The inputs to all molecular simulations involving adsorption are force field / potential parameters (σ, ε/k) between gas and solid, to calculate pair wise potential. Depending on the nature of adsorbent/gas involved and the complexity of the model used in simulations some or all of terms for dispersion, electrostatic interactions, induction, quadrupole etc. are included. In the simplest form, only dispersion interactions are considered (by lumping induced interactions into the dispersion term, if necessary) between pairs of molecules. Pair-wise additivity is assumed to obtain the total potential of the system.

Usually, these parameters are estimated from experimental data of Henry's constants for excess amount adsorbed and isosteric heats.However, this procedure is not straight forward, since the comparison of results between experiments and simulation needs pore volume information of the adsorbent. Moreover, since it is assumed that helium is non-adsorbing to estimate this pore volume, comparison of helium adsorption results between simulation and experiments was ambiguous so far.

In this work, we propose a novel approach to compare results between experiments and simulations. A simple optimization technique to obtain the effective potential parameters is also presented. Using this method helium adsorption results between experiments and simulations can also be compared without ambiguity.

The proposed method was validated by investigating helium adsorption in silicalite. The potential parameters obtained for He-O interactions in silicalite are 3.104 Å and 48.7 K. The helium pore volume for silicalite was found to be 0.155 cm³/g. We further demonstrate the use of these potential parameters by comparing appropriate simulation results with experimental observations for helium adsorption in silicalite over wide temperature (93-515 K) and pressure (0-3300 kPa) range. The experimental observations used were taken from one of our previous works. The average difference between simulation and experimental results was less 2% using the parameters proposed in this work. All other potential parameter sets used in literature so far yield larger average deviations (greater than 17% in appropriate domain).