Work

Revealing interaction mechanisms of substitutional and interstitial solutes in metals through machine learning

Year: 2026

Type: article

Abstract: A machine learning method was used to uncover interaction mechanisms between substitutional and interstitial solutes in metals, using binding energy as the indicator of interaction strength. By applying classification and regression algorithms, both the sign (indicating attraction or repulsion) and magnitude of binding energy were investigated. SHapley Additive exPlanations (SHAP) analysis reveals that the atomic volume of substitutional solutes is the most influential factor, followed by the number of unpaired electrons in the host metal and the electronegativity of substitutional solutes. SHAP-driven visualizations elucidate the decision-making process of the model, providing multi-scale insights from global patterns through local interactions down to individual sample predictions. This approach delivers intuitive, quantitative representations of feature contributions in both magnitude and direction, surpassing traditional analysis methods. Moreover, the model generates thousands of new binding energy predictions with minimal computational overhead, significantly enhancing computational efficiency.

Source: Transactions of Nonferrous Metals Society of China

DOI: 10.1016/s1003-6326(26)67090-2

Authors: Kang-ni HE, Rui WANG, Guo-zheng FENG, Liang CHEN, Xiang-shan KONG

Institution: Shandong University

Language: English

Topic: Machine Learning in Materials Science

Subfield: Materials Chemistry

Field: Materials Science

Domain: Physical Sciences

Open Access status: diamond

Funders: National Natural Science Foundation of China, Natural Science Foundation of Shandong Province, Key Technology Research and Development Program of Shandong, National Key Research and Development Program of China

Awards: G3782433543, G7100168729, G7789709983, G7874267904