Prescreening of Liquid Density and Surface Tension for Synthetic Aviation Turbine Fuels by Nuclear Magnetic Resonance Atom Types

Prescreening of Synthetic (“Sustainable”) Aviation Turbine Fuel, referred to as SAF in this work, is becoming regarded as an essential technique to secure the early-stage development of SAF candidates. Prescreening refers to the study of small volumes of SAF at the early stages of product development for the purpose of assessing the compatibility of the SAF candidate with the American Society for Testing and Materials (ASTM) D4054 evaluation procedure for new aviation turbine fuels. The basis of the prescreening process relies on the availability of numerical models that can predict the fit-for-purpose properties of the ASTM evaluation procedure without the volumes of the SAF candidate needed to perform all the necessary measurements being available. In this work, such models are developed that predict two important fit-for-purpose ASTM D4054 properties, the temperature dependent liquid density and temperature dependent surface tension of complex hydrocarbons mixtures. The basis of the method is the determination of atom type compositions by a 1H 13C Heteronuclear Single Quantum Coherence (HSQC) Nuclear Magnetic Resonance (NMR) spectroscopy methodology. The dependence of each of, temperature dependent liquid density, and temperature dependent surface tension, on the range of atom types that comprise typical aviation fuel hydrocarbons is learned by the multiple linear regression modelling to a database of pure component and pure component mixture data gathered from the literature. A data base of 1,241 entries is gathered for density, and a data base of 1,260 entries gathered for surface. Each database is divided in an 80:20 ratio for model training and model testing. Extensive iterations of model training and testing objectively identify the same seven atom types as being most influential to each property. The collection of seven atom types, in addition to a temperature dependent term, is shown to be sufficient to determine each property quite precisely. The models are then applied to the atom type compositions of two fossil-derived aviation fuels, four synthetic aviation fuels, and four reference gasoline fuels, as determined by 1H 13C HSQC NMR. The methodology shows high performance, predicting the unseen liquid density of these real fuels within an absolute error range of 0.00 – 5.59%, and surface tension within an absolute error range of 0.29 - 4.41%. It also successfully differentiates the relative magnitudes of each property across these fuels. This indicates that the method is valuable for prescreening applications.