(515m) An Experimental and Modelling Study of n-Decane Autoxidation

Introduction
Fuel reactivity is a key element of fuel/engine adequacy governing engine efficiency and emission.
While gas-phase reactivity has been thoroughly studied, literature studies on liquid-phase reactivity are
scarce. From storage to injection in the combustion chamber, significant variation of temperature and
pressure occurs that leads to fuel autoxidation (aging). This unwanted phenomenon leads to changes in
physical and chemical characteristics of the fuel because of the formation of oxidation products and
deposits. Autoxidation can negatively affect engine components such as injectors, pumps, and filter
system, and significantly deteriorates the combustion quality [2] and enhances particle emissions [3].
Currently, the rigorous simulation of fuel autoxidation processes remains a challenge in the literature [5-
8]. A better understanding of liquid-phase oxidation is required to have a better control on this natural
phenomenon. This study aims at experimentally and numerically investigating the reactivity of n-decane
in the liquid phase. The liquid-phase oxidation mechanism of n-decane developed in this work will be a
building block for the development of a detailed chemical kinetic model of a surrogate fuel autoxidation
model.
Experimental method
In this study, the liquid-phase oxidation of n-decane was conducted in a PetrOxy apparatus, which is
part of the ASTM D7545 standard test. Initial oxygen pressure was set at 700 kPa. The temperature of
PetrOxy was held constant. Fuel oxidation stability is quantified by an induction period (IP). The latter
parameter corresponds to the time required to observe a decrease from the gas-phase maximum pressure
by 10%. IP measurements were conducted at two temperatures, 408 and 413 K. During a PetrOxy
experiment, liquid phase samplings were also conducted. These samples were then analyzed by
iodometric titration. This method allows to determine the total concentration of hydroperoxides formed
during the autoxidation process. IP and hydroperoxide speciation as a function of time are valuable data
used to validate the liquid-phase kinetic model of n-decane.
Numerical method
Numerical efforts have been performed in the literature to investigate the autoxidation of n-decane.
Chatelain et al. [5] developed a detailed liquid-phase chemical mechanism of n-decane using an
automated mechanism generator, RMG [9]. For the mechanism generation, RMG adapts an initial gasphase
kinetic model to the liquid-phase by correcting the gas-phase thermodynamic data with free
energies of solvation. The latter correction is estimated based on a linear solvation energy relationship
(LSER) [10]. The model of Chatelain et al. overestimated IP of n-decane with in a factor of 3 [5].
In this study, the development of the liquid-phase oxidation kinetic model of n-decane is based on the
gas-phase one. The modelling work in this study consists of three main steps. In the first step, the gasphase
mechanism of n-decane is generated by the EXGAS software [11], which is an automatic
generator of detailed kinetic model of combustion developed in our group. This software has been
successfully employed to generate robust gas-phase kinetic models [12-14]. The second step allows
adapting the gas-phase model to a liquid-phase model using corrections of thermochemistry data and
rate constants. Thermodynamic data are corrected with free energies of solvation computed with the
UMR-PRU equation of state [15]. Diffusion rate constants are calculated to correct bimolecular rate
constants using the Stoke-Einstein equation, with molecular radii calculated with the UNIFAC
molecular parameters and temperature dependent viscosities taken from experimental databases. Rate
constants for the most sensitive reaction for n-decane conversion and hydroperoxide formation were
determined based on literature review and theoretical chemistry calculations to propose new reaction
rate rules. The final step is dedicated to validating the liquid-phase kinetic model using PetrOxy data
generated during the experimental work and other experimental data obtained from well-defined reactors
in the literature. PetrOxy experiments can be modeled as a homogenous batch reactor [5,6].