(211g) Correlation of Liquid Thermophysical Properties Via Virial-Based Mixing Rules

Over the last several decades, a large number of methods have been proposed for calculating the thermophysical properties of pure liquids (and gases). These include corresponding states techniques and statistical thermodynamics approaches. In modern simulators these methods are usually referred to as stand-alone or library methods (i.e., not based on multi-property P-V-T equations of state), and are available as a standard set of ?temperature-dependent correlations?. The component-specific parameters required by these correlations are implemented as built-in or user-defined libraries. A typical example of a built-in library is one that is derived from the DIPPR-801 database - a collection of high-quality peer-reviewed component-specific parameters.

Although, these library correlations provide high quality estimates for several pure component liquid properties of interest, the science for describing mixtures has consistently trailed behind. For example, the mixing rules typically implemented in process simulators for liquid viscosity and thermal conductivity, are the Kendall-Monroe (1917) and Li (1976) correlations, respectively. While the attractive feature of these mixing rules is that they do not require interaction parameters (IP's), their downside is that the do not allow IP's, even when necessary. In many applications, they incorrectly predict only a small deviation from ideal mixing, often with incorrect sign. Thus, while these mixing rules perform reasonably for hydrocarbon systems, their efficacy for non-ideal systems is questionable.

The state-of-the-art is the Redlich-Kister correlation (RKC), which can incorporate additional higher-order terms to improve the fit of data with higher precision. The RKC is often the correlation of choice of the applied thermodynamics research community. It has been successfully applied to correlate a variety of thermophysical properties for binary mixtures, including density, enthalpy, and viscosity. However, the RKC does not naturally extend to multi-component systems. In addition, cross-correlation versus temperature compromises its ability to describe composition-dependence.

This investigation draws upon the exact science of mixing rules that has been established for the virial coefficients of gases. Perturbations from ideal mixing are modeled akin to successive terms in the virial expansion, resulting in the name ?Virial-Based Mixing Rules? (VBMR). Each perturbation reduces to zero for pure components, and adopts the composition order of the corresponding virial coefficient it represents. For example, the first perturbation represents the second virial coefficient and implements a quadratic mixing rule, while allowing one adjustable binary interaction parameter. The next term employs cubic mixing and allows multiple interaction parameters, including a ternary interaction parameter which may be fine-tuned for ternary and higher mixtures. All interaction parameters are initialized to zero.

Several binary and ternary systems (featuring high-precision as well as noisy data) are investigated. The target properties in this study are liquid density, enthalpy, viscosity and thermal conductivity. The necessity of the higher-order terms in correlating precise data within measurement uncertainty is established. In contrast, it is demonstrated how data with excessive noise do not warrant correlation with higher-order terms, while noting that the proposed approach has a low susceptibility to correlate noise. Wherever appropriate, the VBMR are compared with the RKC (for example, in the correlation of precise data). Cross-correlation with temperature resulting in a compact set of parameters is also demonstrated.