An informatics guided classification of miscible and immiscible binary alloy systems

Area of Science:

• Materials Science
• Computational Materials Science
• Chemical Physics

Background:

• Alloy classification is crucial for designing new materials.
• Existing methods rely on extensive experimental data and theoretical models.

Purpose of the Study:

• To develop a comprehensive classification of alloying behavior for binary alloys.
• To predict the miscibility of unknown alloy systems using data mining and machine learning.

Main Methods:

• Mining experimental phase diagrams and thermodynamic data.
• Utilizing physics-based descriptors, including a modified Pettifor chemical scale.
• Applying artificial neural networks and elemental similarity for prediction.

Main Results:

• A 2D map based on the Pettifor scale effectively separates miscible and immiscible alloy systems.
• High agreement (95%) with Miedema's theory and good agreement (90%) with high-throughput first-principles (HTFP) calculations.
• A complete miscibility map for 813 binary alloy systems of transition and lanthanide metals was generated.

Conclusions:

• Physics-guided data mining offers an efficient approach for materials discovery.
• The developed miscibility map aids in understanding and predicting alloying behavior.
• This methodology facilitates the design of next-generation multicomponent alloys.