What is it about?
Spin glasses are complex disordered systems that serve as models for investigating phenomena as diverse as brain function, random lasers, and quantum dynamics. However, calculating the energy of the equilibrium states of spin glasses is challenging. The authors of this study fashioned an optical method to calculate the energy of a spin glass state. The authors represented the system’s state using an adaptive-optics mirror, which separates a laser input into multiple segments, and read the energy of the spin glass state by collecting the intensities of the scattered laser segments. The authors’ optical method is advantageous over digital algorithms for parallel energy calculation for large-scale disordered systems. The findings carry potential implications for improving neural networks for sound recognition, image reconstruction, and high-definition fiber endoscopy in medical diagnosis.
Photo by MARIOLA GROBELSKA on Unsplash
Why is it important?
Optics is a unique platform to speed up computation, providing advantages especially for inherently parallel operations. Here we demonstrate an optical experiment enabling the simulation of a paradigmatic prototype model for statistical mechanics: the spin glass (SG). SGs are systems presenting all the aspects of complexity, such as many equilibrium states producing multiple relaxation times and nontrivial phenomenology. With the optical approach, the spin state is mapped on the mirrors of a fast adaptive optics device illuminated by a laser, and the energy calculation is performed by collecting the intensity after all the controlled light rays interfered in strongly scattering medium. The optical simulation enables beating the digital calculation in terms of speed for large-sized SG.
Read the Original
This page is a summary of: Optical computation of a spin glass dynamics with tunable complexity, Proceedings of the National Academy of Sciences, May 2021, Proceedings of the National Academy of Sciences, DOI: 10.1073/pnas.2015207118.
You can read the full text:
The following have contributed to this page