(428f) Towards the Prediction of the Liquid Phase Oxidation of Aromatics – an Experimental and Modeling Study for Toluene Autoxidation

Due to adequate physicochemical properties as well as high octane number, aromatics are key components for fuel formulation. They limit abnormal combustion, that is knock for gasoline fuel, and also ensure seal swell and lubrication in jet fuels . However, the same aromatics are also subject to different discussions regarding their impact on other types of abnormal combustion such as low speed pre-ignition, where the reactivity in the liquid phase for fuel/lubricant mixes could be an important parameter. Literature also focuses on their effect on particulate formation in combustion as well as deposits in liquid phase in an autoxidation regime. Due to an important need for better simulation tools regarding fuel autoxidation as well as experimental data that can be simulated, this work focuses on these two aspects. Toluene being the simplest alkyl aromatic, its autoxidation at high temperature in liquid phase is investigated in this study.

Experimental work

The experimental work relies first on a PetroOxy which is employed in a standardized oxidation test method. It is designed for the assessment of automotive fuels according to test methods ASTM D7545, D7525 and EN 16901. This device has been previously used for a similar analysis [1]. This is a closed batch reactor that includes a gold plated dish on which a thin film of liquid is placed. The head-space is pressurized with oxygen to 7 bar and heated to the targeted temperature. The pressure is then recorded while the system evolves. Fuel can then be discriminated according to their induction period (IP) which correspond to a 10% pressure drop. In this study, two types of measurements are employed : (1) single induction period measurements at three individual temperatures of 120°C, 140°C and 160°C to characterize the temperature dependence of the system; (2) The experiment is stopped at half IP so that a liquid sample is analyzed and the toluene consumption can be monitored.

Secondly, an autoclave is set-up to enable a systematic investigation of intermediate autoxidation products at the highest temperature studied in the PetroOxy. A 250 cm³ reactor is filled with 50 mL of sample and agitated with pure oxygen bubbling through the liquid. The system is pressurized at 10 bar to the targeted temperature and the liquid is sampled at different times to follow products build-up.

The oxidation products analysis is carried out using gas chromatography coupled with mass spectrometry for identification. A gas chromatograph CP3800 Varian with a FFAP column of 30 m length, 250 μm and 0.25 μm film thickness is used. Quantification is performed with a flame ionization detector and a different gas chromatograph. It is an Agilent 6890N with the same type of FFAP column. Finally, qualitative measurements are also performed with a HP5 column to assess the feasibility of additional product identification.

Modeling approach

The Reaction Mechanism Generator RMG is an automated code, designed to propose chemical kinetics mechanisms for both the gas and liquid phase [2]. The generation requires a set of initial conditions including reactant species, their structure and concentration, the temperature as well as, for liquid phases, the solvent used. The mechanism generation is based on a library of reaction types such as hydrogen abstraction or radical recombination, with relevant kinetic parameters, for defined molecular graphs. For every molecule in the mechanism, possible reactions from the library, based on structural information of the molecule are evaluated. The highest reaction flux enables the code to select the next species that moves into the core mechanism. This process is repeated until the defined end criteria is reached. This can be a certain time or reactant conversion for example. Where no kinetic data is available in the database, an average, based on related parameters, is calculated in order to obtain an estimate for the reaction. Besides the kinetic estimation, thermochemical parameters are also estimated based on Benson group additivity for the gas phase. A liquid phase system implies solvation effects, RMG use corrective methods in order update the thermochemical parameters. The first one is based on a linear solvation energy relationship (LSER) which presents a parametrized description of solvent effects to estimate the Gibbs free energy of solvation. This method takes aspects such as viscosity and polarity into account as well as solute descriptions. This approach has been shown to provide a good accuracy for most solvents at low computational cost [3]. The second correction applied in RMG uses Mintz model which aims to obtain the enthalpy of formation in the liquid phase [4]. This model is quite similar in principle to the LSER parametrization method and has been evaluated for a large set of closed shell species. The remaining liquid phase entropy is then estimated by RMG using the two corrections cited above. Finally, the code also ensures that reactions do not exceed their diffusion limit. This is achieved by estimating an effective rate constant that takes into account the radii of the involved molecules as well as the diffusion rates in the solvent.

