Multitask Learning and Photonic Neuromorphic Computing Driven by Highly Aligned IPZO Nanofiber-Based Transistors

Abstract

Developing 1D nanofiber networks to act as the potential building blocks for use in fundamental elements of transistors is considered to be a promising approach to realize high-performance 1D electronics. This study successfully developed highly aligned indium praseodymium zinc oxide (IPZO) nanofiber field-effect transistors (FETs) through advanced electrospinning technology, exploring their significant potential in optoelectronic synaptic applications. A key finding was that doping praseodymium (Pr) into the indium zinc oxide (IZO) system effectively suppressed the formation of oxygen vacancies, which typically optimize material performance. This enhancement resulted in an impressive carrier mobility of 12.2 cm²/V·s and exceptional bias stability, essential for reliable device functionality. Leveraging the optoelectronic synaptic characteristics of the IPZO nanofiber-based FETs, biological synaptic plasticity was successfully simulated, enabling dynamic modulation between short-term and long-term memory states while implementing bioinspired signal processing operations, including high-pass filtering. Notably, a multimodal reservoir neural network based on these devices achieved 84.48% accuracy in dual-task classification of clothing type and size, maintaining exceptional robustness even under noisy conditions. This achievement showcased their capability for multitask processing and high-pass filtering, indicating promising applications in neuromorphic computing. Overall, this research not only lays the foundation for the study of neural morphological computing systems but also opens up avenues for intelligent electronic devices.