Inverse modeling of few-mode fiber for high-speed optical communication networks

Few-mode fibers have been used in contemporary communication with mode multiplexing and space-division-multiplexing techniques to enhance the capacity crunch in recent communication links. The design parameters of the proposed fiber are predicted through machine learning-based inverse design approach, using regression model. The proposed few-mode fibre profile parameters are predicted with an accuracy of 99.95% to guide five to ten modes with weak coupling among the guided modes. In second phase of this work, the authors used a finite difference method-based solver to obtain the modal characteristics of the proposed fibre with predicted parameters for six guided modes, namely, LP01, LP11, LP21, LP31, LP41, and LP51. The numerical simulation results show that the predicted profile parameter maximizes effective mode-area and minimizes the inter-channel crosstalk for mode division multiplexing transmission over C-band. Besides this, the proposed ring-core few-mode fiber also exhibits nearly zero-dispersion for LP01 mode at 1550 nm along with low dispersion for other higher-order modes. Finally, an intensity-modulation and direct-detection mode multiplexed transmission link without erbium-doped fiber amplifier is established with six-spatial channels over 50 km and an attenuation of 0.18 dB/km to achieve minimum bit-error-rate of 4.45×10-9