:

• Fabrication of an In2O3/ZnO heterojunction transistor using a solution-based approach.
• Incorporation of Li+-doped zirconium oxide (ZrOx) thin films as a solid electrolyte to induce synaptic plasticity.
• Implementation of the heterojunction transistor in a convolutional neural network for handwritten digit recognition and simulation of a nociceptor neuron.

Main Results:

• The In2O3/ZnO heterojunction structure exhibited energy band bending, leading to electron accumulation and enhanced transistor mobility.
• The device demonstrated high accuracy (93%) in a convolutional neural network for handwritten digit recognition.
• Successful simulation of a sensory neuron (nociceptor) response within the synaptic transistor was achieved.

Conclusions:

• Solution-processed In2O3/ZnO heterojunction transistors offer a viable pathway for creating efficient optoelectronic devices.
• The developed synaptic transistors show promise for cost-effective, three-terminal thin-film applications in neuromorphic computing.
• This research advances the development of materials and devices for next-generation artificial intelligence hardware.