Ab-initio calculations and ensemble learning prediction of the structural properties of solution combustion processed strontium-doped hydroxyapatite

Abstract

In this study, density functional theory (DFT) based ab initio calculations were performed to complement the experimentally obtained structural properties of strontium doped hydroxyapatite (Sr-HAp) prepared through the solution combustion synthesis (SCS) route. The structural properties of the as-prepared strontium ion doped powders were collected on an X-ray diffractometer and the structure-property relation of the datasets was explored using machine learning-assisted ensemble learning techniques such as CatBoost, LightGBM, Random Forest, and XGBoost. The results show that the Random Forest technique surpassed all other ensemble learning techniques, indicating better generalization capacity. The simulated XRD signatures agree well with the experiments.

Keywords

• Hydroxyapatite
• Machine learning
• DFT
• Solution combustion synthesis
• X-ray diffractometer