(123a) Model Development for the Gas-to-Liquid Process Based On Molecular Simulation and Experimental Measurements of Physical Properties

The Gas-to-Liquid (GTL) process is a highly promising technology for the production of high value fuels with a broad range of applications.  Recently, Shell launched the biggest GTL plant worldwide in Qatar which in full capacity will be processing 140,000 bpd.  GTL is based on Fischer-Tropsch synthesis where carbon monoxide hydrogenation takes place at elevated temperatures and pressures.  For optimum process design, various physical properties are required.  Some of the most critical properties include the diffusion coefficient of hydrogen (H₂), carbon monoxide (CO) and water (H₂O) in n-alkanes for the evaluation of mass transport process. 

In this work, Molecular Dynamics (MD) simulations were performed to predict the self-diffusion coefficient of the three gases in various n-alkanes, from n-C₈ up to n-C₉₆ and their mixtures, at temperature and pressure conditions similar to the actual GTL process.  Realistic atomistic force-fields were used to model gas and n-alkane intra- and intermolecular interactions.  The force-fields were initially validated for the prediction of pure component thermodynamic properties.  Subsequently, they were used for the prediction of gas diffusivity in pure n-alkanes and in binary and multicomponent n-alkane mixtures.  In the latter case, the mixture compositions were chosen so that they represent closely the actual GTL mixtures.  MD calculations are in excellent agreement with limited experimental data for the diffusion of gases in medium size n-alkanes.  In addition, the Maxwell-Stefan diffusion coefficient of H₂ and CO in n-C₁₂, n-C₂₀ and n-C₂₈ was calculated for different solute concentrations and used subsequently for the calculation of the Fick diffusion coefficient.

A high pressure Dynamic Light Scattering (DLC) unit was built in the University of Erlangen-Nuremberg to measure the diffusion coefficient of the gases in n-alkanes.  Experimental data for the diffusion coefficient of the three gases in n-C₁₂ and n-C₂₈ in a wide temperature range agree remarkably well with the MD predictions.

MD simulations are proven to be very accurate for property predictions but they are computationally demanding and thus cannot be used routinely for engineering calculations needed for process modeling. Consequently, an empirical phenomenological model was developed that correlates the diffusion coefficient of a gas with respect to other physical properties, such as the n-alkane viscosity.  The latter was also calculated based on MD simulations using the same force-field.  Calculations based on the empirical model are, on the average, within 10 % from MD simulations.

Finally, work is under way to predict the diffusion coefficient of gases in 1-alcohols which may be present in small quantities in the GTL process.