(515l) Automatic Generation of Liquid-Phase Kinetic Models for Fuel Autoxidation

The European regulations impose an increased share of biofuels in the transport sector, which will be blended with conventional transportation fuels. One of the main obstacles to the widespread use of biofuels is their tendency to downgrade the oxidation stability (resistance to aging) of conventional fuels. During storage, transport and injection phases in engine operations, liquid fuels are in contact with air, and undergo significant variations in temperature and pressure, which can favor the fuel autoxidation. This phenomenon leads to changes in the physical and chemical fuel characteristics through the formation of oxidation products, which are gum and deposits precursors. For example, the viscosity of n-alkanes could be raised by a factor of up to 65 [1] due to autoxidation, which can affect negatively engine components as injector, pump, and filter system. The oxidation products can also alter the combustion quality [2] and enhance the particle emission [3]. In the literature, the chemical process governing fuel autoxidation remains poorly understood [4-7]. The chemical mechanism underlying the liquid–phase oxidation of fuels is characterized by a mechanism of free-radical chain reaction. The rigorous modeling of this phenomenon requires the use of detailed chemical kinetic models composed of hundreds to thousands elementary reactions. Automatic generation of detailed kinetic models is a powerful tool to create and handle such large mechanisms. However, if their development is well- established for the generation of gas-phase kinetic models (combustion), this is not the case for liquid- phase oxidation. Currently, RMG software from MIT [8] is the only tool able to generate kinetic models adapted to the liquid phase. The methodological approach used by these authors relies on automated corrections applied to the generated gas-phase oxidation models. In this work, we present new approaches to capture solvent effects on thermodynamic, diffusion and intrinsic rate constants, and their implementation in our gas-phase automatic generator of detailed kinetic models named EXGAS [9]. Validation tests are performed against experimental data of the literature for the liquid-phase oxidation of n-butane.

Numerical method

The automatic generation of a liquid-phase oxidation kinetic model for n-butane starts with the generation of a gas-phase mechanism in EXGAS. This software has been extensively validated for combustion kinetics [10-12]. The generated kinetic model contains a list of several thousands of elementary reactions as well as the associated thermodynamic and kinetic data (ideal gas). New procedures have been coded in EXGAS to adapt these data to the liquid phase.

Thermodynamic data. Gas-phase thermodynamic data are adapted to the liquid phase using free energies of solvation (ΔGsolv) corrections. This approach is also used in RMG, where ΔGsolv is calculated using an empirical linear free energy relationship (LFER) from the literature, which can be applied to 25 solvents. In our work, ΔGsolv is automatically computed for all the species of the mechanism using the UMR-PRU (Universal Mixing Rule Peng-Robinson UNIFAC) equation of state [13,14]. The computed ΔGsolv values were shown to accurately predict thousands of experimental values, with a mean absolute deviation of 0.36 kcal/mol, for a large variety of solute/solvents mixtures at different temperatures [15]. This new approach therefore considerably expands the number of fuels (solvents) and the temperature range that can be treated with our code, compared to LFER approaches defined at ambient temperatures.

Diffusion corrections. Bimolecular gas-phase rate constants are corrected with a diffusion rate constant calculated using a Stokes-Einstein approach for the diffusion of spherical particles through liquids. Molecular radii are calculated based on the UNIFAC molecular size parameter included in UMR-PRU equation of state and temperature dependant viscosities that are read in database.

New rate rules for kinetic parameters. A set of new reaction rate rules, adapted to the liquid-phase oxidation of n-alkanes, is established for the decomposition of hydroperoxides, the self-termination

reactions of peroxy radicals, and the H-atom abstractions by a peroxy and an alkoxy radical. These reaction rate rules are defined based on literature data. When there was a lack of experimental data, theoretical chemistry calculations have been used. Gas-phase rate constants (kgas) are calculated based on potential energy surfaces computed at the CBS-QB3 level of theory and transition state theory with rigid rotor harmonic oscillator approach (1-DHR-U treatment of internal rotations) and Eckart tunnelling [16]. The computed kgas were then corrected with ΔGsolv of activation computed at the M06-2X/6- 31+G(d) level of theory with the SMD solvation model [17].

Results and conclusions

Based on the numerical method described above, the liquid-phase kinetic model of n-butane has been generated with EXGAS. The validation of this model is performed against the experimental data of Mill et al. [18] who conducted experiments on the liquid-phase oxidation of n-butane in a batch reactor. A homogeneous liquid batch reactor (constant T, P and V) model has been developed based on Chemkin II library [19]. The developed model reasonably simulates the conversion of n-butane as shown in Figure 1. Table 1 summarizes the initial conditions of the experiments presented in Figure 1.

The good simulation results obtained in the case of n-butane confirm the robustness of the method proposed in this study to simulate liquid-phase oxidation. On-going works are performed to expend and validate our generator of liquid-phase oxidation mechanisms to other hydrocarbons and oxygenated fuels. Once validated, this tool will be used to simulated and understand the fundamental phenomena that affect the oxidation stability of conventional fuels when biofuels are added.

References

[1] Blaine, S. et al., Ind. Eng. Chem. Res. 30 (1991) 2185−2191.
[2] Kuboyama, T. et al., SAE Int. J. Engines 8 (2015) 529–537.
[3] Pourkhesalian, A. M. et al., Atmos. Chem. Phys., 15 (2015) 9099–9108.
[4] Chatelain K. et al., Energy Fuels 30 (2016) 1294-1303.
[5] Mielczarek D.C. et al., Energy Fuels 31 (2017) 12893-12913.
[6] Chatelain K. et al., Energy Fuels 32 (2018) 9415-9426.
[7] Alves-Fortunato M. et al., Fuel 268 (2020) 117074.
[8] Green et al., RMG−Reaction Mechanism Generator; Department of Chemical Engineering, MIT, 2013.
[9] Warth V. et al., Computers & chemistry 24 (2000) 541-560.
[10] Schönborn M. et al, Combustion and Flame 205 (2019) 466-483.
[11] Schönborn M. et al, Energy Fuels 33.12 (2019) 12704 – 12713.
[12] Gornay J. et al., Journal of Analytical and Applied Pyrolysis 146 (2020), 104767
[13] Voutsas E. et al., Fluid Phase Equilib. 241, 216 (2006).
[14] Voutsas E. et al., Ind. Eng. Chem. Res. 43, 6238 (2004).
[15] Moine E. et al., Pure and Applied Chemistry 91 (2019) 1295-1307
[16] Lizardo-Huerta J. C. et al., Phys. Chem. Chem. Phys., 2016, 18, 12231– 12251
[17] Marenich A. V. et al., J Phys Chem B 113.18 (2009) 6378-6396
[18] Mill, T et al., Journal of the America Chemical Society 94.19 (1972) 6802-6811
[19] Kee R. J. et al., CHEMKIN-II: a Fortran chemical kinetics package for the analysis of gas-phase chemical kinetics. Sandia Report SAND89-8009, Sandia National Laboratory, Livermore, CA, 1989