Abstract
Magnetic nanostructures are fundamental to the functionality, performance, and reliability of Magneto-resistive Random Access Memories (MRAMs). MRAMs offer nanosecond-rate write times, read times, and data rates with write endurance >1015 cycles and read endurance >1015 cycles across wide operating temperature ranges, e.g., -40 °C to +125 °C, with low error rates, e.g., less than 10-10 upsets/bit-day equivalent. MRAMs are non-volatile memories that retain data when unpowered, with data retention times >15 years. In addition, MRAMs are radiation hard to Total Ionizing Dose; proton, neutron, and heavy-ion irradiation; and Single Event Effects. As such, MRAMs offer high performance, high reliability, and long life for space applications. The Magnetic Tunnel Junction (MTJ) is a multilayered magnetic nanostructure that uses exchange, anisotropy, Zeeman, magnetostatic, magnetodynamic, magnetoelectronic, magneto-transport, and spin-dependent effects to enable these desired characteristics including when used with radiation-hardened silicon-on-insulator (SOI) CMOS transistors and metallization. To achieve reliable writing, the MTJ uses either modified inductive switching or electron spin-dependent spin-transfer torque switching which can also be achieved with spin-orbit torque and voltage-controlled magnetic anisotropy approaches. Reliable reading occurs by using tunneling magnetoresistance, in which parallel and antiparallel magnetization, respectively, across a nanometer to sub-nanometer tunneling barrier provide high conductance and low resistance or low conductance and high resistance via quantum mechanical spin-dependent tunneling. The use of coupled metallic systems in the MTJ provides radiation hardness, robustness, and scalability. MTJs can be configured and operated with flexibility to implement Von Neumann and non-Von Neumann architectures. Magnetic systems serve as physical representations of artificial neural network, machine learning, and associative computing architectures.
