Design of hybrid spintronic devices at scaled technologies for non-volatile memory applications
Abstract
The ever-increasing demand for embedding more on- and off-chip memories to increase the bandwidth of high performance systems has led to a significant amount of research directed towards several potential high density memory technologies. With aggressive technology scaling, the researchers are incessantly confronted with various overwhelming challenges associated with the design of low power, ultra-high density and robust memory blocks. An alternative to all currently available memory technologies, spin-transfer torque (STT) Magnetic Random Access Memories (MRAM) offer many desirable memory-attributes. Data non-volatility, unlimited endurance, low power, high performance and high integration capabilities have stimulated an overwhelming interest for STT-MRAM among memory researchers. In an attempt to address the issues associated with parametric process variations and high switching energy consumptions, different genres of magnetic tunnel junction (MTJ) structures, memory bit-cells, and architecture are proposed. Unlike state-of-the-art tri-layer MTJ devices, the multi-port/multi-pillar structures provide the option to eliminate the self-conflicting design requirements for memory read, write and hold. Techniques to reduce thermal fluctuation induced delay spreads is discussed for reliable and deterministic magnetic switching characteristics in both in-plane and perpendicular anisotropy devices. The effect of thermal spin-transfer torque on high speed magnetic switching is discussed in the context of designing low power, robust, and reliable MRAM devices. Based on thermally initiated magnonic spin-transfer torque, we propose three new genres of multi-port MRAMs for low energy, high speed, and reliable magnetic switching. The proposition of several new genres of magnetic tunnel junctions (MTJ) based on both electric and thermal spin-transfer torque, the corresponding bit-cells, and memory architectures make STT-MRAM a promising choice as future universal memories.
Degree
Ph.D.
Advisors
Roy, Purdue University.
Subject Area
Electrical engineering|Physics
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.