High Entropy Alloys Mined From Binary Phase Diagrams.

Area of Science:

• Materials Science
• Metallurgy
• Computational Materials Science

Background:

• High entropy alloys (HEAs) offer tailored properties due to vast compositional freedom.
• Current HEA phase prediction methods lack reliability, especially for single solid solution and composite phases.
• Compositional complexity in HEAs poses significant design challenges.

Purpose of the Study:

• To develop a novel phenomenological method for predicting HEA phases.
• To leverage binary phase diagrams and machine learning for accurate HEA phase prediction.
• To accelerate the design and discovery of high entropy alloys with desired properties.

Main Methods:

• Analysis of binary phase diagrams to derive phase-diagram inspired parameters.
• Application of machine learning (ML) models for classifying HEA phases based on these parameters.
• Validation of the predictive model on a diverse set of HEA compositions.

Main Results:

• The developed ML model achieved over 80% accuracy in predicting overall HEA phases.
• Single-phase HEA prediction accuracy also exceeded 80%.
• Validation demonstrated an 81% success rate in predicting solid solution phases in complex HEA compositions.

Conclusions:

• The phenomenological, phase-diagram-inspired approach provides reliable and detailed HEA phase predictions.
• This method offers a significant improvement over existing large-database statistical models.
• The approach can accelerate HEA design by complementing computation-intensive methods like CALPHAD.