Energetically deposited tin oxide: characterization and device applications

Semiconductor oxides are promising materials that have made impressive progress in recent years, challenging the dominance of silicon not only in conventional devices including field-effect transistors but being amenable to next-generation electronic devices such as memristors. Although a variety of oxides have been explored, tin oxide has been an interesting material for researchers when offering p-type characteristics of tin monoxide SnO and n-type characteristics in tin dioxide SnO2. While SnO2 is easy to grow and well suited for a wide range of applications, it is difficult to form p-type SnO due to its metastability where it forms into the more stable phase SnO2. The work presented in this Doctoral Dissertation focus on exploring the characteristics and applications of energetically deposited tin oxide thin films. The tin oxide film deposited using high-power impulse magnetron sputtering was found to be mixed-phase nanocrystalline SnO and SnO2 in which SnO2 is dominant. The high resistivity, low carrier concentration and low mobility in the as-deposited and annealed samples hindered the application of the high-power impulse magnetron sputtering (HiPIMS) SnOx in thin film transistors, however, suggested suitability for these films as a memristive material. A small but quantifiable variation in film stoichiometry (Sn:O) resulting from the off-axis deposition led to the formation of two different types of memristive devices, namely filamentary and nanoparticle network memristors. Both devices exhibited stable volatile bidirectional resistive switching with a ratio between high resistance and low resistance of more than two orders of magnitude. However, their underlying resistive switching mechanisms and device characteristics were significantly different. Synaptic-like behaviours were observed on both filamentary devices (FDs) and nanoparticle network devices (NNDs), highlighting their potential for information processing in neuromorphic computing systems. While a FD can become only an individual cell in reservoir computing circuits, an NND can be implemented as a reservoir due to their available inter-connectivity which is required for reservoir computing