(76d) Liquid-Phase Mechanism Generation for Predicting Fuel Oxidation

We have added functionality to automatic reaction mechanism generation software, RMG¹, to more accurately predict the reaction rates of detailed, liquid-phase chemical mechanisms. One industrial application of liquid-phase mechanism generation, explored here, is in developing models for fuel oxidation. Other technical applications for this software include the upgrading of heavy oils with supercritical water, and the processing of microalgae to produce fuels.

Mechanistic similarities exist between radical reactions in the gas phase and liquid phase, but the presence of a solvent changes both the thermodynamics and reactivity of each chemical species in a kinetic model. For systems which may include hundreds or thousands of reactions, we use automatic mechanism generators such as RMG, which was first built for the gas-phase, to generate models faster and minimize errors. Estimates for thermodynamic and diffusion corrections to the gas-phase values have been added to RMG², but so far no systematic corrections to the reaction rates exist for when there are no experimental data available.

For a set of hydrogen abstraction reactions, M06-2X/MG3S calculations were used to find the geometries of reactants and transition states in both gas and liquid phase. A continuum solvation model, SMD, was used to find the energy of the liquid-phase structures in eight different solvents. Using these energies, the difference in barrier height between gas and liquid phase, ΔE, was computed; a positive ΔE corresponds to a higher barrier in solvent than in gas. Trends in ΔE based on reactant molecular structure have been deduced, and these trends hold across various solvents. Specifically, these trends include: (1) In hydrogen abstraction from alcohols, ΔE increases with proximity of abstraction site to --OH group; (2) Hydrogen abstraction from molecules with increasing number of carbons beyond C2 does not change ΔE; (3) ΔE is greater in abstraction from aromatic rings than the corresponding saturated ring.

We have implemented molecular group contributions to ΔE into a hierarchical tree in RMG in order to estimate this difference in barrier for reactions without data. Using the new data, rates of hydrogen abstraction in a fuel oxidation model³were modified. Reactor simulations of the resulting model were compared with those of the previous model, which did not include liquid-phase kinetic corrections. For other reaction types, solvation kinetic trends can be similarly developed.

¹ Green, W.H., Allen, J.W. et al. RMG – Reaction Mechanism Generator, Python Version. http://rmg.mit.edu, 2013.

² Jalan, A. et al. An extensible framework for capturing solvent effects in computer generated kinetic models. J. Phys. Chem. B, 117: 2955-2970, 2013.

³ Ben Amara, A. et al. Toward Predictive Modeling of Petroleum and Biobased Fuel Stability: Kinetics of Methyl Oleate/ n-Dodecane Autoxidation. Energy & Fuels, 27: 6125-6133, 2013.