Nanomagnetic Memory and Logic: Energy-delay-reliability Trade-Off
Munira, Kamaram, Electrical Engineering - School of Engineering and Applied Science, University of Virginia
Ghosh, Avik, Department of Electrical and Computer Engineering, University of Virginia
The ability to rotate the magnetization of a single domain nanomagnet using spin polarized current or uniaxial strain leads to exciting possibilities for low-power embedded memory and logic applications. Realizing those applications for real life usage requires addressing a complex and interlinked set of problems: material properties of the ferromagnet-oxide heterostructure, spin transport, micromagnetics and thermal stochasticity of the free layer. A particular challenge the STT-RAM industry faces is maintaining a high thermal stability while trying to switch within a given voltage pulse with an acceptably low error rate and energy cost. While operating at lower barrier increases the static error in STT-RAMs, it decreases the dynamic write error rate associated with the spins freezing around stagnation points along the potential energy landscape of the nanomagnets. We introduce a comprehensive and predictive STT-RAM modeling platform that operates at different levels of complexities, ranging from a quasi-analytical model for the energy-delay-reliability trade-offs to a fully atomistic, chemistry based multi-orbital model for predictive material design and optimization. Using this platform, we identify suitable alloys for perpendicular, in-plane and partially perpendicular magnets, identify the advantages and trade-offs with double barrier junctions, and underscore the dual role of thermal fluctuations, both in hindering rotation and also in releasing spins from their stagnation points. A similar set of challenges confronts ‘straintronics based multiferroic logic, where once again thermal perturbations play a decisive role on the dynamic writing error rate. In presence of stagnation points, applied stress, demagnetization field and dipole-dipole interactions, the error rate and switching delay can be controlled by material design and by engineering the stress profile on the nanomagnets.
PHD (Doctor of Philosophy)
All rights reserved (no additional license for public reuse)