Phase transition mechanism and property prediction of hafnium oxide-based antiferroelectric materials revealed by artificial intelligence

Abstract

Constrained by the inefficiency of traditional trial-and-error methods, especially when dealing with thousands of candidate materials, the swift discovery of materials with specific properties remains a central challenge in contemporary materials research. This study employed an artificial intelligence-driven materials design framework for identifying dopants that impart antiferroelectric properties to HfO₂ materials. This strategy integrates density functional theory (DFT) with machine learning (ML) techniques to swiftly screen HfO₂ materials exhibiting stable antiferroelectric properties based on the critical electric field. This approach aims to overcome the high cost and lengthy cycles associated with traditional trial-and-error and experimental methods. Among 30 undeveloped dopants, four candidate dopants demonstrating stable antiferroelectric properties were identified. Subsequent DFT analysis highlighted the Ga dopant, which displayed favorable characteristics such as a small volume change, minimal lattice deformation, and a low critical electric field after incorporation into hafnium oxide. These findings suggest the potential for stable antiferroelectric performance. Essentially, we established a correlation between the physical characteristics of hafnium oxide dopants and their antiferroelectric performance. The approach facilitates large-scale ML predictions, rendering it applicable to a broad spectrum of functional material designs.