Efficient sorting of orbital-angular-momentum states with large topological charges and their unknown superpositions via machine learning

TL;DR

This paper demonstrates that neural networks can efficiently classify and discriminate orbital-angular-momentum (OAM) modes with large topological charges and unknown superpositions, greatly enhancing the practical use of OAM in optical applications.

Contribution

The authors develop a neural network-based method capable of recognizing large-charge OAM modes and their superpositions with high accuracy, surpassing previous limitations.

Findings

Neural networks accurately classify large topological charge OAM modes.

The method generalizes well to unknown superpositions of OAM states.

Significant reduction in experimental cost and data requirements.

Abstract

Light beams carrying orbital-angular-momentum (OAM) play an important role in optical manipulation and communication owing to their unbounded state space. However, it is still challenging to efficiently discriminate OAM modes with large topological charges and thus only a small part of the OAM states have been usually used. Here we demonstrate that neural networks can be trained to sort OAM modes with large topological charges and unknown superpositions. Using intensity images of OAM modes generalized in simulations and experiments as the input data, we illustrate that our neural network has great generalization power to recognize OAM modes of large topological charges beyond training areas with high accuracy. Moreover, the trained neural network can correctly classify and predict arbitrary superpositions of two OAM modes with random topological charges. Our machine learning approach only requires a small portion of experimental samples and significantly reduces the cost in experiments, which paves the way to study the OAM physics and increase the state space of OAM beams in practical applications.