Force Field Methods and Representative Applications in Material Science
|School||Shanghai Jiaotong University|
|Keywords||Molecular Mechanics Force Field Metal Oxide Transport Properties Reactive Force Field Non-equilibrium Molecular Dynamics Monte Carlo Method|
This dissertation focuses on molecular mechanics force field methods and their representative applications of material science. Three types of force fields have been investigated. First, a novel force field for solid metal oxides and interfaces between the metal oxide solids and organic molecules has been developed. Second, the classic force fields have been applied to predict transport properties of liquids. Finally, the AI-REBO reactive force field has been investigated for modeling chemical reactions in bulk. Associated with the force field development, validation and application, there is a significant work focused on the simulation methods. Various molecular dynamics and Monte Carlo methods have been employed in this work. Among them, the reverse non-equilibrium molecular dynamics (RNEMD) and a Monte Carlo method using the AI-REBO reactive force field have been implemented. The major novelties and conclusions can be summarized as follows:1. A novel force field has been designed and parameterized for both metal oxide bulk and interface simulations. In this force field, partial charges instead of formal charges are used to represent the electrostatic interactions, and bond and non-bond interactions are described using different functions. Applications of the force field in bulk metal oxide show that the accuracy of the force field is at the same level of the classical previously developed force fields. More importantly, the new force field can be used together with general classical organics force field to simulate interfaces between the metal oxide and organic molecules. Simulations of methanol on MgO surfaces have been carried out using the constant pressure Gibbs ensemble Monte Carlo simulations. The results shed light on the mechanisms of the complex interfacial interactions and structures which are otherwise very difficult to reveal using experimental methods.2. In the predictions of shear viscosities of liquids using non-equilibrium molecular dynamics method, the calculated shear viscosities are found to be systematically underestimated by the periodic perturbation method. A perturbation index has been proposed to measure the perturbation extent, which is more accurate than the criterion used in the literature that overemphasizes on the periodic perturbation wavelength only. With careful analyses of the simulation results, the cause of the systematically underestimates has been identified. It has been identified that an acoustic wave of density distribution generated by the PPM causes the underestimates. Based on this finding a linear extrapolation scheme for predict shear viscosity at zero-perturbation is formulated. The method has been applied to more than 20 molecular liquids. The predictions are in good agreement with the experimental data.3. The transferability of classical force fields in prediction of shear viscosities for different molecules and for different states has been examined. It has been demonstrated that carefully derived force field based on ab initio data and static properties of liquid (such as density and enthalpy of vaporization) can be used to predict shear viscosities with average error less than 10%. The force field parameters are generally transferable among different thermodynamic states. However, transfers of force field parameters in different structures must be executed with caution. In calculations of the thermal conductivity with the RNEMD method, we found the quantum effects must be considered. A simple approach is to use the united atom model in which the fast vibrations are excluded.4. The applicability of the AI-REBO reactive force field for hydrocarbons has been assessed in a much broader range than those reported in the literature. It has been identified that the force field works reasonably well for predicting energies of molecules in stable configurations, but it does a poor job for predicting energies for highly distorted structures and transition states. It tends to overestimate the energy barrier heights. Therefore, this force field can be used to study the equilibrium concentration of reactants and products of chemical reactions. The force field needs to be revised extensively in order to work with kinetic consideration. 5. To use the potential of predicting products of the AI-REBO force field, a Monte Carlo procedure has been implemented to simulate the chemical reactions. This software has been tested on pyrolysis reactions of single-walled carbon nano-tube and cubanes. Preliminary results indicate that the program works reasonably well.