Deep Learning of Accurate Force Field of Ferroelectric HfO$_2$

Jing Wu,Yuzhi Zhang,Linfeng Zhang,Shi Liu

DOI: https://doi.org/10.1103/PhysRevB.103.024108

2020-10-30

Abstract:The discovery of ferroelectricity in HfO$_2$-based thin films opens up new opportunities for using this silicon-compatible ferroelectric to realize low-power logic circuits and high-density non-volatile memories. The functional performances of ferroelectrics are intimately related to their dynamic responses to external stimuli such as electric fields at finite temperatures. Molecular dynamics is an ideal technique for investigating dynamical processes on large length and time scales, though its applications to new materials is often hindered by the limited availability and accuracy of classical force fields. Here we present a deep neural network-based interatomic force field of HfO$_2$ learned from {\em ab initio} data using a concurrent learning procedure. The model potential is able to predict structural properties such as elastic constants, equation of states, phonon dispersion relationships, and phase transition barriers of various hafnia polymorphs with accuracy comparable with density functional theory calculations. The validity of this model potential is further confirmed by the reproduction of experimental sequences of temperature-driven ferroelectric-paraelectric phase transitions of HfO$_2$ with isobaric-isothermal ensemble molecular dynamics simulations. We suggest a general approach to extend the model potential of HfO$_2$ to related material systems including dopants and defects.

Materials Science

What problem does this paper attempt to address?

This paper focuses on how to develop an accurate HfO2 (hafnium oxide) force field using deep learning. HfO2 is a ferroelectric material with potential applications in silicon-based electronic devices, such as low-power logic circuits and non-volatile memory. However, the understanding of its dynamic response is still limited, especially regarding its phase transitions under different temperatures and external stimuli.

The paper introduces a deep neural network-based interatomic force field model that learns from first-principles data through concurrent training. This model can accurately predict various structural properties of HfO2, including elastic constants, equation of state, phonon dispersion relationships, and phase transition barriers, which are comparable to density functional theory calculations. Additionally, the simulation reproduces the ferroelectric to paraelectric phase transition of HfO2 observed in experiments.

The researchers emphasize the need for a tool to investigate the atomic-level dynamic properties of HfO2, considering its existence in multiple polymorphic forms and the complex factors affecting its ferroelectric properties, such as doping and defects. They successfully address this issue by employing the Deep Potential Molecular Dynamics (DeePMD) approach, which generates a model capable of describing the thermodynamic properties of HfO2 and can be extended to related material systems, including doping and defects.

In conclusion, the paper aims to address how to establish an accurate and transferable HfO2 force field model through deep learning to better understand and simulate the dynamic behavior of this material, especially in ferroelectric transitions.