Spatial-division-multiplexing, as a promising way to enhance the fiber transmission capacity of optical networks, has been widely researched, and still suffers from the problem of device size increasing along with the space-mode order increasing. In order to solve this problem, an optical mode field compression method is needed, however, it is difficult to realize for traditional optical waveguide. To achieve a highly dense photonic integrated circuit for the transmission of higher order space-mode, a space-mode compressor using nano-pixel structure was proposed and investigated in this work. A supervised machine learning model using a deep neural network was used to design and optimize the space-mode compressor based on nano-pixel structure. As a simulated result, we achieved possible nano-pixel configurations for the purpose of mode-compression and there is a more than 50% size compression for the 0th and 1st order modes, as well as a 39.3% size compression for the 2nd and 35% for 3rd order modes.
!!!All Science Journal Classification (ASJC) codes