Dynamical Stability of Ce-based Intermetallic Ternary Compounds Predicted By Machine-Learning-Guided Studies

Ce-based intermetallic compounds are interesting as they can vary in valence from the non-magnetic Ce⁴⁺ to magnetic Ce³⁺ and form heavy fermion and mixed valent materials with exotic quantum criticality, superconductivity and magnetism. Recently, several new Ce-based ternary intermetallic compounds containing transition metal (Fe, Co) and other metallic elements (Cu, Bi, Pb, In, Ag) were predicted to be stable using an energy-based crystal graph convolutional neural network (CGCNN) machine learning model and first-principles density functional theory (DFT) calculations. In this work, we further investigate the dynamical stability of these newly predicted structures. Phonon calculations by first-principles DFT at zero temperature using harmonic approximations show that majority of these predicted structures exhibit imaginary vibrational modes. However, ab initio molecular dynamics (AIMD) simulations show that some structures are stable at a finite temperature of 500K. Vibrational density-of-states and phonon modes at high symmetry k-points calculated at finite temperature through the AIMD simulations suggest that anharmonic interactions may play an important role in stabilizing these Ce-based intermetallic compounds.