Development and Validation of Neural Network Potentials for Multicomponent Oxide Glasses

Takahiro Ohkubo,Yaohiro Inagaki,Ryuta Matsubara,Keisuke Ishida,Ryuki Kayano

DOI: https://doi.org/10.26434/chemrxiv-2024-hb50k

2024-07-02

Abstract:A neural network potential (MLP) for molecular dynamics simulation (MD) of multicomponent oxide glasses was developed with special consideration of structural reproducibility. The MLP was constructed through pre-training using the dataset of density functional theory (DFT) data provided by the Open Catalysis project and fine-tuning for the dataset of nine-component glasses. A thorough validation based on previous experimental and DFT-MD data was performed for the glass structure derived from neural network molecular dynamics simulation ({\MLMD}) with the developed {\MLP}. The accuracy of {\MLMD} was investigated in terms of the local structure of the glass, comparing it to the glass derived from MD with conventional potentials. The composition dependence of the local structure in \ce{Na2O}--\ce{SiO2} and \ce{Na2O}--\ce{B2O3} glass systems was well reproduced for the {\MLMD}-derived glass. The ability to reproduce the glass structure was demonstrated in the population of four-coordinated boron population and formation of superstructures in alkali borate glasses, and the Al local structure in the novel Al-rich binary aluminoborosilicate glass. The importance of pre-training was revealed by comparing {\MLMD} results using {\MLP} developed with and without pre-training. Although better metric scores for {\MLP} without pre-training can be achieved, the resultant structure was not realistic. This is an important lesson that the metric score alone is inadequate to construct accurate {\MLP} for glasses. Finally, the developed {\MLMD} was applied to the modeling of the reference nuclear waste glass (60.1{\Si}--3.84{\Al}--15.97{\B}--12.65{\Na}--2.87{\Ca}--2.86{\Mg}--1.72{\Zr}), and the charge compensation mechanism of the cations was discussed.

Chemistry

What problem does this paper attempt to address?

This paper focuses on the development and validation of neural network potentials (NNPs) for molecular dynamics simulations of multi-component oxide glasses. The researchers specifically consider structure reproducibility by pre-training the NNP using density functional theory (DFT) data sets provided by the Open Catalyst Project and fine-tuning it with data from nine-component glasses. They extensively validate the NNP by studying the local structure of glass structures simulated by NNMD based on previous experiments and DFT-MD data.

The paper mentions that although unpretrained NNPs can achieve better scores on certain metrics, the generated structures are not realistic. This indicates that relying solely on metric scores is insufficient for constructing accurate NNPs, especially for glasses. Additionally, the researchers explore the ability of the NNP to predict superstructures in tetracoordinate borate and alkali borate glasses, as well as aluminum local structures in novel aluminum borosilicate glasses. Finally, they apply the developed NNMD to model reference nuclear waste glasses and discuss the charge compensation mechanism of cations.

The paper points out that while the applicability of NNPs to simple glass systems has been widely acknowledged, modeling routes for complex glasses containing five or more components still pose a challenge. Existing potential parameters may not ensure applicability across all composition ranges, and CMD simulations may not be accurate enough for predicting complex multi-component systems. Therefore, the researchers propose a method of developing NNPs using large-scale DFT pre-training databases to simulate multi-component oxides and achieve more accurate results through NNMD.

In conclusion, this paper aims to address the accuracy issue in simulating multi-component oxide glass structures by developing and optimizing neural network potentials to improve simulation accuracy and applying them to the study of complex materials such as nuclear waste glasses.