Stability Assessment of Mbene Materials Using Scikit-Learn Inspired Application Programming Interface: Gplearn and Particle Swarm Optimization
Abstract
The stability of materials is paramount, serving as a foundational element for the successful deployment and operation of energy storage systems. It influences reliability, safety, efficiency, cost-effectiveness, environmental sustainability, regulatory compliance, and overall system performance. MBenes are a class of 2D materials that have the general formula MxBy , where M represents a transition metal (Co, Cr, Fe, Hf, Ir, Mn, Mo, Nb, Ni, Os, Pd, Sc, Ta, Tc, Ti, V, W, Y, Zr), and B represents Boron with remarkable electrical, mechanical, and electrochemical properties, require stability assessment to determine their suitability for diverse applications such as catalysis and battery electrodes. In this study, we have used Gplearn, a Scikit-learn-inspired API, to investigate the stability of MBenes with the chemical formula of the form M2B1, M2B2, with and without doping Pd, Co, achieving an accuracy of 0.97, Area Under the Curve (AUC) value of 0.93 from the Receiver Operating Characteristic (ROC) analysis, and a near perfectly fitting calibration curve, surpassing the AUC values obtained from conventional classification algorithms such as Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Gaussian Naïve Bayes (GNB), Gradient Boosting, Voting, Linear Discriminant Analysis (LDA), and Bagging Classifiers. This study has showcased the applicability of genetic programming through Gplearn in materials science, particularly for solving classification problems such as determining the material's suitability as a battery electrode or an electrocatalyst.