Neural evolution structure generation: High entropy alloys

Area of Science:

• Materials Science
• Computational Materials Science
• Alloy Design

Background:

• Generating high-entropy alloy (HEA) structures computationally is challenging due to the large number of atoms and complex configurations.
• Special Quasi-Random Structures (SQSs) are a common method but are computationally intensive and limited in scale.
• Developing efficient methods for generating large, representative HEA structures is crucial for materials discovery.

Purpose of the Study:

• To introduce a novel Neural Evolution Structure (NES) generation methodology.
• To enable the efficient and scalable generation of high-entropy alloy structures.
• To reduce the computational cost and time associated with generating large alloy structures.

Main Methods:

• Combining artificial neural networks and evolutionary algorithms for inverse design.
• Utilizing pair distribution functions and atomic properties as the basis for the model.
• Training the model on smaller unit cells to generate larger, representative structures.

Main Results:

• Achieved a speed-up factor of approximately 1000 compared to SQS methods.
• Enabled the generation of very large structures (over 40,000 atoms) within hours.
• Demonstrated the ability of a single model to generate multiple unique structures with the same composition, unlike SQSs.

Conclusions:

• The NES methodology offers a significant advancement in generating high-entropy alloy structures.
• This approach drastically reduces computational demands, making large-scale HEA structure generation feasible.
• NES provides a versatile and efficient tool for accelerating materials discovery in HEAs.