Y. Zhou, Y. Zheng, S. Wang, S. Sayin, Z. Li, M. Zaghloul
George Washington University,
United States
Keywords: monolayer MoS2, deep neural network (DNN), optical microscope, 2D material Identification
Summary:
Two-dimensional (2D) materials are layered crystalline materials consisting of a single layer or multiple layers of atoms which have strong in-plane bonds but weak coupling van der Waals bonds between layers. The monolayer 2D materials usually exhibit unique electronic, transport and optical properties in contrast to multilayer or bulk material. Thus, a fast and reliable tool that using machine learning to identify monolayer 2D material on various substrates is presented in this paper. In this work, the monolayer MoS2, which is one of the novel 2D semiconductor materials in the Transitional Metal Dichalcogenide (TMDC) group, is chosen as target monolayer material. The monolayer MoS2 is fabricated on various substrates including Sapphire, Quartz, GaAs, and Si substrates coated with 100, 300, or 500nm SiO2 using Chemical Vapor Deposition (CVD). The red (R), green (G), and blue (B) values of the Monolayer MoS2 and the nearby substrate are read by Image J software. These R, G, B values can be used for identifying the monolayer material. However, the R, G, B values at different locations of a wafer can be easily affected by the light source of the microscope and the camera. Thus, we propose to use deep neural network to improve the accuracy and reliability. This method used in the identification of monolayer MoS2 on various substrates is fast, straightforward, cost-effective, and potentially to be used in large-scale sample production.