The transport sector is responsible for a significant fraction of the world’s total greenhouse gas (GHG) emissions. When derived from relatively abundant non-food renewable sources, such as lignocellulosic biomass, biofuels are a viable option to reduce GHG net-emissions. Hydrothermal liquefaction (HTL) is a process that transforms biomass into a wide range of products, including bio-oil; this in turn can be converted into direct replacements for fossil fuel-derived gasoline and diesel, being considered “drop-in” biofuel. The HTL process, which has already been demonstrated at pilot scale, is facilitated by the marked variation in the properties of water as it nears its critical point, allowing the fine-tuning of, e.g., density, acidity, and solubility of organics.
HTL of lignocellulosic biomass is a complex process due to the variability in the composition of the naturally occurring feedstocks, intricate reactional network, and strong dependence on operating conditions. This objective of the work presented in this thesis was to close the gap in knowledge between the kinetic mechanisms for the HTL of complex lignocellulosic biomass, and relatively simple model compounds. A methodology was developed to achieve this objective and tested using feedstocks of dimers that represent structures found in woody biomass: dibenzyl ether (DBE), representative of lignin, and sucrose, representative of cellulose.
Experiments were conducted in a configuration expected to be scalable for industrial applications, consisting of a continuous process, with operating pressure controlled independently from temperature. A confined jet mixer generated rapid mixing at the reactor’s inlet. Reactions occurred in a tubular section, being halted by quenching with water, followed by heat exchange with chilled water. For experiments using DBE as feed, HTL products were distributed between gas and liquid phases, with the latter clearly divided in organic and aqueous fractions. When sucrose was fed to the reactor, solid particles were also produced alongside compounds in gas and liquid phases. Separation and analytical methods were used to measure the amount of each phase and quantify selected products.
The mixing patterns around the confined jet mixer were analysed using computational fluid dynamics (CFD) simulations, which were representative of the performed experiments. All simulations with water fed under supercritical conditions achieved 99.9% of average outlet composition 16 mm after the jet’s outlet, confirming a very fast mixing. The conversion and yields of each simulation were compared against equivalent CFD models which did not consider feed mixing. No significant variations were observed between the outputs of the two types of models. Therefore, modelling the confined jet mixer in detail was not required to analyse the experimental results. Instead, each experiment could be simulated as a plug flow reactor with a single inlet consisting of the hot mixture of water and organic feed. The mathematical model developed included the calculation of: reaction rates, heat losses through the tube wall, thermodynamic and transport properties, and dispersive transport.
Those simulations were used to estimate kinetic parameters by minimising the difference between measured and estimated reaction yields, accounting for the respective experimental uncertainty. The hydrothermal cleavage of the ether bond in dibenzyl ether was studied at temperatures in the range 287 – 369 °C, pressures from 245 to 254 bar, and residence times between 5 and 7 seconds. Simulations accounted for hydrolysis and degradation reactions. Detailed hydrolysis rates including two observed kinetic constants provided a better fitting between experimental data and calculated values. However, the estimated parameters exhibited a large uncertainty. Based on simplified reaction rates, DBE hydrolysis most likely occurred through an SN2 mechanism with OH- as the substitution nucleophile, being affected by the concentration of this ion, as well as H+. For this reaction, the apparent activation energy was estimated as 180 ± 2 kJ/mol.
The degradation of sucrose in the hydrothermal medium was investigated at temperatures, pressures and residence times in the ranges 282 – 372 °C, 246 – 254 bar, and 3 – 7 seconds, respectively. The system was modelled as a homogeneous liquid, with organic compounds divided in three groups: soluble organics, volatiles, and solids. The first group included glucose, fructose, 5-(Hydroxymethyl)furfural (5-HMF), pyruvaldehyde, and 1,6-anhydro-β-D-glucose, resulting in a total of 16 possible reactions. A stepwise approach was developed to obtain kinetic parameters under increasing levels of model complexity and considering the effect of water’s self-ionisation product in subcritical conditions when relevant. Analysing the standard deviation of the obtained kinetic parameters identified the ones that overparameterized the estimation, resulting in the simplification of the 16 reactions to 9. Those reactions included: (1) isomerisation between glucose and fructose, which was driven towards the latter (activation energy of 73 ± 2 kJ/mol); (2) 1,6-anhydro-β-D-glucose formation from glucose and (3) degradation (activation energies of 45 ± 7 and 45 ± 10 kJ/mol, respectively); (4) formation of pyruvaldehyde from fructose (activation energy of 71 ± 5 kJ/mol); (5) 5-HMF production from fructose and (6) degradation (activation energies of 52 ± 9 and 220 ± 60 kJ/mol, respectively). From those, only the rates of (3) and (5) were expected to be significantly affected by [OH-] or [H+].