Insight into the stability in cation substitution of Magneli phase Ti4O7

Doping Magneli phase Ti4O7 by cation substitution has attracted some interest for modulating structure and properties enhancement, but it remains a big problem to understand how doping elements impact the thermodynamic and structural stability of Ti4O7. We utilized first-principles calculations based on density functional theory (DFT) combined with machine learning (ML) to forecast the stability of doped Ti4O7. DFT calculations are used to model the thermodynamic and structural stability, as well as the electronic structure, of doped (Ti,M)(4)O-7 complexes (M = Sc, Y, La, Ce, Zr, Hf, V, Nb, Ta, Cr, Mo, and W). The results reveal that even if all (Ti,M)(4)O-7 are thermodynamically stable, the introduction of rare earth elements Y, La, and Ce causes great structural distortion. Employing Zr, Nb, Mo, and W can improve Ti4O7 thermodynamic stability due to strong bond strength and minimal lattice distortion. The relevance of 78 doping element qualities and one processing feature (doping site) for (Ti,M)(4)O-7 stability is discovered using ML. The results show that modulus of rigidity and entropy of solid of doping atoms have the greatest influence on the thermodynamic and structural stability of doped Ti4O7, which is useful for predicting additional (Ti,M)(4)O-7 stability without DFT calculations. At a low doping concentration, Ce-doped Ti4O7 with massive lattice distortion was synthesized, supporting the DFT results. This study not only applies to all doped Ti4O7 complexes, setting the groundwork for stability of the planned high-performance cation substitution in defect Ti4O7, but also introduces a unique way of predicting stability in defect engineering.

Automated summary

In this study, the impact of doping on the thermodynamic and structural stability of Ti4O7 was investigated using density functional theory and machine learning. The results showed that certain rare earth elements and transition metals can improve the stability of Ti4O7, and machine learning can be used to predict the stability of doped materials, providing a new approach for defect engineering.