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.