Information processing with spin-coupled multi-magnet networks

Angik|Fairbairns Sarkar, Purdue University

Abstract

The speed and efficiency of information processing in conventional charge based field effect transistors have progressed dramatically in the last 50 years due to scaling of transistor dimensions. However, the fundamental scaling limits of this technology are threatening to halt this progress; leading to an enormous interest in alternative computation schemes and devices. All-spin logic (ASL) is one such alternative approach to information processing where information is stored in the magnetization of nanomagnets coupled by spin coherent channels. We developed an experimentally benchmarked spin-coupled multi-magnet simulator to investigate two major aspects of ASL operation. Firstly, we evaluated the energy (E) delay (τ) performance metric of ASL switches and found that similar to transistors, Eτ = Q²R where R is the resistance of the device and Q, total charge supplied by the power supply in the switching process is proportional to the number of Bohr Magnetons comprising the magnet. Secondly, we found that the inbuilt non-reciprocity in ASL allows cascading them to form spin-coupled multi-magnets networks (SMN) where universal logic gates, ring oscillator and other interesting circuits can be implemented. These circuits seem difficult to implement experimentally at this time, since the large magnets used today require very high threshold spin signal to switch. Hence, for ease of experimental implementation, we proposed a different class of devices which could operate at sub-threshold bias; but still rely on the spin torque based interaction between nanomagnets. We have also developed a generic framework that can be used to understand the results of such sub-threshold experiments on spin-coupled multi-magnets networks.^

Degree

Ph.D.|Ph.D.

Advisors

Supriyo Datta, Purdue University.

Subject Area

Engineering, Electronics and Electrical|Nanoscience|Information Science

COinS