Fluid Phase Equilibria 355 (2013) 81–86 Contents lists available at ScienceDirect Fluid Phase Equilibria j ourna l ho me page: www.elsevier.com/locate/fluid Development of a group contribution method for estimating the thermal decomposition temperature of ionic liquids Farhad Gharagheizi a , Poorandokht Ilani-Kashkouli a , Amir H. Mohammadi a,b,∗ , Deresh Ramjugernath a,∗∗ , Dominique Richon a,c a Thermodynamics Research Unit, School of Engineering, University of KwaZulu-Natal, Howard College Campus, King George V Avenue, Durban 4041, South Africa b Institut de Recherche en Génie Chimique et Pétrolier (IRGCP), Paris Cedex, France c Department of Biotechnology and Chemical Technology, School of Science and Technology, Aalto University, P.O. Box 16100, 00076 Aalto, Finland a r t i c l e i n f o Article history: Received 30 January 2013 Received in revised form 22 June 2013 Accepted 28 June 2013 Available online 8 July 2013 Keywords: Thermal decomposition temperature Ionic liquid (ILs) Group contribution Reliable model Dataset a b s t r a c t In this communication, a reliable group contribution (GC) method is presented for the estimation of the thermal decomposition temperature (T d ) of ionic liquids. A dataset comprising experimental T d data for 613 ionic liquids (ILs) that covers a temperature range from 374 to 740 K was collated from various literature sources. Approximately 80% of the dataset (T d data for 489 ILs) was used to develop the model and the remaining 20% (T d data for 124 ILs) was implemented to evaluate the predictive capability of the obtained model. The method uses a total of 30 substructures or structural functional groups to estimate the T d . In order to distinguish the effects of the anion and cation on the T d , 10 sub-structures related to the chemical structure of the anion, and 20 substructures related to the chemical structure of the cation were implemented. The results of this method show an average relative deviation (AARD%) of about 4.4% from dataset values. © 2013 Elsevier B.V. All rights reserved. 1. Introduction Ionic liquids (ILs) are special salts, which are typically liquid below the normal boiling point of water or in case of room temper- ature ILs, below room temperature. These ILs which are entirely composed of ions generally consist of a combination of a large organic cation with smaller sized and more symmetrical anion. Their most important characteristic is that their properties can be significantly manipulated for any particular application by chang- ing their combination of anion and cation. This latter attribute makes them “designable materials” [1,2]. In order to design a new IL, one of the first steps requires the estimation of the elementary physico-chemical properties of the substance. One of the most significant properties of ILs, which determines their processing temperature range, is their liquidus range which is bounded by their normal melting temperature (T m ) as the lower limit and their thermal decomposition temperature (T d ) as the upper limit of temperature. Although the T m of ILs has been carefully studied and numerous models have been suggested ∗ Corresponding author at: Institut de Recherche en Génie Chimique et Pétrolier (IRGCP), Paris Cedex, France. ∗∗ Corresponding author. E-mail addresses: a.h.m@irgcp.fr, amir h mohammadi@yahoo.com (A.H. Mohammadi), ramjuger@ukzn.ac.za (D. Ramjugernath). to date for its prediction [3–16], T d has not been appropriately investigated and only two models have been recently proposed for its estimation [17,18]. One of the estimation methods mentioned above, which was developed by Lazuss [17], is a group contribution method (GC). The model is based on a dataset containing 198 experimental T d data for correlation and prediction (120 data points for developing the model and tuning the model parameters, and the remaining 78 for its validation). Using a combination of a genetic algorithm as an optimizer method and least square error as an objective function, the method introduced a collection of 58 sub-structures (27 cation- based and 31 anion-based) to estimate the T d . The average absolute relative deviations of the model results from experimental data for the correlation set and the prediction set were calculated as 4.3% and 4.2%, respectively. The second estimation method which was proposed by Yan et al. [18] is a quantitative structure–property relationship (QSPR) model using a dataset of 158 experimental T d data (126 data points for developing the model and tuning the model parameters, and the remaining 32 values for its validation). The average absolute rela- tive deviations of this 25-parameter model from experimental data for the training set and the test set were approximated as 3.1% and 3.5%, respectively. In addition to reviewing the estimation methods available for T d , an extensive literature survey undertaken during this study 0378-3812/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.fluid.2013.06.054