Reverse Nonequilibrium Molecular Dynamics Calculation of the Soret Coefficient in Liquid Heptane/Benzene Mixtures

Abstract

We studied the thermal diffusion behavior of mixtures of benzene and heptane isomers by reverse nonequilibrium molecular dynamics. For n-heptane/benzene mixtures we investigated the concentration dependence of the Soret coefficient. The Soret coefficient for equimolar mixtures of the three heptane isomers 3-methylhexane, 2,3-dimethylpentane and 2,4-dimethylpentane in benzene has been calculated. Compared to the experimental data, the simulation results show the same trend in dependence of the mole fraction and degree of branching. The negative Soret coefficient indicates the enrichment of alkanes in the cold side. In the case of the heptane isomers in benzene we could study the influence of the difference in shape and size on the thermal diffusion behavior at constant mass. In the simulation as well as in the experiment we found that the Soret coefficients becomes higher with increasing degree of branching. Such behavior can not be explained only by mass and size effects. The effect of the molecular shape needs to be considered additionaly.

Non-Equilibrium Molecular Dynamics Used to Obtain Soret Coefficients of Binary Hydrocarbon Mixtures

“…Comparing with results obtained with other force fields (anisotropic united atoms -AUA - Toxvaerd, 1990), the self-diffusion coefficients obtained here are 25% higher for n-pentane and 42% higher for n-decane. Polyakov et al (2008) also observed values 20% higher or more than experimental data for self-diffusion coefficients using the TraPPE-UA force field. The same authors found a systematic deviation of 40% for the mutual diffusion coefficients for heptane-benzene mixtures.…”

“…Despite these advantages, several characteristics are required for the method to be used properly. Because the simulation box is divided along one direction into several (commonly 20) slabs, the number of atoms inside each slab must be large enough for it to behave as a thermodynamic system (Hafskjold, et al, 1993;Hafskjold and Ratkje, 1995;Zhang and Müller-Plathe, 2005;Polyakov et al, 2008). Another problem is related to the relatively large temperature gradients needed to obtain a suitable estimate of the Soret coefficient.…”

“…Other authors employed the BD-NEMD method to determine thermal diffusion factors and/or Soret coefficients by using a weakly coupled thermostat to avoid the energy drift (Reith and Müller-Plathe, 2000;Zhang and Müller-Plathe, 2005;Polyakov et al, 2008"). In addition, other methods that use two independent thermostats coupled to different regions of the simulation box may be used to generate the temperature gradient, also avoiding the energy drift problem (Maier et al, 2011;Maier et al, 2012).…”

“…To take into account the influence of a thermostat on the Soret coefficient determination, the method was applied with a Berendsen thermostat with a weak-coupling time of 50 ps for equimolar systems with 1200 molecules and time step equal to 2 fs. This weak coupling factor was chosen following other studies in the literature (Zhang and Müller-Brazilian Journal of Chemical Engineering Plathe, 2008;Polyakov et al 2008) in order to minimize the influence of the thermostat on the system. We evaluated whether the addition of a thermostat, even with a weak coupling factor, can change the dynamics of the system in order to affect the determined Soret coefficient.…”

“…The exchange periods of thermostated systems were 300 and 400 fs. The time step of 2 fs was chosen following studies published in the literature that used 2 fs or higher time steps (Simon et al, 1998;Zhang and Müller-Plathe, 2005;Polyakov et al, 2008). To avoid energy local minima, the systems were heated and cooled in the same way described for self-diffusion coefficient determinations.…”