Speaker
Description
With a rapidly increasing rate of global data production, high-performance and low-power computing is an inevitable necessity. The SOCRATES project investigates Self-Organizing Computational substRATES in the form of artificial spin ice and biological neural networks. The goal is to develop an understanding of the computational properties of the substrates, making way for a radical shift in how we do computing today. The inherent self-organizing properties of biological neural networks, cultivated on a multielectrode array, will serve as inspiration for a nanomagnetic computational system that can mimic key organizational properties of the neurons. Initially, we will investigate a square artificial spin ice, comprised of a square array of vertical and horizontal bistable nanomagnets. These systems exhibit emergent behavior typically found in networks of nonlinear nodes, similar to biological neural networks. In this work, the network response to externally applied fields is explored through micromagnetic modeling, revealing complex dynamics. We observe that the specific collective behavior is highly tunable through parameters such as magnet geometry, stacking and thickness. The inherent non-volatility of each computational node (i.e., nanomagnet) provides local memory. This reduces energy dissipation related to data movement and effectively eliminates the von Neumann bottleneck. We propose that carefully engineered artificial spin ice presents a promising solution towards efficient and powerful data analysis. By combining the complex dynamics of spin-ice systems with insights from biological neural networks the SOCRATES project provides a powerful approach toward novel computational paradigms.
