Two-dimensional Indium Oxide and Chalcogenide-based Memory Systems for Neuromorphic Devices

Two-dimensional (2D) transition metal oxides and chalcogenides have been immensely investigated after the discovery of graphene for various applications. Over conventional transition metal-based atomically thin semiconductors post-transition metals due to their lower melting point can be easily incorporated into different synthesis techniques enabling their obtainment in an atomically thin 2D (layered and non-layered) format. This has resulted in the synthesis of many post-transition metal-based compounds with inherently exceptional electronic and optoelectronic properties for semiconducting technologies. However, this new category of atomically thin semiconductors remains underexplored in many aspects of semiconducting technology in contrast to its more mature transition metal counterparts. Indium (In) a metal belonging to the post-transition metal group has been a widely sought-after material in the semiconductor industry. Its low melting point not only allows for the realization of atomically thin semiconductors using different syntheses but also allows for the feasible formation of alloys with varied properties. The distinct electronic and optoelectronic properties of In-based layered/non-layered compounds with ambient stability have led to their application in transistors and photodetectors. The presence of good electron mobility, high photoresponsivity, low dark current values and the feasibility of doping makes them a viable choice for investigating/engineering efficient electrically/opto-electronically stimulated memory systems. This can result in the devising of atomically thin memory devices for neuromorphic computation capable of addressing the bottleneck faced by traditional silicon-based CMOS technology. Hence, this PhD thesis looks into investigating the applicability of 2D In-based oxide and chalcogenide semiconductors for devising artificial memory systems that can address the issues faced by von Neumann computing architecture. The role of cation and anion diffusion in observing non-volatile resistive switching-based memory in indium selenide (InSe) was investigated. Antimony (Sb) doped In2O3 was engineered into a focal plane array acting like an artificial retina for an optoelectronic synaptic array operating in the UV domain. Defect passivated annealed In2O3 was used for demonstrating proof of concept visible light-activated artificial synapses. For the first part of the thesis, the author used micromechanically exfoliated InSe flakes to devise cross-bar resistive switches with an electrochemically active silver (Ag) electrode. The device exhibited resistive bipolar switching 103 occurring from the diffusion of cationic silver into the InSe stack. The device had a memory retention period of 24 hours and the presence of Ag for filamentary switching was demonstrated microscopically. The role of Ag and oxygen diffusion-based switching in the device and the barriers in van der Waals layers to different ionic migration in the stack was demonstrated experimentally and theoretically. This offered an insight into preferential ion-based resistive switching in 2D materials with van der Waals forces. For the second part of her research, the author investigated the synaptic application of Sb-doped In2O3 demonstrating long photocurrent retention in the UV domain at low operating voltages. The 3 nm thin devices at a minimal voltage of 50 mV were able to replicate all the key synaptic and multi-synaptic functionalities that form a part of the cognition process of the brain with 285 nm optical stimuli. The large area sheets were further engineered into a proof of concept retinal system as a 4x4 UV sensing pixel array. The array demonstrated the capability of pattern recognition and memorization with only 120 training cycles and retained the pattern memory for over 60 minutes. Moreover, stimulation frequency-dependent pattern memorization and high contrast image generation at low power were also demonstrated. The feasibility of the system with neural networks was investigated. Thus, demonstrating the applicability of this material-based pixel array for next-generation UV-based smart neuromorphic devices with reduced power requirements and fast recognition accuracy. After investigating the compatibility of In-based atomically thin semiconductors for electrical and UV-active optoelectronic memory systems for future neuromorphic devices. The author successfully demonstrated the use of annealed 2D In2O3 as an oxide semiconductor feasible for visible active synaptic applications. The resultant material demonstrated the ability to detect short pulses of 455 nm and 565 nm at an operating bias of 200 mV. The individual device demonstrated a responsivity of 6.67103 A/W at 455 nm and was able to replicate short-term potentiation (STP) and long-term potentiation (LTP)-based synaptic phenomenon using blue light. Moreover, the 2.2 nm thin devices were able to replicate multi-synaptic systems demonstrating their feasibility for next-generation ultra-thin transparent oxide-based visible active optoelectronic synapses. In conclusion, the author demonstrated significant analysis and findings during their PhD candidature, which has demonstrated the applicability of 2D In-based oxide and chalcogenide compounds for non-volatile and volatile memory systems feasible for neuromorphic computation. It can be concluded that In-based atomically thin semiconductors are a material of significant consideration for devising energy-efficient atomically scalable electrically and opto-electronically stimulated memory systems for next-generation neuromorphic technology.