Speckle-based wavelength recognition using a defect-engineered multimode fiber

Accurate identification of light wavelengths with high spectral resolution is crucial for precision measurements, bio-imaging, metrology, and various other applications. In speckle-based wavelength recognition systems using multimode fibers (MMFs), spectral resolution can be enhanced by increasing the fiber length. However, longer fibers compromise system stability and reliability. In this paper, we propose a speckle-based wavelength recognition technique that employs defect-engineered multimode fibers prepared with a femtosecond laser. Our method introduces random defect arrays within standard multimode fibers to excite additional higher-order modes, thereby significantly improving spectral resolution. Specifically, incorporating 30 random defect arrays into a 5 cm multimode fiber results in a spectral resolution enhancement from approximately 250 pm to 100 pm. Further, integrating this technique with neural networks enables the classification of light into three bands (1500 nm, 1550 nm, and 1600 nm) with a prediction accuracy exceeding 99% on a dataset with 20 pm intervals. The fibers utilized in this study are characterized by their short length, compact size, stability, and ease of operation, making them well-suited for integration into miniaturized devices. The defect-engineered multimode fiber approach offers novel insights into highly integrated and reliable wavelength recognition systems.