Sensitivity analysis of hydrate thermodynamic reference properties using experimental data and ab initio methods

Various sets of thermodynamic reference properties currently available in the literature were examined by applying the van der Waals and Platteeuw model to predict monovariant three-phase equilibria for hydrate-forming binary mixtures from 260 to 320 K. Two recent and well-accepted experimental studies of cyclopropane hydrates by Dharmawardhana et al. (1980, 1981) with analysis by Holder et al. (1984) (Deltamu(w)(0) = 1299 J/mol, DeltaH(w)(0) = 1861 J/mol) and the NMR study of xenon hydrates by Handa and Tse (1986) (Deltamu(w)(0) = 1287 J/mol, DeltaH(w)(0) = 931 J/mol) were revisited. The deviations introduced by experimental uncertainties were found to be large enough to cause significant changes in the prediction of dissociation pressures. The methodology and parameters fitted to various pure and binary hydrate systems provided by Sloan (1998) were also employed in our calculations for methane hydrates using the Lennard-Jones and Devonshire (LJD) approximation and the Soave-Redlich-Kwong (SRK) or the Peng-Robinson (PR) equation of state. The comparison between the experimental data and software prediction provided by Sloan (1998) is in good agreement over the low-temperature range (<290 K), but there are significant differences at high temperatures (290-320 K). The methane-water intermolecular potential from our previous ab initio study was also used to obtain the reference properties with small deviations (Deltamu(w)(0) = 1236 4 J/mol, DeltaH(w)(0) = 1703 +/- 62 J/mol), and the resulting parameters were able to give the best prediction over the entire temperature range using the ab initio potential. More importantly, the use of the ab initio potential removes the need for adjusting fitted potential parameters to make up for errors in the reference parameters. Statistical variations (+/-) at the 95% confidence limit computed for our data and those from previous researches are found to be significantly smaller using our method. Note that these statistical variations do not take into account systematic error, which is possible in both the experimental and ab initio cases. Sensitivity analyses based on both the LJD approximation and our ab initio study were performed. The value of Deltamu(w)(0) was found to be much more sensitive to the three-phase equilibria prediction of methane hydrates than DeltaH(w)(0). For example, a variation in Deltamu(w)(0) of +/-10-20 J/mol results in an error of 5%-10% in the predicted dissociation pressure.