Date of Award

Spring 2026

Language

English

Embargo Period

5-5-2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

College/School/Department

Department of Nanoscale Science and Engineering

Program

Nanoscale Engineering

First Advisor

Nathaniel Cady

Committee Members

Alain Diebold, Christophe Valee, Vincent LaBella and Manuel Smeu

Keywords

Emerging NVM devices, Resistive Random Access Memory (ReRAM), Artificial Intelligence (AI), Tantalum Oxide (TaOx), Semiconductor

Subject Categories

Electronic Devices and Semiconductor Manufacturing | Nanoscience and Nanotechnology | Nanotechnology Fabrication | Other Materials Science and Engineering

Abstract

The rapid development of artificial intelligence, machine learning, and data-intensive computing has exposed the fundamental limitations of conventional von Neumann architectures, in which energy and time are continuously lost transferring data between physically separate memory and processing units. In contrast, the human brain performs complex computations directly at the point of memory storage through billions of parallel synaptic connections, a paradigm known as in-memory computing. Realizing this in hardware requires memory devices that are fast, energy-efficient, non-volatile, and capable of storing multiple resistance levels in an analog manner. Resistive Random Access Memory (ReRAM) based on tantalum oxide (TaOx) is one of the most promising candidates to meet these requirements, owing to its compatibility with standard CMOS manufacturing processes, low operating voltages, and demonstrated endurance of billions of switching cycles.

This dissertation presents a systematic investigation of TaOx-based ReRAM devices, with a focus on device stack engineering as the primary lever for performance optimization. Three key variables were studied: the oxygen partial pressure during switching layer (SL) deposition, the role of an inert metal capping layer (CL) above the oxygen exchange layer (OEL), and the material composition and thickness of the OEL itself. The OEL is a thin reactive metal layer positioned between the switching oxide and the top electrode that acts as a controllable oxygen reservoir, absorbing oxygen ions during the set operation and releasing them during reset, which might be contributing to stable resistive switching.

Devices fabricated at an optimum oxygen partial pressure of 0.14 mTorr demonstrated forming voltages below 3 V, an order-of-magnitude separation between the high and low resistance states (HRS/LRS), and reduced cycle-to-cycle variability. X-ray photoelectron spectroscopy (XPS) confirmed that prior condition (before devices switch) generates a favorable tantalum suboxide composition that promotes effective conductive filament formation. Inert capping layers of iridium (Ir) and ruthenium (Ru) were shown to prevent parasitic oxygen migration into the tungsten top electrode, preserving long-term switching performance and device reliability. Among four OEL metals, tantalum, titanium, hafnium, and zirconium, studied at thicknesses ranging from 1 to 7 nm, titanium produced the largest initial memory window owing to its favorable metal-oxygen bond reversibility. All OEL materials converged to a similar memory window after 10,000 switching cycles, indicating a transition from an interface-controlled regime to a bulk-controlled steady-state switching regime. OEL thickness was also found to influence forming voltage uniformity, with the 2.5 nm hafnium OEL producing the narrowest distribution.

All experimental work was conducted on a Memory Test Vehicle (MTV) platform developed at NYCREATES/University at Albany, enabling statistically meaningful characterization across hundreds of devices simultaneously. The findings of this dissertation provide practical design guidelines for engineering, high-performance, energy-efficient TaOx ReRAM devices and advance the fundamental understanding of how nanoscale material engineering governs resistive switching behavior, energy efficiency, and multi-level storage capacity in next-generation in-memory computing hardware.

License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Download

Share

COinS