Nanoscale Fluid Flow - Molecular Dynamics Simulations with Solid-Liquid Interfaces

摘要

For the cases of flows with very high shear rate, or in the case of dimensions on the order of nm, the continuum or Newtonian hypothesis bfeaks down. Molecular dynamics (MD) method emerges as a viable approach to address fluid flow problems. MD potentially can address the issues such as solidfluid interfaces/interactions arising in the nanoscale-regime. Furthermore, large scale simulations based on atomistic models may also address the transition from atomistic to continuum scale approaches. In this study, we run two different sets of MD simulations to elucidate these issues. First, we'analyze the size and time averaging effects by performing large size periodic boundary condition MD simulations toevaluate the convergence of average velocity, temperature, and density corresponding to the length of averaging time and size of spatial bins. Based on the convergence, we observe that the error in fluid properties is within an acceptable range in spite of the discrete behavior of the molecules. Second, most previous MD simulations of nano-scale, shear driven fluid flows use specular type walls, consisting of wall molecules with infinite mass. However, the thermal equilibrium at the wall-fluid interface is often neglected in the system. To consider the thermal equilibrium at the wall-fluid interface, we have used the crystal wall model with atomic bonding stiffness and assume that the wall interacts with the fluid molecules and exchanges momentum. Therefore, the net energy of fluid is conserved by exchanging energy with the wall-fluid interface. Hence, our model is a more accurate description of the real physics of the problem, without the need for an imaginary external heat bath connected to the fluid. Our model, with wall-fluid interactions matches the results of existing models. However, since our model implements the energy exchange through the wall-fluid interface, we observe the temperature jump at the wall-liquid interface, which is not possible with previous models.