Alexander Ako Khajetoorians

_{Institute for Molecules and Materials, Radboud University, Nijmegen, The Netherlands}

Can we use complex magnetic states in materials for neural computation?

The unmeetable demands on the electricity consumption in information and computing technology, as well as the rapid growth of pattern recognition and deep learning have created a new challenge in materials-based research: the creation of materials which are capable of energy-efficient pattern recognition. To this end, magnetic phenomena in solid-state materials offers a plethora of options which meet the conditions for neural-based computation, such as memory, non-linearity, and weighted interactions. In this talk, I will review the basic concepts of neural computation and its relationship to solid- state material systems. I will focus on a newly founded type of spin-glass order, or the so-called spin Q glass, which we have recently discovered experimentally in an elemental crystal utilizing ultra-low temperature spin-resolved scanning tunneling microscopy in magnetic fields. I will then discuss new theoretical work, in which such states of matter can be tuned and utilized as associative memory in a Hopfield scheme, derived from tunable and ordered spin arrays. As an outlook, I will also present on- going experimental work in novel material systems, which may be utilized to mimic synaptic behavior.