A Frequency-Phase Potential for a Forced STNO Network: an Example of Evoked Memory

TL;DR

This paper introduces a frequency-phase potential framework for forced spin-torque nano-oscillator (STNO) networks, revealing how external forcing induces evoked memory and organizes network dynamics for applications like pattern recognition.

Contribution

It derives new amplitude and phase potentials for forced STNO networks, demonstrating how external forcing induces evoked memory and network organization.

Findings

Phase potential describes in-phase/anti-phase oscillations and resonances.

External forcing creates evoked memory only present under forcing.

Network can perform pattern recognition and logic operations.

Abstract

The network studied here is based on a standard model in physics, but it appears in various applications ranging from spintronics to neuroscience. When the network is forced by an external signal common to all its elements, there are shown to be two potential (gradient) functions: One for amplitudes and one for phases. But the phase potential disappears when the forcing is removed. The phase potential describes the distribution of in-phase/anti-phase oscillations in the network, as well as resonances in the form of phase locking. A valley in a potential surface corresponds to memory that may be accessed by associative recall. The two potentials derived here exhibit two different forms of memory: structural memory (time domain memory) that is sustained in the free problem, and evoked memory (frequency domain memory) that is sustained by the phase potential, only appearing when the system is illuminated by common external forcing. The common forcing organizes the network into those elements that are locked to forcing frequencies and other elements that may form secluded sub-networks. The secluded networks may perform independent operations such as pattern recognition and logic computations. Various control methods for shaping the network's outputs are demonstrated.