From Material Understanding to Volatile Threshold Switches Using Crystalline Zinc Oxide: Selector Device for X-point Memory Application

A cross-point array is considered a promising architecture that accelerates neuro-inspired machine learning algorithms. Leakage current arising from neighboring un/half-selected memory cells is the main source of power dissipation in the cross-point array and it also increases the read/write disturbance when not properly suppressed. Leakage current is considered as one of the main hurdles that must be overcome to increase the density of cross-point memory arrays, where highdensity is essentially required to achieve the neuromorphic network. A filament-type selector has been suggested as a threshold switching selector that holds the potential for applications in largescale integration that reduces read/write disturbance, and power consumption because of its ultralow leakage current compared to other types of threshold switching selectors. This dissertation focuses on the characterization of the surface and interface of crystalline zinc oxide (ZnO) switching layers, to propose a way to mend the common drawbacks associated with the filament-type threshold switching selectors; threshold voltage variability and poor cycling endurance. Initially, a variety of materials characterization techniques, such as Raman, X-ray photoelectron spectroscopy (XPS), X-ray diffraction (XRD), scanning electron microscope (SEM), four-point probe, Hall measurement, including in-situ electrical monitoring using UHV cluster tool, have been employed to understand the material properties of atomic layer deposited (ALD) crystalline ZnO. Afterward, a technique so-called “super-cycle ALD”, alternating ZnO ALD and Ag metal ALD, to lightly dope ZnO with Ag, has been proposed. Switching parameters are evaluated, however, unstable volatility of silver precursor made it nearly impossible to reproduce the results using ALD. Thus, to lightly dope ZnO with silver, the electrochemical deposition (ECD) process is then adopted as it is highly beneficial in controlling the doping concentration in ZnO with atomic percent precision. ECD process helps in the understanding of volatile switching behavior in lightly doped ZnO with Ag, however, inherently has difficulties in controlling the morphology and thickness. To circumvent the shortcomings associated with both processes, the co-sputtering process (with two different targets sputtered in synchronized phases) has been employed. The significantly improved switching parameters are explained based on Raman, XPS, XRD, AFM, HR-TEM, and semiconductor parameter analyzer.