Arbitrary control of orbital angular momentum entanglement using a diffractive deep neural network-based inverse design technique

Abstract

We study a pre-processing technique to manipulate the orbital angular momentum (OAM) state of the entangled photon pair generated through the spontaneous parametric down-conversion process. In particular, by combining overlap integral relation determining the OAM state of entangled photon pair, we constructed a diffractive deep neural network (DDNN) to predict a pump structure for creating any desired entanglement states. Firstly, we realized each optical element as a neural network layer using the Keras framework in Python and constructed the optical neural network having phase modulation layers. Finally, using the DDNN-based inverse design method, we numerically and experimentally demonstrated the generation of maximally entangled states in OAM space.