Metal Oxide-Based Synaptic Devices for Flexible Neuromorphic Electronics
This thesis investigates metal-oxide-based memristive and transistor devices as building blocks for flexible neuromorphic electronics and integrated memory systems, with particular emphasis on resistive switching, optoelectronic synaptic functionality, device reliability, mechanical flexibility, and transistor selected memory integration.
Overview
The rapid growth of artificial intelligence, edge computing, and wearable electronics has created increasing demand for energy-efficient hardware capable of integrating sensing, memory, and information processing.
First, a wafer-scale ITO/MnOx/Pt resistive random-access memory (RRAM) platform was developed and transformed into a flexible architecture. The devices exhibited stable resistive switching under mechanical deformation together with wavelength-dependent optoelectronic responses. Synaptic functions, including excitatory and inhibitory postsynaptic currents, paired-pulse plasticity, short-term to long-term memory transition, learning-forgetting-relearning, and associative learning, were demonstrated, alongside neuromorphic vision and reconfigurable logic functionalities. To establish a reliable transistor platform, a-IGZO thin-film transistors (TFTs) were subsequently investigated, with HfO$_2$ passivation employed to improve environmental and operational stability by protecting the sensitive IGZO back-channel surface.
Building on this platform, flexible a-IGZO optoelectronic synaptic transistors were developed using an HfO$_2$ gate dielectric. Defect-mediated carrier trapping/de-trapping and persistent photoconductivity enabled optical synaptic plasticity and multimodal information processing with ultra-low energy consumption of approximately 1.6 fJ per synaptic event at 0.01 V. The multifunctionality of the devices was further demonstrated through heatstroke risk assessment and temperature-resilient neuromorphic vision, establishing a connection between oxide defect physics and system-level intelligent sensing.
Finally, to address the leakage-current limitations of the earlier passive RRAM platform, a lower-leakage SiOₓ RRAM was developed and electrically integrated with an IGZO TFT to realize a 1T1R memory cell. The integrated architecture retained resistive-switching functionality while enabling transistor-controlled selection and current regulation, providing a pathway toward suppressing unintended current conduction in future memory arrays.
Presenters
Brief Biography
Manoj Kumar Rajbhar is a Ph.D. student in Electrical and Computer Engineering at King Abdullah University of Science and Technology (KAUST), under the supervision of Professor Nazek El-Atab. Before joining KAUST, Manoj worked as senior research fellow at Indian Institute of Technology Jodhpur (IITJ).