There is a great need for metallic alloys with exceptional mechanical properties at high temperatures for various applications in different industries such as aerospace, power generation, chemical processing, etc. With the development of new technologies, metallic materials that can operate at even higher temperatures than current Ni-based superalloys are required. Refractory high entropy alloys (RHEAs) have emerged as a very promising group of materials for these applications, owing to their high melting points and strength at elevated temperatures, making them suitable for use in extreme environments.

We will show the results of the systematic exploration of RHEAs' hyperspace from the Cr-Mo-Nb-Ta-V-W system. For this study, we synthesized a material library (ML) via physical vapor deposition (PVD) technology on a single silicon wafer with a gradient of each element in the range of 30-45 at.%. Assuming that the alloys should differ from each other by a total of 1 at. %, this translates to ~35,000 alloys. It would take ~136 years to produce the same amount of alloys at a rate of 1 alloy/day. The magnetron co-sputtering process was calibrated in such a way as to obtain an equimolar composition and maximum configurational entropy in the center of the ML (Fig. 1).

In the first stage of the research, selected ML regions were characterized by X-ray fluorescence to determine the chemistry. Structure studies were performed using X-ray diffraction. The obtained diffractograms were subjected to a detailed analysis using the parametric Le Bail refinement, which is a whole diffraction pattern profile fitting technique that allows the determination of the properties of crystalline materials, such as lattice parameters and crystallite size. High-throughput nanoindentation was then used to determine the mechanical properties. The material database created in this way was used to train the artificial neural network algorithm (ANN) to predict the mechanical properties of alloys that extend beyond the compositional space of the tested ML.