In the context of this work, quantum chemical calculations have been performed to asses and improve the quality of the RMG-predicted thermochemical parameters for individual species in the liquid phase. The functional used in this study is M06-2X which has proven its capabilities in terms of accuracy since, including in the application of organic chemistry. A modern Aldrich basis set and the RI-J auxiliary basis set is coupled for lower cost calculations to evaluate a large number of species with reasonable accuracy. Indeed, in the GMTKN30 benchmark, M06-2X performed best for basic properties with a weighted total mean absolute deviation of 3.2 kcal/mol and is one of the best performing hybrid functional [5]. Finally, solvation is considered using the SMD model [6] which has shown to perform well when compared to other implicit solvation methods. Explicit solvation may provide better predictions however such a calculation is beyond the scope of this study.

Results and conclusion

From an experimental side, the IP of toluene has been measured at three different temperatures using the Petrooxy apparatus. The toluene species profile has been measured by sampling the liquid at different times. A more systematic and robust sampling and analysis method has been applied with the autoclave experiments. This second set of data supports the previous findings for major oxidation products of toluene autoxidation at a moderate temperature, that is to say benzaldehyde, benzyl alcohol, benzoic acid as well as benzyl hydroperoxide. In addition, a detailed analysis of termination species for toluene autoxidation is also given. That way a set of potential deposit precursors is established.

These experimental results are used to generate and improve a detailed kinetic mechanism with RMG. This generator provides a first skeletal mechanism for which the thermochemistry is updated with quantum chemical calculations for thermochemistry parameters. Major changes applied to the thermochemistry impacted the equilibrium of key reactions such as the well-known ROO + RH = ROOH + R propagation step, that controls the system at these temperatures. To evaluate the modifications applied to the mechanism, the resulting rate constant for that key step is compared to Denisov’s data. The update of the thermodynamic parameters decreases the rate constant by one order of magnitude relative to RMG, leading to a better agreement with experimental data from literature [7]. This contributes to the fact that the updated mechanism reproduces within less than an order of magnitude the IP obtained with the PetroOxy reactor. On the speciation aspect, species observed experimentally are included in the mechanism. The toluene species profile is well simulated by the model and the qualitative trends for major products are also reproduced [8,9]. Quantitatively, additional work is ongoing to improve the simulation of the species profile.

This methodology demonstrates the feasibility of autoxidation global reactivity modeling with a detailed kinetic mechanism for a simple system. Potential improvements of RMG are targeted through the quantum calculations performed for the liquid phase mechanism generation. In addition, due to its interest for many aspects, this study with toluene provides a first building block to evaluate the capability of detailed kinetic mechanisms in specific co-oxidation phenomena as well as deposit formation that is of interest for many applications.

References

[1] Chatelain, K. et al., Energy & Fuels, 2016, 30 (2), 1294 – 1303

[2] Gao, C. W. et al., Computer Physics Communications, 2016, 203, 212 – 225

[3] Jalan, A. et al., The Journal of Physical Chemistry B, 2013, 117 (10), 2955 – 2970

[4] Mintz, C. et al., QSAR & Combinatorial Science, 2008, 27, 179 – 186

[5] Goerigk, L and Grimmie, S., Physical Chemistry Chemical Physics, 2011, 6670 – 6688

[6] Marenich, A. et al., The Journal of Physical Chemistry B, 2009, 113, 6378 – 6396

[7] Denisov, E. T. and Afanas'ev; I. B., Taylor & Francis, CRC Press, 2005

[8] Hoorn, J. et al., International Journal of Chemical Reactor Engineering, 2005, 3, 1 – 19

[9] Hermans, I. et al., ChemPhysChem, 2007, 8, 2678 – 2688