Pure Compound Self-Diffusivity Correlation With Residual Entropy

Two
models of self-diffusivity in terms residual entropy have been
proposed in the literature. A generalized residual entropy model is
proposed here by merging aspects of these two models. The generalized
model enables inference of interaction and scaling effects by probing
the significance of related terms when minimizing deviations of each
model relative to experimental data. As a result, the interaction
effect between residual entropy and temperature is found to be negligible
using a database of 688 experimental measurements for 13 n-alkanes. Similarly, scaling analysis disfavors the “Rosenfeld”
scaling. Instead, the scaling proposed by Dzugutov is favored. While
implementing these two simplifications, the reduced 3-parameter model
is shown to provide improved accuracy relative to the previous 6-parameter
model. On the other hand, comparison to a previously proposed 2-parameter
self-diffusivity model shows that the 2-parameter model provides practically
the equivalent accuracy to the 3-parameter model. The 2-parameter
model does not apply the principle of residual entropy. Instead, it
applies a corresponding states approach, where a model is fit to the
self-diffusivity of the Lennard-Jones model potential and Lennard-Jones
parameters are treated as adjustable parameters. By considering accuracy
relative to the number of parameters, the 2-parameter and 3-parameter
models are deemed equally acceptable. The root-mean-square logarithmic
deviation for the 2-parameter model is 7.5% compared to 6.9% for the
3-parameter model and 10.2% for the 6-parameter model. An open-source
database of experimental measurements for 34 compounds at 1685 state
points is used in all evaluations. The analysis here is limited to
nonassociating compounds. The mean average percentage error was less
than 5% for all three models.