Machine Learning-Aided High-Throughput First-Principles Calculations to Predict the Formation Energy of μ Phase

Area of Science:

• Materials Science
• Computational Materials Science
• Alloy Design

Background:

• The μ phase is a critical, yet brittle, component in high-temperature alloys.
• Accurate formation energy data is essential for designing advanced high-temperature alloys.
• Traditional first-principles calculations are computationally intensive and time-consuming.

Purpose of the Study:

• To develop an efficient and accurate machine learning (ML) model for predicting the formation energy of the μ phase.
• To reduce the computational burden associated with calculating the formation energy of the μ phase.
• To create a user-friendly tool for accessing μ phase formation energy data.

Main Methods:

• Utilized six machine learning algorithms and two evaluation methods.
• Trained models on a dataset of 1036 binary configurations using 10-fold cross-validation.
• Employed the multilayer perceptron (MLP) algorithm for optimal performance.

Main Results:

• Achieved a mean absolute error (MAE) of 23.906 meV/atom for binary configurations and 32.754 meV/atom for ternary configurations.
• Reduced computational time by at least 52% compared to traditional methods.
• Predicted lattice parameters (a and c) with low error rates (0.479% and 0.578%, respectively).

Conclusions:

• The developed ML model accurately predicts μ phase formation energy and lattice parameters.
• This approach significantly accelerates the material design process for high-temperature alloys.
• A graphical user interface (GUI) was created to improve data accessibility for researchers.