Antiferromagnetic materials for fast and energy-efficient computing

University of Cambridge

What if the world’s computers ran one thousand times faster? This would be a huge boon to a wide range of fields: allowing accelerated drug discovery, broad access to artificial intelligence, and improved financial modelling, as just a few examples. But as our current computing technologies require a constant input of energy – which is mostly wasted as heat – any improvements in speed are met with increased energy consumption, which is unacceptable in an era of soaring energy costs and concern about climate change. My vision is to create a best-of-all-worlds data storage device that combines high speeds, energy-efficiency and non-volatility, all at a low cost.

Antiferromagnets should have ideal properties for such an application. On the atomic scale, these materials possess spins arranged in a periodic counteracting fashion, leading to theoretical switching speeds at picosecond timescales while still offering non-volatile, energy efficient storage. The opposing arrangement of spins also allows for high on-chip packing density, bringing down production costs. Unfortunately, antiferromagnets are also notoriously difficult to characterise – simply imaging the spin structure in an antiferromagnet ordinarily requires high-energy X-rays produced at a synchrotron facility. The impossibility of practically accessing the local spin state in a realistic device has so far hampered exploitation of these useful properties.

This project stands at the interface between Physics and Materials science, and seeks to resolve this problem by developing memory devices that instead incorporate unconventional antiferromagnets with a triangular spin structure. These materials host a topological electron wavefunction phase effect, known as Berry curvature, that manifests as anomalous physical properties not normally present in antiferromagnets and creates a measurable electrical response. Such properties provide new opportunities to visualise and control the underlying spin state quickly and easily using a large variety of external stimuli such as electric current, heat, light and strain. The proposal seeks to harness these properties in real-world devices, in order to build up a picture of the spin structure on the nanoscale and explore the most efficient ways to facilitate ultrafast switching.