Thermal Diffusion in Liquid Mixtures and Polymer Solutions by Molecular Dynamics Simulations
This thesis is focused on simulating the transport processes of heat and matter under a sufficiently weak temperature gradient where the system linearly responds. The systems we are interested in are binary isotropic liquids with no convection and no viscous flows. Four related transport coefficient...
| Summary: | This thesis is focused on simulating the transport processes of heat and matter under a sufficiently weak temperature gradient where the system linearly responds. The systems we are interested in are binary isotropic liquids with no convection and no viscous flows. Four related transport coefficients of the systems, the thermal conductivity of heat conduction, the diffusion coefficient of matter transfer, the Soret coefficient and the thermal diffusion coefficient of the cross effect between the heat and mass transfer, are investigated by the method of molecular dynamics. The reverse non-equilibrium molecular dynamics (RNEMD) method is the tool to compute the thermal conductivity and the Soret coefficient, while the equilibrium molecular dynamics is used to obtain the mutual diffusion coefficient which is needed for the calculation of the thermal diffusion coefficient. The influences of the simulation parameters are investigated in benzene, cyclohexane and their mixtures. These parameters include the perturbation intensity of the RNEMD method, the cutoff length, the system size, and the presence of a thermostat. These molecules are represented with all-atom models. The perturbation intensity has only a small impact on the thermal conductivity, while it affects the Soret coefficient significantly. Above a certain value, longer cutoff length does not yield substantial difference for the Soret coefficients. A system of several hundred to some thousand molecules and of several nanometers in length is sufficient to avoid size effects in the calculations of the thermal conductivity and the mutual diffusion coefficient. The presence of the Berendsen thermostat is harmless for the calculation of the thermal conductivity. The force field potentially affects the results largely. Two groups of force fields slightly different in the non-bonded parameters produce thermal conductivities for cyclohexane which differ by 30%, and lead to about 20% deviation for the Soret coefficient of a benzene-cyclohexane mixture. The degrees of freedom of the model are found to affect the thermal conductivity significantly. Eliminating the vibrational freedom of the aliphatic and aromatic hydrogens yields thermal conductivities closer to the experimental values. Most of the thermal conductivities we obtain have deviations at around 30-50% from the experimental values. Such deviations are quite common in the calculations of transport coefficients. Although the Soret coefficients were (3-5) × 10-3 K-1 larger than the experimental values, our simulation yielded the best results compared to previous simulations. Furthermore, our results reproduced the dependence of the mole fraction and the temperature of the Soret coefficients. Thermal diffusion in dilute polymer solutions has also been studied for the first time by the RNEMD method. The polymers are represented by a generic bead-spring model. The influence of the solvent quality on the Soret coefficient is investigated. At constant temperature and monomer fraction, a better solvent quality causes a higher affinity for the polymer to the cold region. This may even go to thermal-diffusion-induced phase separation. The experimentally known independence of the thermal diffusion coefficients of the molecular weight is reproduced for three groups of polymers with different chain stiffnesses. The thermal diffusion coefficients reach constant values at chain lengths of around 2-3 times the persistence length. Moreover, rigid polymers have higher Soret coefficients and thermal diffusion coefficients than more flexible polymers. Our simulations validate the applicability of the RNEMD methods to investigate heat conduction and matter transport. Potentially, the method can be extended to more systems to study the microscopic mechanisms of transport processes. |
|---|