Microchannel cooling, in which a coolant flows through a microchannel, is a efficient met-hod for heat removal. The heat removal can be improved by having the coolant evaporate inside the microchannel. Although it is clear that this method can achieve large local cooling, the process is not fully understood yet. CFD descriptions fail at the small length scales, but particle-based method still work. Two of the most important particle-based techniques are Direct Simulation Monte Carlo (DSMC) and Molecular Dynamics (MD). DSMC is computationally faster, but is less accurate and only works for gases. MD is more time-consuming, but also works in the liquid and solid phase. In the simulation of microchannels, the channel walls can also be simulated explicitly with MD, which in-creases the accuracy. The methods (CFD, DSMC and MD) are analyzed on their weak-nesses and strengths, and compared in three standard problems: a Poiseuille flow, a gas with a temperature gradient between two walls, and lubrication flow around a heated obstacle. The results show that the modelling of the fluid-wall interaction crucial for the overall behavior. In the simulation of evaporation inside microchannels, the treatment of fluid-wall heat transfer is the most important. In gases close to walls, density fluctuations occur, which have a large effect on the heat transfer between wall and gas; these fluctuations have been analyzed in more detail. For heat transfer between a micro channel wall and the coolant, the explicit wall model can be used in MD with great accuracy, but also with great computational costs. Other wall models that are computationally cheaper exist, but they are less accurate, or have parameters that are unknown a priori. To overcome this problem, a new model was introduced, based on a vibrating wall model, that cuts back on computation time but has an accuracy comparable to the explicit wall model. Evaporation can be modeled in MD, but to be sure that the simulations give the right result, it has to be validated with experimental results. First, the intermolecular interaction was validated, by looking at vaporation of Argon. Then, the bonds were validated, by analyzing Oxygen evaporation. In this way, more complex molecules consisting of atoms with internal bonds have also been checked. The two problems of fluid-wall interaction and evaporation come together at the microregion. This microregion is important in the analysis of a microchannel, as the most heat is transferred here. To validate the MD simulation of this microregion, simulation results are compared to the results of a model of this region based on a continuous description. With the models now available, a complete microchannel with evaporative cooling can be simulated. However, it is computationally too expensive to useMDfor the full domain, so MD is only used for the most relevant parts where CFD methods are inaccurate (close to the walls, at the evaporation interface and in the microregion), and CFD methods are used for the bulk.
|Qualification||Doctor of Philosophy|
|Award date||26 May 2010|
|Place of Publication||Eindhoven|
|Publication status||Published - 2010|