Homogeneous water vapor condensation with a deep neural network potential model.
J Chem Phys
; 160(12)2024 Mar 28.
Article
em En
| MEDLINE
| ID: mdl-38516980
ABSTRACT
Molecular-level nucleation has not been clearly understood due to the complexity of multi-body potentials and the stochastic, rare nature of the process. This work utilizes molecular dynamics (MD) simulations, incorporating a first-principles-based deep neural network (DNN) potential model, to investigate homogeneous water vapor condensation. The nucleation rates and critical nucleus sizes predicted by the DNN model are compared against commonly used semi-empirical models, namely extended simple point charge (SPC/E), TIP4P, and OPC, in addition to classical nucleation theory (CNT). The nucleation rates from the DNN model are comparable with those from the OPC model yet surpass the rates from the SPC/E and TIP4P models, a discrepancy that could mainly arise from the overestimated bulk free energy by SPC/E and TIP4P. The surface free energy predicted by CNT is lower than that in MD simulations, while its bulk free energy is higher than that in MD simulations, irrespective of the potential model used. Further analysis of cluster properties with the DNN model unveils pronounced variations of O-H bond length and H-O-H bond angle, along with averaged bond lengths and angles that are enlarged during embryonic cluster formation. Properties such as cluster surface free energy and liquid-to-vapor density transition profiles exhibit significant deviations from CNT assumptions.
Texto completo:
1
Coleções:
01-internacional
Base de dados:
MEDLINE
Idioma:
En
Revista:
J Chem Phys
/
J. chem. phys
/
Journal of chemical physics
Ano de publicação:
2024
Tipo de documento:
Article
País de afiliação:
China