Enabling Scalable Photonic Tensor Cores with Polarization-Domain Photonic Computing

TL;DR

This paper introduces a silicon-photonic tensor core leveraging polarization- and wavelength-domain computing with ferroelectric materials, achieving significant accuracy improvements over traditional coherent networks.

Contribution

The work presents a novel polarization-domain photonic tensor core using 2D ferroelectric materials, enabling scalable and more accurate photonic computing.

Findings

Up to 83% improvement in computation accuracy

Experimental validation based on characterized material properties

Demonstration of polarization- and wavelength-domain computing capabilities

Abstract

We present a silicon-photonic tensor core using 2D ferroelectric materials to enable wavelength- and polarization-domain computing. Results, based on experimentally characterized material properties, show up to 83% improvement in computation accuracy compared to coherent networks.