Methodology to Predict Thermodynamic Data from Spectroscopic Analysis

Sustainable processes, often dealing with complex mixtures, would benefit from the availability of more accurate and predictive thermodynamic models. Most existing models are (semi)empirical and require extensive input, while application to complex mixtures is cumbersome. In this work, the potential of extracting information about nonideal behavior directly from spectroscopic information as a sole source is studied. A methodology framework is proposed and 45 binary component mixtures with a broad nonideality range were evaluated. Excess infrared absorbance spectra were successfully correlated with Gibbs excess energy using multivariate data analysis. For most binary mixtures, experimental vapor-liquid equilibrium literature data could be predicted accurately based on a model (UNIQUAC) using thermodynamic parameters obtained from the spectroscopic results. This also applied to binary mixtures that were not involved in the correlating step. Potential benefits of the investigated method are cost-effective, accurate, and quick measurement of nonideality information, and improved predictive models, even for complex mixtures. The principle is demonstrated, and suggestions for further developments are provided.