Md Golam Morshed, a Ph.D. student of electrical engineering, has joined the search for high-speed, high-density, low energy consumption memory technology.
The burgeoning internet of things has made the search for a solution more urgent. For several decades, hardware miniaturization and Moore’s law have driven innovation and made digital electronics widely accessible. Nowadays, a slowdown in silicon based Complementary Metal Oxide Semiconductor (CMOS) hardware, the rapid growth of software, and accelerating migration to edge computing are creating a strong impetus to re-examine the limits of computing.
Magnetic devices play a potentially important role here – they can be integrated onto silicon. Ideas like Spin Transfer Torque Based Random Access Memory (STTRAM) and Heat Assisted Magnetic Recording (HAMR) have been commercialized, and magnets have unique properties like non-volatility, meaning they retain their bit information when switched off. However, they need novel designs to read-write and move information stored in electron spins rather than charges.
Morshed’s research, published in Physical Review Applied and Physical Review B, offers an innovative approach to next-generation spintronics memory and computing technology. Morshed explores the idea of computing with nanosized vortex-like excitations in magnetic films, called skyrmions, that can be moved at high speed with modest electrical currents, like beads on an abacus. This characteristic makes skyrmions a promising information carrying bit for some classes of analog applications.
Morshed also contributed to a Perspective article in the Journal of Applied Physics, first-authored by Hamed Vakili (Ph.D., Physics, ‘22). Morshed and Vakili worked together in the Virginia Nano-Computing Group led by Avik Ghosh, a professor of electrical and computer engineering who also holds a courtesy appointment in physics.
Morshed and Vakili collaborated on both material and device modeling, focusing on a “racetrack” memory for skyrmions. A magnetic film serves as the track, which sits on a thin layer of heavy metall. Skyrmions move along the track when a current is applied to the underlying layer.
Exploiting and controlling skyrmionic movement enables two types of memory processing. One approach, for high density storage, encodes digital information based on the presence (bit 1) or absence (bit 0) of a skyrmion at a site along the track. The second approach, known as temporal memory, encodes the arrival time of an incoming pulse directly onto the skyrmion’s positional coordinate, without the need for any analog to digital conversion (ADC).This is a distinct advantage because analog data processing with digital silicon CMOS needs such ADCs that involve a large footprint and energy.
One challenge was to hold the skyrmion in place against thermal jitter. Morshed and Vakili conducted numerical experiments to show it is possible to control skyrmions’ movement for short- and long-term memory storage. They designed a notch or defect in the racetrack that can serve as an energy barrier, holding the skyrmion in place, stable in its original position. The skyrmions can easily clear the barrier at low power with the application of a modest current pulse, referred to as an unpinning current.
Wavefront-based analog computing can be useful for applications like DNA alignment and time series matching with applications in robotics and speech. While this work is still exploratory, the research offers some flexibility to future chip designers interested in integrating magnetic devices onto CMOS. The notch size and shape, the racetrack’s magnetic parameters, and physical parameters such as the material thickness of the racetrack, represent separate knobs that can be used to tune the energy barrier that regulates skyrmions’ movement. By optimizing this energy barrier, the hold time of skyrmions can be modulated to prevent corruption of the stored data.
Morshed continues to build on his notable achievements by collaborating with experimentalists such as Prof. Andy Kent at NYU whose experiments validate his computational predictions. Their paper, Interplay between Spin-Orbit Torques and Dzyaloshinskii-Moriya Interactions in Ferrimagnetic Amorphous Alloys, is published in Advanced Science. Morshed plans to tune the skyrmion properties by scavenging through the material phase space, including machine learning techniques, to optimize their design and fabrication. A cross-university grant led by Ghosh, with partners from UVA, NYU and MIT, are conducting this research with funding from the Defense Advanced Research Projects Agency’s Topological Excitations in Electronics Program.