Magnetic Tunnel Junctions for Enabling Novel and Neuromorphic Computing

Magnetic tunnel junctions (MTJs) are versatile devices with multiple modes of operation. Commonly thought of as memory elements in MRAM applications, magnetic tunnel junctions can also be used as stochastic bitstream generators or nanoscale oscillators. Among modern nanodevices of interest, magnetic tunnel junctions are also technologically mature and are already available in the back-end-of-line setting. In this talk, I discuss our group's recent work on utilizing these devices for alternative computing concepts, using properties like randomness and signal timing to generate energy-efficient brain-inspired computational patterns. We briefly consider the challenges of scaling device research from single-device experiments to the large numbers of devices needed for realistic computing demonstrations; to that end, I discuss NIST's Nanotechnology Accelerator Program and our upcoming hardware AI testbed that researchers can use to quickly close the technology transfer gap in moving from single-devices measurements to computing demonstrations utilizing advanced materials.