Near-infrared speckle wavemeter based on nonlinear frequency conversion

Abstract

The wavemeter is an important instrument for spectrum analysis, widely used in spectral calibration, remote sensing, atomic physics, and high-precision metrology. However, near-infrared (NIR) wavemeters require infrared-sensitive detectors that are expensive and less sensitive compared to silicon-based visible light detectors. To circumvent these limitations, we propose an NIR speckle wavemeter based on nonlinear frequency conversion. We combine a scattering medium and the deep learning technique to invert the nonlinear mapping of the NIR wavelength and speckles in the visible wave band. With the outstanding performance of deep learning, a high-precision wavelength resolution of 1 pm is achievable in our experiment. We further demonstrate the robustness of our system and show that the recognition of power parameters and multi-spectral lines is also feasible. The proposed method offers a convenient and flexible way to measure NIR light, and it offers the possibility of cost reduction in miniaturized wavemeter systems.