Accelerating First Principles Calculation of Multi-Component Alloy Steady-State Structure and Elastic Properties in...

Area of Science:

• Materials Science
• Computational Materials Science
• Alloy Design

Background:

• The FeNiCrAlCoCuTi alloy system offers excellent mechanical properties, but traditional experimental methods are inefficient for exploring its vast composition space.
• Designing multi-component alloys with specific properties requires understanding complex material characteristics-property relationships.

Purpose of the Study:

• To develop an accelerated research methodology for designing multi-component alloys with targeted properties.
• To demonstrate a combined first principles calculation and machine learning approach for predicting alloy properties.

Main Methods:

• Utilizing first principles calculations to generate property data for the FeNiCrAlCoCuTi alloy system.
• Employing machine learning models, optimized for each property (RMSE < 1.1), to predict alloy elastic properties.
• Performing interpretable analysis to understand feature-property relationships and spatial transformation for full-component prediction.

Main Results:

• Validated machine learning models against experimental data with a relative error below 5%.
• Successfully predicted full-component performance across binary to multiple components.
• Identified specific alloy compositions with desired Young's and shear moduli, such as Fe$_{0.23}$Cr$_{0.23}$Al$_{0.23}$Ni$_{0.03}$Cu$_{0.28}$ and Fe$_{0.01}$Cr$_{0.01}$Al$_{0.01}$Ni$_{0.44}$Co$_{0.53}$.

Conclusions:

• The integrated first principles and machine learning approach significantly accelerates alloy design and reduces costs compared to traditional methods.
• This methodology enables direct identification of optimal element compositions and intervals for superior alloy performance.
• The study provides a powerful framework for efficient exploration of complex multi-component alloy systems.