Thermo-diffusion is a coupled heat and mass transfer phenomena observed in Liquid mixtures. Enhancing Thermo-diffusion in liquids can improve the micromachining processes of silicon semiconductor manufacturing. Semiconductors are used in communication devices, military systems, clean energy and transportation among many other applications. Finding ways to improve semiconductor manufacturing will aid several industrial and research institutions. The figure shows the effect of thermo-diffusion, imagine a box with a liquid mixture of argon and krypton inside it, one end of the box is maintained at a temperature of 115K and the other end is at 95K, then the heavier molecules will have a preferential separation at the cold side of the box. This project aimed to apply the skills learned in the Molecular Simulation of Materials class to understand the mechanism of thermo-diffusion in liquid mixtures.  

In the literature review, I carried out, we realized that there is no knowledge of the underlying mechanism of Thermo-diffusion. We used the LAMMPS software package to model a mixture of Argon-Krypton liquid. Each simulation contains 1500 atoms interacting with the Lennard Jones potential, and the domain is partitioned into slabs of equal thickness along the z-direction. The architecture of the simulation cell is shown in the figure. We analyze the values of concentration and temperature in each slab to evaluate the concentration and temperature gradient to calculate the value of the Soret coefficient. We use this methodology to find the Soret coefficient of multiple different materials that interact with the LJ potential and understand the mechanism of Thermo-diffusion. 

I worked on setting up the simulations of different materials to calculate the Soret coefficient in LAMMPs. I created the different materials by changing the mass and the interaction energy or epsilon of LJ for the atoms in the simulation. I developed a Python script to further analyze the results of the simulation which are shown in the figure. From the results, we achieved excellent linearity in the temperature profiles but the concentration profile was highly sensitivity to changes in particle count. In the findings, we see an increase in the Soret coefficient with decreasing particle mobility (i.e with higher mass, epsilon)