Exploring and utilising the effects of environmental factors on memristor behaviour

Electronic devices are sensitive to environmental conditions, the most well studied of these parameters being temperature, leading to the need for full characterisation to determine any environmental dependencies. Memristors are a new class of two-terminal metal-insulator-metal device that show great promise in applications such as high density data storage and physical neural networks. Being a new class of device, with many different combinations of materials means that a method and system for rapidly characterising the environmental dependencies of these devices needs to be developed. Humidity has been demonstrated to affect the behaviour of some memristors (most notably hafnium based), however, the effects of this on titanium based devices have not been investigated. To this end, a low-cost, miniaturised environmental control chamber has been developed, capable of reaching temperatures up to 90°C while simultaneously controlling atmospheric humidity. This system has then been used to investigate the impact of temperature and humidity of memristors constructed with a gold/titanium oxide/platinum stack at four resistivities, from “pristine” to very low resistivity. The data retrieved from these investigations indicates that TiO2 based devices would make for effective resistive temperature sensors. Characteristics of these devices such as the IV, switching, and volatile recovery behaviour have been demonstrated to be affected by changes in temperature, and temperature dependant models of these behaviours have been developed and integrated with existing models (IV, volatility and switching behaviour), allowing for a more comprehensive understanding of how memristors may be affected by real world conditions when integrated into systems that experience thermal cycling. Leveraging the thermal dependencies of the studied devices, the emulation of a homeostatic neuron was undertaken, with a number of memristors acting as the synapses, first in simulation as a proof of concept. For this project, the temperature of the devices was altered as a method of providing global synapse weight control. An initial investigation into the impact of humidity on these devices was conducted, showing evidence that exposure to atmospheric humidity can result in the restoration of switching behaviour on previously broken devices