Machine-learning recognition of light orbital-angular-momentum superpositions

Abstract

We developed a method to characterize arbitrary superpositions of light orbital angular momentum (OAM) with high fidelity by using astigmatic tomography and machine learning processing. In order to define each superposition unequivocally, we combine two intensity measurements. The first one is the direct image of the input beam, which cannot distinguish between opposite OAM components. This ambiguity is removed by a second image obtained after astigmatic transformation of the input beam. Samples of these image pairs are used to train a convolution neural network and achieve high fidelity recognition of arbitrary OAM superpositions with dimension up to five.