Nano Structure Data Collection for AI-Driven Alloy Design in Al-Cr-Fe-Ni-Ti Complex Concentrated Alloy System

Abstract

Finding suitable alloys for high-temperature applications is crucial in the fields of power generation and aerospace. Currently, nickel-based superalloys are used, but they are economically unfavorable, so researchers are studying other materials. However, finding the right elements and composition is challenging, so we are attempting to develop materials through artificial intelligence. This research primarily targets the collection of data to train this AI system. The development of complex concentration alloys involves the addition of various elements, primarily Al-Cr-Fe-Ni-Ti, which are expected to yield high strength at elevated temperatures. These alloys take the form of a matrix based on the BCC structure with L21 precipitates [1]. Consequently, data pertaining to these precipitates including size, shape, chemical composition and interfacial strain was collected. The samples were prepared using focused ion beam (FIB) milling from grains aligned with the (001) plane to optimize the visibility of precipitate morphology in TEM. Utilizing dark-field (DF) images obtained through transmission electron microscopy (TEM), we precisely determined the shape and size of these precipitates. DF TEM images offered superior clarity in distinguishing precipitates from the surrounding matrix, greatly enhancing our data collection process. To obtain precise information about their dimensions, we measured both the length and width of these rectangular-shaped precipitates, calculating deviations to ensure accuracy. We further investigated the chemical composition of the precipitates and the matrix using atom probe tomography (APT), enabling us to compute the volume fraction of the precipitates based on the lever rule. Through APT, information on composition could be obtained more accurately in nanoscale than any other analysis method. In order to measure the interfacial strain between the precipitates and the matrix, we utilized high-resolution scanning transmission electron microscopy (HRSTEM) method visualizing atomic lattice plane and precession electron diffraction technique combined with TopSPIN software (NanoMegas). As size of the precipitates and distance between them gets smaller, applying HRSTEM method became challenging due to complex formation of dislocations. Thus, we turned to utilizing TopSPIN method, which enabled us to measure internal strain and create 2-dimensional strain distribution maps. Consequently, we tried to elucidate correlations of the structure parameters with mechanical properties such as yield strength at room/high temperatures and toughness. By blending artificial intelligence with advanced material characterization techniques, it paves the way for the development of materials crucial to sustainable energy and aerospace technologies. The research emphasizes the vital role of comprehending microstructure and mechanical behavior in achieving these objectives.

[1] W.C. Kim, M.Y. Na, H.J. Kwon, Y.S. Na, J.W. Won, H.J. Chang, K.R. Lim, Acta Materialia, 2021, 211, 116890.