Search

Dissertation, Rheinisch-Westf��lische Technische Hochschule Aachen, 2021; Aachen : RWTH Aachen University 1 Online-Ressource : Illustrationen, Diagramme (2021). = Dissertation, Rheinisch-Westf��lische Technische Hochschule Aachen, 2021, The further development of today's memory technologies faces several technological barriers, which yields the demand for new approaches. One emerging memory type that has the potential to overcome these barriers is the valence change memory (VCM), which may also be usable for neuromorphic applications or in-memory computations. In VCM devices, the information is stored within different resistive states, namely a high resistive state (HRS) and a low resistive state (LRS), which can be programmed with electrical stimuli. The transition from the HRS to the LRS is referred to as SET and the opposite transition as RESET. In this thesis, the SET and RESET times of TaOx-, ZrOx-, and HfOx/TiOx-based VCM devices were studied in the time regime from 50 ps to 100 ns. Signals in this time regime contain high frequency components in the gigahertz regime. The signals require proper impedance matching up to the VCM device, which are, therefore, integrated into coplanar waveguide (CPW) structures. Nevertheless, an integrated VCM device constitutes a parallel plate capacitor with a considerable electrical charging time, which delays the measured SET and RESET times. To estimate the electrical charging time, an experimental approach was used, with which the time-dependent effective voltage at the VCM device could be determined. This approach used the Fourier transformation of the applied voltage pulse and the VCM device's scattering parameters. From the resulting effective voltage at the VCM device, the electrical charging time of all tested VCM devices was determined. The results indicate that conducted optimizations of the CPW structure decreased the electrical charging time significantly and allowed fast measurements on a sub-100 ps time scale. By employing dedicated hardware and integrating this hardware into a software, the measurements could be automated. This allowed collecting comprehensive data sets on the SET and RESET kinetics of the tested VCM devices and, thereby, allowed resolving their transient resistance on a picosecond time scale. The measured SET kinetics revealed that all studied devices can switch within 50 ps from the HRS to the LRS. To the author's knowledge, this is the fastest switching time reported for VCM devices. By comparing the SET kinetics of differently sized TaOx-based VCM devices and using the determined electrical charging times, it could be shown that the SET kinetics of VCM devices in the subnanosecond regime are mainly limited by the electrical charging time. The migration of mobile donors (e.g. oxygen vacancies), which limits the SET kinetics on slower time scales, and the heating time of the VCM device have only a minor influence. Achieving similar fast RESET times proved to be more difficult. In all studied VCM devices, the coexistence of a unipolar switching mode could be shown. This unipolar switching mode is triggered at higher voltages, which would be required for faster RESET times. Nevertheless, on some HfOx/TiOx-based VCM devices 50 ps fast RESET times could be measured repeatedly. From these kinetics measurements, RESET programming windows could be determined and, finally, approaches were derived to achieve faster RESET times., Published by RWTH Aachen University, Aachen

Future development of the modern nanoelectronics and its flagships internet of things and artificial intelligence as well as many related applications is largely associated with memristive elements. This technology offers a broad spectrum of functionalities, however, it follows predominantly a phenomenological approach and crucial challenge/limit for further development remains variability and lack of fundamental materials' design strategy. Here we demonstrate the vital importance of materials' purity for determining memristors' functionalities, showing that part per million foreign elements significantly change the performance. By appropriate choice of chemistry and amount of doping material we can selectively enhance desired operation mode. We highlight how dopant dependent structure and charge/potential distribution in the space charge layers and the cell capacitance determine the device kinetics and functions. We evidence for first time experimentally the relation between materials properties and switching/neuromorphic performance, thus providing rules and directions for a rational design of memristive devices.

APL materials 7(9), 091105 (2019). doi:10.1063/1.5108654, Published by AIP Publ., Melville, NY

Faraday discussions 213, 197-213 (2019). doi:10.1039/C8FD00117K, Published by Royal Society of Chemistry, London, Cambridge [u.a.]