Fluid Phase Equilibria 382 (2014) 21–30 Contents lists available at ScienceDirect Fluid Phase Equilibria jou rn al h om epage: www.elsevier.com/locate/fluid A group contribution model for the prediction of the freezing point of organic compounds Farhad Gharagheizi a,b , Poorandokht Ilani-Kashkouli a,b , Arash Kamari a , Amir H. Mohammadi a,c,∗ , Deresh Ramjugernath a,∗ a Thermodynamics Research Unit, School of Engineering, University of KwaZulu-Natal, Howard College Campus, King George V Avenue, Durban 4041, South Africa b Department of Chemical Engineering, Buinzahra Branch, Islamic Azad University, Buinzahra, Iran c Institut de Recherche en Génie Chimique et Pétrolier (IRGCP), Paris Cedex, France a r t i c l e i n f o Article history: Received 3 March 2014 Received in revised form 19 August 2014 Accepted 20 August 2014 Available online 28 August 2014 Keywords: Group Contribution Freezing Point Sequential Search Model Database a b s t r a c t The freezing point is a fundamental thermo-physical property which is important in describing the tran- sition between the liquid and solid phases. As this property is required for describing phase behavior and the design of separation unit operations, an efficient, applicable and reliable method which can predict it is of great importance, especially for compounds where there are no experimental data available. In this article, an efficient and reliable group contribution (GC) model is developed for the determination of the freezing point of organic compounds. The sequential search mathematical approach is used in this study to select an optimal collection of functional groups (112 functional groups) and subsequently to develop the model. A large dataset of freezing point data for about 17,000 pure mostly organic com- pounds was used to develop and validate the model. A comparison between the model results and the database shows a squared correlation coefficient of 0.735 (R 2 ). Moreover, the proposed group contri- bution model is able to predict the freezing point of organic compounds to within an average absolute relative deviation of 10.76%, which is of adequate accuracy for many practical applications. Furthermore, the leverage approach (Williams plot) is used to determine the applicability domain of the model and to detect probable erroneous data points. © 2014 Elsevier B.V. All rights reserved. 1. Introduction At the freezing point of a solution, solid solvent is in equilib- rium with the solvent in solution. As with the melting point, an increased pressure normally raises the freezing point. The freezing point is lower than the melting point, in the case of mixtures and for certain organic compounds such as fats [1]. As a mixture freezes, the solid that forms first normally has a composition different from that of the liquid, and formation of the solid changes the composi- tion of the remaining liquid, normally in a way that steadily lowers the freezing point. This principle is utilized in purifying mixtures, successive melting and freezing gradually separating the compo- nents [1]. Consequently, the fusion heat (heat required to melt a solid) must be removed from the liquid to freeze it. Some liquids ∗ Corresponding authors at: Institut de Recherche en Génie Chimique et Pétrolier (IRGCP), Paris, France. E-mail addresses: a.h.m@irgcp.fr, amir h mohammadi@yahoo.com (A.H. Mohammadi), ramjuger@ukzn.ac.za (D. Ramjugernath). can be supercooled (cooled below the freezing point) without solid crystals forming. The addition of a seed crystal into a supercooled liquid triggers freezing, whereupon the release of the heat of fusion raises the temperature rapidly to the freezing point [1]. Freezing point and/or melting point (depending on some consid- erations in their descriptions) are fundamental physical property specifying the transition temperature between liquid and solid phases [2]. Furthermore, they have been used for the prediction of other physical properties such as aqueous solubility [3–5]. Hence, accurate prediction of this fundamental thermo-physical property seems an essential necessity. To date, there have been a few quan- titative structure-property relationships (QSPR) methods, such as the property–property relationships (PPR) [6], and group contribu- tion methods [7–9] applied in attempt to estimate freezing/melting point. There are some successful estimations of melting points, e.g. for 24 normal alkanes (R 2 = 0.998) using topological indices like the carbon number, Wiener index, and the Balaban distance sum con- nectivity index [10]. Nevertheless, some models such as the QSPR models proposed by Needham et al. [11] indicate poor predictabil- ity (R 2 = 0.570) for their use of 56 normal and branched alkanes. A http://dx.doi.org/10.1016/j.fluid.2014.08.025 0378-3812/© 2014 Elsevier B.V. All rights reserved.