Machine-learning-assisted design and dehydration optimization of Er3 +/Yb3+ co-doped phosphate laser glasses for eye-safe emission

Abstract

Eye-safe solid-state lasers operating near 1.5 µm are essential for telecommunications, remote sensing, and lidar, yet their performance in phosphate glass hosts is often limited by hydroxyl-related losses and the trade-off between optical gain and thermo-mechanical stability. In this work, a new Er3+/Yb3+ co-doped metaphosphate laser glass is developed using a machine-learning-assisted compositional design strategy to optimize both spectroscopic efficiency and thermal robustness. A controlled dehydration process is implemented as a key processing step, reducing OH⁻ impurities to an absorption coefficient as low as 0.20 cm−1 at 3.33 µm, surpassing previously reported laboratory-scale glasses and commercial phosphate laser glass products. Systematic optimization of rare-earth dopant concentrations establishes clear composition–property relationships governing gain characteristics and laser performance. After dehydration, the optimized glass exhibits a high emission cross section of 6.8 × 10−21 cm2, a long fluorescence lifetime of 9.9 ms, and a broad positive optical gain spanning 1420–1620 nm, yielding optical gain comparable to commercial glasses while maintaining a high emission cross section-lifetime product of 67.2 × 10−21 cm2·ms. These spectroscopic properties are supported by excellent thermo-mechanical characteristics, including a low thermal expansion coefficient of 79.1 × 10−7 K−1, and a wide thermal stability window. Laser experiments confirm the benefits of dehydration, yielding a more than 2-fold reduction in the lasing threshold pump power in dehydrated samples. Overall, this work demonstrates a transferable strategy for designing high-gain, low-loss phosphate laser glasses for next-generation eye-safe laser and near-infrared amplifier applications. © 2026 Elsevier B.V.