Nd³⁺-doped phosphate fibers for 0.9 μm single-frequency lasers: machine learning-driven compositional design for suppressing competitive emission

The 0.9 μm single-frequency fiber laser has emerged as a promising source for doubling frequency to generate blue lasers, which are essential for applications in underwater communication, detection, and laser display. However, the ⁴F_3/2→⁴I_9/2 transition at 0.9 μm encounters significant competition from the four-level structure transition (⁴F_3/2→⁴I_11/2). Designing a fiber matrix to mitigate competition presents an effective strategy without compromising device integration or performance. To address this issue, a machine learning approach is applied to develop the composition-structure descriptor-branching ratio model in laser glasses, which demonstrates high accuracy and showcases impressive extrapolation capabilities from ternary to quaternary glass systems. The phosphate fiber design features an enhanced branching ratio at 0.9 μm (0.40) and a high net gain coefficient (4.21 dB/cm at 915 nm). A compact single-frequency laser at 915 nm (fiber length of 8 mm) with an environmentally friendly pump threshold of 11.5 mW and suppressed competitive emission (signal-to-noise ratio > 65 dB) has been achieved. This methodology holds significant potential for guiding the compositional design of novel fibers targeting other performance metrics.