Vapor-liquid equilibria of model alkanes

Abstract

In this communication, the results of computer simulations of the vapor-liquid equilibria of n-pentane and n-octane are reported. We have used a novel Monte Carlo technique to simulate directly phase equilibria of chain molecules. The calculations were performed for a united-atom model using the Lennard-Jones parameter set of Jorgensen and co-workers.' Comparison of the calculated phase diagrams with experimental data shows very good agreement, in particular for the shorter alkane. Computer simulations have contributed to the general understanding of the behavior of chain molecules. The success of a simulation depends to a large extent on the judicious choice of molecular potentials. An example of a popular potential field is the OPLS model (optimized potentials for liquid simulations) which has been used to study systems ranging from simple alkanes'J and surf act ants'^^ to polymers5 and proteins.6 It is important to realize that the OPLS model has been fitted to liquid properties of small molecules at room temperature. In practice, however, these models are used beyond the conditions at which they have been fitted. An ideal test of model potentials is the comparison with experimental phase diagrams. Unfortunately, determining a phase diagram is an extremely difficult task,' in particular, for chain molecules. Over the last few years there has been an impressive improvement in the development of simulation techniques for phase equilibria. Most noticeable is the introduction of the GibbsEnsemble Monte Carlo (GEMC) technique by Panagiotopoulos.*.9 In the GEMC technique two separate boxes are utilized. Besides the conventional random displacement of particles, additional types of Monte Carlo moves are applied to ensure that the two boxes are in equilibrium, Le., have equal temperature, pressure, and chemical potential. The particular advantage of the Gibbs ensemble is that if conditions are such that the system wants to phase separate, the simulation yields a vapor phase in one box and a liquid phase in the other. As a result, the coexistence properties can be determined directly. One of the Monte Carlo steps involves the swapping of a molecule from the vapor to the liquid box and vice uersa. While this step is relatively easy for systems containing atoms or small molecules, it is virtually impossible to insert chain molecules into their liquid phase using the conventional techniques. To enhance the sampling of chain insertions we have used a novel technique based on the configurational-bias Monte Carlo method (CBMC).loJ1 The random insertion is replaced with a scheme in which a chain molecule is inserted atom by atom such that conformations with favorable