Defect-Engineered MoS2 for Brain-Inspired Computing and Optoelectronic Applications

Abstract

Two-dimensional (2D) transition metal dichalcogenides (TMDs) have garnered significant attention for next-generation computing and optoelectronic applications due to their atomic thickness, tunable electrical and optical properties, presence of various defects and strong light-matter interactions. Among them, MoS2 -based memristors are especially promising for brain-inspired computing, as they offer nanoscale footprints and simplified architectures for implementing synapses and neurons. However, practical realization is hindered by challenges such as low switching ratio, nonlinear and asymmetric synaptic weight updates, limited dynamic range, and high device-to-device and cycle-to-cycle variability. This thesis addresses these limitations through defect engineering approaches, including Ar plasma treatment, dielectric engineering and substitutional doping. We show that optimal Ar plasma treatment of MoS2 improves memristive characteristics by enhancing the switching ratio by two orders of magnitude, enabling more linear and symmetric weight modulation. Dielectric engineering using SiNx as a gate dielectric in monocrystalline MoS2 memtransistors eliminates the need for grain boundaries or post-growth treatments, resulting in devices with a high switching ratio (104), broad dynamic range (>90), low device-to-device and cycle-to-cycle variability. Further, a solvent-based cation-exchange morphotaxy method was employed to dope Cu atoms into CVD-grown MoS2 monolayers. Cu-doped MoS2 memtransistors exhibit multifunctionality, offering both non-volatile memory behaviour and reliable synaptic plasticity. These devices feature wide memory windows (>30 V), extinction ratios exceeding 102, and more than 30 stable conductance states. Linear and symmetric potentiation and depression behaviours enable precise weight control during synaptic updates. Implementation of these enhancements is demonstrated through Artificial Neural Network (ANN) simulations, yielding a learning accuracy of 97\% on the modified national institute of standards and technology (MNIST) hand-written digits dataset. Beyond neuromorphic functionality, the Cu-doped MoS2 memtransistor also demonstrates excellent optoelectronic performance. The induced p-type doping effectively suppresses dark current by four orders of magnitude and enhances the light-to-dark current ratio by over 1,000-fold. The photodetectors achieve specific detectivity values approaching 1014 Jones and exhibit fast photoresponse, outperforming many previously reported doped TMDs devices. This thesis demonstrates how defect engineering in MoS2 can unlock high-performance functionalities in both memristive and optoelectronic devices. The findings provide a pathway for developing multifunctional, energy-efficient components for future computing and sensing technologies.