Design of High-Capacity Multimode Erbium-Doped Optical Fibers Based on Neural Network

In this work, we present a multimode erbiumdoped fiber (MM-EDF) design scheme optimized by neural network, which effectively reduces inter-mode-group crosstalk and achieves high gain and low differential mode gain (DMG). The designed fiber supports the transmission of up to 100 modes across 10 mode groups, and by optimizing the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\Delta\mathbf{n}_{\mathbf{eff}}$</tex> within and between groups, we significantly lower inter-mode-group crosstalk. The demonstrated erbium ion distribution structure we designed shows a high degree of overlap with the mode distribution, achieving mode gains greater than 29 dB and DMG less than 1.38 dB. This design holds promise for low-crosstalk, high-gain, large-capacity online transmission systems.