Computation via Interacting Magnetic Memory Bites: Integration of Boolean Gates

Arrays of interacting, switchable, magnetic nano-islands---often called artificial spin ices---have been recently employed to demonstrate deliberate, exotic, collective behaviors not seen in natural materials. They have also been seen as potential novel platforms for memory encoding, while recent advances in the direct manipulation of these local moments suggest that these systems can be naturally interpreted as nanopatterned, interacting memory media. Exploiting their interaction, we propose here to employ them for computation {\it within} a magnetic memory. The magnetization states of each elongated nano-island can be represented as a binary degree of freedom, or bit. However, unlike in traditional magnetic memory, these bites interact. We show that they can be assembled into elementary 2-input/1-output boolean gates, such as AND, OR, NAND, and NOR, depending on their mutual geometric arrangement. In a first step of a larger program, we demonstrate numerically the physical feasibility of gate integration at least into tree-like circuits, by checking that logical functionality is obtained by interaction. We then discuss conditions for the existence of phase transitions that can limit computational efficiency. While we confine ourselves here to tree-like structures and to the reproduction of existing computational frameworks, the final aim of this effort should involve developments in terms of neuromorphic computation within an interacting memory.