Water is an important substance on Earth. The dielectric constant of water is a crucial quantity affecting solubility of salts in water. However, it is experimentally challenging to study water under high pressure and high temperature. A typical method to study the dielectric constant of water under extreme conditions is molecular dynamics. Here, I applied the machine learning method combined with molecular dynamics to compute the dielectric properties of water. Our neural network model achieves both high accuracy and high computational efficiency. In addition, I performed first-principles calculations to study the elastic properties of ice VIII and ice X with different exchange-correlation functionals. The elastic constant could be an indicator for the phase transition between...[ Read more ]

Water is an important substance on Earth. The dielectric constant of water is a crucial quantity affecting solubility of salts in water. However, it is experimentally challenging to study water under high pressure and high temperature. A typical method to study the dielectric constant of water under extreme conditions is molecular dynamics. Here, I applied the machine learning method combined with molecular dynamics to compute the dielectric properties of water. Our neural network model achieves both high accuracy and high computational efficiency. In addition, I performed first-principles calculations to study the elastic properties of ice VIII and ice X with different exchange-correlation functionals. The elastic constant could be an indicator for the phase transition between ice VIII and ice X.