Deep Learning Infrared Holography with Transformer for Crystal Material Characterization
Zijian Li, Haochong Huang, Dexin Sun, Zhiyuan Zheng, Fang Wang, Spozmai Panezai, Jie Xing, Yehua Yang, Kunfeng Qiu
Abstract
This investigation presents a novel approach for the nondestructive, and real-time analysis of crystalline structures, including transition metal dichalcogenides renowned for their optoelectronic capabilities. The methodology employs a synergistic blend of infrared digital holography and deep learning, utilizing an in-line system and Transformer-based deep learning algorithms, to provide detail in the material microstructure. The article investigates the effects of different parameters on reproduction fidelity, with a particular focus on phase accuracy. A holography-guided training strategy is proposed to enhance the framework’s performance. By demonstration applications such as evaluating the dielectric characteristics of ReS 2, detecting material thickness layers in MoS 2, and monitoring the microstructure evolution during the growth of NaCl and CuSO 4 crystals. Not only addresses existing limitations in material characterization but also offers avenues for exploration.