Date of Award

Summer 2017

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Electrical/Computer Engineering

Committee Director

Helmut Baumgart

Committee Member

Kin P. Cheung

Committee Member

Gon Namkoong

Committee Member

Zhili Hao

Committee Member

Christopher Bailey

Abstract

Current state-of-the-art memory technologies such as FLASH, Static Random Access Memory (SRAM) and Dynamic RAM (DRAM) are based on charge storage. The semiconductor industry has relied on cell miniaturization to increase the performance and density of memory technology, while simultaneously decreasing the cost per bit. However, this approach is not sustainable because the charge-storage mechanism is reaching a fundamental scaling limit. Although stack engineering and 3D integration solutions can delay this limit, alternate strategies based on non-charge storage mechanisms for memory have been introduced and are being actively pursued.

Resistive Random Access Memory (RRAM) has emerged as one of the leading candidates for future high density non-volatile memory. The superior scalability of RRAMs is based on the highly localized active switching region and filamentary conductive path. Coupled with its simple structure and compatibility with complementary metal oxide semiconductor (CMOS) processes; RRAM cells have demonstrated switching performance comparable to volatile memory technologies such as DRAMs and SRAMs. However, there are two serious barriers to RRAM commercialization. The first is the variability of the resistance state which is associated with the inherent randomness of the resistive switching mechanism. The second is the filamentary nature of the conductive path which makes it susceptible to noise.

In this experimental thesis, a novel program-verify (P-V) technique was developed with the objective to specifically address the programming errors and to provide solutions to the most challenging issues associated with these intrinsic failures in current RRAM technology. The technique, called Compliance-free Ultra-short Smart Pulse Programming (CUSPP), utilizes sub-nanosecond pulses in a compliance-free setup to minimize the programming energy delivered per pulse. In order to demonstrate CUSPP, a custom-built picosecond pulse generator and feedback control circuit was designed. We achieved high (108 cycles) endurance with state verification for each cycle and established high-speed performance, such as 100 ps write/erase speed and 500 kHz cycling rate of HfO2-based RRAM cells. We also investigate switching failure and the short-term instability of the RRAM using CUSPP.

ORCID

0000-0002-6200-7035

Share

COinS