Characteristics And Applications Of TiO₂ Based Memory Impedance Devices

Memristor is a newly discovered passive circuit element, which is completely different from traditional resistor, capacitor and inductor as it can memorize the whole history of applied stimulus. Compared to typical CMOS devices, memristors possess excellent features including simple architecture, nanoscale dimension, low power consumption, high program/read speed, as well as their novel multi-state storage capability, and have thus attracted increasingly more research interests. It is believed that memristors will change the fundamentals of IT science, technology and related industry, and introduce revolutions of non-volatile memory, high-performance multi-state computing architecture and single-device neuristors. To date, memristor-related research has become one of the key interdisciplinary direction of electronic science, nano science and material science. This dissertation mainly focus on research of fabrication, characteristic analysis and applications of TiO₂ based memory impedance devices. The main contents of this dissertation include:In Chapter 2, a thorough literature review of memristor-related research is given.In Chapter 3, characteristics of TiO₂ based memory impedance devices are explored.Initially, in section 3.2, it introduces the fabrication flowchart of TiO₂ based MIM devices. By employing the electrical characteristics analysis and filamentary mechanism analysis, the as-fabricated TiO₂ based MIM devices are demonstrated to be typical memristors. Among I- V characteristics test, this dissertation experimentally demonstrated the existence of Type II crossing behavior. Then in section 3.3, for the first time, this dissertation verified coexistence of memristive, memcapacitive and meminductive behaviors within as fabricated TiO₂ devices, by employing a measuring scheme combining pulse series and C-V test. Finally in section 3.4, by biasing the tested devices with AC stimuli of different frequencies, the crossing point of I- V curves are observed to drift between the first and third quadrants, while the detailed drift positions are determined by the co-effect of stimuli frequency and memory impedance contents. Through this work, it proves that the observed non-zero crossing behavior stems from the concurrent memcapacitive and meminductive behaviors other than the imperfect of classical memristor theory.In Chapter 4, two non-ideal characteristics of TiO₂ based memory impedance devices are investigated. It is shown that even provided with identical activation energy, two devices with the same initial state can still possess stochastic resistive switching（section4.2）. By employing measuring scheme comprising of pulse series and C-V test, the high capacitive state exhibits significant variability, which is correlated to the observed resistive state variability（section 4.3）. Finally, by employing filamentary mechanism, origins of above two non-ideal characteristics are verified to be randomly distributed conductive defects within devices’ active volume.In Chapter 5, a new volatile model for memory impedance devices is established.Based on the similarity between synaptic long-term depression and measured volatile memristance dynamics, a new volatile model for memory impedance devices is established and realized in SPICE environment, which can be applied in synaptic characteristics emulating（section 5.3）. It should be highlighting that the proposed model in this thesis can significantly reduce the complexity and energy consumption of artificial synaptic emulators.