Research on High-throughput Characterization Method of Materials Microstructure Based on Glow Discharge Sputtering Sample Preparation

Yu Xing^1,2*, Wan Weihao^1,2, Shen Xuejing^1,2, Wang Haizhou^1,2

¹ Central Iron and Steel Research Institute, Beijing, 100081, China；

² Beijing Key Laboratory of Metal Materials Characterization

ABSTRACT: The microstructure of materials is one of the four major concerns by materials science and engineering. Its observation and quantitative characterization are an important part of materials research, and it is also a precondition for understanding of the relationship between microstructure and material properties. The actual materials, especially the most widely used structural materials, are all non-uniform, multiple, and complex, and the composition and structure of each spatial location are not the same, resulting in different performance at different scales. The essential properties of materials heterogeneity require a large range of quantitative statistical distribution characterization of the materials, so that the conclusions of the structure characterization of the tissue can be more consistent with the actual situation of the materials. Before the characterization of the microstructure of the materials, the surface treatment of the sample is very important, and the surface of the treated samples must truly reflect the microstructure of materials. At present, the sample surface treatment procedure is usually complicated and time-consuming, and dangerous chemicals are widely used for polishing and corrosion. On the other hand, accelerated ions such as focused ion beam (FIB) are used to etch a sample, which has disadvantages such as expensive equipment, small sample preparation area, long time, and the formation of amorphous layer. In addition, existing methods for characterization of microstructure of materials, such as scanning electron microscope (SEM), are characterized by small region and slow speed, and random and qualitative observations on the material observation surface are difficult to obtain the overall quantitative and statistical information of the whole surface. In this paper, glow discharge (GD) sputtering was used to prepare samples for the characterization of microstructure of materials. In glow discharge, ions impact the sample surface at a wide angle and in a large range to achieve uniform sputtering in a large area on the sample surface. Compared with other ion beam methods, the ion energy of glow discharge is low and does not cause changes in the structure of the materials. It can gently but quickly remove scratches and deformed layers which are caused by polishing of the sample surface within a few minutes, and show the materials microstructure. It realizes the preparation of large area samples of mm-cm level. The conditions of the glow discharge sputtering preparation were optimized from the perspective of the influence on the surface flatness and the structure of the materials. It was found that the glow discharge voltage of 1000V and the discharge current of 5mA were suitable, and the single crystal superalloy samples were prepared. High-throughput scanning electron microscope was used to acquire large-scale, fast and full-scale microstructure images of samples prepared by glow discharge sputtering. Based on full electronic detection technology and a precise mobile platform, the scanning electron microscope can collect images with high efficiency, high speed and high resolution, which is more than 50 times the speed of ordinary commercial scanning electron microscopes. By optimizing of the image acquisition conditions of the high-throughput scanning electron microscope, the emission voltage of the electron gun and the magnification are selected to be 7 kV and 20 kX. Tens of thousands of images with the size of 4096 pixels and the resolution of 3.54 nm have been collected successfully for the γ' phase structure of single crystal superalloy samples. For the collected massive organization structure images, the features of the original images are extracted by U-NET image segmentation algorithm which is based on deep learning. The processing speed can be improved by taking advantage of GPU's powerful parallel processing power. The relevant characteristic parameters of γ' phase and γ matrix were obtained quickly, and the quantitative statistics and distribution characterization of all feature parameters on a large-size area are performed. The establishment of this method is of great significance for the large-scale acquisition, identification and quantitative characterization of the materials microstructure. The characteristics of glow discharge layer by layer sputtering and the rapid image acquisition by high-throughput scanning electron microscope make it possible to carry out large-scale three-dimensional reconstruction of materials.

Keywords:Glow discharge sputtering; sample preparation; microstructure; high-throughput characterization