Magnetic and Spintronic Device for Computing and Memory

Abstract

Magnetic devices promise intriguing design paradigms where electron spin is used as the information token instead of its charge counterpart. While magnetic random-access memory (MRAM) is considered one of the most mature nonvolatile memory technologies for next generation computers, spin-based devices can be a game-changing option for beyond-CMOS and in-memory computing. In the future cognitive era, nonvolatile memories hold the key to overcome the bottleneck in the computational performance due to data shuttling between the processing and the memory units. The application of spintronic devices for cognitive applications requires versatile, scalable device design that is adaptable to emerging material physics. Spin-orbit torque driven magnetic tunnel junction has emerged as one of the most promising candidates for energy-efficient nonvolatile logic and memory devices. In this talk, I will discuss the design-space of spintronics devices as the key building blocks for in-memory computing and benchmark the performance metrics with other state-of-the-art non-volatile memories. I will show the first experimental demonstrations of linear synaptic weight generator and the nonlinear activation function generator integrated in a single device and operating with sub-10 ns pulses. The introduction of antiferromagnetic materials in these devices promises to enable even picosecond operations. A complete neuromorphic hardware accelerator using nonvolatile magnetic devices can revolutionize computer architectures by embedding memory into logic circuits in a fine-grained fashion.