Deep Learning-Based Multimodal Analysis for Transition Metal Dichalcogenides

Deep learning (DL) algorithms use a high generalization approach to recognize and interpret images, enabling an efficient identification of properties of materials^1–4. Electronic properties, such as band gaps and electron affinities, have been predicted using machine learning (ML) and DL models based on the structure-property relationship^5–7. This study presents a novel approach to enable high-throughput characterization of transition metal dichalcogenides (TMDs) across various layers, including mono-, bi-, tri-, four, and multilayers. We demonstrate a DL-based image-to-image translation approach with generative conditional adversarial networks, trained with optical labeled images to enable intelligent characterization of mechanically exfoliated and CVD-grown TMDs. Unlike existing AI-based approaches on TMDs, in this method, the DL-based pix2pix cGAN network was trained using a small set of labeled optical images, translating optical images of TMDs into labeled images that map each layer with a specific color and give a visual representation of the number of layers. Multiple segmentation techniques were implemented to extract graphical features from the optical images, including contrast, color, shapes, flake sizes, and their distributions. The trained model was adapted to characterize other 2D materials not initially included in the dataset. To assess the performance of the model, we conducted quantitative measurements using structural similarity index measure (SSIM), peak signal-to-noise ratio (PSNR) and mean square error (MSE) scores. We further trained it on MoS₂ and WS₂ samples and successfully tested it on WSe₂ samples to demonstrate its capability to adapt to different materials. Lastly, we applied the model to characterize heterostructures, highlighting its ability to analyze complex material structures. Our model is solely based on optical images, capturing complex variations, categorizing layers into five different classes and demonstrating adaptability across a diverse range of materials.<br/><br/><b>References</b><br/>1. Han, B. <i>et al.</i> Deep Learning Enabled Fast Optical Identification and Characterization of 2D Materials. <i>Adv. Mater.</i> <b>32</b>, 2000953 (2020).<br/>2. Masubuchi, S. <i>et al.</i> Deep-learning-based image segmentation integrated with optical microscopy for automatically searching for two-dimensional materials. <i>Npj 2D Mater. Appl.</i> <b>4</b>, 3 (2020).<br/>3. Yang, J. & Yao, H. Automated identification and characterization of two-dimensional materials via machine learning-based processing of optical microscope images. <i>Extreme Mech. Lett.</i> <b>39</b>, 100771 (2020).<br/>4. Sanchez-Juarez, J., Granados-Baez, M., Aguilar-Lasserre, A. A. & Cardenas, J. Automated system for the detection of 2D materials using digital image processing and deep learning. <i>Opt. Mater. Express</i> <b>12</b>, 1856 (2022).<br/>5. Alibagheri, E., Mortazavi, B. & Rabczuk, T. Predicting the Electronic and Structural Properties of Two-Dimensional Materials Using Machine Learning. <i>Comput. Mater. Contin.</i> <b>67</b>, 1287–1300 (2021).<br/>6. Bhattacharya, A., Timokhin, I., Chatterjee, R., Yang, Q. & Mishchenko, A. Machine learning approach to genome of two-dimensional materials with flat electronic bands. <i>Npj Comput. Mater.</i> <b>9</b>, 101 (2023).<br/>7. Cheng, Z. <i>et al.</i> 2D Materials Enabled Next-Generation Integrated Optoelectronics: from Fabrication to Applications. <i>Adv. Sci.</i> <b>8</b>, 2003834 (2021).