All spin logic: Modeling multi-magnet networks interacting via spin currents

Srikant Srinivasan, Purdue University

Abstract

The increasing level of power dissipation in today's transistors, due to their continued downscaling, has led to an interest in alternatives to charge-based electronics for information processing. All-spin logic (ASL) represents one such new approach where the roles of charges and capacitors in CMOS are now played by spins and magnets. Available experiments utilizing this principle show operating voltages of the order of few tens of milli-volts, far below today's transistors. However, before an alternative logic scheme — like ASL — can be employed to build logic circuits, certain characteristics have to first be exhibited at the device level such as directionality of information transfer, implementing universal logic gates, cascading and fan-out. In order to devise and analyze ASL based strategies that can incorporate these device characteristics, this report first introduces a novel 4-component Spin-Circuit formalism, which is then coupled to an existing model for magnetization dynamics. This coupled model can simultaneously describe two distinct physical phenomena: (1) spin torque switching of magnets and (2) generation and transport of non-collinear spin currents in spin diffusive channels. The model is first benchmarked against available experimental data and is then used to provide key insights at the ASL device level, such as how to incorporate inbuilt directionality of information transfer and to propose scaling laws. Towards the end of this report, the model is extended to simulate multi-magnet ASL networks interacting via spin currents. In particular, examples of an ASL ring oscillator and a universal NAND gate are presented, which form the basis for designing large scale ASL circuits.

Degree

Ph.D.

Advisors

Datta, Purdue University.

Subject Area

Electrical engineering|Nanotechnology

COinS