Optical polarization for ultrafast computing

Introduction

Computing impacts every aspect of modern science and society. Von-Neumann computers are inefficient in addressing important combinatorial optimizations widespread from networks to artificial intelligence. Finding alternative paradigms to tackle these intractable problems is now vital as digital hardware approaches its ultimate performance limit.

Background

A promising strategy maps the optimization in spin models implemented on a physical setup that evolves as a spin system to the solution. This fascinating scheme has recently attracted huge interest, leading to various electronic and optical Ising machines that, however, remain challenging to scale and speed up.

Project Overview

LOOP is an experimental proposal that aims at creating ultrafast optical hardware for combinatorial optimization by exploiting the polarization of light. The optical polarization is a new ingredient in computing that is absent in electronics.

Concept

The original idea is to use the polarization as an ultrafast nonlinear oscillator that encodes the spin. We demonstrate the concept by controlling polarization bifurcation in third-order nonlinear crystals by electro-optics, enabling a low-power computing mechanism that operates at picoseconds.

Goals

The main goal is the realization of a polarization Ising machine that solves NP-hard optimization problems with thousands of spins in microseconds. The all-optical machine operates orders-of-magnitude faster than state-of-the-art digital and analog hardware.

Features

• The scheme is fully parallel and scalable by combining polarization dynamics with spatial light modulation.
• The vectorial nature of the polarization allows surpassing the binary operation of Ising machines.
• We exploit orbits on the Poincaré sphere to realize an ultrafast XY machine for continuous optimization, opening a new direction in optical computing.

Interdisciplinary Approach

LOOP is an interdisciplinary program that spans from nonlinear optics to complex science to establish a new computing paradigm for optimization in the post-Moore's law era.