NMR Characterisation and Physicochemical Property Prediction of Sustainable Aviation Fuels and Lignocellulosic Biomass

Abstract

This thesis develops and applies advanced Nuclear Magnetic Resonance (NMR) spectroscopy methods to address two central challenges in sustainable energy: (1) reducing certification barriers for sustainable aviation fuels (SAFs), and (2) enabling molecular-level characterisation of lignocellulosic biomass for bioenergy applications. Both problems require accurate, scalable, and minimally invasive analytical techniques capable of linking molecular-scale structure to bulk properties and performance. Aviation Turbine Fuels: The first strand of this research focuses on developing pre-screening tools for SAF certification. Currently, the ASTM D4054 approval process is resource-intensive and time-consuming. This work advances ¹H ¹³C (Heteronuclear Single-Quantum Coherence – Nuclear Magnetic Resonance (HSQC-NMR) spectroscopy, a technique correlating protons to their bonded carbon heteroatoms, as a quantitative tool for compositional analysis of jet fuels and introduces methods to directly predict key physicochemical properties. • A methodology was developed to quantify 19 distinct atom types, including paraffins, cycloparaffins, alkylbenzenes, and, critically, naphthalenes with ring-junction carbons. These structural classes are important for predicting viscosity, density, and thermal stability. Validation with model fuels of increasing complexity showed mean errors as low as 0.28 mol% C, comparable to state-of-art detailed hydrocarbon analysis (DHA). • By coupling atom type profiles with machine learning (ridge predictive framework for dynamic viscosity was constructed. Trained on a database of 75,000 virtual fuels, the model achieved an R² of 0.98 and mean errors approximately 5%. Applied to both fossil-derived (Jet A-1, JP-8) and synthetic fuels (Sasol-IPK, Gevo-ATJ), predictions remained within ~0.4–0.9 mPa·s (3–20%) error of measured values, demonstrating pre-screening viability. Higher relative errors were observed at lower dynamic viscosities as the model was trained to minimise absolute error and the area of greatest interest was at the higher certification cut-off viscosity (8 mPa·s @ -20 °C). • HSQC-NMR proved sensitive enough to detect aromatic resonances at levels below 1 mol% and reliably quantify olefins-classes critical for combustion behaviour but difficult to measure with conventional chromatography. Together, these advances regression), a predictive framework for dynamic viscosity was constructed. Trained on a database of 75,000 virtual fuels, the model achieved an R² of 0.98 and mean errors approximately 5%. Applied to both fossil-derived (Jet A-1, JP-8) and synthetic fuels (Sasol-IPK, Gevo-ATJ), predictions remained within ~0.4–0.9 mPa·s (3–20%) error of measured values, demonstrating pre-screening viability. Higher relative errors were observed at lower dynamic viscosities as the model was trained to minimise absolute error and the area of greatest interest was at the higher certification cut-off viscosity (8 mPa·s @ -20 °C).